Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Available online at www.sciencedirect.

com

ScienceDirect
Scripta Materialia 76 (2014) 37–40
www.elsevier.com/locate/scriptamat

Revisiting the Considère criterion from the viewpoint of dislocation


theory fundamentals
I.S. Yasnikov,a A. Vinogradova,⇑ and Y. Estrinb,c
a
Laboratory for the Physics of Strength of Materials and Intelligent Diagnostic Systems, Togliatti State University,
Togliatti 445667, Russia
b
Centre for Advanced Hybrid Materials, Department of Materials Engineering, Monash University, Clayton 3800, Australia
c
Laboratory of Hybrid Nanostructured Materials, NITU MISiS, Moscow 119490, Russia
Received 9 November 2013; revised 13 December 2013; accepted 14 December 2013
Available online 18 December 2013

We demonstrate that the Considère condition for plastic instability, which is traditionally obtained from solid mechanics con-
siderations, also follows from the intrinsic evolution laws for dislocation density. Taking strain-rate effects into account, a modified
instability condition emerges in a form that resembles Hart’s criterion, but is not identical to it. The Considère strain predicted from
the dislocation-based model shows good agreement with experimental data, highlighting the primary role played by dynamic recovery
in the mechanical response.
Ó 2013 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Plastic deformation; Modelling; Dislocation dynamics; Instability

Plastic deformation in virtually all metals, dislocation density. A recent revival of interest in devel-
ranging from ductile single crystals to relatively brittle oping the foundations for microstructure-based model-
metallic glasses, results finally in strain localization and ling of plastic deformation of conventional and
fracture. Although plastic deformation is obviously emerging materials has prompted us to revisit this classic
inhomogeneous on the dislocation scale, it appears to subject. We will approach the onset of tensile instability
be reasonably homogeneous at a macroscopic scale until from the viewpoint of dislocation-density-based model-
necking sets in at a certain stage of deformation, at which ling. On that basis we shall demonstrate that the Consid-
point the plastic flow becomes localized, the cross-sec- ère criterion is not just a mechanistic condition for loss
tional area of the specimen shrinks and the load–dis- of mechanical stability of a plastically deforming body
placement curve begins to drop off. The historically under uniaxial tension. Rather, it follows as a natural
first and undoubtedly most popular criterion of plastic consequence of the evolution of dislocation ensembles
instability, known as the Considère criterion [1], can be during plastic deformation. In other words, the Consid-
deduced from a variety of considerations of deformation ère condition will be proven to mark the onset of an
mechanics near the maximum point on the stress–strain intrinsic instability of an evolving dislocation ensemble.
diagram [2]. The Considère criterion signifies the onset Several other instability criteria reviewed in Refs. [3–7]
of necking at the point where the strain-hardening coef- have been proposed to describe the critical conditions un-
ficient h drops below the value of the flow stress r at a der which plastic deformation becomes unstable with re-
given plastic strain rate e_ . It is expressed by the condition: spect to local perturbations of uniform plastic flow. A
 recent review of plastic strain localization phenomena
@r
h   ¼ r: ð1Þ was provided by Antolovich and Armstrong [8]. A general
@e e_ basis for a classification of the types of unstable behavior of
Thus, this instability criterion is based entirely on the solids in terms of constitutive characteristics is well estab-
relation between stress and strain—despite the fact that lished [9–11]. The term “plastic instability” refers to a situ-
strain cannot be regarded as a state variable. Indeed, Eq. ation in which a small local deviation from spatially
(1) does not involve intrinsic state variables, such as the uniform deformation grows with time. The particular ap-
proaches may differ in the detail of the constitutive formu-
⇑ Corresponding author. Tel.: +7 8482545603; fax: +7 8482539522; lation or the choice of the quantity whose deviations from
e-mail addresses: alexei.vino@gmail.com; vinogradov@tltsu.ru uniformity are monitored, but the type of the stability

