Modelling SPD

Progress in Materials Science 95 (2018) 172–242
Contents lists available at ScienceDirect
Progress in Materials Science

journal homepage: www.elsevier.com/locate/pmatsci
Analytical and numerical approaches to modelling severe

plastic deformation
Alexei Vinogradov a,b, Yuri Estrin c,d,e,⇑
a
Department of Mechanical and Industrial Engineering, Norwegian University of Science and Technology – NTNU, 7491 Trondheim, Norway
b
Institute of Advanced Technologies, Togliatti State University, Togliatti 445020, Russia
c
Department of Materials Science and Engineering, Monash University, Clayton, VIC 3800, Australia
d
Department of Mechanical Engineering, The University of Western Australia, Nedlands, WA 6009, Australia
e
Laboratory of Hybrid Nanostructured Materials, NUST ‘‘MISIS”, 119049 Moscow, Russia
a r t i c l e i n f o a b s t r a c t
Article history: Severe plastic deformation (SPD) has established itself as a potent means of producing bulk
Received 6 November 2017 ultrafine grained and nanostructured materials. It has given rise to burgeoning research
Received in revised form 28 January 2018 that has become an integral part of the present day materials science. This research has
Accepted 2 February 2018
received a broad coverage in literature, and several recent publications (including reviews
Available online 03 February 2018
in Progress in Materials Science) provide a very good introduction to the history, the current
status, and the potential applications of SPD technologies. There is one aspect of SPD-
Keywords:
related research, though, which despite its great importance has not been covered by
Severe plastic deformation
Modelling
any substantive review, viz. the modelling and simulation work. Due to the complexity
Dislocation kinetics of SPD processing and the specificity of material behaviour at the extremely large strains
Finite element methods involved, analytical and computational studies have been indispensable for process design,
parameter optimisation, and the prediction of the microstructures and properties of the
ultrafine grained materials produced. They have also provided a better understanding of
the physical mechanisms underlying SPD and the mechanical response of the materials
that underwent this kind of processing. The pertinent literature is vast and often difficult
to navigate. The present article addresses this aspect of SPD and provides a commented
exposé of a modelling and numerical simulation toolkit that has been, or can potentially
be, applied in the context of severe plastic deformation.
Ó 2018 Published by Elsevier Ltd.
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
2. Modelling severe plastic deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
2.1. Estimates of equivalent strain during ECAP. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
2.2. Constitutive modelling approaches to simulations of ECAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
2.3. Effects of heat generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
2.4. Strain localisation during ECAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
2.5. Strain rate and temperature effects during ECAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
2.5.1. Estimates for strain rate during ECAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
⇑ Corresponding author at: Department of Materials Science and Engineering, Monash University, Clayton, VIC 3800, Australia.
E-mail address: yuri.estrin@monash.edu (Y. Estrin).
https://doi.org/10.1016/j.pmatsci.2018.02.001
0079-6425/Ó 2018 Published by Elsevier Ltd.
A. Vinogradov, Y. Estrin / Progress in Materials Science 95 (2018) 172–242 173
2.5.2. Estimates for temperature effects during SPD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188

2.5.3. Microstructure evolution during SPD in non-isothermal conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
3. Microstructure-based phenomenological modelling of severe plastic deformation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
3.1. Evolution of dislocation ensembles: An irreversible thermodynamics approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
3.2. One-internal-variable models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
3.2.1. Constitutive modelling of materials with a high Peierls stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
3.2.2. Constitutive modelling of hexagonal materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
3.2.3. Inclusion of deformation twinning in the constitutive modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
3.3. Two-internal-variable models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
3.4. Constitutive modelling of multi-phase materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
4. Possible scenarios for grain refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
5. Numerical simulations of SPD processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
5.1. FEM simulations of SPD processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
5.2. Softening at large strains and dynamic recrystallisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
6. Phase mixture modelling of nanocrystalline materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
7. Gradient plasticity for SPD processed materials. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
8. Simulations of synthesis of architectured materials by SPD techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
9. Texture evolution during SPD processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
10. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
1. Introduction
Deformation-induced microstructure refinement down to submicron range has come to the fore in the past decades as a
promising strategy for design of novel metallic materials with superior properties. A group of materials processing methods,
which enable such extreme microstructure refinement by virtue of plastic deformation to very large strains hardly achiev-
able with traditional metal forming operations, are commonly referred to as severe plastic deformation (SPD) techniques [1].
Not only do these techniques give rise to a radical improvement of the properties of metals and alloys by producing
microstructures with exceptionally small grain size, but they also provide an interesting test bed for validation of models
for large strains. Details of the SPD processing methods, the microstructures they produce, and the ensuing properties of
the processed materials can be found in several comprehensive surveys [1–4]. The complexity of the physical mechanisms
underlying the microstructure development under severe plastic deformation and the large number of the processing
parameters that may affect this development and the ensuing properties of the processed materials make it necessary to rely
on mathematical modelling of SPD. Given the relatively young age of the area of SPD technologies, the literature on the mod-
elling of SPD processes is astonishingly vast and sometimes difficult to navigate. With the diversity of the arsenal of mod-
elling approaches available for numerical simulation of severe plastic deformation in mind, we would like to take the reader
on a ‘guided tour’ through this extensive literature. The relative advantages and weaknesses of the various modelling tech-
niques will be considered and commented on. By no means do we see this overview as a collection of juxtaposed modelling
tools for a numerical simulation practitioner. It is rather intended as an informative introduction for researchers embarking
on modelling of SPD, while also providing in-depth insights into the matter for the more experienced readers.
SPD techniques, which go back to the pioneering work by Bridgman [5], combine severe shear deformation with high
hydrostatic pressure, which is now known to lead to extreme grain refinement, down to a deep submicron range. The impor-
tance of grain size as a pre-eminent characteristic of the microstructure has long been recognised [1,4,6,7]. Pronounced
strengthening achievable through grain refinement is commonly associated with the well-known Hall–Petch relation whose
origin has been reviewed on many occasions. Recent comprehensive reviews by Armstrong [8,9] and a new physical re-
interpretation of this relation by Langer [10] are particularly noteworthy. Originally employed to describe the empirical scal-
ing of the yield stress with the inverse square root of the average grain size of iron-based materials, this relation has provided
a general framework for many strengthening strategies employing the increasing fraction of grain boundaries as a core ele-
ment in both a ‘top-down’ and a ‘bottom-up’ approach to synthesis of novel materials [11–14]. It has also served as a touch-
stone in testing the soundness of models devised to rationalise the grain size dependence of the mechanical response of
polycrystalline materials [15–19].
Grain refinement is undoubtedly pivotal to the enhanced mechanical performance of SPD manufactured material
[1,2,4,9], and understanding the various mechanisms that may give rise to the Hall-Petch relation is a formidable task.
The abundance of grain boundaries in ultrafine-grained (UFG) materials is certainly one of the key features defining their
mechanical behaviour, and such mechanisms as the formation of dislocation pile-ups [20] or the activation of Frank-Read
dislocation sources at grain boundaries [21,22] do give rise to the Hall-Petch relation. Such direct grain boundary strength-
ening mechanisms are not sole contributors to strength of UFG materials, though. Indirect effects entering the flow stress
through the influence of grain boundaries on the dislocation density evolution (dislocation hardening) [4,23] can be equally
significant, if not even more important. This has been demonstrated in numerous publications, see e.g. [24–31]. Recently
174 A. Vinogradov, Y. Estrin / Progress in Materials Science 95 (2018) 172–242
Fig. 1. Hierarchy of length scales in metal plasticity ranging from atomic (dislocation cores) to dislocation patterns to multiple grains to macroscopic scale.
A major gap in multiscale modelling and simulation lies between the scales of atomistic simulations and dislocation pattern modelling. Attempts to close
this gap by discrete dislocation simulations play an increasingly greater role but phenomenological statistical theories are still predominant (from [34],
reproduced with permission).
Starink [32] has re-visited this subject and summarised the findings by showing that in many SPD processed metals and
alloys the dislocation strengthening is the predominant strengthening mechanism and that it gives rise to the same kind
1=2
of Hall-Petch dependence, r d (where r is the flow stress and d is the average grain size), as the direct grain boundary
strengthening. Hence, the description of strain hardening of SPD-processed ultrafine grained materials can be based on the
time proven modelling approaches involving dislocation density evolution.
Although extraordinarily high strains are imparted onto a workpiece during SPD processing, the overall behaviour of the
material can still be described in terms of the general laws of solid mechanics and the theory of plasticity. In a recent essay,
Osakada [33] presented an overview of plasticity in relation to the analysis of metal forming. McDowell [34] also reviewed
the research trends in metal plasticity in a historical retrospective. He highlighted two transformational trends which have
moved and developed the subject in new directions in the not-so-remote past: (i) the use of potent high resolution charac-
terisation tools capable of measuring attributes of microstructures directly related to plastic flow (such as crystallographic
orientation and misorientation between adjacent domains, grain/phase size and shape distribution and dislocation density),
and (ii) development of computational modelling and simulation tools that address inelastic deformation phenomena over a
broad range of length scales - from the atomistic to the macroscopic ones. A historical overview of one of the most popular
severe plastic deformation techniques, high pressure torsion, from its inception in the work by Bridgman [5] to the late
1980s was published by Edalati et al. [35].
Metal plasticity is fundamentally associated with nucleation, migration and interaction of a broad variety of crystal lattice
defects with different dimensionality, such as vacancies, dislocations, grain or interphase boundaries, as well as voids and
pores. With dislocations being the principal carriers of plasticity in most cases, the dislocation theory constitutes a basis
for understanding the evolution of microstructure during plastic flow. The classical treatises written by such nestors of
the dislocation theory as A. Cottrell, T. Mura, J.P. Hirth, J. Lothe, J. Friedel, and F.N.R. Nabarro [36–40] provide an in-depth
introduction to dislocations. Contemporary work based on discrete dislocation dynamics addresses the issue of dislocation
structure formation and evolution more directly, utilising the basics of dislocation theory in a sophisticated computational
framework (cf. [41]).
The picture of plasticity drawn schematically in Fig. 1 is very multifaceted and requires viewing from various angles and
at different length scales. So do modelling and computational simulations, as demonstrated convincingly in the ‘‘Handbook
of Materials Modelling” edited by S. Yip [42]. A good introduction to the key theoretical concepts underlying multiscale mod-
elling of materials from atomistic simulations to continuum mechanics and thermodynamics was presented by Tadmor and
Miller [43,44]. Models represented in Fig. 1 operate at various length scales and microstructure levels. They involve dynam-
ics of individual atoms (as described by ab initio quantum mechanics and molecular dynamics [45–47]), linear elasticity
based theory of discrete dislocations [48,49] and disclinations [50]. Dislocations may form pile-ups at the boundaries [38]
or organise themselves into spatially ordered patterns or substructures [41,51–54] within individual grains, heterogeneous
flow of aggregates of crystallites (polycrystal plasticity [55,56]), and, finally, collective effects at large scale (macroscopic the-
ory of plasticity [57,58]). At an intermediate scale of continuously distributed dislocation density, dislocation patterning was
considered in terms of the reaction-diffusion approach [54,59–64].
Ideally, research into the mechanical behaviour of solids should be done by progressing from left to right in Fig. 1, a model
at a smaller scale informing that at the next level of the length-scale hierarchy. In practice, however, this ‘hand-shake’ of the
models does not necessarily occur, as historically the models targeting the various scales were developed independently
from each other and mostly in the opposite direction – from right to left. Current efforts are focused on multiscale modelling
bridging different length scales with their distinctly different and often disparate ‘languages’. Early approaches to multiscale
modelling were presented in [65–67]. Excellent, more recent accounts of computational modelling incorporating different
length scales in plasticity offer a comprehensive picture of the state-of-the-art in this area [41,68–78].
Although theories at larger length scales attempt to subjugate the smaller scale phenomena into ‘effective’ properties or
‘constitutive rules,’ macroscopic phenomena associated with large strains ultimately depend on the details of smaller scale
processes. The number of degrees of freedom associated with a mathematical description at each scale decreases for the
same volume of material as one progresses from left to right in Fig. 1. This reduction of the number of the degrees of freedom
for a given material volume with ascending level of hierarchy is a fundamental goal of multiscale modelling. Information
necessary to calibrate model parameters can flow bottom–up, top–down, or in both directions. An alternative approach is
the so called concurrent multiscale modelling, which is the opposite of the hierarchical one in that a larger length scale
model does not subjugate a smaller scale one [71,79]. Rather, the models (commonly just two of them) are treated in parallel
and bridging occurs by using matching procedures in some overlapping domain [71]. In this way, some length scales may be
skipped and, for example, continuum level simulations can be coupled directly to the atomistic scale ones. Another useful
concept is quasi-continuous modelling, in which atomistic simulations are carried out in a region of interest, which is
embedded in a medium modelled as a continuum [80]. A most up-to-date summary of multiscale materials modelling defin-
ing the model taxonomy and outlining the conceptual and computational challenges can be found in [81].
Representing the top of the hierarchical pyramid, solid mechanics is the most popular instrument for obtaining both ana-
lytical and numerical solutions for the mechanical behaviour of a workpiece subjected to SPD processing. Solid mechanics
nowadays is a mature theory comprising the governing physical laws, sets of computational techniques, and numerical
methods for virtually all scales shown in Fig. 1 that can be used to predict the response of a solid to mechanical loading. Fun-
damentals of solid mechanics can be found in many excellent textbooks, e.g. [56,82–86], systematically covering all aspects
of modern approaches from theoretical background and basic mathematical principles to practical implementation of finite
element codes.
Choosing the right modelling approach, which determines the governing equations to describe the material behaviour
under specific loading conditions, is crucial for setting up a solid mechanics calculation. The toolbox developed in solid
mechanics includes a variety of conceptual models with specific applications listed below, and this is by no means an
exhaustive list.
1. Linear elasticity [87] is an important tool for describing small deformations of a polycrystalline material subjected to
loads that are much smaller than the material’s macroscopic yield stress, e.g. during very high cycle fatigue. Mechanics
of elasticity is of crucial significance for the theory of defects such as dislocations or disclinations in solids [88]. Despite its
fundamental importance, it is often neglected in simulations of metal forming and large deformations at macroscopic
scale [89]. It is, however, an appreciated instrument in discrete dislocation dynamics simulations owing to the applica-
bility of the elasticity theory to describing the dislocation stress and strain fields outside of the dislocation core.
2. Viscoelastic models [90], which include the Maxwell model, the Kelvin-Voigt model, and the Standard Linear Solid Model,
are widely used to predict the response of materials that exhibit both viscous and elastic behaviour under different load-
ing conditions. Elastic and viscous components are represented by linear combinations of springs and dashpots, respec-
tively. Depending on the arrangements of these elements resulting in different rheological equations, these models have
enabled a description of the behaviour of materials under creep or stress relaxation conditions. Such models are well sui-
ted for materials which exhibit a strongly rate dependent behaviour and are therefore most widely used in simulation of
the mechanical response of polymers. For example, Zaïri et al. [91] have calculated the plastic response of a polymer dur-
ing equal channel angular pressing at room temperature with different extrusion velocities, friction conditions and die
geometry. This group of models has limited applicability to metals and will not be considered in the present review. Inter-
ested readers are referred to [90].
3. Rate independent plasticity represents a wide class of models describing the deformation response of solids loaded above
their yield point up to large strains. The simplest model is that of a rigid elastic-perfectly plastic solid that deforms elas-
tically at stresses below the yield point and then flows plastically at a constant stress if a constant strain rate is applied.
An elastic-perfectly plastic solid thus exhibits a sharp transition from perfectly elastic to perfectly plastic deformation
when the yield stress is reached. Such models are often used to calculate the forces on tools in metal forming, for example
during equal channel angular pressing (ECAP) or high pressure torsion (HPT), cf. [92]. More realistic models involve strain
hardening in some way [93]. The models of this kind are used in simulation of plasticity at large strains and of low cycle
fatigue behaviour under fairly large strain amplitudes. Finally, the most sophisticated plasticity models attempt to track
the microstructure development in a plastically deforming metal. This group of models are popular in metal forming sim-
ulations, particularly simulations of SPD, as will be discussed in the next sections.
4. Viscoplasticity is similar in structure to rate independent plasticity, but it accounts for the strain rate dependence of the
flow stress, which commonly tends to increase with increasing strain rate, particularly under dynamic loading [94]. For
example, Anand [95] developed a model for the elastic–viscoplastic response of ductile single crystals deforming by crys-
tallographic slip. Numerical simulation of the deformation response of face centred cubic (FCC) polycrystalline materials
was conducted by assigning a finite element to each crystallite under the assumption of initially isotropic crystallographic
texture. The simulation results were shown to be in good agreement with experiment on copper. This refers both to the
strain hardening behaviour, which was found to be anisotropic, and to the evolution of crystallographic texture. An exam-
ple illustrating an excellent agreement between experimental observations of texture evolution during deformation to
large strains by simple shear and theoretical predictions [96] provided by the self-consistent viscoplastic polycrystal plas-
ticity model backed by a dislocation density based strain hardening model is shown in Fig. 2. The details of these mod-
elling efforts will be considered in the following sections.
5. In their finite element computations, Acharya and Beaudoin [97] predicted grain size dependent strain hardening beha-
viour of FCC and BCC polycrystals at relatively small strains of 2–30% by applying a viscoplasticity model (see also
[98,99]). At the individual grain level, isotropic Voce-type hardening was assumed. To account for the material’s resis-
tance to plastic flow, lattice incompatibility associated with the presence of lattice dislocations was introduced in a con-
tinuum description. Although the constitutive model for dislocation density evolution did not include the grain size
explicitly, its effect on the plastic flow was considered by analysing the results of modelling of the mechanical response
of FCC nickel and BCC HY-100 steel. An inverse relationship between the flow stress and the grain diameter was found.
The predicted strain hardening behaviour of polycrystalline nickel with a grain size below 100 lm agreed well with the
experimental data of Narutani and Takamura [100]. A transition to Stage IV hardening following saturation of the Voce-
type response inherent in Stage III hardening was found to be governed by a build-up of crystal lattice incompatibility due
to dislocation storage. This constitutive model was further extended to include temperature and strain rate effects
[101,102].
6. The crystal plasticity method has evolved recently as a framework integrating the extensive knowledge gained from
experimental and theoretical studies of single crystal deformation, atomistic fundamentals of dislocations, and contin-
uum mechanics of deforming solids. In crystal plasticity models, this knowledge of the physics of the deformation pro-
cesses at different length scales [69,103] is incorporated in the computational tools of continuum mechanics [84,85,104].
This approach was reviewed by Roters et al. [78] in a comprehensive monograph dedicated to diverse aspects of the plas-
tic behaviour of crystalline solids. Crystal plasticity starts with a premise that crystals are mechanically anisotropic and
therefore the instantaneous and time-dependent deformation of crystalline aggregates is an innately non-homogeneous
process, which depends on the direction of the mechanical loads and geometrical constraints imposed. Macroscopically
directional properties of a polycrystal arise when the orientation distribution of the grains, which is commonly charac-
terised by crystallographic texture, is non-random. Quantitative prediction of crystallographic texture evolution during
SPD that would account for concurrent grain refinement is therefore one of the major tasks for crystal plasticity imple-
mentations in simulations of SPD processes [94,105,106]. Most current models of texture evolution under severe plastic
deformation do not consider this continual variation of the grain population – with some exceptions ([107,108]) that will
be discussed below.
7. Strain Gradient Plasticity is a kind of hybrid approach motivated by advances in dislocation mechanics since the 1990s.
Classical plasticity theories do not account for non-local effects and fail to predict strain localisation or deformation
microstructures with a distinct pattern of spatially arranged dislocations with a well-defined length scale [109]. Unlike
conventional constitutive models, strain gradient plasticity is capable of accounting for non-local effects on the deforma-
tion behaviour and strength of polycrystalline aggregates [17,110,111]. The approach by Fleck and Hutchinson [112,113]
has advanced the concept in which an explicit distinction is made between dislocations stored during uniform straining
(statistically stored dislocations) and those necessitated by gradients of strain. The latter are referred to as geometrically-
necessary dislocations (GNDs) first introduced by Nye [114]. GNDs develop from inhomogeneities of plastic flow due to
loading conditions, texture, second phases, etc. to accommodate lattice misorientations. A main drawback of most strain
gradient plasticity theories is that the so-called internal length scale entering gradient plasticity formulations is vaguely
defined. Being introduced as a phenomenological variable, it serves as a ‘free’ parameter that is to be determined from
experimental data by appropriate fitting procedures. Numerous attempts to establish a relationship between the internal
length scale and the microstructure were reviewed by Zhang and Aifantis [115]. In the studies referenced therein, this
Fig. 2. Comparison of simulated crystallographic texture for simple shear during ECAP processing of pure copper to different numbers of passes (from 1 to
3) via Route A assuming no rotation of the billet between the passes (adapted from [96], reproduced with permission).
parameter has been associated with the dislocation source length, average dislocation spacing, size of pile-ups at grain
boundaries or interfaces, grain size, width of slip zones, or specimen size. Despite all efforts, the physical origin of the
internal length scale is still a matter of debate, however.
8. Discrete Dislocation Plasticity (DDP) is a theory originating from the recognition that plastic flow in crystalline metals
arises primarily from the collective motion of large numbers of dislocations. It emerged as a computational technique
tracking the nucleation, motion and annihilation of individual dislocations interacting with each other in a solid under
load [41,48,52,116–119].
9. Atomistic models are seeing a booming development owing to a rise in the available computation power on one hand and
the advancement of fundamental quantum-mechanical concepts enabling ab initio calculations of realistic interatomic
potentials and elastic constants on the other hand. In this approach phenomenological stress-strain relations are replaced
with a direct calculation of the stress-strain response using atomic scale simulations. A most recent example of such fully
dynamic atomistic simulations of single-crystal plasticity was provided by Zepeda-Ruiz et al. [120] for a specific case of a
BCC metal tantalum. Macroscopic stress-strain curves were calculated from these atomistic simulations directly, avoiding
multiscale modelling.
In what follows, we shall give an overview of the available modelling techniques – both analytical and numerical. On that
basis we shall discuss the salient results they have delivered and eventually single out the approaches that in our view are
most suitable in the context of modelling of severe plastic deformation.
2. Modelling severe plastic deformation
Great expectations are being put in achieving superior mechanical properties of bulk nanostructured and ultrafine-
grained materials manufactured by SPD. Despite a steady growth of the research area of SPD and the promise it offers, an
overly cautious attitude towards this group of processing techniques prevailed in metal forming industry until recently. Var-
ious barriers on the way of SPD processing from laboratory to shop floor include real technological challenges, such as the
need for upscaling the processes and making them continuous or semi-continuous. Some examples of processing routes of
this kind, which promise successful transfer to industry scale manufacturing [121–125] were given in the recent reviews
[1,4,126].
Regardless of the specifics of the processing schedule used, the microstructures and the mechanical properties of the
deformed materials are governed by the degree of plastic deformation (although not by it alone). Therefore, understanding
the stress and strain development in a workpiece is pivotal for efficient and sound SPD process design. The primary factors
governing the efficacy of SPD processing for microstructure refinement have long been understood and highlighted in many
publications. In the case of ECAP, which is illustrated schematically in Fig. 3, these include the die channel shape and dimen-
sions, the angle between the entrance and exit channels, the inner and outer corner radii, the coefficient of friction, the strain
path (or processing route), and the processing parameters, such as back pressure, ram speed, and temperature. The intrinsic
material properties of the billet have, of course, a decisive influence on the processability by ECAP and the ensuing material
properties, as well. The significance of all these process and material variables cannot be overestimated. A huge body of
experimental data was summarised in recent reviews [4,126,127]. Medeiros et al. [128] performed a variance analysis based
on the central composite factorial design to quantify the relevance of processing parameters to the ECAP load and the effec-
tive plastic strain. From the statistical analysis of variance (ANOVA), the ECAP parameters affecting the load the most can be
Fig. 3. Schematic illustration of ECAP showing the channel angle U and the corner angle W: (a) corner angle W = 0, (b) corner angle W > 0 and (c) U = 90o;
ED, TD and ND denote extrusion, transverse and normal directions, respectively. (Adapted from [135], reproduced with permission.)
ranked in the following order of importance: (1) the coefficient of friction, (2) the die channel angle, (3) the outer fillet radius,
(4) the inner fillet radius, and (5) the ram velocity. Also, with regard to the equivalent plastic strain the ranked order of sig-
nificance of the die geometry parameters was determined to be as follows: (1) the die channel angle, (2) the inner fillet
radius, and (3) the outer fillet radius. However, despite substantial efforts none of the experimental research aiming at char-
acterising the processing factors and their roles in the properties of a final product can claim to be comprehensive. As a mat-
ter of fact, the assessment of all these factors in their entirety requires enormous experimental efforts. Indeed, if one were to
limit oneself to just the most important factors and to aim at characterising the effect of, say, two channel shapes, two dif-
ferent cross-section sizes, five different friction conditions (corresponding, for example, to different lubricants, the quality of
the die wall machining, or the die design), five different materials, four most common strain paths (A, B, C and Bc, see [2] for
the definition of the ECAP routes), three different pressing velocities, three levels of back pressure, and three different tem-
peratures, one would need to perform over 10,000 pressings followed by microstructure characterisation and mechanical
testing. Therefore, the use of modelling tools in either mechanistic analytical formulations, cf. [129–133], or finite element
analysis [84–86,134] is indispensable for optimisation of the processing schemes and conditions for a given material in a
cost- and time-effective way. Of course, the mathematical models discussed in the previous section, particularly the analyt-
ical models, provide useful insights in the deformation behaviour. However, while providing the analytically tractable and
elegant solutions, these models commonly oversimplify the processing conditions and provide estimates of average equiv-
alent strains across the workpiece. As such, they do not account for strain and stress inhomogeneity, which inevitably occurs
during SPD processing. Therefore, the analytical solutions are not capable of simulating a real SPD process in sufficient detail.
Nowadays, it is the finite element method (FEM) based modelling that plays a crucial role in understanding, critical assess-
ment and optimisation of the existing SPD processes, visualisation of strain distributions across the sample, and prediction of
the mechanical properties of the processed material. They are also invaluable in developing new SPD processes by enabling
virtual process simulations prior to committing to the use of expensive tooling and machinery.
In this section we look at the ways in which FEM was applied to the simulations of an archetypal SPD process – equal
channel angular pressing (Fig. 3) – arguably the most commonly used technique in the context of SPD processing. Further
SPD methods will be considered in subsequent sections, particularly in Section 5.
2.1. Estimates of equivalent strain during ECAP
Various analytical calculations [131,136–142] and numerical finite element simulations were used for modelling and
prediction of strains induced by ECAP. The models differ in the die geometry represented in Fig. 3, incorporation of fric-
tional effects, and the techniques used to calculate the strain distribution. Sharp corner dies, dies with an external arc
of curvature, and round corner dies are some of the various geometries employed for ECAP. In the dies with an external
arc of curvature, the centre of curvature is located at the inner corner of the die. The round corner dies have a fillet at
the outer corner whose centre of curvature does not lie at the inner corner. Detailed understanding of the kinematics of
deformation and the effects the die geometry plays in it is crucial for rationalising the basic mechanisms that control
the grain refinement in the ECAP process. Eivani and Taheri [141,142] presented a upper-bound solution in which both
the friction conditions and the nonlinear strain hardening behaviour were considered for a die geometry where only the
outer die had a corner curvature. The consideration of nonlinear strain hardening of the material in the calculations of
the ECAP pressure, assuming a frictionless condition, with an outer die corner curvature, was first proposed by Alkorta
and Gil Sevillano [137]. Their analytical solution was based upon the upper bound theorem and provided a good agreement
with numerical predictions determined from a plane-strain finite element model. Paydar et al. [138] also exploited the
upper bound approach to analyse the equal channel angular pressing with circular cross-sections of the die and the billet.
The expressions for the total dissipated power and the applied pressure required for processing were established by con-
sidering velocity discontinuities and non-zero friction at all surfaces. The model developed predicted an increase in size of
the deformation zone with increasing friction, similar to Segal’s original model [131]. In a series of publications Pérez and
Luri [140,143–148] obtained theoretical expressions for the shear strain accounting for all the possible die configurations
and including the friction effects for perfectly plastic materials. Aiming at optimisation of the die geometry, these authors
showed a gain close to 11% in the equivalent plastic strain per pass when the inner radius was 2.67 times larger than the
fillet radius of the outer die. They also demonstrated that an increase of the inner fillet radius led to a greater equivalent
plastic strain and higher pressure levels. Building on the recognition that an extended arc curvature has a detrimental
effect on strain homogeneity in the workpiece during ECAP, Sßimsßir et al. [149] performed an upper bound analysis by
FEM and showed that the process performance in terms of the magnitude and uniformity of strain can be improved by
modifying the shape of the die corner.
Segal [129,150,151] was the first to exploit a widely known slip line field theory for solutions of the boundary value prob-
lems in mechanics of ECAP. The slip line kinematic approach simplifies the governing equations for plastic solids by making
several restrictive assumptions including [83]: (1) plane-strain deformation, (2) quasi-static loading, (3) isothermal condi-
tions, (4) absence of body forces, and (5) idealisation of the material as a rigid-ideally plastic von Mises solid whose mechan-
ical properties are characterised by the yield stress in uniaxial tension. With these assumptions Segal obtained slip line field
solutions for different ECAP die geometries and friction conditions. For a simple die with sharp corners and an angle U
between the channels, Fig. 3a, the von Mises equivalent strain eeq imposed onto a billet per each ECAP pass is given by [129]
Fig. 4. Flow pattern in a cross-section of commercially pure aluminium AA 1050 billet deformed by ECAP (a). Etching reveals a fan-shaped plastic
deformation zone (PDZ) between undeformed and deformed parts in the entrance and exit channels, respectively. The experimentally observed flow lines
on the TD plane are shown in (b); the flow line shape parameter nfl entering the model due to Tóth et al. [156] calculated for individual flow lines is shown
on the right (adapted from [157], reprinted with permission).

2 U
eeq ¼ pffiffiffi cot ð1Þ
3 2
In the case of a die with a round outer corner geometry, Fig. 3b, which owes its great popularity to the simplicity of design
and manufacture, the strain is given by the following relation proposed by Iwahashi et al. for the frictionless conditions [152]

1 U W U W
eeq ¼ pffiffiffi 2cot þ þ Wcosec þ ð2Þ
3 2 2 2 2
Goforth et al. [153] proposed an alternative formula for the equivalent plastic strain:

1 U W
eeq ¼ pffiffiffi 2cot þ þW ð3Þ
3 2 2
Alkorta and Gil Sevillano [137] derived the same expression in the upper bound approximation for the frictionless die
with a curved outer corner, while Milind and Date [154] arrived at Eq. (3) using a kinematics based approach. Aida et al.
[155] reconciled the latter two equations and showed that both yield the same results at the upper and the lower bounds
for the arc angle W = U p and W = 0, respectively, and that they differ by less than 5% under all other conditions for the
channel angle U equal to or exceeding 90°. Segal’s slip line model assumes that plastic flow in simple shear deformation
in the 90° sharp corner dies is confined to the plane of intersection between the channels. To avoid this limitation, Tóth
et al. [105,156] developed an analytical flow line model which they applied – with considerable success - to calculating tex-
ture evolution under ECAP [96]. In this model a flow line is described by the following equation:
n nfl nfl
ðddie xÞ fl þ ðddie yÞ ¼ ðddie x0 Þ ð4Þ
where ddie is the diameter of the die, x0 denotes the current position on the flow line, and nfl is a free parameter describing the
flow line shape in Cartesian ðx; yÞ coordinates corresponding to the rectangular die as illustrated schematically in Fig. 3c. The
condition nfl ¼ 2 defines an idealised circular flow line; higher nfl values approximate well the flow pattern arising under
different friction conditions; in the limiting case of nfl ! 1, the flow line is simply represented by two straight intercepts
connected at the shear plane of the die, as in Segal’s model. The experimentally observed flow patterns can be approximated
by the proposed model very well, as shown in Fig. 4. A generalised expression for the equivalent von Mises strain per ECAP
pass in the U ¼ 90 die derived from the model reads
2 pðnfl 1Þ
eeq ¼ pffiffiffi ð5Þ
3 n2fl sinðp=nfl Þ
pffiffiffi
This equation yields the same total strain, eeq ¼ 2= 3, as Segal’ model for the limit case of nfl ! 1. Eq. (5) implies that the
accumulated strain is uniform across a section of the die, provided nfl is constant. The latter assumption is quite restrictive
and does not have general validity. The flow line model proposed by Tóth et al. predicts a varying deformation field along the
flow line in good agreement with FE simulations [156]. Using a large 50 50 mm2 ECAP die set-up, Panigrahi et al. [157]
demonstrated experimentally that the flow line exponent nfl increases significantly and almost linearly from about 6 at
the top to about 20 at the bottom side of a billet. This behaviour of nfl reflects the occurrence of a strain gradient in the
extended plastic deformation zone, Fig. 4. Using the experimentally determined nfl values, the flow line model [156] cor-
rectly predicts the distribution of texture components in a billet’s cross-section from top to bottom. A further generalisation
of this analytical model was provided by Hasani et al. [158], who considered the elliptical modification of the flow line equa-
tion, Eq. (4), and used it for modelling the plastic flow and texture evolution in the nonequal channel angular pressing, in
which process the exit channel has smaller dimensions than the entrance channel.
Motivated by the same desire to avoid the simplistic representation of the ECAP process as sharply localised shear and to
account for inhomogeneity of plastic flow in a spread plastic deformation zone, Beyerlein and Tomé [159] proposed another
form of analytical solutions for the strain and velocity gradients tensors as functions of the position in the die and deforma-
tion time for three typical deformation situations, including ideal localised simple shear, a central fan-shaped plastic defor-
mation zone, and a central deformation fan combined with an outer region of rigid rotation. Not only is this analytical model
capable of successfully describing texture evolution in the multipass ECAP processing, but it also enables a thorough analysis
of the deformation kinematics.
Analytical modelling was also used to describe HPT processing. The approach goes back to Bridgman [35,160,161] who, in
an idealised way, expressed the velocity field in the specimen (which is independent of the torsion angle of the anvil, b) as
r xz
V z ¼ V r ¼ 0; Vh ¼ ð6Þ
H
with x the angular velocity, and the von Mises equivalent strain as
rb
eeq ¼ pffiffiffi ð7Þ
3H
In the cylindrical coordinate system ðr; h; zÞ used, r is the distance from the specimen axis; H denotes the specimen thick-
ness [162,163]. This type of kinematic model was also used by Khoddam et al. [132]. Beygelzimer et al. [164] provided a strict
mathematical analysis to identify conditions under which the simplified description given by the above equations is valid.
Their conclusion was that in the rigid plastic formulation, a power-law hardening in a rotation angle interval would lead to
self-similar regimes of HPT if the friction with the lateral wall of the die is not too high. In such an interval, the above math-
ematical description holds. Outside of this interval it breaks down, and the plasticity problem needs to be solved for each
value of b. This is an important message that should be considered in calculations of the HPT process.
It should be noted that it is customary in the SPD community to use the von Mises equivalent strain as a measure of strain,
as was also done in deriving the above equations. The ongoing discussion in literature on whether it is the von Mises or the
Hencky strain [165,166] that provides the appropriate measure of deformation involving simple shear, of which SPD pro-
cesses are typical examples, was reviewed in a recent publication [167]. Based on group theory analysis, it was demonstrated
that the Hencky equivalent strain does not satisfy group theory properties for simple shear, while the von Mises equivalent
strain does. This analysis provides a theoretical underpinning for the use of the von Mises platform, which is adopted here.
2.2. Constitutive modelling approaches to simulations of ECAP
As was emphasised above, the microstructure and the concomitant mechanical properties of SPD-manufactured materials
show a strong dependence not only on the amount of plastic strain imparted to the material but also on the strain path. This
is particularly the case for multi-step processing schedules where the results depend on the sequence of strain increments
associated with the processing steps. Substantial modelling efforts have been invested in understanding the strain develop-
ment during various SPD processes.
ECAP has received most attention in finite element modelling, which has finally led from laboratory investigations to
technologically relevant ideas of possible upscaling of SPD techniques, and we shall dwell on this process in some detail.
When a rectangular workpiece is subjected to ECAP under isothermal conditions with ideal lubrication, the strain along
the direction normal to the flow plane is zero, i.e. plane strain deformation dominates the plastic flow. This justifies a
two-dimensional treatment of plastic strain distribution over a rectangular billet during ECAP as a first approximation
[168]. Owing to its simplicity, 2D finite element modelling enjoys the greatest popularity even though the significance of
boundary conditions at the side walls is admittedly disregarded in the 2D approach. Experimental arguments in favour of
nearly isothermal condition during ECAP at low processing speed were provided by Yamaguchi et al. [169,170] and Berbon
et al. [171], at least for the relatively slow ram velocity of 0.18 mm/s. An essential input in any FEM simulation that strongly
influences the results is a constitutive model specifying the response of the material to an applied stress under given con-
ditions, i.e. a functional dependence of the flow stress on strain and further variables such as strain rate e_ and temperature T,
r ¼ rðe; e_ ; TÞ. For this purpose, when strain hardening is considered, a Ludwik type dependence of the flow stress r on plastic
strain epl is assumed, most commonly in the form:
r ¼ K enpl ð8Þ
where K and n are material parameters, which may generally be strain rate and temperature dependent and are supposed to
be known from a set of experiments, e.g. from uniaxial tensile tests performed at different strain rates and temperatures. A
variant of this constitutive equation, which takes into account the strain rate dependence of stress in a power-law form, is
the Ludwik-Hollomon equation

r ¼ K enpl e_ mpl ð9Þ
Here the exponent m characterises the strain rate sensitivity of the flow stress and may be temperature dependent. Obvi-
ously, the validity of the Ludwik or the Ludwik-Hollomon equation implies continual increase of stress with strain. Assuming
a Ludwik type strain hardening, the Considère instability criterion predicts the onset of strain localisation (necking) at a crit-
ical strain en ¼ n under uniaxial tensile deformation. An excellent critical review on plastic strain localisation in materials has
been provided recently by Antolovich and Armstrong [172] with the focus on the influence of the microstructure on different
scales and loading conditions – temperature and strain rate – on the strain hardening behaviour and the appearance of plas-
tic instabilities. The onset of necking limits the validity of Eq. (8) to a relatively small strain range below the necking point.
Although this type of strain hardening law is rather common in solid mechanics, it does not apply to large deformations
where the stress tends to saturate or even drops with strain. For example, after several ECAP passes the microhardness
(and thus the strength) tends to saturate [136,173,174] or exhibits a maximum and then decreases slightly with increasing
strain [175,176]. This apparent breakdown of the Ludwik-Hollomon power law may be corrected by using an alternative
constitutive law first proposed by Voce [177] as an empirical relation between stress r and plastic strain epl (cf. also
[178–180]):

rðepl Þ rs epl
¼ exp ð10Þ
ry rs ec
where ry is the yield stress and ec is the quantity that characterises the rate of variation of stress with strain towards its
saturation level rs . The corresponding strain hardening rate h ¼ @ r=@ e is expressed as:

