X-RAY OBSERVATIONS OF THE NEW UNUSUAL MAGNETAR SWIFT J1834.9−0846

Of the 20 Swift/XRT observations listed in Table 1, 4 were carried out in the Photon Counting (PC) mode and 16 in the Window Timing (WT) mode which provides much better temporal resolution (1.8 ms) at the expense of imaging. We used the HEASOFT⁸ analysis tools to reduce and analyze the data. We extracted spectra from the Level 2 event data using the standard grade selection of 0–12 and 0–2 for the PC and WT mode data, respectively. For the PC mode data, we used an r = 15'' circle as the source region and an annulus with the same center and inner and outer radii of 30'' and 45'' as the background region. For the WT mode data, we extracted the source spectra using a box centered on the CXO location with a length of 30'' aligned to the one-dimensional image. The background spectra were extracted with a similar size box centered far away from the source. We then generated the ancillary response files with xrtmkarf for each spectrum and regrouped the source spectra with a minimum of 15 counts bin⁻¹. The spectral fitting was done in XSPEC 12.6.0. Since the source was relatively bright at the onset of the outburst episode, the first XRT observation in PC mode (performed during two separate spacecraft orbits) was split into two parts to uncover early spectral variations. Three observations in WT mode and three observations in PC mode were too short to allow determination of spectral parameters. These observations were, therefore, excluded from our spectral analysis.

Swift J1834.9−0846 was observed with RXTE in eight pointings with a total exposure time of about 50 ks spanning over 30 days (see Table 1). The RXTE data were collected with the PCA (Jahoda et al. 1996) operating with two out of the five available proportional counter units in most of the observations. All data were collected in the GoodXenon mode, where each photon is time tagged with a minimum time resolution of about 1 μs. We used the PCA data primarily for timing analysis as it is not an imaging instrument, and the source intensity is relatively dim compared to the bright background X-ray emission (e.g., diffuse Galactic ridge emission and bright point sources in the 1° field of view of RXTE). However, we extracted the pulse peak spectrum using the longest RXTE pointing to investigate the source spectral behavior in a joint PCA and CXO analysis (see Section 6.3).

We observed Swift J1834.9−0846 on 2011 August 22 with the CXO ACIS operated in the Timed Exposure mode. The target was imaged near the aim point on the S3 chip using the 1/8 subarray (8' × 1' field of view). The data of an archival CXO observation (see Misanovic et al. 2011 for a description) were also analyzed, taking into account the different angular resolution and sensitivity. In our analysis, we worked with the pipeline-produced Level 2 event files (with standard filtering applied) and utilized CIAO 4.3 with CALDB 4.4.5. The spectral fitting was done in XSPEC 12.6.0.

Swift J1834.9−0846 was observed by RXTE on eight occasions with a total exposure time of ∼50 ks, spanning a time baseline of over 30 days (see Table 1). For our timing analysis we used data collected in the 2–10 keV range. For each observation, we first inspected the light curve with 0.03125 s time resolution and filtered out the times of short spikes and instrumental artifacts. We then converted the event arrival times to that of the Solar System Barycenter in Barycentric Dynamical Time using the JPL DE200 ephemeris and the Swift-derived coordinates of the source.

Next, we employed a Fourier-based pulse profile folding technique to determine the spin ephemeris of Swift J1834.9−0846. We first generated a template pulse profile by folding the longest PCA observation (Observation ID: 96434-01-02-00) at the pulse frequency determined with a Z²₁ search (Buccheri et al. 1983). Then, we generated pulse profiles for all PCA observations as well as for the CXO pointing, and cross-correlated them with the template profile to determine the phase shifts with respect to the template. We obtain the spin ephemeris of the source by fitting the phase shifts with a first or a higher order polynomial. We find that the phase drifts of Swift J1834.9−0846 are best described with a second-order polynomial (χ² = 7.3 for 7 degrees of freedom, dof) that yields a spin period P = 2.4823018(1) s and a period derivative $\dot{P}=7.96(12) \times 10^{-12}$ s s⁻¹ (epoch: 55783 MJD). In Figure 4, we present the drift of the pulse phase with respect to the template and the quadratic trend curve (upper panel), and the fit residuals in cycles (lower panel). The measured values of P and $\dot{P}$ correspond to the following spin-down parameters: age $\tau = P/2\dot{P} = 4.9$ kyr, power $\dot{E}=4\pi ^2I\dot{P}P^{-3}=2.1\times 10^{34}$ erg s⁻¹, and magnetic field $B=3.2\times 10^{19} (P\dot{P})^{1/2} = 1.4\times 10^{14}$ G.

