From owner-chemistry@ccl.net Mon Feb 29 22:46:00 2016 From: "Robert Molt r.molt.chemical.physics##gmail.com" To: CCL Subject: CCL:G: SCF convergence problem with M06L functional in Gaussian 09 Message-Id: <-52081-160229223058-24730-HntZAFwxcmuGWKnvpTZAvQ|a|server.ccl.net> X-Original-From: Robert Molt Content-Type: multipart/alternative; boundary="------------000209040700050200010102" Date: Mon, 29 Feb 2016 22:30:49 -0500 MIME-Version: 1.0 Sent to CCL by: Robert Molt [r.molt.chemical.physics++gmail.com] This is a multi-part message in MIME format. --------------000209040700050200010102 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: quoted-printable No, in short. There are many complexities to the system. For one, choosing "stable" just checks if the single determinant=20 expression is a minimum. It's not clear that the electronic state=20 corresponding to the minimum is the target state of interest? Second, a frequency calculation specifies if the *geometry* is a local=20 minimum for that electronic state. It does not tell you about other=20 electronic state, nor does it even tell you if that is the geometric=20 minimum in which one is interested (there may be many). Three, I am stating as a matter of experience on simple molecules and=20 complex molecules, *more cycles is not better*. You're just wasting=20 computer time. Very rarely do more cycles help; the problem is=20 describing the electronic structure properly. "How many scf cycles it takes is mostly a function of how good the=20 initial guess is, and that is done by the computer, not the user, so I=20 don't think it reflects on the user's intelligence. " This is categorically untrue. You have to know what electronic state is=20 of interest so it does not go to the wrong state, you have to know what=20 geometric minimum is needed, you have to know if there is vibronic=20 coupling that negates the ability to describe the system easily. The=20 user has to know what (s)he is doing carefully, especially with a=20 metallic system loaded with degeneracies. You have to know what are the=20 limitations of the electronic structure method that may interfere with=20 the SCF process, basis set effects. It's also untrue that the user cannot change the initial guess. The=20 initial density guess is changed all the time by quantum chemists. On 2/29/16 7:58 PM, David Shobe avidshobe]~[yahoo.com wrote: > Can't one use Gaussian's stable command to check for this type of probl= em? > And similarly, wouldn't a freq job show whether the tetrahedral Pd4 is = a minimum? > With transition metals like Pd, I've often used up to even scf(qc,maxcy= c=3D999), but I've always checked the wavefunction with stable or stable=3D= opt. How many scf cycles it takes is mostly a function of how good the i= nitial guess is, and that is done by the computer, not the user, so I don= 't think it reflects on the user's intelligence. :-) [Although a really = bad initial guess can often be improved by guess=3D(read,alter), which do= es require user input]. > --David Shobe > P.S. I'll have to take a look at H4 if I ever get Gaussian working agai= n. :-( > > On Monday, February 29, 2016 3:52 PM, V=C3=ADctor Lua=C3=B1a Cabal= victor::fluor.quimica.uniovi.es wrote: > =20 > > =20 > Sent to CCL by: =3D?iso-8859-1?Q?V=3DEDctor_Lua=3DF1a?=3D Cabal [victor= ^fluor.quimica.uniovi.es] > On Mon, Feb 29, 2016 at 12:08:49PM -0500, Robert Molt r.molt.chemical.p= hysics||gmail.com wrote: >> Sent to CCL by: Robert Molt [r.molt.chemical.physics:-:gmail.com] >> I would encourage you to be very careful. The suggestions given to hav= e >> 500 cycles of optimization or SCF cycle are, IMHO, not good. In genera= l, >> if you cannot converge your SCF of a small molecule in 100 steps, you >> need to be smarter, not more persistent. It speaks more to the idea th= at >> you have a fundamental problem in your description of the electronic >> structure in some regard. I have never encountered a situation in whic= h >> more than 75 steps were needed for 4 atoms. >> >> Just check that your electronic state is really what you want to >> converge to and not some other spurious electronic state. I also echo >> the other comment about concern for using an unrestricted reference. >> > Bulumoni, > > Robert Molt words a quite wise. The case of a tetrahedral molecule > like your Pd4 input is a clear candidate for Jahn-Teller efects in whic= h > the electronic state and the vibrational mode interact to produce > complex phenomena. > > You don't even need to consider Pd4, which can be a nightmare by the > participations of high-L electronic levels. Just H_4 is a good candidat= e > to study the complexity of Jahn-Teller effects. > > Good Luck, > V=C3=ADctor Lua=C3=B1a > -- > . . "De la cuna a la tumba es una escuela, por eso lo que llam= as > / `' \ problemas son lecciones." -- Facundo Cabral, Cuna > /(o)(o)\ "From the cradle to the grave life is a school, ..." > /`. \/ .'\ "Lo mediocre es peor que lo bueno, pero tambi=C3=A9n es = peor > / '`'` \ que lo malo, porque la mediocridad no es un grado, es una > | \'`'`/ | actitud" -- Jorge Wasenberg, 2015 > | |'`'`| | (Mediocre is worse than good, but it is also worse than > \/`'`'`'\/ bad, because mediocrity is not a grade, it is an attitud= e) > =3D=3D=3D(((=3D=3D)))=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D+=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > ! Dr.V=C3=ADctor Lua=C3=B1a, in silico chemist & prof. ! > ! Departamento de Qu=C3=ADmica F=C3=ADsica y Anal=C3=ADtica ! > ! Universidad de Oviedo, 33006-Oviedo, Spain ! > ! e-mail: victor[A]fluor.quimica.uniovi.es ! > ! phone: +34-985-103491 fax: +34-985-103125 ! > +--------------------------------------------+ > GroupPage : http://azufre.quimica.uniovi.es/ > (being reworked) > > > > -=3D This is automatically added to each message by the mailing script = =3D-------=3D_Part_1100286_15699725= 47.1456793890084 > Content-Type: text/html; charset=3DUTF-8 > Content-Transfer-Encoding: quoted-printable > >
Can't one use Gaussian's= stable command to check for this type of problem?