r
h ¼ h0 1 ð11Þ
rs
providing a linear stress dependence of the strain hardening rate. Here h0 ¼ rs =ec is the strain hardening rate extrapolated to
zero stress. It is commonly associated with Stage II hardening. Mecking and Lücke [181], Kocks [182] and Mecking [183]
derived such a linear relation for the tensile stress-strain curves of FCC metals by considering the dislocation density evo-
lution. Possible extensions of this approach to large strains are discussed in the review paper by Kocks and Mecking
[184]; see also [58] for an in-depth discussion of the constitutive strain hardening behaviour at large strains. Unlike the
Ludwik-Hollomon equation, Eq. (10) does account for stress saturation with increasing strain. For this reason, it has been
widely used in mechanistic descriptions of large strain deformation. Chinh et al. [185,186] employed the Voce equation
to describe the strain hardening behaviour in ECAP-processed aluminium over a broad range of testing temperatures, includ-
ing both the region of plasticity controlled by thermally activated dislocation glide at low temperatures and the diffusion-
controlled plasticity regime at high temperatures. In an attempt to fit the experimental data, they proposed a different type
of the constitutive law in the form of

enpl
r ¼ r0 þ r1 exp ð12Þ
ec
where r0 , r1 , ec and the exponent n are constants which may be used to fit the experimental data. Eq. (12) fulfils the various
requirements on the r vs. epl relation, as it reduces to the Voce-type equation when n = 1 and its Taylor expansion at small
strains gives an equation of the form r ¼ r0 þ K enpl , which recovers the Ludwik equation, Eq. (8). The relation expressed by
Fig. 5. Illustration of the flow patterns predicted by 2D FEM for different frictional conditions and die configurations: (a, c) well-controlled shearing
facilitated by extremely low friction or a sliding exit channel floor; and (b, d) inefficient, nonuniform shearing and a dead-metal zone caused by high friction
levels in a simple (static) die (here a and b were adapted from [190] and c and d from [196] (reproduced with permission)).
Eq. (8) is purely phenomenological. Its deficiency is a lack of an explicit or implicit relation with the microstructure and dis-
location structure evolution. Being convenient as a mathematical description of the strain hardening behaviour of materials,
it is often used in FEM codes. However, neither is Eq. (8) capable of describing different stages of strain hardening nor can it
be easily justified for both small and large strain ranges.
In 1997, Prangnell et al. performed the first 2D plain strain FEM simulation of the ECAP process [187] emphasising the
significance of friction for the kinematics of plastic flow in the dies with different channel geometries. Friction is undoubt-
edly of great importance in general metal forming and that is why it is routinely included in modern FEM codes [188]. Fig. 5
illustrates the flow patterns predicted by 2D FEM for different frictional conditions and die configurations. Nearly ideal sim-
ple shear was predicted in well-lubricated conditions with low friction, whilst non-uniform shearing and a dead-metal zone
caused by high friction levels in a simple (static) die was found. These computational results are in fair agreement with
experiments performed by Segal and co-workers [150,151,189]. The role of friction in SPD processing was highlighted in
many publications [139,190–199]. The significance of friction for ECAP processing was also confirmed by 3D FEM simula-
tions [200].
As seen in Fig. 3, the channel geometry can be varied by using different values of the channel angle [201,202], or the outer
corner angle [139,192,193,198,203–205], or both [137,194,206,207]. The shape of the entry channel can also be made differ-
ent from that of the exit channel (thus deviating from pure ECAP) [208,209]. The vast majority of FEM simulations were car-
ried out with a view to assess the influence of the die geometry, process conditions, and the intrinsic properties of the billet
material on strain distribution in the ECAP-processed billets. For instance, in a suit of publications Nagasekhar and co-
authors investigated the influence of tool angles, strain hardening behaviour of the material (adopting the Ludwik-
Hollomon model) and friction between the billet and die (adopting the Coulomb friction law) in 2D plane strain simulations
with elastic-plastic material properties of aluminium alloy AA1100 [204] and 3D simulations for copper [210] for conven-
tional ECAP as well as ECAP with specific channel shapes, such as V-shape [211], T-shape [212] and X-shape die configura-
tions [213]. The effect of the type of the friction model - Coulomb vs. shear friction - in numerical analysis of the ECAP
process was considered by Balasundar and Raghu [214]. They demonstrated that the shear friction is preferable in FEM codes
because the predictions of this model are closer to experimentally observed deformation pattern and strain distribution
across the processed billet.
With friction taken into account, it was shown in particular that an optimum in strain homogeneity in the sample with a
lesser propensity for dead zone formation can be achieved, without any detrimental side effects, for the channel angle U =
90° and the outer corner angle W = 10°, cf. also [215–217]. A typical simulated flow pattern and the equivalent stress distri-
bution in the deformation zone of the billet passing through the die with different values of the channel angle are shown in
Fig. 6. As predicted by Segal [131,151,189] considering mechanics of a rigid-plastic body in the ECAP process, deformation is
Fig. 6. Results of FEM simulation showing the effect of the shape of the outer corner on the flow pattern and the equivalent von Mises stress distribution in
the deformation zone. Simulation performed with U = 90° and (a) W = 0°, (b) W = 10°, (c) W = 20°, (d) W = 28° (adopted from [204], reproduced with
permission).
confined to a thin layer at the plane where the two channels meet. Within that layer deformation by simple shear occurs. For
strain hardening materials, under the same zero friction condition, the deformation zone spreads out through an arc around
the crossing plane of the channels due to the formation of a dead zone [218,219]. When both strain hardening and friction
are included, the shape of the deformation zone changes considerably. With W = 0° (cf. Fig. 3a), the deformation zone is very
wide and is no longer in the form of an arc. This strain pattern happens to be closer to the computational predictions (clearly
observable with deformed mesh in Fig. 5a). FEM simulations commonly show that with an increase in the outer corner angle,
the arc of the deformation zone spreads as shown in Fig. 6b–d [204,215,216,220]. Hence, by reducing the outer corner angle a
narrowing of the deformation zone can be achieved.
To raise the process efficiency and reduce the number of processing steps, ECAP tool designs with either U- or S-shape
channel configuration were considered. In view of drastically increased loads, modelling is essential in assessing the viability
of the proposed tool design and its optimisation [195,201–203,208,220,221].
Another factor, which has long been identified as a major one in ECAP processing and the resultant material properties, is
back pressure [129,131,150,222–225]. Back pressure was shown to be crucial for enabling uniform simple shear deformation
[129], but it is comparatively rarely considered in numerical simulations [137,193,196,197,203,226].
Summarising the outcomes of numerous FEM simulations of ECAP, it can be concluded that the strain distribution in the
deformed billet is strongly affected by the channel shape, tool angle, friction, and back pressure. The strain is most uniform if
the deformation zone is as narrow as possible. This is best achieved with a sharp die corner and appropriate back-pressure,
as was recognised in the original work by Segal et al. [129]. Using a slip line model, these authors explicitly formulated the
hydrostatic pressure condition for ‘‘ideal” simple shear deformation with complete filling of the outer die corner, which
according to them is achieved if the friction between the workpiece and the die walls is negligibly small and the operating
punch pressure in the inner channel p1 is related to the back pressure p0 through p1 ¼ p0 þ 2k cot U, where k is the material’s
shear yield stress. Extending Segal’s approach, Lee proposed an upper bound analytical solution taking into account friction
during ECAP [227]. If a ‘razor-blade thin’ deformation zone is formed, the simple shear conditions assumed by Segal et al.
[129] in deriving Eq. (1) prevail and the magnitude of simple shear the billet undergoes agrees well with the predicted val-
ues. In all other cases the shear strain is significantly lower than expected from Eq. (1). This is due to a partial bending of the
billet, which results in the strain distribution being non-homogeneous (by up to 40% across the billet’s width for a 120° die).
Eq. (1), therefore, corresponds to an upper bound that can only be achieved under ideal conditions. It follows from the above
analysis that dies designed with a smoothed internal corner will certainly promote undesirable bending of the billet and
reduce its homogeneity, as well as the overall level of shear.
2.3. Effects of heat generation
The tool temperature and the heat generation effects during SPD were considered in several simulations [135,190,193,
195,197,219,228,229]. DeLo and Semiatin in their early FEM analysis of ECAP evaluated the effect of friction in a systematic
way. More importantly, beside considering the plastic flow behaviour of the workpiece, they accounted for the effects of pre-
heat temperature, heat generation due to mechanical work and interfacial heat transfer between the workpiece and the die
[190]. Since many engineering alloys experience flow softening under hot-working conditions and, consequently, are prone
to flow localisation, the motivation to perform realistic simulations accounting for these effects is very clear and important.
Consideration of heat transfer and of friction between the workpiece and the side walls of the tooling required a 3D approach
to full analysis of hot ECAP. Strain rate sensitivity of the processed material was included into FEM codes for both two-
dimensional and three-dimensional non-isothermal models. The heat generation resulting from deformation was computed
and flow stress data were adjusted in the simulation routines accordingly. The temperature increment DT in the deformation
zone due to plastic deformation was calculated in a simplified way using the adiabatic approximation (see below):
Z
bT ðe_ Þ
DT ¼ r de ð13Þ
qd cv
where T is the absolute temperature and r is the flow stress. (Henceforth a superposed dot denotes time derivative and
a bar refers to the average value). The material constants qd and cv are the mass density and specific heat, respectively;
bT represents the fraction of mechanical work that is converted into heat. The result of a correction for deformation
heating in FE simulation is an increased amount of flow softening compared to that observed in ‘uncorrected’ compres-
sion data [190].
Despite the demonstrated efficacy of 3D simulations in modelling heat flow, they were less successful than the 2D FEM
simulations, Fig. 7, in predicting overall deformation and equivalent strain contours [190], see also [135]. These difficulties of
3D FEM simulations were attributed to problems with mesh density and element behaviour highlighting the challenges in
realistic FEM simulations. The simple shear deformation concentrated in a narrow band during ECAP requires a very fine
mesh in the deformation zone and frequent re-meshing which can be more easily implemented in 2D FEM codes. Conse-
quently, a much greater number of elements, memory and computation time are required for an equivalent 3D simulation.
This is a technical issue, though, which does not undermine the recognition that 3D simulations are more appropriate in
principle [146,195,198,228,230–232].
Fig. 7. FEM-predicted grid distortions, temperature contours, and effective strain contours from a 2D nonisothermal ECAP simulation of Ti-6Al-4V with
flow softening at 985 °C (from [190] reproduced with permission).
2.4. Strain localisation during ECAP
In the absence of heat transfer between the workpiece and tooling, which was tacitly assumed in deriving Eq. (13), the
tendency to flow localisation was estimated from the flow localisation parameter aS defined as [191]:
( )
1 de_ 1 @ r @ r 1 dT 1
aS _ ¼ þ ð14Þ
e de r @ e e_ ;T @T e;e_ r de m
It was demonstrated that materials with a high degree of softening and a low strain rate sensitivity m, which is defined in
the same sense as that in Eq. (9) as m ¼ ð@ ln r=@ ln e_ Þje;T , tend to undergo severe shear localisation. Flow localisation sets in
only if as is positive. Semiatin et al. noted that the development of appreciable strain localisation occurs for aS of 3 at typical
cold-working temperatures [233] or 5 at hot-working temperatures [234].
Lapovok et al. [235] studied strain localisation during ECAP in terms of a simplified power-law constitutive model by con-
sidering gradient plasticity. This was done by introducing a second-order gradient term associated with the incompatibility
stresses between neighbouring grains, giving rise to the following expression for the shear stress
@2c
sðc; c_ Þ ¼ l~ ðc_ Þc_ m ðc0 þ cÞp~ðc_ Þ ~c ð15Þ
@y2
Here, in accordance with a gradient plasticity model due to Estrin and Mühlhaus [236], the coefficient ~c was assumed to
be proportional to the Young’s modulus E and the square of the grain size d: ~c ¼ a ~ is a constant). The deforma-
~Ed (where a 2
tion dynamics was modelled using the classical continuity equation of solid mechanics:
@u @ s
qd ¼ ð16Þ
@t @ c
with a shear stress gradient entering on the right-hand side and qd denoting the density of the material. Linear stability
Rt
analysis of the fundamental solution of Eqs. (15) and (16) with c_ ¼ @u=@y and c ¼ 0 jc_ j dt was performed by introducing
small perturbations, in the exponential form d/ ¼ d/0 egðtt0 Þ ei2py=kw , of the fluctuating variables / ¼ ðu; c; sÞ at a time to.
Fig. 8. Deformation curves used in isothermal FEM simulations with different strain hardening models [191] (reproduced with permission).
The wavelengths kw for which the growth rate g of the perturbations turns out to be positive were determined in dependence
on the material properties (including grain size and strain rate sensitivity) and processing conditions (strain, strain rate,
strain gradient, etc.). These wavelengths define the spatial periodicity of an emerging strain localisation pattern. The model
was verified for magnesium alloy AZ31 deformed by ECAP at 250 °C with different ram speeds and different back pressure
values. The predicted deformation band patterns agreed well with the observed ones, including such detail as the shift of the
spectrum of the preferred wavelengths upon the variation of back-pressure.
Although most published FEM calculations assumed the Ludwik-Hollomon or Voce type strain hardening behaviour of a
material, some simulations were also conducted under the assumption of ideal plasticity [198,206,208] or viscoplasticity
[237]. In some hybrid models, ideal plasticity was considered as a limiting case for strain hardening [197,221] or as a model
to be compared with strain hardening ones [191,192,194]. Two different strain hardening curves for reversed loading (taking
place for 180° billet rotation between two consecutive passes of ECAP) were used by Figueiredo et al. [238]. Three different
kinds of hypothetical material stress-strain behaviour, typical of those encountered in cold and hot working, were investi-
gated by Semiatin et al. [191]: (i) Ludwik-Hollomon type strain hardening, (ii) rigid-perfectly plastic response, and (iii) strain
softening response, cf. Fig. 8 where the deformation curves are shown for these three situations.
Using similar assumptions, Kim [239] simulated plastic flow in a strain hardening material (with 1100Al taken as an
exemplary case) and a nearly perfect elastic-plastic material (6061Al-T6). It was found, Fig. 9a and b, that a larger corner
gap is formed in the material with a higher strain hardening rate because the softer outside part of the workpiece in the
deforming region flows faster in a strain hardening material than in a non-hardening one. The results of these simulations
are in good agreement with experimental data presented by Segal [189], Fig. 9c and d. Besides, Kim demonstrated that the
corner gap formation reduces the strain in the outside region of the workpiece, increases the strain in the inside region, and
diminishes the average strain. A computational result already mentioned above and well-supported experimentally is that
the strain distribution in a workpiece with a larger die corner gap becomes more non-homogeneous.
Fig. 9. FEM-predicted grid distortions for (a) a strain hardening material (1100Al) and (b) a nearly perfectly plastic material (6061Al-T6) (adapted from
[239], reprinted with permission). (c) and (d) represent experimental data for a U = 90° and W = 0° die showing a distortion of the coordinate grid for a
strain hardening material and non-strain hardening material, respectively (adapted from Segal [189], reproduced with permission).
SPD processability of hard-to-deform materials was addressed in an experimental study accompanied with FEM
modelling by Semiatin et al. [240,241]. They investigated the deformation behaviour of commercial purity (CP) titanium
(grade 2) and AISI 4340 steel during ECAP at temperatures between 25 °C and 325 °C and nominal strain rates between
0.002 and 2 s1. CP titanium underwent segmented failure under all conditions except at low strain rates and high temper-
atures, while 4340 steel deformed uniformly except at the highest temperature and greatest strain rate, at which condition it
also exhibited segmented failure. CP Ti was found particularly susceptible to shear localisation during ECAP; uniform flow
occurred only at high temperatures and low strain rates. By contrast, 4340 steel exhibited uniform deformation under all
conditions, apart from the high temperature and high strain rate case mentioned. Observed shear banding and shear failure
were interpreted in terms of the tendency for strain concentration as quantified by the flow localisation parameter aS intro-
duced in Eq. (14). Viscoplastic modelling and FE simulations were applied to elucidate failure modes under non-uniform
plastic flow during non-isothermal ECAP and the model predictions were found to be in good agreement with the experi-
mental findings. Based on the understanding of the effect of material properties on the tendency for flow localisation sup-
ported by both FEM calculations and experiments, Semiatin et al. proposed countermeasures to inhibit non-uniformity of
plastic flow during ECAP [240,241]. Localised shear banding was also found by very recent FEM simulations of ECAP of
AA6060 using a phenomenological constitutive model that included both isotropic and kinematic hardening [242]. The pre-
ponderant role of the latter in strain localisation under ECAP was highlighted by the authors.
Extreme grain refinement of metals usually involves multiple ECAP passes [2,3,126,243]. Such processing can follow var-
ious strain paths, for which the strain hardening behaviour of the material being deformed may be different
[2,4,129,136,150,152,244,245]. Although accounting for plastic anisotropy is essential for accurate simulations of large plas-
tic deformations during metal forming processes, the strain path effects are relatively rarely considered in FEM modelling
[246–249]. Rather, a universal equivalent stress vs. equivalent strain curve of the material is employed for all processing
steps. An example is multiple pass deformatin by route A ECAP considered by Rosochowski [203]. This approach was mod-
ified by Figueiredo et al. [238] who presented a methodology of FEM analysis for two-pass ECAP processing of copper fol-
lowing route C, which involves a strain reversal. The consideration of strain path effects changed the final strain
distribution in the material and led to a lower ram force in the second pass compared with the results based on a universal
stress–strain curve for all passes. Mahallawy et al. [237] applied a 3D rigid-viscoplastic finite element analysis to Al–Cu
alloys for ECAP following routes A and Bc. Experimental and FEM results were found to be in agreement, both showing that
route A leads to greater strain homogeneity across the workpiece than route Bc. Strain localisation and tensile stresses lead
to fracture, which was simulated in the works cited above [190,191,250]. Strain rate sensitive material models were consid-
ered by several groups [191,217,228,230]. In particular, Kim et al. [220] showed that the strain rate decreases with the chan-
nel angle and the die corner angle. The magnitude of the corner angle was found to exert a much stronger effect on the strain
pattern than that of the channel angle, owing to an expansion of the deformation zone and reduced deformation time caused
by an increase in the corner angle.
Finite element analysis that included the Gurson model of damage was carried out by Lee et al. [251] to rationalize the
densification behaviour and the elimination of residual porosity in the aluminium alloy 6061 matrix containing SiC particles
by ECAP. Damage models were also included in the simulations of the ECAP processing by Lapovok [252]. The evolution of
damage (porosity) in cast Al6061 alloy was successfully predicted in terms of a stress index, which represents the ratio of
hydrostatic pressure to equivalent shear stress.
Simulations using models that account for microstructure in the context of FEM analysis are scarce. In Ref. [253], strain
distribution results were used to obtain the textures developed during ECAP. A dislocation-based strain hardening model due
to Tóth et al. [254,255] was implemented in a finite element code which enabled the dislocation cell size evolution in copper
to be predicted [256]. In a companion paper, texture evolution was calculated using the same modelling frame [257]. The
model was also successfully used in simulations of the response to ECAP processing of Al [256] and IF steel [258]. Most
recently, this model was also used for simulations of ECAP deformation of Cu and Al alongside two other dislocation-
density based models [259], and the benefits of the model by Estrin et al. [255] were highlighted. Although stress and
reaction forces are routinely available from FEM simulations, only a few workers used these results to evaluate material
behaviour, tool pressure, and the process force under severe plastic deformation [192,203,210,217,228,260]. The consequent
and successful use of the dislocation density based models in FE simulations of severe plastic deformation by the research
group of Hyoung Seop Kim (see, for instance, [261,262]) should be mentioned, but this is an exception rather than the rule.
As mentioned above, most of ECAP simulations used a two-dimensional model assuming plane strain, which had obvious
benefits in terms of the cost of re-meshing and computational time in general. However, for billets with a circular or rect-
angular cross-section in situations when friction and heat effects are significant, three-dimensional simulation is more
appropriate, as shown in several publications [146,195,198,228,230–232].
As the microstructure evolves and gets refined with increasing strain, conventional FEM codes face the mentioned neces-
sity of re-meshing and the associated difficulties with precision and efficiency, especially in handling large deformations. To
minimise this mesh dependency problem, the finite volume method (FVM) was proposed and used successfully, for example,
in multi-pass ECAP simulations [198,228,230,263]. Another problem which grows ever more challenging in the conventional
computational methods such as finite element or finite volume analysis is the problem of simulation of nucleation and prop-
agation of discontinuities such as cracks during SPD processing. In most cases, these techniques cannot handle such beha-
viour because of their strong reliance on the mesh. Consequently, their inherent structure is not well suited to dealing with
discontinuities which do not coincide with the original mesh lines. Re-meshing in each step of the simulation to keep the
mesh lines coincident with the discontinuities throughout the simulation inevitably introduces many computational
problems.
A group of numerical techniques which aim at addressing the re-meshing problem in many FEM simulations is repre-
sented by the so-called mesh-free or meshless methods. A variety of these techniques has emerged and developed to a highly
matured state in the past two or three decades [264]. Originating from particle hydrodynamics and gas dynamics, they rely
on the nodes (particles or points), not on lines (or meshes). The application of these methods to simulation of SPD processes
is still scarce, but a few successful examples of application of a rigid-plastic/viscoplastic element free Galerkin method to
simulate the plastic flow during ECAP are encouraging [265,266]. The most popular mesh-free methods are smoothed par-
ticle hydrodynamics (SPH) [267,268] and the material point method (MPM) [269]. Such models do not suffer from the prob-
lem of large mesh deformation and the necessity of re-meshing common to FE models in a Lagrangian framework, or the
difficulty of tracing the material’s history in the Eulerian formulation. Studies of ECAP processing by both SPH [270] and
MPM [271] showed the efficacy of these methods and the consistency of the results obtained by the meshless techniques
and the FE analysis.
From the above discussion it is evident that the shape of the deformation zone, strain homogeneity and ultimately the
microstructure evolution in the deformed samples are affected by several factors. These include the intrinsic material prop-
erties, particularly strain hardening behaviour, the die design, friction, and processing parameters, such as back pressure,
ram speed, and temperature. Being reasonable as a first approximation, the 2D FEM simulations assuming plain strain con-
ditions have obvious limitations when (i) very large plastic deformations are involved, (ii) the shape of the billet is other that
rectangular, (iii) friction effects are of concern, and (iv) heat effects cannot be neglected. Thus, for realistic simulations 3D
FEM is the computational tool of choice [228,272].
Because of the significance of ECAP among the SPD processing technologies, its simulations have been considered in this
section at length. FEM simulations of SPD processes other than ECAP are less numerous, while being similar to those of ECAP
in essence. Some of them will be considered in Section 5.
2.5. Strain rate and temperature effects during ECAP
Microstructural evolution during SPD involves a plethora of mutually interacting mechanisms, including dislocation glide,
recovery by dislocation annihilation, and possibly recrystallisation and grain growth. These are chiefly thermally activated
processes and, as such, they are strongly influenced by two extrinsic variables – strain rate and temperature, which are often
interrelated through thermomechanical coupling. In what follows we examine the significance of these two variables in
modelling of SPD processing in some detail.
2.5.1. Estimates for strain rate during ECAP

Berbon et al. [171] showed that the microstructure of pure Al and an Al-1% Mg alloy processed by ECAP is dependent on
the pressing speed. There have been several attempts to estimate the strain rate using different modelling approaches. Semi-
atin et al. [191] evaluated the strain rate in the ECAP process by FEM modelling, though only roughly because of the coarse
mesh size they used. Assuming the die geometry shown in Fig. 3b, Kim [135,220] provided simplified estimates of the aver-
age strain rate. The results were then compared with the accurate 3D FEM analysis performed in isothermal conditions cor-
responding to low ram velocities.
The average deformation time during ECAP can be defined in a first order approximation as the dwell time of the material
within the localised deformation zone in the geometry shown in Fig. 3b as:
Fig. 10. Average strain rate as a function of the corner angle W. Straight lines represent the results obtained from Eq. (12) for two values of the channel angle
U. The circles correspond to the data for the following combinations of the angles: (A) U = 90°, W = 7°; (B) U = 90°, W = 45°; (C) U = 90°, W = 90°; (D) U =
135°, W = 40° and (E) U = 135°, W = 45°. The ram speed to die diameter ratio ðu=ddie Þ was set at 0.0033 s1 (after [135], reprinted with permission).
ddie W
Dt ¼ pffiffiffi ð17Þ
u 2
Medeiros et al. [273] provided a more detailed derivation of the dwell time. The resultant values for Dt they obtained are
quite similar to those estimated by Kim [135], however. By combining Eq. (2) for the strain and Eq. (17) for the deformation
time, a strain rate equation for ECAP is obtained:
pffiffiffi
1 U W U W 2 u
e_ ¼ pffiffiffi 2cot þ þ Wcosec þ ð18Þ
3 2 2 2 2 W ddie
The analytical calculations according to Eq. (18) were found to be in good agreement with the results of FEM modelling, as
shown in Fig. 10.
The analytical mechanistic approaches to deformation histories during ECAP processing developed by Tóth et al. [156]
and Beyerlein and Tomé [159] are quite appealing in that they allow for explicit calculation of velocity gradients and asso-
ciated strain rate tensor components. Using the flow line representation given by Eq. (4), Tóth et al. derived the following
transparent expressions for the strain rate e_ ij components accounting for the curvature of the flow lines and the size of
the die:
e_ xx ¼ uð1 nfl Þðddie xÞnfl 1 ðddie yÞnfl 1 ðddie x0 Þ12nfl ;

e_ yy ¼ uð1 nfl Þðddie xÞnfl 1 ðddie yÞnfl 1 ðddie x0 Þ12nfl ; ð19Þ
1
e_ xy ¼ uð1 nfl Þðddie x0 Þ12nfl ½ðddie xÞnfl ðddie yÞnfl 2 ðddie yÞnfl ðddie xÞnfl 2 ;
2
(with u denoting the ram speed).
For the equivalent (von Mises) strain rate it follows
rffiffiffiffiffiffiffiffiffiffiffiffiffi
2 1
e_ eq ¼ e_ ij e_ ij ¼ pffiffiffi uðnfl 1Þðddie xÞnfl 2 ðddie yÞnfl 2 ðddie x0 Þ12nfl ½ðddie xÞ2 þ ðddie yÞ2 ð20Þ
3 3
The strain rate profile along a flow line predicted by Eq. (19) is shown in Fig. 11 as a function of the angular position of the
flow line, defined in Fig. 4. The model predicts a significant increase of the strain rate along the symmetry plane of the die
and the distribution of the strain rates appears more pointed for larger nfl . Considering that the nfl values increase nearly
monotonically from top to bottom, the model correctly captures the inhomogeneity of the strain rate distribution in the
ECAP billet. Furthermore, at variance with Segal’s original model, Eq. (19) shows that the shear strain e_ xy is non zero.
2.5.2. Estimates for temperature effects during SPD

While the strain rate can easily be calculated and modelled numerically from simple geometry and kinematics consider-
ations, the picture is more complicated with regard to the thermal effects, which require careful approximations to be made.
Since the mechanical properties of engineering materials are temperature dependent, estimates of the deformation-
induced temperature rise are important for determining a material’s response and microstructure formation. Two thermal
conditions, which determine the temperature field in a billet, are naturally distinguished in attempts to interpret and model
experimental results. In one extreme case, the heat generated within the deformation zone as a result of energy dissipation
there is efficiently removed from the deformation zone either to an external ‘thermal bath’ or directly to the bulk of a mas-
sive billet. The temperature rise in this case, which is referred to as the isothermal one, is negligible and the deformation tem-
perature is adequately represented by the initial temperature. In the other extreme, when adiabatic conditions prevail, say at
very high strain rates, the temperature can rise appreciably. As an example, in a structural HY-100 steel deformed dynam-
ically at a strain rate of 1.4 103 s1 the maximum temperatures measured by Marchand and Duffy [274] during localised
shear band formation was as high as 460 ± 50 °C. The real thermal conditions during most metal forming operations, includ-
ing SPD processing, usually lie between these two extremes. The average strain rates within the de/ormation zone during
SPD processing do not exceed 0.2–1 s1 (and usually are much lower), as can be estimated /or typical deformation condi-
tions using Eq. (18). For this strain rate range, the adiabatic conditions are hardly fulfilled and the deformation, even in shear
bands that are sometimes observed during SPD, cannot be commonly regarded as adiabatic.
The most widely used, although not always adequately argued, approach towards modelling deformation-induced ther-
mal effects and energy storage during microstructure evolution is based on a simplest energy balance analysis that follows
from the first law of classical thermodynamics. The energy balance approach has been comprehensively reviewed by Bever
et al. [275] and more recently by Stainier [276]. It states that the elementary mechanical work dA done by the external forces
is split into an increment of heat, dQ , and an increment of the internal energy of the system, dU:
dA ¼ dQ þ dU ð21Þ
Here the kinetic energy of the plastically deforming body has been ignored. It should also be born in mind that the
mechanical work is generally path-dependent and thus the elementary increments of work and heat are represented by
incomplete differentials. For simplicity, we retain the symbol d to denote the differential in all terms in Eq. (21). A portion
Fig. 11. Strain rate profile (solid lines) along a flow line for ECAP deformation for nfl = 6. An excellent agreement with the results of FEM simulations
(symbols) is seen (from[96], see also [156] for results with other nfl values; reprinted with permission).
dW el of the total mechanical work dA expended on a deforming solid is elastic and recoverable, while the remainder - the
plastic work dW pl - is irrecoverable. The latter is mostly, but not entirely, converted into heat dQ . The part of the plastic
work that is not converted into heat represents the stored energy of mechanical work dU. This energy is stored in the
crystal lattice defects created in the course of plastic deformation. Disregarding the elastic part, Eq. (21) can thus be
re-written as
dW pl ¼ dQ þ dU ð22Þ
Taylor, together with Farren and Quinney, [277,278] were the first to perform a series of experiments on the latent energy
remaining stored in copper after cold working. Without specifying microstructural mechanisms for energy storage and strain
hardening, which were unknown at that time, they concluded that ‘‘the fact that the absorption of latent energy and the
increase in strength with increasing strain both cease when the same amount of cold work has been applied suggests that the
strength of pure metals may depend only on the amount of cold work, which is latent in them.” Bever et al. [275] and Benzerga
et al. [279] emphasised that the partitioning of the plastic work into heat generation and energy storage is of interest in a
wide range of contexts. The defect structure in a cold worked ductile metal, particularly in a severely deformed one, is gen-
erally metastable so that, with time, it evolves so as to reduce its stored energy. In this sense, the stored energy of cold work
plays a significant role in driving ‘‘static” recovery and recrystallisation in polycrystalline metals [280,281]. Particularly, it
has been shown that pure FCC metals such as copper or aluminium are prone to relatively quick recrystallisation after
SPD [282,283]. For example, high purity (5N) Cu subjected to eight passes of route Bc ECAP recovered completely to the con-
ventional coarse grain state after one year storage at ambient temperature. Knowledge of the fraction of the mechanical
work converted to heat is essential for prediction of thermomechanical coupling phenomena such as thermal softening that
promotes plastic instabilities [284,285], including necking and shear banding. Thermomechanical coupling plays a critical
role in selection of the deformation mode prevalent in rapid SPD processes, such as machining and projectile penetration
where the adiabatic approximation is well justified [286]. The popular constitutive relations for high strain rate deformation
proposed by Johnson and Cook [287] from purely mechanistic considerations and by Zerilli and Armstrong [288], who
included the grain size dependence of stress and dislocation-based effects of strain hardening, strain-rate hardening, and
thermal softening, provide the necessary thermomechanical coupling. Furthermore, since the stored energy is associated
with the internal stress state of the material, it has a direct connection with the occurrence of the Bauschinger effect. Indeed,
the large energy stored in the material during SPD results in a pronounced Bauschinger effect observed under cyclic loading
in many ECAP-manufactured ultrafine grained materials, including Cu, Ti, Ni, Fe-Ni, and Al-Mg-Sc alloys, etc. [289,290].
Because of the importance of the stored energy of cold work for many physical phenomena in deformed materials, there
is a large literature aimed at its experimental determination, see, e.g., [275,291–294]. The stored energy of cold work in met-
als is associated with the production and evolution of a variety of crystal lattice defects including, primarily, lattice disloca-
tions and grain or sub-grain boundaries, as well as twins and stacking faults. The rate of accumulation of stored energy varies
with strain and strain rate, and its magnitude can usually be related to stress. Thus, the evolution of the stored energy in a
material is intimately connected with the evolution of its defect structure.
As stated above, a number of assumptions are often made for the analysis of the energy balance of a deforming solid to be
tractable analytically. These simplifying assumptions usually include an additive decomposition of a strain increment into an
elastic and a plastic part, a relation between stress and elastic strain adopted from isotropic linear thermoelasticity, and
the use of the linear Fourier heat conduction law. Under these assumptions, the first law of thermodynamics is reduced,
for the case of uniaxial loading, to the following form [292] (see also [295] for details of the derivation of this equation
and the underlying assumptions):
@T @2T
qd cv ¼ bT re_ pl aT ET e_ el þ j 2 ð23Þ
@t @x
with the total heat flux rate on the left-hand side and the elastic strain eel and the plastic strain epl being functions of the
coordinate x and time t. The material constants j, aT , and E are the thermal conductivity, thermal expansion coefficient,
and Young’s modulus, respectively. Eq. (23) implies that the heat removal from the deformation zone is controlled by heat
conduction in the material, rather than heat exchange with the environment through the specimen surface, which is
assumed to be flat. In the opposite case when the latter process is the slower of the two, the last term in Eq. (23) should
be replaced with ð2h=RÞðT T 0 Þ, where T 0 is the ambient temperature, h is the heat exchange coefficient, and R is the radius
of the specimen (assumed to be cylindrical) [296]. In this case, deformation-induced temperature increment in steady state
reads
bT re_ pl R
T To ¼ ð24Þ
2h
There is a consensus that most of the energy of plastic deformation, W pl , yet not all of it, is converted to heat. The con-
version coefficient, i.e. the fraction of the plastic energy rate bT ¼ Q_ =W
_ pl , is material and process specific, but it can typically
be set at about 0.9 [278]. The limitations on the applicability of this equation, e.g. those associated with the assumed con-
stancy of the material parameters cv , j and bT , were discussed by Tsagrakis and Aifantis [297] in connection with a gradient
thermo-viscoplasticity model for adiabatic shear banding.
The deformation-induced heat release can cause microstructural changes and affect the microstructure evolution, e.g.
through the temperature-dependent parameters governing the dislocation recovery rate (see Section 3.2).
The increment in plastic work per unit volume during deformation can be expressed as
dW pl ¼ re_ dt ð25Þ
R R
and the integral plastic work per unit volume is obtained as W pl ¼ re_ dt ¼ rde. In the case of SPD processes, this
energy per pass is quite large. Referring to ECAP of a material with a yield stress ry of, say, 300 MPa and a strain per
ECAP pass of e 1, Korn et al. [298] provided and estimate W pl 3 106 J
m3, or 0.3 J
g1, which is a considerable
figure (although the latent heat of fusion for copper is still three orders of magnitude higher). A mathematically more
rigorous solution for the power dissipated during ECAP process in different regions of the die was obtained by Paydar
et al. [138] based on the upper-bound theory. To attain ultimate grain size reduction, SPD processing is usually
performed at as low homologous temperature as possible, say, at ambient temperature, which is roughly 0.3Tm, (where
Tm is the melting temperature) for such metallic materials as Al, Cu, Ni, or Fe. Due to generation of a significant density of
dislocations in a deforming body at this temperature, SPD triggers a sequence of structural transformations that convert a
random distribution of dislocations to bundles, dislocation cells with low misorientation angles across their boundaries,
and eventually - and most importantly - to highly refined and highly misoriented grain structures [299–307]. This chain
of processes leads inevitably to a progressive increase in flow stress and an ensuing rise in the rate of energy dissipation
for SPD processing at low homologous temperatures [169]. With regard to the question of how seriously the grain size
distribution is affected by the heat developed during ECAP, Korn et al. [298] highlighted three most significant factors: (1)
the maximum temperature attained locally in the shear deformation zone during ECAP, (2) the time an element of
material spent at the site of this maximum temperature, and (3) the rate of decay of temperature during cooling of
the ECAP-processed billet.
Only few direct measurements and indirect estimations of temperature changes during SPD processing were reported.
These were summarised recently by Zhilyaev et al. [308]. For example, Yamaguchi et al. [169] used thermocouples embed-
ded in billets and measured temperature increments of 25–30 °C during an ECAP pass in initially annealed pure aluminium.
Zhilyaev et al. [309] gave an estimate of 140 °C for the temperature rise during HPT of an as-cast Na-added Al-7 wt% Si alloy
from indirect evidence, based on the observed changes in precipitate distributions. Shear plane temperatures as high as the
melting temperature Tm were reported for plane strain machining under certain process conditions [310].
Frictional losses in the ECAP process are a further source of heat compounding the deformation-induced heat release. Kim
[311] considered the friction-related heat generation as an additive term on the left-hand side of Eq. (23) and used a simple
Newtonian viscous friction model as an approximation under well lubricated conditions. In this case, the rate of heat flux
generated by friction between the die walls and the workpiece is proportional to the normal stress f s and the relative velocity
u between the bodies in contact. The respective differential is given by
pffiffiffi
dqf ¼ f s du ¼ mf ðr= 3Þdu ð26Þ
where mf is the coefficient of friction. Ignoring the contribution from the thermoelastic effect and all temperature gradients
within the main deformation zone (i.e. dropping the second and the third terms on the right-hand side of Eq. (23)) reduces
the energy balance to an elementary relation between average quantities:
pffiffiffi
qd cv V DT ¼ bT repl V 0:5mf ðr= 3ÞuSA Dt SA hT DT Dt ð27Þ
Fig. 12. Comparison of the calculated and experimental temperature increments during ECAP of Al-based materials (u = 18 mm/s or 0.18 mm/s) (after
[311], reproduced with permission).
Here V is the volume of the main deformation zone and SA its surface area. The dwell time Dt is given by Eq. (17). In writ-
ing the first term on the right-hand side it was tacitly assumed that during ECAP the stress saturates quickly to a level rep-
resented by r. The factor 0.5 in the second term stems from the assumed equipartitioning of the friction heat between the
workpiece and the die. The third term accounts for the heat transfer from the workpiece (considered to be heated adiabat-
ically) to the die. The coefficient hT, which is proportional to the thermal conductivity j, describes the efficiency of this heat
transfer. Accordingly, the following simple relation provides an estimate for the temperature increment DT in the shear
deformation zone:
pffiffiffi
bT repl þ 0:5mf ðr= 3ÞuðSA =VÞDt
DT ¼ ð28Þ
qcv þ ðSA =VÞhT Dt
From the geometrical considerations the volume and the surface area of the primary deformation zone are
3 2
V ¼ ðp2 =4Þddie ðU=2pÞ and SA ¼ ðp2 ddie ÞðU=2pÞ, respectively, with ddie denoting the diameter of the die. Using this simple
analytical formula, with SA and V estimated for W ¼ p=4 and U ¼ p=2, employing Eq. (2) for the plastic strain epl , and taking
bT = 0.9, mf = 0.2 (a typical value for cold forming of metals with conventional lubricants), and hT = 2 103 W m2 K1, Kim
[311] obtained a fair agreement between the calculated values of DT and the experimental results of Yamaguchi et al. [169]
for Al and Al-Mg alloys deformed by ECAP with different ram speeds u, Fig. 12.
Fig. 13. FEM simulation of the temperature distribution in a Cu billet during ECAP, computed with hT = 2 103 W m2 K1. The figures show the effect of
the ram speed u: (a) u = 1 mm
s1, (b) u = 10 mm
s1 (after [298], reproduced with permission).
Fig. 14. Summary of the results of FEM simulations of the maximum temperature Tmax in the shear deformation zone and temperature Tend at the deformed
end of a Cu billet, plotted as a function of the ram speed u for two different values of the heat transfer coefficient (hT = 2 103 W m2 K1 and
hT = 105 W m2 K1) (from [298], reproduced with permission).
It follows from a comparison of the first and the second terms in the numerator of Eq. (28) that the main contribution to
DT arises from the work of deformation, whereas the heat of friction is almost negligible. This result highlights the signifi-
cance of adiabatic heating in the ECAP deformation process unequivocally.
Korn et al. [298] applied the same procedure to estimate the temperature increment DT in the case of a copper billet.
Assuming a yield stress of 350 MPa (representative for UFG copper [176,290]) and a typical value of e 1, and with the same
other variables as those used by Kim, the temperature rise DT was estimated at approximately 64 °C, 89 °C and 94 °C for
ECAP speeds of 1 mm/s, 5 mm/s and 10 mm/s, respectively. These values correspond to Tmax of 84 °C, 109 °C and 114 °C. They
are in fair agreement with the values of Tmax obtained by FEM simulation, cf. Figs. 13 and 14.
Temperature increments under HPT are typically less pronounced. Thus, Edalati et al. [229] performed FEM modelling of
the temperature profiles for aluminium, copper, iron and molybdenum subjected to HPT at realistic anvil rotation speeds.
In their simulations, slippage and the associated friction heat was accounted for. A temperature rise was observed at the
early stages of straining which then saturated to steady levels of a few tens of degrees at large strains. The temperature
increment was found to be higher for harder materials, in reasonable agreement with Eq. (13). For all materials, the tem-
perature increase became more significant with increasing processing time, rotation speed, applied pressure and the
distance from the disc centre. Despite the giant plastic work done during processing by HPT, the magnitude of the temper-
ature increments for the selected model metals relative to their melting temperatures was not significant, however. It
should be noted that these advanced simulations render essentially the same conclusions as those drawn by Bridgman
in his landmark paper [160].
The simplistic thermodynamic analysis complemented with FEM simulations has shown that depending on the material
and process conditions SPD may involve an appreciable temperature increase. This is particularly true for ECAP processing. It
can thus be concluded that microstructure evolution during SPD at low homologous temperatures can be influenced not only
by the initial processing temperature but also by deformation-induced heating effects.
2.5.3. Microstructure evolution during SPD in non-isothermal conditions

