Zn²⁺ Engineered Low-Barrier LiNbO₃ Enables Visible-Light Programmable Ferroelectric Memristors for Noise-Immune...

Lithium niobate (LiNbO₃), owing to its unique ferroelectric polarization and excellent optical properties, has shown great potential in high-performance optoelectronic integrated devices. However, the high polarization switching energy barrier makes it difficult to achieve polarization reversal under low-power visible light, severely limiting its practical applicability. Here, Zn²⁺ ions were doped into the LiNbO₃ lattice to modulate the local lattice structure via valence-state imbalance, effectively suppressing the formation of Nb_Li ⁴⁺ antisite defects and reducing electron-trap density. Meanwhile, the narrowed bandgap enhanced carrier excitation efficiency and improved depolarization-field screening, lowering the polarization switching energy barrier by approximately 69% and enabling polarization reversal under low-energy visible light illumination (10 mW cm^-2). Accordingly, the fabricated Pt/Zn-LiNbO₃/Nb:SrTiO₃ optoelectronic bimodal memristor exhibits ultra-stable switching voltage characteristics, with a voltage coefficient of variation as low as 2.2%-3.2%; a high on/off ratio of approximately 10³; 2⁴ clearly distinguishable resistance states; retention exceeding 10⁴ s; and excellent endurance up to 10⁸ cycles. Under visible light stimulation, the device emulates multiple representative synaptic functions, including short-term to long-term memory (STP-LTP) transition, paired-pulse facilitation (PPF), and associative learning. Moreover, an optical reservoir computing neural network constructed from the device's multilevel optical memory and synaptic features achieves a high recognition accuracy of 98.6% on the noise-corrupted MNIST dataset, demonstrating robustness and visual recognition capability comparable to biological systems. This study proposes a new materials design paradigm for constructing low-barrier, high-performance ferroelectric optoelectronic systems with integrated sensing, storage, and computation functionalities.