# Optical Cross-Match of SRG/eROSITA X-ray Sources Using the Deep Lockman Hole Survey as an Example

## TL;DR

This paper introduces a neural network-based method for optical identification of X-ray sources from wide-field surveys, achieving high accuracy in cross-matching with optical catalogs, exemplified on the SRG/eROSITA Lockman Hole survey.

## Contribution

The paper presents a novel neural network approach for probabilistic cross-matching of X-ray and optical sources, improving identification accuracy in large sky surveys.

## Key findings

- Cross-match precision reaches 94% overall.
- Precision improves to 97% for brighter X-ray sources.
- Method validated using Chandra and XMM-Newton catalogs.

## Abstract

We present a method for the optical identification of sources detected in wide-field X-ray sky surveys. We have constructed and trained a neural network model to characterise the photometric attributes of the populations of optical counterparts of X-ray sources and optical field objects. The photometric information processing result is used for the probabilistic cross-match of X-ray sources with optical DESI Legacy Imaging Surveys sources. The efficiency of the method is illustrated using the SRG/eROSITA Survey of Lockman Hole. To estimate the accuracy of the method, we have produced a validation sample based on the Chandra and XMM-Newton catalogues of X-ray sources. The cross-match precision in our method reaches 94% for the entire X-ray catalogue and 97% for sources with a flux $F_{\rm x, 0.5-2}>10^{-14}$ erg/s/cm$^2$. We discuss the further development of the optical identification model and the steps needed for its application to the SRG/eROSITA all-sky survey data.