Discovery of thermonuclear (Type I) X-ray bursts in the X-ray binary Swift J1858.6-0814 observed with NICER and NuSTAR
D. J. K. Buisson, D. Altamirano, P. Bult, G. C. Mancuso, T. Güver, G. K. Jaisawal, J. Hare, A. C. Albayati, Z. Arzoumanian, N. Castro Segura, D. Chakrabarty, P. Gandhi, S. Guillot, J. Homan, K. C. Gendreau, J. Jiang, C. Malacaria, J. M. Miller, M. Özbey Arabacı, R. Remillard, T. E. Strohmayer, F. Tombesi, J. A. Tomsick, F. M. Vincentelli, D. J. Walton
MMNRAS , 1–14 (2020) Preprint 9 September 2020 Compiled using MNRAS L A TEX style file v3.0
Discovery of thermonuclear (Type I) X-ray bursts in theX-ray binary Swift J1858.6–0814 observed with
NICER and
NuSTAR
D. J. K. Buisson , (cid:63) D. Altamirano , P. Bult , , G. C. Mancuso , , T. G¨uver , ,G. K. Jaisawal , J. Hare † , A. C. Albayati , Z. Arzoumanian , N. Castro Segura ,D. Chakrabarty , P. Gandhi , S. Guillot , J. Homan , , K. C. Gendreau ,J. Jiang , , C. Malacaria , ‡ , J. M. Miller , M. ¨Ozbey Arabacı , ,R. Remillard , T. E. Strohmayer , F. Tombesi , , , , J. A. Tomsick ,F. M. Vincentelli and D. J. Walton Affiliations at end.
Accepted 2020 September 7. Received 2020 September 7 in original form 2020 June 12
ABSTRACT
Swift J1858.6–0814 is a recently discovered X-ray binary notable for extremely strongvariability (by factors >
100 in soft X-rays) in its discovery state. We present the de-tection of five thermonuclear (Type I) X-ray bursts from Swift J1858.6–0814, implyingthat the compact object in the system is a neutron star. Some of the bursts show pho-tospheric radius expansion, so their peak flux can be used to estimate the distance tothe system. The peak luminosity, and hence distance, can depend on several systemparameters; for the most likely values, a high inclination and a helium atmosphere, D = 12 . +0 . − . kpc, although systematic effects allow a conservative range of 9-18 kpc.Before one burst, we detect a QPO at 9 . ± . . ± .
2% (0 . −
10 keV), likely due to marginally stable burning of helium; sim-ilar oscillations may be present before the other bursts but the light curves are notlong enough to allow their detection. We also search for burst oscillations but do notdetect any, with an upper limit in the best case of 15% fractional amplitude (over1 − Key words: accretion, accretion discs – stars: neutron – X-rays: binaries – X-rays:bursts
A key aspect of accreting systems is the object onto whichthe accretion is occurring; in X-ray binaries (XRBs) this iseither a neutron star (NS) or black hole (BH). Many ob-servable properties are similar in either case, so determiningwhich is present is often a challenging task.There are several properties which can divide NSs andBHs as populations and some features which empirically ap-pear to occur in only one type of system. Firstly, outbursts of (cid:63)
Email: [email protected] † NASA Postdoctoral Fellow ‡ NASA Postdoctoral Fellow the different classes of source follow different tracks in grossproperties such as the hardness-intensity or colour-colour di-agrams (e.g. van der Klis 2006). However, this requires mon-itoring of the full outburst and some sources do not followthe typical patterns. Additionally, quasi-periodic oscillations(QPOs) are only found at kHz frequencies in neutron starsystems (van der Klis et al. 1996; Strohmayer et al. 1996),although there is not yet a universally accepted model fortheir production (e.g. review by van der Klis 2006).Also, BH and NS systems can be separated in theradio/X-ray luminosity plane (while in the hard state), withBH systems being radio brighter (Migliari & Fender 2006;Gallo et al. 2018). Similarly, the hard Comptonised com- c (cid:13) a r X i v : . [ a s t r o - ph . H E ] S e p D. J. K. Buisson et al. ponent tends to have a higher temperature in BH systems(Burke et al. 2017). However, the loci of BHs and NSs over-lap in these properties, so they cannot be used to determinethe accretor definitively in an individual source, particularlywhere a source shows unusual properties.Other properties of an accreting system can give adefinitive determination of whether the accreting object isa black hole or neutron star. To confirm a black hole accre-tor requires a dynamical mass measurement which is greaterthan possible neutron star masses (e.g. Webster & Murdin1972; Bolton 1972; Orosz & Bailyn 1997), since there areno particular accretion properties which are unique to blackholes. Conversely, there are several properties which are con-firmed as unique to neutron stars, since the neutron starsurface can provide an additional location for emission com-ponents and they can support large scale magnetic fields.The emission from this surface may be detected directly asa soft (0 . − . . − k e V R a t e ( c t s / s ) Figure 1.
NICER light curve of Swift J1858.6–0814 since leavingSun constraint in 2020, showing times of observed Type I bursts(purple). In addition to the long-term flux decrease, several dipsand eclipses are visible; these will be considered in detail in futurework. The full
NICER light curve is shown in black at a resolutionof 40 s and the bursts (purple) extend to their maximum countrate at 0.1 s resolution. The zero-point for the time axis is thestart of 2020 February 25 (MJD 58904). the highest accretion rates, both hydrogen and helium burn-ing occur stably so no Type I bursts are observed. However,close to the transition to stability, the burning is marginallystable and has an oscillatory mode (Heger et al. 2007), whichhas been used to explain the millihertz QPOs observed priorto some Type I bursts (Revnivtsev et al. 2001; Altamiranoet al. 2008; Lyu et al. 2016; Mancuso et al. 2019).There are also other effects which can affect the occur-rence and type (fuel) of thermonuclear bursts. Each burstmay not burn all of the available fuel, so some hydrogen andhelium will remain after one burst and can affect propertiesof following bursts. Similarly, the burnt material will containadditional carbon, nitrogen and oxygen from helium burn-ing. These nuclei catalyse hydrogen burning so can affectlater bursts as well. The neutron star spin (Spitkovsky et al.2002; Galloway et al. 2018) and the geometry of where onthe star the material is accreted (Kajava et al. 2014) canalso affect burst properties.Type I bursts can sometimes be used as standard can-dles, as they can be bright enough to reach the Eddingtonlimit. In this situation, the radiation pressure lifts material inthe NS surface and the atmosphere expands in PhotosphericRadius Expansion (PRE; Tawara et al. 1984; Lewin et al.1984). Since the Eddington limit is only weakly dependenton radius, this produces a period during which the luminos-ity remains constant at the (known) Eddington value. ThisPRE phase may be identified (and distinguished from a sim-ple plateau in the burning rate) by measuring the change inphotospheric radius from the time-resolved X-ray spectrum.The measured flux during the PRE phase may then be usedwith this standard candle to estimate the distance to thesource (van Paradijs 1978; Kuulkers et al. 2003).
