2 Observations and data reduction

We observed Cygnus X-3 with the Infrared Space Observatory (ISO, see Kessler et al. 1996) on April 7, 1996 corresponding to JD 2 450 180.8033 to 2 450 180.8519. The subsequent observing modes were: ISOCAM imaging photometry at 11.5 $\mu$m (LW10 filter, bandwith 8 to 15 $\mu$m), ISOPHOT-S spectrophotometry in the range 2.4-12 $\mu$m, for 4096 s, covering the orbital phases 0.83 to 1.04 (according to the parabolic ephemeris of Kitamoto et al. 1992); ISOPHOT multi-filter photometry at central wavelengths 3.6, 10, 25 and 60 $\mu$m. Observing modes and observation times are summarized in Table 1. Preliminary results were presented in Koch-Miramond et al. (2001).

2.1 ISOPHOT-S data reduction

A low resolution mid-infrared spectrum of Cygnus X-3 was obtained with the ISOPHOT-S sub-instrument. The spectrum covered the 2.4-4.9 and 5.9-11.7 $\mu$m wavelength ranges simultaneously with a spectral resolution of about 0.04 and 0.1 $\mu$m, respectively. The observation was performed in the triangular chopped mode with two background positions located at $\pm$120'', and with a dwelling time of 128 s per chopper position. The field of view is $24\hbox{$^{\prime\prime}$ }\times 24\hbox{$^{\prime\prime}$ }$. The whole measurement consisted of 8 OFF1-ON-OFF2-ON cycles and lasted 4096 s.

The ISOPHOT-S data were reduced in three steps. We first used the Phot Interactive Analysis (PIA, Gabriel et al. 1997) software (version 8.2) to filter out cosmic glitches in the raw data and to determine signals by performing linear fits to the integration ramps. After a second deglitching step, performed on the signals, a dark current value appropriate to the satellite orbital position of the individual signal was subtracted. Finally we averaged all non-discarded (typically 3) signals in order to derive a signal per chopper step. Due to detector transient effects, at the beginning of the observation the derived signals were systematically lower than those in the consolidated part of the measurement. We then discarded the first $\sim$800 s (3 OFF-ON transitions), and determined an average [ON-OFF] signal for the whole measurement by applying a 1-dimensional Fast Fourier Transformation algorithm (for the application of FFT methods for ISOPHOT data reduction see Haas et al. 2000). The [ON-OFF] difference signals were finally calibrated by applying a signal-dependent spectral response function dedicated to chopped ISOPHOT-S observations (Acosta-Pulido & Ábrahám 2001), also implemented in PIA.

In order to verify our data reduction scheme (which is not completely standard due to the application of the FFT algorithm) and to estimate the level of calibration uncertainties, we reduced HD 184400, an ISOPHOT standard star observed in a similar way as Cygnus X-3. The results were very consistent with the model prediction of the star, and we estimate that the systematic uncertainty of our calibration is less than 10$\%$.

2.2 ISOPHOT spectral energy distribution

Figure 1: Observed spectrum of Cygnus X-3 in the 2.4-12 $\mu$m range, obtained with the FFT method.

The observed spectral energy distribution is shown in Fig. 1. The observed (not dereddened) continuum flux in the range 2.4-7 $\mu$m is $20\pm10$ mJy in good agreement with that observed by Ogley et al. (2001) with ISOCAM on the same day (the dereddened fluxes are shown in Fig. 2) ; the observed flux decreases to about $10 \pm8$ mJy around 9 $\mu$m.

An unresolved line is observed at about 4.3 $\mu$m peaking at 57 $\pm$ 10 mJy. The linewidth is 0.04 $\mu$m, consistent with the instrumental response and corresponding to $\sim$2500 km s^-1. Note that the measured line flux might be underestimated because the ISOPHOT-S pixels are separated by small gaps, and a narrow line might falls into a gap.

2.3 ISOPHOT-P data analysis and results

The data reduction in the multi-filter mode was performed using the Phot Interactive Analysis (Gabriel et al. 1997) software. After corrections for non-linearities of the integration ramps, the signal was transformed to a standard reset interval. Then an orbital dependent dark current was subtracted and cosmic ray hits were removed. In case the signal did not fully stabilize during the measurement time due to the detector transients, only the last part of the data stream was used. The derived flux densities were corrected for the finite size of the aperture by using the standard correction values as stated in the ISOPHOT Observer's Manual (Klaas et al. 1994). The flux detected at 3.6 $\mu$m (bandwidth 1 $\mu$m) in the 10'' diameter aperture is 8.1 $\pm$ 3.3 mJy at a confidence level of 2.4 $\sigma$. No detection above the galactic noise was obtained at 10, 25 and 60 $\mu$m with 23'', 52'' and 99'' diameter aperture, respectively.

2.4 ISOCAM data reduction and results

The LW10 filter centered at 11.5 $\mu$m was used with the highest spatial resolution of $1.5''\times1.5''$ per pixel. The ISOCAM data were reduced with the Cam Interactive Analysis software (CIA) version 3.0, following the standard processing outlined in Starck et al. (1999). First a dark correction was applied, then a de-glitching to remove cosmic ray hits, followed by a transient correction to take into account memory effects, using the inversion algorithm of Abergel et al. (1996), and a flat-field correction. Then individual images were combined into the final raster map, whose pixel values were converted into milli-Jansky flux densities. No colour correction was applied. A point source is clearly visible at the Cygnus X-3 position on the ISOCAM map at 11.5 $\mu$m. The measured flux is $7.0\pm2.0$ mJy above a uniform background at a level of about 1.2 mJy, in good agreement with our ISOPHOT result. This flux is lower than the $15.2\pm1.6$ mJy (at 11.5 $\mu$m) measured by Ogley et al. (2001), on the same day, using ISOCAM/LW10 with a $6''\times 6''$ aperture.

The high resolution configuration of the ISOCAM camera has been used to constrain the spatial extension of the infrared source. The measured FWHM for the source is $3.90\pm0.45$ arcsec (mean value of the four individual images composing the final raster map). This can be compared to the ISOCAM catalogued point spread functions at these energy and configuration which show a FWHM mean value of $3.44\pm0.45$ arcsec, including the effects of the satellite jitter and of the pixel sampling. The slightly larger value for Cygnus X-3, though only marginally significant, might therefore indicate an extended source. The deconvolved extension would be $1.84 \pm 0.64$ arcsec which at a distance of 10 kpc corresponds to a linear extension of $\sim$ $2.7\times10^{17}$ cm. The extended infrared source may be the result of the heating of the surrounding medium by the radio jets whose existence have been now clearly demonstrated both at arcsec (Martí et al. 2000, 2001) and sub-arcsec (Mioduszewski et al. 2001) scales, but it clearly deserves confirmation.