G.M. de La Penha, L.A. Medeiros (eds.) Contenporary Developments i n Continuum Mechanics and P a r t i a l D i f f e r e n t i a l Equations Worth-Holland Publishing Company (1978)
NON LINEAR CONSTITUTIVE RELATIONS AND HOMOGENIZATION
L. TARTAR University of Paris
-
Sud,
France
1. Introduction. - A few years ago there were two classes of methods to
handle non linear partial differential equations arising i n Continuum Mechanics o r Physics: the first one used a compactness argument and started with the w o r k of Leray, the second one a monotonicity or convexity argument initiated by Minty and Zarantonello and developped by BrGzis, Browder, Lions and others (cf. Lions Ell).
Although Leray's work was motivated
by Mechanics (cf. Leray [l])
the origins of the other method
were numerous but its applications to Mechanics were soon recognized and investigated (cf. Duvaut-Lions [l]).
It was
known, but not really emphasized, that these methods were inadequate for most
practical problems.
After a few years of darkness light came with the work of Ball [l]; adding ideas from homogenization gave rise to a new concept: compensated compactness (cf. Murat El], [l])
Tartar
and then t o the principles of the method presented here
(cf. Tartar [ a ] ) . Weak convergence plays an essential role and will be related to microscopic and macroscopic properties.
This
method seems perfectly adequate f o r non linear partial differential equations coming f r o m Mechanics and Physics (it was
NON LINEAR CONSTITUTIVE RELATIONS A N D HOMOGENIZATION
developed
473
with this purpose) but the approach is philosophic-
ally quite different from the classical one (except for Statistical Mechanics and Quantum Mechanics perhaps). Although the method could be used for stationary
or
evolutionary equations, it cannot yet be used for nonlinear hyperbolic systems: indeed one has first to know what a solution is, the concept o f entropy (which is not completely understood) being a n important restriction added to the equations.
We will avoid this problem here and accept all
solutions (in the sense of distributions) of the equations that we will consider. We will work on partial differential equations and not o n boundary value problems, but the method can o f course be
used to treat non linear boundary conditions.
2. Microscopic-macroscopic properties and weak convergence.
Partial differential equations are used as a mathematical model to describe the values o f physical quantities and
their adequacy is tested through prediction
o f the output of an experience knowing
the input.
Although there is no mathematical difficulty to work with functions there is a physical impossibility t o know everywhere its values (except for analytic cases); measurements o r identifications of some type will be used giving a finite number of parameters and from them an approximation of the function will be derived.
There is no a priori reason
that this approximate function satisfy any equation.
Of
course a large number of measurements may give a better know-
474
L.
TARTAR
ledge of the function but one is usually forced to work above some length (or time) scale in order to avoid perturbing the phenomenon under study (cf. Heisenberg principle of incertitude in Quantum Mechanics). The information one can obtain by measurements is called macroscopic, but the equations are usually valid for a physical quantity which is microscopic.
There is an implicit
belief that, if the phenomenon has been well analysed, all the relevant quantities appear in the equations and the macroscopic quantities will satisfy the same equations as the microscopic ones.
[It is my belief that some classical
equations have not been derived with sufficient care]. T h e numerical analysis approach is similar: to compute the solution of a n equation an approximation, depending o n a finite number of parameters, will be derived;
by increasing
the number of parameters it is hoped that this approximation converges(in
a weak o r strong sense) to the solution.
If the
macroscopic quantities were satisfying a different equation than the microscopic ones the limit of numerical approximations will usually satisfy the macroscopic equations (this fact will in this case depend upon the numerical method used).
Weak convergence seems to be the appropriate mathmatical tool to handle the above situation.
As,
o n bounded sets of
functions, the weak topology can usually be defined by a metric we will interpret the above analysis by saying that if two functions are nearby for the weak topology they are almost undiscernable through measurements.
NON LINEAR C O N S T I T U T I V E R E L A T I O N S A N D H O M O G E N I Z A T I O N
jt7
5
A s a n example c o n s i d e r a measurement o f t h e e l e c t r i c field
u
E(x)
i n a conductor.
is the electrostatic potential.
x0
w i l l c o n s i s t i n measuring
example
u(xo)
and
u(xo+h),
Suppose t h a t
u(x)
if
0.
