Dear Raphael,
In complement to Jean-Jules's answer I'd add
this:
calcfc is an option that calculate the full
hessian matrix (second derivatives of the energy with respect to nuclear
coordinates) for your system. As previously mentioned, It can be very
cpu-time/memory consuming.
Here are some answers to your
questions:
1- No.
You only rarely need to start by calculating force constants, at least when
optimizing a system toward a
minimum of the PES.
By default, Gaussian estimates the force
constants (eq. to opt=newestmfc) that will be used in the minimization
algorithm at the beginning (with a valence force field, which is not time
consuming). Furthermore, along the
optimization, the hessian matrix is updated. 2nd derivatives are used in
methods such as ones based on Newton-Raphson approach, e.g.
2- Same answer as above but with a precision:
for a minimum search, things happen generally well, in most of the cases, calcfc
is not needed.
When you are looking for a transition state,
things are different: you have to provide a proper hessian matrix (see Gaussian
User's manual for TS search, which can be a tricky task, sometimes), which means
that either you
provide it from a previous calculation or calculate it with calcfc.
3- Almost a linear combination of the two
previous answers:
When looking for a minimum, the use of calcfc
is rarely (even almost never) needed. But when looking for a TS, this is needed
if you don't provide the hessian from a previous calculation.
To finish, I absolutely want to
"modulate" my answers: every chemical system is particular and will
behave a particular way. Hence, the calculation of the hessian matrix before a
minimum search could be needed, as already
written by Jean-Jules.
I hope this clarified things.
Anyway,I suggest you to have a look at
Gaussian09 User's reference, especially:
http://www.gaussian.com/g_tech/g_ur/k_opt.htm which describes very
precisely the
use of different keywords related to optimization methods. I would also
suggest some further readings about optimization techniques and algorithms like
in chapter 12 (p380, 2nd edition) of F. Jensen's "Introduction to
Computational Chemistry" or section 4
of chapter 2 (pp 40-46) in C. Cramer's Book: "Essentials of Computational
Chemistry".
Best Regards,
Yohann
Yohann Moreau
Maître de Conférence, Université Joseph Fourier
iRTSV/CBM/MCT, CEA Grenoble
17 avenue des Martyrs
38 054 Grenoble Cedex 09
Tel. : (33) 4 38 78 29 62
Fax : (33) 4 38 78 54 87