Long-term spectral and timing properties of the soft gamma-ray repeater SGR 1833-0832 and detection of extended X-ray emission around the radio pulsar PSR B1830-08

SGR 1833-0832 was discovered on 2010 March 19 thanks to the Swift detection of a short hard X-ray burst and follow-up X-ray observations. Since then, it was repeatedly observed with Swift, Rossi X-ray Timing Explorer, and XMM-Newton. Using these data, which span about 225 days, we studied the long-term spectral and timing characteristics of SGR 1833-0832. We found evidence for diffuse emission surrounding SGR 1833-0832, which is most likely a halo produced by the scattering of the point source X-ray radiation by dust along the line of sight, and we show that the source X-ray spectrum is well described by an absorbed blackbody, with temperature kT=1.2 keV and absorbing column nH=(10.4+/-0.2)E22 cm^-2, while different or more complex models are disfavoured. The source persistent X-ray emission remained fairly constant at about 3.7E-12 erg/cm^2/s for the first 20 days after the onset of the bursting episode, then it faded by a factor 40 in the subsequent 140 days, following a power-law trend with index alpha=-0.5. We obtained a phase-coherent timing solution with the longest baseline (225 days) to date for this source which, besides period P=7.5654084(4) s and period derivative dP/dt=3.5(3)E-12 s/s, includes higher order period derivatives. We also report on our search of the counterpart to the SGR at radio frequencies using the Australia Telescope Compact Array and the Parkes radio telescope. No evidence for radio emission was found, down to flux densities of 0.9 mJy (at 1.5 GHz) and 0.09 mJy (at 1.4 GHz) for the continuum and pulsed emissions, respectively, consistently with other observations at different epochs.


INTRODUCTION
Soft gamma-ray repeaters (SGRs; seven confirmed members) and anomalous X-ray pulsars (AXPs; twelve confirmed members) 1 are two classes of X-ray pulsating sources with no evidence for companion stars which share a number of properties. These include rotation periods of several seconds (P ∼ 2-12 s), rapid spin down (Ṗ ∼ 10 −11 s s −1 ), large and variable X-ray luminosities (exceeding the rate of rotational energy loss), and the emission of flares and short bursts (see Woods & Thompson 2006;Mereghetti 2008 for reviews). SGRs and AXPs (note that the distinction is becoming increasingly blurred) are currently interpreted as observational manifestations of magnetars, namely neutron stars powered by their huge magnetic field (e.g. Paczynski 1992;Duncan & Thompson 1992;Thompson & Duncan 1995Thompson, Lyutikov & Kulkarni 2002). This picture is supported by the fact that the dipole magnetic fields inferred for SGRs and AXPs from their period and period derivative 2 are above, or at the high end of, those of the radio pulsars. Surface magnetic fields in SGRs/AXPs in fact often exceed 10 14 G, although an upper limit as low as 7.5×10 12 G has been recently reported for SGR 0418+5729 .
Between periods of activity, characterised by bursts and significant variability in flux, spectrum, pulse shape, and spin-down rate, magnetars go through long stretches of quiescence. The discovery of the first 'transient ' AXP, XTE J1810-197 (Ibrahim et al. 2004, 3 showed that during quiescence magnetars can be faint and not dissimilar from hundreds of unidentified sources present in various X-ray catalogues (such as the ROSAT, XMM-Newton, and Chandra ones). This suggested that a potentially large number of magnetars had not been discovered yet and may manifest themselves in the future. Thanks also to the effectiveness of the Swift and Fermi satellites in catching magnetar outbursts, five new magnetars (all of them transients) were discovered in the last few years and several major outbursts were observed from known sources (see Rea & Esposito 2011 and references therein).
A recent addition to the magnetar family is SGR 1833-0832. It was discovered on 2010 March 19 when, at 18:34:50 UT, the Swift Burst Alert Telescope (BAT) triggered on a short (<1 s) hard Xray burst and localised it in a region close to the Galactic plane ; Barthelmy et al. 2010;Gögüş et al. 2010a). The Swift X-ray Telescope (XRT) started observing the BAT field one minute after the trigger and unveiled the existence of a previously unknown bright X-ray source. Given the proximity to the Galactic plane and the burst properties, the X-ray source was immediately suggested to be an SGR. The SGR nature of the source, now catalogued as SGR 1833-0832, has been confirmed shortly after by the discovery with Swift and Rossi X-ray Timing Explorer (RXTE) of pulsations at 7.57 s (Gögüş et al. 2010b;Esposito et al. 2010;Palmer & Gelbord 2010). Subsequent observations also allowed the determination of the spin-down rate of SGR 1833-0832 1 See the McGill Pulsar Group catalogue at the webpage http://www.physics.mcgill.ca/∼pulsar/magnetar/main.html. 2 According to the usual magnetic braking formula, for a neutron star of 10 km radius and 1.4 solar masses, B dip ≈ 3.2 × 10 19 (PṖ ) 1/2 G (e.g. Lorimer & Kramer 2004). 3 Actually, at that time another magnetar with a transient behaviour, SGR 1627-41 (Kouveliotou et al. 2003;Esposito et al. 2008), was already known, but its characteristics were not well established yet. Significant flux variability was reported also in the AXP candidate AX J1845.0-0300 (Torii et al. 1998;Vasisht et al. 2000;Tam et al. 2006).  (Esposito et al. 2010b;Gögüş et al. 2010a). Following the onset of the outburst, the source flux remained fairly constant for about 20 days (Gögüş et al. 2010a). No radio, optical or infrared counterparts have been detected (Burgay et al. 2010;Wieringa et al. 2010;Gögüş et al. 2010a).
Here we report on the spatial and long-term spectral and temporal behaviour of the persistent X-ray emission of SGR 1833-0832 using new XMM-Newton, RXTE and Swift observations. In Sections 2 and 3 we describe the X-ray observations used in our study and we present the results of our analysis. In Section 4 we give more details on the radio observations presented in Burgay et al. (2010) and Wieringa et al. (2010). In Section 5 we report on the analysis of the field of PSR B1830-08 (J1833-0827), which was serendipitously imaged by the XMM-Newton observations, and discuss its possible association with the unidentified TeV source HESS J1834-087. Discussion follows in Section 6.

