Chebyshev spectral collocation method approximations of the Stokes eigenvalue problem based on penalty techniques (Q2273045)

The authors study numerical methods for solving the Stokes eigenvalue problem: Find \([\pmb{u},p,\lambda]\), such that \(-\nu\Delta\pmb{u} + \nabla p=\lambda\pmb{u}\) in \(\Omega\), \(\nabla \cdot \pmb{u}=0\) in \(\Omega\) and \(\pmb{u}=\pmb{0}\) on \(\partial\Omega\). Hereby denotes \(\pmb{u}\) the displacement or velocity field and \(p\) the pressure, \(\nu>0\) is a physical parameter and \(\Omega\) is a bounded polyhedral domain in \(\mathbb{R}^{d}\), \(d=2,3\). For the spatial discretisation of this problem a Chebyshev spectral collocation method is used. Two penalty methods for solving the eigenvalue problem are discussed, namely the classical penalty method and the iterative penalty method. For solving the nonlinear problem in each iteration step, Newton's method is used. The convergence behaviour of the presented methods is demonstrated by two numerical examples, the Stokes eigenvalue problem in a square and in a cube. The influence of the choice of the penalty parameter on the convergence is studied.