Vibrational analysis

[mengchao shi]

Dear all

     I want to calculate the atomic frequency of diamond silicon by CP2K.
     My supercell contains N=8 atoms (a PBC system), and I will get the 3N=24 frequencies.
      

     However, I only get the  21 frequencies,  The frequencies of the rotations  are obtained, but the part of translations are not obtained.

     I use the VASP to calcualte the vibrational analysis, I get the 24 frequencies.

     I  compare the two results:

    

 VASP: (unit: cm-1,  24 frequencies )    

 513.100912      513.097046       513.092040    
 461.327377       461.230874     461.217489     461.210747     461.189027     461.185948

 409.422422        409.414429     409.406072    409.397086     409.393930    409.387741  

  140.194017       140.189188     140.187673    140.182783     140.175562   140.172242 

    0.006451           0.010431         0.055547
 

  CP2K: (unit: cm-1,  21 frequencies)    

    515.944308    515.944308     515.944308

    463.042397     463.042397    463.042397    463.042397  463.042397     463.042397

    412.993772     412.993772    412.993772     412.993772  412.993772    412.993772

    139.536054      139.536054    139.536054    139.536054  139.536054    139.536054

   

    The results are similar, but the number is different, how do I get 24 frequencies by CP2K.    

   The part input:

    "

     &GLOBAL

      PROJECT system
      RUN_TYPE VIBRATIONAL_ANALYSIS
      PRINT_LEVEL MEDIUM
   &END GLOBAL

    &VIBRATIONAL_ANALYSIS
        NPROC_REP 3
        DX 0.001
        FULLY_PERIODIC .TRUE.
&END VIBRATIONAL_ANALYSIS

       "

       Thank you for your help.

        Best,

        Mengchao.

[Lucas Lodeiro]

Hello Mengchao,

AFAIK your results are consistent. Of course there are 24 normal modes, and 3 of them are translational ones. CP2K removes the translational modes before printing the data, you can find this in the CP2K output (search "VIB| Eigenvectors before removal" and " VIB| Frequencies after removal"), where the complete set of normal modes and frequencies are printed before the translational removal. This does not change significantly the results, so you can think about 21 normal modes and 3 extra translation modes with zero frequency.

In VASP the translational removal is not done and the low frequency modes are the translational ones... but at Gamma point they must be zero frequency. The resulting non zero value is an artifact of the PW rupture of translational invariance, which is due to the non complete PW basis set (infinite energy plane wave). As a guess, taking into account that in VASP the PW cutoff is lower compared with used in CP2K, and PW are used to wavefunctions, the error due to rupture of translational invariance is higher in VASP.

Regards - Lucas

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/902b47be-94ff-4145-ad9f-365cb6d7317en%40googlegroups.com.

[Jürg Hutter]

Hi

adding to this correct explanation of the problem. Unfortunately, the translational
invariance will be broken more severely in CP2K (using the GPW method)
than in other codes using equidistant grid integration. If you want to reduce this effect
in CP2K you need to run with the GAPW method or GPW with NLCC pseudopotentials.

regards

JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Lucas Lodeiro <eluni...@gmail.com>
Sent: Saturday, October 15, 2022 6:17 PM
To: cp...@googlegroups.com
Subject: Re: [CP2K:17871] Vibrational analysis

Hello Mengchao,

AFAIK your results are consistent. Of course there are 24 normal modes, and 3 of them are translational ones. CP2K removes the translational modes before printing the data, you can find this in the CP2K output (search "VIB| Eigenvectors before removal" and " VIB| Frequencies after removal"), where the complete set of normal modes and frequencies are printed before the translational removal. This does not change significantly the results, so you can think about 21 normal modes and 3 extra translation modes with zero frequency.
In VASP the translational removal is not done and the low frequency modes are the translational ones... but at Gamma point they must be zero frequency. The resulting non zero value is an artifact of the PW rupture of translational invariance, which is due to the non complete PW basis set (infinite energy plane wave). As a guess, taking into account that in VASP the PW cutoff is lower compared with used in CP2K, and PW are used to wavefunctions, the error due to rupture of translational invariance is higher in VASP.

Regards - Lucas

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/902b47be-94ff-4145-ad9f-365cb6d7317en%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/902b47be-94ff-4145-ad9f-365cb6d7317en%40googlegroups.com?utm_medium=email&utm_source=footer>.

--
You received this message because you are subscribed to the Google Groups "cp2k" group.

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAOFT4PLYq6AN59wwRqKmSnbzbnCYGiaE07N26iVkCe0jq_ONFQ%40mail.gmail.com<https://groups.google.com/d/msgid/cp2k/CAOFT4PLYq6AN59wwRqKmSnbzbnCYGiaE07N26iVkCe0jq_ONFQ%40mail.gmail.com?utm_medium=email&utm_source=footer>.

[mengchao shi]

Thank you for your reply, it is very useful for me.

[mengchao shi]

Dear JH
Thank you so much.