A symmetry-preserving and transferable representation for learning the Kohn-Sham density matrix

Key Points

• The aim is to develop a representation that preserves symmetry and is transferable for learning the Kohn-Sham density matrix.
• Developed a new representation approach for the Kohn-Sham density matrix.
• Utilized density functional theory (DFT) principles in the modeling.
• Assured symmetry preservation in the representation.
• Demonstrated significant improvements in representing Kohn-Sham density matrix properties.
• Revealed that the new representation can be applied across different physical scenarios.