Using graph neural networks to predict many-body interactions in amorphous materials
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866914615770742784 |
|---|---|
| author | Ghomsheh, Mehryar Jannesari Koch, Donald L. Hormozi, Sarah |
| author_facet | Ghomsheh, Mehryar Jannesari Koch, Donald L. Hormozi, Sarah |
| contents | Many-body interactions govern the complex behavior of many amorphous materials, from metallic glasses to biological tissues, yet are often replaced by pairwise additive frameworks for computational efficiency. Here, we use classical density functional theory (DFT) to study a model soft glass of solvent-free polymer-grafted nanoparticles (PGNs), where the absence of solvent forces grafted chains to uniformly fill the interstitial space, generating strong angular-dependent many-body interactions between the cores. We show that NequIP, an equivariant message-passing graph neural network (GNN), learns the high-dimensional, rugged potential energy landscape of the system and reproduces classical DFT energies across a range of PGN design parameters at four orders of magnitude lower cost. Systematic analysis of GNN hyperparameters offers physical insights into the range, anisotropy, and effective body order of interactions. GNN-driven Monte Carlo simulations reveal locally favored icosahedral-like structures at equilibrium, and strikingly, recover equilibrium structures in agreement with experiments, despite the network being trained only on high-energy, out-of-equilibrium configurations. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_30822 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Using graph neural networks to predict many-body interactions in amorphous materials Ghomsheh, Mehryar Jannesari Koch, Donald L. Hormozi, Sarah Disordered Systems and Neural Networks Soft Condensed Matter Many-body interactions govern the complex behavior of many amorphous materials, from metallic glasses to biological tissues, yet are often replaced by pairwise additive frameworks for computational efficiency. Here, we use classical density functional theory (DFT) to study a model soft glass of solvent-free polymer-grafted nanoparticles (PGNs), where the absence of solvent forces grafted chains to uniformly fill the interstitial space, generating strong angular-dependent many-body interactions between the cores. We show that NequIP, an equivariant message-passing graph neural network (GNN), learns the high-dimensional, rugged potential energy landscape of the system and reproduces classical DFT energies across a range of PGN design parameters at four orders of magnitude lower cost. Systematic analysis of GNN hyperparameters offers physical insights into the range, anisotropy, and effective body order of interactions. GNN-driven Monte Carlo simulations reveal locally favored icosahedral-like structures at equilibrium, and strikingly, recover equilibrium structures in agreement with experiments, despite the network being trained only on high-energy, out-of-equilibrium configurations. |
| title | Using graph neural networks to predict many-body interactions in amorphous materials |
| topic | Disordered Systems and Neural Networks Soft Condensed Matter |
| url | https://arxiv.org/abs/2605.30822 |