On the regularity of the polar factorization for time dependent maps (Q2575178): Difference between revisions

Let \(\Omega\) be a bounded open set in \({\mathbb R}^d\) with \(| \partial\Omega| =0\) where \(| \cdot| \) denotes the \(d\)-dimensional Lebesgue measure. \textit{Y.~Brenier} [Comm. Pure Appl. Math. 44, 375--417 (1991; Zbl 0738.46011)] has proved that for every Lebesgue measurable mapping \(X\in L^2(\Omega,{\mathbb R}^d)\) satisfying the nondegeneracy condition \[ X^{-1}(B)=0 \quad\text{for all measurable \(B\subset{\mathbb R}^d\) with \(| B| =0\)}, \leqno(*) \] there is a unique factorization \(X=\nabla\Phi\circ g\) where \(\Phi\) is a convex function and \(g\) belongs to \(G(\Omega)\), the set of measure preserving mappings of \(\Omega\). Recall that \(g\in G(\Omega)\) if and only if \[ \int_{\Omega} f(g(x))\,dx = \int_{\Omega} f(x)\,dx \] for all \(f\in C_b(\Omega)\), the set of bounded continuous functions on \(\Omega\). Let \(\rho=X\# da\) denote the pushforward by \(X\) of the Lebesgue measure \(da\) on \(\Omega\). Then \((*)\) is equivalent to the absolute continuity of \(\rho\) with respect to Lebesgue measure, and \(\Phi\) satisfies the Monge-Ampère equation \[ \rho(\nabla\Phi)\det D^2\Phi(x) = 1 \] in the weak sense \[ \int_{\Omega} h(\nabla\Phi(y))\,dy = \int_{{\mathbb R}^d} h(x)\, d\rho(x) \quad\text{for all \(h\in C_b({\mathbb R}^d)\).} \] Furthermore, the Legendre transform \(\Psi\) of \(\Phi\), defined by \[ \Psi(y) = \sup_{x\in\Omega} \{ x.y-\Phi(x) \}, \] satisfies the Monge-Ampère equation \[ \det D^2\Psi(x) = \rho(x) \] in the weak sense that \[ \int_{{\mathbb R}^d} h(\nabla\Psi(x))\,d\rho(x) = \int_{\Omega} h(y)\,dy \quad\text{for all \( h\in C_b(\Omega)\). } \] Now suppose we have a family of mappings \(\{X(t,\cdot)\}_{t\in I}\) from \(\Omega\) into \({\mathbb R}^d\) (with \(\Omega\) now a bounded convex open set in \({\mathbb R}^d\)) for some open interval \(I\subset{\mathbb R}\), with each \(X(t,\cdot)\) satisfying \((*)\). Then for each \(t\in I\) there is a unique factorization \(X(t,\cdot)=\nabla\Phi(t,\cdot)\circ g(t,\cdot)\) as above. The author investigates the regularity of the curve \(t\mapsto(g(t,\cdot),\Phi(t,\cdot),\Psi(t,\cdot))\). Several results are proved. The first states that if \(\rho=X\# da\), \(X\) and \(\partial_t X\) are bounded in space and time, then \(\partial_t \nabla\Phi\) and \(\partial_t g\) are bounded as measures. A second result is that if \(\rho\) is close to \(1\) (or to a continuous positive function) in \(L^\infty\), then \(\partial_t\Phi\) belongs to \(C^\alpha\) for some \(\alpha>0\). Key tools used in the proofs are a local maximum principle for solutions of degenerate elliptic equations proved by \textit{M. K. V.~Murthy} and \textit{G.~Stampacchia} [Ann. Mat. Pura Appl. 80, 1--122 (1968; Zbl 0185.19201)] and \textit{N. S.~Trudinger} [Ann. Scuola Norm. Sup. Pisa 27, 265--308 (1973; Zbl 0279.35025)], and a Harnack inequality of \textit{L. A.~Caffarelli} and \textit{C. E.~Gutiérrez} [Am. J. Math. 119, 423--465 (1997; Zbl 0878.35039)] for solutions of the homogeneous linearized Monge-Ampère equation. The results have an immediate application to the semigeostrophic equations, a system used in meteorology to model the formation of fronts. Namely, they enable the velocity to be defined in physical space, a fact that was not known previously for weak solutions.