We present the analysis of a large sample of early-type galaxies ( ETGs ) at 0 < z < 3 aimed at tracing the cosmic evolution of their size and compare it with a model of pure dissipationless ( dry ) merging in the \Lambda CDM framework .
The effective radius R _ { e } depends on stellar mass { \cal M } as R _ { e } ( { \cal M } ) \propto { \cal M } ^ { \alpha } with \alpha \sim 0.5 at all redshifts .
The redshift evolution of the mass- or SDSS-normalized size can be reproduced as \propto ( 1 + z ) ^ { \beta } with \beta \sim - 1 , with the most massive ETGs possibly showing the fastest evolutionary rate ( \beta \sim - 1.4 ) .
This size evolution slows down significantly to \beta \sim - 0.6 if the ETGs at z > 2 are removed from the sample , suggesting an accelerated increase of the typical sizes at z > 2 , especially for the ETGs with the largest masses .
A pure dry merging \Lambda CDM model is marginally consistent with the average size evolution at 0 < z < 1.7 , but predicts descendants too compact for z > 2 progenitor ETGs .
This opens the crucial question on what physical mechanism can explain the accelerated evolution at z > 2 , or whether an unclear observational bias is partly responsible for that .