Abstract

Identifying active galactic nuclei (AGNs) using only one wavelength region may miss AGNs with characteristics unfavorable for the identification method, which introduces bias in the sample. To better understand these selection effects, we compare the multiwavelength properties of 2584 X-ray- and IR-selected AGNs at z < 3 in the COSMOS field. The X-ray-selected sample consists of the 2219 AGNs with intrinsic 0.5-10 keV X-ray luminosity greater than 10(43) erg s(-1); the 993 IR-selected AGNs are identified by their high-contrast, power-law IR colors using Spitzer/IRAC observations. We find that both selection methods miss a significant fraction of AGNs, as only 24.3% of AGNs (628 out of the total of 2584 AGNs) are recovered by both selection criteria. X-ray selection can capture the most diverse range of multiwavelength spectral emission features, while IRAC selection is biased against AGNs with a steeply decreasing power flux (lambda F-lambda) between 1 and 6 mu m. For AGNs with X-ray detections, IRAC-selected sources are 2.5x more X-ray luminous than AGNs not identified by their IR colors. IRAC-selected AGNs without detectable X-rays are obscured, as they have a mean X-ray correction factor log(L-X,L-obs/L-X,L-int)= -1.4; 28.5%-42.2% of these AGNs are Compton-thick (according to estimated column densities). Only 1.4% of X-ray-selected AGNs are Compton-thick. The AGN selection technique used significantly impacts the overall characteristics of the resulting AGN sample.