2009ApJ...703.2160T

We compare the optical spectral types with the X-ray spectral properties for a uniformly selected (sources with fluxes greater than the 3σ level and above a flux limit of f_2-8 keV_> 3.5x10^–15 erg/cm2/s), highly spectroscopically complete (>80% for f_2-8 keV_> 10^–14 erg/cm2/s and >60% below) 2-8 keV X-ray sample observed in three Chandra fields (CLANS, CLASXS, and the CDF-N) that cover ∼1.2 deg². For our sample of 645 spectroscopically observed sources, we confirm that there is significant overlap of the X-ray spectral properties, as determined by the effective photon indices, Γ_eff, obtained from the ratios of the 0.5-2 keV to 2-8 keV counts, for the different optical spectral types. For example, broad-line active galactic nuclei (AGNs) are expected to be unobscured and hence X-ray soft (Γ_eff≥ 1.2), yet we find 20%± 3% have Γ_eff<1.2. Non-broad-line AGNs are expected to be obscured and hence X-ray hard (Γ_eff < 1.2), yet we find 33%±4% have Γ_eff≥ 1.2. Thus, one cannot use the X-ray spectral classifications and the optical spectral classifications equivalently. Since it is not understood how X-ray and optical classifications relate to the obscuration of the central engine, we strongly advise against a mixed classification scheme, as it can only complicate the interpretation of X-ray AGN samples. We confirm the dependence of optical spectral type on X-ray luminosity, and for z < 1, we find a similar luminosity dependence of Γ_eff. However, this dependence breaks down at higher redshifts due to the highly redshift-dependent nature of Γ_eff. We therefore also caution that any classification scheme which depends on Γ_eff is likely to suffer from serious redshift bias.