H.E.S.S. Observations of the Binary System PSR B1259-63/LS 2883 around the 2010/2011 Periastron Passage

H.E.S.S. (High Energy Stereoscopic System) is an array of four 13 m diameter imaging atmospheric Cherenkov telescopes located in the Khomas Highland, Namibia at an altitude of 1800 m above sea level (Hinton 2004). The telescopes are optimized for the detection of VHE $\gamma$-rays in the range from  100 GeV to several tens of TeV by imaging Cherenkov light emitted by charged particles in Extensive Air Showers. The total field of view of H.E.S.S. is $5^{\circ}$. The angular resolution of the system is $\lesssim 0.1^{\circ}$ and the average energy resolution is about $15\%$. The H.E.S.S. array is capable of detecting point sources with a flux of $1\%$ of the Crab nebula flux at a significance level of $5\sigma$ in 25 hours when observing at low zenith angles (Aharonian et al. 2006).

PSR B1259-63/LS 2883 observations were scheduled to cover the post-periastron period from January to March 2011. The source was not visible for H.E.S.S. before and at the periastron passage. The observations resulted in a rather small dataset due to unfavorable weather conditions. The collected data correspond to 6 h of livetime after the standard quality selection procedure (Aharonian et al. 2006). These data were taken in five nights, namely January 9/10, 10/11, 13/14, 14/15 and 15/16. The observations were performed at a relatively high average zenith angle of 48^∘ and with a mean offset angle of 0.55^∘ from the test region centered at $\alpha_{\mathrm{J2000}}=13^{\mathrm{h}}02^{\mathrm{m}}48^{\mathrm{s}}$, $\delta_{\mathrm{J2000}}=-63\mbox{${}^{\circ}$}50\mbox{${}^{\prime}$}09\mbox{${}^{\prime\prime}$}$ (Wang et al. 2004).

The data were analysed using the model analysis¹¹1ParisAnalysis software version 0-8-18. technique with standard cuts (de Naurois & Rolland 2009). The test region was a priori defined as a circle with radius 0.1^∘ (i.e. $\theta^{2}<0.01^{\circ^{2}}$, where $\theta$ is defined as the angular distance between the $\gamma$-ray event and the nominal target position) which is the standard size for point-like sources. The Reflected Region Background technique was used for the background subtraction. The analysis results were cross-checked with an alternative analysis chain²²2H.E.S.S. Analysis Package (HAP) version 11-02-pl04. using a standard Hillas reconstruction (Aharonian et al. 2006) method for $\gamma$/hadron separation and an independent calibration of the raw data. Both analysis chains yielded consistent results.

The source was detected at a 11.5 $\sigma$ level (Li & Ma 1983) (see Table 1). A spectral analysis of the detected excess events shows that the differential energy spectrum of photons is consistent with a simple power law $\mathrm{d}N/\mathrm{d}E=N_{0}\left(E/1\mathrm{TeV}\right)^{-\Gamma}$ with a flux normalisation at 1 TeV of $N_{0}=(1.95\pm 0.32_{\mathrm{stat}}\pm 0.39_{\mathrm{syst}})\times 10^{-12}\,\textrm{TeV}^{-1}\textrm{cm}^{-2}\textrm{s}^{-1}$ and a spectral index $\Gamma=2.92\pm 0.30_{\mathrm{stat}}\pm 0.20_{\mathrm{syst}}$ (see Fig. 1 and Table 1) with a fit probability of 0.64. The integral flux above 1 TeV averaged over the whole observation period is $F(E>1\mathrm{TeV})=(1.01\pm 0.18_{\mathrm{stat}}\pm 0.20_{\mathrm{sys}})\times 10^{-12}$ cm^-2s^-1.

In order to check for variability of the source a lightcurve has been produced on a night-by-night basis assuming the photon spectral index obtained in the spectral fit (Fig. 2). The spectral index was fixed at the value obtained in the spectral analysis of the total data because of the low statistics for each individual night. The lightcurve is consistent with a constant resulting in a mean flux of $(0.77\pm 0.13)\times 10^{-12}$ cm^-2s^-1 (horizontal line in Fig. 2) with $\chi^{2}/NDF=6.35/4$ (probability of 0.17) yielding no evidence for variability in the 7-night observation period. For each individual night the source is detected at a statistical significance level $>3\sigma$ except the last point which significance is only $1.5\sigma$.

For a comparison with the GeV flare (see Section 4.2) the whole dataset was divided into two datasets: ”pre-flare” ($t_{\mathrm{p}}+26$ d to $t_{\mathrm{p}}+29$ d) and ”flare” ($t_{\mathrm{p}}+30$ d to $t_{\mathrm{p}}+32$ d). These two datasets were analysed independently revealing similar fluxes and significance levels (see Table 1). A spectral analysis of the two datasets shows that both spectra are consistent with a simple power law, yielding similar values of the spectral index (see Table 1). Both spectral indices are consistent with the one obtained for the total dataset. These results are discussed in Section 4.2.

