XMM-Newton and NuSTAR Simultaneous X-Ray Observations of IGR J11215-5952

The XMM-Newton Observatory (Jansen et al. 2001) carries three X-ray telescopes, each with a European Photon Imaging Camera (EPIC; 0.4–12 keV) at the focus: two with MOS CCDs (Turner et al. 2001) and one with a pn CCD (Strüder et al. 2001). Reflection Grating Spectrometer (RGS; 0.4–2.1 keV) arrays (den Herder et al. 2001) are placed at two of the telescopes.

The XMM-Newton observation targeted on IGR J11215–5952 started on 2016 February 14 at 21:25 and ended the day after at 03:36, resulting in net exposure times of 22 ks and 15.5 ks for the MOS and the pn, respectively. All EPIC cameras were operated in small window mode and adopted the medium filter.

We reduced the data using version 15.0.0 of the Science Analysis Software (SAS) with standard procedures and the most updated calibration files. We used the HEASoft version 6.18 to perform the data analysis.

At first, we extracted EPIC source counts from circular regions of 30'' radius, adopting PATTERN selection from 0 to 4 in the pn and from 0 to 12 in MOS spectra. We evaluated the possible presence of pile-up in EPIC data with epatplot, finding moderate pile-up only in MOS data. This suggested that we should excise the core of the point-spread function (PSF) in both MOS data, extracting MOS1 and MOS2 spectra from annular regions with inner radius of 5'' and an outer radius of 30''. Background spectra were obtained from similar sized regions, offset from the source positions. The full exposure time was exploited and no further filtering was applied to the data.

The SAS task rgsproc was used to extract the RGS1 and RGS2 spectra, which were later combined into one single grating spectrum using rgscombine in SAS.

Appropriate response matrices were generated using the SAS tasks arfgen and rmfgen. EPIC spectra were analyzed in the energy range 0.4–12, while the RGSs, which operated in spectroscopy mode (den Herder et al. 2001), were analyzed in the range 0.4–2.1 keV.

To ensure applicability of the χ² statistics, the spectra were rebinned such that at least 20 counts per bin were present. Free relative normalizations between the spectra were included to account for uncertainties in instrumental responses, always fixing the normalization of the EPIC pn spectrum at 1. Spectral uncertainties are given at 90% confidence for one interesting parameter.

In the spectral fitting, we used the photoelectric absorption model TBabs in xspec adopting the interstellar abundances of Wilms et al. (2000) and photoelectric absorption cross-sections of Balucinska-Church & McCammon (1992).

The final, time-averaged, source net count rates were the following: 5.67 ± 0.02 counts s⁻¹ (EPIC pn), 1.143 ±0.007 counts s⁻¹ (MOS1, PSF core excised), 1.130 ± 0.007 counts s⁻¹ (MOS2, PSF core excised), and 0.027 ± 0.001 counts s⁻¹ (combined RGS1 and RGS2 1st order spectra).

The high absorption toward the source (column density, N_H, around 10²² cm⁻²) implied significant detection in the RGSs only above 1 keV. This limited the available RGS band to the range 1–2 keV. The low statistics leave us with an apparent featureless RGS spectrum. Therefore we will not discuss it further in the paper.

IGR J11215–5952 was observed by NuSTAR as part of the joint XMM-Newton/NuSTAR programme approved during XMM-Newton Announcement of Opportunity AO14, with the main aim of searching for cyclotron lines.

NuSTAR observations started on 2016 February 14 at 17:11 and ended the day after at 05:01, for a net exposure time of 20.3 ks.

NuSTAR (Harrison et al. 2013) is the first focusing hard X-ray telescope above 10 keV and was launched in 2012. It carries two identical co-aligned telescopes that focus X-ray photons onto two independent Focal Plane Modules, called A and B (hereafter FPMA and FPMB). Each FPM contains four (2 × 2) solid-state cadmium zinc telluride (CdZnTe) pixel detectors which provide a spatial resolution of 18'' (full width at half maximum, FWHM), a spectral resolution of 400 eV (FWHM) at 10 keV, and a 12' × 12' field-of-view (FOV). The effective area is calibrated in the energy range 3–78 keV (Madsen et al. 2015). The low satellite orbit (with a ∼90 minute period) produces data gaps lasting about 30 minutes for every orbit in the NuSTAR observations.

We processed NuSTAR data (obsID 30101008002) with version 1.5.1 (2015 June 9) of the NuSTAR data analysis software (NuSTAR DAS), using CALDB version 20150316. Spectra and light curves were extracted with nuproducts on the cleaned event files. Circular extraction regions with a 60'' radius were used around IGR J11215–5952 and for the background, well away from the PSF source wings. NuSTAR observation was not contaminated by any stray light produced by sources outside the FOV. Since the background was stable and constant along the observation, no further filtering was applied. NuSTAR source light curves with arrival times corrected to the Solar System barycenter were extracted with the NuSTAR DAS tool nuproducts and "barycorr = yes." When needed, good time intervals were generated by using xselect and then running nuproducts using the "usrgtifile" keyword to correctly extract the temporal selected spectra.

The source net count rates were the following (3–78 keV): 1.783 ± 0.009 counts s⁻¹ (FPMA) and 1.678 ± 0.009 counts s⁻¹ (FPMB). Spectra were rebinned in order to have at least 20 counts per bin to ensure the applicability of the χ² statistics. Spectra from FPMA and FPMB were simultaneously fitted in xspec, without merging them together. Cross-calibration constants were used during the spectral analysis to take into account calibration uncertainties.

