Robust virtual element methods for 3D stress-assisted diffusion problems
Andrés E. Rubiano · 2025
Read Robust virtual element methods for 3D stress-assisted diffusion problems by Andrés E. Rubiano (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-2502-01851.txt.
Introduction
Andrés Rubiano, writing solo, develops a numerical method for a coupled physics problem in a 2025 paper. The problem joins diffusion, the spread of a substance, with mechanical stress in a solid. Stress speeds or slows the spread, and the spread changes the stress. Solving the pair in three dimensions is demanding. Rubiano works inside the virtual element method, a scheme that handles odd shaped mesh cells with ease.
He builds the needed function spaces with norms weighted by the problem's parameters. A perturbed saddle point theory and a fixed point argument carry the proof that the full nonlinear system is well behaved. Custom projection, interpolation, and stabilization pieces make the discrete scheme sound. Computer tests show the method is accurate and performs well. The author frames the work as a first step for this method class on such problems.
Catalog record
| CREATOR | Andrés E. Rubiano |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 30 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-01851
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Robust virtual element methods for 3D stress-assisted diffusion problems Andrés E. Rubiano Abstract This paper presents an initial exploration of stress-assisted diffusion problems in three dimensions within the framework of the virtual element method (VEM). Hilbert spaces enriched with parameter-weighted norms, the extended Babu s ˇ \check{\text{s}} ka–Brezzi–Braess theory for perturbed saddle-point problems, and Banach fixed-point theory play a crucial role in performing a robust analysis of the fully coupled non-linear system. The proposed virtual element formulations are provided with appropriate projection, interpolation, and stabilisation operators that ensures the well-posedness of the discrete problem. Numerical simulations are conducted to show the accuracy, performance, and applicability of the method. 2020 MSC codes: 65N99; 65Y99. Keywords: virtual element methods, stress-assisted diffusion, diffusion-induced stress, perturbed saddle-point problems. 1 Introduction Scope. Stress-assisted diffusion appears broadly in applications like Lithium-ion battery cells, silicon rubber and hydrogen diffusion, polymer-based coatings, semiconductor fabrication, oxidation of silicon nanostructures, and enhancing of conductivity properties in soft living tissue. In these processes, the stress of the material involved affects the patterns of diffusion. This paper addresses one of the simplest models for such an interaction, incorporating stress effects inside the diffusion coefficient, and recovering Fick’s law in the absence of these phenomena (see, e.g., [ 4 ] ). Moreover, we extend the analysis done