The ABINIT Users Mailing List ( CLOSED )

[Marc Torrent <marc.torrent@cea.fr>]

Hello,

Here is the solution:

if you want max(forces)<1.d-5, you have to converge them at least at 
1.d-5 (but 1.d-6 recommended) !
You can do it by putting toldff=1.d-5 is the input file, instead of toldfe.
In passing, note that toldfe is not a suitable convergence criterium in 
most cases...
If you absolutely want to keep this keyword, you have to notice that the 
total energy has to be much more converged than the accuracy you asked 
for forces ! (you can see, in the results shown below that Detotal<1.d-9 
when DForces<1.d-5...)

So I modified your input file as follow:
   nband 75  (you can even take less bands)
   ecut 15  (sufficient for Fe, I think, to get De<1mHa)
   pawecutdg 30
   diemix 0.6 iscf 17 (a good compromise)
   toldff 1.d-5 (see above !)

and I got the following output:
     iter   Etot(hartree)     deltaE(h) residm    nres2   diffor 
maxfor   magn
Broyden step 0:
 ETOT 57  -619.84033889131    -9.14E-11 2.95E-05 1.00E-10 2.3E-06 
4.28E-03 12.29
Broyden step 1:
 ETOT 22  -619.84037782638    -3.56E-10 7.36E-07 3.06E-11 2.6E-06 
4.78E-03 12.30
Broyden step 2:
 ETOT 24  -619.84042352227    -1.01E-10 2.57E-08 3.77E-12 2.1E-06 
5.05E-03 12.29
Broyden step 3:
 ETOT 21  -619.84047391982    -5.58E-10 5.71E-10 6.08E-11 8.0E-06 
5.49E-03 12.30
Broyden step 4:
 ETOT 23  -619.84055650396     3.61E-10 9.03E-11 2.39E-11 9.9E-07 
5.87E-03 12.30
Broyden step 5:
 ETOT 25  -619.84068865326    -2.36E-10 1.19E-11 6.09E-11 4.3E-06 
5.63E-03 12.31
Broyden step 6:
 ETOT 29  -619.84089139027     5.92E-10 4.27E-12 3.98E-11 5.1E-06 
2.24E-03 12.35
Broyden step 7:
 ETOT 26  -619.84093135043     4.66E-10 4.01E-13 5.64E-11 2.1E-06 
8.62E-04 12.34

... and so on... until convergence on max(forces).


Marc Torrent


taokun76@hotmail.com a écrit :
> Dear Marc Torrent,
>
> Thank you for your advice.The version of abinit I used is 4.6.5 and the empty bands in my input files are about 40. After changing some parameters,the program crashes after several itrations. 
>
> 1, When I changed the iscf=17, occopt=3, ngkpt=4*4*1 and diemix=0.5, the log 
> files reminds me to increase the spheres overlap. The program stopped after  
> pawovlp was increased to 10. In log file, an error was reported:
>
>  chkpawovlp : ERROR -
>   PAW SPHERES ARE OVERLAPPING !
>    Distance between atoms   2 and   4 is  :   1.23492
>    Radius of the sphere around atom   2 is:   1.60105
>    Radius of the sphere around atom   4 is:   1.60105
>    This leads to a (voluminal) overlap ratio of 45.02 %
>   Action: 1- decrease cutoff radius of pseudopotential(s)
>       OR  2- ajust "pawovlp" input variable to allow overlap
>
> Moreover, during the iterations, we can also find many warrnings:
>
>  mkdenpos : WARNING -
>   Density went < 0 at      2948 points
>   and was set to  1.00E-14.  Lowest was  -0.16E-08.
>   Likely due to too low boxcut or too low ecut for pseudopotential core 
> charge.
>  
> However, in other input files if boxcut is larger than 2.2 (in this input file, the boxcut is 2.02),it always reminds me to improve ecut to decrease boxcut! How to deal with it?
>
> 2,If the iscf=17, occopt=7,ngkpt=6*6*1 with diemix=0.5,the SCF iterations also stopped 
> and the log file reminds that "the voluminal overlap ration is about 98%"!!! 
> When the diemix was decreased to 0.01, the SCF iterations also stopped for the same 
> reason!
>
> 3,may I use prtdos=3 to get DOS of Fe?
>
> Perhaps there are still many parameters wrong in my input file.Any other 
> comments would be appreciated, thank you in advance!