Spatio-temporal integral equation methods with applications Public Deposited

Downloadable Content

Download PDF
Last Modified
  • March 21, 2019
Creator
  • Chen, Dangxing
    • Affiliation: College of Arts and Sciences, Department of Mathematics
Abstract
  • Electromagnetic interactions are vital in many applications including physics, chemistry, material sciences and so on. Thus, a central problem in physical modeling is the electromagnetic analysis of materials. Here, we consider the numerical solution of the Maxwell equation for the evolution of the electromagnetic field given the charges, and the Newton or Schr\\"odinger equation for the evolution of particles. By combining integral equation techniques with new spectral deferred correction algorithms in time and hierarchical methods in space, we develop fast solvers for the calculation of electromagnetism with relaxations of the model in different scenarios. The dissertation consists of two parts, aiming to resolve the challenges in the temporal and spatial direction, respectively. In the first part, we study a new class of time stepping methods for time-dependent differential equations. The core algorithm uses the pseudo-spectral collocation formulation to discretize the Picard type integral equation reformulation, producing a highly accurate and stable representation, which is then solved via the deferred correction technique. By exploiting the mathematical properties of the formulation and the convergence procedure, we develop some new preconditioning techniques from different perspectives that are accurate, robust, and can be much more efficient than existing methods. As is typical of spectral methods, the solution to the discretization is spectral accurate and the time step-size is optimal, though the cost of solving the system can be high. Thus, the solver is particularly suited to problems where very accurate solutions are sought or large time-step is required, e.g., chaotic systems or long-time simulation. In the second part, we study the hierarchical methods with emphasis on the spatial integral equations. In the first application, we implement a parallel version of the adaptive recursive solver for two-point boundary value problem by Cilk multithreaded runtime system based on the integral equation formulation. In the second application, we apply the hierarchical method to two-layered media Helmholtz equations in the acoustic and electromagnetic scattering problems. With the method of images and integral representations, the spatially heterogeneous translation operators are derived with rigorous error analysis, and the information is then compressed and spread in a fashion similar to fast multipole methods. The preliminary results suggest that our approach can be faster than existing algorithms with several orders of magnitude. We demonstrate our solver on a number of examples and discuss various useful extensions. Preliminary results are favorable and show the viability of our techniques for integral equations. Such integral equation methods could well have a broad impact on many areas of computational science and engineering. We describe further applications in biology, chemistry, and physics, and outline some directions for future work.
Date of publication
Keyword
DOI
Resource type
Rights statement
  • In Copyright
Advisor
  • Marzuola, Jeremy
  • Huang, Jingfang
  • Newhall, Katherine
  • Lu, Jianfeng
  • Kanai, Yosuke
Degree
  • Doctor of Philosophy
Degree granting institution
  • University of North Carolina at Chapel Hill
Graduation year
  • 2017
Language
Parents:

This work has no parents.

Items