TL;DR
This paper introduces discrete versions of the Hessian and divdiv complexes, crucial for understanding differential structures in physics and engineering, on triangulated domains. By generalizing existing finite element methods and Discrete Exterior Calculus, the authors establish that the cohomology of these new discrete complexes accurately reflects their continuous counterparts, offering a powerful tool for numerical simulations.
What Problem Does This Paper Address?
Accurately solving partial differential equations (PDEs) that involve vector-valued functions and differential forms relies heavily on preserving underlying mathematical structures, particularly the de Rham complex. While discretizations of the de Rham complex are well-established (e.g., Whitney forms), extending these robust discrete topological structures to other fundamental differential complexes like the Hessian and divdiv complexes has been a significant challenge. These latter complexes are essential for areas like elasticity, Riemannian geometry, and fluid dynamics. Existing approaches often require high-order polynomials or specialized meshes, limiting their applicability. This paper aims to bridge this gap by constructing discrete versions of the Hessian and divdiv complexes on triangulations, using distributional shape functions and Dirac measures, and demonstrating their cohomological equivalence to their continuous counterparts.
Methods
The authors construct discrete versions of the Hessian and divdiv complexes on triangulations in 2D and 3D. Their approach is inspired by the Bernstein-Gelfand-Gelfand (BGG) construction, which derives new complexes from existing ones while preserving cohomology. They generalize concepts from the de Rham complex (Whitney forms) and elasticity (Regge calculus) by employing finite elements with local polynomial shape functions and, crucially, distributional shape functions and Dirac delta measures on subsimplices. This allows for a more flexible construction that doesn't strictly adhere to conforming finite element spaces. The paper leverages Discrete Exterior Calculus (DEC) perspectives, particularly the idea of discretizing forms on dual meshes, to motivate their construction. They then rigorously prove that the cohomology of these newly constructed discrete complexes is isomorphic to the cohomology of their continuous counterparts, which in turn is related to de Rham cohomology.
Datasets
Not discussed in the paper.
Key Insights and Ideas
- Generalizing Whitney Forms: The paper successfully extends the concept of Whitney forms, which are canonical conforming finite elements for the de Rham complex, to the Hessian and divdiv complexes. This provides a discrete topological interpretation for these more complex differential structures.
- Distributional Shape Functions: The use of distributional shape functions and Dirac delta measures is key to constructing these discrete complexes on triangulations without requiring overly restrictive conformity conditions or high-order polynomials.
- Cohomological Equivalence: A central finding is the demonstration that the cohomology of the constructed discrete Hessian and divdiv complexes is isomorphic to their continuous counterparts. This is crucial for ensuring that numerical solutions retain essential topological and geometric properties of the underlying continuous problem.
- DEC Connection: The paper highlights a strong connection between Discrete Exterior Calculus (DEC) and distributional finite elements, suggesting that DEC's dual mesh perspective can inform the design of finite element spaces for more complex differential operators.
Key Results
The paper establishes the existence and properties of discrete Hessian and divdiv complexes on triangulated domains in 2D and 3D. The primary result is the proof that the cohomology of these discrete complexes is isomorphic to the de Rham cohomology with appropriate coefficients. This is demonstrated through detailed constructions of finite element spaces and differential operators, followed by rigorous mathematical proofs involving diagram chases and duality arguments. Specific theorems are presented for the 2D and 3D Hessian and divdiv complexes, detailing their cohomological properties.
Limitations and Future Work
The authors acknowledge that the resulting spaces for the BGG complexes are less regular than standard Whitney forms. This might pose challenges for direct application in solving PDEs and would require further numerical analysis. They suggest that their approach could be extended to discretizing high-dimensional and high-order tensors on triangulations, with potential applications in areas like numerical geometric PDEs, exterior calculus on graphs, and computer graphics. Computing the cohomology of the 3D Regge sequence is also mentioned as a future direction. The paper also notes that while the constructed complexes are dual to each other, further investigation into their properties and applications, particularly in solving PDEs, is needed.
Summary updated