The clustering properties of Lyman break galaxies at redshift z∼3

We present a new measure of the angular two-point correlation function of Lyman break galaxies (LBGs) at zsimilar to3, obtained from the variance of galaxy counts in two-dimensional cells. By avoiding binning of the angular separations, this method is significantly less affected by shot noise than traditional measures and allows for a more accurate determination of the correlation function. We used a sample of about 1000 galaxies with Rless than or equal to25.5 extracted from the survey by Steidel and collaborators and found the following results. At scales in the range 30 less than or similar to theta less than or similar to 100 the angular correlation function w(theta) can be accurately described as a power law with slope beta = 0.50(-0.50)(+0.25) (1 sigma random)(-0.10) (systematic), shallower than the measure presented by Giavalisco and collaborators. However, the spatial correlation length, derived by Limber deprojection, is in very good agreement with the previous measures, confirming the strong spatial clustering of these sources. We discuss in detail the effects of both random and systematic errors, in particular of the so-called integral constraint bias, to which we set a lower limit using numerical simulations. This suggests that the current samples do not yet provide a fair representation of the large-scale distribution of LBGs at zsimilar to3. An intriguing result of our analysis is that at angular separations smaller than thetaless than or similar to30 the correlation function seems to depart from the power law fitted at larger scales and become smaller. This feature is detected at the similar to90% confidence level and, if real, can provide information on the number density and spatial distribution of LBGs within their host halos as well as the size and the mass of the halos.