Re: nonscc dftb problem

60 views
Skip to first unread message
Message has been deleted

Xin Wu

unread,
Sep 16, 2019, 3:52:45 AM9/16/19
to cp2k
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!


simin pahlavi

unread,
Sep 16, 2019, 5:42:21 AM9/16/19
to cp2k
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

Thomas Kühne

unread,
Sep 16, 2019, 8:32:03 AM9/16/19
to cp...@googlegroups.com
Dear Simin, 

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.
also no, it doesn’t make any difference if you are using scc or not. 

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.
All regtests in the nonscc folder are with PERIODIC NONE, i.e. isolated clusters.

Cheers, 
Thomas

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!



-- 
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/b79b32cf-efcc-4110-87b4-efb6a00a7a35%40googlegroups.com.




simin pahlavi

unread,
Sep 16, 2019, 2:05:22 PM9/16/19
to cp2k

Dear Thomas,
Thanks so much for your reply and help.
Regards,
Simin
Dear Simin, 

To unsubscribe from this group and stop receiving emails from it, send an email to cp...@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages