[Pw_forum] relax question

[Stefano de Gironcoli]

a tipical bond stiffness, if I recall correctly, is in the range of 0.1-1.0
Ry/bohr^2 thus a threshold of 1.d-6 Ry/bohr in the force means that you are
asking a numerical accuracy in your interatomic distances of the order of
1.d-5-1.d-6 bohr... are you sure you really need it ?
best regards,
stefano deGironcoli, SISSA & DEMOCRITOS

Quoting xu yuehua <njuxuyuehua at gmail.com>:

> hello all
> i am doing relax.
> in my inputfile,
> i use  etot_conv_thr=1.D-6
>   forc_conv_thr=1.D-6
> and the output is
>    atom   1 type  1   force =     0.00000290    0.00000095   -0.00000364
>      atom   2 type  1   force =     0.00000081   -0.00000108    0.00000033
>      atom   3 type  1   force =    -0.00000435   -0.00000425   -0.00000160
>      atom   4 type  1   force =    -0.00000312   -0.00000321   -0.00000080
>      atom   5 type  1   force =    -0.00000446   -0.00000036   -0.00000273
>      atom   6 type  1   force =    -0.00000048    0.00000230    0.00000112
>      atom   7 type  1   force =     0.00000164    0.00000118    0.00000060
>      atom   8 type  1   force =     0.00000131   -0.00000040    0.00000145
>      atom   9 type  2   force =    -0.00000362    0.00000154   -0.00000201
>      atom  10 type  2   force =    -0.00000029    0.00000191    0.00000302
>      atom  11 type  2   force =     0.00000576   -0.00000134    0.00000194
>      atom  12 type  2   force =     0.00000390    0.00000276    0.00000233
> 
> Total force =     0.000015     Total SCF correction =     0.000054
>      SCF correction compared to forces is too large, reduce conv_thr
> 
>      The maximum number of steps has been reached.
> my qustion is ,force at every atom is ok ,but why  the program asks to
> reduce conv_thr
> my conv_thr is not enough small ?
> or my total relax step (50) is not adequate?
> is there any principle about these parameter?
> need your help
> thanks a lot
> -- 
> Xu Yuehua
> physics Department of Nanjing university
> China
> 




----------------------------------------------------------------
  SISSA Webmail https://webmail.sissa.it/
  Powered by Horde http://www.horde.org/