Energy-consistent pseudopotentials for group 11 and 12 atoms: adjustment to multi-configuration Dirac–Hartree–Fock data Detlev Figgen a , Guntram Rauhut a , Michael Dolg b , Hermann Stoll a, * a Institut fu¨ r Theoretische Chemie, Universita¨ t Stuttgart, D-70550 Stuttgart, Germany b Institut fu¨ r Theoretische Chemie, Universita¨t zu Ko¨ ln, D-50939 Ko¨ ln, Germany Received 19 May 2004; accepted 4 October 2004 Available online 6 November 2004 Abstract Two-component relativistic pseudopotentials (i.e., scalar-relativistic and spin–orbit (SO) potentials) of the energy-consistent vari- ety have been adjusted for the group 11 and 12 atoms Cu, Zn; Ag, Cd; Au, Hg, replacing the 1s–2p;1s–3d; and 1s–4f cores, respec- tively. The adjustment has been done for the valence-energy spectrum of (near-)neutral atoms, to reference data from numerical all-electron four-component multi-configuration Dirac–Hartree–Fock (MCDHF) calculations, including the two-electron Breit interaction. For use in molecular calculations, the potentials have been supplemented by energy-optimized (12s12p9d3f2g)/ [6s6p4d3f2g] valence basis sets. First benchmark applications of the potentials and basis sets are presented for atomic excitation energies and SO splittings at a correlated level, and for ground and excited state spectroscopic properties of group 11 monohalides and group 12 dimers. Ó 2004 Elsevier B.V. All rights reserved. 1. Introduction Pseudopotentials (PP) [or, synonymously, effective core potentials (ECP)] are used as standard tools nowa- days in valence ab initio quantum-chemical calculations [1,2]. They relieve such calculations from the burden of explicitly treating a plethora of chemically inactive core electrons for heavy atoms, they provide an efficient means of implicitly treating core-valence and valence scalar-relativistic effects in a formally non-relativistic framework, and with two-component extensions of the potentials even a fully relativistic treatment including spin–orbit effects can be simulated. Current demands on pseudopotentials can be charac- terized as follows: (i) with increasing availability of spin– orbit (SO) codes in quantum-chemical packages, requirement for SO extensions of the pseudopotentials will become more or less standard in the near future; (ii) with increasing feasibility of very accurate high-level valence correlation calculations, there is a growing num- ber of cases where errors of PPs with ÔchemicalÕ core sizes turn out to be non-negligible, cf., e.g., [3–6], and development of more accurate PPs with reduced core size (small-core PPs) will be needed therefore; and (iii) with the availability of series of one-particle basis sets allowing for a systematic improvement of accuracy in all-electron calculations (like DunningÕs series of corre- lation-consistent polarized valence n-tuple-zeta basis sets [7]), similar tools for PP calculations promise to be extremely useful. In our own work on energy-adjusted pseudopoten- tials (cf., e.g., [8] for a recent overview), we try to meet these demands as follows: (i) while we previously ad- justed scalar-relativistic pseudopotentials (average rela- tivistic PPs, ARPP) and SO potentials in separate steps – the ARPP to data from quasi-relativistic Wood–Boring (WB) [9] calculations for (near-) neutral 0301-0104/$ - see front matter Ó 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.chemphys.2004.10.005 * Corresponding author. E-mail address: stoll@theochem.uni-stuttgart.de (H. Stoll). www.elsevier.com/locate/chemphys Chemical Physics 311 (2005) 227–244