E
then
x
C
ul(:)
u(x)
-
h
0’
for
E(xo) the
and t h a t t h e
where 1.
u
is
Within t h e
w i l l be i d e n t i f i e d with which i s a l m o s t
duo du1 x - dx (F); dx compared
to
-
uo(x)
-
the difference
du1 x - z(z)
- b u t i t s mean dx
0
i t is quite f a r f r o m i n a weak t o p o l o g y
X
i n any s t r o n g
0
(if
cp
T ( ~ ) c o n v e r g e s weakly t o t h e mean v a l u e o f to
near
T h i s d i f f e r e n c e i s a l m o s t i m p o s s i b l e t o measure
topology b u t n e a r
goes
is
i s t o o small;
6
= uo(x) +
11
important v a r i a t i o n s
value i s
a t two p o i n t s n e a r
and t h u s d e r i v e f o r
uo ( x o ) - u o ( x o + h )
with
But t h e e x a c t has
E
i s periodic with period
u1
a c c u r a c y o f measurement
E(x)
where
A measurement of
i s s m a l l compared t o
C
exact p o t e n t i a l i s
and
(x)
u(x,)-u(x,+h) h
approximation
smooth and
u
- du d x
E(x) =
We h a v e
i s periodic
cp
when
G
0).
I t seems t h a t a l l measurements a r e done t h r o u g h a v e r a g e s of p h y s i c a l q u a n t i t i e s ;
from t h e s e measurements
i d e n t i f i c a t i o n o f p h y s i c a l parameters
can b e o b t a i n e d :
other
n o t i o n s of c o n v e r g e n c e a r e r e l a t e d t o t h i s f a c t as i n homogenization
(cf. Tartar
11 )
.
The r e a d e r w i l l h a v e remarked t h a t
t h e above
c o n s i d e r a t i o n s a r e more p h i l o s o p h i c a l t h a n m a t h e m a t i c a l : u s i n g a compactness r e s u l t weak c o n v e r g e n c e i n some s p a c e may i m p l y s t r o n g c o n v e r g e n c e i n an o t h e r ;
as t h e above a n a l y s i s
d e p e n d s on t h e s p a c e u s e d i t h a s no i n t r i n s i c v a l u e arid e a c h
47 6
L.
TARTAR
m a t h e m a t i c i a n w i l l want t o c h o o s e i t s p r e f e r r e d s p a c e .
But
e q u a t i o n s w i t h d i s c o n t i n u o u s c o e f f i c i e n t s and h o m o g e n i z a t i o n show t h a t t h e r e a r e w e l l d e f i n e d s p a c e s a s s o c i a t e d t o a g i v e n p a r t i a l d i f f e r e n t i a l e q u a t i o n coming f r o m Mechanics or Physics;
o u r a n a l y s i s r e l i e s on t h i s f a c t .
3 . A p p l i c a t i o n s t-.o. __. n o.____ nlinear e l a s t i c i t y . The p a r t i a l d i f f e r e n t i a l e q u a t i o n s g o v e r n i n g t h e motion of a n e l a s t i c body a r e ( c f . G u r t i n [ 11 )
(1) (2)
bo(x,t) = (det F ( x , t ) ) b ( x , t , r ( x , t ) )
(3)
Fij
(4)
S
(5)
det F 7 0
--
i n general
a ri
ax. J
= :(F)
A
The f u n c t i o n a l
S
(6)
satisfies
;(F)
(7)
F~ = F
G(QF) = Q$(F)
for a l l for a l l
F
F
and
Q E Orth'
Of c o u r s e t h e r e a r e u s u a l l y some b o u n d a r y c o n d i t i o n s . Remark 1: I f
one adds t h e s t r o n g e l l i p t i c i t y c o n d i t i o n on t h e
e l a s t i c i t y tensor
A(F)
= DS(F):
( a @ b ) * A ( F ) ( a0 b ) > 0
for
a 0 b
f 0,
t h e above s y s t e m becomes h y p e r b o l i c a n d , by a n a l o g y w i t h other
more or l e s s u n d e r s t o o d s i t u a t i o n s , o n l y p a r t i c u l a r s o l u t i o n s
477
NON L I N E A R C O N S T I T U T I V E R E L A T I O N S AND I I O M O G E N I Z A T I O N
of this system are believed to be physical: some kind of
inequalities, called entropy conditions, are added;
at the
moment only a few examples are understood and the above system seems out of reach. Quite naturally one is
led to consider the stationary
equations, where functions only depend on stable
x.
