ADMM with Mixing

[Matt W]

On Monday, March 9, 2015 at 5:59:45 PM UTC, M. Brehm wrote:
>
> Hi Matt,
>
> thanks for your quick and helpful reply. Then I will just use no 
> purification, and perform some tests on accuracy depending on the size of 
> the aux basis. It is Ok for me to use a larger aux basis set. The most 
> important thing is that forces and stress tensor works somehow, because 
> otherwise no CELL_OPT will be possible at all.
>
> It should work fine. Maybe if you have a cubic system, you can try 
comparing a series of calcs at different volumes to the cell opt calc to be 
certain all is well.

 

> Or do you think I should try to calculate my lattices with OT and without 
> smearing? Is there at least a chance that this might give useful results?
>
> You could try if you like, but without smearing it is very difficult to 
really converge the SCF cycle so you can get quite some noise in the 
forces. If the density of states is small around the fermi level you might 
get away with it...

Matt

 

> Best regards,
> Martin
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20150310/28415f67/attachment.htm>