[Pw_forum] Mo pseudopotential

Ary Junior aryjunior at gmail.com
Tue Mar 26 22:38:01 CET 2013


Dear all,

I was looking for a PBE USPP for molybdenum and checked that the .UPF files
available in the QE website are the ones provided in pslibrary 0.2... So,
based in the input file I created a test for the
Mo.pbe-spn-rrkjus_psl.0.2.UPF file like this:

 &input
   title='Mo',
   zed=42.0,
   rel=1,
   config='[Kr] 4d5.0 5s1 5p0',
   iswitch=2,
   dft='PBE',
   rlderiv=2.93, eminld=-3.0, emaxld=3.0, deld=0.01, nld=3
 /
 &test
   file_pseudo='Mo.pbe-spn-rrkjus_psl.0.2.UPF',
   nconf=1,
   configts(1)='4d5.0 5s1 5p0',
   ecutmin=50, ecutmax=200, decut=50
 /

Besides some warnings after the "All-electron run" section, the difference
between the AE and the PS eigenvalues are zero and it is ok...

     n l     nl             e AE (Ry)        e PS (Ry)    De AE-PS (Ry)
     3 2     4D   1( 5.00)       -0.27582       -0.27582        0.00000
     1 0     5S   1( 1.00)       -0.30033       -0.30033        0.00000
     2 1     5P   1( 0.00)       -0.07669       -0.07669        0.00000

However, when compare these eigenvalues with an atomic calculation using
the spherical Bessel functions I got a weird result, even with a high
cutoff energy:

     Cutoff (Ry) :  200.0
                           N = 1       N = 2       N = 3
     E(L=0) =       -23.1157 Ry  -13.3019 Ry   -9.0812 Ry
     E(L=1) =       -20.7569 Ry  -11.9179 Ry   -7.7609 Ry
     E(L=2) =       -17.0812 Ry   -9.9089 Ry   -6.4930 Ry

Moreover, the AE vs PS logarithmic derivatives are quite discrepant, even
at lower r values... The point is that after analyzing the shape of the AE
orbitals I couldn't understand the so low rcut values in the pslibrary
input file in both versions, 0.2 and 0.3... I have done the same test for
other PBE USPPs available in the QE website like
Ni.pbe-n-rrkjus_psl.0.1.UPF, O.pbe-n-rrkjus_psl.0.1.UPF and
S.pbe-n-rrkjus_psl.0.1.UPF and they are fine... So, I tried to reset the
input file with radii more compatible with the AE wavefunctions in my
conception, besides change the core corrections like this:

 &input
   title='Mo',
   prefix='Mo',
   zed=42.0,
   config='[Kr] 5s1 4d5 5p0',
   iswitch=3,
   dft='PBE',
   rlderiv=2.93, eminld=-3.0, emaxld=3.0, deld=0.01, nld=3,
 /
 &inputp
   pseudotype=3,
   file_pseudopw='Mo.PBE-spn-rrkjus_aryjr.UPF',
   author='ADC',
   lloc=-1,
   rcloc=3.2,
   which_augfun='PSQ',
   rmatch_augfun=1.3,
   nlcc=.true.,
   new_core_ps=.true.,
   rcore=1.0,
   tm=.true.
 /
6
5S  1  0  1.00  0.00  2.70  3.30  0.0
5S  1  0  0.00  7.00  2.50  3.30  0.0
5P  2  1  0.00  0.00  3.20  3.30  0.0
5P  2  1  0.00  7.00  3.20  3.30  0.0
4D  3  2  5.00  0.00  1.50  1.70  0.0
4D  3  2  0.00 -0.30  0.70  1.70  0.0

Then, I think that everything is fine with the AE and the PS eigenvalues:

     n l     nl             e AE (Ry)        e PS (Ry)    De AE-PS (Ry)
     1 0     5S   1( 1.00)       -0.30033       -0.30033        0.00000
     3 2     4D   1( 5.00)       -0.27582       -0.27582        0.00000
     2 1     5P   1( 0.00)       -0.07669       -0.07669       -0.00000

And after compare these eigenvalues with an atomic calculation using the
spherical Bessel functions again, things look better now:

     Cutoff (Ry) :   50.0
                           N = 1       N = 2       N = 3
     E(L=0) =        -0.3002 Ry   -0.0021 Ry    0.0380 Ry
     E(L=1) =        -0.0766 Ry    0.0213 Ry    0.0681 Ry
     E(L=2) =        -0.2681 Ry    0.0313 Ry    0.0692 Ry

     Cutoff (Ry) :  200.0
                           N = 1       N = 2       N = 3
     E(L=0) =        -0.3002 Ry   -0.0021 Ry    0.0380 Ry
     E(L=1) =        -0.0766 Ry    0.0213 Ry    0.0681 Ry
     E(L=2) =        -0.2683 Ry    0.0313 Ry    0.0692 Ry

The AE vs PS logarithmic derivatives are being well reproduced in energies
lower than 1 Ry.... I also tested chemical states like the ionized Mo4+ and
Mo6+, and the results are listed below:

     1 0     5S   1( 1.00)       -0.30033       -0.30033        0.00000
     3 2     4D   1( 5.00)       -0.27582       -0.27582        0.00000
     2 1     5P   1( 0.00)       -0.07669       -0.07669       -0.00000
     Etot =   -8097.243530 Ry,   -4048.621765 Ha, -110168.600949 eV
     Etotps =   -32.753518 Ry,     -16.376759 Ha,    -445.634282 eV

     1 0     5S   1( 0.00)       -2.69065       -2.78173        0.09108
     3 2     4D   1( 2.00)       -3.40343       -3.51010        0.10667
     2 1     5P   1( 0.00)       -2.17762       -2.27343        0.09581
     dEtot_ae =       6.511981 Ry
     dEtot_ps =       6.625031 Ry,   Delta E=      -0.113050 Ry

     1 0     5S   1( 0.00)       -4.30796       -4.58873        0.28077
     3 2     4D   1( 0.00)       -5.60739       -5.87530        0.26791
     2 1     5P   1( 0.00)       -3.66423       -3.99180        0.32756
     dEtot_ae =      15.474602 Ry
     dEtot_ps =      15.954369 Ry,   Delta E=      -0.479767 Ry

My question: is my procedure described above wrong? Study and understand
the generation of PSPs is not easy for me and I'm trying to learn it
grounded in some textbooks as well as some documents spread on the internet
and of course the input files provided in the pslibrary project. But this
PBE USPP for molybdenum made me a little bit confused. I tested both in a
MoS2 unit cell with PBE-D and the cell volume as well as the interatomic
distances are fine if compared with the experiments, with differences lower
than 3%... One more question: how can I improve the result for the Mo4+ and
Mo6+ ionized states?

Thank you very much!

Ary Junior

Doctoral Student
Chemistry Department
Universidade Federal de Juiz de Fora - MG - Brasil

-- 
http://lattes.cnpq.br/8221674673413336
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20130326/e8657765/attachment.html>


More information about the users mailing list