[Wien] Fwd: not getting convergence with MBJ potential

Laurence Marks L-marks at northwestern.edu
Mon Dec 12 15:28:44 CET 2011


If your structure is a metal in GGA in 99% of cases it will be a metal
in GGA+U and mBJ. And, it is wrong. Just because others reported the
structure does not mean that it is right! There are many cases where
someone reported a structure, others followed their lead and only
later someone pointed out that it was wrong.

Look at your GGA (or GGA+U) result, the overall DOS which is in
case.output2up/dn (or do a standard DOS calculation). If the positions
are close to correct you should see a gap between the different bands.
If you have electrons at E_F at the bottom of a band (metal d) your
structure is reduced; you will need some metal vacancies. If there are
holes at the top of the oxygen band your structure is oxidized -- you
will need some oxygen vacancies. Note that vacancies can be ordered or
disordered (partial occupancies) and simulating these is not easy.

Check the bond-valence sums, they are probably given by the powder
diffraction fitting program or there are many codes to do this. You
can even do a "grep -e bond case.outputnn" to get a decent number. All
the O's should be close to 2 (perhaps in the range 1.8-2.2) and all
the metals should be in reasonable states. For instance a value of 3.5
for Ti means that it is reduced, and a value of 4.5 means that
something is badly wrong. As a shameless plug look at Surface Science
doi:10.1016/j.susc.2011.10.018 and references therein, or for instance
I.D. Brown, Chem. Rev. 109 (2009). In my opinion BVS is a simple and
very quick way to do a sanity check and can save you from weeks of not
useful DFT calculations.

