HSE06 band structure

[zoowe]

I want to calculate the band structure. System has hexagonal structure.

I perform the same way as with regular LDA/GGA. (I know how to do it)

But after a long run (with ICHAGE=11), after many electronic iteration, It always ended up with the error:

LAPACK: Routine ZPOTRF failed!

Anyone has any tip? I think there is an unconventional way to make KPOINTS file.

Thank you for any input!

[tommy91779]

zoowe,

I too have been struggling with this issue. However, I cannont even get through any electronic iterations!! I was wondering if you would be willing to share your INCAR and KPOINTS file for the band structure calculation using HSE06 along with your general procedure. In particular, whether you used WAVECAR from DFT bandstructure calculation or from the self consistent HSE calculation as well as from which calculation you used for the CHGCAR file. In that way we might be able to tell if it is only a LAPACK issue. By the way, would you be willing to tell me what you used for your BLAS and LAPACK routines (i.e. MKL, netlib, etc)

-Tom

[zoowe]

It always takes lot of time in order to get into 1st electronic iteration. So I think you should be patient to wait for it.

I use MKL10, ifort 10.1. I am going to use LAPACK from vasp_lib to test.

Sure:
1. I don't think I need to show INCAR and KPOINTS for PBE. (To get initial WAVECAR for HSE)
2. HSE06 scf:
INCAR:
ISTART = 1
ICHARG = 2
EDIFF = 0.0001
#EDIFFG = -0.01
ENCUT = 450
#ENAUG = 800
LREAL = .FALSE.
#ISPIN = 2
#MAGMOM = 80*0 16*1.7 16*1.8 1*0
#PREC = ACCURATE

LHFCALC = .TRUE. ; HFSCREEN = 0.2
ALGO = All ; TIME = 0.4
ENCUTFOCK = 0
NKRED = 2

#LSDA+U
#LDAU = .TRUE.
#LDAUTYPE = 2
#LDAUL = 2 -1
#LDAUPRINT = 0
#LDAUU = 5.2 0
#LDAUJ = 1.0 0
#NWRITE = 3

#RWIGS =1.323 0.82
#NPAR=1
#LORBIT = 11

LWAVE = .FALSE.
LCHARG = .TRUE.
NELM = 200
NSW = 0
IBRION = -1
LMAXMIX = 4
ISMEAR = 0
SIGMA = 0.1

# IMPROVE ELECTRONIC CONVERGENCY
NSIM = 4
IALGO = 48
#AMIX = 0.2
#BMIX = 0.0001
#AMIX_MAG = 0.8

KPOINTS:
Auto
0
G
8 8 6
0 0 0

3. For band:
INCAR:
ISTART = 0
ISYM = 0
ICHARG = 11
EDIFF = 0.0001
#EDIFFG = -0.01
ENCUT = 450
#ENAUG = 800
LREAL = .FALSE.
#ISPIN = 2
#MAGMOM = 80*0 16*1.7 16*1.8 1*0
#PREC = ACCURATE

LHFCALC = .TRUE. ; HFSCREEN = 0.2
ALGO = All ; TIME = 0.4
#ENCUTFOCK = 0
#NKRED = 2

#LSDA+U
#LDAU = .TRUE.
#LDAUTYPE = 2
#LDAUL = 2 -1
#LDAUPRINT = 0
#LDAUU = 5.2 0
#LDAUJ = 1.0 0
#NWRITE = 3

#RWIGS =1.323 0.82
#NPAR=1
LORBIT = 11

LWAVE = .FALSE.
LCHARG = .FALSE.
NELM = 200
NSW = 0
IBRION = -1
LMAXMIX = 4
ISMEAR = 0
SIGMA = 0.1

# IMPROVE ELECTRONIC CONVERGENCY
NSIM = 4
IALGO = 48
#AMIX = 0.2
#BMIX = 0.0001
#AMIX_MAG = 0.8

KPOINTS:
K_POINTS crystal
30
rec
0.3276803390 0.3361549153 0.5000000000 1.0
0.3163810169 0.3418047458 0.5000000000 1.0
0.3050816949 0.3474545763 0.5000000000 1.0
0.2937823729 0.3531044068 0.5000000000 1.0
0.2824830508 0.3587542373 0.5000000000 1.0
0.2711837288 0.3644040678 0.5000000000 1.0
.......

