Exascale transport simulations for the understanding of the switching mechanism in atomically thin memristors

Non-volatile resistive switching (NVRS) has emerged as an important concept in the development of high-density information storage and computing. The recent discovery of NVRS in two-dimensional (2D) monolayer structures, such as MoS2 and hexagonal boron nitride (hBN), open a new avenue for memory/computing devices at the ultrathin scale. The fundamental switching mechanism in 2D monolayers, however, is not yet fully understood. It is hypothesized that vacancies in 2D monolayers mediate formation of conducting filamentary channels leading to a high to low resistance state. However, questions remain as to why the current on/off ratio is strongly device-dependent and vary significantly among different experimental works. To address these questions, it is highly desirable to simulate the electronic transport in a realistic device geometry using ab initio approaches for comparison with experimental data. This is rather challenging as quantum transport simulations are computationally demanding. Here, for the first time, we carried out electronic transport simulations of systems consisting of a hBN monolayer sandwiched by top and bottom gold electrodes with the number of atoms up to 3,600. These large transport simulations are made possible by implementing the non-equilibrium Green’s function (NEGF) method in a highly scalable first-principles DFT code: the Real-space MultiGrid (RMG) that runs efficiently in the first exascale supercomputer, Frontier, at Oak Ridge National Laboratory. Systematic calculations reveal that experimental devices exhibit a wide range of on/off ratios (10⁰ to 10⁷) due to variations in interface distances between the electrode and h-BN that significantly modulates the gold/h-BN wavefunction overlap. Our work provides a deeper understanding of the resistive switching mechanism in atomically thin memristors and demonstrates the significance of interface distance in governing the current on/off ratio.