VII International Conference on Computational Plasticity

COMPLAS VII
E. Oñate and D.R.J. Owen (Eds)
c CIMNE, Barcelona, 2003

MICROSTRUCTURE DEVELOPMENT IN STANDARD

DISSIPATIVE SOLIDS BASED ON INCREMENTAL
ENERGY MINIMIZATION PRINCIPLES

Christian Miehe

Institute of Applied Mechanics (Civil Engineering)

University of Stuttgart
Pfaffenwaldring 57, 70550 Stuttgart, Germany
e-mail: cm@mechbau.uni-stuttgart.de

Key words: Microstructures, Energy Minimization, Relaxation, Homogenization, Stan-

dard Dissipative Materials.

Abstract. The lecture provides an overview about recent developments in the formula-
tion and numerical implementation of incremental minimization principles for inelastic
solids and their exploitation with regard to the analysis of deformation microstructures.
The point of departure is a general internal variable formulation for standard dissipative
materials. Consistent with this type of finite inelasticity we outline a distinct incremen-
tal variational formulation of the local constitutive response where an incremental stress
potential is obtained from a local minimization problem with respect to the internal vari-
ables. The existence of the incremental stress potential allows the formulation of IBVPs
for standard dissipative solids as a sequence of incremental minimization problems. The
stability of these incremental problems is controlled by weak convexity properties of the
incremental stress potential. Micro–structure developments in incrementally non–convex
dissipative solids can be resolved by relaxation methods based on convexification analyses.
The relaxed problems provide a well–posed overall response of the instable dissipative solid
as close as possible to the original problem. We consider the basic set up of these relaxation
analyses for dissipative standard materials in terms of incremental minimization problems
for exact and approximated quasi– and rank–one convexifications and discuss details of
their algorithmic implementations. The incremental energy minimization is applied to
two conceptual model problems which treat microstructure developments in homogeneous
and heterogeneous dissipative solids.

1
 Christian Miehe

1 INTRODUCTION
The paper overviews recent developments in the formulation and numerical implemen-
tation of incremental minimization principles for inelastic solids and their exploitation
with regard to the analysis of deformation microstructures. Key aspect is the outline of
minimization principles for standard dissipative materials that govern the microstructure
development of a priori heterogeneous materials as well as deformation phase decompo-
sitions in initially homogeneous materials. These principles allow to recast incremental
initial boundary–value problems for inelastic solids into minimization problems. The mini-
mization structure provides a fundamental canonical approach to inelastic solid mechanics
under quasistatic conditions, with important consequences to almost all subsequent as-
pects of the modelling and the numerical implementation: (i) The existence of solutions
of the incremental IBVPs depends on weak convexity properties of a stress potential.
(ii) Stress update algorithms in plasticity, viscoplasticity and damage mechanics appear
in a natural format in the form of energy minimizers. (iii) Micro–to–macro transitions
for the modelling of the overall response of a priori given heterogeneous microstructures
can be defined in terms of a principle of minimum averaged incremental energy. (iv)
The material stability of incremental plastic deformations can be defined in terms of the
quasi–convexity of incremental stress potentials in analogy to elasticity, which provides
a more general criterion than the classical Hadamard condition. (v) Deformation micro–
structure developments in non–stable solids can be interpreted as phase decompositions
and be resolved by energy minimization principles of quasi– and rank–one convexifications
of the incremental stress potential.
Constitutive Minimization Problem (C). The point of departure is a general inter-
nal variable formulation for the constitutive response of standard dissipative materials in
terms of an energy storage and a dissipation function. Nonlinear standard materials cover
a broad spectrum of constitutive models in finite elasticity, viscoelasticity, plasticity or
damage mechanics, see for example [39], [10] and [31]. Consistent with this type of finite
inelasticity we outline a distinct incremental variational formulation of the local constitu-
tive response as proposed in [18] and [22], where a quasi–hyperelastic stress potential is
obtained from a local minimization problem with respect to the internal variables. It is
shown that this local minimization problem determines the internal state of the material
for finite increments of time. The approach is conceptually in line with earlier works on
incremental variational formulations and variational updates in plasticity by [32], [34] and
[4].
Minimization Problem (R) of Relaxation. The existence of this variational formulation
allows the definition of the material stability of an inelastic solid based on weak convexity
conditions of the incremental stress potential in analogy to treatments of finite elasticity,
see for example [1], [6], [14] and [36]. Furthermore, material instability phenomena may
be interpreted as deformation microstructure developments associated with non–convex
incremental stress potentials in analogy to elastic phase decomposition problems. These

2
 Christian Miehe

micro–structures can be resolved by the relaxation of non–convex energy functionals based

on a convexification of the stress potential, see for example [29], [16], [7], [8], [32], [33].
The relaxed problem provides a well–posed formulation for a mesh–objective analysis
as close as possible to the non–convex original problem. Following the recent works
[20], [15] and [21], we give a conceptual outline of a relaxation analysis for non–convex
standard dissipative solids and specify a procedure to a one–level rank–one convexification
procedure.
Global Minimization Problem (H) of Homogenization. The existence of the quasi–
hyperelastic incremental stress potentials for standard dissipative materials also allows
the extension of homogenization approaches of elasticity such as outlined in [37], [30], [28],
[35] to the incremental setting of inelasticity. Following the works [18], [19] and [22], we
outline an incremental variational formulation of the homogenization problem for standard
dissipative materials at finite strains, where a quasi–hyperelastic macro–stress potential is
obtained from a global minimization problem with respect to the fine–scale displacement
fluctuation field. It is shown that this global minimization problem determines the state
of the heterogeneous micro–structure for finite increments of time.
In the subsequent three sections we successivly treat the above mentioned minimization
problems (C), (R) and (H), where Box 1 provides a guide. The Boxes 1, 2, 3 outline
algorithmic treatments of the problems (C), (R) and (H) and point out interesting formal
analogies with regard to the representation of the stress updates and the consistent tangent
moduli. The performance of the formulations is demonstrated for two conceptual model
problems which treat microstructure developments in homogeneous and heterogeneous
materials.

2 MINIMIZATION PROBLEM FOR DISSIPATIVE MATERIALS

2.1 Internal variable formulation for standard dissipative materials
Let F ∈ GL+ (3) with det[F ] > 0 be the homogeneous deformation of a material at
time t ∈ R+ . Focussing on purely mechanical problems, the homogeneous stress response
of the material is physically constrained by the so–called Clausius–Planck inequality for
the dissipation of the material
D := P : Ḟ − ψ̇ ≥ 0 , (1)
where P denotes the nominal (first Piola) stress tensor. ψ is an objective energy storage
function that is assumed to depend on the deformation F and a generalized vector I ∈ G
of internal variables. G indicates a vector space Rn of n scalar functions of internal
variables which may be constrained to a manifold, e.g. the Lie group SL(3) of unimodular
tensors in isochoric finite inelasticity. Insertion into (1) yields, by a standard argument,
the constitutive equation for the stresses

P = ∂F ψ(F , I) (2)

3
 Christian Miehe

and the reduced dissipation inequality

D = F · İ ≥ 0 with F := −∂I ψ(F , I) , (3)
where F ∈ Rn is a generalized vector of internal forces conjugate to the internal variables
I. A broad spectrum of materials is covered by the so–called standard dissipative materi-
als where the evolution İ of the internal variables on G is governed by a scalar dissipation
function φ, depending on the flux İ of the internal variables and the internal variables
I themselves, see for example [39] and [10]. It governs the evolution of I in time by the
constitutive differential equation

∂I ψ(F , I) + ∂İ φ(İ, I) = 0 with I(0) = I 0 (4)

We denote this equation as Biot’s equation of standard dissipative systems, referring to

[2]. The two constitutive equations (2) and (4) determine the stress response of a smooth
normal–dissipative material in a deformation–driven process where F is prescribed. Plas-
ticity and dry friction are time–independent or non–viscous irreversible processes, gov-
erned by non–smooth dissipation functions. These functions are positively homogeneous
of degree one with respect to the flux İ, i.e. φ(İ, I) = φ(İ, I) for all  ∈ R+ . Such
a function has a cone–like graph and is not differentiable at the point İ = 0. A gener-
alization of the above formulation needs a generalization of the differential operator of
smooth functions to the notion of a sub–differential operator ∂ of non–smooth convex
functions as follows, see [26], [10] and references therein. Using the definition of the sub–
differential, in what follows we formally understand Biot’s equation (4) to be generalized
to 0 ∈ ∂I ψ(F , I) + ∂İ φ(İ, I).

