🔭 astrophysics

Searching for core-collapse supernovae in binaries with ZTF

This paper presents a computationally efficient method using spline fits and Lomb-Scargle periodograms to search for binary-induced periodic oscillations in core-collapse supernovae within ZTF data, revealing that while long observation baselines can recover about half of such signals, shorter durations often miss them, leading to the identification of two potential candidates among 212 observed supernovae.

Original authors: Sylvia J. Zhu, Jakob Nordin, Emma de Oña-Wilhelmi, Andrea Ercolino

Published 2026-08-04

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Sylvia J. Zhu, Jakob Nordin, Emma de Oña-Wilhelmi, Andrea Ercolino

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 the universe as a grand, chaotic dance floor where stars are the dancers. Most stars, like solo performers, live their lives alone until they run out of fuel and explode in a spectacular finale called a supernova. But many stars are actually pairs, holding hands in a binary dance. When one of these partners explodes, it doesn't just vanish; it leaves behind a dense, compact core (like a neutron star) and a surviving partner. Sometimes, this new core starts "stealing" gas from its companion, creating a rhythmic, pulsing heartbeat in the light we see from Earth. Astronomers recently spotted this heartbeat in a specific supernova, SN 2022jli, which blinked with a perfect 12-day rhythm. This discovery is exciting because it proves that some exploding stars are still interacting with their partners long after the explosion, offering a rare glimpse into how these cosmic couples evolve and potentially merge to create gravitational waves. However, finding these rhythmic blinks in the vast, noisy data of the universe is like trying to hear a specific drumbeat in a stadium full of cheering fans.

This paper is essentially a recipe book and a stress test for a new method to find these hidden cosmic rhythms. The authors, led by Sylvia J. Zhu, wanted to know: if we look at thousands of supernovae with powerful telescopes like the Zwicky Transient Facility (ZTF), can we reliably spot these binary "heartbeats," or are they too easy to miss? To find out, they didn't just look at real stars; they built a massive virtual universe in their computers. They simulated 10,000 fake supernovae, injecting them with different types of rhythmic oscillations—some fast, some slow, some bright, some faint—to see if their search tools could find them.

The search method they developed is clever and computationally cheap. Imagine the light from a supernova as a wavy line on a graph. The main explosion creates a big, smooth hill that goes up and then slowly fades away. The binary rhythm is a tiny, fast wiggle riding on top of that hill. The authors' method uses a flexible mathematical "spline" (think of it as a bendy ruler) to trace the smooth hill and subtract it away, leaving only the tiny wiggles. Then, they use a tool called a Lomb-Scargle periodogram, which acts like a frequency analyzer, to see if those remaining wiggles form a repeating pattern.

The results of their simulations reveal a harsh reality about time. The method works beautifully if you have a long time to watch the star. If you observe a supernova for a long stretch—about 200 days with observations every three days—you can successfully recover about 50% of the binary rhythms they simulated. However, the success rate plummets if you don't have enough time. If you only watch for 30 days (which is typical for many surveys), you can only find a tiny fraction of the signals. Even with 75 days of observation, you only recover about 10%. This suggests that a huge number of binary supernovae might be hiding in plain sight, simply because we stop watching them too soon to catch their full rhythm.

The authors also tested how the "cadence" (how often you take a picture) affects the search. They found that watching every two days is slightly better than every three days, but watching every seven days makes it much harder to find short rhythms. They also discovered that long gaps in observation (like when clouds block the telescope for a week) create "ghost signals" that look like rhythms but aren't, leading to false alarms.

To prove their method works, they ran it on real data from the ZTF telescope. Out of 212 moderately bright supernovae, they found two strong candidates that showed signs of periodic oscillations, though they need more observation to be sure. They also successfully recovered the known 12-day rhythm of SN 2022jli and the 32-day rhythm of SN 2022esa, showing their tool can find what is already known. However, they struggled to find the rhythm in SN 2015ap, likely because the data was too sparse and full of gaps.

In conclusion, the paper suggests that while we have the tools to find these binary interactions, we need to be much more patient with our observations. Most supernovae are currently watched for too short a time to reveal their secret rhythms. The authors propose that future searches, perhaps with the upcoming LSST telescope, should focus on keeping the "eyes" on these stars for at least 100 days to catch the heartbeat. Until then, many of these cosmic couples might remain silent, their rhythmic dance undetected in the vast noise of the universe.

