Probabilistic Computing using Metal Nanoparticles-Embedded Mott Threshold Switches for True Random Number Generator

VO₂ threshold switch, which undergoes a insulator-to-metal transition above a threshold voltage, shows promise as a candidate for stochastic switching due to its capacity to determine the overall electrical transition in devices through the percolative formation of metallic filaments [1]. However, geometric evolution of the switchable (metallic or insulating) domains significantly influences the stochastic switching in VO₂; the performance of VO₂ based stochastic devices (e.g., true random number generators) could be tuned by the evolution of metallic domains in the insulating matrix and random formation of conducting filaments [2].<br/><br/>Here, we achieve drastic enhancement of intrinsic stochastic operation under pump-probe procedure in Pt-nanoparticle-embedded VO₂ (Pt-VO₂) two-terminal devices. In particular, metallic Pt nanoparticles enable an extension of memorization time from pump pulse, which enhances probability of certain selection to show random firing of on/off (1 or 0) states under probe pulse; the extended recovery time to the insulating state presents a wide range of options in terms of probe voltage amplitude and delay time, thereby enhancing the randomness of the switching behavior upon application of the second probe pulse. Moreover, this device has the advantage of high speed (667 kbits/s), low power consumption (12 nJ/bit), high endurance (&gt;105200 bits), and mass production, which is key requirement to realize true random number generator (TRNG). Therefore, our result provides a new strategy to improve stochastic switching characteristics in Mott threshold switches for high performance TRNG.<br/><br/>References<br/>[1] Valle, Javier del, et al., <i>Nano letters</i> 22, 1251 (2022)<br/>[2] Jo, Minguk et al., <i>Nature Communications</i> 13, 4609 (2022)