Since spectroscopy relies on dispersing the light from an object according to wavelength, accurate redshifts, which require sufficient signal-to-noise in individual emission and/or absorption lines, can only be obtained for relatively bright objects. An alternative, although less reliable, technique to measure redshifts relies on broad-band photometry. By measuring the flux of an object in a relatively small number of wavebands, one obtains a very crude sampling of the object’s SED. As we have seen, the SEDs of galaxies reveal a number of broad spectral features (see Fig. 1). An important example is the break, which is due to a sudden change 4000Å break, which is due to a sudden change in the opacity at this wavelength in the atmospheres of low mass stars, and therefore features predominantly in galaxies with stellar population ages ≳ 108 yr. Because of this 4000 Å break ˚ and other broad spectral features, the colors of a population of galaxies at a given redshift only occupy a relatively small region of the full multi-dimensional color space. Since this region changes as function of redshift, the broad-band colors of a galaxy can be used to estimate its redshift.
In practice one proceeds as follows. For a given template spectrum, either from an observed galaxy or computed using population synthesis models, one can determine the relative fluxes expected in different wavebands for a given redshift. By comparing these expected fluxes with the observed fluxes one can determine the best-fit redshift and the best-fit template spectrum (which basically reflects the spectral type of the galaxy). The great advantage of this method is that photometric redshifts can be measured much faster than their spectroscopic counterparts, and that it can be extended to much fainter magnitudes. The obvious downside is that photometric redshifts are far less reliable. While a spectroscopic redshift can easily be measured to a relative error of less than 0.1%, photometric errors are typically of the order of 3–10%, depending on which and how many wavebands are used. Furthermore, the error is strongly correlated with the spectral type of the galaxy. It is typically much larger for star-forming galaxies, which lack a pronounced 4000 Å break, than for galaxies with an old stellar population.
A prime example of a photometric redshift survey, illustrating the strength of this technique, is the COMBO-17 survey (Wolf et al., 2003), which comprises a sample of ∼ 25,000 galaxies with photometric redshifts obtained from photometry in 17 relatively narrow optical wavebands. Because of the use of a relatively large number of filters, this survey was able to reach an average redshift accuracy of ∼ 3%, sufficient to study various statistical properties of the galaxy population as a function of redshift.
Fig.1. Spectra of different types of galaxies from the ultraviolet to the near-infrared. From ellipticals to late-type spirals, the blue continuum and emission lines become systematically stronger. For early-type galaxies, which lack hot, young stars, most of the light emerges at the longest wavelengths, where one sees absorption lines characteristic of cool K stars. In the blue, the spectrum of early-type galaxies show strong H and K absorption lines of calcium and the G band, characteristic of solar type stars. Such galaxies emit little light at wavelengths shorter than 4000A ˚ and have no emission lines. In contrast, late-type galaxies and starbursts emit most of their light in the blue and near-ultraviolet. This light is produced by hot young stars, which also heat and ionize the interstellar medium giving rise to strong emission lines. [Based on data kindly provided by S. Charlot]
