Linking the morphology of young stellar objects to their evolutionary stages with self-organizing maps

¹ Universität Wien, Institut für Astrophysik, Türkenschanzstrasse 17, 1180 Wien, Austria
² Natural History Museum Vienna, Burgring 7, 1010 Vienna, Austria
³ Université de Genève, Department of Astronomy, Chemin Pegasi 51, 1290 Versoix, Switzerland
⁴ Konkoly Observatory, Research Centre for Astronomy and Earth Sciences, Hungarian Research Network (HUN-REN), Konkoly Thege Miklós Út 15-17, 1121 Budapest, Hungary
⁵ CSFK, MTA Centre of Excellence, Budapest, Konkoly Thege Miklós út 15-17, 1121 Budapest, Hungary
⁶ HITS gGmbH, Astroinformatics, Schloss-Wolfsbrunnenweg 35, 69118 Heidelberg, Germany
⁷ Department of Experimental Physics, Institute of Physics, University of Szeged, Dóm tér 9, 6720 Szeged, Hungary

^★ Corresponding author: This email address is being protected from spambots. You need JavaScript enabled to view it.

Received: 12 December 2024
Accepted: 22 September 2025

Abstract

Context. Many studies in the past few decades have investigated the evolution of young stellar objects based on their spectral energy distribution. This distribution is heavily affected not only by the evolutionary stage, but also by the morphology of the forming star. This study is part of the NEMESIS project, which aims to revisit star formation with the aid of machine-learning techniques and provides the framework for this work.

Aims. In a first effort toward a novel spectro-morphological classification, we analyzed the morphologies of young stellar objects and linked them to the currently used observational classes. Thereby, we laid the foundation for a spectro-morphological classification and applied the insights learned in this study in a future revisited classification scheme.

Methods. We obtained archival high-resolution survey images from VISTA for approximately 10000 young stellar object candidates from the literature toward the Orion star formation complex. Using a self-organizing map algorithm, which is an unsupervised machinelearning method, we created a grid of morphological prototypes from near- and mid-infrared images. Furthermore, we determined the prototypes that best represent the different observational classes we derived from the infrared spectral index via Bayesian inference.

Results. We present our grids of morphological prototypes of young stellar objects in the near-infrared. The prototypes were created from observational data alone. They are thus independent of theoretical models. In addition, we show maps that indicate the probability for a prototype to belong to any of the observational classes.

Conclusions. Self-organizing maps created from near-infrared images are a useful tool, with limitations, for identifying the characteristic morphologies of young stellar objects in different evolutionary stages. This first step lays the foundation for a spectro-morphological classification of young stellar objects that is to be developed in the future.