Description

According to Soraya, an international group of physicists used artificial intelligence to compress a highly complex quantum problem involving more than 100,000 equations into an equation that only needs to solve four equations.Interestingly, this compression did not change the accuracy of the result and could help revolutionize research systems in the field of quantum physics.

Hubbard Model

The Hubbard model was first presented in 1963 and tries to explain the behavior of electrons when placed on a grid. According to this model, when two electrons occupy a place in the grid, they interact and their destiny is intertwined as quantum mechanics, even if they are far apart.

Studying the behavior of electrons helps physicists explain the different states and stages of matter. However, since electrons are entangled in terms of quantum mechanics, physicists should consider all electrons together in their calculations. This turns calculations into a complex mathematical challenge in which the higher the number of electrons in question, the harder the calculations become exponentially harder.

To simplify this, physicists used a mathematical device called a "re-normalization group" that could help track all electron interactions. A re-normalization group could eventually contain between tens of thousands and millions of equations that need to be solved.

Using artificial intelligence to simplify

Di Sante and his colleagues wondered whether artificial intelligence could be used to simplify the issue. They turned to neural networks, where the software first made connections between the re-normalization group and then changed the strength of those connections to find a small set of equations that would create the same solution as the original group.

The program needed a lot of computational power to understand the complexity of the Hubbard model and ran for weeks, but its final output summed up hubbard's model in only four equations.

"It's basically a machine that has the power to explore hidden patterns," Di Sante said. When we saw the result, we said, "Wow, that's more than we expected." We were really able to understand and summarize the relevant physics.

Now that the app is trained to search for such patterns, it can be adapted to view other similar problems without having to start from scratch.

If this program can be scaled up for other problems, scientists would like to use it to design materials that provide superconductivity, where electrons pass through matter without any resistance.

In addition, Di Sante and his colleagues are now exploring how machine learning works in this sample to provide insight into how it works and what physicists have lost so far.

The findings are published in the journal Physical Review Letters.