Kuhlmann-Wilsdorf [312,313] has systematically advanced the principles of classic equilibrium thermodynamics to the
so-called low-energy dislocation structure (LEDS) theory of low temperature dislocation-based crystal plasticity. This theory is
essentially based on Taylor’s theory of work hardening which assumes that the external stress is related to the dislocation
structure in a unique way. The deformation-induced microstructure is supposed to correspond to a minimum-energy con-
figuration and to be in equilibrium at a given applied stress during deformation.
Although the equilibrium thermodynamics approach propagated by Kuhlmann-Wilsdorf has been criticised as one disre-
garding the dissipative, far-from-equilibrium nature of the deformation processes [314], the minimum-energy principle
seems to work well in many cases. It was tacitly employed by Zhilyaev et al. [308] who proposed a simple model for grain
size evolution under SPD-induced adiabatic heating, incorporating the temperature- and rate dependent processes. The fol-
lowing basic assumptions were made explicitly or implicitly. (1) The evolution of the mean grain size d is controlled by two
competitive processes - grain refinement with increasing strain (or time) and elastic energy storage in grain boundaries,
leading to grain coarsening that would reduce their free energy:
_ ¼ d
d _ þ
_ þ d ð29Þ
Here the time derivatives d_ and d

_ þ denote the rate of grain refinement and grain coarsening, respectively. (2) Adiabatic
conditions prevail so that the thermoelastic heating term in Eq. (23) is negligible compared with the heating due to plastic
work. (3) Based on the analysis by Kim [311] and Korn et al. [298], the friction effects can be neglected as well, provided the
appropriate lubrication conditions are fulfilled. Assuming further that most of the part of the plastic work that gets stored in
the defect structure goes into the free energy of grain boundaries (i.e. neglecting such energy storage channels as generation
of vacancies and dislocations) Zhilyaev et al. [308] found the following kinetic equation for the rate of decrease of the grain
size:

dd 2
cGB rd
¼ ð30Þ
de 3cGB
Here cGB is the fraction of the plastic work stored in the grain boundaries and cGB is the grain boundary energy per unit
area. The driving force for grain growth at a given temperature was considered to be controlled by diffusion and the kinetic
equation for grain growth with respect to time or strain can be written as
dd þ 2c X D þ 2c X D
dd
¼ GB ; ) ¼ GB ð31Þ
dt d w kB T de e_ d w kB T
where X is the atomic volume, w is the width of the grain boundary and kB the Boltzmann constant. The diffusion coefficient
D is given by D ¼ DV þ 3wd DGB , where DV is the bulk diffusivity and DGB is the grain boundary diffusivity. We note that the first
term in D gives rise to a linear and the second one to a parabolic growth law. The overall kinetics of the evolution of the
with strain is obtained by combining Eqs. (30) and (31):
average grain size d
2c X D
dd 2
cGB rd
¼ GB ð32Þ
de _ed w kB T 3cGB
The temperature variation within the sample under the assumed adiabatic conditions is then expressed as
dT ð1 cGB Þr dT dT ð1 cGB Þr
¼ n ; ) ¼ ð33Þ
de qc v de de ð1 þ nÞ qcv
where n is a purely phenomenological coefficient describing heat transfer to the exterior of the workpiece (note that it is not
identical to hT in Eq. (28)). Eq. (33) is a compromise between the general heat conduction equation, Eq. (23), and its simpli-
fied version, Eq. (28).
It should be mentioned that Eq. (32) predicts an inverse power-law relation between the flow stress and the average grain

size in steady state, when dded ¼ 0, which Pougis et al. [315] refer to as the Derby equation. The exponent of 1/3 following from
Eq. (32) is different from the one obtained by Pougis et al., which was found to be close to 1/2, thus replicating a Hall-Petch
type of grain size dependence.
In their model, Zhilyaev et al. do not restrict themselves to steady state conditions, however. Rather, they assume
that quite generally, the dependence of the flow stress on the average grain size is expressed in terms of the Hall-Petch
relation [8]:
o Þ with strain combining grain refinement and

Fig. 15. Schematic of the variation of the average grain size (normalised with respect to its initial value d
grain growth during SPD (from [308], reproduced with permission). The solid line corresponds to the resultant grain size.
r ¼ r0 þ K HP d1=2 ð34Þ
Here r0 is a ‘friction’ stress that encompasses the contributions of strengthening effects other than the grain size one. The
Hall-Petch parameter K HP is generally a strain rate and temperature-dependent quantity, which has been interpreted in
terms of dislocation processes as a microstructure-sensitive material parameter. Three major groups of models are to me
mentioned in this context: (i) the historically first group of pile-up models [20], work hardening models [316–318] and grain
boundary source models [21,319,320], which are a natural corollary of the dislocation-based kinetic approach to strain hard-
ening considered below. The Hall-Petch relation was also recovered in a study on the grain boundary Frank-Read (FR) mech-
anism [22] where it was established that thermally activated operation of FR sources, which was earlier invoked to explain
the grain size dependence of the strain rate sensitivity of the flow stress, is to be ruled out. Gryaznov [321] proposed a gen-
eralised form of the Hall-Petch relation to extend its applicability to nanostructured materials by accounting for dislocation
slip, dislocation storage and the influence of disclination-type defects. Armstrong and Rodrigues [322] have explained the
dependence of K HP on temperature and strain rate (as observed, e.g., in the work cited above [100]) by combining the dis-
location mechanics-based thermal activation analysis and the dislocation pile-up model, which results in the following
expression for the Hall-Petch parameter
1=2
MS pGbsC
K HP ¼ M ð35Þ
2ad
Here M and M S are the Taylor and Sachs orientation factors, respectively, ad 0:8 is a numerical constant and sC is the
shear stress associated with local obstacles to dislocations. The strain rate and temperature dependence of this quantity for
FCC metals with relatively high stacking fault energy, e.g. for Cu and Ni, is associated with the cross-slip shear stress, and it is
this dependence that shows in K HP . Since a large body of indirect data provides evidence that in the temperature range (0.1–
0.5)Tm the value of K HP is only weakly dependent on temperature in FCC metal and alloys, this dependence can be ignored in
modelling as a first approximation [323].
Eqs. (32)-(34) form a closed set of equations for the evolution of the average grain size and temperature during deforma-
tion (ECAP in the case considered by Zhilyaev et al.), provided the stress can be considered as a slow variable compared to d
and T. Using this model, various regimes of grain size evolution were considered with reference to the magnitude of the
parameter n. Of particular interest is the case of n ! 0 when adiabatic conditions prevail and an initial strain interval of grain
refinement may be followed by grain coarsening at larger strains. This is illustrated schematically in Fig. 15, which shows the
average grain size as a function of the cumulative strain. The results of numerical simulations for pure Cu and Al are shown in
Fig. 16 where the experimental data points and the calculated curves for the strain dependence of the average grain size are
plotted. They indicate that the original grain size is reduced by a factor of >102 at a strain of 5–6 before it increases slightly
upon further straining. The results of the simulations are seen to be in fair agreement with the bulk of experimental data for
the two materials.
Overall, the model developed in [308] shows qualitatively (and even semi-quantitatively) that SPD under isothermal con-
ditions should result in a continuous refinement of the grain size. This reflects the strengthening effect of grain refinement
and a corresponding increase in flow stress as SPD straining proceeds. Saturation of grain size refinement reflects a balance
between grain refinement due to straining and grain growth due to the sample temperature rise associated with
deformation-induced adiabatic heating.
Despite some doubts one may have about the validity of the physical argumentation underlying Eq. (33) (regarding both
the thermodynamics and the kinetics of the processes involved), the advantage of this approach is that it allows evaluating
the role of the model parameters, particularly the heat exchange coefficient n, in the evolution of the grain size during SPD.
Uncertainties and limitations used in this modelling approach stem from the simplifications adopted to make the analysis
tractable. Although this can be accepted as a first approximation, a more elaborate approach accounting for non-
equilibrium character of virtually all processes underlying the microstructure formation has yet to be proposed. Further-
more, disregarding the role of dislocations in the energy dissipation and storage is another oversimplification, and this needs
to be rectified within a dislocation based approach. In addition, such processes as grain fragmentation under SPD processing
are to be included in an explicit way, as will be discussed below, in Section 3.2.
3. Microstructure-based phenomenological modelling of severe plastic deformation
Microstructure-related constitutive models of strength and plasticity of metals and alloys are preferable to purely phe-
nomenological models. Due to their physical foundations, the former are usually more economical than most phenomeno-
logical models in terms of their architecture and the number of adjustable parameters involved. Besides, these parameters
have a clear physical interpretation. This provides the microstructure-based models with a greater predictive capability, par-
ticularly with the ability to predict how these parameters depend on the alloy composition of the material and its processing
history. A reliable foundation for microstructure-related constitutive models has been laid by the dislocation theory. Indus-
tries as diverse as automotive, aerospace, and electronics have been benefiting from the use of integrated computational
models which employ dislocation theory based approaches [324]. Because of the complexity of the mechanisms underlying
dislocation plasticity, some unresolved problems with dislocation theory based constitutive modelling still remain, however.
Fig. 16. Average grain size as a function of cumulative true strain for copper (a) and aluminium (b). The solid lines show the results of numerical
simulations (from [308], reproduced with permission).
The words of Cottrell about modelling of strain hardening, who described it some 65 years ago [325] as ‘‘the first problem to
be attempted by dislocation theory and the last to be solved” are still relevant.
Due to the progress in constitutive modelling achieved by materials scientists over the last two decades, the outlook on its
ability to tackle the SPD processing both analytically and computationally is optimistic. In this section, we shall be focusing
on the microstructure-related modelling ‘toolbox’ available for description and prediction of the behaviour of metallic mate-
rials under SPD. Process optimisation in this area relies more and more on computational modelling. Microstructure-based
models capturing the physics of the deformation processes, and yet robust enough to be used efficiently in finite element
codes, are in demand. Here we present modelling approaches that satisfy this demand. In our earlier review [326], we pro-
vided some insights into the specifics of modelling of SPD processes. In the present section we shall discuss a broader range
of modelling techniques based on the microstructure-based approaches to constitutive modelling. It will also be demon-
strated how irreversible thermodynamics of plasticity leads to the classical constitutive equations that go back to the work
of Kocks [182] (see also [23,184,327]). More complexity will then be introduced in the constitutive equations where it is
needed for an adequate description of the detail of severe plastic deformation. Examples of calculations of the material beha-
viour during various SPD processes will then be given. Throughout this section, problems that need to be resolved by future
model development will be identified and discussed.
An important ingredient of any microstructure model is a reliable description of the dislocation density evolution. In the
next section, we give an overview of an approach to such a description using irreversible thermodynamics to provide a new
interpretation of existing semi-phenomenological models.
3.1. Evolution of dislocation ensembles: An irreversible thermodynamics approach
The foregoing analyses of the grain size evolution were footed on the first law of classical equilibrium thermodynamics. In
reality, however, a plastically deforming solid is an open system exchanging energy with its surroundings. Therefore, regard-
ing plastic deformation as a highly dissipative process [328] suggests that the irreversible thermodynamics principles for-
mulated by Onsager, Prigogine and Bridgman [329–332] should be applied as the governing concept. As distinct from
equilibrium thermodynamics, the second law of thermodynamics of irreversible processes is formulated in terms of the
entropy production. A concept of fundamental importance in thermodynamics of irreversible processes is the notion of a
non-equilibrium stationary state, which plays a role analogous to that of the equilibrium state in classical equilibrium
thermodynamics. According to the second law, all internal processes must terminate when collectively they can no longer
produce entropy. A stationary state has then been reached and it represents a minimum in the rate of entropy production,
which is zero.
Being a highly dissipative process, plastic deformation involves many scales and manifests itself in a variety of patterns.
Characterising and predicting patterning in the dislocation population that accompanies plastic deformation of crystalline
solids is a particularly challenging problem addressed in numerous publications [60,333–335]. The patterns may be different
with regard to their inner length scale or, by contrast, exhibit a scale-free behaviour [328].
For a system undergoing multiple irreversible processes, the entropy di S generated per unit volume during a time interval
dt is the sum of the individual contributions of all irreversible processes involved. This is expressed mathematically as
X
N
di S ¼ di SK > 0 ð36Þ
K¼1
where N is the number of irreversible processes taking place in the system and di SK is the entropy generated due to the Kth
process. di SK is determined by the dissipated energy per unit volume W Kdis associated with the Kth process [280]:
K
dW dis
di SK ¼ ð37Þ
T
In Onsager’s theory [329,330] the rate of entropy production in a thermodynamic system out of equilibrium, but where
local equilibrium exists, is expressed as the sum of the products of generalised forces X K and their corresponding thermo-
dynamic fluxes J k for the individual irreversible processes involved:
di S X N
¼ X k Jk > 0 ð38Þ
dt k¼1
Onsager proposed a linear relation between thermodynamic fluxes and the generalised forces in the form
P
Jk ¼ Lki X i ðLki ¼ Lik Þ where Lki is the matrix of kinetic coefficients, which are independent of J k and X i . Finding these coef-
ficients is a formidable task. To derive a relation between mutual interactions of simultaneous processes in a system, an
extremum principle of irreversible thermodynamics is usually invoked, e.g. Onsager’s principle of maximum rate of entropy
production or Prigogine’s principle of minimum rate of entropy production.
Assuming a local equilibrium in the system, Ziegler [336,337] proposed a maximum entropy production principle to find
X k ðJ k Þ in the explicit form: if irreversible force X i is prescribed, then the actual flux J i , which satisfies the condition (36), maximises
the entropy production. A similar principle, originally proposed by von Mises in the solid mechanics context, appeared in the
mathematical theory of plasticity developed by Drucker [338], Hill [339] and others in the 1950s. Referred to as the von
Mises principle of maximum plastic dissipation [340], it states that the total power of plastic deformation per unit volume
is maximised subject to the yield condition expressed in terms of the stress and strain hardening characteristics for a given
strain rate.
Ghoniem et al. [341] adapted the general irreversible thermodynamics formalism to the discrete dislocation dynamics
(DDD) modelling framework. Benzerga et al. [279] combined the irreversible thermodynamics with the discrete dislocation
dynamics approach, which enabled them to calculate the stored energy of cold work for planar single crystals under quasi-
static tensile loading.
A central theme of a series of recent communications by Langer [10,342,343] was the need for physics-based, nonequi-
librium analyses in developing predictive dislocation-based theories of strain hardening of polycrystalline materials. He and
his co-authors developed a statistical thermodynamic dislocation theory, which has then been used for the analysis of the
mechanical response of pure copper and the effect of grain size on the strain hardening curves. It was proposed that the
description of the dynamic behaviour of polycrystalline aggregates can be furnished with two main intrinsic variables -
the total dislocation density and the characteristic temperature v defined for any dislocation configuration in the statistical
thermodynamic sense as v ¼ dU C =dSC with U C and SC being the energy calculated for a given ensemble of dislocations and
the corresponding configurational entropy, respectively. A good predictive capability of the elegant theory put forward by
Langer will certainly provide a strong impetus to further modelling work.
The irreversible thermodynamics framework was also used in our treatment of plastic deformation [344]. The
deforming crystal was considered as an open system evolving towards a steady state. The differential of the total entropy
flux was presented as dS ¼ di S þ de S, where di S is the entropy production term associated with the changes in the inner
microstructure (and is always positive) and de S is the term corresponding to heat exchange with the exterior. Huang et al.
[345,346] applied Prigogine’s principles of irreversible thermodynamics [331] to dislocation-based modelling of strain
hardening. This approach was modified in our work using a composite-like microstructural model, in which the
dislocation cell interiors and dislocation cell walls are treated as separated ‘phases’ of the material. In both models,
the elementary entropy production di S during a shear strain increment dc is related to the energy dissipated due to three
primary dislocation reactions - generation, motion, and annihilation entering the right-hand side of Eq. (38) additively,
yielding

1 1 2 þ 1 2
di S ¼ Gb dq þ sbhkidqþ þ Gb dq ð39Þ
T 2 2
Here dqþ and dq are the increments of the total dislocation density associated with dislocation production and recovery
(annihilation), respectively. (Note that dq is a negative quantity.) Further quantities entering Eq. (39) are the magnitude of
the dislocation Burgers vector, b, the shear modulus, G, the resolved shear stress, s, acting in the dislocation glide plane, and
pffiffiffiffi
the dislocation mean free path, hki. The latter variable scales with the average dislocation spacing as hki 1= q with
R R þ
R
q ¼ dq ¼ dq þ dq being the total dislocation density [38]. The dislocation density is the principal internal variable
pffiffiffiffi
that determines the resolved shear stress s, which scales with q according to the Taylor relation:
pffiffiffiffi
s ¼ aGb q ð40Þ
Here a is a numerical factor depending on the dislocation arrangement and the mode of deformation (typically 0.1–0.4)
[184,314]. The value of a tends to decrease as deformation proceeds and the dislocation pattern becomes heterogeneous
with increasing lattice misorientations arising from an increasing density of GNDs as has been discussed by Mughrabi
[347]. A ‘friction stress’, which stems from interactions of a gliding dislocation with the Peierls relief and crystal lattice
defects other than dislocations, can be assumed to be negligible, at least for pure FCC metals after a sufficiently large strain,
for which reason it has not been included in the last expression. Eq. (40) is ubiquitous: virtually all conceivable dislocation-
dislocation interaction mechanisms lead to this relation (with a being dependent on the specific mechanism). It is also
supported by a huge body of experimental data and discrete dislocation dynamics computer simulations [119]. Combining
Eqs. (39) and (40) one obtains the entropy production term:
2 2
ð1 þ 2aÞ Gb ð2 þ 2aÞGb
di S ¼ dq þ dq ð41Þ
2T 2T
The entropy flux de S is produced by the dissipation of most of the plastic work sdc as heat [331] and is given by
sdc
de S ¼ ð42Þ
T
In terms of the evolution of a dislocation population, strain hardening is seen as a result of the competing processes of
generation and annihilation of dislocations. The average distance a dislocation travels before it gets annihilated (‘the mean
free path’) is a key characteristic that governs strain hardening. It may be related to the average dislocation spacing or some
geometry features of microstructure extraneous to the dislocation population. Dislocation patterning is known to start
developing early on, already after a small strain following the onset of plastic flow, and it obviously affects the evolution
of the mean free path. Details of these processes are not well understood [348], but activation of multiple slip systems is
seen as a major factor leading to patterning, cf. [119]. Multiple slip systems are activated already in Stage II of strain hard-
ening giving rise to tangled dislocation networks, which further evolve to stable dislocation cell configurations. These
patterns, which emerge as a result of irreversible self-organisation, comprise dislocation-rich ‘cell walls’ separating cell
interiors that have a lower dislocation density. This kind of structure offers itself to a description in terms of the mentioned
‘composite’ model [349,350] that was employed in our work [344].
According to the model [344], the dislocation structure is composed of the cell-wall ‘phase’ with the volume fraction fw
and the cell interior ‘phase’ with the volume fraction fc. Here the subscripts c and w stand for cell interiors and cell walls,
respectively. Obviously, f w þ f c ¼ 1 holds. Applying a rule of mixtures and the Taylor relation, Eq. (40), for each of the
two phases, the shear stress s is expressed as a weighted sum of the contributions of the two ‘phases’
pffiffiffiffiffiffi pffiffiffiffiffi
s ¼ f w sw þ f c sc ¼ f w aGb qw þ f c aGb qc ð43Þ
and qw and qc are the dislocation densities in the two phases.
An empirical fact based on a vast body of experimental evidence, mainly from transmission electron microscopy (TEM)
work, is that the average dislocation cell size dc scales inversely with the square root of the total dislocation density
q ¼ f w qw þ f c qc . This relation can be expressed [351] as q ¼ a=d2c , where a is a proportionality coefficient. A similar scaling
2
relation holds for the dislocation density in the cell walls, qw ¼ aw =dc , where aw is another proportionality constant.
Combining the last three equations, one obtains qc ¼ f qw ¼ bqw - a proportionality relation between qc and qw , which
aaw
c
replicates that reported by Mughrabi [349]. Replacing qw with q in Eq. (43), which is legitimate due to the condition
qw qc , now leads to the following expression for the shear stress:

f w pffiffiffi pffiffiffiffiffiffi
s ¼ aGb þ bf c qw ð44Þ
dc
Accordingly, Eq. (42) can be rewritten as:

pffiffiffi pffiffiffiffi f w aGbdc
de S ¼ bf c q þ ð45Þ
dc T
By combining Eqs. (41) and (45) one obtains

dS ð1 þ 2aÞ Gb dq ð2 þ 2aÞGb dq
2 2 pffiffiffi pffiffiffiffi f w aGb
¼ þ bf c q þ ð46Þ
dc 2T dc 2T dc dc T
As an elementary annihilation event involves two dislocations, the annihilation rate obeys an equation for second-order
annihilation kinetics:
dq 2 DG
¼ q2 b m0 e kB T ð47Þ
dt
The pre-exponential factor m0 in this equation is of the order of the Debye frequency mD . The activation energy DG is asso-
ciated with that for dislocation climb, which is considered to be the governing annihilation mechanism. Hence, DG can be
identified with the activation energy for self-diffusion. A simplifying assumption hidden in Eq. (47) is that annihilation
occurs primarily in the cell walls. The density qw of participating dislocations was replaced with q. Relating the shear rate
to the average dislocation glide velocity hv i through the Orowan relation c_ ¼ qbhv i, one obtains
dq qbm0 kDGT
¼ e B ð48Þ
dc hv i
An equation for the rate of variation of the overall entropy is finally obtained by substituting of Eq. (48) in Eq. (46):
pffiffiffi
dS ð1 þ 2aÞ Gb dq ð2 þ 2aÞGb qbm0 kDGT aGb
2 2
pffiffiffiffi f
¼ þ e B bf c q þ w ð49Þ
dc 2T dc 2T hv i dc T
As seen from Eq. (49), this rate is determined by the instantaneous value of the dislocation density and its shear strain
derivative.
The thermodynamic system under consideration evolves to its steady-state. Even though this steady-state is a non-
equilibrium one, it is defined by the condition ddSc ¼ 0, thus yielding:
pffiffiffi pffiffiffiffi
dq 2af w 1 2a bf c q 2 þ 2a bm0 kDbGT
¼
þ

e q ð50Þ
dc 1 þ 2a bdc 1 þ 2a b 1 þ 2a h v i
Eq. (50) has a familiar form: it recovers the evolution equation for the dislocation density obtained by Kocks, Mecking and
Estrin (KME) [23,184,327,352] in a semi-phenomenological way:
pffiffiffiffi
dq k0 k1 q
¼ þ k2 q ð51Þ
dc bhki b
with the notation
pffiffiffi
2af w 2a b f c 2 þ 2a bm0 kDGT
k0 ¼ ; k1 ¼ ; k2 ¼
e B ð52Þ
1 þ 2a 1 þ 2a 1 þ 2a hv i
The coefficients k0, k1 and k2 of the KME model can now be re-interpreted in terms of the parameters of the two-phase
model via Eq. (52). The dislocation density evolution model in this simplified single internal variable form was presented in a
number of publications [184,352]. It has generated a substantial literature, in which the model was further developed, cf.
[353–355].
In the one-internal-variable approach presented, the variation of the single internal variable, the average dislocation den-
sity, with strain can be found by integrating Eq. (51). The stress–strain curve in the Stage III strain hardening range can then
be obtained using Eq. (40). The parameters k1 and k2 entering these equations can easily be found from strain hardening data
[23]. Further aspects of the dislocation-density based, one-internal-variable models are discussed in Section 3.2. An adequate
description of deformation stages beyond Stage III requires the use of a two-internal variable approach [255] where the evo-
lution of qw and qc is considered in greater detail, cf. Section 3.3.
Concluding this section, it can be stated that the irreversible thermodynamics approach to dislocation-based plasticity of
metals leads to a simple description of the dislocation density evolution and strain hardening. It recovers the well-known
evolution equation of the one-internal-variable KME model [23,182,184,327,356,357], but redefines the coefficients in that
equation.
3.2. One-internal-variable models
A seminal model, which has set the scene for microstructure-based plasticity modelling and remains influential to the
present day, is the Kocks-Mecking model [23,182,184,327] that has been mentioned above repeatedly. The model hinges
on the idea that the shear stress s is determined by a dislocation contribution proportional to the square root of the total
dislocation density q. This dependence is given by Eq. (40), whose origin in the thermally activated dislocation glide is
not obvious from the commonly used power-law strain rate dependence of the factor a:
1=m
c_
a ¼ ao _ ð53Þ
c0
(Note that the strain rate sensitivity index m is the inverse of the exponent m⁄ in the Ludwig-Hollomon Eq. (9). The con-
stant a0 represents the athermal resistance to dislocation glide, i.e. the strength of dislocation-dislocation interactions at
absolute zero temperature. The thermally activated character of the dislocation glide [288,352,358] is accounted for by
the temperature dependent parameters m and c_ 0 .
pffiffiffiffi
It is easily recognised [356] that the relation between c_ and s that follows from Eqs. (40) and (53), s ¼ ao Gb qaðc_ =c_ 0 Þ ,
1=m
is equivalent to a physically motivated Arrhenius equation [358], if the exponent m is linked to the activation volume V a for
the underlying thermally-activated process:
m ¼ V a s=kB T ð54Þ
This interpretation means that at sufficiently low temperatures m is inversely proportional to the absolute temperature T
and is much greater than unity, while c_ 0 is temperature independent. This low temperature case (roughly below half the
melting temperature) corresponds to the dislocation glide controlled regime. At sufficiently high temperatures, where dis-
location controlled plasticity prevails m can be taken to be a constant; an Arrhenius-type temperature dependence then
resides in c_ 0 [23,182,184,327,356].
pffiffiffiffi
In addition to the kinetic equation, s ¼ ao Gb qaðc_ =c_ 0 Þ , a further equation describing the evolution of the dislocation
1=m
density is required. Eq. (51) provides a general structure of such an evolution equation. The main microstructure-related
quantity in this equation, the dislocation mean free path hki, can be associated with the average spacing between the dislo-
cations. Alternatively, once a dislocation cell structure has been formed, hki can be associated with the dislocation cell size.
pffiffiffiffi
Both characteristic lengths are inverse in q, so that the evolution equation for q can be written as
dq k1 pffiffiffiffi
¼ q k2 q ð55Þ
dc b
The proportionality coefficient k1 is, of course, different for these two cases, but the mathematical form of the evolution
equation is the same. As a matter of fact, the transition from a random dislocation distribution to a self-organised cell struc-
ture can be describes through a suitable ansatz for a strain dependence of k1 .
Here a historical note is due. While this type of model is commonly associated with the names of Kocks and Mecking,
some other authors also deserve credit for introducing the dislocation density evolution approaches. In particular, a tribute
should be paid to Janusz Klepaczko who has been developing similar concepts independently [359]. A cognate model was put
forward by Malygin [360]. Yngve Bergström has also been promoting the use of an evolution equation akin to Eq. (51) in its
simplest form:

1
dq=de ¼ M k2 q ð56Þ
bhki
considering the case of a constant dislocation mean free path hki [361]. Gottstein and Argon [362] have extended Eq. (56) by
considering the effect of subgrain boundary motion on the dislocation density evolution.
The dislocation mean free path hki in this equation can be governed by obstacles of various kinds at which gliding
dislocations get immobilised. This is the term through which a lot of physical metallurgy can be captured. Thus, hki can
Fig. 17. Results of least square fitting of the experimental true stress vs. true strain curve for Cu deformed by 12 ECAP passes (route Bc) at room
temperature to the curves predicted by the Kocks-Mecking and Estrin-Mecking models (after [176], reproduced with permission).
be determined by the average spacing between non-shearable particles, the average distance between twins, the lamellae
spacing in a lamellar structure, or the grain size, to name just the most common types of obstacles to dislocation glide.
The latter case is of particular interest in the context of deformation of ultrafine grained materials produced by severe plastic
deformation. However, identifying hki with the average grain size in modelling SPD processing is only sensible if the grain
size is already saturated and does not vary in the process of deformation or if its variation with strain is known or can be
modelled [363]. Holmedal [364] has recently redefined the mean free slip length enabling a straightforward geometrical
interpretation that can be directly compared to the grain size. He suggested that the distribution of obstacles in the slip plane
can be characterised by a fractal dimension so that the athermal storage rate of dislocations in Eq. (56) is proportional to
hkiDF 2 , which is particularly relevant for fine grained materials (a similar assumption was made earlier by Hähner et al.
[365] who argued that dislocation cell structures can be regarded as self-similar fractals characterised by power-law distri-
butions of cell sizes, thus modifying the dislocation storage term in the KME equation. This idea was further promoted by
Vinogradov et al. [344] who demonstrated the significance of the fractal dimension as an informative index to follow the
spatial evolution of dislocation structures. With increasing grain size the dependence on the fractal dimension becomes
weaker and it can be neglected for coarse grained pure metals where the classical interpretation of hki applies.
Examples of successful application of the KME model that employs an evolution equation for the dislocation density in
the form of Eq. (51) can be found in [23,182,184,327,356]. Dalla Torre et al. [176] have tested the Kocks-Mecking and
Estrin-Mecking model approach in an attempt to fit experimental tensile stress-strain curves of commercially pure copper
deformed by ECAP to various strains (up to the cumulative equivalent strain of 16). As demonstrated, both models are appli-
cable almost equally well, and both show an excellent agreement of model predictions with experimental data in Fig. 17.
An important aspect of the mechanical behaviour of UFG materials is their tensile ductility. The one-internal-variable
model turns out to work well in predicting this property. If elongation to failure is controlled by the onset of necking, the
corresponding necking strain can be calculated with the aid of linear stability analysis of Eq. (51), which governs the evolu-
tion of the dislocation density [366]. In a recent paper [367] we looked at the condition for tensile instability following from
the KME model and found that for strain-rate insensitive materials (1=m ! 0) it reduces to the well-known Considère cri-
terion (see [172] for a general review of plastic instabilities and their modelling). For non-zero 1=m, the instability condition
is akin to (yet not identical with) Hart’s criterion. For the case when Eq. (56) holds, the tensile instability condition, in its
simplified form, can be written as
h 1 r0
þ 1 ¼1 ð57Þ
r me BEM r
The new acronym ‘BEM’ introduced here refers to the model by Bergström [361] and Estrin & Mecking [23], correspond-
ing to k1 ¼ 0, which is most appropriate for the case of small grain size d. (For this case, the dislocation mean free path hki is
identified with d.) The quantity eBEM denotes the strain at the onset of necking and r0 , as above, is a ‘friction’ stress at the
onset of plastic flow (i.e. at the elastic-plastic transition). In the Kocks and Mecking (KM) formulation, k0 ¼ 0 in Eq. (51),
the instability criterion reads as

h 1 1 r0
1 þ 1 ¼1 ð58Þ
r m me KM r
The main point made in Ref. [367] is that both conditions, Eqs. (57) and (58), are primarilty determined by the dynamic
recovery coefficient k2 . Analysis of the stress-strain curves performed by Dalla Torre et al. [176] using a non-linear curve fit-
ting has demonstrated that the magnitude of k2 for copper specimens after ECAP is an order of magnitude larger than for the
initial (non-deformed) coarse grained material. The role of the dynamic recovery coefficient k2 was evaluated in Ref. [367] for
UFG copper processed by ECAP to different numbers of passes, up to N = 12 [176,368]. The experimental data and the model
predictions were consistent. Both showed an increase in ductility of UFG copper with increasing number of ECAP passes. The
effect was attributed to the observed decrease of k2 with the progress of straining leading to a UFG microstructue with a large
proportion of high-angle grain boundaries. In the analysis given in Ref. [367], all model parameters enter the expressions for
the onset of necking. Remarkably, only one of them, k2 , appears in the pre-logarithmic factor, while the other ones (r0 , k0 ,
and k1 ) enter logarithmically. This becomes obvious from the explicit relations for the necking strain, eN0 , predicted by Yas-
nikov et al. Ref. [367] for the case when the Kocks-Mecking model applies:
0 1
k2 M
2 B 1þ C
eKM
N0 ¼ ln @ 2
r0 A ð59Þ
k2 M 1þ k =bhki
aGbM 1 k
2
An equation for the BEM model similar to Eq. (59) in spirit, but more cumbersome mathematically, can be found in the
original paper [367].
Thus, quite generally, among all model parameters, k2 is the primary parameter that controls the necking strain. This is
reflected in the graph for the necking strain eN0 as a function of k2 shown in Fig. 18. The graph compiles the experimental
data on copper available in literature [176,369] and in the authors’ own database. All data points are seen to fall on the same
master curve, which is represented by a hyperbola, with good accuracy. An excellent agreement between the observed
Fig. 18. The calculated strain to the onset of necking (solid line) and the experimental data points as a function of the dynamic recovery coefficient k2 for
coarse grain (CG) and ultrafine grained (UFG) copper processed by ECAP. The calculated curve corresponds to a hyperbolic relation between eN0 and k2 M, cf.
Eq. (59). The low tensile ductility of UFG copper after processing is seen to improve with the number of ECAP passes (cf. similar diagrams in [367,383]).
trends and the model predictions is evident. Similar results confirming the veracity of model predictions were reported for
UFG copper and 316L austenitic steel [370,371].
In line with the early work [372,373] and more recent detailed investigations [344,374–376], these results underscore
the importance of the dynamic recovery parameter k2 in the evolution of the dislocation density towards a steady state
and the onset of necking that aborts this development. The experimental value of the phenomenological coefficient k2 can
readily be obtained from uniaxial tensile test data, e.g. by using a recipe proposed in [23] (cf. also [366,367,370,377]).
Assuming a specific microscopic dynamic recovery mechanism, the analytical form of the coefficient k2 ¼ k2 ðc_ ; TÞ can also
be obtained from theory. For example, Huang et al. [378], who modelled plastic deformation of metals in terms of the
Friedel-Escaig compact cross-slip mechanism [379], arrived at the following expression for the rate of dynamic recovery
in FCC metals:
" ! #
4 4
Gb mD Gb mD s Va
k2 ¼ exp A ln þ ð60Þ
8p vSFE V a c_ 16p vSFE V a c_ Gb
3
Here vSFE denotes the temperature dependent stacking fault energy and A is a numerical constant (of the order of unity)
for a given temperature.
With regard to the main object of this review, viz. ultrafine grained materials, it is instructive to note that very high values
of k2 , on the order of several hundred, were reported for UFG materials produced by SPD, for which the KM approach to mod-
elling of strain hardening of Cu [176], Ti [380], Al-Mg alloy AA5056 [381,382], and 316L austenitic stainless steel [370] was
applied. The predicted trends can easily be verified using the experimental data for these materials.
Because of the thermally activated nature of dynamic recovery, the coefficient k2 is temperature and strain rate depen-
dent. This dependence is represented in the form
n1
c_
k2 ¼ k2 ðc_ ; TÞ ¼ k2 ðTÞ ð61Þ
c_ 0
The parameter n in the exponent is, like m, much greater than unity [384]. To validate Eq. (61), we tested well annealed
coarse grained copper and iron polycrystals with approximately the same grain size of 200 lm. In the strain rate range from
5 104 to 5 102 s1, k2 did obey the k2 c_ 1=n relation. The value of n⁄ was found to be about 100, which is close to the

