Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
This research was carried out under the project number MC2.03158 in the framework of
the Research Program of the Materials innovation institute M2i (www.m2i.nl), the former
Netherlands Institute for Metals Research.
P ROEFSCHRIFT
ter verkrijging van de graad van doctor
aan de Technische Universiteit Eindhoven,
op gezag van de rector magnificus, prof.dr.ir. C.J. van Duijn,
voor een commissie aangewezen door het College voor Promoties
in het openbaar te verdedigen
op donderdag 20 oktober 2011 om 16.00 uur
door
Tuncay Yalcnkaya
Contents
Summary
ix
1 Introduction
1.1 Crystal plasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Objective and outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
1
3
2 A finite strain BCC single crystal plasticity model and its experimental
identification
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Slip mechanisms in BCC metals . . . . . . . . . . . . . . . . . . . . . . .
2.3 Violation of Schmids law in BCC metals . . . . . . . . . . . . . . . . . .
2.4 A BCC crystal plasticity model at material point level . . . . . . . . . .
2.4.1 Kinematics in crystal plasticity . . . . . . . . . . . . . . . . . . .
2.4.2 Constitutive model . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5 Modeling some intrinsic properties of BCC single crystals . . . . . . . .
2.5.1 Orientation dependence . . . . . . . . . . . . . . . . . . . . . . .
2.5.2 Example: -Fe single crystal . . . . . . . . . . . . . . . . . . . .
2.5.3 Example: molybdenum single crystal . . . . . . . . . . . . . . .
2.5.4 Temperature dependence . . . . . . . . . . . . . . . . . . . . . .
2.6 Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . .
5
5
8
10
12
12
13
17
17
18
19
19
22
25
25
28
31
34
34
36
37
vi
Contents
3.5
3.6
Numerical examples . . . . . . . . . . . . . . . . . . . . . . .
3.5.1 Example 1: monotonic deformation of single crystals
3.5.2 Example 2: strain path change of single crystals . . .
3.5.3 Example 3: strain path change of polycrystals . . . .
Summary and Conclusion . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
38
38
39
40
43
45
45
49
50
54
54
55
57
58
59
60
66
68
68
68
71
5 Non-convex rate dependent strain gradient crystal plasticity and deformation patterning
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Strain gradient crystal plasticity and finite element implementation . .
5.3 Latent hardening based non-convex plastic potential . . . . . . . . . . .
5.3.1 Conditions for plastic slip patterning . . . . . . . . . . . . . . .
5.4 Numerical analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.1 Convex strain gradient crystal plasticity . . . . . . . . . . . . . .
5.4.2 Non-convex strain gradient crystal plasticity . . . . . . . . . . .
5.5 Summary and conclusion . . . . . . . . . . . . . . . . . . . . . . . . . .
73
73
75
80
80
84
84
88
92
95
Bibliography
99
Contents
vii
Dankwoord / Acknowledgements
109
Curriculum Vitae
111
viii
Summary
bands, etc.) exist as well. These intergranular and intragranular deformation patterns are stated to be inherent minimizers
of the free energy (including the microstructurally trapped plastic energy). These
microstructures may macroscopically manifest themselves through softening of the
material or through plastic anisotropy in hardening under strain path changes. These
effects are crucial with respect to the mechanics of the materials under consideration
and should be taken into account in the constitutive modeling.
In this thesis, the computational modeling of microstructure evolution (with softening or plastic anisotropy) is covered in different crystal plasticity frameworks. The
scope is basically two-fold. First, in the chapters 2 and 3 the plastic anisotropy of
Body Centered Cubic crystals is studied from the onset of deformation due to an
intrinsic orientation dependence from non-planar dislocation core structures, to the
anisotropy upon a strain path change owing to resulting dislocation cell formation.
In this part of the thesis, after developing a proper BCC crystal plasticity framework
taking into account the intrinsic anisotropy, a composite cell model was established
where the evolution of dislocation cells was modeled under monotonic and nonproportional loading histories. Here, the existence of a dislocation microstructure
is introduced into the model in terms of internal variables and the evolution was
described by phenomenologically based evolution equations. However, this phenomenological approach is not able to incorporate the formation stage of the microstructure. Hence, the crystal plasticity framework called for an extension in order
to capture the evolution of the microstructure driven by the free energy of the mateix
Summary
rial.
In order to complete the missing link between the formation of the microstructure and its evolution in crystal plasticity frameworks, the second part of the thesis concentrates on the development of a non-convex rate dependent crystal plasticity model, which reveals a rate dependent dislocation microstructure formation
and evolution together with macroscopic hardening-softening-plateau stress-strain
responses. To this end, non-convexity is treated as an intrinsic property of the plastic
free energy of the material. First, this non-convex contribution is incorporated into a
strain gradient crystal plasticity framework with a double-well character, which results in a computational routine partially dual to the Ginzburg-Landau type of phase
field modeling approaches (with high and low slipped regions representing the different phases). In this model, both the displacement and the plastic slip fields are
considered as primary variables. These fields are determined on a global level by
solving simultaneously the linear momentum balance and the slip evolution equation which is rederived in a thermodynamically consistent manner. In chapter 4, the
analysis is conducted in a 1D mathematical setting in order to illustrate the ability of
the model to capture the patterning of plastic slip. In chapter 5 and inspired by the
literature, the non-convexity originates from latent hardening in a multi-slip strain
gradient crystal plasticity framework. Hence, the 1D approach pursued in chapter 4
is extended to a 2D plane strain setting. Even though the phenomenological doublewell free energy function used in the 1D approach allows to track non-equilibrium
states during microstructure evolution, it does not rely on a physically based expression for non-convexity, but presents a generic formulation. Instead, chapter 5
concentrates more on the physical reasons of plastic slip localization, where a slip
interaction potential is analyzed and incorporated into the rate dependent strain gradient crystal plasticity framework. The non-convexity due to the slip interactions is
explicitly illustrated and the possibility of deformation patterning in the material is
discussed in a boundary value problem. The last part of the thesis, chapter 6, presents
a discussion and conclusions.
Chapter one
Introduction
Abstract / The physics and the basic principles behind the general crystal plasticity modeling are explained. An overview of strain path change related anisotropy and dislocation
microstructure evolution in crystal plasticity approaches is presented. The objectives and
outline of the thesis are given.
1 Introduction
DISLOCATION CELLS
GRAINS
Figure 1.1 / Crystal plasticity bridges the deformation of bulk material and the
movement of dislocations (chapter 2). Clustering of dislocations is accomplished
via non-convex strain gradient crystal plasticity (chapter 4 and 5). The phenomenological evolution of dislocation structures is simulated via a dislocation cell model
(chapter 3).
agating process, a part of the crystal gradually slips one interatomic distance with
respect to the other part. Instead of the ideal fictitious strength associated with the
movement of an entire slip plane, the dislocations enable only sections of the slip
plane to shear, resulting in the observed decimated strengths necessary for plastic
deformation.
The stress directly affecting the motion of dislocations is the projected shear stress
on the specific slip systems, which is also called the resolved Schmid stress. When
the resolved shear stress is larger than the resistance on the respective slip system,
glide is activated. The viscous type of crystal plasticity theories as used in this thesis
employs a power law relation between the rate of plastic slip and the ratio between
the shear stress and the slip resistance. Conceptually, all the slip systems are active
but the most favorable ones carry most plastic deformation. Crystal plasticity is a
mesoscopic modeling approach, bridging the stresses on the slip systems to the total
amount of plastic slip in the bulk material (see Fig. 1.1), eventually affecting the total
amount of macroscopic plastic strain.
In addition to their role of accommodating the plastic deformation in metals, the
work (strain) hardening behavior can also be attributed to the dislocations. The
distinct stages of strain hardening are related to their multiplication or mutual in-
1 Introduction
crystal plasticity framework to describe the dislocation cell structure evolution and
its macroscopic anisotropic effects. The computational framework departs from a
composite aggregate with a cell structure, consisting of a soft cell interior component
and hard cell wall components. The constitutive response of each component has
been obtained from crystal plasticity simulations, while a set of phenomenological
evolution equations for the cell size, the wall thickness and the dislocation density
captures the evolution of the microstructure under complex strain paths. Numerical
examples study both the intrinsic orientation dependence and the anisotropy due to
cell structure evolution.
Chapter four focuses on one of the origins of self-organizing dislocation structures
(driven by the deformation) rather than imposing the cell structure evolution as done
in the previous chapter. To this purpose, a rate dependent strain gradient plasticity
framework for the description of plastic slip patterning in a system with non-convex
energetic hardening is developed. Both the displacement field and the plastic slip
field are considered as primary variables. The slip law differs from classical ones in
the sense that it includes a stress term originating from a non-convex double-well free
energy, which enables patterning of the deformation field. The derivations and implementations are performed in a single slip 1D setting, which allows for a thorough
mechanistic understanding, not excluding its extension to multidimensional cases.
The numerical examples illustrate both the homogeneous and inhomogeneous plastic slip distributions as well as the stress-strain response in relation to the imposed
boundary conditions and the applied rate of deformation.
In chapter five the non-convex strain gradient crystal plasticity formulation is extended to the 2D plane strain case, including multiple slip systems. In order to
capture the effect of dislocation interactions on the non-convexity of the plastic slip
dependent free energy function, a more physically based free energy expression is
incorporated. Attention is focused on the inhomogeneous plastic slip distribution
and deformation patterning due to dislocation slip interactions.
The thesis concludes with a final chapter, summarizing the main achievements and
results obtained as well as an outlook to open challenges.
Chapter two
Abstract / A crystal plasticity model for body-centered-cubic (BCC) single crystals, taking into account the plastic anisotropy due to non-planar spreading of screw dislocation
cores is presented. In view of the long-standing contradictory statements on the deformation of BCC single crystals and their macroscopic slip planes, recent insights and developments are reported and included in this model. The flow stress of BCC single crystals
shows a pronounced dependence on the crystal orientation and the temperature, mostly
due to non-planar spreading of a/2h111i type screw dislocation cores. The main consequence here is the well-known violation of Schmids law in these materials, resulting in
an intrinsic anisotropic effect which is not observed in e.g. FCC materials. Experimental confrontations at the level of a single crystal are generally missing in the literature.
To remedy this, uniaxial tension simulations are done at material point level for -Fe,
Mo and Nb single crystals and compared with reported experiments. Material parameters, including non-Schmid parameters, are calibrated from experimental results using
a proper identification method. The model is validated for different crystal orientations
and temperatures, which was not attempted before in the open literature.
2.1 Introduction
In the present paper, attention is focused on the low temperature (room temperature and lower) properties of single BCC crystals including -Fe, metals of the group
VA (V, Nb, Ta) and of the group VIA (Mo, V, Cr) and certain alkali metals. These
materials show a peculiar mechanical behavior, mostly resulting from their screw
1
2 A finite strain BCC single crystal plasticity model and its experimental
identification
dislocation core configuration. They have a relatively high yield stress which is
strongly temperature, rate and orientation dependent. They exhibit complex slip
modes, dominated by the cross slip of a/2h111i screw dislocations. Due to their dislocation core structure, they show a severe glide direction sensitive behavior (slip
asymmetry) and the well-known Schmid law using the critical resolved shear stress
(CRSS) is violated. Another pronounced phenomenon is the anomalous slip (activation of an unexpected slip system at a certain orientation) observed in pure BCC
metals which, however, is only marginally dealt with in the present work.
Various discussions on the behavior of BCC crystals, reveal a number of contradictions with respect to the slip plane activity. Even though there is no generally accepted explanation to the dislocation behavior of these materials recent studies provide a good basis for the constitutive model presented in the current paper. Seeger
(2001) states that the slip nature of BCC crystals depends highly on the temperature, upon which dislocations may accommodate either a straight {110} slip or a
wavy type {112} cross slip pattern. At room temperature a/2h111i type of screw
dislocations move on {112} slip planes, which enables cross slip. The cross slip phenomenon will not be modeled explicitly but the resulting effects are taken into account in the slip and hardening laws (for example (Pichl (2002)), showing how cross
slip can be included in a model).
The value of the critical resolved shear stress (CRSS) is independent from the slip system and the sense of the slip of FCC metals. For these metals, it is generally accepted
that the only stress component affecting the glide is the Schmid stress. However,
BCC metals show an asymmetry in their slip behavior: the slip resistance in one
direction is different from the resistance in the opposite direction, indicated as the
twinning/anti-twinning asymmetry. Moreover, due to small edge fractional dislocation components in the screw dislocation core, stress components other than the
resolved Schmid stress affect the glide or the CRSS of the material. Both of these effects result from the non-planar spreading of the dislocation cores. For these reasons
Schmids law is not applicable to BCC metals. In the presently proposed crystal plasticity model these two types of intrinsic anisotropy effects will be taken into account.
The formulation of the constitutive crystal plasticity model departs from the papers
of Bronkhorst et al. (1992) and Kalidindi et al. (1992) related to FCC metals. Their developments are extended by including the intrinsic properties of BCC metals,using
a physical description of the slip law based on thermally activated dislocation kinetics. The barriers to dislocation movement are discriminated according to their
short-range or long-range nature. The short-range barrier can be overcome by thermal activation, whereas the long-range barrier is affected slightly through changes
of the elastic moduli. Non-Schmid effects are included in the model by incorporating non-Schmid terms in the slip activation. The model is implemented at a material
2.1 Introduction
2 A finite strain BCC single crystal plasticity model and its experimental
identification
1000
(101)
slip
800
600
straight
slip lines
400
200
100
200
300 T
Temperature T (K)
400
TK 500
Figure 2.1 / Flow stress vs. temperature curve for a pure Mo single crystal at a
plastic shear strain rate 8.6 104 s1 , from Seeger (2001).
tions in kink pairs as the mobility of screw dislocations is lower than the mobility
of edge components. In this temperature range, the flow stress (the stress to maintain plastic deformation after yield) of the metal is highly temperature and strain
rate dependent. The flow stress decreases with increasing temperature and increases
with increasing strain rate. Above the knee temperature, the flow stress decreases
considerably due to self diffusion and recovery processes and the mobilities of screw
and non-screw dislocations are no longer substantially different. The high temperature range is out of the scope of the present work. Attention will be focused on the
behavior at lower temperatures, including the behavior at room temperature.
The flow stress dependence on the temperature and the slip mechanism in this temperature range is visualized in Fig. 2.1, where the presented numerical values refer
to molybdenum single crystals. Below the lower limit T (70 K for Mo and 120 K for
-Fe) the dislocation glide is confined to {110} planes. Screw dislocation glide produces straight step patterns. In this temperature range, dislocations are stated to be
in their ground state. They show a threefold symmetry and they are able to slip on
any of the three {110} slip planes. As a result of mirror symmetry and the absence of
cross slip, no plastic anisotropy is observed.
the dislocation core configuration changes and dislocations undergo a tranAbove T,
sition from their low temperature configuration (slipping on {110} planes) to their
high temperature configuration (slipping on {112} planes). The dislocations show
a wavy type of structure which results from the cross slip, a characteristic property
2 A finite strain BCC single crystal plasticity model and its experimental
10
identification
associated with {112} slip. In BCC metals three {110} and three {112} slip planes intersect on a common h111i direction and screw dislocations can therefore distribute
their core on these planes. This spreading is non-planar and nearly all peculiarities
in the mechanical properties of BCC metals can be attributed to this phenomenon.
The cross slip accommodated by {112} slip planes is the main source of the nonSchmid behavior (orientation dependence of the flow stress). Similar explanations
of the change in the slip mechanism in BCC materials were presented by others. For
example Christian et al. (1990) focused on the mean jump distance of screw dislocations in BCC metals, which decreases considerably above a critical temperature. The
explanation is again a change in the slip mechanism of BCC metals.
The above paragraphs emphasized some recent developments in the understanding
of the slip system activity of BCC crystals and the connection between the core structure and the intrinsic properties. The development of an adequate model will be
based on these considerations.
11
pling of the screw dislocation to the BCC lattice, which constrains the core and its
properties to adopt the symmetry of the lattice (Duesbery and Vitek (1998)). From
the modeling point of view, this phenomenon is included in the models by taking a
different slip resistance for the twinning and anti-twinning planes.
Table 2.1 / {112} slip systems of BCC crystals, T and A referring to twinning/antitwinning planes
(4) A (112)[111]
(1) A (112)[111]
(2) A (112)[1
11]
(5) A (121)[111]
111]
(3) A (112)[
(6) A (121)[11
1]
(7) A (121)[111]
(10)T
111]
(8) A (121)[
(11)T
(9) T (211)[111]
(12)T
(211)[111]
(211)[11
1]
11]
(211)[1
2 A finite strain BCC single crystal plasticity model and its experimental
12
identification
(2.1)
The tensor F p defines the stress-free intermediate configuration. In this configuration, resulting from plastic shearing along well-defined slip planes of the crystal lattice, the orientation of the crystal lattice is identical to the orientation in the reference
state (see Fig. 2.2). The tensor F e reflects the lattice deformation and local rigid body
rotations. The slip systems are labeled by a superscript , with = 1, 2..., ns where
m = Fe . m 0
T
n = Fe . n
F = Fe. Fp
Fe
n0
m 0
Fp
n0
m 0
ns is the total number of slip systems. The vectors m0 and n0 denote the slip direction and the slip plane normal in the reference and intermediate configurations. In
the current state they are represented by m and n , respectively.
