Approximate Solutions to Fractional Subdiffusion Equations: The heat-balance integral method

TL;DR

This paper introduces an integral method for approximating solutions to fractional subdiffusion equations, offering an alternative to hypergeometric function solutions, with demonstrated applications and error analysis.

Contribution

It develops a novel integral approach using prescribed profiles to solve fractional subdiffusion equations without requiring initial derivatives, improving computational efficiency.

Findings

The integral method accurately approximates solutions to fractional subdiffusion equations.

Profiles can be optimized to minimize approximation errors.

Numerical results compare favorably with known solutions.

Abstract

The work presents integral solutions of the fractional subdiffusion equation by an integral method, as an alternative approach to the solutions employing hypergeometric functions. The integral solution suggests a preliminary defined profile with unknown coefficients and the concept of penetration (boundary layer). The prescribed profile satisfies the boundary conditions imposed by the boundary layer that allows its coefficients to be expressed through its depth as unique parameter. The integral approach to the fractional subdiffusion equation suggests a replacement of the real distribution function by the approximate profile. The solution was performed with Riemann -Liouville time-fractional derivative since the integral approach avoids the definition of the initial value of the time-derivative required by the Laplace transformed equations and leading to a transition to Caputo derivatives. The method is demonstrated by solutions to two simple fractional subdiffusion equations (Dirichlet problems): 1) Time-Fractional Diffusion Equation, and 2) Time-Fractional Drift Equation, both of them having fundamental solutions expressed through the M-Write function. The solutions demonstrate some basic issues of the suggested integral approach, among them: a) Choice of the profile, b) Integration problem emerging when the distribution (profile) is replaced by a prescribed one with unknown coefficients; c) Optimization of the profile in view to minimize the average error of approximations; d) Numerical results allowing comparisons to the known solutions expressed to the M-Write function and error estimations.