= And similarly, wouldn't a freq job show whether the tetrahedral Pd4 is a = minimum?
With transition metals like Pd, I've often= used up to even scf(qc,maxcyc=3D999), but I've always checked the wavefu= nction with stable or stable=3Dopt.  How many scf cycles it takes is= mostly a function of how good the initial guess is, and that is done by = the computer, not the user, so I don't think it reflects on the user's in= telligence. :-)  [Although a really bad initial guess can often be i= mproved by guess=3D(read,alter), which does require user input].

--David Shobe

P.S. I'll have to take = a look at H4 if I ever get Gaussian working again. :-(


On Monday, February 29, 2016 3:52 PM, = V=C3=ADctor Lua=C3=B1a Cabal victor::fluor.quimica.uniovi.es <owner-ch= emistry.:.ccl.net> wrote:



Sent to CCL by: = =3D?iso-8859-1?Q?V=3DEDctor_Lua=3DF1a?=3D Cabal [victor^fluor.quimica.uni= ovi.es]
On Mon, Feb 29, 2016 at 12:08:49PM -0500, Robert Molt r.molt.c= hemical.physics||gmail.com wrote:
>
> Sent to CCL by: Robert = Molt [r.molt.chemical.physics:-:gmail.com]
> I would encourage you = to be very careful. The suggestions given to have 
> 500 cycl= es of optimization or SCF cycle are, IMHO, not good. In general,  > if you cannot converge your SCF of a small molecule in 100 steps, = you 
> need to be smarter, not more persistent. It speaks mor= e to the idea that 
> you have a fundamental problem in your = description of the electronic 
> structure in some regard. I = have never encountered a situation in which 
> more than 75 s= teps were needed for 4 atoms.
>
> Just check that your electr= onic state is really what you want to 
> converge to and not = some other spurious electronic state. I also echo 
> the othe= r comment about concern for using an unrestricted reference.
>
<= br>Bulumoni,

Robert Molt words a quite wise. The case of a tetrahe= dral molecule
like your Pd4 input is a clear candidate for Jahn-Teller= efects in which
the electronic state and the vibrational mode interac= t to produce
complex phenomena.

