A Ag2S-island network is fabricated with surrounding electrodes to enable it to be used as a reservoir for unconventional computing. Local conductance change occurs due to the growth/shrinkage of Ag...

A physical reservoir that accepts direct light irradiation as input was developed using a Ag2S island network. Short-term memory and nonlinearity required for reservoirs are achieved by diffusion of Ag+ cations in each Ag2S island and growth of Ag filaments between Ag2S islands. We found that direct light irradiation to Ag2S islands changes local conductivity in a reservoir, which enhances the performance in short-term memory and nonlinearity of the reservoir. Using the effect, we performed a pattern classification of light that was irradiated to a Ag2S island network reservoir through a rectangular slit, which resulted in the accuracy over 95%.

Water-adsorbed states on Si, native oxide and thermal oxide surfaces are investigated by means of thermal desorption spectroscopy. D2O is used to detect the water signals from these surfaces, which are separated from background H2O signals. The water desorption is 1/10 less than the hydrogen desorption. The amount of hydrogen adsorbed is largest on the Si surface and smallest on the thermal oxide surface. Four types of binding states of hydrogen are assigned to be SiO-H and Si-H on the silicon surface, Si-H in the native oxide, and Si-H in the interface between the native oxide and the silicon substrate.

The rapid growth in demand for edge artificial intelligence increases importance of physical reservoirs that work at low computational cost with low power consumption. A Ag2S island network also works as a physical reservoir, in which various physicochemical phenomena contribute to a reservoir operation. In this study, we investigated its frequency dependence and found that diffusion of Ag+ cations in a Ag2S island, which has a relaxation time of about 100 µs, plays a major role when performance is improved. MNIST classification task using an input pulse width of 100 µs resulted in the accuracy of 91 %. Iterative operations up to 10 million cycles revealed a small enough standard deviation of output, suggesting a potential for practical use of a Ag2S island network as a reservoir.

No abstract