stationary solutions will be observed;
Of course only but stability
involves the complete system so we prefer to forget about this point. A more curious point is that, when dealing with
stationary solutions, nobody
thinks of restricting the class
of solutions with entropy conditions as if they were automatically satisfied f o r stationnary discontinuities;
to be
sure o f this point one should know what these entropy inequalities are and this is not the case, but as for some simple hyperbolic systems the analog is false we have to be careful.
In order to avoid this question we add the follow-
ing Postulate: all discontinuous solutions (F or S, not of the stationary
r)
system of elasticity are accepted.
F r o m this and the philosophical approach of paragraph
2 we will derive a necessary condition on the function
A
S;
it is an interesting fact that this condition implies that stationary discontinuities (along a smooth surface) satisfy the entropy inequalities whatever they are (because they cannot rule out discontinuities in the linear case).
Of
course this does not prove (8).
If we do not want to accept (8) we may as well work
47 8
L.
TARTAR
d i r e c t l y w i t h t h e c o m p l e t e e v o l u t i o n problem.
Presumably we
have t o do s o f o r e l a s t i c f l u i d s . Our p h i l o s o p h i c a l a p p r o a c h t e l l s u s t h a t a weak l i m i t
(1) t o ( 5 ) i s a l s o a s o l u t i o n .
of stationnary solutions of
T o e x p r e s s t h a t we f i r s t h a v e t o p r e c i s e what weak t o p o l o g y
we u s e . I n l i n e a r e l a s t i c t h e o r y , u s i n g v a r i a t i o n a l methods and d i s c o n t i n u o u s c o e f f i c i e n t s , we know t h a t a n a t u r a l s p a c e is
~ ‘ ( n ) , sij
F.. E 1J
E L2(61);
by u s i n g m o r e s o p h i s t i c a t e d
r e s u l t s one c a n f i n d t h a t t h e s o l u t i o n s a t i s f i e s a n e s t i m a t e
E LP(n)
Fij,Sij
(2 s
p
f o r some
p 5:
a t the best
+a);
(when
c o e f f i c i e n t s a r e d i s c o n t i n u o u s ) we may e x p e c t a l l f u n c t i o n s
F . ., I J
S
t o b e bounded.
,
i J
We a r e l e d t o s a y t h a t ( a s s u m i n g ( 8 ) ) i f
i s a s e q u e n c e of weak
*
s o l u t i o n s of then
(r,F,S)
to
(rn,Fn,Sn)
( I ) t o ( 5 ) converging i n is a l s o a solution.
(r,F,S)
A
T h i s i s a n i m p l i c i t h y p o t h e s i s on t h e f u n c t i o n motivates
the
Definition 1
,.
-
~ ~ ( 6 2 )
which
S
i s an a d m i s s i b l e c o n s t i t u t i v e r e l a t i o n i f
S
i t s a t i s f i e s the preceding conditions. Let us n o t e f i r s t t h a t t h e only r e a l d i f f i c u l t y i s t o know i f
F
and
theorem o f B a l l [l], det F 2 0;
(5),
s t r o n g l y and
det Fn-det
F
L”
in
I n d e e d , by a
Ig(G)l +
+
det F
det G
As
r
n
4
r
0
w i l l imply
s t r o n g l y we have
bz(x,t)-bo(x,t)
in
L”
*
weak
a n a t u r a l growth c o n d i t i o n
s t a y s bounded and c a r e of
S = g(F).
a r e r e l a t e d by
S
>
+m
0
giving when
G
taking
b(x,t,rn)+b(x,t,r)
weak
*
taking care
479
N O N L I N E A R CONSTITUTIVE RELATIONS AND IIOMOGENIZATION
of
(2).
(1) a n d
('4)
remains;
Thus o n l y a
( 3 ) b e i n g l i n e a r p r e s e n t no d i f f i c u l t y .
Then
result
as f o r
of Murat-Tartar
(6)
note than
,
( c f . M u r a t [ 13
SnFnT--SFT
T a r t a r [ 11
by
).