You don't even need to conside= r Pd4, which can be a nightmare by the
participations of high-L electr= onic levels. Just H_4 is a good candidate
to study the complexity of J= ahn-Teller effects.

Good Luck,
        V=C3= =ADctor Lua=C3=B1a
--
    .  .    "De la = cuna a la tumba es una escuela, por eso lo que llamas
    / = `' \    problemas son lecciones." -- Facundo Cabral, Cuna
&n= bsp; /(o)(o)\  "From the cradle to the grave life is a school, ..."=
  /`. \/ .'\  "Lo mediocre es peor que lo bueno, pero tambi= =C3=A9n es peor
/  '`'`  \ que lo malo, porque la mediocr= idad no es un grado, es una
|  \'`'`/  | actitud" -- Jorge = Wasenberg, 2015
|  |'`'`|  | (Mediocre is worse than good, = but it is also worse than
  \/`'`'`'\/  bad, because mediocr= ity is not a grade, it is an attitude)
=3D=3D=3D(((=3D=3D)))=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D+=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D
! Dr.V=C3=ADctor Lua=C3=B1a, in silico chemist &= ; prof. !
! Departamento de Qu=C3=ADmica F=C3=ADsica y Anal=C3=ADtica= !
! Universidad de Oviedo, 33006-Oviedo, Spain !
! e-mail: = victor[A]fluor.quimica.uniovi.es  !
! phone: +34-985-103491&n= bsp; fax: +34-985-103125 !
+-----------------------------------------= ---+
GroupPage : http://azufre.quimica.uniovi.es/
      &nb= sp;     (being reworked)



-=3D This is automatic= ally added to each message by the mailing script =3D-
To recover the e= mail address of the author of the message, please change
the strange c= haracters on the top line to the .:. sign. You can also
look up the X-= Original-From: line in the mail header.

E-mail to subscribers: CHEMISTRY.:.ccl.net or use:
      http://www= =2Eccl.net/cgi-bin/ccl/send_ccl_message

E-mail to administrato= rs: CHEMISTRY= -REQUEST.:.ccl.net or use
      http://www.ccl= =2Enet/cgi-bin/ccl/send_ccl_message
      =

Before posting, c= heck wait time at: http= ://www.ccl.net

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

Sea= rch Messages: http://www.ccl.net/chemistry/searchccl/index.shtml<= /a>
&nbs= p;     http://www.ccl.net/spammers.txt

RTFI: http://www.ccl.net/chemistry/aboutccl/instr= uctions/




--------------000209040700050200010102 Content-Type: text/html; charset=utf-8 Content-Transfer-Encoding: 8bit No, in short. There are many complexities to the system.

For one, choosing "stable" just checks if the single determinant expression is a minimum. It's not clear that the electronic state corresponding to the minimum is the target state of interest?

Second, a frequency calculation specifies if the geometry is a local minimum for that electronic state. It does not tell you about other electronic state, nor does it even tell you if that is the geometric minimum in which one is interested (there may be many).

Three, I am stating as a matter of experience on simple molecules and complex molecules, more cycles is not better. You're just wasting computer time. Very rarely do more cycles help; the problem is describing the electronic structure properly.

"How many scf cycles it takes is mostly a function of how good the initial guess is, and that is done by the computer, not the user, so I don't think it reflects on the user's intelligence. "

This is categorically untrue. You have to know what electronic state is of interest so it does not go to the wrong state, you have to know what geometric minimum is needed, you have to know if there is vibronic coupling that negates the ability to describe the system easily. The user has to know what (s)he is doing carefully, especially with a metallic system loaded with degeneracies. You have to know what are the limitations of the electronic structure method that may interfere with the SCF process, basis set effects.

It's also untrue that the user cannot change the initial guess. The initial density guess is changed all the time by quantum chemists.