1359-6462/$ - see front matter Ó 2013 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.scriptamat.2013.12.009
38 I. S. Yasnikov et al. / Scripta Materialia 76 (2014) 37–40

analysis used is common to them. In most cases, the start- namics applied to dislocation plasticity [17,18]. The
ing point is the differential form of a constitutive equation phenomenological constants k1 and k2 can be reinter-
that relates the flow stress to the plastic strain e and the preted on that basis [18].
plastic strain rate e_ as dr ¼ hde þ Sd ln e_ . Here Let dq and d_c be small fluctuations of the dislocation
S ¼ @r=@ ln j_ejje denotes the strain-rate sensitivity of the density and shear strain rate, respectively. They are to be
flow stress. Introducing small local deviations from a understood as local deviations from the otherwise uni-
spatially uniform solution of this equation and performing form quantities q and c_ . According to Eq. (3), the fluc-
linear stability analysis, one arrives at the Considère tuating variables obey the following equation:
condition [9–11]. pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dðq þ dqÞ
A time-proven approach, which dominates the philos- ¼ ð_c þ d_cÞðk 1 q þ dq  k 2 ðq þ dqÞÞ; ð5Þ
ophy of constitutive modelling to the present day, was dt
proposed by Kocks and Mecking (KM) [12–15]. The which, after linearization with respect to dq, takes the
structure of their model is based on the relation between form:
the shear stress s and the total dislocation density q:  
1 dq dq
 1 dq_ ¼  k 2 q c_ dq þ d_c: ð6Þ
pffiffiffi c_ m 2q dc dc
s ¼ ða0 Gb qÞ ; ð2Þ
c_ 0 Using Eq. (2) and the relation dr ¼ rde; which fol-
where c_ is the plastic shear strain rate, G the shear modulus, b lows from plastic incompressibility for the case of uniax-
the magnitude of the dislocation Burgers vector, and ao a ial tension [9], one obtains:
constant. The first factor on the right-hand side represents M M
dc ¼ dq þ d_c; ð7Þ
the Taylor relation for the glide resistance due to disloca- 2q m_c
tion–dislocation interactions in the absence of thermal acti- where M is the Taylor factor converting the shear-re-
vation, at zero absolute temperature [12–15]. The thermally lated quantities to the axial ones: dc ¼ Mde and
activated character of dislocation glide underlying Eq. (2) is ds ¼ dr=M. A solution of the set of coupled equations
cast in a power-law form often used in the engineering liter- (6) and (7) is sought in the standard exponential form:
ature. This form of relation between c_ and s is equivalent to  
an Arrhenius equation, if the exponent m is uniquely related dq ¼ x ¼ x0 expðktÞ dq_ ¼ x_ ¼ x0 k expðktÞ
ð8Þ
to the activation volume V a for the underlying thermally dc ¼ y ¼ y 0 expðktÞ d_c ¼ y_ ¼ y 0 k expðktÞ
activated process: m ¼ V a s=k B T [16], where kB denotes
the Boltzmann constant and T is the absolute temperature. with x0 and y 0 being constants defining the initial devia-
The dislocation density dependence given by Eq. (2) holds tions of the respective quantities. After plugging this
quite generally regardless of the geometrical arrangement solution into Eq. (7), one obtains a characteristic equa-
or the detail of the dislocation–dislocation interaction mech- tion for the parameter k:
     
anism. The temperature dependence associated with ther- 2q 2
k þ
dq 2q
 
1 dq
 k2 q k þ
c_ dq
 k 2 q ¼ 0: ð9Þ
mally activated dislocation glide resides in the parameters m_c dc M m dc M dc
c_ 0 and/or m (usually m >> 1 holds). Thus, at low tempera- Stability of the uniform deformation with respect to
tures (below half the melting point, say), c_ 0 may be consid- local perturbations given by Eq. (8) assumes that these
ered to be constant, while m is inversely proportional to T. perturbations vanish with time. This is the case if all roots
Conversely, at sufficiently high temperatures, m can be re- of the quadratic expression in k, Eq. (9), have negative real
garded as a constant, while c_ 0 is temperature dependent parts, so that the perturbations decay with time. Moreover,
[12–16]. As the dislocation density evolves during the pro- both roots must have negative real parts. In general, a poly-
cess of straining, a general structure of the equation describ- Pm
nomial P m ðkÞ ¼ k¼0 ak kk is stable if and only if the leading
ing this evolution in time is given by:
principal minors Di of the Hurwitz determinant D are all
dq pffiffiffi positive, so that the necessary stability conditions are ex-
¼ ðk 1 q  k 2 qÞ_c; ð3Þ
dt pressed as a1 > 0; a1 a2 > 0 (plastic instabilities of the
where k1 is a constant accounting for dislocation storage Portevin–Le Chatelier type that may be associated with
due to interaction of gliding dislocations with forest dis- negative strain-rate sensitivity (m < 0) have been reviewed
locations. The underlying assumption is that the mean extensively [8,10] and will not be considered here). Hence,
free path of gliding dislocations is controlled by disloca-
pffiffiffi the coefficient in the quadratic term is positive, a2 > 0,
tion–dislocation interactions and scales with 1= p Eq. and the stability condition is reduced to:
(3) can be rewritten as:    
dq 2q 1 dq
   k 2 q > 0: ð10Þ
dq pffiffiffi dc M m dc
¼ k 1 q  k 2 q: ð4Þ
dc Using Eq. (2), one obtains:
The coefficient k2 is associated with the dynamic dq 2r dr
¼  : ð11Þ
recovery leading to a loss of dislocation density due to dc 1 2
de
annihilation by cross-slip of screw dislocations (at low a0 Gb ð_c=_c0 Þm M 3
temperatures) or climb of edge dislocations (at high tem- Substitution of Eq. (11) into inequality (10) yields the
peratures). Accordingly, k2 is strain-rate and tempera- stability condition in the form:
ture dependent. The original equation for the    
dislocation density evolution, Eq. (4), can be recovered dr 1 dr k 2 Mr
r   > 0: ð12Þ
in terms of a general formalism of irreversible thermody- de m de 2
I. S. Yasnikov et al. / Scripta Materialia 76 (2014) 37–40 39

