Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

J Mater Sci (2015) 50:6539–6551

DOI 10.1007/s10853-015-9219-2

REVIEW

Low-temperature creep in pure metals and alloys


M. E. Kassner1 • K. K. Smith1 • C. S. Campbell1

Received: 27 May 2015 / Accepted: 27 June 2015 / Published online: 7 July 2015
Ó Springer Science+Business Media New York 2015

Abstract Many crystalline materials are known to exhibit plasticity at lower temperatures. However, some materials
creep at low temperatures (T \ 0.3Tm). Here, we review do demonstrate significant creep at T \ 0.3Tm and, further-
and analyze the phenomenological relationships that more, at stresses both above and below the macroscopic yield
describe primary creep. The discussion focuses on the stress (r0:002
y ). These materials include Ti-alloys and steels
controversy as to whether power-law or logarithmic [1–10], Al–Mg [11], a-Brass [12], ionic solids [13], pure Au,
descriptions better describe the experimental database. We Cd, Cu, Al, Ti, Hg, Ta, Pb, Zn [14–28], and precipitation
compile data from the literature as well as new copper data hardened alloys [29] (as well as glass and rubber [28]).
recently taken by the authors. Depending on the material, it Materials may undergo plasticity that affects its intended
appears that the logarithmic form can somewhat better (e.g., structural and electrical [30–32]) performance. How-
describe creep behavior at low temperatures, while the ever, the mechanism(s) of low-temperature creep are not yet
power-law behavior manifests at intermediate tempera- established. Seeger et al. [19, 33, 34] proposed a dislocation
tures. The basic mechanism(s) of low-temperature creep intersection mechanism for pure FCC metals under polyslip.
plasticity is examined, as well. Others have suggested that creep occurs due to quantum
mechanical tunneling of dislocations at very low tempera-
ture [13, 20, 22, 24, 26, 27], but there is no consensus as to the
Introduction actual mechanisms involved.
In an empirical analysis, Neeraj et al. [1] assumed the
Creep of metals and ceramics occurs over three broad tem- Dorn and Holloman flow equations:
perature ranges: high (T [ 0.6Tm), intermediate (0.3Tm \ - r ¼ Ken e_m ; ð1Þ
T \ 0.6Tm), and low (T \ 0.3Tm). This paper concerns itself
where K is the strength parameter, n is the strain-hardening
largely with the low-temperature range. Less attention has
exponent, and m is the strain-rate sensitivity exponent. This
been paid to this range due to the fact that many materials
equation was found to reasonably describe some Ti-alloy
generally do not experience significant time-dependent
behavior. They showed
de  r m1  n
e_ ¼ ¼ e m; ð2Þ
dt K
& M. E. Kassner Z Z  1=m
kassner@usc.edu r
en=m de ¼ dt: ð3Þ
K. K. Smith K
kamiasmi@usc.edu
So that under constant stress,
C. S. Campbell  r mþn
1 
m þ nmþn m
m
campbell@usc.edu
e¼ tmþn : ð4Þ
K m
1
Department of Aerospace and Mechanical Engineering,
OHE430, University of Southern California, Los Angeles, Equation 4 now fits the general form, e ¼ atb . This
CA 90089-1453, USA represents the power-law behavior, which will be discussed

123
6540 J Mater Sci (2015) 50:6539–6551

at length, later. This model suggests more pronounced low- Eq. (11) is the proper form. This contention is not well
temperature creep in materials with low values of n and established and will be discussed in detail in this paper.
moderate values of m.
This conclusion makes sense based on the arguments
concerning the mechanisms of creep, in general [35]. Experimental procedure
Creep is often associated with constant stresses, as
opposed to constant strain rates. For a constant strain-rate The Metron Group in Scottsville, Virginia provided the
test there is an observed yield stress. A ‘‘drop’’ in the copper rod used in this study with the measurements of
(constant) stress below this yield stress will still cause 1.219 m length 9 15.875 mm diameter and a purity of
time-dependent plasticity, or ‘‘creep,’’ at the lower stress 99.999 %. The samples were machined into threaded tensile
and lower creep rate, which becomes more substantial for specimens with a gage section of 34 mm and a diameter of
higher strain-rate sensitivity (m) and lower strain hard- 6.55 mm. Specimens were tested at a fixed load on a creep
ening (n). rupture test machine made by the Arcweld Manufacturing
Generally, but not always, low-temperature creep is Company. Strain was measured using Omega full-bridge
considered to be primary creep that never reaches a gen- strain gages. The samples were annealed at 420 °C for 2 h in
uine mechanical steady state. Long ago, Andrade [36] and a vacuum furnace with the lowest pressure reaching
Orowan [37] suggested that primary creep at high tem- 6.4 9 10-6 mbar. The average grain size was measured as
peratures obeys a power-law behavior with an exponent of 301 lm by etching with a solution consisting of 7.5 mL
1/3: 35 wt% ammonium hydroxide (NH4OH), 7.5 mL DI H2O,
and 1.5 mL 3 wt% Hydrogen Peroxide (H2O2).
e ¼ bt1=3 þ c1 : ð5Þ
Evans and Wilshire [38] reviewed the high-temperature
primary creep equations and suggested a refinement that Low-temperature creep behavior of various metals
led to an equation of the form and alloys
e ¼ at1=3 þ ct þ dt4=3 : ð6Þ
In this section, we examine the phenomenological trends in
Variations to this equation include [39] the literature for various materials with an objective toward
e ¼ at1=3 þ ct ð7Þ trying to understand how well, and to what extent, they
obey power-law or logarithmic behavior.
and [40]
Titanium Alloys
e ¼ at1=3 þ bt2=3 þ ct ð8Þ
or [41] Neeraj et al. [1] carefully described the low-temperature
creep behavior of Ti-alloys. Figure 1 (taken from Neeraj
e ¼ atb þ ct ; ð9Þ
et al.) plots ambient-temperature creep data [2, 10, 44] for Ti-
where 0 \ b \ 1 or simply, alloys with different microstructures and different compo-
sitions. A particularly interesting characteristic of this data is
e ¼ atb ; ð10Þ the extension of creep tests to longer times (sometimes over a
where 0 \ b \ 1 month). Equation (10), the power-law relationship, appears
Note that Eqs. (5–10) are all of a similar, power-law to better describe the data. In contradiction to some of the
form, as suggested in the Neeraj et al. [1] work discussed earliest phenomenological equations (albeit at higher tem-
earlier. peratures) [36], the exponent of b = 1/3 used in Eqs. (5–8)
Another set of phenomenological equations, with a does not fit the data, and a smaller value b = 0.2 seems to be
logarithmic rather than power-law form, was suggested by better. Also, note that the applied stresses are all below the
Phillips [28], Laurent and Eudier [42], and Chévenard [43] macroscopic yield stress (determined at a conventional strain
for low temperatures, rate) where creep is less expected. As Neeraj et al. point out,
other literature confirms that 0.03 \ b \ 1. Cottrell [45–47]
ep ¼ alnt + c2 : ð11Þ
and Nabarro [48] recently tried to theoretically justify a value
Wyatt [18], long ago, suggested for pure metals, such as for b = 1/3 for power-law creep at low temperatures but that
Al, Cd, and Cu, that at higher temperatures, Eq. (5) is the value does not appear to be universally valid. Other creep
proper descriptive equation, but at lower temperatures, work on titanium has been performed by others as well [49].

