STP 415-1967

Fatigue Crack Propagation
A symposium
presented at the
Sixty-ninth Annual Meeting
AMERICAN SOCIETY FOR
TESTING AND MATERIALS
Atlantic City, N. J., 26 June-1 July, 1966
ASTM SPECIAL TECHNICAL PUBLICATION NO. 415
List price $30.00; 30 per cent discount to members
Copyright by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
Downloaded/printed by
published by the
University of Washington (University of Washington) pursuant to License Agreem
AMERICAN SOCIETY FOR TESTING AND MATERIALS
1916 Race Street, Philadelphia, Pa. 19103
© BY AMERICAN SOCIETY FOR TESTING AND MATERIALS 1967
Library of Congress Catalog Card Number: 67-14532
NOTE
The Society is not responsible, as a body,

for the statements and opinions
advanced in this publication.
University of Washington (University of Washington) pursuant to License Agreement. No furthe
Printed in Baltimore, Md.

June, 1967
Foreword
The Symposium on Fatigue Crack Propagation was presented in four

sessions during the 69th Annual Meeting of the Society, in Atlantic City,
N.J., 26 June-1 July, 1966. The symposium was sponsored by Com-
mittee E-9 on Fatigue with the cooperation of Committee E-24 on Frac-
ture Testing of Metals. The symposium chairman was J. C. Grosskreutz,
Midwest Research Institute. Presiding at the four sessions were Mr.
Grosskreutz; H. F. Hardrath, National Aeronautics and Space Adminis-
tration; G. M. Sinclair, University of Illinois; and Paul C. Paris, Lehigh
University.
University of Washington (University of Washington) pursuant to License Agreement. No further rep
Related
ASTM Publications
Five Year Bibliography on Fatigue, STP 9BB (1966),

$17.00
Fatigue Tests of Aircraft Structures: Low-Cycle, Full-

Scale, and Helicopters, STP 338 (1963),
$10.50
Fracture Toughness Testing and Its Applications, STP

381 (1965), $19.50
Plain Strain Crack Toughness Testing of High Strength

Metallic Materials, STP 410 (1967), $5.50
University of Washington (University of Washington) pursuant to License Agreement. No
Contents
Introduction 1
Fatigue Crack Growth in Structures
Limitations of Fatigue-Crack Research in the Design of Flight Vehicle
Structures—R. H. CHRISTENSEN AND M. B. HARMON 5
Discussion 24
Crack Propagation and Residual Static Strength of Fatigue-Cracked
Titanium and Steel Cylinders—w. J. CRICHLOW AND R. H.
WELLS 25
Fatigue Crack Propagation in Structures with Simulated Rivet Forces
1. E. FIGGE AND J. C. NEWMAN, JR 71
Low Cycle Fatigue Crack Propagation Resistance of Materials for
Large Welded Structures—T. w. CROCKER AND E. A. LANGE . . . . 94
Discussion 126
Microstructural Aspects of Fatigue Crack Growth
The Influence of Metallurgical Structure on the Mechanisms of Fatigue
Crack Propagation—CAMPBELL LAIRD 131
Discussion 169
Effect of Environment on Fatigue Cracks—M. R. ACHTER 181
Discussion 203
Fatigue Fracture Surface Appearance—R. w. HERTZBERG 205
Discussion 224
Microstructures at the Tips of Growing Fatigue Cracks in Aluminum
Alloys J. C. GROSSKREUTZ AND G. G. SHAW 226
Discussion 242
The Continuum Approach to Fatigue Crack Growth
Mechanics of Crack Tip Deformation and Extension by Fatigue—j. R.
RICE 247
Discussion 310
Crack Propagation in Clad 7079-T6 Aluminum Alloy Sheet Under
Constant and Random Amplitude Fatigue Loading—s. R.
SWANSON, F. CICCI, AND W. HOPPE 312
Discussion 360
Investigation of Cyclic Crack Growth Transitional Behavior—D. p.
WILHEM 363
Discussion 380
University of Washington (University of Washington) pursuant to License Agreement. No further reproduc
Application of a Double Linear Damage Rule to Cumulative Fatigue
S. S. MANSON, J. C. FRECHE, AND C. R. ENSIGN 384
Discussion 412
VI CONTENTS
Review, Analysis, and Discussion of the Fatigue Crack Growth Problem

Significance of Fatigue Cracks in Micro-Range and Macro-Range—
J. SCHIJVE 415
Discussion 458
Fatigue-Crack Propagation in Some Ultrahigh-Strength Steels—R. P.
WEI, P. M. TALDA, AND CHE-YU LI 460
Discussion 480
The Effect of Grain Size on Fatigue Crack Propagation in Copper—
D. W. HOEPPNER 486
Fatigue Crack Propagation Under Program and Random Loads—J. c.
MCMILLAN AND R. M. N. PELLOUX 505
Discussion 533
Summary
Summary 537
Introduction
Crack propagation is the one phase of fatigue failure which can be

treated quantitatively without the added complication of statistical
fluctuations. Moreover, in many materials, and especially in notched
structures, the crack propagation phase occupies a major fraction of the
useful fatigue life. There is, in fact, one school of thought which views
all of fatigue damage as the growth of cracks; cracks which grow from
preexisting flaws in real structures.
In recent years an enormous amount of effort has been devoted to the
study of fatigue crack propagation. Some of the first fruits of this effort
were reported at the Crack Propagation Symposium held in Cranfield,
England, in 1961. Since that time several powerful techniques for the
study of crack growth have been developed more fully. Electron frac-
tography has been applied to a wide variety of materials, and the distin-
guishing "fatigue striation" has been studied in detail. Transmission
electron microscopy has been used for the observation of localized
damage in the vicinity of cracks. The science of fracture mechanics,
which treats the stress concentration, energy balance, and plasticity at
crack tips, has found fruitful application to the fatigue problem. Careful
microscopic observation of growing fatigue cracks has led to semi-
quantitative models for the crack extension process. Test apparatus for
application of random loads has been developed, and crack growth un-
der these service-simulated loads has been observed. Finally, the results
of crack propagation research are becoming useful tools which the
engineer can use in designing real structures. In the light of these signifi-
cant advances, the members of Committees E-9 and E-24 of ASTM
felt that the Society could perform a valuable service to the scientific and
engineering community by bringing together representative people to
review recent accomplishments in this field.
This book, which details the proceedings of the Fatigue Crack Propa-
gation Symposium, is divided into four sections. The first three sections,
Fatigue Crack Growth in Structures, Microstructural Aspects of Fatigue
Crack Growth, and Continuum Approach to Fatigue Crack Growth, are
each introduced with a review paper which is followed by several shorter
papers which report current research activities in the field. The last sec-
tion of the book contains a comprehensive paper by J. Schijve, who,
through his broad knowledge of the field, manages to bring together the
diverse elements of crack growth research into a coherent whole. Dr.
Schijve also concludes the volume with a critical survey of the results of
the symposium, in which he defines those areas most in need of additional
work. The last section also includes papers which report very recent
1
2 FATIGUE CRACK PROPAGATION
experimental results which became available after the original Sym-

posium Program had been finalized.
The excellent coverage of the field which we obtained in this sym-
posium was due to the concerted efforts of the program committee. This
committee consisted of: H. F. Hardrath, NASA, Langley Research
Center; A. J. McEvily, Scientific Laboratory, Ford Motor Co.; and P. C.
Paris, Lehigh University. The committee and ASTM gratefully acknowl-
edge the financial support made possible by the Ford Motor Co., the
Boeing Co., and the Timken Roller Bearing Co.
J. C. Grosskreutz
Midwest Research Institute,
Kansas City, Mo.
University of Washington (University of Washington) pursuant to License Agreement. No further repr
Fatigue Crack Growth in Structures
University of Washington (University of Washington) pursuant to License Agreement. No furth
This page intentionally left blank
Limitations of Fatigue-Crack Research in
the Design of Flight Vehicle Structures
REFERENCE: R. H. Christensen and M. B. Harmon, "Limitations of

Fatigue-Crack Research in the Design of Flight Vehicle Structures,"
Fatigue Crack Propagation, ASTM STP 415, Am. Soc. Testing Mats.,
1967, p. 5.
ABSTRACT: This paper surveys the current state of fatigue-crack re-
search and the effect of this work on structural design. The subject areas
discussed include: (1) fatigue-crack growth equations, (2) multiaxial
strain effects, (3) temperature effects, (4) frequency effects, (5) aniso-
tropic effects, (6) variable stress field effects, (7) fractography, (8) crack
growth under random loads, and (9) critical elements and fracture
modes. The current limited application of the available fatigue informa-
tion to design is recognized. Practical objectives for fatigue research
from the standpoint of structural design are itemized.
KEY WORDS: fatigue (materials), crack propagation, anisotropy, struc-
tural design, flight vehicles, aircraft, rockets, missiles
It is well known that the phenomenon of progressive fracture caused

by fatigue can severely degrade the strength and life of flight-vehicle
structures. Therefore, it is very important to collect and provide fatigue
data in a form that is useful to designers.
Fatigue cracks have been recognized as a serious problem in design
for the past 30 years. Deforest and Magnuson [I]3 supervised one of the
earliest investigations to study the growth of fatigue cracks. The experi-
ments were the first to indicate that fatigue studies were often not con-
clusive. The lack was caused primarily by the failure to distinguish be-
tween (1) the stresses and cycles that initiate cracks and (2) the stresses
and cycles that propagate the cracks to complete fracture. Deforest's
experiments set about to determine these distinctions and, in addition, to
define the residual strength of the metal once fatigue cracks had been
1
Chief, Structural Research Branch, Missiles and Space Systems Div., Douglas
Aircraft Co., Inc., Santa Monica, Calif.
2
Design specialist, Vehicle Systems Branch, Missiles and Space Systems Div.,
Douglas Aircraft Co., Inc., Santa Monica, Calif.
3
The italic numbers in brackets refer to the list of references appended to this
paper.
5
formed. The experiments were extremely well planned and achieved the
objectives of the research.
Currently, the problem is that many fatigue evaluation programs
either fail to develop accurate estimates of probable structure life or fail
to define the rates of nonlinear cumulative damage. Current experiments
usually establish the fatigue characteristics of structural components by
subjecting test specimens of the components to repeated or alternating
TABLE 1—Fatigue-crack growth equations.
Date Investigators Equation Reference
1935 to 1936.. . DeForest and No equation, but one of earliest ex- 1

Magnuson periments designed to measure
fatigue-crack growth.
1938 . Langer and No equation, but first suggestions 2
Peterson concerning the inclusion of crack
propagation data into cumulative
damage rules.
1953 . Head dl/dn = 0 3 / 2 3
1956 . McClintock [d(l/h)W(6h)] = l/47/[ln h/2P - 4
J£(l - 4p 2 /A 2 )]---in torsion
1957 to 1958 . . Frost and Dug- dl/dn = constant- • -constant stress 5
dale
dl/dn = Atrsl- • -constant load 6
1958 . McEvily and Illg dl/dn = 2«/3/2, where « = M/~ 3 / 2 7
1961 . Hardrath and logi,,-1 (0.0051 Knanft - 5.472 - 8
McEvily 34/(Knffnet - 34))
1960 . Schijve dl/dn = Klaa^ 9
1954 10
&
1956 . Weibull dl/dn = KW<n - • -constant cyclic <r 11
1960 12
[d(l/w)]/dn = AW)/[(1 - I/*)']--
constant cyclic load
1961 . Valluri dl/dn = Kl/[(\ - //^)4]o-crk 13
(1 - <^/<rcrk)2(l - *)2
1961 . Denke and dl/dn =fi(///e)/[(l- ///c)4] 14
Christensen 15
1961 . Paris, Gomez, dl/dn = Ka = trVUZx 16
and Anderson 17
1963 . Liu dl/dn = c^l 18
1965 . Manson dl/dn = c(AePV/)s 19
loads. Design data are then collected to evaluate the effect of the alter-
nating loads on the proposed structure. Materials are evaluated by the
same method. Agreement between the calculated results and the actual
service results has been poor in both cases. One reason is that the load
cycles in these programs are usually applied in a regular manner, whereas
in actual practice, the fluctuating loads on the structure are random. In
addition, laboratory tests are conducted for the easier, less expensive,
and faster evaluation condition that is represented by uniaxial loading.
Actual flight structures are subject to multiaxial loads and strains. The
CHRISTENSEN AND HARMON ON FLIGHT VEHICLE STRUCTURES 7
FIG. 1—Agreement between test and theory—cycles required to generate

cracks of length 1.
FIG. 2—Variation in crack rate with length of crack.
characteristic properties of uniaxial and biaxial loading can be very dif-

ferent.
The purpose of this paper, therefore, is to review current fatigue-
crack technology and to suggest a means to improve fatigue design data.
The paper discusses the generation
University of Washington (University ofand application
Washington) oftoempirical
pursuant equa-
License Agreement. No further
tions to demonstrate how well fatigue behavior is predicted. Temperature
and frequency effects are also discussed. The phase of observable crack-
ing in service as a function of temperature is illustrated, as well as the
crack rate as a function of material anisotropy, and the growth of cracks
in realistic structure stress fields. In addition, the paper outlines the re-
search still needed to insure accurate prediction of structural behavior
under random loads through simple, quick tests with discrete loading.
The paper also outlines the research necessary to increase the usefulness
of fractography as applied to vehicle structures problems. Finally, the
goals and objectives of fatigue tests from the standpoint of design are
reviewed, since a survey of the literature shows that few investigators
ever present the real objectives or purposes of the investigation reported.
Without a knowledge of these test goals and objectives, useful design
guides can not be formulated.
FIG. 3—Stress intensity versus cracking rate (small change in stress and cyclic
range).
Fatigue-Crack Growth Equations

Without exception, the fatigue-crack equations known to the authors
appear to be valid and express the accumulation of damage when ap-
plied within the limits of the particular test conditions from which the
equations were derived. A representative selection of the equations ap-
pear in Table 1, while Fig. 1 shows the good correlation achieved be-
tween test and theory in the prediction of the number of cycles required
for the propagation of cracks to different lengths [15]. The majority of
equations developed by fatigue investigators show similar correlation.
Within the restrictions (that is, stress ranges), in which the equations
were formulated, they may be applied to other materials, and meaningful
comparisons can be made so that candidate metals may be selected for
use in Downloaded/printed
a specific design.byTo the authors knowledge, however, no single
methodUniversity
has been of devised
Washingtonto (University
predict theof characteristics of allto ofLicense
Washington) pursuant the Agreeme
variables which must be considered for design. The fatigue crack
growth rates in aluminum panels (Fig. 2), show that the rate of growth of
a crack at any time is dependent on the instantaneous length of the crack
at that particular point in time. The crack growth rate, dl/dn, is the slope
of the / versus n data curve at the particular point in tune. All the tests
were conducted at a single stress ratio. The example shows that the rate
FIG. 4—Variation in crack rate with stress intensity.
FIG. 5—Crack growth rate as a junction of range of stress (It = 2.0 in.).
of growth for a crack of a specific length is accelerated twenty-fold if

the applied cyclic stress is doubled.
Fatigue crack growth equations of other investigators can be found
in Refs 76-20.
A few years ago a promising technique was introduced [16, 17] which
attempted to normalize
University as much
of Washington data ofasWashington)
(University possible from different
pursuant contrib-
to License Agreement. No fur
utors. The method plots (Fig. 3) the crack growth rate, dl/dn, as a
function of stress intensity K, which is proportional to the product of

the maximum cyclic^ stress and the square root of half the cracklength.
Originally this plot was for a limited range of maximum stress levels, and
for a rather narrow variation between minimum and maximum stress.
Unfortunately, many recent studies and comments on experimental data
seem to have overlooked this limitation. To improve the method, a de-
crease in the relatively small scatter was subsequently achieved by the
introduction of the parameter &K, that is, the stress intensity calculated
from the maximum stress in a cycle, less the stress intensity based on the
minimum stress in the cycle. The limitations on the universality of Fig.
3 are clearly demonstrated by some recent data (Fig. 4), taken from an
Air Force contract [22]. The maximum test stress levels were varied by
a factor of 2, and a large change was made in the cyclic stress ratios. The
FIG. 6—Crack growth rate as a function of stress field.
ratio of minimum to maximum stress varied from 10 to 65 per cent. It

can be seen that dl/dn as a function of the AK level can not be represented
as a single line. In fact, dl/dn can change by a factor of 3 for a twofold
change in stress of a given level of A£. Another possible point of confu-
sion for designers unfamiliar with the stress intensity criterion concerns
the variation of growth rate with stress level that can occur at a constant
level of stress intensity. For example, in Fig. 4 there is apparently a
higher crack growth rate at the 30,000 psi maximum stress curve at a
ratio of 0.6 as compared to the growth rate for the same maximum stress
at a ratio of 0.1. This is untrue, but, plotted against AK, it appears to be
true. This occurs because AK is a ratio of the maximum stress times the
square root of half the crack length, and the two points in the above
example are at by
Copyright far ASTM
different
Int'l crack lengths
(all rights for the
reserved); Monsame
Dec value of AK.
7 14:40:45 EST 2015
To give a direct comparison,
Downloaded/printed by the same data are replotted in Fig. 5 for
the instant in time
University when the(University
of Washington value of ofk Washington)
= 2 in. It pursuant
can be toseen that Agreement.
License the No
crack growth rate increases as the maximum stress increases. Plots for
shorter or longer crack lengths show a family of curves. It can also be

noted that the 15,000, 20,000, 25,000, and 30,000 psi constant stress
level curves are the reverse of those in Fig. 4.
It has been the purpose of this discussion to show some, features of the
stress intensity approach for the presentation of fatigue-crack growth
data. These features are sometimes omitted in reports and have perhaps
led to misinterpretation by designers. Nevertheless, there is great promise
in this technique. However, caution must be exercised to define the
limits of the data reported clearly in order to avoid errors in flight-
vehicle design.
FIG. 7—Effect of temperature.
The Effect of Multiaxial Strain

In service, the majority of flight vehicle structures are subjected to
some degree of multiaxial or biaxial stress. However, the major empha-
sis in the evaluation of structures for fatigue resistance in the laboratory
is on uniaxial loading methods. In many hardware applications, the
indiscriminate use of uniaxial data will result in unconservative designs
and premature structural failure.
Metal fatigue under multiaxial strain is a complex phenomenon.
Losses in ductility and anisotropy, as well as texture hardening of ma-
terials, contribute to the complexity of the problem, but the effects of
these factors on biaxial fatigue properties have not been properly in-
vestigated. Empirical characteristics defined for one material do not
necessarily apply for other materials since: (1) some materials have
greater tensile strengths in the longitudinal rolling direction, while in
others, it is in the transverse direction; (2) many unflawed materials with
higherDownloaded/printed
biaxial strengthsbythan uniaxial strengths show a higher uniaxial
than biaxial strength
University when flawed;
of Washington andof (3)
(University for certain
Washington) stress
pursuant to ratios,
License the
Agreement. No
maximum principal strain may initially be normal to the weakest axis of
the material but, when the stress ratio is altered, become normal to the
strongest strength axis of the material. In a few cases in the past, al-
though the material property peculiarities were known, they were ne-
glected in design with catastrophic results to the flight structure.
A practical example is given in Fig. 6. The figure shows recent data
FIG. 8—Fatigue life as a function of test temperature (typical for most metals).
FIG. 9—Effect of frequency at elevated temperature.
obtained in a Douglas independent research program and presents crack

growth as a function of multiaxial strain. The crack growth rate is plotted
as a function of &K. Differences in stress intensity for both uniaxial and
biaxial states of strain and the maximum stresses in both cases are the
same. Copyright
The fatigue crack
by ASTM Int'lgrowth
(all rightsrate is much
reserved); greater
Mon Dec for biaxial
7 14:40:45 strain,
EST 2015
which Downloaded/printed
approximates the by actual conditions more closely than uniaxial
strain. University
The designer must (University
of Washington thereforeofbe made aware
Washington) pursuantofto the restrictions
License Agreement. No further
associated with the application of uniaxial data to design.
The Effect of Temperature

The effect of temperature on fatigue and fatigue-crack growth must
also be considered in design. An example is the S-N curve for a titanium
alloy shown in Fig. 7. As would be expected, the fatigue life and physical
properties degrade as temperature increases. All tests were conducted at
a constant test frequency and a constant strain rate. Figure 8 shows that
fatigue characteristics improve as the temperature is lowered to the
cryogenic region. In general, most, but not necessarily all, materials be-
have this way.
FIG. 10—Growth of fatigue cracks as a function of temperature (schematic).
The Effect of Frequency

An additional point to be considered in design is the effect of fre-
quency. The previous discussion confined itself solely to the effect of
temperature on fatigue life and crack growth rate. It is important to
realize the effect of temperature at different strain rates on fatigue
characteristics. Figure 9 shows that, at elevated temperatures, the fatigue
lives achieved at low frequencies and slow strain rates are far different
from the fatigue lives achieved at high frequencies. Other damaging
mechanisms such as creep and creep cracking act jointly with the fatigue
action.
It is of interest to note that if laboratory data on the rate of crack
growth are plotted as a function of stress cycles and temperature, the
result is a series of curves similar to those shown in Fig. 10. The curves
show that at the higher temperatures the crack nucleation period, that
is, theCopyright
process by
of ASTM
getting Int'l
the (all
crackrights
started, is short,
reserved); Mon while
Dec the crack
7 14:40:45 EST 20
growthDownloaded/printed
propagation periodby is much greater. As temperatures are de-
University of Washington (University of Washington) pursuant to License Agreeme
creased, the nucleation period increases and the propagation period
T4 FATIGUE CRACK PROPAGATION
FIG. 11—Anisotropic flaw growth.
FIG. 12—Fracture strength as a function of flaw orientation.
decreases. In general, critical crack lengths decrease as temperature de-

creases. A knowledge of such trends would be vital in the design of
cryogenic pressure vessels for space vehicles.
The Effect of Anisotropy

Another design consideration
Downloaded/printed by is the significant effect of anisotropy on
fatigueUniversity
crack growth, that is,(University
of Washington how rapidly the fatigue
of Washington) cracktopropagates
pursuant License Agreement. N
as a function of the crack's orientation with respect to the principal
rolling direction of the commercial sheet used to build the actual struc-
ture. For example, Fig. 11 shows crack growth as a function of stress
cycles for a crack oriented at different directions with respect to the
rolling direction of the sheet. It is not possible to predict how a crack will
start in service, or if the orientation of the rolling direction on a particular
FIG. 13—Growth of cracks as a function of type of test (laboratory) loading.
FIG. 14—Typical damage accumulation for laboratory tests of complex struc-

ture.
sheet will be normal or parallel to the principal loading direction. Further,

the cracks will not always be perpendicular or parallel to the rolling
direction. It is hoped that additional studies will be undertaken to in-
vestigate the basic mechanism of the phenomenon of the fatigue crack
growthCopyright
as a function
by ASTMof Int'l
the (all
various crystallographic
rights reserved); Mon Dec slip systems
7 14:40:45 in 2015
EST the
basic metal.
University of Washington (University of Washington) pursuant to License Agreement. No fu
Figure 12 shows the differences in the residual strengths, or rupture
strengths, for panels that contain flaws. The flaws were oriented at
various angles to the principal rolling direction. The tests were con-
ducted on aluminum alloy at room and at liquid nitrogen temperatures
with crack angles that ranged from 0 to 90 deg from the rolling direction.
The differences in strength from 31,000 psi to over 36,000 psi in the
aluminum alloy are significant and must be accounted for in design.
The Effect of Variable Stress Fields

Another problem in the use of fatigue characteristics derived from
laboratory tests for design is the result of the technique used to obtain
the growth of the fatigue crack. Figure 13 shows a comparison of the
crack growth as a function of stress cycles where: (1) the cyclic growth
N, STRESS CYCLES
FIG. 15—Damage rate as a function of design and working stress.
section stress or the maximum cyclic load is constant throughout the

tests, and (2) a constant cyclic net section stress is maintained by periodic
reductions of the test panel load. For the former, the graph in Fig. 13
shows an exponential growth in the crack length and in the rate of
damage accumulation. In the latter, the author and many other observers
have noticed the rather linear crack growth shown in the figure. This
difference in methods must be understood. Many analytical developments
and equations have been proposed as representative of the actual be-
havior in service, based on the premise that a small crack in a large
structure approximates a nominal stress field; however, this is not neces-
sarily true. Figure 14 shows actual test data from a large built-up wing
structure TheASTM
[21]. by
Copyright damage accumulation
Int'l (all shown
rights reserved); is typical
Mon Dec of laboratory
7 14:40:45 EST 2015
tests ofDownloaded/printed
complex structures.
by The curve has been normalized for three dif-
ferent University
tests at three different (University
of Washington stress levels. The load pursuant
of Washington) factor differences are
to License Agreement. No fur
0.35, 0.625, and 1.0 on a mean load factor of 1.0. The crack grows in an
accelerated fashion. The average growth rate, that is, average accumula-
tion of physical damage, is exponential.
Another important aspect of the fatigue crack propagation problem
is the damage rate as a function of design as well as working stress
(Fig. 15). A nucleation period occurs wherein no crack growth is ob-
FIG. 16—Tension area cracked as a function of elapsed fatigue life.
FIG. 17—Fatigue crack growth in titanium tension field web.
served; this is followed by the crack propagation period. For a given

design with a nominal stress concentration factor, K f , of 1.4, at 45,000
psi maximum stress, there is an increase in life over a 90,000 psi maxi-
mum stress.
Copyright The percentage
by ASTM Int'l (all rights of life Mon
reserved); in the
Dec 7crack
14:40:45propagation
EST 2015 period is
also greater at the lower
Downloaded/printed by than at the higher stress level. However, for a
designUniversity
with aofhigh stress
Washington concentration
(University factor,
of Washington) pursuant such asAgreement.
to License 3.0, and with reproductions
No further a aut
reduced working stress for increased safety, the result is a short nuclea-
tion period and an extremely long crack propagation period (Fig. 15). An
odd conclusion could be drawn in that a high stress concentration is not
desirable in design but would increase safety because a longer inspection
period to find the crack would exist. This could obviously be carried to
extremes. A poorly designed joint at low and inefficient working stresses
could result in an almost 100 per cent crack propagation period; but its
life would be unacceptably low.
FIG. 18—Fatigue crack striations due to programmed stress cycling.
Figure 16 shows normalized data of tension area cracked as a function

of fatigue life. The design with the low stress concentration factor will
naturally have a greater life but its propagation period is approximately
10 per cent of its over-all life. The design with the high stress concentra-
tion factor has a crack propagation period of approximately 90 per cent
of its over-all life. The damage rate is therefore shown to be strongly in-
fluenced by design as well as by stress level.
In Figure 17, the crack growth is plotted for some annealed titanium
tension field panels. The crack growth is measured at ±6000-psi princi-
pal stress as a function of the stress cycles and appears to be linear.
This is not contradictory to the previous discussion on exponential
growth for simple panels. In the tension field web, the crack is growing
into a diminishing stress field at the heavy caps. This example reinforces
the importance of evaluating design conditions to use fatigue information
properly in contrast to attempting to establish rigid rules.
Fractography
Fractography is another research area to be considered for use in im-
proving flight-vehicle structures. During the past decade, interest has
grown in the use of the electron microscope for crack-growth studies.
The emphasis has been on the investigation of laboratory samples, but
it has been hoped that the method could also be used in post-fracture
FIG. 19—Striation spacing, microscopic versus macroscopic growth.
analysis of actual flight-vehicle components. Some studies have even

identified the occurrence of a single application of a load cycle.
In Figure 18 [22], the micro-progressive growth for an aluminum
alloy fatigue fracture is shown from a replication of the fractured sur-
face of the metal. The figure shows the increment in length that the
crack grows with the application of each load. In addition, the degree
of growth is shown as a function of the applied stress level. In the test,
three levels of stress were programmed for a duration of 10 cycles each.
The growth spacings are small at low stresses and gradually increase as
the stress increases. Only the high and medium stress growth striations
can be identified. The low stress did not propagate the crack. Investiga-
tors originally hoped that the spacing between striations could be asso-
ciated with the applied stress, but it now appears that both the stress and
the instantaneous crack length must be known.
Figure 19 is a plot of growth in microns, measured by the electron
University of Washington (University of Washington) pursuant to License Agreement. No furt
microscope as a function of the macro measurement, calculated from
FATIGUE CRACK PROPAGATIOGATIONN
macrogrowth measurements. As stress level increases, so does the growth

or spacing per cycle. There is fair correlation between the measurements
on both the macroscopic and microscopic scales for low stress regions.
As the stress increases, the 1-to-l ratio no longer applies. At high stress,
the micro-progressive growth does not propagate evenly. There are
small regions of uniform growth and large bursts in growth distance. The
average distance per cycle calculated for n cycles therefore appears to
be much greater than the individually observed spacings between the
striations. The striations represent temporary arrest lines in the process
of cracking. In addition, striation spacings can be the same for combina-
tions of stress level and stress range. At the current level of fractography
FIG. 20—Comparison of predicted and measured crack growth under random

vibration.
development, it is quite difficult to identify operating stress levels in a

post service fracture analysis. More work is required to develop addi-
tional techniques.
Crack Growth Under Random Loading

To improve the accuracy of design techniques for flight vehicles, it
is necessary to define the response of structures to random alternations
of stress approximating the disordered load distribution encountered
within the real vibration, acoustic, and turbulent atmospheric environ-
ments of the vehicle. Several analytical methods have been developed
to calculate or predict the characteristics under random loading from the
resultsUniversity
under sine wave or (University
of Washington discrete loading. Therepursuant
of Washington) are two schoolsAgreement.
to License of No f
thought on the subject. One school believes that an equivalency exists in
random loading if the proper sine wave or single-level discrete load test
could be selected. The other school holds that a complete series of tests
must be performed to realistically integrate the damage produced by a
multiload level situation. Figure 20 shows a comparison of predicted
fatigue-crack growth data with data measured in a pure random load test
[23]. This is a pure random vibration fatigue test. There is close agree-
ment between the predicted and experimental results. The test shows
that a series of discrete load tests can be used to predict the behavior of
structures subjected to random loads. No single stress appears to pro-
duce damage at a rate equal to that measured under random loading.
FIG. 21—Fracture modes.
Additional experimentation is required to develop accurate techniques

for the simulation of the damage from the real environment.
Critical Element and Fracture Mode

In addition to the subject areas previously mentioned, the problems
created by complex structures that possess many modes of fracture
should be investigated in future vehicle programs. It should be deter-
mined if a series of discrete load tests can predict (1) the fatigue life,
(2) the crack propagation characteristics, and (3) the critical fractured
member in a large built-up structure for conditions of pure random load-
ing. Figure 21 shows a panel test in which there was one starter notch
and, therefore, only
Copyright by ASTMoneInt'l
location inreserved);
(all rights which fracture
Mon Dec could occur.
7 14:40:45 ESTAt2015
each
stress level, the locationbyof the fracture is the same. However, it is known
Downloaded/printed
that inUniversity
complexofbuilt-up structures
Washington (Universitysuch as webs, splices,
of Washington) and
pursuant to spar Agreement.
License mem- No furth
bers, the stress level varies when different members in the structure be-
FATIGUE CRACK PROPAGGATIONATION
come the critical member. For example, in Fig. 21, the spar fails at the
high load level and the web cracks at the low load level. Test panels
simulating actual complex structures should be evaluated under com-
pletely random loads to guide vehicle design.
Conclusions
Additional understanding of the micro-progressive growth of frac-
tures caused by fatigue action is needed in the field of structural design
and analysis. The knowledge can be acquired by both experimental and
analytical means. The greatest contribution fatigue investigators can
make are in the following areas:
Assist the designer in the selection of the most promising materials
for a particular design.
Assist the designer in the evaluation of the behavior of candidate de-
signs in service (comparative tests of various joint designs).
Provide data on the various modes of fatigue-cracking as a function of
various structural environments.
Provide data on the fatigue-critical element of a composite design
as a function of the level of repeated stresses.
Contribute to a better understanding of the basic mechanism in fatigue.
Assist in the formulation of semiempirical formulas for use in struc-
tural design.
Provide experimental samples and evidence for microscropic examina-
tion (such observations are useful in service fracture analyses).
Evaluate the effects of various artificial and natural environments on
materials and structures (for example differences between simple and
complex loading).
Finally, researchers in the field of fatigue crack growth are cautioned
to protect unwary designers by clearly reporting all basic assumptions
and limitations associated with future research discoveries.
A cknowledgment
The subject matter presented in this paper represents a general survey
of current fatigue-crack research; much of the information is from the
open literature. However, a large portion of the data represent the re-
sult of specific programs conducted by the Douglas Aircraft Co., Inc.,
Missile and Space Systems Div., under company-sponsored independent
research and development funds (Account No. 81391-010).
References
[/] A.Copyright by ASTM
V. Deforest Int'lW.
and F. (all Magnuson,
rights reserved);
"TheMon
RateDecof7 14:40:45
Growth EST 2015
of Fatigue
Downloaded/printed
Cracks," by
Journal of Applied Mechanics, March, 1936, p. A-23.
[2] B.University
F. Langerof and R. E. Peterson,
Washington (University"Fatigue Failurepursuant
of Washington) from Stress Cycles
to License of
Agreement. No fu
Varying Magnitude," Journal of Applied Mechanics, December, 1937, p.
A-160.
[3] A. K. Head, 'The Growth of Fatigue Cracks," The Philosophical Magazine,

Vol 44, Series 7, 1953, p. 925.
[4] F. A. McClintock, "The Growth of Fatigue Cracks Under Plastic Torsion,"
International Conference on Fatigue of Metals, IME and ASME, London
and New York, September through November, 1965.
[5] N. E. Frost and D. S. Dugdale, "Fatigue Tests on Notched Mild Steel Plates
with Measurements of Fatigue Cracks," Journal of the Mechanics and Physics
of Solids, Vol 5, 1957, pp. 182-192.
[6] N. E. Frost and D. S. Dugdale, 'The Propagation of Fatigue Cracks in Sheet
Specimens," Journal of the Mechanics and Physics of Solids, Vol 6, 1958, pp.
93-110.
[7] A. J. McEvily and W. Illg, "The Rate of Fatigue-Crack Propagation in Two
Aluminum Alloys," NACA TN4394, September, 1958.
[8] H. F. Hardrath and A. J. McEvily, "Engineering Aspects of Fatigue Crack
Propagation," Proceedings of the Crack Propagation Symposium, Vol II,
Cranfield, England, October, 1961.
[9] J. Schijve, "Fatigue Crack Propagation in Light Alloy Sheet Material and
Structures," NLL Report, MP195, National Cuchtvaartiaboratorium, Amster-
dam, The Netherlands, August, 1960.
[10] W. Weibull, "Size Effects on Fatigue Crack Initiation and Propagation in
Aluminum Sheet Specimens Subjected to Stress of Nearly Constant Ampli-
tude," FFA Report 86, June, 1960.
[11] W. Weibull, "The Propagation of Fatigue Cracks in Light-Alloy Plates,"
SAAB TN 25, January, 1954.
[12] W. Weibull, "Effects of Fatigue Crack Length and Stress Amplitude on
Growth of Fatigue Cracks," FFA Report 65, Stockholm, May, 1956.
[13] S. R. Valluri, "A Unified Engineering Theory of High Stress Level Fatigue,"
ARL Tech, GALCIT SM61-1, October, 1961.
[14] Douglas Aircraft Company, Inc., "Crack Strength and Crack Propagation
Characteristics of High Strength Metals," Air Force Technical Document
ASD TR 61-207, January, 1961.
[15] Douglas Aircraft Company, Inc., "Notch Resistance and Fracture Toughness
Characteristics of High Strength Metals," Air Force Technical Document
ASD-TDR-63-494, September, 1963.
[76] P. C. Paris, M. P. Gomez, and W. E. Anderson, "A Rational Analytic Theory
of Fatigue," The Trend in Engineering, Vol 13, No. 1, University of Wash-
ington, January, 1961.
[17] P. Paris and F. Erdogan, "A Critical Analysis of Crack Propagation Laws,"
Transactions of ASME, Journal of Basic Engineering, December, 1963, pp.
528-534.
[18] H. W. Liu, "Fatigue Crack Propagation and Applied Stress Range—An En-
ergy Approach," Journal of Basic Engineering, Transactions of ASME,
Series D, Vol 85, 1963, p. 116.
[19] S. S. Manson, "Fatigue: A Complex Subject—Some Simple Approximations,"
Experimental Mechanics, July, 1965.
[20] S. T. Rolfe and W. H. Munse, "Fatigue Crack Propagation in Notched Mild
Steel Plates," The Welding Journal, June, 1963, pp. 252-255.
[21] R. E. Whaley, M. J. McGuigan, Jr., and D. F. Bryan, "Fatigue-Crack-Prop-
gation and Residual-Static-Strength Results on Full-Scale-Transport-Airplane
Wings," NACA TN 3847, December, 1956.
[22] A. Phillips, "Improved Electron Fractographic Techniques," Contract No. AF
33(615^-3014, 1966.
[23] R. H. Christensen, 'The Growth of Fracture Under Random Cyclic Load-
ing," International Conference on Fracture, Sendai, Japan, September, 1965.
University of Washington (University of Washington) pursuant to License Agreement. No further
FATIGUE CRACK PROPAGAGATIONTION
DISCUSSION
Gerhard H. Jacoby1—Program tests and tests with a random sequence

of load cycles, performed in Germany with test pieces (Kth =3.1) from
2024 aluminum alloy, proved that the percentage of the crack propaga-
tion stage of the total life depends also on the sequence of load cycles,
even if different program schemes and the random sequence lead to the
same fatigue life. This fact may be of great practical importance, if
inspection periods have to be assessed on the basis of such tests.
1
German Research Association and Aerospace Sciences (OVL), at present
Columbia University, New York City.
Crack Propagation and Residual Static
Strength of Fatigue-Cracked Titanium
and Steel Cylinders
REFERENCE: W. J. Crichlow and R. H. Wells, "Crack Propagation

and Residua] Static Strength of Fatigue-Cracked Titanium and Steel
Cylinders," Fatigue Crack Propagation, ASTM STP 415, Am. Soc.
Testing Mats., 1967, p. 25.
ABSTRACT: Flight safety of the U. S. supersonic transport (SST) re-
quires design data on the rate of crack propagation and the residual
static strength of fatigue cracked cylinders. To provide these data an
experimental test program was conducted on stiffened cylinders and curved
and flat panels of titanium sheet (8Al-lMo-lV duplex annealed) in the
environments representative of SST operations (—110 F, room tempera-
ture, and +550 F) and on similar specimens of a candidate steel sheet
material (PH-14-8Mo SRH 1050) at room temperatures. Cyclic pressure
loading was applied to the cylinders and curved panels. Variation of hoop
cross-section area and spacing on longitudinally stiffened cylinders pro-
vided design criteria for the crack-arresting properties of hoops in typical
fuselage applications. Tension tests of flat panels with and without longi-
tudinal straps provided data on material residual strengths and on the
strap attachment qualities in the presence of transverse fatigue cracks. The
experimental data were analyzed by means of several analytical formulas
so that the design data could be interpreted and presented in compact and
usable form.
KEY WORDS: fatigue (materials), crack propagation, titanium alloys,
steels, cylinders, low temperature, high temperature, pressure cycling,
structural analysis
Nomenclature
A, Ah , Ast Area, area of hoop, area of stiffener, in.2
a Spacing of hoops, in.
a Appendix I—increment of crack growth from initial X0 to
critical, Xcr
1
Department manager, Advanced Materials and Structural Mechanics Depart-
ment, Lockheed-California Co., Burbank, Calif.
2
Research and development engineer, Structures Test Group, Structures and
Materials Laboratory, Lockheed-California Co., Burbank, Calif.
25
b Spacing of straps or stiffeners, in.

Co Buckling correction coefficient
d Density, lb/in.3
E Modulus of elasticity, psi
e Ultimate tensile elongation, in./in.
Ftu Ultimate tensile strength, ksi
Fty Yield tensile strength, ksi
Fa Allowable gross area stress, ksi
Fh Allowable hoop tension stress, ksi
F0 Residual hoop tension stress, ksi, of unstiffened cylinder
with crack size desired for use in stiffened cylinder
f» pR/t, hoop tension stress, ksi
./max Maximum cyclic stress, ksi
7m in Minimum cyclic stress, ksi
Kc Fracture toughness constant, ksi in.1/2
Kt Theoretical stress concentration factor
•Kw ([1 - X/W\/[\ + X/W])1'2 finite width correction factor
L Length of cylinder, in.
L* Effective length beyond which initial stress distribution of
the uncracked specimen prevails
m0 Coefficient in curvature equation
N Number of cycles
P Internal pressure (differential), psi
R Radius of cylinder, in.
Rf /minX/rax , stress ratio
r AX/&N, rate of crack growth, in./cycle
T Temperature, deg F
t Thickness, in.
W Width of flat panel, in.
X Crack length, in.
X 0 Initial crack length, in.
xcr Critical crack length, in.
V Poisson's ratio
p' Neuber's constant, in.
f,1? Cartesian coordinates in crack growth rate analysis
*iTO Function in crack growth rate analysis, increment of crack
growth from initial (X0} to critical (Xer) (Appendix I,
Eq 26)
$2(*0) Function in crack growth rate analysis, reciprocal of critical
residual strength, Fcr , at a specific initial crack length, X0
* CopyrightRatio of maximum stress level for nonpropagating crack to
bycritical residual
ASTM Int'l (all rights strength, , at7 14:40:45
FcrDec
reserved); Mon a specific initial crack
EST 2015
length, X
University of Washington0 (University of Washington) pursuant to License Agreement. No further reprod
m,n Exponents in crack growth rate analysis, empirically deter-
mined
CRICHLOW AND WELLS ON TITANIUM AND STEEL CYLINDERS 27
Material Designations
8-1-1 DA Titanium 8Al-lMo-lV duplex annealed
8-1-1 TA Titanium 8Al-lMo-lV triplex annealed
13-11-3 Titanium 13V-llCr-3Al (B 120 VGA)
14-8 Stainless Steel PH 14-8 Mo SRH 1050
The flight regime of the supersonic transport requires special attention
to the pressure cabin design to prevent catastrophic failure. Fail-safe
design principles and eminently satisfactory structural configurations
have been developed over a number of years for the contemporary sub-
sonic transports. The properties of aluminum materials and the configura-
tion of structural details have demonstrated a high degree of damage
tolerance for subsonic operating conditions [/].3 However, the use
of new materials, required for the temperatures of supersonic flight, pose
the familiar problem of generating a considerable body of engineering
data:
1. on the basic properties of the new materials,
2. on the behavior of these new materials in SST structural configura-
tions, and,
3. on the effects of the environmental temperatures on the properties
of material and structure.
This paper describes a portion of the research and development pro-
grams recently undertaken to provide the necessary engineering data for
the damage tolerance design of the supersonic transport pressure cabin.4
The specific objectives of the program to be described were [2,3]:
1. to provide comparative static fracture tolerance and rate of crack
growth data on a candidate stainless steel (PH14-8Mo SRH 1050) and
titanium (8Al-lMo-lV duplex annealed and 13V-llCr-3Al) materials
in flat and cylindrical specimens,
2. to provide fundamental data on the size (cross section) and spacing
of hoops required to prevent unstable crack growth in longitudinally
stiffened steel and titanium cylinders,
3. to investigate these properties on the chosen titanium material
under the environmental extremes of temperature ranging from —110 F
to +550 F,
4. to investigate the effect of curvature on the residual static strength
and rate of crack propagation of the titanium material, and
3
paper.
4
Some of the data presented in this paper were developed on the National
Supersonic Transport Research Programs sponsored by the Federal Aviation
AgencyCopyright by ASTM Int'l
with administrative and(alltechnical
rights reserved);
supportMon Dec 7by
provided 14:40:45 EST 2015 of
the Department
Defense,Downloaded/printed
Research and Technology
by Division, AFSC. Contributing basic research
and technical support
University was also(University
of Washington provided by the Nationalpursuant
of Washington) Aeronautics andAgreement.
to License Space No furth
Administration. Permission of the FAA Supersonic Transport Developmsnt Office
to publish these data is gratefully acknowledged.
5. to investigate the effects of various attachment methods: rivets,

spot and continuous roll seam weld, and intermittent and continuous
fusion welds on the residual static strength of strap reinforced flat panels
of steel and titanium materials.
The paper presents a brief outline of the test program, a description
of the specimens, test equipment and procedures, and provides the re-
sults of the experimental program along with comparisons of the experi-
mental results with analytical representations of the behavior of some
of the more important parameters.
TABLE 1—Design configuration for stiffened cylinders, Titanium 8Al-lMo-lV

duplex annealed.
Skin Z-Stringer Hoops (Straps)
Design Configuration Grain
Figs. 2, 3, and 4 Gage, Direc- Gage, Spac- Gage, Width, Spacing,
in. tion in. ing, in. in. in. in.
1 0.050 Ta 0.032 2.5 0.025 1.0 20.0

2 0.050 T 0.032 2.5 0.050 1.0 20.0
3 0.050 T 0.032 2.5 0.100 1.0 20.0
4 0.050 T 0.032 2.5 0.100 2.0 20.0
5 0.050 L» 0.032 2.5 0.050 1.0 20.0
6 0.050 T 0.032 2.5 0.025 0.33 6.67
7 0.050 T 0.032 2.5 0.100 0.33 6.67
8 0.050 T 0.032 5.0 0.050 1.0 20.0
9C . 0.050 T 0.032 2.5 0.050 1.0 20.0
" T = transverse (grain direction normal to hoop tension).
6
L = longitudinal (grain direction parallel to hoop tension).
c
Crack located at stringer.
Outline of the Test Program

The test program consisted of 61 crack propagation tests on cylinders
and large curved panels of titanium 8-1-1 DA sheet material, one crack
propagation test on a curved panel and one on a cylinder of titanium
13-11-3 (B 120 VGA) and 23 crack propagation tests on similar speci-
mens of stainless steel 14-8 under cyclic internal pressure. In addition,
38 static tear tests were conducted on flat specimens of titanium 8-1-1
DA with a central crack under uniaxial tension load, and 24 static tear
tests were conducted on similar specimens of stainless steel 14-8. Standard
tensile coupon data provided mechanical properties of the materials.
Description of the Test Specimens
Flat Panels
Forty-eight flat-sheet specimens, 12 by 36 in. with straps, were fabri-
cated from titanium
Copyright 8-1-1
by ASTM Int'lDA and reserved);
(all rights stainlessMon
steel
Dec14-8. Tables
7 14:40:45 EST 1 and 2
2015
give theDownloaded/printed
design configurations
by for these specimens. Figure 1 shows a
typical flat panel of
University with straps. (University
Washington Twelve flat panels 9 by
of Washington) 27 in.towith
pursuant noAgreement.
License straps No furthe
were fabricated identically from 0.030 gage titanium 8-1-1 DA.
TABLE 2—Design configuration for stiffened cylinders, Stainless Steel PH 14-8Mo

SRH 1050.
Skin Z-Stringer Hoops (Straps)
Design Coigura-n
tion Figs, 2' 3> Gage, Grain Gage, Spac- Gage, Width, Spacing,
and 4 n. Direc- in. ing, in. in. in. in.
tion
1 0. 025 T° 0. 016 2.5 0. 010 1. 0 20.0

2 0. 025 T 0. 016 2.5 0.,025 1..0 20.0
3 0. 025 T 0. 016 2.5 0. 050 1..0 20.0
4 0. 025 T 0. 016 2.5 0 .050 2 .0 20.0
5 . o ,025 L6 0. 016 2.5 0..025 1 .0 20.0
6 .. o 025 T 0. 016 2.5 0..010 0 .33 6.67
7 0..025 T 0.,016 2.5 0..050 0 .33 6.67
8 0..025 T 0..016 5.0 0 .025 1.0 20.0
9C 0,.025 T 0..016 2.5 0 .025 1.0 20.0
a
T = transverse (grain direction normal to hoop tension).
b
L = longitudinal (grain direction parallel to hoop tension).
c
Crack located at stringer.
FIG. 1—12-in. flat sheet specimens with straps.

The Copyright
grain direction
by ASTMwas parallel
Int'l (all to the short
rights reserved); dimension.
Mon Dec 7 14:40:45 Each panel
EST 2015
had a central saw cut orbya thin Elox cut transverse to the load direction.
Downloaded/printed
Two additional
University ofcrack propagation
Washington (Universitytests were conducted
of Washington) on large
pursuant to License flat No furth
Agreement.
panels of titanium 8-1-1 DA. The nominal dimensions of the panels
were 0.032 by 36 by 108 in. The panel was first tested unstiffened and
without guide plates. After failure, the panel was repaired (double lap-
joint spot-welded), and a new central saw cut was made some distance
from the transverse joint. To represent a stiffened structure, four trans-
verse Z-stringers were added in the area of the crack as shown in Fig. 2.
FIG. 2—36-in. flat sheet specimen configurations, Titanium 8-1-1 DA.
Unstiffened Curved Panels

The 70-in.-radius panel consisted of several sheets of 0.030 gage
titanium 8-1-1 DA material joined together with % 2 - m - diameter
brazier-head Monel rivets in butt joints with inner and outer splice
doublers. The center panel with the grain running lengthwise was the
test panel. The entire assembly was fitted with steel edge doublers and
end bulkheads. The required saw cut was lengthwise (with the grain) in
the test panel and was sealed internally with a rubber bladder.
The 33.3-in.-radius panel consisted of several sheets of 0.030 gage
titanium 8-1-1 DA material fusion butt-welded. Other details were
similarUniversity
to thoseofofWashington
the 70-in.-radius
(University panels as indicated
of Washington) in toFig.
pursuant 3. Agreement.
License No fu
Unstifffened Cylinders
The design of the unstiffened cylindrical test specimens is similar to
the stiffened cylinder shown schematically in Fig. 4 without the stiffeners.
The skin consists of four panels of equal size which are butt-welded in
FIG. 3—Fail-safe tests on curved panels.
FIG. 4—Stiffened cylinders with hoops.
the axial (longitudinal) direction. Five specimens were fabricated from

titanium 8-1-1 DA: t = 0.050 in., one specimen from titanium 13-11-3:
t = 0.050 in., and five specimens from stainless steel 14-8: t = 0.025
in. The unstiffened cylindrical specimens were constructed with the grain
direction of the sheets in the axial direction, transverse to the hoop ten-
sion stress as shown
Downloaded/printed by in Fig. 4. One stiffened cylinder was constructed
with circumferential grain
University of Washington direction
(University parallel
of Washington) pursuantto hoopAgreement.
to License tension.NoMore than authorized
further reproductions
FIG. 5—Titanium 8-1-1 DA fuselage fail-safe test panel.
one test was accomplished on many of the cylinders by repairing the

initial test failure and conducting additional tests on undamaged quad-
rants.
StiffenedCylindeersrs
The required size of crack-stopper hoops in stiffened cylinders was in-
vestigated at two different stress levels for each configuration. Nine de-
sign configurations were tested for titanium 8-1-1 DA and for 14-8 stain-
less steel. Duplicate specimens were tested to assess scatter of test results.
The cross-sectional area of the hoops and the spacing was varied to ob-
tain hoop failure as well as skin failure alone without fracturing the
hoops. Copyright
Tables by1 ASTM
and 2Int'l
list(allconfiguration dimensions. The design is
rights reserved); Mon Dec 7 14:40:45 EST 2015
shown Downloaded/printed
schematically inbyFig. 4. The total number of stiffened specimens
was 36. University of Washington (University of Washington) pursuant to License Agreement. No fu
Several of the configurations were chosen for spot-checking various
effects on crack propagation behavior, including circumferential grain

direction of the skin material, wider stringer spacing, and the location of
the crack in the skin immediately adjacent to the stringer. More than
one test was obtained from many of the cylinders by repairing the initial
failure and testing an undamaged quadrant. One stiffened titanium
cylinder of Configuration 9 was modified, after patching, by spot-welding
PIG. 6—Test jig assembly with cylinder specimen.
two external hoops, each 0.100 by 0.333 in. This resulted in hoop
spacing of 6.67 in. One of the unfailed panels was slot-cut and sealed
for testing.
SST Fuselage Panels

Two panels, 66 by 117 in. with a 75-in. skin radius, as shown in
Fig. 5 Copyright by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
,were tested. One panel was fabricated from titanium 8-1-1 DA and
the other
University of Washington 13-11-3.
from titanium (University ofFor both panels
Washington) the
pursuant skin gage
to License was No further re
Agreement.
0.032 in., the stringer gage 0.032 in., and the rings were 0.050 in.
Significant structural details simulated the current version of the super-

sonic transport at the time of testing. The stringer profile was simulated
accurately at the skin attachment. The ring profile was accurately
simulated in all details including cutouts for stringer passage through
the ring.
Test Equipment and Procedures
Flat Panel Tests

An initial transverse Elox cut or saw cut was made in all flat panels
prior to testing. The panels were cycled under axial tension loading to
generate a naturally propagating crack at the ends of the cut. The 9 by
27-in. and 12 by 36-in. panels were then loaded statically to failure in a
Baldwin Universal test machine. Measurement of slow crack growth
to failure was made directly with a scale taped to the surface of the
panel, and checked by observation of the fracture surface. For those
panels tested at extreme temperature (+550 F or —110 F), thermo-
couples were attached to monitor the skin temperature in the region of
the crack. For hot tests, the specimens were enclosed in an oven and
heated with hot air while clamped in the test machine. For the cold
tests, the specimens were enclosed in a cold box while in the test machine.
Cooling was effected by circulating liquid carbon dioxide and air inside
the box. All testing of the 36 by 108-in. panels was conducted in a
fatigue-testing machine. An initial transverse saw cut was made, and the
specimens were installed in the machine with clamp supports at the third
points. Axial load was cycled until failure. The crack length was recorded
as a function of the number of cycles applied.
Cylinder Tests
Room Temperature—The unstiffened and stiffened cylinders were
tested at room temperature in the setup shown in Fig. 6. The specimen
was mounted between steel bulkheads, one fixed and the other free to
move with the specimen. Thus, a biaxial stress condition was created in
the specimen with internal pressurization. The specimen was completely
filled with water for pressurizing. The initial saw cuts ranged from 2 to
8 in. in length, depending on the material and the stress level. The speci-
men was sealed against leakage in the region of anticipated crack growth.
The first cylinder test was instrumented with three electrical strain gages,
15 in. apart, to check the uniformity of the hoop tension stress within
the test section. Average readings were 91 and 97.7 per cent of theoreti-
cal values at 20 to 30-psi pressures, respectively. The tests were con-
ducted Downloaded/printed
by cycling the internal
by pressure until failure occurred. The crack
length University
was measured continuously until of
of Washington (University failure.
Washington) pursuant to License Agreement.
Elevated Temperature—For the elevated temperature tests the pres-
surizing medium was room temperature air; therefore, a sealed internal

steel tank was included to reduce the volume to be pressurized. High-
pressure air was introduced inside the test cylinder, between the skin
and internal tank, through a system of pressure-sensitive regulators and
solenoid-operated valves. A sheet steel shroud (oven) was supported over
the cylinder. High-temperature air was manifolded between the cylinder
surface and the shroud. Thermo-couples were located to monitor the
skin temperature and also to control the temperature and volume of the
incoming hot air. Cylinder skin temperature was held constant while the
internal pressure was cycled. The shroud included a removable door
over the saw cut crack area. Each cylinder was cyclically pressurized at
room temperature to obtain a naturally propagating crack before record-
ing data. Load cycling was begun at +550 F. At appropriate intervals,
pressurization was stopped and direct readings of crack length were made.
Sub-Zero Temperature—A commercial heat-exchanger fluid (Lexcol)
was the pressurizing medium. Carbon dioxide from a liquid carbon
dioxide system cooled the fluid to approximately —110 F. The cold
fluid was continually circulated between the cylinder and tank at low
pressure (approximately 5 psi). The test cylinder and plumbing were
wrapped in thermal insulation blankets. The target temperature of —110
F was achieved during the sub-zero testing of the smaller flat test panels
(9 by 27 in.) and the standard tensile coupons. However, for the large
cylinder test sections (approximately 30 in. in diameter by 60 in. long),
heat losses were such that the actual test temperatures averaged —90 F.
The same method of pressurization, load, and temperature monitoring
was applied as for the room temperature testing.
To determine the crack length without periodically removing the in-
sulating blanket, a set of 20 failure wires, at lA-m. spacing, was bonded
to the cylinder surface in the area of anticipated crack growth. A small
light bulb was wired in series with each failure wire. As each successive
light went out, the cycle number was noted. Measurement was made of
the actual distance of each wire from the center of the saw cut.
Curved Panel Tests

Room Temperature—The test panels were bolted to a flat steel table.
The 33.3-in.-radius specimen was filled with water within' 5 in. of the
top. Pressurizing air at room temperature was introduced inside the
specimen through a port in the table top. The pressure cycling technique
was the same as for the cylinder tests at elevated temperature described
previously.
Elevated Temperature—The curved test panels and the SST fuselage
panels Downloaded/printed
were bolted to a steel
by table as shown in Figs. 3 and 5 and covered
with a University
sheet metal shroud. The
of Washington shroud was
(University spaced approximately
of Washington) 3 in.Agreement
pursuant to License
outside the skin and included a. window for viewing and measuring
FIG. 7—Mechanical properties of Titanium 8-1-1 DA materials.
crack length. Pressurizing air at room temperature was introduced inside

the specimen, and high-temperature air was introduced between the
shroud and the specimen. The pressure cycling test procedure was the
same as for the cylinder tests at elevated temperature.
Test Results and Analyses

In the interest of brevity, the test results are presented graphically
References 2 and 3 contain the tabulated test data; those interested ir
original test data are referred to these sources. Analyses of these tesl
data are based on selected equations listed and discussed in Appendix I
The approach is to determine material parameters from the lowest order
of test and predict higher order performance. The complexity of the
fracture phenomenon and difficulties of nonlinear plasticity effects pre-
clude any but this semiempirical approach at this time. Indeed the carefu]
substantiation required
Downloaded/printed by at each step of aircraft design makes this essen-
tial empiricism
University of mandatory in anyof case.
Washington (University The pursuant
Washington) mechanical properties
to License from
Agreement. No further reproductio
TABLE 3—Properties of sheet materials.

Temperature
_90 F +70 F +550 F
Material Property
Cylinders and Curved Panels
Cylinders Strap Stiffened and Unstiffened Cylinders
Flat Panels Flat Panels
SAl-lMo-lV Ftu , ksi 167 144 148.5 108

Titanium sheet E, psi 19 X 106 17.5 X 106 17.9 X 106 14.8 X 106
Duplex annealed e,% 13.4 13.4 13.6 13.4
V7 (in.1/2) 0.170 0.275 0.310 0.725
cylinder flat panels

with straps
pH 14-8 Mo Ftu , ksi 207.7 215

SRH 1050 E, psi 29 X 106 29 X 106
Steel sheet e,% 5.3 4
V7 (in.1/2) 0.286 0.425
FIG. 8—Residual
University static
of Washington strengthofofWashington)
(University flat unstiffened titanium
pursuant panels.
to License Agreement. No further repro
FIG. 9—Normalized residual static strength of flat unstiffened titanium panels.
FIG. 10—Critical crack length versus initial crack length of flat unstiffened ti-
tanium panels.
standard tension
Copyright coupon
by ASTM Int'ltests of the
(all rights materials
reserved); Mon utilized in thisEST
Dec 7 14:40:45 program
2015
are plotted in Fig. 7 by
Downloaded/printed and tabulated in Table 3. The Neuber constant
-\Xp' isUniversity
determined from static
of Washington residual
(University tear tests. pursuant
of Washington) These data, presented
to License Agreement. No furth
in Table 3, are the basis for the analysis to follow.
CR1CHLOW AND WELLS ON TITANIUM AND STEEL CYLINDERS 39
FIG. 11—Computed Neuber constant \/p' values for titanium materials.
University of Washington (University of Washington) pursuant to License Agre
FIG. 12—Residual to ultimate strength ratio versus degree of cracking, 36-in.
panel, Titanium 8-1-1 DA.
FATIGUE CRACK PROPAGATIOGASTIOPNN
Flat Panels
Flat panel data are analyzed for two types of information: (1) rate of
crack growth and (2) residual static strength at a specific crack size.
Unstiffened Titanium Flat Panels—The static residual strength data
for the flat unstiffened panels are plotted versus the initial crack length
(Fig. 8); the normalized data are shown in Fig. 9. The relation between
the initial and the final crack length at fast propagation to total failure
is graphically presented for the panels in Fig. 10.
Each static residual strength test was used to compute the characteris-
tic Neuber constant vV (m- 1/2)- The results are plotted in Fig. 11 to-
FIG. 13—Crack growth in 36-in. panels, Titanium 8-1-1 DA.
gether with corresponding NASA data for titanium 8-1-1 triplex-an-

nealed, t = 0.040 in. from Ref 5.
The gross area stress of a cracked flat panel of width W with a crack
length X0 (measured before loading) is given by the stress concentration
equation of Kuhn (Appendix I, Eq 14)
Based Downloaded/printed by
on the twelve panel tests and material coupon strength data,
University of Washington (University of Washington) pursuant to License Agreement
average values required for the prediction of residual strength curves for
the titanium material are defined from Figs. 7 and 11.
Wide Flat Panels of Titanium—Two crack propagation tests were con-

ducted on wide, flat titanium panels for two purposes:
1. to compare the rate of crack growth in wide flat-sheet specimens
under cyclic uniaxial tension with that in unstiffened cylinders under
cyclic internal pressure, and
2. to determine the buckling correction factor for the residual
strength of an unsupported flat-sheet specimen with central crack
(Appendix I).
The results are presented graphically in Figs. 12 to 14. The completely
unstiffened and unguided panel failed during cycling when the crack
FIG. 14—Rate of crack growth in 36-in. panels, Titanium 8-1-1 DA.
reached a length of 10.5 in. The stiffened panel was failed statically after
the crack had grown to a length of 12 in. (1A of the width). The residual
strength was 38.4 ksi. In both cases, buckling occurred along the edges
of the crack. This buckling phenomenon has a pronounced effect on the
rate of crack growth (Fig. 14).
The buckling correction factor was determined by Eq 2 (Appendix I,
Eq 24)
The first panel test gives
Downloaded/printed by C0 = 0.00165, which is of the order of magnitude
0.001.
= University For
of Washington (Universitythe transversely
of Washington) pursuant to License Agreement. No further re
FIG. 15—Residual
University of Washington (University of toWashington)
ultimate strength
pursuantratio, flat panels
to License with straps,
Agreement. Titanium
No further 8-1-1 DA.authorized.
reproductions
Downloaded/printed by FIG. 16—Residual to ultimate strength ratio, flat panels with straps, Steel 14
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized.
ATIGUE CRACK PROPAGATION
stiffened panel, c0 = 0.0037. These values of the coefficient can only be

considered tentative; more tests are required to establish reliability.
Flat Panels With Straps — Forty-eight fatigue-cracked flat-sheet speci-
mens 12 by 36 in. with two longitudinal straps were tested to determine
the fracture toughness and deformation characteristics of attachments.
For each candidate material, twelve design configurations were chosen.
Three designs had mechanical fasteners (Monel rivets), six were spot-
welded, and three configurations had straps fusion-welded to the skin,
as shown in Fig. 1.
Table 3 gives the mechanical properties of the skin and straps. The
residual strength values are graphically presented in Figs. 15 and 16 in a
nondimensional form, Fg/Ftu versus Asiia.p/bt. Each test has been dupli-
cated to obtain some indication of the scatter of test results.
These figures also show curves computed by Eq 3 for the residual
strength of cracked flat panels with straps. This equation was derived in
Appendix I, Eq 16
The values of F0 have been predicted by use of the stress concentration

formula, with the material property values from Table 3.
Considering the simplicity of the derivation of Eq 3, and the scatter
inherent in this type of test, the agreement between test results and pre-
diction is satisfactory. The test points are well scattered around the pre-
diction curves. No definite superiority of any one attachment system is
recognizable, except for a slight superiority of mechanical fasteners and
spot welds compared with fusion welds for the titanium alloy. The stain-
less steel 14-8 does not show appreciable differences between these
types of attachment. However, it appears that closer spot-weld spacings
are of somewhat higher efficiency.
Unstiffened Curved Panels and Cylinders

The purpose of the curved panel crack propagation tests was to obtain
additional experimental substantiation of the curvature correction for-
mula:
From the five crack propagation tests on unstiffened cylinders and the
unstiffened curved panels, correlating data for three radii (R = 15 in.,
33 in.,Copyright
and 70 byin.)ASTM
and two
Int'l thicknesses (t = 0.050
(all rights reserved); in. and
Mon Dec 0.030 in.)
7 14:40:45 ESTare
2015
available. The results by
Downloaded/printed are summarized in Fig. 17. The allowable stress,
Fg, fiat for an infinitely wide cracked sheet was computed by using the
average values for titanium 8-1-1 DA sheet material from Table 3.
Peters and Dow [6] suggested a curvature formula to correct flat panel
data to cylinder data of the general form (Appendix I, Eq 20):
FIG. 17—Comparison between test results and curvature correction, Eq 5.
where the factor m0 has to be determined from tests for each material.
Figure 17 shows a graphical representation of the titanium and steel
test results on unstiffened cylinders along with curved panels with differ-
ent radii and thicknesses. The unstiffened titanium 8-1-1 DA cylinder
test data at room temperature were best fit for an average value of
m0 = 4.25; the steel cylinder data (from Ref 7) fit an average value of
m0 = 3.0. The titanium curved panels require a value of m0 = 12. These
data indicate an effect of radius and sheet thickness which is not ade-
quately accounted
Copyright by ASTMforInt'l
in (all
therights
empirical EqMon
reserved); 5. Dec
This7 formula also2015
14:40:45 EST includes
buckling and pressure bending effects along the edges of the crack,
University of Washington (University of Washington) pursuant to License Agreement. No further r
especially for the large crack lengths under consideration

1.5). Buckling aspects and bending from pressure loading on the lip of
the crack are governed to a degree by the ra
A new formula for curvature correction discussed in Appendix I, Eq 23,
introduces the needed X/t ratio:
Figure 18 shows considerably improved correlation of titanium data
FIG. 18—Improved empirical curvature correction, Eq 6.
which best fits the equation with k = 0.48, n = 0.75. Additional tests
with variations of R and t are required to fully correlate the steel data.
Figure 19 gives the residual static hoop tension strength, Fh for the
cracked titanium cylinders as a function of the critical crack length,
Xcr. The curves of Eq 6 indicate a better fit with the experimental data
than was attainable with Eq 5. The empirical curves include local buck-
ling and bending effects and are also valid for large crack lengths
(X/VRt up to 25).
To calculate the residual static strength curves for the steel cylinders,
the Neuber constant for steel was estimated by using the residual strength
tests of wide specimens reported in Ref 7.
The Copyright
residual bystrength
ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
of cracked unstiffened cylinders of various ma-
terials University
is comparedof Washingtonstrength-weight
on a basis inpursuant
(University of Washington) Fig. 20. The data
to License for No furth
Agreement.
F I G.
cylinders and curved panels.
FIG. 20—Residual hoop strength-to-weight ratio for unstiffened cylinders of

aluminum, steel, and titanium.
aluminum alloy 2024-T3 cylinders have been taken from Ref 8. Titanium
8-1-1 Copyright
DA shows a by
marked
ASTM superiority.
Int'l (allThe residual
rights strengthMon
reserved); variations
Dec 7 14:4
with temperature for titanium cylinders
Downloaded/printed by is shown in Fig. 21.
Ra University of Washington (University of Washington) pursuant
GATION fatigue crack propagation
FIG. 21—The effect of temperature on the residual static strength of Titanium

8-1-1 DA unstiffened cylinders.
TABLE 4—Functions for rate of crack growth equation.

Function Titanium 8-1-1 DA Steel 14-8
perimental data with the predicted curves for the rate of crack growth
of the titanium and steel unstiffened cylinders under cyclic pressure are
made using Eq 7 (derived in Appendix I.)
Rate of crack growth =
The required material coefficients determined from flat panel data are
given in Table 4.
The correlation of prediction with experiment for the rate of crack
growth of titanium 8-1-1 DA cylinders at cyclic stress levels of 15 to 30
ksi is Downloaded/printed
plotted byin Fig. 22. Correlation of predicted and experimental rate
of crack
Universitygrowth for ofthe
of Washington (University steel
Washington) pursuant 14-8 cylinders
to License Agreement. at cyclic
No further reproductions authorized. stress levels of 20 to
FIG. 22—Rate of crack growth in unstiffened cylinders, Titanium 8-1-1 DA.
FIG. 23—Rate of crack growth in unstiffened cylinders, Steel 14-8.
40 ksi is plotted in Fig. 23. Agreement between test results and predicted
values is seen to be surprisingly good, considering the wide scatter in-
herent in this type of physical phenomenon.
Rate of Crack Growth—Unstiffened Curved Panels—Figure 24 shows
the computed crack growth rates for the four curved panels together
with Downloaded/printed
predicted curves by which were computed by using Eqs 6 and 7 with
the following
University ofvalues for(University
Washington titaniumof Washington)
materials:pursuant to License Agreement. No further reproduction
Correlation with the experimental data is seen to be good considering

the inherent scatter of this type of data. Comparing the correlation of
crack growth rates of cylinders predicted with the simple linear curvature
correction (Eq 5) and the rate correlation of the curved panels predicted
with the improved curvature correction (Eq 6) may indicate a standoff
within the scatter present. However, interpolating to other radii and
thicknesses will be more confidently achieved by use of the improved
formula.
FIG. 24—Rate of crack growth for curved panels under cyclic internal pressure.
Stiffene
Tita
properties of skins and hoops for the nine design configurations described
in Table 1 were determined from standard tensile coupons. The results
of the crack propagation tests are summarized graphically in Figs. 25
through 29. Using a saw cut as a starter, the crack in the skin progressed
completely straight between the Z-stringers. Figures 25 through 27 show
the crack growth histories for stiffened cylinders with 20-in. hoop spac-
ing and various hoop areas. These figures show clearly the slower rate of
crack growth with increasing hoop area. Failure of hoops as well as non-
failureCopyright
of hoopsby occurred.
ASTM Int'l For the heaviest
(all rights reserved); configuration with Ahoop
Mon Dec 7 14:40:45 =
EST 2015
in.2, tested at fh by= 15 ksi, the running crack was arrested at the
0.200Downloaded/printed
hoopsUniversity
and became stationary
of Washington for several
(University hundred pursuant
of Washington) pressureto cycles
License until
Agreement. No
FIG. 25—Crack growth in stiffened Titanium 8-1-1 DA cylinders, temperature = 70 F.
FIG. 26—Crack growth in stiffened titanium cylinders at sub-zero temperature

- -90 F.
FIG. 27—Crack growth in stiffened titanium cylinders at elevated temperature

= 550 F.
failure of the skin along the spot welds of the hoops. The effects of sub-
zero and elevated temperature are indicated in Figs. 26 and 27.
Figures 28 and 29 show data for design configurations with corre-
sponding values
Copyright of the
by ASTM Int'lratio (hoop
(all rights area)/(skin
reserved); Mon Dec 7 area)
14:40:45but
ESTwith
2015 a smaller
hoop Downloaded/printed
spacing of 6.67by in. For the same area ratio the total number of
University of Washington (University of Washington) pursuant to License Agreement. No further reprodu
FIG. 28—Crack growth in stiffened titanium cylinders, hoop spacing — 6.67

in., temperature = 70 F.
pressure cycles required to fail the stiffened cylinder increases consider-

ably with decreasing hoop spacing. Sub-zero and elevated temperature
effectsCopyright
on the by data forInt'l
ASTM 6.67-in.
(all rightshoop spacing
reserved); are7shown
Mon Dec 14:40:45in Fig.
EST 201529.
TheDownloaded/printed
critical crack by length in stiffened titanium cylinders under internal
pressure loading
University is graphically
of Washington (Universitypresented in pursuant
of Washington) Fig. 30 for the
to License nine con-
Agreement. No further reprod
FIG. 29—Crack growth in stiffened titanium cylinders, hoop spacing — 6.67

in., sub-zero and elevated temperature.
figurations. The influence of sub-zero and elevated temperatures is
shown in Fig. 31. For the same area ratio but varying hoop spacing, the
residual (static)
Copyright strengths
by ASTM Int'l are
(all approximately the same.
rights reserved); Mon Dec 7For the same
14:40:45 hoop
EST 2015
area Downloaded/printed
but varying spacing,
by the residual (static) strength increases with
smaller hoop ofspacing.
University Washington (University of Washington) pursuant to License Agreement. No furth
Steel 14-8—Stiffened Cylinders—The crack-growth histories for the

stiffened steel 14-8 cylinders are summarized graphically in Figs. 32 and
33. As the cyclic pressure tests of the first two configurations (1 and 2)
were conducted, it became apparent that final failure always occurred
along the spot welds of the hoops and not through the hoops even for
these small hoop areas. It was recognized that this mode of failure was
due to the close spacing of the spot welds (0.30 and 0.50 in.) which
created an almost continuous zone of heat-affected material in the skin.
Heat-treating the spot-welded steel assemblies was not possible; there-
fore, it was decided to increase the spot-weld spacings for the remaining
configurations (5 through 9). The spacings chosen were 0.83 in. for
FIG. 30—Residual strength of stiffened cylinders, Titanium 8-1-1 DA, tempera-

ture = 70 F.
Configurations 5, 6, 7, and 9 (stringer spacing b = 2.5 in.), and 1.00 in.

for Configuration 8 (stringer spacing b = 5.0 in.). The final failure mode
of these specimens with increased spot-weld spacings was straight through
the hoops.
The data have been reduced so that each set of crack-growth curves
has the same common initial crack length. For this steel material the
test scatter is considerable; some of the heavier configurations have a
faster rate of crack growth than the lighter ones. The range of scatter is
also reflected in Fig. 34 which shows the critical crack length versus hoop
tension-stress level for the nine steel cylinder configurations.
TheCopyright
crack-propagation tests conducted on pressurized stiffened steel
by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
cylinders have shown that
Downloaded/printed by for the chosen type of hoop attachments (spot-
welding), titanium
University 8-1-1 DA
of Washington material
(University is superiorpursuant
of Washington) to the tosteel 14-8.
License For No furthe
Agreement.
this and other reasons, sub-zero and elevated temperature data were not
obtained on the steel cylinder specimens.
Analysis—Stiffened Cylinders—Two important design criteria can be
established in a simple manner. The crack-length criterion states that
for a given hoop tension-stress
in a stiffened cylinder is at least equal to that of an unstiffened cylinder of
the same radius (Figs. 30, 31, and 34). In a longitudinally stiffened
FIG. 31—Residual strength of stiffened Titanium 8-1-1 DA cylinders at sub-

zero and elevated temperature.
cylinder with hoops (straps), the actual skin stress in the direction of hoop
tension is always lower than the nominal hoop tension stress, and drops
noticeably at the hoops; hence, the resultant larger critical crack length.
In addition, longitudinal stiffeners reduce the buckling and bending
along the edges of the longitudinal crack, thus further increasing the
critical crack length.
The hoop-size criterion presents a prediction as to whether the rup-
ture ofCopyright
the skin will be
by ASTM Int'lexplosive or confined.
(all rights reserved); Mon Dec In the first
7 14:40:45 ESTcase,
2015 the run-
ning crack in the skinbycannot be stopped at the hoops and causes hoop
Downloaded/printed
failure.University
In theof second case,
Washington under
(University cyclic loading
of Washington) pursuantthe propagating
to License Agreement. skin
No further reprodu
FIG. 32—Crack growth in stiffened Steel 14-8 cylinders, temperature = 70 F.
crack either will be arrested completely at the hoops or the crack growth
will be retarded (gradual failure).
From equilibrium conditions at failure, Eq 9 and design criterion Eq
10 have been derived in Appendix I (Eqs 45 and 46):
Applied stress f explosive failure

marginal
Copyright by ASTM Int'l (allf rights
hretarded
reserved); Mon Dec 7 14:40:45 EST 2015
Figure 35 shows
University the graphical
of Washington (Universityrepresentation of Eqs
of Washington) pursuant 9 andAgreement.
to License 10 for the
No further repro
58 58
FIG. 33—Crack growth in stiffened Steel 14-8 cylinders, hoop spacing — 6.67
in., temperature — 70 F.
stiffened titanium cylinder tests. The values for F0 have been taken from
Fig. 21. By comparing the crack growth histories of the plotted design
configurations
Copyright by(Figs.
ASTM25Int'l through 29) withMon
(all rights reserved); theDecpredicted
7 14:40:45behavior
EST 2015 shown
in Figs. 35 and 36, it can be seen that Eq 9 and the design criterion Eq
10 areUniversity of Washington (University of Washington) pursuant to License Agreement. No further rep
well substantiated for the titanium cvlinders.
FIG. 34—Residual strength of pressurized stiffened cylinders, R = 15 in., Steel

14-8.
FIG. 35—Criteria for stable and unstable crack growth in stiffened cylinders,
Titanium 8-1-1 DA, temperature = 70 F.
SST Fuselage Panels

The crack-resistance tests described in previous sections were con-
ductedCopyright
with specimens in which
by ASTM Int'l one
(all rights or more
reserved); Mon dimensions wereESTscaled
Dec 7 14:40:45 2015
down Downloaded/printed
from the full-sizedby fuselage. Two fuselage panels were constructed
in which all important
University dimensions
of Washington (Universityand details were
of Washington) maintained
pursuant in Agreement.
to License scale, No fu
including radius, ring detail, and the like. One panel was constructed of
titanium 8-1-1 DA material, while the other panel was dimensionally
identical but constructed of titanium 13-11-3 material. A sketch of the
panel is shown in Fig. 5.
The test temperature ranged from 550 to 650 F. Initial damage was
produced by a saw cut approximately 4 in. long, centrally located in
FIG. 36—Criteria for stable and unstable crack growth in stiffened cylinders,
Titanium 8-1-1 DA, sub-zero and elevated temperature.
the skin between two rings and two stringers. A sufficient number of
pressure cycles were first applied at room temperature to produce true
fatigue cracks at the ends of the saw cut before testing at elevated
temperature. This procedure was repeated once (at 8 in.) in the titanium
13-11-3 panel and twice (at 8 and 10 in.) in the titanium 8-1-1 DA panel
when the growing crack was extended by saw cuts in order to accelerate
the test.
Titanium 13-11-3 Fuselage
Downloaded/printed by Panel—The pressure was corrected to the
actualUniversity
skin gage (0.032) and
of Washington density
(University of the titanium
of Washington) pursuant13-11-3
to LicensetoAgreement.
provide No further r
data comparable on an equal weight basis. The resulting pressure was

14.3 psi.
The crack history is plotted in Fig. 37 for comparison with the titanium
8-1-1 DA data. The last measured crack length was 22.7 in. after 1444
cycles. At an estimated length of 23.7 in. after 1458 cycles (from an
8-in. saw cut), the panel was failed statically at a pressure of 14.25 psi
FIG. 37—Crack growth at elevated temperature in titanium fuselage panels.
and 550 F. Figure 38 is a photograph of the rupture which can be seen to

follow a substantially straight path to the ends of the panel.
Titanium 8-1-1 DA Fuselage Panel—After the crack in the skin had
progressed substantially beyond the two adjacent rings, the redistribution
of the load caused these two rings to fail, and the crack growth rate in-
creased. However, after the crack had extended over three bays (crack
lengthCopyright
= 33.4byin.,ASTMFig.Int'l
37),(allthe crack
rights growth
reserved); MonwasDec arrested byEST
7 14:40:45 the 2015
next
pair ofDownloaded/printed
rings. The panel by was then failed statically at room temperature.
TheUniversity
test showed that, even
of Washington with ofthisWashington)
(University extent ofpursuant
damage, the titanium
to License Agreement. No furth
FIG. 38—Failure of Titanium 13-11-3 fuselage panel.
FIG.
University 39—Cycles
of Washington to fail
(University Titanium pursuant
of Washington) 8-1-1 DA fuselage
to License panel.
Agreement. No further reproductions authoriz
8-1-1 DA design configuration would withstand an internal pressure of

12.25 psi. The entire crack growth history of this test is shown in Fig.
37. From these data the curve of Fig. 39 was derived. This curve gives
an estimate of the number of pressure cycles required to fail the titanium
8-1-1 DA fuselage structure from any initial crack length. For example,
a 1-in. initial crack would require approximately 46,000 cycles, or
flights, to reach unstable proportions; or a 10-in. initial crack would re-
quire approximately 2100 cycles, or flights, to reach unstable proportions.
The 46,000 flights for a 1-in. crack are of the order of the required life-
time of the SST vehicle, while the 2100 cycles to failure for a 10-in.
crack would approximate a year's flying. This test provides substantia-
tion of the excellent fail-safe characteristics of titanium 8-1-1 DA mate-
rial for the SST environment.
Summary and Conclusions
A body of experimental data has been presented and analyzed to form
the basis for the damage-tolerant design of the U.S. supersonic transport
fuselage. Stiffened and unstiffened flat panels, curved panels, and cylin-
ders were tested for crack growth rate and residual static strength under
the temperature environments of the SST. Materials investigated were
titanium 8-1-1 DA, titanium 13-11-3, and stainless steel 14-8. Analyses
of the data provided design formulas and failure criteria, which gave
good correlation with the experimental results. The titanium 8-1-1 DA
material was demonstrated to be the best choice for the application. From
prior work, a general formula was derived for predicting the rate of crack
growth under cyclic load, based on the mechanics of stable crack growth
to the critical size. When provided with material data from a few simple
static experiments, good correlation with measured rates of crack growth
of cylinders under cyclic pressure was shown for both the titanium and
steel materials tested.
Tests of two fully detailed SST fuselage panels demonstrated success-
ful application of the design procedures to the full-scale structure. How-
ever, considerable additional testing and engineering development is nec-
essary and is planned for the design of the requisite safety and damage
tolerance into the U.S. supersonic transport.
APPENDIX I
Semi-empirical Equations for Critical Crack Length and Rate of Crack Growth
The data obtained from the research program were reduced by means of
several equations. These are derived or discussed in the following paragraphs.
Critical Crack Length

Much of the present-day knowledge in the field of fracture mechanics is
based upon work of Griffith, who more than 40 years ago first proposed an en-
ergy concept of fracture to explain certain discrepancies in the then-accepted
hypothesis of rupture. For a flat, cracked sheet under uniaxial tension, Griffith's
formula for the critical value of the gross-area tensile stress Fg at which the crack
of length X0 begins to propagate can be written in the form:
Irwin and his associates modified Griffith's formula to its present form:
which, for small crack lengths, becomes:
The fatigue notch strength concept has been developed by Paul Kuhn and others
[4,9,10]. It assumes that a crack is an extreme form of a notch. For a flat sheet
specimen with a centrally located crack under uniaxial tension, the residual
gross-area strength is given by the following expression:
where the Neuber constant
Flat Panel with Straps

If Fo is defined as the failure gross-area stress of the flat-sheet specimen
without straps, with a crack of the length X0 = b (strap spacing), then, the total
load P at failure is given by
which,Downloaded/printed by
with W = 3b, gives:
For simplification, it has been assumed that failure of the skin (fast propagation
of the crack) and failure of the straps occur simultaneously and that both com-
ponents have the same ultimate material tensile strength.
Curvature Correction
The curvature correction for cylinders and curved panels with a longitudinal
crack (perpendicular to the hoop tension load) can be introduced as follows [6]:
which gives for short cracks, say X0 < 5t:
and for large cracks (aircraft pressure cabin):
The numerical coefficient of 5 in Eq 19 has been confirmed by tests of aluminum

cylinders [6\. However, tests of steel and titanium cylinders in this program sug-
gest variations in this factor for best fit with experimental data. A more general
form of the equation can be:
where m0 is to be determined from test data for each material.

The formula (Eq 20), however, does not account for dimensional parameters
X/t significant to buckling, and to bending from pressure loading on the lip of
the crack.
Theoretical investigations by Folias [11] have shown that the stress distribu-
tion around an axial crack in a pressurized cylinder leads to a curvature correc-
tion formula of the type
where the expression in parentheses is a positive quantity. Another theoretical

formula by Lurie by
Copyright for smallInt'l
[12] ASTM circular holes inreserved);
(all rights cylinders gives
Monthe curvature
Dec 7 14:40:45 ES
correction as
University of Washington (University of Washington) pursuant to Lice
66 66
where k = 0.527 for hoop stress alone (axial load Nx = 0) and k = 0.575 for
the usual "boiler" case (N
cracks, X2/(4Rt) « 1.0.
Based on these derivations, therefore, it is proposed to assume the following
curvature correction for titanium alloy sheet material:
Buckling Correction
A flat-sheet specimen with a centrally located crack will buckle along the
crack when the sheet is subjected to tension load; the buckling deformation
reduces the strength of the sheet. The residual strength of the cracked sheet as
predicted by Eq 14 is based on the assumption that buckling is completely re-
strained, for instance, by guide plates. In actual cracked thin-walled structures,
some buckling will occur, even when stiffened. To allow for this effect, the
residual strength as given by Eq 14 should be reduced by the empirical factor
where the constant c0 depends on the material. Equation 24 is based on a very

small number of tests and consequently of questionable reliability. For aluminum
alloys, the order of magnitude of c0 has been found to be approximately 0.001.
The buckling correction should not be applied when the curvature corrections,
Eqs 17 through 19 are applied.
Rate of Crack Growth

From prior work at Lockheed [13], a simple, but rather general and useful,
equation for the rate of crack growth can be derived as follows: £,17 are Cartesian
coordinates as shown in Fig. 40, and the shape of the curve for the slow stable
crack enlargement under increasing tension load is assumed in the simple form:
The total crack enlargement under increasing tension load can be assumed to
be a function of the initial crack length; hence,
In most materials, an initial crack does not enlarge under increasing load until
a certain stress level has been exceeded (ordinate c in Fig. 40). The critical stress
Per is also a function of the initial crack length (for example, Eq 14; hence,
Copyrig
Downlo
Univers
The functions $i(X0), $2(^0), the ratio SF, and the exponents m, n have to be
determined from tests.
Then, if /max is defined as the maximum cyclic tension stress, and assuming
/min = 0, the crack growth for one cycle (rate of crack growth, in. /cycle) is
given by
for
FIG. 40—Notation and coordinate systems for slow crack growth.
Under these assumptions, Eq 25 becomes:
which gives:
Instead of assuming a completely straight portion in the shape of the curve for
slow crack enlargement under increasing tension load, we can also assume ap-
proximately ^ = 0 and a much higher exponent, mi :» m. Under this assump-
tion, Eq 32 becomes
Well-known existing semiempirical formulas are easily derived from Eq 34 by

assuming the appropriate functions 3>i(Xo), ^z(X0}, and the exponents mi, n.
For instance, assuming
and
n - 4, (37)
Equation 34 becomes
which can be brought into the form
This is Valluri's equation for the special case of stress ratio, R/ = /mh,//max = 0,
and endurance stress, <r\ = 0 [14].
Using another set of assumptions: For Eq 11, define K3 = constant
Hence,
With n = 3, Eq 34 becomes:
This is Head's equ.
Hoop Design Criterion

A simple prediction formula was derived from equilibrium conditions as
follows: Assuming that the crack in the skin extends between two adjacent intact
hoops,Copyright
X = a,byanASTM equilibrium
Int'l (all equation can be
rights reserved); Monformulated at theEST
Dec 7 14:40:45 instant
2015 of hoop
failure:Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further reproduct
where:
F0 = residual hoop tension strength of the unstiffened cylinder with the same
radius and the assumed crack length, and
L* = effective length beyond which the initial stress distributions of the un-
cracked specimen prevail. L* has been tentatively assumed
L* = 3X = 3a.
Then, it follows from Eq 44
For a stiffened cylinder (ratio Ah/at and F0 given) under cyclic pressure loading
with an applied hoop tension stress level, fh = pR/t, the design criterion can be
formulated as follows:
f h > Fa explosive failure
fh = Fh marginal (46)
f h < Fh gradual failure
A similar empirical criterion has been established by NACA [6,16].
References
[1] W. J. Crichlow, 'The Ultimate Strength of Damaged Structure—Analysis
Methods with Correlating Test Data," Proceedings, ICAF-AGARD Sym-
posium on Full-Scale Fatigue Testing of Aircraft Structures, Amsterdam,
Netherlands, June, 1959 (published for AGARD by Pergamon Press, Ox-
ford, London, New York, 1960).
[2] FDL-TDR-64-80, "Fuselage Fail-Safe Design Data and Bomb-Resistant
Analysis for the Supersonic Transport," Supersonic Transport Research
Program, Air Force Flight Dynamics Laboratory, Wright-Patterson Air
Force Base, Ohio, June, 1964.
[3] FDL-TDR-65-165, "Fuselage Fail-Safe Design Data for the Supersonic
Transport," Supersonic Transport Research Program, Air Force Flight
Dynamics Laboratory, Wright-Patterson Air Force Base, Ohio, October, 1965.
[4] P. Kuhn, "Notch Effects on Fatigue and Static Strength," NASA Langley
Research Center, Langley Station, Hampton, Va., presented at the Symposium
on Aeronautical Fatigue sponsored by the International Committee on
Aeronautical Fatigue (ICAF) and the Structures and Materials Panel of the
Advisory Group for Aeronautical Research and Development (AGARD),
Rome, Italy, April, 1963.
[5] I. E. Figge, "Residual Static Strength of Several Titanium and Stainless
Steel Alloys and One Super Alloy at -109°F, 70°F, 550°F," NASA TN
D-2045, Nat. Aeronautics and Space Administration, December, 1963.
[6] R. W. Peters and N. F. Dow, "Failure Characteristics of Pressurized Stif-
fened Cylinders," NACA TN 3851, Nat. Advisory Commission for Aero-
nautics, Washington, D.C., December, 1956.
[7] "Fracture Toughness and Tear Tests," ML-TDR-64-238, Supersonic Trans-
port Research Program sponsored by the Federal Aviation Agency, Con-
tract No. AF 33(657)-! 1461, Air Force Materials Laboratory, Research and
Technology Division, Air Force Systems Command, Wright-Patterson Air
Force Base, by
Copyright Ohio, October,
ASTM 1964.
Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
[8] R.Downloaded/printed
W. Peters and P. Kuhn,
by "Bursting Strength of Unstiffened Pressure Cylin-
ders with Slits," NACA TN 3993, Nat. Advisory Commission for Aeronautics,
1957.
[9] P. Kuhn and I. E. Figge, "Unified Notch-Strength Analysis for Wrought
Aluminum Alloys," NASA TN D-1259, Nat. Aeronautics and Space Ad-
ministration, 1962.
[10] P. Kuhn and H. F. Hardrath, "An Engineering Method of Estimating
Notch-Size Effects in Fatigue Tests on Steel," NACA TN 2805, Nat. Advisory
Commission for Aeronautics, 1952.
[11] S. E. Folias, "The Stresses in a Cylindrical Shell Containing an Axial Crack,"
Aerospace Research Laboratories ARL 64-174, 1964.
[12] P. Kuhn, "Note on the Analysis of Longitudinal Cracks in Pressure Vessels,"
private communication.
[13] F. M. Mueller, Crack Growth Under Uniaxial Random Loads, LR 15027,
Lockheed-California Company, Burbank, Calif., January, 1961.
[14] R. S. Valluri, "A Unified Engineering Theory of High-Stress Level Fatigue,"
ARL Technical Note GALCIT SM 61-149-1843, October, 1961.
[15] A. K. Head, "The Growth of Fatigue Cracks," Philosophical Magazine,
Series 7, Vol 44, No. 356, September, 1953, pp. 925-929.
[76] P. Kuhn and R. W. Peters, "Some Aspects of Fail-Safe Design of Pressur-
ized Fuselage," NACA TN 4011, Nat. Advisory Commission for Aeronautics,
Washington, June, 1957.
University of Washington (University of Washington) pursuant to License Agreement. N
I. E. Figge1 and J.C. Newman, Jr.1
Fatigue Crack Propagation in Structures

with Simulated Rivet Forces
REFERENCE: I. E. Figge and J. C. Newman, Jr., "Fatigue Crack

Propagation in Structures with Simulated Rivet Forces," Fatigue Crack
Propagation, ASTM STP 415, Am. Soc. Testing Mats., 1967, p. 71.
ABSTRACT: Analytical and experimental studies were conducted on
rates of fatigue crack propagation in sheet specimens containing either
symmetric or nonsymmetric cracks subjected to uniform external load-
ings, simple concentrated loadings, or combinations of both. The .con-
centrated loads simulated the rivet forces hi an actual structure and
were applied by a special hydraulic fixture which is described. An experi-
mentally determined relation between the stress state near the tip of the
crack and the fatigue crack propagation rate for a given material was em-
ployed to predict the crack growth for specimens subjected to simple
loading conditions with good agreement. Superposition of stress-inten-
sity factors for simple loading was used sucessfully to predict crack
growth for specimens subjected to complex loadings.
KEY WORDS: fatigue (materials), crack propagation, riveted joints,
bolted joints, stress gradients, sheet (metal), aluminum alloy
Nomenclature
The units used for the physical quantities defined in this paper are
given both in U.S. Customary Units and in the International System of
Units (SI) [2]. Appendix I presents factors relating these two systems of
units.
a Half-crack length, in. (cm)
ae Equivalent half-crack length, in. (cm)
fli Initial half-crack length, in. (cm)
b Half-width of specimen, in. (cm)
D Pin diameter, in. (cm)
e Distance from the center line of a crack to the vertical
center line of the specimen, in. (cm)
Fi( ) , ( ? ( ) A function of ( )
1
Aero-space technologist, Fatigue Branch, Structures Research Div., National
Aeronautics and Space Administration, Langley Research Center, Langley Station,
Hampton, Va.
71
h Half-height of specimen, in. (cm)

k Stress-intensity factor, Ibf-in-3/2 (MN-m"3/2)
&max Stress-intensity factor corresponding to maximum load,
Ibf-in-3/2 (MN-m-*/2)
L Length of crack from one side of a hole, in. (cm)
Li Initial length of crack from one side of a hole, in. (cm)
N Number of load cycles
P Concentrated force, Ibf (N)
Pm&x Maximum concentrated force during cyclic loading, Ibf (N)
Pmin Minimum concentrated force during cyclic loading, Ibf (N)
R Load ratio, minimum load to maximum load
S Perpendicular distance from plane of crack to point of
application of concentrated force (midpoint of bearing
interface), in. (cm)
t Plate thickness, in. (cm)
Ja/fi?n Fatigue crack growth rate, in./cycle (m/cycle)
<r Uniform extensional stress, ksi (MN/m2)
<rmax Maximum stress during cyclic loading, ksi (MN/m2)
<r(x) Normal stress applied to crack surface, ksi (MN/m2)
Aa Incremental crack length, in. (cm)
AN Incremental number of cycles
v Poisson's ratio
p Radius of hole, in. (cm)
Fatigue failures are caused by the initiation and cyclic growth of one or
more cracks from areas of stress concentrations. The rate at which the
crack grows depends primarily on the material, specimen configuration,
the cyclic loads applied, and the environmental conditions. The cracks
continue to propagate under repeated loading until a combination of
crack length and load is reached such that static failure occurs.
In the last decade various crack-propagation relations have been devel-
oped which were based on and applied to the results of tests on uniformly
loaded sheet specimen. However, very little effort has been directed to-
ward developing analytical methods of treating crack growth in built-up
structures.
In the present investigation analytical and experimental studies were
conducted on configurations similar to those occurring in built-up struc-
tures. The tests ranged from the simple cases of cracked panels subjected
to either uniform or concentrated loads to more complex cases where
combinations of uniform and concentrated loads were applied to simulate
the loading in a riveted skin-stiffener panel.
The method of prediction was based upon calculation of a stress-in-
tensityCopyright
factor near the crack
by ASTM tip rights
Int'l (all and upon an experimentally
reserved); determined
Mon Dec 7 14:40:45 EST 2015
relation between stress-intensity
Downloaded/printed by and rate of crack propagation observed
FIGGE AND NEWMAN ON STRUCTURES WITH SIMULATED RIVET FORCES 73
in previous studies [1].2 The calculation methods for complex configu-

rations are described. In general, predictions are shown to be in good
agreement with test data.
Basic Consideration
In order to provide a suitable stress parameter to correlate the crack-
growth rates under cyclic loading, the stress-intensity concept was used in
the analysis. The concept is derived from the elastic-stress distribution
near the tip of a crack [3-6] as given by the equations in Fig. 1. The
form of these equations is identical for any case in which the loading is
perpendicular to and symmetric about the plane of the crack. The stress-
intensity factor k is linearly dependent upon load and is a function of
crack length and specimen configuration.
^•^^
FIG. 1—Coordinate system and equations for elastic stresses in vicinity of a

crack.
The correlation between stress-intensity factor and crack-growth rate,

hereafter called A-rate curve, for 7075-T6 aluminum alloy tested at a
load ratio of 0.05 is presented in Fig. 2. The solid and dashed curves in
Fig. 2 represent the mean and extremes of experimental data, respec-
tively. The data shown in the figure were obtained from fatigue tests con-
ducted on two types of simple specimens: uniformly loaded 8 by 24-in.
(20 by 61-cm) panel containing a central crack and 12-in.-wide (30-cm)
wedge-force panel (concentrated forces on the crack surface (see Fig. 2),
ranging in height from 5 to 25 in. (13 to 64 cm). The panel dimensions
are given in Table 1.
The stress-intensity factor for a panel with uniformly applied stress is:
2
The italic numbers
Copyright by ASTMin Int'l
brackets refer reserved);
(all rights to the listMon
of references appended
Dec 7 14:40:45 ESTto2015
this
paper.Downloaded/printed by
where Fi(a/b) is a boundary-correction factor which modifies the stress

intensity for an infinite plate to account for the influence of the finite width
of the panel specimen. The boundary-correction factor was obtained from
the elastic-stress-concentration solution for an elliptic hole in a strip of
infinite height and finite width under a uniformly applied stress [7]. The
influence of specimen height 2h and the grip constraint were assumed to
be of minor importance when the height-to-width ratio was greater than
two.
FIG. 2—Characteristic crack-growth rate for 7075-T6 aluminum alloy sheet

at R = 0.05.
The stress-intensity factor for the wedge-force is:
where F2(a/b, a/h) is a boundary-correction factor accounting for both

finite width and height. The boundary-correction factor was approxi-
mated by superposing the solutions for a strip of infinite height and finite
width [8] and one of infinite width and finite height [9] and resulted in
the following expression:
University of Washington (University of Washington) pursuant to License Agreement.
TABLE 1—Specimen configurations and dimensions.
where
In the uniformly loaded panel, the crack-growth rate increases as the

crack grows, while in the wedge-force panel the rate decreases as the
crack grows. However, the stress-intensity factors for these two cases ac-
count for this difference, as is demonstrated by the close agreement of the
data inDownloaded/printed
Fig. 2. by
In general,
University ofsolutions existed in
Washington (University the literature
of Washington) pursuantfor cracked
to License panels
Agreement. sub- reproduction
No further
jected to either uniform or concentrated forces. For complex loading

cases, the stress intensity was computed by superposition of stress inten-
sities for simple cases. For cases where a stress-intensity solution was not
available, stress-intensity values were calculated from strain measure-
ments made on the panel prior to introducing the crack. (See Appendix
II for detailed discussion.)
In cases where the configuration was such that the crack grew from
only one side of a hole, it was necessary to use an equivalent crack since
finite-boundary corrections were not available for a combination of hole
and crack. In some cases where pins were used to apply the forces, it was
necessary to modify the stress-intensity factor to account for the influ-
ence of the finite size of the pin on the stress distribution. These modifi-
cations are discussed in detail in later sections.
The curve of crack-length against cycles was then obtained as follows.
The values of fcmax at crack lengths of (a) and (a + Aa) were calculated
and the corresponding values of crack-growth rates were determined
from Fig. 2. The number of load cycles required to produce an increment
Aa of crack growth was computed by the following equation:
The total number of cycles required to propagate a crack to a given

length is then the sum of AAPs to that crack length.
Equipment and Test Procedure
Five specimen configurations approximating the complexities that
might be encountered in built-up structures were studied. The panels
were constructed of 0.90-in. (2.28-cm) thick 7075-T6 aluminum alloy.
The dimensions of each configuration are presented in Table 1. The ma-
terial was obtained from a special fatigue stock [10] retained at Langley
Research Center for fatigue testing.
A crack starter perpendicular to the direction of loading in the form
of a slit was produced in each specimen by either a saw cut or by a
spark-discharge technique. A grid was photographically printed on each
specimen to facilitate measurement of crack growth. The grid spacing
was 0.05 in. (1.27 mm). Both metallographic and tension tests revealed
that the grid had no detrimental effect on the material. In order to follow
the crack growth, the fatigue cracks were observed through a 30-power
microscope while illuminated by stroboscopic light.
The axial-load fatigue testing equipment used in this investigation in-
cluded a subresonant machine [11], an inertia-force compensating ma-
chine, and a combination hydraulic and subresonant machine [72]. The
subresonant machine had an operating frequency of 1800 cpm (30 Hz)
and a University
load capacity of ±20,000
of Washington Ib (89
(University kN). Thepursuant
of Washington) inertia-force
to Licensecompen-
Agreement. No furth
sating machine had an operating frequency of 1200 cpm (20 Hz) and a
load capacity of ±20,000 pounds (89 kN). The combination hydraulic
and subresonant was operated only in the hydraulic mode. In this mode
the operating frequency was 50 cpm (0.8 Hz) and the load capacity
132,000 Ib (586 kN).
The concentrated forces were applied with special fixtures described
in Appendix III.
Analysis and Experimental Results for Panels Subjected to Various

Combinations of Load
In the following section the modifications made on the theoretical
stress-intensity factors for each configuration are discussed in detail.
Comparisons are made between the experimental and predicted curves
of crack-length against cycles for each loading configuration. The crack-
propagation data are presented in Table 2. The predicted mean and the
probable scatter in crack-length against cycles data are presented in the
figures as solid lines and shaded areas, respectively. The bounds of this
area were predicted from the extremes of the experimental data as shown
by the dashed lines in Fig. 2. In general, the observed behavior fell within
these bounds.
Case A—Growth of a Fatigue Crack from One Side of a Hole Located

Eccentrically in a Uniformly Loaded Panel
For this case (Fig. 3), the assumption was made that the combination
of crack and hole could be represented by an equivalent crack. Thereby,
the problem of an eccentric hole and single crack is reduced to the case
of an eccentric crack in a panel. The reason for using an equivalent
crack was that finite-boundary corrections did not exist in the literature
for the case of an eccentric hole and crack, but did exist for the case of
an eccentric crack in a panel.
The equivalent crack length was obtained by equating the stress inten-
sities for an infinite plate containing a hole and single crack with that for
an infinite plate with a crack. The stress intensities were obtained from
Refs 13 and 8, and are, respectively:
where Fs(L/p) is a function which describes the influence of the hole on

the crack-tip stress intensity. Equating these solutions and solving for ae
results in the relation for the equivalent crack length.
TABLE 2—Crack propagation characteristics of specimens tested.
Case A
Number of cycles required to propagate a crack from total length of 0.15 in. (0.38 cm) to a length of:
in .0.35 0.55 0. 75 0.95 1.15 1.35 1.55 1.75 2.15 2.35 2.55 2.75 3.15 3.35 3. 55 3.75 3.95 4.35
cm .0.89 1 40 1 91 2.41 2.92 3.43 3.94 4.45 5.46 5.97 6.48 6.99 8.00 8.51 9.02 9.53 10.03 11.05
Specimen 1 ..314 512 660 768 877 953 1016 1073 1162 1199 1234 1270 1332 1356 1374 1393 1410 1439 X 103
Specimen 2 ..302 508 649 765 864 940 1012 1070 1172 1219 1258 1296 1362 1393 1420 1444 1468 1510 X 103
Case B
in .1.20 1.40 1.60 1.80 2.00 2.20 2.40 2.60 2.80 3.00 3.20 3.40 3.60 3.80 4.00
cm ..3.05 3.56 4.06 4.57 5.08 5.59 6.10 6.60 7.11 7.62 8.13 8.64 9.14 9.65 10.16
Specimen 1 ...271 508 688 849 998 1150 1230 1355 1442 1543 1638 1726 1822 1910 1984 X 102
Specimen 2 406 686 914 1116 1274 1418 1578 1768 1876 1963 2079 2210 2308 2425 X102
Case C
Number of cycles required to propagate a crack from total length of 1 .00 in. (2,,54 cm) for Specimen 1 and 0.80 in. (2.03 cm) for Specimen 2
to a length of:
in 1.00 1.20 1.40 1.60 1.80 2.00 2.20 2.40 2.60 2.80 3.00 3.20 3.40
cm Copyright by ASTM2.54Int'l (all
3.05 3.56 Mon
rights reserved); 4.06Dec 74.57
14:40:455.08
EST 20155.59 6.10 6.60 7.11 7.62 8.13 8.64
Specimen 1
112 300 427 494 543 592 604 707 781 855 933 1020 X103
Specimen 2 . .93.4 170 345 500 579 621 668 719 112 835 917 1004 1079 X103
Case D
in 1.20 1.40 1.60 1.80 2.00 2 .20 2.40

cm 3.05 3.56 4 .06 4.57 5.08 5 .59 6.10
Specimen 1 115 213 275 326 360 390 413 X102
Specimen 2 H0 18 25 31 36 40 42 XlO 2 (without concentration forces)
Case E
in 0.20 0 60 1 00 1 48 2.00 2 42 2 62 2.80 2 98 3.02 3 12 3 25 3 55 3 93 4.45 5.00 5.50

cm 0.51 1.52 2.54 3.76 5.08 6.15 6.65 7.11 7.57 7.65 7.92 8.25 9.00 9.98 11.30 12.70 13.97
Specimen 1 0.0 250 450 630 820 1100 1350 1650 2000 2340 2660 3000 3420 3680 3880 4000 4050 X 102
The equivalent crack length calculated by Eq 8 is very nearly equal to

the sum of the actual crack length plus hole diameter. The stress-intensity
factor for Case A is then:
where Fê/b, ajb) is the boundary-correction factor which accounts

for the crack eccentricity and finite width [14].
The predicted crack propagation curve for this specimen is shown in
Fig. 3. The symbols are plots of data obtained from two identical tests.
The shaded area shows the probable scatter that could have been obtained
FIG. 3—Fatigue crack propagation from one side of a hole located eccen-
trically in a uniformly loaded panel.
from these tests. The agreement between the actual and predicted was
considered good.
Case B—Growth of a Symmetrical Fatigue Crack in a Panel Loaded with
Concentrated Forces
The simulation of rivet forces in the testing laboratory is important to
the understanding of the parameters which influence fatigue crack growth
in the vicinity of rivets in structures. Some of the major parameters are
rivet size, rivet pitch, and load.
Case B (Fig. 4) simulated the growth of a fatigue crack in a riveted
doubler. Cyclic loads were applied by two pins located along the vertical
center Copyright
line of theby panel.
ASTM InInt'lthis
(all case,
rights the stress-intensity
reserved); Mon Dec equation
7 14:40:45for
EST 2015
loads applied at a point was
Downloaded/printed by assumed to be applicable since the pins were
locatedUniversity
more than of five diameters
Washington away from
(University the plane of
of Washington) the crack.
pursuant to License Agreemen
The stress-intensity factor for a crack in an infinite plate with sym-

metric point loads was obtained from Ref 75. The solution for a finite
panel is:
where F2(a/b, a/h) is an approximate boundary-correction factor (see

Eq 3).
The prediction of the crack length-cycles curve for this specimen is
shown hi Fig. 4. The symbols represent the data obtained from tests of
I -2
a
FIG. 4—Fatigue crack propagation for a symmetrically cracked panel sub-

jected to concentrated forces.
two identical specimens. In general, the agreement between the data and
predicted results was considered reasonable.
Case C—Growth of a Fatigue Crack from One Side of a Hole Located

Eccentrically in a Panel Subjected to Concentrated Forces
This specimen configuration (Fig. 5) is similar to that of Case A, ex-
cept that the loading was by concentrated forces. In this case, the pins
were located reasonably close to the plane of the crack and thus the finite
pin size was expected to influence the crack growth. The crack length-
cycles predictions are presented as two curves; one was obtained by using
the theoretical solution based on the stress-intensity factors for load ap-
plied at a point, and the other was obtained by using stress-intensity
values calculated from strain-gage measurements made on the panel
prior Copyright
to introducing
by ASTM a fatigue
Int'l (all crack
rights (see Fig. Mon
reserved); 6). Dec 7 14:40:45 EST 2015
The theoretical stress-intensity
Downloaded/printed by factor for a crack (which is eccentric
with University
respect toofthe line of load
Washington application)
(University in an pursuant
of Washington) infinite toplate subjected
License Agreement. No furt
82 TIGUE CRACK PROPAGATION
trically in a panel subjected to concentrated forces.
FIG. 6—Measured and theoretical stress distribution in a panel subjected to

pin loads.
to concentrated forces was obtained from Ref 16. This solution was modi-
fied by using the equivalent crack length as in Case A. The stress intensity
for Case C is:
The quantities /i and

