Performance of relativistic effective core potentials in DFT calculations on actinide compounds.

Density functional theory (DFT) calculations using relativistic effective core potentials (RECPs) have emerged as a robust and fast method of calculating the structural parameters and energy changes of the thermochemical reactions of actinide complexes. A comparative investigation of the performance of the Stuttgart small-core and large-core RECPs in DFT calculations has been carried out. The vibrational frequencies and reaction enthalpy changes of several uranium(VI) compounds computed using these RECPs were compared to those obtained using DFT and a four-component one-electron scalar relativistic approximation of the full Dirac equation with large all-electron basis sets (AE). The relativistic AE method is a full solution of the Dirac equation with all spin components separated out. This method gives the "correct" answer (with respect to scalar relativity) which should be closest to experimental values when an adequate density functional is used and in the absence of significant spin-orbit effects. The small-core RECP always show better agreement with the four-component scalar- relativistic AE method than the large-core RECP. We conclude that the 5s, 5p, and 5d orbitals are of great importance in determining the chemistry of actinide complexes. Instances in which large-core RECPs give better agreement with experimental data are attributed to either experimental uncertainties or error cancellations.