exponent m entering the strain rate dependence of the flow stress through Eqs. (40) and (53) [184]. Similar results were also
obtained for Cu tested over a wider range of strain rates from 104 to 1 s1, Fig. 19 [184] (cf. Fig. 3 therein). The validity of Eq.
(61) can thus be regarded as confirmed.
The differences in the deformation behaviour of FCC and BCC materials stem primarily from different crystallography of
dislocation glide and cross-slip, as well as from the significance of the double-kink mechanism of dislocation motion in the
Peierls relief in BCC metals and alloys that will be discussed below. As a result, dislocations in BCC metals show greater
propensity to annihilate (cf. the thermostatistical analysis of recovery processes in FCC and BCC metals performed in
[376] and [375], respectively). This implies that the coefficient k2 for BCC metals should be higher, which is, indeed, observed
experimentally. Indeed, as shown in Fig. 20 where data for polycrystalline Cu and Fe are compared, an excellent agreement
with Eq. (59) is found: the smaller k2 , the greater is the uniform elongation up to the onset of necking, as given by eN0 . The
expected strain rate dependence of k2 (Eq. (59), see also [367] and Fig. 19), is also confirmed by Fig. 20.
Fig. 19. The dependence of the dynamic recovery coefficient k2 on the strain rate for polycrystalline copper showing agreement with Eq. (61) (from [366],
reprinted with permission, experimental data adopted from [184]).
Fig. 20. Experimentally found relation between the strain to the onset of necking, eN0 , and the dynamic recovery parameter k2 for pure polycrystalline
copper and iron tested at different strain rates indicated on the graph. An excellent agreement with predictions by [366] is noted: the smaller the magnitude
of k2, the greater is the uniform elongation as represented by eN0 (from [366], reprinted with permission).
Obviously, Eq. (55) does not apply when the dislocation mean free path is controlled by extraneous obstacles, rather than
dislocation-dislocation interactions. The length scale determining hki can be associated with such microstructural features as
twin spacing, particle spacing in dispersion-hardened materials, lamellae spacing in lamellar materials, etc. In this case, the
evolution equation, Eq. (55), assumes the form dq=dc ¼ 1=bhki k2 q [380,382,384], which is mathematically equivalent to
Eq. (56). For the ultrafine-grained materials, which are of main interest here, hki can be identified with the average grain size
d or dislocation cell size dc .
In this case, as well, the onset of necking is governed by the dynamic recovery coefficient k2 . The conclusions about the
preponderance of k2 as the control parameter in the instability condition still apply. Hence, high k2 values are responsible for
extremely small uniform elongation commonly observed for UFG metals [4,384].
The significance of the work by Yasnikov et al. [366,367] is seen in the recognition that the Considère condition for neck-
ing can be derived from the evolution laws for the dislocation density – the governing internal variable of a microstructure-
based constitutive model. The explicit form of the necking strain predicted on the basis of the Kocks-Mecking dislocation
model highlighted the leading role of dynamic recovery rate in the onset of necking. By applying this approach to the case
of strain rate sensitive materials, a modified instability condition, Eq. (58), with an added stabilising term proportional to
1=m, was derived.
It is remarkable that essentially the same one-internal-variable approach that was considered for monotonic deformation
turns out to be applicable to describing the material behaviour under cyclic loading. As a matter of fact, despite the complex-
ity of the microstructures created by SPD, the cyclic behaviour of metals with a resulting UFG microstructure permits a sim-
pler description than in the case of conventional coarse-grained poly- or monocrystalline materials. The reason is the
absence of dislocation patterning within the UFG structures. Indeed, the characteristic dimensions of the major structural
elements - grains or dislocation cells - are smaller than the characteristic length scale of would-be dislocation structures that
would form by self-organisation during cyclic loading. Vinogradov et al. [380,382] suggested that in light of this argument it
is sensible to describe the shape of a stable hysteresis loop and the cyclic stress–strain curve of UFG materials in terms of the
one-internal variable approach. Assuming, as above, that the dislocation mean free path hki is controlled by the grain (or dis-
location cell) size, integration of Eq. (51) with k1 ¼ 0 and the initial condition qð0Þ ¼ q0 yields:
k0
qðeÞ ¼ ð1 eMk2 e Þ þ q0 eMk2 e ð62Þ
k2 bd
Combined with Eq. (31), this relation provides a simple analytical expression for the flow stress as a function of strain:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
k0
rðeÞ ¼ MaGb ð1 eMk2 e Þ þ q0 eMk2 e ð63Þ
k2 bd
where M is the Taylor factor, which connects the shear-related quantities with the axial ones: dc ¼ Mde and ds ¼ dr=M. To
approximate the cyclic hysteresis loop, Vinogradov et al. [380,382] applied a slightly different variant of the KME model. It
reduces to the model originally proposed by Essmann and Mughrabi [372] for dislocation annihilation:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 expð2yc=bÞ
s ¼ aGb ð64Þ
dy
Here y is the ‘annihilation length’ for the dislocation density evolution during stable cyclic deformation. It represents a
capture radius of a dislocation residing in a grain boundary within which trapping of a passing dislocation would occur.
The model involves the assumption that mobile dislocations generated at a grain boundary pass through the grain and dis-
appear at the opposite grain boundary. This assumption is supported by molecular dynamics simulations [385–388] which
demonstrated the efficacy of grain boundaries as sources and sinks for dislocations. A similar picture is provided by an elab-
orate analytical dislocation bow-out model for dislocation nucleation at grain boundaries in ultra-fine grain materials [320].
It considers the following stages of the process: (i) a dislocation is emitted from a grain boundary source, (ii) it bows out
between the pinning points located at the boundary, (iii) yielding occurs when the dislocation takes a semi-circular shape
Fig. 21. Ascending parts of the stable hysteresis loop of pure 99.96% copper poly- and single-crystals subjected to one ECAP pass through a 90° die at
ambient temperature. Curve fits by Eq. (63) are shown by solid lines. The references to the data used are given in [380,382].
Fig. 22. Fragments of the cyclic hysteresis loops obtained for a {110} oriented copper single crystal after one ECAP pass. Cyclic softening is illustrated by the
reduction in the stress amplitude with increasing number of cycles; curve fits by Eq. (63) are shown by solid lines (a); the EBSD map (b) and the secondary
electron image (c) of the same single crystal after the first ECAP pass do not exhibit any high angle grain boundaries; the inset shows a (111) pole figure
[26,28] (reproduced with permission).
and (iv) the dislocation breaks away from the source and travels across the grain thus producing an increment in plastic
strain. TEM evidence also suggests that during cyclic loading no dislocation accumulation occurs within the fine grains.
Indeed, TEM observations did not reveal substantial differences between the initial and the post-fatigue structures in pure
FCC metals (provided no fatigue-induced dynamic grain coarsening occurs [389,390]). A good agreement of the model pre-
dictions and the experimental results for UFG Al-Mg alloy AA5056 [380,382] and titanium [380,382] produced by ECAP was
found, cf. Figs. 21 and 22. This is surprising in view of the simplifications made. A major simplification was that back stresses
associated with plastic incompatibility between the two ‘phases’ were not considered. A characteristic ascending part of the
stable hysteresis loop of the ECAP-processed poly- and monocrystalline copper, cf. Fig. 21a, is similar to that reported in Refs.
[380,382] for AA5056. An excellent agreement between the experimental hysteresis loops and the experimental data fitted
by the functions given by Eq. (63) or Eq. (64) is remarkable.
An ideal object for trialling various dislocation-based strain hardening models is a single crystals whose crystallographic
orientation with respect to the processing tool is well controlled in experiment. With this in mind, copper single crystals
with a {110} initial crystallographic orientation were pressed through a 90° ECAP die [26,28]) and then subjected to uniaxial
push-pull cyclic loading. Figs. 21 and 22 show the results of these tests. It is seen that during a single ECAP pass no high-
angle grain boundaries were formed, Fig. 22b. The observed microstructure characterised by a well-developed cell structure
[391,392] is consistent with fragmented structures formed at large strains that were found by early experimental observa-
tions [393,394]. A very large strengthening effect, which was found comparable in magnitude with that for copper polycrys-
tals of the same purity deformed by ECAP under the same conditions, was obviously achieved entirely due to dislocation
storage in a network of cell walls with low angles of misorientation. Not surprisingly is that such dislocation ensembles tend
to recover during cyclic plastic deformation giving rise to a well-known cyclic softening. The KME model correctly captures
this phenomenon as well by appropriate setting of initial conditions for the dislocation density [380,382]. Furthermore, the
cyclic stress-strain curve characterising the cyclic hardening or softening response for conventional polycrystals and fine-
grained specimens manufactured by SPD can also be approximated by an equation of the same type, cf. Fig. 21b. Klemm,
Mughrabi and Höppel [395,396] applied essentially the same model for cyclic deformation of UFG nickel processed by var-
ious SPD techniques. Both groups of researchers arrived at the same, in effect, conclusions. They also obtained similar values
for the dislocation mean free path entering Eq. (64) explicitly as d and the annihilation distance, y, which is contained in k2
implicitly. A further example of a successful interpretation of the stable cyclic hysteresis loops in terms of the KME approach
was reported for UFG titanium fabricated under different SPD processing schedules [4,397]. A reasonably good agreement
between the model fit and the experimental data was observed, similar to a good accord between modelling and experiment
found in Refs.[380,382,395,396]. An interesting detail is that in the low strain limit, Eq. (63) reduces to
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðk0 k2 q0 ÞM e
r MaGb þ q0 ð65Þ
bd
1=2
which represents the Hall–Petch type strengthening: a r / d dependence of the flow stress at a fixed strain is recovered,
in agreement with the experimentally observed stress-strain behaviour for both monotonic and cyclic modes of deformation
and in harmony with a group of strain hardening models of the Hall-Petch relation [316–318]. The fatigue limit of many met-
als with UFG structure: produced by SPD also follows the same trend [380].
Fig. 23. Size of a structural element in Cu deformed by HPT to giant strains at room temperature as a function of equivalent strain (adapted from [402],
reprinted with permission).
An approach more rigorous methodologically would require accounting for the composite-like structure comprised of
dislocation-rich walls and dislocation-depleted cell interior. This would encourage invoking two-internal variable models
distinguishing between at least two kinds of dislocations involved: ‘cell’ and ‘wall’ (or ‘mobile’ and ‘immobile’) dislocations
differed by their density and mobility, as outlined above. A proper account of back stresses is also a necessary ingredient of
any model aiming at describing cyclic plastic deformation as has been illustrated in a model by Estrin et al. [398] for ordinary
polycrystals).
As mentioned above, the quality of the model predictions based on the assumed dislocation density evolution expressed
by Eq. (51) is surprisingly good, given that this evolution equation does not reflect any specificity of UFG materials. Bouaziz
et al. [399] modified this model to provide it with the specificity the original one-internal-variable model was lacking. The
premise they used was that there exists a critical grain size dcrit below which no further grain refinement would occur
regardless of how severe the deformation is. This assumption is well supported by abundant experimental data
[400–402], cf. Fig. 23. Bouaziz et al. [399] defined this critical quantity as the grain size for which the time for a dislocation
to traverse a grain is equal to the time it takes the dislocation arriving at the opposite grain boundary to be accommodated
there. The accommodation was assumed to be furnished by diffusion-controlled relaxation processes. The corresponding
expression for the critical grain size reads
1=3
DGB b
dcrit ¼ ð66Þ
c_
where DGB is the grain boundary diffusivity. For d dcrit , the diffusional relaxation is predominant and grain boundaries lose
their ability to impede dislocation motion. Obviously, in this case the Hall-Petch relation loses its validity, as well. In this
regime, strain hardening vanishes, too. The critical grain size is governed by strain rate and, through the grain boundary dif-
fusivity, temperature. More specifically, dcrit can be reduced by increasing the strain rate or dropping the temperature.
To ‘smoothen’ the transition from strain hardening to no-strain-hardening behaviour when the average grain size passes
through the d ¼ dcrit ‘watershed’, Bouaziz et al. [399] suggested modifying Eq. (51) in the following way:
pffiffiffiffi
dq ko k1 q
¼ Pþ k2 q ð67Þ
dc bd b
The factor
" 3 #
dcrit
P ¼ exp ð68Þ
d
switches off the effect of dislocation storage at grain boundaries for d dcrit by eliminating the first term on the right-hand
side of Eq. (67). This modification of the KME model, whilst retaining all its features, including an increase of strain hardening
with grain refinement as long as d > dcrit holds, predicts a precipitous decline in the dislocation storage (and hence a loss of
strain hardening) once the grain size drops below dcrit .
3.2.1. Constitutive modelling of materials with a high Peierls stress

The constitutive models outlined above tacitly assume that the Peierls stress, which is always present in a crystalline
material [38], is sufficiently small, which is the case for FCC metals and alloys. However, the inherent flexibility of the con-
stitutive modelling allows for easy adaptation to other systems where the Peierls barrier for dislocation motion can no longer
be neglected, e.g. in body centred cubic (BCC) metals. An example was provided in [403], where the strain hardening of
ferritic steels was modelled. To a first approximation, the Peierls stress rp entered Eq. (40) for the flow stress as the additive
contribution along with the dislocation-related stress:
1=m
pffiffiffiffi e_
r ¼ rP þ Ma0 Gb q _ ð69Þ
e0
The Peierls stress represents an activation barrier for dislocation motion in the crystal lattice, and as such it is tempera-
ture and strain rate dependent. This has to be accounted for in the model, along with the possible influence of the dislocation
density. Considering the thermally-activated kink mechanism for the dislocation motion in a BCC lattice, an appropriate
relation is obtained for the average dislocation velocity entering the Orowan equation for the plastic strain rate as [38]:
0 !
qm b 2
2bhk ak rP F ðrP =MÞ þ W m ~L
e_ ¼ vD exp k
~L þ X
ð70Þ
M kB T M kB T
Here F 0k ðrP =MÞ denotes the Peierls stress dependent energy required to create a kink on a dislocation, W m stands for the
activation energy for kink migration along the dislocation line, and ak and hk are kink geometry parameters (both are of the
order of the Burgers vector b). The parameter X defines the average distance between individual kinks on the dislocation line,
which, in thermal equilibrium, is given by the equation
X ¼ 2ak exp½F 0k ðrP =MÞ=kB T ð71Þ
The quantity ~L stands for the distance between the pinning points on the dislocation line. In the case when the pinning
points are associated with dislocation forest junctions, ~
L scales with the inverse square root of the dislocation density, as
noticed above. Accordingly, for small dislocation density, when ~ L X holds, the last factor in Eq. (70) reduces to unity
and the single kink formation energy enters the exponential in Eq. (70). In the opposite limit case, ~L X, the last factor
in Eq. (70) is given by ~L=X, so that the double kink formation energy turns out to be rate controlling. Using a particular form
of the stress dependence of the kink formation energy [404]
" 0:8 #1:3
s
F 0k ðsÞ ¼ F k 1 ð72Þ
sp
where sp is the Peierls shear stress at absolute zero temperature, this model has been successfully applied to describing the
mechanical behaviour of a TRIP steel with co-existing ferritic and austenitic phases [405].
Although the above model for high Peierls stress materials represents the strain hardening behaviour of conventional
coarse-grained BCC materials quite well, the additive form of Eq. (62) is a simplification. A more elaborated theory, which
does not rely on a simple ‘de-coupling’ of the contributions of the localised obstacles and the Peierls relief to the overall
stress, was presented in [406,407]. Still, the above simplified additive model can be applied to account for differences in
the deformation behaviour of ultrafine-grained FCC and BCC materials. Specifically, striking differences in the strain rate sen-
sitivity of their flow stress [408] can be rationalized in this way. Indeed, for UFG materials ~L may be identified with the grain
size d suggesting pinning of dislocations at grain boundaries.
Assuming d X, Eqs. (70)-(72) yield
!
bh ~L r
2 1:3
qm b 2F k ½1 ðr=rp Þ0:8 þ Wm
e_ ¼ mD k exp ð73Þ
M kB T M kB T
It is noted that through X, a second single kink formation energy enters the exponential. Hence, the total activation energy
is controlled by the double kink energy, as distinct from the case of X minfd; q1=2 g, which applies for conventional, coarse-
grained materials, for which the expression for the strain rate reads
!
2bhk ~L r
2 1:3
qm b F k ½1 ðr=rp Þ0:8 þ Wm
e_ ¼ mD exp ð74Þ
M kB T M kB T
By inverting these equations one obtains the following expression for the UFG case
(
1=1:3 )1=0:8
W m kB T e_
r ¼ rp 1 þ ln ð75Þ
Fk 2F k e_ o
For the coarse grain case, the relation between the flow stress and the logarithm of the strain rate reads
(
1=1:3 )1=0:8
W m kB T e_
r ¼ rp 1 þ ln ð76Þ
Fk Fk 2e_ o
Here the notation

qm bv D bh2k ak r
e_ 0 ¼ ð77Þ
M kB T M
has been introduced.
A comparison of Eqs. (75) and (76) shows that the strain rate sensitivity of the flow stress, S ¼ @ r=@ ln e_ , for a body-
centred cubic UFG metal is just about one half of that for its coarse-grained counterpart (a weak logarithmic dependence
of e_ 0 on stress can, of course, be neglected). The above considerations provide an alternative to the screw dislocation star-
vation argument of Cheng et al. [408], or, for that matter, may work in concert with it.
3.2.2. Constitutive modelling of hexagonal materials

The mechanical response of materials with HCP lattice represent a significant challenge to constitutive modelling. Indeed,
the scarcity of independent slip systems active in HCP metals at low homologous temperatures leads to plastic incompati-
bilities between the adjacent grains, which need to be considered in the model. This can be addressed by introducing the
incompatibility-induced back stress rB and including it on the right-hand side of Eq. (69) as an additive term, as suggested
in [409,410]:
1=m
pffiffiffiffi e_
r ¼ rP þ Ma0 Gb q _ þ rB ð78Þ
e0
The kinetic equation describing the evolution of the back stress is written as
! !
drB ~
¼ K ~ þR
~ Q rB ð79Þ
de e_
where the phenomenological parameters introduced have the following meanings: K, ~ which scales with the elastic modulus,
~ is a temperature dependent parameter
represents the rate of linear increase of the incompatibility stress with strain, Q
which characterises the rate of dynamic relaxation of the incompatibility stress, and R ~ controls the rate of time-driven,
diffusion-controlled static relaxation. The KME model extended in this way, with a slight modification, was shown to predict
the deformation behaviour of an HCP alloy Zircaloy-4, including its response to monotonic loading and strain rate jumps
[409], in an excellent way, cf. Fig. 24.
3.2.3. Inclusion of deformation twinning in the constitutive modelling

In order to account for twinning in a constitutive description of the hardening behaviour of cubic or hexagonal metals, the
dislocation-based model was extended further by Bouaziz et al. [411–413]. They considered the freshly formed twin bound-
aries bounding the deformation twins as effective obstacles to dislocation motion, thus causing a ‘dynamic Hall-Petch effect’
due to grain fragmentation by twin lamellae. The model predictions were found to be in good agreement with tensile test
results for ferritic and austenitic steels. However, Gil Sevillano et al. [414,415] questioned this model because it only took
into account the effect of internal back stresses caused by dislocations in the structure. They proposed an alternative model
that was built on the assumption that strength of the thin twin lamellae represents a significant contribution to the overall
strength of the twinned aggregate. The strength of a twin lamella was considered to be very large compared to that of the
Fig. 24. Mechanical behaviour of recrystallised Zircaloy-4 at 470 °C: stress-strain curve at constant strain rate and strain rate jump testing (a) and creep rate
curve at 40 MPa (b). Experimental data and model results are represented by open circles and solid lines, respectively [409].
600
-4 -1
1.0x10 s
500
-3 -1
1.0x10 s
400 -2 -1
1.0x10 s
Stress, MPa
-3 -1
1.0x10 s
300 -3 -1
1.0x10 s
200 -4 -1
1.0x10 s
100
Experimental strain rate jump test
Simulation strain rate jump test
0
0.00 0.05 0.10 0.15
Strain
Fig. 25. A comparison between the calculated and the experimental stress-strain curves for a-Ti deformed in tension at room temperature [410].
matrix owing to its small (nanometre scale) thickness. The twin lamellae thus experience a forward internal stress while the
matrix is under a back stress - in line with the two-phase ‘composite’ model due to Mughrabi [350]. However, this model still
awaits experimental verification in terms of the internal stress distribution between the twins and the matrix. With regard to
HCP metals, a contribution to plastic strain due to twinning was included [410] in the way suggested by Bouaziz et al. [416]
for low stacking-fault alloys (TWIP steels). Following that work, evolution of the twin volume fraction F was considered as a
sigmoidal function of strain:
h ee i
onset
F ¼ F o þ ðF 1 F o Þ 1 exp ð80Þ
~e
The parameter ~e controls the rate of growth of F from an initial value F 0 at the strain eonset , at which twinning sets in, to a
saturation one, F 1 , reached asymptotically as strain increases. The total strain is represented as a sum of two components:
the strain due to dislocation glide, eg , and that produced by twinning et . In the differential form this reads as
de ¼ ð1 FÞdeg þ et dF ð81Þ
Since twin boundaries can serve as sources of dislocations, the evolution equation for the total dislocation density, Eq.
(51), was modified by including an additive twin-induced contribution on the right-hand side:

dq pffiffiffiffi 1 F
¼ M k1 q þ þ Hðe eonset Þ k2 q ð82Þ
deg bdm 2ebð1 FÞ
Here e is the twin thickness and dm is the mean grain size.
For an experimental verification, the above version of the one-internal-variable model was tuned for a-titanium [410].
Fig. 25 represents an example of good agreement between the experimental data and calculated deformation curves (includ-
ing strain rate jump episodes) is shown in It should be mentioned that a provision for accounting for small grain-size effects
is included in the model through the second term on the right-hand side of Eq. (80). Application of the model to UFG tita-
nium and other fine-grained HCP metals are thus possible.
Inclusion of twinning as an important ingredient of constitutive modelling may appear to be less relevant in the context
of UFG materials, as the role of twinning is reduced upon grain refinement [417,418]. However, experimental evidence sug-
gests that twinning does occur in UFG metals with sufficiently low stacking-fault energy [419–423] and, even more surpris-
ingly, in high stacking-fault energy materials such as aluminium [424] and nickel [425], which do not exhibit twinning under
normal testing conditions. Thus, twinning should be considered as a possible deformation mode in an extended constitutive
model for UFG metals. As suggested above, this can be done through an appropriate evolution equation for the dislocation
density, such as Eq. (82), and an additional equation describing the evolution of the twin volume fraction. The common
approach suggested in earlier models and expressed by Eq. (80) is to consider the variation of F with strain. In our current
thinking, however, this evolution is better represented in terms of the variation of F with stress, rather than strain. The
stress-driven variation of the twin volume fraction was a basis for our recent model [426]. The main assumptions of the
model can be summarised as follows. (i) As often observed in experiment [420], a twin nucleated in a fine grain shoots
through the entire grain. Accordingly, its length is dictated by the grain size d. (ii) The average thickness of the nucleated
twin, e, is considered to be grain size independent. (iii) The generated twin is assumed to be in equilibrium for which case
Friedel’s formula for the shear stress for twinning [40] can be applied:
G e
sffi s ð83Þ
2 d
Fig. 26. Twinning model verification for Mg alloy ZK60 for two values of the average grain size, dm = 73 lm and 17 lm (after [426], reproduced with
permission).
Here s is the amount of shear strain produced by the twin. (iv) Twinning is not considered to be a significant contributor
to plastic strain. It is activated gradually as the stress rises owing to strain hardening due to dislocation glide controlled plas-
ticity. Smaller and smaller grains get engaged in twinning with growing stress according to Eq. (83). This gives rise to a con-
tinual increase of the twin volume fraction. (v) The grain size distribution in the material is described by a function f ðdÞ,
which in [426] was taken to be log-normal. This form of the distribution function, while not being binding, is supported
by experimental evidence abundant in literature, see, e.g. [427].
Under these assumptions, the growth rate of the twin volume fraction is governed by the rate with which the grain pop-
ulation is ‘scanned’ for those grains which are ripe for twinning at a given stress level. In terms of the derivative with respect
to strain, this relation reads
2
dF ed dr
~ 3 gðrÞ
¼v ð84Þ
de dm de
~ is a parameter that contains as a factor the fraction of grains capable of undergoing twinning, which is texture-
Here v
2
dependent. Further quantities in Eq. (84) are as follows: gðrÞ ¼ f ðdðrÞÞ M2rGse
2 and d ¼ dðrÞ is a function of the current flow
stress r ¼ Ms with s given by Eq. (83).
The model outlined was shown to work very well for the case of a conventional, coarse grained material. It was validated
by tensile tests, scanning electron microscopy, and acoustic emission measurements on Mg alloy ZK60 [426]. Fig. 26 demon-
strates a very good agreement between modelling results and experimental data on the evolution of the twin volume frac-
tion assessed by the advanced acoustic emission technique employed. Model verification for UFG materials is yet to be
provided.
3.3. Two-internal-variable models
As demonstrated in the foregoing sections, it can be considered as established that one-internal-variable models based on
the Kocks-Mecking approach provide a reliable description of mechanical response of metals and alloys, especially for low to
moderate strains. However, they do not account for the occurrence of late stages of strain hardening beyond Stage III
[58,428]. Stages IV and V extend well into a range of very large strains pertinent to SPD processing. This calls for more sophis-
ticated models that would adequately describe a material’s behaviour at very large strains. Two-internal-variable models
have a better chance to satisfy this expectation, as they possess greater flexibility in describing microstructure evolution dur-
ing SPD processing and the mechanical properties of the processed materials [429]. As discussed in Section 3.1, a dislocation
cell structure emerged during plastic deformation can be seen as a pre-cursor of the final grain structure to be formed at
large strains, particularly under SPD. Splitting the dislocation population into cell interior and cell wall dislocations
[254,255,350,430–433] already mentioned above appears to make more sense in the case of cell-forming materials than sub-
dividing it into mobile and immobile dislocations, as was done in Refs. [185,429,434]. Our first-hand experience with con-
stitutive models involving two dislocation densities [254,255] gives us confidence that they are well suited for modelling in
the SPD realm. These models are presented below, along with other models based on similar concepts.
The underlying assumption is that a dislocation cell structure already exists. The overall share stress is then given by a
linear combination of two components (cell walls and cell interiors), each of them being related to the respective dislocation
density (qw in the walls and qc in the interior). The key characteristic of the dislocation structure, viz. the average dislocation
cell size, dc , is assumed to be inversely proportional to the square root of the total dislocation density q:
pffiffiffiffi
dc ¼ K o = q ð85Þ
where K o is a constant. This is a safe assumption, which follows from dimensional considerations repeatedly proven by TEM
observations [351]. Similarly to Eq. (43), q, is given by a weighted sum of the dislocation densities in the two ‘phases’ of the
material:
q ¼ f w qw þ ð1 f w Þqc ð86Þ
There is an obvious relation between the volume fraction of the cell walls, f w , and the geometrical parameters of the cell
structure, i.e. the cell size and the wall thickness, w. Regardless of the cell morphology, f w can quite generally be expressed as
f w ffi 3w=dc ð87Þ
where dc represents the characteristic size of the cells. For instance, for spherical grains it can be interpreted as the
average sphere diameter, while for cuboidal grains it can be interpreted as the average cube size. As seen from
Eqs. (85) and (86), the cell size will typically decrease during the deformation process, owing to an increase in the dis-
location density. His would mean that f w is an increasing function of strain. Experiment shows an entirely different
behaviour, however. In the case of copper deformed to large strain in torsion, f w was found to continuously drop from
an initial value f wo to a lower saturation value f w 1 . The variation of f w with the plastic shear strain c is well represented
by a heuristic equation [254,255]:
~Þ
f w ¼ f w1 þ ðf wo f w1 Þ expðc=c ð88Þ
The rate of decrease of the cell volume fraction described by this equation is determined by a model parameter c ~. Con-
sidering Eq. (87), the observed decrease of f w represented by Eq. (88) can only mean that the wall thickness must decrease
with strain fast enough to compensate for the decrease in the cell size. This ‘slimming’ of cell walls may be associated with
progressive annihilation of dislocations, which a random or ‘redundant’ in the cell walls in the sense that they do not con-
tribute to misorientation between the cells separated by the walls. We note that the above heuristic relation can be replaced
pffiffiffiffiffiffi pffiffiffiffiffiffi
by another one, based on the idea that w should scale with 1= qw . With the ansatz w ffi 1= qw one obtains by combining
pffiffiffiffiffiffiffiffiffiffiffiffi
Eqs. (85) and (86): f w ffi ð3=K o Þ q=qw . The ratio of the dislocation densities in this formula is not a constant, but rather con-
tains f w . Accordingly, it can be regarded as an implicit equation that needs to be solved in order to calculate the variation of
the volume fraction of cell walls with strain. Within a reasonable range of values of the model parameters, a decline of f w
with strain is predicted.
A practical assumption in applying the two-internal-variable model under consideration is to assume that Taylor-type
(iso-strain) condition holds for the cell interiors and the cell walls, i.e. that the plastic strain in these two ‘phases’ is the same.
A set of coupled differential equations that describe the evolution of the dislocation densities in the two ‘phases’ reads
pffiffiffi pffiffiffiffiffiffi 1=n
dqw 6b ð1 f w Þ c_
2=3
3b ð1 f w Þ qw
¼ þ k2 qw ð89Þ
dc bdfw bfw c_ 0
pffiffiffiffiffiffi 1=n
dqc 1 qw 6 c_
¼ a pffiffiffi b k2 qc ð90Þ
dc 3 b bdð1 f w Þ
1=3 c_ 0
In earlier publications [254,255], a physical interpretation of the various terms in Eqs. (89) and (90) was offered. For
instance, the loss of cell interior dislocations to the walls is accounted for by the first term on the right-hand side of
Eq. (89). The concomitant gain in the wall dislocation density is captured by the second term in Eq. (90). Dynamic
recovery by dislocation cross-slip in the low-temperature regime typical for SPD processing finds its representation in
the last term in both equations. Both exponents entering the model, m⁄ and n , can be taken to be inversely proportional

to the absolute temperature T. Further model parameters, a⁄, b⁄, and k2 , are considered constant. If dynamic recovery in
cell walls is controlled by dislocation climb, as assumed by Zehetbauer and co-authors [428,431,435], c_ o in Eq. (89) is no
longer a constant. Rather, it is given by an Arrhenius equation. The activation energy in this equation is that for self-
diffusion; the exponent n⁄ is a constant. In modelling of SPD, whose signature feature is a high hydrostatic pressure,
the model needs to be modified to include the pressure dependence of dynamic recovery by climb [436]. This was done
in [222] where an exponential pressure dependence of the dynamic recovery term was included in the last term in
Eq. (89). With this modification, the model was used to elucidate the effect of back pressure on the ECAP processing
of an Al alloy [222].
A further modification of the set of the evolution Eqs. (89) and (90) for the dislocation densities in cell forming metals was
proposed by Hosseini and Kazeminezhad [437]. Keeping the basic assumptions the same as in the general two-internal vari-
able model considered above, their model makes a distinction between edge and screw dislocations and divides the whole
population of dislocations into three categories: mobile and immobile dislocations in the cell interior and immobile dislo-
cations in cell walls. The authors assumed that the recovery in the cell interior occurs by cross-slip of screw dislocations
while the recovery kinetics in cell walls is governed by climb of edge dislocations. The set of the evolution equations for
the cell interior and cell wall dislocation densities according to Ref. [438] reads as follows:
pffiffiffiffiffiffi pffiffiffiffiffi " s0c

