MoS₂ Channel-Enhanced High-Density Charge Trap Flash Memory and Machine Learning-Assisted Sensing Methodologies for...

Area of Science:

• Materials Science
• Electrical Engineering
• Computer Engineering

Background:

• Growing demand for high-density, low-power, reliable nonvolatile memory for AI edge computing.
• Limitations of traditional 3D NAND Flash with polycrystalline silicon (Poly-Si) channels, including short-channel effects and cell-current constraints.
• Need for alternative channel materials to overcome Poly-Si bottlenecks.

Purpose of the Study:

• To investigate molybdenum disulfide (MoS2) as a novel channel material for 3D NAND Flash cells.
• To evaluate the performance and reliability of MoS2-based nonvolatile memory.
• To demonstrate the suitability of MoS2 for next-generation AI-centric edge devices.

Main Methods:

• Fabrication and characterization of 3D NAND Flash cells using MoS2 as the channel material.
• Electrical measurements including thickness-dependent characteristics and temperature-dependent conduction studies.
• Technology Computer-Aided Design (TCAD) simulations and deep reinforcement learning-driven Berkeley Short-channel IGFET Model (BSIM) parameter calibration for circuit-level verification.

Main Results:

• MoS2 enables hole-injection-based erase with a broader memory window at moderate voltages due to its low bandgap.
• A low-k tunneling layer enhances the gate-coupling ratio, reducing program/erase voltages and improving reliability (10^4 cycles endurance, 10^5 s retention).
• MoS2 channel thickness was correlated with endurance and retention metrics, validated through simulations and circuit-level testing.

Conclusions:

• MoS2-based nonvolatile memory effectively meets the demands for high-density, low-power, and reliable storage.
• This material presents a viable and promising solution for AI-driven edge computing applications.
• The developed methodology is applicable for evaluating new channel materials in next-generation memory devices.