Zoom In Zoom Out Reset image size

Finally, Figure 5 shows the pulse profiles obtained from several RXTE observations folded together using the derived ephemeris. We note the appearance of additional harmonics in the low-energy pulse profile of the source (2–5 keV) in RXTE data taken at later times.

Zoom In Zoom Out Reset image size

We searched for pulsations in the CXO/ACIS data obtained in the 2011 August 22 observation. We used the 733 counts extracted from the r = 1'' circle around the CXOU J183452.1−084556 position in the 2–10 keV band (there are only four counts below 2 keV, likely from the background). The time resolution in this observation was 0.44104 s (0.4 s frame time plus 0.04104 s charge transfer time). The photon arrival times were transformed to the Solar System Barycenter using the CIAO axbary tool. The ACIS observation started at epoch 55795.6489 MJD and continued for T_span = 13.02 ks.

We calculated the Z²₁ statistic as a function of trial frequency with a step of 0.35 μHz (which is about 0.05 T⁻¹_span) and found the maximum Z²₁ = 467 at $\nu = 0.4028512\ \rm {Hz} \pm 2.0\ \mu \rm {Hz}$,¹⁰ implying a very high significance of the pulsed signal. We also calculated Z²_n for n > 1 but did not find a strong contribution of higher harmonics.

Figure 6 (upper panel) shows the pulse profiles with 5 and 10 phase bins. We used these profiles to measure the pulsed fraction,¹¹ p = 85% ± 10%. We estimated the uncertainty of the pulsed fraction using Monte Carlo simulations and bootstrapping, also accounting for the time resolution and dead time in the 1/8 subarray mode. We also performed randomization of the arrival times within the 0.4 s frame time and re-calculated the pulsed fraction, which remained within the uncertainty range estimated above.

Zoom In Zoom Out Reset image size

The pulsed fraction can also be defined as $\tilde{p}=[2(Z_n^2 - 2n)/N]^{1/2}$, where n is the number of harmonics that give a significant contribution, and N is the number of counts.¹² In our case, $\tilde{p} = 1.13$ exceeds 100%, which might be due to dead-time effects and the relatively large (≈0.18) ratio of the time resolution to the period. To measure the pulsed fraction more accurately, the target should be observed with a better time resolution.

Figure 6 (lower panel) shows a 20-bin pulse profile averaged over the reference phase.¹³ We also produced a pulse profile for the surrounding extended emission but did not find a statistically significant pulsed signal.

We have fitted all XRT spectra (2–10 keV) jointly with two continuum models: a PL and a single blackbody (BB), both with interstellar absorption. In the first case, we find that the photon index remains the same within the uncertainties; therefore, we forced all observations in our joined fit to have the same varying photon index, while the normalizations were allowed to vary individually. We obtained a good fit (χ²_ν = 1.01 for 62 dof) with the best model parameters N_H = 10.5^+1.9 _{− 1.8} × 10²² cm⁻², and photon index Γ = 3.2 ± 0.4. The absorbed BB model resulted in temperatures that also remained consistent within their uncertainties; we then linked the temperatures and allowed the normalizations to vary. We again obtained a good fit (χ²_ν = 1.04 for 62 dof) with N_H = 4.4^+1.3 _{− 1.2} × 10²² cm⁻² and kT = 1.1 ± 0.1 keV. The temperature is higher than those measured in most other magnetars (typically around 0.5 keV; Woods & Thompson 2006).

