Re: [QE-users] Problem with PBE0 calculation
Dear Giuseppe and M.V Kondrin,
Thank you very much for your suggestion and such a fruitful discussion.
Looks like there is no direct way to plot the band structure using a hybrid
functional. However, as you suggested this can be done with some tricks and
techniques. I never heard about the scissor operator which I need to learn
how to use. But as suggested by Mr. MV Konrin, I could get the DOS using
PBE0 which is also very useful for the band gap. Thanks a lot for such
wonderful guidance.
Rameswar
On Thu, Jan 12, 2023 at 5:20 AM mkondrin wrote:
> Dear Giuseppe,
>
> Thank you for very pertinent comment. Of course, I have meant
> stoichiometric materials, not the defect structures. Indeed, in the
> latter case band structure is not well defined notion, but probably DOS
> for some supercell approximation of defect structure is more robust.
>
> Sincerely yours,
> M. V. Kondrin
>
> On 01/12/2023 01:06 PM, Giuseppe Mattioli wrote:
> >
> > Dear M. V. Kondrin
> > The "scissors operator" strategy can be appropriate in the case of
> > stoichiometric materials, but it may be not when you investigate
> > defects or dopants in a host matrix, where the band structure (and the
> > DOS) can depend on the localization of electronic charge on defects
> > that can be seen as "deep" or "shallow" by different GGA or EXX
> > functionals. The best method to use depend on Rameswar's purposes.
> > Best
> > Giuseppe
> >
> > Quoting mkondrin :
> >
> >> I also performed calculations using HSE06 functional for obtaining
> >> electron DOS (using projwfc.x). It is compatible with hybrid
> >> functionals and usually provide correct values of electron bandgaps.
> >> After that I just used "scissors" operation on electron band
> >> structure obtained using standard PBE to match bandgap obtained
> >> previously.
> >>
> >> Approximately in the same words I described this operation in the
> >> recent paper which is under review now.
> >>
> >> Hope this helps.
> >>
> >> Sincerely yours,
> >> M. V. Kondrin
> >>
> >>
> >>
> >> On 01/12/2023 11:03 AM, Giuseppe Mattioli wrote:
> >>>
> >>> AFAIK, nscf calculations are not compatible with EXX functionals.
> >>> There are a few workarounds, like manually adding the band path as
> >>> "fake" k-points with weight=0 to the regular grid obtained by the
> >>> automatic distribution used in the scf calculation.
> >>> I generally perform EXX calculations of isolated or very large
> >>> systems @Gamma. More experienced users may add more accurate
> >>> information.
> >>> HTH
> >>> Giuseppe
> >>>
> >>> Quoting Rameswar Bhattacharjee :
> >>>
> Hi Giuseppe,
> Thank you very much for your detailed reply. I followed the
> instructions
> and was able to complete the SCF calculation. Also, I did an NSCF
> calculation using the same combination you suggested as a pre-step to
> perform the band calculation. My ultimate aim is to get the band
> gap using
> a hybrid functional such as PBE0.
>
> However, I am not able to run the band calculation with this
> combination
> got an error like below:
>
>
> %%
>
> Error in routine exx_grid_init (1):
>
> wrong EXX q grid
>
>
>
> %%
>
> It seems the k-points grid I am using is not suitable for EXX
> calculation
> although I used the same grid for PBE and it was perfect. I am sure
> someone
> in the forum had performed band structure using hybrid functional
> and would
> be able to help me resolve this. Any help would be highly appreciated.
