My Community

Hi Tom
I have the same "WARNING" problem with you when using Line-modes for Band-structure Calculation, but also perfectly able to do a regular non-band structure HSE06 calculation.

Any suggestion from admin???

[quote="westmonster"]Hi Tom
I have the same "WARNING" problem with you when using Line-modes for Band-structure Calculation, but also perfectly able to do a regular non-band structure HSE06 calculation.

Any suggestion from admin???
[/quote]
It's simple. When you do regular SCF, you sample the brouillon zone by a mesh of k-points. This warrants that all k-points are different.
When doing band structure, for some reason, you use some identical k-point repeatedly.

Solution: If you want to use line-mode, just perform one line at a time.

It is also convenient because:
1. When doing band structure, all k-points are treated independently (no couple between them).
2. HSE06 is extremely expensive. Reducing number of k-point in one calculation will help the convergence and will reduce the required memory.

band structures using hybrid functionals have to be calculated the following way:
-) First perform a selfconsistent Hartree-Fock/HSE calculation using a conventional KPOINTS file.
-) Copy the IBZKPT file to KPOINTS, and explicitely add all desired k-points along high-symmetry lines of the BZ that are needed for the bandstructure plot. Add the points at the end of the KPOINTS file, but set the weights of these added k-points to 0.
-) Do not forget to set the number of k-points in KPOINTS correctly (to the number of the k-points used in the standard mesh PLUS the number of the k-points along the lines), such that all k-points are used for the calculations.
-) NKRED can not be used.
-) Perform a second VASP run:
It is recommended to use the Davidson algorithm, since it converges that eigen energies at the new k-points fastest. Since VASP terminates when the total energy is converged to a certain threshold, it is important to force VASP to do a minimum number of steps, so that the orbitals at the new k-points are fully converged (note: since their weight is zero, they do not contribute to the total energy).
This can be done using e.g.

ALGO = N ; NELMIN = 5 ! Davidson, minimum 5 scf-steps
IBRION = 1 ! Use simple charge mixer, since
Pulay might blow up

-) The KS-eigenvalues of the states along the high-symmetry lines are written in OUTCAR, EIGENVAL, vasprun.xml; please cut the k-points required for the bandstructure from one of these files and proceed as usual (using p4vasp or any other graphics tool you usually use to produce bandstructure-plots)

Note: A Hartee-Fock calculation can NOT be continued from an existing CHGCAR file, since the non-local exchange is not determined by the charge density but by the density matrix and/or the KS-orbitals.

[quote author= N ; NELMIN = 5 ! Davidson, minimum 5 scf-steps
IBRION = 1 ! Use simple charge mixer, since
Pulay might blow up

-) The KS-eigenvalues of the states along the high-symmetry lines are written in OUTCAR, EIGENVAL, vasprun.xml; please cut the k-points required for the bandstructure from one of these files and proceed as usual (using p4vasp or any other graphics tool you usually use to produce bandstructure-plots)

Note: A Hartee-Fock calculation can NOT be continued from an existing CHGCAR file, since the non-local exchange is not determined by the charge density but by the density matrix and/or the KS-orbitals.
[/quote]

Hi, professor ,I have a question about the HSE06 and B3LYP .In the forum ,the administer gives us a suggestion about the computation of the band of the B3LYP and HSE06 ,and his advise is as follows :
1. First do static computation to give birth to the WAVECHAR ,IBZKPT .
2. Band structure is like this :cp IBZKPT to the KPOINTS ,and add the points with high symmetry to the end of the old KPOINTS ,which is also the IBZKPT ,then do a computation .And Read from the WAVECAR of the first step.
My question is like this :
The role of the KPoints which come from the IBZKPT of the first state ,I don’t know how it works for the computation of band structure .Because I ‘m really interested in the kpoints with high symmetry, besides that it costs lots of time to compute the part of the KPONTS (IBZKPT) .It looks as if it was not a good idea.

I have a pinion like this :
1. The first state is same as the above .
2. The second is use quite sparse KPOINTS just to produce the IBZKPT which has less K-points, and when I do computation of the band structure I use this IBZKPT to be the first part of the KPOINTS.
3. Read from the old WAVECAR of the first state , and add points with high symmetry to the end of the IBZKPT . And compute the band .

Does it works ? Waiting for your suggestion !!


