An Artificial Memristor Synapse by Transferring and Stacking Freestanding Single-Crystalline SrTiO

Transition metal oxide (TMO) crystalline epitaxial thin film-based resistive switching (RS) devices are proven to be promising candidates for future data storage, information processing, and artificial neuromorphic computing. However, the development of TMO epitaxial film devices is hindered by strict epitaxial preparation requirements, including coherent substrates and high temperatures. The emerging freestanding TMO films, with the advantages of room-temperature transferring and stacking, are desired to be a promising platform for building new oxide devices, while it is still a challenging endeavor to transfer high-quality freestanding oxide films to assemble high-performance devices. Here, we report a prototype synaptic memristor based on freestanding SrTiO3-δ membranes (model perovskite) hosted on indium tin oxide via transferring and stacking. The assembled memristor displays excellent RS performance with a large ON/OFF ratio, good endurance, and superior retention time compared to the reported epitaxially grown memristors. Moreover, the memristor is proved to be an excellent synaptic emulator with multiple synaptic plasticity and is successfully employed in a reservoir computing system to realize pattern recognition with high accuracy. The newly assembled oxide synaptic memristor could theoretically be built on arbitrary substrates, breaking the restriction of epitaxial growth and revealing the huge application potential in next-generation high-density memory and neuromorphic computing.