DISCOVERY OF HIGH-ENERGY GAMMA-RAY EMISSION FROM THE BINARY SYSTEM PSR B1259−63/LS 2883 AROUND PERIASTRON WITH FERMI

The pulsar system PSR B1259−63 was discovered at Parkes in 1989 and comprises a 47.76 ms radio pulsar orbiting a massive star (LS 2883) in a highly elliptical (e ≈ 0.87) orbit with a period of ≈3.4 years (Johnston et al. 1992; Negueruela et al. 2011). Recent optical spectroscopy (Negueruela et al. 2011) yields an updated distance estimate to this source of 2.3 ± 0.4 kpc, in reasonable agreement with the dispersion measure (DM) derived distance of 2.7 kpc using the NE2001 model (Cordes & Lazio 2002), so we adopt D = 2.3 kpc. The companion shows evidence for an equatorial disk in its optical spectrum and has generally been classified as a Be star (Johnston et al. 1994). The pulsar comes within ∼0.67 AU of its companion star at periastron, which is roughly the size of the equatorial disk (Johnston et al. 1992). The orbital plane of the pulsar is believed to be highly inclined with respect to this disk and so the pulsar crosses the disk plane twice each orbit, just before and just after periastron (Melatos et al. 1995). Shock interaction between the relativistic pulsar wind and the wind and photon field of the Be star is believed to give rise to the variable unpulsed X-ray emission observed throughout the orbit (Cominsky et al. 1994; Chernyakova et al. 2009) and the unpulsed radio and TeV γ-rays observed within a few months of periastron (Chernyakova et al. 2006).

At energies around 1 GeV, the Energetic Gamma-Ray Experiment Telescope (EGRET) provided only an upper limit for the 1994 periastron passage (F_γ ⩽ 9.4 × 10⁻⁸ cm⁻² s⁻¹ for E ⩾ 300 MeV, 95% confidence; Tavani et al. 1996). In TeV γ-rays the system was detected during the 2004 and 2007 periastron passages and flux variations on daily timescales were seen for energies >0.38 TeV in 2004 (Aharonian et al. 2005, 2009).

For the 2010/2011 passage, the time of periastron t_p was on 2010 December 15. By comparison to previous passages, the unpulsed radio and X-ray emission was expected to start rising in mid 2010 November, peaking around t_p − 10 days in the pre-periastron phase and reaching another peak around t_p + 15 days in the post-periastron phase. By 2011 April these emissions are expected to go back to their levels when the pulsar is far from periastron.

Abdo et al. (2010a) reported the first discovery of GeV γ-ray emission from this system which was detected during the first disk passage. A flaring GeV γ-ray activity during the second disk passage was reported in Kong et al. (2011) and in Abdo et al. (2011). Recently, Tam et al. (2011) reported with further details the GeV γ-ray activity from this system. We have assembled a multiwavelength campaign to monitor the system in radio, optical, X-rays, GeV, and TeV γ-rays during the 2010/2011 periastron passage. Here we describe the Fermi Large Area Telescope (LAT) detection of PSR B1259−63 in the E ⩾ 100 MeV range. We also present a preliminary analysis of a portion of the radio and X-ray data to determine if there was any anomalous multiwavelength behavior compared with previous periastron passages. We have analyzed LAT data over the entire time period from the beginning of the Fermi mission (2008 August 4; at which time the pulsar was nearing apastron) through periastron up until 2011 April 22 which is after the passage of the pulsar through the dense equatorial wind of the massive star. Full analyses and interpretation of the multiwavelength data are deferred to subsequent papers.

Analysis of the Fermi-LAT data was performed using the Fermi Science Tools 09-21-00 release. The high-quality "diffuse" event class was used together with the P6$\_$v3$\_$diffuse instrument response functions. To reject atmospheric γ-rays from Earth's limb, we selected events with zenith angle <100°. We performed standard binned maximum likelihood analysis using events in the range 0.1–100 GeV extracted from a 20° × 20° region centered on the location of PSR B1259−63. The model includes diffuse emission components as well as γ-ray sources within 20° of the source (based on an internal catalog created from 18 months of LAT survey data). The Galactic diffuse emission was modeled using the gll$\_$iem$\_$v02 model and the isotropic component using isotropic$\_$iem$\_$v02.⁶⁹ To better constrain the diffuse model components and the nearby sources, we first generated a model using two years of data between 2008 August 4 and 2010 August 4, a period during which the pulsar was far away from periastron. We fixed spectral parameters of all the sources between 5° and 15° from the source and left free the normalization factor of all the sources within 5° that were flagged as variable source in the 1FGL catalog (Abdo et al. 2010b). Normalizations for the diffuse components were left free as well. For this time period, the source was not detected with the LAT and we place a 95% upper limit on the photon flux above 100 MeV F₁₀₀ < 9 × 10⁻⁹ cm⁻² s⁻¹ assuming a power-law spectrum with a photon index Γ = 2.1.

