2 Observations and data reduction

2.1 ISO data

The observations with the ISOPHOT (Lemke et al. 1996) photometer onboard the ISO satellite (Kessler et al. 1996) were performed in January 1998 in raster mode (AOT PHT22) with the C200 detector, a $2\times2$ pixel array of stressed Ge:Ga with a pixel size of $89\hbox{$.\!\!^{\prime\prime}$ }4$, in conjunction with the C_160 broad band filter (reference wavelength 170 $\mu\rm m$, equivalent width 89 $\mu$m). Due to the rather large area of the SMC on the sky, the ISOPHOT observations had to be split into a mosaic of nine separate parts, each of which was accompanied by two observations of the ISOPHOT Fine Calibration Source (FCS). Adjacent parts of the whole map were designed to be slightly overlapping, while the raster step size of each part was a full detector size without any overlap.

Since there was little redundancy in the data, cosmic ray hits could mimic compact sources in the map. Therefore, instead of the standard ramp slopes from first-order polynomial fits, the pairwise differences of consecutive ramp readouts were used to derive the detector signals. This allowed a larger distribution to be analysed, leading to considerably more robust results.

To get rid of pairwise differences affected by cosmic ray hits, the robust outlier-insensitive myriad estimator (Kalluri & Arce 1998) was computed and 20% of the most deviant signals as measured by the absolute deviation were cut off. This outlier removal is similar to a median absolute deviation trimming, but instead of the initial median, the sample myriad is used to determine the outliers. The sample myriad value in turn is a robust estimator of the mode (most common value) of a distribution but does not require binning of the actual data set, and is easily computed by simply minimizing a particular cost function with a tuning constant set to a small value (for details see Kalluri & Arce 1998). After rejecting the outliers, the trimmed set of pairwise differences was linearly interpolated to a ten times finer grid and the value giving the minimum value for the myriad cost function accepted as the final signal for each raster point. This interpolation scheme was used as an approximation to a full numerical minimization of the myriad cost function.

Figure 1: False color representation of the 170 $\mu$m ISO SMC map. Logarithmic contours are overplotted at intensities 0.04, 0.06, 0.08, 0.13, 0.20, 0.30, 0.47, 0.72, 1.11, 1.70, 2.63, 4.04, 6.22, 9.58, 14.76, 22.73, and 35.00 Jy/pix (pixel size is 40''). In order to ensure the compatibility with the IRAS HiRes data, the map was constructed from nine smaller ISOPHOT C200 maps with a resulting pixel size of 40'' after the restauration process. Note that in contrast to the IRAS maps, the effective resolution of the restored ISO map is perfectly symmetric. The image is morphologically dominated by the main body of the SMC (the bar), to the east the so-called "bridge'' (connecting the SMC with the LMC) with numerous bright and extended sources is visible. The FIR emission is dominated by several bright star forming regions in the SMC main body (dark blue/white), surrounded by regions of moderate intensity (green). The ISOPHOT observations did not cover the SW region to full extent.

The derived detector signals at each raster position were subsequently corrected for signal dependence on ramp integration times to be consistent with calibration observations (Laureijs et al. 2000), dark-current subtracted, and finally flux calibrated with PIA Version 9.1/Cal G Version 6.0 (Gabriel et al. 1997). For the conversion to an absolute flux level, the observations of the ISOPHOT Fine Calibration Source (FCS) obtained at the beginning and end of each raster in each filter were used.

The flux calibrated data streams of the detector pixels still showed differences in the overall levels of up to 20%, mostly due to inappropriately corrected pixel-to-pixel sensitivities (flat field). If not removed, these varying brightness levels would lead to striping and chessboard-like patterns in the maps. Robust morphological filtering techniques (Sternberg 1986) were used to extract the overall level of the four data streams, which were then brought to a common mean level, thereby giving the relative pixel scaling factors.

Eventually, a complete map of the whole SMC was produced from the flatfielded flux-calibrated data streams of all pixels by using the Drizzle Mapping Method (Hook & Fruchter 2002) within IRAF, which took into account the pixel sizes and inter-pixel distances of the C200 detector and the detector roll angle. The pixel size used for the final full map was $80\hbox{$^{\prime\prime}$ }$, the smallest possible size which did not produce uncovered holes. This final map was restored using the modified Richardson-Lucy-Algorithm (Hook et al. 1994) with an additional subsampling of two pixels and a point spread function approximated by a Gaussian with a FWHM of $2\hbox{$^\prime$ }$, giving a final restored map with $40\hbox{$^{\prime\prime}$ }$ pixel size. This map is shown in Fig. 1.

To allow for a direct comparison with the shorter wavelength IRAS HiRes maps (see below), restored ISOPHOT (sub-)maps with the same center and size as the IRAS sub-fields were created as well as with a pixel size of $30\hbox{$^{\prime\prime}$ }$. This is twice the IRAS HiRes pixel size, so that IRAS maps exactly aligned with the ISO sub-fields could be constructed by a simple $2\times2$ pixel block-averaging.

Although a more quantitative analysis of the FIR properties of the SMC will be performed in subsequent paper, Table 1 lists some preliminary global quantities. Remarkable are the numbers found for the global star formation rate, and the gas-to-dust ratio resulting from the additional cold dust component entering the total dust mass.

