[CP2K-user] nonscc dftb problem

[simin pahlavi]

Hello Xin,
Thanks so much for your detailed comments. I have question about them and I 
appriciate if you could help me.

1) What about scc dftb? do we need grids there? I mean as far as I 
undrestoond where ever we use wave-function in cp2k we need grids as it 
uses amixd of Gaussian s and plane-waves.
2) sorry, before I copy the input here,  I cleaned up my comments  so it 
gets easier to read here, and I deleted END QS by mistake
3) But in most of nonscc cases like 100% of cp2k regtest I see that the 
ewald sum is not used so the electrostatic part is calculated form the 
direct equation and not ewald sum.
4) and about mixing I totally agree with you, but in many of co2k regtests 
in nonscc-DFTB there is this mixing part, why?
I appreciate any help in advance,
Simin Pahlavi
On Monday, September 16, 2019 at 9:52:45 AM UTC+2, Xin Wu wrote:
>
> Hello,
>
> On Sunday, September 15, 2019 at 11:49:57 AM UTC+2, simin pahlavi wrote:
>>
>> 1) If my input is set up correctly for these type of calculation.
>>
>
> No. There are several errors in the FORCE_EVAL section:
>
> 1. MGRID is only relevant to DFT, but not NONSCC-DFTB (please see the DFT 
> section in the CP2K manual).
>
> 2. You have an open QS section, please &END QS.
>
> 3. In the POISSON section, you are requesting periodic boundary 
> conditions. Please enable FORCE_EVAL%DFT%QS%DFTB%DO_EWALD also.
>
> 4. There may be some other errors in your input file.
>  
>
>> 2) Normally what the range of cuttof in nonscc dftb should be?
>>
>
> Normally there is no cutoff in DFTB.
>  
>
>> 3) how should I decide on type of mixing , based on convergence speed up 
>> only?
>>
>
> You're using NONSCC-DFTB. As the name has already implied, there is no SCF 
> iteration at all.
>
> The MIXING approach is only for the SCF procedure. It generates new 
> density matrix by mixing the electron density in the current iteration with 
> that from the preceding SCF iteration, so that the variation of electron 
> density during the SCF procedure becomes smooth and therefore it leads 
> accelerated convergence. The default values are robust enough for most of 
> applications. Hence, it's better to use these default settings.
>  
>
>> Thanks in advance,
>>
>
> You're welcome!
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190916/b45535a7/attachment.htm>