The ABINIT Users Mailing List ( CLOSED )

[matthieu verstraete <matthieu.jean.verstraete@gmail.com>]

>   thm9 : WARNING -
> 4032   No atomic factor tensor is printed out :
> 4033   the tolerance level that was asked
> 4034   has not been match with the grids specified.
> 4035   Action : in the input file, increase the resolution
> 4036   of grids ng2qpt, or decrease the accuracy requirement thmtol.
This is a bit of a pain: by default, you specify a target grid of q
interpolation and anaddb tries several subgrids to check convergence,
then stops once the DOS changes by less than some tolerance from one
grid to the next finer one. If you just want one calculation you can
force the use of a single grid.

This happens often when you have some negative modes (you should check
this explicitly from the interpolation on a band structure in the
whole zone), whose contribution varies with q-mesh refinement as the
negative-frequency q-points are neglected.

>
> I see where thmtol gets checked and it looks like its no where close to
> converging.
>
>   thm9 : check if the thermodynamic functions
> 3254     change within tolerance.
> 3255  thm9 : free energy relative changes     1.7626E+02
this is huge - you should look more closely at the interpolated phonons.

> 3256         are larger than thmtol     3.2000E-01
> 3257  thm9 : energy relative changes     1.7629E+02
> 3258         are larger than thmtol     3.2000E-01
> 3259  thm9 : iwchan=     4
> 3260  thm9 : compute g, f, e, s, c
>
> a few questions:
>
> 1:  what does this mean?  should I be concerned about this warning?   the
> actual thermal output does not seem to change much from a 100 100 100 grid
> to a 200 200 200 grid
that's good. See above for recommendations

>
> 2.  why is it that sometimes an out_THERMO file is created and other times
> not?
If you turn the thmflg on it should always be created, but only with
newer versions of abinit.

>
> 3.   I set my thmtol to 0.320.    how can i interpret that interms of
> accuracy of the calculation?
it's easier to interpret the changes in free energy or entropy as a
function of convergence.

>
> 4.   when choosing a gpt grid,  is it necessary to pick high resolution qpt
> grids for metals just because a large kpt grid was required for the
> electronic structure?
not necessarily: the features in Al phonons are quite smooth. However,
Kohn anomalies and other features in the phonon bands can make dense
sampling important. For thermodynamics, the most important is getting
the low q acoustic modes right (this is where the interpolation can
make convergence difficult, even though the bulk of the DOS etc... is
very well converged).

Matthieu


>
>
> thanks very much!  following is my anaddb input file.
>
> !Input file for the ifc code. Analysis of bulk Al
>   3
>   4 !Flags
>   5  ifcflag   1     ! Interatomic force constant flag
>   6  thmflag   1     ! Thermodynamical properties flag
>   7
>   8 !Wavevector grid number 1 (coarse grid, from DDB)
>   9   brav    1      ! Bravais Lattice : 1-S.C., 2-F.C., 3-B.C., 4-Hex.)
>  10   ngqpt   6 6 6  ! Monkhorst-Pack indices
>  11   nqshft  1         ! number of q-points in repeated basic q-cell
>  12   q1shft  3*0.0
>  13
>  14 !Effective charges
>  15      asr   1     ! Acoustic Sum Rule. 1 => imposed asymetrically
>  16   chneut   1     ! Charge neutrality requirement for effective charges.
>  17
>  18 !Interatomic force constant info
>  19   dipdip  1      ! Dipole-dipole interaction treatment
>  20
>  21 !Wavevector grid number 2 (series of fine grids, extrapolated from
> interat f    orces)
>  22   ng2qpt   200 200 200  ! sample the BZ up to ngqpt2
>  23   ngrids   5         ! number of grids of increasing size
>  24   q2shft   3*0.0
>  25
>  26 !Thermal information
>  27   nchan   2500   ! # of channels for the DOS with channel width 1 cm-1
>  28   nwchan  5      ! # of different channel widths from this integer down
> to 1     cm-1
>  29   thmtol  0.320  ! Tolerance on thermodynamical function fluctuations
>  30   ntemper 60     ! Number of temperatures
>  31   temperinc 20.  ! Increment of temperature in K for temperature
> dependency
>  32   tempermin 20.  ! Minimal temperature in Kelvin
>  33
>  34 # This line added when defaults were changed (v5.3) to keep the
> previous, ol    d behaviour
>  35 symdynmat 0

-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Dr. Matthieu Verstraete

European Theoretical Spectroscopy Facility (ETSF)
Dpto. Fisica de Materiales,
U. del Pais Vasco,
Centro Joxe Mari Korta, Av. de Tolosa, 72,   Phone: +34-943018393
E-20018 Donostia-San Sebastian, Spain        Fax  : +34-943018390

Mail : matthieu.jean.verstraete@gmail.com
http://www-users.york.ac.uk/~mjv500