I. Selection and optical properties of a [Ne ]-selected sample

Aims. The application of multi-wavelength selection techniques is essential for obtaining a complete and unbiased census of active galactic nuclei (AGN). We present here a method for selecting z ∼ 1 obscured AGN from optical spectroscopic surveys. Methods. A sample of 94 narrow-line AGN with 0.65 < z < 1.20 was selected from the 20k-Bright zCOSMOS galaxy sample by detection of the high-ionization [Ne ] λ3426 line. The presence of this emission line in a galaxy spectrum is indicative of nuclear activity, although the selection is biased toward low absorbing column densities on narrow-line region or galactic scales. A similar sample of unobscured (type 1 AGN) was collected applying the same analysis to zCOSMOS broad-line objects. This paper presents and compares the optical spectral properties of the two AGN samples. Taking advantage of the large amount of data available in the COSMOS field, the properties of the [Ne ]-selected type 2 AGN were investigated, focusing on their host galaxies, X-ray emission, and optical line-flux ratios. Finally, a previously developed diagnostic, based on the X-ray-to-[Ne ] luminosity ratio, was exploited to search for the more heavily obscured AGN. Results. We found that [Ne ]-selected narrow-line AGN have Seyfert 2-like optical spectra, although their emission line ratios are diluted by a star-forming component. The ACS morphologies and stellar component in the optical spectra indicate a preference for our type 2 AGN to be hosted in early-type spirals with stellar masses greater than 109.5−10 M , on average higher than those of the galaxy parent sample. The fraction of galaxies hosting [Ne ]-selected obscured AGN increases with the stellar mass, reaching a maximum of about 3% at ≈2 × 1011 M . A comparison with other selection techniques at z ∼ 1, namely the line-ratio diagnostics and X-ray detections, shows that the detection of the [Ne ] λ3426 line is an effective method for selecting AGN in the optical band, in particular the most heavily obscured ones, but cannot provide a complete census of type 2 AGN by itself. Finally, the high fraction of [Ne ]-selected type 2 AGN not detected in medium-deep (≈100–200 ks) Chandra observations (67%) is suggestive of the inclusion of Compton-thick (i.e., with NH > 1024 cm−2) sources in our sample. The presence of a population of heavily obscured AGN is corroborated by the X-ray-to-[Ne ] ratio; we estimated, by means of an X-ray stacking technique and simulations, that the Comptonthick fraction in our sample of type 2 AGN is 43 ± 4% (statistical errors only), which agrees well with standard assumptions by XRB synthesis models.