[Paper Review] The OGLE Collection of Variable Stars. Eclipsing Binaries in the Magellanic System

M. Pawlak, I. Soszyński+11 more|arXiv (Cornell University)|Dec 19, 2016

Astronomical Observations and InstrumentationEngineering22 citations

TL;DR

This paper presents the largest catalog of eclipsing binaries in the Magellanic Clouds to date, containing 48,605 systems (40,204 in the LMC, 8,401 in the SMC), using I- and V-band time-series photometry from the OGLE-IV survey. The catalog was created via a two-step machine learning approach based on Random Forest classification, achieving over 80% completeness and detecting 16,374 new systems, with detailed light curves and photometric parameters provided for all objects.

ABSTRACT

We present the collection of eclipsing binaries in the Large and Small Magellanic Clouds, based on the OGLE survey. It contains 48605 systems, 40204 belonging to the LMC and 8401 to the SMC. Out of the total number of the presented binaries, 16374 are the new ones. We present the time-series photometry obtained for the selected objects during the fourth phase of the OGLE project. The catalog has been created using a two step machine learning procedure based on the Random Forest algorithm.

Motivation & Objective

• To compile a comprehensive, high-completeness catalog of eclipsing binaries in the Large and Small Magellanic Clouds using deep, long-baseline photometry.
• To improve the detection of eclipsing binaries, especially those with non-sinusoidal light curves, by overcoming limitations of traditional Fourier-based periodogram methods.
• To identify and characterize rare systems, such as those with Type II Cepheids, to support distance scale and stellar population studies.
• To provide a publicly accessible, high-precision dataset of time-series photometry and derived parameters for future astrophysical analysis.

Proposed method

• Period determination was performed using the Box-fitting Least Squares (BLS) algorithm, optimized for detecting eclipsing events with sharp, narrow dips.
• A two-step machine learning pipeline was employed: first, a Random Forest classifier was trained on known eclipsing binaries from the OGLE-III catalog; second, the model classified candidates from OGLE-IV data.
• Photometric data were collected in I- and V-bands using the 1.3-m Warsaw telescope at Las Campanas Observatory, with 100–700 epochs per star over 5.5 years (2010–2015).
• The BLS algorithm provided not only orbital periods but also eclipse depth, duration, and fractional eclipse depth as input features for the Random Forest classifier.
• Visual inspection was performed on all machine learning candidates to validate the final catalog and ensure high reliability.
• Photometric uncertainties were corrected using established methods from Skowron et al. (2016), improving photometric accuracy.

Experimental results

Research questions

• RQ1 What is the completeness and reliability of detecting eclipsing binaries in the Magellanic Clouds using a machine learning approach on OGLE-IV time-series photometry?
• RQ2 How do the period and magnitude distributions of eclipsing binaries in the Magellanic Clouds compare to those in the Milky Way, and what does this reveal about observational biases?
• RQ3 Can the machine learning pipeline detect rare or complex systems, such as eclipsing binaries with a Type II Cepheid component?
• RQ4 What is the spatial and color-magnitude distribution of eclipsing binaries in the LMC and SMC, and what does it imply about their stellar population?

Key findings

• The catalog contains 48,605 eclipsing binary systems, with 40,204 in the LMC and 8,401 in the SMC, making it the largest such collection in the Magellanic System.
• Of these, 16,374 systems are newly discovered, significantly expanding the known sample of eclipsing binaries in the region.
• The search completeness exceeds 80%, with high reliability up to I-band magnitude 19.6, and remains significant up to 20.4 mag.
• The period distribution peaks between 1 and 10 days, reflecting a bias toward longer periods due to the difficulty in detecting faint, short-period W UMa-type binaries.
• Spatial distribution and color-magnitude diagrams indicate that most systems are young, main-sequence A/B-type stars, though a significant population of red giant binaries exists, especially among ellipsoidal variables.
• The catalog includes OGLE-LMC-ECL-26653, the 14th known Type II Cepheid in an eclipsing binary system in the Magellanic Clouds, discovered via automated detection.