Spontaneous High-Dynamic-Range\nRandom Current Spiking\nin BaF<sub>2</sub> Memristors

The need for robust cryptographic measures has become\nincreasingly\nimportant in various applications, including secure communication,\nartificial intelligent model training, and stochastic artificial neural\nnetworks. Among these measures, true random number generators (TRNGs)\nare essential for creating unpredictable cryptographic keys. In this\nwork, we utilize BaF<sub>2</sub> as the active layer of the memristors.\nThe switching mechanism of these memristors originates from the randomly\nformed and ruptured conductive filaments, but the devices possess\na random telegraph noise (RTN)-type random current spiking at a single\noperation voltage. This spontaneous spiking behavior features a very\nhigh on/off ratio of 10<sup>3</sup>, several orders higher than traditional\nRTN-based TRNGs. Moreover, we confirm that the resistive switching\nmechanism results from Ag filament formation through time-of-flight\nsecondary ion mass spectrometry and provide a comprehensive understanding\nof the Ag diffusion during spiking thorough a Monte Carlo simulation\nwith results well-consistent with the experimental data. The randomness\nof the random number string generated from the BaF<sub>2</sub> memristors\nis confirmed by passing all 15 National Institute of Standards and\nTechnology (NIST) randomness tests. The high on/off ratio, simple\noperation, and room-temperature fabrication of the BaF<sub>2</sub> TRNG device provide great potential for stochastic hardware implementation\nin various applications.