[VASP001]

YOU can check this message from the source code 'IF (INFO/=0) THEN
WRITE(*,*) 'LAPACK: Routine ZPOTRF failed!',INFO '

And it may help you ! This error can be found in the subprogram called "choleski2.F " to get more information about this error ! No need to gab about it

[panda]

When I do band structure calculations I first find the exact space group number of my system and for that system input the KPOINTS as groups of points located in high symmetry zones like this:

k-points along high symmetry lines
20 ! 20 intersections
Line-mode
rec
0.0 0.0 0.0 ! gamma
0.5293375114 0.0 0.0 ! K

0.5293375114 0.0 0.0 ! K
0.3974005341 0.2294393053 0.0 ! M

0.3974005341 0.2294393053 0.0 ! M
0.0 0.0 0.0 ! gamma

0.0 0.0 0.0 ! gamma
0.0 0.0 0.913624129 ! A

0.0 0.0 0.913624129 ! A
0.5293375114 0.0 0.913624129 ! H

0.5293375114 0.0 0.913624129 ! H
0.3974005341 0.2294393053 0.913624129 ! L

0.3974005341 0.2294393053 0.913624129 ! L
0.0 0.0 0.913624129 ! A

0.0 0.0 0.913624129 ! A
0.0 0.0 0.0 ! gamma

You might find this resource helpful in determining your kpoints for a hexagonal system:

http://cst-www.nrl.navy.mil/users/mehl/ ... index.html

Remember that a high symmetry line in reciprocal space is a physical distance and so for a line that is longer in real space the reciprocal space length will be shorter. The navy site should clear things up for you, because you can see how the vectors are calculated. You should go through and calculate the vectors for your system by hand at first to make sure you understand what is going on, I think the error is that your K-space mesh for band structure calculations is just not correct and this is why you are getting the error, because the iterations are not converging because you are not in the correct region in reciprocal space.

Also, can we please be a bit nicer to one another? zoowe posted his question as well as his input, so telling him no need to gab (which in English is slang for talk) about it is kind of rude, because obviously if he already knew the solution to his problem he would not need to ask the question in the first place. This purpose of the forum is for us to help one another, not to make other people feel stupid.

[zoowe]

Thank you Panda for the input. However, I bet you are not understanding my question.

I think I need input from someone who ACTUALLY did the band structure with HSE06 functional.

I've been working with VASP for more than 5 years so I do know how to make the correct KPOINTS file (required by vasp)

About the idea of VASP001, I did check the code, and it turned out that I need to deal with ZPOTRF routine of LAPACK. It seems to be lot of work.

[panda]

I did not mean to suggest you did not know what you were doing, and there is no way that I can know you have been using VASP for five years since you only recently joined the forums, and I felt like you were being given a hard time for simply asking a question.

In your original post you never even mentioned the HSW06 functional, so you need to be clear. You simply asked about the band structure and the KPOINTS. Specifically, you stated "Anyone has any tip? I think there is an unconventional way to make KPOINTS file." which is why I responded regarding the fact that perhaps your KPOINTS needed to be checked.

I have tried all of the functionals and pseudopotentials available in V5.2 and have not experienced this error, I am using the mkl 10.3 with intel compiler.

[panda]

Oh, just saw the topic of your post! Which is why you did not state the functional you were using, sorry I missed that when I was reading! Just ignore me, long weekend...

Do you get the same problem when using mkl and LAPACK or only when using strictly LAPACK? I would not think going into the LAPACK routine to change things is the way to go...I agree that seems to be a lot of work. I will try running this specific functional for a test with band structure this week and let you know what happens.

