You are on page 1of 34

Finite Elastic-Plastic Deformation of Polycrystalline Metals T. Iwakuma; S. Nemat-Nasser Proceedings of the Royal Society of London.

Series A, Mathematical and Physical Sciences, Vol. 394, No. 1806. (Jul. 9, 1984), pp. 87-119.
Stable URL: http://links.jstor.org/sici?sici=0080-4630%2819840709%29394%3A1806%3C87%3AFEDOPM%3E2.0.CO%3B2-R Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences is currently published by The Royal Society.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/rsl.html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers, and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take advantage of advances in technology. For more information regarding JSTOR, please contact support@jstor.org.

http://www.jstor.org Wed Aug 8 12:54:02 2007

Proc. R. Soc. Lond. A 394, 87-119 (1984) Printed in Great Britain

Finite elastic-plastic deformation of polycrystalline metals


B Y T. I W A K U MA A N D S. N E M A T - N A S S E R

The Technological Institute, Department of Civil Engineering,


Northwestern University, Evanston, Illinois 60201, U.S.A.
(Communicated by Rodney Hill, F.R.S.
- Received

29 July 1983)

Applying Hill's self-consistent method to finite elastic-plastic deformations, we estimate the overall moduli of polycrystalline solids. The model predicts a Bauschinger effect, hardening, and formation of vertex or corner on the yield surface for both microscopically non-hardening and hardening crystals. The changes in the instantaneous moduli with deformation are examined, and their asymptotic behaviour, especially in relation to possible localization of deformations, is discussed. An interesting conclusion is that small second-order quantities, such as shape changes of grains and residual stresses (measured relative to the crystal elastic moduli), have a first-order effect on the overall response, as they lead to a loss of the overall stability by localized deformation. The predicted incipience of localization for a uniaxial deformation in two dimensions depends on the initial yield strain, but the orientation of localization is slightly less than 45" with respect to the tensile direction, although the numerical instability makes it very difficult to estimate this direction accurately.

Our objective is to estimate the overall instantaneous moduli of polycrystalline solids or composites from the mechanical behaviour of their constituents. Many models have been developed for this purpose (Voigt 1889; Reuss 1929 ; Boas & Schmid 1934; Bruggeman 1934; Huber & Schmid 1934; Boas 1935; Taylor 1934, 1938). Taylor (1938), in a pioneering effort, proposed a rigid-plastic model for polycrystals, which was later generalized by Bishop & Hill (1951). For elastic composites, Voigt (1889) and Reuss (1929) each developed a model, and their results were examined and expanded by Hill (1952). An essential requirement is an averaging process whereby, for polycrystalline solids, for example, the rate stress-strain relations are deduced from the 'elastic ' lattice distortion, and from the 'plastic' deformation by slip on crystallographic slip planes, of the single crystals in the aggregate. Averaging methods that have been used seek to account for the interaction of a grain and its surroundings, which must result in a compatible overall deformation. The early work of Batdorf & Budiansky (1949) addressed deformation regimes with elastic and plastic strains of comparable magnitudes. A significant step has been the development of 'selfconsistent ' approaches originated by Hershey (1954), Kroner (1958, 1961), and [ 87 1

88

T. Iwakuma and S. Nemat-Nasser

Budiansky & Wu (1962) and, with considerably greater bearing on later developments, by Budiansky (1965) and Hill (1965a, b). A somewhat different technique has been used by Lin (1957, 1971)who generalized Taylor's work to include elastic effects. Hutchinson (1970)has given a detailed illustrative account of three major averaging techniques attributed to Lin, Budiansky-Kroner-Wu, and Hill, and has revealed that Hill's self-consistent model provides less stiff overall responses. Another requirement for the estimate of the overall properties of polycrystalline solids is the description of the behaviour of the single crystal constituent. This has been pioneered by Taylor & Elam (1923, I 925, 1926) and examined extensively by others, both theoretically and experimentally, as reviewed by Kocks (1970) (see also Kocks 1958). Various hardening rules have been discussed by Hill (1966),who proposed a general type of cross-coupling between slip and hardening on different slip systems, and studied the relation between the hardening rule and the manner in which the stress rates bear on the corresponding slip rates. A key role in most of these contributions is played by Eshelby's solution of the state of stress, strain, and energy of a single ellipsoidal inclusion that undergoes a uniform transformation strain in an infinitely extended linearly elastic homogeneous solid (Eshelby 1957, 1959). With the aid of this solution, or an equivalent method due to Hill (1965a), or other corresponding solutions for anisotropic materials (see, for example, Bhargava & Radhakrishna 1964; Willis 1964; Chen 1967), the local stresses and strains are related to the overall applied stresses and strains. I n the self-consistent approach of Budiansky-Hill-Kroner, the grain is assumed to be within a matrix that possesses the overall average macroscopic moduli, and in this manner the interaction and compatibility of deformation are taken into account to a certain extent. For elastic composites the estimate obtained by this method seems to become less accurate as the volume fraction of the inclusions (or voids) increases. I n a recent paper, Kemat-Nasser, Iwakuma & Hejazi have shown that an excellent prediction of the experimental results is obtained if a periodic structure is assumed, in which case the transformation strains are no longer uniform within each inclusion (Nemat-Nasser et al. 1982).The mathematical problem for the periodic structure can be solved completely and as accurately as desired. The non-uniform transformation strains, although they complicate the calculation, do not present an insurmountable mathematical task (Asaro & Barnett 1975); see Mura (1982) for a detailed discussion and references. All these developments pertain to infinitesimal deformation theories. Little or almost no progress has been made in the calculation of the overall elasto-plastic behaviour of composites and polycrystalline solids when strains are finite, and the domiilallt components of the overall elasto-plastic instantailcous moduli are of' the order of magnitude of the applied stresses; the fundamental work of Hill (1972), however, does provide a general framework and basic guidance. This is notwithstanding the considerable progress that has been made in quantifying the behaviour of single crystals at finite strains by Alandel (1965),Hill (1966),Hill &

Finite elastic-plastic deformation of polycrystals

89

Rice (1972), Zarka (1973), Asaro & Rice (1977), Havner & Shalaby (1977), Asaro (1979), Nemat-Nasser, Mehrabadi & Iwakuma (1980), and Havner (1982a). At finite elasto-plastic deformations, the shapes of individual grains are altered during the course of flow and lattice orientation changed because of finite rotations; this leads to 'texture'. The plasticity of textured solids recently has been addressed from a phenomenological point of view by Bassani (1977)and Hill (1979). But no micromechanically based calculations with due account of the aspects of finite deformation seem to exist. The main purpose of the present work is to calculate the instantaneous moduli of polycrystalline solids (and composites?) which undergo (quasi-statically and isothermally) finite deformations, and cause residual stresses of differing magnitudes to develop within grains or constituents of different orientations. The grains may attain different geometries and orientations in the course of plastic flow. These then may lead to overall instability by localized deformation. The analysis is based on Hill's averaging theorems, a general formulation of the finite elasto-plastic deformation of single crystals by lattice distortion and crystallographic slip, and the calculation of Green's function for incremental deformations superimposed on an initially uniformly (but not isotropically) and arbitrarily stressed solid. The rate problem is cast in terms of the nominal stress rate. The corresponding instantaneous moduli that relate the overall uniform nominal stress rate to the corresponding uniform overall velocity gradient are estimated. Detailed numerical results for uniaxial loading of polycrystals are presented for a two-dimensional illustrative example, where two slip systems within each grain are assumed. This is the simplest case which involves all aspects of the considered problem and, in particular, brings into focus factors that are inherent in a complete finite deformation formulation of the problem. Since the overall loading is uniaxial and the overall deformation is planar, eight out of the total sixteen global moduli remain non-zero. An interesting feature of the solution is that, even though the changes in the shapes and the orientations of grains are very small, and the residual stresses in grains are of second order relative to the crystal elastic moduli, their effect on the overall response is of the first order, as it is precisely these factors that cause the loss of overall stability by localized deformations. Indeed, if the changes in shape and orientation, and the residual stresses in the crystals are neglected, no loss of stability is predicted. Notation Throughout, componentswithrespect to a fixed rectangular Cartesian coordinate system with coordinate axes xi and unit base vectors e,, i = 1 , 2 , 3 , are used. Local fields (within grains) are denoted by lower case letters, and the overall (uniform) quantities are designated by the corresponding capital letters; for example velocity, nominal stress rate, deformation rate, and spin are represented by vi,nij,d i j , and zuii respectively over typical grains, and by 4, ATij,D i j , and yj for the overall solid. 2 when the Various moduli are denoted by capital letters, with a superimposed i

To limit the size of this paper, results for co:nposites ~villbe reported elsewhere.