The crystallographic split of the plastic flow rate is given by
ns
Lp =
m0 n0
=1
(2.2)
13
1
(C e I ) , C e = F eT F e
2
(2.3)
with C e the elastic right Cauchy-Green tensor and I the second order unity tensor.
The second Piola-Kirchhoff stress is the pull-back of the Kirchhoff stress tensor,
T
1
S = F
e Fe
(2.4)
where the Kirchhoff stress can be written in terms of the Cauchy stress using the
Jacobian according to
= J e with J e = det ( F e )
(2.5)
= m n = m0 C e S n0
(2.6)
The slip systems of the {112} family in BCC crystals are given in Table 2.1, which is
the relevant set within the considered temperature range.
Plastic slip and hardening
In order to include the previously described crystallographic intrinsic properties, a
physical description of the slip law will be used instead of a classical phenomenological (power law) relation. Physically based slip laws have been formulated in
crystal plasticity models for materials with a symmetric planar slip dependency (e.g.
Kothari and Anand (1998)). The anisotropy in BCC crystals is included by adapting
the slip law accordingly.
The physical interpretation given hereafter, relies on the thermally activated dislocation kinetics. Plasticity occurs by dislocation motion on certain slip planes in an
2 A finite strain BCC single crystal plasticity model and its experimental
14
identification
energetically favorable direction. The actual flow stress is determined by the resistance to this dislocation motion. The motion is obstructed by short-range and longrange barriers. The short-range barriers in general are generated by the Peierls stress
(periodic resistance of the lattice) and the local forest of dislocations. The long-range
barriers originate from the elastic stress field due to grain boundaries, far field forests
of dislocations and other defects. The total resistance can be split accordingly
s = st + sa
(2.7)
where the short-range barriers are responsible for the first part st , referred to as the
thermal part since thermal activation is sufficient to overcome this resistance. The
athermal part of the resistance sa is related to the long-range barriers. Although
this contribution slightly decreases at higher temperatures (through a decrease of
the elastic moduli), this effect is negligible compared to the change of the thermal
resistance with varying temperature.
During their motion, dislocations are obstructed by a quasi-periodic short-range resistance. The Helmholtz free energy required to isothermally cross a barrier is denoted by F and the mechanical work of st can be written as W. The energy difference between these two,
G = F W
(2.8)
(2.9)
with l representing the distance between the barriers, 0 the attempt frequency, k
the Boltzmann constant and T the absolute temperature. The relation between the
slip rate and the average velocity is given by the Orowan relation, = bm v where
m and b represent the mobile dislocation density and Burgers vector, respectively.
Substituting the velocity expression (2.9) into the Orowan relation leads to the slip
law according to,
0
if ef f 0
=
(2.10)
G
0 exp { kT
} sign ( )
if 0 < ef f
where ef f = | | sa is the driving force for the dislocation motion and 0 =
bm l 0 is the reference strain rate, which is different for the different BCC slip plane
families since the distance between the barriers depends on the family. The energy
G to be supplied by the thermal fluctuations at constant temperature is calculated
as (see Kocks et al. (1975))
"
p #q
e f f
G = G0 1
(2.11)
st
15
where G0 is the activation free energy needed to overcome the obstacles without
the aid of an applied stress. The quantities p and q lie in the range 0 p 1
and 1 q 2, and in the numerical examples of this paper they are taken as 1.
The equations (2.10) and (2.11) constitute the slip law for materials that do not show
crystallographic asymmetry effects. The non-Schmid effects are included in equation
(2.10) by pursuing a similar strategy as introduced by Dao and Asaro (1993), where
the Schmid stress as defined by (2.6) is extended to account for the non-Schmid contribution:
n = + :
(2.12)
with the Kirchhoff stress and representing the tensor governing the non-Schmid
effects for slip system aligned with m , n and z = m n defined as,
= mm (m m) + nn (n n) + zz ( z z)+
mz (m z + z m) + nz (n z + z n)
(2.13)
(2.14)
The hardening moduli h determine the rate of strain hardening on slip system
due to slip on slip system . This self and latent hardening are phenomenologically
described by (Asaro and Needleman (1985))
h = q h
(2.15)
1 qn . . . qn
qn 1 . . . qn
q = .. .. . .
(2.16)
.
. .
. ..
qn qn . . . 1
2 A finite strain BCC single crystal plasticity model and its experimental
16
identification
Fe
Fp
S, , n
Fc = Fe F p
R = F Fc
F e := F e + F e
Figure 2.3 / Schematic overview of the BCC crystal plasticity model.
where qn represents the ratio of the latent hardening with respect to the self hardening for non-coplanar slip systems.
Finally the specific form of the self hardening rate, which is motivated by Brown
et al. (1989), reads
a
sa
sa
h = h0 1 sign 1 ,
(2.17)
ss
ss
where h0 , sa and ss are the initial hardening rate, the actual athermal slip resistance
and the saturation value of the slip resistance, respectively. The exponent a is considered as a constant material parameter.
Implementation of the constitutive model
17
2 A finite strain BCC single crystal plasticity model and its experimental
18
identification
applied stress affects the yielding of a crystal. The associated non-Schmid parameters
will be identified together with the other parameters in the model.
[111]
[001]
Stress (MPa)
120
100
80
60
[011]
40
20
0
3
Strain %
] oriented Figure 2.4 / Tensile orientation dependence of [001], [011] and [111
Fe single crystals. Solid lines are the simulation results and dashed lines represent
experiments.
results show the pronounced influence of the orientation on the yielding and hardening behavior of the crystal. The orientations selected in the example constitute the
] directions. It is a wellcorners of a unit triangle mapping the [001], [011] and [111
established fact that at the corners and at the edges of the unit triangle multislip is
observed, while the deformation starts with single slip for directions mapped inside
the triangle. For these orientations the initial rate of work hardening is relatively
high and decreases rapidly with ongoing deformation.
19
Material parameters have been identified using a least-square optimization procedure which minimizes an objective function that equals the sum of squares of the
differences between experimental and simulation results. Most of the parameters are
presented in Table 2.2. The remaining parameters are C11 = 236GPa, C12 = 134GPa,
Table 2.2 / Material parameters for -Fe single crystals
h0
ss
a
st
s a0
s a0
mm
nn
zz
0
G0
697.88 MPa
132.10 MPa
1.5
13.91 MPa
9.59 MPa
5.75 MPa
0.0544
-0.0293
-0.0267
1.07 106 s1
2.95 1018 J
C44 = 119GPa (Adams et al. (2006)), qn = 1.4, k = 1.3807 1023 J/K, while mz and
nz are taken zero..
2 A finite strain BCC single crystal plasticity model and its experimental
20
identification
40
[111]
35
Stress (MPa)
30
25
20
[101]
15
10
[010]
5
0
0.2
0.4
0.6
0.8
Strain %
1.2
1.4
Figure 2.5 / Orientation dependence of [010], [101] and [111] oriented molybdenum single crystals. Solid lines are the simulation results and dashed lines represent
experiments.
h0
ss
a
st
s a0
s a0
mm
nn
zz
0
G0
251.37 MPa
76.90 MPa
1.05
11.89 MPa
7.23 MPa
4.34 MPa
-0.0528
0.0896
-0.0369
1.4 107 s1
0.1554 1018 J
et al. (1983)), Ta (e.g. Werner (1987)), -Fe (e.g. Brunner and Diehl (1987), Brunner
and Diehl (1997)). Below the so-called knee temperature, where the flow stress is
controlled by the mobility of screw dislocations, the critical shear stress of BCC metals rises progressively with decreasing temperature (Seeger (1981)). The flow stress
in FCC metals on the contrary, only shows a moderate increase when the temperature is lowered below room temperature. Most of the work reported on BCC metals
concentrates on the behavior of pure single crystals in order to eliminate secondary
21
50
45
40
Stress (MPa)
35
30
25
20
15
77 K sim.
77 K exp.
113 K sim.
113 K exp.
175 K sim.
175 K exp.
10
5
0
0.2
0.4
0.6
0.8
1
Strain %
1.2
1.4
1.6
1.8
In Fig. 2.6, the tensile stress-strain curves for a [001] oriented Nb single crystal are
presented at three different temperature levels, at a strain rate of 1.3 104 s1 . The
experimental data is taken from Duesbery and Foxall (1969). The true stress and true
)[111
] primary
strain curves are reproduced from the shear values reported for (112
2 A finite strain BCC single crystal plasticity model and its experimental
22
identification
slip. The Schmid factor for the [001] orientation equals 0.471. The experiment and the
crystal plasticity simulations show an adequate agreement. The material parameters
for this material are: C11 = 250GPa, C12 = 135GPa, C44 = 30GPa (Carroll (1965)),
0 = 1.2 107 s1 , qn = 1.4, G0 = 1.12 1018 J, k = 1.3807 1023 J/K, a = 1.2,
s a0 = 4MPa (atwin sense) and s a0 = 2.4MPa (twin sense). Due to the lack of experimental evidence the identification of the non-Schmid parameters is disregarded and
the effect is excluded here. The variation of most parameters with the temperature is
negligible, with the exception of the parameters presented in Table 2.4.
Table 2.4 / Material parameters for Nb single crystals at three different temperatures
Temperature
h0
ss
st
77 K
1500 MPa
21.7 MPa
15.3 MPa
113 K
922 MPa
19.9 MPa
8.06 MPa
175 K
800 MPa
13.42 MPa
3.18 MPa
23
characteristics result from the actual formulation of the slip rate equation. Additional
to the pre-mentioned existing crystal plasticity frameworks, in this contribution the
non-Schmid behavior is introduced in the slip law by modifying the effective shear
stress, where the non-Schmid contribution represents the dislocation cores spreading
in a non-planar manner. Actually, the emphasis was put on this intrinsic anisotropy
effect, even though other anisotropy effects such as texture development and dislocation sub-structure evolution may be included in the model as well.
The necessity for this research results from the fact that there is a lack of published
works, that confront recent BCC single crystal plasticity models to single crystal experimental data that reveals the intrinsic orientation and temperature dependence of
these crystals.
24
Chapter three
Abstract / Sheet metal forming processes are within the core of many modern manufacturing technologies, as applied in e.g. automotive and packaging industries. Initially flat sheet material is forced to transform plastically into a three dimensional shape
through complex loading modes. Deviation from a proportional strain path is associated
with hardening or softening of the material due to the induced plastic anisotropy resulting from the prior deformation. The main cause of these transient anisotropic effects at
moderate strains is attributed to the evolving underlying dislocation microstructures. In
this paper, a composite dislocation cell model, which explicitly describes the dislocation
structure evolution, is combined with a BCC crystal plasticity framework to bridge the
microstructure evolution and its macroscopic anisotropic effects. Monotonic and multistage loading simulations are conducted for a single crystal and polycrystal BCC metal,
and obtained macroscopic results and dislocation substructure evolution are compared
qualitatively with published experimental observations.
3.1 Introduction
For each car, the automotive industry manufactures more than 500 parts by multistage forming operations, involving complex deformation paths. Deviation from
a proportional strain path is commonly associated with a change in the hardening
(or softening) behavior of the material. In order to achieve a first-time-right design,
modern predictive tools relying on the finite element method are commonly used
nowadays. The anisotropy induced by complex deformation paths, which may lead
1
25
to premature failure (e.g. Sang and Lloyd (1979)), is crucial in this sense and should
be included in the constitutive models used in the analysis.
The overall plastic anisotropy in BCC metals, induced by the imposed deformation,
originates from different sources at different length scales. Slip asymmetry and intrinsic anisotropy effects caused by the non-planar spreading of screw dislocation
cores are active at the micro level (e.g. Bassani et al. (2001), Duesbery and Vitek
(1998), Ito and Vitek (2001), Yalcinkaya et al. (2008)) whereas the development of dislocation substructures is relevant at the meso level (e.g. Rauch and Schmitt (1989),
Wagoner and Laukonis (1983), Rao and Laukonis (1983), Wilson and Bate (1994),
Gardey et al. (2005)). At the macro level, the texture development of polycrystalline
metal is contributing dominantly (e.g. Bacroix et al. (1994), Bacroix and Hu (1995),
Nesterova et al. (2001)). Upon switching strain paths, the intrinsic properties obviously have a substantial effect on the observed anisotropy due to changes in the
dislocation activity. However, the evolution of dislocation microstructures has been
recognized as a main driver triggering the observed anisotropic material behavior. In
a recent report Li et al. (2006) commented on the strong anisotropy, i.e., larger than
expected from the texture, induced by the dislocation structure in IF steel increasing
with the rolling prestrain. The prediction of dislocation microstructures within the
individual dislocation descriptions and continuum theories has been a challenging
subject in the last decades in the material science community (see Groma (1997) for an
overview). While transmission electron microscopy (TEM) observations have been a
powerful tool to understand their origin and to derive the physical parameters that
govern their evolution (e.g. Fernandes and Schmitt (1983)), discrete dislocation models and atomistic considerations improved the understanding of the formation and
the evolution of dislocation microstructures and the related plastic anisotropy. However, only a limited number of micromechanical modeling approaches have been addressing the anisotropy induced by evolving dislocation cells with a crystal plasticity
framework.
Among the attempts to develop plastic anisotropy models that incorporate the microstructure evolution for complex deformation histories, the most remarkable one is
the constitutive model proposed by Teodosiu and Hu (1995). This phenomenological
model uses the Hill criterion for the onset of yielding while the hardening is associated with the dislocation structures. The polarity of dislocation walls, the back-stress
and the strength of the dislocation structure are accounted for by internal variables.
Recently, Wang et al. (2008) presented an improvement of this model especially concentrating on continuous loading path changes from uniaxial tension to simple shear
without unloading the material.
Another attempt to describe the occurring phenomena is presented by Peeters et al.
(2000) dealing with a polycrystal plasticity model that incorporates more details of
27
3.1 Introduction
the microstructure evolution at the grain scale, where cell boundary dislocation densities, cell block boundary dislocation densities, and directionally movable dislocation densities are taken as internal variables. This model attributes a major part of the
strain path change effects to the evolution of cell block boundaries and polarization
of these structures. Additional to above mentioned models, Hoc and Forest (2001),
Mollica et al. (2001) and Tarigopula et al. (2008) presented some other approaches
dealing with the anisotropic strain path change effects. In the present paper the con-
w
w
r
Figure 3.1 / Composite representation of the cell structure.
cell anisotropy. The accordance of the results with respect to published experimental
results is discussed. Finally, concluding remarks are given in section 6.
Cartesian tensors and associated tensor products will be used throughout this paper,
making use of a Cartesian vector basis {e1 e2 e3 }. Using the Einstein summation rule
for repeated indices, the following conventions are used in the notations of vectors,
tensors, related products and crystallography:
scalars a
vectors a = ai ei
second-order tensors A = Ai j ei e j
fourth-order tensors 4 A = Ai jkl ei e j ek el
C = a b = ai b j ei e j
C = A B = Ai j B jk ei ek
C = 4 A : B = Ai jkl Blk ei e j
crystallographic direction, family [uvw], huvwi
crystallographic plane, family (hkl ),{hkl }
slip system, family (hkl )[uvw], {hkl } huvwi
29
sheets is not always manifest and sometimes the microstructure is partitioned by ordinary cell boundaries having no particular crystallographic or macroscopic orientation (Hansen and Huang (1997)). For that reason, depending on the grain orientation
and the strain direction, either parallel dislocation walls or more equiaxed closed
cells are observed (e.g. Rauch and Schmitt (1989)) in low carbon steels. Besides, different materials end up in different type of microstructures. It is neither experimentally nor computationally an easy task to identify the type of evolving dislocation
microstructure, yet formation of dislocation cells is mostly observed. Thereof, this
paper concentrates on the formation and evolution of these dislocation cell structures.
MONOTONIC
CROSS TEST
STRESS REVERSAL
Figure 3.2 / Schematic evolution of a dislocation cell structure under strain path
change.
As discussed above, dislocation cell structures develop upon plastic strain in most
metals, and evolve in a distinct way depending on the applied strain path. The main
features are visualized in Fig. 3.2. Under monotonic deformation a dislocation cell
structure appears and evolves towards a decreasing cell size r, and wall thickness w
accompanied by an increasing dislocation density in the cell walls w (e.g. Fernandes
and Schmitt (1983)). After a strain path change, the developed cell structure adjusts
to the new loading and the dislocation microstructure induced by the prestrain becomes unstable. It is disrupted and dissolved, and a new dislocation structure typical
of the new strain path forms (Barlat et al. (2003)). The characteristic features of the
initial cell structure disappear as the deformation proceeds in the new direction. Unfortunately, there is no clear interpretation of what is occurring with the dislocation
microstructure during the adaptation nor is there a unique terminology to describe
this evolution. Here we distinguish between two different scenarios; dissolution of
cells as in the cross test and disruption as occurring under reversed loading. There
appears to be no consistency in the literature in the use of the dissolution, disruption
and disintegration of dislocation cells, and most of the time any cell evolution after
a strain path change is described as a dissolution process (e.g. Rauch and Schmitt
(1989), Rauch (1992), Rauch (1991) Gardey et al. (2005), Rao and Laukonis (1983)).
