# MIFAL: fully automated Multiple-Image Finder ALgorithm for strong-lens modelling -- proof of concept

## TL;DR

This paper presents MIFAL, an automated method for identifying multiple images in strong lensing clusters, combining arc detection, photometric redshifts, and initial lens models to improve mass distribution analysis.

## Contribution

The paper introduces a fully automated procedure for finding multiple images in strong lensing clusters, reducing reliance on manual identification and enabling large-scale analyses.

## Key findings

- Successfully recovered all known systems in the first cluster
- Recovered about half of the known systems in the second and third clusters
- Identified some false images depending on parameter thresholds

## Abstract

We outline a simple procedure designed for \emph{automatically} finding sets of multiple images in strong lensing (SL) clusters. We show that by combining (a) an arc-finding (or source extracting) program, (b) photometric redshift measurements, and (c) a preliminary light-traces-mass lens model, multiple-image systems can be identified in a fully automated (`blind') manner. The presented procedure yields an assessment of the likelihood of each arc to belong to one of the multiple-image systems, as well as the preferred redshift for the different systems. These could be then used to automatically constrain and refine the initial lens model for an accurate mass distribution. We apply this procedure to \emph{Cluster Lensing And Supernova with Hubble} observations of three galaxy clusters, MACS J0329.6-0211, MACS J1720.2+3536, and MACS J1931.8-2635, comparing the results to published SL analyses where multiple images were verified by eye on a particular basis. In the first cluster all originally identified systems are recovered by the automated procedure, and in the second and third clusters about half are recovered. Other known systems are not picked up, in part due to a crude choice of parameters, ambiguous photometric redshifts, or inaccuracy of the initial lens model. On top of real systems recovered, some false images are also mistakenly identified by the procedure, depending on the thresholds used. While further improvements to the procedure and a more thorough scrutinisation of its performance are warranted, the work constitutes another important step toward fully automatising SL analyses for studying mass distributions of large cluster samples.