The NF2 data were taken from the following publication:
H. S. P. Müller, K. Löblein, H. Hübner, W. Hüttner,
and J. M. Brown,
The Rotational Spectrum of the NF2 Free Radical:
Determination of Molecular Structure,
J. Mol. Spectrosc. 251 (2008) 185 – 193.
View abstract.
The submitted
manuscript has been
provided if one does not have access to the Journal of Molecular
Spectroscopy.
Note: NF2 is a radical
with three non-zero nuclear spins. Therefore, SPFIT
and SPCAT need seven quantum numbers to describe
each rotational level. Older versions of the programs use aggregate spin
numbers to cope with the problem that the maximum number of quantum
numbers for each state was six, see the
old version of the line file
as well as the
old version of the fit file.
As one can see, only five quantum numbers were used this way for each state:
three asymmetric rotor quanta, the aggregate spin number, and F, the
final spin quantum number.
It is possible to use up to 10 quantum numbers for each state in recent
versions of the programs. The is done by putting a minus sign in front of
the number of lines, see the
final parameter file.
See also the
final line file
and the
final fit file.
The article mentioned above provides some information on the choice of
the spectroscopic parameters which have been determined in the final fit.
In contrast to the earlier papers, the S reduction was used in
the present work because the fits with the S reduction were
generally slightly better than those with the A reduction.
In this context it is worthwhile to point out an article on
SiF2. The authors found out for several three-atomic
molecules with C2v symmetry that the vibrational
corrections to the centrifugal distortion parameters are smaller with
an S-reduced parameter set than with an A-reduced one:
L. Bizzocchi, C. Degli Esposti,
Millimeter and submillimeter-wave spectroscopy of silicon
difluoride,
J. Mol. Spectrosc. 235 (2005) 117 – 124.
View abstract.
Thus, it appears to be useful to employ the S reduction for such
molecules even if the fits are not better than those with the A
reduction.
In the case of a very complex spectrum such as that of NF2 it is important to try to keep the number of parameters reasonable. Usually, this can be achieved by including only parameters in the fit which are determined with significance (uncertainties at least a factor of five smaller than the values) and which reduce the rms error of the fit considerably. The choice of low oder parameters (number 1–8, 15–17, and 21–28) is straightforward if the line list is not too short anymore. In later stages of the fits it may be worthwhile to check if replacing one higher order parameter by another one improves the fit. In the final fits it may be useful to see if certain parameters are reasonably well determined even if their inclusion in the fit do not reduce the rms error of the fit considerably. Some parameters may well have effects only in few transitions or very small contributions throughout. In the case of NF2, parameters 32 and 33 were determined reasonably well. In addition, compared with parameter 31, they had appropriate values. Therefore, they were maintained in the final fit even though they contributed only minutely to the reduction of the rms error.