Download as pdf or txt
Download as pdf or txt
You are on page 1of 63

Chapter VII

Thermoelasticity

K y Word

initial state.

J. Salençon, Handbook of Continuum Mechanics


© Springer-Verlag Berlin Heidelberg 2001
Chapter VII. Thermoelasticity 297

In Brief

Experienee shows thermoelasticity to be a eommon type of behaviour. It is


eharacterised by reversibility of the response of the material to the exeitation
it undergoes (Seet. 2).
The model for sueh behaviour, leading to the eorresponding eonstitutive
equations, is based on very general principles that apply to the behaviour of
any material, and feature in partieular the laws of eontinuum thermodynam-
ies. The first law yields the energy equation and the seeond, the fundamental
inequality (Seet. 3).
Thermoelastie behaviour is modelIed by assuming that eurrent values of
the temperature and strain tensor in the material element are sufficient to
define its physieal state. The fundamental inequality then provides the equa-
tions governing thermoelastie behaviour in the Lagrangian representation.
The free energy arises as the thermodynamic potential, a function of the eur-
rent values of the temperature and strain tensor. The stress tensor is obtained
by differentiating with respect to the strain tensor (Seet. 4).
When the deformation is infinitesimal and the temperature variations
small, this eonstitutive equation ean be linearised. This is the physieal lin-
earisation of the relation between stresses, strains and temperature varia-
tions. If in addition the transformation is infinitesimal, linearisation ean be
taken furt her. This is the geometrieal linearisation, leading to a linear rela-
tion between the Cauehy stress tensor, the linearised strain tensor and the
temperature variation (Seet. 5).
The linearised eonstitutive equation involves the elastic constants and the
thermal expansion coefficients. These are intrinsie physieal eharaeteristies of
the material, whose number ean be redueed by taking material symmetries
into aeeount. They also satisfy astability condition. In an isotropie material,
linear thermoelastieity is thus eharaeterised by just two elastic eonstants and
one thermal expansion eoeffieient (Seet. 5).
Chapter VII. Thermoelasticity 299

Main Notation

Notation Meaning First cited


E Internal energy of the system (3.1)
0

Q Heat input (3.1)


r(;!';., t) Internal heat source per unit volume (3.5)
e(;!';., t) Internal energy per unit mass (3.9)
9.(;!';.' t) Outward heat flux for "'t (3.14)
T(;!';., t) Absolute temperature (3.21 )
S Entropy of the system (3.21)
s(;!';., t) Entropy per unit mass (3.23)
1/J Free energy per unit mass (3.25)
iP Dissipation per unit volume (3.27)
9.o(X,t) Outward heat flux for "'0 (3.35)
1/J* Legendre-Fenchel transform of 1/J (4.20)
'Pp(~) = 0 , internal constraint (4.28)
7]p Lagrange multiplier (4.36)
for an internal constraint
I~, I~,I~ Invariants of ~ (4.42)
A Elasticity tensor (5.3)

ls. Tensor of thermal coefficients (5.3)


T Temperature variation (5.3)
>. Lame constant (5.12)
p" G Shear modulus (5.12)
E Young modulus (5.15)
1/ Poisson ratio (5.15)
a Linear thermal expansion coefficient (5.15)
K Bulk modulus (5.39)
300 Chapter VII. Thermoelasticity

1 From Experienee to a Constitutive Law . . . . . . . . . . . . . . . . . . . . . . . . . .. 301


2 Experimental Observations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 302
2.1 Generalities ............................................ 302
2.2 Simple Tension Test. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 303
2.3 Further Experimental Results. . . . . . . . . . . . . . . . . . . . . . . . . . . .. 304
2.4 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 305
3 Thermodynamies of Continuous Media. . . . . . . . . . . . . . . . . . . . . . . . . . .. 306
3.1 First Law: Energy Equation .............................. 306
3.2 Seeond Law: Fundamental Inequality . . . . . . . . . . . . . . . . . . . . .. 311
3.3 Lagrangian Expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 313
4 Thermoelastie Constitutive Laws. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 315
4.1 Elasticity Hypothesis .................................... 315
4.2 Uneonstrained Thermoelastieity ........................... 316
4.3 Thermoelastieity with Internal Constraints . . . . . . . . . . . . . . . .. 321
4.4 Material Symmetries .................................... 324
4.5 Isotropie Thermoelastie Material
in the Referenee Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . .. 325
4.6 Internal Constraints from the Eulerian Standpoint .......... 327
5 U neonstrained Linear Thermoelasticity ........................... 328
5.1 Introduction ............................................ 328
5.2 Physieal Linearisation ................................... 328
5.3 Isotropie Linear Thermoelastic Material. . . . . . . . . . . . . . . . . . .. 332
5.4 Infinitesimal Transformation and Geometrieal Linearisation .. 334
5.5 Stability of Thermoelastie Materials. . . . . . . . . . . . . . . . . . . . . .. 340
5.6 Typieal Values for Materials in Common Use ............... 343
5.7 Examples of Anisotropie Thermoelastie Materials ........... 343
6 Historical Perspeetive. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 346
Summary of Main Formulas ......................................... 349
Exereises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 351
1. From Experience to a Constitutive Law 301

Thermoelasticity

1. From Experience to a Constitutive Law

The thermoelastic behaviour of materials is revealed by everyday experience


in the form of linear or bulk expansions and contractions under the effect
of temperature changes, applications of the elastic properties of met als and
even polymers to the production of springs, pins, clips and the like. From a
phenomenological point of view, thermoelasticity is thereby linked to a not ion
of reversibility. The material responds to mechanical or thermal excitation in
an instantaneous manner and, when the excitation is removed, returns to its
initial state without showing any memory of the recent changes.
In the present chapter, our aim will be to formulate loeal relations be-
tween strain field, internal force fields and temperature field when the mate-
rial under consideration is thermoelastic, in the context of the 3-dimensional
classical continuum model constructed in previous chapters. In other words,
we shall obtain the eonstitutive equations for a thermoelastie material.
In Chap. VI (Sect. 4.2), we already invoked certain principles governing
the form of any constitutive law, namely, spatial isotropy (or material frame
indifference) and symmetries inherent in the material. However, the above
statement now brings out a key principle, the prineiple of loeal action. This
says in the present case that internal forces represented via the stress model,
which correspond to purely local actions (see Chap. V, Sect. 3.4), are related
by the thermoelastic constitutive law only to quantities involving the particle
in question and its nearest neighbours.
Needless to say, the equations governing thermoelastic behaviour must
be formulated, like any constitutive equations, in accordance with the laws
of thermodynamics, that is, obeying the inequality imposed by the second
law. For this reason, some thermodynamic not ions will be introduced in the
context of continuous media, whilst the reader is advised to consult more
specialised works for furt her information on these topics.
302 Chapter VII. Thermoelasticity

2. Experimental Observations

2.1 Generalities

A constitutive law is in the first instance the result of experimental observa-


tion, even though it must obey certain very general principles which prede-
termine in some sense the way it is expressed.
We saw in Chap. V (Beet. 3.3) that the equations of motion can be written
in the form of
• three first order partial differential equations for the field equations,
• three boundary conditions,
whilst {[ is determined by finding six independent scalar fields (independent
components of the symmetrie d~, t)). If we assurne that the fields, F, Q, 'Ln
are known, for example in the static (or quasi-static) case where Q(~, f) =
0, V~ E nt , we see that these equations do not determine the stress field {[
in the system under investigation (test piece). In a like manner, boundary
conditions on displacements are not sufficient to determine the strain field.
Hence the problem faced here is not strictly speaking to determine the
constitutive law from experiment, but rather to identify a locallaw, within
the framework of the geometrical and mechanical model, based on experi-
mental results that are of a global nature. In a completely general context,
such a problem is extremely complex, if not insoluble. In practice, therefore,
we seek to carry out experiments in which the relevant fields can be treated
as homogeneous. Clearly, we must somehow guarantee that the test piece
is homogeneous on the maeroseopie seale under eonsideration, and by suffi-
ciently restrictive boundary conditions, notably through the geometry chosen
for the test piece, but also by the type of loading or excitation, we must be
sure that the stress and strain fields are either homogeneous or vary only
slightly around the homogeneous state. For example, with this problem in
mind, many classical experiments aim to achieve homogeneous stress fields
in which the stress state at each point is one of the simple stress states stud-
ied in Chap. VI, Beet. 3, i.e., simple tension, simple compression, pure shear,
and so on.
The torsion of a thin tube with circular cross seetion (Fig. 1) thus aims to
attain a pure shear state at each point of the test piece, whose components
in the orthonormal basis of cylindrical coordinates are constant: lJ(}z = IJz(} =
constant, other lJij = 0. 1 More complex stress fields can also be obtained,
e.g., tension-torsion testing of a thin tube (Fig. 4).
Note that in all these experiments, attention must be paid to thermal
conditions. As a general rule, the temperature is taken to be constant in time
and also throughout the material making up the test piece.
1 Such an experiment must nevertheless be analysed very carefully in terms of
symmetry properties of the material (isotropy, anisotropy, see Sect. 5.7).
2. Experimental Observations 303

Fig. 1. Torsion of a thin tube

2.2 Simple Tension Test

Figure 2 shows a typical test piece for a simple tension test. Made from a
homogeneous metal, it is characterised by:
• thickening at the ends (to avoid end effects),
• curved connections between the central part and the ends (to avoid the
stress concentration effects that occur near sharp angles),
• a cylindrical central part in which the stress field is assumed to be a simple
homogeneous tension parallel to the axis of the test piece. 2
In its simplest form, the experiment involves recording the change in the
tensile force F or the ratio F / So, where So is the initial value of the cross
section in the central part of the test piece, as a function of the relative
extension

(R - Ro)/Ro = t::.R/Ro ,
where Ra is the initial length marked on the useful part of the test piece.
Figure 3 gives an example of the type of results obtained, for a stainless
steel. In this case, the following features are immediately observable.
a) Independence of the graph from the loading rate, i.e., independence from
(t::.R/ Ra).
b) Reversibility of the portion GA in the loading diagram. After loading up
to a level below the threshold 0"0 marked by the point A on the curve,
total or partial unloading causes the system to run back down the same
curve in the opposite direction.
c) Linearity of the reversible part of the curve.
2 The eonstitutive material (metal) is assumed isotropie. In the ease of anisotropie
materials, the simple off-axis tension experiment requires special preeautions to
be taken, partieularly eoneerning the mounting of the test piece.
304 Chapter VII. Thermoelasticity

FIS. in IO' MPa

ilt

Fig. 2. Test piece for a tension test Fig. 3. Stress-strain diagram for a
stainless steel under tension

d) Beyond the threshold 0"0, i.e., when the initial load exceeds that corre-
sponding to point A, subsequent unloading is represented on the graph
by a different curve to the loading curve. In other words, after following
the loading curve GAB, the system then moves down Be as the load is
removed. In particular, when the load has been completely removed, that
is, when F is reduced to zero, the test piece retains a certain degree of
extension.
The reversible part of the loading diagram represents, by definition, the
elastic behaviour of the material. The point A and the value 0"0 correspond
to the initial elastic limit or yield point in this experiment (see Chap. VI,
Sect. 4.1).
The linearity observed for the portion GA characterises linear elastic be-
haviour.
A furt her experimental observation must also be added to this list, al-
though it is not apparent from the graph in Fig. 3. During traction, the
cross-sectional radius of the test piece is observed to diminish in a linear way
as the system moves up the portion GA of the loading curve. This variation
in the cross section is also reversible for all points along GA.

2.3 Further Experimental Results

Other experiments reveal the same characteristic behaviour. The simple com-
pression of a test piece or the simple torsion of a thin tube are cases where
loading depends on a single parameter (force or couple). In the tension-
torsion of a thin tube, the load depends on two parameters (simultaneously
applied force and couple). It induces a stress field in the test piece whose
components in the orthonormal basis of cylindrical coordinates are assumed
2. Experimental Observations 305

0" in MPa

0 " in MPa
~-+---+--~~--~---r---4~~r-~~.

Fig. 4. Test piece for the tension-compression and torsion of a thin tube. Example
of an experimentally determined yield surface (H.D. Bui, 1969)

to take the form: u zz = constant, UO z = U zO = constant, other Uij zero.


The elastic domain of the material can then be investigated (see Chap. VI,
Sect. 4.1), inside which deformations suffered by the test piece are reversible.
Figure 4 gives an example of an elastic domain determined in this way
(tension-compression and torsion experiment).

2.4 Remarks

Historically, the discovery of linear elastic behaviour has been attributed


to Robert Hooke (British mechanicist, physicist, astronomer and naturalist,
1635-1703) in the 1660s. He made experiments with spiral springs and the
traction of metal wires.
Ut tensio sic vis
was the phrase chosen by Hooke to announce his results in a work published
in 1678. 3 ,4
3 Extract from The New Science of Strong Materials by J.E. Gordon: "Hooke, like
Horace, did not suffer unduly from modesty and he staked his claim to priority
in a number of fields by publishing in 1676 A decimate of the centesme of the
inventions I intend to publish among which was The true theory of elasticity or
springiness. This heading was followed simply by the anagram 'ceiiinosssttuu'.
The scientific public were left to make what they could of this untiI, in 1679,
Hooke published De potentia restitutiva, or of aspring where the anagram was
revealed as OUt tensio* sic uis' - 'As the extension, so the force'. *Tensio means,
generally, not tension but extension in Latin. The truth seems to be that the
Romans muddled up the two ideas. Literary writers probably never thought
about the matter at all."
4 Recall that Galileo Galilei had also had recourse to an anagram when secretly
informing Johannes Kepler that he had discovered the phases of Venus: "Haec
306 Chapter VII. Thermoelasticity

Fig. 5. Illustration taken from Lectur'es de


Potentia Restitutiva, or of Spring, Explain-
ing the Power of Springing Bodies, R. Hooke
(1678)

French physicist and abbot, Edme Mariotte (1620-1684) should also be


remembered in this context. He independently made the same discovery as
Hooke, in about 1680.

