Physics-inspired computer architecture solves complex optimization problems

A line of engineering research seeks to develop computers that can tackle a class of challenges called combinatorial optimization problems. These are common in real-world applications such as arranging telecommunications, scheduling, and travel routing to maximize efficiency.

Unfortunately, today’s technologies run into limits for how much processing power can be packed into a computer chip, while training artificial-intelligence models demands tremendous amounts of energy.

Researchers at UCLA and UC Riverside have demonstrated a new approach that overcomes these hurdles to solve some of the most difficult optimization problems. The team designed a system that processes information using a network of oscillators, components that move back and forth at certain frequencies, rather than representing all data digitally. This type of computer architecture, called an Ising machine, has special power for parallel computing, which makes numerous, complex calculations simultaneously. When the oscillators are in sync, the optimization problem is solved.

In the study, published in Physical Review Applied, the investigators reported on a device based on certain quantum properties that link electrical activity with vibrations traveling through a material. However—unlike most current quantum applications in computing, which require extremely low temperatures to maintain their “quantumness”—the researchers’ device is capable of operation at room temperature.

“Our approach is physics-inspired computing, which has recently emerged as a promising method to solve complex optimization problems,” said corresponding author Alexander Balandin, the Fang Lu Professor of Engineering and distinguished professor of materials science and engineering at the UCLA Samueli School of Engineering. “It leverages physical phenomena involving strongly correlated electron–phonon condensate to perform computation through physical processes directly, thus achieving greater energy efficiency and speed.”

The research showed that oscillators naturally evolve to a ground state, in which they’re synced up, allowing the machine to solve combinatorial optimization problems.

Balandin and his colleagues used a special material to bridge the gap between quantum mechanics—counterintuitive rules governing interactions between subatomic particles—and the more familiar physics of everyday life. Their prototype hardware is based on a form of tantalum sulfide, a “quantum material” that makes it possible to reveal the switching between electrical and vibrational phases.

The new technology has the potential for low-power operation; at the same time, it can be compatible with conventional silicon technology.

“Any new physics-based hardware has to be integrated with the standard digital silicon CMOS technology to impact data information processing systems,” said Balandin, a member of the California NanoSystems Institute at UCLA, or CNSI. “The two-dimensional charge-density-wave material that we selected for this demonstration has the potential for such integration.”

The coupled oscillators in this research were built at the UCLA Nanofabrication Laboratory, jointly run by CNSI and UCLA Samueli, and tested in UCLA’s Phonon Optimized Engineered Materials laboratory.