Gaia DR3 and ZTF timing of 11,668 Mira-like stars (2014-2025) shows a weak persistent period-change signal and no confirmed rapid period change; a simple O-C extrapolation test is biased
Question: do the pulsation periods of faint Gaia DR3 long-period variables change between Gaia DR3 (2014.6-2017.4) and ZTF DR24 (2018.2-2025.8), and does any star show a rapid change (|dP/dt|/P >= 1%/yr) like the thermal-pulse Mira T UMi? Method: we made ZTF r O-C curves of 11,668 large-amplitude LPVs (Dec > -28, G 12.5-18.5) and timed their Gaia G epochs with each star's ZTF template. We applied two population tests. Test A compares the Gaia phase offset with the extrapolated ZTF O-C curvature. Test B compares the within-ZTF period change with the Gaia-to-ZTF change and has no shared Gaia/ZTF term. We compared both tests with light-curve-level simulations (real sampling; cycle-length, amplitude and mean-level jitter) and with injections. Results: Test A gave rho = 0.28 in a 360-star pilot. A Gaussian O-C null gave 0.00 +- 0.06, but noise-only light-curve simulations give correlations of the same size and sign pattern as the full data (data 0.195 vs null 0.122 in well-sampled stars). Thus Test A is dominated by a sampling artifact. In Test B, well-sampled stars give rho = 0.080 [0.049, 0.109] (0.095 with DR3 catalogue periods), against -0.047 for noise only and 0.81 for a constant dP/dt population. Thus a small part of the measured period change persists over about 10 years. The spread of the within-ZTF period change is not larger than in the noise-only simulation. The conservative 95% bound is sd(dP/dt/P) <= 0.39%/yr. Four stars pass a strict rapid-change rule, against 0 of 2,594 simulated noise-only stars. But all four are faint, very red, poorly sampled plane or bulge Miras, two have close Gaia neighbours, and their Gaia phases and earlier OGLE/VSX periods do not follow a monotonic change. We list them as a watch list, not as thermal-pulse candidates. Limitations: simulation jitter settings are approximate, blending and censoring of faint minima are not simulated, and bright Miras are outside the sample.
Mira variables · long-period variables · period change · thermal pulse · O-C diagram · Gaia DR3 · ZTF · epoch photometry · null simulation · Eddington-Plakidis
References
7- useshttps://arxiv.org/abs/astro-ph/0504527
Templeton et al. 2005: period changes of 547 AAVSO Miras; context for the expected fraction of rapid changes.
- https://arxiv.org/abs/1905.00597
Molnar et al. 2019: T UMi thermal pulse; sets the 1%/yr rapid-change scale.
- useshttps://arxiv.org/abs/2206.05745
Gaia DR3 long-period variable catalogue (sample and periods).
- usesdoi:10.1051/0004-6361/202243940
Gaia DR3 data release (epoch photometry).
- usesdoi:10.1088/1538-3873/aaecbe
Zwicky Transient Facility (ZTF DR24 light curves).
- useshttps://db.ztf.snad.space
SNAD ZTF DR24 mirror used for light-curve access.
- useshttps://ui.adsabs.harvard.edu/abs/1929MNRAS..90...65E
Eddington and Plakidis random cycle-length model used as the O-C noise model.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesLong secondary periods of 2,877 LMC red giants lose phase between OGLE 1997-2009 and Gaia DR3 2014-2017, while short-period ellipsoidal binaries keep phase
Lesson that phase-persistence tests need light-curve-level nulls; AIP epoch-photometry recipe.
- usesZTF 2018-2025 period changes of 1,706 Gaia DR3 Cepheids are mostly not evolutionary: Gaia 2014-2017 timing confirms about 40% of the curvature, and signs are balanced
We adapted its ZTF DR24 (SNAD) and AIP Gaia epoch-photometry retrieval scripts and followed its warning that phase-extrapolation tests need light-curve-level null simulations.