The same procedure was used when XMM-Newton and NuSTAR spectra were jointly fitted together.

After acceptance of the joint XMM-Newton/NuSTAR AO14 observations, we asked for a monitoring with Swift around 2016 February 14. Five snapshots (2 ks duration each), from 2016 February 11 to February 19 were performed, with the main aim of monitoring the source light curve being to put the XMM-Newton and NuSTAR IGR J11215–5952 intensity into context of the outburst evolution.

The Swift/XRT source light curve (Figure 1, top panel) was built making use of the XRT data products generator available at the UK Swift Science Data Centre (Evans et al. 2007).

Zoom In Zoom Out Reset image size

In Figure 1 we show the source light curve observed by Swift, NuSTAR, and XMM-Newton during the outburst. In particular, the Swift monitoring demonstrates that both NuSTAR and XMM-Newton observations caught the transient at its peak, as expected.

In Figure 2 we show only the light curves as observed by XMM-Newton (EPIC pn, 0.4–12 keV) and NuSTAR (FPMA, for clarity, 3–78 keV) binned on the known spin (187 s, see below), to avoid variability due to the neutron star rotational period. The intensity variability is large, in excess of one order of magnitude. The uninterrupted XMM-Newton light curve clearly shows that each satellite gap in the NuSTAR observation is actually filled by bright flares, repeating on a timescale of 2–2.5 ks.

Zoom In Zoom Out Reset image size

We performed timing analysis on EPIC (all three cameras, 2–12 keV) and NuSTAR data (FPMA and FPMB, 3–78 keV) after correcting arrival times to the Solar System barycenter, applying epoch folding techniques (Leahy et al. 1983) to measure the known neutron star spin period. We determined a periodicity of 187.0 ± 0.4 s, consistent with what was observed in 2007 with XMM-Newton (Sidoli et al. 2007) and with RXTE/PCA (Swank et al. 2007).

In Figure 3 we show the pulse profiles extracted from the strictly simultaneous XMM-Newton and NuSTAR data, in different energy ranges, together with four hardness ratios. The pulse profile is energy dependent, being dominated by an asymmetric main peak below 12 keV, which becomes symmetric at harder energies. The hardness ratios display different behavior along the pulse phase, as shown in the bottom panels in Figure 3.

Zoom In Zoom Out Reset image size

The pulsed fraction, PF, calculated as PF ≡ (F_max − F_min)/(F_max + F_min), where F_max and F_min are the count rates at the maximum and at the minimum of the pulse profile, respectively, increases with energy, from 24 ± 3% (0.4–3 keV) to 60 ± 14% (20–78 keV). Over the whole XMM-Newton energy band PF = 26 ± 2% (0.4–12 keV), while PF = 29 ± 3% over the whole NuSTAR band (3–78 keV).

3.2.1. Time-averaged XMM-Newton and NuSTAR Spectrum

We extracted NuSTAR spectra overlapping in time with the whole XMM-Newton observation to investigate the broadband (0.4–78 keV) time-averaged IGR J11215–5952 emission. This translated into a net exposure time of ∼9 ks for each NuSTAR FPM.

The joint fit of the five spectra (EPIC pn, MOS1, MOS2, FPMA and FPMB) with a single absorbed power-law model provided unacceptable results (${\chi }_{\nu }^{2}=2.89$ for 3292 degrees of freedom, dof; Figure 4), because of an evident rollover at hard energies, together with a soft excess (below 1.5 keV) and positive residuals at ∼6.4 keV, consistent with a K_α emission line from neutral iron. We tried different double-component continuum models. The best fit was obtained by adopting a blackbody (with a temperature, kT_BB, of ∼1.6 keV, and a radius, R_BB, of a few hundred meters, calculated assuming a distance of 7 kpc), together with a hard power law (photon index, Γ, of 0.6) modified by a cutoff at high energy (hereafter PLCUT), defined as M(E) = (exp[(E_cut − E)/E_fold]) when E ≥ E_cut, while M(E) = 1 at E ≤ E_cut. An additive Gaussian line in emission around 6.4 keV was also needed. The results are reported in Table 1 (Model 1) and shown in Figure 5. The presence of a blackbody component was already known from previous XMM-Newton observations of IGR J11215-5952 performed in 2007 (Sidoli et al. 2007) and it usually accounts for soft excesses detected in several accreting pulsars (La Palombara & Mereghetti 2006). We also tried an alternative decomposition of the broadband spectrum, replacing the blackbody component with a partial covering fraction absorption (pcfabs model in xspec). This model consists of an additional intrinsic absorption (N_Hpcfabs) with a covering fraction f_pcfabs (0 < f_pcfabs < 1), so that the continuum is:

$$\begin{eqnarray*}&&\mathrm{Intensity}={e}^{-\sigma {N}_{{\rm{H}}}}\,[{{fe}}^{-\sigma {N}_{\mathrm{Hpcfabs}}}\,+\,(1-f)]\,({{\rm{I}}}_{\mathrm{PLCUT}}).\end{eqnarray*}$$

A partially covered PLCUT model (hereafter Model 2) provided an equivalently good description of the data.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 1.  Broadband Time-averaged Spectrum (0.4–78 keV)

