normal mode analysis on an ab initio trajectory

[Emma Rossi]

Dear all,

I would need to assign the bands of an IR spectrum, obtained through AIMD simulation. 

I would like to perform a normal mode analysis (NMA) on the trajectory I have already produced to properly take into account solvation effects as well as the correct weight of the sampled conformers. 

I guess I could do so retracing the traj.xyz and using the vibrational analysis section in the input, but I would be very greatful if you could clarify a few issues about this strategy:

1) is it the most efficient and correct way to do NMA over a trajectory or do you suggest any other ways?

2) how the output with eigenfrequencies and eigenvectors should look like? I can figure the typical output for NMA on single configurations, but I have no ideas of what should I expect for NMA on trajectories (I suppose that the hessian will be diagonalized at each step, but how these info will be put together at the end of the process? )

I have read the original paper of CP2K and the pdf presentations available online, but I couldn't fix my problem.

I am very willing to understand and learn from anyone who will help.

Thank you very much in advance for your support.

Best regards,

Emma Rossi

[Giacomo Melani]

Dear Emma,

You can certainly do a vibrational analysis through the standard input for such calculation (https://manual.cp2k.org/cp2k-2023_1-branch/CP2K_INPUT/VIBRATIONAL_ANALYSIS.html). That will read your selected configuration and perform a diagonalization of the Hessian matrix to extract the corresponding normal mode frequencies and eigenvectors. This type of calculation will read a specific geometry. It can also provide you additional information, like IR/Raman intensities, but that requires the calculation of transition dipole moments and polarizabilities.

However, if you already have an AIMD trajectory, you might also want to do a vibrational analysis by computing a time correlation function. For instance, you can extract a vibrational density of states by calculating the time-autocorrelation function of nuclear velocities. For sure, the combination of NMA and correlation functions will give you the best assessment of vibrational modes in your system.

I hope this helps.

Best,

Giacomo Melani

--
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/6edca95f-7fb7-4a27-97cd-5868c9ae611en%40googlegroups.com.

[Lucas Lodeiro]

Hi Emma,

I do not know how you managed to obtain the IR spectrum from your trajectory, but I suggest the TRAVIS program to do it: http://www.travis-analyzer.de/

There is many ways to accomplish it, and the results are very good. Also, there is the possibility to obtain the "normal modes" of the trajectory which are consistent with the trajectory IR/Power spectrum.

There are some tutorials, and the program is self explaining: https://brehm-research.de/spectroscopy.php

In the normal modes section you can feed the program with one or more minimum energy conformer where the trajectory is projected, and approximated normal modes are obtained.

Regards - Lucas

--

[Emma Rossi]

Dear Giacomo,

thank you very much for your prompt reply and for your suggestion. 

I think that the strategy you are prosposing is useful for sure to have an idea about the contribution of the different bond vibrations to the normal modes of a specific conformer. However, I have still doubts concerning how to weight the NMA of the different conformers in order to interpret the whole IR spectrum. 

In particular, this would be important in case different conformers have different normal modes. I think that VDOS analysis could be useful as additional check, but information about the weights would be still missing

However, I will try with your strategy and in the meanwhile I'll think about the next step.

Thank you very much for the stimulating discussion.

Best regards,

Emma Rossi

[Emma Rossi]

Dear Lucas,

Thank you very much for your suggestion. I obtained the IR spectrum through the dipole moment of the whole box computed at each step of the trajectory by CP2K. Unfortunately, I couldn't use TRAVIS since I had not printed the electron density during the simulation.

However, it would be great if I could use TRAVIS as post processing tool to perform NMA on the trajectory. As far as I have seen in the documentation,  the online tutorials/presentations deal only with the simulation of the spectra, which, indeed, require electron density. I couldn't find any instruction about how to perform the NMA (I've found just the original paper). 

However, if you have expertise in TRAVIS usage, I would be very grateful if you could answer a doubt about TRAVIS NMA tool.

Can I perform NMA on my trajectory without performing as a preliminary step the calculation of the IR spectrum? As I mentioned above, I have not the electron density.

Thank you very much for your kind support.

[Lucas Lodeiro]

Dear Emma,

AFAIK compute the IR spectra through dipole (Berry phase), Wannier or Voronoi integration must result in almost the same spectra, I remember Dr. Brehm shows the unique difference between them is the relative intensity between peaks.and Wannier could be problematic sometimes when there is high symmetry. So there is no problem. BTW, you can compute the properties of your trajectory using the REFTRAJ option in the MD.

About the IR spectrum on TRAVIS, it can be computed with Voronoi integrated electron density (which currently can be done directly in CP2K on the fly), with fixed charges or fluctuating charges calculated along the AIMD. I used the last one with self-consistent Hirschfeld charges and it worked very well. 

But the point is, with just your trajectory you can compute the power spectrum which can show you some "missing" signals in IR (maybe), and the NMA without problem. The other spectral properties need electron densities and more.

In the program, the NMA is called NC (normal coordinate analysis) . When the "List of functions" is displayed, in the Spectroscopic Functions section there is "nc". This is what you need. Then you need give some information and feed with the conrformer's structures (previously optimized at the same level). 

I suggest computing first the power spectra with TRAVIS and "optimize/play" with the variables and become familiar with them (the defaults are good, but sometimes a little change can upgrade the spectrum), because for the "nc" the same procedure is done + the reference structures.

