[molpro-user] rs2 frequency calculcation complains about large gradient norm

[Benj FitzPatrick]

Hello,
I have tried ss-sr-caspt2 and ms-mr-caspt2 geometry optimizations of transition
states followed by frequency calculations.  I changed the tolerances for the
optimizations so that the step=1.d-7 and gradient=1.d-7.  The optimization
completes fine, but when the frequency calculation finishes it complains of a
non-zero gradient norm.  This has me baffled because the gradient norm is much
larger than that calculated in the final step of the optimization.  Is there a
reason it changes so much?  Also, how much should I worry about this?
Thanks,
Benj FitzPatrick
University of Chicago

---geometry optimization summary
 END OF GEOMETRY OPTIMIZATION.    TOTAL CPU:     13977.1 SEC

 ITER.   ENERGY(OLD)    ENERGY(NEW)      DE          GRADMAX     GRADNORM   
GRADRMS     STEPMAX     STEPLEN     STEPRMS
   1  -191.92739935  -191.92901706    -0.00161772  0.03179216  0.04014669 
0.00191175  0.16286050  0.29999982  0.01428571
   2  -191.92901706  -191.92910986    -0.00009280  0.00328012  0.00411807 
0.00019610  0.14362552  0.20840683  0.00992413
   2  -191.92901706  -191.92910986    -0.00009280  0.00328012  0.00411807 
0.00019610  0.14362552  0.20840683  0.00992413
   3  -191.92910986  -191.92911807    -0.00000820  0.00195023  0.00322124 
0.00015339  0.07660351  0.13617987  0.00648476
   4  -191.92911807  -191.92911369     0.00000438  0.00030322  0.00060006 
0.00002857  0.16490991  0.28181425  0.01341973
   4  -191.92911807  -191.92911369     0.00000438  0.00030322  0.00060006 
0.00002857  0.16490991  0.28181425  0.01341973
   5  -191.92911369  -191.92912132    -0.00000762  0.00067278  0.00103621 
0.00004934  0.11735469  0.20131804  0.00958657
   6  -191.92912132  -191.92912153    -0.00000022  0.00008001  0.00018761 
0.00000893  0.00657972  0.01200790  0.00057180
   6  -191.92912132  -191.92912153    -0.00000022  0.00008001  0.00018761 
0.00000893  0.00657972  0.01200790  0.00057180
   7  -191.92912153  -191.92912154    -0.00000001  0.00001077  0.00002055 
0.00000098  0.00463685  0.00794725  0.00037844
   8  -191.92912154  -191.92912154     0.00000000  0.00000169  0.00000323 
0.00000015  0.00069815  0.00119906  0.00005710
   8  -191.92912154  -191.92912154     0.00000000  0.00000169  0.00000323 
0.00000015  0.00069815  0.00119906  0.00005710
   9  -191.92912154  -191.92912154     0.00000000  0.00000002  0.00000005 
0.00000000  0.00000260  0.00000455  0.00000022


---start of frequency calculation summary

Frequencies dumped to record   5401.2

 Gradient norm at reference geometry: 0.35112E-02

 ?WARNING: GRADIENT NORM NOT ZERO!

 Normal Modes