Parameter	Model 1	Model 2
NH (1022 cm−2)	1.05 ± 0.06	${1.27}_{-0.09}^{+0.07}$
NHpcfabs (1022 cm−2)	none	${4.5}_{-0.9}^{+0.6}$
fpcfabs	none	0.51 ± 0.04
kTBB (keV)	${1.61}_{-0.07}^{+0.06}$	none
RBB (m)	460 ± 30	none
Γa	${0.56}_{-0.11}^{+0.08}$	${0.97}_{-0.08}^{+0.11}$
Ecut (keV)	${9.7}_{-1.0}^{+0.9}$	${5.5}_{-0.4}^{+0.8}$
Efold (keV)	14.3 ± 1.4	${15.3}_{-1.4}^{+1.8}$
Eline (keV)	${6.44}_{-0.03}^{+0.03}$	${6.45}_{-0.03}^{+0.03}$
σline (keV)	<0.09	<0.09
Fluxline (ph cm−2 s−1)	(3.2 ± 0.8) × 10−5	(${3.2}_{-0.4}^{+0.9}$) × 10−5
EWlineb (eV)	40 ± 10	${41}_{-5}^{+12}$
Fluxc (erg cm−2 s−1)	(1.7 ± 0.1) × 10−10	(1.8 ± 0.1) × 10−10
Luminosityc (erg s−1)	(1.0 ± 0.1) × 1036	(1.1 ± 0.1) × 1036
${\chi }_{\nu }^{2}$/dof	1.042/3285	1.057/3285

Notes.

^aPower-law photon index. ^bEquivalent width. ^cFluxes (corrected for the absorption) and luminosities are in the energy range 0.1–100 keV. The flux reported for Model 2 is corrected for both absorbing components.

Download table as:  ASCIITypeset image

Note that consistent spectral results were obtained from strictly simultaneous XMM-Newton and NuSTAR time-averaged spectra, but we considered here the whole XMM-Newton exposure time to better constrain the parameters of the iron emission line.

3.2.2. Temporal-selected Spectroscopy

Cyclotron resonant scattering features in X-ray pulsars with standard neutron star magnetic fields (a few 10¹² Gauss) can be detected in the X-ray spectrum at energies E > 10 keV. Thus, in principle, if they are also strong enough in IGR J11215–5952, they should be found in the NuSTAR spectra alone. Therefore, we started concentrating on NuSTAR data only, extracting spectra from the peaks of the flares, as shown in Figure 2 (numbers in the lower panel).

In the temporal-selected spectroscopy using only NuSTAR data, the soft blackbody component, the iron line parameters, and the absorption column density could not be constrained. Since we were interested in searching for absorption features at hard energies (E > 10 keV), we preferred not to include the blackbody component and the iron line in the spectral model. The absorbing column density was a free parameter in the fit. In this manner, it naturally converged to very high (and unrealistic) values, with respect to the time-averaged spectroscopy of XMM-Newton and NuSTAR simultaneous data. As a first attempt, we simply adopted this absorbed PLCUT model as a pure phenomenological and statistically acceptable description of the continuum emission. The results from fitting the eight time-resolved 3–78 keV spectra in this way are reported in Table 2.

Table 2.  Time-resolved Spectroscopy (3–78 keV), Fitting Only NuSTAR Data of the Flare Peaks

Parameter	Spec. 1	2	3	4	5	6	7	8
NH (1022 cm−2)	${7}_{-4}^{+5}$	8 ± 3	6 ± 3	${5}_{-5}^{+6}$	9 ± 4	9 ± 4	${5}_{-5}^{+6}$	${3}_{-3}^{+6}$
Γ	${1.4}_{-0.3}^{+0.5}$	${1.6}_{-0.2}^{+0.2}$	${1.04}_{-0.22}^{+0.22}$	${1.27}_{-0.30}^{+0.37}$	${1.69}_{-0.61}^{+0.15}$	${1.64}_{-0.10}^{+0.11}$	${1.6}_{-0.7}^{+0.4}$	${1.28}_{-0.60}^{+0.19}$
Ecut (keV)	6 ± 6	${7}_{-2}^{+3}$	${3}_{-3}^{+4}$	${7}_{-7}^{+5}$	${15}_{-15}^{+8}$	${22}_{-3}^{+13}$	${9}_{-9}^{+14}$	${8}_{-8}^{+5}$
Efold (keV)	${27}_{-8}^{+35}$	${33}_{-13}^{+20}$	${15}_{-3}^{+5}$	${25}_{-8}^{+22}$	${40}_{-20}^{+50}$	${19}_{-19}^{+28}$	${30}_{-12}^{+40}$	${20}_{-9}^{+9}$
F (10−10 erg cm−2 s−1)a	3.0 ± 0.3	3.6 ± 0.4	2.5 ± 0.2	3.2 ± 0.3	4.6 ± 0.3	3.1 ± 0.2	2.3 ± 0.7	1.7 ± 0.2
L (1036 erg s−1)a	1.8 ± 0.2	2.1 ± 0.2	1.5 ± 0.1	1.9 ± 0.2	2.7 ± 0.2	1.8 ± 0.1	1.3 ± 0.4	1.1 ± 0.1
${\chi }_{\nu }^{2}$/dof	1.046/239	0.949/347	0.998/282	1.121/155	0.900/164	0.887/160	0.602/89	0.839/94

Note.

^aFluxes (corrected for the absorption) and luminosities are in the energy range 1–100 keV.

