Li-Well ZnO Memtransistors: High Reliability for Neuromorphic Applications

Area of Science:

• Materials Science
• Solid State Physics
• Device Physics

Background:

• Memtransistors integrate transistor functionality with nonvolatile memory using ionic memristive channels.
• Existing memtransistor technologies like FeFETs and charge-trap flash have limitations.
• Ionic memtransistors have historically lagged behind in performance compared to other nonvolatile memory devices.

Purpose of the Study:

• To introduce a facile and extendable lithium-well oxide memtransistor (LWOM) as a high-performance ionic memtransistor candidate.
• To demonstrate analog memory characteristics via Schottky barrier modulation induced by Li-ion migration.
• To evaluate LWOM's potential for nonvolatile memory and artificial neural network (ANN) acceleration.

Main Methods:

• Fabrication of LWOM devices using mature oxide semiconductor technology with a 230°C thermal budget.
• Induction of Li⁺-ion migration via write V_DS to modulate the Schottky barrier.
• Characterization using 3D secondary ion mass spectrometry (SIMS) to confirm Li-ion redistribution and resistance-switching mechanism.
• Testing of a 21 × 21 crossbar array for operational yield and weight update accuracy.

Main Results:

• LWOM exhibits low-voltage weight updates and precise gate-controlled weight update characteristics.
• 3D SIMS analysis confirmed Li-ion redistribution and the underlying resistance-switching mechanism.
• A 21 × 21 crossbar array achieved a 99.31% operational yield with successful weight updates to target conductance values.

Conclusions:

• LWOM is a promising candidate for next-generation nonvolatile memory due to its facile fabrication and performance.
• The device's analog memory characteristics and precise control make it suitable for artificial neural network (ANN) acceleration hardware.
• LWOM leverages mature oxide semiconductor technology, offering a viable path for scalable memory solutions.