A DeepRitzSplit neural operator trained on energy-split variational forms enforces dissipation in phase-field models and outperforms data-driven training in generalization while running faster than Fourier spectral methods on Allen-Cahn and dendritic growth cases.
Rethinking materials simulations: Blending direct numerical simulations with neural operators
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
math.AP 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
DeepRitzSplit Neural Operator for Phase-Field Models via Energy Splitting
A DeepRitzSplit neural operator trained on energy-split variational forms enforces dissipation in phase-field models and outperforms data-driven training in generalization while running faster than Fourier spectral methods on Allen-Cahn and dendritic growth cases.