Conceptual

Rate-Optimal Adaptive Fictitious Domain Solver for Elliptic PDEs via Inexact Uzawa Iteration

A fictitious domain formulation embeds an irregular physical domain in a simple rectangular mesh and imposes the Dirichlet boundary condition with a Lagrange multiplier, producing a symmetric saddle-point problem whose solution is non-smooth across the embedded interface. This concept shows how to solve that saddle-point system with a nested, inexact, wavelet-preconditioned Uzawa iteration whose inner elliptic problems — which carry H^-1 forcing data supported on the interface curve — are approximated by an adaptive finite element method driven by bulk-chasing on a residual estimator. The payoff a student should be able to argue is that the combined outer/inner scheme still converges at the best rate the trial spaces allow, and how that best rate degrades in three or more dimensions because of the interface jump.