90

T. Iwakuma and S. Nemat-Nasser

quantity refers to the typical grain R ; for example the instantaneous moduli in v,, and = jk,V,,,,where repeated indices are summed. the relations ri, = Averages taken over all crystal orientations and shapes are represented by (. . ; e.g. = (nij).

FS,, , iyj

a )

Features essential for the description of single crystals which flow by slip on crystallographic planes, accompanied by the accommodating elastic lattice distortion (both a t finite strains) have been disussed by Mandel (1965),Hill (1966), Hill & Rice (1972), Zarka (1973),Asaro & Rice (1977), Havner & Shalaby (1977), Asaro (1979), Havner (1979), and Nemat-Nasser, Mehrabadi & Iwakuma (1980) whose studies follow early developments by Taylor & Elam (1923, 1925, 1926) and Kocks (1958) (see also Kocks 1970). Single crystal relations for polycrystal calculations are briefly reviewed in this section. The presentation follows recent work by Nemat-Nasser, Mehrabadi & Iwakuma (1980). I n the microscopic model of elasto-plastic flow of a single crystal, the deformation rate, dij, and the spin, wij, associated with the velocity gradient,

are each thought to consist of a plastic (denoted by a superscript p ) and an elastic (denoted by a superscript e) contribution, dij=d7j+diPj and W..=U.'?.+U.'?. 1 . 3 23 2 32 (2.2a, b) where the additive decomposition of rates follows rigorously from the choice of variables and the structure of the theory (Hill 1950, 1958; Hill & Rice 1972; Ifandel 1974, 1981; Nemat-Nasser 1979, 1982, 1983). I n (2.1) the comma followed by the index signifies partial differentiation with respect to the corresponding coordinate. Consider a deformed crystal and let sz be the unit vector in the glide direction on the a t h slip plane of the unit normal n?. Let j" be the slip rate of the a t h system. Then dB = p Q y , wFj = wri y, (2.3) where a is summed over all active slip systems, and the notation wzj = &(s;ng - sj"n;), with no sum over a , (2.4a, b) p a = &(s;nj"+ ST,:), is used. Traditionally, plastic volume changes have been ignored, as they are often negligibly small. Recent experimental observations on super-alloys and other highstrength metals reveal that small voids are often generated a t the intersection of slip bands and non-metallic inclusioils on grain boundaries, which result in a finite, though small, plastic volumetric expansion (see Dyson, Loveday & Rodgers 1976; Kikuchi & Weertman 1980; Saegusa, Uemura & Weertman 1980).To account for this and, a t the same time, to i)rovide potential application to non-metallics, i.e. jointed rocks and granular materials, terms n%y tan v, + sf sj"tan v, + S i j tan v,

Finite elmtic-plastic deformation of polycrystals

91

(no sum over a ) may be added to the right side of (2.4a),where Sij is the Kronecker delta (Nemar-Nasser I 983). Let uij denote the Cauchy stress. It is natural to employ stress rates co-rotational with elastic lattice distortion for the local variation of tractions, as observed by Hill (1966) and used by Asaro (1979) to describe the local elastic response,

The local stress variation is then expressed by

where LGklrepresents the elastic moduli of the crystal with the usual symmetries and with, a t most, 21 independent components (see, however, Hill 1975) which may be reduced further as a result of crystal symmetries. The dynamics of slip must be in accord with assumed plastic rules (Bishop & Hill 1951). A simple model emerges from the Schmid law, and results in normality (Hill 1966). According to this model, a slip system is inactive in unloading or when the shear stress falls below the current yield stress. Then, for a typical active slip,

where ra = uijs$ nja (no sum over a) is the resolved shear stress in the slip direction, h a p is the hardening parameter on the ath system produced by plastic flow of the Pth system, and /3 in (2.7) is summed over all active systems. To obtain f a , various assumptions may be used to calculate the rate of change of vectors na and s" (Asaro & Rice 1977). The calculation based on the lattice elastic spin, i.e. (2.8) v simplifies the results and has natural appeal. Indeed, then ia = uijpFj, and, with 7% denoting the current yield stress in shear, one arrives a t the following flow conditions : j" = 0 if 0 < u..p" 2 3 23 < I-$, v = 0 if 0 < a. 23.pa. z~ = r$ and uijpFj < h a p y a , (2.9) v p > o if 0 < uijp$ = r$ and uijpz7 = hapjp,
3

i'L - w;j sj", h$ = we.m? 23 3

where p$j is defined by (2.4a). If potential deviation from the Schmid law, inherent in frictional or pressuresensitive materials, is to be included, then one defines

"1 (2.10) qij - (8% nja sja n$) n$ nja tan 7 , + 6ij tan v and sets ia = uijq$ so that pZj everywhere in (2.9) is now replaced by q2j. For application to geomaterials, tan^, in (2.10) may be viewed as the coefficient of the sliding friction? (being zero when normal stress is tensile, i.e. when uijn$nja > 0).

r2,

t For a macroscopic formulation of plastic flow of frictional materials, see Nemat-Nasser & Shokooh (1980) (see also Nem~t-Nasser 1983).

92

T. Iwakuma and S. Nemat-Nasser

For metals, the last term in (2.10)is in accord with the suggestion that hydrostatic pressure affects the motion of dislocations and hence the corresponding plastic slip. To render the final equations more general, in what follows the tensor qtj is used in place of prj to characterize the flow conditions (2.9). The relation between a prescribed crystal strain rate and the associated slip rates and, particularly, on the depends on the structure of the hardening matrix ha@ magnitude of the off-diagonal elements relative to the diagonal ones (Hill 1966). Here, the possibility of a non-unique relation is noted where other considerations must govern a specific selection among all possible ones (Havner 1982b); see Appendix A. Substitute from ( 2 . 3 ) and (2.6) into the right side of ( 2 . 2 a ) , and use the flow conditions (2.9) with pTj replaced by qFj to arrive a t

Denote the inverse of Nap by Map, and from ( 2 . 2 ) to (2.11) obtain

where the Jaumann rate of the Cauchy stress is

and a and /3 are summed over all active slip systems. Since the nominal stress rate nij is? nij = $ i j + u i j d k k - d i k u k j + ~ j k u k i , it follows that
n 1.3. = F c k l ~ k , l , (2.15) (2.14)

where
p$kz

= Lgkl-

$(Sik&n t 8 i 1 8 k m ) umj
~

-(L$pg~$~-~{i&pf

+ $(8j;.k81m-8jlSkm) umi ? upj p + u& u p + )MaPq!nn(L%nkl - urn, akl)

(2.16)

and where some symmetries of LCkl are used.$

Within a continuum approximation, micro-heterogeneities a t grain level are averaged to arrive a t 'material neighbourhoods' which, to some linear scale, are regarded as locally homogeneous. The linear scale then depends on the dimensions

t n,, is the jt11 component of the tractions transmitted across an elementary area which In the reference state 1s normal to the t t h coordinate direction. n,, is not symmetric, In general. When the reference state colncldes with the current state, then the nomlnal stress equals the Cauchy stress but their rates are not the same, as is evldent from (2.14). $ Note that p,4, 1s zero ~f ( 2 . 4 ~is ) used, but ~tis non-zero when dllatancy 1s included; see the comments that follow (2.4).

Finite elastic-plastic deformation of polycrystals

93

of the constituent grains a t the microscale. For elasto-plastic polycrystals the matter has been discussed by Bishop & Hill (1951);Hill (1972); Havner (1974),and, more recently, Havner ( 1 9 8 2 ~ )I.n what follows, a representative volume subjected on its boundaries either to uniform t'raction rates or to a velocity field compatible with uniform overall velocity gradients, is considered. Theri the overall nominal stress rate measured with respect to the current configuration as reference is given by

qj
( 4 .

where Let

K, be the overall velocity gradient defined by

.) denotes volume average taken over the reference volume.

and write in terms of that of The aim is t o estimate the instantaneous modulus tensor Fijkl e quation (2.16), by suitable averaging over all crystal oriensingle crystals Fz.kl, tations and shapes. To this end, Hill's self-consistent method is used, which requires an examination of the following auxiliary problem. under the overall nominal Consider an extended solid, D, with overall moduli Fijk, stress rate &j, containing a single ellipsoidal crystal, Q, of moduli, Fskl. The stress rate is

and across the boundary aQ of the crystal, traction rates must remain continuous. It turns out that, as in linear elasticity, in the present general setting the velocity gradient in the crystal is uniform for ellipsoidal shapes ($4). I n view of this, one seeks to calculate the concentration tensor, st$kl,

which relates the uniform velocity gradient in the crystal to the far-field uniform data. Prom (3.6), the nominal stress rate in the crystal is

and, therefore, (3.1) and (3.3) require

in accordance with self-consistency. It remains to express the concentration tensor in terms of the crystal and the overall moduli.