>
> Here is my input for band calculation using PBE0
>
> --
>
> &control
>
> restart_mode='from_scratch'
>
> calculation='bands'
>
> prefix='Ethyne-Anthracene',
>
> pseudo_dir = '/home/rb1820/QE-PP/NC',
>
> outdir='./scratch'
>
> /
>
> &system
>
> ibrav = 0,
>
> tot_charge=0,
>
> nat=36,
>
> ntyp=2,
>
> ecutwfc = 90,
>
> ecutfock= 150,
>
> input_dft='pbe0'
>
> vdw_corr='grimme-d2',
>
> nbnd = 60
>
> /
>
> &electrons
>
> electron_maxstep = 1000
>
> mixing_beta=0.7
>
> diagonalization='david'
>
> diago_david_ndim = 2
>
> conv_thr=1.0d-8
>
> /
>
> &ions
>
> /
>
> &CELL
>
> cell_dynamics='bfgs',
>
> press = 0.0
>
> cell_dofree = 'x',
>
> /
>
> ATOMIC_SPECIES
>
> H 1.008 H_ONCV_PBE-1.2.upf
>
> C 12.0107 C_ONCV_PBE-1.2.upf
>
>
> ATOMIC_POSITIONS {angstrom}
>
> K_POINTS { crystal }
>
> 1
Re: [QE-users] Problem with PBE0 calculation
Dear Giuseppe,
Thank you for very pertinent comment. Of course, I have meant
stoichiometric materials, not the defect structures. Indeed, in the
latter case band structure is not well defined notion, but probably DOS
for some supercell approximation of defect structure is more robust.
Sincerely yours,
M. V. Kondrin
On 01/12/2023 01:06 PM, Giuseppe Mattioli wrote:
Dear M. V. Kondrin
The "scissors operator" strategy can be appropriate in the case of
stoichiometric materials, but it may be not when you investigate
defects or dopants in a host matrix, where the band structure (and the
DOS) can depend on the localization of electronic charge on defects
that can be seen as "deep" or "shallow" by different GGA or EXX
functionals. The best method to use depend on Rameswar's purposes.
Best
Giuseppe
Quoting mkondrin :
I also performed calculations using HSE06 functional for obtaining
electron DOS (using projwfc.x). It is compatible with hybrid
functionals and usually provide correct values of electron bandgaps.
After that I just used "scissors" operation on electron band
structure obtained using standard PBE to match bandgap obtained
previously.
Approximately in the same words I described this operation in the
recent paper which is under review now.
Hope this helps.
Sincerely yours,
M. V. Kondrin
On 01/12/2023 11:03 AM, Giuseppe Mattioli wrote:
AFAIK, nscf calculations are not compatible with EXX functionals.
There are a few workarounds, like manually adding the band path as
"fake" k-points with weight=0 to the regular grid obtained by the
automatic distribution used in the scf calculation.
I generally perform EXX calculations of isolated or very large
systems @Gamma. More experienced users may add more accurate
information.
HTH
Giuseppe
Quoting Rameswar Bhattacharjee :
Hi Giuseppe,
Thank you very much for your detailed reply. I followed the
instructions
and was able to complete the SCF calculation. Also, I did an NSCF
calculation using the same combination you suggested as a pre-step to
perform the band calculation. My ultimate aim is to get the band
gap using
a hybrid functional such as PBE0.
However, I am not able to run the band calculation with this
combination
got an error like below:
%%
Error in routine exx_grid_init (1):
wrong EXX q grid
%%
It seems the k-points grid I am using is not suitable for EXX
calculation
although I used the same grid for PBE and it was perfect. I am sure
someone
in the forum had performed band structure using hybrid functional
and would
be able to help me resolve this. Any help would be highly appreciated.
Here is my input for band calculation using PBE0
--
&control
restart_mode='from_scratch'
calculation='bands'
prefix='Ethyne-Anthracene',
pseudo_dir = '/home/rb1820/QE-PP/NC',
outdir='./scratch'
/
&system
ibrav = 0,
tot_charge=0,
nat=36,
ntyp=2,
ecutwfc = 90,
ecutfock= 150,
input_dft='pbe0'
vdw_corr='grimme-d2',
nbnd = 60
/
&electrons
electron_maxstep = 1000
mixing_beta=0.7
diagonalization='david'
diago_david_ndim = 2
conv_thr=1.0d-8
/
&ions
/
&CELL
cell_dynamics='bfgs',
press = 0.0
cell_dofree = 'x',
/
ATOMIC_SPECIES
H 1.008 H_ONCV_PBE-1.2.upf
C 12.0107 C_ONCV_PBE-1.2.upf
ATOMIC_POSITIONS {angstrom}
K_POINTS { crystal }
11
-0.5 0.0 0.0 1.0
-0.4 0.0 0.0 1.0
-0.3 0.0 0.0 1.0
-0.2 0.0 0.0 1.0
-0.1 0.0 0.0 1.0
0.0 0.0 0.0 1.0
0.1 0.0 0.0 1.0
0.2 0.0 0.0 1.0
0.3 0.0 0.0 1.0
0.4 0.0 0.0 1.0
0.5 0.0 0.0 1.0
CELL_PARAMETERS {angstrom}
6.933462797 0.0 0.0
0.0 20.0 0.0
0.0 0.0 20.0
---
Thanks
Rameswar Bhattacharjee
Georgetown University
On Wed, Jan 11, 2023 at 3:21 AM Giuseppe Mattioli <
giuseppe.matti...@ism.cnr.it> wrote:
Dear Rameswar Bhattacharjee
I don't know if this is the actual source of error, but I strongly
recommend to use norm-conserving pseudopotentials when performing EXX
calculations. You have generally no speed-up with US because ecutrho
is more or less often the same, and in your case you are using
tremendously high (and likely useless) values for ecutwfc and
ecutrho.
