Astron. Astrophys. 346, 359-368 (1999)

3. Photometric and spectroscopic study of H3 and H5

3.1. Lensing properties, spectral energy distribution and morphology of the sources

Hereafter we use the new lens model of A2390 by Kneib et al. (1999), which is a refined version of the earlier model presented in Pierre et al. (1996), based on lensing and X-ray data, taking into account the new constraints given by the HST images. According to this lens modelling, H3 and H5 are multiple images of two high-redshift sources at . This redshift estimate has been confirmed through photometric redshift techniques (filters from B to K) and spectroscopy. Fig. 1 displays a zoom on the different components of H3 and H5, as well as their location in the cluster. H3a-b-c is an impressive cusp arc showing several bright knots which are identifiable in each different image. H5 is a fold arc with the two radial components showing similar morphologies. A third faint image is predicted for this source, but unfortunately it lies on the edge of the Planetary Camera, where the exposures have a poor S/N. Table 2 summarizes the photometry of all these components except H3c, which is too close to the bright cluster galaxy. The different colors have been obtained through aperture magnitudes computed within the same region in all the filters, correcting for sampling and seeing effects. The averaged colors of a bright E/S0 cluster galaxy are also given for comparison. The surface brightnesses and colors are compatible with the multiple-image hypothesis within the photometric errors ( typically, and in the near-IR). The faint images H5a and H3b are surrounded by bright objects (see Fig. 1), and they are hardly detected in the near-IR and g where the sampling ( 0:005 per pixel) and/or the seeing conditions are poor. Moreover H5a and H5b are hardly detected in B. In these particular cases, photometric errors are at least , and these magnitudes are indicated by ":" in Table 2. It is worth noting that the identification of multiple images based on the similarity of the spectral energy distributions (SED) is more easily obtained with extended wavelength coverage.

Table 2. Photometry of the different H3 and H5 components, compared to the mean values for E/S0 cluster galaxies (see text).

The first spectrum of H3 obtained by Bézecourt & Soucail (1997) shows a single emission line which was assigned to [O II]3727 Å, giving , a value which is in fact incompatible with both the lens modelling and the photometric redshift. On the contrary, when it is correctly assigned to the rest-frame Ly, it gives . Subsequent observations of H3 by Frye & Broadhurst (1998) at the Keck Telescope obtained a redshift of through a higher quality spectrum. This time, the redshift is based on several absorption lines, the Ly centroid being slightly shifted redwards with respect to the absorption system, in good agreement with the position of the line found by Bézecourt & Soucail (1997).

In the case of H5a and H5b, both spectra show a strong emission line (Fig. 2). When this line is identified as Ly, the corresponding redshifts are and for the a and b components respectively. Thus, this result confirms that H5a and H5b are two images of the same source, in perfect agreement with the lens model. When taking the Ly centroid to compute the redshift, the two multiple images H3 and H5 are at the same redshift within the errors. Hereafter we take the Ly-based redshift of for both sources, because we have no estimate based on absorption lines for H5. The measured equivalent width of Ly is relatively high for H5, Å for the brightest image, corresponding to a rest frame value of 54 Å. We obtain a rest frame upper limit of Å for the faintest image, where the continuum is hardly detected. These values are similar to the ones observed by Hu et al. (1998) in their sample of emission-line galaxies at to 6. The equivalent width of H3 is smaller, the upper limit obtained from the CFHT spectrum being Å.

Fig. 2. Top: Mean spectra of the b (top ) and a (bottom ) components of H5, each of them showing a strong emission line. Bottom: Averaged spectrum of the two H5 components, showing the emission line identified as Ly at . No other features are clearly identified on the continuum.

The gravitational amplification computed for the brightest images H3a and H5b is magnitudes in both cases, and the predicted difference between the two images are = 0.4 and 0.8 magnitudes for H3 and H5 respectively. The corresponding measured differences (Table 2) are also consistent with this model, and . These amplification factors are surface-averaged values computed with the Kneib et al. (1999) lens model.

