On a Babuška paradox for polyharmonic operators: spectral stability and boundary homogenization for intermediate problems (Q2284357)

In this interesting paper, the authors investigate the spectral convergence of high-order elliptic differential operators subject to singular domain perturbations and homogeneous boundary conditions of intermediate type. In particular, they discuss a Babuška-type paradox for the case of polyharmonic operators \((-\Delta)^m\), with \(m>2\), subject to intermediate boundary conditions. A way to describe the Babuška paradox in a classical formulation consists in considering a circle \(\Omega\) and an invading sequence of indented polygons \(\Omega_n\). Under certain assumptions, the solution \(u_n\) to the problem \[ \begin{cases} \Delta^2 u_n=f & \text{in } \Omega_n\, , \\ u_n=0 & \text{on } \partial \Omega_n\, , \\ \frac{\partial^2 u_n}{\partial n^2}=0 & \text{on } \partial \Omega_n\, , \end{cases} \] does not converge to the solution of the same problem in \(\Omega\). In the present paper, the authors obtain two main results. The first one is a general stability criterion for self-adjoint elliptic differential operators of order \(2m\) with non-constant coefficients and compact resolvents on varying domains featuring a fast oscillating boundary. The second one concerns a smooth version of the Babuška paradox for polyharmonic operators \((-\Delta)^m\), with \(m>2\). Given a sufficiently regular bounded domain \(W\) in \(\mathbb{R}^{N-1}\), \(N \geq 2\), they set \(\Omega \equiv W \times (-1,0)\) and \[ \Omega_\epsilon = \{(\overline{x},x_N) :\overline{x} \in W, -1< x_N <\epsilon^\alpha b(\overline{x}/\epsilon)\}\, , \] where \(\overline{x} =(x_1,\dots, x_{N-1})\) and \(b\) is a non-constant, smooth, positive, periodic function of period \([-1/2,1/2]^{N-1}\). Then they denote by \(u_\epsilon\) the solution to a Poisson problem for \((-\Delta)^m+\mathbb{I}\) subject to strong intermediate boundary conditions and they study the convergence of \(u_\epsilon\) as \(\epsilon \to 0\). They show that the behavior of the limit \(u\) depends on the parameter \(\alpha\). Namely, if \(\alpha >3/2\) they have stability (i.e., the limit function \(u\) solves the same problem in \(\Omega\)), if \(\alpha <3/2\) they have degeneration (i.e., the limit function \(u\) satisfies Dirichlet boundary condtions on \(W \times \{0\}\)), whereas if \(\alpha = 3/2\) a \textit{strange term} appears.