Copyright by ASTM Int'l (all7rights
2 are functions
reserved); Mon Decof equivalent
7 14:40:45 EST 2015 crack length and ac-
countDownloaded/printed
for the coordinateby location of the concentrated forces as given by:
where
The function, F2(ae/b, ae/h), is an approximate boundary-correction fac-

tor that accounts for the fact that the specimen has a finite width and a
finite height (see Eq 3). The influence of the crack eccentricity in this
case was neglected because the ratio of crack eccentricity e to plate di-
mension (b or K) was less than 0.1.
FIG. 7—Stress-intensity factors for a fatigue crack growing from one side of a
hole located eccentrically in a panel subjected to concentrated forces.
The crack length-cycles predictions based on Eq 11 or on the strain-
gage measurements are shown in Fig. 5 as solid and broken curves, re-
spectively. The shape of the latter curve approaches the trend of the data
and produces substantially better agreement than the theoretical solu-
tion.
The theoretical solution, Eq 11, and the stress-intensity values calcu-
lated from the strain-gage readings are shown as solid and dashed lines,
respectively, in Fig. 7. The symbols represent the values of &max obtained
84FATIGUE CRACK PROPAGATION
by measuring the crack-growth rate at a given crack length and determin-

ing from the £-rate curve (solid line, Fig. 2) the stress-intensity factor
which causes the same crack-growth rate. The stress-intensity factors
obtained from the strain-gage readings are in good agreement with the
data when the crack length L is less than 1 Yz in. (4 cm).
The disagreement between the values at crack lengths greater
in. (4 cm) was attributed to a secondary influence of the pinhole and pin
on the crack-tip stress intensity, since the stress-intensity solution for an
arbitrarily stressed crack surface in an infinite plate was used to calculate
the stress-intensity factors.
FIG. 8—Stress-intensity factors for a symmetrically-cracked panel subjected to

uniform load and concentrated forces.
In general, the theoretical solution disagreed with the data over the
entire range of crack lengths.
Case D—Growth of a Symmetrical Fatigue Crack in a Panel Subjected

to Uniform Cyclic Tension Loading and Concentrated Compressive
Forces
This particular case (Fig. 8) is of interest since it is similar to the actual
case of a crack initiating and propagating in the skin material under a
riveted stiffener. The simulated rivet loadings varied from those in the
actual structure in that the forces were applied through pins (simulated
rivets) by a hydraulic jack under steady pressure. The forces fluctuated
slightly as a result of the cyclic uniform loading, but at all times tended
to close the crack. The local stress produced by the combination of uni-
form Copyright
and concentrated forces
by ASTM Int'l (all thus
rights varied between
reserved); tension
Mon Dec and compres-
7 14:40:45 EST 2015
sion. Downloaded/printed by
A test was also
University conducted
of Washington in which
(University only uniform
of Washington) loads
pursuant to were
Licenseapplied
Agreement. No fu
to the panel to study the influence of the concentrated forces on crack

growth.
Several basic assumptions were made in the analysis of the complex
configuration. First, the stress-intensity solution for theoretical point
loads was considered applicable, since the pins were located a considera-
ble distance (4 diameters) away from the plane of the crack. Secondly,
the concentrated forces were assumed to be applied at the midpoint of the
bearing interfaces of the pins and sheet. Finally, the A>rate curve for a
load ratio of 0.05 was considered applicable for the combined loading,
since it has been shown [17, 18] that the compressive portion of the load
cycle does not contribute appreciably to the rate of crack growth.
The stress intensity for this case was obtained by superposing the
stress intensity for a central crack subjected to uniform load [7, 8] and
that for a central crack subjected to concentrated forces [15]. These
solutions were corrected for finite width (see Ref 7) and are, respectively:
Since the forces tended to close the crack, it was necessary to use Eq
13 with a negative sign to account for the direction of loading. The two
solutions and their algebraic sum are represented graphically in Fig. 8.
The experimental results and the predictions obtained by using the re-
sultant stress-intensity expression are presented in Fig. 1. The predicted
number of cycles, based on the mean /:-rate curve, was slightly less than
the experimental results.
Crack propagation in the absence of concentrated forces is, as ex-
pected, much more rapid than with concentrated forces applied as shown
by the symbols on the left of Fig. 9.
Case E—Growth of a Fatigue Crack from One Side of a Hole Located

Eccentrically in a Panel Subjected to Uniform Load and Concen-
trated Forces
This case (Fig. 10) is a combination of loading and configuration,
similar to Case A and Case C, and represents a fatigue crack propagating
from one side of an eccentric hole toward a row of rivets in a stiffened
panel. The concentrated forces were applied in the same manner as in
Case D.
The resultant stress-intensity factor for this case was obtained by su-
perposition of the stress intensities for the uniformly distributed stress
Copyright by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 ES
and for the concentrated forces.
Downloaded/printed by The stress-intensity factor for a single
crack University
growing from one side of an(University
of Washington eccentric hole
of in a panel subjected
Washington) to to Licen
pursuant
86 8FATIGUE CRACK PROPAGATION
uniformly applied stress was obtained by using the equivalent crack

length ae as in Case A, and is:
where F^(e/b, ae/b) is the boundary-correction factor which accounts

for the crack eccentricity and finite width [14].
FIG. 9—Fatigue crack propagation for a symmetrically cracked panel sub-

jected to uniform load and concentrated forces.
The stress-intensity factor for concentrated forces applied to the same

configuration is:
A boundary-correction factor was not available for this loading configu-

ration. However, it was assumed that the correction factor applied to the
uniformly loaded panel would give reasonable results. The quantities /i
and /2 are defined hi Case C.
Since the concentrated forces are tending to close the crack, the stress-
intensity factor for the concentrated forces (Eq 14) is used with a nega-
tive sign. The algebraic sum of the stress intensities for these two cases is
represented graphically in Fig. 10.
While the uniformly applied stress was cycled at a load ratio, R =
0.05, Copyright
the concentrated
by ASTM Int'lforces
(all rightsfluctuated between
reserved); Mon the limits
Dec 7 14:40:45 shown in Fig.
EST 2015
11. However, the fc-rate curve for a load ratio of 0.05 was considered
University of Washington (University of Washington) pursuant to License Agreement. No further reproducti
FIG. 10—Stress-intensity factors for a fatigue crack growing from one side of
a hole located eccentrically in a panel subjected to uniform load and concentrated
forces.
trically in a panel subjected to uniform load and concentrated forces.
applicable for the combined loading, since the compressive portion of the
load cycle does not contribute appreciably to crack growth.
TheCopyright
predicted curve
by ASTM Int'lof(all
crack
rights length against
reserved); Mon Deccycles for this
7 14:40:45 specimen is
EST 2015
shownUniversity
in Fig.of 11, together with the data obtained from one test. The
Washington (University of Washington) pursuant to License Agreement. No further rep
FIG. 12—Measured stress distribution in a uniformly-loaded panel containing

a hole with a pin.
FIG. 13—Measured stress distribution in a panel subjected to pin loads.
agreement between the actual and predicted cycles was considered good
except when the crack tip was located between the concentrated forces.
This disagreement was attributed to the reduction in stress along the
plane of the crack due to the finite pin size in a manner similar to that
encountered in Case C. The effect of these pins on the stress distribu-
tions is presented in Fig. 12 and 13.
Conclusion
Analytical and experimental studies were conducted on the rates of
fatigue crack propagation in sheet specimens subjected to combinations
of uniform and concentrated forces similar to those occurring in built-up
structures.
University of Washington (University of Washington) pursuant to License A
These studies indicated that reasonable predictions of crack length

against cycles were obtained for panels subjected to complex loadings
when the stress state near the tip of the crack was defined by stress-inten-
sity factors obtained from analytical solutions or derived from strain-gage
measurements.
Although the loading conditions studied were considerably more com-
plex than those generally studied in the past, they are still relatively
simple when compared to the loadings that occur in actual aircraft
structures. Further work is needed to evaluate more completely the re-
liability of this method. Ultimately, this analytical tool should prove use-
ful in designing structures that are more resistant to fatigue and less
vulnerable to catastrophic failure due to cracks.
APPENDIX I
Conversion of
The International System of Units (SI) was adopted by the Eleventh Gen-
eral Conference of Weights and Measures, Paris, October 1960, in Resolu-
tion No. 12 (Ref 3). Conversion factors for the units used herein are given in
the accompanying table:
To Convert from U.S. Multiply by To Obtain SI Units

Customary Units
Ibf 4.448222 newton (N)

in 2.54 X 10~2 meter (m) 2
ksi 6.894757 meganewton/meter (MN/rn2)
Ib-fin /"
3 2
1.099 X 10-3 MN-m~ 3 / 2
com 1.67 X 10~2 hertz (Hz)
Prefixes and symbols to indicate multiples of units are shown in table here-
with.
Multiple Prefix Symbol
10-6 micro u
10-3 milli m
10~2 centi c
3
10 kilo k
106 meea M
Copyright by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST
University of Washington (University of Washington) pursuant to License
APPENDIX II
Experimental Determination of Stress-Intensity Factors from Strain-Gage
Measurements on Uncracked Panels
The stress-intensity factors for a panel subjected to loadings perpendicular
to and symmetric about the plane of the crack can be determined experi-
mentally from strain-gage measurements made along the plane of crack prop-
agation prior to introducing a fatigue crack.
For example, in Fig. 14a, the plate is subjected to concentrated forces. The
FIG. 14—Superposition of two cases for the determination of stress-intensity

factors.
strain distribution along the ;t-axis is measured by the strain gages. If the
corresponding stress distribution is reversed and applied as shown in Fig. 146
(compression on crack surfaces), the superposition of Figs. I4a and b gives the
stress solution for the case of a plate with concentrated forces containing a
stress-free crack boundary (see Fig. 14c).
The stress intensity, k, for Fig. 14a is zero; therefore, the stress intensities
for Figs. I4b and c are equivalent. The resultant stress-intensity factor for an
infinite plate [15] is:
The accuracy of this equation in calculating stress-intensity factors from

strain measurements made on a finite panel depends on how much the strain
distribution was influenced by the specimen boundaries.
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions a
APPENDIX III
Test Fixtures
Several specially designed fixtures were used in this investigation to apply
concentrated forces to the test specimens. Cyclic tension forces were applied
FIG. 15—Fixtures used to apply cyclic concentrated forces.
FIG. 16—Schematic of fixture used to apply concentrated compressive forces.
with the sets of straps shown in Fig. 15. Set A was used for the cases in which
the pin holes were remote from the crack. Set B was used when the loads were
applied directly to the crack surface. In both cases, the pins extended through
the specimens with a strap located on each side. In the latter case, the pins
were semicircular in cross section, and two were inserted in the same hole.
The opposite ends of the straps were attached directly to the loading heads
of the testing machine.
Compressive forces were applied using the fixture shown schematically in
Fig. 16. The fixture consisted of a small hydraulic jack and sets of straps
which transmitted the loads through pins to the specimen. The outer straps
were attached to the jack ram by a pin which passed through slots in the inner
straps. The inner straps were attached to the cylinder of the jack. The inner
and outer straps were in the same plane (this feature is not shown in the
schematic). Two view ports were provided in the legs of the outer straps to
permit observations of the crack as it grew under the legs. Strain gages were
mounted on each leg of the outer straps to monitor loads. The jack was
mounted in a hole in the specimen located approximately 6l/z in. from the
plane of the crack. The center of the jack was on the center line of the speci-
men in both the width and thickness direction. Strain-gage measurements in-
dicated that the effect of this hole on the strain in the region where predic-
tions were made was less than 1 per cent and was thus considered negligible.
Pressure was applied to the jack by an accumulator and was set at the desired
value prior to application of the uniform cyclic loads. The pressure, and thus
load, fluctuated as the specimen was cycled. The minirrium and maximum
compressive forces on the specimen occurred when the uniform cyclic forces
were a minimum and maximum, respectively.
References
of Fatigue," The Trend in Engineering, University of Washington, Seattle,
Wash., Vol 13, No. 1, January, 1961, p. 9.
[2] E. A. Mechtly, "The International System of Units—Physical Constants and
Conversion Factors," NASA SP-7012, 1964.
[3] I. N. Sneddon, "The Distribution of Stress in the Neighborhood of a Crack
in an Elastic Solid," Proceedings, Royal Society, London, Vol A-187, 1946.
[4] G. R. Irwin, "Analysis of Stresses and Strains Near the End of a Crack
Traversing a Plate," Transactions, Am. Society Mechanical Engrs., 1957.
[5] G. R. Irwin, "Fracture," Handbuch der Physik, Vol VI, Springer-Verlag, Ber-
lin, 1958.
[6] M. L. Williams, "On the Stress Distribution at the Base of a Stationary
Crack," Transactions, Am. Society Mechanical Engrs., Journal of Applied
Mechanics, 1957.
[7] M. Isida, "On the Tension of a Strip with a Central Elliptic Hole," Transac-
tions, Japan Society Mechanical Engrs., Vol 21, No. 107, 1955, p. 511.
[8] P. Paris and G. Sib., "Stress Analysis of Cracks," Fracture Toughness Testing
and Its Applications, ASTM STP 381, Am. Soc. Testing Mats., 1964.
[9] W. B. Fichter, "Effects of Strip Width on Stresses at the Tip of a Longitudi-
nal Crack in a Plate Strip," paper presented at 5th U.S. National Congress of
Applied Mechanics, Minneapolis, Minn., June, 1966.
[10] H. J. Grover, S. M. Bishop, and L. R. Jackson, "Fatigue Strengths of Air-
craft Materials. Axial-Load Fatigue Tests on Unnotched Sheet Specimens of
24S-T3 and 75S-T6 Aluminum Alloys and of SAE 4130 Steel," NACA TN
2324, 1951.
[11] H. J. Grover and W. S. Hayler (Battelle Memorial Inst.); Paul Kuhn and
C. B. Landers (NASA Langley); and F. M. Howell (Alcoa), "Axial-Load
Fatigue Properties of 24S-T and 75S-T Aluminum Alloy as Determined in
Several Laboratories," NACA TR 1190, May, 1953.
[12] C. Michael Hudson and H. F. Hardrath, "Investigation of the Effects of
Variable-Amplitude Loadings on Fatigue Crack Propagation Patterns,"
NASA TN D-1803, August, 1963.
[13] O. L. Bowie, "Analysis of an Infinite Plate Containing Radial Cracks Origi-
nating From a Circular Hole," Journal of Mathematics and Physics (MIT),
Vol 23, No. 1, April, 1956, pp. 60-71.
[14] M. Isida, "Crack Tip Stress Intensity Factors for the Tension of an Eccen-
trically Cracked Strip," Department of Mechanics, Lehigh University, Beth-
lehem, Pa., 1965.
[15] P. Paris, "Application of Mushelishvillis' Methods to the Analysis of Crack

Tip Stress Intensity Factors for Plane Problems," Fracture Mechanics Re-
search for the Boeing Airplane Co., Institute of Research, Lehigh University,
Bethlehem, Pa., 1960.
[16] F. Erdogan, "The Stress Distribution in an Infinite Plate with Two Colinear
Cracks Subjected to Arbitrary Loads in Its Plane," Institute of Research,
Lehigh University, Bethlehem, Pa., June, 1961.
[17] D. R. Donaldson and W. E. Anderson, "Crack Propagation Behavior of Some
Airframe Materials," Crack Propagation Symposium, The College of Aero-
nautics, Cranfield, England, 1961.
[18] W. Illg and A. J. McEvily, Jr., "The Rate of Fatigue Crack Propagation for
Two Aluminum Alloys Under Completely Reversed Loading," NASA TN
D-52, October, 1959.
T. W. Crooked andE. A. Lange*
Low Cycle Fatigue Crack Propagation

Resistance of Materials for Large
Welded Structures
REFERENCE: T. W. Crocker and E. A. Lange, "Low Cycle Fatigue

Crack Propagation Resistance of Materials for Large Welded Structures,"
Fatigue Crack Propagation, ASTM STP 415, Am. Soc. Testing Mats.,
1967, p. 94.
ABSTRACT: The development and application of modern, higher strength
materials for large welded structures, combined with the realization that
flaw-free fabrication cannot be ensured, have created a need for methods
to evaluate the low cycle fatigue crack propagation resistance of structural
metals. In this paper a test method is discussed which has been applied to
a broad spectrum of structural metals, both ferrous and nonferrous, in-
cluding yield strengths from 30 to 200 ksi. The results apply to structural
lives ranging from 100 to 100,000 cycles. Data are presented in terms of
an empirical power-law relationship between fatigue crack growth rate
and total (elastic plus plastic) strain range. Comparisons are made among
the low cycle fatigue crack propagation characteristics of competitive ma-
terials, and a first-order translation of crack growth rates to an index of
structural fatigue life is developed. The influences of aqueous environments
and mean strain are described.
KEY WORDS: fatigue (materials), crack propagation, low-cycle fatigue,
weldments
It is generally accepted that many large welded structures are sus-

ceptible to failure from low cycle fatigue. Examples range from rocket
cases, which must sustain several proof-testing pressurizations, to sub-
mersible-vehicle hulls, which must withstand repeated hydrostatic load-
ing excursions spanning years of service, not to mention the far more
numerous structures in commercial transport and industrial processes
which daily expend a portion of their useful service lives. This paper
describes a continuing research effort aimed at providing new knowledge
1
Research materials engineer, Plastic Flow Section, Strength of Metals Branch,
Metallurgy Div., U.S. Naval Research Laboratory, Washington, D. C. Personal
member ASTM.
2
Head, Plastic Flow Section, Strength of Metals Branch, Metallurgy Div., U.S.
Naval Research Laboratory, Washington, D. C.
94
CROCKER AND LANGE ON WELDED STRUCTURES 95
for the safe and dependable application of modern higher strength ma

terials to large cyclically loaded structures. The information presented is
primarily relevant to fatigue failures in the service life span between
100 and 100,000 cycles.
This work is grounded on the philosophy that the crack propagation
phase of fatigue failure is the most suitable criterion for predicting the
life of large, complex welded structures. Experience has shown that
nondestructive testing techniques cannot ensure flaw-free fabrication ol
large welded structures even in cases where unusual efforts, using every
available technique, are made to achieve a very high quality of fabrica-
tion and inspection. Consequently, there is a high probability that
structural fatigue life will depend upon the growth of undetected sub-
critical flaws. Well-documented examples which substantiate this ap-
proach are to be found in the experiences of the Pressure Vessel Research
Committee's full-size pressure vessel fatigue tests [7]3 and the recent
catastrophic hydrotest failure of a large rocket motor case [2].
A second guiding principle in this work has been to initially empha-
size studies of material characteristics rather than structural models or
components. Model or component fatigue tests are necessary in the
final analysis for specific applications but do not provide the opportunity
for a singular study of materials. Just as it is necessary to separate crack
initiation from crack propagation for definitive purposes, it is also neces-
sary to evaluate the separate roles of material characteristics, structural
design features, and quailty of fabrication. The literature contains an
abundance of fatigue data which inseparably combine two or more
relevant factors into an arbitrary definition of fatigue failure without
attempting to define the contribution of each.
The scope of this fatigue study has included a broad spectrum of
materials: ferrous and nonferrous, and low strength and high strength,
since structural fatigue failures from undetected flaws can occur in
nearly all classes of structural metals. The study also includes the effects
of aqueous environments and mean strain on the growth of fatigue
cracks in the low cycle life region. The ultimate goal of this approach
to the problem of low cycle fatigue is to provide a more universal and
accurate procedure for predicting the low cycle fatigue life of structures.
Experimental Program
Fatigue Machines and Test Specimen

The experimental apparatus employed in this study consists of a bank
of six specially built hydraulic fatigue machines which can separately
3
paper.
cycle plate bend specimens in cantilever fashion. These machines are

shown in Fig. 1, and the test specimen is shown in Fig. 2. A single 1200-
psi hydraulic pump powers the system, and each machine is actuated by
a separate, double-acting, 2.5-in.-diameter hydraulic cylinder at a fre-
quency of approximately 5 cpm.
Automatic operation is obtained by deflection control of the specimen.
Micrometer-adjusted microswitches contact the specimen at the maxi-
mum and minimum points in the loading cycle, causing a signal to re-
verse specimen movement. Deflections are measured by dial indicators
at a common reference point on the specimen moment arm. A precision
load cell is placed in series with the actuating hydraulic cylinder to pro-
FIG. 1—General features of the low cycle fatigue test equipment.
vide a continuous and accurate load readout. These machines can pro-
vide any desired loading cycle, from full-reverse balanced tension-com-
pression to offset tension-tension, with strain, deflection, or load as the
controlled loading parameter.
The plate bend fatigue specimen, shown in Figs. 2 and 3, was originally
developed to study the fatigue characteristics of pressure vessel steels [3],
It is widely used for fatigue crack initiation studies involving plate thick-
ness sections and plastic strains in the full spectrum of structural metals.
This specimen possesses several important advantages for low cycle
fatigue crack propagation studies in plate sections: the specimen can be
loaded to high strain levels in both tension and compression with rela-
tively simple and inexpensive machines, and the test section geometry
permits direct observation of the fatigue crack and direct measurements
of nominal surface strains.
FIG. 2—The plate bend fatigue specimen. Note the machined surface notch
to facilitate crack initiation and the foil-type strain gage for measuring the strain
range.
FIG. 3—Dimensions of the plate bend fatigue specimen.

For fatigue crack propagation studies, the plate bend specimen is

modified by engraving a lA-m. surface notch at the center of the test
section, Fig. 3, to act as a crack starter. The notch is sharpened with a
punch to facilitate rapid crack initiation, usually within a few hundred
cycles of repeated load. With this technique nominal elastic loading is
sufficient to initiate the fatigue crack except in very tough materials, thus
leaving the mechanical properties of the test section unaffected by the
crack initiation process. Once initiated, the fatigue crack propagates
across the width of the test section, following a path along the plane of
minimum section thickness which is perpendicular to the principal bend-
ing strain.
Testing Procedure
Strain is used throughout this paper as the parameter which describes
the nominal intensity of loading hi the material under test. No other
parameter for intensity of loading is universally applicable to a study
where many of the materials tested undergo plastic loading. Neither the
bending stress nor the fracture mechanics stress intensity parameter can
be calculated with validity for the loading conditions employed in this
study. Furthermore, nominal strains are easily measured by common
engineering techniques, both in the plate bend specimen and in many
other specimens or structural configurations where comparisons may be
drawn.
Nominal net section strains are measured with a lA-m. gage-length
resistance strain gage placed at the minimum test section and located
well in advance of the center fatigue crack. Metalfoil strain gages put
down with pressure activating cement are used for this task. Under cyclic
loading, periodic replacement of strain gages is necessary at intervals
which become more frequent with increasing strain levels due to fatigue
of the gage cement. Strain gage signals are combined with the load cell
signals on an X-Y recorder to generate load-versus-strain charts.
The first step in the experimental procedure is to obtain the baseline
strain range-versus-deflection characteristics of an unnotched specimen
of the test material. This baseline information is useful for control pur-
poses in determining the elastic limit of the materials in the plate bend
specimen configuration and in estimating the strain range and deflection
values relevant to low cycle fatigue.
A sample of this sort of baseline information is presented in Fig. 4,
which shows a typical strain range versus deflection amplitude relation-
ship and includes a schematic load-versus-strain chart of the full-reverse
strain cycle employed. The upper curve is the relationship for total
strain range (elastic plus plastic), and the lower curve is the relationship
for plastic strain range. These curves are obtained for each new ma-
terial by slowly cycling an unnotched specimen in full-reverse strain
CROOKER AND LANGE ON WELDED STRUCTURES 99
loading with a strain gage located at the midpoint of the test section. A
load-versus-strain chart is generated by an X-Y recorder, and deflection
values from the dial indicator are tabulated for each cycle. Deflection
limits are increased after each cycle. The development of plastic strains
is indicated by the point of nonlinearity in the strain range-versus-deflec-
tion curve and, more definitively, by an opening of the load-versus-strain
trace which forms a hysteresis loop.
FIG. 4—Schematic illustration of the relationships between total and plastic

strain ranges and deflection amplitude. Note that the proportional limit strain
range is set at 500-/J. in.I in. plastic strain range.
Since only a few cycles are involved in gathering data for these base-
line relationships, this procedure amounts to a measurement of static
materials properties. It is recognized that after several hundred cycles of
repeated load involving plastic strains, the mechanical properties of
metals are changed. However, for the majority of the materials discussed
in this report, the tests were conducted at nominally elastic loading or
in the range of very small plastic strains; consequently, work-hardening
or work-softening effects are of less importance to this study than in low
cycle fatigue crack initiation studies which involve large plastic strains in
smooth specimens, especially in smooth specimens of lower strength

material.
The total strain range at which the plastic strain range reaches 500
pin./in. is arbitrarily defined as the proportional limit in the plate bend
specimen, and this proportional limit is indicated on Fig. 4. The plastic
strain level of 500 //in./in. is a small value of plastic strain, and since it
can be measured with consistent accuracy, it is employed as a benchmark
for indicating the onset of nominal plastic loading. The amplitude of the
proportional limit strain range is less than the strain amplitude at the
0.2 per cent offset yield strength for lower strength materials, but it ap-
proaches yield strength strain in higher strength materials.
FIG. 5—Schematic illustration of the effect of loading conditions, constant

deflection amplitude, and constant total strain range on fatigue crack propagation
in the plate bend specimen.
Figure 5 is a plot of total surface crack length versus cycles of re-

peated load and schematically illustrates fatigue crack propagation be-
havior in the plate bend specimen under two loading conditions: con-
stant deflection and constant total strain range. It was observed that,
under constant deflection loading, a gradual increase in the rate of
growth of the fatigue crack occurs as the crack length increases. However,
under constant total strain range loading, a constant rate of crack growth
is obtained, irrespective of crack length, for the range of crack lengths
relevant to this study. This characteristic of a singular dependence of
fatigue crack growth rate on total strain range in the plate bend specimen
has been verified time and again throughout the course of this work and
eliminates the need to evaluate a crack length parameter for each speci-
men. For this reason, all fatigue crack growth rate data reported herein
were taken under constant total strain range loading conditions. The
procedure has been to gradually reduce the deflection in order to main-
tain constant nominal net section strain in the presence of a growing

crack.
After this characteristic was established, the testing procedure was
standardized, and each specimen is cycled at a series of constant total
strain range intervals. A schematic illustration of this standardized pro-
cedure is shown in Fig. 6, which was drawn from the test results on a
FIG. 6—Illustration showing the experimental procedure of increasing the total

strain range at given intervals.
T-l steel specimen and is representative of this procedure. Cycling is

begun at the lowest desired strain range and is continued until a fatigue
crack is initiated, propagates away from the mechanical notch, and
establishes a constant rate of growth. Then the specimen is loaded to
the next higher strain range level desired and cycled until the new, higher
rate of growth is established. The number of cycles involved in each in-
terval depends upon the rate of growth; several thousand cycles may be
needed to establish a low crack growth rate, whereas only several
hundred cycles may be needed to collect sufficient data to establish a
FIG. 7—Composite photo of fatigue crack profiles (outlined in black) devel-

oped in plate bend specimens at various surface crack lengths. The specimens are
quenched and tempered steels cycled in air.
very high rate of crack growth. The reproducibility of the data obtained
by this procedure has been very good and has substantiated the validity
of this procedure.
The testing of a specimen is terminated for one of several reasons: on
occasion, one specimen gives all the desired data, also on occasion, a
fatigue crack of excessive length is formed (more than 1.5 in.), but most
commonly the material in the test section undergoes fatigue damage re-
sulting in visible microcracks. Such obvious fatigue damage requires
that the testing of the specimen be terminated. A new specimen is in-
stalled if additional data are needed.
Test data are based on observations of fatigue crack growth across the
surface of the test section. Extensions at both ends of the crack are
monitored, and the sum is reported. These observations are made by
an optical micrometer mounted above the specimen. No practical means
of detecting fatigue crack growth vertically down into the specimen is
presently available, and these data are not reported. Generally, the tip
of the fatigue crack is easily visible at the surface, and the crack grows
in a linear fashion across the minimum test section.
Upon completion of fatigue testing, the plate bend specimens are
pulled apart in tension to expose the fatigue surface. No plate bend
fatigue specimen has suffered catastrophic bending mode fracture in this
series of tests. Even specimens of high strength steels with surface flaws
longer than 1 in. at nearly yield strength stress level have remained in-
tact. Figure 7 is a composite of photos of several fatigue crack surfaces,
which illustrates the development of a fatigue crack front. The fatigue
surfaces on Fig. 7 have been outlined in black to distinguish their size
and shape. Basically, the fatigue crack grows radially from the notch
until its downward growth is inhibited by the low stress region near the
neutral bending axis. Beyond this point, the outward growth continues.
The shape of the crack front near the test surface remains nearly con-
stant throughout the various stages of growth. This is probably the reason
for the singular correlation between the fatigue crack growth rate and
total strain range, irrespective of flaw size.
The final piece of apparatus and procedure employed in this study is a
corrosion cell for conducting fatigue crack propagation studies under
aqueous environments. The corrosion cell is placed over the portion
of the test section containing the fatigue crack. It is made of molded
polyurethane which is relatively soft and flexes with the specimen. The
aqueous solution, either distilled water or 3.5 per cent salt water, is
constantly circulated through the corrosion cell from a reservoir by a
small electric pump. The solution undergoes filtering and periodic
monitoring of its pH value and salt content. A glass-covered opening at
the top of the corrosion cell permits optical observation of the fatigue
crack. Periodic removal of tarnish or rust from the test surface is re-
quired of most metals in order to maintain good visibility of the crack.
Materials
The materials considered in this paper include the following: normal-
ized, quenched and tempered, and maraged steels; high strength alu-
minum alloys; low strength unalloyed titanium and high strength titanium
TABLE 1—Chemical composition.
NRL Composition, weight %
Material Code
No. C Mn Si s P Ni Cr Mo V Al Cu Ti
Steels
A201B D42 0.18 0.65 0.21 0.024 0.010 0.010 0.08 0.03 0.02 0.05 0.18
A302B E85 0.20 1.42 0.19 0.022 0.018 0.12 0.07 0.52 0.02 0.18
T-l F42 0.14 0.72 0.22 0.007 0.010 0.78 0.51 0.45
HY-80 E84 0.15 0.20 0.23 0.007 0.007 2.18 1.29 0.30
HY-80 F66 0.20 0.32 0.33 0.012 0.006 3.04 1.58 0.78
HY-80 F60 0.20 0.35 0.32 0.010 0.007 3.20 1.69 0.70
HY-80 F65 0.19 0.32 0.33 0.013 0.006 3.28 1.69 0.70
HY-80 G13 0.14 0.37 0.27 0.009 0.003 3.03 1.72 0.52
HY-80 G14 0.16 0.37 0.29 0.010 0.005 3.03 1.69 0.55
5Ni-Cr-Mo-V H13 0.11 0.86 0.20 0.008 0.005 5.19 0.54 0.57 0.08 0.015
5Ni-Cr-Mo-V H98 0.11 0.76 0.25 0.004 0.007 4.90 0.60 0.49 0.06 0.021
12Ni maraging J8 0.005 0.04 0.05 0.007 0.005 11.8 5.16 3.30 0.13 0.24
12Ni maraging 37 0.005 0.03 0.06 0.007 0.003 12.1 4.83 3.10 0.22 0.24
D6AC G85 0.47 0.80 0.20 0.006 0.005 0.54 1.18 1.04 0.29
4335 G98 0.38 0.85 0.39 0.006 0.008 1.65 0.95 0.36 0.15
Al V Cb Ta Fe Mn Mg Si O2 N2 c Ti
Titanium alloys
Unalloyed Ti T16« 0.1 0.07 0.009 0.029 0.006
Ti-7Al-2Cb-lTa TA2« 6.9 2.5 1.1 0.13 0.063 0.006 0.01
Ti-6Al-4V T27« 5.2 3.9 0.07 Trace 0.0005 0.03 0.06 0.06 0.042 0.74
Ti-6Al-4V T5" 5.2 3.9 0.07 Trace 0.0005 0.033 0.06 0.06 0.042 0.74
Mn Cu Cr Mg Zn
Aluminum alloys
2024-T4 A2 5 0.6 4.5 1.5
7079-T6 A13 6 0.2 0.6 0.2 3.3 4.3
C Mn Fe S Cu Si Ni Ti Al
Nickel alloys
Monel 400 G24 0.17 0.94 1.64 0.008 32.24 0.11 64.84 0.009
Monel
Copyright K-500Int'l (all rights
by ASTM 0.16Mon Dec
G23reserved); 1.57 EST0.005
0.577 14:40:45 2015 30.21 0.14 63.91 0.51 2.89
Downloaded/printed
0 by
Producer's data.
University
6 of Washington
Metals Handbook,(University of Washington)
Am. Society Metals.pursuant to License Agreement. No further reproductions authorized.
TABLE 2—Mechanical properties.

Tension Test Properties
Drop-
NRL Orien- cv Weight 0.05%
Material Code tation Elong- Reduc- at 30 F, Tear Offset, u
No. ation tion of ft.lb Test at in./in."
in 2 Area, 30 F, ft.lb
in., % %
Steels
A201B D42 RW 48.0 67.9 40.0 54.6 21 3 500
A302B E85 RW 58.1 79.5 26.0 54.5 5 800
T-l F42 RW 107.0 118.0 20.0 64.4 33 8 400
H Y-80 E84 WR 85.8 102.7 22.8 67.6 93 5000
H Y-80 E84 RW 90.5 105.7 24.0 70.8 115 6200 7 500
HY-80 F66 WR 116.0 134.4 20.8 68.4 42 2000
H Y-80 F66 RW 117.8 135.3 16.5 50.6 86 6250 9 700
HY-80 F60 WR 131.8 151.6 14.5 45.8 30 1250
HY-80 F60 RW 132.0 151.2 19.0 66.8 82 5750 10 200
HY-80 F65 WR 158.0 175.6 12.5 41.3 23 1000
HY-80 F65 RW 156.6 175.2 17.0 54.6 48 3250 11 700
HY-80 G13 WR 133.0 142.2 17.5 57.1 62 2789 10 000
HY-80 G14 WR 151.4 165.0 17.0 55.2 42 2364 11 500
5Ni-Cr-Mo-V H98 RW 133.2 140.5 20.0 65.1 112 >5000 11 500
5Ni-Cr-Mo-V H13 WR 145.0 153.5 18.7 63.5 80 4573
5Ni-Cr-Mo-V H13 RW 143.8 152.5 20.3 63.7 12 000
12Ni maraging J8 WR 162.6 170.6 15.8 64.1 84 4478 13 800
12Ni maraging J7 WR 181.5 188.5 14.5 62.3 66 3843 14 800
D6AC G85 RW 211.7 228.7 12.0 44.6 19 870 16 000
4335 G98 RW 215.5 244.4 10.5 41.3 19 870 16 000
Titanium alloys
Unalloyed ti-
tanium T16 RW 34.7 46.7 40.5 84.5 163 9625 5 000
Ti-7Al-2Cb-lTa TA2 RW 107.0 120.0 13.0 28.0 39 2675 16 000
Ti-6Al-4V T27 RW 115.8 128.4 12.2 26.4 1228
Ti-6AMV T5 RW 125.5 131.0 12.5 44.5 29 2075 15 000
Aluminum alloys
2024-T4 A2 WR 47.7 71.9 10 367
2024-T4 A2 RW 48.0 71.7 7 668 13 000
7079-T6 A13 RW 77.1 86.0 11.7 23.2 5 339 18 000
Nickel alloy
Monel 400 G24 RW 35.0 86.0 46.0 39.0 4 500
Monel K-500 G23 RW 103.0 154.0 25.0 31.0 9 500
" Full-reverse plate bend proportional limit.
alloys; and low and high strength nickel-copper alloys. These include
many currently popular structural alloys as well as a number of new
alloys under evaluation for future application. Chemical compositions
and mechanical properties of these materials are displayed in Tables 1
and 2, respectively.
In nearly all cases fatigue specimens were machined from l-in.-thick
rolled plate stock. Exceptions are the A201B, A302B, and T-l steel
specimens, which were cut from the 2-in. thick shells of full-size pres-
sure vessels which had previously undergone fatigue testing by the

Pressure Vessel Research Committee [1]. Specimen orientations are
identified with respect to the principal or final rolling direction of the
material. The symbols RW or j_ indicate that the direction of measured
fatigue crack propagation is perpendicular to the rolling direction (some-
times referred to as the "strong" orientation); whereas the symbols WR
or P indicate fatigue crack propagation parallel to the rolling direction
("weak" orientation). The symbols RW and WR are taken from standard
ASTM designations as defined in Ref 5.
Many of these materials have also been the subject of extensive re-
search to determine their fracture toughness, stress-corrosion cracking
susceptibility, weldability, and heat treatability. Detailed information on
these properties can be found in Refs 6 through 14.
Results
Air Environment
The first step in developing an analysis of the data from this test pro-
cedure was the determination that constant total strain range cycling
resulted in a constant rate of crack growth. The next step was to estab-
lish the form of the relationship between crack growth rate and total
strain range, for the strain range values which are pertinent to low cycle
fatigue in large structures.
Referring back to Fig. 6, which is a plot of total crack length versus
cycles of repeated load for a series of constant total strain range inter-
vals, it follows that the slopes of these straight-line segments are directly
proportional to the fatigue crack growth rate at each corresponding
level of total strain range. If these crack growth rate values are plotted
versus total strain range using log scales, the data form a straight-line
plot. The equation between fatigue crack growth rate and total strain
range which is established by this plot has the form
where:
la = total crack length,
N = cycle of repeated load,
= fatigue crack growth rate,
eT = total strain range,
m = numerical exponent, and
c = numerical constant.
Power-law relationships of this general form have been empirically
developed for all materials tested to date. These power-law curves,
generated under full-reverse (tension-to-compression) strain cycling, in

a normal room-air environment, provide a baseline for quantitatively
determining the low cycle fatigue crack propagation characteristics of
structural materials. The baseline obtained for an air environment is
used: (a) for establishing the effects of other, more hostile environments
FIG. 8—Log-log plot of fatigue crack growth rate versus total strain range
showing the consistency of a plot for data from different 5Ni-Cr-Mo-V steels in a
room-temperature air environment.
and mean strain on crack growth rates, (b) for comparing the fatigue
performance of different alloys, and (c) for estimating structural fatigue
life. This method for establishing the fatigue crack propagation character-
istics for a wide spectrum of materials is illustrated by the following
summary of the data collected in this series of tests.
1 08 FATIGUE CRACK PROPAGATION
FIG. 9—Summary of relationships between fatigue crack growth rate and total
strain range for a spectrum of steels in air, which relationships provide the base-
lines for the fatigue crack propagation characteristics,
Figure 8 is a log-log plot of fatigue crack growth rate versus total

strain range data for two closely similar 5Ni-Cr-Mo-V steels. Figure 8
illustrates the consistency of the data for denning baseline fatigue propa-
gation characteristics of a material. A summary illustration of such curves
for all the steels reported in this paper is shown in Fig. 9, and a sum-
mary illustration for nonferrous alloys is shown in Fig. 10.
These illustrations reveal several important fatigue crack propagation
characteristics of materials. First, it can be noted that there is a general
tendency for the slopes of these curves to increase with increasing yield
strength level, in both ferrous and nonferrous alloys. These slopes are
FIG. 10—Summary of relationships between fatigue crack growth rate and

total strain range for a spectrum of nonferrous alloys in air.
indicative of the fatigue crack growth rate sensitivity of a material to

changes in the level of total strain range. The steep slopes exhibited by
high strength materials shows that they can very easily undergo a transi-
tion from a low to a high fatigue crack growth rate with a relatively
small increase in total strain range. By way of contrast, the lower slopes
exhibited by low strength materials indicates a more gradual reduction
in fatigue crack propagation resistance with increasing total strain range
levels.
showing data for 5Ni-Cr-Mo-V steels in a 3.5 per cent salt water environment.
Note the locus of air data indicated by the dashed line.
A second characteristic is that several high strength materials which

possess low fracture toughness and are highly sensitive to environment,
notably D6AC and 4335 steels and 7079-T6 aluminum alloy, exhibit a
distinct transition point at which the slope increases to a much higher
value, as shown on Figs. 9 and 10. Such behavior is also noted in Ti-7Al-
2Cb-lTa alloy in wet environments [75]. This transition is associated
with a change in the fracture mode from the ductile mode of fatigue frac-
ture to a brittle fracture mode.
A third characteristic of these curves is illustrated by the data in Fig.
showing data for 4335 steel in a 3.5 per cent salt water environment. Note the
locus of air data indicated by the dashed line.
8 for 5Ni-Cr-Mo-V steels. It can be seen that the data follow the same
straight-line locus for strain range values above the plate bend propor-
tional limit (denoted by the arrows) into the region of nominal plastic
strains. This behavior has been observed in all tests which have included
plastic strain cycling and broadens the application of this test method to
structural materials of relatively low yield strength levels.
A fourth observation of considerable practical significance is that for
a given alloy, the relationship between fatigue crack growth rate and
showing data for Ti-6Al-4V and Ti-7Al-2Cb-lTa alloys in a 3.5 per cent salt water
environment. Note the locus of air data indicated by the dashed line.
total strain range can be identical for various yield strength levels and
rolling direction orientations. This was first observed in studies on
specially processed HY-80 steels prepared with various degrees of cross-
rolling, ranging from straight-rolled to 1:1 crossrolled, and heat treated
to yield strength levels from 90 to 160 ksi [6]. Similar results have been
obtained in other steels, both quenched and tempered and maraging
[8,11,13]. Further implications of this phenomenon will be discussed in
a later section of this paper. For numerous steels, however, alloy modifi-
cations, yield strength variations through heat treatment, or anisotropy

effects due to rolling do not necessarily alter the intrinsic resistance to
low cycle fatigue crack propagation when comparisons are made using
total strain range as the parameter for load intensity.
Wet Environments
The influence of environment on the structural performance of metals

can be of the utmost importance, especially where sharp flaws and re-
peated loadings are involved. For this reason, nearly all the materials
discussed in this report have undergone testing in wet environments as
well as in air.
The wet fatigue data presented in this paper were taken from tests
where 3.5 per cent salt water solution was the corrosive agent. Dis-
tilled water and a simulated boiler water have also been used in specific
tests. However, the 3.5 per cent salt water solution has been found to be
the severest of these agents and represents the worst case. Generally,
distilled water has nearly the same effect as 3.5 per cent salt water at
high strain range values, above 50 per cent of the plate bend proportional
limit; but it tends to be somewhat less severe than salt water at lower
strain range values.
The effect of a wet environment on fatigue crack growth rates can be
determined by adding a corrosion cell to a fresh specimen of a material
and duplicating the program of total strain range values employed in the
air test. With this procedure the effect of environment for the specific
conditions of the two otherwise identical tests in air and in water can
be determined.
For most structural metals the introduction of a wet environment
causes an increase hi the rate of fatigue crack growth at any given level
of total strain range. This environmentally induced increase is indicated
by a vertical displacement of the curve of wet environment fatigue data
above the curve of air environment fatigue data on the log-log plot of
fatigue crack growth rate versus total strain range. Figures 11, 12, and
13 illustrate several characteristic types of wet fatigue curves for high
strength steels and titanium alloys in 3.5 per cent salt water solution.
Figure 11 is a plot of salt water fatigue data for 5Ni-Cr-Mo-V steels,
with the relationship for air indicated by the dashed line for comparison.
These steels can be categorized as possessing good fatigue crack propaga-
tion resistance in a 3.5 per cent salt water environment. The maximum
increase in crack growth rates due to environment in these steels is well
under an order of magnitude. Experience suggests that an order of
magnitude increase is a good benchmark for a first-order categorization
of wet fatigue performance.
An important characteristic of these 5Ni-Cr-Mo-V steels is that the
salt water curve tends to deflect toward the air curve at higher strain
range levels in the region of small plastic strains. This behavior is in sharp
contrast to materials which are highly sensitive to wet environments.
Other steels which possess good wet fatigue characteristics include
steels of HY-80 chemistry and 12 per cent nickel maraging steels. It is
characteristic of these steels that heat treatments which increase the
yield strength also tend to improve the wet fatigue crack propagation
resistance. This is attributed to reduced plastic deformation at the crack
tip due to higher yield strength.
In contrast to the salt water fatigue crack propagation resistance of
HY-80, 5Ni-Cr-Mo-V, and 12 per cent nickel maraging steels which has
been classified as good, some higher strength steels are known to be very
sensitive to wet environments and therefore possess very poor salt water
fatigue crack propagation resistance [10,16]. An example of this category
is shown in Fig. 12, which is a log-log plot of fatigue crack growth rate
versus total strain range data for 4335 steel in 3.5 per cent salt water
solution. The relationship for air is shown by dashed lines for compari-
son. This 4335 steel, heat treated to a yield strength of 215 ksi, exhibited
a gross increase in fatigue crack growth rates at each level of total strain
range examined. Furthermore, the levels of total strain range associated
with rapid crack growth rates are very low and are equivalent to only
small fractions of the elastic strength of this 4335 steel. Unlike the curves
in Fig. 11 for 5Ni-Cr-Mo-V steels, the salt water curve in Fig. 12 di-
verges from the air curve at higher total strain range levels and repre-
sents an example of very poor salt water fatigue crack propagation resis-
tance.
A third example of wet fatigue behavior is shown in Fig. 13, which
shows salt water fatigue crack propagation data for Ti-6Al-4V and Ti-
7Al-2Cb-lTa alloys. These data have been the subject of a previous
study [75] and are an outstanding example of both good and bad salt
water fatigue performance in structural metals. On one hand, the Ti-
6A1-4V alloys appear to be completely immune to the salt water en-
vironment. The Ti-6Al-4V salt water data show no displacement from
the air curve, indicated by the dashed line. On the other hand, the Ti-
7Al-2Cb-lTa salt water data show a relatively small displacement from
the air data plot at lower strain range levels, leading to a dramatic
breakaway point and sudden increase in crack growth rates at a strain
range value equal to approximately 60 per cent of the plate bend propor-
tional limit. This material has also been found to be sensitive to stress-
corrosion cracking in the presence of a sharp flaw under static loading
[77] and represents another example of very poor salt water fatigue
crack propagation resistance.
The examples that have been presented are only an introduction to
the wet fatigue crack propagation behavior of structural metals. Further
detailed information on the wet fatigue characteristics of the materials
FIG. 14—Relationships between fatigue crack growth rate (log scale) and
mean strain (linear scale) obtained for 5Ni-Cr-Mo-V steel in air. Solid lines indicate
constant values of total strain range; dashed lines indicate constant values of maxi-
mum strain amplitude.
listed in Tables 1 and 2 is contained in Refs 8 through 14. The examples

presented do illustrate, however, that the plate bend fatigue test is a
sensitive tool of determining the sensitivity of structural metals to aqueous
environments when the propagation of fatigue cracks is monitored.
FIG. 15—Relationships between fatigue crack growth rate (log scale) and mean
strain (linear scale) obtained for 7079-T6 aluminum alloy in air. Solid lines indicate
constant values of total strain range; dashed lines indicate constant values of maxi-
mum strain amplitude.
CROOKER AND LANGE ON WELDED STRUCTURES 1 17
Mean Strain
Previous sections of this paper have dealt with the role of cyclic strain
as a parameter in determining the rate of fatigue crack growth in struc-
tural materials. However, cyclic strain alone is not sufficient to describe
the loading which materials undergo when placed in actual service con-
ditions. Static preload or residual stress conditions or both exist in most
structures, creating a mean strain upon which cyclic strain is then super-
imposed. This section describes the results of experiments in which
various mean strains, both tensile and compressive, were introduced
under cyclic loading conditions to determine their influence on the rate
of fatigue crack growth.
The loading cycle employed in these tests differs from that described
in previous section in that a full-reverse (tension-to-compression) cycle
is not used throughout. Instead, the loading cycle is unbalanced so as to
provide a known mean strain for study. Each specimen was tested at a
specific constant value of cyclic strain range, to which various known
mean strains were added.
Care was taken to keep maximum strain amplitudes within the elastic
limit of the material. This procedure minimized hysteresis and permitted
relaxation of the bend specimen to zero strain at zero load, and thus
enabled accurate measurement of mean strains.
Experimental data are shown plotted in Fig. 14 for a 5Ni-Cr-Mo-V
steel and in Fig. 15 for 7079-T6 aluminum alloy. These figures are
semilog plots of fatigue crack growth rate (log scale) versus mean strain
(linear scale). The solid lines represent the response of the fatigue crack
growth rate to the application of mean strain for a constant value of
cyclic strain. The dashed lines represent contours of constant maximum
strain amplitude.
Qualitatively, it can be seen that tensile mean strain accelerates the
growth of fatigue cracks and compressive mean strain retards the growth
of fatigue cracks. The magnitude of the effect of mean strain on fatigue
crack growth rate depends upon the level of cyclic strain and the low
cycle fatigue crack propagation characteristics of the material. These
tests indicate that the sensitivity of the fatigue crack propagation char-
acteristics of metals to mean strain appears to correspond with their
sensitivity to cyclic strain. For example, fatigue crack propagation in
7079-T6 aluminum alloy is a good deal more sensitive to cyclic strain
than in 5Ni-Cr-Mo-V steels, as indicated by the slopes of their curves in
Figs. 9 and 10, and this same relative sensitivity is also exhibited in the
mean strain data for these two materials in Figs. 14 and 15.
Quantitatively, the results of these preliminary studies of the effects
of mean strain show that accelerations in fatigue crack growth under
elastic loading conditions can approach an order of magnitude in 5Ni-
Cr-Mo-V steel and two orders of magnitude in 7079-T6 aluminum alloy,
FIG. 16—Spectrum, view of fatigue propagation characteristics of steels on a

normalized load intensity basis.
from the combined effect of a relatively small cyclic strain range and a
relatively large static tensile mean strain. In terms of structural life this
suggests that mean strains in high strength materials should not be
ignored at cyclic elastic strain levels normally considered safe from
danger of low cyclic fatigue failure. It is at these lower cyclic strain
levels that tensile mean strain can reduce fatigue life by an order of
magnitude, and in some cases two orders of magnitude, below estimates
of fatigue life based only on cyclic strain and zero mean strain. At higher
cyclic strain levels, with tensile strain amplitude approaching the plastic
strain region, the maximum reduction in fatigue life due to tensile mean
FIG. 17—Comparison of fatigue propagation characteristics of a spectrum of

nonferrous alloys on a normalized load intensity basis.
strain appears to be approximately a factor of three for both 5Ni-Cr-

Mo-V steel and 7079-T6 aluminum alloy. Thus, there appears to be a
limiting effect of mean strain on fatigue crack growth rate when plastic
strains are involved.
Interpretation
Comparative Material Performance

A primary application of the test methods and data described in this
paper is to provide criteria for comparing the low cycle fatigue crack
propagation resistance of structural materials. Correlations between the

rate of fatigue crack growth and total strain range, such as those shown
in Figs. 9 and 10, can be a useful tool for this purpose. However, com-
parisons among various materials based on strain can be misleading un-
less differences in yield strength and elastic modulus are taken into con-
sideration.
A more definitive criterion for comparing material performance is
obtained by normalizing load intensity with respect to the proportional
limit strain range. A normalized load intensity factor is thus obtained by
converting strain units to the ratio of total strain range to proportional
limit strain range. Figure 16 shows a summary log-log plot of fatigue
crack growth rate versus the ratio of total strain range to proportional
limit strain range for steels, and Fig. 17 is a similar graphic summary for
nonferrous alloys.
In terms of application to large welded structures, these graphs re-
veal several important materials performance characteristics. With
strain expressed in relation to proportional limit strain range, it becomes
apparent that lower strength materials possess certain advantages over
higher strength materials in then" ability to resist the growth of low
cycle fatigue cracks. This is indicated by the generally lower slopes of
the curves for low strength materials and by the locus of these curves.
It should be noted that in the region of plastic strains, low strength
materials develop relatively low rates of fatigue crack growth. This means
that lower strength materials possess a greater tolerance for plastic strains
which develop due to overload or due to regions of strain concentration.
Since strain concentrations commonly occur around welds where crack-
like defects may be present, the greater resistance of low strength ma-
terials to the growth of low cycle fatigue cracks is an important factor
to consider when designing large, complex welded structures which
cannot be rigorously stress analyzed.
On the other hand, benefits are to be had from higher yield strength
materials, if the added elastic strength is used to good advantage to re-
duce or eliminate plastic strains at points of discontinuity. However, the
curves in Figs. 16 and 17 also point out the hazards of employing higher
strength materials where design codes are based on a fixed percentage
of elastic strength. As mentioned in a previous section, it is not uncom-
mon for a family of materials (for example, martensitic steels) posses-
sing widely different yield strength levels to have nearly the same relation-
ship between fatigue crack growth rate and total strain range. As a
result, the higher strength materials within the given family will propagate
fatigue cracks at a faster rate when cyclic strain levels are based on a
fixed percentage of yield strength. It is important to emphasize that
fatigue crack growth rates increase exponentially with total strain range.
For example, in HY-80 steels fatigue crack propagation varies with the
forth power of cyclic strain, which means that doubling the level of
total strain range results in a crack growth rate 16 times greater.
It is apparent from the relationships displayed on Figs. 16 and 17
that the application of higher strength materials to large welded struc-
tures can seriously aggravate the probability of failure from low cycle
fatigue unless comprehensive engineering considerations involving ma-
terials characteristics, design, and fabrication are fully recognized. Safe
application of new materials in high performance structures can only be
attained with refined design and high quality fabrication.
Index of Structural Fatigue Life

The ultimate objective of this low cycle fatigue crack propagation
program is to establish structural design criteria for fatigue life. As a
first step in this direction, fatigue crack growth rate data for high fracture
toughness materials have been given an arbitrary relation to structural
life in terms of the Index of Structural Fatigue Life.
The Index of Structural Fatigue Life is defined as the number of
cycles of repeated load required to propagate a small subcritical flaw
to 1-in. length at a given intensity of loading, as measured by the total
strain range, in the plate bend fatigue test. This correlation between
fatigue crack growth rates and structural fatigue life is indexed along the
right-hand ordinate of each fatigue data illustration, Figs. 8 through 17.
We will confine our evaluation here to three steels: HY-80, 5Ni-Cr-Mo-
V, and 12 per cent nickel maraging. These steels follow closely similar
fourth power relationships between fatigue crack growth rate and total
strain range, Figs. 8 and 9. Therefore, in effect, the Index of Structural
Fatigue Life states that for these steels in an air environment fatigue life
is inversely proportional to the fourth power of total strain range.
The Index of Structural Fatigue Life definition of fatigue failure is in
contrast to other commonly employed test procedures which either
designate the initiation of a small fatigue crack of arbitrary size as the
failure criterion [4] or which inseparably combine crack initiation plus
crack propagation to terminal fracture as the failure criterion.
Conventional log-log graphs of total strain range versus the number of
cycles to failure for each of the three steels are shown in Figs. 18, 19,
and 20. Each graph contains four curves representing failure by initia-
tion and failure by propagation, each in air and 3.5 per cent salt water
environments. The relationships for fatigue crack initiation shown on
these plots were obtained from recent studies conducted by the United
States Steel Applied Research Laboratory, which employ unnotched
plate bend specimens in full-reverse strain cycling [18]. Initiation life is
defined as the number of cycles to produce a %6-in. fatigue crack.
The life values for the crack initiation and the crack propagation fail-
ure criteria are not radically different, as indicated by the closely grouped
FIG. 18—Comparison of relationships between total strain range and structural

life from crack initiation and crack propagation criteria for HY-80 steel in air and
3.5 per cent salt water. The crack initiation data were obtained by the US. Steel
Applied Research Laboratory [18].

life for 5Ni-Cr-Mo-V steels. The curves represent failure by crack initiation [181
and failure by crack propagation in both air and 3.5 per cent salt water environ-
ments.
curves in Figs. 18, 19, and 20. This is not surprising, since the initiation
and growth of small fatigue cracks have been quite commonly observed
during crack propagation tests. Such occurrences in 5Ni-Cr-Mo-V steel
are described in Ref 11.
The failure curves for the two criteria in air follow closely similar
paths. Where significant differences do appear, the initiation criterion
curve is generally a more conservative estimate of fatigue life. How-
ever, in salt water, the propagation curve lies below rather than above
the initiation curve, indicating that failure by fatigue crack propagation

life for 12 per cent nickel maraging steels. The curves represent failure by crack
initiation [18] and failure by crack propagation in both air and 3.5 per cent salt
•water environments.
is the more critical criterion in salt water. At elastic strain range levels,
the two salt water curves differ by at least a factor of two.
The salt water environment propagation data curves tend to deflect
toward the air environment data curves at higher strain levels above the
respective elastic limits. This phenomena should be carefully examined
where structural life is involved. These salt water fatigue crack propaga-
tion curves were derived from laboratory tests conducted at 5 cpm with
no hold period at peak loads. The trend of the salt water environment
propagation data toward air environment data values may very well be
altered by the introduction of slower rates of cycling and a hold dura-
tion at maximum load. In addition, the effect of mean strain on fatigue
crack propagation in salt water is likely to deleteriously alter the com-

parisons in the data shown. Finally, for materials that are sensitive to
stress-corrosion cracking, the salt water crack propagation curve can be
expected to differ radically from the salt water initiation curve at the
critical higher strain range levels. These aspects require and deserve
further research.
It is recognized that the Index of Structural Fatigue Life is obtained
from an arbitrary criterion for failure. This index does, however, corre-
spond to the life and strain values obtained in the study of low cycle
fatigue of full size pressure vessels under the direction of the Pressure
Vessel Research Committee of the Welding Research Council [1]. Po-
tentially, fatigue crack propagation criteria hold promise for more precise
predictions of structural life. To develop more precise methods for
predicting structural life, more knowledge concerning the strain intensity
factors for structural configurations (design and fabrication) and the
effects of flaw size on strain range is required.
1. The plate bend fatigue test can provide a relatively simple and in-
expensive method of evaluating the low cycle fatigue crack propagation
resistance of the full spectrum of structural materials, from low to high
yield strength levels.
2. Power-law relationships between fatigue crack growth rate and
total strain range have been found to exist for all materials tested in the
plate bend configuration. These relationships provide a baseline for
comparing the low cycle fatigue crack propagation characteristics of
materials and for evaluating the effects of environment and mean strain
on fatigue crack growth rates.
3. The response of fatigue crack propagation to environment or mean
strain varies greatly depending upon the material, the environment, and
the degree of loading. However, increases in fatigue crack growth rates
of more than an order of magnitude due to environment or mean strain
are not uncommon in high strength structural materials, and these as-
pects demand careful consideration.
4. A comparative criterion for evaluating the low cycle fatigue crack
propagation resistance of a wide spectrum of materials can be obtained
from power-law relationships between fatigue crack growth rates and
the ratio of total strain range to proportional limit strain range. The in-
troduction of this ratio provides a common denominator for comparing
materials of various yield strength levels and various elastic moduli.
5. Fatigue crack growth rate data can be used to develop a first-order
approximation of structural fatigue life in high fracture toughness ma-
terials referred to as the Index of Structural Fatigue Life.
Acknowledgments
The authors wish to acknowledge the considerable efforts of R. E.
Morey, S. J. McKaye, and R. J. Hicks who participated in the develop-
ment of these test methods and carried out the experimental work on
which this paper is based.
References
[1] L. F. Kooistra, E. A. Lange, and A. G. Pickett, "Full-Size Pressure Vessel
Testing and Its Application to Design," Transactions, ASME Journal of
Engineering for Power, Vol 86, Series A, No. 4, October, 1964.
[2] J. E. Srawley and J. B. Esgar, "Investigation of Hydrotest Failure of Thiokol
Chemical Corporation 260-Inch-Diameter SL-1 Motor Case," NASA TM
X-1194, January, 1966.
[3] J. H. Gross, S. Tsang, and R. D. Stout, "Factors Affecting Resistance of Pres-
sure Vessel Steel to Repeated Overloading," Welding Journal Research Sup-
plement, Vol 32, No. 1, January, 1953.
[4] M. R. Gross, "Low-Cycle Fatigue of Materials for Submarine Construction,"
Naval Engineers Journal, Vol 75, No. 4, October, 1963.
[5] The ASTM Committee on Fracture Testing of High-Strength Metallic Mate-
rials, "The Slow Growth and Rapid Propagation of Cracks," (second report),
Materials Research & Standards, Vol 1, No. 5, May, 1961.
[6] P. P. Puzak, K. B. Lloyd, R. J. Goode, R. W. Huber, D. G. Howe, T. W.
Crooker, E. A. Lange, and W. S. Pellini, "Metallurgical Characteristics of
High Strength Structural Materials (Third Quarterly Report)," NRL Report
6086, January, 1964
[7] R. J. Good, R. W. Huber, D. G. Howe, R. W. Judy, Jr., T. W. Crooker,
R. E. Morey, E. A. Lange, P. P. Puzak, and K. B. Lloyd, "Metallurgical Char-
acteristics of High Strength Structural Materials (Fourth Quarterly Report),"
NRL Report 6137, June, 1964.
[8] T. W. Crooker, R. E. Morey, E. A. Lange, R. W. Judy, Jr., C. N. Freed,
R. J. Goode, P. P. Puzak, K. B. Lloyd, R. W. Huber, D. G. Howe, and W. S.
Pellini, "Metallurgical Characteristics of High Strength Structural Materials
(Fifth Quarterly Report)," NRL Report 6196, September, 1964.
[9] W. S. Pellini, R. J. Goode, R. W. Huber, D. G. Howe, R. W. Judy, Jr., P. P.
Puzak, K. B. Lloyd, E. A. Lange, E. A. DeFelice, T. W. Crooker, R. E. Morey,
E. J. Chapin, and L. J. McGeady, "Metallurgical Characteristics of High
Strength Structural Materials (Sixth Quarterly Report)," NRL Report 6258,
December, 1964.
[10] R. J. Goode, D. G. Howe, R. W. Huber, P. P. Puzak, K. B. Lloyd, T. W.
Crooker, R. E. Morey, E. A. Lange, R. W. Judy, Jr., and C. N. Freed,
"Metallurgical Characteristics of High Strength Structural Materials (Seventh
Quarterly Report)," NRL Report 6327, May, 1965.
[11] P. P. Puzak, K. B. Lloyd, R. J. Goode, R. W. Huber, D. G. Howe, R. W.
Judy, Jr., T. W. Crooker, R. E. Morey, E. A. Lange, and C. N. Freed, "Metal-
lurgical Characteristics of High Strength Structural Materials (Eighth Quarterly
Report)," NRL Report 6364, August, 1965.
[12] R. J. Goode, R. W. Huber, D. G. Howe, R. W. Judy, Jr., P. P. Puzak, K. B.
Lloyd, T. W. Crooker, R. E. Morey, E. A. Lange, and C. N. Freed, "Metal-
lurgical Characteristics of High Strength Structural Materials (Ninth Quarterly
Report)," NRL Report 6405, November, 1965.
[13] R. J. Goode, R. W. Huber, R. W. Judy, Jr., D. G. Howe, P. P. Puzak, K. B.
Lloyd, T. W. Crooker, R. E. Morey, E. A. Lange, and C. N. Freed, "Metal-
lurgical Characteristics of High Strength Structural Materials (Tenth Quar-
terly Report)," NRL Report 6454, April, 1966.
[14] T. W. Crocker, R. E. Morey, and E. A. Lange, "Low Cycle Fatigue Crack

Propagation Characteristics of Monel 400 and Monel K-500 Alloys," NRL
Report 6218, March 10, 1965.
[75] R. W. Judy, Jr., T. W. Crocker, R. E. Morey, E. A. Lange, and R. J. Goode,
"Low Cycle Fatigue Crack Propagation and Fractographic Investigation of
Ti-7Al-2Cb-lTa and Ti-6Al-4V in Air and in Aqueous Environments,"
Transactions, Am. Society Metals, June, 1966.
[16] E. P. Dahlberg, "Fatigue-Crack Propagation in High-Strength 4340 Steel in
Humid Air," Transactions, Am. Society Metals, March, 1965.
[17] B. F. Brown, "A New Stress-Corrosion Cracking Test for High Strength Al-
loys," Materials Research & Standards, Vol 6, No. 3, March, 1966.
[18] T. L. Boblenz and S. T. Rolfe, "Corrosion-Fatigue Characteristics of HY-80,
5Ni-Cr-Mo-V, 12Ni-5Cr-3Mo, and 18Ni-8Co-3Mo Steels," U.S. Steel Applied
Research Laboratory, Report No. 39.018-002(37), (S-23308-2), Jan. 1, 1966.
DISCUSSION
G. H. Jacoby1 (written discussion)—The tests of the authors were

performed with parent material. In welded structures, however, fatigue
cracks develop mainly at the weld bead and grow into the heat affected
zone or even into the filler material. These areas, however, usually ex-
hibit quite another metallurgical structure than the parent material. To
our experience the crack propagation rate in the heat affected zone and
even more in the filler material may be considerably higher than in the
virgin material. Do the authors think that their values for crack propaga-
tion rate will cover this behavior (for instance by the adaption of a
correction factor) and can be used in practical applications?
T. W. Crooker and E. A. Lange (authors)—Dr. Jacoby has raised a
valid and important consideration. By way of reply, the authors have
confined their studies to plate materials for several reasons. First, the
work presented in this paper represents an initial research program
which involved both the development and application of test methods.
The inherent simplicity of plate materials rather than weldments was
judged to be an important factor during this phase of the program.
Second, although welded joints are prime sources of defects in service,
the propagation of defects in fatigue is largely controlled by the stress
pattern and is not always limited to weld metal or heat affected zones.
Differences between the fatigue crack propagation characteristics of
plate materials and weldments are dependent not only upon metallurgical
structure, but also upon geometry and residual stress. Until more is
learned of fatigue crack propagation in weldments, the authors would
1
German Research Association for Aerospace Sciences (OWL), at present
Columbia University, New York, N. Y.
DISCUSSION ON WELDED STRUCTURES 127
hesitate to recommend the use of prime plate data for weldment design
purposes. However, the authors are hopeful that the test methods and
analysis described in this paper can be applied to weldments, and that
the prime plate data presented will prove useful in evaluating the results
of weldment studies.
Microstructural Aspects of Fatigue Crack
Growth
Campbell Laird1
The Influence of Metallurgical Structure

on the Mechanisms of Fatigue Crack
Propagation
REFERENCE: Campbell Laird, "The Influence of Metallurgical Struc-

ture on the Mechanisms of Fatigue Crack Propagation," Fatigue Crack
ABSTRACT: Studies of fatigue fractures have indicated that crack propa-
gation takes place by two stages. The first stage propagates at 45 deg with
respect to the stress axis, while the second does so at 90 deg. The evidence
for crack growth is reviewed, and it is concluded that the Stage II mode
occurs by plastic blunting of the crack tip during the tensile part of a fa-
tigue cycle followed by resharpening of the crack in the compression part.
The mechanism of Stage I growth occurs on too small a scale to be studied
directly but is deduced to be essentially the same as Stage II. The softness
of fatigue slip bands with respect to the bulk of the material, however,
provides an easy path for crack propagation in Stage I and thus causes prop-
agation at about 45 deg to the stress axis. On the other hand, in Stage II,
the larger stress concentration associated with the longer crack dominates
the structure at the crack tip and orients the path of crack propagation at
90 deg to the stress axis. The morphological details of fracture surfaces are
considered in the light of these mechanisms.
The role of microstructure in influencing the mechanisms is treated
from two aspects. First, it is shown that the various kinds of hardened
structures merely act to control the degree of deformation at a crack tip in
relaxation, and thus the kinetics of crack propagation. Otherwise, micro-
structures hardened to extremes by cold-working or by precipitation can
be the exception to the generalization, acting in conjunction with other test-
ing variables to change the mode of propagation so as to introduce cleav-
age or quasi-cleavage fracture to the crack propagation process.
KEY WORDS: microstructure, fatigue (materials), crack propagation, plas-
tic flow, slip bands, fracture
The theme of this paper is that there is one general mechanism of fa-
tigue crack growth in ductile materials, and, consequently, the micro-
structure of the material undergoing fatigue acts only to alter the kinetics
of crack propagation; it does not change the nature of the process. Direct
observation of the changes occurring at crack tips under cyclic loads indi-
1
Scientific Laboratory, Ford Motor Co., Dearborn, Mich.
131
cates that a crack propagates by plastic blunting of the crack tip during
the tensile part of the fatigue cycle followed by resharpening of the crack
in the compression part. This process will be called hereafter, for brevity,
the plastic blunting process. The paper is thus divided into two parts. The
first part is a review of the evidence both for the plastic blunting process
and for its generality. In the second part, the influence of microstructure
is considered from two aspects. The more important aspect is concerned
with how inhomogeneities in the microstructure and the work-hardening
capacity of the material control the degree of localized plastic deformation
of a crack tip and thus the kinetics of crack propagation. However, in
FIG. 1—Schematic representation of the two stages of crack propagation.
some microstructures hardened to extremes either by cold work or by

precipitation, it is possible that the nature of the microstructure induces
an element of cleavage fracture into the crack propagation process. This
rare exception to the generality is treated as a special case of the plastic
blunting process.
Mechanisms of Crack Propagation
We know now that crack propagation in fatigue takes place by two
stages under a wide range of stressing and environmental conditions
[7-5].2 The first stage, illustrated in Fig. 1, is characterized by propagation
of the crack on a plane oriented at approximately 45 deg to the stress
2
paper.
LAIRD ON INFLUENCE OF METALLURGICAL STRUCTURE 133
axis and by crystallographic fracture facets changing direction slightly3

with orientation at grain boundaries. Subsequently, the propagation
enters Stage II, also illustrated in Fig. 1, where the plane of crack propa-
gation is now at 90 deg to the stress axis and the fracture surface is
covered by striations running parallel to the crack propagation front.
Striations on fracture surfaces look like tidemarks. Failure in high strain
fatigue takes place predominantly by Stage II crack propagation, whereas,
at lower strains at least 90 per cent of the fatigue lives of unnotched speci-
mens is spent in Stage I [2]. Even at low strains, however, most of the
fracture surface of a broken specimen is covered with striations; this indi-
cates that Stage II also plays an important role in low strain fatigue, since
this process actually fractures most of the cross section of the specimen.
Since crack propagation rates in Stage II growth can reach values of
microns per cycle, the phenomena associated with the growth mechanism
are fairly large and relatively easy to observe. Most of the evidence for
growth, therefore, concerns Stage II which has been proved to operate in
the plastic blunting process; the discussion in this section of the paper will
thus be centered around Stage II growth. On the other hand, rates of crack
propagation in Stage I are of the order of angstroms per cycle. This stage
is therefore extremely difficult to study, and my conclusion that it also
operates by the plastic relaxation process is deduced from our knowledge
of Stage II growth. The final part of this section is devoted to a de-
scription of the morphological details of fatigue fractures and to their
explanation in terms of the plastic blunting process.
Plastic Blunting Process of Crack Growth

Evidence for the mechanism of Stage II crack propagation may be
obtained from a study of the profiles of striations on fracture surfaces.
Several types of profiles have been observed in Stage II and are illustrated
schematically in Fig. 2 [4,6,7]. That shown in Fig. 2a is the one most
commonly observed in ductile specimens broken at the highest stresses and
consists of parallel, juxtaposed depressions on both of the fracture sur-
faces. Another type of striation, Fig. 2b, occurs in the same circumstances
as the first type. Ridges on one fracture surface fit depressions on the
other. It is common, also, for small cracks, indicated by arrowheads in
Fig. 2b, to undercut the ridges on the fracture surfaces. The types of stria-
tion shown in Fig. 2c and d occur in specimens cycled at lower stresses.
They are essentially the same as those of Figs. 2a and b, but the ratio of the
depth of the depressions to the distance between them is higher. Examples
of these types of striation are provided in Fig. 3. A mixture of those shown
in Figs. 2a and b may be seen in Fig. 3a and are indicated with arrowheads
3
Slight direction changes at grain boundaries are the rule because the grains of
most materials undergoing fatigue are not randomly oriented but are textured by
working and annealing processes.
marked A and B, respectively. The increase of the striation depth-to-

spacing ratio typical of those in Figs. 2c and d is illustrated in 3b.
A satisfactory theory of fatigue crack propagation in Stage II must ex-
plain these morphologies.
Forsyth and Ryder [8] demonstrated by means of fatigue tests, pro-
grammed in amplitude, that each fracture surface striation is associated
with one stress cycle and proposed a mechanism for their formation. In
this mechanism, the increment of crack propagation per cycle is supposed
to take place by cleavage fracture ahead of the crack tip and the elevation
FIG. 2—Various types of morphology exhibited by striations on ductile fatigue

fracture surfaces, viewed in profile. The stress axis is vertical.
forming the striation occurs by ductile necking. It was later shown [6],
by direct observation of the processes occurring at the crack tip, to be
incorrect in the case of ductile metals. These observations were made by
sectioning specimens which had been strain cycled until they were known
to contain cracks and which were unloaded at various points on the
tension and on the compression sides of the cyclic stress strain hysteresis
loops. Since the metals chosen for this experiment were cycled at high
strains, and since they also had high values of stacking fault energy which
is associated with a lessened Bauschinger effect [9,10], the geometries of
the crack tips on unloading the specimens were representative of their
shapes under load. The mechanism of crack propagation deduced from
these and similar observations made later [77,72] is now called the "plas-
tic blunting process" and is described in Fig. 4.
It is convenient to begin the description from the zero load position of

the cracked specimen, Fig. 4a. As the tensile load of the cycle is applied,
the small double notch at the crack tip serves to concentrate slip zones
along planes at 45 deg to the plane of the crack and to maintain a square
geometry of tip, Fig. 4b. When the specimen is deformed to the maximum
tensile strain, Fig. 4c, the stress concentration effect of the crack is less-
FIG. 3—(a) A longitudinal section through a crack in an aluminum specimen

after 80 cycles at a strain range of 0.052 (number of cycles to failure — 180)
(X 250) and (b) a section cut prependicularly to the striations on the fracture surface
of an aluminum specimen cycled to complete failure at a strain range of 0.007
(X 250).
ened, the slip zones at the tip broaden and the crack tip blunts to a semi-
circular configuration. Upon application of compressional load, Fig. 4d,
the slip direction in the zones is reversed, the crack faces are crushed to-
gether, and the new crack surface created in tension is forced into the
plane of the crack and partly folded by buckling of the very front of the
crack tip into another notch, Fig. 4e. The formation of this notch, which
was not properly understood when the results were first reported [6],
serves to emphasize the 45 deg plastic zones when the stress cycle is re-
peated, Fig. 4/. Recently, McEvily et al [13] observed this mechanism di-
rectly as it occurred in polymer specimens undergoing cyclic strain. The

capacity of this material to deform plastically by considerable amounts,
combined with the choice of sheet specimens, allowed these workers to see
the plastic blunting process with the naked eye.
Formation of Fracture Surface Striations

The question now to be considered is whether or not the plastic blunt-
ing process can explain the various morphologies of fracture striations and
FIG. 4—The plastic blunting process of fatigue crack propagation in the stage
II mode: (a) zero load, (b) small tensile load, (c) maximum tensile load, (d) small
compressive load, (e) maximum compressive load, and (f) small tensile load. The
double arrowheads in (c) and (d) signify the greater width of slip bands at the crack
in these stages of the process. The stress axis is vertical.
the circumstances in which they occur. The answer lies, presumably, in

the mode of slip in the zones at the crack tip. Thus, in pure materials of
single phase cycled at high stresses these zones may be expected to be simi-
lar in both width and length, provided, of course, that the material has
enough slip systems. In this case and for many orientations of the crystal
structure with respect to the stress axis, the plane of the crack should bi-
sect the angle between two operable slip systems. The notch left at the
crack tip during the compression part of the cycle would therefore be sym-
metrical about the plane of the crack, as illustrated in Fig. 5a. This kind
of notch would be changed into the type of striation shown in Fig. la at
the end of the subsequent fatigue cycle. Figure 5b provides an example of
such a situation in the form of a photomicrograph of a section through
the crack tip in a copper single crystal which was strain cycled in a ten-
sion-compression test with the stress axis parallel to the [001] direction.
The uniformly juxtaposed depressions on both fracture surfaces are indi-
cated with arrowheads.
On the other hand, in most commercial polycrystalline materials, the
presence of grain boundaries and nonmetallic inclusions may be expected
to destroy the symmetry of the slip zones at the crack tip. Moreover, the
orientation of the crystal with respect to the stress axis at the crack tip may
be such as to have slip systems at "inconvenient" angles to the plane of
the crack. It will sometimes happen, therefore, that the two parts of the
notch left at a crack tip in compression will be asymmetrical, Fig. 6a; in
FIG. 5—(a) Uniform notches formed on the fracture surfaces of a homogeneous

material and (b) crack tip profile in a copper single crystal cycled in tension-com-
pression with the stress axis parallel to [001] (X250).
the subsequent stress cycle most of the strain occurring at the tip will be
concentrated in the most advanced notch, indicated with an arrowhead in
Fig. 6a. At the maximum tensile stress of this cycle, the crack tip will adopt
the configuration shown in Fig. 6b and on the next compression stroke
will show the profile of 6c, the type of striation of Fig. 2b. An example of
this occurrence, "caught in the act" and indicated with arrowheads, is
provided in Fig. 6d, showing the crack tip in tension at zero strain in poly-
crystalline nickel. Another example showing the starting condition neces-
sary for the formation of this kind of striation, this time in aluminum, is
shown in Fig. 6e; the more advanced part of the crack tip notch is indi-
cated by A, the less advanced by B. The kind of striation illustrated in Fig.
2b and indicated with arrowheads is formed by another variant of these
processes. In this situation, the buckling at the very front of the crack tip
during compression (Fig. 46) forces metal into one of the notches, com-
FIG. 6—T/ze changes occurring at a crack tip with asymmetrical notch during a
fatigue cycle: (a) in compression; (b) in maximum tension; (c) in compression; (d)
the crack tip in an annealed nickel specimen sectioned longitudinally after 360
cycles at a strain range of 0.017 (number of cycles to failure — 465), the specimen
was unloaded from the tension part of the cycle (X&50), and (e) a crack in a fully
compressed aluminum specimen cycled at 0.020 (X350).
pletely filling it. On the next tensile stroke the other notch is selected for
most plastic blunting, and on completion of the cycle the filled notch is
left as a ridge on the fracture surface undercut by a short crack.
The size of the notch left at a crack tip upon compression as a function
of the applied stress range has a bearing on the nature of the striations
formed at lower stresses, Figs. 2c and d. From experience of sectioning
specimens over a range of applied stresses, the author has the impression
that the size of compression notches varies less sensitively with applied
stress than does the striation spacing. At low stresses, therefore, these two
dimensions are nearly equal, and the "sawtooth" arrangement of stria-
tions develops.
Stage I Mode of Crack Propagation

At still lower stresses, when the Stage I mode of crack propagation is
operative, the formation of fatigue-type slip bands in the grains will en-
courage a marked asymmetry of the slip zones at a crack tip. Particularly
when one considers that cracks are nucleated in slip bands and that there
is some evidence [14,15] to indicate that fatigue bands are softer than the
surrounding metal, it is reasonable to suppose that the geometrical changes
at the crack tip will conform to the plane of the slip band. Thus, when
cracks are very shallow, of the order of a few microns in length, the type
of Stage I mechanism suggested by McEvily and Boettner [3] may operate.
At slightly greater crack depths, of the order of a grain diameter, the mech-
anism suggested by Schijve [16], or the one illustrated in Fig. 7 will occur.
In the latter, the increase in crack length per cycle is accomplished by
spreading of the crack tip to the width of the slip zone (Fig. 7b). Presuma-
bly, the required component of deformation normal to the slip band can
occur by duplex slip within the band, rather than by operation of a second
slip band, as is the situation in Stage II crack growth. This mechanism thus
explains the observation that cracks propagate along the common slip
planes while in Stage I and therefore at 45 deg to the stress axis, in con-
trast to Stage II growth. Since all these mechanisms involve crack growth
by localized deformation, they are essentially the plastic blunting process.
The difference between the Stage I and II mechanisms lies in the different
types of dislocation motion in the two cases and thus in the different asso-
ciated geometrical changes of the crack tip.
A major obstacle to understanding the mechanism of fatigue crack
growth in the early stages of failure is the general lack of features on
Stage I fracture surfaces in specimens broken in tension-compression or
bending tests. One might expect that when the amount of plastic blunting
at a crack tip is very small the folding process in compression will be mini-
mized. Resolvable visible striations therefore will not be left on the frac-
ture surface. However, in rare circumstances and especially when Stage I
cracks are deep, striations have been observed on Stage I fracture surfaces.
This is good indirect evidence that Stage I growth does occur by a variant
of the plastic relaxation process. Moreover, the work of Williams [17]
contains a link between striated and unmarked regions of Stage I fracture
surfaces. Williams found that small periodic changes in the mean stress
during a reversed test experiment left a kind of "tide-mark" on the fatigue
fracture surfaces of his /2-brass specimens.4 This mark, which is much
larger than a normal fracture surface striation, was formed by a slight
change in direction of the plane of crack propagation. By repeating the
FIG. 7—The plastic blunting process of crack propagation in Stage I: (a) zero
stress, (b) maximum tensile stress, and (c) compressive stress.
small change in mean stress at short intervals, he was able to delineate the
crack front throughout the fatigue life of his specimens and thus to ac-
count for every group of cycles spent in crack propagation. In the early
stages of crack growth, no feature was observed between the marks due to
the mean stress change in the ordinary way of Stage I growth. Before this
growth became exhausted, however, Williams was able to resolve stria-
tions between the marks due to the stress change by means of replication
electron microscopy. These striations were initially very small, having a
spacing of less than 0.1 /*, but gradually became larger until Stage I
abruptly changed into Stage II. This change was marked by the reorienta-
4
This phenomenon has also been observed in other metals and alloys [11,18].
tion of the crack plane from the normal slip plane of the alloy to one per-
pendicular to the stress axis, again in the ordinary manner. Williams noted
that the change from Stage I to Stage II occurred always when the stress
acting on the section of alloy remaining intact was ±41,000 psi, irrespec-
tive of the overall stress on the specimen. Presumably this stress was suffi-
cient to dominate the local condition of the alloy at the crack tip and to
nucleate the multiplicity of slip bands required for the Stage II plastic
blunting process. Williams could not say from this experiment whether or
not striations were also being formed at the earliest stages of growth since
they were too small to be observed. However, since the spacing of the
markings due to the small mean stress change increased in a smooth man-
ner, it seems reasonable to conclude that they are indeed formed. I con-
clude, therefore, that the operation of the plastic blunting process of crack
propagation is thus supported for the case of Stage I growth.
Markings Between, and Perpendicular to, Fracture Surface Striations

The purpose of this section is to describe two kinds of fracture surface
markings, occurring on a scale finer by an order of magnitude than the
striations, and to discuss their relation to the plastic blunting process.
In every case where fracture surface striations have formed and where
their size is of the order of one micron or greater, that is, during Stage II
growth, fine linear markings may be seen between, and usually roughly
parallel to, the larger striations which mark the progress of the crack in
each cycle. An example of this detail in indicated with arrowheads in the
electron fractograph of Fig. 8. Several explanations [19] have been offered
for the formation of these substriations. One suggestion is that they have
been caused by the rubbing together of the fracture surfaces before com-
plete failure has occurred. Another possibility is the following. During
rounding of the crack tip in the tensile stroke of the fatigue cycle, the "ac-
cidental" presence of free dislocation sources near a portion of the crack
front allows more deformation locally. The drag exerted on this part of
the crack front by an adjacent section deforming with greater difficulty
will slow down the "favored" section, relaxing it and forming a mark par-
allel to the crack gront, presumably by a process similar to that occurring
during compression of a rounded crack tip. Intermittent operation of this
mechanism is suggested to cause a number of these markings which are
similar to those formed on a slightly larger scale on fracture surfaces in
ductile specimens torn apart completely in tension. Yet another origin
for interstriation markings is also suggested by the plastic blunting process
of crack growth; it is possible that the substructure consists simply of slip
lines formed by plastic flow at the crack tip during both tension and com-
pression.
Direct evidence
Copyright for theInt'l
by ASTM validity of at
(all rights least theMon
reserved); lastDec
of these explanations
7 14:40:45 EST 2015
has been obtained recently
Downloaded/printed by [72] in a study of the fracture surfaces of cop-
per single crystals. In this experiment, cylindrical single crystals with

(001) axes parallel to the axis of the cylinders were notched circumfer-
entially and cycled in push-pull at high stresses. The [001] orientation of
the crystals was chosen to induce slip on more than one slip system.
Cracks propagating in Stage II thus did so on the (001) plane but simul-
taneously in all directions contained in the (001) plane as the cracks ad-
vanced at all points on the perimeters. Replicas taken of the fracture sur-
face showed interstriation lines which were parallel to the crack front in
directions where the crack advanced parallel to (110) directions. On the
other hand, where the crack was propagating parallel to [100], there were
FIG. 8—The fracture surface of a cold-worked copper specimen cycled at high

strain, and showing inter-striation markings. Replication electron microscopy. The
large arrow indicates the direction of crack propagation (X4500).
two sets of interstriation markings oriented approximately at 90 deg to

each other and at 45 deg to the crack front. This result is illustrated sche-
matically in Fig. 9; typical replication electronmicrographs of striations
at areas of the fracture surface marked A and B in Fig. 9 are provided in
Figs. 10a and b, respectively. The result may be interpreted on the basis
of the plastic relaxation theory of crack growth, as follows. Plastic round-
ing of the crack tip in areas near (110) directions of advance would occur
by slip on (111) and (111) planes. Slip lines formed on the crack tip by
this deformation would accordingly intersect the fracture surface along
(110) directions and would thus be parallel to the crack front. On the
other Copyright
hand, while deformation
by ASTM at thereserved);
Int'l (all rights crack front where7 the
Mon Dec direction
14:40:45 of
EST 2015
crackDownloaded/printed by
propagation is parallel to an (001) direction would take place on
the same slip systems, slip lines intersecting the fracture surface would
again be parallel to (110) directions and would thus adopt the morphology
observed.
The other prominent type of feature on striated Stage II fracture sur-
faces is a displacement of the crack plane observed as a ridge running
parallel to the direction of crack propagation and which frequently crosses
FIG. 9—(a) single crystal specimen and crack geometry, and (b) the slip line
morphology in a single striation on a (001) fracture plane.
many striations; the size of these ridges generally increases in the direction
of crack propagation. A typical ridge is illustrated in Fig. 11. Again, the
formation of such a feature may be explained by the plastic blunting
process of fatigue crack growth. In this connection, we must consider the
residual notch left at a closed crack tip in compression. It is bound to vary
in asymmetry along a crack front due to inhomogeneities in the slip zones.
Along some lengths of the front, the part of the notch below the overall
crack plane will be prominent, while at other lengths, that above (11).
Since Copyright
the plastic
by relaxation following
ASTM Int'l (all in the next
rights reserved); tension
Mon Dec cycle will
7 14:40:45 EST be con-
2015
centrated in the more by
Downloaded/printed prominent of the two parts of the notch, the crack
FIG. 10—The morphologies of inter-striation markings on the (001) fracture

surface of a copper single crystal. Replication electron microscopy (a) taken from
,Jf™£ marf^ A itl Fig- 9(b) where the crack Propagated in the [010] direction
(X4500), and (b) taken from the area marked B in Fig. 9(b) where the crack prop-
agated in the [110] direction (X5500).
plane will diverge onto slightly different levels. At a transition between

the two planes a shear step will be formed and may be expected to grow
greaterCopyright
in succeeding
by ASTMcycles, in the
Int'l (all rights manner
reserved); observed.
Mon Dec 7 14:40:45 EST 2015
Although these features are incidental to growth of the crack, it is neces-
sary that they should be explained by the plastic blunting process. Its con-
sistency in doing so is worthy of note.
Crack Propagation Mechanisms in Pulsating Tension or Compression
Tests
An essential component of the plastic blunting process of fatigue crack
propagation is the resharpening of a relaxed crack by the compression
FIG. 11—Typical ridges formed on a fatigue fracture surface in a direction

parallel to that of crack propagation (indicated by the arrow).
FIG. 12—The stress-strain behavior of a ductile metal subjected to pulsating

tension fatigue.
stroke of the fatigue cycle. However, fracture surface striations are gener-
ally observed in ductile materials broken in pulsating tension tests since
many tests have been conducted on sheet specimens. The purpose of this
section is to discuss why striations are formed nevertheless in these cir-
cumstances.
The link which resolves this question is the state of the material sub-
jected to pulsating tension tests. Any ductile metal tested in this way will
work-harden rapidly [10,20,21] to a saturation hardness even at stresses
up toDownloaded/printed
the ultimate tensile
by strength (UTS). The pulsating tension stress-
University of Washington (University of Washington) pursuant to License Agreement. No f
146 146 hhhhhhhhhhhh
strain loop will therefore quickly attain equilibrium, as illustrated in Fig.

12 and in this form will be a closed, very narrow loop. Accordingly, any
small degree of localized plastic deformation occurring at a crack tip will
be reversed by the total elastic constraint of the specimen as the applied
stress falls to zero. Indeed, it may be expected that even a small (but finite)
decrease of stress from the maximum applied tensile stress of the cycle will
cause a local reversal of strain at a crack tip, an associated crack tip fold
and, therefore, a fracture surface striation. We may accordingly conclude
that the observation of fracture surface striations on metals broken by
pulsating tension stresses provides further support for the plastic blunting
theory of crack propagation.
Similar kinds of arguments may be used to explain why cracks propa-
gate in specimens subjected to pulsating compression tests. Just as it is not
possible to perform pure pulsating tension tests, elastic constraints between
grains and inhomogeneities in microstructure introduce localized tensile
stresses in compression tests. Since these stresses are likely to be small,
crack propagation would be expected to occur by Stage I growth through-
out the cross section of a specimen. The general observation [11,22] that
cracks do indeed propagate at 45 deg to the stress axis in pulsating com-
pression tests is in line with this prediction. In some circumstances, how-
ever, the crack in a specimen subjected to pulsating compression may ap-
pear to have propagated at 90 deg to the stress axis, as in the case of a
graphitic nodular cast iron pointed out to the author by Hempel [22]. How-
ever, metallographic examination of this material indicated that cracks,
on a microscale, were propagating at 45 deg to the stress axis in the usual
manner, but between graphite nodules.
Stage II Crack Propagation in Thin Sheet Specimens
It is well known that cracks propagating by the Stage II mode in thin
sheets frequently change their plane of propagation from one at 90 deg to
the stress axis to another which is at 45 deg both to the stress axis and to
the plane of the sheet. This phenomenon usually occurs when the crack
has become long and when the plastic zone at the crack tip is accordingly
of large size. In this situation slip bands begin to operate at 45 deg to the
plane of the sheet and not in the usual "hinge" manner of Stage II growth.
The plastic blunting process thus becomes one of "slipping off" by plastic
deformation in combination with the rounding-off process. This changes
the plane of crack propagation to the geometry observed. Striations formed
in such circumstances are not so well pronounced as in bulk specimens,
and in some cases are completely obliterated by the rubbing together of
the fracture surfaces.
Conclusions
Fracture surface striations have been observed on an extremely wide
range University
of metals and alloys,
of Washington plastics,
(University and evenpursuant
of Washington) on strain crystallizing
to License Agreement.elas-
No further reprodu
tomers [23]. In view of this, and the direct and indirect evidence on crack
tip processes discussed above, it seems reasonable to conclude that the
plastic blunting process of fatigue crack propagation is quite general in
Stage II and probably operates in Stage I also but with differences in crack
tip geometry; the mechanism may be used to explain the varieties of frac-
ture surface striations and associated interstriation markings, and the dif-
ferences between the two growth stages. Another mechanism of crack
propagation [24-26], namely, that fracture occurs along the boundaries of
a dislocation cell structure at a crack tip, has been suggested to account for
the indirect observation that fatigue cracks appear to follow subbound-
aries in metal of high stacking fault energy. This may be a mechanism al-
ternative to the plastic blunting process. However, since such dislocation
boundaries are hardly planes of weakness, the role of the cell structure in
crack propagation may be to produce dislocation sources for operation of
the plastic blunting process. Moreover, the high cyclic strains at a crack
tip will form dislocation cell structures in such profusion that any conclu-
sion about the path of the crack with respect to the cells will not be un-
equivocal. Thus the observation of cell structure can be reconciled with the
plastic blunting theory.
Influence of Microstructure on the Mechanism and Kinetics of Crack

Propagation
Clearly, fatigue crack propagation is the result of localized plastic
deformation. Microstructure can therefore influence the phenomenon in
one of two ways. In combination with other variables, such as corrosive
environment and low temperature, it may inhibit or modify plastic defor-
mation to such an extent that the nature of the cracking process is changed
to a cleavage or quasi-cleavage phenomenon. Otherwise, it can only deter-
mine the scale of the plastic deformation and the detailed crack path in a
manner which is related in complex ways with the static properties of the
material.
The conditions under which microstructure can effect crack propaga-
tion will now be reviewed. Since the transition of the cracking process from
one involving plastic deformation to another involving cleavage is a spe-
cial case of the plastic blunting process, it is appropriate to describe this
aspect first. In the latter part of this section, the more general effect of mi-
crostructure on the plastic blunting process will be treated under separate
headings of high and low strain fatigue, since the amplitude of fatigue
stressing can affect the conditions somewhat differently without altering
the character of the process.
Transition from Ductile to Cleavage Crack Propagation Mode

Forsyth andbyRyder
Copyright ASTM[8] Int'lwere the first
(all rights to point
reserved); Monout
Decthat the fracture
7 14:40:45 sur-
EST 2015
faces Downloaded/printed
of fatigue specimens
by could exhibit two very different kinds of stria-
148 hhhhhhhhhhhhhhhhh
tion. This distinction is based mainly upon the appearance of the striation
types, one of which is characterized by very flat fracture facets and by
river patterns running parallel to the direction of crack propagation, the
other type being the subject of the first half of this paper and termed "duc-
tile." The former were called "cleavage" striations from their similarity to
that kind of fracture. Examples of micrographs of these two kinds of
striations are provided in Figs. 13a and b, respectively; both were obtained
from different areas of the same cold-worked, polycrystalline nickel speci-
men broken in high strain fatigue.
Forsyth deduced from the observation of cleavage striations in alumi-
num alloys that they are formed by true cleavage fracture [4] which, sub-
sequently and in the course of the tension stroke of a stress cycle, becomes
changed to a ductile shear fracture. This was necessary to explain the for-
mation of a small crack branching at 45 deg from the main crack and mak-
ing the mark of the striation. Support for the validity of this mechanism
was provided by several results, as follows:
1. The high strength aluminum alloys which have most consistently
formed cleavage striations also break by the cleavage mode in a Charpy
test at the temperature of liquid nitrogen [27].
2. The fracture plane of cleavage striations in aluminum alloys has been
found to be the (100) plane [28], which is the fracture plane for cleavage
in many brittle materials.
3. The formation of cleavage striations is more frequent in the presence
of a corrosive environment [1,29] and, in the case of iron, after the metal
has been charged with hydrogen [30].
However, other evidence which raises doubts about the validity of the
mechanism of formation of cleavage striations comprises the following:
1. The circumstances under which cleavage striations have been ob-
served (in aluminum alloys cycled at room temperature [1,29], in poly-
crystalline nickel also at room temperature [11], and in Udimet 700 at
1400 F [31]) are not such as to produce cleavage fracture under any other
conventional mode of testing.
2. The fracture plane of cleavage striations is not the same as the usual
plane of fracture, the (111) plane, when face centered cubic (fee) mate-
rials fail by cleavage in other circumstances [32,33], such as in the pres-
ence of liquid metal environments.
3. Cleavage, as a rule, is enhanced by corrosive environments. In one
circumstance, however, cleavage striations formed in vacuo, as follows
[11]. Heavily cold-swaged nickel, cycled at high strains, failed by simul-
taneous propagation of cracks from the surfaces of the specimens inward
and from interior flaws outward. All cracks which propagated from the
surface of the specimen were associated with the common ductile stria-
tion; the typical striations
Copyright by ASTM shownInt'l
in Fig. 13rights
(all b were taken from this
reserved); Monarea Dec
of 7 14
the fracture surface. All cracks by
Downloaded/printed starting from internal flaws showed the
University of Washington (University of Washington) pursua
FIG. 13—Optical fractographs from different areas of the same work-hardened

nickel specimen cycled at a strain range of 0.001 (Nf = 9430): (a) "cleavage" stri-
ations on a crack originating from a flaw in the center of the specimen (X830), and
(b) "ductile" striations near the point of crack initiation at the surface of the spec-
imen (X670).
cleavage type of striation (Fig. 13d). Since metallographic inspection of

the metals prior to testing revealed that the flaws were of a type formed
by the swaging process at nonmetallic inclusions and were not associated
with Copyright
porosity,byitASTM
seems
Int'lreasonable to conclude,
(all rights reserved); Mon Dec 7therefore,
14:40:45 ESTthat
2015the inter-
nal cracks were propagating
Downloaded/printed by in an effective vacuum.
University of Washington (University of Washington) pursuant to License Agreement. No further reprod
FIG. 14—Cleavage striations observed in longitudinal :section; stress axis ver-

tical: (a) age-hardened aluminum alloy, Forsyth [4] (X1320). (Courtesy of Pergamon
Press) and (b) Udimet 700 alloy tested at 1400 F, Wells and Sullivan [31] (X1100)
(Courtesy of Am. Society Metals).
4. The tendency for cleavage striations to be formed increases with de-

creasing frequency of stressing [29]; cleavage fracture is normally sensi-
tive to strain rate in the opposite sense.
5. The profiles of cleavage striations, of which two examples from dif-
ferent materials are provided in Fig. 14, invariably have the same form,
and appear closely similar to the kind of ductile striation illustrated in Fig.
2b.
6. There appears to be no significant difference in the rates of propa-
gationCopyright
of cracks by the ductile or the cleavage striation mechanism [11].
In Downloaded/printed
view of this conflicting
by evidence and of the extreme difficulty of ob-
LAIRD ON INFLUENCE OF METALLURGICAL STRUCTURE 1 51
taining direct evidence of such a small-scale dynamic process, complete

certainty as to the exact mechanism of the formation of the cleavage stria-
tion is presently unobtainable. It would only appear reasonable to con-
clude, however, that there is a considerable probability that cleavage plays
a part in the process. Furthermore, it is clear that environment is an im-
portant agent in the process, but to judge from the result on polycrystal-
line nickel, not an essential agent; it is possible that in this case the high
hydrostatic stresses associated with the internal flaws made up for the ex-
ternal stress concentrations caused by the local attack of a corrosive en-
vironment. However, the dominant connecting link in the circumstances
promoting cleavage striations appears to be the high strength associated
with the microstructure. The nature of the strengthening process, whether
by cold work or by precipitation, does not seem to be important.
On the other hand, I am not convinced that cleavage is really playing
a part in this process. Another possibility is that quasi cleavage is occur-
ring. In this mechanism, the usual plastic blunting process is operating;
slip on the {111} planes of the fee system would cause the resultant
plane crack of propagation to be symmetrically contained by* the opera-
tive slip planes. This explains why the (100) plane is selected for crack
propagation.3 The occurrence of river patterns can also be explained by
the plastic blunting phenomenon through application of the argument in
the section on "Markings Between, and Perpendicular to, Fracture Sur-
face Striations."
Although cleavage striations may involve a different mechanism of
fracture, this does not unduly affect the overall rate of crack propagation.
Two observations provide evidence for this conclusion. Firstly, the ductile
and cleavage striations along the same crack front in aluminum alloys show
no pronounced difference in spacing [1]. Secondly, the striations on the
fracture surfaces originating at the internal flaws of the polycrystalline
nickel specimens referred to above were equal in number, within a 15 per
cent experimental error, to those on cracks originating at the surfaces of
the same specimens [77]. Invoking Forsyth's proof that each striation cor-
responds to one cycle of stress and the further observation that the lengths
of crack traversed by each mode of fracture were also roughly equal sub-
stantiates the conclusion. Since plastic deformation must operate for both
ductile and cleavage striations, in the former to allow crack propagation
and in the latter to stop the cleavage fracture, it must be the controlling
factor of the overall rate of crack propagation in both modes of fracture.
In connection with the plastic deformation occurring at the end of this
cleavage part of fracture in the formation of a cleavage striation, the mech-
anism by which the striation is marked on the fracture surface remains to
5
A Copyright
(100) plane is frequently
by ASTM normal
Int'l (all rights to the
reserved); Monaxis
Dec of the specimen
7 14:40:45 EST 2015in aluminum
alloys, Downloaded/printed
because of texturing.
by
1523 hhhhhhhhhhhhhhhhhhhhhhhh
be discussed. Forsyth's suggestion [4] that a branching crack formed by a

shear fracture6 at a plane of 45 deg to the main crack may be valid. How-
ever, a careful examination of the profiles of tips of cracks propagating
by the cleavage mode reveals that there are two small cracks branching
from the main crack instead of only one as predicted by Forsyth's mecha-
nism (see those indicated by arrowheads in Fig. 14). It seems, therefore,
that the mark of the striation is probably formed by a variant of the process
FIG. 15—The propagation of fatigue cracks by the mode which involves cleav-
age fracture; stress axis vertical: (a) zero load, (b) and (c) tensile load, (d) maximum
tensile load, and (e) maximum compressive load.
known to form the ductile striation. In the case of cleavage striations, the
process may be expected to occur in the manner illustrated in Fig. 15. At
zero stress the crack will probably have the configuration shown in Fig.
15a. Assuming, then, that cleavage fracture ensues, Fig. I5b, and that it
is stopped by plastic blunting, shear will occur in a narrow band on either
side of the dotted lines drawn at the tip in Fig. \5c. Since the material is
6
It is not clear whether the type of fracture Forsyth is adopting is one of "shear
decohesion" originally suggested by Crussard et al [34] or else a ductile shear tear
fracture. If the by
Copyright latter,
ASTM then
Int'lhis
(allsuggestion is contained
rights reserved); Mon Decin7 essentials by 2015
14:40:45 EST the discussion
following.
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh 153
hard, it will have little remaining work-hardening capacity. Shear defor-

mation will therefore continue along the same band and will probably in-
duce the kind of spreading of the crack tip and, locally, of the bulk of the
specimen shown in Fig. I5d. Subsequent compression of such a crack will
give the configuration of Fig. I5e. On the next cycle, cleavage fracture
will start in that branch of the crack most nearly oriented towards the
cleavage fracture plane. The observation that the 45 deg subcracks usually
lie on the same side of the fracture (Fig. 14) is thus explained. If cleavage
fracture does not really occur, then the quasi-cleavage mode of the plastic
blunting process will occur by the same mechanism as that shown in Fig.
15 but with the omission of the cleavage increment illustrated in 156. The
geometrical relationship of slip planes to the plane of the crack will also
cause the 45 deg subcracks to lie on the same side of the fracture. In either
event, growth is occurring by a variant of the plastic blunting process.
Influence of Structure on Crack Propagation Mechanism High Strain Fa-

tigue
(a) Cold-Worked Structures—A prerequisite to considering the effect
of microstructure on the plastic blunting process of crack propagation is a
treatment of the effect of cyclic strains on the microstructure itself. In this
connection, it has long been known that cold-worked structures soften un-
der the action of cyclic strains, provided these are large enough [35-39].
Moreover, it has been observed quite recently that the flow stress attained
by a material with a high stacking fault energy is quite independent of its
prior annealing or strain history [10], over a range of cyclic strains giving
lives of from 102 to 104 cycles to failure. Since the metal in the path of a
crack will be undergoing a gradual increasing cyclic strain to one which
is many times that of the applied strain, it is reasonable to expect that the
effect of cold work on crack propagation will be negligible in high strain
fatigue. Direct support for this hypothesis has been obtained by Feltner
and Laird [10] in a current study of the fracture surfaces of annealed and
cold-worked copper specimens broken in lives of less than 104 cycles. They
found no significant differences in the numbers of striations on the fracture
surfaces of these specimens. A similar conclusion was also obtained for
annealed and cold-worked nickel [11].
Plots of cyclic-strain versus cycles to failure are well known from the
work of Coffin and his co-workers [39—41] to be insensitive to the degree
of cold work in many diverse materials. Bearing in mind that most of the
lives of specimens cycled at high strains are spent in crack propagation
[2], we must interpret this result as again indicating a negligible effect of
cold work on crack propagation.
(6) Structures Containing Second Phases—Second phases are usually
stableCopyright
in materials undergoing
by ASTM Int'l (all rightshigh strain
reserved); Monfatigue carried
Dec 7 14:40:45 ESTout
2015at room tem-
perature, because thebylifetimes of such materials are relatively short. The
Downloaded/printed
University of Washington (University of Washington) pursuant to License Agreement. No further reproductio
role of these phases in affecting crack propagation might be expected to

depend mainly on three factors: (1) their strength in relation to that of the
matrix, (2) their size, and (3) the nature of the bond between the second
phase and the matrix. Large second-phase particles, stronger than, and
well bonded to, their matrix should have little effect on crack propagation
since any crack would seek out the easiest path through the matrix. In some
circumstances, second-phase particles bonded by the metallic type of
bond to their matrix and having a much higher modulus and flow
stress than their matrix would exert a stress concentration and attract a
propagating crack. Under constant strain cycling conditions such stress
FIG. 16—A longitudinal section through the tip of a crack in a mild steel speci-
men after 47 cycles at a strain range of 0.127 (X250).
inhomogeneities would quickly be overcome by the establishment of the

dislocation cell structure typical of high strain fatigue. On the other hand,
a distribution of hard particles in a soft matrix would raise the overall
strain range of cycling in the matrix of the material. An increase in crack
propagation rate should therefore be expected to occur.
When the bond between large particles and their matrix is weak, as in
the situation of nonmetallic inclusions in metals or alloys of a single phase,
holes will be formed round the particles as they feel the stress concentration
of a crack. An example of such an occurrence is shown in Fig. 16, which
is a micrograph of a crack tip in a mild steel (0.15 per cent carbon) speci-
men containing pearlite colonies and nonmetallic inclusions cycled at a
strain range of 0.127 and having an associated life to failure of 65 cycles.
This phenomenon, which
Copyright by ASTM is similar
Int'l (all to the Mon
rights reserved); process
Dec 7by14:40:45
which EST
holes
2015are
formedDownloaded/printed
at the center ofbya specimen broken in direct tension, is still essen-
tially the plastic blunting process. However, the opening of holes com-
pletely suppresses the formation of striations, and such fracture surfaces
appear like the ordinary fracture surfaces of specimens broken in tension.
At lower testing strains, when the strain field associated with the crack be-
comes smaller than the interparticle spacing, striations are again observed.
It is interesting to note that the crack propagation rate in this particular
mild steel when holes were formed ahead of a crack tip was approximately
equal to that of a single-phase material, polycrystalline nickel, cycled at
an equivalent strain range [11]. This result held in spite of the fact that an
increase of crack propagation rate would be expected on the basis of the
argument given above for a distribution of hard particles in a soft matrix.
It should be emphasized at this point that the comparison between nickel
and iron was made on the basis of strain-cycling. The flow stresses of the
metals under these conditions reflected, of course, the inherent differences
of the metals.
To judge by the experiment of counting striations on the fracture sur-
faces of high strength aluminum alloys broken in high strain fatigue [12],
small, coherently bonded particles likewise have little effect either on the
mode of crack propagation or on its rate, as compared with single-phase
metals cycled at equivalent strains.
(c) Structures Hardened by Substitutional Alloying—While substitu-
tional alloys generally have higher yield strengths than the elements of
their composition, the really significant increase in strength of these alloys
is derived from their superior work-hardening properties. The effect of
the work-hardened structure on crack propagation mechanism should
thus be considered.
Since it has recently been observed that work-hardened alloys of
copper-7.5 per cent aluminum do not soften under cyclic strains to the
flow stress attained by cycling the alloy from the annealed condition [70],
a certain decrease in crack propagation rate might reasonably be predicted
in work-hardened alloy as compared with the annealed. This should re-
sult from a smaller degree of deformation in the work-hardened alloy at a
crack tip per cycle, under the same constant stress cycling conditions. If
the strain were the independent variable of testing, no difference (at worst
a very small difference) would be observed in crack propagation rate be-
tween these different materials as, for example, in the case of nickel and
iron mentioned above. However, McEvily et al [42], who investigated the
rates of crack propagation in annealed and cold worked sheets of «-brass
(copper—30 per cent zinc) under constant stress cycling conditions, found
no appreciable difference between them.7 The extreme roughness of the
7
The stresses which McEvily et al used in their tests were appropriate to the long-
life range of fatigue. However, since they notched their specimens to induce imme-
diate propagation
Copyright by by the Stage
ASTM II rights
Int'l (all mode,reserved);
these experiments
Mon Dec 7 qualify
14:40:45for
ESTconsideration
2015
in the high strain range. by
Downloaded/printed
University of Washington (University of Washington) pursuant to License Agreement. No further re
fracture surfaces of this alloy prevented them from measuring the micro-
hardness near enough to the fracture surface to gage whether or not the
flow stress of the alloy was independent of its initial condition. They con-
cluded tentatively, however, that the initially work-hardened alloy was
harder at the region of the crack tip than the annealed, and suggested that
some additional factor must be contributing to the similarity in crack
growth rates. They thought that the concentration of deformation in the
cold worked material in localized zones at the crack tip would allow
crack propagation per cycle equal to that in the annealed alloy, where the
crack tip deformation may be more spread. While it does seem likely that
the deformation resistance of the material will govern the geometry of a
crack tip, more work on a less exacting alloy will have to be carried out to
establish this and also whether or not the hardness of the alloy at a crack
tip is indeed dependent on the initial hardness, in the same way as the
bulk properties.
(d) Effect of Grain Size—The saturation flow stress attained by a metal
under a given cyclic strain has recently been found to be independent of
grain size as well as of initial cold work [10], provided the metal has a
high stacking fault energy, since such materials form a characteristic dis-
location cell structure dependent only on the amplitude of straining and the
temperature of testing. This cell structure produced by high cyclic strains
is the order of 1 /x in diameter [10,43,44] and is much smaller than the
grain size of the commercial metals. Therefore, the flow stress associated
with this structure (the saturation stress under a given cyclic strain) will
be larger than the yield stress of any usual annealed metal even when its
grain size is small. It should be pointed out also that the static strengths of
such metals are not strongly dependent on grain size. Therefore the grain
size will have virtually no effect on the capacity of metal at a crack tip to
resist deformation. Thus grain size will have no effect on the rate of crack
growth. This conclusion is supported by the observation of equal numbers
of fracture surface striations in unnotched aluminum specimens of differ-
ent grain size, all of which were cycled at the same stress range [19].8
A grain boundary may, of course, influence the path of a crack front
locally to produce an asymmetry in a fracture surface striation, or even to
change the direction of crack propagation. An example of such a process
is shown in the electron fractograph of Fig. 17. In this situation, a crack
propagating transgranularly in a copper specimen changed the direction
of its crack front along an annealing twin lying parallel to the direction of
crack propagation. It will be noted from Fig. 17, in further support of the
conclusion of independence of growth rate on grain size, that the spacing
of the fracture surface striations does not change in the vicinity of the twin.
8
The variation in the lives of these specimens, which increased with decreasing
grain size, was attributed to delayed crack nucleation in specimens of small grain
size. Downloaded/printed by
On the other hand, materials of low stacking fault energy should be

expected to show a dependence of crack propagation rate on grain size.
Since these materials do not form a dislocation cell structure at high cyclic
strains but a uniformly banded structure [10], and since the cyclic stress-
strain curve is dependent on grain size [10], the amount of plastic defor-
mation occurring at a crack tip in such a material should decrease in pro-
portion to decreasing grain size. Crack propagation occurring by the
plastic relaxation process would accordingly be slowed down. However,
FIG. 17—The fracture surface of a work-hardened copper specimen cycled at

a stress of ± 20,000 psi. Replication electron microscopy (X 10,000). The arrow
shows the direction of crack propagation.
there exist no data to my knowledge in the appropriate range of cyclic

strains by which this hypothesis may be tested.
Influence of Structure on Crack Propagation Mechanism: Low Strain

