Gradually Varied Functions and λ-Connectedness for Structure-Guided Finite Element Approximation
Year: 2026
Type: preprint
Abstract: Finite element methods provide powerful tools for solving partialdifferential equations once a discretization and approximation spacehave been specified. In problems with spatially heterogeneous solutionbehavior, however, an additional question arises: how can informationcontained in discrete function values be used to identify appropriatelocal approximation structures? This paper investigates graduallyvaried functions (GVF) and λ-connectedness as a discrete structurallayer for this purpose. GVF provides a global C 0 extension of sparsefunction values on a graph, while λ-connectedness analyzes the resultingfield through a hierarchy of connectivity levels determined by localvariation. A persistence measure derived from this hierarchy is usedto construct data-dependent structural partitions without requiringderivative information.The framework is studied using the Cahn–Hilliard equation dis-cretized by a mixed finite element formulation, together with a manufac-tured example exhibiting heterogeneous polynomial and exponential be-havior. Numerical experiments examine the evolution of λ-connectivity,automatic λ-based partitions, sparse-data GVF reconstruction, andvalidated local selection among polynomial and exponential approx-imation families. An additional experiment demonstrates the use ofλ-partition boundaries to guide local mesh refinement and subsequentfinite element recalculation. The experiments indicate that λ-derivedregions can exhibit different local approximation behavior and provideuseful structural information for adaptive numerical computation.
Author: Chen, Li
Language: English
Subfield: Computational Mechanics
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7236054132
