Microsoft Research has integrated its Skala AI model into the open-source CP2K software, bringing high-accuracy simulations to large molecular systems.
The AI-based exchange-correlation functional Skala (developed by Microsoft Research AI for Science) has been natively integrated into the open-source quantum chemistry and solid-state physics simulation platform CP2K.
The integration was built through a joint effort initiated in early 2026 between Microsoft Research AI for Science and the CP2K team at the Centre for Advanced Systems Understanding (CASUS) at Helmholtz-Zentrum Dresden-Rossendorf (HZDR).
Results of the integration and numerical verification were published in a joint mid-August 2026 arXiv preprint titled ‘Molecular Implementation of the Machine-Learned Skala Exchange-Correlation Functional in CP2K through GauXC’.
Upcoming Skala releases in CP2K will expand support to periodic solids (such as metals and semiconductors) and liquids.
Microsoft Research has integrated its Skala AI model into the open-source CP2K software, bringing high-accuracy simulations to large molecular systems.
The AI-based exchange-correlation functional Skala (developed by Microsoft Research AI for Science) has been natively integrated into the open-source quantum chemistry and solid-state physics simulation platform CP2K. The integration was built through a joint effort initiated in early 2026 between Microsoft Research AI for Science and the CP2K team at the Centre for Advanced Systems Understanding (CASUS) at Helmholtz-Zentrum Dresden-Rossendorf (HZDR).
Unlike traditional mathematical density functional theory (DFT) approximations, Skala uses a neural network trained to model how electron densities in different atomic regions influence one another. Skala achieves a higher level of accuracy for molecular system simulations, outperforming traditional meta-GGA and hybrid functionals on benchmarks at a fraction of the computational cost.
Combined with CP2K’s parallel processing capabilities, the integration enables accurate quantum-mechanical simulations of dynamic systems containing thousands to tens of thousands of atoms (for example, proteins, battery materials, semiconductors and catalysts).
Results of the integration and numerical verification were published in a joint mid-August 2026 arXiv preprint titled ‘Molecular Implementation of the Machine-Learned Skala Exchange-Correlation Functional in CP2K through GauXC’. The teams created a comprehensive suite of numerical verification tests. Upcoming Skala releases in CP2K will expand support to periodic solids (such as metals and semiconductors) and liquids.
