Pseudooptimization?
Axel
akoh... at gmail.com
Tue Nov 17 20:56:20 UTC 2009
On Nov 17, 12:22 pm, Teodoro Laino <teodor... at gmail.com> wrote:
> Centering coordinates after each restart is slightly moving your system.
> This creates spurious forces (because of the presence of the grid ->
> see old messages) and the optimization may go on over forever..
>
> Take out the CENTER_COORDINATES and everything should go in the way you
> think it should.
this will get you consistently into the same minimum, but
perhaps, one should add to this, that because of theses "ripples"
there is a limit to the accuracy of the energy of a geometry
optimization. this is a limitation of using plane waves and
does not only affect CP2k with QS.
this doesn't matter for MD, but if you are after very small
differences, it should be taken under consideration
as well as the inherent other accuracy limitations (e.g.
using DFT, basis set, BSSE, and so on). at some point small
differences are just random noise.
cheers,
axel.
> Teo
>
> Ilya wrote:
> > Dear CP2K users.
>
> > I run into a problem few days ago.
>
> > I have performed a small (60 atoms) piece of hydrated quartz QS DFT
> > relaxation.
>
> > Cp2k could not find energy minimum at a time.
>
> > If I take optimized coordinates, put them in the input file again and
> > restart the optimization with the SCF_GUESS ATOMIC, cp2k will continue
> > optimization and will find some lower minimum.
>
> > The amount of opt steps needed to reach the minimum is much fewer. The
> > energy may differ by ~0.01 but in my case it really matters.
>
> > Launching 2-3 sequential optimizations I managed to find the exact
> > minimum value.
>
> > Anyway it looks suspicious. As it makes up a new wave function
> > approximation for an "optimized" coords the whole optimization goes in
> > a slightly different way each restart. As I got.
>
> > I tried to apply different optmizers - CG, BFGS, LBFGS. I also tried
> > to use OT, Mixing/Diag, changing SCF optimizers CG, DIIS. All those
> > influenced slightly. Still it does new optimization each restart,
> > instead of saing "nothing to do everything is optimized".
>
> > Now trying to change CUTOFF value, may be this could help.
> > The question might be stupid, I feel that the answer is simple, just
> > can't get it.
>
> > If someone have some ideas, I will appreciate you advice.
>
> > Input
>
> > &GLOBAL
> > PRINT_LEVEL LOW
> > PROJECT_NAME ps-t-11
> > RUN_TYPE GEO_OPT
> > PREFERRED_FFT_LIBRARY FFTW
> > &END GLOBAL
> > &MOTION
> > &GEO_OPT
> > OPTIMIZER CG
> > MAX_DR 3.00000000E-04
> > MAX_FORCE 4.50000000E-05
> > RMS_DR 1.50000000E-04
> > RMS_FORCE 3.00000000E-05
> > MAX_ITER 1000
> > &END GEO_OPT
> > &END MOTION
> > &FORCE_EVAL
> > METHOD QS
> > &DFT
> > BASIS_SET_FILE_NAME GTH_BASIS_SETS
> > &SCF
> > SCF_GUESS ATOMIC
> > MAX_SCF 1000
> > &END SCF
> > &XC
> > &XC_FUNCTIONAL PBE
> > &END XC_FUNCTIONAL
> > &END XC
> > &POISSON
> > PERIODIC XZ
> > POISSON_SOLVER WAVELET
> > &END POISSON
> > &END DFT
> > &SUBSYS
> > &CELL
> > ABC 9.8724 25.0 10.08094
> > UNITS ANGSTROM
> > PERIODIC XZ
> > &END CELL
> > &KIND Si
> > BASIS_SET TZVP-GTH
> > POTENTIAL GTH-PBE-q4
> > &END KIND
> > &KIND O
> > BASIS_SET TZVP-GTH
> > POTENTIAL GTH-PBE-q6
> > &END KIND
> > &KIND C
> > BASIS_SET TZVP-GTH
> > POTENTIAL GTH-PBE-q4
> > &END KIND
> > &KIND B
> > BASIS_SET TZVPP-GTH
> > POTENTIAL GTH-PBE-q3
> > &END KIND
