Cycle shape and noise blocks change the four-orbit period test in 2012–2026 BAT data for 4U 1538-52
We tested the known long X-ray cycle of 4U 1538-52 against four times a specified binary period. We used 45,155 Swift/BAT observations outside eclipse from 2012 to 2026. The exact clock was 14.91324 days. We fit two harmonics, orbital terms, and changing means. With one shared cycle shape, the best period was 14.899849 days. With separate shapes in three published source intervals, it was 14.888345 days. The loss gain over the exact clock fell from 13.471 to 4.477. Wild bootstrap tests used 15–120 day noise blocks. They gave probabilities of 0.001–0.004 for the shared shape and 0.022–0.065 for separate shapes. The separate-shape result crossed the 0.05 threshold as block length changed. Its best period also changed with weights and the eclipse cut. These BAT data do not give a rejection that is stable across the tested shape and noise assumptions. The tests assume independent noise blocks and a fixed shape inside each source interval. They do not prove a continuous four-orbit clock or a physical resonance. We release the public input snapshot, provenance, code, bootstrap draws, numerical checks, and figures.
4U 1538-52 · Swift BAT · X-ray binaries · superorbital period · orbital commensurability · time series · wild bootstrap · correlated errors · reproducible research
References
8- usesA constant period fits mean half-amplitude BAT eclipse centers of 4U 1538-52 from 2005 to 2026 under a ramp model sci:sha256:06a8a46afe230dd9f4cf61eafd98c39943bc152dff985db1ea721a0b06e26693
The verified eclipse study informed the input quality rules and midpoint time treatment. We wrote new long-period regression and bootstrap code.
- usesdoi:10.3847/1538-4357/abc477
This prior work established the long cycle, its proximity to four binary periods, and its changing harmonic content. We test the stability of an exact-clock comparison.
- usesdoi:10.3847/1538-4357/ae5530
We use the published time divisions at MJD 57650 and 58800. Our BAT-only shape and noise tests differ from this combined BAT and MAXI analysis.
- usesdoi:10.3847/1538-4357/ab03d3
The paper supplies the distinct orbital and eclipse periods, reference epoch, and conditional period derivative for the specified clock checks.
- usesdoi:10.1088/0067-0049/209/1/14
We use public 15–50 keV pointing light curves from the Swift/BAT transient monitor.
- usesdoi:10.3847/1538-4357/ac7c74
Astropy reads the FITS input and supplies the time conversion and built-in barycentric ephemeris.
- usesdoi:10.1038/s41592-019-0686-2
SciPy supplies bounded scalar optimization for the local period fit.
- usesdoi:10.1162/rest.90.3.414
This is a reference for wild cluster bootstrap inference. We apply common random residual signs in time blocks to a local period-profile gain. We do not claim a new general resampling method.
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesCycle profiles and date cuts change tests of four fixed eclipse periods in 2005–2026 BAT data for 4U 1538−52
This earlier study tests the same broad clock question. We use it for comparison and add a four-profile date-cut test. The FITS input bytes are the same. We wrote our code separately.