Summary of Gaussian quirks

From: young ^at^ slater.cem.msu.edu (Dave Young)
Subject: Summary of Gaussian quirks
Date: Tue, 1 Feb 94 08:45:07 -0500
 Hi,
 	Here is the summary of messages received about Gaussian quirks.
 The original posing was
 *******************************************************************
 >Hi,
 >
 >	I'm sure many of the subscribers to this list are fluent users
 >of the Gaussian XX programs.  Although these are wonderful programs,
 >the number of revsions and authors has created a number of places
 >where the manual is either wrong or too vague.
 >
 >	I don't mean to offend the authors.  I couldn't keep on top
 >of that volume of code either.
 >
 >	However, my question is this does anyone have a compilation
 >of the various work arounds for these quirks?  Following are two
 >that I have found.
 >
 >1.  In order to request a tighter convergence criteria, the command
 >"conv=7" for 10**-7 convergence doesn't work in our version, but
 the
 >command "iop(5/6=7)" will have the same result.
 >
 >2.  When doing a CISD calculation, the command "pop=no" does not
 give
 >a complete set of natural orbitals with occupation numbers.  (By complete
 >set, I desire all orbitals with non-zero occupations.)  However, if
 >you are creating a cube file from a CI calculation and have the
 >commands "pop=no density=ci cubedensity=density" THEN you get the
 >natural orbitals.
 >
 >	Please send comments directly to me young ^at^ slater.cem.msu.edu
 >and I will summarize for the list.
 Replys are
 **********************************************************************
 From: rossi ^at^ t1.chem.umn.edu (Ivan Rossi)
 Dear Dave,
 	I don`t know what version of gaussian you have, but on my g92 rev b
 this command work both with scf and casscf :
 	scf=(conv=7)
 that is equivalent to the iop(5/6=7) you reported.
 I don`t use ci so i can't be helpful for the other case
 	good luck
 	Ivan
 Ivan Rossi, Ph.D.                         |
 Department of Chemistry                   |   EXPERT : (n)
 University of Minnesota                   |   someone who avoids all the
 207 Pleasant St. SE,Minneapolis, MN 55455 |   little errors, going straight
 Tel. 612-624-6099                         |   towards the catastrophe
 e-mail : rossi ^at^ t1.chem.umn.edu            |
 ********************************************************************
 From: rbw ^at^ msc.edu (Richard Walsh)
 Dave,
 The one that I would like to point out
 is the description of the assignment
 of additional basis functions to atoms using the MASSAGE
 key word. The manual indicates that one should use the
 'Center number' on page 78. The output uses the center
 number inconsistently and if you have a system with dummy
 atoms watch out because the output marks them as 'centers'.
 If you do not get things exactly right you may end up
 with d-orbitals on hydrogen atoms! Anyway the best way to
 avoid this problem is to count real atoms only and use
 the GFPRINT key word to confirm that you got what you
 wanted. I reported this to Doug Fox and he seemed to
 take an interest, but a fix was not in the most recent
 release of the manual. I think that to some extent they
 are relying on the availability of the new book to
 compensate for what is lacking in the manuals.
 Richard Walsh
 Minnesota Supercomputer Center
 Chemistry Applications
 ****************************************************************
 From: FOX ^at^ cmchem.chem.cmu.edu
   Dr. Young,
   Your first quirk is exactly as you describe, additional options were
 added which make CONV no longer unique.  You need to use CONVER for
 post-HF and SCF=CONVER=n for SCF convergence.
   Your second quirk is one I don't recognize.  I tried to duplicate it
 here and did not find a difficulty.  Can you send an input which shows
 the difficulty to
 help ^at^ gaussian.com
    In general we stand ready to address questions/problems you have with
 Gaussian at that address.
 Doug Fox
 Director of Technical Support
 *********************************************************************
 From: FOX ^at^ cmchem.chem.cmu.edu
    Dr. Young,
    For the largest part we have been able to correct the code to back up
 the manual so most of the quirks are taken care of.  One feature which is
 commonly missed is that you can only specify a keyword once on the route
 card, i.e.
 # MP2/6-31G* MP2=Stingy
 will produce an MP4 calculation instead of MP2.
    On the density question I now understand the confusion.  The CI is not
 formed by default for a CISD energy calculation.  This requires an additional
 step to solve coupled perturbed equations and applies to any of the post-HF
 methods which can produce a gradient, MP2, QCISD, CIS etc.  This step can
 be fairly costly if all you require is the energy so the default is to analyze
 the HF density unless you include the DENSITY keyword.  Thus the NO analysis
 in you first case is of the HF density and the NO's have the all or nothing
 population you observed.  In the second case you have truly analyzed the CI
 density.
   My general recommendation is to use DENSITY or DENSITY=CURRENT which selects
 the DENSITY of your energy keyword, if it is available, rather than using
 DENSITY=CI which would fail if you did an MP2 for example.  You could also
 consider adding DENSITY=CURRENT into the Default.Route file on your system
 if you always want to produce the density.
    Doug Fox
    help ^at^ gaussian.com
 				Dave Young
 				young ^at^ slater.cem.msu.edu
 				youngdc ^at^ msucem