We obtained our observations over three nights in August 1994 at
the Canada-France-Hawaii Telescope (CFHT) with the multi-object
spectrograph (MOS). The MOS is an imaging, multi-slit
spectrograph that employs a grism as the dispersing
element (see
Le Fèvre et al. 1994 for
details). Objects are selected for spectroscopy using focal plane
masks that are constructed on-line from previously acquired
images. The detector was the Loral3 CCD, a thick CCD with
m
square pixels in a
format, coated to enhance the
quantum efficiency in the blue. The Loral3's read noise was 8
electrons and its gain was set to 1.9electrons/ADU. For the
observations of both M 31 and M 32, we used slits
15
long by 1
wide. No order-sorting filter was
used for any of these observations.
Table 1 presents a log of our observations. During the course of the observations, we used three different grism set-ups in order to optimize throughput, wavelength coverage, and spectral resolution. We used the B600 grism only because of the disappointing throughput of the U900 grism. Although the precise dispersion and wavelength coverage depend upon each object's position within the field of view, Table 1 lists typical values for all three grisms (minimal ranges for the wavelength coverage).
![]() ![]() |
Extracting the spectra proved challenging on account of the nature and faintness of the sources, and on account of the characteristics of the spectrograph. The planetary nebulae in M 31 and M 32 are sufficiently faint that we were unable to detect their continuum emission. Only the emission lines were visible, appearing as a sequence of dots, so it was impossible to trace these spectra. Furthermore, the spectra spanned the full width of the detector, so they suffered from geometric distortion (pin-cushion) introduced by the optics of the spectrograph. Fortunately, we had to include star apertures when defining the spectrograph's focal plane mask to permit accurate re-alignment on the field when ready to do spectroscopy. We used these stars (6 for M 31, 3 for M 32) to map the geometric distortion imposed by the optics, and corrected this distortion using the tasks in the noao.twodspec.longslit package (Anderson 1987). At this point, we had images in which the wavelength axis was parallel to the rows of the CCD, and we could use the brightest line in each spectrum to define an extraction aperture (e.g., Massey et al. 1992). Except for the U900 spectra, the individual spectra were extracted from each image and then combined to produce the final combined spectra. To better define the extraction apertures for the U900 spectra, the spectra were combined first, after verifying that the individual images had the same spatial coordinate scales. In all cases, extraction involved local subtraction of the underlying galaxy and sky spectra.
Establishing a consistent sensitivity scale across all three grism set-ups was a primary consideration of our data reduction. We calibrated the instrumental sensitivity for each set-up using observations of the spectrophotometric standard stars listed in Table 1. We verified that our slitlet-to-slitlet sensitivity scale was secure in three ways. First, the observations of the standard stars were made in pairs through two different slitlets. These slitlets were cut at the red and blue extremes of the field of view to ensure that our standard star observations spanned the full wavelength range of our planetary nebula observations. These paired observations of the standard stars had 500 Å, 800 Å, and 1900 Å of spectrum in common for the U900, B600, and O300 grisms, respectively. In these overlap regions, the sensitivity functions for each grism (on each night) were in agreement. Second, we obtained a spectroscopic sky flat through the standard star mask with the B600 grism on the last night. This mask contained two slitlets in addition to those used for the standard star observations. Comparing the night sky spectra through these four slitlets indicates that variations in the wavelength sensitivity between different slitlets are less than 4.5% (rms). Finally, observations of NGC 6720 were obtained through a different mask than the standard stars, and no wavelength-dependent trends are seen in its sensitivity calibration (see Table 3 below). Therefore, though we did not observe the standard stars through the slitlets used for our program objects, we have no reason to believe that our sensitivity calibration is slitlet-dependent.
We then chose the O300 observations of the planetary nebulae in
M 32 as our reference data set. This choice was
motivated by a number of considerations. First, these
planetary nebulae were observed with all three grisms. Second,
the O300 grism has good sensitivity over the H -
H
wavelength range
(Le Fèvre et al. 1994), which contains the strongest lines in
the spectra. Third, our reddening values for these planetary
nebulae (see Tables 6, 7, and
8) were reasonable, typically E(B-V)<0.2mag, and
invariably positive. These reddenings were consistent with
previous observations of PN1 in M 32
(Ford et al. 1978). The reddening towards M 32 is also
expected to be small if it is in front of the disk of
M 31 (e.g.,
Burstein & Heiles 1984).
We ensured that there were no systematic differences between
the B600 and O300 data sets by comparing the intensities
of H, H
, [O III]
4959, and
He I
5876 measured relative to
[O III]
5007 for the planetary nebulae in
M 32. In making these comparisons, we considered only
those objects for which we had the best detections of these
lines. For these objects, we computed the ratio of the line
intensity in the B600 spectrum to that in the O300 spectrum. Table
2 lists the mean value of this ratio, the standard
error in the mean, and the objects we considered for each line.
