A gradient-enhanced approach for stable finite element approximations of reaction-convection-diffusion problems Journal: arXiv
Soheil Firooz, B. Daya Reddy, Paul Steinmann · 2025
Read A gradient-enhanced approach for stable finite element approximations of reaction-convection-diffusion problems Journal: arXiv by Soheil Firooz, B. Daya Reddy, Paul Steinmann (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-2506-01873.txt.
Introduction
Soheil Firooz, Daya Reddy, and Paul Steinmann present a steadier way to solve a family of flow equations on a computer in a 2025 paper. The equations mix three effects: spread, drift with a current, and local reaction. Standard grid methods can wobble or fail when drift or reaction grows strong.
Their fix adds a helper field that tracks the slope of the main field, borrowing an idea from the theory of materials with inner structure. The two fields are solved together. The authors prove the paired problem is well posed and give an error bound for the grid version. Tests in one and two dimensions show sharp, stable answers across a wide range of drift and reaction strengths, even on problems where plain methods struggle.
Catalog record
| CREATOR | Soheil Firooz, B. Daya Reddy, Paul Steinmann |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 41 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-01873
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A gradient-enhanced approach for stable finite element approximations of reaction-convection-diffusion problems Journal: arXiv Soheil Firooz, B. Daya Reddy, Paul Steinmann Abstract We develop a micromorphic-based approach for finite element stabilization of reaction-convection-diffusion equations, by gradient enhancement of the field of interest via introducing an auxiliary variable. The well-posedness of the coupled-field approach is established, together with an error estimate. Through a set of 1D and 2D numerical examples the high accuracy and enhanced stability of the approach in approximating solutions associated with complex problems is demonstrated, for situations of varying reactivity and convection. 1 Introduction Reaction-convection-diffusion equations play a central role in physical and mathematical modeling of transport and reaction phenomena across a wide range of disciplines from heat transfer and fluid flow to mass transport in environmental and chemical/biological systems. Diffusion commonly represents the spreading of species within the domain due to concentration gradients. Convection accounts for the transport of species driven by the movement of the fluid medium. Reaction refers to the chemical, biological or physical interactions through which species are produced or consumed during the process. In reaction-convection-diffusion problems, if the diffusive effects are negligible compared to convective or reactive effects, the problem is referred to as convection-dominated or reaction-dominated, respectively [ 1 , 2 , 3 ] . A well-known challenge that arises in such problems is that the solutions develop sharp gradients and steep internal or boundary layers. In such scenarios, standard Galerkin finite element methods (FEM) fail to fully resolve the localized features leading to nonphysical, oscillatory solutions with spurious overshoots and undershoots [ 4 , 5 ] . Over the past few decades, a variety of stabilization techniques have been developed to address the instabilities associated particularly with convection-diffusion equations. A classical remedy to eliminate such oscillations is to introduce artificial diffusion in the equations. However, excessive artificial diffusion can smear the solution which affects the solution accuracy. One of the most well-established stabilization techniques to overcome instabilities in convection-dominated problems is the streamline-upwind Petrov–Galerkin (SUPG) introduced by Brooks and Hughes [ 5 , 6 ] . The SUPG method adds a perturbation term aligned with the flow/streamline direction to the test functions which effectively introduces a minimal targeted diffusion along streamlines. The resulting formulation maintains consistency and greatly reduces global oscillations for high Peclet number flows. In narrow regions such as sharp internal layers or boundary layers, the SUPG solution may still exhibit small overshoot/undershoot profiles because the stabilization is primarily in the flow direction and lacks dissipation in the crosswind direction [ 7 ] . Another approach developed to address the numerical instabilities that arise in convection-dominated and reaction-dominated problems is the Galerkin/Least-Squares (GLS) method [ 8 , 9 ] . The GLS method enhances the standard Galerkin formulation via incorporating additional least-squares terms involving the residual of the governing equations [ 10 ] . Unlike upwinding or artificial diffusion methods, the GLS method retains the variational consistency of the formulation. A critical and nontrivial aspect of the GLS method is the choice of the stabilization parameter which typically depends on local mesh size, the magnitude of convection