PSR J0537-6910: Exponential recoveries detected for 12 glitches

TL;DR

PSR J0537−6910 is the most frequently glitching pulsar, and this study tests whether exponential post-glitch recoveries are common or hidden by cadence. The authors fuse RXTE and NICER timing data with a glitch timing model that allows a single exponential recovery, evaluating many glitches with a Bayesian model comparison to decide if an exponential term is warranted. They identify 12 glitches with exponential recoveries (11 with strong/moderate/decisive evidence) and add six NICER glitches, bringing the total to 66; exponential recoveries occur only for the largest glitches, with timescales $4\le\tau_d\le37$ days and decaying frequency changes $\Delta\nu_d$ in $(0.05,0.8)\ \mu$Hz. Incorporating these recoveries significantly affects persistent spin-down parameters and braking indices, revealing a nuanced interaction between superfluid re-coupling and external torques and underscoring the need for high-cadence monitoring of glitching pulsars.

Abstract

Pulsar glitches are unresolved increments of the rotation rate that sometimes trigger an enhancement of the spin-down rate. On occasions, the augmented spin-down decays gradually in an exponential manner, particularly after the largest glitch events. The young pulsar PSR J0537-6910 exhibits the highest known glitching rate, with 60 events detected in nearly 18 years of monitoring. Despite most PSR J0537-6910 glitches being large, only one exponential recovery has been reported, following its first discovered glitch. This is puzzling, as pulsars of similar characteristics typically present significant exponential recoveries. We aim to determine whether this reflects an intrinsic difference in PSR J0537-6910 or a detectability issue, for example due to its high glitch frequency. The full dataset, including recent NICER observations, was systematically searched for exponential relaxations. Each glitch was tested for evidence of a recovery over a broad range of trial timescales. Promising candidates were investigated further by comparing recovery models with and without an exponential term using Bayesian evidence. We discovered six new glitches, bringing the total to 66. Our criteria strongly indicates the presence of 11 previously undetected exponential recoveries. We presente updated glitch and timing solutions. Exponential recoveries are detected only for the largest glitches, though not all of them. The inferred timescales range from 4 to 37 d, with the decaying frequency increment generally below $1\%$ of the total. We find that $\ddotν$ can remain stable across several glitches, with persistent changes associated with only some events. In particular, it tends to be lowest after glitches with exponential recoveries, yielding inter-glitch braking indices between 6 and 9. Following glitches without recoveries, $\ddotν$ is higher, leading to braking indices between 10 and 35.

Figures (13)

• Figure 1: Example of the exploratory process to determine possible values for an exponential relaxation timescale. The blue dots corresponds to the $\chi^2_\mathrm{red}$ values of glitch #64, for a subset of the total range of trial $\tau_\mathrm{d}$ values, focused around the timescale that returns the minimum $\chi^2_\mathrm{red}$ (marked by the red line). Orange dashed lines correspond to $1\sigma$ uncertainty. These, rather large $\chi^2_\mathrm{red}$ values are only used to identify a possible timescale. Further analyses are performed for the final measurements (Table \ref{['tab:glitch_solutions']}).
• Figure 2: Examples of exponential glitch recovery detections, for glitches 1 (left) and 56 (right). The panels show, from top to bottom: Phase residuals relative to Model 1, with the residuals of the most likely exponential recovery model (Model 2 relative to Model 1) superimposed (blue curve); Phase residuals relative to Model 2; Frequency residuals relative to Eq. \ref{['eq:timing-model']} fitted to TOAs up to the glitch epoch, with the post-glitch data all lowered by a certain amount (the mean post-glitch frequency residual) for better visualisation; $\dot\nu$ evolution with Model 2 shown as the blue curve.
• Figure 3: Difference in inferred glitch parameters between Model 1 (blue points) and Model 2 (red points), for glitches with detected exponential recovery. Gray points represent values for glitches without detectable exponential terms (Model 1).
• Figure 4: Distribution of $\tau_\mathrm{d}$ values for the 12 glitches with a detected recovery as represented by a histogram of 10000 samples (weighted) from the posterior distributions. The red ticks correspond to the values yield by the final solutions, as reported in Table \ref{['tab:glitch_solutions']}.
• Figure 5: Time to the next glitch as a function of $\Delta \nu_p$, for 66 glitches of PSR J0537$-$6910. Red markers show glitches for which an exponential recovery was detected. Using the Model 2 solutions in these cases improves the size-waiting time correlation. Glitch 45, with a detected recovery, is the last glitch in the RXTE dataset, which ends $107$ d after. Thus the time to the next glitch is unknown and it is plotted with a vertical arrow indicating the lower limit. The latest glitch that we report in this work is glitch 66, which we are able to include in this plot as we know glitch 67 occurred on MJD 60592.