Moreover, using NC you can finely tune the density basis set used to
build Fock integrals with ecutfock, otherwise stuck to the ecutrho
value. This combination of values is probably responsible for a very
high memory request (the code prints an estimate at the beginning of
the output). For example, *using C and H ONCV pseudopotentials*, you
should be on the reasonably accurate side using
ecutwfc=90.0
ecutfock=135.0~180.0
ecutrho is automatically fixed at ecutwfc*4=360.0 in th
Re: [QE-users] Problem with PBE0 calculation
Dear M. V. Kondrin
The "scissors operator" strategy can be appropriate in the case of
stoichiometric materials, but it may be not when you investigate
defects or dopants in a host matrix, where the band structure (and the
DOS) can depend on the localization of electronic charge on defects
that can be seen as "deep" or "shallow" by different GGA or EXX
functionals. The best method to use depend on Rameswar's purposes.
Best
Giuseppe
Quoting mkondrin :
I also performed calculations using HSE06 functional for obtaining
electron DOS (using projwfc.x). It is compatible with hybrid
functionals and usually provide correct values of electron bandgaps.
After that I just used "scissors" operation on electron band
structure obtained using standard PBE to match bandgap obtained
previously.
Approximately in the same words I described this operation in the
recent paper which is under review now.
Hope this helps.
Sincerely yours,
M. V. Kondrin
On 01/12/2023 11:03 AM, Giuseppe Mattioli wrote:
AFAIK, nscf calculations are not compatible with EXX functionals.
There are a few workarounds, like manually adding the band path as
"fake" k-points with weight=0 to the regular grid obtained by the
automatic distribution used in the scf calculation.
I generally perform EXX calculations of isolated or very large
systems @Gamma. More experienced users may add more accurate
information.
HTH
Giuseppe
Quoting Rameswar Bhattacharjee :
Hi Giuseppe,
Thank you very much for your detailed reply. I followed the instructions
and was able to complete the SCF calculation. Also, I did an NSCF
calculation using the same combination you suggested as a pre-step to
perform the band calculation. My ultimate aim is to get the band gap using
a hybrid functional such as PBE0.
However, I am not able to run the band calculation with this combination
got an error like below:
%% Error in routine exx_grid_init
(1):
wrong EXX q grid
%% It seems the k-points grid I am using is not suitable for EXX
calculation
although I used the same grid for PBE and it was perfect. I am sure someone
in the forum had performed band structure using hybrid functional and would
be able to help me resolve this. Any help would be highly appreciated.
