By Mark Lowey
Is it possible to measure something very precisely – an atomic clock driven by oscillating electrons, for example – when we live in a quantum world of uncertainty?
University of Calgary physicist Barry Sanders and his research team employed artificial intelligence to devise a strategy that avoids limits of measurement imposed by the quantum world.
“The main ‘message’ in my work is that artificial intelligence methods work for quantum-related problems and work very well,” says Sanders, professor of Physics and Astronomy, iCORE Chair of Quantum Information Science, and director of the university’s Institute for Quantum Science and Technology.
He and his research team, which included post-doctoral researcher Neil Lovett and European undergraduate students Cécile Crosnier and Martí Perarnau-Llobet, had their new approach published as the cover story in Physical Review Letters, a top-ranked physics journal.
Search for measurement certainty
Quantum physics is a branch of science that deals with discrete, indivisible units of energy called “quanta” as described by the Quantum Theory.
We live in a quantum world. Or, as Sanders puts it: “The universe is quantum.”
His team’s new paper involves “metrology,” the science of measurement. More specifically, the work involves quantum metrology: making the best use of quantum properties to measure things as precisely as possible, and to surpass the quantum limit imposed by the “uncertainty principle.”
This principle says that in the quantum world, it is impossible to measure both the momentum and the position of a particle perfectly precisely. So there is a limit or an uncertainty to measurement.
Artificial intelligence provides way around human limitation
To “beat the limit,” Sanders and his team used an approach called adaptive quantum metrology. The challenge is that this approach requires measurement, feedback and control sequences that are beyond human ability to devise.
The team turned to artificial intelligence for a solution. Using western Canada’s WestGrid supercomputer system, they ‘taught’ the machine how to use each measurement of a particle in order to measure the next particle, and to adapt this succession of measurements to be “infinitely precise.”
This “reinforcement learning” method is similar to how a computer learns to play chess, Sanders explains. “Each time it plays, it is effectively rewarded or punished, depending on whether it wins or loses. And it then gets better and better.”
Laying the groundwork for later discoveries
The work may lead to ultra-precise atomic clocks, extremely accurate Global Positioning System instruments or a way to measure “gravitational waves” – ripples in the curvature of spacetime for which there’s indirect evidence but which have yet to be detected.
“This research will matter down the road,” Sanders says. “As usual in physics (the discovery of the laser, for example), the most important implications are unforeseeable.”