Transition Between Digital Switching and Analog Synapse in AgOx/SnSe2 Memristor Through Thickness Engineering

Why It Matters

Enabling a single memristor to serve both high‑density digital storage and analog neural computing simplifies hardware architectures and accelerates neuromorphic system integration.

Key Takeaways

• Thick AgOx yields >10^6 ON/OFF ratio
• Thin AgOx enables gradual analog conductance
• Subthreshold swing drops below 10 mV/dec
• Device retains state for 5000 seconds
• Simple thermal oxidation fabricates neuromorphic memristor

Pulse Analysis

Memristors have emerged as a bridge between conventional memory and brain‑inspired computing, yet most devices are locked into either digital or analog behavior. The ability to switch modes without redesigning the circuit addresses a critical bottleneck in scaling neuromorphic processors. By focusing on the AgOx layer thickness, the researchers provide a straightforward, temperature‑compatible method that can be integrated into existing semiconductor lines, reducing both cost and fabrication complexity.

In the digital regime, a 75 nm AgOx layer forms robust oxygen‑vacancy filaments that break and reform abruptly, delivering an ON/OFF ratio exceeding one million and a subthreshold swing under 10 mV per decade. These metrics rival or surpass commercial resistive‑RAM, while the device maintains data for over 5,000 seconds and remains stable in ambient air. Conversely, a 35 nm AgOx film weakens filament formation, producing a smooth, incremental resistance change that mimics synaptic weight updates, enabling on‑chip learning and real‑time signal classification such as electrocardiogram waveform detection.

The broader impact lies in consolidating storage and processing functions within a single nanoscale element, a key step toward eliminating the von Neumann bottleneck. Industries ranging from edge AI to biomedical monitoring can benefit from compact, low‑power chips that adapt their operation on demand. Future work will likely explore scaling the thickness‑gradient approach to arrays, integrating with CMOS drivers, and extending material systems, positioning silver‑oxide memristors as a versatile platform for next‑generation intelligent hardware.