EDP Sciences
Free Access
Issue
A&A
Volume 617, September 2018
Article Number L8
Number of page(s) 5
Section Letters to the Editor
DOI https://doi.org/10.1051/0004-6361/201834160
Published online 01 October 2018

© ESO 2018

1. Introduction

Accreting fast-rotating neutron stars (NS) in low-mass X-ray binaries have been investigated for almost two decades. This class of objects, also known as accreting millisecond X-ray pulsars (AMXPs), currently includes 21 sources with spin periods ranging between 1.7 ms and 9.5 ms (see Burderi & Di Salvo 2013; Patruno & Watts 2012; Campana & Di Salvo 2018 for extensive reviews). The characteristic short spin periods observed in AMXPs are the result of long-lasting mass transfer from an evolved sub-solar companion star via Roche-lobe overflow onto a slow-rotating NS (recycling scenario; Alpar et al. 1982), making them the progenitors of rotation-powered millisecond pulsars that emit from the radio to the gamma-ray band. Almost one-third of the AMXPs are ultra-compact binary systems (Porb < 1 h), and the remainder show short orbital periods on average (Porb < 12 h), except for the intermittent pulsar Aql X–1 (Casella et al. 2008), whose period is Porb∼18 h (Welsh et al. 2000). Short orbital periods suggest small, low-mass companion stars, consistent with donor masses of <0.2 M on average.

Here, we report on the detection of millisecond X-ray pulsations from IGR J17591–2342, an X-ray transient that was first detected by the INTernational Gamma-Ray Astrophysics Laboratory (INTEGRAL) during a scan of the Galactic centre on 2018 August 10 (Ducci et al. 2018). An archival search in the Neil Gehrels Swift Observatory (Swift) Burst Alert Telescope data revealed the source to be active since 2018 July 22 (Krimm et al. 2018). Pointed Swift/XRT observations of the source region on 2018 August 12 revealed a point-like X-ray source and determined its precise X-ray position (Bozzo et al. 2018), which was refined by a Chandra observation on 2018 August 23 to be RA = 17h59m02.83s Dec = −23° 43′ 10.2″ (J2000) with an astrometric uncertainty of 0.6″ (Nowak et al. 2018). On 2018 August 14, the Australia Telescope Compact Array (ATCA) detected IGR J17591–2342 with a flux density Sν = 1.09 ± 0.02 mJy and Sν = 1.14 ± 0.02 mJy (at 68% c.l.) at 5.5 GHz and 9.0 GHz, respectively, at a position RA = 17h59m 02.83s ± 0.04s Dec = −23° 43′ 08.3″ ± 0.1″. This is consistent with the X-ray determinations (Russell et al. 2018).

In this Letter, we describe a coherent timing analysis of the Nuclear Spectroscopic Telescope Array (NuSTAR) and Neutron Star Interior Composition Explorer (NICER) observations that provided the pulsar spin period and binary ephemeris. We also analyse the X-ray spectral modelling obtained from NuSTAR, Swift, INTEGRAL, and NICER data.

2. Observations and data reduction

IGR J17591–2342 was observed by IBIS/ISGRI on board the INTEGRAL satellite from 2018 August 10 at 15:50 to August 11 at 12:30 UT for a total exposure time of 66 ks. We performed the data reduction and analysis using the off-line science analysis (OSA) software provided by the INTEGRAL Science Data Centre. We produced a mosaic image from the combined individual images of the available data set, where the source was detected with a significance >7σ in the 20–80 keV energy band. We extracted the average IBIS/ISGRI spectrum in 8 bins from 25 to 150 keV with equal logarithmic spacing. No JEM-X data were available because the source is located at the edge of its field of view.

