Hi!

I played around with transition metal oxides and LDA+U calculations during my diploma thesis a couple of years back, comparing the method by Dudarev et al. (where you have at least only one parameter to tune, as in this approach only U-J enters the equation) with hybrid functional calculations (HSE) as a benchmark. I found that it is very hard to select a specific U that correctly replicated the HSE DOS, the lattice constant and the magnetic moment. Really, you mostly have to select one thing that you are interested in most and then tune the parameter U (assuming you keep J fixed, for example to 0) to a value that replicates this parameter sufficiently accurate without producing unrealistic results elsewhere. For example a large U will blow up the lattice constant to completely unrealistic values if relaxations are turned on.

This is my suggestion:

1) Select LDAUTYPE=2 to only work with one effective parameter

2) Do some reference calculations on a small system with hybrid functionals (e.g. HSE) or study previous calculation of higher accuracy or experimental data to get benchmark values.

3) Vary U in about 0.2 eV steps from 0.0 to ~8.0 and compare the results to the benchmark. (In my experience values between 1 and 4 usually offer the best overall results, but it depends on what you are looking fore I guess.

I attached a link to my Diploma thesis for more details. I hope this can help you and would be glad to provide more feedback or answer further questions,

cheers, Michael

http://catalogplus.tuwien.ac.at/primo_library/libweb/action/display.do?dscnt=0&elementId=3&recIdxs=3&frbrVersion=5&scp.scps=scope:(UTW_O_SFX),scope:(UTW_aleph_acc),primo_central_multiple_fe&frbg=&tab=default_tab&displayMode=full&dstmp=1428394515758&srt=rank&ct=display&mode=Basic&dum=true&indx=4&recIds=UTW_aleph_acc000502969&renderMode=poppedOut&doc=UTW_aleph_acc000502969&vl(freeText0)=wolloch&vid=UTW&fn=search&tabs=detailsTab&vid=UTW&backFromPreferences=true

Hai Wolloch 

can i use the same U values for a transition metal that u already bench marked Or these values are different for system to system  

Dear naga venkateswara

U is a system dependent property that means the value of U is not necessarily same for a particular transition metal in all systems. So you have to follow the 3rd suggestion of  Michael Wolloch to get the U value of your systems.

Ashis

Thanks for your suuggestion Ashis

Hi again!

I wanted to weight in again by telling you all that there is of course also a way to determine Hubbard Parameters by constrained RPA (random phase approximation) calculations. I have no experiance with this method, but it is probably superior to the method described by me above, certainly it is more rigorous. 

Plese check the links I provided for details,

cheers, Michael

http://www.cond-mat.de/events/correl11/manuscripts/aryasetiawan.pdf

Article Calculations of Hubbard U from first-principles

Thank you Mr.Michael Wolloch for the information . 

I am using bellow INCAR for HSE06 method , But i am unable to get output. Can anyone please check these below files and suggest for further.

Thanks

INCAR:::

# SCF input for VASP

# Note that VASP uses the FIRST occurence of a keyword

SYSTEM = HSE06

PREC = High

ENCUT = 600

IBRION = 2

NSW = 1000

ISIF = 3

ALGO = D ; TIME = 0.5

NELM = 60

NELMIN = 5

EDIFF = 1.0e-06

EDIFFG = -0.01

VOSKOWN = 1

NBLOCK = 1

ISPIN = 2

MAGMOM = 4*2 4*-2 4*0 12*0

LHFCALC = .T.

HFSCREEN = 0.2

PRECFOCK = fast

NKRED = 2

LDIAG = .TRUE.

INIWAV = 1

ISTART = 0

ICHARG = 2

LWAVE = .FALSE.

LCHARG = .FALSE.

ADDGRID = .FALSE.

ISMEAR = 0

SIGMA = 0.01

GGA = PE

LREAL = .FALSE.

What do you mean you do not get output? Does the simulation crash? Are there error messages? Does it even start? What K-points grid are you using?

On a quick glance your INCAR seams to be OK, but I would write out some commands just to be sure:

ALGO = DAMPED instead of ALGO = D and

HFCALC = .TRUE.

Be also aware that hybrid calculations take about 50 times longer than a GGA calculation, so maybe the output is just coming later than expected...

Intially Job was started running , but after few minutes it was stoped and given a OUTCAR like this....i could not able to understand what is the problem???

OUTCAR  ending lines are given here...

 k-point 61 :  -0.4375 0.3750 0.4375  plane waves:   11092

 k-point 62 :  -0.3125 0.3750 0.4375  plane waves:   11078

 k-point 63 :  -0.1875 0.3750 0.4375  plane waves:   11114

 k-point 64 :  -0.0625 0.3750 0.4375  plane waves:   11097

 maximum and minimum number of plane-waves per node :     11137    11071

 maximum number of plane-waves:     11137

 maximum index in each direction:

   IXMAX=   11   IYMAX=   19   IZMAX=   12

   IXMIN=  -11   IYMIN=  -20   IZMIN=  -12

 NGX is ok and might be reduce to  46

 NGY is ok and might be reduce to  80

 NGZ is ok and might be reduce to  50

 serial   3D FFT for wavefunctions

 parallel 3D FFT for charge:

    minimum data exchange during FFTs selected (reduces bandwidth)

 FFT grid for exact exchange (Hartree Fock)

  NGX = 24; NGY = 42; NGZ = 28

 Radii for the augmentation spheres in the non-local exchange

 for species   1 augmentation radius   1.265 (default was   0.979)

       energy cutoff for augmentation   2400.0

 for species   2 augmentation radius   1.022 (default was   0.959)

       energy cutoff for augmentation   2400.0

 for species   3 augmentation radius   1.034 (default was   0.970)

       energy cutoff for augmentation   2400.0

Is HSE calculation will effect with POTCAR file????? i have used PAW_PBE  and LDA for above HSE calculations, but it was failed.

Hej!

The PBE POTCAR should be perfectly fine, the OUTCAR does not look too bad eighter, but that does not mean much. What about the OSZICAR or stdout? Is it possible that you might run out of memory? On what kind of machine are you performing this calculation, how big is your cell? What version of VASP are you using? Maybe you can upload all of your input files and the most important output files (INCAR, KPOINTS, jobfile, POSCAR, POTCAR, OUTCAR, OSZICAR, error files, etc...) in a zipped format?

With the little information you give I am unable to help, sorry

Dear Sir,

Here i am attaching INPUT files for graphene. with graphene also i am facing same problem. Can U Please check these inputs and suggest me for further.

Thanks in advance

ANJI