Optical identification of X-ray sources in the SRG/eROSITA survey of Lockman Hole

These are scripts and notebooks which are used to identify optical counterparts for X-ray sources detected in the SRG/eROSITA Lockman Hole survey (Bykov, Belvederskiy, Gilfanov 2022).

The analysis results I used in my paper. Necessary python packages: numpy, scipy, matplotlib, seaborn, pandas, astropy, sklearn, tensorflow (keras), tqdm, nway.

Note that the data (e.g. from Chandra, XMM, DESI) is not included and needed to be downloaded separately. eROSITA data from Lockman Hole will be available in the future. All data should be placed in ./0_data folder

The structure of the code is as follows:

• ./scripts
  the main scripts for the analysis are placed here. It containt a few files to manage the cross-match problem.
  • ./utils.py is used to set up pathes and contains some utility/plotting functions
  • ./cross_match_scripts.py contains functions for managing the catalogs, building machine learning models, and calculating the needed survey metrics
  • ./viewer.py contains functions for visualising optical fields around X-ray sources.

• ./notebooks
  the main notebooks for the analysis are stored here. It contains a few folders and separate notebooks to make the identification and its validation. The explanations are given in each notebook. The content of the folder is as follows (roughly in the order of the paper sections):

  • ./1_desi-photo-prior contains notebooks to build photometric priors with machine learning models.

    • 0_train-catalogs.ipynb creates a training data for prior model from Chandra and DESI LIS catalogs.
    • 1_train-catalogs-preprocessing contains notebooks to preprocess the training data.
    • 2_prior-learning-keras-nnmag.ipynb trains the prior models with keras and scikit-learn, and saves the models.
    • 3_prior-learning-fain-features.ipynb trains the prior based only on the distibution of the features (for faint sources).

  • ./2_desi-validation-catalog contains notebooks to cross-match the X-ray (eROSITA, Chandra, XMM) and optical (DESI LIS) catalogs to create a validation sample in the Lockman Hole area.

    • 0_catalog-preparation-ero-csc-xmm.ipynb prepares the catalogs from eROSITA, Chandra, and XMM.
    • 1_validation-counterparts.ipynb creates a sample of robust indentifications (true counterparts).
    • 2_validation-hostless.ipynb creates a sample of robust hostless X-ray sources (thue hostless).

  • ./3_desi-crossmatch contains notebooks for cross-match with the priors learned.

    • 0.0_catalog-preparation-erosita.ipynb prepares eROSITA data for cross-match with NWAY code.
    • 0.1_catalog-preparation-desi.ipynb prepares DESI LIS data for cross-match with NWAY code. Adds photometric priors according to the learned models and available data.
    • 1_magnn-match.ipynb cross-matches the eROSITA and DESI LIS catalogs with NWAY code. It uses slightly modified version of the code, which is available here.
    • 2_distance-match.ipynb the same as above but without photometric priors.
    • 3.1_results.ipynb presents the cross-match results and metrics.
    • 3.2_viewer.ipynb shows example of optical fields around X-ray sources for correct and incorrect identifications.