Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Automated all-sky detection of γ Doradus / δ Scuti hybrids in TESS data from positive unlabelled (PU) learning

Mykyta Kliapets, Pablo Huijse, Andrew Tkachenko, Alex Kemp, Dario J. Fritzewski, Daniel Hey, Conny Aerts

TL;DR

This paper tackles all-sky detection of γ Doradus/δ Scuti hybrids in TESS data using positive unlabelled (PU) learning, addressing the rarity of these objects among AF stars. It systematically evaluates TGLC light curves against QLP and Kepler baselines for robust frequency extraction, and introduces a dynamic binning feature set combined with PU bagging to identify new hybrids. The authors report a high recall of $93.04\%$ for known hybrids and deliver a catalog of 62,026 new candidate hybrid light curves with per-sector probabilities, enabling targeted follow-up and population studies. The work demonstrates that TGLC is competitive with QLP for hybrid studies, expands the candidate pool substantially, and paves the way for unsupervised clustering and advanced asteroseismic analyses, including preparation for PLATO field applications.

Abstract

The Transiting Exoplanet Survey Satellite (TESS) mission has observed hundreds of millions of stars, substantially contributing to the available pool of high-precision photometric space data. Among them are the relatively rare $γ$ Doradus / $δ$ Scuti ($γ$ Dor / $δ$ Sct) hybrid pulsators, which have been previously studied using Kepler data. These stars are perfect laboratories to probe both inner and outer interior stellar layers thanks to them exhibiting both pressure and gravity modes. We seek to classify an all-sky sample of AF stars observed by TESS to find previously undiscovered hybrid pulsators and supply them in a catalogue of candidates. We also aim to compare the light curves produced with the TESS-Gaia Light Curve (TGLC) pipeline, currently underused in variability studies, with other publicly available light curves. We compared dominant and secondary frequencies of confirmed hybrid pulsators in Kepler, extended mission Quick Look Pipeline (QLP) data, and nominal and extended mission TGLC data. We then used a feature-based positive unlabelled (PU) learning classifier to search for new hybrid pulsators amongst TESS AF stars and investigated the properties of the detected populations. We find that the variability of confirmed hybrids in TGLC agrees well with the one occurring in QLP light curves and has a high recovery rate of \kepler-extracted frequencies. Our `smart binning' method allows for robust extraction of hybrids from large unlabelled datasets, with an average out-of-bag prediction for test set hybrids at 93.04\%. The analysis of dominant frequencies in high-probability candidates shows that we find more pressure-mode dominant hybrids. Our catalogue includes 62,026 new candidate light curves from the nominal and extended TESS missions, with individual probabilities of being a hybrid in each available sector.

Automated all-sky detection of γ Doradus / δ Scuti hybrids in TESS data from positive unlabelled (PU) learning

TL;DR

This paper tackles all-sky detection of γ Doradus/δ Scuti hybrids in TESS data using positive unlabelled (PU) learning, addressing the rarity of these objects among AF stars. It systematically evaluates TGLC light curves against QLP and Kepler baselines for robust frequency extraction, and introduces a dynamic binning feature set combined with PU bagging to identify new hybrids. The authors report a high recall of for known hybrids and deliver a catalog of 62,026 new candidate hybrid light curves with per-sector probabilities, enabling targeted follow-up and population studies. The work demonstrates that TGLC is competitive with QLP for hybrid studies, expands the candidate pool substantially, and paves the way for unsupervised clustering and advanced asteroseismic analyses, including preparation for PLATO field applications.

Abstract

The Transiting Exoplanet Survey Satellite (TESS) mission has observed hundreds of millions of stars, substantially contributing to the available pool of high-precision photometric space data. Among them are the relatively rare Doradus / Scuti ( Dor / Sct) hybrid pulsators, which have been previously studied using Kepler data. These stars are perfect laboratories to probe both inner and outer interior stellar layers thanks to them exhibiting both pressure and gravity modes. We seek to classify an all-sky sample of AF stars observed by TESS to find previously undiscovered hybrid pulsators and supply them in a catalogue of candidates. We also aim to compare the light curves produced with the TESS-Gaia Light Curve (TGLC) pipeline, currently underused in variability studies, with other publicly available light curves. We compared dominant and secondary frequencies of confirmed hybrid pulsators in Kepler, extended mission Quick Look Pipeline (QLP) data, and nominal and extended mission TGLC data. We then used a feature-based positive unlabelled (PU) learning classifier to search for new hybrid pulsators amongst TESS AF stars and investigated the properties of the detected populations. We find that the variability of confirmed hybrids in TGLC agrees well with the one occurring in QLP light curves and has a high recovery rate of \kepler-extracted frequencies. Our `smart binning' method allows for robust extraction of hybrids from large unlabelled datasets, with an average out-of-bag prediction for test set hybrids at 93.04\%. The analysis of dominant frequencies in high-probability candidates shows that we find more pressure-mode dominant hybrids. Our catalogue includes 62,026 new candidate light curves from the nominal and extended TESS missions, with individual probabilities of being a hybrid in each available sector.

Paper Structure

This paper contains 27 sections, 15 figures.

Table of Contents

  1. Introduction
  2. TGLC for hybrid pulsator studies
  3. The TGLC pipeline
  4. Dominant and secondary variability of labelled hybrids
  5. Effect of Gaia-informed light curve extraction: TGLC vs. QLP light curves
  6. Longer baselines for g-mode detection: TESS TGLC and QLP vs. Kepler light curves
  7. Finer sampling for p-mode detection: primary- vs extended-mission TGLC light curves
  8. PU learning
  9. Application of PU learning to TESS data
  10. The dataset
  11. Positive (labelled) dataset of confirmed hybrids
  12. Unlabelled dataset of AF stars in TESS
  13. Classification features
  14. PU bagging classifier
  15. Classification evaluation & error analysis
  16. ...and 12 more sections

Figures (15)

  • Figure 1: Aperture (top) and PSF (bottom) light curve (grey) and Lomb-Scargle periodogram (overplotted in burgundy, clipped at 30 d$^{-1}$ for visibility) of a confirmed hybrid pulsator (DR3 2147267632621883776) detected in TESS sector 40. Note how this target is p- and g-mode dominated on the aperture and PSF light curves, respectively.
  • Figure 2: Comparison of Kepler$f_{1}$ (purple histogram) recovery efficiency in QLP (gold), TGLC (genuine and binned to 30-minute cadence in teal and grey, respectively).
  • Figure 3: Graphic representation of binning approaches applied to a confirmed hybrid pulsator (DR3 1653802003712212992). Top plot: bins from skarka2022 and skarka2024, centre plot: bins from hey2024, bottom plot: this work.
  • Figure 4: Combined out-of-bag predictions for the unlabelled instances from nominal and extended missions with 15% of the total data (golden) and hidden hybrids from nominal (teal) and extended (purple) missions. Teal (94.25%) and purple (90.98%) dotted lines represent mean predictions for hidden hybrids from nominal and extended missions, respectively. Golden dotted line represents the grand mean for all unlabelled targets (7.57%) used for training.
  • Figure 5: Out-of-bag predictions for pure $\delta$ Sct stars (purple), eclipsing binaries (burgundy), pure $\gamma$ Dor / SPB pulsators (teal), and rotational variables (golden) from TESS sectors 1 and 14 from hey2024.
  • ...and 10 more figures