In a breakthrough that redefines the boundaries of statistical physics, researchers at the University of New Mexico and Los Alamos National Laboratory have unveiled THOR (Tensors for High-dimensional Object Representation). This new artificial intelligence system has managed to solve—within seconds—a materials‑calculation problem that has kept some of the world’s most powerful supercomputers busy for weeks.
The innovation tackles an exceptionally rigid mathematical challenge: the evaluation of configurational integrals. These calculations are essential for predicting the thermodynamic behavior of materials and the interactions between atoms. The difficulty stems from what mathematicians call the curse of dimensionality. As the number of atoms increases, the complexity of the equation grows exponentially. A classical integration across thousands of dimensions would require more computing time than the age of the universe.
To break through this barrier, THOR AI abandons brute force and instead relies on tensor‑network algorithms and a technique known as tensor train cross interpolation. Rather than processing the entire multidimensional space at once, the algorithm compresses the equation into a sequence of smaller matrices and automatically detects the material’s crystallographic symmetries—drastically reducing computational load.
For decades, estimating these integrals meant depending on slow, repetitive indirect simulations such as Monte Carlo sampling or molecular dynamics. THOR AI removes the need for such approximations, enabling direct calculation from first principles.
Tested on materials including tin, copper, and high‑pressure argon, the system proved 400 times faster than state‑of‑the‑art traditional methods, all while maintaining exceptional accuracy.
By combining tensor networks with machine‑learning models that predict atomic motion, THOR emerges as a revolutionary tool. It promises to dramatically accelerate research and innovation in metallurgy, chemistry, and materials science—unlocking the ability to simulate matter with precision even under the most extreme physical conditions.
Sources:
- ScienceDaily – article „THOR AI solves a 100-year-old physics problem in seconds” (Publicat pe 15 martie 2026)
https://www.sciencedaily.com/releases/2026/03/260315004344.htm
The Original Scientific Paper:
- Journal: Physical Review Materials, 2025; Volumul 9 (8).
- Title: Breaking the curse of dimensionality: Solving configurational integrals for crystalline solids by tensor networks.
- Authors: Duc P. Truong, Benjamin Nebgen, Derek DeSantis, Dimiter N. Petsev, Kim Ø. Rasmussen, Boian S. Alexandrov.
- DOI: 10.1103/xrbw-xr49
Institutions:
- The University of New Mexico
- Los Alamos National Laboratory