!#
dqc a qw _ 6b v mD qc Q cross-slip
¼ pffiffiffi cw c_
1=3 c 0 exp G
dt 3 b bdð1 f w Þ l kB T
p ffiffiffi ð91Þ
2=3 pffiffiffiffiffiffi 3
dqw 6b ð1 f w Þ 3b ð1 f w Þ qw d Gb DL q2w
¼ c_ c þ c_ c
dt bd f w bfw kB T
0
where DL is the lattice diffusivity, Q cross-slip is the activation energy for the cross-slip of screw dislocations, l is the size of a
potential site for cross-slip and d and v are numerical constants. Although this attempt to provide more microscopic details
of the recovery process is appealing, the way in which recovery was treated in that model is questionable. Indeed, the recov-
ery effects enter both Eq. (91) as static recovery terms, since a factor representing the strain rate is missing in these terms.
This is radically different from the original Eqs. (89) and (90) where the recovery was introduced as a dynamic, strain-rate
driven process, which is consistent with the underlying philosophy of the KME approach. Despite this deficiency with han-
dling the recovery processes, in their subsequent work [438] the same authors demonstrated that their approach is capable
of accounting for strain softening at large strains - a phenomenon which will be discussed in more detail in Section 5.2.
The applicability of the model outlined above hinges on the validity of Eq. (85) that establishes a relation between the
dislocation cell size and the total dislocation density. As mentioned above, this scaling relation follows from dimensional
considerations and is universally accepted. It is also supported by stochastic modelling of dislocation populations [63]. How-
ever, the validity of Eq. (85) for non-steady state deformation still awaits a rigorous proof, cf. [4].
The dislocation-based constitutive modelling based on Eqs. (89) and (90) or their modifications were used to account for
experimental data over a wide range of strain, including giant strain values attained during SPD. While having many com-
monalities, the models proposed in literature are different in detail of the microscopic mechanisms of dislocation processes
considered, cf. early work by Argon and Haasen [439] and more recent publications by Nes and Marthinsen [354,355].
For example, to model dislocation density evolution in deformed pure Al, Chinh et al. [185] have recently employed a
two-internal-variable approach proposed by Kubin and Estrin [284,429,434], which was originally developed to account
for plastic instabilities associated with the Portevin–Le Chatelier effect in solid solutions. In this approach, which was not
intended specifically for description of large deformations, the cellularity of the dislocation structure was not considered.
The two dislocation densities comprising the total one, q ¼ qm þ qf , are the densities of mobile (qm ) and forest (qf ) disloca-
tions. They obey the following differential equations [434]
dqm
¼ C 1 C 2 qm C 3 qf1=2
dc
ð92Þ
dqf 1=2
¼ C 2 qm þ C 3 qf C 4 qf
dc
The individual terms in Eq. (83) have a transparent physical meaning, as they can be interpreted in terms of various dis-
location reactions analogous to those underlying Eqs. (89)-(91). What distinguishes this set of equations from that given by
Eqs. (89) and (90), is that the former is ‘structureless’. While Eqs. (92) and (69) do provide an adequate description of the
deformation behaviour of metals at small and moderate strains (which can be extended to large strains, as done in Figs. 27
and 28, they do not reflect any development of dislocation cell or grain structure, which is obviously their deficiency.
The two-internal-variable formulation can also be used for modelling high-speed severe deformation processes where
heat release plays a substantial role. The model has a provision for considering the thermomechanical coupling through
including a temperature dependence in the exponents m⁄ and n⁄ (or the reference strain rates e_ 0 and c_ 0 ), as was done
recently in modelling impact deformation [440].
Fig. 27. Experimental data for Al and the calculated stress-strain curve obtained by solving Eqs. (92) and (69) (from [185], reprinted with permission).
Fig. 28. The r vs e curve for polycrystalline Al over a wide range of strain calculated by using the differentiated Voce equation, Eq. (10) (from [185],
reprinted with permission).
The two-internal-variable formulation can also be used for modelling high-speed severe deformation processes where
heat release plays a substantial role. The model has a provision for considering the thermomechanical coupling through
including a temperature dependence in the exponents m⁄ and n⁄ (or the reference strain rates e_ 0 and c_ 0 ), as was done
recently in modelling impact deformation [440].
3.4. Constitutive modelling of multi-phase materials
The above considerations referred to single-phase materials. Constitutive modelling for multi-phase materials can b
based on the same principles. One can distinguish between two principal cases, as can be illustrated for a two-phase
material. In the first case, a second phase is represented by small dispersed particles and does not have any substantial
load-bearing capability, so that stress sharing between the phases can be disregarded. The particles then play a ‘passive’
role, constituting obstacles to mobile dislocations of the majority phase (‘matrix’). If, by way of example, the particles are
non-shearable and are to be overcome by moving dislocations via the dislocation bowing-out (Orowan) mechanism, they
qffiffiffiffiffi
contribute to the flow stress an additive term of the order of the Orowan stress, rOrowan ffi MGb=Lp ffi MGb f p =R [441,442].
Here Lp is the average particle spacing, R their radius, and f p the volume fraction of the second phase. Similarly, shearable
particles add a contribution of the order of MpRcs =bLp (with cs denoting the surface energy) to the flow stress. There is
another, more subtle, effect of second-phase particles that enters through the evolution of the dislocation density in the
dislocation cells, e.g. Eq. (90), as a positive additive storage term inverse in Lp and also as a multiplicative factor in the
dynamic recovery term (the last term in that equation) [443]. The latter factor accounts for a reduction of the dynamic
recovery rate due to the need for a dislocation in contact with a particle to be detached from it by a mechanism proposed
by Rösler and Arzt [444].
In the opposite case where a second phase does have a sizeable load-bearing capacity, the overall stress of a phase mix-
ture encompassing the contributions from all phases has to be calculated. The simplest approach is to use a rule of mixtures,
whereby the stress of the multi-phase material is given by a weighted sum of the constituent phases. The weight factors are
taken as the volume fractions of the respective phases, while the stress in a particular phase is calculated using a model for a
single-phase material presented above in Sections 3.1–3.3. A more sophisticated treatment of the multi-phase nature of a
material would involve an appropriate homogenisation technique, and the one originally proposed by Molinari et al.
[445] is believed to be both versatile and reliable [446,68].
As a convincing illustration of this approach, Ardeljan et al. [447] have developed a multiscale model for anisotropic
deformation of multi-phase polycrystalline aggregates to large plastic strains. The proposed elasto-plastic strain rate- and
temperature-sensitive model is footed on a dislocation-based hardening law in the spirit of Eqs. (89) and (90) combined with
crystal plasticity-based finite element modelling, which bridges the single-crystal and polycrystal mechanical response. The
model was successfully applied to study the texture evolution and the deformation mechanisms in a HCP-Zr/BCC-Nb layered
composite under severe plastic deformation by accumulative roll bonding. Not only did the model predict the texture in both
co-deforming materials to very large plastic strains, but it also provided an accurate simulation of the tensile behaviour of
both constituent phases (HCP-Zr and BCC-Nb). Fig. 29 represents a distribution of local stresses and strains obtained with the
full account of the evolving dual-phase microstructure, including the variation of dislocation density and crystallographic
grain reorientation in both constituents during deformation.
Thus, it can be concluded that the modelling approaches developed for single-phase materials can be extended to multi-
phase ones.
Fig. 29. Initial FE mesh and deformed FE models of the Zr/Nb composite showing Von Mises stress and equivalent strain contours at strain levels indicated
in the figures (adapted from [447]).
Fig. 30. Schematic showing the microstructural evolution during ECAP processing: (a) diffuse low angle boundaries dominating after a few (say, two)
passes, (b) transition state from low to high angle boundaries (e.g. after 8 passes), (c) stabilized microstructure (say, 16 passes). Note that the average grain
size (including both high and low angle boundaries) stabilises after only a few passes (from [453], reprinted with permission).
4. Possible scenarios for grain refinement
The scenario of grain refinement tacitly assumed in the foregoing section infers that the underlying process is the dislo-
cation cell formation followed by cell size evolution according to Eq. (85). That is to say, the dislocation cell structure is
assumed to be a pre-cursor of the new grain structure that develops with accumulation of misorientations across the cell
boundaries. Evolution of misorientation between the dislocation cells in the process of straining was modelled by Estrin
and Kim [448]. They considered the growth of the misorientation for low-angle boundaries with the influx of edge disloca-
tions, assuming a local disbalance in the density of edge dislocations of opposite sign. A similar treatment, yet through a
more involved dislocation density model, was proposed by Alexandrov et al. [449], who introduced the excess dislocation
density in the boundaries. This simplistic consideration of the evolution of cell/grain misorientations was further improved
by Pantleon [450] and Estrin et al. [451] on the basis of a probabilistic approach. The results of the latter studies account
reasonably well for the misorientation angles associated with the so called incidental dislocation boundaries [452], but
the models used do not apply to high-angle (geometrically necessary) grain boundaries [452] whose fraction grows as a
result of SPD processing. The Langevin approach leading to a Fokker-Planck equation for the distribution of misorientation
angles [451,452] appears to provide a promising frame for modelling the evolution of grain misorientations.
The above scenario is illustrated by Fig. 30 [453] that shows the transformation of diffuse cell walls to small-angle grain
boundaries, which, upon increase in misorientations, are converted to large-angle ones.
In the grain refinement scenario in which the emerging grain structure inherits the length scale of the parent dislocation
cell structure, there is an obvious limit to the attainable degree of grain refinement. Indeed, by combining the expression for
the dislocation cell size, Eq. (85), with the Taylor relation, Eq. (40), one arrives at a simple formula for the smallest attainable
min
dislocation cell size dc (i.e. the minimum grain size):
min G
dc ffi K o Ma ð93Þ
rtheor
This relation shows that for typical values of K o of the order of 10, M ffi 3, and a ffi 0:5; the achievable grain size cannot be
smaller than about 100b even for the stress level close to the theoretical strength rtheor . That is to say, the grain refinement
mechanism, in which grain subdivision occurs via the cell structure formation and its refinement during straining, cannot
bring the average grain size below the 100 nm mark that defines a nano material. This is consistent with a huge body of
experimental evidence on grain refinement by severe plastic deformation [4]. While the model predictions with regard to
the mechanical behaviour and the emerging grain size based on the two-internal-variable model [254,255] have been ver-
ified for various materials and SPD processes, there is still a pressing need to improve modelling with regard to grain frag-
mentation. A recently proposed approach [107] offers a modelling frame suitable for a simple description of grain
fragmentation in terms of lattice curvature evolution. It was suggested [107] that lattice curvature develops within a grain
due to kinematic constraints imposed by the neighbouring grains. More specifically, the rotation of the crystallographic
planes in the grain due to dislocation slip was considered to be retarded near the grain boundaries. The grain subdivision
was associated with the emergence of geometrically necessary dislocations required to accommodate the resulting lattice
curvature. Agglomeration of these geometrically necessary dislocations to new grain boundaries was proposed as the mech-
anism of fragmentation of a grain. It was suggested that the process of grain subdivision into new grains (owing to the occur-
rence of a zone near the grain boundaries where crystallographic plane rotations are inhibited by the grain boundaries and a
middle part of the grain where they are not) leads to the emergence of a grain population with the grain size of d/3. This
process of grain subdivision can go on and on leading to multiple generations of grains of ever decreasing size until the dis-
location glide mechanism causing the rotation of crystallographic planes ceases to operate. As mentioned above, this is the
case when the grain boundaries no longer act as impenetrable obstacles to dislocations, as dislocations arriving at grain
boundaries get accommodated there with the aid of grain boundary diffusion [399]. This occurs when the grain size drops
below a critical level, dcrit , given by Eq. (66), whose estimated magnitude for room temperature deformation amounts to sev-
eral hundred nanometres [4]. Toth et al. [454] also looked at the evolution of the misorientation distribution function during
severe plastic deformation. The predicted shift of the distribution towards large misorientation angles, which was confirmed
by experiments on ECAP processing of Cu, was attributed to the shape variation of the initial grains, effectively raising the
volume fraction of the ‘old’ high angle boundaries. This concept, which is at variance with the commonly shared view that
new grain boundaries emerging during SPD acquire large misorientations in the process, calls for further investigations.
An approach to grain fragmentation associated with the rotation of slip planes was also taken by Kratochvíl et al. [455]
who proposed a crystal plasticity based model to describe the material response to severe plastic deformation under HPT.
Specifically, they assumed uniform deformation by double slip in plane-strain and considered the rotations of the slip sys-
tems caused by the imposed shear strain. They found local variations in the crystal lattice orientation and claimed these vari-
ations were responsible for microstructure fragmentation. Aoyagi and Shizawa [99] introduced geometrically necessary
dislocations and geometrically necessary incompatibility tensorial terms corresponding to isolated dislocations and disloca-
tion pairs into a strain gradient crystal plasticity model reproducing a grain subdivision process at large strains under pure
share cold rolling deformation. In their model, new grain boundaries with large angles of misorientation were nucleated
along the GNBs due to enhanced strain gradients associated with dislocation activity at the boundaries, which is in fair agree-
ment with experimental observations. In contrast with the ‘mainstream’ notion of the grain subdivision process with gradual
accumulation of misorientations, Seefeldt et al. [456] modelled the evolution of misorientations across the cell boundaries as
a ‘nucleation and growth’ process triggered by disclination nuclei randomly induced in the cell walls. In their model, these
nuclei represented by dipoles of partial disclinations interact with statistically stored dislocations and spread along the walls
by capturing incoming lattice dislocations, thus increasing the misorientation between the neighbouring cell blocks.
The above considerations of a grain refinement scenario based on gradual subdivision of the grains suggest that there are
two limit grain sizes, viz. those given by Eqs. (66) and (85) (or (93)). Grain refinement will cease to occur once the greater of
the two critical values of the grain size is reached. Both are in the range of several hundred nanometers, i.e. slightly above the
true nano material range. The above scenario of a ‘peaceful’ transformation of dislocation cell structure to a grain structure
was contrasted by a picture of a more ‘violent’ grain refinement down to nano scale by the so called dynamic plastic
deformation (DPD) [457], which involves high strain rates and/or low deformation temperature (large values of the
Zener–Hollomon parameter Z) [4]. The corresponding mechanism of grain refinement involves formation of nanotwin bun-
dles [458], which transform to nanograins by fragmentation of twin/matrix lamellae due to interaction of twin boundaries
with dislocations or shear banding. No model of these processes has been offered so far, and this is certainly an area to be
considered in the future.
Saturation of grain fragmentation implied by the models discussed cannot be taken for granted, and there has been a
long-standing discussion in literature on whether there is such a thing as ‘saturation’. The question is related to that of
whether saturation of the flow stress (i.e. asymptotic vanishing of strain hardening) occurs at large strains. The latter can
be answered affirmatively. Indeed, the concepts based on dislocation density evolution do account for the occurrence of late
strain hardening stages (including Stage IV and beyond) that eventually terminate in saturation of the flow stress
[167,255,431,459,460], as do numerous experimental studies, see, e.g. [459–462]. It has been claimed, however, that despite
saturation in the intrinsic strain hardening capability of a material, some microstructural changes may still occur at very
large strains [463], so the question of whether microstructural variation would eventually saturate still remains. More
specifically, in the context of SPD processing, it is of interest to understand whether saturation of grain fragmentation does
or does not occur. Pippan et al. [462,464] strongly advocate the notion of saturation and provide convincing evidence for
that. Moreover, it was shown in their work that the saturated grain structure was not very sensitive to strain path chosen
(HPT vs. HPT + rolling), which speaks for certain ‘universality’ of microstructure development under very large strains.
However, substantial differences in the saturated microstructure between monotonic and cyclic loading were established.
A prominent role of grain boundary migration in the scale of the saturated grain size was also highlighted. The relative
‘universality’ of the grain structure development appears to be at odds with the analysis of Tomé et al. [465] who argued
that strain hardening behaviour should be strain path dependent. A similar view is held by Beygelzimer et al. [167], whose
point is that the intrinsic strain hardening derived from deformation tests for different strain paths should be different due to
differences in texture and in the complexity of dislocation slip. At this stage, the issue is not fully resolved and more work is
needed to establish whether saturation in grain fragmentation under SPD does take place and if so, whether it is strain path
dependent or not.
5. Numerical simulations of SPD processes
5.1. FEM simulations of SPD processing
Literature on FEM simulations of SPD processing is quite substantial. Important contributions to this area by the research
group of Prof. Hyoung Seop Kim of Postech, Korea – especially in relation to modelling of high-pressure torsion and equal
channel angular pressing - should be noted. A representative sample of this work can be found in [466]. Further publications
in this field have been cropping up recently, cf. [467]. Most of them are based on phenomenological constitutive models, as
represented by a recent publication on HPT modelling by FEM [92]. The authors used an elastic-perfectly plastic constitutive
model to study the effects of contact friction conditions on the evolution of the distributions of the components of the stress
tensor and the plastic strain in copper and highlighted strong heterogeneity of the plastic strain both along the radius and
thickness of the sample within the first revolution of the anvils.
In Section 2.2 we considered simulations of the ECAP process in some detail. FEM simulations were also applied to a range
of other SPD techniques. Thus, ECAP with a rotary-die was considered by Yoon et al. [468]. Cyclic extrusion-compression
(CEC) was analysed by Rosochowski et al. [469]. High pressure torsion (HPT) of disks was considered by Kim [470], Figueir-
edo et al. [261,262,471,472] and Molotnikov et al. [384,473]. As new SPD technologies and the areas of their possible appli-
cation keep emerging, the significance of FEM methods for quick and efficient design of processing routes and tools increases.
This has been successfully demonstrated in a recent study of a novel industrially scalable extrusion process for curved pro-
files [474] or a simple shear extrusion process [475–477]. In particular, FEM is indispensable in the modelling and simulation
of deformation processing under extreme conditions such as low temperatures during cryo-SPD processing [478], or very
high strain rate deformation in the high-speed machining [479–481].
Constrained groove pressing and constrained groove rolling were simulated by Lee and Park [482]. Following the FEM
simulation-based strategy, new optimized processing schemes different from known SPD processes were proposed. For
example, Luri et al. [145] offered a modified ECAP die configuration while Rosochowski et al. [483] developed an ECAP-
like incremental process called incremental ECAP (I-ECAP) suitable for continuous processing. A large body of computational
(chiefly FEM-based) work on the SPD process known as twist extrusion and cognate processes has been discussed and
evaluated in a recent review [484]. A successful exercise in analytical modelling of an axisymmetric forward spiral extrusion
process [485] – yet another recent variety of SPD processing - employed an upper bound approach, developed earlier in the
context of twist extrusion modelling [486].
Substantial efforts were also put in design and simulation of SPD processing of thin products [487,488], notably thin-
walled tubes [315,489,490]. A process dubbed tube channel pressing [491] was studied by Farshidi and Kazeminezhad
[492] by employing FEM simulations. An assessment of a prospective Tube Twist Pressing process as a new SPD method
for tubular materials by FEM simulations was presented by Babaei et al. [493]. All these techniques rely on a combination
of large shear deformation and high hydrostatic pressure and in that sense they are akin to the ‘mainstream’ SPD processes,
such as ECAP or HPT. A related technique for manufacturing thin-walled cone-shaped products [494] which utilises this kind
of imposed conditions, was assessed on the basis of FEM simulations.
These examples of successes with simulation of ‘classical’ and emerging SPD processes demonstrate the power of solid
mechanics in the analytical and computationally-assisted engineering of SPD and the potential of such approaches for future
industry scale applications.
As repeatedly emphasised above, microstructure-based constitutive models have substantial advantages over the phe-
nomenological, mechanistic ones, and it is encouraging that this view is becoming more and more accepted by the research
community. As an example of the application of the two-internal-variable model [254,255] we show the results for ECAP of
Al [256], Fig. 31. The calculated curves representing the evolution of the dislocation cell size and the equivalent stress as a
function of the cumulative equivalent strain for the ECAP die (which for the channel angle of 90° practically coincides with
the number of ECAP passes) are in good agreement with the measured values. As discussed above, the emerging grain size
was identified with the dislocation cell size.
In several numerical simulation exercises addressing SPD processing, see, e.g. [495], it was established that the basic two-
internal-variable model [254,255] provides a potent and robust tool for predicting the material behaviour during processing
as well as the properties of the processed material. For instance, in the above example of ECAP of Al, such aspects as nonuni-
formity of the microstructure produced was studied by finite element simulations [256].
Fig. 31. Refined grain structure of pure Al after four ECAP passes (a); the evolution of the dislocation cell size (b) and the variation of equivalent stress with
equivalent strain for commercial purity Al (c) [256].
Fig. 32. Evolution of the dislocation cell size across an Al billet with the number of ECAP passes. The quantity ŝ denotes the distance from the bottom of the
billet normalised with respect to the billet thickness [256].
Fig. 33. TEM image showing the microstructure of Cu after 8 ECAP passes (a) and evolution of the yield strength of copper as a result of ECAP processing to
different numbers of passes via routes Bc (b) and A (c) (after [496] (a) and [96] (b), reproduced with permission).
Fig. 34. Dislocation cell size variation with the cumulative equivalent strain during high-pressure torsion of copper (a) [384]. The schematic of the process
is shown in (b).
The results shown in Fig. 32 demonstrate how the cell size distribution (which, as mentioned above, is regarded to be
tantamount to that of the grain size distribution) across the billet thickness progressively becomes more and more uniform
with the number of ECAP passes. The average cell size (which is to become the average grain size when the cumulative strain
is large enough for sufficient cell misorientations to develop) tends to saturate with strain at a level of about 800 nm. For
other materials and with different SPD techniques, significantly smaller values of grain size are achievable [4]. Thus, for cop-
per of commercial purity, the average cell/grain size was shown to drop to about 250 nm after eight ECAP passes by Route Bc
as illustrated in Fig. 33a [496]. The concomitantly increasing gain in the yield strength with the number of passes seen in
Fig. 33b is quite impressive. Fig. 33c illustrates a similar behaviour of Cu deformed by Route A ECAP as predicted by Tóth
[96] who used the two-internal-variable cellular model based on Eqs. (89) and (90), which was combined with the flow line
model [156] discussed above and the self-consistent viscoplastic model in the spirit of the Molinari and Tóth approach [497].
Compared to the results shown in Fig. 33b, the modelled flow curve is considerably more detailed. It also accounts for stress
drops due to the inter-pass breaks.
An example of even more extreme grain refinement is shown in Fig. 34, which displays the evolution of the calculated
dislocation cell size in copper under high-pressure torsion [384] – a process discussed above and schematically illustrated
in the same figure. For the vertical pressure of 8 GPa the cell/grain size attained a saturation value of about 120 nm, in close
agreement with experimental data included in the diagram. The simulations [384] were based on the two-internal-variable
model and additionally included a gradient plasticity term. The provision for gradient plasticity [473], which will be dis-
cussed in Section 7, helps explaining the development of nearly uniform, refined microstructure in an inherently non-
uniform process such as HPT.
The realisation that the details of the strain hardening behaviour in terms of microstructure evolution are important for
realistic FEM modelling has motivated a series of studies by Hosseini et al. [272,437,438,498–502] in which FEM modelling
was based on dislocation density evolution. They exploited a flow line model originally proposed by Tóth et al. [156] and
extended by Hasani et al. [158] which was backed by a two-internal-variable model involving two dislocation densities
of the kind described in Section 3.3. It was used to investigate the deformation behaviour of Cu and Al in the route Bc ECAP
process with sharp and curved die corners, Fig. 3. The strain and strain rate were obtained from the flow line model to inform
a modified version of the Estrin-Tóth -Mollinari-Bréchet model [255] for the evolution of a cellular microstructure. The pre-
dicted characteristics of the microstructure were found to be in fair agreement with the commonly reported experimental
data: (i) the dislocation density for the material pressed in the sharp-corner die was higher than that in the die with a corner
curvature; (ii) the dislocation density in processed Al was lower than that in processed Cu; (iii) the cell size in the processed
Cu was predicted to be smaller than that in the processed Al; and (iv) the sharp-corner die produced a finer cell size in both
materials than the curved corner die.
Following the success of this modelling approach, that group of authors incorporated constitutive modelling of disloca-
tion hardening into the FEM code which was applied to simulating the groove pressing/repetitive corrugation (GPRC) and
repetitive corrugation straightening (RCS) processes [500]. The set of coupled differential equations for the evolution of
the cell interiors and the cell walls dislocation densities similar to Eqs. (89) and (90), combined with Eqs. (40) and (86) pro-
vide a full constitutive description of dislocation behaviour. To apply this constitutive model to SPD processing, the cumu-
lative strain and the resolved shear strain rate for a particular process, e.g. ECAP, is required. Hosseini and Kazeminezhad
[499] calculated these quantities from the flow function results.
A powerful tool to model the deformation behaviour of polycrystalline materials is polycrystal plasticity [68,94,503]. A
way to incorporate microstructure evolution in the polycrystal plasticity framework is by considering the actual active slip
systems with their individual slip rates and the resolved shear stresses that also vary from one slip system to another. For
simplicity, it can be assumed that the dislocation cells within a grain are identical and their mechanical response can be
characterised by a unique common resolved shear strain rate c_ r . In this didactically simplified way, misorientations between
cells within a grain are disregarded, while the orientation of each individual grain is considered. Assuming the shear resis-
tance is the same for all slip systems, c_ r is derived as follows [504]:
c_ r ¼ Me_ eq
1 X
N
ð76Þ
M¼ c_ rn
e_ eq n¼1
where N is the number of slip systems and c_ rn denotes the strain rate for the nth slip system. The equivalent (von Mises)
qffiffiffiffiffiffiffiffiffiffiffiffi
strain rate can be calculated via e_ eq ¼ 23 e_ ij e_ ij .
Embedding the two-internal-variable model based on dislocation density evolution [429] in the crystal plasticity frame-
work was considered by Tóth [96] and Shanthraj and Zikry [505]. Hosseini and Kazeminezhad [499] used a similar approach,
yet with a specific consideration for the dislocation cell structure. The average dislocation densities within the cell interiors
and cell walls as well as the total dislocation density calculated by Hosseini and Kazeminezhad [499] are presented in Fig. 35.
In agreement with general expectations and experimental observations the model predicts a rapid increase of dislocation
density in both cell interiors and cell walls from 1013 m2 and 1014 m2 to 1.54 1015 m2 and 5.11 1015 m2, respectively,
after 6 ECAP passes. Over the following passes, the calculated dislocation density in cell walls increases continuously at a low
rate and reaches a level of 6.29 1015 m2 at the end of pressing. This kind of dislocation density evolution was also pre-
dicted for Al by McKenzie et al. [222] and reported experimentally by Estrin et al. [473]. In the hybrid model [499] the
dislocation density in the cell interior reached a maximum and then dropped slightly with the increasing number of ECAP
passes (see Fig. 36). This characteristic reduction in the dislocation density in the cell interior was attributed to the
transformation of cells with low angles of misorientation to fine grains with high angle grain boundaries. This trend was
confirmed by the experiments carried out for copper severely deformed by high pressure torsion by Estrin et al. [473].
Fig. 35. Calculated evolution of the total dislocation density (a) and corresponding flow stress (b) in Cu and Al subjected to different numbers of passes
through a sharp-corner and a curved ECAP die (from [498], reprinted with permission).
Fig. 36. Evolution of (a) the fraction of cell walls and (b) strength as a function of the number of passes in the model by Hosseini and Kazeminezhad (from
Hosseini [499], reprinted with permission).
Fig. 37. Contour maps of the total stored dislocation density (in units of m/m3) at different levels of sheet thickness reduction: (a) 0–10%, (b) 10%, (c) 10–
20%, (d) 20%, (e) 20–30%, and (f) 30% (from [506], reprinted with permission).
Modelling of the variation of dislocation density has shown unequivocally that, depending on the material, intensive mul-
tiplication and storage of dislocations during SPD occurs up to cumulative imposed strain of 3 to 4. This finding is in excellent
agreement with the most known and well documented feature of the SPD processed metals, viz. rapid strain hardening dur-
ing the first ECAP passes (one to four, depending on the material, tool geometry and processing conditions), followed by sat-
uration or even some small, yet noticeable, reduction of hardness and flow stress.
A dislocation density based crystal plasticity formulation in the spirit of Mughrabi’s cell wall/cell interior composite
model [350] and a finite-element computational method was used by Tóth [96] and Rezvanian et al. [506] in conjunction
with the dislocation density evolution approach proposed by Nix et al. [507] and Estrin et al. [255]. Their aim was to inves-
tigate grain subdivision into inhomogeneous deformation bands in an aluminium single crystal in cube orientation under
large rolling strains. Evolution equations relating to the dislocation densities in cell interiors and cell walls as well as their
dimensions were formulated in line with the above works and coupled to the multiple-slip crystal plasticity formalism. The
formation of an inhomogeneous dislocation miscrostructure predicted by Rezvanian et al. is illustrated in Fig. 37. The com-
putational analysis clearly shows the non-uniformity of deformation, with localised deformation bands forming and devel-
oping with strain. These bands are comprised of two families including relatively wide matrix bands (MBs) and narrow
transition bands (TBs) separating MBs, which were initiated and evolved across the thickness of the deforming aluminium
crystal. Analysis of slip systems showed that deformation and structure evolution in the TBs was mediated by a combination
of four active slip systems with equal activity, while in the MBs two dominant slip systems operated. These different com-
binations of active slip systems resulted in the microstructures with distinctly different morphologies. The TBs were shown
Fig. 38. High Pressure Tube Shearing: Tube (1) is continuously drawn through the opening between the die (2) and a mandrel (3) using draw-bench (4).
Shear strain is imposed within the thickness of the tube wall by rotation of either the die (2) or the mandrel (3), or both (from [489], reproduced with
permission).
to have a uniform dislocation-cell microstructure, whereas MBs appeared to have a cell block structure with narrower dis-
location cells. These predictions were found to be in good agreement with experimental measurements and observations
related to formation and evolution of MBs and TBs in cube-oriented single crystals subjected to large strain rolling. The anal-
ysis performed in Ref. [506] underscores that models based on dislocation density evolution in conjunction with crystal plas-
ticity can successfully account for the formation of deformation band patterns under large strain processing.
FEM tools based on dislocation density modelling were used for numerical simulation of tube channel pressing (TCP)
[508] - a variant of an SPD technique introduced for production of tubes with refined grain structure. The effects of the pro-
cessing routes, back pressure, and friction were included in the investigation of the deformation behaviour of commercial
purity aluminium tubes. Implementing a modified Estrin-Tóth-Molinari-Bréchet constitutive model for large strains [255]
in an FEM program allowed evaluating the mechanical response of the material to TCP deformation in fairly accurate agree-
ment with experiment.
Quite generally, strengthening of tubes by severe plastic deformation is attracting a great deal of attention due to the
importance of tubular products for a broad range of engineering applications. FEM simulations provide important assistance
in evaluating the behaviour of metallic materials during such processes and the mechanical properties of the tubes strength-
ened by SPD. Particular examples include the so called Cone-Cone method (CCM) [494] in which a cone-shaped tube is
deformed by severe twist under high hydrostatic pressure, and the techniques for producing gradient structures in thin-
walled cylindrical tubes, such as High Pressure Tube Twisting (HPTT) [315,509] and High Pressure Tube Shearing (HPTS)
[489]. An important advantage of the latter technique, which is illustrated schematically in Fig. 38, is its continuous charac-
ter, which makes it amenable to large-scale tube manufacturing. FEM simulations of this process were conducted using the
QForm software [510]. The results for low carbon steel are shown in Fig. 39. The picture on the left is a snapshot that captures
a narrow deformation zone (indicated by an arrow) moving along the tube. The picture on the right shows the distribution of
the equivalent von Mises strain within the tube wall. In the example seen in Fig. 39, the equivalent strain has a value of about
32 at the outer surface and 16 at the inner surface of the tube wall. This gradient of strain does not necessarily translate to a
pronounced gradient of microstructure characteristics, as even the lower value of strain is large enough for saturation of
Fig. 39. FEM simulation of the room temperature HPTS process for low carbon steel using the QForm software (from [489], reproduced with permission).
microstructure development to have taken place. Thus, reasonably uniform mechanical characteristics can be achieved
throughout the entire thickness of the wall. If required, a gradient in hardness and strength can be produced for smaller
degrees of shear deformation. FEM simulations provide a very useful means for fine-tuning the processing parameters for
the desired outcome in terms of uniformity or deliberately set non-uniformity of microstructure.
Another SPD process that has been studied at length by FEM simulations is twist extrusion (TE). In a series of publications,
cf. [511–514], which have been reviewed recently [484], various aspects of TE processing, including the role of friction, die
geometry, etc. were investigated by FEM in combination with experimental studies. In particular, these numerical studies
helped identifying the limitations associated with the assumption of simple shear that is often made in the TE process
modelling.
SPD techniques commonly target large-scale production of bulk UFG materials. Interesting niche applications of SPD pro-
cessing can be found in manufacturing of parts for microelectronics, micro-electro-mechanical systems (MEMS), and other
miniaturised engineering structures. Rosochowska et al. [515] pointed out that successful forming of metallic micro-
components requires particularly careful design of the processing route using small-scale tooling, especially considering lim-
ited handling capability. They demonstrated that FEM simulations can be helpful in identifying the best process configura-
tion before engaging in expensive experimental trials. Simulations were used to study material flow, the required force, and
the tool contact stress in micro-extrusion of a single conical pin. Different process configurations (forward vs. backward
extrusion), tool geometries (pin angles) and material models (coarse grained vs. ultrafine grained) were tested to gain a bet-
ter understanding of the process conditions relevant to microforming. Kim and Nam [516] performed a quantitative FEM
simulation and characterisation of the size effects in microscale coining process of copper where the grain size was compa-
rable with the die cavity. In this way, the initial microstructure could be optimised before the actual micro-processing.
Another successful exercise in FEM modelling of microforming by cup drawing was presented in Ref. [517]. 3D FEM simu-
lations were carried out using the ABAQUS software package in which the dislocation based constitutive model expressed by
Eqs. (89) and (90) was embedded. It was demonstrated that the drawing behaviour of metallic materials, particularly the
occurrence of a blank thickness effect, is captured by the model very well. The occurrence of the size effect was shown to
be governed by the ratio of the blank thickness to the grain size of the material, as illustrated by Fig. 40, which shows
the model prediction and the experimental data for the limit drawing ratio. This quantity is defined as the ratio of the largest
blank diameter drawn without failure to the punch diameter, and its calculated dependence on the ratio of the blank thick-
ness to the grain size shows the right kind of trend consistent with experiment.
A metal forming process at microscale, which does not fall in the category of SPD processing in a strict sense, yet has a
clear relevance to this group of techniques, is filling of vias for microelectronic devices by the so called forcefill [518]. In this
process, aimed at connecting a metallisation level with another circuit level through channels in a semiconducting layer, a
metal from a blanket film is forced into the via channels by applied pressure. The metal is in solid state and undergoes a large
plastic deformation when filling the vias that commonly have a submicron diameter. A prediction of the process conditions,
in terms of the applied pressure and temperature, for efficient via filling with aluminium was made with the aid of FEM sim-
ulations employing a commercial software package DEFORM 2D in conjunction with a phenomenological constitutive model
having some dislocation underpinning. A useful map in the pressure-temperature plane specifying the regions where effi-
cient via filling occurs was calculated on that basis.
Fig. 40. The limit drawing ratio as a function of the thickness t to grain size d ratio for a blank of ultrafine grained copper (after [517], reproduced with
permission).
5.2. Softening at large strains and dynamic recrystallisation
Coming back to the variation of strength with the cumulative equivalent strain (or the number of passes) under ECAP, we
should mention that the progressive rise of strength with a trend to saturation seen in Fig. 33 and accounted for by the two-
internal-variable model for copper represents only part of the story. Deformation to still larger strains, beyond the range
where monotonic rise of strength was observed, cf. Fig. 33, was reported to result in a decrease in strength of copper
[176]. This behaviour is illustrated in Fig. 41 which shows the results of uniaxial tensile testing of 99.95% pure copper by
up to 16 passes via ECAP route Bc. After the very first pass through the die, which corresponds approximately to a strain
of 1, a strong increase in strength with a significant decrease in ductility is observed, as shown in Fig. 41. With further passes
the yield strength r02 and the ultimate tensile stress rUTS increase and reach a maximum for the specimens subjected to four
ECAP passes. This behaviour is similar to that seen in Figs. 28 and 33. With further processing beyond four passes, however,
both r02 and rUTS drop. This is believed to be associated with development of inhomogeneous microstructure, as confirmed
by TEM observations and XRD measurements of the width of X-ray peak profiles. The total elongation is low (8–10%) and
remains nearly constant throughout all passes. However, as discussed above in connection with Fig. 41, the uniform elon-
gation (i.e. the strain to the onset of necking) increases with the number of passes from 0.75% to up to 2.5%. Qualitatively
similar behaviour was observed by Gubicza et al. [27] for several FCC metals, including Ag, Al, Au, Cu, and Ni, deformed
to different strains.
As is obvious from Fig. 41, we are dealing here with two strain softening effects. The first one is a decrease of the yield
strength of copper with pre-straining by ECAP for sufficiently large numbers of passes. The second one is a decrease of stress
with strain after a short stage of initial strain hardening under uniaxial tensile deformation of the ECAP-processed material.
One possible reason for strain softening during plastic deformation is dynamic recovery and/or dynamic recrystallisation
which is common at high homologous temperatures and/or large strains [519,520]. The available explanations for the strain
softening behaviour observed after severe plastic deformation are diverse and mostly qualitative, although considerable pro-
gress has been achieved in understanding the microstructural factors affecting this trend relating it to recovery and dynamic
recrystallisation processes [521–523].
Although the processes of interest - recovery, recrystallisation (and possibly grain growth) - are implicitly hidden in the
coefficients of the KME model based on dislocation density evolution (particularly in the parameters k0 and k2), the disloca-
tion density tends to saturate, but not to decrease, with strain in the one-internal-variable model [176,184,327,366,431,
520,524,525]. To account for the softening behaviour at large strain, Wei et al. [175] proposed a combined approach where
the dislocation density evolution in the spirit of the KME model and the Taylor relation for the friction stress rf are comple-
mented with a softening fraction of the flow stress X S which can be defined as:
rf r
XS ¼ ð94Þ
rf rs
Thus, the strain dependence of the flow stress can be expressed as r ¼ rf ðeÞ ðrf ðeÞ rs ÞX S ðeÞ. The softening kinetics
reflected in the strain dependence of X S can be described using an Avrami type equation [526] or an empirical relation pro-
posed by McQueen et al. [519] and then modified by Oudin et al. [527] and Wei et al. [175] to
q
e
X S ¼ 1 exp r ð95Þ
~e
Fig. 41. Stress–strain curves obtained by uniaxial tensile tests on pure Cu that has been processed by up to 16 ECAP passes performed at room temperature.
Fig. 42. Dependence of the yield strength of pure copper on the pre-strain produced by different numbers of ECAP passes (adopted from [175,176]).
The quantities r and q, together with ~e, which is close to the peak strain, are parameters that govern the softening kinetics
and are to be determined by fitting Eq. (95) to the experimental data. Although a satisfactory fit can be obtained for a variety
of experimental data, cf. Fig. 42, this heuristic approach to strain softening behaviour does not claim to be explanatory. The
model would also have difficulty accounting for the data of the kind presented in Fig. 43, which displays a non-monotonic
evolution of the dislocation density and the fraction of large-angle grain boundaries in copper to the backdrop of a mono-
tonic decrease of the grain size under ECAP processing. While the form of the intuitively chosen equations is suggested
by the theories of dynamic recrystallisation treated as a phase transformation [528–530], this blend of a dislocation-
based approach and an empirical one is somewhat artificial. Moreover, the parameters entering Eq. (95) are loosely defined
and cannot be related directly to microstructural characteristics of the material.
A survey of the existing modelling efforts towards predicting the grain size distribution obtained by SPD would be incom-
plete without a brief remark on the dynamic recrystallisation. Indeed, grain refinement during deformation at high homol-
ogous temperatures [529,531,532] (or very high strain rates [533]) is often governed by dynamic recrystallisation (DRX). This
phenomenon occurs in a deformed material through the nucleation and migration of high-angle grain boundaries that con-
stitute a distinct interface between the previously deformed material and the new fine-grained structure. Once recrystalli-
sation is triggered, the ensuing microstructural changes will affect the dislocation behaviour in the material. The
recrystallised microstructure typically comprises residual pre-existing grains with a high dislocation density, newly formed
fine grains which are initially dislocation-free, and evolving subgrains. As the average grain size decreases, the volume frac-
tion of grain boundary area rises. The DRX kinetics is usually described by an S-shaped exponential curve for the time depen-
dence of the volume fraction of the recrystallised material:
X DRX ¼ 1 exp½ðK DRX =d0 ÞtnDRX ð96Þ
Fig. 43. Characteristic evolution of the dominant microstructural parameters for copper with the number of ECAP passes (from [453], reprinted with permission).
Fig. 44. Results showing the average grain size (a) and relative dislocation density (b) distribution across the working Al billet during two ECAP passes via
route C (after Hallberg [534], reproduced with permission).
where K DRX and nDRX are constants determining the rate of DRX and d0 is the initial grain size of the non-recrystallised mate-
rial. With the progress of DRX, the internal boundaries, with their growing total area, will reduce the dislocation mobility.
Following this empirical approach, Hallberg et al. [534] described the evolution of the average grain size d from the initial
value d0 to the final saturation one, df , as a function of plastic strain
d ¼ d0 ðd0 df Þ exp½K X hepl epl c inX ð97Þ
with another pair of rate-controlling constants K X and nX . Here the McCauley brackets h
i indicate that no recrystallisation
will occur until the critical plastic strain epl c is attained, i.e. until epl > epl c . The strain-dependent average grain size d then
enters the KME equation for the dislocation density evolution, Eq. (51), as hki. The FEM simulations of room-temperature
ECAP with the die geometry defined by U ¼ 90o and W ¼ 20o showed, Fig. 44, that after one pass neither the grain size
nor the dislocation density are homogeneously distributed in the 20 20 150 mm3 Al billet. Largest deviations from
homogeneity were observed in the bottom part of the billet. After the second pass, the distribution of both grain size and
dislocation density was found to be almost homogeneous. To avoid empirical formulations including loosely defined con-
stants K DRX and nDRX or K X and nX , Galindo-Nava and Rivera-Díaz-del-Castillo [530] developed a thermodynamic approach
coupled with classical grain nucleation and growth formulations describing the grain size evolution during discontinuous
dynamic recrystallisation, which can potentially be incorporated in a more rigorous numerical modelling framework. With
the same motivation, Le and Kochmann [535] proposed a thermodynamically based model for the DRX process during SPD,
which also provided explicit evolution equations for the grain size and the dislocation density. Despite the simplicity of the
model, evaluation of these quantities on its basis showed reasonable (though still only qualitative) agreement with exper-
imental data.
6. Phase mixture modelling of nanocrystalline materials
The discussion in the foregoing sections made it clear that most SPD processing techniques produce an ultrafine grained
structure that falls short of being qualified as nanocrystalline one. However, modelling of the mechanical behaviour of true
nanocrystalline structures – regardless of the way in which it has been produced – is of interest in the context of this review.
Here we give a brief account of a model that is well suited to describe this behaviour. It was first proposed in 2000 [536] and
was further developed in subsequent work [537–539]. The main features of the model are as follows. The interior of the
grains and the grain boundaries are regarded as two separate phases. It is assumed that plastic strain rate in the grain bound-
aries is controlled by diffusion fluxes therein and is given by
X DGB
e_ GB ¼ A~ rGB ð98Þ
kB T d2
~ is a constant. This equation resem-
where rGB is the stress within the grain boundaries, X denotes the atomic volume and A
bling the familiar Nabarro-Herring creep equation [540,541] obtained from semi quantitative considerations [536] was con-
~ for copper to be
firmed by molecular dynamic calculations of Yamakov et al. [542] who found the value for the coefficient A
approximately 55.5. The plastic strain rate in the grain interior (GI) is controlled by an additive combination of dislocation
glide and diffusion processes of the Nabarro-Herring and Coble type:
e_ GI ¼ e_ disl þ e_ NH þ e_ Coble ð99Þ
with
m m=2
jrGI j q
e_ disl ¼ e_ 0 signðe_ ÞHðd dcrit Þ ð100Þ
r0 q0
X D
e_ NH ¼ 14 rGI ð101Þ
kB T d2
and
X w DGB
e_ Coble ¼ 14p rGI ð102Þ
kB T d d2
The Heaviside function Hðd dcrit Þ in Eq. (100) accounts for the fact that dislocation glide ceases to contribute to plastic
deformation for grain sizes below dcrit defined by Eq. (66). In Eq. (102), w denotes the grain boundary thickness, cf. Fig. 45.
The variation of the dislocation density entering Eq. (100) obeys an evolution equation similar to Eq. (56) with hki identified
with the average grain size d. In the phase-mixture model under consideration, the overall flow stress r is taken as the
weighted sum of the stresses in the grain interiors and grain boundaries:
r ¼ ð1 f ÞrGI þ f rGB ð103Þ
where f is the grain size dependent volume fraction of the grain boundaries.
As in the two-internal-variable model considered above, the iso-strain assumption is made. Accordingly, the strain (and
strain rate) in the grain interior and the grain boundaries is taken to be the same, as seen in the dashpot-spring diagram in
Fig. 45, which provides a recipe for the calculation of the flow stress described above. The diagram was augmented by a
sequential member associated with gradient terms [543], which were absent in the original model and will be considered
below.
The phase-mixture model has been quite successful with describing the mechanical response of nanocrystalline metals in
terms of their strain hardening behaviour and the grain size dependence of the yield stress. An example of the calculated
diagram truthfully predicting the Hall-Petch dependence of the r0.2 proof stress of copper and accounting for the ‘inverse’
Hall-Petch behaviour equally well is given in Fig. 46.
Fig. 45. Mechanisms underlying plastic flow in the interior of the grain and the grain boundaries. (a) nanocrystalline material with grain size d and grain
boundary width w. In the phase mixture model a nanocrystalline grain is approximated as being cube-shaped. (b) Rheological interpretation of phase
mixture model according to Eq. (99) including a gradient effect. Each dashpot represents a viscoplastic mechanism and the spring accounts for the gradient
effects (after [543], reproduced with permission).
Fig. 46. The grain size dependence of the r0.2 proof stress of nanocrystalline copper calculated using the phase-mixture model combined with the elastic-
viscoplastic (Taylor-Lin) approach [538]. The experimental data by Sanders et al. [544] are shown for comparison.
7. Gradient plasticity for SPD processed materials
While a lot of effort in SPD research went into developments leading to maximising homogeneity of the microstructure of
the processed material, there is also great value in processes capable of producing non-uniform, gradient microstructures
[545]. In the cited work, an excellent combination of strength and ductility of interstitial-free steel was achieved by surface
mechanical attrition treatment (SMAT). This was associated with a microstructure gradient induced by SMAT in a near-
surface layer of the material. The benefits of gradient structures were also discussed in a recent work on strengthening thin
walled tubes by a process referred to as continuous high pressure tube shearing [489]. Modelling of the mechanical beha-
viour of such materials calls for new theoretical tools that account for gradient structures. Of special interest in this regard
are gradient plasticity models that are microstructure-based.
A physically based model relating higher (second order) gradient terms to plastic strain incompatibility between the
adjoining grains was proposed by Estrin et al. [236,473]. The model results in an additive second derivative (Laplacian) term
in an expression for the flow stress, which is similar in structure to that proposed earlier by Aifantis [335], but has the benefit
of providing a link between the coefficient of this term to the grain size. It was shown [473] to account for homogenisation of
microstrucure and hardness across a specimen undergoing severe plastic deformation by HPT.
The role of gradient terms in providing a more uniform strain and enabling a more homogeneous strengthening effect of
SPD is demonstrated by the measurements of microhardness distribution in a specimen deformed by high pressure torsion.
By the very nature of deformation by torsion, the shear strain produced varies from zero to a maximum value with the radial
Fig. 47. Progressive rise of homogeneity of microhardness in an HPT specimen of aluminium with the number of turns (after Zhilyaev et al. [547]).
Fig. 48. Progressive increase of shear strain uniformity in an HPT specimen of copper with the number of anvil revolutions as calculated using a gradient
plasticity model [473].
distance from the specimen axis to its rim. However, with the growing twist angle, or the number of revolutions, the
microstructure and the microhardness were shown to become more and more uniform throughout a cross-section of an
HPT aluminium specimen [546], cf. Fig. 47.
This paradox is resolved in terms of the mentioned model by Estrin et al. [473] according to which strain gradients act to
equalise the dislocation cell size at different locations within the specimen. This is associated with progressive equalisation
of shear stress across an HPT specimen with the growing number of revolutions of the anvil, as illustrated in Fig. 48, which
shows the variation of the cumulative shear strain with the distance from the HPT specimen axis [473].
Another physically sensible variant of gradient plasticity associates a gradient term with mobile dislocation exchange
between adjoining regions of the material via dislocation cross slip [284]. Both variants of gradient plasticity were consid-
ered in the context of localised strain pattern formation in Mg alloy AZ31 under ECAP deformation [235]. Susceptibility of
ECAP processing to strain localisation, particularly due to heat release, was mentioned in Section 2.3, but even without ther-
momechanical coupling, strain localisation may occur, and gradient plasticity is a suitable way to account for it. Using linear
stability analysis, localised strain patterning in the shear zone of the alloy undergoing ECAP deformation was considered for
both variants of gradient plasticity theory in conjunction with the two-internal-variable model [254,255]. In this kind of
analysis, conditions for growth of small periodic perturbations of a uniform solution of the constitutive equations with
Fig. 49. Results of linear stability analysis for strain localisation under ECAP with back pressure of 130 MPa (a) and 260 MPa (b) in magnesium alloy AZ31
[235]. The perturbation growth rate is shown as a function of the wave number. The estimated average shear strain rates in the tests are shown for each
curve. The diamonds correspond to the experimentally observed wave numbers and are placed on theoretical curves for the respective strain rates. The
vertical dotted line on the left marks the position of the smallest theoretically possible wave number, while the short vertical dashed lines on the right mark
the largest possible wave numbers for the respective strain rates (after [431], reproduced with permission).
different wave numbers are determined. The temporal and spatial variation of the perturbations are represented by a factor
expðgt þ inxÞ; solutions for which the real part of the ‘growth parameter’ g is positive correspond to unstable behaviour.
This may be the case for various values of the wave number n. The one that corresponds to the highest rate of growth
(i.e. n ¼ nmax corresponding to a peak in the dispersion curve g ¼ f ðnÞ) obviously defines the length scale, i.e. the character-
istic wave length k ¼ 2p=nmax of the non-uniform strain pattern that develops to become predominant. A comparison
between the experimentally observed wave length of the strain localisation patterns with the calculated dispersion curves
in Fig. 49, demonstrates a very good agreement between the predictions and the measurements done for two different back
pressure levels used in ECAP processing.
In their model of gradient plasticity, Klusemann et al. [543] went beyond the second gradients and extended the phase-
mixture model presented in the foregoing section by including a fourth-order gradient. They augmented Eq. (103) for the
flow stress with gradient terms in the following form:
r ¼ f rGB þ ð1 f ÞrGI þ hg ðl2 @ 2x e þ 2=5l4 @ 4x eÞ ð104Þ

