TAMkin with CP2K - the wavenumbres (cm-1) in the cp2k output file are different from the TAMkin wavenumbers

56 views
Skip to first unread message

Torstein Fjermestad

unread,
May 8, 2020, 9:43:19 AM5/8/20
to TAMkin
Dear all, 

I am trying to use TAMkin with CP2K. 
For the time being, I use no special treatment. The idea is just to check if I can get the same result with TAMkin as with CP2K. 

I made a CP2K version of the script shown here https://molmod.github.io/tamkin/tutorial/getting_started.html, see attached py file. 
However, the frequencies (in cm-1) that TAMkin produces are different form those that are found in the CP2K output file.
A comparison of the eight first frequencies are shown here:

TAMkin: 31.1    38.3    47.9    74.7    82.0   111.0   223.7   245.6
CP2K:   27.3     40.8    51.0    74.0    80.0   109.9  224.1   246.3

What could be the cause of this difference?
The Python script and the CP2K single point and frequency calculation are attached.

Thanks a lot for your help. 

Regards,
Torstein Fjermestad

cp2k-tamkin.zip

Toon Verstraelen

unread,
May 11, 2020, 8:13:44 AM5/11/20
to tam...@googlegroups.com
Dear Torstein,

There could be two reasons: (i) CP2K prins out Hessian information in limited precision, which may affect the low frequencies and (ii) the constraints imposed in the vibrational analysis my differ. E.g. one code may project out external rotations and translations, while another one may not. On your tamkin script, this is not done.

Best regards,

Toon

--
You received this message because you are subscribed to the Google Groups "TAMkin" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tamkin+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/tamkin/774838ff-ae52-488a-a329-cbfe6a6cfef6%40googlegroups.com.


--
Prof. Dr. ir. Toon Verstraelen
Center for Molecular Modeling
Technologiepark-Zwijnaarde 46 (before Jan 1, 2019, this was building 903)
Ghent University
Tech Lane Ghent Science Park Campus A
9052 Zwijnaarde
Belgium
How to reach my office? See https://molmod.ugent.be/where-find-us (2nd floor, CMM, door next to Xerox)
E-mail: Toon.Ver...@UGent.be


Torstein Fjermestad

unread,
May 13, 2020, 5:28:04 AM5/13/20
to TAMkin
Dear Toon,

I have now looked into the two possible reasons you mention:

(i) CP2K prints the Hessian with six digits after the decimal point. An example of the first few lines is here:

 VIB| Hessian in cartesian coordinates

                    1            2            3            4            5
     1  H       28.709095    -1.146315    -1.020252     0.009869     0.011226
     2  H       -1.138645    42.763219     3.995350     0.124385    -0.141326
     3  H       -1.012371     4.015084    50.855425     0.093292    -0.117246
     4  H        0.010317     0.119539     0.089114    59.156790   -26.388589
     5  H        0.010004    -0.120816    -0.097750   -26.392370    42.890280

Do you suspect this is not sufficiently accurate? 
I have not found out whether it is possible to control the precision with which the Hessian is printed, but I could ask the CP2K forum.

(ii) The vibrational analysis in CP2K is a Partial Hessian calculation of a periodic zeolite structure. Only the adsorbate + the Brønsted proton are being moved. There are therefore no external rotations or translations to project out. 

May there be other causes of the discrepancy between CP2K and TAMkin?

Thanks a lot for your help.
Regards,
Torstein
  

mandag 11. mai 2020 14.13.44 UTC+2 skrev Toon Verstraelen følgende:
Dear Torstein,

There could be two reasons: (i) CP2K prins out Hessian information in limited precision, which may affect the low frequencies and (ii) the constraints imposed in the vibrational analysis my differ. E.g. one code may project out external rotations and translations, while another one may not. On your tamkin script, this is not done.

Best regards,

Toon

On Fri, May 8, 2020 at 3:43 PM Torstein Fjermestad <tfjer...@gmail.com> wrote:
Dear all, 

I am trying to use TAMkin with CP2K. 
For the time being, I use no special treatment. The idea is just to check if I can get the same result with TAMkin as with CP2K. 

I made a CP2K version of the script shown here https://molmod.github.io/tamkin/tutorial/getting_started.html, see attached py file. 
However, the frequencies (in cm-1) that TAMkin produces are different form those that are found in the CP2K output file.
A comparison of the eight first frequencies are shown here:

TAMkin: 31.1    38.3    47.9    74.7    82.0   111.0   223.7   245.6
CP2K:   27.3     40.8    51.0    74.0    80.0   109.9  224.1   246.3

What could be the cause of this difference?
The Python script and the CP2K single point and frequency calculation are attached.

Thanks a lot for your help. 

Regards,
Torstein Fjermestad

--
You received this message because you are subscribed to the Google Groups "TAMkin" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tam...@googlegroups.com.

Toon Verstraelen

unread,
May 13, 2020, 6:19:04 AM5/13/20
to tam...@googlegroups.com
Dear Torstein,

I agree regarding the external degrees of freedom. There should be None.

I checked several potential causes and my best guess is that CP2K is not symmetrizing the Hessian prior to diagonalization. If I comment out the symmetrization from TAMkin, I get these frequencies:

28.3    38.7    45.8    75.0    84.1   112.3   223.2   244.7

I'm sure how exactly CP2K is doing the calculation, so I cannot tell for sure. By symmetrizing, the finite difference error in the Numerical Hessian from CP2K is somewhat reduced, so I would recommend it.

Best regards,

Toon

To unsubscribe from this group and stop receiving emails from it, send an email to tamkin+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/tamkin/58edaaca-1ccc-4146-93ef-fafe41e518ea%40googlegroups.com.


--
Prof. Dr. ir. Toon Verstraelen
Center for Molecular Modeling
Technologiepark-Zwijnaarde 46 (before Jan 1, 2019, this was building 903)
Ghent University
Tech Lane Ghent Science Park Campus A
9052 Zwijnaarde
Belgium
How to reach my office? See https://molmod.ugent.be/where-find-us (2nd floor, CMM, door next to Xerox)
Mobile: +32 471 66 68 83 (Consider email instead. If you leave no message, I will not call you back.)
E-mail: Toon.Ver...@UGent.be


Reply all
Reply to author
Forward
0 new messages