On the basis set selection for molecular core-level $GW$ calculations
The $GW$ approximation has been recently gaining popularity among the method for simulating molecular core-level X-ray photoemission spectra. Traditionally, $GW$ core-level binding energies have been computed using either the cc-pV$n$Z or def2-$n$ZVP basis set families, extrapolating the obtained results to the complete basis set limit, followed by a an element-specific relativistic correction. Despite of achieving good accuracy, these binding energies are chronically underestimated. By using first-row elements and standard techniques known to offer good cost-accuracy ratio in other theories, we show that the cc-pV$n$Z and def2-$n$ZVP families show large contraction errors and lead to unreliable complete basis set extrapolations. On the other hand, we demonstrate that uncontracted versions of these basis sets offer vastly improved convergence. Even faster convergence can be obtained using core-rich, property-optimized, basis sets families like pcSseg-$n$, pcJ-$n$ and ccX-$n$Z. Finally, we also show that the improvement over the core properties does not degrade the calculation of the valence excitations, and thus offer a balanced description of both core and valence regions.