Jihoon Yang1,Soon-Yong Kwon1,Zonghoon Lee1,2,Donghyeok Lim1,Hongsik Jeong1,Aram Yoon1,2,Donghyun Lee1,Iljohn Jung1
Ulsan National Institute of Science and Technology1,Institute for Basic Science2
Jihoon Yang1,Soon-Yong Kwon1,Zonghoon Lee1,2,Donghyeok Lim1,Hongsik Jeong1,Aram Yoon1,2,Donghyun Lee1,Iljohn Jung1
Ulsan National Institute of Science and Technology1,Institute for Basic Science2
Two-dimensional (2D) materials have generated attention for neuromorphic computing applications, and are promising for use in low-power synaptic devices at the atomic scale. However, 2D material-based memristors are disadvantaged by the large stochastic forming process, which results in switching variability. In this study, we present a 2-inch wafer scale memristor array that contains nanometer-sized grains of highly polycrystalline 2H-MoTe<sub>2</sub> as an active medium and exhibits reliable resistive switching. The polycrystalline 2H-MoTe<sub>2</sub> films were synthesized on a 2-inch SiO<sub>2</sub>/Si wafer by tailoring the Te flux through a Te vapor-confined method using a eutectic alloy. The synthesized films contain uniformly sized nanograins (~60 nm) and exhibit ultrahigh-density (1.37<b>×</b>10<sup>11 </sup>cm<sup>-2</sup>) grain boundaries (GBs). These GBs provide confined defective paths to enable conduction, facilitating reliable resistive switching. Compared to single-crystalline 2H-MoTe<sub>2</sub>-based memristors, the polycrystalline 2H-MoTe<sub>2</sub>-based memristor (PMM) arrays show improved resistive switching uniformity and stable multi-level resistance states, along with high device yields (>83.7%), small device-to-device variations (<13.8%), and long retention times (>10<sup>5 </sup>s). Finally, this PMM shows linear analog synaptic plasticity under repeated pulses more than 2,500 and exhibits a learning accuracy of 96.05% for MNIST handwritten digit classification. The introduction of nanograins in the PMM represents a novel route to accelerate the use of 2D memristors in practical neuromorphic computing applications.