RE: force constants of diatomics in GAUSSIAN-03



A few loose ends here (now that my interest has been kindled).  To aid the
 discussion, let's call R the ratio of [the force constant with respect to
 Gaussian's mass-weighted definition of a unit step] to [the force constant with
 respect to interatomic distance].
 	fc Gaus	fc dist	R
 HF	10.66		9.70		1.10
 HCl	5.22		4.80		1.09
 HBr	4.19		4.10		1.02
 HI	3.23		3.20		1.01
 CO	41.34		18.60		2.22
 NO	39.29		15.30		2.57
 1. One thing that puzzled me was the difference between the H-X molecules for
 which Gaussian and the web site http://hyperphysics.phy-astr.gsu.edu/hbase/molecule/vibspe.html
 R is close to 1, and on the other hand NO and CO where R is much larger.
 Given your explanation [first attached message], it seems that the mass ratio of
 the two atoms is the important thing here, and that the ratio R between the
 mass-weighted comes between two limits.  The upper limit is the homonuclear case
 with R=2.  The lower limit occurs with an infinite mass ratio, in which case R=1
 because the normal-mode step is purely the movement of the light atom.
 2. But if that's true, why is R>2 for CO and NO?
 3. What definition does GAMESS(US) use for the unit step [second attached
 message]?  It doesn't agree with either Gaussian or the web site!
 --David Shobe, Ph.D., M.L.S.
 S|d-Chemie, Inc.
 phone (502) 634-7409
 fax (502) 634-7724
 Don't bother flaming me: I'm behind a firewall.
 -----Original Message-----
 From: Computational Chemistry List [mailto:chemistry-request$at$ccl.net]On
 Behalf Of Michael Frisch
 Sent: Wednesday, July 21, 2004 10:24 AM
 To: chemistry$at$ccl.net
 Subject: CCL:force constants of diatomics in GAUSSIAN-03
 On Wed, Jul 21, 2004 at 03:15:48PM +0400, Dmitry Rozmanov wrote:
 Dmitry Rozmanov wrote
 > If this is the case, then I guess this is just a wrong way of doing things
 > and the force constants got by Gaussian are not correct at all. There is a
 > definition of the thing and there is no two way of calculation.
 >
 > ---Dmitry.
 >
 This is nonsense.  The details are in a white paper on our web site,
 but the key point is that the force constant is the second derivative
 with respect to a normal mode displacement and the units for the
 normal mode, or equivalently the convention for what consititues a
 unit step, are arbitrary.
 For polyatomic molecules, one typically diagonalizes the force
 constant matrix in mass-weighted coordinates, so the natural unit step
 is a normalized displacement in these coordinates.  This approach is
 general and applicable to any polyatomic molecule.  In the particular
 case of H2 with the molecule along the x-axis, this normalized step
 would be (1/sqrt(2),0,0,1/sqrt(2),0,0) in the 6 cartesian coordinates.
 This unit step changes the H-H distance by sqrt(2).  For the particular
 case of diatomic molecules when people calculate by hand, they use the
 distance between atoms as the coordinate, which simpler for diatomics
 but doesn't apply to polyatomics.  In that coordinate system, a unit
 displacement changes the distance by 1 rather than sqrt(2), so the force
 constants (second derivatives of the energy) differ by a factor of 2.
 The corresponding reduced masses for the mode also differ by a factor
 of two and the frequency is the same.
 For the diatomic, the "by hand" coordinates give a reduced mass for
 the mode which is the same as the overall reduced mass for the
 molecule.  For a general polyatomic molecule, the reduced mass
 corresponding to a particular mode is not an observable quantity and
 is not defined until one adopts a convention for the (arbitrary) size
 of a unit normal mode displacement.
