If a very accurate pseudopotential has to be created it is savest and simplest to create 2 projectors for each l-quantum number and chose the truncated AE-potential as local potential. As cutoff radius use half the nearest neighbor distance, and a relatively small value for the cutoff of the local potential. For Hg we have used the following reference potential:
LULTRA = T use ultrasoft PP ? RCLOC = 2.0 use i.e 2/3 of the radial cutoff Description l E TYP RCUT TYP RCUT 2 .000 7 2.800 23 2.900 2 -.790 7 2.800 23 2.900 0 .000 15 2.900 23 2.900 0 -.400 15 2.900 23 2.900 1 .000 15 2.900 23 2.900 1 -.400 15 2.900 23 2.900This PP is much better than for example a standard BHS pseudopotentials, and the convergence speed is also reasonable. To improve efficiency it is possible to increase the radial cutoffs for the US-part in our example up to 3.2 a.u., and that one of the normconserving part to 3.0 a.u., without loss of accuracy.
Second, it is not always necessary to include two projectors per l-quantum number, for instance there is no need to make the s and p-part ultrasoft for the transition metals, and first row elements do not require an accurate description of the d-electrons. Examples are given below.