A posteriori error analysis of a virtual element method for stress-assisted diffusion problems
Franco Dassi, Rekha Khot, Andrés E. Rubiano, Ricardo Ruiz-Baier · 2025
Read A posteriori error analysis of a virtual element method for stress-assisted diffusion problems by Franco Dassi, Rekha Khot, Andrés E. Rubiano, Ricardo Ruiz-Baier (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-2504-00648.txt.
Introduction
Franco Dassi and three co-authors build error checks for a math method used on spread problems tied to stress. The model links how a solid bends with how a substance spreads through it, a pair that pulls on each other both ways. Their 2025 work asks how a user can trust the numbers such a method gives.
The team derives checks from what is left over after each run, weighted so the checks stay sound across settings. They prove the checks both flag real error and avoid false alarms. The proofs lean on split tools for the flow fields, with new steps spelled out for three dimensions. Tests in two and three dimensions show the checks track true error well. Mesh fixes guided by them should be safe.
Catalog record
| CREATOR | Franco Dassi, Rekha Khot, Andrés E. Rubiano, Ricardo Ruiz-Baier |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 146 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-00648
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A posteriori error analysis of a virtual element method for stress-assisted diffusion problems Franco Dassi, Rekha Khot, Andrés E. Rubiano, Ricardo Ruiz-Baier Abstract We develop and analyse residual-based a posteriori error estimates for the virtual element discretisation of a nonlinear stress-assisted diffusion problem in two and three dimensions. The model problem involves a two-way coupling between elasticity and diffusion equations in perturbed saddle-point form. A robust global inf-sup condition and Helmholtz decomposition for 𝐇 ( div , Ω ) \mathbf{H}(\mathrm{div},\Omega) lead to a reliable and efficient error estimator based on appropriately weighted norms that ensure parameter robustness. The a posteriori error analysis uses novel quasi-interpolation operators for Stokes and edge virtual element spaces, and we include the proofs of such operators with estimates in 3D for completeness. Finally, we present numerical experiments in both 2D and 3D to demonstrate the optimal performance of the proposed error estimator. MSC Classification]65N30, 65N12, 65N15, 74F25. keywords A posteriori error analysis in 2D and 3D, virtual element method, stress-assisted diffusion, perturbed saddle-point problems, interpolation operators. pacs [ 1 Introduction The physical interaction between solute diffusion, driven by chemical reactions, in a deformable medium was first formally addressed in Unger and Aifantis (1983) ; Wilson and Aifantis (1982) known as the stress-assisted diffusion model. This phenomenon appears in a variety of scientific, medical, and engineering applications such as the lithiation of ion batteries Taralov (2015) , cardiac tissue electromechanics Loppini et al. (2018) , semiconductor fabrication Shaw (2007) , biomechanics of brain tissue Prevost et al. (2011) , among others. These applications operate across multiple scales, and motivating the need of robust models. The mathematical analysis and discretisation of this model have been conducted using mixed-primal and mixed-mixed finite element methods (FEM) in Hilbert spaces Gatica et al. (2018) , pseudo-stress-based formulations using Banach spaces Gatica et al. (2022b) , and well-posedness results for a primal formulation Lewicka and Mucha (2016) ; Malaeke and Asghari (2023) . In the virtual element method (VEM) framework, this model was first studied in Khot et al. (2025) , which introduced a robust mixed-mixed formulation with parameter-weighted norms ensuring the unique solvability