6.2.1. CXOU J183452.1−084556

We collected a total of 733 counts (2–10 keV) from a circular region of r = 1'' centered at CXOU J183452.1−084556; the background contribution is expected to be only 0.25 counts (background was measured an 20'' < r < 33'' annulus). We then grouped the source spectrum requiring a minimum of 15 counts per spectral bin. The resulting spectrum is shown in Figure 7 (black error bars). The source pileup is negligibly small (≲ 1%), as the total source count rate of 0.057 counts s⁻¹ corresponds to 0.025 counts per frame. The point source spectrum can be fitted equally well with both the absorbed PL and BB models (see Table 2). The observed (absorbed) flux (2–10 keV) is F_point = (3.0 ± 0.1) × 10⁻¹² erg s⁻¹ cm⁻² (corrected for the finite extraction aperture and 9% dead time). Table 2 contains the values of the N_H and photon index for the best-fit absorbed PL model, and the N_H and temperature (kT) of the BB model. From the BB fit we estimate the emitting area radius to be 0.26 km (assuming that the source is at the same distance of 4 kpc as the SNR W 41; Tian et al. 2007). The corresponding unabsorbed PL and BB fluxes (2–10 keV) are 1.6^+0.6 _{− 0.4} and (5.8 ± 0.6) × 10⁻¹² erg s⁻¹, respectively (see also Table 2).

Zoom In Zoom Out Reset image size

Table 2. Spectral Fit Parameters of Swift J1834.9−0846 (Point Source) and its Extended Emission (Halo)

Model	NH, 22	$\mathcal {N}$a or Rb	Γc or kTc	χ2ν/dof d	LXe
CXO, PL (point)	20.3+2.7 − 2.5	4.8+5.9 − 2.4	3.5+0.5 − 0.4	0.72/31	3.1
CXO, BB (point)	12.0+1.8 − 1.7	0.26+0.14 − 0.07	1.1 ± 0.1	0.75/31	0.33
CXO, PL (halo)	10.0+2.2 − 2.0	0.6+1.1 − 0.3	3.7 ± 0.6	0.86/17	0.31
CXO, PL (point)	15 (fixed)	0.83+0.19 − 0.16	2.7 ± 0.2	0.86/32	1.7
CXO, PL (halo)	15 (fixed)	4.8+2.2 − 1.5	5.0 ± 0.3	1.05/18	0.67
CXO, PL (pre-outburst, halo)	4.0+3.1 − 1.9	5.5+4.0 − 2.5 × 10−4	1.0+0.8 − 0.5	0.72/7	∼0.01
CXO, PL (pre-outburst, halo)	15 (fixed)	4.7+3.2 − 1.0 × 10−2	3.5 ± 0.1	0.80/8	∼0.03
CXO (point) and RXTE, PL	20.7+2.2 − 2.0	4.8+5.2 − 2.0	3.6+0.4 − 0.3	0.77/41	2.8
CXO (point) and RXTE, BB	10.0+1.7 − 1.5	0.18+0.10 − 0.06	1.15+0.09 − 0.08	0.88/41	0.18

Notes. The uncertainties are given at 68% confidence level for a single interesting parameter. ^aSpectral flux in units of 10⁻² photons cm⁻² s⁻¹ keV⁻¹ at 1 keV. ^bBB radius in units of km². ^cPhoton index or BB temperature in keV. ^dReduced χ² and the number of degrees of freedom. ^eUnabsorbed PL luminosity in the 2–10 keV band or bolometric BB luminosity (πR²σT⁴) in units of 10³⁴ ergs s⁻¹.

Download table as:  ASCIITypeset image

Guided by the pulse profiles shown in Figure 6, we have extracted spectra from two different phase intervals: 0.15–0.35 (peak; indicated with the shaded region in Figure 6, upper panel) and the rest (off-peak). These spectra are shown in Figure 8; we again used the PL and BB models in each case, fixing the N_H at the best-fit value of the phase-integrated spectrum. The peak and off-peak spectral parameters (Γ or kT; see Table 3) are consistent within their uncertainties.

