Keywords:Reservoir Computing, Atomic Switches, Ag-Ag2S core-shell nanoparticles
Neuromorphic devices are expected to have a high-performance arithmetic circuit with very low power consumption to be applied in many fields, such as brain-like computer. In the present study, we demonstrated a preliminary study of reservoir computing (RC) hardware using the Ag-Ag2S core-shell nanoparticles aggregation for speech recognition. The Ag-Ag2S core-shell nanoparticles were synthesized by modified Brust-Schiffrin procedure at room temperature with Ag/S molar ratios of 0.25/1 as described in [1, 2]. The RC device was then fabricated by drop-casting highly concentrated nanoparticles in ethanol on to 50 ℃ of multi electrodes device and characterized echo state properties, which exhibit phase shifting of output signal owing to the short-term memory effect as depicted in Figure 1. Another important key point of RC is that the device maps input signals into higher dimensional computational spaces through the dynamics of a fixed, non-linear system that indicated by the generation of higher harmonics at the output as shown in Figure 2. Our study on RC devices using nanoparticles suggested a great potential for further time series prediction tasks such as speech recognition. The details will be presented at the conference.
 C. Battocchio et al., J. Phys. Chem. C 116, 19571 (2012)
 Hadiyawarman et al., Jpn. J. Appl. Phys. 59, 015001 (2020