[CP2K:5585] DX in vibrational analysis

[hut... at chem.uzh.ch]

Hi

maybe it would be a good idea to test your settings with a
smaller molecule.

It looks to me like you should considerably tighten the 
convergence criteria, both for the SCF and the geometry
optimization. Especially, if you use a small DX you will need
a very well converged energy.

I would use a RMS_FORCE of 5. 10-5 or smaller and a
SCF convergence of 1. 10-7 or better.

regards

Juerg Hutter
 
--------------------------------------------------------------
Juerg Hutter                         Phone : ++41 44 635 4491
Institut für Chemie C                FAX   : ++41 44 635 6838
Universität Zürich                   E-mail: hut... at chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp... at googlegroups.com wrote: -----
To: cp... at googlegroups.com
From: Tobias Kraemer <161brun... at gmail.com>
Sent by: cp... at googlegroups.com
Date: 08/05/2014 04:11PM
Subject: [CP2K:5585] DX in vibrational analysis

Dear all,


following my previous post (huge number of imaginary frequencies...) on this issue, I have in the meantime
tried to break down to problem into a smaller one.Basically, instead of using the full bulk system, I have tried
a simple optimization and frequency calculation on one of the cationic portions (organometallic complex) of the
system. I have used a cubic cell with a = 20 Ang. I have made sure that the structure for the subsequent vibrational
analysis job does indeed correspond to the final structure of the optimization run (I have included ATOMIC_COORDINATES ON
to print out the geometry in the output).  Again, in the first attempt (using DX 0.001) I end up with a huge number (~50) of imaginary
frequencies, including some very large ones.

 VIB|                        NORMAL MODES - CARTESIAN DISPLACEMENTS
 VIB|
 VIB|                         1                    2                    3
 VIB|Frequency (cm^-1) -4133.188102         -2206.847291         -2036.718924
 VIB|Intensities           0.326216             0.395802             0.125908
 VIB|Red.Masses (a.u.)     1.746454             1.699628             1.961778
 VIB|Frc consts (a.u.)    -1.304878            -0.103208            -0.086426


Now, if I repeat the same calculation with a larger DX of 0.01, I end up with less imaginary modes, but still too many (17) to trust the result.
Also, massive change now in the magnitude of the wavenumbers, which drops to -957 cm-1 for imaginary mode nr 1. Think the details here
are not so relevant, but I am quite confused. So the problem is not related to the bulk system, but intrinsically related to the vibrational
analysis protocol I am following at the moment. Also, as far as the SCF is concerned, I am using inner / outer SCF  in combination with DIIS.
Apparently the very first SCF never converges after 100 steps, which is when the outer SCF kicks in and from there everything seems to converge
smoothly. Maybe my thinking is wrong here? This is what it looks like (in condensed form) from one of the replica outputs: 

rhdcpenba_mol_vib-r-9.out:  *** SCF run NOT converged ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    15 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in     4 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in     6 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in     6 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in     8 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in     9 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in     8 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in     9 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in     9 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    10 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    10 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    13 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    13 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    14 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    15 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    16 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    15 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    15 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    15 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    15 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    15 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    14 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    14 steps ***
rhdcpenba_mol_vib-r-9.out:  *** SCF run converged in    15 steps ***

So my question really is, how do I choose an appropiate value for DX (and other parameters), and what is potentially going wrong here. I am absolutely clueless I must say, since I considered this as an 'easy' task. I feel that I am just simply doing something wrong, although I thought I was following standard procedure. I am keeping all other values such as SCF convergence and so on untouched, in order to not introduce any inconsistencies here. Would a more rigorous optimisation perhaps be helpful?


Thanks for your help, much appreciated. Files attached. 

Best

Tobi
  
  -- 
 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... at googlegroups.com.
 To post to this group, send email to cp... at googlegroups.com.
 Visit this group at http://groups.google.com/group/cp2k.
 For more options, visit https://groups.google.com/d/optout.
 

[attachment "rhdcpenba_vib0001.out" removed by Jürg Hutter/at/UZH]
[attachment "rhdcpenba_vib0001.inp" removed by Jürg Hutter/at/UZH]