When the strain-rate sensitivity is negligible, i.e. when Moreover, this dislocation density evolution approach is
1=m ! 0 and the flow stress is determined by the rate- inherently capable of predicting the critical strain corre-
independent variant of Eq. (2), inequality (12) reduces sponding to the Considère instability point in a simple and
to the familiar Considère stability condition rh > 1 (cf. explicit manner. The critical strain for the onset of necking,
Eq. (1)). We note that this condition was obtained from en ; which is given by the condition rðen Þ ¼ @r=@ejen , can be
the dislocation density evolution in the KM approach, calculated from  Eqs. (15)  and (16), yielding:
without any further assumptions or restrictions. For 2 k2M
strain-rate-dependent plasticity (1=m–0), Hart’s analy- en ¼ ln 1 þ : ð19Þ
k2M 2
sis [5] leads to the following instability condition: This result is remarkable in that it predicts that en is gov-
h 1 erned primarily, if not entirely, by the rate of dynamic recov-
þ 6 1: ð13Þ ery k2. Indeed, the larger k2, the greater is the resistance of the
r m
This condition shows that a positive strain-rate sensi- material to strain localization and the larger is en . Hence, in
tivity of the flow stress stabilizes plastic flow against line with several investigations [18–23], the present result high-
strain localization. In our analysis, cf. Eq. (12), the lights unequivocally that during plastic deformation of
instability condition takes a somewhat different form: metals, a key factor influencing dislocation evolution in qua-
  si-steady-state plastic flow and up to the point of instability is
h 1 k2M h
þ  6 1: ð14Þ the dislocation annihilation rate controlled by k2. The value of
r m 2 r the phenomenological coefficient k2 can readily be recovered
Importantly, for all practical purposes the second from analysis of the stress–strain curve using either a method
term in inequality (14) can be considered to be positive, proposed in Ref. [14] or from a non-linear fit of the stress–
see below. In other words, in our approach, like in that strain data by Eq. (15). Despite its simplicity, this equation
of Hart, the positive strain-rate sensitivity of the flow matches the experimental data obtained for conventional
stress plays a stabilizing role. We consider the sign of coarse-grained materials [15,19] quite well, at least for suffi-
the second term in inequality (14) by making use of ciently small strains prior to necking (Fig. 1). The necking
the explicit solution of Eq. (4): strain en corresponding to an intercept of the r vs. e and the
   h vs. e curves is marked in the figure and is found in this tradi-
aGMk 1 k 2 Me tional way (0.235 ± 0.004). This value is very close to that
r¼ 1  exp  ; ð15Þ
k2 2 found from the KM model-based Eq. (19) (0.242) with k1
combined with the corresponding form of the strain- and k2 recovered from the non-linear curve fit of the stress–
hardening coefficient: strain curve in the early deformation stage marked in Figure 1
    by a rectangle. Dynamic recovery of dislocations is considered
@r aGM 2 k 1 k 2 Me aGM 2 k 1 k2 r
h¼ ¼ exp  ¼ 1 : ð16Þ to be a thermally activated process, so that the coefficient k2 is
@e 2 2 2 aGMk 1
strain-rate and temperature dependent:
The instability condition (14) can then be rewritten as:  1n
   c_
h 1 1 d ln q k 2 ¼ k 2 ð_c; T Þ ¼ k 2 ðT Þ : ð20Þ
þ k2M  6 1: ð17Þ c_ 0
r m 2 dc
Here the exponent n associated with the dynamic recov-
Near the point of instability, where the strain-harden- ery process [16] is, like m, much greater than unity [24]. For
ing rate is low, the dislocation density does not vary sig- annealed copper and iron polycrystals with approximately
nificantly, so that the term which involves a derivative of the same grain size of 200 lm tested at nominal strain rates
the logarithm of q can be neglected. As a result, the ranging from 5  104 to 5  102 s1, we found that k2
instability condition assumes the following form: does indeed obey the k 2  c_ n relation, as anticipated from
h k2M Eq. (20). The value of n was found to be about 100, which is
þ 6 1: ð18Þ close to the exponent m entering the strain-rate dependence
r m
of the flow stress through Eq. (2) [15]. Qualitatively and
Obviously, the second term in inequality (14), or its quantitatively similar results were also obtained when we
equivalent, inequality (18), is then positive. It should be used experimental data presented in Ref. [15] (cf. Fig. 3
noted that a limiting case of k 2 ! 0 that would lead to therein) for Cu tested over a wider range of strain rates
a negative sign of the second term in inequality (14), thus from 104 to 1 s1 (Fig. 2).
destabilizing uniform deformation, is not relevant, since The distinctions between the deformation behavior of
in that case Eqs. (15) and (16) yield h=r ¼ 1=e , so that face-centered cubic and body-centered cubic (bcc)
h=r > 1 applies for reasonable values of strain. Accord-
ingly, there is no chance for instability condition (14) to
be fulfilled, regardless of the sign of the second term in
that inequality. We note that for a reasonable range of
values of k2 inequality (14) is well represented by inequal-
ity (18). While this instability condition is rather close to
that of Hart both in its analytical form and quantita-
tively, it is not identical to the latter.
Figure 1. Stress–strain diagram for conventional polycrystalline cop-
Hence, we have shown that the instability criterion com- per with a grain size of 70 lm and a model prediction based on the KM
monly used in the form of the Considère condition appears approach. The dashed line shows a fragment of the strain dependence
as a natural consequence of the evolution of dislocation of the strain-hardening coefficient h used to determine the onset of
ensembles described by the KM equations [12–15]. necking corresponding to the Considère criterion.
40 I. S. Yasnikov et al. / Scripta Materialia 76 (2014) 37–40