For the data taken around the 2004 periastron passage the energy spectrum had been measured only up to $\sim 3$ TeV (Aharonian et al. 2005b) while for the much smaller dataset of 2011 observations the spectrum was measured up to $>4$ TeV and for the comparable dataset of 2007 observations up to $>10$ TeV (Aharonian et al. 2009) using more advanced analysis techniques with a better understanding of weak fluxes. In order to fully compare observation results around different periastron passages (see below), the 2004 data were re-analysed with the current analysis techniques, the same as used for the analysis of the 2011 data described above. The re-analysis results are compatible with the published ones. The differential energy spectrum measured up to $\sim 10$ TeV is consistent with a power law with a spectral index $\Gamma=2.8\pm 0.1_{\mathrm{stat}}\pm 0.2_{\mathrm{syst}}$ and a flux normalisation at 1 TeV of $N_{0}=(1.29\pm 0.08_{\mathrm{stat}}\pm 0.26_{\mathrm{syst}})\times 10^{-12}\,\textrm{TeV}^{-1}\textrm{cm}^{-2}\textrm{s}^{-1}$. The new analysis of the 2004 data is therefore compatible with the published results when extrapolated above 3 TeV.

In Fig. 3, the integrated photon flux above 1 TeV as a function of time with respect to periastron (indicated by the dashed vertical line) is shown. The lightcurve compiles the data from all three periastron observation campaigns spanning from 100 days before to 100 days after the periastron. The observed flux from the 2010/2011 observation campaign is compatible with the flux detected in 2004 at the similar orbital phases. Observation periods from 2004 and 2007 were separated in time with respect to the periastron position, i.e. observations in 2004 were performed mainly after and in 2007 mainly before the periastron. Therefore, it was impossible to directly confirm the repetitive behaviour of the source by comparing observations of PSR B1259-63/LS 2883 at the same orbital phases. In this perspective, although the 2011 observations do not exactly overlap with the orbital phases of previous studies, they cover the gap in the 2004 data post-periastron lightcurve and the integrated flux follows the shape of the lightcurve yielding a stronger evidence for the repetitive behaviour of the source.

The spectral shape of the VHE $\gamma$-ray emission from PSR B1259-63/LS 2883 around the 2010/2011 periastron passage is similar to what was observed during previous periastron passages (Fig. 4). The photon index of $2.92\pm 0.25_{\mathrm{stat}}\pm 0.2_{\mathrm{syst}}$ inferred from the 2011 data is well compatible with previous results. The spectrum measured for the 2011 data can be resolved only up to $\sim 4$ TeV, which is explained by a very low statistics at higher energies due to a short exposure of the source.

The absence of the flux enhancement during the GeV flare at radio and X-ray wavebands indicates that the GeV flare may be created by different physical processes from the ones responisble for the emission at other wavelengths. The VHE post-periastron data obtained with H.E.S.S. around the 2004 periastron passage do not show any evidence of a flux outburst at orbital phases at which the GeV flare is observed. However, the H.E.S.S. observations around the 2004 periastron passage do not comprise the orbital phase when the GeV flare starts. Also, in order to compare H.E.S.S. 2004 data with the GeV flare observed after the 2010 periastron passage, one has to assume that the GeV flare is a periodic phenomenon which may not be the case. The H.E.S.S. data taken between 9th and 16th of January in 2011 provide a three-day overlap in time with the GeV flare. Therefore, it is possible to directly study any flux enhancement in the VHE band on the timescale of the HE flare. In order to improve the sensitivity of the variability search the whole period of the H.E.S.S observations was divided into two time periods of almost equal length: before (”pre-flare”) and during (”flare”) the HE flare (see Section 3.4). The pre-flare and flare datasets analysis results are presented in Table 1.

In order to search for variability, the flux as a function of time was fitted with a constant resulting in a mean flux of $(0.91\pm 0.18)\times 10^{-12}$ cm^-2s^-1 (black horizontal dashed line in Fig. 5 (Top)). The fit has a $\chi^{2}$ to NDF ratio of $0.73/1$ which corresponds to a $\chi^{2}$ probability of 0.39, showing no indication for a flux enhancement. Note that the spectral parameters obtained by an independent fit of each of the two periods have been used here.

If one assumes that HE and VHE emission are created in the same scenario, i.e. the same acceleration and radiation processes and sites, then a flux enhancement of the same magnitude as observed at HEs should be also seen at VHEs. In order to investigate this hypothesis, the flare coefficient $\kappa$ is introduced as the ratio of the fluxes during the flare period and the pre-flare period. The ratio of the HE ($E>0.1$ GeV) flux averaged over the three days interval between $(t_{\mathrm{p}}+30\,\mathrm{d})$ and $(t_{\mathrm{p}}+32\,\mathrm{d})$ to the upper limit on the HE pre-flare emission (see Fig. 5 (Top)) yields a lower limit on the HE emission flare coefficient $\kappa_{\mathrm{HE}}\geq 9.2$. An upper limit on the VHE flare coefficient can be estimated using the profile likelihood method. The likelihood function is defined as a product of two Gaussian distributions of the pre-flare and flare flux measurements $\phi_{1}$ and $\phi_{2}$ correspondingly, stating that the flare measurement $\phi_{2}$ varies around $\tilde{\kappa}\tilde{\phi_{1}}$, where the tilde denotes the true value for a parameter. The profile likelihood $\lambda$ is then built as a function of $\kappa$:

where the hat denotes the maximum likelihood estimate for a parameter. The variable $-2\log{\lambda}$ follows a $\chi^{2}$ distribution with 1 degree of freedom, which allows to calculate the 99.7 % confidence level (equivalent of 3$\sigma$) upper limit of $\kappa_{99.7\%}<3.5$. The obtained upper limit is lower than the observed lower limit on $\kappa_{\mathrm{HE}}$.