3. Thermodynamics of Continuous Media

3.1 First Law: Energy Equation

We shall consider a system S occupying a region V with volume D t and


boundary {)Dt in the current configuration "'t, the convective transport of
V o, Do, {)Do in the reference configuration "'0.
We assurne that the thermodynamic state of the system is determined by
certain fields defined on "'t (or "'0) .
For the sake of simplicity, the discussion will be restricted to the case
where the system Sexchanges only heat and work with its surroundings.
The first law of thermodynamics postulates the existence of a function of
the thermodynamic state of the system called the internal energy, denoted
by E, with dimensions of work and the property that:
• at any moment of time, the sum of the material derivative of the internal
energy E of Sand the material derivative of the kinetic energy K of S
is equal to the sum of the rate of work done by external forces acting on
the system in its actual motion, p(e) (U), also called the power input, and
immatura a me iam frust ra leguntur 0 y" for "Cynthiae figuras aemulatur mater
amorum" . ("These things that are unready far disclosure are read by me" for
"The phases of Cynthia [the Moon] are imitated by the mother of love [Venus]" .)
3. Thermodynamics of Continuous Media 307

o
the heat input Q for the system. In symbols,

1 E + k = p(e) an + QI· (3.1)

This statement is equally true for any subsystem S' of S:

+ K = p(e)(U) + Q
• I • I I 0 I
E . (3.2)
In (3.1) and (3.2), the symbol 0 on the heat input indieates that this
quantity is not a material derivative (in the sense of Chap. III, Seet. 4) of
some funetion measuring the quantity of heat reeeived and whieh might be
explicitly defined as a funetion of variables eharaeterising the state of the
system and its evolution at the time t. Indeed, in the differential form, with
whieh the reader may be more familiar, (3.1) ean be written

dE + dK = d'W + d' Q ,

where W stands for work and the two terms on the right hand side are not
exaet differentials.
For the 3-dimensional eontinuum, external farees were deseribed in Chap. V
(Seet. 2.2 and 3.1), assuming that the particles in the system exert no action
at a distanee upon one another:
• body forees defined by a density per unit mass F(;]2, t),
• surfaee forees aeting on the boundary of S, or S', and defined by a surfaee
density 'Ln(;]2, t) for S, 'Ln' (;]2, t) for S'.
We dedueed the expression for the virtual rate of work by external forces, in
sueha way that far the aetual motion we have, for an arbitrary subsystem,

P Ce ) (U) = r
in;
p F(;]2, t) . U dilt + r
ion;
'Ln' (;]2, t) . U da . (3.3)

The kihetie energy is given, as deseribed in Chap. IV (Seet. 7.5), by

(3.4)

o
For the heat input Q into the system S, we make the hypothesis that
there ean be no heat exchange at a distance between particles of Sand that
o
Q is foundas the result of two eontributions.
• A volume term, expressing the rate at whieh heat is supplied at a distanee
to the particles of S from outside the system and the loeal heat produetion
by possible ehemieal or nuclear reactions, and so on. This ean be expressed
in the form of an integral over ilt of a density per unit volurne r(;]2, t),
also ealled the internal heat souree per unit val urne.
308 Chapter VII. Thermoelasticity

• A surface term, integral over the boundary of S of a surface density


hn(:r, t) .
Hence, we find

r hn(:r, t)da +
Q= lant 1 nt
r(:r, t)dnt . (3.5)

o
For Q' , referring to an arbitrary subsystem S', a similar decomposition
is postulated.
• The volume term for S' is the integral over .0;
of the same volume density
r(:r, t) as defined for S , whatever the subsystem S' , since there is no heat
exchange at a distance between the particles of S.
• The surface term for S' is the integral over an; of a surface density
hn' (:r, t) depending only on first order local elements of an; , i.e., h n , (:r, t)
depends on .0;
only through :rr(:r), so that we may write

hn' (:r, t) = h(:r, t, :rr(:r)) . (3.6)


Consequently,

r h(:r, t, :rr(:r)) + 1
cl = lan; n;
r(:r, t) dn
da t . (3.7)

We assume additivity of the internal energy E. More precisely, we assume


that for two complementary subsystems S{ and S~ within S, we have

(3.8)
also written

E~ +E~ = E,
since our formulas only involve vari.ations (material derivatives in thiscase).
We may then introduce the density per unit mass of internal energy
e(:r, t), also called the specific internal energy. We have

E' = 1
n't
p(:r, t) e(:r, t)dnt and E' = 1
n't
p(:r, t) e(:r, t)dnt . (3.9)

Equation (3.2) can then be written, simplifying the notation,5

VS' eS,
{
1 (e +
ddt n; p U
2
2
) dnt = 1n;
(pF. U + r) dnt + r (Ln"
lan;
U + h(:rr)) da .
(3.10)
5 The dependence of the various quantities on :r and t is dropped under the integral
sign.
3. Thermodynamics of Continuous Media 309

If the real velo city field U is continuous and continuously differentiable,


we can apply the kinetic energy theorem (Chap. IV, Sect. 7.5) in (3.1) and
(3.2). This says that

VS' eS,
{
p(e)(U) + P(i)(U) = K' ,
so that, substituting into (3.1) and (3.2), we obtain
• 0

E =Q- P(i) (U) , (3.11)

(3.12)

P(i) (U) can then be replaced by the expression in which internal forces
are represented by the Cauchy stresses, yielding

VS' eS,
r p(J;., t) e(J;., t)
Jn~
dilt = r (g(J;., t) : g(J;., t) + r(J;., t))
Jn~
dilt
(3.13)
+ r h(J;., t, 11(J;.)) da .
Jan~
Equation (3.10), or (3.13) via the kinetic energy theorem, expresses the
first law of thermodynamics in global form (balance equation).
From (3.13), the local form of the first law can be obtained by using the
fact that this equation is true for any subsystem S'.
Indeed, it can be shown that, since (3.13) is true VilL the surface density
h(J;., t, 11(J;.)) , assumed to be continuous and dass Cl, necessarily takes the
form of a flux which we write

(3.14)

where ~(J;., t) is the outward heat flux vector 6 at point M. Heat exchanges
between particles in the system occur via conduction.
o
Hence, as a consequence of the first law, Q' given by (3.7) necessarily
takes the form

Q' = - r
Jan~
~(J;., t) . 11(J;.) da + r r(J;., t)
Jn~
dilt . (3.15)

Applying the divergence theorem, the last integral of (3.13) is, by (3.14),
transformed into an integral over il~. We thus deduce the local form of the
6 The adjective 'outward' reminds us that h corresponds to the heat received by
the system with outward normal!! and that (3.14) has a minus sign on the right
hand side. This is the most commonly adopted convention.
310 Chapter VII. Thermoelasticity

first law of thermodynamics in the Eulerian representation, which we call the


energy equation:
P(;12, t)e(;12, t) = g(;12, t) : ~(;12, t) + r(;12, t) - div 9:(;12, t) .
In the abbreviated notation,

pe = g : ~ + r - div 9: I. (3.16)

Comments
• Equation (3.14) is sometimes called the tetrahedron lemma, because of its clas-
sical proof, which closely resembles the argument given in Chap. V, Sect. 3.6 to
prove that stresses form a tensorial quantity. Equation (3.13), valid \fD~ , is applied
to an infinitesimal subsystem containing an arbitrary point M in the interior of
D t . This subsystem has the shape of a tetrahedron constructed on the three axes
of an orthonormal coordinate system. The edges of the tetrahedron have lengths
dXl, dX2 and dX3 (Fig. 6).

Fig. 6. Tetrahedron lemma

For this infinitesimal subsystem, the volume integral in (3.13) is third order
in dXj. The surface integral is the sum of four terms corresponding to the four
faces MA1A2,MA2A3,MA3Al and A 1A 2A 3 of the tetrahedron. Equation (3.13)
requires the sum of these four terms to be third order, so that the sum of second
order terms must be zero. Simplifying, it then follows that

~dX2dX: h( -~l)+ ~dX3dxl h( -~2)+ ~dXldx2 h( -~3) + dS h(nJ =0,


{ (3.17)
rrdS = 2(~ldx2dx3 + ~2dx3dxl + ~dXldx2).

Setting
.2. = ~l h( -~l) + ~2h( -~2) + ~3h( -~3) , (3.18)
equation (3.17) becomes
h(rr) = -.2. . rr , (3.19)
which proves (3.14).
3. Thermodynamics of Continuous Media 311

• Equation (3.14) proven above shows that, in the region nt , there exists a vector
field q(;r:, t), called the outward heat flux vector. On the boundary ant of S, the
field ~ satisfies, as a boundary condition,

(3.20)

• The result (3.14) is based upon the additivity hypothesis for internal energy.
Physically speaking, appealing to (3.8), it can be seen that the additivity of internal
energy merely intro duces a further hypothesis in addition to those already made
to exclude both action at a distance and heat exchange at a distance between
the particles of S. At the boundary between two complementary subsystems, the
overall balance of heat and work exchanges between the two subsystems is zero, so
that the rate of work by internal contact forces between the two subsystems in any
discontinuity of the actual velocity field is compensated locally by an equivalent
heat input for the two systems.
A quite different approach might have been adopted here. We could have made
the hypothesis that heat exchanges between particles in the system take place only
by conduction according to (3.14), and then deduced the additivity of internal
energy and the energy equation (3.16) as a consequence.

3.2 Second Law: Fundamental Inequality

The second law of thermodynamics for continuous media postulates the exis-
tence of a universal temperature scale known as absolute temperature, denoted
by T, which is always positive, and an additive function ofthe thermodynamic
state of the system called the entropy, denoted by 8, such that
at any time, for the system S, and for any subsystem S' of S, we have
the fundamental inequalities

d8 = 15 > r r(;r:, t) dD r t _ 'l(;r:, t) . n(;r:, t) da (3.21 )


dt - Jn t T(;r:,t) Jan t T(;r:,t)
and

{ ., 1
VS' eS,

8 >
-
r(x, t)
---dDt
n; T(;r., t)
-
ian;
q(;r:, t) . n(;r:, t)
-
T(;r., t)
da.
(3.22)

Let 8 denote the entropy per unit mass (or specific entropy). Then,

8' =
Jrn , p(;r., t) 8(;r., t) dDt
t
and 15' = r p(;r., t) 8(;r., t) dD
Jn;
t .

Applying the divergence theorem and simplifying the notation, (3.21) and
(3.22) transform to
312 Chapter VII. Thermoelasticity

'<IS' eS,
{
L: (ps + div(g/T) - r /T) dDt ~0,
from which we may deduce the local form of the fundamental inequality

ps + div(51./T) - r/T ~ 0 . 7 (3.23)

Taking the energy equation (3.16) into account in this formula, we obtain
(since T > 0)

g: g+ p(Ts - e) - ~. gradT ~ 0 . (3.24)

This basic inequality can be transformed by introducing the thermody-


namic function called the Helmholtz free energy. The free energy per unit
mass 'ljJ is defined by

'ljJ=e-Ts. (3.25)

We then obtain the Clausius-Duhem inequality 8

(3.26)

The left hand side of (3.24) and (3.26) is often denoted <P. This is the
dissipation per unit volume in the current configuration. It can be written

(3.27)

where

(3.28)

and

<P2 =- Tq .grad T (3.29)

are respectively the intrinsie dissipation per unit volume and the thermal
dissipation per unit volume in the current configuration.
The fundamental inequality thus states that the dissipation is non-
negative.

7 The left hand side of (3.22) is the rate of entropy increase, whilst the right hand
side may be called the entropy input rate. The right hand side of (3.23) is the
internal entropy production per unit volume.
8 R. Clausius (1822-1888). P. Duhem (1861-1916).
3. Thermodynamics of Continuous Media 313

The evolution of a system S is said to be thermodynamically reversible if


at any time and at all points of the system

cJh = 0 and <[>2 =0. (3.30)

Note that for an adiabatic evolution, in which q = 0 at any time and at


all points, and for an isothermal evolution, in which-T is constant throughout
the system as weIl as in time, we have <[>2 = O. (But clearly, these are not the
only situations in which <[>2 = 0.)

3.3 Lagrangian Expressions

All arguments in the preceding sections were based on the current configura-
tion "'t. Equations (3.16) and (3.24), or (3.26), are local Eulerian expressions
of the first and second laws of thermodynamics for continuous media.
The Lagrangian formulation of these laws is obtained, either for the whole
system or for a subsystem, by transporting the integrals in the equality (3.10)
and the inequality (3.22) into the reference configuration. Local expressions
can then be deduced in a like manner to the arguments produced above.
More simply, note that the left hand sides of the local Eulerian expressions
(3.16), (3.24) and (3.26) are just the volume densities in the integrals (3.10)
and (3.22) over the current configuration. The local Lagrangian express ions
are the same equality and inequality as (3.16), (3.24) and (3.26) for the
volume densities of the same quantities in the reference configuration, once
written in terms of Lagrangian variables.
Let us consider in particular the fundamental inequality (3.24) and the
Clausius-Duhem inequality (3.26). The Lagrangian volume density corre-
sponding to (3.24) follows from the expression for mass conservation dm =
Po(X) dDo = p(J;., t) dDt . We then have

Po Q. : 4 + Po (T s - e) - Po :1. . grad T ~ 0 , (3.31 )


P-- pT--

which should be expressed in terms of Lagrangian variables.


The second term in (3.31) is the simplest in this respect. The functions
T, e and 8 are physical quantities related to the material. The Lagrangian
expressions for the temperature, and for the internal energy and entropy per
unit mass in the reference configuration at time t are

T(X, t) : T(;r, t) , ;r = 1!.(X, t) ,


{ e(X, t) - e(;r, t) , ;r = 1!.(X, t) , (3.32)
8(X, t) = 8(;r, t) ;r = 1!.(X, t) .

The material derivative is conserved in the transition from the Eulerian to


the Lagrangian description.
314 Chapter VII. Thermoelasticity

The Lagrangian expression for the first term in (3.31) follows from the
definition of the Piola-Kirchhoff stress tensor (see Chap. V, Sect. 4.1). Recall
that, at corresponding points of ""0 and ""t,

KCK, t) : ~(K, t) _ d~, t) : 4(~, t)


{ po(X) - p(~, t) , (3.33)
~=pJX,t).

Finally, for the third term in (3.31), the correspondence between La-
grangian and Eulerian gradients at corresponding points9 together with
(3.32), implies
Po(X) CJ.(~' t)
-(-) - T () .gradT(~,t)
P ~,t ~,t--

I 1
T(K, t) (J(X, t) F- (X, t). CJ.(~' t)). 'VT(X, t) . (3.34)

We thus introduce, in ""0, the vector CJ.o(X, t) defined by

CJ.o(X, t) = J(X, t) F-1(X, t). CJ.(~' t) . (3.35)

This vector CJ.o(X, t) can be interpreted as the outward heat fiux vector
in the reference configuration at time t, since we may check that, for surface
elements dA and da corresponding under convective transport, we have

(3.36)

Finally, the Lagrangian expression for the fundamental inequality (3.24)


is

(3.37)

Likewise the Lagrangian formulation ofthe Clausius-Duhem inequality (3.26)


becomes

I ~ : f=. - Po ('ljJ. + sT)


. - qo .S!I. ? 0 I .
T (3.38)

9 Recall the relations established in Chap. II at corresponding points:


= 1!.(X, t) , ~(X, t) = 'Vqy(X, t) , po(X)j P(;f, t) = det ~(X, t) = J(X, t) ,
;f

'VT(X, t) = grad T(;f, t) . ~(X, t) , da = J(X, t) t~-l(X, t) . dA.


4. Thermoelastic Constitutive Laws 315

4. Thermoelastic Constitutive Laws

4.1 Elasticity Hypothesis

Thermoelastic behaviour is first defined in the framework of the Lagrangian


description. We shall postulate that, for the particle X in the material under
consideration, at time t, the values of the specific internal energy e(X, t),
the specific entropy s(X, t) and the Piola-Kirchhoff stress tensor n::(X, t) are
completely determined once we know the temperature T(X, t) and-the strain
tensor f(X, t) for this particle and at this time. Clearly the same will then
be true-for the specific free energy 'ljJ(X, t).

e(X, t) : e(T(X, t), ~(X, t)) ,


{ s(X,t) - s(T(X,t),~(X,t)), (4.1)
~(X, t) = ~(T(X, t), ~(X, t)) ,
'ljJ(X, t) = 'ljJ(T(X, t), ~(X, t)) .

