摘要
The need for robust cryptographic measures has become increasingly important in various applications, including secure communication, artificial intelligent model training, and stochastic artificial neural networks. Among these measures, true random number generators (TRNGs) are essential for creating unpredictable cryptographic keys. In this work, we utilize BaF2 as the active layer of the memristors. The switching mechanism of these memristors originates from the randomly formed and ruptured conductive filaments, but the devices possess a random telegraph noise (RTN)-type random current spiking at a single operation voltage. This spontaneous spiking behavior features a very high on/off ratio of 103, several orders higher than traditional RTN-based TRNGs. Moreover, we confirm that the resistive switching mechanism results from Ag filament formation through time-of-flight secondary ion mass spectrometry and provide a comprehensive understanding of the Ag diffusion during spiking thorough a Monte Carlo simulation with results well-consistent with the experimental data. The randomness of the random number string generated from the BaF2 memristors is confirmed by passing all 15 National Institute of Standards and Technology (NIST) randomness tests. The high on/off ratio, simple operation, and room-temperature fabrication of the BaF2 TRNG device provide great potential for stochastic hardware implementation in various applications.