[molpro-user] Problem regarding IRC with method QSDPATH

G Mahesh gmahesh at students.iiserpune.ac.in
Tue Aug 27 18:02:16 BST 2013


Dear Ramon,

Thanks for the suggestions, Now the problem is almost solved. First I have
done transition state optimization by using central differences method to
calculate Hessian as you suggested and got transition state with one
imaginary frequency and I also have started IRC with this transition state
geometry. Now, though it's still running but, it's doing well which means
it has done 6 steps yet. So, with this I don't have any problems.

Now, I have small doubt which I'll explain below as much clear as I can.

First, by using  the above mentioned procedure I got transition state, here
is the partial output which is showing that the energy is converged.

  Numerical MULTI000 hessian completed. CPU-time: 14959.95 sec, Elapsed:
15756.64 sec

 MULTI hessian saved to record  5300.2
   4  -353.92204742  -353.92205201    -0.00000458  0.00262178  0.00270543
0.00030060  0.01644770  0.02796584  0.01250670  30537.85

   5  -353.92205201  *-353.92205134*     0.00000067  0.00016687
0.00021778  0.00002420  0.02154643  0.02706698  0.01210472  31486.08

 END OF GEOMETRY OPTIMIZATION.


Now I have started IRC with the coordinates of TS which I got from above
output. Here (IRC input), first it'll do transition state optimization and
after that it'll start IRC. Since I am doing transition state optimization
with transition state geometry it supposed to be done within single step.
But, it hasn't, it took 4 steps again to converge. Let me post the part of
IRC output too.


   1  *-353.92205134*  -353.92205323    -0.00000190  0.00048402
0.00056518  0.00006280  0.05066241  0.05339754  0.02388011  13641.47
   2  -353.92205323  -353.92205329    -0.00000005  0.00067958  0.00073450
0.00008161  0.01391002  0.01394040  0.00623434  14417.26
   3  -353.92205329  -353.92205336    -0.00000008  0.00032270  0.00033121
0.00003680  0.00648028  0.00668994  0.00299183  15217.03

 Numerical MULTI000 hessian completed. CPU-time: 13239.33 sec, Elapsed:
13652.13 sec

 MULTI hessian saved to record  5300.2
   4  -353.92205336  -353.92205342    -0.00000006  0.00014249  0.00015385
0.00001709  0.01205479  0.01211916  0.00541985  28400.79

 END OF GEOMETRY OPTIMIZATION.

Is this usual......? How could we explain this ....?




On Fri, Aug 23, 2013 at 10:17 PM, Ramon Hernandez <ramon at uaem.mx> wrote:

> Dear Mahi,
>
> I see two problems with your calculations:
>
> a) The estimate of frequencies with central differences should be more
> reliable and
>  this calculation shows four negative hessian eigenvalues so doing an
> IRC at this
>  geometry is not correct. Look at the manual page for frequency
> calculations.
>
> b) The program is complaining about the gradient value it seems it is not
> small
>  enough to be considered converged for a stationary point.
>
> I think first you have to make sure that you have really found a
> transition state
> with only one imaginary frequency doing the calculation with central
> differences.
>
> Cheers
>
> Ramon
>
> 2013/8/23, G Mahesh <gmahesh at students.iiserpune.ac.in>:
> > Dear Ramon,
> >
> > I thought, the problem arising in IRC was because of negative Hessian
> > eigenvalues but, now I realized that's not the case. Earlier I have done
> > transition state optimization with Hessian calculated by forward
> > differences and IRC with central differences so that, it was giving 4
> > negative Hessian values in the case of IRC.
> >
> > Now I ran IRC with calculating Hessian using forward differences and  all
> > positive Hessian eigenvalues except one got too. Even now also
> surprisingly
> > giving the same error which I got earlier. Here, I also have posted the
> > echo of output having error.
> >
> > ***********Partial out put************
> >
> >  Geometry optimization using default procedure for command MULTI
> >
> >  Geometry written to block  1 of record 700
> >
> >  Generated   9 symmetrical displacements
> >
> >  Numerically approximating hessian using forward gradient differences
> >
> >  Task list generated. Total number of displacements:      9
> >
> >     5 tasks completed, CPU=1h 3m 15s Elapsed=1h 10m 13s
> >
> >  Numerical MULTI000 hessian completed. CPU-time:  6877.21 sec, Elapsed:
> > 7645.49 sec
> >
> >  MULTI hessian saved to record  5300.2
> >
> >  Quadratic Steepest Descent -  Reaction Path Following using exact
> Hessian
> >
> >  Hessian eigenvalues:    -0.012734  0.000786  0.011948  0.018747
>  0.075415
> > 0.141616  0.303108  1.646533  1.810536
> >
> >  IDIR= 1 requested but starting point is not a stationary point. Gradient
> > norm=  0.98D-04  Step length=  0.36D-01
> >  ? Error
> >  ? Not a stationary point
> >  ? The problem occurs in QSDPATH1
> > ***********
> >
> > What's wrong I am doing here....??
> >
> > Thanks,
> >
> > Mahi.
> >
> >
> >
> > On Wed, Aug 21, 2013 at 9:18 PM, Ramon Hernandez <ramon at uaem.mx> wrote:
> >
> >> Dear Mahi,
> >>
> >> I suggest you double-check that indeed your initial CASSCF calculation
> >> corresponds
> >> to a transition state, perhaps run a frequency calculation with tight
> >> parameters. Then
> >> when you run the IRC make sure you indeed converge to the same energy
> >> as the original
> >> CASSCF calculation. If you are sure about both these points then the
> >> problem might really
> >> have to do with a bug in IRC.
> >>
> >> Also if you run again the initial frequency calculation you can save
> >> that information and
> >> then read it directly in the next IRC job so you avoid calculating it
> >> twice.
> >>
> >> Cheers
> >>
> >> Ramon
> >>
> >> 2013/8/21, G Mahesh <gmahesh at students.iiserpune.ac.in>:
> >> > Dear all,
> >> >
> >> > I am familiar with MOLPRO from last two months. Since then I am trying
> >> > to
> >> > find a transition state for the reaction of reactants and products
> >> > having
> >> > multi reference character so, I am using CASSCF method for my problem.
> >> > After a long struggle, I have found the transition state with one
> >> imaginary
> >> > frequency (150 cm-1) and this normal mode having imaginary frequency
> >> seems
> >> > like reaction mode which will connect reactants to products.
> >> >
> >> > Now the problem is with IRC, here, I started IRC with the geometry
> >> > which
> >> I
> >> > got from the transition state optimization with QSDPATH method and
> >> > IDIR=1
> >> > then, after Hessian calculation, it's complaining that, the given
> >> structure
> >> > is not stationary with 4 negative Hessian eigenvalues. Does this
> means,
> >> my
> >> > initial structure is not real transition state structure.....??
>  Should
> >> > I
> >> > have to try for the transition state structure again ..?? Any ideas to
> >> > solve the problem....?? Pls help me...!
> >> >
> >> > Below additional details may be required to understand the problem.
> >> >
> >> > We are using the MOLPRO version 2012.
> >> > Our system has 5 atoms. We are using 13 active orbitals and 15 active
> >> > electrons as my cas space and AVTZ as basis.
> >> >
> >> > Thanks,
> >> > Mahi
> >> >
> >>
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20130827/432d294a/attachment.html>


More information about the Molpro-user mailing list