next up previous contents
Next: 2 Step-by-step Pseudopotential generation Up: 1 Introduction Previous: 1.2 About similar work   Contents

1.3 Pseudopotential generation, in general

In the following I am assuming that the basic PP theory is known to the reader. Otherwise, see Refs.[1,4,7,8,9] and references quoted therein for NC-PP's; Refs.[2,3] for US-PP's and PAWsets. I am also assuming that the generated PP's are to be used in separable form [10] with a plane-wave (PW) basis set.

The PP generation is a three-step process. First, one generates atomic levels and orbitals with Density-functional theory (DFT). Second, from atomic results one generates the PP. Third, one checks whether the reesulting PP is actually working. If not, one tries again in a different way.

The first step is invariably done assuming a spherically symmetric self-consistent Hamiltonian, so that all elementary quantum mechanics results for the atom apply. The atomic state is defined by the "electronic configuration", one-electron states are defined by a principal quantum number and by the angular momentum and are obtained by solving a self-consistent radial Schrödinger-like (Kohn-Sham) equation.

The second step exists in many variants. One can generate ``traditional'' single-projector NC-PP's; multiple-projector US-PP's, or PAW sets. The crucial step is in all cases the generation of smooth ``pseudo-orbitals'' from atomic all-electron (AE) orbitals. Two popular pseudization methods are presently implemented: Troullier-Martins [7] and Rappe-Rabe-Kaxiras-Joannopoulos [9] (RRKJ).

The second and third steps are closer to cooking than to science. There is a large arbitrariness in the preceding step that one would like to exploit in order to get the "best" PP, but there is no well-defined way to do this. Moreover one is often forced to strike a compromise between transferability (thus accuracy) and hardness (i.e. computer time). These two steps are the main focus of these notes.


next up previous contents
Next: 2 Step-by-step Pseudopotential generation Up: 1 Introduction Previous: 1.2 About similar work   Contents
Filippo Spiga 2015-09-24