From owner-chemistry@ccl.net Tue Nov 23 07:33:01 2021 From: "Sumangla Arora arorasumangla[]gmail.com" To: CCL Subject: CCL: Why on increasing the SCF maxiter, the undesired modes vanish? Message-Id: <-54516-211123073055-24467-nZvo4YgxuPBLulvEAcQ1hQ\a/server.ccl.net> X-Original-From: Sumangla Arora Content-Type: multipart/alternative; boundary="000000000000202ecf05d173e917" Date: Tue, 23 Nov 2021 18:03:06 +0530 MIME-Version: 1.0 Sent to CCL by: Sumangla Arora [arorasumangla%x%gmail.com] --000000000000202ecf05d173e917 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable Dear Neese and Christoph, I have completed the full project using Orca-4.2.1. Only two transition states are not converged. Can I use Orca-5.0 to find these two transition states although all geometries are calculated using Orca-4.2.1? Thanks Sumangla On Mon, Nov 22, 2021 at 11:29 PM Neese, Frank neese * kofo.mpg.de < owner-chemistry~!~ccl.net> wrote: > Dear Herbert, Dear Brian, > > We had indeed discussed that point recently among the developers. My > opinion is that an unconverged SCF at the beginning of an optimization is > fine for the reasons that you stated. However, an optimization should nev= er > end with an unconverged SCF. In addition, an unconverged SCF during the > numerical Hessian calculation must lead to an aborted job as the results > cannot possibly be reliable. > > This is how it will be. > > Fortunately, with the new SCF converger and the overall reduced numerical > noise from the new grids, this will hardly be an issue anymore. > > Thanks for the input! > Best wishes, > Frank > > Am 22.11.2021 um 13:00 schrieb Brian Skinn brian.skinn|-|gmail.com < > owner-chemistry+*+ccl.net>: > > Christoph, > > Indeed, please make sure to discuss this with Prof. Neese before > implementing. I can't find the post now, but I'm pretty sure there was a > discussion on the ORCA forum a while back about this: there may be cases > where users want/need to proceed to a property calculation even in the > absence of a converged-SCF flag, as with Prof. Fruchtl's geometry > optimization example, and such calculations then become impossible if the > software refuses to continue. > > -Brian > > > On Mon, Nov 22, 2021 at 6:51 AM Herbert Fruchtl herbert.fruchtl[A] > st-andrews.ac.uk wrote: > >> >> Sent to CCL by: Herbert Fruchtl [herbert.fruchtl=3D-=3Dst-andrews.ac.uk] >> Hi Christoph, >> >> > A meaningful warning was printed by ORCA, but fully agreed, the >> > calculation should have stopped. We'll change that for the next ORCA >> > (bugfix) release. >> > >> I agree that this should happen in a frequency calculation, but please >> don't do it in geometry optimizations. I sometimes had the problem (with >> older ORCA versions) that when starting at a bad geometry, the SCF >> wouldn't converge. An approximately correct gradient that points in the >> roughly right direction at the start of an optimization is better than >> the program stopping. >> >> Cheers, >> >> Herbert >> -- >> Herbert Fruchtl >> Senior Scientific Computing Officer / HPC Administrator >> School of Chemistry, IT Services >> University of St Andrews >> -- >> The University of St Andrews is a charity registered in Scotland: >> No SC013532 >> >> >> >> -=3D This is automatically added to each message by the mailing script = =3D- >> >> >> E-mail to subscribers: CHEMISTRY,ccl.net or use:>> >> E-mail to administrators: CHEMISTRY-REQUEST,ccl.net or use>> >> >> > > -------------------------------------------------------------------------= --- > Prof. Dr. Frank Neese > Department of Molecular Theory and Spectroscopy > Max-Planck Institut f=C3=BCr Kohlenforschung > Kaiser-Wilhelm-Platz 1 > D-45470 M=C3=BClheim an der Ruhr > Germany > E-Mail: Frank.Neese+*+kofo.mpg.de > > -------------------------------------------------------------------------= --- > > --000000000000202ecf05d173e917 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable

Dear Neese and=C2=A0Christoph,

I have completed the full project using Orca-4.2.= 1. Only two transition states are not converged. Can I use Orca-5.0 to find these two transition states although all geometries are calculated using Orca-4.2.1?<= /p>


Thanks

Sumangla<= /p>


On Mon, Nov 22, 2021 at 11:29 PM Neese, Frank neese * kofo.mpg.de <owner-chemistry~!~ccl.net> wrote:
Dear Herb= ert, Dear Brian,

We had= indeed discussed that point recently among the developers. My opinion is t= hat an unconverged SCF at the beginning of an optimization is fine for the = reasons that you stated. However, an optimization should never end with an = unconverged SCF. In addition, an unconverged SCF during the numerical Hessi= an calculation must lead to an aborted job as the results cannot possibly b= e reliable.=C2=A0

This is how it will be.<= br style=3D"color:rgb(0,0,0)">
Fortunately, with the new SCF converger and the overall = reduced numerical noise from the new grids, this will hardly be an issue an= ymore.=C2=A0

Thanks for the input!
Best wishes,=
Frank

Am = 22.11.2021 um 13:00 schrieb Brian Skinn brian.skinn|-|gmail.com <owner-chemistry+*+ccl.net>:
Christoph,

Indeed, please make sure to discuss this with Prof. Neese before impl= ementing. I can't find the post now, but I'm pretty sure there was = a discussion on the ORCA forum a while back about this: there may be cases = where users want/need to proceed to a property calculation even in the abse= nce of a converged-SCF flag, as with Prof. Fruchtl's geometry optimizat= ion example, and such calculations then become impossible if the software r= efuses to continue.

-Brian


On = Mon, Nov 22, 2021 at 6:51 AM Herbert Fruchtl herbert.fruchtl[A]st-andrews.ac.uk <owner-chemistry,ccl.n= et> wrote:
st-andrews.ac.uk]=
Hi Christoph,

> A meaningful warning was printed by ORCA, but fully agreed, the
> calculation should have stopped. We'll change that for the next OR= CA
> (bugfix) release.
>
I agree that this should happen in a frequency calculation, but please
don't do it in geometry optimizations. I sometimes had the problem (wit= h
older ORCA versions) that when starting at a bad geometry, the SCF
wouldn't converge. An approximately correct gradient that points in the=
roughly right direction at the start of an optimization is better than
the program stopping.

Cheers,

=C2=A0 =C2=A0Herbert
--
Herbert Fruchtl
Senior Scientific Computing Officer / HPC Administrator
School of Chemistry, IT Services
University of St Andrews
--
The University of St Andrews is a charity registered in Scotland:
No SC013532



-=3D This is automatically added to each message by the mailing script =3D-=


E-mail to subscribers: CHEMISTRY,ccl.net or use:
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message

E-mail to administrators: CHEMISTRY-REQUEST,ccl.net or use
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message


=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/chemistry/sub_un= sub.shtml

Before posting, check wait time at: http://www.ccl.net

Job: http://www.ccl.net/jobs
Conferences: http://server.ccl.net/chemist= ry/announcements/conferences/

Search Messages: http://www.ccl.net/chemistry/sear= chccl/index.shtml


=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/spammers.txt

RTFI: http://www.ccl.net/chemistry/aboutccl/ins= tructions/



----------------------------------------------------------------------= ------
Prof. Dr. Frank Neese=C2=A0
Department of Mole= cular Theory and=C2=A0Spectroscopy
Max-Planck Institut f=C3= =BCr Kohlenforschung=C2=A0
Kaiser-Wilhelm-Platz 1
D-45470 M=C3=BClheim an der Ruhr=C2=A0
Germany
E-Mail: F= rank.Neese+*+kofo.mpg.de=C2=A0
--------------------------= --------------------------------------------------

--000000000000202ecf05d173e917--