in [ 8 ] pointing out the definition of VEM spaces in 3D and the polynomial, interpolation, and stabilisation operators used in this framework. Outline. The content of the paper is organised as follows: The remainder of this section presents the strong form of the coupled fully coupled stress-assisted diffusion problem. Section 2 is devoted to presenting the weak formulation and well-posedness analysis using perturbed saddle-point theory and fixed-point arguments. In Section 3 we define the discrete problem, show the existence and uniqueness of discrete solution, and outline the a priori error analysis. Finally, in Section 4 we provide numerical results that illustrate the theoretical results presented, and we present as an application of the model the lithiation of a perforated cylindrical anode particle. Model statement. Let us consider the following nonlinear formulation for stress-assisted diffusion of a solute that interacts with an elastic material \displaystyle-\mathop{\mathbf{div}}\nolimits(2\mu\boldsymbol{\varepsilon}(\boldsymbol{u})-p\mathbb{I})=\boldsymbol{f} \displaystyle\hskip 10.00002pt\text{in $\Omega$},\hskip 10.00002pt\boldsymbol{u}=\mathbf{0}\hskip 10.00002pt\text{on $\Gamma_{D}$}, (1.1a) \displaystyle p=-\lambda\mathop{\mathrm{div}}\nolimits\boldsymbol{u}\,+\ell(\varphi) \displaystyle\hskip 10.00002pt\text{in $\Omega$},\hskip 10.00002pt(2\mu\boldsymbol{\varepsilon}(\boldsymbol{u})-p\mathbb{I})\boldsymbol{n}=\boldsymbol{0}\hskip 10.00002pt\text{on $\Gamma_{N}$}, (1.1b) \displaystyle\boldsymbol{\zeta}=\mathbb{M}(\boldsymbol{\varepsilon}(\boldsymbol{u}),p)\nabla\varphi \displaystyle\hskip 10.00002pt\text{in $\Omega$},\hskip 10.00002pt\varphi=\varphi_{D}\hskip 10.00002pt\text{on $\Gamma_{D}$}, (1.1c) \displaystyle\theta\varphi-\mathop{\mathrm{div}}\nolimits(\boldsymbol{\zeta})=g \displaystyle\hskip 10.00002pt\text{in $\Omega$},\hskip 10.00002pt\boldsymbol{\zeta}\cdot\boldsymbol{n}=0\hskip 10.00002pt\text{on $\Gamma_{N}$}. (1.1d) Equations ( 1.1a ) and ( 1.1b ) constitute a Herrmann-like mixed formulation for linear elasticity. The coupling between the displacement \boldsymbol{u} of the material, the pressure p , and the solute concentration \varphi follows Hooke’s law to express the Cauchy stress tensor as 2\mu\boldsymbol{\varepsilon}(\boldsymbol{u})-p\mathbb{I} , \mu and \lambda are the Lamé parameters, \ell is the active stress coefficient, and \boldsymbol{f} is a vector of external body loads. On the other hand, equations ( 1.1d ) and ( 1.1c ) correspond to the reaction-diffusion equation written in mixed form where \boldsymbol{\zeta} is the diffusive flux, g is a given net volumetric source of solute, \theta is a positive model parameter, and \mathbb{M}(\boldsymbol{\varepsilon}(\boldsymbol{u}),p) is the stress-assisted diffusion coefficient. We adopt mixed loading boundary conditions for the coupled problem: the structure is clamped and has a given concentration on \Gamma_{D} , where the boundary subset \Gamma_{D}\subset\partial\Omega is of positive surface measure. In addition, a given traction and zero solute flux are prescribed on \Gamma_{N}:=\partial\Omega\setminus\Gamma_{D} . The nonlinear terms. We assume that \mathbb{M}(\cdot,\cdot) is symmetric, positive semi-definite and uniformly bounded in \mathbb{L}^{\infty}(\Omega) , likewise for \mathbb{M}^{-1}(\cdot,\cdot) . More explicitly, for all \boldsymbol{w}\in\mathbf{H}^{1}(\Omega),r\in\text{L}^{2}(\Omega)\text{ and }\boldsymbol{x},\boldsymbol{y}\in\mathbb{R}^{3} there exists M\in\mathbb{R} such that 0<M^{-1}\leq M with M^{-1}\boldsymbol{x}\cdot\boldsymbol{x}\leq\boldsymbol{x}\cdot[\mathbb{M}^{-1}(\boldsymbol{\varepsilon}(\boldsymbol{w}),r)\boldsymbol{x}] , and \boldsymbol{y}\cdot[\mathbb{M}^{-1}(\boldsymbol{\varepsilon}(\boldsymbol{w}),r)\boldsymbol{x}]\leq M\boldsymbol{x}\cdot\boldsymbol{y} . In addition, we assume that \ell:\text{L}^{2}(\Omega)\to\text{L}^{2}(\Omega) and satisfies \lVert\ell(\vartheta)\rVert_{0,\Omega}\lesssim\lVert\vartheta\rVert_{0,\Omega} for all \vartheta\in\text{L}^{2}(\Omega) . Moreover, we assume that \mathbb{M}^{-1}(\cdot,\cdot) and \ell(\cdot) are Lipschitz continuous with Lipschitz constants L_{\mathbb{M}} and L_{\ell} . Examples of these terms can be found in [ 7 ] for the stress-assisted diffusion and [ 10 ] for the active stress. 2 Weak formulation In view of the boundary conditions, we define the Hilbert spaces \displaystyle\mathbf{H}^{1}_{D}(\Omega):=\{\boldsymbol{v}\in\mathbf{H}^{1}(\Omega):\boldsymbol{v}=\boldsymbol{0}\,\text{on }\Gamma_{D}\}, \displaystyle\mathbf{H}_{N}(\mathop{\mathrm{div}}\nolimits,\Omega):=\{\boldsymbol{\xi}\in\mathbf{H}(\mathop{\mathrm{div}}\nolimits,\Omega):\boldsymbol{\xi}\cdot\boldsymbol{n}=0\,\text{on }\Gamma_{N}\}, with the boundary assignment understood in the sense of traces, and consider the following weak formulation: for given \boldsymbol{f}\in\mathbf{L}^{2}(\Omega) , g\in\text{L}^{2}(\Omega) , and \varphi_{D}\in H^{1/2}(\Gamma_{D}) , find (\boldsymbol{u},p,\boldsymbol{\zeta},\varphi)\in\mathbf{H}_{D}^{1}(\Omega)\times\text{L}^{2}(\Omega)\times\mathbf{H}_{N}(\mathop{\mathrm{div}}\nolimits,\Omega)\times\text{L}^{2}(\Omega) such that \displaystyle 2\mu\int_{\Omega}\boldsymbol{\varepsilon}(\boldsymbol{u}):\boldsymbol{\varepsilon}(\boldsymbol{v})-\int_{\Omega}p\mathop{\mathrm{div}}\nolimits\boldsymbol{v}=\int_{\Omega}\boldsymbol{f}\cdot\boldsymbol{v},\;\forall\boldsymbol{v}\in\mathbf{H}_{D}^{1}(\Omega) (2.1a) \displaystyle-\int_{\Omega}q\mathop{\mathrm{div}}\nolimits\boldsymbol{u}-\lambda^{-1}\int_{\Omega}pq=\lambda^{-1}\int_{\Omega}\ell(\varphi)q,\;\forall q\in\text{L}^{2}(\Omega) (2.1b) \displaystyle\int_{\Omega}\mathbb{M}(\boldsymbol{\varepsilon}(\boldsymbol{u}),p)^{-1}\boldsymbol{\zeta}\cdot\boldsymbol{\xi}+\int_{\Omega}\varphi\mathop{\mathrm{div}}\nolimits\boldsymbol{\xi}=\langle\varphi_{D},\boldsymbol{\xi}\cdot\boldsymbol{n}\rangle_{\Gamma_{D}},\;\forall\boldsymbol{\xi}\in\mathbf{H}_{N}(\mathop{\mathrm{div}}\nolimits,\Omega) (2.1c) \displaystyle\int_{\Omega}\psi\mathop{\mathrm{div}}\nolimits\boldsymbol{\zeta}-\theta\int_{\Omega}\varphi\psi=-\int_{\Omega}g\psi,\;\forall\psi\in\text{L}^{2}(\Omega). (2.1d) Unique