2025-11-10T03:15:01.076525

Construction of Basis Functions for the Geometry Conforming Immersed Finite Element Method

Adjerid, Lin, Meghaichi
The Frenet apparatus is a new framework for constructing high order geometry-conforming immersed finite element functions for interface problems. In this report, we present a procedure for constructing the local IFE bases in some detail as well as a new approach for constructing orthonormal bases using the singular value decomposition of the local generalized Vandermonde matrix. A sample implementation in MATLAB is provided to showcase the simplicity and extensionability of the framework.
academic

Construction of Basis Functions for the Geometry Conforming Immersed Finite Element Method

Basic Information

  • Paper ID: 2510.12018
  • Title: Construction of Basis Functions for the Geometry Conforming Immersed Finite Element Method
  • Authors: Slimane Adjerid (Virginia Tech), Tao Lin (Virginia Tech), Haroun Meghaichi (Ohio State University)
  • Classification: math.NA cs.NA
  • Publication Date: October 15, 2025
  • Paper Link: https://arxiv.org/abs/2510.12018

Abstract

This paper proposes a novel framework based on the Frenet apparatus for constructing high-order geometry-conforming immersed finite element (IFE) basis functions for solving interface problems. The paper details the process of constructing local IFE basis functions and introduces a new method for constructing orthonormalized bases using singular value decomposition (SVD) of local generalized Vandermonde matrices. MATLAB implementation examples are provided, demonstrating the simplicity and extensibility of the framework.

Research Background and Motivation

Problem Background

This research addresses elliptic interface problems: −∇⋅(β∇u)=f, on Ω−∪Ω+-\nabla \cdot (\beta\nabla u) = f, \text{ on } \Omega^- \cup \Omega^+ where the computational domain Ω⊂R2\Omega \subset \mathbb{R}^2 is divided by interface Γ\Gamma into two subdomains Ω−\Omega^- and Ω+\Omega^+, and the diffusion coefficient β\beta is a piecewise constant function.

Research Motivation

  1. Limitations of existing IFE methods: Traditional immersed finite element methods can only approximately satisfy interface jump conditions
  2. High-order accuracy requirements: Need to construct IFE functions for arbitrary-order polynomials
  3. Geometry conformity requirements: Need IFE functions that exactly satisfy interface jump conditions
  4. Computational stability: Need basis function systems with well-conditioned properties

Significance

Interface problems arise widely in multiphase flows, materials science, biomedical engineering, and other fields. High-precision numerical solution methods are of significant importance for engineering applications.

Core Contributions

  1. Proposed a GC-IFE construction framework based on the Frenet apparatus: Utilizes Frenet frames from differential geometry to simplify interface jump conditions
  2. Developed two basis function construction methods:
    • Initial construction method based on special formats
    • Generalized construction method based on extension ideas
  3. Proposed two orthogonalization reconstruction algorithms:
    • Method based on mass matrix SVD (Approach 1)
    • Method based on generalized Vandermonde matrix SVD (Approach 2)
  4. Provided complete MATLAB implementation: Demonstrates the practicality and extensibility of the method

Methodology Details

Frenet Transformation

The core idea is to use Frenet transformation to convert physical coordinates (x,y)(x,y) to local coordinates (η,ξ)(\eta,\xi): [ηξ]→PΓ[x(η,ξ)y(η,ξ)]=g(ξ)+ηn(ξ)\begin{bmatrix} \eta \\ \xi \end{bmatrix} \xrightarrow{P_\Gamma} \begin{bmatrix} x(\eta,\xi) \\ y(\eta,\xi) \end{bmatrix} = g(\xi) + \eta n(\xi)

where g(ξ)g(\xi) is the parametrization of the interface curve and n(ξ)n(\xi) is the normal vector.

Interface Jump Condition Transformation

In the local coordinate system, interface jump conditions become:

  1. ⟦u^⟧Γ^KF=0\llbracket \hat{u} \rrbracket_{\hat{\Gamma}_{K_F}} = 0
  2. ⟦β^u^η⟧Γ^KF=0\llbracket \hat{\beta}\hat{u}_\eta \rrbracket_{\hat{\Gamma}_{K_F}} = 0
  3. ⟦β^∂j∂ηjL(u^)⟧Γ^KF=0,j=0,1,…,m−2\llbracket \hat{\beta}\frac{\partial^j}{\partial\eta^j}\mathcal{L}(\hat{u}) \rrbracket_{\hat{\Gamma}_{K_F}} = 0, j = 0,1,\ldots,m-2

Basis Function Construction Methods

Method 1: Special Format Construction

Construct basis functions of the form: ϕ^i,j(η,ξ)=1β^(η,ξ)qj(ηηh)pi(ξ−ξmidξh)\hat{\phi}_{i,j}(\eta,\xi) = \frac{1}{\hat{\beta}(\eta,\xi)}q_j\left(\frac{\eta}{\eta_h}\right)p_i\left(\frac{\xi-\xi_{mid}}{\xi_h}\right)

For 1≤j≤m,0≤i≤m1 \leq j \leq m, 0 \leq i \leq m, these functions automatically satisfy jump conditions.

Method 2: Generalized Extension Construction

Adopt a more general form:

\hat{\lambda}_j^-(\eta,\xi) = \sum_{i=1}^{(m+1)^2} C_{i,j}^- R_i(\eta,\xi), & \eta < 0 \\ \hat{\lambda}_j^+(\eta,\xi) = \sum_{i=1}^{(m+1)^2} C_{i,j}^+ R_i(\eta,\xi), & \eta > 0 \end{cases}$$ The coefficient matrices $C^{\pm}$ are determined by solving linear systems. ### Orthogonalization Reconstruction #### Approach 1: Mass Matrix SVD Given $M_q = V_1\Lambda V_1^T$, select $Q_1 = V_1\Lambda^{-1/2}$ so that the reconstructed mass matrix becomes the identity matrix. #### Approach 2: Vandermonde Matrix SVD Perform SVD decomposition on $\tilde{V} = \sqrt{W}V$: $\tilde{V} = U_2\Sigma V_2^T$, and select $Q_2 = V_2\Sigma^{-1}$. Theoretical analysis shows that Approach 2 is more numerically stable in implementation, as it avoids SVD decomposition of matrices with squared condition numbers. ## Experimental Setup ### Test Problem Consider domain $\Omega = [-1,1]^2$ with interface $x^2 + y^2 = r_0^2$ ($r_0 = 1/\sqrt{3}$), and test function: $$u(x,y) = \begin{cases} \frac{1}{\beta^+}\cos(2\pi r^2), & r > r_0 \\ \frac{1}{\beta^-}\cos(2\pi r^2) + \cos(2\pi r_0^2)\left(\frac{1}{\beta^+} - \frac{1}{\beta^-}\right), & r < r_0 \end{cases}$$ ### Evaluation Metrics - Convergence order of $L^2$ projection error - Condition number of mass matrix - Numerical stability of basis function construction ## Experimental Results ### Convergence Verification For different polynomial orders $m = 1,2,3,4$, the $L^2$ projection error exhibits optimal convergence orders: | $m$ | $N = 16$ | $N = 32$ | $N = 64$ | $N = 128$ | $N = 256$ | Convergence Order | |-----|----------|----------|----------|-----------|-----------|-------------------| | 1 | 8.14E-02 | 2.08E-02 | 5.23E-03 | 1.31E-03 | 3.28E-04 | ≈2.0 | | 2 | 9.29E-03 | 1.19E-03 | 1.50E-04 | 1.88E-05 | 2.36E-06 | ≈3.0 | | 3 | 8.80E-04 | 5.74E-05 | 3.62E-06 | 2.27E-07 | 1.42E-08 | ≈4.0 | | 4 | 7.55E-05 | 2.39E-06 | 7.49E-08 | 2.34E-09 | 7.33E-11 | ≈5.0 | ### Condition Number Improvement Comparison of mass matrix condition numbers before and after reconstruction shows: - Initial construction condition numbers grow exponentially with polynomial order - Approach 2 stably reduces condition numbers to 1.0 - Approach 1 exhibits numerical instability at high orders ### Preconditioning Effects For coefficient matrices $A$ and $\tilde{A}$, both Jacobi and row normalization preconditioners significantly improve condition numbers, particularly in high-order and small mesh size cases. ## Related Work ### Development of IFE Methods - Traditional IFE methods: Can only approximately satisfy interface conditions - Extended IFE methods: High-order methods based on Cauchy extension - GC-IFE in this paper: Geometry-conforming methods that exactly satisfy interface conditions ### Orthogonalization Techniques - Orthogonal polynomial construction in virtual element methods - Orthogonal basis functions in DG methods - This paper introduces orthogonalization ideas into the IFE framework ## Conclusions and Discussion ### Main Conclusions 1. **Theoretical contribution**: Establishes a GC-IFE theoretical framework based on the Frenet apparatus 2. **Algorithmic innovation**: Proposes two construction methods and two orthogonalization strategies 3. **Numerical verification**: Demonstrates optimal convergence and numerical stability of the method 4. **Practical value**: Provides complete MATLAB implementation ### Limitations 1. **Geometric restrictions**: Requires sufficiently smooth interface curves ($C^3$) 2. **Computational complexity**: Higher matrix assembly costs for high-order cases 3. **Extensibility**: Currently applicable only to two-dimensional elliptic interface problems ### Future Directions 1. Extension to three-dimensional problems and complex geometries 2. Application to time-dependent interface problems 3. Extension to elasticity, Stokes systems, and other applications 4. Development of more efficient implementation algorithms ## In-Depth Evaluation ### Strengths 1. **Theoretical rigor**: Frenet apparatus from differential geometry provides solid theoretical foundation 2. **Methodological innovation**: First implementation of high-order IFE methods that exactly satisfy interface jump conditions 3. **Numerical stability**: Orthogonalization reconstruction significantly improves computational stability 4. **Complete implementation**: Provides detailed algorithm descriptions and MATLAB code 5. **Sufficient verification**: Theoretical analysis and numerical experiments are mutually corroborating ### Weaknesses 1. **Limited applicability**: Currently restricted to two-dimensional elliptic interface problems 2. **Computational cost**: Insufficient analysis of computational complexity for high-order cases 3. **Strict geometric requirements**: Requires high smoothness of interface curves 4. **Lack of practical applications**: Missing validation on complex engineering problems ### Impact 1. **Academic value**: Provides new theoretical framework and technical approach for IFE methods 2. **Practical value**: Offers effective tools for high-precision interface problem solving 3. **Reproducibility**: Complete code implementation facilitates use and extension by other researchers 4. **Development potential**: Establishes solid foundation for subsequent research ### Applicable Scenarios 1. **Multiphase flows**: Numerical simulation of different fluid interfaces 2. **Materials science**: High-precision computation of composite material interface problems 3. **Biomedical engineering**: Heat and mass transfer problems at tissue interfaces 4. **Geophysics**: Wave propagation problems at geological layer interfaces ## References The paper cites 22 relevant references covering important works in differential geometry, numerical analysis, finite element methods, and other fields, providing a solid theoretical foundation for this research.