Submegaparsec individual photometric redshift estimation from cosmic web constraints

Abstract

We present a method, PhotoWeb, for estimating photometric redshifts of individual galaxies, and their equivalent distance, with megaparsec and even sub-megaparsec accuracy using the Cosmic Web as a constraint over photo-z estimates. PhotoWeb redshift errors for individual galaxies are of the order of delta_z = 0.0007, compared to errors of delta_z = 0.02 for current photo-z techniques. The mean redshift error is of the order of 0.00005-0.0004 compared to mean errors in the range delta_z =z 0.001-0.01 for the best available photo-z estimates in the literature. Current photo-z techniques based on the spectral energy distribution of galaxies and projected clustering produce redshift estimates with large errors due to the poor constraining power the galaxy's spectral energy distribution and projected clustering can provide. The Cosmic Web, on the other hand, provides the strongest constraints on the position of galaxies. The network of walls, filaments and voids occupy ~%10 of the volume of the Universe, yet they contain ~%95 of galaxies. The cosmic web, being a cellular system with well-defined boundaries, sets a restricted set of intermittent positions a galaxy can occupy along a given line-of-sight. Using the information in the density field computed from spectroscopic redshifts we can narrow the possible locations of a given galaxy along the line of sight from a single broad probability distribution (from photo-z) to one or a few narrow peaks. Our first results improve previous photo-z errors by more than one order of magnitude allowing sub-megaparsec errors in some cases. Such accurate estimates for tens of millions of galaxies will allow unprecedented galaxy-LSS studies. In this work we apply our technique to the SDSS photo-z galaxy sample and discuss its performance and future improvements.