MNRAS000
MNRAS000 , 1–14 (2020) ype I bursts in Swift J1858.6–0814 The low-mass X-ray binary Swift J1858.6–0814 has beenin its first observed outburst since late 2018 (Krimm et al.2018). The X-ray emission in the initial phase of the outburstwas highly variable as was the emission in other wavebands(Ludlam et al. 2018; van den Eijnden et al. 2020, Fogantini etal. in prep.): the
NICER . −
10 keV count rate peaks at over650 cts/s within 200 s of intervals at ≈ . < > . α line and K edge(Reynolds et al. 2018; Hare et al. 2020) and soft X-ray emis-sion lines (Buisson et al. 2020a). It also shows P-Cygni linesin its optical spectra, which look similar to those seen inseveral BH XRBs (Munoz-Darias et al. 2019; Mu˜noz-Dariaset al. 2020, Castro-Segura in prep.), as well as strongly vari-able optical emission (Paice et al. 2018). These propertieshave led to Swift J1858.6–0814 being viewed (Hare et al.2020) as an analogue of V404 Cyg (Gandhi et al. 2016; Wal-ton et al. 2017; Motta et al. 2017) and V4641 Sgr (Wijnands& van der Klis 2000; Revnivtsev et al. 2002), which havebeen dynamically confirmed as hosting black holes (Casareset al. 1992; Orosz et al. 2001, respectively). Swift J1858.6–0814 also lies within the range occupied by BHs in the radio-X-ray plane (van den Eijnden et al. 2020). However, recentobservations of Swift J1858.6–0814 have shown qualitativelydifferent X-ray properties, suggesting a state change whilethe source was unobservable due to Sun constraint (between2019 November and 2020 February), although the proper-ties of the initial phase were not typical of a canonical state(e.g. van der Klis 1994). In the 2020 observations, the fluxlevel is much steadier and the strong iron line and edge areabsent (Buisson et al. 2020b, and Figure 1), . These observa-tions have also shown Type I X-ray bursts in both NICER and
NuSTAR data (Buisson et al. 2020c), unambiguouslyidentifying the compact object as a neutron star.In this paper, we analyse the Type I X-ray bursts de-tected in
NICER and
NuSTAR data of Swift J1858.6–0814.
We have inspected the
NICER (Gendreau et al. 2016)light curves from 2020 by eye. Apparent Type I bursts arepresent in OBSIDs 3200400106, 3200400111, 3200400114,3200400121 and 3200400122, corresponding to March 6, 11,14, 21 and 22.We begin with the calibrated, unfiltered events file fromHEASARC ( event_cl/ni32004001**_0mpu7_ufa.evt ). Weuse the standard filters to produce good time intervals(GTIs) apart from the undershoot range, which we relaxfrom ≤
200 s − to ≤
300 s − for the first Type I burst and For further information on the filters, seeheasarc.gsfc.nasa.gov/lheasoft/ftools/headas/nimaketime.html ≤
250 s − for the second. This is required due to high opti-cal loading due to the relatively low Sun angle. Additionally,to include the peak of the second burst, we relax the off-set from the nominal target direction slightly, using 0.0155 ◦ rather than 0.015 ◦ . This is a small change from the standardvalue, so data during this time are unlikely to show signifi-cant deviations from the standard calibration. Further, thefourth burst occurs during passage through the South At-lantic Anomaly (SAA) and the overshoot rate reaches closeto 5 s − , so is removed by standard filtering. We removethese filters in order to show the light curve but note thatthe spectrum may be affected.We then use nicerclean to produce a clean events list,which we then barycentre to the ICRS reference frame andJPL-DE200 ephemeris. From this, we extract spectra andlight curves using xselect .We use NuSTAR (Harrison et al. 2013) OB-SID 90601308002, which overlaps with
NICER
OBSID3200400106. We reduce this using the standard nupipeline and nuproducts software, version 1.9.0. We use a source re-gion of a circle of radius 2 arcmin centred on the centroid ofthe detected counts. We use a background region of a circleof radius 2 arcmin from a source-free area of the detector.
We show the light curve of Swift J1858.6–0814 since leav-ing Sun constraint on 2020 February 25 in Figure 1. Thecount rate shows a secular decrease throughout the whole ofthis period, punctuated by short dips and eclipses as well asthe five Type I bursts analysed here. The drop in persistentcount rate from the first to last burst was by a factor ofaround 4 and bursts were in general brighter at lower per-sistent count rate, with only the fourth burst not followingthis trend. Around the time of the last observed burst, therate started decreasing more rapidly, before flattening oncemore. The period of fastest flux drop extended considerablybefore and after the final burst, so the coincidence in timeis probably only by chance. As well as the Type I bursts,several dips are present, many of which are due to eclipses(Buisson et al. 2020c); these will be analysed in detail infuture work.The times between Type I bursts are 4.5, 3.6, 6.3 and1.4 days (we summarise lists, such as this, of properties ofeach burst in Table A1). Since the coverage of Swift J1858.6–0814 is not continuous, there may have been other bursts be-tween those observed, in observation gaps. Therefore, thesegaps are an upper limit to the recurrence time. The duty cy-cle of
NICER observations is low ( ≈ .
9% over the 37 daysshown in Figure 1, but not evenly across this time) so itis very likely that other bursts did occur outside timesof observation. Furthermore, we can consider the α -value,the ratio of inter-burst (persistent) fluence to burst fluence,which is typically ≈
40 for hydrogen fuelled bursts and ≈ −
200 for helium (e.g. Gottwald et al. 1986; Gallowayet al. 2004). Here, the lowest observed α ≈
500 (for burst 5,integrating the fluxes found in Section 3.4) is higher, mean-ing more emission occurs between bursts than would be ex-pected. This suggests that other, intermediate bursts did
MNRAS , 1–14 (2020)
D. J. K. Buisson et al. occur and/or substantial nuclear burning occurred betweenbursts.There is also a period longer than any gap betweenobserved bursts at the start of the
NICER monitoring( ≈
10 days) where no bursts are observed; again, it is pos-sible that bursts did occur during this period but that theyoccurred during gaps in the
NICER monitoring (which ob-served only 0.24 days of this time). The count rate and spec-tral shape show no large changes during this time, so thereis no obvious reason for bursts not to have occurred. An al-ternative explanation for the lack of bursts in this period isthat it followed a superburst, which quenched the normalType I bursts (Keek et al. 2012); however, there is no evi-dence in the
NICER monitoring or
MAXI data (which coverearlier times) for a superburst having occurred.
