Mesh-Intrinsic GFEM: High-Order Smoothness on $C^0$ Unstructured Meshes
For engineers and researchers using standard C0 finite element codes, this method provides a unified high-order route for solving high-order PDEs without modifying existing mesh infrastructures.
The paper introduces a mesh-intrinsic generalized finite element method (MiGFEM) that achieves high-order smoothness on standard C0 unstructured meshes without extra global degrees of freedom, enabling pointwise strong-form evaluation of high-order PDEs. Numerical experiments demonstrate machine-precision patch tests and robust performance on distorted meshes.
High-order partial differential equations (PDEs) require derivative regularity that standard $C^0$ finite element infrastructures do not directly provide on unstructured meshes. We propose a mesh-intrinsic generalized finite element method (MiGFEM) that reconstructs local polynomial fields on overlapping nodal patches from shared nodal unknowns and blends them by a partition of unity, without introducing extra global degrees of freedom. The core analysis establishes a partition-of-zero (PoZ) smoothness-transfer mechanism driven by interface coherence: derivative jumps cancel exactly for polynomial reproduction and decay as $O(h^{p+1-|α|}) $for smooth nonpolynomial fields. On this basis, we define a PoZ-consistent intrinsic derivative that is polynomial-exact and approximation-order consistent, enabling pointwise strong-form evaluation of high-order PDEs on $C^0$ meshes. For derivative-type/free boundary conditions in strong-form collocation,we introduce a boundary absorption constrained weighted least-squares strategy (BA-CWLS), which embeds boundary constraints into local patch reconstruction. This avoids globally overdetermined boundary augmentation and penalty tuning, while preserving a square sparse global system. Numerical experiments show machine-precision patch tests,jump-decay rates consistent with theory, and robust performance on highly distorted meshes. The same mesh-intrinsic trial space supports both weak-form Galerkin and strong-form collocation discretizations, providing a unified high-order route on standard $C^0$ mesh infrastructures.