Finally you will have something like this:

     Mode         Integral (K)       Center (cm^-1)
     ----------------------------------------------
        1            -0.112446              2735.82
        2           -0.0757895              1377.51
        3           -0.0508332               529.47
        4          -0.00656595               615.81
        5            0.0120252               542.87
        6             0.177628              1415.45
     ----------------------------------------------
        7                327.5               528.37
        8              335.364               535.27
        9              338.557               561.62
 ...
       23              319.248              2958.93
       24              330.046              3163.23
     ----------------------------------------------

Which list the modes frequencies and their thermalization... The mode temperature must be similar to the AIMD mean temperature (the first six must be near zero)... and a molden file is printed with the normal modes for each reference structure where you can see them. The normal mode frequencies must be similar to the peaks in the power spectrum.

In my experience when a time step over 1fs is used the nc starts to fail... also if you have the velocities "trajectory" you can use it (with -vel option) to obtain a better velocity autocorrelation function when the time step is equal or bigger than 0.5fs.

If you have doubts or need help do not hesitate to write to me, I am not an expert but I can help or learn on the way :P.

Regards - Lucas

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/6bc02dbb-0824-46a7-8edc-8d98c5c87f21n%40googlegroups.com.

[Francisco Gámez]

Dear all

What I frequently found with the NMA is that I the "normal modes coordinates" in the molden file after TRAVIS evaluation might be Ok and the IR spectra is remarkably good, but the visualization of the NM in MOLDEN shows that the assignments are disparate and make no sense from a chemical point of view and, obviously, do not match with DFT calculations. Does somebody has some tips to get correct mode assignment from the BOMD trajectory?

Kind regards

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAOFT4PLFUJ3KjuE%2BstwXiCONooLJx%3DCBBYA_1UuXM7cFL%2B6V%2BA%40mail.gmail.com.

[Matthew Graneri]

Hi Lucas,

Sorry for resurrecting this issue, but I'm in a similar situation to Emma, trying to run the normal coordinate analysis through TRAVIS and am absolutely stuck. I get up to the question asking for the reference structures and, when I specify the file containing my reference geometry, I get told: 'Some atoms are not connected to this molecule // Molecule recognition failed'.

My AIMD simulation contains several different molecular species, and I am trying to work out which peak corresponds to which vibration for each molecule in the system. I've tried using a normal .xyz file containing only the molecule I want TRAVIS to analyse, and I've tried using an .xyz file containing the entire system from the simulation, but nothing has worked.

The only thing I can think of that might be an issue is that I'm using the .emp files from the IR spectrum simulation as the input for the analysis. I can't think of what other file TRAVIS would like, though...

If you (or anyone else) has any idea why the analysis might not be working, I would very much appreciate any advice you can provide!

Regards,

Matthew

[Luca]

I am currently developing a procedure to resolve complex vibrational spectra. Although the procedure is not yet freely available, and the related paper is currently under review, I can provide you with the preprint of the article. Additionally, I am willing to apply the software to resolve your computed spectra. I am looking for some applications.

The procedure requires Cartesian coordinates and velocities from a cp2k molecular dynamics (MD) simulation. The vibrational density of states (VDOS) is calculated from the Fourier transform of the velocity autocorrelation function, obtained by processing ab initio (DFT) molecular dynamics simulations.
Within this framework, I have developed a computational tool designed to assign the vibrational mode associated with a specific frequency. The method involves projecting velocities along a specific set of internal coordinates, such as stretching or bending, to calculate a partial VDOS (pVDOS). This partial VDOS accounts only for the vibrations associated with the selected set of internal coordinates. It aids in the interpretation of the total VDOS and the experimental spectra in a meaningful way.

Cheers,

Luca

[Matthew Graneri]

Hi Luca,

I'd be happy to give it a go, but I would still like to get the normal coordinate analysis from TRAVIS working first, considering that's what I generated my IR spectra with. I could definitely use your program as a validation/verification method, though. I imagine my systems (there are three) would be a good test case, to see how robust your program is.

Perhaps get back to me when the article gets published? There's another paper I need to publish before I can start working on this one, so there's no serious urgency. Plus, I'm still trying to work out what the AIMD data is telling me—this is the only AIMD work anyone in my department (possibly the entire university) has ever done, so I'm flying blind (people have only ever done proper ab initio/DFT or MD calcs here), so I'd really value any advice on getting the normal coordinate analysis on TRAVIS to work...

Regards,

Matthew

[Lucas Lodeiro]

Hi Matthew,

To run the normal coordinate analysis you have to feed TRAVIS with the trajectory file and the velocity file if you have it (this is just an option to avoid numerical differentiation, which introduce some errors). Running the normal coordinate analysis needs the same procedure as the power spectra calculation. In the middle of run, TRAVIS ask you for a reference structure: AFAIK this structure must be written in the same atom order as is in  the AIMD  molecule atomic order, and the structure must be optimized at the same level of theory. With it you can complete the calculation. Some comments, if your molecule has many and pretty different accessible conformers, this could be a problem. Also take into account that hydrogen bonding active modes cannot be protected properly on one molecule, because those are collective modes.

Also if you want to check from which molecule is the normal mode in the IR, you can compute the power spectra for the whole system, and also for each molecule type. This also can be done for the IR, so you can check which molecule contributes to a given signal... but you cannot know how is that normal mode, for that you need the normal coordinate analysis.

Regards - Lucas Lodeiro

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/2e3cbdb1-6112-4584-b658-2d797a741665n%40googlegroups.com.