 Mike Frisch
 -= This is automatically added to each message by the mailing script =-
 To send e-mail to subscribers of CCL put the string CCL: on your Subject: line
 and send your message to:  CHEMISTRY$at$ccl.net
 Send your subscription/unsubscription requests to: CHEMISTRY-REQUEST$at$ccl.net
 HOME Page: http://www.ccl.net
 | Jobs Page: http://www.ccl.net/jobs
 If your mail is bouncing from CCL.NET domain send it to the maintainer:
 Jan Labanowski,  jlabanow$at$nd.edu (read about it on CCL Home Page)
 -+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 Earlier, Dmitry Rozmanov wrote:
 This is what Gaussian says:
   Harmonic frequencies (cm**-1), IR intensities (KM/Mole), Raman scattering
   activities (A**4/AMU), depolarization ratios for plane and unpolarized
   incident light, reduced masses (AMU), force constants (mDyne/A),
   and normal coordinates:
                       1
                      SG
   Frequencies --  2117.8243
   Red. masses --    14.8673
   Frc consts  --    39.2882
   IR Inten    --    86.4086
   Atom AN      X      Y      Z
     1   7     0.00   0.00   0.75
     2   8     0.00   0.00  -0.66
 This is what GAMESS (US) says:
 -------------------------------------------------------------------------------
           INTRINSIC VIBRATIONAL FREQUENCIES AND FORCE CONSTANTS
   FREQUENCIES AND FORCE CONSTANTS IN PARENTHESES SCALED BY 1.000 AND 1.000,
 RESP.
 
 -------------------------------------------------------------------------------
   INTERNAL COORDINATE        INTRINSIC FREQUENCIES
                                   (CM**-1)
   --------------------------------------------------
   STR.  2  1               2117.8       (2117.8    )
   --------------------------------------------------------------------
   INTERNAL COORDINATE                INTRINSIC FORCE CONSTANTS
                              (HARTREES/BOHR**2)        (MDYN/ANG)
   --------------------------------------------------------------------
   STR.  2  1                  1.2673  (  1.2673)     19.731  ( 19.731)
 You see, the frequencies are perfectly the same and the units are the same.
 However, I was pointed to a document which describes some pecularities of how
 Gaussian treats the reduced masses. I saw it before and I cannot still get how
 the same value with definite units can be different with two ways of calculation
 anyway.
 Here is the link:
 http://www.gaussian.com/g_whitepap/vib.htm#SECTION00036000000000000000
 If this is the case, then I guess this is just a wrong way of doing things and
 the force constants got by Gaussian are not correct at all. There is a
 definition of the thing and there is no two way of calculation.
 ---Dmitry.
 Per-Ola Norrby wrote:
 >     Hi Dmitry
 >
 >> I do no agree here. In fact I can do the same calculation with GAMESS
 >> (US) easily and get the meaningful result with the same method
 >> (ROHF/6-31G):
 >
 >
 >     OK, I think you have two problems.  First, if GAMESS and G98 differ
 > that much, it's probably because you located two different states.  Have
 > you compared the total energies of the molecules, and the orbital
 > output?  I had a look in Jaguar, using C2v symmetry (I have to stay with
 > Abelian subgroups).  In that notation, the 2a1 state is high in energy,
 > whereas the 2b1 and 2b2 states are degenerate.  Watch out for this type
 > of situation!  The final wavefunction is completely dependent on the
 > first guess, it's very possible that Gaussian and GAMESS locate
 > different states (you can modify it in Gaussian using the GUESS
 > keyword).  And you can be quite sure that the 2b1 and 2b2 states will
 > not be correctly represented by a simple ROHF.
 >
 >     I get my output as frequencies.  For the 2a1 state (the wrong one,
 > since it's high in energy), I get 943 cm-1, whereas 2b1 gives 2278
 > cm-1.  My guess is that GAMESS actually is the culprit here, locating
 > the high-energy state and thus giving a reasonable frequency for the
 > completely wrong reason.  You'd expect that an ROHF solution for a
 > problem that is inherently multideterminant would be expected to give a
 > too high force constant, like Gaussian is doing.
 >
 >     And I know the problems of applying MCSCF to "real"
 situations, I'm
 > currently working with CASPT2 on some metal complexes.  Sometimes you
 > have to, when you run into problems that not even B3LYP or CCSD(T) can
 > handle...
 >
 >     Regards,
 >
 >     Per-Ola
 -= This is automatically added to each message by the mailing script =-
 To send e-mail to subscribers of CCL put the string CCL: on your Subject: line
 and send your message to:  CHEMISTRY^at^ccl.net
 Send your subscription/unsubscription requests to: CHEMISTRY-REQUEST^at^ccl.net
 HOME Page: http://www.ccl.net
 | Jobs Page: http://www.ccl.net/jobs
 If your mail is bouncing from CCL.NET domain send it to the maintainer:
 Jan Labanowski,  jlabanow^at^nd.edu (read about it on CCL Home Page)
 -+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+