We have applied a lens inversion procedure to restore the morphology of these sources on the source plane at . This method is close to the LensClean algorithm (Kochanek & Narayan 1992; Kneib et al. 1994). Fig. 3 displays the resulting morphology and location of the two objects on their source plane. The true separation between H3 and H5 in the source plane would be , corresponding to a linear separation of kpc with . Note that these sources are not resolved in their width, the lens inversion being limited by the resolution of the composite HST images ( in and ). Both sources are extremely clumpy. H5 consists of an alignment of two compact and bright blobs, less than apart ( kpc), and a faint extended component of ( kpc), the total length being ( kpc). H3 is more elongated than H5, and it displays four small and bright subclumps. Each one of these subclumps has less than of diameter, and the total length of the structure is about ( kpc). The orientation of the two sources is similar (see Fig. 3). Compared to the morphologies and sizes of the sample by Steidel et al. (1996a), typically - for the resolved cores ( to kpc with ; see also Giavalisco et al. 1996), the bright subclumps of H3 and H5 are more compact, but the total length of the emitting region is of the same order. From the morphological point of view, H3 and H5 are similar to the lensed sources found behind Cl0939+4713 (Trager et al. 1997).

Fig. 3. Restored image of the two objects H3 (left ) and H5 (right ) on the source plane at , as obtained from the lens-inversion procedure. Both sources are clumpy, elongated and exhibit the same orientation. The distance between H3 and H5 is . At , corresponds to a linear separation of kpc with . Isocontour plots display the center and profile of the different clumps, according to a linear scale which is identical for the two objects.

3.2. The photometric redshift approach

H3 and H5 are the first two high-redshift objects spectroscopically confirmed in our sample of photometrically selected candidates at in this cluster. A photometric redshift method has been used to identify such candidates in the cluster core, close to the critical lines. Photometric redshifts were derived according to the standard minimization method described by Miralles et al. (1999) and Miralles (1998). The observed SED of each galaxy, as obtained from its multicolor photometry, is compared to a set of template spectra. The new Bruzual & Charlot evolutionary code (GISSEL98, Bruzual & Charlot, 1993, 1998) was used to build 5 different synthetic star-formation histories, each with solar metallicity (): a burst of 0.1 Gyr, a constant star-formation rate, and 3 µ models (exponential-decaying SFR) with characteristic time-decays matching the present-day sequence of colors for E, Sa and Sc galaxies. The template database includes 255 synthetic spectra. The intergalactic absorption in the Lyman forest is modelled using the average flux decrements and , according to the original definition by Oke & Korycansky (1982). and correspond respectively to the continuum depression between Ly and Ly, and between Ly and the emission Lyman limit. The prescriptions for the redshift distribution of and are taken from Giallongo & Cristiani (1990), and they are in good agreement with those given by Madau (1995) in the common redshift domain . When applied to our data, the photometric redshift method identifies sources at in this field, most of them too faint to be confirmed spectroscopically using 4m telescopes. The high gravitational amplification of H3 and H5, and their strong Ly emission makes it possible in these two particular cases. The results obtained on the whole field of A2390 and two other cluster-lenses will be discussed in more details elsewhere (see a preliminary version in Pelló et al., 1998).

In the case of H3 and H5, we have examined the sensitivity of our photometric redshift estimate as a function of the relevant parameters, namely SFR, age and metallicity of the stellar population. This exercise is especially important for H5, because the spectroscopic redshift is based on a single line. Fig. 4 presents a likelihood map for this object, showing a good agreement between the spectroscopic () and the photometric redshift (, where the error bar corresponds to a level). This map was obtained using the set of 10 different SFRs presented in the next section, representing templates with a range of metallicities. Each point on the redshift-age map corresponds to the best fit of the SED obtained across the SFR-metallicity space. The dark regions in Fig. 4 (above confidence level) result from the overlap of the 5 metallicities considered (age vs. metallicity degeneracy). The redshift region around appears as the most likely solution for this object. The map obtained when using the solar-metallicity set above mentioned is qualitatively the same. In the case of H3, we obtain a photometric redshift of using the same procedure, thus in good agreement with the spectroscopic value, but with a larger error bar.

Fig. 4. Photometric redshift likelihood-map of H5 showing the excellent agreement with the spectroscopic redshift of this object, which is contained within the region at 68% confidence level. The scaling displayed at the top represents the likelihood value associated to the . The shaded area enclose the contour (confidence level of , or a likelihood value of ). The shaded region on the lower part of the map is excluded because of age-limit considerations for the stellar population (stars cannot be older than the age of the universe, with = 50 km s-1 Mpc-1 and ).

© European Southern Observatory (ESO) 1999

Online publication: May 21, 1999
helpdesk.link@springer.de