Surface Acoustic Wave Physical Reservoir Computer Using Thermoacoustic Phase Modulator

Machine learning tasks were conducted by a surface acoustic wave (SAW) thermoacoustic phase modulator using physical reservoir computing architecture to realize an on-chip mechanical computing device. The SAW nonlinear modulator was fabricated on a 128°Y-X cut piezoelectric lithium niobate wafer. The interdigital transducers and waveguide were defined by deposited gold film. The modulator acts as a physical reservoir, consisting of the time-expanded system. The performance was evaluated using NARMA2 benchmark task and the normalized mean square error was 0.102, as good as other simple physical reservoir computers.