The first burst was observed by both
NuSTAR and
NICER .We show a
NuSTAR image of the sky around Swift J1858.6–0814 in Figure 2. This shows that, to the resolution availableto
NuSTAR , only one source is apparent in the
NICER fieldof view and the location of the Type I burst flux is consistentwith the location of the persistent emission. The offset be-tween the
NuSTAR position and the nominal
NICER point-ing is around 15 arcsec, which is less than the 1 arcmin nom-inal pointing stability of
NICER (Arzoumanian et al. 2014).This shows that the X-ray bursts are from Swift J1858.6–0814.
The light curves for each Type I burst are shown in Fig-ure 3. Each burst has a fast rise, lasting (cid:46) ≈
40 s. The decay of each burst, except thefirst, has an initial fast drop (within ≈ − ≈
40 s) when the burst is observableover the persistent flux. This fast drop is by a greater fac-tor in brighter bursts (Figure 4); for example, this drop isby a factor of ≈ . ≈ ≈ ≈
290 cts/s over 0 . −
10 keV; the next faintest, burst 4,peaks at ≈
600 cts/s) and shorter than the others. Apartfrom the fourth, each burst is stronger than the previousone, while the persistent count rate decreases; this couldbe due to partial burning of the accreted material occur-ring outside bursts producing more H-poor fuel in the latterbursts, if more inter-burst burning occurred due to a longerrecurrence time.
We extract time-resolved spectra for each Type I burst us-ing time intervals containing a minimum number of photons.First, we estimate the persistent emission from the intervalfrom 200 s to 50 s before the burst peak. We then definethe start of the burst: we take a light curve binned to 0.1 sand find the final bin before the burst which is not abovethe persistent rate. We define the burst as starting at theend of this bin. Starting from this point, we extract spectrafrom time intervals containing at least 300 counts in excessof that expected from the persistent rate. We then fit thespectrum of the burst emission as the difference betweeneach burst spectrum and the persistent spectrum (this isperformed by treating the persistent emission as the back-ground). We initially model the burst emission with a singleblackbody. Apart from around the burst peaks, the spectraare described well by this model. However, spectra aroundthe peaks of the second, third and fifth bursts are broaderthan a simple blackbody and in the fits show excess emis-sion at low energies. We test two alternative phenomenologi-cal models to explain this excess: allowing the normalisationof the persistent emission to change by a factor (1 + f a )(Worpel et al. 2013) or adding a second blackbody. The for-mer case requires a model for the persistent emission; weuse tbabs × (diskbb+bbody) , representing an absorbed discand blackbody (we also use this model for the accretion rateestimates in Section 3.5). This soft state model gives a goodfit to the persistent spectra for each burst (the worst case χ / d . o . f . = 226 . /
212 = 1 . , p = 0 .
24) and agrees withvarious properties of the bursts (mentioned throughout thiswork) which match other bursts observed during the softstate.Both of these burst models provide good fits to all spec-tra (Figure 5) and provide similar peak fluxes and qualitativebehaviour of the first blackbody component’s radius aroundthe burst peak. The total fit statistics for burst 5 for thespectra from times where a single black body gives a poorfit ( χ ν >
2) are: χ / d . o . f . = 78 . /
76 for the double blackbody model; and χ / d . o . f . = 73 . /
83 when varying the per-sistent emission. This implies a weak preference for a changein the strength of the persistent component but both mod-els are statistically acceptable so we regard both options aspossible. Parameters of the fits of each of these models toburst 5 are shown in Figure 6. Bursts 2 and 3 show similarfeatures with lower signal; burst 1 has much lower signal;and we do not analyse burst 4 in detail due to the enhancedbackground from the SAA.The area of the blackbody increases around the Type Iburst peak before reducing and settling to a steady value forthe majority of the burst tail, characteristic of photosphericradius expansion. Both well-fitting models (two blackbodiesor additional persistent emission) show a similar degree ofexpansion, by around a factor of 2 over the radius in thetail of the burst, once the radius has settled to a steadyvalue. The dip in blackbody radius after the burst peak tobelow the tail value is typical of bursts while accreting inthe soft state, which agrees with our identification from thepersistent spectrum, and is likely due to a changing colour-correction factor (G¨uver et al. 2012a,b; Kajava et al. 2014).These fits show a fast rise and smooth decay in bolomet-ric flux. The apparent double peak in the flux curve for the
MNRAS000
MNRAS000 , 1–14 (2020) ype I bursts in Swift J1858.6–0814 h m s s s s s -8°12'14'16' Right Ascension D e c li n a t i o n C o un t s p e r p i x e l h m s s s s s -8°12'14'16' Right Ascension D e c li n a t i o n C o un t s p e r p i x e l Figure 2. −
50 keV
NuSTAR image of the sky around Swift J1858.6–0814. Left: over the full (26.7 ks on source time) observation; onlya single point source is apparent, at the position of Swift J1858.6–0814. Right: during the Type I burst only; the source position matchesthe position during the full observation. The
NICER field of view is shown by the black circle and the nominal pointing direction by theblack cross. R a t e ( c t s / s ) NuSTARNuSTAR − × NICER . − Figure 3.
Light curves of each Type I burst, in order of occurrence. Purple: 0 . −
10 keV
NICER ; red: 3 −
10 keV
NuSTAR , scaled(increased by a factor of 5) and offset (by +600 cts s − ). Each burst has had the persistent rate (the mean rate from 50 −
200 s beforethe burst) subtracted. The shaded regions are the 1 σ Poisson uncertainties. single blackbody model is likely due to the poor fit aroundthis time, although double peaks in bolometric luminosityhave been seen in other PRE bursts (Jaisawal et al. 2019).The comparatively smooth flux profile contrasts with thefast drop in count rate after the peak; the difference beingdue to the higher temperatures early in the decay producinga lower count rate for a given flux (when convolved with theinstrument response, given the
NICER effective area curveand the temperatures concerned).Near the times of the Type I burst peaks (within about2 s), there is an excess of soft emission over the simple blackbody model. Similar excesses have been seen in Type I burstsin many other sources observed with
NICER , e.g. Aql X-1(Keek et al. 2018a), 4U 1820–30 (Keek et al. 2018b) andSAX J1808.4–3658 (Bult et al. 2019). This could be dueto other extra components such as re-emission from the disc (corresponding to the extra black body, Keek et al. 2018a) orenhanced accretion through Poynting-Robertson drag (cor-responding to the change in persistent emission normalisa-tion, Worpel et al. 2013). There may also be deviation from asimple blackbody due to Comptonisation (Keek et al. 2018b)or scattering processes in the atmosphere (Romani 1987).The data for the X-ray bursts presented here are not sensi-tive enough to distinguish between these possibilities clearly.
