Abstract
This paper presents a systematic framework for constructing finite element de Rham and Stokes complexes in three dimensions with customizable levels of smoothness ($C^m$ continuity). By employing a novel non-overlapping decomposition of simplicial lattices based on geometric distances to sub-simplices, the authors derive stable, exact sequences of functional spaces. These constructions are vital for robust numerical solutions to complex partial differential equations (PDEs), such as the Stokes and Maxwell equations, ensuring both exactness and divergence stability in 3D simulations.
Introduction
In numerical simulation, discretizing physics equations requires more than just approximating values; the discrete spaces used for different physical quantities (e.g., velocity and pressure) must maintain specific mathematical relationships. This is formalized as a Hilbert complex—a sequence of spaces where the range of one differential operator is exactly the kernel of the next. While these complexes are well-established in 2D, extending them to 3D with high-order smoothness (ensuring continuity of derivatives across element boundaries) has historically been a significant challenge. This work addresses the need for a unified, systematic construction of these complexes in 3D.
Methodology: Geometric Decomposition and Barycentric Calculus
The core of the proposed framework is a sophisticated geometric decomposition of simplicial lattices. Instead of treating a tetrahedron as an monolithic block, the authors partition the lattice of polynomial degrees of freedom (DoFs) based on their algebraic "distance" to sub-simplices like vertices, edges, and faces.
As shown in the figure above, this distance metric allows for precise control over $C^r$ continuity at each geometric level. By categorizing nodes based on their proximity to specific boundaries, the construction turns a complex geometric continuity problem into a manageable combinatorial one.
To enforce these continuity requirements across shared faces of tetrahedra, the authors leverage Bernstein polynomials and Barycentric Calculus. This approach makes it easier to define and manipulate functions within the barycentric coordinate system of the simplex.
The use of local tangential vectors and barycentric coordinates (illustrated above) facilitates the construction of interpolation operators that satisfy commutative properties. This ensures that the discrete model faithfully represents the continuous physics—for instance, ensuring that taking the gradient of an interpolated function yields the same result as interpolating the gradient itself.
Key Insights and Results
Unified Construction
A major contribution of this work is the demonstration that $H(\text{div})$ and $H(\text{curl})$ elements can be derived from the same scalar smooth finite element framework. This unified approach simplifies the creation of complete de Rham complexes.
Stability and Exactness
The authors provide rigorous mathematical proofs for the exactness of their constructed sequences on domains homeomorphic to a ball, meaning the methods are free of "spurious modes." Furthermore, they establish div-stability (the LBB condition) for velocity-pressure pairs in the Stokes complex by carefully constructing "bubble spaces"—polynomials that vanish on the boundaries—to fill gaps in the divergence operator's range.
New 3D Families
The paper successfully generalizes famous 2D elements, such as the Hermite and Argyris families, into full 3D de Rham and Stokes complexes. While the focus is on 3D, the authors also provide a roadmap for generalizing this lattice decomposition to arbitrary $n$-dimensions.
Conclusion and Future Work
This framework provides a robust theoretical foundation for high-order smooth finite elements in 3D. While these elements offer superior accuracy and stability for fluid and electromagnetic simulations, they come with increased computational complexity due to the high number of degrees of freedom per element. Future research will focus on optimizing the implementation of these elements in standard software and exploring their performance in domains with complex topologies or non-trivial boundary conditions.
Summary updated