UC San Diego

In this thesis we explore convergence theory for adaptive mixed finite element methods. In particular, we introduce an a posteriori error-indicator, and prove convergence and optimality results for the mixed formulation of the Hodge Laplacian posed on domains of arbitrary dimensionality and topology in R/n. After developing this framework, we introduce a new algorithm and extend our theory and results to problems posed on Euclidean hypersurfaces. We begin by introducing the finite element exterior calculus framework, which is the key tool allowing us to address the convergence proofs in such generality. This introduction focuses on the fundamentals of the well- developed a priori theory and the results needed to extend the core of this theory to problems posed on surfaces. A basic set of results needed to develop adaptivity in this framework is also established. We then introduce an adaptive algorithm, and show convergence using this infrastructure as a tool to generalize existing finite element theory. The algorithm is then shown to be computationally optimal through a series of complexity analysis arguments. Finally, a second algorithm is introduced for problems posed on surfaces, and our original convergence and optimality results are extended using properties of specific geometric maps between surfaces