Download table as:  ASCIITypeset image

In some of these spectra, negative residuals appeared at energies between 15 and 40 keV. To account for these absorption features, we modified the PLCUT continuum multiplying it with a cyclotron absorption model (cyclabs in xspec package). At first, we calculated the line significance by increasing the Δχ² value until the confidence region boundary of the depth (Depth_cycl in Table 3) of the cyclotron line crosses 0. The significance of these lines was greater than 2σ in five (of eight) spectra (flare numbers 1, 2, 3, 4, and 6), so we have reported absorption line parameters only for these spectra in Table 3 for clarity. The absorption lines were measured with a significance ranging from σ = 2.3 to σ = 4.4 (with cyclabs model), with the best value obtained during the first flare observed by NuSTAR. The centroid energy was different: in spectra 1, 4, and 6 the feature was in a narrow range around 17–18 keV, while during spectrum 3 it was at ∼33 keV. In spectrum 3, the line at ∼17 keV was undetected: we could place an upper limit (Depth_cycl ≲ 0.18) to the depth of the cyclabs line with the centroid energy fixed at 17 keV (and line width of 2 keV).

Table 3.  Same as Table 2 (NuSTAR Only), Considering Only Flare Peaks 1, 2, 3, 4, and 6, Where the Inclusion of an Absorption Feature Resulted in a Line Significance Larger than 2σ

Parameter	Spec. 1	2	3	4	6
NH (1022 cm−2)	10 ± 3	${7}_{-7}^{+4}$	6 ± 4	${4}_{-4}^{+6}$	${6}_{-4}^{+4}$
Γ	${1.6}_{-0.2}^{+0.2}$	${1.40}_{-1.39}^{+0.19}$	${1.4}_{-0.4}^{+0.3}$	${1.15}_{-0.32}^{+0.31}$	${1.49}_{-0.14}^{+0.14}$
Ecut (keV)	${19}_{-9}^{+4}$	${6.3}_{-6.3}^{+2.2}$	${7}_{-7}^{+4}$	${6.4}_{-6.4}^{+1.8}$	${20}_{-3}^{+3}$
Efold (keV)	${21}_{-8}^{+21}$	${27}_{-13}^{+16}$	${31}_{-15}^{+95}$	${23}_{-7}^{+12}$	${15}_{-5}^{+9}$
Depthcycla	${0.44}_{-0.17}^{+0.18}$	${0.16}_{-0.11}^{+0.16}$	${0.74}_{-0.46}^{+2.9}$	${0.38}_{-0.26}^{+0.31}$	${0.46}_{-0.25}^{+0.27}$
Ecycla (keV)	${16.7}_{-0.9}^{+1.0}$	${15}_{-5}^{+3}$	${32.5}_{-3.4}^{+4.2}$	18.0 ± 1.0	17.2 ± 0.7
Widthcycla (keV)	${2.6}_{-1.5}^{+3.2}$	${1.6}_{-1.6}^{+14}$	${4.9}_{-4.9}^{+11.3}$	${1.0}_{-1.0}^{+2.4}$	${1.0}_{-1.0}^{+1.2}$
Line Signif.b	4.4σ	2.3σ	2.7σ	2.5σ	3.25σ
${\chi }_{\nu }^{2}$/dof	1.012/236	0.942/344	0.984/279	1.103/152	0.837/157

Notes.

^a cyclabs model in xspec. ^bThis line significance was calculated by increasing the Δχ² value until the confidence region boundary of the depth (Depth_cycl) of the cyclotron line crosses 0.

Download table as:  ASCIITypeset image

In order to test the significance of the absorption line, since the F-test could bring unreliable results (Protassov et al. 2002), we used Monte Carlo simulations to confirm the presence of the cyclotron line in spectrum 1 (our best case, Figure 6), following the approach described by several authors (Barrière et al. 2015; Bhalerao et al. 2015; Lotti et al. 2016). We used the simftest routine in xspec with 100,000 simulations,⁶ fixing the N_H to the values reported in Table 3 to speed up the process, and found that the probability of such a line to be required by the data is P = 99.15% for spectrum 1, corresponding to 2.63σ (normalized to 8 trials). The significance of the line did not improve if a blackbody component was included in the fitting model to the flare 1 spectrum, with temperature, radius, and absorption fixed to the broadband time-averaged values.

Zoom In Zoom Out Reset image size

To obtain a global view of the source behavior as observed by NuSTAR, we combined the two FPM source event files and built an "energy versus time" plot, which was then normalized to the time-averaged NuSTAR spectrum and the energy-integrated light curve. Each pixel of this image represents the count rate in a 187 s time interval and 1 keV energy bin, divided by the time averaged count rate in the same energy bin during the whole observation and by the ratio of the 3–40 keV rate in the same time interval and in the full observation. The result is shown in Figure 7: the different colors in the image mark the residuals of the source counts at different energies with respect to the time-averaged spectrum and broadband light curve (color level at 1), along the NuSTAR observation (the gaps due to the eight satellite revolutions are clearly visible). We note that this is the only time where we combined the two NuSTAR FPMs (here not barycentered) events, to get better statistics. From the visual inspection of Figure 7, some dark horizontal lines (deficit of counts with respect to the time-averaged NuSTAR spectrum) appear in the first four satellite orbits, between 15 and 35 keV, confirming the hint of absorption features suggested by the spectral analysis. We note that, although this plot is not able to assess the significance of any negative residual, it nevertheless shows that we did not miss other eventual spectral features with our particular time (and intensity) resolved spectral extraction.

