arXiv2006
(ABRIDGED) Motivated by the desire to reliably and automatically classify structure of thousands of COSMOS galaxies, we present ZEST, the Zurich Estimator of Structural Types. To classify galaxy structure, ZEST uses: (i) Five non-parametric diagnostics: asymmetry, concentration, Gini coefficient, 2nd-order moment of the brightest 20% of galaxy pixels, and ellipticity; and (ii) The exponent n of single--Sersic fits to the 2D surface brightness distributions. To fully exploit the wealth of information while reducing the redundancy present in these diagnostics, ZEST performs a principal component (PC) Analysis. We use a sample of ~56,000 I<24 COSMOS galaxies to show that the first three PCs fully describe the key aspects of the galaxy structure, i.e., to calibrate a three-dimensional classification grid of axis PC_1, PC_2, and PC_3. We demonstrate the robustness of the ZEST grid on the z=0 sample of Frei et al. (1996). The ZEST classification breaks most of the degeneracy between different galaxy populations that affects morphological classifications based on only some of the diagnostics included in ZEST. As a first application, we present the evolution since z~1 of the Luminosity Functions of COSMOS galaxies of early, disk and irregular galaxies and, for disk galaxies, of different bulge-to-disk ratios. Overall, we find that the LF up to a redshift z=1 is consistent with a pure-luminosity evolution (of about 0.95 magnitudes at z \~0.7). We highlight however two trends, that are in general agreement with a down-sizing scenario for galaxy formation: (1.) A deficit of a factor of about two at z~0.7 of MB>-20.5 structurally--classified early--type galaxies; and (2.) An excess of a factor of about three, at a similar redshift, of irregular galaxies.