Add the possibility restart the vibrational modes calculation

[lipelopesoliveira]

Motivation

The current version of the workflow uses CP2K to calculate the vibrational modes and FTIR and Raman spectra of a material. Unfortunately, it does not allow this calculation to be restarted if it is not completed. As the number of steps required for the calculation is considerably large (6N where N is the number of atoms) it is extremely likely that it will be necessary to restart the calculations for some materials.

What should be done

Implement an alternative for calculating the vibrational modes and the FTIR/Raman spectra that allows the correct restart at any stage of the simulation.

[lipelopesoliveira]

Here we have two different paths:

1. Calculate the Raman spectra based on a molecular dynamics simulation using the TRAVIS software.^[1,2]

   • Pros:
     • The simulation can be easily restarted several times
     • The use of the ASPC algorithm to extrapolate the electron density of a new MD step that CP2K uses makes the calculation relatively fast.
     • Adds information on the structure dynamics and temperature dependency to the Raman/FTIR spectrum
     • Allow the use of TRAVIS for other kinds of analysis
   • Cons:
     • The number of MD steps required for a well-converged spectrum is not known a priori
     • Increased difficulty in peak identification within the vibrational spectrum.
     • Generates a substantial volume of files (several terabytes per simulation). Adds an additional software component to the workflow, necessitating further testing and benchmarking.

2. Developing custom code to generate the different "images" needed for the finite difference method and post-process the results.

   • Pros:
     • Parallelization over the "images" would allow easy recover and speed-up
     • Minor modifications on the current workflow
     • Offers flexibility to compute additional information, such as anharmonic corrections.
   • Cons:
     • Demands more extensive development effort.
     • Requires thorough testing and benchmarking.
     • Lacks the capacity to provide insights into structural dynamics and temperature effects on the simulated spectra.

@neumannrf @m-estevesf I would like your input on this issue. I'm initially leaning towards option 2, but option 1 could also add interesting extra information, specially the temperature dependency. In any case, I will only have time to actually dedicate to implementing one of the two options after the SBQT (27/10), so we have time to think calmly about this matter and decide the best path.

[m-estevesf]

Personally I also would also lean towards option 2 in case it is possible.

For option 1 I see two main concerns:

1. Considering the discussions with Prof. Hélio, I'm not sure how good the results from the molecular dynamics would be. My biggest concern is creating a new implementation from scratch and it not yielding as good results as the method we are currently using.
2. If each simulation with TRAVIS can take a few Tb of data, we may face problems running these simulations on the machines we have available. On our previous work on MicroCT images we had severe limitations due to the storage requirements on CCC/Virtual machines.

[neumannrf]

@m-estevesf @lipelopesoliveira I don't know if you have seen this already, but there is a similar approach implemented in "A data-science approach to predict the heat capacity of nanoporous materials" (see the Supplementary Information) using AiiDA and made available here.

PS: by the way, is our current workflow able to calculate the heat capacity already? Maybe this should become its own issue later...

[lipelopesoliveira]

Thank you @neumannrf for remembering this repository. AiiDA uses Phononpy to generate displacements and calculate normal vibration modes via the harmonic approximation, however it is not possible to calculate Raman/FTIR intensities in Phononpy. There is a tool called Phonopy-Spectroscopy that should be able to calculate it for us, but I remember testing it a while ago and not being able to run it. But I think it's worth trying one more time, so we don't have to reinvent the wheel.

No, our current workflow does not calculate the heat capacity or any other thermodynamic property. I tried to implement this when I was still at IBM but ran out of time.

Yes, I think we can create a new issue to implement the calculation of the heat capacity and other relevant thermodynamic properties.

[lipelopesoliveira]

A Critical Assessment on Calculating Vibrational Spectra in Nanostructured Materials

Alexander E. J. Hoffman, Wim Temmerman, Emma Campbell, Alessandro Ali Damin, Ines Lezcano-Gonzalez, Andrew M. Beale, Silvia Bordiga, Johan Hofkens, and Veronique Van Speybroeck* DOI: 10.1021/acs.jctc.3c00942

This paper answer several methodological questions that we raised in the past.

[neumannrf]

High-throughput computation of Raman spectra from first principles

"Raman spectroscopy is a widely-used non-destructive material characterization method, which provides information about the vibrational modes of the material and therefore of its atomic structure and chemical composition. Interpretation of the spectra requires comparison to known references and to this end, experimental databases of spectra have been collected. Reference Raman spectra could also be simulated using atomistic first-principles methods but these are computationally demanding and thus the existing databases of computational Raman spectra are fairly small. In this work, we developed an optimized workflow to calculate the Raman spectra efficiently and taking full advantage of the phonon properties found in existing material databases. The workflow was benchmarked and validated by comparison to experiments and previous computational methods for select technologically relevant material systems. Using the workflow, we performed high-throughput calculations for a large set of materials (5099) belonging to many different material classes, and collected the results to a database. Finally, the contents of database are analyzed and the calculated spectra are shown to agree well with the experimental ones."

The Computational Raman Database (CRD) contains calculated IR/Raman spectra (associated CIF files) for 10,032 materials from the Materials Project database.