Array-Level Integration of MoS2-Hf0.5Zr0.5O2 Ferroelectric Memristors for in-Memory Image Preprocessing in Neural Networks.

In-memory computing using floating-gate memristor (FGMEM) arrays with 2D channels enables large-scale parallelism, low energy consumption, efficient hardware-level matrix-vector multiplications. However, tunneling-based FGMEM faces high operating voltage, slow programming, low on/off ratio, and a small memory window. Here, we demonstrate a 16 × 16 ferroelectric memristor (FeMEM) array integrating a 2D MoS2 channel with a hafnium-zirconium oxide (HZO) layer, enabled by metal-insulator-metal (MIM) annealing and top-electrode etching for direct HZO-MoS2 coupling. This enables ±1.5 V, 10 µs array-level operation, outperforming FGMEMs (>±4 V, ∼100 ms). The array achieves 91% yield and excellent uniformity in ON-OFF current and forward/reverse threshold voltage (Vth), with Gaussian distributions within ±3σ of the mean. The FeMEM exhibits a large memory window (32.3%) and a high Ion/Ioff ratio of 1.02 × 105-about 2 and 1000 times higher than an FGMEM, respectively. It shows high linearity (β = 0.14/0.68 for long-term potentiation/depression). Parallel-multiply (PM) operations between input voltages (V) and conductance (G) yield an output current map (I = G × V) corresponding to a preprocessed image achieving quality gains of +13.03 dB PSNR, +0.226 SSIM, and 20 times MSE reduction after preprocessing. As an in-memory preprocessing for a CNN, it boosts classification accuracy from 55.7% (noisy inputs) to 95.5% (excluding failed devices).