A Regularity Result for a Class of Anisotropic Systems Antonia Passarelli di Napoli and Francesco Siepe () Sommario.R - Si prova la regolarita parziale dei minimi del funzionale I (u) = G(Du), con G integrando convesso a crescita anisotropa. Non si fanno ipotesi speciali sulla struttura di G. Summary.R - We prove partial regularity of minimizers of the functional I (u) = G(Du), where G is a convex integrand satisfying anisotropic growth condition. No special structure assumption is needed on G. 1. Introduction In this paper we study the partial regularity of minimizers of integral functionals of the type I (u) = Z G(Du(x))dx (1:1) u : IRn ! IRN , N 1, where G is a C 2 convex integrand satisfying the growth condition: C j jq G( ) L(1 + j jp) (1:2) (*) Indirizzi degli Autori: Antonia Passarelli di Napoli: Dipartimento di Matematica e Applicazioni \R. Caccioppoli", Universita di Napoli \Federico II", Complesso Monte s.Angelo, Via Cintia - 80126 Napoli, e-mail: [email protected]. Francesco Siepe: Dipartimento di Matematica \U. Dini", Universita di Firenze, Viale Morgagni 67/A - 50134 Firenze, e-mail: [email protected]. Key words and phrases: regularity, minimizers, convexity. This work has been supported by M.U.R.S.T. (40%). 14 A. PASSARELLI DI NAPOLI and F. SIEPE with p > q . Few years ago it was observed that even in the scalar case, i.e. N = 1, minimizers of (1.1) may fail to be regular (see [M2], [G2]), when p is too large with respect to q . On the other hand, one can prove regularity of scalar minimizers of (1.1) if p is not too far away from q (see e.g. [M3], [FS] and the references given in [M3]). More precisely, in [M3] it is shown that if one writes down the Euler equation for the functional I , under suitable assumptions on p and q , the Moser iteration argument still works, thus leading to a sup estimate for the gradient Du of the minimizer. Clearly this approach can not be carried on in the vector valued case, i.e. when N > 1. As far as we know, the only regularity results for systems are proved under special structure assumptions (see [AF2], [M4]). Namely, the model case covered in [AF2] is the functional Z jDujp + k X =1 jDujp with u : IR n ! IRN , N 1, 1 k n, 2 p < p, and p not too far from p, while in [M4], it is proved everywhere regularity of minimizers of (1.1) when G( ) = f (j j). In this paper we prove that if G satises (1.2) and the strong ellipticity assumption hD2G(); i (1 + jj2) q jj2 2 and 2 qn (1:3) 2 q < p < min q + 1; n 1 ; a minimizer u 2 W 1;q ( ; IRN ) of functional (1.1) is C 1; for all < 1 in an open set 0 such that meas( n 0) = 0. We point out that a part from condition (1.3), no special structure assumption is needed on G. The proof of our result goes through a more or less standard blow-up argument aimed to establish a decay estimate on the excess function for the gradient