Indeed, both dissolved and disrupted structures appear as disorganized structures
with a higher degree of homogeneity compared to the state before the strain path
change. Nevertheless, there are indications that there is a morphological difference
between the two microstructure evolution scenarios mentioned above (e.g. Gardey
et al. (2005)) in correspondence with the difference between the driving forces and
their physical origins.
Figure 3.3 / Left: Experimental results for mild steel DC06 subjected to monotonic
simple shear and simple shear followed by load reversal (10 % and 30 %) [Bouvier
et al. (2006)] (reversed loading) Right: Experimental results for IF-steel subjected to
monotonic simple shear and tensile tests (10 % and 20 %) followed by shear [Peeters
et al. (2000)] (orthogonal loading).
In the example shown in Fig. 3.2, two different strain path changes have been considered where the two types of evolution phenomena can be distinguished. A cross test,
e.g. tension followed by simple shear or a tension test followed by tension in a dif-
31
ferent direction, reveals progressive cell evolution (e.g. Rao and Laukonis (1983)). It
has been observed that after a strain path change cell walls become thicker while the
dislocation density in the walls becomes smaller (e.g. Schmitt et al. (1991)). Hence,
the dislocation distribution is more uniform and the cell structure is less organized.
This evolution process can cause partial or complete dissolution of the existing cell
structure, while concurrently a new cell structure develops with a morphology related to the new loading direction. The cell structure evolution resulting from a
stress reversal has received more attention in the context of the analysis of the wellknown Bauschinger effect (e.g. Rauch (1991)). The evolution of the cell structure
under stress reversal can be characterized by the disruption of cell walls (e.g. Viatkina (2005), Christodoulou et al. (1986)). The thickness of the cell walls does not
change significantly, however, the walls tend to disconnect. Experimental observations (e.g. Christodoulou et al. (1986)) also report a strong flux of dislocations from
the walls to the cell interiors, decreasing the wall dislocation density and increasing
the density in the cell interiors. Accordingly, the descriptive modeling of the dislocation distribution relies on an increase of cell size and a dislocation redistribution
(Viatkina (2005)). With ongoing deformation cell walls reappear and a new cell structure originates.
Upon sustained loading after a strain path change, the microstructure always evolves
such that transient effects disappear and the macroscopic stress-strain curve saturates to the monotonic deformation curve (see Fig. 3.3).
F i = F ie F ip ,
(3.1)
where the superscript i indicates the specific component (w: wall, c: cell) and tensor
F ip defines the stress-free intermediate configuration. In this configuration, resulting
from plastic shearing along well-defined slip planes of the crystal lattice, the orientation of the slip systems is unaltered. The tensor F ie reflects the lattice deformation
and local rigid body rotations. The slip systems are labeled by a superscript , with
= 1, 2..., ns where ns is the total number of slip systems. The vectors m0 ,i and n0 ,i
denote the slip direction and the slip plane normal in the reference and intermediate
configurations. In the current state they are represented by m ,i and n ,i , respectively.
i
The crystallographic split of the plastic flow rate Lip = F p F ip 1 is given by
Lip =
ns
,im0 ,i n0 ,i,
(3.2)
=1
with
E ie =
1 iT i
( F F e I ),
2 e
(3.3)
with I the second order unity tensor and 4 C the fourth order tensor consisting of
elastic moduli.
From the second Piola-Kirchhoff stress the Kirchhoff stress in the current configuration can be determined by a push-forward operation,
T
i = F ie Si F ie .
(3.4)
From the Kirchhoff stress the Cauchy stress can be derived according to,
i =
1 i
Jei
with
Jei = det( F ie ).
(3.5)
The Schmid resolved shear stress is the projection of the Kirchhoff stress on the slip
systems, i.e.
,i = m ,i i n ,i = m0 ,i C ie Si n0 ,i
with
C ie = F ie F ie ,
(3.6)
which is the driving force for the dislocation movement on a certain slip system .
There has been various discussions and contradictions considering the active slip
systems of BCC crystals, and recent studies shows that the slip system activation
is highly temperature dependent (see Yalcinkaya et al. (2008)). At room temperature the {112} slip system family is dominantly active. The effect of the non-Schmid
stresses on the non-planar screw dislocation cores which contributes to the orientation dependence of BCC crystals at single crystal level can be taken into account
as an additional contribution to the driving stress in equation (3.6) (Yalcinkaya et al.
(2008)), however, this contribution affects the initial anisotropy of these metals rather
than the transient effects observed during the strain path changes (see Fig. 3.3). Including this effect would increase the material parameters while it does not contribute to the aim of this paper. Hence it was decided not to account for this effect
here.
33
,i
kT
st
where ef,if = | ,i | sa ,i is the effective driving stress on the slip systems, G0 is the activation free energy, k is Boltzmanns constant and T is the absolute temperature, 0 ,i
is a reference strain rate. For isothermal cases, the thermal part st ,i of the slip resistance is taken constant and the a-thermal slip resistance is related to the dislocation
densities on all slip systems through,
s
sa ,i
ns
= Gb
Au |u,i |,
(3.8)
u=1
where G is the shear modulus, b is the magnitude of the Burgers vector, A u are
the interaction coefficients between the slip systems and u, and u,i the dislocation
density on the slip system u of component i. The dislocation interaction coefficients
of the matrix A u depend on the type of interaction between dislocations on different
slip systems (e.g. Franciosi and Zaoui (1982), Queyreau et al. (2008)). Because of the
lack of data on {112} slip systems, only the interactions between the dislocations
belonging to the same slip system, i.e. = u, and different slip systems, i.e. 6= u,
will be distinguished for A u .
The macroscopic mechanical response of the composite model is obtained by applying a Taylor averaging assumption where the deformation in each component is
assumed to be equal to the macroscopic deformation and where the rule of mixtures
gives the macroscopic stress from the local stresses in each component according to,
= f w + (1 f ) c .
(3.9)
In this equation f represents the actual volume fraction of the cell walls, expressed
in terms of the microstructural morphology parameters w and r according to:
w 2 w 3
Vw
w
f =
= 3 3
+
,
V
r
r
r
(3.10)
where V and V w are the volumes of the entire composite and the wall component,
respectively.
Experimental studies (e.g. Fernandes and Schmitt (1983)) suggest that the wall thickness w, the cell size r (see Fig. 3.1), the dislocation densities in the cells ,c, and
the walls ,w evolve with increasing applied strain. Moreover, these quantities are
dependent on the deformation history, and therefore they are taken into account as
internal variables in this framework. Corresponding evolution equations are to be
formulated that describe the cell structure development during monotonic loading
as well as complex strain path histories. This is done in the following section, where
the incorporation of a dislocation cell structure evolution model into the crystal plasticity framework is presented.
1 ,w
,c ,w
,w
,w
=
I R
f
| | +
b
f
(3.11)
35
The following empirical relation between the cell size r and the flow stress y is
suggested in the literature (e.g. Barker et al. (1989), Mughrabi (1987)),
y =
CGb
,
rm
(3.12)
,w
s a
CGb
+ (1 f ) sa ,c
(3.13)
The evolution of the wall thickness w is adopted from Viatkina et al. (2003) and assumed to be governed by an effective plastic strain rate measure | | according
to
w = km (win f w) | |
with
| | = f | ,w | + (1 f ) | ,c|.
(3.14)
In this evolution law a decrease of the wall thickness with a saturation factor km is
incorporated, with a final saturation value equal to win f , which is consistent with
experimental observations (e.g. Fernandes and Schmitt (1983)).
The implementation of the model presented above follows an incremental-iterative
solution procedure, which is applied for each of the composite components with the
same imposed deformation (Taylor approach). The first step in this procedure is the
initial estimate for the elastic part F ie , resulting in an estimate for the plastic part F ip
through (3.1). With the kinematics defined, both the stress and the Schmid stress is
calculated. These values together with the slip resistance (3.8) (which is calculated
from dislocation density evolution (3.11)) enter the slip law (3.7) resulting in the slip
rates on each slip system. The updated plastic part of the deformation gradient is
obtained from the calculated slip rates through a time integration scheme. Generally, the calculated and the imposed deformation will be different, which results in a
residual. Iteration on the residual leads to updated values of variables including F ip
and F ie . With the current values of r and w the volume fraction f is calculated with
(3.10), which is used to determine the macroscopic stress (3.9). The procedure is repeated for all time steps, which results in the entire history of stress, slip and internal
variable evolution.
(3.15)
where kd is the dissolution factor. The saturation value of the wall thickness logically
equals the cell size r corresponding to a complete dissolution of the cell. When both
the dissolution and the redevelopment processes are taken into account, the wall
evolution becomes
w = pkd (r w) | | + (1 p)km (win f w) | |,
(3.16)
where the first contribution on the right-hand side reflects equation (3.15) accounting
for the effect of the loading in the new direction, i.e. the dissolution process. The
second contribution on the right-hand side of (3.16) represents the development of
the wall structure according to equation (3.14). The transition parameter p, to be
specified in the following, defines the relative contribution of the dissolution process
in the evolution of w. It is characterized by taking into account: (i) the dissolution
process depends only on the angle between successive deformation paths and (ii)
the dissolution effect disappears as the deformation proceeds in the new direction.
In this context, the following expression for p is taken:
,
(3.17)
p = (1 ||) exp B | | | pre |
where is a scalar measure that identifies the strain path change and |pre | indicates the accumulated plastic deformation prior to the strain path change and B is a
material parameter. To characterize the strain path change measure , a commonly
used definition is adopted here:
L p1 : L p2
,
| L p1 || L p2 |
(3.18)
37
where L p1 and L p2 are the macroscopic plastic velocity gradient tensors prior to and
after the strain path change. Here = 1 refers to monotonic deformation, = 0 to a
cross test, and = 1 to a reverse test.
Equations (3.16) and (3.17) describe the evolution of the wall thickness during the
whole deformation process. During monotonic deformation where = 1, equation (3.16) reduces to equation (3.14) describing cell wall thinning. After a strain
path change, the dissolution is initiated with an intensity proportional to (1 ||).
The dislocation walls start widening, governed by the competition between the new
structure development and the old structure dissolution. As the deformation proceeds in the new direction, the dissolution process fades out and accordingly p approaches 0 due to (3.17) and the wall thickness tends to decrease again: a new dislocation structure is developing.
,w
s a
CGb
+ A exp [kc ( | | |pre |)],
+ (1 f ) sa ,c
(3.19)
where A defines the degree of disruption and kc is a constant reflecting the recovery
speed (governed by slip). The entire term in the right-hand side (added to (3.13))
determines the immediate increase of the cell size. This effect vanishes with ongoing
deformation depending on the value of the parameter kc . As soon as the disruption
contribution gradually disappears, the cell size again follows the evolution as given
for the monotonic deformation case. To reflect the fact that the cell disruption is
triggered by a stress reversal, the coefficient A depends on the strain path change
through
a||
if < 0
A=
(3.20)
0
if 0
where a is a fitting parameter. Consequently, the Bauschinger test triggers the highest disruption. For more complex strain path changes with negative values of both
equations (3.16) and (3.19) have non-zero strain path change contributions (p > 0
and A > 0), describing a process with coexisting dissolution and disruption of cells.
When p = 0 and A = 0, the model describes the evolution under monotonic deformation. In this way the resulting system of equations effectively unifies different
types and combinations of strain path changes.
39
Fernandes and Schmitt (1983) give data on the wall thickness and dislocation cell
size of low-carbon steel under various types of loading. In the current framework,
the parameters in the evolution equations are identified to retrieve this characteristic
behavior.
Figure 3.4 / Evolution of the cell structure variables during monotonic deformation
for -Fe. Left: wall thickness w and cell size r. Right: volume fraction of the walls.
In Fig. 3.4 the evolution of the microstructure of an -Fe single crystal during a
uniaxial tension simulation at 298 K and a strain rate of 5 104 s1 is presented
with the material parameters given in Table 3.1 2 . The cell size shows a decreasing
trend with increasing strain as expected. The wall thickness decreases quickly and
stabilizes at a constant value. The volume fraction approaches a value of around 0.1
after a sharp decrease. This trend can also be found in the literature, where it is stated
that the volume fraction of dislocation walls remains at a constant value (e.g. Peeters
(2002)).
Note that the number of papers focusing on the determination of the interaction coefficients in
BCC single crystals are very limited. Moreover, the actual value strongly depends on the actual BCC
crystal considered and impurities present and the initial dislocation density. Values may therefore
differ considerably, where a difference of a factor 10 can be easily find throughout the literature.
Youngs modulus
Shear modulus
Poissons ratio
Reference strain rate
Burgers vector length
Interaction coefficient (self)
Interaction coefficient (latent)
Initial dislocation density
Activation free energy
Dislocation multiplication parameter
Dislocation annihilation rate parameter
Boltzmann constant
Thermal dislocation resistance
Saturation factor
Saturation value
Material constant
E
G
0
b
A
A u
0
G0
I
R
k
st
km
win f
C
139 GPa
64 GPa
0.362
1.07 106 s1
0.248 109 m
0.00072
0.001
0.18 1014 m2
2.95 1018 J
0.228
5.1 109
1.3807 1023 J/K
15 MPa
150
0.18 106 m
20
Table 3.1 / Material parameters for monotonic loading case. Some of the parameters, i.e. 0 , G0 and st have already been identified in previous work (Yalcinkaya
et al. (2008)). The parameters b and G are taken from Frost and Ashby (1982). The
value of 0 is presented by Krejci and Lukas (1971). The latent interaction coefficient
A u is identified and the self interaction coefficient A is obtained by assuming a
ratio of 1.4 between latent and self hardening. The remaining parameters are identified by comparing the experimental trends with computational results. Additional
parameters needed for complex strain paths are commented in the text.
is not incorporated, i.e. transient hardening and softening effects due to the strain
path change (see solid line in Fig. 3.5) are absent. The dashed line presents the
outcome of the full microstructure evolution computation, in which both the crystal slip anisotropy and the dislocation cell anisotropy are revealed. The purpose of
this example is to discriminate these two intrinsic sources of anisotropy at the single
crystal level. Obviously, both mechanisms here contribute to a larger yield stress after reloading. Whereas this increase is systematic for dislocation cell contribution, it
is obviously orientation dependent for the slip anisotropy contribution.
41
60
50
Stress (MPa)
40
30
20
10
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Strain
Figure 3.5 / Stress-strain curve for [001] uniaxial tension followed by [011] uniaxial
tension with (dashed line) and without (solid lines) microstructure evolution effect
for -Fe single crystal.
As explained previously, at a single crystal level the effect of intrinsic crystallographic anisotropy during a strain path change is noticeably high. The initial
anisotropy, the so-called orientation dependence of BCC single crystals has been
studied before (e. g. Yalcinkaya et al. (2008)) and this effect adds up to anisotropy
due to the dislocation microstructure evolution. In this example the main interest
focuses on the anisotropy due to substructure evolution during multi-stage loading
processes for the case where the intrinsic orientation effect is known to contribute
less. To this purpose polycrystal simulations have been conducted, where 100 randomly oriented crystals are considered interacting according to a Taylor averaging
scheme.
First, the cell dissolution process and its macroscopic cross effect are analyzed. The
characteristic feature of the stress-strain curve in Fig. 3.3 is the transient change induced by a change of the deformation path. An initial increase in the yield stress is
followed by moderate softening. The cross effect vanishes gradually and the curve
saturates towards to the monotonic case. In order to measure this effect experimentally either tension followed by simple shear or two successive orthogonal tensile
experiments need to be conducted. With respect to the latter approach Schmitt et al.
(1991) presented clear experimental results where various tensile sequences were examined, with different angles between the succeeding tensile directions equal to 15 ,
45 and 90 with different amounts of prestrain. In Schmitt et al. (1991), it was reported that no evolution of cell-blocks was observed, supporting the case examined
here, where the cell structure development is assumed to be the main mechanism
accompanying the strain path change. The sequence of two uniaxial tests with 45
between the tensile axes is, according to equation (3.18), characterized by = 0.25,
and it is rather close to a cross test exhibiting the highest cell dissolution. Additional
to the parameters used during monotonic deformation (see Table 3.1), kd and B are
identified as 300 [-] and 20 [-] respectively. The obtained typical cross effect is presented in Fig. 3.6. This transient effect is observed in many materials and the size
of the effect is determined by the amount of applied prestrain while the shape of the
hardening and softening zones depends on the material.
