A hybrid high-order method for the biharmonic problem

Yizhou Liang and Ngoc Tien Tran · 2025

Read A hybrid high-order method for the biharmonic problem by Yizhou Liang and Ngoc Tien Tran (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-2504-16608.txt.

Introduction

Yizhou Liang and Ngoc Tien Tran propose, in 2025, a new numerical method for the biharmonic problem, a fourth-order equation that models plate bending and slow fluid flow. Their hybrid high-order method adds a fresh ingredient: extra unknowns living on lower-dimensional pieces of the mesh, not just inside cells and on faces.

That addition restores a commuting property the method family is prized for, in any number of dimensions. The team proves guaranteed lower bounds for the related vibration eigenvalues, at higher order than before. Error estimates hold under weak smoothness assumptions, and they drive a mesh refinement scheme that recovers top convergence rates even for rough solutions.

Catalog record

CREATORYizhou Liang and Ngoc Tien Tran
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE81 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-16608

Browse the full library · About the Tower of Babel

Full text

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

A hybrid high-order method for the biharmonic problem Thanks: The research of the first author was partly supported through a Royal Society University Research Fellowship (URF\R1\221398, RF\ERE\221047). The second author received funding from the European Union’s Horizon 2020 research and innovation programme (project RandomMultiScales, grant agreement No. 865751) Yizhou Liang and Ngoc Tien Tran Abstract . This paper proposes a new hybrid high-order discretization for the biharmonic problem and the corresponding eigenvalue problem. The discrete ansatz space includes degrees of freedom in n − 2 n-2 dimensional submanifolds (e.g., nodal values in 2D and edge values in 3D), in addition to the typical degrees of freedom in the mesh and on the hyperfaces in the HHO literature. This approach enables the characteristic commuting property of the hybrid high-order methodology in any space dimension. The main results are guaranteed lower eigenvalue bounds of higher order. Furthermore, we derive quasi-best approximation estimates as well as reliable and efficient a posteriori error estimators under minimal regularity assumptions on the exact solution. The latter motivates an adaptive mesh-refining algorithm that empirically recovers optimal convergence rates for singular solutions. 2020 Mathematics Subject Classification 65N30, 65N25, 65N15 1. Introduction Let \Omega\subset\mathbb{R}^{n} , n\in\{2,3\} , be a polyhedral Lipschitz domain with boundary \partial\Omega and unit outer normal vector \nu_{\partial\Omega} . The main focus of this paper is the approximation of the eigenpairs (\lambda,u) of the biharmonic eigenvalue problem (1.1) \displaystyle\Delta^{2}u \displaystyle=\lambda u\quad\text{in }\Omega, \displaystyle u \displaystyle=0\quad\text{on }\partial\Omega, \displaystyle\partial_{n}u \displaystyle=0\quad\text{on }\partial\Omega. Here and throughout this paper, the subscripts n and t are associated with differential operators in the normal and tangential directions, respectively. The variational formulation of ( 1.1 ) seeks (\lambda,u)\in\mathbb{R}_{\geq 0}\times V with V\coloneqq H^{2}_{0}(\Omega) such that (1.2) \displaystyle(\nabla^{2}u,\nabla^{2}v)_{\Omega}=\lambda(u,v)_{\Omega}\quad\text{for any }v\in V. For the source problem, there are numerous finite element methods for the approximation of ( 1.2 ). Conforming methods include the piecewise quintic Argyris element in 2D [ 2 ] or C^{1} conforming elements in a subtriangulation, also known as Hsieh-Clough-Tocher splits in 2D [ 20 , 37 ] and Worsey–Farin splits in 3D [ 47 , 37 ] . Due to the complicated nature of C^{1} conforming elements in the implementation, in particular in 3D, many nonconforming methods have been developed. On simplicial meshes, we mention the lowest-order Morley element [ 41 ] as well as its higher-order generalization in 2D [ 7 ] . Finite elements of arbitrary order on general meshes include the virtual elements (VE) of [ 10 , 51 , 1 ] in 2D, the weak Galerkin (WG) methods of [ 43 , 49 ] , the hybrid high-order (HHO) methods of [ 8 , 29 , 30 ] , and the discontinuous Galerkin (dG) methods of [ 42 , 36 ] . The degrees of freedom of the aforementioned methods are different and lead to varying computational efficiency. In structured meshes, the performance of all methods is fairly comparable; cf. [ 29 ] for a comparison between some HHO, WG, dG and the Morley finite element methods. The focus of this paper is the HHO methodology introduced in [ 26 , 27 ] ; we refer to the monographs [ 28 , 21 ] for an overview of