Here is my input for band calculation using PBE0
--
&control
restart_mode='from_scratch'
calculation='bands'
prefix='Ethyne-Anthracene',
pseudo_dir = '/home/rb1820/QE-PP/NC',
outdir='./scratch'
/
&system
ibrav = 0,
tot_charge=0,
nat=36,
ntyp=2,
ecutwfc = 90,
ecutfock= 150,
input_dft='pbe0'
vdw_corr='grimme-d2',
nbnd = 60
/
&electrons
electron_maxstep = 1000
mixing_beta=0.7
diagonalization='david'
diago_david_ndim = 2
conv_thr=1.0d-8
/
&ions
/
&CELL
cell_dynamics='bfgs',
press = 0.0
cell_dofree = 'x',
/
ATOMIC_SPECIES
H 1.008 H_ONCV_PBE-1.2.upf
C 12.0107 C_ONCV_PBE-1.2.upf
ATOMIC_POSITIONS {angstrom}
K_POINTS { crystal }
11
-0.5 0.0 0.0 1.0
-0.4 0.0 0.0 1.0
-0.3 0.0 0.0 1.0
-0.2 0.0 0.0 1.0
-0.1 0.0 0.0 1.0
0.0 0.0 0.0 1.0
0.1 0.0 0.0 1.0
0.2 0.0 0.0 1.0
0.3 0.0 0.0 1.0
0.4 0.0 0.0 1.0
0.5 0.0 0.0 1.0
CELL_PARAMETERS {angstrom}
6.933462797 0.0 0.0
0.0 20.0 0.0
0.0 0.0 20.0
---
Thanks
Rameswar Bhattacharjee
Georgetown University
On Wed, Jan 11, 2023 at 3:21 AM Giuseppe Mattioli <
giuseppe.matti...@ism.cnr.it> wrote:
Dear Rameswar Bhattacharjee
I don't know if this is the actual source of error, but I strongly
recommend to use norm-conserving pseudopotentials when performing EXX
calculations. You have generally no speed-up with US because ecutrho
is more or less often the same, and in your case you are using
tremendously high (and likely useless) values for ecutwfc and ecutrho.
Moreover, using NC you can finely tune the density basis set used to
build Fock integrals with ecutfock, otherwise stuck to the ecutrho
value. This combination of values is probably responsible for a very
high memory request (the code prints an estimate at the beginning of
the output). For example, *using C and H ONCV pseudopotentials*, you
should be on the reasonably accurate side using
ecutwfc=90.0
ecutfock=135.0~180.0
ecutrho is automatically fixed at ecutwfc*4=360.0 in the case of NC
HTH
Giuseppe
Quoting Rameswar Bhattacharjee :
Hi Everyone,
I am trying to get a band structure using a hybrid functional. As
suggested
previously in this forum I choose PBE0 functional with pbe-ultrasoft
pseudopotential. The calculation starts fine and scf iteration converges
smoothly. But the job is stuck and does not complete with the last
Re: [QE-users] Problem with PBE0 calculation
I also performed calculations using HSE06 functional for obtaining
electron DOS (using projwfc.x). It is compatible with hybrid functionals
and usually provide correct values of electron bandgaps. After that I
just used "scissors" operation on electron band structure obtained using
standard PBE to match bandgap obtained previously.
Approximately in the same words I described this operation in the recent
paper which is under review now.
Hope this helps.
Sincerely yours,
M. V. Kondrin
On 01/12/2023 11:03 AM, Giuseppe Mattioli wrote:
AFAIK, nscf calculations are not compatible with EXX functionals.
There are a few workarounds, like manually adding the band path as
"fake" k-points with weight=0 to the regular grid obtained by the
automatic distribution used in the scf calculation.
I generally perform EXX calculations of isolated or very large systems
@Gamma. More experienced users may add more accurate information.
HTH
Giuseppe
Quoting Rameswar Bhattacharjee :
Hi Giuseppe,
Thank you very much for your detailed reply. I followed the instructions
and was able to complete the SCF calculation. Also, I did an NSCF
calculation using the same combination you suggested as a pre-step to
perform the band calculation. My ultimate aim is to get the band gap
using
a hybrid functional such as PBE0.
However, I am not able to run the band calculation with this
combination
got an error like below:
%%
Error in routine exx_grid_init (1):
wrong EXX q grid
%%
It seems the k-points grid I am using is not suitable for EXX
calculation
although I used the same grid for PBE and it was perfect. I am sure
someone
in the forum had performed band structure using hybrid functional and
would
be able to help me resolve this. Any help would be highly appreciated.