2.2 Incremental variational formulation of inelastic materials

We now proceed with the construction of an integrated version of constitutive equations
giving a consistent approximation of the continuous differential equation (4) in a finite
increment [tn , tn+1 ] ∈ R+ of time. Conceptually following the works [32] and [4] on
plasticity, the key point is the definition of an incremental stress potential function W
depending of the deformation F n+1 := F (tn+1 ) at time tn+1 that determines the stresses
at tn+1 by the quasi–hyperelastic function evaluation

P n+1 = ∂F W (F n+1 ) (5)

Clearly, this function must cover characteristics of the storage function ψ and the dissi-
pation function φ introduced above. To this end, consider the minimization problem
Z tn+1
W (F n+1 ) = inf [ ψ̇ + φ ] dt with I(tn ) = I n (6)
I∈G tn

4
 Christian Miehe

(M) Constitutive Model. F ∈ GL+ (3) at X ∈ B is the local deformation of the

material and I ∈ G a generalized vector of internal variables. A normal–
dissipative set of local material equations has the structure
stresses P = ∂F ψ(F , I; X)
evolution equation ∂I ψ(F , I; X) + ∂İ φ(İ, I; X) = 0 , I(0) = I 0
defined in terms of an energy storage function ψ and a dissipation function φ.

(C) Incremental Variational Formulation of Constitutive Model. In a finite time

increment [tn , tn+1 ], the minimization problem of the constitutive response
stresses P n+1 = ∂F W (F n+1 ; X)
Rt
stress potential W (F n+1 ; X) = inf I tnn+1 [ ψ̇ + φ ] dt , I(tn ) = I n
determines the current internal state I n+1 ∈ G and provides a potential for
the stresses at time tn+1 . Box 2 provides a time–discrete version.

(S) Stability of Incremental Constitutive Response. In [tn , tn+1 ] the material is

locally stable if the incremental stress potential W is quasi–convex
1 R
stable response |D| D
W (F n+1 + ∇w(y)) dV ≥ W (F n+1 )
for all possible fluctuations w(y) on the domain D.

(R) Microstructure Development in Homogeneous Materials. For an instable non–

convex response, the incremental minimization problem of convexification
macro–stresses P Qn+1 = ∂F WQ (F n+1R)
1
relaxation WQ (F n+1 ) = inf w |V| B W (F n+1 + ∇w(y))dV

provides a relaxed convex hull WQ of W and determines the current microstruc-

ture fluctuation field w(y). Box 3 outlines a first–order rank–one relaxation.

(H) Microstructure Development in Heterogeneous Materials. F̄ n+1 ∈ GL+ (3) is

the macro–deformation of a heterogeneous microstructure B at time tn+1 . The
incremental minimization problem of homogenization
macro–stresses P̄ n+1 = ∂F̄ W̄ (F̄ n+1 )
1 R
homogenization W̄ (F̄ n+1 ) = inf w |V| B W (F̄ n+1 + ∇w(X); X)dV
determines the homogenized potential W̄ and the fluctuation field w(X) on
the heterogeneous microstructure. Box 4 provides a FE discretization.

Table 1: Overview: minimization principles for standard dissipative solids

5
 Christian Miehe

for dissipative standard materials proposed in [18]. The principle provides an alternative to
the formulation proposed by [32] where a variational principle minimizes the incremental
Rt Rt
work tnn+1 P : Ḟ dt = tnn+1 [ψ̇ + F · İ]dt. Starting with the given initial condition I(tn ) =
I n , the minimum problem defines an optimal path of the internal variables I(t) for
t ∈ [tn , tn+1 ], including the right boundary value I n+1 := I(tn+1 ). We refer to [17] and
the recent work of [25] for a discussion of extremum paths in linear and nonlinear plasticity.
The two equations (5) and (6) provide an approximative variational counterpart of the
continuous setting (2) and (4) of the constitutive equations in the discrete time step
[tn , tn+1 ] under consideration. In order to show the consistency, we at first recast (6) into
the form
 Z tn+1 
W (F n+1 ) = inf [ψ(F , I)]ttn+1
n
+ φ(İ, I) dt . (7)
I∈G tn

The necessary condition for the minimum problem is that the variation with respect to
the internal variables of the term in brackets vanishes. For smooth functions, integration
by parts yields the expression
Z tn+1 d
[( ∂I ψ + ∂İ φ ) · δI]ttn+1 + [− (∂ φ) + ∂I φ ] · δI dt = 0 . (8)
n
tn dt İ
Thus the variational problem (6) yields Biot’s equation (4)

∂I ψ + ∂İ φ = 0 for t = tn+1 (9)

at the discrete right boundary of the interval [tn , tn+1 ]. The minimizing path of the internal
variables inside the interval is determined by the Euler equation
d
− (∂ φ) + ∂I φ = 0 for t ∈ [tn , tn+1 ] . (10)
dt İ
For the limit tn+1 → tn , the form of the minimization path becomes irrelevant, because the
time increment degenerates to a discrete time tn+1 . Because (9) still holds in this case, it
is shown that the proposed variational formulation (6) represents a consistent point–wise
approximation of Biot’s normal–dissipative evolution equation (4). Furthermore, taking
the derivative of the incremental potential function with respect to the deformation F n+1 ,
we have

∂F W (F n+1 ) = ∂F ψ(F n+1 , I n+1 ) , (11)

where I n+1 is considered to be given by the minimization problem (6). A comparison

with (2) then shows the consistency of the potential equation (5) with the continuous
setting.

6
 Christian Miehe

1. Database {F n+1 , I n } given. Initialize algorithmic parameter γ = 0.

2. Determine current internal state I n+1 = I h (F n+1 , γ) by implicit integration
algorithm of evolution equation

Ah (F n+1 , γ, I h (F n+1 , γ)) = 0

3. Evaluate local minimization function

Z γ m∆t 1/m γ (1+m)/m
h h
W (F n+1 , γ) = ψ(F n+1 , I ) − ψn + c(I h )dγ + η ( )
0 1+m ∆t
and its derivatives W h,F , W h,γ , W ,F
h h h
F , W ,γγ , W ,F γ .
4. Incrementally elastic response: If W,γh (F n+1 , γ = 0) > 0 set P n+1 = W ,F
h
,
h
Cn+1 = W ,F F and exit.
5. Convergence check: If (|W,γh | ≤ tol) go to 7.
h −1
6. Perform Newton update γ ⇐ γ − [ W,γγ ] [ W,γh ] and go to 2.
7. Set micro–stresses and compute micro–moduli
h h h h −1 h
P n+1 = W ,F and Cn+1 = W ,F F − [ W ,F γ ][ W ,γγ ] [ W ,γF ]

Table 2: Discretization of the local constitutive minimization problem (C)

2.3 Numerical implementation for plasticity and viscoplasticity

For an inelastic response with an elastic domain such as in plasticity, viscoplasticity
and damage mechanics, one usually models a convex elastic domain