NuSTAR observed IGR J17591–2342 (Obs.ID. 90401331002) on 2018 August 13 starting from 22:36 UT for a total exposure of ∼30 ks. We performed standard screening and filtering of the events using the NuSTAR data analysis software (NUSTARDAS) from HEASOFT version 6.24. We extracted source events from a circular region of radius 80″ centred on the source position. We extracted the background with a similar extraction area, but located in a region far from the source with similar degree of straylight contamination. The average source count rate per instrument is ∼9 counts s−1 in the energy range 3–80 keV. We corrected for spacecraft clock drift by applying the latest clock correction file (version 84, valid up to 2018-08-14).

Swift observed IGR J17591–2342 (Obs.ID. 00010804002) on 2018 August 14 from 00:38 UT for a total of ∼0.6 ks with Swift/XRT operated in photon-counting mode. We reduced and processed the XRT data with XRTPIPELINE version 0.13.4, extracting source events from a circular region of radius 64″. We estimated a source count rate of ∼1 counts s−1 above the pile-up threshold (see, e.g. Romano et al. 2006). Thus, we extracted the source spectrum using an annular region centred at the source position with inner and outer radius of 10″ and 64″, respectively. Similarly, we extracted the background spectrum from an annular region with inner and outer radius of 102″ and 230″, respectively.

NICER observed IGR J17591–2342 on 2018 August 15 from 00:00 to 14:08 UT for a total exposure of ∼7.3 ks (Obs.ID. 1200310101, hereafter NICER-1), on August 18 from 02:08 to 03:56 UT (exposure of ∼1.5 ks, Obs.ID. 1200310102, NICER-2), on August 23 from 00:58 to 22:57 UT (exposure of ∼11.5 ks, Obs.ID. 1200310103, NICER-3), and on August 24 from 01:51 to 14:44 UT (exposure of ∼7 ks, Obs.ID. 1200310104, NICER-4). We filtered events in the 1–12 keV band applying standard screening criteria with NICERDAS version 4.0. We removed short intervals that showed background flaring in high geomagnetic latitude regions. The spectral background was obtained from blank-sky exposures.

We did not detect any Type-I thermonuclear burst during the observations analysed in this work.

Table 1.

Orbital parameters and spin frequency of IGR J17591–2342 with uncertainties on the last digit quoted at 1σ confidence level.

3. Results

3.1. Timing analysis

To search for coherent signals, we first corrected the NICER and NuSTAR photon arrival times to the solar system barycentre with the tool BARYCORR, using the DE-405 solar planetary ephemeris, and adopting the source coordinates of Table 1. Power density spectra (PDS) obtained by averaging either NuSTAR or NICER 150 s data segments showed a prominent (>4σ) excess at a frequency of 527.3 Hz. The feature exhibited a double-peaked structure characteristic of orbital Doppler broadening. Inspection of the NuSTAR PDS over short segments (500 s) revealed pulse frequency modulation typical for pulsars in binary systems (see Fig. 1). The best fit is found for orbital period Porb = 31694(67) s, projected semi-major axis of the NS orbit x = 1.233(9) lt-s, time of passage through the ascending node TNOD = 58343.704(1) (MJD), and spin frequency ν0 = 527.4253(6) Hz.

Starting from this provisional solution, we folded data segments of ∼400 s and ∼300 s for the NuSTAR and NICER observations, respectively, into eight phase bins at the preliminary spin frequency ν0. We modelled each pulse profile with a constant plus two sinusoidal functions, representing the pulse frequency and its second harmonic. We retained only profiles for which the ratio between the pulse frequency amplitude and its 1σ uncertainty was higher than three. We modelled the pulse phase with a linearly varying frequency plus a circular Keplerian orbital model separately for NuSTAR and NICER. A more detailed description of the procedure can be found in Sanna et al. (2016, and references therein).

