Impact of subsolar metallicities on photometric redshifts

If you are interested in all the details you can find all information in our paper:

Impact of sub-solar metallicities on photometric redshifts
Ralf Kotulla & Uta Fritze
MNRAS Letters 393, 55 (2009) - arXiv:0811.3217

Most of current photometric redshift codes can be roughly separated into three types: 1) machine learning and neural networks; 2) template fitting using empirical templates; 3) template fitting using theoretical templates; All three methods have their pros and cons, and among the pros of the third - which is among the most commonly used ones - is its capability to derive physical parameters in a relatively straightforward way. However, common pratice is to use templates that use a fixed metallicity, usually using solar metallicity.

In this paper we showed the impact that this approach has on the derived photometric redshifts when it is applied to simulated, but more realistic galaxies that have an intrinsic metallicity dispersion. So we created two grids of models, one using our chemically consistent (CC) models to simulate the galaxies, and another grid of fixed, solar metallicity (Z3).

Fig. 1: Chi² values as goodness-of-fit indicator, with larger values representing increasingly poor fits.

After adding dust and photometric noise to simulate realistic observations, I then used my photometric redshift code GAZELLE and analyzed the mock-observations with both grids. If analyzed with a matched grid, i.e. comparing chemically consistent galaxies with chemically consistent templates, or solar metallicity galaxies with solar metallicity templates, the chi²values as a measure of the goodness-of-fit indicate very good fits with chi² near the ideal value of 1. In the test-case of analyzing "real" galaxies with solar-metallicity templates the fits are much worse with chi² values larger by orders of magnitudes.

Fig. 2

Fig. 3

The next step was then to look at the deviation of the derived photometric redshifts. This is done individually for Elliptical-like galaxies (Fig. 2) and spiral models (Fig. 3). In both cases, even for the Elliptical for which the metallicity is near solar, we found a significant bias in that photometric redshifts that are generally underestimated. The reason for this bias is simple: All real galaxies OBand also all out chemically consistent galaxies have some contribution of low-metallicity stars that were formed when the galaxy was still young and the gas not yet heavily enriched. These stars make the integrated spectrum bluer, and these bluer colors are in turn compensated for by a smaller redshift.