Burkholder integrals, Morrey's problem and quasiconformal mappings (Q2879891)

A long-standing open problem in the calculus of variations is whether every rank-one convex function \(E:\mathbb R^{2\times 2}\to\mathbb R\) is quasiconvex. Rank-one convexity means that \(t\mapsto E(A+tX)\) is convex for every \(A\in \mathbb R^{2\times 2}\) and every rank \(1\) matrix \(X\). Quasiconvexity requires that for every smooth function \(f\) with compact support in a bounded domain \(\Omega\subset \mathbb R^2\) the following holds: \(\int_{\Omega}E(A+Df) \geq \int_{\Omega}E(A )\). An important example of a rank-one convex function whose quasiconvexity remains unknown is \(-B_p\), where NEWLINE\[NEWLINEB_p(A)=((p/2)\det A+(1-p/2)|A^2|)|A|^{p-2},\quad p\geq 2.NEWLINE\]NEWLINE Here \(|A|\) is the operator norm.NEWLINENEWLINEThe main result, Theorem 1.2, shows that \(\int_{\Omega}B_p(I+Df) \leq \int_{\Omega}B_p( I )\) holds under the additional assumption that \(B_p(I+Df)\geq 0\) pointwise. Thus, one can say that \(-B_p\) is quasiconvex in a neighborhood of the identity matrix \(I\). Among many interesting consequences of this result is the following sharp inequality (Corollary 1.7):NEWLINENEWLINE If \(f\) is a self-homeomorphism of \(\Omega\) such that \(f(z)-z\in W_0^{1,2}(\Omega)\), then NEWLINE\[NEWLINE\int_{\Omega}(1+\log|Df|^2)\,\det Df \leq \int_\Omega |Df|^2.NEWLINE\]NEWLINE A remarkable feature of this and other inequalities obtained in the paper is that they are not only sharp, but admit a very large family of extremal maps -- so-called piecewise radial maps -- sketched in Figure 1 in the paper.