On 2/29/16 7:58 PM, David Shobe avidshobe]~[yahoo.com wrote:
Can't one use Gaussian's stable command to check for this type of problem?
And similarly, wouldn't a freq job show whether the tetrahedral Pd4 is a minimum?
With transition metals like Pd, I've often used up to even scf(qc,maxcyc=999), but I've always checked the wavefunction with stable or stable=opt.  How many scf cycles it takes is mostly a function of how good the initial guess is, and that is done by the computer, not the user, so I don't think it reflects on the user's intelligence. :-)  [Although a really bad initial guess can often be improved by guess=(read,alter), which does require user input].
--David Shobe
P.S. I'll have to take a look at H4 if I ever get Gaussian working again. :-( 

    On Monday, February 29, 2016 3:52 PM, Víctor Luaña Cabal victor::fluor.quimica.uniovi.es <owner-chemistry.:.ccl.net> wrote:
 

 
Sent to CCL by: =?iso-8859-1?Q?V=EDctor_Lua=F1a?= Cabal [victor^fluor.quimica.uniovi.es]
On Mon, Feb 29, 2016 at 12:08:49PM -0500, Robert Molt r.molt.chemical.physics||gmail.com wrote:

Sent to CCL by: Robert Molt [r.molt.chemical.physics:-:gmail.com]
I would encourage you to be very careful. The suggestions given to have  
500 cycles of optimization or SCF cycle are, IMHO, not good. In general,  
if you cannot converge your SCF of a small molecule in 100 steps, you  
need to be smarter, not more persistent. It speaks more to the idea that  
you have a fundamental problem in your description of the electronic  
structure in some regard. I have never encountered a situation in which  
more than 75 steps were needed for 4 atoms.

Just check that your electronic state is really what you want to  
converge to and not some other spurious electronic state. I also echo  
the other comment about concern for using an unrestricted reference.


Bulumoni,

Robert Molt words a quite wise. The case of a tetrahedral molecule
like your Pd4 input is a clear candidate for Jahn-Teller efects in which
the electronic state and the vibrational mode interact to produce
complex phenomena. 

You don't even need to consider Pd4, which can be a nightmare by the
participations of high-L electronic levels. Just H_4 is a good candidate
to study the complexity of Jahn-Teller effects.

Good Luck,
        Víctor Luaña
--
    .  .    "De la cuna a la tumba es una escuela, por eso lo que llamas
    / `' \    problemas son lecciones." -- Facundo Cabral, Cuna
  /(o)(o)\  "From the cradle to the grave life is a school, ..."
  /`. \/ .'\  "Lo mediocre es peor que lo bueno, pero también es peor
 /  '`'`  \ que lo malo, porque la mediocridad no es un grado, es una
 |  \'`'`/  | actitud" -- Jorge Wasenberg, 2015
 |  |'`'`|  | (Mediocre is worse than good, but it is also worse than
  \/`'`'`'\/  bad, because mediocrity is not a grade, it is an attitude)
===(((==)))==================================+=========================
! Dr.Víctor Luaña, in silico chemist & prof. ! 
! Departamento de Química Física y Analítica ! 
! Universidad de Oviedo, 33006-Oviedo, Spain ! 
! e-mail:  victor[A]fluor.quimica.uniovi.es  ! 
! phone: +34-985-103491  fax: +34-985-103125 ! 
+--------------------------------------------+
 GroupPage : http://azufre.quimica.uniovi.es/
            (being reworked)               Job: http://www.ccl.net/jobs      ------=_Part_1100286_1569972547.1456793890084
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