XMM-Newton
The three focal plane CCD cameras of the XMM-Newton EPIC instrument, pn (Strüder et al. 2001), MOS1 and MOS2 (Turner et al. 2001), cover the 0.1-12 keV energy range with an effective area of roughly 1400 cm 2 for the pn and 600 cm 2 for each MOS. After the discovery of SGR 1833-0832, XMM-Newton pointed its mir-rors towards the new SGR three times (see Table 1). All observations were performed with the thick optical filter and in full frame mode, 4 except for the first one, carried out with the MOS cameras in large window mode. Moreover, XMM-Newton serendipitously imaged the field of SGR 1833-0832 on 2006 September 16 during an observation targeting the nearby supernova remnant G23.5-0.0 (obs. ID: 0400910101, exposure: 12.4 ks; all the detectors were in full frame mode with the medium filter).
The data were processed using version 10.0 of the XMM-Newton Science Analysis Software (SAS) and standard screening criteria were applied. Source events were accumulated for each camera from circular regions with a 36 arcsec radius. We selected this aperture, corresponding to ∼85% of the encircled energy fraction at 5 keV for a point source, in order to minimise the contamination from the diffuse emission surrounding SGR 1833-0832 (see Section 3.1). The background counts were extracted from sourcefree regions far from the position of the SGR. The ancillary response files and the spectral redistribution matrices for the spectral analysis were generated with the SAS tasks ARFGEN and RMFGEN, respectively.

Swift
The X-Ray Telescope (XRT; Burrows et al. 2005) on-board Swift uses a CCD detector sensitive to photons between 0.2 and 10 keV with an effective area of about 110 cm 2 . Twenty-seven observations of SGR 1833-0832 were performed (see Table 1), starting right after its discovery, in both photon counting (PC) and windowed timing (WT) modes (see Hill et al. 2004 for more details on the XRT readout modes). The results of the first 18 Swift observations have already been published in Gögüş et al. (2010a), while the others are reported here for the first time.
The data were processed and filtered with standard criteria using the FTOOLS software package. We extracted the PC source events from a circle with a radius of 15 pixels (one pixel corresponds to about 2. ′′ 36) and the WT data from a 25-pixel-wide strip. To estimate the background, we extracted PC and WT events from source-free regions distant from the position of SGR 1833-0832. For the spectral fitting we used the latest available spectral redistribution matrix in CALDB (v011), while the ancillary response files were generated with XRTMKARF, and they account for different extraction regions, vignetting and point-spread function corrections.

RXTE
The Proportional Counter Array (PCA; Jahoda et al. 1996) onboard RXTE consists of an array of five collimated xenon/methane multi-anode Proportional Counter Units (PCUs) operating in the 2-60 keV energy range, with a total effective area of approximately 6500 cm 2 and a full width at half-maximum field of view (FOV) of about 1 • . The 80 RXTE/PCA pointed observations of SGR 1833-0832 (obs. ID: 95048) reported here span from 2010 March 19 to 2010 December 04. The exposures range from 1 ks to 20 ks, for a total of about 466.4 ks. The results of the first 31 RXTE observations have already been published in Gögüş et al. (2010a), while the analysis of the others is reported here for the first time.
The raw data were reduced using the FTOOLS package. Given the non-imaging nature and wide FOV of the PCA instrument and Figure 1. Surface brightness radial profiles of the background-subtracted X-ray emission of SGR 1833-0832 from the XMM-Newton MOS1 and MOS2 cameras (3-6 keV data from the three observations combined). The red solid line shows for each camera the best-fit point-spread function (Ghizzardi 2001). The radial profiles show clear evidence for extended emission starting from ∼35 arcsec. the relatively low flux of SGR 1833-0832, the RXTE data were used only to study the timing properties of the source. We thus restricted our analysis to the data in Good Xenon mode, with a time resolution of 1 µs and 256 energy bins. They were extracted in the 2-10 keV energy range from all active PCUs (in a given observation) and all layers, and binned into light curves of 10-ms resolution.

