6 Photometric redshift determination

We have estimated photometric redshifts for the four unidentified sources 14Z, 84Z, 486A, 607 with broad-band photometry in several filters, described in Sect. 4. Due to the hardness of 14Z, 84Z and 486A (see Table 3) and their large $R-K^{\prime }$ colours we argue that they are probably obscured AGNs. The photometric redshift of these objects is based on the assumption that their spectral energy distribution (SED) in the optical/near-infrared is due to stellar processes. If emission from an obscured AGN is contributing significantly at some wavelength, the following results should be taken with caution.

We used a standard photometric redshift technique (see e.g., Cimatti et al. 1999 and Bolzonella et al. 2000). In our version of the software the templates consist of a set of synthetic spectra (Bruzual & Charlot 1993), with different star formation histories and spanning a wide range of ages (from 10⁵ to 2 10¹⁰ yrs); the basic set of templates includes only solar metallicity and Salpeter's IMF. The effect of IGM attenuation (Madau 1995), extremely important at high redshift, is included, along with the effect of internal dust attenuation, using a dust-screen model and the SMC extinction law with E(B-V) ranging from 0 to 0.5. In total 768 synthetic spectra have been used.

Figure 5: Galaxy template fits to the VRIzJHK photometry of 14Z, 84Z, 486A, and 607 (36Z). The wavelength scale is in Ångstrom, the logarithm of the flux is in units of 10-20 erg s-1 cm-2 Å-1

The "best" photometric redshift ( $z_{\rm phot.}$) for each galaxy is computed by applying a standard, error-weighted $\chi^{2}$ minimization procedure. Moreover we have computed error bars to $z_{\rm phot.}$ corresponding to 90% confidence levels, computed by means of the $\Delta \chi^2$ increment for a single parameter (Avni 1976). The observed spectral energy distribution (SED) of each objects, obtained from broad-band photometry in several filters (V,I from 8K UH, R from Keck+LRIS and z,J,H,K from Keck+NIRC), is compared to our set of template spectra. The V,I 8K UH observations are described by Wilson et al. (1996).

In agreement with their very red colours ( 4.6 < R-K < 5.7) we obtained relatively high $z_{\rm phot.}$, ranging from $1.21<z_{\rm phot.}<2.71$, for all sources (Fig. 5). For objects 14Z and 486A we estimate $z_{\rm phot.}=1.94^{+0.18}_{-0.10}$ and $z_{\rm phot.}=1.21^{+0.10}_{-0.14}$, respectively. Both observed SEDs are consistent with an old stellar population ( ${\rm age}=2.5 \div 5$ Gyrs), while the content of dust is badly constrained because of the absence of U and B band photometry.

The formal best estimate for redshift of object 607 (36Z) is $z_{\rm phot.}=1.36^{+0.07}_{-0.12}$, but the resulting fit is very poor; the upturn in the V photometry at $\lambda<6000$indicates the presence of a young stellar population (0.3 Gyrs) and the absence or a low dust extinction or the presence of an underlying AGN component. The photometric redshift of 607 (36Z) is therefore very uncertain. Data in bluer bands (U and B) and combined AGN and galaxy templates would probably be needed for a more reliable redshift estimate for this object. We therefore have not included the value of 607 in Table 4.

Object 84Z shows a quite clear break between J and H photometry, consistent with a best fit model at $z_{\rm phot.}=2.71^{+0.29}_{-0.41}$, while the decreasing flux towards shorter wavelength indicates the presence of a moderate dust content ( E(B-V)=0.3) in a young stellar population ( ${\rm age}=0.1$ Gyrs). As seen in Fig. 5, the resulting fit is very good, with a $\chi^{2}$value of the order of unity. However, due to the relatively large magnitude errors, especially in the I and z bands, also lower redshifts (down to $z \sim1.5$) would be statistically acceptable.