Fatigue
(a) Cold-Worked Structures—The negligible effect of microstructure
either on crack propagation rate or on mechanism in metals and alloys
cycled at high strains is reflected in the generality of the Coffin-Hanson
law of high strain fatigue [19,40,45]. At low strains of cycling, however,
a large variation in fatigue properties with respect to microstructure has
been observed [46] when stress, rather than strain, has been used as the
testing variable. One of the variables which has affected the forms of plots
of stress versus
Copyright numbers
by ASTM of cycles
Int'l (all to failure
rights reserved); Monin
Deca most striking
7 14:40:45 way is, of
EST 2015
course, the degree of initial
Downloaded/printed by cold work in the material. Although there have
been many investigations of the effect of cold work on crack nucleation

mechanisms, particularly the work of Gough, Forsyth, Wood, Kemsley,
and their co-workers [38,47-51] who studied the morphology of fatigue
slip bands in several metals and alloys, the phenomenon of fatigue crack
propagation has received less attention in this connection.
It is not difficult, however, to explain the much longer lives of heavily
cold-worked specimens cycled at low strains, in comparison with the an-
nealed [52]. The difference must again lie in how the initial condition of
the material is modified by cyclic strains, and thus affects the degree of
deformation occurring in the plastic blunting process. We note that work
softening is much reduced at low strains, even to negligible proportions.
A schematic plot (Fig. 18) of percentage change in flow stress of a work-
hardened metal versus the number of cycles to failure associated with a
wide range of cyclic strains, which has been taken from several sources
FIG. 18—The softening produced in typical cold-worked metals by cyclic strain

as a function of the number of cycles to failure associated with that strain.
[10,11,53], illustrates this behavior. Although the degree of work soften-

ing varies considerably with the type of metal, alloying content, degree of
initial cold work and temperature of testing, several metals show a marked
decrease in the amount of softening towards the long life end of the high
strain fatigue range, 105 cycles to failure. In low strain fatigue, therefore,
a crack propagating in cold-worked metal will encounter material of
higher flow stress which will offer greater resistance to the plastic blunting
process. Although the higher cyclic strains in the region of the crack tip
should induce much more work softening there than in the bulk of the
specimen, it is unlikely that enough cycles will occur to bring the material
to equilibrium at that point. The crack propagation rate will therefore be
slower. Moreover, since Stage I crack propagation is associated with tests
having small amounts of deformation per cycle, lives will be further en-
hanced on that account. A single result obtained on annealed and cold-
worked nickel supporting this conclusion [11] is shown in Fig. 19 in the
form Copyright
of a plotby of
ASTMcrack depth
Int'l (all rightsversus number
reserved); Mon Decof7 cycles
14:40:45 applied.
EST 2015 Crack
measurements were made
Downloaded/printed by on unnotched specimens so as to maximize the
amount of Stage I growth; except for the final stages of growth in which
fracture surface striations were used as a measure of crack depth on com-
pletely broken specimens, each point on the graph represents one speci-
men cycled partially to failure and then electropolished in a controlled way
to eliminate surface cracks and thus to measure the deepest crack. It is
quite clear from Fig. 19 that the final stages of crack propagation have ap-
proximately equal rates, and the majority of increase in life of the work-
hardened specimen is associated with Stage I growth in the depth of a few
grain diameters.
(b) Structures Containing Second Phases—Large inclusions have a
considerable influence on the mechanism and kinetics of crack nucleation,
particularly in steels [54]. Their influence on crack propagation, on the
other hand, is much less significant. The details of their effects are com-
FIG. 19—The growth of fatigue cracks in annealed and cold-worked nickel

cycled in tension-compression at a stress of ±50,000 psi.
plicated in that they can both accelerate and inhibit crack propagation
rates in various circumstances [55,56]. The largest effect of nonme-
tallic inclusions is to allow crack nucleation at their interface with the
matrix in a region ahead of an advancing crack, with subsequent propaga-
tion of this subcrack towards, and into conjunction with, the main crack
[1]. In some circumstances, particularly in constant stress tests of precipi-
tation hardened metals containing nonmetallic inclusions, the rate of crack
propagation may be doubled by this mechanism [1,12].
The influence of microconstituents in precipitation hardened alloys on
crack propagation offers a much more interesting problem. The large num-
ber of reports dealing with the fatigue properties of such materials under-
line the well-known result that crack growth rates are unduly high even at
stresses which are small in relation to the static strengths of the materials.
An explanation for this has come out of some recent studies of the problem
using transmission electron microscopy [57,55]. Two mechanisms seemed
to beCopyright
involved.
by First, there
ASTM Int'l (allisrights
some evidence
reserved); Mon to
Decindicate that
7 14:40:45 ESTreversion
2015 of
coherent precipitates,bysuch as Guinier-Preston (GP) zones, may occur un-
Downloaded/printed
der the action of cyclic stresses or else the process of overaging may take
place [57,55]. In either event, soft regions of alloy are formed9 in the mi-
crostructure and provide both crack nucleation sites and easy paths of
crack propagation. However, such zones are not only formed by the action
of cyclic stresses, they occur naturally in the normally aged microstruc-
ture both at grain boundaries, as is well known, and also within grains.
Examples of such transgranular soft zones in an aluminum-4 per cent
copper alloy are provided in Fig. 20 comprising two electronmicro-
graphs taken from specimens of slightly different heat treatments, namely
160 C for 16 hr and 200 C for 1 hr. The background precipitates are GP
II zones, and the larger plates within and on either side of the depleted
zones are of the & phase.10 Thus, locally, the transition of the metastable
precipitates towards equilibrium is well advanced, even though the maxi-
mum bulk hardness to be expected under these aging conditions would
not have been achieved until several more hours of aging had elapsed
[60]. The eradication of such structural inhomogeneities by thermo-me-
chanical treatments suggests a solution to the problem of the poor fatigue
properties of precipitation-hardened materials. Thus, the relatively good
fatigue properties of ausformed steels [61,62] in relation to iron, as com-
pared to those of aluminum alloys in relation to pure aluminum, may re-
flect this state of affairs. However, these kinds of treatments are only a
partial solution because of the inherent instability of age-hardened struc-
tures to cyclic strains.
(c) Structures Hardened by Substitutional Alloying—The importance
of cross slip has a long history of comment in studies of fatigue [3,63-66].
This stimulus has led to an understanding of the behavior of alloys with
low stacking fault energy in which two results have been emphasized:
firstly, that the fatigue properties of such alloys are uncommonly good
[65] as compared with precipitation hardened alloys11 and secondly,
that crack propagation rates are considerably lower than those of the
solvent metal with a higher stacking fault energy and a capacity for cross-
slip [3]. Initially it was uncertain whether or not these properties were a
discrete function of the increased yield stress of these alloys or of their
stacking fault energy [3,65]. However, recent work by Miller [67] has a
9
A recent study [15] and the early work of Kenyon [59] indicate that fatigue-
softened bands can be formed in annealed and cold-worked metals, respectively. It
seems, therefore, that this phenomenon is general, and all kinds of materials may
show an associated weakness for fatigue crack propagation.
10
The circular structures indicated with arrowheads in the electron micrograph
of Fig. 206 are not entirely composed of dislocation loops. Many segments are, in
fact, tf precipitates, but the contrast conditions are such as to image only their perim-
etral misfit dislocations. The identity of the plates was determined by tilting exper-
iments.
11
A normalized parameter for comparing the fatigue properties of different
materials in this by
Copyright context
ASTMmight be rights
Int'l (all takenreserved);
as the ratio
MonofDec
the 7pulsating
14:40:45 stress required
EST 2015
to cause failure in 10" cycles
Downloaded/printed byto the static flow stress of the material.
FIG. 20—Transmission electron-micrographs of aged Al-4%Cu alloy. The

arrowheads indicate 0' plates, (a) 160 C for 16 hr and (b) 200 C for 1 hr.
bearing on this problem. He studied crack propagation rates in two series

of alloys: one of them, copper-nickel alloys, showed a large variation in
yield stress without an appreciable corresponding change of stacking fault
energy and the other, copper-aluminum alloys, by way of contrast, having
increasing yield stress and simultaneously decreasing stacking fault energy
with increasing amounts of solute. Miller found that, for a given strain
amplitude, crack propagation rates in copper-nickel alloys were constant
with respect
Copyrighttobyyield
ASTMstress.
Int'l (allOn thereserved);
rights other hand, in copper-aluminum
Mon Dec 7 14:40:45 EST 2015 alloys,
the crack propagation by
Downloaded/printed rate decreased markedly with both decreasing stack-
ing fault energy and increasing yield stress. Miller inculpated stacking fault
energy, therefore, as the factor exerting the most influence on crack propa-
gation rate.
Miller interpreted this finding as supporting the argument that stacking
fault energy influences growth rate through its effect on substructure de-
velopment, citing the supposed tendency of cracks to propagate by separa-
tion of subgrain boundaries [26]. He therefore concluded that since sub-
structure formation would be minimized in alloys of low stacking fault
energy, crack propagation must accordingly be inhibited. However, Miller
observed striations on the fracture surfaces of his alloys in the ordinary way
of ductile metals, indicating operation of the plastic blunting process of
growth. Moreover, recent studies of the dislocation structures of alloys
with low stacking fault energy [10] indicate that substructure formation is
completely suppressed even at high cyclic strains; a high cyclic strain
would surely occur at a crack tip. An alternative interpretation of Miller's
results therefore is the following. Since the stress concentration at a sharp-
ened crack tip is extremely high [68], a very moderate stress would induce
plastic deformation at that point. Thus the yield stress of an alloy would
have little influence on crack propagation. The total amount of plastic
deformation which occurs there, however, will depend on the rate at
which the material can be work hardened locally. The much greater work-
hardening capacity of alloys with low stacking fault energy will thus resist
crack tip deformation and induce a lower crack propagation rate.
(d) Effect of Grain Size—It is well known that the lives of many mate-
rials cycled at low strains (cycles to failure 105 and greater) increase mark-
edly with decreasing grain size [69-71]. This contrast to the grain size in-
dependence of life at high cyclic strains is a consequence, presumably, of
two factors. First, the yield stress of a material increases with decreasing
grain size. Therefore, in a series of tests conducted at the same stress am-
plitude on specimens of varying grain size, the ratio of applied stress to
yield stress will decrease with decreasing grain size. Since Stage I crack
growth is emphasized at small values of this ratio, specimens of small
grain size will show lower crack propagation rates and, therefore, longer
lives. The second factor concerns the rate at which slip can spread from
one grain to another and thus affect the kinetics of Stage I growth. We
know by means of transmission electron microscopy that Stage I cracks
propagate through dislocation cell structures confined to the slip bands of
their crack planes [72-73]. These structures are different than the dislo-
cation loop structures in the bulk of the metal associated with low strain
cycling [74]. Thus a propagating Stage I crack is constantly promoting a
dislocation -cell structure ahead of it. Grain boundaries will act as barriers,
therefore, to cracks propagating at low strain amplitudes in contrast to
the behavior
Copyright observed
by ASTM atInt'lhigh
(all strains where the
rights reserved); Mon general
Dec 7 dislocation
14:40:45 ESTcell
2015
structure is controllingbythe plastic blunting process. The macroscopic ten-
Downloaded/printed
sile stress necessary to propagate plastic strains across a grain boundary,

(and thus the dislocation cell structure associated with a Stage I crack)
depends, in a complicated way, on the reciprocal of the square root of
the grain size [75]. Thus Stage I crack propagation will be inhibited in
materials of small grain size.
The slip behavior of the material plays an important role in this respect.
A capacity for cross slip, in particular, increases the rate at which slip can
spread from one grain to another along the line of impingement of a slip
band against a grain boundary [75]. The recent observation of Forrest
and Tate [70] that grain boundaries in 70-30 brass, having a low stacking
energy and a planar slip mode, are strong barriers to Stage I cracks sup-
ports this interpretation. Furthermore, we should expect that the grain
boundaries in materials having a wavy slip mode should offer less resis-
tance to Stage I crack propagation than those in materials of planar slip
mode. There are no data, to my knowledge, by which this prediction may
TABLE 1—The influence of grain size on the fatigue properties of materials with
different slip modes.
Source Grain Stress

-r^ M
at 106 cycles• /
e fffl/Up
e
0.1% Proof
Size, mm to Failure, ffe , psi ' Stress, psi
Forrest and Tate [70], 70-30

brass 0.016 24 200 ((rel) 1.00 ~22 400
0.022 21 800 1.11
0.040 18 200 1.33 ~13 700
Klesnil et al [71], a-iron 0.011 32 600 ( ff .i) 1.00 40 000
0.020 30 700 1.06 33 600
0.073 27 400 1.19 24 700
be tested directly. However, some S-N plots showing the effect of grain
size on fatigue life may be analyzed to bear upon the problem. Thus, in
Table 1, data have been tabulated from the results of Forrest and Tate
[70] who studied the influence of grain size on the fatigue behavior of
70-30 brass, and from those of Klesnil et al [71] who studied the same
problem in iron. The fact that these workers used tension-compression
tests for their studies simplifies the comparison of their data. The table
contains values of grain size, the stress required to cause failure in 106
cycles for each grain size, the 0.1 per cent proof stress, and the ratio of
stress at 106 cycles for the smallest grain size to the same stress for the
particular grain size under consideration. While the data are meagre and
must therefore be treated with reserve, the stress ratios increase more in
the case of the brass than in the case of iron for which the data extend
over a larger grain size interval. The conclusion that grain boundaries in
materials of planar slip mode offer a greater barrier to Stage I crack
propagation
Copyrightisbythus supported,
Influence of Structure on Crack Propagation Mechanism: Summary

At high cyclic strain amplitudes, the microstructures of most materials
are disturbed to such an extent that differences between their initial
conditions tend to be minimized. The flow stress of the material thus
depends on the testing conditions. In consequence the plastic blunting
process occurs uniformly with respect to material under constant strain
cycling conditions, although, of course, the stress required to enforce
the strain amplitude will depend on material to a large extent.
On the other hand, in tests conducted at low strain amplitudes, a
wider spectrum of behavior has been observed. This may be due in part
to our past inability to measure and control small cyclic strains pre-
cisely. Thus, the stress amplitude has generally been used as the inde-
pendent variable of testing. Now it is reasonable to expect that those
cyclic stress-strain curves which vary with respect to the initial condition
of the material will show a greater variation when testing amplitudes are
low [10]. Thus, under constant stress cycling conditions the strain in a
specimen per cycle is going to depend on the nature of its microstructure.
Therefore, great variations in crack propagation rates will be observed.
It seems likely that the uniformity of behavior shown by all materials
tested at high strain amplitudes will also be observed at low amplitudes
provided that the latter tests are compared on the basis of plastic strain
amplitude. The recent work of Morrow [62] who used accurate electronic
means of measuring strain in the long life range of fatigue and found
that the Coffin-Manson law was valid at longer lives provides support
for this conclusion.
General Conclusion
Both the direct evidence for geometrical changes occurring at a crack
tip during the course of each fatigue cycle and the qualitative depend-
ence of crack propagation rate on microstructure support the plastic
blunting theory of fatigue crack growth; the strain-hardening associated
with the localized plastic deformation appears to be the rate controlling
factor. Apart from Grosskreutz's recent attempt [76] to explain the
Coffin-Manson law from first principles, this result has not yet been
used to derive a quantitative expression for fatigue crack propagation and
its relation to microstructure. Thus the most recent mathematical theory
of fatigue by Weertman [77], who applied the dislocation formalism of
Bilby et al [78] to the fatigue problem, assumes the summation of cyclic
displacements at a crack tip until an arbitrary critical value is attained;
at this stage the crack is assumed to extend itself. In this assumption
Weertman's theory is similar to that of Head's theory of fatigue [79,80]
whichCopyright
has been discredited by the discovery of the phenomenon of work-
softening under cyclicbystrains. Thus Weertman's theory must also be con-
Downloaded/printed
LAIRD ON INFLUENCE OF METALLURGICAL STRUCTURE 1 65
sidered inadequate for its failure to express reality, although it does pre-
dict the correct dependency of crack propagation rate on stress, in some
circumstances.
However, the value of Weertmari's theory is undeniable in that it has
stimulated McEvily and Johnston [7] to express crack propagation rate
as a function of both stress and microstructure; in this formalism the
work-hardening capacity of the material is accounted for through applica-
tion of its ultimate tensile strength. McEvily and Johnston have thus de-
rived an equation for the rate of crack propagation in any material, using
the parameters of the static properties. The limited success of this
equation when tested against crack propagation data for aluminum
alloys, copper and nickel alloys and steels, indicates that while con-
siderable progress has been made in recent years towards understanding
the mechanism of fatigue failure, both the laws and the theory of
fatigue crack propagation are still in primitive form.
A cknowledgments
I wish to thank C. E. Feltner for many valuable discussions and Y. C.
Liu for the gift of some copper single crystals. My thanks are also due
to the following colleagues: H. I. Aaronson, C. E. Feltner, J. J. Harwood,
T. L. Johnston, and A. J. McEvily for their helpful criticism of the
manuscript.
References
[/] P. J. E. Forsyth, "A Two Stage Process of Fatigue Crack Growth," Proceedings,
Crack Propagation Symposium, Cranfield, 1961.
[2] C. Laird and G. C. Smith, "Initial Stages of Damage in High Stress Fatigue,"
Philosophical Magazine, Vol 8, 1963, pp. 1945-1963.
[3] A. J. McEvily and R. C. Boettner, "On Fatigue Crack Propagation in fee Me-
tals," Acta Metallurgica, Vol 11, 1963, pp. 725-744.
[4] P. J. E. Forsyth, "Fatigue Damage and Crack Growth in Aluminum Alloys,"
Acta Metallurgica, Vol 11, 1963, pp. 703-716.
[5] H. C. Burghard and D. L. Davidson, "Fracture Mechanisms and Fracture Sur-
face Topography," International Conference on Fracture, Sendai, September,
1965.
[6] C. Laird and G. C. Smith, "Crack Propagation in High Stress Fatigue," Phila-
sophical Magazine, Vol 7, 1962, pp. 847-857.
[7] A. J. McEvily, Jr. and T. L. Johnston, "On the Role of Cross-Slip in Brittle
Fracture and Fatigue," International Conference on Fracture, Sendai, Sep-
tember, 1965.
[8] P. J. E. Forsyth and D. A. Ryder, "Some Results of the Examination of Alu-
minum Alloy Specimen Fracture Surfaces," Metallurgia, Vol 63, 1961, pp.
117-124.
[9] K. U. Snowden, "Dislocation Arrangements during Cyclic Hardening and
Softening in Al Crystals," Acta Metallurgica, Vol 11, 1963, pp. 675-684.
[10] C. E. Feltner and C. Laird, Scientific Reports Nos. SL-66-93 and SL-66-115,
Ford Motor Co., 1966.
[11] C. Laird, "Studies of High Strain Fatigue," Ph.D. thesis, Cambridge Univer-
sity, 1962. by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
Copyright
Laird, unpublished
[12] C,Downloaded/printed by work.
[13] A. J. McEvily, R. C. Boettner, and T. L. Johnston, "On the Formation and

Growth of Fatigue Cracks in Polymers," 10th Sagamore Army Materials Re-
search Conference, 1963.
[14] R. K. Ham and T. Broom, 'The Mechanism of Fatigue Softening," Philo-
sophical Magazine, Vol 7, 1962, pp. 95-103.
[75] O. Helgoland, "Cyclic Hardening and Fatigue of Copper Single Crystals,"
Journal, Institute of Metals, Vol 93, 1965, pp. 570-575.
[76] J. Schijve, "Analysis of the Fatigue Phenomenon in Aluminum Alloys,"
NLR-TR M2122, National Aero and Astronautical Research Inst., Amsterdam,
April, 1964.
[77] H. D. Williams, "The Fatigue of /3-Brass," Ph.D. thesis, Cambridge University,
1961.
[75] G. Jacoby, "Observation of Crack Propagation on the Fracture Surface,"
Symposium on Current Aeronautical Fatigue Problems, Rome, April, 1963.
[79] R. C. Boettner, C. Laird, and A. J. McEvily, "Crack Nucleation and Growth
in High Strain-Low Cycle Fatigue," Transactions, Am. Institute Mining,
Metallurgical, and Petroleum Engrs., Vol 233, 1965, pp. 379-387.
[20] E. Hein and R. A. Dodd, "Softening of Strain-Hardened Polycrystalline Copper
during Reversed Stress Fatigue and Tensile Fatigue," Transactions, Am. In-
stitute Mining, Metallurgical, and Petroleum Engrs., Vol 221, 1961, pp. 1095-
1098.
[27] C. E. Feltner and G. M. Sinclair, "Cyclic Stress Induced Creep of Close-Packed
Metals," Proceedings, International Conference on Creep, New York, 1963.
[22] M. Hempel, private communication.
[23] E. H. Andrews, "Crack Propagation in a Strain-Crystallizing Elastomer,"
Journal of Applied Physics, Vol 32, 1961, pp. 542-548.
[24] J. Holden, "The Formation of Sub-grain Structure by Alternating Plastic
Strain," Philosophical Magazine, Vol 6, 1961, pp. 547-558.
[25] J. Holden, "Observations of Cyclic Structure at Large Ranges of Plastic Strain,"
Acta Metallurgica, Vol 11, 1963, pp. 691-701.
[26] D. H. Avery and W. A. Backofen, "Fracture of Solids," John Wiley & Sons,
Inc., New York, 1963, pp. 339-382.
[27] P. J. E. Forsyth, "Cleavage and Interrupted Crack Growth in Aluminum
Alloys," Journal, Institute Metals, Vol 93, 1965, p. 456.
[25] P. J. E. Forsyth, C. A. Stubbington, and D. Clark, "Cleavage Facets Observed
on Fatigue Surfaces in an Aluminum Alloy," Journal, Institute Metals, Vol
90,1962, p. 238.
[29] C. A. Stubbington, "Some Observations on Air and Corrosion Fatigue of an
Aluminum-7.5% Zinc-2.5% Magnesium Alloy," Metallurgia, Vol 65, 1963,
pp. 109-121.
[30] D. A. Ryder, private communication.
[31] C. H. Wells and C. P. Sullivan, "Low Cycle Fatigue Damage of Udimet 700
at 1400°F," Transactions, Am. Society Metals, Vol 58, 1965, pp. 391-402.
[52] R. G. Davies and N. S. Stoloff, "On the Yield Stress of Aged Ni-Al Alloys,"
Transactions, Am. Institute Mining, Metallurgical, and Petroleum Engrs., Vol
233,1965,pp. 714-719.
[33] N. S. Stoloff, R. G. Davies, and T. L. Johnston, "Slip Character and Liquid
Metal Embrittlement," Report No. SL65-42, Ford Motor Co., July, 1965.
[34] C. Crussard, J. Plateau, R. Tamhankar, G. Henry, and D. Lajeunesse, "A Com-
parison of Ductile and Fatigue Fractures," Proceedings, International Con-
ference on the Atomic Mechanisms of Fracture, Swampscott, 1959.
[35] H. N. Polakowski, "Restoration of Ductility of Cold-Worked Aluminum, Cop-
per, and Low-Carbon Steel by Mechanical Treatment," Proceedings, Am. Soc.
Testing Mats., Vol 52, 1952, pp. 1086-1097.
[36] H. N. Polakowski and A. Polchoudhuri, "Softening of Certain Cold-Worked
Metals under the Action of Fatigue Loads," Proceedings, Am. Soc. Testing
Mats., Vol 54, 1954, pp. 701-712.
[37] D.Copyright
S. Dugdale, "Stress-Strain
by ASTM Int'l (allCycles
rights of Large Amplitude,"
reserved); Mon Dec 7Journal
14:40:45of EST
the 2015
Mechanics and Physicsby
Downloaded/printed of Solids, Vol 7, 1959, pp. 135-142.
[38] W. A. Wood and R. L. Segall, "Softening of Cold-Worked Metal by Alter-

nating Strain," Journal, Institute Metals, Vol 86, 1958, pp. 225-228.
[39] L. F. Coffin and J. H. Read, "A Study of the Strain Cycling and Fatigue Be-
havior of a Cold-Worked Metal," International Conference on Fatigue of
Metals, London, 1956.
[40] L. F. Coffin and J. Tavernelli, "Cyclic Straining and Fatigue of Metals/' Re-
port No. 58-RL-2100, General Electric Research Laboratory, 1958.
[41] E. E. Baldwin, G. J. Sokol, and L. F. Coffin, "Cyclic Strain Fatigue Studies on
A1S1 Type 347 Stainless Steel," Proceedings, Am. Soc. Testing Mats., Vol 57,
1957, pp. 567-586.
[42] A. J. McEvily, Jr., R. C. Boettner, and A. P. Bond, "On Cold Work and Fa-
tigue-Crack Propagation in a-Brass," Journal, Institute Metals, Vol 93, 1965,
pp. 481-482.
[43] J. C. Grosskreutz and P. Waldow, "Substructure and Fatigue Fracture in
Aluminum," Acta Metallurgica, Vol 11, 1963, pp. 717-724.
[44] J. C. Grosskreutz, "Development of Substructure in Polycrystalline Aluminum
During Constant Strain Fatigue," Journal of Applied Physics, Vol 34, 1963,
pp. 372-379.
[45] S. S. Manson, T. N. 2933, National Advisory Committee for Aeronautics, 1953.
[46] N. Thompson and N. J. Wadsworth, "Metal Fatigue," Advances in Physics,
Vol 7,1958, pp. 72-166.
[47] H. J. Gough, "Crystalline Structure in Relation to Failure of Metals—Es-
pecially by Fatigue," Proceedings, Am. Soc. Testing Mats., Vol 33, 1933, pp.
3-114, Part II.
[48] P. J. E. Forsyth, 'The Basic Mechanism of Fatigue and Its Dependence on the
Initial State of a Material," International Conference on Fatigue of Metals,
London, 1956.
[49] P. J. E. Forsyth, "Some Metallographic Observations on the Fatigue of Metals,"
Journal, Institute Metals, Vol 80, 1951, pp. 181-186.
[50] D. S. Kemsley, "Crack Paths in Fatigued Copper," Journal, Institute Metals,
Vol 85,1956, pp. 420-421.
[51] D. S. Kemsley, "The Behavior of Cold-Worked Copper in Fatigue," Journal,
Institute Metals, Vol 87, 1958, pp. 10-15.
[52] N. E. Frost, 'The Effect of Cold Work on the Fatigue Properties of Two Steels,"
Metallurgia, Vol 62, 1960, pp. 85-90.
[55] R. J. Warrick, private communication.
[54] F. B. Stulen, H. N. Cummings, and W. C. Schulte, "Relation of Inclusions to
the Fatigue Properties of High Strength Steels," International Conference on
Fatigue of Metals, London, 1956.
[55] R. M. N. Pelloux, "Fractographic Analysis of the Influence of Constituent
Particles on Fatigue Crack Propagation in Aluminum Alloys," Transactions,
Am. Society Metals, Vol 57, 1964, pp. 511-518.
[56] L. H. Glassman and A. J. McEvily, "Effect of Constituent Particles on the
Notch-Sensitivity and Fatigue-Crack-Propagation Characteristics of Alumi-
num-Zinc-Magnesium Alloy," National Aeronautics and Space Administra-
tion TD D-328, April, 1962.
[57] J. B. Clark and A. J. McEvily, "Interaction of Dislocations and Structures in
Cyclically Strained Aluminum Alloys," Acta Metallurgica, Vol 12, 1964, pp.
1359-1372.
[58] C. A. Stubbington and P. J. E. Forsyth, "Some Observations on Microstruc-
tural Damage Produced by Fatigue of an Aluminum-7.5% Zn-2.5% Mag-
nesium Alloy at Temperatures between Room Temperature and 250°C," Acta
Metallurgica, Vol 14, 1966, pp. 5-12.
[59] J. N. Kenyon, "The Reverting of Hard-Drawn Copper to Soft Condition under
Variable Stress," Proceedings, Am. Soc. Testing Mats., Vol 50, 1950, pp. 1073-
1084.
[60] J. M. Silcock, T. J. Heal, and H. K. Hardy, "Structural Aging Characteristics
ofCopyright by ASTM Int'l (allAlloys,"
Binary Aluminum-Copper rights reserved); Mon DecMetals,
Journal, Institute 7 14:40:45 EST1954,
Vol 82, 2015
pp. 239-248.
[61] F. Borik, W. M. Justusson, and V. F. Zackay, "Fatigue Properties of an Aus-

formed Steel," Transactions, Am. Society Metals, Vol 56, 1963, pp. 327-338.
[62] J. Morrow, "Fatigue Properties of Metals," Manual, Society Automotive Engrs.,
ISTC, Division 4, April, 1964.
[63] A. J. McEvily and E. S. Machlin, "Critical Experiments on the Nature of Fa-
tigue in Crystalline Materials," International Conference on Atomic Mech-
anisms of Fracture, Swampscott, 1959.
[64] T. Broom and R. K. Ham, "The Hardening of Copper Single Crystals by Fa-
tigue," Proceedings, Royal Soc., Vol A251, 1959, pp. 186-199.
[65] J. A. Roberson and J. C. Grosskreutz, "Fatigue of Copper-Zinc Alloys at
100°K," Acta Metallurgica, Vol 11, 1963, pp. 795-798.
[66] N. F. Mott, "A Theory of the Origin of Fatigue Cracks," Acta Metallurgica,
Vol 6, 1958, pp. 195-197.
[67] G. A. Miller, "Fatigue Crack Growth in Some Copper Base Alloys," Ph.D.
thesis, Massachusetts Institute Technology, 1965.
[65] F. A. McClintock, "The Growth of Fatigue Cracks under Plastic Torsion,"
International Conference on Fatigue, Institute Mechanical Engr., London,
1956.
[69] G. M. Sinclair and W. J. Craig, "Influence of Grain Size on Work-Hardening
and Fatigue Characteristics of Alpha Brass," Transactions, Am. Society
Metals, Vol 44, 1952, pp. 929-948.
[70] P. G. Forrest and A. E. L. Tate, "The Influence of Grain Size on the Fatigue
Behavior of 70/30 Brass," Journal, Institute Metals, Vol 93, 1965, pp. 438-444.
[71] M. Klesnil, M. Holzmann, P. Lukas, and P. Rys, "Some Aspects of the Fatigue
Process in Low-Carbon Steel," Journal, Iron and Steel Inst., Vol 203, 1965,
pp. 47-53.
[72] E. E. Laufer and W. N. Roberts, "Dislocation Structures in Fatigued Copper
Single Crystals," Philosophical Magazine, Vol 10, 1964, pp. 883-885.
[73] M. Klesnil and P. Lukas, "Dislocation Arrangement in the Surface Layer of
a-Iron Grains During Cyclic Loading," Journal, Iron and Steel Inst., Vol 203,
1965, pp. 1043-1048.
[74] C. E. Feltner, "A Debris Mechanism of Cyclic Strain Hardening for fee Met-
als," Philosophical Magazine, Vol 12, 1965, pp. 1229-1248.
[75] T. L. Johnston, R. G. Davies, and N. S. Stoloff, "Slip Character and the Duc-
tile to Brittle Transition of Single-Phase Solids," Philosophical Magazine, Vol
12, 1965, pp. 305-317.
[76] J. C. Grosskreutz, "A Theory of Stage II Fatigue Crack Propagation," Air
Force Materials Laboratory, Technical Report No. 644-15, 1965.
[77] J. Weertman, "Rate of Growth of Fatigue Cracks as Calculated from the
Theory of Infinitesimal Dislocations Distributed on a Plane," International
Conference on Fracture, Sendai, September, 1965.
[75] B. A. Bilby, A. H. Cottrell, and K. H. Swinden, "The Spread of Plastic Yield
from a Notch," Proceedings, Royal Soc., London, Vol A272, 1963, pp. 304-314.
[79] A. K. Head, "The Growth of Fatigue Cracks," Philosophical Magazine, Vol 44,
1953,pp. 925-938.
[50] A. K. Head, "The Propagation of Fatigue Cracks," Journal of Applied Me-
chanics, Vol 78, 1956, pp. 407-410.
DISCUSSION ON INFLUENCE OF METALLURGICAL STRUCTURE 169
DISCUSSION
/. Schijve1 (written discussion)—The author has associated "Stage I"

with crack growth along a single slip plane (or a slip band) and with the
first period of the fatigue life. I personally believe that crack growth
along a slip plane is promoted at the surface of the material due to the
lower restraint on plastic flow. Moreover, there is also less restraint on
the path of the crack since the crack front is terminating at the surface.
Consequently, crack growth along slip bands need not be restricted to
Stage I if this is slow growth during the first part of the fatigue life. It
can occur at high crack rates, at least that is what we found on bare
aluminum alloys,2 while at the same time crack growth at some depth
below the surface was apparently noncrystallographic in nature. Would
Dr. Laird please comment on this?
Campbell Laird (author)—Dr. Schijve has suggested that Stage I
growth along slip bands may occur at the surface of a material while the
overall rate of crack growth is high, and while, at the center of the speci-
men, the crack is propagating by the Stage II mechanism, presumably
the plastic blunting process. He also believes that there is less restraint
on the path of the crack at the surface, citing experiments on aluminum
alloys.
The phenomenon which Dr. Schijve is describing has been commonly
observed in crack propagation experiments with sheet specimens and
varies considerably depending on stress amplitude. For example, I
have observed a similar phenomenon in pure aluminum where, in a crack
propagation test on a notched specimen !/4 in. thick, fracture surface
striations were formed in the center of the specimen, the crack propaga-
gation rate was of the order of 1 p. per cycle, the crack length about 1A in.,
and the gross applied stress ±5,000 psi. At the surface of the specimen,
extending to a depth of about 0.050 in., the crack propagated at an angle
of 45 deg to both the stress axis and the average plane of the crack. The
crack front was curved considerably, the radius of curvature being some-
what less than the thickness of the specimen and the center of curvature
on the crack nucleation side of the front. This indicates that, in contrast
to Dr. Schijve's situation, there was considerable restraint on the propa-
gation of the crack at the surface. The gross slip bands forming in asso-
ciation with this cracking were visible to the naked eye and must, there-
1
Director of technical services, National Aerospace Laboratory, NRL, Amster-
dam, Holland.
2
J. Downloaded/printed
Schijve, "Analysis by
of the Fatigue Phenomenon in Aluminum Alloys," NRL-
TR M 2122, Amsterdam, April, 1964.
fore, have traversed several grains. I think that the term "Stage I growth"
should be reserved for growth along a slip band within a single grain and
where the associated slip band is also restricted to one grain, although
rates of crack propagation may approach a large fraction of a micron per
cycle, as Williams has observed in /?-brass.8 If the crack propagation
which Schijve is describing appears more similar to that observed in the
aluminum than in William's brass, it would perhaps be more accurate to
term it: Stage II crack propagation under conditions of plane stress,
rather than as Stage I growth. Since I believe that both Stages I and II
involve irreversible geometrical changes due to localized plastic flow and
that they are therefore fundamentally similar, any distinction between
them may appear arbitrary. Making a distinction based mainly on the
different scale of the phenomena is useful, however, when considering
the fatigue of an unnotched, ductile specimen which has failed in a
million cycles, for example. The fracture surface of such a specimen will
show large numbers of striations having an average spacing of about 1 //,.
For a laboratory specimen of ordinary size, for example, a diameter of
about l/2 in., this observation indicates that only several thousand cycles
are spent in Stage II crack propagation. Since it is well known that
fatigue cracks are nucleated very early in the fatigue life over the whole
range of applied stress, the average rate of crack propagation in the
early stages of cracking, that is, Stage I, is at least two orders of magni-
tude slower than that in Stage II.
F. A. McClintock4 (written discussion)—Dr. Laird strongly supports
the theory that fatigue crack growth is controlled primarily by localized
plastic deformation. Liu5 has pointed out that if this is so, dimensional
analysis dictates that the growth rate must vary linearly with crack length,
since for a crack in a large part the only significant length parameter is
the crack length itself. Numerous observations have shown a stronger
dependence. Apparently, this paradox can be resolved by considering the
data of Pelloux (Ref 55 of the paper). He found that the microscopic
crack growth rate, as measured by the striation spacing, was indeed
linearly proportional to crack length, as indicated by the log-log plot in
Fig. 21. This is consistent with the theory that striation growth is con-
trolled solely by localized plastic deformation.
The mechanism that causes the exponent of the dependence of macro-
scopic growth rate on crack length to be greater than unity appears to
by one of re-nucleation of cracks at inclusions ahead of the main crack.
3
H. D. Williams, "The Fatigue of /3-brass," Philosophical Magazine, Vol 13, 1966,
pp. 835-854.
* Department of Mechanical Engineering, Massachusetts Institute of Technology,
Cambridge, Mass.
5
H. W. Liu, "Fatigue Crack Propagation and Applied Stress Range—An Energy
Approach," Transactions, Am. Society Mechanical Engrs., Vol 85D, 1963, pp. 116-
122. Downloaded/printed by
University of Washington (University of Washington) pursuant to License Agreement. No further repro
If growth is produced only by such damage, there should be a second-

power dependence of crack growth rate on crack length.6 This is not
inconsistent with the 1.5-power dependence of growth rate on crack
length for the macroscopic growth rate shown in Fig. 21, since the
macroscopic rate in reality depends on both the striation mode and the
inclusion-nucleated mode of growth. These ideas may be put in more
quantitative form as follows:
1. Striation mode—If the fatigue cracks grow simply by kinematically
irreversible plastic flow in the neighborhood of the tip, with no other
FIG. 21—Rates of crack growth in two 7178 aluminum alloys (taken from
Pelloux [55]).
fracture mechanism, one would expect the growth per cycle to be some
fraction of the crack opening displacement at the crack tip. Hult and
McClintock7 have shown that in shear the extent of the plastic zone due
to reversal can be calculated from the equations for monotonic loading
with the applied stress range in place of the applied stress and the stress
range for reversed plastic flow (twice the flow stress) substituted for the
yield strength. Assuming that similar relations hold for the tensile case,
one finds the extent of the plastic zone Ra in terms of the crack half
6
F. A. McClintock, "On the Plasticity of the Growth of Fatigue Cracks,"
Fracture of Solids, Interscience Publishers, New York, 1963, pp. 65-102.
7
J. A. H. Hult and F. A. McClintock, "Elastic-Plastic Stress and Strain Dis-
tributions AroundbySharp
Copyright ASTMNotches
Int'l (allUnder
rights Repeated
reserved);Shear,"
Mon DecIX6 7Congres
14:40:45Inter-
EST 2015
national de Mechanique Appliquee, Actes 8, 1956, pp. 51-58.
University of Washington (University of Washington) pursuant to License Agreemen
length c, the applied tensile stress amplitude <ra and the tensile strength,
TS (as a measure of flow stress):
The crack opening displacement has been given for the shear solution by
McClintock and Irwin.8 Again changing to the stress range and, by
analogy, to tension, one obtains the flank-to-flank displacement 2uy at
the crack tip due to a load reversal in a material with modulus E:
If the rate of growth is related to the displacement by a proportionality

constant C, one obtains
Perhaps some hint of C, the fraction of the crack opening that con-
tributes to crack growth, can be obtained theoretically by considering
the instability in uniaxial tension.6 Consider initial differences in cross-
sectional area, leading to a difference a£ in strain amplitude for two adja-
cent elements of the specimen. The rate at which this inequality in strain
builds up is a measure of the rate at which the geometry changes with
cycling. This rate was expressed in terms of the applied true stress S, the
rate of strain hardening, h (assumed linear), and a Bauschinger coeffi-
cient b, which gives the loss in flow stress on stress reversal per unit
strain in the previous half cycle. Two extreme cases may be considered.
Suppose that steady-state hysteresis loops have been attained; then h =
b and the rate of increase of strain inequality per cycle is given by
If, as another limiting case, it is assumed that the strain amplitude at

the crack tip is building up rapidly enough so that the Bauschinger effect
may be neglected (b « h), the rate of increase of strain inequality per
cycle is given by
where ea is the strain amplitude.

In either case the rate of increase of strain inequality becomes infinite
as the applied stress approaches the rate of strain hardening, as is ex-
pected for necking under uniaxial tension. At a crack tip, a reinforcing
8
F.Copyright
A. McClintock
by ASTMandInt'l
G. (all
R. Irwin, "PlasticityMon
rights reserved); Aspects
Dec 7of14:40:45
FractureEST
Mechanics,"
2015
Fracture Toughness Testing and Its Applications, ASTM STP 381, Am. Soc. Test-
ing Mats., 1965, pp. 84-113.
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh 173
effect of surrounding material will prevent extreme necking, but it

does seem possible that the net^hapejcharjês^j^Md-heôLihe.order
of magnitude of the crack opening displacement, so the coefncientj? of
EqJ3 could be_ofjh^_order..Qj_unily>.As shown in Fig. 21, Eq 3 gives
the observed striation spacing if the coefficient C is taken to be %>
a not unreasonable value.
2. Re-nucleation mode—The rate of growth due to inclusion fracture
under low cycle fatigue may be calculated from a Coffin low-cycle dam-
age rule, neglecting both the striation growth of the main crack and the
cycling required for nucleated cracks to join the main crack. McClintock6
derived the growth rate for such a case in terms of the extent of the
plastic zone Ra and the following parameters of the material:
ps = structural size (here the inclusion spacing),
yy = yield strain in shear, and
7/p = plastic strain at fracture under monotonically increasing shear.
Substitution of Eq 1 for the extent of the plastic zone and again con-
verting to tension by analogy gives:
where e/p is the true strain at fracture in monotonic tension.

The spacing between inclusions can be estimated from the inclusion
density found in Fig. 1 of Pelloux [55]. Let the density of inclusions
on a metallographic section be 1//TO2 = N/A. Then since the crack will
deviate to go through the inclusions, the fracture surface density will
be greater than the density on the metallographic section by roughly the
ratio of inclusion spacing to diameter:
The inclusion spacing is therefore
which gives p = 0.00059 and 0.00044 in. for the Alloys A and B, re-
spectively. For Alloy B, this number corresponds closely to Pelloux's
electron fractographs, but the difference between the alloys is much
less than expected from the volume fraction of inclusions, which were
reported to differ by a factor of 13. Other properties needed for Eq 5
are E = 107 psi and TS = 87,000 psi for both alloys. For lack of better
information, the fracture strain is assumed to be e/p = 0.40 for both
alloys. The resulting predictions for crack growth rates are shown in
Fig. 21.
Considering the numerous assumptions and the approximations made,
as well as the lack of empirical coefficients, the agreement is remarkably
close Downloaded/printed
for both theby striation and the re-nucleation modes of growth.
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized
It would be of interest to compare this analysis with other data to see

how general the agreement is, and how the theory must be modified to
take into account other mechanisms which may occur.
Dr. Laird—Professor McClintock has demonstrated most encouraging
agreement between Peloux' measurements of crack propagation and
theory based on both crack propagation models, namely, the plastic
blunting process and the renucleation mode by "damage." It should be
pointed out, however, that the renucleation model, where the Coffin-
Manson law of low cycle fatigue is used for a fracture criterion, is very
artificial. Since failure in low-cycle fatigue generally takes place by nu-
cleation of cracks at a free surface and their subsequent propagation
by the plastic blunting process,9-10 it does not seem realistic to use this
law as the basis of a different model. The agreement between theory and
experiment in this case is therefore probably fortuitous.
One must allow, however, the difficulty of such problems in fracture
mechanics and the expedient of simplification through extremes. For the
renucleation model, an extreme case would require a very dirty material
having a low fracture strain in a tension test. Such materials do fail by
a true "delayed static fracture" in high strain fatigue, that is, by the
opening of holes round inclusions and their linking up by plastic flow,
but only when the applied cyclic strain is large enough to cause failure
in less than about 10 cycles.10 This kind of fracture occurs throughout the
cross section of the material, and it therefore seems extremely unlikely
that the relatively small region at the tip of a crack would contain
sufficient inclusions to fail by this means, as required by the renucleation
theory. In practice, any inclusion would simply promote the plastic
blunting process by providing a linking hole, thereby increasing the rate
of crack propagation in the manner indicated by McClintock's treat-
ment of the plastic blunting process.
G. A. Miller11 (written discussion}—Dr. Laird favors the plastic blunt-
ing process to crack propagation along subgrain boundaries as an
explanation of fatigue-fracture topography. However, those authors
who proposed the model for subgrain cracking did not do so to provide
an explanation of fracture topography, but rather as a rationale for
the observation that fatigue cracks in high stacking-fault energy y
materials do appear to follow subboundaries.12-16 I agree with Laird's
9
C. Laird and G. C. Smith, "Crack Propagation in High Stress Fatigue,"
Philosophical Magazine, Vol 7, 1962, pp. 847-857.
10
C. Laird, "Studies of High Strain Fatigue," Ph.D. thesis, Cambridge University,
Cambridge, 1962.
11
Engineer, Bethlehem Steel Corp., Bethlehem, Pa.
12
D. H. Avery and W. A. Backofen, Fracture of Solids, John Wiley & Sons,
Inc., New York, 1963, pp. 339-382.
13
J.Copyright
Holden, by'The Formation
ASTM Int'l (allof Sub-grain
rights Structure
reserved); by 7Alternating
Mon Dec Plastic
14:40:45 EST 2015
Strain," Philosophical Magazine,
Downloaded/printed by Vol 6, 1961, pp. 547-558.
observation that any explanation based on subboundary cracking is

complicated because a dislocation boundary does not necessarily
constitute a plane of weakness. On the other hand, weakness per se
may not be involved. Rather, the boundary may serve to constrain
plastic deformation. Perhaps the high dislocation density in the bound-
aries, about lOVcm2, inhibits long-range dislocation motion to the
extent that fracture and plastic deformation are brought in closer com-
petition.
Another point at issue relates to Laird's interpretation of the influence
of stacking-fault energy on fatigue-crack propagation. He argues that
the lower growth rate for low y materials is primarily the result of a
higher work-hardening capacity, rather than an absence of substructure.
Now the question is whether Laird's argument relates to static or cyclic
hardening. A separation of the effects of substructure and cyclic harden-
ing might not be straightforward. One problem is that the dependence of
cyclic hardening on 7 is a function of the strain amplitude. For low-
amplitude, cyclic hardening and the propensity to form substructure
both increase with7.12-17 However, for high-amplitude cycling, substruc-
ture formation is still favored by high 7, but now hardening is less as
7 increases.18 Another difficulty is that for low 7 materials cyclic hard-
ening depends on strain history, whereas cyclic hardening of high 7
materials is somewhat insensitive to strain history.18 It is likely that as the
crack approaches a given volume of material the strain amplitude will
increase, and the strain history will therefore comprise both high- and
low-amplitude cycles. However, since there is little detailed information
on the actual strains in the vicinity of a crack, there is no basis yet for
concluding that the improved resistance to crack growth for low 7
materials is the result of higher cyclic work hardening. A judgment about
the problem must wait until cyclic hardening can be evaluated for actual
growth conditions.
Recently, McEvily and Johnston17 have investigated the relation be-
tween crack-growth resistance and static strain hardening. They consider
propagation to be a unidirectional process of plastic instability, and the
implication is that cyclic straining prior to the arrival of the crack can be
neglected. Their expression for growth rate is:
14
G. Y. Chin and W. A. Backofen, "Some-Metallographic Observations on the
Fatigue of Aluminum Bicrystals," Journal, Institute Metals, Vol 90, 1961, pp. 13-17.
15
J. C. Grosskreutz and P. Waldow, "Substructure and Fatigue Fracture in
Aluminum," Acta Metallurgica, Vol 11, 1963, pp. 717-724.
18
G. A. Miller and W. A. Backofen, Fatigue—an Interdisciplinary Approach,
Syracuse University Press, Syracuse, 1964, pp. 59-62.
17
A. J. McEvily and T. L. Johnston, 'The Role of Cross-Slip in Brittle Fracture
and Fatigue," International Conference on Fracture, Sendai, September, 1965.
18
C. E. Feltner and C. Laird, "Cyclic Stress-Strain Response of F.C.C. Metals,"
Fatigue Symposium,byPhysical
Copyright ASTM Metallurgy Soc., Am.
Int'l (all rights Inst. Mining,
reserved); Mon Metallurgical &
Dec 7 14:40:45 EST
Petroleum Engrs, Detroit, 1965.
University of Washington (University of Washington) pursuant to Licen
176 hhhhhhhhhhhhhhhhhhhhhhhh
where:
1
growth rate,
dN =
a = gross section stress,
f* crack length,
^—
<rv = yield stress,

0U = ultimate tensile strength,
fu = strain at maximum load, and
E = Young's modulus.
McEvily and Johnston argue that the beneficial effects of strain
hardening are incorporated in the expression for KI through [(ay + <r«)/
2]eu , the area under the stress-strain curve. If static properties are
a useful indication of crack-growth resistance, then KI should de-
crease as y is lowered. A useful test of the expression for KI is provided
by the experimental results of a study by Miller et al19 in which growth
rate was measured for copper-base alloys with a wide range of y. From
their experimental results dc/dN = Kep2, where ep is the plastic strain
amplitude. If ep2 and <r*c2 are equivalent representations, then K and KI
should be proportional if both depend on static properties only. K and
KI are presented in the table below where eu has been replaced by the
true strain at fracture, e/ , computed from the tensile reduction of area.
The constant 1.65 was chosen so that K = KI for copper, y' is the ratio
of y for any alloy to that of copper.
y', y alloy K 1.65 X 1021

Material hhhhh
•y copper XKi
Cu 1.00 3.5 2.12 3.50

Cu-lONi 1.04 6.5 4.16 6.87
Cu-20Ni 0.96 4.6 2.66 4.39
Cu-30Ni 0.74 4.3 1.24 2.04
Cu-2Al 0.50 3.9 2.26 3.73
Cu-4Al 0.29 1.1 1.28 2.11
Cu-SAl 0.04 0.7 0.51 0.84
On the basis of a comparison of the measured K values with those pre-

19
G. A. Miller, D. H. Avery, and W. A. Backofen, "Fatigue Crack Growth in
Some Copyright byAlloys,"
Copper-Base ASTM Int'l (all rights Am.
Transactions, reserved); Mon
Institute Dec 7 Metallurgical
Mining, 14:40:45 ESTand
2015
Petroleum Engrs., Vol. 236,by1966, pp. 1667-1673.
Downloaded/printed
dieted from the expression for KI , KI appears to be an accurate de-

scription of the crack-growth resistance of a material. With the excep-
tion of Cu-4Al and Cu-30Ni, agreement between K and 1.65 KI is good,
and even for these two alloys the divergence between measured and pre-
dicted is a factor of only two.
In summary, these results support the view that tensile instability is
a useful model of the crack extension process and that strain history
prior to the crack arriving at a given location has little effect on growth
rate. Also, cyclic hardening during crack growth appears to bear a rela-
tionship to static strain hardening. However, the necessity for further
testing must be emphasized since the validity of KI expression is ques-
tionable at least for some of the high-strength steels.20
Dr. Laird—Dr. Miller defends the subgrain fracture mechanism of
fatigue crack propagation irrespective of the plastic blunting process on
the grounds that, in high stacking fault energy materials, cracks appear
to follow subboundaries.12'16'21 Since subgrain formation presumably
occurs when the plastic blunting process is operating, it is necessary to
reconcile the observations on which the respective models are based. It
is also important to notice that the types of observation giving rise to
these mechanisms are different in kind. On the one hand, those support-
ing the subgrain cracking model are indirect. The micrographs depicting
the evidence, and shown in Fig. 22 of footnote 12, Fig. 2c of footnote
14, Fig. la of footnote 16, and Fig. 14 of footnote 21 do not demon-
strate unequivocally that cracks are propagating along subgrain bound-
aries because the crack size is so gross in relation to the cells. One can
as easily claim that, in some regions, the cracks seem to cross subgrains.
In the case of the specimen surface micrograph,14 the cracks appear to
run too regularly parallel to the general slip band to be associated with
subgrain boundaries. Since subgrains are frequently equiaxed, especially
in fatigue,15 some cracking at a large angle to the plane of the general
slip band, as well as within it, would be expected if these boundaries were
indeed regions of weakness, but this was not observed. Furthermore, all
such observations were made post facto, and cannot account for the
possibility that subboundaries in the neighborhood of a crack could
escape to the crack surface, thus giving the appearance of subboundary
cracking. Since recent work22'23 has shown that such subboundaries are
frequently composed of dislocation arrays in glissile configuration, the
20
See p. 460.
21
J. C. Grosskreutz, "A Critical Review of Micromechanisms in Fatigue," Pro-
ceedings, 10th Sagamore Army Materials Research Conference, Syracuse University
Press, Syracuse, 1964, pp. 27-59.
22
E. E. Laufer and W. N. Roberts, "Dislocations and Persistent Slip Bands in
Fatigued Copper," Philosophical Magazine, Vol 14, 1966, pp. 65-78.
23 Copyright by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
C. E. Feltner and C. Laird, Scientific Reports, Nos. SL-66-93 and SL-66-115,
Downloaded/printed
Ford Motor Co., 1966. by
178 hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
probability that they will escape to the crack surface just subsequent to
passage of the crack and before the given volume of the material passes
into its stress shadow seems to be high. The other references cited by
Miller13- 15 contain evidence which is more indirect. Thus, in Holden's
work,13 cell structure was detected by an X-ray microbeam technique, and
no observations of cracking were made. The observations by Grosskreutz
and Waldow15 of thin foils stressed in tension while contained in an
electron microscope actually support the plastic blunting process rather
than the subboundary cracking model, since tearing of the foil oc-
curred by "slipping-off" in the region of subboundaries which supplied
the dislocations for the mechanism. No fracture, in the sense of atomic
bond rupture over an appreciable area, was involved.
On the other hand, the evidence for the plastic blunting process9'10
is entirely direct, and this model can explain the observations which
suggest subboundary cracking. For example, the micrograph given in
Fig. la of footnote 16, cited to support the subboundary cracking model,
shows much branching of minor cracks from the main one in a specimen
of copper. On the subboundary cracking model, such cracks could be
supposed to progress along available subboundaries. I also have observed
such branching phenomena in polycrystalline nickel10 and in a situation
where I knew from fracture surface observations that the plastic blunting
process was operating. An alternative explanation for the branching
cracks is, therefore, as follows. It frequently happens that the notch left
at a crack tip, Fig. 4 in the paper, when the compression stroke of a
fatigue cycle has been completed, is symmetrical with respect to the plane
of the crack. Subsequent propagation of the crack can therefore occur at
an equal rate in both notches, producing divergence of the main crack
for several cycles, and sometimes for many cycles. Eventually one diver-
gent crack dominates the other which drops into the stress shadow of the
first and becomes a dormant branch. The factors which promote this be-
havior are those which cause three-dimensional plastic deformation at a
crack tip and therefore symmetrical residual notches, that is, high stack-
ing fault energy. Even materials of "normally" low stacking fault energy
will behave in this manner if cycled at high temperature, for example,
the stainless steels studied by Johannsson24 at temperatures between 20
and 500 C showed a greater tendency for branching with increase of
temperature.
The cell structure normally observed in materials of high stacking
fault energy at a crack tip is a natural consequence of the large strain
amplitude of cycling in that region concomitant with the high stress con-
centration of the crack and the plastic blunting process. The subboundaries
may be expected to promote the plastic blunting process by provid-
24
A.Downloaded/printed
Johannsson,by"Fatigue of Stainless Steels at Constant Strain Amplitude and
Elevated Temperature,"
University Stockholm
of Washington (University Colloquium
of Washington) on Fatigue,
pursuant to License May,
Agreement. No further1955.
reproductions authorized.
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh 179
ing dislocation sources, rather than to exert a constraint on plastic defor-

mation as Miller has suggested. In this connection, it seems extremely
unlikely that the stress required for fracture, a relatively large fraction
of Young's modulus, would be attained without causing dislocations in
subboundaries to move, particularly in materials of high stacking fault
energy which are capable of dynamic recovery via cross-slip.
The second point at issue raised by Miller concerns relative rates of
crack propagation in materials of high and low stacking fault energy.
He reasonably states that as a crack approaches a given volume of ma-
terial, the localized strain amplitude will increase and the strain history
will therefore be comprised of both high- and low-amplitude cycles.
Miller's conclusion, however, that there is no basis for arguing im-
proved crack resistance in materials with low stacking fault energy be-
cause of their higher work-hardening capacity, is unacceptable for the
following reasons. Although there is little detailed information on the
actual strains in the vicinity of crack tip, it is known from observation
of sections of cracked specimens9 that the strain at a crack tip is between
one and two orders of magnitude greater than the overall strain applied
to the specimen. Thus, even if a specimen is cycled at a low strain, the
strain at its crack tip is in the high amplitude regime. In a current in-
vestigation of cyclic strain hardening and softening in materials with both
low and high stacking fault energy,23 Feltner and Laird have demon-
strated by phenomenological measurements and structural observations
that low- and high-strain fatigue hardening is mechanistically similar to
that which occurs respectively in Stages I and II of the unidirectional
deformation of a single crystal. They have also shown that, when ap-
plied cyclic strains are sufficiently large, a situation analogous to Stage
III develops. At cyclic strain amplitudes giving Stage II deformation,
rates of hardening are approximately equal in materials of high and
low stacking fault energy.18'23 If the strain amplitude at a crack tip re-
mained at such a level, equal rates of crack propagation would therefore
be expected in both types of material. The strain in a volume of material
approached by a crack is increasing continuously, however, to a very
large value, and eventually the deformation mechanism will change to
that typical of Stage III. It is the difference in the Stage II to Stage III
transitional properties of materials with high and low stacking fault
energies which, I am suggesting, controls the relative crack propagation
rates in these materials. Where the stacking fault energy is high, the
Stages II-III transition occurs at a strain lower than that in materials
with low stacking fault energy. Less deformation will therefore be occur-
ring in the latter type of material at a crack tip and the crack propagation
rate (assuming the plastic blunting process) will be lower. I accordingly
conclude thatbythere
Copyright is (all
ASTM Int'l a sound basis Mon
rights reserved); forDec
ascribing
7 14:40:45 the improved crack
EST 2015
growth resistance ofby such materials to their higher cyclic (and static)
Downloaded/printed
University of Washington (University of Washington) pursuant to License Agreement. No further reproduction
work hardening. More work is required, of course, to prove this theory in

secure fashion.
The above considerations are entirely consistent with the other con-
clusions Miller derived from his application of the McEvily-Johnston17
equation to certain crack propagation data. First he noted that the re-
sults support a tensile instability model of crack extension which the
plastic blunting process is. Second, they indicate a relationship between
cyclic and static hardening which Feltner and Laird have demon-
strated directly. Third, Miller's conclusion that "strain history prior to
the crack arriving at a given location has little effect on growth rate"
also follows when we consider, on the one hand, the phenomenon of
work softening in fatigue which tends to obliterate unidirectional strain
history and, on the other, that very high strains, whether cyclic or unidi-
rectional, at a crack tip or in bulk, overcome the effects of previous
strain.23
M. R. Achter1
Effect of Environment on Fatigue Cracks
REFERENCE: M. R. Achter, "Effect of Environment on Fatigue

Cracks," Fatigue Crack Propagation, ASTM STP 415, Am. Soc. Testing
Mats., 1967, p. 181.
ABSTRACT: In this review of the fatigue of metals in controlled gaseous
environments, particular emphasis is placed on the mechanism of crack
propagation as it is affected by the test variables. The crack growth
rates of some metals are accelerated more by oxygen than by water
vapor, while for others the reverse is true. Increases of cyclic frequency
and of stress decrease the magnitude of the effect of environment. It is
generally agreed that the mechanism is more an increase of the rate of
crack propagation than of crack initiation. Of the two explanations pro-
posed, the process of corrosive attack of the crack tip is favored over
that of the prevention of rewelding of crack surfaces by the formation
of oxide layers. Curves of fatigue life, or of crack growth rate, versus
gas pressure show regions of little or no dependence, connected by a
transition region of steep slope. In a quantitative treatment of the shape
of the curve, the significance of the location of the transition region is dis-
cussed.
KEY WORDS: fatigue (materials), environment, corrosion, oxidation,
cracks, temperature, adsorption, crack propagation
For over thirty years, since the first studies by Gough and Sopwith
[I],2 it has been known that air is a corrosive medium in which the fa-
tigue life of metals can be shorter than in vacuum. Lately, an impetus was
given to the study of the details of this phenomenon by the demonstration
that traces of oxygen or water vapor in the test chamber can have as much
effect as air at atmospheric pressure. The demonstration that water vapor
could have a substantial influence on the fatigue strength raised the pos-
sibility that variations in atmospheric humidity could be contributing to
the scatter in experimental results. By means of direct measurements of
crack propagation rate and by means of post-test examination, it is
pretty well agreed that the reaction of the gases with the metal surface
increases the rate of crack propagation.
1
Head, High Temperature Alloys Branch, Metallurgical Div., Naval Research
Laboratory, Washington, D.C.
2
paper.
181
Although the variables in the process have been carefully studied in an

impressive number of investigations, the mechanism of the effect is not
known with any degree of confidence. The uncertainty derives from the
necessity of describing the detailed reaction of the reactive gas with the
atoms at the crack tip, which is usually at the end of a fine capillary chan-
nel. Therefore, a complete description of the phenomenon involves not
only adsorption of the gas on the crack surfaces but also the diffusion
from the specimen surface to the tip.
In this paper, the work undertaken to investigate the mechanism in
gaseous environments will be reviewed and a model will be discussed
which can serve as a basis for the design of experimental programs.
Oxygen or Water Vapor

A number of investigators have addressed themselves to the task of
determining whether it is oxygen or water vapor which shortens the fa-
tigue life. Wadsworth and Hutchings [2] reported that the life of copper
in a vacuum of 7 X 10~6 torr was 20 times longer than in air. In dry air
the ratio was about 10 but in wet oxygen it was approximately 30. How-
ever, in water vapor alone the life was about the same as in vacuum. They
concluded that oxygen is the effective component in air and that water
vapor has no effect of its own but increases that of oxygen. Snowden
[3], working with lead, also found that oxygen was the damaging compo-
nent, but that its effect was not increased by water vapor; in fact, the life
was about twice as long in wet oxygen as in dry.
In four investigations with steel it was reported that water vapor re-
duces the fatigue life. In a moist argon atmosphere Mantel et al [4] found
that the life of 52100 steel in the hardened condition was statistically al-
most ten times shorter than in dry argon. Dahlberg [5], Fig. 1, reported
that the rate of fatigue crack propagation of 4340 steel was increased by
an order of magnitude by the addition of water vapor to dry air. Shives
and Bennett [6], reported that moisture reduced the life of a low alloy
steel, brass, a titanium alloy, and a magnesium alloy.
Frost [7] has reported results which are at variance with those just
discussed. In Fig. 2 he shows that cracks in mild steel, tested in push-
pull fatigue, grow faster in air than in tap water. He reasons that tap
water excludes oxygen from the crack tip and, since the rate was the same
in oil as in water, that water vapor has no effect and that oxygen is re-
sponsible for the faster rate in air. When, however, he resorted to a load-
ing cycle with a mean stress, so that the cracks never close, he found the
same life in water as in ah*. From this result he concludes, since it was
previously shown that oxygen is excluded by water, that the water vapor
reduced the life with respect to that in oil. He also found a longer life in
dry air than inby normal
Copyright air.(allIt rights
ASTM Int'l is difficult
reserved);toMon
reconcile the results
Dec 7 14:40:45 of Frost
EST 2015
on push-pull fatigue with
Downloaded/printed by those of Shives and Bennett [6] and Mantel et
ACHTER ON EFFECT OF ENVIRONMENT ON FATIGUE CRACKS 183
al [4], who used reversed bending fatigue in which the cracks are closed
during each cycle. I wonder about the effectiveness of oil as a barrier to
water vapor since the lives in that fluid are no more than a factor of two
greater than in air. It would be illuminating to be able to see curves, such
as in Fig. 2, comparing humid and dry air, using completely reversed
stress.
With aluminum, there is agreement that water vapor shortens the life.
Wadsworth and Hutchings [2] observed slightly shorter lives in water
vapor or humid air than in dry air. Broom and Nicholson [8] compared
the life of an aluminum alloy in vacuum with that in gas atmospheres. In
FIG. 1—Effect of humidity on the crack growth rate of 4340 steel tested in
tension-tension loading [5].
Table 1 it is shown that the longest life is obtained when a cold trap is
used in a vacuum of 2 X 10~6 torr. In the same vacuum, but without a
cold trap, the life is reduced by a factor of two. In water vapor, the life is
lower by almost an order of magnitude. In atmospheric pressure dry oxy-
gen, however, the life is about the same as in the best vacuum. That water
vapor accelerates the rate of crack propagation of an aluminum alloy was
demonstrated by Bradshaw and Wheeler [9], in Fig. 3. In normal air, or
in water vapor at 15 torr, the rate is roughly an order of magnitude
higher than in a vacuum of 4 X 10-8 torr, while in dry oxygen it is only
slightly higher than in vacuum.
To explain the effectiveness of water vapor and the lack of it for oxy-
gen, Broom and Nicholson [8] advanced the proposal that the mechanism
involves the diffusion by
Downloaded/printed inward of hydrogen ions produced by the reaction
University of Washington (University of Washington) pursuant to License Agreement. No fur
FIG. 2—Crack growth on edge-notched mild steel plate specimens tested at

±5 tons /in2 in four environments [7].
TABLE 1—Fatigue of electropolished aluminum-4 per cent copper alloys under

various conditions [8].
(Stress ±12 tons/in2)
Temp, of Approx.
Test Environment Trap, Pressure, Tests Logtf Deviation cycles
degC mm Hg
Air 20 760 33 4.9891 0.1556 0.98 X 105

Vacuum 20 2 X ID-6 6 5.5719 0.1166 3.7
Vacuum . -196 2 X 10-« 12 5.9486 0.1228 8.9
Nitrogen . -196 760 12 5.7993 0.1229 6.3
Oxygen . -183 760 7 5.9252 0.2191 8.4
Hydrogen . -196 760 12 5.5010 0.2044 3.2
Nitrogen . -78 760 6 5.4120 0.0509 2.6
Oxygen . -78 760 6 5.3478 0.1228 2.2
Water vapor 20 17.5 4 4.773 0.1822 0.59
of water vapor with the metal. Such a mechanism is supported by the

demonstration by Bennett [10] that gas bubbles were formed under a tape,
applied to the surface of an aluminum specimen, only in a humid atmos-
phere. Similarly, Stubbington and Forsyth [11] detected gas bubbles
evolving from the surface of an aluminum alloy under salt water only
when it was being fatigued.
It is too soon to attempt to generalize about the relative effect of oxygen
and water vapor. It is well known that many metals, including steel and
aluminum, react with water vapor and are sensitive to hydrogen. But the
results on copper [2], that water vapor is effective only when combined
with oxygen, defy explanation as yet. What is needed are more investiga-
tions Copyright
in whichbythe
ASTMfatigue
Int'l (alllives
rightsin dry oxygen,
reserved); Mon Decand in water
7 14:40:45 EST vapor,
2015 are
compared to those in vacuum.
ACHTER ON EFFECT OF ENVIRONMENT ON FATIGUE CRACKS 1 85
FIG. 3 — Crack rate versus half crack length for DTD 5070A tested in tension-
tension loading in air, oxygen, and vacuum [9].
Corrosive Attack or Rewelding

In a study of the fatigue of copper, using push-pull loading, Thompson
et al [12] reported life to be about five times longer in nitrogen than in air.
As one possible mechanism to explain the shorter life in air, they sug-
gested that the oxide layer, formed there, might prevent rewelding of the
crack Copyright
surfacesbyduring
ASTM the
Int'l compressive part of
(all rights reserved); theDeccycle.
Mon Other EST
7 14:40:45 investiga-
2015
tors have favored thebycorrosive attack mechanism; oxygen reacts with
Downloaded/printed
the atoms at the tip of the crack and weakens the bonds there, making
crack propagation easier. It would appear that there are enough objec-
tions to the rewelding hypothesis to obviate the need for pursuing it fur-
ther. For one, it is known that oxygen accelerates crack propagation in
tensile creep where there are no compressive stresses [13]. Secondly,
even in a vacuum of 10~6 torr, the surface is covered with gas in about
one second, so that only the atoms at the very tip of the crack could par-
ticipate in the rewelding.
However, to resolve the question, several investigators have designed
experiments to distinguish between these mechanisms. Laird and Smith
[14] performed push-pull fatigue on copper and aluminum in a vacuum
of 5 X 10~6 torr and examined the specimens after stopping the test
during the compression cycle. Because they found no evidence for
welding of the cracks, they favor the corrosive attack hypothesis. Mar-
tin [15] used an oscilloscope to monitor the stress-strain behavior during
push-pull fatigue in a vacuum of 1 X 10~6 torr and found some evidence,
which he interprets as indicating a possible reweldment of the crack
surfaces in stainless steel, copper, and aluminum, but a negligible amount
in 1018 and 1113 steel. He does not, however, consider that there was
enough rewelding to account for the effect of environment on the fatigue
life.
Frost [7] attacked the problem by using two types of loading in ten-
sion fatigue. In one series he used fully reversed stress, and in the other
he used a tensile mean stress. In the first, the crack surfaces are com-
pressed during each cycle, and in the second the cracks were always open.
In a series of involved, not easily understood experiments, he interpreted
the dependence of the environmental effect on the type of loading to
mean that there was some rewelding of the crack surfaces, when the load-
ing cycle involves a compressive stress. Bradshaw and Wheeler [9], who
called the results of Frost unexpected, set out to determine the effect of
the type of loading in tensile fatigue. Whether with a tensile mean stress,
or with a fully reversed stress, they found the same effect of water vapor
on the rate of crack propagation in aluminum. Therefore, they favor the
mechanism of hydrogen generation and diffusion.
It would appear that the weight of evidence is against the rewelding
hypothesis. However, it would be worth while to clear up the uncertainty
by means of an experiment carried out under carefully controlled condi-
tions. The ratios of life in vacuum to that in dry oxygen should be com-
pared with and without a tensile mean stress. If rewelding is involved, a
larger ratio would be expected when there is a compressive stress in the
loading cycle.
Crack Initiation and Propagation
Since the fatigue process is generally considered to be separated into
two stages, theofinitiation
University and
Washington propagation
(University of cracks,
of Washington) it was
pursuant to natural to look No furthe
License Agreement.
for two separate effects of environment. Broom and Nicholson [8] inter-
rupted tests on aluminum at various stages and found cracks occurring
sooner in air than in vacuum, an observation which led them to conclude
that crack initiation was affected more than growth. From the finding
that crack propagation rate in an aluminum alloy was affected more by
environment than was the fatigue life, Bradshaw and Wheeler [9] arrived
FIG. 4—Growth of fatigue cracks in air and in vacuo, in annealed and in

work-hardened nickel. Push-pull loading at ±50,000 psi and 72 cpm [14].
at the opposite conclusion, that the major effect was on the rate of propa-
gation.
A careful measurement of crack depth in nickel was performed by
Laird and Smith [14]. In Fig. 4, it is demonstrated clearly that, in their
tests, cracks are present in the early stages, in a vacuum of 5 X 10~6
torr, and that the effect of environment on the life time can be attributed
to a difference in crack propagation rate.
Christensen [16] measured the rates of crack propagation of an alumi-
num alloy and found increasing rates with increases of pressure in short-
time tests. But, in long-time tests, after prolonged pumping to outgas the
system, he reported
Copyright by ASTMthat
Int'lthe
(all vacuum environment
rights reserved); Mon Dec no longer EST
7 14:40:45 suppressed
2015
crack University
growth, ofa Washington
result he termed
(Universitythe
of "low-torr
Washington) paradox." To verify
pursuant to License this
Agreement. No further
result, Bradshaw and Wheeler [9] also resorted to prolonged pumping

on aluminum but did not confirm the "paradox." Neither was this "para-
dox" observed on work with nickel [22] and, unless it is confirmed, it
must be regarded the result of some unknown peculiarity of the equip-
ment used.
In a different type of experiment, Wadsworth [17], and Frost [7], found
that the fatigue limit for steel is higher in a neutral environment than in
an aggressive one. Cracks were found in the unfailed specimens fatigued
in the neutral environment at the fatigue limit. At this same stress the
specimens in air failed. From these results, it is concluded that the stress
required to initiate a crack is not affected by environment but that the
rate of propagation is.
FIG. 5—Change in resonant frequency for notched specimens of Type 316

stainless steel at 800 C, tested in reversed bending at a strain of 0.071 per cent
[22].
A different result was obtained by Grosskreutz and Bowles [18] on

single crystals of aluminum. In an atmosphere of low pressure oxygen,
the rate of surface strain hardening is higher than in a vacuum of 10 ~ 9
torr. In later work, Grosskreutz [79] reports that crack nuclei are found
earlier in oxygen than in vacuum. This finding, I think, may explain the
slight effect of oxygen reported by Bradshaw and Wheeler [9] on alumi-
num. Shen et al [20] attribute the shorter life of aluminum in air than in
vacuum to the trapping of dislocations under an oxide film, which then
form a debris layer. I wonder, however, how to reconcile this explana-
tion with the evidence, mentioned earlier, that the reduction of life in air
is the result of hydrogen diffusion.
At elevated temperature it is difficult to observe crack depths directly,
and we, therefore, had to develop a method for accomplishing it. It con-
sists of a fatigue
Copyright by machine (all in
ASTM Int'l[21] which
rights the frequency
reserved); Mon Dec of7 vibration is au-
14:40:45 EST 2015
tomatically kept at resonance
Downloaded/printed by and recorded. Curves, such as in Fig. 5,
showing the decrease in frequency with time, are a convenient means of

determining the relative rates of crack propagation at 800 C in vacuum
and in a partial pressure of oxygen [22]. Crack depths, measured on inter-
rupted specimens, are given in Table 2 and demonstrate the faster growth
in oxygen. A specimen interrupted after 3 hr in vacuum had cracks too
small to be measured, attesting to the insensitivity of crack initiation to
environment.
There is no doubt that the major effect of environment, at least for
TABLE 2—Measurements of fatigue crack growth in stainless steel in vacuum

and oxygen.
Vacuum, 7 X 10~7 torr Oxygen, 1 torr
Time in Crack Length, Frequency Time in Crack Length, Frequency
Test, hr in. Decrease, cps Test, hr in. Decrease, cps
17 0.010 0.03 1.7 0.032 0.35

42 0.030 0.35 2.8 0.040 0.60
47 0.039 0.60
FIG. 6—Fatigue of nickel in air and in vacuum at 1500 F. Tested in reversed

bending [23].
most metals so far examined, is on the rate of crack propagation. With

regard to the sensitivity of the rate of initiation, it could be a matter of
the definition of a crack nucleus. As the available methods of detection
become more sensitive, the beginnings of cracks are found at earlier
stages. It then becomes a matter of deciding whether a localized site of
deformation is a crack.
Effect of Temperature
At elevated temperature and low rates of crack propagation, the ef-
fect of environment
Copyright by ASTMmay berights
Int'l (all reversed; theMon
reserved); fatigue
Dec 7 strength may
14:40:45 EST be greater
2015
in airDownloaded/printed
than it is in vacuum
by or a neutral environment. In Fig. 6, it is seen
190 hhhhhhhhhhhhhhhhhhhhhhhhh
that nickel [23] at 1500 F has a longer fatigue life in air than in vacuum
at the lowest rate of crack propagation. Although tests with type 316
stainless steel and Inconel X, Fig. 7, do not show this reversal, there is
one indicated by extrapolation for very long lifetimes. Nachtigall and
his co-workers [24], also at 1500 F, do show an actual reversal for In-
conel 550 but not for S-816 except by extrapolation, Fig. 8.
We have explained the reversal in terms of two competing processes
[23]. Surface adsorption of gas makes crack propagation easier, but bulk
oxidation hardens the metal and delays failure. At elevated temperature,
if the test is long enough, the metal may be stronger in air than in vac-
uum. In a short-time test, where there is no opportunity for extensive
oxidation, the reverse is true.
FIG. 7—Fatigue of Type 316 stainless steel and Inconel X at 1500 F in air and
in vacuum tested in reversed bending [23].
Effect of Frequency and Strain

If the effect of environment can be likened to corrosion fatigue, it
would be expected that cyclic frequency and strain would affect the re-
sults in the same way. An increase of each would shorten the life and,
consequently, reduce the time available for corrosive attack. Those in-
vestigators who have considered the influence of frequency have found a
larger environmental effect at lower speeds [6,20,28].
This result necessitates a word of caution about the interpretation of
work in which coatings, such as oil, are evaluated for their protective
properties. These tests are usually conducted at high frequencies to
shorten the time of the experiment. Under service conditions, where long
times are involved, diffusion through the coating e£ oxygen and water
vaporCopyright
might significantly reduce
by ASTM Int'l the protective
(all rights reserved);properties
Mon Dec 7of14:40:45
oil coatings.
EST 2015
Testing at higher strains
Downloaded/printed by also reduces the effect of environment. From
University of Washington (University of Washington) pursuant to License Agreeme
(a) Ratio of cyclic to mean stress, 0.125.
(b) Ratio of cyclic to mean stress, 0.667.

FIG. 8—Axial tensile fatigue properties of S-816, a, and Inconel 550, b, at
1500 F in vacuum and air [24].
curves such as Figs. 2-5, it is evident that, when they are long, cracks
tend to grow at rates which are independent of environment.
Metallography of Cracks
In examining failed specimens for evidence of the mechanism of crack
propagation, it must be remembered that the lifetime in the neutral en-
vironment is usually longer than in a corrosive gas and that there is, there-
fore, more opportunity for damage to be introduced. Accordingly, in a
number of investigations
Copyright more
by ASTM Int'l (all surface rumplingMon
rights reserved); andDec
plastic deformation
7 14:40:45 EST 2015
have been observed after
Downloaded/printed by fatigue in vacuum than in a reactive gas [2,9,
25,55], Similarly, it has been reported by Wadsworth and Hutchings [2]

and by Shen et al [20] that there are more cracks when the testíng is
done in vaeuum than in air.
Four ínvestigators have reported that cracks, which are transcrystal-
line in a neutral environment, show a tendency to be along grain bound-
aries in a corrosive atmosphere [5,9,25,55], Clearly, this result demón-
strales the greater xeactivity of the grain boundaries with respect to the
FIO. 9—-Fatigue crack surface of 2024 T3 aluminum tested under a stress in-
tensity of 9000 psi "Vin, -» direction of propagation [27]. The striated surface rep-
resenis fatigue in air and the other, fatigue in vacuum.
bulk material. As a consequence of this property of grain boundaries, as

reported by Snowden [26], single crystals showed a smaller dependence
on environment than did polycrystals,
Microfractography shows promise of being able to contribute infor-
mation about the mechanism of the environmental effect Figure 9, rep-
resentative of preliminary results by Meyn [27], shows the change in the
appearance of the fatigue surface of an aluminum alloy when the atmos-
phere is cycled between air and vacuum. At the stress used the rate in air
was four times
Copyright by faster than(allinrights
ASTM Int'l vacuum. AtMon
reserved); lower
Dec stresses the
7 14:40:45 ratio
EST 2015was
higher.
Effect of Gas Pressure

Some of the most useful information for quantitative discussion of en-
vironmental effects is furnished by measurements of fatigue life, or
crack propagation rate, as a function of the pressure of the reactive gas.
Figure 10 shows the results of this application on copper, arrived at by
Wadsworth and Hutchings [2] who first determined the effect of gas pres-
FIG. 10—Variation of fatigue life of copper with air pressure, tested in re-
versed bending [2].
i
sure on the fatigue life. As the pressure of air is increased there is a
linear decrease in fatigue life. They also showed a linear plot for alumi-
num. Snowden [3,30], working with lead, was the first to publish an S-
shaped curve, Fig. 11, with plateaus at high and low pressures and a
transition region at approximately 10~2 torr of air. A suggestion of an
S-shaped curve was reported by Ham and Reichenbach [25], with the
transition at 10~3 torr.
Three other
Copyright investigators
by ASTM foundreserved);
Int'l (all rights S-shapedMoncurves for aluminum.
Dec 7 14:40:45 EST 2015 Brad-
shawDownloaded/printed
and Wheeler [9] by reported some very useful data. They plotted
crack growth rate versus the pressure of water vapor, Fig. 12, and deter-
mined that the transition pressure was at about 0.5 torr and increased
with increasing crack size. Shen et al [20], in Fig. 13, and Horden and
Reed [29] found the transition pressure for the fatigue life to be at 10~2
torr pressure of air.
At elevated temperature the shape of the S-curve is somewhat differ-
ent. At 300 C, Fig. 14, the curve of the dependence of life of nickel on
the pressure of oxygen has a finite slope at high pressure instead of a
level plateau [31]. The transition is at approximately 10~4 torr, and there
is only a suggestion of a low pressure plateau. For copper, there is no
FIG. 11—Variation of fatigue life of lead with air pressure. Tested in reversed
bending [3].
sharp transition region, and the high pressure plateau, if it exists, appears
to have been displaced to oxygen pressures higher than atmospheric. At
816 C, Fig. 15, the transition pressure appears to have been moved to
the 10~5 torr range. In the lO"1 torr range, the descending portion of the
curve is attributed to diffusion of oxygen in grain boundaries ahead of the
crack at this elevated temperature. The increase in fatigue life with in-
creasing oxygen pressure in the IO2 range is caused by oxidation
strengthening, as discussed previously. For type 316 stainless steel at
800 C, Fig. 16, the transition pressure is at about 10"1 torr [22]. In an
attempt to determine whether there is a low pressure plateau for nickel
at 300 C, the measurements were extended to lower pressures, as shown
in Fig. 17, but the fatigue life was still improving at the best vacuum ob-
tained, 7 X 10~9 torr. However, the curvature suggests the presence of a
plateau at still lower pressures.
FIG.by 12—Crack rate versus water vapor pressure for DTD 5070A at fixed half crack lengths, tested in tension-tension loading [9].
Downloaded/printed
FIG. 13—Variation in fatigue life of aluminum 1100 with air pressure, tested
in reversed bending at 50 C [20].
FIG.Copyright
14—Cycles-to-failure,
by ASTM Int'l (allN,rights
at 300 C asMon
reserved); a function of oxygen
Dec 7 14:40:45 EST pressure,
2015 p.
The plastic bending strainbyfor nickel is 0.170 per cent; for copper, 0.162 per cent,
Downloaded/printed
tested University
in reversedofbending [31].(University of Washington) pursuant to License Agreement. No further r
Washington
Discussion
A mechanism for the effect of environment on the propagation of
cracks would have to explain the features of the S-shaped curves of gas
pressure versus fatigue life or crack propagation rate. As yet, there are
not enough data available for the task, and some of the published data
are contradictory. In this section an attempt will be made to set up a
model which will explain as much of the data as possible, and which can
serve as a basis for designing future experiments.
FIG. 15—Cycles-to-failure, N, at 816 C as a function of air pressure, p, for

nickel tested in reversed bending, 0.146 per cent plastic bending strain [31].
The salient features of the S-shaped curves are the two plateaus, along
which the fatigue properties are independent of gas pressure and the
transition region where the properties are decreased with increasing pres-
sure. To explain the portion of the curve involving the transition region
and the high pressure plateau, my co-workers and I [31] and Snowden
[3] assumed a model which had the crack growth rate increasing with in-
creasing pressure, until a critical pressure was reached at which the sur-
face of the crack was saturated with gas, in a time equal to half the pe-
riod of vibration. Once a monolayer of gas was formed, there would be
no further effect of increases in gas pressure.
Using the elementary theory of gas kinetics, we calculated a critical
pressure, which
University was lower (University
of Washington than the observed valuepursuant
of Washington) for nickel, Fig. Agreement.
to License 14, No
by a factor of 50, and Snowden, Fig. 11, reported a factor of 103 in the
same direction. We advanced the suggestion that the discrepancy could
be accounted for by the sticking probability, and Snowden considered
the possibility that the necessity for diffusion along a capillary crack de-
layed the arrival of gas atoms and raised the observed critical pressure.
FIG. 16—Cycles-to-failure for notched specimens of Type 316 stainless steel

as a function of oxygen pressure, tested in reversed bending at 800 C and 10 cps
under a strain of 0.071 per cent [22].
Bradshaw and Wheeler [9] were able to refine the model further by
using their data on crack propagation rate, Fig. 12. They said that, at
the critical pressure, the gas must be delivered to the crack tip at a rate
just fast enough to cover the fresh surfaces, as they are being produced
by the advance of the crack front. They considered various crack geom-
etries to rationalize their data.
It appears
Copyright to
by me thatInt'l
ASTM the(all
impedance of theMon
rights reserved); crack
Dec to7gas diffusion
14:40:45 does
EST 2015
not change the value by
Downloaded/printed of the critical pressure as much as expected. Con-
sider the following model: The event which accelerates crack growth is
the reaction of water vapor with the intermetallic bonds at the tip of the
crack. From this, the critical pressure is that which delivers molecules of
gas to each intermetallic bond at a rate which is twice that at which
bonds are being broken.3 (Assume two molecules react with each bond.)
According to this scheme, gas is being used up only in the interval during
which the crack is moving. One does not know how to estimate this in-
terval; they used tensile fatigue with a mean tensile stress so that one half
FIG. 17—Cydes-to-failure for nickel as a function of oxygen pressure at 300 C,

tested in reversed bending under a plastic strain of 0.170 per cent Q A two batches
of specimens [22].
seems reasonable. Therefore, the effective crack propagation rate would
be twice that measured. As a result, the predicted critical pressure would
be doubled. In effect, the impedance of the crack is halved; during the
portion of the cycle that the crack is not moving, its volume is being re-
charged. This treatment would predict the same critical pressure as does
Bradshaw and Wheeler, except for the factor of two to account for the
effective growth rate. Bonds are broken along a line parallel to the
growth direction at a rate of 2da/dN/x per cycle, where da/dN is the
measured growth rate per cycle, and x is the lattice constant. The pre-
3
ADownloaded/printed
similar model wasbyused to calculate gas adsorption during creep [34].
dieted bombardment rate per lattice site is then obtained by multiplying

by 2/, where / i s 100 cps. Reading 2 X 10~6 cm/cycle from the lowest
curve in Fig. 12, one calculates 0.05 torr for the critical pressure to be
compared to the measured value of about 0.4 torr.
There is another test which I applied to their model. At faster rates of
crack growth the critical pressure should be increased proportionately.
Referring again to Fig. 12, the ratio of the highest to the lowest plateau
is about ten, to be compared to the observed ratio of critical pressures
of about five.
Of course, it is realized that the readings of the critical pressure are
subject to uncertainty. However, I believe that these calculations sup-
port the Bradshaw and Wheeler model. One cannot, however, conclude
that it holds true for other metals, tested in the same way or in reversed
stress. Further verification will await crack propagation rate data for
other materials. I think, however, that the model looks reasonable. Of
course, the fact that Snowden [30] did not observe a shift in critical pres-
sure with changes in frequency argues against this model. But judgment
will have to be suspended until tests can be conducted with larger differ-
ences in frequency.
It remains to consider the reason for the lack of a pressure dependence
along the low pressure plateau. Certainly, even at these low pressures,
gas atoms which hit the surface stick to it and cannot be pumped off, as
described by Schlier and Farnsworth [32]. They report adherence even at
10~10 torr. A possible explanation is that there is a subsurface crack
initiation mechanism, perhaps similar to the process described by Wood
et al [33], which controls the crack propagation rate when the pressure is
low. At the transition pressure, the surface mechanism of crack propaga-
tion becomes the dominant one. Arguing against this scheme is the ob-
servation by Bradshaw and Wheeler [9] that there is no tunneling ahead of
the cracks. However, it would be worth while to examine the cracks
carefully, below the surface, with the new, improved techniques available,
for evidence that the mechanism of crack propagation changes in going
from one plateau to the other.
More data on adsorption of gases on metals are needed for the inter-
pretation of the S-shaped curves. For example, one needs to know the
sticking coefficient to be able to compare the predicted and observed
critical pressure. More fatigue work at elevated temperature would be
helpful, since sticking coefficients are known to increase with tempera-
ture and should affect the results. In this respect, we wondered why the
transition in Fig. 16 appeared at such a high pressure, whereas an in-
creased sticking coefficient would predict a lower value. It is possible
that oxide formation in the cracks at 800 C restricts the access of oxygen.
Measurements at lower temperatures should resolve this point.
A consideration of the
Downloaded/printed by S-shaped curves of fatigue versus pressure of
gas makes it evident that one must exercise great care in the planning of
experiments. For one thing, if the variations in gas pressure are confined
to one of the plateau regions, one would get the false impression that the
metal under test is not affected by environment. For another, small
amounts of impurities in inert gas environments may have as much effect
as air at atmospheric pressure. Therefore, it is mandatory that careful
techniques be applied not only to the control of the composition of the
atmosphere but also to its measurement.
Acknowledgments
I wish to record my gratitude to my co-workers, P. Shahinian, H. H.
Smith, and R. L. Stegman, for helpful discussions.
References
[1] H. J. Gough and D. G. Sopwith, "Atmospheric Action as a Factor in
Fatigue of Metals," Journal, Institute of Metals, Vol 49, No. 2, 1932, pp.
92-112.
[2] N. J. Wadsworth and J. Hutchings, "The Effect of Atmospheric Corrosion
on Metal Fatigue," Philosophical Magazine, Series 8, Vol 3, No. 34, October,
1958, p. 1154.
[3] K. U. Snowden, "The Effect of Atmosphere on the Fatigue of Lead," Acta
Metallurgica, Vol 12, 1964, p. 295.
[4] E. R. Mantel, G. H. Robinson, and R. F. Thompson, "Influence of Atmos-
pheric Moisture on Fatigue of Hardened Steel," Metals Engineering
Quarterly, Am. Society Metals, Vol 1, 1961, p. 57.
[5] E. P. Dahlberg, "Fatigue-Crack Propagation in High-Strength 4340 Steel
in Humid Air," Transactions, Am. Society Metals Quarterly, Vol 58, 1956,
p. 46.
[6] T. R. Shives and J. A. Bennett, 'The Effect of Environment on the Fatigue
Strength of Four Selected Alloys," NASA CR-267.
[7] N. E. Frost, 'The Effect of Environment on the Propagation of Fatigue
Cracks in Mild Steel," Applied Materials Research, Vol 3, 1964, p. 131.
[8] T. Broom and A. Nicholson, "Atmospheric Corrosion—Fatigue of Age-
Hardened Aluminum Alloys," Journal, Institute of Metals, Vol 89, 1960-
1961, p. 183.
[9] F. J. Bradshaw and C. Wheeler, "The Effect of Environment on Fatigue
Crack Propagation," Applied Materials Research, Vol 5, No. 2, 1966, p. 112.
[10] J. A. Bennett, "Changes in the Influence of Atmospheric Humidity During
Fatigue of an Aluminum Alloy," Journal of Research, Nat. Bureau Standards,
Vol 68C, April-June, 1964, p. 91.
[11] C. A. Stubbington and P. J. E. Forsyth, "Some Corrosion-Fatigue Observa-
tions on a High-Purity Al-Zn-Mg Alloy and Commercial DTD 683 Alloy,"
Journal, Institute of Metals, Vol 90, 1961-62, p. 329.
[12] N. Thompson, N. Wadsworth, and N. Louat, "The Origin of Fatigue Frac-
ture in Copper," Philosophical Magazine, Vol 1, Series 8, No. 2, February,
1956, pp. 113-126.
[13] P. Shahinian and M. R. Achter, "Creep-Rupture of Nickel of Two Purities
in Controlled Environments," Proceedings, Joint International Conference on
Creep, 1963.
[14] C. Laird and G. L. Smith, "Initial Stages of Damage in High Stress Fatigue in
Some Pure Metals," Philosophical Magazine, Vol 8, 1963, p. 1945.
[/5] D.Copyright
E. Martin, "PlasticInt'l
by ASTM Strain Fatigue
(all rights in Air and
reserved); MonVacuum,"
Dec 7 14:40:45 EST 2015Am.
Transactions,
Soc. Mechanical Engrs.,
Downloaded/printed by Series D, Vol 87, 1965, p. 850.
[16] R. H. Christensen, "Fatigue of Metals Accelerated by Prolonged Exposure

to High Vacuum," Transactions, Am. Society Metals Quarterly, Vol 57, 1964,
p. 373.
[17] N. J. Wadsworth, "The Influence of Atmospheric Corrosion on the Fatigue
Limit of Iron-0.5% Carbon," Philosophical Magazine, Vol 6, No. 63, 1961,
pp. 397-401.
[18] J. C. Grosskreutz and C. Q. Bowles, "Effect of Environmental Gases on the
Surface Deformation of Aluminum and Gold in Fatigue," to be published
in RIAS Conference on Surface Effects.
[/9] J. C. Grosskreutz private communication.
[20] H. Shen, S. E. Podlaseck, and I. R. Kramer, "Effect of Vacuum on the
Fatigue Life of Aluminum," Acta Metallurgica Vol 14, 1966, p. 341.
[21] M. R. Achter, H. H. Smith, R. J. Riley, and R. L. Stegman, "Flexural
Fatigue Machine for High Temperature Operation at Resonance in Vacuum,"
The Review of Scientific Instruments, Vol 37, No. 3, March, 1966, pp. 311-
315.
[22] H. H. Smith, P. Shahinian, and M. R. Achter, unpublished research, U.S.
Naval Research Laboratory.
[23] G. J. Danek, Jr., H. H. Smith, and M. R. Achter, "High Temperature
Fatigue and Bending Strain Measurements in Controlled Environments,"
Proceedings, Am. Soc. Testing Mats., Vol 224, 1961, p. 1115.
[24] A. J. Nachtigall, S. J. Klima, J. C. Freche, and C. A. Hoffman, 'The Effect
of Vacuum on the Fatigue and Stress-Rupture Properties of S-816 and
Inconel 550 at 1500°F," NASA TN D-2898, June, 1965.
[25] K. U. Snowden and J. N. Greenwood, "Surface Deformation Differences
between Lead Fatigued in Air and in Partial Vacuum," Transactions, Am.
Institute Mining, Metallurgical, and Petroleum Engrs., October, 1958, p. 626.
[26] K. U. Snowden, "The Effect of Oxygen Adsorption on the Fatigue Behavior
of Lead Single Crystals and Polycrystals," Proceedings, Sendai Conference,
September, 1965.
[27] D. A. Meyn, unpublished research, Naval Research Laboratory.
[28] J. L. Ham and G. S. Reichenbach, "Fatigue Testing of Aluminum in
Vacuum," Materials for Aircraft, Missiles, and Space Vehicles, ASTM STP
345, Am. Soc. Testing Mats., 1963, pp. 3-13.
[29] M. J. Hordon and M. E. Reed, "Study of the Mechanism of Atmospheric
Interaction with the Fatigue of Metals," NRC Project No. 86-1-0609,
August, 1965.
[30] K. U. Snowden, "The Effect of Cycle Frequency and Atmospheric Cor-
rosion on the Fatigue of Lead," Philosophical Magazine, Vol 10, 1964, p. 435.
[31] M. R. Achter, G. J. Danek, Jr., and H. H. Smith, "Effect on Fatigue of
Gaseous Environments Under Varying Temperature and Pressure," Trans-
actions, Am. Institute Mining, Metallurgical, and Petroleum Engrs., Vol 227,
1963, p. 1296.
[32] R. E. Schlier and H. E. Farnsworth, "Low-Energy Electron Diffraction
Studies of Cleaned and Gas-Covered Germanium (100) Surfaces," Semi-
conductor Surface Physics, Proceedings of a Conference, University of
Pennsylvania Press, Philadelphia, 1957.
[33] W. A. Wood, S. McK. Cousland, and K. R. Sargant, "Systematic Micro-
structural Changes Peculiar to Fatigue Deformation," Acta Metallurgica,
Vol 11, 1963, p. 643.
[34] M. R. Achter and H. W. Fox, 'The Effect of Surface Adsorption of Gas on
Crack Propagation," Transactions, Am. Institute Mining, Metallurgical, and
Petroleum Engrs., Vol 215, April, 1959, p. 295.
[55] K. U. Snowden, "Atmospheric Corrosion Fatigue of Indium," Journal of the
Less-Common Metals, Vol 7, 1964, p. 84.
DISCUSSION ON EFFECT OF ENVIRONMENT ON FATIGUE CRACKS 203
DISCUSSION
R. P. Wei1 (written discussion)—Dr. Achter is to be complimented on

his comprehensive review on the effect of environment on fatigue crack-
ing. In his review Dr. Achter pointed out that in aluminum the fact that
crack propagation rates are increased appreciably only by water vapor
and not by oxygen is interpreted as evidence that the effect is due to hy-
drogen which diffused into the metal. Although this viewpoint is widely
accepted also for the case of the enhancement of crack growth by atmos-
pheric moisture in high-strength steels, recent experimental results2'5
raise serious question to this interpretation, particularly regarding the
relative effects of "hydrogen embrittlement" and water-metal surface re-
action (oxidation) in promoting crack growth.
For a 2024-T3 aluminum alloy, Hartman2 observed that both water
vapor and dry oxygen increased the rate of fatigue crack propagation,
and concluded that these effects could be explained on the basis of a sur-
face reaction between the environment and the crack surface. The experi-
mental results of Bradshaw and Wheeler3 for an aluminum-copper-mag-
nesium type alloy showed that water vapor was much more effective in
accelerating fatigue crack propagation than dry hydrogen. On the other
hand, Hancock and Johnson4 showed the opposite to be true for the case
of a high-strength steel tested under constant static load (the behavior
being consistent with the influences of these environments on the rate of
fatigue crack propagation in high-strength steels5'6). It is evident that the
"hydrogen embrittlement" mechanism for the enhancement of crack
propagation does not adequately explain the observed moisture effect on
these materials. The importance of the water-metal surface reaction itself
in promoting crack growth must also be considered.2'5 Additional sup-
port for the surface-reaction concept is given by the results of the dis-
cussor and his colleagues5 on fatigue-crack propagation of a high-strength
steel tested in an atmosphere of bromine vapor and dry argon in which
both hydrogen and hydrogen-producing compounds are absent.
1
Research associate professor, Lehigh University, Bethlehem, Pa.
2
A. Hartman, "On the Effect of Oxygen and Water Vapor on the Propagation
of Fatigue Cracks in 2024-T3 Alclad Sheet," International Journal of Fracture
Mechanics, Vol 1, No. 3, September, 1965, p. 167.
S
F. J. Bradshaw and C. Wheeler, "The Effect of Environment on Fatigue
Crack Growth in Aluminum and Some Aluminum Alloys," Applied Materials
Research, April, 1966, p. 112.
*G. G. Hancock and H. H. Johnson, "Hydrogen, Oxygen and Sub-Critical
Crack Growth in a High Strength Steel," Transactions, Am. Institute Mining,
Metallurgical, and Petroleum Engrs., April, 1966.
5
See p. 460.
8 Downloaded/printed by and W. A. Spitzig, unpublished results.
R. P. Wei, P. M. Talda,
It is hoped that this discussion along with the fine review by Dr.
Achter will stimulate further research in this important area of material
behavior.
M. R. Achter (author)—Professor R. P. Wei is to be thanked for
drawing attention to the work of A. Hartman on aluminum.
It is agreed that surface reactions can increase crack propagation rates.
In the present review the large effects of the surface adsorption of oxygen
are mentioned prominently for copper, lead, nickel, and stainless steel.
Oxygen can have an appreciable effect on pure aluminum as shown by
Bradshaw and Wheeler (Ref. 9 of the paper), but on the commercial al-
loys studied in both investigations, as stated by Hartman, "water vapor
had a much more severe effect than oxygen." They both agree with Broom
and Nicholson (Ref 8 of the paper) that the dominant mechanism is the
diffusion of hydrogen.
It is well that Professor Wei has pointed out that oxygen adsorption on
aluminum can also have an effect.
R. W hERTZBERG1.
Fatigue Fracture Surface Appearance
REFERENCE: R. W. Hertzberg, "Fatigue Fracture Surface Appear-

ance," Fatigue Crack Propagation, ASTM STP 415, Am. Soc. Testing
Mats., 1967, p. 205.
ABSTRACT: The macroscopic and microscopic appearance of fatigue
fracture surfaces is strongly affected by the prevailing stress intensity
conditions at the moving crack tip. The nature of the fatigue striation
found during plane strain propagation is examined and crystallographic
considerations introduced to explain different topographical striation
arrangements. Program loading studies reveal the importance of stress
intensity factor range and maximum or mean values in the determina-
tion of striation spacing. Evidence of fatigue damage accumulation is
presented. "Elongated dimples" found in the plane stress region of crack
propagation are not related to stress intensity conditions or macroscopic
growth rate.
KEY WORDS: fatigue (materials), crack propagation, fracture surface,
striations, plane strain, plane stress, microstructure
Observations of a flat to shear fracture mode transition have been

made in the course of fatigue crack propagation studies. Irwin [7]2 has
associated this with a transition from plane strain to plane stress condi-
tions. Recent developments in electron fractography have shed consider-
able light on the fatigue fracture surface morphology of both regions. A
major characteristic feature observed in plane strain are parallel mark-
ings, referred to as striations. Forsyth and Ryder [2] found these stria-
tions could be associated with the position of the advancing crack inter-
face as a result of each load excursion. Others [3-7] have shown that
striations can both identify fatigue fracture as such and render important
quantitative information concerning fatigue crack propagation rates.
While much useful information concerning fatigue failure can be ob-
tained from the striations, doubt still exists about the mechanism(s) by
which striations are formed. Several models describing the formation
mechanism have been proposed, based upon different interpretation of
1
Assistant professor, Department of Metallurgy and Materials Science, Lehigh
University, Bethlehem, Pa.
2
paper.
205
observations. It is not clear why some regions in the plane strain region
show no signs of striations while contiguous areas show them to be clearly
defined. In addition, considerable uncertainty arises as to the appearance
of striations under conditions of complex sinusoidal loading paterns.
In the region of plane stress crack propagation fatigue striations are
generally not found. Their absence has been explained in terms of the
stress distribution associated with plane stress conditions [3]. Without
striations little is known about whether quantitative information may be
gleaned from the morphological features in the plane stress region.
The objective of this research was, therefore, threefold: (1) to examine
the nature of fatigue striations in some detail, (2) to examine striations
morphology under varying sinusoidal loading conditions, and (3) to de-
termine whether fracture surface characteristics in the plane stress re-
gion can be quantitatively related to observed fatigue crack propagation
rates and the existing stress intensity conditions.
Nature of Fatigue Striations

While differences exist in previously proposed models for fatigue
striation formation, there are some areas of general agreement. It is
agreed that striations are undulations on the fracture surface; while most
fatigue fracture surfaces are macroscopically flat, they are microscopi-
cally rippled. This fact has been experimentally verified by means of
metallographic sectioning [8] and electron fractography [9]. Based upon
the experiments of Forsyth and Ryder [7], it is generally agreed that each
striation is formed during one load cycle and that it defines the position
of the advancing crack front at a point during the fatigue life of the spec-
imen; it follows that striations are generally parallel to the advancing
crack front.
At this point differences in interpretation of observations become
marked. The undulations on the fracture surface which delineate each
striation have been described in terms of peak-to-peak or peak-to-valley
or valley-to-valley mating of matching fracture surfaces [6,8,10,11]. Pro-
posed models of striation formation have been based upon both crystal-
lographic and noncrystallographic considerations. It is possible that the
former models were proposed since most deformation processes are
crystallographic in nature. Noncrystallographic arguments may have
been proposed because striations were observed in face centered cubic
(fee), body centered cubic (bcc), and hexagonal close packed (hep) met-
als which would tend to cast doubt on the applicability of one striation
formation mechanism for three different types of crystal structures. Fi-
nally, some doubt exists as to whether fatigue crack growth is confined
to the loading portion of the load cycle or whether growth (or equivalent
damage) can occur
Copyright during
by ASTM the(allunloading
Int'l portionMon
rights reserved); as well.
Dec 7 14:40:45 EST 2015
HERTZBERG ON FATIGUE FRACTURE SURFACE APPEARANCE 207
It is clear that a critical evaluation of already proposed striation forma-

tion models is needed. The following discussion and experimental find-
ings are directed toward that end.
Previously Proposed Models

Laird3 has reviewed the features of the Forsyth and Ryder model and
the plastic relaxation model [6] which consider striations to be of the
peak-to-peak and generally of the matching trough morphology, respec-
FIG. 1—Large angle between adjacent packets of fatigue striations (X4800).
lively. Experimental results have shown the plastic relaxation model to

represent a more probable mechanism of crack propagation for high
purity aluminum and nickel [6] and also partially crystalline polymers
[12] which were subjected to large cyclic plastic deformation.
Published photographs in the literature [8,13] have suggested that stri-
ations can also take on a sawtooth appearance wherein peaks on one
fracture surface mate with valleys on the matching fracture surface. The
existence of the sawtooth type of striation has been established by Stubb-
ington [8] by means of metallographic taper sections and also suggested
p. 131.
SeeDownloaded/printed by
by Jacoby [10]. Stubbington considered one wall of the striation to have

been caused by plastic deformation while the other wall was considered
to have been produced by brittle cleavage. Several variables would ap-
pear to determine the relative amounts of each fracture component. One
objection to the Stubbington model is that it requires cleavage fracture
in fee metals and alloys which is not a likely event.
One may summarize the Stubbington and Jacoby model as depicting
sawtooth type striations, suggesting the need for crystallographic consid-
erations and not clearly defining growth with respect to specific portions
of the loading cycle.
Discussion of Previously Proposed Striation Mechanisms

In an earlier discussion of the plastic relaxation model [6], emphasis
was placed upon continuum aspects and little attention given to details of
crystallography. In Laird's paper, however, crystallographic considera-
tions were introduced resulting in a clearer evaluation of the deformation
process.
Since deformation in metals is highly anisotropic due to the limited
number of available slip systems, it is possible that striation formation
can be interpreted in terms of crystallographic considerations. Large
angles found between adjacent packets of striations (Fig. 1) are possibly
caused by slip system orientation differences in adjacent grains. Similar
reasoning would explain why striations are not always found on the frac-
ture surface (that is, grains not being favorably oriented for the forma-
tion of striations) [3,14,15]. Finally local changes in striation size could
be explained by elastic moduli and yield strength dependence on crystal-
lographic orientation within each grain.
Experimental Results
Differences in striation formation models may be traced in part to dif-
ferences in interpretation of striation morphology as observed with the
aid of electron fractography or metallographic sectioning. When working
with fracture replicas it is important that all uncertainties of replication
techniques be well understood. There are two major sources of possible
error in interpretation. First, unless stereographic techniques are exten-
sively employed, one is forced to analyze three-dimensional structures
with two-dimensional photographs. Second, the direction and angle of
shadowing with respect to the striations can lead to confusion when in-
terpreting the structure of the replica. In an attempt to clarify this point,
paper models of sawtooth striations with both equal and unequal wall
size were prepared and examined. Photographs of the paper striations
FIG. 2—(a) Paper model with oblique lighting, (b) Fractograph. Symmetrical
sawtooth type striations (X9300).
were taken with oblique lighting to simulate the effect of the shadowing
material on the replica surface. Figure 2a shows that with oblique light,
striations of equal size appear on the fracture surface as alternating light
and dark bands (one light and one dark band constituting one striation).
The adjacent fractograph (Fig. 2b) is similar in appearance and inter-
FIG. 3—(a) Paper model with oblique lighting, (b) Fractograph. Nonsym-
metric sawtooth type fatigue striations (X 13,200).
preted as a sawtooth type striation having equal sides. Figure 3a reveals

the two-dimensional photographic appearances of the paper model of
sawtooth striations of unequal size. When the light is directed from one
side, the two-dimensional appearance suggests the surface to be composed
of flatCopyright
darkly by
colored areas separated by thin lightly colored bands. In
this case the surface appears
Downloaded/printed by to be composed of flat bands with crevices
while in actual fact it is a misrepresentation of a sawtooth type striation

with unequal sides. Figure 3 b reveals a similar appearance on the frac-
ture surface. One may conclude from these observations that it is difficult
to correctly describe the nature of the fatigue striation found on the frac-
ture surface without resorting to stereographic techniques. Furthermore,
it appears that the sawtooth striation might be the general form found on
the fracture surface since its two-dimensional representation describes
both its own morphology and is similar to that which has been associated
with the flat type striation.
FIG. 4—Sketch of model describing mechanism of fatigue striation formation.