Clearly, the main wavelength-dependent trend in Table
2 is a systematic decrease in the B600 sensitivity
relative to the O300 sensitivity as one goes to longer
wavelengths.
![]() ![]() |
![]() ![]() |
![]() ![]() ![]() |
The U900 data required no correction to put them on the
O300 sensitivity scale. We deduced this from direct
comparison with the B600 and O300 data (Tables 6,
7, and 8), and
independently using a spectrum we obtained of the Galactic
planetary nebula NGC 6720. Table 3 lists the
intensities and
reddening values for hydrogen lines in three regions of NGC
6720.
The reddening values we derive from H, H
,H9, H10,
H11, and H12 are in very good agreement in all three
apertures, indicating that our U900 sensitivity calibration
is good to 3750 Å. Our reddening values at H
are
consistently 0.16mag lower than calculated from H
, so our
U900 sensitivities may be under-estimated by 15% near 4100 Å.
Our H8 reddening values are consistently high, but H8 was
blended with He I
3889. We corrected the
blend for the
He I
3889 contribution using the
He I
4471 intensity assuming
no radiative transfer correction, thereby removing the
maximum possible He I
3889 contribution (e.g.,
Aller 1987).
Thus, it is perhaps not surprising that our H8 reddenings are
too high. Overall, our Balmer line intensities for NGC 6720
indicate that our U900 sensitivity calibration is secure from
3750 Å to H
. Similarly, for the planetary nebulae in
M 32
(Tables 6, 7, and 8),
the U900 line intensities for [O II]
3727,
[Ne III]
3869, and
He II
4686 are in excellent agreement with
their B600 and O300 counterparts.
Figures 1 through 6 display the O300,
B600, and U900 spectra of the planetary nebulae in M 32, while
Figs. 7 through 12 display the
B600 spectra of the planetary nebulae
in the bulge of M 31.
The object designations
(Ciardullo et al. 1989)
are shown next to the spectra.
Normally, the
spectra are scaled such that H occupies the full intensity
scale, so stronger lines from adjacent spectra overlap, but
some of the U900 and B600 spectra are scaled such that H
and
H
, respectively, occupy the full intensity scale. This
scaling allowed the best compromise in demonstrating the
signal-to-noise for various lines and an assessment of the
background sky and galaxy subtraction. The full wavelength
range is shown for the B600 and U900 spectra, but only the
wavelength range below 7350 Å is shown for the O300 spectra.
Cosmic rays were not removed unless they interfered with the
measurement of line intensities, and many remain in the
spectra displayed in Figs. 1 through 12.
![]() |
Figure 1:
The O300 spectra for PN8, PN11, PN2, PN7,
H II1,
and PN25 in the M 32 field. The spectra are displayed such
that H![]() ![]() |
![]() |
Figure 2: The O300 spectra for PN24, PN6, PN5, PN1, PN4, and PN17 in the M 32 field. The format is identical to Fig. 1. Like PN25, PN24 is also very close to M 32's nucleus and suffers from somewhat poorer background subtraction. Note that PN4 and PN17 are background planetary nebulae in the disk of M 31 (Ford & Jenner 1975) |
![]() |
Figure 3:
The B600 spectra for PN8, PN11, PN2, PN7, H II 1,
and PN25 in the M 32 field. For PN8, PN11, and PN25, the
scaling is such that H![]() ![]() |
![]() |
Figure 4:
The B600 spectra for PN24, PN6, PN5, PN1, PN4,
and PN17 in the M 32 field. For PN24 and PN17 (M 31),
H![]() ![]() |
![]() |
Figure 5:
The U900 spectra for PN8, PN11, PN2, PN7, H II 1,
and PN25 in the M 32 field. Only for PN11 and PN25 does H![]() ![]() |
![]() |
Figure 6:
The U900 spectra for PN24, PN6, PN5, PN1, PN4,
and PN17 in the M 32 field. H![]() ![]() |
![]() |
Figure 8: The B600 spectra for PN28, PN23, PN12, PN10, and PN1 in the M 31 bulge field. The intensity and wavelength scales are as in Fig. 7. Note that the background subtraction is poorer for PN12 than is normally the case |
![]() |
Figure 9: The B600 spectra for PN3, PN38, PN36, PN53, and PN52 in the M 31 bulge field. The intensity and wavelength scales are as in Fig. 7 |
![]() |
Figure 10: The B600 spectra for PN42, PN45, PN43, PN48, and PN95 in the M 31 bulge field. The intensity and wavelength scales are as in Fig. 7 |
![]() |
Figure 11: The B600 spectra for PN47, PN408, PN93, PN92, and PN91 in the M 31 bulge field. The intensity and wavelength scales are as in Fig. 7. Note that the signal-to-noise is poor for the very faint object PN408 |
Copyright The European Southern Observatory (ESO)