Thermodynamical and self consistent

approach to inelastic ferromagnetic

polycrystals
0
M. MI

CUNOVI

C
Faculty of Mechanical Engineering, University of Kragujevac,
Sestre Janjica 6a, 34000 Kragujevac, Serbia and Montenegro
e-mail: mmicun@kg.ac.yu
Dedicated to memory of Professor Henryk Zorski
with gratitude and admiration
Geometric and kinematic aspects of intragranular as well as intergranular plas-
tic deformation of ferromagnetic polycrystals are discussed. Elastic strain is
covered by the eective eld homogenization method inside a representative
volume element (RVE). By applying this method an eective magnetostriction
4-tensor is determined. Evolution equation formed by tensor representation
having incremental form is postulated to model inelastic metals. The rate de-
pendence takes place by means of stress rate dependent value of the initial
yield stress. Concept of M. Zorawski deformation geometry is extended on the
base of constrained micro and free macro rotations in intermediate reference
conguration. This has as a consequence that evolution equation for plastic
spin of RVE is an outcome of evolution equation for plastic stretching. The
macroscopic evolution equation is based on the Vakulenkos concept of ther-
modynamic time. A tensor representation for magnetomechanical interaction
is proposed and susceptibility coecients for iron are calibrated.
Keywords: Anisotropic Eshelby tensor, Tensor representation, Vakulenkos
thermodynamic time, Susceptibility coecients.
1 Introduction
The principal objective of this paper is to give a simplied approach to in-
elasticity of (inherently polycrystalline) ferromagnetic materials aimed to serve
primarily to nondestructive magnetic examination of inelastic behavior of reac-
tor steels (cf. [30, 56, 47]). The subject is complicated and requires a careful
examination of geometry as well as thermodynamics of deformation process.
As the starting point the geometry of deformation with multiplicative de-
formation gradient must be extended in order to include magneto-mechanical
interaction. This is done in the easiest way following the Kroners incom-
patibility method which in the paper [15] was applied to magnetostriction.
0
Submitted to Archives of Mechanics, Warszawa 2005
1
The authors assumed that total incompatibility composed of purely elastic and
magnetostrictive strain is zero while its constituents are not. They assumed
isoclinicity of magnetization vectors in the natural state elements (intermedi-
ate local reference conguration) and their inhomogeneity in instant deformed
conguration which is responsible for magnetostrictive strains. Such an as-
sumption is very much in accord with our geometrical approach (cf. [32]) and
is accepted henceforth. An extension of their reasoning to the more general
case of thermo-elasto-magneto-plastic strain is allowed if Kroners formula
(1) F = F
E
F
P
is understood as follows. The incompatible tensor F
E
is obtained by cutting
the body into innitesimal pieces which are free of neighbors, brought to the
reference temperature in the absence of external magnetic eld. The other
constituent on right hand side of (1), namely F
P
, contains strain caused by
irreversible magnetization as well as by pure plastic strain. This will be further
elaborated in the next section.
Our approach to evolution equations does not take as granted associativ-
ity of ow rule i.e. the normality of the plastic stretching tensor onto a yield
surface [34, 35, 28] even if such an approach is accepted in the majority of the
papers dealing with the subject. Such a normality is seriously questioned not
only by the theoretical but by experimental results as well.
1
For these reasons
the normality is at rst abandoned and instead of such an assumption evolu-
tion equations (exposed in the second subsection of this section) are based on
the appropriate geometry of deformation and tensor representation. Each rea-
sonable constitutive theory must have a tight relationship to thermodynamics.
Here we mention some possible approaches.
A very attractive approach to the extended thermodynamics has been pro-
posed in [42] with a rational analysis of thermodynamic processes leading
to the desired thermodynamic restrictions of general constitutive equa-
tions. This approach with the Lius theorem [25] was applied by this
author to viscoplastic materials in [34] and to inelastic composite materi-
als in [35]. In spite of its beauty an inherent coldness function (which is
not quite clear from the experimental point of view) is inevitable.
An alternative approach to extended thermodynamics following [8] was
applied to thermoplasticity of irradiated materials in [28] and viscoplas-
ticity of single ferromagnetic crystals in [37].
1
In the paper [16] a comparison between tension and torsion was one of the rst signs of
such a discrepancy. This subject has been discussed in detail in [41] where also experiments
dealing with cruciform specimens [1] are included. In so called J
2
-theories with corners where
a lot of normals to yield surface exist this normality is in fact abandoned.
2
The approach of endochronic thermodynamics with properly chosen ther-
modynamic time is the succeeding choice. It will be presented here in the
version of Vakulenko [53].
Statistical thermodynamics is a mighty tool in treating choice of internal
variables. It will be partly used here in dening magnetization distribu-
tions. In this eld developments of Zorski [57] and Kroner [22] are very
important.
The analysis in this paper is aimed to a description of fast multiaxial ex-
periments on austenitic steels like AISI 316H having face centered cubic lattice
(compare [36]) as well other steels with body centered cubic lattice. For this
sake it is essential to reduce the number of material constants to be found from
the available experiments. In other words, the general desire is always to make
evolution equations with minimal number of material constants even if these
equations originate form very general functionals. The evolution equations usu-
ally comprise of plastic stretching (often named by experimentalists as plastic
strain rate tensor) as well as plastic spin. The later is understood by some
authors as a trigger in localization behavior while some others require indepen-
dence of these two evolution equations which greatly complicates identication
problem. This issue will be touched as well.
Means applied in the paper to realize the stated goal are listed below. Micro-
evolution equation having incremental form is postulated. The rate dependence
takes place by means of stress rate dependent value of the initial yield stress.
Thus, such materials are quasi-rate dependent. Such an approach follows re-
sults of dynamic experiments on AISI-steels performed at JRC-Ispra, Italy at
strain rates in the interval [0.001, 1000] 1/s (described in [36]. The still con-
troversial plastic spin issue is treated by extending the concept of M. Zorawski
deformation geometry postulating constrained micro and free macro rotations
in intermediate reference conguration. The self-consistent method (eective
eld variant) resulting from paper [24] provides eective stiness tensors for
RVE from individual grains and leads to a simplication of tedious calculations
of their eective stiness fourth rank tensors of individual grains.
Throughout the paper thermal, elastic and magnetostrictive strains are as-
sumed to be small being approximately additive. This does not hold for plastic
strains which are nite in general. However, in the fourth section devoted to
the self consistent method plastic strains are also considered as small due to
assumptions in eective eld theory developed in the papers [18, 19, 24].
At present, only rst deformation gradients are considered. Thus, higher
order eects like mechanical couple stresses (requiring a gradient viscoplasticity
treated in [39]) as well as phase transitions, ferroelectric and ferrimagnetic
eects, intrinsic spin, exchange forces and gyromagnetic eects are ignored
(with negligible precessional velocity of magnetization). The considered process
3
is electromagnetically slow enough such that ratio of particle velocity and speed
of light is negligible. Due to size of paper boundary conditions are not analyzed.
The notation used in this paper might be briey shown by summation over
repeated Cartesian indices:

H

M = H
a
M
a
, (

H

M)
ab
= H
a
M
b
, (AB)
ab
=
A
ac
B
cb
, tr(AB) = A
ac
B
ca
, (DE)
ab
= D
abcd
E
dc
. The usual notation: 2 symA =
A+A
T
and 2 skwA = AA
T
is also applied here. Due to limited space some
second order eects (treated in detail in [39]) have to be dropped from the con-
sideration. For the same reason many important references cannot be included
into already long list of references.
2 Micro and macro-geometry
2.1 Magneto-thermo-mechanical distortions
As a prerequisite, a correct geometric description of an inelastic deformation
process analyzed is necessary. Consider a polycrystalline body in a real con-
guration (k) (k(t)) with dislocations, magnetization

M(X, t) and an inho-
mogeneous temperature eld T(X, t) (where t stands for time and X for the
considered particle of the body) subject to external stress (i.e. surface tractions)
and external magnetic eld

H. Corresponding to (k) there exists, usually, an
initial reference conguration (K) = (k (t
0
)) with dislocations at a homoge-
neous temperature T
0
without surface tractions. Due to these defects such a
conguration is not stress-free but contains an equilibrated residual stress (so
called back-stress). It is generally accepted that linear mapping function
F(., t) : (K) (k) is compatible second rank total deformation gradient ten-
sor. In the papers dealing with continuum representations of dislocation distri-
butions conguration (k) is imagined to be cut into small elements denoted by
(n) (n(t)), these being subsequently brought to the temperature of (K) free
of neighbors. The deformation gradient tensor F
E
(., t) : (n) (k) obtained
in such a way is incompatible and should be called the thermoelastic distortion
tensor whereas (n)-elements are commonly named as natural state local refer-
ence congurations (cf. for instance [21, 32, 15]). Of course, the corresponding
plastic distortion tensor is not compatible. Here F is found by comparison of
material bres in (K) and (k) while F
E
is determined by crystallographic vec-
tors in (n) and (k). Multiplying (1) from the right hand side by F
E
(., t
0
)
1
we reach at slightly modied Kroners decomposition rule
2
. The formula (1)
is often wrongly named as Lees decomposition formula. It is worthy of note
that curlF
E
(., t)
1
,= 0 and this incompatibility is commonly attributed to an
2
This modication, representing mapping (k(t
0
)) (k(t)) and introduced by Teodosiu
[51] is necessary to account for Dashner replacement invariance [12].
4
asymmetric second order tensor of dislocation density
3
.
Taking into account the above discussion, it is reasonable to decompose
irreversible magneto-plastic distortion as follows:
(2) F
P
= F
R