In order to simplify the equations, we shall adopt the following notation:

'ljJ = 'ljJ(T,~), ~ = ~(T,~) , etc. (4.2)

From a mathematical standpoint, the notation used in (4.1) and (4.2) is


somewhat misleading. In fact, taking the 'ljJ in (4.2) as an example, this func-
tion is not generally a function of the tensor f alone, in the sense explained
in Chap. VI, Sect. 2.7 and 4.2, despite appearances in (4.2). This would only
be the case if the material were isotropie (Sect. 4.5). Otherwise the function
also depends on the spatial orientation of some specific set ofaxes attached
to the material particle at the reference time. A rigorous notation would re-
quire mention of this information amongst the arguments of (4.2), or use of
a matrix notation such as (4.3), whilst maintaining the intrinsic character of
the formulas under change of coordinates. We shall return to this point in
Sect. 4.4.
Motivated by experimental results such as those described in Sect. 2.2
concerning the simple tension test, the assumptions (4.1) are formulated in
such a way that they satisfy the principle of spatial isotropy.

Indeed, if we take the function e as an example, and noting that the experi-
ment does not appear to indicate any time dependence, it would appear reasonable
apriori to write e as a function of temperature T and the gradient E of the trans-
formation which the element undergoes. Having chosen a coordinate system R in
the given frame, we calculate e from the components of ~:

(4.3)

where E. denotes the matrix of E in the coordinate system R, which we shall assurne
orthonÜrmal, for the sake of simplicity.
316 Chapter VII. Thermoelasticity

The principle of spatial isotropy, already discussed in Chap. VI (Sect. 4.2),


here implies that the specific internal energy of the element does not depend on
its orientation in space. More exact1y, if the same element of matter were made to
undergo two transformations, starting from the same initial configuration, which
differed only by an isometry, then the values of e in the two corresponding current
configurations would be equal. The function eR therefore possesses the following
mathematical property

V~, V~ such that ~.~ = ~,


{ (4.4)
eR(T,K,) = eR(T,g.K,).

Recalling the polar factorisation of K, established in Chap. 11 (Sect. 3.4 and 4.5),

{ K,=I1.g,
tE . E = 1), , det E = +1, tri = ri , (4.5)
- - -
we see that it is possible, in (4.4), to choose g =t lJ.,
thereby bringing out the fact
that eR (T,~) depends only on the matrix §" of the stretch tensor, or alternatively
t.. or f:
(4.6)

The result obtained here is clearly valid for the functions sand 'Ij;.
As far as stresses are concerned, the arguments are analogous. In the coordinate
system R, the dependence most naturally suggested by experience is

(4.7)

The principle of spatial isotropy supplies the mathematical property that the matrix
function gß must satisfy in the co ordinate system R, viz.,

VE. , VfJ. such that tfJ.. fJ.


= Ä ,
{ (4.8)
g-;;(T,g.~) = g.g;(T,~)~tg.

Then, choosing g =tlJ. defined by (4.5), we have the relation

or, introducing the Piola-Kirchhoff stress tensor,

(4.9)

thereby justifying the claim in (4.1).

4.2 Unconstrained Thermoelasticity

Thermodynamic Potential
We start with the Lagrangian formulation (3.38) of the Clausius-Duhem
inequality. This inequality governs all real thermoelastic evolution for the
particle under consideration. From its state specified by the variables T and
4. Thermoelastic Constitutive Laws 317

~, in which the outgoing heat flux is '10 with temperature gradient VT, such
an evolution is defined by i' and ~.
The inequality (3.38) involves the material derivative "j; of 'lj;(T,~) . In an
arbitrary basis, this can be written in terms of the components eij of ~:

· 8'lj; (T, ~). 8'lj; (T, ~) .


'lj; = 8T - T + 8eij - eij . (4.10)

Quite generally, if fW is a differentiable scalar function of the components of


f
an arbitrary tensor = t ij ~i ® ~j' the second rank tensor 8 f 8~ is defined W/
by the relations

(4.11a)

so that
· 8fW· 8fW·
f = - t··
8 - tij = - 8
- : 1:. .
t
(4.11b)
tJ =-
Applying this general definition to 'lj; (T,~) introduces the second rank tensor
8'lj; (T,~)/8~, viz., -

(4.12a)

and the material derivative (4.10) becomes


· 8'lj; (T, ~). 8'lj; (T, ~)
'lj; = 8T T + 8~ - : f.;, • (4.12b)

We assurne here that there are no internal constraints associated with the
material. This means that the geometrical evolution of the material element
is not subject to any restriction such as incompressibility, inextensibility in
one or more directions, and so on, which might arise from its microstructure,
chemical constitution, or other. (Such cases will be discussed in Sect. 4.3).
The tensor ~ is an arbitrary symmetrie tensor and any evolution is described
by arbitrary i' and symmetrie f.;,.
The Clausius-Duhem inequaJity (3.38) thus beeomes

VT, V~ symmetrie, VVT ,


vi', V~ symmetrie,
. (8'lj; (T,~). 8'lj; (T,~) . .) (4.13)
lb (T,~) : ~ - Po 8T - T + 8~ - : ~ + s (T,~) T

VT > 0 .
- '1T0 . - -
318 Chapter VII. Thermoelasticity

The first line of this inequality, representing the intrinsie dissipation per
unit volume, depends only on T, §., T and f:.. The seeond line represents the
thermal dissipation per unit volume. By Fourier's law '10 = - K 0 (T,~) . V'T
we see that it depends on T, §. and V'T. From (4.13), it follows that eaeh of
these two lines separately must be either positive or zero.
Indeed, putting T = 0 and f:. = 0 in (4.13), we obtain the (non-striet)
positivity of the thermal dissipation:

VT, V~ symmetrie, VV'T ,


{ (4.14)
-'10 . V'T ~ 0 .

This is the conduction inequality.


Then setting V'T = 0, we obtain the (non-striet) positivity of the intrinsie
dissipation:

VT , V~ symmetrie , VT , V~ symmetrie ,
{ ö1jJ (T, §.)). ( ö1jJ (T, §.)) .
( ~ (T,~) - Po ö~ - : ~ - Po s (T,~) + öT - T ~0.

(4.15)

Taking ~ =0 and using the faet that T ean be ehosen arbitrarily, we then
have

VT, V§. symmetrie,


= _ ö1jJ (T,~) (4.16)
(T )
s,~ öT

It follows that

(~(T,~) - Po ö1jJ~:, §.)) : ~ ~ 0, V~ symmetrie,


so that, sinee ~(T,~) is symmetrie,

Ö1jJ(T,§.))
~(T,~) = Po ( ö~ s '
(4.17)

the symmetrie part of the tensor Poö1jJ(T, §.) / ö§..


It is not surprising to find that only the symmetrie part of the tensor
ö1jJ(T, §.) / ö§. enters the expression in (4.17). This results artifieially from the
way we made the definition (4.12a). Indeed, the thermodynamie function
1jJ(T, §.) is only physieally defined on the set of symmetrie tensors §.. Apply-
ing the general definition (4.11a), only the symmetrie part of ö1jJ(,T,§.)/ö§.
is physically determined by (4.12a), sinee in any evolution of the material
4. Thermoelastie Constitutive Laws 319

element, §. is symmetrie. lO The antisymmetrie part of ö'ljJ(T,f:.)/öf:. only has


a mathematieal signifieanee. It eorresponds to the arbitrary continuous and
differentiable extension of the physieal function 'lj)(T, f:.) to the whole set of
seeond rank tensors, i.e., to an expression for 'ljJ(T, f:.) as a function of the
ni ne independent eomponents of f:.. This artifieial diffleulty ean be removed
by taking the following formula as the definition of 'ljJ for ~ an arbitrary seeond
rank Euelidean tensor:
(4.18)
It implies that the eomponents eij and eji for i i- j play the same role in this
symmetrie expression for 'ljJ. We then have
Ö'ljJ(T,f:.))
(
ö~ s

and (4.17) assumes the simple form

VT, V~ symmetrie,
ö'ljJ(T,f:.) (4.19)
lI:(T,~) = Po Öf:.-

Equations (4.16) and (4.19) show that, with the hypotheses (4.1), the
function 'ljJ determines the behaviour of the thermoelastie material. The form
of these equations justifies ealling 'ljJ(T, f:.) a thermodynamie potential. Note
that they imply that the inequality (4.15) is aetually an equality. In fact, the
intrinsie dissipation per unit volume is zero. It follows that an adiabatie or
isothermal evolution (see Seet. 3.2) of the thermoelastie material is thermo-
dynamically reversible.
We shall see in Sect. 5.5 for the ease of a linear thermoelastie model that
the stability of the material implies that the function 'ljJ is strietly eonvex l l
with respeet to symmetrie f:.. This striet eonvexity means that, at T fixed,
the eorrespondenee between- ~ and lI: is bijeetive, and hence implies that de-
formations are reversible in the sense indieated when diseussing Fig. 3.

Legendre-Fenchel Transform
The convexity result allows us to introduee the transformed function 'ljJ* ob-
tained by the Legendre-Fenchel transform.
For values ofT, f:. and 1[ related by the thermoelastie eonstitutive law (4.19),
the function 'ljJ* with arguments T and 1[ is defined by

(4.20)

10 •
Indeed, deeomposmg .
(4.11a), weobtam 8fW : t. = (8fW)
-",- .
-",- : t + (8fW)
-",- : t. .
ut - u~ s -s u~ a-a
11 See Chap. X (Seet. 1.5) for the definition ~f eonvexity.
320 Chapter VII. Thermoelasticity

The material derivative of 'lj;* (T,~) is

· ö'lj;* (T, 7I). ö'lj;* (T, 7I) .


'lj;* = - T + - : 7r . (4.21)
öT Ö7I =

It can also be obtained by differentiating (4.20) to give

· 1 1 .
'lj;* = - K: ~ + - 7I : f. - 'lj; . (4.22)
Po-- Po--
Using (4.12b), (4.16) and (4.19), equation (4.22) becomes

· . 1
'lj;* = sT + - ~ : K. (4.23)
Po - -
Comparing (4.23) and (4.21), we find that, if we defined the thermody-
namic state by T and 7I, we have from 'lj;* (T, 7I) the counterparts to (4.16)
and (4.19),which can b~ written 12 -

Ön/.* (T ,7r)
(T ) = 'f/ -
s,~ öT'
{
(4.24)
ö'lj;*(T,7r)
f!:(T,~) = Po ö~-

Moreover, 'lj;* is itself convex in ~ (see Chap. X, Seet. 1.6).

Symmetry
In terms of components, (4.19) becomes

ij ö'lj; (T, ~)
7r =Po ö . (4.25)
eij

This implies the remarkable symmetry

Ö7r ij (T, f!:) Ö7r kR (T, f!:)


(4.26)
öekR Öeij

which is only valid if we make the hypothesis (4.18), that is, if the 7r ij are
expressed symmetrically in terms of ekR and eRk.

12 Same convention for the symmetrie expression for 'IjJ* as a function of ;g, and for
'IjJ as a function of f!:.
4. Thermoelastic Constitutive Laws 321

4.3 Thermoelasticity with Internal Constraints

Definition of Internal Constraints


The behaviour of the material is said to be subjeet to internal eonstraints
if its thermomeehanieal evolution is bound from a geometrieal point of view
by eertain restrietive eonditions. These eonditions must be integrated into
the equations governing the behaviour of the material. They originate in
the material mierostructure whieh und er lies the meehanieist's model. They
are expressed by one or more independent relations referring to the gradient
of the transformation F(X, t). They must satisify the prineiple of spatial
isotropy (see Seet. 4.1) and henee affeet F(X, t) only through the strain
tensor f (X, t) or expansion tensor C (X, t)~ There ean be no more than 6
indeperi""dent relations (and if n = 6,the material is undeformable).
The most eommon example of an internal eonstraint is ineompressibility,
expressing invariability of the volume of a material element under any real
transformation for the material under eonsideration. Simplifying notation, it
is formulated by requiring

det F = 1 or det (~ + 2~) =1. (4.27)

Another example is inextensibility in a eertain direetion (determined rela-


tive to the material), the length of the material remaining eonstant in this
direetion.
In the present diseussion, we shall suppose that the internal eonstraints
are expressed by n independent relations in terms of ~. These will be written

p = 1, ... ,n (1:<:::: n :<: : 6) . (4.28)

Like 'ljJ(T, f) in Seet. 4.2, the functions 'Pp(f) are only physieally defined on
the set of symmetrie tensors f. We shall write them treating the 9 eomponents
of f to be distinet and ehoosing symmetrie forms in eij and eji whieh thus
satTsfy an analogous equation to (4.18). It follows that tensors ä'Pp(~)/ä~
arising in the diseussion below are symmetrie.

Writing internal constraints in the form (4.28), we may make the same com-
ments as in Seet. 4.1 with regard to the expression for 'IjJ in the form (4.2). These
criticisms can be illustrated by the two examples cited. Incompressibility is clearly
isotropie and its expression in the form (4.28) is fuHy justified by the fact that 'P is
a genuine funetion of the tensor ~ alone, viz.,

'P(~) = det (Jk + 2~) - 1= 0. (4.29)

In contrast, inextensibility with respect to some direction defined by the unit vector
Tl in/i;o is expressed by

(4.30)

whieh shows that 'P is not strictly speaking a funetion of the tensor ~ alone. However,
with this proviso, (4.28) should not lead to any confusion.
322 Chapter VII. Thermoelasticity

Obtaining the Constitutive Law


The Clausius-Duhem inequality (3.38) is written for an arbitrary state of the
particle, defined by the variables T and ~ whieh satisfy the internal eonstraints
(4.28). It refers to a thermomeehanieal evolution defined by T and symmetrie
~, and satisfying the internal eonstraints, i.e., such that

. -- o<pp(f})
<Pp 0 ..
. ~ -_ 0 , p = 1,..., n (1 <
_ n <
- 6) . (4.31)
~ -

Henee, instead of (4.13), we now have

'VT, 'V!l:. symmetrie such that (4.28), 'V\JT ,

'V T , 'V ~ symmetrie such that (4.31) ,

.
JI:(T,!l:.):!l:.-PO
(O'ljJ(T,~). o'ljJ(T,~) .
8T T+ 8!l:.
c.) 9.0
:!l:.+s(T'!l:.)T -T·\JT?O.
(4.32)

The arguments in the last seetion ean be repeated and lead onee again to
the eonduetion inequality

'V T , 'V!l:. symmetrie such that (4.28), 'V \J T ,


{ (4.33)
-9.0 . \JT ? 0 ,
and the non-striet positivity of the intrinsic dissipation, from whieh we begin
by dedueing an expression for the entropy:

'VT , 'V!l:. symmetrie such that (4.28),


{ (4.34)
(T ) = _ 8'ljJ(T,~)
s ,!l:. oT

The rest of the argument is based on the inequality

'V T , 'V!l:. symmetrie such that (4.28) ,

'V T , 'V ~ symmetrie such that (4.31),


(4.35)
o'ljJ(T,e))
( JI:(T,!l:.)-Po 0!l:. = :~?O,
4. Thermoelastie Constitutive Laws 323

from which we deduce,13 given the symmetry of K, o'ljJ(T, f:.)/Of:. and 0<pp(f:.)/Of:.,
that the tensor K(T, f:.) - Po (o'ljJ(T, f:.) /Of:.) is an-arbitrary linear combination
of the n tensors8<pp (f:.) /Of:. . --
We can now deduce the counterpart to (4.19), expressing the thermoelas-
tic constitutive law in the presence of internal constraints:

VT , V ~ symmetrie such that <pp(~) =0, p = 1, ... ,n (1 :s; n :s; 6) ,


{
o'ljJ(T,~) o<pp(~) 14
K(T,
-
f:.)
-
= Po 0f:. + 'T/p-o-
f:.
,
- -
(4.36)

where the 'T/p are n arbitrary scalars (Lagrange multipliers associated with
each internal constraint). We see that K is indeterminate through the n terms
'T/p o<pp (f:.) /Of:., which are n symmetrietensors. Independence of the internal
constrafnts fmplies that the n symmetrie tensors o<pp(f:.)/Of:. are linearly in-
dependent, for any value of f:. respecting the internal constraints (4.28).
Equation (4.36) expressing the behaviour of the material emphasises the
fact that internal constraints (4.28) are an integral part of this behaviour.
In this eonstitutive equation, the n terms 'T/p o<pp(f:.)/Of:. eompensate at the
stress level for the restrietions imposed on deformatiom. This idea of eom-
pensation in the constitutive law is more clearly brought out when we analyse
a system made from sueh a material, that is, if we take a global view of a
thermoelastie evolution problem of the kind to be exposed in Chap. VIII.
Equation (4.36) written out for eaeh point of the system then introduees n
indeterminate scalar fields 'T/p. For a weIl posed quasi-static evolution prob-
lem in which kinematie boundary eonditions are eompatible with internal
eonstraints, these fields are determined, in the eonstruction of a solution, by
the set of all field equations and boundary eonditions of the problem, as we
shall see in Chap. VIII (Sect. 1.3).
It is worth making several remarks with regard to (4.36).
• It is clear that, for a given internal eonstraint, the representation in the
form (4.28) is not unique. It follows that distinet sets of functions <Pp
are equivalent as far as (4.28) is eoneerned. This arbitrariness has no
eonsequenees for the form (4.36) of the eonstitutive equation, sinee the
tensors o<pp(f:.)/Of:., with <pp(f:.) = 0, arising from the different sets of <Pp,
are mutually-proportional and thus lead to the same indeterminacy.
• It should also be noted that, although the presence of internal constraints
(4.28), restrietions imposed on f:., now introduees a degree of arbitrari-
ness into the symmetrie expression for 'ljJ(T,~) as a funetion of the 9
13 The inequality (4.35) yields an equality by eonsidering pairwise opposite values
of ~. It then expresses orthogonality of (~ - po (o'ljJ / of:)) to all the symmetrie
tensors f respeeting (4.31). (1I - po (o1j;/Of)) is thus an element of the veetor
spaee ge;erated by the tensor; o<pp (f) /Of .-
14 With the summation eonvention for the ~epeated index p.
324 Chapter VII. Thermoelasticity

eomponents of ~ treated as independent, equation (4.36) expressing the


eonstitutive equation is insensitive to this indeterminaey. Its effeets are
absorbed in some sense by the presenee of the terms TJp(acpp(~)/a~) .15
• If we eonsider two tensors 1[1 and 1[2 satisfying the eonstituti~e e<1uation
(4.36) for the same values of T and ~, and astrain rate f. satisfying the
internal eonstraints (4.31), it is clear-that -

(4.37)

In other words, the rates of work by internal forees in astrain rate eom-
patible with the internal eonstraints are identieal. The tensor (1[1 _1[2) is
an ineffective tensor throughout any evolution from the present state re-
speeting the internal eonstraints. The eonstitutive law (4.36) thus defines
the internal eonstraints of the material and determines the stresses up to
addition of an ineffeetive tensor.
• Note also that, as in Seet. 4.2, the relations (4.34) and (4.36) imply that
the intrinsie dissipation per unit volume is zero. The internal eonstraints
are ideal in this sense, a eonsequenee of the thermoelasticity hypotheses
(4.1).

4.4 Material Symmetries

In Chap. VI (Seet. 4.2), we introdueed the principle of material symmetries


when diseussing the form of the yield funetion. The thermoelastie eonstitu-
tive law is also subjeet to this prineiple, whieh expresses the invarianee of
the equation under any isometrie transformation belonging to the material
symmetry group, also ealled the isotropy group.
The thermoelastic constitutive law is expressed in the Lagrangian formu-
lation by equations (4.19) or (4.36), whilst the symmetries of the material
are defined in the eorresponding referenee eonfiguration.
The group 9 of material symmetries in the eonfiguration "'0 can be ehar-
aeterised by the following two equivalent statements:
15 Let 'Ij;(T,~) and 'lj;1 (T,~) be two such equivalent express ions in the context of
these internal constrairlts, so that

'Ij;(T,~) = 'lj;1 (T,~), V ~ symmetrie such that 'Pp(~) = 0, p = 1, ... ,n .

Then the derivatives of 'Ij; and 'lj;1 with respect to the 9 components of ~ treated
as independent are related by -

V ~ symmetrie such that 'Pp(~) = 0, p = 1, ... , n ,


where the a p are n scalars determined for each value of (T,~).
4. Thermoelastic Constitutive Laws 325

• It is not possible in the configuration "'0 to distinguish two elements re-


lated by an isometry belonging to the group Q. They reaet in precisely
the same way under the same arbitrary exeitation.
• If we apply, to a given element, two distinet exeitations related by an
isometry in the group Q, the responses are different but related by the
same isometry.
It is interesting at this juneture to write 'ljJ in a more explieit form relative
to some eoordinate system R, chosen orthonormal for the sake of simplieity
las in (4.3)]:
'ljJ='ljJR(T,~), (4.38)
where f the matrix of f. in this eoordinate system.
Respeet for the symmetries of the material in the configuration "'0 requires
'ljJ to satisfy the mathematieal eondition
(4.39)
It ean then be eheeked that we do indeed have, from the eonstitutive equation
(4.19), the relation

Vf,ViJ.EQ,
{ (4.40)
~R(T,tg.~.g) = tg.~R(T,~).g.

Depending on the situation, it may be preferable when expressing and


exploiting this prineiple of material symmetries either to argue in terms of
the funetion 'ljJ and to apply (4.39) as we shall in Seet. 4.5, or to write (4.40)
in terms of the stress and strain tensors (see Seet. 5.7).
Analogous eonsiderations to those given above obviously apply to the
expression for the internal eonstraints.

4.5 Isotropie Thermoelastie Material


in the Reference Configuration

For a material whieh is isotropie (in the Cauehy sense) in the referenee config-
uration, the group Q is the group of alt isometries, both direet and indireet.
This means that we eannot distinguish by their response to identical excita-
tions either two elements with different orientations, or one element and its
mirror image. 16
We return to the arguments already developed in Chap. VI (Seet. 4.2) in
the eontext of the yield funetion. Henee,

Vf, ViJ. such that tiJ. . iJ. = ], ,


{ (4.41 )
'ljJR(T,~) = 'ljJR(T, tg.~.g) .
16 A material whieh is not isotropie is said to be aeolotropie.
326 Chapter VII. Thermoelasticity

The function 7jJ is a sealar (isotropie) function of the tensor ~, whieh does
indeed express the fact that the material has no favoured direetions that
might orient its response. 17
Sinee ~ is symmetrie, we ean apply the representation theorem (Chap. VI,
Seet. 2.7,-Appendix I, Seet. 5.7), whieh says that 7jJ ean be expressed as a
function of T and the invariants of the tensor ~.
Setting -

, 1 ,
I~ = tr ~ 12 = -2 tr e
2
13 = -1 tr e3 (4.42)
= 3 =
7jJ ean be written

(4.43)

We are now in a position to understand how we may exploit the prineiple


of material symmetries, to use the term mentioned at the end of Seet. 4.4. It
allows us to speeify the form that the function 7jJ must assurne. As we shall see
in Seet. 5.3, this will in turn lead to a reduction at the out set in the number
of independent eoeffieients eharaeterising the behaviour of the material and
which remain to be determined experimentally.
Sinee isotropy clearly eorresponds to the biggest group of material sym-
met ries possible, i.e., the group of all isometries, the expression (4.43) repre-
sents the most redueed form that ean be obtained for 7jJ (without introdueing
furt her hypotheses ).
From (4.42), the following formulas ean be dedueed without diffieulty:

81' 8I~ 2
_2 =e ~=~ =~.~, (4.44)
8~ = u~ - --

so that by (4.19), for a material without internal eonstraints (and simplifying


the notation),

(4.45)

in whieh 87jJ / f)J~, 87jJ / 8I~ and 87jJ / 8I~ are functions of T, I~, I~,I~.
This form of the eonstitutive equation for an isotropie thermoelastie ma-
terial brings out the following remarkable result:

for a thermoelastie material that is isotropie in the referenee eonfigumtion,


the stress tensor ~ and stmin tensor ~ share the same prineipal axes.

17 The notation 7jJ(T, f) for the thermodynamic potential is, in this case, thoroughly
justified from the mathematical stand point.
4. Thermoelastic Constitutive Laws 327

For a material with internal constraints, if the isotropy hypothesis is vali-


dated in a physieal sense, it eoneerns all functions expressing the behaviour of
the material and thus, in partieular, the internal eonstraints themselves. The
functions 'Pp (~) also depend only on the invariants If,I~,I~ and we obtain
for 1I -

It still follows that the prineipal axes of 1I and ~ eoineide.


Let us just add that there is no mathematiCal eontradiction involved in
expressing the thermoelastie behaviour of a material via a potential 'lj;(T,~)
whieh is an isotropie function of ~ with anisotropie internal eonstraints. The
eorresponding material is clearly- anisotropie. Physieally realistie models of
reinforeed eomposite materials possess these features.

4.6 Internal Constraints from the Eulerian Standpoint


In terms of the current state satisfying (4.28), internal constraints are expressed in
the Eulerian framework by conditions on the (Eulerian) strain rate deduced from
(4.31) by applying the correspondence formula established in Chap. 111, viz.,

~(X, t) =t!fo(X, t) .g(:r., t) .!fo(X, t) ,

(!
{ (4.47)
:r. = 1!..(X, t) .
It thus follows that

!fo(X, t).
8ipp Ü;;(X , t))
a~ .t;.(X, t)
t )
: g(:r., t) = 0,

:r. = 1!..(X, t) , (4.48)

P= 1, ... ,n (1:::; n :::; 6) .

The constitutive equation (4.36) provides an expression for the Cauchy stress tensor
g(:r., t) when we apply the correspondence formula given in Chap. V:

) p(:r., t) ( ) t
{ g~,t = po(X)!fo X,t .~(X,t). t;.(X,t) , (4.49)
:r. - 1!..(X, t) .
It then follows for g(:r., t) that
328 Chapter VII. Thermoelasticity

VT, V~(X,t) symmetrie such that (4.28),

o'lj;(T,f.(X, t)) t
g(;f, t) = P(;f, t) ~(X, t) . Of. . ~(X, t)
(4.50)
+17~~(X,t) O'PP(~~X,t)) .t~(X,t),
P = 1, ... ,n (1::; n ::; 6) ,

where the 17~ are arbitrary scalars related to the 17P of (4.36) by 17~ = 17P P(;f, t)/ po(X).
Comparing (4.48) and (4.50), we obtain the expected result: the indeterminacy
introduced in the expression (4.50) for the Cauchy stress tensor is just an ineffective
tensor for any evolution respecting (4.48). As an example, if the internal constraint
is incompressibility (4.29), the Eulerian expression (4.48) is simply tr 4(;f, t) = 0
and it follows that the indeterminacy of g(;f, t) in (4.50) is an ineffective tensor,
hence of the form 17' ~ .

5. Unconstrained Linear Thermoelasticity

5.1 Introduction

The experimental observations deseribed in Seet. 2 of this ehapter raised two


basic points:
• reversibility of the loading diagram (or deformations),
• linearity of the loading diagram in all or part of the elastie phase
(Seet. 2.2) when deformations remain smalI.
We saw in Seet. 4.2 that the eonstitutive equation (4.19) aeeounts for
reversibility in a rat her precise manner. Let us now linearise the equation, so
as to obtain a better understanding of the observed linearity.

5.2 Physical Linearisation

Linearised Form of the Constitutive Equation


The referenee eonfiguration is taken to be the so-ealled initial eonfiguration
"'0, and the diseussion is restrieted to infinitesimal deformations from this
configuration, as in Chap. II Sect. 7.1, defined by

lid« 1, (5.1)

and temperature variations which are small in some sense to be determined,

T = (T - T o) small . (5.2)
The idea is then, with these hypotheses and in the absence of internal
constraints, to linearise the constitutive equation (4.19), that is, to write it
in the form of a linear (affine) expression in ~ and T. Such an expression is ob-
5. Unconstrained Linear Thermoelasticity 329

tained by differentiating the second order polynomial expansion of Po 'Ij;(T,~)


in terms of 7 and the components of~. -
Dropping the additive constant which contributes not hing to the deriv-
ative, Po 'Ij;(T,~) is therefore put into the form (5.3) which must, moreover,
respect the symmetry condition (4.18), so that

1 Po 'Ij; = lJo : ~ - Po So 7 + ~ ~ : ~ : ~ - !g : ~ 7 - ~ Po b 72 1. (5.3)

In this formula, So and bare scalar physical constants. The corresponding


terms each have a straightforward interpretation.
• The term lIo : ~ is a linear function of the components of ~. In order to
satisfy the conditfon (4.18) concerning the symmetrie expression for 'Ij;, the
constant physical tensor lJ:0 must be symmetrie.
• The term k : ~ 7 is bilinear as a function of 7 and the components of ~.
In order to satis[y the condition (4.18), the constant physical tensor k must
also be symmetrie. -
• The term ~ ~ : 4 : ~ is quadratic in the components of ~. The tensor 4 is
a constant fourth rank physical tensor and the expression ~ ~ : 4 : ~ denotes
the total contraction over indices 1 and 4, 2 and 3, 5 and 8, 6 and 7, of the
tensor product ~ ~ ® 4 ® ~. Hence, in an orthonormal basis, the quadratic
term is

(5.4)

• Note that in the explicit form (5.4) for the doubly contracted product, 81
terms occur, including 9 square terms and 72 rectangular terms, whilst the
most general expression for a quadratic form in the 9 components of ~ only
requires 45 (= 9 + 36) terms. This is the classic doubling ofrectangular terms.
For (i, j) =I- (k, l), the terms eji A ijkC eCk and eCk A kRij eji are distinguished
and we set

A ijkR = A kCij , V(i, j) =I- (k, l), (5.5)

so that there is symmetry over the two pairs of indices (i,j) and (k,C). By
this means, we achieve a compact expression (5.4) for the quadratic form
and, by the convention (5.5), a simple expression for its derivative (5.7).
• In order to satisfy the condition (4.18) concerning the symmetrie form of
'Ij;, the tensor 4 must be symmetrie in the indices i and j, and also in the
indices k and C:

(5.6)
330 Chapter VII. Thermoelasticity

• Hence, in total, given the 60 symmetry relations expressed by (5.5) and


(5.6), the 81 components of 4 depend on only 21 independent components. 18
The linear thermoelastic constitutive equation is obtained by applying
(4.16) and (4.19) to 'ljJ expressed in the form (5.3), where the derivative of
the quadratic form is given by the simple formula 19

aa~ (!2-~ :==A. :~) A. : ~ .