U (x0; r) = Z Br (x0 ) jDu (Du)x ;r j2 + jDu (Du)x ;rjq dx: 0 0 The essential tool in the case we consider, is a lemma due to Fonseca and Maly (see [FM] and also Lemma 2.3 below) which makes possible to connect in the annulus Br n Bs two W 1;q functions v and w with a function z 2 W 1;p(Br n Bs ) if q < p < nqn1 . A REGULARITY RESULT etc. 15 2. Statements and preliminary Lemmas Let us consider the functional I (u) = Z G(Du(x))dx where is a bounded open set of IRn , n 2. Let G : IRnN ! IR, N 2, satisfy the following assumptions: G 2 C2 (H 1) C j jq G( ) L(1 + j jp) (H 2) q hD2G(); i (1 + jj2) jj2 (H 3) 2 2 2 q < p < min q + 1; nqn 1 where It is well known that jDG()j c(1 + jjp 1): (H 4) We say that u 2 W 1;q ( ; IRN ) is a minimizer of I if I (u) I (u + v ) for any v 2 u + W01;q ( ; IRN ). Remark 1. If u is a local minimizer of I and 2 C01 ( ; IRN ) from the minimality condition one has for any " > 0 0 Z [G(Du + "D) G(Du)]dx Z Z1 @G (Du + "tD)D i dt 0 @i Dividing this inequality by ", and letting " go to zero, from (H4) and the assumption p q + 1 we get Z @G (Du)D i dx 0 @ i and therefore by the arbitrariness of the usual Euler-Lagrange =" dx system holds: Z @G (Du)D i dx = 0 @ i 8 2 C01( ; IRN ) 16 A. PASSARELLI DI NAPOLI and F. SIEPE We prove the following Theorem 2.1. Let G be as above and let u 2 W 1;q ( ; IRN ) be a minimizer of I . Then there exists an open subset 0 of such that meas( n 0) = 0 and u 2 C 1;( 0; IRN ) for all < 1 : N 1;q RIn the following, we will denote by u a W ( ; IR ) minimizer of G(Du)dx and assume that G satises (H1), (H2), (H3). We set for every Br (x0 ) U (x0; r) = where Z Br (x0 ) Z Br (x0 ) jDu (Du)x ;r j2 + jDu (Du)x ;rjq dx; 0 0 g = (g )x ;r = meas(B1 (x )) Z Br (x0 ) r 0 0 g: The next Lemma can be found in [FM], (Lemma 2.2), in a slightly dierent form. Lemma 2.1. Let v 2 W 1;q (B1 (0)) and 0 < s < r < 1. There exists a linear operator T : W 1;q (B1 (0)) ! W 1;q (B1 (0)) such that Tv = v on (B1 n Br ) [ Bs and for all > 0 , for all p < q n n 1 jjTvjjW ; (BrnBs) + jjTvjjW ;p(Br nBs) 12 ( 1 h C (r s) sup (t s) jjvjjW ; (BtnBs) + 1 2 t2(s;r) 12 i + sup (r t) jjv jjW ; (Br nBt) + t2(s;r) 1 2 12 h + (r s) sup (t s) q jjv jjW ;q (BtnBs) + t2(s;r) 1 1 + sup (r t) q jjv jjW ;q (Br nBt ) t2(s;r) 1 i) 1 where C = C (n; p; q ) > 0, = (n) > 0 and = (n; p; q ) > 0. A REGULARITY RESULT etc. 17 Let us recall an elementary lemma also proved in [FM]. Lemma 2.2. Let be a continuous nondecreasing function on an interval [a; b], a < b. There exist a0 2 [a; a + 31 (b a)], b0 2 [b 13 (b a); b] such that a a0 < b0 b and (t) (a0) 3 (b) (a) t a0 b a (2.3) (b0) (t) 3 (b) (a) b0 t b a for all t 2 (a0; b0). Finally the next result is a straightforward generalisation to our case of Lemma 2.4 in [FM]. We give the proof here for completeness. Lemma 2.3. Let v; w 2 W 1;q (B1 (0)) and 41 < s < r < 1. Fix