A team of researchers at Queen’s University has developed a powerful new kind of computing machine that uses light to take on complex problems such as protein folding (for drug discovery) and number partitioning (for cryptography). Built from off-the-shelf components, it also operates at room temperature and remains remarkably stable while performing billions of operations per second.
This breakthrough shows that it is possible to build a practical and scalable machine that can tackle extremely difficult problems.
The project, led by Bhavin Shastri, Canada Research Chair in Neuromorphic Photonic Computing and professor in the Department of Physics, Engineering Physics, and Astronomy, with a team of his graduate students including Nayem Al Kayed and Hugh Morison, uses commercially available lasers, fibre optics, and modulators – the same technology that powers today’s internet infrastructure. The team partnered with McGill University researcher David Plant and his graduate student Charles St-Arnault.
The research was recently published in Nature – one of the world’s most prestigious scientific journals.
Since the machine operates at room temperature it consumes significantly less energy than other advanced computing systems.
Throughout testing, the Shastri Lab’s machine has proven to be stable for long periods, operating for hours at a time, which makes it well-suited for solving problems that require repeated steps.
A century-old concept, reimagined
The Queen’s processor is based on the Ising model, which represents problems as interacting magnets with “spins” that point up or down. Much like how magnets naturally align when brought closer, the Ising searches for the lowest-energy state – mathematically equivalent to finding the best solution to a difficult optimization problem. Though simple, the model is powerful for solving problems with many interconnected binary (up/down or yes/no) choices.
The Queen’s system instead uses pulses of light that act like the magnets – but instead of a binary system of up or down, there is either a light pulse, or the absence of one. The pulses move through a loop, interact, and gradually settle into a configuration that represents a good solution, much like a group reaching a consensus after many quick exchanges.
“It’s a way to turn light into a problem solver,” Dr. Shastri says.