Hi all,
I’m interested in calculating the absolute band edge positions. And since you guys have implemented the class BranchPointEnergy in matminer/featurizers/bandstructure.py. I was wondering if that routine can be used for the calculation and how are you planning to do it?
Are you already using the routines? And from which kind of calculation (GGA, Hybrid, GW, …) do you plan to take the Bandstructure? Thanks for your work, it is always interesting going through your routines and seeing what kind of thinks your codes can do or are planned.
Thanks
Hendrik
Hi Hendrik,
We have two different codes:
-
atomate is used for actually doing calculations (interfaces mostly w/ VASP) - see https://hackingmaterials.github.io/atomate/
-
matminer is taking data (from calculations, expts, etc) and doing various post-processing or featurizing. It does not help you perform simulations.
Matminer starts with a pymatgen Bandstructure object and can calculate branch point energies from that. It is up to the user as to where the Bandstructure object comes from - pymatgen Bandstructure objects can be created from any type of band structure simulation. It could be GGA, Hybrid, GW, etc. - that is up to the user. Matminer does not provide any guidance as to which to use, that is really up to the user’s discretion and available data set.
Best,
Anubhav
Hi Anubhav,
thanks for your reply. So in principal the HSE-Bandstructure workflow from atomate can be used in order to generate the bandstructure object, which can than be post-processed in the BranchPointEnergy routine from matminer. I will test it next week. Sorry, if I have posted this topic in the wrong forum. I now realise that my questions were more about the workflow, so the atomate forum was probably better suited for asking this questions.
Best
Hendrik
···
So in principal the HSE-Bandstructure workflow from atomate can be used in order to generate the bandstructure object, which can than be post-processed in the BranchPointEnergy routine from matminer
Yes - just make sure you have a “uniform” band structure object since that samples k-space (not the “line” mode). Normally the atomate bandstructure workflows compute both, but the HSE band structure routines built into atomate (e.g., atomate/vasp/workflows/base/library/bandstructure_hse.yaml) only compute the line mode band structure in HSE. This is because computing a dense uniform k-mesh is very expensive in HSE since one is not able to do non self-consistent runs with hybrid functionals. You might be able to get away with using the band structure generated from an HSE static run, although this would be coarse mesh and probably better suited towards quick screening than detailed publication.
Sorry, if I have posted this topic in the wrong forum
I think you are fine, I am actually glad someone started using the matminer list, hopefully this will encourage others. If you have further questions about the workflow you should post in atomate. Probably, you will in the end get me responding no matter where you post 
···
On Sun, Feb 18, 2018 at 6:26 AM, [email protected] wrote:
Hi Anubhav,
thanks for your reply. So in principal the HSE-Bandstructure workflow from atomate can be used in order to generate the bandstructure object, which can than be post-processed in the BranchPointEnergy routine from matminer. I will test it next week. Sorry, if I have posted this topic in the wrong forum. I now realise that my questions were more about the workflow, so the atomate forum was probably better suited for asking this questions.
Best
Hendrik
–
You received this message because you are subscribed to the Google Groups “matminer” group.
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].
To view this discussion on the web visit https://groups.google.com/d/msgid/matminer/c53749e2-2579-49d5-a2f6-682373bd14cf%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
–
Best,
Anubhav
Hi all,
I wanted to let you know that the procedure is working. We are getting quite good results compared to experiment. Especially when we tune the fraction of exchange in such a way that it reproduces the experimental band gap. With the exception of one amorphous material (the others are polycrystaline). I’m not sure if I’m allowed to show the comparison. Some not compared results can be found on my Poster on Researchgate.
