[Pw_forum] Smearing Problem

[Nicola Marzari]

Hi Paresh,


all good - but are you sure the units are mRy? A difference of
0.06 mRy /atom is super small (11604K=1eV), so we are talking about
10K difference, lost in the noise of the errors of approximate functionals,
entropic effects, etc...

Anyhow, if you want the computational "truth" you could try halving or
doubling the fermi-dirac smearing (or m-v, or m-p), using the largest 
sampling,
and see if you get the same number.

				nicola


On 01/07/2015 21:15, paresh rout wrote:
> Sorry sir in the previous plot the energy difference was not per atom.
> So here I am attaching the Corrected plot. My system contains 20 atom.
>
> On Wed, Jul 1, 2015 at 6:09 PM, paresh rout <paresh.rout88 at gmail.com
> <mailto:paresh.rout88 at gmail.com>> wrote:
>
>     Thank you very much sir. That was a very helpful reply. Now I
>     understand  smearing. Just one more query, I did a convergence test
>     with k-point sampling  by fixing the F-D smearing value to 0.00184
>     and the energy difference between C-type and G-type  seems to be
>     converged (I have attached the convergent test plot). If I
>     understand your explanation correctly then this result should also
>     be reliable. Please let me know if I am wrong.
>
>       I will also do the convergent test with M-V(0.01) and the second
>     solution that you have suggested.
>     (The K-points grids are 8 8 6 , 10 10 8, 12 12 10, 14 14 12,
>     previously I was using 8 8 6)
>
>     Kind Regards
>     Paresh Chandra Rout
>     Research Scholar
>     INDIAN INSTITUTE OF SCIENCE EDUCATION AND RESEARCH, BHOPAL
>
>
>     On Wed, Jul 1, 2015 at 12:56 PM, Nicola Marzari
>     <nicola.marzari at epfl.ch <mailto:nicola.marzari at epfl.ch>> wrote:
>
>
>
>         Dear Paresh,
>
>
>         maybe this is easier to follow:
>         http://theossrv1.epfl.ch/Main/ElectronicTemperature
>
>         The main concept is that with very small smearing you need *a
>         lot* of
>         k-points
>         to perform the correct Brillouin zone integrations - so with a
>         0.00184
>         smearing
>         you have a *large* error due to insufficient sampling, not a
>         small error.
>
>         Solution 1 - if you do not want to think, ever, in your life,
>         use 0.01
>         Ry of m-v
>         or m-v smearing, and increase the k-point sampling until the
>         difference
>         in energy
>         between G and C type is converged.
>
>         Solution 2 - plot the energy difference between G and C as a
>         function of
>         smearing,
>         for m-v smearing of 0.005, 0.010, 0.015, 0.020, 0.025, and with
>         at least
>         3 or
>         4 k-point sampling. Use the discussion from
>         http://theossrv1.epfl.ch/Main/ElectronicTemperature
>         to understand what to do.
>
>                                          nicola
>
>
>
>         On 30/06/2015 17:07, paresh rout wrote:
>         > Dear all,
>         > I am dealing with a transition metal oxide compound (ABO3), Where I use
>         > cold smearing *m-v(0.02)* for geometry optimization and for total energy
>         > calculation I use *f-d (0.00184*) at the final optimized structure. I
>         > have calculated the total energy for various type of magnetic ordering
>         > like (G-type, C-type). The total energy (SCF) for G-type and C-type are
>         > in reverse ordering for different smearing technique (*G-type is the
>         > ground state for* *MV smearing and C-type is the ground state for FD*).
>         > But when I reduced the value of degauss of *MV to 0.008 *I got the same
>         > ordering as FD.
>         > I am very confused about the smearing technique . I have also read Prof.
>         > Nicola's thesis about smearing but there I did not get rid  of my
>         > confusion also I did not find about FD technique . Can anybody kindly
>         > tell me about the right smearing technique which I can use for the spin
>         > polarized calculation ? Any reply would be highly appreciated.
>         >
>         >
>         > Kind Regards
>         > Paresh Chandra Rout
>         > Research Scholar
>         > INDIAN INSTITUTE OF SCIENCE EDUCATION AND RESEARCH, BHOPAL
>         >
>         >
>         >
>         > _______________________________________________
>         > Pw_forum mailing list
>         >Pw_forum at pwscf.org <mailto:Pw_forum at pwscf.org>
>         >http://pwscf.org/mailman/listinfo/pw_forum
>         >
>
>         --
>
>         ----------------------------------------------------------------------
>         Prof Nicola Marzari, Chair of Theory and Simulation of
>         Materials, EPFL
>         Director, National Centre for Competence in Research NCCR
>         MARVEL, EPFL
>         _______________________________________________
>         Pw_forum mailing list
>         Pw_forum at pwscf.org <mailto:Pw_forum at pwscf.org>
>         http://pwscf.org/mailman/listinfo/pw_forum
>
>
>
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>

-- 

----------------------------------------------------------------------
Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL
Director, National Centre for Competence in Research NCCR MARVEL, EPFL