2011/12/12 shamik chakrabarti <shamikphy at gmail.com>:
> Dear Dr. Laurence Marks Sir,
>
> I am giving the answers all of your queries below,
>
> a) If R-factor means reliability factor then we have achieved a parameter
> called goodness of fit whose value was 1.0038. But this material was
> prepared previously by researchers and they have also got the same
> structure. Although the structure is based only on the Rietveld fitting of
> experimental XRD pattern. There was no rigorous TEM analysis on this
> structure so far.
>
> b) We have done UV-visible band gap measurement on this material and it is
> 3.26 eV. But as you have said properly it is indeed coming as metal after
> getting converged in 80 iterations. But we have noticed that there are
> presence of O p orbital between upper and lower Hubbard band of transition
> metal atom and hence GGA+U will not improve the situation.
>
> c) At the moment we are a bit handicapped for doing Geometrical (both
> lattice parameter and atomic coordinates) optimization of 56 atoms unit cell
> structure as we do not have the facility of cluster. Hence as a first
> approximation we are first trying to reach as near as possible to the
> accurate electronic structure. We are soon going to get the facility and
> then we can use geometrical optimization.
> And hence we really do have large forces on the structure.
>
> d) We have used a greed of 0.2 and 0.15 both for PRATT mixing and we have
> not used MSR1 mixing for mbj potential yet.
>
> Sir, as far as the above discussion is concerned we might have our structure
> wrong (need more accurate TEM analysis!) or Geometrical optimization may
> solve the problem. But as we are now trying to experiment with test run we
> and as it seems GGA+U can not solve the problem we have opted for mbj
> potential.
>
> But as you have said if the structure is wrong (we are not yet confirmed
> though) we may never reach the convergence or desired electronic structure!
>
> Thank you very much for your help Sir.
>
> On Mon, Dec 12, 2011 at 6:52 PM, Laurence Marks <L-marks at northwestern.edu>
> wrote:
>>
>> I wonder whether you really have the structure right.
>>
>> a) When you say that you get the same structure, what R-factor? Powder
>> diffraction is not very sensitive to oxygen positions in the presence
>> of heavy metal atoms.
>>
>> b) If GGA takes about 80 iterations I suspect that you have a metal.
>> If you do not have the correct number of oxygen versus metal atoms you
>> will get this, and no amount of U or mBJ will change this (in 99% of
>> cases). Check the bond-valence sums of your model, are the numbers
>> reasonable? This is a standard test in solid-state chemistry.
>>
>> c) Do you have large forces (as Peter asked)? If you do this also
>> indicates that you have the structure wrong.
>>
>> d) When you say that you were using a greed of 0.2 (please do not use
>> the term "mixing factor", it is misleading) was this for Pratt or
>> MSR1? A change from 0.2 to 0.15 with MSR1 will not matter much.
>>
>> 2011/12/12 shamik chakrabarti <shamikphy at gmail.com>:
>> > Dear Dr Peter Blaha and Dr Pascal,
>> >
>> >                               The material is a ferrimagnetic transition
>> > metal compound and a charge transfer insulator and the gap is between d
>> > and
>> > O p states. Yes we have started MBJ potential calculation from the
>> > converged
>> > GGA charge density. So what we can do, we can start from the scratch
>> > with
>> > MBJ potential and after few iterations will get shifted to MSR1 mixing
>> > to
>> > improve the convergence. We hope that will help. Thank you for all your
>> > helps.
>> >
>> > with regards,
>> >
>> >
>> > On Mon, Dec 12, 2011 at 1:31 PM, pascal boulet
>> > <pascal.boulet at univ-provence.fr> wrote:
>> >>
>> >>
>> >> Dear Dr. Chakrabarti,
>> >>
>> >> All right. So, another idea: could it be that you restarted your m-BJ
>> >> calculation using the converged GGA or GGA+U density? I ran into these
>> >> kinds of troubles several times. I found it preferable to start from
>> >> scratch...
>> >>
>> >> Hope this help,
>> >> Pascal
>> >>
>> >> On 12/12/2011 08:45, shamik chakrabarti wrote:
>> >> > Dear Dr. Pascal,
>> >> >
>> >> >                 As we were getting Ghost band error due to large
>> >> > charge fluctuation during initial cycle *we have reduced the mixing
>> >> > parameter from 0.2 to 0.15 & hence thereafter we have not received
>> >> > ghost band error yet.*
>> >> >
>> >> > As long as variation of RMT is concerned the variation is only from
>> >> > 1.7 to 2.01 in the structure.
>> >> >
>> >> > So far there is no energy equivalence between a local orbital and
>> >> > overall energy parameter which can induce ghost band error.
>> >> >
>> >> > Also we have achieved convergence of -0.000034 with plain GGA and
>> >> > only
>> >> > with MBJ potential we are getting the convergence problem and again
>> >> > plain GGA and also GGA+U have not given the proper electronic
>> >> > structure.
>> >> >