Zoom In Zoom Out Reset image size

A further spectral investigation was performed considering flare peaks simultaneously observed by XMM-Newton and NuSTAR. This led to four flare peaks (marked with numbers 4, 5, 6, and 7). We extracted strictly simultaneous spectra (0.4–78 keV) and performed a new broadband spectroscopy. We adopted the same models used to fit the time-averaged 0.4–78 spectrum (Table 4). When required by the presence of negative residuals, we also report the fit results obtained with the inclusion of a cyclabs model, although in no case is the line significant. We note that similar results were obtained using a Gaussian line in absorption (gabs in xspec) instead of cyclabs, and a Comptonization continuum model (compTT) instead of a PLCUT model.

Table 4.  Broadband (0.4–78 keV) Spectroscopy of the Flares Simultaneously Observed by XMM-Newton and NuSTAR

Model 1a Parameters	flare 4	4	5	6	6	7
NH (1022 cm−2)	${1.08}_{-0.29}^{+0.28}$	${0.94}_{-0.30}^{+0.31}$	${0.89}_{-0.24}^{+0.25}$	${1.15}_{-0.20}^{+0.21}$	${1.00}_{-0.24}^{+0.24}$	${1.3}_{-0.1}^{+0.1}$
kTBB (keV)	${1.68}_{-0.22}^{+0.43}$	${1.47}_{-0.18}^{+0.27}$	${1.51}_{-0.16}^{+0.39}$	${1.88}_{-0.10}^{+0.13}$	${1.71}_{-0.12}^{+0.15}$	${1.94}_{-0.07}^{+0.07}$
RBB (m)	${480}_{-120}^{+180}$	${600}_{-160}^{+250}$	${630}_{-120}^{+200}$	${450}_{-50}^{+60}$	${520}_{-80}^{+90}$	${380}_{-10}^{+10}$
Γ	${0.54}_{-0.36}^{+0.23}$	${0.29}_{-0.94}^{+0.31}$	${0.43}_{-0.40}^{+0.25}$	${0.85}_{-0.14}^{+0.13}$	${0.64}_{-0.29}^{+0.19}$	${1.11}_{-0.05}^{+0.05}$
Ecut (keV)	6 ± 6	${6.4}_{-1.6}^{+1.2}$	${7}_{-7}^{+11}$	${21}_{-2.2}^{+2.5}$	${18.6}_{-3.2}^{+2.3}$	${29}_{-2}^{+3}$
Efold (keV)	${15}_{-4}^{+6}$	${12}_{-4}^{+4}$	${13}_{-3}^{+5}$	${12}_{-4}^{+7}$	${9.9}_{-2.5}^{+3.5}$	<25
Depthcycl	none	${0.45}_{-0.28}^{+0.30}$	none	none	${0.64}_{-0.33}^{+0.32}$	none
Ecycl (keV)	none	${18.1}_{-1.8}^{+1.0}$	none	none	${17.4}_{-0.67}^{+0.64}$	none
Widthcycl (keV)	none	${1}_{-1}^{+4}$	none	none	${1}_{-1}^{+1.4}$	none
Line Signif.b	none	2.8σ	none	none	3.35σ	none
F (10−10 erg cm−2 s−1)c	2.8 ± 0.4	2.8 ± 0.4	3.1 ± 0.4	2.4 ± 0.2	2.4 ± 0.2	1.6 ± 0.1
L (1036 erg s−1)c	1.6 ± 0.2	1.6 ± 0.4	1.8 ± 0.2	1.4 ± 0.1	1.4 ± 0.1	0.94 ± 0.06
${\chi }_{\nu }^{2}$/dof	0.989/457	0.978/454	1.018/503	0.869/492	0.851/489	0.929/271
Model 2a Parameters	flare 4	4	5	6	6	7
NH (1022 cm−2)	${1.47}_{-0.38}^{+0.28}$	${1.37}_{-0.42}^{+0.30}$	${1.30}_{-0.43}^{+0.28}$	${1.37}_{-0.26}^{+0.24}$	${1.6}_{-0.2}^{+0.2}$	${1.52}_{-0.79}^{+0.38}$
NHpcfabs (1022 cm−2)	8 ± 4	${7}_{-3}^{+4}$	7 ± 4	${8}_{-4}^{+6}$	${10.6}_{-2.3}^{+2.8}$	${10}_{-9}^{+14}$
fpcfabs	${0.55}_{-0.14}^{+0.09}$	${0.51}_{-0.13}^{+0.12}$	${0.55}_{-0.13}^{+0.08}$	${0.49}_{-0.15}^{+0.18}$	${0.66}_{-0.05}^{+0.04}$	${0.4}_{-0.3}^{+0.4}$
Γ	${1.16}_{-0.24}^{+0.20}$	${1.05}_{-0.21}^{+0.24}$	${1.16}_{-0.27}^{+0.19}$	${1.04}_{-0.24}^{+0.40}$	${1.45}_{-0.10}^{+0.10}$	${1.1}_{-0.4}^{+0.6}$
Ecut (keV)	${6.7}_{-1.3}^{+1.2}$	${6.2}_{-1.1}^{+1.4}$	${6.7}_{-1.5}^{+1.0}$	${5}_{-5}^{+4}$	${20}_{-3}^{+2}$	${5}_{-5}^{+5}$
Efold (keV)	20 ± 5	${20}_{-4}^{+6}$	${20}_{-6}^{+7}$	${18}_{-5}^{+16}$	${15}_{-5}^{+8}$	${15}_{-5}^{+20}$
Depthcycl	none	${0.43}_{-0.29}^{+0.28}$	none	none	${0.52}_{-0.26}^{+0.26}$	none
Ecycl (keV)	none	${18.0}_{-1.8}^{+1.0}$	none	none	${17.2}_{-0.7}^{+0.6}$	none
Widthcycl (keV)	none	${1}_{-1}^{+5}$	none	none	${1}_{-1}^{+1.8}$	none
Line Signif.b	none	2.8σ	none	none	3.87σ	none
F (10−10 erg cm−2 s−1)c	3.17 ± 0.11	3.15 ± 0.1	3.6 ± 0.1	2.8 ± 0.1	2.8 ± 0.1	2.0 ± 0.1
L (1036 erg s−1)c	1.86 ± 0.07	1.85 ± 0.06	2.1 ± 0.06	1.6 ± 0.1	1.6 ± 0.1	1.2 ± 0.1
${\chi }_{\nu }^{2}$/dof	0.998/457	0.987/454	1.029/503	0.871/492	0.863/489	0.943/271