94

T. Iwakuma and S. Nemat-Nasser

I n view of the linearity of the boundary-value problem described by (3.4) and (3.5),it is expedient to consider a solution uniform everywhere,

and superimpose on it a perturbation solution,

where v&( x) is the Eshelby-type transformation velocity gradient, identically zero outside of Q, and, as remarked before, turning out constant within R for ellipsoidal regions. Let %&(x- x') be the Green function in an unbounded uniform medium, satisfying

where 8(x - x') is the Dirac delta function, and observe that the solution to the field equations, n$!!i= 0, is given in view of (3.10) by where

It is convenient to introduce a Hill-type constraint tensor, F z k l ,by

and observe that the nominal stress rate within the crystal is, by superposition, Substitution into ( 3 . 4 ~yields ) (3.6),where the concentration tensor is The concentration tensor is affected by the overall and the local moduli, as well as by the constraint tensor FGkZ which itself depends on the overall moduli and on the aspect ratios and the orientation of the ellipsoid fi : equations (3.13) and (3.14). On the other hand, from (3.15),the self-consistency conditions (3.1)and (3.2)require that { F Z k l v k , r ) = W z k l ) %,l , which is satisfied when 9 & is common to all crystals. Observe that this requirement is also shared by the small-deformation, self-consistent account of the polycrystal, where, as pointed out by Hill (1965 a, p. 93), single crystals are regarded as spheres or as ellipsoids with corresponding axes aligned. Fortunately, in the present finite deformation case, the changes of shape and orientation induced by deformation of initially spherical grains are rather small compared with variations in the moduli from grain to grain, and therefore the use of an average (in some sense) concentration tensor may be viewed as justified. At any rate, this or an equivalent requirement is an integral part of the self-consistent method.

Finite elastic-plastic deformation of polycrystals


Based on these comments, the concentration tensor is redefined as
d $ k ~= ( g $ m n

95

+F t . m n ) - '

(g%nkl+

Snlnkl),

(3.17)

where we choose to identify g$,, by Direct calculation now shows that (4ykl) = Iijkl,the identity tensor. An alternative approach is to normalize directly the concentration tensor .01$kl, d $ k l = d$mn(dmfikl)-l, (3.19) and to use to calculate the local velocity gradient. This method has been used by Walpole (1969) to ensure self-consistency in estimating the overall moduli of a linearly elastic polycrystal of non-aligned ellipsoidal constituents. Equations (3.17) and (3.8) (or (3.16), (3.19), and (3.8), if we employ Walpole's method) complete the sought objective. The actual calculations, however, will require numerical effort, even in the simplest case, since the unknown overall moduli, Fijkl, also occur in the right side of (3.8) through the specification of the concentration tensor by (3.17) (or (3.16)). I n the self-consistent method used by Budiansky & Wu (1962) and Hutchinson (1964), the concentration tensor is calculated on the basis of an inclusion embedded in a solid whose moduli coincide with the elastic moduli of the matrix. I n this approach the concentration tensor is given in terms of Hill's constraint tensor, calculated on the basis of the elastic moduli of the polycrystal or the composite. This means that, in place of FntnkZ in (3.13) and the similar quantity in (3.14), the elastic moduli of the polycrystal are used. Note that in the present case, and in general, the overall instantaneous elastic moduli do not equal those of the single crystal. I n the phenomenological approach, the elasto-plastic constitutive relations are generally stated in terms of a relation between the co-rotational (or Jaumann) rate of change of the Kirchhoff stress and the deformation rate tensor,

Stkl

where

Tij is the Kirchhoff stress, and where

where LCij is the Cauchy stress,

and qjkZ is a homogeneous function of degree zero in Dij (Hill 1959).Whereas (3.22) directly follows from (3.2),equation (3.21), in general, does not hold for the average of the corresponding local quantities. This is clear from the expression for the local material rate of change of the Kirchhoff and Cauchy stresses,

96

T. Iwakuma and S. Nemat-Nasser

Similarly, the material and Jaumann rates of the overall Cauchy stress, iij a nd i i j , when defined globally, are not the averages of the corresponding local quantities. These stress measures must be deJined in terms of the overall nominal stress rate and the overall velocity gradient. Guided by (3.23), one may introduce the material rate of the overall Kirchhoff stress by (3.24) T~~ = iij D kk= flji 5,,Jki),

+ zij

$(aij + + K, kzkj +

which ensures the symmetry of both the Kirchhoff and the Cauchy stresses. These stresses are then calculated incrementally from definition (3.24). For example, with t as the load parameter and the Cauchy stress, Zij(t),known, after an increment of loading At, one obtains

Zij(t+ At)

= &(t)

+ &(t)
+zkj

At + O(At2).

(3.25)

From (3.3), (3.20), and (3.24), it follows that

%jkl

= 8 ( 6 j k l +Fjikl

&il

+ zkiajl)

K,l.

(3.26)

Since the left side is independent of the overall spin, (3.26) yields the following restriction on the overall moduli Sijkl:
&jkl

Fjikl

+ Zkjail+Z;ci 8jl =

+Fjilk + Zliaik+ Zli Bjk.

(3.27)

Then

is defined by

Note that these moduli are not necessarily symmetric with respect to the exchange of the leading and the terminal pairs of indices. This is expected, because the corresponding local moduli also lack this symmetry unless a special stress state is involved. With reference to (2.12), observe that terms involving the stress in the right-hand tensor of the instantaneous modulus render this modulus non-symmetric with respect to the exchange of the leading (ij)and the terminal (kl) pairs of indices. Hill (1972) has shown that if the symmetry exists locally, then it will also exist globally. Since the current configuration is used as the reference one, 4. = Tii = Zii and nij = 7.. 1 I = cij. Consistent with uniform tractions prescribed on the boundary of a representative macro-sample, we thus have This suggests that in place of (3.25) one may first compute efj(t + At) from

efj(t + At) = cij(t) + $ij(t) At + O(At2)


and then calculate

Zii(t+ At)

(cii(t + At)).

Consistent with this, the overall Cauchy stress rate may be defined by

Finite elastic-plastic deformation of polycrystals

97

which, in general, does not satisfy (3.24). Hence the overall Cauchy stress and its rate obtained from (3.25) and (3.24),respectively, will not in general coincide with those calculated by using (3.29) and (3.31), although the differences ought to be rather small. Note that operations (3.25) and (3.30) are valid for the Cauchy stress but not, for example, for the nominal stress. The quantity nij(t)+ni,(t)At + O(At2)is the nominal stress a t time t At, but referred to the configuration of the solid a t time t. The numerical results reported in 7 are all based on definitions (3.24)and (3.25).

