suggestions and issues

[phyjz]

Dear developers,

I have been using TDEP for some time, and I would like to share a few suggestions and report some issues I have encountered:

1. the current TDEPtools doesn't seem to support parsing vasprun.xml for machine learning AIMD since vasp6. my version of TDEP is 24.01.
2. I hope to add commands in TDEP software to convert force constants from TDEP format to phonopy and shengbte format.
3. I would like to know the difference between the following three ways of calculating thermal conductivity, mainly in terms of accuracy, calculation cost and the range of applicable materials.
   a. Pure AIMD
   b.AIMD+Samples from MD
   c. Canonical configuration module to generate the structure.

Thank you for your time and efforts, and I look forward to your response!

[mjv500]

Hello Phyjz Thanks for the feedback - my 2 cents below

On Fri, Dec 27, 2024, 09:16 phyjz ***@***.***> wrote: Dear developers, I have been using TDEP for some time, and I would like to share a few suggestions and report some issues I have encountered: 1. the current TDEPtools doesn't seem to support parsing vasprun.xml for machine learning AIMD since vasp6. my version of TDEP is 24.01. Good to know, but not clear what will be maintained in the package long

term. TDEPTools was created by Florian Knoop who has now accepted a job in industry

1. I hope to add commands in TDEP software to convert force constants from TDEP format to phonopy and shengbte format. Great. See the utility

https://tdep-developers.github.io/tdep/program/dump_dynamical_matrices/ where I did the same for abinit. It will be a good template.

1. I would like to know the difference between the following three ways of calculating thermal conductivity, mainly in terms of accuracy, calculation cost and the range of applicable materials. a. Pure AIMD b.AIMD+Samples from MD Don't see a distinction here. You always subsample the MD with a limited

number of snapshots. If you mean from a classical ML potential fit to the AIMD, it should be identical if the fit is good, just faster.

1. c. Canonical configuration module to generate the structure. Here the configurations are taken from a gaussian sampling of the harmonic

phonons. Taken to self consistency this is optimal within the framework of SCHA but can not generate a distribution which is not gaussian, and the exact one should not be gaussian. This is the extra detail you get from MD sampling, which is not limited in shape.

…

Thank you for your time and efforts, and I look forward to your response! — Reply to this email directly, view it on GitHub <#120>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABWOT3Z6OMCMXLWQUHCX3BT2HUEEDAVCNFSM6AAAAABUIJD7X2VHI2DSMVQWIX3LMV43ASLTON2WKOZSG43DANJWGQYTINY> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[Guodonglin-cqust]

I hope TDEP can directly output the phonon dynamics matrix directly used by EPW program in QE code, studying the influence of anharmonic phonons renormalization on electrical transport.

[flokno]

Hi @phyjz ,

thank you for your interest in TDEP!

Matthieu already commented, but let me add the following to the question

the current TDEPtools doesn't seem to support parsing vasprun.xml for machine learning AIMD since vasp6. my version of TDEP is 24.01.

tdeptools use ASE to parse vasprun.xml. ASE, however, does not yet support parsing the VASP MLFF output, see https://gitlab.com/ase/ase/-/issues/1491

I strongly encourage you to implement the parser in ASE, then it will automatically work in tdeptools!

[phyjz]

tdeptools use ASE to parse . ASE, however, does not yet support parsing the VASP MLFF output, see https://gitlab.com/ase/ase/-/issues/1491`vasprun.xml`

I strongly encourage you to implement the parser in ASE, then it will automatically work in tdeptools!

Thank you very much, I will try what you mentioned above!