E := { F |f (F, I) ≤ c(I) } (12)

in the space of the thermodynamical forces F . Here, f (F , I) is a level set function

which is convex an positively homogeneous of the degree one with repect to the forces F .
c(I) ≥ is a threshold function. Then a dissipation function may be obtained by a viscous
regularization of the principle of maximum dissipation
1
φ(İ, I) = sup[ F · İ − (f (F , I) − c(I))+ (1+m) ] (13)
F η(1 + m)

that penaltyzes the overstresses (f − c)+ . Here, (x)+ = 12 ( x + |x| ) denotes the ramp
function. The inverse penalty parameter η ∈ R+ may physically be interpreted as a

7
 Christian Miehe

viscosity of the material and m ∈ R+ is a power exponent. Interpreting the penalty term
in (13) as the dual dissipation function φ∗ (F , I), we obtain the classical normality rule

İ = ∂F φ∗ (F , I) = λ∂F f (F , I) , (14)

where we have introduced per definition the parameter

1
λ := (f − c)+ m ≥ 0 . (15)
η
Insertion of (14) into (13) and exploiting the homogeneity of the function f identifies the
dissipation D = F · I = f λ ≥ 0. Furthermore, after some algebraic manipulations we
obtain the simple representation of the dissipation function
m 1/m (1+m)/m
φ = c(I)λ + η λ (16)
1+m
in terms of the plastic parameter λ introduced above. Observe that this dissipation
functions additively splits up into a rate–independent plastic part and a viscous overstress
part.
The discretization of the variational principle can be based on a convential integration
of the evolution equation (14). Such an update determines implicitely the current internal
state I n+1 = I h (F n+1 , γ), for example by a backward Euler algorithm

I h = I n + γ∂F f (−∂I ψ(F n+1 , I h ), I h ) . (17)

With this notion at hand, the discrete counterpart of the constitutive minimization prob-
lem (6) can be transformed into a minimization problem for the incremental parameter
γ := λn+1 ∆t ≥ 0
W (F n+1 ) = inf W h (F n+1 , γ) (18)
γ≥0

for the scalar function

Z γ m∆t 1/m γ (1+m)/m
W h (F n+1 , γ) = ψ(F n+1 , I h ) − ψn + c(I h )dγ + η ( ) . (19)
0 1+m ∆t
The Karush–Kuhn–Tucker conditions of the constrained optimization problem (18) read

W h,γ ≥ 0 , γ≥0 , W h,γ γ = 0 . (20)

In case of inelastic loading, the incremental parameter γ > 0 is computed at frozen

deformation F n+1 from the condition
ηγ 1/m
W,hγ = −F · I , γ + c + ( ) = 0, (21)
∆t

8
 Christian Miehe

which can be considered as a (negative) pseudo–yield condition. The sensitivity of the

2
internal state with respect to γ is obtained in an implicit form I , γ = ∂F f − γ(∂F Ff ·
2 2 −1
∂II ψ − ∂F I f ) · I ,γ from the algorithm (17), yielding the relationship I ,γ = A · ∂F f
2 2 2
with the fourth–order matrix A := I + γ(∂F F f · ∂II ψ − ∂F I f ). Taking into account
the homogeneity of the level set function f , which induces the relationships F · ∂F f = f ,
2 2
F ·∂F F f = 0 and F ·∂F I f = ∂I f , we may recast (21) into the energetic quasi–consistency
condition
ηγ
W,hγ = −(f + γ∂I f · A−1 · ∂F f ) + c + ( )1/m = 0 . (22)
∆t
Thus the minimization problem (18) in connection with the fully implicit update (17)
for the internal variables yields the standard quasi–consistency condition χ := f − c −
ηγ 1/m
( ∆t ) = 0 of viscoplasticity at time tn+1 only for level set functions in (12) which
depend on the forces F only, i.e. for ∂I f = 0 in (22). This is usually not the case in
finite–strain models of plasticity. Then the variational update (18) cannot be interpreted
as a general return mapping scheme, but should considered to be the canonical symmetry–
preserving algorithmic treatment of associative plasticity. A Newton–type solver including
the representation of the stresses and consistent tangent moduli is summarized in Box 2.
For further details and applications of this type of variational update algorithm to isotropic
and anisotropic models of plasticity we refer to the recent papers [18] and [22].

3 MINIMIZATION PROBLEM FOR DISSIPATIVE SOLIDS

3.1 Incremental variational formulation of standard dissipative solids
The existence of the constitutive minimization problem (6) allows the introduction of an
incremental minimization formulation of the boundary–value problem of finite inelasticity
for standard dissipative solids. Let ϕ : B × R+ → R3 denote the nonlinear deformation
map of an inelastic continuum B ⊂ R3 at a material point X ∈ B and time t ∈ R+ . Then
F := ∇ϕ is the deformation gradient with Jacobian J := det[F ] > 0. Now consider a
functional of the current deformation field ϕn+1 at the right boundary of the increment
[tn , tn+1 ]
Z
I(ϕn+1 ) = W (F n+1 ) dV − [ Πext (ϕn+1 ) − Πext (ϕn ) ] (23)
B

with the global load potential function Πext (ϕ) = B ϕ · γ dV + ∂Bt ϕ · t dA of dead body
R R

forces γ(X, t) in B and surface tractions t(X, t) on ∂Bt . W is the incremental stress
potential function defined in (6). The current deformation map of inelastic standard
dissipative materials can then be determined by a principle of minimum incremental
energy for standard dissipative solids

I(ϕ∗n+1 ) = inf I(ϕn+1 ) (24)

ϕn+1 ∈W

9
 Christian Miehe

subject to ϕn+1 ∈ W := {ϕ ∈ W 1,p (B)|ϕ = ϕ̄(X) on ∂Bϕ } associated with prescribed

deformations ϕ̄ at X ∈ ∂Bϕ on the boundary. The minimization problem has a structure
identical to the principle of minimum potential energy in finite elasticity. The discretiza-
tion of the minimization principle (24) can be performed in a straightforward manner by
a displacement–type finite element method.

3.2 Incremental stability of standard dissipative materials

With the incremental minimization problem (24) at hand, existence results of finite
elasticity as outlined in [1] and [6], can straightforwardly be applied to the incremental
response of standard dissipative materials. The internal part I(ϕn+1 ) := B W (F n+1 ) dV
R

of the functional (23) is sequentially weakly lower semicontinuous (s.w.l.s.), if the incre-
mental stress potential W (F n+1 ) defined in (6) is quasiconvex, see [27] and [6]. This is
considered to be the key property for the existence of sufficiently regular minimizers of
the variational problem (24). Hence, the quasiconvexity
1
Z
W (F n+1 + ∇w(y)) dV ≥ W (F n+1 ) (25)
|D| D

of the incremental stress potential W is considered to be the fundamental criterion for

a local material stability of the incremental response of standard dissipative solids. The
condition states that for all fluctuations w at y ∈ D on an arbitrarily chosen part D of the
inelastic solid with support w = 0 on ∂D the homogeneous deformation given by F n+1
provides an absolute minimizer of the incremental potential. Thus (25) rules out internal
buckling, the development of local fine–scale microstructures and phase decompositions
of a homogeneous local deformation state. Integral condition (25) is difficult to check in
practical computations. However, recall that weak convexity conditions are related via
convexity ⇒ polyconvexity ⇒ quasiconvexity ⇒ rank–one convexity (26)
and that the rank–one convexity condition is a close approximation of the quasiconvexity
condition, see for example [6]. As outlined in [14], the infinitesimal form of the rank–one
convexity condition is the classical Legendre–Hadamard or strong ellipticity condition
(m ⊗ M ) : A : (m ⊗ M ) > 0 with A := ∂F P = ∂F2 F W (F ; X) . (27)
A is the consistent nominal tangent modulus which describes the sensitivity of the nominal
micro–stresses P with respect to the micro–deformation F . Material instabilities of the
constituents of the micro–structure in the form of localizations and phase decompositions
may occur if condition (27) is violated for arbitrary unit vectors m and N .