4. GREEN'SF U N C T I O N Equation (3.12)requires explicit knowledge of the quantity


J.. umn =

Jn

%%,mi($-

xl)dx',

which, as will be shown below, is independent of x when Q is ellipsoidal. Although Green's functions for unbounded media have been calculated for isotropic as well as for various anisotropic solids, the case in question presents considerably greater analytical difficulties, as the number of moduli involved depends on the state of stress and varies in the course of deformation. None the less, the analytical procedure outlined for anisotropic elasticity by Stroh (1958, 1962),Il'illis (1961),and Kinoshita & Mura (1971) (see also Mura 1982) provides guidance in the present context. With the aid of a Fourier transform, from (3.11) it follows that

where 4.3~) ~ ~ ~ (Kij), K ~ ~ K ~ (~ Nij = $ E ~ ~ ~ , =Ecofactor

D
K4j

= r, 6 ikmejln
= %fjl
6k

KklKnrn

det

(Kfj),

&>

(4.3b) (4.3~)

where eijk is the permutation symbol. Equations (4.3) suggest a natural contact between the structure of the Green function and the possibility of loss of stability of the uniformly stressed overall polycrystal, because the nature of the roots of the determinant D in (4.3b) defines the ellipticity, parabolicity, or hyperbolicity of the operator in the left side of (3.1I ) . Necessary conditions for the possibility of jumps of certain components of the velocity gradient across a discchtinuity plane coincide with the existence of the real roots of D = 0. This rather interesting observation emerges as an inherent part of the present finite deformation theory. It has no counterpart in previous works on polycrystals and composites, which have been based on infinitesimal deformation. As discussed in the next section, if the changes in the grains' geometries and orientations and their residual stresses are included in the calculation, then indeed at a certain strain level in uniaxial extension the overall moduli tensor

98

T. Iwakuma and S. Nemat-Nasser

ceases to correspond to an elliptic operator, and the overall uniform deformation state ceases to be stable, leading to possible localized deformation. (For general discussions of discontinuity and localized deformation, see Thomas (1961) and Hill (1961, 1962)). Substitution from (4.2) into (4.1) yields

where fJ(4) is a unit sphere in the <-space, and

E = E L C2={.5. (4.5) I n deriving (4.4), the property that lVn,({) and D({) are fourth and sixth degree polynomials in 6 is used. If the global coordinate axes are taken parallel to the principal axes of the ellipsoid whose semi-principal radii are a,, i = 1,2,3, then integral (4.4) can be further simplified by first identifying the components of the vector 4 as (,/ai (no sum over i) and then defining and in terms of as follows:

Cl

( I - ci)*cos 8,

c2

c2

c,,

(1 - [!I* sin 8.

(4.6)

I n this manner, it follows that

Define

and, with z = eie, obtain

where y is the unit circle for fixed

I(,[

< 1. Equation (4.7) then gives

~ijrnnti3)d~. (4.10) 2 -1 Explicit results emerge for two-dimensional problems whose corresponding Green functions are deduced by the limiting process of a, -t co which renders the integral Thus, in two dimensions, (4.7)independent of
-

Ji jmn

c,.

JijnLn =

Hijmn(C3

01,

(4.11)

l 8,a ~ l s i 0,O). n The where the components of ( now are identifietl with ( a ~cos details are summarized in Appendix B. 5. L O C A L I Z ED D EFORMATION Necking and localized shear bands are of common occurrence in the finite deformation of metal specimens. The analysis of these phenomena under various constitutive assumptions has received considerable attention. Within the framework

Finite elmtic-plastic deformation of polycrystals

99

presented by Hill ( I 962) one seeks conditions under which the governing equations for the rate of deformation from a given initially stressed state cease to be elliptic (Hill & Hutchinson 1975; Storen & Rice 1975; Rice 1976; Needleman & Tvergaard 1977; Hutchinson & Neale 1978; Nemat-Nasser, Mehrabadi & Iwakuma 1980; Iwakuma & Nemat-Nasser 1 9 8 2 ) . I n the present work this has a fundamental bearing on the calculation of the overall moduli by means of Hill's self-consistent method, because such loss in ellipticity prevents further calculation of Green's functions and the corresponding concentration tensors. Locally, the instability condition is attained a t individual grains during the course of plastic flow, before the loss of overall stability. However, the constraint imposed by the surrounding grains renders the overall response stable. As the deformation proceeds, more grains become locally unstable, and the constraining effect of the surrounding grains diminishes accordingly, leading eventually to global instability. This global instability is closely related to the location of the poles that enter the calculation of the concentration tensor. Let v be the unit normal to the surface across which certain components of the velocity gradient may admit jumps. The continued equilibrium condition necessarily requires the continuity of the traction rates across such discontinuity surfaces, which in the present case is guaranteed by where [.. -1 denotes the jump of the enclosed quantity across the discontinuity surface. If q is the magnitude of the jump, then ( 5 . 1 )yields

[v,,21 = r k V 2 ,
%Fiiklvi vlTk = 0.

(5.2) (5.3)

For non-trivial values of q, the determinant of the coefficients in ( 5 . 3 )must vanish,

The necessary condition for localization therefore is that ( 5 . 4 ) yields real orientations v. This is precisely equivalent to the existence of the real roots of D = 0 in the calculation of the Green function, ( 4 . 3 6 , c ) . Localized deformations become possible, therefore, when a pole in either the complex c-plane or the complex {-plane is located on the real axis; see (4.4)-(4.9). For the two-dimensional case, the necessary condition ( 5 . 4 )becomes
el vt - (ell c18)vf v2 - (c13 c17)V: V ; (cg c14)v1V $ c2V: = 0, (5.5) where the coefficients, ci, are given by ( B 3 ) of Appendix B . As is evident from ( B 5 ) and ( B 6), if 2; is a root of Do = 0, then the characteristic equation ( 5 . 5 ) admits real roots for ( v l / v 2 ) ,when lzEl -t 1. Since el is expected to remain non-zero during the course of deformation, ( 5 . 5 ) may be regarded as quartic in ( v 1 / v 2 )If . the instantaneous moduli change smoothly with the deformation, then the condition for the existence of a double real root for

+ +

100

T. Iwakuma and S. Nemat-Nasser

( 5 . 5 ) defines the incipience of localization. Indeed, for the two-dimensional illustrative example of uniaxial loading considered in 5 7, the symmetry reduces ( 5 . 5 )to c1(vl/v2)4 - (el3 + el,) ( v 1 / v J 2+ c2
=

0.

(5.6)

If the instantaneous moduli change smoothly as the axial loading continues, then the inception of localization is marked by the vanishing of the discriminant d = (c13 - 4c1c2, (5.7) associated with the existence of a double root for ( 5 . 6 ) .I n terms of the location of poles for the calculation of Green's function, the vanishing of d corresponds to the state when two poles in the z2-plane, one inside and the other outside of the unit circle, coalesce on the circle.
6. S H A P EC H A N G E

AND ROTATION OF GRAINS

Since the grains are assumed t o be ellipsoidal in shape, their orientation can be defined by the orientation of the ellipsoid's principal axes. The method does not include the size effect in the absolute sense. However, the effect of a change in shape can be included by means of the change in the relative dimensions of the ellipsoid. The self-consistent method assumes uniform deformation of each grain. This suggests that the principal axes of the ellipsoidal grain can be assumed to coincide with the principal axes of the Eulerian triad associated with the corresponding local deformation gradient. Initially, all grains are assumed to be spherical, and the deformation gradient is
Xi, A = S ~ A , (6.1) where x measures the current position of a particle which initially is a t X; xi are components of x, and X, are those of X, i ,A = 1,2,3. Let the deformation gradient be xi,,(t) a t a certain time t for a given crystal. The deformation gradient after incremental loading a t time t + At then is

(6.2) xi, A (t At) [Sij vi,j At] x j,A ( t ) . With the aid of polar decomposition, define the rotation R and the right-stretch tensor U as xi,^ = R ~ B U ~ ~ , (6.3) R - l = RT, detIRI = $ 1 . (6.4) The current orientation of the principal axes of the ellipsoid is given by

where N(a', a

1,2,3, are the principal directions of U ,

here A(,) are the corresponding principal values. Thus, the current orientation of , the corresponding shape is given the grain is given by the Eulerian triad d a )and by the ratios of the principal stretches, e.g. h(,)/A(,)and h(.,)/h(,). As deformation

Finite elastic-plastic deformation of polycrystals

101

proceeds, grains that are assumed to be spherical initially deform by different amounts and attain different orientations. Observe that in the context of the present rather general formulation, such effects as initial anisotropy or initial texture can be included by considering suitably orientated grains of suitable relative dimensions.

I
FIGURE 1. Definitions of kinematic quantities in a typical grain.

It should be emphasized that the assumption of an ellipsoidal grain with a uniform


deformation gradient used in the self-consistent calculation is appropriate, as the most general finite deformation (6.3) maps materials within an initially spherical neighbourhood into an ellipsoid. I n this manner the continuum dimension of the local neighbourhood' is defined by the average grain size of the polycrystal.

