High-Performance Ferroelectric Field-Effect Transistors Based on Ultrathin Indium Oxide for Neuromorphic Computing

Area of Science:

• Materials Science
• Computer Engineering
• Artificial Intelligence

Background:

• Traditional von Neumann computing struggles with AI's computational demands.
• In-memory computing (IMC) offers a solution to the von Neumann bottleneck.
• Field-effect transistors (FeFETs) are promising for IMC hardware but face challenges.

Purpose of the Study:

• To demonstrate a complementary metal oxide semiconductor (CMOS)-compatible In2O3 FeFET array for neuromorphic computing.
• To evaluate the performance and potential of In2O3 FeFETs for IMC applications.

Main Methods:

• Fabrication of an In2O3 FeFET array compatible with CMOS processes.
• Characterization of FeFET performance metrics: on-off ratio, memory window, endurance, retention, variation, uniformity, and linearity of potentiation/depression.
• Evaluation of the FeFET array's performance in image classification tasks.

Main Results:

• Achieved an ultrahigh on-off ratio (10^7) and a large memory window (>6 V).
• Demonstrated high endurance (10^7 cycles), long retention (>10 years), low variation (1.1%), and high uniformity.
• Exhibited linear and symmetrical long-term potentiation (LTP) and long-term depression (LTD).
• Attained 92.5% accuracy in image classification.

Conclusions:

• In2O3 FeFETs show excellent performance characteristics for IMC hardware.
• The developed FeFET array holds significant potential for constructing neuromorphic computing systems.
• This work paves the way for advanced AI hardware by addressing computational bottlenecks.