3.3 Microstructure development of non–stable (non–convex) materials

The variational formulation of the incremental boundary value problem for standard
dissipative solids outlined above opens up the opportunity to apply the concept of relax-
ation to non–convex problems where an energy–minimizing deformation microstructure

10
 Christian Miehe

∂D

a. D b. c.
PSfrag replacements

PSfrag replacements
D

D
∂D

PSfrag replacements

∂D
Figure 1: Interpretation of incremental energetic stability conditions of an inelastic material. A given
homogeneous deformation state F n+1 of the material such as the pure shear mode of Figure a.) is stable
if superimposed fine–scale fluctuation patterns b.) (quasiconvexity) with support on the boundary ∂D or
first–order laminates c.) (rank–one convexity) increase the averaged incremental stress potential on D.

develops. A relaxation is associated with a convexification of a non–convex function W

by constructing its convex envelopes WQ as discussed below, see [12], [13], [6], [29] and
[8] for a sound mathematical basis. The convexification procedure determines the devel-
oping microstructure. We consider the quasi–convexified incremental energy functional of
standard dissipative solids
Z
IQ (ϕn+1 ) = WQ (F n+1 ) dV − [ Πext (ϕn+1 ) − Πext (ϕn ) ] , (28)
B

where the internal part of the relaxed energy functional is obtained by replacing the
non–convex integrand by its quasiconvex envelope. The current deformation field of the
elastic–plastic solid is then determined by the relaxed incremental variational principle

IQ (ϕ∗n+1 ) = inf IQ (ϕn+1 ) (29)

ϕn+1 ∈W

that minimizes the relaxed incremental potential energy. This relaxed problem (29) is
considered to be a well–posed problem as close as possible to the ill–posed problem (24).
In (29), the quasi–convexified incremental stress potential is defined by the minimization
problem of relaxation with respect to the microscopic fluctuation field w

1
Z
WQ (F n+1 ) = inf W (F n+1 + ∇w(y)) dV (30)
w∈W0 |D| D

with y ∈ D subject to the constraint w ∈ W0 := {w ∈ W 1,∞ (D)|w = 0 on ∂D}

providing a support on ∂D. It determines a micro–fluctuation field w for an arbitrarily
chosen domain D ⊂ R3 .

3.4 First–order rank–one convexification analysis

The quasiconvexification procedure (30) can be approximated by a rank–one convex-
ification based on the development of sequential laminates. The rank–one convexified

11
 Christian Miehe

  F n+1 Macro–Level

 


W R1





 

W R2 F+ F− Micro–Level 1









A+ A− B+ B− Micro–Level 2

Figure 2: Rank–one convexification and development of sequential laminates. The rank–one convexifi-
cation WRk (F n+1 ) based on Kohn–Strang’s recursion formula implies the development of a sequential
laminate. Starting from the homogeneous deformation state F n+1 any phase of level k − 1 decomposes
into two phases (+) and (−) of level k. As a consequence, a typical binary tree structure emerges.

envelope is obtained by
WR (F n+1 ) = lim WRk (F n+1 ) (31)
k→∞

with the recursion formula

WRk (F n+1 ) = inf {ξ + WRk−1 (F + ) + ξ − WRk−1 (F − )} (32)

ξ + ,ξ − ,F + ,F −

for the scales k = 1, 2, 3..., starting with WR0 = W (F n+1 ). This procedure implies
that any deformation phase of order k − 1 decomposes into two phases (+) and (−)
of order k, where the difference between the two phases gives a rank–one tensor, i.e.
rank[F + − F − ] ≤ 1. The volume fractions satisfy the condition ξ + + ξ − = 1 and can
be understood to play the role of probability measures in the sense of [38], see [5] and
[29] for further details. Figure 2 shows the developing micro–phases form a sequential
rank–2 laminate. The instable macroscopic deformation state F n+1 decomposes into two
micro–phases F + and F − of micro–level 1 which again split into two pairs of micro–
phases A+ , A− and B + , B − of micro–level 2. The rank–one convexified potential WR2
then consists of the volume average of the stress potentials W at the root of the tree, i.e.
+ + +
WR2 (F n+1 ) = ξ ε [ξ A W (A+ ) + ξ A W (A− )] + ξ ε [ξ B W (B + ) + ξ B W (B − )]. Sequential
− − −

laminate phase decompositions have been investigated by [32] and [33] in the context of
subgrain dislocations structures in single crystals.
Following the works [20] and [21], we approximate the exact rank–one convexification
procedure outlined above by a two–phase analysis that takes into account only the first
micro–level of Figure 2. To this end, we introduce the ansatz

L+ := 1 +(1 − ξ)d M ⊗ N
(
± ±
F := F n+1 L with . (33)
L− := 1 − ξd M ⊗ N

It models a first–order laminate in terms of the two Lagrangian unit vectors M and
N . The scalar d describes the intensity of the bifurcation on the micro–scale. ξ is the

12
 Christian Miehe

1. Database {F n+1 , I + −
n , I n } and starting value q 0 := {ξ, d, N, M }0 given.
2. Set micro–deformation phases

L+ := 1 +(1 − ξ)d M ⊗ N
(
F ± := F n+1 L± with
L− := 1 − ξd M ⊗ N
3. Evaluate minimization function

W̄ h (F n+1 , q) = ξW (F + ) + (1 − ξ)W (F − )
h
and its derivatives W̄,F , W̄,qh , W̄,F
h h h
F , W̄ ,qq , W̄ ,q F̄ .

4. Convergence check: If (k W̄,qh k ≤ tol) go to 6.

h −1
5. Newton update of micro–variables q ⇐ q − [ W̄,qq ] [ W̄,qh ]
6. Set relaxed macro–stresses and tangent macro–moduli

P̄ n+1 = W̄ h,F̄ and C̄n+1 = W̄ h,F̄ F̄ − [ W̄ h,F̄ q ][ W̄ h,qq ]−1 [ W̄ h,qF̄ ]

Table 3: Approximation of minimization problem (R) of relaxation

volume fraction of the phase (+). For the 2D description of the rank–one laminate, the
deformations of the two phases (+) and (−) on the micro–scale are characterized by the
micro–variables
q := {ξ, d, N, M } . (34)
In the 2D case, the unit vectors N and M can be parameterized by two angles ϕ and
χ according to M (ϕ) = [cosϕ sin ϕ]T and N (χ) = [cos χ sin χ]T . Then the rank–one
laminate is determined by four scalar variables q = {ξ, d, ϕ, χ}, which are constrained to
the admissible domain Q := {q | 0 ≤ ξ ≤ 1 , d ≥ 0 , 0 ≤ ϕ ≤ π , 0 ≤ χ ≤ π}. Then the
minimization problem (32) is approximated by

WR1 (F n+1 ) = inf {W̄ h (F n+1 , q)} . (35)

q∈Q

in terms of the function

W̄ h (F n+1 , q) := ξW (F + (F n+1 , q)) + (1 − ξ)W (F − (F n+1 , q)) . (36)

The solution of the minimization problem (34) by a Newton–type algorithm is outlined

in Box 3. The obtained rank–one convexified potential WR1 (F n+1 ) replaces the quasi–

13
 Christian Miehe

convexified potential WQ (F n+1 ) in (28). The overall stresses and consistent tangent mod-
uli are then simply defined by by the quasi–hyperelastic function evaluations

P̄ n+1 = ∂F̄ WR1 (F n+1 ) and C̄n+1 = ∂F̄2 F̄ WR1 (F n+1 ) (37)

The solution of the minimization problem (34) by a Newton–type algorithm and the
representation of the relaxed stresses and tangent moduli is outlined in Box 3. For further
details, we refer to the recent works [15], [20] and [21].

