Theoretical arguments and indirect observational evidence suggest that the stellar initial mass function ( IMF ) may evolve with time , such that it is more weighted toward high mass stars at higher redshift .
Here we test this idea by comparing the rate of luminosity evolution of massive early-type galaxies in clusters at 0.02 \leq z \leq 0.83 to the rate of their color evolution .
A combined fit to the rest-frame U - V color evolution and the previously measured evolution of the M / L _ { B } ratio gives x = -0.3 ^ { +0.4 } _ { -0.7 } for the logarithmic slope of the IMF in the region around 1 M _ { \odot } , significantly flatter than the present-day value in the Milky Way disk of x = 1.3 \pm 0.3 .
The best-fitting luminosity-weighted formation redshift of the stars in massive cluster galaxies is 3.7 ^ { +2.3 } _ { -0.8 } , and a possible interpretation is that the characteristic mass m _ { c } had a value of \sim 2 M _ { \odot } at z \sim 4 ( compared to m _ { c } \sim 0.1 M _ { \odot } today ) , in qualitative agreement with models in which the characteristic mass is a function of the Jeans mass in molecular clouds .
Such a “ bottom-light ” IMF for massive cluster galaxies has significant implications for the interpretation of measurements of galaxy formation and evolution .
Applying a simple form of IMF evolution to literature data , we find that the volume-averaged star formation rate at high redshift may have been overestimated ( by a factor of 3 - 4 at z > 4 ) , and the cosmic star formation history may have a fairly well-defined peak at z \sim 1.5 .
The M / L _ { V } ratios of galaxies are less affected than their star formation rates , and future data on the stellar mass density at z > 3 will provide further constraints on IMF evolution .
The formal errors likely underestimate the uncertainties , and confirmation of these results requires a larger sample of clusters and the inclusion of redder rest-frame colors in the analysis .