F
p
.
It contains irreversible residual magnetic distortion F
R

and pure plastic distor-

tion F
p
. On the other hand, magneto-thermo-elastic distortion may be decom-
posed by means of:
(3) F
E
= F
e
F

F
r

with pure elastic distortion F

e
, thermal distortion F

as well as reversible mag-

netic distortion F
r

. Now it is reasonable to dene thermo-magnetic quasi-plastic

distortion tensor as the product of thermal and magnetic distortion:
(4) F

= F

F
r

F
R

.
The name quasi-plastic was introduced by Anthony in [2]. Suppose now
that thermal as well as magnetostrictive strains are much smaller than plastic
strain. This is conrmed by experiments even for plastic strains of the order
of 1%. Then for thermo-magnetic quasi-plastic strain a linear decomposition
approximately holds:
(5) E

+ E
r

+ E
R

+ E

.
Here E

= (F
T

1)/2, , , are Lagrangian strains and E

is called
magnetostrictive strain. Due to magnetic symmetry it is bilinear function of
magnetization vector [31]. Its constituents are reversible magnetostrictive strain
and irreversible magnetostrictive strain. Their explicit forms are given in the
next section.
2.2 Polycrystal strains
Let us imagine now that a typical (n)-element (called in the sequel represen-
tative volume element and denoted by RVE) is composed of N
g
single crystal
grains such that each -th grain has N
s
slip systems A

=

s

. For
instance, for fcc-crystals N
s
= 24. Here

s

is the unit slip vector and

n

3
This denition covers only geometrically necessary dislocations (GND) without statisti-
cally stored dislocations (SSD) appearing at dislocation loops and dipoles. For this reason in
the paper [22] a more precise denition is given by innite number of correlation functions
composed by fundamental dyadic of Burgers vector and dislocation line tangent vector. On
the other hand, F
P
and its rst gradients allow diverse non-Euclidean interpretations covering
not only GND but implanting Eshelbian strains as well (cf. details in [32, 40]).
5
is the unit vector normal to the slip plane. Comparing a RVE in (n(t)) and
(n(t
0
)) we may write a formula similar to Kroners formula (1) holding for the
microplastic distortion tensor
(6)

:=
E

P
,
whose components are the residual microelastic distortion tensor
E
and mi-
croplastic distortion tensor
P
. Then the polar decomposition gives

E
= R

U
E
.
Here microrotation satises the relations R
T

= 1 and its time rate equals to

D
t
R

. Therefore, slip systems dyadics evolve according to: A

(t) =
R

(t)A

(t
0
)R
T

(t) as microrotations must be constrained inside each RVE. By

making use of these dyadics as well as microrotations we may write
U
E
= diag(1 +
k
), k 1, 2, 3
as well as

P
:= 1 +

,
where

( 1, N
s
, 1, N
g
) are plastic microshears inside the th
grain. If a RVE has the volume V =

and the microplastic deforma-

tion tensors for individual grains are
C

=
T
P
U
2
E

P

_
1 +

A
T

_
U
2
E
_
1 +

_
,
then their volume average named macroplastic deformation tensor C
P
:= F
T
P
F
P
has the following form:
(7) C
P
= C

) =

_

1
V

.
Moreover, in the corresponding polar macro-decomposition F
P
= R
P
U
P
the macroplastic rotation tensor R
P
is free to be taken arbitrary (according to
Zorawski [58] and might be xed either to be a unit tensor or to have Mandels
isoclinicity property (details are given in [41]). For a denition of isoclinicity
we should have to nd average crystal directions in RVE(t) and RVE(t
0
) and
to make them coincident. Accepting henceforth the rst choice we acquire the
relationship
(8) R
P
= 1 F
P
= U
P
= C
1/2
P
,
6
0
RVE(t)
0
RVE(t)
F (t)
p
U (t)
p
=
(K)
(t)
0
F
E
(t) F
E
(t) F
(t))
(k
(n)
0
(n(t))
Figure 1: Principal congurations of a polycrystal body with illustration of free
macro and constrained micro-rotation
which greatly simplies macroplastic spin issue (cf. [41]).
The above introduced microrotations of grains permit the exact relationship
for material time rate of microplastic distortion tensor (cf. gure 1)
D
t

P
=

D
t

D
t
A

.
Here the aforementioned constrained microrotations must full the relationship
D
t
A

+A

,
such that microplastic stretching and microplastic spin tensors read:
D

= R
T

(D
P
+D
t
log U
E
) R

, W

= R
T

W
P
R

+W
E
,
with
W
E
= D
t
R
T

7
and
D
t
log U
E
= diag
_
D
t

k
(1 +
k
)
1
_
, k 1, 2, 3
as well as next notations
2D
P
= D
t

1
P
+
T
P
D
t

T
P
, 2W
P
= D
t

1
P

T
P
D
t

T
P
.
The corresponding macroplastic stretching and macroplastic spin tensors follow
now directly from (8) in the following form:
(9) 2D
P
= D
t
U
P
U
1
P
+U
1
P
D
t
U
P
, 2W
P
= D
t
U
P
U
1
P
U
1
P
D
t
U
P
.
It is worthy of note that such a representation considerably reduces num-
ber of necessary material constants if some evolution equations for macro-
quantities D
P
and W
P
are chosen in such a way to follow from tensor rep-
resentation. Connection of the macroplastic stretching with (7) by means of
2D
P
= U
1
P
D
t
C
P
U
1
P
is then straightforward and is obtained from:
D
t
C

=
T
P
D
t
U
2
E

P
+
_

D
t

D
t
A

_
U
2
E

P
+

T
P
D
t
U
2
E
_

A
T

D
t

D
t
A
T

_
by means of the spatial averaging throughout a RVE i.e. D
t
C
P
= D
t
C

) .
In the paper [38] an initial attempt has been made to model transition of
plastic strain from a grain to its neighbors. Yet an application of such an idea
to computer simulation of inelastic behavior of RVE would require too large
computing time. Instead of that here a self-consistent method is applied. It
will be explained in more detail in the subsequent section.
3 Macroscopic evolution and constitutive equations
3.1 Evolution equations by extended thermodynamics
Let us consider only magnetomechanical terms in balance laws and constitutive
equations. More precisely the scope of this subsection is limited by the next
assumption:
Assumption 3.1 Ferroelectric and ferrimagnetic eects, intrinsic spin, ex-
change forces and gyromagnetic eects are ignored (with negligible precessional
velocity of magnetization). The considered process is slow enough such that
ratio of particle velocity and speed of light is negligible.
4
4
Details of such a situation are explained in [31, 37].
8
Then, the next reduced set of objective and Galilei-invariant state variables
[30, 46] should be introduced in general
(10) := E, E
P
, A, T, GRADT,

Q,

/,

/
R
, (,
The tensorial quantities are chosen here in such a way to be connected to the
convective material X-coordinates
5
in the deformed instant (k)-conguration
(cf. gure 1).
Herein
2 E = F
T
F 1 C1 is the Lagrangean total strain tensor,
2E
P
= F
T
P
F
P
1 C
P
1 - Lagrangean plastic strain tensor,
GRADT F
1
gradT - temperature gradient,

Q JF
1
q - the heat ux vector ,

/ = JF
1
m - the magnetization vector,

/
R
- the corresponding irreversible (residual) magnetization vector,
A - the volume dened dislocation density (number of dislocation
lines per unit volume).
Capital letters are reserved for such a convective representation. Dierential
operator GRAD

C
K

K
is referred to such coordinate frame, whereas
grad

g
a

a
is used to indicate the same operator in spatial (possibly Carte-
sian) coordinate frame of (k)-conguration. Accordingly, GRADT,

Q,

/,

/
R
have convective material either covariant or contravariant components. The
above set may be otherwise understood as a point belonging to the ex-
tended conguration (deformation-temperature-magnetic) space (. Its subset
E
P
, A,

/
R
collects internal variables responsible for irreversible behavior.
To this conguration point there corresponds a reaction point represented
by the set
(11)
1
:= T
K
, u, s,

S,

c,

B,
1
T,
where
6
5
Another choice is to accept convective structural coordinates in one of local natural state
congurations depicted on the gure 1. However, this is more dicult here since it is necessary
to account for non-Euclidean expressions of dierential operators (cf. [41] for details).
6
Our convective magnetic vectors coincide with those in the comprehensive reference
([31, p. 169]) with the exception of magnetic induction and magnetization where

M =
J C
1

M
Maugin
and

B = J
1
C

B
Maugin
holds.
9
T
K
= JF
1
T
k
F
T
- the symmetric Piola-Kirchho stress tensor of
the second kind related to the material convective coordinates of (k)-
conguration wherein T
k
is the Cauchy stress,
u and s - the internal energy and the entropy densities,

S JF
1
s - the entropy ux vector,

B =

bF - the magnetic induction vector.