Explicit results are presented in this section for plane deformation of face-centred cubic (f.c.c.) polycrystals. For all crystals in the aggregate, the plane of deformation is defined by the local (100)- and (011)-orientationswith two slip lines (and hence four slip systems) symmetrical with respect to the (100)-axis (denoted by itl) a t an angle close t o $ = 35" which follows from the fact that the angle between two slip planes is approximately 70"; see figure 1. It has been suggested (Asaro 1979) that this may be a reasonable two-dimensional model for f.c.c, single crystals. I n what follows we first summarize expressions for this two-dimensional problem, and then present numerical results, where the effect of deviation from $ = 35" is also considered.

t It wlll be ~llustratedby numerical examples that Q slightly larger (smaller) than 35" leads to a 'stiffer' ('softer') polycryst~11111~ response; see figure 2 b .

1 02

T. Iwakuma and S. Nemat-Nasser

Let be the angle between the local ?,-axis and the global xl-direction. The unit vectors defining the slip systems are

7.1. Constitutive equations for single c~ystals

s1 = - s3 =
s2 =

- sin ($- $1

(7.1)
- s4 =

Substitution into (2.4) yields

The spin induced by elastic lattice distortion is

which follows from kinematic assumption (2.8)and E 0. I n the course of deformation, the individual crystal orientations are calculated by updating the angle $, similar to (3.25),as $(t + At) 2 +(t) $(t) At. (7.4)

The elastic moduli of a cubic crystal are usually defined in terms of three elastic constants Cll, C,,, and C4, (in Voigt's notation). Let ZRbe the elasticity tensor of a typical crystal in the usual crystal directions in three dimensions. Then

Since the local El, E,-coordinate system is defined by rotating the crystal coordinates about the (100)-direction by 45", the elasticity tensor in the local ?{-coordinate system, En, is expressed as

Then L" in (2.6) and (2.16)is the crystal elasticity tensor in the global xi-coordinate. It is related to by

z"

Lz'kl =

ail, ajs aknt

@q?nn,

(7.7)

Finite elmtic-plastic deformation of polycrystals


where For the isotropic crystal, cos y9 sin y9 -sin $ cos y9

103

wherep and h are the Lam6 constants. The difference, Cll - (C12+ 2C4,), characterizes the degree of elastic anisotropy in cubic crystals. For small deformations, where the residual stresses are ignored, the overall elastic parameters of the polycrystal can be calculated directly from the elasticity of the single crystal constituents (Hershey 1954; Kroner 1958).For finite plastic deformations, where residual stresses of differing magnitudes accumulate in crystals of different orientations and shapes, the estimate of the effective overall elasticity requires an incremental elastic unloading. The overall elasticity tensor changes in the course of deformation and, in fact, becomes anisotropic (formation of texture), even though i t is initially isotropic because of the random distribution of the orientation of the crystals. To render the governing equations dimensionless, it is convenient to introduce the yield shear strain, Yo ~$/c44, (7.9) and set S i j = f l i j / ~ $ , 5" = ?'"/Yo, dij = dij/yo, (7.10) = dzj/yo, %,pj = dFj/yo, kaP = haP/C4,,

Gj

where 7% is the common initial yield shear stress. All second-order terms will now be proportional to the parameter yo. These terms are not present in the small strain theory of crystal plasticity (Hutchinson 1970). Hence to retrieve results for the case where residual stresses as well as the effects of crystal rotation and shape change are ignored, simply exclude terms which are proportional t o yo. From (2.16), the non-dimensional instantaneous moduli are expressed as follows:

where the fact that pTi

0 is used and where gal is the inverse of m"P defined by

Note that we are excluding possible pressure sensitivity and plastic dilatancy. I n the case of elastically isotropic single crystals,

104

T. Iwakuma and S. Nemat-Nasser


f l a p = kaP + 2 p f jpfj.

where v is the Poisson ratio and g a p is the inverse of

The flow conditions ( 2 . 9 ) , with

7% = 7%

+ haPyB, become
v sij p f j < kaPgP, v 82.1 p f j = kaP[P,
,

p =0 &=0 &B0

if if if

0 < s i j pfj < I

+ kaP[P,

0 < s i j p$

1 + kaP<P and

(7.15)

0 < sii pfj = 1 + kaP<P and

and thus the plastic part of the deformation rate is

while the elastic part is related to the stress rate as

I n particular, for the elastically isotropic crystal, (7.17) can be written as

are obtained when terms As is seen, the equations used by Hutchinson (1970) proportional to yo are excluded. Therefore, terms with y o in the above equations characterize the effects due to residual stresses and the shape change and rotation of individual grains.
7.2. Constitutive relations for polycrystals

The overall instantaneous moduli of the polycrystal are given by (3.8),in which the concentration tensor is given by ( B 9 ) of Appendix B as

depending on whether we impose self-consistency? through (3.17)or through ( 3 . 1 9 ) ; in (7.19)


FijkZ
%jk~/~44

(7.20) (7.21)

and

Jijkl

-- (C44hjkl),

where Ajkl is given by ( B 7 ) . The quantity Jjmn Fm,,, corresponds to gijkl defined in (3.19). The expressions in Appendix B are with respect to the local Eulerian deformation 8 define the orientaaxes; see 6. To transform to the global coordinate system, let ,
Note that normalization of the concentration t,ensor is required only when grains are unaligned ellipses (in two dimensions), i.e. when shape changes are inclriclrd.

Finite elastic-plastic deformation of polycrystals

105

tion of the major principal axis of the local Eulerian ellipse (in two dimensions) relative to the global xl-direction; figure 1. Then

where a$ is obtained from (7.7) by replacing y? with P and by taking its transpose. Finally, the condition for the objectivity of the global instantaneous moduli, (3.27),becomes (7.23) F1212 +F2112 +7oX1l = F1221 +F2121f 7 0 8 2 2 9 where Sij Zii/r;. (7.24)

7.3. Localization The necessary condition for the inception of localized deformations is given in two dimensions by (5.5).If we can estimate the limiting values of xi, the poles for the integrand in (4.9)on the unit circle, then the orientation of localization relative to the local x,-axis, 0, becomes tan 0
=

v1/v2

Im [zi]/a(l -Re [xi])

(7.25)

(compare (5.5)with Do= 0 in Appendix B), where a is the aspect ratio of the grain, (B 1). Since the evaluation of the integral (4.9)is done in the local coordinates, the localization orientation relative to the global x,-axis, 0 , is given by

7.4. Results Numerical results are obtained for uniaxial tension. All quantities are referred to the current configuration and are then updated incrementally after each incremental loading. For each crystal, local deformation and stresses are calculated step by step, and then the overall instantaneous moduli are estimated by (3.8). First, typical results are discussed below for the case in which the residual stresses and rotations and shape changes of grains are ignored, to compare the numerical results with those presented by Hutchinson (1970). Then the general case is examined. To neglect residual stress and grain rotation and shape changes, exclude secondorder terms; these are proportional to yo. For isotrbpic single crystals, the Poisson ratio v = 4 is considered. For the anisotropic case, the elastic parameters associated with copper, C12/Cll = 0.722, C4,/Cll = 0.447, (7.27) are used. Figure 2 shows the stress-strain curves for the locally non-hardening model and compares results with the three-dimensional calculations by Hutchinson (1970) Table 1 shows some examples of local quantities and several cases of the overall histories. In these tables, , I T denotes the number of grain orientations chosen

T. Iwakuma and S. Nemat-Nasser

---1

Hutchinson (1970) Bishop & Hill (195 I)

%I70
FIGURE 2a, 6 . Tensile stress-strain curves; N = 46, second-order terms
are excluded, no hardening.

between 0" and 90, and 1 and 6, are the elongations in the xl- and x,-directions, evaluated by = n(l+DllAt)-1, 6, = n(l +DZ2At)-1, (7.28) where II denotes the product taken over all preceding (variablel-) loading steps, At; the parameter 7 , is the effective yield shear strain defined by

where ,Z is the initial overall shear modulus of the polycrystal calculated in a selfconsistent manner, as will be discussed later. I n the case of elastic isotropy, 7, is equivalent to yo of (7.9).

The 'loading' steps are adjusted to ensure accuracy t o within a prescribed limit.

Finite elastic-plastic deformation of polycrystals


(Second-order terms are excluded ; elastically isotropic grains.)

107

TABLE I b. GLOBAL QUANTITIES ; N

46
no. of grains in plastic state 0 10 14 18 22 24 26 28 30 30 32 34 36 38 38 40

(Second-order terms are excluded; elastically isotropic grains, no hardening.)


