Dear Mihal,
A few suggestions:
1). Compare a few types of PPs: PSLibrary (QE defaults), GBRV
(https://www.physics.rutgers.edu/gbrv/), ATOMPAW
(http://users.wfu.edu/natalie/papers/pwpaw/PAWDatasets.html).
2). The cutoff depends on the PPs, I'd not go below the value
recommended for the PPs used.
3). Are there any compounds chemically similar to yours, with good known
structures?
If not - cell dimension is at least some quality indicator, good
dimensions mean good bond lengths, approximately.
Smooth EOS (energy-volume curve) is an indicator that your setup is
self-consistent (basis is good, etc).
4). Gamma-only is not good for geometries. I'd use 3x3x3 grid (if your
systems is approximately a cube), or 2x2x2 (yes, K_POINTS automatic 2 2
2 0 0 0 ) if 3x3x3 is too much.
Best regards.
Andrii Shyichuk, University of Wrocław
W dniu 2020-11-16 22:46, Husak Michal napisał:
Hi all ...
Before I continue I would like to mentioned I had done the homework on
this topic (checked mail list, studied about 20 related articles, done
a few hundred test calculations by QE and CASTEP) ...
The target of my work is to verify (enhance quality) of X-ray based
molecular crystal structure solution
by DFT methods ...
Target molecules: primary C,H,N,O composed middle size molecular
crystals (100-500 atoms per unit cell).
Insulators. Non magnetic. Other atoms potentially: F,Cl,S,B,Na,
Goal: Confirm the geometry from X-ray is correct (by results RMS
comparison with original) ..
Questions:
1) What pseudopotential is suitable ? pbe-n-rrkjus_psl.1.0.0.UPF Like
one ?
2) Some sources suggest 38 Ry cutf off (study done in VASP),
H-ultrasoft potential suggest 46 Ry cutoff,
QE related study (M. Lund) suggest 55 Ry cut off ...
I need to screen a lot of structures = be fast = low cut off . Opinion
?
3) How can i determine I had chosen correct parameters ?
Comparison of RMS with X-ray data is not reliable - X-ray shows
position of max. electron density,
DFT shows positions of cores (especially for H-atoms) ...
Any idea how to benchmark, before doing some large screening ?
Does exist some database of "100% reliable determined" reference
structures ?
4) Gamma K-point only usage will be nice.
M. Lund work suggest 2x2x2 k-point to be OK (is it identical to
K_POINTS automatic 2 2 2 0 0 0 ) ?
What do you thing ?
Thanks for any ideas
Michal Husak
UCT Prague
_______________________________________________
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list [email protected]
https://lists.quantum-espresso.org/mailman/listinfo/users
_______________________________________________
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list [email protected]
https://lists.quantum-espresso.org/mailman/listinfo/users