By means of T the extended stress space is indicated, whose objective and
Galilei-invariant elements are listed hereinabove in (11). At this place the
constitutive equations are simply stated by the bijective mapping:
(12)
1
= ()
1
() or : ( T,
which is too general so that the thermodynamic analysis presented henceforth is
aimed to supply restrictions concordant with the second law of thermodynamics.
The evolution functions are proposed here in such a way to be compatible
with (10)-(12) and are collected into the set
(13)
2
:=

, E

,

M

, A

,
2
T,
such that objective evolution equations simply read:
(14) D
t

Q =

Q

(),
(15) D
t
E
P
= E

(),
(16) D
t

/
R
=

M

(),
(17) D
t
A = A

(),
where the material time derivative is designated by D
t
. The simplicity of left
hand sides of (14-17) owing to the absence of corotational time derivatives
has the origin in the accepted material convective description of constitutive
functions and variables listed in (10)-(11)
7
.
A thermodynamic process occurring in the considered body is described by
the evolution equations and by the following balance laws which are equivalent
but slightly modied with regard to those of [31]:
D
t
u
_
T
K
+
_
1(

B

/)

B

/
_
C
1
_
: D
t
E

BD
t

/+DIV

Q
0
h = 0,
(18)
7
It is noted here that right hand sides of (14-17) do not include material time derivatives
of internal variables - elements of the set
1
in (11). In other words, quasi rate independent
ferromagnetics (cf. [40]) are not covered by the above evolution equations (14)-(17). In the
subsections 3.2 and 3.4 this approach is extended to include such materials as well.
10
(19)
0
J = 0,
(20) D
t
v

f

f
em

1
J
DIV (T
K
F
T
) =

0,
(21) skwT
K
= skw
_
C
1
(

B

/

P

c)
_
,
wherein
0
and are mass densities in (K) and (k), v is the velocity of the
particle, while the conventional notation:
J

f
em
F J GRAD(
1
J

BF
T
)F
T

/,
is used. The above equations (18)-(21) are, respectively, the equation of energy
balance, the mass conservation law, the equation of balance of momentum and
the equation of balance of angular momentum.
Let us remind that electric quantities are not considered in this paper. Then
one of Maxwell equations becomes identity whilst the others read:
(22) D
t

B = 0,
(23) DIV

B = 0,
(24) CURL(

B J
1
C

/) =

0.
Further consequence of the assumption 3.1 is a simplication of the set of
internal variables loosing from it gradient of the magnetization vector assuming
in such a way that balance law for magnetization [45] (i.e. balance of angular
momentum of spin continuum in wording of [31]) is identically satised.
The above listed balance laws imply constraints on the elements of the set
D
t
causing breaking of their independence which is the essence of the
Lius theorem (given in [25]).
Nonetheless, there is still another constraint on these elements in the case of
inelastic deformation process: the essential notion of yield surface which divides
sharply two regions of material behavior. In other words, the elastic and plastic
strain ranges are separated by the yield surface.
8
The dynamic and static scalar
yield functions are here dened by means of
(25) f = f(T
K
, T, E
P
,

/
R
) h(),
8
An essentially important question to be posed here reads: whether such a frontier between
reversible and irreversible magnetization exists or not. If the answer is armative, then it
leads to next questions: which irreversible process (mechanical or magnetic) advances and
what is shape of magnetic yield surface? It is tacitly assumed in this subsection that both
processes are simultaneously triggered. In subsection 3.4 some other possibilities are discussed.
11
(26) f
#
= f(T
#
K
, T, E
P
,

/
R
) h
0
(),
where T
#
K
is static stress corresponding to the dynamic viscoplastic stress T
K
.
Their dierence is usually termed as the overstress tensor and in the simplest
case it may be represented by a linear function of D
t
E
P
as follows:
(27) T
K
:= T
K
T
#
K
= P() : D
t
E
P
,
with P() being fourth rank tensor of plastic viscosity coecients. Introducing
the plastic strain rate intensity by
(28) D
t
p := (D
t
E
P
: D
t
E
P
)
1/2
|D
t
E
P
| 0,
the classication:
f > 0, f
#
= 0, Dp > 0 - viscoplastic behavior;
f = f
#
= 0, Dp = 0 - elastoplastic frontier;
f = f
#
< 0, Dp = 0 - elastic behavior;
and the kinematic constraint:
9
(29) < f > D
t
f
#
= 0,
will be useful in the sequel.
All thermodynamic processes must obey the master law of nature i.e. the
second law of thermodynamics which in material convective coordinate frame
of (k)-conguration reads:
(30) D
t
s +DIV

S
r
T
= 0,
where r/T is the entropy source. Precisely dened a thermodynamic process is
a solution of evolution and balance equations which obeys (30). The analysis
of the above entropy inequality (30) by the Lius theorem may be described
as follows. Replacing s

() and

S

() into (30) this becomes a dierential

inequality linear with respect to the elements of the set D
t
GRAD
namely:
9
Here < f >= 1 for f > 0 and f = 0 otherwise. This notation should not be confused
with averaging sign < > used in the previous section.
12

0
D
t
s +DIV

S
r
T

u
_

0
D
t
u

BD
t

/+DIV

Q
0
h
D
t
E :
_
T
K
+
_
1(

B

/)

B

/
_
C
1
__

: skw
_
T
K
C
1

B

/
_

f
< f > D
t
f
#

v
_
D
t
v

f

f
em
J
1
DIV (T
K
F
T
)
_

A
_
D
t
AA

()
_

E
:
_
D
t
E
P
E

()
_

M
_
D
t

/
R


M

()
_

2
D
t

B
3
DIV

B

4
CURL(

B J
1
C

/) 0.
(31)
By introducing Lagrange multipliers into (30) all the elements of the set
D
t
GRAD except solely GRADT (which is already included into )
become mutually independent. Hence, in thus extended inequality all the co-
ecients with the elements of the set D
t
GRAD must vanish. This
gives rise to the consecutive constitutive restrictions (cf. [28]):
(32)

S =
u
(T)

Q

Q/T,
(33) T
K
= C
1
_

/

B + (

B

/)1
_
+
0

E
g+ < f > T
f

E
f
#
_
,
(34) s =
T
g+ < f > T
1

T
f
#
,
(35)

B =
0

M
g < f > T
f

M
f
#
,
(36)

0 =
GRADT
g+ < f > T
1

GRADT
f
#
,
as well as to the residual dissipation inequality
(37)

M

M

() +
E
: E

() +
A
A

() T
2

QGRADT 0,
where g := us(
u
)
1
uTs is the free energy density. If the thermodynamic
process is very near to equilibrium (cf. [28]) then the above residual inequality
permits the direct application of Onsager-Casimir reciprocity relations. Some
of the above Lagrange multipliers are explicitly given by:

M
=
0
T
1

M
g < f >
f

M
f
#
,

E
=
0
T
1

E
g < f >
f

E
f
#
,

A
=
0
T
1

A
g < f >
f

A
f
#
,
whereas the others vanish

v
= 0,

2
= 0,
3
= 0,

4
= 0,

= 0.
The details of the above given procedure are presented in the reference [28]
where thermoplasticity of neutron irradiated steels was considered.
13
3.2 Small magnetoelastic strains of isotropic plastically deformed
insulators and generalized normality
In order to illustrate the above derived constitutive and evolution equations
we accept in this subsection the following very simplifying assumptions for an
isotropic body:
Assumption 3.2 Elastic and thermal strain as well as strain induced by re-
versible and irreversible magnetization are small of the same order but plastic
strain itself is nite (cf. also [28])
Such an assumption corresponds to the so called piezomagnetism processes
when magnetization is generated by straining processes (illustrated in [13, 26]).
Let us take into account that by its very nature the mechanical stress dis-
appears when pure elastic strain vanishes and, similarly, the local magnetic
eld equals to zero if the reversible magnetization vanishes. Then, according
to [2, 15] and the above relationship (5) it is reasonable to introduce magne-
tostrictive quasi-plastic strain by means of
(38) E

= Lg.
Here L is the fourth rank tensor of magnetostriction constants symmetric only
in indices of the rst as well as the second pair whereas the notation: g

stands for diadics of the unit vector of the magnetization vector (trg = 1).
For convenience the elastic strain tensor expressed in material convective
coordinates, accepted in previous subsection, then reads:

E
e
:=
1
2
F
T
p
F
T

F
T

(F
T
e
F
e
1)F

F
p
EE
p
F
T
p
(E

+E

)F
p
.
(39)
As already discussed in detail in the rst section, the constituents of the La-
grangian strain tensor, namely,

E
e
, E

, E

as well as E
p
are incompatible.
With these facts taken into account and the above introduced assumption
3.2 mechanical and magnetic constitutive equations are presented henceforth.
First, mechanical part of the stress tensor must be linear in elastic strain (39).
In the case of elastic and thermal isotropy case we would have the relationship:
T
K
=

D(E
p
)

E
e
(c
1
1 +c
2
E
p
+c
3
E
2
p
) tr

E
e
+ 2c
4

E
e
+ (40)
c
5
(E
p

E
e
+

E
e
E
p
) +c
6
(E
2
p

E
e
+

E
e
E
2
p
),
which clearly exhibits a plastic strain induced anisotropy.
Before formulating a magnetic constitutive equation, let us transform vec-
tors of magnetization and magnetic eld into the corresponding second rank
14
skew-symmetric tensors. This is done by means of the product with material
convective Ricci third rank permutation tensor as follows:
(41) H c

H = H
T
, M
a
c

/
a
= M
T
a
, a r, R.
In the above replacements
(42)

/
r
:=

/

/
R
is the reversible magnetization vector while the skew-symmetric second rank
tensors H, M
r
and M
R
should replace corresponding vectors.
As for magnetic constitutive equation we rst remark that we have accepted
Heavisisde-Lorentz form of Maxwell equations (cf. table on page 59. of [31]).
Thus from the last of these equations i.e. (24) (i.e. CURL

H =

0 ) we see that
the relationship between magnetic induction eld vector

B and the internal
magnetic eld vector

H (opposing the local magnetic eld vector) reads:
(43)