123
J Mater Sci (2015) 50:6539–6551 6541

(a)
0.018 Logarithmic
Ti-Alloys

0.016 T = 0.15Tm

0.014
Ti-6211 Basketweave
Ti-5-2.5 R = 0.93447
0.012
Ti-6211 Colony = 667 MPa
y= 834 MPa
0.01 Ti-6-4 Colony = 8.3 x 10-4 s-1
Strain

R = 0.82161
= 637 MPa
0.008 y= 796 MPa
= 5 x 10-3 s-1

0.006 R = 0.92414
= 608 MPa
y= 760 MPa
0.004 = 5 x 10-3 s-1
R = 0.98529
= 870 MPa
0.002
y = 1088 MPa
= 8.3 x 10-4 s-1
0
1 10 100 1,000 10,000 100,000 1,000,000 10,000,000
Time (sec)

(b) 0.1 Power-Law


Ti-Alloys

T = 0.15Tm R = 0.86482
= 667 MPa
y = 834 MPa
= 8.3 x 10-4 s-1
Ti-6211 Basketweave R = 0.8773
Ti-5-2.5 = 637 MPa
0.01 y= 796 MPa
Ti-6211 Colony = 5 x 10-3 s-1
Ti-6-4 Colony
Strain

R = 0.94396
= 608 MPa
y = 760 MPa
= 5 x 10-3 s-1

0.001
R = 0.98529
= 870 MPa
y= 1088 MPa
= 8.3 x 10-4 s-1

0.0001
1 10 100 1,000 10,000 100,000 1,000,000 10,000,000
Time (sec)

Fig. 1 Ambient-temperature creep for various Ti-alloys of different log plot. Based on the R2 values, a power law may better describe the
microstructures and different compositions. From Neeraj et al. [1]. creep behavior
a plots the data in a log–log plot while b plots the same data in a semi-

Steels macroscopic yield stress. One of the tests was conducted at an


applied stress of just half the yield stress (again, assessed at a
AISI 4340 steel conventional strain rate between 10-5 and 10-4 s-1). The
duration of the tests is relatively short, performed over, at
Oehlert and Atrens [5] performed creep studies on steel at most, 20 min and this is unfortunately short. The nature of the
ambient temperature where the applied stress is below the phenomenological equations may not be fully assessed with