The results of this fit were used to constrain the background source model for analyses on shorter timescales starting in 2010 November. In the source model, the normalization of the isotropic component was fixed to the two-year value, while the normalization for the Galactic diffuse component and three variable sources were left free.

We searched for γ-ray emission from this source on daily and weekly timescales during the first disk passage (2010 mid November to mid December). No detection at the level of 5σ was observed from the source on these timescales. Integrating from t_p − 28 days (the typical start of enhanced X-ray and unpulsed radio flux) to periastron yielded a clear detection of excess γ-ray flux from the source with a test statistic (TS) of ∼24 which corresponds to a detection significance of ∼5σ (Mattox et al. 1996). To estimate the duration of this enhanced emission and to get the best fit for the spectrum, we looked at the cumulative TS as a function of time for integrations starting at t_p − 28 days (Figure 1). Inspection of this plot reveals that the TS drops monotonically for integrations ending after t_p + 18 days, so we use this as the end of the integration for this initial period of detected γ-ray flux. For the rest of the paper, we will refer to this period as the "brightening." During the brightening, the detected γ-ray signal was in the energy range 0.1–1 GeV and no significant emission was detected above 1 GeV. The spectrum in the energy range 0.1–1 GeV is best described by a simple power law with photon index Γ = 2.4 ± 0.2_stat ± 0.5_sys with average photon and energy fluxes of F = (2.5 ± 0.8_stat ± 0.8_sys) × 10⁻⁷ cm⁻² s⁻¹ and F = (0.9 ± 0.3_stat ± 0.4_sys) × 10⁻¹⁰ erg cm⁻² s⁻¹, respectively. Because of the low signal-to-noise ratio during this period spectral fits to an exponentially cutoff power law were not constraining. This period is shown as the shaded region on the top panel of Figure 2, which shows the γ-ray flux in weekly time bins in the period t_p − 131 days to t_p + 128 days.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

At about t_p + 30 days, the source brightened rapidly, reaching a flux that was ∼10 times higher than the integrated flux measured during the first disk passage (Figure 2). This flare lasted for about 7 weeks. The spectrum during this period (t_p + 30 days to t_p + 79 days) is best described by a power law with exponential cutoff. The best-fit result is obtained for a photon index Γ = (1.4 ±0.6_stat ± 0.2_sys) and a cutoff energy E_c = (0.3 ±0.1_stat ± 0.1_sys) GeV. The average photon and energy fluxes above 100 MeV during this period are F₁₀₀ = (1.3 ± 0.1_stat ± 0.3_sys) × 10⁻⁶ cm⁻² s⁻¹ and G₁₀₀ = (4.4 ± 0.3_stat ± 0.7_sys) × 10⁻¹⁰ erg cm⁻² s⁻¹, respectively.

In addition to the significant difference in flux and spectral shape between the brightening and the flare periods, the weekly time bins show clear evidence for a change in the photon index for a power-law fit on these timescales (bottom panel of Figure 2). The phonon index softens from 2 to 2.5 during the brightening to a value of 3.5 around the peak of the flare. After that, the index hardens over the rest of the flare period to its values during the brightening.

The top panel of Figure 3 shows γ-ray flux as a function of time in 1 day time bins during the flare. The daily source flux during this period is variable at the 99.9% confidence level according to the method outlined in Abdo et al. (2010b). During this strong-variability flaring period, the source flux varied by a factor of 2–3 on daily timescales.

Zoom In Zoom Out Reset image size

The previous upper limits from EGRET are fully consistent with the flux observed by Fermi during the same orbital phase (Tavani et al. 1996). At the time of the bright emission seen by Fermi, which was well above the sensitivity level of EGRET, EGRET was pointed elsewhere.

