Enhancement of NbO2-based oscillator neuron device performance via cryogenic operation.

The Niobium Dioxide (NbO2) oscillator neuron has garnered significant interest because of its simple structure compared to conventional CMOS-based circuits. However, the limited on/off resistance ratio narrows the range of series resistances that satisfy the self-oscillation conditions and limits its use in large-scale synaptic arrays. In this study, we report the possibility of improving the performance of NbO2-based oscillator neuron devices through cryogenic operation. The study emphasizes two crucial parameters: the on/off resistance ratio and the oscillation amplitude, both of which are essential for accurate weighted sum classification. The data suggest that these parameters can be effectively enhanced under cryogenic conditions. In addition, we revealed that 120 K is the optimal temperature for cryogenic operation, as it represents the temperature where the on/off resistance ratio ceases to increase. As a result, we revealed that the series resistance range satisfying the self-oscillation condition in a single oscillator increases from 20 to 126 kΩ. The research also probes the maximum possible array size at each temperature. At 300 K, representation is only possible for a 5 × 5 array, but at 120 K, a 30 × 30 array can be represented as a frequency. The evidence implies that the 120 K conditions not only broaden the range of series resistors that can be connected to a single oscillator but also increases the array size, thereby representing different weighted sum currents as frequencies. The research indicates that using carefully optimized cryogenic operation could be a viable method to enhance the necessary NbO2properties for an oscillator neuron device.

Sodium-based nano-ionic synaptic transistor with improved retention characteristics.

We propose an all-solid-state Na ion-based synaptic transistor (NST) to overcome the low retention problem of the Li ion-based synaptic transistor (LST). Through our analysis, it was found that the retention instability in an ionic synaptic transistor originated from its high ionic diffusivity. As confirmed by cyclic voltammetry analysis, Na ions have a lower ionic diffusivity than Li ions in the WOx layer. The state retention of NST was found to be improved to 20 times that of LST. Furthermore, near-ideal synaptic behaviors, such as linear weight update and linear I-V characteristics, were also obtained by material engineering.