applications. The HHO methodology is a class of hybridizable methods that can be defined for arbitrary polytopal meshes and polynomial degree – a flexibility shared by many nonconforming methods from above. In fact, close relationship between hybridizable methods such as HDG, HHO, and WG exists, cf. [ 23 , 22 , 21 ] for further discussions. In certain applications, hybridizable methods allow for additional benefits. We mention guaranteed error control in convex minimization problems by duality techniques, where HHO methods provide a cheap post-processing of the dual variable [ 18 , 44 ] . For eigenvalue problems, upper bounds can be obtained by conforming methods. However, applications, e.g., in safety analysis require guaranteed lower eigenvalue bounds (LEB), which is accessible by hybridizable methods [ 19 , 15 , 45 ] . Here, direct lower bounds can be obtained with a fine-tuned stabilization, i.e., the discrete eigenvalue is a lower bound of the exact one. This leads to higher-order convergence rates and provides access to adaptive mesh-refining algorithm. Thus, the HHO eigensolver of this paper can be considered as a higher-order generalization of [ 17 ] . It turns out that these benefits are related to the design of the reconstruction operator, which satisfies the following commuting property. Let V_{h} denote the discrete ansatz space and I_{h}:V\to V_{h} is a canonical interpolation of V onto V_{h} . The Hessian of a continuous function is approximated by the piecewise Hessian of a reconstruction operator R_{h}:V_{h}\to P_{k+2}(\mathcal{T}) , where k\geq 0 is the degree of approximation and \mathcal{T} is the underlying mesh, such that the L^{2} orthogonality (1.3) \displaystyle\nabla^{2}_{\mathrm{pw}}(v-R_{h}I_{h}v)\perp\nabla^{2}_{\mathrm{pw}}P_{k+2}(\mathcal{T})\quad\text{for any }v\in V holds. This property is available in 2D for [ 29 ] and in arbitrary space dimensions for [ 8 ] , although the design of [ 29 ] appears more sophisticated as it utilizes less (non-hybridizable) degrees of freedom. To achieve ( 1.3 ) in any space dimensions and derive LEB for ( 1.2 ), we utilize an ansatz similar to [ 39 ] . The point of departure is the local integration by parts formula (1.4) \displaystyle(\nabla^{2}v,\nabla^{2}p)_{T}=(v,\Delta^{2}p)_{T}-(v,\partial_{n}\Delta p)_{\partial T}+(\partial_{n}v,\partial_{nn}p)_{\partial T}+(\partial_{t}v,\partial_{nt}p)_{\partial T} for any v\in H^{2}(T) and p\in P_{k+2}(T) in a mesh element T from [ 29 ] . For the Poisson equation, lowest-order hybridizable and Crouzeix-Raviart methods are closely related. Thus, considering the degrees of freedom of Morley finite elements motivates another integration by parts for the term (\partial_{t}v,\partial_{nt}p)_{\partial T} along the n-1 dimensional hyperfaces F of T with outer normal vector \nu_{\partial F} . We arrive at (1.5) \displaystyle\begin{aligned} (\nabla^{2}v,\nabla^{2}p)_{T}&=(v,\Delta^{2}p)_{T}-(v,\partial_{n}\Delta p)_{\partial T}+(\partial_{n}v,\partial_{nn}p)_{\partial T}\\ &\quad-(v,\Delta_{t}\partial_{n}p)_{\partial T}+\sum_{F\in\mathcal{F}(T)}(v,\partial_{tn}p\cdot\nu_{\partial F})_{\partial F},\end{aligned} where \partial F denotes the n-2 dimensional boundary of the face F . (In 2D, integrals over vertices are understood as the point evaluation.) This formula leads to a discrete ansatz space that also involves degrees of freedom to approximate the restriction of v along \partial F similar to [ 39 ] and the virtual elements [ 10 , 51 , 1 ] . These additional degrees of freedom ensure that the resulting reconstruction operator R_{h} satisfies ( 1.3 ). From a computational point of view, the degrees of freedom along the edges lead to a larger stencil compared to [ 29 ] and, thus, higher computational cost is expected, cf. Remark 2.5 below for further details. This is, however, justified for the task at hand. In combination with a new stabilization, we derive the first LEB with higher convergence rates for the biharmonic problem in the current literature with the arguments of [ 19 , 45 ] . To be precise, we establish that (1.6) \displaystyle\mathrm{LEB}(j)\coloneqq\min\{1,1/(\alpha+\beta\lambda_{h}(j))\}\lambda_{h}(j)\leq\lambda(j) is a lower bound for the j -th exact eigenvalue \lambda(j) . Here, \lambda_{h}(j) is the j -th discrete eigenvalue, \alpha=\sigma(1/\pi^{4}+c_{\mathrm{tr}}/\pi^{2}+c_{\mathrm{tr}}+c_{\mathrm{tr}}(2/\pi+n/\pi^{2})) with a parameter \sigma chosen by the user and the constant c_{\mathrm{tr}} from the continuous trace inequality, and \beta=h^{4}/\pi^{4} . Notably, \alpha and \beta are explicit and independent of the polynomial degree k . In particular, if \alpha<1 and the mesh size h is sufficiently small, then \lambda_{h}(j)\leq\lambda(j) , which provides higher-order convergence rates

Read the complete text