Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2509.11030 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866908660052000768 |
|---|---|
| author | Feng, Zhihao Randolph, Christian T. Martin, Tyler B. Gartner III, Thomas E. |
| author_facet | Feng, Zhihao Randolph, Christian T. Martin, Tyler B. Gartner III, Thomas E. |
| contents | Polymer reference interaction site model (PRISM) theory, a descendent of Ornstein-Zernike liquid state theory, is a powerful tool to predict the structure and thermodynamics of equilibrium polymer systems, but its accuracy and applicability can be limited in some important cases. Typically, these shortcomings are traced to the analytical closure relationships used to solve the integral equations. Here, we propose a machine learning (ML)-based closure relation trained on a dataset of coarse-grained molecular dynamics simulations of homopolymer melts and solutions. PRISM theory with the ML closure outperforms traditional atomic closures (e.g., Percus-Yevick) in predicting the structure of typical coarse-grained model systems. We also use the ML closure to accurately model the results of small-angle neutron scattering experiments. This ML-enhanced PRISM theory can therefore enable rapid soft materials discovery and design efforts. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_11030 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | A Machine Learning Closure for Polymer Integral Equation Theory Feng, Zhihao Randolph, Christian T. Martin, Tyler B. Gartner III, Thomas E. Soft Condensed Matter Polymer reference interaction site model (PRISM) theory, a descendent of Ornstein-Zernike liquid state theory, is a powerful tool to predict the structure and thermodynamics of equilibrium polymer systems, but its accuracy and applicability can be limited in some important cases. Typically, these shortcomings are traced to the analytical closure relationships used to solve the integral equations. Here, we propose a machine learning (ML)-based closure relation trained on a dataset of coarse-grained molecular dynamics simulations of homopolymer melts and solutions. PRISM theory with the ML closure outperforms traditional atomic closures (e.g., Percus-Yevick) in predicting the structure of typical coarse-grained model systems. We also use the ML closure to accurately model the results of small-angle neutron scattering experiments. This ML-enhanced PRISM theory can therefore enable rapid soft materials discovery and design efforts. |
| title | A Machine Learning Closure for Polymer Integral Equation Theory |
| topic | Soft Condensed Matter |
| url | https://arxiv.org/abs/2509.11030 |