Crack propagation direction is parallel to long arrow. Striations seen to form on
(111) planes and slip occurs in (110) directions.
Discussion
The sawtooth morphology immediately suggests the interaction of
two competing planes during the fracture process, and the triangular
shape implies the appropriateness of crystallographic considerations.
When metals plastically deform only specific slip systems are operative.
In the FCC crystal system, for example, slip will occur on {111} type
planes and in (110) type directions. If one assumes that the sides of
Striations are parallel to the operative slip systems, then the striation sides
should be parallel to {111} slip planes in fee metals and alloys. The posi-
tion of the {111}
Copyright planes
by ASTM in rights
Int'l (all a given crystal
reserved); Monshould determine
Dec 7 14:40:45 the angle
EST 2015
that the Striations make
Downloaded/printed by with the advancing crack front. From this model
and its dependence upon the particular crystallographic orientation of

each grain, it is possible to explain why large angles are sometimes made
between adjacent packets of fatigue striations (Fig. 1). Furthermore, un-
favorable crystallographic orientations can give rise to the formation of
poorly defined striations or, perhaps, even none at all [3,14,15]. Such
areas are, indeed, observed on the fatigue fracture surface. The observa-
FIG. 5—Fractograph of 7075-T6 aluminum alloy showing slip markings on

side of striation wall which are not parallel to direction of crack propagation
(X8900).
tion that fatigue striations hi aluminum alloys are often curved is not
considered to be in contradiction with crystallographic arguments con-
cerning the mechanism of striation formation. While curved striations
have been found in high stacking fault energy aluminum alloys, straight
striations occur hi low stacking fault energy copper and copper base al-
loys. McEvily and Boettner [14] have attempted to explain this morpho-
logical difference hi terms of the effect of stacking fault energy on cross
slip. Copyright by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
ForDownloaded/printed
striations to formby exactly perpendicular to the crack direction, the
grain would have to be oriented as in Fig. 4 [3]. In this case the cross-
hatched {111} planes would be the fracture planes, while only two of
the three possible slip directions would be operative. From Fig. 4 it is to
be noted that the direction of slip is not parallel to the direction of crack
propagation. Consequently, an examination of the fracture surface should
reveal the presence of slip markings on the fatigue striations which are
not parallel to the direction of crack growth. Figure 5 taken from the
FIG. 6—Schematic diagram of special loading sequence applied to specimen.

Fracture surface appearance is described for conditions of tensile growth only and
for both tensile and compressive growth.
fracture surface of a 7075-T6 sheet specimen subjected to one peak load-

ing cycle reveals the presence of these slip markings along the wall of the
large striation, thereby, lending further credence to the proposed model.
Based upon the model described schematically in Fig. 4, the two com-
peting {111} planes are inclined at an angle of approximately 55 deg to
the horizontal plane where plane strain fracture occurs macroscopically.
It has been found that in plane strain the direction in which slip is most
likely to occur is along planes oriented at an angle of 60 to 70 deg to the
horizontal by This
[16].
Copyright ASTMsuggests that reserved);
Int'l (all rights the {111} Monplanes, when oriented
Dec 7 14:40:45 EST 2015in the
manner described in Fig.
Downloaded/printed by 4, are favorably oriented for slip. With such an
orientation, {100} plane is parallel to the horizontal fracture plane. It

is interesting to note that after polishing away fatigue striations Stubbing-
ton [8] found as a result of etch pit and X-ray diffraction studies that a
(100} plane was parallel to the horizontal fracture surface. This obser-
vation is consistent with the proposed model.
Effect of Variable Loading Pattern on Fracture Surface Morphology

Having interpreted the sawtooth striation as the product of two com-
peting fracture planes, there remains the task of determining when the
striation forms during the loading and unloading sequence. One of two
possible mechanisms exist: (1) crack propagation switches from one
plane to the other during the tensile portion of the loading cycle to form
the complete striation, and (2) the crack propagates along one plane dur-
ing the loading portion and then continues along the other plane during
the unloading portion of the cycle.
A disturbing factor associated with the first possible explanation is
that the crack must suddenly switch fracture planes while the load is
continually rising. This might occur in isolated instances, but it does not
appear likely that the switch can be a continually repetitious process so
as to produce the countless numbers of equally spaced striations found on
the fracture surface. Consequently, it is suggested that at least damage
and perhaps growth occurs separately during both loading and unloading.
It is important to note that compressive damage is implied to occur dur-
ing the unloading portion of a tension-tension loading cycle. It has been
observed that compression plasticity occurs at a crack tip for the unload-
ing portion of a tension-tension loading cycle or any other type of load-
ing [17].
To determine whether damage occurs during both loading and unload-
ing, block loading was employed. Approximately 30 load cycles with a
given stress range and maximum value were followed by one cycle with
the same stress range but higher maximum value. This loading scheme is
shown in Fig. 6. It was assumed that the major portion of growth would
occur during the half cycles where the change in the stress range was the
greatest. If damage or growth occurred only during the loading portion
of the cycle then the fracture surface should reveal the following: 30
small striations (formed during the tensile portions of the 30 cycles), one
large striation, 31 smaller ones, and so forth. The larger striation and the
additional small striation are considered to have formed during the load-
ing portions of their respective cycles (Fig. 6b). However, if damage or
growth occurred during both loading and unloading then one should see
a sequence consisting of 30 small striations one large band, one small
striation, one large band, 30 small striations, and so forth (Fig. 6c). Each
large band was considered to be produced by each half of the peak load-
FIG. 7—Fractograph of 2024-T3 showing two large bands associated with
growth during the peak loading cycle (X 13,000).
FIG. 8—Fractograph of 2024-T3 showing four large bands associated with

Copyright
growth during by
theASTM
peak Int'l (all rights
loading cyclereserved); Mon Dec 7 14:40:45 EST 2015
(X 13,000).
University of Washington (University of 215
Washington) pursuant to License Agreement. No further repro
FIG. 9—Fractograph of 2024-T3 showing six large striations associated with
growth during the peak loading cycles (X7300).
FIG. 10—Fractograph of 2024-T3 showing eight large striations associated

with Copyright
growth during the Int'l
by ASTM peak(allloading cycles (X7300).
216
ing cycle and the single small striation formed from the tensile and com-
pressive portion of the intermediate cycle. The appearance of this
sequence would suggest that damage or growth did occur during both the
tensile and compressive portions of the loading cycle. Specimens of
2024-T3 and 7075-T6 aluminum alloy were subjected to this loading
sequence and the fracture surfaces examined. Figure 7 suggests that
damage or growth did occur separately since two large bands separated
by a smaller striation were noted on the fracture surface. However,
further tests have cast some doubt on the above conclusion. Similar tests
FIG. 11—Plot of normalized striation size versus striation location after high
mean load block. Scatter band and mean values are plotted.
have resulted in the formation of four equally spaced bands on the frac-
ture surface (Fig. 8). These bands appear to be the equal sides of two
large striations caused by the loading sequence. Based upon this observa-
tion it appears that the maximum stress intensity value or the mean value
in addition to the range strongly affects striation size. Pelloux and Mc-
Millan4 have made similar fractographic observations in their program
loading fatigue studies. Therefore, the loading sequence used in this in-
vestigation did not clearly define whether growth occurred during both
loading and unloading. Pelloux and McMillan concluded from their
work that crack growth was restricted to the loading portion.
Further testing with three and five cycle blocks separated by 15 to 20
4
See p. 505.
cycle blocks were employed to further investigate the generality of the

striation size dependence on the mean or maximum stress intensity
values. In these loading blocks there were five cycles or two adjacent
blocks of three cycles each at the same range but higher mean stress in-
tensity values and six or eight peak stress intensity values, respectively.
If growth were controlled only by the range values, the fracture surface
would reveal two large bands separated by five smaller striations for the
FIG. 12—Fractograph of 7075-76 AI alloy revealing increased striation spac-

ing after peak load cycles. Note slip markings on side of large striations. Crack
propagation from bottom to top of photograph (X9000).
first case and two large bands separated by an array of three small
striations, one very large striation and three small striations for the
second case. However, Figs. 9 and 10 clearly show that six and eight large
striations result from the test conditions. These results suggest that
striation formation and growth are definitely affected by the stress in-
tensity maximum or mean values.
Accumulated Damage
Using specially programmed loading patterns as discussed previously
the question of fatigue damage accumulation and its dependence on stress
history arises. Hudson and Hardrath [18], for example, reported that
large intermediate stress cycles in a programmed loading sequence
severely retarded crack propagation during subsequent cycling at lower
stress levels. Residual compressive stresses at the fatigue crack tip from
the peak stress cycles were considered to be the cause of crack arrest at
lower stresses [18,19]. Paris [20], on the other hand, reported that no de-
lays occurred during random loading since a corresponding number of
fatigue striations found on the fracture surface could be traced to each
FIG. 13—Fractograph of 2024-T3 showing "elongated dimple" formation on

plane stress fracture surface. Direction of crack propagation parallel to arrow
(X7500).
loading peak in the loading spectrum. From this fractographic observa-

tion, it would appear that crack growth under random loading was a
reasonably continuous process.
To investigate the question of possible crack arrest due to intermediate
peak load cycling, the fracture surfaces described in the above section
were examined for signs of total crack arrest or damage accumulation.
As discussed in the Appendix, the ratio between the high-mean and low-
mean load blocks was 1.31 for the 2024-T3 alloy. A count of striations
causedCopyright
by the by
lower mean load block showed that no crack arrest had
occurred as a result of by
Downloaded/printed the prior peak load cycles. To the contrary, meas-
urement of each striation immediately following many peak striations

revealed that an accumulation of damage had occurred. Figure 11 repre-
sents a plot of striation size (normalized to the striation spacing imme-
diately preceding the peak load tycles) as a function of distance from the
larger striations. Though scatter is considerable, a trend is apparent.
The striations formed at the lower load levels immediately after the peak
loads are larger than those that follow (Fig. 12). One may infer from
this plot that damage was accumulated from the peak loads and caused
an increase in striation spacing with increasing amounts of damage from
the prior peak load cycles.
It would be of great interest to determine whether other block loading
combinations also lead to the observed results.
Plane Stress Fracture Surface Morphology
Microfractographic examination of plane stress fatigue fracture sur-
faces revealed "elongated dimples" to be the predominant feature on the
surfaces examined. However, McEvily5 has observed areas of striations
on plane stress fracture surfaces in copper and copper base alloys. It is
not clear at this time whether the difference in fracture appearance is due
to different material behavior or rather to a different stress environment
such as mean stress. The "dimples" observed in this investigation were
aligned with their parabolic axes parallel to the advancing crack inter-
face; the axes are perpendicular to the direction of crack propagation
(Fig. 13). It is of interest and importance to determine whether the size
of the dimples can be related to the applied stress intensity conditions at
the crack tip and whether such a variation in dimple size could be related
to macroscopically determined crack growth rates. To this end, a 0.126-
in.-thick test specimen of 2024-T3 Al alloy exhibiting a large region of
plane stress crack propagation was examined fractographically. The
pertinent dimple measurement was considered to be that of the dimple
width since this dimension was parallel to the direction of crack propaga-
tion and would give a measure of the microscopic growth rate. Measure-
ments were taken across the entire plane stress region from its origin at
relatively low stress intensity conditions to the point of total failure at
the highest stress intensity condition. In this region the macroscopically
determined growth rate increased from 2.3 X 10~4 in./cycle to about
1 X 10~2 in./cycle. No significant change in dimple size (approximately
0.35 to 0.70 ju.) was observed over the entire region examined. The only
perceptible change in the fracture surface appearance was that dimples
became more prevelant on the fracture surface toward the latter stages of
A.Downloaded/printed
J. McEvily, privateby
communication.
fatigue Ufe. This may simply be due to the fact that less surface rubbing
and damage occurred near the final fracture site so fewer dimples would
have been obliterated. Figure 14 reveáis the fracture surface morphology
attrition resulting from repeated contact of the two fractured halves of
the specimen.
Discussion
It is concluded that the growth rate of plañe stress fatigue crack propa-
gation can not be determined by dimple size since the latter appears to be
independent of both growth rate and stress intensity conditions. Since
dimples are usually associated with fracture of secondary particles in
FIO. 14—Plañe stress fatigue fracture surface reveáis rubbing marks (X7000),
regions of large strain concentration, the dimple size and distribution

might be more of a function of the distribution of impurity particles in
the strained región than any imposed stress intensity field. Furthermore,
crack growth raíe during one loading cycle in plañe stress cannot be re-
lated to the extent of one fracture marking as was the case in plañe
strain propagation where each striation was related to one loading excur-
sión. In the plañe stress región the observed dimple size was consistently
smaller than the average growth rate for one loading cycle. Therefore,
growth by dimple formation in the plane stress region is controlled by

some multiple of an apparently invariant unit fracture size. This multi-
ple will increase with increasing growth rate. These observations show that
the crack is stable even though many fracture units are traversed during
one loading cycle. This would tend to cast some doubt on the Krafft
model [21] which considers fracture instability to occur when one process
zone (considered to be related to dimple size) is ruptured.
The fact that dimples are characterized by a fairly constant size sug-
gests that they might be related to some characteristic microstructural or
continuum zone size. It is doubtful that the plastic zone size would.be the
key factor since it is strongly related to the stress intensity state while
dimple size appears to be independent of this. On the other hand, it is
very tempting to relate dimple size to deformation induced cell structures
that are observed in thin film electron microscopy. Not only is cell size
stabilized after large amounts of deformation (recall that dimples also
appear to be a constant size) but also the absolute cell dimensions are
of the same order of magnitude as dimple sizes. Whether these observa-
tions show that such a relationship does exist or is purely fortuitous must
certainly await a detailed fracture analysis employing both fractographic
techniques and thin film electron microscopy of material immediately
adjacent to the fracture surface.
Conclusions
1. Several different observed striation morphologies can be interpreted
as being variations in appearance of sawtooth type striations. On this
basis, the sawtooth type is considered to be a commonly produced form.
2. Crystallographic considerations can be used to explain the forma-
tion of sawtooth type striations, rationalize large angle differences ob-
served between adjacent striation packets, and offer reason for the ab-
sence or presence of striations in the plane strain region.
3. Striation size has been found to be dependent upon both stress
intensity range and maximum values. Evidence of fatigue damage ac-
cumulation is presented.
4. Elongated dimples found in the plane stress region do not vary
with increasing stress intensity conditions or macroscopic growth rate.
It is concluded that dimple size can not give a measure of crack propaga-
tion rate.
A cknowledgment
The author wishes to thank the National Aeronautics and Space
Agency (Grant Nsg 410) and the Institute of Research of Lehigh Uni-
versity for support of the investigation. Grateful acknowledgment is made
to Professor Paul Paris of Lehigh University for his many contributions
to this manuscript and to Stanley Shapiro for many enlightening discus-

sions of fractographic interpretation.
APPENDIX
Experimental. Procedure
All tests were conducted using 0.126-in. gage 2024-T3 and 0.100-in. gage
7075-T6 aluminum alloys. Test specimens were of the single-edge notch
pin loaded type with overall dimensions of 3 in. by 12 in. An initial side notch
1 in. in length was machined and sharpened with a scalpel.
In testing the 2024-T3 alloy, the load range for the low-load cycles was
from 500 to 2100 Ib while the peak loads were 2500 Ib. Consequently the
mean-load ratio between the low-load cycles and the peak-load cycles was
1.31. For the 7075-T6 alloy the load range for the low-load cycles was from
500 to 2350 Ib, while the peak loads were 3000 Ib. Therefore, the mean-load
ratio was 1.45.
Fracture surface replicas were prepared by the two-stage technique using
Faxfilm and carbon. Replicas were shadowed parallel to the direction of
crack propagation using carbon-platinum pellets.
References
[1] G. R. Irwin, "Fracture Mode Transition for a Crack Traversing a Plate,"
Transactions, Am. Society Mechanical Engrs., Vol 82, Series D, No. 2,
1960, p. 417.
[2] P. J. E. Forsyth and D. A. Ryder, "Fatigue Fracture," Aircraft Engineering,
Vol 32, No. 374, 1960, p. 96.
[3] R. W. Hertzberg, "Application of Electron Fractography and Fracture
Mechanics to Fatigue Crack Propagation in High Strength Aluminum
Alloys," Ph.D. dissertation, Lehigh University, Bethlehem, Pa., June, 1965.
[4] D. E. Piper, W. E. Quist, and W. E. Anderson, "The Effect of Composition
on the Fracture Properties of 7178-T6 Aluminum Alloy Sheet," presented
at AIME Fall Meeting, Philadelphia, Pa., October, 1964.
[5] C. Carmen and M. Shuler, "Low Cycle Fatigue Properties of 18 Ni-Co-Mo
250 Maraging Steel," presented at ASTM Subcommittee on Electron
Fractography, Schenectady, N. Y., September, 1964.
[6] C. Laird and G. C. Smith, "Crack Propagation in High Stress Fatigue,"
Philosophical Magazine, Vol 7, 1962, p. 847.
[7] R. M. H. Pelloux, "Fractographic Analysis of the Influence of Constituent
Am. Society Metals, Vol 57, No. 2,1964, p. 511.
Aluminum-7.5% Zinc-2.5% Magnesium Alloy," Metallurgia, Vol 68, 1963,
p. 109.
_ [9] C. Crussard, J. Plateau, R. Tamhanger, and D. Lajeunesse, "A Comparison
of Ductile and Fatigue Fractures," Fracture, New York Technology Press
M.I.T. and John Wiley & Sons, Inc., New York, 1959, p. 524.
[10] G. Jacoby, "Fractographic Methods in Fatigue Research," Experimental
Mechanics, Vol 5, No. 3, March, 1965, p. 65.
[11] P. J. E. Forsyth and D. A. Ryder, "Some Results of the Examination of
Aluminum Alloy
Copyright by ASTMSpecimen Fracture
Int'l (all rights Surfaces,"
reserved); 7 14:40:45 Vol
Metallurgia,
Mon Dec 63, 1961,
EST 2015
117.
p.Downloaded/printed by
[12] A. J. McEvily, Jr., R. C. Boettner, and T. L. Johnston, "On the Formation

and Growth of Fatigue Cracks in Polymers," Fatigue—An Interdisciplinary
Approach, Syracuse University Press, Syracuse, N.Y., 1965.
[13] P. J. E. Forsyth, "Fatigue Damage and Crack Growth in Aluminum
Alloys," Acta Metallurgica, Vol 11, July, 1963, p. 703.
[14] A. J. McEvily and R. C. Boettner, "On Fatigue Crack Propagation in F.C.C.
Metals," Acta Metallurgica, Vol 11, No. 7, 1963, p. 703.
[15] R. W. Hertzberg and P. C. Paris, "Application of Electron Fractography and
Fracture Mechanics to Fatigue Crack Propagation" presented at Interna-
tional Conference on Fracture, Sendai, Japan, September, 1965.
[16] M. L. Williams, "On the Stress Distribution at the Base of a Stationary
Crack," Journal of Applied Mechanics, Am. Society Mechanical Engrs., Vol
24, 1957, p. 109.
[77] P. C. Paris, "The Growth of Cracks Due to Variations in Load," Ph.D.
dissertation, Lehigh University, Bethlehem, Pa., September, 1962.
[18] C. M. Hudson, and H. F. Hardrath, "Investigation of the Effects of Variable-
Amplitude Loadings on Fatigue Crack Propagation Patterns," NASA
Technical Note D-1803, August, 1963.
[19] A. J. McEvily and W. nig, "An Investigation of Nonpropagating Fatigue
Cracks," NASA TN D-208, December, 1959.
[20] P. C. Paris, 'The Fracture Mechanics Approach to Fatigue," Fatigue—An
Interdisciplinary Approach, Syracuse University Press, Syracuse, N. Y., 1964,
p. 107.
[21] J. M. Krafft, "Correlation of Plane Strain Crack Toughness with Strain
Hardening Characteristics of a Low A Medium, and a High Strength Steel,"
Applied Materials Research, April, 1964, p. 88.
DISCUSSION
/. Schijve1 (written discussion)—The effect of a high load cycle on sub-

sequent crack growth was an acceleration of the growth. This is an in-
teresting result since this effect was not apparent from crack growth
records based on visual observations made previously. Apparently it
requires fractographic observations with the electron microscope to re-
veal this interaction effect. The effect could be indicated as an unfavor-
able interaction effect since it implies an acceleration of crack growth.
Until now only favorable interaction effects due to high peak loads had
been noted, involving delays of crack growth. The delays were generally
attributed to residual stresses. It appears questionable whether the latter
argument could be applicable to the author's results. The author
mentioned a "conditioning" of the material as an explanation. In this
respect I can think of strain hardening only the more so since Dr.
1
Director of technical
Copyright by ASTMservices,
Int'l (allNational AerospaceMon
rights reserved); Laboratory,
Dec 7 NRL, Amster-
14:40:45 EST 2015
dam, Holland.
DISCUSSION ON FATIGUE FRACTURE SURFACE APPEARANCE 225
Grosskreutz in his paper showed the precipitates to be stable under

cyclic deformation.
R. W. Hertzberg (author)—The observation of crack growth accelera-
tion measured fractographically is in direct contrast with crack growth
retardation observed on the surface of test panels. As Dr. Schijve has
stated, a favorable residual stress pattern has been proposed to explain
crack growth retardation at the surface. Furthermore, "conditioning" due,
perhaps, to strain hardening might be argued as the cause for crack
growth acceleration internally. It is important to verify whether both
arguments may be involved simultaneously as the data would imply.
The main underlying factor might be found in a consideration of the pre-
vailing stress conditions (that is, plane strain versus plane stress) through
the panel thickness.
/. C. Grosskreutz1 and G. G. Shaw*
Microstructures at the Tips of Growing

Fatigue Cracks in Aluminum Alloys*
REFERENCE: J. C. Grosskreutz and G. G. Shaw, "Microstructures at

the Tips of Growing Fatigue Cracks in Aluminum Alloys," Fatigue Crack
ABSTRACT: The technique of transmission electron microscopy has
been used for direct observation of the microstructures at fatigue crack
tips in 1100-O, 2024-T4, and 7075-T6 aluminum alloys. The dislocation
morphology and density, and the metallurgical structures which are
formed in the plastic zone were studied. In 1100-O, the dislocations are
distributed in a subgrain structure, while in 2024-T4 they are distributed
uniformly throughout the grain, and many dislocation loops are found.
In 7075-T6, dislocations are tightly packed and show little sign of motion
or interaction. No correlation was found between the dislocation struc-
ture just beneath the fracture surface and the regularly spaced crack
growth striations found on the surface. Dislocation densities 6 /j, from the
crack range from about 1010 lines/cm2 in 1100-O to about 1011 lines/cm2
in 7075-T6. No localized reversion or overaging was observed to occur
in the plastic zones of these alloys. The possibility of a general change in
metallurgical structure is discussed, and it is concluded that reversion of
very small Guinier-Preston zones or a limited, general overaging would
not have been detected by our techniques.
KEY WORDS: fatigue (materials), crack propagation, microstructure,
aluminum alloys, dislocations, plastic flow
It is now well established that fatigue failure is the result of a series

of highly localized events [1].B Crack nucleation normally occurs at the
surface of a material either in slip bands or at the interface between an
inclusion and the matrix. Growth of the microcrack then proceeds by
means of repeated plastic deformation caused by the stress concentration
at the tip of the crack. Currently, there is widespread effort devoted to the
study of the plastic zone which exists at crack tips. The size and shape of
* Supported by Air Force Materials Laboratory, Wright-Patterson Air Force

Base, Ohio.
1
Senior advisor for physics, Midwest Research Inst, Kansas City, Mo. Personal
member ASTM.
2
Senior electron microscopist, Midwest Research Inst., Kansas City, Mo.
3
paper.
226
GROSSKREUTZ AND SHAW ON MICROSTRUCTURES 227
the zone, the strain distribution within the zone, and the dislocation
interactions which lead to hardening and eventual fracture within the
zone are all subjects of intense interest. A number of crack propagation
theories have been advanced which are based on specific models of the
deformation within the plastic zone [2-5].
Direct observation of the microstructure which develops within the
plastic zone has heretofore been restricted to surface etching patterns.
The interpretation of these patterns in terms of dislocation morphology
and metallurgical structure is not always straightforward. There is, in
fact, a definite need for a direct, unambiguous observation of these
FIG. 1—Schematic of fatigue crack in partially failed specimen. Direct ob-

servation of microstructure is made at point P, a distance r below the fracture
surface. Earlier position of crack (length 1) and associated plastic zone is shown as
dashed lines.
microstructures within the volume of the plastic zone. The information

thus obtained would be of value in establishing the mechanisms of crack
growth, in interpreting fracture surface features (fractographic analysis),
and in determining the degree of metallurgical stability under the high
strains which exist at the tip of a growing crack.
TABLE 1—Nominal composition (in weight per cent) of aluminum alloys.
Alloy Si Fe Cu Mn Mg Cr Zn Ti
1100 05 05 02 005 01
2024 05 05 45 06 15 0.1 025
7075 05 07 16 03 25 03 56 0.2
In this work, the technique of transmission electron microscopy has

been applied for the first time to the direct observation of the microstruc-
tures at fatigue crack tips. The general principles of the technique are
illustrated in Fig. 1. The point of observation, P, lies in a plane below
and parallel to the crack surface. The dislocation morphology and
metallurgical changes which are observed result from the large strains
which occurred earlier in the region when the crack tip passed close by
(dashed lines, Fig. 1). The technique could be extended to map the en-
tire plastic
Copyrightzone;
by ASTMsuch anrights
Int'l (all extension would
reserved); Mon Dec 7involve considerably
14:40:45 EST 2015 more
finesseDownloaded/printed
than the elementary
by sampling of the zone which is reported here.
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions
FATIGUE CRACK PROPAGATIAGATIONON
Three commercial aluminum alloys were chosen for study: 1100-O

(commercially pure aluminum), 2024-T4, and 7075-T6. The latter two
are age-hardened, multiphase alloys. The nominal compositions of the
three materials are given in Table 1.
Experimental Details
Flat, notched specimens were machined from sheet stock of each of
the three alloys. The gage section was nominally l/s in. thick by l/z in.
wide. Specimens were then heat treated to the appropriate condition and
fatigued under constant, fully reversed push-pull load in a Sontag SF-U-
1 machine. The faces of the specimen were replicated at intervals, and
after the appearance of a crack, the specimen was observed closely until
FIG. 2—Illustration of technique for obtaining thin foils adjacent to fracture

surface. Thin slab (dashed lines) is first cut by spark erosion; disk Ys in. in diam-
eter is then removed by spark machining and thinned from one side by electro-
polishing.
the crack had grown to approximately 5 mm in length. At this point, the

test was stopped and the specimen removed for observation in the
electron microscope.
Thin foils, suitable for electron transmission (1000 to 4000 A thick),
were obtained from the plastic zone by a combination of spark machin-
ing and electropolishing. First, a thin section containing the fracture
surface was cut from the specimen by spark erosion (dashed lines, Fig.
2). The fracture surface was then covered with a protective lacquer, and
a disk l/s in. in diameter was spark-machined from the section. With the
protective lacquer still in place, the disk was thinned by electropolishing
(at IOC) from the rear until perforation occurred at the fracture surface.
Observation at various distances from the crack surface was achieved by
first electropolishing the desired amount from the fracture surface before
beginning the thinning process from the rear. The details of the thinning
procedure have been described elsewhere [6].
FIG. 3—Dislocation morphology in 1100-O aluminum: (a) uncycled, (b) 6 p

below fracture surface, and (c) 2.5 mm ahead of final crack tip. (3 X 10e cycles at
±4160 psi.)
The perforated disk was mounted directly into the specimen holder of
an electron microscope. Examination of the foils was carried out at an
accelerating voltage of 100 kv. Dislocations were observed directly in
the thin area surrounding the perforation by means of diffraction con-
trast; Guinier Preston (GP) zones and precipitates were observed pri-
Copyright by ASTM Int'l (all rights reserved); Mon Dec
marilyDownloaded/printed
by means of structure
by
factor and orientation contrast [7]. Average
University of Washington (University of Washington) pursuant to
FIG. 4—Dislocation morphology and metallurgical structure in 2024-T4

aluminum: (a) uncycled, (b) 6 /j, below fracture surface. Dislocations impeded by
precipitates at X; tangles at inclusion, Y. (6.5 X 10s cycles at ±14,000 psi.)
dislocation densities were measured after the method of Ham [8]. Foil
thickness, necessary
Copyright by ASTM for Int'l
this (all
calculation, was obtained
rights reserved); Mon Decfrom dislocation
7 14:40:45 EST 2015
traces Downloaded/printed
in 1100 aluminum, by from the projected lengths of helical disloca-
GROSSKREUTZ~AND SHAW ON MICROSTRUCTURES 231
FIG. 5—Dislocation morphology and metallurgical structure in 7075-T6

5
aluminum: (a) uncycled,
Copyright (b) (all
by ASTM Int'l 2 ^rights
below fracture
reserved); Monsurface.
Dec 7 (2.7 X 10
14:40:45 EST cycles
2015 at
±13,400 psi.)
FATIGUE CRACK PROPAGATIONAGATION
tions in 2024 aluminum, and from the projected width of large 77 pre-
cipitates in 7075 aluminum.
The dislocation morphologies, dislocation densities, and metallurgical
structures which were observed are described in the succeeding para-
graphs.
Dislocation Morphology
1. 1100-O—The dislocation morphology in uncycled material is com-
pared in Fig. 3 with that observed at 6 and 2500 p. ahead of a fatigue
crack. The specimen had accumulated a total of 3 X 106 cycles under a
nominal stress of ±4160 psi. The length of the final crack was such that
the net section stress was above the yield stress for 1100-O aluminum.
This fact accounts for the appreciable dislocation density in Fig. 3c. The
TABLE 2—Dislocation densities, A, observed near fatigue cracks.

Material Nominal Stress, psi la, mm ra, microns A, lines/cm2
1100-O ±4 160 ~3.5 6 1 X 1010

2024-T4 ±14 000 1.5 6 2 X 1010
7075-T6 ±13 400 ~1.7 6 8 X 1010
7075-T6 ±13 400 6.5 2 2 X 1011
7075-T6 >±73 000 ? 6 4 X 1011
a
I and r are defined in Fig. 1.
structure at the crack tip shown in Fig. 3b can be accurately described as

subgrains; appreciable misorientation exists between adjacent cells as
evidenced by the abrupt change in diffraction contrast at the boundaries.
2. 2024-T4—The dislocation arrangement in an uncycled specimen
and 6 /* from the crack surface is shown in Fig. 4a and b. The dislocations
in Fig 4a were produced by the quenching treatment. The structure 3
mm ahead of the crack was identical with the uncycled specimen. This
specimen had accumulated 650,000 cycles at a nominal stress level of
±14,000 psi. No subgrain or cell structure was observed in the fatigued
specimen. A close look at Fig. 4b shows several dislocations which have
been impeded in their motion by the presence of GP zones (X in Fig.
4b). Other dislocations appear tangled about the large inclusions at Y
(probably iron + silicon). Dislocation loops are widely distributed.
3. 7075-T6—The dislocation morphology in an uncycled specimen
and 2 p. from the crack tip in a fatigued specimen are shown in Fig. 5a
and b, respectively. The dislocation structure 2 mm ahead of the crack
was identical with the uncycled specimen. This specimen had accumu-
lated Copyright
2.7 X 10 by5 ASTM
cyclesInt'l
at (all
±13,400 psi. Although
rights reserved); Mon Dec 7subgrains are2015
14:40:45 EST found
extensively in as-received 7075-T6 sheet (and are not removed by solu-

tion heat treatment), no additional subgrain or cell structure is intro-
duced in this alloy by the large plastic strains which exist near a fatigue
crack. Rather, the dislocations are evenly distributed and very dense.
There is little evidence that the dislocations have moved far enough to
interact extensively with themselves or with the precipitate structure.
Dislocation Densities
The dislocation densities, observed at various distances, /, from the
point of crack nucleation and removed a distance, r, from the crack sur-
face, are given in Table 2.
The specimen listed on the last line of Table 2 was loaded well into
the plastic range by means of the preload motor on the Sontag machine.
Plastic strains of the order of 4 to 5 per cent were applied, and the speci-
men failed after 9l/z cycles.
It is difficult to assess the accuracy of the dislocation densities quoted
in Table 2. Certainly the number given for 1100-O aluminum may be too
low by as much as 50 per cent [9,10]. The values given for 2024 and
7075 should be fairly reliable; the dislocations in these materials are too
well immobilized by the precipitate for much loss to occur during thin-
ning of the specimen.
Metallurgical Structures
In 2024 aluminum, the T4 condition is characterized by a uniform
dispersion of GP zones and transition phase precipitates with an average
size of approximately 150 A (Fig. 4a). Scattered throughout this disper-
sion are much larger inclusions and precipitates whose dimensions are
approximately 3000 by 900 A. The section shown in Fig. 4a was taken
transverse to the long direction of these large particles, and, hence, they
appear as circular spots. Within the plastic zone near a fatigue crack,
the general metallurgical structure does not appear appreciably changed
(Fig. 46). The dispersed precipitate structure can still be observed, al-
though it is not as well resolved as in the photographs of unstrained
specimens. This difficulty arises because of the uneven nature of the frac-
ture surface which prevents obtaining large fields of view of uniform
thickness. The structure factor contrast by which these small precipi-
tates are imaged becomes weaker and finally disappears completely as
the thickness of the foil increases. No localized re-solution of precipitates
was observed following fatigue, nor was there any evidence of slip zones
in which large amounts of dislocation motion had occurred. No attempt
was made to detect subtle changes in structure which might have oc-
curred, for example, a small, general reversion or overaging.
Copyright
Because of thebypoor ASTM
Downloaded/printed
fatigue
by
Int'l
behavior (all rights
of 7075-T6 reserved);
aluminum, Mon
a de- Dec
University of Washington (University of Washington) pursuant

tailed search was made for possible metallurgical changes near the crack
tip. The normal aging sequence in this alloy is [11,12]
GP zones- • •?/• • -J?(MgZn 2 ).
In the state of maximum hardness, the alloy contains a fine dispersion
FIG. 6—Effect of fatigue on metallurgical structure at the crack tip in 7075-

T6 aluminum: (a) 2 /* from fracture surface, dislocations in contrast; (b) same
area as (a), but with dislocations out of contrast (arrows denote identifying pre-
cipitate in (a) and (b)); and (c) uncycled control specimen.
of GP zones and T\ precipitates whose average size is approximately

150 A (Fig. 5a). Long lath shaped 77 precipitates about 300 X 1000 A
can also be seen. It is characteristic of this alloy [12] that the dispersion
of precipitates after a uniform heat treatment cycle is highly inhomogene-
ous both within single grains and from grain-to-grain. Therefore, in
attempting to byassess
Copyright ASTMboth
Int'l the
(all uncycled and Mon
rights reserved); fatigued
Dec states, a large
7 14:40:45 EST num-
2015
ber ofDownloaded/printed
photographs must by be taken of many different foils. The photo-
graph in Fig. 5a was chosen to illustrate all three of the phases which
are present in 7075-T6. Many regions can be found in which one or the
other of these phases predominates. Precipitate-free zones along primary
grain boundaries are typical of this alloy and extend approximately 150
A on either side of the boundary. These zones were not observed to be
preferred paths for crack growth. In our specimens, which were oriented
such that cracks grew transverse to the original rolling direction, fatigue
cracking was predominantly transgranular.
A careful study of photographs taken well within the plastic zone in
this material has not revealed any qualitative differences in the metal-
lurgical structure over the uncycled state. There are no localized regions
of obvious reversion or overaging. In Fig. 6, the structure in an uncycled
control is compared with an area 2 ju from a crack which developed
after 267,000 cycles at ±13,400 psi. (The control piece was cut from the
end of the fatigued specimen.) The dislocations are shown in contrast
in Fig. 6a, while they have been tilted out of contrast in Fig. 6b to reveal
only the precipitate structure which has survived intense plastic strains.
There is no qualitative difference between Fig. 6b and the control shown
in Fig. 6c. Specimens which had been subjected to extreme cyclic stress
conditions were also examined. One specimen developed a crack and
failed within 9% cycles; another developed a crack after 5% million
cycles of stressing. In neither case could any metallurgical changes be
detected by comparing micrographs from the plastic zone with those
from uncycled specimens.
It is useful to estimate the sensitivity of our qualitative comparison
technique to a general reversion or overaging (over several square mi-
crons) of the precipitate structure. From measurements on an uncycled
foil, the average number of second phase particles (GP + 77') per unit
volume, Nv, is 5 X 1015, and the average particle dimension, K, is 170
A. We estimate that an increase or decrease of 50 per cent in this average
dimension would certainly have been observable, and that an increase
or decrease in the total number per unit volume by a factor of two would
also have been observable. Any smaller changes than these would have
to be determined by quantitative measurements of Nv and K on a large
number of foils. Qualitative comparison would almost certainly fail to
detect a general reversion of very small (<60 A) GP zones. These par-
ticles do not occur uniformly throughout a control foil, and the added
problems of resolution brought about by the unevenness of the fracture
surface makes it difficult to detect them over extended areas in a fatigued
specimen.
Discussion
In Copyright
none ofbythe
ASTMthree materials
Int'l (all wereMon
rights reserved); anyDec
microstructures
7 14:40:45 EST 2015observed be-
neath Downloaded/printed
the fracture surface
by which would correlate with the regularly spaced
growth striations which are normally found on the surface. There is, for
example, no sawtooth variation in dislocation density which might cor-
respond to a discontinuous mechanism of crack growth. Rather, the three
alloys develop quite different dislocation distributions under fracture sur-
faces which all exhibit comparable crack growth striations. These obser-
vations support the original model proposed by Laird and Smith [13],
in which striations are formed as a result of the periodic blunting and re-
sharpening of the crack tip.
There are at present no suitable theories of crack tip yield with which
to compare the experimental values of dislocation densities quoted in
Table 2. Bilby et al [14] have treated the yield at the tip of a two-dimen-
sional crack in terms of a one-dimensional array (continuum) of edge
dislocations extending in front of the crack. However, to obtain quantita-
tive values of the number of dislocations per unit length requires that a
number be assigned to o-i, the frictional stress which opposes dislocation
motion. Because this parameter is not known for most materials, it ap-
pears that a better procedure might be to use the experimental values of
dislocation density to obtain an estimate for ai. We have not yet carried
out this calculation.
Some of the additional observations which have been made are best
discussed separately for each material.
1100-O
The observation of extensive subgrain formation within the plastic
zone in this material correlates well with previous observations. Segall
et al [-75], and Grosskreutz and Waldow [16] showed that subgrains
formed throughout high purity aluminum provided the cyclic strain level
exceeded a certain threshold. Moreover, Grosskreutz [17] has shown
that fatigue crack growth in 1100-O single crystals proceeds along sub-
grain boundaries. These boundaries apparently represent a least energy
path for the advancing crack. The formation of subgrains can also be
viewed as a stress relaxation mechanism whereby a uniform distribution
of dislocations relaxes into a cell structure with a corresponding absorp-
tion of energy. This mechanism permits crack growth rates in 1100-O
aluminum stressed near the yield stress which are comparable to those in
much harder materials when stressed at only 10 to 20 per cent of the
yield stress.
2024-T4
The presence of a uniform dispersion of second phase precipitates
prevents the relaxation of dislocations into a cell structure in this mate-
rial. Nevertheless, there is evidence for considerable ductility at the crack
tip from
Downloaded/printed by morphology of Fig. 4a. The dislocations show
the dislocation
evidence of motion and interaction which implies that appreciable work

hardening has also taken place.
McEvily et al [18] have suggested that reversion of the precipitate
structure in this alloy occurs during prolonged fatigue stressing, thereby
explaining, at least partially, the low ratio of endurance limit to ultimate
strength. Subsequent transmission microscopy of a high purity alumi-
num-4 per cent copper alloy fatigued in torsion confirmed that reversion
had occurred in localized bands of high dislocation activity [79]. The
present investigation has not revealed either localized or widespread
evidence of reversion of the GP zone and transition precipitate structure
in 2024-T4 aluminum. To subject the hypothesis of McEvily et al to
an exacting test would require a more exhaustive study of this alloy by
transmission electron microscopy. Such a study should include a quanti-
tative determination of the distribution of precipitate size and the num-
ber of second phase particles per unit volume before and after fatigue.
7075-T6
The dislocation morphology which is observed near the crack in this
alloy (Fig. 5£) illustrates a greater degree of dislocation pinning than in
2024-T4. There is little evidence of dislocation interaction which would
lead to work hardening. Furthermore, the dislocation density is 3 to 5
times higher than that observed in 2024. Apparently, the plastic strains
at the crack tip are accommodated more by the generation of additional
dislocations than by the continued motion of those already present. Wal-
dron [20] has observed a similar increase in dislocation density for a
given strain in aluminum-magnesium alloys as the yield stress is raised
by the addition of magnesium.
There has been considerable interest over the past decade in the metal-
lurgical stability during fatigue of alloys based on the ternary aluminum-
zinc-magnesium system. Hanstock [21] measured the energy dissipation
per cycle during torsion fatigue of commercial DTD 683 aluminum (simi-
lar to 7075) and proposed that the increased dissipation which occurred
after several million cycles was due to a softening brought on by overag-
ing. A reheat treatment served to restore the energy dissipation to its
original value. Sections of a specimen were also prepared which con-
tained a group of small surface cracks; etching with nitric acid revealed
a set of bands parallel to the cracks. Hanstock concluded that the etching
pattern was caused by localized precipitation brought about by the
prolonged fatigue cycling. Later, Broom et al [22] prepared superpurity
alloys similar in composition and mechanical properties to DTD 683
and performed metallographic examination of sections taken from fa-
tigued specimens. After etching, they observed "dark bands" which cor-
relatedCopyright
in spacingbyand direction
ASTM
with
Int'l
surface
(all
sliprights
bands. reserved);
Such regionsMon De
were Downloaded/printed
not formed when the by alloy was fatigued at 78 K. These authors
FATIGUE CRACK PROPAGATIOAGATIONN
concluded that re-solution had occurred within these bands, thus pro-
ducing soft regions which eventually became crack sites. They were not
able to reproduce the etching results of Hanstock which had indicated
bands of overaging.
Since these early papers, a number of authors [23-25] have presented
direct evidence including transmission electron micrographs to show that
re-solution does, in fact, occur around slip zones produced during fatigue
of superpurity aluminum-zinc-magnesium alloys. The effect is best dem-
onstrated [25] after prolonged torsion cycling which allows extensive
development of slip bands and zones without appreciable tensile-mode
crack propagation. It has been proposed that repeated shearing of GP
zones and partially coherent if precipitates by moving dislocations re-
duces these particles to a size which is unstable at the temperature of
testing, and that re-solution then occurs.
In this paper, extensive examination of the crack tip plastic zone in
7075-T6 aluminum has not revealed any localized bands of fatigue de-
formation or of precipitate re-solution. Apparently it is not possible to cre-
ate the conditions of repeated dislocation-precipitate interaction in this
alloy which occur during fatigue of superpurity alloys. The presence of
impurities and trace elements in the commercial alloy (Table 1) either
makes the GP zones and rf precipitates less penetrable or they contribute
to a general dislocation pinning throughout the matrix. In any event, slip
bands are not observed to develop on 7075-T6 except at very high cyclic
strain levels. Even in this case only a few hundred cycles can be accumu-
lated before complete failure occurs. In the present work, one 7075-T6
specimen was cycled in bending at very high strains to produce surface
slip bands. Subsequent electropolishing revealed that these bands pene-
trated less than 1 /* into the material. Thin foils taken at the surface and
containing slip bands showed only a general, uniform dislocation distri-
bution with no localized concentration correlating with the surface slip
and no localized re-solution of precipitates. We must therefore conclude
that the fatigue deformation of commercial 7075-T6 aluminum differs
strongly from that of the superpurity alloys of similar composition. Con-
centrated slip into zones or bands and consequent precipitate re-solution
is not an important feature.
One might assume that sufficient dislocation mobility exists on a gen-
eral scale to insure that some precipitate shearing occurs near a crack tip.
However, the number of cycles during which this could occur is quite
small. The size of the plastic zone in 7075-T6 under the stress conditions
reported here is probably in the range 10 to 100 p.. The rate of crack
propagation observed in these tests was about 1 X 10~5 cm/cycle in the
region of interest. Therefore, the precipitate structure in the plastic zone
will have endured, at most, about 1000 cycles of stress reversal before
local Downloaded/printed
fracture occurs. by
It is not likely that much re-solution could occur
in such a short interval, given the limited mobility of dislocations in this

material (Fig. 5b). The results of this paper support this conclusion, al-
though the possibility of a limited reversion, especially of very small GP
zones, cannot be ruled out until detailed, quantitative measurements are
made.
The only other experimental evidence on the metallurgical stability of
commercial alloys of the 7075 type is that of Broom et al [26], who inves-
tigated the fatigue strength of DTD 683 fatigued at room temperature
and at 78 K. From a comparison of the relative fatigue strengths and ten-
sile strengths at these temperatures, they concluded that vacancy pro-
duction in the room temperature tests caused overaging during the test,
while such was not the case at the lower temperature. No direct evidence,
for example, etching patterns, was presented to support this hypothesis.
The evidence of Fig. 6, and many other similar micrographs, is that any
overaging within the crack tip plastic zone is beyond the sensitivity of a
qualitative comparison with uncycled material. An upper limit to the
amount of overaging which could have occurred can be obtained in the
following way. From measurements of precipitate density and size, we
estimate that about 15 per cent of the available zinc and magnesium is
in the GP + t{ precipitate in the uncycled state. If all of the remaining sol-
ute were to be precipitated during fatigue onto the existing particles,
their size would increase by about 100 per cent—certainly observable by
qualitative comparison. Since we estimate a 50 per cent increase to be a
conservative lower limit to our sensitivity, we compute that less than one
third of the remaining solute has precipitated to cause overaging. Any
more precise estimate would require detailed, quantitative measurements
of particle density and size.
Conclusions
The conclusions to be drawn from the work reported here can be sum-
marized as follows:
1. Dislocations near the tips of growing fatigue cracks in 1100-O
aluminum are distributed in a subgrain structure. In contrast, a rather
uniform distribution of dislocation loops are found in 2024-T4; in 7075-
T6, the dislocations are tightly packed in a uniform distribution and show
little sign of motion or interaction.
2. There is no correlation between the dislocation structure just be-
neath the fracture surface and the regularly spaced growth striations nor-
mally found on this surface. The mechanism of striation formation
appears to be noncrystallographic in nature and a function of the macro-
scopic plastic strains which accompany the blunting and resharpening of
the crack tip.
3. Copyright
Dislocation by
densities 6 //. from
ASTM Int'lthe tips
(all of fatigue
rights cracks grown over
reserved); Mon Dec
105 to 106 total elapsed cycles range from about 1010 lines/cm2 in 1100-
O to about 1011 lines/cm2 in 7075-T6 aluminum.
4. No localized reversion or overaging is observed to occur near fa-
tigue cracks in either 2024-T4 or 7075-T6 aluminum. This result is due
to the lack of localized slip zones in these alloys in which prolonged dis-
location-precipitate interaction might occur.
5. The possibility of a general reversion of small GP zones, or of a
limited, general overaging near fatigue cracks is not eliminated by this
study. Quantitative measurements of precipitate size and density in a
large number of specimens would be necessary to settle this point.
A cknowledgments
The authors would like to thank Gordon Gross for helpful discussions
concerning the interpretation of the data.
References
[1] J. C. Grosskreutz, "A Critical Review of Micromechanisms in Fatigue,"
Fatigue—An Interdisciplinary Approach, Syracuse University Press, Syracuse,
1964, p. 27.
[2] F. A. McClintock, "On the Plasticity of the Growth of Fatigue Cracks,"
Fracture of Solids, Interscience Publishers, Inc., New York, 1963, p. 65.
Theory of Infinitesimal Dislocations Distributed on a Plane," International
Conference on Fracture, Sendai, Japan, 1965, p. A-109.
[4] P. C. Paris, "The Fracture Mechanics Approach to Fatigue," Fatigue—An
Interdisciplinary Approach, Syracuse University Press, Syracuse, 1964, p.
107.
[5] J. C. Grosskreutz, "A Theory of Stage II Fatigue Crack Propagation," Air
Force Materials Laboratory, Technical Report 64-415, March, 1965, Wright-
Patterson Air Force Base, Ohio.
[6] G. G. Shaw, "Techniques for Transmission Electron Microscopy at a Fatigue
Crack Tip," Electron Microscopy, Vol 1, Maruzen Co., Ltd., Tokyo, 1966,
p. 325.
[7] Hirsch et al, Electron Microscopy of Thin Crystals, Butterworth, London,
1965.
[8] R. K. Ham, "The Determination of Dislocation Densities in Thin Films,"
[9] R. K. Ham, "On the Loss of Dislocations during the Preparation of a Thin
Film," Philosophical Magazine, Vol 7, 1962, p. 1177.
[10] J. C. Grosskreutz and G. G. Shaw, "Dislocation Rearrangement in Fatigue-
Hardened Aluminum during Preparation for Transmission Electron Micros-
copy," Philosophical Magazine, Vol 10, 1964, p. 961.
[11] G. Thomas and J. Nutting, "The Ageing Characteristics of Aluminum
Alloys," Journal, Institute of Metals, Vol 88, 1959-1960, p. 81.
[12] J. D. Embury and R. B. Nicholson, "The Nucleation of Precipitates: The
System Al-Zn-Mg," Acta Metallurgica, Vol 13, 1965, p. 403.
[14] B. A. Bilby, A. H. Cottrell, and K. H. Swinden, "The Spread of Plastic
Yield from a Notch," Proceedings, Royal Society London, Vol A272, 1963,
304.
p.Copyright by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
[15] R.Downloaded/printed
L. Segall, P. G. by
Partridge, and P. B. Hirsch, "The Dislocation Distribu-
tion in Face-Centered Cubic Metals after Fatigue," Philosophical Magazine,

Vol 6, 1961, p. 1493.
[16] J. C. Grosskreutz and P. Waldow, "Substructure and Fatigue Fracture in
Aluminum," Acta Metallurgica, Vol 11, 1963, p. 717.
[17] J. C. Grosskreutz, "Fatigue Crack Propagation in Aluminum Single Crys-
tals," Journal Applied Physics, Vol 33, 1962, p. 1787.
[18] A. J. McEvily, J. B. Clark, E. C. Utley, and W. H. Herrnstein, "Evidence
for Reversion during Cyclic Loading of an Aluminum Alloy," Transactions,
Am. Institute Mining, Metallurgical, and Petroleum Engrs., Metallurgical
Soc., Vol 227, 1963, p. 1093.
[19] J. B. Clark and A. J. McEvily, "Interaction of Dislocations and Structures
in Cyclically Strained Aluminum Alloys," Acta Metallurgica, Vol 12, 1964,
p. 1359.
[20] G. W. J. Waldron, "A Study by Transmission Electron Microscopy of the
Tensile and Fatigue Deformation of Aluminum-Magnesium Alloys," Acta
Metallurgica, Vol 13, 1965, p. 897.
[21] R. F. Hanstock, "The Reaction of High-Strength Aluminum Alloys to
Alternating Stress," Proceedings, International Conference on Fatigue of
Metals, Institute of Mechanical Engrs., London, 1956, p. 425.
[22] T. Broom, J. A. Mazza, and V. N. Whittaker, "Structural Changes Caused
by Plastic Strain and by Fatigue in Al-Zn-Mg-Cu Alloys Corresoonding to
DTD 683," Journal, Institute of Metals, Vol 86, 1957-1958, p. 17.
[23] I. J. Polmear and I. F. Baindridge, "Precipitate Instability during Unidi-
rectional Extension of an Age-Hardened Al-Zn-Mg Alloy," Philosophical
Magazine, Vol 3, 1958, p. 655.
[24] I. J. Polmear and I. F. Baindridge, "Fatigue Deformation in the Interior of
Aged Ternary Al-Mg-Zn Alloys," Philosophical Magazine, Vol 4, 1959, p.
1293.
[25] C. A. Stubbington, "Some Observations on Microstructural Damage Pro-
duced by Reversed Glide in an Al-7.5% Zn-2.5% Mg Alloy," Acta Metal-
lurgica, Vol 12, 1964, p. 931.
[26] T. Broom, J. H. Molineux, and V. N. Whittaker, "Structural Changes during
the Fatigue of Some Aluminum Alloys," Journal, Institute of Metals, Vol
84, 1955-1956, p. 357.
DISCUSSION
/. F. Throop1 (written discussion)—We are measuring fatigue crack

propagation rates in a 4340 steel, using single-edge notch tension speci-
mens, and monitoring the size of the plastic zone by means of photographs
of Newton rings from optical interference patterns. We find that the plas-
tic zone size at any given crack length is such that about 1000 cycles are
required to advance the crack edge to a point which was initially at the
far side of the plastic zone. This behavior in the steel is in agreement with
your statement that in the aluminum alloys a given site is subjected to
the plastic straining for only about 1000 cycles.
We find that the plastic zone size is related to the crack propagation
rate by a constant factor, at least at stress intensity factor levels less than
the plane strain fracture toughness KIC . For stress intensity factor levels
greater than KIC it may be related to the crack rate by a different con-
stant factor. This difference should be expected, because the shape of the
plastic zone at low stress intensity levels is nearly round in cross section
and corresponds to plane strain conditions, while at high levels the shape
changes to approach plane stress conditions and the effects of greater
shearing action become evident.
I believe that the microstructure observed at the crack tip, particularly
along a line extending ahead of the crack, will depend upon whether the
stress intensity factor at the given crack length is less than or greater
than the fracture toughness of the metal. The latter case should show
the greatest effects of plastic deformation. It would be of interest to know
if your observations show any distinction between low- and high-stress
intensity levels.
/. C. Grosskreutz and G. G. Shaw (authors)—We have observed the
microstructure near the fracture surface of a specimen of 7075-T6 which
was fatigued to failure in 9l/% cycles. Comparison with micrographs taken
from specimens failed at much lower stresses (Nf about 106 cycles) shows
only a greatly increased dislocation density, so dense in fact that quanti-
tative measurements could not be made from the pictures. With the dis-
locations out of contrast, no qualitative difference in metallurgical struc-
ture could be distinguished between the 9]/2 cycle specimen and the 106
cycle specimen.
G. H. Jacoby2 (written discussion)—In Germany we carried out a
1
Research mechanical engineer, Watervliet Arsenal, Watervliet, N. Y.
2
German Research Association and Aerospace Sciences, OVL; at present,
Columbia University, New York, N. Y.
DISCUSSION ON MICROSTRUCTURES 243
comparison of several aluminum alloys (2024, 2020, 2218, 2219, RR58,

and SAP 8G5) under fatigue loads. In this study single level and program
tests with unnotched- and notched-test pieces made from sheets and ex-
truded bars were performed. These tests proved that in general the nickel-
containing aluminum alloys exhibit a higher (under certain conditions a
considerably higher) fatigue life (especially 2218) than 2024 and the
other alloys mentioned above. This fact has lead us to a more detailed
study of the fatigue behavior of nickel-containing aluminum alloys in
several conditions (sheets, clad and unclad, extruded bars with and with-
out extrusion effect) with and without notches.3 In this study 2024 as the
main aircraft material has been included for a direct comparison. Single
level and program tests, as well as tests at low temperature, under con-
trolled high humidity and at different test frequencies were performed. In
the analysis of the tests, life crack initiation and crack propagation were
considered. The results of this study strongly support overaging theories
for the fatigue mechanism in aluminum alloys. The good properties of
nickle-containing aluminum alloys may be attributed to the fact that
overaging under fatigue loads is rendered more difficult than in 2024,
which coincides with the behavior of these materials at higher tempera-
tures. This statement has been deduced only from the materials behavior
under the different test conditions. We were not able to support it for
the materials under test by direct evidence of overaging, for example,
etching patterns, as we did in an earlier study with an aluminum-zinc-
magnesium alloy.
I have learned from the paper of the authors that they were not able
to detect by their method evidence for overaging (or re-solution) in 2024
and 7075 aluminum alloys. I would like to ask them if they nevertheless
would support theories which attribute the fatigue mechanism in alumi-
num alloys mainly to metallurgical instabilities?
Messrs. Grosskreutz and Shaw—We would have to state that metal-
lurgical instability is not an important part of Stage II crack propagation
in 2024-T4 or 7075-T6 alloys, unless the instability involves reversion
or growth of second phase particles below about 60 A. Unpublished
work carried out in our laboratory also shows that there is no observable
(on an electron microscope level) overaging of 2024-T4 during crack
nucleation after as much as 106 cycles. In view of these results, is it diffi-
cult to support theories which attribute the poor fatigue properties of
these alloys exclusively to metallurgical instability.
3
G. Jacoby and R. Luxemburger, "Zerriittungsfestigkeit Ni-haltiger Aluminium-
legierunger," to be published in Aluminum.
The Continuum Approach to Fatigue
Crack Growth
J. R. Rice1
Mechanics of Crack Tip Deformation and

Extension by Fatigue
REFERENCE: J. R. Rice, "Mechanics of Crack Tip Deformation and

Extension by Fatigue," Fatigue Crack Propagation, ASTM STP 415, Am.
Soc. Testing Mats., 1967, p. 247.
ABSTRACT: This paper surveys the continuum mechanics of deformation
near cracks and its application to propagation by fatigue loadings. Re-
sults of various elastic-plastic models are summarized and compared in
relation to hardening behavior, size effects, and large scale plastic yield-
ing. The role of the elastic stress intensity factor variations in governing
local plastic flow, and thus crack growth rates, is emphasized for the
common high-cycle low-stress fatigue situations. General features of
crack propagation are discussed, and theories are examined which seek
to relate continuum analyses to separation mechanisms.
KEY WORDS: fatigue (materials), crack propagation, fracture mechanics,
plasticity, strain concentrations, size effect, hardening
This work views crack propagation as primarily a problem in con-

tinuum mechanics. Part I surveys the elastic and elastic-plastic stress
analyses of cracked bodies, with emphasis on the plasticity. In addition
to well-known results based on models of perfectly plastic anti-plane
shearing and discrete surfaces of tensile yielding or slip (equivalently,
continuous dislocation arrays), some recently obtained results on work-
hardening and anisotropic perfect plasticity are summarized, and methods
are presented for the modeling of plane strain yielding. Emphasis is
placed on the common result of all plasticity analyses that the coefficient
of a characteristic singularity in elastic solutions determines the plastic
deformation in situations of small scale yielding. The influences of hard-
ening behavior, finite width effects, and large scale yielding are illustrated
and the predictions of various models compared.
Part II considers the mechanics of fatigue crack propagation. Elastic-
plastic responses to cyclic loading are determined for perfectly plastic
1
Assistant professor of engineering, Brown University, Providence, R. I.
247
and a type of stable hardening behavior. Effectively, the yield stress is

doubled so that cyclic flow zones and variations in plastic deformations
are smaller than for monotonic loading. Crack tip blunting by large de-
formations and related effects are treated approximately. General features
of fatigue crack growth are surveyed, and the extensive evidence is cited
supporting a primary conclusion of continuum analyses: that crack
growth rates are determined by elastic stress intensity factor variations
for the small scale yielding situation common in low-stress high-cycle
fatigue. Results pertaining to mean load, sheet thickness, and mode transi-
tion effects, delays in crack extension due to overloads, growth under
bending loads, and growth by random applied loads are also noted and
interpreted in continuum terms. Theories of crack growth relating con-
tinuum considerations to "damage" accumulation and material separation
are examined. Further progress requires better continuum analyses, in-
corporating crack blunting by deformation, and clearer ideas of separa-
tion mechanisms.
Part I Continuum Mechanical Description of Deformation Near

Cracks
The fundamental importance of the origin, stable growth, and final
propagation of cracklike flaws in material failure has stimulated a rapidly
increasing amount of research on continuum mechanical descriptions of
the stress and deformation fields in cracked bodies. Advances in the
elastic and elastic-plastic modelling of crack tip deformations for various
geometries and methods of loading are summarized in this part of the
paper, and applications of these results in establishing a mechanics of
fatigue crack propagation are discussed in Part II. Primary attention
is given to elastic-plastic stress analyses, as no survey of the considerable
(but as yet far from complete or even reasonably satisfactory) work in
this area is currently available and several excellent surveys of the
purely elastic treatment of stress concentrations (Neuber [I],2 Savin [2]),
and cracks in particular (Irwin [3], Paris and Sih [4]), have appeared.
The objectives of continuum analysis in developing a theoretical frame-
work for crack propagation by fatigue or catastrophic fracture are essen-
tially twofold. First, for various methods and histories of loading struc-
tures containing cracks of different sizes and orientations, we seek to
determine functions of the applied load and geometry which describe the
local deformations suffered by material near the crack tip. Postulating
dependence of fracture on this local deformation, parameters of im-
portance in crack propagation and expected characteristics may be iden-
tified for the organization and analysis of experimental results, and if
2
paper.
RICE ON MECHANICS OF CRACK TIP DEFORMATION 249
similarities exist in the local deformation fields for a class of configura-

tions, experimental data on one configuration may then be rationally ex-
tended to the prediction of behavior of others. As a second objective, we
envision continuum solutions as setting the boundary conditions on
microstructural processes submerged in the intense deformation field
near the crack tip and resulting in material separation, and seek to pre-
dict crack propagation behavior on the basis of models for accumulation
of "damage" and final separation. As will be seen, considerable progress
has and is being made along the lines of this first objective. Although
many theories have been proposed (see Part II), it appears that presently
the second objective of relating microstructural models of separation to
FIG. 1—Crack of length 2a in infinite plane; uniform stress at infinity.
continuum deformation analyses, with subsequent predictions of crack

growth laws, is inadequately treated.
We begin by considering the elastic analysis of cracks and then take up
various plasticity models for crack tip deformation, starting with the case
of anti-plane strain, a case of little direct physical interest but one in
which the mathematical simplicity allows gaging the effect of various
types of plastic behavior. Then models based on discrete surfaces of
yielding or slip, which appear most appropriate for plane stress, and a
new model based on a slip line field for plane strain are considered.
Originally elastic-plastic solutions for monotonically increasing loads
are given. Solutions for unloading, cyclic loading, and general load-time
histories, as of interest in fatigue crack growth, are then given separately
in Part II. Only two-dimensional problems are considered, and primarily
the case of a single finite crack of length 2a in an infinite plane subject to
a uniform stress state at infinity (Fig. 1) is treated in detail, although ref-
erence is made to several other configurations.
Linear Elastic Stress Analysis of Cracked Bodies

The characteristic feature emerging from all linear elastic solutions
to problems of cracks in homogeneous isotropic materials is the inverse
square root character of the crack tip stress singularity [4]. In particular,
components of the stress tensor (Tij(<Txx = ax , <jxy = Txy , <jyy = ay , etc.)
are representable in the form [4]:
+ other terms nonsingular at the crack tip

where, as for example in Fig. 1 , r and 6 are polar coordinates introduced
at the crack tip. Here the symbols I, II, and III pertain, respectively,
to parts of the crack tip region stress field corresponding to displacement
discontinuities induced along the crack surface in the tensile, in-plane
shear, and anti-plane shear directions respectively, as in Fig. 2. The
dimensionless functions/^ ,/" , and /"* depend on the orientation angle
FIG. 2—Modes of crack tip deformation.
6 only, and K ^ , Ku , and Km are the stress intensity factors, introduced

into fracture mechanics by Irwin, and again pertaining to the three
modes of crack tip deformation. The stress intensity factors may be
determined from complete solutions to boundary value problems after
an examination of the crack tip stress field and comparison with Eq 1;
coming from a linear theory, they must be linear functions of the applied
loads. Further, from dimensional considerations the K's have dimensions
of F L~*12 (F denoting force and L denoting length). Thus if a cracked
body is loaded with a remotely applied stress (dimensions of F L~2)
and crack length is the only characteristic length, the K's are propor-
tional to the stress times square root of crack length. If a crack is loaded
by a concentrated force per unit thickness (dimensions of F L"1) applied
at the crack center, as for crack branches of equal length emanating
from a rivet hole in a sheet, so that crack length remains the only char-
acteristic length, the K's are proportional to the force per unit thickness
times the inverse square root of crack length. The general conclusion
following from Eq 1 is that, to the extent linear elasticity is appropriate,
all crack tip stress fields are of identical functional form with the in-
fluence of the particular magnitude and method of loading, and geome-
try of the cracked body, sensed in the crack tip region only through the
stress intensity factors K ^ , Ku , Km .
For Mode I (tensile) crack tip deformations, the flj(6) are readily
identified from the equations [4] (with axes as in Fig. 1)
(generalized plane stress)

(plane strain, v is Poisson ratio)
and similar equations may be written for Modes II and III. The stresses
(Ty(x, 0), rxy(x, 0), and rzy(x, 0) represent the tractions acting (for
x > 0) in the y, x, and z directions along the prospective fracture surface.
For Mode I
for Mode II
and for Mode III
For example, with remotely applied stresses and a crack of length la as

in Fig. 1,
Other stress intensity factors for a wide variety of crack configurations,

including several typical experimental specimens, as well as an exhaustive
list of references to workers on elastic crack stress analyses, are cata-
loged in the paper by Paris and Sih [4].
Limiting attention to two-dimensional problems in homogeneous iso-
tropic materials, Muskhelishvili [5] has shown that for in-plane loadings
(Modes I and II) the stresses and displacements are expressible as
where the equations are given in cartesian and polar form, the u's are
displacement components, $(£") and \f/(£) are analytic functions of the
complex variable f = x + iy, a bar over a quantity denotes its complex
conjugate, Re means real part, G is the shear modulus, and K = 3 — 4v
for plane strain and K = (3 — v ) / ( l + v) for generalized plane stress.
All other stresses vanish except in the case of plane strain for which
<rz = v(0x + <Ty). Similarly, for anti-plane loadings (Mode III)
with all other stress and displacement components vanishing.

As an example, for the configuration of Fig. 1 with a uniform remote
stress state, the complex stress potentials are [5]
where 3>(f = $(f) and the branch cut of f~1/2(f + 2a)~1/2 is chosen
along the crack line so that the combination behaves as ^ for large £.
That the stress intensity factors for this configuration are correctly given
by Eq 4 may be checked by computing the crack tip stresses from Eqs
5 and 6 and comparing with the singular forms of Eqs 3 which serve to
define the stress intensity factors. Paris and Erdogan [6] have pointed
out the relation between the stress concentration factors popularized
by Neuber [7] and Irwin's stress intensity factors. For an ellipse of length
2a and end radius of curvature p subjected to an in-plane tension (<ry)x ,
the maximum concentrated tensile stress is
Similarly, the maximum concentrated shear stress due to an anti-plane

shear is
Comparing these with Eqs 4, for p « a (narrow ellipse) the maximum

concentrated stresses and stress intensity factors are related by
in the tensile mode, and by
in the anti-plane shear mode.

To the extent that inelastic behavior is of no major influence, the
surrounding elastic crack tip stress fields are identical for two configura-
tions if their stress intensity factors are equal, and thus if material prop-
erties are identical, the stress and deformation distributions in the in-
elastic regions are presumably identical. This conclusion of elastic stress
analysis is of obvious importance; it embodies a large part of the progress
FIG. 3—Approximate calculation of plastic zone size.
of modern fracture mechanics. The deficiencies of purely elastic analysis

are equally clear, necessitating analyses based on more realistic material
models allowing for plastic flow. Quantitative estimates of when the
elastic singularity controls local deformations are needed and new pa-
rameters must be established as controlling these deformations when
plasticity is on a large enough scale to wipe out the characteristic sur-
rounding elastic field. The physical differences between plane strain and
plane stress states, very significant in the presence of plastic yielding, are
inadequately reflected in purely elastic treatments. Details of stress and
strain distributions in separation prone material at the crack tip ob-
viously require a plasticity treatment for reasonable descriptive accuracy.
Serious crack blunting can only result from large strains of plastic, not
elastic, magnitude. Finally, plastic deformations are history dependent
and elastic deformations are not of particular significance in fatigue
where separations result from nonmonotonic deformation histories.
Rough estimates of the scale of plasticity are obtainable from simple
manipulations based on the elastic stress field. Consider the tensile mode
as depicted in Fig. 3 and suppose that the plastic region, of linear di-
mension o>, is imbedded in the field of the elastic singularity. Let a0 be an

effective yield stress, as modified by a degree of hardening or by hydro-
static stresses. As in Fig. 3, the yielded region should extend over roughly
twice (due to stress redistribution) the distance from the crack tip at
which the dashed line singular elastic stress field equals the yield stress.
Then from Eq 3a,
This estimate of plastic zone size is surprisingly accurate for well con-
tained plasticity. Similar formulas follow for the shear modes with a0
replaced by r0, a yield stress in shear.
The plastic zone size (Eq 9) establishes a geometric dimension indi-
FIG. 4—5/naW scale yielding near crack; crack may be viewed as semiinfinite
with inverse square root stresses approaches at large distances.
eating the region over which deviations from elastic behavior occur.
Now the characteristic length associated with the elastic stress field is a
dimension such as crack length, uncracked width of a finite specimen,
distance from crack tip to points of load application, and so forth. Thus
at load levels sufficiently low so that the plastic zone size computation
of Eq 9 gives a length small compared to all such dimensions, the plas-
ticity may be expected controlled by the elastic stress intensity factor.
We call this situation "small scale yielding." Conversely, when the length
predicted by Eq 9 is comparable to or greater than such geometric
dimensions, a correction to Eq 9 is required as the stress intensity factor
may no longer be expected to control the plasticity. These conclusions
are borne out in all the plasticity models examined subsequently. As
might be expected, the nonlinearity inherent in elastic-plastic analysis
causes considerable complexity in the determination of deformation
distributions even for the simplest of models. Since our interest is fre-
quently in situations of small scale yielding where the stress intensity
factor dominates, it is of interest to inquire as to whether simpler analyti-
cal^ methods may be established yalid only, for this case and not for the
entire range of large scale .yielding. A concept of a boundary layer ap-
proach has emerged in this connection in some recent work by the author
[7-70]. In the 1'mit when plastic region dimensions are negligible com-
pared to geometric dimensions, the surrounding elastic singularity sets
the boundary conditions on the elastic-plastic boundary value problem,
in the sense that the plastically yielding material only "sees" the sur-
rounding stress field through the inverse square root term in the elastic
solution, this stress field being approached by the elastic-plastic solution
at distances large compared to the plastic zone size but still small com-
pared to other geometric dimensions. Thus the small scale yielding
solution for any loading (Fig. 40) may be obtained by considering a
semi-infinite crack (Fig. 4b) with the asymptotic boundary conditions
that the inverse square root elastic stress field is approached at large
distances:
Here K is the stress intensity factor pertaining to the particular manner

of loading and deformation mode.
Letting $s(f), ^«G"), and Q,(f) denote the form of the complex stress
potentials when only the singular term is included,
where the crack tip is at f = 0 and the crack surface extends along the
negative real f axis. The asymptotic boundary condition for small scale
yielding is equivalent to requiring that the complex potentials approach
the appropriate above form as | f | —> °o.
Anti- Plane Shearing

The third mode of crack tip deformation, anti-plane shearing, is of
little practical importance as fatigue cracks generally tend to initiate
(if not already present) through an in-plane shearing mode involving
repeated deformation in slip bands [11,12], and to propagate in a tensile
mode. The motivation for dealing with this case is primarily the lack of
comparable progress in the analysis of the tensile mode. Exact mathe-
matical solutions are now available for the anti-plane shearing of cracks
in perfectly plastic materials, including both plastically isotropic (Tresca
or Mises yield condition [13] governing) and anisotropic (including the
special case of single crystals with discrete slip directions) materials,

and for work hardening materials with arbitrary relations between
principal shear stress and shear strain in the work hardening range.
Attempts at drawing direct analogies between detailed features of anti-
plane shearing and tensile crack tip deformation fields appear largely
unjustified. However, the two cases share commonly the principal
features of load transmission around a traction free surface in materials
which will support limited stresses, and since Eqs 3, 4, and 8 predict
notable similarities in purely elastic response, one confidently expects
analogies between gross features of plasticity effects in the anti-plane
and tensile modes. Thus, for example, anti-plane predictions of the
dependence of plastic zone sizes and crack opening displacements on
applied loads and geometrical dimensions, as well as general conclusions
drawn from variations of the yield surface and introduction of hardening
FIG. 5—Yielded region near crack in perfectly plastic Tresca or Mises material
subjected to anti-plane shearing.
behavior, are likely good approximations for the tensile case. McClin-
tock and Irwin [14] have recently discussed this point.
Hult and McClintock [75] gave the form of the stress and strain
distribution in the plastic region adjoining the crack tip for a perfectly
plastic material satisfying the Tresca or Mises yield condition (coinci-
dent in this case),
that the principal shear stress not exceed the yield stress r0. Referring
to Fig. 5, in the plastic zone
where y0 = r0/G is the yield strain and R(0) is the radial distance to the
elastic-plastic boundary. The plastic zone size, co, and crack opening dis-
placement, u0 , are
The small scale yielding solution for this case was first found by Hult and
FIG. 6—Plastic zone size as a junction of net section stress for various crack
length to width ratios; anti-plane shearing of perfectly plastic Tresca or Mises
material.
McClintock [15] and further elaborated by Irwin and Koskinen [16] and
Rice [8], The plastic zone is circular in shape with
Shear stresses in the elastic region outside the circular plastic zone are
given by [8,16]
Comparing with Eq lie, the effect of plasticity is to shift the purely

elastic stress distribution ahead by half the plastic zone size.
For the crack of length la in an infinite body configuration of Fig. 1,
FIG. 7—Crack opening displacement as a function of net section stress for

various crack length to width ratios; anti-plane shearing of perfectly plastic Tresca
or Mises material.
loaded with a single remotely applied anti-plane shear stress (Tyz)x = T^ ,

the plastic zone size and crack opening displacement are [8,15]
where s = rjr0 and E\, Ez are complete elliptic integrals of the first
and second kind. Noting that KUI = rîra)112, these may readily be
shown to reduce to the small scale yielding results of Eq 13 a at low stress
levels. The shape of the plastic zone is initially circular and it elongates
much as in Fig. 5 at higher stress levels until at the limit load T^ = r0
the zone extends to infinity in the x direction with a height in the y
direction approaching [8] asymptotically to 4a/ir. Koskinen [17] first
treated the configuration of Fig. 1 for the case of a finite, rather than
infinite, width in the x direction. While his solution was based on a
numerical finite-difference scheme, Rice [8] later provided an analytic
solution. When the width is 2b and the crack is centrally located the
stress intensity factor is [4]
The same solution applies for an edge crack of length a in a plane of

width b, or for double edge notches of depth a in a plane of width 2b.
These configurations are only approximately equivalent in the tensile
case (for example, K^ tt 1.1 a-îra}112 for an edge crack in an infinite
plane). The bracketed term of Eq I5a provides a good approximation,
which can be improved upon [4], to the finite width correction in the
tensile case. The graphs of Figs. 6 and 7 give dimensionless plots of the
plastic zone size and crack opening displacement for several crack length
to thickness ratios, a/b, in terms of the ratio of the net section stress,
r
n = T M (1 — a/b~)~l, to the yield stress. Using Eqs 13a, \3b, and I5a,
the small scale yielding solution for which the stress intensity factor
dominates leads to
The dimensionless plastic zone size
appears as a function of dimensionless net section stress, rn/T0 , in Fig. 6

for crack length to plane thickness ratios, a/b, of 0, 1/5, and 3/5. The
dimensionless crack opening displacement
appears as a function of the same parameters in Fig. 7. From Eq 156,

both of these dimensionless parameters equal the square of the net
FIG. 8—(a and b) General formulation for anti-plane shearing of elastic per-
fectly plastic material with crack; (c and d) special features of solution when yield
surface contains straight line segments corresponding to restricted slip directions
(discrete slip lines formed at crack tip).
section stress to yield stress ratio according to the small scale yielding
solution, and this is shown by the dashed lines. The heavy lines are the
results of exact computations [8] which do not make the small scale
yielding approximation. The bracketed factor in the above two-dimen-
sionless expressions equals 1.00 for a/b = 0, 1.51 for a/b = 0.20, and
4.28 for a/b = 0.60, as indicated on the graphs.
Considering Fig. 6 first, prediction of the plastic zone size by the small
scale yielding solution (Eq 156) is seen to be accurate up to 30 per cent
of the limit load (T»/TO = 1) for a/b - 0, up to 40 per cent for a/b =
0.20, and up to 50 per cent for a/b = 0.60. At higher stress levels the
stress intensity factor is not even approximately descriptive of local
conditions at the crack tip. It should be cautioned that these results
are for monotonic loading only. As will be seen later, cyclic loadings as
for fatigue produce a smaller zone of cyclic plastic deformation for which
the same curves are valid but with Tn/r0 replaced by the cyclic variation
in net section stress divided by twice the yield stress, so that the stress
intensity factor is descriptive of local cyclic conditions up to much
higher stress levels. The crack opening displacement (Fig. 7) is seen to
be accurately predicted by the small scale yielding formula (Eq 15&)
up to much higher stress levels, with the stress intensity factor failing to
be descriptive of local conditions above about 60 to 70 per cent of the
limit load.
Rice [10] has recently generalized the procedures employed [8,15,17]
for obtaining the solutions discussed above, by formulating the anti-
plane shearing problem for perfectly plastic materials with yield surfaces
of arbitrary convex shape in the two-dimensional TXZ , ryz shear stress
space (the Tresca or Mises yield condition (Eq 12a) then being the special
case of a circle of radius TO in this space). The principal features of the
solution for an edge crack of depth a in an infinite body (or equivalently
the crack of length la in an infinite body configuration of Fig. 1), sub-
jected to a remotely applied anti-plane shear stress TW , are summarized
in Fig. 8. A yield surface appears in Fig. 8a, the cracked body in Fig.
86. For the remotely applied stress in the direction shown, stresses in the
plastic zone correspond to points on the upper part (ryz > 0) of the
yield surface. When the yield surface contains no straight line segments
the stress components are constant along radial lines in the plastic zone,
so that along a line from the crack tip making an angle 9 with the x
axis the stresses have those values corresponding to the point on the
yield surface for which the tangent line makes the same angle 9 with the
TXZ axis. Strains exhibit a \/r singularity at the crack tip, and the plastic
part of the strain has a direction perpendicular to the radial lines. The
elastic portion of the physical x, y plane may be shown [10] to map
into the region of the stress plane between the rxz axis and the upper
part of the yield surface, with corresponding points as labeled by the
capital letters in Figs. 8a and b. An inverse solution for physical co-
ordinates in the elastic region as a function of stress follows by intro-
ducing a potential function 00 = <J>O(TXZ , ryz). Then
and
and the harmonic function </>„ vanishes on the TXZ axis and the upper part
of the yield surface, as in Fig. 8a. Along the TVZ axis from the origin out
to a stress distance equal to the remotely applied stress,
FIG. 9—Small scale yielding solutions for anti-plane shearing of perfectly

plastic materials: (a) Tresca-Mises material, (b) orthogonal allowed slip directions
at 45 deg with crack line, and (c and d) single allowed slip directions.
The distance R from the crack tip to a point on the elastic plastic bound-
ary, corresponding to a given point on the yield surface, is found from
[10]
where d(f>0/(dn) is the derivative of <f>0 in the direction normal to the yield
surface. A membrane analogy is readily established allowing eifective
visualization of the solution; <£0 may be viewed as the deflection of a
membrane subjected to zero transverse pressure and stretched out over
an opening in a sheet corresponding to the region between the upper
part of the yield surface and the rxz axis of Fig. 8a. According to Eq I6b,
the membrane is loaded by bringing a thin wire of length corresponding
to TX into contact with the membrane along the ryz axis, such that the
wire has a downward slope corresponding to crack length a. Equation
16c then indicates that the distance R to a point on the elastic plastic
boundary corresponds to the slope of the membrane in the direction
normal to its fixed boundary. Thus, for example, it is readily understood
that as the remotely applied stress is increased toward the limit load,
the plastic zone tends to elongate in a direction corresponding to the
yield surface tangent at the point where the yield surface intersects the
ryz axis.
When the yield surface contains straight line segments, as in Fig. 8c,
the above formulation of Eqs 16 remains valid, but it is now meaning-
less to speak of plastic strains as the plastic zone coalesces into discrete
slip lines across which the anti-plane displacement uz has a discontinuous
jump. As in Figs. 8c and d, a discrete slip line emanating from the
crack tip corresponds to each straight line segment of the yield surface;
a continuous field of plastic strain joins separate discrete slip lines when
a corner of the yield surface is rounded as at point H of Fig. 8c. Straight
line segments on the yield surface result when only certain directions
of plastic shearing are allowable, as in a single crystal or for events on
the scale of a single grain rather than a polycrystalline aggregate. The
allowed slip surfaces have the direction of the straight line segment, and
the perpendicular stress distance from the origin of the stress plane to
the straight line segment is the resolved shear stress required for slip.
Studies are currently underway on the relevance of similar conclusions
for the tensile deformation of materials with restricted slip directions.
Figure 9 pictures four yield surfaces and the corresponding plastic
zones obtained by Rice [10] in a solution method developed for small
scale yielding. The plastic zone dimension, co, and crack opening dis-
placement, u0, have already been given in Eqs 13 for the circular Tresca-
Mises criterion of Fig. 9a. For the diamond-shaped yield surface of
Fig. 9b, corresponding to slip under a resolved shear stress TO on planes
inclined at 45 deg with the crack line, plastic zone size (slip line length)
and crack opening displacement are [10]
where X « 1.8541 is the complete elliptic integral of the first kind with
modulus %> Allowable slip surfaces are parallel to the crack surface in
Fig. 9c; in this case
Similarly, for allowable slip surfaces perpendicular to the crack surface

as in Fig. 9d,
FIG. 10—(a and b) Elastic-work hardening plastic anti-plane shear stress-

strain relation, and (c) polar strain co-ordinate system (y is magnitude of strain
vector, <f> is angle between strain vector and y axis).
Comparing these last three results and Eq 13a, the particular shape of the
yield surface is seen not to be of great influence in determining the extent
of plasticity, in spite of the radically different plastic zones formed.
Neuber [18] first pointed out the possibility of obtaining anti-plane
shear stress distributions in materials with nonlinear stress-strain rela-
tions of the elastic work-hardening plastic type. His results were limited
to the "small scale yielding" case, in our present terminology, although
this has unfortunately not been realized by some investigators (see the
subsequent discussion on stress concentrations due to smooth ended

notches). Rice [9] provided a hardening anti-plane shear solution for
sharp edge notches, including the finite crack configuration of Fig. 1
as a special case, through a procedure valid so long as the remotely
applied shear stress TM does not exceed the initial yield stress r0, which
marks the onset of plastic flow. These solutions are of the deformation
plasticity type (that is, indistinguishable from nonlinear elastic solutions)
rather than the physically appropriate incremental type. The principal
shear stress and principal shear strain,
FIG. 11—Geometry of small scale yielding near a crack tip for arbitrary
relation between anti-plane shear stress and strain in the work hardening range.
are assumed uniquely related to one another by

for for
where r = r(y) is the equation of the plastic portion of the stress strain
curve (Fig. 10a). Component forms of the stress strain relations are
equivalent to assuming that the stress and strain vectors are colinear
(Fig. 10ft).
A polar strain coordinate system is employed (Fig. lOc) where y is the

magnitude of the strain vector and </> the angle between the strain vector
and the y axis. Much as in the perfectly plastic case, physical coordinates
x and y are related to derivatives of a potential function ^ = \{/(j, 0)
by [9,18]
anti-plane displacements are
The potential function satisfies
and appropriate boundary conditions may be set [9] as suggested by the

labeling of Fig. lOc.
The small scale yielding solution views the crack as semi-infinite and
imposes boundary conditions that the elastic inverse square root sin-
gularity (Eq lie) be approached at large distances. In this case for
7 > To (that is, for points in the plastic zone) the potential function is
[9J8]
A geometrical interpretation of this solution is given in Fig. 11. Upon

introducing the functions of strain
physical coordinates are related to the strain vector and its orientation by
As in Fig. 11, these require that lines of constant strain 7 in the plastic
region be circles of radius R(y) with center a distance ^(7) ahead of the
crack tip. The direction angle 0 of the strain vector at any point on a
constant strain circle is one half the angle made with the x axis by a line
from the center of the circle to that point. Similarly, the elastic-plastic
boundary is a circle with center at X(y0) and radius R(yo) = Klu/(2irT^) ;
this radius is independent of the form of the work hardening stress strain
curve. The plastic zone extends a distance R(y0) + X(y0} ahead of the
crack tip and R(y0} — X(y0) behind the crack tip. From Eq 2lb, since
0 = 0 on the line in front of the crack, distance x from the crack tip
and strain yyz(x, 0) are related by
Stresses in the elastic region outside the circular plastic zone are given by
so that, as noted earlier for the special case of perfect plasticity, the effect
of yielding is to shift the elastic singularity stress distribution ahead a
distance equal to that between the crack tip and center of the plastic
zone. Lines of constant strain 7 in the elastic region remain circular, but
not concentric with the elastic plastic boundary (Fig. 11).
As an example, for a stress-strain relation following a power law
in the work hardening range,
Thus the plastic zone extends a distance
ahead of the crack tip, and on the line ahead of the crack in the plastic
region (Eq 22)
When the small scale yielding approximation is inappropriate, the

solution for the potential function \f/ (from which physical coordinates
are obtained by Eq \9a) in the plastic region (y > y0} takes the form
[9]
Here the set of functions/„(7), n = 1, 2, 3, • • • , satisfy the ordinary

differential equations
FIG. 12—Anti-plane shear of power law hardening material; N = 0.1.
FIG. 13—Anti-plane shear of power law hardening material; N = 0.3.
with boundary conditions

The set of coefficients Dn has been determined [9] for the edge crack
configuration of Fig. lOc when the remotely applied stress is below the
initial yield stress; they are linear in crack length and rather complicated
functions of the form of the hardening stress-strain curve and the ratio
of remotely applied stress to initial yield stress. Extensive numerical
tabulations of the coefficients Dn have been published [9] for materials
hardening according to the power law of Eq 24. Figures 12 and 13 show
some of the final results in graphical form. In Fig. 12 the hardening
FIG. 14—Plastic zone dimension as a function of net section stress and harden-
ing exponent; anti-plane shear of material following T = TO (7/70)*' in plastic range.
exponent is N = 0.1; the position of the elastic plastic boundary (lower

right quartile of figure) and strain distribution in the plastic region
ahead of the crack tip (upper right quartile) are shown for TX = 0.6 TO
and T^ = 0.8 r 0 . These remotely applied stress levels were chosen for
illustration as they nicely typify the transition from the circular plastic
zones of small scale yielding to the highly elongated plastic zone extending
out to infinity as the remotely applied stress approaches the initial yield
stress. The same results are graphed in Fig. 13 for a higher hardening
exponent of N = 0.3.
The variation of plastic zone dimension co with applied stress and
hardening exponent N is shown in Fig. 14 for the edge crack of depth a
in an infinite body of material hardening according to the power law of
Eq 24. At low stress levels the small scale yielding Eq 256 applies,
with Km = T00(7ra)1/2. Significant deviations occur above about 30 per
cent of the initial yield stress, the range for which the curves are drawn.
The curve labeled N = 0 corresponds to perfect plasticity, and is simply
Eq 140. The other extreme, N = 1.0, corresponds to linear elastic be-
havior. Here the plastic zone dimension was computed directly from the
elastic solution of Eq 7 by setting w equal to the distance ahead of the
crack tip at which ryz(x, 0) = TO , resulting in
It is of interest to note that the relatively high hardening exponent of

N = 0.3 gives a result about as close to the linear elastic prediction as
to the perfectly plastic.
FIG. 15—The anti-plane solution for a crack or sharp notch also provides a
solution for a family of smooth ended notches.
Neuber [18] has indicated that an anti-plane stress distribution for a

crack or sharp ended notch also provides the stress distribution for a
family of smooth ended notches. The stress free boundary condition is
that the stress vector be tangent to the boundary. Thus the family of
stress trajectories for, say, the crack solution is the family of correspond-
ing smooth ended notches. From Eqs I9a, x and y are known as a
function of 7 and <£. Thus the stress trajectory equation (see Fig. 15)
becomes a differential equation for 7 as a function of </> along a trajectory.

Inserting initial conditions 7 = 7max (maximum strain, occurring at
notch tip) when 0 = 0, 7 is determined as a function of <£ and thus the
x and y coordinates of points on the notch surface are determined as a
function of (/>. Upon solving for the radius of curvature p at the notch
tip, one finds after some computations that
For the small scale yielding solution, this turns out to be (Fig. 15)
Thus, defining rmax = r(7max) as the maximum concentrated strain, one

finds Neuber's result [18] that for the family of smooth ended notches
generated in the way described above and having the same root radius
of curvature, the product of the jrnaximum concentrated stress and
strain is a constant.independent of the stress:strain curve; ^
Equivalently, the product of the stress and strain concentration factors

has the same value as for the linear elastic case. Neuber unfortunately
FIG. 16—The effect of yielding is viewed as yield level stresses restraining an

extended portion of the crack surface.
failed to adequately emphasize the restriction of his result to what is

here called the small scale yielding case; some investigators [19,20]
have attempted to employ this result in the net section yielding range
where it is incorrect. Rhee and McClintock [21] have pointed out that
in the perfectly plastic case the family of stress free boundaries are circu-
lar arcs in the plastic region and
Were the Neuber result to hold at all stress levels, u> would have to be
given always by the small scale yielding expression. We see from Fig. 6
that this is not the case, and indeed at 80 per cent of the limit stress
Neuber's result (Eq 296) is incorrect by a factor of about three for the
edge notch in an infinite body. More generally, employing the solution
of Eq 26a valid also at high stress levels, Eq 286 leads to
for the relation between root radius and maximum concentrated strain.
Discrete Surfaces of Slip or Tensile Yielding

Several authors have proposed approximate treatments of crack tip
plasticity based on a supposition that plastic flow is adequately modelled
as either slip or tensile yielding on discrete surfaces emanating from the
crack tip. Barenblatt [22] first considered models of this type, although
his original application was to the influence of molecular "cohesive"
forces on the form of deformation near a crack tip in brittle materials.
Later Dugdale [23] replaced the cohesive forces of the Barenblatt theory
with yield level stresses restraining the crack opening over an extended
portion of the crack surface equal to the plastic zone dimension. Bilby,
Cottrell, and Swinden [24] examined the same model but approached the
mathematical problem from the point of view of a continuous array of
dislocations distributed on a plane containing the crack and its yield
zone. Further studies of models based on discrete surfaces of slip or
tensile yielding have been made by Goodier and Field [25], Field [26],
Smith [27], Rice [7], Keer and Mura [28], and Hahn and Rosenfield
[29].
Figure 16 pictures the basic idea for the case of a discrete surface of
tensile yielding ahead of the crack, as appropriate for the tensile mode.
The influence of yielding is viewed as effectively extending the crack a
distance co ahead of its tip, with yield level stresses a0 acting to restrain
the extended crack surfaces. The computation is then entirely elastic, and
the plastic zone size is determined as that length co which makes the
stresses bounded at the outer tip of the plastic zone. For the tensile
mode, the small scale yielding solution which asymptotically approaches
the crack tip elastic singularity leads to [7]
The complex potentials from which stresses and displacements may be

computed (Eq 5) are
taking the branch cut along the crack line and plastic region,
— «> •< x < a. Displacements uy(x, 0) of the extended crack surface
in the plastic zone are given by [7]
and the crack opening displacement uy(Q, 0) = u0 is
Here displacements are measured from zero along the crack line so that
2uy(x, 0) is the total displacement between the upper and lower crack
surfaces.
Similar results follow from the small scale yielding solution for in-
plane shear. Replacing the tensile yield stress <TO in Fig. 16 by a shear
yield stress TO , the plastic zone size is
An equation identical to Eq 31c results for the sliding displacement

ux(x, 0) in the plastic region, with a0 replaced by r0 and w as above.
The crack opening displacement wx(0, 0) = u0 is
Similarly for the case of anti-plane shear [7],
The sliding displacements uz(x, 0) in the plastic zone are again given by
an equation identical to Eq 31c, but with the factor outside the bracket
replaced by 4T0w/TrG, and the crack opening displacement «2(0, 0) = u0
is
This solution is, of course, identical to that of Eqs lib resulting from
the highly anisotropic anti-plane shear yield surface of Fig. 9c. It is seen
that anti-plane shear results, in this case, are also correct for in-plane
shear if \/G is replaced by (K + 1)/4G in the displacement formulae.
At high stress levels the small scale yielding solution is no longer
useful and recourse must be made to complete solutions. For the finite
crack of length la configuration of Fig. 1, subjected to a remotely applied
tensile stress (crj^ = ax , the resulting plastic zone size for discrete
planes of tensile yielding at the ends of the crack is [13-25]
Identical formulas apply for in-plane or anti-plane shearing, with

aja0 replaced by rjTQ . The complex stress potentials (Eq 5) are
For in-plane shear, a0 is replaced by —ir0 in the expression for <fr(^)

and the constant is replaced by +*Y00/2, and ^(f) = — 2<J>(f) — f<t»'(f)
+ /TM . The complex potential fi(f) for the anti-plane shear case is identi-
cal to the above expression for $(£"), but with <TO replaced by 2r0 and the
constant dropped. The crack opening displacement uy(Q, 0) = u0 is
FIG. 17 —Plastic zone size and crack opening displacement as a function of

applied stress; Discrete surfaces of tensile yielding at crack tips.
Again, the same expression applies for in-plane or anti-plane shear with
\/G replacing (K + 1)/4G in the latter case.
The plastic zone size and crack opening displacement given by the
above expressions are graphed as a function of applied stress in Fig. 17.
These appear in the dimensionless forms
both of which equal (7r2/8) (o-^/o-,,)2 according to the small scale yielding
solution of Eqs 3\a and d upon noting that KI = o-^Tra)112 in this
case. The solid lines of Fig. 17 are Eqs 34a and c above; the dashed
line is the result of the small scale yielding solution. As is similar to the
anti-plane shear case of Figs. 6 and 7, the surrounding elastic singularity
controls the plastic zone size up to about 30 to 40 per cent of the limit
load, and the crack opening displacement up to about 40 to 50 per cent
of the limit load.
The discrete surface of tensile yielding models for finite width planes
with single edge, double edge, or central cracks are solved approximately
by cutting appropriate segments from an infinite array of identical colin-
FIG. 18—The infinite colinear array of identical cracks provides an approxi-

mate solution of the discrete surfaces of tensile yielding models for the double
edge cracks, central crack, and single edge crack configurations.
ear cracks, as in Fig. 18. The plastic zone size for this approximation
is [27]
and the crack opening displacement is given by the integral
where a = IT (a + co)/(26) and / * = ( ! — aJo^Tf/1. Numerical results

have been given by Bilby and Swinden [30]. Noting that the stress
intensity factor is
for the infinite array, these results may be shown to reduce to the small
scale yielding solution at low stress levels. The plasticity models for the
single edge, double edge, and central crack configurations, as based on
the infinite array, are likely accurate to about the same order that the elas-
tic stress intensity factors are given by the infinite array expression (Eq
35c). From the discussion of Paris and Sih [4], the approximation would
be best for short central cracks and worse for single edge cracks due to
bending in the latter case. In fact, the limit load is even given incorrectly
in the single edge crack case as some compressive yielding must
occur to offset the bending induced by the lack of symmetry. Smith
[27] has discussed the discrete surfaces of yielding model for other
configurations involving more than one crack; Rice [7] has given general
methods through which plasticity solutions may be determined directly
FIG. 19—Plane stress plasticity in thin sheet may be modeled as discrete

surface of tensile yielding at crack tip: (a) two 45 deg slip bands ahead of crack and
(b) fully developed 45 deg shear Up crack with one slip band.
from known elastic solutions for single cracks in infinite bodies, and has
given the detailed solution for wedge forces on a crack surface.
Hahn and Rosenfield [29] have discussed the particular relevance of
the discrete surface of tensile yielding model for plane stress plasticity.
As in Fig. 190, plastic flow ahead of flat through-the-thickness cracks
in thin sheets tends to consist of two intersecting 45-deg shear bands.
The plasticity is then localized to a narrow region of height roughly
equal to the sheet thickness. Presumably for an inclined 45-deg crack
(Fig. 196) as may occur after a plane stress transition in fatigue, a single
narrow 45-deg shear band would appear ahead of the crack. The average
plastic extensional strain, e/, would equal 2ua(x, Q)/t in either case,
where / is the sheet thickness and uy(x, 0) the displacements calculated
from the discrete surface of tensile yielding model. A plane stress analysis
based on digital computer solutions of the governing elastic-plastic
equations has recently been presented by Swedlow, Williams, and
Yang [36], While their formulation was two-dimensional and thus
naturally does not reflect the inclined 45-deg shear band patterns ob-
served in steels [29,37], their techniques show the promise of highly
accurate numerical solutions to configurations such as cracked bodies
exhibiting steep stress and strain gradients. Details of the deformation
very near the crack tip remain obscure. Approximate methods of
determining the plane stress distributions near cracks, based on photo-
elastic analyses, have been proposed by Dixon and Strannigan [38].
Keer and Mura [28] treated the penny shaped crack of radius a in an
infinite solid subjected to the uniform remote tension <rx , employing the
model of a discrete annular surface of tensile yielding surrounding the
crack. In this case the plastic zone size is given by
and the crack opening displacement by
FIG. 20—Discrete in-plane slip lines near crack under tension; plane strain
mode of yielding.
It is of interest to note that for the penny shaped crack, the crack open-
ing displacement approaches a finite limit as the remotely applied stress
approaches the limit value, ^ = a0 . Since Kt = 2(r08(a/ir)1/2 for the
penny shaped crack [4], both co and u0 may being shown to take the form
of the small scale yielding results (Eqs 3 la and d) at low stress levels
(provided the plane strain value of K = 3 — 4v is chosen), even though
these were derived for the planar case.
A model for tensile deformation near a crack tip, by in-plane sliding
on two discrete slip surfaces inclined at angles ±0 with the crack line, is
pictured in Fig. 20. As has been seen in the last section, such plasticity
distributions constitute an exact solution for anti-plane shearing of ma-
terials with single crystalline yield surfaces. Whether analogous results
occur in the tensile case is currently unknown; however, such discrete
slip models may be useful in determining gross features of plane strain
yielding, just as the discrete surface of tensile yielding is useful for the
plane stress situations. An exact method of solution would be to deter-
mine the stresses due to the application of external loads and yield level
shear stresses r0 along the slip lines, imposing also the condition of zero
normal displacement discontinuity, and then to choose the slip line
length so that stresses are bounded at the outer tips of the slip lines.
The mathematical difficulties are considerable and therefore an approxi-
mate solution is given based on the in-plane shear results of Eqs 320
and b.
Employing the singular terms of the complex potentials (Eq lla) and
Eqs 5, the elastic singularity gives the shear stress
along the prospective slip line. Now Eq 32a gives the slip line length
required to relax a shear stress An(27r/')~1/2. It is plausible to assume that
the inclined slip line length in Fig. 20 required to relax the shear stress of
Eq 37a is given approximately by the same expression, but with Ktt re-
placed by Y^Ki sin 0 cos (0/2). Thus
is the length of a slip line inclined at angle 0. It is also plausible to assume

that each slip line produces a total slip displacement jump twice the
crack opening displacement given by Eq 32b (recall that for a slip line
ahead of the crack, the opening displacement is one half the total dis-
placement discontinuity), so that the slip opening displacement u0 pro-
duced by sliding along each of the two slip lines is given approximately
by
While these approximations may appear somewhat arbitrary, it is in-

teresting to note that should results for a slip line ahead of the crack be
similarly employed to approximate the inclined slip line cases in anti-
plane shear, the plastic zone size and opening displacement of Eqs
lie for the 90-deg slip lines (Fig. 9cf) would be given exactly, and for
the 45-deg slip lines (Fig. 9&) the zone size and opening displacement of
Eqs \la would be given to within the errors of about 4 and 40 per cent,
respectively.
If 6 is taken as 90 deg in the above equations, the yield stress in shear
set equal to one half the yield stress in tension and the plane strain value
of K chosen with a Poisson ratio of 0.3, one obtains for plane strain
conditions
Here E = 2(1 + v)G is Young's modulus. The values when 6 = 45 deg

are about 15 per cent lower. For comparison, using Eqs 3la and d
as descriptive of plane stress conditions,
These results suggest that in the small scale yielding range, the maximum
plastic zone dimension and crack opening displacement are roughly half
as large for plane strain conditions as compared to plane stress condi-
FIG. 21—Slip line field immersed in plane strain plastic zone surrounding
crack tip; constant stress regions joined by centered fans.
tions. Another approach to plane strain yielding is considered in the

next section. While one generally anticipates that plane strain plastic
yielding will produce a "spread out" plastic region, a discrete slip line
model may be particularly appropriate for unstable materials exhibiting
upper and lower yield points.
Plane Strain Slip Line Field

While models based on discrete surfaces of slip or tensile yielding give
at best gross features of tensile plasticity, some of the detailed features
of plane strain yielding can be studied through recourse to the slip line
theory [13,31]. F. A. McClintock has recently shown the author some
yet unpublished experimental results on yielding in doubly edge notched
steel bars. Long bars with side grooves were compressed and later sec-
tioned in the center where plane strain conditions should prevail.
Etching revealed that the plastic region completely surrounded the crack
tip, much as in Fig. 21. At what appears to be a transition to limit
conditions at high stress levels, a few discrete slip lines inclined at
roughly 45 deg to the crack line (as in Fig. 20 with 6 — 45 deg) shoot
out from the smooth plastic region of Fig. 21.
The plane strain plastic slip line theory is exact only when Poisson's
ratio is ^ or when plastic strains are much larger than elastic strains
[31]. Neither of these conditions are met exactly; nevertheless, we pro-
ceed to examine the stress and strain distribution near the crack tip on
the basis of this theory. The stress free boundary conditions on the
crack surface determine the entire stress fields in the largest isosceles
right triangles, labeled A in Fig. 21, which may be fit in the plastic zone.
The stresses are constant in regions A and
Now any slip line emanating in region A, and finding its way to the line
in front of the crack, must cross that line at 45 deg. Thus, the same
hydrostatic stress buildup [13] occurs on each slip line, so that stresses
are constant on the line ahead of the crack. Therefore, another constant
stress region C is determined ahead of the crack, and in this region
ira-o/2 being the hydrostatic stress buildup in excess of the tensile yield
strength. Centered fans, regions B of Fig. 21, must join such constant
state regions, and employing polar coordinates the stresses in the upper
fan are [73]
Making the assumption of incompressible elastic behavior (v = 3/2),

strain and displacement distribution in the slip line net may be worked
out in terms of displacements on the outer boundary of the three regions.
Once the normal displacements un are specified at each point of the outer
boundary, the interior strain and displacement fields are determined.
Alternately, such a slip line net permits the arbitrary specification of
only the normal displacements. Letting unc(s) be displacements in the
normal direction on the boundary of region C, with s measured as shown
in Fig. 21, employing symmetry restrictions and following the treat-
ment of Prager and Hodge [73], displacements in region C are
where w 0 is the radius of the centered fans. Differentiating to obtain

strains,
Setting y = 0, the strain ej/(x,0) along the line ahead of the crack is
The extensional strain at distance jc ahead of the crack tip is the deriva-
tive of normal displacement with respect to arc length on the outer
boundary evaluated at 5 = co0 — x/\/2. Note that no singularity occurs
as the crack tip is approached along the line ahead of the crack. Letting
UHB(B} be normal displacements along the outer boundary of the cen-
tered fans and imposing continuity conditions along the common
boundary of B and C, displacements in the fan region B are
The corresponding strain field in the centered fan is
Just as in the anti-plane shear case, the centered fan focuses a 1/r shear
strain singularity into the crack tip, but now the focusing is from above
and below rather than from in front of the crack tip.
While the general features of the plane strain yielding are qualitatively
clear, further work remains to be done before quantitative results are
developed. Studies are currently underway on the matching of a slip
line net with the surrounding elastic field so that the dimension o>0 and
normal displacements of the outer boundary may be estimated. Several
questions remain. The assumption of elastic incompressibility is un-
realistic, and since plastic strains are clearly not enormous in comparison
to elastic strains in regions A and C, some perturbation of the above

solution will occur due to plastic flow in the z direction. Further, it is
presently unclear as to whether the slip line net of Fig. 21 occupies a
major portion of the plastic region or a small fraction, and also as to
whether the discrete slip lines alluded to earlier constitute simply a
transition to large scale yielding or an intrinsic part of the yielding even
at low stress levels.
The large stresses generated in front of a crack under plane strain con-
ditions likely extend to high applied stress levels, the regime in which
yielding is controlled by the elastic singularity. This would be consistent
with the greater success of elastic fracture mechanics for plane strain
failures [14]. Judging from the limit analysis solutions summarized by
Drucker [32] and McClintock [33], at applied stress levels causing net
section yielding the stress and strain distributions near the crack tip are
strongly dependent on geometry; the double edge notch configuration
permits the persistence of hydrostatic stress buildup, whereas the single
edge notch and central configurations do not.
Part II Mechanics of Fatigue Crack Propagation
The elastic-plastic mechanics developed in Part I is applied to fatigue
crack propagation in this part of the paper. Solutions of the various
models for unloading and cyclic loadings are given, and parameters
which may rationally be expected important in fatigue crack propagation
are identified. Theories of crack growth, seeking to relate cyclic crack
tip plasticity to microstructural "damage" accumulation and material
separation, are critically surveyed in the light of experimental results
and general consistency with the elastic-plastic analyses.
Elastic-Plastic Response to Cyclic Loadings

All of the elastic-perfectly plastic models discussed in Part I have the
common feature of involving proportional plastic flow; components of
the plastic strain tensor (or displacement discontinuity components
where discrete surfaces of slip or yielding are involved) remain in con-
stant proportion to one another at each point of the plastic region. This
permits a general treatment of the response to unloading, reloading, and
cyclic loading through the plastic superposition method developed in
special cases by Hult and McClintock [75] and Rice [7]. Undoubtedly,
the persistence of proportional flow is much more a commentary on
our mathematical ingenuity than on the physical situation. Compressi-
bility effects in plane strain yielding as well as the transition from in-plane
deformation to inclined 45-deg shear bands in plane stress yielding con-
stitute deviations from proportional flow; these may, nevertheless, be
expected insignificant for large plane strains and well developed plane
stress yielding.
Suppose a cracked body is loaded by a system of stresses proportional

to some parameter L, and that the loading parameter is reduced by an
amount AL to a lower level L — AL. To the extent that crack tip rounding
by plastic deformation is neglected, the stress concentration factor is
effectively infinite, and reverse plastic flow commences with the first
increment of load reduction, creating a new plastic zone of reversed
deformation imbedded in the plastic zone accompanying the original
FIG. 22—Plastic superposition for unloading. Adding (b) for load —AL with
a doubled yield stress to (a) gives the solution (c) resulting after unloading from
L to L-AL. Reloading, L-AL to L, restores (a). .
loading (Fig. 22). When flow is proportional, the effect of unloading is

to reverse the direction of stresses in the reversed flow region, without
otherwise affecting their magnitude or distribution. The changes in
stresses, strains, and displacements due to load reduction are then given
by a solution identical to that for original monotonic loading, but with
the loading parameter replaced by the load reduction AL and the yield
strain and stress replaced by twice their values for original loading, so
that stresses have the correct magnitude and direction in the reversed
zone when the changes due to load reduction are subtracted from the
distributions corresponding to the original monotonic loading. The pro-
cedure is illustrated in Fig. 22. Neglecting the possibility of crack closure,
the plastic superposition is valid up to the point where the reversed

plastic zone is equal in size to the original plastic zone accompanying
monotonic loading, which corresponds to complete load reversal
(AL = 2L). Crack closure always intervenes before complete load
reversal, as discussed subsequently. For unloadings L to L — AL, re-
loadings L — AL to L, and subsequent load cycles which do not cause
crack closure, the reversed plastic zone size and cyclic variations in
stresses, strains, and displacements depend only on the load fluctuation
AL and are independent of the maximum load L.
As an example, for the finite crack in a large body configuration of
Fig. 1, stress, strain, and displacement results from the elastic-perfectly
plastic models considered in Part I may be generally represented in the
form
for monotonic loadings, where r, 6 are polar coordinates centered at the

crack tip, ax is a remotely applied stress (shear or tension), o0 and e0
are a representative yield stress and strain, and ^,-y, EH , and Ui are
dimensionless functions of their arguments, reversing sign with sign
reversals of <rm . The distance to to the elastic-plastic boundary along any
radial line at angle 9 may be represented as
Now if the applied stress is reduced in magnitude by A<rx , the above

formulas apply for the changes in stress, strain, and displacement, as
well as for the position of the reversed plastic region boundary, provided
ax is replaced by Avx and a0, e0 replaced by 1a0, 2e0. Thus the distance
co* to the reversed flow region boundary along any radial line at angle 6
is
The stress, strain, and displacement after unloading are a a — Ao-^ ,

tij — Aaj, U{ — Aui, where the variations due to load reduction from
*<» to a^ — Ao-^ are
Complete load removal (Ao-^, = O leaves residual stresses and dis-

placements; from Eqs 42 and 43, their values are those at full load ax
minus twice those at half load ajl. Reloading, o-M — Ao-^ to <rx , re-
stores stresses, strains, and displacements to values taken before un-
loading. Cycling over a stress range Ao-^ produces alterations A<r i ; ,
Ae t -y, and Aw; with each load rise and fall as given by Eqs 436, with
Eq 430 giving co*, the size of the zone of cyclic plasticity.
The small scale yielding range is doubled for cyclic loadings since
ojo0 is now replaced by Ao-M/2a-0. Thus if the elastic stress intensity
factor is found to control some aspect of the plasticity up to a certain
per cent of the limit load, the cyclic variation in elastic stress intensity
factor controls the corresponding cyclic aspect of the plasticity up to
load variations which are double that per cent of the limit load
The maximum plastic zone dimension and crack opening displacement
have been plotted in Figs. 6 and 7 for anti-plane shearing of a Tresca-
Mises material and in Fig. 17 for a discrete surface of tensile yielding
(or, with appropriate substitutions, slip) ahead of the crack. The same
figures apply for load cycling with co replaced by co*, u0 replaced by
Au0/2, and TX (or O replaced by Ar^/2 (or Ao-^/2). Taking from these
figures 40 per cent of the fully plastic load as the upper limit on the
small scale yielding range, the corresponding upper limit on the small
scale yielding range for load cycling extends to load variations which
are 80 per cent of the limit load. Thus, for example, in a tension-tension
cyclic loading with a maximum stress cr^ = 0.8 a0, the total plastic
flow is poorly described by the Irwin stress intensity factor, while the
embedded cyclic plastic flow is determined to within a small error by the
variation in the stress intensity factor.
Small scale yielding solutions for load cycling are obtained directly
from those for monotonic loading by replacing the stress intensity fac-
tor by its variation and doubling the yield stress and strain. Thus, for
example, in the anti-plane shear of a Tresca-Mises material one obtains
from the monotonic loading solutions of Eqs 12 and 13
for the cyclic plastic zone dimension, cyclic strain variation ahead of the
crack tip in the cyclic plastic zone, and cyclic variation in crack opening
displacement. Similarly, for a discrete surface of tensile yielding one
obtains from Eqs 31
In these two examples, as for all perfectly plastic small scale yielding
situations, a cyclic variation in the stress intensity factor from zero to
some maximum value results in a cyclic plastic zone of reversing defor-
mation one quarter the maximum plastic zone size, and a variation in
crack opening displacement over one half the total opening.
Since a residual displacement remains at the crack tip after load re-
moval, the crack surfaces remain propped apart by the plastic flow.
A rough estimate of the stress variation required to initiate contact of
the crack surfaces may be had by examining the anti-plane shear case.
Cracks do not open or close in this case, but rather slide. However, the
load variation at which the shear displacement jump returns to zero at
the crack midpoint of Fig. 1 should give a good estimate of the tensile
load variation required to initiate closure. Supposing plastic flow to
give the crack an effective length of 2l(>2a), from the elasticity solution
of Eq 7 displacements along the upper crack surface are
Choosing / to give the small scale yielding crack opening displacement

result of Eq 13, which is noted from Fig. 7 to be fairly accurate up to
about 60 to 70 per cent of the limit load,
At low stress levels this agrees with the interpretation of considering the
crack to be longer by half the plastic zone size. The change in displace-
ments of the crack surface due to a load reduction ATX is then
where
Setting Auz = uz at the crack midpoint x = — a, one finds that closure

initiates at a stress reversal satisfying
the formula being valid for rw less than 60 to 70 per cent of r0. Conse-
quently, closure initiates at Ar^/r^ = 1.015 for rjr0 = 0.2, at ATOO/TOO =
1.065 for rjr0 = 0.4, and at ATK/TX = 1.105 for TJr0 = 0.6. The
closure ratio must increase rapidly as the remote stress approaches the
limit load, as it is clear that its value would then approach two for the
crack in an infinite solid. Once closure of a tensile crack initiates, rela-
tively large compressive stresses would be expected necessary to cause
much further reverse flow; thus, the above numerical results set approxi-
mate limits on the amount of load variation which should be considered
responsible for cyclic plastic flow under tension-compression load cycles.
A more exact analysis, based, say, on the discrete surface of tensile
yielding model and following through the calculations after closure ini-
tiates at the crack midpoint, would be useful for the interpertation of
fully reversed loading fatigue results.
FIG. 23—Stable hysteresis loop.
The anti-plane shear work-hardening results for monotonic loading

may similarly be extended to cyclic loadings, when stress-strain relations
in the cyclic plasticity zone may be described as the traversal of stable
hysteresis loops (Fig. 23) for which the stress and strain variations fol-
lowing a load reversal are related by a law independent of the maximum
amplitudes of the loops. Morrow [34] has suggested such relations
between stress and strain variations as good approximations to experi-
mental results for cyclic plastic straining under zero mean load after a
small per cent of the low cycle fatigue life is passed. Whether similar
results hold in the presence of a mean stress is unclear. As in Fig. 23,
T0* and y0* are the yield stress and strain in a coordinate system AT,
Ay measured from a point of load reversal; in the plastic range, Ay > y0*,
AT = r*(A7). The solution for stress and strain variations is identical
in form to that for monotonic loading. Thus, employing a polar strain
coordinate system Ay, 0 analogous to Fig. We, the small scale reverse
yielding solution follows the form of Eqs 21; in the reverse plastic zone:
where
The geometric interpretation of Fig. 11 applies for cyclic flow also.