- = ==-
(5.7)

It then follows that

(5.8)

1
S = So + -Po -l5. : -~ + b T (5.9)

or, in component form

(5.10)

It is worth commenting on this linear expression for the components of 1[ in


terms of T and the components of ~. It is clear that we could have written it
down directly, by the very definition of linearisation, and then imposed the required
symmetries. Symmetry in i and j is immediately obvious, because of the symmetry
of 1[. The complete sum over all components of symmetrie f distinguishes the term
in Aijkf efk and the term in Aijfkekf. Only the sums of co~ponents (A ijkf + Aijfk)
are independent, so that a symmetry can be imposed on indices k and e, thereby
obtaining for 7rij a symmetrie expression in efk and ekf. Then the symmetry relation
(4.26), arising from the existence 0/ the potential po'lj; which gives rise to 1[, can
be applied. It yields the symmetry between the pairs of indices (i, j) and (k, C). We
thus retrieve all the symmetries listed in (5.6).

Physical Interpretation of Coefficients


The constants introduced in (5.3) can now by physically interpreted and the
not ion of 'small' temperature variations can be specified.
18 No sorcery is involved in this reduction! It is simply the number of terms in a
quadratie form on a 6-dimensional space (the independent components of f), and
this is simply the number of independent entries in asymmetrie 6 x 6 matrix.
19 By the condition (5.5), we have Vt., A. t. = t. A.. Then with the definition
(4.11a), it follows that - - - -

a
-
at-
(1 - t :A
2= -==
: t
=
) 1
= - (t : A
2 = -
+ A : t)
== -== =
= A: t .
-== =
5. Uneonstrained Linear Thermoelasticity 331

• Initial Values
The symmetrie tensor zr.0 is the tensor of initial stresses. These are stresses
that correspond to zero deformation and zero temperature difference with
respect to the reference configuration. The linear formula (5.8) states that,
in the Lagrangian representation, the stress zr. results from superposition of
the initial stress '/J0, the stress 4 : ~ induced by the deformation ~ alone,
at constant temperature, and the stress -k T induced by the temperature
difference T at constant deformation. 20 -
So is the initial specijic entropy defined in the same way.

• Elastic Constants
The tensor 4 is the elasticity tensor which, at constant temperature, relates
stresses to cleformations. The above analysis shows that, in an arbitrary co-
ordinate system, the tensor .Ac corresponds to 21 independent coefficients.
This refers to the general case of a material with no specific symmetries.
The 21 coefficients are defined in terms of 18 elastic moduli, proper physical
constants of the material, and 3 Euler angles defining its orientation in the
chosen coordinate system.

This should be eompared with the diseussion in Chap. VI (Seet. 4.2) eoneerning
the yield function and the principle of spatial isotropy. In an orthonormal eoordi-
nate system R bearing a speeifie physieal relations hip to the material, the funetion
'ljJR(T,~) in (4.38) is expressed in the linearised eontext as a function of 18 eo-
efficients that are physical eonstants of the material. The expression for 'IjJ in an
arbitrary orthonormal eoordinate system R' is then obtained from 'IjJ R by

where the change of eoordinate matrix Q is defined by the Euler angles character-
ising the orientation of the material rel~ive to R'.

The elastic constants have dimensions of stress and are given in pascals
(Pa). However, for typical materials, the unit commonly used is the mega-
pascal (MPa).

• Thermal Coefficients
From (5.9), it is clear that

1
Ts=-Tk:e+Tbi. (5.11)
Po - -
20 This result is not maintained for the Cauehy stress tensor Q.. The initial stress
tensor zr.0 induees in the expression for Q. the additive term (pi po)E.. !Io . tE. which
is not ;;-onstant. We shall return to th~ point in Seet. 5.4, diseussi~g i~partic­
ular the ease where this term ean be treated as eonstant for the infinitesimal
transformation.
332 Chapter VII. Thermoelasticity

The left hand side of this equality, with the dimensions of power per unit
mass, is the incoming specific heat input for strain rate f. and temperature
change rate T, setting aside thermal irreversibility. Hence~

(1/ Po) T k is the tensor of specific latent heat of deformation. (6 in-


dependellt coefficients defined from 3 physical constants and 3 Euler
angles giving the orientation of the material).
T b is the specific heat at constant deformation.

• Small Temperature Variations


We can use (5.8) to specify what constitutes a 'small' temperature variation.
The stresses -k T generated by a temperature variation T must be of the
same order of magnitude as the stresses 4 : ~ generated by a deformation ~
satisfying the condition (5.1) for infinites~al deformations.

5.3 Isotropie Linear Thermoelastic Material

The constitutive law for a linear thermoelastie material that is isotropie in


the referenee eonfiguration is obtained by eombining the results of Seets. 4.5
and 5.2.
The polynomial expansion of Po 'IjJ(T, f.) to second order in T and the eom-
ponents of f. must now involve only the invariants of f. defined by (4.42). Given
the order 07 I~, I~, I~, the most general expression for Po'IjJ in the isotropie lin-
ear thermoelastie material is

(5.12)

Note that, in this formula, 1[0, So, A, M, k and bare all scalar eonstants.
Applying (4.16), (4.19) and (4.45), we may deduee that

1
= So + - k I1 + bT,
I
S
Po
or alternatively,

~ = 1[0 lk + A (tr ~) lk + 2 M~ - kT lk (5.13)


1
S = So + - k tr f. + bT (5.14)
Po -
5. Unconstrained Linear Thermoelasticity 333

It can be seen that, in this eonstitutive law, the initial stress tensor lIo of
(5.8) has to be an arbitrary isotropie tensor, i.e., :JI:.0 = 7r0l!:, beeause of the
hypothesis that the material is isotropie.
The elastieity tensor A depends on only two independent eoeffieients, ).
and p,.21 -
These two eoeffieients are genuine elastic moduli, physieal eonstants of
the material, beeause the isotropy of the material implies that its orientation
is irrelevant.

The linear elasticity 0/ an isotropie material is eharaeterised by two elastie


moduli.

The eoefficients ). and p, in (5.12) and (5.13) are ealled Lame constants 22
and p, is the shear modulus.
Finally, we note that the tensor k. must also be isotropie, i.e., k. = k 11..
There thus remains only one thermal coeffieient after the transition from (5.S)
to (5.13). It is a straightforward matter to invert the eonstitutive equation
(5.13) and express f. in terms of 1[ and T. It is eommon praetiee to put the
inverted relation info the form -

(~ - °) = -1+1/
~ y~- E Q
I
1/ (tr :JI:.) l!: + Tl!:, (5.15)

whieh involves 3 elastie and thermal eonstants E, 1/, Q.


• Eis Young's modulus whieh has the dimensions of stress,23
• 1/ is the Poisson ratio, a dimensionless number,24
• Q is the linear thermal expansion coefficient.

f. 0 is the deformation in the zero stress state (1[ = 0) and for zero tem-
perature differenee with respect to the referenee eonfiguration (or state). It
is given by 25

e
=
°= -1-21/
--E
- 7r°-11.. (5.16)

It will be useful to reeord the relations eonverting from the Lame eonstants
to Young's modulus and the Poisson ratio:
21 In fact, A;jkl = A(.5ij .5kl) + p,(.5ik .5jl + .5il .5jk ).
22 G. Lame (1795-1870). In the literature, the shear modulus p, is often denoted by
G and A is called the Lame constant.
23 Doctor, physicist (Young's fringes) and mechanicist, T. Young (1773-1829) was
also an outstanding linguist, interested in both aneient and modern languages,
including the deciphering of Egyptian hieroglyphs.
24 D. Poisson (1781-1840).
25 The similarity of notation may be misleading. Indeed, lIo and f.0 do not corre-
spond to the same state of the material, except when they are -both zero (zero
stress state in the reference configuration).
334 Chapter VII. Thermoelasticity

E- (3)'+2fl) >.
1/ = -::c-;-:----,-
{ -fl (>'+fl) , 2(>' + fl) ,
(5.17)
>'=E 1/
E
(1 + 1/)(1 - 21/) fl = 2(1 + 1/) ,

k=~ (5.18)
1-21/'

The form of (5.15) chosen to express the inverse ofthe linear relation (5.13) may
look surprisingly complicated. The explanation for this choice lies in the meaning
and hence the experimental accessibility of the elastic constants E and 1/. The result
just obtained for the isotropie material is indeed quite remarkable if we reflect for
a moment on the tension-compression test. It means that a single simple tension
(or compression) test, with infinitesimal deformation in an arbitrary direction f. x
relative to the material, in which we impose a stress tensor of form ~ = 'Ir x x f. x ®f. x '
starting from the initial state 7l = 0 and maintaining a temperature variation T =
0, is sufficient to completely determine the behaviour of an isotropic elastic material
through measurement of ~. Indeed, ~ assurnes the form ~ = e xx f. x 181 f. x + (e yy =
e zz ) (f. y 181 f. y + f. z 181 f. z )' where f. y and f. z form an arbitrary orthonormal triad
with f. x . We thus have

where e xx ' e yy and e zz are the linearised unit extensions along f. x ' f. y and f. z for
infinitesimal deformations (Chap. 11, Sect. 7.1). This experiment and its interpre-
tation are clearly facilitated in the case of an infinitesimal transformation, which
will be examined in Sects. 5.4 and 5.5.
Furthermore, a thermal expansion (or contraction) experiment, in which a par-
ticular value of T is imposed, whilst ~ = ~o = 0, leads to adetermination of a:

a = tr ~/3T = exx/T,

where tr f. '::: (dS2t - dS2o) / dS20 is just the linearised volume dilatation for infinites-
imal defo"i=mations.
It should be mentioned that the result established here by exploiting symmetries
ofthe material, expressed via the principle (4.39), starting from the thermodynamic
potential po'IjJ, can also be obtained by applying the principle in the form (4.40) to
equation (5.10) (as in Sect. 5.7). We then observe that the argument is clearly based
on the symmetries between indices i and j and between indices k and C, although
it does not involve the symmetry between the pairs of indices (i, j) and (k, C).

5.4 Infinitesimal Transformation and Geometrical Linearisation

Let us reeall the eorrespondenee between the Cauehy and Piola-Kirehhoff


stress tensors (Chap. V, Seet. 4.1):

Q = !!.... F .1[. tF, where F = (1l, + V'~) . (5.19)


- Po- - -
5. Uneonstrained Linear Thermoelastieity 335

We thus deduee the following expression for !Z. from the linear thermoelastie
eonstitutive equation (5.8): -

(5.20)

We shall now examine eertain key practical situations in whieh the above
formula ean be linearised. This is the geometrical linearisation of the eonsti-
tutive law.

Infinitesimal Transformation
One hypothesis shared by all the above-mentioned situations is that the trans-
formation should be infinitesimal (Chap. II, Seet. 5.1)

11 V'~ 11 « 1. (5.21)

This means that, neglecting seeond order terms in 11 V'~ 11, we may write ~
in its linearised form ~ and identify the Lagrangian and Eulerian gradients,
V'~ and grad ~, respectively, of the displaeement f at eorresponding points
X and ;r. in the referenee and eurrent eonfigurations: 26

(5.22)

In addition, to the same order,

pi Po = dDo/dD ~ (1 + tr~)-l ~ (1 - tr~) . (5.23)

By this hypothesis, retaining only first order terms in (5.20),

g ~ ~o (1 - tr~;) + grad ~ . ~o + ~o . tgrad ~ + 4 : ~ - /fiT . (5.24)

In terms of the symmetrie and antisymmetrie parts, ~ and Mi: of grad ~,

(5.25)

this ean now be expressed in the form

g ~ ;gO + [Mi:.;gO - ;gO . Mi:] + {_;gO tr ~ + ~ .;gO + ;gO . ~ + 4 : ~} - !g T .

(5.26)

In this formula, the term between square braekets says that, when there
is no deformation of the material element and no temperature variation, a
26 As in Seet. 4.6, it should not be forgotten that the quantities in these equations
have spaee and time dependenee. It seems preferable at this stage to saerifiee their
explieit mention, which would soon render the formulas exeessively eomplieated,
in favour of an implicit dependenee that should not eause any great ambiguity.
336 Chapter VII. Thermoelasticity

tensor Q. corresponding to a given tensor lIo is modified by a rigid body


transformation. The terms between curly bnlckets represent a linear mapping
of ~ that can be written

(5.27)

where the B ijk € exhibit the same symmetries as the A ijk € and are defined
from lIo by

B ijk€ = (-7l'?k 6€j + 1f~j 6ik + 1f?€ 6kj + 1fgj 6i€) /2 . (5.28)
The following expression is thus obtained for Q.:

g ~ JI,0 + [Jg .JI,0 - JI,0 . Jg] + (.:i + B - JI,0 ® E) : ~ - !g T . (5.29)

This can be furt her linearised by making more assumptions.

Natural Initial State


We shall say that the initial state, taken as the reference state, is natural if
the initial stress tensor is zero, i.e.,

1I = 0 for f. = 0 and T =0. (5.30)


Then, by (5.8),

lIo =0. (5.31)


Returning to the expression (5.29) for g, it now follows that

Q.~A:~-kT. (5.32)

We shall replace the approximation sign ~ by a strict equality, following


common usage:

Q.=A:~-kT I. (5.33)

We have arrived at an Eulerian formulation in terms of Q., ~,T which is


identical to the Lagrangian expression (5.8) in terms of 1I, f.:;- :;ith lIo = O.
For an isotropie material: - - -

(5.34)

with inverse

~
1 +v
= --y v
g- E (tr g) :& + CH:&
I' (5.35)

the counterparts of (5.13) and (5.15).


5. Unconstrained Linear Thermoelasticity 337

It is eonvenient in eertain applieations to transform these expressions


by introdueing the deviator !i of Q. defined in Chap. VI (Seet. 2.8) and the
deviator f;.d of f;. defined in the same manner:

!i = Q. - (tr Q.j3) 11. ,


{ (5.36)
~d =~- (tr~/3)!.

We then obtain the following formulas (5.37) and (5.38), whieh are equivalent
to (5.34) and (5.35), respeetively, and mutually inverse:

tr g = (3)''+2p,) tr~-3kT,
{ (5.37)
~=2p,~d,

1- 2/.1
~ trg+3aT,
(5.38)
1+/.1
--so
E =
The deviators !i and f;.d are proportional to one another through the shear
modulus p,. Reealling that, aeeording to (5.23), trf;. c:::; (d5tt - d5to)/d5to
represents the linearised volume dilatation, we see that this quantity depends
linearly on the temperature differenee T, through the bulk thermal expansion
eoeffieient 3 a, and on the traee of the stress tensor.
It is usual to set
E
3K=3>..+2p,= - - . (5.39)
1- 2/.1
We see from (5.38) that the elastie volume dilatation in an isothermal uniform
eompression experiment with Q. -
= -p 11.,
-
is equal to

tr~ = trg/3K = -piK . (5.40)

The eoeffieient K defined by (5.39) is ealled the bulk modulus.