X-RAY DATA ANALYSIS AND RESULTS
We inspected all observations for the presence of bursts by a careful examination of the light curves binned with different time resolutions. A few were found (besides those reported in Gögüş et al. 2010a, we found another short and weak burst in the RXTE data of 2010 May 18). In the analyses that follow, we removed the bursts from the event lists by applying intensity filters. Swift and RXTE data from a few contiguous observations carried out with the same instrumental setup were combined in order to achieve better statistics and higher signal to noise ratio. For the timing analysis, the data were corrected to the barycentre of the Solar system using the Chandra/UKIRT position reported in Gögüş et al. (2010a).

Spatial analysis
As can be seen in Fig. 1, some diffuse emission around SGR 1833-0832 was detected during the three XMM-Newton observations, mainly in the 3-6 keV energy band. Considering the large absorption derived from the X-ray spectrum, this is likely a halo produced by the scattering of the point source X-ray radiation by dust along the line of sight (Overbeck 1965;Tiengo et al. 2010; Rivera-Ingraham & van Kerkwijk 2010).
By analysing the spectrum from different annular regions, we have found that, as expected from a dust halo, the diffuse X-ray emission is significantly softer than the point source. Moreover, the XMM-Newton observation of the SGR 1833-0832 field performed in 2006, where no point-like nor diffuse emission was detected at the SGR position (see Section 3.4), favours the dust halo hypothesis with respect to steady diffuse emission, as expected, for example, in a supernova remnant. However, the data quality of the outburst observations is not good enough to detect significant variability in the extended emission at the level expected for a dust scattering halo.

Spectral analysis
For the spectroscopy (performed with the XSPEC 12.6 fitting package; Arnaud 1996) we concentrate first on the XMM-Newton spectra which, owing to the EPIC instrument high throughput and long observing time, are those with the best statistical quality. We fit the spectra from the three observations simultaneously with the hydrogen column density tied between all data sets. Photons having energies below 2 keV and above 10 keV were ignored, owing to the very few counts from SGR 1833-0832. This resulted in about 6100 ± 80, 5800 ± 90, and 4000 ± 70 net EPIC-pn counts in the three observations in chronological order. The abundances used are those of Anders & Grevesse (1989) and photoelectric absorption cross-sections are from Balucinska-Church & McCammon (1992).
Using the first twelve Swift/XRT observations in PC mode, Gögüş et al. (2010a) could not discriminate between a power-law and a blackbody as the model that better describes the spectrum of the SGR persistent emission. A fit of the XMM-Newton pn and MOS data with an absorbed power law yields a relatively high χ 2 value [χ 2 ν = 1.21 for 673 degrees of freedom (dof)] and structured residuals, while a better fit (χ 2 ν = 1.05 for 673 dof) is obtained using a blackbody model (corrected for the absorption). No additional spectral components are statistically required. The results of this simultaneous modelling are presented in Table 2 (the values of the spectral parameters were not significantly different when each observation was fit separately). The best-fitting hydrogen column density is (10.4 ± 0.2) × 10 22 cm −2 (here and in the following uncertainties are at 1σ confidence level, unless otherwise noted) and the blackbody temperature is consistent within the errors with being constant at kT ≃ 1.2 keV, while the observed flux slightly (but significantly, see Table 2) decreased from F ≃ 3.9 × 10 −12 erg cm −2 s −1 to 3.1×10 −12 erg cm −2 s −1 (in the 2-10 keV band) during the ∼20 days that separate the first and last XMM-Newton pointings. Assuming an arbitrary distance of 10 kpc, the corresponding XMM-Newton maximum and minimum luminosities are ∼9×10 34 erg s −1 and ∼7×10 34 erg s −1 (Table 2). For each XMM-Newton observation we performed phase-resolved spectroscopy by extracting the spectra for different selections of phase intervals. No significant variations with phase were detected, all the spectra being consistent with the model and parameters of the phase-averaged spectrum, simply re-scaled in normalisation.
Most magnetars, especially when in outburst, exhibit more complex spectra, usually fit by the superposition of two/three blackbodies or a blackbody and a high-energy power law (e.g. Mereghetti et al. 2005;Esposito et al. 2009;Bernardini et al. 2009Bernardini et al. , 2011Israel et al. 2007a;Enoto et al. 2009;Rea et al. 2009). The spectrum of SGR 1833-0832 is reminiscent of that reported by Esposito  Newton data, now public, clearly show that a more complex model is required for the emission of SGR 0418+5729. 5 In order to achieve better statistics and higher signal to noise, we merged the data from the first two XMM-Newton observations (since there is no evidence for variations in the flux and spectrum of the source between them) and accumulated a combined spectrum which is presented in Fig. 2. Again, the single-blackbody model provides an excellent fit (χ 2 ν = 1.00 for 352 dof), and no additional components are required. By including in the model a power-law component with photon index fixed to 3 (see e.g. Rea et al. 2009), we can set 3σ upper limits of ≈25% and ≈30% on the contribution of this component to the total observed and unabsorbed fluxes, respectively. This shows that the relatively low flux and high absorption of SGR 1833-0832 make even our deep XMM-Newton observations not very sensitive to the presence of a second spectral component in this SGR. We repeated the phase-resolved spectral analysis on the combined dataset. Once more, we see no evidence of spectral shape evolution with the rotational phase; in particular, no variation larger than the 1σ error (kδT ≃ 0.02 keV) was found in the blackbody temperatures at different phases.
To obtain flux measurements over the outburst, the Swift/XRT data were fit in the same way as the XMM-Newton ones, simultaneously and with all parameters left free to vary, except for the absorption column density, that this time was fixed at the value measured with XMM-Newton. This resulted in an acceptable fit (χ 2 ν = 1.01 for 602 dof) with spectral parameters similar to those reported in Table 2 (see Fig. 3). We plot the resulting long-term light curve in the top panel of Fig. 3.
The blackbody radius is calculated at infinity and for an arbitrary distance of 10 kpc. b In the 2-10 keV energy range.
c In the 2-10 keV energy range and for an arbitrary distance of 10 kpc.
uncertainties on the fluxes and moderate time-span, we observe that similarly good fits can be obtained with either an exponential function or a broken power-law model. For an exponential function of the form F (t) = A exp(−t/τ ), the best-fitting values (χ 2 ν = 0.84 for 23 dof) are A = (3.92 ± 0.05) × 10 −12 erg cm −2 s −1 and τ = (108 ± 9) day. 6 Adopting a broken power law (χ 2 ν = 0.73 for 22 dof;), the break occurs at (20 ± 3) d, when the index changes from α1 = −0.01 ± 0.02 to α2 = −0.54 ± 0.09; the flux at the break time is (3.7 ± 0.2) × 10 −12 erg cm −2 s −1 . In both cases we assumed as t = 0 the time of the Swift/BAT trigger. From the broken power-law fit in particular, it is apparent that the source flux is consistent with a constant value for the first ∼20 days. The two models considered are plotted in Fig. 3.

