CCL:G: APT And Mulliken

 Sent to CCL by: Orlin Blajiev [blajiev,,vub.ac.be]
 Hi everybody,
 I have a system which I charge and discharge by adding or subtracting electrons.
 Then when I looked at the output I saw that the total number of electrons is
 correctly given by the Mulliken approach and was not so by the APT.
 For example two electrons out, then:
 Sum of Mulliken charges=   2.00000
 But
 Sum of APT charges=   2.14530
 Can someone tell me why APT charge is not an integer one?
 Thanks.
 Orlin
 >Sent to CCL by: frisch^_^gaussian.com (Michael Frisch)
 >On Wed, Jun 18, 2008 at 02:17:41PM -0400, Steve Williams
 willsd]|[appstate.edu wrote:
 >>
 >> Sent to CCL by: Steve Williams [willsd^_^appstate.edu]
 >> Raman intensities are available in G03, but since they require
 numerical
 >> derivatives for MP2 and DFT methods they are very slow to compute and
 by
 >> default they are not calculated on a frequency job.  If you change your
 >> input to include freq=raman on the route line, they will be computed.
 >> Six separate calculations will be done for each atom (+/- displacements
 >> in x, y, z) to compute the needed polarizability derivatives.  You can
 >> estimate how long this will take by doing a single point polarizability
 >> calculation then multiplying the time required for this by 6*N where N
 >> is the number of atoms in your molecule (the portion you want to
 compute
 >> with b3lyp).
 >>
 >
 >This is not correct.  In G03, MP2 and B3LYP Raman intensities (static
 >polarizability derivatives, "Freq=Raman" on the route card) are
 done
 >by numerically differentiating the analytic dipole derivatives with
 >respect to an applied electric field.  So there are only 6 additional
 >calculations regardless of the size of the molecule.  The cost of each
 >of the 6 calculations in a finite electric field is somewhat more than
 >a polarizability calculation but much less than the calculation of the
 >second derivatives for the basic frequency calculation.  So for very
 >small molecules the total cost of Freq=Raman with DFT or MP2 is around
 >than 3x the cost of just doing Freq, but this drops to 1.5x for larger
 >systems.
 >
 >Freq=Raman does work with ONIOM.  With ONIOM(MO:MM) and mechanical
 >embedding the cost is about the same as for the MO calculation on the
 >model system.  With ONIOM(MO:MM) and electronic embedding, the cost is
 >higher than just doing the model system (with or without Raman)
 >because derivatives with respect to all the nuclei have to be included
 >in the QM calculation.
 >
 >Mike Frisch>
 >
 >
 >
 Orlin Blajiev
 Department of Metallurgy, Electrochemistry
 and Materials Science
 Faculty of Applied Science
 Vrije Universiteit Brussel
 Pleinlaan 2, B-1050 Brussels
 Belgium
 http://www.vub.ac.be/META/
 tel.: 32-(0)2-6293538
 fax : 32-(0)2-6293200