H :=

B J
1
C

/.
Due to the assumption 3.2 the magnetic constitutive equation for an isotropic
ferromagnetic must have the form:
(44) H = c
7
M
r
+c
8
(E
p
M
r
+M
r
E
p
) +c
9
(E
2
p
M
r
+M
r
E
2
p
),
which is linear in reversible magnetization. The hereinabove constitutive ex-
pression for H has been derived from (35) and (43) under assumption 3.1 by
means of tensorial representations for the proper orthogonal group [49].
10
Equation (40) is the generalized Hookes law accounting for plastic strain
induced mechanical anisotropy. It is noteworthy that the constitutive equation
for internal magnetic eld predicts magnetic anisotropy induced by the same
cause.
According to the assumption 3.2 the free energy function generating (40)
and (44) must be quadratic in elastic strain and reversible magnetization i.e.:
g =
1
2
c
1
i
2
1
+
1
2
c
2
i
2
2
+
1
2
c
3
i
2
3
+c
4
i
4
+c
5
i
5
+
c
6
i
6
+
1
2
c
7
i
7
+
1
2
c
8
i
8
+
1
2
c
9
i
9
, (45)
where the consecutive proper and mixed invariants appearing in the above scalar
function must be introduced (cf. [49]):
i
1
= tr

E
e
, i
2
= trE
p

E
e
, i
3
= trE
2
p

E
e
,
i
4
= tr

E
2
e
, i
5
= trE
p

E
2
e
, i
6
= trE
2
p

E
2
e
, (46)
i
7
= trM
2
r
, i
8
= trE
p
M
2
r
, i
9
= trE
2
p
M
2
r
.
10
The skew-symmetric tensors are favored instead of the corresponding vectors for conve-
nience and more compact representation. Of course, an equivalent formulation using products
of vectors

M
r
and

M
R
with symmetric second rank tensor E
p
is also possible.
15
In the sequel inverse forms of (40) and (44) will be useful. They can be written
in the following way:

E
e
= (
1
1 +
2
E
p
+
3
E
2
p
) trT
K
+ 2
4
T
K
+ (47)

5
(E
p
T
K
+T
K
E
p
) +
6
(E
2
p
T
K
+T
K
E
2
p
),
and
(48) M
r
=
7
H+
8
(E
p
H+HE
p
) +
9
(E
2
p
H+HE
2
p
).
The relationships between sets c
1
, ..., c
9
and
1
, ...,
9
can be found as fol-
lows. Let us multiply (40) as well as (47) by the tensors 1, E
p
and E
2
p
nding
traces of both sides. If we introduce notations:
s
1
= trT
K
, s
2
= trE
p
T
K
, s
3
= trE
2
p
T
K
,
s
4
= trT
2
, s
5
= trE
p
T
2
K
, s
6
= trE
2
p
T
2
K
, (49)
s
7
= trH
2
, s
8
= trE
p
H
2
, s
9
= trE
2
p
H
2
,
then such a procedure will produce relationships between c
1
, ..., c
6
and
1
, ...,
6
.
Of course, the same procedure applied to (44) as well as (48) would connect
sets c
7
, ..., c
9
and
7
, ...,
9
.
Similarly, the evolution equations for plastic strain rate and residual mag-
netization rate are explicitly stated by the following formulae:
D
t
E
p
=< f >
_
d
1
1 +d
2

E
e
+d
3
(

E
e
E
p
+E
p

E
e
) +
d
4
(

E
e
E
2
p
+E
2
p

E
e
) +d
5
E
p
+d
6
E
2
p
+
d
7
(M
r
E
p
E
p
M
r
) +d
8
(M
r
E
2
p
E
2
p
M
r
) +
d
9
(M
R
E
p
E
p
M
R
) +d
10
(M
R
E
2
p
E
2
p
M
R
) +
d
11
(E
p
M
r
E
2
p
E
2
p
M
r
E
p
) +
d
12
(E
p
M
R
E
2
p
E
2
p
M
R
E
p
)
_
, (50)
D
t
M
R
= e
1
M
r
+e
2
(M
r
E
p
+E
p
M
r
) +e
3
_
M
r
E
2
p
+
E
2
p
M
r
_
+e
4
M
R
+e
5
(M
R
E
p
+E
p
M
R
) +
e
6
(M
R
E
2
p
+E
2
p
M
R
) +e
7
(E
p

E
e


E
e
E
p
) +
e
8
(E
2
p

E
e


E
e
E
2
p
) +e
9
(E
2
p

E
e
E
p
E
p

E
e
E
2
p
). (51)
16
It should be noted here that all the scalar coecients in above relationships (40)-
(44) and (50)-(51) are functions of the principal invariants of the plastic strain
tensor E
p
. Of course, if plastic strain itself is small, then the corresponding
complete linearization of constitutive and evolution equations is straightforward
which might be of interest especially if dynamic eects are considered i.e. wave
equations of the linearized problem written (cf. [28]). Evolution equations then
would reduce to Onsager-Casimir reciprocity relations.
The above evolution equations (50) and (51) become very much simplied
if a generalized loading function is assumed. Such a function would have the
consecutive orthogonality properties (cf. [30, 37])
(52) D
t
E
p
= D
t

T
K
and D
t
M
R
= D
t

H
.
where the material time rate of a scalar function vanishes if the yield func-
tions f as well as f
#
are either negative or zero (cf. (29)). These constitutive
equations (52) include magneto-mechanical reversible and irreversible interac-
tions. To underline that the evolution equations are of quasi rate independent
type we use here D
t
instead of [40, 41]. A development of into a power
series whose terms are products of invariants (49) would give rise to dierent
magnetomechanical irreversible couplings. Such a procedure has been applied
in [41] for thermomechanical processes. It is worthy of note that even in the
simplest case when loading function is quadratic function of H and T:
(53) =
1
2

1
s
2
1
+
1
2

2
s
4
+
1
2

3
s
7
,
such a coupling may appear through D
t
. This function leads by means of (52)
into the consecutive two evolution equations
(54) D
t
E
p
= D
t
(
1
1 trT
K
+
2
T
K
),
(55) D
t
M
R
= D
t

3
H,
whose simplicity follows from the above very special loading scalar function .
They are oversimplied as will be seen from the following considerations.
3.3 Magnetoelasticity of cubic crystals by tensor representation
In order to show simplest magnetomechanical interaction let us consider in this
subsection magnetoelasticity of iron crystals. If we neglect the exchange energy,
then the potential energy for a magnetic domain 1, . . . , 6 inside a single
crystal grain reads (cf. [31], p. 378):
w

=
1
2
e
E
De
E
+M
2
s
e
E
Bg

0
M
s

+
1
4
M
4
s
g

K
1
g

+
1
4
M
6
s
g

(g

K
2
g

)
(56)
17
with the notation: g

(cf. (38)). Following [11] it has been assumed
here that strain e
E
and local magnetic eld

H are uniform for the considered
grain. Strictly speaking, all the terms on the right hand side of (56) except

0
, M
s
and

H should have subscript to show grain dependence. For simplicity

all these subscripts are not written. In the above equation the rst term on
the right hand side is the magnetoelastic energy w

, the second term shows the

magneto-strictive energy w

, magneto-static energy w
mag

is the next term,

whereas the last two terms are responsible for the magneto-crystalline energy
w
an

. If we subtract the magnetostrictive strain from magnetoelastic strain by

means of e

e
e
E
e

(cf. also (38)) and take into account that square of

magnetostrictive strain is negligible (cf. also Appendix in [11]), then we get:
(57) w

+w
mag

tr(

)
0
M
s

.
For cubic crystals magnetocrystalline 4-tensor K
1
and 6-tensor K
2
in the
crystallographic frame (CF) have the special forms [31]:
(K
1
)

= K
1

,
(K
2
)

= K
2

3
.
(58)
The base of (CF) is determined by the easy directions [100], [010] and [001]
while unit vectors

, 1, . . . , 6 are either parallel or antiparallel to this
base.
For the sake of illustration the next gure shows the energy (56) for iron as
function of direction in the mentioned CF (cf. also [10]).
An issue of high importance for polycrystals is concerned with volume frac-
tions of magnetisations inside magnetic sub-domains. Throughout a typical
cubic -grain it is assumed along with [6, 10] that they full the Boltzmann
distribution:
(59) f

=
exp(A
D
)w

exp(A
D
)w

, , 1, . . . , 6.
It is important to note here that the rst and last two terms in (56) are constant
throughout the -grain. Therefore, only energies (57) changing from domain
to domain actually appear in (59).
The coecient A
D
was calibrated for iron by Daniel in his PhD-thesis [10] on
the basis of measurements by Webster performed in 1925. Such an identication
means that the coecient A
D
depends either on experiment or material or
both
11
. Having found these volume fractions magnetisation vector inside -
11
As explained by Sommerfeld in [48] for small body immersed into a huge closed thermo-
18
Figure 2: Magneto-crystalline energy of iron
grain reads:
(60)

M

:=
6

=1
f

,
where constituents

M

are directed along the easy magnetic directions

,
namely, [100], [010], [001], [100], [010] and [001].
In order to see how multiaxial stress inuences magnetic susceptibility ten-
sor let us apply (59) and (60) to a grain of iron cubic crystal acted upon by
arbitrarily oriented external magnetic eld inside CF-plane [100] & [010]

