CCL:G: G09: Convergence Issues for H-Bonded Systems



 Sent to CCL by: Soren Eustis [soreneustis\a/gmail.com]
 Sam,
     It looks like you are taking some good steps to solve this issue.
 Firstly, the  -1.9 wavenumber frequency should not be taken as an
 indication of a saddle point.   The energy is simply too low.  Look at the
 whitepapers on the Gaussian site if in doubt.
    Your second question would then be negated by my answer to your first
 question, as it seems you have a valid converged geometry (opt-calcall,
 maxstep=5).  If you were to continue to try the second method with
 opt=tight, you should use int=ultrafine.  This is always a good idea for
 shallow potential wells and is recommended anytime opt=tight is used.
 Best,
 Soren
 Soren N. Eustis, Ph.D.
 ETH ­ Zürich
 Environmental Chemistry Group
 Institute for Biogeochemistry and Pollutant Dynamics
 Universitätstrasse 16
 CHN F33
 8092 ZürichSWITZERLAND
 +41 44 632 93 48 (office)
 +41 44 632 14 38 (fax)
 soren a env.ethz.ch
 On 9/22/11 17:25 , "Sam Abrash sabrash]_[richmond.edu"
 <owner-chemistry a ccl.net> wrote:
 >
 >Sent to CCL by: "Sam   Abrash" [sabrash_+_richmond.edu]
 >Hi Folks,
 >
 >I have a job involving hydrogen bonding of 3 acetylenes to the benzene
 >cation.  I'm having trouble with the geometric convergence.  I was able
 >to get the system to converge with normal convergence criteria,
 >opt=calcall, and MaxStep=5, but then I got an imaginary frequency of -1.9
 >wavenumbers.
 >
 >Subsequently, I changed to Opt=tight, tried both calcfc and calcall, and
 >have steadily reduced MaxStep to 1, but it still won't converge.  The
 >force constant criteria have been met, but the problem is the
 >displacement criteria.  Both this observation and looking at the
 >structures with JMOL show that the problem is finding the minimum in a
 >very shallow potential.
 >
 >Two questions.  First, is the 1.9 wavenumber imaginary frequency real or
 >an artifact?  Second, any advice on how to force the system to converge?
 >Model chemistry is M062X/6-311++G**.
 >
 >Thanks!
 >Sam>
 >