We extracted from the reduced ISOCAM data cubes spectra integrated over various regions: a "starburst core'' centered on each galaxy's nucleus and covering most of the emission source, and a smaller region at the position of the peak observed for the integrated MIR emission. For M 82 and NGC 253, we further included a "disk'' corresponding to an annulus outside of the starburst core and the ISOCAM field of view. Due to the higher noise level in the NGC 1808 data, only the starburst core and MIR peak spectra are of sufficient quality for analysis. The disk, core, and MIR peak regions were chosen to sample relatively quiescent to intense star-forming activity as traced by e.g. the 15 $~\mu$ m continuum emission (see Sects. 3.2 and 5.2). The starburst cores correspond to the apertures generally used in previous studies when referring to the global properties of the starburst in each galaxy.

$\begin{figure} \par\includegraphics[width=5.8cm,clip]{h3938f2.ps}\end{figure}$

Figure 2: Selected regions in M 82 (top), NGC 253 (middle), and NGC 1808 (bottom), shown on $\lambda = 14.8$ - $15.2~{\rm \mu m}$ continuum maps. The starburst core and MIR peak regions are enclosed within the black and white circles, respectively. The disk regions for M 82 and NGC 253 correspond to the hatched areas. The parameters for all regions are given in Table 2. The horizontal bar and the filled circle at the bottom of each panel indicate the physical linear scale and the FWHM of the PSF.

Figure 2 indicates the selected regions on maps of the 14.8- $15.2~{\rm \mu m}$ emission. Figure 3 shows the spectra, all normalized to a total 6.0- $6.6~{\rm\mu m}$ flux density of unity to facilitate comparison of the relative strength of the emission features at $\lambda \la 11~{\rm\mu m}$ and of the continuum intensity at longer wavelengths. The formal effective uncertainties are plotted along with the spectra. Table 2 gives the parameters for the synthetic apertures, the normalizing fluxes, and the average and median uncertainties for each spectrum. Figure 3 also shows for M 82 and NGC 253 the average spectrum of all individual resolution elements within the ISOCAM field of view, normalized as described above, together with the dispersion around the mean, the full range observed, and the typical (median) uncertainties. The resolution elements for M 82 and NGC 253 correspond to rebinned pixels of size $2 \times 2$ and $4 \times 4$ in original detector pixels, respectively.

Table 2: Synthetic apertures for selected regions.
Source Region Position $\Delta\alpha$ ^a $\Delta\delta$ ^a Aperture $f_{\rm 6.0-6.6~\mu m}$ ^b Uncertainties ^c

(arcsec) (arcsec) (arcsec) (Jy) (%)

M 82 CAM FOV ... 0.0 +5.0 $81 \times 81$ at PA of $212^{\circ}$ $45.0 \pm 0.3$ 6 (1)

Disk Nucleus 0.0 0.0 $20 \leq r \leq 30$ $9.37 \pm 0.06$ 3 (2)

Core Nucleus 0.0 0.0 $r \leq 15$ $19.5 \pm 0.2$ 2 (2)

MIR peak MIR Peak -3.0 -1.5 $r \leq 6$ $5.74 \pm 0.10$ 5 (5)

NGC 253 CAM FOV ... -4.0 +1.0 $40.5 \times 40.5$ at PA $290^{\circ}$ $8.34 \pm 0.73$ 20 (3)

Disk Nucleus 0.0 0.0 $10 \leq r \leq 15$ $1.75 \pm 0.09$ 11 (4)

Core Nucleus 0.0 0.0 $r \leq 7.5$ $3.90 \pm 0.05$ 4 (3)

MIR peak MIR Peak 0.0 0.0 $r \leq 3$ $1.14 \pm 0.03$ 8 (7)

NGC 1808 Core Nucleus 0.0 0.0 $r \leq 15$ $1.77 \pm 0.15$ 57 (4)

MIR peak MIR Peak +1.5 -1.5 $r \leq 6$ $0.74 \pm 0.02$ 9 (8)

^a
Right ascension and declination offsets relative to the position of the nucleus of each galaxy. For M 82: $\alpha_{2000} = 09^{\rm h} 55^{\rm m} 52\hbox{$.\!\!^{\rm s}$ }2$ , $\delta_{2000} = +69^{\circ} 40^{\prime} 46\hbox{$.\!\!^{\prime\prime}$ }6$ (Dietz et al. 1986). For NGC 253: $\alpha_{2000} = 00^{\rm h} 47^{\rm m} 33\hbox{$.\!\!^{\rm s}$ }1$ , $\delta_{2000} = -25^{\circ} 17^{\prime} 18\hbox{$.\!\!^{\prime\prime}$ }3$ (Sams et al. 1994). For NGC 1808: $\alpha_{2000} = 05^{\rm h} 07^{\rm m} 42\hbox{$.\!\!^{\rm s}$ }3$ , $\delta_{2000} = -37^{\circ} 30^{\prime} 46\hbox{$.\!\!^{\prime\prime}$ }3$ (Collison et al. 1994).
^b
Integrated flux density between 6.0 and 6.6 $~\mu$ m used for the normalization of the spectra plotted in Fig. 3. The integration area reported for the ISOCAM field of view of M 82 and NGC 253 excludes the edges which were not illuminated or are highly noisy. Quoted uncertainties are computed from the formal effective uncertainties for individual pixels and wavelength channels (see Sect. 2).
^c
Average of the formal effective uncertainties per wavelength channel for the integrated spectra (median values are given in parenthesis).