Zoom In Zoom Out Reset image size

6.2.2. Halo

We collected 314 counts (2–10 keV) from the 2'' < r < 10'' annulus (hereafter "halo"), centered at CXOU J183452.1−084556, where the contribution of the point source is expected to be small (<10%). We subtracted the background (estimated from a much larger region away from the source) and obtained a net total of ≃ 300 counts. To separate the halo from the point source, we simulated a point-spread function (PSF) using MARX.¹⁴ The comparison of the data with the PSF simulation (Figure 9) shows a good agreement within a small aperture (approximately up to 1'' radius), while the extended emission dominates at larger radii. Based on our simulation, we estimate that ∼33 photons come from the point source, after taking into account the extended PSF wings. The final halo spectrum was also binned, requiring a minimum of 15 counts bin⁻¹. The best-fit PL slope is approximately the same as that of the point source spectrum, while the best-fit N_H is a factor of two lower (see Table 2). The N_H–Γ confidence contours for the halo spectrum, together with those for the point source spectrum, are shown in Figure 10. The absorbed and unabsorbed fluxes (2–10 keV) of the halo emission are F_halo = (4.7 ± 0.2) × 10⁻¹³ and (1.6 ± 0.4) × 10⁻¹² erg s⁻¹ cm⁻², respectively.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 3. Power-law Fits to the Phase-resolved ACIS Spectra of Swift J1834.9−0846

Phases	NH, 22	$\mathcal {N}$ a	Γ b	χ2ν/dof c
Peak	20.3	1.5 ± 0.2	3.8 ± 0.3	0.94/40
Off-peak	20.3	0.30 ± 0.04	3.5 ± 0.2	0.94/40

Notes. N_H was held fixed during the fit. The uncertainties are given at 68% confidence level for a single interesting parameter. ^aSpectral flux in units of 10⁻² photons cm⁻² s⁻¹ keV⁻¹ at 1 keV. ^bPhoton index. ^cReduced χ² and the number of degrees of freedom.

Download table as:  ASCIITypeset image

The extended emission is well described by the same dust halo model as the one used by Misanovic et al. (2011) for another nearby source, CXOU J183434.9−084443, according to which most of the dust must be located relatively close to the source (within 1/4 of the distance). At least part of this dust could be associated with the molecular cloud that appears to be interacting with W41 (Leahy & Tian 2008), in agreement with the very large absorption column that we find in our spectral analysis (see Table 2).

6.2.3. Pre-outburst CXO/ACIS Data Taken on 2009 June

We analyzed the 2009 CXO observation covering the Swift J1834.9−0846 field and found zero photons within the error circle (r = 06) of CXOU 183452.1−084556 (see Figure 11). The off-axis angle of ≈46 during that observation is, however, large enough for the angular resolution to be substantially degraded compared to on-axis. Hence, to estimate the 2009 upper limit on the source flux, we used a larger radius, r = 2'', which would contain about 50% of the flux of a point source at this off-axis angle. We found five and four photons in the 0.5–10 and 2–10 keV bands, respectively. The mean local background surface brightness is 0.24 ± 0.02 and 0.18 ± 0.01 counts arcsec⁻² in 0.5–10 and 2–10 keV, respectively. Thus, within the r = 2'' extraction aperture we would expect to detect about two counts from the background. This translates into an upper limit of 0.15 counts ks⁻¹ in 2–10 keV, which corresponds to an absorbed flux limit of F_point < (2–4) × 10⁻¹⁵ erg s⁻¹ cm⁻².

Although we do not detect a point source, we notice extended emission on larger scales around the position of Swift J1834.9−0846. We demonstrate this by plotting the radial profile of the surface brightness (see Figure 11, top panel). The 10 annuli used to extract the radial profile are centered at the position of Swift J1834.9−0846, while the background is measured from 10 circular (r = 20'') regions surrounding the source. One can see from the figure inset that most of the excess over the background is within r ≲ 12'' and it corresponds to a detection significance of ≈5.1σ. However, there is also marginal (≈3σ) evidence for extended emission at larger scales (between r = 12'' and 30''; see Figure 11, top). There might be an even more extended (primarily toward southwest from the Swift J1834.9−0846), fainter asymmetric emission, but its significance can only be established with deeper CXO/ACIS observations.

