PSR B1931+24: Long-Term Intermittent Pulsar

• PSR B1931+24 is a long-term intermittent pulsar that alternates between active and quiet radio states on a ~38-day cycle with a ~50% difference in spin-down rates.
• Extensive 13-year monitoring using high-cadence observations and advanced methods like weighted wavelet Z-statistics and Monte Carlo analyses has detailed its emission and rotational dynamics.
• Sensitive FAST observations uncovered weak off-state emissions and dwarf pulses, bolstering magnetospheric switching models over alternative orbital interpretations.

PSR B1931+24 is the prototypical long-term intermittent pulsar, exhibiting strikingly regulated switching between radio-emitting and quiescent phases that is tightly coupled to its rotational dynamics. A 13-year timing and monitoring campaign established that it alternates between two discrete emission states and two discrete spin-down states on an average timescale of approximately $38 \pm 5$ days, while later FAST observations showed that the radio-off state is not fully silent but contains continuous weak emission and occasional dwarf pulses (Young et al., 2012, Rusul et al., 7 Jul 2025).

1. Observational basis and basic parameters

The long-baseline characterization of PSR B1931+24 is based on approximately 13 years of observations from 29 April 1998 to 19 May 2011, largely with the 76-m Lovell Telescope and the 28$\times$25-m Mark II at Jodrell Bank, with additional observations from the 94-m Nançay Radio Telescope to bridge maintenance gaps. Two backends were employed at Lovell: the Analogue Filter Bank to May 2010 and the Digital Filter Bank from January 2009. Typical center frequencies and bandwidths were 1402 MHz/32 MHz, 1520 MHz/384 MHz, 1396 MHz/32 MHz, and 1368 MHz/64 MHz, with intensified twice-daily monitoring starting in 2006 to sharpen transition constraints. The analysis used a one-bit “activity” time series for activity duty cycle estimation with bootstrap resampling ($10^{6}$ resamples), a weighted wavelet Z-statistic with tuning constant $c=0.001$, and timing residual fits that explicitly allow two spin-down rates, augmented by Monte Carlo propagation of transition-time uncertainties ($10^{5}$ trials per interval). A stride-fitting approach was used in both emission and rotational analyses to enhance temporal resolution (Young et al., 2012).

The pulsar period is $P \approx 0.814$ s, corresponding to a spin frequency $\nu = 1/P \approx 1.228965 \pm 0.000001$ Hz as measured in the timing fits. The long-term study adopted a distance $d \approx 4.6$ kpc to convert fluxes to pseudo-luminosities. Where needed, period and frequency derivatives are related via

$$\nu = \frac{1}{P}, \qquad \dot{\nu} = -\frac{\dot{P}}{P^{2}}.$$

A concise parameter summary is given below.

Quantity	Value
Spin period $P$	$\times$0 s
Spin frequency $\times$1	$\times$2 Hz
Mean radio-on duration	$\times$3 d
Mean radio-off duration	$\times$4 d
Mean cycle timescale	$\times$5 d
Radio-emitting duty cycle	$\times$6
$\times$7	$\times$8
$\times$9	$10^{6}$0
Torque ratio $10^{6}$1	$10^{6}$2

Later FAST observations used the catalog value $10^{6}$3, and the faint off-state pulses line up at the same DM as the on state, confirming astrophysical origin and rejecting radio-frequency interference as the cause (Rusul et al., 7 Jul 2025).

2. Intermittency cycle and state-duration phenomenology

The radio-on and radio-off durations span broad ranges. In high-cadence subsets, radio-on durations are typically 1–19 days, with average $10^{6}$4 days, while radio-off durations are typically 4–39 days, with average $10^{6}$5 days. There is no evidence for emission cessations shorter than a day in the long-term monitoring data. Across the full data set, PSR B1931+24 cycles between phases on an average timescale of about $10^{6}$6 days (Young et al., 2012).