The statistical tests presented above give two main results:

• •

  A flare of similar magnitude as observed in the HE band can be firmly rejected in the VHE band at the same orbital phase.

• •

  There is no significant difference between the pre-flare and the flare flux in the VHE band.

These two results suggest that the HE flare emission has a different nature than the VHE emission.

This conclusion is also further supported by the inconsistency of HE and VHE emission spectra (see Fig. 5 (Bottom)). The joint fit of the Fermi and H.E.S.S. data points with the simple power law (the dashed line on Fig. 5 (Bottom)) results in a fit probability of 0.004 and, hence, fails to explain the combined HE/VHE emission even ignoring Fermi upper limits which cannot be taken into account in the fit procedure. Moreover, the Fermi upper limits at $1-100$ GeV violate any reasonable model to explain the HE and VHE emission together. The Fermi spectrum alone is consistent with the the power law with exponential cut-off $E^{2}\mathrm{d}N/\mathrm{d}E=N_{0}\left(E/0.1\mathrm{GeV}\right)^{-p}e^{-E/E_{\mathrm{cutoff}}}$ with the index $p=0.16\pm 0.32$, the cut-off energy of $E_{\mathrm{cutoff}}=0.5\pm 0.2$ GeV and the normalisation $N_{0}=(4\pm 0.4)\times 10^{-10}$ erg cm^-2 s^-1. The fit probability is 0.27.

Several models have been proposed to explain the VHE emission from the source. In a hadronic scenario, the VHE $\gamma$-ray emission could be produced in the interaction of the ultrarelativistic pulsar wind particles with the dense equatorial disk outflow with subsequent production of $\pi^{0}$ pions and hence VHE $\gamma$-rays (Kawachi et al. 2004; Neronov & Chernyakova 2007). However, the detection of the source before the expected disk passage in 2007 put the hadronic scenario under doubt, suggesting that the VHE emission should be created at least partly by leptonic processes (Aharonian et al. 2009). Within the leptonic scenario, VHE emission from PSR B1259-63/LS 2883 is explained by the inverse Compton (IC) scattering of shock-accelerated electrons on stellar photons (Tavani & Arons 1997; Kirk et al. 1999; Dubus 2006; Khangulyan et al. 2007).

Few possible explanations for the nature of the HE flare are discussed in the literature. One of them suggested by Khangulyan et al. (2012) is based on IC scattering of the unshocked pulsar wind on the stellar and circumstellar disk photons. While the pulsar is inside the disk the IC scattering of the unshocked pulsar wind is supressed due to the high ram pressure. But right after the pulsar escapes the disk in the post-periastron phase the unshocked pulsar wind zone towards the observer increases significantly, while the density of the circumstellar disk photons is still high enough for the efficient IC scattering. Therefore, the enhancement of the HE flux is observed. This is not expected in the pre-periastron phase because the termination shock should expand towards the direction opposite to the observer. This model also predicts the difference between the HE and VHE emission, the latter expected to result from the upscattering of the stellar photons by the electrons accelerated at the termination shock between pulsar and stellar winds. Another explanation of the HE flare can be the Doppler boosting of the radiation created by the shocked pulsar wind (Bogovalov et al. 2008; Dubus et al. 2010; Kong et al. 2012). It is unclear though why the flare is not detected at other wavebands, since the Doppler boosting should enhance also X-ray and VHE $\gamma$-ray fluxes. This issue, however, can be resolved assuming a specific anisotropy of the pulsar wind and the difference of the emission behavior in different regions of the termination shock, the isotropic emission in the apex and the beamed emission in the tail of the shock (Kong et al. 2012). In this particular case the HE flare is explained by the Doppler boosting of the synchrotron emission. Abdo et al. (2011) suggested that the flare can also be explained by an anisotropy of the pulsar wind and/or stellar material. The anisotropy of electrons with the highest energies would cause an anisotropy of the synchrotron radiation at high energies. In this interpretation, the HE emission is produced by the synchrotron mechanism. The local increase of the stellar wind density would increase the Bremsstrahlung component which may also cause the HE flare. Regardless of which mechanism is responsible for the HE flare, the fact that it is observed only after the periastron indicates either a strong dependency of the HE emission on the geometry of the system, i.e. its configuration with respect to the direction to the observer, or some local perturbation of the stellar material. A detailed interpretation of the HE-VHE emission is beyond the scope of this paper and will be discussed in the joint MWL paper.