Elliptic complexes in the calculus of variations

From MaRDI portal





This paper summarizes the author's study (with Fiorenza, Iwaniec, Onninen and Verde) on elliptic couples. Let \[ {\mathcal D}'(\mathbb{R}^N, U) @>{\mathcal P}>> {\mathcal D}'(\mathbb{R}^N,V) @>{\mathcal Q}>> {\mathcal D}'(\mathbb{R}^N,W), \] be an elliptic complex. An elliptic couple associated to this complex is \({\mathcal F}= [{\mathcal P}\alpha, {\mathcal Q}^* \beta]\). The norm and the Jacobian of this couple are defined by \[ |{\mathcal F}|= \bigl(|{\mathcal P}\alpha |^2+|{\mathcal Q}^* \beta |^2 \bigr)^{1/2},\;J(x,{\mathcal F})= \langle {\mathcal P}\alpha, {\mathcal Q}^* \beta\rangle. \] \({\mathcal F}\) is called \(K\)-quasiharmonic with \(1\leq K= K(x)< \infty\) if \[ \bigl|{\mathcal F}(x)\bigr |^2\leq{\mathcal K}(x)J(x,{\mathcal F}), \;{\mathcal K}(x)= K(x)+K^{-1}(x)\geq 2. \] After preparing on Orlicz space, Schwartz distribution, maximal operator and Hardy space in Chap. 2, it is shown that quasi harmonic fields are closed under weak convergence in Chap. 3 (Th. 3.12). Then variational integrals \[ I[{\mathcal F}]= \int_{\mathbb{R}^N} f(X,Y), \quad{ \mathcal F}= [X,Y]\in L^p(\mathbb{R}^N, V\times V), \] are considered. Basic definitions on \(f\) are quasiconvex; \[ \int_{\mathbb{R}^N}\bigl|f(A+{\mathcal P}\alpha, B+{\mathcal Q}^* \beta)- f(A,B)\bigr |dx\geq 0, \] for any constant \(A,B\) and Lipschitz functions with compact support \(\alpha,\beta\), convex in singular directions; \(t\to f(A+tX, B+tY)\) is a convex function if \(X\) and \(Y\) is orthogonal in \(V\), and polyconvex; \(f(X,Y)= g(X,Y,\langle X,Y\rangle)\) where \(g\) is convex. Polyconvexity implies quasiconvexity. Quasiconvexity does not need to imply convexity in singular direction. But if the elliptic complex satisfies \(\cup_{ |\xi |=1}\ker {\mathcal Q}(\xi)=V\), then a quasiconvex function is convex in singular direction [Th. 3.18. \textit{F. Giannetti} and \textit{A. Verde}, Stud. Math. 140, 79-98 (2000; Zbl 0968.58019)]. In Chap. 4, integrability of the Jacobian determinant of some Sobolev meanings is treated, and it is shown that \(f\in W^{1,N-1} (\Omega,\mathbb{R}^n)\), \(N>2\), is an orientation preserving (reserving) mapping whose cofactor matrix \(|D^\sharp f|^{n/(N-1)}\) is in Orlicz space \(L^P(\Omega)\), where \[ \int^\infty_1 {P(s)ds\over s^2}= \infty,\quad \left[{P(t) \over t}\right]'\leq 0\leq\left [{P(t)\over t^s} \right]',\quad s={N^2-2N+ 1 \over N^2-N-1}, \] then the Jacobian determinant of \(f\) is locally integrable and obeys the rule of integration by parts (Th. 4.6). As for further results, the author referes to \textit{F. Giannetti}, \textit{T. Iwaniec}, \textit{J. Onnien} and \textit{A. Verde} [J. Geom. Anal. 12, 223-254 (2002; Zbl 1053.42024)]. Chap. 5 deals with regularity results for vector fields of bounded mainly taken from \textit{A. Fiorenza} and \textit{F. Giannetti}, JIPAM, Inequal. Pure Appl. Math. 1, No. 2, Paper No. 14 (2000; Zbl 0962.35057). They gave a family of reverse Hölder inequalities (Th. 5.6, Prop. 5.8) and applied to the theory of quasiconformal mappings and PDEs. In particular, Bojarski's regularity result: \(f=(f_1,f_2): \Omega\subset \mathbb{R}^2\to \mathbb{R}^2\) with bounded distorsion holds \[ f\in W^{1,2- \varepsilon}(\Omega, \mathbb{R}^2)\Rightarrow f\in W^{1,2+ \varepsilon} (\Omega,\mathbb{R}^2); \] and is derived from Prop. 5.8. Chap. 6, the last Chapter, discusses the lower semicontinuity of the functional \[ F(u)= \int_\Omega f(x,u,{\mathcal L}u) dx,\;0\leq f(x,s,\xi)\leq c(1+|\xi|^q), \] where \(u\in W^{1,p}(\Omega, \mathbb{R}^d)\), \(q\geq p>1\) and \({\mathcal L}\) is a linear differential operator of first-order. The main result asserts if \[ \int_Q \biggl [f\bigl({\mathcal L}(Ax+ u(x))\bigr) -f\bigl({\mathcal L}(Ax) \bigr)\biggr] dx\geq 0, \] for any \(A\) and \(u\in C_0^\infty (Q,\mathbb{R}^d)\), where \(Q\) is the unit cube, then \[ \int_\Omega f\bigl({\mathcal L}u(x)\bigr) dx\leq\liminf_{n\to\infty} \int_\Omega f\bigl( {\mathcal L}u_n(x) \bigr) dx, \] for any sequence \(u_n\in W^{1,q} (\Omega, \mathbb{R}^d)\) converges weakly to \(u\) in \(W^{1,p}(\Omega,\mathbb{R}^d)\), provided \(q\geq p >\max \{1,q(N-1)/N\}\) (Th. 6.1. \textit{F. Giannetti} and \textit{A. Verde} [Ann. Fac. Sci. Toulouse Math. 10, 299-311 (2001; Zbl 1017.49016)]). The author notes that it is a BV setting extension of a result of \textit{P. Marcellini} and \textit{M. Gori} [J. Convex Anal. 9, 475-502 (2002; Zbl 1019.49021)], by showing that no growth assumptions on \(f\) are needed as long as we assume to be nonnegative and locally Lipschitz.











This page was built for publication: Elliptic complexes in the calculus of variations

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4425003)