High-efficiency atomic orbital data for first-principles calculations of large atomic systems

This dataset focuses on constructing high-efficiency and high-accuracy numerical atomic orbital basis set data for first-principles calculations, suitable for simulating atomic and molecular structures in large-scale systems (with thousands of atoms or more). Two different optimization algorithms were employed in this construction, using wave function data calculated by the plane-wave basis as reference states, to generate atomic orbital data suitable for three sets of norm-conserving pseudopotentials across 68 elements. To meet the requirements for different computational speeds and accuracies, we adopted a hierarchical optimization approach, starting with low-precision orbitals and gradually improving them to higher precision. This approach enabled the batch generation of data with various basis set sizes. In addition, systematic tests were conducted to ensure data quality, using metrics such as equations of state and bond length-energy relationship curves. These tests verified the accuracy differences between the generated atomic orbital basis set data and conventional plane-wave basis set data. The verification results show that the average error between the results obtained using the constructed atomic orbital basis set and the standard plane-wave basis set is as low as 0.66 meV/atom. The dataset has been widely utilized in the ABACUS program, demonstrating high reusability. It plays an important role in fields such as condensed matter physics, materials science, and chemical simulations, greatly enhancing the efficiency and accuracy of complex system simulations.