Negative Feedback in Hf 0.5 Zr 0.5 O 2 Resistive Memory: Feigenbaum Route Toward Two‐Terminal Pseudo‐Random Number Generation and Neural Network Instabilities

ABSTRACT Resistively switching metal‐insulator‐meal (MIM) structures are a strong contender for universal memory technologies and neuromorphic computing applications. Their ability to combine non‐volatility with analogue tuneability enables in‐memory computing, where both data processing and storage occur in the same physical location. A period‐doubling bifurcation was observed in endurance tests of TiN/ Hf 0.5 Zr 0.5 O 2 /Au devices, where sequences of voltage pulses with constant amplitudes are supplied to the stack. Based on this, the memory cell's response to increasingly larger voltage pulses is measured. The extracted parameters are used to deduce a probabilistic non‐linear map describing endurance cycling of the MIM system. The investigation of the dynamics involves the identification of parameters that have been reported to be subject to change during continuous switching, and the investigation of how this drift affects the switching dynamics. The found model thus serves as a design framework to either deliberately develop systems exhibiting chaotic dynamics or circumvent them for classical memory applications.