Quasi-Natural Initial State


We shall say that the initial state is quasi-natural if in physieal terms the
initial stresses are very small compared with the elastic moduli of the mater-
ial under consideration. (This eomparison has meaning beeause the physieal
quantities involved do have the same dimensions.) In praetiee, such an as-
sumption is very often justified, particularly for the kinds of structure and
material generally eneountered in meehanieal eonstructions. Appealing to
(5.27), (5.28) and (5.29), we have

(5.41)
338 Chapter VII. Thermoelasticity

If in addition we can assume that

11 ~ 11 and 11 grad ~ 11 are of the same order, (5.42)

the terms in square brackets in (5.29) are negligible and g reduces to

(5.43)

Replacing the sign ~ by a strict equality as above, the assumptions made


here thus lead to an Eulerian formulation in terms of IZ, ~ , T which is identical
to the Lagrangian expression (5.8) in terms of JI, ~,;, ;here JIo arises as the
initial stress tensor lZo. Hence, - - -

(5.44)

For an isotropie material, (5.44) becomes

g = (J0 Jk + ), (tr~) 11. + 2 J.1 ~ - k T Jk ' (5.45)

with inverse

°= -
l+v -v (tr(J) 11.
c- c
= =
- (J -
E = E == + aT ='
11. where
(5.46)
c°= ----(J
1-2vo 11. .
= E-
Note that the natural initial state is a special case of the quasi-natural initial
state. Equation (5.44) contains (5.33) but the latter does not require the
extra hypothesis (5.42).
The hypothesis (5.42) is generally satisfied in practice, if a suitable choiee
is made for the frame in which the material is observed. The study of problems
associated with slender objects is nevertheless difficult from this point ofview.

Prestressed Initial State


Note that in (5.44) the initial stress tensor lZo = JIo is clearly compelled to
respect material symmetries in the configuration ~o (see Beet. 5.2). This is
what is implied, for example, by the expression lZo = (J0 J! in (5.45), which
refers to an isotropie material. --
In practice, the validity of (5.44) is extended by dropping the rest riet ion
imposed on lZo by material symmetries. This amounts to saying that material
symmetries ~oncern not IZ but rather IZ' = (IZ-IZO) in the linearised expression
for the constitutive law in the Eulerian formulation.
In particular, this is how the constitutive law is written in the case where
the reference configuration is some configuration A;p that is thermoelastically
prestressed in an infinitesimal transformation of the natural initial state:
5. Unconstrained Linear Thermoelasticity 339

(5.47)

where the tensor of prestresses Q.P is not subjeet to any restrietion, and where
t,',~.' and T' are referred to the prestressed eonfiguration K p . Transformations
are-infinitesimal and must respeet the eondition (5.42).
For an isotropie material:

g' = (g - gP) = ). (tr ~') Jk + 2/-l ~' - k T' Jk (5.48)

~
I 1 +-
=- v (J" I V (')
tr (J" n. + 0: T I n. (5.49)
= =
- -
E = E =

Equation (5.47) can be justified in a way that brings out its meaning.
Letting ~o denote the unstressed configuration of the material and applying
(5.20), we have

(5.50)

(5.51)

Introducing the gradient ~P relative to the configuration Kp , it follows that

and since transformations are infinitesimal here,

§. ~ §.p + §.', where ~'= Cs?'pe' + t V' pO /2 .


Using (5.51), we may now deduce that

tz ~.f.(1l, + V'pO ·tzp· (1l, +tV'pO


- po- - - - -

+ ~ (lk + ~pO· {;: (lk + V'e). (~:~' -IF')· (lk +tV'~P)} . (lk +t~pe'),
(5.52)

which has the same structure as (5.20), with ~p in place of ~o and tzP playing the
ro~~~. -
Theinfinitesimal nature of the transformation from ~o to ~p implies that pre-
stresses are very small compared with the elastic moduli of the material. The anal-
ogous condition to (5.41) for tzP is therefore satisfied. Moreover, the term in curly
brackets in (5.52) is equivale~t to (4 : -
~' ~T'). With the analogous condition
340 Chapter VII. Thermoelasticity

to (5.42) for the infinitesimal transformation from ""p to ""t, we thus obtain the
equation corresponding to (5.44), viz.,

(5.53)

This shows that tensors 4 and Jg in (5.47) are those for the configuration ""0.

Summary and Remarks


Without being exhaustive, the above diseussion gathers together some of
the more frequently eneountered examples of geometrieallinearisation of the
thermoelastie eonstitutive law. 27 It should not be forgotten that ':::::' has been
replaeed by =, and that these are in faet approximations at ~ in "'t, valid
only under the eonditions speeified above. If there is any risk of eonfusion, it
is advisable to return to the Lagrangian formulation at the point X in "'0.
Given the praetieal importanee of this partieular ease, we gather here the
main formulas expressing the linearised thermoelastie eonstitutive law for
an isotropie material assuming infinitesimal transformation and prestressed
referenee state (5.21, 5.41, 5.42).

g = gP + ). ( tr ~') lk + 2 JL~' - k T' lk


, 1 + v,
~ = ~
V (')
g - E tr g ! + CH , lk

(5.54)
). k
v = -::-;-:---.,- Cl! = -,---.,---------,-
2().+JL) (3)'+2JL)

).=E E E
v k =
(1 + v)(l - 2v) JL = 2(1 + v) Cl! ...,-------,-
(1-2v)

5.5 Stability of Thermoelastic Materials

Consider an element of thermoelastie material in its unstressed state, taken


as the initial state. A study of the stability of this material involves examining
whether or not the material element tends to evolve spontaneously or rather
to remain in its initial state, when there are no external exeitations (i.e.,
forees, temperature ehanges).
27 It is also worth mentioning the isotropie prestressed reference state. The term in
square brackets in (5.26) is then zero. (See J. Mandel, Introduction la mecanique a
des milieux continus deformables.)
5. Unconstrained Linear Thermoelasticity 341

We shall assurne 28 that a necessary condition for the isothermal stability


of a linear thermoelastic material is that the quadratic form appearing in
(5.3) should be positive definite, i.e.,

A : ~ posztwe
"21 ~ : = .. dfi'
e mte. (5.55)

For an isotropie linear thermoelastic material, we have

(5.56)

In order to obtain conditions under which this term is positive definite, we


express I{ and I~ in terms of the principal values of ~, viz., el, e2 and e3:

(5.57)

This quadratic form can be written as a sum of squares of three independent


linear forms in el, e2, e3,29 for example,

(5.58)

Necessary conditions for stability, implying (5.55), can now be stated. They
consist of two positivity conditions on the Lame constants:

3A+2/-t>0
(5.59)
/-t>0

If we prefer to use Young's modulus and the Poisson ratio, conditions


(5.59) are equivalent, by (5.17), to:

E>O
1 (5.60)
-l<v<-
2

In practice, negative values of v are exceptional. 30


28 Converse of the Dirichlet criterion.
29 The eigenvalues of the associated linear mapping are ((3'>"/2) + J-L) and J-L, a
doubly degenerate eigenvalue corresponding to el + e2 + e3 = O. The condition
(5.55) is equivalent to positivity of these eigenvalues.
30 A value of v = 0.05 has been measured for beryllium and a doubt remains as to
the sign of v for pyrites.
342 Chapter VII. Thermoelasticity

In order to appreciate the scope of the stability condition as imposed at


the out set in the form (5.55), it is worth identifying the physical meaning of
the rest riet ions (5.59) and (5.60) that follow from it in simple experiments
falling with the context of the infinitesimal transformation. 31
• E > 0 implies, by (5.35), that in an isothermal (7 = 0) tension test where
{[ is uniaxial,

(J'xx > 0,
we have

which means that the material is stretched if subjected to an isothermal


tension.
• v < 1/2 with E > 0 implies that in an isothermal, isotropie compression
test, where g = -p 11, we have, by (5.40),

tr~ = -piK = -3p(1- 2v)/E < 0,


which means that the volume of the material is reduced if it is subjected
to an isothermal, isotropie compression.
• v > -1 with E > 0 implies that, in a pure shear test, the direction of
the maximal principal stress coincides with the principal axis of maximal
extension.
It is comforting to observe that, at least in the first two cases, these conse-
quences seem natural enough.
Finally, as mentioned in Sect. 4.2, the stability condition (5.55) confirms
that, in an isothermal evolution, the linearised elastic constitutive law (5.8) is
a bijective correspondence between ~ and the stress 1[. This implies that in this
type of evolution, deformations are reversible, as illdicated when discussing
experimental observations in Sect. 2.

31 The important assumption is that the deformation should be infinitesimal. The


transformation is only assumed infinitesimal in order to allow a more immediate
interpretation of the experiments (see Seet. 5.3).
5. Unconstrained Linear Thermoelasticity 343

5.6 Typieal Values for Materials in Common Use

Material E [Mpa] 1/ Cl: [10- 6 K -1]*

Mild steel 2 x 10 5 0.25-0.30 12


Invar steel (64% Fe, 36% Ni) 1.4 x 10 5 ~O

Aluminium 7.4 x 104 0.34 22


Silver 8.5 x 10 4 0.39 19
Concrete 3.5-4.5 x 10 4 0.2 10
Bronze 1 x 10 5 0.31 17-19
Copper 1.2 x 10 5 0.34 17
Cast iran 8 x 10 4 0.36 10
Granite 8 x 10 4 0.27 9
Brass 9.2 x 104 0.33 18
Gold 8 x 104 0.42 14
Platinum 1.5 x 10 5 0.38 9
Lead 1.7 X 104 0.45 29
Glass 7 x 104 0.22-0.31 6-10

* A value of Cl: = -5 X 10- 6 K- 1 has been measured for (5 phase plutonium, stable
between 315 and 445°C. (See Y. Quere, Physies of Materials.)

The yield point for mild steel under simple tension is of the order of
240 MPa, whilst for high strength steel it is around 1000 MPa.

5.7 Examples of Anisotropie Thermoelastie Materials

Relative to the macroscopic scale chosen to model a material as a continuous


medium, the microscopic structure arising from the composition and production
of this material governs the symmetry properties discussed in Chap. VI (Sect. 4.2)
and in Sect. 4.4. This explains why, in addition to classic mono- or polycrystalline
materials such as metals, for example, composite materials of all types are affected
by considerations of material symmetry on the scale chosen for their mechanical
analysis. Given the ever increasing practical importance of this type of material,
we shall now present, without going into the details, results concerning the form
of the thermoelastic constitutive law for two types of anisotropie material that are
frequently encountered in practice: orthotropic materials and transversely isotropie
materials. More detailed results concerning, in particular, crystalline anisotropies
(monocrystals) can be found in A.E. Love's book A treatise on the mathematical
theory 0 f elastieity.
The method used here consists in exploiting the principle of material symme-
tries as expressed by (4.40), although other methods exist, based on representation
344 Chapter VII. Thermoelasticity

theorems for tensor functions, for example. To this end we adopt the coordinate
system adapted to the material from the point of view of its symmetries. We may
then enumerate the physical constants (elastic or thermal) that characterise its be-
haviour. Geometrical parameters such as Euler angles must clearly be included in
order to write down the constitutive equation in an arbitrary frame. These merely
specify the orientation of the material (see Sect. 5.2).

Orthotropic Material in the Reference Configuration


A material is orthotropic in the reference configuration if, in this configuration,
it has three mutually orthogonal planes of symmetry. This defines the group of
material symmetries 9 for application of (4.40). Using an orthonormal basis of
vectors lying along the intersections of the three planes of symmetry of the material,
as shown in Fig. 7, the constitutive law is written in component form and we can
express the fact that it is invariant under change of orientation of the basis vectors.
We obtain the constitutive law for the orthotropic thermoelastic material in the
special basis (g, Q, Q):

'Traa = 'Tr~a + All e aa + A 12 ebb + A 13 e cc - k1 T ,


'Trbb = 'Tr~b + A 12 e aa + A 22 ebb + A 23 e cc - k2T ,
'Trcc = 'Tr~c + A 13 e a a + A 23 ebb + A 33 e cc - k3 T ,
(5.61)
'Trbc = A 44 ebc ,
'Trac = Ass e ac ,
'Trab = A 66 eab .

There are therefore 9 elastic moduli and 3 thermal constants. Furthermore, the
initial stress tensor is necessarily diagonal in the basis (g, Q, Q). This basis can then
be specified relative to a general set of eoordinate axes by giving the 3 Euler angles.
It is eommon praetiee to transform (5.61) by introducing the Young's moduli,
Poisson ratios and shear moduli. Then, treating the isothermal case and starting
from the natural state to simplify, the eonstitutive law is written in inverse form
(5.63). We note carefully that there is no symmetry between Poisson ratios, but
there is symmetry in the array (5.63), as indieated in Seet. 5.2, i.e.,
(5.62)

Fig. 7. Orthotropic material:


planes of symmetry
5. Unconstrained Linear Thermoelasticity 345

1 V21 V31
e aa = EI 7raa - E2 7rbb - E3 7rcc ,

V12 1 V32
ebb = - EI 7raa + E2 7rbb - E3 7rcc ,
(5.63)
V13 V23 1
e cc = - EI 7raa - E2 7rbb + E3 7rcc ,

1 1 1
2ebc = -C 7rbc, 2eac = -C 7rac , 2ea b = -C 7ra b·
23 13 12

Transversely Isotropie Material in the Referenee Configuration


A material is transversely isotropie in the reference configuration if, in this con-
figuration, it has an axis of symmetry of revolution and also admits as planes of
symmetry any plane through this axis and one plane perpendicular to this axis.
The material is then said to be transversely isotropie about the given axis. We use
an orthonormal basis comprising two vectors Q and Q lying in the plane of sym-
metry orthogonal to the axis and one vector f lying along the axis of symmetry
of revolution (Fig. 8). The constitutive law is written in component form relative
to this basis and invariance expressed under change of orientation of these basis
vectors and under rotation of the basis about the axis of symmetry.
We obtain the constitutive law for the transversely isotropie material in the
basis Q, Q, f (or any equivalent basis):

7raa = 7r 0 + All e aa + A 12 ebb + A 13 e cc - kT ,


7rbb = 7r 0 + A 12 e aa + All ebb + A 13 e cc - kT ,
7rcc = 7r~c + A 13 e aa + A 13 ebb + A 33 e cc - k3 T ,
(5.64)
7rbc = A 44 ebc ,
7r ac = A 44 e ac ,

7ra b = (All - A 12 ) eab .

There are therefore 5 elastic moduli and 2 thermal constants. Furthermore, the
initial stress tensor is necessarily diagonal and invariant under rotation about the
axis of symmetry of the material. This axis can be specified relative to an arbitrary
set of co ordinate axes by giving two angles.

Fig. 8. Transversely isotropie mate-


rial: axis of symmetry and planes of
symmetry
346 Chapter VII. Thermoelasticity

We can also introduce Young's moduli, Poisson ratios and shear moduli. In the
isothermal case, and starting from the natural state, we have
1 V Vi
e aa = E ?Taa - E ?Tbb - E' ?Tee,

V 1 Vi
ebb = -E ?Taa + E ?Tbb - E' ?Tee,
(5.65)
Vi V 1
eee = - E' ?T aa - E' ?Tbb + E' ?Tee,
1 1 2(1 + V)
2 ebe = G ?Tbc , 2 e ae = G ?Tae , 2 eab = E ?Tab .