<html><head></head><body><div style="color:#000; background-color:#fff; font-family:HelveticaNeue, Helvetica Neue, Helvetica, Arial, Lucida Grande, sans-serif;font-size:13px"><div id="yui_3_16_0_1_1456791461898_8236"><span id="yui_3_16_0_1_1456791461898_8353">Can't one use Gaussian's stable command to check for this type of problem?</span></div><div id="yui_3_16_0_1_1456791461898_8236"><span><br></span></div><div id="yui_3_16_0_1_1456791461898_8236"><span id="yui_3_16_0_1_1456791461898_8967">And similarly, wouldn't a freq job show whether the tetrahedral Pd4 is a minimum?</span></div><div id="yui_3_16_0_1_1456791461898_8236"><span><br></span></div><div id="yui_3_16_0_1_1456791461898_8236"><span id="yui_3_16_0_1_1456791461898_8588">With transition metals like Pd, I've often used up to even scf(qc,maxcyc=999), but I've always checked the w
avefunction with stable or stable=opt. &nbsp;How many scf cycles it takes is mostly a function of how good the initial guess is, and that is done by the computer, not the user, so I don't think it reflects on the user's intelligence. :-) &nbsp;[Although a really bad initial guess can often be improved by guess=(read,alter), which does require user input].</span></div><div id="yui_3_16_0_1_1456791461898_8236"><span><br></span></div><div id="yui_3_16_0_1_1456791461898_8236"><span id="yui_3_16_0_1_1456791461898_8449">--David Shobe</span></div><div id="yui_3_16_0_1_1456791461898_8236"><span><br></span></div><div id="yui_3_16_0_1_1456791461898_8236"><span id="yui_3_16_0_1_1456791461898_9498">P.S. I'll have to take a look at H4 if I ever get Gaussian working again. :-(</span></div> <div class="qtdSeparateBR" id="yui_3_16_0_1_1456791461898_8322"><br><
br></div><div class="yahoo_quoted" id="yui_3_16_0_1_1456791461898_8314" style="display: block;"> <div style="font-family: HelveticaNeue, Helvetica Neue, Helvetica, Arial, Lucida Grande, sans-serif; font-size: 13px;" id="yui_3_16_0_1_1456791461898_8313"> <div style="font-family: HelveticaNeue, Helvetica Neue, Helvetica, Arial, Lucida Grande, sans-serif; font-size: 16px;" id="yui_3_16_0_1_1456791461898_8312"> <div dir="ltr" id="yui_3_16_0_1_1456791461898_8321"><font size="2" face="Arial" id="yui_3_16_0_1_1456791461898_8320"> On Monday, February 29, 2016 3:52 PM, Víctor Luaña Cabal victor::fluor.quimica.uniovi.es &lt;owner-chemistry.:.ccl.net&gt; wrote:<br></font></div>  <br><br> <div class="y_msg_container" id="yui_3_16_0_1_1456791461898_8311"><br>Sent to CCL by: =?iso-8859-1?Q?V=EDctor_Lua=F1a?= Cabal [victor^fluor.quimica.uniovi.es]<br>On Mon, Feb 29, 2016 at 12:08:49PM -0500, Robert Molt 
r.molt.chemical.physics||gmail.com wrote:<br>&gt;<br>&gt; Sent to CCL by: Robert Molt [r.molt.chemical.physics:-:gmail.com]<br>&gt; I would encourage you to be very careful. The suggestions given to have&nbsp; <br>&gt; 500 cycles of optimization or SCF cycle are, IMHO, not good. In general,&nbsp; <br>&gt; if you cannot converge your SCF of a small molecule in 100 steps, you&nbsp; <br>&gt; need to be smarter, not more persistent. It speaks more to the idea that&nbsp; <br>&gt; you have a fundamental problem in your description of the electronic&nbsp; <br>&gt; structure in some regard. I have never encountered a situation in which&nbsp; <br>&gt; more than 75 steps were needed for 4 atoms.<br>&gt;<br>&gt; Just check that your electronic state is really what you want to&nbsp; <br>&gt; converge to and not some other spurious electronic state. I 
