A Search for Exoplanet Candidates in TESS 2-min Light Curves using Joint Bayesian Detection
Jamila S. Taaki, Athol J. Kemball, Farzad Kamalabadi · 2025
Read A Search for Exoplanet Candidates in TESS 2-min Light Curves using Joint Bayesian Detection by Jamila S. Taaki, Athol J. Kemball, Farzad Kamalabadi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-18706.txt.
Introduction
Jamila Taaki, Athol Kemball, and Farzad Kamalabadi hunt for planets in data from the TESS space telescope, in a 2025 paper. Their detector looks for the small dips in starlight that mark a planet crossing its star. It works in a Bayesian way, learning the noise and the star's own flicker from priors it estimates as it goes, which guards against overfitting.
In injection tests the detector caught eighty percent of planted signals with a modest false-alarm rate, a hair better than a standard comparison method. On three years of real TESS data it recovered most known planet candidates. The team also ran it as an independent search for new candidates in the telescope's continuous viewing zone.
Catalog record
| CREATOR | Jamila S. Taaki, Athol J. Kemball, Farzad Kamalabadi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 72 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-18706
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A Search for Exoplanet Candidates in TESS 2-min Light Curves using Joint Bayesian Detection Jamila S. Taaki, Athol J. Kemball, Farzad Kamalabadi Abstract Journal: AJ This work used the following external packages: Astropy [ 5 , 6 , 7 ] , Scikit-learn [ 62 ] , Matplotlib [ 34 ] , NumPy [ 30 ] , SciPy [ 88 ] , and transit v0.3.0 ( https://github.com/dfm/transit ). abstract In this work, we apply an exploratory joint Bayesian transit detector [ 76 ] , previously evaluated using Kepler data, to the 2 min simple aperture photometry light curve data in the continuous viewing zone for the Transiting Exoplanet Survey Satellite (TESS) over three years of observation. The detector uses Bayesian priors, adaptively estimated, to model unknown systematic noise and stellar variability incorporated in a Neyman-Pearson likelihood ratio test for a candidate transit signal; a primary goal of the algorithm is to reduce overfitting. The detector was adapted to the TESS data and refined to improve outlier rejection and suppress false alarm detections in post-processing. The statistical performance of the detector was evaluated using transit injection tests, where the joint Bayesian detector achieves an 80.0 \% detection rate and a 19.1\% quasi-false alarm rate at a detection threshold \tau=10 ; this is a marginal, although not statistically significant, improvement of 0.2\% over a reference sequential detrending and detection algorithm. In addition, a full search of the input TESS data was performed to evaluate the recovery rate of known TESS objects of interest (TOI) and to perform an independent search for new exoplanet candidates. The joint detector has a 73\% recall rate and a 63\% detection rate for known TOI; the former considers a match against all detection statistics above threshold while the latter considers only the maximum detection statistic. I Introduction The Transiting Exoplanet Survey Satellite (TESS) 1 1 1 tess.mit.edu [ 65 ] was launched in 2018 to survey \sim 10^{6} nearby stars for planets. TESS is the successor to two decades of wide-field exoplanet transit surveys: primarily CoRoT 2 2 2 sci.esa.int/corot [ 8 ] and Kepler 3 3 3 keplerscience.arc.nasa.gov [ 11 ] . Such surveys are scientifically critical to understanding the population statistics of exoplanets by type and exoplanetary configuration [ 93 ] . Innovations in statistical inference and specialized data processing have been critical in achieving the photometric precision [ 20 ] required to robustly detect exoplanet transits in these data-intensive surveys [ 35 ] . In particular, a number of specialized approaches have been developed to remove residual instrumental systematics that may otherwise mask astrophysical transit signals in light curves [ 81 , 71 , 75 , 74 , 68 , 1 , 25 ] . In this work we adapt a prior algorithm for joint Bayesian transit detection and systematic noise characterization [ 76 ] we previously evaluated on Kepler data, and apply it to TESS 2-min simple aperture photometry (SAP) light