**Table 2:** Synthetic apertures for selected regions.
Source	Region	Position	$\Delta\alpha$ ^a	$\Delta\delta$ ^a	Aperture	$f_{\rm 6.0-6.6~\mu m}$ ^b	Uncertainties ^c
			(arcsec)	(arcsec)	(arcsec)	(Jy)	(%)
M 82	CAM FOV	...	0.0	+5.0	$81 \times 81$ at PA of $212^{\circ}$	$45.0 \pm 0.3$	6 (1)
	Disk	Nucleus	0.0	0.0	$20 \leq r \leq 30$	$9.37 \pm 0.06$	3 (2)
	Core	Nucleus	0.0	0.0	$r \leq 15$	$19.5 \pm 0.2$	2 (2)
	MIR peak	MIR Peak	-3.0	-1.5	$r \leq 6$	$5.74 \pm 0.10$	5 (5)
NGC 253	CAM FOV	...	-4.0	+1.0	$40.5 \times 40.5$ at PA $290^{\circ}$	$8.34 \pm 0.73$	20 (3)
	Disk	Nucleus	0.0	0.0	$10 \leq r \leq 15$	$1.75 \pm 0.09$	11 (4)
	Core	Nucleus	0.0	0.0	$r \leq 7.5$	$3.90 \pm 0.05$	4 (3)
	MIR peak	MIR Peak	0.0	0.0	$r \leq 3$	$1.14 \pm 0.03$	8 (7)
NGC 1808	Core	Nucleus	0.0	0.0	$r \leq 15$	$1.77 \pm 0.15$	57 (4)
	MIR peak	MIR Peak	+1.5	-1.5	$r \leq 6$	$0.74 \pm 0.02$	9 (8)

All spectra look very similar and exhibit the "classical'' characteristics observed towards star-forming regions and galaxies: conspicuous broad emission features (the "unidentified infrared bands'' or UIBs), a featureless continuum rising importantly at $\lambda \ga 11~{\rm\mu m}$ , and an apparent dip near 10 $~\mu$ m. The UIBs are attributed to a family of particles, the nature of which still is debated, stochastically heated by single ultraviolet photons while the long-wavelength continuum is ascribed to very small dust grains between the transient heating and thermal equilibrium regimes depending on grain properties and radiation field intensity (e.g. Léger et al. 1989; Allamandola et al. 1989; Désert et al. 1990; Duley & Williams 1991; Tielens et al. 1999; see also reviews by Puget & Léger 1989; Cesarsky & Sauvage 1999; Genzel & Cesarsky 2000). We will hereafter refer to these ISM components as "PAHs,'' adopting the currently popular model in which the UIB carriers consist of polycyclic aromatic hydrocarbon molecules, and "VSGs.''

A number of weaker emission features are also detected in the spectra of Fig. 3. However, their identification is problematic at such low spectral resolution due to possible feature blends. To emphasize this point and secure the identifications, Fig. 4 shows the spectra of M 82 and NGC 253 obtained at $R \sim 500$ -1000 with the ISO-SWS (from Sturm et al. 2000 and Förster Schreiber et al. 2001). The SWS data are also compared with ISOCAM spectra taken in the same apertures, after convolution to the same spectral resolution. The ISOCAM and SWS data agree very well, confirming the accuracy of the absolute and relative flux calibration for both instruments. For M 82, the differences are 5% on average (24% at most). For NGC 253, they are of 11% on average (30% at most) at $\lambda \geq 5.4~{\rm\mu m}$ while they reach a factor of two at shorter wavelengths, probably due to residual transient memory effects.

From this comparison (see also Sturm et al. 2000), the features at 6.2, 7.7, 8.6, and 11.3 $~\mu$ m are unambiguously identified with PAH emission. The broad feature near 12.7 $~\mu$ m clearly results from the blending of the PAH 12.7 $~\mu$ m band and of the [Ne II] 12.81 $~\mu$ m fine-structure line. The [Ne III] 15.56 $~\mu$ m line is also blended with the nearby PAH 15.7 $~\mu$ m feature. From the SWS data, and with the continuum and integration bandpasses given below, the PAH 12.7 $~\mu$ m accounts for about 50% of the flux in the 12.7 $~\mu$ m blend for both galaxies. The PAH 15.7 $~\mu$ m contains about 30% of the flux in the 15.6 $~\mu$ m blend for M 82 (the noisier SWS spectrum of NGC 253 makes an estimate difficult). The least contaminated fine-structure line detected with ISOCAM is [Ar II] 6.99 $~\mu$ m, with the underlying PAH at 7.0 $~\mu$ m and the ${\rm H_{2}}$ 0-0 S(5) rotational line at 6.91 $~\mu$ m contributing $\approx$ $20\%$ and 5%, respectively, to the blend flux in both galaxies. The weak features at 5.65, 13.55, and 14.25 $~\mu$ m are identified with PAH bands; the latter is definitely not due to the high excitation [Ne V] 14.32 $~\mu$ m line which is undetected in the SWS data. The final identifications in the ISOCAM spectra are given in Fig. 3. Because of the excellent correspondence between features seen in all three galaxies, the identifications for M 82 and NGC 253 are adopted for NGC 1808 as well.

$\begin{figure} \par\includegraphics[width=15cm,clip]{h3938f3.ps}\end{figure}$

Figure 3: Mid-infrared spectra of M 82, NGC 253, and NGC 1808. All spectra are normalized to unit total flux density between 6.0 and $6.6~{\rm\mu m}$ . The three panel pairs to the left show the spectra and relative formal effective uncertainties for selected regions: the ISOCAM field of view, the disk region, the starburst core, and the MIR peak (see labels in each plot); the apertures used, normalizing factors, and typical uncertainties are given in Table 2. The top and middle panel pairs to the right show the average spectrum together with the dispersion, the full range observed, and the median uncertainties derived from all individual resolution elements for M 82 and NGC 253, respectively. The resolution elements correspond to rebinned $2 \times 2$ pixels for M 82 and $4 \times 4$ pixels for NGC 253 (i.e. $6^{\prime \prime } \times 6^{\prime \prime }$ in both cases). The bottom right panel pair compares the spectra of the starburst core regions of each galaxy and their relative uncertainties. The relative uncertainties are expressed as fraction of the measured flux density.