The weighted wavelet Z-statistic shows short-term modulation between $10^{6}$7–50 days but a remarkably stable long-term average periodicity. The integrated WWZ power peaks near $10^{6}$8 ($10^{6}$9 days), with peak frequencies generally in $c=0.001$0–$c=0.001$1 ($c=0.001$2–42 days). Data-windowing WWZ segments repeatedly recover significant ($c=0.001$3) power at periods $c=0.001$4–47 days, and the overall variance around the 38-day mean is $c=0.001$5 days. On average, the neutron star is radio emitting for $c=0.001$6 of the time, consistent with $c=0.001$7 from stride-fitting with $c=0.001$8 d windows offset by 25 d (Young et al., 2012).

Statistical tests indicate no significant temporal evolution in the activity duty cycle, no correlation between the lengths of consecutive radio-on and radio-off intervals across nine high-cadence subsets, and no systematic intrinsic pulse-intensity variation during the radio-emitting phases. This combination of broad duration ranges, quasi-periodic cycling, and long-term stability is central to the classification of PSR B1931+24 as a long-term intermittent pulsar rather than a conventional short-nulling source (Young et al., 2012).

A distinct statistical interpretation was proposed in which the source is described by a two-state Markov model with periodically modulated transition rates. In that framework, the non-monotonic duration histograms and the quasi-period $c=0.001$9 days suggest stochastic resonance in an asymmetric bistable magnetospheric system, rather than the purely exponential residence-time distributions expected from a time-homogeneous two-state chain (Cordes, 2013).

3. State-dependent rotational dynamics

The defining dynamical property of PSR B1931+24 is the stability of its two spin-down rates over the full observing baseline: $10^{5}$0 These values were determined via least-squares fits to timing residuals that explicitly model dual spin-down rates across radio-on and radio-off segments. The residual model is

$10^{5}$1

with $10^{5}$2, $10^{5}$3, and $10^{5}$4 fitted per emission phase, and with transition-time uncertainties propagated by Monte Carlo sampling (Young et al., 2012).

A complementary stride-fit of $10^{5}$5 versus radio-on duration $10^{5}$6 across many three-burst windows yields a strong linear relation,

$10^{5}$7

with $10^{5}$8, reinforcing the stability of the on-state torque. From $10^{5}$9 s, the corresponding period derivatives are

$P \approx 0.814$0

The torque ratio between the two states is

$P \approx 0.814$1

Assuming a canonical moment of inertia $P \approx 0.814$2, the spin-down powers are

$P \approx 0.814$3

reflecting the $P \approx 0.814$4 torque enhancement in the radio-on state (Young et al., 2012).

Higher-order derivatives are not constrained: fits to $P \approx 0.814$5 do not yield significant values because of anti-correlated errors between $P \approx 0.814$6 and $P \approx 0.814$7 arising from finite transition-time uncertainties. Within a pulsar-wind formulation, the off state is treated as magnetic dipole braking with negligible wind contribution, while the on state adds a particle-wind torque; the measured spin-down ratio $P \approx 0.814$8 then yields model-dependent inclination angles and on-state braking indices in the range $P \approx 0.814$9–$\nu = 1/P \approx 1.228965 \pm 0.000001$0, depending on the accelerator prescription (Li et al., 2013).

4. Radio emission in the on and off states

In the 13-year study, 12-minute averaged profiles showed $\nu = 1/P \approx 1.228965 \pm 0.000001$1 and $\nu = 1/P \approx 1.228965 \pm 0.000001$2 in the $\nu = 1/P \approx 1.228965 \pm 0.000001$3 MHz Digital Filter Bank setup, implying that the radio-off phases were consistent with emission cessation. Scaled to 1400 MHz, the corresponding pseudo-luminosities are $\nu = 1/P \approx 1.228965 \pm 0.000001$4 and $\nu = 1/P \approx 1.228965 \pm 0.000001$5. Pulse-shape stability during on phases was supported by a first/second peak ratio of $\nu = 1/P \approx 1.228965 \pm 0.000001$6, with Anderson-Darling and reduced-$\nu = 1/P \approx 1.228965 \pm 0.000001$7 tests finding no significant deviations from normal noise behavior (Young et al., 2012).

