Simulations of nearby (0.015 < z < 0.025) SDSS galaxies have been used to reproduce as accurately as possible the appearance that they would have on COSMOS ACS images if they had been observed at z ~ 0.7 and z ~ 1.2. By adding the SDSS galaxies to random locations in the COSMOS images, we simulate the effects of chance superpositions of high redshift galaxies with unrelated foreground or background objects. We have used these simulated images, together with those of real COSMOS galaxies at these same redshifts, to undertake a blind morphological classification of galaxies to identify those that appear to be undergoing mergers and thus to estimate the change in merger fraction with redshift. We find that real mergers are harder to recognize at high redshift, and also that the chance superposition of unrelated galaxies often produces the appearance of mergers where in reality none exists. In particular, we estimate that 1.5 - 2.0% of objects randomly added to ACS images are misclassified as mergers due to projection with unrelated objects, and as a result, that 40% of the apparent mergers in COSMOS at z=0.7 are likely to be spurious. We find that the fraction of galaxies undergoing mergers increases as (1+z)^3.8+/-1.2 to z ~ 0.7 and that this trend appears to continue to z = 1.2. Merger candidates at z ~ 0.7 are bluer than the parent population, especially when the statistical effects of the chance projections are accounted for. Merger candidates are more asymmetric than the population as a whole, and are often associated with irregular morphology. Nevertheless, the majority (~60%) of the merger candidates appear to be associated with spiral galaxies although in this case we cannot correct for the effects of chance projections.
Download