NGC 1808

   
2 Observations and data reduction

Figure 1: The ISOCAM CVF field of view, indicated by a square, is superimposed on publicly available images in the V band for M 82 (left), and R band for NGC 253 and NGC 1808 (middle and right).

 

 

Table 1: Log of the observations.

Source	Date	Revolution	$\alpha_{2000}$	$\delta_{2000}$	PA a	Pixel scale	$t_{\rm int}$ b
			( ${\rm { }^{h}:{ }^{m}:{ }^{s}}$)	( ${ }^{\circ}:{ }^{\prime}:{ }^{\prime\prime}$)	(degree)	(arcsec)	(s)
M 82	1996 Mar 19	123	09:55:51.8	+69:40:45.8	212	3.0	25.2
NGC 253	1997 Jan 03	414	00:47:32.9	-25:17:18.3	290	1.5	25.2
NGC 1808	1997 Oct 12	697	05:07:42.3	-37:30:46.1	64	3.0	25.2

^a Position angle of the vertical axis of the ISOCAM array relative to the north and increasing counterclockwise.
^b Total integration time per wavelength channel.

We observed M 82, NGC 253, and NGC 1808 with ISOCAM (Cesarsky et al. 1996c) on board ISO (Kessler et al. 1996) as part of the guaranteed time program CAMACTIV (P. I. I. F. Mirabel). We used the CVFs to cover the entire $\lambda = 5.0$- $16.5~{\rm\mu m}$ range with a resolution of $R \equiv \lambda/\Delta\lambda \approx 35$ to 45 and increments $\delta\lambda = 0.06$ to 0.1$~\mu$m from short to long wavelengths. For all three sources, twelve single-frame exposures of 2.1 s were recorded per wavelength channel, for total integration times of about ${\rm 1^{h} 05^{m}}$. We selected the $3^{\prime\prime}~{\rm pixel^{-1}}$scale for M 82 and NGC 1808, and the $1.5^{\prime\prime}~{\rm pixel^{-1}}$ scale for NGC 253; total fields of view of $96^{\prime\prime} \times 96^{\prime\prime}$ and $48^{\prime\prime} \times 48^{\prime\prime}$ are thus covered by the $32 \times 32$ pixels detector array. Table 1 presents the details of the observations. The areas observed with ISOCAM cover almost entirely the MIR sources in each galaxy (Sect. 4) and are indicated on optical images in Fig. 1. This emphasizes the small MIR source sizes compared to the optical extent of the galaxies and that most of the MIR emission originates in their nuclear regions.

We reduced the data using the CAM Interactive Analysis package (CIA) following the procedures described in the ISOCAM User's Manual (Delaney 1997). We first subtracted the dark current using a model of the secular evolution of ISOCAM's dark current (Biviano et al. 1997), removed cosmic ray hits by applying a multiresolution median filtering technique (Starck et al. 1999), and corrected for transient detector memory effects using the "Fouks-Schubert'' algorithm (Coulais & Abergel 2000). We then combined the individual exposures for each wavelength channel with the help of a shift-and-add algorithm accounting for the satellite jitter motions (typical amplitude of $\approx$ $1^{\prime\prime}$) and performed absolute flux calibration based on the sensitivity calibration files provided in CIA version 3.0.

For proper comparison of fluxes and maps obtained at different wavelengths, we smoothed the CVF channels to a common angular resolution and registered them appropriately. The point spread function (PSF) varies throughout the CVF scans, with theoretical full-width at half-maximum ${FWHM} = 3.1^{\prime\prime}$- $5.6^{\prime \prime }$ for the $3^{\prime\prime}~{\rm pixel^{-1}}$ scale and $2.0^{\prime\prime}$- $5.2^{\prime\prime}$ for the $1.5^{\prime\prime}~{\rm pixel^{-1}}$ scale, over the range 5.5- $16.5~{\rm\mu m}$. We convolved the data to a final resolution of 5.6 $^{\prime\prime}$ for M 82 and NGC 1808, and 5.2 $^{\prime\prime}$ for NGC 253, corresponding to a linear resolution of 90, 297, and 62 pc, respectively. We computed the transfer functions (close to Gaussians in profile) using the library of theoretical PSFs for ISOCAM and a CLEAN algorithm. The small-scale structure is thus smeared out in the higher resolution channels but no important spatial feature is lost.