60
50
Stress (MPa)
40
30
20
10
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Strain
Figure 3.6 / Predicted stress-strain curve of a 45 cross tensile test of an -Fe polycrystal.
The second example concerns the Bauschinger effect, which yields a reduction of the
yield strength of the material after a load reversal. A simple shear and reversal simulation is presented in Fig. 3.7 , where the evolution of the cell size r is dominantly
contributing to the anisotropy at this continuum level. Additional to the parameters used during monotonic deformation in Table 3.1, the parameters a and kc are
identified as 1 104 m and 14 [-] respectively, to validate this part of the model.
Even though both the cross effect and the Bauschinger effect are extensively documented in the literature, quantitative data on the evolution of dislocation cell structure during strain path changes remains hard to find. For this reason, a qualitative
analysis rather than a quantitative study has been conducted here.
43
50
45
40
Stress (MPa)
35
30
25
20
15
10
5
0
0
0.05
0.1
0.15
0.2
0.25
Strain
0.3
0.35
0.4
0.45
Figure 3.7 / Bauschinger effect for shear - reverse shear test of an -Fe polycrystal.
Chapter four
Abstract / A rate dependent strain gradient plasticity framework for the description
of plastic slip patterning in a system with non-convex energetic hardening is presented.
Both the displacement and the plastic slip fields are considered as primary variables.
These fields are determined on a global level by solving simultaneously the linear momentum balance and the slip evolution equation which is postulated in a thermodynamically consistent manner. The slip law differs from classical ones in the sense that it includes a non-convex free energy term, which enables patterning of the deformation field.
The formulation of the computational framework is at least partially dual to a GinzburgLandau type of phase field modeling approach. The essential difference resides in the
fact that a strong coupling exists between the deformation and the evolution of the plastic slip, whereas in the phase field type models the governing fields are only weakly
coupled. The derivations and implementations are done in a transparent 1D setting,
which allows for a thorough mechanistic understanding, not excluding its extension to
multidimensional cases.
4.1 Introduction
During forming processes most metals develop cellular dislocation structures due to
dislocation slip patterning from moderate strains onwards. Typical examples of dislocation microstructures are dislocation cells and dislocation walls (see e.g. Young
et al. (1986), Mughrabi (1987), Yalcinkaya et al. (2009)). Patterning typically refers
to the self organization of dislocations with formation of regions of high dislocation
1
45
46
density (dislocation walls) which envelop areas of low dislocation density (dislocation cell interiors), also to be regarded as domains of high plastic slip and low plastic
slip activity. Due to the induced macroscopic anisotropic effects (see e.g. Peeters
et al. (2000), Li et al. (2006), Wang et al. (2008), Yalcinkaya et al. (2009)) the occurrence
of dislocation microstructures and their evolution have been an interesting topic
for the materials science community for decades. Starting with the studies on the
cold-worked sub-structure of polycrystals using transmission electron microscopy
in 1960s (e.g. Bailey and Hirsch (1960), Keh et al. (1963) and Swann (1963)), a vast
amount of experimental results have been collected, and several promising theoretical models were presented dealing with dislocation (or slip) patterning. Nevertheless, a complete descriptive understanding of the occurring phenomena has never
been reached and the necessary input for computational models is still subject of
ongoing discussions.
In the context of the computational modeling of plastic slip pattering (or dislocation
sub-structure formation), different approaches have been pursued in the literature
that can be categorized into three main groups: (i) models using directly the mechanics of single dislocations or populations of dislocations, (ii) phase field modeling
of dislocation patterning, (iii) the incremental variational formulation of inelasticity
which has the advantage of applying the concept of relaxation resulting in fine scale
microstructure evolution.
In the first group, scientists deal either with the problem at a discrete dislocation level
by solving a system with only a limited number of dislocations (e.g. Lubarda et al.
(1993), Groma and Pavley (1993), Kubin and Canova (1992)), or they approach the
problem from a continuum point of view by using homogenized variables like dislocation densities and internal stress fields in a system of coupled balance equations.
One of the fundamental studies in this class was introduced by Kuhlmann-Wilsdorf
(see e.g. Kuhlmann-Wilsdorf and Van der Merwe (1982)) within the so-called lowenergy dislocation structure (LEDS) approach, seeking for dislocation configurations
of minimal free energy under given constraints. In this framework, pattern formation
is driven by the reduction of the system energy. Holt (1980) improved this static description by introducing dynamics into the model, which is based on a conservation
law for the dislocation density by setting up a relation for the evolution of the dislocation density in terms of elastic energy changes related to the dislocation density
fluctuations. Another approach in this class is the reaction-diffusion model (e.g. Walgraef and Aifantis (1985) and Aifantis (1987)), where mobile and immobile dislocations are distinguished and the dynamics of the system is governed by diffusion and
reaction terms. The competition between the mobility and the non-linear interactions
(creation, annihilation and pinning) causes the instability of uniform dislocation distributions versus inhomogeneous ones and leads to the formation and persistence of
dislocation patterns. Even though there have been some improvements on the men-
4.1 Introduction
47
tioned models, they generally rely on assumptions that are difficult to validate at the
discrete dislocation level.
In the second group, there are the phase field modeling approaches of dislocation
patterning, which offer a valuable alternative for discrete dislocation dynamics simulations. These models have the advantage of making less small scale assumptions
on the dislocation interactions, such as multiplication/annihilation of dislocations.
Instead, elastic interactions of numerously interacting individual segments of dislocations are considered, dealing with only a few density functions (see Wang et al.
(2001a), Wang et al. (2001b) and Wang et al. (2001c)). However, these models require
the use of large computational grids and time integration over large numbers of time
steps. These difficulties are overcome by the model of Koslowski et al. (2002) presenting an analytically tractable theory which determines the value of phase field at
point-obstacle sites without using any grid. Even though the framework is successful
in predicting dislocation (slip) patterns, it is rather complicated to incorporate into
a finite-deformation formulation of single-crystal elastic-plasticity, as needed in the
context of large-scale finite element calculations of macroscopic samples.
The third category of methods aims to capture the phenomenological evolution of
dislocation microstructures by modeling a physically deformed crystal in finite plasticity frameworks, which are the most relevant ones considering the present paper.
Ortiz and Repettto (1999) regard the dislocations as manifestations of the incompatibility of the plastic deformation gradient field. Within this framework the incremental displacements of inelastic solids follow as minimizers of a suitably defined
pseudoelastic energy function. Miehe et al. (2004) propose an approach based on
finite-step-sized incremental energy minimization. The boundary value problems
were recast into a principle of minimum incremental energy for standard dissipative solids. The general concepts are applied to analyze evolving deformation microstructures in single-slip plasticity. The derivations of the incremental variational
formulation of inelasticity in both frameworks are conceptionally parallel to each
other. Hackl and Kochmann (2008) employ energy principles as well to analyze the
microstructure formation and evolution as a result of energy minimization and relaxation via lamination. The idea is that, for non-quasiconvex energy potentials the
minimizers are no longer continuous deformation fields but small-scale fluctuations
related to probability distributions of deformation gradients to be calculated via energy relaxation.
The objective of the present paper consists in bridging models of the last two categories described above. To achieve this, a continuum plastic slip field model in a
non-convex strain gradient plasticity framework is proposed to simultaneously predict microstructure formation and the overall macroscopic elastic-plastic response in
a computationally low cost setting. A viscous relaxation scheme in a system with en-
48
ergetic hardening is used, which makes the method comparable to non-local phase
field models (e.g. Ubachs et al. (2004)) as used in diffusional evolving microstructures. Using the adequate formulations for the physics, thermodynamics and kinematics, phase field models are able to predict the evolution of microstructures and
morphologies without explicit tracking of the positions of the phases and interfaces.
Inspired by their success, the present paper extends this concept to the formation
and evolution of plastic slip patterns based on a consistent derivation of the thermodynamic relations and evolution equations. Plastic slip patterns in metals are
typically linked to the evolution of dislocation sub-structures, i.e. dislocation cell
walls and cell interiors. The computational framework yields a stable algorithm for
which the finite element formulation and implementation has been performed in a
fictitious 1D setting in order to preserve simplicity and transparency, and to easily
interpret the results. Moreover, in this simple presentation the theory stays general
and depending on the parameters and their physical interpretation, the model can
potentially describe different dislocation slip microstructures (from Luders
bands
to dislocation cells) at different length scales. The formulation can be extended to
2D or 3D, where the multi-slip character of crystals is often considered as a natural
source of non-convexity due to latent hardening mechanism. In line with models of
the third category, a double-well free energy function is assumed here, which makes
the approach rather phenomenological for the considered 1D case, whereas the nonconvex nature in a single slip deformation state becomes more natural through the
interaction of multiple slip systems (e.g. through latent hardening). In the model, the
plastic slip and the displacement are taken as degrees of freedoms. These fields are
obtained on a global level by solving simultaneously the linear momentum balance
and the slip evolution equation. The thermodynamically consistent slip law is the
crucial part of the model. At first glance, the slip law (see equation (4.18)) is similar
to the one encountered in a classical rate dependent crystal plasticity approach (e.g.
Hutchinson (1976), Peirce et al. (1982), Yalcinkaya et al. (2008)), however the stress
expression in the numerator differs. Three participating stress contributions can be
distinguished: (1) the conventional resolved stress directly related to the external
loading, (2) a surface-like stress depending on the gradient of plastic slip which is
characteristic for strain gradient crystal plasticity models (e.g. Evers et al. (2004b))
and (3) the stress that emanates from a non-convex free energy, which is omitted
in classical convex theories and which triggers the patterning of the dislocation slip
field.
The paper is organized as follows. First, in section 2, the incremental variational approach for microstructure evolution is discussed shortly including some comments
on its connection with the present framework. Then, in section 3, the thermodynamical consistency of the constitutive model and the derivation of the equations to be
solved in a finite element context are studied. Next, in section 4, the finite element
formulation for the 1D strain gradient plasticity framework and the incorporation of
the non-convexity into the problem are summarized. Further, section 5 addresses the
link between the non-convexity and the resulting patterning in phase field models in
general and in the presented model in particular. In section 6 numerical examples are
presented in order to demonstrate the performance of the proposed model. Finally,
some concluding remarks are given in section 7.
50
tn +1
tn
+ D dt
(4.1)
51
additively,
= e + p
(4.2)
into an elastic part e and a plastic part p . In the single slip 1D case considered here,
the total amount of plastic slip and the plastic strain are identical. Note that in
general the plastic slip may be derived from a collection of internal variables, even
for a 1D problem (in particular if multiple slip systems are present). The free energy
is assumed to be a function of the state variables,
state = e , ,
(4.3)
where = /x is the gradient of the plastic slip. In the multi slip crystal plasticity context the definition of plastic slip gradients is not trivial, and Cermelli and
Gurtin (2001) summarize the different options and requirements for such a formulation. Following the arguments of Gurtin (e.g. Gurtin (2000), Gurtin (2002)), the
power expended by each independent rate-like kinematical descriptor is expressible
in terms of an associated force system consistent with its own balance. Yet, the basic
kinematical rate variables, namely e , u and are not independent. It is therefore not
apparent what forms the associated force balances should take, and, for that reason,
these balances are established using the principle of virtual power.
e and
Assuming that at a fixed time the fields u, e and are known, we consider u,
e , ).
as virtual rates which are collected in the generalized virtual velocity V = (u,
The force systems are characterized through their work-conjugated nature with respect to the state variables. Pext is the power expended on the domain P and Pint a
concomitant expenditure of power within P,
Pext ( P, V) =
Pint ( P, V) =
R
R
t(n)udS
+
e
P dV +
(n) dS
dV +
P
(4.4)
dV
P
where , and are the thermodynamical forces conjugate to the internal state variables e , and respectively. In Pext , t(n) is the macroscopic surface traction while
represents the microscopic surface traction conjugate to at the boundary P with
n indicating the normal direction.
Postulation of the principle of virtual power states that given any generalized virtual
velocity V the corresponding internal and external powers are balanced,
Pext ( P, V) = Pint ( P, V)
(4.5)
52
Considering a generalized virtual velocity without slip, the virtual velocity field can
be chosen arbitrarily and this leads to the classical macroscopic force balance
=0
x
(4.6)
and considering the virtual slip field arbitrarily without generalized virtual velocity
leads to the microscopic force balance,
+ = 0
x
(4.7)
In Gurtin (2000), Gurtin (2002), Gurtin et al. (2007) the previous derivations are fully
detailed in a generalized 3D setting while Del Piero (2009) presents an alternative
way to derive similar balance equations.
The local internal power expression can simply be written as,
Pi = e + +
(4.8)
or
D = e + + 0
(4.9)
= e + +
(4.10)
D = e + + e e
de
d
d
e
= e +
+
0
d
d
d
|
{z
}
{z
}
|
0
(4.11)
Additional to the stress , the microforce conjugate to the slip gradient is also
assumed to be energetic:
de
de
d
=
d
The remaining term in the dissipation expression (4.11) reads,
(4.12)
d
D=
0
d
|
{z
}
53
(4.13)
dis
where the term conjugated to the slip rate is identified as the dissipative stress dis,
dis =
d
d
(4.14)
The following constitutive equation which satisfies the inequality (4.13) is next proposed,
dis = sign ( )
(4.15)
(4.16)
In here, 0 and m are the reference slip rate and the rate sensitivity exponent respectively. Furthermore, s is the resistance to dislocation slip which is assumed to
be constant for simplicity. Substitution of equation (4.16) into equation (4.15) and
extracting the slip rate yields the slip law
= 0
| dis |
s
m1
sign ( dis )
(4.17)
As a result, a thermodynamically consistent constitutive relation for the slip evolution is obtained. By using the definition of dis in equation (4.14) and using the
microforce balance according to equation (4.7), equation (4.17) can be written as,
m1
d
| x + d |
| {z }
sign ( d )
= 0
s
d
(4.18)
Each of the stress contributions in the slip law is derived from the free energy, i.e.
= /e , = / and any change in the free energy directly affects the slip
law.
54
= e + =
1 e2 1
E + A( )2
2
2
(4.19)
| dis |
s
m1
sign (
dis
= 0
(4.20)
) = 0
For further elaborations it is more convenient to write the plastic slip evolution in the
following format,
| |m sign ( )
0m dis
=0
s
(4.21)
(4.22)
55
with t representing the time step and n is the plastic slip at the end of the previous
time step.
The dissipative stress in equation (4.22) can be written as
dis = +
(/ )
2
=+
=+A 2
x
x
x
(4.23)
(4.24)
The weak forms of the equations are obtained in a standard manner, using a Galerkin
procedure. Both the balance of linear momentum and the slip equation are tested
with virtual displacement u and virtual slip fields, respectively, and integrated
over the domain 0 < x < L. The obtained variational Galerkin functionals are solved
in a fully coupled manner (monolithic) by means of a Newton-Raphson scheme after
linearization and discretization procedures as outlined in the appendix section. Linear interpolation functions for the slip field and quadratic interpolation functions
for the displacement field u are used. The resulting system of equations is solved
for the increments of the displacement u and the plastic slip . The formulation is
numerically implemented and solutions for the nodal displacements and plastic slips
are obtained in a standard incremental iterative manner. Examples are presented in
section 4.6.
= e + +
1 e2 1
E + A( )2 + (C1 4 + C2 3 + C3 2 + C4 + C5 )
2
2
(4.25)
56
Repettto (1999) present a comprehensive summary on the relation between the nonconvexity of the energy function and the dislocation microstructure evolution. In
crystals exhibiting latent hardening, the energy function is non-convex and has wells
corresponding to single-slip deformations. This favors microstructures consisting locally of single slip. Looking from thermodynamics perspective Kuhlmann-Wilsdorf
LEDS (low energy dislocation structure) theory (e.g. Kuhlmann-Wilsdorf (2001))
completes the explanation by stating that among all potentially accessible dislocation structures, plastic deformation will generate the one with the lowest free energy.
This means that, limited only by dislocation mobility, availability of slip system and
insignificant entropy, dislocation structures always approach the lowest possible mechanical energy of the present dislocation population.
The present study clearly reveals the intrinsic role of the non-convexity in a 1D problem, even though it is a simplification of the underlying (more complex) 3D reality.
The non-convex energy which is coming from the accumulation of trapped dislocations is introduced phenomenologically in terms of a dislocation slip potential representing the microstructurally trapped energy. It is introduced in terms of plastic
slip representing the dislocation movement looking for the minimum energy configuration (in analogy to patterning in phase field models). The convex gradient term
represents the surface energy (penalizing spatial transitions from low to high values
of slip), which in fact regularizes the problem in the mathematical sense. Depending
on the value of the R (hidden in parameter A standing in front of the gradient term),
the presented theory is able to explain the formation and evolution of microstructures at different length scales.
The non-convex function does not have to be a polynomial one. The free energy function adopted is just a simple mathematical representation for a double-well function
motivated from phase field models. The parameters C1 , C2 and C3 are chosen in a
way that they introduce a small modulation from a convex plastic slip potential (see
Fig. 4.3 (a)). Their exact values are therefore not essential if the non-convex modulation is superimposed on a free energy function used for classical hardening laws.
