🔢 mathematics

Error analysis of a fully discrete structure-preserving finite element scheme for a diffuse-interface model of tumour growth

This paper presents and rigorously analyzes a linear, fully discrete, structure-preserving finite element scheme for a diffuse-interface tumour growth model that combines a scalar auxiliary variable formulation with mixed and conforming elements to achieve unconditional energy stability, mass preservation, and optimal first-order convergence rates.

Original authors: Agus L. Soenjaya, Ping Lin, Thanh Tran

Published 2026-07-28

📖 3 min read🧠 Deep dive

CC BY 4.0

Original authors: Agus L. Soenjaya, Ping Lin, Thanh Tran

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to model how a crowd of people moves through a city, but instead of people, you are tracking blobs of cancer cells and the food they need to survive. This is the world of mathematical biology, where scientists use complex equations to predict how tumours grow, spread, and interact with their environment. The challenge is that tumours don't have sharp, hard edges like a marble; they have fuzzy, blurry boundaries where healthy tissue slowly turns into cancerous tissue. To describe this "fuzzy" transition, scientists use a tool called a "diffuse-interface model." Think of it like a smoothie: you can't point to exactly where the strawberry ends and the banana begins, but you can measure the concentration of fruit at every point.

However, simulating these fuzzy tumours on a computer is incredibly tricky. The equations that describe them are like a tangled knot of spaghetti: they are non-linear (meaning small changes can cause huge, unpredictable effects) and they must obey strict physical laws, such as conserving mass (you can't create or destroy cells out of thin air) and dissipating energy (tumours naturally lose energy as they grow, similar to a hot cup of coffee cooling down). If a computer simulation ignores these laws, it might produce nonsense results, like a tumour that grows infinitely fast or disappears into thin air. The big question for mathematicians has been: Can we build a computer program that is fast enough to run, simple enough to solve, but strict enough to never break these physical rules?

This paper, written by Agus L. Soenjaya, Ping Lin, and Thanh Tran, says "Yes, we can." The authors have developed a new, clever way to simulate tumour growth that acts like a digital guardian of physics. They created a method that is "linear," meaning it solves simple, straight-line equations at every step instead of getting bogged down in complex, messy calculations. This makes the simulation much faster and easier to run. But the real magic is that they used a technique called the "Scalar Auxiliary Variable" (SAV) to ensure the simulation never breaks the rules of energy and mass.

Imagine you are playing a video game where your character must stay within a specific energy budget. If the character tries to jump too high, the game automatically adjusts to keep the energy correct. That is what this new method does for tumour simulations. It introduces a "helper variable" (the SAV) that acts like a referee, constantly checking the math to ensure the tumour doesn't magically gain or lose energy or mass. The authors proved mathematically that their method is stable and accurate, showing that as they make the computer grid finer (like zooming in with a better camera), the results get closer and closer to the true answer.

In their experiments, they simulated a single tumour growing steadily, a group of tumours merging into a larger mass, and even a three-dimensional tumour that moves toward food sources (a behavior called chemotaxis). In every case, the simulation held up: the tumours grew realistically, the total amount of "stuff" (cells and nutrients) stayed constant, and the energy levels dropped exactly as nature intended. Even when they tested the method with "messy" starting conditions that the theory didn't strictly guarantee would work, the simulation remained stable and captured the expected behavior. The paper concludes that this approach is a robust, reliable tool for understanding how tumours behave, offering a way to run complex biological simulations without the computer crashing or producing impossible results.

Technical Summary: Error Analysis of a Fully Discrete Structure-Preserving Finite Element Scheme for a Diffuse-Interface Model of Tumour Growth

Problem Statement
This work addresses the numerical approximation of a thermodynamically consistent, four-species model of tumour growth based on the continuum theory of mixtures. The model couples a Cahn–Hilliard type equation for the tumour cell volume fraction ( $u$ ) with a nonlinear reaction-diffusion equation for the nutrient concentration ( $n$ ). The system is governed by a Helmholtz free energy functional and admits a dissipative energy law at the continuous level. While previous studies have proposed various numerical schemes (including nonlinear convex splitting and time-semidiscrete SAV approaches), a fully discrete, linear, and structure-preserving scheme accompanied by a rigorous error analysis for this specific coupled system in three dimensions has been lacking in the literature.

