This thread is locked.Only browsing is available.
Top Page > Browsing
optimization problem
Date: 2020/10/21 16:04
Name: reza   <>

Hello to all
I have faced with the problem in selecting optimized lattice constant. my first calculation results are below. scf criterion: 10-6

a=5.117 -1722.4509 7233 normRD:3*10-5
a=5.107 -1722.4516 2818 normRD:5*10-3
a=5.097 -1722.4516 0752 normRD:10-5
a=5.087 -1722.4513 8141 normRD:10-5

and if we continue calculations to reach higher norm RD with restart file:

a=5.117 -1722.4509 7300 normRD:10-6
a=5.107 -1722.4514 6913 normRD:10-6
a=5.097 -1722.4516 0729 normRD:10-6
a=5.087 -1722.4513 8118 normRD:10-6

wich is true? is norm RD important? or just fulfilling the scf.criterion:10-6 is enough.
could some one give me any idea ?
thanks a lot.

from users manualof openmx:

The keyword 'scf.criterion' specifies a convergence criterion (Hartree) for the SCF calculation. The SCF iteration is ended when a condition, dUele<scf.criterion, is satisfied, where dUele is defined as the absolute deviation between the eigenvalue energy at the current and previous SCF steps. The default is 1.0e-6 (Hartree).

In the NEGF method, the SCF criterion given by the keyword 'scf.criterion' is applied to the residual norm between the input and output charge densities 'NormRD', while in the other cases 'dUele' is monitored.

Page: [1]

Re: optimization problem ( No.1 )
Date: 2020/10/21 16:16
Name: T. Ozaki


For the case of a=5.107 in the former, the SCF convergence may not be enough
compared to the other cases. So, the latter calculations are trustable.

The minimum seems to locate in between 5.097 and 5.107.


Re: optimization problem ( No.2 )
Date: 2020/10/21 20:24
Name: reza  <>

thank you for your reply prof.ozaki

Page: [1]