$\begin{figure} \par\includegraphics[width=14.4cm,clip]{h3938f4.ps}\end{figure}$

Figure 4: Comparison of MIR spectra of M 82 and NGC 253 obtained with ISOCAM and SWS. The top panels show the full resolution SWS spectra ( $R \sim 1000$ to 500 from short to long wavelengths), with the identification of the detected emission features (from Förster Schreiber et al. 2001 for M 82, and Sturm et al. 2000 for NGC 253). For M 82, the peak flux densities for the [NeII] $12.81~{\rm \mu m}$ and [NeIII] $15.56~{\rm \mu m}$ lines are 291.8 Jy and 109.1 Jy. For NGC 253, those for the [ArII] $6.99~{\rm \mu m}$ , [NeII] $12.81~{\rm \mu m}$ , and [NeIII] $15.56~{\rm \mu m}$ lines are 31.2 Jy, 140.6 Jy, and 35.2 Jy. The middle panels compare the SWS data convolved to the resolution of ISOCAM ( $R \sim 35$ to 45 from short to long wavelengths) with the ISOCAM spectra (with uncertainties) integrated over the SWS aperture, while the bottom panels show the ratio of the ISOCAM and smoothed SWS spectra. In the spectral range covered by ISOCAM, the SWS aperture changes from $14^{\prime \prime } \times 20^{\prime \prime }$ shortwards of $\lambda \approx 12.0~{\rm \mu m}$ to $14^{\prime \prime } \times 27^{\prime \prime }$ at longer wavelengths. The long wavelength portions of the spectra have been scaled to match the smaller SWS apertures by factors of 0.83 for M 82 and 0.9 for NGC 253, as determined from the continuum level measured in the ISOCAM data.

We performed various continuum and feature measurements on the spectra. Table 3 gives the intervals used in the computations and Tables 4 and 5 report the results. The flux densities in the 6.0- $9.0~{\rm\mu m}$ and 13.5- $15.0~{\rm\mu m}$ bands are dominated by emission from PAHs and VSGs, respectively. We chose the narrow continuum bands centered at 5.5 and 15.0 $~\mu$ m so as to minimize the contribution from PAHs and other emission lines. In particular, we obtain the same 5.5 $~\mu$ m flux densities within $\approx$ $5\%$ from the ISOCAM data and from the higher resolution SWS spectra for the SWS field of view in both M 82 and NGC 253, indicating negligible contribution from the adjacent PAH 5.65 $~\mu$ m feature and [Fe II] 5.34 $~\mu$ m line in the lower resolution data. We measured the fluxes in the PAH features at 6.2, 7.7, 8.6, and 11.3 $~\mu$ m, in the [Ar II] 6.99 $~\mu$ m line, and in the PAH 12.7 $~\mu$ m + [Ne II] 12.81 $~\mu$ m and PAH 15.7 $~\mu$ m + [Ne III] 15.56 $~\mu$ m blends by integrating the flux under the feature profiles after subtracting a continuum baseline obtained by linear interpolation between adjacent spectral intervals. More sophisticated methods such as profile fitting (e.g. Uchida et al. 2000) are not necessary for our purposes and are difficult to apply to the lower signal-to-noise (S/N) ratio data of individual pixels in generating linemaps (Sect. 3.2).

We did not compute fluxes for the individual [Ne II] 12.81 $~\mu$ m and [Ne III] 15.56 $~\mu$ m lines. The wavelength sampling is too coarse for reliable profile decomposition, with the feature peaks of each blend sampled by adjacent wavelength channels. In addition, our attempts to subtract the PAH contribution by attributing an excess in the blend 12.7 $~\mu$ m/PAH 11.3 $~\mu$ m ratio to the [Ne II] 12.81 $~\mu$ m line proved too sensitive to the definition of "pure'' PAH 12.7 $~\mu$ m/11.3 $~\mu$ m ratio (e.g. as measured outside of the starburst cores where comparatively little fine-structure line emission from H II regions is expected). Complications further arise from possible intrinsic variations in the PAH ratios, extinction effects ( $A_{\rm 12.7~\mu m}/A_{\rm 11.3~\mu m} \approx 0.55$ ), and unconstrained fine-structure line emission from disk H II regions. The fluxes for [Ar II] 6.99 $~\mu$ m are much more reliable because possible contributions by other features in our sample galaxies are substantially smaller, as mentioned above.

3.2 Mid-infrared images: Mapping the features

We obtained broad- and narrow-band images as well as maps of the PAHs and [Ar II] 6.99 $~\mu$ m line emission from the ISOCAM data cubes by applying to each pixel the procedures described above for the spectra. Figures 5-7 present selected images and ratio maps for M 82, NGC 253, and NGC 1808. Contours corresponding to the same levels relative to the peak intensity are plotted for all continuum and emission feature maps for ease of comparison by visual inspection. A $3\sigma$ contour is also shown to delineate regions where observed small-scale structures are reliable. In the following, we describe the various maps source by source; their interpretation will be discussed in subsequent sections.

3.2.1 M 82

The continuum and emission feature maps in M 82 show a globally smooth spatial distribution, centered and peaking roughly 5 $^{\prime\prime}$ southwest of the nucleus. The PAH and 5.5 $~\mu$ m continuum emission are the most extended and symmetric, with disk-like isophotes elongated along the galactic plane ( ${\rm PA} \approx 70^{\circ}$ ). In contrast, the 15 $~\mu$ m continuum and [Ar II] 6.99 $~\mu$ m line emission have more compact distributions which are more asymmetric relative to the nucleus. The 15 $~\mu$ m/5.5 $~\mu$ m ratio map outlines well the difference in peak morphology between the short and long wavelength continuum. The [Ar II] 6.99 $~\mu$ m distribution is the most compact, with centroid (determined from the emission out to a radius of 25 $^{\prime\prime}$ ) displaced 3.5 $^{\prime\prime}$ southwest of that of the PAH and continuum emission, and showing only a slight extension towards the east.

Our ISOCAM maps provide an important complementary dataset to existing MIR images in the literature, which were mostly obtained in different bands or over limited regions (although with higher angular resolution up to $\approx$ $1^{\prime\prime}$ ). The distributions observed in our PAH maps and for the PAH 3.29 $~\mu$ m feature by Normand et al. (1995) and Satyapal et al. (1995) at $\approx$ $1^{\prime\prime}$ resolution are consistent with each other. Our 15 $~\mu$ m continuum map and the 19.2 $~\mu$ m image of Telesco et al. (1991) are similar. Maps of the N-band (10.8 $~\mu$ m) and 11.8- $13.0~{\rm\mu m}$ emission generated from the ISOCAM data cubes agree well with those of Telesco et al. (1991) and Telesco & Gezari (1992) within the regions covered by the latter two images.