Since the later Type I bursts (certainly burst 5, with someevidence also in bursts 2 and 3) show photospheric radiusexpansion, their peak luminosity should be governed by theEddington limit. The observed flux can then be used to es-timate a distance. Initially, we use L Edd = 3 . × erg/s, MNRAS , 1–14 (2020)
D. J. K. Buisson et al. R a t e ( c t s / s × o ff s e t ) Figure 4.
Light curves of each Type I burst, in order of occur-rence from bottom to top. Purple: 0 . −
10 keV
NICER . Eachburst is offset from the previous by a factor of 10 . . Each bursthas had the persistent rate (the mean rate from 50 −
200 s beforethe burst) subtracted. The cooling tail is similar for each burstbut bursts 2, 3 and 5 have a stronger initial peak. The shadedregions are the 1 σ Poisson uncertainties. found empirically by Kuulkers et al. (2003) to be suitablefor neutron stars at known distance, and to have an accu-racy of 15% for source-to-source variation. This matches theEddington limit of a helium atmosphere around a 1 . M (cid:12) object.We take the peak flux from the second, third and fifthType I bursts (which are consistent with each other; differenttemperatures mean that these correspond to different countrates). We use the model of the burst including a scaled per-sistent emission component (see Section 3.4), although thedouble black body model gives very similar results. For eachburst, we use the least squares average of the fluxes from in-tervals which are consistent within 1- σ of the highest value.These values are consistent with each other and their aver-age is 1 . ± . × − erg cm − s − , which gives a distanceof D = 16 . +1 . − . kpc (1 σ ) . This puts Swift J1858.6–0814 at the far side of theGalaxy; given its Sky coordinates ( l = 26 . b = − . . D = 16 . +1 . − . kpc (1 σ ) . From the relative densities of the components of theGalactic model at this position, we infer that Swift J1858.6–0814 is most likely (75%) to be a disc object but could alsobe part of the halo (25%). A bulge origin is highly unlikely( P (Bulge) = 7 × − ). R a t e / C o un t ss − k e V − -9 F l u x d e n s i t y / e r g c m − s − k e V − DataSingle black bodyDouble black bodyVariable persistent emission123 D a t a / m o d e l ( D a t a - m o d e l ) / e rr o r Figure 5.
Comparison of different models for the net burst emis-sion at the peak of the burst (the spectrum with highest countrate in Burst 5). A single blackbody (navy) is a poor fit; two black-bodies (yellow) or a contribution proportional to the persistentflux (green) both give similarly good fits.
There are systematic effects which may affect this dis-tance estimate (e.g. Galloway et al. 2008b). Many of these,such as differences in neutron star mass and photospheremetallicity, are implicitly included by the empirical natureof the critical luminosity (and its uncertainty) measured byKuulkers et al. (2003). However, the effects of obscurationin high inclination sources are not accounted for – Kuulkerset al. (2003) find that in some high inclination sources theobserved PRE luminosity is significantly lower. In this case,the photosphere may be partially obscured by larger compo-nents of the system, principally the disc. For the simple caseof a razor-thin disc, the disc can obscure up to half of theNS so the flux may be underestimated by up to a factor of 2and the distance may actually be smaller by a factor of upto √
2. This factor is mitigated by reflection of the radiationintercepted by the disc but may be increased by a thick discHe & Keek (2016).To show the magnitude of these effects, we show dis-tance estimates for various specific values of metallicity andinclination in Figure 7 and Table 1. We calculate the dis-tances by replacing the empirical peak luminosity from Ku-ulkers et al. (2003) with the theoretical Eddington luminos-
MNRAS000
MNRAS000 , 1–14 (2020) ype I bursts in Swift J1858.6–0814 . − k e V R a t e ( c t s / s ) -9 -8 B o l o m e t r i c f l u x ( e r g c m − s − ) E ff e c t i v e r a d i u s ( k m a t . k p c ) k T ( k e V ) E ff e c t i v e r a d i u s ( k m a t . k p c ) χ ν A dd i t i o n a l p e r s i s t e n t e m i ss i o n ( f a ) Figure 6.
Parameters of time-resolved spectra of Burst 5. Left column: Parameters from modelling the Type I burst emission with 1or 2 blackbodies. Parameters for a single blackbody model are shown in navy; for two blackbodies, values for the complete model areshown in orange; the hotter blackbody in red and the cooler blackbody in yellow. The two-blackbody model is only shown where thesingle blackbody has χ ν > Table 1.
Distance estimates (kpc) for various gas compositionsand inclinations (see text for details).Isotropic i = 70 ◦ i = 80 ◦ Pure helium 16 . +0 . − . . +0 . − . . +0 . − . Cosmic abundances 12 . +0 . − . . +0 . − . . +0 . − . ity (e.g. Lewin et al. 1993) modified by the anisotropy factor( ξ b ) from (He & Keek 2016), L Obs = 8 πGm p M NS cξ b σ T (1 + X )(1 + z ( R ))where G is the gravitational constant, m p is the proton mass, M NS is the neutron star mass, c is the speed of light, σ T isthe Thomson cross section, X is the hydrogen mass fractionand z ( R ) is the gravitational redshift at the photosphericradius R . We show the two extremes of likely metallicity, pure helium ( X = 0) and a cosmic abundance of hydro-gen ( X = 0 . ξ − ) for inclinations of 70 − ◦ is a factor of 0 . − . . M (cid:12) and a photo-spheric radius of 20 km. This allows a significantly largerrange of distances than the Kuulkers et al. (2003) range,due to the lower effective Eddington luminosities for highhydrogen fractions and high inclination. However, even thesmallest distance estimate (9 . +0 . − . kpc) is 60% of the valuederived from the Kuulkers et al. (2003) luminosity and be-yond the average distance of Galactic sources along this lineof sight.To reduce the range of these estimates, we can considerwhether particular values of parameters generating the sys-tematic uncertainty are preferred by other evidence. The MNRAS , 1–14 (2020)
D. J. K. Buisson et al. P D F ( k p c − ) P ( d | Galactic) P ( d | f, Kuulkers) P ( d | f, Helium) P ( d | f, Helium , i = 70 ◦ ) P ( d | f, Helium , i = 80 ◦ ) P ( d | f, Cosmic) P ( d | f, Cosmic , i = 70 ◦ ) P ( d | f, Cosmic , i = 80 ◦ ) Figure 7.