To search for evidence of γ-ray pulsations from PSR B1259−63, we have constructed a radio ephemeris using observations from the 64 m Parkes telescope. Using Tempo2 (Hobbs et al. 2006), we fitted a timing model to 45 TOAs covering the range 2007 October 6 through 2011 January 16. Using this ephemeris, we folded LAT photons in the interval 2008 August 4 through 2010 August 4, a total of 24 months of observation, all well away from the flaring region near periastron where non-pulsed γ-ray emission is observed. We found no statistically significant indication of a pulsed γ-ray signal from this source. The pulsed flux upper limit depends on the unknown pulse shape and assumed spectrum. We therefore use the continuum upper limit on the energy flux above 100 MeV of G₁₀₀ < 1.0 × 10⁻¹¹ erg cm⁻² s⁻¹ in our comparisons with the rest of the γ-ray pulsar population.

Comparing this pulsar to the rest of the LAT-detected pulsars, we find that most detectability metrics predict that this should be a γ-ray pulsar. Although the characteristic age of 333 kyr is fairly large, the spin period is short for a middle-aged pulsar and thus at a distance of 2.3 kpc the $\dot{E}$ of 8.2 × 10³⁵ erg s⁻¹ and magnetic field at the light cylinder of 2.9 × 10⁴ G are well within the range where γ-ray pulsations are typically detected (Abdo et al. 2010c). If we assume that the beaming factor $f_\Omega$ is 1, then the γ-ray efficiency of the pulsar is less than 0.7%, which is considerably lower than the 10% efficiencies that are typical in this $\dot{E}$ range. Determining whether this pulsar is intrinsically underluminous in the γ-ray band or if the low γ-ray luminosity is simply a geometric effect will require detailed modeling that includes geometrical information from radio polarization measurements.

We have also searched for γ-ray pulsations during the brightening and flare periods where continuum γ-ray emission was detected. No pulsations were detected in these intervals. This is consistent with this γ-ray emission originating from the intrabinary shock, which is well outside the light cylinder of the pulsar and thus is not expected to be modulated at the spin period.

Pulsed emission was monitored at Parkes to look for changes in the DM and rotation measure and determine the duration of eclipse of the pulsed signal. Pulses disappeared on t_p − 16 days and reappeared on t_p + 15. In the ∼2 weeks leading up to the disappearance of the pulse, significant changes in the DM were observed.

The PSR B1259−63 system was monitored at frequencies between 1.1 and 10 GHz using the ATCA array. Twelve observations spanning t_p − 31 days to t_p + 55 days were collected. Unpulsed transient radio emission was detected throughout the periastron passage with a behavior similar to that seen in previous observations (Johnston et al. 2005), as shown in Figure 4.

Zoom In Zoom Out Reset image size

X-ray observations of PSR B1259−63 during this passage demonstrated the repeatability of the 1–10 keV light curve as shown in Figure 4. As with the periastron passage of 2004, Swift observed a rapid X-ray brightening starting at ∼t_p − 25 days. These observations confirmed the spectral hardening preceding the pre-periastron flux rise. Similarly to previous periastron passages (see, for example, Chernyakova et al. 2009, and references therein), observations with Swift, Suzaku, and XMM-Newton showed a rise of X-ray flux after periastron.

Emission from the PSR B1259−63 system is produced in the interaction of the pulsar wind with the stellar wind of the companion star. Observations in radio, X-ray, and TeV γ-ray bands (Johnston et al. 1992; Aharonian et al. 2005, 2009; Kawachi et al. 2004; Tavani et al. 1996; Chernyakova et al. 2006, 2009) revealed a characteristic variability of this emission during the periods of periastron passage. Detection of the 0.1–10 GeV band γ-ray emission around periastron was not unexpected. However, Fermi observations reveal puzzling behavior of the source, which was not predicted in any model of γ-ray emission from this system. An unexpected strong flare visible only in the GeV band was observed some 30 days after the periastron passage and after the neutron star passage of the dense equatorial wind of the massive star.

During this flare the source was characterized by an extremely high efficiency of conversion of pulsar spin-down power into γ-rays. The highest day-average flux was F₁₀₀ ∼ 3.5 × 10⁻⁶ cm⁻² s⁻¹ with a spectral index of Γ ∼ 3.0. This corresponds to an isotropic γ-ray luminosity of ≃ 8 × 10³⁵(D/2.3 kpc)² erg s⁻¹, nearly equaling the estimated total pulsar spin-down luminosity L_SD ≃ 8.3 × 10³⁵ erg s⁻¹ (Johnston et al. 1992). This is illustrated in Figure 5 where the horizontal red line shows the flux which would be produced when 100% of the spin-down power is converted into radiation emitted within one decade of energy, not taking into account possible beaming effects.

