Standard Solid State Pseudopotentials (SSSP)
About the SSSP
The two libraries (SSSP Accuracy and SSSP Efficiency, lower and
upper triangle, respectively) cover for each element either the
best in class (among many libraries, or generated by us), or
occasionally allow for a compromise (either expensive and
accurate, or less expensive and a bit less accurate).
The convergence pattern of each element includes for each of the
considered family (represented by a color):
the total number of electrons in the valence, Z;
delta value (the error in the equation of state compared
with all-electron WIEN2k results, developed
by Cottenier
group), at full converged cutoff;
the largest phonon frequency, ωmax, at the
zone boundary (as a number), at full cutoff; and then as a
function of wave function cutoff:
the discrepancy of all phonon frequencies at the zone
boundary, with respect to the converged value (dashed lines indicate
± 1% errors);
the convergence of the formation energy of the solid (with
respect to the isolated atom), δHf, with
respect to the converged value;
the convergence of the pressure (trace of the stress divided
by 3), δP, with respect to the converged value.
The suggested wave function cutoff is indicated. All
normconserving psp have a dual of 4 (i.e. ecutrho=4*ecutwfc);
all ultrasoft/PAW have a dual of 8, with the
exception of
Mn and Fe, that have duals of 12. The tests have been made for
PBE pseudos using the Quantum ESPRESSO package.
Notes
The pressures converge very slowly with cutoff - so a
vc-relax at a cutoff where the error on the pressure is larger
than 1 kbar would be inaccurate. We are working at a simple
correction scheme for this.
We haven't looked at band structures/ghost states yet (we
are starting on this) - but as a worse case scenario these would be
most likely in the conduction. E.g. ~10 of the SG15 pseudos have ghost
states (not those we selected as best), although pslibrary and GBRV
should have been checked for those already by the authors.
All our scripts, inputs and outputs are available in the download page.
We have managed the calculations using AiiDA, an interface to automate, store and share calculations and results.
Selection Criteria for SSSP
|
Efficiency (upper triangle) |
Accuracy (lower triangle) |
| δHf |
< 3 meV |
< 3 meV |
| Δ |
< 1 meV (when possible) |
Smallest |
| Semi-core states |
Optional |
Included |
| Phonons convergence (w.r.t. 120 Ry cutoff) |
Within 1 % |
Within 1 % |
| Computational cost |
Cheap! |
Not too costly |
Computational Details
Structures: most stable elemental system. SiF4 has been
used for fluorine. For the rare-earth elements, the nitrides
have been considered.
Δ calculations:
- energy and wave function cutoffs: at least 150 % of the
value suggested in the pseudopotential file;
- k-points: 20x20x20.
- smearing: Marzari-Vanderbilt, 0.002 Ry;
- Mn: antiferrimagnetic;
- O and Cr: antiferromagnetic;
- Fe, Co, and Ni: ferromagnetic.
Phonon calculations:
SCF part:
- dual: 8 (PAW/US) and 4 (NC);
- k-points: 6x6x6; 10x10x10 (Oxygen);
- smearing: Marzari-Vanderbilt, 0.02 Ry;
- Non spin-polarized calculations.
PH part:
- q-point: (0.5, 0.5, 0.5).
Contact: Ivano E. Castelli
Theory and Simulation of Materials (THEOS) and
National Centre for Computational Design and Discovery of Novel
Materials (MARVEL),
École Polytechnique
Fédérale de Lausanne, CH 1015, Lausanne,
Switzerland.