Here is my input for band calculation using PBE0
--
&control
restart_mode='from_scratch'
calculation='bands'
prefix='Ethyne-Anthracene',
pseudo_dir = '/home/rb1820/QE-PP/NC',
outdir='./scratch'
/
&system
ibrav = 0,
tot_charge=0,
nat=36,
ntyp=2,
ecutwfc = 90,
ecutfock= 150,
input_dft='pbe0'
vdw_corr='grimme-d2',
nbnd = 60
/
&electrons
electron_maxstep = 1000
mixing_beta=0.7
diagonalization='david'
diago_david_ndim = 2
conv_thr=1.0d-8
/
&ions
/
&CELL
cell_dynamics='bfgs',
press = 0.0
cell_dofree = 'x',
/
ATOMIC_SPECIES
H 1.008 H_ONCV_PBE-1.2.upf
C 12.0107 C_ONCV_PBE-1.2.upf
ATOMIC_POSITIONS {angstrom}
K_POINTS { crystal }
11
-0.5 0.0 0.0 1.0
-0.4 0.0 0.0 1.0
-0.3 0.0 0.0 1.0
-0.2 0.0 0.0 1.0
-0.1 0.0 0.0 1.0
0.0 0.0 0.0 1.0
0.1 0.0 0.0 1.0
0.2 0.0 0.0 1.0
0.3 0.0 0.0 1.0
0.4 0.0 0.0 1.0
0.5 0.0 0.0 1.0
CELL_PARAMETERS {angstrom}
6.933462797 0.0 0.0
0.0 20.0 0.0
0.0 0.0 20.0
---
Thanks
Rameswar Bhattacharjee
Georgetown University
On Wed, Jan 11, 2023 at 3:21 AM Giuseppe Mattioli <
giuseppe.matti...@ism.cnr.it> wrote:
Dear Rameswar Bhattacharjee
I don't know if this is the actual source of error, but I strongly
recommend to use norm-conserving pseudopotentials when performing EXX
calculations. You have generally no speed-up with US because ecutrho
is more or less often the same, and in your case you are using
tremendously high (and likely useless) values for ecutwfc and ecutrho.
Moreover, using NC you can finely tune the density basis set used to
build Fock integrals with ecutfock, otherwise stuck to the ecutrho
value. This combination of values is probably responsible for a very
high memory request (the code prints an estimate at the beginning of
the output). For example, *using C and H ONCV pseudopotentials*, you
should be on the reasonably accurate side using
ecutwfc=90.0
ecutfock=135.0~180.0
ecutrho is automatically fixed at ecutwfc*4=360.0 in the case of NC
HTH
Giuseppe
Quoting Rameswar Bhattacharjee :
> Hi Everyone,
> I am trying to get a band structure using a hybrid functional. As
suggested
> previously in this forum I choose PBE0 functional with pbe-ultrasoft
> pseudopotential. The calculation starts fine and scf iteration
converges
> smoothly. But the job is stuck and does not complete with the last
line
"ACE
> projected onto60 (nbndproj) and applied to60 (nbnd)
bands". I am
> not sure what mistake I am making here or if it is just a memory
issue as
> no error message is printing. Any suggestion would be highly
appreciated. I
> am providing last few lines of the output and my input str for
reference.
> Thank you.
>
> Last few lines of the scf output
> ---
>
> highest occupied, lowest
Re: [QE-users] Problem with PBE0 calculation
AFAIK, nscf calculations are not compatible with EXX functionals.
There are a few workarounds, like manually adding the band path as
"fake" k-points with weight=0 to the regular grid obtained by the
automatic distribution used in the scf calculation.
I generally perform EXX calculations of isolated or very large systems
@Gamma. More experienced users may add more accurate information.
HTH
Giuseppe
Quoting Rameswar Bhattacharjee :
Hi Giuseppe,
Thank you very much for your detailed reply. I followed the instructions
and was able to complete the SCF calculation. Also, I did an NSCF
calculation using the same combination you suggested as a pre-step to
perform the band calculation. My ultimate aim is to get the band gap using
a hybrid functional such as PBE0.
However, I am not able to run the band calculation with this combination
got an error like below:
%%
Error in routine exx_grid_init (1):
wrong EXX q grid
%%
It seems the k-points grid I am using is not suitable for EXX calculation
although I used the same grid for PBE and it was perfect. I am sure someone
in the forum had performed band structure using hybrid functional and would
be able to help me resolve this. Any help would be highly appreciated.
