Toward high-speed effective numerical simulation of multiple filamentation of high-power femtosecond laser radiation in transparent medium

TL;DR

This paper introduces an improved numerical simulation method for high-power femtosecond laser filamentation in transparent media, utilizing machine learning to optimize phase screen techniques and accelerate computations.

Contribution

The paper presents a novel numerical approach combining phase screen modifications with machine learning to enhance simulation speed and accuracy of laser filamentation.

Findings

Significantly faster simulation times compared to classical methods.

Maintained high accuracy in modeling filamentation phenomena.

Effective application of machine learning for optimizing numerical solutions.

Abstract

High-power femtosecond laser radiation during the propagation in air (and other transparent media) experiences multiple filamentation. Filamentation is a unique nonlinear optical phenomenon, which is accompanied by a wealth of nonlinear optical effects such as formation of extended plasma channels in the beam wake, generation of higher harmonics and supercontinuum, generation of THz radiation. The manifestations of laser filamentation can be useful for solving atmospheric optics problems related to remote sensing of the environment as well as directed transmission of laser power. The classical numerical methods used for simulating the nonlinear long-range atmospheric propagation of high-power radiation with a sufficiently large laser beam aperture have almost reached their limit regarding the acceleration of calculations. To solve this problem and speed-up the numerical simulations of laser filamentation, we propose an improved numerical technique based on a modified method of phase screens constructed on a sparse spatial grid. Within the framework of this technique, we seek for optimal ansatz (substitution function) to the governing equations using the machine learning technology, which provides for the best correspondence to the numerical solution of the test problem using a denser spatial grid