Table 3: Intervals used to measure the continuum and features.
Property Symbol Continuum points or intervals Integration limits

( $~\mu$ m) ( $~\mu$ m)

PAH emission^a $f_{\rm PAH}$ ... 6.00-9.00

VSG emission^b $f_{\rm VSG}$ ... 13.5-15.0

5.5 $~\mu$ m continuum $f_{\rm 5.5~\mu m}$ ... 5.40-5.52

15.0 $~\mu$ m continuum $f_{\rm 15~\mu m}$ ... 14.8-15.2

PAH 6.2 $~\mu$ m F_6.2 5.81-5.99, 6.57-6.80 6.04 - 6.51

PAH 7.7 $~\mu$ m F_7.7 7.14, 8.22-8.38 7.19-8.17

PAH 8.6 $~\mu$ m F_8.6 8.27-8.33, 8.84-8.89 8.33-8.84

PAH 11.3 $~\mu$ m F_11.3 10.95, 11.7-11.8 11.1-11.7

PAH 12.7 $~\mu$ m + [Ne II] 12.81 $~\mu$ m F_12.7 12.1-12.2, 13.2-13.3 12.3-13.2

PAH 15.7 $~\mu$ m + [Ne III] 15.56 $~\mu$ m F_15.6 15.0-15.2, 16.0-16.1 15.3-15.9

[Ar II] 6.99 $~\mu$ m $F_{\rm [Ar~II]}$ 6.74-6.86, 7.14-7.19 6.86-7.14

^a Integrated flux in the band including the short wavelength PAH emission complex.
^b Integrated flux in the band probing the continuum emission from VSGs free from strong emission lines and PAH features.

**Table 3:** Intervals used to measure the continuum and features.
Property	Symbol	Continuum points or intervals	Integration limits
		( $~\mu$ m)	( $~\mu$ m)
PAH emission^a	$f_{\rm PAH}$	...	6.00-9.00
VSG emission^b	$f_{\rm VSG}$	...	13.5-15.0
5.5 $~\mu$ m continuum	$f_{\rm 5.5~\mu m}$	...	5.40-5.52
15.0 $~\mu$ m continuum	$f_{\rm 15~\mu m}$	...	14.8-15.2
PAH 6.2 $~\mu$ m	F_6.2	5.81-5.99, 6.57-6.80	6.04 - 6.51
PAH 7.7 $~\mu$ m	F_7.7	7.14, 8.22-8.38	7.19-8.17
PAH 8.6 $~\mu$ m	F_8.6	8.27-8.33, 8.84-8.89	8.33-8.84
PAH 11.3 $~\mu$ m	F_11.3	10.95, 11.7-11.8	11.1-11.7
PAH 12.7 $~\mu$ m + [Ne II] 12.81 $~\mu$ m	F_12.7	12.1-12.2, 13.2-13.3	12.3-13.2
PAH 15.7 $~\mu$ m + [Ne III] 15.56 $~\mu$ m	F_15.6	15.0-15.2, 16.0-16.1	15.3-15.9
[Ar II] 6.99 $~\mu$ m	$F_{\rm [Ar~II]}$	6.74-6.86, 7.14-7.19	6.86-7.14

Table 4: Broad- and narrow-band measurements^a.
Source Region $f_{\rm PAH}$ $f_{\rm VSG}$ $f_{\rm 5.5~\mu m}$ $f_{\rm 15~\mu m}$

