Open Shell CASSCF calculations

[Mark Watkins]

A query regarding open shell CASSCF frequency calculations.  This uses the
32-bit version of Molpro (all latest patches applied as of end of March, 2001)
on an IBM RS/6000 (AIX 4.3.3, 4x 375 MHz Power3 procesors, 4 Gb RAM).

We have a problem whereby Molpro always stops without error message during when
numerically calculating CASSCF frequencies - closed shell calculations are fine
for the same molecule.

An example of the input file as follows:

**************************************************
  memory,260,m
 FILE,1,/scratch/molpro/resII.int
 FILE,2,/scratch/molpro/resII.wfu
 geomtyp=xyz
 geometry={                                           
 14
 resII
 C,1.2177622432,-0.2522161661,0.
 C,1.1029785752,1.1968768641,0.
 C,-0.1527390135,1.7937888553,0.
 C,-1.2894437174,0.9931756374,0.
 C,-1.1575699001,-0.4544696847,0.
 C,0.0903139164,-1.0607793504,0.
 H,-0.2442627027,2.8691514621,0.
 H,-2.278091321,1.4320187259,0.
 O,-2.2074362136,-1.2141569258,0.
 H,0.1818140165,-2.1356548054,0.
 H,2.00318307,1.7965780656,0.
 O,2.3809178453,-0.8234579085,0.
 H,3.1141007533,-0.2180354654,0.
 H,-3.0324094927,-0.7413962419,0.
 }

 basis=cc-pvdz
 rhf
 wf,57,2,1
 mcscf;occ,24,8;closed,24;wf,57,2,1
 optg
 put,molden,resII-i-opt.molden
 put,gaussian,resII-i-opt.gauss
 freqencies
 put,molden,resII-i-freq.molden

**********************************************************

The above calculation, and other similar calculations were always found to
stop, with the last few entries in the .out file going something along the
lines of the following:

Incrementing variable OZ12 by   .010 BOHR  to     .0100 BOHR
Incrementing variable HX13 by   .010 BOHR  to   -5.7059 BOHR
Incrementing variable HY13 by   .010 BOHR  to   -1.3893 BOHR
Incrementing variable HZ13 by   .010 BOHR  to     .0100 BOHR
Incrementing variable HX14 by   .010 BOHR  to    4.7763 BOHR
Incrementing variable HY14 by   .010 BOHR  to   -3.0731 BOHR
Incrementing variable HZ14 by   .010 BOHR  to     .0100 BOHR

Total number of displacements:    71

Incrementing variable CX1 by   .010 BOHR  to   -2.1814 BOHR




at which point all calculations and molpro processes cease.  We have calculated
UHf frequencies without problems.

Our questions are:

1)  Has anyone had a similar experience with doublet state CASSCF frequency
calculations, and if so, how were these problems overcome?

2) Is the above some problem with Molpro itself, or an aspect of the IBM
RS/6000 - we have tested the frequency calculations on several RS/6000's, all
with the same fault.  Could this also relate to a problem with the Fortran
compilers (though we have tested installations with both versions 6 and 7 of XL
Fortran)?

3) Is there something wrong with the above input file?

Any info. gretly appreciated!

Susanne Ullrich - su102 at york.ac.uk
Mark Watkins - mjw103 at york.ac.uk