It is clear that if a convex plastic slip potential enters equation (19) it would result in
only hardening (no softening branch) behavior and a homogeneous (constant) distribution of the plastic slip would be obtained.
The strong form of the system description according to the equations (4.20) remains
identical. The dissipative stress term in equation (4.23) can be rewritten as,
2
= + A 2 (4C1 3 + 3C2 2 + 2C3 + C4 )
x
dis = +
(4.26)
57
3
2
+ A 2 (4C1 + 3C2 + 2C3 + C4 ) = 0
s
x
(4.27)
The same procedure as in the previous section is followed to obtain the set of equations to be solved for the displacement and plastic slip increments (see the Appendix
for details).
Figure 4.1 / Free energy curves of a two-phase material, and binodal and spinodal
points.
interest are the so-called spinodal points. These points mathematically reflect the
58
state (e.g. composition) for which the second derivative of the configurational free
energy becomes zero. At these points the sign of the curvature (2 c /c2 ) changes.
In the region where the curvature is positive the system is stable with respect to
small fluctuations in composition. However in the regions where the curvature is
negative (between the spinodal points) the system is locally unstable with respect to
small fluctuations and for that reason phase separation occurs (see e.g. Cahn (1961),
Langer (1971), Nauman and He (2001)). For the case considered here, this issue is
easily shown on the basis of a stability analysis of the steady state solution of the
plastic slip distribution, implying that = 0,
dis = + A
=0
x2
(4.28)
which can be considered as a nonlinear equation in terms of . A spatial wave perturbation is applied to the distribution = 0 + = 0 + gei(kxt) with g the
complex amplitude, k the wave number and the frequency. Linearization of (4.28)
around 0 yields the behavior in terms of the perturbation
A
2 2
2
= 0
x2
=0
(4.29)
(4.30)
(4.31)
(4.32)
and between these two spinodal points the system is unstable and any given perturbation will give rise to patterning of the plastic slip.
59
models. The first example deals with a conventional convex free energy with slip
gradients (see section 4.4), where the influence of the internal length parameter determining the value of the constant A (see equation (4.19)) is elucidated. The second
example studies the absence, onset and evolution of patterning of the plastic slip, departing from a homogeneous distribution, depending on the deformation rate during
monotonic loading of a bar. The homogeneous (non-patterned) case is also compared
with the analytical solution. The third example studies plastic slip patterning in a relaxation test where the 1D bar is deformed to a certain state mapped between the
spinodal points. Constraining the deformation at this point leads to viscous relaxation of the microstructure evolution. The effect of the boundary conditions on the
slip patterning is analyzed in each example by considering hard ( = 0) and soft
( /x = 0) cases.
60
Stress (MPa)
30
R = 0.01 mm
R = 0.02 mm
R = 0.05 mm
25
20
15
10
5
0.2
0.4
0.6
0.8
0
0
0.01
0.02
0.03
0.04
0.05
Strain
ture studying energetic size effects (e.g. Shu et al. (2001), Evers et al. (2004a), Dunstan
and Bushby (2004)). Because of the constrained slip at each end of the bar, plastic slip
gradients develop, resulting in inhomogeneous deformation with the occurrence of
boundary layers. The boundary layer thickness is typically increasing with R. As expected, Fig. 4.2 also reflects the size dependence of the stress vs. strain curve which
is consistent with the literature on this aspect.
61
mm, the rate sensitivity exponent m = 1 and the length of the bar is L = 1 mm. The
overall deformation rate used in this calculation is = 2 s1 . This rather high rate
prevents microstructures from evolving and thereby favors a merely homogenous
solution even for the non-convex case considered here. The plastic slip dependent
free energy is taken, in analogy with phase field approaches, as a double-well
non-convex potential = 1.525 108 4 5.2 106 3 + 5 104 2 MPa. Note,
however that with respect to this item there is insufficient experimental data to specify the precise form and nature of this non-convex contribution. The spinodal points
are identified by 2 / 2 = 0 yielding sp1 = 0.0042 and sp2 = 0.0131. The binodal
points are obtained by extracting the points where the tangent line touches the free
energy curve, bp1 = 0.001 and bp2 = 0.0163 (see Fig. 4.3 (a)).
350
300
Stress (MPa)
250
200
150
100
50
0
0
0.005
0.01
0.015
0.02
Strain
Figure 4.3 / (a) Applied plastic slip dependent part of the free energy density function for the non-convex case (solid line) with spinodal (stars) and binodal (polygons)
points and for the convex case (dotted line). (b) Stress vs. strain response for a homogeneous deformation (solid line: non-convex case, dotted line: convex case).
For the given free energy ( shown in Fig. 4.3 (a)) and the other selected parameters
a homogeneous plastic slip distribution all along the bar is recovered. The resulting
stress vs. strain response for this particular case is shown in Fig. 4.3 (b). In Fig. 4.3
the effect of the convex plastic slip free energy density is also presented with material
parameters: C1 = 0.72 108 MPa, C2 = 2 106 MPa and C3 = 2.1 104 MPa. This
set of parameters leads to a small change in the free energy curve, however results
in a significantly different behavior in the stress-strain response, i.e. the convex case
does not show a softening branch. The peculiar behavior described by Fig. 4.3 (b)
at this scale can be related to Luders
62
2.5
Total
Plastic
Elastic
1.5
0.5
0
0
0.005
0.01
0.015
0.02
Strain
the idealized homogeneous plastic slip distribution case which is not easy to obtain
in experiments due to the fact that plastic deformation always imposes evolution of
dislocation slip microstructures. Therefore, stress vs. strain curves generally include
a plateau corresponding to microstructure evolution (see following examples).
The analytical solution for this case is obtained assuming a steady state, = 0 revealing dis = 0 and using the homogeneity condition A2 /x2 = 0. The slip equation
then simplifies to:
E( ) 4C1 3 3C2 2 2C3 = 0
(4.33)
Equation (4.33) is solved analytically for a set of strain values ranging from 0 to 0.02
which is a different way to recover the homogeneous solution. This solution is equal
to the homogeneous continuous loading case if the strain rates are not very high
(viscosity effects are minor). The contributing terms to the free energy densities are
plotted with respect to strain in Fig. 4.4. The obtained stress vs. strain response is
identical with the one from the finite element calculation (see Fig. 4.3 (b)). Within
the applied strain range, the total free energy also shows a double-well non-convex
behavior, which is an important prerequisite for patterning of the plastic slip.
Patterned solution and rate dependency
In this example, the rate dependent evolution of the plastic slip patterns during
monotonic loading is dealt with. The geometry and material parameters are identical to the ones in the previous example. First, results of hard and soft boundary
conditions are respectively presented in Figs. 4.5 and 4.6 at a strain rate of = 0.02
s1 which is low compared to the value used in the previous subsection. The effect
63
of the loading rate on the stress vs. strain response and slip patterning is shown in
Figs. 4.7 and 4.8.
400
Patterned solution
Homogenous solution
350
Stress (MPa)
300
250
200
150
100
50
0
0
0.005
0.01
0.015
0.02
0.025
Strain
Figure 4.5 / Stress vs. strain response and plastic slip evolution (legend presenting
the global strain) for a low monotonic loading rate with hard boundary conditions.
350
Patterned solution
Homogenous solution
300
Stress (MPa)
250
200
150
100
50
0
0
0.005
0.01
0.015
0.02
0.025
Strain
Figure 4.6 / Stress vs. strain response and plastic slip evolution (legend presenting
the global strain) for a low monotonic loading rate with soft boundary conditions.
In Figs. 4.5 (a) and 4.6 (a), the stress vs. strain curve is presented for the patterned solution together with the corresponding homogeneous solution. The stress response
64
shows a typical plateau for the values of strain where plastic slip patterning is observed. Due to the low values of the elastic strain the total strain in the stress vs.
strain curves can roughly be linked to the plastic slip values. The plastic slip limits of the plateau regime in the stress vs. strain curves coincide with the spinodal
points in the free energy and the distribution (decomposition) of the plastic slip during the inhomogeneous plastic slip evolution matches with the binodal points (see
Fig. 4.3). The constant stress plateau corresponds to the convexified solution of the
non-convex stress potential, where a linear convex envelope yields constant stress
values (see e.g. Lambrecht et al. (2003) and Miehe et al. (2004)).
The patterned solutions in 4.5 (a) and 4.6 (a) converge in the rate independent limit
to Maxwell-lines which can be associated with a global convexification of rate independent problems, such as discussed in Lambrecht et al. (2003). However, while
a convexification approach is able to resolve full Maxwell-lines similar to classical
treatments in phase-decompositions of real gases, the presented non-convex gradient
theory catches the peaks which can be directly related to lower and upper yield phenomena observed experimentally during the formation of microstructures in metals,
as in the case of Luders
band)
and the plateau can be linked to evolution or the movement of the band. However,
the relaxed stress obtained in the convexification approach represents only the perfectly plastic response.
Regarding the distribution of the plastic slip, in case of both hard and soft boundary conditions at low and at high strain levels (outside the binodal region) the same
stable behavior as in the case of convex strain gradient plasticity (see example 1) is
logically recovered. Hard boundary conditions typically trigger slip gradients at the
ends of the bar, resulting in characteristic boundary layers. Soft boundaries on the
other hand induce an initially homogeneous distribution of plastic slip. At strain
levels corresponding to the levels where the stress plateau is observed (between the
spinodal points) pronounced slip patterns develop depending on the boundary conditions (see Fig. 4.5 (b) and 4.6 (b)). This is consistent with the expectations based on
phase field modeling on the one hand, and the incremental minimization procedure
on the other hand. What is new with respect to the latter, is the fact that transitory
regimes are obviously captured as well, highlighting the role of the rate dependent
character of the model and the spinodal characteristic of the free energy. The stress
drop in Fig. 4.6 (a) is essentially due to the sudden (rate dependent) loss of stability
in the spinodal regime. This drop is sharp for the considered idealized case, and it
is not expected to occur if a more gradual loss of stability is invoked through the
intrinsic statistically non-homogeneous nature of the material.
65
The analysis in the present paper concerns single laminates only. Multiple laminates can be observed as in the energy-convexification methods, e.g. Lambrecht
et al. (2003) if a spatial fluctuation or random distribution is given to the material
parameters in the model. Departing from such fluctuations the obtained laminates
would coarsen in time, depending on the stabilizing gradient term in the free energy. Changes of the surface (gradient) term or introduction of higher gradient terms
(as in Cahn-Hilliard models) could give different response to a random fluctuations,
i.e. laminates may become more pronounced and/or stabilize at a particular size,
depending on the stabilizing (gradient) term in the slip evolution equation.
4000
1500
450
3500
400
350
2000
1500
1000
Stress (MPa)
2500
Stress (MPa)
Stress (MPa)
3000
500
1000
200
150
50
0.005
0.01
0.015
0.02
0
0
0.025
Strain
0.005
0.01
0.015
0.02
0
0
0.025
300
300
250
250
Stress (MPa)
250
200
150
100
50
0.01
0.015
Strain
(d) = 2 s1 .
0.02
0.025
Stress (MPa)
350
300
200
150
100
50
0.01
0.015
0.025
0.02
0.025
150
50
0.005
0.02
200
100
0
0
0.015
(c) = 20 s1 .
350
0.005
0.01
Srain
400
0
0
0.005
Strain
Stress (MPa)
250
100
500
0
0
300
0.02
Strain
(e) = 0.2 s1 .
0.025
0
0
0.005
0.01
0.015
Strain
(f) = 0.02 s1 .
The examples in Figs. 4.7 and 4.8 deal with the dependence of stress vs. strain response and the microstructure (plastic slip) evolution on the loading rate under soft
boundary conditions. A high loading rate (i.e. 2 s1 ) tends to inhibit the microstructure evolution leading to a homogenous distribution of plastic slip. The stress vs.
strain response corresponds to the steady state analytical solution presented in section 4.6.2. Higher loading rates (i.e. 20 s1 , 200 s1 , 2000 s1 ) result in a homogeneous plastic slip distribution with a more stiff stress vs. strain response, approaching the elastic limit behavior. Lower loading rates (i.e. 0.2 s1 , 0.02 s1 ) typically
result in patterned plastic slip distributions, which implies that the patterns evolve
only close to the rate independent limit. The slip pattern development as a function
66
of the deformation rate is shown in Fig. 4.8. The deformation is initially homogeneous and the slip patterns after passing the first spinodal point. This slip patterning evolves further with time and finally vanishes after passing the second spinodal
point. The transition from the homogeneous to the inhomogeneous slip distribution
is clearly rate dependent. For low loading rates (Fig. 4.8(b)), the transition from
the homogeneous to the patterned state (after the first spinodal point) and from the
patterned to the homogeneous state (after the second spinodal point) appears considerably faster compared to higher loading rates (Fig. 4.8(a)). This is most obvious
at an overall displacement of 7.2 m and a displacement equal to 14.8 m, where the
different loading rate cases clearly reveal different regimes.
u = 20m [t=0.1 s]
u = 20m [t=1 s]
u = 14.8m [t=0.074 s]
u = 14.8m [t=0.74 s]
u = 13.6m [t=0.068 s]
u = 13.6m [t=0.68 s]
u = 7.8m [t=0.0392 s]
u = 7.8m [t=0.392 s]
u = 7.2m [t=0.36 s]
u = 7.2m [t=0.036 s]
0.2
0.4
0.6
u = 4m [t=0.02 s]
u = 4m [t=0.2 s]
u = 0 [t=0]
u = 0 [t=0]
0.8
0.2
0.4
0.6
0.8
Figure 4.8 / Rate dependent microstructure evolution for the same imposed displacement.
67
0.2
0.4
0.6
0.8
0.2
Figure 4.9 / Stress relaxation test: stress vs. strain response and plastic slip evolution
with hard boundary conditions.
0.2
0.4
0.6
0.8
Figure 4.10 / Stress relaxation test: stress vs. strain response and plastic slip evolution with soft boundary conditions.
the same as used in section 4.6.2. The relaxation starts immediately after the displacement yielding an average overall strain equal to 0.0049 has been reached for the hard
boundary case and 0.0054 for the soft boundary case. The stress vs. strain response
and the evolution of the plastic slip are presented for hard and soft boundaries in
Figs. 4.9 and 4.10, respectively. Note that the soft boundary conditions lead to an initially homogeneous distribution, for which any perturbation may trigger patterning.
In order to stabilize this a small spatial fluctuation is applied to the Youngs modulus E along the bar, which restores uniqueness and triggers a stable evolution of the
68
4.7 Conclusion
Inspired by the efficiency of phase field models for microstructure formation and
evolution, we developed a non-convex strain gradient plasticity model in a system with energetic hardening-softening, which takes a conceptually dual structure
to Ginzburg-Landau type of phase field models with high and low slipped regions
representing different phases. Focusing on the basics and simplicity, the derivation
and the implementation are conducted in a 1D setting in order to illustrate the ability
of the model to capture the patterning of plastic slip similar to a phase decomposition mechanism. The destabilizing non-convex term in the free energy is stabilized
through the gradient term in the free energy and the viscous nature of the thermodynamically consistent slip law. Note that the resulting model typically yields a rate
dependent microstructure evolution. The framework can capture both homogeneous
and inhomogeneous deformations depending on the rate of the applied deformation.
The model is conceptually capable of covering the entire microstructure evolution
process, depending on the externally applied load and boundary conditions. Nonequilibrium microstructures are thereby well at reach.
The non-convexity in the presented model is incorporated by a double-well function,
not addressing particular materials yet. We present a generic formulation in order to
demonstrate the microstructure evolution in a thermodynamical and mathematical
rigorous setting. A more practical emphasis applied to a particular material is obviously needed in future work, provided reliable experimental data to recover the
non-convex term are available.
