Granular memristors with tunable stochasticity

Most realizations of memristive devices exhibit characteristic noise sometimes described as random telegraph noise. These fluctuations in current, ubiquitous in nature, carry significant implications for device performance, reliability, and the broader landscape of memristor technology applications. Here, we study inherent random fluctuations observed in silver based granular memristive devices operating under steady bias conditions. Random telegraph noise observed in our system is characterized in terms of distributions of ON and OFF times of the current flow at a particular bias. We find that these fluctuations adhere to power law statistics with , where τ_OFF/ON denotes the time during which the output value remains below or above a specified threshold. We follow the fluctuations for up to four decades. Significantly, unlike previous studies, we find the emergence of a new regime of behavior where the power law exponent varies as a function of applied bias. We find that our results are best described by the Marcus-Tang expression for diffusion along intersecting parabolae with bias as the driving force. The predictions of this picture of dynamics also provide a satisfactory explanation for the quiescence of the OFF/ON state of our devices.