123
6542 J Mater Sci (2015) 50:6539–6551

such short testing times. Nonetheless, the creep behaviors room temperature (0.22Tm). Figure 3c illustrates how the
were best described by a logarithmic equation (i.e., Eq. 11): behavior (PL vs. Log) changes with time and stress.
ep ¼ epy þaln(t); ð12Þ Overall, at ambient temperature, power law is evident at
smaller times and lower stresses (lower strains), whereas
where epy is the plastic strain on loading and is a function of logarithmic is more evident at longer times and higher
the applied stress. One clumsiness with Eq. (12) is that, at stresses [t [ 105 s and r [ 20 MPa (ry & 20 MPa and
t = 0, an infinite creep rate is predicted although the strain e_ = 10-4 s-1)]. At 77 K, the work of Wyatt [18] (very
should be equal to epy . The clumsiness was eliminated by short times) and Yen et al. [30], evince both logarithmic
modifying the equation to and power-law behavior as confirmed by our R2 analysis.
ep ¼ epy þalnð1 þ btÞ; ð13Þ At 0.3Tm (transition to intermediate-temperature creep), the
data of Jenkins-Digges [50] and Wyatt suggest power law
where a and b are constants. Oehlert and Atrens found that, better describes the data. This is also confirmed by our R2
over the range of stresses studied, the constant a varied by a analysis.
factor of nearly 20 and b by a factor of 2. They also examined
3.5NiCrMoV and AeroMet100 over similar time ranges and
Aluminum
observed a similar logarithmic phenomenology. By contrast-
ing the differences in the logarithmic and power-law fits, our R2
The creep behavior of high-purity aluminum of Wyatt [18]
correlation factor analysis is consistent with their conclusions.
and Sherby et al. [51] is shown in Fig. 4. The 300 K
(0.32Tm) data shown in Fig. 4a, b, exhibits power-law
304 Stainless steel
behavior (Eq. 10) perhaps because the temperature is high
enough to transition from a low-temperature to intermedi-
Figure 2 shows the creep curves for annealed 304 stainless
ate-temperature creep regime as also observed in Cu.
steel (earlier work by the authors) [7], a material that evinces
Again, test times are unfortunately relatively short. How-
some typical features of low-temperature creep. The applied
ever, the 77 K (0.08Tm) data [51] also exhibits a somewhat
stresses are both above and below the yield stress (here,
better fit with a power-law equation from the Sherby et al.
ry ¼ 221 MPa at e_ ¼ 3  104 s-1 at room temperature). data for longer times. However, the Wyatt data for shorter
Figure 2a is a semi-log plot and shows the full time span of times show logarithmic behavior. The lower temperature
the tests. Note that times, in some cases, extend to nearly a tests may be more reliable in assessing low-temperature
year, although tests above the yield stress were generally creep, in that additional restoration mechanisms other than
performed over shorter times (a week or less). For the shorter dynamic recovery may occur [52, 53] toward the higher
tests above the yield stress, strains were measured using an testing temperatures (but still in the low-temperature creep
extensometer and for the longer tests, below the yield stress, regime).
strain was measured using an optical comparator. Based on
the R2 quality-of-fit factor, it appears that the creep data for
this material, at ambient temperature, best follow a loga- Cadmium
rithmic behavior at a fixed stress, r. This is similar to Eq. (12)
that represented the ferritic steel data of Ohlert and Atrens. Figure 5 illustrates the creep behavior of pure Cd at 77 K
epy can be approximated by the usual relationship, (0.13Tm) and 300 K (0.51Tm). The low-temperature
behavior over a period of just one hour may evince loga-
epy ¼ a þ br; ð14Þ
rithmic behavior, while the higher temperature data seems
where a and b are constants, and a basically reflects the strain better described by a power law, again, consistent with Cu
on loading. Also, b is the slope in Fig. 2b and appears to and Al.
decrease with decreasing stress, approximated by, Table 1 summarizes the overall phenomenological
behavior of the metals and alloys discussed. Trends are
b ¼ kr þ Co ; ð15Þ
described for both above and below the conventional
where k and Co are constants. yield strength of the metals as well as for shorter and
longer creep times. Obviously, more data is desirable,
Pure metals especially for longer times. Overall, logarithmic behavior
appears to be more commonly observed. It should be
Copper mentioned that we have performed the graphical analysis
of all the materials that the table is based off of but
Figure 3 illustrates the new data of this study. Figure 3a, b these graphs were not included in light of the brevity of
show logarithmic and power-law axes, respectively, at this article.

123
J Mater Sci (2015) 50:6539–6551 6543

(a) 0.050
Logarithmic
276 MPa 304 Stainless Steel
0.045 293 MPa R = 0.99999
R = 0.99996 259 MPa T = 0.17Tm
R = 0.99999 y= 221 MPa
0.040 = 3 x 10-4 s-1

0.035
241 MPa
R = 0.99998
0.030
R = 0.99955
Strain

0.025 Extensometers

Optical Comparator
0.020 R = 0.99994 Measurements

0.015 224 MPa


R = 0.99976 207 MPa 190 MPa
R = 0.99973 R = 0.9898
0.010
172 MPa 155 MPa R = 0.98928
0.005 R = 0.99373
145 MPa R = 0.99958
138 MPa R = 0.9937
0.000
1 10 100 1,000 10,000 100,000 1,000,000 10,000,000
Time (sec)

(b) 0.1
276 MPa 259 MPa
293 MPa
R = 0.98635 R = 0.98101 241 MPa
R = 0.99743
R = 0.96104
R = 0.99077
R = 0.97846 190 MPa 172 MPa
207 MPa R = 0.99658 R = 0.94276
0.01 R = 0.92614
224 MPa 155 MPa R = 0.95812
R = 0.95734
145 MPa R = 0.96687
Strain

138 MPa R = 0.98409

0.001
Power-Law
304 Stainless Steel
Extensometers
T = 0.17Tm
Optical Comparator y= 221 MPa
Measurements = 3 x 10-4 s-1

0.0001
1 10 100 1,000 10,000 100,000 1,000,000 10,000,000
Time (sec)

Fig. 2 The plastic strain vs. time behavior of annealed 304 stainless steel under different stresses from [7, 59]; a Linear strain vs. log time; b Log
strain vs. log time. Based on the R2 values, a logarithmic equation better describes the data

