Hyperfine Coupling Constants in Local Exact Two-Component Theory

We present a highly efficient implementation of the electron-nucleus hyperfine coupling matrix within one-electron exact two-component (X2C) theory. The complete derivative of the X2C Hamiltonian is formed, i.e. the derivatives of the unitary decoupling transformation are considered. This requires solution of the response and Sylvester equations, consequently increasing the computational costs. Therefore, we apply the diagonal local approximation to the unitary decoupling transformation (DLU). The finite nucleus model is employed for both the scalar potential and the vector potential. Two-electron picture-change effects are modeled with the (modified) screened-nuclear spin--orbit approach. Our implementation is fully integral direct and OpenMP-parallelized. An extensive benchmark study regarding the Hamiltonian, the basis set, and the density functional approximation is carried out for a set of 12--17 transition-metal compounds. The error introduced by DLU is negligible and the DLU-X2C Hamiltonian accurately reproduces its four-component ``fully'' relativistic parent results. Functionals with a large amount of Hartree--Fock exchange such as CAM-QTP-02 and omega-B97X-D are generally favorable. The pure density functional r2SCAN performs remarkably and even outperforms the common hybrid functionals TPSSh and CAM-B3LYP. Fully uncontracted basis sets or contracted quadruple-zeta bases are required for accurate results. The capability of our implementation is demonstrated for [Pt(C6Cl5)4]- with more than 4700 primitive basis functions and four rare-earth single molecule magnets: [La(OAr*)3]-, [Lu(NR2)3]-, [Lu(OAr*)3]-, and [TbPc2]-. Here, the spin--orbit DLU-X2C Hamiltonian results in an excellent agreement with the experimental findings of all Pt, La, Lu, and Tb molecules.