M 82 CAM FOV $61.1 \pm 0.8$ $78.3 \pm 0.3$ $6.19 \pm 1.98$ $80.8 \pm 0.6$

Disk $12.7 \pm 0.1$ $11.3 \pm 0.1$ $1.16 \pm 0.17$ $10.8 \pm 0.1$

Core $26.9 \pm 0.1$ $45.7 \pm 0.3$ $3.15 \pm 0.08$ $49.3 \pm 0.5$

MIR Peak $8.03 \pm 0.06$ $16.3 \pm 0.2$ $1.02 \pm 0.04$ $17.9 \pm 0.4$

NGC 253 CAM FOV $11.9 \pm 0.3$ $25.3 \pm 0.2$ $2.12 \pm 1.31$ $27.2 \pm 0.3$

Disk $2.49 \pm 0.03$ $3.93 \pm 0.03$ $0.41 \pm 0.19$ $4.12 \pm 0.06$

Core $5.63 \pm 0.03$ $15.3 \pm 0.1$ $1.05 \pm 0.11$ $16.8 \pm 0.3$

MIR Peak $1.68 \pm 0.02$ $5.74 \pm 0.09$ $0.35 \pm 0.02$ $6.40 \pm 0.19$

NGC 1808 Core $2.66 \pm 0.05$ $3.13 \pm 0.04$ ... $3.08 \pm 0.07$

MIR Peak $1.10 \pm 0.01$ $1.39 \pm 0.03$ $0.11 \pm 0.01$ $1.41 \pm 0.05$

**Table 4:** Broad- and narrow-band measurements^a.
Source	Region	$f_{\rm PAH}$	$f_{\rm VSG}$	$f_{\rm 5.5~\mu m}$	$f_{\rm 15~\mu m}$
M 82	CAM FOV	$61.1 \pm 0.8$	$78.3 \pm 0.3$	$6.19 \pm 1.98$	$80.8 \pm 0.6$
	Disk	$12.7 \pm 0.1$	$11.3 \pm 0.1$	$1.16 \pm 0.17$	$10.8 \pm 0.1$
	Core	$26.9 \pm 0.1$	$45.7 \pm 0.3$	$3.15 \pm 0.08$	$49.3 \pm 0.5$
	MIR Peak	$8.03 \pm 0.06$	$16.3 \pm 0.2$	$1.02 \pm 0.04$	$17.9 \pm 0.4$
NGC 253	CAM FOV	$11.9 \pm 0.3$	$25.3 \pm 0.2$	$2.12 \pm 1.31$	$27.2 \pm 0.3$
	Disk	$2.49 \pm 0.03$	$3.93 \pm 0.03$	$0.41 \pm 0.19$	$4.12 \pm 0.06$
	Core	$5.63 \pm 0.03$	$15.3 \pm 0.1$	$1.05 \pm 0.11$	$16.8 \pm 0.3$
	MIR Peak	$1.68 \pm 0.02$	$5.74 \pm 0.09$	$0.35 \pm 0.02$	$6.40 \pm 0.19$
NGC 1808	Core	$2.66 \pm 0.05$	$3.13 \pm 0.04$	...	$3.08 \pm 0.07$
	MIR Peak	$1.10 \pm 0.01$	$1.39 \pm 0.03$	$0.11 \pm 0.01$	$1.41 \pm 0.05$

^a: All flux densities are expressed in Jy. The uncertainties result from the formal effective uncertainties of the relative fluxes (see Sect. 2).

Table 5: Emission feature measurements^a.
Source Region F_6.2 F_7.7 F_8.6 F_11.3 F_12.7 F_15.6 $F_{\rm [Ar~II]}$

M 82 CAM FOV $103 \pm 1$ $226 \pm 1$ $31.8 \pm 2.4$ $57.7 \pm 0.9$ $54.0 \pm 0.4$ $5.77 \pm 0.47$ $6.05 \pm 0.31$

Disk $21.5 \pm 0.2$ $46.5 \pm 0.2$ $6.70 \pm 0.08$ $13.1 \pm 0.1$ $9.74 \pm 0.06$ $0.76 \pm 0.02$ $0.67 \pm 0.02$

Core $43.7 \pm 0.5$ $100 \pm 1$ $13.8 \pm 0.4$ $22.0 \pm 0.5$ $26.4 \pm 0.5$ $3.49 \pm 0.39$ $3.96 \pm 0.15$

MIR Peak $12.8 \pm 0.3$ $29.7 \pm 0.4$ $3.94 \pm 0.15$ $5.76 \pm 0.25$ $8.50 \pm 0.35$ $1.43 \pm 0.21$ $1.64 \pm 0.08$

NGC 253 CAM FOV $16.1 \pm 1.1$ $36.3 \pm 0.3$ $4.24 \pm 0.15$ $10.2 \pm 0.2$ $10.6 \pm 0.2$ $0.55 \pm 0.12$ $1.93 \pm 0.12$

Disk $3.55 \pm 0.09$ $7.63 \pm 0.07$ $1.03 \pm 0.02$ $2.41 \pm 0.03$ $2.02 \pm 0.02$ $0.077 \pm 0.007$ $0.36 \pm 0.01$

Core $7.30 \pm 0.11$ $17.6 \pm 0.2$ $1.86 \pm 0.05$ $4.09 \pm 0.11$ $5.71 \pm 0.17$ $0.37 \pm 0.10$ $1.13 \pm 0.03$

MIR Peak $2.02 \pm 0.06$ $4.95 \pm 0.09$ $0.54 \pm 0.02$ $1.07 \pm 0.06$ $1.83 \pm 0.10$ $0.19 \pm 0.05$ $0.37 \pm 0.02$

NGC 1808 Core $4.17 \pm 0.19$ $9.39 \pm 0.12$ $1.38 \pm 0.03$ $3.32 \pm 0.08$ $3.05 \pm 0.06$ $0.096 \pm 0.016$ $0.23 \pm 0.02$

MIR Peak $1.59 \pm 0.05$ $3.74 \pm 0.08$ $0.57 \pm 0.02$ $1.31 \pm 0.05$ $1.34 \pm 0.04$ $0.039 \pm 0.008$ $0.087 \pm 0.007$

**Table 5:** Emission feature measurements^a.
Source	Region	F_6.2	F_7.7	F_8.6	F_11.3	F_12.7	F_15.6	$F_{\rm [Ar~II]}$
M 82	CAM FOV	$103 \pm 1$	$226 \pm 1$	$31.8 \pm 2.4$	$57.7 \pm 0.9$	$54.0 \pm 0.4$	$5.77 \pm 0.47$	$6.05 \pm 0.31$
	Disk	$21.5 \pm 0.2$	$46.5 \pm 0.2$	$6.70 \pm 0.08$	$13.1 \pm 0.1$	$9.74 \pm 0.06$	$0.76 \pm 0.02$	$0.67 \pm 0.02$
	Core	$43.7 \pm 0.5$	$100 \pm 1$	$13.8 \pm 0.4$	$22.0 \pm 0.5$	$26.4 \pm 0.5$	$3.49 \pm 0.39$	$3.96 \pm 0.15$
	MIR Peak	$12.8 \pm 0.3$	$29.7 \pm 0.4$	$3.94 \pm 0.15$	$5.76 \pm 0.25$	$8.50 \pm 0.35$	$1.43 \pm 0.21$	$1.64 \pm 0.08$
NGC 253	CAM FOV	$16.1 \pm 1.1$	$36.3 \pm 0.3$	$4.24 \pm 0.15$	$10.2 \pm 0.2$	$10.6 \pm 0.2$	$0.55 \pm 0.12$	$1.93 \pm 0.12$
	Disk	$3.55 \pm 0.09$	$7.63 \pm 0.07$	$1.03 \pm 0.02$	$2.41 \pm 0.03$	$2.02 \pm 0.02$	$0.077 \pm 0.007$	$0.36 \pm 0.01$
	Core	$7.30 \pm 0.11$	$17.6 \pm 0.2$	$1.86 \pm 0.05$	$4.09 \pm 0.11$	$5.71 \pm 0.17$	$0.37 \pm 0.10$	$1.13 \pm 0.03$
	MIR Peak	$2.02 \pm 0.06$	$4.95 \pm 0.09$	$0.54 \pm 0.02$	$1.07 \pm 0.06$	$1.83 \pm 0.10$	$0.19 \pm 0.05$	$0.37 \pm 0.02$
NGC 1808	Core	$4.17 \pm 0.19$	$9.39 \pm 0.12$	$1.38 \pm 0.03$	$3.32 \pm 0.08$	$3.05 \pm 0.06$	$0.096 \pm 0.016$	$0.23 \pm 0.02$
	MIR Peak	$1.59 \pm 0.05$	$3.74 \pm 0.08$	$0.57 \pm 0.02$	$1.31 \pm 0.05$	$1.34 \pm 0.04$	$0.039 \pm 0.008$	$0.087 \pm 0.007$

^a: All fluxes are expressed in $10^{-14}~{\rm W~m^{-2}}$ . The uncertainties result from the formal effective uncertainties of the relative fluxes (see Sect. 2).