Lines of constant cyclic strain variation are circles of radius R*(Ay~)
centered a distance X*(Ay) ahead of the crack. The perfect plasticity
cyclic solution of Eqs 44 is recovered by setting r*(Ay] = TO* = 2r0.
As an example, for a power law relating stress and strain variations in
the hardening range,
the cyclic strain variation in the reverse flow plastic region ahead of the
crack caused by a load fluctuation AKni is
The stress concentration of a crack has been assumed to remain infinite

in discussing unloading and cyclic loading solutions. Actually the large
strains generated result, as has been seen, in finite crack opening dis-
placements so that a small range of elastic unloading may be expected.
A rough estimate of this range may be had by assuming the crack to
open into a narrow elliptical shape with a root radius of curvature given
by the crack opening displacement corresponding to the maximum
applied stress <rM . Then employing the stress concentration factor of
Eq 8a for the ellipse of length la in a large body, the elastic unloading
range Ao-^ prior to reverse flow at the crack tip is given by
Estimating the opening displacement from the discrete surface of tensile

yielding model, Eq 34c, the ratio of the elastic unloading range to the
maximum applied stress is
At low stress levels this becomes independent of aja0, and for plane
stress conditions the elastic range is given by
where c0 = a0/E is the yield strain in tension. The result is not highly
dependent on the maximum stress over a substantial range; at vja0 = 0.8
the numerical factor is 1.49 and at <jjv0 = 0.95 it is 1.89. There is a
rather fast transition as fully developed plasticity is attained, for the
numerical factor approaches infinity at the limit load for the perfectly
plastic model. This would suggest that overall plastic straining is re-
quired to cause sufficient crack blunting for purely elastic response to
all but the small load variations. Taking e0 = 0.002, as appropriate for a
yield stress of 60 ksi in steel, below limit conditions the elastic range is
roughly 5 per cent of the maximum applied stress. An approximately
equal figure results for the elastic range following the application and
complete removal of a maximum stress 0-^ .
A further approximate analysis of crack blunting may be based on the
anti-plane shear perfect plasticity solution for a rounded end notch of
root radius p. Recalling that stress free boundaries are now semicircular
and taking x = 0 at the rounded notch tip, in the plastic re-
gion 0 < x < co — p,
where u0 is the crack opening displacement for monotonic loading. Here

p is considered of negligible size compared to the plastic zone size. As-
suming for the moment that the radius of curvature is fixed, for a mono-
tonic load increment resulting in du0,
It turns out that this expression is also correct for unloading and subse-
quent reloading if variations in u0 are computed according to Eq 44.
Now if one assumes a similar relation to hold for tensile cracks, the root
radius of curvature would appropriately be chosen as the current crack
opening displacement resulting from the prior deformation history and
computed according to equations based on the neglect of crack tip blunt-
ing, as displacements are an integrated effect. Thus,
At the notch tip this results in dty = dp/p, which would appear to be of
approximately the correct form. Thus integrating under the assumption
that p = u0 = 0 prior to any load application (the only other plausible
assumption is an atomic spacing; any further crack blunting must be
assumed the result of plastic deformation although many authors prefer
ascribing a fixed nonatomic radius to cracks), after any loading se-
quence which results in a final opening displacement u0
Since u0 is on the order of a yield strain times a plastic zone dimension,

the logarithm behaves as u0/x at distances much greater than the yield
strain times plastic zone size, in agreement with the shear result (Eq
510). For a load variation which changes the opening displacement from
MO min to MO m ax, the corresponding strain variation is from Eq 52a
Now, regardless of the model employed, the crack opening displacement

for small scale yielding takes the form
for monotonic loading, where a is a numerical constant. Suppose that a

peak stress intensity factor Kx max is applied and removed, and then fur-
ther cyclic loads are applied over a range from zero to AKt ; then
At the notch tip x = 0,
Setting Aey = 2c0 and taking the stress intensity factor proportional to a
remotely applied stress, this equation closely approximates the estimate
of the elastic unloading range given previously. When KI max = AKT (no
peak load), the strain variation at the notch tip is Aey tt 0.69. While this
result is independent of the load level, the size of the region ahead of the
notch affected by high strains is not. This falls off to Ae^ tt 0.22 when
the peak load is twice the subsequent cyclic amplitude, and Aev » 0.04
when a peak load of five times the subsequent amplitude is applied.
Smaller values develop if the peak is out of the small scale yielding
range. While one hesitates to attempt quantitive comparisons due both
to the highly approximate nature of the calculations and the lack of
data on the strain amplitude necessary to continuously propagate a crack,
this marked reduction in local cyclic strain by a peak loading does
appear consistent with the experimental results of Donaldson and
Anderson [35], who found that high peak loads could effectively stop
crack propagation for a large number of subsequent load cycles.
General Features of Fatigue Crack Growth

The growth of fatigue cracks prior to catastrophic fracture is con-
veniently divided into processes of initiation as microcracks and propaga-
tion as macrocracks. The latter propagation stage is of primary interest
here. Thompson and Wadsworth [39], Avery and Backofen [11], and
Grosskreutz [12] have surveyed work on micromechanisms relating to
crack initiation. Commonly, initiation occurs at a free surface. The cyclic
component of tensile stress at a surface point sets up alternating shear
stresses, maximum on 45-deg planes with the tensile direction. Con-
veniently oriented slip systems respond by the formation of slip bands
resulting in a roughened surface topology in the form of intrusions and
extrusions. These apparently act as stress concentrators so that cracking
occurs along the slip bands. Interior initiation sites at grain boundaries
may occur in the presence of overall cyclic plastic deformation. While
no clear demarcation point exists, cracks initiated at 45 deg with the
tensile direction tend to propagate as macrocracks on planes perpen-
dicular to the tensile direction. Factors such as corrosive environments
and fabrication imperfections in practical structures tend to limit the
importance of initiation processes; cracks may frequently be present
from the start [35].
Striation formation [12,41] is observed on portions of fatigue crack
surfaces perpendicular to the tensile direction, both in metals and
polymers [42}. Estimates of striation spacings agree well with macro-
scopic growth per cycle measurements [41], so that fatigue crack growth
would appear continuous at least in regions of striation formation. This
need not always be the case; Frost [40] reports continuous fatigue inter-
mingled with occasional spurts of short fracture advances in some
cases. Laird and Smith [43] have demonstrated a mechanism of stria-
tion formation through crack tip deformation under large cyclic stresses;
the observed deformation pattern is much like what might be expected
from the slip line field of Fig. 21. A fatigue crack mode transition from a
90 to 45-deg orientation of the crack surface with respect to the tensile
direction is observed in sheet materials at sufficiently high load levels
[40,44], similar to the plane stress fracture mode transition [45] under
monotonic loadings. Hertzberg [41] found no striations on 45-deg por-
tions of the fatigue crack surfaces; ductile "dimples" appear instead,
elongated in a direction perpendicular to the direction of crack advance.
A survey of work on the relation of continuum considerations to
models of microstructural "damage" accumulation and material sepa-
ration is given in the next section. For the present we examine what gen-
eral conclusions on fatigue crack growth follow the elastic-plastic analy-
ses discussed earlier, without specific reference to microstructural
considerations and
Copyright by fine details
ASTM of cyclic
Int'l (all rights stressMon
reserved); andDec
strain distributions.
7 14:40:45 While
EST 2015
conclusions are more limited in detail, the groundwork is firmer. The

varying results of Figs. 8 and 9 raise valid cause for doubt that continuum
plasticity solutions can confidently be extrapolated to the fine scale of the
microstructure. For example, a cyclic stress variation of one fifth the
yield stress sets up a cyclic plasticity zone of roughly one hundredth of
the crack depth in size; this figure is in the size range of one or a few
grains for cracks less than an inch in depth.
One of the central results emerging from the analyses of all the elastic-
plastic models is that, for small scale yielding, the plastic deformation is
FIG. 24—The correlation of crack growth rate in terms of stress intensity

factor variation for 7075 T-6 aluminum (maximum load equal to load range) (Ref
47).
entirely determined by the history of variation in the Irwin stress inten-

sity factor. Thus two different cracked bodies with identical material
properties (and sheet thickness where three dimensional considerations
govern, as in plane stress yielding) will exhibit identical fatigue crack
extensions if each is subjected to the same small scale yielding range time
history of variation in elastic stress intensity factor. Paris, Gomez, and
Anderson [46] first proposed a correlation of this type; a series of sub-
sequent studies by Paris [47], Paris and Erdogan [6], Donaldson and
Anderson [35], Schijve and Jacobs [48], and Schijve et al [49], employing
both their own data and that of several other investigators, have amply
documented the validity of this approach. McEvily and Illg [50] de-
veloped an equivalent approach based on the variation of the elastically
predicted maximum concentrated
Downloaded/printed by stress, viewing a crack as a notch with
a fixed root radius sufficiently small so that the direct relation with stress
intensity factors noted in Eq 8c applies. (Incidentally, most of the litera-
ture on fatigue crack growth employs a definition of the stress intensity
factor which differs by a factor of ir1/2 from that given here.) A cyclic
loading may be characterized by the maximum load value, Z,max , and the
FIG. 25—Crack growth rates in terms of stress intensity factor variation for
several materials (Refs 6 and 47).
load range, AL. When both the maximum plastic zone size and cyclic
plastic zone size are in the small scale yielding range, the loadings and
planar geometry of the cracked body are sensed at the crack tip only
through the maximum, AT max , and range, A.K, in the stress intensity
factor. The maximum value may be represented in the dimensionless
form
the latter following since stress intensity factors, coming from linear
boundary value problems, are directly proportional to applied loads
times functions of the current geometry. Thus, where da/dn is the exten-
sion of a crack tip per cycle of loading,
for the comparisons of crack growth rates when all other material char-
acteristics are held constant. One of the most striking verifications of a
result of this form was provided by Paris [47] on the basis of data on
7075T-6 aluminum with X = 1 . Figure 24 (Fig. 1 of Ref 47) shows crack
growth rates as a function of stress intensity factor range for two cen-
tral crack configurations, one [55] loaded by wedge forces F per unit
sheet thickness and the other [50] by a tensile stress a. Without correc-
tions for finite width (included along lines of the infinite array analysis
of Fig. 18 in plotting the data), the stress intensity factors are
AK increases with crack length for a constant tensile stress variation and
decreases with crack length for a constant wedge force variation. While
the loadings are about as different as could be imagined, both cases fall
on essentially the same curve.
Figures 25a, b, and c, from Refs 6 and 47, demonstrate a similar suc-
cess of the stress intensity factor variation in unifying crack growth rate
data for several materials. Figure 25a represents the results of five inde-
pendent investigations on 2024T-3 aluminum. Figure 25b replots in a
similar fashion that of Fig. 24 on 7075T-6 aluminum, with the results of
another investigation added. The two aluminum alloys plots are repre-
sented by straight lines in Fig. 25c, which shows additional data on mag-
nesium, titanium, molybdenum, and steel. The scatter is largely due to a
failure to distinguish, in these plots, effects due to maximum loads greater
than the load variation range, differences in test frequency, differences in
sheet thicknesses, and mode transition from 90 to 45 deg with the tensile
axis. The plots being log-log, straight lines are indicative of power law
relations. All of the straight lines drawn have a slope of four so that, as
noted by Paris and Erdogan [6], the broad trend of the data is repre-
sented by a crack growth rate proportional to the fourth power of the
stress intensity factor variation. At the same time, it is clear from the
figures that a law of this type may provide a poor approximation over
limited ranges of data which may be of interest in particular applica-
tions.
All perfect plasticity models discussed earlier predict a dependence of
the cyclic plastic flow only on the variation in stress intensity factor; the
same Downloaded/printed
conclusion followsby for work-hardening behavior when stable
hysteresis loops of the type described in connection with Fig. 23 are

formed. Thus, it is not surprising that the maximum load achieved is a
minor variable, as compared to the load range, in determining fatigue
crack growth rates. An obvious exception is when the maximum is less
than the range so that crack closure may intervene, as in the fully re-
versed loading tests by Illg and McEvily [57]. The small range of com-
pressive loading prior to crack closure predicted by Eq 47 would suggest
that for reversed loadings in and somewhat above the small scale yielding
range, the load variation should be counted simply as the tensile part of
the cycle; the interpretation appears justified as crack growth rates are
only slightly faster for fully reversed loadings [57]. Aside from the possi-
bility of crack closure, the effect of different maximum loads with the
same load variation is to increase, with increasing maximum levels, the
average strain distribution about which the cycling takes place. While
perfect plasticity predicts a zero mean stress in the cyclic plastic zone,
the effect of hardening for a hysteresis loop shift in the direction of the
monotonic loading stress, as in Fig. 23, would be to place a biasing aver-
age stress distribution, more intense with increasing maximum level in
the cyclic plastic zone.
Judging from the data of Schijve et al [49] on 2024T-3 under various
ratios X of maximum load to load range, at a given range of variation of
the stress intensity factor the crack growth rate is tripled by doubling X
from 2 to 4. For comparison, a doubling of the stress intensity factor
range of variation increased the crack growth rate by a factor between 10
and 15. Results at a fixed value of X and sheet thickness, as by Schijve
et al [49], greatly reduce the scatter in plots such as Fig. 25a. Paris [52]
has tabulated values of the coefficient in a fourth power law description
of crack growth rates for several materials and conditions. He finds an
increase in the ratio of maximum to range from 1 to 4.5 increases growth
rates, at the same stress intensity factor range, by a factor of roughly 10
in both 2024T-3 and 7076T-5. An increment by a factor of 4.5 in the
stress intensity factor range increases growth rates by a factor of ap-
proximately 200 to 400. Frequency effects are slight in aluminum alloys,
as might be expected from their relatively rate independent stress-strain
behavior. An increase from 20 cpm to 20 cps roughly doubles [52] growth
rates in 2024T-3.
Liu [44] has proposed that the fatigue crack mode transition from a 90
to 45-deg orientation with the tensile direction in sheet materials may be
explained as a transition from plane strain (Figs. 20 and 21) to plane
stress (Fig. 19) plasticity conditions, an extension of Irwin's [45] approach
to fracture mode transition. Plane stress conditions may be expected to
initiate when the plastic region becomes sufficiently large so that 45-deg
shearing is kinematically possible. Equivalently, one then expects mode
transition at a constant ratio of plastic zone dimension to sheet thick-
ness, t, as found by Liu [44] and Hertzberg [41] in cases where ^Tmax =
AK, and the plastic zone dimension was estimated as proportional to the
square of the ratio of stress intensity factor to yield stress. Results re-
ported by Hertzberg would lead to a maximum plastic zone dimension
equal to the sheet thickness and a reversed zone size of about one quarter
the sheet thickness at transition, whether computed according to the
discrete slip line model of Eq 3 80 (with the plastic zone size now counted
as twice the slip line length) or the surface of tensile yielding model of
Eq 310. Broek and Schijve [55] found that a similar idea, for various
cases of a maximum load greater than the load range, did not accurately
predict the mode transition; the ratio of both maximum and cyclic zone
sizes to thickness decreased with increasing thickness. They also report a
FIG. 26—Delay affects in 7075 T-6 aluminum crack propagation due to re-
duction of stress range from <ri to a% (data from Ref 54).
systematic increase in growth rates with increasing sheet thicknesses prior

to transition, although the effect becomes obscured after development of
the inclined 45-deg fatigue crack surface.
The success in correlating crack growth behavior with stress intensity
factor variation, as predicted by the mechanics analyses, may also be
expected in situations of similar variations in stress intensity factors due
to complicated deterministic or random loadings. Loadings with varying
ranges introduce, however, an additional complication in that a sudden
shift from a large to small range of load variation tends to stop crack
extension from a few hundred to a few million load cycles [35,54]. Hard-
rath [54] has reported the results of tests on 7075T-6 aluminum in which
a stress cycling range from zero to o\ is suddenly followed by a cycling
range from zero to <r 2 . As expected, for cr2 > a\ the crack growth rate
immediately adjusts to that corresponding to the latter stress range;
delays result for 0-2 < a\. Hardrath's suggestion that the presence of
residual compressive
Copyright stresses
by ASTM nearrights
Int'l (all the crack tip Mon
reserved); is responsible for delays
Dec 7 14:40:45 EST
2015
appears to lack support from the plasticity analyses. For all the perfectly
plastic models and for the stable hysteresis loops work-hardening model,
the zone of cyclic plastic deformation as well as cyclic strain variations
within this zone are unaffected by the prior load level, so that the residual
stress argument would at most lead to an effect comparable to the slight
consequences of changing the mean load level. A more plausible explana-
tion would be based on the blunting of the crack tip by large deformations.
When both en and o-2 are in the small scale yielding range, the analysis in
connection with Eq 51>d suggests that the strain variations at the blunted
crack tip depend only on the ratios of the stress intensity factors K\ and
Kz or equivalently in this case on the ratio 0-1/0-2, and not on the current
crack length and separate ratios of 01 and o-2 to the yield stress. Supposing
the delay number of cycles to depend only on the reduced strain variation
at the crack tip, a unique relation between delay time and stress ratio
0-1/^2 would be expected from the argument based on crack blunting.
This conclusion appears roughly verified by the replot of Hardrath's
data in Fig. 26. The largest prestress, a\, is 50,000 psi which is about 70
per cent of the yield stress so that Figs. 7 and 17 appear to justify the use
of small scale yielding results as the calculation is based on crack opening
displacements. Perhaps the systematically shorter delays at the same load
ratio for the highest prestress is due to a higher average strain, about
which variations take place, at the blunted crack tip.
For random loadings in the small scale yielding range, crack exten-
sions may be expected identical in separate configurations if each crack
tip experiences statistically identical variations in stress intensity factor.
Paris [47] has displayed data verifying such a postulate for narrow band-
width random loadings of geometrically similar power spectra, but differ-
ent root mean square (rms) stress levels varying from 0.03 to 0.05 of the
yield stress. The scatter is slightly larger than for cyclic loadings, possibly
due to the enhanced likelihood of severe tip blunting at the higher rms
levels. Although no computations of the statistical distribution of plastic
strain variations have appeared (these could easily be provided for the
various perfectly plastic models through a relatively simple computer
program coupled with digitally generated random loadings), it appears
reasonable on the basis of unloading and cyclic loading solutions to
assume that the plastic deformations are controlled primarily by the dis-
tribution of rises and falls [55,56], A.K, in the stress intensity factor.
Rice et al [47,56] and Smith [57] have shown that crack growth rates may
be correlated in terms of averages in stress intensity factor rise and fall
heights, AK, for both narrow bandwidth random loadings and for doubly
peaked (two dominant frequencies) wide bandwidth loadings. When
correlated in this way all the random load crack growth rate data ap-
pear to fall on roughly the same curve, in spite of the great differences in
waveform appearance by
Downloaded/printed for the different random loadings. But neither
random load crack growth rates, nor their general trend of variation
with averages of AAT, agree with cyclic loadings crack growth rates cor-
related in the same way. This suggests that it is not so much the wave-
form appearance, but rather the very introduction of randomness in
load sequencing, which creates differences in comparison to the cyclic
loading case. Hardrath [54] and Hardrath and Naumann [58] have indi-
cated the differences in fatigue life resulting from similar load amplitude
distributions programmed in different nonrandom sequences.
Erdogan and Roberts [59] reported a study of crack growth in thin
plates subjected to fully reversed bending loads, and obtained a similar
correlation of propagation rates in terms of the variation of a stress
intensity factor defined through elastic solutions for plate bending.
Their comparison with the tensile case suggested that for a given re-
motely applied stress range on the surfaces of a plate under bending, the
propagation rate is approximately that which would result from one half
that stress range applied in direct tension to a geometrically similar
cracked plate. For identical stress ranges, propagation rates under direct
tension are about 10 to 16 times faster than those for bending. Erdogan
and Roberts also gave an approximate analysis of plastic yielding, based
on an idea similar to the discrete surface of tensile yielding model, which
suggested that the cyclic plastic zone for a given surface stress range in
bending is equal in size to the zone resulting from one half that stress
range applied in direct tension.
No single parameter plays the role of the elastic stress intensity factor,
in determining the crack tip plasticity, when yielding is on a large scale
compared to planar geometric dimensions. Thus crack growth rates may
be predicted or estimated from data collected in the small scale yielding
range only in conjunction with a reliable theory of crack propagation.
Some progress may still, however, be anticipated from continuum con-
siderations alone. The small scale yielding range for load cycling is double
that for monotonic load application. Thus to the extent that the maxi-
mum applied load is a secondary variable in comparison with load range,
situations may be expected in which the stress intensity factor variation
correlates growth rates even though the total plastic region is outside the
small scale yielding range. Further, although adequate experimental
data to check such hypotheses are not available, one might expect that
in the large scale yielding range crack growth rates may be correlated by
such parameters as the maximum and cyclic plastic zone sizes or crack
opening displacements, both of which are equivalent to the maximum
stress intensity factor and its variation at low load levels. Some orderings
of deviations from small scale yielding range growth results are also pre-
dictable. From Figs. 6 and 7, along with their interpretations for cyclic
loadings, the small scale yielding solution always underestimates
actual
7 14:40:45 EST 2015
resultsDownloaded/printed
for remotely applied
by stress loadings, the underestimate being
more severe with the smaller crack length to width ratios. Thus growth
rates from small scale yielding tests conducted with long cracks and low
net section stresses tend to underestimate rates at the same stress in-
tensity factor when large scale yielding occurs, and for configurations
with the same (high) net section stress and stress intensity factor, cracks
grow faster for the smaller crack length to width ratios. A large scale
yielding solution for wedge loadings as in Fig. 24 may be obtained
through known general methods [7] for the discrete surface of tensile
yielding model. It turns out in this case that the small scale yielding solu-
tion overestimates actual results, so that with wedge force loading growth
rates are slower, with large scale yielding, than those occurring in a small
scale yielding range test at the same stress intensity factor. Large scale
yielding analyses are not available for situations involving net section
yielding. This is an inherent difficulty with perfect plasticity models as
unrestricted flow ensues; for work hardening an apparently tractable
formulation [9] of the net section yielding problem is available in the
anti-plane shear case, although mathematical difficulties have to date
limited solutions to the case discussed in Part I. Consequently crack
propagation under repeated overall plastic straining, as in low-cycle
fatigue, currently is beyond the reach of reliable analytical treatments.
Theories of Fatigue Crack Growth

Obvious importance attaches to the prediction of crack growth rates
in terms of continuum and microstructural variables for arbitrary load
sequences. Only the studies of Head [60,61], McClintock [62], Rice [7,63],
and Weertman [64] have attempted the direct use of elastic-plastic
analyses, and none of these incorporated the possibility of crack blunting
and delays which are known to occur with block loadings or peak over-
loads and probably form an integral part of fatigue crack growth under
more general deterministic or random loading sequences. Thus, the work
surveyed here is limited to nonvariable amplitude cyclic loadings and
should be viewed as at most a start toward the problem of relating con-
tinuum analyses to material separation by fatigue.
Dimensional considerations were introduced by Frost and Dugdale
[65] and Liu [66,67]. The relevant variables for a crack of length a in a
large body are (assuming the maximum stress to equal the stress range)
where the elastic modulus is included implicitly through the yield strain
e0, N is the hardening exponent or some other dimensionless variable
(or set of variables) characterizing the hardening behavior, and e/ is some
characteristic
Copyright fracture strain.
by ASTM Int'l Attention
(all rights is limited
reserved); Mon Dec to plane strain
7 14:40:45 situa-
EST 2015
tions Downloaded/printed
so that sheet thickness
by does not enter. With the assumption that
continuum variables alone control the extension per cycle at a crack tip,,
there results [65-67]
where f\ is a dimensionless function of its arguments. Liu [67] recog-

nized the requirement of applied stress and current crack length entering
only in the form of the stress intensity factor variation, AK = Acr(7ro)1/2,
and therefore proposed
On the basis of a somewhat arbitrary modification of a result of Head dis-

cussed subsequently and the requirement of fitting samples of data with
a growth rate proportional to crack length, Frost and Dugdale proposed
a result contrary to predictions of continuum analyses in the small scale

yielding range typical of high-cycle fatigue, but possibly indicative of the
direction of the large yielding modifications necessary for stress ranges
above about 80 per cent of the net yield stress. Paris and Erdogan [6] and
Schijve and Jacobs [48] have amply documented the failure of these laws
to fit the broad trend of crack growth rate data. For example, in Fig. 24
an increase in AK by a factor of 10, from 4400 to 44,000 lb-in.~3/2, in-
creases the growth per cycle by a factor very nearly equal to 104, from
3 X 10~7 to 2.5 X 10~3 in. Liu's result would predict an increase in
growth per cycle by a factor of only 102, and the Frost and Dugdale result
would increase growth by a factor between 102 and 103. Both fail by one
or two orders of magnitude, although both can admirably fit limited
portions of the available data [6]. Liu [44,68] has suggested that part of
the difficulty may lie in the failure to distinguish between the plane stress
and plane strain modes of crack growth in the general trend of the data,
thus introducing sheet thickness as another length. The maximum stress
intensity factor value cited above leads to a maximum plastic zone size of
cu « 0.16 in. in 7075T-6 aluminum, according to the discrete surface of
tensile yielding model, Eq 310. As noted earlier, this would correspond
to mode transition in all sheets thinner than about 0.16 in.; however,
the general fourth power trend of the data extends to lower load levels
and at half the maximum stress intensity factor value the transition
thickness would already be reduced to about 0.04 in. While Liu does give
evidence [44] of different growth rates before and after transition, he
does not verify (Eq 56&).
Equation 566 is the by
Downloaded/printed logical consequence, for the usual small scale yield-
ing situations, of an assumption that only continuum variables govern

fatigue crack propagation. Its failure may be taken as a definite indica-
tion that another characteristic length, related to material separation,
must enter a theory of crack growth. Calling / this characteristic length,
a dimensional analysis incorporating the small scale yielding result that
loads and geometry enter only through the stress intensity factor leads
to
/2 being dimensionless. If several characteristic lengths enter, their ratios

are understood included. When sheet thickness, t, enters as in the plane
stress mode, an additional dimensionless parameter (A.K}2/(tcr02) is in-
cluded. For maximum loads different from the load range, the dimen-
sionless ratio X = KmSkSL/AK enters. As growth rates generally vary faster
than (AJ£)2, any increase in / while holding other parameters constant
serves to lower crack growth rates. The choice of / is related to the choice
of a separation mechanism. McClintock [62] (see further), for example,
takes / as the characteristic cell size formed by cyclic straining [12] and
achieves a remarkable unity in predicting growth rates by taking this as
the smallest size at which the concept of a homogeneous fracture strain
retains meaning. Other choices lead to physically reasonable results. As
material separation involves bond breakage and, in fatigue, possibly
partial rehealing, / may be related to the surface energy or work per unit
area in separating bonds (surface energy divided by a yield stress or
elastic modulus has length dimensions). Thus picturing environmental
effects as reducing surface energy through the formation of weak easily
broken bonds with the environmental agent, similar to proposals by
Westwood [69], known influences [70,71] of environments in reducing
fatigue life would be expected. Hertzberg [41] reports no distinct surface
markings indicative of a characteristic length in regions of striation
formation, so that surface energy may indeed provide the appropriate
length. Ductile dimples do occur, however, on fully developed plane
stress fatigue surfaces and these might be the result of progressive void
growth and coalescence under repetitive straining, similar to an observed
mechanism of ductile rupture [72], In this case the mean void nucleation
site spacing would provide the characteristic length.
Head's [60,67] work on fatigue was done before treatments of plasticity
near cracks were available and is based on drastic simplifying assump-
tions. A cracked body is viewed as a composite array of three types of
continua. Taking infinitesimal elements of each of these types, material
directly ahead of the crack is viewed as an array of independent rigid-
plasticCopyright
tensile bars, each hardening linearly from a yield stress a to an
by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST0 2015
ultimate stress oy at
Downloaded/printed a by
modulus Ew . The material above and below the
crack and rigid-plastic bar array is viewed as an array of independent

elastic tensile bars each carrying the remotely applied stress a. The re-
motely applied stress is transmitted to the rigid-plastic bars both directly
and through an array of elements transmitting load by shear. Head chose
properties of the shear elements and lengths of the elastic tensile bars so
that an elastic solution of the model properly gave the opening displace-
ment at the center of the crack. The only shear elements activated are
those above and below the crack and plastic zone, so that the stress on
the line ahead of the crack turns out to equal the remotely applied stress
outside the plastic region.
An approximate solution was then given for the crack growth rate
due to a stress range — <r to +0-, under the assumption that separation
occurred when the rigid plastic elements cyclically hardened in the ab-
sence of a Bauschinger effect up to the ultimate stress ay . At low stress
levels, a- <<C <r0,
where 21 is the height of the rigid plastic elements in front of the crack.
Head does not suggest how the constant / should be chosen. Contrary to
the impression created by several summaries [40,6] of Head's work, the
plastic zone size is not constant but rather given by
As the computation is for low stress levels it is appropriate to replace

(T0 — (r by er0 so that Eq 58a gives a growth rate varying as the third power
of the stress intensity factor and Eq 58£ gives a plastic zone size propor-
tional to the first power. The deviation in the latter result from depend-
ence on the square of the stress intensity factor is clearly due to the arti-
ficial introduction of the length /. Frost and Dugdale obtained Eq 56c
by assuming / proportional to crack length, but as Paris and Erdogan
have noted [6] I must be chosen proportional to stress squared times
crack length if it is to be interpreted as a plastic zone dimension. Head
also provided a solution for the remotely applied stress in excess of the
yield stress, a- ^ a0 :
Aside from the drastic idealization of a continuum and the unidentified

length, Head's proposed mechanism of separation by continuous work
hardening up to a fracture stress may be questioned, as materials do not
hardenCopyright
indefinitely
by ASTM under cycling
Int'l (all straining
rights reserved); Dec 7 but
Mon[12,34] rather
14:40:45 EST tend
2015 to
stabilize and in somebycases to cyclically soften.
Downloaded/printed
Rice [7,63] and Weertman [64] considered essentially identical models

for fatigue crack growth, both based on the plasticity model of a dis-
crete surface of tensile yielding or slip ahead of the crack. Tracing the
deformation history of a particular point from when that point is first
encompassed by the plastic zone to when the crack tip advances to that
point, separation is assumed to occur when the total absorbed hysteresis
energy equals a postulated critical value U* per newly created surface
area. Letting Auy(x,Q) be the plastic displacements of the discrete surface
ot tensile yielding per load reversal when the crack tip is at x = 0 and
assuming the growth rate sensibly constant in a traversal of a zone w* of
reversed deformation, the growth rate is given by
Restricting attention to the small scale yielding range and obtaining

AM!/(JC,O) by doubling the yield stress and strain in the monotonic loading
solution, there results
the latter forms using the plane strain value of K = 3 — 4v. While the
fourth power dependence resulting is in accord with the general trend of
experimental results, the model does not provide a direct interpretation
of the hysteresis energy U* required per unit of newly created fatigue
crack surface area. Choosing a best fit fourth power law to the data of
Fig. 25a, U* « 6.3 X 104 lb-in./in.2 for 2024T-3 aluminum. For com-
parison, the energy absorbed in fracture of this material under monotonic
loading is typically over two orders of magnitude smaller at about 3 X
102 lb-in./in.2 It is of interest to note that a difference in total ductility
of the same order of magnitude (typically 100 to 1000 times) occurs in
fully plastic push pull fatigue tests as opposed to monotonic fracture
tests [73,62]. Perhaps the most serious objection here is the simplicity of
the discrete surface of tensile yielding model; while the model no doubt
gives accurate estimates of gross features of the plastic deformation such
as zone size, crack opening displacement, and total hysteresis energy
losses, all fine details of the plastic strain distribution are lost. Also, at
high stress levels plastic regions tend to change shape so that the per cent
of the hysteresis energy absorbed in regions very near the fracture surface
will not be constant. Nevertheless, Wells [74] does find that a similar
failure criterion based on energy absorbed at the crack tip (that is, yield
stress Downloaded/printed
times crack opening
by
displacement) tends to correctly extend the
small scale yielding monotonic fracture criterion into the large scale
yielding range.
McClintock [62,14] has developed a mechanics of crack extension by
fatigue, stable slow growth, and catastrophic propagation, employing
the anti-plane shear perfect plasticity solution and a failure criterion
based on plastic strain accumulation over a "structural size" of the ma-
terial. While valid objections might be raised to the relevance of these
results for tensile loadings, we have seen that gross features of the plastic-
ity are relatively insensitive to the deformation mode, and, indeed, his
studies have provided the most satisfactory conceptual and quantitative
basis for the entire range of observed tensile crack extension behavior in
ductile materials. One of the less obvious results arising is the differentia-
tion between strain increments due to crack extension under fixed loads
as opposed to load variations at a fixed crack length. For example, in
the case of an edge crack of length a in a large body remotely anti-plane
sheared, the strain variation per unit crack length increment at fixed
loads in the plastic region ahead of the crack tip is [62] (approximately)
where the crack tip is at x = 0 and o> is the plastic zone size accompany-
ing monotonic load application. While the shear case probably tends to
somewhat overestimate growth effects in tension, their inclusion results
in a prediction of observed stable crack extensions prior to catastrophic
fracture, more pronounced for increasing fracture ductility. Even at
relatively low stress levels the stress intensity factor fails to uniquely
characterize the plasticity when significant growth under fixed loads
occurs; as a consequence the small scale yielding range for fractures con-
trolled by the stress intensity factor is greatly reduced in very ductile
materials under plane stress conditions where slow growth is most pro-
nounced. Similar eifects do not occur in fatigue crack propagation as
strain variations due to load cycling at sensibly fixed crack lengths over-
whelm the then negligible contributions due to growth. Growth effects
would, however, be important in the few load cycles prior to catastrophic
propagation in which the extension per cycle is comparable to the plastic
zone size, but these compi ise a negligible portion of the fatigue life.
The strain singularity at the crack tip forces one to work with a spe-
cially defined average strain over a finite region or to limit the region over
which strain variations are considered consequential in determining sepa-
ration. McClintock has chosen an average strain defined over a narrow
wedge shaped region of angle 50 and length equal to the structural size, /:
Dropping the negligible crack growth contribution and assuming the

growth rate sensibly constant in traversing the reverse yielding plastic
zone, an adaption of the Coffin [75] criterion to varying amplitude plastic
straining leads to the growth rate
where A7*z(jc,0) is the variation in the plastic portion of the average

plastic strain caused by a load variation, w* is the reversed plastic zone
size, and y/ is the monotonic fracture strain appropriate for the structural
region of size /. There results [62] when co* ^> /
or for small scale yielding when the stress intensity factor variation, AK,
controls the size of the reversed plastic zone,
the latter form for the edge crack of length a or central crack of length
2a in a large body.
The result is not insensitive to the way the singularity is dealt with.
For example, if only strain variations up to the point at which the dis-
tance to the crack tip is the structural size are considered of consequence
in determining separation, which would be consistent with the treatment
of monotonic fracture given by McClintock and Irwin [14], there results
And integrating with Ay£. = 2j0(co*/x — 1),
which resembles a fourth power dependence only for or ^>> /. Converting

the failure criterion into one of a critical hysteresis energy absorption,
T07/* per unit volume,
and
The variety of possibilities
Downloaded/printed by is clearly endless, further pointing to the re-
quirement of a better knowledge of the micromechanisms involved in

material separation. Some results tabulated by McClintock tend to sug-
gest the reasonableness of such estimates. Taking / equal to 5 /u (2 X
10~4 in.), a typical cell or subgrain size, the required ratio of the fracture
strain to yield strain may be computed so as to fit experimental results to
Eq 6\d, replacing shear stresses by analogous tensile stresses. Within
factors of about 2 to 4, known monotonic fracture strains for two alumi-
num alloys, two high strength steels, and mild steel appear accurately
estimated, and the materials appear to be correctly ordered as to ductility
[62].
The fact that Eqs 59 and 61 do not lead to major discrepancies lends
hope that further progress may be made in relating continuum solutions
to microstructural separation. Specifically, better continuum plasticity
analyses and clearer pictures of separation mechanisms are needed.
Crack blunting must be better modelled if varying amplitude loadings
are to be handled. Probably the mathematical simplifications inherent
in the boundary layer approach will make small scale yielding solutions
the first obtained. As has been noted, these are adequate for virtually all
of the usual high-cycle fatigue life. But the effects of large overloads and
low-cycle fatigue loadings causing nonlocalized plastic flow cannot be
handled until complete large scale yielding analyses are available.
Acknowledgment
The financial support of work leading to this report by a National
Science Foundation Research Initiation Grant (GK-286) is gratefully
acknowledged.
References
[1] H. Neuber, Kerbspannungslehre (English translation available from Edwards
Bros., Ann Arbor, Mich.), 1st edition, Berlin, 1937, and 2nd edition, Berlin,
1958.
[2] G. N. Savin, Stress Concentration Around Holes, Pergamon Press, New
York, 1961.
[3] G. R. Irwin, "Fracture Mechanics," Structural Mechanics (Proceedings of
1st Naval Symposium), Pergamon Press, New York, 1960.
[4] P. C. Paris and G. C. Sih, "Stress Analysis of Cracks," Fracture Toughness
Testing and Its Applications, ASTM STP 381, Am. Soc. Testing Mats., 1965.
[5] N. I. Muskhelishvili, Some Basic Problems of the Mathematical Theory of
Elasticity, P. Noordhoff, 1953.
[6] P. C. Paris and F. Erdogan, "A Critical Analysis of Crack Propagation
Laws," Transactions, Am. Soc. Mechanical Engrs., Series D (Journal Basic
Engineering), Vol 85, No. 4, 1963.
[7] J. R. Rice, "Plastic Yielding at a Crack Tip," Proceedings, International
Conference Fracture, Sendai, Japan, 1965.
[8] J. R. Rice, "Contained Plastic Deformation Near Cracks and Notches Under
Longitudinal Shear," International Journal of Fracture Mechanics, June,
1966.
[9] J.Downloaded/printed
R. Rice, "Stresses byDue to a Sharp Notch in a Work Hardening Elastic
Plastic Material Loaded by Longitudinal Shear," Journal of Applied Median^

ics. (to appear, ASME Paper 67-APM-l).
[10] J. R. Rice, "On the Theory of Perfectly Plastic Anti-Plane Straining,"
Technical Report NSF GK-286/2, Brown University, March, 1966. ;
[11] D. H. Avery and W. A. Backofen, "Nucleation and Growth of Fatigue
Cracks," Fracture of Solids, John Wiley & Sons, Inc., New York, 1963.
[12] J. C. Grosskreutz, "A Critical Review of Micromechanisms in Fatigue,''
Fatigue—An Interdisciplinary Approach, Syracuse University Press, Syra-
cuse, N.Y., 1964.
[13] W. Prager and P. G. Hodge, Jr., Theory of Perfectly Plastic Solids, John
Wiley & Sons, Inc., New York, 1951.
[14] F. A. McClintock and G. R. Irwin, "Plasticity Aspects of Fracture Mechan-
ics," Fracture Toughness Testing and Its Applications, ASTM STP 381, Am.
Soc. Testing Mats., 1965.
[15] J. A. Hull and F. A. McClintock, "Elastic-Plastic Stress and Strain Distribu-
tion Around Sharp Notches Under Repeated Shear," 9th International
Congress Applied Mechanics, Vol 8, Brussels, 1956.
[16] G. R. Irwin and M. F. Koskinen, discussion and author's closure to Ref 17.
[17] M. F. Koskinen, "Elastic-Plastic Deformation of a Single Grooved Flat
Plate Under Longitudinal Shear," Transactions, Am. Soc. Mechanical Engrs.,
Series D (Journal Basic Engineer), Vol 85, 1963.
[18] H. Neuber, 'Theory of Stress Concentration for Shear Strained Prismatical
Bodies with Arbitrary Non-Linear Stress Strain Law," Journal Applied
Mechanics, Vol 28, December, 1961.
[19] J. C. Grosskreutz, "A Theory of Stage U Fatigue Crack Propagation,"
Report AFML-TR-64-415, Air Force Materials Laboratory, March, 1965.
[20] V. Weiss, "Analysis of Crack Propagation in Strain-Cycling Fatigue," in
Fatigue—An Interdisciplinary Approach, Syracuse University Press, Syracuse,
N. Y., 1964.
[21] S. Rhee and F. A. McClintock, "On the Effects of Strain-Hardening on
Strain Concentrations," Proceedings, 4th National Congress Applied Mechan-
ics, Am. Soc. Mechanical Engrs., 1962.
[22] G. I. Barenblatt, "Mathematical Theory of Equilibrium Cracks in Brittle
Fracture," in A dvances in Applied Mechanics, Vol VII, Academic Press, New
York, 1962.
[23] D. S. Dugdale, "Yielding of Steel Sheets Containing Slits," Journal of
Mechanics and Physics of Solids, Vol 8, 1960.
[24] B. A. Bilby, A. H. Cottrell, and K. H. Swinden, 'The Spread of Plastic Yield
From a Notch," Proceedings, Royal Society A, Vol 272, 1963.
[25] J. N. Goodier and F. A. Field, "Plastic Energy Dissipation in Crack Propa-
gation," Fracture of Solids, John Wiley & Sons, Inc., New York, 1963.
[26] F. A. Field, "Yielding in a Cracked Plate Under Longitudinal Shear,"
Journal Applied Mechanics, Vol 30, 1963.
[27] E. Smith, "Fracture at Stress Concentrations," Proceedings, International
[25] L. M. Keer and T. Mura, "Stationary Crack and Continuous Distributions of
Dislocations," Proceedings, International Conference Fracture, Sendai, Ja-
pan, 1965.
[29] G. T. Hahn and A. R. Rosenfield, "Local Yielding and Extension of a Crack
Under Plane Stress," Acta Metallurgica, Vol 13, no. 3, 1965.
[30] B. A. Bilby and K. H. Swinden, "Representation of Plasticity at Notches by
Linear Dislocation Arrays," Proceedings, Royal Society A, Vol 285, 1965.
[31] R. Hill, The Mathematical Theory of Plasticity, Clarendon Press, Oxford,
1950.
[32] D. C. Drucker, "A Continuum Approach to the Fracture of Metals," in
Fracture of Solids, John Wiley & Sons, Inc., New York, 1963.
[33] F.Copyright
A. McClintock,
by ASTM "Effect of rights
Int'l (all Root reserved);
Radius, Stress, Crack
Mon Dec Growth, EST
7 14:40:45 and 2015
Rate
on Fracture Instability," Proceedings, Royal Society A, Vol 285, 1965.
[34] J. D. Morrow, "Cyclic Plastic Strain Energy and Fatigue of Metals," In-
ternal Friction, Damping, and Cyclic Plasticity, ASTM STP 378, Am. Soc.
Testing Mats., 1965.
[35] D. R. Donaldson and W. E. Anderson, "Crack Propagation Behavior of
Some Airframe Materials," Proceedings, Crack Propagation Symposium,
Cranfield, College of Aeronautics, 1962.
[36] J. L. Swedlow, M. L. Williams, and W. H. Yang, "Elasto-Plastic Stresses and
Strains in Cracked Plates," Proceedings, International Conference Fracture,
Sendai, Japan, 1965.
[37] R. Rosenfield, P. K. Dai, and G. T. Hahn, "Crack Extension and Propaga-
tion Under Plane Stress," Proceedings, International Conference Fracture,
Sendai, Japan, 1965.
[38] D. R. Dixon and J. S. Strannigan, "Strain Distributions Around Cracks in
Ductile Sheets During Loading and Unloading," Journal Mechanical Engi-
neering Science, Vol 7, No. 3, 1965.
[39] N. Thompson and N. J. Wadsworth, "Metal Fatigue," Advances in Physics
(Supplement to Philosophical Magazine), Vol 7, 1958. '
[40] N. E. Frost, "The Growth of Fatigue Cracks," Proceedings, International
[41] R. W. Hertzberg, "Application of Electron Fractography and Fracture
Mechanics to Fatigue Crack Propagation in High Strength Aluminum Al-
loys," Ph.D. thesis, Lehigh University, Bethlehem, Pa., 1965.
Approach, Syracuse University Press, Syracuse, N. Y., 1964.
Philosophical Magazine, Vol 7, 1962.
[44] H. W. Liu, "Fatigue Crack Propagation and the Stresses and Strains in the
Vicinity of a Crack," Applied Materials Research, Vol 3, No. 4, 1964.
[45] G. R. Irwin, "Fracture Mode Transition for a Crack Traversing a Plate,"
Transactions, Am. Soc. Mechanical Engrs., Series D. (Journal Basic Engi-
neering), Vol 82, No. 2, 1960.
of Fatigue," Trend in Engineering, (University of Washington), Seattle,
Washington, Vol 13, No. 1, 1961.
Interdisciplinary Approach, Syracuse University Press, Syracuse, N. Y., 1964.
[48] J. Schijve and F. A. Jacobs, "Fatigue Crack Propagation in Unnotched and
Notched Aluminum Alloy Specimens," NLR-TR M.2128, Amsterdam, 1964.
[49] J. Schijve, A. Nederveen, and F. A. Jacobs, "The Effect of Sheet Width on
the Fatigue Crack Propagation in 2024-T3 Alclad Material," NLR-TR M.
2142, Amsterdam, 1965.
[50] A. J. McEvily and W. Dig, "The Rate of Fatigue Crack Propagation in Two
Aluminum Alloys," NACA-TN 4394, 1958.
[51] W. Dig and A. J. McEvily, "The Rate of Fatigue Crack Propagation fol
Two Aluminum Alloys Under Completely Reversed Loading," NASA TN-D-
52, 1959.
[52] P. C. Paris, "The Growth of Cracks Due to Variations in Loads," Ph.D.
thesis, Lehigh University, Bethlehem, Pa., 1962.
[53] D. Broek and J. Schijve, 'The Effect of Sheet Thickness on the Fatigue-
Crack Propagation in 2024-T3 Alclad Sheet Materials," NLR-TR M.2129,
Amsterdam, 1963.
[54] H. F. Hardrath, "Cumulative Damage," Fatigue—Interdisciplinary Approach,
Syracuse University Press, Syracuse, N. Y., 1964.
[55] J. R. Rice and F. P. Beer, "On the Distribution of Rises and Falls in a Con-
tinuous Random Process," Transactions, Am. Soc. Mechanical Engrs., Series
DCopyright
(Journal Basic Engineering), Vol 87, No. 2, 1965.
[5(5] J. R. Rice, F. P. Beer, and P. C. Paris, "On the Prediction of Some Random
Loading Characteristics Relevant to Fatigue," Acoustical Fatigue in Aerospace

Structures, Syracuse University Press, Syracuse, N. Y., 1965.
[57] S. H. Smith, "Fatigue Crack Growth Under Axial Narrow and Broad Band
Random Loading," Acoustical Fatigue in Aerospace Structures, Syracuse
University Press, New York, N. Y., 1965.
[55] H. Hardrath and E. C. Naumann, "Variable Amplitude Fatigue Tests of
Aluminum-Alloy Specimens," Fatigue in Aircraft Structures, ASTM STP
274, Am. Soc. Testing Mats., 1959.
[59] F. Erdogan and R. Roberts, "A Comparative Study of Crack Propagation in
Plates Under Extension and Bending," Proceedings, International Conference
Fracture, Sendai, Japan, 1965.
[60] A. K. Head, 'The Growth of Fatigue Cracks," Philosophical Magazine, Vol
44, Series 7, 1953.
[61] A. K. Head, 'The Propagation of Fatigue Cracks," Journal Applied Mechan-
ics, Vol 23, September, 1956.
[62] F. A. McClintock, "On the Plasticity of the Growth of Fatigue Cracks,"
Fracture of Solids, John Wiley & Sons, Inc., New York, 1963.
[63] J. R. Rice, "Fatigue Crack Growth Model: Some General Comments and
Preliminary Study of the Rigid Plastic Strip Model," Lehigh University
Institute Research Report, December, 1962.
Theory of Infinitesimal Dislocations Distributed on a Plane," Proceedings,
International Conference Fracture, Sendai, Japan, 1965.
[65] N. E. Frost and D. S. Dugdale, 'The Propagation of Fatigue Cracks in Sheet
Specimens," Journal Mechanics Physical Solids, Vol 6, No. 2, 1958.
[66] H. W. Liu, "Crack Propagation in Thin Metal Sheets Under Repeated
Loading," Transactions, Am. Society Mechanical Engrs., Series D (Journal
Basic Engineering), Vol 83, No. 1, 1961.
[67] H. W. Liu, "Fatigue Crack Propagation and Applied Stress Range—An
Energy Approach," Transactions, Am. Soc. Mechanical Engrs., Series D
(Journal Basic Engineering), Vol 85, No. 1, 1963.
[68] H. W. Liu, discussion of Ref 47.
[69] A. R. C. Westwood, "Environment Sensitive Mechanical Behavior—Status
and Problems," Technical Report 65-5, RIAS (Martin Co.), June, 1965.
[70] N. E. Frost, 'The Effect of Environment on the Propagation of Fatigue
Cracks in Mild Steel," Applied Materials Research, Vol 3, No. 3, 1964.
[71] J. A. Bennett, "Effect of Reactions with the Atmosphere During Fatigue of
Metals," Fatigue—An Interdisciplinary Approach, Syracuse University Press,
Syracuse, N. Y., 1964.
[72] H. C. Rogers, 'The Tensile Fracture of Ductile Metals," Transactions, Am.
Institute Mining, Metallurgical, and Petroleum Engrs., Vol 218, 1960.
[73] F. A. McClintock and F. J. Ryan, "On the Rate of Growth of Fatigue
Cracks," Journal Applied Mechanics, Vol 76, 1954.
[74] A. A. Wells, "Application of Fracture Mechanics At and Beyond General
Yielding," British Welding Journal, November, 1963.
[75] L. F. Coffin, Jr., "Low Cycle Fatigue—A Review," Applied Materials Re-
search, Vol 1, No. 3, 1962.
DISCUSSION
W. E. Anderson1 (written discussion)- —Have you attempted a plas-

ticity solution for the case where the engineering stress-strain curve
peaks directly after the elastic portion and decreases continuously to
failure?
M . R. Achter2 (written discussion) —Your Class III mode looks as if
it might represent high temperature failure. A line scribed on the un-
deformed specimen surface will be seen, after high temperature deforma-
tion, to have an offset at the point where it crosses a grain boundary.
From this, it is inferred that adjacent grains suffer shear displacement
with respect to each other along the grain boundary plane. Cracks are
initiated in these grain boundaries and are propagated inward from the
surface. Do you think that you could apply your Class III treatment to
the case of intergranular failure at high temperature?
/. C. Grosskreutz3 (written discussion) —Is it possible, at the present
stage of elastic-plastic theory, to quantitatively describe the plastic relaxa-
tion process observed by Dr. Laird? If so, how would the crack length
enter into the crack growth law?
/. R. Rice (author) —Upon examining the elastic work hardening plas-
tic anti-plane strain solution for a crack described herein, I find the
solution method to fail for the unstable class of stress-strain relations
noted by Mr. Anderson. Discrete slip line models appear to be an ap-
propriate model of observed discontinuous yielding when the instability
is abrupt as in upper and lower yield point phenomena, but no analysis
is available for more gradual relaxations of flow stress with strain. Grain
boundary sliding in creep as noted by Dr. Achter could be modelled as
a crack with relative motion of surfaces in the two sliding modes, just as
such models have been useful in describing the dependence of uncon-
strained flow and crack nucleation on grain size in the Petch-Stroh analy-
sis of slip line blocking (dislocation pile-up) at a grain boundary.
Dr. Grosskreutz has indicated a very important point in connection
with Dr. Laird's observations of striations and their interpretation in
terms of a fatigue mechanism. As is clear both from dimensional analysis
and the various elastic-plastic models examined herein, a crack growth
rate proportional to crack length (more generally, stress intensity factor
Senior supervisor, Fail-Safety-Fracture Mechanics, Commercial Airplane

Div., The Boeing Co., Renton, Wash.
2
Head, High Temperature Alloys Branch, Metallurgy Div., Naval Research
Laboratory, Washington, D. C.
Midwest Research Inst.,
Downloaded/printed by Kansas City, Mo.
DISCUSSION ON MECHANICS OF CRACK TIP DEFORMATION 311
squared) will occur if advance is to result purely from a ductile crack

opening by sliding off of material at the tip in each load cycle. Further,
every striation would then have to be geometrically similar to every
other striation (same ratio of spacing to depth). These results are clearly
in violation of observations: the general trend of growth rate data is much
closer to a proportionality to crack length squared, and Dr. Laird has
shown pictures indicating an increasing ratio of striation spacing to
depth as crack length or stress level is increased. This raises the question
as to whether striations are in themselves indicative of a separation
mechanism or are simply observers on the scene. For example, suppose
we imagine a ductile material subjected to a cyclic loading while a crack
is steadily cut through the material with a razor blade. Striations will re-
sult as an inevitable consequence of the cyclic loading combined with a
material capability for permanent deformations, but they then have
nothing to do with the separation mechanism. This is precisely the prob-
lem with Dr. Laird's suggestion that striation markings necessarily indi-
cate a mechanism of separation in fatigue—other interpretations are
possible. In particular one interpretation is that the observation of stria-
tions simply means that crack opening displacements (approximately
equal to striation depth) are comparable in size to the growth increment
per cycle (striation spacing). Indeed, application of Eqs 43b, 44, and 45
for crack opening displacements to the data for the two aluminum alloys
of Figs. 24 and 25 gives a prediction of opening displacements compara-
ble in size to the measured growth per cycle over the middle range of
the data in both cases. From the plane strain slip line field for a sharp
crack in Fig. 21, it would appear that large strains can occur directly in
front of the crack only if the tip is rounded. Thus crack opening displace-
ments are indicative of the size of the highly strained region in front
of the crack, so that a growth rate per cycle roughly comparable in size
to opening displacement is not unexpected. Still, the general trend of
growth rate data shows clearly that the growth per cycle is not directly
proportional to crack opening displacement, as required in Dr. Laird's in-
terpretation. A length parameter characterizing the material must enter as
discussed in the last section of the paper.
S. R. Swanson,1 F. Cicci,1 and W. Hoppe1
Crack Propagation in Clad 7079-T6

Aluminum Alloy Sheet Under Constant
and Random Amplitude Fatigue Loading
REFERENCE: S. R. Swanson, F. Cicci, and W. Hoppe, "Crack Propaga-

tion in Clad 7079-T6 Aluminum Alloy Sheet Under Constant and Random
Amplitude Fatigue Loading," Fatigue Crack Propagation, ASTM STP 415,
Am. Soc. Testing Mats., 1967, p. 312.
ABSTRACT: This paper presents the results of a comprehensive experi-
mental study, mainly using square panels, of a recently introduced high-
strength aluminum alloy sheet material and its resistance to the propaga-
tion of fatigue cracks. An axial-load fatigue machine, applying both
constant and Rayleigh random loads, was developed for the test program,
and fatigue cracks in the panels were continuously monitored by a new
servo-controlled eddy-current crack follower. The resulting traces of crack
length permitted the study of several crack growth laws, by manipulation
of the loads to maintain a theoretically constant crack growth rate. From
the tests where the load levels were not altered, an optimum thickness of
sheet (yielding a minimum average crack growth rate) was indicated. Load-
shedding, a phenomenon inherent in the failure of elements in redundant
(fail-safe) structures, was simulated in many of the tests. It was found that
appreciably different load-shedding histories still yielded the same instan-
taneous crack growth rate for a given instantaneous level of stress inten-
sity factor. It was found that fracture mechanics provided the most effec-
tive approach for the study of fatigue crack propagation, either by constant
amplitude or by Rayleigh random amplitude loading.
KEY WORDS: fatigue (materials), crack propagation, aluminum alloy,
random-amplitude loading, sheet (metal), cumulative damage, fracture
mechanics
Nomenclature
a Half the crack length, in.
di Distance of hole center within a notch from the specimen longi-
tudinal center line
k Constant used in Eq 4
m Stress exponent in Eq 3
1
Formerly senior research engineer (structures and materials), research engineer
(aeronautical), and research engineer (metallurgical), respectively, de Havilland
Aircraft of Canada, Ltd., Downsview, Ontario, Canada.
312
SWANSON ET AL ON CLAD 7079-16 ALUMINUM ALLOY SHEET 313
n Number of cycles, crack length exponent in Eq 3

Yi Notch drill radius
ry Size of the plastic zone at the crack tip, inches
t Sheet thickness, inches
A Constant
B Fixity constant for sheet buckling (Appendix II)
C Material constant in Eq 3
E Modulus of elasticity
max , (-Kp)max , K\ , KZ , K% , Ks
K, K
Values of the stress intensity factor at the crack tip, using the
maximum stress, using the maximum stress with the length of
the center notch, and with crack lengths LI , Z, 2 , L3, Ls, re-
spectively, lb-in.~ 3/2 , psiVin., or ksi\/in.
Kc Stress intensity factor associated with fast fracture
Kt Theoretical stress concentration factor
L, LQ , LI , Z/2, L% , Ls
Instantaneous crack length, length of the center notch, length
associated with the onset of shear lip development, fully estab-
lished shear, re-established plane strain cracking, and onset of
tensile tearing of specimen, respectively, in.
TV Number of cycles
P, Po Instantaneous maximum load, initial maximum load, Ib
W Sheet width, in.
&g j Ga.lt > O"crit
Values of stress amplitude with respect to the gross area, the
alternating component of load, the critical stress for buckling,
respectively, psi or ksi
o-ys Yield stress (0.2 per cent offset)
<f> Angle of shear lip development
045/1/3/5/125/50 Sample specimen identification
045—laboratory drawing number
1—the number of the sheet from which the blank was taken
(Fig. 16)
3—location of the blank within the sheet
5-—nominal specimen width, in.
125—nominal specimen thickness = 0.125 in.
50—specimen number
3/L/1/125 Sample control tension specimen identification
3—sheet number (Fig. 16)
L—longitudinal (grain direction)
1—location within the sheet
125—nominal sheet thickness = 0.125 in.
Initiation Number of cycles required for a fatigue crack nucleus at the
center notch toby grow to such an extent as to become visible on
Downloaded/printed
the untreated front clad surface ot the specimen with the unaided
eye.
Since the fatigue failure of well-designed "fail-safe" aerostructural
materials in service under the action of random loading processes is
usually a matter of progressive cracking, we undertook an investigation
of a recently introduced aerostructures material, clad 7079-T6 aluminum
alloy sheet, to study its crack propagation behavior under cyclic loads.
From a knowledge of the general literature already available in the
field, we found that:
1. Most data have been generated under conditions of constant ampli-
tude (CA) involving dynamic loading with a sine-wave of invariant form
and size [1,7]2.
2. Most of the data on crack propagation (with certain exceptions) was
obtained by manual (visual) recording of crack lengths at discrete inter-
vals with traveling microscopes, and the monitoring was usually stopped
after the crack had progressed to about half the width of the sheet
[1,4,20}.
3. Much of the data has been generated with panels whose free length
between grips is greater than one finds in aerostructures in actual service
[16}.
4. In the majority of testing, the magnitudes of the fatigue loads were
held constant during the test, even though the area of cross section was
decreasing. This practice is so prevalent that it is usually not explicitly
pointed out in the reports. This type of loading will be referred to as No
Load Shedding (NLS) in this report.
If the crack length could be reasonably monitored, Weibull found that
he could achieve, after an initial nonlinear stage, a constant rate of crack
growth up to high percentages of crack length/specimen width. The load
history he used was one in which the load magnitude decreased linearly
to zero as the percentage of width covered by the crack rose to 100
(hereafter referred to as Linear Load Shedding, LLS).
These observations influenced our choice of objectives in the research
project described in this paper, and the program divided naturally into
the following phases:
1. Construct a fatigue machine capable of applying random loading as
well as constant amplitude loading to sheet specimens. Since the design
so far has involved the Gaussian white noise excitation of a single degree
of freedom, all random load results presented in this paper refer to a
Rayleigh distribution of peaks. The process is referred to in this report as
Rayleigh Random Amplitude (RRA). We also obtained a background of
2
paper. Copyright by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
SWANSON ET AL ON CLAD 7079-T6 ALUMINUM ALLOY SHEET 315
ahhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
in the literature. The material is described in Appendix I. Having both
types of data permits cumulative damage studies of the two types of load-
ings (CA, RRA). The machine, designated R3, is described in Ap-
pendix II.
2. Design the sheet specimens with an unsupported length/width ratio
more typical of that found in the geometry of panel elements in redundant
structures. A brief survey of such panels in aircraft revealed that this
ratio is very often about unity, and that unity would be a practical ratio
to examine first. A few tests were also carried out with an unsupported
length/width ratio of 2.5.
3. Design a means of continuous monitoring of crack length. After
considerable study, a servo-controlled eddy-current crack tip follower
was developed. This device is described in Appendix III.
4. Use the output of the continuous crack monitor to permit the shed-
ding of the fatigue load hi the predetermined manner. This idea, sug-
gested by Weibull's work [7], has certain advantages. It eliminates the
necessity of calculating slopes from crack growth curves in determining
which growth rule is applicable. Instead, we postulate that, if the rule
(usually equating crack growth rate to a function of various quantities—
stress, length, etc.) is valid, then it should indicate the correct manipula-
tion of variables to obtain a constant rate of crack growth as the test
progresses. We would then carefully adjust the loading according to the
rule to obtain this constant rate from an initial value. If it failed to
yield a constant rate in the test, the rule would be considered suspect, or
of limited applicability.
For the constant amplitude tests described in the first section of this
paper the stress levels investigated were 7 ± 1 ksi (4.9 ± 4.9 kg/mm2), 7
± 5 ksi (4.9 ± 3.5 kg/mm2), and 7 ± 3 ksi (4.9 ± 2.1 kg/mm2). These
tests covered a range of (averaged) crack growth rates from 10~6 to 10~3
in./cycle.
The Rayleigh random amplitude tests are described in the second
section for the loading 7 + 2.5 ksi rms (4.9 + 1.76 kg/mm2).
The initiation of fatigue cracks was studied intensively and is discussed
in the third section.
Fatigue Crack Propagation—Constant Amplitude
No Load Shedding (CA-NLS)

This popular method of testing sheet material hi crack propagation
studies was undertaken first. In carrying out all our tests the operator
would note the commencement of loading, the time to first visual indica-
tion of a crackbyatASTM
Copyright eitherInt'l
side(allof the reserved);
rights center notch on the
Mon Dec front surface,
7 14:40:45 EST 2015 the
time to initiation on the opposite side of the notch, and finally the cycles
or time to final failure of the specimen. After failure, the specimen frac-
ture surface was examined for further infonnatíon (Fig. 1).
Considering only the information detailed previously, we can assess
time or cycles to initiation and cycles involved in propagation of a crack
from visible indication to final failure. All initiation information is dis-
cussed later, since all load-shedding histories start at the same load
levéis and only differ after a fatigue crack length has been established.
The cycles involved in propagation from initiation to final failure in a
FIO. 1—Fatigue machine wilh specimen and crack follower imtalled.
CA-NLS test yield an average crack propagation rate, which ignores the
actual instantaneous variation in rate which we will discuss later, The
data from these tests are shown in Table 1.
The variation of average crack growth rate with sheet thickness is re-
vealed in Fig. 2. While we were usually able to test only one specimen for
each configuration of loading and thickness and width, the results show
an interesting trend. It appears that íhe average crack growth rate is higher,
for a given loading, at 0.080 and 0.250 in. thickness, than it is for 0.125 or
0.160 in. thickness. The results thus indícate that, for the CA-NLS test,
and considering propagation
Copyright by ASTM Int'l (all only, there is Mon
rights reserved); an optimum thickness
Dec 7 14:40:45 which
EST 2015
will yield a minimum value for average crack growth rate. In certain cases
where the test was repeated (for example, at 0.080 in. thickness) one can
estimate the scatter in growth rate for a particular thickness (see for ex-
ample Table 1).
Plots of the log crack growth rate versus the log of instantaneous
(maximum) stress intensity factor are shown in Fig. 3 for three CA-NLS
tests. It can be seen that the slope is generally about 4 as indicated in
other researches in this field.
After the test, the specimen fracture surfaces were studied for macro-
scopic features. For the CA-NLS tests (and also the RRA-NLS tests),
there appeared to be several successive morphological regimes (Fig. 4).
Because of the effectiveness of applying fracture mechanics principles to
later tests, the concept of crack tip stress intensity factor K will be em-
ployed in describing these changes.
1. For a given test, the value of K associated with the length of the
center notch and the maximum stress amplitude is referred to as (Ko)max •
This is the parameter used later to discuss initiation. From visible initia-
tion the crack progresses in a flat or 90-deg mode across the thickness
perpendicular to the loading. After progressing a certain distance LI ,
shear lips begin to develop at the sheet surface. The K associated with
this length (and corrected for finite width using the tangent formula) is
referred to as KI (Fig. 4). From tests (Table 1) we found that Lx was gen-
erally independent of thickness t.
2. At the onset of shear lip activity the crack enters a mixed-mode
regime. The shear (45 deg to surface) area increases from the surface
until the plane strain region at midthickness is reduced to zero. The
crack lerigth at-which plane strain cracking drops to zero and the shear
regions join is called L2 , and the appropriate K is called K2 .
L2 varied with thickness in such a manner as to imply, for instance,
that the angle <j> made by the increasing area of shear as L progressed
from LI to L2 (see Fig. 4) varied in a manner similar to the average crack
rate in Fig. 2. These findings (contained in Table 5 of the preprint) were
intended to form Table 2 of this paper, but were omitted to reduce the
volume of data. These data are, of course, available from the authors or
the preprints for interested readers.
In developing shear lips at 45 deg to the sheet surface, there seems to
be no correlation between the 45-deg angle chosen on the front surface,
say, and that chosen on the back surface. However, if the same angle is
chosen on both surfaces, the centroid of the crack front is invariably
pulled vertically in the opposite direction to that indicated by the
resulting V or arrow shape. This behavior is shown in Fig. 10. But if op-
posite angles are chosen, this effect cancels out, and the vertical posi-
tion of the crack front remains at the level of the prior plane strain crack-
ing. This behavior is schematically
Downloaded/printed by illustrated in the following sketch.
TABLE 1—Test results at constant amplitude. Values of stress intensity factors KI , K 2 , and K 8 , average propagation rates
and size of plastic zone r y .
Kilocycles to Average0 Relative
Specimen Visual Initiation r
L\
•
, in.
Ki,
. . ?— la. , in. Ki Crack Ls, in. Ks Kilocycles ry , in. ry/t Humidity.
ksi-v/m. Growth to Failure
Left Right Rate %
No Load Shedding 7 ± 3 ksi (CA-NLS)
045/2/3/5/080/100 82.5 150.0 1.78 9.3 2.01 10.0 20.0 4.1 21.8 198.6 0.0111 0.139 38
045/2/3/5/125/103 364.0 ... 2.18 10.6 3.07 14.1 24.3 4.0 20.6 459.7 0.0202 0.164 50
045/2/1/5/160/102 205.0 283.6 2.01 9.7 3.07 14.1 19.0 4.4 26.9 327.4 0.0196 0.121 50
045/1/4/5/250/104 719.0 665.1 1.75 9.2 3.22 14.8 28.5 4.1 21.8 746.8 0.0214 0.086 70
045/2/2/7.5/080/90 68.0 117.0 1.80 9.0 2.49 10.9 38.0 5.8 23.6 159.8 0.0132 0.165 54
045/1/3/7.5/125/91 103.0 1.89 9.3 2.65 11.3 38.1 5.9 24.3 194.7 0.0128 0.103 54
045/2/3/7.5/160/89 126.0 2.23 10.2 3.63 14.2 64.6 6.3 28.4 172.0 0.0202 0.124 56
045/1/5/7.5/250/87 314.0 2.10 9.8 3.59 13.9 55.5 6.1 26.2 376.7 0.0188 0.075 59
045/1/1/10/125/93 102.0 2.12 9.8 3.84 13.8 121.7 8.3 31.7 140.2 0.0190 0.154 58
045/2/1/10/160/94 140.0 2.72 11.2 3.56 13.1 128.1 8.0 29.1 176.3 0.0172 0.107 60
045/2/1/12.5/080/112 63.0 96.3 1.43 7.9 2.15 9.7 60.3 10.2 33.9 159.3 0.0104 0.132 34.5
045/1/2/12.5/125/96 49.2 85.0 2.04 9.5 4.14 13.8 85.1 9.0 27.0 117.5 0.0190 0.153 56
045/2/3/12.5/160/95 50.0 81.0 1.98 9.4 3.39 12.5 85.9 9.5 29.5 117.6 0.0157 0.097 58
No Load Shedding 7± 5 ksi (CA-NLS)
045/1/-/5/080/18 30.9 50.2 0.77 7.0 1.10 8.76 38.9 4.2 27.8 90.6 0.0085 0.106
045/2/2/5/080/57 23.1 27.1 0.78 7.1 1.22 9.0 52.2 4.3 29.7 67.6 0.0090 0.113 48
045/1/4/5/125/58 31.4 34.9 0.83 7.3 1.37 9.6 53.3 4.2 27.8 75.0 0.0092 0.075 43
045/3/1 5/5/125/222* 37.5 31.0 1.00 8.0 2.10 11.2 65.5 4.45 33.5 66.7 0.0131 0.094 63
6
045/3Copyright by ASTM Int'l (all 37.0
/17/5/125/224 40.2 Mon
rights reserved); 1.05Dec 7 14:40:45
8.3 2.10201511.2
EST 74.0 4.3 29.8 68.2 0.0131 0.094 63
045/1Downloaded/printed
/-/5/1 60/56 by 12.5 23.8 1.13 8.5 2.09 12.4 59.8 4.2 27.8 51.4 0.0154 0.101 70
045/1/3/5/250/55 63.7 63.1
University of Washington (University 0.71 pursuant
of Washington) 12.4 No76.0
2.09 Agreement.
6.7 to License 4.0 93.7
24.7authorized.
further reproductions 0.0150 0.060 71
045/2/3/7 5/080/59 21 1 21 1 0.92 7.6 1.29 9.1 89.5 6.4 35.5 60.1 0.0092 0.116 42
045/1 /-/I. 5/080/27 24.7 41.4 1.11 8.4 1.38 9.4 58.2 6.6 39.5 84.7 0.0098 0.123 25
045/1/4/7 5/125/61 27 3 28 3 0.84 7.3 1.83 10.9 56.5 6.4 35.6 89.1 0.0118 0.097 26
045/1 /-/7. 5/1 60/60 33 5 33 1 0.57 6.0 1.81 10.9 65.4 6.35 34.8 86.5 0.0119 0.079
045/1/6/7 5/250/62 27 3 27 5 1.10 8.4 2.11 11.9 130.7 6.53 37.9 54.0 0.0138 0.055 66
045/2/3/10/080/63 12 7 16 2 0.71 6.7 1.32 9.1 126.0 8.5 40.7 49.6 0.0092 0.116 39
045/1/5/10/125/65 22 6 27 2 0.74 6.8 1.82 10.8 108.4 8.7 43.7 65.5 0.0116 0.095 45
045/2/2/10/160/64 27 3 25 3 2.26 12.1 116.0 65.4 0.0147 0.091 37
045/2/6/12 5/080/67 17 5 116 0.97 7.8 1.37 9.2 223.5 8.5 30.0 37.6 0.0094 0.118 62
045/1 /-/I 2 5/080/19 18 0 18 0 0.88 7.4 1.25 8.9 225.2 8.2 28.7 43.8 0.0087 0.109 29
045/1/6/12 5/125/68 19 7 19 7 0.96 7.8 2.03 11.3 199.0 9.1 33.0 48.9 0.0128 0.105 61
045/1 /-/12 5/160/48 22 0 20 0 0.92 7.6 1.75 10.6 177.7 10.0 39.0 52.7 0.0113 0.075 51
045/2/5/12 5/160/69 16 8 18 9 0.88 7.4 2.14 11.6 160.1 9.7 36.7 53.1 0.0135 0.083 43
No Load Shedding 7 ± 7 ksi (CA-NLS)
045/2/5/5/080/79 13.6 12.4 0.56 7.0 0.93 9.0 107.6 4.1 30.5 34.0 0.0090 0.114 59
045/1/2/5/125/80 9.4 10 2 0.37 5.6 1.17 10.2 87.1 4.1 30.5 36.1 0.0104 0.085 54
045/2/4/5/160/81 10.8 12.7 0.67 7.6 1.37 11.2 88.7 4.1 30.5 37.0 0.0126 0.078 52
045/1/1/5/250/82 18.7 11 8 0.40 5.9 1.33 10.9 94.9 4.1 30.5 36.3 0.0116 0.047 48
045/1/9/5/250/111 9.9 12.8 0.55 6.9 1.48 11.6 110.7 3.8 26.0 30.9 0.0131 0.053 50
045/2/6/7 5/080/84 91 10 1 0.79 8.3 0.97 9.1 216.8 5.8 33.0 25.2 0.0092 0.115 58
045/1/1/7 5/125/85 72 85 0.68 7.7 1.23 10.4 151.1 6.0 35.3 30.3 0.0108 0.088 58
045/2/1/7 5/160/86 83 78 0.65 7.4 1.71 12.3 155.8 6.2 38.1 30.2 0.0152 0.094 60
045/2/4/10/080/88 9.9 10 9 0.70 7.7 1.04 9.4 303.9 7.3 34.6 25.5 0.0099 0.125 56
045/1/2/10/125/74 78 84 0.73 7.8 1.32 10.6 214.3 8.0 40.7 29.5 0.0113 0.092 64
045/2/4/10/160/76 10.1 71 0.70 7.7 1.44 11.2 186.0 7.6 37.0 32.1 0.0126 0.078 42
045/2/5/12.5/080/70. .. . 5.3 6.2 0.88 8.68 1.10 9.7 581.0 7.0 28.6 15.3 0.0104 0.130 52
045/1/4/12.5/125/71.
Copyright by ASTM .Int'l 7.4 reserved);
. . (all rights 7.9 Mon 0.88Dec 7 14:40:45
8.7 1.40201510.8
EST 287.6 7.0 28.6 27.6 0.0116 0.094 50
0 Downloaded/printed by
Microinches per cycle.
b University of Washington (University of Washington)
Free length of specimen extended to 12.5 in.pursuant to License Agreement. No further reproductions authorized.
From Table 1 we can see that for a given stress level the values of K2 are
lowest at 0.080 in. thickness, and increase with thickness. There is also
some indication of a flattening out or a fall in the value of K2 at 0.250 in.
thickness, at low stress levels (see Fig. 5). This behavior of K2 is likely
related to the corresponding opposite behavior in average crack growth
FIG. 2—Variation in average crack growth rates with sheet thickness for con-
Copyright by ASTM
stant amplitude—no Int'l (all tests.
load shedding rights reserved); Mon Dec 7 14:40:45 EST 2015
FIG. 3—Log-plotting of crack growth rate versus maximum stress intensity

factor for CA-NLS and CA-KLS tests.
PIG. 4—Typical fatigue fracture surfaces for constant amplitude tests in-
volving: (a) no load shedding; (b) linear load shedding; (c) load shedding for con-
stant K (low K).
rate. That is, a high value of K2 is associated with a low value of crack
growth rate.
Irwin [2] has suggested that the average crack rate is a function of ry ,
the plastic zone size. Since rv is mainly a function of K2 , this correlation
is understandable from an examination of Fig. 5.
FIG. 5—Variation of stress intensity factor K2 for fully developed shear with
thickness. CA-NLS test results.
Similar optimum thickness K2 effects have been noted by Katz and Ab-
bott in their study of the static fracture toughness of steel for rocket
motor cases [5].
3. As the crack now progresses in a fully developed shear mode,
usually referred to as a regime of plane stress cracking, a length Ls is
finally reached which seems to indicate the cessation of what could be
calledCopyright
fatigue, and the beginning of a tensile tearing away of the sheet ma-
terial. Downloaded/printed
The associated value
by of stress intensity K8 was calculated (Table 1).
It was found that the Value of Ks appears considerably less than rough
estimates of the value of Kc for this material (Appendix I).
When we examined the continuous trace of crack length with time, we
found the expected steady increase in crack growth rate from a very
low value at initiation, rising exponentially to very high rates as the
cracking entered the final tear-away region. The sensitivity of the con-
tinuous crack monitor did, however, allow us to detect, in several cases,
a momentary slowdown in crack growth at L2 . A good example of this
is shown in Fig. 6.
FIG. 6—Typical crack length-time trace for a constant amplitude—no load

shedding fatigue test.
In this particular case, the (local) crack growth rate fell from 77.7 /^m./
cycle to 37.2 pin./cycle—about a factor of 2, as suggested by McEvily
for 7075-T6 [20]. The crack growth rate soon increases rapidly again
after L2 due to the relentless rise in net section stress associated with a
CA-NLS test. This behavior is understandable in that much more energy
per unit area is required to create shear lips than to create flat fracture
[3].
Linear Load Shedding (CA-LLS)

Weibull [7] has stated "that the rate of propagation, independently of
the crack length, is determined by the nominal stress amplitude based on
the remaining area of by
Downloaded/printed the specimen and some other parameters."
Accordingly, the first load-shedding test we tried when the crack moni-
tor was perfected was one involving a linear dropoff of load
Indications from the literature were that after an initial transient period
in which the crack growth rate would increase to a stable value, we would
obtain linear crack growth at constant rate. This load shedding history is
shown as constant net section stress in Fig. 7.
From a large number of tests (Table 2) we found that, for the square
panels between grips, the initial transient regime is usually quite large,
and that in the latter 50 per cent of crack length/width there is a definite
FIG. 7—Theoretical curves and experimental increments of load ratio versus

crack length to specimen width ratio for load shedding tests.
slowing down of crack growth, in some cases resulting in a nonpropa-

gating crack. The crack length trace (Fig. 8) is therefore not a straight
line, though the middle portions appear straight because they represent
a region containing a point of inflection from increasing rate to decreasing
rate of crack growth.
A recent paper by Benham and Moag [4] reports similar observations
in crack growth rate (initial increase, constant rate, decreasing rate) for
a CA-LLS type test. A possible correlation between the final slowing
down in crack growth rate and the geometry of the specimens is discussed
in Appendix I.
The morphology of cracking associated with the CA-LLS test was quite
interesting (Fig. 4). After the initial plane strain mode, the transition to
plane stress or the shear mode is unstable. While the load is initially too
high to prevent transition from occurring, once it does occur, the large
amount of fretting which usually shows up indicates that the load level is
not high enough to maintain the shear in a fully developed state. The
cracking usually reverts to the flat mode finally, at L3 , at large percent-
ages of crack length/width.
Figure 8 also shows the variation in stress intensity with crack length
for a typical CA-LLS test. Note that the stress intensity factor Ks associ-
ated with this reestablishment of plane strain cracking is usually of the
same magnitude as KI (Table 2). This indicates the reversible nature of
the K associated with transition from plane strain to plane stress or vice
versa.
Constant K Load Shedding (CA-KLS)
From a study of the relevant principles of fracture mechanics it was
decided to perform crack propagation tests in which the crack tip stress
intensity factor for maximum and alternating stress was held constant,
that is,
P = P0(tan ^Lo/tan AL)1/2 (2)
This type of load shedding history is shown in Fig. 7. A sample specimen
surface is shown in Fig. 4, and an example of the resulting crack length
trace appears in Fig. 9.
From a large number of traces over a range of width, thickness, and
loading (Table 3) it became evident that this rule invariably produced
constant rate(s) of crack growth, for unsupported square panels of sheet.
Figure 9 was quite typical of all the (low K) CA-KLS tests. Small varia-
tions from a constant rate that occurred in some tests always coincided
with asymmetry in cracking from the center notch.
This finding for square panels is unusual, since we are using an ap-
proximate tangent formula first proposed by Irwin [2] for K for a wide
plate with an infinite row of cracks. More exact solutions for the
theoretical case of infinitely long panels [6] would yield a higher effec-
tive value of K than that arising from the tangent formula. Thus we would
expect to see a slightly increasing rate of cracking at values of crack
length/width approaching 100 per cent.
The square dimensions of the unsupported portion of sheet between
the grips is the most likely cause for this discrepancy, since in both KLS
(tangent formula) and LLS tests, cracking is slower than expected in the
latter stages (Appendix I). This suspicion was confirmed by tests on
"long" panels.
For values of K below KI the crack progressed under plane strain con-
ditions (flat 90 deg as in Fig. 4). Just above this value of K± (in certain
tests at 7 ± 7 ksi) the crack developed a shear lip and abruptly began to
progress at abyslower
Copyright (but
ASTM Int'l (all constant) rateMon
rights reserved); inDec
a mixed-mode configura-
7 14:40:45 EST 2015
tion. Downloaded/printed
We did not test by with K high enough to crack the specimen with a
TABLE 2— Constant amplitude test results with linear load shedding with calculated values of stress intensity factors KI , K 2 , and K 3 .
Kilocycles to Visual Relative
Initiation Ki, Kilocycles
Specimen Li , in. Li , in. Kt , ksiVin. La , in. Ki , ksi-v/in. to Failure Humidity,
ksivin. %
Left Right
Constant Amplitude 7 ± 5 ksi (CA-LLS)
045/3/3/5/080/131 32. 0 24.2 0.70 6.2 248.3 34

045/2/6/5/125/132 40. 1 47.7 1.00 6.9 209.2 33
045/3/13/5/125/220° 45..0 38.0 157.5 47
045/3/16/5/125/223° 29..5 29.7 113.3 57
045/1 /-/5/1 60/1 33 35..0 32.7 181.3 39
045/1/19/5/250/211 89..6 54.4 1.10 7.0 229.2 26
045/2/4/7.5/080/135 32,.5 20.0 0.85 6.9 1.49 8.4 5.07 8.0 270.4 29
045/1/5/7.5/125/136 32.,2 29.6 0.95 7.3 2.46 11.9 4.58 8.6 306.8 25
045/2/4/7.5/160/137 35..0 32.6 0.85 6.9 2.81 9.5 4.28 8.9 305.4 22
045/1 1 -/I. 5/160/162 32. 0 32.7 0.87 7.0 2.43 9.4 3.94 9.1 294.2 28
045/2/1/10/080/138 21. 0 22.3 0.70 6.6 1.26 8.4 7.75 7.9 377.1 21
045/1/4/10/125/139 15. 5 18.2 1.10 7.9 1.68 9.3 6.34 9.7 347.5 26
045/2/5/10/125/163 23..4 25.8 0.90 7.3 1.66 9.2 6.58 9.5 381.4 29
045/2/3/10/160/120 20. 0 20.3 1.10 7.9 2.90 10.6 5.40 10.5 151.9 42
Copyright by ASTM Int'l (all rights reserved);19.
045/1/1/12.5/125/122 ,4 Dec17.8
Mon 0.88
7 14:40:45 7.4
EST 2015 1.74 9.7 10.73 7.1 212.1 37
045/2/1/12 5/160/123 by
Downloaded/printed 26,.0 18.8 0.88 7.4 2.42 10.8 9.01 9.7 252.3 29
Constant Amplitude 7 ± 7 ksi (CA-LLS)
045/3/4/5/080/124 12. 3 12.5 0.50 6.3 1.29 8.6 3.59 7.2 167.3 38
045/2/4/5/125/125 9.0 12.8 0.60 6.7 1.29 8.6 4.07 6.0 115.3 36
045/1 /-/5/160/126 10 .0 10.5 0.60 6.7 1.57 8.9 4.02 6.1 106.8 32
045/1/8/5/250/127 17 .3 15.1 0.50 6.3 1.95 9.0 3.99 6.2 119.3 34
045/2/1/7.5/080/116 9. 8 8.5 0.53 6.7 1.09 8.9 5.63 8.4 122.4 33
045/1/2/7.5/125/117 9.,4 9.1 0.53 6.7 1.47 9.8 6.73 5.5 139.3 28
045/2/2/7.5/160/118 10,.7 10.1 0.53 6.7 1.61 10.1 6.29 6.9 114.3 27
045/1/4/7.5/250/119 17..3 19.1 0.53 6.7 2.32 10.8 112.4 26
045/2/2/10/080/115 5. 0 5.0 0.70 7.7 1.26 9.8 8.62 7.3 220. 46 66
045/1/3/10/125/113 7..0 7.0 0.70 7.7 1.92 11.3 8.67 7.2 98.1 54
045/3/2/10/160/114 5,,2 5.2 0.70 7.7 1.86 11.1 8.67 7.2 77.3 51
045/2/2/12.5/080/129 4,.5 5.0 0.88 8.6 1.12 9.6 10.90 7.9 94.5 47
045/1/5/12.5/125/130 9 .0 8.4 0.88 8.6 1.48 10.7 65.4 41
a
b
Free length of specimen extended to 12.5 in.
No failure, crack no longer propagating.
FIG. 8—Composite illustration of a typical crack length-time trace for a con-

stant amplitude-linear load shedding test and the instantaneous value of stress in-
tensity ratio for the same test.
fully developed shear mode throughout propagation. The tendency for

the shear lips to deflect the locus of the vertical position of the crack
front as described in the discussion of CA-NLS tests was also manifest
in these "High K" tests. In the few tests where it was possible to measure
this change in trace slope at transition, the slowing down was of the same
order of magnitude as observed in Fig. 6.
The sudden occurrence of transition in a KLS test was found to be
attributable to the stepwise manual reduction of loading (Fig. 7). At high
values of K, if the load level was left constant for even a small increment
longer in crack length than theoretically necessary, it could, in certain
cases, be just sufficient to initiate transition into the mixed mode. A
similarCopyright
behaviorby has
ASTM been
Int'lnoted by Broek
(all rights et alMon
reserved); [7] recently. Conceivably,
Dec 7 14:40:45 EST 2015
if the Downloaded/printed
shedding of loadbywere ideally smooth (maintaining K constant) the
crack would progress from initiation in either one mode or the other,
subject to the constraints imposed by the geometry of the center notch.
Figure 3 shows the log-log behavior for crack growth rate and (maxi-
mum) stress intensity for CA-KLS tests. This provides an interesting
comparison with the superimposed CA-NLS test results discussed earlier.
For the range of crack growth rates covered, one can see that prior history
(for a given K value) has an insignificant effect for maximum stress
intensity. If there was a prior history effect, a particular relation in Fig.
3 between a given K and the associated crack growth rate for one load
FIG. 9—Typical crack length-time trace for a constant amplitude-constant K

load shedding test.
history would differ from the equivalent relation for the other due to the
large difference between no load shedding and constant K load shedding
histories, as shown in Fig. 7.
Other Load Shedding Histories

Many different workers have studied fatigue crack propagation under
(NLS) constant amplitude cyclic loads. Their general results can usually
be described by the equation:
results are summarized
by in Table 4.
TABLE 3—Constant amplitude test results with load shedding to preserve
a constant value o/K.
Kilocycles to da/dn Relative
Visual Initiati n K, Kilocycles Humid-
Specimen ° to Failure (Trace)
hhhh f/in./cycle ity,
Left Right %
Constant Amplitude 7 ± 7 ksi (CA-KLS)
045/4/2/5/080/174 13.7 13.7 5460 352.5 6.8 27

045/2/7/5/125/158 15.8 17.2 5460 318.1 6.8 21
045/4/6/5/160/209 15.7 18.7 5460 238.5 8.2 27
045/1/16/5/250/176 17.8 19.4 5460 260.0 8.7 18
045/4/1/7.5/080/165 13.6 12.0 6690 423.7 8.6 27
045/2/4/7.5/125/167 14.1 14.5 6690 556.2 11.2 22
045/2/5/7.5/160/168. 12.3 11.1 6690 270.6 10.9 26
045/1/2/7.5/250/177 11.3 15.8 6690 227.4 14.8 18
045/2/6/10/080/98 9.6 11.7 7720 201.3 19.9 40
045/2/6/10/125/166 13.0 11.0 7720 283.7 12.6 22
045/2/5/10/160/75 8.6 9.1 7720 192.0 24.0 54
045/3/3/12.5/080/16 10.0 11.6 8630 234.6 21.1 20
045/1/3/12.5/125/17 8.4 9.5 8630 313.4 17 A 21
Constant Amplitude 7 =b 5 ksi (CA-KLS)
045/4/1/5/080/179 42.0 37.0 4680 585.4 5.3 25

045/3/2/5/125/178 33.0 35.3 4680 429.0 6.7 26
045/4/5/5/160/208 40.0 27.6 4680 331.0 6.5 27
045/1/18/5/250/210 55.0 102.6 4680 373.2 6.5 25
045/3/3/7.5/080/180. 24.0 26.6 5730 513.0 6.2 20
045/2/5 7.5 125/181 35.0 25.5 5730 579.8 5.6 18
045/2/6/7.5/160/182 42.0 36.9 5730 (1000.0) 4.5 14
045/1/1/7.5/250/212 33.5 27.7 5730 338.5 10.6 34
045/3/3/10/080/164 21.6 18.0 6620 528.9 8.4 22
045/1/6/10/125/97 17.8 13.9 6620 178.4 25.6 56
045/3/1/10/160/140 21.2 18.4 6620 380.5 11.5 40
045/2/6/10/160/72 19.6 11.1 6620 192.5 23.6 63
045/2/3/12.5/080/12 14.7 9.6 7390 d 22.2 63
045/2/3/12.5/125/20 21.0 20.6 7390 410.6 13.6 29
045/3/8/5/125/2156 41.0 68.0 4680 231.5 9.9 51
045/3/9/5/125/216* 61.6 52.4 4680 300.0 8.5 42
045/3/14/5/1 25/221 6 36.5 44.1 4680 253.0 9.6 47
045/3/10/5/125/217" 48.7 42.0 4680 308.7 8.1 43

045/3/11/5/125/218' 38.0 49.0 4680 207.4 10.6 63
045/3/12/5/125/219c 35.4 32.5 4680 268.4 8.2 43
0
Nonpropagating crack. Growth rate constant to 37 per cent sheet width then
decreases nonlinearly.
6
Free length of specimen extended to 12.5 in.
c
Free length of specimen extended to 12.5 in. The Isida finite width correction
was used.
d Copyright by ASTM Int'l (all rights reserved);
Test stopped at 220.0—total crack lengthMon
= Dec
10.5 7in.,
14:40:45 EST
due to 2015
excessive trans-
verse vibration.
33Q
Tests ai a Constant Value of al\* L—This type of load shedding is

also shown in Fig. 7. Apparently where this rule is used, it is applied
only to the alternating load. A plot of the load shedding for the alternat-
ing component (or zero mean load), and for maximum stress when we
use 7 ksi mean stress, is shown in Fig. 7. It can be seen that, for appreci-
able values of mean stress, the crack will behave in a manner similar to
a CA-NLS test, while for small values of mean stress, we should expect
TABLE 4— Results of other studies of fatigue crack propagation under

NLS constant amplitude cycling.
Stress Crack Length
Author Exponent, Exponent, Remarks
m n
Head [5] 3 and - 1 3/2 first to set up a mechanistic law

involving hardening at tip;
main flaw — a constant size as-
sumed for the plastic enclave
Liu [9] . . 2 1 elegant similarity argument for
size of enclave proportional to
L
Frost [10] . .. 3 1
Schijve [12].. 2.6 1.5
Paris [77] 4 2 based on "broad trend of data"
Weibull [ / ] « . . (function of 0 for CA-LLS tests only
material)
McEvily and Boett-
ner [13]. . . 2n 2 for low values of a
3 for high values of a
b
Valluri [14] 3 1
0
For all n 7* Q, the general relation implies zero rate for zero length, and a pre-
existing flaw must be assumed.
b
Valluri's model has the feature that the final critical crack length varies with
stress level. Any model which has a constant crack length at failure results in linear
damage accumulation. Frost [10] and Williams [75] have both shown that the
final critical crack length relation for failure has the relation:
o-eL1-*/2 = constant (4)
For most brittle materials, k = 0.9 to 2.0. For glass, it is unity.
a prolonged region of plane strain. In actual tests at 7 ± 5 ksi we ob-

tained crack growth very similar to that of a typical CA-NLS test:
Kilocycles to Initiation Kilocycles

SPeCimen
£S S^T- ^ Failure
045/2/2/10/125/183 23.8 25.7 118.7
Tests at a Constant Value of (aga L?I2)/(A — <rg)—While we were un-

able toCopyright
test with this load
by ASTM shedding
Int'l (all historyMon
rights reserved); (based
Dec 7on work EST
14:40:45 by Head
2015 [8]),
we anticipate that this type of loading will yield failures closely resembling
those for CA-KLS tests. From Fig. 7, one can see that the resulting load
histories for constant crack growth rate are remarkably similar in the
initial stages.
Tests at Constant Lever Displacement—This type of test was carried
out in consideration of the recommendation by Barrois [16] that, for a
practical test for fail-safe redundant structures, constant strain ampli-
tude would be more appropriate than constant stress tests. As it turned
out, the limited testing we carried out of this type (at 7 ± 5 ksi) gave
results almost identical to the CA-NLS test:
Kilocycles to Initiation Kilocycles

Specimen to Failure
Left Right
045/S/-/10/160/5 17.8 29.0 67.5
Possibly this is due to the limited ductility of the high strength alumi-
num alloy sheet, combined with the influence of St. Venant's principle, at
least for the early stages of cracking.
Extended Free Length Specimens

Since cracking in the latter stages of both the LLS and KLS testing
was slower than anticipated, we decided to test several specimens with
an unsupported length/width ratio of 2.5. The specimen configuration
used was identical to that of the 5-in. wide specimen of Fig. 17 except
that the total length became 22.5 in. (571.5 mm). The stress level used
was 7.0 ± 5.0 ksi (4.9 ± 3.5 kg/mm2).
Ten such specimens were tested and the results compared with the
square panel tests. The median values of the number of cycles to visual
initiation agreed well with the results from the cracks in the square panel
specimens, although the latter tended to initiate slightly sooner.
From the values in Table 1 it can be seen that in the CA-NLS tests
the long specimens showed a considerably faster average crack growth
rate, showing that the effective instantaneous values of £max are higher
than at the same length of crack in the square panels.
As expected, the CA-LLS tests on the extended free length specimens
showed an increased crack growth rate (Table 2). Also, the apparently
linear region of crack growth rate was extended over a wider range of
crack length/specimen width ratio (0.2 to 0.8). However, the latter por-
tions of the curves still showed a decreasing crack growth rate, beginning
at 80 per cent length/width ratio. That portion of the crack growth
curves which seemed linear also showed a much higher crack growth rate
than did the equivalent portion of the curves for the square panels.
When the CA-KLS
Copyright by ASTMtests were
Int'l (all carried
rights out Mon
reserved); on the
Decextended
7 14:40:45free
ESTlength
2015
specimens (Table 3), using
Downloaded/printed by the tangent finite-width correction, the crack
growthUniversity
rate wasof found to increase
Washington (Universityslightly with increasing
of Washington) pursuant tocrack
Licenselength.
Agreement. No fu
When the more exact correction suggested by Isida [6] was used, the
crack growth rates were found to be constant.
From Table 3 it can also be seen that for the CA-KLS tests using the
extended free length the crack growth rate is considerably larger than for
the square panels.
FIG. 10—Composite illustration correlating a typical Rayleigh random loading

with no load shedding with the resulting crack length and the crack surface mor-
phology.
Fatigue Crack Propagation—Rayleigh Random Amplitude
No Load Shedding (RRA-NLS)

In this test, the rms of the alternating load (equivalent to 2.5 ksi) and
the mean load (7 ksi) were held constant throughout. The material, sub-
jected to the stationary random process (Appendix II), cracked in a
manner similarbytoASTM
Copyright whatInt'l
one(all
would
rightsexpect from
reserved); MonaDec
CA-NLS test.EST
7 14:40:45 A typical
2015
crack Downloaded/printed
length trace and by
a sample fracture are shown in Fig. 10. In Fig. 10
TABLE 5—Summary of results Rayleigh random amplitude tests (RRA),

7 + 2.5 ksi RMS.
Time to Visual Average
Initiation, min Time to Crack
Specimen Finall failure Growth
Left Right min. Rate,
It in./cycle
No Load Shedding (RRA-NLS)
045/4/5/5/080/143 101.0 75.0 302.9 5.2

045/2/8/5/125/144 53.0 55.0 176.5 9.7
045/2/7/5/160/142 58.5 67.7 184.5 9.5
045/3/7/5/160/204 65.0 73.0 170.3 11.3
045/4/2/7.5/080/149 65.0 64.0 290.3 7.9
045/2/2/7.5/125/146 46.8 52.5 161.0 15.7
045/3/5/10/080/150 51.0 53.0 229.5 13.4
045/3/2/12.5/080/152 36.0 33.0 158.6 23.7
Linear Load Shedding (RRA-LLS)
045/3/9/5/080/202 61.6 62.0 1072.0

045/2/9/5/125/171 51.0 52.5 590.0
045/3/7/5/160/159 56.4 58.0 1249. 03
045/2/6/7.5/125/186 33.0 45.0 724.2
Constant K Load Shedding (RRA-KLS)
045/4/7/5/080/173c 52.2 61.4 1990.0* 0.79

045/3/4/5/160/1720 70.0 95.0 1288.0 1.19
045/1/6/7. 5/1 25/1 84d 65.4 58.0 1164.1 1.41
Other Tests—Constant o-ait L
045/2/1/7.5/125/185 42.0 53.0 249.0

a
No failure, crack no longer propagating.
6
No failure, test stopped.
c
#max = 3690 psi\/in-
d
#max = 4530 psMn.
NOTE—Average frequency of testing = 32.5 cps or 1950 cpm (from sample
traces).
we have included a slow-speed continuous trace of the complete fatigue

loading from the oscillograph, and it is possible to correlate certain bursts
of loading with corresponding surface markings, by first correlating with
the crack length trace. The test data are shown in Table 5.
Note that with RRA tests, the crack tip follower moves ahead in abrupt
steps associated with random bursts of loading and then remains station-
ary for a short period of time. The resulting trace of crack length versus
time invariably is more jagged than for a corresponding CA test.
This irregularity in loading seems to be responsible for the fact that
in RRA-NLS tests only was it possible to see clearly with the unaided
eye the progress
Copyright of theInt'l
by ASTM crack frontreserved);
(all rights on theMon
fracture surfaceEST
Dec 7 14:40:45 over the entire
2015
width of the sheet. These crack front markings (likely due to the fretting
action associated with residual compressive tip stresses after a random
burst of high load cycles) are valuable in that they reveal the through-the-
thickness profile for the crack front in the different modes associated with
an NLS test. They show that there is very little curvature in this front,
once the crack is initiated, in the tensile, shear, or tearing phases. A close
study of the RRA-NLS tests reveals that, despite the more irregular
advance of the crack, there is often a clear indication of a slowing down
at transition. An example of this behavior is shown in Fig. 10.
Finally, when comparing RRA-NLS tests with CA-NLS tests, there is
FIG. 11—Log-plotting of crack growth rate versus maximum stress intensity

factor for RRA-NLS and RRA-KLS tests.
a tendency in the former to exhibit linear crack growth (constant rate)

before transition, compared with the smoothly increasing rate associated
with the latter type.
A log-log plot of the maximum stress intensity factor K versus crack
growth rate for RRA-NLS tests is shown in Fig. 11, for comparison with
the corresponding CA presentation in Fig. 3.
Linear Load Shedding (RRA-LLS)

This type of test, with the rms of the alternating load, and the mean
load being reduced to zero, constitutes the application of a quasi-station-
ary random loading to the specimen. In all respects this type of loading
resulted in crack
Copyright length Int'l
by ASTM traces
(all which behavedMon
rights reserved); (in Dec
a more jaggedEST
7 14:40:45 fashion)
2015
in essentially the same way that the CA-LLS tests performed (Fig. 8).
The surface of the crack was also similar to that shown in Fig. 4 for the
CA-LLS specimen.
These test results are presented in Table 5. Note that in one case the
crack finally became nonpropagating at a high value of crack length/
width. Since the general slowing down in crack rate occurs with both
CA-LLS and RRA-LLS tests, we have further evidence that the con-
straints associated with panels of square dimensions, rather than the type
of loading, are responsible for part of the deviation from expected crack
growth behavior.
Constant K Load Shedding (RRA-KLS)

For the lightly damped single degree of freedom system, we found
that by decreasing the rms level of the alternating stress and the mean
stress according to the tangent-formula constant K load-shedding curve
(Fig. 7), we obtained a constant rate of crack propagation.
The test results (Table 5) cover quite long periods (in one case five
days) for the chosen rms level. Since the stepwise reduction of loading
was manually carried out, these tests often required overnight interrup-
tions. It was found, however, that by removing all loading from the
specimen and re-loading on the following morning, no change (in slope
or position) occurred on the crack length trace. This procedure was also
applied to certain other tests which ran over the normal working day.
The morphology of cracking in the RRA-KLS tests was exclusively
the flat 90-deg mode associated with plane strain fracture, as might be
expected from the low value of K (2.5 ksi rms).
Figure 11 shows the log-log plotting of the test results for RRA:KLS
tests with the RRA-NLS test results. It is interesting that the relations
are consistent, as they were in CA tests (Fig. 3), indicating negligible
prior history effect.
Having both CA-KLS and RRA-KLS data, we can explore the possi-
bility of cumulative damage prediction in our crack propagation tests,
random amplitude versus constant amplitude.
In Fig. 3 we have shown the results of maximum stress intensity versus
crack growth rate for our CA-KLS and CA-NLS tests. It is also possible
to replot these data using stress intensity factors based on the alternating
stress range, rather than maximum stress, thus avoiding the possibility
of crack growth with merely a mean stress present. We then obtain the
relation (by best visual fit) R = (A/T)4 X 10~14, where R is the half-crack
growth rate in microinches per cycle, and AK is the stress (range) in-
tensity factor. We have used this relation with the equation for the Ray-
leigh probability density to obtain the curve of crack growth rate versus
proportion of unit time at a given intensity of stress, for 2.5 ksi rms and
assuming an RRA-KLS test. The final result of this analysis is:
Working with 5-in.-wide specimens, a check of our experimental re-

sults (Fig, 11 and Table 5) shows:
R = 0.79 and 1.19 în./cycle (Specimens 172 and 173)
A similar analysis using (Kmax)* was also carried out, yielding a pre-
diction of the RRA-KLS test of 1.78 nin./cycle.
We believe it is quite significant that both predictions should approxi-
mate so closely the actual Rayleigh test results. This shows that linear
damage accumulation actually applies for crack propagation in our ma-
terial, with no prior history or sequence effect under Rayleigh random
load conditions.
Other Load Shedding Histories

We were able to carry out only one test with a constant value of o-ait L
using RRA loading. As expected, the high mean stress (7 ksi) resulted in
a crack growth trace very similar to the RRA-NLS test (Table 5).
Study of Crack Initiation

The test data for visual initiation shown in Fig. 12 are the result of
constant amplitude tests only, using square panels. Since the original
design of the test specimens was based on maintaining the stress con-
centration factor Kt constant at 7 for all widths, the initial values of
(Ko)max increase with sheet width.
The RRA values (Table 5) are reasonable values to expect if one
considers the rms level roughly equivalent to the alternating stress ampli-
tude.
In Fig. 12 we see that sheet thickness has little effect at high stress
levels but becomes appreciably significant at low stress levels. Since
practical structures are usually designed for relatively low fatigue stresses,
this finding would indicate that for a given stress level, it is worthwhile
to use thicker sheet, if there is a choice.
An example where this would apply is in the design of bottom wing
sections. In this case the critical design case arises from a positive
wing bending
Copyrightinvolving tension,
by ASTM Int'l withreserved);
(all rights a secondary case7 (for
Mon Dec negative
14:40:45 bend-
EST 2015
ing) involving compression. From the above information, the designer

might use stringer-skin combinations which involve proportionately more
cross-sectional area in the skin in satisfying the compression case, to
help him in the tensile (fatigue-sensitive) design case.
A metallographic study of the crack surface at initiation was carried
out by fatigue testing until a crack was sighted, removing the specimen,
cutting away nearly all the uncracked material on either side of the
notch, splitting the remaining section vertically through the center of
the notch, and carefully loading these "edge-notch" specimens to failure
in a tension test machine.
Figure 13 shows typical results of this procedure for a high stress level
and for a lower stress level constant amplitude test. The high stress level
FIG. 12—Variation in cycles to visual initiation with loadings width and thickness.
test appears to generate a multitude of crack nuclei, some not necessarily

right at the plane of maximum stress (under axial loading). As a result,
crack fronts develop quickly along the thickness joining up by shear with
abrupt (but small) changes in vertical elevation from one crack site to the
next. It is understandable, from such a concurrent formation of crack
nuclei across the thickness, that the thickness of the sheet would not
greatly influence the cycles to visual initiation.
At low stress levels we have a different picture. Only a few cracks are
nucleated, and then usually only at the plane of maximum stress. These
few nuclei grow radially, fanning out until they join up to form a through-
the-thickness crack. The resulting crack surface at the edge of the notch is
usually much flatter. Also, since the steady propagation of an isolated
nucleus finally results in the visual observance of a crack at the surface
it is understandable
Downloaded/printedthat
by sheet thickness would be a significant parameter,
since it would take longer on the average to propagate through the thicker
gages of material, from the interior of the notch.
As a corollary to this finding, one can appreciate that actual crack
initiation is a much more symmetrical phenomenon than one would
assume from the recorded values of visible crack initiation. A study of
the fracture surface thumbnail cracks reveals that often there exists an
FIG. 13—Differences in initiation—7 ± 7 ksi (CA) and 7 ± 5 ksi (CA).