and reaction coefficients, and sometimes solution features such as local gradients. Moreover, due to its global formulation which is advantageous for convergence analysis, GLS fails to offer fine local control over the stabilization effects. For such cases, other methods such as SUPG or the discontinuous Galerkin finite element method (DGFEM) [ 11 , 12 ] may provide better adaptability and sharper resolution. The so-called unusual stabilization method [ 13 , 14 ] maintains the original Galerkin framework but augments it with stabilizing terms that are not necessarily derived from a least-squares or residual-based interpretation. A key feature of this method is its variational consistency where the added stabilization terms vanish when the exact solution is substituted into the formulation thereby preserving convergence and error properties. Many studies have been carried out to improve the stability and accuracy of the unusual stabilization method, see [ 15 , 16 , 17 ] among others. Several other techniques have been developed to deal with instabilities associated with convection-dominated and reaction-dominated problems such as the variational multiscale (VMS) method [ 18 , 19 , 20 , 21 ] , local projection stabilization (LPS) method [ 22 , 23 ] , the symmetric stabilization method [ 24 , 25 , 26 ] , the residual bubble free (RFB) method [ 27 , 28 ] , high-resolution Petrov–Galerkin (HRPG) method [ 29 , 30 ] , weak Galerkin FEM [ 31 ] , and variational subgrid scale method [ 32 ] . See [ 33 , 34 , 35 ] for comparative studies on some stabilizing finite element methods. Despite these advances, challenges still remain in achieving stability without compromising accuracy. Some methods require tuning of problem-dependent parameters such as the SUPG weighting parameter or shock-capturing coefficients. Such limitations motivate the development of improved stabilization strategies that selectively target oscillations and internal layers without smearing the sharp gradients. In a recent contribution [ 36 ] , a micromorphic-based artificial diffusion (MMAD) method has been proposed for stable and accurate finite element approximation of convection-diffusion problems. The method is, in principle, a gradient-enhanced extension of the mean-zero artificial diffusion method presented in [ 37 ] , and which is equivalent to a projected artificial diffusion approach. Motivated by the micromorphic approach [ 38 , 39 , 40 ] , the modification entails the introduction of a micromorphic-type variable and enhancement in terms of the variable and its gradient. A number of numerical examples demonstrate the excellent performance of the approach, in the convection-dominated range and for pure advection problems. In this work the formulation in [ 36 ] is extended to treat reaction-convection-diffusion problems. Of particular interest are situations in which reaction or convection dominate. The formulation is shown to be well-posed and convergent for equal-order (linear) approximation of the original variable and that introduced for the enhancement. A number of numerical examples are presented, for situations of reaction- or advection- dominated flows. These illustrate the accurate and stable performance of the method, with oscillations or wiggles largely suppressed in regions of steep gradients or localized layers. The structure of the rest of this work is as follows. Section 2 introduces the problem and presents the governing equations for reaction-convection-diffusion problems. The micromorphic-based artificial diffusion method together with details of the well-posedness and convergence of the method are the subject of Section 3. Section 4 illustrates the performance of the MMAD method through a set of numerical examples. Section 5 summarizes the the key contributions of the work. 2 Reaction-convection-diffusion problems Let \mathcal{B} be a bounded domain with the boundary \partial\mathcal{B} and \boldsymbol{n} be the outward unit normal to \partial\mathcal{B} . For a scalar quantity \varphi , the general dimensionless form of the steady-state reaction-convection-diffusion equation reads \displaystyle\text{Da}\,\varphi+\boldsymbol{u}\cdot\nabla\varphi-\displaystyle\frac{1}{\text{Pe}}\Delta\varphi=F\quad\text{in }\mathcal{B}\,, (1) \displaystyle\text{subject to:}\left\{\begin{aligned} &\varphi=\varphi_{\text{p}}\quad&&\text{on }\partial\mathcal{B}_{\text{D}}\,,\\ &\nabla\varphi\cdot\boldsymbol{n}=\frac{\partial\varphi}{\partial\boldsymbol{n}}=t_{\text{p}}\quad&&\text{on }\partial\mathcal{B}_{\text{N}}\,,\end{aligned}\right. \displaystyle\text{Pe}=\frac{UL}{D}\,,\qquad\text{Da}=\frac{BL}{U}\,,\qquad\boldsymbol{u}=\displaystyle\frac{\boldsymbol{U}}{U}\,, where Pe is the Peclet number, Da the Damköhler number, \boldsymbol{u} the dimensionless velocity vector, \boldsymbol{U} the divergence-free velocity, D the diffusion coefficient, B the reaction coefficient, L a characteristic length of the spatial domain, U a characteristic velocity, \boldsymbol{n} the unit vector normal to the boundary \partial\mathcal{B} , \varphi_{\text{p}} the prescribed value