Here is my input for band calculation using PBE0
--
&control
restart_mode='from_scratch'
calculation='bands'
prefix='Ethyne-Anthracene',
pseudo_dir = '/home/rb1820/QE-PP/NC',
outdir='./scratch'
/
&system
ibrav = 0,
tot_charge=0,
nat=36,
ntyp=2,
ecutwfc = 90,
ecutfock= 150,
input_dft='pbe0'
vdw_corr='grimme-d2',
nbnd = 60
/
&electrons
electron_maxstep = 1000
mixing_beta=0.7
diagonalization='david'
diago_david_ndim = 2
conv_thr=1.0d-8
/
&ions
/
&CELL
cell_dynamics='bfgs',
press = 0.0
cell_dofree = 'x',
/
ATOMIC_SPECIES
H 1.008 H_ONCV_PBE-1.2.upf
C 12.0107 C_ONCV_PBE-1.2.upf
ATOMIC_POSITIONS {angstrom}
K_POINTS { crystal }
11
-0.5 0.0 0.0 1.0
-0.4 0.0 0.0 1.0
-0.3 0.0 0.0 1.0
-0.2 0.0 0.0 1.0
-0.1 0.0 0.0 1.0
0.0 0.0 0.0 1.0
0.1 0.0 0.0 1.0
0.2 0.0 0.0 1.0
0.3 0.0 0.0 1.0
0.4 0.0 0.0 1.0
0.5 0.0 0.0 1.0
CELL_PARAMETERS {angstrom}
6.933462797 0.0 0.0
0.0 20.0 0.0
0.0 0.0 20.0
---
Thanks
Rameswar Bhattacharjee
Georgetown University
On Wed, Jan 11, 2023 at 3:21 AM Giuseppe Mattioli <
giuseppe.matti...@ism.cnr.it> wrote:
Dear Rameswar Bhattacharjee
I don't know if this is the actual source of error, but I strongly
recommend to use norm-conserving pseudopotentials when performing EXX
calculations. You have generally no speed-up with US because ecutrho
is more or less often the same, and in your case you are using
tremendously high (and likely useless) values for ecutwfc and ecutrho.
Moreover, using NC you can finely tune the density basis set used to
build Fock integrals with ecutfock, otherwise stuck to the ecutrho
value. This combination of values is probably responsible for a very
high memory request (the code prints an estimate at the beginning of
the output). For example, *using C and H ONCV pseudopotentials*, you
should be on the reasonably accurate side using
ecutwfc=90.0
ecutfock=135.0~180.0
ecutrho is automatically fixed at ecutwfc*4=360.0 in the case of NC
HTH
Giuseppe
Quoting Rameswar Bhattacharjee :
> Hi Everyone,
> I am trying to get a band structure using a hybrid functional. As
suggested
> previously in this forum I choose PBE0 functional with pbe-ultrasoft
> pseudopotential. The calculation starts fine and scf iteration converges
> smoothly. But the job is stuck and does not complete with the last line
"ACE
> projected onto60 (nbndproj) and applied to60 (nbnd) bands". I am
> not sure what mistake I am making here or if it is just a memory issue as
> no error message is printing. Any suggestion would be highly
appreciated. I
> am providing last few lines of the output and my input str for reference.
> Thank you.
>
> Last few lines of the scf output
> ---
>
> highest occupied, lowest unoccupied level (ev):-3.5627 -3.2294
>
>
> !total energy =-312.15802972 Ry
>
> estimated scf accuracy< 3.2E-09 Ry
>
>
> convergence has been achieved in 12 iterations
>
>
> Using ACE for calculation of exact exchange
>
>
> EXX grid: 4188907 G-vectors FFT dimensions: ( 100, 288, 288)
>
> ACE projected onto60 (nbndproj) and applied to60 (nbnd)
bands
> --
>
> Input str
> --
>
> &control
>
> restart_mode='from_scratch'
>
>
Re: [QE-users] Problem with PBE0 calculation
Hi Giuseppe,
Thank you very much for your detailed reply. I followed the instructions
and was able to complete the SCF calculation. Also, I did an NSCF
calculation using the same combination you suggested as a pre-step to
perform the band calculation. My ultimate aim is to get the band gap using
a hybrid functional such as PBE0.