Methodology
The authors propose a linear, fully discrete finite element method that preserves the structural properties of the continuous model. The methodology combines:

Scalar Auxiliary Variable (SAV) Formulation: To handle the nonlinear free-energy density $g(u)$ , the system is reformulated using an auxiliary variable $r(t) = \sqrt{E_1[u] + B}$ , where $E_1$ is the non-quadratic part of the free energy. This allows for the construction of a linear scheme while maintaining energy stability.
Spatial Discretization:
- A mixed finite element method is employed for the Cahn–Hilliard part to reduce the fourth-order equation into a coupled system of second-order equations.
- Standard conforming finite elements are used for the reaction-diffusion equation.
Temporal Discretization: A first-order semi-implicit time-stepping scheme is utilized.

The resulting algebraic system is linear at each time step. The implementation of the SAV term is handled efficiently via a rank-one correction to the sparse finite element block matrix, solvable using the Sherman–Morrison formula, avoiding the need for dense matrix assembly or nonlinear solvers.

Key Contributions and Analysis
The primary contribution of this paper is the rigorous error analysis of the proposed scheme under suitable regularity assumptions on the exact solution. The analysis establishes:

Structural Properties: The scheme is proven to be unconditionally energy-stable (satisfying a discrete energy dissipation law for a modified energy functional) and mass-preserving.
Well-posedness: The existence and uniqueness of the solution to the linear algebraic system at each time step are established.
Error Estimates: The authors derive rigorous error estimates in $L^2$ $L^{2}$ , $H^1$ $H^{1}$ , and $L^\infty$ $L^{\infty}$ norms. The analysis demonstrates:
- First-order accuracy in time ( $O(\tau)$ ).
- Opt-order accuracy in space ( $O(h^{q+1})$ for $L^2$ and $H^1$ , where $q$ is the polynomial degree).
Technical Innovations in Analysis:
- Unlike previous time-semidiscrete analyses (e.g., in 2D), this work addresses the fully discrete setting in a 3D convex domain.
- The authors overcome the lack of $H^2$ regularity estimates for fully discrete solutions (which are only $H^1$ ) by deriving delicate estimates for nonlinear terms using discrete Sobolev embeddings and discrete Gagliardo–Nirenberg inequalities.
- A key step involves proving the boundedness of numerical solutions in $L^\infty$ and establishing superconvergence estimates for the difference between the finite element solution and its Ritz projection.

Results
Theoretical results are validated through numerical simulations using the FEniCS package. The experiments include:

Single Tumour Growth: Demonstrating steady growth and mass conservation.
Aggregation of Multiple Tumours: Simulating the merging of two tumours.
Convergence Rates: Verifying the theoretical first-order temporal convergence and optimal spatial convergence rates, even for initial data with lower regularity than assumed in the theory.
3D Chemotactic Growth: Simulating a three-dimensional tumour exhibiting chemotactic behavior towards nutrient sources. Notably, the simulation remains stable even when the sufficient coercivity condition ( $\lambda > \chi_0^2 \delta$ ) used in the theoretical analysis is not strictly satisfied, suggesting robustness beyond the proven bounds.

Significance
The paper claims to fill a gap in the literature by providing the first fully discrete, linear, and structure-preserving numerical scheme for this specific tumour growth model, complete with a rigorous error analysis in 3D. The method is significant because it avoids the computational cost of nonlinear solvers while guaranteeing mass conservation and energy stability. The analysis extends the understanding of structure-preserving schemes to fully discrete settings in higher dimensions, where controlling nonlinear terms is substantially more delicate than in time-semidiscrete or lower-dimensional cases. The numerical experiments confirm that the method is robust and capable of capturing characteristic phenomena such as aggregation and chemotaxis.

More like this