Notes.

^aModel 1 is a double-component continuum made of a blackbody plus a PLCUT model. Model 2 is a PLCUT model modified by a partial covering fraction absorption pcfabs. ^bThis line significance was calculated by increasing the Δχ² value until the confidence region boundary of the depth (Depth_cycl) of the cyclotron line crosses 0. ^cFluxes (corrected for the absorption) and luminosities are in the energy range 0.1–100 keV.

Download table as:  ASCIITypeset image

3.2.3. Not Only Flares: The Lowest Intensity State

To investigate the broadband spectrum taken from the lowest intensity state where simultaneous XMM-Newton and NuSTAR data were available, we accumulated a spectrum from the valleys just before and after flare 4. A meaningful spectroscopy could be performed in the energy band 0.8–30 keV. Also in this low luminosity state (L_X ∼ 3 × 10³⁵ erg s⁻¹, 0.1–100 keV, see below), a fit with a single absorbed power law revealed structured residuals, with a low energy excess and a roll-over at high energy (Figure 8). A simple double-component model, with a power law plus a blackbody (equally absorbed), already provided a good fit (${\chi }_{\nu }^{2}/\mathrm{dof}=0.939/267$) to the emission (0.8–30 keV), resulting in an absorbing column density N_H = 2 × 10²² cm⁻², Γ = 1.3, kT_BB = 1.3 keV and a blackbody radius R_BB ∼ 150 m. In order to compare the spectral properties of the low state with other intensity states, we adopted both Model 1 and Model 2, fixing both Fe line properties and the power law cutoff to the time-averaged spectral shapes, since in the low intensity state the latter were unconstrained. The results are reported in Table 5, to be compared with Table 4 (flares) and Table 1 (time-averaged spectrum).

Zoom In Zoom Out Reset image size

Table 5.  Broadband Spectrum during the Lowest Intensity State (0.8–30 keV)

Parameter	Model 1	Model 2
NH (1022 cm−2)	${1.52}_{-0.33}^{+0.37}$	${0.84}_{-0.77}^{+0.93}$
NHpcfabs (1022 cm−2)	none	${2.1}_{-0.7}^{+2.2}$
fpcfabs	none	${0.82}_{-0.52}^{+0.16}$
kTBB (keV)	${1.43}_{-0.15}^{+0.19}$	none
RBB (m)	280 ± 70	none
Γa	${0.78}_{-0.21}^{+0.18}$	${0.98}_{-0.08}^{+0.09}$
Ecut (keV)	9.7 fixed	5.5 fixed
Efold (keV)	14.3 fixed	15.3 fixed
Eline (keV)	6.44 fixed	6.45 fixed
σline (keV)	0 fixed	0 fixed
Fluxline (ph cm−2 s−1)	3.2 × 10−5 fixed	3.2 × 10−5 fixed
EWlineb (eV)	<120	<120
Fluxc (erg cm−2 s−1)	(5.8 ± 0.6) × 10−11	(6.1 ± 0.2) × 10−11
Luminosityc (1035 erg s−1)	3.4 ± 0.4	3.6 ± 0.1
${\chi }_{\nu }^{2}$/dof	0.911/267	0.911/267

Notes.

^aPower-law photon index. ^bEquivalent width. ^cFluxes (corrected for the absorption) and luminosities are in the energy range 0.1–100 keV. The flux reported for Model 2 is corrected for both absorbing components.

Download table as:  ASCIITypeset image

3.2.4. Spin-phase Resolved Spectroscopy

A spin-phase resolved spectroscopy was performed on strictly simultaneous XMM-Newton and NuSTAR data. From the hardness ratios reported in Figure 3, a spin-phase resolved spectral extraction driven by the source hardness would suggest different spectral extractions, depending on the energy bands adopted. So, we simply divided the pulse period into five equally spaced intervals, and extracted five broadband (0.4–78 keV) spectra. Note that we also tried narrower or broader spin-phase intervals, but we found that five is the best compromise to map eventual spectral variability, given the available count statistics.

The same models adopted before (Model 1 and Model 2) were applied to the five spectra, resulting in the spectral parameters reported in Table 6 and summarized in Figure 9. The evolution of the intensity along the pulse profile is dominated by the variability of the power-law flux (2–10 keV), while the spectral variability seems to be mainly due to the power-law slope, which is the hardest in the first spin-phase interval (Δϕ = 0.0–0.2), although the significant variability of the power law cutoff energy and of the blackbody parameters (in Model 1) along the spin phase did not allow us a straightforward interpretation.

