Astron. Astrophys. 363, 517-525 (2000)

1. Introduction

Recent multi-waveband galaxy surveys, carried out from UV to radio wavelengths, have identified a population of rapidly star forming galaxies in the range (Sullivan et al 1999; Lilly et al. 1996; Cowie et al. 1997, Rowan-Robinson et al. 1997; Hughes et al. 1998; Mobasher et al. 1999). In particular, these studies confirm a relatively higher rate of star formation in the past, as supported by the discovery of a population of massive, starforming galaxies at (Madau et al. 1996; Steidel et al. 1996). These objects are likely to be progenitors of the present day galaxies (Giavalisco et al. 1996; Lowenthal et al. 1997) and hence, a statistical study of this population from the epoch of galaxy formation to the present, gives clues towards scenarios of the formation and evolution of galaxies (Fukugita et al. 1996). This can also constrain the star formation history of galaxies out to . Such studies require redshift information for a population of starforming galaxies at faint levels. Recently, the depth of the available surveys is greatly extended by the Hubble Space Telescope (HST) observations of the Hubble Deep Fields (HDF). The large spectral coverage of the HDFs provide a unique opportunity to study evolutionary properties of faint galaxies.

Ground-based spectroscopic measurements of the brighter ( mag.) sub-sample of the HDF have been performed (Cohen et al. 2000; Steidel et al. 1996; Lowenthal et al. 1997; Zepf et al. 1996). However, for the fainter galaxies in the HDF, redshift measurements are more difficult with spectroscopic features almost impossible to identify. For these objects, the photometric redshift technique (Loh & Spillar 1986; Connolly et al. 1995) is faster than its spectroscopic counterpart and applicable to much fainter magnitudes. This is due to a larger bin size in photometry compared to spectroscopy ( Å Å), leading to a shorter exposure time with a trade-off in accuracy of the measured redshifts.

Considering the new generation of 8m class telescopes, the planned instrumentation on the HST and future, high sensitive radio telescopes, a substantial number of deep surveys at different wavelengths will soon become available. Most of these galaxies will be too faint for spectroscopic study and hence, the photometric redshift technique is the only practical way for estimating their redshifts. In a recent assessment of different photometric redshift techniques, using a redshift-limited spectroscopic survey, it was shown that photometric redshifts can be predicted with an accuracy of 0.1 (0.3) for () of the sources examined (Hogg et al. 1998). Therefore, photometric redshifts could provide a powerful tool for statistical studies of evolutionary properties of galaxies and in particular of faint galaxies for which spectroscopic data are difficult to obtain.

The most important step in any study concerning photometric redshift measurement, is the choice of the template Spectral Energy Distributions (SEDs) for different populations of galaxies, with which the observed SEDs should be compared. There are two general ways for adopting the template SEDs:

• a) Empirical templates: in this case one uses the mean observed SEDs corresponding to different types of galaxies. The problem here is that there are not enough information about the observed SEDs for different classes of objects at different redshifts (particularly at high redshifts). Therefore, incorporating the spectral evolution of galaxies of different types on their template SEDs is difficult and uncertain.

• b) Synthetic templates: uses model SEDs for different spectral types of galaxies, shifted in redshift space, assuming evolutionary population synthesis models. The main problem here is to constrain the evolutionary model parameters to produce realistic model SEDs (for different types) as a function of redshift. In particular, the effect of dust at high redshifts (specially in star forming galaxies) is not known.

To avoid these problems, we introduce a combined approach, producing realistic model SEDs based on chemo-photometric Evolutionary Population Synthesis (EPS) models, extending from UV to 1 mm in wavelength. The template SEDs here, are consistently and simultaneously optimised to; a) produce the observed colours of galaxies at ; b) incorporate chemo-photometric evolution for galaxies of different types, in agreement with observations; c) allow treatment of dust contribution and its evolution with redshift, consistent with the EPS models; d) include absorption and re-emission of radiation by dust and hence, realistic estimates of the far-infrared radiation; e) include correction for inter-galactic absorption by Lyman continuum and Lyman forest. The evolutionary models and hence, the template SEDs, are constrained by minimising the scatter between the photometric and spectroscopic redshifts for a calibrating sample of galaxies with known spectroscopic data .

The main advantage of this technique over the previous works is that it simultaneously and self-consistently allows for the treatment of both the photometric and chemical evolution of individual galaxies with time. Moreover, since the synthetic template SEDs cover the range from UV to sub-mm wavelengths, one could consistently use the optimised SEDs to estimate contributions from individual galaxies to the far-infrared and sub-mm wavelengths. Also, the effect of dust and its evolution with redshift is self-consistently accounted for in the template SEDs and optimised to produce the observations. This is crucial in any photometric redshift technique if it is to be applied on high redshift, star-forming galaxies (Meurer et al. 1997; Cimatti et al. 1998).

The new photometric redshift technique is outlined in the next section. In Sect. 3, the EPS models are briefly discussed. Sect. 4 presents the calibration sample. This is followed by the optimised template SEDs in Sect. 5. The uncertainties in the photometric redshifts and spectral types are explored in Sect. 6. The conclusions are presented in Sect. 7.

© European Southern Observatory (ESO) 2000

Online publication: December 11, 2000
helpdesk.link@springer.de