We also found evidence for extended emission in the 2005 XMM-Newton data (see also Mukherjee et al. 2009). The extent, location, and significance of the extended emission in the EPIC/MOS images (which are not affected by the chip gaps and have a low enough background) are similar to those measured from the 2009 CXO/ACIS images (see above). The previously reported large-scale extended emission west–northwest of the SGR (i.e., in the direction toward CXOU 183434.9−084443) could be mainly due to the point sources that are smeared out in the EPIC images because of the coarse angular resolution of XMM-Newton. The two brightest point sources are clearly resolved in the sharper CXO/ACIS images (cf. corresponding panels in Figure 11). We did not attempt to extract the spectra from the 2005 XMM-Newton/EPIC data because the background is much higher and the angular resolution is worse than the one of the CXO observation. No point source is detected in the XMM-Newton images at the position of CXOU J183452.1−084556. We have not estimated an XMM-Newton upper limit on the point source flux, as it would be less restrictive than the one derived using the 2009 CXO/ACIS data.

Zoom In Zoom Out Reset image size

Since the RXTE/PCA is not an imaging instrument, we could not spatially separate the halo and the point source or even subtract a background measured independently from an offset region. However, since the instrument has a broader spectral range than the Swift/XRT and CXO/ACIS, potentially providing valuable source information above 10 keV, we used the latter data to calibrate our PCA spectrum of the longest (9.7 ks) pointing of 2011 August 9 (see Table 1). For the RXTE data, we accumulated the spectrum at the pulse minimum (which contains background, halo, and any unpulsed point source contributions) as the background and subtracted it from the source spectrum integrated over the remaining phases (see shaded regions in Figure 5). The resulting pulsed emission spectrum was then rebinned to have at least 50 counts per spectral bin after the background subtraction.

We performed a joined spectral fit of the RXTE (2–50 keV) and the CXO/ACIS (2–10 keV) data (see Figure 7). We found that the best fits are obtained when the RXTE flux is scaled down by a factor η = 0.6. The resulting PL best-fit parameters are very close to those of the CXO/ACIS fits but somewhat better constrained (see Table 2). A single BB fit is disfavored by systematic residuals at energies >8 keV (see Figure 7, bottom panel). The introduced scaling of the RXTE flux can be interpreted as due to two reasons: (1) the source was brighter at the time of the RXTE observation, and (2) the true background is lower than that estimated from the pulse minimum (see above).

Swift J1834.9−0846 has one of the shortest periods among magnetars¹⁵ and one of the highest pulsed fractions of the persistent X-ray emission, similar to that of 1E 1048.1−5937 on 2000 December 28 (Tiengo et al. 2002). In its timing and spectral properties, Swift J1834.9−0846 strongly resembles the recently discovered SGR J1833−0832, which has a period of 7.6 s and a magnetic field of 1.8 × 10¹⁴ G (Göğüş et al. 2010; Esposito et al. 2011). In particular, similar to SGR J1833−0832 (and unlike most other SGRs with good-quality spectra), the 0.5–10 keV spectrum of Swift J1834.9−0846 can be fitted with a single BB model, whose temperature, kT ≃ 1.1 keV, is the same as that of SGR J1833−0832. The Swift J1834.9−0846 BB radius, R = 0.26 km, is a factor of three smaller than in SGR J1833−0832, which, however, may not be a significant difference given the poorly known distances. Another similarity between Swift J1834.9−0846 and SGR J1833−0832 is the lack of obvious spectral shape evolution with rotational phase. The phase-resolved spectra (see Figure 8) differ only in normalization, and the differences in other model parameters are not statistically significant. Despite these similarities, the post-burst flux decay trend is markedly different for the two SGRs. The unabsorbed flux of Swift J1834.9−0846 decreased as ∝t^{−0.53 ± 0.07} from day 2 after the burst (Figure 3), while the flux of SGR J1833−0832 remained constant for nearly 20 days before the onset of decline. We note, however, that this early constancy of the flux in SGR J1833−0832 is unusual; the enhanced persistent X-ray flux of magnetars following an outburst usually declines as a PL with an index similar to that of Swift J1834.9−0846.