Discussion of the mechanisms of low-temperature used by a variety of investigators but the meaning is
creep unclear. It appears that some common descriptions of
exhaustion creep are that at low temperatures, we generally
Logarithmic creep observe primary creep that never reaches a mechanical
steady state, a balance of hardening and recovery pro-
The earliest explanation of low-temperature creep by Mott cesses. It is often suggested that the decrease in creep rate
and Nabarro [48, 54] was based on dislocation glide and is a result of a decrease in the mobile dislocation density.
dislocation exhaustion. The term ‘‘exhaustion creep’’ is This could be a consequence of an increase in the energy

123
6544 J Mater Sci (2015) 50:6539–6551

Fig. 3 Strain vs. time data for (a) 0.014


copper in this study at various Logarithmic
99.999% Polycrystalline Copper
stresses and at ambient
temperature (T = 0.22Tm). 0.012 T = 0.22 Tm R = 0.93292
y = 20 MPa
a Linear strain vs. log time; = 10-4 s-1
b Log strain vs. log time; c the 0.01 30 MPa
variation in power-law and log
20 MPa
behavior with time and stress.
0.008 20 MPa
The yield stress of copper was

Strain
reported by the authors at a 18 MPa

strain rate of 10-4 s-1 0.006


R = 0.73056
R = 0.82685
0.004
R = 0.91202

0.002

0
1 10 100 1,000 10,000 100,000 1,000,000 10,000,000
Time (sec)

(b) 0.1
Power-Law 30 MPa
99.999% Polycrystalline Copper
20 MPa
T = 0.22 Tm 20 MPa
y = 20 MPa
= 10-4 s-1 18 MPa

R = 0.91943
Strain

0.01

R = 0.73469

R = 0.84272

R = 0.91763

0.001
1 10 100 1,000 10,000 100,000 1,000,000 10,000,000
Time (sec)

(c) Change in Correlation Factor Over Time


0.015 99.999% Polycrystalline Copper
Logarithmic
Dominant
18 MPa
0.01
20 MPa
30 MPa
R2 (Log - PL)

0.005

0
1 10 100 1,000 10,000 100,000 1,000,000 10,000,000

-0.005

Power-Law
Dominant

-0.01
Time (sec)

123
J Mater Sci (2015) 50:6539–6551 6545

Fig. 4 Cryogenic temperature (a) 0.30


(a, b) creep behavior of pure Al Logarithmic
based on data by [18, 51]. Aluminum Sherby et al.
Logarithmic behavior at 77 K = 152 MPa
may dominate at lower stresses. 0.25 T = 0.08 Tm (77K) R = 0.92887
A yield stress at a given strain
rate was not reported by either
reference for their Aluminum
samples 0.20

Strain
0.15

= 59 MPa
0.10
R = 0.9953

0.05 Wyatt et al.


= 19.5 MPa
R = 0.99324
0.00
1 10 100 1,000 10,000 100,000
Time (sec)

(b) 1
Power-Law
Aluminum Sherby et al.
= 152 MPa
T = 0.08 Tm (77K)
R = 0.95109

= 59 MPa
0.1
R = 0.97022
Strain

Wyatt et al.

= 19.5 MPa
R = 0.98094
0.01

0.001
1 10 100 1,000 10,000 100,000
Time (sec)

required to surmount obstacles because of dislocation Burger’s vector, b, the difference between the applied
multiplication (i.e., hardening). Alternatively, the disloca- stress, s, and the ‘‘back,’’ or long-range internal stress
tion density may not change and the distribution of the (LRIS), sG, due to other dislocations, is the energy sup-
obstacle energy increases with plasticity without multipli- plied. The activation area is usually defined as the product
cation, instead arising from a statistical change in the dis- of the width of the obstacle, d, and the obstacle spacing, ‘.
tribution of obstacles. Alternatively, the activation volume, v, equals Ab = ‘b2 .
These descriptions often are placed in the context of the This leads to the classic rate equation [19, 33, 55],
dislocation intersection mechanism by Seeger et al. [33,  
34] where the concept of activation volume and area were DHo  vðs  sG Þ
c_ ¼ NAbmo exp ; ð16Þ
probably first used. The product of the activation area, A, kT

123
6546 J Mater Sci (2015) 50:6539–6551

Fig. 5 Creep behavior of pure (a) 0.07


Cd, at 0.51Tm and 0.08Tm,
based on data by [18]. At low
Logarithmic
99.999% Cadmium
temperatures, logarithmic
0.06
behavior (T \ 0.30Tm) may best (Data from Wyatt et al.)
describe the data but power-law
behavior may be better at higher
temperatures. A yield stress at a 0.05
given strain rate was not
reported by Wyatt [18] for the
Cadmium data 0.04

Strain
= 59 MPa
T = 0.13Tm
0.03 R = 0.98352
= 20 MPa
T = 0.51Tm
0.02 R = 0.83012

0.01

0.00
1 10 100 1,000 10,000
Time (sec)

(b) 1
Power-Law
99.999% Cadmium

(Data from Wyatt et al.)