4 INCREMENTAL MINIMIZATION PROBLEM OF HOMOGENIZATION

4.1 Fluctuation fields on heterogeneous microstructures
Let ϕ : B × R+ → R3 denote the deformation map of a heterogeneous micro–
structure B ⊂ R3 at X ∈ B and time t ∈ R+ . The deformation is assumed to be driven
by a prescribed macro–deformation F̄ (t)

ϕ(X, t) = F̄ (t)X + w(X, t) (38)

consisting of a linear part F̄ (t)X and a superimposed fine–scale fluctuation field w :

B × R+ → R3 . Following [11], the macro–deformation F̄ (t) is Rassumed to be governed
1
by surface data of the microscopic deformation field F̄ (t) = |B| ∂B ϕ(X, t) ⊗ n(X) dA.
Then, as a consequence of the decomposition (38), the fluctuation field w has to satisfy
1 R
the constraint |B| ∂B w(X, t) ⊗ n(X) dA. This constraint is satisfied for the three classes
of constraints
1
Z 
C̄D (w; t) := − t · w dA =0 
|B| Z∂B




1


+ + + −
C̄P (w; t ) := − t · (w − w ) dA = 0 (39)
|B| Z∂B+ 

1


C̄S (w; P̄ ) := − (P̄ n) · w dA =0 


|B| ∂B


associated with the surface ∂B of the micro–structure. The first constraint C̄D in (39)
demands homogeneous deformation with zero fluctuations w = 0 on the boundary ∂B
of the micro–structure, where the traction field t(X, t) is considered to be a Lagrangian
multiplier. The second constraint C̄P states a non–trivial periodicity w + = w− of the
superimposed fluctuation field on ∂B. Here, the boundary ∂B is understood to be decom-
posed into two parts ∂B = ∂B − ∪∂B + with outward normals n+ = −n− at two associated
points X − ∈ ∂B − and X + ∈ ∂B + . t+ (X, t) is a Lagrangian multiplier field that charac-
terizes the antiperiodic tractions on ∂B + . The third constraint C̄S in (39) is associated
with homogeneous stresses t = P̄ n on the boundary ∂B, where the Lagrangian multiplier
P̄ (t) is the macro–stress dual to the macro–deformation F̄ (t). The discrete formulation
of the boundary constraints (39) for finite element discretizations of microstructures is
visualized in Figure 3, see [19] for further details.

14
 Christian Miehe

Xq xq δq

∂B F̄

PSfrag replacements B
a.
Aq σAq
Xq xq

∂B F̄

B
b.

X+ x+ πq
q q

∂B F̄

B
c.
X−
q πq x−
q

Figure 3: Partitioning of nodes and deformation–driven boundary constraints of discretized micro–

structures with Lagrangian multipliers δ, σ and π. a.) Prescribed deformation constraint C D =
1
PM
− |B| q=1 δ q · [xq − F̄ X q ] with forces δ q at q = 1 . . . M nodes on boundary. b.) Uniform traction
1 PM
constraint CS = − |B| q=1 (σAq ) · xq + σ : F̄ with average first Piola stress σ. c.) Periodic deformation
1 PP − + −
constraint CP = − |B| q=1 πq · [(x+ q − xq ) − F̄ (X q − X q )] with forces π q at q = 1 . . . P corresponding
node pairs on boundary.

4.2 Incremental variational formulation of homogenization

Based on the constitutive minimization problem for the incremental response of stan-
dard dissipative materials defined in (6) we obtain the current stresses by the quasi–
hyperelastic function evaluation (5). Hence, the micro–stress P n+1 and micro–moduli
Cn+1 at a local point X ∈ B of the heterogeneous micro–structure at time tn+1 are
defined by

P n+1 = ∂F W (F n+1 ; X) and Cn+1 = ∂F2 F W (F n+1 ; X) . (40)

The analogy to finite elasticity induces the following incremental variational formulation
of the homogenization problem for normal–dissipative inelastic solids. As the key homog-
enization condition, we consider the principle of minimium average incremental energy

15
 Christian Miehe

proposed in [18]

1 Z
W̄ (F̄ n+1 ) = inf W (F̄ n+1 + ∇w n+1 ; X) dV (41)
wn+1 |V| B

for heterogeneous micro–structures with constituents of standard dissipative materials.

The minimization problem is understood to be subject to the three alternative constraints
defined in (39). The minimization problem determines the fluctuation field w n+1 on the
micro–structure at discrete time tn+1 . The macro–stress potential W̄ is assumed to define
the macro–stresses P̄ n+1 and the macro–moduli C̄n+1 by the quasi–hyperelastic function
evaluations

P̄ n+1 = ∂F̄ W̄ (F̄ n+1 ) and C̄n+1 = ∂F̄2 F̄ W̄ (F̄ n+1 ) (42)

in complete analogy to (40). Thus the minimization problem (41) provides a shift of
associated variables from the micro–scale to the macro–scale, often denoted as micro–
to–macro transition. It extends the so–called average variational principle of nonlinear
elasticity, outlined by [28] and [35], to the incremental formulation of finite inelasticity.
The attractive feature of the proposed formulation (41) is that the structure of the average
variational principle is preserved, i.e. the energy storage function ψ of finite elasticity is
replaced by the incremental stress potential W of finite inelasticity. A finite element
discretization of the minimization problem (41) of homogenization is outlined in Box 4.
For further details of the formulation and numerical implementation of homogenization
methods for dissipative materials based on minimization principles we refer to the recent
papers [18], [19] and [22].

4.3 Stability problems of dissipative heterogeneous microstructures

An important problem in the geometrically nonlinear homogenization of heterogeneous
microstructures is the occurance of stability problems. With the homogenized incremental
macro–stress potential W̄ defined in (41) at hand, these stability problem can be defined
as follows. Structural stability of the micro–structure like buckling of fibers may occur if
the infinitesimal structural stability condition
1
Z
W̄ (F̄ , ϕ) − W̄ (F̄ , ϕ + δϕ) = ∇δϕ : A : ∇δϕ dV > 0 (43)
2|V| B

is violated. The condition states an increase of averaged incremental energy for an in-
finitesimal perturbation δϕ of an equlibrium state ϕ of the micro–structure. It is satisfied
for positive definite tangent moduli

∇δϕ : A : ∇δϕ > 0 (44)

in the whole domain of the micro–structure.

16
 Christian Miehe

1. Given Database {F̄ , dn }. Initialize nodal displacements dn+1 = dn .

2. Determine gradient of current fluctuation field w hn+1 ∈ W by the finite element
ansatz
∇whn+1 (X) = B e (X)den+1 in B e ⊂ B

3. Evaluate minimization function

1 ne Z
h
W̄ (F̄ n+1 , dn+1 ) = A W (F̄ n+1 + B e den+1 ; X)dV
|V| e=1 Be

and its derivatives W̄ h,F̄ , W̄ h,d , W̄ h,F̄ F̄ , W̄ h,dd and W̄ h,dF̄ .

h 2 1/2
4. Convergence check: If ([ α∈A W̄,d ] ≤ tol) go to 6.
P

5. Perform Newton update of nodal displacements dn+1 ⇐ dn+1 + ∆dn+1 with

h −1
∆dn+1 := −[ W̄,dd ] [ W̄,dh ]
6. Set nominal macro–stresses and macro–moduli

P̄ n+1 = W̄ h,F̄ and C̄n+1 = W̄ h,F̄ F̄ − [ W̄ h,F̄ d ][ W̄ h,dd ]−1 [ W̄ h,dF̄ ]

Table 4: Discretization of minimization problem (H) of homogenization

The material stability of the homogenized macroscopic response can be defined in

terms of the homogenized incremental energy function W̄ in analogy to the microscopic
formulations of Section 2.8.1. A necessary condition for a sequentially weakly lower semi-
continuous (s.w.l.s.) macroscopic energy functional is the quasi–convexity condition
1
Z
W̄ (F̄ + ∇w̄(ȳ))dV > W̄ (F̄ ) (45)
|D̄| D̄

for ȳ ∈ D̄ and w̄ = 0 on ∂ D̄, where D̄ ⊂ R3 is an arbitrarily chosen part of the homog-

enized material. The condition states that for all fluctuations w̄ on D̄ with support on
∂ D̄ a homogeneous deformation of the homogenized material by F̄ provides an absolute
minimizer of the macroscopic incremental energy in D̄. Close to (45) is the infinitesi-
mal form of the macroscopic rank–one convexity condition, i.e. the classical macroscopic
Legendre–Hadamard or strong ellipticity condition
(m̄ ⊗ N̄ ) : Ā : (m̄ ⊗ N̄ ) > 0 with Ā := ∂F̄ P̄ = ∂F̄2 F̄ W̄ (F̄ ) (46)
for arbitrary unit vectors m̄ and N̄ . Ā is the macroscopic tangent modulus which
describes the sensitivity of the nominal macro–stresses P̄ with respect to changes of
the macro–deformation F̄ .