Distance estimates for Swift J1858.6–0814 based onvarious critical luminosities for PRE. Distance estimates for sev-eral specific luminosity values are shown (details in legend andtext) along with the empirical luminosity range found by Kuulk-ers et al. (2003). The Galactic prior is shown in grey. eclipse duration implies an inclination of at least 70 ◦ (Buis-son et al. 2020c, Buisson et al. in prep.). A more accuratedetermination of the inclination would require detailed mod-elling of optical light curves and spectra, beyond the scope ofthis paper; meanwhile, we regard our calculation using 80 ◦ as a fiducial value. The atmospheric composition of a TypeI burst can be inferred from its light curve. The relativelyfast rise and initial decay of the PRE bursts observed heresuggest a helium burst. Additionally, helium fuelled burstsare also more common during the soft accretion state andthe dip in apparent radius below the final value is moretypical of soft state bursts (Kajava et al. 2014). Further,bursts can reach the Eddington limit for helium even whereaccreted material is hydrogen rich, either by the hydrogenbeing burnt between bursts or the hydrogen rich atmospherebeing blown off by the burst (Bult et al. 2019; Galloway et al.2006). This would imply that the further distance estimates(blue curves in Figure 7) are more likely (12 . +0 . − . kpc for i = 80 ◦ ).With this distance estimate, we can also estimatethe accretion rate at the times of the bursts from thepersistent flux measurements from the modelling in Sec-tion 3.5. We find a bolometric (of the X-ray compo-nents) flux before the bursts of 12 . +0 . − . , 10 . +0 . − . , 8 . +0 . − . ,5 . +0 . − . and 2 . +0 . − . × − erg cm − s − , in chronologicalorder. If the persistent emission has the same anisotropyas the burst, this implies an Eddington fraction ˙ m Edd =0 . +0 . − . , . +0 . − . , . +0 . − . , . +0 . − . and 0 . ± .
004 formaterial in the accretion flow (calculating L Edd for X =0 . ◦ ,the increase is by a factor of 1.2-2. This is similar to therange at which helium fuelled bursts are expected and ob- served (Galloway et al. 2008a) but extends slightly higher,so there could be some influence of residual hydrogen in theburning material.Our distance estimates are all relatively large (Gallowayet al. 2008a; Gandhi et al. 2019) but not unprecedented (e.g.Homan et al. 2014) for an XRB. The absorbing column den-sity ( ≈ × cm − ) is comparatively low for such a distantsource, but the total Galactic column density in the direc-tion of Swift J1858.6–0814 is similar (1 . × cm − , HI4PICollaboration et al. 2016).A large distance can also help in explaining the strongvariability observed in the initial state of Swift J1858.6–0814 (during 2018-9): it is comparatively faint for a binarybut strong winds (Mu˜noz-Darias et al. 2020) and variability(Ludlam et al. 2018) are often explained by a high Eddingtonrate (King & Pounds 2003; Grupe 2004). During the flaringstate but between flares, the observed flux of Swift J1858.6–0814 was ≈ . × − erg cm − s − (Hare et al. 2020) ,which is ≈
5% of the Eddington limit for hydrogen and a1 . M (cid:12) object. Some flares increased count rates by factorsof many tens, so at least during bright flares, the luminositywas above the Eddington luminosity (and correcting for anyanisotropy is likely only to increase the strength of this). Ifmuch of the variability was due to obscuration, the intrinsicluminosity would also have been above Eddington at othertimes. We also looked for millihertz quasi-periodic oscillations(mHz QPOs), which are sometimes found before an X-rayburst (e.g. Revnivtsev et al. 2001; Altamirano et al. 2008;Mancuso et al. 2019). We used 0 . −
10 keV light curvesat 1 second resolution and applied the Lomb-Scargle peri-odogram (Lomb 1976; Scargle 1982) to each gap-less lightcurve, excluding periods of dipping and eclipses. In the fivecases where we detected the type-I X-ray bursts, we searchedfor the oscillations before and after the X-ray bursts. To es-timate the significance level, we followed the approach of(Press et al. 1992), which assumes white noise and takes asa number of trials the number of frequencies explored.We detected a mHz QPO at a significance of 5 . σ inthe 1.8 ks of data before the 5th X-ray burst (Figure 8). ThemHz QPO has an average frequency of 9 . ± . . ± .
2% (0 . −
10 keV). Thereis no evidence of the oscillations in the ≈
600 s of data afterthe X-ray burst, with a 90% upper limit on the rms ampli-tude of 1.2% ruling out that the same strength of oscillationcontinues. We also found marginal evidence of QPOs in atleast three other cases; however the datasets are relativelyshort ( (cid:46) −
700 s), and therefore it is not possible to un-derstand if they are real or the product of red-noise. Theupper limits to the QPO amplitude for the time segmentsprior to the earlier bursts are somewhat lower in fractionalamplitude (0.6, 0.8, 1.0 and 0.8% in chronological order)than for the detected QPO but due to the brighter flux at This was measured for the 3-78 keV band, which for the spec-tral shape of this observation includes the majority of flux; anybolometric correction will only increase the strength of super-Eddington behaviour. MNRAS000
700 s), and therefore it is not possible to un-derstand if they are real or the product of red-noise. Theupper limits to the QPO amplitude for the time segmentsprior to the earlier bursts are somewhat lower in fractionalamplitude (0.6, 0.8, 1.0 and 0.8% in chronological order)than for the detected QPO but due to the brighter flux at This was measured for the 3-78 keV band, which for the spec-tral shape of this observation includes the majority of flux; anybolometric correction will only increase the strength of super-Eddington behaviour. MNRAS000 , 1–14 (2020) ype I bursts in Swift J1858.6–0814 N o r m a li z e d L S p o w e r σ σ R a t e ( . − k e V c t s / s ) Figure 8.