Similar BB temperatures and radii were also found for SGR 0418+5729 (P = 9.1 s, B < 7.5 × 10¹² G; Esposito et al. 2010) from the Swift XRT data taken within ∼10 days after the outburst. Also, SGR 0418+5729 exhibited a ∝t^−0.3 decay during the first 19 days and a much steeper, ∝t^−1.2, decay thereafter.

It is tempting to interpret the small emitting area of Swift J1834.9−0846 (similar to those of SGR J1833−0832 and SGR 0418+5729) as a hot spot on the NS surface. We should note that it would be very difficult to obtain such a high pulsed fraction even for a very small hot spot emitting (nearly isotropic) BB radiation because the pulsations would be washed out by the light bending in the NS gravitational field (see, e.g., Zavlin et al. 1995). If, however, we take into account that the angular distribution of radiation from a NS atmosphere has a narrow peak along the magnetic field direction (Pavlov et al. 1994), such a high pulsed fraction can indeed be explained assuming that the observed radiation emerges from a small hot spot near the magnetic pole of the NS. The fact that the light curve with such a high pulsed fraction shows only one peak per period suggests that the magnetic field configuration is substantially different from a centered dipole (e.g., it could be a strongly decentered dipole, in which case the magnetic fields and the temperatures are substantially different at the two poles). We caution here that the BB model provides only an empirical description of the spectral shape. It can be used for comparison of different sources, but it may be significantly different from the actual spectrum emitted from a NS atmosphere (Pavlov et al. 1995) and possibly modified by the resonance Compton scattering in the NS magnetosphere (Nobili et al. 2008). Comparing the BB fit parameters of the three SGRs (J1834.9−0846, J1833−0832, and J0418+5729), we can conclude that they depend neither on the SGR period (in the range of 2–10 s) nor on the strength of the spin-down magnetic field (in the range of (0.1–2) × 10¹⁴ G).

The nondetection of Swift J1834.9−0846 in the pre-outburst CXO data shows that the SGR flux can vary by at least a factor of ∼10³ between the presumably truly quiescent level in the low state and an elevated level that has persisted, with a slow decay, for at least 6 weeks after the outburst. This suggests that there is a large number of SGRs in a quiescent state undetectable at the current level of sensitivity of X-ray observatories.

At first glance, the extended emission around Swift J1834.9−0846, detected by CXO in 2011, looks rather unusual. Its radial distribution is consistent with that of a dust scattering halo (see Figure 9), but its spectrum shows some peculiarities. In particular, the best-fit hydrogen column density of the halo is a factor of two lower than that of the central source, while their spectral slopes are similar within statistics, instead of being steeper by ΔΓ = 1–2, as expected for the model halo spectrum (see, e.g., Misanovic et al. 2011). A likely explanation can be derived from the breadth of the N_H–Γ confidence contours and the strong correlation of these parameters. Indeed, Figure 10 shows that an intermediate N_H ≈ 1.5 × 10²³ cm⁻² corresponds to the 90% point source and halo confidence contours, and the best-fit photon indices at such N_H are Γ ≈ 3 and ≈5 for the point source and halo, respectively. Thus, we believe that the dust scattering halo is the most plausible interpretation of the extended emission around Swift J1834.9−0846.

We should also note that a fainter extended emission was seen around the magnetar position in the archival CXO data from 2009 in which no point source was detected. Although the best-fit N_H and Γ for this pre-outburst emission are substantially smaller than those in the post-outburst data (see Table 2), the large uncertainties of these parameters make them consistent with the corresponding parameters measured for the halo in 2011. The existence of a halo in the archival data may indicate that Swift J1834.9−0846 experienced an outburst not long before the 2009 June 7 observation.

