6 Ultraviolet observations  

3C 273 was already observed by the International Ultraviolet Explorer (IUE) satellite in the first month of the mission. As soon as the existence of important variability was established (Courvoisier & Ulrich 1985), a long-term monitoring campaign was launched, which started in 1986 and went on up to the end of the mission in 1996. The spectra have been taken in the low-dispersion mode, and the usual observation rate was once every 2-3 weeks during two annual observation periods of about 3 months.

During the whole IUE mission, and taking into account only the low-dispersion observations, we have collected 256 short wavelength (SWP: 1150-1980 Å) spectra and 212 long wavelength (LWP or LWR: 1850-3350 Å) spectra. The spectra presented here are the IUE newly extracted spectra (INES) that were taken from the INES access catalogue publicly available on the WWW at: http://ines.vilspa.esa.es/ines/. This site contains also an important documentation including the quality flag description. The INES flux extraction algorithm (Rodríguez-Pascual et al. 1998) was built to correct some problems found in the spectra of the IUE final archive (IUEFA) reduced with the NEWSIPS software. The comparison of IUE flux extraction by INES and NEWSIPS shows that the INES spectra are generally more reliable than the NEWSIPS spectra in difficult conditions (Schartel & Skillen 1998).

  

Figure 5: a and b. The average IUE spectrum of 3C 273 in the range $\lambda_{\mathrm{\,obs}}$1150-3295 Å. a) the logarithmic $F_{\nu}$ vs. $\nu$representation. b) the $F_{\lambda}$ vs. $\lambda$ representation with the continuum bands from which the UV light curves were extracted

The ultraviolet (UV) continuum light curves (see Table 2) were extracted from eight 50Å continuum bands centered at $\lambda_{\mathrm{\,obs}}$1300Å, 1525Å, 1700Å, 1950Å, 2100Å, 2425Å, 2700Å and 3000Å (see Fig. 5). We considered only the wavelength bins with a quality flag of zero. The spectra having no such bin in a continuum band are therefore not included in the corresponding light curve (e.g. SWP01365LL).

The flux density $F_{\nu}$ and its uncertainty $\Delta\,F_{\nu}$ in a 50Å band were defined as  

$$F_{\nu}=\frac{1}{n}\sum_{i=1}^{n}F_{\nu_i}\quad\mbox{and}\qu... ...\,F_{\nu}=\frac{x}{n}\sqrt{\sum_{i=1}^{n}(\Delta\,F_{\nu_i})^2}$$ (1)

where $F_{\nu_i}$ and $\Delta\,F_{\nu_i}$ are respectively the INES flux density and its uncertainty at the frequency $\nu_i$, and x is a correction factor. The factor x is introduced to ensure that the average uncertainty $\overline{\Delta\,F_{\nu}}$ corresponds to the experimental value derived from all pairs of observations taken within one day. The obtained values of x reported in Table 4 are not simply equal to one, because the bins used to calculate the fluxes are not independent from each other due to the oversampling of the IUE spectra. However, by calculating x for a single INES bin, we obtain on average over the eight values given in Table 4 a value of 0.97$\pm$0.36, which shows that the INES uncertainties for a single bin are in good agreement with the experimental values derived from consecutive spectra. Our estimation of the uncertainties gives average relative values, $\overline{\Delta\,F_{\nu}}/\overline{F_{\nu}}$, around $2-3\%$ in all eight 50Å bands.

  

Table 4: Values of the correction factor x in Eq. (1) determined for a 50Å band and a single INES bin at the observed wavelength $\lambda_{\mathrm{\,obs}}$

In Fig. 4 we show the 1300Å UV light curve extracted from the IUE spectra at $\lambda_{\mathrm{\,obs}}$1275-1325 Å. It can be seen that the six SWP small aperture spectra (namely, SWP01492LS, SWP01498LS, SWP01509LS, SWP01655LS, SWP02100LS and SWP29775LS) all lie well below the light curve drawn only with the large aperture observations. The same occurs with the two LWR small aperture spectra (LWR01447LS and LWR04470LS), clearly illustrating that, even for point sources, the small aperture of IUE gives unreliable fluxes due to the relatively large size of the IUE point spread function compared to the size of the small aperture.

Other points lie well outside the general UV light curves. In three spectra (SWP49812LL, SWP49813LL and LWP27215LL) no significant flux is detected, because of a technical problem. Other spectra show a characteristic 3C 273 spectrum, but with a normalization very different from the typical flux at this period. Most of these spectra have a problem which is reported in the INES header of the spectrum or at least in the SILO file. SWP38038LL and SWP53305LL are out of aperture. SWP43550LL, SWP46633LL, SWP46647LL, SWP46650LL, LWP22191LL, LWP24638LL and LWP24657LL had no tracking during exposure and thus drifted out. LWP27045LL is contaminated by the solar spectrum and LWP24741LL had no guiding and is most probably also contaminated by scattered solar spectrum.

Three other SWP spectra and one LWP spectrum (LWP21967LL) are dubious, although no problem has been reported. The SWP spectra SWP44732LL, SWP44733LL and SWP56622LL have a Ly$\alpha$ emission-line flux which is respectively only 82%, 60% and 73% that of a comparison spectrum obtained at nearby epoch. It suggests that there is a normalization problem with these spectra, and not an unusually fast variability, since the Ly$\alpha$ line was observed to remain constant within $\pm$8% of the average line flux from 1978 to 1992 (Ulrich et al. 1993). This argument cannot be applied to the LWP spectrum LWP21967LL, but it was found to have a flux about 20% lower than that of the following spectrum obtained only three hours later.

All the spectra discussed above are flagged in the UV light curves by a non-zero value of "Flag''. Small aperture spectra are flagged with a value of -2, whereas other spectra with problems or found to be dubious are flagged with a value of -1. The average IUE spectrum of 3C 273 shown in Fig. 5 was constructed by excluding these spectra, but without taking into account the INES quality flag of each wavelength bin.