On certain mixed-type boundary-value problems of elastostatics (with a simple example of Melin's inequality for a system) (Q804337)

In the classical theory of elasticity a typical problem is that of solving the system \[ (1)\quad Au=f\text{ in } \Omega,\quad Bu=\phi \text{ on } \Gamma_ N,\quad u|_{\Gamma}=\psi \text{ on } \Gamma_ D, \] where \(\Omega\) is a bounded domain in \({\mathbb{R}}^ n\), with \(C^{\infty}\)-boundary \(\Gamma =\Gamma_ N\cup \Gamma_ D=\partial \Omega\), and A,B are two differential operators defined on \(\Omega\) and \(\Gamma\), respectively. But sometimes a slightly different problem of the type \[ (2)\quad A_{\lambda}u=f\text{ in } \Omega,\quad B_{\alpha}u=\phi \text{ on } \Gamma_ N,\quad u|_{\Gamma}=\psi \text{ on } \Gamma_ D, \] must be solved, where \(A_{\lambda}=\lambda I+A\), with \(\lambda\geq 0\) and I the identity, \(B_{\alpha}u=\alpha (x)Bu+(1-\alpha (x))u\), with \(\alpha\) (x) a \(C^{\infty}\)-function on \(\Gamma\) such that \(0\leq \alpha (x)<1\) but \(\alpha\) (x)\(\not\equiv 1.\) This new problem is usually treated by applying the Faedo-Galerkin method, but an alternative technique is employed here. It is indeed possible to introduce a pseudodifferential operator \(T_{\alpha}(\lambda)=\alpha (x)T(\lambda)+(1-\alpha (x))I\), defined on \(\Gamma\), such that the originary problem becomes \(T_{\alpha}(\lambda)\psi =\phi\) on \(\Gamma\), with \(\psi \in H^{\sigma -1/2}(\Gamma)\). The solvability of this new problem is based on an inequality for pseudodifferential operators proved by \textit{A. Melin} [e.g.; Ark. Mat. 9, 117-140 (1971; Zbl 0211.171)]. The selection is unique provided that \(\lambda\geq 0\) and \(\sigma\geq 2\).