q < p < nnq1 , for all > 0 and m 2 IN there exist a function z 2 W 1;q (B1 (0)) and 14 < s < s0 < r0 < r < 1 with r0 ; s0 depending on v; w and , such that (2:4) z = v on Bs0 ; z = w on B1 n Br0 ; r s r0 s0 r s m 3m and jjzjjW ; (Br0 nBs0 ) + jjzjjW ;p(Br0 nBs0 ) hZ ( r s ) C m 1 + jDv j2 + jDwj2 + jv j2 + jwj2+ Br nBs 2 (2.5) + m2 j(vr ws)j2 + Z 1 + jDv jq + jDwjq + jv jq + jwjq + + q Br nBs q i + mq jv wjq (r s) where C = C (n; p; q ) > 0 and = (p; q; n) > 0. Proof. As in Lemma 2.4 in [FM], choose m 2 IN and set 2 f = 1 + jDv j2 + jDwj2 + jv j2 + jwj2 + m2 j(vr ws)j2 + q + q 1 + jDv jq + jDwjq + jv jq + jwjq + mq j(vr ws)jq : 12 1 1 2 18 A. PASSARELLI DI NAPOLI and F. SIEPE We may nd k 2 f1; :::; mg such that Z Bs+ k(r s) nBs+ (k m r s) m 1)( Z 1 fdx m Br nBs fdx ; r s) ; s + k(r s) ], Set, for t 2 [s + (k 1)( m m (t) = Z BtnBs fdx which is a continuous nondecreasing function. By Lemma 2.2, there r s) ; s + k(r s) ] such that exists [s0; r0] [s + (k 1)( m m r s r0 s0 r s m 3m and Z BtnBs0 0 )m Z ( t s fdx 3 r s Bs k r s nBs m 0Z t s 3r s fdx; Br nBs + ( ) + ( k r s) m 1)( fdx Z 0 fdx fdx 3 rr st Br nBs Br0 nBt Z (2.6) (2:7) for all t 2 (s0 ; r0). Set 8 v (x) > > > > > < 0 jxj s0 )w(x) u = > (r jxj)v(xr)+( 0 s0 > > > > : w(x) if x 2 Bs0 if x 2 Br0 n Bs0 if x 2 B1 n Br0 . A direct computation shows that juj2 + jDuj2 + q (jujq + jDujq ) Cf : If we apply Lemma 2.1 to the function u, we then nd z 2 W 1;q (B1 ) satisfying (2.4). Moreover, from (2.6) and (2.7) one readily cheks A REGULARITY RESULT etc. 19 that jjzjjW ; (Br0 nBs0 ) + jjzjjW ;p(Br0 nBs0 ) ( ! Z 0 s0 ) ( r c 0 0 jBr0 n Bs0 j f + Br nBs (r s ) !q ) Z 0 s0 ) ( r jBr0 n Bs0 j q + f 0 0 q Br nBs (r s ) 1 12 1 2 1 2 1 2 1 1 1 ( c (r0 s0) Z Br nBs f ! 12 + (r0 s0 ) Z Br nBs f ! 1q ) ; from which (2.5) follows choosing = minf; g. } 3. Proof of Theorem 1 As usual, to get the partial regularity result stated in Theorem 1, we need a decay estimate for the excess function U (x0; r) dened in section 2. Proposition 3.1. Fix M > 0. There exists a constant CM > 0 such that for every 0 < < 41 , there exists = (; M ) such that, if j(Du)x ;r j M 0 and U (x0 ; r) then U (x0 ; r) CM 2 U (x0; r) : Proof. Fix M and . We shall determine CM later. We argue by contradiction. We assume that there exists a sequence Brh (xh ) satisfying Brh (xh ) ; but Set j(Du)xh;rh j M; lim U (xh ; rh) = 0; h U (xh ; rh) > CM 2 U (xh ; rh) : ah = (u)xh;rh Ah = (Du)xh;rh 2h = U (xh ; rh) : (3:1) 20 A. PASSARELLI DI NAPOLI and F. SIEPE [Blow up.] We rescale the function u in each Brh (xh ) to obtain a sequence of functions on B1 (0). Set v (y ) = 1 [u(x + r y ) a r A y ]; Step 1. h h h rh then Clearly we have h h h Dvh (y ) = 1 [Du(xh + rh y ) Ah ] : h (vh )0;1 = 0 Moreover, h Z B1(0) (Dvh )0;1 = 0 : (1 + hq 2jDvh jq 2 )jDvh j2dy = 1 : (3:2) Passing possibly to a subsequence we may suppose