FAST observations revised the observational picture of the off state. Contrary to two decades of null detections, PSR B1931+24 does not fully turn off. Even after removing all obvious “bursting dwarfs,” the residual off-state integrated profile shows a significant detection with peak S/N $\nu = 1/P \approx 1.228965 \pm 0.000001$8–10 and flux density $\nu = 1/P \approx 1.228965 \pm 0.000001$9–2 mJy. Sporadic, narrow dwarf pulses were detected in every off-state epoch, with single-pulse S/N from 5 to 30 and flux densities $d \approx 4.6$0–6 mJy; the occurrence rate is typically around $d \approx 4.6$1 per epoch, with 63 dwarfs out of 4055 pulses at MJD 58790 (Rusul et al., 7 Jul 2025).

The FAST campaign also found a substantial contraction in the integrated pulse width in the off state. The on-state integrated width is $d \approx 4.6$2 in the transition dataset and the catalog value is $d \approx 4.6$3, whereas the off-state value is $d \approx 4.6$4. The $d \approx 4.6$5 shrinkage is therefore $d \approx 4.6$6 using FAST-on versus FAST-off, or $d \approx 4.6$7 using catalog-on versus FAST-off. The mean flux density per individual pulse in the on state is $d \approx 4.6$8 mJy, while off-state dwarfs have $d \approx 4.6$9 mJy, giving $$\nu = \frac{1}{P}, \qquad \dot{\nu} = -\frac{\dot{P}}{P^{2}}.$$0 (Rusul et al., 7 Jul 2025).

Single-pulse statistics further show continuity rather than disjunction between the states. The single-pulse flux distributions for on-state and off-state dwarf pulses are both lognormal and overlap at their margins. Off-state dwarfs populate the full on-pulse longitude window but concentrate around the leading component A. The off-state integrated profile retains the same three-component structure, A, B, and C, at reduced intensity and contracted width. FAST also captured an off-to-on switch within one rotation: after 13 consecutive null pulses, the 14th pulse appeared abruptly with S/N $$\nu = \frac{1}{P}, \qquad \dot{\nu} = -\frac{\dot{P}}{P^{2}}.$$1, flux $$\nu = \frac{1}{P}, \qquad \dot{\nu} = -\frac{\dot{P}}{P^{2}}.$$2 mJy, and $$\nu = \frac{1}{P}, \qquad \dot{\nu} = -\frac{\dot{P}}{P^{2}}.$$3, after which the profile and intensity gradually recovered (Rusul et al., 7 Jul 2025).

5. Magnetospheric interpretations

The persistent, state-dependent $$\nu = \frac{1}{P}, \qquad \dot{\nu} = -\frac{\dot{P}}{P^{2}}.$$4 values and the stable $$\nu = \frac{1}{P}, \qquad \dot{\nu} = -\frac{\dot{P}}{P^{2}}.$$5-day switching strongly indicate two global magnetospheric states: a lower-conductivity, more vacuum-like configuration in the radio-off phase with smaller torque, and a higher-conductivity, plasma-loaded, pair-producing state in the radio-on phase with larger torque. In torque terms, the standard comparison is between an approximate vacuum dipole,

$$\nu = \frac{1}{P}, \qquad \dot{\nu} = -\frac{\dot{P}}{P^{2}}.$$6

and a force-free magnetosphere,

$$\nu = \frac{1}{P}, \qquad \dot{\nu} = -\frac{\dot{P}}{P^{2}}.$$7

with $$\nu = \frac{1}{P}, \qquad \dot{\nu} = -\frac{\dot{P}}{P^{2}}.$$8. The measured ratio $$\nu = \frac{1}{P}, \qquad \dot{\nu} = -\frac{\dot{P}}{P^{2}}.$$9 sits well below the force-free/vacuum extreme, suggesting that the on-state is more conductive than the off-state but not fully force-free (Young et al., 2012).

The long-term study argued that the year-to-year stability of both the switching timescale and the two spin-down values implies a high degree of magnetospheric memory, in spite of comparatively rapid $P$0ms dynamical plasma timescales. This suggests hysteresis-like behavior in which the magnetosphere remains in a quasi-stable configuration until a slow trigger initiates a global reconfiguration. In the same framework, the additional on-state torque corresponds to an estimated plasma charge density $P$1, comparable to the Goldreich–Julian co-rotational value $P$2, indicating that the on-state magnetosphere achieves a substantial fraction of the nominal co-rotation density (Young et al., 2012).