approximately equal amount of crack area on either side of the notch.
Chance differences in the distribution of the cracked area along the
notch bring about differences in the cycles to visible "initiation" at the
front surface.
By sectioning a test specimen parallel to the surface plane, removing
one third of the thickness, it was possible to observe the formation at
the crack tip (Fig. 14) in the core material as the crack formed under
plane strain conditions near the notch. Apparently as the crack tip
passes Downloaded/printed
through an individual
by
grain, plasticity effects become significant,
FIG. 14—Metallograph of the crack tip in plane strain cracking.
FIG. 15—Fractographic study of initiation Rayleigh random amplitude 7 +

2.5 ksi rms Specimen No. 045/3/3/5/.160/194.
and oblique slip bands result from shear strain slip. This crack growth
mechanism was discussed by Laird3 as a form of plastic relaxation, and
also by Schijve4 at this symposium.
Under conditions of random loading at the low rms level chosen (2.5
ksi), we found an interesting combination of the two mechanisms dis-
cussed earlier. The wide variation in load cycle magnitude is probably
responsible for the fact that many nuclei (Fig. 15) are observed at the
edge of the notch (and distributed vertically) as with the high stress level
CA tests. However, once away from the edge, these nuclei coalesce into
3
4
SeeCopyright
p. 131. by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
Downloaded/printed
See p. 415. by
a few major crack fronts. When we studied the fracture surface at high
optical magnification (Fig. 15) we often found examples of striations or
beach marks ahead of the macrocrack front. The occurrence of groups
of striations ahead of the macrocrack front was also observed with CA
test specimens.
The fractographs also showed the characteristic Type 2b brittle frac-
ture morphology described by Forsyth [17], that is, fracture facets nor-
mal to the maximum tensile stress direction. These plateaus or facets are
covered with the characeteristic fatigue striations remaining as waves or
ripples in the direction of crack development. There are two types of
Stage 2 cracks:
Type 2a—Ductile fatigue fracture striations which have the beautifully
regular ripple appearance, like a conical serrated file.
Type 2b—Brittle fatigue fracture striations. This type is similar to
the ductile pattern, but the uniformity of the ripples is marred by super-
imposed cleavage fracture rivers. These brittle striations usually lie near
some well-defined crystallographic plane.
Finally, the tendency of RRA fatigue to result in crack nucleus forma-
tion similar to that found at high stress levels correlates with earlier ob-
servations by Swanson [18] with bar specimens. It was found that the
S-N behavior (especially in terms of scatter of endurances) of random
loading tests was in many ways similar to that observed at high stress
levels under constant amplitude loading. With nonredundant bar speci-
mens, NLS fatigue endurances are largely governed by macrocrack initia-
tion rather than propagation.
Conclusions
1. For the square panels of clad 7079-T6 aluminum alloy studied,
there appears to be an optimum (as-supplied) thickness of sheet yielding
a minimum average crack growth rate for CA-NLS tests. This finding,
probably due to fabricating history, is of practical significance in design
involving nonredundant structures, and should be checked for statistical
soundness by repeat tests. Its interpretation must also be qualified by
the findings on variation of cycles to initiation with thickness, since the
total number of cycles to failure is often the more significant parameter.
The test results also indicate that the stress intensity factor K2 , asso-
ciated with the commencement of fully developed shear fracture, is a
factor in this behavior, since the average crack growth rate tends to vary
inversely with the value of K2 .
The stress intensity factor KI , associated with the onset of shear lips,
was found to be a material constant, independent of sheet thickness, al-
though the angle of shear lip development shows the "optimum thick-
ness"Copyright
effect.by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions a
2. The momentary drop in crack growth rate in both random and

constant amplitude loading (with no load shedding) associated with
transition from tensile mode to shear mode cracking indicates that the
shear mode is intrinsically more resistant to propagation. This behavior
was also confirmed in the CA-KLS tests (at high K values), and in the
usual appearance of a kink in the log-log plots of both CA-NLS and
RRA-NLS test results. This would indicate that, under certain condi-
tions of load shedding in a practical redundant structure, it is conceivable
for cracking to occur at a faster rate than if the load were not shed as
much, simply because the crack stays in the tensile mode.
3. For the KLS type tests it was soon confirmed that we always ob-
tained a constant rate of crack growth, as predicted by Paris and others
working in fracture mechanics, for both constant amplitude and Rayleigh
random amplitude loadings. This result was initially fortuitous, since we
found, at first, that load shedding to maintain constant K, using the ap-
proximate tangent formula rather than more exact solutions to correct
for finite width of the square-panel specimen, led to constant rates of
crack extension with square panels. The greater effective K associated
with the tangent formula approximation was balanced by the drop in
effective K due to using square panels. By testing with an exact solution,
and using long specimens with 21/2 times the square-panel free length, we
again obtained a constant crack growth rate.
4. We also found that, in all other load shedding tests to study the
relative merits of various crack growth rules, the deviations from a
constant rate of crack growth can be explained by the difference in the
instantaneous value of stress intensity factor dictated by the particular
load shedding, history, and the value of K for a constant K test under the
same conditions.
5. From the morphology of fracture in tests involving linear load
shedding, it was found that the stress intensity factor associated with the
transition from plane strain to plane stress had about the same value as
the factor associated with later transition from plane stress to plane strain.
This indicates a reversibility in the morphology which might not have
been expected, considering possible effects of prior stress history.
6. The crack growth rates from constant amplitude tests at constant
crack tip stress intensity factor (CA-KLS), when compared with the
results of the CA-NLS tests, also showed that prior stress history has no
appreciable effect of its own, since the resulting instantaneous CA-NLS
crack growth rates are consistent with the average values from the CA-
KLS tests. This finding also applies to Rayleigh random amplitude test
results.
7. By comparing crack growth rates under CA-KLS and RRA-KLS
Copyright by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 ES
conditions, we found that by
Downloaded/printed a linear summation of CA rates using the
Rayleigh distribution yielded a good estimate of the actual rate of crack

growth observed under Rayleigh random loading.
8. The development of a servo-controlled eddy-current crack tip
follower permits the automatic control of fatigue loading during the
cracking of a panel. This not only permits fundamental study of crack
propagation by manipulating prior history, but allows the test engineer
to simulate the practical situation of a cracking element in redundant
structure characteristic of fail-safe design.
A cknowledgment
This paper forms the final report for Project 772 carried out as part
of the Canadian National Research Council Industrial Research Assist-
ance Program. The authors would like to express their appreciation to
all those responsible for making this work possible. We are also grateful
to Alcoa International (Canada) Ltd. for their cooperation in supplying
the sheet material for our tests.
APPENDIX I
Specimen Material and Preparation

The material for this study of crack propagation was chosen on the basis
of the practical necessity of learning more about a recently introduced alumi-
num alloy with certain promising characteristics for application in the air-
craft industry. This sheet material was supplied in four mill-rolled thick-
nesses. It was possible, fortunately, to obtain all the 4 by 8-ft sheets of
material from just two melts (the 0.125 and 0.250 in. from one cast, the
0.080 and 0.160 in. from another). When these large sheets were received,
the specimens were allocated according to the layout shown in Fig. 16.
From the information in Fig. 16 one can see that the specimen identification
contains the information regarding the original position of any specimen
and the particular sheet it was taken from.
The chemical composition and an outline of the history of fabrication for
the different thicknesses of clad 7079-T6 aluminum alloy sheets are given in
Table 6.
Note that only the 0.080 and 0.125 in. sheet were cold rolled during
fabrication (intuitively assumed a beneficial process), while the 0.080 in.
sheet was annealed after hot rolling, a possible detrimental process for
fatigue. These intuitive factors might account for the observed "optimum
thickness" behavior in the CA-NLS tests, and would be absent if we had
machined out four thicknesses from, say, a single thick sheet.
Table 7 presents the results of tension tests to determine the static me-
chanical properties, comparing our results with those obtained by Alcoa.
Since it is important to have a repeatable and controlled method of crack
initiation, we decided to employ a center-notch specimen rather than an
unnotched configuration. The specimen edges were not milled after shearing,
and the clad surface
Copyright was not
by ASTM Int'l treated in any
(all rights way. Mon
reserved); Four Dec
specimen widthsEST
7 14:40:45 were2015
chosenDownloaded/printed
(5, 7.5, 10, andby 12.5 in.) with each specimen containing a center
FIG. 16—Layout of specimen blanks within a sheet of material.
notch to yield a stress concentration factor of 7.0. The details of the geometry
of these specimens are presented in Fig. 17.
The multiple-hole notch was decided on for several reasons:
1. It is relatively easy to obtain an accurate and repeatable notch tip
radius with drilled holes. Our values of tip radius are given in Fig. 17.
2. A drilled hole introduces little work hardening in the vicinity of the
stress raiser, if carefully drilled.
3. The notch pattern can be completed to an essentially elliptical shape
by progressively larger holes between the two ends of the notch. Real stress
distributions around a central crack [19] are dominated by the length of
TABLE 6—Fabricating history of Alclad 7079-T6 sheet.
Order number INC 977-513 INC 977-513 INC 977-197 INC 977-198
Item 1 Item 2
Specification 0.125 X 48 X 96 in. 0.250 X 48 X 96 in. 0.080 X 48 X 96 in. 0.160 X 48 X 96 in.
Alclad 7079-T6 Alclad 7079-T6 Alclad 7079-T6 Alclad 7079-T6
Lot number 619-461 619-451 627-591 627-581
Hot roll 15 to 0.162 in. 11 to 0.250 in. 15 to 0.162 in. 15 to 0.160 in.
Anneal 1 hr 775 deg
Cold roll 0.162 to 0.125 in. 0.162 to 0.080 in.
Solution heat treat 20 min soak at 830 F 25 min soak at 830 F 15 min soak at 830 F 20 min soak at 830 F
Stretch for flatness permanent set for flatness for flatness
Artificial age 6 hr. at 190 deg plus 6 hr. at 190 deg plus 6 hr. at 190 deg plus 6 hr. at 190 deg plus
24 hr. at 240 F 24 hr. at 240 F 24 hr. at 240 F 24 hr. at 240 F
Cast Number H517-23 H517-23 H673-04 H673-04
Chemical analysis (MIL-A-8923 ASG)
Cu (0.4-0.8) 0.8 0.8 0.7 0.7
Fe (0.4 max) 0.22 0.22 0.18 0.18
Si (0.3 max) 0.12 0.12 0.11 0.11
Mn (0.1-0.3) 0.22 0.22 0.18 0.18
Mg (2.9-3.7) 3.5 3.5 3.4 3.4
Cr (0.10-0.25) 0.16 0.16 0.16 0.16
Zn (3.8-4.8) 4.6 4.6 4.5 4.5
Ti 0.03 0.03 0.03 0.03
the crack and the radius of curvature at the tip, and are relatively insensitive
to the exact shape at the center.
4. With suitable techniques, specimens can be manufactured quickly. A
template method was devised to manufacture the specimens. A master tem-
plate was made from %-in.-thick commercial grade aluminum alloy plate for
each specimen. In it were placed drill bushings to spot the pickup holes in
TABLE 7—Mechanical properties of Alclad 7079-T6 sheet from control tests
carried out at the de Havilland Aircraft of Canada, Ltd. compared with results from
Alcoa International Canada, Ltd.
Sheet Number Ultimate 0.2% Yield Per Cent Modulus of
Grain Direction Tensile Strength, ksi Elongation Elasticity,
and Thickness Strength, ksi in 2 in. 106 psi
l/L/080 75.4 67.3 12.0 10.0

l/T/080 75.7 65.3 11.5 10.2
2/L/080 74.6 68.1 12.0 10.2
2/T/080 75.0 65.5 11.0 10.1
3/L/080 75.6 68.8 11.5 9.9
3/T/080 76.2 66.6 11.0 10.6
4/L/080 75.8 68.3 11.8 10.2
4/T/080 75.6 66.6 10.3 10.4
Alcoa/L/080 75.6 66.3 10.0
l/L/125 77.7 70.9 12.3 10.3
l/T/125 78.7 68.0 12.0 10.4
2/L/125 77.2 70.1 12.5 9.9
2/T/125 78.1 68.3 11.5 9.9
3/L/125 76.5 69.1 11.5 9.9
3/T/125 78.7 67.6 10.8 10.0
Alcoa/L/125 77.7 68.4 12.0
l/L/160 76.8 70.6 12.8 9.7
l/T/160 78.1 68.1 11.3 10.0
2/L/160 74.7 69.3 11.8 10.0
2/T/160 77.3 68.5 10.3 10.6
3/L/160 73.6 67.1 12.5 10.1
3/T/160 78.9 68.5 11.3 10.6
4/L/160 75.6 69.9 12.3 10.0
4/T/160 77.9 67.8 10.3 10.0
Alcoa/L/160 77.0 67.8 10.0
l/L/250 77.8 71.6 13.0 9.9
l/T/250 78.5 68.7 12.5 9.7
Alcoa/L/250 77.1 67.7 12.8
NOTE—Each DHC value is the average of two tests.
the specimen. A drilled and hardened plate was accurately positioned in and
bolted to the master template. This allows the notch tip holes to be drilled
through the hardened plate at the same time that all pickup holes and the
remaining notch holes are spotted. As a result the overall hole pattern of the
specimen is accurate and free of variations. It was found that the work re-
quired to manufacture a specimen was approximately 1 manhour.
As pointed out at the beginning of this paper, a survey of practical designs
for sheet panels asby
Copyright used in aircraft
ASTM Int'lsoon
(allrevealed
rights that unsupported
reserved); Monpanel
Dec ele-
7 14:40:4
mentsDownloaded/printed
were usually close to a by
square configuration, with stiff clamping along
University of Washington (University of Washington) pursuant to L
one pair of opposite edges. While testing exclusively with such a configuration
in our program (see Fig. 16), we began to uncover evidence that such a geom-
etry, practical as it is, may not simulate adequately the theoretical boundary
conditions for cracking from a central notch, especially at large values of
crack length to specimen width.
Accordingly we cemented a 0.080-in.-thick sheet of photoelastic plastic
to the free (square) surface of a specimen (Fig. 18), and fatigue-loaded the
specimen until we had cracks on both sides of the notch yielding a total
length of 3 in.
We then removed the 5-in.-wide specimen from the fatigue machine,
clamped it with pin-jointed grips to yield a 6-in.-free length, loaded the sheet
FIG. 17—Center notch specimen details (stress concentration factor Kt = T).
in tension in a tension test machine, and photographed the result (Figs. 18a
and b). The specimen was then removed and clamped to leave 11 in. of free
length. We loaded the specimen to the same level and photographed the
photoelastic pattern (Figs. 18c and d).
We can see that by almost doubling the free length there is a slight shift
or narrowing of the pattern directly above and below the center, more con-
sistent with theoretical stress distributions for center-cracked sheet [19,22],
The difference is very small and not conclusive, due to the proximity of the
free edge of the photoelastic plastic, which distorts the picture of strains near
the stress level in the vicinity of the crack will tend to be less for square panels
the constraints. However, the indication appears to confirm that qualitatively
than the theoretical value for an infinitely long specimen (of finite width)
for large values of crack length/specimen width.
This would by
Copyright account for (all
ASTM Int'l therights
observed behavior
reserved); in the
Mon Dec LLS and
7 14:40:45 ESTKLS
2015 (tangent
FIG. 18—Photographs of 5-in. wide, 0.160-in. thick specimen coated with

photoelastic plastic containing 3-in. crack at 7000-lb load.
FATIGUE CRAGATIONACK PROPAGATION
formula) tests, using square panels. Testing long specimens with free length
equal to 2Yz times the width resulted in accelerated cracking, using the tan-
gent formula (Tables 2 and 3).
In order to obtain a rough estimate of the fracture toughness of the ma-
terial, a few static unidirectional tension tests were carried out to obtain Kc,
the (finite width) stress intensity factor associated with fast (unstable or
catastrophic) fracture. By "fast" it is understood that the fracture travels at
TABLE 8—Effect of material on crack initiation and propagation in 5-in. wide spec-
mens at constant load amplitude (CA-NLS) (9.8 ± 5.5 ksi) (thickness = 0.725 in.}.
Stress
Specimen Minimum Kilocycles Plane Strain Intensity
Sheet Material Number Kilocycles to Failure Crack Length Factor Kz ,
to Initiation L! , in. ksi-yin.
7079-T6 50 11.9 32.4 1.51 12.9

7079-T6 53 13.6 35.8 1.38 12.2
7079-T6 51 14.5 36.2 1.59 13.3
2024-T3 10 32.1 97.0 1.12 10.9
2024-T3 15 35.3 103.7 1.06 10.6
2024-T3 9 36.2 97.3 1.14 11.0
7075-T6 14 18.4 56.7 1.29 11.8
7075-T6 13 18.5 52.0 1.05 10.6
7075-T6 12 18.6 57.0 1.17 11.2
TABLE 9—Mechanical properties of the 2024-T3 (QQ-A-362) and 7075-T6

(QQ-A-287) material used to compare crack initiation and propagation properties
with 7079-T6.
Speci- Ultimate 0.2% Per Cent
Material Grain Direction men Thickness, Tensile Yield Elonga-
Num- in. Strength, Strength, tion
ber ksi ksi in 2 in.
2024-T3.... longitudinal 88 0.125 69.8 53.7 19

2024-T3.... longitudinal 89 0.1265 68.6 53.2 18.5
2024-T3.... longitudinal -transverse 90 0.125 65.4 44.7 20
2024-T3.... longitudinal-transverse 91 0.1265 66.7 46.3 20.5
7075-T6. . . . longitudinal 84 0.124 80.1 74.0 12.5
7075-T6. . . . longitudinal 85 0.1245 79.9 73.1 13.5
7075-T6. . . . longitudinal-transverse 86 0.1235 81.7 72.6 12.5
7075-T6. . . . longitudinal-transverse 87 0.1235 82.1 72.4 13.0
a reasonable fraction of the speed of sound in the material. The procedure

used was to fatigue test given specimens (5 in. wide) until the fatigue crack
length was about 30 to 50 per cent of the width. The specimen was then in-
serted in a tension test machine, oscillograph ink was inserted into the frac-
ture, and the specimen was pulled apart. The length of the (admittedly sus-
pect) ink stain was taken as a measure of the length at which fast fracture
occurred. The results were:
Sheet Thickness, in. Kc Values Obtained, ksi-v/in.
0 160 41.3, 40.0

0.125 62.2, 54.0
Considering the cladding, these values are in reasonable agreement with

those presented for bare material by Kaufman [23].
Plasticity Effects
The stress intensity factor K is a valid concept when behavior can be
based on elasticity. It is well suited for materials of limited ductility, a com-
mon characteristic of high strength materials. In any real material capable
of plastic flow, a small region of plasticity will exist at the crack tip due to
the elastic singularity there. If this region of plasticity is small, fracture
mechanics solutions will apply. It is therefore necessary to evaluate this
quantity for our material. The size of the plastic zone rv is usually defined
by the ratio of the stress intensity factor K2 to the yield strength of the ma-
terial [21]:
ry = KS/2<rl
The ratio of this value to the thickness appears relatively constant (Table 1),
of the order of 0.1 for most cases. These values indicate that ductility is
sufficiently limited in 7079-T6 aluminum alloy to permit application of frac-
ture mechanics concepts.
Comparison with Other Clad Materials

A series of tests was also carried out to compare the general CA-NLS
propagation properties of clad 7079-T6 with other materials using the same
fatigue machine. The other materials were 2024-T3 and 7075-T6 clad
aluminum alloy sheet. The stress levels used for all three materials were 9.8
± 5.5 ksi. The experimental results are listed in Table 8. The 5-in.-wide
specimens were identical to the type shown in Fig. 17. The mechanical prop-
erties of the 2024-T3 and 7075-T6 material are given in Table 9.
From the tests, it is indicated that cracks initiate sooner and are propa-
gated faster in 7079-T6 material than in either 2024-T3 or 7075-T6 aluminum
alloys, at least for the CA-NLS stress level and geometry studied.
For these tests the plastic zone size rv was also determined:
Specimen Number ry , in. ry/t
10 0.0208 0.165
15 0.0196 0.156
9 0.0211 0.167
Hertzberg [21] has obtained a value of 0.50, while Liu obtained a value of
0.3 (as reportedbyby
Copyright Hertzberg).
APPENDIX II
Fatigue Test Machine R3
The first shaker-lever machine, Rl, used to test bar specimens is described
in Ref 18. In the early stages of the crack propagation program, a "feasibility
model" shaker-lever machine, R2, was designed to test out the principles of
operation of this type of apparatus with sheet specimens. When Machine R2
proved successful and capable of 1000-lb dynamic loading (CA) using a 25-
Ib shaker, no further development was undertaken related to this program,
and it is now being used in fundamental fatigue studies with small copper
specimens.
FIG. 19—Block diagram electronic equipment for the fatigue machine and
crack monitor.
Construction then proceeded with the 12,000-lb dynamic (CA) load

machine, R3, shown in Fig. 1. The electronic equipment associated with its
operation is shown in Fig. 19. It can be seen that the machine is capable of
Rayleigh random loading as well as constant amplitude loading. In service, a
panel in an aerostructure will be loaded by a random process, filtered by the
response characteristics of the structure between the panel in question and
the fatigue loading (for example, atmospheric gusts). If the fatigue energy
enters the panel primarily in a single mode or at a single frequency, the
fatigue machine, R3, is capable of providing a good simulation of this load
pattern to the specimen.
In Rayleigh random loading the machine was found to be capable of
3000-lb-rms dynamic loading.
The mean load, applied by a motorized mean stress spring mounted at the
opposite end of the lever from the specimen (Fig. 1), permits the machine to
apply Copyright
up to 20,000-lb
by ASTMsteady
Int'lload
(all in either
rights tensionMon
reserved); or compression.
Dec 7 14:40:45 EST 2015
developing R3, anby optimization study was carried out based on the
principles of operation of the machine Rl [18], While space does not permit
a description of this study, it resulted in the positioning of the shaker in-
board of the mean stress mechanism, closer to the fulcrum, compared with the
opposite arrangement which exists with Rl.
Machine R3 operates at resonance with the single lever. When excited by
filtered white noise, one obtains a load history with the (random) amplitudes
following a Rayleigh distribution. Figure 20 is a plot of the relative frequency
of peaks taken from a sample trace of the load. This reading of load comes
from a laterally stabilized steel plate beneath the specimen, joining the bot-
tom grips for the specimen to the machine foundation. This steel plate is
fitted with a complete bridge of strain gages whose output is read from an
oscillograph (Fig. 19).
FIG. 20—Relative frequency of peaks Rayleigh random amplitude sample trace.
In order to obtain a statistically sound trace for checking the actual

distribution of stress peaks, the root mean square level was kept at a con-
stant value while first a fast trace (delineating every cycle) was taken. Then
slower and slower oscillograph speeds were used which permitted only manual
assessment of the larger load excursions. However, with the fast trace ana-
lyzed by computer, and supplying the higher probability information, such a
test procedure permits the experimental study of the occurrence of peaks of
progressively lower probability assuming the process remains stationary.
The clipping ratio, which yields the limiting value for load excursions with
this apparatus, compared with the rms level, was found to be 4.32 (Fig. 20).
A companion study of the irregularity factor—the ratio of half the zero
crossings to the peaks—was undertaken using the fast oscillograph traces.
Machine R3 appears to be more lightly damped than Rl, the ratio being
greater than 0.99 with R3 compared with 0.96 with Rl.
Since the possibility
Copyright by existed
ASTM of Int'l
buckling,(all with rights
the occurrence
reserved); of com-
Mon Dec
pressive loads under random
Downloaded/printed by amplitude loading conditions, the critical buck-
ling stress was calculated for the specimens. The value is given by the fol-
lowing equation (Ref 24):
for thin sheets. For the square-panel specimens used, the free length is equal
to the width W. B is a constant which varies with constraint on the sheet.
For clamped ends and free edges B = 3.62. For clamped ends and simply
supported edges, B = 6.4.
At the stress levels of 7.0 ksi mean and 2.5 ksi rms random load com-
pressive stresses of the order of 3.0 to 4.25 ksi occur with an assumed clipping
ratio of 4.0 to 4.5.
For this range of stress levels, the 0.080-in.-thick specimens are unstable
at the 10.0 and 12.5-in. widths, and marginal at 7.5 in. for free edges. The
other marginal specimens were 12.5 in. wide by 0.125 in. thick and 0.160
in. thick. Accordingly we required vertical clamps on either free edge of
the specimen. The simple support edge restraint decided upon consisted of
two vertical pieces of 1.25 by 1.25 by 0.125-in. extruded aluminum angles
clamping the specimen on each side. They extended vertically to within 0.25
in. of the grips at each end, and clamped about 0.5 in. on each side of the
specimen. These were used with all 7.5 in./0.080 in., 10.0 in./0.080 in., 12.5
in./0.080 in. and 12.5 in./0.125-in. specimens, under random loading.
These stiffeners were satisfactory in all cases except the 12.5 in./0.080 in.
specimen, which seemed to buckle several times. For this width and thick-
ness, even the simply supported edge condition is marginal in capability.
In order to calibrate the load applied to the specimen, a set of 0.160-in.-
thick unnotched specimens, one for each width, was set aside and fitted with
a 4-gage bridge at each end of the transverse axis of the specimen at the
center of the specimen. In this way we were able to obtain readings of bend-
ing from differences in strain on either side, and average loading by sum-
mation for the load calibration.
Each instrumented specimen was calibrated statically in a tension test
machine before inserting in the fatigue machine. The bottom grip plates
vary in vertical position depending on the specimen used.
From a photoelastic study of the distribution of strains (Figs. ISe and /)
we saw qualitatively that there was a small amount of bending in the plane of
the sheet specimen when gripped in the machine. The strain gage bridges
detected maximum bending stresses of the following magnitudes:
Specimen Width, in. Bending Stress as Per Cent of Axial Stress
5 2.95
7.5. 1.99
10 2.13
12.5 2.65
This effect was not great enough to cause a noticeable nonrandom varia-
tion in crack growth from one side of the notch to the other.
For the axial loads our procedure was to balance before each test and to
always use the exact cross-sectional dimension of the particular specimen in-
volvedC in
o pcalculating
y r i g h t test
b y loads
A S for
TM theIrequired
n t ' l ( astresses.
ll rights reserved); Mon
During
D o wthe
n l time
o a dperiod
e d / p involved
r i n t e din bthe
y actual test program, October, 1964,
University of Washington (University of Washingto
TABLE 10—Effect of frequency on crack propagation rate of 7079-T6 material

loading 9800 ± 5500 psi (CA-NLS).
Kilocycles Lz Total L, Total
to First Kilocycle Length to Ki i Length to K« >
Specimen Visual to Transition, k si\/in.
Failure Static ksi\/in.
Initiation in. Failure, in.
LAZ-1 Machine— 50 cps Sine Wave— 3000 cpm

045/1 /-/5/080/33 13.0 39.1 0.96 9.9 4.00 31.5
045/1 /-/5/080/35 20.0 47.2 0.89 9.6 3.80 28.6
045/1/-/5/080/34 24.1 49.5 0.88 9.5 4.10 33.4
045/1/-/5/080/32 28.0 55.1 0.81 9.2 4.05 32.3
R3—30.0 cps Sine Wave—1800 cpm
045/1 /-/5/080/28 9.5 29.7 1.09 10.7 4.10 33.4
045/1 /-/5/080/29 10.8 31.7 0.99 10.3 4.15 34.3
045/1 /-/5/080/30 12.8 32.2 0.97 10.1 4.10 33.4
045/1/-/5/080/31 13.7 33.4 0.99 10.3 4.10 33.4
Servohydraulic Rig—2 cps Sine Wave—120 cpm
045/1 /-/5/080/40 10.8 30.5 1.21 11.3 4.00 31.5
045/1 /-/5/080/42 12.8 33.4 1.25 11.5 3.70 27.2
045/1 /-/5/080/43 15.5 36.7 1.39 12.2 3.90 30.0
045/1 /-/5/080/41 16.8 40.1 1.22 11.5 3.90 30.0
Servohydraulic Rig—10 cps Sine Wave—600 cpm

045/2/7/5/080/46 1.5 27.4 1.27 11.6 4.00 31.5
045/2/1/5/080/45 12.5 28.7 1.28 11.8 3.65 26.8
045/2/4/5/080/44 11. 30.9 0.86 9.5 3.85 29.2
045/2/6/5/080/47 3.5 33.6 0.98 10.1 3.85 29.2
Load-Ring Hydraulic Jack—2 cps Triangular Wave—120 cpm
045/1/-/5/080/36 14.5 31.5 1.51 12.9 3.85 29.2
045/1 /-/5/080/39 15.9 35.3 1.46 12.7 3.80 28.5
045/1 /-/5/080/37 17.0 44.4 1.01 10.3 3.75 27.2
045/1 /-/5 080/38 15.9 48.1 1.02 10.4 4.10 33.4
to April, 1966, the load dynamometer was calibrated ten times. The calibra-
tion was found to be quite repeatable and randomly varied by about ll/z per
cent from the average values of pounds per oscillograph division.
Effect of Frequency, Machine, and Waveform on Crack Propagation

A series of twenty CA-NLS tests was carried out to determine the effect
of frequency, machine, and waveform on crack propagation. The specimens
were all 5-in. wide. One strain gaged specimen was used to calibrate all the
test fixtures used, including R3. The test load level was kept constant to create
an initial 9800 psi net mean stress and 5500 psi net alternating stress. Crack
initiation and 0.1-in. increments on each side of the notch were recorded, as
was final
C ofailure.
p y r i g h t b y A S T M I n t ' l ( a l l r i g h t s r e s e
TheDfour
o wtest
n lfixtures
o a d were:
e d / p r i n t e d b y
U n i v e r s i t y o f W a s h i n g t o n ( U n i v e r s i t y o f
1. Machine R3 (30 cps)

2. LAZ-1 machine (50 cps)
3. Hydraulic jack with a triangular wave pattern (2 cps)
4. Servohydraulic actuator with a sine wave pattern (2 and 10 cps)
From Table 10 it can be seen that there is no ordered frequency effect in
the frequency range studied. The endurances from cyclic loads involving a 10-
cps sine wave and a 2-cps sine wave, both tested on the same machine, were
less than and greater than the 30-cps sine wave test results, respectively. The
2-cps triangular waveform results yielded slightly greater endurances than the
other three. This might possibly be explained by the fact that a triangular
wave has a lower rms value than does the sine wave (0.577 <r a it as compared
to 0.707 o-ait) for a given peak load. The endurances using a 50-cps sine wave
were greater than all the others. This is felt to be due to a significantly dif-
ferent "machine factor" for the LAZ-1 machine.
APPENDIX III
Servo-Controlled Eddy-Current Crack Follower

While electron microscopy has made possible the observation of individual
striations on the fatigue crack surface and their local spacing, there still
remains a great deal of investigation of the macroscopic and continuum as-
pects of crack growth (for example, rate of propagation) which will be of
particular use to the designer.
From a review of the techniques of crack detection and crack length moni-
toring, we can summarize the different methods, grouped in order of in-
creasing desirability, into the following sections:
(a) Methods involving contact with the specimen (often with individual
specimen preparation) and possibly affecting the crack process it detects: (1)
ink stain; (2) coatings (thermal, visual, etc.); and (3) bulk measurements (for
example, high current resistance measurement by milliohmeter).
(b) Methods involving specimen contact but probably not affecting the
fatigue process: (1) voltage drops across a given section; (2) low current
(bridge) resistance measurement; (3) compliance gages, strain gages; and (4)
ultrasonics.
(c) Methods not touching the specimen, but not readily amenable to con-
tinuous crack monitoring: (1) visual observances; (2) X-rays; and (3) certain
fringe (interference) methods.
(d) Methods not touching the specimen and amenable to continuous and
automatic crack length recording: (1) optical servomechanisms; (2) eddy-cur-
rent probes (with lift-off compensation); and (3) heat detection.
We did not want to prepare individual specimens nor affect the fatigue
process in any way. We also wanted to make the monitor automatic, since,
from a practical design point of view, it is the low rates of growth which are
of interest in the assessment of the life of reasonably well-designed struc-
tures in practice.
We, therefore, investigated the first two methods in the last category
thoroughly. The
Copyright opticalInt'l
by ASTM extensometer was found
(all rights reserved); to be
Mon Dec usuallyEST
7 14:40:45 an 2015
expensive
method, and the less expensive
Downloaded/printed by models were suspected of not having sufficient
contrast discrimination to keep visual track of the tip of a fatigue crack on a

surface of uniform reflectivity.
This latter point is catered to very well by eddy-current devices, which see
the crack as a large perturbation contrasting strongly with the background
of uncracked sheet. Also they have reasonable power [25] to penetrate the
top surface layers, to some degree, to obtain an accurate and stable signal.
Since the eddy-current device supplied us with an electrical signal of the
presence of a crack, it remained to mount the probe at the end of a servo-
controlled linear actuator and position it at the side of the center notch
FIG. 21—Closeup of Servo-controlled eddy-current crack tip follower.
(Fig. 21). By enclosing the probe in a nylon sheath, it was possible to keep
the probe surface a fixed distance (0.010 in.) from the sheet surface, pre-
venting any damage to the probe when the specimen fractures into two pieces.
As soon as the crack appears, the eddy-current off-null signal is used to
drive the linear actuator horizontally to the left, in Fig. 21. The probe
is then moved physically to the left. When the probe reaches the crack tip
the off-null signal drops to zero, which stops the servoactuator. In this way
the actuator system, having high response (88 in./min velocity), is "locked
on" to the tip of the crack. The crack signal could also be used to actuate the
elevator for the probe (Fig. 21) to "lock on" in the vertical dimension.
A linear potentiometer is coupled physically to the actuator and presents
a continuous
Copyright signalby of actuator
ASTM position,
Int'l for
(all the ll-in.-wide
rights strip chartMon
reserved); re- Dec
corder.Downloaded/printed
By selecting different
by resistors in this circuit, we can make use of
FATIGUE CRACK PROPAGAGATIONATION
nearly the full width of the chart paper regardless of actual specimen width.
The strip chart speed was kept at a constant 4 in./hr in all tests.
While we intend eventually to use two probe systems, one on each side of
the notch, we were forced to take a visual reading of the unmonitored side
in order to make the discrete stepwise changes in load level (Fig. 7) as-
sociated with the load shedding experiments.
It is unfortunate that the automatic presentation of the crack growth trace
(for example, Figs. 6, 8, and 9) has the ordinate and abscissa parameters op-
posite to the traditional presentation in the literature. It was considered
pointless, though, to convert the data, since other workers will no doubt find
this technique more convenient than the earlier methods, and we will see
more plots of this type in the future.
The principles of detection by eddy currents are discussed in Ref 25.
The block diagram of the servosystem is included in Fig. 19. We used a
probe of % in. diameter for all the work reported in this paper, and the nylon
sheath included a 0.030-in. rim around the probe circumference. Smaller
probes (down to at least Ys in. diameter) are also available.
The size of the probe is not really an indication of the sensitivity of the
system, since by balancing for a signal associated with, say, 70 per cent pene-
tration of the probe into uncracked material one can detect increments in
crack growth of less than 0.010 in., by the resulting shift in the balance point
on the probe face along the sheet surface. Further discussions of this
balance in the servo system are presented in Ref 5.
For our tests this balance point was at about 90 per cent penetration by the
probe diameter into uncracked sheet. As a result, when the crack had almost
reached the specimen edge so that the probe received a signal from the edge
as well, it would quickly move to the edge when the signal generated by the
edge became stronger. This behavior of the probe is evident in Fig. 9.
By controlled tests (for example, driving the probe actuator by a worm
gear assembly) we found that continuity and linearity were very good.
This system will find application in other areas of practical engineering
besides fatigue crack monitoring. Such a servo-controlled device, sensitive
to edges as well as to cracks, presents an alternative guide for template
machining and milling.
One can also envisage a great improvement in automatic welding by hav-
ing such a probe mounted ahead of the electrode, and changing the trans-
verse position of the weld assembly to follow the irregularties of the two
edges to be welded, as the welding assembly proceeds automatically along
the general direction of the edges.
Another application, closer to the interest of fracture toughness test en-
gineers, is in the evaluation of the critical plane strain fracture toughness
parameter Klc. By increasing the sensitivity of the balance in the servo
system, and using frequencies appropriate for deep penetration, one can de-
tect the "pop-in" event more accurately, presumably, than with any other
system used so far at least for thin materials. When one considers the com-
paratively low sensitivity with a compliance system at low percentages of
crack length, where very little change in compliance occurs in the initial
5 per cent of crack length/width [26], the preparation associated with ac-
curate compliance readings, and their greater vulnerability to temperature,
the servo-controlled eddy-current probe has definite possibilities.
TheCopyright
servoprobe can also
by ASTM Int'l be
(all considered for Mon
rights reserved); fracture
Dec toughness tests 2015
7 14:40:45 EST to ob-
tain K c, provided the response
Downloaded/printed by of the actuator is made as high as possible.
It is also interesting to note that the system was tested with the probe tip
immersed in water, and that its performance was not hampered in any way.
References
[1] W. Weibull, "Further Investigations into Fatigue Crack Propagation in
Sheet Specimens," Current Aeronautical Fatigue Problems, Pergamon Press,
New York, April, 1963.
[2] G. R. Irwin, "Analysis of Stresses and Strains Near the End of a Crack
Traversing a Plate," Journal of Applied Mechanics, Vol 24, Transactions,
Am. Society Mechanical Engrs., Vol 79, September, 1957, pp. 361-364.
[3] R. N. Katz and K. H. Abbott, 'The Use of the Critical Thickness Concept in
Design," Technical Report AMRA TR 64-51, Materials Engineering Div.,
U.S. Army Materials Research Agency, Watertown, Mass., December, 1964.
[4] P. P. Benham and T. G. J. Moag, "Fatigue Crack Propagation and Hardness,"
The Engineer, April 30, 1965, p. 760.
[5] V. M. Markochev and B. A. Drozdovskii, "A Method for Determining the
Speed of Propagation of Cracks and for Maintaining the Specified Stress,"
(UDC 620.178.4), Zavodskaya Laboratoriya, Vol 31, No. 3, March, 1965,
pp. 345-349.
[6] M. Isida, "Stress Concentration in an Eccentrically Cracked Strip Subject to
Tension," Internal Report, Department of Mechanics, Lehigh University,
Bethlehem, Pa., September, 1965.
[7] J. Broek, P. de Rijk, and P. J. Sevenhuysen, 'The Transition of Fatigue
Cracks in Alclad Sheet," NLR-TR-M2100, November, 1962.
[8] A. K. Head, "The Propagation of Fatigue Cracks," Journal of Applied Me-
chanics, Am. Society Mechanical Engrs., Vol 23, September, 1956.
[9] H. W. Liu, "Crack Propagation in Thin Metal Sheet Under Repeated Load-
ing," Transactions, Am. Society Mechanical Engrs., Vol 83, 1961, p. 23.
[10] N. E. Frost and K. Denton, "The Fatigue Crack Propagation Characteristics
of HS30WP Aluminum Alloy," NEL Report No. 157, August, 1964.
[11] P. C. Paris and F. Erdogan, "A Critical Analysis of Crack Propagation
Laws," ASME Paper 62-WA-234, Am. Society Mechanical Engrs., Novem-
ber, 1962.
[12] J. Schijve, "Fatigue Crack Propagation in Light Alloy Sheet Materials and
Structures," Advances in Aeronautical Sciences, September, 1960, p. 387.
[13] A. J. McEvily and R. C. Boettner, "On Fatigue Crack Propagation in FCC
Metals," Acta Metallurgica, Vol 11, July, 1963.
[14] S. R. Valluri, "A Theory of Metal Fatigue," Acta Metallurgica, Vol 11, 1963,
p. 759.
[15] M. L. Williams, "The Bending Stress Distribution at the Base of a Stationary
Crack," Journal of Applied Mechanics, Vol 28, Transactions, Am. Society
Mechanical Engrs., Series E, Vol 82, 1961.
[76] W. Barrois, Discussion following W. Weibull's paper: "Size Effects on Fatigue
Crack Initiation and Propagation in Aluminum Sheet Specimens," Fatigue
of Aircraft Structures, MacMillan Publishers, New York, 1963, p. 51.
[17] P. J. E. Forsyth, "A Two Stage Process of Fatigue Crack Growth," Proceed-
ings of the Crack Propagation Symposium, Cranfield, England, 1961, p. 76.
[18] S. R. Swanson, "An Investigation of the Fatigue of Aluminum Alloy Due to
Random Loading," UTIA Report 84, February, 1963.
[19] L. R. Jackson, "Some Observations on the Distribution of Stress in the Vi-
cinity of a Crack in the Center of a Plate," DMIC Memorandum 178, Sep-
tember, 1963.
[20] A. J. McEvily, Jr., and T. L. Johnston, "The Role of Cross-Slip in Brittle
Fracture and Fatigue," paper presented at the International Conference on
Fracture, Sendai, Japan, September, 1965 (Fig. 17).
[21] R.Copyright
W. Hertzberg,by "Application
ASTM ofInt'l
Electron
(allFractography
rights and Fracture Me-
reserved); Mon Dec
chanics to Fatigue Crack Propagation in High Strength Aluminum Alloys,"
Report, Lehigh University Institute of Research, May, 1965.
University of Washington (University of Washington) pursuant t
FATIGUE CRACK PROAGATIONPAGATION
[22] L. R. Jackson, "Continued Observations on the Distribution of Stress in the

Vicinity of a Crack in the Center of a Plate," DMIC Memorandum 190,
April, 1964.
[23] J. G. Kaufman and M. Holt, "Fracture Characteristics of Aluminum Alloys,"
Technical Paper No. 18, Alcoa Research Laboratories, 1965.
[24] D. J. Peery, Aircraft Structures, McGraw-Hill Book Co. Inc., New York,
1950, Eq 14.47, p. 370.
[25] R. C. Grubinskas, "Development of Eddy Current Inspection Equipment,"
Technical Report AMRA TR 63-24, U.S. Army Materials Research Agency,
Watertown, Mass., November, 1963.
[26] J. E. Srawley and W. F. Brown, "Fracture Toughness Testing Methods,"
Fracture Toughness Testing and Its Applications, ASTM STP 381, Am.
Soc. Testing Mats., 1965, p. 133.
DISCUSSION
J. G. Kaufman1 (written discussion)—The authors are unquestionably

aware of the great number of data showing that the fatigue strengths of
Alclad aluminum alloy sheet are appreciably lower than those of bare
sheet of the corresponding alloys and tempers.2 Would they care to
speculate on the influence of the Alclad coating on the fatigue crack
growth rate in aluminum alloy sheet or are they aware of other data
showing this influence?
5. R. Swanson, F. Cicci, and W. Hoppe (authors)—Mr. Kaufman has
correctly emphasized the limitations of the present work to the material
used; namely Alclad 7079-T6 aluminum alloy sheet. We have tried to
determine the influence of the 4 to 5 per cent thickness coating by com-
paring our NLS data with the relevant literature. However, since 7079-T6
is a rather recent material, and since the present study showed a signifi-
cant effect due to specimen free length, we have not been able to make
a sound comparison. From the crack initiation study, there seemed to
be little effect of the cladding, since crack nuclei were usually created
at mid-thickness first. We would like to suggest, however, that the KLS
type test is much more suited to the study of second-order effects than is
the NLS type test.
/. Schijve3 (written discussion)—The authors have made an interesting
1
Assistant chief, Mechanical Testing Div., Alcoa Research Laboratories, New
Kensington, Pa.
2
G. W. Stickley and J. O. Lyst, "Effects of Several Coatings on Fatigue
Strengths of Some Wrought Aluminum Alloys," Journal of Materials, Vol. 1, No.
Copyright
1, March, 1966. by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
3 Downloaded/printed
Director of technical by
services, National Aerospace Laboratory, NLR, Am-
sterdam, Holland.of Washington (University of Washington) pursuant to License Agreement. N
University
DICUSSION ON CLAD 7079-T6 ALUMINUM ALLOY SHEET 361
approach to crack propagation and the transition phenomenon by per-

forming constant K value tests. With respect to the geometry of the crack
front, crack propagation tests under random and programmed load
sequences on 2024-T3 and 7075-T6 sheet specimens carried out at
our laboratory4 revealed straight crack fronts in the former alloy and
evidence of tongue shaped crack fronts in the latter alloy (see the fracture
surfaces in Fig. 22). The latter observation was due to "tunneling." An-
other observation was that the major contribution to the crack growth
was made at the higher peak loads. Have the authors observed any evi-
dence of tunneling in the 7079 alloy?
Messrs. Swanson, Cicci, and Hoppe—Dr. Schijve has indicated evi-
(a) 2024-T3 alclad, spec. Cl(a) program load, tested outdoors.

(b) 7075-T6 clad, spec. C76(d) random load + GTAC, tested indoors.
(c) 7075-T6 clad, spec. C73(b) program load, tested outdoors.
FIG. 22—Fracture surfaces (X26).
dence of tongue-shaped crack fronts in 7075-T6 aluminum alloy. Since

7079-T6 is a sister alloy he suggests we might have evidence of a similar
phenomenon. While we can say little about tunneling in constant ampli-
tude tests (except that, from partial tests, just past initiation, the crack
front was straight) the Rayleigh tests do provide us with a definite answer.
Figure 10 of our paper shows a composite illustration in which a
photograph of the crack surface is given for an RRA-NLS test. It is
quite clear that for all stages of fatigue, the crack front is only very slightly
curved across the thickness, with the mid-thickness front about 10 per
cent thickness ahead of the front at the sheet surfaces. This was a com-
mon feature of the RRA-NLS test specimens. There is, of course, a
* J.Copyright
Schijve by
and P. de
ASTM Int'lRijk, 'The
(all rights CrackMon
reserved); Propagation in Two
Dec 7 14:40:45 Aluminum Alloys
EST 2015
in an Indoor and an Outdoor Environment Under Random and Programmed
Load Sequences," NLR Report M.2156, Amsterdam, November, 1965.
FATIGUE CRACK PROPAGAagationTION
vertical tunneling due to the V shaped profile which develops in shear

mode cracking, but this is a different matter.
Regarding crack growth at peak loads, Fig. 10 also shows typical be-
havior for RRA tests. The crack length can be seen to jump ahead under
the action of bursts of high random loading, and then to remain stationary
(with time) for a period before resuming growth. This is an analogous
behavior in random loading, to what Dr. Schijve has found in constant
amplitude loading.
D. P. Wilhem1
Investigation of Cyclic Crack Growth

Transitional Behavior
REFERENCE: D. P. Wilhem, "Investigation of Cyclic Crack Growth

Transitional Behavior," Fatigue Crack Propagation, ASTM 415, Am. Soc.
Testing Mats., 1967, p. 363.
ABSTRACT: Observations of tensile to shear transition behavior of fa-
tigue crack fronts have been recorded numerous times during tension-
tension cycling of sheet materials. Based on recent crack growth data,
correlation of shear lip development and associated crack tip stress in-
tensities has been obtained for a variety of aluminum, steel, and titanium
alloys. Assumption of a crack growth transition zone region allows for
direct application to deviations in crack growth rate versus stress-intensity
relationships at low crack tip intensities. A stress-controlled specimen of
7075-T6 aluminum sheet was fatigue cycled from tensile to shear and
back to tensile within the transition zone; crack front behavior was
microscopically observed. Macroscopic and microscopic correlations were
established with discontinuities in stress intensity versus crack growth
semilog relationships. The beginning of shear lip development in 2024-T3
aluminum appears to be independent of thickness and dependent only on
the stress-intensity factor at transition.
KEY WORDS: fatigue (materials), crack popagation, aluminum alloys,

steels, titanium alloys, crack growth rate, fracture, transitional behavior,
microstructure, stress intensity
Flat to shear-through-the-thickness fracture surfaces are quite com-

mon for the crack front of tension-tension cycled sheet material. These
topographic features have been observed numerous times during post-
fracture examination of the specimen and reported by several investi-
gators [1—6].2 Particular research has been undertaken to explain the
operative mechanisms of the crack front behavior [7-9]. This research
led to the development of several models to describe the topographic
transition [9-11] from a micromechanics approach. In all cases, however,
particular research has been directed to the end or tunneled portion of
1
Senior engineer, Research and Technology, Northrop Corp., Norair Division,
Hawthorne, Calif.
2
paper.
363
FATIGUE CRACK PROagationPAGATION
FIG. 1—Crack growth rate versus crack tip stress intensity factor range, 2024-
T3 clad.
the tensile mode. Correlation of this particular point was then attempted
with crack growth rate, stress intensity, and associated geometric param-
eters. Copyright by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
For practical engineering purposes, the beginning of shear lip de-
velopment appears to be the
University of Washington more important
(University criterion
of Washington) pursuant due to the
to License tunneled
Agreement. No further rep
WILHEM ON CYCLIC CRACK GROWTH TRANSITIONAL BEHAVIOR 365
FIG. 2—Crack growth rate versus crack tip stress intensity factor range, 7075-T6.
portion of the tensile mode being masked by the material thickness. If

a change in cracking rate or crack tip stress intensity could be correlated
with the topographic transition behavior, a general basis of correlation
would exist. Investigation of this particular behavior and the consequent
results are reported in this paper.
Crack Growth Rate—Stress Intensity Relationships

Using the analysis of Paris [12] and his relationship for crack tip stress
intensity factor
Copyright by range, the(alldata
ASTM Int'l from
rights RefsMon
reserved); Decand
2, 4, 13 through
7 14:40:45 18 were
EST 2015
analyzed. These data by(Figs. 1 and 2) represented cracking rate informa-
Downloaded/printed
FATIGUE CRACKagation PROPAGATION
FIG. 3—Crack growth versus crack tip stress intensity factor range slopes, steels
[20,21].
tion for sheet aluminum alloys 2024-T3 and 7075-T6 in thicknesses of

0.032 to 0.130 in. and widths of 3 to 30 in. All data were obtained from
centrally cracked material and encompass six orders of magnitude in
crack growth rate. This analysis uses total crack length, /, and, therefore,
the crack tip stress intensity factor range (A&) relationship is
where o-max is the maximum stress and <r m m the minimum stress per cycle
and
is the finite width correction term for centrally cracked sheet proposed by
Dixon [79], and W is the specimen width.
The important feature of Figs. 1 and 2 is the progressive change in
slope that occurs at a stress intensity of 10 ksi \/in- f°r both alloys.
Two dashed lines have been employed as a guide in establishing the
FIG. 4—Crack growth versus crack tip stress intensity factor range slopes, ti-
tanium alloys [20,21].
approximate midpoint of the knee in the data at that point. Therefore,

a zone of behavior (transition zone) has been established for these two
materials which encompasses the range of 5 to 15 ksi \/in.
As additional crack growth data from other materials became available
[20, 21], similar semilog trends in crack growth were detected, and cor-
responding
Copyrighttransition
by zones
ASTMwere established.
Int'l (all Therights
slopes of the fatigueMon
reserved); D
data Downloaded/printed
from these referencesby for several steel, titanium, and aluminum
FIG. 5—Crack growth versus crack tip stress intensity factor range slopes,
aluminum alloys [2,4,13-18,20,21].
alloys are shown in Figs. 3, 4, and 5. The intersecting straight line rela-
tionships were established in the same manner as for the 2024-T3 and
7075-T6 aluminum alloys.
Relationship of Stress Intensity Zone to Fracture Surface Transition

Topography
A plausible explanation for the knee in the semilog crack growth versus
stress intensity factor relationship (Fig. 6) may be found in a detailed
examination of fracture surface topography. A centrally slotted sheet of
7075-T6 aluminum alloy, 12 in. wide by % in. thick, was tension-tension
fatigue cycled under controlled stress conditions to obtain programmed
crackCopyright
tip stress
by intensity.
ASTM Int'lThe
(all fracture surface
rights reserved); of Dec
Mon this specimen
7 14:40:45 is shown
EST 2015
in Fig. 7. Stress intensity
Downloaded/printed by was controlled to correlate the transition zone
(knee) behavior of the semilog plots with plausible crack front reaction.
An initial growing stress of 20 ksi combined with a ^-in. starter slot
produced an initial stress intensity factor range of 7 ksi \/in.; this is
within the lower range of the proposed transition zone for this material.
At a total crack length of 0.80 in. (Afc = 19 ksi \/in.), a complete traverse
of the stress intensity transition knee was accomplished. The stress was
FIG. 6—Crack growth versus crack tip stress intensity factor range AM 350
(20 per cent CRT).
then reduced at that crack length to 7 ksi which corresponded to a reduc-

tion in stress intensity to 6 ksi \/in., or a value near the low end of the
transition zone. Fatigue cracking progressed at the 7 ksi stress level
through the transition zone and beyond until complete specimen fracture
occurred.
Topographic features and the behavioral trends of the crack front are
shown in Fig. 7. Zones A and B delineate the two complete traverses
Copyright by ASTM Int'l (all rights reserved
of theDownloaded/printed
proposed transition zone.by
The particular topographic features
University of Washington (University of Washing
are easily discernible; tensile or 90-deg crack front progression has

taken place from the starter notch. At a total crack length of % in.
(Afc = 14 ksi \/in.) shear lip development was initiated. At the point
of stress reduction (/ = 0.80 in.) full shear had taken place, double
shear on the left side fracture surface and single shear on the right. The
reduction in stress intensity at that point rotated the crack front almost
immediately into tensile mode progression, producing a noticeable step
on the specimen face normal to the loading direction. As the fatigue
cracking progressed at this reduced stress, shear lip development oc-
FIG. 7—Fracture topography—stress intensity controlled specimen (X2.4).
curred at a crack length of 1.92 in. (Ak = 10 ksi A/in.), the midpoint
of the projected transition zone knee for this material.
The topographic features of this specimen were thus correlated with
the transition zone behavior outlined in crack growth-stress intensity
semilog representations.
Micro-Observations of Stress Intensity Controlled Specimen

Electron fractographs of the stress intensity controlled specimen are
shown in Figs. 8 through 11. The tensile (90-deg) and shear (45-deg)
modes of Zone A are shown in Figs. 8 and 9; Figs. 10 and 11 show these
modes for Zone B. The operating stress level was 20 ksi within Zone A
and 7Copyright
ksi within
by Zone
ASTMB.Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
Relationship of the byindividual fractographs, their location on the
Downloaded/printed
FIG. 8—Electron fractographs of stress intensity controlled specimen—Zone

A tensile mode—nominal stress 20 ksi.
fatigued surface, and crack front direction is as indicated. The tensile

crack front progression of Zone A (Figs. 8a through d) shows the typical
planar transgranular fatigue striations of this region. The spacing between
striations in Figs. 8a and b is approximately 0.3 /*,. As the crack length
increased, thebymicrotopography
Copyright ASTM Int'l (all rights in the planar
reserved); Mon Decmode showed
7 14:40:45 striation
EST 2015
spacing of approximately
Downloaded/printed by 0.7 /x (Fig. 8c). Fractograph 8d shows the
area 0.5 in. beyond
University the area
of Washington illustrated
(University in Fig. pursuant
of Washington) 8c. Figure Sd also
to License gives No further r
Agreement.
agationFATIGUE CRACK PROPAGATION
FIG. 9—Electron fractograph of stress intensity controlled specimen—Zone

A shear mode—nominal stress 20 ksi.
an indication of the rapid increase in the crack progression at the apex

of the 90-deg mode; the average striation spacing here is approximately
2.0 n, which is three times the size of the spacing in Fig. 8c. This large
change is striation spacing within the tensile mode apex region also oc-
curred in Zone B and appears to be typical behavior in this region. The
shear area of Zone A (Fig. 9) indicates transgranular cleavage mixed
with some representation of fatigue striation.
Zone B fractographs
Copyright (Figs.
by ASTM Int'l (all 10 andreserved);
rights 11), representative
Mon Dec 7 of14:40:45
the lower
EST 2015
stressed region (7 ksi),by show planar and shear crack front progression
Downloaded/printed
FIG. 10—Electron fractographs of stress intensity controlled specimen—Zone

B tensile mode—nominal stress 7 ksi.
similar to Zone A. Figures 10a through d correspond to similar behavior

within Zone A. The fatigue striation spacing is of the same magnitude
(approximately 0.3 /*, Fig. 105) as Zone A at identical crack tip stress
intensities, thus confirming the findings of Ref 9. Figure Wd indicates
also the rapid change in planar fatigue striation spacing at the apex of the
tensileDownloaded/printed
mode. These spacings
by
are approximately 2.0 /x, wide, which is
FIG. 11—Electron fractograph of stress intensity controlled specimen—Zone

B shear mode—nominal stress 7 ksi.
identical to those of Zone A. The texture differences are primarily due

to the increased plastic flow associated with Zone A due to the high
stress level fatiguing of that zone as opposed to low stress cracking of
Zone B. Beyond the tensile apex, a typical mixture of fatigue and cleavage
fracture topography is evident within the shear mode (Fig. 11).
Examination of Figs. 8 through 11 supplies additional insight into the
transition zone (knee) behavior of semilog crack growth, stress intensity
data. The rapid change in fatigue striation spacing near the apex of
tensile mode progression
Downloaded/printed by furnishes additional substantiation to the in-
TABLE 1—Transition crack lengths and associated stress

intensities for various materials.
Crack Suggested
Ak at Begin- Transition
Length at ning of A*
Material am, ksi aa , ksi Beginning of Shear Lip, Range, ksi V in.,
45 Deg0Shear ksi \/in. (from Figs. 3,
Lip, in. 4, and 5)
AM 350 (20% CRT) .(40 20 0.52 291 19-29

\40 10 0.83 19/
AM 350 (double aged) . . . 40 10 0.44 17 7-17
PH 14-8 Mo /40 5 1.35 121 8-18

\40 10 0.43 13}
PH 15-7 Mo 40 5 0.91 10 9-19
Rene 41 40 5 0.80 10 11-21
AISI301 40 10 0.32 11 6-16
/40 5 1.10 10
Inconel 718
\40 10 0.26 10} 9-19
Ti-4Al-3Mo-lV (25 5 1.10 101 3-13

125 10 0.41 13/
T1-6AMV 25 10 0.85 19 9-19
20 1.20 221
Ti-8Al-lMo-lV
$ 30 1.02 30/
2-12
5 1.42 121
2024-T3
h5 5 1.35 12/
5-15
15 2 1.05
RR 58 Al I115 5 0.32 % 2-12
2 1.25
%
2024-T81 15
IV15 3.5 0.65 2-12
0
Data on transition crack lengths obtained through courtesy of NASA-Lang-
ley personnel with exception of 2024-T3 [13].
crease in crack growth rate indicated by the empirical data. The increase
in crack growth progression indicated graphically is thus correlated with
microcracking behavior. Furthermore, the feasibility of crack tip stress
intensity control is given further credence on a microscopic as well as
macroscopic scale. This confirms the recent findings of Hertzberg [9].
Constant Transition Stress Intensity Trends for Various Materials

Figures 4, 5, and 6 show that transitional behavior zones do exist for
materials other than aluminum alloys 2024-T3 and 7075-T6. To deter-
mine if the zones proposed in this paper possess direct correlation with
crack Copyright
tip stressby ASTM Int'l (all rights
intensity reserved);
factor rangeMon Dec 7 14:40:45
at the EST 2015 of shear lip devel-
beginning
FATIGUE CRACK PROPAGATIOaaGATIONN
FIG. 12—Crack growth rate versus crack tip stress intensity factor range,
2024-T3 clad [13].
opment, measured crack lengths were required. NASA3 supplied these

data, which were taken from fatigue fractured surfaces of various alloys.
The data were then compared with the transition zones proposed by the
crack growth data from Refs 20 and 21, which were used in establishing
the transition knees. Table 1 shows the calculated values of crack tip stress
intensity factor range based on crack length at the beginning of shear
lip development, as well as proposed transition zone (knee) ranges for
the materials investigated.
Copyright by ASTM Int'l (all The spread in
rights reserved); Montransition stress
Dec 7 14:40:45 intensity factor
EST 2015
1
NASA, Langley
University Field, private
of Washington communication,
(University letter dated
of Washington) pursuant Oct.Agreement.
to License 14, 1965.No further reproduc
FIG. 13—Crack growth versus crack tip stress intensity factor range for two
thicknesses, Ti-8Al-lMo-lV duplex annealed [21].
range corresponds to ±5 ksi -y/in. from the point of intersection of the

two straight lines; this method is similar to that employed for the two
aluminum alloys. It is evident that the calculated values of stress intensity
are contained within the transition zones suggested by the crack growth-
stress intensity data, except for the Ti-8Al-lMo-lV alloy, which indicates
slightly higher transition stress intensities. These data supply additional
verification of the interrelation between: (1) change in fracture surface
topography
Copyrightduring
by ASTMshear lip rights
Int'l (all development andDec
reserved); Mon (2) 7deviation in 2015
14:40:45 EST crack
growth-stress intensity semilog relationships.
FATIGUE CRACK PROPAagationGATION
Geometric Differences and Relationship to Transition Shear Lip De-

velopment
References 2, 4, and 13 through 18 were used to establish the transi-
tion behavior of 2024-T3 aluminum (Fig. 1). The relatively large differ-
ences in specimen thickness (0.032 to 0.130 in.) for these data indicate
that thickness may not be a determining factor in shear lip development
for 2024-T3 aluminum sheet. Tension-tension fatigue crack-propagation
data were obtained from five sheet specimens of 2024-T3 clad aluminum
[72] with average yield and ultimate strengths of 48.7 and 70 ksi, re-
spectively. Thickness ranges for these specimens were 0.038 to 0.250 in.
These crack growth data are shown in Fig. 12 and are also incorporated as
part of Fig. 1. The data obtained at a low cracking stress level (10 ksi)
established the lower portion of the curve, and the fracture surfaces con-
tained the typical tensile-to-shear transition expected from fatigue within
this region. The specimens tested at a higher stress (30 ksi) showed typical
shear mode cracking throughout the fatigue life; therefore, the initial
stress intensity from the starter notch placed the crack tip stress intensity
values above the transition zone, and no transition in surface topography
should develop. It is of interest to note that, for a twofold variation in
thickness, the stress intensity at shear lip development was identical (12
ksi "V/ in.). Those specimens tested at a higher stress level were used to
establish the remainder of the crack growth-stress intensity curve.
Additional data [20] for two thicknesses of titanium 8Al-lMo-lV
(duplex annealed) sheet show trends similar to the aluminum alloy. Figure
13 shows these data for thicknesses of 0.050 and 0.250 in.
The limited amount of available crack growth data for different thick-
nesses appears to indicate that shear lip transition zone behavior is not
dependent on material thickness. However, before any definite conclu-
sions can be drawn, additional research must be accomplished on various
materials and over a wider range of specimen thicknesses. The resulting
crack front behavior must be observed within the transition zone itself.

Analysis of the transition zone in the semilog crack growth versus
stress-intensity data supplies further insight into the behavior of the
fracture surface of tension-tension fatigued sheet. The crack length at the
start of shear lip development has been shown to be the best possible
correlating parameter for engineering usage. Topographic features of
tensile to shear transition zone behavior have been correlated with
graphical deviations in the stress-intensity, crack growth data for alumi-
num, steel, and titanium alloys. A sheet specimen of 7075-T6 aluminum
was fatigued under controlled crack tip stress intensity conditions within
the knee of the
Copyright by proposed
ASTM Int'l transition zone. Macroscopic
(all rights reserved); and microscopic
Mon Dec 7 14:40:45 EST 2015
Downloaded/printed
observations by
of the fracture surface indicate direct relationships between:
(1) stress intensity transition zone (tensile to shear), (2) increase in crack
growth rate (fatigue striation spacing), and (3) corresponding deviation
from linearity in crack growth versus stress intensity semilog relation-
ships. Limited crack growth data for aluminum alloy 2024-T3 indicate
that transition zone stress intensity analysis, based on crack length at the
beginning of shear lip development, may be independent of specimen
thickness. However, additional transition data on more brittle materials
must be obtained for additional confirmation.
Acknowledgment
The author wishes to express his thanks to H. F. Hardrath and C. M.
Hudson of NASA Langley Research Center for their cooperation in
supplying data on shear lip crack lengths for this study and for stimulating
discussions of the problem. Electron fractographic examination was per-
formed by R. Herfert, Northrop Norair Materials Research Group, who
provided support in their interpretation.
References
[1] N. E. Frost and D. S. Dugdale, "The Propagation of Fatigue Cracks in
Sheet Specimens," Journal of the Mechanics and Physics of Solids, Vol 6,
1958.
[2] A. J. McEvily, Jr., and W. nig, 'The Rate of Fatigue-Crack Propagation in
Two Aluminum Alloys," NACA-TN-4394, September, 1958.
[3] H. A. Lipsitt, F. W. Forbes, and R. D. Baird, "Crack Propagation in Cold-
Rolled Aluminum Sheet," Proceedings, Am. Soc. Testing Mats., Vol 59, 1959.
[4] D. E. Martin and G. M. Sinclair, "Crack Propagation Under Repeated Load-
ing," Proceedings, Third National Congress of Applied Mechanics, 1959.
[5] W. Weibull, "A Theory of Fatigue Crack Propagation in Sheet Specimens,"
Acta Metallurgica, Vol II, No. 7, July, 1963, pp. 745-752.
[6] N. E. Frost and K. Denton, "Effect of Sheet Thickness on the Rate of Growth
of Fatigue Cracks in Mild Steel," Journal of Mechanical Engineering Science,
Vol 3, No. 4, 1961.
[7] D. Broek, P. DeRijik, and P. J. Sevenhuysen, "The Transition of Fatigue
Cracks in Alclad Sheet," NLR-TR-M. 2100, Nat. Aero and Astronautical
Research Inst., November, 1962.
[8] D. Broek and J. Schijve, 'The Effect of Sheet Thickness on the Fatigue Crack
Propagation in 2024-T3 Alclad Sheet Material," NLR-TR-M. 2129, Nat.
Aero and Astronautical Research Inst., April, 1963.
[9] R. W. Hertzberg, "Application of Electron Fractography and Fracture Me-
chanics to Fatigue Crack Propagation in High Strength Aluminum Alloys,"
Ph.D. dissertation, Lehigh University, Bethlehem, Pa., May, 1965.
NLR-TR-M. 2122, Nat. Aero and Astronautical Research Inst., April, 1964.
[11] H. W. Liu, "Fatigue Crack Propagation and the Stresses and Strains in the
Vicinity of a Crack," Applied Materials Research, October, 1964.
[12] P. C. Paris, "The Fracture Mechanics Approach to Fatigue," Fatigue an
Interdisciplinary Approach, Proceedings, 10th Sagamore Army Materials
Research Conference, Syracuse University Press, Syracuse, N. Y., 1964, pp.
107-132.
[13] D. P. Wilhem, "Analysis and Prediction of Cyclic Crack Growth Transitions'
2024-T3 and 7075-T6
Copyright by AluminumInt'l
ASTM Alloys," (all
Report rights
NOR 65-135, NorthropMon
reserved);
Norair, July, 1965.
[14] D. Broek and ofJ. Schijve,
University "The Influence
Washington of Mean Stress
(University of on the Propaga- pursuant
Washington)
FATIGUE CRACK PROPagationAGATION
tion of Fatigue Cracks in Aluminum Alloy Sheet," Report NLR-TN-M. 2111,

Nat. Aero and Astronautical Research Inst., January, 1963.
Structures," Advances in Aeronautical Sciences, Vols 3 and 4, 1961.
[16] C. M. Hudson and H. F. Hardrath, "Effects of Changing Stress Amplitude
on the Rate of Fatigue-Crack Propagation in Two Aluminum Alloys,"
NASA-TND-960, September, 1961.
[17] S. Yusuff, "Fracture Phenomena in Metal Plates," Aircraft Engineering, May,
1964.
[18] N. F. Harpur, "Material Selection for Crack Resistance," Proceedings, Crack
Propagation Symposium, Vol II, Cranfield, 1961, pp. 442-466.
[79] J. R. Dixon, "Stress Distribution around a Central Crack in a Plate Loaded
in Tension: Effect of Finite Width of Plate," Journal, Royal Aeronautical
Soc., Vol 64, 1960, pp. 141-145.
[20] C. M. Hudson, "Fatigue-Crack Propagation in Several Titanium and Stain-
less-Steel Alloys and One Superalloy," NASA TND-2331, October, 1964.
[27] C. M. Hudson, "Studies of Fatigue Crack Growth in Alloys Suitable for
Elevated-Temperature Applications," NASA TND-2743, April, 1965.
DISCUSSION
/. A. Dunsby1 (written discussion)—Would the author care to com-

ment on the morphology of cracks in clad as compared with bare ma-
terial? One might expect the cladding to undergo transition at a stress
intensity factor other than that of the core. Are there indications of two
separate transitions, or do they blend?
/. C. Newman, Jr.2 (written discussion)—Mr. Wilhem has analyzed
fatigue crack propagation data for a wide variety of materials to demon-
strate a correspondence between the initial formation of shear lips (tran-
sition) on the fatigue surfaces and the knee in the curve of crack propaga-
tion rate against the stress intensity factor, Afc, plotted on semilog paper.
He associates this transitional point with a specific stress intensity in
the data treated. Unfortunately, the data available to Mr. Wilhem were
limited to cases with a rather narrow range of R values (ratio of minimum
to maximum load). Thus, the influence of this parameter could not be
studied, but may have contributed to the scatter observed. Support for
this suggestion comes from the observation that crack propagation data
show a consistent effect of R. Also, one might expect intuitively that
the formation of shear lips should be dependent upon the maximum
stress intensity rather than on A&.
1
Senior research
Copyright officer, Int'l
by ASTM National Research
(all rights Council,
reserved); MonOttawa,
Dec 7Ont., Canada.
14:40:45 EST 2015
2
Aero-space technologist,
Downloaded/printed by NASA Langley Research Center, Langley Station,
Hampton, Va. of Washington (University of Washington) pursuant to License Agreement. No
University
DISCUSSION ON CYCLIC CRACK GROWTH TRANSITIONAL BEHAVIOR 381
The two figures of this discussion have been prepared from test data
recently obtained by C. M. Hudson of the NASA Langley Research
Center and show the influence of R values from 0 to 0.85 on transition.
The data in Fig. 14 are from fatigue crack propagation tests of 7075-T6
aluminum alloy, and the data in Fig. 15 are for Ti 8Al-lMo-lV titanium
alloy. The data indicate a rather clear influence of R on the A& at tran-
sition. Similarly, the maximum stress intensity at transition can be shown
to vary as a function of R. The number beside each of the symbols in the
FIG. 14—Stress intensity at transition as a function of load ratio.
figures gives the date of crack propagation in microinches per cycle for
that data point. An interesting observation from these tests is that the
transition points occur at very nearly the same crack growth rate for a
given material. Since the rate of crack propagation is likely to be a vari-
able dependent upon stress considerations, as is the formation of shear
lips, a functional relationship of stress that includes the parameter R is
needed to explain the transitional behavior.
Forman et al3 have developed an empirical relation for correlating the
3
R.Copyright
G. Forman, V. E.
by ASTM Int'lKearney, and R. Mon
(all rights reserved); M. Engle, "Numerical
Dec 7 14:40:45 Analysis of Crack
EST 2015
Propagation in Cyclic Loaded
Downloaded/printed by Structures," Paper 66-WA/Met-4, presented at ASME
WinterUniversity
AnnualofMeeting, New
Washington York, of
(University N.Washington)
Y., 1966.pursuant to License Agreement. No further reproductions au
FATIGUE CRACK PROagationPAGATION
FIG. 15—Stress intensity at transition as a function of load ratio.
rates of fatigue crack propagation for a range of A&, R, and kc values.

Their equation has the form:
where:
da/dn = rate of fatigue crack propagation,
kc = fracture toughness value, and
C and n = empirical constants.
The values of C and n were determined to give the best fit for the crack
propagation data mentioned above. The fracture toughness values
were determined from static strength tests and are given on the figures.
This equation may then be used to determine the relation between Ak
and R for a given value of da/dn. This relation has been plotted on each
figureCopyright
for the byaverage
value of da/dn at transition and is shown as a
solid line.
University of Washingtonbroken
In Fig. 14 the (Universitylines show thepursuant
of Washington) upper toand lower
License boundsNo further re
Agreement.
DISCUSSION ON CYCLIC CRACK GROWTH TRANSITIONAL BEHAVIOR 383
on crack propagation rate at transition. The individual values of da/dn

are seen to be in reasonable agreement with the average and the trend
of Ak at transition against R is very nearly as indicated by the equation.
In conclusion, further studies on other materials are needed to see
whether this observation of constant crack growth rate at transition is
a common occurrence.
Incidentally, the stress intensity values used in this discussion were
computed by the relation k = <r\/a, in which a is the half length of
the crack instead of <r\/] as used by Mr. Wilhem.
D. P. Wilhem (author)—The author wishes to thank the discussers for
their comments, particularly Mr. Newman for his enlightening presenta-
tion of the role of stress ratio on transitional behavior.
In reply to Mr. Dunsby: bare aluminum alloys contain complete 90-deg
through-the-thickness crack-front progression up to the start of shear
transition. The clad layer shows total shear throughout the 90-deg mode,
confined to the thickness of the cladding layer. The more brittle base
metal, however, acts in a manner identical to the bare metal, and under-
goes a complete 90-deg progression up to the point of transition. At that
point an abrupt increase in the percent of through-the-thickness shear
takes place, leading to the familiar tongue appearance (Fig. 7) of the
90-deg mode.
The morphological behavior of the cladding layer indicates that due
to higher ductility its transition stress intensity has been exceeded. Two
separate transitions for clad aluminum alloys will therefore not occur, and
none were observed during the experiments on which this paper is based.
It can be hypothesized, however, that two distinct transition stress in-
tensities exist for the core and its cladding layer, the latter having a lower
transition stress intensity compared to the alloyed base material.
Copyright by ASTM Int'l (all rights reserved); Mon

Application of a Double Linear Damage
Rule to Cumulative Fatigue
REFERENCE: S. S. Manson, J. C. Freche, and C. R. Ensign, "Applica-

tion of a Double Linear Damage Rule to Cumulative Fatigue," Fatigue
Crack Propagation, ASTM STP 415, Am. Soc. Testing Mats., 1967, p. 384.
ABSTRACT: The validity of a previously proposed method of predicting
cumulative fatigue damage in smooth ^-in.-diameter specimens based
upon the concept of a double linear damage rule is investigated. This
method included simplified formulas for determining the crack initiation
and propagation stages and indicated that each of these stages could be
represented by a linear damage rule. The present study provides a critical
evaluation of the earlier proposal, further illuminates the principles under-
lying cumulative fatigue damage, and suggests a modification of the origi-
nal proposal. Data were obtained in two stress level tests with maraged
300 CVM and SAE 4130 steels in rotating bending. Two strain level tests
were conducted in axial reversed strain cycling with maraged 300 CVM
steel. The investigation showed that in most cases the double linear dam-
age rule when used in conjunction with originally proposed equations for
determining crack initiation and propagation predicted fatigue life with
greater or equal accuracy than the conventional linear damage rule. An
alternate viewpoint of the double linear damage rule is suggested. This
requires that a limited number of simple two-stress level tests be run to
establish effective fatigue curves for what may be defined as Phases I and
II of the fatigue process. These fatigue curves may then be used in the
analysis of any spectrum of loads involving as loading extremes the two
stresses used for their determination. Only limited verification of the new
method has been obtained to date, and it must presently be limited to the
study of smooth, ^-in.-diameter specimens. However, it may be con-
sidered as a first step in the direction of eventually predicting the effect of
a complex loading history on the life of more complex geometrical shapes.
KEY WORDS: fatigue (materials), crack propagation, cumulative fatigue
damage, linear damage rule, maraging steels, steels
The subject of cumulative fatigue damage is extremely complex and

various theories have been proposed [1-10]2 to predict fatigue life in
1
Chief, Materials and Structures Div., chief, Fatigue and Alloys Research
Branch; and research engineer, respectively, Lewis Research Center, National
Aeronautics and Space Administration, Cleveland, Ohio.
2
paper.
384
MANSON ET AL ON APPLICATION OF DOUBLE LINEAR DAMAGE RULE 385
advance of service. The most widely known and used procedure is the
linear damage rule commonly referred to as the Miner rule [7]. It is
well known that the linear damage rule, which indicates that a summa-
tion of cycle ratios is equal to unity, is not completely accurate; how-
ever, because of its simplicity and because it has been found to be in
reasonable agreement with experimental data for certain cases it is al-
most always used in design. If a new method is to replace the linear
damage rule in practical design it is important that much of the simplicity
of the linear damage rule be retained. The double linear damage rule,
considered herein, retains much of this simplicity and at the same time
attempts to overcome some of the limitations inherent in the conventional
linear rule.
One of the limitations of the linear damage rule is that it does not
take into account the effect of order of loading. For example, in a two-
stress level fatigue test in which the high load is followed by a low load,
the cycle summation is less than unity, whereas a low load followed by a
high load produces a cycle summation greater than unity. The effect of
residual stress is also not properly accounted for by the conventional
linear damage rule, nor does it take into account cycle ratios applied be-
low the initial fatigue limit of the material. Since prior loading can reduce
the fatigue limit, cycle ratios of stresses applied below the initial fatigue
limit should be accounted for [10]. In addition, "coaxing" effects present
in some strain-aging materials [11] in which the appropriate sequence
of loading may progressively raise the fatigue limit are not accounted
for by the linear damage rule. Various methods have been proposed as
alternatives to the linear damage rule. None overcomes all of the defi-
ciencies, and many introduce additional complexities which either pre-
clude or make their use extremely difficult in practical design problems.
The possibility of improving the predictions of a linear damage rule
by breaking it up into two phases, a linear damage rule for crack initia-
tion and a linear damage rule for crack propagation, was first suggested
by Grover [12]. No rational basis for this approach was indicated, nor
were definite expressions provided for separating out the two phases.
One of the authors of this paper considered these aspects in greater
detail [13]. Total life was considered as consisting of two important
phases, one for initiating a crack and one for propagating a crack, and
a linear damage rule was applied to each of these phases. This double
linear damage rule was intended to correct the deficiencies associated
with order of loading; the other limitations cited above are not directly
taken into account. Simplified formulas derived from limited data for
determining the crack initiation and propagation stages were tentatively
presented.
The present study was conducted to provide a critical evaluation of the
proposal of Ref 13, specifically
Downloaded/printed by the analytical expressions for separating
out the two phases. Additional data were obtained in two stress level
tests in rotating bending and two strain level tests in axial reversed
strain cycling. The materials investigated were maraged 300 CVM and
SAE 4130 steels. Fatigue life predictions by the double linear damage
rule and the conventional linear damage rule are compared with experi-
mental data. In addition, instead of using the analytical expression given
in Ref 75 to represent the crack propagation stage in the application of
the double linear damage rule as originally proposed, a more generalized
expression is suggested which involves the separation of the fatigue proc-
ess into two experimentally determined phases. These are not necessarily
the physical processes of crack initiation and propagation.
Concepts of Double Linear Damage Rule
Analytical Application
In Ref 13 it was proposed that the crack propagation period (A-/V)/
and crack initiation N0 can both be expressed in terms of total fatigue
life Nf by the following equations
and
where the coefficient P = 14. The experimental basis for the selection
of this value of coefficient is given in Refs 13 and 14 and will also be
further described later in the text. The equations expressing cumulative
fatigue damage in terms of the double linear damage rule as proposed
in Ref 13 are:
For the crack initiation phase
If any part of the loading spectrum includes a condition where Nf < 730
cycles, an effective crack is presumed to initiate upon application of
that first loading cycle.
For the crack propagation phase, the expression is
where:
No = cyclic life to initiate an effective crack at a particular strain
or stress level,
(AAO/ = cyclic life to propagate a crack from initiation to failure
at a particular strain or stress level,
Nf = cyclic life to failure of specimen, and
n = number of cycles applied at a particular strain or stress level.
An example of the manner of applying these equations for a simple two
APPLIED CYCLE RATIO, P j / N j l
FIG. 1—Fatigue damage in two stress level tests interpreted by linear damage
rules.
stress level loading case is given in Appendix I. Further discussion of the

equation relating crack initiation and propagation to total fatigue life
is presented in Ref 14. It should be emphasized that these equations
were derived on the basis of data obtained with i^-in.-diameter speci-
mens of notch ductile materials and have thus far been shown to be
valid only for this size specimen [14]. Of course, most materials would
be notch ductile for such a small specimen size. This aspect is discussed
more fully in Ref 13.
By comparison the conventional linear damage rule is expressed as
Equation 5 states that a single summation of cycle ratios applied at

different stress or strain levels is equal to unity.
Graphical Representation of Double Linear Damage Rule Applied to

Two-Stress Level Fatigue Test
Figure 1 illustrates the graphical representation of the double linear
damage rule plotted in terms of the remaining cycle ratios, « 2 /jV/, 2 , at a
TABLE 1.—Material description.

Material Nominal Composition Condition Hardness
4130 (soft) . C 0.30, Mn 0.50, P 0.0

1700 F, % hr in salt, Rc 25 to 27
S 0.040, Si 0.28, Cr water quenched;
0.95, Mo 0.20, Fe re- 1200 F, % hr in salt,
mainder air cooled
4130 (hard)... . same as above 1600 F, K hr in salt, Rc 39 to 40
water quenched; 750
F, 1 hr in salt, air
cooled
300 CVM . C 0.03 max, Si 0.10 ma 90 F, 3^4- hr, air Rc 52
Mn 0.10 max, S 0.010 cooled
max, P 0.010 max, Ni
18.50, Co 9.00, Mo
4.80, Al 0.10, Ti 0.60,
B 0.003, Zr 0.02
added, Ca 0.05 added
TABLE 2— Tensile properties of test materials.

Yield
lUotBrial
Matenal Strength,
0.2% .
Fracture ReductionModulus of
strength, ksi Strength, ksi of Area, %
MODULUS OF
Elasticity, psi
offset,ksi
4130 (soft) 113 130 245 67 3 32 X 106

4130 (hard) 197 207 302 54.7 29
300 CVM 295 380 50.7 27
second stress level against the cycle ratios, ni/N/,i, applied at an initial
stress level. Also shown is a dashed 45-deg line which represents the
conventional linear damage rule. The figure is illustrative of the case in
which the prestress condition is the high stress, and this is followed by
operation to failure at a lower stress. The position of lines AB and BC
would be located on the other side of the 45-deg line for the condition of
low prestress followed by operation to failure at a high stress. Referring
to Fig. 1, according to the double linear damage rule, if the cycle ratios
applied (»i/W/,i) are less than the number required to initiate an effective
crack at a particular stress level, then the remaining predicted cyclic life
ratio Copyright
(«2/7V/,by2) ASTM
would lie along AB. The linearity of AB is implicit in the
Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
assumption of a linear
Downloaded/printed by damage rule for crack initiation. Point B repre-
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions author
sents the cycle ratio applied at the first stress level which is sufficient to
initiate an effective crack, so that upon changing to the second stress
level the remaining cycle ratio at that stress level is exactly equal to the
total propagation stage. The coordinates of this point are designated as
N0,\/Nfti, and ANz/N/t. Beyond this initial cycle ratio N0ii/Nf,i, the
(A) KROUSE ROTATING BENDING SPECIMEN
r
0.481
_L
(B) R. R. MOORE R O T A T I N G BENDING SPECIMEN.
( C ) A X I A L F A T I G U E TEST S P E C I M E N .
FIG. 2—Fatigue specimens.
first applied cycle ratio is more than that required to initiate an effective
crack, and the crack propagation phase is entered. This is represented
by the line BC which is also straight reflecting the second assumed linear
relation. The remaining cyclic life ratio then lies along line BC. Thus,
in two-step tests in which a single stress level was applied for a given
cycle ratio and the remainder of the life taken up at a second stress level,
two straight lines positioned
Downloaded/printed by as shown would be expected. It should also
FATIGUE CRACK PROPAGagationATION
be emphasized that Point B is significant since it permits determination

of both the effective crack initiation and propagation periods for both
stress levels used in the test.
A final point should be made with respect to the graphical application
of the double linear damage rule. Since lines AB and BC are straight
and since Points A and C are fixed, ideally only two tests are required to
establish the positions of these lines and consequently the Point J?. The
only requirement for selecting these tests is that in one test the cycle
ratio applied at the initial stress level should be relatively large, and for
the other test it should be relatively small, in order to insure that the
remaining cycle ratios #2/-/V/,2 do not both fall on the same straight line,
either AB or BC. The significance of obtaining Point B in this simple
fashion is apparent in the illustrative examples of Appendixes II and III.
Experimental Procedure
Materials
Two steels, SAE 4130 and maraged 300 CVM, were investigated.
Their compositions, final heat treatments and hardnesses are listed in
Table 1 and their tensile properties in Table 2. Two different types of
test specimens were used to accommodate the Moore and Krouse rotating
bending test machines. A third type of specimen was used for axial strain
cycling tests. All three specimen types are shown in Fig. 2. The 4130
steel test specimens were machined after heat treatment. The maraged
300 CVM specimens were machined prior to aging, and after aging, finish
ground to remove the final 0.015 in. from the test section. In addition all
rotating bending specimens were machine polished with abrasive cloth
of three grit sizes (320, 400, and 500). After final polishing the specimens
were subjected to a microscopic examination at X 20.
Tests
Specimens were subjected to rotating bending in modified Moore and
Krouse rotating beam fatigue machines and to axial reversed strain
cycling in hydraulically actuated axial fatigue machines. In the rotating
bending tests a rotational speed of 5000 rpm was employed at the lower
stress levels. In order to avoid the detrimental effect of severe heat ac-
cumulation due to hysteresis, rotational speeds as low as 100 rpm were
employed at the higher stresses, and jets of cooling air were directed at
the specimens. A specimen runout no greater than 0.001 in. full indicator
reading was permitted upon installation into the fatigue machines. Addi-
tional details regarding the rotating bending test procedure are given
in Refs 9 and 10. Axial fatigue tests were run at 20 cpm. Details of the
test procedure areASTM
Copyright by givenInt'l
in Ref 75. reserved); Mon Dec 7 14:40:45 EST 2015
(all rights
fatigue curves for
by each material were obtained by fairing the best
visual fit curves through the median data points obtained at each stress
or strain range level. The number of data points at each level varied from
a maximum of 25 to a minimum of 2. In conducting the investigation
specimens were prestressed at a single stress (in rotating bending tests)
to the desired percentage of material life as determined from the fatigue
curves of the original material and to a single strain range (axial fatigue
tests) as determined from strain range-life curves of the original material.
The specimens were then run to failure at various stress (or strain range)
FIG. 3—Two stress level rotating bending fatigue tests for determination of
coefficient in expression for crack propagation. Material, maraged 300 CVM steel.
levels. The specific conditions are indicated on the figures that describe
the results of these tests.
Results and Discussion
Comparison of Experimental and Predicted Fatigue Life by Originally
Proposed Double Linear Damage Rule and Conventional Linear
Damage Rule
Figure 3 shows the results reported previously in Ref 13 for maraged
300 CVM steel which were obtained from rotating bending tests. The
stress levels were so chosen that life at the initial stress was approxi-
mately 1000 cycles and at the second stress 500,000 cycles. Experimental
data are shownby by
Copyright the circles.
ASTM Int'l (all The
rights solid linesMon
reserved); represent
Dec 7predicted be- 2015
14:40:45 EST
haviorDownloaded/printed
by the double linear
by damage rule using different values of the
FATIGUE CRACK PRaagationOPAGATION
FIG. 4—Fatigue curves for materials investigated.
coefficient in Eq 1. For a value of coefficient equal to 14 the predicted

behavior was represented by the line ABG; for a coefficient of 12 it was
ACG, etc. If a linear damage rule applied for the total life values, the
behavior would be that shown by the dashed line AG. A reasonable agree-
ment with the experimental data was obtained for a coefficient of 14.
Since these data represent only one material and one combination of
high and low stress, Eq 1 was only tentatively proposed [13] as being
representative
Copyright byofASTM
cumulative
Int'l (all fatigue damage
rights reserved); behavior.
Mon Dec 7 14:40:45 EST 2015
extending this approach
by many additional tests were conducted with
(a) Krouse rotating bending - high to low stress with low initial life.
FIG. 5—Comparison of predicted fatigue behavior by conventional and double

linear damage rules with experimental data for two stress (strain) level tests.
Material, mar aged 300 CVM steel.
University of Washington (University of Washington) pursuant to License A
(b) Krouse rotating bending - high to low stress with high initial life.
FIG. 5—Continued.
Copyright by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST
University of Washington (University of Washington) pursuant to Licen
(c) Axial strain cycling - high to low strain with low initial life.
(d) Rotating bending and axial strain cycling - low to high stress (strain).
FIG. 5—Continued.
the same and with other materials in rotating bending and axial reversed
strain cycling. Figure 4 shows the fatigue curves of these materials,
maraged 300 CVM and SAE 4130 steel, hard and soft. Since both the
Moore and Krouse machines were used for the 300 CVM tests the fa-
tigue curves obtained with each machine are shown (Fig. 4a). The curves
are largely coincident. Figure 4b shows the fatigue curve for maraged
300 CVM steel obtained in axial reversed strain cycling.
Predictions
Copyrightof by
fatigue
ASTMbehavior by rights
Int'l (all the double linear damage
reserved); Mon Decrule7 (using
14:40:45 EST
the expression 14N/0 6 as representing
Downloaded/printed by the crack propagation stage) and
(a) R. R. Moore rotating bending - high to low stress with low initial life.
(b) R. R. Moore rotating bending - high to low stress with relatively high initial life.
FIG. 6—Comparison of predicted fatigue behavior by conventional and

double linear damage rules with experimental data for two stress level tests. Mate-
rial, SAE 4130 steel.
the conventional linear damage rule are compared with experimental

data in Figs. 5 and 6. Different combinations of loading corresponding to
different life levels were chosen. Figure 5a presents the results from ro-
tatingCopyright
bendingbytests
ASTMforInt'l maraged 300 CVM
(all rights reserved); steel
Mon Dec designed
7 14:40:45 to give rela-
EST 2015
tively Downloaded/printed
low fatigue livesbyof 1280, 1870, 2050, and 2350 cycles at the initial
stress level. The loads at the second stress level were chosen to give lives
up to 940,000 cycles. Generally the greater the difference between the
initial and final life level (that is, initial and final stress applied) the
greater the deviation between the experimental data and the predicted
behavior by the conventional linear damage rule shown by the 45-deg
dashed line; also, the steeper is the first (corresponding to line AB, Fig. 1)
of the two solid lines which predict fatigue behavior by the double linear
damage rule. Agreement between predicted fatigue behavior by the double
linear damage rule and experimental data is good for these test conditions.
This might be expected since the higher stress level as well as some of
the lower stress levels are generally of the same order as those selected
originally for determining Eq 1 for this same material in Ref 13.
Figure 5b deals with the same material but considers other combina-
tions of test conditions in which the initial life level is relatively high.
It is apparent that the greatest discrepancies between experimental data
and predicted fatigue behavior by the double linear damage rule as
originally proposed occur when both the initial and final life levels are
high. It would be expected that this double linear damage rule would
predict almost the same fatigue behavior as the conventional linear
damage rule in these cases because the crack propagation period as
determined from Eq 1 would be relatively small. This is readily seen by
using Eq 1 for values of Nf,i of 15,925, 47,625, 44,000, and so forth, the
specific conditions which are considered in Fig. 5b. The experimental
data show appreciably lower values of remaining cycle ratio, n 2/N/,z,
than would be expected by either rule.
Figure 5c illustrates the results obtained under conditions of axial
strain cycling with maraged 300 CVM steel. The initial life level was
chosen in all cases to be less than 730 cycles. For this case the major
part of the fatigue life would be taken up by the crack propagation
period according to the expressions thus far assumed for crack propaga-
tion and initiation in applying the double linear damage rule. Since
there is essentially no crack initiation stage, the predictions by the double
linear damage rule should coincide with those by the conventional linear
damage rule. This was the case for the two conditions in which the final
stress level was chosen so as to give a low value of life Nf,2 and the experi-
mental data agreed well with the predictions. However, when the second
stress level was chosen so as to give a long life, 7V/>2 = 15,950 cycles,
the predicted fatigue life by the double linear damage rule was less than
that obtained experimentally. It is apparent from Figs. 5b and c that
there are deviations of the experimental data on both sides of the predic-
tions made by the double linear damage rule when the expression 14JV/0-6
was used to represent the crack propagation stage.
Thus far consideration
Copyright has(allbeen
by ASTM Int'l rightsgiven onlyMon
reserved); to the
Decgeneral caseEST
7 14:40:45 in which
2015
the high stress or strainby
Downloaded/printed (for strain cycling tests) was applied first. Figure
5d illustrates the opposite case. Except for the single axial strain cycling
test the predictions by the double linear damage rule show general
agreement with the experimental data. Regardless of deviations of
individual data points from the predictions, it is evident from the
figure that the order effect of loading is accounted for by the double
linear damage rule.
The results for SAE 4130 steel are shown in Fig. 6. Part a of the figure
deals with tests in which the initial life level was low, and loads at the
FIG. 7—Effect of stress combinations in determination of crack propagation

period. Material, 4130 soft steel, R. R. Moore rotating bending tests.
second stress level were chosen to give various life values up to 203,000
cycles. Part b of the figure considers cases where the initial life level
was relatively high. In both cases, however, the order of load application
was that of high stress followed by low stress. In general the results ob-
tained with 4130 steel are the same as those obtained with the mar aged
300 CVM steel for similar test conditions. For the most part agreement
between predictions by the double linear rule using (AJV)/ = 14/V/0-6 and
experimental data was good, although deviations between predic-
tions and data are clearly present in some cases. As was the case for the
maraged 300 CVM steel, a more conservative prediction was always
provided by the double linear damage rule, assuming the expression
147V/0-6 as being representative of the crack propagation stage, than by
the conventional linear damage rule when the high stress was applied
first. Copyright by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
FIG. 8—Stress response in axial strain cycling for constant strain amplitude
tests of maraged 300 CVM steel.
FIG. 9—Stress response in axial strain cycling two strain level tests of maraged
300 CVM steel.
Examination of Assumed Relation for Crack Propagation (AN)/

In view of the deviations noted between predictions and experimental
results closer examination of the assumption that the crack propagation
period (AN)/ may be expressed by the relation 147V/0-6 is clearly in order.
However, before considering the possibility of improving this relation by
changing the coefficient or exponent or both, attempts were made to
determine experimentally if the propagation period (AN)/ was indeed
uniquely dependent upon fatigue life to specimen failure, N/ . The results
of oneCopyright
suchbyinvestigation are reserved);
ASTM Int'l (all rights shownMoninDecFig.7 7. Values
14:40:45 of ANi and A7V2
EST 2015
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authorized
were obtained from two stress level tests with SAE 4130 soft steel in
which N/,i was 485 cycles and A7/,2 was 14,000 cycles using the graphical
method previously described and illustrated in Fig. 1. These values are
plotted on Fig. 7 as Points B and A'. The values of ANt and AN2 simi-
larly obtained from another set of data in which N/t\ was 14,000
cycles and A 7 /* was 203,000 cycles are also plotted on Fig. 7 as Points
A and C. Obviously, Points A' and A do not coincide as they would be
expected to if AN were solely a function of A7/ . Thus, whether a given
stress (corresponding to a fixed life) is used as the first or the second stress
in a two-stress level fatigue test is clearly significant and entirely different
results can be obtained. If the representation of the crack propagation
period by the expression 14A7/0-6 were correct the points determined as
above would fall on the line with a slope of 0.6 when A7/ values were
greater than 730 cycles. It must therefore be concluded that the concept
of representing crack propagation by a universal relation in terms of
Nf, whether the coefficient is 14 or any other number, would produce
some discrepancies. Other tests of the same type for other combinations
of stress were also made. These gave similar results to those shown in
Fig. 7.
There are probably several reasons why the crack propagation period
is not uniquely related to total fatigue life (that is, life to failure of the
specimen). Reexamination is in order of the concept that the effective
crack length for crack initiation is the same at all stress levels, and that
extending a crack at a stress level different from that at which it was
initiated is simply a continuation of the same process. Obviously the
mechanisms involved are not so readily explainable. What may corre-
spond to a crack length for effective crack initiation at one stress level
may not be so at another stress level.
Another reason for the discrepancies relates to the hardening and
softening characteristics of materials. Upon changing to a new strain
level in a two-step test, a material that hardens or softens extensively
will not reach the same stress level for a given applied strain as it would
have, had that same strain been maintained throughout the test. This is
illustrated in Figs. 8 and 9. Figure 8 shows the stress response in axial
strain cycling at constant strain amplitude for maraged 300 CVM steel.
Two tests were run at each of two values of total strain. These were
chosen to give lives on the order of 400 and 16,000 cycles. Agreement
between the two tests run at each condition was good and demonstrated
the ability to maintain and control approximately the same strain level
on the fatigue machines used. Figure 9 illustrates the stress response in
axial strain cycling two-strain level tests for maraged 300 CVM when the
higher of these two-strain levels was applied first and the lower strain
level Copyright
subsequently applied. It is evident that maraged 300 CVM is a
strain-softening material.
Downloaded/printed by As continually increasing percentages of the
life were applied at the higher strain level, the stress required to main-
tain that strain level progressively decreased. Also shown on the figure
are the results of running for 5, 25, and 75 per cent of the total life at
the initial strain followed in each case by operation to failure at the lower
strain level. In each case the stress required to maintain the lower level
of constant strain in these two-step tests was lower than that required
to maintain this level of strain in a single strain level test. Thus, the
material was oversoftened as a result of the initial application of a high
strain level. As a consequence one would expect a longer life than would
be predicted by the double linear damage rule using the expression
14/V/0-6 as representing the crack propagation stage. Figure 9 shows
this to be true. The circles represent the predicted lives according to the
FIG. 10—Illustration of fit of two straight lines to rotating bending data ob-
tained from two stress level tests with SAE 4130 steel.
double linear damage rule using (A7V)/ = 147V/0-6; the squares are the
experimentally determined lives.
In order to describe the cumulative fatigue damage process more
accurately methods must be sought to account for the factors discussed.
This can be done while still retaining the double linear damage rule
concept as discussed in the next section.
An Alternate Viewpoint of Double Linear Damage Rule

In the suggested alternate approach the concept of crack initiation
and propagation in the literal sense is altered to represent two effective
phases of the fatigue process which might be designated as Phases I and
II. The assumption of a linear damage rule for each of these two phases,
however, would be retained. That such an assumption is reasonable may
be seen by inspection
Copyright of (all
by ASTM Int'l therights
datareserved);
obtained MoninDec
this7 14:40:45
investigation.
EST 2015This is
particularly evident from
Downloaded/printed by some of the rotating bending test results ob-
tained with SAE 4130 steel shown in Fig. 60. These results are replotted
in Fig. 10 to illustrate how well two straight lines originating at ordinate
and abscissa values of 1.0 fit the data. The coordinates of the intersection
of these lines (as defined in Fig. 1) determine the values of N0 and AN
used to establish the fatigue curves which represent Phase I and Phase
II of the fatigue process. In keeping with this change in concept the form
of the rule would be different for different materials and for different
extreme loads that might be applied in a test. Additional experimental
verification of this approach is still needed; however, it would seem to
FIG. 11—Determination of Phase I and Phase II of fatigue process from two-

stress level tests at highest and lowest stress levels of loading spectrum. Material,
maraged 300 CVM steel.
take into account the complexities discussed in an approximate fashion.