I t i s n o t h a r d t o d e r i v e a n e c e s s a r y c o n d i t i o n for admissibility Theorem 1
-
(cf. n
If
T a r t a r [ 21 )
is admissible then i f
S
d e t Fi
> 0;
F2-F
F1,
F2
satisfy
5
- 1 8
1 -
T h e n we h a v e
(10)
5((1-e) F 1 + 8 F 2 ) = (1-8) g ( F 1 )
Remark 2 : from
As
F1 t o
( 9 ) i s the Rankine-Hugoniot F
2
A
wise
(a.e)
E
[O,ll.
condition f o r a
5
jump
this
n
s
d i s c o n t i n u i t i e s o c c u r o n l y on
l i m i t o f smooth s o l u t i o n s :
contradictory
for 8
i s a f f i n e a n d t h u s c a n b e o b t a i n e d as p o i n t -
S
a r e believed
€I;(F,)
a c r o s s a n h y p e r s u r f a c e of n o r m a l
s h o w s t h a t for a d m i s s i b l e lines where
+
entropy inequalities
t o hold f o r these solutions s o ( 8 ) i s not ( b u t may n o t b e p h y s i c a l ) .
I S the s t r o n g e l l i p t i c i t y condition holds o n l y o c c u r s for
F1
= F2;
then
presumably s o l u t i o n s of
e q u a t i o n s may b e s m o o t h i n t h i s c a s e
(F
and
S
(9)
the
HBlder
continuous), F o r h y p e r e l a s t i c m a t e r i a l s whose s t o r e d e n e r g y f u n c t i o n a l s a t i s f i e s t h e Legendre-Hadamard necessary
condition, the
c o n d i t i o n o f Theorem 1 i s s a t i s f i e d .
I t i s n o t known i f sufficient o r not.
this necessary condition i s
Some s u f f i c i e n t c o n d i t i o n s f o r actnisiibility
480
L. TARTAR
can be obtained but the main problem remains that it is hard to check on particular examples if they apply.
-
Example 1
F
for all
matrix
A sufficient condition for admissibility is that
(satisfying
MF
det F > 0)
such that
5
(We assume o f course that if
G
there exists an invertible
is bounded and
is continuous and I;(G)I
det G + 0).
thesis corresponds to the case
MF
+
+co
The monotonicity hypoI
I.
T o obtain a wider class of admissible conditions we will use the following important notion which is adapted to equations (1)(3). Definition -_
2
-
A functional
(rn,Fn,Sn) converging in
sequence
and such that then
is admissible if for any
cp(F,S)
Div Sn
cp(Fn,Sn) If
L"
is bounded in
converging weakly to depends only on
to the quasiconvexity o f
ep
F
(and
L" J,
++
weak
to
(r,F,S)
Fn = O r n )
implies
~
z q(F,S).
this notion is equivalent
(cf. Ball [l]).
The exact
structure of these admissible functions is not known but simple examples, generalizing Ball's polyconvex functions,
All quadratic admissible functionals are
can be obtained.
known (cf. Tartar 111 , [ 2 ] ) : if
F = X 8
Example 2
-
5
and
S5
they must satisfy
cp(F,S)
2
0
= 0.
A sufficient condition for admissibility is that
there is a family
(cp,)
a€A
of admissible functionals such
that (12)
s
= :(G)
is equivalent to cp ( G , s ) c Q
o
for all
a E A.
481
N O N L I N E A R C O N S T I T U T I V E R E L A T I O N S AND H O M O G E N I Z A T I O N
Then Example 1 is only a particular case where the functions VU
take the form
Example 3 (PU)u:A
-
.
A sufficient condition is that there is a family
of admissible functionals satisfying cpa(F-G,
(13)
T
c p ( G , S ) = MF(:(F)-S(G))'(F-G)
< 0 f o r all F,G and all
;(F)-:(G))
u E
A
and the maximality condition cpu(F-G,
(14)
;(F)-S)
s
0 for all
F and a implies S = : ( G ) .
This is also a particular case of Example 2 but it will he more convenient to handle homogenization. Remark 3: It is not known if hyperelastic material having a polyconvex stored energy functional (cf. B a l l [l]) have an admissible constitutive relation.
