An AI framework now solves complex physics equations in seconds, transforming how scientists study material behavior. Researchers at the University of New Mexico and Los Alamos National Laboratory developed the Tensors for High-dimensional Object Representation (THOR) AI system, which tackles a problem that has stumped statistical physicists for decades.
THOR combines tensor network algorithms with machine learning to efficiently compress and analyze massive configurational integrals and partial differential equations—key for predicting how materials respond to various thermodynamic and mechanical conditions. This approach enables accurate, scalable simulations across diverse physical environments.
“Evaluating the configurational integral, which describes particle interactions, is notoriously slow, especially under extreme pressures or phase transitions,” said Boian Alexandrov, senior AI scientist at Los Alamos. “THOR allows us to precisely model thermodynamic behavior, advancing statistical mechanics and critical applications like metallurgy.”
Overcoming the Limits of Classical Simulations
Scientists have long relied on approximate methods like molecular dynamics and Monte Carlo simulations to estimate the configurational integral. These techniques mimic atomic motion over time to bypass the “curse of dimensionality,” where each added variable exponentially increases computational demands—often overwhelming even the fastest supercomputers. Despite weeks of intensive processing, these simulations yield limited results.
Read More: 20 Unstoppable Productivity Hacks to Supercharge Your Success in 2025
Dimiter Petsev, a UNM chemical engineering professor and frequent collaborator with Boian Alexandrov, saw the potential of the new AI-based methods. Alexandrov’s team’s strategies enable direct solutions to the configurational integral, a feat once deemed impossible.
“Directly solving the configurational integral has been considered unfeasible, as it can involve thousands of dimensions,” Petsev explained. “Classical methods would take longer than the age of the universe. Tensor network approaches set a new benchmark for accuracy and efficiency.”
Fast and Accurate Computation with THOR AI
THOR AI tackles high-dimensional physics problems by breaking the integrand’s massive data cube into smaller, connected components using “tensor train cross interpolation.” A custom variant identifies key crystal symmetries, allowing configurational integrals to be solved in seconds rather than thousands of hours—without sacrificing accuracy.
Tested on metals like copper, high-pressure noble gases such as crystalline argon, and tin’s solid-solid phase transition, THOR AI matches the results of Los Alamos’ top simulations while running over 400 times faster. Its compatibility with modern machine learning atomic models makes it a versatile tool for materials science, physics, and chemistry.
“This breakthrough replaces century-old approximations with first-principles calculations,” said Duc Truong, Los Alamos scientist and lead author of the study in Physical Review Materials. “THOR AI accelerates discoveries and deepens our understanding of materials.”
Frequently Asked Questions
What is THOR AI?
THOR AI is an advanced computational framework that solves complex physics equations, enabling accurate and fast simulations of materials.
How does THOR AI work?
It uses tensor train cross interpolation and machine learning to break high-dimensional problems into smaller, solvable components.
What problems can THOR AI solve?
THOR AI calculates configurational integrals and partial differential equations for metals, gases, and phase transitions, previously impossible with classical methods.
How fast is THOR AI compared to traditional simulations?
It runs simulations over 400 times faster than classical approaches while maintaining high accuracy.
Which materials has THOR AI been tested on?
It has been applied to metals like copper, noble gases such as argon, and tin’s solid-solid phase transitions.
Why is this breakthrough significant for materials science?
THOR AI replaces centuries-old approximations, providing first-principles calculations that accelerate discoveries and improve understanding of material behavior.
Who developed THOR AI?
Researchers from the University of New Mexico and Los Alamos National Laboratory, led by Boian Alexandrov and collaborators, developed THOR AI.
Conclusion
THOR AI represents a transformative leap in computational physics, solving long-standing challenges in material simulations with unprecedented speed and accuracy. By combining tensor network methods with machine learning, it replaces centuries-old approximations with first-principles calculations, enabling researchers to study metals, gases, and phase transitions more efficiently than ever. This breakthrough not only accelerates discoveries in materials science but also deepens our understanding of fundamental physics, opening the door to innovations across chemistry, engineering, and technology.