Convex functions with unbounded gradient (Q817170): Difference between revisions

The author extends to a domain \(\Omega\subset {\mathbb R}^n\) a result previously obtained in the context of affine hypersufaces by \textit{N. S. Trudinger} and \textit{X.-J. Wang} [Invent. Math. 150, No. 1, 45--60 (2002; Zbl 1071.53008)]. He shows that if there exists a convex function \(u:\Omega\subset {\mathbb R}^n\to \mathbb R\), such that \(\nabla u\) becomes unbounded near \(\partial \Omega,\) then \(\Omega\) is convex. The notion of convexity is given for a real function \(u\) defined on a given domain \(\Omega\) as follows: \(u\) is convex if \[ u(\tau x+(1-\tau)y)\leq \tau u(x)+(1-\tau)u(y), \] for all \(x,y\in \Omega,\) and all \(\tau\in (0,1)\) such that \(\tau x+(1-\tau)y\in \Omega.\) The proof makes a brilliant use of basic facts about convexity and standard properties of functions of several variables.