Block triangular preconditioning for inverse source problems in time-space fractional diffusion equations

TL;DR

This paper introduces a block triangular preconditioning method to efficiently solve large, ill-posed inverse source problems in time-space fractional diffusion equations, improving convergence and stability in numerical solutions.

Contribution

The paper develops and analyzes a novel block triangular preconditioner tailored for large-scale inverse problems governed by TSFDEs, enhancing computational efficiency and robustness.

Findings

Preconditioner significantly accelerates GMRES convergence.

Improves robustness and accuracy of inverse problem solutions.

Effective for large-scale fractional diffusion models.

Abstract

The current work investigates the effectiveness of block triangular preconditioners in accelerating and stabilizing the numerical solution of inverse source problems governed by time-space fractional diffusion equations (TSFDEs). We focus on the recovery of an unknown spatial source function in a multi-dimensional TSFDE, incorporating Caputo time-fractional derivatives and the fractional Laplacian. The inherent ill-posedness is addressed via a quasi-boundary value regularization, followed by a finite difference discretization that leads to large, structured linear systems. We develop and analyze a block triangular preconditioning strategy that mimics the coefficient matrix, while simplifying its structure for computational efficiency. Numerical experiments using the GMRES solver demonstrate that the proposed preconditioner significantly improve convergence rates, robustness, and accuracy, making it well-suited for large-scale, real-world inverse problems involving fractional modeling.