The best-fit orbital and pulsar spin parameters for the NuSTAR and NICER data sets are shown in the first and second column of Table 1, respectively. They are derived from the phase delays at the fundamental frequency; a timing analysis using the second harmonic yields compatible results. The top panel of Fig. 2 shows the NuSTAR and NICER pulse phases after correcting for the best orbital solution and assuming a constant frequency. The NuSTAR pulse phases show a large distribution of residuals, which is not compatible with the binary orbital modulation. These can be well described (see residuals in the bottom left panel of Fig. 2) by a strong spin frequency derivative (see Table 1) and an additional sinusoidal term with period P ∼ 193 minutes (close to twice the NuSTAR orbital period). We note that no sinusoidal modulation is present in either NICER data (both harmonics) or the NuSTAR second harmonic pulse phases, suggesting a likely instrumental origin. For the latter, we emphasise that the phase variations of ∼0.5 observed for the fundamental component are close to a complete cycle of the harmonic, tuning down any effect of the phase drift. We also note that the lower statistics in the harmonic would make residuals less evident with respect to the fundamental component.

thumbnail Fig. 1.

Time evolution of the pulsar frequency shifts (with respect to ν = 527.432 Hz) estimated from 500 s NuSTAR data segments. The frequency shifts show a clear modulation that is compatible with Doppler shifts caused by orbital motion in a binary system. The red dashed line represents the best-fitting orbital Doppler modulation model assuming a circular orbit.

Open with DEXTER

The NICER pulse phase residuals (top right panel) show a hint of a positive spin frequency derivative (see Table 1 for the corresponding value and the bottom right panel of Fig. 2 for final residuals that account for this component). We improved the orbital parameter estimates (third column of Table 1) by orbitally phase-connecting the available data sets (see, e.g. Sanna et al. 2017a,c, for more details on the procedure). We note that the uncertainty on the source localisation has an effect on the determination of the frequency and frequency derivative, which is more than one order of magnitude smaller than the uncertainties reported in Table 1 for the corresponding parameters. Therefore we ignore this effect.

Finally, we created pulse profiles by epoch-folding the NuSTAR (3–80 keV) and NICER (1–12 keV) observations with the parameters reported in Table 1. We modelled them with a combination of three sinusoidal components (see Fig. 3). In NuSTAR data, the fundamental, second, and third harmonics have fractional amplitudes of (16.1 ± 0.3)%, (5.1 ± 0.3)%, and (1.5 ± 0.3)%, respectively. NICER pulse profiles are characterised by fractional amplitudes of (13.7 ± 0.2)%, (5.0 ± 0.2)%, and (1.0 ± 0.2)% for the fundamental, second, and third harmonics, respectively.

thumbnail Fig. 2.

Top panel: pulse phase delays as a function of time computed by epoch-folding at a constant frequency ∼400 s and 300 s data intervals of the NuSTAR (left) and the NICER (right) data sets, respectively, after correcting for the best orbital solution. In the same panel we show the best-fit models for each group of data. Bottom panel: residuals in μs with respect to the best-fitting models for the pulse phase delays.

Open with DEXTER

thumbnail Fig. 3.

IGR J17591–2342 pulse profiles (black points) from epoch-folding the NuSTAR (top panel) and the NICER (bottom panel) observations after correcting for the orbital parameters reported in Table 1. The best-fitting model (red line) is the superposition of three sinusoidal functions with harmonically related periods. For clarity, we show two cycles of the pulse profile.

Open with DEXTER

3.2. Spectral analysis

We performed a spectral analysis with Xspec 12.10.0c (Arnaud 1996) after applying an optimal binning (Kaastra & Bleeker 2016). For the first epoch, we used the 0.5–7.5 keV range for Swift/XRT, 3–70 keV for NuSTAR and 30–80 keV for IBIS/ISGRI and grouped the spectra to collect at least 20 counts per bin (Fig. 4). The broad-band (0.5–80 keV) spectrum is well described by an absorbed thermal component (black-body) with temperature kTBB that also provides the seed photons for a thermally Comptonised continuum with electron energy kTe and a weak emission line centred on the iron Kα complex and width fixed to zero. In Xspec, this is implemented as TBabs*[cflux*nthcomp+bbodyrad]+gauss. For the modelling, we left the flux of the Comptonisation component free to vary independently for each instrument to account for cross-instrument calibration offsets and possible source variations that are due to not strictly simultaneous observations.

thumbnail Fig. 4.