Top: Lomb-Scargle periodogram of the light curvesegment immediately prior to burst 5, which shows a QPO at9 . ± . . −
10 keV
NICER light curve of theperiod up to and including burst 5, during which a 9 . ± . earlier times, all but the last of these limits are higher inabsolute amplitude than the detected QPO. Therefore, wecannot definitively rule out a QPO from the same level of os-cillatory burning occurring prior to the other bursts. We alsonote that a mHz QPO was also detected in a NuSTAR ob-servation during 2019 February (Hare et al. 2019), althoughat a frequency of 2.7 mHz, which is lower than other mHzQPOs which have been explained by marginally stable nu-clear burning.Revnivtsev et al. (2001) find that these mHz QPOs areonly found in a narrow range of luminosities, L −
20 keV =5 − × erg s − . The QPO found here occurs while themean flux (from our model for the persistent emission, Sec-tion 3.4) is f −
20 keV = 1 . ± . × erg cm − s − , whichcorresponds to L −
20 keV = 4 . +0 . − . × erg s − at 16 . th NS system that shows this type of QPOs.The fact that we do not detect more episodes of mHz QPOscould either be due to their intrinsic absence or to detec-tion difficulty: the mHz QPOs are not always present in theX-ray light curves (they are state dependent and even ina given state, there is not yet a clear physical trigger forthem, see Altamirano et al. 2008; Mancuso et al. 2019, etc.);in addition, the frequency and amplitude of these QPOs arevery low, and therefore to acquire enough QPO cycles andsufficient signal-to-noise, uninterrupted datasets longer than1000 − We searched each of the X-ray bursts observed with
NICER for the presence of burst oscillations, but did not detect anysignificant signals. To search for oscillations, we constructeda 1/8192 s time resolution light curve for each X-ray burst,using only those events in the 1 − T = 2 , , T /
2. We then calculated the power spectrum associatedwith each window position, and searched the 100 − The identification of the accretor in Swift J1858.6–0814 asa neutron star informs several outstanding questions relat-ing to the properies of Swift J1858.6–0814. It fits with thelow coronal temperature found in Hare et al. (2020), sinceneutron stars tend to have lower coronal temperatures thanblack holes (Burke et al. 2017).However, the neutron star accretor implies an unusual
MNRAS , 1–14 (2020) D. J. K. Buisson et al. location in the radio-X-ray plane: Swift J1858.6–0814 ap-pears relatively X-ray faint for a NS XRB (van den Eijndenet al. 2020). This could imply that Swift J1858.6–0814 hasan intrinsically unusually low X-ray/radio luminosity ratioor that the observed X-ray luminosity is unrepresentativelylow. The latter case would support a model in which theX-ray emission (which may already be comparatively lowdue to anisotropy, e..g. He & Keek 2016) is usually obscuredby the high inclination disc, apart from during the flares,which represent the true intrinsic luminosity, when viewingthe central source directly through a gap in the (irregular)disc surface.Swift J1858.6–0814 has previously been compared withthe black hole XRBs V4641 Sgr and V404 Cyg (e.g. Hareet al. 2020). All of these sources have shown strong vari-ability due to some combination of changes in intrinsic fluxand obscuration, although the relative contribution of thesetwo effects is not yet clear (e.g. compare Walton et al. 2017;Koljonen & Tomsick 2020). The relative radio loudness alsoprovides a further similarity with V404 Cyg, which is unusu-ally radio loud for its inclination (Motta et al. 2018). Theidentification of Swift J1858.6–0814 as a neutron star XRBmeans the flaring behaviour in these sources must now beexplained in a model which is compatible with a neutronstar accretor. In particular, extreme variability from pro-cesses very close to the event horizon may be ruled out, sincea neutron star is significantly larger than its Schwarzschildradius.
Swift J1858.6–0814 had been active for over a year beforeany Type I bursts were detected; there are several meansto explain the non-detection of bursts during this period.Firstly, there may truly have been no bursts, due to thedifferent accretion regime during this period. In a modelwhere variable obscuration causes much of the strong vari-ability, the intrinsic accretion rate was much higher duringthe flaring period, so would likely have induced stable nu-clear burning of both hydrogen and helium. Additionally,in this model, the obscuration between flares would haveimpeded observation of any Type I bursts which occurredwhile the neutron star was obscured (which is the majorityof the duty cycle). It is also possible that bursts were ob-served but not identified if they occurred at the same timeas flares. The observed flares are all different in spectrum,light curve and/or duration to thermonuclear bursts; how-ever, the variety of flares means that it is possible that aburst coincident with a flare would go unnoticed. Finally, itis also possible that bursts did occur during this phase of theoutburst but, by chance, not during
NICER observations ofSwift J1858.6–0814. Overall, it is unsurprising that X-raybursts had not been detected in the flaring state, whetheror not they occurred.
We can also compare the flaring state to other strong vari-ability regimes in neutron stars. Two famous neutron starsystems exhibiting flare-like behaviour are the Rapid Burster(MXB 1730–335, e.g. Hoffman et al. 1978) and Bursting Pulsar (GRO J1744–28, Fishman et al. 1995). The RapidBurster shows many (up to thousands per day) ‘rapid’ burstsin addition to Type I bursts; these rapid bursts are muchshorter ( <
10 s) and more regular in cadence than the flaresof Swift J1858.6–0814, so are probably different phenomena.The Bursting Pulsar is the archetypal example of Type IIX-ray bursts (Kouveliotou et al. 1996). These bursts also dif-fer markedly from the flares observed in Swift J1858.6–0814:the type II bursts are again much shorter and are accompa-nied by a drop in emission following the burst. Therefore,the flaring state of Swift J1858.6–0814 is not explained asan example of these other unusual neutron star XRB states.The high inclination NS LMXB EXO 0748–676 has alsoshown flaring episodes (Homan et al. 2003), although theseare more sporadically interspersed with other light curveshapes and less prominent at harder energies than those inSwift J1858.6–0814.Transitional millisecond pulsars (tMSPs) also have a‘flaring’ accretion mode (de Martino et al. 2013; Bogdanov &Halpern 2015), although this occurs at much lower luminos-ity ( ≈ erg s − ) than the flaring state in Swift J1858.6–0814 ( (cid:38) erg s − observed). The tMSP flaring mode canalso show strong, variable absorption (e.g. Li et al. 2020), socould be an analogue with lower accretion efficiency.There have not yet been measurements of the magneticfield strength in Swift J1858.6–0814; the closer comparisonof the flaring state of Swift J1858.6–0814 with black holethan neutron star systems could be because the magneticfield of its neutron star is low enough to be unimportant inits accretion flow, implying a relatively low magnetic fieldstrength. ACKNOWLEDGEMENTS
We thank Laurens Keek for helpful discussions and thereferee for comments which improved the manuscript. Wethank the NuSTAR operations team for rapid approval andexecution of our Target of Opportunity proposal. D.J.K.B.and D.A. are funded by the Royal Society. T.G. has beensupported in part by the Scientific and Technological Re-search Council (T ¨UBITAK) 119F082, Royal Society New-ton Advanced Fellowship, NAF \ R2 \ NuSTAR mission, a project led by the California Insti-tute of Technology, managed by the Jet Propulsion Labo-ratory, and funded by the National Aeronautics and SpaceAdministration. This research has made use of the
NuS-TAR
Data Analysis Software (NuSTARDAS) jointly devel-oped by the ASI Science Data Center (ASDC, Italy) andthe California Institute of Technology (USA).
NICER is amission of NASA’s Astrophysics Explorers Program. This
MNRAS000
MNRAS000 , 1–14 (2020) ype I bursts in Swift J1858.6–0814 DATA AVAILABILITY
The data underlying this article are available in HEASARC.
