Strain engineering on the insulator–metal transition of VO2-memristor for high-speed and low-power neural firing

Vanadium dioxide (VO2)-memristors exhibit abrupt electrically driven insulator–metal transition (IMT), making them promising for high-speed switching and neuromorphic computing. The device performance (such as operation voltage, operation speed, and energy consumption) of VO2 memristors strongly depend on dynamic IMT, yet systematic understanding of their correlations is lacking. In this work, high-quality VO2 thin films were grown epitaxially on TiO2 substrates with different crystal orientations, which enabled strain engineering in VO2 lattice and tunable critical temperature (Tc) of IMT ranging from 315 to 366 K. Quasi-static current–voltage measurements reveal that the electrically driven IMT in VO2 memristor is closely related to the Tc of IMT, which follows the Joule heating model well. The neural firing behaviors and device performance of various VO2 memristors are evaluated based on electric pulse tests. The VO2 memristor with the lowest Tc (∼315 K) exhibits the best performance with an integration time of 66.4 ns and energy consumption of 14.2 pJ at a relatively low operation voltage of 1.8 V, significantly superior to other VO2 memristors with higher Tc. Furthermore, the effects of extrinsic factors, such as applied voltage, ambient temperature, and load resistance, on device performance are investigated to explore the strategies of optimizing the operation speed and energy consumption of VO2 memristor in spiking neural network applications. Our work not only elucidates the correlations between IMT dynamics and device performance of VO2 memristors but also provides guidance for material selection and device operation in high-speed and low-power neuromorphic electronics.