Generation of Accurate and Transferable Pseudopotentials for Nuclear Magnetic Resonance

Markstein 101

Nuclear Magnetic Resonance (NMR) is a powerful method for investigating the structure of materials by measuring the response of an atom's nucleus to external magnetic fields. Computationally predicting NMR results with Density Functional Theory (DFT) based Gauge-Including Projector Augmented Wave (GIPAW) methods can enable a deeper understanding of the experimental results. The accuracy of these predictions depends strongly on the pseudopotential, a simplification of the core atomic orbitals. In this work, we generate a new series of pseudopotentials and compare their accuracy against reference all-electron calculations. Python scripts help test and optimize over 40 different input parameters for the atomic code of Andrea Dal Corso (Scuola Internazionale Superiore di Studi Avanzati, Trieste, Italy). The primary focus was PAW-type pseudopotentials where the greatest sensitivity was to the radius used for specific atomic orbitals and the electron occupation number of these orbitals. With these parameters carefully assigned, the other inputs can be used to fine tune the pseudopotential for accurate NMR-GIPAW calculations. We will comment on the accuracy of our pseudopotentials on select test materials as well as their transferability to additional systems.