New AI framework solves century-old physics problem hundreds of times faster

According to Science Daily, scientists have developed an artificial intelligence framework capable of solving one of the most stubborn computational problems in physics, cutting calculation times from weeks to seconds without sacrificing accuracy.

The system, called THOR — Tensors for High-dimensional Object Representation — was developed by researchers at the University of New Mexico and Los Alamos National Laboratory. It tackles a class of calculations known as configurational integrals, which describe how atoms interact inside materials and are essential for predicting how substances behave under different physical conditions.

For more than a century, scientists have had to rely on indirect methods such as molecular dynamics and Monte Carlo simulations to work around the sheer complexity of these calculations. The core obstacle is what researchers call the "curse of dimensionality" — as the number of variables grows, the computational load balloons exponentially, pushing even the most powerful supercomputers to their limits and producing only approximate results after weeks of processing.

THOR sidesteps this by breaking the enormous mathematical problem into a sequence of smaller, interconnected pieces using a technique called tensor train cross interpolation. The framework also identifies symmetry patterns within crystal structures, further reducing the amount of computation required. When tested on copper, argon under extreme pressure, and a complex phase transition in tin, THOR reproduced results from advanced Los Alamos simulations while running more than 400 times faster.

The system also works alongside modern machine learning models that simulate atomic behaviour, making it adaptable across a wide range of materials and conditions.

Boian Alexandrov, the Los Alamos senior AI scientist who led the project, said the ability to accurately model thermodynamic behaviour had direct implications for fields ranging from metallurgy to the study of materials under extreme pressures.

"THOR AI opens the door to faster discoveries and a deeper understanding of materials."

The THOR Project is available on GitHub.