Thanks for help and have a nice day
Hendrik
···
Am Sonntag, 18. Februar 2018 16:46:22 UTC+1 schrieb Anubhav Jain:
So in principal the HSE-Bandstructure workflow from atomate can be used in order to generate the bandstructure object, which can than be post-processed in the BranchPointEnergy routine from matminer
Yes - just make sure you have a “uniform” band structure object since that samples k-space (not the “line” mode). Normally the atomate bandstructure workflows compute both, but the HSE band structure routines built into atomate (e.g., atomate/vasp/workflows/base/library/bandstructure_hse.yaml) only compute the line mode band structure in HSE. This is because computing a dense uniform k-mesh is very expensive in HSE since one is not able to do non self-consistent runs with hybrid functionals. You might be able to get away with using the band structure generated from an HSE static run, although this would be coarse mesh and probably better suited towards quick screening than detailed publication.
Sorry, if I have posted this topic in the wrong forum
I think you are fine, I am actually glad someone started using the matminer list, hopefully this will encourage others. If you have further questions about the workflow you should post in atomate. Probably, you will in the end get me responding no matter where you post
On Sun, Feb 18, 2018 at 6:26 AM, [email protected] wrote:
Hi Anubhav,
thanks for your reply. So in principal the HSE-Bandstructure workflow from atomate can be used in order to generate the bandstructure object, which can than be post-processed in the BranchPointEnergy routine from matminer. I will test it next week. Sorry, if I have posted this topic in the wrong forum. I now realise that my questions were more about the workflow, so the atomate forum was probably better suited for asking this questions.
Best
Hendrik
–
You received this message because you are subscribed to the Google Groups “matminer” group.
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].
To view this discussion on the web visit https://groups.google.com/d/msgid/matminer/c53749e2-2579-49d5-a2f6-682373bd14cf%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
–
Best,
Anubhav
Hi Hendrik,
That’s great! Thanks for reporting back.
A couple of FYI:
···
On Monday, April 16, 2018 at 6:30:34 AM UTC-7, [email protected] wrote:
Hi all,
I wanted to let you know that the procedure is working. We are getting quite good results compared to experiment. Especially when we tune the fraction of exchange in such a way that it reproduces the experimental band gap. With the exception of one amorphous material (the others are polycrystaline). I’m not sure if I’m allowed to show the comparison. Some not compared results can be found on my Poster on Researchgate.
Thanks for help and have a nice day
Hendrik
Am Sonntag, 18. Februar 2018 16:46:22 UTC+1 schrieb Anubhav Jain:
So in principal the HSE-Bandstructure workflow from atomate can be used in order to generate the bandstructure object, which can than be post-processed in the BranchPointEnergy routine from matminer
Yes - just make sure you have a “uniform” band structure object since that samples k-space (not the “line” mode). Normally the atomate bandstructure workflows compute both, but the HSE band structure routines built into atomate (e.g., atomate/vasp/workflows/base/library/bandstructure_hse.yaml) only compute the line mode band structure in HSE. This is because computing a dense uniform k-mesh is very expensive in HSE since one is not able to do non self-consistent runs with hybrid functionals. You might be able to get away with using the band structure generated from an HSE static run, although this would be coarse mesh and probably better suited towards quick screening than detailed publication.
Sorry, if I have posted this topic in the wrong forum
I think you are fine, I am actually glad someone started using the matminer list, hopefully this will encourage others. If you have further questions about the workflow you should post in atomate. Probably, you will in the end get me responding no matter where you post
On Sun, Feb 18, 2018 at 6:26 AM, [email protected] wrote:
Hi Anubhav,
thanks for your reply. So in principal the HSE-Bandstructure workflow from atomate can be used in order to generate the bandstructure object, which can than be post-processed in the BranchPointEnergy routine from matminer. I will test it next week. Sorry, if I have posted this topic in the wrong forum. I now realise that my questions were more about the workflow, so the atomate forum was probably better suited for asking this questions.
Best
Hendrik
–
You received this message because you are subscribed to the Google Groups “matminer” group.
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].
To view this discussion on the web visit https://groups.google.com/d/msgid/matminer/c53749e2-2579-49d5-a2f6-682373bd14cf%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
–
Best,
Anubhav