61 atom Fe wire lead, achieving convergance|
- Date: 2010/03/05 01:29
- Name: Gabriel Greene
I have recently been running a calculation of a 61 atom Fe wire structure. I believe am close to achieving convergance, dUele decreases consistently with each iteration, and final values of dUele are usually ~ 0.0004, while the scf.criterion is 1e-6, so i think complete convergance is pretty close. So to achieve convergance I have been trying various values for the mixing parameters, and i have been keeping in mind the advice given on the website concerning the kerker factor and mixing parameters for metal systems.
for scf.init.mixing.weight i have tried:- 0.3, 0.1, 0.03, 0.01, 0.006, 0.003
for the scf.min.mixing.weight i have tried:- 0.001, and 0.002
for scf.max.mixing.weight iv tried:- 0.002, 0.004, 0.006, 0.04, 0.08, 0.4
for scf.mixing.history iv tried:- 30, 40, and 60
the usual mixing schemes are either kerker or rmm-diisk
however despite all these attempts and the different combinations of these parameters, convergance is never fully achieved within 200 iterations, i could increase the max number of iterations, but more than 200 would impracticle.
any advice on speeding up convergance 2would be greatly appreciated, thank you