Parameter identification via optimal control for a Cahn--Hilliard-chemotaxis system with a variable mobility

TL;DR

This paper addresses the inverse problem of parameter identification in a tumor growth model using PDE-constrained optimal control, establishing theoretical conditions and proposing a numerical solution scheme.

Contribution

The paper introduces the first-order optimality conditions for a variable mobility Cahn--Hilliard-chemotaxis system and develops a numerical method for inverse parameter estimation.

Findings

Proved continuous dependence of solutions on parameters in 2D.

Derived first-order optimality conditions for the inverse problem.

Implemented a trust-region Gauss-Newton method for numerical solution.

Abstract

We consider the inverse problem of identifying parameters in a variant of the diffuse interface model for tumour growth model proposed by Garcke, Lam, Sitka and Styles (Math. Models Methods Appl. Sci. 2016). The model contains three constant parameters; namely the tumour growth rate, the chemotaxis parameter and the nutrient consumption rate. We study the inverse problem from the viewpoint of PDE-constrained optimal control theory and establish first order optimality conditions. A chief difficulty in the theoretical analysis lies in proving high order continuous dependence of the strong solutions on the parameters, in order to show the solution map is continuously Fr\'{e}chet differentiable when the model has a variable mobility. Due to technical restrictions, our results hold only in two dimensions for sufficiently smooth domains. Analogous results for polygonal domains are also shown for the case of constant mobilities. Finally, we propose a discrete scheme for the numerical simulation of the tumour model and solve the inverse problem using a trust-region Gauss--Newton approach.