SCF loop can not converge with using hybrid functional

[Marcella Iannuzzi]

Hi Jiabo,

Well, difficult to say whether it was really successful, but I could 
converge PBE0 calculations of molecules adsorbed on gold. 
In your case, I would try for sure starting from the PBE wfn, as suggested 
before. 
Then, considering that it is a metallic system, probably more patience is 
needed. It might require more than 200 iterations to converge.
Try the usual tricks for metals: increase the T of smearing and reduce the 
mixing parameters.

BTW, have you tried to converge the PBE0 SCF for the two subsystems 
independently, i.e., water and metal?

kind regards
Marcella


On Monday, November 2, 2015 at 1:56:05 PM UTC+1, jiabo wrote:
>
> Hi Marcella
>
> Thanks for your reply, I have tried to run some tests according to your 
> suggestions, but failed. Adjusting the EPS_SCHWARZ from 10^-7 to 10^-10 
> will increase huge computer memory, so currently I can not run a job this 
> optimised parameter, and increasing CUTOFF_RADIUS does not help to 
> converge, neither. Do you know anyone who has calculated the metal system 
> with hybrid functional implemented  in CP2K successfully?
>
> ciao
> Jiabo 
>
> On Wednesday, 28 October 2015 16:26:34 UTC, Marcella Iannuzzi wrote:
>>
>> Hi Jiabo,
>>
>> Probably both EPS_SCHWARZ and CUTOFF_RADIUS are not optimal. 
>> I would use something between 10^-10/10^-12 for the first and more than 
>> 10 for the cutoff.
>> best
>>
>> Marcella
>>
>>
>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20151102/ae0b5f50/attachment.htm>