Choosing pseudopotentials

For many elements several pseudopotential choices exist. The usual tradeoff between computational cost and accuracy and transferability applies. The choice of pseudopotential variant is not always straight forward. For the PBE generated latest PAW potential set (potpaw_PBE) we provide some recommendations for calculations involving a significant number of states far above the Fermi energy, and for those that are mainly evolved with states below or around the Fermi level.

Step-by-step instructions

Step 1: Select a pseudopotential release.

Generally we recommend to use the latest release of pseudopotentials (currently potpaw.64), but for consistency reasons or to accurately reproduce another calculation, you might need to use another release. Please consult the Construction:Lists_of_pseudopotentials for a list of all available potentials.

Step 2: Analyze your structure.

Your pseudopotential choice depends on the type of structure you have and the species it contains. Not all variants are available for all elements. E.g. there are no _GW potentials for the lanthanides. Bond lengths are also critical.

Tip: If dimers with short bonds are present in the compound (O2, CO, N2, F2, P2, S2, Cl2), we recommend to use the _h potentials. Specifically, C_h, O_h, N_h, F_h, P_h, S_h, Cl_h.

Step 3: Be sure what you want to calculate.

If you are interested in a quick and rough structure optimization only, soft potentials with minimal valency may be sufficient. The same can be true for phonon calculations, which often require very large supercells. However, when carefully optimizing a magnetic structure, it might be necessary to include a lot of semicore states in the valence, and optical properties should be calculated with potentials that are optimized for the description of empty states.

STEP 4: Consider the calculation method you will be using.

For any calculation involving unoccupied states significantly above the Fermi energy, the _GW variants of potentials are superior and should be used. This is especially true for all kinds of many-body perturbation calculations which need a high number of empty bands, but also "plain" DFT, where a higher number of empty bands are considered. On the other hand, most