The FAST results refine, rather than overturn, this magnetospheric picture. The off state is not a vacuum state; both on and off states are better described as largely force-free, with localized acceleration and pair-cascade regions turning up or down. The narrower off-state beam, lower pair supply, and dyssynchronous component behavior support a picture of spatially inhomogeneous pair cascades in a resistive magnetosphere, with magnetic reconnection near the Y-point intermittently returning charges via the separatrix to the cap, thereby producing dwarf pulses and sustaining the faint continuous off-state emission (Rusul et al., 7 Jul 2025).

A related torque prescription is the pulsar-wind model, in which the on-state spin-down is written as the sum of dipole and particle-wind terms. In that approach the timing-derived wind duty cycle for PSR B1931+24 is $P$3, matching the independently measured radio on-state duty cycle of $P$4. The model also predicts $P$5 and model-dependent $P$6 values below 3, while ruling out the inverse Compton scattering induced space charge limited flow with field saturation for intermittent pulsars because it implies unrealistic parameters (Li et al., 2013).

6. Alternative models, comparative context, and open questions

PSR B1931+24 inaugurated the class of long-term intermittent pulsars, with similar behavior reported for PSR J1832+0029 and PSR J1841−0500. Compared to conventional short-nulling pulsars, whose nulls are typically $P$7 rotations, its week-to-month emission cycles and stable torque bistability point to a more global magnetospheric phenomenon (Young et al., 2012).

Several alternative explanations have been proposed. One is the Markov-plus-stochastic-resonance interpretation, in which the pulsar occupies two metastable states with periodically or quasi-periodically modulated transition rates. In that picture, the binary radio-emission and torque states, wide and non-monotonic duration distributions, and the $P$8 day quasi-period are quantitatively consistent with stochastic resonance in an asymmetric bistable magnetospheric system, possibly driven by feedback from the outer magnetosphere or return currents from an equatorial disk (Cordes, 2013).

A distinct class of proposals invokes orbital dynamics. One paper examined whether a companion orbiting with a period of 35 or 70 days could account for the intermittent behavior and concluded that none of the large-orbit companion configurations explains the whole set of peculiar properties, especially the measured spin-down change, the lack of a measurable period change between on and off, and the absence of timing residuals expected from a massive companion (Mottez et al., 2013). A companion paper instead proposed that PSR B1931+24 may be surrounded by a stream of small bodies of kilometric or sub-kilometric sizes orbiting at close distance to the star, with the recurrence period of 70 days interpreted as the period of precession of the periastron; in that scenario, Alfvén wings from fragments near the light cylinder perturb the magnetospheric current system (Mottez et al., 2013).

Another orbital interpretation treats the pulsar as a hidden ultra-compact binary. In that model, geodetic precession of the pulsar’s spin axis alters the azimuth and latitude at which the line of sight crosses the emission beam, so that timing-noise quasi-periodicity and intermittency arise from the same precessional geometry. For PSR B1931+24, the fitted residuals show rapid oscillations with period $P$9 days and slower modulation $\times$00 days, and the best-fit ultra-compact orbit is $\times$01 hr $\times$02 min, although the authors explicitly note parameter degeneracies and limited data (Gong et al., 2013).

The current observational situation favors magnetospheric-state switching as the governing mechanism, while leaving the trigger mechanism illusive. Higher-cadence, broadband, and multi-messenger observations remain decisive. LOFAR, MeerKAT, and the SKA were identified as facilities that can track transitions at hourly cadence, probe spectra and polarization through transitions, collect single-pulse statistics, and coordinate high-energy coverage. FAST has already shown that sensitive observations can recover off-state emission that was previously inaccessible, suggesting that other intermittent pulsars and even nulling pulsars with detected off-state dwarfs may reveal similarly faint continuous emission when observed with sufficient sensitivity and integration (Young et al., 2012, Rusul et al., 7 Jul 2025).