Probabilistic positional association of catalogs of astrophysical sources: the Aspects code

TL;DR

This paper introduces a probabilistic method and the Aspects code for cross-identifying astrophysical sources in catalogs, accounting for positional uncertainties and different association assumptions, validated through extensive simulations.

Contribution

It presents a novel probabilistic framework and a publicly available Fortran code for source association, handling both several-to-one and one-to-one cases with unknown positional uncertainties.

Findings

The method accurately identifies the correct association assumption.

Estimators reliably recover unknown parameters from simulated data.

The approach is effective for large all-sky catalogs with up to 10^5 objects.

Abstract

We describe a probabilistic method of cross-identifying astrophysical sources in two catalogs from their positions and positional uncertainties. The probability that an object is associated with a source from the other catalog, or that it has no counterpart, is derived under two exclusive assumptions: first, the classical case of several-to-one associations, and then the more realistic but more difficult problem of one-to-one associations. In either case, the likelihood of observing the objects in the two catalogs at their effective positions is computed and a maximum likelihood estimator of the fraction of sources with a counterpart -- a quantity needed to compute the probabilities of association -- is built. When the positional uncertainty in one or both catalogs is unknown, this method may be used to estimate its typical value and even to study its dependence on the size of objects. It may also be applied when the true centers of a source and of its counterpart at another wavelength do not coincide. To compute the likelihood and association probabilities under the different assumptions, we developed a Fortran 95 code called "Aspects" ([asp{\epsilon}], "ASsociation PositionnellE/ProbabilistE de CaTalogues de Sources" in French); its source files are made freely available. To test Aspects, all-sky mock catalogs containing up to 10^5 objects were created, forcing either several-to-one or one-to-one associations. The analysis of these simulations confirms that, in both cases, the assumption with the highest likelihood is the right one and that estimators of unknown parameters built for the appropriate association model are reliable.