Conventional silicon-based CMOS devices encounter severe limitations in high-temperature environments, including functional layer degradation, excessive leakage currents, and the inherent von Neumann bottleneck resulting from the physical separation of memory and computation units. These constraints impede their use in intelligent systems operating under thermally harsh conditions. In this work, we present a thermally robust synaptic transistor based on a lithium phosphorus oxynitride (LiPON) solid-state electrolyte gate dielectric and an indium tin oxide (ITO) semiconductor channel, enabling stable operation over a wide temperature range from 25 °C to 150 °C. Even at 150 °C, it retains a high on/off ratio of (7.56 ± 1.08) × 105 and a subthreshold swing as low as 271 ± 14 mV/dec, demonstrating superior thermal stability compared with conventional ion-gel and proton-conducting gate dielectrics. More importantly, by exploiting the temperature-dependent synaptic plasticity with relaxation timescales tunable over several orders of magnitude, we realize a high-temperature physical reservoir computing (RC) system. This RC system achieves 93.07% accuracy in dynamic pattern recognition and a normalized error of 0.0154 in Mackey-Glass chaotic time-series prediction under high-temperature conditions. This work paves the way for intelligent sensing and edge computing in extreme thermal environments, such as deep-space exploration and geothermal energy development.
Luo et al. (Thu,) studied this question.