4.8 Appendix
4.8.1 Finite element implementation of slip based strain gradient plasticity
The weak forms of the equations are obtained in a standard manner, using a Galerkin
procedure. Firstly, the balance of linear momentum is tested with a field of virtual
69
4.8 Appendix
displacements u and integrated over the domain 0 < x < L, which results in Gu ,
Gu =
dx = 0
x
(4.34)
The slip equation (4.24) is tested with a field of virtual slips and integrated towards
G
m
Z L
n
n
0m
0m 2
G =
sign
A
dx = 0 (4.35)
t
t
s
s x2
0
Using integration by parts equation (4.34) can be written as
Gu =
L
0
du
dx + [u ]|0L = 0
dx
Cb [ ]|0L = 0
x
(4.36)
(4.37)
(4.38)
where Gu and G stand for the values of the previous estimate and with
u Gu =
Gu =
G
R L du
dx
dx
R L du
dx
dx
m
RL
n n
= 0 sign
m
dx
t
t
n
RL
0
u G =
C a
RL
0
d( )
dx + 0L Cb
dx
dx x
C a
dx
(4.39)
70
The equations (4.38) are discretized according to a finite element approach. Linear
interpolation functions N are used for the slip field , and quadratic interpolation
functions N u for the displacement field u.
u = J N JuuJ
u = K NKu uK
(4.40)
L NL L
M N M M
Substitution of the finite element interpolations into the linearized forms results in
the global system of coupled linear equations with element tangent matrices,
R
kuu =
k u
k
dN Ju dNKu
dx
Be J K
dx dx
dN Ju
= Be J M
N dx
dx M
m
R
n n
m
N M dx
= B e L M NL sign
t
t
n
R
dNL dN M
B e L M N L Ca N M
Cb
dx
dx
dx
R
k u =
Be
L K N L Ca
(4.41)
dNKu
dx
dx
where / = E, / = E due to = E( ) and B e is the element domain. Element residual vectors are calculated by using the values of and from
the previous estimate
dN Ju
dx
Be J
dx
m
R
n
n
sign
dx
= B e L M NL N M
t
t
R
dNL dN M
M
B e L NL Ca L M
Cb
dx
dx
dx
ru =
r
(4.42)
The assembly operation gives the global tangent and residual i.e.,
K uu = A {kuu }
K u = A { k u }
R u = A {r u }
K u = A {k u }
K = A {k }
R = A {r }
R + Rext
(4.43)
(4.44)
ext
where Rext
u and R originate from the boundary terms in the equilibrium (4.36) and
slip (4.37) weak forms.
71
4.8 Appendix
4.8.2 Finite element implementation of slip based non-convex strain gradient plasticity
Pursuing the same procedure as in the previous section, the weak form for equilibrium is written as
Gu =
L
0
du
dx [u ]|0L = 0
dx
d
(
sign
G = 0
Ca + Cb
dx
t
t
dx x
R
(4.45)
(4.46)
The variational functionals Gu and G are again solved in a fully coupled manner
(monolithic) by means of a Newton-Raphson scheme. The corresponding element
tangent matrices are identical to (4.41), except the k term which becomes,
m
R
n n
m
k = B e L M NL sign
N M dx
t
t
n
R
dNL dN M
Cb
B e L M NL Ca N M
(4.47)
dx
dx
2
+ NL N M Ca (12C1 + 6C2 + 2C3 ) dx
and r
r =
Be
Be
Be
n m
sign n
dx
t
t
dNL dN M
M
Cb
dx
L NL Ca L M
dx
dx
3
2
L NL Ca (4C1 + 3C2 + 2C3 + C4 ) dx
L M NL N M
(4.48)
which are calculated for the values of and from the previous estimate. After assembly the resulting system of equations in (4.44) is solved in a standard incremental
iterative manner.
72
Chapter five
5.1 Introduction
At the microscopic scale, deformed crystalline materials usually show heterogeneous
plastic deformation, where the amount of plastic strain varies spatially. At moderate
strain levels, regular cellular dislocation structures have been observed. Typical examples of dislocation microstructures are dislocation cells and dislocation walls (see
e.g. Rauch and Schmitt (1989), Gardey et al. (2005), Yalcinkaya et al. (2009)). Patterning typically refers to the self organization of dislocations, yielding regions with a
high dislocation density (dislocation walls) that envelop areas with a low dislocation
1
73
74
density (dislocation cell interiors), also regarded as domains of high and low plastic
slip activity, respectively. In addition to the cellular microstructures at meso-scale,
clear band formation and related plastic flow localization in irradiated materials (see
e.g. Sauzay et al. (2010)) at lower scales and macroscopic plastic slip bands such as
Luders
bands (see e.g. Shaw and Kyriakides (1998)) are also commonly observed
structures due to plastic deformation. These microstructures macroscopically e.g.
manifest themselves through softening of the material or through plastic anisotropy
under strain path changes.
The softening of the material and macroscopic anisotropic effects under strain path
changes (see e.g. Peeters et al. (2000), Yalcinkaya et al. (2009)), resulting from evolving dislocation microstructures, have been an interesting topic for the materials science community and the metal forming industry for decades. Starting with the studies on the cold-worked sub-structure of polycrystals using transmission electron microscopy in 1960s (e.g. Bailey and Hirsch (1960), Keh et al. (1963) and Swann (1963)),
a vast amount of experimental results have been collected, and several promising
theoretical models were presented dealing with dislocation (or slip) patterning. Nevertheless, a complete descriptive understanding of the occurring phenomena has not
been reached and the necessary input for computational models is still subject of
ongoing discussions.
In the context of the computational modeling of plastic slip pattering (or dislocation sub-structure formation), different approaches have been pursued in the literature which can be categorized into three main groups: (i) models using directly
the mechanics of single dislocations or populations of dislocations, (ii) phase field
modeling of dislocation patterning, (iii) the incremental variational formulation of
inelasticity by applying relaxation concepts for fine scale microstructure evolution.
See Yalcinkaya et al. (2011a) for a global overview of the available approaches.
The objective of the present paper is to develop a rate dependent strain gradient crystal plasticity finite element framework for the simulation of dislocation microstructure evolution, where the non-convexity is treated as an intrinsic property of the plastic free energy of the material. The constitutive model aims to simulate deformation
patterning and the macroscopic material behavior in a thermodynamically consistent
manner. Hence, the influence of latent hardening on the dislocation microstructure
evolution is studied through a physically based latent hardening potential proposed
by Ortiz and Repettto (1999). The physically based non-convex formulation is relevant in multi-slip deformation states, accounting for the interaction of slip systems.
In the model, the plastic slip and the displacement are taken as degrees of freedoms.
These fields are determined on a global level by solving simultaneously the linear
momentum balance and the slip evolution equation. The latter, expressed through a
thermodynamically consistent slip law is a crucial part of the model. At first glance,
75
the slip law (see equation (5.18)) is similar to the one encountered in classical rate
dependent crystal plasticity approaches (e.g. Hutchinson (1976), Peirce et al. (1982),
Yalcinkaya et al. (2008)), however the contribution of the actual stress state differs.
Three stress contributions can be distinguished: (1) the conventional resolved shear
stress directly related to the external loading, (2) an internal stress depending on the
gradient of the plastic slip, which is characteristic for strain gradient crystal plasticity models (e.g. Evers et al. (2004b), Yefimov et al. (2004), Bayley et al. (2006)) and
(3) the stress that emanates from the non-convex part in the free energy. The latter
contribution eventually triggers the patterning of the dislocation slip field.
The paper is organized as follows. First, in section 2, the rate dependent strain gradient crystal plasticity and its finite element solution is briefly discussed. Then, in
section 3, the incorporation of non-convexity into the model is presented, using a
physically based latent hardening potential. A detailed analysis of the latent hardening based non-convex function is performed in this section in order to clarify the
conditions enabling microstructure evolution. In section 4, numerical examples are
presented in order to demonstrate the capability of the proposed model. First, the
size effect related to plastic slip gradients and rate dependent deformation evolution
in the context of convex strain gradient crystal plasticity is studied. Then, the rate
dependent microstructure evolution via the physically motivated latent hardening
non-convex potential is addressed in this section. Finally, some concluding remarks
are given in section 5.
= e + p
(5.1)
into an elastic part e and a plastic part p . The plastic strain rate can be written as
a summation of plastic slip rates on the individual slip systems, p = P with
P = 12 (s n + n s ) the symmetrized Schmid tensor, where s and n are the
76
unit slip direction vector and unit normal vector on slip system , respectively. The
state variables are chosen to be given by the set
state = e , ,
(5.2)
where contains the plastic slips on the different slip systems and represents
the gradient of the slips on these slip systems. Following the arguments of Gurtin
(e.g. Gurtin (2000), Gurtin (2002)), the power expended by each independent ratelike kinematical descriptor is expressible in terms of an associated force consistent
with its own balance. However, the basic kinematical fields of rate variables, namely
e , u and are not spatially independent. It is therefore not immediately clear how
the associated force balances are to be formulated, and, for that reason, these balances
are established using the principle of virtual power.
e
Assuming that at a fixed time the fields u, e and are known, we consider u,
and as virtual rates, which are collected in the generalized virtual velocity V =
e , ). Pext is the power expended on the domain and Pint a concomitant
(u,
expenditure of power within , given by
Pext (, V) =
Pint (, V) =
t (n ) u dS +
e d +
:
( (n) ) dS
( )d +
(5.3)
( )d
where the stress tensor , the scalar internal forces and the microstress vectors
are the thermodynamical forces conjugate to the internal state variables e ,
and , respectively. In Pext , t (n ) is the macroscopic surface traction while
represents the microscopic surface traction conjugate to at the boundary S with n
indicating the boundary normal.
The principle of virtual power states that for any generalized virtual velocity V the
corresponding internal and external power are balanced, i.e.
Pext (, V) = Pint (, V)
(5.4)
Considering a generalized virtual velocity without slip, the virtual velocity field can
be chosen arbitrarily and this leads to the classical macroscopic force balance,
= 0
(5.5)
and considering arbitrary virtual slip fields without a generalized virtual velocity
leads to the microscopic force balances,
+ = 0
(5.6)
77
(5.7)
(5.8)
The material is assumed to be endowed with a free energy with different contributions according to
= e + +
(5.9)
The time derivative of the free energy is expanded and equation (5.8) is elaborated
to
e
)
e :
(5.10)
de
= ( e ) : e + ( ) + (
) 0
| {zd }
|
{z
}
D = : e + ( +
The stress and the microstress vectors are regarded as energetic quantities having no contribution to the dissipation
de
d e
(5.11)
( ) 0
(5.12)
The multipliers of the plastic slip rates are identified as the set of dissipative stresses
dis
dis =
(5.13)
In order to satisfy the reduced dissipation inequality at the slip system level the following constitutive equation is proposed
dis
= sign ( )
(5.14)
78
where represents the mobilized slip resistance of the slip system under consideration
s
= | |
(5.15)
0
with s is the resistance to dislocation slip which is assumed to be constant and 0 is
the reference slip rate. Substituting (5.15) into (5.14) gives
s dis
Substitution of dis
according to (5.13) into (5.16) reveals,
0
=
s
Using the microforce balance (5.6) results in the plastic slip equation
0
= +
s
(5.16)
(5.17)
(5.18)
(5.19)
1
A
2
where A is a scalar quantity, which includes an internal length scale parameter, governing the effect of the plastic slip gradients on the internal stress field. It may be
expressed as A = ER2 /(16(1 2 )) as e.g. used in Bayley et al. (2006) and Geers
et al. (2007), where R physically represents the radius of the dislocation domain contributing to the internal stress field, is Poissons ratio and E is Youngs modulus.
The plastic slip dependent free energy will be defined in the following section.
In order to solve the initial boundary value problem for this rate dependent strain
gradient crystal plasticity framework, a fully coupled finite element solution algorithm is used in which both the displacement u and plastic slips are considered
as primary variables. These fields are determined in the solution domain by solving
simultaneously the linear momentum balance (5.5) and the slip evolution equation
(5.18), which constitute the local strong form of the balance equations:
= 0
0 0
0
+
s
s
s
= 0
(5.20)
79
In order to obtain variational expressions representing the weak forms of the governing equations given above, these equations are multiplied by weighting functions u
and and integrated over the domain . Using the Gauss theorem (S is the boundary of ) results in
Gu =
G
u : d
u tdS
0
d +
s
0
d
s
0
A d
s
(5.21)
dS
0
where t is the external traction vector on the boundary S, and = A n.
s
The domain is subdivided into finite elements, where the unknown fields of the
displacement and slips and the associated weighting functions within each element
are approximated by their nodal values multiplied with the interpolation shape functions stored in the N u and N matrices, using a standard Galerkin approach.
u = N u u
u = Nuu
= N
(5.22)
= N
with u, u , and are columns containing the nodal variables. Bilinear interpolation functions for the slip field and quadratic interpolation functions for the displacement field are used. An implicit backward Euler time integration scheme is
used for in a typical time increment [tn , tn+1 ] which gives = [n+1 n ]/t.
The discretized element weak forms read
Z
Z
T
u
u
e
e
e
Gu = u
B d
N t dS
e
G e
"Z
Z
Se
n+1 n
t
0
A B T B de +
0 T
N de
s
0 T
N
de
s
Se
(5.23)
dSe
The weak form of the balance equations (5.23) are linearized with respect to the variations of the primary variables u and and solved by means of a Newton-Raphson
solution scheme for the increments of the displacement field u and the plastic slips
. The procedure results in a system of linear equations which can be written in
the following matrix format,
" uu
# "
#
"
#
u
Ru + Rext
K
K u
u
=
(5.24)
K u K
R + Rext
80
"
| | | |
a 0 0
# 3/ 4
(5.25)
Here 0 and 0 are a reference resolved shear stress and a reference slip value, respectively, and a are interaction coefficients. For the values of the matrix a , a simple
geometrical model is used proposed by Cuitino and Ortiz (1993),
2
=
1 (n n )2
(5.26)
where n and n are normals on the slip planes of the systems considered. Using
(5.26), slip systems do not self-harden in a multi-slip context. The reasoning leading
to (5.26) is based on the fact that the typical resolved shear stress required to deform a well-annealed crystal in single slip tends to be small compared to the stress
required for multi-slip. For the purpose of understanding the morphology of dislocation structures, self-hardening can be neglected at first instance. Ortiz and Repettto
(1999) and Ortiz et al. (2000) employ (5.25) and (5.26) in the context of crystal plasticity to obtain lamellar dislocation structures via a sequential lamination procedure.
The purpose here is to study patterning driven by this latent hardening based nonconvex potential.
81
0.012
0.01
0.012
0.01
0.008
0.008
0.006
0.004
0.006
0.004
0.002
0.002
0
0.06
0.06
0.04
0.04
0.02
0.02
0
0
0.06
0.04
0.02
0.06
0.05
0.04
0.03
0.02
0.01
Figure 5.1 / Two different views of (MPa) for different amounts of slip on two
slip systems oriented with respect to x axis as 60 and 120 where 0 = 1 and 0 = 1
MPa.
view, a patterned microstructure will develop if it has a lower energy than a state
with a homogeneous plastic deformation distribution. Yalcinkaya et al. (2011a) studied this aspect on a double-well potential and showed that plastic slip patterning
occurs at relatively low strain rates due to the existence of an unstable regime in
the double-well potential, where the microstructure evolves in a patterned way to
lower its free energy. Convex energy potentials preserve stability and do not trigger
patterning. Therefore, the presence of non-convexity is the first condition for a heterogeneous microstructure evolution to develop, yielding a lower energy than the
homogeneous state. In what follows, the latent hardening function (5.25), which is
assumed to be the driving force for the evolution of dislocation microstructures, is
examined in this sense.
For this purpose, a single crystal in 2D having 2 slip systems oriented 1 and 2
with respect to x axis is considered. A pure shear deformation is applied with a
macroscopic strain field M , whose components are written in Cartesian coordinates
as
"
#
0
M =
(5.27)
0
which satisfies the plastic incompressibility condition tr ( M ) = 0. Assume that the
plastic deformation is patterned in the solution domain and that there exist two states
with volume fraction f and 1 f . The weighted average strain of the two phases
should be equal to the macroscopic strain, i.e.
f 1 + (1 f ) 2 = M
(5.28)
82
1 = 2 =1 1 P
2 = 2 =1 2 P
(5.29)
with both patterned regions having the same crystal orientation and hence the same
P . The strain tensors, 1 = 1 (11 , 12 , 1 , 2 ) and 2 = 2 (21 , 22 , 1 , 2 ) are symmetric
and incompressible. Using (5.29), equation (5.28) reduces to a set of 2 linear equations with 5 unknowns 11 , 12 , 21 , 22 , f . For a fixed value of f and given two of the
unknown slips 11 and 12 the value of the other plastic slips 21 and 22 can be calculated according to
21 =
22
f
P21
1
11 +
1 f
P22 P11 P12 P21 1 f
1
f
P11
=
12 +
1 f
P22 P11 P12 P21 1 f
(5.30)
with P11 = P1 (1, 1), P12 = P1 (1, 2), P21 = P2 (1, 1), P22 = P2 (1, 2). The energy of this
assumed patterned state can be calculated as,
= f 1 (11 , 12 , 1 , 2 ) + (1 f )2 (21 , 22 , 1 , 2 )
(5.31)
which is plotted in Fig. 5.2. The latent hardening energy corresponding to the macroscopic shear strain M is calculated via (5.25) in terms of the plastic slips 1M and 2M
on given slip systems as,
= ( 1M , 2M , 1 , 2 )
(5.32)
The purpose of this analysis is to investigate if there exists a domain of a lower latent
hardening based energy for the assumed patterned case than the macroscopic
latent hardening based energy (which reflects the homogeneous non-patterned
state). To this end, is calculated for different values of 11 and 12 , and for specific
(chosen) orientations and volume fractions.
