[QE-users] Structure optimization using rvv10-scan

Giovani Rech gio.pi.rech at gmail.com
Thu Jun 6 20:39:59 CEST 2019


Hello all,

I just wanted to give you an update on this matter in case anyone stumble
upon the same issue in the future.

The problem with the calculation using the SCAN+rVV10 functional was in the
pseudo-potential that I was using. Previously I was using a PP built using
PBE. When I changed to a PP built using  SCAN, the calculation of pressure
as well as the structure optimization seemed to work well.

Here is a relevant paper on this: Yao, Y. and Kanai, Y., 2017. Plane-wave
pseudopotential implementation and performance of SCAN meta-GGA
exchange-correlation functional for extended systems. The Journal of
chemical physics, 146(22), p.22410

And, finally, a few pseudo-potentials built using SCAN can be found here:
https://yaoyi92.github.io/scan-tm-pseudopotentials.html

Best regards,
Giovani Rech
Universidade de Caxias do Sul

On Thu, Jun 6, 2019 at 1:04 PM Giuseppe Mattioli <
giuseppe.mattioli at ism.cnr.it> wrote:

>
> Dear José
> I do not know about rvv10, but spin-polarized SCAN (and at least plus
> dft-d2 for sure) is implemented in pw.x AFAIK. In my tests it was
> quite stable but as slow as EXX, anyway...
> HTH
> Giuseppe
>
>
> José Carlos Conesa <jcconesa at icp.csic.es> ha scritto:
>
> > Hi,
> >
> > Are there plans to implement in qe any meta-GGA (or at least
> > rvv10-scan) for the spin-polarized case?
> >
> > José Carlos
> >
> > El 26/04/2019 a las 22:12, Paolo Giannozzi escribió:
> >> Correcting myself: stress for meta-GGA is implemented, but only in
> >> the spin-unpolarized case
> >>
> >> Paolo
> >>
> >> On Thu, Apr 25, 2019 at 8:48 AM Paolo Giannozzi
> >> <p.giannozzi at gmail.com <mailto:p.giannozzi at gmail.com>> wrote:
> >>
> >>    I am not sure that the calculation of stress is implemented with
> >>    meta-GGA.
> >>
> >>    SCAN behaves better than other meta-GGA, but still it is
> >>    numerically unstable. See for instance here:
> >>    https://gitlab.com/QEF/q-e/issues/32. Before trying difficult
> >>    calculations with SCAN you should verify whether you can do simple
> >>    ones.
> >>
> >>    Paolo
> >>
> >>
> >>    On Sat, Apr 20, 2019 at 3:33 AM Giovani Rech
> >>    <gio.pi.rech at gmail.com <mailto:gio.pi.rech at gmail.com>> wrote:
> >>
> >>        Hello all,
> >>
> >>        Have anyone tried structure optimization using rvv10-scan?
> >>
> >>        I'm trying to optimize a structure (graphite) at 0.0 kbar
> >>        taking into account van der Waals interactions. For such, I'm
> >>        using the SCAN+rVV10 by setting "input_dft = 'rvv10-scan'".
> >>        What I'm getting as a result makes no sense, with unreasonable
> >>        pressures. Here's a plot of the pressure and volume as a
> >>        function of optimization step:
> >>        image.png
> >>
> >>        When I got this values I was using version 6.4.0 and then
> >>        tried again with 6.3 and finally with the latest version,
> >>        6.4.1, and got the same values (plotted above). Here's the
> >>        input that I used:
> >>
> >>                &CONTROL
> >>                                 title = "graphite_rvv10_vcrelax" ,
> >>                           calculation = 'vc-relax' ,
> >>                          restart_mode = "from_scratch" ,
> >>                                outdir = "./" ,
> >>                            pseudo_dir = "/home/giovani/graphite/pseudo"
> ,
> >>                                prefix = "gC" ,
> >>                               disk_io = 'default' ,
> >>                             verbosity = 'default' ,
> >>                         etot_conv_thr = 1.0D-4 ,
> >>                         forc_conv_thr = 1.0D-3 ,
> >>                                 nstep = 400 ,
> >>                               tstress = .true. ,
> >>                               tprnfor = .true. ,
> >>                 /
> >>                 &SYSTEM
> >>                                     A = 2.47000e+00 ,
> >>                                     C = 8.68000e+00 ,
> >>                                   nat = 4,
> >>                                  ntyp = 1,
> >>                               ecutwfc = 80 ,
> >>                               ecutrho = 320 ,
> >>                             input_dft = 'rvv10-scan' ,
> >>                                 ibrav = 4 ,
> >>                 /
> >>                 &ELECTRONS