Timing analysis
In Esposito et al. (2010) we reported on our search for periodicities made on the first Swift/XRT observation (00416485000) by calculating a fast-Fourier-transform power spectrum. A very prominent peak occurs in the spectrum of that observation at 7.5653(4) s (the quoted uncertainty indicates the Fourier period resolution). Pulsations were clearly detected also in all the others Swift and XMM-Newton datasets, and in the RXTE ones up to 2010 October 30 (MJD 55499), when presumably the flux became too low for the PCA sensitivity.
In order to obtain a refined ephemeris for the longest possible baseline, we studied the pulse phase evolution in these observations by means of an iterative phase-fitting technique (see e.g. Dall'Osso et al. 2003). The fits were carried out in the range 2-10 keV with a χ 2 minimisation approach using MINUIT (James & Roos 1975).
Throughout the period covered by useful observations (∼225 days) the relative phases and amplitudes were such that the phase evolution of the signal could be followed unambiguously. A second-order polynomial, as employed in the recent analysis by Gögüş et al. (2010a) over the first ∼47 days, provides an unacceptable fit to the data, with χ 2 ν = 9.06 for 86 dof. We tried higher-order polynomials until the addition of a further (higherorder) term was not statistically significant at more than 3σ with respect to the null hypothesis (as evaluated by the Fisher test). The outcome of this process was a fourth-order polynomial (the improvement obtained in the fit with a fifth-order polynomial has 6 An additional component F 0 representing the base flux level [F (t) = F 0 + A exp(−t/τ )] is not required. Including it anyway in the fit (χ 2 ν ≃ 0.83 for 22 dof), we find F 0 = (5 ± 4) × 10 −13 erg cm −2 s −1 . keV energy range (two possible models describing the decay are also plotted: an exponential function with the solid magenta line and a broken power-law with the dot-dashed blue line; see Section 3.2), the blackbody temperature and radius (evaluated at infinity and assuming an arbitrary distance of 10 kpc), and the pulsed fraction. We assumed as t = 0 the time of the Swift/BAT trigger. a statistical significance of only 2.6σ), which we used to fit the phase shifts. The resulting phase-coherent solution is given in Table 3 and plotted in Fig. 4; the best fit (χ 2 ν = 1.07 for 84 dof) gives ν = 0.132 180 571(7) Hz andν = −6.0(5) × 10 −14 Hz s −1 , assuming MJD 55274.0 as reference epoch. We have checked that the positional uncertainty (0. ′′ 3; Gögüş et al. 2010a) does not significantly affect the rotational parameters resulting from our analysis.
In Fig. 5 we show the three XMM-Newton light curves ob- Table 3. Spin ephemeris of SGR 1833-0832 obtained using the combined RXTE, Swift, and XMM-Newton observations. We also give for convenience the corresponding period P and period derivativeṖ , as well as the derived characteristic age τc = P/(2Ṗ ), dipolar magnetic field B ≈ (3c 3 IPṖ /8π 2 R 6 ) 1/2 , and rotational energy lossĖ = 4π 2 IṖ P −3 (here we took R = 10 km and I = 10 45 g cm 2 for the star radius and moment of inertia, respectively).

