3. Near-IR spectroscopy of the lens
3.1. Plausible lens redshift
The galaxy spectrum is shown in Fig. 2. The signal-to-noise ratio is very low, so the spectrum has been smoothed with a box car with 200 Å width. Also plotted are the broadband magnitudes of the lens (C98, R98 and Hjorth, private communication). The lens spectrum is scaled to match the H and K band magnitudes.
The spectrum does not lead to a redshift measurement. However, the broadband colours suggest a significant break in the spectrum between the I and J bands. We have used the publicly available photometric redshift code hyperz (Bolzonella et al. 2000) to estimate the redshift of the lens.
Since there is little evidence for dust in the quasar spectrum (see below), we have fitted dust free models to the data. The best fitting model is a galaxy that was formed in a single burst of star formation. The best fitting redshift is with a 1-sigma range of 0.8 to 1.2. The age of the burst is 1.7 Gyrs. The quoted errors on the redshift do not include systematic errors that could be due to the heterogenous nature of the photometric data, which is derived from a mixture of ground and space based observations.
The estimated redshift is slightly higher than those estimated from the position of the lens on the fundamental plane (; Kochanek et al. 2000) or from lens models and the time delay (; W98). Note however that models including a dark component (see next section) can cope with any redshift between 0.7 and 1.3 and reproduce the observed time delay, assuming for example H0=60 km s-1 Mpc- 1.
As the break between 0.7 and 1.0 µm in the model spectra is very strong, a deep spectrum in this region is probably the key to accurately measure its redshift.
3.2. Evidence for a high redshift cluster-lens?
With only two quasar images available to constrain the lensing potential, the unusual image configuration of HE 1104-1805 (R98) is very difficult to model uniquely. The system can not be reproduced with a Singular Isothermal Sphere. Additional shear and convergence, whatever their origin may be (intrinsic ellipticity of the lens and/or intervening lenses), are required to match simultaneously the positions and flux ratio of the quasar images (=2.8, see Sect. 4) and the measured time delay. In a first model, we introduce an ellipticity in the lens model, i.e., we choose an isothermal ellipsoid (first column in Table 1). Although we can easily obtain a good fit, the resulting model has a very large velocity dispersion, over 300 km s-1, and an unrealistic ellipticity compared with the ellipticity of the associated light distribution. Finally, such a model predicts time delays of days, while the observed value is 265 days, according to W98.
Table 1. Model parameters. Parameters indicated within bracket, as well as the (x,y) position of the main galaxy, correpsond to fixed parameters for the fit. In both cases, the position and flux ratio of 2.8 for the 2 images were used as constraints. In the case of the Galaxy+Cluster model the time delay t= 265 days was added as a supplementary constraint.
The uncertainty on the lens redshift can not explain the discrepancy between the measured and predicted time delays: additional mass is required to describe the image configuration, flux ratio and time delay. We therefore adopt a two component model including (1) the main lensing galaxy, with ellipticity and position angle as constrained by the light distribution of the main galaxy lens, and, (2) a more extended component mimicking an intervening galaxy cluster. For simplicity, we centered the cluster on the main galaxy and assume an elliptical isothermal mass profile with a core radius. The fitted parameters were only the velocity dispersion, the ellipticity and position angle. Both the main lens and cluster components are taken to be at redshift 1.0. Our best fit model is shown in Fig. 3 and the model parameters are summarized in Table 1. It involves a cluster with moderate mass, i.e., a velocity dispersion of km s-1. The error bar given here is obtained by varying the velocity dispersion of the cluster while keeping all the others fixed, and by running the models with different velocity dispersions for the main lens (all lower than 300 km s-1). This error bar is probably underestimated but still allows to show that 1- one need a cluster to model the system and 2- this cluster has a low velocity dispersion. The cluster ellipticity is 0.27 (defined as ]) and PA=4 degrees, which is slightly tilted relative to the axis of the main lens. The PA of the main lensing galaxy also disagree with that measured in the HST images of Lehar et al. (1999), who gives PA=6317 degrees. However, the discrepancy is much smaller than in the single lens model. Adding the cluster component therefore allows one to match better the observed shape parameters of the main lensing galaxy and to derive more realistic velocity dispersions. With the presence of the cluster, the models can accommodate a PA of 46 degrees for the main lensing galaxy and a lower velocity dispersion, 235 km s-1.
If we assume H km s-1 Mpc- 1, and , the galaxy+cluster model reproduces well the measured time delay, giving a value of days. However, we stress that the lens redshift and the velocity dispersion of the cluster are redundant parameters: increasing the cluster's mass or decreasing the lens redshift have the same effect on the time delay. This degeneracy between the two parameters will prevent any estimate of H0 until the redshift of the main lens is measured and more observational constraints are available on the "extended" cluster component of the lensing matter. The time delay now available in HE 1104-1805 can therefore be seen as a new important constraint on the lens model, indicating the presence of a shallower mass component that could be related to a yet undetected cluster, rather than a way to constrain H0. The mass within an area of a given radius is shown in Fig. 4 for different model components.
If real, the cluster we predict in our model is difficult to detect, with only km s-1. At a redshift of 1, it would be even more difficult to see than the more massive clusters involved in other lenses such as AX J2019+112 (e.g., Benitez et al. 1999) or RX J0911.4+0551 (Burud et al. 1998, Kneib et al. 2000). However the velocity dispersion of such a cluster will change depending on the cluster center position. If it is not aligned with the main galaxy lens, its velocity dispersion will increase significantly. Deep X-ray observations and/or deep IR images of this field would be invaluable in constraining further the models.
© European Southern Observatory (ESO) 2000
Online publication: August 23, 2000