solvability with parameter-weighted norms. Let us adopt the following notation for the functional spaces for displacement and total volumetric stress \mathbf{V}_{1}:=\mathbf{H}_{D}^{1}(\Omega) and Q_{1}=Q_{b_{1}}:=\text{L}^{2}(\Omega) , equipped with the scaled norms and semi-norms given by \displaystyle\lVert\boldsymbol{u}\rVert_{\mathbf{V}_{1}}^{2}:=2\mu\lVert\boldsymbol{\varepsilon}(\boldsymbol{u})\rVert_{0,\Omega}^{2},\hskip 10.00002pt\lVert p\rVert_{Q_{1}}^{2}:=\left((2\mu)^{-1}+\lambda^{-1}\right)\lVert p\rVert_{0,\Omega}^{2}, \displaystyle|\boldsymbol{v}|_{1,\mathbf{V}_{1}}^{2}:=2\mu|\boldsymbol{v}|_{1,\Omega}^{2},\hskip 10.00002pt|q|_{1,Q_{b_{1}}}^{2}:=(2\mu)^{-1}|q|_{1,\Omega}^{2}. On the other hand, let us denote the functional spaces for diffusive flux and concentration as \mathbf{V}_{2}=\mathbf{H}_{N}(\mathop{\mathrm{div}}\nolimits,\Omega) and Q_{2}=Q_{b_{2}}:=\text{L}^{2}(\Omega) , furnished with the following norms and semi-norms as \lVert\boldsymbol{\zeta}\rVert_{\mathbb{M},\Omega}^{2}:=\int_{\Omega}\mathbb{M}(\boldsymbol{\varepsilon}(\boldsymbol{u}),p)^{-1}\boldsymbol{\zeta}\cdot\boldsymbol{\zeta} , \displaystyle\lVert\boldsymbol{\zeta}\rVert_{\mathbf{V}_{2}}^{2}:=\lVert\boldsymbol{\zeta}\rVert_{\mathbb{M},\Omega}^{2}+M\lVert\mathop{\mathrm{div}}\nolimits\boldsymbol{\zeta}\rVert_{0,\Omega}^{2},\hskip 10.00002pt\lVert\varphi\rVert_{Q_{2}}^{2}:=\left(M^{-1}+\theta\right)\lVert\varphi\rVert_{0,\Omega}^{2}, \displaystyle|\boldsymbol{\xi}|_{1,\mathbf{V}_{2}}^{2}:=M|\boldsymbol{\xi}|_{1,\Omega}^{2},\hskip 10.00002pt|\psi|_{1,Q_{b_{2}}}^{2}:=M^{-1}|\psi|_{1,\Omega}^{2}. The proposed spaces in conjunction with the extended Babuška–Brezzi–Braess theory for perturbed saddle-point problems and a fixed-point argument show the robust unique solvability of our weak formulation presented in Section 2 . We finish this section by stating the continuous dependence on data (see [ 8 , Section 3] for details). Theorem 1 . Let W=\{w\in Q_{2}\colon\lVert w\rVert_{Q_{2}}\leq C_{2}\sqrt{M}\lVert\varphi_{D}\rVert_{1/2,\Gamma_{D}}+\lVert g\rVert_{0,\Omega})\} . Under the assumptions over the non-linear terms, suppose further that 1\leq\lambda , 1\leq\mu , \theta\leq M^{-1} , and C_{1}L_{\ell}\sqrt{2\mu}M^{2}C_{2}^{2}L_{\mathbb{M}}(\lVert\varphi_{D}\rVert_{1/2,\Gamma_{D}}+\lVert g\rVert_{0,\Omega})<1. Then, for \varphi\in W there is an unique solution (\boldsymbol{u},p,\boldsymbol{\zeta},\varphi)\in\mathbf{V}_{1}\times Q_{1}\times\mathbf{V}_{2}\times Q_{2} of ( 2.1 ) such that \displaystyle\lVert(\boldsymbol{u},p)\rVert_{\mathbf{V}_{1}\times Q_{1}} \displaystyle\leq C_{1}\left(\lVert F_{1}\rVert_{\mathbf{V}^{\prime}_{1}}+\lVert G^{\varphi}_{1}\rVert_{Q^{\prime}_{1}}\right), (2.2a) \displaystyle\lVert(\boldsymbol{\zeta},\varphi)\rVert_{\mathbf{V}_{2}\times Q_{2}} \displaystyle\leq C_{2}\left(\lVert F_{2}\rVert_{\mathbf{V}^{\prime}_{2}}+\lVert G_{2}\rVert_{Q^{\prime}_{2}}\right), (2.2b) where the constants C_{1} and C_{2} do not depend on the physical parameters. 