of two uncoupled systems across different parameter scales (i.e. robustness is achieved), along with a fixed-point argument establishing well-posedness for the fully coupled equations. Our goal here is to derive an adaptive mesh refinement strategy driven by a posteriori error estimators for this type of formulations, and extend reliability and efficiency results to the parameter-robust setting. Adaptive schemes driven by a posteriori error estimators allow optimal convergence recovery in cases such as singular solutions, rough data, and complex geometries (e.g., non-convex domains and sharp corners). The key advantage of the VEM framework in adaptive algorithms is its natural handling of hanging nodes during mesh refinement. However, several persistent challenges remain, such as developing open-source implementations for polytopal conforming mesh refinement Antonietti et al. (2022) . The FEM literature on a posteriori error analysis of this problem can be found in Gatica et al. (2022a) . On the other hand, a posteriori error estimates for VEM were first introduced in Cangiani et al. (2017) , and the extension to displacement-based deformation models and reaction-diffusion equations using VEM can be found in Wang et al. (2020) ; Munar et al. (2024) . To the best of the authors’ knowledge, the present paper is the first one addressing robust a posteriori error estimators for VEM applied to stress-assisted diffusion problems both in 2D and 3D. We first establish residual-based error estimators in 2D and prove the reliability and efficiency under a small data assumption due to the Banach fixed-point argument. Here, we use classical tools such as quasi-interpolation operators, polynomial projections, a Helmholtz decomposition, and the definition of bubble functions to achieve our purpose. Further, we provide “building blocks” for the 3D case. Specifically, we prove novel quasi-interpolation operator for both Stokes- and edge-like virtual element spaces in 3D, together with a stable Helmholtz decomposition for reaction-diffusion equations (in mixed VEM). The associated estimators are implemented in the VEM++ library Dassi (2025) (available upon request) with an extensible design, enabling the incorporation of various a posteriori error estimators for the virtual element (VE) spaces available in the library. Plan of the paper. The contents of this paper have been organised as follows. The remainder of this section contains preliminary notational conventions and useful functional spaces. Section 2 presents the coupled stress-assisted diffusion model, weak formulation consisting of two coupled perturbed saddle-point problems, unique solvability result, and robust global inf-sup result. The mesh assumptions and VE spaces for the coupled problem along with their properties are provided in Section 3 for the 2D case. Section 4 is devoted to deriving a reliable and efficient error estimator. Next, we extend the VE spaces to the 3D case in Section 5 , and construct two novel quasi-interpolation operators for Stokes and edge spaces. Finally, our theoretical results are illustrated via numerical examples in Section 6 . Recurrent notation. Let D be a domain of \mathbb{R}^{d} ( d=2,3 ), with boundary \partial D . Given a tensor function \boldsymbol{\sigma}:D\to\mathbb{R}^{d\times d} , a vector field \boldsymbol{u}:D\to\mathbb{R}^{d} and a scalar field {p}:D\to\mathbb{R} we set the tensor divergence \mathop{\mathbf{div}}\nolimits\boldsymbol{\sigma}:D\to\mathbb{R}^{d} , the vector gradient \boldsymbol{\nabla}\boldsymbol{u}:D\to\mathbb{R}^{d\times d} , the symmetric gradient \boldsymbol{\varepsilon}(\boldsymbol{u}):D\to\mathbb{R}^{d\times d} , the vector divergence \mathop{\mathrm{div}}\nolimits\boldsymbol{u}:D\to\mathbb{R} , the 3D vector rotational \mathop{\mathbf{curl}}\nolimits\boldsymbol{u}:D\to\mathbb{R}^{3} , the 