We present two methods for determining spectroscopic redshifts of galaxies inthe DEEP2 survey which display only one identifiable feature, an emission line,in the observed spectrum ("single-line galaxies"). First, we assume each singleline is one of the four brightest lines accessible to DEEP2: Halpha, [OIII]5007, Hbeta, or [OII] 3727. Then, we supplement spectral information with BRIphotometry. The first method, parameter space proximity (PSP), calculates thedistance of a single-line galaxy to galaxies of known redshift in (B-R), (R-I),R, observed wavelength parameter space. The second method is an artificialneural network (ANN). Prior information, such as allowable line widths andratios, rules out one or more of the four lines for some galaxies in bothmethods. Based on analyses of evaluation sets, both methods are nearly perfectat identifying blended [OII] doublets. Of the lines identified as Halpha in thePSP and ANN methods, 91.4% and 94.2% respectively are accurate. Although themethods are not this accurate at discriminating between [OIII] and Hbeta, theycan identify a single line as one of the two, and the ANN method in particularunambiguously identifies many [OIII] lines. From a sample of 640 single-linespectra, the methods determine the identities of 401 (62.7%) and 472 (73.8%)single lines, respectively, at accuracies similar to those found in theevaluation sets.
展开▼