Zoom In Zoom Out Reset image size

Table 6.  Broadband (0.4–78 keV) Spin-phase-resolved Spectroscopy^a

Model 1 Parameters	Δϕ = 0.0–0.2	0.2–0.4	0.4–0.6	0.6–0.8	0.8–1.0
NH (1022 cm−2)	${0.91}_{-0.14}^{+0.20}$	1.0 ± 0.16	${1.01}_{-0.18}^{+0.13}$	${1.04}_{-0.17}^{+0.16}$	${1.20}_{-0.21}^{+0.23}$
kTBB (keV)	${1.39}_{-0.10}^{+0.08}$	${1.73}_{-0.12}^{+0.28}$	${1.92}_{-0.16}^{+0.09}$	${1.65}_{-0.10}^{+0.12}$	${1.78}_{-0.09}^{+0.20}$
RBB (m)	${630}_{-80}^{+60}$	${490}_{-60}^{+75}$	${400}_{-30}^{+50}$	${430}_{-50}^{+60}$	${430}_{-50}^{+70}$
Γ	$-{0.16}_{-0.57}^{+0.38}$	${0.60}_{-0.19}^{+0.36}$	${0.96}_{-0.21}^{+0.09}$	${0.85}_{-0.17}^{+0.14}$	${0.75}_{-0.24}^{+0.24}$
Ecut (keV)	${7.1}_{-0.9}^{+0.8}$	${9}_{-9}^{+7}$	${21}_{-5}^{+2}$	${13.0}_{-1.7}^{+1.7}$	${10}_{-2}^{+6}$
Efold (keV)	${9.5}_{-2.1}^{+2.4}$	${16}_{-3}^{+8}$	${12}_{-3}^{+6}$	${15}_{-3}^{+3}$	${21}_{-7}^{+20}$
Pow Fluxb (10−11 erg cm−2 s−1)	3.6 ± 0.7	4.4 ± 0.4	2.62 ± 0.15	${2.51}_{-0.29}^{+0.15}$	${2.9}_{-0.3}^{+0.3}$
${\chi }_{\nu }^{2}$/dof	0.988/782	1.014/943	1.011/749	1.043/655	0.997/752
Model 2 Parameters	Δϕ = 0.0–0.2	0.2–0.4	0.4–0.6	0.6–0.8	0.8–1.0
NH (1022 cm−2)	${0.9}_{-0.15}^{+0.41}$	1.38 ± 0.18	${1.26}_{-0.15}^{+0.14}$	${1.45}_{-0.21}^{+0.18}$	${1.50}_{-0.20}^{+0.22}$
NHpcfabs (1022 cm−2)	${3.2}_{-0.19}^{+1.0}$	${7.6}_{-2.6}^{+2.7}$	${7.9}_{-2.1}^{+2.2}$	${8.6}_{-3.0}^{+2.7}$	${6.8}_{-2.1}^{+2.7}$
fpcfabs	${0.77}_{-0.11}^{+0.01}$	${0.54}_{-0.09}^{+0.07}$	${0.57}_{-0.08}^{+0.07}$	${0.56}_{-0.17}^{+0.09}$	${0.58}_{-0.08}^{+0.07}$
Γ	${1.00}_{-0.04}^{+0.09}$	${1.13}_{-0.18}^{+0.17}$	${1.22}_{-0.15}^{+0.15}$	${1.35}_{-0.37}^{+0.21}$	${1.18}_{-0.15}^{+0.17}$
Ecut (keV)	${6.5}_{-0.6}^{+0.7}$	${6.4}_{-0.9}^{+1.1}$	${6.5}_{-0.8}^{+0.8}$	${6.4}_{-6.4}^{+1.6}$	${6.4}_{-0.72}^{+0.89}$
Efold (keV)	${19.8}_{-1.2}^{+2.6}$	${18}_{-3}^{+5}$	${17}_{-3}^{+4}$	${20}_{-6}^{+7}$	${18}_{-3}^{+6}$
Pow Fluxb(10−11 erg cm−2 s−1)	${7.0}_{-0.13}^{+0.16}$	${9.8}_{-0.4}^{+0.5}$	${7.5}_{-0.32}^{+0.38}$	${6.1}_{-0.33}^{+0.39}$	${7.3}_{-0.3}^{+0.4}$
${\chi }_{\nu }^{2}$/dof	0.984/782	1.027/943	1.013/749	1.046/655	0.987/752

Notes.

^aXMM-Newton and NuSTAR strictly simultaneous spectra. Model 1 is a double-component continuum made of a blackbody plus a PLCUT model. Model 2 is a PLCUT model modified by a partial covering fraction absorption pcfabs. ^bPower-law flux (background corrected) in the energy band 2–10 keV.

Download table as:  ASCIITypeset image

When adopting Model 2, E_cut and E_fold parameters remained constant along the pulse profile. For this reason, we performed a second fit, fixing them to their mean value. We show in Figure 10 the resulting confidence contour levels of the photon index and the additional absorbing column density partially covering the power law. The numbers indicate the five spin-phase-intervals. In these fits, the covering fractions in the five spectra were consistent with each other (within 90% uncertainty), in the range 0.52–0.67. Again, spin-phase interval ϕ = 0.0–0.2 showed a harder power-law emission.