In the spirit of the Estrin-Mühlhaus model [236], the intrinsic length scale l can be identified with the average grain size;
hg is a numerical parameter whose value controls the magnitude of the gradient effects. Dealing with fourth-order gradients,
particularly in the context of FE simulations, is computationally challenging, but a great benefit of the model is that the
fourth-order gradient provides a regularisation which results in a stable computational scheme. The analysis based on the
model permits efficient treatment of patterning, as was shown in the computational exercises in [543].
8. Simulations of synthesis of architectured materials by SPD techniques
The use of SPD techniques to produce novel hybrid materials heralded as a new paradigm in materials design in [548] is
seen as an important emerging direction of research in the area of severe plastic deformation. Indeed, SPD methods can be
employed to engineer hybrid (multi-component) materials to a desired inner architecture while simultaneously producing
an ultrafine or nanoscale grain structure of the components of a hybrid [464,549]. This approach to fabrication of archi-
tectered hybrid materials was termed ‘SPD-induced synthesis’ [550,551]. This nascent area of research is a perfect play-
ground for modelling and numerical simulations, as the variants of thinkable inner architectures of such hybrid materials
are sheer countless. Screening of conceived designs and SPD processes leading to hybrids with promising mechanical prop-
erties can best be done by numerical simulations, primarily by FEM.
A hybrid design with potentially attractive properties is that with spiral-shaped armour embedded in a massive metallic
part (‘matrix’). Bouaziz [552] has shown that by embedding a hard spiral-shaped filament in a softer matrix provides the
material with enhanced tensile ductility, as the armour gives rise to extra strain-hardening thus delaying the onset of neck-
ing. This feature offers a possibility to alleviate a common drawback of UFG materials processed by SPD, viz. their low strain
hardening capability and the ensuing low ductility. There are various SPD and cognate techniques by which helical armour
fibres can be inserted into a massive billet. These include HPT [548], torsion [553], twist extrusion [554], screw rolling [555],
and high pressure torsion-extrusion (HPTE) [556] .
By way of example, we present a schematic showing the fabrication of a hybrid material with layered spiral armour archi-
tecture [551], Fig. 50. SPD-induced synthesis of architectured materials illustrated by Fig. 50 involves the following steps.
Prior to processing by SPD, inclusions in the form of fibres, solid particles, or powders are embedded in a workpiece. Con-
trolled material fluxes during the SPD process will transform them to produce a desired inner structure of the hybrid mate-
rial. Knowing this architecture, one can ‘pre-program’ the initial distribution of these embedded objects from the targeted
final architecture computationally, by solving an inverse problem.
An example of such ‘pre-programming’ of the initial configuration of reinforcing fibres to produce a desired final archi-
tecture is shown in Fig. 51 [550]. The SPD process applied is twist extrusion (TE). As seen from Fig. 51 that represents the
results of FEM simulations using the software package DEFORM 3D, five TE passes have transformed an initially straight cop-
per fibre inserted in a titanium matrix to a helical one. The cumulative strain involved in the process is high enough to impart
submicron-scale grain structure on both copper and titanium. In this way, the targeted spiral architecture is achieved with a
simultaneous strengthening of both constituents of the hybrid material due to extreme grain refinement.
With the computational tools available, there are practically no limits to the complexity of the targeted inner architecture
of a hybrid material. An example of ‘pre-programming’ the initial embedded structure to obtain a hyperboloid final structure
is shown in Fig. 52 [550]. On the fabrication side, new additive manufacturing technologies make it possible to produce ini-
tial structures with a pre-programmed geometry that may be required for forming hybrid materials with virtually any
desired inner architecture. Additional compaction and healing of possible defects in an additively manufactured preform
provided by SPD are an extra benefit of this processing strategy.
9. Texture evolution during SPD processing
The radical grain refinement effect of severe plastic deformation should not be considered in isolation from the variation
in the crystallographic texture it may cause. Indeed, the strengthening effect of grain refinement may be negated by unfa-
vourable texture. Hence, predicting texture evolution is an important goal of modelling of SPD processes. That is why already
Fig. 50. Producing a hybrid material with a spiral layered armour by the HPTE process: (a) initial arrangement of straight armour fibres in a billet that is
subjected to HPTE processing and (b) simulated and experimental spiral layered structure in copper armoured with aluminium. (Yellow: Cu matrix; blue:
aluminium fibres (transformed to spiral sheets by HPTE) (after [551], reproduced with permission).
Fig. 51. Shape transformation of an inclined straight fibre corresponding to the initial configuration of copper fibres within a titanium matrix (a) to a spiral
one (b) after five TE passes (after [550] reproduced with permission).
in early work dedicated to simulations of SPD processes of FCC metals, texture evolution was calculated alongside grain
refinement. This is exemplified by the work of Baik et al. [257] who employed the dislocation density-based strain hardening
model expressed by Eqs. (92) with some modifications [255]. The deformation history calculated by 2D FEM simulation with
ABAQUS software was combined with the full constraint Taylor model to compute textures produced by ECAP. A good agree-
ment with the measured textures was achieved. This may appear somewhat surprising, as the model used did not consider
grain subdivision that occurs during SPD processing.
An excellent overview of literature on SPD-induced texture development (specifically, for the ECAP process) was provided
by Beyerlein and Tóth [94]. At the time the Beyerlein-Tóth report was written, the effect of grain subdivision was not con-
sidered in texture simulations associated with severe plastic deformation. Inclusion of the grain subdivision effect in terms of
the model [454] described above improved the agreement between the calculated and the experimental texture emerging
under SPD processing [108]. Fig. 53 shows the results of FEM simulations of texture in the form of a {1 0 0} pole figure
for micro-extrusion of copper [557] vis-à-vis an experimental pole figure. The simulations were based on polycrystal plas-
ticity modelling (the viscoplastic self-consistent (VPSC) approach) and did include grain subdivision. A good agreement
between the computational results and the experimentally measured texture is a testimony to the potential crystal plasticity
modelling has in the context of SPD processing – of course, if the grain structure evolution is included in the texture calcu-
lations, which entails updating of the grain population throughout the simulation process.
Fig. 52. Simulated twist extrusion-based process of fabrication of a hybrid material: (a) ‘pre-programmed’ billet for obtaining the final hyperboloid
structure (b) by four TE passes (simulation by DEFORM 3D, [550]).
Fig. 53. Experimental (left) and predicted (right) {1 0 0} pole figure for micro-extruded UFG copper. Isolevels shown are 1.2, 2.0, 2.5, 3.0, 3.5, 4, 8, 12, 20, 30;
also shown is the texture intensity in terms of the multiples of random distribution (MRD) (after [557], reproduced with permission).
The viscoplastic self-consistent approach mentioned above is a powerful computational technique that permits simulta-
neous tracing of strain hardening and texture development. It was developed by Molinari et al. [445] and further elaborated
for anisotropic materials by Lebensohn and Tomé [558]. In this approach the behaviour of a polycrystalline aggregate is
related to the behaviour of a representative ensemble of monocrystalline grains. The components of the local stress ½r
and strain rate ½e_ tensors in a particular grain are considered to be different from the corresponding macroscopic quantities
_ The problem is reduced to the calculation of microscopic stresses and strain rates for each grain with explicit
½R and ½E.
account of possible active slip systems. The volume average of these quantities determines the mechanical response of
the polycrystalline aggregate. The problem of strain compatibility and stress equilibrium is solved by allowing some local
relaxations of both quantities, while keeping their averages equal to their respective macroscopic values. A particular grain
is considered as an Eshelby-type inclusion embedded in an equivalent homogeneous medium, whose behaviour is expressed
in terms of the volume-weighted average over the ensemble of grains. This yields a relation between the local and the
macroscopic quantities in the general form
e_ ij E_ ij ¼ a~ Mijkl ðrkl Rkl Þ ð105Þ
where ½M is the interaction tensor and a ~ is a constant which parametrises the interaction between an individual grain and
the homogeneous equivalent medium, covering the special cases of a ~ ¼ 0 for the upper-bound model corresponding to
homogenous strain (Taylor limit), a ~ ¼ 1 for the lower-bound corresponding
~ ¼ 1 for the tangent self-consistent model, and a
to homogeneous stress (Sachs limit).
A constitutive description of slip on multiple slip systems at the level of a grain can be furnished in terms of the dislo-
cation density evolution models of the kind presented in Section 3.3. The complexity of the problem stems from the need
to introduce dislocation densities for each slip system and to consider interactions between dislocations of the different slip
systems. This is challenging, but not impossible, as was demonstrated in the mentioned review by Beyerlein and Tóth [94].
Recently, Ayoub et al. [559] modified the crystal plasticity approach to incorporate the grain boundary sliding mechanism
and also account for grain subdivision. Their model provided an accurate prediction of texture evolution and the mechanical
response of the magnesium alloy AZ31B at different temperatures and strain rates.
It is believed that crystal plasticity modelling, which has been championed by Lebensohn and Tomé, cf., e.g. [558], and
advanced by Molinari and Tóth [96,497,560], Skrotzki et al. [561], Beyerlein and Tóth [94], Horstemeyer et al. [562], Roters
et al. [78], and others, offers the most suitable platform for texture simulations for polycrystalline materials. This applies to
modelling of the SPD processes and the mechanical properties of ultrafine-grained materials produced by SPD technology. As
mentioned above, in numerical simulations of the SPD processes their specifics, notably progressive grain fragmentation,
needs to be accounted for by updating the representative grain population.
10. Conclusion
In this article, we attempted to overview the extensive literature on the modelling and computational aspects of severe
plastic deformation processing. Starting from phenomenological models common in solid mechanics, we took the reader
through the various modelling approaches. Certainly, our preferred taste for mechanism-related and microstructure based
models was obvious in this exposé. We hope to have convinced the reader that this type of modelling has substantial benefits
over the more phenomenological approaches that are still common in solid mechanics literature. The entirety of the models
available provide a useful toolkit for both analytical work and numerical simulations, which can be employed ‘on demand’,
depending on the specifics of the SPD-related problem to be solved and the detail to which the microstructu information is
required. Examples of the use of these modelling approaches to particular SPD processes were given and practical advice on
the suitability of one or the other model was offered.
If, from the various elements presented in this overview, we were to construct an ‘ideal’ model to predict the material
behaviour and the load on the tooling in a given SPD process, along with the microstructure, texture, and the mechanical
properties of the processed material under subsequent service conditions, our model of choice would be as follows. We
would certainly opt for a crystal plasticity model of the Lebensohn-Tomé kind [94,563], but would include a provision for
grain subdivision as proposed in Ref. [557]. The description of strain hardening for individual slip systems of a crystallite
would be motivated by a two-internal-variable model presented in Section 3.3. Such a model would have sufficient flexibility
to monitor the microstructure and texture development, yet would be robust enough to be handled within a commercial
FEM code with a facility for implementing a constitutive material model through a user interface. If no microstructural detail
is needed for a particular application, a simple phenomenological constitutive description, such as the Voce [177], Johnson-
Cook [287], or Armstrong-Zerilli [288] model, can be used instead. We would certainly advise against the use of oversimpli-
fied models of the Ludwik or Ludwik-Hollomon type, as they do not capture the essential feature of large strain processing,
viz. a trend to stress saturation. Regardless of the choice of constitutive model, our preference would definitely be for a full
3D FEM simulation.
It is hoped that through this paper we helped eliminating a blind spot in the review literature on severe plastic deforma-
tion, provided the reader with a compendium of suitable modelling tools, and offered an outlook on possible future direc-
tions of analytical and computational work on SPD modelling.
Acknowledgements
The authors would like to thank Ronald Armstrong, Swantje Bargmann, Yan Beygelzimer, Hyoung Seop Kim, Benjamin Kluse-
mann, Roman Kulagin, Rimma Lapovok, Vincent Lemiale, Sebastian Mercier, Alain Molinari, Andrey Molotnikov, Laszlo Tóth,
Yuntian Zhu, and Igor Yasnikov for stimulating discussions. AV wishes to acknowledge financial support from the Russian
Science Foundation through the grant-in-aid No. 15-19-30025 and the State Assignment, contract #3.3881.2017/4.6. YE
acknowledges research funding received from the Ministry of Education and Science of the Russian Federation through grant
#14.A12.31.0001 to the National University of Science and Technology ‘‘MISIS” where part of this work was conducted.
References
[1] Valiev RZ, Estrin Y, Horita Z, Langdon TG, Zehetbauer MJ, Zhu YT. Producing bulk ultrafine-grained materials by severe plastic deformation. J Metals
2006;58:33–9.
[2] Valiev RZ, Islamgaliev RK, Alexandrov IV. Bulk nanostructured materials from severe plastic deformation. Prog Mater Sci 2000;45:103–89.
[3] Valiev RZ, Langdon TG. Principles of equal-channel angular pressing as a processing tool for grain refinement. Prog Mater Sci 2006;51:881–981.
[4] Estrin Y, Vinogradov A. Extreme grain refinement by severe plastic deformation: a wealth of challenging science. Acta Mater 2013;61:782–817.
[5] Bridgman PW. The tensile properties of several special steels and certain other materials under pressure. J Appl Phys 1946;17:201–12.
[6] Carreker RP, Hibbard WR. Tensile deformation of high-purity copper as a function of temperature, strain rate, and grain size. Acta Metall 1953;1:656.
[7] Kawasaki M, Lee H-J, Jang J-i, Langdon TG. Strengthening of metals through severe plastic deformation. Rev Adv Mater Sci 2017;48:13–24.
[8] Armstrong RW. 60 years of Hall-Petch: past to present nano-scale connections. Mater Trans 2014;55:2–12.
[9] Armstrong RW. Plasticity: grain size effects III. Reference module in materials science and materials engineering. Elsevier; 2016.
[10] Langer JS. Yielding transitions and grain-size effects in dislocation theory. Phys Rev E 2017;95.
[11] Gleiter H. On the structure of grain boundaries in metals. Mater Sci Eng 1982;52:91–131.
[12] Gleiter H. Nanocrystalline materials. Prog Mater Sci 1989;33:223–315.
[13] Hansen N. Plastic behavior of metals: large strains. In: Buschow KHJ, Cahn RW, Flemings CR, Ilschner B, Kramer EJ, Mahajan S, et al., editors.
Encyclopedia of materials: science and technology. Oxford: Elsevier; 2001. p. 7040–50.
[14] Hansen N. Hall-Petch relation and boundary strengthening. Scr Mater 2004;51:801–6.
[15] Chokshi AH, Kottada RS. The deformation characteristics of nanocrystalline metals. Trans Indian Inst Met 2005;58:939–50.
[16] Li Z, Hou C, Huang M, Ouyang C. Strengthening mechanism in micro-polycrystals with penetrable grain boundaries by discrete dislocation dynamics
simulation and Hall-Petch effect. Comput Mater Sci 2009;46:1124–34.
[17] Lyu H, Taheri-Nassaj N, Zbib HM. A multiscale gradient-dependent plasticity model for size effects. Phil Mag 2016;96:1883–908.
[18] Zontsika NA, Abdul-Latif A, Ramtani S. Pertinence of the grain size on the mechanical strength of polycrystalline metals. J Eng Mater Technol
2017;139:021017-1–021017-10.
[19] Armstrong RW, Balasubramanian N. Unified Hall-Petch description of nano-grain nickel hardness, flow stress and strain rate sensitivity
measurements. AIP Adv 2017;7:085010.
[20] Eshelby JD, Frank FC, Nabarro FRN. XLI. The equilibrium of linear arrays of dislocations. The London, Edinburgh, Dublin Philos Mag J Sci
1951;42:351–64.
[21] Cheng S, Spencer JA, Milligan WW. Strength and tension/compression asymmetry in nanostructured and ultrafine-grain metals. Acta Mater
2003;51:4505–18.
[22] Estrin Y, Kim HS, Nabarro FRN. A comment on the role of Frank-Read sources in plasticity of nanomaterials. Acta Mater 2007;55:6401–7.
[23] Estrin Y, Mecking H. A unified phenomenological description of work hardening and creep based on one-parameter models. Acta Metall
1984;32:57–70.
[24] Gubicza J, Chinh NQ, Szommer P, Vinogradov A, Langdon TG. Microstructural characteristics of pure gold processed by equal-channel angular
pressing. Scr Mater 2007;56:947–50.
[25] Balogh L, Ungar T, Zhao Y, Zhu YT, Horita Z, Xu C, et al. Influence of stacking-fault energy on microstructural characteristics of ultrafine-grain copper
and copper-zinc alloys. Acta Mater 2008;56:809–20.
[26] Vinogradov A, Maruyama M, Hashimoto S. On the role of dislocation hardening in the monotonic and cyclic strength of severely plastically deformed
metals. Scr Mater 2009;61:817–20.
[27] Gubicza J, Chinh NQ, Lábár JL, Dobatkin S, Hegedüs Z, Langdon TG. Correlation between microstructure and mechanical properties of severely
deformed metals. J Alloy Compd 2009;483:271–4.
[28] Vinogradov A, Maruyama M, Kaneko Y, Hashimoto S. Effect of dislocation hardening on monotonic and cyclic strength of severely deformed copper.
Phil Mag 2012;92:666–89.
[29] Starink MJ, Cheng X, Yang S. Hardening of pure metals by high-pressure torsion: a physically based model employing volume-averaged defect
evolutions. Acta Mater 2013;61:183–92.
[30] Jóni B, Schafler E, Zehetbauer M, Tichy G, Ungár T. Correlation between the microstructure studied by X-ray line profile analysis and the strength of
high-pressure-torsion processed Nb and Ta. Acta Mater 2013;61:632–42.
[31] Chen Y, Gao N, Sha G, Ringer SP, Starink MJ. Microstructural evolution, strengthening and thermal stability of an ultrafine-grained Al–Cu–Mg alloy.
Acta Mater 2016;109:202–12.
[32] Starink MJ. Dislocation versus grain boundary strengthening in SPD processed metals: Non-causal relation between grain size and strength of
deformed polycrystals. Mater Sci Eng, A 2017;705:42–5.
[33] Osakada K. History of plasticity and metal forming analysis. J Mater Process Technol 2010;210:1436–54.
[34] McDowell DL. A perspective on trends in multiscale plasticity. Int J Plast 2010;26:1280–309.
[35] Edalati K, Horita Z. A review on high-pressure torsion (HPT) from 1935 to 1988. Mater Sci Eng, A 2016;652:325–52.
[36] Cottrell A. Theory of crystal dislocations. London, Glasgow; printed in Belgium: Blackie & Son; 1965.
[37] Mura T. American Society of Mechanical Engineers. Applied Mechanics Division. Mathematical theory of dislocations. New York: American Society of
Mechanical Engineers; 1969.
[38] Hirth JP, Lothe J. Theory of dislocations. 2nd ed. New York: Wiley; 1982.
[39] Nabarro FRN. Theory of crystal dislocations. Oxford: Clarendon Press; 1967.
[40] Friedel J. Dislocations. 1st English ed. Oxford, New York: Pergamon Press; 1964.
[41] Kubin L, Oxford University Press. Dislocations, mesoscale simulations and plastic flow. Oxford series on materials modelling 5. Oxford: Oxford
University Press; 2013. p. 1 online resource.
[42] Yip S. Handbook of materials modeling. Dordrecht; [London]: Springer; 2005.
[43] Tadmor EB, Miller RE. Modeling materials: continuum, atomistic and multiscale techniques. Cambridge University Press; 2011.
[44] Tadmor EB, Miller RE, Elliott RS. Continuum mechanics and thermodynamics: from fundamental concepts to governing equations. Cambridge
University Press; 2011.
[45] Li J. Atomistic calculation of mechanical behavior. In: Yip S, editor. Handbook of materials modeling. Netherlands: Springer; 2005. p. 773–92.
[46] Kaxiras E, Yip S. Introduction: atomistic nature of materials. In: Yip S, editor. Handbook of materials modeling. Netherlands: Springer; 2005. p. 451–8.
[47] Jones RE, Weinberger CR, Coleman SP, Tucker GJ. Introduction to atomistic simulation methods. In: Weinberger CR, Tucker GJ, editors. Multiscale
materials modeling for nanomechanics. Cham: Springer International Publishing; 2016. p. 1–52.
[48] Van der Giessen E, Needleman A. Discrete dislocation plasticity. In: Yip S, editor. Handbook of materials modeling. Netherlands: Springer; 2005. p.
1115–31.
[49] Zbib HM, Khraishi TA. Dislocation dynamics. In: Yip S, editor. Handbook of materials modeling. Netherlands: Springer; 2005. p. 1097–114.
[50] Vladimirov VI, Romanov AE. Disclinations in crystalline solids. Amsterdam: North Holland; 1991. p. 191.
[51] Groma I, Bakó B. Dislocation patterning: from micro- to mesoscale description. Phys Rev Lett 2000;84:1487.
[52] Groma I, Bako B. Linking different scales: discrete, self-consistent field, and stochastic dislocation dynamics. Mater Sci Eng, A 2001;309–310:356–9.
[53] Groma I, Balogh P. Investigation of dislocation pattern formation in a two-dimensional self-consistent field approximation. Acta Mater
1999;47:3647–54.
[54] Hähner P. Stochastic dislocation patterning during cyclic plastic deformation. Appl Phys A Mater Sci Process 1996;63:45–55.
[55] Teodosiu C. Large plastic deformation of crystalline aggregates. Wien etc.: Springer; 1997.
[56] Argon AS. Strengthening mechanisms in crystal plasticity. Oxford: Oxford University Press; 2008.
[57] Chakrabarty J. Applied plasticity. 2nd ed. New York: Springer; 2010.
[58] Gil Sevillano J, van Houtte P, Aernoudt E. Large strain work hardening and textures. Prog Mater Sci 1980;25:69–134.
[59] Walgraef D, Aifantis EC. On the formation and stability of dislocation patterns-II: Two-dimensional considerations. Int J Eng Sci 1985;23:1359–64.
[60] Walgraef D, Aifantis EC. Dislocation patterning in fatigued metals as a result of dynamical instabilities. J Appl Phys 1985;58:688–91.
[61] Walgraef D, Aifantis EC. On the formation and stability of dislocation patterns-I: One-dimensional considerations. Int J Eng Sci 1985;23:1351–8.
[62] Walgraef D, Aifantis EC. On the formation and stability of dislocation patterns-III: Three-dimensional considerations. Int J Eng Sci 1985;23:1365–72.
[63] Hähner P. A theory of dislocation cell formation based on stochastic dislocation dynamics. Acta Mater 1996;44:2345–52.
[64] Hähner P, Zaiser M. Dislocation dynamics and work hardening of fractal dislocation cell structures. Mater Sci Eng, A 1999;272:443–54.
[65] Panin VE, Grinyaev YV, Elsukova TF, Ivanchin AG. Structural levels of deformation in solids. Russ Phys J 1982;25:479–97.
[66] Estrin Y. Syntheses: playing scales - a brief summary. Modell Simul Mater Sci Eng 1999;7:747–51.
[67] Raabe D. Computational materials science the simulation of materials, microstructures and properties. Weinheim etc.: Wiley-VCH; 1998.
[68] Roters F, Eisenlohr P, Hantcherli L, Tjahjanto DD, Bieler TR, Raabe D. Overview of constitutive laws, kinematics, homogenization and multiscale
methods in crystal plasticity finite-element modeling: theory, experiments, applications. Acta Mater 2010;58:1152–211.
[69] Curtin WA, Ronald EM. Atomistic/continuum coupling in computational materials science. Modell Simul Mater Sci Eng 2003;11:R33.
[70] Shilkrot LE, Miller RE, Curtin WA. Coupled atomistic and discrete dislocation plasticity. Phys Rev Lett 2002;89:025501.
[71] Horstemeyer MF. Multiscale modeling: a review. Practical aspects of computational chemistry: methods, concepts and
applications. Netherlands: Springer; 2010. p. 87–135.
[72] Horstemeyer MF. Mesoscale/macroscale computational methods. In: Yip S, editor. Handbook of materials modeling. Netherlands: Springer; 2005. p.
1071–5.
[73] Pippan R, Gumbsch P. Multiscale modelling of plasticity and fracture by means of dislocation mechanics. Courses and lectures; no 522. Wien New
York: Springer Verlag; 2010. p. 394.
[74] Weygand D, Mrovec M, Hochrainer T, Gumbsch P. Multiscale simulation of plasticity in bcc metals. Annual review of materials research. Annual
Reviews Inc.; 2015. p. 369–90.
[75] Koester A, Ma A, Hartmaier A. Atomistically informed crystal plasticity model for body-centered cubic iron. Acta Mater 2012;60:3894–901.
[76] Blanckenhagen Bv, Gumbsch P, Arzt E. Dislocation sources in discrete dislocation simulations of thin-film plasticity and the Hall-Petch relation.
Modell Simul Mater Sci Eng 2001;9:157.
[77] Raabe D, Sachtleber M, Zhao Z, Roters F, Zaefferer S. Micromechanical and macromechanical effects in grain scale polycrystal plasticity
experimentation and simulation. Acta Mater 2001;49:3433–41.
[78] Roters F, Eisenlohr P, Bieler TR, Raabe D. Crystal plasticity finite element methods in materials science and engineering. Weinheim: Wiley-Vch; 2010.
[79] Phillips R. Crystals, defects and microstructures: modeling across scales. Cambridge, New York: Cambridge University Press; 2001.
[80] Tadmor EB, Ortiz M, Phillips R. Quasicontinuum analysis of defects in solids. Philos Mag A 1996;73:1529–63.
[81] The Minerals M, Materials S. Modeling Across Scales: a roadmapping study for connecting materials models and simulations across length and time
scales. Warrendale (PA): TMS; 2015.
[82] Lemaitre J, Chaboche J-L. Mechanics of solid materials. Repr. 2000, transferred to digital printing 2002 ed. Cambridge: Cambridge University Press;
2002.
[83] Bower AF. Applied mechanics of solids. Boca Raton: CRC Press; 2010.
[84] Zienkiewicz OC, Cheung YK. The finite element method in structural and continuum mechanics: numerical solution of problems in structural and
continuum mechanics. London, New York etc.: McGraw-Hill; 1967.
[85] Zienkiewicz OC, Taylor RL. The finite element method for solid and structural mechanics. 6th ed. Oxford; Burlington (MA): Elsevier Butterworth-
Heinemann; 2005.
[86] Prathap G. The finite element method in structural mechanics: principles and practice of design of field-consistent elements for structural and solid
mechanics. Dordrecht; Boston: Kluwer Academic; 1993.
[87] Landau LD, Lifshitz EM. Theory of elasticity. 2nd English ed. Oxford, New York: Pergamon Press; 1970.
[88] Kosevich AM. Crystal Dislocations and Theory of Elasticity. In: Nabarro FRN, editor. Dislocations in Solids. Amsterdam: North-Holland; 1979. p. 33.
[89] Kobayashi So OhS-I, Altan T. Metal forming and the finite-element method. New York: Oxford University Press; 1989.
[90] Nowick AS, Berry BS. Anelastic relaxation in crystalline solids. New York, USA: Academic Press; 1972.
[91] Zaïri F, Aour B, Gloaguen JM, Naït-Abdelaziz M, Lefebvre JM. Numerical modelling of elastic–viscoplastic equal channel angular extrusion process of a
polymer. Comput Mater Sci 2006;38:202–16.
[92] Kamrani M, Levitas VI, Feng B. FEM simulation of large deformation of copper in the quasi-constrain high-pressure-torsion setup. Mater Sci Eng, A
2017;705:219–30.
[93] Cerri E, De Marco PP, Leo P. FEM and metallurgical analysis of modified 6082 aluminium alloys processed by multipass ECAP: influence of material
properties and different process settings on induced plastic strain. J Mater Process Technol 2009;209:1550–64.
[94] Beyerlein IJ, Tóth LS. Texture evolution in equal-channel angular extrusion. Prog Mater Sci 2009;54:427–510.
[95] Anand L. Single-crystal elasto-viscoplasticity: application to texture evolution in polycrystalline metals at large strains. Comput Methods Appl Mech
Eng 2004;193:5359–83.
[96] Tóth LS. Modelling of strain hardening and microstructural evolution in equal channel angular extrusion. Comput Mater Sci 2005;32:568–76.
[97] Acharya A, Beaudoin AJ. Grain-size effect in viscoplastic polycrystals at moderate strains. J Mech Phys Solids 2000;48:2213–30.
[98] Aoyagi Y, Tsuru T, Shimokawa T. Crystal plasticity modeling and simulation considering the behavior of the dislocation source of ultrafine-grained
metal. Int J Plast 2013.
[99] Aoyagi Y, Shizawa K. Multiscale crystal plasticity modeling based on geometrically necessary crystal defects and simulation on fine-graining for
polycrystal. Int J Plast 2007;23:1022–40.
[100] Narutani T, Takamura J. Grain-size strengthening in terms of dislocation density measured by resistivity. Acta Metall Mater 1991;39:2037–49.
[101] Beaudoin AJ, Acharya A, Chen SR, Korzekwa DA, Stout MG. Consideration of grain-size effect and kinetics in the plastic deformation of metal
polycrystals. Acta Mater 2000;48:3409–23.
[102] Beaudoin AJ, Acharya A. A model for rate-dependent flow of metal polycrystals based on the slip plane lattice incompatibility. Mater Sci Eng, A
2001;309–310:411–5.
[103] Arsenlis A, Parks DM, Becker R, Bulatov VV. On the evolution of crystallographic dislocation density in non-homogeneously deforming crystals. J Mech
Phys Solids 2004;52:1213–46.
[104] Asaro RJ, Needleman A. Overview no. 42 Texture development and strain hardening in rate dependent polycrystals. Acta Metall 1985;33:923–53.
[105] Tóth LS. Texture evolution in severe plastic deformation by equal channel angular extrusion. Adv Eng Mater 2003;5:308–16.
[106] Kalidindi SR, Donohue BR, Li S. Modeling texture evolution in equal channel angular extrusion using crystal plasticity finite element models. Int J Plast
2009;25:768–79.
[107] Tóth LS, Estrin Y, Lapovok R, Gu C. A model of grain fragmentation based on lattice curvature. Acta Mater 2010;58:1782–94.
[108] Gu CF, Tóth LS. Texture development and grain refinement in non-equal-channel angular-pressed Al. Scr Mater 2012;67:33–6.
[109] Humphreys J, Bate P. Gradient plasticity and deformation structures around inclusions. Scr Mater 2003;48:173–8.
[110] Aifantis EC. Update on a class of gradient theories. Mech Mater 2003;35:259–80.
[111] Aifantis EC. Mechanics aspects in micro- and nano-scales. New York, NY, United States: Publ by ASME; 1992. p. 1–3.
[112] Fleck NA, Hutchinson JW. Strain gradient plasticity. In: John WH, Theodore YW, editors. Advances in applied mechanics. Elsevier; 1997. p. 295–361.
[113] Fleck NA, Muller GM, Ashby MF, Hutchinson JW. Strain gradient plasticity: theory and experiment. Acta Metall Mater 1994;42:475–87.
[114] Nye JF. Some geometrical relations in dislocated crystals. Acta Metall 1953;1:153–62.
[115] Zhang X, Aifantis K. Interpreting the internal length scale in strain gradient plasticity. Rev Adv Mater Sci 2015;41:72–83.
[116] Needleman A, Van Der Giessen E. Discrete dislocation and continuum descriptions of plastic flow. Mater Sci Eng, A 2001;309–310:1–13.
[117] Hiratani M, Zbib HM. Stochastic dislocation dynamics for dislocation-defects interaction: a multiscale modeling approach. J Eng Mater Technol
2002;124:335–41.
[118] Deshpande VS, Needleman A, Van der Giessen E. Finite strain discrete dislocation plasticity. J Mech Phys Solids 2003;51:2057–83.
[119] Kubin LP, Fressengeas C, Ananthakrishna G. Chapter 57 Collective behaviour of dislocations in plasticity. In: Duesbery FRNNaMS, editor. Dislocations
in Solids. Elsevier; 2002. p. 101–92.
[120] Zepeda-Ruiz LA, Stukowski A, Oppelstrup T, Bulatov VV. Probing the limits of metal plasticity with molecular dynamics simulations. Nature 2017.
advance online publication.
[121] Srinivasan R, Cherukuri B, Chaudhury PK. Scaling up of equal channel angular pressing (ECAP) for the production of forging stock. Nanomater Severe
Plastic Deform 2006;503–504:371–8.
[122] Orlov D, Raab G, Lamark TT, Popov M, Estrin Y. Improvement of mechanical properties of magnesium alloy ZK60 by integrated extrusion and equal
channel angular pressing. Acta Mater 2011;59:375–85.
[123] Raab GJ, Valiev RZ, Lowe TC, Zhu YT. Continuous processing of ultrafine grained Al by ECAP-Conform. Mater Sci Eng, A 2004;382:30–4.
[124] Jin YG, Baek HM, Im Y-T, Jeon BC. Continuous ECAP process design for manufacturing a microstructure-refined bolt. Mater Sci Eng, A 2011;530:462–8.
[125] Jin YG, Baek HM, Hwang SK, Im Y-T, Jeon BC. Continuous high strength aluminum bolt manufacturing by the spring-loaded ECAP system. J Mater
Process Technol 2012;212:848–55.
[126] Valiev RZ, Zhilyaev AP, Langdon TG. Bulk nanostructured materials: fundamentals and applications; 2014.
[127] Rosochowski A, Olejnik L. Finite element simulation of severe plastic deformation processes. Proc Inst Mech Eng Part L-J Mater-Des Appl
2007;221:187–96.
[128] Medeiros Nd, Moreira LP, Bressan JD, Lins JFC, Gouvêa JPd. Sensitivity analysis of the ECAE process via 2k experiments design. Matéria (Rio de Janeiro)
2010;15:208–17.
[129] Segal VM, Reznikov VI, Drobyshevkiy AE, Kopylov VI. Plastic metal working by simple shear. Russian Metall 1981;99.
[130] Segal VM. Severe plastic deformation: simple shear versus pure shear. Mater Sci Eng, A 2002;338:331–44.
[131] Segal VM. Slip line solutions, deformation mode and loading history during equal channel angular extrusion. Mater Sci Eng, A 2003;345:36–46.
[132] Khoddam S, Hodgson PD, Zarei-Hanzaki A, Yan Foon L. A simple model for material’s strengthening under high pressure torsion. Mater Des
2016;99:335–40.
[133] Khoddam S. A detailed model of high pressure torsion. Mater Sci Eng, A 2017;683:256–63.
[134] Kleiber M. Incremental finite element modelling in non-linear solid mechanics. Chichester, England: E. Horwood Halsted Press; 1989.
[135] Kim HS. Evaluation of Strain Rate During Equal-channel Angular Pressing. J Mater Res 2002;17:172–9.
[136] Iwahashi Y, Horita Z, Nemoto M, Langdon TG. The process of grain refinement in equal-channel angular pressing. Acta Mater 1998;46:3317–31.
[137] Alkorta J, Gil Sevillano J. A comparison of FEM and upper-bound type analysis of equal-channel angular pressing (ECAP). J Mater Process Technol
2003;141:313–8.
[138] Paydar MH, Reihanian M, Ebrahimi R, Dean TA, Moshksar MM. An upper-bound approach for equal channel angular extrusion with circular cross-
section. J Mater Process Technol 2008;198:48–53.
[139] Luis-Perez CJ, Luri-Irigoyen R, Gaston-Ochoa D. Finite element modelling of an Al-Mn alloy by equal channel angular extrusion (ECAE). J Mater Process
Technol 2004;153–54:846–52.
[140] Luis Pérez CJ, Luri R. Comparative analysis of actual processing conditions in ECAE between FEM and both analytical and experimental results. Mater
Manuf Processes 2011;26:1147–56.
[141] Eivani AR, Karimi Taheri A. An upper bound solution of ECAE process with outer curved corner. J Mater Process Technol 2007;182:555–63.
[142] Eivani AR, Taheri AK. A new method for estimating strain in equal channel angular extrusion. J Mater Process Technol 2007;183:148–53.
[143] Luis Pérez CJ. On the correct selection of the channel die in ECAP processes. Scr Mater 2004;50:387–93.
[144] Luis Pérez CJ, Luri R. Study of the ECAE process by the upper bound method considering the correct die design. Mech Mater 2008;40:617–28.
[145] Luri R, Luis CJ, Leon J, Sebastian MA. A new configuration for equal channel angular extrusion dies. J Manuf Sci Eng-Trans ASME 2006;128:860–5.
[146] Luri R, Luis CJ. Study of ECAE process by using FEM. Adv Mater Process Technol 2006;526:193–8.
[147] Luri R, Luis Pérez CJ. Modeling of the processing force for performing ECAP of circular cross-section materials by the UBM. Int J Adv Manuf Technol
2012;58:969–83.
[148] Luri R, Luis Pérez CJ. Analysis and modelling by finite element method of the equal channel angular extrusion pressure. Proc Inst Mech Eng, Part B: J
Eng Manuf 2010;224:925–35.
[149] Sßimsßir C, Karpuz P, Gür CH. Quantitative analysis of the influence of strain hardening on equal channel angular pressing process. Comput Mater Sci
2010;48:633–9.
[150] Segal VM. Materials processing by simple shear. Mater Sci Eng, A 1995;197:157–64.
[151] Segal VM, Reznikov VI, Kopylov VI, Pavlik DA, Malyshev VF. Processes of plastic structure formation of metals. Minsk, Belarus: Nauka i Tehnika; 1994
[in Russian].
[152] Iwahashi Y, Wang J, Horita Z, Nemoto M, Langdon TG. Principle of equal-channel angular pressing for the processing of ultra-fine grained materials.
Scr Mater 1996;35:143–6.
[153] Goforth RE, Hartwig KT, Cornwell LR. Severe plastic deformation of materials by equal channel angular extrusion (ECAE). In: Lowe T, Valiev R, editors.
Investigations and applications of severe plastic deformation. Netherlands: Springer; 2000. p. 3–12.
[154] Milind TR, Date PP. Analytical and finite element modeling of strain generated in equal channel angular extrusion. Int J Mech Sci 2012;56:26–34.
[155] Aida T, Matsuki K, Horita Z, Langdon TG. Estimating the equivalent strain in equal-channel angular pressing. Scr Mater 2001;44:575–9.
[156] Tóth LS, Arruffat Massion R, Germain L, Baik SC, Suwas S. Analysis of texture evolution in equal channel angular extrusion of copper using a new flow
field. Acta Mater 2004;52:1885–98.
[157] Panigrahi A, Scheerbaum N, Chekhonin P, Scharnweber J, Beausir B, Hockauf M, et al. Effect of back pressure on material flow and texture in ECAP of
aluminum. IOP Conf Series: Mater Sci Eng 2014;63:012153.
[158] Hasani A, Tóth LS, Beausir B. Principles of nonequal channel angular pressing. J Eng Mater Technol, Trans ASME 2010;132:0310011–310019.
[159] Beyerlein IJ, Tomé CN. Analytical modeling of material flow in equal channel angular extrusion (ECAE). Mater Sci Eng, A 2004;380:171–90.
[160] Bridgman PW. Effects of high shearing stress combined with high hydrostatic pressure. Phys Rev 1935;48:825–47.
[161] Bridgman PW. Shearing phenomena at high pressure of possible importance for geology. J Geol 1936;44:653–69.
[162] Dieter GE. Mechanical metallurgy. 3rd ed. New York a. o.: McGraw-Hill; 1986.
[163] Pippan R, Scheriau S, Hohenwarter A, Hafok M. Advantages and limitations of HPT: a review. In: Estrin YMHJ, editor. 4th International conference on
nanomaterials by severe plastic deformation. Goslar, GERMANY: Trans Tech Publications Ltd; 2008. p. 16–21.
[164] Beygelzimer Y, Kulagin R, Toth LS, Ivanisenko Y. The self-similarity theory of high pressure torsion. Beilstein J Nanotechnol 2016;7:1267–77.
[165] Onaka S. Appropriateness of the hencky equivalent strain as the quantity to represent the degree of severe plastic deformation. Mater Trans
2012;53:1547–8.
[166] Jonas JJ, Ghosh C, Toth LS. The equivalent strain in high pressure torsion. Mater Sci Eng, A 2014;607:530–5.
[167] Beygelzimer Y, Toth LS, Jonas JJ. Some physical characteristics of strain hardening in severe plastic deformation. Adv Eng Mater 2015;17:1783–91.
[168] Kim HS, Seo MH, Hong SI. Plastic deformation analysis of metals during equal channel angular pressing. J Mater Process Technol 2001;113:622–6.
[169] Yamaguchi D, Horita Z, Nemoto M, Langdon TG. Significance of adiabatic heating in equal-channel angular pressing. Scr Mater 1999;41:791–6.
[170] Yamaguchi D, Horita Z, Fujinami T, Nemoto M, Langdon TG. Factors affecting grain refinement in equal-channel angular pressing. Aluminium Alloys:
Phys Mech Properties 2000;Pts 1–3(331–3):607–12.
[171] Berbon P, Furukawa M, Horita Z, Nemoto M, Langdon T. Influence of pressing speed on microstructural development in equal-channel angular
pressing. Metall Mater Trans A 1999;30:1989–97.
[172] Antolovich SD, Armstrong RW. Plastic strain localization in metals: origins and consequences. Prog Mater Sci 2014;59:1–160.
[173] Komura S, Horita Z, Nemoto M, Langdon TG. Influence of stacking fault energy on microstructural development in equal-channel angular pressing. J
Mater Res 1999;14:4044–50.
[174] Horita Z, Fujinami T, Nemoto M, Langdon T. Equal-channel angular pressing of commercial aluminum alloys: grain refinement, thermal stability and
tensile properties. Metall Mater Trans A 2000;31:691–701.
[175] Wei W, Wei KX, Fan GJ. A new constitutive equation for strain hardening and softening of fcc metals during severe plastic deformation. Acta Mater
2008;56:4771–9.
[176] Dalla Torre F, Lapovok R, Sandlin J, Thomson PF, Davies CHJ, Pereloma EV. Microstructures and properties of copper processed by equal channel
angular extrusion for 1–16 passes. Acta Mater 2004;52:4819–32.
[177] Voce E. The relationship between stress and strain for homogeneous deformation. J Inst Met 1948;74:537–62.
[178] Panin VE, Likhachev VA, Grinyaev YV. Structural levels of deformation of solids. Novosibirsk, Russia: Nauka Publishers; 1985 [in Russian].
[179] Wetscher F, Vorhauer A, Pippan R. Strain hardening during high pressure torsion deformation. Mater Sci Eng, A 2005;410:213–6.
[180] Csanádi T, Chinh NQ, Gubicza J, Langdon TG. Plastic behavior of fcc metals over a wide range of strain: Macroscopic and microscopic descriptions and
their relationship. Acta Mater 2011;59:2385–91.
[181] Mecking H, Lücke K. Quantitative analyse der bereich III-verfestigung von silber-einkristallen. Acta Metall 1969;17:279–89.
[182] Kocks UF. Laws for work-hardening and low-temperature creep. J Eng Mater Technol-Trans ASME 1976;98:76–85.
[183] Mecking H. Description of hardening curves of fcc single- and polycrystals. In: Thompson AW, editor. Work hardening in tension and fatigue:
proceedings of a symposium, Cincinnati, Ohio, November 11, 1975. The Metallurgical Society of AIME; 1977. p. 67–90.
[184] Kocks UF, Mecking H. Physics and phenomenology of strain hardening: the FCC case. Prog Mater Sci 2003;48:171–273.
[185] Chinh NQ, Horvath G, Horita Z, Langdon TG. A new constitutive relationship for the homogeneous deformation of metals over a wide range of strain.
Acta Mater 2004;52:3555–63.
[186] Chinh NQ, Illy J, Horita Z, Langdon TG. Using the stress–strain relationships to propose regions of low and high temperature plastic deformation in
aluminum. Mater Sci Eng, A 2005;410–411:234–8.
[187] Prangnell PB, Harris C, Roberts SM. Finite element modelling of equal channel angular extrusion. Scr Mater 1997;37:983–9.
[188] Valberg HS. Applied metal forming: including FEM analysis. Cambridge: Cambridge University Press; 2010.
[189] Segal VM. Equal channel angular extrusion: from macromechanics to structure formation. Mater Sci Eng, A 1999;271:322–33.
[190] DeLo DP, Semiatin SL. Finite-element modeling of nonisothermal equal-channel angular extrusion. Metall Mater Trans A 1999;30:1391–402.
[191] Semiatin SL, DeLo DP, Shell EB. The effect of material properties and tooling design on deformation and fracture during equal channel angular
extrusion. Acta Mater 2000;48:1841–51.
[192] Li S, Bourke MAM, Beyerlein IJ, Alexander DJ, Clausen B. Finite element analysis of the plastic deformation zone and working load in equal channel
angular extrusion. Mater Sci Eng, A 2004;382:217–36.
[193] Son I-H, Lee J-H, Im Y-T. Finite element investigation of equal channel angular extrusion with back pressure. J Mater Process Technol
2006;171:480–7.
[194] Dumoulin S, Roven HJ, Werenskiold JC, Valberg HS. Finite element modeling of equal channel angular pressing: Effect of material properties, friction
and die geometry. Mater Sci Eng, A 2005;410–411:248–51.
[195] Krallics G, Budilov IN, Alexandrov IV, Raab GI, Zhernakov VS, Valiev RZ. Computer simulation of equal-channel angular pressing of tungsten by means
of the finite element method. nanomaterials by severe plastic deformation. Wiley-VCH Verlag GmbH & Co. KGaA; 2005. p. 271–7.
[196] Bowen JR, Gholinia A, Roberts SM, Prangnell PB. Analysis of the billet deformation behaviour in equal channel angular extrusion. Mater Sci Eng, A
2000;287:87–99.
[197] Oruganti RK, Subramanian PR, Marte JS, Gigliotti MF, Amancherla S. Effect of friction, backpressure and strain rate sensitivity on material flow during
equal channel angular extrusion. Mater Sci Eng, A 2005;406:102–9.
[198] Suo T, Li Y, Guo Y, Liu Y. The simulation of deformation distribution during ECAP using 3D finite element method. Mater Sci Eng, A 2006;432:269–74.
[199] Kim HS, Seo MH, Hong SI. Finite element analysis of equal channel angular pressing of strain rate sensitive metals. J Mater Process Technol 2002;130–
131:497–503.
[200] Djavanroodi F, Ebrahimi M. Effect of die channel angle, friction and back pressure in the equal channel angular pressing using 3D finite element
simulation. Mater Sci Eng, A 2010;527:1230–5.
[201] Ono M, Mizufune H, Narita M. Development of semicontinuous 4-stage ECAE method. In: Yanagimoto NKHNJ, editor. Proceedings of the 7th
international conference on technology of plasticity. Yokohama, Japan: Japan Society for Technology of Plasticity; 2002. p. 1249–54.
[202] Raab GI. Plastic flow at equal channel angular processing in parallel channels. Mater Sci Eng, A 2005;410–411:230–3.
[203] Rosochowski A, Olejnik L. Numerical and physical modelling of plastic deformation in 2-turn equal channel angular extrusion. J Mater Process
Technol 2002;125–126:309–16.
[204] Nagasekhar AV, Tick-Hon Y. Optimal tool angles for equal channel angular extrusion of strain hardening materials by finite element analysis. Comput
Mater Sci 2004;30:489–95.
[205] Suo T, Li YL, Deng Q, Liu YY. Optimal pressing route for continued equal channel angular pressing by finite element analysis. Mater Sci Eng, A
2007;466:166–71.
[206] Srinivasan R. Computer simulation of the equichannel angular extrusion (ECAE) process. Scr Mater 2001;44:91–6.
[207] Djavanroodi F, Omranpour B, Ebrahimi M, Sedighi M. Designing of ECAP parameters based on strain distribution uniformity. Progr Nat Sci: Mater Int
2012;22:452–60.
[208] Yang F, Saran A, Okazaki K. Finite element simulation of equal channel angular extrusion. J Mater Process Technol 2005;166:71–8.
[209] Toth LS, Lapovok R, Hasani A, Gu CF. Non-equal channel angular pressing of aluminum alloy. Scr Mater 2009;61:1121–4.
[210] Nagasekhar AV, Yoon SC, Tick-Hon Y, Kim HS. An experimental verification of the finite element modelling of equal channel angular pressing. Comput
Mater Sci 2009;46:347–51.
[211] Nagasekhar AV, Tick-Hon Y, Li S, Seow HP. Effect of acute tool-angles on equal channel angular extrusion/pressing. Mater Sci Eng, A
2005;410:269–72.
[212] Nagasekhar AV, Kim HS. Analysis of T-shaped equal channel angular pressing using the finite element method. Met Mater Int 2008;14:565–8.
[213] Nagasekhar AV, Kim HS. Plastic deformation characteristics of cross-equal channel angular pressing. Comput Mater Sci 2008;43:1069–73.
[214] Balasundar I, Raghu T. Effect of friction model in numerical analysis of equal channel angular pressing process. Mater Des 2010;31:449–57.
[215] Suh J-Y, Kim H-S, Park J-W, Chang J-Y. Finite element analysis of material flow in equal channel angular pressing. Scr Mater 2001;44:677–81.
[216] Park J-W, Suh J-Y. Effect of die shape on the deformation behavior in equal-channel angular pressing. Metall Mater Trans A 2001;32:3007–14.
[217] Kim HS. Finite element analysis of equal channel angular pressing using a round corner die. Mater Sci Eng, A 2001;315:122–8.
[218] Kim HS, Hong SI, Seo MH. Effects of strain hardenability and strain-rate sensitivity on the plastic flow and deformation homogeneity during equal
channel angular pressing. J Mater Res 2001;16:856–64.
[219] Kim HS. Finite-element method simulation of severe plastic-deformation methods. bulk nanostructured materials. Wiley-VCH Verlag GmbH & Co.
KGaA; 2009. p. 137–63.
[220] Kim HS. Finite element analysis of deformation behaviour of metals during equal channel multi-angular pressing. Mater Sci Eng, A 2002;328:317–23.
[221] Zuyan L, Gang L, Wang ZR. Finite element simulation of a new deformation type occurring in changing-channel extrusion. J Mater Process Technol
2000;102:30–2.
[222] Mckenzie PWJ, Lapovok R, Estrin Y. The influence of back pressure on ECAP processed. AA 6016: modeling and experiment. Acta Mater
2007;55:2985–93.
[223] Lapovok R. The positive role of back-pressure in equal channel angular extrusion. Nanomater Severe Plastic Deform 2006;503–504:37–44.
[224] Lapovok R. The role of back-pressure in equal channel angular extrusion. J Mater Sci 2005;40:341–6.
[225] Ouyang JH, Sasaki S. Unlubricated friction and wear behavior of low-pressure plasma-sprayed ZrO2 coating at elevated temperatures. Ceram Int
2001;27:251–60.
[226] Kang F, Wang JT, Su YL, Xia KN. Finite element analysis of the effect of back pressure during equal channel angular pressing. J Mater Sci
2007;42:1491–500.
[227] Lee DN. An upper-bound solution of channel angular deformation. Scr Mater 2000;43:115–8.
[228] Chung SW, Somekawa H, Kinoshita T, Kim WJ, Higashi K. The non-uniform behavior during ECAE process by 3-D FVM simulation. Scr Mater
2004;50:1079–83.
[229] Edalati K, Yamamoto A, Horita Z, Ishihara T. High-pressure torsion of pure magnesium: Evolution of mechanical properties, microstructures and
hydrogen storage capacity with equivalent strain. Scr Mater 2011;64:880–3.
[230] Kim WJ, Namgung JC, Kim JK. Analysis of strain uniformity during multi-pressing in equal channel angular extrusion. Scr Mater 2005;53:293–8.
[231] Leo P, Cerri E, De Marco PP, Roven HJ. Properties and deformation behaviour of severe plastic deformed aluminium alloys. J Mater Process Technol
2007;182:207–14.
[232] Alexandrov IV, Budilov IN, Krallics G, Kim HS, Yoon SC, Smolyakov AA, et al. Simulation of equal-channel angular extrusion pressing. Nanomater
Severe Plastic Deform 2006;503–504:201–8.
[233] Semiatin SL, Jones JJ. Formability and workability of metals: plastic instability and flow localization. Metals Park, Ohio: American Society for Metals;
1984.
[234] Semiatin SL, Staker MR, Jonas JJ. Plastic instability and flow localization in shear at high rates of deformation. Acta Metall 1984;32:1347–54.
[235] Lapovok R, Toth LS, Molinari A, Estrin Y. Strain localisation patterns under equal-channel angular pressing. J Mech Phys Solids 2009;57:122–36.
[236] Estrin Y, Mühlhaus HB. From micro- to macroscale, gradient models of plasticity. In: Yuen D, editor. Proceedings of the international conference on
engineering mathematics. IEAust. Australia: AEMC Sydney; 1996. p. 161–5.
[237] Mahallawy NE, Shehata FA, Hameed MAE, Aal MIAE, Kim HS. 3D FEM simulations for the homogeneity of plastic deformation in Al-Cu alloys during
ECAP. Mater Sci Eng, A 2010;527:1404–10.
[238] Figueiredo RB, Pinheiro IP, Aguilar MTP, Modenesi PJ, Cetlin PR. The finite element analysis of equal channel angular pressing (ECAP) considering the
