Adaptive Crouzeix-Raviart finite elements for the first eigenpair of p p -Laplacian

Guanglian Li, Yueqi Wang, Yifeng Xu · 2025

Read Adaptive Crouzeix-Raviart finite elements for the first eigenpair of p p -Laplacian by Guanglian Li, Yueqi Wang, Yifeng Xu (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-2508-02077.txt.

Introduction

Guanglian Li, Yueqi Wang, and Yifeng Xu build an adaptive mesh method in 2025 for the first eigenpair of the p-Laplacian. Such eigenpairs set base tones for a class of bent diffusion problems. Their method uses Crouzeix-Raviart elements that can break at edges, with mesh steps that refine where error runs high.

The team proves error marks shrink to zero and, from a fine start mesh, values near the true first tone. A tightness trait for such elements across changing meshes is the key new step. Number tests show the edge of the adaptive plan.

Catalog record

CREATORGuanglian Li, Yueqi Wang, Yifeng Xu
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE75 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-02077

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Adaptive Crouzeix-Raviart finite elements for the first eigenpair of p p -Laplacian Guanglian Li, Yueqi Wang, Yifeng Xu Abstract In this paper, we propose and analyze an adaptive Crouzeix-Raviart finite element method for computing the first Dirichlet eigenpair of the p p -Laplacian problem. We prove that the sequence of error estimators produced by the adaptive algorithm has a vanishing limit and that, starting from a fine initial mesh, the relevant sequence of approximate eigenvalues converges to the first eigenvalue and the distance in a mesh-dependent broken norm between discrete eigenfunctions and the set composed of relevant continuous eigenfunctions also tends to zero. The analysis hinges on establishing a compactness property for Crouzeix-Raviart finite elements over a sequence of adaptively generated meshes, which represents key theoretical challenges and novelties. We present numerical results to illustrate the advantage of the proposed algorithm. keywords p -Laplacian, first eigenvalue, Crouzeix-Raviart finite element, adaptive finite element method, convergence AMS 65N12, 65N25, 65N30, 65N50, 35P30 1 Introduction Let \Omega be a bounded polyhedral connected domain in \mathbb{R}^{d} ( d=2,3 ), and fix p\in(1,\infty) . Consider the following Dirichlet eigenvalue problem of the p -Laplacian operator: \left\{\begin{aligned} -\mathrm{div}(|\bm{\nabla}u|^{p-2}\bm{\nabla}u)&=\lambda|u|^{p-2}u&&\text{in}~\Omega,\\ u&=0&&\text{on}~\partial\Omega.\end{aligned}\right. (1.1) In the engineering literature, the p -Laplacian operator is often used to describe non-Newtonian fluids, turbulent flow, flow through porous media and power-law materials etc. The weak formulation of ( 1.1 ) is to find (\lambda,u)\in\mathbb{R}\times V:=W_{0}^{1,p}(\Omega) such that \int_{\Omega}|\bm{\nabla}u|^{p-2}\bm{\nabla}u\cdot\bm{\nabla}v\mathrm{d}x=\lambda\int_{\Omega}|u|^{p-2}uv\mathrm{d}x,\quad\forall v\in V. (1.2) The existing theory [ 36 , 42 , 44 ] states that there is a nondecreasing sequence of positive eigenvalues \{\lambda_{n}\}_{n\geq 1} tending to \infty . Moreover, the first eigenvalue \lambda_{1} is simple and isolated [ 44 ] , and can be characterized as the minimum of the Rayleigh quotient: \lambda_{1}=\inf_{v\in V\setminus\{0\}}\mathcal{J}(v):=\int_{\Omega}|\bm{\nabla}v|^{p}\mathrm{d}x/\int_{\Omega}|v|^{p}\mathrm{d}x. (1.3) By the direct method in calculus of variation [ 5 , 44 ] , there exists a minimizer to problem ( 1.3 ) among all L^{p} -normalized functions. This implies an L^{p} -normalized eigenfunction set with respect to \lambda_{1} : E_{\lambda_{1}}:=\left\{u\in V~\Big|~\int_{\Omega}|\bm{\nabla}u|^{p-2}\bm{\nabla}u\cdot\bm{\nabla}v\mathrm{d}x=\lambda_{1}\int_{\Omega}|u|^{p-2}uv\mathrm{d}x,~\forall v\in V,~\|u\|_{L^{p}(\Omega)}=1\right\}. The numerical treatment of problem ( 1.1 ) is challenging, due to the degeneracy of the differential operator and the possible presence of geometric singularities of the domain \Omega , e.g., reentrant corners. One effective strategy to resolve the challenges is to employ a suitable adaptive technique, e.g., adaptive finite element method (AFEM). The AFEM can achieve the desired accuracy with minimal degrees of freedom, making it a popular tool in scientific and engineering computing. The standard AFEM typically consists of the following four modules in every loop: \textrm{SOLVE}\to\textrm{ESTIMATE}\to\textrm{MARK}\to\textrm{REFINE}. (1.4) Since