MA5251: Spectral Methods & Applications

Weizhu Bao

Department of Mathematics
& Center for Computational Science and Engineering
National University of Singapore
Email: bao@math.nus.edu.sg
URL: http://www.math.nus.edu.sg/~bao
 Contents

Introduction and Preliminaries

Some numerical examples
Review of different numerical methods of PDE
Historical background of spectral methods
Some examples of spectral methods
Fourier series and orthogonal polynomials
Review of iterative solvers and preconditioning
Review of time discretization methods
Spectral-Collocation Methods
Introduction
Differentiation matrices
Fourier, Chebyshev collocation methods
 Contents

Spectral-Galerkin methods
Introduction
Fourier spectral method
Legendre spectral method
Chebyshev spectral method
Error estimates
Spectral methods in unbounded domains
Introduction
Hermite spectral method
Laguerre specreal methods,.
 Contents

Applications
In fluid dynamics
In heat transfer
In material sciences
In quantum physics and nonlinear optics
In plasma and particle physics
In biology
..
Dynamics of soliton in quantum physics
Wave interaction in plasma physics
Wave interaction in plasma physics
Wave interaction in plasma physics
Wave interaction of plasma physics
Wave interaction in particle physics
Vortex-pair dynamics in superfluidity
Vortex-dipole dynamics in superfluidity
Vortex lattice dynamics in superfluidity
Vortex lattice dynamics in superfluidity
Vortex lattice dynamics in BEC
 Main numerical methods for PDEs

Finite difference method (FDM) MA5233

Advantages:
Simple and easy to design the scheme
Flexible to deal with the nonlinear problem
Widely used for elliptic, parabolic and hyperbolic equations
Most popular method for simple geometry, .
Disadvantages:
Not easy to deal with complex geometry
Not easy for complicated boundary conditions
..
 Main numerical methods

Finite element method (FEM) MA5240

Advantages:
Flexible to deal with problems with complex geometry and
complicated boundary conditions
Keep physical laws in the discretized level
Rigorous mathematical theory for error analysis
Widely used in mechanical structure analysis, computational fluid
dynamics (CFD), heat transfer, electromagnetics,
Disadvantages:
Need more mathematical knowledge to formulate a good and
equivalent variational form
 Main numerical methods

Spectral method This module

High (spectral) order of accuracy
Usually restricted for problems with regular geometry
Widely used for linear elliptic and parabolic equations on
regular geometry
Widely used in quantum physics, quantum chemistry, material
sciences,
Not easy to deal with nonlinear problem
Not easy to deal with hyperbolic problem
..
 Main numerical methods
Finite volume method (FVM) MA5250
Flexible to deal with problems with complex geometry and complicated
boundary conditions
Keep physical laws in the discretized level
Widely used in CFD
Boundary element method (BEM)
Reduce a problem in one less dimension
Restricted to linear elliptic and parabolic equations
Need more mathematical knowledge to find a good and equivalent
integral form
Very efficient fast Poisson solver when combined with the fast multipole
method (FMM), ..
 Historical background

Method of weighted residuals (MWR) Finlayson & Scriven (1966)

Trial functions (or expansion or approximation functions): are
used as the basis functions for a truncated series expansion of the
solution.
Test functions (or weight functions): are used to ensure that the
differential equation is satisfied as closely as possible by the truncated
series expansion.
This is achieved by minimizing the residual, i.e. the error in the
differential equation produced by using the truncated expansion instead of
the exact solution, with respect to a suitable norm.
An equivalent requirement is that the residual satisfy a suitable
orthogonality condition with respect to each of the test functions.
 Historical background

Trial functions
Spectral method: infinitely differentiable global functions, i.e.
eigenfunctions of singular Sturm-Liouville problems
Finite Element Method (FEM): partition the domain into small elements,
and a trial function (usually polynomial) is specified in each element and
thus local in character & well suited for handling complex geometries.
Finite Difference Method (FDM): similar as FEM.
Test functions
Spectral methods: three different ways
FEM: similar as trial functions
FDM: Dirac delta functions centered at the grid points
 Historical background

Different test functions of spectral methods

Galerkin method: same as the trial functions which are infinitely
smooth functions & individually satisfy the boundary conditions. The
differential equation is enforced by requiring that the integral of the
residual times each test function be zero.
Collocation method: Dirac delta functions centered at the
collocation points. The differential is required to be satisfied exactly at the
collocation points.
Spectral tau method: Similar as the Galerkin method except that no
need the trial and test functions satisfy the boundary conditions. A
supplementary set of equations is used to apply the boundary conditions.
 Historical background

Collocation approach (simplest of the MWR) Slater

(1934); Kantorovic (1934); Frazer, Jones and Skan (1937).
Proper choice of trial functions and distribution of
collocation points Lanczos (1938)
Orthogonal collocation method Clenshaw (1957); Clenshaw
and Norton (1963); Wright (1964).
Earliest application of spectral methods to PDE Kreiss
and Oliger (1972)Fourier method; Orszag (1972) pseudospectral.
Spectral-Galerkin method Silberman (1954) in meteorological
modeling; Orszag (1969, 1970); etc.
 Historical background

Theory of spectral method -- Gottlieb and Orszag (1977)

Symposium Proceedings Voigt, Gottlieb and Hussaini (1984)
First International Conference on Spectral and High
Order Methods (ICOSAHOM) -- Como, Italy in 1989. It
becomes series conference every three years. The next one is
http://www.math.ntnu.no/icosahom/