[Pw_forum] SCF not stop when it reaches the default conv_thr

Stefano de Gironcoli degironc at sissa.it
Mon Jul 11 09:54:39 CEST 2005


Yong Jiang wrote:

>Dear All,
>
>This is a question about conv_thr:
>
>I use the default value for conv_thr (default = 1.D-6)in my supercell 
>(including 32 Ni atoms). Why the scf calculation is still running when it 
>reaches ethr = 6.39E-07 since interation #22 (pls see below)?
>  
>
because conv_thr is the threshold for  total energy self consistent 
accuracy and
your estimated scf accuracy if far above 1.D-6

Stefano de Gironcoli

>----------------------------------
>     iteration # 21     ecut=    27.00 ryd     beta=0.30
>     Conjugate-gradient style diagonalization
>     ethr =  1.52E-06,  avg # of iterations =  3.0
>
>     total cpu time spent up to now is  31997.23 secs
>
>     total energy              = -2748.70062200 ryd
>     estimated scf accuracy    <     0.00204526 ryd
>
>     total magnetization       =    22.86 Bohr mag/cell
>     absolute magnetization    =    26.19 Bohr mag/cell
>
>     iteration # 22     ecut=    27.00 ryd     beta=0.30
>     Conjugate-gradient style diagonalization
>     ethr =  6.39E-07,  avg # of iterations =  3.1
>
>     total cpu time spent up to now is  33483.75 secs
>
>     total energy              = -2748.68225286 ryd
>     estimated scf accuracy    <     0.02254775 ryd
>
>     total magnetization       =    22.86 Bohr mag/cell
>     absolute magnetization    =    26.22 Bohr mag/cell
>
>     iteration # 23     ecut=    27.00 ryd     beta=0.30
>     Conjugate-gradient style diagonalization
>     ethr =  6.39E-07,  avg # of iterations =  3.1
>
>     total cpu time spent up to now is  35054.16 secs
>
>     total energy              = -2748.69429201 ryd
>     --More--(98%)
>--------------------------------------------
>
>Thanks very much in advance for your kind help.
>
>Good weekend,
>
>Yong 
>_______________________________________________
>Pw_forum mailing list
>Pw_forum at pwscf.org
>http://www.democritos.it/mailman/listinfo/pw_forum
>  
>







More information about the users mailing list