0 then r(0) = 0. The results and comments in this section apply with appropriate modifications to incompressible elasticity, in which the deformation u is required to satisfy the pointwise constraint (6.5 j
a.e. .YE n.
det Vu(x) = 1
The constraint (6.5) is conveniently incorporated into our framework by requiring that the stored-energy function g: M3 x3 --t R satisfy g(A ) = 00 if and only if det A # 1; note that such g are Bore1 measurable but not continuous. If, for example, if
g(A) =fl(vy; + cy + v;j =+a
detA =P,L’?P~=
1,
otherwise,
where p > 0, a > 1, then g is WI+-quasiconvex if and only if max(a,p) > 3. In fact, if max(a,p) > ? then g is W’+ quasiconvex by Theorem 4.5(i). If max(u.p) < 3 and i, > 1 let U(X) = (r(R)/R)x be the radial mapping of B= (1x1 < 11 defined by r(R)=
(R” +A’-
1)’ ‘.
Then u E W’%“(B; IF;‘) and U(X) la* = Ax but g(1l) = 03. jH g(Vu(x)) d.u < co. so that g is not W’.P-quasiconvex. 7. COUNTEREXAMPLES CONCERNING WEAK CONTINUITY OF JACOBIANS
Let II > 1 and let R c R” be open. In the preceding sections we used the fact that the mapping u I-+ det Vu is sequentially continuous from WL3”(Q; R”) endowed with the weak topology to L’(a) endowed with the topology of measures. By means of two counterexamples of a different nature we now show that u ++ det Vu is not sequentially weakly continuous from W’*“(Q; R”) to L’(B), that is, uj--\ u in W’*“(R; R”) does not imply that det Vuj - det Vu in L’(Q). This should be contrasted with the fact that if h is convex and lower semicontinuous then J‘* h(det Vu(x)) dx is swlsc on W’*“(Q; F?“) (Th eorem 4.l(ii)), and the fact that, for example, the functional u H jc, a(x) ]det Vu(x)\ dx is swlsc on I+“-“(R; P”) for all a E L”(Q). a > 0 (see Acerbi and Fusco [3, Theorem 2.41). COUNTEREXAMPLE 7.1. Let B = (xE R”: 1x1 < l}, II > 1. Consider for j = 1, 2,..., the radial mappings rW Uj(X) = *
x,
R = 1x1,
246
BALL AND MURAT
where if
0 < R < l/j,
=2-jR
if
l/j < R < 2/j,
=o
if
2/j 1. Consider the space Rad’.“(B) of radial mappings u(x) = (r(R)/R)x belonging to W’*“(B; IR”), with norm N(r) given by (7.3). Supposing that 0 < R < R < q and using Holder’s inequality with exponents II and n/(n - 1) we have It-“(R) - r”(E)1 1. For x = (xl,..., x”) E Rn lxkl and Q = {x E R”: 11x/I< l} = (-1, 1)“. We let =maxlGkGn
COUNTEREXAMPLE
M
w=+
for all
denote
x E Q.
This function is continuous except at zero and vanishes on 3Q. We extend u as a Q-periodic function to the whole of R”. In Q\{O} u = (u’,..., u”) has distributional derivatives
g
(x)’ St,--
II4
a llxll F=
sign xk
=o
xi a llxll _ 6’ k’ ll-#~
(7.5)
if Il.4 = lxkI, if
11x/I= Ixi/, i# k.
We note that
and show that (7.5) defines the derivatives of u in the sense of distributions in Q (and not just in Q\ (0)). In fact, if 4 E 9(Q) and E > 0,
where ra denotes terms tending to zero as E + 0. Finally, since by (7.6)
it follows that u E W’*p(Q; R”) for all p, 1
1, is not easy to find proved completely in the literature. LEMMA
fE L&,(ip”)
A.1. Let QC IF?“’ be an m-cube. Let 1 < p < co and be Q-periodic. Define&(x) =f( jx). Then as j + co
let
in Lp(ll) for any bounded open subset fJ c Rm. Proof. We suppose without loss of generality that Q = (0, 1)“; in this case a function f is Q-periodic provided f(x + ei) =f(x) for all x and i = l,..., m, where (ei} denotes the standard basis of pm.
250
BALLANDMURAT
Suppose first that fE Lm(IRm). Then 4 is bounded in Lm(G) contains a subsequence f, -* 0 in L”O(J2). We show that B(x) = j f(y)
dy = const.
and so
a.e. x E a.