If a macroscopic shear deformation tensor M with = 0.02 is applied on a combinations of slip systems with orientations 1 = 60 and 1 = 120 , the amount of
slips on the two slip systems for the local homogeneous state equals 0.04. Using
these values for the slips and the orientations, the locally homogeneous latent hardening plastic potential is calculated via Equation (5.25). Assuming 0 = 1 MPa and
0 = 1 the characterizing latent hardening energy for the non-patterned state equals
= 0.0057 MPa. Presuming the existence of patterned states for different values of
f and (11 , 12 ), the values of 21 and 22 are calculated via (5.30) and the latent hardening energy follows from (5.31). In order to obtain a patterned microstructure the
83
f = 0.1
x 10
f = 0.2
x 10
x 10
x 10
6.4
6.5
6
6.2
6
6.5
6
5.8
5.6
5.5
5.5
5.8
5.6
5.4
5.4
4.5
5.2
5.2
4.5
4
5
5
0.1
0.08
0.06
11
0.04
0.02
0.05
0.1
0.1
12
f = 0.3
x 10
11
0.05
x 10
12
0.1
x 10
12
0.014
7.5
7
0.05
f = 0.5
0.012
10
6.5
0.01
5.5
5
0.008
3.5
4.8
0.006
4.5
4
0.004
3.5
0.002
3
0
2
0.1
11
0.05
12
0.05
0.1
f = 0.6
2.5
0.1
11
0.05
16
0.04 2 0.06
1
0.08
0.1
f = 0.7
x 10
0.018
0.02
0.022
0.025
0.02
0.016
14
0.018
0.02
0.014
12
0.016
10
0.01
0.008
0.006
0.014
0.015
0.012
0.012
0.01
0.01
0.004
0.008
0.006
0.005
0.004
0.002
2
0
0.002
0
0.1
11
0.05
0.05
0.1
0.1
12
11
0.05
f = 0.8
0.05
12
0.1
f = 0.9
0.05
0.045
0.045
0.04
0.03
0.035
0.03
0.025
0.025
0.04
0.035
0.035
0.03
0.02
0.02
0.015
0.015
0.01
0.01
0.03
0.025
0.025
0.02
0.02
0.015
0.015
0.01
0.005
0.005
0.01
0.005
0.005
0
0.1
11
0.05
0.05
0.1
12
0.1
0.05
11
0.05
0.1
12
Figure 5.2 / (MPa) in terms of a set of given 11 and 12 for different values of f
with 1 = 60 and 1 = 120 .
84
energy of the patterned state should be smaller than the energy of the local homogeneous state < . As seen in Fig. 5.2 there are many combinations satisfying this
inequality for different values of the volume fraction f and the plastic slip values 11
and 12 .
85
900
= 0.25s1
140
800
120
700
600
500
400
300
100
80
60
40
200
20
100
0
0.005
0.01
0.015
0.02
3.5
12
0.015
0.02
= 0.0025s1
14
10
8
6
4
2
0
0.01
= 0.025s1
16
0.005
2.5
2
1.5
1
0.5
0.005
0.01
0.015
0.02
0.005
0.01
0.015
0.02
Figure 5.3 / Rate dependent shear stress vs. the applied shear for the plane strain
shear problem of an infinite strip.
The examples addressed in this subsection are carried out for different applied shear
rates and R/ H ratios for varying R and constant height H, where the relative effect
of the internal length scale parameter R determining A, acting on the higher order
microstresses (see equations (5.11) and (5.19)), is analyzed.
First, the resulting shear stress versus the applied macroscopic shear is presented
in Fig. 5.3 for different overall shear rates , using R = 0.35 m, and H = 20 m.
is the macroscopic shear defined as = u/ H. In this case, the average of the local
strain 12 should be half of the macroscopic shear . If the elastic deformation is
p
small, the average value of the local plastic shear strain 12 will be roughly equal to
half of the macroscopic shear as well. When the applied macroscopic shear rate
increases, the material shows a stiffer (dominantly elastic) response. In Fig. 5.4, the
local rate dependent plastic shear strain evolution is presented at a macroscopic shear
level of = 0.02. There is a clear boundary layer width dependence on the applied
shear rate. For high values of the applied shear rate, e.g. = 2.5s1 we observe
a sharp boundary layer, while for low values of the shear rate, e.g. = 0.0025s1
86
the boundary layer thickness is more diffuse. This type of rate dependent dislocation
slip profile is also observed in Yalcinkaya et al. (2011a) where a high strain rate causes
that the plastic slip has not enough time to evolve.
= 2.5s1
= 0.25s1
3
x 10
4.5
x 10
9
4
8
3.5
7
3
2.5
1.5
0.5
= 0.025s1
= 0.0025s1
3
x 10
12
x 10
10
12
14
10
8
6
6
4
4
2
Figure 5.4 / Rate dependent plastic shear strain 12 distribution for the plane strain
shear problem of an infinite strip at a global shear level of = 0.02.
The next example in this subsection concerns the influence of the internal length scale
parameter R on the mechanical behavior for a fixed height H = 20 m. The value of R
is taken as 0.35 m, 0.7 m and 1.75 m corresponding to a R/ H ratio equal to 0.0175,
0.035 and 0.0875 respectively. The shear stress versus the applied macroscopic shear
response is plotted in Fig. 5.5 for different values of R at = 0.025s1 . Note that
the formulation is intrinsically viscous, corresponding to models using a linear drag
law for dislocation motion to determine the slip. Fig. 5.5 illustrates a significant
effect of the internal length parameter where the strip is exhibiting a stiffer response
for larger values of R. The results are consistent with many strain gradient models
(e.g. Shu et al. (2001), Evers et al. (2004b), Yalcinkaya et al. (2011a)) in which the
influence of the length scale was shown. Physically, this effect results from the fact
that large R values induce a large internal stress and hence penalize high plastic slip
87
gradients (see Fig. 5.6) spreading the geometrically necessary dislocation densities.
In this convex example, the slip gradient is the dominating factor in the plastic and
hardening behavior of the material and therefore a clear size effect is observed. In
Fig. 5.6 a clear dependence of the boundary layer evolution on R is presented at
= 0.02. With increasing R an increased influence of the boundary conditions is
observed with a more diffuse boundary layer.
70
R = 0.35 m
R = 0.7 m
R = 1.75 m
60
50
40
30
20
10
0.005
0.01
0.015
0.02
Figure 5.5 / The effect of the internal length scale parameter R on the stress vs.
applied shear response for the plane strain shear problem of an infinite strip.
R = 0.7m
R = 0.35m
x 10
12
R = 1.75m
x 10
x 10
14
12
10
10
8
12
10
8
6
4
2
Figure 5.6 / The effect of the internal length scale parameter R on the plastic shear
p
strain 12 distribution for the plane strain shear problem of an infinite strip at a global
shear level of = 0.02.
The distribution of the plastic slips on each of the slip systems is illustrated in Fig. 5.7
for the same example with R = 0.35 m and = 0.025s1 at = 0.02. The amount
88
= 120
0.0121
0.0121
0.0242
0.0242
Figure 5.7 / The distribution of the plastic slip on both slip systems at a global shear
level of = 0.02.
of the slip on both slip systems is identical and the evolution is similar to the plastic
shear strain evolution, exhibiting a boundary layer.
89
in the context of the present small strain rate dependent crystal plasticity formulation. Compared to the previous work of Ortiz and Repettto (1999) and Ortiz et al.
(2000) relying on the explicit construction of cellular dislocation microstructures via
a lamination procedure, the present framework is based on the implicit evolution of
microstructures driven by the deformation. Throughout the incremental deformation process the state of the plastic slip enters energetically favorable regimes, (as
illustrated in section 5.3) which might eventually result in deformation heterogeneity.
The numerical study in this subsection concerns a plane strain pure shear problem of
a square representative volume element (RVE) in order to have a direct link with the
study in section 5.3. Locally, two slip systems with orientations 60 and 120 with
respect to the x axis are considered. The material is assumed to be elastically isotropic
with Youngs modulus E = 210 GPa, Poissons ratio = 0.33, slip resistance for both
slip systems s = 35 MPa, and a reference slip rate 0 = 0.15 s1 . The reference slip
strain and resolved shear stress in (5.25) are assumed to be 0 = 0.015 and 0 = 50
MPa, respectively.
In the square RVE the displacements and the plastic slips on the left edge are tied to
the ones on the right edge and the ones on the bottom edge are tied to the ones on the
top edge which makes it a fully periodic configuration. The vertical displacement at
the right bottom corner and the horizontal displacement at the left top corner are prescribed, both equal to u(t). The displacements at left bottom corner are suppressed
together with the horizontal displacement at the right bottom corner and vertical displacement at left top corner. These boundary and loading conditions result in a pure
shear deformation mode. The length of each edge of the square is H = 20 m. Results presented in the following correspond to a discretization with 20x20 rectangular
elements.
In Fig. 5.8 the shear stress versus applied macroscopic shear, defined as = 2u/ H,
is plotted for an applied shear rate = 0.005s1 and an internal length scale parameter R = 0.01 m. In this case, the average of the local strain 12 should be half of
the macroscopic shear . If the elastic deformation is small, the average value of the
p
local plastic shear strain 12 will be roughly equal to half of the macroscopic shear
as well. The corresponding shear strain, plastic shear strain and plastic slips of
the slip systems are plotted in Fig. 5.9. All plotted fields exhibit a strong patterned
response due to the incorporated latent hardening potential. Note that the applied
periodic boundary conditions lead to an initially homogeneous distribution of the
strain and plastic slip fields. In order to properly trigger a small perturbation a small
spatial fluctuation is applied to the Youngs modulus E in the RVE. In order to illustrate the influence of the selected spatial fluctuation of E on the evolution of the
plastic microstructure, two different types of fluctuations are applied in the next two
90
80
70
60
50
40
30
20
10
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
Figure 5.8 / Latent hardening based non-convex shear stress vs. applied macroscopic shear for a plane strain pure shear problem of a fully periodic RVE.
p12
12
0.022
0.021
0.021
0.02
0.02
0.019
0.019
0.018
0.018
0.017
0.017
0.016
0.016
0.015
0.015
0.014
0.014
0.013
0.013
0.012
= 60
= 120
0.01
0.01
0.015
0.02
0.02
0.025
0.03
0.03
0.035
0.04
0.04
0.045
0.05
0.05
0.055
Figure 5.9 / Shear strain (first figure), plastic shear strain (second figure), and plastic
slip (last two figures) distributions at = 0.034 for a plane strain pure shear problem
of a fully periodic RVE.
examples. In Fig. 5.10 the field quantities are plotted for two additional cases with
different fluctuations applied in Youngs modulus. First, the fluctuation is given only
91
p12
12
0.02
0.02
0.019
0.019
0.018
0.018
0.017
0.017
0.016
0.016
0.015
0.015
0.014
0.014
0.013
= 60
= 120
0.015
0.015
0.02
0.02
0.025
0.025
0.03
0.03
0.035
0.035
0.04
0.04
0.045
0.045
0.05
0.05
p1 2
12
0.02
0.02
0.019
0.019
0.018
0.018
0.017
0.017
0.016
0.016
0.015
0.015
0.014
0.014
0.013
= 60
= 120
0.015
0.015
0.02
0.02
0.025
0.025
0.03
0.03
0.035
0.035
0.04
0.04
0.045
0.045
0.05
0.05
Figure 5.10 / Shear strain, plastic shear strain, and plastic slip distributions at =
0.034 for a plane strain pure shear problem of a fully periodic RVE with a fluctuation
applied to the central element (first four figures) and to the middle element on the
right edge (last four figures).
92
to an element in the center of the RVE and in the second case the fluctuation is given
to an element in the middle of the right edge. The applied fluctuation has certainly
an effect on the distribution of the deformation patterns, however, the spacing of
high and low strain areas and the amplitude do not depend on the applied fluctuation. The macroscopic stress versus strain diagram is not plotted in these two cases
because it is identical to the previous case plotted in Fig. 5.8.
It is remarked that the obtained deformation patterns depend considerably on the
applied rate of deformation. Increasing the strain rate results in a stiffer stress versus
strain response while the amplitude of the obtained patterns decreases. Another
important parameter affecting the patterning of deformation fields is the internal
length scale R which should be small enough to obtain a stable numerical solution
and pronounced patterns. These two factors and other material parameters play
an important role in the convergence of the numerical solution as well. Due to the
non-convexity of the latent hardening potential at each increment of the deformation
some combinations of the parameters might give convergence problems and it is not
always possible to reach the same state of deformation with different combinations
of the rate of deformation and material parameters.
As stated before the non-convex potential has been used to recover a cellular kind of
dislocation microstructures by Ortiz and Repettto (1999) and Ortiz et al. (2000) via
an external lamination procedure, while we try to obtain deformation patterns via
a non-convex evolution problem. In the current examples we observe deformation
patterning at low loading rates and a small internal length scale parameter due to
the latent hardening based non-convexity. This shows agreement with the study in
section 5.3 illustrating the capability of the latent hardening function for deformation
patterning.
93
94
Chapter six
Abstract / The main conclusions drawn from the thesis are summarized in this concluding chapter and an outlook to further work related to microstructure evolution in crystal
plasticity models is presented.
The objective of this thesis was the investigation of the macroscopic transient hardening and softening effects due to the dislocation microstructure evolution in BCC metals. The project started with the development of a large strain BCC crystal plasticity
framework where the intrinsic anisotropy due to non-planar spreading of screw dislocation cores was addressed. Next, the evolution of an existing dislocation cell structure has been investigated in a composite cell model which was incorporated into
the BCC crystal plasticity model. This framework starts with an assumed dislocation
structure and studies the evolution of this microstructure via phenomenological dislocation evolution equations. However, the formation of dislocation microstructures
or deformation patterning cannot be simulated by these kind of models. Therefore,
in order to model the formation and evolution of inhomogeneous deformation patterns, a non-convex rate dependent strain gradient crystal plasticity model has been
developed in a small strain context in the last part of the thesis.
The BCC crystal plasticity framework, presented in Yalcinkaya et al. (2008) (chapter
2), reveals some unique characteristics of BCC crystals. A comprehensive summary
of intrinsic properties of these materials are presented, including recent insights in
the activation of different slip systems, violation of Schmids law, temperature and
orientation dependence of the flow stress and resulting stress-strain curves. With
respect to the BCC crystal plasticity model the following conclusions are drawn:
The applied slip evolution equation which is based on the thermally activated
dislocation kinetics plays an essential role in the model since all the intrinsic
characteristics result from the actual formulation of this equation.
95
96
Contrary to the common assumption of using the {110}, {112} and {123} type
of slip planes in all conditions, it is explained that the activation of slip systems
depends on the temperature. At moderate temperatures only {112} slip planes
are activated which affect considerably the mechanical behavior of BCC metals.
The non-Schmid behavior is introduced in the slip law by modifying the effective shear stress, where the non-Schmid contribution represents the dislocation
cores spreading in a non-planar manner.
Deviation from a proportional strain path is associated with hardening or softening
of the material due to the induced plastic anisotropy. At moderate strains the dominating effect is attributed to the evolving underlying dislocation microstructures.
Chapter 3 (Yalcinkaya et al. (2009)) deals with a combination of a composite dislocation cell model, which explicitly describes the dislocation structure evolution, with
the BCC crystal plasticity framework from chapter 2 to bridge the microstructure
evolution and its macroscopic anisotropic effects. The main conclusions from this
part are:
A phenomenological cell structure evolution model embedded into a crystal
plasticity framework is well able to reproduce all essential characteristics of
strain path changes reported, consistently with experimental observations at
two scales.
The model proposed allows to study the interaction between different sources
of anisotropy, where a clear example at the single crystal and polycrystal has
been given.
The level at which the enrichment of the crystal plasticity model was made,
enables its use in more complex microstructures as e.g. multi-phase steels.
In order to complete the missing link between the formation of the microstructure
and its evolution in crystal plasticity frameworks, the second part of the thesis
concentrated on the development of a non-convex rate dependent crystal plasticity model, which may eventually simulate rate dependent dislocation microstructure formation and evolution together with macroscopic hardening-softening stressstrain responses. In chapter 4 (Yalcinkaya et al. (2011a)), inspired by the efficiency
of phase field models for microstructure formation and evolution, a 1D non-convex
strain gradient plasticity model has been developed yielding combined hardeningsoftening. The resulting formulation takes a conceptually dual structure to the
Ginzburg-Landau type of phase field models where high and low slipped regions
represent different phases. The main conclusions related to this part are:
97
The thermodynamically consistent constitutive model is able to capture patterning of plastic slip, qualitatively similar to a phase decomposition mechanism.
The mathematical structure of the employed double-well plastic potential allows to control the amount and the timing of deformation patterning.
The framework can capture both homogeneous and inhomogeneous deformations depending on the rate of the applied deformation.
The model is conceptually capable of covering many aspects of microstructure
evolution processes, depending on the externally applied load and boundary
conditions. Non-equilibrium microstructures are thereby well at reach.
The model could be used for various materials. Especially the clear band formation observed in irradiated materials or Luders
98
introduced externally, in the current approach it becomes a natural product. Moreover, compared to incremental variational principles and relaxation theories for microstructure evolution it is simpler to implement and computationally less expensive. Capturing both equilibrium and non-equilibrium states of the microstructures
makes it a unique framework in computational plasticity. The crucial aspect in the
present models is the plastic potential entering the formulation in a thermodynamically consistent manner.