However, I am not able to run the band calculation with this combination
got an error like below:
%%
Error in routine exx_grid_init (1):
wrong EXX q grid
%%
It seems the k-points grid I am using is not suitable for EXX calculation
although I used the same grid for PBE and it was perfect. I am sure someone
in the forum had performed band structure using hybrid functional and would
be able to help me resolve this. Any help would be highly appreciated.
Here is my input for band calculation using PBE0
--
&control
restart_mode='from_scratch'
calculation='bands'
prefix='Ethyne-Anthracene',
pseudo_dir = '/home/rb1820/QE-PP/NC',
outdir='./scratch'
/
&system
ibrav = 0,
tot_charge=0,
nat=36,
ntyp=2,
ecutwfc = 90,
ecutfock= 150,
input_dft='pbe0'
vdw_corr='grimme-d2',
nbnd = 60
/
&electrons
electron_maxstep = 1000
mixing_beta=0.7
diagonalization='david'
diago_david_ndim = 2
conv_thr=1.0d-8
/
&ions
/
&CELL
cell_dynamics='bfgs',
press = 0.0
cell_dofree = 'x',
/
ATOMIC_SPECIES
H 1.008 H_ONCV_PBE-1.2.upf
C 12.0107 C_ONCV_PBE-1.2.upf
ATOMIC_POSITIONS {angstrom}
K_POINTS { crystal }
11
-0.5 0.0 0.0 1.0
-0.4 0.0 0.0 1.0
-0.3 0.0 0.0 1.0
-0.2 0.0 0.0 1.0
-0.1 0.0 0.0 1.0
0.0 0.0 0.0 1.0
0.1 0.0 0.0 1.0
0.2 0.0 0.0 1.0
0.3 0.0 0.0 1.0
0.4 0.0 0.0 1.0
0.5 0.0 0.0 1.0
CELL_PARAMETERS {angstrom}
6.933462797 0.0 0.0
0.0 20.0 0.0
0.0 0.0 20.0
---
Thanks
Rameswar Bhattacharjee
Georgetown University
On Wed, Jan 11, 2023 at 3:21 AM Giuseppe Mattioli <
giuseppe.matti...@ism.cnr.it> wrote:
>
> Dear Rameswar Bhattacharjee
> I don't know if this is the actual source of error, but I strongly
> recommend to use norm-conserving pseudopotentials when performing EXX
> calculations. You have generally no speed-up with US because ecutrho
> is more or less often the same, and in your case you are using
> tremendously high (and likely useless) values for ecutwfc and ecutrho.
> Moreover, using NC you can finely tune the density basis set used to
> build Fock integrals with ecutfock, otherwise stuck to the ecutrho
> value. This combination of values is probably responsible for a very
> high memory request (the code prints an estimate at the beginning of
> the output). For example, *using C and H ONCV pseudopotentials*, you
> should be on the reasonably accurate side using
>
> ecutwfc=90.0
> ecutfock=135.0~180.0
> ecutrho is automatically fixed at ecutwfc*4=360.0 in the case of NC
>
> HTH
> Giuseppe
>
> Quoting Rameswar Bhattacharjee :
>
> > Hi Everyone,
> > I am trying to get a band structure using a hybrid functional. As
> suggested
> > previously in this forum I choose PBE0 functional with pbe-ultrasoft
> > pseudopotential. The calculation starts fine and scf iteration converges
> > smoothly. But the job is stuck and does not complete with the last line
> "ACE
> > projected onto60 (nbndproj) and applied to60 (nbnd) bands". I am
> > not sure what mistake I am making here or if it is just a memory issue as
> > no error message is printing. Any suggestion would be highly
> appreciated. I
> > am providing last few lines of the output and my input str for reference.
> > Thank you.
