Machine learning is helping UQ researchers control one of nature's strangest states of matter. By optimising how an ultracold quantum fluid is stirred, the team demonstrated a more efficient way to generate persistent currents, a key capability for future quantum sensors, atomtronic devices and other emerging quantum technologies.
These systems can detect tiny changes in gravity, motion and magnetic fields, opening new possibilities for navigation, resource exploration and scientific research.
What's the problem?
Quantum technologies promise breakthroughs in sensing, computing and measurement, but controlling quantum systems remains extremely challenging. Bose-Einstein condensates, a type of quantum fluid, display unusual behaviours such as persistent currents, where atoms flow without friction around a ring-shaped system. Generating these states reliably typically requires researchers to optimise many experimental parameters through labour-intensive trial and error.
As quantum devices become more complex, finding efficient ways to control and optimise these systems becomes increasingly important.
What's UQ's solution?
Researchers from UQ's School of Mathematics and Physics, including Dr Simeon Simjanovski, Associate Professor Tyler Neely, Professor Halina Rubinsztein-Dunlop, , Dr Guillaume Gauthier and Professor Matthew Davis, used machine learning to experimentally control the stirring of a Bose-Einstein condensate and optimise the generation of persistent currents.
The team employed a Gaussian process learner that continually assessed experimental outcomes and adjusted the stirring protocol to achieve specific objectives. The machine-learning system was guided by the strength of the persistent current generated and the number of unwanted vortices introduced during the process.
Rather than relying solely on manual optimisation, the AI system autonomously explored the experimental parameter space and identified effective strategies for producing desired quantum states. The work demonstrated that machine learning can successfully optimise a complex quantum experiment in the laboratory.
What's the impact?
The study showed that persistent currents can be reliably generated through a variety of stirring approaches, highlighting the robustness of the method. By reducing the need for repeated trial-and-error experimentation, machine learning could help researchers develop and control quantum systems more efficiently.
The findings have potential applications in emerging quantum technologies, including atomtronic devices, matter-wave interferometers and quantum sensors that rely on precise control of ultracold quantum matter.
More broadly, the work demonstrates how AI can help scientists optimise complex experiments and accelerate discovery in cutting-edge areas of physics.
These systems can detect tiny changes in gravity, motion and magnetic fields, opening new possibilities for navigation, resource exploration and scientific research.
Did you know?
Atomtronic technologies aim to harness the quantum properties of matter for enhanced sensing, extending beyond what can be achieved with conventional electronics. Ring-trapped superfluids represent a key geometry for atomtronics and can be used for rotation sensing.
Led by
This work brings together researchers from UQ's School of Mathematics and Physics and the ARC Centre of Excellence for Engineered Quantum Systems (EQUS).
UQ Research Team
- Dr Tyler Neely
- Professor Halina Rubinsztein-Dunlop
- Dr Simeon Simjanovski
- Dr Guillaume Gauthier
- Professor Matthew Davis
| AI Research Strength | Scalable and Sustainable AI |
|---|---|
| Industry Portfolio | Advancing AI |
| Key Publications | Simjanovski, S., Gauthier, G., Davis, M.J., Rubinsztein-Dunlop, H. and Neely, T.W. (2023). Optimizing persistent currents in a ring-shaped Bose-Einstein condensate using machine learning. Physical Review A, 108(6), 063306. https://doi.org/10.1103/PhysRevA.108.063306. |