[CP2K-user] Using TAMkin with CP2K - how is CP2K diagonalizing the Hessian matrix?

[Torstein Fjermestad]

Dear all, 

I would like to follow up on this question that I asked on June 10 this 
year.
I realize that I might have expressed myself in a very complicated way, and 
I therefore try to be clearer this time.

The problem is that when I diagonalize the Hessian printed in the cp2k 
output file with the following python code (The complete python code is 
attached)

from numpy import genfromtxt
import numpy as np
import sys

hessianFile=sys.argv[1]
hessian = genfromtxt(hessianFile,delimiter=',')
freq = np.linalg.eigvals(hessian)

 
, I get slightly different frequencies than what is printed in the output 
file (and the .mol file). The frequencies are compared in the table below.
Column 2: Frequencies from CP2K output file
Column 3: Frequencies obtained when diagonalizing the Hessian with numpy

Table 1: The eight lowest frequencies. There are 48 in total.

What am I doing wrong in the diagonalization?

Thanks a lot for your help.

Regards,

Torstein Fjermestad





onsdag 10. juni 2020 kl. 15:56:53 UTC+2 skrev Torstein Fjermestad:

> Dear all, 
>
> I would like to use TAMkin (http://molmod.github.io/tamkin/) to compute 
> thermodynamic properties from the vibrational frequencies computed with 
> CP2K.
> As I have understood it, the way TAMkin works is to read and diagonalize 
> the Hessian matrix printed in the CP2K output.
> When I compare the frequencies obtained with TAMkin with those obtained 
> with cp2k, the values are not the same (Compare column 2 and 3 in the table 
> below). 
>
> The first task is then to understand how CP2K computes the frequencies. To 
> investigate this, I copied the Hessian matrix from the CP2K output (printed 
> below the string "Hessian in cartesian coordinates") and pasted it into 
> Excel and saved it as a csv file. This file I read into python and 
> diagonalized the Hessian with numpy. The frequencies I get are different 
> from those printed in the cp2k output file (compare column 3 and 4 in the  
> table below). 
> I have assumed that the units of the Hessian matrix in the cp2k output 
> file are Ry*bohr-2*kamu-1 (kamu = 1000*amu). This assumption is based on 
> trial and error; it was what got me closest to the cp2k frequencies. 
>
>
> Table 1: The eight lowest frequencies. There are 48 in total.
>
> frequencies
>
> TAMkin / 
> cm-1
>
> cp2k / 
> cm-1
>
> Diag., cp2k Hessian / 
> cm-1
>
> 1
>
> 31.1
>
> 27.3
>
> 32.6
>
> 2
>
> 38.3
>
> 40.8
>
> 40.1
>
> 3
>
> 47.7
>
> 51.0
>
> 50.2
>
> 4
>
> 74.7
>
> 74.0
>
> 78.2
>
> 5
>
> 82.0
>
> 80.0
>
> 85.8
>
> 6
>
> 111.0
>
> 109.9
>
> 116.3
>
> 7
>
> 223.7
>
> 224.1
>
> 234.3
>
> 8
>
> 245.6
>
> 246.3
>
> 257.3
>
> What is wrong with my calculation?
> What steps should be done to go from the Hessian to the frequencies?
>
> The Hessian in csv format, the python script and the cp2k output file are 
> attached. 
>
> Thanks a lot for your help.
>
> Regards,
> Torstein Fjermestad
>  
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20201219/8a4c82b5/attachment.htm>