Description
This computational exercise is based on our previously published work https://doi.org/10.1016/S0009-2614(99)00665-X.
We evaluated the first ionization potential (IP) of the mercury atom using various software packages from the conda-forge channel, including MOPAC, xTB, NWChem, and PySCF.
Python scripts were developed within the Atomic Simulation Environment (ASE) with assistance from online AI tools.
The experimental ionization energy of mercury is 10.437 eV.
MOPAC’s semi-empirical PM7 method (with a UHF reference) produces an accurate value of 10.518 eV, showcasing robust parametrization for heavy elements like Hg. In contrast, the GFN1-xTB extended tight-binding semi-empirical DFT method yields a severe overestimation at 16.768 eV.
We subsequently performed ab initio calculations based on single-determinant ECP-SCF wave functions, utilizing the effective core potential (ECP) from the DK scalar relativistic treatment described in the reference study.
Using the valence def2-SVP basis set, our ECP-ROHF calculation yields an IP of 8.529 eV. This is in close alignment with the DK-SCF value of 8.576 eV reported in the reference study, thereby validating our computational protocol.
Correlated ECP-UHF-CCSD(T) calculations in NWChem demonstrate a systematic improvement with expanding basis set size, yielding 10.018 eV with def2-SVP, 10.139 eV with def2-TZVP, and a peak value of 10.325 eV with def2-QZVP.
In comparison, the PySCF ECP-UHF-CCSD(T)/def2-TZVP calculation—leveraging density fitting and efficient integral screening—produces a value of 10.125 eV.
All working files are available at
https://github.com/miroi/Molecular-and-Materials-Modeling-2026-August/tree/main/molecular-modeling/abstract_ionization_potentials.