0.1
= 59 MPa
T = 0.13Tm
R = 0.89796
Strain

= 20 MPa
0.01 T = 0.51Tm
R = 0.99831

0.001
1 10 100 1,000 10,000
Time (sec)

where c_ = strain rate, N = number of dislocation segments Zc


per unit volume held up at the intersection points of mean sG ¼ soG þ hdc; ð17Þ
spacing, ‘, mo is an atomic frequency of the order of the 0
Debye frequency, DHo = energy required for the inter- where soG is the stress due to the dislocations initially in the
1
section process, i.e., the energy for jog formation & 10 Gb2 crystal and h is the strain-hardening coefficient, which is
[56]. defined as ds=dc.
According to Conrad [19], Conrad further suggested

123
J Mater Sci (2015) 50:6539–6551 6547

Table 1 Materials and regimes where either logarithmic (Log) or troubling aspects to this analysis. The logarithmic behavior
power-law (PL) description apply to low-temperature creep behavior stems from Eq. (17) which leads to the -hc term in the
T \ 0.3Tm numerator of the exponential in (18). sG in Eq. 17 must be
about zero [47, 58] and the inclusion of hc in Eq. 18 is
Overall t\1 h t[1 h
unnecessary. Without the inclusion of hc, Eq. 21 is
r \ ry r [ ry r \ ry r [ ry impossible. Greater detail for the derivation of Eq. (17) is
Ti-alloy Mostly PL Mostly PL presented [58] by the authors. Other attempts to justify this
304 SS Log Log Log Log Log equation were made by Wyatt [18], but the methodology
Ferritic Steel Log Log
was unclear. Welch et al. [60] appeared to attempt a similar
Aluminum Log/PL Log Log PL
approach to that of Wyatt.
Perhaps, a more realistic approach may be found using
Copper Log/PL Log/PL Theo-PL PL/Log
the Seeger intersection mechanism equation in a different
Cadmium Log Log
way. The plastic shear strain is defined as
cp ¼ xqm b: ð22Þ

Assume cp & 0.01.


  
DHo  v s  soG  hc Here qm = mobile dislocation density; x = average
c_ ¼ NAbmo exp : ð18Þ distance a dislocation moves
kT
Assume x ¼ p1ffiffiffiffi
qT ; where qT = total dislocation density.
Integrating pffiffiffiffiffi
    cp 0:01 qT
kT DHo  v s  soG  hc qm ¼ ¼ ; ð23Þ
exp ¼ NAbmo t þ D: xb b
vh kT
b (for Cu) = 2.55 9 10-8 cm.
ð19Þ
In annealed copper, assume qT ¼ qo  2  108 cm2
So, Multiplication of dislocations is likely occurring so we
   can assume
DHo  v s  soG  hc
¼ lnðv0 t þ D0 Þ; ð20Þ qT ¼ Mo c þ qo ; ð24Þ
kT
where where Mo is the dislocation multiplication constant and
 qT ¼ ð0:84  1011 cm2 Þc þ qo for copper [61].
NAbmo vh Starting with Seeger’s Eq. (16), we can rephrase many
v0 ¼ :
kT of the quantities in terms of more fundamental quantities.
Rearranging Consider again Eq. (16) where, A = activation
 area = d0 ‘  ‘b; v = activation volume = Ab ¼ ‘b2 ;
kT DHo þ vðs  soG Þ
c¼ lnðv0 t þ D0 Þ  : ‘ = average dislocation segment length = p1ffiffiffiffi
q ¼
pffiffiffiffiffiffiffiffiffiffiffiffi
1
Mo cþqo
hv hm T

From [57, 58] we must assume sG = backstress & 0.


The constant D0 should be chosen so that cð0Þ = 0. This
N = number of dislocation links & q‘T . mo = vibrational
requires
 frequency & Debye Frequency & 1013 (for Cu)
DHo þ vðs  soG Þ Substituting the above terms into Eq. (16) leads to,
D0 ¼ exp : 8 h i9
kT
< DHo b2 ðMo cþqo Þ1=2 s =
This can be manipulated into the form found by Conrad _ 2 mo ðMo cþqo Þ1=2 exp
c¼b :
: kT ;
[19]:
 ð25Þ
kT
c¼ lnðCt þ 1Þ; ð21Þ
hv Here the hardening occurs due to variation in the
where average dislocation segment length through the ðMo c þ
 qo Þ1=2 term that appears in both the prefactor and in the
DHo þ vðs  soG Þ
C ¼ m0 exp  : exponential. In both cases, due to the negative exponent,
kT
it works to slow the strain rate as c increases and thus acts
Equation (21) is in the form of Eq. (11) and suggests as a hardening term. Therefore, in this analysis, hardening
logarithmic behavior throughout. However, there are is specifically linked to dislocation multiplication, as at

123
6548 J Mater Sci (2015) 50:6539–6551

(a) 0.009
Logarithmic
0.008
Ho = 135 kJ/mol
= 3 MPa
0.007 Predicted Cu Data Line based on Seeger Mechanism

0.006
R = 0.89172

0.005

0.004

0.003

0.002

0.001

0
1 10 100 1,000 10,000 100,000 1,000,000 10,000,000 100,000,000 1,000,000,000
Time (sec)

(b)1.0000 Power-Law

Ho = 135 kJ/mol
= 3 MPa
Predicted Cu Data Line based on Seeger Mechanism
0.1000

R = 0.9846
0.0100

0.0010

0.0001
1 10 100 1,000 10,000 100,000 1,000,000 10,000,000 100,000,000 1,000,000,000
Time (sec)

Fig. 6 The predicted strain vs. time plots for ambient-temperature critical resolved shear stress, s = 3 MPa. Power-law behavior
creep of Cu based on the dislocation intersection mechanism through (a) appears to better describe the predictions of Eq. (29) than
Eq. (29). The most realistic behavior is for DHo = 135 kJ/mol and a logarithmic behavior (b)

elevated temperatures in an earlier article in this journal  ( 1=2


!
2exp DHkT
o
b2 sqo
[62]. t¼ Ei 
Mo mo b2 kT
The solution for cðtÞ is most easily found numeri- !) ð26Þ
cally. However, while complicated, there is an analytic b2 sðMo c þ qo Þ1=2
solution. First, while one cannot integrate (25) to get Ei  ;
kT
cðtÞ, one can easily obtain the inverse solution for tðcÞ:

123
J Mater Sci (2015) 50:6539–6551 6549

Fig. 7 Activation energy for Activation Energy Ratios for Creep and Self-Diffusion vs. Melting Temperature Ratios
(steady-state) creep of Ag, Ni, of Pure Metals
Cu, and Al as a function of
1.4
temperature. Adapted from [63,
64]
1.2

0.8
Qcreep/QSD
Silver
0.6
Nickel
Aluminum
0.4 Copper

0.2

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
T/Tm

where Ei is the exponential integral which is roughly a factor of 4 lower than expected. If a
Z1 more reasonable resolved shear stress is used, such as
ey 35 MPa, unreasonably high (factor of 10) activation ener-
EiðxÞ ¼ : ð27Þ
y gies of 1350 kJ/mol are produced to achieve the experi-
x
mental observations. Additionally, we note for the most
Substituting EiðxÞ into tðcÞ reasonable case, power-law behavior is predicted. Of
b2 sq
1=2 course, the power-law behavior is merely an approximation
DH   Z kTo
of the fundamental Eq. (29). The Seeger model appears,
2exp kT o
ey
t¼ : ð28Þ overall, to very reasonably predict the Cu room tempera-
M o m o b2 y
b2 sðMo cþqo Þ1=2
ture creep behavior. If other metals, such as aluminum, are
 kT
examined, unrealistic stresses and/or strains are predicted
Technically, since EiðxÞ is multiple valued, it is not for Eq. (25). In Aluminum, cross-slip may be an important
invertible. However, in this problem the argument is thermally activated mechanism for plasticity.
always negative and in that range, EiðxÞ is single valued Other creep mechanisms considered include quantum
and invertible. Therefore, it is proper in this context to mechanical tunneling, which predicts an athermal creep
assume that Ei1 ðxÞ exists, especially as EiðxÞ and its behavior at low temperatures. Early proponents included
inverse for negative arguments, Ei1 ðxÞ, must ultimately be [20, 24, 26, 27]. Conversely, subsequent work by [13, 22,
determined numerically: 23] suggests that, even to 4 K, creep is time-dependent and
" !!#2 is a result of the thermal activation of dislocations. Dislo-
1=2
1 kT 1 Mo mo b2 t b2 sqo q cation kink mechanisms were suggested for BCC metals
c¼ Ei DH   Ei   o:
Mo b2 s 2exp kT o kT M o [23], however, it is unclear how this mechanism, by itself,
explains the observed creep behavior. In materials with
ð29Þ
solutes, it is suggested that creep occurs through thermal
Figure 6 shows creep predictions based on Eq. (29) for activation past pinning solutes [11] in Al–Mg. The precise
combinations of applied stress (s) and activation energy mechanism by which the strengthening variables super-
(DHo ), the two principal variables in Eq. (29). If a rea- impose is unclear. In HCP metals, single slip may occur
sonable value of DHo is used (see Fig. 7), reasonable creep and the Seeger intersection model may not apply.
strains and times are predicted for a resolved shear stress of Thus, polyslip Cu may represent one of the few cases
3 MPa. This translates to an applied tensile stress for examined in this study where the Seeger mechanism is
polycrystalline Cu of about (3.06 9 3 Mpa & 9 Mpa), applicable. For Cu, the fundamental physics may be the