4. Homogenization. Homogeneous materials are very
often heterogeneous at
a microscopic level (we stay o f course far above the molecular level).
If the different components are small enough compared
to the experience scale the material will behave like a homogeneous material. as
(Of course this is the same approach
in paragraph 2). T o avoid technicalities we will work with a material
having a periodic structure of size
6
and assume that there
are no exterior forces; we have functions satisfying
(15)
Div Sc = 0
( r E , F E, s e )
L. TARTAR
482
a Tie
F6. . =
ax. J
IJ
Sc
,(:
=
Fe)
> 0
det F' G
,;(: If when (r,F,S)
QF ) = US(%, F')
g o e s to
E
we expect that
for
Q F Orth'.
(re ,FG , S ' )
0
(r,F,S)
converge weakly to
will satisfy the equations (Of c o u r s e
corresponding to some homogeneous material. does not go to
0
but if
almost undiscernable f r o m
(rE ,FE ,S')
is small enough
E
e
is
-
(r,F,S)).
We will obtain the homogenized constitutive relation S
1
by saying that
(21)
-
S = S(F)
(re ,FE, S @ )
if there exists a sequence
m
satisfying (15)(16)(17) converging i n L
-S
It is a belief that such an
weak
exists: a priori
-S
*
to (r,F,S)
may be
multivalued. Properties (18)(20) f o r
5;
corresponding properties of automatic by its definition. obtain information on
5:
-S
follow easily from
admissibility of
5
is
The only kind of question is to
i f for every
x,
...
S(x,F)
is of
the type of example 1, 2 , 3 ( o r in any other interesting class) what can be said on same family
To.
-S.
F o r the class of Example 3 , if the
is used f o r all
inequality ( 1 3 ) is true for
4 . ,
S;
x,
then the same
on the contrary Example 1
does not seem to b e a good setting for homogenization.
NON LINEAR CONSTITUTIVE RELATIONS AND HOMOGENIZATION
483
5. Comments. Non linear partial differential equations o f Continuum Mechanics or Physics should be stable under some kind of weak convergence, the natural spaces being pointed out by the case o f discontinuous coefficients (heterogeneous materials) and
homogenizati on. The main open problem is related to the so c a l l e d entropy inequalities and consists in asking which are the physical discontinuous solutions of the equations.
If one accepts all weak solutions o f the equations, which we have done here, w e are led to a simple necessary condition and some implicit sufficient conditions which are difficult to check.
An important problem is to derive a
necessary and sufficient condition or at least to give a simple way to check a sufficient condition. The same analysis can be done with boundary conditions and of course writing all this f o r manifolds with or without boundary is left as a good exercise for specialists in translations. Existence theorems are now reduced to construct approximations with a good a priori estimate, the admissibility property dealing with the passage to limit. The philosophical approach used (which is more or less known in Statistical and Quantum Mechanics) is i n opposition
with the classical use of strong topology, implicit function theorem and other local results (in n o r m topology).
484
L. TARTAR
Bibliography
Ball,
-
J.M. [l]
Convexity conditions and existence theorems
in nonlinear elasticity,
63, Duvaut, G.
337-403
-
-
Arch. Rational Mech. Anal.
(1977)
Lions, J.L.
[l]
-
Inequalities in Mechanics and
in Physics, Paris, Dunod 1972 (in French); Springer
1974.
-
Gurtin, M.E. [l]
O n the n o n linear theory of elasticity
in this volume. Leray, J. [l]
-
Etude de diverses Qquations int6grales n o n
lin6aires et de quelques probl5rnes que pose l'hydrodynamique, J. Math. Pures et AppliquGes
a, 1-82
(1933). Lions, J . L .
-
[l]
Quelques d t h o d e s de r6solutions des
probl6rnes aux limites n o n lin6aires.
Paris, Dunod-
Gauthier Villars, 1969. Murat, F. [ 13
-
Cornpacite par compensation, t o appear in
Annali d i Pisa. Tartar, L . [ll
-
HomogenGisation dans les Gquations aux
d6rivGes partielles.
[a]
Cours Peccot
1977, to appear.
Weak convergence in non linear partial differential equations in "Existence theory in nonlinear
elasticity", Austin
1977.