The automatic solution of partial differential equations using a global spectral method

Abstract

A spectral method for solving linear partial differential equations (PDEs) with variable coefficients and general boundary conditions defined on rectangular domains is described, based on separable representations of partial differential operators and the one-dimensional ultraspherical spectral method. If a partial differential operator is of splitting rank $2$, such as the operator associated with Poisson or Helmholtz, the corresponding PDE is solved via a generalized Sylvester matrix equation, and a bivariate polynomial approximation of the solution of degree $(n_x,n_y)$ is computed in $\mathcal{O}((n_x{n}_y{)}^{3/2})$ operations. Partial differential operators of splitting rank $\ge 3$ are solved via a linear system involving a block-banded matrix in $\mathcal{O}(\min (n_x^3{n}_y,n_x{n}_y^3))$ operations. Numerical examples demonstrate the applicability of our 2D spectral method to a broad class of PDEs, which includes elliptic and dispersive time-evolution equations. The resulting PDE solver is written in MATLAB and is publicly available as part of CHEBFUN. It can resolve solutions requiring over a million degrees of freedom in under $60$ seconds. An experimental implementation in the Julia language can currently perform the same solve in $10$ seconds.

22 pages

Cited by in corpus (21)

• Tensor calculus in polar coordinates using Jacobi polynomials
• Geometry-dependent viscosity reduction in sheared active fluids
• The ultraspherical spectral element method
• Functional Tensor-Train Chebyshev Method for Multidimensional Quantum Dynamics Simulations
• FEAST for differential eigenvalue problems
• Computing Equilibrium Measures with Power Law Kernels
• On symmetrizing the ultraspherical spectral method for self-adjoint problems
• Fast global spectral methods for three-dimensional partial differential equations
• A fast sparse spectral method for nonlinear integro-differential Volterra equations with general kernels
• Computing with functions in the ball
• Approximation in the extended functional tensor train format
• A static memory sparse spectral method for time-fractional PDEs
• A sparse spectral method for Volterra integral equations using orthogonal polynomials on the triangle
• Two spectral methods for 2D quasi-periodic scattering problems
• Anderson localisation in two dimensions: insights from Localisation Landscape Theory, exact diagonalisation, and time-dependent simulations
• Polar differentiation matrices for the Laplace equation in the disk subjected to nonhomogeneous Dirichlet, Neumann and Robin boundary conditions and the biharmonic equation subjected to nonhomogeneous Dirichlet conditions
• A fast and spectrally convergent algorithm for rational-order fractional integral and differential equations
• Hierarchical adaptive low-rank format with applications to discretized PDEs
• On well-conditioned spectral collocation and spectral methods by the integral reformulation
• Computing the Tracy-Widom Distribution for Arbitrary $\beta >0$
• Rational function approximation with normalized positive denominators