physics.comp-ph papers | Gist.Science

Computational physics bridges the gap between abstract theory and real-world observation by using powerful computers to solve complex physical problems. This field allows scientists to simulate everything from the collision of subatomic particles to the swirling dynamics of galaxies, offering insights that traditional experiments alone cannot provide.

On Gist.Science, we continuously process every new preprint in this category from arXiv to make these breakthroughs accessible to everyone. Each entry is accompanied by both a clear, plain-language explanation and a detailed technical summary, ensuring that researchers and curious readers alike can grasp the significance of the latest findings without getting lost in dense equations.

Below are the latest papers in computational physics, curated to keep you at the forefront of this rapidly evolving discipline.

When Attention Beats Fourier: Multi-Scale Transformers for PDE Solving on Irregular Domains

This paper introduces the Multi-Scale Attention Transformer (MSAT), a deep learning architecture that outperforms Fourier-based neural operators and other baselines in solving partial differential equations on irregular domains by leveraging learned attention mechanisms, while also establishing theoretical and empirical boundaries for when physics-informed regularization improves or hinders generalization.

Brandon Yee, Pairie Koh, Jack Rodriguez, Mihir Tekal2026-05-12🤖 cs.LG

A meshfree exterior calculus for generalizable and data-efficient learning of physics from point clouds

This paper introduces MEEC-Net, a meshfree exterior calculus framework that learns structure-preserving physics on point clouds through a differentiable, mesh-free discretization, enabling highly data-efficient, generalizable surrogate modeling that significantly outperforms baseline neural operators on unseen geometries and physical parameters.

Benjamin D. Shaffer, Brooks Kinch, M. Ani Hsieh, Nathaniel Trask2026-05-12🔬 physics

CrystalREPA: Transferring Physical Priors from Universal MLIPs to Crystal Generative Models

CrystalREPA is a plug-and-play framework that enhances the stability, validity, and fidelity of generated crystals by aligning generative model representations with frozen universal machine learning interatomic potentials (MLIPs) through a contrastive objective, revealing that an MLIP's effectiveness for transfer depends more on its representation distinguishability than its standard accuracy benchmarks.

Chengqian Zhang, Yucheng Jin, Duo Zhang, Tiejun Li, Han Wang2026-05-12🔬 cond-mat.mtrl-sci

Nonlinear GENERIC Informed Neural Networks (N-GINNs): learning GENERIC dynamics with non-quadratic dissipation potentials

This paper introduces Nonlinear GENERIC Informed Neural Networks (N-GINNs), a deep learning framework that enforces thermodynamic consistency through convex dissipation potentials to accurately discover evolution equations for systems exhibiting both conservative dynamics and non-quadratic dissipation.

Vojtěch Votruba, Zequn He, Weilun Qiu, Celia Reina, Michal Pavelka2026-05-12🔬 physics

Semi-Supervised Neural Super-Resolution for Mesh-Based Simulations

The paper introduces SuperMeshNet, a semi-supervised neural framework that utilizes complementary learning and inductive biases to efficiently reconstruct high-fidelity mesh-based simulation solutions from low-resolution data while requiring 90% less high-resolution training data than fully supervised benchmarks.

Jiyeon Kim, Youngjoon Hong, Won-Yong Shin2026-05-12🔬 physics.app-ph

Constitutive Priors for Inverse Design

This paper presents an end-to-end framework for the inverse design of elastic networks that optimizes spatially varying material properties in the space of constitutive behaviors by leveraging a thermodynamically consistent latent prior, homotopy-based continuation, and neural network smoothness constraints to robustly solve PDE-constrained optimization problems without requiring mesh correspondence.

Jinkyo Han, Bahador Bahmani2026-05-12🔬 physics

Teaching Molecular Dynamics to a Non-Autoregressive Ionic Transport Predictor

This paper proposes a non-autoregressive learning framework that utilizes atomic trajectories as an auxiliary modality during training to enable fast, accurate, and dynamic ionic transport prediction from static structures without requiring sequential inference or trajectory data at inference time.

Jiyeon Kim, Byungju Lee, Won-Yong Shin2026-05-12🔬 physics.atom-ph

Classification of Chimera States via Fourier Analysis and Unsupervised Learning

This paper proposes a novel method combining Fourier analysis and unsupervised clustering of normalized total variations to precisely detect and classify various types of chimera states in networks of coupled Rayleigh oscillators, overcoming the limitations of existing detection techniques.

Rommel Tchinda Djeudjo, Riccardo Muolo, Thierry Njougouo, Timoteo Carletti2026-05-12🌀 nlin

Accuracy assessment of scalar wave propagation methods for diffractive optics design: from thin elements to thick binary grating

This paper systematically evaluates the accuracy of thin-element approximation, beam propagation, and wave propagation methods against a rigorous reference for binary diffractive gratings, generating accuracy maps to guide the selection of appropriate forward models in inverse design pipelines based on spatial frequency and grating thickness.

Nicolas Barré2026-05-12🔬 physics.optics

Fast Evaluation of the Azimuthal Fourier Modes of the 3D Helmholtz Green's Function and Their Derivatives

This paper presents an $O(M)$ algorithm that efficiently and accurately evaluates the azimuthal Fourier modes of the 3D Helmholtz Green's function and their derivatives for any real wavenumber by combining contour deformation with stable recurrence relations, enabling high-performance axisymmetric acoustic scattering simulations.

Hanwen Zhang2026-05-12🔬 physics

← Previous Next →