[Pw_forum] Run using a relaxed structure as initial one NOT converge in fir st few ionic steps.

Paolo Giannozzi p.giannozzi at gmail.com
Tue Dec 20 11:27:51 CET 2016


If forces are small: not a big deal. Note that "conv_thr = 1.0D-6" is
typically good to start a structural optimization, but the convergence
error on forces might be significant if forces are close to zero.

If forces are not small, there is something not right in your procedure.

Paolo

On Tue, Dec 20, 2016 at 11:05 AM, quantum <quantum at mail.ncyu.edu.tw> wrote:
> Dear Giovanni and Stefano:
>
> Thank you very much for your response.
>
> Listed below are the details of input and output.
>
> ----- The first input file --------------------------------------
> &control
>   title = 'Gypsum (CaSO4.2H20 Z=4)'
>   prefix = 'gyp-pbe-mt_fhi-exp1.01G-opt'
>   outdir = '/work2/u2equa02/tempdir/gyp-pbe-mt_fhi-exp1.01G-opt'
>   pseudo_dir = '/home/u2equa02/quantumhome/pseudo'
>   calculation = 'relax'
>   restart_mode = 'from_scratch'
>   disk_io = 'low'
>   verbosity = 'high'
>   wf_collect = .true.
>   tstress = .true.
>   tprnfor = .true.
>   etot_conv_thr = 1.0D-4
>   forc_conv_thr = 1.0D-3
>   max_seconds = 466560
> /
> &system
>   space_group = 15
>   uniqueb = .true.
>   origin_choice = 1
>   celldm(1) = 11.79378074341055 ! bohr
>   celldm(2) = 2.42429097900977
>   celldm(3) = 0.904502483576350
>   celldm(5) = -0.413422293216829
>   ntyp = 4
>   nat = 7
>   ecutwfc = 240
>   ecutrho = 960
>   vdw_corr = 'grimme-d2'
> /
> &electrons
>   electron_maxstep = 200
>   conv_thr = 1.0D-6
>   startingpot = 'atomic'
>   startingwfc = 'atomic+random'
>   mixing_mode = 'plain'
>   mixing_beta = 0.5
>   mixing_ndim= 8
>   diagonalization = 'david'
> /
> &ions
>   ion_dynamics = 'bfgs'
> /
>
> ATOMIC_SPECIES
>    Ca   40.0780000000  Ca.pbe-mt_fhi.UPF
>     S   32.0650000000  S.pbe-mt_fhi.UPF
>     O   15.9994000000  O.pbe-mt_fhi.UPF
>     H    1.0079000000  H.pbe-mt_fhi.UPF
>
> ATOMIC_POSITIONS crystal_sg
>     Ca  0.00000   0.16960   0.25000
>     S   0.00000   0.32680   0.75000
>     O   0.08400   0.27200   0.59000
>     O   0.20000   0.38200   0.91500
>     O  -0.21300   0.06300  -0.08000
>     H  -0.25800   0.08700  -0.23400
>     H  -0.24400   0.02000  -0.07700
>
> K_POINTS automatic
>     12 12 12 0 0 0
> ----------------------------------------------
>
> The last section of "ATOMIC_POSITIONS (crystal)" ----------------
>
> ATOMIC_POSITIONS (crystal)
> Ca       0.831507586   0.168492414   0.250000000
> Ca       0.168492414   0.831507586   0.750000000
> S        0.670838834   0.329161166   0.750000000
> S        0.329161166   0.670838834   0.250000000
> O        0.814789681   0.361192199   0.593221086
> O        0.638807801   0.185210319   0.906778914
> O        0.185210319   0.638807801   0.406778914
> O        0.361192199   0.814789681   0.093221086
> O        0.817560716   0.587016680   0.921747159
> O        0.412983320   0.182439284   0.578252841
> O        0.182439284   0.412983320   0.078252841
> O        0.587016680   0.817560716   0.421747159
> O        0.713851138   0.844837552   0.919788047
> O        0.155162448   0.286148862   0.580211953
> O        0.286148862   0.155162448   0.080211953
> O        0.844837552   0.713851138   0.419788047
> H        0.650318830   0.820912679   0.741084444
> H        0.179087321   0.349681170   0.758915556
> H        0.349681170   0.179087321   0.258915556
> H        0.820912679   0.650318830   0.241084444
> H        0.749428151   0.753450611   0.911642228
> H        0.246549389   0.250571849   0.588357772
> H        0.250571849   0.246549389   0.088357772
> H        0.753450611   0.749428151   0.411642228
>
> ----------------------------------------------------------------
>
>
> I understand, when using space_group to specify atomic positions
> a conversion between coordinates of convectional and centered cells
> is performed.  (Thanks to Andrea Dal Corso)
>
> So, I manually reverted the conversion, following the PW code below.
> ============================================================
>  espresso-5.3.0/PW/src/wyckoff.f90:
>
>        SUBROUTINE ccord(outco,sym_n,not_eq,ibrav,unique)
>
>        CASE (-13) !Centrato B unique MONOCLINO
>        DO k=1,not_eq
>           DO i=1,sym_n
>              tmpx=outco(1,i,k)
>              tmpy=outco(2,i,k)
>
>              outco(1,i,k)=tmpx-tmpy
>              outco(2,i,k)=tmpy+tmpx
>           END DO
>        END DO
> ===========================================================
>
> After manually reverting the above process, I identify the positions of the 7 inequivelent atoms, and use them for the new run.
>
> The new input file --------------------------------------------------------
> &control
>   title = 'Gypsum (CaSO4.2H20 Z=4)'
>   prefix = 'gyp-pbe-mt_fhi-exp1.01G-opt-cont'
>   outdir = '/work2/u2equa02/tempdir/gyp-pbe-mt_fhi-exp1.01G-opt-cont'
>   pseudo_dir = '/home/u2equa02/quantumhome/pseudo'
>   calculation = 'relax'
>   restart_mode = 'from_scratch'
>   disk_io = 'low'
>   verbosity = 'high'
>   wf_collect = .true.
>   tstress = .true.
>   tprnfor = .true.
>   etot_conv_thr = 1.0D-4
>   forc_conv_thr = 1.0D-3
> /
> &system
>   space_group = 15
>   uniqueb = .true.
>   origin_choice = 1
>   celldm(1) = 11.79378074341055 ! bohr
>   celldm(2) = 2.42429097900977
>   celldm(3) = 0.904502483576350
>   celldm(5) = -0.413422293216829
>   ntyp = 4
>   nat = 7
>   ecutwfc = 240
>   ecutrho = 960
>   vdw_corr = 'grimme-d2'
> /
> &electrons
>   electron_maxstep = 200
>   conv_thr = 1.0D-6
>   startingpot = 'atomic'
>   startingwfc = 'atomic+random'
>   mixing_mode = 'plain'
>   mixing_beta = 0.5
>   mixing_ndim= 8
>   diagonalization = 'david'
> /
> &ions
>   ion_dynamics = 'bfgs'
> /
>
> ATOMIC_SPECIES
>    Ca   40.0780000000  Ca.pbe-mt_fhi.UPF
>     S   32.0650000000  S.pbe-mt_fhi.UPF
>     O   15.9994000000  O.pbe-mt_fhi.UPF
>     H    1.0079000000  H.pbe-mt_fhi.UPF
>
> ATOMIC_POSITIONS crystal_sg
>   Ca       0.000000000   0.168492414   0.250000000
>   S        0.000000000   0.329161166   0.750000000
>   O        0.087990940   0.273201259   0.593221086
>   O        0.202288698   0.384727982   0.921747159
>   O       -0.220655655   0.065493207  -0.080211953
>   H       -0.264384246   0.085296924  -0.258915556
>   H       -0.248560619   0.002011230  -0.088357772
>
>
> K_POINTS automatic
>     12 12 12 0 0 0
>
> ------------------------------------------------------------
>
> The new run does not converge after an ionic step.
>
>
>
> Sincerely,
> Tsung-Lung Li
>
>
>
>
>
>
>
>
> -----Original message-----
> From:stefano de gironcoli <degironc at sissa.it>
> To:pw_forum at pwscf.org
> Date:Tue, 20 Dec 2016 17:34:17
> Subject:Re: [Pw_forum] Run using a relaxed structure as initial one NOT converge in fir st few ionic steps.
>
> dear Tsung-Lung Li,
>
> Your expectations are correct.
> If this is what you experience you probably are not doing what you think
> you are doing but it is difficult to make any suggestion without further
> information
>
> stefano
>
>
> On 20/12/2016 10:26, quantum wrote:
>> Dear QE Friends:
>>
>> I did a "relax" calculation with PW (version 5.3.0)
>> and obtained the relaxed atomic positions.
>>
>> Then, I took the atomic positions
>> in the last "ATOMIC_POSITIONS (crystal)"
>> section of the output file to replace the
>> atomic positions in the original input file
>> without changing any other things except new
>> prefix and outdir.
>>
>> The atoms in the relaxed structure should
>> all satisfy etot_conv_thr and forc_conv_thr
>> thresholds.  With this in mind, I started
>> the new run.
>>
>> I expected the run to
>> converge after the first or two ionic steps.
>> But it did not stop, and continued running
>> for many ionic steps without convergence.
>>
>> Is my expectation correct? or it is a problem
>> of QE?
>>
>> Sincerely,
>> Tsung-Lung Li
>>
>>
>> Tsung-Lung Li, Ph. D.
>> Professor
>> Department of Applied Physics
>> National Chia-Yi University
>> 300 Hsueh-Fu Road, Chiayi 60004, Taiwan
>> Phone: 886-5-2717904.  FAX: 886-5-2717909.
>> E-mail:quantum at mail.ncyu.edu.tw
>> URL:http://web.ncyu.edu.tw/~quantum
>>
>> _______________________________________________
>> Pw_forum mailing list
>> Pw_forum at pwscf.org
>> http://pwscf.org/mailman/listinfo/pw_forum
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>
>
> Tsung-Lung Li, Ph. D.
> Professor
> Department of Applied Physics
> National Chia-Yi University
> 300 Hsueh-Fu Road, Chiayi 60004, Taiwan
> Phone: 886-5-2717904.  FAX: 886-5-2717909.
> E-mail:quantum at mail.ncyu.edu.tw
> URL:http://web.ncyu.edu.tw/~quantum
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum



-- 
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



More information about the users mailing list