CCL:G: Projected frequencies

 Sent to CCL by: "luis simon" []
 Few days ago I posted a similar question, so maybe it is not such an unusual
 question (please take a look at March 13th messages). I got some interesting
 answers. The most promising in gaussian was to include "-1" on frozen
 when the molecule is specified by the cartesian coordinates. Although it
 worked, gaussian calculated the full hessian matrix, so it still requires a
 lot of time. Nevertheless, I guess taht the results would be probably valid,
 as this hessian is calculated on a real stationary point on the subspace of
 the not frozen coordinates. I think that you would make a similar work if
 you write a script to read the hessian matrix (in cartesian coordinates) and
 then set to 0.0 all those elements concerning derivatives of the position of
 the frozen atoms, but using the "-1" trick allows you to obtain a .chk
 that could be helpful for further thermochemistry and vibrational analysis
 (you can also write the algorithms to diagonalize the hessian, but there is
 no need if you can use those implemented in gaussian).
 Nevertheless, the ideal solution would be that the program really skip the
 calculation of those derivatives.  Maybe it is not possible with gaussian if
 it uses internal coordinates to build the hessian, but I have observed that
 even if you specify gaussian to work with numerical derivatives, it
 calculates second derivatives concerning "-1" tagged atoms.
 I hope that you receive more useful answer as it will probably be useful to
 many other people.
 ----- Original Message -----
 > From: "Ramon Crehuet rcsqtc[a]"
 To: "Simón, Luis Manuel "
 Sent: Thursday, March 16, 2006 2:42 PM
 Subject: CCL:G: Projected frequencies
 > Sent to CCL by: "Ramon  Crehuet" [rcsqtc()]
 > Dear CCLers,
 > I have a relatively large system where I keep some (external) atoms fixed.
 After doing an optimisation I want to check that I have a minimum or a TS
 candidate. So I would like to do a frequency calculation on the free atoms.
 I think that makes sense from a theoretical point of view. So I do not want
 to calculate the second derivative with respect to fixed atoms coordinates,
 although they have to be there for the other derivatives to make sense.
 Otherwise, I could get the full hessian and project it.
 >   I was using gaussian and don't know how to do it with this software. Is
 is possible? Is is possible with GAMESS-US (which I use regularly)? Is it
 possible with any other software (that can also do DFT and ONIOM)?
 > And last, but not least... am I asking something very unusual?
 > Thanks in advance,
 > Ramon>