strain path dependence of the work hardening of metals. J Mater Process Technol 2006;180:30–6.
[239] Kim HS, Seo MH, Hong SI. On the die corner gap formation in equal channel angular pressing. Mater Sci Eng, A 2000;291:86–90.
[240] Semiatin SL, DeLo DP, Segal VM, Goforth RE, Frey ND. Workability of commercial-purity titanium and 4340 steel during equal channel angular
extrusion at cold-working temperatures. Metall Mater Trans A 1999;30:1425–35.
[241] Semiatin SL, DeLo DP. Equal channel angular extrusion of difficult-to-work alloys. Mater Des 2000;21:311–22.
[242] Horn T, Silbermann C, Frint P, Wagner M, Ihlemann J. Strain localization during equal-channel angular pressing analyzed by finite element
simulations. Metals 2018;8:55.
[243] Valiev R. Nanostructuring of metals by severe plastic deformation for advanced properties. Nat Mater 2004;3:511–6.
[244] Furukawa M, Iwahashi Y, Horita Z, Nemoto M, Langdon TG. The shearing characteristics associated with equal-channel angular pressing. Mater Sci
Eng, A 1998;257:328–32.
[245] Azushima A, Kopp R, Korhonen A, Yang DY, Micari F, Lahoti GD, et al. Severe plastic deformation (SPD) processes for metals. CIRP Ann Manuf Technol
2008;57:716–35.
[246] Teodosiu C, Hu Z. Evolution of the intragranular microstructure at moderate and large strains: modelling and computational significance. In: Shen S-F,
Dawson P, editors. Simulation of materials processing: theory, methods and applications. Proceedings of the 5th international conference
NUMIFORM’95, Ithaca, New York, 18–21 June 1995. Rotterdam: A.A.; 1995. p. 173–82.
[247] Peeters B, Seefeldt M, Teodosiu C, Kalidindi SR, Van Houtte P, Aernoudt E. Work-hardening/softening behaviour of b.c.c. polycrystals during changing
strain paths: I. An integrated model based on substructure and texture evolution, and its prediction of the stress–strain behaviour of an IF steel during
two-stage strain paths. Acta Mater 2001;49:1607–19.
[248] Li S, Hoferlin E, Bael AV, Houtte PV, Teodosiu C. Finite element modeling of plastic anisotropy induced by texture and strain-path change. Int J Plast
2003;19:647–74.
[249] Wen W, Borodachenkova M, Tomé CN, Vincze G, Rauch EF, Barlat F, et al. Mechanical behavior of low carbon steel subjected to strain path changes:
experiments and modeling. Acta Mater 2016;111:305–14.
[250] Figueiredo RB, Aguilar MTP, Cetlin PR. Finite element modelling of plastic instability during ECAP processing of flow-softening materials. Mater Sci
Eng, A 2006;430:179–84.
[251] Lee SC, Ha SY, Kim KT, Hwang SM, Huh LM, Chung HS. Finite element analysis for deformation behavior of an aluminum alloy composite containing
SiC particles and porosities during ECAP. Mater Sci Eng, A 2004;371:306–12.
[252] Lapovok R. Damage evolution under severe plastic deformation. Int J Fract 2002;115:159–72.
[253] Li SY, Beyerlein IJ, Necker CT, Alexander DJ, Bourke M. Heterogeneity of deformation texture in equal channel angular extrusion of copper. Acta Mater
2004;52:4859–75.
[254] Toth LS, Molinari A, Estrin Y. Strain hardening at large strains as predicted by dislocation based polycrystal plasticity model. J Eng Mater Technol
2002;124:71–7.
[255] Estrin Y, Tóth LS, Molinari A, Bréchet Y. A dislocation-based model for all hardening stages in large strain deformation. Acta Mater 1998;46:5509–22.
[256] Baik SC, Estrin Y, Kim HS, Hellmig RJ. Dislocation density-based modeling of deformation behavior of aluminium under equal channel angular
pressing. Mater Sci Eng, A 2003;351:86–97.
[257] Baik SC, Estrin Y, Hellmig RJ, Jeong HT, Brokmeier HG, Kim HS. Modeling of texture evolution in copper under equal channel angular pressing.
Zeitschrift Fur Metallkunde 2003;94:1189–98.
[258] Baik SC, Estrin Y, Kim HS, Jeong HT, Hellmig RJ. Calculation of deformation behavior and texture evolution during equal channel angular pressing of IF
steel using dislocation based modeling of strain hardening. Textures Mater 2002;408–4(Pts 1 and 2):697–702.
[259] Bratov V, Borodin EN. Comparison of dislocation density based approaches for prediction of defect structure evolution in aluminium and copper
processed by ECAP. Mater Sci Eng, A 2015;631:10–7.
[260] Nagasekhar AV, Tick-Hon Y, Li S, Seow HP. Stress and strain histories in equal channel angular extrusion/pressing. Mater Sci Eng, A 2006;423:143–7.
[261] Lee DJ, Yoon EY, Lee SH, Kang SY, Kim HS. Finite element analysis for compression behavior of high pressure torsion processing. Rev Adv Mater Sci
2012;31:25–30.
[262] Lee DJ, Yoon EY, Ahn D-H, Park BH, Park HW, Park LJ, et al. Dislocation density-based finite element analysis of large strain deformation behavior of
copper under high-pressure torsion. Acta Mater 2014;76:281–93.
[263] Chung SW, Kim WJ, Kohzu M, Higashi K. The effect of ram speed on mechanical and thermal properties in ECAE process simulation. Mater Trans
2003;44:973–80.
[264] Belytschko T, Krongauz Y, Organ D, Fleming M, Krysl P. Meshless methods: an overview and recent developments. Comput Methods Appl Mech Eng
1996;139:3–47.
[265] Lu P, Zhao G, Guan Y, Wu X. Bulk metal forming process simulation based on rigid-plastic/viscoplastic element free Galerkin method. Mater Sci Eng, A
2008;479:197–212.
[266] Yanjin G, Guoqun Z, Ping L. Numerical study of equal-channel angular pressing based on the element-free Galerkin method. Int J Mater Res
2012;103:1361–75.
[267] Monaghan JJ. An introduction to SPH. Comput Phys Commun 1988;48:89–96.
[268] Cleary PW, Prakash M, Ha J, Stokes N, Scott C. Smooth particle hydrodynamics: status and future potential. Prog Comput Fluid Dyn 2007;7:70–90.
[269] Tan H, Nairn JA. Hierarchical, adaptive, material point method for dynamic energy release rate calculations. Comput Methods Appl Mech Eng
2002;191:2123–37.
[270] Fagan T, Das R, Lemiale V, Estrin Y. Modelling of equal channel angular pressing using a mesh-free method. J Mater Sci 2012;47:4514–9.
[271] Lemiale V, Nairn J, Hurmane A. Material point method simulation of equal channel angular pressing involving large plastic strain and contact through
sharp corners. CMES Comput Model Eng Sci 2010;70:41–66.
[272] Kazeminezhad M, Hosseini E. Modeling of induced empirical constitutive relations on materials with FCC, BCC, and HCP crystalline structures: Severe
plastic deformation. Int J Adv Manuf Technol 2010;47:1033–9.
[273] Medeiros N, Lins JFC, Moreira LP, Gouvea JP. The role of the friction during the equal channel angular pressing of an IF-steel billet. Mater Sci Eng A-
Struct 2008;489:363–72.
[274] Marchand A, Duffy J. An experimental study of the formation process of adiabatic shear bands in a structural steel. J Mech Phys Solids
1988;36:251–83.
[275] Bever MB, Holt DL, Titchener AL. The stored energy of cold work. Prog Mater Sci 1973;17:5–177.
[276] Mechanically Stainier L, Heat Generated. In: Hetnarski R, editor. Encyclopedia of thermal stresses. Netherlands: Springer; 2014. p. 2958–64.
[277] Farren WS, Taylor GI. The heat developed during plastic extension of metals. Proc R Soc Lond Ser A 1925;107:422–51.
[278] Taylor GI, Quinney H. The Latent Energy Remaining in a Metal after Cold Working. Proc R Soc Lond Series A, Math Phys Charact (1905–1934)
1934;143:307–26.
[279] Benzerga AA, Bréchet Y, Needleman A, Van der Giessen E. The stored energy of cold work: PREDICTIONS from discrete dislocation plasticity. Acta
Mater 2005;53:4765–79.
[280] Poliak EI, Jonas JJ. A one-parameter approach to determining the critical conditions for the initiation of dynamic recrystallization. Acta Mater
1996;44:127–36.
[281] Hansen N, Vandermeer RA. Recovery, recrystallization, and grain growth. In: Wyder FBLL, editor. Encyclopedia of condensed matter
physics. Oxford: Elsevier; 2005. p. 116–24.
[282] Mishin OV, Godfrey A, Juul Jensen D, Hansen N. Recovery and recrystallization in commercial purity aluminum cold rolled to an ultrahigh strain. Acta
Mater 2013;61:5354–64.
[283] Yamasaki T, Miyamoto H, Mimaki T, Vinogradov A, Hashimoto S. Corrosion fatigue of ultra-fine grain copper fabricated by severe plastic deformation.
ultrafine grain metals II TMS:USA; 2002. p. 361–70.
[284] Estrin Y, Kubin LP. Plastic instabilities: phenomenology and theory. Mater Sci Eng, A 1991;137:125–34.
[285] Kubin LP, Spiesser P, Estrin Y. Computer simulation of the low temperature instability of plastic flow. Acta Metall 1982;30:385–94.
[286] Rogers HC. Adiabatic plastic deformation. Annu Rev Mater Sci 1979;9:283–311.
[287] Johnson GR, Cook WH. Fracture characteristics of three metals subjected to various strains, strain rates, temperatures and pressures. Eng Fract Mech
1985;21:31–48.
[288] Zerilli FJ, Armstrong RW. Dislocation-mechanics-based constitutive relations for material dynamics calculations. J Appl Phys 1987;61:1816–25.
[289] Vinogradov A, Kaneko Y, Kitagawa K, Hashimoto S, Stolyarov V, Valiev R. Cyclic response of ultrafine-grained copper at constant plastic strain
amplitude. Scr Mater 1997;36:1345–51.
[290] Vinogradov A, Hashimoto S. Multiscale phenomena in fatigue of ultra-fine grain materials - an overview. Mater Trans, JIM 2001;42:74–84.
[291] Mason JJ, Rosakis AJ, Ravichandran G. On the strain and strain rate dependence of the fraction of plastic work converted to heat: an experimental
study using high speed infrared detectors and the Kolsky bar. Mech Mater 1994;17:135–45.
[292] Hodowany J, Ravichandran G, Rosakis AJ, Rosakis P. Partition of plastic work into heat and stored energy in metals. Exp Mech 2000;40:113–23.
[293] Oliferuk W, Maj M. Stress–strain curve and stored energy during uniaxial deformation of polycrystals. Eur J Mech A Solids 2009;28:266–72.
[294] Kostina A, Iziumova A, Plekhov O. Energy dissipation and storage in iron under plastic deformation (experimental study and numerical simulation).
Frattura ed Integrità Strutturale 2014;27:28–37.
[295] Rosakis P, Rosakis AJ, Ravichandran G, Hodowany J. A thermodynamic internal variable model for the partition of plastic work into heat and stored
energy in metals. J Mech Phys Solids 2000;48:581–607.
[296] Estrin Y, Kubin LP. Criterion for thermomechanical instability of low temperature plastic deformation. Scr Metall 1980;14:1359–64.
[297] Tsagrakis I, Aifantis EC. On the effect of strain gradient on adiabatic shear banding. Metall Mater Trans A 2014.
[298] Korn M, Lapovok R, Böhner A, Höppel HW, Mughrabi H. Bimodal grain size distributions in UFG materials produced by SPD - Their evolution and
effect on the fatigue and monotonic strength properties. Kovove Mater 2011;49:51–63.
[299] Rybin VV. Bolshie Plasticheskie Deformacii i Razrushenie Metallov. Moscow, Russia: Metallurgiya; 1986.
[300] Huang X, Winther G. Dislocation structures. Part I. Grain orientation dependence. Phil Mag 2007;87:5189–214.
[301] Winther G, Huang X. Dislocation structures. Part II. Slip system dependence. Phil Mag 2007;87:5215–35.
[302] Hughes DA, Hansen N, Bammann DJ. Geometrically necessary boundaries, incidental dislocation boundaries and geometrically necessary
dislocations. Scr Mater 2003;48:147–53.
[303] Hansen N, Barlow CY. 17 - Plastic deformation of metals and alloys. In: Hono DEL, editor. Physical metallurgy. Oxford: Elsevier; 2014. p. 1681–764.
[304] Kozlov EV, Zhdanov AN, Popova NA, Pekarskaya EE, Koneva NA. Subgrain structure and internal stress fields in UFG materials: problem of Hall-Petch
relation. Mater Sci Eng A-Struct Mater Properties Microstruct Process 2004;387–89:789–94.
[305] Koneva NA, Kozlov EV. Deformation-induced ordering of dislocation structures. Mater Sci Eng A-Struct Mater Properties Microstruct Process
2004;387–89:64–6.
[306] Pantleon W. The evolution of disorientations for several types of boundaries. Mater Sci Eng, A 2001;319:211–5.
[307] Pantleon W. Disorientations in dislocation structures. Mater Sci Eng, A 2005;400:118–24.
[308] Zhilyaev AP, Swaminathan S, Pshenichnyuk AI, Langdon TG, McNelley TR. Adiabatic heating and the saturation of grain refinement during SPD of
metals and alloys: experimental assessment and computer modeling. J Mater Sci 2013;48:4626–36.
[309] Zhilyaev AP, Garcia-Infanta JM, Carreno F, Langdon TG, Ruano OA. Particle and grain growth in an Al-Si alloy during high-pressure torsion. Scr Mater
2007;57:763–5.
[310] Shaw MC. Metal cutting principles. 2nd ed. New York; Oxford: Oxford University Press; 2005.
[311] Kim HS. Prediction of temperature rise in equal channel angular pressing. Mater Trans 2001;42:536–8.
[312] Kuhlmann-Wilsdorf D, Duesbery FRNNaMS. Chapter 59 The LES theory of solid plasticity. Dislocations in Solids. Elsevier; 2002. p. 211–342.
[313] Kuhlmann-Wilsdorf D. The theory of dislocation-based crystal plasticity. Philos Mag A 1999;79:955–1008.
[314] Sauzay M, Kubin LP. Scaling laws for dislocation microstructures in monotonic and cyclic deformation of fcc metals. Prog Mater Sci 2011;56:725–84.
[315] Pougis A, Tóth LS, Bouaziz O, Fundenberger JJ, Barbier D, Arruffat R. Stress and strain gradients in high-pressure tube twisting. Scr Mater
2012;66:773–6.
[316] Li JCM, Chou YT. The role of dislocations in the flow stress grain size relationships. Metall Mater Trans B 1970;1:1145.
[317] Thompson AW, Baskes MI, Flanagan WF. The dependence of polycrystal work hardening on grain size. Acta Metall 1973;21:1017–28.
[318] Ashby MF. The deformation of plastically non-homogeneous materials. Phil Mag 1970;21:399–424.
[319] Li JCM. Petch relation and grain boundary sources. Trans Metall Soc AIME 1963;277:239–47.
[320] Kato M, Fujii T, Onaka S. Dislocation bow-out model for yield stress of ultra-fine grained materials. Mater Trans 2008;49:1278–83.
[321] Gryaznov VG, Gutkin MY, Romanov AE, Trusov LI. On the yield stress of nanocrystals. J Mater Sci 1993;28:4359–65.
[322] Armstrong RW, Rodriguez P. Flow stress/strain rate/grain size coupling for fcc nanopolycrystals. Phil Mag 2006;86:5787–96.
[323] Kopylov VI, Chyvil’deev VN. The limit of grain refinement on equal cannel angular deformation. Metally 2004;22–35.
[324] Gottstein G. Integral materials modeling: towards physics-based through-process models. Weinheim: Wiley-VCH; 2007. Chichester: John Wiley
[distributor].
[325] Cottrell A. Dislocations and plastic flow in crystals. Oxford: Clarendon Press; 1953.
[326] Zehetbauer MJ, Estrin Y. Modeling of strength and strain hardening of bulk nanostructured materials. Bulk Nanostructured Materials. Wiley-VCH
Verlag GmbH & Co. KGaA; 2009. p. 109–36.
[327] Mecking H, Kocks UF. Kinetics of flow and strain-hardening. Acta Metall 1981;29:1865–75.
[328] Mandelbrot BB. The fractal geometry of nature. Updated and augm. New York: W.H. Freeman; 1983.
[329] Onsager L. Reciprocal Relations in Irreversible Processes. I. Phys Rev 1931;37:405–26.
[330] Onsager L. Reciprocal relations in irreversible processes. II. Phys Rev 1931;38:2265–79.
[331] Prigogine I. Introduction to thermodynamics of irreversible processes. 2nd rev. ed. New York: Interscience Publishers; 1961.
[332] Bridgman PW. The thermodynamics of plastic deformation and generalized entropy. Rev Mod Phys 1950;22:56.
[333] Kubin LP, Estrin Y, Canova G. Dislocation patterns and plastic instabilities. In: Walgraef D, Ghoniem NM, editors. Patterns, defects and materials
instabilities. Netherlands: Springer; 1990. p. 277–301.
[334] Aifantis EC. Deformation and failure of bulk nanograined and ultrafine-grained materials. Mater Sci Eng, A 2009;503:190–7.
[335] Aifantis EC. The physics of plastic deformation. Int J Plast 1987;3:211–47.
[336] Ziegler H. An introduction to thermomechanics. 2d rev. ed. Amsterdam; New York: North-Holland Pub. Co.; 1983.
[337] Ziegler H. Eidgenössische Technische Hochschule. Some extremum principles in irreversible thermodynamics, with application to continuum
mechanics. Zürich: Swiss Federal Institute of Technology; 1962.
[338] Drucker DC. Some implications of work hardening and ideal plasticity. Q Appl Math 1950;7:411–8.
[339] Hill R. The mathematical theory of plasticity. Oxford: Clarendon Press; 1950.
[340] Rabotnov IUN, Novozhilov VV. Mechanics of deformable solids and structures. Moskva: Mashinostroenie; 1975.
[341] Ghoniem NM, Tong SH, Sun LZ. Parametric dislocation dynamics: a thermodynamics-based approach to investigations of mesoscopic plastic
deformation. Phys Rev B 2000;61:913.
[342] Langer JS, Bouchbinder E, Lookman T. Thermodynamic theory of dislocation-mediated plasticity. Acta Mater 2010;58:3718–32.
[343] Langer JS. Statistical thermodynamics of strain hardening in polycrystalline solids. Phys Rev E 2015;92:032125.
[344] Vinogradov A, Yasnikov IS, Estrin Y. Evolution of fractal structures in dislocation ensembles during plastic deformation. Phys Rev Lett
2012;108:205504.
[345] Huang M, Rivera-Díaz-Del-castillo PEJ, Bouaziz O, Zwaag SVD. Predicting the strength and ductility of ultrafine grained interstitial free steels using
irreversible thermodynamics. Buenos Aires 2008: 805–12.
[346] Huang M, Rivera-Díaz-del-Castillo PEJ, Bouaziz O, van der Zwaag S. Irreversible thermodynamics modelling of plastic deformation of metals. Mater
Sci Technol 2008;24:495–500.
[347] Mughrabi H. The a-factor in the Taylor flow-stress law in monotonic, cyclic and quasi-stationary deformations: Dependence on slip mode, dislocation
arrangement and density. Curr Opin Solid State Mater Sci 2016;20:411–20.
[348] Zaiser M. Scale invariance in plastic flow of crystalline solids. Adv Phys 2006;55:185–245.
[349] Mughrabi H. Implications of linear relationships between local and macroscopic flow stresses in the composite model. J Mater Res 2006;97:594–6.
[350] Mughrabi H. Dislocation wall and cell structures and long-range internal-stresses in deformed metal crystals. Acta Metall 1983;31:1367–79.
[351] Holt DL. Dislocation cell formation in metals. J Appl Phys 1970;41:3197–201.
[352] Kocks UF, Argon AS, Ashby MF. Thermodynamics and kinetics of slip. Prog Mater Sci 1975;19:1–281.
[353] Hansen N, Kuhlmann-Wilsdorf D. Low energy dislocation structures due to unidirectional deformation at low temperatures. Mater Sci Eng
1986;81:141–61.
[354] Marthinsen K, Nes E. A general model for metal plasticity. Mater Sci Eng, A 1997;234–236:1095–8.
[355] Nes E. Modelling of work hardening and stress saturation in FCC metals. Prog Mater Sci 1997;41:129–93.
[356] Estrin Y. Unified constitutive laws of plastic deformation. San Diego; London: Academic Press; 1996.
[357] Alexandrov IV, Chembarisova RG. Development and application of the dislocation model for analysis of the microstructure evolution and deformation
behavior of metals subjected to severe plastic deformation. Rev Adv Mater Sci 2007;16:51–72.
[358] Caillard D, Martin J-L. Thermally activated mechanisms in crystal plasticity. Amsterdam: Pergamon; 2003.
[359] Klepaczko J. Thermally activated flow and strain rate history effects for some polycrystalline f.c.c. metals. Mater Sci Eng 1975;18:121–35.
[360] Malygin GA. Dislocation density evolution equation and strain-hardening of fcc crystals. Phys Status Solidi A-Appl Res 1990;119:423–36.
[361] Bergström Y. Rev Powder Metall Phys Ceram 1973;2:79.
[362] Gottstein G, Argon AS. Dislocation theory of steady state deformation and its approach in creep and dynamic tests. Acta Metall 1987;35:1261–71.
[363] Starink MJ, Qiao XG, Zhang J, Gao N. Predicting grain refinement by cold severe plastic deformation in alloys using volume averaged dislocation
generation. Acta Mater 2009;57:5796–811.
[364] Holmedal B. On the basic relation between mean free slip length and work hardening of metals. Phil Mag 2015;95:2817–30.
[365] Hahner P, Bay K, Zaiser M. Fractal dislocation patterning during plastic deformation. Phys Rev Lett 1998;81:2470.
[366] Yasnikov IS, Vinogradov A, Estrin Y. Revisiting the Considère criterion from the viewpoint of dislocation theory fundamentals. Scr Mater
2014;76:37–40.
[367] Yasnikov IS, Estrin Y, Vinogradov A. What governs ductility of ultrafine-grained metals? A microstructure based approach to necking instability. Acta
Mater 2017;141:18–28.
[368] Torre D, Pereloma EV, Davies CHJ. Strain rate sensitivity and apparent activation volume measurements on equal channel angular extruded Cu
processed by one to twelve passes. Scr Mater 2004;51:367–71.
[369] Dao M, Lu L, Shen YF, Suresh S. Strength, strain-rate sensitivity and ductility of copper with nanoscale twins. Acta Mater 2006;54:5421–32.
[370] Vinogradov A, Yasnikov IS, Matsuyama H, Uchida M, Kaneko Y, Estrin Y. Controlling strength and ductility: dislocation-based model of necking
instability and its verification for ultrafine grain 316L steel. Acta Mater 2016;106:295–303.
[371] Vinogradov A. Mechanical properties of ultrafine-grained metals: new challenges and perspectives. Adv Eng Mater 2015;17:1710–22.
[372] Essmann U, Mughrabi H. Annihilation of dislocations during tensile and cyclic deformation and limits of dislocation densities. Philos Mag A
1979;40:731–56.
[373] Blum W. Role of dislocation annihilation during steady-state deformation. Phys Status Solidi (b) 1971;45:561–71.
[374] Rivera-Díaz-del-Castillo PEJ, Huang M. Dislocation annihilation in plastic deformation: I. Multiscale irreversible thermodynamics. Acta Mater
2012;60:2606–14.
[375] Galindo-Nava EI, Rivera-Díaz-del-Castillo PEJ. Modelling plastic deformation in BCC metals: dynamic recovery and cell formation effects. Mater Sci
Eng, A 2012;558:641–8.
[376] Galindo-Nava EI, Rivera-Díaz-del-Castillo PEJ. Thermostatistical modelling of hot deformation in FCC metals. Int J Plast 2013;47:202–21.
[377] Galindo-Nava EI, Sietsma J, Rivera-Díaz-del-Castillo PEJ. Dislocation annihilation in plastic deformation: II. Kocks-Mecking analysis. Acta Mater
2012;60:2615–24.
[378] Huang M, Rivera-Díaz-del-Castillo PEJ, Bouaziz O, van der Zwaag S. Modelling plastic deformation of metals over a wide range of strain rates using
irreversible thermodynamics. IOP Conf Series: Mater Sci Eng 2009;3:012006.
[379] Bonneville J, Escaig B. Cross-slipping process and the stress-orientation dependence in pure copper. Acta Metall 1979;27:1477–86.
[380] Vinogradov AY, Stolyarov VV, Hashimoto S, Valiev RZ. Cyclic behavior of ultrafine-grain titanium produced by severe plastic deformation. Mater Sci
Eng, A 2001;318:163–73.
[381] Patlan V, Higashi K, Kitagawa K, Vinogradov A, Kawazoe M. Cyclic response of fine grain 5056 Al-Mg alloy processed by equal-channel angular
[382] Patlan V, Vinogradov A, Higashi K, Kitagawa K. Overview of fatigue properties of fine grain 5056 Al-Mg alloy processed by equal-channel angular
[383] Vinogradov A, Yasnikov IS, Estrin Y. Irreversible thermodynamics approach to plasticity: dislocation density based constitutive modelling. Mater Sci
Technol 2015;31:1664–72.
[384] Molotnikov A. Application of strain gradient plasticity modelling to high pressure torsion. Mater Sci Forum 2008;584–586:1051–6.
[385] Kumar KS, Van Swygenhoven H, Suresh S. Mechanical behavior of nanocrystalline metals and alloys. Acta Mater 2003;51:5743–74.
[386] Wolf D, Yamakov V, Phillpot SR, Mukherjee A, Gleiter H. Deformation of nanocrystalline materials by molecular-dynamics simulation: relationship to
experiments? Acta Mater 2005;53:1–40.
[387] Van Swygenhoven H, Derlet PM, Frøseth AG. Nucleation and propagation of dislocations in nanocrystalline fcc metals. Acta Mater 2006;54:1975–83.
[388] Chang L, Zhou C-Y, Li J, He X-H. Investigation on tensile properties of nanocrystalline titanium with ultra-small grain size. Comput Mater Sci
2018;142:135–44.
[389] Höppel HW, Zhou ZM, Mughrabi H, Valiev RZ. Microstructural study of the parameters governing coarsening and cyclic softening in fatigued
ultrafine-grained copper. Philos Mag A 2002;82:1781–94.
[390] Höppel HW, Mughrabi H, Vinogradov A. Fatigue properties of bulk nanostructured materials. Wiley-VCH Verlag GmbH & Co. KGaA; 2009.
[391] Miyamoto H, Fushimi J, Mimaki T, Vinogradov A, Hashimoto S. Dislocation structures and crystal orientations of copper single crystals deformed by
equal-channel angular pressing. Mater Sci Eng, A 2005;405:221–32.
[392] Miyamoto H, Fushimi J, Mimaki T, Vinogradov A, Hashimoto S. The effect of the initial orientation on microstructure development of copper single
crystals subjected to equal-channel angular pressing. In: Horita Z, editor. Nanomater Severe Plastic Deform. p. 799–804.
[393] Rybin VV. Large plastic deformations and fracture of metals. Moscow: Metallurgiya; 1986.
[394] Rybin VV. Physical model of the phenomenon of loss of mechanical stability and necking. hys Met Metallogr 1977;44:149.
[395] Klemm R. Zyklische Plastizität von mikro- und submikrokristallinem Nickel. Technische Universität Dresden; 2004.
[396] Mughrabi H, Höppel HW. Cyclic deformation and fatigue properties of very fine-grained metals and alloys. Int J Fatigue 2010;32:1413–27.
[397] Polyakova VV, Semenova IP, Valiev RZ, Vinogradov A. [unpublished work].
[398] Estrin Y, Braasch H, Brechet Y. A dislocation density based constitutive model for cyclic deformation. J Eng Mater Tech ASME 1996;118:441–57.
[399] Bouaziz O, Estrin Y, Bréchet Y, Embury JD. Critical grain size for dislocation storage and consequences for strain hardening of nanocrystalline
materials. Scr Mater 2010;63:477–9.
[400] Schafler E, Pippan R. Effect of thermal treatment on microstructure in high pressure torsion (HPT) deformed nickel. Mater Sci Eng, A 2004;387–
89:799–804.
[401] Hebesberger T, Stuwe HP, Vorhauer A, Wetscher F, Pippan R. Structure of Cu deformed by high pressure torsion. Acta Mater 2005;53:393–402.
[402] Pippan R, Wetscher F, Hafok M, Vorhauer A, Sabirov I. The limits of refinement by severe plastic deformation. Adv Eng Mater 2006;8:1046–56.
[403] Reichert B, Estrin Y. Steel Res Intl 2007;78:791–7.
[404] Baufeld B, Petukhov BV, Bartsch M, Messerschmidt U. Transition of mechanisms controlling the dislocation motion in cubic ZrO2 below 700 C. Acta
Mater 1998;46:3077–85.
[405] Lee S, Estrin Y, De Cooman BC. Effect of the strain rate on the TRIP-TWIP transition in austenitic Fe-12 pct Mn-0.6 pct C TWIP Steel. Metall Mater Trans
A 2014;45:717–30.
[406] Petukhov BV. Phenomenological description of material plasticity in the region of transition from the kink mechanism of dislocation motion to the
mechanism of their local pinning. Crystallogr Rep 1996;41:181–8.
[407] Dour G, Estrin Y. Dislocation motion in crystals with a high peierls relief: a unified model incorporating the lattice friction and localized obstacles. J
Eng Mater Technol 2002;124:7–12.
[408] Cheng GM, Jian WW, Xu WZ, Yuan H, Millett PC, Zhu YT. Grain size effect on deformation mechanisms of nanocrystalline bcc metals. Mater Res Lett
2013;1:26–31.
[409] Dunlop JW, Bréchet YJM, Legras L, Estrin Y. Dislocation density-based modelling of plastic deformation of Zircaloy-4. Mater Sci Eng, A
2007;443:77–86.
[410] Ahn D-H, Kim HS, Estrin Y. A semi-phenomenological constitutive model for hcp materials as exemplified by alpha titanium. Scr Mater
2012;67:121–4.
[411] Bouaziz O, Guelton N. Modelling of TWIP effect on work-hardening. Mater Sci Eng, A 2001;319–321:246–9.
[412] Allain S, Chateau JP, Bouaziz O. A physical model of the twinning-induced plasticity effect in a high manganese austenitic steel. Mater Sci Eng, A
2004;387–389:143–7.
[413] Bouaziz O, Allain S, Scott C. Effect of grain and twin boundaries on the hardening mechanisms of twinning-induced plasticity steels. Scr Mater
2008;58:484–7.
[414] Gil Sevillano J. Geometrically necessary twins and their associated size effects. Scr Mater 2008;59:135–8.
[415] Sevillano JS. An alternative model for the strain hardening of FCC alloys that twin, validated for twinning-induced plasticity steel. Scr Mater
2009;60:336–9.
[416] Bouaziz O, Allain S, Scott CP, Cugy P, Barbier D. High manganese austenitic twinning induced plasticity steels: a review of the microstructure
properties relationships. Curr Opin Solid State Mater Sci 2011;15:141–68.
[417] Armstrong RW, Worthington PJ. A constitutive relation for deformation twinning in body centered cubic metals. In: Rohde RW, Butcher BM, Holland
JR, Karnes CH, editors. Metallurgical effects at high strain rates. Boston (MA): Springer US; 1973. p. 401–14.
[418] Meyers MA, Vöhringer O, Lubarda VA. The onset of twinning in metals: a constitutive description. Acta Mater 2001;49:4025–39.
[419] Wu XL, Youssef KM, Koch CC, Mathaudhu SN, Kecskés LJ, Zhu YT. Deformation twinning in a nanocrystalline hcp Mg alloy. Scr Mater 2011;64:213–6.
[420] Beyerlein IJ, Capolungo L, Marshall PE, McCabe RJ, Tomé CN. Statistical analyses of deformation twinning in magnesium. Phil Mag 2010;90:2161–90.
[421] Zhu YT, Liao XZ, Wu XL. Deformation twinning in nanocrystalline materials. Prog Mater Sci 2012;57:1–62.
[422] Ueno H, Kakihata K, Kaneko Y, Hashimoto S, Vinogradov A. Nanostructurization assisted by twinning during equal channel angular pressing of
metastable 316L stainless steel. J Mater Sci 2011;1–8.
[423] Lu L, Chen X, Huang X, Lu K. Revealing the maximum strength in nanotwinned copper. Science 2009;323:607–10.
[424] Liu MP, Roven HJ, Yu YD. Deformation twins in ultrafine grained commercial aluminum. Int J Mater Res 2007;98:184–90.
[425] Wu XL, Ma E. Dislocations and twins in nanocrystalline Ni after severe plastic deformation: the effects of grain size. Mater Sci Eng, A 2008;483:84–6.
[426] Vinogradov A, Vasilev E, Merson D, Estrin Y. A phenomenological model of twinning kinetics. Adv Eng Mater 2017;19. 1600092-n/a.
[427] Fátima Vaz M, Fortes MA. Grain size distribution: the lognormal and the gamma distribution functions. Scr Metall 1988;22:35–40.
[428] Zehetbauer M, Seumer V. Cold work hardening in stages IV and V of F.C.C. metals—I. Experiments and interpretation. Acta Metall Mater
1993;41:577–88.
[429] Kubin LP, Estrin Y. Evolution of dislocation densities and the critical conditions for the Portevin-Le Châtelier effect. Acta Metall Mater
1990;38:697–708.
[430] Prinz FB, Argon AS. The evolution of plastic resistance in large strain plastic flow of single phase subgrain forming metals. Acta Metall
1984;32:1021–8.
[431] Zehetbauer M. Cold work hardening in stages IV and V of F.C.C. metals—II. Model fits and physical results. Acta Metall Mater 1993;41:589–99.
[432] Nes E, Marthinsen K. Modeling the evolution in microstructure and properties during plastic deformation of f.c.c.-metals and alloys – an approach
towards a unified model. Mater Sci Eng, A 2002;322:176–93.
[433] Malygin GA. Kinetic mechanism of the formation of fragmented dislocation structures upon large plastic deformations. Phys Solid State
2002;44:2072–9.
[434] Estrin Y, Kubin LP. Local strain hardening and nonuniformity of plastic deformation. Acta Metall 1986;34:2455–64.
[435] Les P, Zehetbauer Mj. Evolution of microstructural parameters in large strain deformation: description by Zehetbauer’s Model. Key Eng Mater
1994;97–97:335–40.
[436] Zehetbauer MJ, Stüwe HP, Vorhauer A, Schafler E, Kohout J. The role of hydrostatic pressure in severe plastic deformation. nanomaterials by severe
plastic deformation. Wiley-VCH Verlag GmbH & Co. KGaA; 2005. p. 433–46.
[437] Hosseini E, Kazeminezhad M. Integration of physically based models into FE analysis: homogeneity of copper sheets under large plastic deformations.
Comput Mater Sci 2010;48:166–73.
[438] Hosseini E, Kazeminezhad M. A new microstructural model based on dislocation generation and consumption mechanisms through severe plastic
deformation. Comput Mater Sci 2011;50:1123–35.
[439] Argon AS, Haasen P. A new mechanism of work hardening in the late stages of large strain plastic flow in F.C.C. and diamond cubic crystals. Acta
Metall Mater 1993;41:3289–306.
[440] Lemiale V, Estrin Y, Kim HS, O’Donnell R. Grain refinement under high strain rate impact: a numerical approach. Comput Mater Sci 2010;48:124–32.
[441] Martin JW. Precipitation hardening. 2nd ed. Oxford: Butterworth-Heinemann; 1998.
[442] Nembach E. Particle strengthening of metals and alloys. New York etc.: Wiley; 1997.
[443] Estrin Y. Constitutive modelling of creep of metallic materials: Some simple recipes. Mater Sci Eng, A 2007;463:171–6.
[444] Rösler J, Arzt E. A new model-based creep equation for dispersion strengthened materials. Acta Metall Mater 1990;38:671–83.
[445] Molinari A, Canova GR, Ahzi S. A self consistent approach of the large deformation polycrystal viscoplasticity. Acta Metall 1987;35:2983–94.
[446] Mercier S, Molinari A, Berbenni S, Berveiller M. Comparison of different homogenization approaches for elastic–viscoplastic materials. Modell Simul
Mater Sci Eng 2012;20:024004.
[447] Ardeljan M, Beyerlein IJ, Knezevic M. A dislocation density based crystal plasticity finite element model: application to a two-phase polycrystalline
HCP/BCC composites. J Mech Phys Solids 2014;66:16–31.
[448] Estrin Y, Kim H. Modelling microstructure evolution towards ultrafine crystallinity produced by severe plastic deformation. J Mater Sci
2007;42:9092–6.
[449] Alexandrov IV, Chembarisova RG, Sitdikov VD. Analysis of the deformation mechanisms in bulk ultrafine grained metallic materials. Mater Sci Eng, A
2007;463:27–35.
[450] Pantleon W. Formation of Disorientations in Dislocation Structures during Plastic Deformation. Solid State Phenom 2002;87:73–92.
[451] Estrin Y, Tóth L, Bréchet Y, Kim HS. Modelling of the evolution of dislocation cell misorientation under severe plastic deformation. Mater Sci Forum
2006:503–4.
[452] Hughes DA, Hansen N. Microstructure and strength of nickel at large strains. Acta Mater 2000;48:2985–3004.
[453] Maier HJ, Gabor P, Gupta N, Karaman I, Haouaoui M. Cyclic stress-strain response of ultrafine grained copper. Int J Fatigue 2006;28:243–50.
[454] Tóth LS, Beausir B, Gu CF, Estrin Y, Scheerbaum N, Davies CHJ. Effect of grain refinement by severe plastic deformation on the next-neighbor
misorientation distribution. Acta Mater 2010;58:6706–16.
[455] Kratochvíl J, Kružík M, Sedláček R. A model of ultrafine microstructure evolution in materials deformed by high-pressure torsion. Acta Mater
2009;57:739–48.
[456] Seefeldt M, Delannay L, Peeters B, Kalidindi SR, Van Houtte P. A disclination-based model for grain subdivision. Mater Sci Eng, A 2001;319–
321:192–6.
[457] Zhang Y, Tao NR, Lu K. Mechanical properties and rolling behaviors of nano-grained copper with embedded nano-twin bundles. Acta Mater
2008;56:2429–40.
[458] Lu K, Lu L, Suresh S. Strengthening materials by engineering coherent internal boundaries at the nanoscale. Science 2009;324:349–52.
[459] Olivares FH, Gil Sevillano J. A quantitative assessment of forest-hardening in f.c.c. metals. Acta Metall 1987;35:631–41.
[460] Gil Sevillano J, Aernoudt E. Low energy dislocation structures in highly deformed materials. Mater Sci Eng 1987;86:35–51.
[461] Zhang HW, Huang X, Hansen N. Evolution of microstructural parameters and flow stresses toward limits in nickel deformed to ultra-high strains. Acta
Mater 2008;56:5451–65.
[462] Pippan R, Scheriau S, Taylor A, Hafok M, Hohenwarter A, Bachmaier A. Saturation of fragmentation during severe plastic deformation. Annu Rev Mater
Res 2010;40:319–43.
[463] Huang Y. Steady state and a general scale law of deformation. 1st ed. Institute of Physics Publishing; 2017.
[464] Bachmaier A, Pippan R. Generation of metallic nanocomposites by severe plastic deformation. Int Mater Rev 2013;58:41–62.
[465] Tomé C, Canova GR, Kocks UF, Christodoulou N, Jonas JJ. The relation between macroscopic and microscopic strain hardening in F.C.C. polycrystals.
Acta Metall 1984;32:1637–53.
[466] Lee DJ, Kim HS. Finite element analysis for the geometry effect on strain inhomogeneity during high-pressure torsion. J Mater Sci 2014;49:6620–8.
[467] Borodachenkova M, Wen W, Pereira AMdB. High-pressure torsion: experiments and modeling. In: Cabibbo M, editor. Severe plastic deformation
techniques. Rijeka: InTech; 2017. p. Ch. 04.
[468] Yoon SC, Seo MH, Krishnaiah A, Kim HS. Finite element analysis of rotary-die equal channel angular pressing. Mater Sci Eng, A 2008;490:289–92.
[469] Rosochowski A, Rodiet R, Lipinski P. Finite element simulation of cyclic extrusion-compression. In: Pietrzyk JK M, Majta J, editors. Metal forming 2000
Krakow, Poland. Rotterdam, Brookfield: A. A. Balkema; 2000. p. 253–9.
[470] Kim HS. Finite element analysis of high pressure torsion processing. J Mater Process Technol 2001;113:617–21.
[471] Figueiredo RB, Pereira PHR, Aguilar MTP, Cetlin PR, Langdon TG. Using finite element modeling to examine the temperature distribution in quasi-
constrained high-pressure torsion. Acta Mater 2012;60:3190–8.
[472] Figueiredo RB, Cetlin PR, Langdon TG. Using finite element modeling to examine the flow processes in quasi-constrained high-pressure torsion. Mater
Sci Eng, A 2011;528:8198–204.
[473] Estrin Y, Molotnikov A, Davies CHJ, Lapovok R. Strain gradient plasticity modelling of high-pressure torsion. J Mech Phys Solids 2008;56:1186–202.
[474] Zhou W, Lin J, Dean TA, Wang L. Feasibility studies of a novel extrusion process for curved profiles: experimentation and modelling. Int J Mach Tools
Manuf 2018;126:27–43.
[475] Pardis N, Ebrahimi R. Deformation behavior in Simple Shear Extrusion (SSE) as a new severe plastic deformation technique. Mater Sci Eng, A
2009;527:355–60.
[476] Bagherpour E, Ebrahimi R, Qods F. An analytical approach for simple shear extrusion process with a linear die profile. Mater Des 2015;83:368–76.
[477] Sheikh H, Ebrahimi R, Bagherpour E. Crystal plasticity finite element modeling of crystallographic textures in simple shear extrusion (SSE) process.
Mater Des 2016;109:289–99.
[478] Mohebbi MS, Akbarzadeh A. Constitutive equation and FEM analysis of incremental cryo-rolling of UFG AA 1050 and AA 5052. J Mater Process
Technol 2018;255:35–46.
[479] List G, Sutter G, Bi XF, Molinari A, Bouthiche A. Strain, strain rate and velocity fields determination at very high cutting speed. J Mater Process Technol
2013;213:693–9.
[480] Sevier M, Yang HTY, Lee S, Chandrasekar S. Severe plastic deformation by machining characterized by finite element simulation. Metall Mater Trans
B-Process Metall Mater Process Sci 2007;38:927–38.
[481] Sevier M, Yang HTY, Moscoso W, Chandrasekar S. Analysis of severe plastic deformation by large strain extrusion machining. Metall Mater Trans A-
Phys Metall Mater Sci 2008;39A:2645–55.
[482] Lee JW, Park JJ. Numerical and experimental investigations of constrained groove pressing and rolling for grain refinement. J Mater Process Technol
2002;130–131:208–13.
[483] Rosochowski A, Olejnik L. Incremental equal channel angular pressing for grain refinement. Mater Sci Forum 2011;674:19–28.
[484] Beygelzimer Y, Kulagin R, Estrin Y, Tóth LS, Kim HS, Latypov MI. Twist extrusion as a potent tool for obtaining advanced engineering materials: a
review. Adv Eng Mater 2017;19.
[485] Khoddam S, Farhoumand A, Hodgson PD. Upper-bound analysis of axi-symmetric forward spiral extrusion. Mech Mater 2011;43:684–92.
[486] Salehi Seyed M, Serajzadeh S. A new upper bound solution for the analysis of the twist extrusion process with an elliptical die cross-section. Proc Inst
Mech Eng Part C J Mech Eng Sci 2009;223:1975–81.
[487] Lapovok R, Pougis A, Lemiale V, Orlov D, Tóth L, Estrin Y. Severe plastic deformation processes for thin samples. J Mater Sci 2010;45:4554–60.
[488] Lapovok R, Estrin Y, Djugum R, Lerk A. Severe plastic deformation processes with friction induced shear; 2011. p. 25–30.
[489] Lapovok R, Qi Y, Ng HP, Tóth LS, Estrin Y. Gradient structures in thin-walled metallic tubes produced by continuous high pressure tube shearing
process. Adv Eng Mater 2017;1700345-n/a.
[490] Faraji G, Babaei A, Mashhadi MM, Abrinia K. Parallel tubular channel angular pressing (PTCAP) as a new severe plastic deformation method for
cylindrical tubes. Mater Lett 2012;77:82–5.
[491] Zangiabadi A, Kazeminezhad M. Development of a novel severe plastic deformation method for tubular materials: Tube Channel Pressing (TCP).
Mater Sci Eng, A 2011;528:5066–72.
[492] Farshidi MH, Kazeminezhad M. The effects of die geometry in tube channel pressing: Severe plastic deformation. Proc Inst Mech Eng Part L J Mat Des
Appl 2016;230:263–72.
[493] Babaei A, Jafarzadeh H, Esmaeili F. Tube Twist Pressing (TTP) as a new severe plastic deformation method. Trans Indian Inst Met 2017.
[494] Bouaziz O, Estrin Y, Kim HS. A new technique for severe plastic deformation: the cone-cone method. Adv Eng Mater 2009;11:982–5.
[495] Joo S-H, Kim H. Comparison of deformation and microstructural evolution between equal channel angular pressing and forward extrusion using the
dislocation cell mechanism-based finite element method. J Mater Sci 2010;45:4705–10.
[496] Baik SC, Hellmig RJ, Estrin Y, Kim HS. Modeling of deformation behavior of copper under equal channel angular pressing. Zeitschrift Fur Metallkunde
2003;94:754–60.
[497] Molinari A, Tóth LS. Tuning a self consistent viscoplastic model by finite element results—I. Modeling. Acta Metall Mater 1994;42:2453–8.
[498] Hosseini E, Kazeminezhad M. The effect of ECAP die shape on nano-structure of materials. Comput Mater Sci 2009;44:962–7.
[499] Hosseini E, Kazeminezhad M. A hybrid model on severe plastic deformation of copper. Comput Mater Sci 2009;44:1107–15.
[500] Hosseini E, Kazeminezhad M. Implementation of a constitutive model in finite element method for intense deformation. Mater Des 2011;32:487–94.
[501] Hosseini E, Kazeminezhad M, Mani A, Rafizadeh E. On the evolution of flow stress during constrained groove pressing of pure copper sheet. Comput
Mater Sci 2009;45:855–9.
[502] Kazeminezhad M, Hosseini E. Coupling kinetic dislocation model and Monte Carlo algorithm for recrystallized microstructure modeling of severely
deformed copper. J Mater Sci 2008;43:6081–6.
[503] Beyerlein IJ, Lebensohn RA, Tomé CN. Modeling texture and microstructural evolution in the equal channel angular extrusion process. Mater Sci Eng,
A 2003;345:122–38.
[504] Roters F, Raabe D, Gottstein G. Work hardening in heterogeneous alloys—a microstructural approach based on three internal state variables. Acta
Mater 2000;48:4181–9.
[505] Shanthraj P, Zikry MA. Dislocation density evolution and interactions in crystalline materials. Acta Mater 2011;59:7695–702.
[506] Rezvanian O, Zikry MA, Rajendran AM. Microstructural modeling of grain subdivision and large strain inhomogeneous deformation modes in f.c.c.
crystalline materials. Mech Mater 2006;38:1159–69.
[507] Nix W, Gibeling J, Hughes D. Time-dependent deformation of metals. Metall Mater Trans A 1985;16:2215–26.
[508] Zangiabadi A, Kazeminezhad M. Computation on new deformation routes of tube channel pressing considering back pressure and friction effects.
Comput Mater Sci 2012;59:174–81.
[509] Tóth LS, Arzaghi M, Fundenberger JJ, Beausir B, Bouaziz O, Arruffat-Massion R. Severe plastic deformation of metals by high-pressure tube twisting.
Scr Mater 2009;60:175–7.
[510] QForm. Software for Metal Forming Simulation.
[511] Latypov MI, Alexandrov IV, Beygelzimer YE, Lee S, Kim HS. Finite element analysis of plastic deformation in twist extrusion. Comput Mater Sci
2012;60:194–200.
[512] Latypov MI, Lee M-G, Beygelzimer Y, Kulagin R, Kim HS. Simple shear model of twist extrusion and its deviations. Met Mater Int 2015;21:569–79.
[513] Beygelzimer Y, Kulagin R, Latypov MI, Varyukhin V, Kim HS. Off-axis twist extrusion for uniform processing of round bars. Met Mater Int
2015;21:734–40.
[514] Chen C, Beygelzimer Y, Toth LS, Estrin Y, Kulagin R. Tensile yield strength of a material preprocessed by simple shear. J Eng Mater Technol
2016;138:031010–31014.
[515] Rosochowska M, Rosochowski A, Olejnik L. FE simulation of micro-extrusion of a conical pin. Int J Mater Form 2010;3:423–6.
[516] Kim HS, Nam JS. Quantitative modeling and characterization of the size effects in microscale coining process of copper. Precis Eng 2018;51:490–8.
[517] Molotnikov A, Lapovok R, Gu CF, Davies CHJ, Estrin Y. Size effects in micro cup drawing. Mater Sci Eng, A 2012;550:312–9.
[518] Estrin Y, Kim HS, Kovler M, Berler G, Shaviv R, Rabkin E. Modeling of aluminum via filling by forcefill. J Appl Phys 2003;93:5812–5.
[519] McQueen HJ, Yue S, Ryan ND, Fry E. Hot working characteristics of steels in austenitic state. J Mater Process Technol 1995;53:293–310.
[520] Huang C, Hawbolt EB, Chen X, Meadowcroft TR, Matlock DK. Flow stress modeling and warm rolling simulation behavior of two Ti–Nb interstitial-free
steels in the ferrite region. Acta Mater 2001;49:1445–52.
[521] Shih MH, Yu CY, Kao PW, Chang CP. Microstructure and flow stress of copper deformed to large plastic strains. Scr Mater 2001;45:793–9.
[522] Li YJ, Valiev R, Blum W. Deformation kinetics of ultrafine-grained Cu and Ti. Mater Sci Eng, A 2005;410–411:451–6.
[523] Mazilkin AA, Straumal BB, Rabkin E, Baretzky B, Enders S, Protasova SG, et al. Softening of nanostructured Al-Zn and Al-Mg alloys after severe plastic
deformation. Acta Mater 2006;54:3933–9.
[524] Estrin Y, Mecking H. A unified constitutive model with dislocation densities as internal variables. In: Boehler J-P, Khan A, editors. Anisotropy and
localization of plastic deformation. Netherlands: Springer; 1991. p. 385–8.
[525] Estrin Y. 2 - Dislocation-density-related constitutive modeling. In: Krausz AS, Krausz K, editors. Unified constitutive laws of plastic deformation. San
Diego: Academic Press; 1996. p. 69–106.
[526] Kong LX, Hodgson PD, Wang B. Development of constitutive models for metal forming with cyclic strain softening. J Mater Process Technol 1999;89–
90:44–50.
[527] Oudin A, Hodgson PD, Barnett MR. EBSD analysis of a Ti-IF steel subjected to hot torsion in the ferritic region. Mater Sci Eng, A 2008;486:72–9.
[528] Porter DA, Easterling KE, Sherif MY. Phase transformations in metals and alloys. 3rd ed. Boca Raton, FL: CRC Press; 2009.
[529] Sakai T, Belyakov A, Kaibyshev R, Miura H, Jonas JJ. Dynamic and post-dynamic recrystallization under hot, cold and severe plastic deformation
conditions. Prog Mater Sci 2014;60:130–207.
[530] Galindo-Nava EI, Rivera-Díaz-del-Castillo PEJ. Grain size evolution during discontinuous dynamic recrystallization. Scr Mater 2014;72:1–4.
[531] Sakai T, Jonas JJ. Overview no. 35 Dynamic recrystallization: Mechanical and microstructural considerations. Acta Metall 1984;32:189–209.
[532] Kaibyshev OA, Kaibyshev R, Salishchev G. Formation of submicrocrystalline structure in materials during dynamic recrystallization. Mater Sci Forum
1993;113–115:423–8.
[533] Murr LE, Pizana C. Dynamic recrystallization: The dynamic deformation regime. In: 4th Symposium on the dynamic behavior of materials held at the
2007 TMS annual meeting and exhibition. Orlando (FL): Minerals Metals Materials Soc; 2007. p. 2611–28.
[534] Hallberg H, Wallin M, Ristinmaa M. Modeling of continuous dynamic recrystallization in commercial-purity aluminum. Mater Sci Eng, A
2010;527:1126–34.
[535] Le KC, Kochmann DM. A simple model for dynamic recrystallization during severe plastic deformation. Arch Appl Mech 2008;79:579.
[536] Kim HS, Estrin Y, Bush MB. Plastic deformation behaviour of fine-grained materials. Acta Mater 2000;48:493–504.
[537] Kim HS, Estrin Y, Bush MB. Constitutive modelling of strength and plasticity of nanocrystalline metallic materials. Mater Sci Eng, A 2001;316:195–9.
[538] Mercier S, Molinari A, Estrin Y. Grain size dependence of strength of nanocrystalline materials as exemplified by copper: an elastic-viscoplastic
modelling approach. J Mater Sci 2007;42:1455–65.
[539] Kim HS, Estrin Y. Strength and strain hardening of nanocrystalline materials. Mater Sci Eng, A 2008;483:127–30.
[540] Mukherjee AK. An examination of the constitutive equation for elevated temperature plasticity. Mater Sci Eng, A 2002;322:1–22.
[541] Langdon TG. An analysis of flow mechanisms in high temperature creep and superplasticity. Mater Trans 2005;46:1951–6.
[542] Yamakov V, Wolf D, Phillpot SR, Gleiter H. Grain-boundary diffusion creep in nanocrystalline palladium by molecular-dynamics simulation. Acta
Mater 2002;50:61–73.
[543] Klusemann B, Bargmann S, Estrin Y. Fourth-order strain-gradient phase mixture model for nanocrystalline fcc materials. Modell Simul Mater Sci Eng
2016;24:085016.
[544] Sanders PG, Eastman JA, Weertman JR. Elastic and tensile behavior of nanocrystalline copper and palladium. Acta Mater 1997;45:4019–25.
[545] Wu XL, Jiang P, Chen L, Zhang JF, Yuan FP, Zhu YT. Synergetic strengthening by gradient structure. Mater Res Lett 2014;2:185–91.
[546] Zhilyaev AP, Langdon TG. Using high-pressure torsion for metal processing: Fundamentals and applications. Prog Mater Sci 2008;53:893–979.
[547] Zhilyaev AP, Oh-ishi K, Langdon TG, McNelley TR. Microstructural evolution in commercial purity aluminum during high-pressure torsion. Mater Sci
Eng, A 2005;410:277–80.
[548] Bouaziz O, Kim HS, Estrin Y. Architecturing of metal-based composites with concurrent nanostructuring: a new paradigm of materials design. Adv
Eng Mater 2013;15:336–40.
[549] Raabe D, Choi P-P, Li Y, Kostka A, Sauvage X, Lecouturier F, et al. Metallic composites processed via extreme deformation: toward the limits of
strength in bulk materials. MRS Bull 2010;35:982–91.
[550] Beygelzimer Y, Estrin Y, Kulagin R. Synthesis of hybrid materials by severe plastic deformation: a new paradigm of SPD processing. Adv Eng Mater
2015;17:1853–61.
[551] Beygelzimer Y, Kulagin R, Estrin Y. Severe plastic deformation as a way to produce architectured materials. In: Estrin Y, Dunlop J, Brechet Y, Fratzl P,
Dendievel R, editors. Architectured materials. Springer; 2018.
[552] Bouaziz O. Geometrically induced strain hardening. Scr Mater 2013;68:28–30.
[553] Khoddam S, Estrin Y, Kim HS, Bouaziz O. Torsional and compressive behaviours of a hybrid material: Spiral fibre reinforced metal matrix composite.
Mater Des 2015;85:404–11.
[554] Latypov MI, Beygelzimer Y, Kulagin R, Varyukhin V, Kim HS. Toward architecturing of metal composites by twist extrusion. Mater Res Lett
2015;3:161–8.
[555] Diez M, Kim H-E, Serebryany V, Dobatkin S, Estrin Y. Improving the mechanical properties of pure magnesium by three-roll planetary milling. Mater
Sci Eng, A 2014;612:287–92.
[556] Ivanisenko Y, Kulagin R, Fedorov V, Mazilkin A, Scherer T, Baretzky B, et al. High pressure torsion extrusion as a new severe plastic deformation
process. Mater Sci Eng, A 2016;664:247–56.
[557] Gu CF, Tóth LS, Lapovok R, Davies CHJ. Texture evolution and grain refinement of ultrafine-grained copper during micro-extrusion. Phil Mag
2011;91:263–80.
[558] Lebensohn RA, Tomé CN. A self-consistent anisotropic approach for the simulation of plastic-deformation and texture development of polycrystals -
application to zirconium alloys. Acta Metall Mater 1993;41:2611–24.
[559] Ayoub G, Rodrigez AK, Shehadeh M, Kridlia G, Young JP, Zbib H. Modeling the rate and temperature dependent behavior and texture evolution of the
Mg AZ31B alloy TRC sheets. Phil Mag 2017. in press.
[560] Tóth LS, Molinari A. Tuning a self consistent viscoplastic model by finite element results—II. Application to torsion textures. Acta Metall Mater
1994;42:2459–66.
[561] Skrotzki W, Scheerbaum N, Oertel CG, Arruffat-Massion R, Suwas S, Toth LS. Microstructure and texture gradient in copper deformed by equal
channel angular pressing. Acta Mater 2007;55:2013–24.
[562] Horstemeyer MF, Potirniche GP, Marin EB. Crystal plasticity. In: Yip S, editor. Handbook of materials modeling. Netherlands: Springer; 2005. p.
1133–49.
[563] Lebensohn RA, Tomq CN. A self-consistent anisotropic approach for the simulation of plastic deformation and texture development of polycrystals:
application to zirconium alloys. Acta Metall Mater 1993;41:2611–24.

