Astron. Astrophys. 360, 853-860 (2000)

4. Near-IR spectroscopy of the source at z = 2.319

The 0.95-2.50 µm spectrum of the quasar pair is shown in Fig. 5. The spectra are on a relative flux scale. Regions of high atmospheric absorption are set to zero. The spectra show clearly the Balmer lines: H, H, H and a partially obscured H, the [OIII] doublet and several broad FeII features (Francis et al. 1991). From the Balmer lines, the redshift is 2.323 for the brighter component (component A) and 2.321 for the fainter (component B). The measurement error is z, so the redshift of the two components agree with each other, but are slightly larger than the determination at optical wavelengths (, Smette et al. 1995). As with most quasars (McIntosh et al. 1999b), the [OIII] doublet is slightly blue shifted () with respect to the Balmer lines.

Fig. 5. One dimensional near-IR spectra of components A and B of HE 1104-1805.

Following Wisotzki et al. (1993), we subtracted a scaled version of the fainter component from the brighter one, that is . The scale is set so that the Balmer lines vanish. We find that we require for the red spectrum and for the blue spectrum. Wisotzki et al. and Smette et al. (1995) have used . The slight difference between Wisotzki's value and ours may only reflect systematic differences in the way the object was observed and the way the data were reduced rather than anything real. For example, the IR observations were done with a one-arc-second slit, and any small error in the alignment angle could cause such a difference.

The difference spectra are plotted in Fig. 6. Here we plot the raw difference spectra as the dotted line, and a smoothed version of this as the continuous line. The spectrum of the brighter component is also displayed. The difference spectra are featureless. The residual after subtracting the strong H line is less than 1%. Not only are the broad hydrogen features removed from the spectra, but the broad iron features and the [OIII] doublet are removed as well. As noted by Wisotzki et al. (1993) there appears to be excess continuum in the brighter component.

Fig. 6. Difference spectra of the two quasar images for the blue grism (top) and red grism (bottom). The raw difference spectra are plotted as the dotted lines, a smoothed version of these as the continuous lines. The spectra of the brighter component is also displayed.

4.1. Extinction

The Balmer decrement is around 4 for both components, and this is well within the range expected for unreddened quasars (e.g., Baker et al. 1994). Thus, there is no evidence for absolute reddening. However, the limits we can set on this are weak as the range of values for the Balmer decrement in quasars is rather broad.

The limits for differential reddening are considerably stronger. The ratio of the emission lines in the brighter and fainter components is . The error brackets the measured variation of this ratio over time (six years of observations) and over wavelength. It is not clear if this variation is real or the result of measurement error. This ratio is remarkably constant over a large wavelength range, from CIV at 1549 Å to H, and we can used it to place an upper bound on the amount of differential extinction between the two components. If we assume that the lens is at and if we assume that the standard galactic extinction law (Mathis 1990) is applicable, then the differential extinction between the two components is magnitudes.

Recently, Falco et al. (1999) measure a differential extinction of for HE 1104-1805 in the sense that the B component has a higher extinction. However, their measurements rely on broad band photometry and their results can be mimicked by chromatic amplification of the continuum region by microlensing. If we were to repeat the experiment by comparing the relative strength of the continuum at 1.25 µm and 2.15 µm, we would derive a differential extinction of magnitudes and we would find also that B component was differentially reddened.

4.2. Emission line properties of the source

The emission line properties of high redshift quasars have been examined for correlations between line ratios and equivalent widths (McIntosh et al. 1999a,b; Muramaya et al. 1999). As the signal-to-noise ratio and spectral coverage of our IR data are considerably better, we have re-measured the emission line parameters for HE 1104-1805.

Fitting of the spectrum was done in a similar way to that used in McIntosh et al. (1999a), but with extended spectral coverage. The model spectrum is a sum of Gaussian lines superposed on an exponential continuum to which is added a numerical optical FeII template (4250 Å and 7000 Å). The iron template consists of the optical spectrum of I Zw 1 obtained by Boroson & Green (1992). Before computing the model spectrum the template is smoothed to the resolution of our observations by convolving it with a Gaussian line which has the same FWHM than the broad emission lines of the quasar (CIV, in the present case), i.e., a rest-frame width of 6400 km s^-1, or 14 Å. A systemic redshift of is determined from the [OIII] 5007 emission line and applied to the data to obtain a rest-frame spectrum (multiplied by to conserve flux). As the positions of all other emission lines are redshifted relative to the [OIII] line by different amounts, their wavelengths are adjusted independently of each other. We measure a mean redshift of from the H 4340, H 4861, and H 6562 emission lines. We used a sum of Gaussians to fit each Balmer line. This arbitrary decomposition is certainly not aimed at being representative of any physical model but still allows us to measure fluxes. One single Gaussian was used to fit the H line while two Gaussians are required to fit H and three to fit H which shows very wide symmetrical wings. The [OIII] doublet is represented by two Gaussians with a fixed line ratio of three (between [OIII] 4959 and [OIII] 5007). All line widths are fixed during the fit and all intensities are adjusted simultaneously (with the conjugate gradient algorithm) with the strength of the iron template and exponential continuum. The results of the fit are reported in Table 2 and Fig. 7. The best fit spectrum has a power law continuum of the type , with . One-sigma errors were estimated by running the fit with different line widths. In addition to these errors, one should consider the error introduced by the continuum determination. Changing the index of the exponential continuum by 10% can affect iron flux measurement by up to 20%. The other, much narrower emission lines, are less affected, but we stress the need for continuum fitting over a very wide wavelength range in order to minimize such systematic errors. This was pointed out by Murayama et al. (1999). It is now obvious on our better data.

Fig. 7. Left: The top left panel displays the rest-frame spectrum of component A and its fit as described in the text. It is the entire spectrum (multiplied by ), with the model spectrum superposed as a bold line. The lower two left panels are both zooms in the left and right halves of the top panel. The vertical dashed lines are drawn at the rest-frame wavelength of each emission line, considering a systemic redshift of . Right: A zoom in the optical FeII region, again with the fitted spectrum superposed. The bottom panel is the difference between the data and the fit, in units of the photon noise.

Table 2. Rest-frame emission line properties of HE 1104-1805, as measured from the fit performed in Sect. 4.2.

The quality of the fit is overall very good; however, there are some regions where significant differences exist, as indicated by the residuals shown in the bottom left panel of Fig. 7. Most notably the FeII complex red-wards of [OIII] is relatively stronger that the FeII complex blue-wards of H.

The ratio of the EWs of [FeII] to H is 0.18. This is slightly lower than that measured by McIntosh et al. (1999a), who report . The difference is probably not significant but there are two systematic biases that make a direct comparison difficult. Firstly, the continuum in this fit is well determined, whereas the small spectral coverage of the previous work may mean that EWs are underestimated (Muramaya et al. 1999). Secondly, and more fundamentally, EWs measured in lensed quasars are susceptible to microlensing which preferentially amplifies the continuum rather than the larger emission line regions. In fact, the continuum for unlensed quasars also varies. This means that EWs are a poor measure to use. Line fluxes are not susceptible to continuum variations, no matter how the continuum varies, whether it is intrinsic to the AGN or caused by microlensing. From our fit, we measure F(FeII(4810-5090)) / F(H) = 0.18  0.04 and F(FeII(4434-4685)) / F(H) = 0.32  0.04 which, according to Lipari et al. (1993) makes of HE 1104-1805 a rather weak FeII emitter.

© European Southern Observatory (ESO) 2000

Online publication: August 23, 2000
helpdesk.link@springer.de