Our [Ar II] 6.99 $~\mu$ m linemap globally resembles those of other tracers of ionized gas from H II regions at mid- and near-infrared wavelengths such as [Ne II] 12.81 $~\mu$ m, [Ar III] 8.99 $~\mu$ m, [S IV] 10.51 $~\mu$ m, Br $\alpha$ , Br $\gamma$ , and Pa $\beta$ (Achtermann & Lacy 1995; Satyapal et al. 1995). At a resolution of $1^{\prime\prime}$ - $2^{\prime\prime}$ , these maps reveal a rich sub-structure dominated by prominent sources $\approx$ $6^{\prime\prime}$ and 12 $^{\prime\prime}$ southwest of the nucleus and $\approx$ $6^{\prime\prime}$ to the northeast (labeled W1, W2, and E1 by Achtermann & Lacy 1995). While their intensity ratio depends somewhat on the emission line considered, W1 and W2 are together about three times brighter than E1. Radial velocity data of the [Ne II] 12.81 $~\mu$ m and Br $\gamma$ emission (Larkin et al. 1994; Achtermann & Lacy 1995) suggest that most sources reside in a nearly edge-on rotating ring at radius coinciding in projection with W1 and E1 as well as along the stellar bar at larger radii, where the most recent starburst episode took place about 5 Myr ago (e.g. Förster Schreiber et al. 2003). Within the positional uncertainties and resolution limitations, the spatial distribution of our [Ar II] 6.99 $~\mu$ m map peaks between W1 and W2 and encompasses E1, and thus traces well the youngest starburst regions.

Little differences are seen between the PAH maps but the PAH 6.2 $~\mu$ m/7.7 $~\mu$ m and PAH 8.6 $~\mu$ m/11.3 $~\mu$ m ratio maps reveal spatial variations at the 20% and 60% level, respectively. The variations are significant within the brighter emission regions along the disk, where the relative uncertainties are < $15\%$ . Structures seen towards the map edges are much less reliable as they become comparable in amplitude to the uncertainties. The PAH 6.2 $~\mu$ m/7.7 $~\mu$ m ratio is lower along the disk and reaches minima on each side of the nucleus. The overall morphology appears to curve northwards away from the nucleus and extensions are hinted at above and below the galactic plane. Similar shape and spurs are observed notably in the large-scale distribution of the molecular and ionized gas line emission and of the radio continuum emission (e.g. Shen & Lo 1995; Achtermann & Lacy 1995; Wills et al. 1999). The spatial variations in the PAH 8.6 $~\mu$ m/11.3 $~\mu$ m ratio match roughly those of the PAH 6.2 $~\mu$ m/7.7 $~\mu$ m ratio, with higher values along the disk and maxima flanking the nucleus. Noticeably, the western peak lies closer to the nucleus than the western PAH 6.2 $~\mu$ m/7.7 $~\mu$ m minimum, and the apparent curving and large-scale extensions have no counterpart in the PAH 8.6 $~\mu$ m/11.3 $~\mu$ m map.

Figure 8 compares the PAH ratio maps with the CO J: $1 \rightarrow 0$ millimetric emission of Shen & Lo (1995) convolved at the resolution of the ISOCAM maps. The overall correspondence of the PAH 6.2 $~\mu$ m/7.7 $~\mu$ m minima and PAH 8.6 $~\mu$ m/11.3 $~\mu$ m maxima with the peaks in CO emission, as well as the curved shape and northeastern extension for the PAH 6.2 $~\mu$ m/7.7 $~\mu$ m ratio, is quite striking. We believe that the observed variations in PAH ratios are mostly real. Characteristic patterns expected for ghosts are not seen (unresolved ring- or arc-like features most prominent in the presence of point-like sources). Artifacts due to the uncorrected flat field and straylight could produce extended lobes on each side of an axis at ${\rm PA} \approx 150^{\circ}$ for the M 82 data, i.e. roughly the minor axis (Biviano et al. 1998a,1998b; Okumura 2000). However, such lobes would have a much larger extent than the features seen in our PAH ratio maps and the chromatic dependence between 6.2 and 7.7 $~\mu$ m, and 8.6 and 11.3 $~\mu$ m is predicted and observed to be smaller than the measured variations at 20% and 60% levels, respectively.

3.2.2 NGC 253

In NGC 253, the emission in the continuum bands, PAH features, and [Ar II] 6.99 $~\mu$ m line is characterized by a strong peak within $1^{\prime\prime}$ - $2^{\prime\prime}$ of the nucleus, embedded in a diffuse envelope elongated along the galactic plane ( ${\rm PA} \approx 50^{\circ}$ ). The 15 $~\mu$ m continuum and [Ar II] 6.99 $~\mu$ m line distributions do not seem to extend as far as the PAH emission in the outer parts of the source. The noisy $\lambda < 6~{\rm\mu m}$ channels prevent reliable assessment of the lower level, large-scale 5.5 $~\mu$ m continuum emission. The centroids in the various images are essentially indistinguishable (differences < $0.5~{\rm pixel}$ ).

