SandboxAQ’s peer-reviewed AQCat25 research, published in npj Computational Materials, brings the magnetic behavior of catalysts into large-scale AI modeling and is now available publicly to researchers.
PALO ALTO, Calif., June 16, 2026 /PRNewswire/ — SandboxAQ has published peer-reviewed research in npj Computational Materials, a Nature Portfolio journal, introducing a machine learning approach that lets industries simulate, screen, and optimize catalysts with physics-based accuracy. The research addresses a long-standing gap in computational chemistry: the treatment of magnetism in the earth-abundant metals that drive industrial catalysis.
At the center of the work is AQCat25, a dataset of 13.5 million density functional theory calculations spanning 47,000 catalyst systems. It is the first large-scale catalysis dataset to incorporate spin polarization, and it was generated using roughly 400,000 GPU-hours on NVIDIA DGX Cloud.
“Proud of the team for their new paper in the peer-reviewed journal Nature Computational Materials,” SandboxAQ CEO Jack Hidary wrote on LinkedIn. “The SandboxAQ team presents a breakthrough in catalyst discovery and computational chemistry.”
Catalysts underpin more than 80 percent of manufactured goods, including fertilizers, fuels, and chemicals. Many depend on iron, cobalt, and nickel, whose magnetic behavior strongly affects how molecules bind to a surface. Because those effects were costly to compute, earlier datasets often omitted them, which limited accuracy on industrially important materials. AQCat25 closes that gap. Researchers can explore material discovery and request access at sandboxaq.com.
“Catalysis drives the global economy, from the fuels that power our world to the materials that shape it,” Hidary wrote. “With our AQCat model, industries can now simulate, screen, and optimize catalysts with physics-based accuracy, unlocking performance and sustainability breakthroughs at unprecedented scale.”
Key highlights:
- 13.5 million DFT calculations across 47,000 catalyst systems
- Spin polarization for 12 magnetic elements, plus 6 new elements: barium, cerium, fluorine, lithium, lanthanum, and magnesium
- Up to 20,000 times faster than first-principles simulation, according to SandboxAQ, making high-throughput virtual screening practical
- Dataset and models released publicly on Hugging Face under a Creative Commons license
About SandboxAQ
SandboxAQ delivers solutions at the intersection of AI and quantum techniques. The company’s Large Quantitative Models (LQMs) apply physics-based AI to deliver advances in life sciences, chemicals and materials, financial services, navigation, and cybersecurity. Learn more at sandboxaq.com.
The full peer-reviewed paper is available in npj Computational Materials.
View original content to download multimedia:https://www.prnewswire.com/news-releases/sandboxaq-led-by-ceo-jack-hidary-publishes-spin-aware-catalysis-ai-research-302802130.html
SOURCE SandboxAQ
