Luminous red galaxies ( LRGs ) are much rarer and more massive than L _ { * } galaxies .
Coupled with their extreme colours , LRGs therefore provide a demanding testing ground for the physics of massive galaxy formation .
We present the first self-consistent predictions for the abundance and properties of LRGs in hierarchical structure formation models .
We test two published models which use quite different mechanisms to suppress the formation of massive galaxies : the Bower et al .
( 2006 ) model , which invokes “ AGN-feedback ” to prevent gas from cooling in massive haloes , and the Baugh et al .
( 2005 ) model which relies upon a “ superwind ” to eject gas before it is turned into stars .
Without adjusting any parameters , the Bower et al .
model gives an excellent match to the observed luminosity function of LRGs in the Sloan Digital Sky Survey ( with a median redshift of z = 0.24 ) and to their clustering ; the Baugh et al .
model is less successful in these respects .
Both models fail to match the observed abundance of LRGs at z = 0.5 to better than a factor of \approx 2 .
In the models , LRGs are typically bulge dominated systems with stellar masses of \approx 2 \times 10 ^ { 11 } h ^ { -1 } M _ { \odot } and velocity dispersions of \sigma \sim 250 { kms } ^ { -1 } .
Around half of the stellar mass in the model LRGs is already formed by z \sim 2.2 and is assembled into one main progenitor by z \sim 1.5 ; on average , only 25 % of the mass of the main progenitor is added after z \sim 1 .
LRGs are predicted to be found in a wide range of halo masses , a conclusion which relies on properly taking into account the scatter in the formation histories of haloes .
Remarkably , we find that the correlation function of LRGs is predicted to be a power law down to small pair separations , in excellent agreement with observational estimates .
Neither the Bower et al .
nor the Baugh et al .
model is able to reproduce the observed radii of LRGs .