its first introduction in [ 4 ] , there has been significant progress in mathematical analysis of AFEM. The use of a posteriori estimation, a key component in the ESTIMATE module, is well-documented [ 2 , 57 ] . Over the past three decades, there has been substantial progress on the convergence and complexity of AFEM; see the reviews [ 9 , 17 , 48 ] and the references therein. In this work, we develop a novel adaptive nonconforming linear FEM, using the Crouzeix-Raviart (C-R) FEM [ 26 ] for computing the first eigenpair of ( 1.2 ) (cf. Algorithm 1 ), and establish the convergence of the sequence of discrete eigenpairs \{(\mu_{k},u_{k})\}_{k\geq 0} (cf. Theorems 9 and 11 ). We derive a computable quantity inspired by [ 49 ] that measures the discontinuity of the discrete eigenfunction u_{k} , and prove the weak convergence of the adaptively generated sequence \{u_{k}\}_{k\geq 0} up to a subsequence (see Lemma 2 ) if the computable quantity tends to zero as k\to\infty . Moreover, the L^{p} weak convergence in Lemma 2 can be lifted to a strong one (see Lemma 4 ). This is the discrete compactness of C-R elements over a sequence of adaptively generated meshes, thereby generalizing the result in [ 54 ] (see also [ 51 , Chapter 10.3] ) in the case of uniform mesh refinements, and the compactness property of adaptive C-R elements in the L^{2} -version [ 39 ] . Moreover, we prove that a residual functional associated with the discrete eigenpair (\mu_{k},u_{k}) vanishes in the limit as a computable quantity related to ( 2.4 ) tends to zero (see Lemma 6 ). This result relies on an interpolation operator from W_{0}^{1,p}(\Omega) to the C-R FE space [ 26 , 49 ] , which also provides a guaranteed lower bound for \lambda_{1} in Theorem 5 . Then we establish the existence of a strong limit pair (\mu,w)\in\mathbb{R}\times W_{0}^{1,p}(\Omega) of a subsequence \{(\lambda_{k_{j}},u_{k_{j}})\}_{j\geq 0} , provided that two relevant computable quantities converge to zero (see Theorem 8 ). With an appropriate assumption on the initial mesh, this gives the desired convergence of Algorithm 1 (see Theorem 9 ). Hence, the convergence analysis motivates the two estimators in the ESTIMATE module of Algorithm 1 . Furthermore, by leveraging the argument from adaptive conforming FEMs [ 52 ] , the sufficient condition in Theorem 8 is proved to hold (see Lemma 10 and Theorem 11 ). Now we discuss related works in existing studies. For linear or semi-linear (with nonlinearity in low-order terms) eigenvalue problems, there is a large body of literature on the linear Laplacian, diffusion, and bi-Laplacian [ 8 , 10 , 18 , 19 , 20 , 21 , 23 , 24 , 25 , 28 , 31 , 33 , 32 , 34 , 35 , 37 ] . Moreover, several works [ 6 , 22 , 30 , 45 , 46 , 56 ] have investigated a posteriori error estimation for p -Laplacian boundary value problems. Despite these impressive advances, the AFEM approximation of the p -Laplacian eigenvalue problem ( 1.2 ) remains very limited. Compared with boundary value problems, the eigenvalue problem itself is nonlinear even for the linear Laplacian, and hence ( 1.2 ) associated with a nonlinear operator is naturally more computationally involved and theoretically challenging. The only existing work on problem ( 1.2 ) is [ 43 ] , which investigated the AFEM using conforming linear elements for the first eigenpair of ( 1.2 ). For a sufficiently fine initial mesh, Li et al. [ 43 ] proved that the sequence of the first eigenvalues generated by the adaptive algorithm converges to \lambda_{1} from above and that the W^{1,p} distance between the sequence of associated discrete eigenfunctions and E_{\lambda_{1}} tends to zero. The present work presents an extension of the work [ 43 ] on the conforming method to a nonconforming one, aiming at approximating the first eigenpair from below. The extension is highly nontrivial due to the nonconformity of C-R FEs, and requires several new technical tools, especially discrete compactness on a sequence of discrete solutions from adaptive meshes. Note that the approximate first eigenpair from [ 43 ] converges to the exact one from above due to its conformity, while the current approach with C-R FEs can provide a lower bound. The rest of the paper is structured as follows. In Section 2 , we describe the C-R FEM for problem ( 1.3 ). In Section 3 , we establish a sufficient condition for the convergence of discrete solutions using the AFEM, and in Section 4 , we propose an adaptive algorithm and give

Read the complete text