Although it seems certain that most of the extended emission is the dust-scattered emission from the magnetar, we cannot exclude the possibility that it may contain some kind of a PWN due to synchrotron radiation from relativistic electrons/positrons accelerated in the NS magnetosphere and shocked in the ambient medium. We know that in the case of rotation-powered pulsars, a typical X-ray PWN luminosity is about $10^{-4} \dot{E}$, albeit with a large scatter (Kargaltsev & Pavlov 2008). If the same relationship is valid for magnetars, we would expect L_pwn ∼ 10³⁰ erg s⁻¹, which would be undetectable at the presumed distance of 4 kpc. It might happen, however, that a "magnetar wind nebula" is more efficient than one created by a rotation-powered pulsar, in which case we would expect a detectable contribution. To separate it from the dust scattering halo, one should analyze several data sets obtained at different times after the outburst. We expect that the halo component flux would be changing in proportion to the point source flux (with a time lag), while the PWN component would remain constant.

The distance to Swift J1834.9−0846 still remains an open issue. As this source is located within W41, association with this SNR is certainly plausible (other SGRs were found near SNR centers; Woods & Thompson 2006), but it has not been firmly proven. Similar extreme absorption (N_H = 3 × 10²³ cm⁻²; Misanovic et al. 2011) has been measured for the neighboring CXOU J183434.9−084443, indicating that such an absorption is not a unique feature of Swift J1834.9−0846 and hence not intrinsic to it. However, the distance to (and the origin of) CXOU J183434.9−084443 is also unknown. It could be a pulsar associated with W41 or a background active galactic nucleus located much farther. Leahy & Tian (2008) presented evidence for molecular clouds near W41, which are likely interacting with the SNR. The large absorbing column could be attributed to those clouds. At this point we can only conclude with certainty that Swift J1834.9−0846 is at the distance of 4 kpc or farther (the line of sight in that direction intersects several spiral arms). To better constrain the distance, the method of Durant & van Kerkwijk (2006) could be used; however, it requires grating observations, which are only feasible when the source is in the bright state. We note, however, that the conclusion by Durant & van Kerkwijk (2006), that all AXPs have more or less standard luminosity of 1.3 × 10³⁵ erg s⁻¹, cannot hold for quiescent SGRs because otherwise they would have been easily seen even at the most extreme distance of 20 kpc.

The field surrounding Swift J1834.9−0846 is rich in high-energy sources (see Figures 12 and 13). This magnetar is located at the heart of SNR W41 and nearly at the center of the extended TeV source HESS J1834−087, which itself is confined to the SNR interior (see Figure 13). In addition, there is a somewhat offset Fermi source, 2FGL J1834.3−0848, located nearby (see Figure 13). Since the extent of HESS J1834−087 is significantly smaller than that of the SNR, the TeV emission cannot be coming from the SNR shell as it does in some other cases (Bochow et al. 2011). The only other plausible explanation is that the TeV emission is powered by relativistic electrons injected by the compact object formed after the SNR explosion. There are currently several candidates for such an object. First, a few SGRs are known to be associated with shell-type SNRs (Hurley 2000), and the central location of Swift J1834.9−0846 certainly supports such a hypothesis. On the other hand, there is no firm evidence so far that SGRs can produce copious amounts of relativistic particles similar to young rotation-powered pulsars. While there is a convincing evidence that pulsars can power relic PWNe emitting TeV γ-rays, such evidence is currently lacking for magnetars.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Among other sources possibly related to HESS J1834−087 and W41 are CXOU J183434.9–084443 (a PWN candidate discussed in detail by Misanovic et al. 2011) and the 2XMM J183417.2−084901, which is located right at the center of the Fermi error circle (see Figure 12, top panels). Further longer observations of this region are required to understand the connection between the sources observed in different energy domains.