Zoom In Zoom Out Reset image size

Broadband spectra of emission around periastron are shown in Figure 5. Strong increases in GeV flux and changes in γ-ray spectrum during the flare were not accompanied by noticeable spectral variations in the X-ray band.

Several possible mechanisms of production of 0.1–10 GeV γ-ray emission from the system were previously discussed: synchrotron, inverse Compton (IC), bremsstrahlung, or pion decay emission (Tavani & Arons 1997; Kawachi et al. 2004; Chernyakova et al. 2006; Khangulyan et al. 2007). Electrons with energies E_e ∼ 100 TeV produce synchrotron emission in the energy range E_γ ∼ 10⁹[B/1 G][E_e/10¹⁴ eV]² eV. Alternatively, electrons with energies E_e ∼ 1–10 GeV could produce γ quanta with energies E_γ ≃ 10⁸[E_e/1 GeV]² eV via IC scattering of Be star photons. Bremsstrahlung emission in the GeV band could be produced by the GeV electrons. Finally, the dense equatorial stellar wind could provide a sufficiently dense target for proton–proton interactions followed by decays of neutral pions into photons.

Figure 5 shows example model fits to the persistent emission data. The model shown in the upper panel assumes that high-energy particles escape with the speed of the stellar wind, as in the model of Chernyakova & Illarionov (1999) and Chernyakova et al. (2006). Slow escape of the high-energy particles leaves enough time for the efficient cooling of electrons via IC and/or bremsstrahlung/Coulomb loss mechanisms. In the lower panel, high-energy particles are assumed to escape with the speed 10¹⁰ cm s⁻¹, as in the model of Tavani & Arons (1997). In this case only synchrotron cooling is efficient. The code used for the calculations is described in Zdziarski et al. (2010).

In general, the flare could be explained either by anisotropy of the γ-ray emission or by an abrupt change of physical conditions in the emission region or by the appearance of a new emission component.

Several possible sources of anisotropy are present in the system: relativistic beaming of the γ-ray emitting outflow from the system (Bogovalov et al. 2008; Dubus et al. 2010), anisotropy of pulsar wind, or anisotropy of radiation field of the massive star. The model for the flare spectrum (cyan data points) shown in the lower panel of Figure 5 assumes a particular type of anisotropy which could appear at the high-energy end of the synchrotron spectrum, at the energies at which the synchrotron cooling distance is comparable to the gyroradius. In such a situation the electron distribution could not be isotropized within the synchrotron cooling timescale. The assumption that highest energy electrons with anisotropic initial velocity distribution cool before being isotropized results in the increase of apparent luminosity by a factor 4π/Ω ∼ 10, where Ω is the solid angle into which most of the highest energy synchrotron power is emitted.

Another possible explanation of the flare is considered in the model shown in the upper panel of Figure 5. It assumes a local increase of the density of stellar wind by a factor of ∼10 which results in the increase of the bremsstrahlung component of emission spectrum.

Clarification of the physical mechanism of the puzzling flare discovered by Fermi requires a more complete view of the properties of the flare, including information on system behavior in optical and TeV bands and on orbit-to-orbit variations of the GeV flaring pattern.

We thank Mallory Roberts for helpful contributions.

The Fermi-LAT Collaboration acknowledges support from a number of agencies and institutes for both the development and the operation of the LAT as well as scientific data analysis. These include NASA and DOE in the United States, CEA/Irfu and IN2P3/CNRS in France, ASI and INFN in Italy, MEXT, KEK, and JAXA in Japan, and the K. A. Wallenberg Foundation, the Swedish Research Council, and the National Space Board in Sweden. Additional support from INAF in Italy and CNES in France for science analysis during the operations phase is also gratefully acknowledged. The Parkes radio telescope is part of the Australia Telescope which is funded by the Commonwealth Government for operation as a National Facility managed by CSIRO. We thank our colleagues for their assistance with the radio timing observations. This work was supported in part by a NASA Fermi Guest Investigator Program.