1. Conditions of the CH3CN laboratory measurements

The CH3CN pseudo-linelist was created from 29 laboratory spectra
taken at the Pacific Northwest National Laboratory (PNNL) and provided
by Steven Sharpe. It should be noted that these are not exactly the same as the
CH3CN cross-sections on the HITRAN website (http://cfa-www.harvard.edu/HITRAN/),
which had been modified by converting all negative values to zero.
The measurements and the absorption cross sections (incl. assignments
of major bands) are described by Rinsland et al. (2005).

The measurement conditions for each of these spectra are tabulated in Tab. 1.
Each measurement used the same cell of 8.1576 m length. Each spectrum
covers a region between 600 and 6500 cm-1 with a resolution of 0.1125 cm-1
and a spectral point spacing of 0.0603 cm-1. 


2. Derivation of the CH3CN pseudo-linelist

2.1. Spectral fitting

For the spectral fitting of the laboratory measurements the absorption
cross-sections were converted back into transmittance
spectra from knowledge of the cell length and gas concentrations.
The resulting laboratory transmittance spectra were then
fitted using the same nonlinear least squares algorithm (GFIT)
that was used to analyze the atmospheric spectra. The
linelist was created by iteratively adjusting the strengths and
ground-state energies of the pseudo-lines according to the fits.

Due to the resolution of the laboratory spectra of 0.1125 cm-1 and the
spectral point spacing of 0.06 cm-1 a pseudo-line spacing
of 0.05 cm-1 was considered to be appropriate. Fitting was
performed in the frequency regions from 800-970 cm-1 (where
the nu_4 band is located), from 960-1170 cm-1 (where the nu_7 band
is located), and from 1160-1700 cm-1 (for the nu_3, nu_6, and nu_7+nu_8 bands).
These regions include the two bands with the strongest infrared absorption
features. A zero level offset of 0.2% has been assumed throughout,
based on fits to spectra in which the absorption feature at 1463 cm-1
was saturated. As the linestrengths below 870 cm-1 and above 1650 cm-1
were negligibly small, the resulting pseudo-linelist has been truncated
to a range between 870 and 1650 cm-1, thus containing 15601 lines.


2.2. Calculation of line strength, ground-state energy, and width

At each line frequency, an effective strength and ground-state energy (E")
was derived by simultaneous non-linear least squares fitting to the 29
spectral residuals. There is insufficient information in the PNNL spectra
to also determine an air broadened half width (ABHW). So the ABHW was calculated
from the ground-state energy using the formula

ABHW = 0.04 * (E" + 2000)/(E" + 1000)

(where E" is given in cm-1), giving an ABHW of 0.08 cm-1/atm for E"->0
and an ABHW of 0.04 cm-1/atm for E"->oo. This formulation seemed to be the
most appropriate to fit the feature at 1042.3 cm-1, which is the narrowest
feature in the considered frequency region. These widths are smaller
than those measured by Drouin (2003) in the microwave region, but we found that
using larger widths produced significantly poorer fits to the sharp spectral
features. For the the self broadened half width a value of 0.9 cm-1/atm
was assumed. As part of the fitting, the strengths and ground-state energies
were both constrained to be positive.


2.3 Partition function

The rotational partition function for CH3CN was assumed to be (296/T)^1.5.
The vibrational partition function was calculated in the way it had been
done for the ATMOS experiment, as described e. g. by Norton and Rinsland (1991).
The assumed vibrational frequencies and degeneracies are given in Tab. 2.


3. Accuracy of the CH3CN pseudo-linelist

To estimate how well the pseudo-linelist represents the PNNL spectra,
test retrievals were performed in which the laboratory spectra were fitted
using the pseudo-linelist. Two tests were performed: one in which the
full frequency range covered by the linelist was used, and one in which
only the narrow window (1462.0-1464.4 cm-1) chosen for the atmospheric
analyses was used.

The top plot in Fig. 3 shows the fit to the laboratory spectrum #21,
which is the one with the lowest absorption. It can be seen that the
pseudo-linelist fits the spectrum essentially down to the noise level.
The bottom plot in Fig. 3 shows the fit to the spectrum #9, the one
with the highest absorption. The residual of the fit to the band
around 900 cm-1 also reaches down to the noise level. In the band
around 1450 cm-1 some residuals in close proximity to strong lines remain
but they are typically below 1%. In the band around 1050 cm-1 there
are stronger residuals, indicating the assumed linewidths do not
represent the true linewidhts perfectly. However, the residuals
stay below ~2.5%. Fig. 4 shows the fits to these two spectra in
a narrow region around the strong feature at 1463.3 cm-1, which is used
for the atmospheric retrievals. The residuals stay below 0.25% in both
cases.

The retrieved scale factors for the CH3CN abundances in the different
spectra are given in Tab. 3. The pseudolines correctly represent the
PNNL spectra to within about 0.7% of the given CH3CN amount in both cases.


4. References

Drouin, B. J., Temperature dependent rotational transition lineshape
parameters for O3, O2, SO2, CH3CN and CO, International Symposium on
Molecular Spectroscopy, Columbus, OH, USA, June 16-20, 2003.

Norton, R. H. and C. P. Rinsland, ATMOS data processing and science
analysis methods, Appl. Opt., 30, 389-400, 1991. 

Rinsland, C. P., S. W. Sharpe, and R. L. Sams, Temperature-dependent
infrared absorption cross-sections of methyl cyanide,
J. Quant. Spectrosc. Radiat. Transfer, 96, 271-280, 2005.