H = H
0
_
cos(
H
) sin(
H
) 0
_
with
H
[0, /4] while H
0
[20, 2000]A/m.
To choose typical and representative stress histories let us remind that only
deviatoric stress inuences magnetostriction. Then the following two cases are
worthwhile to be considered.
dynamic system the coecient A
D
= 1/kT where k = 1.3806505 10
23
[J/K] is Boltzmann
universal constant and T the absolute temperature. It is important to underline that this
coecient does not depend on process and energies. Such an approach was applied in the
paper [33] where eld equations of Zorskis statistical theory of dislocations [57] were closed
by Boltzmanns partition function in a very special case of 2D screw dislocations. Since a
magnetic domain is much smaller than RVE a good preliminary choice would be to accept
Sommerfelds suggestion. The essential dierence between these two approaches is the ques-
tion of universality of A
D
. Final judgement must be drawn from experiments.
19
Case 1. Let the stress tensor acting on the grain be
=
0
_
_
cos
2
(

) sin(

)cos(

) 0
sin(

)cos(

) sin
2
(

) 0
0 0 1
_
_
with

[0, /4] while stress magnitudes are varied in the interval

0

[0, 160]MPa. Thus its compressive component is along CF-[001] direction
while the tension component is arbitrarily oriented inside plane [100] &
[010].
Case 2. Let the stress tensor be situated in the same CF-plane as

H i.e.
=
0
_
_
cos(2

) sin(2

) 0
sin(2

) cos(2

) 0
0 0 0
_
_
with

[0, /4] and

0
[0, 160]MPa. Since, in general,

,=
H
di-
rections of external magnetic eld vector and major principal stress dier
in general. In this case both principal stresses (tension and compression)
are also arbitrarily oriented inside plane [100] & [010].
From the numerical results of the above two programs of magnetomechanical
histories with calculated values for

M at this point we propose an identication
of stress and magnetic eld dependent susceptibilities. Such a formula could be
useful for fast nding inuence of stress on susceptibilities in each grain when
magnetic constitutive equation for magnetization of a polycrystal is analyzed.
The fact that reversible magnetization depends on stress and magnetic eld
and it disappears with external magnetic eld suggests tensor generators in the
tensor representation formula (cf. [49]):
(61)

M =

H

.
Thus, symmetric and skew-symmetric tensor generators are formed by products
of the tensors 1, and

H

H. As a result we obtain
(62)

M =
_
1
_

1
+
2
H
2
+
3

H
_
+
_

4
+
5
H
2
_
_

H.
Such a reduced representation is so chosen to have the smallest number of
material constants
1
, . . .
5
. It has odd powers of magnetic eld in order to
maintain magnetic symmetry requirement [31, 46]. Its linearity in stress is the
simplest approach taking into account magnetomechanical interaction. When
stress disappears it fulls symmetry requirement for cubic crystal.
20
Figure 3: Principal magnetic susceptibilities
1
/1000, 1000
2
,
3
, 1000
4
and

5
of mono-iron for diverse magnetic eld H[A/m] and biaxial stress magnitude

0
[MPa]. Out-of-plane case.
Input data for the numerical simulation are taken here for iron from [9, 11]
as follows:
[100]
= 21 10
6
- magnetostriction constant for the easy direc-
tion,
0
= 4 10
7
- vacuum magnetic permeability,
0
= 2000 - initial
magnetic susceptibility determined in Websters experiments (cf. [11]) and
A
D
= 1.6 10
3
[m
3
J

1] is BuironDaniel constant appearing in distribution

(59). From these data the corresponding saturated magnetization equals to:
M
s
= (3
0
/
0
A
D
)
1/2
[A/m]. Calibration results for both above cases are shown
on the following two gures. Each (H
max
,
0
) point encompasses tting for all
-coecients including all the points inside the 4-domain: H [0, H
max
],
H

[0, /4], [
0
,
0
],

[0, /4]. Obviously, the proposed approximation

(62) is satisfactory in domains where -coecients are only slightly variable.
In order to present the results for -coecients in a more explicit way let
us introduce scaling coecients
H
H/H
max
and

/
0
. For simplicity,
consider special case
H
=

. Then dependence of correlation coecient

21
Figure 4: Principal magnetic susceptibilities
1
/1000, 1000
2
,
3
, 1000
4
and

5
of mono-iron for diverse magnetic eld H[A/m] and biaxial stress magnitude

0
[MPa]. In-plane case.
22
of magnetization calculated by (59) and (60) as well as by (62) on scaling coe-
cient can be determined. As an example let us take = 0.5 corresponding to
H = 1000A/m and
0
= 80MPa. For in-plane case calibrated constants read
= 1726, 5.5710
4
, 6.9710
4
, 803, 1.2110
3
, = 0.944 whereas for
out-of-plane stress we have = 1661, 7.43 10
4
, 1.12 10
4
, 2250, 1.55
10
3
, = 0.896. The later correlation is not acceptable whereas the rst is
much better. From these two cases it may be concluded that the proposed
tensor representation (62) is more suitable for in-plane case
12
. The special
stress-free case is easily observed from the gures 3 and 4.
At the end of this subsection let us remark that the formula (62) might
be further improved introducing Boehlers structural tensors for cubic crystals
[4] at the expense of additional susceptibility coecients
13
. Of course, such a
result is physically more justied due to correct material symmetries.
Moreover in a polycrystal each grain orientation either is random or depends
on orientation function originating from texture. A comparison of simplied
representation (62) with the corresponding cubic crystal representation for full
randomness is a worthwhile task.
It should be mentioned that Motogi and Maugin in [44] considered sub-
domain distribution from convexity property of stocked energy g
p
(cf. (63))
without introducing micro-distribution (59).
3.4 Approach by endochronic thermodynamics
1. Let us rst briey discuss purely mechanical inelastic irreversible behavior
of steels given in [38]. The specic free energy of the considered body is taken
to be of the form
(63) g = g
E
(E
E
, T) +g
p
(, T) ,
where is the isotropic hardening parameter. Its time rate is given by
(64) D
t
:= T
K
: D
t
E
p
,
having the meaning of plastic power. Since the free energy is assumed in the
form (63) we have the plastic part of dissipation

p
= (1

g) D
t
.
The total thermoplastic dissipation appearing in the second law of thermody-
namics is denoted by , namely T
_
D
t
s + div(q/T)
_
0, where is the
12
A further improvement of the results for the susceptibilities could be obtained if rotations
of domain magnetisations from

H towards easy directions are calculated according to [7].
13
Even the isotropic representation (62) predicts induced magnetic anisotropy i.e. dierent
orientations of

H and

M (cf. [9] section 13.) unless stress disappears.
23
mass density, T is the absolute temperature, q is the heat ux vector and s is
the specic entropy.
The plastic dissipation served Vakulenko to introduce his thermodynamic
time [53] by the next hereditary function
(65) (t) :=
t
_
0

p
(t

)
_
dt

.
The function (t) is piecewise continuous and nondecreasing in the way that
D
t
(t) = 0 within elastic ranges and D
t
(t) > 0 when plastic deformation takes
place. Splitting the whole time history into a sequence of innitesimal segments
Vakulenko represented the plastic strain tensor as a functional of stress and
stress rate history.
Moreover, in the paper [38] the accumulated plastic strain
(66)
p
eq
()
_

0
|D
t
E
p
()| d,
as the important inelastic history parameter, was included into the memory
kernel, extending in such a way formerly mentioned Vakulenkos arguments.
Another important generalization of his model in [38] was extension of the
function to have the nonlinear power form:
(67) (
p
) = (
p
)
a
.
The exponent a is of a great importance since it shows the speed of ageing. For
example a < 1 may be named decelarated ageing whereas a > 1 would dene
accelarated ageing. By such a classication the Vakulenkos value a = 1 might
be termed as steady ageing.
Now, according to Vakulenkos postulate we have:
(68) E
p
() =

_
0

_
, T
K
(), D

T
K
(),
p
eq
()

d.
Of course, this integral equation is adopted to our case of nite plastic strains
and absence of plastic rotation. Dierentiation of (68) with respect to the
thermodynamic time gives

E
p
=
_
0, T
K
(), D

T
K
(),
p
eq
()

_
0

_
, T
K
(), D

T
K
(),
p
eq
()

d. (69)
24
Further analysis of the above integral equation is given in the next subsection.
2. Let us apply now the above explained concept to evolution of irreversible
magnetization. Again we have non-steady ageing speed dened by the exponent
a by means of:
(70) D
t
= (
PM
)
a

HD
t

M
R
+tr(T
K
D
t
E
p
)
_
a
,
but now irreversible power induced by magnetisation must be taken into ac-
count. It is included into the magnetoplastic dissipation
PM
. Suppose now
that magnetomechanical interaction is only through equivalent plastic strain
history. Then the magnetic evolution equation in its integral form may be
taken as
(71)

M
R
() :=

_
0

(
p
eq
(z), z,

H(z)) dz,
where the corresponding endochronic memory is characterized by the thermo-
dynamic time (70). Choosing a special form of the integral kernel as follows
14
(72)

=

H(z) (
p
eq
) exp ( z)
we would arrive at the following simple explicit evolution equation for residual
magnetization vector:
(73) D

M
R
=

H

M
R
.
However, in this equation derivative is taken not with respect to real but to
thermodynamic time. In order to transform it to real time let us rst introduce
irreversible magnetic power by means of
(74) D
t

:=

HD
t

M
R
,
following the same notation in (64). Now, when we replace (74) into (70) and
multiply this by magnetic eld vector and time derivative of thermodynamic
time we get a nonlinear algebraic equation:
(75)
_
D
t

