A Search for Supernovae That Fade to a Rhythm

Supernova light curves usually trace a broad rise and decline. But a few stripped-envelope supernovae—explosions whose stars had lost much of their outer hydrogen, and sometimes helium, before collapse—show smaller, repeated changes during the fading phase. Such periodic modulations may be linked to binary progenitor systems. To find out whether these cases are rare curiosities or part of a larger group, Asaf Horowicz, Eran O. Ofek and Avishay Gal-Yam searched observations from the Zwicky Transient Facility (ZTF).

The study focused on 34 supernovae with usable measurements in both of ZTF’s optical bands, g and r. The researchers first gathered light curves for a broader set of candidates, then required at least 20 detections per band for a two-band search. This left a selected sample: objects with adequate observations, not a complete census of all stripped-envelope supernovae.

Separating a rhythm from a fading trend

A challenge is that a supernova’s overall decline can masquerade as a slow oscillation—or absorb part of a real one. Rather than first removing a smooth trend and then searching what remains, the team fitted the trend and a trial repeating signal together. They tested many possible periods, measuring whether adding a sinusoid substantially improved the fit. A candidate also had to show a sufficiently large modulation and compatible periods in both bands.

The researchers checked the search with bootstrap tests and simulations. In injection-recovery tests, they added artificial signals to the actual observations, preserving their irregular timing, gaps and measurement uncertainties, then checked how often the analysis recovered them. This matters because the survey is more sensitive to strong, relatively short-period signals than to weak or long-period ones. Long rhythms are especially hard to test when a supernova is visible for only a limited time.

Three two-band candidates, with different levels of confidence

The pipeline recovered the previously reported rhythms in SN 2022jli and SN 2022esa, and identified SN 2020sgf as another candidate. SN 2022jli’s roughly 12-day modulation is recovered consistently across the tested polynomial baselines. SN 2022esa’s signal is near the study’s significance cutoff in one band, though its matching signal in the other band and additional checks led the authors to include it. Its fitted period also shifts somewhat with analysis choices.

SN 2020sgf has a roughly 30–36-day signal in the two bands, but its period is less secure: another analysis favors a period near twice as long. The authors also found that its brightness residuals track the Moon’s illumination in ZTF observations and in an independent ATLAS band. That association does not establish that moonlight caused the apparent rhythm; the paper says the available data cannot distinguish a real signal, an observational effect, or a combination. The candidate is included in the population analysis, but warrants caution.

The team therefore did not simply treat three detections among 34 objects as the true fraction. Instead, it combined simulated survey sensitivity with population-synthesis models that predict possible period distributions. Under those assumptions, models with intrinsic periodic fractions on the order of 20% can be consistent with the detections. This is an illustrative, model-dependent inference, not a precise measurement: it depends on the assumed period and amplitude distributions, and the models do not predict observable amplitudes directly.

A promising clue, not a population census

Important uncertainties remain. The sample favors brighter events, and the search is primarily sensitive to periods of a few tens of days; it cannot adequately test some much longer periods predicted by models. Only three two-band candidates underpin the population inference, and two are less robust to choices in the analysis than SN 2022jli. The authors also note that two candidate periods lie near the Moon’s monthly cycle, a possible systematic that needs further investigation. These limitations mean the result should not be read as proving that one in five stripped-envelope supernovae has a periodic signal.

Periodic light changes could offer indirect evidence about binary progenitors, but this survey does not establish that every candidate’s rhythm has the same physical cause. Its contribution is both a set of candidates and a method for testing them across larger samples. More events, denser observations and longer monitoring should help determine how often these rhythms occur and whether they share an origin. The paper is a preprint, so its claims should be considered in that context rather than as a settled population estimate.