(A.1)
Q
In fact, let Q, c R be an m-cube with edges parallel to the axes and of side a. If pus > 1 then PQ, is a union of ([pa - 11)” unit cubes of the fundamental lattice and a set E, with meas E, < ([pa + 21)” - ([pa - l])m, where [ ] denotes “integer part.” Thus I!,,
(.twjQf(r)dv)d~~
= lpi,,,
(f(4-jQf(YPY)
dzi
meas E,,
< Cp-”
which tends to zero as p --t co. Thus JQ, (0(x) - J&(Y) dy) dx = 0 for all Q, and (A.l) follows by a density argument. Since 0 is uniquely determined byf it follows that the whole sequence fj - * JQf(y) dy in L”O(R). Next suppose that fE Lo,, with 1
O, lRmp(x)dx = 1. For E > 0 let P,(x)=~~’ E-“‘p(x/&), and let f” = pE*J Then f ‘+ f in L”(Q) and clearly f E is Q-periodic. We claim that fi-+ fj in L”(R) as E + 0, uniformly inj. We have that
Jo If;(x) -fj(x)l”
dx =jPm ho If”(y) -f (y)l’ dy
k
1
QkVYY) -f(y)l”
dy,
where the sum JJk is over cubes Qk = ak + Q of the fundamental lattice, the number of these cubes being less than Cjm. But
jQ, lf"(~> -f (YIP dy = jQ In”
-f (y)l’ dy
since f”, f are Q-periodic. Therefore ]Ifj” -fj]lLe(q) < C ]]f” - f ]lLPcQ,,which proves the claim. Since 7 is smooth we know that f;-* 1 f’(y) dy in L”O(s2) as j + co. Writing 4 = fj +fj - fj” we deduce that 4 - 7Qf(y) dy in Lp(12) as required. 1 COROLLARY A.2. Let Q c I?“’ be an m-cube. Let 1 < p < cq let v E W:,$(Rm; R”) be such that Vu is Q-periodic, and define u/(x) = j- ‘v(jx). Then v.---& (jQ Vv(y)dy) x (-* if-P= a) 3
in W’qp(12;R”) for any bounded open subset l2 c R”.
QUASICONVEXITYANDVARIATIONALPROBLEMS
251
ProoJ Let Q = (0, 1)” and consider the function zi(x) = v(x + ei) - P(X). Since Vz,(x) = Vtl(x + ei) - VU(X) = 0, zi(x) is actually independent of X: let A E I’W’~“’ be defined by Aei = zi(x) and w by w(x) = v(x) -Ax. Then w(x + ei) - w(x) = t’(x + ei) - c(x) - Aei = 0. Hence u’ is Q-periodic. But c’~(x) = Ax + + w(jx), Vcj(x) = A + VW($). in ,!,P(R; IR”), and by the lemma VC __-\ Therefore cj+Ax ,fv (A + VW(Y)) dq’ = IQ VV(JJ) 4~ in LP(R; MnYfl) (-* if P = co), which completes the proof. 1 The final result is a lower semicontinuity
theorem.
PROPOSITION A.3. Let R c iR”’ be bounded open, and let H: 7” + E be convex, lower semicontinuous and bounded below. Let 6,, 8 E L ‘(0; F, ‘) with ei -* 0 in the sense of measures (i.e., ./‘()Bjq4dx + .(‘Ij $4 dx for all continuous functions I$: f2 + IR with compact support in Q). Then
\ H(B(x)) dx < lim inf (_ H(B,(x)) dx. i-a .n . I>
(A.21
Remark A.4. This proposition was proved by Reshetnyak [ 17, p. 805 1 for the case when H depends also on x, but takes only finite values. In our case we are able to give a slightly simpler proof. but still based on Reshetnyak’s idea. (Note added in proof: see also Marcellini 121 1.) Proof of Proposition A.3. We suppose first that H(z) = max, 41s,,, (Ui + @iy Z)) is P’iecewise aftine, convex and nonnegative. For each z E I!‘\ we choose an element A(z) E aH(z). There exists a sequence ck --f 0 a.e. in R and strongly in L’(l2; IR”) with the property that z’~ is constant on each open s-cube Qk,,, = k-‘[m + (0, l)‘], m E Z’. Also, there is a sequence $k E Ci(O1 such that 0 < 4k < 1, #k = 0 in a neighborhood of each aQk,, and tik + I a.e. in R. By the convexity of H,
H(ej(x))> H(u~(x))+ (A(v/ct-x)), e,(X)- u/c(X)) a.e. x E R.
252
BALL ANDMURAT
We multiply
this inequality
by $,Jx) and integrate over ~2 to obtain
!, H(ej(X)) dx> I, #/c(X) ff(ej(X))dx 2 I, $/c(X) H(u/c(X)) dx + !, #L(X)@ (u/c(X)>, ej(X)- u/c(X)) dx. Fix k and let j+ co. Since #,&x) A(n,Jx)) support in Q we obtain
is continuous
with compact
lim inf J H(ej(X)) dx > I, #/c(X) H(u/AX)) dx j-100 0
+ j, !h(-w(%@))~ e(X)- Q(X))dx. Let k-+co. By Fatou’s lemma lim inf,,, j”c #,Jx) H(D~(x)) dx > J”, H(d(x)) dx. S’mce A(.) is uniformly bounded a simple estimate shows that
;;% j, h(X)(wdX))y e(X)- Q(X))dx = 0. Hence we obtain (A.2). Now let H: IRS+ R be convex, lower semicontinuous and bounded below. Any such function is the supremum of a countable family of afftne functions. Assuming without loss of generality that H > 0 it follows that H can be written as the limit of an increasing sequence H, of piecewise affne, convex, nonnegative functions. For each I 1 H/(0(x)) dx < li,m_&f j H,(6Jj(x)) dx < liz&f R R
j
R
H(Bj(x)) dx.
