MAGNETAR-LIKE ACTIVITY FROM THE CENTRAL COMPACT OBJECT IN THE SNR RCW103

The Swift X-ray Telescope (XRT) has been monitoring 1E 161348–5055 almost monthly, starting from 2006 April (Esposito et al. 2011). We have analyzed all Swift-XRT observations in photon counting mode from 2006 April 18 until 2016 July 20 (93 pre-burst and 20 post-burst observations, for an exposure of 236.2 ks). The last Swift-XRT observation prior to the burst was performed on 2016 June 22 from 01:30 to 01:42 UT (finished ∼20 minutes before the burst trigger) and showed the source already in an enhanced X-ray state (1–10 keV observed flux of ∼1.2 × 10⁻¹⁰ erg cm⁻² s⁻¹), while the previous observation was on 2016 May 16 from 13:47 to 15:47 UT with the source still at a low flux rate (1–10 keV observed flux of ∼1.7 × 10⁻¹² erg cm⁻² s⁻¹; see Figures 1 and 2).

Zoom In Zoom Out Reset image size

The Swift data were processed and analyzed with usual procedures using the standard tasks included in the heasoft software package (v.6.19) and the calibration files in the 2016 May 02 caldb release. The Swift-XRT source counts were extracted from a circular region centered on the most accurate position of the CCO (${\rm{R}}.{\rm{A}}.={16}^{{\rm{h}}}{17}^{{\rm{m}}}36\buildrel{\rm{s}}\over{.} 23$, decl. = −51°02'246; De Luca et al. 2008) with a radius of 10 pixels (1 pixel = 236), and the background events from an annulus of radii of 10–20 pixels. Only the observation ∼20 minutes before the BAT trigger, which yielded a severe pile-up, had to be extracted excising the inner 3.5 pixels of the extraction region.

We analyzed the Swift-BAT data of the burst (trigger 700791, obs ID: 00700791000). The T90 duration of the event (the time during which 90% of the burst counts were collected) was 0.009 ± 0.001 s, and its total duration was ∼10 ms. These durations were computed by the Bayesian blocks algorithm battblocks on mask-weighted light curves binned at 1 ms in the 15–150 keV (Scargle 1998), where essentially all the emission is contained. For the burst only, mask-tagged light curves, images, and spectra were created. We extracted a 15–150 keV sky image and performed a blind source detection over the whole duration of the burst: a single, point-like source was detected at high significance (14.5σ) at the best-fit coordinates ${\rm{R}}.{\rm{A}}.={16}^{{\rm{h}}}{17}^{{\rm{m}}}29\buildrel{\rm{s}}\over{.} 62,\,\mathrm{decl}.\,=\,-51^\circ 03^{\prime} 07\buildrel{\prime\prime}\over{.} 9$, with an uncertainty radius of 1.5 arcmin (1σ, including a systematic error of 0.25 arcmin; Tueller et al. 2010). This position is consistent with a single known X-ray source: 1E 161348–5055 (see Figure 1). No other X-ray source was detected within the burst error circle in the XRT data, with a 3σ 0.5–10 keV detection limit of <0.003 counts s⁻¹. Together with the exceptionally high flux of 1E 161348–5055 at the epoch of the burst, this strongly points to the CCO in RCW 103 as the origin of the burst.

After the burst trigger, 1E 161348–5055 was observed with the Advanced CCD Imaging Spectrometer spectroscopic array (ACIS-S; Garmire et al. 2003) on board the Chandra X-Ray Observatory, starting on 2016 June 25 at 09:20:07 until 22:00:38 UT, for an on-source exposure time of 44.2 ks (obs ID: 18878). The ACIS-S was configured in continuous clocking (CC) mode with FAINT telemetry format, yielding a readout time of 2.85 ms at the expense of one dimension of spatial information. The source was positioned on the back-illuminated S3 chip.

