[CP2K-user] [CP2K:13675] Energy convergence failed on large surface system?

[jt Yang]

Hi Thomas,

Thanks for the reply,

To more clarify, I post this question because there may be more functional 
will be tested and I want a relatively high accuracy(*e-6 a.u.* say) at the 
same time, so a big safe cutoff is needed.

I'll try EPS_SCF and EPS_DEFAULT later, thank you!

Best wishes,
jtYang

On Wednesday, July 29, 2020 at 10:19:45 PM UTC+8, tkuehne wrote:
>
> Dear jt Yang, 
>
> the basis set convergence you observed is already very good! Increasing 
> the cutoff values further, but don’t forget that other parameters such as 
> EPS_SCF and EPS_DEFAULT may have a much bigger impact. 
> In general, you may ask yourself if your target is to single out the PBE 
> limit, 
> or to do the most accurate calculation within given computational 
> resources. 
> In the case of the latter, available computer time is most likely better 
> spend 
> by simply increasing the level of theory beyond GGA-based DFT than 
> painstakingly trying to improve upon an already converged basis. 
>
> Cheers, 
> Thomas
>
> Am 28.07.2020 um 08:23 schrieb jt Yang <jt... at gmail.com <javascript:>>:
>
> Hi,
>
> I got a energy converge problem on a surface system.
>
> - system: TiO2 slab with MeCN adsorbed at it, 192 atoms, vaccum layer 20 
> angs.
> - with `NGRIDS=4, REL_CUTOFF=80, CUTOFF=1200` and NLCC basis, as adviced 
> by the post <https://groups.google.com/forum/#!topic/cp2k/GwdlTOVgBYc>, 
> the convergence i got is e-5 a.u.
> - could I push it to e-6 a.u. with higher `NGRIDS` or `REL_CUTOFF`, is 
> that make sense?
> - or for a system like this, it's not realistic to that high accuracy?
>
> here is my result:
> relcutoff cutoff E a.u. (PBE)
> 80 600 -4644.229268091786
> 80 700 -4644.229265371339
> 80 800 -4644.229242358915
> 80 900 -4644.229242904181
> 80 1000 -4644.229242703446
> 80 1100 -4644.229243509899
> 80 1200 -4644.229242879066
>
> Best wishes,
> jtYang
>
> -- 
> 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 c... at googlegroups.com <javascript:>.
> To view this discussion on the web visit 
> https://groups.google.com/d/msgid/cp2k/3515c775-9af8-4c3f-a6b2-c6c4153e1d35o%40googlegroups.com 
> <https://groups.google.com/d/msgid/cp2k/3515c775-9af8-4c3f-a6b2-c6c4153e1d35o%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>
>
>
>
> ==============================
> Thomas D. Kühne
> Dynamics of Condensed Matter
> Chair of Theoretical Chemistry
> University of Paderborn
> Warburger Str. 100
> D-33098 Paderborn
> Germany
> td... at mail.upb.de <javascript:>
> +49/(0)5251/60-5726
>
>
On Wednesday, July 29, 2020 at 10:19:45 PM UTC+8, tkuehne wrote:
>
> Dear jt Yang, 
>
> the basis set convergence you observed is already very good! Increasing 
> the cutoff values further, but don’t forget that other parameters such as 
> EPS_SCF and EPS_DEFAULT may have a much bigger impact. 
> In general, you may ask yourself if your target is to single out the PBE 
> limit, 
> or to do the most accurate calculation within given computational 
> resources. 
> In the case of the latter, available computer time is most likely better 
> spend 
> by simply increasing the level of theory beyond GGA-based DFT than 
> painstakingly trying to improve upon an already converged basis. 
>
> Cheers, 
> Thomas
>
> Am 28.07.2020 um 08:23 schrieb jt Yang <jt... at gmail.com <javascript:>>:
>
> Hi,
>
> I got a energy converge problem on a surface system.
>
> - system: TiO2 slab with MeCN adsorbed at it, 192 atoms, vaccum layer 20 
> angs.
