The analysis of the variability of active galactic nuclei (AGNs) at different wavelengths and the study of possible correlations among different spectral windows are nowadays a major field of inquiry. Optical variability has been largely used to identify AGNs in multivisit surveys. The strength of a selection based on optical variability lies in the chance to analyze data from surveys of large sky areas by ground-based telescopes. However the effectiveness of optical variability selection, with respect to other multiwavelength techniques, has been poorly studied down to the depth expected from next generation surveys. Here we present the results of our r-band analysis of a sample of 299 optically variable AGN candidates in the VST survey of the COSMOS field, counting 54 visits spread over three observing seasons spanning > 3 yr. This dataset is > 3 times larger in size than the one presented in our previous analysis (De Cicco et al. 2015), and the observing baseline is ~8 times longer. We push towards deeper magnitudes (r(AB) ~23.5 mag) compared to past studies; we make wide use of ancillary multiwavelength catalogs in order to confirm the nature of our AGN candidates, and constrain the accuracy of the method based on spectroscopic and photometric diagnostics. We also perform tests aimed at assessing the relevance of dense sampling in view of future wide-field surveys. We demonstrate that the method allows the selection of high-purity (> 86%) samples. We take advantage of the longer observing baseline to achieve great improvement in the completeness of our sample with respect to X-ray and spectroscopically confirmed samples of AGNs (59%, vs. ~15% in our previous work), as well as in the completeness of unobscured and obscured AGNs. The effectiveness of the method confirms the importance to develop future, more refined techniques for the automated analysis of larger datasets.
Download