Atomic relativistic approximation for <i>ab initio</i> prediction of excited-state potential energy surfaces and X-ray spectra

The atomic relativistic approach is a formally simple and linear scaling ansatz that exploits the locality of the relativistic effect. Studies have shown that this approximation introduces moderate error on ground state absolute energy in the presence of short bonds between heavy nuclei. However, scientists have not investigated its accuracy on computing excited states for absorption spectra and excited-state potential energy surfaces. In this work, we demonstrate atomic exact two-component (X2C) predictions of the L_2,3-edge X-ray absorption spectra of five representative heavy-atom-centered molecules and the excited-state potential energy curves of the platinum dimer (Pt₂). Surprisingly, not only the errors to full X2C results in core excitation energies are negligible on the order of 0.01 eV, but the oscillator strengths also agree well with those from full X2C computations. Meanwhile, the atomic X2C potential energy curves and crossings of Pt₂ are almost indistinguishable from the full X2C ones.