[Pw_forum] Convergence during vc-relax

Varadharajan Srinivasan vsriniva at Princeton.EDU
Sat Oct 16 18:55:32 CEST 2004


Hello,

Could this be because you are using the default number of steps (nsteps
 = 50 in &control) for relaxation? If you increase this number of
ions+electron steps the calculation should proceed to convergence.

Regards,
Vardha.



On Sat, 16 Oct 2004, Sergey Lisenkov wrote:

> Dear PWscf authors,
>
> I am performing variable-cell relaxation of unit cell of my structure. Firstly. I performed atomic relaxation. My calculations were finished without any informations:
> ...
>
>    atom  11 type  1   force =    -0.00001671    0.00011974   -0.00001286
>      atom  12 type  1   force =     0.00000952    0.00008130   -0.00004912
>      atom  13 type  2   force =     0.00004235    0.00005128   -0.00004270
>      atom  14 type  2   force =     0.00005058   -0.00004839   -0.00005341
>      atom  15 type  2   force =     0.00004395   -0.00000478    0.00006488
>      atom  16 type  1   force =    -0.00000257    0.00001235   -0.00008447
>
>      Total force =     0.000348     Total SCF correction =     0.001318
>
>
>      entering subroutine stress ...
>
>           total   stress  (ryd/bohr**3)                  (kbar)     P=   -2.02
>   -0.00001869   0.00000015  -0.00000016         -2.75      0.02     -0.02
>    0.00000015  -0.00001119  -0.00000001          0.02     -1.65      0.00
>   -0.00000016  -0.00000001  -0.00001138         -0.02      0.00     -1.67
>
>
>      NEW-OLD atomic charge density approx. for the potential
>   NEW K-POINTS
>    0.0833333   0.0000000   0.0000000   0.6666667
>    0.2500000   0.0000000   0.0000000   0.6666667
>    0.4166667   0.0000000   0.0000000   0.6666667
>
>      Writing file      40bn.save for program phonon
>
>      PWSCF        :    14h39m CPU time
>
>      init_run     :     4.90s CPU
>      electrons    : 36045.98s CPU (    1000 calls,  36.046 s avg)
>      forces       :  3611.79s CPU (    1000 calls,   3.612 s avg)
>      stress       :  8443.11s CPU (    1000 calls,   8.443 s avg)
>
>      electrons    : 36045.98s CPU (    1000 calls,  36.046 s avg)
>      c_bands      : 21691.54s CPU (    4007 calls,   5.413 s avg)
>      sum_band     :  6934.47s CPU (    4007 calls,   1.731 s avg)
>      v_of_rho     :  1456.36s CPU (    8016 calls,   0.182 s avg)
>      newd         :  2189.58s CPU (    4009 calls,   0.546 s avg)
>      mix_rho      :   191.46s CPU (    4007 calls,   0.048 s avg)
>
>      c_bands      : 21691.54s CPU (    4007 calls,   5.413 s avg)
>      init_us_2    :   245.33s CPU (   30042 calls,   0.008 s avg)
>      cegterg      : 21414.15s CPU (   12021 calls,   1.781 s avg)
>
>      sum_band     :  6934.47s CPU (    4007 calls,   1.731 s avg)
>      becsum       :     4.53s CPU (   12021 calls,   0.000 s avg)
>      addusdens    :  1508.49s CPU (    4007 calls,   0.376 s avg)
>
>      cegterg      : 21414.15s CPU (   12021 calls,   1.781 s avg)
>
>
> Where is output geometry, output cell parameters? As I remember, in past versions it was printed during each step, in this version (2.1) no information at all. Does it mean convergence or not? What is criteria of convergence in vc-relax?
>
>  Thank you very mcuh,
>    Best wishes,
>      Sergey
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>



More information about the users mailing list