Soft and accurate norm conserving pseudopotentials and their application for structure prediction

Structure prediction and discovery of new materials are essential for the advancement of new technologies. This have been possible due to the developments in Density Functional Theory (DFT) and increase in computational power of the supercomputers. One of the key aspect is the reliability of the structures predicted by the DFT codes. In this regard pseudopotentials are essential for both fast and accurate predictions. Through the addition of softness constraints on the pseudo valence orbitals along with the non-linear core correction and semicore states, new soft and accurate dual space Gaussian type pseudopotentials have been generated for the Perdew Burke Ernzerhof (PBE) and PBE0 functionals. Despite being soft, these pseudopotentials were able to achieve chemical accuracy necessary for the production runs. These pseudopotentials have been benchmarked against the most accurate all-electron (μHa accuracy) reference data of molecular systems till date which has been obtained using the Multi-Wavelets as implemented in the MRCHEM. In addition the pseudopotentials for the PBE functional show remarkable accuracy in the Delta tests. These new soft and accurate pseudopotentials have been used for structure prediction of large clusters.