Though the images reveal little spatial structure because of the intrinsically small source size and the limited angular resolution, they are consistent with expectations at this resolution based on previously published MIR maps (mostly obtained at $\sim$ $1^{\prime\prime}$ resolution). These include broad-band images at 10.8 (N band) and 19.2 $~\mu$ m, narrow-band images at 8.5, 10.0, 12.5, and 20.2 $~\mu$ m tracing PAH and/or continuum emission, maps of the 3.29 and 11.3 $~\mu$ m PAH emission, and images of the [Ne II] 12.81 $~\mu$ m line and of the underlying continuum + PAH emission (Piña et al. 1992; Telesco et al. 1993; Keto et al. 1993,1999; Kalas & Wynn-Williams 1994; Böker et al. 1998). We note that the [Ne II] linemaps of Böker et al. (1998) and Keto et al. (1999) likely contain a contribution from PAH 12.7 $~\mu$ m emission because the observations were made at low spectral resolution ( $FWHM \approx 0.2~\mu\rm m$ ).

The spatial distribution seen in all our maps is evidently dominated by a very prominent compact MIR source, better outlined in arcsecond resolution images. Keto et al. (1999) associated this source with a bright off-nucleus super star cluster resolved by optical Hubble Space Telescope observations. Based on their data, the source has a size of $20~{\rm pc}$ (marginally resolved) and accounts for $\approx$ 20% of the total continuum emission at 12 and 20 $~\mu$ m, and 12% of the total [Ne II] 12.81 $~\mu$ m flux. From our own measurements, nearly 25% of the total continuum emission detected with ISOCAM between 13.5 and 15 $~\mu$ m originates in the 70 pc-diameter MIR peak; this fraction varies between 10% and 30% for our various broad- and narrow-band and emission feature measurements (Tables 4 and 5). NGC 253 is quite remarkable in the compactness of its main MIR emitting region, with a $FWHM \la 150~\rm pc$ while the optical disk extends over $\sim$ $20~{\rm kpc}$ . Although of a different nature, the interacting system NGC 4038/4039 offers a similar example, with 15% of its 12.5- $18~{\rm\mu m}$ luminosity being produced within a 100 pc-size star-forming knot in the overlapping region between the galaxies (Mirabel et al. 1998).

Additional small-scale structure of NGC 253 at MIR wavelengths includes a secondary much fainter peak in the 12 and 20 $~\mu$ m continuum emission nearly coinciding with the nucleus, about 2 $^{\prime\prime}$ northeast of the prominent source discussed above. The [Ne II] 12.81 $~\mu$ m line emission differs somewhat from the continuum, except possibly for the brightest peak, showing a bilobal or arc-like structure also hinted at in Br $\gamma$ images and suggestive of a circumnuclear star-forming ring (Böker et al. 1998; Engelbracht et al. 1998; Keto et al. 1999). None of these features, however, is resolved with ISOCAM.

The PAH 6.2 $~\mu$ m/7.7 $~\mu$ m and PAH 8.6 $~\mu$ m/11.3 $~\mu$ m ratio maps show variations of about 40% and nearly a factor of 2, respectively. These are statistically significant in view of the corresponding formal uncertainties of $\leq$ $20\%$ and < $30\%$ . The ratio maps differ markedly, with the PAH 6.2 $~\mu$ m/7.7 $~\mu$ m image indicating a general increase from south to north of the nucleus whereas the PAH 8.6 $~\mu$ m/11.3 $~\mu$ m ratio appears more centrally concentrated, peaking near the nucleus and slightly more extended northeast. The reality of the arc-like feature $\approx$ $10^{\prime\prime}$ northwest of the nucleus in the PAH 8.6 $~\mu$ m/11.3 $~\mu$ m map is dubious because it is barely resolved and could be due to ghost effects given the strong unresolved peak of the emission. No corresponding structure that could perhaps support this feature is seen in maps of the molecular gas emission, of the radio continuum emission, of tracers of ionized gas or even of extinction (e.g. Peng et al. 1996; Ulvestad & Antonucci 1997; Engelbracht et al. 1998). The limited region for which $S/N \geq 3$ prevents us from examining the larger-scale distribution in PAH ratios.

$\begin{figure} \par\includegraphics[width=15cm,clip]{h3938f5.ps}\end{figure}$

Figure 5: Selected ISOCAM maps of M 82. The maps are identified at the top of each panel while the physical linear scale and the FWHM of the PSF are indicated at the bottom left and right. The grayscale is linear, with white and black tones corresponding to the minimum and maximum values displayed. For the continuum and emission feature images, the solid contours are 10%, 25%, 40%, 55%, 70%, 85%, and 95% of the maximum fluxes. For the 5.5 and $15.0~{\rm\mu m}$ continuum maps, the peaks are 10.3 and $187~{\rm mJy~arcsec^{-2}}$ , respectively. For the [ArII] $6.99~{\rm \mu m}$ and the PAH 6.2, 7.7, 8.6, and 11.3 $~\mu$ m maps, they are 2.05, 12.5, 28.9, 3.84, and 5.54, respectively, in units of $10^{-16}~{\rm W~m^{-2}~arcsec^{-2}}$ . The contour levels for the ratio maps are as follows: from 10 to 18 in steps of 2 for the ${\rm 15.0~\mu m/5.5~\mu m}$ continuum ratio, from 0.42 to 0.48 in steps of 0.02 for the PAH ${\rm 6.2~\mu m/7.7~\mu m}$ ratio, and from 0.45 to 0.75 in steps of 0.10 for the PAH ${\rm 8.6~\mu m/11.3~\mu m}$ ratio. The dotted contours in the continuum and feature maps enclose the regions where data values exceed $3\sigma$ , and ratio maps are displayed for the area where this is satisfied by both images involved. The cross at relative coordinates $(0^{\prime \prime }, 0^{\prime \prime })$ indicates the location of the galaxy nucleus.

