Dear abinit users,
I would like to calculate the complete band structure using hybrid functionals (HSE, PBE0), is this possible in v8.0.8? Can anyone provide a help on the general procedure to follow? That will be much appreciated.
Thanks!
Hazem
band structure using hybrid functionals
Moderator: bguster
Re: band structure using hybrid functionals
Dear Hazem,
From the last release note (http://www.abinit.org/doc/helpfiles/for-v8.0/release_notes/release_notes.html), you have access to hybrid functionals as follows:
Best wishes,
Eric
From the last release note (http://www.abinit.org/doc/helpfiles/for-v8.0/release_notes/release_notes.html), you have access to hybrid functionals as follows:
Code: Select all
B.10 Achievements related to the treatment of the Fock operator, and hybrid functionals :
Hartree-Fock is available with PAW (self-consistent, based on plane waves). However, forces and strains are not yet available.
The Spencer-Alavi scheme if used for Coulomb operator
Test v7#70.
The HSE functional is available in the case of norm-conserving pseudopotentials, via libXC.
Note that forces are available in this case (not yet the strains).
The PBE0 functional is also available in the case of norm-conserving pseudopotentials (via libXC as well as via ixc=41/42).
Work by C. Martins, F. Arnardi, M. Torrent and F. Jollet.Best wishes,
Eric
Re: band structure using hybrid functionals
Dear Eric,
Thank you for your kind reply.
I would like to know on how to technically apply that in band structure calculations.
In the test v7 #t65, one dataset is for a usual SCF run and then a second dataset with ixc = 41 (PBE0), this works fine and gives the total energy.
When I add a third dataset to calculate the band structure (with k-points paths) and getden -1 it gives the DFT band structure from dataset 1, which is not the desired result.
For HSE, I tried just setting ixc=-524 (via libXC) with the two datasets usually needed for BS calculations, the run completes, but I am not sure about the band gap, it does not seem to be corrected as expected. I will need to do better tests.
Does one need to do the calculations it with the gwcalctyp=105/205 (HSE06) or 205/225 (PBE0)? and then for the BS one will probably need to interpolate it using Wannier functions?
Thanks again and best regards,
Hazem
Thank you for your kind reply.
I would like to know on how to technically apply that in band structure calculations.
In the test v7 #t65, one dataset is for a usual SCF run and then a second dataset with ixc = 41 (PBE0), this works fine and gives the total energy.
When I add a third dataset to calculate the band structure (with k-points paths) and getden -1 it gives the DFT band structure from dataset 1, which is not the desired result.
For HSE, I tried just setting ixc=-524 (via libXC) with the two datasets usually needed for BS calculations, the run completes, but I am not sure about the band gap, it does not seem to be corrected as expected. I will need to do better tests.
Does one need to do the calculations it with the gwcalctyp=105/205 (HSE06) or 205/225 (PBE0)? and then for the BS one will probably need to interpolate it using Wannier functions?
Thanks again and best regards,
Hazem
Re: band structure using hybrid functionals
Dear Hazem,
For the time being, it is not possible to restart a non-self-consistent calculation with Hybrid functional (Fock) from a density file. It make sense because Fock operator depends directly on the WF.
In a future realease of Abinit, it will (for sure) be possible to restart from the WF.
Meanwhile, to plot a band structure, your only possibility it to compute manually each k-point (have a whole set of key-points containing the ones you want to plot).
Regards,
For the time being, it is not possible to restart a non-self-consistent calculation with Hybrid functional (Fock) from a density file. It make sense because Fock operator depends directly on the WF.
In a future realease of Abinit, it will (for sure) be possible to restart from the WF.
Meanwhile, to plot a band structure, your only possibility it to compute manually each k-point (have a whole set of key-points containing the ones you want to plot).
Regards,
Marc Torrent
CEA - Bruyères-le-Chatel
France
CEA - Bruyères-le-Chatel
France