that vh * v weakly in W 1;2(B1 ; IRN ) and, since 8h jAh j M , Ah ! A : Step 2. (3:3) (3:4) Now we show that Z @ 2G (A)D v j D i dy = 0 B (0) @i @j 1 8 2 C01(B1; IRN ) : (3:5) Since we assume p 1 q we can write the usual Euler-Lagrange system for u (see Remark 1). Then, rescaling in each Brh (xh ), we get for any 2 C01 (B1; IRN ) and any 1 i N Z @G (A + Dv )D i dy = 0 : h h h B (0) @i 1 Then Z @G (A + Dv ) @G (A )]D i dy = 0 : [ h B (0) @i h h h @i h 1 1 (3:6) A REGULARITY RESULT etc. 21 Let us split B1 = Eh+ [ Eh = fy 2 B1 : h jDvh (y )j > 1g [ fy 2 B1 : h jDvh (y )j 1g ; then by (3.2) we get jEh+j Z Eh+ 2h jDvhj2dy 2h Z B1 (0) jDvhj2dy c2h: (3:7) Now, by (H4) and Holder inequality, we observe that 1 j Z [DG(A + Dv ) DG(A )]Ddy j h h Eh c jEh+j + cph + h ch + c Z 2 h Z Eh+ h jDvhjp 1dy q 2jDvh jq dy E+ h h h !p q 1 p q q 2 h 2 jEh+j q p+1 q where we used the assumption p 1 q . From this it follows that Z 1 [DG(Ah + h Dvh ) DG(Ah )]Ddy = 0: lim h h Eh+ ch (3:8) On Eh we have 1 Z [DG(Ah + h Dvh ) DG(Ah )]Ddy Eh Z Z1 D2G(Ah + shDvh )Dvh Ddsdy = Eh 0 Z Z1 [D2G(Ah + sh Dvh ) D2G(Ah )]Dvh Ddsdy + = Eh 0 Z + D2G(Ah )Dvh Ddy : Eh that (3.7) ensures that Eh ! B1 in Lr (B1 ) for all r < h Note and by (3.2) we have, passing possibly to a subsequence, h Dvh (y ) ! 0 a:e: in B1 : 1 22 A. PASSARELLI DI NAPOLI and F. SIEPE Then, by (3.3), (3.4) and the uniform continuity of D2 G on bounded sets, we get Z lim 1 [DG(A + Dv ) DG(A )]Ddy Rh h Eh h h h h = B D2 G(A)DvDdy : By (3.6), (3.8) and the above equality, we obtain that v satises equation (3.5), which is elliptic by (H3). We have for any 0 < < 1 1 Z Z B jDv (Dv) j2dy c 2 Moreover we have and Step 3. q B1 jDv (Dv)1j2dy c 2: v 2 C 1 (B1 ; IRN ) : (3:9) (3:10) 2 1;q (B ; IRN ) h q (vh v ) * 0 weakly in Wloc 1 [Upper bound.] We set Gh ( ) = 12 [G(Ah + h ) G(Ah) h DG(Ah) ] and for every r < 1 h Ih;r (w) = Z Br Gh (Dw)dy : Note that by the strong ellipticity assumption (H3) it follows that Gh ( ) 0, for any . Fix 14 < s < 1. Passing to a subsequence we may always assume that lim [I (v ) Ih;s (v )] h h;s h exists. We shall prove that lim [I (v ) Ih;s (v )] 0 : h h;s h (3:11) Consider r > s and x m 2 IN . Observe thatp , since v 2 W 1;q (B1 ) and vh 2 W 1;q (B1 ), Lemma 2.3, with = h p , implies that there exist zh 2 W 1;q (B1) and 14 < s < sh < rh < r < 1 such that zh = v on Bsh zh = vh on B1 n Brh 2 A REGULARITY RESULT etc. and p 23 2 jjzhjjW ; (Brh nBsZh ) + h p jjzhjjW ;p(Brh nBsh ) h C (r ms) (1 + jDv j2 + jDvh j2 + jv j2 + jvh j2 + Br nBs 2 + m2 jv vh j2 )+ (3.12) (r s) p qZ + h p (1 + jDv jq + jDvh jq + jv jq + jvh jq + Br nBs q i + mq j(vr vsh)jq ) Since by (3.10), Dv is locally bounded on B1 we get Ih;s (vh ) Ih;s (v ) Ih;rh (vh) Ih;rh (v) + IZh;rh (v) Ih;s(v) = Ih;rh (vh ) Ih;rh (v ) + Gh (Dv ) (3.13) Brh nBs Ih;r ( z ) I ( v ) + c ( r s ) h h;r h Zh c [Gh (Dzh ) Gh (Dv )] + c(r s) : 12 1 2 1 2 Brh nBsh where we used the minimality of vh . As jGh ( )j c(j j2 + ph 2 j jp) (see [AF], Lemma II.3), we get by (3.12) Ih;rh Z(zh ) Ih;rh (v ) c jDzhj2 + hp 2jDzhjp Brh nBsh Z 2 h ( (1 + jDv j2 + jDv j2 + jv j2 + jv j2+ C r s) m2 h Br nBs h 2 ip + m2 j(vr vsh)j2 ) + (1 + jDv jq + jDvh jq + jv jq + jvh jq + 2 2 h p 2 qZ Br nBs + C (r m2s) h p q ip + mq j(vr vsh)jq ) 2 = Jh;1 + Jh;2 : Since vh ! v in L2 (B1; IRN ) we have, using (3.2) lim sup Jh;1 Cm 2 : h!1 24 A. PASSARELLI DI NAPOLI and F. SIEPE Moreover, since q(p 2) Z p h and 2( B1 q(p 2) Z p h jDvhjq = h B1 Z p q) p q 2 2( B1 jDvhjq Ch B1 jDvhjq ch h q(p 2) Z p jvh vjq ch we have 2( p q) p p q) p lim J = 0: h h;2 Hence we conclude letting rst m ! 1 and then r ! s in (3.13). 1 < r < 1, Step 4. [Lower bound.] We shall prove that, for a.e. 4 2 if t < r then lim sup h Z Bt jDv Dvhj2(1 + hq 2jDv Dvhjq 2) lim [I (v ) Ih;r (v )] : h h;r h For any Borel set A B1 , let us dene Z h (A) = (jvh j2 + jDvhj2)dx : A Passing possibly to a subsequence, since h (B1 ) c, we may suppose h * weakly in the sense of measures; where is a Borel measure over B1 . Then for a.e. r < 1 (@Br ) = 0 and let us choose such a radius r. Consider 41 < t < s < r, also such that (@Bs ) = 0, and x m 2 IN . Observe that , as vh 2 W 1;q (B1 ) Lemmas 2.3 implies that there exist zh 2 W 1;q (B1 ) and 14 < s < sh < rh < r < 1 such that zh = vh on Bsh zh = vh on B1 n Brh rh sh r3ms A REGULARITY RESULT etc. and p 25 2 jjzhjjW ; (Brh nBsh ) + h p jjzhjjW ;p(Brh nBsh ) hZ C (r ms) (1 + jDvh j2 + jvh j2)+ B n B r s p qZ i p (1 + jDvh jq + jvh jq ) + h 1 12 ( 2) (3.14) 1 2 Br nBs Passing possibly to a subsequence, we may suppose that weakly in W 1;2(B1 ) : zh * vr;s and vr;s = v in (B1 n Br ) [ Bs Moreover from (3.14) it is clear that Z q 2 h jDzhjq B1 c (3:15) Consider h 2 C01 (Brh ) such that 0 h 1, h = 1 on Bsh and jDhj rhCsh and set h = h (zh ); vr;s = ?vr;s , and is the usual sequence of molliers. Now, where vr;s setting v = ? v , we observe that Ih;rh (vh ) Ih;rh (v ) + )+ = Ih;rh (vh ) Ih;rh (zh ) + Ih;rh (zh ) Ih;rh (vr;s h (3.16) + Ih;rh ( h + vr;s ) Ih;rh (vr;s ) Ih;rh ( h )+ Ih;rh (vr;s) Ih;rh (v ) + Ih;rh ( h ) = Rh;1 + Rh;2 + Rh;3 + Rh;4 + Rh;5 To bound Rh;1 we observe that Ih;rh (vh) Ih;rh (zh ) = Z Brh nBsh Z Gh (Dvh) Brh nBsh Gh(Dzh) Z Brh nBsh Gh (Dzh ) + 26 A. PASSARELLI DI NAPOLI and F. SIEPE on the other hand we have Z Brh nBsh Gh (Dzh ) Z jDzhj2 + hp 2jDzhjp Brh nBsh hZ cm 2 1 + jDvh j2 + jvh j2 + Br nBs p 2 qZ ip 1 + jDvh jq + jvh jq 2 + h p Br nBs and then arguing as we did in Step 3 to bound Jh;1 we get lim sup Z Brh nBsh h Gh(Dzh) Cm 2 hence, letting h ! 