17
 Christian Miehe

As pointed out in [28] and [9] for periodic heterogeneous micro–structures in finite elas-
ticity, microscopic structural instability phenomena like the buckling of fibers may induce
a macroscopic instability phenomenon in the form of a discontinuity–type localization if
the buckling modes δϕ in (43) have the so–called shear deformation form

∇δw = m̄ ⊗ M̄ + ∇q(X) (47)

modulated by a periodic contribution q + = q − on ∂V. However, the occurence of struc-

tural instabilities on the micro–structure for violations of conditions (43) and (44) raise
the fundamental problem that the size of the unit cell cannot a priori be defined based
on the periodicity of the material as depicted in Figure 2a. The cell must be enlarged
such that it catches the current energy–minimizing buckling modes. As outlined in [28],
[9] and [23] for problems of finite elasticity, this induces a full–space analysis based on an
ensemble of nc unit cells
nc
[
E(nc ) := Vk with k = 1, 8, 27, 64, ... (48)
k=1

The homogenized incremental energy is then defined by an extension of (41)

 
W̄ (F̄ ) = inf inf W̄ (ϕ; nc ) , (49)
nc ϕ

which minimizes the ensemble–averaged incremental energy also with respect to the size
of the RVE or the ensemble of unit cells. The solution of (49) then catches the currently
energy–minimizing buckling mode. A computational investigation of this property of
non–convex homogenization is considered in [23] for problems of finite elasticity.

5 NUMERICAL MODEL INVESTIGATIONS OF MICRO–STRUCTURES

5.1 Micro–structure development in a homogeneous crystalline strip
We demonstrate the performance of the one–level rank–one relaxation technique sum-
marized in Box 3 by means of a numerical example that analyzes the deformation of a
crystalline strip in tension under plane strain conditions. The constitutive response of
the strip is assumed to governed by a model of single slip crystal plasticity. For details
of the constitutive formulation and its numerical implementation by the above outlined
minimization principles, we refer to [21]. The main goals of the numerical investigations
are the analysis of the developing microstructures and the demonstration of the mesh–
invariance of the relaxation technique proposed. The boundary conditions are such that
the specimen is able to perform unconstrained displacements perpendicular to the cross
section. The specimen is discretized with 6 × 12, 8 × 16, 12 × 24, 16 × 32 and 20 × 40
enhanced strain finite elements. We treat the problem in a deformation–driven analy-
sis with increments ∆u = 0.05 mm. For the finest mesh the increments are reduced to

18
 2 2

P
Poten

Poten
PSfrag replacements4 PSfrag replacements
4

Stress

Stress
6 6
8 8
10 10
F−12 12
F−
F+14 14
F+
WR1 (F n+116
) WR1 (F n+1
16 )
0.0
W (F n+1 ) W (F0.0
n+1 )
0.2
F n+1 F0.2
n+1
0.4 Christian0.4Miehe
Intensity β [−]0.6 Intensity β0.6[−]
0.8 0.8
1.0 1.0 a]
τ11 [GP

Stress P m · N [N/mm2 ]
1.2 200 1.2 1500
Potential W [N/mm2 ]

τ22 [GP a]
1.4 180 1.4
1.6 1.6 0 1000
1.8 160 1.8 2
2 140 2 4 500
F n+1
1 1 6
F n+1 0
0 120 0 8
F+ F−
−1 100
−110
-500
−2 −212
−3 80 F− F+ −3 14
-1000
−4 60
−416
−5 −5 -1500
−6 -0.4 -0.2 0 0.2 0.4 −6 -0.4 -0.2 0 0.2 0.4
−7 a. −7 b.
Intensity β [−] Intensity β [−]

16
16000 2
2000
14
14000 1
1000

12
12000 0
τ22 [GP a] −1
-1000
τ11 [GP a]

10
10000
2 −2
-2000
8000
8
1 −3
-3000
0 6000
6 Relaxed −4
-4000 Relaxed
−1
4
4000 −5
-5000
−2
−3 2
2000 −6
-6000
−4 0 −7
-7000
−5 0 0.2
0.0 0.4 0.6
0.2 0.4 0.6 0.8
0.8 1.0
1 1.2
1.2 1.4 1.6 1.8
1.4 1.6 1.8 0 0.2
0.0 0.2 0.4
0.4 0.6
0.6 0.8
0.8 1.0
1 1.2
1.2 1.4
1.4 1.6 1.8
1.6 1.8
−6
−7 c. Loading d. Loading
Figure 4: Crystalline strip in tension. Details of the rank–one convexification for the orientation angle
α = 90o of the slip–system at Λn+1 = 0.075. β parametrizes the intensity of the laminate F ± =
F n+1 + β ± m ⊗ N . a.) At F n+1 the potential is not rank–one convex (dashed line). F n+1 decomposes
into micro–phases F ± (solid line). b.) The relaxed stress-strain relation characterizes a snap–through
Maxwell–line behavior between the micro–phases F ± . c.) The shape of the relaxed and non–relaxed
Kirchhoff stress components τ11 d.) and τ22 .

∆u = 0.025 mm. In order to provoke a loss of material stability and a phase–decay of

the homogeneous deformation into first–order laminates F ± , we choose a counterclock-
wise orientation α = 10◦ of the slip system with respect to the horizontal. Figure 4a
shows the shape of a non–convex incremental stress potential evaluated at an integration
point of the finite element mesh. The variable β parameterizes the intensity of the lam-
inate F ± = F n+1 + β ± m ⊗ N . Obviously, the incremental stress potential W (F n+1 )
is greater than the interpolation of the potentials W (F + ) and W (F − ) corresponding to
the micro–phases (+) and (−). As a consequence, the homogeneous deformation state is
not stable and decomposes into the micro–deformations F ± which minimize the function
W̄ h with respect to the variables q. The relaxed stress–deformation relation plotted in
Figure 4b characterizes a snap–through behavior (Maxwell–type line) between the two
micro–phases F ± due to the constant slope of the rank–one convex envelope. The arising
microstructures are resolved by the rank–one convexification of the non–convex potential.
The in–plane Kirchhoff stress coordinates of the relaxed and the non–relaxed response

19
 Christian Miehe

a.

b.

