Error analysis of a fully discrete structure-preserving finite element scheme for a diffuse-interface model of tumour growth
Agus L. Soenjaya, Ping Lin, Thanh Tran · 2025
Read Error analysis of a fully discrete structure-preserving finite element scheme for a diffuse-interface model of tumour growth by Agus L. Soenjaya, Ping Lin, Thanh Tran (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-14486.txt.
Introduction
Agus L. Soenjaya, Ping Lin and Thanh Tran analyse errors in 2025 for a finite element scheme that simulates tumour growth with a diffuse interface. The model couples a Cahn-Hilliard equation for tumour phase with a reaction diffusion equation for nutrients. Its energy should decay in a controlled way.
Their fully discrete method uses a scalar auxiliary variable trick with mixed finite elements in space and simple Euler steps in time. The scheme is unconditionally energy stable and preserves mass, matching the continuous energy law. The error analysis proves how accuracy depends on mesh and time step sizes. Structure preserving schemes matter in medicine models because a solver that gains fake energy can invent tumour growth.
Catalog record
| CREATOR | Agus L. Soenjaya, Ping Lin, Thanh Tran |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 137 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-14486
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Error analysis of a fully discrete structure-preserving finite element scheme for a diffuse-interface model of tumour growth Agus L. Soenjaya, Ping Lin, Thanh Tran Abstract . We develop a fully discrete structure-preserving finite element method for a diffuse-interface model of tumour growth. The system couples a Cahn–Hilliard type equation with a nonlinear reaction-diffusion equation for nutrient concentration and admits a dissipative energy law at the continuous level. For the discretisation, we employ a scalar auxiliary variable (SAV) formulation together with a mixed finite element method for the Cahn–Hilliard part and standard conforming finite elements for the reaction-diffusion equation in space, combined with a first-order Euler time-stepping scheme. The resulting method is linear, unconditionally energy-stable, mass-preserving, and inherits a discrete energy dissipation law associated with the SAV-based approximate energy functional, while requiring the solution of only linear systems at each time step. Under suitable regularity assumptions on the exact solution, we derive rigorous error estimates in L 2 L^{2} , H 1 H^{1} , and L ∞ L^{\infty} norms, establishing first-order accuracy in time and optimal-order accuracy in space. A key step in this analysis is the proof of boundedness of the numerical solutions in L ∞ L^{\infty} . Numerical experiments validate the theoretical convergence rates and demonstrate the robustness of the method in capturing characteristic phenomena such as aggregation and chemotactic tumour growth. 1. Introduction Based on the continuum theory of mixtures [ 21 ] , Hawkins-Daarud, van der Zee, and Oden develop a thermodynamically consistent four-species model of tumour growth [ 13 ] (see also [ 5 , 10 ] for related models and [ 30 ] for a simplified derivation). The formulation introduces several unknowns, which we now describe. Let \mathscr{D}\subset\mathbb{R}^{3} be an open and bounded domain, and let u:\mathscr{D}\to\mathbb{R} and n:\mathscr{D}\to\mathbb{R} denote the tumour cell volume fraction and the nutrient-rich extracellular water volume fraction, respectively. Physical consideration dictates that the Helmholtz free energy of the system is given by (1.1) \displaystyle\mathcal{E}[u,n]:=\int_{\mathscr{D}}\left(\frac{\epsilon^{2}}{2}\left|{\nabla u}\right|^{2}+\frac{\lambda}{2}u^{2}+f(u)+\frac{1}{2\delta}n^{2}-\chi_{0}un\right)\mathrm{d}x, where f(u) is the classical Cahn–Hilliard free-energy density function. Here, \epsilon^{2} is a small diffusivity parameter related to the thickness of interfacial layers, \lambda>0 is a constant to be specified later, \chi_{0}>0 is a chemotaxis parameter, and \delta\in(0,1) is a small parameter that determines the relative interaction strength between the cells and nutrient species. The chemical potentials associated to the variables u and n , respectively denoted by \mu and \sigma , are given by the variational derivative of the Helmholtz free energy ( 1.1 ) with respect to each variable. The unknowns described above satisfy a system of coupled nonlinear PDEs which can be formulated as follows: Find (u,\mu,n):\mathscr{D}\to\mathbb{R}^{3} satisfying (1.2a) \displaystyle\partial_{t}u=\Delta\mu+P(u)\cdot(\sigma-\mu)\qquad for (t,\boldsymbol{x})\in(0,T)\times\mathscr{D} , (1.2b) \displaystyle\mu=-\epsilon^{2}\Delta u+\lambda u-\chi_{0}n+f^{\prime}(u)\qquad for (t,\boldsymbol{x})\in(0,T)\times\mathscr{D} , (1.2c) \displaystyle\partial_{t}n=\Delta\sigma-P(u)\cdot(\sigma-\mu)\qquad for (t,\boldsymbol{x})\in(0,T)\times\mathscr{D} , (1.2d) \displaystyle\sigma=\delta^{-1}n-\chi_{0}u\qquad for (t,\boldsymbol{x})\in(0,T)\times\mathscr{D} , (1.2e) \displaystyle u(0,\boldsymbol{x})=u_{0}(\boldsymbol{x}),\quad n(0,\boldsymbol{x})=n_{0}(\boldsymbol{x})\qquad \displaystyle\text{for }\boldsymbol{x}\in\mathscr{D}, (1.2f) \displaystyle\partial_{\boldsymbol{\nu}}u=\partial_{\boldsymbol{\nu}}\mu=\partial_{\boldsymbol{\nu}}n=0\qquad \displaystyle\text{for }(t,\boldsymbol{x})\in(0,T)\times\partial\mathscr{D}, where \boldsymbol{\nu} is the outward pointing normal vector and P is a non-negative proliferation function. The exact assumptions on the functions P and f will be given later. As described in [ 13 ] , the model ( 1.2 ) has a natural four-constituent interpretation: a tumorous phase u\approx 1 , a healthy cell phase u\approx 0 , a nutrient-rich extracellular water phase \sigma\approx 1 , and a nutrient-poor extracellular water phase \sigma\approx 0 . Relevant mathematical results on the problem ( 1.2 ) pertinent to the present work will be reviewed next. Unique existence of weak and strong solutions, as well as the existence of global attractor are established in [ 12 ] (see also [ 9 ] for a simplified model). A characterisation of the singular limit of the problem to the corresponding sharp-interface model is discussed in [ 15 ] , with a rigorous proof given more recently in [ 24 ] . An optimal control problem for ( 1.2 ) is studied in [ 16 ] . On the numerical side, several schemes to solve ( 1.2 ) are described in [ 13 ] based on the convex splitting approach. A linear second-order scheme is further proposed in [ 31 ] , however no error analysis is performed for the schemes in any of these papers. A Fourier-spectral first-order scheme utilising the scalar auxiliary variable (SAV) approach is later proposed in [ 28 ] . While this scheme is linear, mass-preserving, and energy-dissipative, the method is applicable mainly to rectangular domains with periodic boundary conditions, and again no error analysis is given. More recently, numerical analysis of a first-order scheme to solve ( 1.2 ) employing a mass-lumping technique is performed in [ 11 ] , where convergence along a subsequence to the exact solution is shown, albeit without a convergence rate. However, this scheme is nonlinear and does not necessarily guarantee mass conservation or energy dissipation at the discrete level. In contrast, structure-preserving numerical schemes employing Fourier-spectral spatial discretisation combined with the SAV approach are proposed in [ 30 ] to solve the problem under periodic boundary conditions. For this setting, an error analysis of the time-semidiscrete case for the first-order scheme is performed, establishing an optimal temporal order of convergence, although it is still limited to rectangular domains \mathscr{D}\subset\mathbb{R}^{2} with periodic boundary conditions. To the best of our knowledge, a fully discrete , linear , and structure-preserving numerical scheme for ( 1.2 ), together with a rigorous error analysis, has not yet been developed in the literature. In this work, we aim to fill this gap by proposing a fully discrete, linear, and structure-preserving numerical method for ( 1.2 ), and providing its rigorous error analysis. Specifically, we employ an SAV formulation [ 26 ] together with a mixed finite element method for the Cahn–Hilliard part and standard conforming finite elements for the reaction-diffusion part, combined with a first-order Euler time-stepping scheme. This mass-preserving linear scheme inherits a discrete energy dissipation law from the SAV formulation, but differs from the one introduced in [ 30 ] to facilitate the analysis in the fully discrete setting considered here. Furthermore, we consider a three-dimensional convex domain. Despite the strongly coupled and nonlinear nature of the underlying equation, we establish unconditional energy stability and prove optimal-order convergence of the numerical scheme, assuming adequate regularity of the exact solution. The convergence analysis carried out here also differs fundamentally from that in [ 30 ] . In particular, our finite element functions possess only H^{1} spatial regularity, which prevents us from using the H^{2} regularity estimates that play a central role in the time-semidiscrete analysis of [ 30 ] . To overcome this difficulty, we derive a number of delicate estimates for the nonlinear terms, relying on discrete Sobolev embeddings and discrete Gagliardo–Nirenberg inequalities. A key ingredient is the derivation of L^{\infty} bounds for the finite element solution, together with estimates for its discrete time derivative in negative-order Sobolev spaces. Furthermore, we employ the Ritz projection of the solution and establish superconvergence estimates for the difference between the finite element solution and its Ritz projection. These tools allow us to obtain optimal error estimates not only in H^{s} for s\in\{0,1\} , but also in L^{\infty} . The remainder of the paper is organised as follows. In Section 2 , we introduce the notations and assumptions, present a formulation of the problem