Calculation of the Scattering Matrix in Nanoscale Systems using Physics-Informed Neural Networks
Calculation of the Scattering Matrix in Nanoscale Systems using Physics-Informed Neural Networks
Mohammad Hadi Ahouei1 Ehsan Noruzifar2
1) K. N. Toosi University of Technology, Tehran, Iran,
2) K. N. Toosi University of Technology, Tehran, Iran,
IOI Code
:
AAHD-ACZBFB
Publication :
National Conference on Information Technology, Nanotechnology, Artificial Intelligence and Technological Futures Studies(itnaf.ir)
Abstract :
Calculating physical interactions and responses in nanoscale systems, particularly those with complex geometries, requires the accurate solution of the governing differential equations and the determination of scattering coefficients. The scattering matrix is a key tool for describing the response of a single particle or structure to an external field, as well as for studying the interaction between multiple particles. In this research, a method based on Physics-Informed Neural Networks (PINNs) is proposed for computing the scattering matrix in nanoscale systems. In this approach, the neural network is trained without the need for training data obtained from numerical simulations, by utilizing the governing differential equation, boundary conditions, and far-field conditions. The derivatives required to form the residual of the governing equation are automatically calculated using automatic differentiation, and the network s loss function is composed of the sum of the PDE residual and the errors associated with the boundary conditions. After obtaining the scattered field, the coefficients corresponding to different modes are extracted, and the matrix is computed. To evaluate the method, a circular system is considered as a test problem, and the results obtained from the neural network will be compared with the analytical solution. This framework can pave the way for the development of artificial intelligence-based methods for calculating scattering properties and studying the interactions of nanoscale systems.