123
6550 J Mater Sci (2015) 50:6539–6551

dislocation intersection mechanism as in Eq. (29). Equa- of reliable creep behavior to longer times. Nonetheless, the
tions (10) and (11) are merely simple forms that approxi- power-law and logarithmic behavioral trends are simple
mately fit the data. Nonetheless, Eqs. (10) and especially curve fits and not reflections of the fundamental physics.
(11), reasonably describe and predict low-temperature
creep behavior in all the metal and alloy systems examined. Acknowledgements The authors are grateful for support from the
NSF under Grant DMR-1401194.
Increasing creep resistance at low temperatures appears to
be accomplished in similar ways as at elevated tempera-
tures. For example, cold work increased the creep resis-
tance in 304-stainless steel at both low and elevated References
temperatures [31]. Others have suggested the role of other
1. Neeraj T, Hou D-H, Daehn GS, Mills MJ (2000) Phenomeno-
features [3] such as twin boundaries. logical and microstructural analysis of room temperature creep in
titanium alloys. Acta Mater 48:1225–1238
2. Thompson AW, Odegard BC (1973) The influence of
Activation energies and mechanisms microstructure on low temperature creep of Ti-5 Al-2.5 Sn.
Metallurg Trans 4:899–908
of low-temperature creep 3. Kameyama T, Matsunaga T, Sato E, Kuribayashi K (2009)
Suppression of ambient-temperature creep in Cp-Ti by cold-
Figure 7 [63, 64] shows that the activation energy rolling. Mater Sci Eng A 510–1:364–367
decreases with temperature below 0.30Tm. The values are 4. Liu C, Liu P, Zhao Z, Northwood DO (2001) Room temperature
creep of a high strength steel. Mater Des 22(4):325–328
much lower than at higher temperatures where the activa- 5. Oehlert A, Atrens A (1994) Room temperature creep of high
tion energy, Q, is associated with lattice self-diffusion strength steels. Acta Mater 42(5):1493–1508
(dislocation climb). The lower temperature activation 6. Krempl E (1979) An experimental study of room-temperature
energies suggest some mechanism other than dislocation rate-sensitivity, creep and relaxation of AISI type 304 stainless
steel. J Mech Phys Solids 27:363–375
climb to be active. The different metals may be associated 7. Kassner ME, Geantil P, Rosen RS (2011) Ambient temperature
with different rate-controlling processes. In Cu, for exam- creep of type 304 stainless steel. J Eng Mater Technol 133:021012
ple, the dislocation intersection mechanism (Seeger) where 8. Castro R, Tricot R, Rousseau D (1969) Steel strengthening
polyslip is expected in grains and negligible solute mechanisms. Climax Molybdenum Co., Greenwich, pp 117–134
(in French)
strengthening is likely occurring. In other polyslip mate- 9. Kassner ME, Kosaka Y, Hall J (1999) Low cycle dwell time
rials (e.g., steels), solute strengthening in addition to dis- fatigue in Ti 6242. Metallurg Trans 30A:2383–2389
location intersection mechanisms may be active. How these 10. Thompson AW, Odegard BC (1974) Low temperature creep of
superimpose in a mathematical description is unclear. If Ti-6 Al-4 V. Metallurg Trans 5:1207–1213
11. Prasad YVRK, Sastry DH, Vasu KI (1970) Mechanism of low-
single slip occurs (e.g., pure HCP at low temperatures), temperature deformation in quenched aluminum-magnesium
glide-controlled mechanisms may be relevant. alloys. Mater Sci Eng 6:327–333
12. Prasad YVRK, Ramchandran T (1971) Low-temperature creep in
polycrystalline a-brass. Scr Metall 5:411–416
13. Tesh JR, Whitworth RW (1970) Plastic deformation and creep if
Conclusions sodium chloride crystals at liquid helium temperatures. Physica
Status Solidi 39:627–633
The experimental literature on low-temperature creep 14. Matsunaga T, Takahashi K, Kameyama T, Sato E (2009)
(T \ 0.30Tm) of a variety of metals and alloys was Relaxation mechanisms at grain boundaries for ambient-temper-
ature creep of H.C.P. metals. Mater Sci Eng A 510–1:356–358
examined and combined with new low-temperature creep 15. Li JCM (1963) A dislocation mechanism of transient creep. Acta
tests performed on pure Cu. In addition, theoretical studies Metall 11:1269–1270
were examined and a new derivation, based on the dislo- 16. Thornton PR, Hirsch PB (1958) The effect of stacking fault
cation intersection model by Seeger was performed that energy on low temperature creep in pure metals. Phil Mag
31:738–761
lead to power-law behavior, rather than the logarithmic 17. Karanjgaokar N, Stump F, Geubelle P, Chasiotis I (2013) A
behavior found by Conrad. Although both power-law and thermally activated model for room temperature creep in
logarithmic models can reasonably model the creep nanocrystalline Au films at intermediate stresses. Scripta Mater
behavior in various conditions, neither is perfect. The 68:551–554
18. Wyatt OH (1953) Transient Creep in Pure Metals. Proc Phys Soc
logarithmic form generally appears to more reliably B 66:459–480
describe the low-temperature creep behavior of most of the 19. Conrad H (1958) An investigation of the rate controlling mech-
metals and alloys analyzed (Cd, ferritic steel, stainless anism for plastic flow of copper crystals at 90 K and 170 K. Acta
steel). Most of the Ti-alloys are better described by power- Metall 6:339–350
20. Mott NF (1956) Creep in metal crystals at very low temperatures.
law equations and Cu and Al may be approximately Phil Mag 1(6):568–572
equally well described by logarithmic and power-law 21. Zeyfang R, Martin R, Conrad H (1971) Low temperature creep of
equations. The equation forms may allow the extrapolation titanium. Mater Sci Eng 8:134–140