The source position varies slightly with wavelength due to telescope motions and changes in optical path, with the largest offset occurring near 9.2$~\mu$m between the short and long wavelength segments of the CVFs. Since no unresolved point source is detected in the data, we determined the position offsets from cross-correlation over the entire field of view, taking as references images integrated over various wavelength intervals. Effects due to the different spatial distributions of emission features were easily identified and ignored in the determination of the instrumental shifts. We fitted straight lines to the cross-correlation offsets for each CVF segment. For the $3^{\prime\prime}~{\rm pixel^{-1}}$ scale, the horizontal, or x-axis offsets relative to 9.2$~\mu$m in the short and long wavelength segments did not exceed 0.3 pixel and 0.5 pixel, respectively, with a jump of 0.7 pixel at the transition wavelength. Shifts along the y-axis were negligible throughout the entire wavelength range (< $0.2~{\rm pixel}$). For the $1.5^{\prime\prime}~{\rm pixel^{-1}}$ scale, both x- and y-axis offsets were $\leq$ $0.3~{\rm pixel}$ within each CVF segment, with jumps at 9.2$~\mu$m of 0.2 pixel only.

We did not attempt to correct for instrumental flat-field, straylight, and ghost images generated by multiple reflections between the detector and filters since appropriate calibration files for the CVF mode were not available. Based on theoretical models for point-like and extended sources as well as observations of stars and of the zodiacal light (Biviano et al. 1998a,b; Okumura 2000), the combined flat-field, straylight, and ghost effects integrated over the entire detector array vary between $\approx$$35\%$ and 10% of the source flux from 5 to 16$~\mu$m, and exhibit qualitatively similar spatial structures at different wavelengths. The rms noise estimated from existing flat-field images obtained with ISOCAM's broad-band filters is $\sim$10% for both pixel scales (Biviano et al. 1998b).

Based on available calibration accuracy reports (Blommaert & Cesarsky 1998 and references therein), the systematic errors of the data processing and flux calibration described above sum up to total uncertainties on the absolute photometry of $\approx$$50\%$. For our analysis, we will consider formal effective uncertainties on the relative fluxes estimated as follows. We measured the dispersion around the mean flux among the individual reduced exposures for each wavelength channel at each detector pixel, included an additional 10% rms noise for intrinsic pixel-to-pixel variations in sensitivity, and accounted for uncertainties in relative spectral calibration. For M 82 and NGC 253, the latter were evaluated from the comparison with data from the ISO Short Wavelength Spectrometer (SWS; de Graauw et al. 1996) in Sect. 3, amounting to 5% and 11%, respectively. For NGC 1808, very low flux levels and high noise around 10$~\mu$m, below 6$~\mu$m, and outside of the main emission region led to poor correction for transient memory effects because of the difficulty in determining the initial and stabilized fluxes and of the bad behaviour of the algorithm wherever fluxes become negative. In this case, we estimated the errors in relative spectral calibration by comparing results when applying the transient correction to the data as they were and to the data where all values lower than 0.01 were set to 0.01, adding a conservative 10%. The median and average differences are of 16% and a factor of 2. Experimentation showed that the choice of the threshold is of little consequence as long as it is sufficiently smaller than the signal where the transient correction is satisfactory.

Overall, the formal effective uncertainties per pixel and wavelength channel within the brighter emission regions of interest (defined by the "starburst cores'' in Sect. 3) have median values of 18%, 29%, and 45% for M 82, NGC 253, and NGC 1808. Computed over the entire ISOCAM field of view, the medians increase to 32%, 88%, and a factor of 2.6 as a result of the higher noise towards the edges of the array where little flux is detected.

NGC 1808