Letting I+ co we deduce from the monotone convergence theorem that 1 H(O(x)) dx < lirn&f R as required.
1 H(Bj(x)) dx R
I ACKNOWLEDGMENTS
We thank P. Marcellini for interesting discussions. The research of JMB was supported by a U.K. Science and Engineering Research Council Senior Fellowship. Parts of the research were also done as JMB was visiting the Universitk Pierre et Marie Curie and then as FM visited Heriot-Watt University as an S.E.R.C. Visiting Fellow.
QUASICONVEXITY ANDVARIATIONALPROBLEMS
253
REFERENCES 1. E. ACERBI, G. BUTTAZZO AND N. Fusco, Semicontinuity in L” for polyconvex integrals, Atti Accad. Nar. Lincei, Rend. Cl. Sci. Fis. Mat. Natur.. in press. 2. E. ACERBI, G. BUTTAZZO, AND N. Fusco, Semicontinuity and relaxation for integrals depending on vector valued functions, J. Math. pures et appl. 62 (1983). 371-387. 3. E. ACERBI AND N. Fusco, Semicontinuity problems in the calculus of variations, Arch. Rat. Mech. Anal., in press. 4. J. M. BALL, Convexity conditions and existence theorems in nonlinear elasticity. ilrch. Rat. Mech. Anal. 63 (1977). 337403. 5. J. M. BALL. Constitutive inequalities and existence theorems in nonlinear elastostatics, in “Nonlinear Analysis and Mechanics: Heriot-Watt Symposium Vol. I” (R. J. Knops. Ed.), pp. 187-241, Pitman, London, 1977. 6. J. M. BALL, Discontinuous equilibrium solutions and cavitation in nonlinear elasticity. Philos. Trans. Roy. Sot. London A 306 (1982), 557-611. 7. J. M. BALL, J. C. CURRIE, AND P. J. OLVER, Null Lagrangians. weak continuity. and variational problems of arbitrary order, J. Funcf. Anal. 41 (1981). 135-174. 8. B. DACOROGNA, Quasiconvexity and relaxation of non convex problems in the calculus ol variations, J. Funct. Anal. 46 (1982), 102-l 18. and weak lower semi-continuity of nonlinear 9. B. DACOROGNA, “Weak continuity functionals,” Lecture Notes in Mathematics, Vol. 922. Springer-Verlag, Berlin. 1982. 10. N. Fusco, Remarks on the relaxation of integrals of the calculus of variations, in “Systems of Nonlinear Partial Differential Equations” (J. M. Ball, Ed.), pp. 401408. Reidel, Dordrecht. 1983. Il. R. V. KOHN AND G. STRANG, Explicit relaxation of a variational problem in optimal design, Bull. Amer. Math. Sot. 9 (1983), 211-214. 12. N. G. MEYERS, Quasi-convexity and lower semicontinuity of multiple variational integrals of any order, Trans. Amer. Math. Sot. 119 (1965). 125-149. 13. C. B. MORREY, Quasi-convexity and the lower semicontinuity of multiple integrals. Pacific J. Math. 2 (1952) 25-53. 14. C. B. MORREY, “Multiple Integrals in the Calculus of Variations.” Springer, Berlin, 1966. 15. F. MURAT, Compacite par compensation II, in “Proc. International Meeting on Recent Methods in Non-linear Analysis, Rome 1978” (E. de Giorgi. E. Magenes. and U. Mosco. Ed.), pp. 2455256, Pitagora, Bologna, 1979. 16. Y. G. RESHETNYAK, On the stability of conformal mappings in multidimensional spaces. Siberian Math. J. 8 (1967), 69-85. 17. Y. G. RESHETNYAK, General theorems on semicontinuity and on convergence with a functional, Siberian Math. J. 8 (1967), 801-8 16. 18. Y. G. RESHETNYAK, Stability theorems for mappings with bounded excursion, Siberiatr Mafh. J. 9 (1968), 499-5 12. 19. S. SAKS. “Theory of the integral,” Hafner, New York. 1937. 20. L. TARTAR. Compensated compactness and partial differential equations. ifr “Nonlinear Analysis and Mechanics: Heriot-Watt Symposium Vol. IV” (R. J. Knops. Ed.). pp. 136-212, Pitman, London, 1979. 21. P. MARCELLINI, Approximation of quasiconvex functions and lower semicontinuity of multiple integrals, to appear.