Best wishes

Guanzy

USTC

(I'm deleting what I wrote because I posted after only seeing the first page of the thread and didn't see admin's answer.)
<span class='smallblacktext'>[ Edited Fri Jun 04 2010, 09:04AM ]</span>

hmm. I don't know how to delete this post, so I'll just edit away the contents.
Much better contents are already given in my next post.
<span class='smallblacktext'>[ Edited Fri Jun 04 2010, 09:08AM ]</span>

VASP001, the sparse k-point grid will not work for two reasons. First, it will not work (correctly anyway) to read in the WAVECAR when the grid part of your k-poinits is more sparse than it was when the WAVECAR was created. To read the WAVECAR properly you need to have the first N kpoints in the same position as before, where N is the number of kpoints used to create this WAVECAR.

The second reason is that even if the original WAVECAR could be interpolated onto the sparse grid, the calculation would not be any more accurate than starting with the sparse grid. This is because it is a self-consistent calculation. As confirmed by admin's post, there is no way to do a non-self-consistent calc with a hybrid functional. It's all scf because the wavefunctions solved for go directly into the exchange correlation potential.
<span class='smallblacktext'>[ Edited Fri Jun 04 2010, 09:01AM ]</span>

[quote="westmonster"]Hi Tom
I have the same "WARNING" problem with you when using Line-modes for Band-structure Calculation, but also perfectly able to do a regular non-band structure HSE06 calculation.

Any suggestion from admin???
[/quote]
Unfortunately ,I have do 6 tests ,and the points used in the bandstructrues ban be repeated ,it doesn't make any mistake .

[quote="http://moviesonlinefree.biz"]Watch Splice Online Free[/url]

it will not cost more than 1st step if you add 30 more K-points

[quote="admin"]band structures using hybrid functionals have to be calculated the following way:
-) First perform a selfconsistent Hartree-Fock/HSE calculation using a conventional KPOINTS file.
-) Copy the IBZKPT file to KPOINTS, and explicitely add all desired k-points along high-symmetry lines of the BZ that are needed for the bandstructure plot. Add the points at the end of the KPOINTS file, but set the weights of these added k-points to 0.
-) Do not forget to set the number of k-points in KPOINTS correctly (to the number of the k-points used in the standard mesh PLUS the number of the k-points along the lines), such that all k-points are used for the calculations.
-) NKRED can not be used.
-) Perform a second VASP run:
It is recommended to use the Davidson algorithm, since it converges that eigen energies at the new k-points fastest. Since VASP terminates when the total energy is converged to a certain threshold, it is important to force VASP to do a minimum number of steps, so that the orbitals at the new k-points are fully converged (note: since their weight is zero, they do not contribute to the total energy).
This can be done using e.g.

ALGO = N ; NELMIN = 5 ! Davidson, minimum 5 scf-steps
IBRION = 1 ! Use simple charge mixer, since
Pulay might blow up

-) The KS-eigenvalues of the states along the high-symmetry lines are written in OUTCAR, EIGENVAL, vasprun.xml; please cut the k-points required for the bandstructure from one of these files and proceed as usual (using p4vasp or any other graphics tool you usually use to produce bandstructure-plots)

Note: A Hartee-Fock calculation can NOT be continued from an existing CHGCAR file, since the non-local exchange is not determined by the charge density but by the density matrix and/or the KS-orbitals.
[/quote]

Agree with u, maybe a spelling mistake. The note has told us use simple charge mixer, since Pulay might blow up.

I found this:

http://cms.mpi.univie.ac.at/vasp-forum/ ... php?4.9205

where sankh states "Exact exchange energy is not determined through the local charges but through density matrix over KS orbitals, and hence you cannot use ICHARG = 11"

also this:

http://cms.mpi.univie.ac.at/vasp-forum/ ... php?4.5896

Thoughts?

Dear all,

I studied the above recipies and try to do some HSE band of some monolayer system, such TiS2, VS2 etc.