Parameter Value
Range ( 3.2 × 10 32 tained folding the high time-resolution (nominal frame time of 73.4 ms) EPIC-pn data at our phase-coherent ephemeris. The pulse profile is sinusoidal and the pulsed fraction, that we define as the semi-amplitude of sinusoidal modulation divided by the mean source count rate, was consistent through the first two observations [(69.0 ± 1.4)% and (67.4 ± 1.5)%, respectively], while it decreased to (57.0 ± 1.7)% in the third one. We have investigated the morphology of the pulse phase distribution as a function of energy by comparing the XMM-Newton pulse profiles in different energy bands. The measured pulsed fractions in the three observations are (66 ± 2)%, (64 ± 2)%, and (56 ± 2)% in the soft (2-5 keV) energy band, and (75 ± 2)%, (71 ± 2)%, and (59 ± 3)% in the hard (5-10 keV) band. Apart from this marginal indication for an increasing trend of the pulsed fraction with energy, no significant pulse shape variations (such as phase shifts of the maxima) were found as a function of energy by cross-correlating or comparing through a two-sided Kolmogorov-Smirnov test the soft and hard folded profiles. The 2-10 keV pulsed fractions measured in the individual observations obtained with XMM-Newton and Swift are shown in the bottom panel of Fig. 3 (we did not consider the RXTE data as the non-imaging PCA instrument does not ensure reliable background subtraction). A constant fit of the pulsed fraction values derived with XRT 7 (and shown in Fig. 3) does not adequately describe the data (χ 2 ν = 4.52 for 24 dof); however no particular trend is apparent in the pulsed fraction evolution and the simple functions we tried do not yield significantly better fits.

Archival searches
No X-ray source was detected at the Chandra position of SGR 1833-0832 (Gögüş et al. 2010a) in the XMM-Newton observation performed in 2006. In a 20-arcsec circle centred at the SGR coordinates we detected in the 2-10 keV energy band 24 EPIC/pn 7 We did not attempt to fit simultaneously the XMM-Newton and Swift data in order to avoid possible effects due to the different responses and energy dependence of the effective areas of the of the EPIC and XRT detectors. counts (we did not use the MOS data since the source position fell near CCD gaps in both cameras). The net exposure time was 9.2 ks. Considering the expected background counts (estimated from large surrounding source-free regions) and assuming Poissonian fluctuations, we set a 3σ upper limit on the 2-10 keV countrate of SGR 1833-0832 of 2.3 × 10 −3 counts s −1 . Assuming a blackbody spectrum with kT = 1.2 keV and an absorption of NH = 10.4 × 10 22 cm −2 , this translates into an upper limit on the 2-10 keV observed flux of ∼4 × 10 −14 erg cm −2 s −1 and of ∼8 × 10 −14 erg cm −2 s −1 on the 2-10 keV unabsorbed flux. If instead we assume a much softer blackbody spectrum with kT = 0.3 keV, as it might be expected from a magnetar in quiescence (e.g. Lamb et al. 2002;Gotthelf et al. 2004;Bernardini et al. 2009 from the spectrum. The array was in the hybrid configuration H168 (maximum baseline ∼4.5 km). 8 Given the location of the source in the Galactic plane, images including the short baselines were too confused by the emission of extended sources to yield useful limits. Continuum images combining all usable frequency channels were then made considering only the long baselines, i.e. those including the antenna CA06. These images were still somewhat confused by other sources in the field and the rms noise near the field centre was twice that outside the ∼11 ′′ × 80 ′′ wide beam. In particular, the local rms at the SGR position turned out to be 0.3 mJy, whence the 3σ flux density limit for an unresolved continuum radio emission from SGR 1833-0832 is 0.9 mJy.

Parkes
The source was observed on 2010 March 25 (MJD 55279.954) for 90 minutes at a centre frequency of 1374 MHz with the central beam of the Parkes 20 cm multibeam (MB) receiver (Staveley-Smith et al. 1996). The total 288 MHz bandwidth was split into 96 channels, each 3 MHz wide, and the time series was 1 bit sampled every 1 ms. Data have been searched for periodicities around the value obtained from the X-ray observations over a wide range of dispersion measure values (DM < 6000 pc cm −3 , given the very high NH value derived from X-ray observations). A search for single de-dispersed pulses was also carried out.
No radio pulsation with a period matching (within ±2.5 ms) the X-ray period, nor half of it, has been found. The single pulse search only revealed the presence of a bright known pulsar in the beam (the 85-ms J1833-0827/B1830-08, at DM 411 pc cm −3 ; see Section 5). The upper limit on the pulsed flux density at 20 cm wavelength is of 0.09 mJy for a duty cycle of 10 per cent, calculated for the maximum DM investigated and assuming a minimum signal-to-noise ratio of 9. The latter relatively high value for the adopted minimum signal-to-noise ratio is due to the presence of strong radio interferences at ∼7.5 s, that affect the signal folded at the SGR period.

