Vetting TESS Exoplanet Candidates and Refining Their Transit Predictions
ISEF · 2026 Physics and Astronomy
Overview
When transit exoplanets pass in front of their parent stars, they produce periodic but faint dips in the star's light curve. To schedule follow-up observations with ground-based telescopes and space observatories it is essential to know exactly when these transits occur. Discovery catalogs like the TESS Objects of Interest (TOI) catalog originally publish these predicted ephemerides. Unfortunately, over time the precision of these ephemerides is known to worsen as the errors in the orbital period accumulate across many revolutions. An initial uncertainty of only a few minutes at discovery can grow to deviations of several hours after just a few years. However, precise ephemerides are critical for scheduling follow-up observations on expensive facilities like JWST. In this work I developed a pipeline that refines TESS ephemerides by fitting individual transits across multiple observation sectors and implemented a machine learning classifier to separate true exoplanets from false signals such as eclipsing binaries. I then applied the pipeline to 377 TESS Objects of Interest, where 19,506 individual transits were fitted, resulting in a median precision 4.2 times better than current catalog values. Planets with shorter orbital periods showed the highest refinements, exceeding thirteenfold improvements. Also, the classifier reached 95.9% accuracy when tested on 2,616 annotated samples. To confirm the real-world reliability of the improved ephemerides, I compared the results against 125 verified exoplanets with published periods, finding a median difference of just 645 milliseconds, with 76% matching within five seconds. This system will streamline observation scheduling, and similar methods could be adapted for thousands of remaining TESS candidates needing characterization.
Competition history
- ISEF 2026
Resources
Related projects
ISEF · 2018
Challenging Limitations: Using Deep Learning, Time Series Analysis, and Statistical Methods for Noise Reduction to Develop an Innovative Approach to Exoplanet Candidate Detection Using Earth-Based Telescopes
ISEF · 2019
Ground-based Followups of TESS Exoplanet Candidates
ISEF · 2023
A Novel Method in Improving the Accuracy of Exoplanet Classification Using Machine Learning
ISEF · 2022
A Specialized Searching Algorithm Towards the Improvement in Detection of Single-Transit and Long-Period Exoplanets in Brightness Time-Series Data
Closest projects by meaning, across every fair and year in the corpus.
Source: Regeneron International Science and Engineering Fair