Looking for the best constant in a Sobolev inequality: a numerical approach (Q607691)

Let \(\Omega\) be a convex bounded domain with a smooth boundary in \(\mathbb{R}^d\), \(d\in\{1,2,3\}\), let \(H^2(\Omega)\) be the Sobolev space and \(V\equiv H^2(\Omega)\cap H^1_0(\Omega)\) so that \(V\) is a closed subspace of \(H^2(\Omega)\). Let us assume that \(\phi: V\to C^0(\overline\Omega)\) is an injection from \(V\) into \(C^0(\overline\Omega)\). Then \(\phi\) is continuous and therefore exists a constant \(C\) such that \(\|\phi\|_\infty\leq C\|\phi\|_V\), where \(\|\phi\|_\infty\equiv \max_{x\in \overline\Omega}|\phi(x)|\). The smallest constant \(\gamma\) satisfying the inequality above is the norm of \(\phi\). The computation of \(\gamma\) is equivalent to solving a non-smooth eigenvalue problem. The authors propose two methods for computing \(\gamma\). The first one uses an augmented Lagrangian method together with finite element approximations, the second one is based on the usage of the Green's function. Numerical experiments presented in the concluding part of the paper show good convergence properties of the augmented Lagrangian based method.