Abstract:The density functional theory has become a popular tool for electronic structure calculations in various fields include chemistry, material science and biology in recent years. The thesis is devoted to the mathematical analysis and algorithm developments for density functional theory. The dissertation consists of three parts. In the first part, we introduce the density functional theory, in particular the Kohn-Sham density functional theory. We begin with the mathematical formulation and structure of density functional theory. By exploiting the invariance properties of the model, we reformulate density functional theory as a variational problem on subspaces, the minimizer is the occupied space in quantum mechanics terminology. We will then focus on the difference between metallic and insulating systems, in particular, the locality principle for insulators: The exponential decay of the Green's function, the density matrix and the Wannier functions. The second part is some mathematical results related with the Kohn-Sham density functional theory. To gain better understanding of the Kohn-Sham map arising from the theory, we analyze the linearized Kohn-Sham map in detail, in particular its action on different function spaces. The rest of the second part is devoted to the study of the electronic structure of elastically deformed crystals. For a reduced model, we validate the Cauchy-Born rule in the continuum limit under stability conditions. The notion of Wannier functions is generalized to elastically deformed systems, and we also establish the Cauchy-Born rule for Wannier functions, and hence for the electronic structure. We discuss about numerical issues related with density functional theory in the third part. We first discuss the self-consistent iteration and show different behaviors in convergence for insulating and metallic systems. Then we focus on the developments of more efficient algorithms for solving the model: Linear scaling algorithms for insulating systems based on the locality principle, and also efficient algorithms for metallic systems based on better representation of the Fermi operator and fast algorithm to extract the diagonal of an inverse matrix. We then discuss the sublinear scaling algorithms based on the knowledge of elastically deformed system. This is an example of multiscale methods, and as such, we present a thorough discussion about the treatment of boundary between the fine-scale and the coarse-scale regions and how to correctly formulate the problem to avoid artifacts on the boundary.

MultiShape: A Spectral Element Method, with Applications to Dynamic Density Functional Theory and PDE-Constrained Optimization

Shape optimisation with multiresolution subdivision surfaces and immersed finite elements

MultiMesh Finite Element Methods: Solving PDEs on Multiple Intersecting Meshes

Shape optimization with virtual element method

Shape Optimization using the Finite Element Method on Multiple Meshes

Numerical Methods for Shape Optimal Design of Fluid-Structure Interaction Problems

A pseudo-spectral multiscale method: Interfacial conditions and coarse grid equations

Optimization of piecewise smooth shapes under uncertainty using the example of Navier-Stokes flow

A multi-physics compiler for generating numerical solvers from differential equations

Space-Time-Parallel Poroelasticity Simulation

A Generic MATLAB Toolbox to Approximate PDEs Using Computational Geometry

A Lagrangian shape and topology optimization framework based on semi-discrete optimal transport

Multiscale Analysis of Density Functional Theory

Automatic shape derivatives for transient PDEs in FEniCS and Firedrake

A product shape manifold approach for optimizing piecewise-smooth shapes

Multi-physics and Multi-scale Electromagnetic Modeling and High Performance Algorithms

A Spectral Multiscale Method for Wave Propagation Analysis: Atomistic-Continuum Coupled Simulation

Adaptive selection of shape parameters for MQRBF in arbitrary scattered data: enhancing finite difference solutions for complex PDEs

A multi-mesh finite element method for Lagrange elements of arbitrary degree

A dual-pairing summation-by-parts finite difference framework for nonlinear conservation laws

A Monotonic Algorithm for Eigenvalue Optimization in Shape Design Problems of Multi-Density Inhomogeneous Materials