PSfrag replacements

c.
Figure 5: Crystalline strip in tension. Comparison of different finite element meshes for relaxed and
non–relaxed analyses at u = 30mm. a.) deformed meshes for non–relaxed analysis, b.) deformed meshes
with interface–directions, c.) distribution of instable regions where microstructures arise.

are given in Figures 4c and 4d. The convexified stress–loading curves for τ11 and τ22 show
a non–linear snap–through behavior within the non–convex range. After the recovery of
the stable homogeneous state the relaxed and the non–relaxed stress responses coincide
again.
In Figure 5a the deformed meshes of the non-relaxed analysis are plotted for the five
mesh–discretizations considered. The blocking of the slip–system leads to a stiffer response
of the non–convex formulation, documented by partially strong distortions of the finite
element meshes. In particular the flexibility of the deformation depends on the mesh
discretization applied. Figure 5b reports on the deformed meshes and the orientation of

20
 Christian Miehe

30000 30000

25000 25000
Load F [N]

Load F [N]
20000 20000

15000 15000
PSfrag replacements non–relaxedPSfrag replacements relaxed
10000 solutions 10000 solutions
relaxed 5000 5000

0 0
0 5 10 15 20 25non–relaxed
30 0 5 10 15 20 25 30
a. Vertical Top Displacement u [mm] b. Vertical Top Displacement u [mm]

Figure 6: Crystalline strip in tension. Load–displacement curves for five different finite–element meshes
in terms of a.) the non–relaxed (non–objective) formulation and b.) the proposed relaxation technique.

the directions of the laminate–interfaces which result from the relaxation analysis. The
development of the first–order laminates smoothes out the stress response and leads to
more flexible and less distorted finite element meshes. The instable regions where the first–
order laminates develop are given in Figures 5c. The mesh–dependent response of the non–
relaxed formulation is also evident by considering the load–displacement curves plotted
in Figure 6a. In contrast to the non–relaxed formulation, application of the proposed
relaxation technique yields a mesh–invariant response. The load–deflection curves do not
depend on the mesh–size, but are identical for all different mesh–densities, see Figure 6b.
It turns out that the resolution of the microstructure as a first–order laminate is sufficient
with regard to an objective stress response. Note that the objectivity of the material
behavior is obtained without introduction of an internal length scale parameter. In Figure
7 the microstructures at the central integration points of the red–marked elements are
magnified for the 6 × 12 and the 20 × 40 finite element meshes at u = 16 mm. For both
mesh–densities the arising microstructures are identical. This confirms the accuracy of
the first–order rank–one convexification of the non–convex stress potential.

5.2 Micro–structure development in heterogeneous polycrystals

We demonstrate the performance of the homogenization technique summarized in Box
4 by means of a texture analysis for a polycrstalline aggregate, as conceptually used by
micro–to–macro transitions as visualized in Figure 8. The example demonstrates the
performance of the three types of micro–to–macro transitions when applied to a 3d–
simulation of the overall response in polycrystals. The micro–structure considered is a
copper polycrystal discretized using 73 = 343 mixed finite elements. Individual crystal
orientations, determined by a random distribution of quaternions as suggested in [24],
are assigned to each integration point in order to simulate a polycrystalline aggregate of
2744 f.c.c. single crystal grains. For a detailed representation and numerical implementa-
tion of single crystal plasticity models for the grains and the material parameters chosen,

21
 Christian Miehe

a.

PSfrag replacements
b.

Figure 7: Crystalline strip in tension. Visualization of microstructures at selected integration points

a.) for 6x12 and b.) for 20x40 mesh discretizations at u = 16mm.

we refer to [24] and [22] and the works cited therein. We consider a plane strain com-
pression test of the polycrystalline micro–structure where the macroscopic deformation
gradient is prescribed by F̄ = [F̄11 ; F̄22 ; F̄33 ; ...]T = [1; exp(vt); exp(−vt); 0; 0; 0; 0; 0; 0]T
with v = 1.0s−1 . During the simulation the time is linearly increased in an interval
t ∈ [0, 1.54s] in equal time steps ∆t = 2.5 · 10−4 s. The final state at time t = 1.54s
corresponds with a compression of the micro–structure to 21% of its initial height. The
plot sequences in Figure 9 describe the deformed geometry for three different constraints
when the three types of micro–to–macro transitions are applied. The RVE–type uniform
traction condition on the boundary represents the softest response of the polycrystalline
aggregate, yielding the heavily distorted finite element meshes in Figure 9a. The texture
development in the specimen is reported for the three types of micro–to–macro transitions

22
 Christian Miehe

P̄3
PSfrag replacements
P̄1

micro-structure pole figure

P̄2

F F

Figure 8: Necking of a polycrystalline bar with an attached heterogeneous micro–structure to model a

polycrystalline aggregate. The texture plot visualizes the developed deformation micro–structure.

in Figure 10. We plotted the h111i–pole figures with an equal–area projection. The re-
sults of the simulations are in a good qualitative agreement with the textures observed in
experiments performed by [3]. The RVE–type linear displacement condition in Figure 10b
yields the highest contrast and therefore represents in this sense an upper bound of the
pole figures. With increasing relaxation of the deformation constraint on the boundary
the textures become more smeared. A lower bound is provided by the RVE–type uniform
stress condition yielding the pole figures in Figure 10a. Notice that during the final stage
of deformation the texture becomes increasingly diffuse due to the heavily distorted fi-
nite element mesh depicted in Figure 9a. The unit–cell–type periodic solution Figure 10c
bounds the contrast of the pole figures in Figure 10a–b and is considered to be close to
the experimental results in Figure 10d.

a.

b.

c.
Figure 9: Plane strain compression test of polycrystalline aggregate for different micro–to–macro transi-
tions. Deformed micro–structures after a reduction of 18%, 41%, 63% and 79%. a.) Uniform tractions,
b.) linear displacements, c.) periodic displacements and antiperiodic tractions on the boundary. Micro–
structure stiffness (c) is bounded by (a) and (b).

6 CONCLUSION
The formulation of minimization principles for dissipative standard media and their
exploitation with regard to the analysis of deformation microstructures in homogeneous

23
 Christian Miehe

a.

b.

c.

d.
Figure 10: Plane strain compression test. Sequences of h111i–pole figures in the 12–plane for different
micro–to–macro transitions corresponding to the deformation states of Fig. 9. a.) Uniform tractions,
b.) linear displacements, c.) periodic displacements and antiperiodic tractions on the boundary, d.)
experimental results from [3]. Pole figure contrast (c) is bounded by (a) and (b).

and heterogeneous materials was overviewed. A constitutive minimization problem (C)

for dissipative materials in an incremental setting allows the definition of incremental
stability (S), relaxation (R) of non–convex homogeneous materials and homogenization
(H) of heterogeneous materials including energetic definitions of structural instabilities.
As summarized in Box 1, these minimization principles provide a canonical guide to
literally all relevant ingredients of the theroretical formulation and numerical implemen-
tation of inelastic solid materials for quasistatic mechanical problems. Boxes 2, 3 and 4
show interesting formal analogies of the Newton–type algorithmic settings of problems
(C), (R) and (H), respectively, which are exclusively governed by the micro– and macro–
potentials and their first and second derivatives. The minimization structure provides a
fundamental canonical approach to inelastic solid mechanics under quasistatic conditions,
with important consequences to almost all aspects of the modelling and the numerical
implementation.
Acknowledgement. Thanks go to Matthias Lambrecht, Ercan Gürses and Jan
Schotte for their support and many helpful discussions.

24
 Christian Miehe

REFERENCES
[1] Ball, J.M. [1977]: “Convexity Conditions and Existence Theorems in Nonlinear
Elasticity”, Archive of Rational Mechanics and Analysis, Vol. 63, 337–403.

[2] Biot, M.A. [1965]: “Mechanics of Incremental Deformations”, John Wiley & Sons,
Inc., New York.

[3] Bronkhorst, C.A.; Kalidindi, S.R.; Anand, L. [1992]: “Polycrystalline Plas-

ticity and the Evolution of Crystallographic Texture in f.c.c. Metals”, Phil. Trans.
R. Soc. Lond. A, Vol. 341, pp. 443–477.

[4] Carstensen, C.; Hackl, K.; Mielke, A. [2002]: “Nonconvex Potentials and Mi-
crostructures in Finite–Strain Plasticity”, Proceedings of the Royal Society London,
Series A, Vol. 458, 299–317.

[5] Carstensen, C.; Roubíček, T. [2000]: “Numerical Approximation of Young Mea-

sures in Non–Convex Variational Problems”, Numerische Mathematik, Vol. 84, 395–
415.