123
J Mater Sci (2015) 50:6539–6551 6551

22. Arko AC, Weertman J (1969) Creep deformation of Cd and Hg at 44. Miller WH, Chen RT, Starke EA Jr (1987) Microstructure, creep,
liquid helium temperatures. Acta Metall 17:687–699 and tensile deformation in Ti-6Al-2Nb-1Ta-0.8Mo. Metall Trans
23. Arsenault RJ (1966) An investigation of the mechanism of ther- A 18A:1451–1468
mally activated deformation in tantalum and tantalum-base 45. Cottrell AH (1996) Andrade creep. Philos Mag Lett 73(1):35–37
alloys. Acta Metall 14:831–838 46. Cottrell AH (1996) Strain-hardening in andrade creep. Philos
24. Glen JW (1956) The creep of cadmium crystals at liquid helium Mag Lett 74(5):375–379
temperatures. Phil Mag 1(5):400–408 47. Cottrell AH (1997) Logarithmic and andrade creep. Philos Mag
25. Sastry DH, Prasad YVRK, Vasu KI (1971) Low-temperature Lett 75(5):301–308
deformation behavior of polycrystalline copper. J Mater Sci 48. Nabarro FRN (1997) Thermal activation and andrade creep.
6:1433–1440 Philos Mag Lett 75:227–233
26. Startsev VI, Soldatov VP, Natsik VD, Abraimov VV (1980) Role 49. Oberson PG, Ankem S (2009) The effect of time-dependent
of quantum mechanisms and thermal heating in low-temperature twinning on low-temperature (\0.25 Tm) creep in an alpha-tita-
creep of metals. Phys Status Solidi A 59:377–388 nium alloy. Int J Plast 25:881–900
27. Osetskii AI, Soldatov VP, Startsev VI, Natsik VD (1974) Tem- 50. Jenkins WD, Digges TG (1950) Creep of high-purity copper.
perature dependence and activation parameters of creep in Zn in J Res Natl Bureau Stand 45(2):153–173
the temperature range 1.5 to 80 K. Physica Status Solidi A 51. Sherby OD, Lytton JJ, Dorn JE (1957) Activation energies for
22:739–748 creep of high-purity aluminum. Acta Metall 5:219–227
28. Phillips FP (1905) The Slow stretch in india rubber, glass, and 52. Kassner ME, Pollard J, Evangelista E, Cerri E (1994) Restoration
metal wires when subjected to a constant pull. Phil Mag 9:513 mechanisms in large-strain deformation of high-purity aluminum
29. Olds GCE (1954) Mechanisms of creep in a precipitation hard- at ambient temperature, and the determination of the existence of
ened alloy. Proc Phys Soc B 67:832–842 a steady-state. Acta Metall Mater 42:3223–3230
30. Yen C, Caulfield T, Roth LD, Wells JM, Tien JK (1984) Creep of 53. McQueen HJ, Blum W, Straub S, Kassner ME (1993) Dynamic
copper at cryogenic temperatures. Cryogenics 24(7):371–377 grain growth—a restoration mechanism in 99.999% Al. Scr
31. Kassner ME, Rosen RS, Henshall GA (1990) Delayed-mechani- Metall Mater 28:1299–1304
cal failure of silver interlayer diffusion bonds’. Metallurg Trans 54. Mott NF, Nabarro FRN (1948) Dislocation theory and transient
21A:3085–3100 creep. Report on Strength of Solids. In: Bristol Physical Society
32. Kassner ME, Kennedy TC, Schrems KK (1998) The mechanism Conference, pp 1–19
of ductile fracture in constrained thin silver films. Acta Mater 55. Seeger A (1958) Encyclopedia for physics, vol 7/2. Springer,
46:6445–6458 Berlin
33. Seeger A (1955) The generation of lattice defects by moving 56. Hull D, Bacon DJ (2011) Introduction to dislocations, 5th edn.
dislocations and its application to the temperature dependence of Elsevier, Amsterdam
the flow-stress of F.C.C. crystals. Phil Mag 46:1194–1217 57. Levine LE, Geantil P, Larson BC, Tischlers JZ, Kassner ME, Liu
34. Seeger A, Diehl J, Mader S, Rebstock H (1957) Work-hardening W, Stoudt MR, Tavazza F (2011) Disordered long-range internal
and work-softening of face-centered cubic metal crystals. Phil stresses in deformed copper and the mechanisms underlying
Mag 2:323–350 plastic deformation. Acta Mater 59(14):5803–5811
35. Kassner ME (2014) Fundamentals of creep in metals and alloys, 58. Kassner ME, Pérez-Prado M-T, Vecchio KS, Wall MA (2000)
3rd edn. Elsevier, Amsterdam Determination of internal stresses in cyclically deformed Cu
36. Andrade ENdaC (1910) On the viscous flow in metals and allied single crystals using CBED and dislocation dipole separation
phenomena. Proc R Soc A 84:1–12 measurements. Acta Mater 48:4247–4254
37. Orowan E (1947) The creep of metals. Glasgow, West Scotland 59. Kassner ME, Smith KK (2014) Low-temperature creep plasticity.
Iron and Steel Instiute, p 45 J Mater Res Technol 3(3):280–288
38. Evans HE, Wilshire B, Institute of Metals (1985) Creep of metals 60. Welch DO, Smoluchowski R (1972) Exhaustion theory of creep
and alloys. Institute of Metals, London metals at low temperatures. J Phys Chem Solids 33:1115–1127
39. Cottrell AH, Aytekin V (1947) Andrade’s creep law and the flow 61. Challenger KD (1973). Irradiation-induced void swelling in cold-
of zinc crystals. Nature 160:328–329 worked type 316 stainless steel resulting from 5 MeV nickel ion
40. Conway JB (1967) Numerical methods for creep and rupture bombardment. General Electric Report NEDO-13990 73NED101,
analysis. Advanced materials research. Gordon and Breach, New Class I
York, p 22 62. Kassner ME (1990) A case for Taylor hardening during primary
41. Garofalo F, Richmond C, Domis WF, Von Gemmingen F (1965) and steady-state creep in aluminum and stainless steel. J Mater
In: Joint International Conference on Creep, 1963. Inst. Mech. Sci 25:1997–2003
Engineers, London. pp 1–31 63. Landon PR, Lytton JL, Shepard LA, Dorn JE (1959) The acti-
42. Laurent P, Eudier M (1950) Crystalline interaction and the brit- vation energies for creep of polycrystalline copper and nickel.
tleness of metals. Revue de Métallurgie 47:582–587 Trans ASM 51:900–910
43. Chévenard P (1934) Experimental study of the viscous defor- 64. Kassner ME (1989) The rate dependence and microstructure of
mation of iron and nickel wires—I, II. Revue de Métallurgie high-purity silver deformed to large strains between 0.16 and 0.30
31:473–535 Tm. Metallurg Trans 20A:2001–2010

123

You might also like