2014A&A...567A.111M

Strong gravitational lensing provides fundamental insights into the understanding of the dark matter distribution in massive galaxies, galaxy clusters, and the background cosmology. Despite their importance, few gravitational arcs have been discovered so far. The urge for more complete, large samples and unbiased methods of selecting candidates increases. Several methods for the automatic detection of arcs have been proposed in the literature, but large amounts of spurious detections retrieved by these methods force observers to visually inspect thousands of candidates per square degree to clean the samples. This approach is largely subjective and requires a huge amount of checking by eye, especially considering the actual and upcoming wide-field surveys, which will cover thousands of square degrees. In this paper we study the statistical properties of the colours of gravitational arcs detected in the 37 deg² of the CFHTLS-Archive-Research Survey (CARS). Most of them lie in a relatively small region of the (g'-r',r'-i') colour-colour diagram. To explain this property, we provide a model that includes the lensing optical depth expected in a ΛCDM cosmology that, in combination with the sources' redshift distribution of a given survey, in our case CARS, peaks for sources at redshift z∼1. By furthermore modelling the colours derived from the spectral energy distribution of the galaxies that dominate the population at that redshift, the model reproduces the observed colours well. By taking advantage of the colour selection suggested by both data and model, we automatically detected 24 objects out of 90 detected by eye checking. Compared with the single-band arcfinder, this multi-band filtering returns a sample complete to 83% and a contamination reduced by a factor of ∼6.5. New gravitational arc candidates are also proposed.