🔭 astrophysics

Finding Circumbinary Planets: A Semi-Automated Transit Search of TESS Eclipsing Binaries

This paper introduces a semi-automated framework called ${\tt mono-cbp}$ for detecting circumbinary planets in TESS eclipsing binary light curves, which successfully recovers over 50% of known transiting circumbinary planets and identifies one new candidate while demonstrating strong performance in simulated tests.

Original authors: Benjamin D. R. Davies, David J. A. Brown, Samuel Gill, Jenni R. French

Published 2026-04-13

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Benjamin D. R. Davies, David J. A. Brown, Samuel Gill, Jenni R. French

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to find a tiny, dark moth flying past a pair of massive, flashing spotlights. That is essentially what astronomers are doing when they hunt for circumbinary planets—planets that orbit two stars at once.

This paper introduces a new, semi-automated "moth-hunting" tool called mono-cbp, designed to scan data from NASA's TESS space telescope to find these elusive worlds. Here is a breakdown of how they did it, using simple analogies.

The Challenge: Finding a Needle in a Haystack of Flashlights

Most planets orbit a single star. But some orbit two stars (like Tatooine in Star Wars). These are called circumbinary planets.

The Problem: The two stars are constantly eclipsing each other (one blocking the other), creating huge, bright dips in the light curve. A planet passing in front of them is like a tiny moth fluttering past a lighthouse; its shadow is incredibly faint compared to the stars' own "flashing."
The Complication: Unlike planets around single stars, these planets don't have a strict schedule. Because the two stars are dancing around each other, the planet's transit times shift wildly (like a train that is always late or early). This makes standard search algorithms, which look for perfect patterns, useless.

The Solution: The "Mono-CBP" Framework

The authors built a software pipeline to find these planets by looking for single, isolated events rather than a repeating pattern. Think of it like looking for a single drop of rain in a storm rather than waiting for a specific pattern of raindrops.

Here is how their "Mono-CBP" tool works, step-by-step:

1. The Noise-Canceling Headphones (Masking & Detrending)

First, the software has to ignore the massive "flashing" of the two stars.

Masking: They use a known schedule of the stars' eclipses to digitally "black out" those parts of the data. It's like putting a piece of tape over the bright parts of a photo so you can see the tiny details in the shadows.
Detrending: Stars aren't perfectly steady; they twinkle, spot, and pulse. The software uses a "cosine" and "biweight" filter (think of it as a sophisticated noise-canceling headphone) to smooth out the background wiggles so the tiny planet signal doesn't get lost.

2. The "One-Two Punch" Detector (The Search)

The software scans the cleaned-up light curve for any sudden dip in brightness.

It doesn't care if the dip happens once or twice. It just looks for a "threshold crossing event" (TCE)—a moment where the light drops significantly.
It can even spot a "one-two punch," where a planet passes in front of both stars in the same orbit, creating two dips close together.

3. The Bouncer at the Club (Vetting)

The search is very sensitive, so it catches a lot of "false alarms" (like a glitch in the camera or a star spot). The software acts as a bouncer with a checklist:

Is it a glitch? If the dip happens at the exact same time for thousands of other stars, it's a camera error, not a planet.
Is it a trend? If the dip looks like a slow slope rather than a sharp box, it's probably just the star breathing, not a planet.
The Model Match: The software tries to fit the dip into different shapes: a planet transit, a sine wave, or a sudden jump. If the "planet shape" fits best, it gets a "Candidate" badge.

The Results: One New Hope

The team applied this tool to 512 binary star systems from the TESS catalog.

The Catch: They found one promising candidate (TIC 319011894).
The Mystery: This system has two stars eclipsing every 8.3 days. There is a tiny, extra dip in the light that might be a planet. However, it could also be a trick of the light or a second, very faint eclipse. It needs more observation to confirm.
The Good News: They tested their tool on 14 known circumbinary planets (discovered by the Kepler mission). Even though TESS data is "noisier" than Kepler's, their tool successfully found more than 50% of the transits for every planet, and 75% for most of them. This proves the tool works!

