2 The observations

Figure 1: TIMMI2 image of $\eta$ Car in the 11.9 $\mu$m narrow band filter. We applied a square root function to enhance the intensity contrast. The "+'' indicates the central source.

We observed $\eta$ Car between 3 and 6 March 2001, with the TIMMI2 (Reimann et al. 2000; Käufl et al. 2000) mid-IR camera attached to the ESO 3.6 m telescope at La Silla, Chile. The camera is equipped with a 320$\times$240 pixel array; we applied a pixel scale of 0.2 arcsec/pixel in both the N and the Q band. We used a chop throw of 22 arcsec north-south and a nod of 30 arcsec east-west. This allowed for both the chopped as well as the nodded positions to fall onto the detector. In the final images, we combined the positive and negative images resulting from the chopping and nodding positions. $\eta$ Car was observed in the M band (4.8 $\mu$m), at several wavelengths in the Nband using narrow-band filters (7.9, 11.9, 12.9 $\mu$m), and in the Qband (20 $\mu$m). The observing conditions were variable with a high humidity and seeing ranging between 0.3 and about 1 arcsec. Apart from the images, a number of long-slit N-band spectra covering the entire nebula were taken, as well as a few Q-band spectra. In this paper, we concentrate on the 7.9, 11.9, 12.9 and 20 $\mu$m images. We will report on the other images, as well as on the spectroscopy in a subsequent study.

The images were reduced using a shift-and-add technique, where the shift between individual frames was determined from a least-squares comparison between the images. The resulting images were then deconvolved using an empirically determined point spread function obtained by observing $\gamma$ Cru with the same set-up as $\eta$ Car. For the Q band image we have no good empirical point spread function. However, given that in the Q band TIMMI2 at the 3.6 m is diffraction-limited, we have adopted a Gaussian beam of 1.25 arcsec to deconvolve the Q band image.

We show the final images in Figs. 1 and 2. Figure 3 shows a temperature map, derived assuming that the total flux in the images is equal to that seen in the ISO-SWS spectrum (Mor99). We realize that the variability of $\eta$ Car may introduce errors in this calibration. However, we verified the calibration at 7.9 and 11.9 $\mu$m using $\gamma$ Cru and found values of $1.6\times10^4$ and $5.2\times10^4$ Jy. The agreement between the two methods is within 10-20 per cent. We stress that unless the shape of the spectrum of $\eta$ Car has changed considerably between 1996 and 2001, our method should result in reasonable estimates of the temperature. Note that the strong 10 $\mu$m silicate band (e.g. Mor99) may introduce some errors ($\sim$15%) in the temperature map.