X-RAY EMISSION AROUND PSR B1830-08
Inspection of the XMM-Newton EPIC images reveals an excess of X-ray emission north of SGR 1833-0832, encompassing the position of the middle-aged radio pulsar PSR B1830-08 (spin-down age τc ∼ 150 kyr; Taylor, Manchester & Lyne 1993).
PSR B1830-08 is located outside, and possibly associated to, the asymmetric shell-type supernova remnant W41 (G23.3-0.3) despite an angular separation of 24 ′ to the centre of the remnant (Kassim 1992;Gaensler & Johnston 1995). Its distance estimated through the dispersion measure (DM = 411 pc cm 3 ) and the NE2001 model of the Galactic electron density distribution by Cordes & Lazio (2002) is 4.6 kpc. PSR B1830-08 might be also linked to the unidentified TeV source HESS J1834-087 lying inside W41 (Aharonian et al. 2006;Albert et al. 2006). In this case, the extended TeV source would be a relic pulsar wind nebula (PWN) generated by electrons ejected by PSR B1830-08 at its birth place and up-scattering the surrounding low-energy photons (e.g., Bartko & Bednarek 2008). However, different explanations have been suggested: the TeV emission of HESS J1834-087 could result from the π 0 decay of shock-accelerated hadrons interacting with a giant molecular cloud along the line of sight (Tian et al. 2007) or from the inverse Compton radiation of electrons in a different PWN recently discovered near the centre of W41 (Mukherjee et al. 2009;Misanovic, Kargaltsev & Pavlov 2011). The detection of a large-scale X-ray nebula trailing the pulsar in the direction of the supernova remnant and of the TeV source would support the association between PSR B1830-08 and HESS J1834-087.
We refined the imaging analysis by considering only EPIC-MOS data from the observations 0605852001 and 0605852101, whereas PSR B1830-08 was located in the unread regions of the MOS detectors operating in Large Window mode during observation 0605851901. The EPIC-pn data of the three observations were not used either, because of a higher level of residual soft proton contamination indicated by the ratio of the surface brightness inside the field of view to the one outside the field of view (De Luca & Molendi 2004). We selected only 1-and 2-pixel events with the default flag mask.
We built background maps using the Filter Wheel Closed (FWC) event lists provided by the XMM-Newton background working group. 9 As the profile of the FWC maps changed since the launch of the mission, we filtered only the events collected after January 2007 to produce a more accurate template of the particleinduced background. The background maps were re-scaled in nar- row energy bands according to the counts in the non-vignetted MOS corners. Finally, the images were mosaiced, backgroundsubtracted, exposure-corrected, and adaptively smoothed to achieve a signal-to-noise ratio of 6. Diffuse emission around PSR B1830-08 is apparent in the 2-10 keV band mosaic, extending ∼2 ′ north-east and south of the pulsar (Fig. 6, left panel). A closer inspection shows a compact structure at the pulsar position embedded in a halo of lower surface brightness (Fig. 6, right panel). The compact structure is ∼1 ′ elongated towards south-east, whereas the halo is mostly extending in the perpendicular direction. The bulk of the halo emission lies closer to PSR B1830-08, while the the south-western tip is apparently connected with SGR 1833-0832; this is most likely due to the smoothing procedure. Both the halo and the compact structures are hardly visible in the soft band (0.5-2 keV), suggesting an absorbed non-thermal emission. This is confirmed by the spectral analysis.
We produced the spectra also including the three EPIC-pn data sets. The source counts were extracted from the mutually exclusive elliptical regions shown in Fig. 6 (right panel). The background counts were estimated from 60 ′′ apertures placed on the same CCD of each camera, and at the same distance to the readout node as the source for the EPIC-pn. For each EPIC camera, we merged the individual spectra from the different observations into a single spectrum and combined the response matrices. The compact and halo regions contained 560±30 and 2800±160 background-subtracted counts, respectively, corresponding to an average surface brightness of 0.2 and 0.085 counts arcsec −2 . The spectra were rebinned with a minimum of 20 counts per spectral channel for the compact emission and 50 for the halo.
The hard diffuse emission around PSR B1830-08 is suggestive of a PWN. With a spin-down power ofĖ = 5.8×10 35 erg s −1 , PSR B1830-08 is energetic enough to produce a synchrotron nebula detectable in X-rays for a broad range of distances. For an assumed distance of 4.6 kpc, the measured fluxes imply a luminosity in the 2-10 keV energy range of ∼4×10 32 erg s −1 for the compact structure and of ∼1.5×10 33 erg s −1 for the halo. This corresponds to a conversion efficiency LX/Ė of 7 × 10 −4 and 3 × 10 −3 , within the observed range of pulsar/PWN systems (Possenti et al. 2002;Kargaltsev & Pavlov 2008).
The inner compact structure looks one-sided, and its brightness peak is coincident with the pulsar position. Such a morphology is reminiscent of a bow-shock PWN, in which the wind of a supersonically moving pulsar is confined by the ram-pressure of the surrounding medium (see Gaensler & Slane 2006, for a review). This phase occurs once the pulsar escapes its remnant, several ten thousand years after the supernova event, which is compatible with the age of PSR B1830-08. Furthermore, the proper motion of PSR B1830-08 (33 ± 5 mas yr −1 , corresponding to a linear velocity of 730d4.6 km s −1 ; Hobbs et al. 2004) lies in the high-velocity end of the pulsar population and it is most likely supersonic. The compact emission seems to be aligned with the pulsar proper motion (Fig. 6, right), the brightness profile declining gradually from the pulsar position to south-east and sharply to north-west.
The location opposite to the direction of the pulsar motion, the hard extended emission, and the column density suggest that the halo is also linked to PSR B1830-08. However, its morphology is different from observed (e.g., the 'Mouse' G359.23-0.82, Gaensler et al. 2004) and simulated (e.g., Bucciantini et al. 2005) bow-shock PWNe, in which the emission is tightly confined along the proper motion direction. The halo forms instead a broad diffuse region behind the pulsar, almost perpendicular to the proper motion direction.