1 we get limhinf Rh;1 Cm 2 (3:17) We obtain that Rh;2 = Z BZrh nBsh c ) Gh(Dzh) Gh (D h + Dvr;s j2 + p 2jD + Dv jp jD h + Dvr;s h r;s h ZBrh nBsh j2 + jDzhj2 + ph 2jDzhjp + jDvr;s Z jz v j2 c m2 (hr sr;s)2 + Br nBs jp ! h h j z v h + mp ph 2 (r sr;s)p = Sh;1 Sh;2 c Brh nBsh jp + hp 2 jDvr;s (3.18) where we used the bound rh sh r3ms . By (3.15), since p < q , we get Z p 2jzh jp B1 h cph 2 cph 2 Z B1 (Z jzh (zh)0;1 B1 jzh jp + j(z (zh )0;1jq jp h)0;1 qp Z + B1 jzhj p ) A REGULARITY RESULT etc. chp 2 2( ch (Z B1 jDzh qp q j Z + pq Z p q) q q 2 jDzhjq R 1 h B1 B1 27 jzh j2 p2 ) + cph 2: where we used (3.14) to bound B jzh j2 .Therefore 2 1 2 lim sup Sh;2 c (r m s)2 Z h!1 B1 j2 : jvr;s vr;s 2 To bound Sh;1 , observe that for every h Z Brh nBsh Z c j2 jDvr;s Br nBs limjinf c = c limj inf jDvr;s j2 + c Z Z Br nBs jDzj Z j2 jDvr;s Dvr;s B1 2 j2 + c Z B1 j2 jDvr;s Dvr;s 2 jDvj j2 + (Br nBs)n(Brj nBsj ) Z Z jDzj j2 + c + c lim sup B1 Brj nBsj j j2 jDvr;s Dvr;s 2 We control the second integral as usual using Lemma 2.3, while the rst is less or equal than c(Br n Bs ). Moreover we can estimate Z Brh nBsh jDzhj2 + ph 2jDzhjp as we did in Step 3 to bound Jh;1 . Hence limhinf Rh;2 cm 2 c(Br n Bs)+ Z j2 + c jDvr;s Dvr;s B1 cm2 Z jv v j2 (r s)2 B r;s r;s 2 1 2 (3.19) 28 A. PASSARELLI DI NAPOLI and F. SIEPE To bound Rh;3 we observe that Gh (A + B) Gh (A) Gh (B) = and Z 1Z 1 0 0 D2Gh (sA + tB )ABdsdt + tD ) = D2 G(A + s Dv + t D ) D2 Gh (sDvr;s h h r;s h h h is bounded and converges to D2 G(A) a.e.. Since Z Rh;3 = Br h dx Z [0;1][0;1] + th D and we may suppose that Z B1 jD j2 + D2G(Ah + sh Dvr;s h )Dvr;s D h dsdt 1;2 h * weakly in W (B1), where 2 Z j2+ c (r m s)2 jvr;s vr;s B1 Z 2 (3.20) j2 + c jDvr;s Dvr;s B1 2 we get easily jj lim sup jRh;3j c(M )jjDvr;s L (B ) jjD jjL (B ) : 2 h To bound Rh;4 we observe that nRh;4 = Then Z Brh nBsh Z 1 2 2 1 2 ) G (Dv )] [Gh (Dvr;s h Gh (Dv ) Brh nBs cjBr n Bs j : limhinf Rh;4 cjBr n Bs j : Moreover (H3) implies jRh;5j = ZIh;rh ( h ) = Gh (D h ) BZrh (3:21) (1 + hq 2 jDv Bt Dvh jq 2)jDv (3:22) Dvh j2 (3.23) A REGULARITY RESULT etc. 29 for small enough. Passing to a subsequence we may suppose that lim sup Rh;5 = lim R : h h;5 h Therefore returning to the (3.16), from (3.17), (3.19), (3.21), (3.22) and (3.23) we get limhinf [Ih;r (vh ) Ih;r (v )] Z lim sup (1 + hq 2jDv Dvhjq 2)jDv Dvh j2 + Bs h jj cjBr n Bs j c(Br n Bs ) cjjDvr;s L (B ) jjD jjL (B ) + Z 2 Z m 2 2 j2 : cm jDvr;s Dvr;sj c (r s)2 jvr;s vr;s B B 2 1 2 2 1 2 1 2 1 2 Passing to the limit as ! 0+ we get easily limhinf [Ih;r (vh ) Ih;r (v )] Z lim sup (1 + hq 2jDv Dvh jq 2)jDv Dvhj2 + Bs h cjBr n Bs j c(Br n Bs ) cm 2 then passing to the limit as m ! 