curves. The goals of this paper are to explore the modifications required and the numerical performance of this algorithm when applied to data with different residual instrumental systematics and to independently search for new candidate exoplanets in the TESS data. The TESS and Kepler missions differ significantly in their observing strategy and cadence. TESS has an instantaneous field-of-view (FOV) of 24^{\circ}\times 96^{\circ} spanned by four individual cameras each containing four 2048\times 2048 pixel CCDs in their focal planes [ 67 ] . Individual pointings align the primary axis of the FOV along ecliptic longitude spanning a range of ecliptic latitudes from \sim 6^{\circ} to the ecliptic pole [ 67 ] . Each pointing of the FOV is of \sim 27 d duration and is known as a TESS sector [ 67 ] . Each observing year in the primary mission spans one ecliptic hemisphere over 13 sectors equispaced in ecliptic longitude and alternates to the other hemisphere the following year [ 67 ] . Approximately 2 \% of targets are continuously observed for the full year in the overlapping polar region known as the Continuous Viewing Zone (CVZ) [ 9 , 21 ] . TESS produces full-frame images (FFI) with a 30-min cadence which are analysed by the Quick-Look-Pipeline (QLP) [ 33 ] . The TESS Science Processing Operations Center (SPOC) pipeline [ 80 , 42 ] also produces SAP light curves with 2-min cadence for a list of targets defined in the TESS Input Catalog (TIC 4 4 4 tess.mit.edu/science/tess-input-catalogue/ ) [ 72 ] . To date, there have been 7525 TESS Objects of Interest (TOI) identified, 2262 of which are from the SPOC pipeline and a further 728 have been promoted from Community TESS Objects of Interest (CTOI) to TOI. Currently, 620 TOI have been confirmed as exoplanets 5 5 5 tess.mit.edu/toi-releases/ . In this work we apply our prior Bayesian detection method to the 2-min SAP light curves [ 82 , 42 , 59 ] obtained for the TESS CVZ targets over the first three years of observation. Each year of data is searched independently and where a target is observed in multiple years light curve data were not combined. A total of 5535 year-long light curves were searched for exoplanets with orbital periods in the range 1-100 days. Our search is complementary to the SPOC search for exoplanet candidates and other searches of TESS 2-min SAP data. These include the Planet-Hunters-TESS project [ 21 ] which identified 90 promising candidates in the first two years of TESS operations using a combination of visual inspection and semi-automated vetting. The Weird detector Wheeler & Kipping [91] , Chakraborty et al. [13] was used to search the first year of TESS data for any interesting periodic phenomena using a technique based on phase dispersion minimization and report 28 planet-like signals. Further Wong et al. [92] performed a phase curve study of 22 known transiting systems with light curves from the first year to obtain secondary-eclipse, brightness modulation measurements, as well as other transit properties. Holdsworth et al. [32] perform a search for, and identify 12, rapidly oscillating stars. We note the implementation of the Systematics-Insensitive Periodogram [ 3 , 31 ] , a Python tool to search TESS targets from the CVZ for long-period signals and identify stellar rotation rates. This method uses a noise model that incorporates systematics alongside sinusoid signal terms, which are jointly fitted to form a periodogram. TESS light curves have also been used to study stellar properties including an eclipsing binary (EB) catalog 6 6 6 http://tessebs.villanova.edu/ [ 63 ] , stellar flare statistics [ 24 ] and asteroseismology [ 53 ] . Wide-field transit surveys retain unmodeled residual instrumental errors after initial calibration and processing that compromise their high target photometric sensitivity [ 20 ] . The instrumental origins are diverse [ 37 , 87 ] and the systematic noise in the light curves is generally correlated both temporally and spatially across the imaging sensor [ 77 ] . The removal of residual systematics (detrending) is known to risk overfitting and distorting transit features [ 16 , 25 ] , particularly if the systematic and transit variability occur over similar timescales and are correlated. In addition, the residual systematics may mimic transit signatures, and consequently lower