Why This Matters

Finding these planets is hard because:

They are rare: We only know of about 15 of them so far.
They are unpredictable: Their orbits wobble, making them hard to track.
They are faint: The planet's shadow is tiny compared to the stars.

This new framework is like a high-tech metal detector that can scan a beach for a specific type of coin, even if the waves are crashing and the sand is shifting. It allows astronomers to scan massive amounts of data quickly without needing a supercomputer for every single star.

The Bottom Line

The authors have built a reliable, semi-automated way to hunt for planets in binary star systems. While they only found one new candidate so far, the method is proven to work. As TESS continues to collect data and future telescopes (like PLATO) come online, this tool will help us build a bigger picture of how planets form in these chaotic, double-star neighborhoods.

In short: They built a smart filter to ignore the noise of dancing stars so we can finally hear the whisper of a planet passing by.

1. Problem Statement

Circumbinary planets (CBPs)—planets orbiting the center of mass of a stellar binary—are rare in the known exoplanet population, with only 14 confirmed systems to date (12 from Kepler, 2 from TESS). Detecting them presents unique challenges:

Aperiodicity: Strong Transit Timing Variations (TTVs) caused by the binary's orbital motion and dynamical precession make standard periodic search algorithms (e.g., Box Least Squares, BLS) ineffective.
Transit Duration Variations (TDVs): The transit duration changes significantly depending on the binary's orbital phase, complicating shape-based detection.
Data Limitations: The TESS mission observes in 27-day sectors, providing shorter continuous baselines than Kepler, which reduces the probability of capturing multiple transits for long-period CBPs.
Light Curve Complexity: Eclipsing binary (EB) light curves are dominated by deep stellar eclipses and intrinsic variability (spots, ellipsoidal modulation), which can obscure shallow planetary transits.

The authors aim to develop a robust, semi-automated framework to search for single-transit events (monotransits) in TESS EB light curves to identify new CBP candidates without requiring prior knowledge of binary masses or strict periodicity.

2. Methodology: The `mono-cbp` Framework

The authors developed a four-stage semi-automated pipeline named mono-cbp:

A. Stellar Eclipse Masking

Approach: Instead of modeling and subtracting eclipses (which risks creating artifacts), the pipeline masks the data.
Implementation: Using known ephemerides, data within $\pm w/2$ of the primary and secondary eclipse midpoints are removed. This prevents stellar eclipses from dominating the search but reduces sensitivity to "syzygy" events (where a planet transits during a stellar eclipse).

B. Detrending

Challenge: EBs exhibit complex variability (spots, reflection, Doppler beaming) that standard detrending might remove along with the transit signal.
Two-Step Process:
1. Iterative Cosine Detrending: Uses the wotan package to remove quasi-periodic variability. It fits cosine functions with decreasing window lengths (starting at 12 days) until no significant periodic peaks (False Alarm Probability < 1%) remain in the Lomb-Scargle periodogram.
2. Biweight Detrending: Applies Tukey's biweight time-windowed slider across a grid of 21 window lengths (1.0 to 3.0 days).
Independence Check: An event is flagged as "detrending-independent" only if detected in $\ge 18/21$ of the different window-length light curves. This filters out artifacts caused by specific detrending choices.

C. Transit Search (Threshold Crossing Events)

Algorithm: Scans for 3 consecutive cadences (binned to 30 minutes) falling below a threshold of $3 \times$ Median Absolute Deviation (MAD) of the local flux.
Output: Identifies Threshold Crossing Events (TCEs) and estimates their depth ( $\delta$ ) and duration ( $t_{dur}$ ).

D. Automated Vetting & Classification

Initial Filters:
- Skye Metric: Flags events correlated with other targets (instrumental systematics).
- SNR Cut: Requires SNR > 5 (calculated based on depth, duration, and local scatter).
- Duration Cut: Rejects events > 1 day (likely systematics).
Model Comparison: For remaining candidates, the pipeline fits four models: Transit, Sinusoid, Step, and Linear.
- Uses Markov Chain Monte Carlo (MCMC) to sample posteriors.
- Classifies events based on the Akaike Information Criterion (AIC) and Root Mean Square Error (RMSE).
- Events best fit by the transit model (with low RMSE) are flagged for manual inspection.

3. Key Contributions

Novel Framework: Introduction of mono-cbp, a flexible search method specifically designed for the aperiodic and variable nature of CBP transits in EBs.
Single-Transit Focus: Unlike methods requiring multiple transits or known periods, this approach targets individual high-SNR events, making it suitable for TESS's fragmented observing windows.
Robust Vetting: A multi-layered vetting process combining statistical metrics (Skye), signal-to-noise thresholds, and physical model comparison to minimize false positives.
Validation: The framework was rigorously tested against known CBPs and through extensive injection-recovery simulations.

4. Results

A. Search of TESS Eclipsing Binary Catalogue (TEBC)

Sample: 512 detached EBs with periods > 7 days (selected from 4,584 in TEBC).
Search Volume: Analyzed 3,808 light curves (up to Sector 79).
Outcome:
- Identified 7,176 initial TCEs.
- After automated vetting and model comparison, 1,647 events were classified as Transit/Ambiguous Transit.
- Manual Inspection: Reduced the list to one candidate: TIC 319011894.
  - Candidate: A single event in Sector 7 with a depth of ~0.2% and duration of ~0.29 days.
  - Status: The host is an 8.3-day EB. The event is astrophysical and on-target but lacks a second observed transit. It requires ground-based follow-up to confirm if it is a CBP or a secondary eclipse.

B. Validation on Known CBPs

Test Set: Applied to 14 known transiting CBPs (12 Kepler, 2 TESS).
Recovery Rate:
- Achieved $\ge 50\%$ recovery for all 14 planets.
- Achieved $> 75\%$ recovery for 9 of the 14 planets.
- Performance correlated strongly with Transit Signal-to-Noise Ratio (SNR); systems with SNR > 10 were recovered nearly 100%.

C. Injection-Recovery Tests

Method: Injected 49,000 synthetic transits (varying depths 0.1–1.0% and durations 0.1–1.0 days) into the sample light curves.
Findings:
- Sensitivity: Best recovery for deeper ( $\ge 0.46\%$ ) and intermediate-duration ( $0.25–0.55$ days) transits.
- Limitations: Long-duration transits ( $> 0.85$ days) suffer from signal loss due to the detrending process (biweight windows remove long trends). Shallow transits are harder to detect in high-noise light curves.
- False Positive Rate: Automated vetting reduced the raw false positive rate from ~0.84 per light curve to ~0.007 for confirmed transit candidates.

D. Detection Sensitivity Estimates

For a typical binary in the sample ( $T \approx 10.2$ $T \approx 10.2$ , low mass ratio), mono-cbp can detect:
- Sub-Saturns ( $R \approx 0.67 R_J$ ): ~61% efficiency (for $t_{dur} \le 0.85$ d).
- Neptunes ( $R \approx 0.46 R_J$ ): ~43% efficiency.
- For equal-mass binaries (maximum dilution), sensitivity drops to Jupiter-sized planets for short durations.

5. Significance and Future Directions

Scalability: The framework is computationally efficient and applicable to the hundreds of thousands of EBs identified in TESS Full Frame Images (FFIs).
Candidate Generation: While single-transit detections cannot confirm a CBP without follow-up, mono-cbp effectively identifies high-priority targets for spectroscopic and photometric follow-up.
Future Missions: The methodology is well-suited for future surveys like PLATO, which will offer longer baselines (2+ years) ideal for capturing multiple transits of long-period CBPs.
Scientific Impact: By expanding the sample of CBP candidates, this work aids in constraining the formation and dynamical evolution of planets in binary systems, particularly addressing the "pile-up" of planets near the inner stability limit and the lack of CBPs in short-period binaries.

In conclusion, mono-cbp provides a validated, semi-automated tool for the next generation of circumbinary planet searches, successfully balancing the detection of real signals against the high noise and variability inherent in eclipsing binary light curves.