US lab solves 100-year-old physics puzzle with new AI framework

Scientists at Los Alamos National Laboratory and the University of New Mexico have developed an AI framework called Tensors for High-dimensional Object Representation (THOR) that solves a century-old physics challenge: efficiently computing the configurational integral. This integral is fundamental for understanding particle interactions within materials, crucial for predicting properties like strength and stability under extreme conditions. THOR uses tensor-network mathematics to drastically reduce the computation time from weeks on supercomputers to mere hours or seconds, while maintaining high accuracy. This advancement enables more precise modeling of metals and crystals, particularly under high pressure and during phase transitions. THOR tackles the "curse of dimensionality" by decomposing complex, high-dimensional data into smaller, linked components, akin to reorganizing billions of Lego bricks into manageable chains. When combined with a custom interpolation algorithm, this tensor-train technique achieves speeds up to 400 times faster than traditional molecular dynamics simulations. Real-world tests on copper, argon, and tin demonstrated THOR’s ability to accurately reproduce therm