also echo&nbsp; <br>&gt; the other comment about concern for using an unrestricted reference.<br>&gt;<br><br>Bulumoni,<br><br>Robert Molt words a quite wise. The case of a tetrahedral molecule<br>like your Pd4 input is a clear candidate for Jahn-Teller efects in which<br>the electronic state and the vibrational mode interact to produce<br>complex phenomena. <br><br>You don't even need to consider Pd4, which can be a nightmare by the<br>participations of high-L electronic levels. Just H_4 is a good candidate<br>to study the complexity of Jahn-Teller effects.<br><br>Good Luck,<br>&nbsp; &nbsp; &nbsp; &nbsp;  Víctor Luaña<br>--<br>&nbsp; &nbsp;  .&nbsp; .&nbsp; &nbsp; "De la cuna a la tumba es una escuela, por eso lo que llamas<br>&nbsp; &nbsp; / `' \&nbsp; &nbsp; problemas son lecciones." -- Facundo Cabral, Cu
na<br>&nbsp;  /(o)(o)\&nbsp; "From the cradle to the grave life is a school, ..."<br>&nbsp; /`. \/ .'\&nbsp; "Lo mediocre es peor que lo bueno, pero también es peor<br> /&nbsp;  '`'`&nbsp;  \ que lo malo, porque la mediocridad no es un grado, es una<br> |&nbsp; \'`'`/&nbsp; | actitud" -- Jorge Wasenberg, 2015<br> |&nbsp; |'`'`|&nbsp; | (Mediocre is worse than good, but it is also worse than<br>&nbsp; \/`'`'`'\/&nbsp; bad, because mediocrity is not a grade, it is an attitude)<br>===(((==)))==================================+=========================<br>! Dr.Víctor Luaña, in silico chemist &amp; prof. ! <br>! Departamento de Química Física y Analítica ! <br>! Universidad de Oviedo, 33006-Oviedo, Spain ! <br>! e-mail:&nbsp;  victor[A]fluor.quimica.uniovi.es&nbsp;  ! <br>! phone: +34-985-103491&nbsp; fax: +34-985-103125 ! <br>+----------------
----------------------------+<br> GroupPage : <a href="http://azufre.quimica.uniovi.es/" target="_blank">http://azufre.quimica.uniovi.es/</a><br>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;  (being reworked)<br><br><br><br><br><br>the strange characters on the top line to the .:. sign. You can also<br><br><br>E-mail to subscribers: <a ymailto="mailto:CHEMISTRY.:.ccl.net" href="mailto:CHEMISTRY.:.ccl.net">CHEMISTRY.:.ccl.net</a> or use:<br>&nbsp; &nbsp; &nbsp; <a href="http://www.ccl.net/cgi-bin/ccl/send_ccl_message" target="_blank">http://www.ccl.net/cgi-bin/ccl/send_ccl_message</a><br><br>E-mail to administrators: <a ymailto="mailto:CHEMISTRY-REQUEST.:.ccl.net" href="mailto:CHEMISTRY-REQUEST.:.ccl.net" id="yui_3_16_0_1_1456791461898_8781">CHEMISTRY-REQUEST.:.ccl.net</a> or use<br>&nbsp; &nbsp; &nbsp; <a href="http://www.ccl.net/cgi-bin/ccl/send_ccl_message" target="_blank"
>http://www.ccl.net/cgi-bin/ccl/send_ccl_message</a><br><br<br>&nbsp; &nbsp; &nbsp; <a href="http://www.ccl.net/chemistry/sub_unsub.shtml" target="_blank">http://www.ccl.net/chemistry/sub_unsub.shtml</a><br><br>Before posting, check wait time at: <a href="http://www.ccl.net/" target="_blank">http://www.ccl.net</a><br><br>Job: <a href="http://www.ccl.net/jobs" target="_blank">ht
tp://www.ccl.net/jobs </a><br>Conferences: <a href="http://server.ccl.net/chemistry/announcements/conferences/" target="_blank">http://server.ccl.net/chemistry/announcements/conferences/</a><br><br>Search Messages: <a href="http://www.ccl.net/chemistry/searchccl/index.shtml" target="_blank">http://www.ccl.net/chemistry/searchccl/index.shtml</a><br><br><br>&nbsp; &nbsp; &nbsp; <a href="http
://www.ccl.net/spammers.txt" target="_blank">http://www.ccl.net/spammers.txt</a><br><br>RTFI: <a href="http://www.ccl.net/chemistry/aboutccl/instructions/" target="_blank" id="yui_3_16_0_1_1456791461898_8315">http://www.ccl.net/chemistry/aboutccl/instructions/</a><br><br><br><br><br></div>  </div> </div>  </div></div></body></html>

--------------000209040700050200010102--