Identifying highly magnetized white dwarfs: A dimensionality reduction framework for estimating magnetic fields

TL;DR

This paper introduces a machine learning framework using UMAP and DBSCAN to classify white dwarf subpopulations and estimate magnetic fields, aiding the discovery of highly magnetized white dwarfs without direct magnetic measurements.

Contribution

It presents a novel application of unsupervised ML techniques for classifying white dwarfs and estimating magnetic fields, enhancing detection of magnetized white dwarfs beyond traditional methods.

Findings

Successfully classified white dwarf subpopulations

Differentiated magnetic from non-magnetic white dwarfs

Estimated magnetic fields for unmeasured white dwarfs

Abstract

Magnetic fields play a crucial role in compact object physics, particularly in white dwarfs (WDs), where high densities can sustain strong magnetic fields. Observations have revealed magnetized WDs (MWDs) with surface fields reaching approximately $10^9\rm\,G$, although high-field MWDs are fewer in number in current catalogs owing to their intrinsic faintness and limitations in conventional electromagnetic surveys. In this study, we apply unsupervised machine learning (ML) techniques to systematically analyze a sample of hydrogen-atmosphere (DA) WDs. Using Uniform Manifold Approximation and Projection (UMAP) for dimensionality reduction and Density-Based Spatial Clustering of Applications with Noise (DBSCAN) for cluster identification, we classify distinct subpopulations within the DA WD sample. Each cluster exhibits unique intrinsic properties such as mass, surface gravity, temperature, and age. Our analysis further reveals that these subgroups effectively differentiate MWDs from non-magnetic or weakly magnetic counterparts. Moreover, utilizing a set of previously confirmed MWDs, we estimate the field strengths of all other MWDs lacking magnetic field measurements. This study underscores the effectiveness of ML-based approaches in astrophysical discovery, particularly detecting magnetized compact objects when direct measurements are unavailable.