>> >> > As far as the structure is concerned we are very much sure about its
>> >> > structural parameters as they are giving the same XRD as experiment
>> >> > when the positional coordinates and lattice parameters of the 56 atom
>> >> > cell were put into a software powdcell.
>> >> >
>> >> > looking forwards to your comments.
>> >> >
>> >> > with regards,
>> >> > On Mon, Dec 12, 2011 at 12:26 PM, pascal boulet
>> >> > <pascal.boulet at univ-provence.fr
>> >> > <mailto:pascal.boulet at univ-provence.fr>> wrote:
>> >> >
>> >> >     Hello,
>> >> >
>> >> >     It is strange that you have such a bad convergence on the energy
>> >> >     and the
>> >> >     charge after that many cycles.
>> >> >     My feeling is that something is wrong in the input.
>> >> >
>> >> >     Have a look at the user manual. From page 205 on, there is an
>> >> >     explanation of the meaning of ghost bands. You have to experiment
>> >> > on
>> >> >     your chemical system to get rid of them.
>> >> >
>> >> >     Pascal
>> >> >
>> >> >
>> >> >     On 12/12/2011 05:21, shamik chakrabarti wrote:
>> >> >     > Dear wien2k users,
>> >> >     >
>> >> >     >                        Any response of my previous mail will be
>> >> > very
>> >> >     > helpful for us. Eagerly waiting for your comments.
>> >> >     >
>> >> >     > with regards,
>> >> >     >
>> >> >     > ---------- Forwarded message ----------
>> >> >     > From: *shamik chakrabarti* <shamikphy at gmail.com
>> >> >     <mailto:shamikphy at gmail.com>
>> >> >     > <mailto:shamikphy at gmail.com <mailto:shamikphy at gmail.com>>>
>> >> >     > Date: Sun, Dec 11, 2011 at 5:42 PM
>> >> >     > Subject: not getting convergence with MBJ potential
>> >> >     > To: A Mailing list for WIEN2k users
>> >> >     <wien at zeus.theochem.tuwien.ac.at
>> >> >     <mailto:wien at zeus.theochem.tuwien.ac.at>
>> >> >     > <mailto:wien at zeus.theochem.tuwien.ac.at
>> >> >     <mailto:wien at zeus.theochem.tuwien.ac.at>>>
>> >> >     >
>> >> >     >
>> >> >     > Dear wien2k users,
>> >> >     >
>> >> >     >                        I am trying to simulate electronic
>> >> >     structure of
>> >> >     > a charge transfer insulator. Our system consists of a unit cell
>> >> >     having
>> >> >     > 56 atoms and we are using 14 k-points. As GGA+U method was
>> >> > failed
>> >> > to
>> >> >     > reproduce the proper band gap (in fact much less than the
>> >> >     > experimentally obtained one) we have opted for MBJ potential.
>> >> >     > *Initially the mixing parameter was set to 0.2 and we got Ghost
>> >> > band
>> >> >     > error after few iterations. Hence we reduce the mixing
>> >> > parameter
>> >> > to
>> >> >     > 0.15.* Until now still it has shown no error except some extra
>> >> >     message
>> >> >     > in case.dayfile. I am giving those messages below. So far e.c.
>> >> > &
>> >> > c.c
>> >> >     > were achieved up to 2.239... & 5.4599 respectively and already
>> >> > 36
>> >> >     > iterations are over. My questions are:
>> >> >     >
>> >> >     > (1) Is it ok to set up a mixing parameter of 0.15 that is with
>> >> >     it, is
>> >> >     > it possible to reach convergence?
>> >> >     >
>> >> >     > (2) As plain GGA takes around 80 iterations to converge, is it
>> >> > that
>> >> >     > MBJ potential will take much higher than that as we know it is
>> >> >     always
>> >> >     > hard to get convergence with this method?
>> >> >     >
>> >> >     > (3) what are the meanings of the given messages below:
>> >> >     >
>> >> >     >   lapw2 -c -up   -vresp (17:12:13) >   lapw1  -c -dn
>> >> >  (16:19:33)
>> >> >     > 10848.369u 98.026s 52:39.78 346.4%0+0k 0+1099920io 0pf+0w
>> >> >     > >   lapw1  -c -up    (15:26:49) 10864.462u 99.082s 52:44.39
>> >> >     346.4%0+0k
>> >> >     > 0+1100232io 0pf+0w
>> >> >     > 460.324u 151.313s 6:52.62 148.2%0+0k 0+732368io 0pf+0w
>> >> >     >  -5.956509016536202E-002
>> >> >     >   3.325214690385789E-002 -9.166205529297161E-002 tauwrong=
>> >> >     >  int:rho,tauw,grho,g2rho  2.146196813781874E-002
>> >> >      1.287982148952303E-002
>> >> >     >  -4.564940741427861E-002
>> >> >     >   3.061850458947012E-002 -5.357427363703868E-002 tauwrong=
>> >> >     >  int:rho,tauw,grho,g2rho  1.778719641126464E-002
>> >> >     >  1.317651193616000E-002        *what is tauwrong?*
>> >> >     >  -4.561631481999567E-002
>> >> >     >   3.051935798833875E-002 -5.372842413400847E-002 tauwrong=
>> >> >     >  int:rho,tauw,grho,g2rho  1.783702549056333E-002
>> >> >      1.305474408433668E-002
>> >> >     >  -2.872847342414767E-002
>> >> >     >   2.625149916992486E-002 -2.271097056901070E-002 tauwrong=
>> >> >     >  int:rho,tauw,grho,g2rho  1.456340274765583E-002
>> >> >      1.183001700580401E-002
>> >> >     >  -2.843529510619935E-002
>> >> >     >   2.607602777746414E-002 -2.279192386903682E-002 tauwrong=
>> >> >     >  int:rho,tauw,grho,g2rho  1.462794826548238E-002
>> >> >      1.162089194449066E-002
>> >> >     >  -1.211083415464120E-002
