From chemistry-request@server.ccl.net  Sat May 22 15:29:54 1999
Received: from neon.chem.ucla.edu (neon.chem.ucla.edu [128.97.35.251])
	by server.ccl.net (8.8.7/8.8.7) with ESMTP id PAA19350
	for <chemistry@server.ccl.net>; Sat, 22 May 1999 15:29:53 -0400
Received: from localhost (kuwata@localhost) by neon.chem.ucla.edu (8.8.0/8.8.0) with SMTP id MAA16922 for <chemistry@server.ccl.net>; Sat, 22 May 1999 12:30:51 -0700 (PDT)
X-Authentication-Warning: neon.chem.ucla.edu: kuwata owned process doing -bs
Date: Sat, 22 May 1999 12:30:51 -0700 (PDT)
From: Keith Kuwata <kuwata@chem.ucla.edu>
To: chemistry@server.ccl.net
Subject: CASMP2(10,10) Problems
Message-ID: <Pine.GSO.3.95.990522115833.16061B-100000@neon.chem.ucla.edu>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

Hi CCLers,

I have converged CASSCF(10,10)/6-31G(d,p) wave functions (95 basis
function total).  I have been unable to obtain single-point
MP2 corrections (using the keyword MP2=full) 
with either Gaussian 94 or Gaussian 98.

G94 has failed on both an SP2 cluster and the NCSA HP Exemplar
If I use the keyword iop(4/43=1), the job crashes before the CAS begins
with the error message "OPERATION ON FILE OUT OF RANGE"
If I omit this keyword, the job crashes with a "Segmentation
Fault" right after the route information prints.  This happens
even when 4 GB of RAM are allotted.

With G98 on the Exeplar, CASSCF(10,10) runs successfully with the 
battery of additional keywords posted by Wolfgang Roth last month:
  casscf(10,10,nofulldiag) iop(5/6=8) iop(4/46=1,5/39=000001)
  iop(5/16=1)

Upon adding the keyword mp2=full, G98 at least doesn't crash
(unlike G94).  However, in the MP2 routine, all the configurations
are skipped and there is zero MP2 correction to the CASSCF energy.

I know that CASPT2 is arguably better, and I am pursuing those
calculations as well.  But it would be very useful to have CASMP2
(10,10) energies for comparison.  If anyone knows how to get this
to run on either G94 or G98, that'd be great.

Much thanks,

Keith Kuwata

P.S. [Needless to say], Gaussian has failed to respond to my three
requests for help.

_________________________________________________________________

Keith T. Kuwata
kuwata@chem.ucla.edu