X-ray photoelectron spectroscopy (XPS) is a powerful
experimental
technique for probing the electronic structure of molecules and materials;
however, interpreting XPS data requires accurate computational methods
to model core-ionized states. This work proposes and benchmarks a
new approach based on the generalized active space-driven similarity
renormalization group (GAS-DSRG) for calculating core-ionization energies
and treating correlation effects at the perturbative and nonperturbative
levels. We tested the GAS-DSRG across three data sets. First, the
vertical core-ionization energies of small molecules containing first-row
elements are evaluated. GAS-DSRG achieves mean absolute errors below
0.3 eV, which is comparable to high-level coupled cluster methods.
Next, the accuracy of GAS-DSRG is evaluated for larger organic molecules
using the CORE65 data set, with the DSRG-MRPT3 level yielding a mean
absolute error of only 0.34 eV for 65 core-ionization transitions.
Insights are provided into the treatment of static and dynamic correlation,
the importance of high-order perturbation theory, and notable differences
from density functional theory in the predicted energy ordering of
core-ionized states for specific molecules. Finally, vibrationally
resolved XPS spectra of diatomic molecules (CO, N2, and
O2) are simulated, showing excellent agreement with experimental
data.