5 Optical observations  

Since 1985, 3C 273 is observed about once a week (when the Moon is not too bright) during its visibility season, from end of December to beginning of July with the Swiss telescope at La Silla, Chile. Our database contains presently 376 observations made in the seven-colour Geneva photometric system from 1985 to 1997: U (3439Å), B₁ (4003Å), B (4213Å), B₂ (4466Å), V₁ (5395Å), V (5479Å) and G (5798Å). The quoted wavelength are effective wavelength calculated with the spectrum of 3C 273. The stability of the Geneva photometry is extremely good, and the uncertainty on the magnitudes can be as small as 0.01, which is about 1% in flux.

The response functions of the seven filters to an equiphotonic flux are shown in Fig. 3. The B and V filters are broad-band filters (width $\sim$300 Å) relatively close to the corresponding Johnson's filters. The other filters are narrower (width $\sim$180Å). The U filter is comparable to Johnson's U filter, apart from the fact that its "red'' wing is cut below the Balmer discontinuity (3647Å). B₁ and B₂ are roughly the "blue'' and the "red'' parts of B, and V₁ and G are roughly the "blue'' and the "red'' parts of V.

Broad-band photometry has the disadvantage of including emission-line contamination in the continuum measurements. The observed position of the main emission-lines in 3C 273 is shown in Fig. 3. Unfortunately, the flux measurements in all filters are contaminated by line emission. Using the line parameters given by Wills et al. (1985), we calculated that the MgII $\lambda$2798 line contributes by 2.8% in the U filter and that the broad H$\beta$ line contributes respectively by 9.1%, 9.7% and 17.1% in the V₁, V and G filters. The other lines are expected to contribute much less to the V₁, V and G filters. The contamination in the U, B₁, B and B₂ filters due to the FeII pseudo-continuum and to the Balmer lines is difficult to estimate. Since the ultraviolet emission-lines (Ly$\alpha$ and CIV $\lambda$1549) are nearly constant in 3C 273 (Ulrich et al. 1993; Türler & Courvoisier 1998), we do not expect significant variations of the optical lines. Therefore, the optical flux variations are expected to be nearly pure continuum variations.

The normal photometric reduction produces magnitudes in the seven filters, which are transformed into fluxes following the calibration of Rufener & Nicolet (1988) (Eq. (6) with the $E_{\nu}$constants of Table 8). The uncertainties have been estimated by taking two consecutive observations in about 40% of the nights; the average of the deviation in flux during these nights provided the uncertainties of the order of 1%, even in the less sensitive filters. The systematic uncertainty introduced by the absolute calibration has been estimated to be about 5 times the photometric accuracy, i.e. of the order of 5% (Rufener & Nicolet 1988).

Other optical observations in the UBV filters from the literature were included, but only when they satisfy the following criteria. We included only photo-electric measurements, since the photographic observations have usually much greater uncertainties (see the B band light curve of 3C 273 from 1887 to 1980 shown by Angione & Smith 1985). We considered only sets of magnitude measurements made with the same instrument and having at least a few observations during the period from 1970 to 1985. It means that we did not include isolated magnitudes, as well as sets of observations that are contemporaneous with the Geneva photometric observations, since they would only add artificial scatter to the well sampled light curves due to the use of different instruments, filter passbands and calibrations.

From all references given by Belokon' (1991), the five following sets of magnitudes satisfy our criteria. The two main sources of data are those of Burkhead (Burkhead 1969, 1980; Burkhead & Hill 1975; Burkhead & Lee 1970; Burkhead & Parvey 1968; Burkhead & Rettig 1972; Burkhead & Stein 1971) and the Crimean observations of Lyutyi (Lyutyi 1976; Lyutyi & Metlova 1987), which cover respectively the periods from 1968 to 1979 and from 1971 to 1986. The uncertainty of each measurement in these two data sets was assumed to be either 0.03 or 0.05 magnitude depending on the remarks in Burkhead's papers and the presence of a ":'' sign in Lyutyi's papers. The other sets of magnitudes we considered are those from Mount Lemmon (Cutri et al. 1985; Elvis et al. 1994; O'Dell et al. 1978; Smith et al. 1987), from Turku (Sillanpää et al. 1988), and the magnitudes from Las Campanas (Impey et al. 1989) with uncertainties not exceeding 0.1 magnitude.

  

Table 3: List of the magnitude corrections applied to the UBV sets of magnitudes. The number N of nearly simultaneous ($\Delta\,t\leq 5$days) observation pairs we used is given in parentheses

Since the Geneva photometry set of observations is the most complete and the most accurate, we chose to rescale the other sets of magnitudes according to the Geneva photometric system. Similarly to Belokon' (1991), we adjusted each set of magnitudes by comparing pairs of nearly simultaneous observations ($\Delta\,t\leq 5$days). We first determined with 35 to 40 such pairs of observations taken mainly during 1986, that the Crimean magnitudes are consistent (average deviation smaller than 0.01 magnitude) with the Geneva photometry flux densities if their zero-magnitude fluxes are 1700Jy (U), 3900Jy (B) and 3600Jy (V). These two sets of observations being consistently linked by these values, we corrected the other sets of magnitudes according to this combined reference data set. The magnitude corrections we applied are given in Table 3, together with the number N of observation pairs we used. We then converted the Crimean original magnitudes and the other corrected magnitudes to flux densities using the zero-magnitude fluxes given above. Finally, we completed the UBV light curves by including a few isolated flux densities reported by Landau et al. (1983, 1986) and Sadun (1985).

  

Figure 4: Four characteristic light curves of the infrared-to-X-ray behaviour of 3C 273 from 1974 to 1998. Three flares are clearly visible in the H and V band light curves in 1983, 1988 and 1990. The dashed line is the contribution from the host galaxy (see Sects. 4 and 5). In the 1300Å ultraviolet light curve, we indicated by crosses the six small aperture IUE observations and by stars the 11 other dubious spectra (see Sect. 6). We added to the 2keV light curve the BATSE 20-350 keV light curve (solid line), which was rebinned into 0.1 year bins and was extrapolated to 2keV assuming a power law of spectral index $\alpha =0.6$

The obtained V band light curve is shown in Fig. 4. The contribution from the host galaxy in the V band is only 1.0mJy. This value was obtained from the host galaxy's V magnitude of 16.4 (Bahcall et al. 1997) by using the V band zero-magnitude flux of 3647Jy given by Elvis et al. (1994). The relative contribution of the stars in the host galaxy is only about $3-4\%$ in the V band, which is about four times less than in the H band (see Sect. 4 and Fig. 6).

The database contains also observations from the literature in the R (7000Å) and I (9000Å) filters, which are not included in the Geneva photometry. The R and I magnitudes added to the database are from Cutri et al. (1985), Elvis et al. (1994), Hamuy & Maza (1987) (24$\hbox{$^{\prime\prime}$}$aperture), Impey et al. (1989), Moles et al. (1986), O'Dell et al. (1978), Smith et al. (1987), Takalo et al. (1992) and Valtaoja et al. (1991). All these magnitudes were converted into flux densities using the zero-magnitude fluxes given by Elvis et al. (1994), which are 2791Jy (R) and 1871Jy (I). We also added the R and I band flux densities of 3C 273 reported by Landau et al. (1983, 1986), Lichti et al. (1995), Sadun (1985) and von Montigny et al. (1997).