2011ApJ...738...97B

The positive correlation between planet detection rate and host star iron abundance lends strong support to the core accretion theory of planet formation. However, iron is not the most significant mass contributor to the cores of giant planets. Since giant planet cores are thought to grow from silicate grains with icy mantles, the likelihood of gas giant formation should depend heavily on the oxygen and silicon abundance of the planet formation environment. Here we compare the silicon and oxygen abundances of a set of 76 planet hosts and a control sample of 80 metal-rich stars without any known giant planets. Our new, independent analysis was conducted using high resolution, high signal-to-noise data obtained at McDonald Observatory. Because we do not wish to simply reproduce the known planet-metallicity correlation, we have devised a statistical method for matching the underlying [Fe/H] distributions of our two sets of stars. We find a 99% probability that planet detection rate depends on the silicon abundance of the host star, over and above the observed planet-metallicity correlation. We do not detect any such correlation for oxygen. Our results would thus seem to suggest that grain nucleation, rather than subsequent icy mantle growth, is the important limiting factor in forming giant planets via core accretion. Based on our results and interpretation, we predict that planet detection should correlate with host star abundance for refractory elements responsible for grain nucleation and that no such trends should exist for the most abundant volatile elements responsible for icy mantle growth.