Nanoscaffold Ba0.6Sr0.4TiO3:Nd2O3 ferroelectric memristors crossbar array for neuromorphic computing and secure encryption

Abstract

Recent advancements in AI have spurred interest in ferroelectric memristors for neuromorphic chips due to their ability to precisely control resistive states through polarization flip-flop without electroforming. However, oxygen vacancies in these devices often cause high leakage current, low endurance, and dispersed switching voltages. Here, we introduce a silicon-based integrated (Ba_0.6Sr_0.4TiO₃)_0.5(Nd₂O₃)_0.5 (BSTN) nanoscaffolded ferroelectric thin film memristor with a vertically self-assembled nanocomposite structure (VSNs) optimally oriented on La_0.67Sr_0.33MnO₃/SrTiO₃/P–Si substrates. This device demonstrates a widely tunable ferroelectric domain range (0°–180°), high remnant polarization (21.04 μC/cm²), and a greater number of unitary states (16 states or 4 bits). It exhibits high durability, enduring over 10⁹ switching cycles. The switching mechanism combines ferroelectric polarization and oxygen vacancy migration, enabling the simulation of biological synaptic functions via bi-directional conductance tunability. Additionally, we implemented a low-power (0.57 pJ per event) multi-factor secure encryption system for smart locks using 16×16 BSTN memristor crossbar arrays and a pressure sensor. Under multiple factors (disordered inputs, specific users, and corresponding passwords) the system recognized passwords with 97.6% accuracy and a 3.8% loss rate after 500 iterations. Overall, this work establishes a robust foundation for advancing multilevel storage, neuromorphic computing, and AI chip applications based on ferroelectric memristors.