Observations were made using the LWS in grating mode (L01, L02, 43-197
µm, /
~ 200). The LWS consists
of ten detectors with spectral overlap for adjacent detectors. In the
grating mode of the LWS, the spectral resolution is about 0.29
µm for the 10 µm-wide short-wavelength detectors
(SW1-SW5) and 0.60 µm in the 20 µm-wide
long-wavelength detectors (LW1-LW5). The L01 Astronomical Observation
Template (AOT) is a range scan of the grating that results in 10 spectra
covering a significant range of the LWS. The L02 AOT produces spectra
for up to ten wavelengths, specified by the observer. In this mode, data
are recorded for all ten detectors while the specified wavelengths are
being scanned, producing spectra with significant gaps across the range
of the LWS.
All guaranteed and open time observations for 227 galaxies were extracted from the ISO Data Archive and processed through the LWS Pipeline Version 7.0 or 8.7. Slight improvements in the photometric model are made beyond Pipeline Version 7.0, but these changes have minor effects on the calibration of the L01 and L02 grating mode AOTs. These changes yield improvements in the flux accuracies by a few percent but do not significantly alter the line and continuum fluxes that are derived from Pipeline 7.0.
Further data manipulation is then carried out using the LWS Interactive
Analysis (LIA;
Hutchinson et al. 2001)
and the ISO Spectral Analysis Package (ISAP;
Sturm et al. 1998).
The continuum fluxes in the
LWS spectra are significantly affected by the uncertainties in the dark
current, which can be of the same order as the source continuum. Many of
the galaxies in this sample are in this faint flux regime
(f(60) < 50 Jy in the 75" LWS beam). As the
dark currents are only additive in nature across the whole band, they do
not affect the line flux estimates. The dark currents are re-estimated
and removed one at a time by hand through visual inspection using the
LIA. The data are then corrected detector by detector for any evident
instrumental responsivity variations and flux calibrated to the LWS
calibration source Uranus, applied using LIA. Glitches due to cosmic
rays are removed by hand from the data using ISAP by plotting spectral
scans as a function of time and identifying bad data points through the
characteristic appearance of falling glitch trails. Depending on the
quality of the observation of a galaxy, between 15% and 20% of the data
are typically discarded. Spectral scans are co-added and averaged
together using a 3
clip
in spectral bins of about 0.05
µm. For extended sources or for sources that are off-center
with respect to the LWS aperture, a sinusoidal fringe associated with
internal reflection and interference within the LWS instrument may arise
(Gry et al. 2003;
Swinyard et al. 1998).
The fringes are usually less
than 5% of the continuum and do not severely affect the line and
continuum measurements. For full-grating L01 observations, these fringes
can be removed using a defringing algorithm available within ISAP. The
LWS data also suffer from transients. When the grating is scanned
between the forward and reverse directions, a small (< 5%) detector
memory effect
(Gry et al. 2003)
may be visible between the two scan
directions. This memory effect is due to different response times for
the detectors depending on whether the signal increases or decreases
with time and is most visible in the SW1, SW2, and LW2 detectors during
L01 observations. No correction is applied for these memory
effects. When these memory effects are present in the data, each scan
direction is averaged separately, and the line and continuum fluxes for
each scan direction are measured before estimating the final fluxes and
uncertainties (see Sections 4 and
5 for further details). An
additional source of uncertainty occurs for extended sources where the
variation in the LWS beam from detector to detector might cause a
mismatch between adjacent detectors by up to 30% depending on the extent
and structure of the galaxy. With the application of an extended source
correction, this mismatch can be partially corrected. The data presented
in this paper are based on the point source calibration of the pipeline
and no correction for extended sources has been applied due to the
uncertainty in this correction. See the Appendix for the definition and
discussion of the extended source correction.
Through the use of LIA and ISAP, the improvement in the overall quality
of the data from the original pipeline Auto-Analysis Result product is
substantial. By re-estimating the dark currents, the appearance of
negative fluxes in most of these observations is removed. Through the
re-estimation of the dark currents and gain corrections and careful
glitch removal, the match between overlapping detectors is improved,
thus producing more continuous spectra, shown in
Figure 7. Any remaining spectral mismatch between
adjacent detectors may be the result of residual errors in the dark
current subtraction or beam uncertainties from detector to
detector. Using LIA and ISAP, the line and continuum calibration
uncertainties decrease from 20%-30% to 10%-20%, on average, for faint
sources (f(60) < 50 Jy) as illustrated in
Figure 7.