<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<br>
Dear Gabriel and others,<br>
<br>
Having now attempted to change the value of idir to other numbers I
still have the same problem. The issue arises because of the varying
step length taken depending on the curvature. When the gradient is
at its largest value molpro takes a larger step. This should
ofcourse correspond to a large change in the energy and thus
correspondingly a large change in the CAS. As a consequence of the
fact that it takes the previous points orbitals as an initial guess,
which are no longer suitable, it fails to converge the CAS within
the maximum number of iterations (set to 40 the maximum molpro
allows for a single CAS). Changing idir to other values increased
the number of steps taken before arriving at the problem region but
it does not solve the problem as the change between the two regions
is still to large.<br>
<br>
Regards<br>
<br>
Phillip<br>
<br>
<br>
<br>
On 22/11/2011 15:30, Gabriel Freitas wrote:
<blockquote
cite="mid:CAOr-wM2Py8z0zvEBDSZCcasYfSJqRPtkrfZRiPZ45aQjeN_-9A@mail.gmail.com"
type="cite">Dear Valérie and other collegues,<br>
<br>
I would also appreciatte the modification in the code regarding
the freedom in choosing guesses for IRCs / geometry optimzations /
frequencies calculations.<br>
<br>
Regarding the IRC step, you can choose it with the DIRECTION
option in optg directive. In addition to choosing the values +1 /
-1 for positive and negative directions, you can choose +n / -n (n
natural), which exactly determines the step lenght.<br>
<br>
I had also a similar problem with some convergence problems with
ROHF / CCSD(T) IRCs, and I found out that it had a crossing of
states near the problematic point. Although CAS can sometimes
change smoothly between sates, but not always. So I think it's
worthy to check the proximity of excited states near these points
(with SA-CAS)<br>
<br>
I also had convergence problems calculation quasi-symmetrical
structures (for example, numerical frequencies near structures
with symmetry higher than C1), and it had nothing to do with
excited states.<br>
<br>
Regards,<br>
<br>
-- <br>
Gabriel do Nascimento Freitas<br>
D.Sc. Student - Graduate Program of Chemistry<br>
Molecular Modelling and Theoretical Chemistry Laboratory - Room
412<br>
Chemistry Institute - Universidade Federal do Rio de Janeiro
(UFRJ) - Brazil <br>
(+5521)8830-9971 / (+5521)2562-7179<br>
<br>
<div class="gmail_quote">2011/11/22 Valérie Vallet <span
dir="ltr"><<a moz-do-not-send="true"
href="mailto:valerie.vallet@univ-lille1.fr">valerie.vallet@univ-lille1.fr</a>></span><br>
<blockquote class="gmail_quote" style="margin:0 0 0
.8ex;border-left:1px #ccc solid;padding-left:1ex;">
Dear Joachim,<br>
<br>
We have tried the failsafe option in the casscf block, but
this does not improve the CASSCF convergence along the IRC.<br>
Could you please modify the molpro code so that we can force
the casscf step to restart from HF orbitals?<br>
Would it also be possible to offer the possibility to change
the step size in the IRC calculation, since this may help
improving the CASSCF convergence along the IRC?<br>
<div class="im HOEnZb"><br>
Thanks a lot for your help,<br>
<br>
Valérie<br>
<br>
</div>
<div class="HOEnZb">
<div class="h5">On Nov 20, 2011, at 5:33 PM, Hans-Joachim
Werner wrote:<br>
<br>
> Dear Valérie and Jacky,<br>
> I will look at this. This was once changed to make
sure that in geometry optimizations it always<br>
> uses the same options and starting orbitals from the
previous calculation, which is normally most<br>
> sensible. I will make an option to disable this
behavior. In the meantime you could try the<br>
> "FAILSAFE" option, which enables stricter convergence
criteria. Perhaps that helps finding the correct<br>
> solution.<br>
> Best wishes<br>
> Joachim<br>
><br>
><br>
> --<br>
> Prof. Hans-Joachim Werner<br>
> Institute for Theoretical Chemistry<br>
> University of Stuttgart<br>
> Pfaffenwaldring 55<br>
> D-70569 Stuttgart, Germany<br>
> Tel.: (0049) 711 / 685 64400<br>
> Fax.: (0049) 711 / 685 64442<br>
> e-mail: <a moz-do-not-send="true"
href="mailto:werner@theochem.uni-stuttgart.de">werner@theochem.uni-stuttgart.de</a><br>
><br>
><br>
><br>
> Am 18.11.2011 um 16:22 schrieb Valérie Vallet:<br>
><br>
>> Dear MOLPRO users,<br>
>><br>
>> We are trying to run an IRC path at the CASSCF
level, and we are facing a case where the automatic CASSCF
calculation that uses previous CASSCF orbitals along the
reaction coordinate converges to the wrong energy, that is
an energy higher than the that one would obtain with a
single point calculation using Hartree-Fock orbitals.<br>
>><br>
>> We thus would like to force the CASSCF to use HF
orbitals instead at each geometry steps. However, it seems
that the optg keyword ignores "start" commands in the
MCSCF step. Is there an option that allows us to force
optg to read "start" instructions?<br>
>><br>
>> Thanks a lot for your help,<br>
>><br>
>> Valérie Vallet<br>
>><br>
>><br>
>><br>
>> ---<br>
>> Valérie Vallet<br>
>> Research assistant<br>
>> Laboratoire PhLAM, CNRS UMR 8523<br>
>> Université des Sciences et Technologies de Lille<br>
>> Bât P5<br>
>> F-59655 Villeneuve d'Ascq, France<br>
>> Tel: <a moz-do-not-send="true"
href="tel:%2B33%203%202033%205985" value="+33320335985">+33
3 2033 5985</a><br>
>> Fax: <a moz-do-not-send="true"
href="tel:%2B33%203%202033%207020" value="+33320337020">+33
3 2033 7020</a><br>
>> E-mail: <a moz-do-not-send="true"
href="mailto:valerie.vallet@univ-lille1.fr">valerie.vallet@univ-lille1.fr</a><br>
>> <a moz-do-not-send="true"
href="http://www.phlam.univ-lille1.fr/pub/f/themas/pcmt/"
target="_blank">http://www.phlam.univ-lille1.fr/pub/f/themas/pcmt/</a><br>
>><br>
>> _______________________________________________<br>
>> Molpro-user mailing list<br>
>> <a moz-do-not-send="true"
href="mailto:Molpro-user@molpro.net">Molpro-user@molpro.net</a><br>
>> <a moz-do-not-send="true"
href="http://www.molpro.net/mailman/listinfo/molpro-user"
target="_blank">http://www.molpro.net/mailman/listinfo/molpro-user</a><br>
><br>
<br>
_______________________________________________<br>
Molpro-user mailing list<br>
<a moz-do-not-send="true"
href="mailto:Molpro-user@molpro.net">Molpro-user@molpro.net</a><br>
<a moz-do-not-send="true"
href="http://www.molpro.net/mailman/listinfo/molpro-user"
target="_blank">http://www.molpro.net/mailman/listinfo/molpro-user</a><br>
</div>
</div>
</blockquote>
</div>
<br>
<br>
</blockquote>
<br>
</body>
</html>