Upper panel: Swift (blue), NuSTAR FPMA and FPMB (black and red, respectively) and IBIS/ISGRI (green) unfolded spectra of IGR J17591–2342. The dashed and dash-dotted lines represent black-body and Comptonisation components, respectively. The sharp dotted line is a Gaussian with width set to zero. Additional rebinning is for displaying purposes. Lower panel: residuals from the best-fit model in units of standard deviations.

Open with DEXTER

The NICER observations were fitted independently by minimising the pgstat statistic in Xspec1. Because of uncertainties in the low-energy response, we ignored the NICER data below 1.4 keV. We adopted the same model, but fixed the electron temperature at the value found from NuSTAR and checked that the result remained insensitive as long as kTe > 12 keV. After verifying the compatibility of single spectra, we performed a joint fit of NICER-1 and -2 observations, leaving only the flux to be independently determined between them. We applied the same procedure to NICER-3 and -4. The best-fit parameters are shown in Table 2 with uncertainties at the 90% confidence level. Regardless of the statistical test used to determine the best-fit parameters, we report the value.

Table 2.

Best-fit spectral parameters of IGR J17591–2342 with uncertainties at the 90% confidence level.

4. Discussion

We reported the newly discovered AMXP IGR J17591–2342, for which we detected coherent X-ray pulsations at ∼527 Hz in the NuSTAR and NICER observations performed almost 25 days from the beginning of the outburst, with a pulse fraction of 15%. We modelled the NS spin frequency drift as the Doppler shift induced by the binary orbital motion, discovering the binary nature of the system, which is characterised by an orbital period of almost 8.8 h, very similar to the intermittent AMXP SAX J1748.9–2021 (see, e.g. Altamirano et al. 2008; Sanna et al. 2016) and the eclipsing AMXP SWIFT J1749.4–2807 (see, e.g. Markwardt & Strohmayer 2010; Altamirano et al. 2011; Ferrigno et al. 2011).

The analysis of the NuSTAR pulse phase residuals revealed a large spin-up derivative (2.6 × 10−10 Hz s−1) and an oscillation close to twice the satellite orbital period. We suggest that both effects are associated with the time drift of the internal clock instrument (see, e.g. Madsen et al. 2015). Similar spurious frequency derivatives have been reported for NuSTAR observations of AMXPs such as MAXI J0911–655 (Sanna et al. 2017b), SAX J1808.4–3658 (Sanna et al. 2017a), and IGR J00291+5934 (Sanna et al. 2017c). The large discrepancy on the spin frequency ((∼1 × 10−4 Hz) between NuSTAR and NICER is also a direct consequence of the NuSTAR clock issue. These considerations are reinforced by the analysis of NICER observations, which do not show any significant frequency drift.

A spin-up frequency derivative of (2.0 ± 1.6) × 10−13 Hz/s is observed from the phase-coherent timing analysis of the NICER observations. For a broad-band (0.1–100 keV) absorbed flux of ∼7 × 10−10 erg/s/cm2 and a source distance of 8.5 kpc (assumed near the Galactic centre, see, e.g. Kerr & Lynden-Bell 1986), we estimate an accretion rate of ≃ 5.2 × 10−10 M/yr (for an NS radius and mass of 1.4 M and 10 km). Assuming the accretion disc to be truncated at the co-rotation radius, the observed mass accretion rate yields a maximum spin-up derivative of a few 10−13 Hz s−1, fully consistent with the observed one.