Zoom In Zoom Out Reset image size

No significant negative residuals around 16–18 keV were found during the spin-phase resolved spectroscopy.

3.2.5. Hardness-ratio-selected Spectroscopy

We investigated the spectral variability along the observation plotting different hardness ratios versus time. Whatever the choice of hardness ratio (in both XMM-Newton and NuSTAR data), some level of scatter around a mean value was always present, but with no clear trend suggesting a specific and meaningful hardness-ratio-selected spectroscopy. The only exception was in the case of a hardness ratio of 4–12 keV to 1–4 keV count rates (Figure 11): a hard peak lasting ∼500 s in the hardness ratio was found after 10⁴ s from the start of the XMM-Newton observation. We extracted an XMM-Newton spectrum from this time interval ("hard dip," hereafter), as it was not simultaneously covered by NuSTAR. Meaningful spectroscopy was possible in the energy range 1–10 keV. A single absorbed power-law model resulted in an absorbing column density N_H = (3.2 ± 0.6) × 10²² cm⁻², ${\rm{\Gamma }}={0.8}_{-0.15}^{+0.16}$, and a flux corrected for the absorption F = 3.9 × 10⁻¹¹ erg cm⁻² s⁻¹ (1–10 keV). Adopting more complicated models (like Model 1 and Model 2) is meaningless, since several spectral parameters are unconstrained.

Zoom In Zoom Out Reset image size

After ten years since the discovery of SFXTs, huge progress has been made in understanding their behavior.

At first, it was proposed that Bondi–Hoyle direct accretion from dense clumps in the supergiant wind could explain the sporadic flaring activity in SFXTs (in't Zand 2005). A comparison of the clump properties implied by the SFXT X-ray light curves with what was predicted by the theory of line-driven hot stellar winds indicates that direct Bondi–Hoyle accretion from clumpy winds cannot reproduce the X-ray data (see Martínez-Núñez et al. 2017 and references therein for a comprehensive review). One of the difficulties of the clumpy wind scenario is that the accreted mass derived from the SFXT X-ray flares (10¹⁹–10²² g) is typically larger than what is expected from a single clump in stellar winds (∼10¹⁸ g). The IGR J11215–5952 flare peak luminosity implies an accreted mass of ∼10¹⁹ g onto the NS, confirming once more this discrepancy. Note that this mass should be considered as a lower limit to the clump mass, since a wind blob, if it is larger than the accretion radius, might be only partially accreted by the pulsar.

Currently, two physical mechanisms might explain the SFXT flares in a 187 s pulsar like IGR J11215–5952: either a centrifugal barrier (Grebenev & Sunyaev 2007; Bozzo et al. 2008), or a quasi-spherical settling accretion regime (Shakura et al. 2012, 2014). In the first case, rescaling the Grebenev & Sunyaev (2007) equations to IGR J11215–5952 (spin period of 187 s and orbital period of 165 days) we derived a magnetic field of 1.2 × 10¹³ G needed to have an operating centrifugal barrier (although we note that the equations reported by Grebenev & Sunyaev (2007) assume a circular orbit). If we adopt the average outburst X-ray luminosity of 10³⁶ erg s⁻¹ instead of the wind parameters in this scenario, we found an even higher estimate for the magnetic field of ∼5 × 10¹³ G needed for the centrifugal barrier to operate. Note that also the source distance of 7 kpc is not accurate, but it should be considered as a lower limit (Lorenzo et al. 2014). In the second case, the settling accretion scenario sets in in slow pulsars (spin period >70 s) when the X-ray luminosity is below 4 × 10³⁶ erg s⁻¹ (Shakura et al. 2012). Below this threshold the matter captured inside the Bondi radius is not able to efficiently cool down by Compton processes, such that it cannot penetrate the NS magnetosphere by means of Rayleigh–Taylor instabilities. Hot matter accumulates above the NS magnetosphere forming a quasi-static shell that cools down only by inefficient radiative cooling. Shakura et al. (2014) suggested that the bright SFXT flares are produced by the collapse onto the NS of this hot shell, triggered by reconnection events between the magnetic field carried by the accreting matter and the neutron star magnetosphere.

Quasi-periodic flaring activity that has often been observed to punctuate the SFXT outbursts (e.g., SAX J1818.6-1703, Boon et al. 2016) would be produced when the shell, after a collapse, is replenished by new wind capture; so the flares repeat on a regular timescale, as long as the mass accretion into the magnetosphere is allowed. The energy normally released in an SFXT bright flare (∼10³⁹ erg) is in agreement with estimated mass of the hot shell (Shakura et al. 2014), which is also consistent with the accreted mass we have derived here from the luminosity of the flares in IGR J11215–5952 (∼10¹⁹ g).

Our XMM-Newton and NuSTAR observations of IGR J11215–5952 allowed us also to observe, for the first time in this source, X-ray flares repeating every ∼2–2.5 ks—a property that in the settling accretion scenario is explained in a natural way. Note that the IGR J11215–5952 X-ray luminosity during outburst is fully compatible with the luminosity threshold predicted by Shakura et al. (2012) for the onset of the settling accretion. All these observational facts strongly favor an interpretation of the IGR J11215–5952 SFXT properties being within the quasi-spherical settling accretion model.

Finally, we note that if the hint of a cyclotron line at 17 keV is confirmed by future observations (implying a NS surface magnetic field of 2 × 10¹² G) it will reinforce the interpretation of the SFXT behavior falling within the settling accretion scenario.