It is interesting to note that the use of a new fatigue curve different from
that of the original material in predicting remaining fatigue life after
prestressing is not inconsistent with other methods such as that of Corten
and Dolan [4], In general such previous approaches have assumed that
the modified fatigue curves are best determined from a consideration
of the highest and lowest stress levels of the applied spectrum. This
basic approach will also be adopted in the following treatment.
To apply the double linear damage rule in the light of this revised
concept to any anticipated loading spectrum for a given material a
decision must first be made as to which are the highest and lowest loads
of importance. Stress
Copyright by ASTM levels
Int'l below the fatigue
(all rights reserved);limit
Monwill
Decnot 7be14:40:45
considered
EST 2015
for theDownloaded/printed
present. A series by of two stress level tests would be run in which
the highest stress level would be applied first followed by operation to

failure at the lowest stress level of significance within the loading spec-
trum. This should bring into play the important variables such as any
extremes of hardening or softening of the material and extremes of crack
length involved in initiating and propagating an effective crack. From
such a series of tests it is possible to determine for that particular com-
bination of stress levels the values of N0 and AN for both stresses by
using the graphical procedure for applying the double linear damage
rule as previously described. These values may then be plotted as shown
in Fig. 11 at the two stress levels and curves sketched between these
points that are consistent with the appearance of the original fatigue
curve. It would then be possible to analyze the effect of block or spectrum
loading of any pattern that could also include loadings between the
highest and lowest levels by the double linear damage rule. The N0 and
AN curves would be used for determining the effective values of Phase
I and Phase II of the fatigue process. These curves would replace the
expression (AN)/ = 14./V/0-6 or any other variation of such a formula.
One example of applying this procedure is given in Appendix II.
Although at this early stage of the development of this approach, it
seems most convenient to use a two-level test as illustrated in Fig. 1
to determine the effective values of Phases I and II for given extremes of
stress or strain level within a given loading spectrum, further considera-
tion may reveal better approaches for particular circumstances. For
example, it may be desirable to use a block loading in which the highest
and lowest significant stresses in the cycle are more typical of the spec-
trum of service loading for establishing the point of effective transition
between the two phases (Point B in Fig. 1), rather than merely following
the high stress cycles by continuous loading at the lower stress. Since
only two unknowns are involved (the points of effective transition from
Phase I to Phase II for each of the two stress levels when applied in
conjunction with the other) only two tests would be required to deter-
mine the two unknowns. This approach is further discussed in the
Concluding Remarks.
Limited Experimental Verification of Alternate Viewpoint of Double

Linear Damage Rule
In order to provide an indication of the validity of the alternate view-
point of the double linear damage rule, a series of repetitive alternating
two stress level block tests was conducted. Such a test may be considered
as the next step in complexity to the single block two stress level test
which provided the bulk of the data obtained in this investigation. The
manner of conducting the test is fully described in Appendix III. Briefly,
a twoCopyright
stress level by
single block base was selected. Equal fractional portions
ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST
of theDownloaded/printed
number of cycles at each
by stress level in the block were applied in a
University of Washington (University of Washington) pursuant to Licens
(b) Summation of cycle ratios applied at low stress.
(c) Total summation of cycle ratios.
FIG. 12—Comparison of experimental and predicted summation of cycle ratios

for alternating two stress level tests. Predictions made by double linear damage
rule using experimental data to define Phase I and Phase II of fatigue process.
repetitive fashion. The double linear damage rule was applied to pre-
dict the summation of the cycle ratios required to cause failure using
experimentally determined curves representing Phase I and Phase II of
the fatigue process. A numerical example illustrating the use of this
method in making these predictions is also given in Appendix III.
The experimental results of these tests as well as the predictions are
shownCopyright
in Fig. by
12.ASTM
The summations
Int'l (all rightsof reserved);
the cycleMon
ratios
Decare7plotted
14:40:45against
EST 2015
the fractions of the basic block considered. Part a of the figure deals
with the summation of cycle ratios applied at the high stress; Part b with
the summation of cycle ratios at the low stress; and Part c with the
total summation. The experimental data shown represent the arithmetic
averages of three data points obtained at each fraction of the block con-
sidered. In general, there is reasonable agreement between the predicted
results and the experimental data. The irregularity in the predicted re-
sults is probably associated with failure in either the high or low portions
of the block loading pattern. In all cases the predictions by the double
linear damage rule are conservative. Of course, it is important to note
that much additional experimental verification is needed to fully estab-
lish the usefulness of the double linear damage rule in predicting re-
maining fatigue life for more complex loading spectra. The single series
of tests contained in Fig. 12 serve more to illustrate the approach than to
prove validity of the method.
Concluding Remarks
Cumulative fatigue damage in two stress level tests was predicted
with reasonable accuracy for smooth ^-in.-diameter specimens by a
previously proposed method based on the concept of a double linear
damage rule. However, it can be concluded that while a double linear
damage rule involving the assumption that (A7V)/ = 147V/0-6 gives better
results than the conventional linear damage rule it is not adequate
where crack initiation and propagation are expressed solely in terms of
total life. Other representations of crack initiation and propagation
might be more accurate, but they must in some way take into account
the hardening and softening characteristics of the material and more
particularly the effect of the stress levels involved. An alternate view-
point of the double linear damage rule in which the concept of crack
initiation and propagation in the literal sense is altered to represent two
effective phases of the fatigue process, designated as Phases I and II
which can be determined experimentally, appears to overcome some of
the limitations of the original proposal. The form of the rule then be-
comes different for different materials and for different extreme loads
that might be applied. In principle only two tests are required to deter-
mine the point of transition between the two effective phases of the
fatigue process. The actual number of tests employed can of course be
greater, depending upon the degree of accuracy desired. A suggested
approach is to conduct the first test by applying a cycle ratio n\/N/,\ of
approximately 0.2 at the high stress and then operate to failure at the
low stress. For the second test, apply a cycle ratio ni/N/,i of approxi-
mately 0.5 at the high stress before running to failure at the low stress.
Two straight lines emanating from ordinate and abscissa values of one
on a Copyright
plot of nz/Nfj against «i/JV/,i may then be drawn, each of which
by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST
passesDownloaded/printed
through one of the by
data points describing these tests. The co-
ordinates of the intersection of these two straight lines will determine

the values of N0 and A# used to establish the fatigue curves which repre-
sent Phases I and II of the fatigue process.
It is emphasized that the purpose of the two-step test is to provide a
determination of the transition between these two phases of fatigue,
and thereby to enable the investigator to construct effective fatigue
curves for each phase as shown in Fig. 11. Once these two curves are
constructed, their application is intended for all spectrums of loading
involving as loading extremes the two stresses used for their determina-
tion. Thus, use of these curves is not limited to two-step tests. Until
further research is conducted to extend the application, the procedure
must be limited to the study of smooth }/±-in.-diameter specimens. It
may, however, be regarded as a first step in the direction of eventually
predicting the effect of a complex loading history on the life of more
complex structures.
APPENDIX I
Application of Double Linear Damage Rule to a Two-Stress Level Test Using
Relation 14 TV/0-6 to Define Crack Propagation Period
Given two stress levels 1 and 2, at which total life of the original material is
Nfti and Nft2, respectively, and a prestress cycle ratio ni/A'/,! ; it is desired to
find the number of cycles that can be applied at the second stress level. The
values of AM and AA^ are first determined from Eq 1. The values of N0il and
N0,z can then be obtained by subtraction using Eq 2. Next, determine the ratio
N0ii/Nfri. For the case where W/,i > 730 cycles, if /ii/W/.i is equal to N0,i/Nf,i,
the crack initiation stage has just been completed and the cyclic life remaining
at the second stress level is exactly equal to that making up the crack propaga-
tion period, or
If the ratio «i/A/-,i > N0,i/Nf,i , the life remaining at the second stress level may
be expressed as
If the ratio /ii/AT/,i < N0,i/Nf,\, the life remaining at the second stress level may
be expressed as
For the case where

Copyright by ASTM 730(all
Nf,i <Int'l cycles it isreserved);
rights assumedMon
that Dec
there7is 14:40:45
no lengthy
ESTcrack
2015
initiation period, but rather
Downloaded/printed by that total life consists only of crack propagation.
Then the life remaining at the second stress level can be determined from the
expression
In effect then for the latter case total life at stress 2 is determined from the linear
damage rule for crack propagation only.
APPENDIX II
Application of Double Linear Damage Rule Using Experimental Data to Define
Phases I and II of Fatigue Process
In this appendix detailed examples will be given to show how the Phase I and
Phase II curves of Fig. 11 were obtained and how these curves might possibly
be used to predict the life of a three-stress level fatigue test.
To define the two phases of the fatigue process some two-stress level tests
must first be conducted using the highest and lowest stresses of importance in
the particular loading spectrum under consideration. For purposes of this il-
lustration the material chosen was maraged 300 CVM steel, and the two stresses
chosen were 290,000 and 120,000 psi. From the original fatigue curve of Fig. 11
for this material (obtained on a Krouse machine) Nf,i and Nf,2 equal 1280 and
244,000 cycles, respectively. The data obtained from a series of tests conducted
Copyright by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45

FIG. 13—Two straight lines fitted to data from two stress level tests of maraged
Downloaded/printed
300 CVM steel. by
University of Washington (University of Washington) pursuant to L
by applying various cycle ratios «i/M/,i at the high stress and operating to failure
at the low stress, are plotted in Fig. 13. Straight lines were then fitted through the
data. These were required to originate from cycle ratio values of 1.0 on the or-
dinate and abscissa. The coordinates of the intersection Point B are «i/AT/,i and
«2/M/,2 and have numerical values of 0.25 and 0.24. Since these ratios are equiv-
alent to M0>i/M/,i and AM/M/,2 as shown in Fig. 1, the values of M,i, AM and
N0,2 and AM were calculated to be 320, 960, 185,000, and 59,000 cycles, respec-
tively. These values were then plotted at their corresponding stresses as shown
in Fig. 11 and were connected by curves which approximate the shape of the
original fatigue curve. These curves may then be used in separate linear summa-
tions for Phase I and Phase II of the fatigue process.
As a numerical example of the method of applying the double linear damage
rule using these Phase I and Phase II curves consider a three-stress level test in
which the highest and lowest stresses are 290,000 and 120,000 psi. It is required
to predict the remaining life at a third stress level, 200,000 psi, after 200 and
40,000 cycles, respectively, have been applied at the highest and lowest stresses.
Values of M0,3 and AM3 can be obtained from the Phase I and Phase II curves of
Fig. 11. Thus, M,s equals 5900 and AM3 equals 6100. The application of 200
cycles at stress 1 results in a ratio of
indicating that Phase I has not been completed and that it is continued at the
second stress level. The application of 40,000 cycles at the second stress results in
a ratio of
Summing up the cycle ratios applied at stresses 1 and 2 results in
The portion, x, of the number of cycles applied at stress 3 needed to complete

Phase I is, from Eq 3,
or
x = 885 cycles
The portion of the number of cycles applied at stress 3, needed to complete
Phase II is, from Eq 4,
or y = 6100 cycles
Then, the total number of cycles remaining at the third stress level is equal to
x
APPENDIX III
Application of Double Linear Damage Rule to Alternating Two-Stress Level Test
in which Experimental Data Are Used to Define Phase I and II of Fatigue
Process
Alternating two-stress level tests were conducted as follows: First, two-stress

level tests were conducted at various cycle ratios, /zi/W/,1 at a high stress, 190,000
psi, and the remaining cyclic life ratios n^/Nf^ were determined at a second
stress, 110,000 psi. At 190,000 psi, Nf.i was found to be 8000 cycles. At 110,000
psi, Nf ,2 was found to be 625,000 cycles. These N/ values were obtained with
specimens from a different heat of maraged 300 CVM steel than the Moore
FIG. 14—Experimental determination of Phase I and Phase II transition used

in conjunction with double linear damage rule. Material, maraged 300 CVM steel.
machine data previously described in this paper for this material. The results of
the tests in which different cycle ratios were applied at 190,000 psi were plotted
as shown in Fig. 14. Best visual fit straight lines were drawn through the data,
again meeting the requirement that they originate from a value of cycle ratio of
1.0 on the ordinate and abscissa. From the coordinates of the intersection Point
B, and the values of N/,i and W/,2, the Phase I and Phase II parameters were
determined. Thus, N0,i equaled 1300 cycles, AM 6700 cycles, N0,z 537,000 cycles,
and AAT2 88,000 cycles. Several alternating two-stress level block tests were then
specified such that various fractions of 1300 cycles were applied at the high
stress of 190,000 psi and identical fractions of 88,000 cycles were applied at the
low stress.
The following numerical example illustrates the manner of applying the double
linear damage rule to an alternating two-stress level test. The sample considers
the case of an alternating block test in which the alternating or repeated block
is taken to be one by
Copyright halfASTM
of the number ofrights
Int'l (all cycles reserved);
at each stressMon
levelDec
in the7 base
14:40:45 ES
block.Downloaded/printed
Both the base block and the
by alternating block example are shown diagram-
matically in Fig. 15. The base block for this example (as well as all tests of Fig.
12) is defined as consisting of 1300 cycles at the first stress and 88,000 cycles at
the second stress. To determine the number of cycles to complete Phase I apply
Eq 3. Since
FIG. 15—Diagram of loading pattern for numerical example of Appendix HI.
it is apparent that Phase I has not been completed in the first loading block. To
determine if Phase I is completed in the high stress portion of the second loading
block, again apply Eq 3
or x = 543 cycles. Phase I has then been completed. Next, determine the number
of cycles needed to complete Phase II. Apply Eq 4 to determine first whether
Phase II is completed in the second loading block. This gives
indicating that Phase II has not been completed in Block 2. Therefore, determine
if Phase II is completed in the high stress portion of Block 3. Thus
and
y = 3243
Since y > 650, Phase II has not been completed in the high stress portion of
Block 3, and the next step is to determine if it is completed in the low stress
portion of this block. Thus
and z = 34,000. Since z < 44,000 cycles, Phase II has been completed and failure
occurs during the low stress portion of Block 3. The total summation of cycle
ratios for this example then is
In the same manner other apportionments of the cycles sustained at the inter-
section point of a two stress-level test (analogous to Point B of Fig. 1) can be
computed and the expected number of cycles to failure predicted for alternating
block loading applications.
References
[1] L. Kaechele, "Review and Analysis of Cumulative-Fatigue-Damage Theories,
Memorandum RM-3650-PR, Rand Corp., August, 1963.
[2] F. E. Richart and N. M. Newmark, "An Hypothesis for the Determination of
Cumulative Damage in Fatigue," Proceedings, Am. Soc. Testing Mats., Vol
48, 1948, pp. 767-800.
[3] S. M. Marco and W. L. Starkey, "A Concept of Fatigue Damage," Transac-
tions, Am. Society Mechanical Engrs., Vol 76, 1954, pp. 627-632.
[4] H. T. Corten and T. J. Dolan, "Cumulative Fatigue Damage," Paper No. 2
of Session 3, from International Conference on Fatigue of Metals, Vol 1,
1956, Institute of Mechanical Engrs.
[5] A. M. Freudenthal and R. A. Heller, "On Stress Interaction in Fatigue and a
Cumulative Damage Rule, Part I, 2024 Aluminum and SAE 4340 Steel
Alloys," TR 58-69, AD No. 155687, Wright Air Development Center, June,
1958.
[6] Palmgren, "Die Lebensdauer von Kugellagern," ZVD1, Vol 68, 1924, pp. 339-
341.
[7] M. A. Miner, "Cumulative Damage in Fatigue," Journal of Applied Me-
chanics, Vol 12, 1945, pp. A159-A164.
[8] D. L. Henry, "A Theory of Fatigue Damage in Steel," Transactions, Am. So-
ciety Mechanical Engrs., Vol 77, 1955, pp. 913-918.
[9] S. S. Manson, A. J. Nachtigall, and J. C. Freche, "A Proposed New Relation
for Cumulative Fatigue Damage in Bending," Proceedings, Am. Soc. Testing
Mats., Vol 61, 1961, pp. 679-703.
[10] S. S. Manson, A. J. Nachtigall, C. R. Ensign, and J. C. Freche, "Further In-
vestigation of a Relation for Cumulative Fatigue Damage in Bending,"
Transactions, Am. Society Mechanical Engrs., February, 1965.
[11] F. C. Rally and G. M. Sinclair, "Influence of Strain Aging on the Shape of
the S-N Diagram," Report No. 87, Department of Theoretical and Applied
Mechanics, University of Illinois, Urbana, 111., June, 1955.
[72] H. J. Grover, "An Observation Concerning the Cycle Ratio in Cumula-
tive Damage," Fatigue in Aircraft Structures, ASTM STP 274, Am. Soc.
Testing Mats.,
Copyright by1960,
ASTMpp. Int'l
120-124.
(all rights reserved); Mon Dec 7 14:40:45 EST 2015
[13] S.Downloaded/printed
S. Manson, "Interfacesbybetween Fatigue, Creep, and Fracture," Proceedings,
University of Washington (University of Washington) pursuant to License Agreem
International Conference on Fracture, Sendai, Japan, September 14, 1965,

and International Journal of Fracture Mechanics, March, 1966.
[14] S. S. Manson and M. H. Hirschberg, "Prediction of Fatigue of Notched
Specimens by Consideration of Crack Initiation and Propagation," proposed
NASA technical note.
[75] R. W. Smith, M. H. Hirschberg, and S. S. Manson, "Fatigue Behavior of
Materials Under Strain Cycling in Low and Intermediate Life Range," NASA
TN D-1574, 1963.
DISCUSSION
C. R. Smith1 (written discussion)—In going from high cyclic loading to

low cyclic loading, did you notice whether the sign of the previous high
load had any effect on life at subsequent low loading? In the case of
notched specimens, a high tension load should leave a residual com-
pressive stress which would be beneficial, while a high compressive load
would result in residual tensile stresses which would be detrimental at
subsequent lower level loading.
S. S. Manson, J. C. Freche, and C. R. Ensign (authors')—In answer
to Mr. Smith's comment, several points should be noted. First, most
of the tests described herein were in rotating bending. The pre-stress load
was removed manually while the specimen continued to rotate. Rotation
was stopped immediately after load removal. Such a procedure ap-
preciably reduced any residual stresses of the type mentioned by Mr.
Smith. In the axial strain cycling tests the stress and strain are essentially
uniform across the section unless there is a notch present. For the
smooth specimen tests removal of the load reduces the uniform stress to
zero. Only in notched specimens which were not investigated in this
program would residual stresses be appreciable.
1
Design specialist, General Dynamics/Convair, San Diego, Calif.
Review, Analysis, and Discussion of the
Fatigue Crack Growth Problem
.Schijve1
Significance of Fatigue Cracks in

Micro-Range and Macro-Range
REFERENCE: J. Schijve, "Significance of Fatigue Cracks in Micro-

Range and Macro-Range," Fatigue Crack Propagation, ASTM STP 415,
ABSTRACT: The fatigue phenomenon can be described as a sequence of

four phases: (1) crack nucleation, (2) microcrack propagation, (3) macro-
crack propagation, and (4) final failure. The last phase is not discussed.
Technical problems related to the other phases are indicated. The con-
tinuum mechanics approach is discussed, including the application of the
stress intensity factor to microcracks and macrocracks. Several factors
are mentioned that cannot be accounted for by continuum mechanics.
Questions regarding the microstructural aspects of crack growth are ana-
lyzed, and a tentative model for crack growth is outlined. The discussion
is illustrated by results of recent NLR investigations regarding the growth
of microcracks and the influence of several factors on the propagation of
macrocracks, including mean stress, variable amplitude loading, sheet
thickness, sheet width, heat treatment, manufacture, batch, rolling direc-
tion, environment, and frequency. All data apply to aluminum alloys and
mainly to 2024-T3. The main results are summarized in a number of con-
clusions, and problems for further research are listed.
KEY WORDS: fatigue (materials), crack nucleation, crack propagation,

microcracks, macrocracks, continuum mechanics, stress intensity factor,
microstructure, dislocations, inclusions, environment
Nomenclature
k Stress intensity factor
ka k based on Sa
frmax k based on m8X
kmia k based on Smin
oretical stress concentration

/ Half crack length for central crack (21 = crack length from tip to
tip), and crack length for edge crack or corner crack
n Number of load cycles applied
1 National Aerospace Laboratory NLR, Amsterdam, Holland.
415
nx Number of load cycles to obtain a crack length / = x

dl/dn Crack rate
N Fatigue life
NI Fatigue life until a crack with a length 7 is obtained
r, e Polar coordinates
R Stress ratio = Smia/Sm&*
S Stress 1
Sa Stress amplitude
Sm Mean stress j- gross stress, unless otherwise specified
•Smax Maximum stress
Smin Minimum stress
Su Ultimate tensile strength
So.2 Yield strength
t Sheet thickness
T Temperature
w Half width of specimen
5 Elongation
More than 30 years ago Bacon [I]2 expressed the following view: "The
major mystery of fatigue is not so much why cracks form at unsafe limits,
but why they take so long about it. This will never be cleared up so long
as the total of cycles to rupture is all that is observed. Indeed, it is not
going too far to describe the standard method of fatigue test of plain
polished specimens as how to discover a minimum of information with a
maximum expenditure of time, labor, and expense."
The present symposium is just one way of showing that the situation has
drastically changed since then. Recent publications show excellent elec-
tron micrographs of fatigue cracks and fractures with interpretations based
on dislocations and other physical concepts. One might now get the im-
pression that we are intensively interested in everything that occurs be-
tween the start and the end of a fatigue test except perhaps for the fatigue
load applied and the number of cycles until failure. The truth is that fa-
tigue has both technical and physical aspects, whereas a technical solution
of the fatigue problem requires a physical understanding. This constitutes
the theme of the present paper.
After an indication of technical problems, a discussion of two different
approaches to the growth of fatigue cracks is made, namely, the con-
tinuum mechanics approach and the microstructural approach. The dis-
cussion is illustrated by test results of recent National Aerospace
Laboratory investigations on the propagation of both microcracks and
macrocracks. These investigations were performed on aluminum alloys,
2
The italic numbers
Copyright by ASTMin Int'l
brackets refer reserved);
(all rights to the listMon
of references appended
Dec 7 14:40:45 ESTto2015
this
paper. Downloaded/printed by
SCHIJVE ON CRACKS IN MICRO- AND MACRO-RANGE 417
and this restriction in principle applies to the present paper as well. Conse-
quently, the technical problems discussed are related to aircraft structures.
Fatigue Phenomenon
As a result of many microscopical investigations it became clear that
microcracks may be nucleated very early in the fatigue life of a specimen.
Secondly, the investigations, supplemented by other kinds of observations,
FIG. 1—Percentages of the fatigue life at which small cracks (0.1 to 1 mm)
were present. Tests at R = 0 on unnotched specimens of bare 2024-T3. Results
from Ref. 9.
showed that the nucleation had to be associated with cyclic slip. This has
led to a division of the fatigue life until failure into the following phases:
cyclic slip, crack nucleation, growth oflmcrocrack, growth of macrocrack,
and final failure.
It will be obvious that the transition from one phase to the subsequent
one cannot easily be defined. For instance, which local changes of the
material constitute a crack? The vagueness of defining the limits of the
phases need not bother us too much. The important point is that fatigue
successively passes through the above phases.
An indication of the duration of the phases may be obtained from Fig.
1. ForCopyright
unnotched aluminum alloy specimens it shows the percentage fa-
tigue Downloaded/printed
life expired to obtain
by certain values of the crack length, the smallest
418 418IGUE CRACK PROPAGTION
value being 0.1 mm. If stresses near the fatigue limit are disregarded, the
graph suggests that the major portion of the fatigue life was probably cov-
ered by microcrack propagation.
Of course, Fig. 1 cannot be generalized to all materials and loading
conditions, but two features may still have a broader application: (1) For
finite lives (say N < 106 cycles) the period of cyclic slip and crack nuclea-
tion will be a relatively very short one and its duration may be practically
negligible. (2) For stress amplitudes just above the fatigue limit, visible
cracks turn up at a relatively late stage of the fatigue life (see also Ref
FIG. 2—Schematic picture of crack propagation curves, indicating the range

of values for the crack length.
72) and either the nucleation or the first part of the microcrack growth
appears to be an extremely slow process.
The specimen of Fig. 1 had a width of 16 mm. Cracks with a much
greater length are important for structures, in particular for aircraft struc-
tures. A schematic survey of the values of the crack length that may have
technical significance is given in Fig. 2, including some indications on
relevant dimensions of the material or the structure. It is needless to say
that Fig. 2 shows trends rather than accurate data. Cracks smaller than
the atomic distance (about 3 A) can obviously not be considered. Figure
2 already indicates that the various phases of the fatigue phenomenon may
have a different significance for different technical problems. It also illus-
tratesCopyright
the large range of values of the crack length, which covers a ratio
1 to 108 or 109. by
fromDownloaded/printed
Technical Problems
Technical fatigue problems all have in common that predictions are
required about the behavior of a specimen or a structure under cyclic
loading. The predictions may be concerned with fatigue lives or crack
propagation (fail-safe problem).
The estimation of fatigue properties is essentially a problem of correlat-
ing fatigue properties of specimens of the same material under different
conditions. Problems that could be indicated as being almost classical are
listed in Table 1. The first four problems are concerned with the prediction
of fatigue lives, the major part of which is occupied by the nucleation and
propagation of microcracks. Problems 5a to 5c involve macrocrack propa-
gation only. It appears that a rational approach to the problems listed in
TABLE 1—Some fatigue problems regarding the correlation of fatigue

properties under different conditions.
Problem Comparison Made Between
1 Effect of stress amplitude different Sa or Sm (same speci-

and mean stress mens)
2 Notch effect notched and unnotched speci-
mens
3 Size effect large and small specimens with
similar geometry
4 Cumulative damage constant-amplitude and variable-
amplitude loading (same speci-
mens)
5a Macrocrack propagation different Sa or Sm
5b constant-amplitude and variable-
amplitude loading
5c different specimens or structures
Table 1 requires the determination of correlations between the stress

environment and the fatigue phenomenon, including crack nucleation,
microcrack growth, and macrocrack growth. Continuum mechanics are
the only means to accomplish such an approach.
Continuum Mechanics Approach
Correlation Methods
Before discussing the potential applications of continuum mechanics, it
is useful to sketch the fatigue process in some more detail. Fatigue is a
consequence of cyclic slip, and this applies to both nucleation and propa-
gation. The crack rate depends on the amount of cyclic slip in the crack
tip region and on the conversion of cyclic slip into crack extension. The
latter depends on the tensile stress in the crack tip region (crack opening,
strainCopyright
energy byrelease), including
ASTM Int'l (all rightsresidual
reserved);stress.
Mon DecThe amountEST
7 14:40:45 of 2015
cyclic slip
depends on the shear stress
Downloaded/printed by and the local strain hardening.
This sketch of the fatigue phenomenon, which is discussed later, is in

rather broad terms only. Nevertheless, it clearly indicates the questions to
be solved by continuum mechanics. It implies that stress and strain distri-
butions have to be calculated for notched specimens (crack nucleation)
and around cracks. The calculations should account for plasticity and
cyclic strain hardening and should also incorporate residual stresses. If it
is also realized that the problems will have a three-dimensional character,
it will be clear that there is still a formidable task reserved for continuum
mechanics.
Fortunately, problems may be easier if it is possible to correlate results
from different situations, which can be adequately characterized by simple
means without requiring complete and exact solutions. A well-known
example is the stress intensity factor for characterizing the stress field
around the tip of a crack. Another example is the stress concentration
factor, which may be used for correlating crack nucleation in different
specimens. Some elementary comments will first be given on the latter
issue before discussing aspects of the stress intensity factor.
The duration of the crack nucleation period is relatively short, except
for stress levels near the fatigue limit (see Fig. 1). This limit is generally
defined as the highest amplitude at which the fatigue life is still infinite. An
alternative definition is: the highest stress amplitude that does not lead to
crack nucleation. The latter definition assumes that a crack will grow to
complete failure if it has been nucleated. This assumption is thought to be
acceptable for mild notches and positive mean stresses, although the work
of Frost [2,3] showed the assumption to be incorrect for sharp notches and
zero mean stress (nonpropagating cracks). With the second definition it is
a natural approach to consider the stress at the root of the notch as the
characteristic stress level that determines whether crack nucleation will
or will not occur. This approach leads to the well-known relation Kf = Kt.
Deviations from this relation were frequently noted, and there are indeed
several factors that could account for it, for instance, the biaxial stress at
the root of a notch compared with the uni-axial stress in an unnotched
specimen, the size effect (statistical effect and stress gradient effect) and
plasticity, invalidating the elastic stress concentration factor. A noteworthy
proposal with respect to the plasticity effect was due to Gunn [4] who
introduced the plastic stress concentration factor [Kpi] as developed by
Hardrath and Ohman [5], in order to account for both exceeding the yield
stress at the root of the notch and the effect of mean stress. Gunn's pro-
posal neglects cyclic strain hardening (or softening). In this respect, in-
vestigations such as those recently started by Crews and Hardrath [6] may
shed new light on the problem of correlating fatigue strength data. These
authors measured the strain history at the root of a notch under cyclic
loading. It is thought that the estimation of fatigue limits is a field where
SCHIJVE ON CRACKS IN MIC21RO- AND MACRO-RAN421
useful research based on continuum mechanics can be done.3 Although

fatigue limits are perhaps not too important for aircraft structures, they
are important in various other branches of engineering, where the fatigue
limit can have the meaning of an allowable (cyclic) stress level.
In the finite-life range microcracks are nucleated early in the fatigue
life. This complicates correlations based on local stresses that are present
in the absence of cracks. Some further comments are given in the next
section.
The stress intensity factor [7] appears to be so well known by now that
it is unnecessary to explain it in great detail. For specimens or structures
with a crack the asymptotic solution for the stress field (elastic theory)
around the tip of the crack can be written as
with r,6 being polar coordinates at the tip of the crack and k being the
stress intensity factor. For symmetric loadings perpendicular to the
crack the function /(0) does not depend on the load or the geometry of
the specimen. For a specimen loaded by a gross stress S the following
relation for k applies:
k = CSV! ............ /• .......... (2)
The constant C depends on the dimensions of the specimen and on the
crack length / [8], From Eq 1 it follows that k completely determines the
stress distribution in the crack tip region. The assumption is then made
that the same crack rates will be obtained if the same k values apply, or
in other words, the crack rate is a function of k only.
If this relation is determined by experiments, its application does not

require any knowledge of a fracture criterion.
Since the stress intensity factor is based on the theory of elasticity, one
might fear that local plastic deformation at the crack tip would invalidate
the approach. However, if the plastic zone is small the stress redistribu-
tion outside the zone may be small too, and k can still be indicative for
the stress field, although Eq 1 is invalid at the very apex of the crack.
3
It is remarkable that efforts to correlate fatigue limits always compare fatigue
limits of notched and unnotched specimens, although the amount of highly stressed
material is small in the former and large in the latter. This involves a severe risk
of a size effect, and secondly it implies an extrapolation (from unnotched to notched
data) as large as possible. It appears to be a more reasonable approach to compare
fatigueCopyright
limits of by
different
ASTM types of notched
Int'l (all specimens.
4FATIGUE CRACK GATIONPROPAGATION
Stress Intensity Factor Applied to Microcracks

As explained in the section on fatigue phenomenon, microcrack
propagation is important for several technical problems concerning
estimations of fatigue lives. However, so far the stress intensity factor was
applied to macrocracks only, because quantitative data on the propaga-
tion of microcracks were hardly available. In a recent NLR investigation
[9] such data were obtained for unnotched and notched specimens of
aluminum alloy sheet material (bare 2024-T3). Propagation records were
made during the fatigue tests by observing the growing crack through a
FIG. 3—Unnotched specimen and specimens notched by a hole, cut from bare
2024-T3 sheet material. Specimens were used for the observation of the growth of
very small cracks [9].
microscope, employing stroboscopic light. The dimensions of the speci-

mens are given in Fig. 3, which shows that two notched specimens were
used, having a similar geometry but different sizes. All tests were per-
formed with Smin = 0 (R = 0). Crack growth started at the corners of
the minimum section. The corner cracks had the shape of a quarter
circle, and the crack length / was measured from the edge of the hole or
the edge of the unnotched specimen along the specimen surface. For the
corner cracks the crack rate has been plotted as a function of the stress
intensity factor k in Fig. 4. The results plotted were obtained at different
stress levels, tabulated in Fig. 4 with the corresponding fatigue lives, and
at three valuesby
Copyright ofASTM
the crack length,
Int'l (all rightswhich are Mon
reserved); / = 0.2,
Dec 0.5, and 1 mm.
7 14:40:45 The
EST 2015
sheet Downloaded/printed
thickness was 2 mm.by Although there are a few deviating data points
SCHIJVE ON CRACKS IN MICRO- AND MACRO-RANG423E
in Fig. 4, it is thought that the graph demonstrates a surprisingly good

correlation for such small cracks.
The stress intensity factors4 were taken from Ref 1, considering the
corner cracks as edge cracks. Width corrections were not applied. After
the corner cracks had completely penetrated the sheet thickness, the
crack front was perpendicular to the sheet surface. For these larger edge
cracks (/ > 2 mm) similar plots as given in Fig. 4 were presented in
FIG. 4—The crack rate as a function of k for small corner cracks in un-
notched and notched 2024-T3 specimens [9].
Ref 9, which indicated a reasonable correlation for each of the three

types of specimens. The three average curves, presented in Fig. 5, show
systematic differences between the three types of specimens, although the
differences are small. The curve for the corner cracks in Fig. 4 has been
4
It was noted that the stress intensity factor for a hole with a single edge crack
was approximately equal to S ^fl' with 21' = d + I for l/d > 0.05, d being the hole
diameter. That means that the stress intensity factor was approximately the same as
for a crack with by
Copyright a total
ASTM length I. It was
d +rights
Int'l (all surprising
reserved); Mon to see7that
Dec this still
14:40:45 ESTheld
2015for
fairly Downloaded/printed
small cracks. by
FATIGUEAGATION CRACK PROPAGATION
replotted in Fig. 5 and shows that the crack rate of a corner crack is
much smaller than the crack rate of an edge crack. It is clear that it must
be smaller in view of the restraint on crack opening. It was then assumed
that the ratio between the stress intensity factors for corner cracks and
edge cracks is the same as the ratio between the k values for a central
FIG. 5—The crack rate as a function of k for edge cracks in unnotched and
notched 2024-T3 specimens [9]. Comparison with corner cracks.
(through) crack and a circular crack in an infinite body. This ratio is

2/i" [8], and application led to a satisfactory agreement between the
dl/dn — k plots for the small corner cracks and the larger edge cracks,
see Fig. 5.
Stress Intensity Factor Applied to Macrocracks

In Copyright
the literature
by ASTMthe
Int'lstress intensity
(all rights reserved); factor
Mon Decwas applied
7 14:40:45 EST to
2015numerous
crackDownloaded/printed
propagation data, by notably by the group of Paris [7,10]. Good
SCHIJVE ON CRACKS IN MICRO- AND MACRO-RANGE425
correlations between k and dl/dn were obtained, especially for aluminum

alloys (see also Ref 11). In an NLR investigation [12] on the effect of
mean stress on the crack propagation in 2024-T3 and 7075-T6, it turned
out that a good correlation was obtained provided that the results for a
constant stress ratio (R = Smia/Smax) were considered. The results for
the 2024-T3 material are shown in Fig. 6.
If Smin T* 0 the stress distributions around the tip of the crack occurring
FIG. 6—The effect of the stress ratio R on the function dl/dn = f(k). Results
from tests on 2024-13 Alclad sheet material [12] (t = 2 mm, 2w = 160 mm).
at ,Smax and Smin are both significant. Similar to Eq 1 they can be written
as:
and
For a certain R value, A:max determines the asymptotic stress distributions

at both the maximum and the minimum of the load cycle. The fact that
correlations between dl/dn and k are possible for constant R values
should then be
Copyright considered
by ASTM Int'l (allasrights
a further confirmation
reserved); Mon Dec 7 of the conception
14:40:45 EST 2015
of the stress intensitybyfactor being a reasonable approach.
Downloaded/printed
FATIGUE CRACK PROPAGATIOagationN
Figure 6 implies that the function/of Eq 3 depends on the stress ratio:
In other words, the crack rate will not depend on ,S1Eax only but also on
Smin , and this is hardly a surprising result. It would require a fracture
criterion for fatigue crack propagation to account for the effect of R in a
rational way. Empirical relations, however, can be determined [13].
The usefulness of the stress intensity factor can further be checked by
considering the effect of the sheet width on the crack propagation. The
NLR has recently carried out [14] tests on 2024-T3 sheet specimens with
four different values of the width, namely, 80, 160, 300, and 600 mm
(3.15 in., 6.3 in., 11.8 in., and 23.6 in., respectively). The stress intensity
factor for a central crack in an unstiffened sheet of finite width is [15]
with
where 21 is the length of the central crack and 2w is the width of the sheet
(see Fig. 7). Some values of the correction factor C for finite width are:
l/w 0 0.1 0.2 0.3 0.4 0.5

C 1 1.004 1.017 1.039 1.077 1.127
Although the values of C increase for increasing crack length, they remain
close to 1 for fairly large cracks. Equations 7 and 8 then predict that k,
and thus the crack rate, at a certain cyclic stress depend on / but hardly
on the width of the sheet. Only for large values of l/w the width will
become of importance. Results of the tests, which were carried out at
Sm = 8 kg/mm2 and Sa = 2.5, 4 and 6.5 kg/mm2, respectively, were in
good agreement with these predictions. For Sa = 4 kg/mm2 this is il-
lustrated by Figs. 7, 8, and 9. Figure 7 confirms the small width effect on
the crack propagation curves (averages of three tests each) and Fig. 8
shows that the crack rates for high values of / are larger in the smaller
specimens. The latter feature is well accounted for by the width correction
factor C, as shown by Fig. 9.5
Finally, further support for the stress intensity factor stems from ex-
periments reported by Donaldson and Anderson [11], who also tested
cracked specimens of 7075-T6 with the fatigue loads transmitted as point
loads on the edges of the crack. Values of k were known for these loading
5
For a singlebydeviating
Copyright ASTM Int'lpoint in Fig.
(all rights 9 l/wMon
reserved); wasDec
27.5/40 ~ 0.7.
7 14:40:45 ESTFor
2015this relatively
large crack the net stress at the maximum load was 38.4 kg/mm2, and since So.z —
Downloaded/printed
2 by
35.8 kg/mm one can no longer expect elastic conceptions to be valid.
SCHIJVE ON CRACKS IN MICRO- AND MACRO-RANGE
cases, and a reasonable correlation between the crack rate and the stress
intensity factor was obtained. More experiments of this type will be
worth while.
Application of Stress Intensity Factor to Technical Problems
The examples of the correlation between the stress intensity factor and
the crack propagation rate as presented in the two preceding sections
indicate that there may be certain technical applications of this concep-
FIG. 7—The effect of the sheet width on the crack propagation in 2024-T3
Alclad. Results of Ref. 14.
tion. Considering the list of problems in the section on technical prob-

lems, some comments will be given below regarding possible applications,
limitations, and extensions of the theory.
In Table 1 problems have been mentioned concerning the prediction
of fatigue lives, the major part of which is occupied by nucleation and
propagation of microcracks. In the section on "The Stress Intensity
Factor Applied to Microcracks" the correlation between dl/dn and k for
microcracks
Copyright was illustrated
by ASTM by rights
Int'l (all the results
reserved);ofMon
a recent
Dec NLR investigation
7 14:40:45 EST 2015
(specimens of Fig. 3 and results in Fig. 4). The results of the same in-
vestigation can now be used to check the possibilities of application.

The curve of Fig. 4 was assumed as being obtained with unnotched speci-
mens only, and it was subsequently used to predict the fatigue life covered
by crack extension from 0.1 to 1 mm in the two notched specimens. For
this purpose k was calculated for various values of / and the correspond-
ing crack rate values were read from the curve in Fig. 4. The number of
cycles Aî — N0.i was obtained by integrating the inverse of dl/dn. The
results in Table 2 show a good agreement between the test data and the
FIG. 8—The effect of the sheet width on the crack rate as a function of the
crack length. Graph based on same results as Fig. 7.
predictions for the large notched specimens, whereas for the small
notched specimens the predictions are somewhat lower than the test
results. The agreement is, of course, not unexpected in view of the fairly
narrow scatter band of the results of all specimens in Fig. 4. It is, never-
theless, encouraging to see that the stress intensity factor can apply to
cracks as small as 0.1 mm. Considerations on the size of the plastic zone
and the numbers of dislocations and GP zones in the plastic zone, as
given in Ref 16, sustain the view that application of continuum mechanics
could still be allowed for such small cracks.
Table 2 further
Copyright by ASTMshows that
Int'l (all Mreserved);
rights — N0.iMonis only
Dec 7a 14:40:45
small part of the entire
EST 2015
fatigue life N. Another
Downloaded/printed by small part is covered by crack propagation from
5CHIJVE ON CRACKS IN MICRO- AND MACRO-RANGE429
1 mm to failure, but the major part is occupied by crack nucleation and

propagation until / = 0.1 mm. Unfortunately, no reasonable predictions
for this phase of the fatigue life could be made. In order to match the
predictions to the test data it had to be assumed that small cracks with a
length depending on the applied stress level were present in the specimen
at the beginning of the test. A possible alternative assumption was that
FIG. 9—The crack rate as a function of the stress intensity factor. The same
data were used as for Fig. 8.
the crack rates until / = 0.1 mm were much larger than predicted by Fig.
4. Both assumptions could not be given a satisfactory evaluation in the
absence of experimental information.
It was subsequently tried to estimate 7V0.i on the basis of the peak
stress at the root of the notch. For the large notched specimens the stress
gradient is so small that at a distance of 0.1 mm from the edge of the
hole the tensile stress has decreased no more than 3 per cent below the
peak stress KtSmax . Predictions for the large notched specimens were
made Copyright
by deriving TVo.iInt'l
by ASTM values fromreserved);
(all rights the unnotched
Mon Dec 7specimen data
14:40:45 EST 2015for the
same Downloaded/printed
values of the peakby stress (Kt = 2.66 and 1.085 for the notched and
FATIGUE CRACK agationPROPAGATION
the unnotched specimens, respectively). The results are given in Table 3,

which shows poor agreement between test data and predictions. The
main cause is probably that the nucleation was enhanced by inclusions.
In the unnotched specimens cracks were sometimes nucleated away from
the corners and slightly below the surface. Cracks can nucleate in the
TABLE 2—Prediction of the number of cycles for crack growth from 1 = 0.1 mm to
\ — 1 mm in the small and the large notched specimens, based on the results of
the unnotched specimens.
(Prediction employing stress intensity factor)
c*
omax »
a Ni - No.i , kc
apecimen, kg/mm2 AT \re*
In Test Prediction
Large notched specimen 24 14.2 5.0 4.3

16 62 16 17
12 180 36 46
11 300 66 63
10.5 480 72 75
Small notched specimen 24 28 8.4 6.4
16 150 34.5 26
14 350 60 42
13 850 85 54
0
Net stress.
TABLE 3 — Prediction of the fatigue life until a crack of 0.1 mm in the large
notched specimen, based on the results of the unnotched specimens.
(Prediction based on the peak stress
no.1,kc
smax,akg/mm2
In Test Prediction
16 35 13
12 110 40
11 195 50
10 5 360 60
10 ( 75
9 >104 135
8 I 3604
7 [>10
0 Net stress.
unnotched specimens over a much larger area than in the notched speci-
mens, and this may well explain the discrepancies in Table 3. A statistical
size effect emerges from the above reasoning. Obviously, it will not be
easy to account for this in a continuum mechanics theory. A more precise
knowledge of the statistical aspects controlling the nucleation is required,
and microscopic studies on this topic appear to be worth while. They also
could throw more light on such problems as scatter in fatigue life and
improvements
Copyright by ASTMof fatigue
Int'l (all properties.
rights reserved); The
Mon Dec 7 14:40:45 question at which locations
EST 2015
cracksDownloaded/printed
do nucleate by
is certainly an intriguing one, the more so if we realize
hchijv
that unnotched specimens with a long fatigue life generally exhibit one
visible crack nucleus only.
The prediction of macrocrack propagation appears to be of prime
importance for fail-safe problems. Until now information on the crack
propagation in a new aircraft design is obtained in full-scale tests on a
structure or on components. Much would be gained if reasonably ac-
curate estimates could be calculated from Eq 6
dl/dn = fR(k) (6)
after determining the function fR by simple tests on sheet specimens. It is
thought that Eq 6 will account to an acceptable degree of accuracy for
the crack length, the stress level (Sa or Sm&^) and the geometrical con-
figuration, provided that k can be calculated. Variables that may affect
the function fR will be discussed later on.
If Eq 6 takes the form proposed by Paris and Erdogan [10]:
dl/dn = Ck* (9)
it is easily calculated from Eq 2 that a reduction of the stress level in a
structure with 10 per cent will lead to a gain of about 50 per cent in the
crack propagation life. Such indications may already be helpful when
considering the fail-safe properties of an aircraft is the design stage. The
figures also indicate that an underestimation of k with 10 per cent implies
an overestimation of the crack propagation life with 50 per cent.
With respect to the geometrical configuration of an aircraft structure the
complexity of the structure is undoubtedly a major obstacle to the applica-
tion of the stress-intensity-factor conception. Skins are reinforced at one
side only, introducing an asymmetry. Cut-outs and local reinforcements
are further complications. Moreover, in a fuselage the state of stress is
biaxial and due to the cabin pressure the edges of the crack will bulge.
Calculations on the effect of stringers on the stress distribution in a
cracked sheet have been published [17-19]. Although the calculations
were made for relatively simple structural configurations, they are still
informative with respect to the effect of stringers on the stresses at the tip
of the crack and hence on the crack rate. Extension of this type of research,
supplemented by fatigue crack propagation experiments will be very
worth while. It may ultimately turn out that the effect of certain dimen-
sional parameters on the crack propagation can be studied by calculations
and this could yield useful background for developing design philosophies.
Before discussing some of the limitations of the stress-intensity-factor
conception it may be said that it has removed much of the mystery about
the relation between cyclic stress and crack propagation, which is cer-
tainly a step forward on its own. Also the combined development of stress
analysis and by
Copyright computer
ASTM Int'ltechniques will certainly
(all rights reserved); Mon Dec broaden
7 14:40:45 the possible
EST 2015
applications of the stress
Downloaded/printed by intensity factor in the near future.
4FATIGUE CRACK PROPAGATIOhationN
The preceding discussion was restricted to constant-amplitude loading.

In Table 1 reference was made to variable-amplitude loading with regard
to the estimation of fatigue lives (Problem 4) as well as macrocrack propa-
gation (Problem 5b). For both cases it has been known for long that inter-
actions between load cycles of different magnitudes may considerably
affect the increase of fatigue damage. The interaction was attributed to
residual stresses set up around the tip of the crack. An extreme example
of the interaction for macrocracks is shown in Fig. 10, which illustrates
the delaying effect on crack propagation caused by intermittent high loads
[20]. The residual stresses are thought to be responsible for ^ n/W-values
FIG. 10—The delaying effect on the crack propagation in 2024-T3 Alclad

material as caused by intermittent peak loads. Results from Ref. 20 (t = 2 mm,
2w = 160 mm).
larger than unity when program loading with a positive mean stress is ap-
plied. This appears to hold for both fatigue lives [27] and macrocrack
propagation [22]. Since the stress intensity factor is an elastic conception
it is obviously incapable of accounting for interaction effects due to re-
sidual stresses, which are caused by plastic deformation. This is indeed a
serious limitation, which is only partly offset by the fact that neglecting
the interaction will lead to conservative estimates for many practical
problems.
An extension of continuum mechanics to account for residual stress will
be extremely complex. It has to consider the plastic deformation in a
nonhomogeneous stress field and also the strain hardening behavior of the
material, which may differ from cycle to cycle (cyclic strain hardening or
cyclicDownloaded/printed
strain softening).byProgress has been made regarding the calculation
of the plastic stress and strain distributions around a crack during a

monotonically increasing load [23,24], and this may lead to useful results
for residual strength problems. However, for fatigue the problem is more
complicated due to the cyclic nature of the load and the increasing crack
length. Moreover, contrary to the use of the stress intensity factor, a frac-
ture criterion will be required.
In conclusion, attention has to be drawn to a number of variables that
could affect the function
dl/dn = f R ( k } (6)
It was already pointed out that the function depends on the stress ratio.
Unfortunately, there are several other variables that more or less affect the
crack propagation, such as loading frequency, environment, sheet thick-
ness, ductility of the material, manufacture, batch, and rolling direction.
Although these variables do not invalidate the stress-intensity-factor con-
ception, it is still a pity that the value of available data may be limited by
systematic influences of the above variables. These influences are discussed
later.
Some Factors Affecting Crack Propagation

In this section a brief recapitulation will be given of results obtained
in some recent NLR investigations on crack propagation in aluminum al-
loy sheet material. The data will illustrate certain effects on crack growth,
which could be of technical importance and which are not easily accounted
for by continuum mechanics.
In Ref 12 the effect of mean stress on the crack rate in 2024-T3 Alclad
and 7075-T6 clad material was studied. Figure 6 (discussed in the section
on "Stress Intensity Factor Applied to Microcracks") was taken from this
reference. Results of the same tests plotted as a modified Goodman dia-
gram are given in Fig. 11. This figure shows the effect of Sm in a more
direct way, and it also shows the 2024-T3 alloy to be superior to the
7075-T6 alloy. As an average result it could be said that the crack rate in
the latter alloy was 3 to 4 times faster than in the former, more ductile
alloy.
The faster crack propagation in a less ductile alloy was confirmed [25]
when comparing 2024-T81 with 2024-T3, both obtained from the same
manufacturer, see Fig. 12. In Ref 26, where the crack propagation in
2024-T3 material from seven different manufacturers is compared, it
turned out, however, that the crack rates were not correlated with the
ductility alone. This is illustrated by Fig. 13, showing crack propagation
lives obtained with sheet materials from two different batches for each of
the two manufacturers C and F (letter code from Ref 26). For two batches
of theCopyright
same manufacture
by ASTM Int'l (allthe lower
rights lives
reserved); Monare
Dec again associated
7 14:40:45 EST 2015 with the lower
ductility. However, bycomparing Curve 1 with Curve 3, or Curve 2 with
Downloaded/printed
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions auth
FATIGUE CRACK PROPAGATAGATIONION
FIG. 11—Fatigue diagram for the crack propagation life N c (t = 2 mm, 2 w

160 mm, data from Ref. 12).
FIG. 12—Comparison of the crack propagation lives for 2024 Alclad mate-
rial in the T3 and the T81 condition. (Crack growth interval from 21 = 6 mm to
2 1=80 mm, 2^-160 mm, t = 2 mm, Ref. 25).
Curve 4, one cannot attribute the differences to differences in ductility but

has to associate them with differences between manufacturing procedures.
The comparison of results of the materials of seven different manufacturers
has revealed maximum life ratios between the most inferior and the most
superior material in the order of 1:2. If the batch to batch variation is
added, the ratio
Copyright becomes
by ASTM about
Int'l (all rights 1:3; see Mon
reserved); Fig. Dec
13. 7It14:40:45
shouldEST
be2015
borne in
mind Downloaded/printed
that these ratios by
are concerned with average crack propagation data.
Similar data on these sources of scatter were recently published [27] by

Piper et al for 7178-T6 material. It appears that the generalization of data
from a single batch from one manufacture could lead to technically sig-
nificant errors.
All the above data were obtained on sheet specimens with a thickness of
2 mm, loaded in the rolling direction. Specimens loaded transverse to the
FIG. 13—The crack propagation life for three crack propagation intervals and
t
rolling direction [26] on the average showed a 40 per cent faster crack
rate. These tests were carried out on sheet materials from six different
manufacturers at the same stress levels as indicated in Fig. 13.
The effect of sheet thickness was checked on 2024-T3 Alclad material
with five different thicknesses: 0.6, 1, 2, 3, and 4 mm (0.024, 0.04, 0.08,
0.12, and 0.16 in., respectively). Although the results were not fully sys-
tematic, higher crack rates were found in the thicker material [28], The
trendsCopyright
are illustrated
by ASTMby Table
Int'l 4, which
(all rights shows
reserved); an average
Mon Dec 7 14:40:45life
ESTratio
2015between
the thickest and the thinnest
Downloaded/printed by material of 0.54:1. A similar trend was found
by Raithby and Bebb [29]. It thus appears that the rolling direction and
the sheet thickness are additional variables restricting the generalization
of crack propagation data.
The effect of the loading frequency was studied for 2024-T3 Alclad in
Refs 20 and 30, for frequencies of about 20 and 2000 cpm. The results
are summarized in Table 5, including some data from Refs 12 and 31
(specimens from the same sheets). The table shows higher crack rates at
TABLE 4—Comparison of the crack propagation lives for crack growth from 21 =
6 mm to 21 = 50 mm in 2024-T3 Alclad sheet materials with five different thick-
nesses (2w = 100 mm). Data from Ref28.
Sa , kg/mm2 —> 2 .5 4 6. 5
average
t, mm Life, Ratio Life, kc Ratio Life, kc Ratio Ratio
kc
0.6 277 1 98.6 1 33.0 1 1

1 266 0.96 81.0 0.82 25.9 0.78 0.85
2 213 0.77 85.3 0.87 29.5 0.89 0.84
3 176 0.64 58.8 0.61 19.5 0.59 0.61
4 . 179 0.65 49.8 0 51 15 4 0.47 0.54
TABLE 5—Comparison between the crack propagation lives for 2024-T3 Alcttd
sheet material at a low and a high loading frequency (2w = 160 mm,
t — 2 mm).
Crack Propagation
Life, kc, fi•om 11 = 10
Manufacturer sm,kg/mm2 asakgmm2 T, deg C mm to 21 = 60 mm Ratio Reference
20 cpm 2000 cpm
c 8.2 2.4 86.5 103.2 1.19

3.3 20 39.3 52.0 1.32 20
5.5 10.8 14.2 1.31
F 9 3 20 68.6 115 1.68 12, 31
F 7 2.5 96.4 125.3 1.30
4 150° 36.7 51.5 1.40 30
6.5 14.3 19.3 1.35
0 Specimens presoaked at 150 C for 100 hr under load (S = 1 kg/mm 2 ).
the lower frequency. It is noteworthy that the frequency effect for Ma-
terial F appeared to be larger than for Material C. A second remarkable
trend is that the frequency effect for Material F was lower at 150 C than
it was at room temperature. Fractographic observations indicated that
the cracking mechanism might be different at the two temperatures, al-
though it was still transgranular for both.
One could guess that the crack rate at room temperature will increase
further at still lower frequencies, but no data are available. At 150 C
Lachenaud
Copyright by ASTMand Jaillon
Int'l (all rights 7 14:40:45 ESTwho
reserved); Mon Dec[32], 2015 tested RR 58 (aluminum alloy for high
temperature
Downloaded/printed byapplication), indeed arrived at crack rates for 0.3 cpm that
were 2.5 times as high as the rates at 2000 cpm. It will be clear that the
frequency effect should be kept in mind when crack propagation data ob-
tained in conventional high-frequency fatigue machines are to be used
for practical applications concerning much lower frequencies.
Another aspect to be considered is the test environment. In the labora-
tory, room-temperature air with a high humidity is the normal environ-
ment. Chemically active constituents and moisture, such as rain or
dew, which are absent in the laboratory may be present in service. An
exploratory investigation on the influence of an outdoor environment was
carried on both 2024-T3 Alclad and 7075-T6 Clad material [33]. Tests
were conducted in two test rigs, one located indoors and the other one
outdoors. The tests were run concurrently with full-scale fatigue tests on
tension skins [34], and the same random and programmed load sequences,
TABLE 6—Comparison between the crack propagation lives simultaneously
obtained in an indoor and an outdoor test rig (results from Ref33) (2w =
100 mm, t = 2 mm).
indoor crack-propagation life
Ratio = ——: ; : 777-
outdoor crack-propagation life
Cr
Material Type of Loading ^k g^^0™ Ratio
2024-T3 random 5 to 12 1.2

random + GTAC 8 to 18 0.9
program 7 to 24 1.1
program + GTAC 6 to 14 1.2
7075-T6 random + GTAC 4 to 28 1.6
program 4 to 9 2.1
program + GTAC 6 to 22 1.5
representing a severe gust spectrum, were employed. Tests with and with-
out ground-to-air cycles (GTAC) were performed. The frequencies of
loading were in the order of 20 cpm. The results summarized in Table 6
show small differences between the crack rates in the 2024 alloy tested
indoors and outdoors. However, for the 7075 material the outdoor crack
propagation was significantly faster than indoors. This is again an effect
to be considered when applying crack propagation data.
Microstructural Approach
Some Fundamental Questions

The previous section has illustrated the influence of several factors that
cannot easily be introduced in a continuum mechanics theory. In order to
account for these factors when dealing with practical problems we have
to extrapolate from existing empirical evidence, such as presented previ-
ously.Copyright
This requires a qualitative
by ASTM understanding
Int'l (all rights reserved); Mon of
Decthe7 trends
14:40:45observed
EST 2015
and thus of fatigue crack extension as a micrpstructural phenomenon. The

problem of fatigue in aluminum alloys was analyzed in Ref 16, and the
present discussion is mainly based on that analysis. The following ques-
tions were raised there:
1. Which part of the fatigue life is covered by crack propagation?
2. Is there a fundamental difference between crack nucleation and
crack propagation?
3. Does crack propagation occur along crystal planes?
4. Does crack extension occur in every load cycle, and what are the
orders of magnitude of the crack length and the crack rate to be consid-
ered?
5. What is the effect of the stress amplitude on the propagation mecha-
nism? Is there a fundamental difference between low-amplitude and high-
amplitude fatigue, as is sometimes postulated in the literature?
6. How is the stress-strain behavior of aluminum alloys under cyclic
loading? The interaction of dislocation movements and precipitation is a
major aspect of this question.
Brief thoughts on these questions are given in the following. Some com-
ments on inclusions and chemical aspects are added.
A simple model for crack growth, which for obvious reasons has a tenta-
tive character, is described next.
Crack Nucleation and Propagation

Question 1 was already discussed under the heading "Fatigue Phenome-
non" where it was said that the major part of the fatigue life was occupied
by crack growth, except perhaps for stress levels near the fatigue limit.
Question 2, concerning fundamental differences between nucleation and
propagation, can be answered only if the mechanisms for both are known,
which in fact is not true. Observations indicate that nucleation starts at
the surface of the material and that both nucleation and propagation are a
consequence of cyclic slip. Differences should then be associated with
differences between the surface and the crack tip regarding the state of
stress or the environment or both.
The state of stress at the surface of the material is characterized by:
(1) the stress distribution is approximately homogeneous, (2) the volume
of highly stressed material is large, (3) a tendency towards triaxiality is
absent, and (4) the restraint on plastic flow is low. All these circumstances
are different in a crack tip region. The differences will certainly affect the
amount of plastic deformation, but they do not necessarily imply a differ-
ent cracking mechanism.
In Ref 16 the first model of Wood [55] for crack nucleation was ex-
tended to crack propagation. The basic idea is illustrated by Fig. 14, which
shows that crack growth is considered to be a geometric consequence of
to-and-fro slip movements in the crack tip region. If slip planes A and B
change places, an extrusion instead of an intrusion would form and crack
extension would not occur. It is thought, however, that the forming of
intrusions is preferred to that of extrusions in view of the energy release
associated with intrusion. If the planes A and B coincide, the slip would
be reversible. The occurrence of crack growth obviously proves that at
least a part of the slip is irreversible. Irreversibility may be caused by oxi-
dation of the freshly exposed material at the tip of the crack [36,37].
Another mechanism is indicated in Fig. 15, based on dislocation climb or
cross slip.
FIG. 14—Four subsequent stages of a simple model of crack extension during

one cycle [16].
Although Figs. 14 and 15 suggest crack growth due to dislocations mov-

ing into the tip of the crack, it was proposed that dislocations may also be
emitted by the tip of the crack, see Fig. 16. This mechanism is expected to
be promoted by high stresses.
According to the elementary conceptions of Figs. 14—16 crack growth
should occur along a single slip plane. However, if slip at the crack tip oc-
curs on two differently orientated slip planes, crack growth in any direction
between the orientations of the two planes is possible in essentially the
same way, see Fig. 17. The essential feature is that crack extension is a
consequence of "sliding-off."
The cyclic sliding-off mechanism could start right from the beginning of
a fatigue test, and it appears that both nucleation and propagation may be
considered as being caused by the same mechanism. For crack nucleation
at the surface the model requires a component of the slip vector perpendic-
ular to the surface. Observations on single crystals of copper [38,39] sup-
port this view.
TheCopyright
cyclicbysliding-off
ASTM Int'l (allmechanism
rights reserved);implies
Mon Dec 7a14:40:45
conversion
EST 2015of cyclic slip
into crack extension. This could be true even when the actual occurrence
is more complicated than the one sketched in Figs. 14-17. It is expected
that the amount of cyclic slip primarily depends on the local shear stress
amplitude and hence on the amplitude of the stress exerted on the speci-
men. The conversion of cyclic slip into crack extension is thought to be
Oxide film at tip under tension,

which facilitates dislocation flo\v
into crack
a. Rising part of the load cycle.
Oxide film at tip under com-

pression, which hinders flow
of dislocation into crack
b. Falling part of the load cycle. Similar pictures apply

to screw dislocations. Dislocation climb is then replaced
oy cross slip, promoted by strain energy release involved
in crack extension.
FIG. 15—Further evaluation of the crack growth mechanism presented in Fig.
14 [16].
mainly dependent on the local tensile stress, including any residual stress
being present. This implies that for a certain Sa the magnitude of 5"max
(or Sm , or R) will be important for crack growth, as illustrated by Fig.
11, and it is also in agreement with the delayed crack growth due to
residual compressive
Copyright by ASTMstress shown
Int'l (all in Fig. 10.
a Increase of stress b Decrease of stress.

+ oxide film under tension
— oxide film under compression.
FIG. 16—Dislocation emission by the tip of the crack on different slip planes
during the rising and falling part of a load cycle [16].
FIG. 17—Crack extension in one load cycle by dislocation movements on two

different sets of crystallographic planes [16].
Crack Propagation Along Crystal Planes

Crack growth will occur along a slip plane if slip is mainly restricted
to that plane, which does not exclude the occurrence of cross slip. Ob-
servations on pure aluminum and bare 2024-T3 material [16] indicated
crack growth along a slip plane at the surface of the specimens even at
high crack rates. It was thought that the weak restraint on plastic flow at
the surface allowed the selection of a single plane for crack growth, most
favorably orientated for slip.
At the interior of the material crack growth did apparently not occur
along a single plane because it followed a more or less irregular path,
crack branching being not at all rare. Slip in the interior of the material
will not easily be restricted to a single slip plane. The restraint on plastic
flow is larger, and, moreover, the continuity of the crack front through the
various grains with random orientations does not allow the crack to stick
to a single plane in each grain.
Similar observations regarding crack growth along slip planes at the
surface and an apparently noncrystallographic growth in the interior were
made by Forsyth and Stubbington for an aluminum-zinc-magnesium alloy
[40-42] and by McEvily and Boettner [43] for pure aluminum. Forsyth
[40] has indicated the two modes as Stage I and Stage II crack growth.
Although he associated Stage I with nucleation and slow growth and Stage
II with a continuation of Stage I, both stages can occur simultaneously,
and Stage I need not be restricted to slow growth. If the principle of the
model sketched in the preceding section would be correct, the difference
between the two modes is a matter of slip on one or more than one slip
plane contributing to the crack growth.
Crack Extension Occurring in Each Load Cycle

Fractographic observations on macrocracks have clearly shown that
crack extension occurs in each load cycle. The most conclusive evidence is
coming from constant-amplitude tests with single, more severe load cycles
being periodically inserted. Growth lines of the latter cycles can be recog-
nized on the fracture surface, and the number of finer growth-lines in be-
tween corresponds exactly to the number of load cycles between two suc-
cessive more severe load cycles [31,44-46]. An example is shown in Fig.
18. In Ref 31 the smallest spacing observed between successive growth
lines in 2024-T3 was 300 A. Christensen [47] reports a smallest value for
7075-T6 of 200 A and a similar value was reported by Hertzberg [48].
The values correspond to 70 to 100 atomic distances. Probably growth
lines for still smaller spacings are beyond the limit of observation. On the
one hand there are limitations to the microscopical techniques, and on the
other hand it is possible that the cyclic fracturing hardly leaves any growth
markings if the crack extension is so small.
Orders of magnitude of the crack rate, as derived from the tests on
unnotched and notched 2024-T3 specimens discussed previously, are
shown in Table 7. Since cyclic slip will occur in every load cycle, it is
thought that crack growth will in principle also occur in each load cycle.,
Table 7 shows values of the crack rate that are smaller than one atomic
distance per cycle, which is physically impossible. It is thought that crack
extension then does not simultaneously occur along the entire crack front,
LOADING:
Specimen All5, fracture surface at / ~ 14 mm.
Crack rate derived from picture: dl/dn — 0.55 /i/cycle.
Crack rate derived from growth record: dl/dn — 0.50 ^/cycle.
FIG. 18—Optical micrograph (X1700) of a fracture surface of a 2024-T3
specimen, to which a more severe downward load was applied after each ten load
cycles. (Ref 31, 2w = 160 mm, t = 2mm).
so that the average crack rate can be lower than 1 atomic distance per
cycle. It is expected that the continuity of crack growth will be more
easily disturbed for small microcracks than for macrocracks due to local
inhomogeneities of the metal structure. It was also observed for micro-
cracks [16,49,50] that crack growth temporarily stopped when the direc-
tion of crack growth changed.
The fatigue model sketched previously predicts continuous crack growth
(that is, in every cycle) as a consequence of continuous cyclic slip. This
does not necessarily apply to two groups of alternative crack growth
mechanisms. The groups can be characterized as follows:
1. Crack extension is a localized failure due to a local increase of

cyclic stress as a consequence of cyclic strain hardening around the tip of
a crack. Theories of Orowan [57] and Head [52] belong to this group.
2. Crack extension is a localized failure due to a deterioration of
strength as a consequence of cyclic plastic strain in the crack tip region.
Obviously, different mechanisms could apply to different materials in
view of the differences in the crystal structure, the number of easily acti-
vated slip systems, the stacking fault energy (cross slip), the dislocation
density (strain hardening), a precipitation hardening, etc. The two above
groups have in common that a cyclic "preparation" of the material ahead
of the crack is necessary for crack extension and that the crack growth is
not necessarily a continuous process, occurring in every load cycle. Dis-
continuous growth of macrocracks was observed in aluminum [53], soft
brass (unpublished NLR results), and in aluminum-zinc alloys (especially
TABLE 7—Orders of magnitude of the crack rate in unnotched and in notched
specimens of bare 2024-T3 sheet material in the beginning of the fatigue tests
(R = 0) and towards the end of the tests, based on results of Ref. 9.
Order of Magnitude of Crack Rate, dl/dn
Type of Specimen Stress Amplitude Beginning of Test At 95% N
A/c b/c* A/c b/c°
Unnotched low Sa 0.3 0.1 100 30

high Sa 30 10 10 000 3000
Notched lOW Sa 1 0.3 1000 300
high Sa 30 10 30 000 10 000
0
b = size of Burgers vector = 2.86 A for the slip systems {111} (110) in alu-
minum.
in the heavier sections, Ref 54). Apparently, cyclic plastic strain during a
number of cycles had led to a condition of the material prone to brittle
crack extension. This may be further enhanced by a small amount of fa-
tigue crack growth, implying crack tip sharpening and an increased tri-
axiality of the stress.
In pure metals, such as aluminum and copper, cyclic strain hardening
and cyclic strain softening are both possible [55]. The formation of a
substructure due to cyclic loading was observed [37,56-58,67], and it was
argued that crack growth occurred along the subgrain boundaries, assisted
by submicrocracks formed in these boundaries. This conception belongs
to the second group mentioned above. Forming of pores [59] and vacancy
condensation [60] can also be regarded as a local deterioration of strength'
preluding crack nucleation and growth. It is true that on the atomic level
several dislocation mechanisms can explain crack growth, and it is proba-
bly also true that different mechanisms are active in different materials
and under different stressing conditions.
Stressing Conditions
Brief comments will be made on the following three topics: (1) differ-
ences between tension and torsion loading, (2) differences between high-
amplitude and low-amplitude loading, and (3) differences between crack
propagation in the tensile mode and in the shear mode.
Wood et al [59] performed alternating torsion tests on copper and
brass specimens. Crack nucleation and growth were preceded by the form-
ing of pores, which multiplied and finally linked up to cracks. Under alter-
nating torsion there is no tendency to open the crack and hence a much
lower tendency towards conversion of cyclic slip into a crack. It then ap-
pears that other mechanisms than the sliding-off model are more relevant.
A noteworthy difference between torsion and tension was observed by
Forsyth [42] when studying fatigue in an aluminum-zinc-magnesium alloy.
He found resolution of the precipitates and substructure formation in slip
bands where subsequent slip was concentrated, leading to pits and finally
to cracks. However, if a tensile stress was present slip bands were con-
verted into cracks with virtually no prior slip band deterioration.
In the literature it has frequently been suggested that fatigue at a high
amplitude should be different from fatigue at low amplitudes. However,
also for high stress amplitudes leading to short endurances cracks are
nucleated early in the fatigue life and crack propagation is observed [9].
Noteworthy results were obtained by Raymond and Coffin [67] when test-
ing pure aluminum at high strain amplitudes. The surface was heavily
wrinkled during the tests. If the surface was periodically reshaped by
removing a small surface layer, an indeterminate extension of the fatigue
life was obtained. Cracks were observed before the removals. This con-
firms that cyclic plastic deformation per se did not constitute fatigue dam-
age. A similar indication was offered by Alden and Backofen [62]. In
tests on annealed aluminum specimens a thick anodic coating (1000 A)
prevented dislocations from cutting through the surface. Fatigue failures
could not be obtained at a strain amplitude that would have given a fatigue
life of about 105 cycles if the anodic coating had not been present. Still
there was considerable cyclic plastic straining at the interior of the ma-
terial. After removal of the coating and subsequent fatigue loading, the
normal fatigue life was again obtained. This shows that slip movements at
the free surface are essential for nucleating cracks and, at the same time,
that cyclic strain-hardening does not necessarily imply fatigue damage.
It is concluded that if there is a difference between high-amplitude and
low-amplitude fatigue, this has to be primarily a difference between crack
propagation mechanisms. Considering the values of the crack rate in Table
7, ranging from 0.1 to 10,000 atomic distances per cycle, we should
wonder whether a single mechanism could be valid for such a large range
of crack rates. Figure 8 even shows values up to about 100,000 atomic
distances per cycle. It was already said before that sliding-off could occur
by dislocations flowing towards the tip of the crack and at higher crack
rates by dislocation emission from the tip of the crack. It appears that a
crack extension in the order of 104 to 105 atomic distances per cycle is, on
the atomic level, a rather chaotic process, which cannot be very precisely
described. It is especially for the higher crack rates that crack growth in
each load cycle was clearly confirmed by microscopical observations. It
is then thought that the two alternative mechanisms are sliding-off (glide
plane decohesion) and brittle crack extension (cleavage). Indications of
cleavage were obtained by Stubbington [41] on an aluminum-zinc-mag-
nesium alloy, mainly under corrosive conditions. This alloy also exhibits
the so-called tunneling [54]. The problem is that cleavage does not appear
to be a likely process in aluminum alloys in view of its face centered cubic
structure and the relatively large number of equivalent slip systems. This
does not preclude the occurrence of brittle fracture in the sense that little
plastic deformation occurs during crack extension. Such brittle fractures
can still be considered as fast ductile fractures and could still be caused by
some sliding-off mechanism.
Macroscopic observation of a fracture surface shows a difference be-
tween slow and rapid crack propagation. Slow crack growth occurs in the
tensile mode with the fracture surface perpendicular to the maximum
principal stress. Rapid crack propagation occurs in the shear mode, the
fracture surface making an angle of 45 deg with both the maximum princi-
pal stress and the material surface. In Figs. 19 and 20 corner cracks in
unnotched specimens are shown, which start in the tensile mode and then
change over to the shear mode. The transition occurs gradually by the
development of shear lips (shear mode) at the surface. The shear lips be-
come broader when the length of the crack increases, until the whole
crack grows in the shear mode. The tensile mode area is smaller for
higher fatigue stresses.
The development of the shear lips and the transition from the tensile
mode to the shear mode can be associated with the state of stress [16].
Plane strain promotes the tensile mode and plane stress the shear mode.
It is noteworthy that the stress intensity factor correlates crack propagation
in both the tensile mode and the shear mode (see, for instance, Fig. 6; the
transition point refers to the crack length at which the transition to the
shear mode is fully completed).
With the optical microscope the growth lines are easily detected in the
tensile mode, and they are hard to detect in the shear mode. With the
electron microscope growth lines in the shear mode, if visible, showed up,
as blurred bands. There is no doubt that crack growth in the shear mode
occurs in every load cycle, since growth lines remain visible in the cladding
[16]. For a double shear fracture the lines also remain visible in the
center of the sheet. Single shear fractures frequently show the well-known
black powder on the fracture surface whereas double shear fractures do
not. Rubbing of the fracture surfaces may perhaps obliterate growth lines
in the shear mode.
The question whether the transition from the tensile mode to the shear
mode implies a fundamental change of the fracture mechanism cannot be
ignored. In view of the change of the orientation of the fracture surface
(90 deg —> 45 deg) it is thought that screw dislocations omitted by the tip
of the crack will become more important after the transition, whereas
edge dislocations were perhaps predominant during the tensile mode.
Specimen nr 3A35 1A12 3B36 1A35

Sm.x (kg/mm 2) 45.5 47 24 18 '
N (cycles) 13500 5500 141000 662000
FIG. 19—The transition from the tensile mode to the shear mode in unnotched
specimens of bare 2024-T3 material [9]. The cracks started as corner cracks (see
Fig. 20).
Whether this could account for the above observations is not known as
yet, since the formation of growth lines is still largely a matter of specula-
tion. The problem was recently discussed by Hertzberg [48]. The model
as sketched earlier predicts a blunting and a resharpening of the crack in
every load cycle. It is thought that the growth lines are formed during
the latter phase. More research on this intriguing topic is desirable.
Cyclic Strain Hardening
The cyclic strain hardening in aluminum alloys is a very complex prob-
lem concerning the interactions between dislocations and GP-zones. More-
over, it may be different for aluminum-copper base alloys and for alumi-
num-zince base alloys, although cyclic strain hardening occurs in both

alloys [55]. For the 2024 alloy it appeared that the cyclic strain hardening
stabilizes in a very small number of cycles [16}.
Cyclic plastic deformation will occur ahead of the crack, even when
there is crack extension in every cycle. Hence the crack will propagate in
material that has been changed. There are two opposing effects to be as-
sociated with the cyclic strain hardening. (1) The amount of slip will
FIG. 20.—Sketches of fractures shown in Fig. 19.
decrease, and (2) the relief of tensile stress peaks at the tip of the crack will
be less effective. According to the fatigue model outlined before, the
former will be favorable and the latter unfavorable. In view of the ductility
effect mentioned earlier it is expected that the latter aspect will be pre-,
dominant. The results in Fig. 1 3 have shown that two materials with ap-
parently the same static properties could still show significantly different
crack propagation characteristics. In Ref 26 this was attributed to differ-
ences in the heat treatment, which may lead to differences in the disloca-
tion densities, the distribution, and the sizes of the precipitated zones. The
differences need not have the same effect on static properties and on
cyclic strain hardening and hence on crack propagation. Further study
of this difficult issue is certainly worth while.
Inclusions
The effect of inclusions (intermetallic particles) on the propagation of
macrocracks was considered by various authors [26,27,40,63-65]. Re-
sults of Glassman and McEvily [63] showed that the inclusions can be
cracked in a tension test by 3 per cent plastic strain. Pelloux [65] and Piper
et al [27] concluded that the crack propagation was not accelerated by
inclusions as long as the crack rate was low. In an NLR investigation on
2024-T3 sheets from seven different procedures [26] inclusions could
easily be detected on the fracture surface. The growth lines on the tensile
mode fracture gave the impression that the effect of the inclusions was
small and of a rather local nature. For larger cracks (shear mode) the
number of inclusions picked up by the crack increased, and it may well be
that fast crack propagation was accelerated by them. Nevertheless, a
clear correlation between the inclusion pattern and the propagation char-
acteristics could not be established. It seems most worth while to conduct
further study of the question whether high-purity aluminum-copper base
alloys could have improved properties with respect to fast crack propaga-
tion and residual strength.
The small effect of inclusions on crack propagation in the tensile mode
might mean that inclusions will not be important for microcracks. Hunter
and Fricke [72] found that inclusions did not act as focal points for crack
nucleation. However, DeLange [66] reported several cracks starting from
inclusions. Inclusions were mentioned earlier as a possible explanation for
size effects. Although inclusions may perhaps not noticeably affect the
rate of microcrack growth, the possibility of crack nucleation at inclusions
certainly deserves more attention.
Chemical Aspects
The environmental effect on fatigue will obviously depend on the type
of material and the type of environment. The effect may be different for
crack nucleation, microcrack growth, and macrocrack propagation. The
effect of vacuum, water vapor, and oxygen on crack nucleation and micro-
crack growth was analyzed, among others, by Bennett et al [68]. For
aluminum alloys water vapor appeared to have a detrimental effect.
For the propagation of macrocracks in aluminum alloys reference may
be made to recent papers of Hartman [69] and Bradshaw and Wheeler
[70]. In both publications it was shown that water vapor rather than pure
oxygen had a detrimental effect on crack propagation, which was tenta-
tively attributed to some kind of hydrogen embrittlement. The effect was
large for small values of the crack length (life ratios in the order of 10
were obtained), and it decreased at high crack rates. Fortunately, the

effect of the water vapor was very small for the range of water vapor con-
tents that may occur in normal room-temperature air.
A noteworthy result of the two above publications was that the transi-
tion of the tensile mode to the shear mode was hardly affected by the en-
vironment. It is thought that the environment did not affect the amount of
cyclic plastic deformation, but rather the conversion from cyclic slip into
crack extension.
The results of both Hartman and Bradshaw and Wheeler further indi-
cate that oxygen becomes important for very small crack rates and hence
might be important for microcrack nucleation and growth. It is well known
that oxide layers may prevent dislocations from cutting through the surface
and thus could prevent crack nucleation. Since cracks nucleate early in
the fatigue life, the effect of oxide films on the crack surfaces will be of
greater interest. The latter allows ample room for speculation. Oxide
layers will prevent rewelding. The possibility of rewelding was recently
shown for the 2024-T3 alloy by Martin [77]. Secondly, the oxide layer
may affect the cyclic slip at the tip of the crack. Since the layer is brittle it
will break during an increasing tensile load. It is difficult to see whether
this will locally enhance or decrease the amount of slip. Further, the
oxide layer has a larger volume than the parent metal. During a decrease
of the load this may prevent crack sharpening. The hydrogen embrittling
mentioned above was also based on the presence of oxide layers.
For the sake of completeness electrochemical attack has to be men-
tioned in this section. The attack at the tip of the crack will be assisted by
breaking of the oxide layer and by the tensile stress (stress corrosion
aspects). The results mentioned earlier concerning crack propagation
in 7075-T6 sheet material in an indoor and an outdoor environment were
probably associated with stress corrosion.
Discussion
In the previous sections various aspects of crack propagation in alumi-
num alloys were summarized. It will now be tried to give a rough picture
of the present situation and some recommendations for future research.
Continuum mechanics aims at the quantitative prediction of fatigue
lives and crack propagation rates, whereas the microstructural approach is
mainly concerned with the understanding of the material behavior under
cyclic load. The latter serves several purposes: it will indicate limitations
to the application of continuum mechanics; it may lead to improvements
of materials with respect to fatigue; it may contribute to the knowledge of
metal physics.
As a result of microstructural investigations we have obtained a de-
tailed picture of the fatigue phenomenon, including cyclic slipi crack
nucleation, and crack growth. The precision of the picture is steadily in-
creasing. Unfortunately, this has taught us that the phenomenon is de-
pending on a large variety of factors. Fatigue can be different from one

material to another, and it can even be different for one material per se,
depending on loading circumstances. We are in the contradictory situation
that the more we learn about fatigue the more the complexity of the
phenomenon becomes evident. Although the electron microscope has
added much new information, pertinent observations on the atomic level
have not yet been reached.
Several models for fatigue were published in the literature. The model
described earlier is a relatively simple one. Fatigue models may be helpful
for planning further research and they also may serve as a basis for
continuum mechanics.
It is thought that for aluminum alloy, the following subjects could be
worth while for further microstructural research:
1. Crack nucleation. At which preferred sites are cracks nucleated?
2. Interaction between slip and precipitates under cyclic load. Various
heat treatments should be considered.
3. The influence of inclusions on slow and fast crack propagation (and
on residual strength).
4. The formation of growth lines in the tensile-mode fracture and the
geometry of crack growth in the shear-mode fracture.
5. A combined study of the effects of the environment and the fre-
quency.
Studies of the above aspects may yield useful results both for a better
understanding of fatigue and improvement of structural materials.
With respect to continuum mechanics it appears that the stress intensity
factor, k, holds good promise for correlating crack rates, dl/dn, in differ-
ent types of specimens and structures. There are, however, certain re-
strictive conditions that have to be met, namely: (1) constant-amplitude
loading, (2) the same stress ratio, and (3) the same sheet material. The
second and the third restrictions can easily be overcome by performing
crack propagation tests on unstiffened sheet material, which is relatively
simple. The stress ratio effect could also be accounted for by empirical
relations. However, the problem of variable-ampltiude loading as it occurs
in service and additional aspects, such as the effects of loading frequency
and environment and the determination of applicable k values, will pre-
clude an accurate estimation of the crack propagation in an aircraft struc-
ture.
Since such estimates are vital for assessing fail-safe characteristics, it
is thought that crack rate estimates cannot yet replace full-scale testing.
This testing, to be fully relevant, has to be as realistic as possible, which
is extensively discussed in Ref 34. Estimations of the crack propagation
based on the stress intensity factor may be an informative supplement to
full-scale testing. Estimates for different types of structures may give use-
ful indications for design purposes.
Earlier it turned out that stress intensity factors could correlate crack
rates for cracks as small as 0.1 mm. Efforts to estimate the fatigue life
until a crack of 0.1 mm is obtained were unsuccessful, and, unfortunately,
this may cover the major part of the fatigue life.
It now appears that the following subjects could be added to the above
list of worth-while aspects for further research.
6. Calculation of the stress intensity factors for more complicated
structures and fatigue tests to check the applicability of these factors.
7. Extension of the elastic theory to include plasticity. It is realized
that this is a formidable problem for cyclic loading. The estimation of the
static strength of notched or cracked specimens involves similar difficulties
except for the cyclic nature of the load.
8. In view of crack extension under variable-amplitude loading it may
be useful to explore the potential use of models such as proposed by Head
[52] or as adopted in the redundant force analysis. Cyclic strain-hardening
and residual stress should be included as additional parameters describing
the fatigue damage. Extensive computer capacity will be required.
9. The precipitation of crack nucleation is partly a matter of the sta-
tistical aspects concerning weak sites for crack nucleation. To include the
statistical aspect in continuum mechanics, Problem 1 should be explored
first.
Conclusions
The conclusions are restricted to aluminum alloys.

1. A good correlation between the crack rate and the stress intensity
factor was obtained for corner cracks and edge cracks in unnotched and
notched specimens of 2024-T3 sheet material. This applies to corner
cracks as small as 0.1 mm (0.004 in.). Unsatisfactory results were ob-
tained in estimates of the fatigue life to reach a crack length of 0.1 mm,
when adopting either the stress intensity factor or the stress concentration
factor.
2. The relation between the crack rate and the stress intensity factor
is significantly dependent on the stress ratio R.
3. Tests on 2024-T3 Alclad specimens with a width varying from 80 to
600 mm (3.15 to 23.6 in.) indicated a small width effect, in agreement with
predictions based on the stress intensity factor.
4. Tests on 2024-T3 Alclad specimens with a thickness varying from
0.6 to 4 mm (0.024 to 0.16 in.) revealed faster crack rates in the thicker
sheets, the maximum ratio being about 1:2.
5. Tests on 2024-T3 Alclad specimens at 2000 cpm and at 20 cpm1
showed an approximately 30 per cent faster crack rate at the lower fre-
quency for the material of one manufacture, whereas the percentage was
about 70 per cent for the material of another manufacture.
6. Tests on 2024-T3 Alclad specimens taken from sheets of different
manufactures and from sheets of different batches of the same manufacture

indicated systematic differences. The average crack rate could vary as
much as 1:3. Only a part of the variation could be associated with a differ-
ence in ductility.
7. Specimens of 2024-T3 Alclad loaded transverse to the rolling direc-
tion showed a 40 per cent faster crack propagation than specimens loaded
parallel to the rolling direction.
8. Comparative tests in an outdoor atmosphere and in the laboratory
under program and random loading with and without ground-to-air cycles
indicated a small environmental effect for 2024-T3 Alclad. However, for
7075-T6 Clad the outdoor crack-propagation was about 1.5 to 2 times
faster than in the laboratory.
9. The trends summarized in conclusions 4 to 8 do not invalidate the
stress-intensity-factor conception. They imply, however, a warning that
the generalization of data from the literature may lead to inaccurate predic-
tions of crack rates. A major restriction of the stress intensity factor is that
it cannot account for interaction effects between load cycles of different
magnitudes occurring in variable-amplitude loading. For accurate data for
an aircraft structure full-scale tests remain indispensable.
10. A simple and tentative fatigue model outlines crack growth as a
geometric consequence of cyclic sliding-off at the crack tip. This involves
dislocation flow towards the tip of the crack at low crack rates and disloca-
tion emission by the tip of the crack at high crack rates. The conversion of
cyclic slip into crack extension is promoted by a tensile stress normal to the
crack in view of strain energy release. The tensile stress includes any re-
sidual stress being present. It is thought that this cyclic sliding-off mecha-
nism could apply to cyclic tensile loading and to those materials for which
a tendency to brittle fracture and substructure formation under cyclic
load are both absent. Aluminum alloys could be in this group, apart from
aluminum-zinc base alloys under corrosive conditions, in view of their
tendency to brittle fracture.
11. Crack growth along slip planes as observed at the surface, even for
high crack rates, is promoted by the weak restraint on plastic flow. At the
interior of the material crack growth occurs in apparently noncrystallo-
graphic directions as a consequence of the required continuity of the crack
front and of slip occurring on more than one slip plane due to the higher
restraint on plastic flow.
12. Crack extension in principle occurs in every load cycle as a conse-
quence of cyclic slip occurring in every load cycle. This also applies to
microcracks for which growth lines have not been observed.
13. Estimation of the minimum crack rates during the nucleation period
of fatigue cracks in unnotched and notched specimens yielded an order of
magnitude as small as 0.1 atomic distance per cycle. The crack growth
cannot be a continuous process along the entire crack front at this rate.
14. The nucleation of a microcrack under cyclic torsion is different

from that under cyclic tension due to the absence of crack opening.
15. Crack nucleation early in the fatigue life occurs at both high-ampli-
tude and low-amplitude fatigue loading.
16. The effect of inclusions on crack propagation is probably small
when the crack rate is low and it may be larger for fast crack propagation
and residual strength.
17. Although our understanding of fatigue is increasing steadily, there
are still many important aspects which are not known in sufficient detail.
A list of topics for further research is given in the discussion, concerning
both continuum mechanics and microstructural aspects.
References
[7] R. A. McGregor, W. S. Burn, and F. Bacon, 'The Relation of Fatigue to
Modern Engine Design," Transactions, N. E. Coast Institution for Engineers
and Shipbuilders, Vol 51, 1935, p. 161.
[2] N. E. Frost, "Crack Formation and Stress Concentration Effect in Direct
Stress Fatigue," The Engineer, Vol 200, 1955, p. 464.
[3] N. E. Frost, "Notch Effects and the Critical Alternating Stress Required to
Propagate a Crack in an Aluminum Alloy Subject to Fatigue Loading,"
Journal Mechanical Engineering Science, Vol 2, 1960, p. 109.
[4] K. Gunn, "Effect of Yielding on the Fatigue Properties of Test Pieces Con-
taining Stress Concentrations," The Aeronautical Quarterly, Vol 6, 1955,
p. 277.
[5] H. F. Hardrath and L. Ohman, "A Study of Elastic and Plastic Stress Con-
centration Factors Due to Notches and Fillets in Flat Plates," NACA TN
2566, Nat. Advisory Committee for Aeronautics, December, 1951.
[6] J. H. Crews, Jr., and H. F. Hardrath, "A Study of Cyclic Plastic Stresses at a
Notch Root," Paper No. 963, Soc. Exp. Stress Analysis, May, 1965. (See also
NASA TN D-3152, December, 1965.)
[7] P. C. Paris, M. P. Gomez, and W. E. Anderson, "A Rational Analytic The-
ory of Fatigue," The Trend in Engineering, Vol 13, 1961, p. 9.
[8] P. C. Paris, "A Handbook of Crack Tip Stress Intensity Factors," Lehigh
University, Institute of Research, Bethlehem, Pa., June, 1960.
[9] J. Schijve and F. A. Jacobs, "Fatigue Crack Propagation in Unnotched and
Notched Aluminum Alloy Specimens," NLR-TR M. 2128, Nat. Aerospace
Laboratory, Amsterdam, May, 1964.
[10] P. C. Paris and F. Erdogan, "A Critical Analysis of Crack Propagation Laws,"
Transactions, Am. Soc. Mechanical Engrs., Series D, Vol 85, December, 1963,
p. 528.
[11] D. R. Donaldson and W. E. Anderson, "Crack Propagation Behaviour of
Some Airframe Materials," Proceedings Crack Propagation Symposium,
Cranfield, 1961, Vol II, p. 375. The College of Aeronautics, 1962.
[12] D. Broek and J. Schijve, 'The Influence of the Mean Stress on the Propaga-
tion of Fatigue Crack in Aluminum Alloy Sheet," NLR-TR M. 2111, Nat.
Aerospace Laboratory, Amsterdam, January, 1963. Also NLR-Report MP.
229, August, 1964.
[13] R. G. Forman, V. E. Kearney, and R. M. Engle, "Numerical Analysis of
Crack Propagation in Cyclic Loaded Structures," paper to be presented at the
1966 Annual Winter Meeting of the Am. Soc. Mechanical Engrs.
[14] J. Schijve, A. Nederveen, and F. A. Jacobs, "The Effect of the Sheet Width
on the Fatigue Crack Propagation in 2024-T3 Alclad Material," NLR-TR M.
2142, Nat. Aerospace Laboratory, Amsterdam, March 1965.
[75] G. R. Irwin, "Fracture," Encyclopaedia of Physics, Vol 6, p. 565, Springer-
Verlag, Berlin, 1958.

NLR-TR M.2122, Nat. Aerospace Laboratory, Amsterdam, April, 1964.
[17] R. Greif and J. L. Sanders, Jr., 'The Effect of a Stringer on the Stress in a
Cracked Sheet," ONR TR No. 17, Harvard University, Cambridge, Mass.,
June, 1963.
[18] J. M. Bloom, 'The Effect of a Riveted Stringer on the Stress in a Sheet with
Crack or a Cutout," ONR TR No. 20, Harvard University, Cambridge, Mass.,
June, 1964.
[79] R. T. Hunt, "Crack Propagation and Residual Static Strength of Stiffened
and Unstiffened Sheet," Current Aeronautical Fatigue Problems, Pergamon
Press, London, 1965, p. 287.
Structures," Advances in Aeronautical Sciences, Vol 3, p. 387, Pergamon
Press, London, 1962. Also NLR-Report MP. 195, Nat. Aerospace Labora-
tory, Amsterdam, August, 1960.
[21] J. Schijve, "Estimation of Fatigue Performance of Aircraft Structures,"
Fatigue of Aircraft Structures, ASTM STP 338, Am. Soc. Testing Mats., 1962,
p. 193. Also NLR-Report MP. 212, Nat. Aerospace Laboratory, Amsterdam,
June, 1962.
[22] J. Schijve and D. Broek, "Crack Propagation under Variable-Amplitude
Loading," Aircraft Engineering, Vol 34, 1962, p. 314. Also NLR-Report MP.
208, Nat. Aerospace Laboratory, Amsterdam, December, 1961.
[23] J. R. Dixon and J. S. Strannigan, "Effect of Plastic Deformation on the
Strain Distribution Around Cracks in Sheet Materials," Journal Mechanical
Engineering Science, Vol 6, 1964, p. 132.
[24] J. R. Dixon, "Stress and Strain Distributions Around Cracks in Sheet Mate-
rials Having Various Work-Hardening Characteristics," International Journal
of Fracture Mechanics, Vol 1, 1965, p. 224.
[25] D. Broek, "Crack Propagation Properties of 2024-T8 Sheet Under Static and
Dynamic Loads," NLR-Report M. 2161, Nat. Aerospace Laboratory, Amster-
dam, March, 1966.
[26] J. Schijve and P. de Rijk, "The Fatigue Crack Propagation in 2024-T3 Alclad
Sheet Materials from Seven Different Manufacturers," NLR-Report M. 2162,
Nat. Aerospace Laboratory, Amsterdam, May, 1966.
[27] D. E. Piper, W. E. Quist, and W. E. Anderson, 'The Effect of Composition
on the Fracture Properties of 7178-T6 Aluminum Alloy Sheet," paper pre-
sented at the Fall Meeting of the Metallurgical Society of Am. Institute
Mechanical Engrs., October, 1964.
[28] D. Broek and J. Schijve, "The Effect of Sheet Thickness on the Fatigue-
Crack Propagation in 2024-T3 Alclad Sheet Material," NLR-TR M. 2129,
April, 1963. Also NLR-Report MP, 230, Nat. Aerospace Laboratory, Am-
sterdam, November, 1964.
[29] K. D. Raithby and M. E. Bebb, "Propagation of Fatigue Cracks in Wide
Unstiffened Aluminum Alloy Sheets," RAE-TR Structures 305, Royal Air-
craft Est., September, 1961.
[30] J. Schijve and P. de Rijk, "The Effect of Temperature and Frequency on the
Fatigue Crack Propagation in 2024-T3 Alclad Sheet Material," NLR-TR M.
2138, Nat. Aerospace Laboratory, Amsterdam, January, 1965.
[31] J. Schijve and P. de Rijk, 'The Effect of Ground-to-Air Cycles on the Fa-
tigue Crack Propagation in 2024-T3 Alclad Sheet Material," NLR-TR M.
2148, Nat. Aerospace Laboratory, Amsterdam, July, 1965.
[32] R. Lachenaud and P. Jaillon, "Influence de la Frequence et de la Tempera-
ture sur la Vitesse de Propagation des Criques en Fatigue,"Sud-Aviation,
Proces-Verbal No. 26-712/4, Paris, September, 1964.
[33] J. Schijve and P. de Rijk, 'The Crack Propagation in Two Aluminum Alloys
in an Indoor and an Outdoor Environment under Random and Programmed
Load Sequences," NLR-Report M. 2156, Nat. Aerospace Laboratory, Amster-
dam, November, 1965.
[34] J. Schijve, D. Broek, P. de Rijk, A. Nederveen, and P. J. Sevenhuysen,
"Fatigue Tests with Random and Programmed Load Sequences, with and
without Ground-To-Air Cycles. A Comparative Study on Full-Scale Wing

Center Sections," NLR-Report S. 613, Nat. Aerospace Laboratory, Amster-
dam, December, 1965.
[35] W. A. Wood, "Recent Observations on Fatigue Failure in Metals," Basic
Mechanisms of Fatigue, ASTM STP 237, Am. Soc. Testing Mats., 1958,
p. 110.
[36] N. Thompson, N. Wadworth and N. Louat, "The Origin of Fatigue in Cop-
per," Philosophical Magazine, (8), Vol 1, 1956, p. 113.
[37] D. H. Avery and W. A. Backofen, "Nucleation and Growth of Fatigue
Cracks," Fracture of Solids, D. C. Drucker and J. J. Oilman, editors, Inter-
science Publishers, New York, 1963, p. 339.
[38] W. A. Backofen, "Formation of Slip-Band Cracks in Fatigue," Fracture,
John Wiley and Sons, Inc., New York, 1959, p. 435.
[39] N. Thompson, "Some Observations on the Early Stages of Fatigue Fracture,"
Fracture, John Wiley and Sons, Inc., New York, 1959, p. 354.
[40] P. J. E. Forsyth, "A Two Stage Process of Fatigue Crack Growth," Proceed-
ings Crack Propagation Symposium, Cranfield, Vol 1, p. 77. The College of
Aeronautics, 1961.
Aluminum-7.5% Zinc-2.5% Magnesium Alloy," Metallurgica, Vol 68,
1963, p. 109.
Acta Metallurgica, Vol 11, 1963, p. 703.
[43] A. J. McEvily and R. C. Boettner, "On Fatigue Crack Propagation in f.c.c.
Metals," Acta Metallurgica, Vol 11, 1963, p. 11.
[44] D. A. Ryder, "Some Quantitative Information Obtained from the Examina-
tion of Fatigue Fracture Surfaces," RAE TN Met. 288, Royal Aircraft Est.,
1958.
Philosophical Magazine, (8), Vol 7, 1962, p. 847.
[46] A. Matting and G. Jacoby, "Ueber das Verhalten von Schweissverbindungen
aus Aluminiumlegierungen bei Schwingbeanspruchung," Teil II: "Beitrag
zum Mechanismus des Schwingungsbruches," Aluminum, Vol 38, 1962, p.
309.
[47] R. H. Christensen, "Cracking and Fracture in Metals and Structures," Pro-
ceedings Crack Propagation Symposium, Cranfield, Vol 2, p. 326, The College
of Aeronautics, 1961.
chanics to Fatigue Crack Propagation," thesis, Lehigh University, Bethlehem,
Pa., May, 1965.
[49] M. S. Hunter and W. G. Fricke, "Fatigue Crack Propagation in Aluminum
Alloys," Proceedings, Am. Soc. Testing Mats., Vol 56, 1956, p. 1038.
[50] F. Girard, "Etude Microscopique de la Fissuration par Fatigue a Haute Fre-
quence," Advances in Aeronautical Sciences, Vol 3, Pergamon Press, Lon-
don, 1960, p. 357.
[51] E. Orowan, 'Theory of the Fatigue of Metals," Proceedings, Royal Soc., (A),
Vol 171, 1939, p. 79.
[52] A. K. Head, "The Growth of Fatigue Cracks," Philosophical Magazine, Vol
44, 1953, p. 925.
[55] H. A. Lipsitt, F. W. Forbes, and R. B. Baird, "Crack Propagation in Cold-
Rolled Aluminum Sheet," Proceedings, Am. Soc. Testing Mats., Vol 59, 1959,'
p. 734.
[54] Proceedings, Crack Propagation Symposium, Vols 1 and 2, Cranfield 1961,
papers of Forsyth, Hardrath and McEvily, Christensen, and Frost, Holden,
and Philips.
[55] L. F. Coffin and J. F. Tavernelli, 'The Cyclic Straining and Fatigue of
Metals," Transactions, Am. Institute Mining, Metallurgical, and Petroleum
Engrs., Vol 215, 1959, p. 794.
[56] J. Holden, "The Formation of Sub-Grain Structure by Alternating Plastic
Strain," NEL-Report PM 298, Nat. Engineering Laboratory, East Kilbride,

1960.
[57] J. C. Grosskreutz, "Research on the Mechanics of Fatigue," WADD-TR-60-
313, Part II, Research and Technology Div., Air Force Systems Command,
December, 1963.
[58] J. C. Grosskreutz, "Fatigue Crack Propagation in Aluminum Single Crystals,"
Journal Applied Physics, Vol 33, 1962, p. 1787.
[59] W. A. Wood, S. McK. Cousland, and K. R. Sargant, "Systematic Microstruc-
tural Changes Peculiar to Fatigue Deformation," Acta Metallurgica, Vol 11,
1963, p. 643.
[60] I. A. Oding, "Ueber den Mechanismus der Zerstorung bei der Zyklischen
Belastung von Metalen," 1UTAM Colloquium on Fatigue—Stockholm,
Springer Verlag, Berlin, 1955, p. 178.
[61] M. H. Raymond and L. F. Coffin, "Geometrical Effects in Strain Cycled
Aluminum," Journal Basic Engineering, Transaction, Am Soc. Mechanical
Engrs., Vol 85, Series D, 1963, p. 548.
[62] T. H. Alden and W. A. Backofen, "The Formation of Fatigue Cracks in
Aluminum Single Crystals," Acta Metallurgica, Vol 9, 1961, p. 352.
[63] L. H. Glassman and A. J. McEvily, "Effects of Constituents Particles on the
Notch-Sensitivity and Fatigue-Crack-Propagation Characteristics of Al-Zn-
Mg Alloys," NASA TN D-928, Nat. Aeronautics and Space Administration,
April, 1962.
[64] H. de Lairis, E. Mencarelli and J. Poulignier, "Contribution de la Micro-
Fractographie Electronique a FEtude de 1'Influence de la Frequence des
Cycles sur le Developpement des Cassures Fatigue dans les Alliages d'Alu-
minium," ON ERA, T.P., No. 117, Paris, 1964.
[65] R. M. N. Pelloux, "Fractographic Analysis of the Influence of Constituent
Am. Soc. Metals, Vol 57, June 1964, p. 511.
[66] R. G. DeLange, "Plastic Replica Methods Applied to a Study of Fatigue
Phenomena," Transactions, Metallurgical Soc., Am. Institute Mining, Metal-
lurgical, and Petroleum Engrs., Vol 230, 1964, p. 644.
[67] W. A. Wood, W. H. Reimann, and K. R. Sargant, "Comparison of Fatigue
Mechanisms in bcc and fee Metals," Transactions, Metallurgical Soc., Am.
Institute Mining, Metallurgical, and Petroleum Engrs., Vol 230, 1964, p. 511.
[68] J. A. Bennett, W. L. Holshouser, and H. P. Utech, "The Importance of En-
vironment in Fatigue Failure of Metals," Fatigue of Aircraft Structures,
W. Barrois and E. L. Ripley, editors, Pergamon Press, London, 1963, p. 1.
[69] A. Hartman, "On the Effect of Oxygen and Water Vapor on the Propagation
of Fatigue Cracks in 2024-T3 Alclad Sheet," International Journal of Frac-
ture Mechanics, Vol 1, 1965, p. 167. Also NLR-Report MP. 225, Nat. Aero-
space Laboratory, Amsterdam.
[70] F. J. Bradshaw and C. Wheeler, "The Effect of Environment on Fatigue
Crack Growth in Aluminum and Some Aluminum Alloys," Applied Mate-
rials Research, Vol 5, April 1966, p. 112.
[71] D. E. Martin, "Plastic Strain Fatigue in Air and Vacuum," Journal Basic
Engineering, Transactions, Am. Soc. Mechanical Engrs., Series D, Vol 87,
1965, p. 850.
[72] M. S. Hunter and W. G. Fricke, "Metallographic Aspects of Fatigue Behav-
ior of Aluminum," Proceedings, Am. Soc. Testing Mats., Vol 54, 1954, p. 717,
DISCUSSION
F. A. Smith} (written discussion)—Your work appears to show a varia-

tion in crack propagation for several heats of the same alloy. There is also
evidence to suggest that oxygen at the seat of the crack aids in crack propa-
gation. There appears a real need to understand crack propagation as re-
lated to the variables of temperature, environment, or reaction rates of the
specimens in the environment of test. I should appreciate the author's com-
ment on this subject.
7. Schijve (author)—If the environment has an influence on the crack
propagation one should expect that temperature and frequency would
also affect the propagation. Unfortunately, varying the temperature will
not only affect the environmental influence, but at the same time the
"resistance" of the material against cyclic slip will also be altered. The
best method for improving our understanding of the environmental effect
would therefore be to include the environment and the frequency as
variables in an experimental program. Some work along these lines is
now being performed at our laboratory on aluminum alloy sheet material.
It turned out that the influence of water vapor on the crack propagation
increased at lower frequencies. Although this result is probably not unex-
pected, it has to be admitted that our understanding of the environmental
effect is still of a speculative nature. More research on this issue is cer-
tainly worthwhile.
At the same time it cannot be denied that certain effects of frequency
and environment as mentioned in the paper have some direct bearing on
certain practical aircraft problems. The main reason for performing tests
such as reported in Tables 5 and 6 was to obtain a first estimate of the
magnitude of the effects of frequency and environment when comparing
laboratory conditions with more practical conditions.
S. R. Swanson2 (written discussion)—I should like to make just two
short comments on this excellent review paper. The first is that the delays
in crack growth exhibited in Fig. 10 of Dr. Schijve's paper correlate with
similar behavior in Fig. 10 of our paper,3 which showed the continuous
crack length trace with time for a typical Rayleigh random loading fa-
tigue test. In cases where a random burst of high load cycles occurred,
there was, immediately following, a cessation of crack growth. This can
be seen clearly in the specimen bursts of loading marked on our figure.
1
Physicist, Argonne National Laboratory, Argonne, 111.
3
Senior research engineer, The de Havilland Aircraft of Canada, Ltd., Downs-
view, Ont. Canada.
3
See p. 333.
DISCUSSION ON CRACKS IN MICRO- AND MACRO-RANGE 459
The second point is that, for the work referred to in Table 4 of Dr.
Schijve's paper we have noticed that the minimum sheet thickness consid-
ered had a greater crack growth rate and length of crack at transition than
the next (greater) thickness. This would indicate some support for the
possibility of an optimum thickness effect, at least at transition, similar to
that found in our work.
Mr. Schijve—The first comment of Dr. Swanson is probably related
to his Fig. 12 (instead of Fig. 10). I agree that the delays of crack growth
he found in his tests were probably caused by the same effect that we had
to such an exaggerated extent in our tests, namely, the introduction of
compressive residual stresses by a high tensile peak load.
Regarding Dr. Swanson's second point, let me say first that the data
in Table 4 were presented in the first place in view of their technical
significance. They are not pointing to an optimum thickness. Dr. Swan-
son's remark on the crack length and the crack growth rate at the moment
that the transition was completed is correct. Nevertheless, our plots of
crack rate versus crack length do not reveal an optimum thickness. If
there is such an optimum for our material it has to be at a very low
thickness, lower than 0.6 mm (0.024 in.).
R. P. Wei,1 P. M. Talda,1 and Che-Yu Li1
Fatigue-Crack Propagation in Some

Ultrahigh-Strength Steels
REFERENCE: R. P. Wei, P. M. Talda, and Che-Yu Li, "Fatigue Crack

Propagation in Some Ultrahigh-Strength Steels," Fatigue Crack Propaga-
tion, ASTM STP 415, Am. Soc. Testing Mats., 1967, p. 460.
ABSTRACT: Fatigue-crack propagation studies were carried out on some
ultrahigh-strength steels in controlled environments using the fracture-
mechanics approach. The steels investigated include a medium-carbon low-
alloy ultrahigh-strength steel tempered to two strength and toughness
levels, a 250-grade 18Ni maraging steel, and a 300-grade 18Ni steel. The
results indicate that, in the inert environment, the rates of fatigue-crack
propagation in these steels are about the same and show no obvious cor-
relation with other mechanical properties of the steels. The sensitivity of
the rates of fatigue-crack propagation to the presence of moisture in the
test environment shows rough correlation with plane-strain fracture tough-
ness for steels of similar composition. The role of water vapor in promot-
ing fatigue-crack propagation is considered. The results are also dis-
cussed in terms of existing theories and metallographic observations.
KEY WORDS: fatigue (materials), crack propagation steels, fracture
mechanics, maraging steels, nickel steels, water vapor, environmental
effects
The role of fracture toughness, (GJc or X"ic) in determining the size of

crack that would produce failure of a high-strength material (with yield-
strength/specific-weight ratio in excess of 700,000 in.) in a structure
under given loading conditions (critical crack size) has been established
by fracture mechanics analysis [I].2 However, cracks smaller than the
critical size for given loading conditions are able to grow under both
static loads and cyclic loads, and the rate of growth may be strongly in-
1
The authors were affiliated with the Applied Research Laboratory, United
States Steel Corp., Monroeville, Pa. at the time this research was conducted. R. P.
Wei is now with the Department of Mechanics, Lehigh University, Bethlehem,
Pa., P. M. Talda is now with the Research Laboratory, Aluminum Company of
America, New Kensington, Pa., and Che-Yu Li was on leave from the Department
of Materials Science and Engineering, Cornell University, Ithaca, N. Y. Mr. Wei
is a personal member of ASTM.
2
paper.
460
WEI ET AL ON PROPAGATION IN ULTRAHIGH-STRENGTH STEELS 461
fluenced by the surrounding environment. Thus, a knowledge of the fac-

tors that affect the growth of these subcritical cracks, or flaws, is needed
to properly assess the suitability of a given material for service.
The influence of environments on slow crack growth under static loads
in some ultrahigh-strength steels has been studied by several investigators
using a fracture mechanics approach (for example, Johnson and Willner
[2], Hancock and Johnson [3,4], Hanna et al [5], and Hanna and Steiger-
wald [6]). These results, in general, showed that delayed failures under
static loads are the result of slow crack growth induced by the action of
aggressive environments: hydrogen, water vapor, or aqueous solutions.
The sensitivity to these aggressive environments varied among the ma-
terials investigated. In fact, the results of Hancock and Johnson [4]
showed that aggressive environments had negligible effect on crack growth
in an AM-350 steel (a precipitation-hardened stainless steel) and indi-
cated that subcritical-crack growth under constant static loads can also
occur in the absence of aggressive environments. These results serve to
point out the importance of controlling the test environments in any
crack-growth studies.
Although the influence of environments on the fatigue life of "smooth"
metal specimens has been studied by several investigators (for example,
Wadsworth and Hutching [7], Wadsworth [8], Bennett [9]), most of the
studies on fatigue-crack propagation (see Paris [10] and Krafft [11,12])
were not carried out under controlled environments. Recently, Dahlberg
[13] and Li et al [14] showed that water vapor has a significant effect on
the rate of fatigue-crack propagation in some medium-carbon, low-alloy
ultrahigh-strength steels. Hartman [15] showed that water vapor also
has a strong effect on the rate of fatigue-crack propagation in a 2024-T3
(Alclad) aluminum alloy. Li et al [14] showed that the sensitivity of the
rate of fatigue-crack propagation to an aggressive environment appears
to be related to the fracture toughness of the material. In view of these
recent experimental results, it was considered desirable to carry out addi-
tional fatigue-crack propagation studies under controlled environments,
on materials with different compositions, microstructures, and properties.
The aims of these additional studies were: (1) to assess the effects of
these variables on the rate of fatigue-crack propagation, and on the sus-
ceptibility of the material to aggressive environments; and (2) to develop,
under controlled environments, fatigue-crack-propagation data that can
be used for a more meaningful comparison with several current "theories"
of fatigue [11,12,16,17].
In this paper the results of a study of fatigue-crack propagation in con-
trolled gaseous environments for two currently important classes of ultra-
high-strength steels (quenched and tempered, and maraging steels) are
presented. The effects of composition, fracture toughness, microstructure,
and moisture on the rate of fatigue-crack propagation are considered.
Comparison of these results with the proposed correlations of Krafft [11,

12], Weertman [16], and McEvily and Johnston [17] is discussed.
Material and Experimental Procedures
Experimental, vacuum-melted 18Ni (250) and 18Ni (300) mar aging
steels were selected for this study to complement data already reported
by Li et al [14] on an experimental, vacuum-arc-remelted 0.45C-Ni-
Cr-Mo steel. Pertinent data from Li et al [14] are included for the sake of
completeness. Chemical composition, heat-treatment, and mechanical
properties for all three steels are given in Tables 1, 2, and 3, respectively.
These steels represent microstructures associated with two principal
strengthening mechanisms used in ultrahigh-strength steels: a tempered
TABLE 1—Chemical composition of steels investigated.
Material c Mn P s Si Ni
0.45C-Ni-Cr-Mo
steel 0.45 0.26 0.007 0.016 0.25 2.04
18Ni (250) maraging
steel 0.003 0.002 0.002 0.004 0.017 17.8
18Ni (300) maraging
steel 0.007 <0.02 0.001 0.001 0.010 17.6
Material Cr Mo Co Ti Al N o
0.45C-Ni-Cr-Mo
steel 1.49 0.44 0.029 0.009 0.001
18Ni (250) maraging
steel 4.95 7.60 0.45 0.055 0.002 0.002
18Ni (300) maraging
steel 4.93 9.01 1.33 0.095 0.002 0.004
carbon martensite structure and a precipitation-hardened nickel mar-

tensite structure. The microstructures of these two important classes of
ultrahigh-strength steels have been examined in detail by Baker et al
[18] and Baker and Reisdorf [79]. The levels of plane-strain fracture
toughness, Kle, for the steels investigated, ranging from about 40 to 110
ksi \/in. [20], are representative of the range of ATIO levels for commercial
ultrahigh-strength steels.
Centrally notched specimens, Fig. 1, were used for the crack-propaga-
tion studies. The initial center notch (about 0.02 in. wide) was introduced
by spark machining in each specimen. The dimensions of the specimens
and the initial notch lengths are given in Table 4. Fatigue precracking
and the fatigue-crack-propagation experiments were conducted at room
temperature in an Amsler High Frequency Vibrophore under zero-to-
tension loading, with a maximum gross section stress of about 20,000
psi (the actual stress level varied slightly among the different materials).
Operating frequencies and ranges of stress-intensity parameters (AK)
TABLE 2—Processing and heat treatment.

Material Processing Heat Treatment
0.45C-Ni-Cr-Mo
steel 13/^-in. -thick plate hot- normalization for 45 min at
rolled, straightaway, to 1650 F in a neutral, pro-
0.215 in.; finished to 0.125- tective atmosphere; cool-
in.-thick specimens by- ing in air; austenitization
surface grinding after for 45 min at 1550 F in a
heat treatment neutral, protective atmos-
phere, quenching in oil;
tempering for 1 hr at 400
or 800 F; and cooling in
air
18Ni (250) maraging
steel 4-in.-thick by 12-in. by 18- solution annealing for 1 hr
in. ingot hot-rolled, at 1700 F in a neutral, pro-
straightaway, to 0.35 in.; tective atmosphere; cool-
finished to 0.250-in.-thick ing in air; aging for 3 hr
specimens by surface at 900 F, and cooling in air
grinding after heat treat-
ment
18Ni (300) maraging
steel 1-in.-thick plate hot-rolled, solution annealing for 1 hr
straightaway, to 0.20 in.; at 1650 F in a neutral,
finished to 0.125-in.-thick protective atmosphere;
specimens by surface quenching in water; solu-
grinding after heat treat- tion annealing for 1 hr at
ment 1550 F in a neutral, protec-
tive atmosphere; quench-
ing in water, aging for 3 hr
at 900 F, and cooling in air
TABLE 3 —Mechanical properties of steels investigated.

0.2% Offset Tensile Engineering
Material Yield Strain at S train-Hardening Klc,
Strength, Strength,
ksi Maximum Exponent, » ksi \/in.
E, psi
ksi Load, tmax
0.45C-Ni-Cr-Mo
steel (tempered
for 1 hr at
400 F) 208 312 0.08" 0.08 ± 0.01 37fc 29 X 106
0.45C-Ni-Cr-Mo
steel (tempered
for 1 hr at
800 F) 215 234 0.06° 0.06 ± 0.01 56" 29 X 106
18Ni (250) marag-
ing steel 246 257 0.015° 0.015 ± 0.005 110" 27 X 106
18Ni (300) marag-
ing steel 305 315 0.015° 0.015 ± 0.005 53* 27 X 106
° Based on strain-hardening exponent «.

6
From 1-in.-diameter notched round specimens [20].
c
From slow-bend test (0.7 by 1 J£ by 7-in. specimens) data of a similar heat of
steel.
d
From % by 2 by 8-in. center-notched specimens.
also are given in Table 4. Each specimen was precracked in air to pro-
vide a fatigue crack about 0.03 in. in length from the ends of the starter
notch. An environment chamber similar to that used by Johnson and
Willner [2] was then clamped in place. The environmental gas was
started through the system about 20 min before the actual test run to
purge the environment system. At intervals of 104 cycles, fatigue cycling
was interrupted and crack-length measurements were made with an
electrical potential method, the specimen being under the static mean
load. The detailed experimental procedure and the calibration of this
method have been described elsewhere [2,21,22].
The reference (dry) environment was obtained by passing Matheson
FIG. 1—Schematic view cf experimental setup.
"research grade" argon (99.9995 per cent purity) through a dehumidifier

(Matheson Model 460 purifier with Model 461-R cartridge for moisture).
The residual impurity in this gas consisted of 5 ppm of nitrogen accord-
ing to tank analysis supplied by Matheson. The humid environment
consisted of Matheson "pre-purified grade" argon saturated with
moisture.
Results and Discussions
Several methods for analyzing the experimental fatigue-crack-propaga-
tion data were tried. It was found that the following modified form of the
empirical relationship suggested by Paris and Erdogen [23] was most
convenient to use, and gave the best fit to most of the test data.
TABLE 4—Specimen dimensions and other pertinent information.

Nominal Specimen Dimension, Initial Notch AK, Test
Material thick (B) X width (W) X Length, f,
ksi rv/Ba Speed, cps
length (L), in. 2 a0 , in. ksi \/m.
0.45C-Ni-Cr-Mo steel (tem-

pered for 1 hr at 400 F) 0.125 X 2 X 8 0.24 22.0 ~15 to 30 0.0065 to 0.026 125
0.45C-Ni-Cr-Mo steel (tem-
pered for 1 hr at 800 F) 0.125 X 2X8 0.24 21.0 —15 to 30 0.0062 to 0.025 125
(0.125 X 1.75 X 8) b (0.24)6 (20. 0)6 (~14 to 28)6 (0.0054 to 0.022)b
18Ni (250) maraging steel 0.250 X 2.75 X 12 0.40 18.3 ~15 to 35 0.0024 to 0.013 150
18Ni (300) maraging steel 0.125 X 2X 8 0.24 19.5 -15 to 30 0.0031 to 0.012 125
0
ry/B = (AA"/ffYs)2/2ir5, where o-ys = 0.2 per cent offset yield strength.
* Tested in dehumidified research-grade argon.
where :
c = a-f(a/W] = equivalent half-crack length [24,25],
a = actual half-crack length,
N = number of cycles,
AAT = Ao- \STTC = range of crack-tip stress-intensity parameter, and
A = proportionality constant (or rate constant) .
FIG. 2—Typical results showing the effect of moisture on fatigue-crack growth

in an 18Ni (250) maraging steel.
Since the range of gross section stress Ao- was maintained constant,
Eq 1 may be simply integrated to give
where c0 is the initial equivalent half-crack length. The rate constant A

was then obtained from a least-square (straight-line) regression analysis
of the experimental data (\/c0 — l/c versus N). The magnitude of A
represents the ease of fatigue-crack propagation in the material; the
ratio of A values for the different tests indicates the relative sensitivity
of the rate of fatigue-crack propagation to various environments and to
different properties of the material.
Typical fatigue-crack-propagation data on the 18Ni maraging steels
are presented in Figs. 2 and 3 as plots of l/c 0 — l/c versus JV. Data on
the 0.45C-Ni-Cr-Mo steel for the same test environments from Li et al
FIG. 3—Typical results showing the effects of moisture on fatigue-crack growth

in an 18Ni (300) mar aging steel.
FIG. A—Typical results showing the effect of moisture on fatigue-crack growth

in an 0.45C-Ni-Cr-Mo steel tempered for 1 hr at 400 F.
[14] are shown in Figs. 4 and 5.3 These data cover a range of stress-
intensity parameters (AK) from about 15 to 35 ksi -\/in., with corre-
3
Data on the 0.45C-Ni-Cr-Mo steel (tempered at 800 F) tested in dehumidified
"research grade" argon are new and were not reported in Ref 14. Specimen width
was 1.75 in.
spending crack-propagation rates ranging from 6 X 10~7 to 2 X 10~5

in./cycle. For this range of AK, the ratios of crack-tip plastic-zone
correction factor to specimen thickness, ry/B, were less than %Q, Table
4. Thus, these data represent fatigue-crack propagation under essen-
tially plane-strain conditions. No plastic-zone corrections were incor-
FIG. 5—Typical results showing the effect of moisture on fatigue-crack growth

in an 0.45C-Ni-Cr-Mo steel tempered for 1 hr at 800 F.
TABLE 5—Coefficient A [in dc/dN = A(AK) 4 ] for the steels investigated.