A rough estimate of the NS dipolar magnetic field can be obtained by assuming the condition of spin equilibrium for the X-ray pulsar. The magnetic field can then be estimated as

(1) where ζ represents a model-dependent dimensionless factor (between 0.1 and 1) corresponding to the ratio between the magnetospheric radius and the Alfvén radius (see, e.g. Ghosh & Lamb 1979; Wang 1996), Pspin is the pulsar spin period in ms, and M is the NS mass. Assuming a 1.4 M NS and the value of reported above, we obtain a range for the dipolar magnetic field of 1.4 × 108 < B < 8 × 109 G, consistent with the average magnetic field of known AMXPs (see, e.g. Mukherjee et al. 2015).

With the NS mass function f(m2, m1, i) ∼ 1.5 × 10−2 M, we can constrain the mass of the companion star. Since neither total eclipses nor dips have been observed in the light curve, we can assume a binary inclination lower than 60° (see, e.g. Frank et al. 2002). As shown in Fig. 5, the upper limit on the inclination angle (represented in blue) allows us to limit the companion star mass m2 ≳ 0.42 M (for a 1.4 M NS), which increases up to m2 ≳ 0.52 M if we consider a 2 M NS. Introducing the contact condition (R2RL2) required to activate Roche-lobe overflow, we can express the donor radius as a function of its mass as , where m2 is the companion mass in units of M and Porb,9 h represents the binary orbital period in units of 9 h. In Fig. 5 we report the companion mass-radius relation (black solid line) assuming a 1.4 M NS. For comparison, we show numerically simulated mass-radius relations for zero-age main-sequence stars (ZAMS; purple crosses; see, e.g. Tout et al. 1996), as well as isochrones for stars of 8 (red diamonds) and 12 Gyr (blue circles; see, e.g. Girardi et al. 2000). From the intersections with the mass-radius companion curve, we can infer that the donor is compatible with either a ZAMS with mass ∼1.1 M (corresponding to an inclination angle of i ∼ 24°) or an old main-sequence star with mass 0.85–0.92 M (i ranging between 28 and 30°) for a stellar age between 8 and 12 Gyr. We note, however, that the a priori probability of observing a binary system with inclination i ≤ 30° is of the order of 13%. Mass values for which the main-sequence radius is smaller than the companion Roche lobe could still be acceptable if we consider the possibility of a bloated donor star. In that case, the thermal timescale () should be much longer than the evolutionary timescale (M2/2).

thumbnail Fig. 5.

Radius-mass plane showing the size constraints on the companion star Roche-lobe of IGR J17591–2342 (black solid line) obtained from the orbital parameters of the neutron star. The hatched region represents the constraints on the companion mass from the binary mass function, and the blue area defines the mass constraints for inclination angles between 60 and 90°. The other curves represent theoretical mass-radius relations for zero-age main-sequence stars (purple crosses) and isochrones for stars of 8 (red diamonds) and 12 Gyr (blue circles). The top x-axes represent the corresponding binary inclination angle in degrees assuming a 1.4 M.

Open with DEXTER

Finally, the broad-band energy spectrum of IGR J17591–2342 is well described by an absorbed soft black-body-like component (kT ∼ 0.8 keV) with a relatively small emitting area that is compatible with emission from the neutron star surface (or part of it) plus a Comptonised component (Γ ∼ 1.8) with a seed photon temperature compatible with the soft thermal component. The spectral properties of the source are consistent with those of other AMXPs observed in the hard state (see, e.g. Falanga et al. 2005; Gierliński & Poutanen 2005; Papitto et al. 2009, 2013; Sanna et al. 2017b,c). We found marginal evidence (Δχ2 = 21 for 2 d.o.f.) of a weak emission line that is compatible with the iron K-α transition. Even if marginally significant (an F-test gives a 4σ line significance), its introduction removes positive residuals around the expected energy, and such lines are not unusual for this type of source (see, e.g. Papitto et al. 2009, 2013; Sanna et al. 2017b,c). Table 2 shows that at later times, when the source re-brightens, the additional black body seems to disappear but the asymptotic Comptonisation power-law index increases. This might be due to more effective cooling of the seed photons by a thicker accretion stream above the stellar surface.

