Optimization of Nonreciprocal Transmission Through Dissipative Phononic Crystals With Machine Learning Techniques

Abstract

Transmission through a phononic crystal of metallic rods in a viscous environment is numerically calculated. The cross-section of the rods is selected to be asymmetric to provide very different transmission in opposite directions along a given crystallographic line. Difference in transmission contains the reciprocal part, caused by asymmetry of the scatterers, and the truly nonreciprocal part, related to nonequal viscous losses for sound waves propagating in opposite directions. The rectification ratio for different levels of asymmetry is evaluated and optimized over its value at a fixed frequency, with various machine learning models. The possibility of using asymmetric phononic crystals as acoustic diodes is discussed.