%/Yo 0.673 0.678 0.683 0.698 0.723 0.758 0.808 0.878 0.968 1.028 1.098 1.293 1.60 2.04 2.32 3.02 %/YO - 0.337 - 0.339 - 0.343 - 0.354 - 0.374 - 0.404 - 0.448 -0.510 - 0.592 - 0.648 - 0.713 - 0.897 - 1.19 - 1.61 - 1.89 - 2.58

z11/7;
2.020 2.035 2.044 2.066 2.093 2.124 2.162 2.207 2.256 2.284 2.313 2.378 2.458 2.546 2.587 2.650

Fllll
4.000 3.539 3.411 3.308 3.238 3.202 3.170 3.143 3.119 3.107 3.086 3.067 3.051 3.036 3.022 3.017
=

F121,,,,
1.OOO 0.994 0.981 0.956 0.858 0.827 0.785 0.733 0.668 0.538 0.466 0.386 0.300 0.212 0.117 0.0885

TABLE 1 C. GLOBAL QUANTITIES; N

46
no. of grains in plastic state 0 10 14 18 22 24 26 28 30 32 34 36 38 40 41

(Second-order terms are excluded; elastically anisotropic grains, no hardening.)

108

T.Iwakumn and S. Nemat-Nasser

As is seen from figure 2a, the two-dimensional results are in good accord with the three-dimensional ones. Figure 2 b illustrates the effect of varying the angle 4 on the overall response. Hence it is possible to choose an appropriate 4 to make the two-dimensional results analogous to the three-dimensional ones. The choice of 4 z 35" seems to be reasonable. The volumetric change and the change in the global instantaneous moduli are shown in figures 3 and 4, respectively.

isotropic

cl/rO FIGURE 3. Volume changes; N = 46, second-order terms are excluded, no hardening.

FIGURE 4. Changes of global moduli for elastically isotropic case; N = 46, second-order terms are excluded, no hardening.

For the elastically isotropic single crystals, the global instantaneous moduli initially satisfy

The first equation in (7.30) holds throughout the deformation history, whereas

Finite elastic-plastic deformation of polycrystals

109

as el becomes large; see table I . Hence the apparent Poisson ratio tends to the value +, as el increases, and thus the overall response becomes incompressible. Locally, however, the moduli no longer correspond to an isotropic tensor, as illustrated in table I for ~ ~ = 4.12. / y The ~ asymptotic results ( 7 . 3 0 ) and ( 7 . 3 1 ) conform to expected macroscopic observations. Observe that, even if individual grains are elastically anisotropic, the global initial modulus tensor is isotropic, because of the random distribution of the grains. i , i?, and ti denoting the overall shear modulus, bulk modulus, and Indeed, with j Poisson ratio, respectively, results shown in table 1 c indicate that

These estimates deviate slightly from those obtained by Hutchinson (1970), mainly because of the difference between bhe three-dimensional and two-dimensional models used. From ( 7 . 2 7 ) and ( 7 . 3 2 ) ,the effective yield shear strain becomes

As has been mentioned before, no localization of deformation is predicted since the effects of residual stresses and of crystal rotations and shape changes are ignored in the calculations above. However ( 7 . 3 0 ) and ( 7 . 3 1 )indicate that

Fllll

l / ( l - 2 v ) a,,

q12, l / ( l - 2 v ) - a,)

F1212

a,)

(7.34)

where a, and a, are positive parameters that become very small a t larger axial strains. Moreover, figure 4 suggests that a, approaches zero much faster than a,. Therefore by putting a1 elfrn, a, 6, rn > 0 , (7.35) and substituting ( 7 . 3 4 ) and ( 7 . 3 5 ) into the necessary condition ( 5 . 5 ) ,we find

which implies that instability may occur as ellyo -t oo, and that, from ( 7 . 3 6 ) ,the orientation of the localization, 0, equals 45". As is indicated by ( 7 . 3 5 ) ,the change is more gradual than that of the other compoin the shearing components of Fij,, nents. This effect must be included in the phenomenological constitutive models, otherwise these models will not predict localized deformations even for very large axial strains. Since only a finite number of crystal orientations is used in the calculation, the stress-strain curves are actually piecewise linear, and for large strains especially the behaviour depends on the number N of the orientations used. Figure 5 shows results for the Taylor hardening and elastically isotropic grains, where the 'plastic' part of the elongation is defined by

9 = 1-{(I

-v

) / ~ , UCll. )

(7.37)

The yield surface of the polycrystal can be obtained by unloading from various

110

T. Iwakuma and S. Nemat-Nasser

stress states. The evaluation is simplified if elastic isotropy is assumed and the residual stresses and the crystal rotations and shape changes are neglected. Locally the stress state in each grain is not uniaxial, and, therefore, the overall yield surface, as an envelope of the local yield surfaces, exhibits a corner and a Bauschinger effect consistent with the changes of the global instantaneous moduli. A typical result is shown in figure 6.

3.0 -

--0

Hutchinson (1970)

4'/7O
FIGURE 5. Stress-strain curve for elastically isotropic case with Taylor hardening; N = 46, second-order terms are excluded, ka@ = 0.02.

211/7$

FIGURE 6. Yield surfaces for elastically isot~opic case a t s,/y, = 0, 2.02, and 4.12; A' = 46, second-order terms are excluded, no hardening.

Consider now the case when residual stresses and crystal rotations and shape changes are included, i.e. when second-order terms (proportional to yo)are included. The calculation follows the same procedure as before, but close to the inception of localization the evaluation of the Green function breaks down, its the operator Pijkla2/axiax,
tends to lose its ellipticity. This is manifested by the loss of con-

Finite elastiqlastic deformation of polycrystals

111

vergence of the iterative scheme necessary for the evaluation of the overall instantaneous moduli. However, the critical axial stress and the orientation of localized deformation can be estimated. To illustrate this and related points, consider several values for yo, e.g. yo = 0.1,0.05,0.01. (7.38) These are rather large, but serve to identify the main features of the problem.

FIGURE 7. Tensile stress-strain curves for indicated values of yo (yo= 0 when second-order terms are excluded); N = 46, no hardening. ( a ) elastically isotropic case; (b) elastically anisotropic case.

Figures 7a, b show the corresponding stress-strain curves for both elastically isotropic and anisotropic crystals. These results are based on the concentration tensor normalized according to (7.19a). The numerical calculation becomes unstable when e,/yo attains the values given in tables 2-4. This instability is associated with the finite movement of the poles in the complex z2-planeand with one or more additional grains becoming plastic during the iteration procedure. The erratic variation in the location of poles stems in part from the fact that only a finite number of crystal directions, N, are used. Before instability, however, the poles, for the case of elastically isotropic crystals, move continuously along a straight line emanating from the origin of the complex plane, i.e, on
-

6 ' a const.

(7.39)

Values of this angle are given in tables 2-4 for indicated values of yo. Based on this, we may estimate the localization angle from (7.25) and (7.26) as
@
=

arctan [sin B/a( 1 - cos B)]-

p.

(7.40)

Note that (7.40) yields O = 45" for yo = 0, i.e. when residual stresses and crystal rotations and shape changes are ignored; in (7.40),put a = 1, P = O0, and 8 = 90".

112

T. Iwakuma and S. Nemat-Nasser

Since the effects of rotation and residual stresses are taken into account in local constitutive relations, the overall instantaneous moduli show non-symmetry but tend to have forms similar to the local ones. Note that the objectivity condition (3.27) or (7.23) is always satisfied (within the accuracy of the numerical estimates).