H[
2

H

M
R
_ a
1a
D
t

:= D
t
.
This equation explicitly characterizes magnetoplastic interaction. Its validity
should be checked by experiments where simultaneously stress, plastic strain,
magnetic eld and residual magnetization are measured. The two interesting
special cases may be drawn from this equation:
14
Such an exponential kernel is typical for endochronic theories (cf. [54]).
25
If plastic power is approximately equal to zero, then we would have a ther-
moelastic irreversible magnetization. Since D
t
0 the equation (75)
gives a simplied time rate of the thermodynamic time:
(76) D
t
= (D
t

)
a
=
_
[

H[
2

H

M
R
_ a
1a
.
Another special case of interest would be choice a = 1 which might be called
Vakulenkos coupled magneto-viscoplasticity. For such a choice time rate
of reads:
(77) D
t
= D
t

+D
t
=
D
t

1 [

H[
2
+

H

M
R
.
In both cases evolution equation for residual magnetization in real time domain
has the form:
(78) D
t

M
R
=
_

H

M
R
_
D
t
.
Suppose for simplicity that constitutive equation (62) holds not only for a
grain but for RVE as well. Then

M
r
=

H and using (43) we arrive at the

integral evolution equation connecting magnetic induction and magnetic eld
vectors:

B() =

H() +J()
1
C()
_
()

H()+

_
0

(
p
eq
(z), z,

H(z)) dz
_
,
(79)
where the explicit form of the right Cauchy-Green total deformation tensor
15
C = 1 + 2E is found from the relationship (39). Here magnetomechanical in-
teraction appears through total as well as plastic strain history and fact that
susceptibilities depend on stress. Obviously, the oversimplied equation (55)
might hold only in the case of negligible . We believe that the above integral
equation could be used for some nondestructive experimental check of order of
magnitude of magnetomechanical interactions at low cycle fatigue or at some
other experiments designed to establish characteristic points of inelastic behav-
ior of steels or some other ferromagnetic materials.
15
Terminology is taken from [52] Sec. 24.
26
3.5 Low-cycle fatigue of ferromagnetics
The constitutive and evolution equations described in previous subsections
might be used to describe piezomagnetic behavior induced by low-cycle fatigue
of ferromagnetics.
Such a process has been investigated in the paper [13]. A cylindrical speci-
men of AISI 1018 was uniaxially treated by push-pull tests on MTS-810 servo-
hydraulic testing machine such that total strain was periodic and triangularly
shaped |E| 0, 0.009 with cycle duration of 2 s. Magnetic induction due
to piezomagnetic eect was also almost periodic with very slight changes of
periodicity with increase of relative number of cycles N/N
f
and cumulation of
phase delay with respect to strain with growth of accumulated plastic strain.
Maxima and minima of total Lagrangean strain E are displaced with respect
to minima and maxima of the magnetic induction vector
16
. According to (66)
we calculate accumulated plastic strain as a function of time by means of
(66)
p
eq
(t) :=
_
t
0
|D
t
E
p
()| d.
Now, if uniaxial components of tensors E,

B,

M
r
,

M
R
are denoted by means
of E
11
, B
1
, M
r1
, M
R1
then the following memory-type equation emerging from
(55) and (71)
(80) B
1
(t) :=
_
t
0
J(
p
eq
, t ) D

() H
1
() d,
could cover the delay between measured functions E
11
(t) and B
1
(t). Time
dierentiation of the above relationship gives rise to the expression:
D
t
B
1
(t) := J(
p
eq
, 0) D
t
(t)H
1
(t)+
_
t
0

t
J(
p
eq
, t ) D

() H
1
() d.
(81)
In the above integro-dierential equation the second term on the right hand side
is responsible for the above mentioned change of time delay and the deection
of pure periodicity of B
1
(t). Therefore, it is much smaller than the rst part.
On the other hand, if the constitutive equation

B =
0
_
(1 +)

H +

M
R
_
(by
means of (62)) is used, then we would have
D
t
B
1
=
0
D
t
M
R1
+
0
D
t
_
(1 +
11
)H
1
_
,
D
t
E
11
= D
t
E
e11
+D
t
E
11
+D
t
E
p11
.
(82)
16
Here for convenience again magnetic induction is represented by the vector

B.
27
Now, the relationships (77) and (64) in our case lead to:
(83) D
t
M
R1
=
_
H
1
M
R1
_

11
D
t
E
p11
1 H
2
1
+H
1
M
R1
.
The above three equations show a clear magnetomechanical interaction. They
may serve for identication of material functions from LCF uniaxial tension
experiments (like those in [13]). Simultaneous zeros of D
t
E
p
and D
t

M
R
(en-
suing from (54) and (55)) are consequence of the model where it is assumed
that irreversible magnetisation and plastic strain are triggered at the same time
instant.
Instead of using (82) and (83), it is possible to nd a more explicit en-
dochronic kernel in (81). Let us remind that the Langevin function is suggested
in many references as the best approximant for anhysteretic curve
17
(cf. for in-
stance [43]). Then using the data in (18.137) and gure 22.1 from [9] for such
cubic crystals
18
it is possible to depict the consecutive gure by translating
the anhysteretic curve by coercive magnetic eld H
c
either to the right or to
the left depending on sign of time rate of magnetic eld. Dropping indices for
simplicity this may be represented by the following kernel
(, z) = M
0
_

_
H(z) H() +H
c
()sgn(D
t
H)
_

_
H(z) H()
_
_
L
_
H(z)
_
D
z
H,
(84)
0.06 0.04 0.02 0 0.02 0.04 0.06
2
1.5
1
0.5
0
0.5
1
1.5
2
H [T]
M [T]
H
c
M
R
0
Anhysteretic curve
Figure 5: Soft ferromagnetic steel behavior approximated by Langevin function
according to [9]
17
Let us recall that this function has the form L(H) = coth(H) 1/H.
18
In the case of a soft ferromagnetic steel this author suggests the following data: (M)|
H=0
=
0.832M
sat
, (dM/dH)|
max
= 0.6M
sat
/H
c
where H
c
= 0.0063T is the coercive magnetic eld
and M
sat
= 2.15T is the value of saturated magnetization value.
28
where sgn(x) = 1 for x > 0 and sgn(x) = 1 for x < 0. If this kernel is
inserted into the integral equation (71) then it gives rise to the following explicit
expression for residual magnetization:
M
R
= M M
r
=
M
0
L
_
H H
c
sgn(D
t
H)
_
M
0
L(H).
(85)
It is worthy of note that herein the magneto-mechanical interaction is taken
into account by dependence of coercive eld on plastic power whose time rate
is given by (64).
Turning again to the paper [13] it may be concluded that the dependence
H
c
() and (85) permit taking into account complex disturbances of shape from
simple periodicity of E
11
(t) towards more complicated shape of B
1
(t) as well as
their relative delays of minima and maxima. It may be thus concluded that such
an equation could be fruitful for description of magneto-viscoplastic phenomena
at low cycle fatigue.
4 Eective eld method for ferromagnetic polycrys-
tals
From this point we will assume that in the relationship (39) all the strains are
small such that an approximate additivity holds. Then Eulerian strains full
the following relation:
(86) e = e
e
+e

+e

+e
p
e
e
+e
p
,
with: 2e

= 1 F
T

F
1

, e, , , p. Here the term e

p
is the inelastic
eigen-strain (with thermal, magnetostrictive and plastic parts) representing
a source of internal stresses (cf. [18]). Suppose that a RVE is composed of
many grains which may be modeled by randomly oriented ellipsoids. If such
a RVE has homogeneous elastic strain, magnetization and temperature then it
is customary for the eective eld method (cf. [18]) to consider such a grain
as an inclusion implanted (by eigen-strains) into a hypothetical matrix with
average properties over grain orientations (with notation
1
N
g

)):
(87) D
0
= D

), L
0
= L

),
0
=

),
Two main hypotheses of eective eld method (according to [23, 18]) are:
Hypothesis 4.1 Inside each ellipsoidal inclusion strain is homogeneous.
Hypothesis 4.2 Ergodic property holds i.e. properties of inclusions are sta-
tistically independent of their spatial distribution.
29
The rst hypothesis in [27] is named step-constant approximation. If these
two hypotheses are fullled and the RVE is acted upon by some external
stress
0
, then performing the procedure applied in [24, 23] the stress and
strain inside a -grain become:
(88)

=
0
+

S
0

_
[[D
1

]]

+ [[e
p

]]
_

,
e

= e
0

K
0

_
[[D

]]e
e
D
0
[[e
p

]]
_

,
where (A a)(x) :=
_
A(x x

)a(x

)dx

, A K
0
, S
0
, [[a

]] a

a
0
is
the jump of a across -grain boundary and

is the characteristic function of

-grain being unity when position vector points to -grain and zero otherwise.
The corresponding characteristic function for whole RVE is obtained by sum-
mation of characteristic functions of all the grains belonging to the considered
RVE. The kernel S
0
was determined in [18]:
(89) S
0
(x) = D
0
K
0
(x)D
0
D
0
(x)
by means of the kernel (in Kunins notation [23]):
(90) K
0
= defG
0
def
built by the Green function G
0
corresponding to stiness D
0
(cf. (87)) and
def

a sym(

a ).
For simplicity, let us concentrate our attention to the special case of magne-
toelastic strain leaving out plastic and thermal strains. Then eective stiness
and eective magnetostriction tensor are dened by means of
(91)

) = D
e
e
e
), e

) = L
e
g

),
Here volume averages give macroscopic elastic strain, macroscopic stress, macro-
scopic magnetostrictive strain and magnetization dyadic:
e
e
= e
e
) , T =