REFERENCES
Altamirano D., van der Klis M., Wijnands R., Cumming A., 2008,ApJ, 673, L35Arzoumanian Z., et al., 2014, The neutron star interior com-position explorer (NICER): mission definition. p. 914420,doi:10.1117/12.2056811Bildsten L., 1995, ApJ, 438, 852Bildsten L., 1998, in Buccheri R., van Paradijs J., Alpar A., eds,NATO Advanced Science Institutes (ASI) Series C Vol. 515,NATO Advanced Science Institutes (ASI) Series C. p. 419( arXiv:astro-ph/9709094 )Bogdanov S., Halpern J. P., 2015, ApJ, 803, L27Bolton C. T., 1972, Nature Physical Science, 240, 124Brown E. F., Bildsten L., Rutledge R. E., 1998, ApJ, 504, L95Buisson D. J. K., et al., 2020a, MNRAS,Buisson D. J. K., Altamirano D., Remillard R., ArzoumanianZ., Gendreau K., Gandhi P., Vincentelli F., 2020b, The As-tronomer’s Telegram, 13536, 1Buisson D. J. K., et al., 2020c, The Astronomer’s Telegram,13563, 1Bult P., et al., 2019, ApJ, 885, L1Burke M. J., Gilfanov M., Sunyaev R., 2017, MNRAS, 466, 194Casares J., Charles P. A., Naylor T., 1992, Nature, 355, 614Cornelisse R., Heise J., Kuulkers E., Verbunt F., in’t ZandJ. J. M., 2000, A&A, 357, L21Cumming A., Bildsten L., 2001, ApJ, 559, L127Cumming A., Macbeth J., 2004, ApJ, 603, L37Dehnen W., Binney J., 1998, MNRAS, 294, 429Fishman G. J., Kouveliotou C., van Paradijs J., Harmon B. A.,Paciesas W. S., Briggs M. S., Kommers J., Lewin W. H. G.,1995, IAU Circ., 6272, 1Fujimoto M. Y., Hanawa T., Miyaji S., 1981, ApJ, 247, 267Gallo E., Degenaar N., van den Eijnden J., 2018, MNRAS, 478,L132Galloway D. K., Keek L., 2017, arXiv e-prints, p.arXiv:1712.06227Galloway D. K., Cumming A., Kuulkers E., Bildsten L.,Chakrabarty D., Rothschild R. E., 2004, ApJ, 601, 466Galloway D. K., Psaltis D., Muno M. P., Chakrabarty D., 2006,ApJ, 639, 1033Galloway D. K., Muno M. P., Hartman J. M., Psaltis D.,Chakrabarty D., 2008a, ApJS, 179, 360Galloway D. K., ¨Ozel F., Psaltis D., 2008b, MNRAS, 387, 268Galloway D. K., et al., 2018, ApJ, 857, L24Gandhi P., et al., 2016, MNRAS, 459, 554Gandhi P., Rao A., Johnson M. A. C., Paice J. A., MaccaroneT. J., 2019, MNRAS, 485, 2642Gendreau K. C., et al., 2016, in Space Telescopes and In-strumentation 2016: Ultraviolet to Gamma Ray. p. 99051H,doi:10.1117/12.2231304 Gottwald M., Haberl F., Parmar A. N., White N. E., 1986, ApJ,308, 213Grimm H. J., Gilfanov M., Sunyaev R., 2002, A&A, 391, 923Grindlay J., Gursky H., Schnopper H., Parsignault D. R., HeiseJ., Brinkman A. C., Schrijver J., 1976, ApJ, 205, L127Grupe D., 2004, AJ, 127, 1799G¨uver T., Psaltis D., ¨Ozel F., 2012a, ApJ, 747, 76G¨uver T., ¨Ozel F., Psaltis D., 2012b, ApJ, 747, 77HI4PI Collaboration et al., 2016, A&A, 594, A116Hare J., Gandhi P., Paice J. A., Tomsick J., 2019, The As-tronomer’s Telegram, 12512, 1Hare J., et al., 2020, arXiv e-prints, p. arXiv:2001.03214Harrison F. A., et al., 2013, ApJ, 770, 103He C. C., Keek L., 2016, ApJ, 819, 47Heger A., Cumming A., Woosley S. E., 2007, ApJ, 665, 1311Hewish A., Bell S. J., Pilkington J. D. H., Scott P. F., CollinsR. A., 1968, Nature, 217, 709Hoffman J. A., Marshall H. L., Lewin W. H. G., 1978, Nature,271, 630Homan J., Wijnands R., van den Berg M., 2003, A&A, 412, 799Homan J., Fridriksson J. K., Wijnands R., Cackett E. M., Dege-naar N., Linares M., Lin D., Remillard R. A., 2014, ApJ, 795,131Jaisawal G. K., et al., 2019, ApJ, 883, 61Kajava J. J. E., et al., 2014, MNRAS, 445, 4218Keek L., Heger A., in’t Zand J. J. M., 2012, ApJ, 752, 150Keek L., et al., 2018a, ApJ, 855, L4Keek L., et al., 2018b, ApJ, 856, L37Kennea J. A., Krimm H. A., 2018, The Astronomer’s Telegram,12160, 1King A. R., Pounds K. A., 2003, MNRAS, 345, 657Koljonen K. I. I., Tomsick J. A., 2020, arXiv e-prints, p.arXiv:2004.08536Kouveliotou C., van Paradijs J., Fishman G. J., Briggs M. S.,Kommers J., Harmon B. A., Meegan C. A., Lewin W. H. G.,1996, Nature, 379, 799Krimm H. A., et al., 2018, The Astronomer’s Telegram, 12151, 1Kuulkers E., den Hartog P. R., in’t Zand J. J. M., VerbuntF. W. M., Harris W. E., Cocchi M., 2003, A&A, 399, 663Lewin W. H. G., Vacca W. D., Basinska E. M., 1984, ApJ, 277,L57Lewin W. H. G., van Paradijs J., Taam R. E., 1993, Space Sci.Rev., 62, 223Li K.-L., Strader J., Miller-Jones J. C. A., Heinke C. O., ChomiukL., 2020, arXiv e-prints, p. arXiv:2004.14573Linares M., Altamirano D., Chakrabarty D., Cumming A., KeekL., 2012, ApJ, 748, 82Lomb N. R., 1976, Ap&SS, 39, 447Ludlam R. M., et al., 2018, The Astronomer’s Telegram, 12158,1Lyu M., M´endez M., Altamirano D., 2014, MNRAS, 445, 3659Lyu M., M´endez M., Zhang G., Keek L., 2015, MNRAS, 454, 541Lyu M., M´endez M., Altamirano D., Zhang G., 2016, MNRAS,463, 2358Mancuso G. C., Altamirano D., Garc´ıa F., Lyu M., M´endez M.,Combi J. A., D´ıaz-Trigo M., in’t Zand J. J. M., 2019, MNRAS,486, L74McClintock J. E., Remillard R. A., 2006, Black hole binaries. pp157–213Migliari S., Fender R. P., 2006, MNRAS, 366, 79Motta S. E., Kajava J. J. E., S´anchez-Fern´andez C., Giustini M.,Kuulkers E., 2017, MNRAS, 468, 981Motta S. E., Casella P., Fender R. P., 2018, MNRAS, 478, 5159Mu˜noz-Darias T., et al., 2020, arXiv e-prints, p. arXiv:2003.12073Munoz-Darias T., Jimenez-Ibarra F., Armas Padilla M., CasaresJ., Cuneo V., Panizo-Espinar G., Sanchez-Sierras J., TorresM. A. P., 2019, The Astronomer’s Telegram, 12881, 1Narayan R., Heyl J. S., 2003, ApJ, 599, 419MNRAS , 1–14 (2020) D. J. K. Buisson et al.