SGR 1833-0832
In this work, we presented a new timing solution for SGR 1833-0832 based on a phase-coherent analysis of XMM-Newton, Swift, and RXTE data and valid in the range from MJD 55274.775 to 55499.170. Under the standard assumption that the neutron star slows down because of magnetic braking, it implies a characteristic age τc ≃ 35 kyr, a dipolar magnetic field B ≈ 1.6 × 10 14 G and a rotational energy loss rateĖ ≃ 3.2 × 10 32 erg s −1 (see Table 3).
The modelling of the phase shifts required a fourth-order polynomial which in general is an indication for timing noise ('polynomial whitening'). However, it is worth noting that the second derivative we measured is unlikely to be related to a random change of the pulse profiles, which is expected to introduce only a random distribution of the phase residuals, rather then a cubic term. Recent studies on a sample of 366 radio pulsars showed that cubic terms in the phase residuals are actually possible but smaller by several orders of magnitudes than those detected in SGR 1833-0832 10 and recorded on time-scales longer (years) than those we 10 The indicator commonly used to quantify the amount of timing noise in radio pulsars is the ∆ 8 parameter introduced by Arzoumanian et al. (1994). The value computed with the ephemeris of SGR 1833-0832 given in Table 3 is ∆ 8 ≃ 4.2, which is much higher than the typical values measured for radio pulsars (roughly between 0 and −5; Arzoumanian et al. 1994; are sampling in our dataset (Hobbs, Lyne & Kramer 2010). We also note that the (long-term) timing noise of young radio pulsars (age < 10 5 yr) can be better understood as resulting from the recovery from previous glitch events (Hobbs, Lyne & Kramer 2010), making it unlikely that the frequency second derivative is actually due to random noise.
A possibility is therefore that the higher-order frequency derivatives we observed are manifestations of a glitch recovery. In the AXP 1RXS J1708-4009, for instance, negative second frequency derivatives in the −(0.01÷1.3)×10 −20 Hz s −2 range has been detected just after a glitches Israel et al. 2007b). We found no evidence for glitches in the time spanned by our observations, but this does not exclude the possibility that such events occurred before the first observation of the outburst.
Another possibility is that the frequency second derivative we measured (or at least a significant fraction of it) is linked to the magnetospheric activity of SGR 1833-0832. In the magnetar scenario the spin derivative is in fact expected to increase while the magnetospheric twist is growing, i.e. for instance in periods preceding large outbursts, and to decrease in the aftermath (see e.g. Thompson, Lyutikov & Kulkarni 2002;Beloborodov 2009). On the other hand, as pointed out by Beloborodov (2009), a negativeν following a period of bursting activity (as that detected here) can be still accounted for if a magnetospheric twist is suddenly implanted but its strength is moderate (twist angle less than ∼1 rad). Then the twist may still grow for a while in spite of the luminosity released by dissipation monotonically decreases. Only after the twist angle has reached its maximum value, it will start to decay, together with the torque.
Nevertheless, the timing properties of the source appear consistent with those typically observed in the AXP/SGR class. The spectral characteristics of SGR 1833-0832 are instead somewhat less usual. The X-ray spectrum is well described by a single blackbody with temperature kT ∼ 1.2 keV, definitely higher than what generally observed both in transient and 'persistent' magnetars (kT 0.7 keV, with the exception of SGR 0418+5729; Esposito et al. 2010a;Rea et al. 2010). We cannot however exclude that the high temperature of SGR 1833-0832 partially results from a bias in the spectral fits that, given the paucity of counts below ∼2-3 keV (owing to the large absorption), acts in the direction of increasing the temperature, in order to account for a second spectral component which cannot be properly modelled (see Section 3.2).
The large increase in flux, a factor 20 above the quiescent level, is indicative of a rather powerful event, not dissimilar from those observed in other transient sources. This, however, seems to be in contrast with the low level of activity seen in SGR 1833-0832, from which only few bursts were detected (Gögüş et al. 2010a). A scenario in which we are presently witnessing the later stages of an outburst which was caught during its decay rather than at its onset, could account for the low activity but seems difficult to reconcile with the current high value of the temperature (if the thermal component in the spectrum of SGR 1833-0832 is indeed at kT ∼ 1.2 keV). A further possibility might be that SGR 1833-0832 experienced a (crustal) heating episode which, however, was not accompanied by (or did not trigger) a twisting of the external field, as possibly suggested by the non-detection of a hard X-ray tail. In this Hobbs, Lyne & Kramer 2010). Moreover, considering the correlation between timing noise and spin-down rate, one would expect for SGR 1833-0832 ∆ 8 to be approximately −0.6 (Hobbs, Lyne & Kramer 2010). case, however, an alternative explanation, not related to the twist evolution, for the presence of a negativeν must be sought (perhaps a long-term postglitch recovery, see above).
Immediately after the BAT trigger that led to the discovery of SGR 1833-0832, its observed flux was of ∼3.8 × 10 −12 erg cm −2 s −1 , and it decreased by ≈75% during the following 5 months (which is not unusual for magnetars; e.g. Rea & Esposito 2011), to ∼1.1 × 10 −12 erg cm −2 s −1 . Although present data do not allow us to discriminate among different decay patterns, the flux evolution can be satisfactorily described by a broken power law, similarly to the case of SGR 0418+5729. The flux of SGR 1833-0832 remained fairly constant for about 20 days (α1 ∼ 0), when the decay index changed to α2 ∼ −0.5. Interestingly, also the decay of SGR 0418+5729 exhibited a break at ∼20 days (Esposito et al. 2010a). As in the case of SGR 0418+5729 (see the discussion in Esposito et al. 2010a), the presence of a break, together with the rather flat decay indices, might be difficult to reconcile with the predictions of the deep crustal heating scenario (Lyubarsky, Eichler & Thompson 2002), F ∝ t −n/3 with n ∼ 2-3. This model provides a satisfactory description of the flux decay in SGR 1627-41 (Kouveliotou et al. 2003, although subsequent analyses found no evidence for the plateau around ∼ 400-800 d predicted by the model; Mereghetti et al. 2006) and AXP 1E 2259+586 (Zhu et al. 2008). At the same time, however, it cannot explain the variety of decay patterns which emerged from recent observations of SGRs/AXPs (e.g. Rea et al. 2009). This may reflect the intrinsic limitations of the model (which relies an a simplified, one-dimensional treatment), or point towards the existence of different heating mechanisms which are at work during the outbursts of magnetars, like in the case of SGR 1833-0832.
Prompted by previous detections of a transient (pulsed) radio emission following X-ray transient activity in other magnetars (Camilo et al. 2006(Camilo et al. , 2007Burgay et al. 2009), we observed SGR 1833-0832 with the ATCA and the 64-m Parkes radio telescope (Burgay et al. 2010;Wieringa et al. 2010). No evidence for radio emission was found, down to flux densities of 0.9 mJy and 0.09 mJy for the continuum and pulsed emissions, respectively. Similar upper limits on the pulsed radio emission, 0.1 mJy at 1.38 GHz and 0.2 mJy at 2.28 GHz for a 10% duty cycle, were obtained with the Westerbork Radio Synthesis Telescope in the same days (but not simultaneously with our observations; Gögüş et al. 2010a). Despite this rather intensive coverage, the negative results of radio searches performed so far on SGR 1833-0832 cannot anyway be taken as conclusive because of the rapid variability of the pulsed flux shown by the known radio magnetars (see e.g. Burgay et al. 2009). Moreover, for a distance d = 10 kpc these limits translates into a pseudo-luminosity L = Sd 2 ≈ 10 mJy kpc 2 , which is significantly smaller than the 1.4 GHz luminosity of the other known radio magnetars at their peak (∼100-400 mJy kpc 2 ; Camilo et al. 2006Camilo et al. , 2007Levin et al. 2010) but still much larger than the luminosities of some known ordinary radio pulsars. 11