) , e

= e

) , g = g

) ,
respectively.
Even in this linear case without thermal and plastic strains for ellipsoids
randomly oriented with crystallographic frames inside them incoincident with
their semi-axes solution of the set of coupled integral equations (88) towards
homogenization formulae (91) is tremendous. For this reason some reasonable
simplications are inevitable.
In homogenization theories for composites with particulate inclusions there
are two distinct self-consistent approaches:
30
eective medium approach where it is assumed that each inclusion behaves
as isolated and immersed into a medium having eective constants D
e
and
eective eld approach with an assumption that again each inclusion be-
haves approximately as isolated and situated into the matrix with elas-
ticity constants D
M
while inuence of neighboring inclusions is taken into
account by means of the eective strain eld e
e
acting on the considered
inclusion [27].
In this paper the second approach is employed. The mentioned eective strain
in the case of pure elastic strain covers in the second of (88) all the terms in
the sum with ,= under the assumption that correlation eld induced by all
the other inclusions on -grain is of the same shape as -grain itself but with
larger dimensions. This assumption is simply illustrated by the gure 6. Under
(RVE)
e
0
e, A
,A e
eff F
Figure 6: Eective self consistent eld homogenization
such assumption the authors in [19] found (I is unit 4-tensor):
(92) e
e
(x) = (I p

[[D

]]M

))
1
e
0
,
with M

= (I+A(x

)[[D

]])
1
, A(x) =
_
K
0
(xx

)dx

and A

=
_
K
0
(x)(x)dx.
The function (x) describes correlation distribution inside RVE. Having found
e
e
the average strain equals to
(93) e

) = M

e
e
).
Substituting this expression into (88) leads to the eective stiness 4-tensor:
(94) D
e
= D
0
+

[[D

]]
_
I +
_
A(x

) p

_
[[D

]]
_
1
.
31
In the special case when all the inclusions have parallel semi-axes coinciding
with their crystallographic frames we have A(x

) = A

( 1, N
g
) and
the above relationship takes the form of Mori-Tanaka eective stiness: D
e
=
D
0
+

[[D

]] (I + (1 p

)A[[D

]])
1
.
Suppose now that throughout a RVE magnetostrictive strains are balanced
in the way that average stress induced magnetostrictive grain strains equals to
zero. Then by making use of the ergodic hypothesis 4.2 we obtain
(95) L
T
e

0
L
T
0

0
L
T

) = 0,
where (L
T
)
cdab
= (L)
abcd
. Due to linearity of (88) we may introduce the stress
concentration 4-tensor by means of the substitution

= N

0
. Knowing this
stress concentration tensor we nd the eective magnetostriction 4-tensor by:
(96) L
T
e
= L
T
0
+ [[L
T

]] N

).
This concentration tensor is found from the reduced version of rst of (88) when
inelastic terms are dropped, i.e.
(97) N

= I +

S
0
[[D
1

]]N

.
Omitting details of the derivation we just quote result for the tensor L
T
e
derived
in the same way as eective thermal expansion tensor in [24]:
(98) L
T
e
= L
T
0
+ B

)
1

[[L
T

]] )
with B

=
_
I + D
0
_
I A(x

)D
0
_
[[D
1

]]
_
1
. Let us nally quote the eective
thermal expansion tensor derived by Levin in [24]:
(99)
e
=
0
+ B

)
1

[[

]] ).
In this way we have completed all the necessary eective constitutive ten-
sors entering Hookes law for RVE based on microstructural thermo-magneto-
mechanical properties of individual grains. It is noteworthy that eective mag-
netostriction as well as eective thermal expansion tensors are derived from
pure elasticity consideration. According to Taylors assumption plastic strain
is assumed to be homogeneous throughout RVE being equal for all the grains.
19
19
Let us note that the eective eld approach to eective susceptibilities by means of grain
based constitutive equation is more complicated than the above analysis due to essential
nonlinearity of (62). It seems that a variational approach like in [55] is more suitable. Anyway,
if we suppose that

