Scalable Ising machine composed entirely of Si transistors

Area of Science:

• Novel semiconductor device physics
• Emerging computing architectures
• Materials science in electronics

Background:

• Ising machines are crucial for solving complex combinatorial optimization problems.
• Conventional Ising machines often rely on specialized or heterotypic device architectures.
• Scalability and stability challenges hinder the widespread adoption of current Ising machines.

Purpose of the Study:

• To introduce a fully silicon-based complementary metal-oxide semiconductor (CMOS) Ising machine (CIM).
• To leverage the dual characteristics of metal-oxide semiconductor field-effect transistors (MOSFETs) for oscillator and coupler functions.
• To demonstrate a scalable and robust hardware implementation for advanced computing.

Main Methods:

• Utilizing MOSFETs in an unconventional manner as both oscillators and tunable couplers.
• Employing gate biasing to precisely control coupling strengths between oscillators.
• Implementing gate voltage tuning to mitigate frequency variability in large-scale CIMs.

Main Results:

• Achieved effective synchronization within the CIM through dual MOSFET functionality.
• Demonstrated full compatibility with standard CMOS fabrication for scalability.
• Successfully applied the CIM to solve a MaxCUT problem, showcasing its practical utility.

Conclusions:

• The developed CIM offers a promising pathway toward large-scale, energy-efficient Ising machines.
• Compact cell size, high scalability, and robust operation are key advantages of this MOSFET-based approach.
• This technology represents a significant advancement in hardware implementations for optimization problems.