6. Historical Perspective
Without aiming at an exhaustive review, it seems pertinent at this point
to mention the historieal appearanee of the main eoneepts diseussed in this
ehapter, devoted to elastie behaviour, deseribing the more signifieant elements
of the work earried out in this field. The reader wishing to go furt her with
these questions is strongly reeommended to refer to the historie al introduetion
and appendixes in A.E.H. Love's A treatise on the mathematical theory of
elasticity.

Uniaxial Elasticity
As indicated in Sect. 2.4, the uniaxial elastic constitutive law, relating the extension
of an elastic solid to the tension force applied to it, was discovered independently by
Hooke and Mariotte. The main application was to the study of beams. Mariotte thus
took up a problem first posed by Galileo (1638) who hoped to evaluate the carrying
capacity of a supporting beam. When studying the elastica (an elastic rod, fixed at
one end and deforming by flexion) Jakob Bernoulli 32 associated the resistance of
the rod to flexion with the extension and contraction of its longitudinal fibres, thus
arriving at a bending moment proportional to the corresponding curvature of the
rod. The problem of the elastica was also studied by Euler, Lagrange, and Daniel
Bernoulli, among others, and led in particular to the first studies of elastic stability
(see Chap. XII, Sect. 3.1). For beams with finite cross section, it was Coulomb 33
who put forward a theory of flexion, applying Hooke's law to the longitudinal fibres.

Continuous Media and Three-Dimensional Elasticity


Thomas Young was the first to recognise shear as an elastic deformation. (Coulomb
only discussed it in relation to rupture. ) He observed that elastic resistance to shear
and elastic resistance to tension-compression are generally different within a given
body. He introduced the idea of elastic modulus of a substance from which arose
the currently used Young's modulus. 34
32 Jakob Bernoulli (1654-1705). Johann Bernoulli (1667-1748) was Jakob's brot her
and Daniel Bernoulli (1700-1782) was Johann's son.
33 C.A. Coulomb (1736-1806).
34 In A Cou.,.se of Lectu.,.es on Natural Philosophy and the Mechanical Ans (London,
1807), Thomas Young gave the following definition: "The modulus of elasticity of
6. Historieal Perspective 347

N avier 35 sought the equilibrium equations for elastic solids through a molecular
theory. Considering matter to be made up of point 'molecules' (material points)
exerting axial forces upon one another, he obtained the elastic forces as a result
of variations in these 'molecular' forces which were themselves due to the relative
displacement of the 'molecules'. Assuming the material to be isotropie, Navier found
the equilibrium equations for an elastic solid, expressed in terms of displacements
of the material points, but involving only one material-dependent constant, similar
to Young's modulus. 36
We have already referred in Chap. V (Sect. 3.6) to the paper communicated
by Cauchy to the French Academie des Sciences in September 1822. Apart from
introducing the idea of stress, as mentioned previously, this paper also ciearly laid
out the notion of deformation characterised by its 6 components or by the principal
axes of deformation and the corresponding principal stretches. From the equilibrium
equations expressed in terms of stresses, Cauchy hoped to arrive at the displace-
ment equations governing the equilibrium of the elastic solid. To this end, he used
the stress-strain relation whieh he obtained, for isotropic materials, by assuming
linearity and also that the principal axes of stresses and strains coincide. Introduc-
ing 2 material constants in this way, Cauchy produced the equilibrium equations
for an elastie solid expressed in terms of displacements. 37 In a subsequent extension
of this work, Cauchy became interested in crystalline solids, making the hypothesis
that material points interact via attractive and repulsive forces. In a first paper
treating this topie, Cauchy was able to retrieve the equations for the isotropic case
that he had found previously (with 2 material constants). In a second paper shortly
afterwards, he arrived at a stress-strain relation with a single elastic constant in
the isotropie case and also the equations discovered by Navier. For the general
anisotropie material, Cauchy found 21 material constants, of which 6 represented
initial stresses, leaving just 15 elastic constants.
We should also mention papers presented at the French Academie des Sciences
by Poisson. These were based on a molecular theory and retrieved Navier's equa-
tions.

The Energy Approach


It was Green 38 who first obtained the elasticity equations from energy considera-
tions. "In whatever way the elements of any material system may act upon each
other, if all internal forces exerted be multiplied by the elements of their respective
directions, the total sum for any assigned portion of the mass will always be the ex-
any substance is a column of the same substance capable of producing apressure
on its base which is to the weight causing a certain degree of compression, as the
length of the substance is to the diminution of its length". (Quoted by S. Timo-
shenko in History of the Strength of Materials). The complexity of this statement
explains J.E. Gordon's remark in The New Science of Strong Materials: "Young
published the idea of his modulus in a rather incomprehensible paper in 1807
after he had been dismissed from his lectureship at the Royal Institution for not
being sufficiently practical. Thus perhaps the most famous and the most useful
of all concepts in engineering, was not generally understood or absorbed into
engineering practice until after Young's death."
35 C. Navier (1785-1836), in a paper read at the French Academie des Sciences in
1821, published in 1827.
36 The equation obtained was not the one given in Chap. VIII (Sect. 5.3).
37 Exact equations given in Chap. VIII (Sect. 5.3).
38 G. Green (1793-1841).
348 Chapter VII. Thermoelasticity

act differential of some function." Assuming that this function might be expanded
in powers and products of the strain components, and restricting to second order for
small deformations, Green obtained the constitutive equations for linear elasticity
with the 21 elastie constants in the general anisotropie case (see Sect. 5.2 above).
Note that the quest ion of the number of elastie constants (1 or 2 for an isotropie
material, 15 or 21 in the general case) was long the subject of controversy, until
undisputed experimental results and sound theoretieal considerations finally led to
rejection of the thesis upholding a reduced number of constants. It was thus that in
the isotropie case, for example, Navier's molecular theory led to the unique value
of 0.25 for the Poisson ratio, agreeing with the theories of Cauchy, Poisson, etc.,
but disagreeing with experience for a very large number of materials.
Summary of Main Formulas 349

Summary of Main Formulas

• Fir. t law 0/ thennodynamics


•• 0
E +K = p(e)(!l.) +Q
E = Q -P(i) (!l.)
Q= - { ~ .!! da + ( rdS1t
ian. in,
Energy equation
pe = g : g+ r - div ~

0/ thermodynamies
. 1q·n + 1
• Second law

S >- =--da r
-dS1,
- an, T n, T

Fundamental inequality

p. + div( ~) - ~ ~ 0
. . q
g: g - p(1/I + sT) - T .grad T ~ 0
. . 20
~: ~ - Po(1/I + sT) - T . VT ~ 0

• Themwelastic constitutive law (with internal constraint )


<pp(~)=O, p=l, ... ,n 1::;n::;6
&tp(T,~)
s=- f1I'
&tp(T,_) 8<pp(i)
~ = Po 8~ + 'Ip 8r. - ('1P arbitrary scalars)
1/1 and <pp symmetrie in e.] and c]'
Phy ical !in arisation
~ = ~o + ~ :~ _~T
11"0] = 11"~ +" A'}kl elk - k.] T
A,}kl = A,}lk = A}'k! = A},fk = Akt,}

lsotropy
~= 1I"0! + >. (tr ~) ! + 2 Jt ~ - k T!
350 Chapter VII. Thermoelasticity

• Infinite. lmal tmn4ormatiofl. otropic matenal


Gl'Ometrical lillt'arisatioll

g = ~" + ~ (tr ~/)! + 211 ~' - k T'!


+
~ = -y- g - E
I ) V I V (')
tr g ! +0 T
I
!
g' = g - gP

E - (3~+21l) ~ k
v= 0 = .,.......,,--...,...--:-
(~+ 11 ) + 21l)
~--:-
J - JA 2(~+1l) (3~

~=E v E E
(1 + v)(l - 2v) 11 = 2(1 + v} k =0 (1- 2v)

3K = a~ + 211 = (1 !2V)
3~ + 2JA > 0 I' > 0
1
E>O -l<v<i
Exercises 351

Exercises

VII.1 Linear Thermoelasticity. Write down the linear thermoelastie


eonstitutive law for an isotropie material under infinitesimal transformations
starting from an arbitrary isotropie initial stress.

Solution
~ = Jr 0 ~ + >.(tr ~) ~ + 2!L~ - kT ~.
To first order in the infinitesimal transformation, this yields the farm of (5.29) for
an isotropie material:
g = Jr 0 ~ + (>. -Jr°)(tr g;)~ + 2(!L+ Jr0)g; - kT~,
or alternatively,
g = 17° ~ + >.'(tr g;)~ + 2!L'g; - kT~.

Remarks
In the Eulerian representation, the eonstitutive law has the same form as in the
Lagrangian representation, but the eoeffieients are modified by the initial stress (see
Seet. 5.2). If the initial state is quasi-natural, we have to first order >.' = >. and
!L' = !L as in the ease of an unstressed initial state.

VII.2 Second Order Thermoelasticity. Consider a thermoelastie mate-


rial that is isotropie in the referenee eonfiguration and free from any internal
eonstraints. Write down the thermoelastie eonstitutive law for this material
to seeond order.

Solution
The polynomial expansion of po 'I/J to third order is, taking isotropy into aeeount,
°, >.( , 2 , , 1
po'I/J=Jr 11-POSOT+2 11) +2!L12-kl1T-"2pobT
2

+ :3a (1')3
1 + ß I'1 I'2 + 'Y I'3 + "20 (1')2
1 T + E I'2 T + <P I'1 T 2 + 'TJ T3 .

Henee, ~ = (Jro+>'I~ -kT+a(In2+ßl~+0I{ T+<PT2)~+(2!L+ßl{ +cT)~+T~2 .


There are 9 seeond order thermoelastic eonstants for this isotropie material, includ-
ing 5 isothermal elastie eonstants.

VII.3 Incompressible Material. Study the internal eonstraint imposed


by volume invarianee also referred to as ineompressibility of the material.
Express 'P(~) and determine the ineffective tensors.

Solution
• Expression far <P(f):
the restriction on dclormations is det E = 1;
using the expression for (det t:/
give~ in Chap. II (Seet. 3.3), we obtain for the
internal eonstraint
<p(~) = l{ + (In 2 - 21~ + ~(In3 - 4I~ I~ + 4I~ = o.
352 Chapter VII - Thermoelasticity

• Ineffective tensors:
in the Eulerian representation, the invariant volume constraint is written tr ~ = 0,
so that, in the Lagrangian representation, (lk + 2~)-1 : ~ = 0 ;

~~ = (lk + 2~)-1 (whieh can also be obtained directly from <p(~) and the Cayley-

Hamilton theorem).
Ineffective tensors, with Tl an arbitrary scalar:
in the Lagrangian representation,
Tl (lk + 2~)-1 = Tl «1 + 2I~ + 2(ID 2 - 4I~)lk - (2 + 4ID~ + 4~2) ,
and in the Eulerian representation, Tllk (see Sect. 4.6).

VII.4 First Order Thermoelasticity for an Incompressible Isotropie


Material. Write down the eonstitutive law for an incompressible, isotropie
thermoelastie material to first order in 11 ~ 11.

Solution
• The internal constraint was studied in Ex. VII.3. It implies that I~ is second
order in 11 f.11 and equivalent to 2 I~. We expand Po'IjJ in such a way that, after
differentiation, there remain only terms to first order in 11 ~ 11, and ~ is given by

~= po ~~ + Tl ~~ , Tl an arbitrary scalar.
- -
Hence, using the results of Ex. VII.1 and Ex. VII.2,
po 'IjJ = 7r 0 I~ - Po So T + 2 J-L I~ - k I~ T
~ = 7r 0 lk + 2 J-L ~ + Tl (lk - 2~) - k T lk '
where <p(~) = det (lk + 2~) - 1 = 0,
and (lk - 2~) is the first order expression for (lk + 2~) -1 .
We mayaiso write, to first order, ~ = 2(7r° + J-L) ~ + (Tl + 7r 0 - k T)(lk - 2~) ,
so that ~ = 2(7r° + J-L) ~ + welk - 2~), w an arbitrary scalar.

• For the infinitesimal transformation,


g=2(7r°+J-L)~+wlk with tr~=O.

Remark
The term w ~ (w arbitrary) in this equation can be interpreted as a result of the
indeterminacy in the term >.' (tr ~) ~ introduced in Ex. VII.l when tr ~ ---> 0 and
>.' ---> 00. The constant 7r 0 (and theterm 7r 0 ~) no longer have the me~ning of an
initial stress (whieh justified the notation in (5."3), (5.12) and Ex. VILl). Indeed, for
f. = 0, the initial stress is an isotropie tensor left undetermined by the constitutive
equation because of the internal stress. In this first order expression, the constant
7r 0 plays the same role as J-L. This is consistent with the fact that we can take account
of the internal constraint expressed in Ex. VII.3 (see Seet. 4.3) when writin~ po 'IjJ.
In po 'IjJ for the first order constitutive equation, we can replace the term 7r I~ by
Exercises 353

the equivalent term 2 7r 0 I~. For this reason, the thermoelastie constitutive law for
an incompressible material can be reduced, to first order, to
g = 2 f1 ~+ro Jk where tr ~ = 0, ro an arbitrary scalar, with the condition f1 > 0 .
The fact that the roles of 7r 0 and f1 are confused in the express ions for zr: and Q. is
clearly related to the infinitesimal character of the deformation (cf. Ex. VII. 11 f
It is interesting to note that, in the formulation chosen here, the internal con-
straint of incompressibility applies to the deformation of the material, whatever
its cause. In other words, the material is both mechanically incompressible and
thermally undilatable. This agrees with our intuition concerning the stability of
the material and whieh is confirmed by analysis. It follows that, taking the limit
(A -+ 00) in (5.18) for an incompressible material, the coefficient k remains finite.

VII.5 Inextensibility. Consider a material whose microstructure imposes


length invariance (also called inextensibility) in some material direction.
Study the consequences of such an internal constraint.

Solution
The internal constraint is defined in the Lagrangian representation. Letting X de-
note the relevant direction,
<p(~) = 0 {o} e xx = 0 .

The ineffective tensors take the form: 'f] ~~ = 'f] ~x ® ~x' 'f] an arbitrary scalar.

VII.6 Hexagonal Symmetry. Let R = (O'~1'~2'~3) be orthonormal co-


ordinate axes. Consider a linear elastic material whose microstructure induces
the following material symmetries in the reference configuration: symmetry in
the plane (~1'~2)' and isometries in the invariance group of a regular hexagon
in the plane (~1' ~2)' Determine the number of linear thermoelastic constants
of this material.

Solution
The material is orthotropic. g, Q, ~3 define the intersections of the orthotropic planes,
where g is a radius vectar of the hexagon, and Q the apothem orthogonal to it.
We must make use of the ternary symmetry to reduce still furt her the number of
constants. Equation (4.40) is written relative to the basis (g, Q, ~), far the isometry
of rotation through an angle <p = 27r /3 about ~3' We obtain the furt her relations:
A 13 = A 23 , All = A 22 , A 44 = A 55 , All - A 12 = A 66 , k 1 = k 2 ,
which are identieal to those obtained for a transversely isotropie material, so that
we end up with 7 independent linear thermoelastic coefficients (5 elastie and 2
thermal constants).

