2009MNRAS.392..998E

Using a sample of 21 damped Lyman alpha systems (DLAs) and 35 sub-DLAs, we evaluate the D-index from high-resolution spectra of the Mgii λ 2796 profile. This sample represents an increase in the sub-DLA statistics by a factor of 4 over the original D-index sample. We investigate various techniques to define the velocity spread (Δv) of the Mgii line to determine an optimal D-index for the identification of DLAs. The success rate of DLA identification is 50-55 per cent, depending on the velocity limits used, improving by a few per cent when the column density of Feii is included in the D-index calculation. We recommend the set of parameters that are judged to be most robust, have a combination of high DLA identification rate (57 per cent) and low DLA miss rate (6 per cent) and most cleanly separate the DLAs and sub-DLAs (Kolmogorov-Smirnov probability 0.5 per cent). These statistics demonstrate that the D-index is the most efficient technique for selecting low-redshift DLA candidates: 65 per cent more efficient than selecting DLAs based on the equivalent widths of Mgii and Feii alone. We also investigate the effect of resolution on determining the N(Hi) of sub-DLAs. We convolve echelle spectra of sub-DLA Lyα profiles with Gaussians typical of the spectral resolution of instruments on the Hubble Space Telescope and compare the best-fitting N(Hi) values at both the resolutions. We find that the fitted Hi column density is systematically overestimated by ∼0.1 dex in the moderate-resolution spectra compared to the best fits to the original echelle spectra. This offset is due to blending of nearby Lyα clouds that are included in the damping wing fit at low resolution.