We analyzed the data following the standard analysis threads⁹ with the Chandra Interactive Analysis of Observations software (ciao, v. 4.8; caldb v. 4.7.2). We accumulated the source photon counts within a box of dimension 3 × 3 arcsec² centered on the position of the CCO. The background was estimated by collecting photons within two rectangular regions oriented along the readout direction of the CCD, symmetrically placed with respect to the target and both lying within the remnant, whose spatial extension is ∼9 arcmin in diameter (Frank et al. 2015). The average source net count rate was 3.352 ± 0.009 counts s⁻¹, which guarantees no pile-up issues in the data set.

We have also analyzed all archival Chandra observations pointing at <30'' from our target (24 observations from 1999 September 26 until 2015 January 13; see Figure 2). Photons from TE mode observations were extracted from a 2'' circular region, and the background from an annulus with radii 4''–10''. These observations were used for the timing and spectral long-term analysis (see below). When necessary, we corrected for pile-up effects by using the model of Davis (2001).

Zoom In Zoom Out Reset image size

The NuSTAR (Harrison et al. 2013) observed 1E 161348–5055 starting on 2016 June 25 at 06:46:47 UT until June 26 at 18:42:50 UT, for a total on-source exposure time of 70.7 ks (obs ID: 90201028002), simultaneously with the Chandra observation (Section 2.2). The data were processed using version 1.6.0 of the NuSTAR Data Analysis Software (NuSTARDAS; using version 59 of the clock file to account for NuSTAR clock drifts caused by temperature variation). We used the tool nupipeline with default options for good time interval filtering to produce cleaned event files, and we removed time intervals corresponding to passages through the South Atlantic Anomaly. We ran the nuproducts script to extract light curves and spectra and generated instrumental response files separately for both focal plane modules (FPMA and FPMB). We collected the source counts within a circular region of 40'' radius around the CCO position. The background subtracted source count rate in the 3–79 keV was 0.27 ± 0.03 counts s⁻¹. We checked that a 30'' extraction region gives consistent results. Background was estimated from two 60'' circular regions in the same chip, one inside and one outside the ghost-rays-contaminated area. We verified that the two background estimations did not significantly affect spectral modeling (see Figure 3).

The light curve of the Swift-BAT burst shows a double-peak profile (Figure 1). We fit the spectra of the two peaks with single-component models typically used for magnetar bursts: a power law, a blackbody, and a bremsstrahlung component (e.g., Israel et al. 2008). Only the blackbody model provided an acceptable fit for both peaks. The first ∼5 ms of the event can be fit by a blackbody with kT = 9.2 ± 0.9 keV (χ²_ν = 1.03 (36 dof), null hypothesis probability (nhp) = 0.42), while for the second peak the blackbody temperature is kT = 6.0 ± 0.6 keV (χ²_ν = 1.22 (36 dof), nhp = 0.16). The corresponding total burst flux is (1.6 ± 0.2) × 10⁻⁶ erg cm⁻² s⁻¹ in the 15–150 keV range (corresponding to a luminosity of 2 × 10³⁹ erg s⁻¹). All errors are given at 1σ confidence level throughout the Letter, and we assume a 3.3 kpc distance (Caswell et al. 1975).

For the timing analysis, photon arrival times were reported to the solar system barycenter frame, using the DE200 ephemerides and the Chandra CCO position (see above). We performed a blind search both for fast periodic and aperiodic signals using our new Chandra and NuSTAR data sets, using the Xronos timing package as well as the Z_n² test (Buccheri et al. 1983). We did not find any periodic signal via Fourier transform, but in both observations we detected the known ∼6.67 hr periodic modulation (see Figure 3). We inferred 3σ pulsed fraction upper limits (as explained in Israel & Stella 1996), of 5% (0.01–10 Hz), 6% (10–100 Hz), and in the 7%–9% range for the highest sampled frequencies (100–200 Hz), for the Chandra observation. A similar analysis carried out on the NuSTAR data resulted on 3σ upper limits of 12% (0.01–3 Hz) and in the range 26%–34% at higher frequencies (3–1000 Hz).