> - with `NGRIDS=4, REL_CUTOFF=80, CUTOFF=1200` and NLCC basis, as adviced 
> by the post <https://groups.google.com/forum/#!topic/cp2k/GwdlTOVgBYc>, 
> the convergence i got is e-5 a.u.
> - could I push it to e-6 a.u. with higher `NGRIDS` or `REL_CUTOFF`, is 
> that make sense?
> - or for a system like this, it's not realistic to that high accuracy?
>
> here is my result:
> relcutoff cutoff E a.u. (PBE)
> 80 600 -4644.229268091786
> 80 700 -4644.229265371339
> 80 800 -4644.229242358915
> 80 900 -4644.229242904181
> 80 1000 -4644.229242703446
> 80 1100 -4644.229243509899
> 80 1200 -4644.229242879066
>
> Best wishes,
> jtYang
>
> -- 
> 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 c... at googlegroups.com <javascript:>.
> To view this discussion on the web visit 
> https://groups.google.com/d/msgid/cp2k/3515c775-9af8-4c3f-a6b2-c6c4153e1d35o%40googlegroups.com 
> <https://groups.google.com/d/msgid/cp2k/3515c775-9af8-4c3f-a6b2-c6c4153e1d35o%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>
>
>
>
> ==============================
> Thomas D. Kühne
> Dynamics of Condensed Matter
> Chair of Theoretical Chemistry
> University of Paderborn
> Warburger Str. 100
> D-33098 Paderborn
> Germany
> td... at mail.upb.de <javascript:>
> +49/(0)5251/60-5726
>
>
On Wednesday, July 29, 2020 at 10:19:45 PM UTC+8, tkuehne wrote:
>
> Dear jt Yang, 
>
> the basis set convergence you observed is already very good! Increasing 
> the cutoff values further, but don’t forget that other parameters such as 
> EPS_SCF and EPS_DEFAULT may have a much bigger impact. 
> In general, you may ask yourself if your target is to single out the PBE 
> limit, 
> or to do the most accurate calculation within given computational 
> resources. 
> In the case of the latter, available computer time is most likely better 
> spend 
> by simply increasing the level of theory beyond GGA-based DFT than 
> painstakingly trying to improve upon an already converged basis. 
>
> Cheers, 
> Thomas
>
> Am 28.07.2020 um 08:23 schrieb jt Yang <jt... at gmail.com <javascript:>>:
>
> Hi,
>
> I got a energy converge problem on a surface system.
>
> - system: TiO2 slab with MeCN adsorbed at it, 192 atoms, vaccum layer 20 
> angs.
> - with `NGRIDS=4, REL_CUTOFF=80, CUTOFF=1200` and NLCC basis, as adviced 
> by the post <https://groups.google.com/forum/#!topic/cp2k/GwdlTOVgBYc>, 
> the convergence i got is e-5 a.u.
> - could I push it to e-6 a.u. with higher `NGRIDS` or `REL_CUTOFF`, is 
> that make sense?
> - or for a system like this, it's not realistic to that high accuracy?
>
> here is my result:
> relcutoff cutoff E a.u. (PBE)
> 80 600 -4644.229268091786
> 80 700 -4644.229265371339
> 80 800 -4644.229242358915
> 80 900 -4644.229242904181
> 80 1000 -4644.229242703446
> 80 1100 -4644.229243509899
> 80 1200 -4644.229242879066
>
> Best wishes,
> jtYang
>
> -- 
> 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 c... at googlegroups.com <javascript:>.
> To view this discussion on the web visit 
> https://groups.google.com/d/msgid/cp2k/3515c775-9af8-4c3f-a6b2-c6c4153e1d35o%40googlegroups.com 
> <https://groups.google.com/d/msgid/cp2k/3515c775-9af8-4c3f-a6b2-c6c4153e1d35o%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>
>
>
>
> ==============================
> Thomas D. Kühne
> Dynamics of Condensed Matter
> Chair of Theoretical Chemistry
> University of Paderborn
> Warburger Str. 100
> D-33098 Paderborn
> Germany
> td... at mail.upb.de <javascript:>
> +49/(0)5251/60-5726
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20200729/dbdf270b/attachment.htm>