> > &KIND H
> > BASIS_SET TZVP-GTH
> > POTENTIAL GTH-PBE-q1
> > &END KIND
> > &TOPOLOGY
> > COORD_FILE_FORMAT XYZ
> > COORD_FILE_NAME /home/kupchenko/job/ps/t/ps.xyz
> > CENTER_COORDINATES T
> > &END TOPOLOGY
> > &END SUBSYS
> > &END FORCE_EVAL
>
> > Coordinate file ps.xyz:
>
> > 60
> > i = 1, E = -570.7446060489
> > H 1.2164405235 15.4842136302 0.4143744831
> > H 6.1121896899 15.5038087647 0.4144334876
> > H 1.2707602262 15.3655789278 5.4048093021
> > H 6.1271079554 15.4026946050 5.4235580700
> > O 0.4551586659 15.1183862048 0.9207263709
> > O 5.3651001781 15.1154121407 0.9229189539
> > O 0.5077183637 15.1235599574 5.9915222988
> > O 5.3762635138 15.1158339064 6.0005346429
> > H 3.8570937184 15.1545145022 5.1023747623
> > H 8.8344418767 15.1584752804 5.0833260666
> > H 3.8786236331 15.1947700953 10.3189982581
> > H 8.8423420430 15.2102731172 10.2764809174
> > O 3.0957286169 14.9472184410 4.5045744980
> > O 8.0582595943 14.9554451515 4.5037182386
> > O 3.0983543419 14.9962628173 9.7165567981
> > O 8.0637432247 15.0292909929 9.6668423665
> > O 2.4682398721 13.9468761359 2.1126453081
> > O 7.4016859412 13.9535260868 2.1076981879
> > O 2.4331459358 13.9452444285 7.4094370497
> > O 7.3474209604 13.9487720010 7.3996943622
> > Si 0.9174823157 13.6452453328 1.6331964708
> > Si 5.8506776815 13.6460140438 1.6312922823
> > Si 0.8945731285 13.7654223514 6.8801727781
> > Si 5.8090768385 13.7639843442 6.8815442106
> > Si 3.3445442865 13.6624717732 3.4584593898
> > Si 8.2735277184 13.6625861281 3.4589630814
> > Si 3.3211393429 13.6848373669 8.7566224332
> > Si 8.2481557358 13.6840992310 8.7402594584
> > O 4.9261822520 13.4850120463 3.0058435744
> > O 9.8492612051 13.4823424700 2.9989183455
> > O 4.8578563490 13.4964580997 8.1955274091
> > O 9.7798690555 13.4911169759 8.1692673736
> > O 0.8797033120 12.3830946804 0.6001141356
> > O 5.7933420692 12.3831395817 0.6001632671
> > O 0.8817757579 12.3795984391 6.0019735587
> > O 5.7953154313 12.3796068132 6.0019844251
> > O 2.7873912363 12.2645536217 4.1156930416
> > O 7.7011498729 12.2647096129 4.1156209075
> > O 2.8124720845 12.2891736146 9.5066763280
> > O 7.7263380707 12.2894528459 9.5065558302
> > Si 1.8767330997 11.4234259421 -0.2249206571
> > Si 6.7904244881 11.4233716970 -0.2249530592
> > Si 1.8906925923 11.4346717365 5.1661113227
> > Si 6.8044045172 11.4346717256 5.1660918387
> > O 2.8683800441 10.6886833092 0.8130047883
> > O 7.7820554561 10.6886976079 0.8130092380
> > O 2.8661225434 10.6907181430 6.2154513092
> > O 7.7798155302 10.6907252061 6.2154563221
> > O 0.9993480464 10.3826593668 4.3304611630
> > O 5.9130553067 10.3826768100 4.3304808739
> > O 0.9879650764 10.3694140533 9.7308331231
> > O 5.9016086863 10.3694428080 9.7308465605
> > H 2.5220936594 9.9015909978 1.2556566675
> > H 7.4361354814 9.8984291273 1.2500346062
> > H 3.8152371223 10.6874649056 5.9934984006
> > H 8.7297691935 10.6932243995 5.9965260796
> > H 1.3674986990 9.4962228514 4.1919909263
> > H 6.2998681174 9.5107427878 4.1552631113
> > H 0.0234329886 10.3965414804 9.7770309728
> > H 4.9370093473 10.3979787214 9.7924189287
>
> > --
>
> > You received this message because you are subscribed to the Google Groups "cp2k" group.
> > To post to this group, send email to cp... at googlegroups.com.
> > For more options, visit this group athttp://groups.google.com/group/cp2k?hl=.
More information about the CP2K-user
mailing list