[molpro-user] Starting point is not a critical point

[Gabriel Freitas]

Dear molpro users/developers,

I'm trying to do an IRC calculation from a transition state previously
calculated. But by the time the qsdpath takes its first step, I get the
message:

*QSDPATH2: Starting point is not a critical point.*

I've tried many things, but unsuccessfully, and I'm afraid there can be a
bug in the code.

Here follows one example, in which I reoptmize the TS (which converges in a
single step, since the starting geometry is already converged) and ask for
IRC using the hessian calculated during the TS search.


Part of the input

XXXXXXXXXXXX
 nosym
 R1=1.34210982 ang,
 R2=1.76235720 ang,
 R3=1.30142476 ang,
 a=92.77261716 degree,
 b=113.97606866 degree,
 g=0.00000 degree
 geometry={ }
 rhf,maxdis=50,maxit=100;
 rccsd(t),maxit=70;
 {optg,root=2,method=qsd,maxit=100,saveact=...,rewind;print,history}
 {optg,method=qsdpath,hessrec=5300.2,dir=+3,numhess=0,hesscentral,maxit=100;print,history}

XXXXXXXXXXXXXXX

*Part of the output*

XXXXXXXXX

 PROGRAM * OPT (Geometry optimization)     Authors: F. Eckert and H.-J.
Werner

 Geometry optimization using default procedure for command RCCSD(T)

 Numerically approximating hessian using central energy differences

 Task list generated. Total number of displacements:     42

 ....

 Numerical RCCSD(T) hessian completed. CPU-time: 31364.93 sec, Elapsed:
33653.29 sec

 RCCSD(T) hessian saved to record  5300.2

 Combined Powell-Murtagh-Sargent Update of Hessian

 Quadratic Steepest Descent -  Transition State Search

 Optimization point 1

 Variable                      Last           Current        Next
Gradient       Hessian

 E(RCCSD(T)) / Hartree       0.00000000  -548.43957924     0.00000000
 R1 / ANG                    0.00000000     1.34210982     1.34215444
 -0.00004976     0.97545936
 R2 / ANG                    0.00000000     1.76235720     1.76307008
 -0.00025977     0.46872502
 A / DEGREE                  0.00000000    92.77261716    92.75401930
0.00000022     0.00006113
 R3 / ANG                    0.00000000     1.30142476     1.30118943
0.00017736     1.53165483
 B / DEGREE                  0.00000000   113.97606866   113.97094779
0.00000020     0.00009433
 G / DEGREE                  0.00000000     0.00000000    -0.06878260
0.00000000     0.00000090
 Convergence:                0.00000000  (line search)     0.00189055
0.00016936  (total)

 END OF GEOMETRY OPTIMIZATION.    TOTAL CPU:     31240.2 SEC


 PROGRAM * OPT (Geometry optimization)     Authors: F. Eckert and H.-J.
Werner


 Geometry optimization using default procedure for command RCCSD(T)


 Number of displacements for numerical gradient:     12

 Starting numerical gradient for RCCSD(T)

 Numerical gradient completed. CPU-time:  8776.29 sec, Elapsed:  9365.30 sec

 RCCSD(T) hessian read from record  5300.2

 Combined Powell-Murtagh-Sargent Update of Hessian

 Quadratic Steepest Descent -  Reaction Path Following using updated Hessian

 Hessian eigenvalues:     0.002870  0.102644  0.198864  0.270747  0.292713
 0.478791

 QSDPATH2: Starting point is not a critical point.
 Performing a regular QSD step. Stepsize =  0.075068    Curvature =
.49831D+04

 Optimization point 1

 Variable                      Last           Current        Next
Gradient       Hessian

 E(RCCSD(T)) / Hartree       0.00000000  -548.43957912     0.00000000
 R1 / ANG                    0.00000000     1.34215444     1.34215192
 -0.00000415     0.97545936
 R2 / ANG                    0.00000000     1.76307008     1.76311166
 -0.00002081     0.46872502
 A / DEGREE                  0.00000000    92.75401930    92.75259887
0.00000000     0.00006113
 R3 / ANG                    0.00000000     1.30118943     1.30117610
0.00000355     1.53165483
 B / DEGREE                  0.00000000   113.97094779   113.97050327
 -0.00000001     0.00009433
 G / DEGREE                  0.00000000    -0.06878260    -0.14458311
0.00000007     0.00000090
 Convergence:                1.00000000  (line search)     0.00132580
0.00001209  (total)

XXXXX

Although the hessian is calculated for a geometry slightly displaced from
the one used for the gradient calculation, it can be seen that both
gradients are smaller than the default converge threshold,

Any help would be very appreciated.

Regards,

-- 
Gabriel do Nascimento Freitas
D.Sc. Student - Graduate Program of Chemistry
Molecular Modelling and Theoretical Chemistry Laboratory - Room 412
Chemistry Institute - Universidade Federal do Rio de Janeiro (UFRJ) -
Brazil
(+5521)8830-9971 / (+5521)2562-7179
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20120108/4a3e2271/attachment.html>