Twenty-two years of X-ray timing suggest the disk around pulsar 4U 1626−67 may have flipped from rotating backward to forward in 2008, turning years of spin-down into spin-up — though the evidence remains indirect.

For decades, the ultracompact X-ray binary 4U 1626−67 has behaved like a machine that periodically changes gear. Its neutron star spun faster for years, slowed for roughly 18 years, then began accelerating again in 2008. A new analysis of 22.7 years of pulse timing says the middle era is more naturally explained if the accretion disk was rotating backward and then returned to forward rotation.

That is an intriguing interpretation, not a direct sighting of a disk changing direction. The Astrophysical Journal study reports only weak preference for the flip after accounting for the full range of model parameters. The timing data used in the analysis do not cover the reversal itself, leaving the pivotal event inside a gap between two observatories.

The result matters because accretion disks are usually expected to inherit the forward orbital motion of the binary system feeding them. A durable disk turning the other way would demand an explanation of how its flow was reorganized. It could also provide a direct way to account for a neutron star losing angular momentum while it continued to accrete.

A 7.67-second clock in a 42-minute binary

4U 1626−67 was discovered during the early years of X-ray astronomy by the Uhuru satellite. It lies at an estimated distance of about 3.5 kiloparsecs and contains a neutron star with a surface magnetic field near three trillion gauss. Its very low-mass, hydrogen-depleted companion is known optically as KZ TrA.

The pair completes an orbit in approximately 42 minutes, placing it among ultracompact X-ray binaries. Material pulled from the companion forms a disk around the neutron star. Near the star, the magnetic field disrupts the disk and channels plasma toward the magnetic poles. The hot infalling gas emits X-rays in a beam that sweeps past Earth as the star rotates.

Those pulses repeat at about 130 millihertz, equivalent to one rotation every 7.67 seconds. The period can be measured with extraordinary precision. When successive pulses arrive slightly earlier, the star is spinning up. When they drift later, it is spinning down. That turns a distant point of X-ray light into a sensitive record of the torque exerted by incoming material.

Three long torque eras

The historical record shows 4U 1626−67 spinning up from the late 1970s until about 1990. It then reversed into an approximately 18-year spin-down episode. Around 2008 the sign changed again, and the neutron star began accelerating. Each switch happened much faster than the long, comparatively steady intervals on either side.

The 2008 event was not merely a change in the slope of a timing graph. A long-term observational study found that the X-ray flux rose by more than a factor of two around the second reversal. The pulse profile changed, and a quasi-periodic oscillation near 48 millihertz that had been seen in the spin-down era disappeared.

Those accompanying changes suggest a reorganization of the inner accretion flow. They do not identify its direction. Brightness can track the rate at which matter reaches the star, while pulse shape can respond to the geometry of the emission region. Neither alone says whether gas in the disk travels with or against the binary orbit.

What 3,340 timing measurements can reveal

The new work combines 3,340 publicly available pulse-frequency measurements. Of those, 706 came from the Burst and Transient Source Experiment, or BATSE, aboard the Compton Gamma Ray Observatory. They cover about 8.2 years of deceleration. Another 2,634 measurements from the Fermi Gamma-ray Burst Monitor pulsar project cover roughly 14.5 years of acceleration.

The combined span is 22.7 years, although 4U 1626−67 has a much longer historical record. The researchers used an unscented Kalman filter, a state-space method designed to estimate changing quantities that cannot be observed directly. Pulse frequency supplied the measured signal. The model then reconstructed hidden histories of accretion torque, mass inflow and the interaction between the disk and magnetosphere.

This approach does more than fit one straight line before 2008 and another after it. It asks whether a physical torque prescription can follow the smaller fluctuations within both long episodes. That time-resolved requirement gives the data more leverage over competing descriptions of the disk.

Two routes to spin-down

The first tested configuration kept the disk prograde before and after the reversal. Prograde means its material orbits in the same sense as the binary and neutron star. Although incoming gas normally adds angular momentum, a forward-moving disk can still produce spin-down if the magnetosphere at the inner edge rotates faster than the local disk.

Magnetic stresses can then transfer angular momentum from the neutron star back to the flow, and sufficiently strong centrifugal action can inhibit accretion. A previous SpaceDaily account of the propeller effect in X-ray pulsars describes how the balance between disk and magnetosphere can sharply reduce the material reaching a star. Negative torque therefore does not, by itself, prove backward rotation.

The alternative configuration treated the disk as retrograde during the long spin-down interval and prograde during the subsequent spin-up. Retrograde material carries angular momentum opposite to the star’s rotation, so accreting it applies a braking torque directly. The 2008 change then marks a reversal in disk orientation rather than a large jump between two magnetic coupling regimes.