The discovery of another transient by INTEGRAL and its characterisation as the 22. accreting millisecond pulsar by NuSTAR and NICER enriches the census of these key objects in the understanding of the late stages of stellar evolution.


1

pgstat is suited for Poisson distributed data with Gaussian distributed background; see https://heasarc.gsfc.nasa.gov/xanadu/xspec/manual/XSappendixStatistics.html.

Acknowledgments

We thank Fiona Harrison for accepting a NuSTAR observation in the Director Discretionary Time. We thank the Swift, NuSTAR, and NICER teams for scheduling and performing these target-of-opportunity observations on short notice. This work was supported by NASA through the NICER and NuSTAR missions and the Astrophysics Explorers Program. It made use of data and software provided by the High Energy Astrophysics Science Archive Research Center (HEASARC). It is also based on observations with INTEGRAL, an ESA project with instruments and science data centre funded by ESA member states (especially the PI countries: Denmark, France, Germany, Italy, Switzerland, and Spain) and with the participation of Russia and the USA. DA acknowledges support from the Royal Society. LD acknowledges grant FKZ 50 OG 1602.

References

All Tables

Table 1.

Orbital parameters and spin frequency of IGR J17591–2342 with uncertainties on the last digit quoted at 1σ confidence level.

Table 2.

Best-fit spectral parameters of IGR J17591–2342 with uncertainties at the 90% confidence level.

All Figures

thumbnail Fig. 1.

Time evolution of the pulsar frequency shifts (with respect to ν = 527.432 Hz) estimated from 500 s NuSTAR data segments. The frequency shifts show a clear modulation that is compatible with Doppler shifts caused by orbital motion in a binary system. The red dashed line represents the best-fitting orbital Doppler modulation model assuming a circular orbit.

Open with DEXTER
In the text
thumbnail Fig. 2.

Top panel: pulse phase delays as a function of time computed by epoch-folding at a constant frequency ∼400 s and 300 s data intervals of the NuSTAR (left) and the NICER (right) data sets, respectively, after correcting for the best orbital solution. In the same panel we show the best-fit models for each group of data. Bottom panel: residuals in μs with respect to the best-fitting models for the pulse phase delays.

Open with DEXTER
In the text
thumbnail Fig. 3.

IGR J17591–2342 pulse profiles (black points) from epoch-folding the NuSTAR (top panel) and the NICER (bottom panel) observations after correcting for the orbital parameters reported in Table 1. The best-fitting model (red line) is the superposition of three sinusoidal functions with harmonically related periods. For clarity, we show two cycles of the pulse profile.

Open with DEXTER
In the text
thumbnail Fig. 4.

Upper panel: Swift (blue), NuSTAR FPMA and FPMB (black and red, respectively) and IBIS/ISGRI (green) unfolded spectra of IGR J17591–2342. The dashed and dash-dotted lines represent black-body and Comptonisation components, respectively. The sharp dotted line is a Gaussian with width set to zero. Additional rebinning is for displaying purposes. Lower panel: residuals from the best-fit model in units of standard deviations.

Open with DEXTER
In the text
thumbnail Fig. 5.

Radius-mass plane showing the size constraints on the companion star Roche-lobe of IGR J17591–2342 (black solid line) obtained from the orbital parameters of the neutron star. The hatched region represents the constraints on the companion mass from the binary mass function, and the blue area defines the mass constraints for inclination angles between 60 and 90°. The other curves represent theoretical mass-radius relations for zero-age main-sequence stars (purple crosses) and isochrones for stars of 8 (red diamonds) and 12 Gyr (blue circles). The top x-axes represent the corresponding binary inclination angle in degrees assuming a 1.4 M.

Open with DEXTER
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.