Take TiS2 monolayer as example, at first, I get converged HSE total energy, CHGCAR and WAVECAR etc. using the following INCAR:

SYSTEM = Ti S !TiS2
#
ISTART = 1
ICHARG = 1
PREC = normal
ENCUT = 500

NELM = 200
NELMIN = 6
EDIFF = 1E-5
EDIFFG = -0.02

ADDGRID=.TRUE.
NSW = 0
IBRION = -1
ISIF = 2

ISMEAR = 0
SIGMA = 0.05
ISYM = 3
ISPIN = 1

LORBIT = 11

LHFCALC = .TRUE.
HFSCREEN = 0.2
ALGO = Damped
TIME = 0.4
ENCUTFOCK = 0

AMIX = 0.2
BMIX = 0.0001
AMIX_MAG = 0.8
BMIX_MAG = 0.0001

the POSCAR file is:
Ti S
1.00000000000000
1.7035000000000000 -2.9505485506935822 0.0000000000000000
1.7035000000000000 2.9505485506935822 0.0000000000000000
0.0000000000000000 0.0000000000000000 22.7800000000000000
Ti S
1 2
Direct
0.0000000000000000 0.0000000000000000 0.0000000000000000
0.3333332999999996 0.6666667000000004 0.0625000000000000
0.6666667000000004 0.3333332999999996 -0.0625000000000000

The KPOINTS used is gamma centered 8x8x1.

And then I copy the IBZKPTS to KPOINTS and append the k-path with weight equal to 0, chang the total number of k-points to correct one,
1) if I use both the above converged CHGCAR and WAVECAR, for the k-points in the original IBZKPTS file, the eigenvalues are almost the same to the previous normal HSE calculation to the second digit after the decimal point, but for the eigenvalues for the same k-point with weight equal to 0 in the band calculation itself, sometimes they are different to even the first digit, for example for gamma point:

0.0000000E+00 0.0000000E+00 0.0000000E+00 0.1562500E-01 | 0.0000000E+00 0.0000000E+00 0.0000000E+00 0.0000000E+00
1 -19.207405 | 1 -19.202116
2 -17.954867 | 2 -17.949591
3 -9.362146 | 3 -9.351653
4 -6.835580 | 4 -6.710608
5 -6.835579 | 5 -6.773375
6 -5.791374 | 6 -5.751144
7 -3.505488 | 7 -3.280321
8 -3.505487 | 8 -3.345879
9 -2.690800 | 9 -2.687493
10 -2.690800 | 10 -2.674736
11 -2.132242 | 11 -2.111599
12 0.601448 | 12 0.903134
13 0.601449 | 13 0.917483
14 1.910214 | 14 2.422735
15 2.046518 | 15 2.493371
16 2.427028 | 16 8.648083

2) if i use CHGCAR only, although the eigenvalues for the same k-points with different weights are exactly the same, but for spin polarized case, such as VS2, the eigenvalues are now different with the ones in the previous converged HSE normal calculation, also take gamma point as example:
#vimdiff EIGENVAL band/EIGENVAL
0.0000000E+00 0.0000000E+00 0.0000000E+00 0.1562500E-01 | 0.0000000E+00 0.0000000E+00 0.0000000E+00 0.1562500E-01
1 -19.058755 -19.261436 | 1 -19.026963 -19.292982
2 -17.692630 -17.988281 | 2 -17.670474 -18.039900
3 -9.578964 -9.104816 | 3 -9.346534 -9.118215
4 -6.512799 -6.112080 | 4 -6.423356 -6.017819
5 -6.512798 -6.112080 | 5 -6.367105 -6.017818
6 -5.528627 -5.566813 | 6 -5.480533 -5.600647
7 -3.735544 -1.623466 | 7 -4.782012 -1.270216
8 -3.735544 -1.623466 | 8 -4.525995 -1.270214
9 -4.516553 -2.841982 | 9 -3.280320 -1.275925
10 -2.100262 -2.841982 | 10 -2.037262 -2.908562
11 -2.100263 -0.891297 | 11 -2.024924 -2.908561
12 1.043219 1.231823 | 12 1.148018 1.359465
13 1.043220 1.231824 | 13 1.157030 1.359465
14 2.360690 2.499889 | 14 2.563717 2.478484
15 2.554327 2.584961 | 15 2.392196 2.580055
16 3.039150 3.081379 | 16 3.410338 3.069216
and the Fermi energy from these two calculations are different. The most top eigenvalues are always differ much. So I can not use this trick to spin polarized case, how to do HSE energy bands calculations for spin polarized system?

3) sometimes, the eigenvalues are not in the increasing order, they are mixed, as shown in point 2. Although I can sort it by myself when plot bands, I do not know if the mixed eigenvalues here affect the Fermi energy in OUTCAR.

Anybody here could you please help to have a look at it that if I did something wrong in the above calculations.

Thank you!