A Bernstein problem for special Lagrangian equations (Q1401445): Difference between revisions

The classical Bernstein result says that the only minimal surfaces in \({\mathbb R}^3\) which are the graphs of a function on \({\mathbb R}^2\) are the planes. This article derives a similar result for special Lagrangian submanifolds of \({\mathbb R}^{2n}\) which are graphs over \({\mathbb R}^n\). By results of \textit{R. Harvey} and \textit{H. B. Lawson} [Acta Math. 148, 47-157 (1982; Zbl 0584.53021)] such a submanifold can always be written as the graph of \(\nabla u:{\mathbb R}^n\to{\mathbb R}^n\) for a \(C^2\) function \(u:{\mathbb R}^n\to{\mathbb R}\) which must satisfy the elliptic pde \[ \text{ Im}(\det(I+iD^2u))=0 \] where \(D^2u\) denotes the Hessian of \(u\). This equation can also be represented in the form \[ \text{ arg}(\det(I+iD^2u)) = \arctan(\lambda_1)+\ldots+\arctan(\lambda_n)= \text{ const.} \] where the \(\lambda_j\)'s are the eigenvalues of \(D^2u\): this is the form given in the paper under review. The principal result of the paper is that if \(M=(x, \nabla u)\) is a minimal surface in \({\mathbb R}^n\times{\mathbb R}^n\) (which is equivalent to \(M\) being special Lagrangian) and \(u\) is a smooth convex function then \(M\) is a plane. The paper actually proves a slightly stronger result, where the condition of convexity is weakened to \(D^2u\geq \epsilon(n)I\) for a small dimensional constant \(\epsilon(n)\).