Aims. The clustering properties of a large sample of U-dropouts are investigated and compared to very precise results for B-dropouts from other studies to identify a possible evolution from z=4 to z=3. Methods. A population of ~8800 candidates for star-forming galaxies at z=3 is selected via the well-known Lyman-break technique from a large optical multicolour survey (the ESO Deep Public Survey). The selection efficiency, contamination rate, and redshift distribution of this population are investigated by means of extensive simulations. Photometric redshifts are estimated for every Lyman-break galaxy (LBG) candidate from its UBVRI photometry yielding an empirical redshift distribution. The measured angular correlation function is deprojected and the resulting spatial correlation lengths and slopes of the correlation function of different subsamples are compared to previous studies. Results. By fitting a simple power law to the correlation function we do not see an evolution in the correlation length and the slope from other studies at z=4 to our study at z=3. In particular, the dependence of the slope on UV-luminosity similar to that recently detected for a sample of B-dropouts is confirmed also for our U-dropouts. For the first time number statistics for U-dropouts are sufficient to clearly detect a departure from a pure power law on small scales down to ~2 reported by other groups for B-dropouts.
Download