Modeling the color evolution of luminous red galaxies - improvements with empirical stellar spectra

Predicting the colors of Luminous Red Galaxies (LRGs) in the Sloan Digital Sky Survey (SDSS) has been a long-standing problem. The g,r,i colors of LRGs are inconsistent with stellar population models over the redshift range 0.1<z<0.7. The g-r colors in the models are on average redder than the data while the r-i colors in the models are bluer towards low redshift. Beyond redshift 0.4, the predicted r-i color becomes instead too red, while the predicted g-r agrees with the data. We provide a solution to this problem, through a combination of new astrophysics and a fundamental change to the stellar population modeling. We find that the use of the empirical library of Pickles (1998) instead of theoretical spectra modifies the predicted colors exactly in the way suggested by the data. The reason is a lower flux in the empirical libraries, with respect to the theoretical ones, in the wavelength range 5500-6500 AA. The discrepancy increases with decreasing effective temperature independently of gravity. This result has general implications for a variety of studies from globular clusters to high-redshift galaxies. The astrophysical part of our solution regards the composition of the stellar populations of these massive Luminous Red Galaxies. We find that on top of the previous effect one needs to consider a model in which ~3% of the stellar mass is in old metal-poor stars. Other solutions such as substantial blue Horizontal Branch at high metallicity or young stellar populations can be ruled out by the data. Our new model provides a better fit to the g-r and r-i colors of LRGs and gives new insight into the formation histories of these most massive galaxies. Our model will also improve the k- and evolutionary corrections for LRGs which are critical for fully exploiting present and future galaxy surveys.