> >
> > Last few lines of the scf output
> > ---
> >
> > highest occupied, lowest unoccupied level (ev):-3.5627 -3.2294
> >
> >
> > !total energy =-312.15802972 Ry
> >
> > estimated scf accuracy< 3.2E-09 Ry
> >
> >
> > convergence has been achieved in 12 iterations
> >
> >
> > Using ACE for calculation of exact exchange
> >
> >
> > EXX grid: 4188907 G-vectors FFT dimensions: ( 100, 288, 288)
> >
> > ACE projected onto60 (nbndproj) and applied to60 (nbnd)
> bands
> > --
> >
> > Input str
> > --
> >
> > &control
> >
> > restart_mode='from_scratch'
> >
> > calculation='scf'
> >
> > prefix='Test',
> >
> > pseudo_dir = '/home/QE-PP',
> >
> > outdir='./scratch'
> >
> > /
> >
> > &system
> >
> > ibrav = 0,
> >
> >tot_charge=0,
> >
> > nat=36,
> >
> > ntyp=2,
> >
> > ecutwfc = 140,
> >
> > ecutrho = 1400.0,
> >
> > input_dft='pbe0'
> >
> > v
Re: [QE-users] Problem with PBE0 calculation
Dear Rameswar Bhattacharjee
I don't know if this is the actual source of error, but I strongly
recommend to use norm-conserving pseudopotentials when performing EXX
calculations. You have generally no speed-up with US because ecutrho
is more or less often the same, and in your case you are using
tremendously high (and likely useless) values for ecutwfc and ecutrho.
Moreover, using NC you can finely tune the density basis set used to
build Fock integrals with ecutfock, otherwise stuck to the ecutrho
value. This combination of values is probably responsible for a very
high memory request (the code prints an estimate at the beginning of
the output). For example, *using C and H ONCV pseudopotentials*, you
should be on the reasonably accurate side using
ecutwfc=90.0
ecutfock=135.0~180.0
ecutrho is automatically fixed at ecutwfc*4=360.0 in the case of NC
HTH
Giuseppe
Quoting Rameswar Bhattacharjee :
Hi Everyone,
I am trying to get a band structure using a hybrid functional. As suggested
previously in this forum I choose PBE0 functional with pbe-ultrasoft
pseudopotential. The calculation starts fine and scf iteration converges
smoothly. But the job is stuck and does not complete with the last line "ACE
projected onto60 (nbndproj) and applied to60 (nbnd) bands". I am
not sure what mistake I am making here or if it is just a memory issue as
no error message is printing. Any suggestion would be highly appreciated. I
am providing last few lines of the output and my input str for reference.
Thank you.
Last few lines of the scf output
---
highest occupied, lowest unoccupied level (ev):-3.5627 -3.2294
!total energy =-312.15802972 Ry
estimated scf accuracy< 3.2E-09 Ry
convergence has been achieved in 12 iterations
Using ACE for calculation of exact exchange
EXX grid: 4188907 G-vectors FFT dimensions: ( 100, 288, 288)
ACE projected onto60 (nbndproj) and applied to60 (nbnd) bands
--
Input str
--
&control
restart_mode='from_scratch'
calculation='scf'
prefix='Test',
pseudo_dir = '/home/QE-PP',
outdir='./scratch'
/
&system
ibrav = 0,
tot_charge=0,
nat=36,
ntyp=2,
ecutwfc = 140,
ecutrho = 1400.0,
input_dft='pbe0'
vdw_corr='grimme-d2',
nbnd = 60
/
&electrons
electron_maxstep = 1000
mixing_beta=0.7
diagonalization='david'
diago_david_ndim = 2
conv_thr=1.0d-8
/
&ions
/
&CELL
cell_dynamics='bfgs',
press = 0.0
cell_dofree = 'x',
/
ATOMIC_SPECIES
H 1.008 H.pbe-rrkjus_psl.1.0.0.UPF
C 12.0107 C.pbe-n-rrkjus_psl.1.0.0.UPF
ATOMIC_POSITIONS {angstrom}
K_POINTS automatic
12 1 1 0 0 0
CELL_PARAMETERS {angstrom}
6.933462797 0.0 0.0
0.0 20.0 0.0
0.0 0.0 20.0
Rameswar Bhattacharjee
Georgetown University
GIUSEPPE MATTIOLI
CNR - ISTITUTO DI STRUTTURA DELLA MATERIA
Via Salaria Km 29,300 - C.P. 10
I-00015 - Monterotondo Scalo (RM)
Mob (*preferred*) +39 373 7305625
Tel + 39 06 90672342 - Fax +39 06 90672316
E-mail:
___
The Quantum ESPRESSO community stands by the Ukrainian
people and expresses its concerns about the devastating
effects that the Russian military offensive has on their
country and on the free and peaceful scientific, cultural,
and economic cooperation amongst peoples
___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users
