XMMPZCAT: A catalogue of photometric redshifts for X-ray sources

TL;DR

This paper presents XMMPZCAT, a large catalog of photometric redshifts for X-ray sources from the 3XMM catalogue, using machine learning on multi-wavelength data to estimate distances for half of the sources.

Contribution

Introduction of a new photometric redshift catalog for 3XMM X-ray sources utilizing machine learning and multi-wavelength data, covering about 50% of the sources.

Findings

Photometric redshifts estimated for 100,178 X-ray sources.

Outlier rate varies from 4% to 40% depending on data availability.

The catalog enhances the scientific potential of the 3XMM X-ray catalogue.

Abstract

The third version of the XMM-Newton serendipitous catalogue (3XMM), containing almost half million sources, is now the largest X-ray catalogue. However, its full scientific potential remains untapped due to the lack of distance information (i.e. redshifts) for the majority of its sources. Here we present XMMPZCAT, a catalogue of photometric redshifts (photo-z) for 3XMM sources. We searched for optical counterparts of 3XMM-DR6 sources outside the Galactic plane in the SDSS and Pan-STARRS surveys, with the addition of near- (NIR) and mid-infrared (MIR) data whenever possible (2MASS, UKIDSS, VISTA-VHS, and AllWISE). We used this photometry data set in combination with a training sample of 5157 X-ray selected sources and the MLZ-TPZ package, a supervised machine learning algorithm based on decision trees and random forests for the calculation of photo-z. We have estimated photo-z for 100,178 X-ray sources, about 50% of the total number of 3XMM sources (205,380) in the XMM-Newton fields selected to build this catalogue (4208 out of 9159). The accuracy of our results highly depends on the available photometric data, with a rate of outliers ranging from 4% for sources with data in the optical+NIR+MIR, up to $\sim$40% for sources with only optical data. We also addressed the reliability level of our results by studying the shape of the photo-z probability density distributions.