Best,

Amanda

[tommy91779]

Zoowe,


I have tried your input files and I am having the same problems I was having before, namely when I attempt to do a HSW06 calculations using the appropriate reciprocal space coordinates in Line Mode I receive the following error:

distr: one band on 1 nodes, 8 groups
vasp.5.2.2 15Apr09 complex
POSCAR found : 2 types and 4 ions
LDA part: xc-table for Ceperly-Alder, standard interpolation
error in IBZKPT_HF: two k-points are equivalent 1 40
this will cause problems in the HF routine
error in IBZKPT_HF: two k-points are equivalent 1 40
this will cause problems in the HF routine
error in IBZKPT_HF: two k-points are equivalent 1 40
this will cause problems in the HF routine
error in IBZKPT_HF: two k-points are equivalent 1 40
this will cause problems in the HF routine
error in IBZKPT_HF: two k-points are equivalent 1 40
this will cause problems in the HF routine

-----------------------------------------------------------------------------
| |
| W W AA RRRRR N N II N N GGGG !!! |
| W W A A R R NN N II NN N G G !!! |
| W W A A R R N N N II N N N G !!! |
| W WW W AAAAAA RRRRR N N N II N N N G GGG ! |
| WW WW A A R R N NN II N NN G G |
| W W A A R R N N II N N GGGG !!! |
| |
| Your generating k-point grid is not commensurate to the symmetry |
| of the lattice. This can cause slow convergence with respect |
| to k-points for HF type calculations |
| suggested SOLUTIONS: |
| ) if not already the case, use automatic k-point generation |
| ) shift your grid to Gamma (G) (e.g. required for hex or fcc lattice) |
| |
-----------------------------------------------------------------------------

error in IBZKPT_HF: two k-points are equivalent 1 40
this will cause problems in the HF routine
error in IBZKPT_HF: two k-points are equivalent 1 40
this will cause problems in the HF routine
error in IBZKPT_HF: two k-points are equivalent 1 40
this will cause problems in the HF routine

-----------------------------------------------------------------------

Further I have included my KPOINTS and INCAR file for my HSE band structure calculation

SYSTEM = CdSe

ISTART = 1

LHFCALC = .TRUE.
HFSCREEN = 0.2

NBANDS = 24
ALGO = All
TIME = 0.4

! ENCUTFOCK = 0
! NKRED = 2

NSW = 0

! LREAL = .FALSE.

ICHARG = 11
LORBIT = 12
ISMEAR = 0
SIGMA = 0.02


-----------------------------------------------------------------------

Hexagonal BZ
10 ! 10 intersections
Line-mode
rec
0.000 0.000 0.500 ! A
0.000 0.500 0.500 ! L

0.000 0.500 0.500 ! L
0.000 0.500 0.000 ! M

0.000 0.500 0.000 ! M
0.000 0.000 0.000 ! Gamma

0.000 0.000 0.000 ! Gamma
0.000 0.000 0.500 ! A

0.000 0.000 0.500 ! A
0.333 0.333 0.500 ! H

0.333 0.333 0.500 ! H
0.333 0.333 0.000 ! K

0.333 0.333 0.000 ! K
0.000 0.000 0.000 ! Gamma


It seems to me based on your error, you should try to link your VASP agains the MKL library used for BLAS. While this will depend on your version, for mine I do the following:

LAPACK=-L/opt/intel/mkl/10.2.1.017/lib/em64t -lmkl_intel_lp64 -lmkl_sequential -lmkl_core -lpthread

Panda,


From your email I get the sense that you have been able to do HSE band structure calculations. Any thoughts on what I am doing wrong? I am perfectly able to do a regular non-band structure HSE calculation without issue.


Thanks in advance,

-Tom

[zoowe]

Hi Tom,
You should have ISYM = 0 in order to avoid the mentioned problem.


<span class='smallblacktext'>[ Edited Mon Jan 25 2010, 09:30PM ]</span>

[tommy91779]

zoowe,

Thanks for the suggestion. I have tried it and unfortunately, I still have the same problem. I am quite at my wits end with this issue and would welcome any suggestions on the aforementioned problem,

Thanks in advance

-Tom

[zoowe]

It will work!!!
The problem is from your KPOINTS file. Unlike regular LDA/GGA band calculation, you should not do 2 or more directions that contain the same k-point.

For instant, from your KPOINTS file, the 1st k-point is A, the 40th K-point is also A, 41st is also A. This is not allowed in IBZKPT_HF routine.

[zoowe]

I think I figured out by playing with mixing tag and/or changing electronic minimisation algorithm.

[tommy91779]

zoowe,

As per your previous post, I have been trying different mixing tags, and minimization algorithms. Can you please relay the tags you used to get yours calculations to complete?

-Tom

[zoowe]

Hi Tom,
I don't think I solve the problem although the calculation was improved a little bit. I was able to do one direction but not all. I will update later.
D.