/Tic, 0 A X 1023, in. 7 Itr4 cycle-i
Material
ksi -\/in. Dry Argon Humid Argon
0.45C-Ni-Cr-Mo steel (tempered 1 hr—

400 F) 37 1.3 2.0
800 F) 56 1.6 1.7
18Ni (250) maraging steel 110 1.6 1.6
18Ni (300) maraging steel 53 1.6 2.0
0
See Table 3.
porated in the crack lengths shown in Figs. 2 to 5. The error in calculating

fatigue-crack-propagation rates from this source is less than 1 per cent.
The values of the constant A (based on data from at least two test
specimens) are tabulated in Table 5. The typical reproducibility of data
in these experiments is illustrated in Fig. 6 by experimental results from
a set of duplicate test specimens. Scatter in growth-rate measurements
(in terms of A) is less than 10 per cent. The absolute accuracy of A,
determined by the combined accuracies in load and crack-length meas-
urements, is estimated to be ±10 per cent. These results (Figs. 2 to 5,

and Table 5) show that there is no obvious correlation between the
plane-strain fracture toughness parameter KIC and the rates of fatigue-
crack propagation in an inert environment. However, the rates of fatigue-
crack propagation in the aggressive (humid) environment showed an
inverse relationship with KIC values; the steel with the lowest value of
KIC showed the highest sensitivity to environment-induced crack growth.
In the following discussion, it is necessary to refer the present data
(based on dc/dN] to da/dN, the growth rate based on actual crack length.
The conversion may be made most conveniently, with sufficient accuracy,
FIG. 6—Typical reproducibility of experimental data as shown by fatigue-

crack-growth results from duplicate test specimens.
by using Irwin's [7] tangent correction for a central crack of length la

in a plate of finite width W. Thus, Eq 1 becomes
where a is the fatigue-growth-rate coefficient corresponding to da/dN.

It should be noted that since a. is dependent on the ratio a/ W, compari-
sons on the basis of da/dN for the different steels and test environments
should be made only at some specific value of a/ W. The above expression
was used for converting the present results on 18N1 maraging steels to
permit comparison with the data of Carman and Katlin [26]. For the
comparisons with current theories of fatigue-crack propagation, an
average value of a = A cos2 ira/W appropriate for the range of a/W
of the present data is used. With an a/W value approximately at mid-

range of the experimental data (a/W = >£), the average value of a is
given by
Comparison with Results of Carman and Katlin

Carman and Katlin [26] reported fatigue-crack-propagation data on
34-in.-thick 18Ni (250) and 18Ni (300) maraging steels tested in "room"
air at 250 cpm or about 4 cps. The yield strength, tensile strength, and
FIG. 7—Comparison of crack-propagation rates from Carman and Katlin [26]

with present experiments on 18Ni (250) maraging steels.
Klc were 252 ksi, 266 ksi, and 73.6 ksi \/\n., respectively, for the 18Ni
(250) maraging steel, and 295 ksi, 303 ksi, and 43.2 ksi \/m., respec-
tively, for the 18N1 (300) maraging steel. Since the present results indi-
cate that the rate of fatigue-crack propagation does not depend strongly
on the values of Klc, a direct comparison with the data of Carman and
Katlin [26] can be made. Test results obtained in the humid argon
environment are used for this comparison because Li et al [14] showed
that these results would be comparable to those obtained in room air.
Figures 7 and 8 show fatigue-crack-propagation rate data computed by
Eq 3 (from A values given in Table 5) on the 18Ni (250) and 18Ni (300)
maraging steels in comparison with those of Carman and Katlin [26].
It is apparent that the present results (with da/dN from about 6 X 10~7
471
to 10~5 in. per cycle) merge well with the data of Carman and Katlin
[26] for the higher rates of growth. Furthermore, the fatigue-crack-
propagation rate (about 7 X 10~6_m. per cycle) for the 18Ni (250)
maraging steel at AK « 30 ksi \/in. is in good agreement with that
reported by Schuler and Carman [27] who used a fractographic tech-
nique. This agreement indicates that cycling speeds in the range of 4 cps
(Carman and Katlin [26]) to 150 cps (present experiment) have little
influence on the rate of fatigue-crack propagation.
FIG. 8—Comparison of crack-propagation rates from Carman and Katlin [26]

with present experiments on 18Ni (300) maraging steels.
The Effect of Test Environments

It is evident from Table 5 that the sensitivity of the rate of fatigue-
crack propagation to water vapor varies among the steels tested. This
environmental sensitivity can be roughly correlated with the plane-
strain fracture toughness, Klc, for materials having similar chemical
compositions. For the 0.45C-Ni-Cr-Mo steel tempered at 400 and 800 F
with respective Klc levels of 37 and 56 ksi \/in., the rates of fatigue-
crack propagation in an environment of argon saturated with water
vapor were, respectively, 50 and 10 per cent above those in a dry argon
environment [14]. For the 18Ni (300) maraging steel having a KIC of
53 ksi \/in., the increase in growth rate was about 30 per cent; whereas
the 18Ni (250) maraging steel (Klc =110 ksi \/m.) showed little sensi-
tivity to the moist environment. These results are consistent with the
results of Hanna and Steigerwald [6] for crack growth under static
loads on some 0.40 per cent carbon low-alloy ultrahigh-strength steels
and 18Ni maraging steels tested in an aqueous environment.
The precise mechanism by which water vapor affects crack growth
is not known. However, some qualitative observations can be made on
the basis of the results obtained from these experiments. It is tempting
to speculate that the increased rate of fatigue-crack propagation is the
result of hydrogen produced by the reaction of water vapor with the
freshly created crack surfaces. The hydrogen thus produced may enhance
crack growth by means of the several mechanisms already suggested for
hydrogen embrittlement [3,5,28] or by some other mechanism that may
operate in the zone of material immediately ahead of the crack tip. The
fact that this reaction does occur is strongly suggested by the fractographic
results of Spitzig4 which showed that larger amounts of oxide were
formed on the fracture surfaces of specimens tested in the moist environ-
ment than were found on specimens tested in the inert environment. How-
ever, the possibility that the enhancement of crack growth may be, in part,
a consequence of the water-metal surface reaction itself cannot be dis-
counted. Hartman [75] observed that both water vapor and dry oxygen
affect fatigue-crack propagation in a 2024-T3 (Alclad) aluminum alloy,
and concluded that these effects could be explained on the basis of a
surface reaction between the environment and the crack surface. Further
support of the surface-reaction concept is given by unpublished results
of the authors on fatigue-crack propagation of the 0.45C-Ni-Cr-Mo steel
(tempered for 1 hr at 400 F) in an atmosphere of bromine vapor and
dry argon. The magnitude of increase in the rate of fatigue-crack propaga-
tion produced by bromine vapor is similar to that produced by water
vapor and seems to depend on the concentration of bromine vapor in
the system. Since both hydrogen and hydrogen-producing compounds
are absent from this environment, the only possible explanation of the
observed bromine effect is that of surface reaction. Undoubtedly both
the effect of hydrogen and the influence of water-metal surface reaction
could be important. The efficacy of each in promoting crack growth
would likely depend on the microstructure of the material.
The observed differences in sensitivity to moist environment between
the 0.45C-Ni-Cr-Mo steel tempered at 400 F and at 800 F and between
the two maraging steels suggest this microstructural dependence. This
effect of microstructure may be considered in terms of surface reaction,
as well as hydrogen embrittlement. From the point of view that the en-
hancement of crack growth is the result of the water-metal surface reac-
tion itself, the difference in chemical activity between the 0.45C-Ni-Cr-Mo
* Unpublished work conducted at the Applied Research Laboratory, U.S. Steel

Corp., Monroeville, Pa.
steel tempered at 400 F and at 800 F, and between the two maraging steels
should be small and would not account for the observed differences in
sensitivity. However, the microstructures of the steels are different. For
the 0.45C-Ni-Cr-Mo steels, Baker et al [18] have shown the types of tem-
pered carbon martensite that are developed in this class of steels by dif-
ferent tempering treatments. For the 400 F tempering treatment, the
martensite has a predominantly platelike morphology, with a high density
of dislocations and many microtwins in the martensite plates. There is
an almost continuous film of e-carbide at the martensite and twin bound-
aries, providing preferred paths for crack growth. Upon tempering at
800 F, a tougher microstructure is obtained through the reduction of
embrittling carbide films at the boundaries by spheroidization (to form
iron carbide (Fe3C) cementite particles) and the redistribution and re-
moval of lattice defect structure by recovery processes. Baker et al [18]
suggested that the low levels of îc exhibited by these steels at the lower
tempering temperatures can be attributed to the presence of the brittle
carbide films at the boundaries. Thermodynamically, these brittle bound-
aries (for the 400 F tempering treatment) are high-energy sites and
should be more reactive to water vapor. This may be the reason that this
0.45C-Ni-Cr-Mo steel tempered at 400 F is more susceptible to the effect
of water vapor than the same steel tempered at 800 F. For the two pre-
cipitation-hardened maraging steels, Baker and Reisdorf [79] have studied
the precipitation structure in relation to the mechanisms of strengthening
in the 18Ni (250) and 18Ni (300) maraging steels. However, the de-
pendence of Klc on the microstructures in these steels is not clear. Even
though the microstructural differences are less obvious, it is possible
that their sensitivity to environment, as shown by the experimental data,
Table 5, reflects differences in structure. From the viewpoint of hydrogen
embrittlement, on the other hand, it is reasonable to expect that hydrogen
would operate in regions of high stress concentration [3,29]. Since the
stress field around second-phase particles, and near internal boundaries,
could attain very high values [30], it is likely that the influence of hydro-
gen on crack propagation will also depend on the microstructure of a
material.
Recent experimental evidence suggests that subcritical-crack growth
under static loads may be controlled by thermally activated processes.
For example, Johnson and Willner [2] reported that the rate of subcritical-
crack growth of H-ll steel under static loads in an aqueous environment
depends on the test temperature; Li et al [31] observed subcritical-crack
growth in several ultrahigh-strength steels under static loads in the ab-
sence of aggressive environments, the crack-growth behavior being similar
to (thermally activated) creep deformation. It is quite possible that fatigue-
crack propagation, especially with the influence of aggressive environ-
ments, is also controlled by thermally activated processes. Since the exact

mechanisms for fatigue-crack propagation is not known, further experi-
mental work to test this type of hypothesis seems worth while.
Comparison with Current Theories

Several theories [11,12,16,17] have been proposed recently that relate
the rates of fatigue-crack propagation to the crack-tip stress-intensity
parameter K (or the range A^Q and to certain properties of the material.
A comparison of these theories with the present fatigue-crack-propagation
data (obtained in controlled environments) on a series of ultrahigh-strength
steels with different properties seems worth while. Since none of these
theories specifically account for environmental influences, only data ob-
tained in the inert dry argon environment will be used in the following
comparisons.
Krafft [11,12] proposed the following relationship for the rate of fatigue-
crack propagation, on the basis of an empirical correlation between the
plane-strain fracture toughness and the plastic-flow property of a mate-
rial, and on the a priori assumption of a fourth-power dependence on K,
where :
a = actual half-crack length in inches,
N = number of cycles,
n = strain-hardening exponent,
E = Young's modulus in psi,
&K = range of crack-tip stress-intensity factor,
-Kmax = maximum value of K in one cycle,
Klc = critical crack-tip stress-intensity factor (plane-strain), and
dT = (Klc/(Eri)Y/2ir = process-zone size [32].
This relationship is based on the further assumption that fatigue-crack
growth proceeds according to the same model as that proposed for the
onset of plane-strain fracture instability under a monotonically increas-
ing load. This model [32] states that the onset of plane-strain fracture
instability is determined by the necking and tensile rupture of small
elemental fracture "cells" (with fracture-cell size dF tt dT] lying along
the crack front. For convenience in comparing Krafft's relationship
with the present experimental results (for zero-to-tension loading where
AAT = Km&x) , Eq 5 may be rewritten as
and
Table 6 shows a comparison between the aK values calculated on the

basis of data given in Table 3 and the measured aavg values (Eq 4) for
the steels tested in dry argon environment. Although the agreement
between the aK and aavg is better than a factor of four, some of the experi-
mental data are not consistent with the predictions from Krafft's rela-
tionship: (a) For the 0.45C-Ni-Cr-Mo steel, Eq 5 predicts that the rate
of fatigue-crack propagation in the 800 F tempered material should be
40 per cent lower than in the 400 F tempered material, whereas the
experimental data indicate a 20 per cent higher rate, (b) For the two
maraging steels, with the same value of strain-hardening exponent n,
the predicted Ki<r2 dependence of crack-propagation rate was not
observed, (c) Similarly, the dependence of crack-propagation rate on
n~l was not observed for the 18Ni (300) maraging steel and the 0.45C-
Ni-Cr-Mo steel tempered at 800 F, with similar levels of KIC .
The general applicability of the model proposed by Krafft for onset
of plane-strain fracture instability [32] to all types of fractures has been
questioned [33]. In light of the above observations, the extension of
this model to fatigue-crack growth is also subject to question.
Weertman [16] derived an expression for the rate of fatigue-crack
propagation as an extension of the Bilby, Cottrell, and Swinden dis-
location model of a crack [34]. Bilby et al. considered an idealized
model whereby the displacement at and near the tip of a stationary
crack can be calculated from a continuous two-dimensional distribution
of dislocations. Bilby et al [34] assumed that unstable fracturing will
occur when the displacement at the crack tip reaches some critical
value D*, and showed that the critical stress ac for unstable fracturing
is given by a relationship of the same form as that proposed by Irwin
[1] and Orowan [35], where o-0D* is identified as a plastic-work term,
(<r0 = yield strength).
Weertman [76] assumed that this critical displacement criterion for

failure under monotonically increasing (static) loads can be directly
applied to fatigue, and that the displacements calculated from disloca-
tion distributions for the stationary crack can be simply added to ap-
proximate the accumulation of displacements ahead of a running
fatigue crack. On the basis of these assumptions, the following expression
for the rate of fatigue-crack propagation for applied stresses below yield
stress was derived [16],
Since zero-to-tension loading is considered, a4a2 in Eq 8 may be replaced

by (AAT)4. The various constants in Eq 8 may be regrouped and defined
TABLE 6—Comparison of experimental results (in dry argon) with
the prediction of Krafft [11,12].
2
«av,g X 10 » (present «K X7 102'4 (Krafft),
Material re suits), in.*1 lb-« in. lb~ cycle"1
cycle"

400 F) 1.0 0.60
800 F) 1.2 0.35
18Ni (250) maraging steel 1.2 0.45
18Ni (300) maraging steel 1.2 1.9
TABLE 7—Comparison of experimental results (in dry argon) with theories of

Weertman [16] and McEvily and Johnston [17].
Present Results Weertman McEvily and
Johnston,
Matenal «avg X 10^3,
in.' lb-< Gio , in. rip , in. . <roD*, L, in.
cycle-i •lb/in.2 in. -Ib/in.2
0.45C-Ni-Cr-Mo steel
(tempered 1 hr—
400 F) 1.0 43 0.0036 5100 0.25
0.45C-Ni-Cr-Mo steel
(tempered 1 hr—
800 F) 1.2 99 0.0076 7500 0.56
18Ni (250) maraging
steel 1.2 410 0.022 6700 1.8
18Ni (300) maraging
steel 1.2 95 0.0034 4500 0.97
by «w for the purpose of comparison with the experimental data on the

ultrahigh-strength steels.
Weertman [16] suggested that <TO be identified with the ultimate strength
of the material. However, it is not clear what values for (r0D* should be
used. Thus, a direct comparison with the present experimental results
is not possible. On the other hand, for the ultrahigh-strength steels
tested in this work, a0D* must correspond to the value of Gc appropriate
to the state of stress at the tip of the advancing fatigue crack (see Eq 7).
Since the crack-propagation rates in this work are obtained under
essentially plane-strain conditions, <r0D* values computed from Eq 9
(by setting aw = aavg) should be compared with the (?ic of each steel,
Table 7. It is apparent that the <r0D* values needed to account for the
observed crack-growth rates are one to two orders of magnitude too

large in comparison with the Gjc values. This suggests that fatigue-crack
growth may be a more complicated phenomenon, especially with the
existence of grain boundaries and second-phase particles in the material,
and that it may be inappropriate to apply a failure criterion for static
loads directly to the case of crack growth under cyclic loading.
McEvily and Johnston [17] modified the expression derived by Weert-
man [76] by replacing a0D* in Eq 8 with more familiar engineering
properties of the material. The rationale was that <r0D* should be pro-
portional to the area under the engineering tensile stress-strain curve
up to the onset of plastic instability or necking. The rate of fatigue-
crack propagation is therefore proportional to (A7C)4 and a correlation
parameter (see Eq 8),
where:
OYS = 0.2 per cent offset yield strength,
o-TS = ultimate tensile strength, and
e™ = engineering strain at maximum load.
For consistency, a parameter L having the dimension of length is intro-
duced here. L, which was not included in the original analysis by McEvily
and Johnston [77], should correspond to the dimension of the plastic
zone at the onset of unstable fracturing for the ultrahigh-strength steels
considered here. Thus, a M j may be defined in a manner as before:
By setting « M j = «avg again, the dimensions of L required to account

for the observed crack-growth data are computed from Eq 11 (based
on data given in Table 3) and are compared with the plane-strain plastic-
zone sizes [36]
in Table 7. The dimensions of L are much larger than the corresponding

plastic-zone size rlp . This is, of course, a reflection of the inadequacies
of the original Weertman analysis and no further comments will be
necessary.
Summary
The results of this study indicate that the rates of fatigue-crack propaga-
tion for these ultrahigh-strength steels tested in an inert environment are
about the same and are relatively insensitive to strength, strain-hardening
exponent, and fracture toughness. However, the rates of fatigue-crack

propagation for these steels tested in aggressive environments appear to
depend sensitively on their fracture toughness, KIC and chemical composi-
tion. The environmental sensitivity for steels with similar chemical com-
position appears to be related to differences in microstructure. Because
of this sensitivity to aggressive environments (including moisture normally
present in air), environmental control is very important, particularly if
the experimental data are intended for use in correlating with theories of
fatigue-crack propagation.
Data based on the limited range of materials tested in this experiment
show that the current theories of fatigue-crack propagation are not satis-
factory for general use in predicting crack-growth rates. Refinements and
further development in the theories of fatigue-crack propagation should
include, certainly, the influence of environments.
Acknowledgment
The authors are grateful for the helpful discussions with J. C. M. Li of
the Edgar C. Bain Laboratory for Fundamental Research, U.S. Steel Corp.
References
[1] G. R. Irwin, "Fracture Mechanics," Structural Mechanics, Pergamon Press,
New York, 1960, p. 557.
[2] H. H. Johnson and A. M. Willner, "Moisture and Stable Crack Growth in
a High Strength Steel," Applied Materials Research, Vol 4 January, 1965,
p. 34.
[3] G. G. Hancock and H. H. Johnson, "Hydrogen, Oxygen, and Sub-Critical
Crack Growth in a High Strength Steel," Transactions, Metallurgical Soc.,
Am. Institute Mining, Metallurgical, & Petroleum Engrs., April, 1966.
[4] G. G. Hancock and H. H. Johnson, "Subcritical Crack Growth in AM-350
Steel," Materials Research & Standards, Vol 6, No. 9, September, 1966, p.
431.
[5] G. L. Hanna, A. R. Troiano, and E. A. Steigerwald, "A Mechanism for the
Embrittlement of High Strength Steels by Aqueous Environments," Trans-
actions, Am. Society Metals, Vol 57, 1964, p. 658.
[6] G. L. Hanna and E. A. Steigerwald, "Influence of Environment on Crack
Propagation and Delayed Failures in High-Strength Steels," Technical Docu-
mentary Report RTD-TDR-63-4225, Air Force Materials Laboratory, Janu-
ary, 1964.
[7] N. J. Wadsworth and H. Hutching, "The Effect of Atmospheric Corrosion
on Metal Fatigue," Philosophical Magazine, Vol 3, 1958, p. 1154.
[8] N. J. Wadsworth, "The Effect of Environment on Metal Fatigue," Internal
Stresses and Fatigue in Metals, Elsevier Publishing Co., Amsterdam & New
York, 1959, p. 382.
[9] J. A. Bennett, "Effect of Reactions with the Atmosphere During Fatigue of
Metals," Fatigue—An Interdisciplinary Approach, Syracuse University Press,
Syracuse, 1964, p. 209.
Interdisciplinary Approach, Syracuse University Press, Syracuse, 1964, p.
107.
[11] J. M. Krafft, "On Prediction of Fatigue Crack Propagation Rate from
Fracture Toughness and Plastic Flow Properties," Transactions, Am. Society
Metals, Vol 58, 1965, p. 691.
[12] J. M. Krafft (U. S. Naval Research Laboratory), private communication,
1966.
[13] E. P. Dahlberg, "Fatigue-Crack Propagation in High-Strength 4340 Steel in

Humid Air," Transactions, Am. Society Metals, Vol 58, 1965, p. 46.
[14] Che-Yu Li, P. M. Talda, and R. P. Wei, "The Effect of Environments on
Fatigue-Crack Propagation in an Ultra-High-Strength Steel," International
Journal of Fracture Mechanics, 1967.
[75] A. Hartman, "On the Effect of Oxygen and Water Vapor on the Propaga-
tion of Fatigue Cracks in 2024-T3 Alclad Sheet," International Journal of
Fracture Mechanics, Vol 1, No. 3, September, 1965, p. 167.
Theory of Infinitesimal Dislocations Distributed on a Plane," paper presented
at the International Conference on Fracture, Sendai, Japan, Sept. 12-17, 1965.
[77] A. J. McEvily and T. L. Johnston, "On the Role of Cross-Slip in Brittle
Fracture and Fatigue," paper presented at the International Conference on
Fracture, Sendai, Japan, Sept. 12-17, 1965.
[75] A. J. Baker, F. J. Lauta, and R. P. Wei, "Relationships Between Micro-
structure and Toughness in Quenched and Tempered Ultra-High-Strength
Steels," Structure and Properties of Ultrahigh-Strength Steels, ASTM STP
370, Am. Soc. Testing Mats., 1965, p. 3.
[79] A. J. Baker and B. G. Reisdorf, "Strengthening Mechanisms of the 18 Per-
cent Nickel Maraging Steels," 4th Maraging Steel Project Review, Technical
Documentary Report ML-TDR-64-225, Air Force Materials Laboratory,
1964.
[20] R. P. Wei, "Fracture Toughness Testing in Alloy Development," Fracture
Toughness Testing and Its Applications, ASTM STP 381, Am. Soc. Testing
Mats., 1965, p. 279.
[27] H. H. Johnson, "Calibrating the Electrical Potential Method for Studying
Slow Crack Growth," Materials Research & Standards, Vol 5, No. 9, Sep-
tember, 1965, p. 442.
[22] Che-Yu Li and R. P. Wei, "Calibrating the Electrical Potential Method for
Studying Slow Crack Growth," Materials Research & Standards, Vol 6, No.
8, August, 1966, p. 392.
[23] P. C. Paris and F. Erdogen, "A Critical Analysis of Crack Propagation
Laws," Journal of Basic Engineering, Am. Society Mechanical Engrs., Decem-
ber, 1963.
[24] M. Isida and Yoshio Itagaki, "Stress Concentration at the Tip" of a Trans-
verse Crack in a Stiffened Plate Subjected to Tension," Proceedings, 4th U.S.
Congress of Applied Mechanics, Berkeley, Calif., 1962.
[25] P. C. Paris and G. C. M. Sih, "Stress Analysis of Cracks," Fracture Tough-
ness Testing and Its Applications, ASTM STP 381, Am. Soc. Testing Mats.,
1965, p. 31.
[26] C. M. Carman and J. M. Katlin, "Low Cycle Fatigue Crack Propagation of
High Strength Steels," 66-MET-3, Am. Society Mechanical Engrs., April,
1966.
[27] M. Schuler and C. M. Carman, "Low Cycle Fatigue Properties of ISNiCoMo
250 Maraging Steel," note for the meeting of Subcommittee II of ASTM
Committee E-24 on Fracture Testing of Metals, Schenectady, N.Y., Sep-
tember, 1964.
[25] F. Garofalo, Y. T. Chou, and V. Ambegaokar, "Effect of Hydrogen on
Stability of Micro Cracks in Iron and Steel," Acta Metallurgica, Vol 8, No.
8, 1960, p. 504.
[29] J. C. M. Li, R. A. Oriani, and L. S. Darken, "The Thermodynamics of
Stressed Solids," Z. Physik. Chemie, 1966.
[JO] Che-Yu Li and R. A. Oriani, "Some Considerations on the Stability of
Oxide Dispersions," paper presented at the Second Bolton Landing Confer-
ence on Oxide Dispersion Strengthening, Bolton Landing, N.Y., June, 1966.
[31] Che-Yu Li, P. M. Talda, and R. P. Wei, "Subcritical Crack Growth in an
Inert Environment," note for ASTM Committee E-24 meeting, Washington,
D.C., Jan. 31-Feb. 3, 1966.
[32] J. M. Krafft, "Correlation of Plane Strain Crack Toughness with Strain
Hardening Characteristics of a Low, a Medium, and a High Strength Steel,"

Applied Materials Research, Vol 3, April, 1964, p. 88.
[33] A. J. Birkle, R. P. Wei, and G. E. Pellissier, "Analysis of Plane-Strain
Fracture in a Series of 0.45C-Ni-Cr-Mo Steels with Different Sulfur Contents,"
Transactions, Am. Society Metals, Quarterly, Vol 59, Dec. 1966, p. 981.
[34] B. A. Bilby, A. H. Cottrell, and K. H. Swinden, "The Spread of Plastic
Yield from a Notch," Proceedings, Royal Soc., London, A, Vol 272, 1963, p.
304.
[35] E. Orowan, "Fundamentals of Brittle Behavior in Metals," Fatigue and
Fracture of Metals, Technology Press of MIT and John Wiley & Sons, Inc.,
New York, 1952.
[36] G. R. Irwin, "Crack Toughness Testing Using Circumferentially Notched
Round Bars," note for ASTM-FTHSMM Committee (now Committee E-24)
Meeting, May, 1961.
DISCUSSION
A. J. McEvily1 and T. L. Johnston1 (written discussion)—One of the

authors' conclusions is that current theories of fatigue crack propaga-
tion, including our own semi-empirical approximation,2 are not satis-
factory for general use in predicting fatigue crack propagation rates.
We do not agree with this conclusion and in fact will show that their
results for the rate of propagation, d2l/dN, conform to predictions made
by means of the equation
In this equation / is the half-length of a crack (see Fig. 9), ag is the peak
gross stress in a cycle, N is the number of cycles, E is Young's modulus,
ov is the yield stress, <ru is the engineering tensile stress, e is the strain at
maximum load, and A is a proportionality constant of dimensions in."1.
A is of magnitude 2 when the other units are psi and inches.
Equation 12 can be written as
or as
1
Metallurgy Dept., Scientific Laboratory, Ford Motor Co., Dearborn, Mich.
2
A. J. McEvily and T. L. Johnson, "On the Role of Cross-Slip in Brittle Frac-
ture and Fatigue," International Conference on Fracture, Sendai, Japan, September,
1965.
DISCUSSION ON PROPAGATION IN ULTRAHIGH-STRENGTH STEELS 481
FIG. 9—Normalized fatigue crack growth rates for 18NI (250) maraging
steel, 18Ni (300) maraging steel, and other alloys.
From the mechanical properties given in the authors' Table 3, Eq 13a

can be expressed as
for the 18Ni (250) maraging steel, and
for the 18Ni (300) maraging steel. Here &K is expressed in ksi\/irL
units (AK — 0-0V7T/)- Note that the value of e for both steels is estimated
by the authors to be 0.015 ± 0.005. The uncertainty will limit the relia-
bility of the computations.
From the authors' Figs. 6 and 7 we estimate that if a straight line of
slope y± is fitted to their data then at a dl/dN value of 10~5 in./cycle
the corresponding AK value would be 38 ksi-y/irl for the 250 maraging
steel and 35 ksi\/in. for the 300 maraging steel. Figure 9 shows the
corresponding lines plotted in the coordinate system used to compare
the behavior of a number of quite different alloys. It is seen that the
results for the maraging steels fall within the scatter band previously
determined. Therefore on the basis of the stress-strain properties of these
alloys a reasonable estimate of the resistance to fatigue crack propaga-
tion can still be made.
/. Weertman (written discussion)—The authors have concluded that
a crack propagation theory of mine fails to explain their experimental
data by a factor of 102.1 wish to point out in this discussion that another
interpretation of their data leads to a disagreement of less than an order
of magnitude. Hence, in view of the crudeness of the theory, experi-
mental and theoretical results are essentially in agreement.
The authors have expressed the theoretical equation I derived as
follows:
where a is the crack width, a is the amplitude of the cyclic stress, N is

the number of cycles, and aw is a constant which is given approximately
by the equation:
where E is a modulus, o-0 is identified with the ultimate strength of the

material, and D is a critical displacement. The product a0D can be identi-
fied with the Irwin crack extension force GZc .
The authors used the fatigue data to determine a "theoretical" value
of the term (u0D) which appear in Eq 15. They let aw equal the experi-
mental value, assumed that <r0 is equal to the tensile strength (I will
question this identification later), and then through Eq 15 evaluated
the term (cr0D). When this "theoretical" value of (o-0£>) was compared
with the experimental value of (jic it was found to be a factor of about
102 too large. As a result of this discrepancy the authors concluded that
the theory was inadequate.
Suppose that instead of making a comparison between theory and
experiment by their procedure, the following (equivalent) method is
used. Let (a0D~) be given the value equal to an experimentally determined
3
Materials Science Dept., Northwestern University, Evanston, 111.
(JK- . A "theoretical" value of a0 then is found with the aid of Eq 15 and

the experimental value of aw • The values of GO so determined are listed
in the table herewith.
Tensile Engineering
Material <TO , ksi Strength, Strain at a/Tensile
ksi Maximum Strength
Load
0.45C-Ni-Cr-Mo steel tempered 1 hr

at 400 F 3400 312 0.08 11
Above steel but tempered 1 hr at
800 F 2000 234 0.06 8.6
18Ni (250) maraging steel 1000 257 0.015 4
18Ni (300) maraging steel 2100 315 0.015 7
The tensile strengths measured by the authors also are listed. It can
be seen that the theoretical values of a0 determined from the fatigue
data are a factor of 4 to 11 greater than the tensile strengths.
Is this factor of 4 to 11 inconsistent with my theory and evidence
against it? My answer to this question is that it is not. In my paper I
suggested that the values of the ultimate strength to use in the theory
should be obtained from compression rather than tension tests. The
reason for this suggestion is that tension specimens always break pre-
maturely because of such extraneous factors as the formation of a neck
in ductile specimens and failure from internal preexisting flaws in more
brittle materials. The materials tested by the authors had only limited
ductility. (The engineering strain at maximum load is given in the fore-
going table.) If these materials had been tested under conditions in
which plastic strains of the 100 per cent could occur, it is clear that their
ultimate strengths would have been considerably higher. I think it not
unreasonable to suppose that the ultimate compressible strength of
these materials is at least a factor of 2 or 3, or larger, than the tensile
strength. If so, the disagreement between theory and experiment is only
a factor of about 3 or less.
J. Schijve* (written discussion)—The apparent agreement between
the growth in different types of steel can qualitatively be explained on
the model outlined in my own paper. For low yield strength materials
there is more cyclic slip. However, the conversion into crack extension
is less efficient due to the peak stress at the tip of the crack being released
more than in the high yield strength steel. These two factors may cancel
one another. The explanation does not work very well for aluminum
alloys, but it may be that it could be qualitatively applicable to the results
of the authors.
J. M. Krafft5 (written discussion)—It might be useful here to publish
* Director of technical services, National Aerospace Laboratory, NLR, Amster-
dam, Holland.
B
U.S. Naval Research Laboratory, Washington, D.C.
the content of the private communication designated as authors' Ref

12. The idea behind the growth rate estimate of Eq 5 is that the plane
strain fracture instability occurs with the establishment of strain for
tensile rupture over a small process zone distance df ahead of the crack.
In a single cycle up to Klc, a crack propagation df should occur. The
process zone size dF was reasoned to be substantially smaller than the
distance ahead of the crack over which the strain for tensile instability
prevails, as estimated from correlations between K^c and the strain
hardening exponent n. This discrepancy was accepted because the fracto-
graphic dimples which were examined appeared smaller than dT . It has
subsequently developed from work of Birkle et al of Ref 33 that fracto-
graphic dimples in close correspondence with dT can indeed be observed.
With a result indicating that dp ~ dT, a growth rate law fixed at df =
dT/l was obviously unrealistic. Correspondingly the cyclic damage
factor hypothesised to restore the growth rate after it was so reduced
had also become questionable. Hertzberg, also contributing to this
Symposium, had seen the striations indicative of propagation increments
to respond immediately to changes in &K levels, no delay as would be
expected if it were necessary for the material to be plastically cycled,
and thus "damaged," to effect a change in pace.
In view of these developments, I am inclined to revert to a simpler
modeling based on dT , formerly used in Ref 32, with minimal adjustment
(the n/E term for cyclic softening and a simpler though equivalent form
of/(7))
where 7 is the K or stress ratio kK/K^^ .

This formula [Eq 5 of text] reduced the calculated crack propagation
rates by a factor of %6. Relative to the available data, this would appear
as a lower limit to which influences of environment, which the authors
have treated, could account for measured positive departures.
With regard to the correlations with this formula of Table 6 it is
possible that a better agreement would be found at higher growth rate
levels, or as &.K approaches Klc where the model has its "fix" point.
R. P. Wei, P. M. Talda, and Che-Yu Li (authors)—The authors wish
to thank Professor Weertman, Drs. McEvily, Johnston, Schijve, and
Krarft for their interest and their discussions.
Strictly speaking, in accordance with the Bilby, Cottrell, and Swinden
dislocation model and in accordance with one of the specific conditions
imposed by Professor Weertman in his derivation, a0 must be identified
with an appropriate yield stress. The authors' choice in using the ultimate
tensile strength for <TO was made to favor Professor Weertman's analysis.
The authors, therefore, question Professor Weertman's justification in
claiming <TO to be the ultimate strength in compression in his discussion,

and question the precise physical meaning of this strength value in rela-
tion to the dislocation model used in his analysis.
With reference to the discussion by Drs. McEvily and Johnston, the
authors have re-examined their data on the four steels tested in the dry
argon environment in a manner consistent with that suggested by the
discussers' Eq 13a and Fig. 9. Based on the mechanical properties given
in Table 3 and the values of aavg given in Table 7, the results for the
two maraging steels would be slightly to the left of the lines indicated
by the discussers in Fig. 9 (being on either side of the left-hand limit of
the indicated scatter band); the line for the 0.45C-Ni-Cr-Mo steel tem-
pered at 800 F would be nearly coincident with that shown for the
7075-T6 aluminum alloy, and the results for the 0.45C-Ni-Cr-Mo steel
tempered at 400 F would lie to the left of the 7075-T6 aluminum alloy
data. On the basis of this re-examination and the authors' original
considerations, the following points appear pertinent. First, it is clear
from the foregoing discussion that not all of the authors data fall within
the scatter band indicated by the discussers. The disagreement would
be somewhat larger if data for the moist environment were included in
the comparison. Second, whereas the discussers' relationship predicts
an order of magnitude difference between the rates of fatigue-crack
growth in the 18Ni (250) maraging steel and the 0.45C-Ni-Cr-Mo steel
tempered at 400 F, the authors' experimental results showed that the
growth rates were about the same. Third, by comparing the discussers'
Eq 12 with the form used by the authors, Eqs 10 and 11, a characteristic
length of the order of 1 in. is obtained when the constant A in Eq 12
is given a value of 2 as suggested by the discussors. It is difficult to
attach any physical significance to this dimension in relation to the
fatigue-crack growth process. Therefore, the authors can only conclude
that the current theories of fatigue-crack propagation are not satis-
factory for general use in "predicting" fatigue-crack growth rates,
particularly since none of these theories account for the influence of
test environments which has been shown to be important.
With regard to the comments of Dr. Schijve, since they are qualitative
in nature, it is not possible for the authors to comment on them specifi-
cally.
The authors are very pleased to have Dr. KrafTt include the content
of his private communication designated as authors' Ref 12 here. They
agree that perhaps a better agreement between Dr. Krafft's formula
and experiment could be found at growth rate levels higher than those
observed in this experiment. However, the results of Carman and Katlin,
shown in authors' Figs. 7 and 8, for an 18Ni (250) and an 18Ni (300)
maraging steel cast some doubt on this possibility.
D. W. Hoeppner1
The Effect of Grain Size on Fatigue Crack

Propagation in Copper
REFERENCE: D. W. Hoeppner, "The Effect of Grain Size on Fatigue

Crack Propagation in Copper," Fatigue Crack Propagation, ASTM 415,
ABSTRACT: A study was made of the influence of grain size on the
propagation of fatigue cracks in electrolytic tough pitch copper. Copper
specimens possessing grain sizes ranging from 0.060 to 7 mm were pro-
duced by annealing at various temperatures for preselected periods of
time. Center-notched specimens were subjected to a positive mean stress
and an alternating stress superimposed. Measurements of crack length
were taken continually throughout the test to determine fatigue crack
length as a function of number of cycles. Observations were made, using
optical and electron fractography, on the mode of propagation and the
influence of grain boundaries on the crack as it extends through the matrix.
At a given stress level the initiation period of a fatigue crack at the starter
notch was observed to increase with decreasing grain size. However,
grain size was observed to have little effect on the propagation rate.
KEY WORDS: copper, fatigue (materials), crack propagation, grain size,
grain boundary, microstructure
Studies pertaining to the influence of grain size on fatigue behavior

have not been extensive. One such study was made by Sinclair and Craig
[I].2 Among other things, they concluded that the finite fatigue life of a
metal increases as the grain size is reduced in 70-30 brass. Sinclair and
Craig dealt with the fatigue life as exemplified by rotating cantilever
beam specimens. Consequently, they dealt with total life and were not
specifically concerned with initiation or propagation. More recently, For-
rest and Tate [2] published the results of a detailed study on the factors
which influence the initiation and propagation of fatigue cracks in 70-30
brass. They found that fatigue cracks initiated in a brass at stresses be-
low the fatigue limit and that the cracks propagated in a transcrystalline
manner until they contacted a grain boundary, at which point their mo-
1
Research metallurgist, Mechanical Engineering Dept., Battelle Memorial
Inst., Columbus Laboratories, Columbus, Ohio. Personal member ASTM.
* The italic numbers in brackets refer to the list of references appended to this
paper.
486
KOEPPNER ON EFFECT OF GRAIN SIZE ON PROPAGATION IN COPPER 487
tion would be blocked. They thus concluded that the fatigue limit is more
directly related to the stress required to propagate a crack across a
boundary than to the stress required to initiate a crack.
Albeit the initiation of fatigue cracks is important; during recent years
increased emphasis has been placed on developing an understanding of
fatigue-crack propagation. A great deal of progress has been made under-
standing fatigue-crack propagation using metallographic and fracto-
graphic studies. However, many of the studies which have been made
thus far have not taken the metallurgical structure of a material into ac-
count. Furthermore, the "laws" which have been developed to describe
fatigue-crack propagation behavior, with but one exception, have ne-
glected the metallurgical structure. It was with these points in mind that
the present study was initiated.
Several factors influence the behavior of a polycrystalline aggregate as
compared with a single crystal. One of the most important is the restraint
to slip. Restraint to slip offered by mutually interacting grains in a poly-
crystal is caused by two factors [3], First, the grain boundaries hinder
slip, and, second, deformation in a polycrystal is more complex than in
a single crystal.
In addition to the hindering and complexity effect, the role of grain
boundaries in a polycrystalline matrix is dependent on the orientation of
the grains adjacent to the boundary [4] and the orientation of the bound-
ary with respect to the applied stress [5,6]. For example, Hoeppner and
Vitovec [4] found, using tricrystals of copper, that the mode of initiation
and propagation changed with the misorientation angle at grain bound-
aries. Both Shrier et al [5] and Hempel [6] have found that the introduc-
tion of boundaries has an effect on the mode of fatigue failure—the effect
being dependent on the orientation of the boundary with respect to the
stress field. Thus, the importance of grain size and grain boundaries in
the overall fatigue process has been established. It now seems important
to determine how significant each of these aspects is in the propagation
stage. In this study, the influence of a wide variation in grain size on the
propagation of fatigue cracks in electrolytic tough pitch copper has been
made.
Specimen Preparation
Sheet specimens Vs-ui. thick were selected as the geometry to be used
for this study. The specimens were sectioned from one large sheet of
electrolytic tough pitch copper in the longitudinal direction and were
machined to the dimensions shown in Fig. 1. A starter notch, possessing
the dimensions shown in the figure, was electric-discharge-machined
through each sheet specimen. Prior to testing and final polishing, the
starter notches were polished using a wire and jeweler's rouge slurry to
ensure that all burrs and rough spots were removed.
Based upon a series of trial experiments, specimen blanks were en-

capsulated in argon-filled, Vycor tubes and heat-treated at the tempera-
ture and time indicated in Table 1. Specimens were supported in the
Vycor tubes to prevent warpage, and the specimens treated at 1830 F
were wrapped with iron wire, outside the gage sections, to prevent diffu-
sion bonding.
FIG. 1—Specimen used for crack propagation studies.
TABLE 1—Heat treating time and temperature and resulting grain size.
Heat Treatment
Average Grain Diameter, mm
Time, hr Temperature, deg F
2^ 420 0.058
1 570 0.064
1 1000 0.107
64 1100 0.127
4 1470 0.133
64 1470 0.150
(16 14701
7.0
+
\64 1830J
The grain size corresponding to a given heat treating condition is also

shown in Table 1. The grain sizes of the sheet specimens, heat treated
according to the regimes mentioned in Table 1, were determined by the
line intercept method. In addition to the line intercept method, compari-
son was made between the actual grain sizes and ASTM grain size photo-
micrographs. The determination of grain size was in all instances difficult
due to the presence of annealing twins. It was often difficult to distinguish
between grain boundaries and twin boundaries.
Subsequent to heat treating, tension tests were conducted to deter-
HOEPPNER ON EFFECT OF GRAIN SIZE ON PROPAGATION IN COPPER 489
mine the typical tensile properties for the various grain sizes. The results
are shown in Table 2. It is important to note in this table that the range
in ultimate tensile strength is 7.5 ksi, with the strength decreasing with
increasing grain size. The 0.2 per cent offset yield strength, however,
decreases significantly from the grain size values of 0.064 to 0.107 mm
and then decreases slightly with increasing gram size.
Fatigue Testing
All fatigue tests were conducted in a Krouse direct-stress fatigue ma-
chine of 5000-lb capacity at a frequency of 1725 cpm. The machine is
equipped with an automatic hydraulic load maintainer that monitors
test loads on the specimen. Accuracy of load setting and maintenance is
approximately 3 per cent of the maximum test load.
Prior to fatigue testing each specimen was mechanically polished in
the metallography laboratory. Subsequent to polishing and immediately
TABLE 2—Yield strength and ultimate tensile strength for the various
grain sizes.
Average Grain Diameter, mm Yield Strength, ksi, 0.2% offset0 Ultimate Tensile Strength, ksi"
0.058 28.8 34.7

0.064 25.6 33.9
0.107 5.0 31.4
0.127 5.4 31.2
0.133 4.45 31.03
0.150 4.47 30.5
7.0 3.5 27.2
" Average of three values.
prior to test, each specimen was chemically etched for 1 min using the
following etchant: 100 cm3 distilled water; 2 g potassium dichromate;
8 cm3 sulfuric acid; and 1 drop hydrochloric acid. This procedure was
performed to allow microscopic observations to be made throughout the
crack propagation test.
During the test the machine was stopped and the specimen removed.
Measurements of crack length were then taken and microscopic obser-
vations made on the specimen surface. Measurements of crack length
were made using a replication procedure. The technique involves the
application of ethyl acetate to the specimen surface and pressing a small
sheet of cellulose nitrate over the acetate.
Crack Propagation Results and Discussion

Crack propagation tests for the various grain sizes were conducted at a
mean stress of 5000 psi and an alternating stress of 5000 psi. Results for
eight of the crack propagation tests are shown in Fig. 2. The specimen
number and average grain diameter in millimeters are indicated above
FIG. 2—Fatigue crack propagation curves for the various grain size specimens (mean stress 5000 psi, alternationstress5000 psi).
each curve. It will be noted in the figure that the two larger-grained speci-
mens possessed longer starter notch lengths than the remainder of the
specimens. This procedure was necessary in order to ensure initiation
of the crack at the starter notch rather than at a grain boundary.
The curves shown in Fig. 2 can essentially be grouped in three grain-
size categories, namely:
FIG. 3—Cycles of propagation from a crack length of 0.2 in.
1. Grain size from 0.058 to 0.064 mm

2. Grain size from 0.107 to 0.150 mm
3. Grain size of 7.0 mm
The data plotted in Fig. 2 indicate that the first group of specimens
spent a great deal of lifetime in initiation of the crack at the starter notch
and the early stages of propagation. Upon increasing the grain size, the
initiation and initial growth period decreased. The curves hi Fig. 2 es-
sentially shift from right
Downloaded/printed by to left for the three grain-size categories. The
finite fatigue life is greater as the grain size in decreased. This observa-
tion is consistent with that of Sinclair and Craig [1],
In all but one case, that being Specimen D-6, the propagation of the
crack was continuous. The crack in Specimen D-6 propagated in a some-
what erratic manner and was repeatedly blocked by grain boundaries,
thus providing the delay periods in the growth curve.
The data thus indicate that grain size has a significant effect on the
initial development of the crack. As the crack grew, grain size had no
apparent effect on the propagation of the crack. This is further demon-
strated in Fig. 3, where total crack length versus cycles of propagation
from a crack length of 0.2 in. is plotted. With the exception of the data
for Specimen A-3, all of the curves fall within a narrow scatter band.
FIG. A—Observations on the surface of Specimen C-4; average grain diameter

0.107 mm. (a) 800,000 cycles and (b) 1,400,000 cycles.
These two figures indicate that the variation in grain size produced little
observable effect on the later stages of crack propagation.
It is apparent that little influence of grain size is observed after the
crack reaches some point in propagation. Up to this point the effect of
grain size is relatively pronounced. In the early stages of propagation the
crack is being influenced greatly by local crystallographic conditions.
Presumably the hindering and complexity effects are greatest in the fine-
grained material and the crack has greater difficulty "getting started."
This is not the case in the coarse-grained material. Slip is relatively easy
and the crack can develop much more freely.
Metallographic Observations
Observations which were made on the surfaces of various specimens
are shown in Figs. 4 through 9. In all cases the stress direction is perpen-
dicular to the horizontal
Downloaded/printed by axis of the figures. In the discussion of the
FIG. 5—Observation on the surface of Specimen B-5; average grain diameter

0.127 mm, 1,000,000 cycles.
FIG. 6—Observations on the surface of Specimen C-7; average grain diameter

7.0 mm. (a) 311,000 cycles and (b) 1,000,000 cycles.
figures, reference is made to important features on the micrographs of

the surfaces.
Figure 4 shows observations made on the surface of Specimen C-4.
In Fig. 4a the crack is shown in its early development. The path of the
crack appears to be influenced markedly by local crystallographic condi-
tions. Direction changes occur when the crack contacts a boundary, and
the crack may propagate in either an intercrystalline or transcrystalline
manner. In Fig. 4b significant branching of the crack is observed. In
addition, crystallographic features appear to have a smaller influence on
the crack path as the paíh is becoming more nearly perpendicular to the
applied stress field.
Figure 5 shows even more clearly how the crack path was influenced
by grain boundaries. Considerable grain boundary cracking can be ob~
served in this photomicrograph. The extent of slip ín the vicinity of the
crack can also be observed in this figure.
Figure 6 shows observations made on the surface of Specimen C-7.
In Fig. 6a, rather intensive slip is observed in the vicinity of the starter
notch. As can be observed in Fig. 6b, the crack began propagating along
the slip bands. However, after propagating approximately 30 /* the crack
FIO. 7—Observations on the surface of Specimen D-2; average grain diameter

7.0 mm. (a) 103,000 cycles and (b) 300,000 cyctes.
changed direction, and the path became somewhat irregular. These de-
velopments are dramatized in Figs. la and b. Presumably, the crack is
switching from one slip system to another, and its mode of propagation is
dependent on local crystallographic conditions. In Fig. la a rather in-
tensive slip área formed as shown. As niay be seen in Fig. Ib, the crack
started at a srnall nick in the starter notch and propagated approximately
perpendicular to the stress field. Also, in Fig. Ib slip can be seen on at
least three systems.
In Fig. 8 observations made on the surface of Specimen D-5 are
shown. Note in Fig. 8a that a small group of slip lines has formed at the
midpoint of the starter notch. In Fig. 86 the crack has initiated and
propagated
Copyrightinby the transverso
ASTM boundary
Int'l (all rights to the
reserved); Mon triple
Dec 7point, where
14:40:45 EST it
2015
changed direction and
Downloaded/printed by attempted to propágate along an incline in a
Tnscrystalline man Intensive slip is seen to have formed along thhener.

inclined boundary in Fig. Sc, and it may be noted that the crack eventu-
ally propagated through the triple point and that its subsequent propaga-
tion was transcrystalline and perpendicular to the applied stress field.
To further illustrate the influence of boundaries on the crack front,
Fig. 9 was prepared. In Fig. 9a the crack has propagated through a grain,
contacted a boundary, and propagated along the boundary. Figure 9b
FIG. 8—Observations on the surface of Specimen D-5; average grain diameter

7.0 mm. (a) 600,000 cycles, (b) 1,200,000.
shows how the crack has extended itself along the boundary and propa-
gated to the terminus of the boundary, at which point it changes direction.
In Fig. 9c the crack is seen to be propagating approximately perpendicular
to the applied stress field, but it is still influenced by crystallographic con-
ditions as is seen in Fig. 9d. Figure 9c shows a twin boundary at x50 in the
crack path. Figure 9d shows an enlarged view of the effect of the twin
boundary on the crack path.
The many by
Copyright features observed
ASTM Int'l in reserved);
(all rights Figs. 4 through
Mon Dec9 7can be summarized
14:40:45 EST 2015
as follows:
As the crack initially developed, its path was quite irregular, and it
propagated in either a transcrystalline or intercrystalline mode. The
propagation mode is quite dependent on the local orientation of the
grains. Presumably, when a grain is not favorably oriented for slip, the
crack propagates in a boundary, since it prefers, at least in the early
FIG. 9—Observations on the surface of Specimen D-6; average grain diameter

7.0 mm. (a) 700,000 cycles, (b) 753,000 cycles, (c) 950,000 cycles, and (d) 950,000
cycles.
stages, to propagate along favorably oriented slip bands. It always ap-

peared that the crack was searching out the most favorable path in the
initial stages. It was stated that as the crack progressed, the direction of
propagation became nearly perpendicular to the stress field. However,
the crack path was considerably influenced by grain size. For example,
in the smallest grain size specimens, the crack path was very nearly per-
pendicular to by
Copyright theASTM
applied
Int'l tensile stress.
(all rights In addition,
reserved); Mon Dec as7 the crackEST
14:40:45 grew
2015it
was influenced less and
Downloaded/printed by less by grain boundaries. This was not the case
for the large-grained material. In this case, the crack path did not always
become perpendicular to the applied stress field. Rather, the crack was
influenced by local crystallographic conditions over much of its route.
The amount of slip which preceded the crack increased as the crack
extended; it was difficult to obtain photomicrographs of the specimen sur-
FIG. 10—Electron fractograph of area near the notch of Specimen D-4

(0.058 mm) showing crystalline facets and slip lines on the fracture surface.
face after the crack extended beyond a certain length. The plastic zone
size in these materials, at least on the surface, appeared to be a function
of grain size. However, quantitative measurements on the plastic zone
size were not made in this study.
It is not clear in this study how the distinction between Stage I and
Stage II propagation is clearly established—since the propagation path
was Copyright
quite dependent on local crystallographic conditions even at rather
long Downloaded/printed
crack lengths. Forsyth
by [9] has stated that Stage I cracks propagate
along crystallographic planes of maximum shear stress and change direc-

tion with orientation at grain boundaries. Stage II cracks, however, prop-
agate approximately perpendicular to the applied tensile stress in a regu-
lar manner and are characterized by striations on the fracture surface.
In this study the crack path in the large-grained specimens was dependent
FIG. 11—Electron fractograph of area located Vi in. from starter notch of

of Specimen D-4 (0.058 mm) showing crystalline facets and intersecting slip.
on crystallographic factors over almost its entire length. The extent of

crystallographic influence was dependent on grain size, increasing as the
grain size increased.
In an attempt to more clearly ascertain the stage of crack propagation,
electron fractographic studies were employed. Figures 10 through 17
show typical regions on the fracture surfaces of the specimens. Arrows on
the fractographs indicate the direction of propagation.
Copyright
Figure 10 isbyanASTM Int'l (all
electron rights reserved);
fractograph of theMon Dec
area 7 14:40:45
near ESTof
the notch 2015
Speci-
men D-4 showing crystalline facets and slip lines on the fracture surface.
The observation of slip lines on the fracture surface, also reported by
Hertzberg [10], suggests the propagation of the crack in this specimen was
dependent on local crystallography. The mechanism of their formation
is not clear at this time.
FIG. 12—Striated fracture surface of Specimen C-4 (0.107 mm}. Area located
Vi in. from starter notch.
Areas such as those shown in Fig. 10 were typical of the fracture surface
appearance near the notch of the fine-grained specimens. Few regions
were striated. Even at magnifications of the order of X 20,000 striations
were not resolvable on the crystalline facets. Presumably, the crack was
propagating between grain, twin, and sub-boundaries, resulting in the
faceted appearance. Figure 11 is another fractograph from Specimen D-4
located about !/2-in. from the starter notch. Again, crystalline facets and
intersecting
Copyrightslip
by are
ASTMvisible. Striated
Int'l (all regions were
rights reserved); found7 at14:40:45
Mon Dec this distance
EST 2015
from the notch, but the appearance of faceted regions and slip lines at
this distance from the starter notch is of interest.
Figure 12 shows a striated fractograph of Specimen C-4. This fracto-
graph is typical of fatigue fracture surfaces, and it was anticipated that
this appearance would be found frequently on the surfaces of the fine-
grained specimens. As has already been shown, this was not the case. The
FIG. 13—Striated fracture surface of Specimen C-4 (0.107 mm) showing fur-
rows. Area located Vi in. from starter notch.
fractograph shown in Fig. 13 also was taken of Specimen number C-4. An

interesting development here is the observation of parallel furrows per-
pendicular to the propagating crack front. As will be shown, these became
quite common as the grain size was increased.
It is generally accepted that striations form parallel to the advancing
crack front Figure 14 illustrates the random manner in which striations
can form. It isbyquite
Copyright ASTMprobable, considering
Int'l (all rights thatDecthe7 grain
reserved); Mon size
14:40:45 ESTof2015
this speci-
men was 0.150 mm and that the magnification was X3780, that all of
these striations formed within one, or at most two, grains. Thus, any
model which proposes that the formation of striations is dependent on
crystallography must be able to explain the random manner in which
striations may develop locally.
Figure 15 shows a region located l/z-m. from the starting notch of Speci-
men D-6. The appearance of two striated regions separated by an un-
FIG. 14—Fractograph showing multiple crack paths as evidenced by striated

path directions. Specimen D-7 (0.150 mm) at Vi in. from starter notch.
striated region is quite evident. Regions such as this were typical of those
detected in the large-grained specimens. Stereo pairs of these regions were
taken, and the unstriated regions connecting the striated ones were found
to be ledges. A pictorial sketch of this observation is shown in Fig. 16. It
appears that the crack was propagating by fatigue along the striated re-
gions, and the ledge regions between appear to have fractured by some
form of cleavage, but since cleavage is not observed at room temperature
in face centered
Copyright cubic Int'l
by ASTM metals(all this
rightsdoes not appear
reserved); likely.
Mon Dec FigureEST
7 14:40:45 17 2015
shows
a second fractographbyfrom Specimen D-6. The furrows pointed out in
Downloaded/printed
FIG. 15—Specimen D-6 (7.0 mm) Vi in. from starter notch. Fractograph shows
striated region in center and an unstriated ledge connecting the lower striated re-
gion.
FIG. 16—Pictorial
Copyright sketch
by ASTM Int'l showing
(all rights bandsMon
reserved); of Dec
striated regionsEST
7 14:40:45 joined
2015 by un-
striated ledges.
University of Washington (University of502
Washington) pursuant to License Agreement. No further
Fig. 13 are quite evident in this fractograph. It appears as if the crack is

extending itself along several local fronts and these fronts are tilted with
respect to one another.
Many of the points indicated in Figs. 10 through 17 cannot be readily
explained. They are reported here as significant observations on the frac-
ture surfaces of this study. The models thus far proposed for striation
FIG. 17—Striated fracture surface showing striated region and furrows. Speci-
men D-6 (7.0 mm) Vi in. from starter notch.
formation [10] do not account for most of these observations. The observa-
tions made do not allow a clear distinction to be drawn between the stages
of crack propagation. In the fine-grained material the cracks, after reaching
a length of }/2-in., were macroscopically nearly perpendicular to the ap-
plied stress field, but microscopically striations could not be detected. In
the coarse-grained materials, the fracture surfaces were striated a con-
siderable amount, but macroscopically the crack path was irregular and
quiteUniversity
dependent on local crystallographic
of Washington conditions.
(University of Washington) It thus
pursuant appearsAgreement.
to License that No fur
additional experimental programs are needed to clearly define the charac-

teristics of Stage I and Stage II crack propagation if, indeed, they are the
only stages of consequence.
Conclusions
As a result of this preliminary study, the following conclusions can
be drawn:
1. In electrolytic tough pitch copper a variation in grain size from 0.058
to 7.0 mm results in a decrease in the cycles required to initiate a crack at
the starter notch, and thus the overall fatigue lifetime decreases with an
increase in grain size.
2. Even though there are substantial differences in the details of
propagation in the fine-grained compared to the coarse-grained material,
grain size had relatively little influence on the rate of propagation after
the crack reached a length of 0.2 in. It may be that this conclusion is true
only for materials of high stacking fault energy which form cell structures.
This paper has dealt primarily with a presentation of observations made
on ten fatigue specimens. Further work is contemplated from which it
may be possible to resolve some interesting questions posed by the study.
References
[7] G. M. Sinclair and W. J. Craig, "Influence of Grain Size on Work Hardening
and Fatigue Characteristics of Alpha Brass," Transactions, Am. Society
Metals, Vol 44, 1952, pp. 929-946.
[2] P. G. Forrest and A. E. L. Tate, "The Influence of Grain Size on the Fatigue
Behavior of 70-30 Brass," Journal, Institute of Metals, Vol 93, 1964-65, pp.
438-444.
[3] D. McLean, Grain Boundaries in Metals, Clarendon Press, Oxford, England,
1957, pp. 150-199.
[4] D. W. Hoeppner and F. H. Vitovec, "Initiation and Propagation of Fatigue
Cracks in Tricrystals of Copper," Transactions, Metallurgical Society of the
AIME, Vol 230, October, 1964.
[5] A. Shrier, H. Yamamoto, and S. Weismann, "Fatigue of Metal Crystals,"
Technical Report No. AFML-TR-65-86, April 1965.
[6] M. R. Hempel, "Slipband Formation and Fatigue Cracks Under Alternating
Stress," Basic Mechanisms of Fatigue, ASTM STP 237, Am. Soc. Testing
Mats., 1959, pp. 52-82.
Interdisciplinary Approach, Syracuse University Press, Syracuse, N. Y., 1964,
pp. 107-127.
[8] D. R. Donaldson and W. E. Anderson, "Crack Propagation Behavior of
Some Airframe Materials," Proceedings, Crack Propagation Symposium,
Cranfield, England, Vol 2, September, 1961, pp. 375-441.
[9] P. J. E. Forsyth, "A Two-Stage Process of Fatigue Crack Growth," Proceed-
ings, Crack Propagation Symposium, Cranfield, England, Vol 1, September,
1961, pp. 76-87.
chanics to Fatigue Crack Propagation in High-Strength Aluminum Alloys,"
Ph.D. dissertation, Lehigh University, Bethlehem, Pa., May, 1965.
Fatigue Crack Propagation Under Program
and Random Loads
REFERENCE: J. C. McMillan and R. M. N. Pelloux, "Fatigue Crack

Propagation Under Program and Random Loads," Fatigue Crack Propa-
gation, ASTM STP 415, Am. Soc. Testing Mats., 1967, p. 505.
ABSTRACT: The influence of maximum stress, stress range, and sequence
of load application on the rate and mechanism of fatigue crack propaga-
tion in 2024-T3 aluminum alloy was studied by means of electron fractog-
raphy. Variable-amplitude loading programs were designed to provide tests
under the following conditions: (1) constant maximum stress with three
different levels of stress range, (2) constant stress range with three and
four levels of maximum stress, (3) pseudo-random load application
achieved by random distribution of the load spectra defined in Items 1
and 2, and (4) uniform maximum stress with peak overloads and under-
loads.
The macroscopic growth rates were determined on center-notched
crack growth panels, and the fracture surfaces were examined by elec-
tron fractography. The analysis of the influence of program loads on the
rate and mechanism of fatigue crack growth was accomplished by:
1. Comparing plots of crack length versus measured rates of crack
propagation for the different programs. The measured rates were also
compared with rates calculated by a computer program.
2. Relating the count and spacing of the fatigue striations observed on
the fracture surfaces with the applied load program.
KEY WORDS: aluminum alloys, fatigue (materials), crack propagation,
crack growth rate, fractography, fatigue striations, program loads, ran-
dom loads
The fatigue fracture surfaces of many alloys show well-defined crack

front arrest lines commonly called fatigue striations. These striations were
first observed with the optical microscope but can best be studied with the
electron microscope at magnifications ranging from xlOOO to X 20,000.
In the case of aluminum alloys, the striations are remarkably well defined,
at least for crack growth rates ranging from 1 to 1000 /xin. per cycle.
Nearly all of the fracture surface shows striations at the slowest growth
1
Research engineer, Commercial Airplane Div., The Boeing Co., Renton, Wash.
2
Research specialist, Boeing Scientific Research Laboratory, The Boeing Co.,
Seattle, Wash.
505
rates, and the percentage of fracture surface occupied by stnations de-

creases with increasing growth rates.
The fact that the striations represent successive positions of the crack
front enables us to study the mechanisms and rates of fatigue crack propa-
gation as a function of crack length, load history, stress intensity factor,
and environment. So far, most studies of fatigue fracture surfaces by
electron microscopy have been limited to qualitative observations of the
striation shapes and profiles and to striation interactions with grain
boundaries or second-phase particles. A limited amount of quantitative
work has shown that for uniform cyclic loads and for crack growth rates
greater than 1 to 10 /iin. per cycle, there is a one-to-one correlation be-
TABLE 1—Chemical composition of 2024-T3 aluminum alloy.

Program Material Cu Mg Mn Zn Si Cr Fe Ti Al
PI to P7, Pll,
P12.. 4,.10 1.65 0.56 <0.03 0.07 <0.03 0.21 <0.03 Bal
P8, P9, P10. 4 .44 1.56 0.66 <0.05 0.08 <0.01 0.17 <0.03 Bal
TABLE 2—Mechanical properties of 2024-T3 aluminum alloy (0.16 in.

gage sheet, transverse grain direction).
Reduction of
Program Material FTU , psi FTY . psi Elongation, % Area, %
PI to P7, Pll, P12 68 400 46 900 17.0 23.0

68 400 47 700 18.0 24.0
69 000 48 000 19.5 26.0
P8, P9, P10 70 200 46 700 20.0 27.0
69 700 45 200 20.0 28.0
70 500 45 100 19.0 26.0
tween striations and load cycles [/].3 Also, it has been shown that the
measurements of striation spacing, which represent the microscopic rate
of crack propagation, correlate very well with the macroscopic measure-
ments of crack propagation rates [2,3].
In recent studies we used optical and electron fractography to look at
the fatigue fracture surfaces of many structural test components tested
under programmed loads designed to simulate service conditions. In each
case, the load program could be readily related to the general topography
of the fracture surface. However, the complexities of the test components
and of the load spectra were such that it was not possible to make a direct
correlation between each load and the observed striations, or to explain
the unusual striation spacings often observed. As a consequence, we ran
a Copyright by ASTM Int'l (all rights reserved); Mon Dec 7 14:40:45 EST 2015
paper.Downloaded/printed by
MCMILLAN AND PELLOUX ON PROGRAM AND RANDOM LOADS 507
TABLE 3—Programs with constant maximum stress (maximum stress

Sg = 12,000 psi).
Programs
Spectrum
PI ABC P2 CBA P3 CBA P4 ABCD PS Random
A
n cycles 9 9 6 12
0 1.2 1.2 2 2
B
« cycles 8 8 8 8
ft 3 3 3 3
C
« cycles 7 7 10 20
ft 20 20 20 20
D
n cycles 8
0 3
TABLE 4—Programs with constant stress amplitude (stress amplitude

AS = 7000 psi).
Programs
Spectrum P6 P7 P8 P9 J"10 Pii P12
Ran
ABCD DCBA ABC PR A
CBA ' AB AB
dom
A
n cycles . . 4 4 20 20 4 3
Smax , ksi . . 14 14 14 14 12
18 2.0 2.0 2.0 2.0 2.0 2.4
B
n cycles 5 5 16 16 20 21
Smax . ksi 12 12 12 12 12 14
/3 2.4 2.4 2.4 2.4 2.4 2.0
C
« cycles 6 6 12 12
Smax, Ksi 10 10 10 10
j3 3.33 3.33 3.33 3.33
D
n cycles. . 7 7
Smax, ksi 8 8
0 11.4 11.4
a number of fatigue crack propagation tests with different spectra of

loads. These loads were programmed in such a way as to lead to an unam-
biguous interpretation of the microfractographic observations.
The purpose of this research program was to develop a better under-
standing of the following problems:
1. The influence of maximum load, load amplitude, and load sequence
on crack propagation rates—Two basic types of load programs were
studied: (a) constant maximum load with variable load amplitudes and (b)
constant load by
Copyright amplitude
ASTM Int'lwith variable
(all rights maximum
reserved); Mon Decloads.
7 14:40:45 EST 2015
2. Downloaded/printed
The comparisonbyof pseudorandom loading with program loading—
FIG. l(a)—Program PL (b}—Program P3. (c)—Program P4. (d)—Program P5.
The two basic types of load programs defined in the foregoing were
randomized and the crack growth rates were compared.
3. The occurrence of crack arrest and crack acceleration—Striations
were counted, and striation spacings were measured following changes of
maximum loads or load amplitudes.
4. The mechanisms of fatigue cracking and striation formation—Some
of the load sequences were planned to distinguish clearly between loading
and unloading behavior at the crack tip.
5. Copyright
The prediction of programmed or pseudorandom load crack propa-
gationDownloaded/printed by growth data from constant maximum load and
rates—Using crack
FIG. 1—Continued.
constant load amplitude tests, a computer program calculated the expected

crack growth rates for all the programmed and random load tests.
Experimental Procedure
Materials
The material chosen for this program was 2024-T3 aluminum alloy
in theCopyright
form of bybare 0.160-in.-gage
ASTM Int'l (all rightssheet. TheMon
reserved); 2024-T3 alloy was
Dec 7 14:40:45 selected
EST 2015
because its fatigue crack
Downloaded/printed bygrowth rates are less susceptible to environmental
FIG. 2(a}—Program P6. (b)—Program P8. (c}—Program P10.
effects such as humidity than are alloys of the 7000 series [4]. Also, the
fatigue striations of the 2024 alloy are always clearly defined and are of
the ductile type [5]. Thus, they lend themselves to a detailed striation pro-
file and spacing analysis. A number of similar tests were also conducted
on 7075-T6 aluminum alloy to confirm the generality of relationships
established for the 2024 alloy. The chemical composition and mechanical
properties of the transverse grain direction of the 2024-T3 material used
in this program
Copyright by are
ASTMshown
Int'l in
(allTables 1 and 2, Mon
rights reserved); respectively.
Dec 7 14:40:45 EST 2015
FIG. 2—Continued.
FIG. 3 (a)—Program PH. (b}—Program P12.
Test Procedure
The test specimen used in the fatigue testing portion of the program was
a 24 by 9-in. center-notched fracture panel with the transverse grain di-
rection parallel to the tensile axis. Fatigue testing was accomplished in a
vertical 125-kip electrohydraulic fracture jig of Boeing design at a cyclic
rate of 90 cpm.
Program and random loads were introduced by punched-tape digital
programming through a Boeing-designed forced closed loop, servo system
random loadbycontroller.
Copyright Previous
ASTM Int'l (all work has
rights reserved); Monshown
Dec 7this testing
14:40:45 ESTsystem
2015
to be capable of applying loads with an absolute error of ± 1 per cent of
the maximum programmed load for the random load case at cyclic rates
up to 90 cpm.
The initial center crack (2d) was 0.5 in. and crack growth was monitored
with a X50 traveling microscope. Specimens were normally cycled to
failure. All testing was conducted in a laboratory environment. Tempera-
ture and relative humidity, recorded periodically throughout the tests,
ranged from 65 to 79 F and 17 to 52 per cent relative humidity. The
averages were approximately 70 F and 40 per cent relative humidity.
Standard two-stage replicating techniques were used to prepare replicas
for examination with the electron microscope. The shadowing material was
germanium, and in every case the shadowing direction was parallel to
the overall crack propagation direction. Replicas of the fracture surface
were examined at average crack lengths of la = 0.6, 0.9, 1.3, and 1.7 in.
The crack length corresponding to each electron micrograph was recorded
with an accuracy of Aa = ±0.02 in. The critical fatigue crack lengths for
all specimens ranged from la = 6 to 7 in.
Test Programs
Twelve load programs, PI to PI2, were run. Each program was broken
down into spectra identified as A, B, C, and D. In each spectrum the
maxima and minima of loads were numbered in sequence so that any
cycle or part of a cycle could be identified and referred to without con-
fusion. For instance, P1-A9-B1 refers to the compression part of the first
cycle of Spectrum B of Program PI. The cycle digits should always be
read in the forward direction.
Tables 3 and 4 list the programs, the spectra sequence, the number of
cycles in each spectrum, the maximum stress and stress amplitude, and
y8-faCtOrS (/? = Smax/Smln).
The programs are shown in Figs. 1, 2, and 3. Load was a sine function
of time represented as triangular waves on the sketches for simplicity.
Programs PI, P2, P3, P4, and P5 represent a constant maximum load with
variable load amplitudes. P2 is the reverse of PI, and P3 was obtained
by modifying P2 in order to obtain crack propagation with the lower load
amplitude level. P4 was made up by juxtaposing P3 and a reversed P3
to measure the influence of the sequence of load spectra. Program P5
was obtained by "randomizing" P3. The first three cycles were purposely
planned with a constant AS = 11,400 psi in order to have a reference
marker on the microfractographs.
Programs P6, P7, P8, P9, PI 1, and P12 represent a constant load ampli-
tude with variable maximum loads. P7 and P9 are the reverse of P6 and
P8, respectively. P8 and P9 were designed after it was found that P6 and
P7 gave a striation profile difficult to analyze. PI 1 and PI2 were used to
identify unambiguously
Copyright by ASTM Int'lthe
(allcracking sequence
rights reserved); Mon in
Deceach cycle and
7 14:40:45 EST also
2015 to
measure the influence
Downloaded/printed by of overloads or underloads. P10 was obtained by
"randomizing" the same number and levels of Smax and 5mln values present
in P9. Four pseudorandom programs were produced by alternately draw-
ing a maximum and minimum stress value from separate boxes. Program
P10 was then arbitrarily picked from among the four pseudorandom pro-
grams. In this paper, Programs P5 and P10 will be called random programs
although we know that they are not truly random.
CRACK PROPAGATION SEQUENCE
SHADOWING CONTRAST
FIG. 4—Sketches showing the formation of a striation. Crack advance occurs
only during the load rise part of the load cycle.
Results
Fatigue Cracking and Striation Formation

All the fatigue striations observed in this work were of the ductile
type [5]. The fatigue striations, which are well resolved at high magnifica-
tion, show a ridge or "saw tooth" profile with light and dark sides on each
side of the ridge. All the replicas were the two-stage plastic-carbon type
shadowed in the direction of crack propagation. In each fractograph the
local Copyright
crack propagation
by ASTM Int'l (alldirection is Mon
rights reserved); indicated by theESTblack
Dec 7 14:40:45 2015 arrow. By com-
paring the contrastby resulting from shadowing, we can determine unam-
Downloaded/printed
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authoriz
biguously the slope orientation on each side of a striation, Fig. 4. The

dark side of the striation (on either of the fracture halves) always faces
the crack front. By opposition, the brighter part of the fatigue striation
faces away from the crack front. The dark side of the striation always has
a rumpled appearance with fine, wavy, "slip" lines parallel to the crack
front, indicating heavy deformation. The bright side of a striation is flat
and generally featureless. The relative width of the dark to bright sides
FIG. 5—Striation profiles and spacing for parts of Programs P3, P7, and P9.
of a striation is quite uniform for a constant cyclic load. Changes of maxi-

mum load or amplitude result in marked changes in striation spacing,
striation contrast, and relative width of the bright to dark sides of a stria-
tion. A systematic analysis of these changes and correlation with the ap-
plied program, Fig. 5, led us to the following conclusions:
1. Fatigue cracking occurs only during a load rise or opening of the
crack, and accounts for the entire crack extension. Figures 6 and 7, show-
ing striations developed during Programs Pll and PI2, demonstrate this
finding clearly. In Fig. 6 the large bright striation is due to the load rise
B21-A1 at thebybeginning
Copyright ASTM Int'lof(all
Spectrum A, andMon
rights reserved); in Fig.
Dec 7 7it 14:40:45
is due toEST
the2015
load
rise A4-B1 at the end of
Downloaded/printed by Spectrum A. Our finding is in disagreement with
FIG. 6—Typical fracture surface resulting from Program Pll. Note the large
striation spacing (marked with small arrows) due to the load amplitude B21-A1
(9000 Copyright
psi), followed by three
by ASTM large
Int'l (all striations
rights corresponding
reserved); Mon Dec 7to14:40:45
the load cycles
EST 2015 of
Spectrum A (2a — 1.7 in.).
FIG. 7—Typical fracture surface resulting from Program P12. Note the large
striation spacing (marked with small arrows) due to the load amplitude A4-B1
(9000 Copyright
psi), preceded by three
by ASTM Int'lsmaller striations
(all rights corresponding
reserved); to the load
Mon Dec 7 14:40:45 ESTcycles
2015 of
Spectrum A (2a = 1.7 in.).by
Downloaded/printed
Hertzberg's finding [2] that cracking also occurs during unloading. This
can be explained by the assumption that Hertzberg mistook cracking due
to the tension part of the ripple he had programmed for cracking during
unloading.
2. During unloading, or closing of the crack, the two fracture surfaces
FIG. 8—Typical fracture surface due to Program P2. Only Spectra A and
B striations are observed (2a = 1.3 in.).
created during the preceding stress rise are heavily deformed near the
crack tip. This leads to the formation of the dark, rumpled side of the
striation and erases in part the bright appearance of the fracture face cre-
ated during the loading part of the cycle. The ratio of dark to bright sides
of the striation depends on loading versus unloading load amplitudes. But,
even when the unloading load amplitude is much larger than the preceding
load rise, the deformed width of the striation (dark side) is never greater
than Copyright
the crackbyadvance
ASTM Int'ldue
(all to thereserved);
rights preceding Monload
Dec rise. At this
7 14:40:45 ESTpoint
2015 we do
not clearly understand
Downloaded/printed bythe deformation mechanisms operating at the crack
tip during unloading which create the striation dark side. If crack tip
sharpening occurs during unloading, as observed by Laird and Smith [6]
and McEvily et al [7], we feel it is not associated with cracking during un-
loading.
Figures 5a, b, and c illustrate different fatigue striation profiles and
the corresponding load sequences. Since the exact shape and size of the
FIG. 9—Typical striation profile corresponding to Program P3. Striations due

to Spectra A, B, and C are observed (2a = 0.9 in.).
crack tip are unknown, they are represented by dotted lines in each sketch.
The height of the Striations is magnified for better illustration. These
sketches were confirmed by observation of matching areas on opposite
fracture faces and by stereo electron fractography.
Constant Maximum Load and Variable Load Amplitude Programs

The plots of
Copyright crackInt'l
by ASTM length versus
(all rights growth
reserved); rates7 14:40:45
Mon Dec were theESTsame
2015 for Pro-
gramsDownloaded/printed
PI and P2. Thebyfracture surfaces did not show any Striations corres-
University of Washington (University of Washington) pursuant to License Agreement. No further reprod
ponding to Spectrum A (AS = 2000 psi) of either program, and the pro-
grams could not be differentiated by looking at photomicrographs alone.
The only significant feature besides the change in striation spacing between
Spectrum B and C was a brighter striation corresponding to C77-A-B91
in PI and to C77-B71 in P2. In P2, this striation had the same spacing as
FIG. 10—Typical fracture surface due to Program P4. The sequence of appli-
cation of the load amplitudes does not seem to change the striation spacing mark-
edly (2a = 0.9 in.).
the other striations in Spectrum C but appeared brighter because of the

smaller width of the dark part of the striation, which corresponded to a
smaller unloading amplitude (B71). Figure 8 illustrates a typical P2 frac-
ture surface. In PI, the bright striation had the same appearance as in P2
but appeared to be larger than the other C striations, which would mean
that Spectrum A resulted in some limited crack extension but in no re-
solvable striation formation. Spectrum A does not show up even at the
longestCopyright
crack bylengths
ASTM where
Int'l (all striations are Mon
rights reserved); still Dec
observed for Spectra
7 14:40:45 EST 2015 B
and C.Downloaded/printed by
Programs P3, P4, and P5 were designed to determine the influence of

load amplitude and sequence at constant maximum load on the crack
growth rates. The characteristic fractographs of these programs are shown
in Figs. 9, 10, and 11.
Figure 12 gives a log-log plot of observed and calculated crack length
versus growth rate per program for P3, P4, and P5. (For P4, the rates
FIG. 11—Strlation profile corresponding to pseudorandom Program P5. The

marker shows the program repeat interval (2a = 0.9 in.).
plotted correspond to a half program.) The crack growth rates for the
random case, P5, were the same as for the programmed spectra, P3 and
P4. This shows that in the tests at constant maximum load the sequence of
load application did not measurably influence the overall crack growth
rates. This was confirmed by the fractographic analysis with the electron
microscope:
1. There was no marked crack front advance when the load amplitude
changed frombyone
Copyright ASTMspectrum
Int'l (all to another.
2. Downloaded/printed
After a change by of load amplitude, the spacing of the striations for
FIG. 12—Crack length versus crack growth rates for Programs P3, P4, and P5. The x indicates the crack length at final
fracture for each program. f;-
Copyright by ASTM
University of Washington
the next load amplitude sequence reached a stable and uniform value on
the first cycle of the new load amplitude sequence.
The average rates of crack propagation per load cycle for each load
amplitude sequence in Programs P3 and P4 were measured for various
crack lengths between 2a = 0.5 in. and 2a = 0.9 in. The growth rates per
cycle in each field of view were compared with the larger growth rate
(taken as unity) corresponding to AS1 = 11,400 psi.
Figure 13 is a plot of log AS1 versus log relative crack growth rate per
cycle. Paris [8] has proposed a growth rate equation of the form growth
FIG. 13—Relative crack growth rates at constant S ma x for Programs PI, P2,
P3, and P4.
rate = C (AS)*1. Figure 13 shows that for tests under programmed load
the exponent n varies markedly between 1 and 4.
Constant Load Amplitude and Variable Maximum Load Programs
Programs P6 and P7 caused markedly similar macroscopic crack
growth rates. On the fracture surfaces, the striations were never as well
defined as in Programs PI to P5. Spectra A, B, and C of P6 could be
easily identified. Spectrum D did not show any striation growth but
did create a very narrow and sharp compression groove before the large
crack jump at D7-A1, Fig. 14; this was the major load rise in P6. The
crackCopyright
jumpbydue ASTMtoInt'lthis load
(all rights change
reserved); was7 14:40:45
Mon Dec equivalent
EST 2015to the total crack
advance due to the three following cycles, A12, A23, and A34.
University of Washington (University of Washington) pursuant to License Agreement. No further reproductions authoriz
FIO. 14—Typical fracture surface due to Program P6. Note the large crack
jump (indicated by the small arrows) corresponding to D7-A1 (2a ~ 0.9 in.).
In the fractographs of P7, Spectra C, B, and A were easily identified,

Fig. 15, while Spectrum D did not give any visible cracking. Each change
of maximum load at C66-61-B11 and B55-51-A11 resulted in a larger
striation spacing between Cycle A112 and the following Cycles A223,
A334, and A441, although all these cycles had the same load amplitude
and the same maximum load.
FIG. 15—Typical fracture surface due to Program P7. Striations due to

Spectra A, B, and C are observed (2a = 1.7 in.).
Let us suppose that for a given crack length, at each maximum load
level S1 there is a corresponding stable crack tip radius r, and that for
Sz > Si we have rz > r\. The sequence P7-B55-A11 shows that when a
load level S2 (14 ksi) follows a load level Si (12 ksi), the first load cycle
All at level S2 is applied to a crack with the smaller tip radius ri. This
sequence results in a larger crack jump than normally expected from
level S2, followed by an immediate change of the crack tip radius from
ri to r-2 . Subsequent growth at S2 then occurs at a uniform rate. This
observation
Copyrightpoints
by ASTMout
Int'lthat an important
(all rights relationship
reserved); Mon Dec 7 14:40:45must exist between
EST 2015
FIG. 16—Crack growth rates versus crack length for Programs P8, P9, and P10.
FIG. 17—Fracture surface topography due to Program P9. The small arrows
define the large crack jump due to B16-A1 (2a = 0.9 in.).
crack tip shape or radius and the amount of cracking due to changes in
load level and load amplitude.
Programs P6 and P7 proved difficult to analyze, with too many levels
and too few cycles
Copyright at each
by ASTM loadrights
Int'l (all level. For thisMon
reserved); reason,
Dec 7Programs P8, 2015
14:40:45 EST P9,
and PIO were designed by
Downloaded/printed and tested. Figure 16 presents a plot of log crack
528 hhhhhhhhhhhhhhhhh
FIG. 18—Fracture surface topography due to Program PIO. The marker indi-
cates the program repeat internal (2a = 0.6 in.).
length versus log macroscopic crack growth rates for P8, P9, and PIO.
At lower growth rates, P9 was faster than P8, while the random program,
PIO, was always faster than P8 or P9. Figures 17 and 18 show typical
fractographs of P9 and PIO. Spectra A and B are well resolved, with all
the striations
Copyrightaccounted
by ASTMfor.Int'l
The(all
cracking
rightsdue to Spectrum
reserved); Mon C cannot
Dec 7 14:40:45 E
be resolved on the micrographs.
Downloaded/printed by The crack jumps due to change in
University of Washington (University of Washington) pursuant to Lic
maximum load levels P9-A20-C1, P9-C12-B1, and P9-B16-A1 are

clearly identified.
The average microscopic rates of crack growth per load cycle for
Spectra A and B of Programs P6, P7, P8, and P9 were measured within
each program and compared with the larger growth rate (taken as
unity) corresponding to Sma* = 14 ksi. The effect of the maximum load
amplitude on the relative crack growth rate per cycle is shown in Fig.
19. The scatter is quite large, but the plot shows that for the same load
amplitude a small change in maximum load results in large change in
growth rate. In the same figure, the relative microscopic rates confirm
the fact that P9 had a larger growth rate than P8. This is explained by a
crack growth retardation in Spectrum B of P8 because Spectrum B fol-
FIG. 19—Relative crack growth rates at constant load amplitude, variable

maximum loads for Programs P6, P7, P8, and P9.
lowed the higher load level of Spectrum A. The crack growth retardation
could be due to a work softening of the plastic zone by Spectrum B and
a consequent lower crack growth rate.
Program P10, as expected, was much faster than P8 and P9 [4]. The
large number of high load rise amplitudes coupled with many changes of
load levels accounts for this large difference in growth rates. In Fig. 18,
the striation spacings and profiles can be related to the applied random
loads if 5max , AS, and prior Smax are considered. With few exceptions,
only striations due to Smax = 14,000 psi are clearly resolved. This implies
that Smax effects are more important than AS or prior Smax effects.
Pll and PI2 were designed to show that cracking takes place only
during the opening of the crack. Spectrum A in Pll did not result in any
crack arrest since all the cycles of Spectrum B could be counted; on the
contrary, the first striations of Spectrum B were often observed to have
a slightly larger
Copyright spacing
by ASTM Int'lthan the reserved);
(all rights last one.Mon
This
Deccrack acceleration
7 14:40:45 EST 2015 at a
lower level following a high load level was not clearly observed in P9.
Computed Crack Growth Rates

An available computer program was used to calculate the crack growth
rates of all the programs tested. The crack growth data used in the com-
puter program were taken from previous tests at constant .Smax and con-
stant /?-values. These "in house" data correspond to different heats of
material and the difference between computed and measured growth
rates can be attributed to chemistry, environment, and testing variables.
However, since one practical aim of fatigue cracking tests is to predict
random crack growth from programmed or constant load crack growth
data, an attempt was made to use these data for our programs. The com-
puter calculated the crack advance for each load cycle and, consequently,
took into account the large changes in load amplitude at each change of
maximum load level. Figures 12 and 16 show plots of log crack length
versus computed rates for Programs P3, P4, P5 and P8, P9, P10. All the
computed growth rates were slower than the observed macroscopic rates
for the same crack length between la = 0.5 in. and 2a = 2 in. At larger
crack lengths, the computed rates were equal to the observed rates, and,
at very large crack lengths, the computed rates in many cases were faster
than the observed rates. We see that the computed growth rates differ
from the measured rates by a factor less than two, but the calculated
rates are not conservative in the long-life region. This can be explained
by the fact that although the cracking due to all the AS rises is accounted
for by the computer, the sequence in which the rises occur will give more
cracking than we would expect from the normal constant load, constant
amplitude data. Paris [8] and Smith [9] have also observed a similar
difference in crack growth behavior between sinusoidal and random
loading, random loading having a faster growth rate below 10~5 or 10~4
in./cycle. The computed crack growth curves are based on sinusoidal
crack growth data and therefore are equivalent to Smith's sinusoidal
load growth curves. The comparison between our results and Smith's
is therefore valid.
Discussion
The different program load sequences have clearly demonstrated that
crack extension in 2024-T3 aluminum alloy occurs only during the stress
rise part of a load cycle, but the striation profile and appearance depend
on both the loading and unloading amplitudes of the load cycle. The
smallest striation spacing observed was of the order of 1 /an. or 250 A.
The resolving power of the replicating technique is approximately 100 A.
In all the programmed load tests, whenever the striations were denned
well enough to be easily counted, we found a perfect one-to-one correla-
tion between the number of striations and the number of load cycles in
a spectrum.
CopyrightIn
bythe
ASTMrandom programs,
Int'l (all P5 and
rights reserved); P10,
Mon Deccorrelating striation
7 14:40:45 EST 2015
spacing to load cyclesbyin the program was more difficult, but with some
Downloaded/printed
time and effort it was done for all the micrographs taken. With training,
one can, by looking at a random load spectra, select the load levels and
load amplitudes which will account for the largest striation spacings
observed on the fracture surface.
Although we partly understand the sequence of striation formation
right at the crack tip, we still do not know the exact shape and size of
the crack tip. It is difficult to see how we can get a better crack tip image
with the existing tools and techniques. At this stage it is convenient to
assume a uniform crack tip shape, with the crack tip radius r becoming an
important variable which determines the intensity of the stress concentra-
tion at the tip of the crack. In a first approximation, it seems that r could
be closely related to the width of the part of the striation created during
the closing mode of the crack, which probably is a measure of the "re-
sharpening" of the crack tip during unloading. In this way, the im-
portance of a change of crack tip radius on the spacing was argued for
Program P7, Fig. 5b.
All our interpretations of fractographs show that for a given crack
length the amount of crack advance due to a change of load is an inverse
function of the crack tip radius estimated from the load history preceding
the change of load. It seems that fatigue cracking under random loads
will be better understood if some relationships among striation spacing,
crack tip radius, stress concentration factor, crack length, and load
sequence are established.
The microscopic crack growth rate measurements for Programs PI
to P5 and P6 to PI2 show a certain scatter. However, it is clear that a
simple relationship between growth rate, AS, and 5max does not exist.
The sequence of load application results in crack arrest and crack
retardation, and these sudden changes in growth rates cannot be ac-
counted for by a mathematical equation. The large dependence of the
crack growth rate on the maximum stress level is very striking and ex-
plains the large crack jumps observed on the fracture surface of random
load programs whenever there is a large peak overload. It also explains
the absence of striations due to the lower ,Smax applications.
In this work an attempt was made to separate the influence of AS and
Smax °n the microscopic crack growth rates. Paris [8] suggested that the
factor 7 = Smean/AS is a better variable. This factor varies between 0.5
and 1.5 for most of the test programs. Measurements of the microscopic
growth rates for test programs with constant 7 but having wide differences
in ,Smean and AS are certainly desirable.
Summary
In this paper it has been shown that fatigue crack propagation tests
underCopyright
program and random
by ASTM loads,
Int'l (all rights combined
reserved); with7 14:40:45
Mon Dec a study EST
of 2015
striation
counts and spacing measurements
Downloaded/printed by by electron microscopy, have given
us a better understanding of the mechanism of fatigue cracking. The

main findings with respect to 2024-T3 aluminum alloy are as follows:
1. The advance of the fatigue crack front takes place only during the
stress rise portion of a cycle.
2. The profile or sides of a striation are related to the loading and
unloading sequence of a load cycle.
3. The measured crack growth rate of program and random load
crack propagation tests were compared with rates calculated by a com-
puter program. The computed rates are not conservative but do not
differ from the measured rates by a factor of more than two.
4. Microscopic crack growth rates were measured and compared as a
function of load amplitudes AS and maximum loads 5"max .
5. In random load programs with constant Sm&s_, the growth rate is
the same as the programmed spectra rates. No sequencing effect is ap-
parent in changes in AS alone.
6. In random load programs with variable Smax , the sequence of load
application can markedly influence the crack growth rate on any one
cycle. The random crack growth rates were higher than the programmed
spectra rates.
A cknowledgment
The authors wish to thank S. H. Smith and W. E. Anderson for many
valuable discussions, T. Porter for preparation of the computer programs,
and W. C. Larson and W. D. Sump for conducting the mechanical tests.
References
[7] P. Forsyth and D. Ryder, "Fatigue Fracture," Aircraft Engineering, Vol 32,
April, 1960, p. 96.
[2] R. W. Hertzberg, Application of Electron Fractography and Fracture Me-
chanics to Fatigue Crack Propagation in High Strength Aluminum Alloys,
Ph.D. dissertation, Lehigh University, Bethlehem, Pa., May 15, 1965.
[3] J. C. McMillan, "Fractographic Analysis of Fatigue Crack Propagation,"
presented at technical sessions of Fractography Subcommittee of ASTM
Committee E-24, Lehigh University, Bethlehem, Pa., Oct. 5, 1965.
[4] J. Schijve, "Fatigue Life and Crack Propagation under Random and Pro-
grammed Load Sequences," Current Aeronautical Fatigue Problems, Per-
gamon Press, New York, 1965.
Ada Metallurgica, Vol 11, July, 1963, p. 703.
[6] C. Laird and G. C. Smith, "Crack Propagation in High Stress Fatigue," Philo-
sophical Magazine, Vol 7, 1962, p. 847.
Approach, Proceedings of the 10th Sagamore Conference, Syracuse University
Press, Syracuse, N. Y., 1964.
Interdisciplinary Approach, Proceedings of the 10th Sagamore Conference,
Syracuse University Press, Syracuse, N. Y., 1964.
[9] S.Copyright
H. Smith,by"Fatigue
ASTM Int'l (all rights
Crack Growth reserved); Mon Dec
under Axial 7 14:40:45
Narrow EST 2015
and Broad Band
Random Loading," Acoustical
Downloaded/printed by Fatigue in Aerospace Structures, Syracuse Uni-
versity Press,ofSyracuse,
University N. Y.,
Washington 1965. of Washington) pursuant to License Agreement. No fu
(University
DISCUSSION ON PROGRAM AND RANDOM LOADS 533
DISCUSSION
/. Schijve1 (written discussion)—The authors should be congratulated

on presenting most informative electrongraphs of fatigue fracture surfaces.
The pictures clearly illustrate interaction effects between load cycles of
different magnitudes and give conclusive eyidtnce ofjgrack growth oc-
curring during^ the increasing part of theJoad cycle cmly.
The authors refrained from describing in any detail the deformation
process at the tip of the crack during unloading. Personally, I believe that
FIG. 20—Crack tip sharpening during unloading.
crack tip sharpening occurs during unloading, although not exactly in

the way described by Dr. Laird in his paper, but rather as depicted in
Fig. 20.
The mechanism is in full agreement with the observations of the authors
and was in fact based on some similar work on 2024-T3 sheet material
recently carried out at our laboratory although on a rather limited scale.
It is also easily reconciled with the model outlined in my own paper.
The crack sharpening as described by Dr. Laird is probably more relevant
in the case of pure metals with strain hardening characteristics different
from those of the 2024-T3 alloy.
It is thought that the formation of the saw tooth profile of the striations
as outlined above is supported by the regularly repeating character of
1 Director of technicalbyservices, National Aerospace Laboratories, NLR, Am-
Downloaded/printed
sterdam, Holland.
successive striations. That implies that the crack tip sharpening process
is practically reproduced in successive load cycles. It is thought that this
will occur only if crack tip sharpening starts at the extreme apex of the
crack and then leads to a further closing of the crack tip. The direction of
closing is opposite the growing direction.
The direction of shadowing in our tests was opposite the growing
direction. The dark side of the striation (not closed during unloading
according to Fig. 20) was not completely featureless but showed deforma-
tion markings parallel to the crack front. The brighter part (closed during
unloading according to Fig. 20) showed similar markings and in addition
it made the impression of being "squeezed." Have the authors any experi-
ence with shadowing in a direction opposite the growing direction?
/. C. McMillan and R. M. N. Pelloux (authors)—We have shadowed
specimens in the direction opposite to the crack propagation direction to
verify our picture of striation profiles. The distinct differences between the
two sides of the striation were still apparent. Parallel markings on the
featureless side mentioned by Dr. Schijve have been observed under both
shadowing conditions, but the density of markings is much less than on
the rumpled side.
G. H. Jacoby2 (written discussion)—At the OVL (Deutsche Versvch-
sanstalt fur Luftund Raumfahrt) in Germany we have been studying for
several years the cumulative damage behavior under fatigue loading by
microfractographic methods. In the course of these studies we have also
found that a random sequence of load cycles leads to a higher crack propa-
gation rate than program loads.3 I think that the interesting paper of the
authors is a starting point for the more basic understanding of this very
important fact. If, however, we try to develop suitable program schemes
by applying microfractography to the crack propagation stage, it has to be
considered that the crack propagation and crack initiation stages may be
influenced in different ways. In our tests we have found that a specific
program and a random sequence of load cycles may lead to identical fa-
tigue lives, although in the latter case the crack propagation rate was
quite higher. This means that in the case of random loading the crack
initiation stage has been prolonged, while the crack propagation stage
has been shortened. More details including the definition of crack initia-
tion and crack propagation stage will be given elsewhere.4 From the
described studies it has to be deduced that the development of program
schemes can be based not only on the crack propagation stage alone, but
2
German Research Association and Aerospace Sciences, OVL; at present, Co-
lumbia University, New York, N. Y.
3
E. Gassner and G. Jacoby, "Experimentelle und rechnerische Lebensdauer-
beurteilung von Bauteilen mit Start-Lande-Lastwechsel," Raumfahrttechuik 11,
1965, Copyright
pp. 138-148.
4
G. Jacoby, "Application of Microfractography to the Study of Crack Propaga-
tion under Fatigue Stresses," to be published as an AGARD report.
DISCUSSION ON PROGRAM AND RANDOM LOADS 535
has to consider life and crack initiation as well. Nevertheless microfrac-

tography is a very useful tool for studying the influencing factors of pro-
gramming in detail, because no other method can record such small
differences in crack propagation rate as they occur from load cycle to load
cycle.
Messrs. McMillan and Pelloux—We agree with Dr. Jacoby's comments
and recognize the dangers in designing fatigue test programs only on the
basis of crack propagation rate comparisons. Our purpose in this paper
was to specifically separate initiation from propagation and to concentrate
on gaining an understanding of the crack growth stages.
Summary
Summary
It is a challenge of its own to recapitulate the major conclusions which

this symposium has brought us. No doubt it has added a lot of most valua-
ble experimental data, but it is not my intention to survey this material.
Rather, I would prefer to emphasize some of the highlights regarding how
to interpret available data and how to apply them to practical problems.
The symposium has certainly offered new views in this respect.
An outstanding feature of the collection of papers is its divergency of
approaches to the propagation of fatigue cracks. The ultimate goal of our
effort should be a synthesis of the various approaches in order to arrive
at a full understanding of the crack growth phenomenon. Technically
speaking, this should be interpreted as coming to accurate predictions on
crack growth, which obviously cannot be accomplished if we do not under-
stand the fatigue mechanisms.
The divergency of the papers may be illustrated by contrasting the
various starting points adopted by the authors:
1. Engineering approach versus the theoretical approach.
2. Empirical data versus microstructural studies.
3. Technical materials versus pure metals.
4. Macroscopic observations versus microscopic observations.
Although the divergency of approaches could be formulated in other
ways, the above contrasts, more or less, sum up the different standpoints.
It appears to me that one of the major contributions of the symposium is
that the authors have arrived at a bridging of the different points of view
on fatigue. This coupling has revealed new insights and will certainly
stimulate further investigations. Along the lines of the above contrasts, I
would like to summarize the advances emerging from the papers, and at
the same time to indicate certain gaps that still have to be filled. It is need-
less to say that such a summary cannot be complete since this would imply
rewriting parts of the papers. The selection of topics is naturally my per-
sonal one.
Engineering Approach Versus Theoretical Approach

Structures are generally very complex as compared with specimens
tested in the laboratory. A part of the fatigue problem for a structure can
be eliminated by the engineer by careful detail design and by incorporating
fail-safe features. Nevertheless, predictions on life, crack growth, and
strength in the cracked condition have to be made, and, for this purpose,
the designer obviously would prefer to adopt calculation rules. The paper
of Crichlow and Wells clearly illustrates that the designer cannot rely on
existing rules only. This is partly due to the complexity of the structure, but
539
also to the complexity of the fatigue process and the load history in
service. It appears to me that full-scale testing is going to be indispensable
for a good many years to come. Nevertheless, the papers of Crichlow and
Wells and of Christensen and Harmon show that the designer is aware of
the fact that he may considerably benefit from understanding his test
results. This understanding can lead to analytic functions (as shown by
Crichlow and Wells) that not only indicate trends, but also allow certain
predictions to other circumstances not directly covered by the tests. Since
full-scale testing is extremely expensive, it will be clear that efforts to solve
prediction problems under complex conditions are anyhow worth while.
If we now look at what the present theoretical approach may add to the
above problems, this symposium indicates certain advances that have
been made. The prediction of macrocrack growth rates using the stress-
intensity-factor concept has offered promising results. A new stimulus to
this concept comes from Rice, who considered the elastic-plastic problems
of crack tip deformation, and who showed that the stress intensity factor
could be applicable at high gross stresses. Second, the results of Swanson
show that the stress intensity factor is capable of correlating crack growth
rates under Rayleigh random load sequences of different intensities. In
my own paper it turned out that the stress intensity factor was capable of
correlating crack growth for small cracks in the microrange (0.1 mm).
Nevertheless, there are still serious limitations to the application of the
stress intensity factor to actual structures. A first step was successfully
made by Figge and Newman who predicted the crack growth in panels
with both end-loads and concentrated loads. A major obstacle appears
to be the calculation of K values for more complicated structures. It is
thought, however, that advances in stress analysis could be made with
the computers presently available.
Another major obstacle is the application to variable-amplitude loading.
Load cycles of different magnitude which are applied sequentially lead to
interaction effects, and these effects cannot be accounted for by the
stress intensity factor. The fractographic observations of McMillan and
Pelloux, of Hertzberg, and of Christensen and Harmon have shown that
we are certainly going to learn more about these effects. At the same time,
we could hope that this problem may come within reach of an analytic
treatment by means of the elastic-plastic analysis discussed by Rice.
Studies along both lines are worth while and will be mutually helpful.
Fatigue damage, by now, means more than just crack length. Residual
stress and strain hardening have to be accounted for, as illustrated by
Manson et al.
The nucleation period is less satisfactorily understood than the crack
propagation. The stress-intensity-factor concept fails here, and other
methods, probably based on the peak stresses at the notch root, will have
to be developed. One major issue in this respect, is the limited knowledge
SUMMARY 541
of the apparently weaker sites at which cracks are nucleated. For some
materials inclusions and intermetallic particles may be important. One
other very practical aspect that was not discussed in this symposium, is the
contribution of fretting corrosion to crack nucleation.
Empirical Information Versus Microstructural Studies
The empirical approach has provided a wealth of data, generally indicat-
ing trends that show how certain conditions can affect crack propagation.
In various papers, the effects of such factors as type of material, type of
alloy, heat treatment, thickness of the material, loading frequency, load-
ing direction, biaxial stress, etc., were investigated and systematic trends
were observed. The designer should take notice of these results and feed
them into his design philosophy. The most difficult factor is probably the
environment, which was analyzed by Achter and by Wei et al. Results were
presented by Crooker and Lange, by Wei et al, and in my own paper.
Further studies including the frequency effect will be most worth while
for both practical and theoretical reasons. It is realized, however, that this
is far from easy.
It is the task of the microstructural approach to arrive at explanations
of the above influences on crack growth. This, however, cannot be done
without an elementary understanding of the fatigue crack extension
process. Laird proposed the plastic relaxation model based on crack blunt-
ing and sharpening. In my own paper I have suggested that crack growth
may be considered as a consequence of cyclic slip, that is, cyclic sliding-off
at the tip of the crack. The conversion of cyclic slip into crack extension
is promoted by the local tensile stress normal to the crack. This model may
qualitatively account for mean stress and residual stress effects. The model
of Laird and my own are not too much different. They both can be
reconciled with the finding of McMillan and Pelloux that crack extension
occurs only during the loading part of the cycle. However, to arrive at a
quantitative evaluation, one has to know the local stresses and strains at
the tip of the crack. How difficult this problem is becomes evident in the
paper of Rice, who surveyed and analyzed the literature. The problems are
not only of a mathematical nature; in fact, our information on the
processes occurring at the tip of the crack are not yet sufficiently detailed.
It is expected that the electron microscope will give us more valuable
information in the future. The paper of Grosskreutz has indicated new
possibilities in this respect. He found practically no evidence of interactions
between dislocations and precipitated zones in aluminium alloys. His data
on dislocation densities in the immediate vicinity of the crack may stimu-
late further thoughts on fatigue mechanisms in these alloys.
Technical Materials Versus Pure Metals
The complexity of the technical materials is well known. They are
strengthened by various means. Apart from cold work, the pure metals
are less complex and their plastic behavior is better understood. Although
results on these metals may not be directly applicable to the technical ma-
terials, they may still be useful since it can lead to a better understanding
of cyclic slip. Laird has summarized and analyzed existing information
starting from the pure metals and subsequently considering various
strengthening mechanisms. In Hoeppner's paper, the crack propagation
in copper and the effect of grain size on the growth are described. Although
it has been frequently said before, it may be repeated here: fatigue can be
different from material to material. Even in one material, it may exhibit
different forms, for example, under tensile loading and under torsion
loading. For the metal physicist, it is generally instructive to study a
variety of materials. For practical problems, on the other hand, it is de-
sirable that fundamental studies on engineering materials not be neglected.
Macroscopic Observation Versus Microscopic Observations
The differences between macroscopic observation and microscopic
observation were most prominently manifested by studies of the transition
from the tensile mode to the shear mode fatigue fracture on the one hand
and fractographic observations on the other hand. It is evident that
information obtained both on the macroscale and the microscale is re-
quired for a continuum mechanics approach.
The transition from the tensile mode to the shear mode is now generally
attributed to a transition from plane strain to plane stress. Interesting ob-
servations were made by Swanson in tests where a constant stress intensity
factor was maintained throughout the test. Also, the analysis of data by
Wilhem supports the idea that the crack rate is different during the two
modes of failure. Herzberg's fractographic work seems to confirm this.
An important microscopic observation was offered in Laird's paper where
indirect, but still convincing, evidence was shown that crack growth in
Stage I occurs in each load cycle in spite of the apparent absence of stria-
tions.
Since the macroscopic and the microscopic observations are both con-
tributing to the description of the fatigue phenomenon, further studies
are worth while in order to minimize the amount of speculation on various
aspects.
Several authors, in conclusion to their papers, have stated that they had
not solved the problem put forward in their introductions. Nevertheless,
I personally prefer to hold the optimistic view that the symposium has
shown that our grip on the fatigue crack propagation problem is increas-
ing and that we may expect this to continue in the future. The interaction
between various disciplines must continue to be fruitful to accomplish this
end.
J. Schijve
National Aero- and Astronautical Institute
Amsterdam
THIS PUBLICATION is one of many
issued by the American Society for Testing and Materials
in connection with its work of promoting knowledge
of the properties of materials and developing standard
specifications and tests for materials. Much of the data
result from the voluntary contributions of many of the
country's leading technical authorities from industry,
scientific agencies, and government.
Over the years the Society has published many tech-
nical symposiums, reports, and special books. These may
consist of a series of technical papers, reports by the
ASTM technical committees, or compilations of data
developed in special Society groups with many organiza-
tions cooperating. A list of ASTM publications and
information on the work of the Society will be furnished
on request.

STP 415-1967

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

STP 415-1967

Uploaded by

Copyright:

Available Formats

Fatigue Crack Propagation

ASTM SPECIAL TECHNICAL PUBLICATION NO. 415

List price $30.00; 30 per cent discount to members

The Society is not responsible, as a body,

Printed in Baltimore, Md.

The Symposium on Fatigue Crack Propagation was presented in four

Five Year Bibliography on Fatigue, STP 9BB (1966),

Fatigue Tests of Aircraft Structures: Low-Cycle, Full-

Fracture Toughness Testing and Its Applications, STP

Plain Strain Crack Toughness Testing of High Strength

Review, Analysis, and Discussion of the Fatigue Crack Growth Problem

Crack propagation is the one phase of fatigue failure which can be

experimental results which became available after the original Sym-

REFERENCE: R. H. Christensen and M. B. Harmon, "Limitations of

It is well known that the phenomenon of progressive fracture caused

1935 to 1936.. . DeForest and No equation, but one of earliest ex- 1

FIG. 1—Agreement between test and theory—cycles required to generate

FIG. 2—Variation in crack rate with length of crack.

characteristic properties of uniaxial and biaxial loading can be very dif-

Fatigue-Crack Growth Equations

FIG. 4—Variation in crack rate with stress intensity.

of growth for a crack of a specific length is accelerated twenty-fold if

function of stress intensity K, which is proportional to the product of

FIG. 6—Crack growth rate as a function of stress field.

ratio of minimum to maximum stress varied from 10 to 65 per cent. It

shorter or longer crack lengths show a family of curves. It can also be

FIG. 7—Effect of temperature.

The Effect of Multiaxial Strain

FIG. 9—Effect of frequency at elevated temperature.

obtained in a Douglas independent research program and presents crack

The Effect of Temperature

FIG. 10—Growth of fatigue cracks as a function of temperature (schematic).

The Effect of Frequency

FIG. 11—Anisotropic flaw growth.

FIG. 12—Fracture strength as a function of flaw orientation.

decreases. In general, critical crack lengths decrease as temperature de-

The Effect of Anisotropy

FIG. 13—Growth of cracks as a function of type of test (laboratory) loading.

FIG. 14—Typical damage accumulation for laboratory tests of complex struc-

sheet will be normal or parallel to the principal loading direction. Further,

The Effect of Variable Stress Fields

section stress or the maximum cyclic load is constant throughout the

FIG. 16—Tension area cracked as a function of elapsed fatigue life.

FIG. 17—Fatigue crack growth in titanium tension field web.

served; this is followed by the crack propagation period. For a given

FIG. 18—Fatigue crack striations due to programmed stress cycling.

Figure 16 shows normalized data of tension area cracked as a function

FIG. 19—Striation spacing, microscopic versus macroscopic growth.

analysis of actual flight-vehicle components. Some studies have even

macrogrowth measurements. As stress level increases, so does the growth

FIG. 20—Comparison of predicted and measured crack growth under random

development, it is quite difficult to identify operating stress levels in a

Crack Growth Under Random Loading

FIG. 21—Fracture modes.

Additional experimentation is required to develop accurate techniques

Critical Element and Fracture Mode

[3] A. K. Head, 'The Growth of Fatigue Cracks," The Philosophical Magazine,

Gerhard H. Jacoby1—Program tests and tests with a random sequence

REFERENCE: W. J. Crichlow and R. H. Wells, "Crack Propagation

b Spacing of straps or stiffeners, in.

5. to investigate the effects of various attachment methods: rivets,

TABLE 1—Design configuration for stiffened cylinders, Titanium 8Al-lMo-lV

1 0.050 Ta 0.032 2.5 0.025 1.0 20.0

Outline of the Test Program