3 Virtual element discretisation Mesh assumptions. Let \mathcal{T}^{h} be a decomposition of \Omega into polyhedral elements P with diameter h_{P} , let \mathcal{F}^{h} be the set of faces f with length h_{f} and \mathcal{E}^{h} be the set of edges e of \mathcal{T}^{h} with length h_{e} . The following mesh assumptions are considered throughout this paper. We assume that there exists a universal constant \rho>0 such that ( M1) Each polyhedral element P is star-shaped with respect to a ball of radius \geq \rho h_{P} , ( M2) Every face f of P is of diameter h_{f} and is star-shaped with respect to a disk of radius \geq \rho h_{P} , ( M3) Every edge e of P has length \geq \rho h_{P} . Polynomial spaces. Given an integer k\geq 0 the space of polynomials of degree \leq k on P is denoted by \mathcal{P}_{k}(P) (resp. for faces f ). The space of the gradients of polynomials of grade \leq k+1 on P is denoted as \mathcal{G}_{k}(P):=\nabla(\mathcal{P}_{k+1}(P)) with standard notation \mathcal{P}_{-1}(P)=\{0\} for k=-1 . The space \mathcal{G}_{k}^{\oplus}(P) denotes the complement of the space \mathcal{G}_{k}(P) in the vector polynomial space (\mathcal{P}_{k}(P))^{3} , that is, (\mathcal{P}_{k}(P))^{3}=\mathcal{G}_{k}(P)\oplus\mathcal{G}_{k}^{\oplus}(P) . In particular, following [ 2 ] , we set \mathcal{G}_{k}^{\oplus}(P):=\boldsymbol{x}\wedge(\mathcal{P}_{k-1}(P))^{3} with \boldsymbol{x}=(x_{1},x_{2},x_{3})^{\tt t} . Likewise, the space that defines the rotational of polynomials with degree \leq k+1 is denoted as \mathcal{R}_{k}(P):=\mathop{\mathbf{curl}}\nolimits(\mathcal{P}_{k+1}(P)) where the associated complement space \mathcal{R}_{k}^{\oplus}(P) fulfills the property (\mathcal{P}_{k}(P))^{3}=\mathcal{R}_{k}(P)\oplus\mathcal{R}_{k}^{\oplus}(P) with \mathcal{R}_{k}^{\oplus}(P)=\boldsymbol{x}\mathcal{P}_{k-1}(P) . Let \boldsymbol{x}_{P}=(x_{1,P},x_{2,P},x_{3,P})^{\tt t} denote the barycentre of P and let \mathcal{M}_{k}(P) be the set of scaled monomials \mathcal{M}_{k}(P):=\left\{\left(\frac{\boldsymbol{x}-\boldsymbol{x}_{P}}{h_{P}}\right)^{\boldsymbol{\alpha}},0\leq|\boldsymbol{\alpha}|\leq k\right\}, where \boldsymbol{\alpha}=(\alpha_{1},\alpha_{2},\alpha_{3})^{\tt t} is a non-negative multi-index with |\boldsymbol{\alpha}|=\alpha_{1}+\alpha_{2}+\alpha_{3} and \boldsymbol{x}^{\boldsymbol{\alpha}}=x_{1}^{\alpha_{1}}x_{2}^{\alpha_{2}}x_{3}^{\alpha_{3}} . In particular, we can take the basis of \mathcal{G}_{k}(P) and \mathcal{G}_{k}^{\oplus}(P) as \mathcal{M}_{k}^{\nabla}(P):=\nabla\mathcal{M}_{k+1}(P)\setminus\{\mathbf{0}\} and \mathcal{M}_{k}^{\oplus}(P):=\boldsymbol{x}_{P}\wedge(\mathcal{M}_{k-1}(P))^{3} , where (\mathcal{M}_{k}(P))^{3}=\mathcal{M}_{k}^{\nabla}(P)\oplus\mathcal{M}_{k}^{\oplus}(P) holds. In addition, we introduce