the basis of the KM dislocation kinetics approach. This


result highlights the defining role that dynamic recovery
of dislocation density plays in the onset of instability. If
strain-rate sensitivity effects are included, the instability
condition is modified to inequality (18), which resembles
Hart’s criterion. However, the two instability conditions
for strain-rate-sensitive materials are not identical. There
Figure 2. Dependence of the dynamic recovery coefficient k2 on the is no universal convention on the exact procedure for lin-
strain rate, showing agreement with Eq. (20). (Experimental data ear stability analysis, and so either condition has the right
adopted from Ref. [15].) to be considered as the correct one, unless proven other-
wise. In fact, Havner [29] stated that criteria for necking
materials stem primarily from the differences in the crystal- for rate-dependent materials based on an analysis that as-
lography of dislocation slip and cross-slip, as well as from sumes initially homogeneous deformation cannot be con-
the significance of the double-kink mechanism of disloca- sidered as proven. Both Hart’s condition and our
tion motion in Peierls relief in bcc metals. As a result, dis- instability condition were obtained under the latter pre-
locations in bcc metals show a greater propensity to mise and are thus subject to validation. Further studies
annihilate. This implies that the coefficient k2 in bcc metals in this direction are planned.
should be higher, which is, indeed, observed experimen-
tally, as shown in Figure 3 where data for polycrystalline Financial support from the Russian Founda-
Cu and Fe are compared and an excellent agreement with tion for Basic Research under project No. 13-08-00259
the predictions by Eq. (19) is found: the smaller k2, the and from the Russian Ministry of Education and
greater is the uniform elongation stage governed by en . Science through the Grants-in-aid 11.G34.31.0031 and
The trends expected for the strain-rate dependence of k2 14.A12.31.0001 is gratefully appreciated.
(Eq. (20) and Fig. 2) are also confirmed in Figure 3. It
should be noted that very high values of k2, of the order
of several hundreds, were reported for ultrafine-grained [1] A. Considère, Annales des ponts et chaussées I sem.(1885) 574.
(UFG) materials produced by severe plastic deformation, [2] M.F. Ashby, Engineering Materials. 2 An Introduction to Micro-
structures, Processing and Design, Elsevier, Oxford, 2006, p. 451.
for which the KM approach to the modelling of strain
[3] I.H. Lin, J.P. Hirth, E.W. Hart, Acta Metall. 29 (1981) 819.
hardening applies as well [24–27]. Evidently, Eq. (4) does [4] J.J. Jonas, R.A. Holt, C.E. Coleman, Acta Metall. 24
not apply to the case when the dislocation mean free path (1976) 911.
is controlled by length scales not related to dislocation–dis- [5] E.W. Hart, Acta Metall. 15 (1967) 351.
location interactions. This is the case with UFG materials, [6] M. Zaiser, P. Hähner, Mater. Sci. Eng., A 238 (1997) 399.
