Search
On July 27, the Tsinghua University News Network announced a significant breakthrough achieved by Xu Yong and Duan Wenhui's team from the Physics Department at Tsinghua University in the interdisciplinary realm of artificial intelligence and first-principles calculations. They have introduced a new deep learning Hamiltonian method, termed DeepH-R. This innovative approach shifts the focus of deep learning predictions from the Hamiltonian matrix, represented in the atomic orbital basis, to the real-space Kohn-Sham potential. Notably, this potential is rotationally invariant and independent of any specific basis, thereby streamlining the deep learning process in theory. Consequently, it promises a substantial enhancement in the accuracy of electronic structure predictions and bolsters the model's ability to generalize across different materials.
-
C114 Communication Network -
Communication Home