2D scalar rotational \mathop{\mathrm{rot}}\nolimits\boldsymbol{u}:D\to\mathbb{R} , and the 2D vector rotational \mathop{\mathbf{rot}}\nolimits{p}:D\to\mathbb{R}^{2} , as (\mathop{\mathbf{div}}\nolimits\boldsymbol{\sigma})_{i}:=\sum_{j}\partial_{j}\boldsymbol{\sigma}_{ij} , (\boldsymbol{\nabla}\boldsymbol{u})_{ij}:=\partial_{j}\boldsymbol{u}_{i} , \boldsymbol{\varepsilon}(\boldsymbol{u}):=\frac{1}{2}\left[\boldsymbol{\nabla}\boldsymbol{u}+(\boldsymbol{\nabla}\boldsymbol{u})^{\tt t}\right] , \mathop{\mathrm{div}}\nolimits\boldsymbol{u}:=\sum_{i}\partial_{i}\boldsymbol{u}_{i} , \mathop{\mathbf{curl}}\nolimits\boldsymbol{u}:=(\partial_{2}\boldsymbol{u}_{3}-\partial_{3}\boldsymbol{u}_{2},\partial_{3}\boldsymbol{u}_{1}-\partial_{1}\boldsymbol{u}_{3},\partial_{1}\boldsymbol{u}_{2}-\partial_{2}\boldsymbol{u}_{1}) , \mathop{\mathrm{rot}}\nolimits\boldsymbol{u}:=\partial_{1}\boldsymbol{u}_{2}-\partial_{2}\boldsymbol{u}_{1} , and \mathop{\mathbf{rot}}\nolimits{p}:=(\partial_{2}{p},-\partial_{1}{p})^{\tt t} , respectively. The component-wise inner product for vectors \boldsymbol{u},\boldsymbol{v}\in\mathbb{R}^{d} and matrices \boldsymbol{\sigma},\,\boldsymbol{\tau}\in\mathbb{R}^{d\times d} are defined by \boldsymbol{u}:\boldsymbol{v}:=\sum_{i}\boldsymbol{u}_{i}\boldsymbol{v}_{i} , and \boldsymbol{\sigma}:\boldsymbol{\tau}:=\sum_{i,j}\boldsymbol{\sigma}_{ij}\boldsymbol{\tau}_{ij} . For s\geq 0 , we denote the Sobolev space of scalar functions with domain D as \mathrm{H}^{s}(D) , and their vector and tensor counterparts as \mathbf{H}^{s}(D) and \mathbb{H}^{s}(D) , respectively. The norm in \mathrm{H}^{s}(D) is denoted \lVert\cdot\rVert_{s,D} and the corresponding semi-norm |\cdot|_{s,D} . We also use the convention \mathrm{H}^{0}(D):=\mathrm{L}^{2}(D) and let (\cdot,\cdot)_{\mathrm{D}} be the inner product for in \mathrm{L}^{2}(D) (similarly for the vector and tensor counterparts). The space \text{H}^{\frac{1}{2}}(\partial D) contains traces of functions of \text{H}^{1}(D) , \text{H}^{-\frac{1}{2}}(\partial D) denotes its dual, and \langle\cdot,\cdot\rangle_{\partial D} stands for the duality pairing between them. The Hilbert space \mathbf{H}(\mathop{\mathrm{div}}\nolimits,D) of vectors in \mathbf{L}^{2}(D) with divergence in \mathrm{L}^{2}(D) equipped with the norm \lVert\cdot\rVert^{2}_{\mathop{\mathrm{div}}\nolimits,D}:=\lVert\cdot\rVert_{0,D}^{2}+\lVert\mathop{\mathrm{div}}\nolimits\cdot\rVert^{2}_{0,D} . Similarly, the Hilbert spaces \mathbf{H}(\mathop{\mathrm{rot}}\nolimits,D) and \mathbf{H}(\mathop{\mathbf{curl}}\nolimits,D) are equipped with the norms \lVert\cdot\rVert^{2}_{\mathop{\mathrm{rot}}\nolimits,D}:=\lVert\cdot\rVert_{0,D}^{2}+\lVert\mathop{\mathrm{rot}}\nolimits\cdot\rVert^{2}_{0,D} and \lVert\cdot\rVert^{2}_{\mathop{\mathbf{curl}}\nolimits,D}:=\lVert\cdot\rVert_{0,D}^{2}+\lVert\mathop{\mathbf{curl}}\nolimits\cdot\rVert^{2}_{0,D} . The outward unit normal vector and unit tangential vector to \partial D are denoted respectively by \boldsymbol{n} and \boldsymbol{t} . Throughout this paper, we shall use the letter C to denote a generic positive constant independent of the mesh size h and physical constants, which might stand for different values at its different occurrences. Moreover, given any positive expressions X and Y , the notation X\,\lesssim\,Y means that X\,\leq\,C\,Y (and similarly for X\gtrsim Y ). 2 The stress-assisted diffusion problem This section recalls from Khot et al. (2025) the weak formulation and its well-posedness analysis based on the Babuška–Brezzi–Braess theory and a fixed-point argument. Model problem. Let \Omega be a polytopal (polygonal in 2D and polyhedral in 3D) bounded domain with boundary \Gamma=\Gamma_{\mathrm{D}}\cup\Gamma_{\mathrm{N}} , such that \Gamma_{\mathrm{D}}\cap\Gamma_{\mathrm{N}}=\emptyset , and consider the following coupled PDE with mixed