Forming-Free Multilevel\nResistive Switching in a ZnO@β-SiC\nComposite for Neuromorphic Computing

Recent attention has been focused on developing artificial\nsynaptic\ndevices for in-memory computing, with the aim of long-term device\nstability, low power consumption, and high performance. Oxides and\nperovskites have been explored in this context to address limitations\nof the Von Neumann architecture. However, these materials face individual\nconstraints that hinder their full potential. This study introduces\na unique approach using a ZnO@β-SiC composite for low-power,\nhigh-performance, and forming-free bipolar resistive switching devices.\nNotably, these devices exhibit switching from high to low resistance\nstates at a very low voltage of ∼100 mV with a fast response\ntimes of ∼40 ns and 50 ns for positive and negative pulses,\nrespectively, and consume a very low power of ∼100 μW.\nChemical and microstructure analyses reveal Zn<sub>2</sub>SiO<sub>4</sub> nanocrystals embedded in an amorphous layer, and it is found\nto be suitable for enhancing device stability over 10<sup>4</sup> cycles\nwith ∼10<sup>4</sup> s retention. The phenomenon is explained\nby the formation and dissolution of oxygen vacancy and metal cation-driven\nconductive filaments. Moreover, the devices effectively replicate\nversatile synaptic functions such as excitatory postsynaptic current,\npair pulse facilitation, potentiation/depression, long-term memory/short-term\nmemory, and learning/forgetting behavior. This work thus presents\na promising avenue for the sustainable development of artificial intelligence\nthrough in-memory neuromorphic computing.