H

, then the Wiener upper bound dependent on RVE-average of

stress could be obtained (cf. also [11]).
32
5 Some concluding comments
The subject has been treated by tensor representation applying either non-
associativity of ow rule with extended thermodynamics or generalized normal-
ity which includes orthogonality of residual magnetization rate on generalized
loading surface which includes mechanical as well as magnetic state variables.
While plastic strain is nite thermoelastomagnetostrictive strain is assumed
to be small. Small magneto-elasto-viscoplastic strains are then considered in
detail in order to analyze magnetomechanical interaction at low-cycle fatigue.
Furthermore, endochronic thermodynamics with Vakulenkos thermodynamic
time made possible an account to (experimentally observed) time delay between
stress and magnetic eld histories. Such a result could be useful in inelastic
testing with magnetic elds either induced or applied.
Geometrical approach based on early papers in the eld of continuum theory
of dislocations leads to the essential dierence between micro and macro-spin
having as the origin constrained micro-rotations of grains inside a representative
volume element. Here an Eshelbian approach is applied assuming that quasi-
plastic (thermomagnetoplastic) strain is unconstrained whereas elastic strain is
constrained. Since a RVE, having volume of an innitesimal volume element,
cannot be disintegrated any more, micro-spin does not follow from plastic micro-
stretching. In previous papers of the author (reviewed shortly in [41]) treating
purely thermomechanical strain histories of viscoplastic polycrystals such an
idea has been proved to be very successful. Namely, purely elastic micro-strains
have been assumed to be covered by self consistent method (eective medium
or eective eld approach by Levin) whereas for plastic stretching as well as
residual magnetization rate quasi rate independent incremental macro-evolution
equations are postulated. The rate dependence takes place by means of stress
rate dependent value of the initial yield stress. The macroscopic magneto-
inelastic evolution equations obey the Vakulenkos concept of thermodynamic
time. The macro-evolution equation for plastic spin of RVE is an outcome of
the corresponding evolution equation for plastic stretching. The same does not
hold true for plastic micro-spin.
This paper has been dealt with viscoplasticity of ferromagnetic materials.
Evolution equations have been derived either from inelastic materials of dier-
ential type or from loading function generalized normality. In both cases tensor
representation is applied to such a set of evolution equations. Restrictions to
the set of eld equations are established by means of the extended irreversible
thermodynamics (version which follows exposition in [8]). Small magnetoelas-
tic strains of isotropic insulators are considered in detail in two special cases
of nite as well as small plastic strain. As one example low-cycle fatigue of
ferromagnetics is considered with special account to time delay between stress
and magnetic eld histories. To describe such an experimental evidence an
33
integro-dierential equation is proposed whose equivalent plastic strain depen-
dent kernel covers the observed delay.
Concluding this section it is inevitable to compare the foregoing results with
existing achievements in the eld. The major contributions to viscoplasticity
of ferromagnetic materials have been given by Maugin and his collaborators in
[30, 31]. The principal assumptions accepted in this section are closer to the
scope of the rst of these references where
small strain case together with absence of exchange forces and gyromag-
netic eects has been assumed;
the accent on hysteresis eects has been given and
evolution equations have been derived by normality of plastic strain rate
and residual magnetization rate onto a loading surface.
On the other hand we presented here the following results:
in the case of nite plastic strains magnetic anisotropy induced by plastic
strain is predicted by (51) where development of residual magnetization
by mechanical terms is also evident;
the inuence of magnetization on plastic strain rate is obtained even in
the case of isotropic ferromagnetic materials;
the extended thermodynamics procedure allows for more general history
eects with inhomogeneities of magnetization taken into account;
the obtained relationships with couplings allow for magnetic measure-
ments of inelastic phenomena but the measurements will show their order
of magnitude and practical measurability of these phenomena;
in general, the developed theory is of non-associate type for plastic strain
rate and residual magnetization rate are not perpendicular to the yield
surface;
albeit a generalized normality is much simpler with smaller number of
material constants, a careful examination of the experiments on piezo-
magnetism and magnetostriction processes would give the nal judgement
which theory should be applied;
endochronic thermodynamical approach is less general than the approach
by extended thermodynamics but it is much more suitable for explicit
description and calibration of inelastic magneto-mechanical experiments
like low cycle fatigue stress-strain-induction histories.
34
Acknowledgment
The author is grateful to Professors Gerard Maugin and Vukota Babovic for
valuable discussions concerning some aspects of electromagnetism.
Reviewers criticism has helped very much to clarify and organize the text
of this paper.
References
[1] Albertini,C., Montagnani,M. and Micunovic,M., (1989) Viscoplastic Behavior of
AISI 316H - Multiaxial Experimental Results, In: Transactions of SMIRT-10, ed. Had-
jian, Los Angeles, 31-36.
[2] Anthony, K.-H. (1970) Nonmetric connexions, quasidislocations and quasidisclinations.
A contribution to the theory of nonmechanical stresses in crystals, Fundamental Aspects
of Dislocation Theory, Vol. I, National Bureau of Standards Special Publication, 317,
(eds. Simmons, J. A., de Wit, R. & Bullough R.), pp. 637-649, Washington, D.C.
[3] Bilby, E. (1960). Continuous distributions of dislocations, Progress in Solid Mechanics,
1, 329398.
[4] Boehler, J-P. (1979) A simple derivation of representations for non-polynomial consti-
tutive equations in some cases of anisotropy, ZAMM, 59, 157167.
[5] Brown, W. F. Jr. (1966) Magnetoelastic Interactions, Springer, Berlin.
[6] Buiron, N., Hirsinger, L. and Billardon, R. (1999) A micro-macro model for magne-
tostriction and stress eect on magnetisation, J. Magn. Magn. Mats, 196-197, 868870.
[7] Buiron, N., Hirsinger, L. and Billardon, R. (2001) A multiscale model of magne-
tostriction strain and stress eect, J. Magn. Magn. Mats, 226-230, 10021004.
[8] Casas-Vazquez, J., Lebon, G. & Jou, D. (2001) Extended Irreversible Thermodynam-
ics, Berlin: Springer.
[9] Chikazumi, S. (1984) Physics of Ferromagnetism, Vol. 2, Shokobo, Tokyo. (Translation
into Russian, 1987, Mir Publ., Moscow)
[10] Daniel, L. (2003) Modelisation multi-echelle du comportement magneto-mecanique des
materiaux ferromagnetiques textures, PhD thesis, Ecole Normale Superieure de Cachan,
France.
[11] Daniel, L., Buiron, N., Hubert, O. and Billardon, R., (2005) Reversible magneto-
elastic behavior Parts I & II, J. Mech. Phys. Solids, submitted .
[12] Dashner, P. A. (1986). Invariance considerations in large strain elastoplasticity, J.
Appl. Mech., 53 ,5560.
[13] Erber, T., Guralnick, S. A., Desai, R. D. and Kwok,W., (1997) Piezomagnetism
and Fatigue, J. Phys. D: Appl. Phys., 30, 2818-2836.
[14] Eshelby, J. D. (1957). The determination of the elastic eld of an ellipsoidal inclusion,
and related problems, Proc. Roy. Society, 241, 376396.
[15] Faehnle, M., Furthmuehler, J. and Pawellek. R., (1991) Continuum Models of
Amorphous and Polycrystalline Ferromagnets: Magnetostriction and Internal Stresses,
In: Proc. Symp. Continuum Models and Discrete Systems (CMDS-6), Vol.2., ed. G.
Maugin, Longman, Harlow, 120-126.
[16] Gil Sevilano, J.F., Van Houtte, P. and Aernoudt, E.,(1981) Large Strain Work
Hardening and Textures, In: Progress in Material Science, eds. Chalmers et al, Pergamon,
Oxford
35
[17] Hill R. (1965). A self-consistent mechanics of composite materials, J. Mech. Phys.Solids,
13, 213222.
[18] Kanaun S. K. and Levin V. M. (1993). Eective Field Method in Mechanics of Com-
posites (in Russian), Petrozavodsk University Edition.
[19] Kanaun S. K. and Jeulin, D. (2001) J. Mech. Phys. Solids, 49, 2339-2367.
[20] Kondo, K., (1958). RAAG Memoirs, II(D), Tokyo: Gakujutsu.
[21] Kr oner E., (1960) Arch. Rational Mech. Anal.,4, 273.
[22] Kr oner E. (1970) Initial studies of a plasticity theory based upon statistical mechanics,
Inelastic Behavior of Solids, (ed. Kanninen, M. Adler, W. Rosenfeld A. & Jofee R.), pp.
137-148, New York: McGraw Hill.
[23] Kunin, I. A., (1983) Elastic Media with Microstructure, Berlin: Springer Series in Solid
State Sciences.
[24] Levin V. M. (1982). On thermoelastic stresses in composite media (in Russian), Appl.
Math. Mech. (PMM), 46/3, 502506.
[25] Liu, I.-S., (1972) Arch.Rational Mech.Anal. 46, 131.
[26] Makar, J. M. and Atherton, D. L., (1995) Eects of Isoeld Uniaxial Cyclic Stress
on the Magnetization of 2% Mn Pipeline Steel - Behavior on Minor Hysteresis Loops and
Small Major Hysteresis Loops, IEEE Trans. Magn., 31/3, 2220-2227.
[27] Markov, K.Z., (2001) Justication of an eective eld method in elasto-statics of het-
erogeneous solids J. Mech. Phys. Solids, 49, 2621-2634.
[28] Maruszewski, B. and Micunovic, M., (1989) Int. J. Engng. Sci. 27, 955.
[29] Maugin, G. A. and Fomethe, A., (1982) Int. J. Engng. Sci. 20, 885.
[30] Maugin, G. A., Sabir, M. and Chambon,P., (1987) Coupled Magnetomechanical Hys-
teresis Eects: Applications to Nondestructive Testing, In: Electromagnetic Interactions
in Deformable Solids and Structures, eds. Y.Yamamoto and K.Miya, North Holland,
Amsterdam.
[31] Maugin, G. A., (1988) Continuum Mechanics of Electromagnetic Solids, North Holland,
Amsterdam.
[32] Mi cunovi c, M. (1974). A geometrical treatment of thermoelasticity of simple inhomo-
geneous bodies : I - geometric and kinematic relations, Bull. Acad. Polon. Sci.,Ser. Sci.
Techn., 22/11, 579-588.
[33] Mi cunovi c, M. (1979). Kinetic stress tensor in statistical theory of screw parallel dislo-
cations, Theoret. Appl. Mech., 5, 91-96.
[34] Micunovic, M., (1987) A Thermodynamic Approach to Viscoplasticity, In: Proceedings
of the 1st National Congress on Mechanics, eds. A.Kounadis et al, HSTAM, Athens,
364-376.
[35] Micunovic. M., (1987) A Thermodynamic Viscoplasticity Theory of Ideal Fibre-
Reinforced Materials, In: Mechanical Behavior of Composites and Laminates, eds.
W.A.Green and M.Micunovic, Elsevier, London, 199.
[36] Mi cunovi c, M., Albertini, C. & Montagnani, M. (1997). High strain rate viscoplas-
ticity of AISI 316 stainless steel from tension and shear experiments, Solid mechanics,
(ed. Miljanic P.), Serbian Acad. Sci. Meetings - LXXXVII, Dept. Techn. Sci., 3, 97106.
[37] Mi cunovi c, M. (2001) On viscoplasticity of ferromagnetics, Theoret. Appl. Mech., 26,
107126.
36
[38] Mi cunovi c, M. (2002a) Some issues in polycrystal viscoplasticity of steels, Int. Symp.
Structured Media - in memory of E. Kroner, (ed. Maruszewski, B. T.), Poznan: Poznan
Univ. Technology Publ., 196-207,
[39] Mi cunovi c, M. (2002b) Low order inelastic micromorphic polycrystals, Theoret. Appl.
Mech., 28-29, 235259.
[40] Mi cunovi c, M. (2005) Self-consistent method applied to quasi-rate dependent polycrys-
tals Phil. Magazine, 85 (33-35), 40314054.
[41] Mi cunovi c, M. (2006) Some Issues of Thermomechanics of Viscoplasticity AMMA -
III,(ed. Ogden, R. W. & Gao, D.),187348, in print, Berlin: Springer.
[42] M uller, I. (1971)Arch.Rational Mech.Anal. 41, 319.
[43] Motogi, S. & Fukuda, T. (1990) Modelling of E eect in polycrystalline ferromag-
nets, Mechanical Modellings of New Electromagnetic Materials, (ed. Hsieh R. K. T.),
Amsterdam: Elsevier. 281288.
[44] Motogi, S. & Maugin, G. A. (1993) Nonlinear stress-strain behavior of multi-domain
ferromagnets, Mechanics of Electromagnetic Materials and Structures, (eds. Lee, J. S.,
Maugin, G. A. & Shindo, Y.), ASME, AMD-Vol. 161 / MD-Vol. 42, 99114.
[45] Naerlovic-Veljkovi c, N. (1977). Field equations for micropolar current and a heat
conducting magnetically-saturated solid, Arch. Mech., 29/4, 625630.
[46] Novozhilov, Yu.V. and Yappa, (1981) Electrodynamics, Mir, Moscow.
[47] Ruuskanen, P., (1988) Magnetomechanical Eect in Polycrystalline Iron, In: Strength
of Metals and Alloys, eds. P.O.Kettunen et al, Pergamon, Oxford
[48] Sommerfeld, A., (1955) Thermodynamics and Statistical Physics, Academic Press,
New York.
[49] Spencer, A.J.M., (1971) Theory of Invariants, In: Continuum Physics, ed. A.C.Eringen,
Academic Press, New York.
[50] Stojanovi c, R. (1962). On the reference-state problem in the non-linear elasticity
theory of continua with dislocations, Physica Status Solidi, 2, 566575.
[51] Teodosiu, C., (1970). A dynamic theory of dislocations and its applications to the
theory of the elastic-plastic continuum, Fundamental Aspects of Dislocation Theory,
Vol. II, National Bureau of Standards Special Publication, 317, (ed. Simmons, J. A., de
Wit, R. & Bullough R.), pp. 837877, Washington, D.C.
[52] Truesdell, C. & W. Noll. (1965). The non-linear eld theories of mechanics, Hand-
buch der Physik, III/3, (ed. Fl ugge S.), Berlin: Springer.
[53] Vakulenko, A. A., (1970). Superposition in Continuum Rheology (in Russian), Izv.
AN SSSR Mekhanika Tverdogo Tela, 1, 69-74.
[54] Valanis, K. C., (1971). A theory of viscoplasticity without a yield surface, Archives of
Mechanics, 23, 517533.
[55] Willis, J. R., (1982) Elasticity theory of composites. In: Mechanics of Solids, eds.
Hopkins, H. G., Sewell, M. J., Pergamon, Oxford, 653-686.
[56] Zhou, S. A. and Hsieh, R. K. T.,(1988) Int.J.Engng.Sci. 26/1,13.
[57] Zorski, H. (1968) Statistical theory of dislocations, Int. J. Engng. Sci., 4, 959-974.
[58] Zorawski, M., (1974) A private communication.
37