[6] Dacorogna, B. [1989]: “Direct Methods in the Calculus of Variations”, Springer–

Verlag Berlin Heidelberg.

[7] DeSimone, A.; Dolzmann, G. [2000]: “Material Instabilities in Nematic Elas-

tomers”, Physica D, Vol. 136, 175–191.

[8] Dolzmann, G. [2003]: “Variational Methods for Crystalline Microstructure – Anal-

ysis and Computation”, Springer–Verlag Berlin Heidelberg.

[9] Geymonat, G.; Müller, S.; Triantafyllidis, N. [1993]: “Homogenization

of Nonlinearly Elastic Materials, Microscopic Bifurcation and Macroscopic Loss of
Rank–One Convexity”, Archive of Rational Mechanics and Analysis, Vol. 122, 231–
290.

[10] Halphen, B.; Nguyen, Q.S. [1975]: “Sur les Matéraux Standards Généralisés”,
Journal de MécaniqueVol. 40, 39–63.

[11] Hill, R. [1972]: “On Constitutive Macro–Variables for Heterogeneous Solids at

Finite Strain”, Proceedings of the Royal Society London, Series A, Vol. 326, 131–147.

[12] Kohn, R.V.; Strang, G. [1983]: “Explicit Relaxation of a Variational Problem in

Optimal Design”, Bulletin of the American Mathematical Society, Vol. 9, 211–214.

[13] Kohn, R.V.; Strang, G. [1986]: “Optimal Design and Relaxation of Variational
Problems I, II, III”, Communications on Pure and Applied Mathematics, Vol. 39,
113–137, 139–182, 353–377.

25
 Christian Miehe

[14] Krawietz, A. [1986]: “Materialtheorie. Mathematische Beschreibung des phäno-

menologischen thermomechanischen Verhaltens”, Springer–Verlag, Berlin.

[15] Lambrecht, M.; Miehe, C.; Dettmar, J. [2003]: “Energy Relaxation of Non–
Convex Incremental Stress Potentials in a Strain–Softening Elastic–Plastic Bar”,
International Journal of Solids and Structures, Vol. 40, 1369–1391.

[16] Luskin, M. [1996]: “On the Computation of Crystalline Microstructure”, Acta Nu-
merica, Vol. 36, 191–257.

[17] Martin, J.B. [1975]: “Plasticity. Fundamentals and General Results”, MIT press,
Cambridge, Massachusetts.

[18] Miehe, C. [2002]: “Strain–Driven Homogenization of Inelastic Microstructures and

Composites Based on an Incremental Variational Formulation”, International Jour-
nal for Numerical Methods in Engineering, Vol. 55, 1285–1322.

[19] Miehe, C. [2003]: “Computational Micro–To–Macro Transitions for Discretized

Micro–Structures of Heterogeneous Materials at Finite Strains Based on the Mini-
mization of Averaged Incremental Energy”, Computer Methods in Applied Mechanics
and Engineering, Vol. 192, 559–591.

[20] Miehe, C.; Lambrecht, M. [2003]: “A Two–Scale Finite Element Relaxation

Analysis of Shear Bands in Non–Convex Inelastic Solids: Small–Strain Theory for
Standard Dissipative Materials”, Computer Methods in Applied Mechanics and En-
gineering, Vol. 192, 473–508.

[21] Miehe, C.; Lambrecht, M.; Gürses, E. [2002]: “Analysis of Material Instabil-
ities of Inelastic Solids Based on Incremental Variational and Relaxation Methods.
Application to Single Slip Crystal Plasticity”, Computer Methods in Applied Me-
chanics and Engineering, submitted.

[22] Miehe, C.; Schotte, J.; Lambrecht, M. [2002]: “Homogenization of Inelas-

tic Solid Materials at Finite Strains Based on Incremental Minimization Principles.
Application to the Texture Analysis of Polycrystals”, Journal of the Mechanics and
Physics of Solids, Vol. 50, 2123–2167.

[23] Miehe, C.; Schröder, J.; Becker, M. [2002]: “Computational Homogenization

Analysis in Finite Elasticity. Material and Structural Instabilities on the Micro– and
Macro–Scales of Periodic Composites and their Interaction”, Computer Methods in
Applied Mechanics and Engineering, Vol. 191, 4971–5005.

[24] Miehe, C.; Schröder, J.; Schotte, J. [1999]: “Computational Homogenization

Analysis in Finite Plasticity. Simulation of Texture Development in Polycrystalline

26
 Christian Miehe

Materials”, Computer Methods in Applied Mechanics and Engineering, Vol. 171, 387–
418.

[25] Mielke, A. [2002]: “Finite Elastoplasticity, Lie Groups and Geodesics on SL(d)”,
in Geometry, Dynamics, and Mechanics, P. Newton, A. Weinstein, P.J. Holmes (ed-
itors), Springer Verlag, 61–90.

[26] Moreau, J.J. [1976]: “Application of Convex Analysis to the Treatment of Elasto-
plastic Systems”, in Application of Methods of Functional Analysis to Problems in
Mechanics, P. Germain, B. Nayroles (editors), Springer Verlag.

[27] Morrey, C.B. [1952]: “Quasi–Convexity and the Lower Semicontinuity of Multiple
Integrals”, Pacific Journal of Mathematics, Vol. 2, 25–53.

[28] Müller, S. [1987]: “Homogenization of Nonconvex Integral Functionals and Cellu-

lar Elastic Materials”, Archive of Rational Mechanics and Analysis, Vol. 99, 189–212.

[29] Müller, S. [1998]: “Variational Models for Microstructure and Phase Transisi-
tions”, Springer Lecture Notes in Mathematics, in press.

[30] Nemat–Nasser, S.; Hori, M. [1993]: “Micromechanis: Overall Properties of Het-

erogeneous Materials”, North–Holland Series in: Applied Mathematics and Mechan-
ics, Vol. 37.

[31] Nguyen, Q.S. [2000]: “Stability and Nonlinear Solid Mechanics”, John Wiley &
Sons, LTD, Chichester.

[32] Ortiz, M.; Repetto, E.A. [1999]: “Nonconvex Energy Minimization and Dislo-
cation Structures in Ductile Single Crystals”, Journal of the Mechanics and Physics
of Solids, Vol. 47, 397–462.

[33] Ortiz, M.; Repetto, E.A.; Stainier, L. [2000]: “A Theory of Subgrain Disloca-
tion Structures”, Journal of the Mechanics and Physics of Solids, Vol. 48, 2077–2114.

[34] Ortiz, M.; Stainier, L. [1999]: “The Variational Formulation of Viscoplastic

Constitutive Updates”, Computer Methods in Applied Mechanics and Engineering,
Vol. 171, 419–444.

[35] Ponte Castañeda, P.; Suquet, P. [1998]: “Nonlinear Composites”, Advances

in Applied Mechanics, Vol. 34, 171–303.

[36] Šilhavý, M. [1997]: “The Mechanics and Thermodynamics of Continuous Media”,

Springer–Verlag, Berlin, Heidelberg, New York.

27
 Christian Miehe

[37] Suquet, P.M. [1987]: “Elements of Homogenization for Inelastic Solid Mechanics”,
in: Homogenization Techniques for Composite Materials. E. Sanchez–Palenzia, A.
Zaoui (editors), Lecture Notes in Physics, Vol. 272, 193–278.

[38] Young, L.C. [1969]: “Lectures on the Calculus of Variations and Optimal Control
Theory”, Saunders, London.

[39] Ziegler, H. [1963]: “Some Extremum Principles in Irreversible Thermodynamics

with Application to Continuum Mechanics”, in Progress in Solid Mechanics, Vol. IV,
I.N. Sneddon, R. Hill (editors), North–Holland Publishing Company, Amsterdam.

28