ECE scientists demonstrate quantum computing using silicon technology
ECE scientists led by Distinguished McKnight University Professor Chris Kim successfully demonstrated the solving of an optimization problem using a chip fabricated using current day semiconductor technology. The research titled
“A 1,968 node coupled ring oscillator circuit for combinatorial optimization problem solving,” is published in Nature Electronics. The demonstration is significant as it shows that computationally challenging problems that are considered to be the domain of quantum computers, can be solved using conventional complementary metal-oxide-semiconductor (CMOS) technology. While quantum computing has held a lot of promise, it is constrained by issues related to the building of a practical large-scale error-tolerant hardware platform. The outcome of the team’s research holds the promise of delivering on critical factors such as speed and energy efficiency that are several orders of magnitude higher than those of current state of the art digital computers.
Kim who is also the Louis John Schnell Professor in Electrical and Computer Engineering commented on the challenge of using silicon technology to resolve problems considered to be in the realm of quantum computing: "The key research question was how to emulate quantum properties such as superposition and entanglement in a silicon chip operating at room temperature."
New opportunities in conventional technology
A wide range of problems in disparate fields including operations and logistics, finance, communication network design, and drug discovery are characterized by combinatorial optimization problems. Conventional computing technology based on the von Neumann architecture struggles with solving these problems because of the demands on computational power, which results in delays and energy inefficiencies. Unconventional and relatively newer approaches such as quantum computing and optical-based methods are affected by mass-manufacturing and other challenges. For instance, quantum computers require near zero absolute temperature conditions and therefore kilowatts of power per chip to maintain those conditions. Optical Ising machines (an Ising machine is a device that can solve combinatorial optimization problems) need a kilometer long optical fiber, which limits power efficiency, and density.
Although digital computers are ubiquitous and manufactured at a large scale, their use in solving computationally complex problems is hindered by critical factors such as time to solution and energy consumption. Currently, the iterative nature of optimization problems makes computation time and energy consumption prohibitive in the case of conventional digital computers. For the Kim research team, the situation presented an opportunity. They took on the challenge of combining speed and energy efficiency with the ubiquity of silicon technology.
They worked on the premise that quantum properties can have parallels in classical integrated circuits. If ring oscillators stand in for qubits, then superposition and entanglement (quantum properties that are key to solving complex problems quickly) can be emulated. For instance, entanglement (which allows a quantum computer to quickly complete an exponentially larger number of calculations than a regular digital computer) is comparable to the phases of oscillators pulling or pushing each other in a coupling interaction. Significantly, the signal to noise ratio of a CMOS ring oscillator circuit is high enough to allow it to operate at room temperature, a move away from the near absolute zero operating temperature requirements of quantum computers.
Taking advantage of the scalability of silicon chips, the team constructed an Ising machine from ring oscillators, starting with 6 oscillators in 2018, and building up to 1,968 in 2021, on a single silicon chip. Along with the increase in the number of oscillators, other specifications such as the readout speed, phase sampling accuracy, and weight precision were improved with each generation of the Ising chip design.
Another significant outcome of their work is that they were able to measure the relaxation time of a coupled oscillator array comprising thousands of nodes. This is a first in itself, but a point of additional significance is that the value was less than 50 oscillation cycles. It points to a considerable reduction in time required to solve a problem, and therefore the energy consumed to accomplish the solution. These are compelling results that prove that complex problems originally believed to be solvable only by quantum computers, can indeed be solved using a silicon chip fabricated in a standard semiconductor foundry.
What the current research demonstrates is that a system based on coupled CMOS ring oscillators can solve the same complex combinatorial optimization problems that are the target of quantum computers and other emerging technologies, while also avoiding the manufacturing challenges that typically plague them. While the researchers point to areas of improvement their current design could benefit from at the software and hardware ends, and additional work that will need to be done to reach commercial potential, their work has reliably demonstrated the immense promise it holds. By deploying these chips in devices as varied as a remote data center or a mobile device such as a smartphone, they can solve problems in areas as diverse as resource allocation, traffic and vehicle routing, machine learning, and beyond.
Acknowledgements: The project was supported by the National Science Foundation (NSF) and the Semiconductor Research Corporation (SRC).