Why the flipped-disk history looks smoother

Under the retrograde-to-prograde interpretation, the inferred mass-accretion rate remained fairly smooth across the two observed eras. It varied by no more than 0.34 dex, equal to a factor of about 2.2. That scale is consistent with the substantial, but not enormous, X-ray brightening reported around the reversal.

The model’s fastness parameter also stayed in a similar range. Fastness compares the magnetosphere’s angular speed with the orbital motion at the inner disk. The inferred value was about 0.25 during deceleration under the retrograde assumption and about 0.30 during the later acceleration.

For an always-prograde disk, the same timing record required fastness near 3.0 during spin-down and about 0.30 during spin-up. In that account the system crosses from a rapidly rotating magnetosphere, capable of braking the star, into a much slower relative state. The authors found no compelling astrophysical mechanism for such an abrupt shift, but they did not rule one out.

The inferred angular acceleration in the flipped-disk model was broadly minus 9 to minus 5 trillionths of a radian per second squared during deceleration, and plus 2 to plus 9 trillionths after the reversal. The sign change is measured through the pulse history. Disk direction remains the model’s explanation for it.

The evidence is suggestive, not decisive

The retrograde-prograde model achieved a maximum-posterior log likelihood advantage of 2.5. At the single best-fitting point, that corresponds to a likelihood ratio of about 12. It sounds strong, but a best fit does not account for how much parameter space each explanation needs or how the alternatives perform away from their optimum.

After integrating across the allowed parameters, the natural-log Bayes factor was only 0.44 in favor of the flip. Exponentiating that figure gives odds of roughly 1.55 to one. That is a weak preference, far short of decisive evidence. The distinction between the best point and the fully marginalized result is essential to interpreting the study.

The paper therefore describes the result as the first indirect, time-resolved support for a retrograde disk during the deceleration era. It explicitly says the analysis does not prove the disk’s direction and calls for independent corroboration. The title’s phrase “may have flipped” reflects that boundary.

The transition itself falls in a data gap

The strongest limitation is easy to state: the model did not watch the disk reverse. BATSE stopped operating years before Fermi began its survey, so the public series used in the analysis have a gap. About 100 measurements around the transition, spanning roughly Modified Julian Dates 53,300 to 54,900, were not available in the post-processed public form the researchers required.

The study compares a long spin-down record with a long spin-up record and asks which disk history joins them most naturally. A state-space reconstruction can infer hidden variables from each side, but it cannot manufacture a direct observation of the missing event. The retrograde-to-prograde change could have occurred during that interval, yet the timing archive does not show gas slowing, stopping and reversing.

The system’s nearly face-on geometry also makes orbital direction hard to establish spectroscopically. In a more edge-on disk, the approaching and receding sides can imprint clearer Doppler shifts. For 4U 1626−67, signatures tied to circulation are likely to be subtler and more dependent on detailed models of where emission lines form.

Earlier X-ray spectra add context, not a verdict

Suzaku observations obtained in 2006 and 2010 bracketed the reversal from another angle. The later observation found an X-ray luminosity about 2.8 times higher and a much stronger complex of emission near one kiloelectronvolt. The cyclotron feature near 37 kiloelectronvolts, which traces the neutron star’s powerful magnetic field, did not change significantly.

That analysis favored a geometrically thin disk during both spin-down and spin-up. It showed that an accretion structure could persist across the torque change, but it could not determine the sign of the disk’s angular momentum. A thin retrograde disk and a thin prograde disk can both feed a magnetized neutron star.

SpaceDaily has previously described how matter spiralling through an accretion disk can power up a pulsar. 4U 1626−67 presents the inverse puzzle as well: how can an actively accreting pulsar brake for nearly two decades, then resume acceleration without a catastrophic change in its fuel supply?

What could decide which way the disk turned

A persuasive confirmation would require an observable that responds to orbital direction rather than only to the net torque. High-resolution, phase-resolved X-ray spectroscopy could search for consistent line asymmetries linked to the approaching and receding disk. Polarization and detailed pulse-profile modelling may also constrain the geometry of the inflow, although neither offers a simple directional label.

Continuous monitoring through a future torque reversal would be especially valuable. It could show whether brightness, pulse shape, quasi-periodic oscillations and inferred inner-disk radius change in the sequence expected for a reorientation. The historical record indicates that 4U 1626−67 can switch torque states more than once, so another test may be possible.

For now, the new analysis demonstrates that a backward disk followed by a forward disk can explain 22.7 years of timing with smoother hidden parameters than the traditional always-prograde picture. It turns a long-standing qualitative possibility into a quantitative, time-resolved hypothesis. The case remains open because the statistical advantage is modest and the most dramatic event in the proposed history lies exactly where the observations are thinnest.