for which the dislocation mean free path is determined by [7] A.K. Ghosh, Acta Metall. 25 (1977) 1413.
the grain size. In order to account for the latter effect, the [8] S.D. Antolovich, R.W. Armstrong, Prog. Mater Sci. 59 (2014) 1.
governing evolution equation (4) was replaced with [9] Y. Estrin, L.P. Kubin, Res Mech. 23 (1988) 197.
dq=dc ¼ 1=bL  k 2 q, where L denotes the grain or cell size [10] Y. Estrin, L.P. Kubin, Mater. Sci. Eng., A 137 (1991) 125.
[24–27]. It can be shown that in this case too the Considère [11] Y. Estrin, Solid State Pnenom. 384 (1988) 417–428.
[12] U.F. Kocks, J. Eng. Mater. Tech. Trans. ASME 98 (1976) 76.
instability point is determined solely by the rate of dynamic
[13] H. Mecking, U.F. Kocks, Acta Metall. 29 (1981) 1865.
recovery and the above conclusions regarding the signifi- [14] Y. Estrin, H. Mecking, Acta Metall. 32 (1984) 57.
cance of k2 to the instability condition apply. Thus, in line [15] U.F. Kocks, H. Mecking, Prog. Mater Sci. 48 (2003) 171.
with the expected trend, the observed high k2 values in [16] Y. Estrin, in: A.S. Krausz, K. Krausz (Eds.), Unified
UFG metals are the cause for a very short uniform elonga- Constitutive Laws of Plastic Deformation, Academic
tion stage (e.g. 0.0025–0.005 for copper, depending on the Press, San Diego, 1996, pp. 69–106.
equivalent strain imposed [24]), which is a very common [17] M. Huang, P.E.J. Rivera-Dı́az-del-Castillo, O. Bouaziz,
feature of these materials [28]. S. van der Zwaag, Mater. Sci. Technol. 24 (2008) 495.
In conclusion, it has been shown that the Considère [18] A. Vinogradov, I.S. Yasnikov, Y. Estrin, Phys. Rev. Lett.
condition for necking under uniaxial tensile deformation, 108 (2012) 205504.
[19] U. Essmann, H. Mughrabi, Philos. Mag. A 40 (1979) 731.
which is traditionally obtained from solid mechanics con-
[20] W. Blum, Physica Status Solidi (b) 45 (1971) 561.
siderations, also follows from intrinsic evolution laws for [21] P.E.J. Rivera-Dı́az-del-Castillo, M. Huang, Acta Mater.
the governing internal variable—the total dislocation 60 (2012) 2606.
density. The critical strain corresponding to the Consid- [22] E.I. Galindo-Nava, P.E.J. Rivera-Dı́az-del-Castillo,
ère point of instability is predicted in explicit form on Mater. Sci. Eng., A 558 (2012) 641.
[23] E.I. Galindo-Nava, J. Sietsma, P.E.J. Rivera-Dı́az-del-
Castillo, Acta Mater. 60 (2012) 2615.
[24] Y. Estrin, A. Molotnikov, C.H.J. Davies, R. Lapovok, J.
Mech. Phys. Solids 56 (2008) 1186.
[25] A.Y. Vinogradov, V.V. Stolyarov, S. Hashimoto, R.Z.
Valiev, Mater. Sci. Eng., A 318 (2001) 163.
[26] V. Patlan, K. Higashi, K. Kitagawa, A. Vinogradov, M.
Kawazoe, Mater. Sci. Eng., A 319 (2001) 587.
[27] V. Patlan, A. Vinogradov, K. Higashi, K. Kitagawa,
Figure 3. Experimental relation between en and k2 for pure polycrys- Mater. Sci. Eng., A 300 (2001) 171.
talline copper and iron tested at different strain rates as indicated on the [28] Y. Estrin, A. Vinogradov, Acta Mater. 61 (2013) 782.
graph. An excellent agreement with predictions by Eq. (19) is noted. [29] K.S. Havner, Int. J. Plast. 20 (2004) 965.

You might also like