Convergence analysis of a numerical scheme for a tumour growth model

TL;DR

This paper analyzes a numerical scheme for a complex tumour growth model involving coupled hyperbolic, elliptic, and parabolic equations, proving convergence to a weak solution and validating results with numerical tests.

Contribution

It introduces a reformulation to handle time-dependent boundaries and proves convergence of a mixed finite volume and finite element scheme for the model.

Findings

Existence of a convergent subsequence of numerical approximations.

The limit is a weak solution of the tumour growth model.

Numerical tests confirm theoretical convergence and solution properties.

Abstract

We consider a one--spatial dimensional tumour growth model [2, 3, 4] that consists of three dependent variables of space and time: volume fraction of tumour cells, velocity of tumour cells, and nutrient concentration. The model variables satisfy a coupled system of semilinear advection equation (hyperbolic), simplified linear Stokes equation (elliptic), and semilinear diffusion equation (parabolic) with appropriate conditions on the time-dependent boundary, which is governed by an ordinary differential equation. We employ a reformulation of the model defined in a larger, fixed time-space domain to overcome some theoretical difficulties related to the time-dependent boundary. This reformulation reduces the complexity of the model by removing the need to explicitly track the time-dependent boundary, but nonlinearities in the equations, noncoercive operators in the simplified Stokes equation, and interdependence between the unknown variables still challenge the proof of suitable a priori estimates. A numerical scheme that employs a finite volume method for the hyperbolic equation, a finite element method for the elliptic equation, and a backward Euler in time--mass lumped finite element in space method for the parabolic equation is developed. We establish the existence of a time interval $(0,T_{\ast})$ over which, using compactness techniques, we can extract a convergent subsequence of the numerical approximations. The limit of any such convergent subsequence is proved to be a weak solution of the continuous model in an appropriate sense, which we call a threshold solution. Numerical tests and justifications that confirm the theoretical findings conclude the paper.