Ultra-large dynamic range synaptic indium gallium zinc oxide transistors

Abstract

Three-terminal neuromorphic transistors have garnered considerable attention owing to their superior learning and recognition capabilities in neural computing. For efficient and rapid computing of parallel arithmetic and unstructured large-sized data, a high-synaptic channel conductance (G) (i.e., synaptic weight) and large dynamic range (DR) (i.e., weight update) are necessary. In this study, we successfully fabricated high-performance and all-solution-processed synaptic transistors by employing a thiol–ene photoclick chemistry-based, cross-linked vinyl addition polynorbornene copolymer dielectric layer, poly(norbornene-co-5-vinyl-2-norbornene) (P(NB/VNB)) and sol-gel-derived indium gallium zinc oxide (IGZO) semiconducting channel layer. The molecular behaviors of free hydroxyl group in the dielectric layer formed via ultraviolet photo-induced thiol–ene click reactions of the polar cross-linker pentaerythritol tetrakis(3-mercaptopropionate) (PETMP) allows slow dipole polarization effects on the IGZO channel layer. An exhaustive and systemic investigation of the surface physicochemical and electrical properties confirmed that the mutual interface coupling between the cross-linked P(NB/VNB) dielectric and IGZO semiconductor exhibited remarkable synaptic functionality with a high G (∼540 μS), large DR (∼6213), and long-term operational stability following excitation with 104 successive pulses that resulted in neural recognition accuracy of 87.11% based on the use of MNIST database set which is close to ideal value of 88%. This high-performance synaptic transistor based on all-solution-derived functional organic–inorganic hybrid stack layers can serve as a basis for the successful implementation of novel neuromorphic computing systems that satisfy the requirements for strong neuronal signal transmission and distinguishable multi-level data states.