The degenerate special Lagrangian equation (Q520375)

Let \(D\subset \mathbb{R}^n\) be a bounded domain with smooth boundary \(\partial D\), and let \(\mathcal{D}=(0,1)\times D\subset \mathbb{R}^{n+1}\). The coordinates in \(\mathcal{D}\) are denoted by \((t,x)\). Let \(I_n\) be the \((n+1)\times (n+1)\) diagonal matrix with diagonal entries \((0,1,\dots,1)\). A function \(u\in C^2(\mathcal{D})\) satisfies the degenerate special Lagrangian (DSL) equation of phase \(\theta\in (-\pi,\pi]\) if \[ \text{Im} \left( e^{-\sqrt{-1} \theta} \det(I_n+\sqrt{-1} \nabla^2 u ) \right)=0, \qquad \text{Re} \left( e^{-\sqrt{-1} \theta} \det(I+\sqrt{-1} \nabla^2_x u ) \right)>0. \] Geometrically the DSL equation governs geodesics in the space of positive Lagrangians. Such geodesics play a crucial role in the work of the second author [Math. Ann. 357, No. 4, 1389--1424 (2013; Zbl 1282.53067); Geom. Funct. Anal. 24, No. 2, 670--689 (2014; Zbl 1296.53157)] on the existence and uniqueness of special Lagrangians in Calabi-Yau manifolds, the geometry of the space of positive Lagrangians, and stability conditions for Lagrangian submanifolds in the context of mirror symmetry. Another geometric motivation for this work is the observation that whenever a pair of positive Lagrangians is connected by a sufficiently regular geodesic, the number of intersection points is bounded below by the number of critical points of a function on one of them. This can be viewed as a first step in a new approach to Arnold's conjecture on fixed points of Hamiltonian symplectomorphisms. Previously the second author and \textit{A. M. Yuval} [``Geodesics of positive Lagrangians in Milnor fibers'', Int. Math. Res. Not. 2017 (3), 830--868 (2017)] constructed such geodesics under an \(O(n)\) symmetry assumption to reduce the problem to a Hamiltonian flow ODE. Here the authors construct geodesics of positive Lagrangians in \(\mathbb{C}^n\) in the absence of symmetry assumptions using the theory of fully nonlinear degenerate elliptic PDE. The authors' main PDE results are the existence and uniqueness of weak solutions of the Dirichlet problem for \((*)\) in the sense of \textit{F. R. Harvey} and \textit{H. B. Lawson jun.}'s Dirichlet duality theory [Commun. Pure Appl. Math. 62, No. 3, 396--443 (2009; Zbl 1173.35062)]. Namely, one needs to construct a proper closed subset \(F\) of the set of symmetric matrices \(\text{Sym}^2(\mathbb{R}^m)\) satisfying \(F+\mathcal{P}\subset F\), where \(\mathcal{P}\subset \text{Sym}^2(\mathbb{R}^m)\) is the set of nonnegative matrices. Such a set \(F\) is called a subequation for a PDE of the form \(f(\nabla^2 u(x))=0\) for \(u\in C^2(U)\), \(U\subset \mathbb{R}^m\), if the solutions \(u\) of class \(C^2(U)\) satisfy \(\nabla^2 u(x)\in \partial F\) for each \(x\in U\). Further, \(u\in C^2(U)\) is a subsolution if \(\nabla^2 u(x)\in F\) for each \(x\in U\). Harvey and Lawson have proved the existence and uniqueness of continuous solutions to the \(F\)-Dirichlet problem under suitable assumptions on \(\partial U\). Finally, if a solution in the sense of Harvey and Lawson is in \(C^2(U)\), then it is a solution in the classical sense. The authors construct a subequation for the DSL equation by associating to each \(u\in C^2(\mathcal{D})\) the \(S^1\) valued function \( \Theta_u(t,x)= \text{arg} \det (I_n+\sqrt{-1} \nabla^2 u(t,x))\), appropriately lifting this to \(\mathbb{R}\) to obtain another function \(\widehat\Theta\), and then defining \(\widetilde\Theta\) to be the minimal upper semi-continuous extension of \(\widehat\Theta\) to \(\text{Sym}^2(\mathbb{R}^{n+1})\). The result is that for each \(c\in (-(n+1)\pi/2, (n+1)\pi/2)\) such that \(c\equiv \theta \) mod \(2\pi\), the set \(\mathcal{F}_c=\{A\in \text{Sym}^2(\mathbb{R}^{n+1}): \widetilde \Theta(A) \geq c\}\) is a subequation of the DSL equation of phase \(\theta\). The authors then modify the Harvey-Lawson theory to deal with domains with corners and which do not satisfy a certain convexity assumption required in that theory (the domain \(\mathcal{D}\) does not satisfy these conditions). The main theorem is the existence and uniqueness of solutions for all branches of the \(\mathcal{F}_c\)-Dirichlet problem, and hence, for the endpoint problem for geodesics. Further results concerning the existence of a calibration measure associated to solutions of the DSL equation in the outermost branches, and a partial Lipschitz estimate for the solution \(u\), are also proved. These results are used to show that the length of the geodesic of graph Lagrangians corresponding to \(u\) is well defined. Furthermore, by integrating the calibration measure along paths of Lagrangians, the authors define the calibration functional, which plays a similar role in the geometry of positive Lagrangians as the Monge-Ampère measure plays in convex and complex geometry.