From NWChem
Viewed 130 times, With a total of 2 Posts

Just Got Here
Threads 1
Posts 2


2:13:00 AM PDT  Tue, Apr 30th 2019 

Dear NWChem developers and users:
I have been using NWChem 6.6 to perform computations on organometallic compounds using B3LYP/Lanl2DZ/6311+G* for a couple of months now but I keep running in geometry convergence issues, i.e. I end up with one or two imaginary frequencies. I know this topic has been discussed numerous times in this forum but all solutions that seemed to work for the other users did not work for me. First of all here is my input file
start ProductCompound+Pd_ActivNewTry
title "ProductCompound+Pd_Activ in 6311+g* basis set"
scratch_dir ../../../../var/scratch/user
memory total 8000 mb
charge 1
geometry units angstroms print xyz autosym
C 0.84169 4.67838 0.78686
C 1.65397 3.38660 0.96403
H 1.71396 3.11420 2.02404
H 2.68067 3.53152 0.62818
C 0.81108 5.06104 0.70418
H 1.82691 5.25146 1.06680
H 0.24681 5.98894 0.84871
C 0.59057 4.45649 1.29716
H 1.18915 5.36596 1.17793
H 0.58450 4.22342 2.36829
C 1.03586 2.19919 0.15858
C 0.16400 3.91760 1.50747
H 0.13525 4.18115 2.57003
C 1.23527 3.30783 0.50083
H 2.25590 3.13864 0.86420
C 1.00137 2.63356 1.33562
C 0.40263 1.99950 0.73850
C 1.26997 3.69073 0.99150
H 2.02947 2.81257 1.66299
H 1.85682 4.60796 1.10790
H 1.78896 2.93546 1.59039
H 0.26814 1.91138 1.82192
H 0.62254 1.83117 1.97726
H 1.32011 5.47621 1.36413
C 2.03374 1.02677 0.30303
N 1.65887 0.28979 0.39970
O 3.23714 1.27466 0.35105
C 2.68492 1.24806 0.05181
Pd 0.18915 1.02023 0.40189
C 0.19606 3.93530 0.45343
C 0.93663 5.05841 0.77117
N 0.77952 2.74485 0.21089
H 3.49750 1.29078 0.68187
H 3.14971 0.89072 0.97802
C 2.32714 4.94388 0.83541
C 2.11951 2.61368 0.27843
C 2.92018 3.71491 0.58740
H 0.43566 5.99866 0.96547
H 2.93777 5.80587 1.07986
H 3.99642 3.59503 0.63422
H 0.88556 3.96331 0.39216
C 1.23672 0.78444 0.35938
C 1.99822 0.17545 1.42220
C 1.56525 0.36838 0.96422
C 2.90352 0.83525 1.21601
C 2.42046 0.72670 1.18037
C 3.09822 1.33314 0.10341
H 1.83795 0.53558 2.43275
H 1.14869 0.86934 1.82439
H 3.44415 1.25444 2.05408
H 2.63506 1.06956 2.18525
O 3.91464 2.33607 0.41531
C 4.68995 2.99342 0.60493
H 5.37414 2.28919 1.08103
H 4.03828 3.45909 1.34701
H 5.25736 3.75752 0.08147
end
basis
H library 6311+G*
O library 6311+G*
N library 6311+G*
C library 6311+G*
Pd library lanl2dz_ecp
end
ecp
Pd library lanl2dz_ecp
end
stepper
maxiter 100
end
scf
thresh 1.0 d8
end
dft
xc b3lyp
grid fine
tolerances tight
convergence energy 1.0d8
iterations 100
end
task dft optimize
task dft freq
This was my last attempt to reach convergence, without success. Before I also tried
1. the "default" driver module with settings tight
2. did computations with and without "autosym"
3. changing the structure randomly to approach the minimum from different directions
4. use the normal vectors of the imaginary frequencies to change the structure along these normal vectors
All attempts without success and I always end up with imaginary frequencies (about 60 cm^1). Does someone has another idea what to try and could this also come from an incorrect compilation of the NWChem code and corresponding libraries?
Best regards,
StarkEffect




Forum Regular
Threads 56
Posts 268


7:41:33 PM PDT  Tue, Apr 30th 2019 

The following reference from Tang, et al. may be helpful but it might be unsuitable in your case.
The Electrolyte Genome project: A big data approach in battery materials discovery.
COmputational Materials Science. 103(2015) 5667
"The results, shown in Fig. 4, demonstrate that if the frequency is very low (<39.0 cm1), a highaccuracy frequency calculation can remove up to 42% of the imaginary frequencies. However, high accuracy reoptimization is less helpful for larger imaginary frequencies. In contrast, the two molecular geometry perturbation strategies work extremely well: they remove at least 21 out of the 24 imaginary frequencies for the whole range of imaginary frequencies."

Edited On 7:43:12 PM PDT  Tue, Apr 30th 2019 by Xiongyan21




Just Got Here
Threads 1
Posts 2


3:29:19 AM PDT  Thu, Jun 13th 2019 

Dear Xiongyan21 and NWChem developers:
Thanks for your reply. Unfortunetely, I was unable to resolve the problem posted above.
Additionally, I tested other quantum chemistry codes with their implementation of the same level of theory but in these codes the geometry optimization and subsequent frequency computations revealed that the structure has converged and is a local minimum.
To double check I asked a friend to run the same computation on another machine with the newest NWChem version. The computation also failed.
Therefore, my followup question: is anything known about numerical inaccuracies for some special cases? Are there ways to check for probelms or even resolve these?
Best regards,
StarkEffect



AWC's:
2.5.10 MediaWiki  Stand Alone Forum Extension
Forum theme style by: AWC