[Pw_forum] ambiguity in stress calculation

Javad hashemifar jhashemifar at gmail.com
Mon Dec 18 12:23:28 CET 2006


Dear Pwscf users

Concerning the "tstress" input variable in the namelist 'control', I
found the following explanation in the prepared sample INPUT_PW file:

tstress     LOGICAL
               calculate stress. Set to .TRUE. if calculation='vc-md'

Does it mean that the calculated stress for other type of calculations
(such as 'scf' and 'relax') is meaningless and is incorrect. I mean if
we set tstress=.TRUE. in calculations='relax' then obtained stress
values are correct or not.

Thanks in advance for your answers
Javad Hashemifar

On 12/17/06, Malgorzata Wierzbowska <wierzbom at ts.infn.it> wrote:
> On Sun, 17 Dec 2006, Amit Kumar wrote:
>
>
>   In the k-points input we read  a line with numbers
>   nk1 nk2 nk3 k1 k2 k3
>   where (k1,k2,k3) is a shift-vector from the origin.
>   For the el-ph calculation (and I guess also for the integration
>   with the tetrahedra method) we need to specify k1=k2=k3=0
>
>   ---- So your "32 32 4 1 1 1" for sure is wrong.
>        You need to take "32 32 4 0 0 0".
>
>   You are right that all k and k+q have to belong to the dense grid.
>   But see: All your meshes are in the same BZ and they are generated
>   by division of the reciprocal lattice vectors b1,b2,b3 by numbers
>   nk1, nk2, nk3 (above is true also for q-point grid nq1, nq2, nq3).
>
>   To check whether your your k and k+q are in dense k-mesh,
>   You need to see whether  dense nk1 divided by rare nk1 gives integer
>   and whether dense nk1 divided by nq1 gives integer.
>   If above 2 conditions are satisfied, you check direction b2 and b3
>   in your cell: I mean dense_nk2/nq2 and dense_nk3/nq3
>   and dense_nk2/rare_nk2 etc. If all above give integers --
>   which is your case -- your error does not occur.
>
>   -- So let us check your meshes:
>      dense    32 32 4 0 0 0
>      rare      8  8 1 0 0 0
>      phonon    4  4 1 (it should be default here that they are not shifted)
>
>    direction b1:   32/8=integer, 32/4=integer, also k+q is OK because 8/4=integer
>    the same for b2
>    direction b3:  4/1=integer
>
>    For me it works for hehagonal Mg. Please do my excercise I attach the script.
>
>    Gosia
>
>
>
>
> >     Dear Malgorzata,
> >
> >     It's very nice to recieve your useful suggestions.
> >     I tried all the combinations--->
> >     32 32 4 0 0 0,        32 32 32,
> >     32 32 4 1 1 1 --->>> every possible combinations before posting this
> >     problem on the forum.
> >     But I got the same following error  again and again.
> >
> >     electron-phonon interaction  ...
> >
> >
> > %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
> >    from lint : error #        23
> >    cannot remap grid on k-point list
> > %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
> >
> >    stopping ...
> >
> >
> >   I think the source of the above error is something different.
> >
> >   My guess::::
> >
> >   The dense grid must contain all k  and k+q grid points used in the
> >   e-p calculation.
> >
> >   How do we know the above condition has been satisfied for automatic
> >   generated k and q points with nk1=32, nk2=32, nk3=4 and
> >   nq1=4, nq2=4, nq3=1??????
> >
> >    Here is my q-point grid::::
> >
> >   Calculation of the dynamical matrices for ( 4, 4, 1,) uniform grid of
> > q-points
> >    (   7q-points):
> >      N       xq(1)       xq(2)       xq(3)
> >      1     0.00000     0.00000     0.00000
> >      2     0.00000     0.28868     0.00000
> >      3     0.00000    -0.57735     0.00000
> >      4     0.25000     0.14434     0.00000
> >      5     0.25000     0.43301     0.00000
> >      6    -0.50000    -0.28868     0.00000
> >      7    -0.50000     0.00000     0.00000
> >
> >    And my k-points grid is here::::
> >
> >        number of k points=  166  gaussian broad. (ryd)=  0.0100     nga
> > uss =   1
> >                      cart. coord. in units 2pi/a_0
> >       k(    1) = (   0.0000000   0.0000000   0.0000000), wk =   0.0
> > 312500
> >       k(    2) = (   0.0000000   0.2886751   0.0000000), wk =   0.0
> > 000000
> >       k(    3) = (   0.0000000   0.1443376   0.0000000), wk =   0.0
> > 104167
> >       k(    4) = (   0.0000000   0.4330127   0.0000000), wk =   0.0
> > 000000
> >
> >   *
> >   *
> >   *  k(  166) = (   0.2500000   0.8660254   0.0000000), wk =   0.0000000
> >
> > See!!!! 0.28868  and 0.2886751 are not exactly equal.
> >
> > Is this the possible reason for the error like
> >    "cannot remap grid on k-point list"?????
> >
> > I'm not sure any other there might be any other.
> > Please help me to find out the unknown reason of this well known
> > error  and  give me some clue to get rid of this error.
> >
> > Thank you again for your very very useful letter.
> >
> > Regards,
> > Amit
> >
> >
> >
> >>
> >
>
>


-- 
======================================
Seyed Javad  Hashemifar,    Ph.D.
[current:]
Tel:+49-203-3794743        Fax:+49-203-3794742
Fachbereich Physik (AG Kratzer), Universitat Duisburg-Essen
47048 Dusiburg, Germany
[permanent:]
Tel:  +98-311-3912375       Fax: +98-311-3912376
Physics Department, Isfahan University of Technology
84156 Isfahan, Iran
---------------------------------------------------------------------------



More information about the users mailing list