PSR B1830-08 and HESS J1834-08
The discovery of the X-ray PWN generated by PSR B1830-08, suggestive of a pulsar bow-shock and a possible diffuse emission trailing the pulsar proper motion, provides new elements to the identification of HESS J1834-087. Accounting for a systematic 20% uncertainty, the distance of PSR B1830-08 is compatible with the one of W41, 4.0 ± 0.2 kpc (Tian et al. 2007), which is likely associated to HESS J1834-087 given the low probability of a chance superposition. The pulsar spin-down age is also consistent with the dynamical age of remnant, roughly 100 kyr (Tian et al. 2007). As pointed out by Mukherjee et al. (2009), the pulsar moves fast enough to have reached its current position starting only 40 kyr from the geometric centre of the remnant (see also Fig. 6, left  panel).
However, the Gamma-ray to X-ray flux ratio, F (E > 1 TeV) /F (2−10 keV) ∼ 50 for the compact emission and ∼10 for the total X-ray emission, is lower than the one expected from the inverse correlation between this quantity andĖ (Mattana et al. 2009), which should be around 600 for PSR B1830-08. If HESS J1834-08 is the relic PWN generated by PSR B1830-08, it is not very efficient in radiating in TeV gamma rays. Otherwise, it may be unrelated to PSR B1830-08. The nondetection of a more extended X-ray emission trailing the pulsar does not support the association with HESS J1834-087 either. Higher resolution and longer X-ray observations are needed to establish the nature of the halo and the detailed morphology of the compact emission, as well as singling out a possible contribution from the pulsar magnetosphere to the compact emission.