Modelling SPD

Uploaded by

Copyright:

Available Formats

Modelling SPD

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Modelling SPD

Uploaded by

Copyright:

Available Formats

Progress in Materials Science 95 (2018) 172–242

Contents lists available at ScienceDirect

Progress in Materials Science

Analytical and numerical approaches to modelling severe

2.5.2. Estimates for temperature effects during SPD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188

2. Modelling severe plastic deformation

2.1. Estimates of equivalent strain during ECAP

2.2. Constitutive modelling approaches to simulations of ECAP

2.3. Effects of heat generation

2.4. Strain localisation during ECAP

2.5. Strain rate and temperature effects during ECAP

2.5.1. Estimates for strain rate during ECAP

e_ xx ¼ uð1 nfl Þðddie xÞnfl 1 ðddie yÞnfl 1 ðddie x0 Þ12nfl ;

2.5.2. Estimates for temperature effects during SPD

2.5.3. Microstructure evolution during SPD in non-isothermal conditions

Here the time derivatives d_ and d

o Þ with strain combining grain refinement and

3. Microstructure-based phenomenological modelling of severe plastic deformation

3.1. Evolution of dislocation ensembles: An irreversible thermodynamics approach

3.2. One-internal-variable models

3.2.1. Constitutive modelling of materials with a high Peierls stress

X ¼ 2ak exp½F 0k ðrP =MÞ=kB T ð71Þ

Here the notation

3.2.2. Constitutive modelling of hexagonal materials

3.2.3. Inclusion of deformation twinning in the constitutive modelling

3.3. Two-internal-variable models

pffiffiffiffiffiffi pffiffiffiffiffi " s0c

3.4. Constitutive modelling of multi-phase materials

4. Possible scenarios for grain refinement

5. Numerical simulations of SPD processes

5.1. FEM simulations of SPD processing

5.2. Softening at large strains and dynamic recrystallisation

d ¼ d0 ðd0 df Þ exp½K X hepl epl c inX  ð97Þ

6. Phase mixture modelling of nanocrystalline materials

7. Gradient plasticity for SPD processed materials

r ¼ f rGB þ ð1 f ÞrGI þ hg ðl2 @ 2x e þ 2=5l4 @ 4x eÞ ð104Þ

8. Simulations of synthesis of architectured materials by SPD techniques

9. Texture evolution during SPD processing

e_ ij E_ ij ¼ a~ Mijkl ðrkl Rkl Þ ð105Þ

You might also like

X ¼ 2ak exp½F 0k ðrP =MÞ=kB T ð71Þ

d ¼ d0 ðd0 df Þ exp½K X hepl epl c inX ð97Þ