[Pw_forum] Total energy vs PP and some beginner questions

Giovanni Cantele giovanni.cantele at spin.cnr.it
Fri Apr 10 12:50:19 CEST 2015

> On 10 Apr 2015, at 11:33, Konrad Gruszka <kgruszka at wip.pcz.pl> wrote:
> Dear Users,
> Recently while making some tests on simple systems I've found very 
> strange behaviour of pw.x. It is obvious that choosing different pseudo 
> potentials should give to some extent different total energies. Since 
> I'm a beginner,  I would therefore ask a few questions:
> 1. To what extent has the pseudopotentials of two different elements 
> must be the same? Can I use spin polarized together with none spin 
> polarized PP? Or can I for example use Fe.pbe-sp-van.upf with 
> Y.pbe-nsp-van.UPF? Or even combine for example Fe.pbe-sp-van.upf with 
> Y.pbe-mt_fhi.UPF ?
> 2. I have noticed that combining some PP together gives very different 
> total energies: for example for the same system first run with Fe and Y 
> with ",pbe-mt_fhi.UPF" gives total = -200 Ry and when using
> Fe.pbe-sp-van.upf with Y.pbe-nsp-van.upf gives total energy = -1333 Ry. 
> Why those two PP give so big difference?
> 3. The last one noob-like question:  While doing a 'relax' calc. only 
> the X and Y coordinates change during calculation, the Z always is such 
> as set on the begining. Is that normal? (or maybe it is only fortuity 
> for my atomic system?)
> The lecture of pwscf manual did not answer my questions. I regret to 
> note that the beginnings with QE are wery hard for novice...
> Thank you in advice for your kindly answer
> Konrad Gruszka, PhD.
> Czestochowa University of Technology, Poland
> -- dr inż. Konrad Gruszka Politechnika Częstochowska Instytut Fizyki
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum

actually, the lecture of the pwscf documentation does answer your questions:

1. I guess (but maybe I’m wrong!) that you are wondering whether it is possible to mix spin polarised and non spin polarised pseudo potentials because
you are mislead by the “sp “ and “nsp” found in the pseudo file name. As you can read here:
the “s”, “p” and “n” stand for
	• s: semicore state s in valence
	• p: semicore state p in valence
	• n: nonlinear core-correction
So you can mix the mentioned pseudos, provided you know the features of the pseudo you are using (norm conserving versus ultrasoft,
with versus without non linear core correction, ect.)

2. see here: http://www.quantum-espresso.org/faq/pseudopotentials/#2.4
as explained in the above link, mostly when two different pseudos of the same elements are built with a different number of electrons, the two total energies differ a lot. If you look carefully inside the pseudo files, you (not that hardly) discover that Y.pbe-mt_fhi.UPF has 3 valence electrons, with only 5s and 4d orbitals occupied,
    5s  5  0  2.00      0.000      0.000     0.000000
    5p  5  1  0.00      0.000      0.000     0.000000
    4d  4  2  1.00      0.000      0.000     0.000000
    4f  4  3  0.00      0.000      0.000     0.000000
whereas Y.pbe-nsp-van.UPF has 11 valence electrons with 4s and 4p orbitals occupied as well
Wavefunctions         nl  l   occ
                       4S  0  2.00
                       4P  1  6.00
                       4D  2  1.00
                       5S  0  2.00

3. see here: http://www.quantum-espresso.org/faq/self-consistency/#6.14
so it is hard to give the correct answer without your input file(s), but what is likely going on is that you system has a symmetry that forces the z component of the forces to zero.
Beware: when finding a symmetry, pw.x always keeps that symmetry during the run. How? For example, all the forces that, consistently with the found symmetry, should be zero, are NOT calculated but set to zero. A typical example is try to run a relax run for an H2O molecule. If you start with the following positions:
H    0.0    0.0   -1.0
O    0.0    0.0    0.0
H    0.0    0.0    1.0
then pw.x relaxes to
H        0.000000000   0.000000000  -0.945242812
O        0.000000000   0.000000000   0.000000000
H        0.000000000   0.000000000   0.945242812
because the molecule initially starts with a linear geometry (in this case along the z axis). The found symmetry is such that any force between atoms cannot have either an x or an y component, so pw.x is able (unless otherwise instructed) to output a relaxed structure constrained to lie along the z axis (which in this case is not the lowest energy one!).
But, introducing an even small perturbation to the linear geometry, such as
H    0.0    0.1   -1.0
O    0.0    0.0    0.0
H    0.0    0.1    1.0
the right experimental geometry is found:
H        0.000000000   0.264981368  -0.771676756
O        0.000000000  -0.329962735   0.000000000
H        0.000000000   0.264981368   0.771676756
(OH bond: 0.97 A, H-O-H angle: 104.7)

In the first case 16 symmetry operations are found, whereas in the second only 4!



Giovanni Cantele, PhD
c/o Dipartimento di Fisica
Universita' di Napoli "Federico II"
Complesso Universitario M. S. Angelo - Ed. 6
Via Cintia, I-80126, Napoli, Italy
e-mail: giovanni.cantele at spin.cnr.it
Phone: +39 081 676910
Skype contact: giocan74

ResearcherID: http://www.researcherid.com/rid/A-1951-2009
Web page: http://people.na.infn.it/~cantele

More information about the users mailing list