New technique for formulation of domain decomposition algorithms (Q2207511)

The present work introduces a new method for constructing domain decomposition algorithms based on optimal control theory, the theory of inverse and ill-posed problems, application of adjoint equations, and modern iterative processes. In a first step, the original domain is divided into subdomains. Each subproblem stated in subdomains is equipped with matching conditions on their boundaries. Some of these matching conditions are subsequently reformulated in terms of ``boundary functions.'' These functions are treated as additional controls that have to be determined alongside the solutions of each individual subdomain. The remaining matching conditions are used as closure equations. Overall, an optimal control problem is obtained that can be solved by traditional methods. Starting from the overall problem formulation, first order (necessary) optimality conditions are derived (i.e., first variations of the control problem). Based on these optimality conditions, a generalized formulation is developed to obtain generic operator equations. Subsequently, uniqueness results of the considered problem are provided. Based on the derived operator equation, a (convergent) iterative algorithm for domain decomposition is presented, and assessed empirically through numerical experimentation. The work concludes with a discussion of of other approaches to the formulation of domain decomposition algorithms based on the proposed approach.