Microsoft AI's new electronic structure computation framework M-OFDFT debuts in Nature subissue with both accuracy and efficiency

Microsoft Research has developed a new AI-based electronic structure computation framework, M-OFDFT, which improves the computational efficiency of electronic structure methods while maintaining the comparable computational accuracy of KSDFT. M-OFDFT utilizes orbital-free density flooding theory and deep learning models to significantly improve the accuracy and efficiency of electronic structure computation, especially when dealing with large-scale molecular systems, and demonstrates excellent extrapolation performance. The successful application of M-OFDFT on small molecules and protein systems verifies its accuracy and efficiency, demonstrating the potential of AI in advancing the development of electronic structure computation and opening up new research directions and application scenarios. (ScienceAI)