1 and s ! r we get lim sup Z h Br jDv Dvhj2(1+q 2jDv Dvhjq ) lim [I (v ) Ih;r (v )]: h h;r h [Conclusion.] From the two previous steps we conclude that, for any B , with 0 < < 41 Step 5. lim h Z B jDv Dvhj2(1 + q 2jDv Dvhjq ) = 0 : Now, from this equality and by (3.9) we get lim U (xh ; rh) h 2h Z 1 = lim h 2 h Brh (xh ) (jDu (Du)rh j2 + jDu (Du)rh jq )dx 30 A. PASSARELLI DI NAPOLI and F. SIEPE = lim h = Z Z B (jDu (Du) j2 + hq 2 jDu (Du) jq )dy (jDv (Dv ) j2)dy B CM 2 which contradicts (3.1) if we choose CM = 2CM . } The proof of Theorem 1 follows by proposition 3.1 by a standard iteration argument, see [G1]. Remark 2. Notice that the proof of Proposition 3.1 and of Theorem nq 1 still works if, beside assuming p < n 1 , we have p q + 1. References [AF1] Acerbi E. and Fusco N., A regularity theorem for minimizers of quasiconvex integrals , Arch. Rational Mech. Anal. 99 (1987), 261{ 281. [AF2] Acerbi E. and Fusco N., Partial regularity under anisotropic ( ) growth conditions , J. Di. Equat. 107 (1994), 46{67. [D] Dacorogna B., Direct methods in the calculus of variations, Appl. Math. Sci. 78, Springer Verlag 1989. [Ek] Ekeland I., Nonconvex minimisation problems, Bull. Amer. Math. Soc. 1 (1979), 443{474. [Ev] Evans L. C. , Quasiconvexity and partial regularity in the calculus of variations, Arch. Rational Mech. Anal. 95 (1986), 227{252. [EG] Evans L. C. and Gariepy R. F., Blow-up, compactness and partial regularity in the calculus of variations, Rend. Circolo Mat. Palermo, suppl. no. 15 (1987), 101{108. [FM] Fonseca I. and Maly J., Relaxation of multiple integrals in Sobolev spaces below the growth exponent for the energy density, to appear. [FH] Fusco N. and Hutchinson J., A direct proof for lower semicontinuity of polyconvex function, preprint 1994. [FS] Fusco N. and Sbordone C., Higher integrability of the gradient of minimizers of functionals with nonstandard growth condition, Communications on Pure and Appl. Math. XLIII (1990), 673{683. p; q A REGULARITY RESULT etc. 31 [G1] Giaquinta M., Multiple integrals in the calculus of variations and nonlinear elliptic systems, Annals of Math.Studies 105 (1983), Princeton Univ. Press. [G2] Giaquinta M., Growth conditions and regularity, a counterexample, Manu. Math. 59 (1987), 245{248. [GM] Giaquinta M and Modica G., Partial regularity of minimizers of quasiconvex integrals, Ann. Inst. H. Poincare, Analyse non lineaire 3 (1986), 185{208. [Gi] Giusti E., Metodi diretti in calcolo delle variazioni, U.M.I. (1994). [M1] Marcellini P., Approximation of quasiconvex functions and lower semicontinuity of multiple integrals, Manuscripta Math. 51 (1986), 1{28. [M2] Marcellini P., Un example de solution discontinue d' un probeme variationel dans le cas scalaire, Preprint Ist. U. Dini, Firenze, 1987{ 88. [M3] Marcellini P., Regularity and existence of solutions of elliptic equations with ( ) growth conditions , J. Di. Equat. 90 (1991), 1{30. [M4] Marcellini P., Everywhere regularity for a class of elliptic systems without growth conditions, Preprint Ist. U. Dini, Firenze 1993. [Mo] Morrey C. B., Multiple integrals in the calculus of variations, Springer, New York (1966). p; q Pervenuto in Redazione il 3 Febbraio 1996.
© Copyright 2024 ExpyDoc