In this thesis two different plastic potentials have been formulated. The double-well
based non-convexity in the fourth chapter presents a phenomenological description
of plastic slip patterning. The coefficients of the potential and the rate of deformation control the whole patterning mechanism. Even though the material parameters are not identified for real materials it offers a broad range of application. A
more practical emphasis applied to a particular material is obviously needed in future work, provided reliable experimental data to recover the non-convex term are
available. On the other hand, the latent-hardening based non-convex potential used
in the last chapter addresses a physical phenomenon, i.e. slip interactions in crystals.
It induces a clear implicit deformation patterning in the present non-convex strain
gradient crystal plasticity framework, compared to its previous usage for external
microstructure evolution via lamination procedure. A more detailed study on the
available functions and possible other descriptions of the non-convexity for specific
materials is obviously needed as a future work.
Bibliography
A CKERMANN , F., M UGHRABI , H., and S EEGER , A. (1983). Temperature and strainrate dependence of the flow stress of ultrapure niobium single crystals in cyclic
deformation. Acta Metall., 31, 13531366.
A DAMS , J. J., A GOSTA , D. S., L EISURE , R. G., and L EDBETTER , H. (2006). Elastic
constant of monocrystal iron from 3 to 500 K. J. Appl. Phys., 100, 113530.
A IFANTIS , E. C. (1987). The Physics of plastic deformation. Int. J. Plast., 3, 211247.
A RGON , A. S. (2008). Strengthening mechanisms in crystal plasticity. Oxford University Press.
A SARO , R. and N EEDLEMAN , A. (1985). Texture Development and Strain Hardening
in Rate Dependent Polycrystals. Acta Metall., 33, 923953.
A SARO , R. J. and R ICE , J. R. (1977). Strain localization in ductile single crystals. J.
Mech. Phys. Solids, 25, 309338.
B ACROIX , B. and H U , Z. (1995). Texture evolution induced by strain path changes
in low carbon steel sheets . Metall. Mater. Trans. A, 26, 601613.
B ACROIX , B., G ENEVOIS , P., and C., T EODOSIU (1994). Plastic anisotropy in low
carbon steels subjected to simple shear with strain path changes. Eur. J. Mech.
A/Solids, 13, 661675.
B AILEY , J. E. and H IRSCH , P. B. (1960). The dislocation distribution, flow stress, and
stored energy in cold-worked polycrystalline silver. Philos. Mag., 5, 485.
B ARDELLA , L. (2006). A deformation theory of strain gradient crystal plasticity that
accounts for geometrically necessary dislocations. J. Mech. Phys. Solids, 54, 128160.
B ARDELLA , L. (2007). Some remarks on the strain gradient crystal plasticity modelling, with particular reference to the material length scales involved. Int. J. Plast.,
23, 296322.
B ARKER , I., H ANSEN , N., and R ALPH , B. (1989). The development of deformation
substructures in face-centered cubic metals. Mat. Sci. Eng. A, 113, 449454.
99
100
Bibliography
B ARLAT , F., F ERREIRA D UARTE , J. M., G RACIO , J. J., L OPES , A. B., and R AUCH ,
E. F. (2003). Plastic flow for non-monotonic loading conditions of an aluminum
alloy sheet sample. Int. J. Plast, 19, 12151244.
B ARRETT , C. S., A NSEL , G., and M EHL , R F (1937). Slip, twinning and cleavage in
iron and silicon ferrite. Trans. Am. Soc. Met., 25, 702733.
B ASSANI , J. L. (1994). Plastic flow of crystals. Adv. Appl. Mech., 30, 191257.
B ASSANI , J. L., I TO , K., and V ITEK , V. (2001). Complex macroscopic plastic flow
arising from non-planar dislocation core structures. Mat. Sci. Eng. A, 319-321, 97
101.
B AYLEY , C. J., B REKELMANS , W. A. M., and G EERS , M. G. D. (2006). A comparison
of dislocation induced back stress formulations in strain gradient crystal plasticity.
Int. J. Solids Struct., 43, 72687286.
B OLEF , D. I. and K LERK , J. D. (1962). Elastic Constants of Single-Crystal Mo and W
between 77 and 500 K. J. Appl. Phys., 33, 23112314.
B OUVIER , S., G ARDEY , B., H ADDADI , H., and C., T EODOSIU (2006). Characterization of the strain-induced plastic anisotropy of rolled sheets by using sequences of
simple shear and uniaxial tensile tests. J. Mater. Process. Tech., 174, 115126.
B RONKHORST , C. A., K ALIDINDI , S. R., and A NAND , L. (1992). Polycrystalline
plasticity and the evolution of crystallographic texture in FCC metals. Philos. T.
Roy. Soc. A, 341, 443477.
B ROWN , S. B., K IM , K. H., and A NAND , L. (1989). An internal variable constitutive
model for hot working of metals. Int. J. Plast., 5, 95130.
B RUNNER , D. and D IEHL , J. (1987). The use of stress-relaxation measurements for
investigations on the flow stress of alpha-iron. Phys. Stat. Sol. A, 104, 145155.
B RUNNER , D. and D IEHL , J. (1997). The effect of atomic lattice defects on the softening phenomena of high-purity alpha-iron. Phys. Stat. Sol. A, 160, 355372.
C AHN , J. W. (1961). On spinodal decomposition. Acta Metall., 9, 795801.
C ARROLL , K. J. (1965). Elastic Constants of Niobium from 4.2 to 300 K. J. Appl. Phys.,
36, 36893690.
C ERMELLI , P. and G URTIN , M. E. (2001). On the characterization of geometrically
necessary dislocations in finite plasticity. J. Mech. Phys. Solids, 49, 15391568.
C HEN , N. K. and M ADDIN , R. (1954). Slip planes and energy of dislocations in a
body centered cubic structure. Acta Metall., 2, 4951.
C HRISTIAN , A., K ANERT, O., and D E H OSSON , J. T H . M. (1990). Dislocation dynamics in Vanadium: A nuclear magnetic and transmission electron microscopic
study. Acta Metall. Mater., 38, 24792484.
Bibliography
101
102
Bibliography
G ARDEY , B., B OUVIER , S., R ICHARD , V., and B ACROIX , B. (2005). Texture and
dislocation structures observation in a dual-phase steel under strain-path changes
at large deformations. Mater. Sci. Eng., A, 400-401, 136141.
G EERS , M. G. D., B REKELMANS , W. A. M., and B AYLEY , C. J. (2007). Second-order
crystal plasticity: internal stress effects and cyclic loading. Model. Simul. Mater. Sci.
Eng., 15, 133145.
G OUGH , H. J. (1928). The behavior of a single crystal alpha-iron subjected to alternating torsional stresses. P. Roy. Soc. Lond. A, 118, 498534.
G R OGER
, R. and V ITEK , V. (2005). Breakdown of the Schmid law in BCC molybdenum related to the effect of shear stress perpendicular to the slip direction. Mater.
Sci. Forum, 482, 123.
G ROMA , I. (1997). Link between the microscopic and mesoscopic length scale description of the collective behavior of dislocations. Phys. Rev. B, 56, 58075813.
G ROMA , I. and PAVLEY , G. S. (1993). Role of the secondary slip system in a computer
simulation model of the plastic behaviour of single crystals. Mater. Sci. Eng. A, 164,
306311.
G UIU , F. (1969). Slip asymmetry in molybdenum single crystals deformed in direct
shear. Scr. Metall., 3, 449454.
G URTIN , M. E. (2000). On the plasticity of single crystals: free energy, microforces,
plastic-strain gradients. J. Mech. Phys. Solids, 48, 9891036.
G URTIN , M. E. (2002). A gradient theory of single-crystal viscoplasticity that accounts for geometrically necessary dislocations. J. Mech. Phys. Solids, 50, 532.
G URTIN , M. E., A NAND , L., and L ELE , S. P. (2007). Gradient single-crystal plasticity
with free energy dependent on dislocation densities. J. Mech. Phys. Solids, 55, 1853
1878.
H ACKL , K. and K OCHMANN , D. M. (2008). Relaxed potentials and evolution equations for inelastic microstructures. IUTAM Symposium on Theoretical, Computational
and Modelling Aspects of Inelastic Media, 2741.
H AHNER
, P. (1994). Theory of solitary plastic waves. Part I: Luders
Bibliography
103
104
Bibliography
K UHLMANN -W ILSDORF , D. (2001). Q: Dislocation structures - how far from equilibrium? A: Very close indeed. Mater. Sci. Eng. A, 315, 211216.
K UHLMANN -W ILSDORF , D. and VAN DER M ERWE , J. H. (1982). Theory of dislocation cell sizes in deformed metals. Mater. Sci. Eng., 55, 7983.
L AMBRECHT , M., M IEHE , C., and D ETTMAR , J. (2003). Energy relaxation of nonconvex incremental stress potentials in a strain-softening elastic-plastic bar. Int. J.
Solids Struct., 40, 13691391.
L ANGER , J. S. (1971). Theory of spinodal decomposition in alloys. Ann. Phys., 65,
5386.
L EE , E. H. (1969). Elastic-plastic deformation at finite strains. J. Appl. Mech., 36, 16.
L EE , Y. J., S UBHASH , G., and R AVICHANDRAN , G. (1999). Constitutive modeling of
textured body-centered-cubic (BCC) polycrystals. Int. J. Plast., 15, 625645.
L I , Z. J., W INTHER , G., and H ANSEN , N. (2006). Anisotropy in rolled metals induced by dislocation structure. Acta Mater., 54, 401410.
L IAO , K. C., F RIEDMAN , P. A., and PAN , J. (1998). Texture development and plastic
anisotropy of BCC strain hardening sheet metals. Int. J. Solids Struct., 35, 5205
5236.
L UBARDA , V. A., B LUME , J. A., and N EEDLEMAN , A. (1993). An Analysis of Equilibrium Dislocations Distributions. Acta Metall. Mater., 41, 625642.
M IEHE , C. (2002). Strain-driven homogenization of inelastic microstructures and
composites based on an incremental variational formulation. Int. J. Numer. Methods
Eng., 55, 12851322.
M IEHE , C., S CHOTTE , J., and L AMBRECHT, M. (2002). Homogenization of inelastic
solid materials at finite strains based on incremental minimization principles. Application to the texture analysis of polycrystals. J. Mech. Phys. Solids, 50, 21232167.
Bibliography
105
106
Bibliography
S AUZAY , M., B AVARD , K., and K ARLSEN , W. (2010). TEM observations and finite
element modelling of channel deformation in pre-irradiated austenitic stainless
steels Interactions with free surfaces and grain boundaries. J. Nucl. Mater., 406,
152165.
S CHMID , E. and B OAS , W. (1935). Kristallplastizitat. Springer.
S CHMITT , J. H., F ERNANDES , J. V., G RACIO , J. J., and F., V IEIRA M. (1991). Plastic
behavior of copper sheets during sequential tension tests. Mat. Sci. Eng. A, 147,
143154.
S EEGER , A. (1981). Temperature and strain-rate dependence of the flow stress of
body-centered cubic metals. A theory based on kink-kink interactions. Z. Metallkd.,
72, 369380.
S EEGER , A. (2001). Why anomalous slip in body-centered cubic metals? Mat. Sci.
Eng. A, 319-321, 254260.
S EEGER , A. and WASSERBACH , W. (2002). Anomalous slip - A feature of high-purity
body-centered cubic metals. Phys. Stat. Sol. A, 189, 2750.
S EVILLANO , J. G., VAN H OUTTE , P., and A ERNOUDT , E. (1981). Large strain work
hardening and textures. Prog. Mater Sci., 25.
S HAW , J. A. and K YRIAKIDES , S. (1998). Initiation and propagation of localized
deformation in elasto-plastic strips under uniaxial tension. Int. J. Plast., 13, 837
871.
S HU , J. Y., F LECK , N. A., VAN DER G IESSEN , E., and N EEDLEMAN , A. (2001).
Boundary layers in constrained plastic flow: comparison of nonlocal and discrete
dislocation plasticity. J. Mech. Phys. Solids, 49, 13611395.
S PITZIG , W. A. (1979). Effect of hydrostatic pressure on plastic properties - flow
properties of iron single crystals. Acta Metall., 27, 523534.
S UN , H. B., Y OSHIDA , F., M A , X., K AMEI , T., and O HMORI , M. (2003). Finite element simulation on the propagation of Luders
Bibliography
107
TAYLOR , G. I. and E LAM , C. F. (1926). The distortion of iron crystals. P. Roy. Soc.
Lond. A, 112, 337361.
T EODOSIU , C. (1992). Materials Science Input to Engineering Models. Modeling of
Plastic Deformation and its Engineering Applications (Proc. 13th Riso), 125146.
T EODOSIU , C. and H U , Z. (1995). Simulation of Materials Processing: Theory and Applications (Proc. of NUMIFORM 95), 173182.
U BACHS , R. L. J. M., S CHREURS , P. J. G., and G EERS , M. G. D. (2004). A nonlocal phase field model for microstructure evolution of Sn-Pb solder. J. Mech. Phys.
Solids, 52, 17631792.
V IATKINA , E. (2005). Micromechanical modelling of strain path dependency in FCC
metals. Ph.D. thesis, Eindhoven University of Technology.
V IATKINA , E. M., B REKELMANS , W. A. M., and G EERS , M. G. D. (2003). Strain path
dependency in metal plasticity. J. Phys. IV, 105, 355362.
V ITEK , V. (2004). Core structure of screw dislocations in body-centred cubic metals:
Relation to symmetry and interatomic bonding. Philos. Mag., 84, 415428.
108
Bibliography
DE
Acknowledgements
Firstly, I would like to express my deep gratitude to my promoter Marc Geers, who
gave me the opportunity to start my PhD research in his group and who has been
continuously supporting me scientifically and socially throughout the whole PhD
process. He was not only the person to ask questions about the convergence problems in the models or the thermodynamical relations but also he has been the example researcher widening my horizon considerably. His positive approach during the
stressful times has always ended up with high motivation. Marc, thank you.
I would like to thank my co-promoter Marcel Brekelmans for his valuable input in the
process of developing each chapter and his fruitful discussions and criticism during
the development phase of the models and also on writing articles. Even though he
has been partially stopped with the university he never gave up supporting me.
I sincerely thank to the committee members; Prof.Dr. Klaus Hackl, Prof.Dr. Fionn
Dunne, Prof.Dr. Bob Svendsen, Prof.Dr. Marc Peletier and Dr. Henk Vegter for their
interest and constructive comments on the thesis.
A special thanks goes to Dr. Izzet Ozdemir who has always been there for discussions
and input. He has taught me how to be stubborn on debugging process and how to
overcome numerical problems during the implementation of some of the models.
I am grateful to Prof.Dr. Jeff de Hosson whom I visited in Groningen to discuss on
BCC crystals and who supported me via emails when I needed physical insight for
my computational models. Even though I have never met him, Prof.Dr. Ali Argon
from MIT has supported me remotely by answering my emails on the difficulties I
have faced in crystal plasticity modeling. I would like to thank him for his valuable
comments.
My sincere thanks goes to the IT administrators Patrick and Leo for their neverending support related to computer issues. And I would like to thank Alice van
Litsenburg in Mate and Monika Hoekstra from M2i for the support on solving practical matters faced throughout the PhD years in the Netherlands.
109
110
Acknowledgements
I would like to thank Karl Fredrik Nilsson, Peter Haehner and Vesselina Rugnelova
at the Institute of Energy in Joint Research Centre of European Commission, where I
have already started my post-doctoral research, for their understanding and support
on wrapping up my thesis. Considering the fact that chapter 5 of the thesis has
been completed at the Institute of Energy, the support of the institute is gratefully
acknowledged.
Finally, I am extremely grateful to my parents Duran and Tazegul
and my sister
Derya. I am such a lucky person to have such lovely and open-minded family. They
have always supported me for all the decisions I have taken.
Tuncay Yalcnkaya,
Petten, August 2011.
Curriculum Vitae
Tuncay Yalcnkaya was born on January 31, 1980 in Ankara, Turkey. After completing
his high school education in Batkent High School in 1998, he studied Aerospace
Engineering at Middle East Technical University, Ankara and got his B.Sc. degree in
2003. Directly thereafter, he continued his studies with a full DAAD scholarship at
the University of Stuttgart and got his M.Sc. degree on computational mechanics of
materials and structures (COMMAS) in 2005. After his graduation he was employed
by the Materials Innovation Institute (M2i) in Delft and has been working on his PhD
thesis on strain path change effects and crystal plasticity modeling of BCC metals at
the Mechanical Engineering Department of the Eindhoven University of Technology
under the supervision of Prof.Dr.Ir. M.G.D. Geers and Dr.Ir. W.A.M. Brekelmans.
He is currently a post-doctoral researcher at the Institute for Energy - Joint Research
Centre of European Commission.
111