A Unified Spectral Method for FPDEs with Two-sided Derivatives; A Fast Solver

TL;DR

This paper introduces a unified spectral Petrov-Galerkin method for solving complex fractional PDEs with two-sided derivatives, employing eigenfunction bases for efficiency and a novel fast solver for high-dimensional systems, validated through numerical tests.

Contribution

The paper develops a unified spectral method with a fast solver for high-dimensional FPDEs with two-sided derivatives, independent of fractional orders, and demonstrates its efficiency and accuracy.

Findings

The method achieves optimal complexity of lops for high-dimensional problems.

Numerical tests confirm the convergence and efficiency of the proposed approach.

The solver significantly reduces computational time compared to traditional methods.

Abstract

We develop a unified Petrov-Galerkin spectral method for a class of fractional partial differential equations with two-sided derivatives and constant coefficients of the form $ _{0}{\mathcal{D}}_{t}^{2\tau}u^{} + \sum_{i=1}^{d}$ $[c_{l_i}$ $_{a_i}{\mathcal{D}}_{x_i}^{2\mu_i} u^{} +c_{r_i}$ $_{x_i}{\mathcal{D}}_{b_i}^{2\mu_i}$ $u^{} ] +$ $\gamma$ $u^{} = \sum_{j=1}^{d} [ \kappa_{l_j}$ $_{a_j}{\mathcal{D}}_{x_j}^{2\nu_j} u^{}$ $+\kappa_{r_j}$ $_{x_j}{\mathcal{D}}_{b_j}^{2\nu_j}$ $u^{} ]$ $+ f$, where $2\tau \in (0,2)$, $2\mu_i \in (0,1)$ and $2\nu_j \in (1,2)$, in a ($1+d$)-dimensional \textit{space-time} hypercube, $d = 1, 2, 3, \cdots$, subject to homogeneous Dirichlet initial/boundary conditions. We employ the eigenfunctions of the fractional Sturm-Liouville eigen-problems of the first kind in \cite{zayernouri2013fractional}, called \textit{Jacobi poly-fractonomial}s, as temporal bases, and the eigen-functions of the boundary-value problem of the second kind as temporal test functions. Next, we construct our spatial basis/test functions using Legendre polynomials, yielding mass matrices being independent of the spatial fractional orders ($\mu_i, \, \nu_j, \, i, \,j=1,2,\cdots,d$). Furthermore, we formulate a novel unified fast linear solver for the resulting high-dimensional linear system based on the solution of generalized eigen-problem of spatial mass matrices with respect to the corresponding stiffness matrices, hence, making the complexity of the problem optimal, i.e., $\mathcal{O}(N^{d+2})$. We carry out several numerical test cases to examine the CPU time and convergence rate of the method. The corresponding stability and error analysis of the Petrov-Galerkin method are carried out in \cite{samiee2016Unified2}.