>> >> >     > ..........................................
>> >> >     > ..........................................
>> >> >     > >   lapw0 (15:19:56)  sphere:rho,tauw,grho,g2rho
>> >> > 7.46875999972237
>> >> >     >      180.956718336004
>> >> >     > >   lapw0 -grr (15:15:10) 315.375u 93.253s 4:45.73 143.0%0+0k
>> >> >     > 0+................
>> >> >     >
>> >> >     >
>> >> >     > Any response in this regard will be very helpful for us. Thanks
>> >> > in
>> >> >     > advance.
>> >> >     >
>> >> >     > with regards,
>> >> >     > --
>> >> >     > Shamik Chakrabarti
>> >> >     > Senior Research Fellow
>> >> >     > Dept. of Physics & Meteorology
>> >> >     > Material Processing & Solid State Ionics Lab
>> >> >     > IIT Kharagpur
>> >> >     > Kharagpur 721302
>> >> >     > INDIA
>> >> >     >
>> >> >     >
>> >> >     >
>> >> >     > --
>> >> >     > Shamik Chakrabarti
>> >> >     > Senior Research Fellow
>> >> >     > Dept. of Physics & Meteorology
>> >> >     > Material Processing & Solid State Ionics Lab
>> >> >     > IIT Kharagpur
>> >> >     > Kharagpur 721302
>> >> >     > INDIA
>> >> >     >
>> >> >     >
>> >> >     > _______________________________________________
>> >> >     > Wien mailing list
>> >> >     > Wien at zeus.theochem.tuwien.ac.at
>> >> >     <mailto:Wien at zeus.theochem.tuwien.ac.at>
>> >> >     > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>> >> >
>> >> >     --
>> >> >     Dr. pascal Boulet, computational chemist
>> >> >     University of Aix-Marseille I
>> >> >     Laboratoire Chimie Provence, UMR 6264
>> >> >     Group of Theoretical Chemistry
>> >> >     Avenue Normandie-Niemen
>> >> >     13397 Marseille Cedex 20
>> >> >     France
>> >> >     **********
>> >> >     Tel. (+33) (0)413.55.18.10
>> >> >     Fax. (+33) (0)413.55.18.50
>> >> >     **********
>> >> >     http://www.lc-provence.fr
>> >> >     https://sites.google.com/a/univ-provence.fr/pb-comput-chem
>> >> >     %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
>> >> >
>> >> >        * Français - détecté
>> >> >        * Anglais
>> >> >        * Français
>> >> >
>> >> >        * Anglais
>> >> >        * Français
>> >> >
>> >> >     <javascript:void(0);>
>> >> >     _______________________________________________
>> >> >     Wien mailing list
>> >> >     Wien at zeus.theochem.tuwien.ac.at
>> >> >     <mailto:Wien at zeus.theochem.tuwien.ac.at>
>> >> >     http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>> >> >
>> >> >
>> >> >
>> >> >
>> >> > --
>> >> > Shamik Chakrabarti
>> >> > Senior Research Fellow
>> >> > Dept. of Physics & Meteorology
>> >> > Material Processing & Solid State Ionics Lab
>> >> > IIT Kharagpur
>> >> > Kharagpur 721302
>> >> > INDIA
>> >> >
>> >> >
>> >> > _______________________________________________
>> >> > Wien mailing list
>> >> > Wien at zeus.theochem.tuwien.ac.at
>> >> > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>> >>
>> >> --
>> >> Dr. pascal Boulet, computational chemist
>> >> University of Aix-Marseille I
>> >> Laboratoire Chimie Provence, UMR 6264
>> >> Group of Theoretical Chemistry
>> >> Avenue Normandie-Niemen
>> >> 13397 Marseille Cedex 20
>> >> France
>> >> **********
>> >> Tel. (+33) (0)413.55.18.10
>> >> Fax. (+33) (0)413.55.18.50
>> >> **********
>> >> http://www.lc-provence.fr
>> >> https://sites.google.com/a/univ-provence.fr/pb-comput-chem
>> >> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
>> >>
>> >> _______________________________________________
>> >> Wien mailing list
>> >> Wien at zeus.theochem.tuwien.ac.at
>> >> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>> >
>> >
>> >
>> >
>> > --
>> > Shamik Chakrabarti
>> > Senior Research Fellow
>> > Dept. of Physics & Meteorology
>> > Material Processing & Solid State Ionics Lab
>> > IIT Kharagpur
>> > Kharagpur 721302
>> > INDIA
>> >
>> > _______________________________________________
>> > Wien mailing list
>> > Wien at zeus.theochem.tuwien.ac.at
>> > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>> >
>>
>>
>>
>> --
>> Professor Laurence Marks
>> Department of Materials Science and Engineering
>> Northwestern University
>> www.numis.northwestern.edu 1-847-491-3996
>> "Research is to see what everybody else has seen, and to think what
>> nobody else has thought"
>> Albert Szent-Gyorgi
>> _______________________________________________
>> Wien mailing list
>> Wien at zeus.theochem.tuwien.ac.at
>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>
>
>
>
> --
> Shamik Chakrabarti
> Senior Research Fellow
> Dept. of Physics & Meteorology
> Material Processing & Solid State Ionics Lab
> IIT Kharagpur
> Kharagpur 721302
> INDIA
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>



-- 
Professor Laurence Marks
Department of Materials Science and Engineering
Northwestern University
www.numis.northwestern.edu 1-847-491-3996
"Research is to see what everybody else has seen, and to think what
nobody else has thought"
Albert Szent-Gyorgi


More information about the Wien mailing list