$\begin{figure} \par\includegraphics[width=14.8cm,clip]{h3938f6.ps}\end{figure}$

Figure 6: Selected ISOCAM maps of NGC 253. The maps are identified at the top of each panel while the physical linear scale and the FWHM of the PSF are indicated at the bottom left and right. The grayscale is linear, with white and black tones corresponding to the minimum and maximum values displayed. For the continuum and emission feature images, the contours are 10%, 25%, 40%, 55%, 70%, 85%, and 95% of the maximum fluxes. For the 5.5 and $15.0~{\rm\mu m}$ continuum maps, the peaks are 13.6 and $269~{\rm mJy~arcsec^{-2}}$ . For the [ArII] $6.99~{\rm \mu m}$ and the PAH 6.2, 7.7, 8.6, and $11.3~{\rm \mu m}$ maps, they are 1.50, 7.66, 19.1, 2.33, and 4.08, respectively, in units of $10^{-16}~{\rm W~m^{-2}~arcsec^{-2}}$ . The contour levels for the ratio maps are as follows: from 0.40 to 0.55 in steps of 0.05 for the PAH ${\rm 6.2~\mu m/7.7~\mu m}$ ratio, and from 0.30 to 0.60 in steps of 0.10 for the PAH ${\rm 8.6~\mu m/11.3~\mu m}$ ratio. The dotted contours in the continuum and feature maps enclose the regions where data values exceed $3\sigma$ (for the $5.5~{\rm \mu m}$ continuum map, the $3\sigma$ limit lies between the contours of 55% and 70% of maximum flux), and ratio maps are displayed for the area where this is satisfied by both images involved. The cross at relative coordinates $(0^{\prime \prime }, 0^{\prime \prime })$ indicates the location of the galaxy nucleus.

$\begin{figure} \par\includegraphics[width=10.4cm,clip]{h3938f7.ps}\end{figure}$

Figure 7: Selected ISOCAM maps of NGC 1808. The maps are identified at the top of each panel while the physical linear scale and the FWHM of the PSF are indicated at the bottom left and right. The grayscale is linear, with white and black tones corresponding to the minimum and maximum values displayed. The contours correspond to 10%, 25%, 40%, 55%, 70%, 85%, and 95% of the maximum fluxes. For the $15.0~{\rm\mu m}$ continuum map, the peak is $18.7~{\rm mJy~arcsec^{-2}}$ . For the [ArII] $6.99~{\rm \mu m}$ and the PAH 6.2, 7.7, 8.6, and 11.3 $~\mu$ m maps, they are 0.13, 1.82, 4.23, 0.71, and 1.58, respectively, in units of $10^{-16}~{\rm W~m^{-2}~arcsec^{-2}}$ . The dotted contours in the continuum and feature maps enclose the regions where data values exceed $3\sigma$ . The cross at relative coordinates $(0^{\prime \prime }, 0^{\prime \prime })$ indicates the location of the galaxy nucleus.

3.2.3 NGC 1808

The starburst in NGC 1808 covers a region of comparable physical size to that of M 82 but being three times more distant, less structural details are resolved by ISOCAM. The PAH emission appears the most extended and oriented parallel to the major axis of the galaxy ( ${\rm PA} \approx 140^{\circ}$ ), peaking at the nucleus. The 15 $~\mu$ m emission follows closely the PAH emission. The [Ar II] 6.99 $~\mu$ m line emission is the most distinct in that the peak clearly is off-nucleus, about 5 $^{\prime\prime}$ to the southeast, and the centroid of the emission region lies 2.5 $^{\prime\prime}$ southeast of that for the 15 $~\mu$ m continuum and PAH emission. The [Ar II] 6.99 $~\mu$ m map agrees well in peak position and extent with the global distribution of the most intense star-forming regions, or hot spots, as traced by H recombination lines and radio continuum emission (e.g. Saikia et al. 1990; Krabbe et al. 1994; Kotilainen et al. 1996). Due to the limited number of pixels with $S/N \geq 3$ , no useful ratio maps could be made for NGC 1808.

Previously published N band images of NGC 1808 (Telesco et al. 1993; Krabbe et al. 2001) show an overall similar morphology as our 15 $~\mu$ m continuum and PAH feature maps with, at $\approx$ $1^{\prime\prime}$ resolution, a strong point-like source at the nucleus and a southeastern extension covering the starburst regions. ISOCAM broad-band LW4 (6 $~\mu$ m) polarisation observations were presented by Siebenmorgen et al. (2001) along with a CVF spectrum of the central 25 $^{\prime\prime}$ of NGC 1808 which is essentially identical in shape to ours of the starburst core (30 $^{\prime\prime}$ aperture; Fig. 3). Siebenmorgen et al. (2001) assigned all detected features to PAHs and successfully reproduced them with PAH emission alone in their radiative transfer models. As also noted by these authors, the contribution of nebular gas emission lines blended with PAH features is uncertain for NGC 1808, and we cannot constrain it using, e.g., SWS data as for M 82 and NGC 253. However, our [Ar II] 6.99 $~\mu$ m map clearly differs from those of the PAH emission and is consistent with the spatial distribution of the bulk of H II regions, supporting the idea that the 7 $~\mu$ m feature is indeed dominated by the [Ar II] line instead of PAHs.

$\begin{figure} \par\includegraphics[width=5.6cm,clip]{h3938f8.ps}\end{figure}$

Figure 8: Comparison of ISOCAM PAH ratio maps of M 82 with the distribution of CO J: $1 \rightarrow 0$ millimetric emission. The PAH ${\rm 6.2~\mu m/7.7~\mu m}$ and PAH ${\rm 8.6~\mu m/11.3~\mu m}$ maps (top and bottom panels, respectively) are displayed as grayscale images as for Fig. 5. The contours show the CO J: $1 \rightarrow 0$ emission from the map of Shen & Lo (1995) convolved at the resolution of the ISOCAM images (from the original $2.5^{\prime \prime }$ to $5.6^{\prime \prime }$ ). The physical linear scale and the FWHM of the ISOCAM PSF are indicated at the bottom left and right. The cross at relative coordinates $(0^{\prime \prime }, 0^{\prime \prime })$ marks the location of the galaxy nucleus.

3 Results

3.1 Mid-infrared spectra and feature identification

3.2 Mid-infrared images: Mapping the features

3.2.1 M 82

3.2.2 NGC 253

3.2.3 NGC 1808

3.3 Salient features