TABLE 2. GLOBAL AND

LOCAL QUANTITIES AT LOCALIZATION, AND

INSTANTANEOUS AND ELASTIC MODULI

elastically isotropic

anisotropic

global
0.756 2.09 86 44

local
0 0 1.113 0 2.10 .0.0107 0 14.0 13.5 1.125 0.61 2.06 0.0376 -0.0121 16.0 15.6 1.124 0.72 2.07 0.0300
-0.0137

instantaneous and elastic moduli


near yield point
pllll

Fzzzz
PI122

F Z Z l l

Flzlz
p1221

3.80 4.00 2.00 2.00 0.901 1.10

in at localization unloading 3.17 3.79 3.23 4.01 2.77 2.01 2.63 1.99 0.851 0.896 1.06 1.10

near yield at in point localization unloading 2.53 1.99 2.49 2.73 2.04 2.74 1.39 2.06 1.39 1.39 1.84 1.39 0.569 0.461 0.551 0.783 0.716 0.797

The strong (induced) anisotropy of the instantaneous moduli, as well as that of the elastic moduli a t the bifurcation point, is also shown in tables 2-4. The elastic moduli are obtained by unloading. Also the typical order of magnitude characterizing shape changes and grain rotations is given in these tables. Observe that the axial strain in figures 7 a , b is normalized with respect to 7, (to yo in the isotropic case and to 1.48y, in the anisotropic case). Hence, the estimated axial strains a t localization for y, = 0.1, 0.05, and 0.01, respectively, are e, = 7.6, 5.1, and 2.2 % in the anisotropic case. Data presented in figures 7 a , b and tables 2 to 4 are based on (7.19a). Since, in principle, (2Zkl) must equal the identity tensor, deviation from this may be regarded as a measure of the numerical inaccuracies. For the elastically isotropic crystal, the value of e l l y o to localization increases as yo (the yield shear strain) is decreased, especially when hardening is included. Calculations reveal greater

Finite elastic-plastic deformation of polycrystals


INSTANTANEOUS AND ELASTIC MODULI

113

elastically isotropic global 1.01 2.25


87
44
local 0 14.0 22.0 0 12.6 20.9 1.098 1.104 1.092 0 1.77 4.00 2.19 2.14 2.24 0.0744 0.123 0.0361 0 - 0.0375 - 0.0654 instantaneous and elastic moduli near yield point
Fllll

anisotropic

Fzzzz
E;l22 FZZll

F ~ z ~ 2
Flzzl

3.90 4.00 2.00 2.00 0.950 1.05

at in localization unloading 3.06 3.89 3.08 4.01 2.92 2.00 2.83 1.99 0.525 0.944 0.638 1.06

near yield at in localization unloading point 2.64 2.11 2.63 2.74 2.18 2.75 1.39 1.94 1.39 1.39 1.84 1.39 0.622 0.299 0.616 0.730 0.416 0.734

TABLE 4. GLOBALAND

LOCAL QUANTITIES AT J,OCALIZATION, AND

INSTANTANEOUS AND ELASTIC hlODULI


(S= 46, yo = 0.01.)

elastically isotropic global

anisotropic
1.56 2.64 87 45

eliro
zll/70,

8/deg O/deg Ilr,,,ti,,/deg $Id%


0 0 1.053 0 2.36 0.217 0

2.15 2.54 89 44 14.0 12.8 1.054 2.53 2.29 0.261 - 0.0799

a
PId%
&I
8 2 2 812

local 30.0 29.2 1.038 11.8 2.70 -0.132 - 0.174

0 0 1.052 0 2.32 0.173 0

12.0 11.0 1.056 2.82 2.25 0.241 - 0.161

20.0 29.5 1.035 12.4 2.80 -0.234 - 0.286

Fllll

Fzzzz
F 1 1 2 2 FZZll

&,lz

F,,.,,

instantaneous and elastic moduli near yield in at at in near yield point localization unloading localization unloading point 3.02 3.97 2.71 2.72 2.06 3.98 3.03 4.01 2.74 2.06 2.75 4.00 2.97 2.00 1.39 2.02 1.39 2.00 1.39 2.95 2.00 1.99 1.39 2.00 0.201 0.985 0.169 0.665 0.662 0.990 0.227 1.02 0.197 0.686 0.689 0.101

T. Iwakuma and S. Nemat-Nasser

cllro FIGURE8a. Tensile stress-strain curves for indicated values of yo (upper dot-dash curve is for the case where second-order terms are excluded) ; N = 46, no hardening, elastically isotropic case.

i
o*

4 -

10

20

30

%/To

FIGURE 8b. Tensile stress-strain curves for indicated values of y o ; L V = 46, Taylor
hardening with k f f b= 0.01, elastically isotropic case.

numerical stability when the concentration tensor is normalized in accordance with Walpole's method, i.e. when (7.19b) is used. Figures 8a, b are obtained in this manner. Table 5 compares some of the data estimated on the basis of the two normalization methods, ( 7 . 1 9 ~and ) (7.19b), for yo = 0.01 and no hardening. Results presented in figures 7 a , b and 8a, b show that instability by localized

Finite elastic-plastic deformation of polycrystals


TABLE 5 . EFFECT O F NORMALIZATION
WITH
METHOD,

115

( 7 . 1 9 ~COMPARED )

( 7 . 1 9 b ) , ON

RESULTS AT LOCALIZATION

(N = 46, yo = 0.01, elastically isotropic crystals.)

deformation may start while the sample exhibits overall positive hardening; in the case of figure 8 b , instability occurs in the presence of local (and hence a strong global) positive hardening. This work has been supported by the National Aeronautics and Space Administration under grant NAG 3-134 to Northwestern University.

REFERENCES
Asaro, R. J. 1979 Acta metall. 27, 445-453. Asaro, R. J. & Barnett, D. M. 1975 J. Mech. Phys. Solids 23, 77-83. Asaro, R . J. & Rice, J. R. 1977 J. Mech. Phys. Solids 25, 309-338. Bassani, J. L. 1977 Int. J . mech. Sci. 19, 651-660. Batdorf, S. B. & Budiansky, B. 1949 A mathematical theory of plasticity based on the concept of slip. Tech. Note no. 1871, Nat. Adv. Comm. Aeronaut. Bhargava, R. D. & Radhakrishna, H . C. 1964 J. phys. Soc. Japan 19, 396-405. Bishop, J. F. W. & Hill, R . 1951 Phil. Mag. 42, 414-427; 1298-1307. Boas, W. 1935 Helv. phys. Acta 8, 674-681. Boas, W. & Schmid, E. 1934 Helv. phys. Acta 7, 628-632. Bruggeman, D. A. G. 1934 Z. Phys. 92, 561-588. Budiansky, B. 1965 J . Mech. Phys. Solids 13, 223-227. Budiansky, B. & Wu, T. T. 1962 I n Proc. 4th U.S. National Congress of Applied Mechanics, pp. 1175-1185. New York: A.S.M.E. Chen, W. T . 1967 Q. JI Mech. appl. Math. 20, 307-313. Dyson, B. F., Loveday, M. S. & Rodgers, M. J. 1976 Proc. R. Soc. Lond. A 349, 245-259. Eshelby, J. D. 1957 Proc. R. Soc. Lond. A 241, 376-396. Eshelby, J. D. 1959 Proc. R. Soc. Lond. A 252, 561-569. Havner, K. S. 1974 2. angew. iWath. Phys. 25, 765-781. Havner, I<.S. 1979 J . Mech. Phys. Solids 27, 415-429. Havner, K. S. 1982a I n Mechanics of solids (ed. H . G. Hopkins & M. J. Sewell), pp. 265-302. Oxford: Pergamon Press. Havner, K. S. 1982b Mech. Muter. 1, 97-111. Havner, K. S. & Shalaby, A. H . 1977 Proc. R. Soc. Lond. A 358, 47-70. Hershey, A. V. 1954 J . appl. Mech. 21, 236-240. Hill, R . 1950 The mathematical theory oj plasticity. The Dxford Engineering Science Series. Oxford : Clarendon Press. Hill, R. 1952 Proc. phys. Soc. A 65, 349-354. Hill, R. 1958 J. Mech. Phys. Solids 6, 236-249. Hill, R. 1959 J. Mech. Phys. Solids 7, 209-225. Hill, R. 1961 I n Progress i n solid mechanics (ed. I . N . Sneddon & R. Hill), vol. 2, pp. 247-276. Amsterdam : North-Holland. Hill, R . 1962 J. Mech. Phys. Solids 10, 1-16. Hill, R. 1965a J. Mech. Phys. Solids 13, 89-101. Hill, R. 1965b J. Mech. Phys. Solids 13, 213-222. Hill, R . 1966 J. Mech. Phys. Solids 14, 95-102.

