Summary: G94/98: How to skip SCF completely?

 Dear CCL,
 this is the summary for my earlier query on how to use input ("guess")
  MOs to calculate properties like nmr in G94/98. The essential answer
 is, it doesn't work (cf. the comments by Mike Frisch -- and he should
 know ...).
    Quite a few people suggested the keywords "Guess=(Read,Only)". It
 is possible to calculate population stuff and related properties with this
 feature, but it is impossible to go into the CPHF equations (l1002).
    Further, I'd like to single out the response by Rene Fournier: when
 posting my question, I was hoping for something along the lines of his
 suggestion. In this respect, all I found in the code is the old link 503 that
 seems to do the direct minimazation that Rene suggested. However, it
 is, apparently, restricted to 256 basis functions (I have a lot more), and
 I don't really know how to use it anyways.
    Finally, I wish to pass on the comment by Marta ForÈs i Bocsh
 <marta.,at,.stark.udg.es>: She had posted -- and summarized -- a related
 question some weeks ago. You can search the CCL archives, or get her
 summary from me.
 Thanx to all who took the time to reply! And, please feel free to comment
 on this further.
 Best regards, Georg
 --
 Dr. Georg Schreckenbach           Tel:     (USA)-505-667 7605
 Theoretical Chemistry T-12        FAX:     (USA)-505-665 3909
 M.S. B268, Los Alamos National      E-mail:  schrecke.,at,.t12.lanl.gov
 Laboratory, Los Alamos, New Mexico, 87545, USA
 Internet:    http://www.t12.lanl.gov/~schrecke/
 ============================
 Here are the detailed answers:
 ============================
 Original question:
 Dear CCL,
 here is a GAUSSIAN related question: I would like to calculate some
 properties (like forces or NMR) based solely on the "input guess". How
 ~is this best done? (Or is it possible at all in GAUSSIAN?)
 More specifically, I imagine to have some input MOs using the
 "Guess=Read" command. I would specify a certain DFT method, say,
 and then go on to calculate the desired properties.
 What I have tried so far is the following.
 A) specify a "non-standard route" that resembles a standard route but
 contains no 500 links.
    Didn't work because the subsequent links requested some files that
 had not been written. Perhaps one can do this in a more clever way?
 B) Specify "SCF(MaxCycle=1)" as well as the appropriate flags to
 avoid the error termintation due to incomletely converged SCF. These
 flags are IOp(8/7=1,5/13=1).
    This procedure comes close to what I want, in particular if I use
 SCF=QC. However, the one SCF cycle is still too much because it does
 modify the input MOs as is clearly visible from the output. Hence, I would
 want to avoid it!
 Any comments or suggestions? -- I will summarize as usual.
 Thank you very much!
 Georg
 ============================
 1) response by Mike Frisch:
 From: frisch.,at,.lorentzian.com (Mike Frisch)
 You could probably hack the program to do this, but all of the properties
 in Gaussian are based on derivative expressions which are derived assuming
 that the appropriate SCF equations have been solved.  In order to use
 orbitals and or a density that comes from somewhere else, the CPHF/CPKS
 equations and possibly other terms would need to be modified to reflect
 how the density was determined and how it changes in response to the
 perturbations of interest.
 Mike
 ============================
 2) response by Rene Fournier:
 From: Rene Fournier <rene.,at,.mountain.chem.yorku.ca>
 Hello,
    In most programs the obvious way would be to use
 simple mixing convergence procedure with a mixing
 parameter equal to zero.  Can you do that in Gaussian?
  --------------------------------------------------------------------
  | Rene Fournier                  | Bureau/Office  303 Petrie       |
  | Chemistry Dpt, York University |      (416) 736 2100 Ext. 30687  |
  | 4700 Keele Street,  Toronto    | FAX: (416) 736-5936             |
  | Ontario, CANADA   M3J 1P3      | e-mail: renef.,at,.yorku.ca          |
  --------------------------------------------------------------------
 ============================
 3) One example of a response suggesting Guess=(Read,Only):
 (Quite a few people proposed the same thing.)
 From: Darko Babic <dbabic.,at,.rudjer.irb.hr>
         Dear George,
         This is what I've found in the G98 manual:
 "Guess(Only,Read):
 May also be used to produce population and other post-calculation analyses
 from the data in a checkpoint file. For example, these options alone will
 produce a population analysis using the wavefunction in the checkpoint file.
 Guess(Only,Read) Prop will cause electrostatic properties to be calculated
 using the wavefunction in the checkpoint file."
         I hope this helps.  If it doesn't, I am afraid that there is no simple
 way to achieve this.  Please, let me know if this happens.
                                         Sincerely,
                                                                 Darko Babic
                                                         Institute "Rudjer
 Boskovic"
                                                         HR-10001 Zagreb,
 P.O.B. 1016
                                                                   Croatia
 ============================
 4) Response by Neil Henson:
 From: Neil Henson <neil.henson.,at,.lanl.gov>
 What does it do with "SCF(MaxCycle=0)"?
 [Comment by G.S.: It goes to the default]
 How about setting the SCF convergence criteria to something ridiculous
 so that it is already converged before it starts?
 [Comments by G.S.: a) there is some maximum possible value for the
 convergence criteria, and it is still to low. b) it should do one cycle anyhow]
 Neil.
 --
 * Neil Henson                               Hartlepool Renaissance: a marina, *
 * neil.henson.,at,.lanl.gov/chimera.,at,.hubwest.com  a hung monkey and much
 much more  *
 * http://www.t12.lanl.gov/~njh           Tel: 505-667-7795
 Fax: 505-665-3909 *
 *  Group T-12, Mail Stop B268, Los Alamos National Laboratory, NM 87545, USA  *