Technical Summary: Searching for core-collapse supernovae in binaries with ZTF

Problem Statement
Stripped-envelope core-collapse supernovae (CC SNe) often originate from massive stars in binary systems. Recent observations of SN 2022jli revealed significant periodic oscillations in its lightcurve, attributed to accretion from a binary companion onto a newborn compact object. While such periodicities have been identified in a few bright, well-observed transients (e.g., SN 2022esa, SN 2015ap), the population properties of these binary interactions remain poorly understood. Existing searches are often limited to individual, bright objects. There is a need for a systematic, computationally inexpensive method to search large datasets (such as those from the Zwicky Transient Facility, ZTF, and the future LSST) to determine how common these binary interactions are and to identify candidates for follow-up.

Methodology
The authors developed a search pipeline designed for real-time implementation within the AMPEL framework. The methodology involves three primary stages:

Simulation: The team simulated 10,000 CC SN lightcurves using sncosmo templates, augmented with binary interaction features. These features included periodic oscillations (modeled as sinusoids or skewed sawtooth waves) and rebrightening episodes (modeled by a fast-rise exponential decay function). Parameters varied included redshift, peak magnitude, oscillation period ( $T_{osc}$ ), amplitude ( $A_{osc}$ ), delay from peak ( $\Delta_{osc}$ ), and rebrightening characteristics. Simulations were performed for ZTF $g$ and $r$ bands with realistic noise and cadence (including random gaps and weather-related dropouts).
Signal Isolation: To isolate periodic signals from the underlying supernova evolution, the authors applied spline fits to the lightcurves. They tested various spline stiffness parameters ( $\lambda$ ) and start times relative to the peak ( $t_0$ ). The residuals were then analyzed using a generalized Lomb-Scargle periodogram.
Recovery Criteria: A signal was considered "recoverable" if the frequency of the maximum periodogram peak ( $P_{max}$ ) fell within a specific tolerance ( $\pm 0.0021 \text{ d}^{-1}$ ) of the true oscillation frequency or its aliases. The study evaluated performance across different observation durations ( $t_{obs}$ ), cadences (2, 3, and 7 days), and spline configurations.

Key Contributions and Results

Observation Duration and Cadence: The study finds that detection sensitivity is heavily dependent on the total observation duration.
- With a 200-day baseline and 3-day cadence, approximately 50% of the simulated binary oscillations are recoverable.
- This rate drops precipitously with shorter baselines: only ~10% are recoverable with a 75-day observation, and only a few percent with a 30-day observation.
- The authors conclude that at least 100 days of observations are necessary to effectively reject noise and recover a significant portion of the parameter space.
Spline Stiffness ( $\lambda$ ): The choice of spline stiffness is critical and depends on the oscillation period.
- A moderate stiffness ( $\lambda = 1$ ) is optimal for recovering shorter periods ( $T_{osc} < 50$ days).
- A stiffer spline ( $\lambda = 10$ ) is required to recover longer periods ( $T_{osc} > 50$ days) but risks subtracting out shorter-period signals or retaining too much underlying structure.
- The authors recommend using both $\lambda$ values to maximize sensitivity across the full range of potential binary periods.
Impact of Gaps and Cadence: Irregular observation gaps (common in real data) introduce aliases and complex structures in periodograms, increasing false alarms. While a 2-day cadence offers slightly better sensitivity than a 3-day cadence, a 7-day cadence (representative of LSST) significantly reduces sensitivity to short periods.
Follow-up Strategy: The authors demonstrate that high-cadence follow-up observations (e.g., 1-day cadence for 30 days) following a standard 75-day survey can achieve noise rejection levels comparable to a full 200-day survey, though sensitivity to long-period signals remains limited by the total time baseline.
Application to ZTF Data:
- The search was applied to 212 moderately bright ZTF CC SNe ( $0.03 < z < 0.06$ ).
- The method successfully recovered the known 12.4-day period of SN 2022jli and the 32-day period of SN 2022esa.
- The 8.4-day period of SN 2015ap was not recovered, likely due to the method's sensitivity to uneven sampling compared to Gaussian process approaches.
- Among the 212 candidates, two objects (SN 2022pss and SN 2022oeh) were identified as promising candidates for periodic oscillations, though both showed outlier peaks in only a single band.

Significance and Claims
The paper presents a "proof of concept" for a simple, computationally inexpensive search for periodic oscillations in CC SN lightcurves. The authors modestly claim that their method is suitable for setting up real-time searches in current and future surveys like ZTF and LSST.

The primary significance lies in the quantitative finding that a large fraction of binary interactions may have been missed in previous surveys due to insufficient observation durations (typically <30-75 days). The authors suggest that many SNe with binary companions may simply lack the necessary baseline to reveal periodicity. They propose that their search should serve as a filter to identify candidates for deeper, targeted follow-up (e.g., spectroscopy to detect line velocity shifts) rather than as a definitive detection tool on its own. The work highlights the trade-offs between observation duration, cadence, and spline parameters, providing a framework for optimizing future searches to characterize the population of binary-interacting supernovae.

More like this