T. Iwakuma and S. Nemat-Nasser


Hill, R. 1972 Proc. R . Soc. Lond. A 326, 131-147. Hill, R . 1975 Math. Proc. Camb. phil. Soc. 77, 225-240. Hill, R . 1979 Math. Proc. Camb. phil. Soc. 85, 179-191. Hill, R. & Hutchinson, J. W. 1975 J . Mech. Phys. Solids 23, 239-264. Hill, R . 8E Rice, J. R . 1972 J. Mech. Phys. Solids 20, 401-413. Huber, A. & Schmid, E. 1934 Helv. phys. Acta 7, 620-627. Hutchinson, J. W. 1964 J . Mech. Phys. Solids 12, 11-24; 25-33. Hutchinson, J . W. 1970 Proc. R. Soc. Lond. A 319, 247-272. Hutchinson, J. W. & Neale, K . W. 1978 I n Mechanics of sheet metal forming: material behaviour and deformation analysis (ed. D. P . Koistinen & N.-M. Wang), pp. 127-153. New York, London: Plenum Press. Iwakuma, T. & Nemat-Nasser, S. 1982 Int. J.Solids Struct. 18, 69-83. Kikuchi, M. & Weertman, J. R. 1980 Scr. metall. 14, 797-799. Kinoshita, N. & Mura, T. 1971 Physica Status Solidi A 5, 759-768. Kocks, U. F. 1958 Acta metall. 6, 85-94. Kocks, U. F. 1970 Metall. Trans. 1, 1121-1143. Kroner, E. 1958 2. Phys. 151,504-518. Kroner, E. 1961 Acta metall. 9, 155-161. Lin, T. H . 1957 J. Mech. Phys. Solids 5, 143-149. Lin, T . H . 1971 I n Advances i n applied mechanics (ed. C.-S. Yih), vol. 11, pp. 255-31 1. New York: Academic Press. Mandel, J. 1965 Int. J . Solids Struct. 1, 273-295. Mandel, J. 1974 I n Foundations of continuum thermodynamics (ed. J . J . D. Domingos, M. N. R. Nina & J. H. Whitelaw), pp. 283-304. London: MacMillan Press Ltd. Mandel, J. 1981 Int. J . Solids Struct. 17, 873-878. Mura, T. 1982 ~Uicromechanics of defects i n solids. Boston, The Hague: Martinus Nijhoff. Needleman, A. & Tvergaard, V. 1977 J . Mech. Phys. Solids 25, 159-183. Nemat-Nasser, S. 1979 Int. J . Solids Struct. 15, 155-166. Nemat-Nasser, S. I 982 Int. J. Solids Struct. 18,857-872. Nemat-Nasser, S. 1983 J. appl. Mech. 50, 1114-1126. Nemat-Nasser, S., Iwakuma, T. & Hejazi, M. 1982 Mech. Mater. 1, 239-267. Nemat-Nasser, S., Mehrabadi, M. M. & Iwakuma, T. 1980 I n Three-dimensional constitutive relations and ductile fracture (ed. S. Nemat-Nasser), pp. 157-1 72. Amsterdam : NorthHolland. Nemat-Nasser, S. & Shokooh, A. 1980 Int. J. Solids Struct. 16,495-514. Reuss, A. 1929 2. angew. Math. Mech. 9, 49-58. Rice, J. R. 1976 I n Proc. 14th Int. Congress of Theoretical and Applied Mechanics, vol. 1 , pp. 207-220. Amsterdam: North-Holland. Saegusa, T., Uemura, M. & Weertman, J. R. 1980 Metall. Trans. A 11, 1453-1458. Storen, S. & Rice, J. R . 1975 J. dlech. Phys. Solids 23, 421-441. Stroh, A. N. 1958 Phil. Mag. 3, 625-646. Stroh, A. N. 1962 J . math. Phys. 41, 77-103. Taylor, G.I . 1934 Proc. R. Soc. Lond. A 145, 362-404. Taylor, G.I. 1938 J . Inst. Metals 62, 307-324. Taylor, G.I . & Elam, C. F. 1923 Proc. R . Soc. Lond. A 102,643-667. Taylor, G.I. & Elam, C. F. 1925 Proc. R . Soc. Lond. A lb8,28-51. Taylor, G.I. & Elam, C. F. 1926 Proc. R . Soc. Lond. A 112, 337-361. Thomas, T. Y. 1961 Plastic $ow and fracture i n solids. New York: Academic Press. Voigt, W. 1889 Wied. Ann. 38, 573-587. Walpole, L. J . 1969 J. Mech. Phys. Solids 17,235-251. Willis, J. R. 1964 Q . Jl Mech. appl. Math. 17, 157-174. Zarka, J. 1973 J. Me'c. 12,275-318.

Finite elastic-plastic deformation of polycrystals

117

Depending on the hardening rule considered, Nap in (2.11)may become singular, which would lead to possible non-uniqueness in the choice of the active slip systems. For example, the case of Taylor's hardening, hl1 = h12, with 4 = 45", renders Nap singular. Hence there are an infinite number of solutions for ya, and all combinations give the same rate of plastic work on the active slip systems. Another example is the case where

in which the matrix Nap is not singular, but there are several alternatives arising from flow rules (2.9).Usually there are three possibilities: either one of the two slip systems is inactive or both are active. The latter case yields a smaller rate of plastic work than the former. Havner (1982b) discusses these and related cases and suggests that the choice should be made on the basis of the minimum rate of plastic work.

[For Appendix B see overleaf.]

118

T. Iwakuma and S. Nemat-Nasser


I N TWO DIMENSIONS

For simplicity in notation, let the principal axes of a typical elliptical grain coincide with the global coordinate directions. The results for other cases are then obtained by the usual tensor transformation; see $7. Define

a
1 '5 '9

= a,/a,,

where a, and a, are the principal radii of the elliptical grain, and set
= '1111
'2 '6 '10 '14

= '2111,
= '2121, = '2112,

'3 '7

= '1211,
= '1221,

'4 '8 '12 '16

= '2211,

= '1121, = '1112,

= '2221, = '2212,
= '2222'

' 1 1 = '1212, '15

'13 = '1122,

= '2122,

= '1222,

(B 1)

(112)

From ( B 2 ) , define the following additional non-dimensional coefficients:

c1 = Fl F, - F3F5,
C3 = C5

' 2 = '10 '16 - '12 ' 1 4 ,

F7F13-F'F15,
'10

C4

= '6'16-'8'14,
-' 3
'9,

= '4

-' 2
-'4

'12,

' 6 = ' 1 '11

C7=F5F16+F6F15-F7F14-F8F13, Cg = '2'16 c11 = F5F;l '13 '14 '14,

C8=F4F9+F3F.F;O-F2F11-F1F12,
C l= ~'3'13

-' 1
-'6

'15, '129

-'7%

'12

= '8

'10

= F5F12+?ll(F6+F13)-F7F10-F9~F8+F15)~ = 'gFl6+
'10('8+ '6('4+ '15)

-'11

'14-

'12('6+

F13),
'9)y

(B 3)

' 1 5 = '5'12+

Fll)-F7F10-F8(F2+

'16 = F 3 F 6 + F 5 ( F 4 + F l l ) - F l F ~ - F 7 ( F 2 + F 9 ) ,
'17 = F3?14+F4(F6+F13)-F1 Cl8 '19 '16-'2('8+'15),

= '4'5+'3('6+?l3)-'2'7-'1('8+'15),
= '3'14+
'13('4+ '11)'1'16'15('2+ '13 'S),
'11).

'20 = '10'15

+ '16('2

+' 9 )

- '12

- '14('4+

The calculation of the concentration tensor requires in (3.13) or Jijkl in (4.1), where Jijk, in two dimelisions can be calculated directly by complex integration ( 4 . 9 ) which involves the following line integrals on the unit circle y :'

aijkl

where

D,

Do(z)= D l z s + D 2 z 6 + D 3 z 4 + D ~ z 2 + D ~ .

(B5)

Finite elastic-plastic deformation of polycrystals


Here the superimposed stands for the complex conjugate, and

119

Dl = (el + a2(c13 c,,) + a 4 c 2 ) ia(cll + c18+ a 2 ( c g+ c 1 4 ) ) ,


D2 = 4 ( c 1- a 4 c 2 ) 2ia{cll + c18- a 2 ( c g+ c 1 4 ) ) ,
D, = 6c1- 2a2(c13 + e l , ) + 6a4c,.

Using (B 2 ) and ( B 4 ) , we express Jijklexplicitly as

+ +

are: Therefore, from ( 3 . 1 3 ) ,the components of .%7.jk,

corresponds to the Eshelby tensor S i j k lin the As is clear from the context, gijkl infinitesimal deformation theory. For the numerical calculation, from ( 3 . 1 3 ) , ( 3 . 1 4 ) , and ( 4 . 1 ) , the concentration tensor ( 3 . 1 6 ) is expressed as

You might also like