Remark
Although the hypothesis is weaker than transverse isotropy, it turns out that the
linear elastie material studied is actually transversely isotropie. In order to define
the orientation of the material, it is therefore sufficient to specify ~3' that is, we
need only specify 2 angles. The same conclusion is reached for any higher order
symmetry about ~3' This result cannot be generalised to other types of constitutive
law.
354 Chapter VII - Thermoelasticity

VII. 7 Square Symmetries. Let R = (0, Q, !u;;) be orthonormal eoordinate


axes. Consider a linear material whose symmetries in the referenee eonfigura-
tion are: symmetry in the plane (Q, Q), and isometries in the invariance group
of a square with sides Q and Qin the plane (Q, Q). Determine the number of
linear thermoelastie eonstants of this material.

Solution
The material is orthotropic with respect to the planes (g, Q) , (Q, f) , (f, g). We
must make use of the other invariance properties of the square: invariance under
rotation 'P = k 7r /2 and symmetry in the diagonal. One rotation or one symmetry
is sufficient.
Equation (4.40) is expressed in the basis (g, Q, f), for such an isometry. We obtain
the furt her relations A 13 = A 23 , Au = A 22 , A 4 4 = A 55 , k 1 = k 2 .
There remain 8 independent linear thermoelastic coefficients, including 6 elastie
and 2 thermal constants.

Remark
Note that, in contrast to the transverse isotropy case, the shear modulus in the plane
(g, Q) remains an independent elastic constant. From the mechanical standpoint,
we must give the 3 Euler angles of the basis (g, Q, f) to define the orientation of
the material relative to arbitrary co ordinate axes, whilst from the thermal point of
view, it is enough to specify the orientation of f (i.e., 2 angles).

VII.8 Cubic Symmetry. Let R = (0, Q, Q, f) be orthonormal eoordinate


axes and eonsider a linear thermoelastic material whose mierostrueture in-
duees the following material symmetries in the referenee eonfiguration: isome-
tries in the invarianee group of the eube with sides Q, Q, f. Determine the
number of linear thermoelastic constants of this material.

Solution
Applying the arguments presented in Ex. VII. 7 to the squares with sides (Q, f)
or (g, f), we obtain the following extra relations, in addition to those for the or-
thotropic material:
Au = A 22 = A 33 , A 12 = A 13 = A 23 , A 44 = A 55 = A 66 , k 1 = k2 = k 3 . There thus
remain 4 independent linear thermoelastic coefficients, including 1 thermal and 3
elastie constants.

Remark
There remain 3 isothermal elastic constants instead of the 2 for an isotropie mater-
ial. Isotropy corresponds to the extra relation A 44 = (A u -A 12 ), which is obtained
in the same way as for a transversely isotropie material (5.64). From a mechanical
standpoint, the 3 Euler angles of the basis (g, Q, f) must be given in order to define
the orientation of the material relative to an arbitrary set of coordinate axes, whilst
from a thermal point of view, the material is isotropie.

VII.9 Other Formulations of the Thermoelastic Constitutive Law.


Consider a thermoelastie material without internal constraints. Maintaining
the tensor notation used in (4.2), with the proviso explained in Sect. 4.1, let
'Ij; denote the function defined by
Exercises 355

VT, VF, '!fJ(T, F) = 'lj;(T, !~) .


In terms of &'!fJ(T, F)/&F, give the expressions for!L and B corresponding to
F by the thermoelastic constitutive law. Consider 'the case of materials with
internal constraints, and in particular, incompressible materials.

Solution
• Using the general definition (4.11a) applied to ~~ , vr, Vf;,:
. o1jJ· o1jJ· . o'lj;· o'lj;. . t' t'
1jJ = oT T + oE : t;. and 'Ij; = oT T + Of : ~ where 2f= E.E+ E·E·

Now we say that vr, Vf;, : ;p = ?j; ,


o1jJ(T, E) o'lj;(T, f)
whenee oT = oT
. o1jJ· o'lj; t' o'lj; t •
and Vt;., 2 oE : t;. = Of (~.t;.) + Of : (~.t;.).
It follows that

. 2 ot;.
Vt;., o~· t)
o1jJ : t;.. = ( o'lj; ~ : t;.. + (t (o'lj;)
o~· t)~ : t;.. ,

and henee, ~~ = ( ~~). . tt;. .


Then, with the eonvention for the symmetrie expression for 'Ij;(T, ~),
o1jJ(T,E) o'lj;(T,f) t
oE - = Of . t;..
. . o1jJ(T,E) t - 1 . . -1 t(01jJ(T,E))
It lS now ObVlOUS that ot;. - . ~ lS symmetrIe and equal to ~. ol!;. .

• From
-Q. = ~ E .E:.
po- - tE- , the thermoelastic eonstitutive law relating -Q. and -E is
o1jJ(T,E)
g=p~. oE-'

• Likewise, from .ft. = t;.. 7J. (Chap. V, Seet. 4.2),


t(01jJ(T,E))
.ft. = po ot;. and ~ : t;. = 7J. : ~.
t ' .

• In the same manner, internal eonstraints are eharaeterised by funetions 'Pp of t;.
such that
'Pp(t;.) = 0 {o} rpp(~) = 0, p = 1, ... , n .

This yields g = pt;.. ~~ +1]p (det t;.)-1 t;.. O'P;l~) with 'Pp(~) = 0, p = 1, ... , n,
356 Chapter VII - Thermoelasticity

and also 14 = po t (81/J)


8~ + 'f/p
t (8t.pp)
8~ ('f/p arbitrary sealars ).

• For the ineompressible material, the internal eonstraint is written in the form
t.p(~) = det ~ - 1 = 0 .

d(det F) 1
From Chap. III, Seet. 3.5, we have dt - (det ~)- = tr g= ~: ~-1 ,

whenee ~~ = (det ~) ~-1 and

g = p~.
81/J + 'f/Jk
8E with det E = 1, 14= po t (81/J)
8~ +'f/~
-1

('f/ arbitrary sealar).

VII.IO Isotropie Thermoelastic Material: Principal Axes of Stress


and Principal Stresses. Consider an isotropie thermoelastie material with-
out internal eonstraints. Show that the prineipal axes of ~ ean be found from
the prineipal axes of JI. by eonveetive transport and give the relation be-
tween prineipal stresses (Tl, (T2, (T3 of ~ and prineipal stresses 7rl, 7r2, 7r3 of JI..
Show that for this material, 'IjJ(T, F) defined in Ex. VII.9 ean be written as
asymmetrie function "iß(T, Al, A2, A3) of the prineipal stretehes and give the

expressions for (Ti and 7ri as functions of ;~. Consider the ease of isotropie
materials with internal eonstraints, and in particular, ineompressible mate-
rials.

Solution
• Beeause of the isotropy of the material, 1!: has the same prineipal axes as f. Let
fl., }C, W be the eorresponding unit veetors, so that 1!: = 7rl fl. ® fl. + 7r2 }C 0}C +
~W®W. -
:!J., Q, :!Q obtained by eonveetive transport from fl., }C, W ean be written
:!J.=~.fl., Q=~.}C, :!Q=~.W.

From sr = .!!...
-
E . 1!: . t - , we find sr
p o - -E -
= .!!...
po
(7r1 :!J. ® :!J. + 7r2 Q ® Q + 7r3:!Q ®:!Q) ,

which shows what was required eoneerning the principal axes.

• We have I:!J.I = Al, 111.1 = A2 , (principal stretehes) ,


A3
0"3 = 7r3 Al A2 .

• As the material is isotropie, 1/J(T, E) = 'Ij;(T, f) is a funetion ofthe invariants of f


or, ifwe prefer, a symmetricfunetion-;;fthe prin~ipal values of f, viz., ei = (Ar -1)/2~
It is therefore asymmetrie funetion of the Ai: -
Exereises 357

1f;(T,f) = -J;(T, el, e2, e3) = :fij(T, Al, A2, A3) .


Let eij be the eomponents of ~ in its principal basis Jl, J-::., W frozen at time t.
Caleulating j~, j~, j~, it follows that
el1 + e22 + e33 = el + e2 + e3 ,
el el1 + e2 e22 + e3 e33 = el el + e2 e2 + e3 e3 ,
(eI? el1 + (e2)2 e22 + (e3)2 e33 = (eI? el + (e2)2 e2 + (e3? e3 ,
whieh shows that el1 = el, e22 = e2, e33 = e3 .

Sinee ~= po ~~ has prineipal axes Jl, J-::., and W, we may deduee that

O-J; 1 o:fij \ o:fij. p 1


and henee, 7ri = po oei = Ai po OAi an d (Ti = Ai P 0 \ slnee - = \ \ \
Ai po Al A2 A3

• Beeause the internal eonstraints must be isotropie, we have similarly


'P p([;) = <pp(Al, A2, A3) asymmetrie function of Al, A2, A3. It follows that

7ri
1 (o:fij
= Ai po OAi + T/p [)<Pp) .
OAi ,T/p arbltrary sealars, and

o:fij p O<Pp) P 1
(Ti = Ai ( POAi + T/p po OAi ' where po

• For an ineompressible material, this internal constraint is expressed by


<p(Al, A2, A3) = Al A2 A3 - 1 = O.
1 o:fij 1
Henee, 7ri = Ai po OAi + T/ A2
1
'

VII.11 Modelling the Elastic Behaviour of Rubber. Variousmodels


have been proposed to explain experimental observations made of the isother-
mal elastie behaviour of highly strained rubber. The material is eonsidered
to be isotropie and ineompressible with (see Ex. VII.lO for the notation):

Neo-Hookean model Po'l/J = J-lli ,


Varga model Po ~ = + A2 + A3 - 3) ,
2J-l (Al
Mooney-Rivlin model Po'l/J = wO Ii + ~(I? - 2I~) ,

Ogden model

Po ~ = J-ll (Ar l + A~l + A~l - 3) + J-l2 ()..<};2 + A~2 + A~2 - 3)


al a2
+ J-l3 (Ar3 + A~3 + A~3 - 3) .
a3
Write down the eorresponding eonstitutive laws.
358 Chapter VII - Thermoelastieity

Solution
As the isotropie material is ineompressible, we ean use the results in Sect. 4.3
and Exs. VII.3, VII.9 and VILlO. cp(~) det (~+ 2~) - 1, <p(.f:) detE - 1,
'ip('\'1, '\'2, '\'3) ='\'1'\'2'\'3 - 1.

• Neo-Hookean model.
~ = 11~ + 1] (~+ 2~)-1 with cp(~) = 0, 1] arbitrary sealar.

g = 11.f:.. t.f:. + 1] ~ with det E - 1= 0,


- 11(,2 2 2) ,2
Po'l/J = 2' Al + '\'2 + '\'3 - 3 ,O'i = 11 Ai + 1]

• Varga model.
1 1
7l'i = 211 '\'i + 1] ,\,2 with '\'1'\'2 '\'3 = 1, 1] arbitrary sealar, O'i = 211'\'i + 1] .
1

• Mooney-Rivlin model.
~= 7l'0 ~+ 00 I~ ~ - oo~ + 1] (~+ 2 ~)-1, with cp (~) = 0, 1] arbitrary sealar.
-1 0 222 00222222
po'l/J = 2(7l' - 00)('\'1 +'\'2 +'\'3 - 3) + 4('\'1'\'2 +'\'2'\'3 +'\'3'\'1 - 3) ,

or, taking the internal eonstraint '\'1 '\'2 '\'3 = 1 into aeeount,
- 1 0 2 2 2
Po'l/J= 2(7l' -00)('\'1+'\'2+'\'3- 3 )+4
00(1,\,2 1 1 )
+ ,\,2 + ,\,2 -3 ,
123

7l'i = (7l'0 - 00) - ~ ,14 + 1]' ,\ ' with '\'1'\'2'\'3 = 1, 1]' arbitrary sealar,
2 Ai Ai

o 1
0"
1
= (7l' -
2 00
oo)'\'·1 - -2 -,\,~ + 1], .
For the rubber studied by Mooney, 7l' = 0.502 MPa, 00 = 0.136 MPa.

• Ogden model
O'i = 111 ,\,~' + 112 ,\,~2 + 113 ,\,~3 + 1], '\'1'\'2'\'3 = 1, 1] arbitrary sealar.
The Ogden model based on experiments earried out by Treloar eorresponds to

111 = 6.3 Mpa, 112 = 0.012 Mpa, 113 = - 0.1 Mpa,


001 = 1.3, 002 = 5, 003 = -2.

• Linearisation of these laws for infinitesimal transformations.


Neo-Hookean model: g = 211 ~ + ro ~, with tr ~ = 0, ro arbitrary sealar.
Varga model: g = 211 ~ + ro Jl:, with tr ~ = 0, ro arbitrary sealar.
Mooney-Rivlin model: g = 2( 7l'0 - ~) ~ + ro Jl:, with tr ~ = 0, ro arbitrary sealar.
Ogden model: g = (001111 + 002112 + 003113) ~ + ro Jl:, with tr ~ = 0, ro arbitrary
sealar.
Exercises 359

Remarks
For each of the various laws, linearisation clearly leads to the single form established
in Ex. VII. 4, viz., g = 2 j.J ~ + r:v Jk. (For the Mooney-Rivlin model j.J = 1r0 - 0/2
and for the Ogden model 2 j.J = 01 j.J1 + 02 j.J2 + 03 j.J3 .) The present exercise was
inspired by the article Sur les densites d'energie en elasticite non lineaire . .. by
J.L. Davet in Annales des Ponts et Chaussees 35, 1985, pp. 2-33. In particular,
this article considers the experimental validation of the various models. (See also
Ex. IX.8.)

VII.12 Incremental Formulation of the Constitutive Law. Write the


relation between the (Eulerian) strain rate and the intrinsic derivative of the
stress {l (see Chap. VI, Sect. 5.2) for the isothermal evolution of a thermoelas-
tic material. Examine the particular case of an infinitesimal transformation
from the natural or quasi-natural initial state.

Solution
• The material derivative of the constitutive law in an isothermal evolution is
~ = (o~(T,rg)/o,!) : ~.
Using (4.47), ~ = tK:..g.K:. and (5.6) in Chap. VI, (Dg/Dt)T = r 1 K:..~. ~,
we obtain (Dg/Dt)T =r 1 K:.. (o~(T,rg)/org: (~.g.K:.». ~.
Put 4 (T,rg) = o~(T,rg)/org = PO(021jJ/orgorg), which has the same symmetries (5.5)
and (5.6) as 4 in the case of linear elasticity, and define ~ (T, K:.) = ~ijuv fi ® fj ®
- -
f u ® f v in an-orthonormal basis by Aijuv(T, K:.) = J- 1 Fi:FjnFupFvqAmnpq(T, rg) .
~ (T, K:.) has the same symmetries (5.5) and (5.6) as 4. The equation relating
(Dg/Dt)T to g is then
(Dg/Dt)T = ~ (T, K:.) : g, a linear relation between (Dg/Dt)T and g . Note that
g = °implies (Dg/Dt) T = °,that is, g then undergoes the same rigid body motion
as the material element.

• When the transformation is infinitesimal and starts from the natural or quasi-
natural initial state, we have
(Dg/Dt)T = g- grad U.g - g. t grad U co:: g,
~(T,~) co::~ (=4(T,O»,

gco:: ~.

Hence, !i co:: A : ~ .

Remarks
The above calculation shows how the Truesdell rates (or Jaumann rates by the
linear relation (5.13) in Chap. VI) arise naturally in the incremental formulation of
the constitutive law.

You might also like