| |
 |
UNIVERSITY OF BUCHAREST
FACULTY OF PHYSICS Guest
2024-11-22 2:06 |
 |
|
|
|
Conference: Bucharest University Faculty of Physics 2024 Meeting
Section: Polymer Physics
Title:
Machine Learning Assisted Calibration of the Lebwohl-Lasher Model for Liquid Crystal Simulations
Authors:
Adrian BERLIC (1,2), Catalin BERLIC (1)
Affiliation:
1) University of Bucharest, Faculty of Physics, 405 Atomistilor Street, 077125, Magurele, Romania
2) National Meteorological Administration, 97 Soseaua Bucuresti - Ploiesti, Bucharest, Romania
E-mail
cataliniulian.berlic@g.unibuc.ro
Keywords:
liquid crystals, computer simulation, Lebwohl-Lasher, machine learning, neural network
Abstract:
The accurate simulation of liquid crystal behaviors is important for advancing materials science and engineering applications. Traditional computational models, while robust, often require extensive calibration to align with experimental data, a process that can be both time-consuming and computationally expensive. These include the scenarios where the characteristics of the surface, confined geometries, the size, location, and quantity of inclusions, as well as temperature, electric or magnetic fields cand lead to complicated behaviors. This study introduces a novel approach that employs machine learning techniques to enhance the calibration of the Lebwohl-Lasher model, which may become important for simulating nematic liquid crystals. We developed a hybrid framework combining the Lebwohl-Lasher model with supervised learning algorithms to adjust model parameters based on previous simulations or experiment. The initial phase involves training a neural network on a dataset derived from high-fidelity simulations under varied conditions. The trained models predicted model parameters that optimize the alignment of simulation outputs with observed data, significantly reducing the manual parameters adjusting traditionally required.
|
|
|
|