A scarcity of data is a major issue in scientific research. If you want to use machine learning, it seems almost impossible because the ML technology relies heavily on large data to produce significant output.
So, the question that arises is, how do we use AI effectively to support their investigations?
With this question in mind, a team of researchers has already started work on finding a solution for a field called condensed matter physics. This field studies how materials behave, especially weird or complex materials.
The area of interest
The team was specifically interested in frustrated magnets, where the magnetic parts of the material don’t line up individually. These magnets tend to behave unusually, and this could pave the way for humans to understand things like quantum computers.
However, a major roadblock in this study is that frustrated magnets are difficult to simulate. This condition arises due to constraints arising from the way magnetic ions interact with each other.
Dissecting the problem
The researchers were particularly curious about what happens to a type of magnet when it gets super cold, almost down to absolute zero.
This magnet enters a strange state called a spin fluid at low temperatures, where the magnetic parts constantly change. It is similar to how molecules in liquid water move around. However, the researchers hit a dead end as they couldn’t determine what the spin liquid turns into when it freezes further.
“Recently, physicists have been excited about a type of quantum spin liquid which could help us to understand fault-tolerant quantum computers,” explained Professor Nic Shannon, Head of the Theory of Quantum Matter Unit at the Okinawa Institute of Science and Technology (OIST), and co-author on this study.
“In 2020, we realized that this spin liquid could occur naturally in a class of magnetic materials called ‘breathing pyroclores’. But we couldn’t figure out what happened to that spin liquid at low temperatures,” he continued.
A quantum task for AI
The researchers from OIST teamed up with ML experts from LMU Munich, who had developed an ML algorithm that could classify conventional magnetic orders.
“Our method is highly interpretable, meaning it’s easy for humans to decipher the decision-making processes, and doesn’t rely on prior training of the model. This makes it better suited for such applications where data is limited, compared to other forms of machine learning,” said Professor Lode Pollet of LMU Munich, a co-author on this study.
“Before we teamed up with OIST, we had never applied it to a spin liquid, so we were excited to see if it could be useful in gaining insights into such difficult physics problems where all other approaches had failed,” he explained further.
How did they do it?
The team used Monte Carlo simulations to model their spin liquid cooling. They ran the simulation data through the ML algorithm and spotted patterns they had missed earlier.
In the next step, the team used this data to run the Monte Carlo simulations in reverse and ended up heating the unknown magnetic state instead of cooling it. This activity helped confirm the state of the magnet, solving a puzzle they couldn’t figure out before.
“What was interesting is that neither man nor machine alone were able to solve this problem—it was more like colleagues collaborating, with the algorithm spotting something we hadn’t, and vice versa, building together towards this complete picture of understanding,” added Dr. Ludovic Jaubert of CNRS, University of Bordeaux.
This experiment shows that humans and AI have their strengths, but neither can solve every problem alone. Breakthroughs can be achieved only by combining human insight with AI pattern recognition in certain areas.