Soft Matter Molecular Simulations

Hybrid Particle-Field MD goes neural! A new cover on Macromolecules.

Neural networks can be efficiently trained using local density fields as physical order parameter. The multiscale workflow integrates coarse graining from atoms to beads and voxelization into a 3D density field as the physical order parameter for deep learning in block copolymer/carbon nanotube nanocomposites.

Link to the paper in the first comment

Physics-Informed Graph Learning for Field-Based Molecular Models

In two of our recent works, we introduced a new paradigm that integrates hybrid particle–field molecular dynamics (hPF-MD) with graph neural networks, moving beyond black-box machine learning toward physically grounded representation learning.

Links to the two papers are in the first two comments.

Università di Salerno 20 Febbraio 2026