Orosz J. A., Bailyn C. D., 1997, ApJ, 477, 876Orosz J. A., et al., 2001, ApJ, 555, 489Paice J. A., Gandhi P., Dhillon V. S., Marsh T. R., Green M.,Breedt E., 2018, The Astronomer’s Telegram, 12197, 1Patruno A., Watts A. L., 2012, arXiv e-prints, p. arXiv:1206.2727Press W. H., Teukolsky S. A., Vetterling W. T., Flannery B. P.,1992, Numerical recipes in FORTRAN. The art of scientificcomputingRevnivtsev M., Churazov E., Gilfanov M., Sunyaev R., 2001,A&A, 372, 138Revnivtsev M., Gilfanov M., Churazov E., Sunyaev R., 2002,A&A, 391, 1013Reynolds M. T., Miller J. M., Ludlam R. M., Tetarenko B. E.,2018, The Astronomer’s Telegram, 12220, 1Romani R. W., 1987, ApJ, 313, 718Scargle J. D., 1982, ApJ, 263, 835Spitkovsky A., Levin Y., Ushomirsky G., 2002, ApJ, 566, 1018Strohmayer T., Bildsten L., 2006, New views of thermonuclearbursts. pp 113–156Strohmayer T. E., Brown E. F., 2002, ApJ, 566, 1045Strohmayer T. E., Smith E. A., 2011, The Astronomer’s Tele-gram, 3258, 1Strohmayer T. E., Zhang W., Swank J. H., Smale A., TitarchukL., Day C., Lee U., 1996, ApJ, 469, L9Strohmayer T. E., et al., 2018, ApJ, 865, 63Tawara Y., et al., 1984, ApJ, 276, L41Walton D. J., et al., 2017, ApJ, 839, 110Webster B. L., Murdin P., 1972, Nature, 235, 37Wijnands R., van der Klis M., 2000, ApJ, 528, L93Worpel H., Galloway D. K., Price D. J., 2013, ApJ, 772, 94Zdziarski A. A., Lubi´nski P., Smith D. A., 1999, MNRAS, 303,L11de Martino D., et al., 2013, A&A, 550, A89in’t Zand J., 2017, in Serino M., Shidatsu M., Iwakiri W., MiharaT., eds, 7 years of MAXI: monitoring X-ray Transients. p. 121( arXiv:1702.04899 )van Paradijs J., 1978, Nature, 274, 650van den Eijnden J., et al., 2020, MNRAS, 496, 4127van der Klis M., 1994, ApJS, 92, 511van der Klis M., 2006, Rapid X-ray Variability. pp 39–112van der Klis M., Swank J. H., Zhang W., Jahoda K., MorganE. H., Lewin W. H. G., Vaughan B., van Paradijs J., 1996,ApJ, 469, L1
APPENDIX A: SUMMARY OF BURSTPROPERTIES
MNRAS000
MNRAS000 , 1–14 (2020) ype I bursts in Swift J1858.6–0814 Table A1.
Summary of burst properties. B u r s t R ec u rr e n ce t i m e a P e a k r a t e b P e r s i s t e n t flu x c P e r s i s t e n t a cc r e t i o n r a t e d m H z Q P O f r e q u e n c y e m H z Q P O a m p li t ud e e nu m b e r ( d a y s )( . − k e V c t s / s )( × − e r g c m − s − )( ˙ M E dd )( m H z ) − ± . + . − . . + . − . − < . % ≤ . ± . + . − . . + . − . − < . % ≤ . ± . + . − . . + . − . − < . % ≤ . ± . + . − . . + . − . − < . % ≤ . ± . + . − . . ± . . ± . . ± . % a Recurrence times are upper limits as bursts may have occurred between observations. b Peak rate is the highest value in the0 . −
10 keV light curve binned to 0.1 s. c Persistent flux is measured bolometrically. d The persistent accretion rate is not corrected forthe effects of anisotropy, which likely increase it by a factor of 1.2-2. e The mHz QPO is measured over 0.5-10 keV.MNRAS , 1–14 (2020) D. J. K. Buisson et al.
AFFILIATIONS Department of Physics and Astronomy, University ofSouthampton, Highfield, Southampton, SO17 1BJ Department of Astronomy, University of Maryland, CollegePark, MD 20742, USA NASA/Goddard Space Flight Center, Code 662, Green-belt, MD 20771, USA Instituto Argentino de Radioastronom´ıa(CCT-La Plata,CONICET; CICPBA), C.C. No. 5, 1894 Villa Elisa, Ar-gentina Facultad de Ciencias Astron´omicas y Geof´ısicas, Universi-dad Nacional de La Plata, Paseo del Bosque s/n, 1900 LaPlata, Argentina Istanbul University, Science Faculty, Department ofAstronomy and Space Sciences, Beyazıt, 34119, Istanbul,Turkey Istanbul University Observatory Research and ApplicationCenter, Istanbul University 34119, Istanbul Turkey National Space Institute, Technical University of Denmark,Elektrovej 327-328, DK-2800 Lyngby, Denmark MIT Kavli Institute for Astrophysics and Space Research,Massachusetts Institute of Technology, Cambridge, MA02139, USA IRAP, CNRS, UPS, CNES, 9 avenue du Colonel Roche,BP 44346, F-31028 Toulouse Cedex 4, France Eureka Scientific, Inc., 2452 Delmer Street, Oakland, CA94602, USA SRON, Netherlands Institute for Space Research, Sorbon-nelaan 2, 3584 CA Utrecht, The Netherlands Department of Astronomy, Tsinghua University,Shuangqing Road 30, Beijing 100084 China Tsinghua Center for Astrophysics, Tsinghua University,Shuangqing Road 30, Beijing 100084 China NASA Marshall Space Flight Center, NSSTC, 320 Spark-man Drive, Huntsville, AL 35805, USA Universities Space Research Association, Science andTechnology Institute, 320 Sparkman Drive, Huntsville, AL35805, USA Department of Astronomy, University of Michigan, 1085South University Avenue, Ann Arbor, MI 48109-1104, USA Department of Astronomy & Astrophysics, Atat¨urkUniversity, Erzurum, Turkey Astrophysics Science Division and Joint Space-ScienceInstitute, NASA’s Goddard Space Flight Center, Greenbelt,MD 20771, USA Department of Physics, Tor Vergata University of Rome,Via della Ricerca Scientifica 1, I-00133 Rome, Italy INAF - Astronomical Observatory of Rome, Via Frascati33, I-00078 Monte Porzio Catone (Rome), Italy Space Sciences Laboratory, 7 Gauss Way, University ofCalifornia, Berkeley, CA 94720-7450, USA Institute of Astronomy, Madingley Road, Cambridge,CB3 0HA
This paper has been typeset from a TEX/L A TEX file prepared bythe author. MNRAS000