Physics Informed Neural Networks applied to liquid state theory

•The Physics Informed Neural Network is used to solve Ornstein–Zernike equation.•Results for PY and HNC are compared with the ones obtained by the iterative method.•The pair correlation function decreasing behaviour is garanteed in the new method.•This method enable to recover the pair correlation f...

Full description

Saved in:
Bibliographic Details
Published in:Journal of molecular liquids 2022-12, Vol.367, p.120504, Article 120504
Main Authors: Carvalho, Felipe Silva, Braga, João Pedro
Format: Article
Language:English
Subjects:
Correlation functions
Experimental data
Machine Learning
Ornstein–Zernike equation
Physics Informed Neural Networks
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•The Physics Informed Neural Network is used to solve Ornstein–Zernike equation.•Results for PY and HNC are compared with the ones obtained by the iterative method.•The pair correlation function decreasing behaviour is garanteed in the new method.•This method enable to recover the pair correlation function from experimental data. The calculation of correlation functions for liquids is a problem that can be solved using different methods such as molecular dynamics, integral equation theories or from experimental data. Several Machine Learning algorithms, which are extremely powerful to solve classification and regression methods, have been applied to chemistry and physics. In this work, the recently proposed Physics Informed Neural Networks will be applied to solve the Ornstein–Zernike equation and also to retrieve the radial distribution function from experimental data. This model enables one to solve both direct and inverse problems by defining properly the problem cost function. The results found demonstrate the robustness of this method, when applied to liquid state theory problems.
ISSN:0167-7322
1873-3166
DOI:10.1016/j.molliq.2022.120504