 

 

Table 1: Global properties of the SMC, derived from the 170 $\mu$m ISO and the 100 $\mu$m IRAS map. Both data sets were used to construct a color temperature map the referring average dust temperature $<T_{{\rm D},~170/100}>$ was derived from. Dust masses were computed using a $\kappa(\beta , 170~\mu{\rm m})=21.6~{\rm cm}^2~{\rm g}^{-1}$, according to the method presented in Klaas et al. (2001). An average grain size of $a=0.1~\mu$m was used, the grain density was assumed to be $3~{\rm g~cm}^{-3}$. Note that we used $\beta =2$ for our computations. A lower emissivity index would lead to larger values for $\kappa$, hence to dust masses $M_{\rm D}$ which are smaller than for $\beta =2$.

F40-220	$5.6\times 10^{-10}~{\rm W/m}^2$
F1-1000	$7.6\times 10^{-10}~{\rm W/m}^2$
L1-1000	$8.5\times 10^7~{L}_{\odot}$
${\it SFR}$	$0.0148~{M_{\odot}/\rm yr}$
$<T_{{\rm D},~170/100}>$	$20.5~{\rm K}$
$M_{\rm D}$	$3.7\times 10^5~{M_{\odot}}$
$M_{\rm gas}/M_{\rm dust}$	$\approx$1140

2.2 IRAS data

While 12 $\mu$m and 25 $\mu$m images trace mainly hot HII regions and foreground stars with little diffuse IR emission, FIR maps (60 $\mu$m and 100 $\mu$m IRAS observations) show a lot of diffuse emission with a wealth of filamentary structure.

Figure 2: Orientation of the $2^{\circ }\times 2^{\circ }$ degree IRAS HiRes fields in the sky. Centering the SMC on field 0 meant that it covered most of the central body, though not the total area of the SMC. In order to make our source catalogs as complete as possible, all fields containing SMC objects (0, 1, 3, 5, 7) were therefore used for further analysis. Fields 2, 4, 6, and 8 turned out to be devoid of any objects and have therefore not been considered.

Although source catalogs for this wavelength range were published in the past already (for an extensive comparison of IRAS SMC source catalogs with other wavelength bands see, e.g., SI89), we decided to re-analyse the IRAS high resolution (HiRes hereafter) data in order to treat ISO results and IRAS data in an identical and reproducible way. Since the area in the sky covered by the SMC is comparably large ( $6^\circ\times 8^\circ$ in the optical) and IRAS HiRes data can be requested with maximum field sizes of $2^{\circ }\times 2^{\circ }$ only, we split up the SMC field into 9 single $2^{\circ }\times 2^{\circ }$ fields with the central one covering most of the central body of the SMC in the FIR (for the orientation of the fields in the sky see Fig. 2). This ensured that the borderlines between two adjacent fields (though covered by a small overlap) would not be located in the brightest parts of the SMC. The central coordinates for the requested $2^{\circ }\times 2^{\circ }$ fields (equinox B1950.0) are given in Table 2. Pixel sizes were 15''for all fields.

 

 

Table 2: Requested fields for IRAS HiRes processing. Numbers are the same as in Fig. 2, where the orientation of the fields on the sky is given. All four bands (12 $\mu$m, 25 $\mu$m, 60 $\mu$m, 100 $\mu$m) were retrieved for each field and checked for the presence of possible MIR/FIR SMC sources.

Field	RA	DEC
0	13.320000	-72.937271
1	20.093601	-72.937271
2	20.093601	-70.949771
3	13.320000	-70.949771
4	6.5463982	-70.949771
5	6.5463982	-72.937271
6	6.5493982	-74.924771
7	13.320000	-74.924771
8	20.093601	-74.924771

The data reduction for the IRAS scans at IPAC comprises several steps: first, the calibrated, reconstructed detector data is deglitched (which removes spurious non-source-like signals originating from radiation impacts on the detector) and destriped (which corrects for different detector responsivities during different scans, i.e., additive offsets of certain strips). The zodiacal emission model was then subtracted from all data scans individually. For the HiRes data fields a maximum correlation method (MCM) for the reconstruction of the original image is applied to the single scans which not only iteratively builds a reliable model of the sky brightness, but also enhances the resolution to 15'' pixel size. For a more detailed description of IRAS data reduction routines see Assendorp et al. (1995), Bontekoe et al. (1994), and Aumann et al. (1990). All resulting IRAS maps are calibrated in MJy/sr, two of them (central 60 $\mu$m and 100 $\mu$m) are shown in Fig. 3.

The visual inspection of all nine fields yielded the result that only in five of them objects belonging to the SMC are located. These fields are oriented in a cross-like pattern in the sky with numbers 0, 1, 3, 5, and 7 (see Fig. 2). All other fields were neglected for the subsequent analysis.

Figure 3: IRAS HiRes maps of the SMC. Left: 60 $\mu$m map, right: 100 $\mu$m. Since requests for IRAS HiRes maps are limited to $2^{\circ }\times 2^{\circ }$ for an individual field, the whole area of interest was split up into 9 single fields. This figure shows the central field only, but the surrounding eight other fields (see Fig. 2) were studied as well. As expected, the 100 $\mu$m image shows a lower resolution than the 60 $\mu$m data due to the PSF FWHM increase. Consequently, more isolated small-scale structure is seen in the left image.