A study of cross-correlation and breakfinding algorithms applied to the measurements of redshifts in very-low-resolution spectra

In this article we compare the cross-correlation and breakfinder techniques applied to the measurements of redshifts from low-resolution spectra. We assume spectra obtained from multinarrowband imagery, a technique for multiobject spectrophotometry. Comparing the cross-correlation with the breakfinder, we find that neither is intrinsically superior to the other. They have comparable precision for early-type galaxies but the cross-correlation is clearly superior for later-type galaxies. On the other hand, unlike the breakfinder, the cross-correlation is sensitive to instrumental errors as well as the shape of the templates used. We carried out simulations with two types of global spectral errors: a slope error and sinusoidal errors. They can introduce serious systematic errors in redshifts measured with the cross-correlation. The breakfinder is a spectrally local test and does not suffer from this type of error. We therefore conclude that the two techniques are complementary and that they should both be used to help flag objects for which they give abnormally discrepant redshifts.