> >>                      electron_maxstep = 200,
> >>                              conv_thr = 1.00000e-06 ,
> >>                           startingpot = "atomic" ,
> >>                           startingwfc = 'atomic' ,
> >>                           mixing_mode = "plain" ,
> >>                           mixing_beta = 7.00000e-01 ,
> >>                           mixing_ndim = 8,
> >>                       diagonalization = 'david' ,
> >>                        diago_thr_init = 1e-4 ,
> >>                 /
> >>                 &IONS
> >>                          ion_dynamics = 'bfgs' ,
> >>                         ion_positions = 'from_input' ,
> >>                               upscale = 100 ,
> >>                      trust_radius_max = 1.0D-3 ,
> >>                 /
> >>                 &CELL
> >>                         cell_dynamics = 'bfgs' ,
> >>                                 press = 0.0 ,
> >>                        press_conv_thr = 0.05 ,
> >>                           cell_factor = 1.2 ,
> >>                 /
> >>                ATOMIC_SPECIES
> >>                C  12.0107 C.SR.ONCVPSP.PBEsol.stringent.upf
> >>                ATOMIC_POSITIONS crystal
> >>                C 0.0000000000 0.0000000000 0.000000000
> >>                C 1/3          2/3          0.000000000
> >>                C 1/3          2/3          1/2
> >>                C 2/3          1/3          1/2
> >>                K_POINTS automatic
> >>                  6 6 2   0 0 0         I then tried the same
> >> optimization using PBE, by just
> >>        commenting the 'input_dft' line, and got values of both
> >>        pressure and volume converging to fairly reasonable values (as
> >>        plotted below) which makes me think that the problem might be
> >>        with the rVV10-scan option. Have anyone else had this kind of
> >>        problem? Any ideas on how this could be fixed?
> >>        image.png
> >>
> >>        Also, when testing and comparing the results of both
> >>        approaches with verbosity=high to investigate which
> >>        contribution to the pressure was wack, I noticed that almost
> >>        all the pressure matrices were more or less similar, except
> >>        for 'exc-cor stress', that was of the same order of magnitude
> >>        but opposite signs, and 'core-core stress', which was off both
> >>        in magnitude and in sign. I'm not sure if this is relevant to
> >>        the problem, but I thought it could help in finding a solution.
> >>
> >>        Thank you for your attention,
> >>        Best regards,
> >>        Giovani Rech
> >>
> >>        Universidade de Caxias do Sul,
> >>        Caxias do Sul - RS, Brazil
> >>
> >>        _______________________________________________
> >>        Quantum Espresso is supported by MaX
> >>        (www.max-centre.eu/quantum-espresso
> >>        <http://www.max-centre.eu/quantum-espresso>)
> >>        users mailing list users at lists.quantum-espresso.org
> >>        <mailto:users at lists.quantum-espresso.org>
> >>        https://lists.quantum-espresso.org/mailman/listinfo/users
> >>
> >>
> >>
> >>    --     Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e
> Fisiche,
> >>    Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
> >>    Phone +39-0432-558216, fax +39-0432-558222
> >>
> >>
> >>
> >> --
> >> Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
> >> Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
> >> Phone +39-0432-558216, fax +39-0432-558222
> >>
> >>
> >> _______________________________________________
> >> Quantum Espresso is supported by MaX (
> www.max-centre.eu/quantum-espresso)
> >> users mailing list users at lists.quantum-espresso.org
> >> https://lists.quantum-espresso.org/mailman/listinfo/users
> >
> > --
> > José C. Conesa
> > Instituto de Catálisis y Petroleoquímica, CSIC
> > Marie Curie 2, Madrid, Spain
> > www.icp.csic.es
> > Tel. (+34)915854766
>
>
>
> GIUSEPPE MATTIOLI
> CNR - ISTITUTO DI STRUTTURA DELLA MATERIA
> Via Salaria Km 29,300 - C.P. 10
> I-00015 - Monterotondo Scalo (RM)
> Mob (*preferred*) +39 373 7305625
> Tel + 39 06 90672342 - Fax +39 06 90672316
> E-mail: <giuseppe.mattioli at ism.cnr.it>
>
> _______________________________________________
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso)
> users mailing list users at lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20190606/8c02c80c/attachment.html>


More information about the users mailing list