Zoom In Zoom Out Reset image size

In Figure 3, we show the determination of the ∼6.67 hr period using the longest data sets in the X-ray archives, with the light curves of the two most extreme cases of a pure single peak (from XMM-Newton in 2005; De Luca et al. 2006), and a clear double peak (in 2016 June; this work). The 3–79 keV light curve of the NuSTAR data and the simultaneous 1–8 keV Chandra data were fit by two sinusoidal harmonics with fundamental periods 24095 ± 167 s (at TJD 17565.0) and 23983 ± 263 s (at TJD 17564.7), respectively. We also studied the profile as a function of the energy in the 1–25 keV band and found that the profile may smooth to a single peak with increasing energy (see the middle panel of Figure 3). Pulsed fractions (defined as the profile Max−Min/Max+Min) are: 40 ± 1% in the 1–8 keV Chandra band and 41 ± 5% for the 10–20 keV NuSTAR data.

Studying the timing properties of 1E 161348–5055 is complicated by the changing pulse profile. However, if we assume the ephemeris of Esposito et al. (2011; solution "A"; constant period: 24 030.42(2) s; see the paper for a discussion of the assumptions and the validity of the solution) and extrapolate the phase of the minimum predicted for the fundamental harmonic, this is consistent within 2σ (Δϕ = 0.03 ± 0.02 for the 1–8 keV Chandra profile, and Δϕ = 0.08 ± 0.04 for the 10–15 keV NuSTAR profile) with that of the second minimum in Figure 3, around phase ϕ ∼ 0.9 (implying $| \dot{P}| \lt 7\times {10}^{-10}$ s s⁻¹).

We started the spectral analysis (always using xspec v.12.9) by simultaneously fitting the new Chandra and NuSTAR observations (see Figure 3). We found that although the Chandra spectrum alone is well fit with two blackbodies, this is not the case when taking into account also the NuSTAR hard X-ray spectrum of 1E 161348–5055. A good fit is found for a model comprised of two absorbed (N_H = 2.05(5) × 10²² cm⁻²) blackbodies with temperatures of kT₁ = 0.52 ± 0.01 keV and kT₂ = 0.93 ± 0.05 keV, with radii of R₁ = 2.7 ± 0.7 km and R₂ = 0.4 ± 0.2 km, plus the addition of a power-law component with photon index Γ = 1.20 ± 0.25 (adding a constant between the two instruments to account for inter-calibration uncertainties, which was always within 10%). The total observed flux in the 0.5–30 keV energy range is (3.7 ± 0.1) × 10⁻¹¹ erg cm⁻² s⁻¹, and the joint fit gives χ²_ν = 1.04 (660 dof; nhp = 0.2). A model with a blackbody plus two power laws results in a slightly worse fit of χ²_ν = 1.2 (660 dof; nhp = 2 × 10⁻⁴) and bad residual shape.

To study the outburst history of 1E 161348–5055, we reanalyzed all the available Chandra, XMM-Newton, and Swift data of the source acquired from 1999 until 2016 July (see Figure 2). All spectra were fit by fixing the absorption column density to the value derived using Chandra (N_H = 2 × 10²² cm⁻²) plus two blackbody components (because in the 1–8 keV range the hard X-ray power law is not required by the fit and contributes less than 10% to the flux in this band). We show the extrapolated 0.5–10 keV luminosity in Figure 2. This source underwent two outbursts in the past ∼17 years. The first outburst can be empirically fit by a constant plus three exponential functions, resulting in a total (impulsive plus persistent) emitted energy in the 0.5–10 keV band of ${E}_{1\mathrm{st} \mbox{-} \mathrm{out}}\sim 9.9\times {10}^{42}$ erg. This outburst was characterized by heating of two different regions on the surface, with the two blackbodies in the X-ray spectra cooling and shrinking from the outburst peak until quiescence: kT₁ ∼ 0.6–0.4 keV (R₁ ∼ 5–1 km), and kT₂ ∼ 1.4–0.7 keV (R₂ ∼ 1.4–0.1 km). This new second outburst, which started <1 month before the SGR-like burst (see Section 2.1), shows similar energetic and spectral decomposition so far (${E}_{2\mathrm{nd} \mbox{-} \mathrm{out}}\sim 1.6\times {10}^{42}$ erg). Furthermore, our NuSTAR observation shows for the first time, that during the outburst peak this source emits up to ∼30 keV (certainly modulated until ∼20 keV; Figure 3).