Distances to Galactic X-ray Binaries with Gaia DR2
R. M. Arnason, H. Papei, P. Barmby, A. Bahramian, M.D. Gorski
MMNRAS , 1–16 (2021) Preprint 5 February 2021 Compiled using MNRAS L A TEX style file v3.0
Distances to Galactic X-ray Binaries with
Gaia
DR2
R. M. Arnason, ★ H. Papei , P. Barmby , , A. Bahramian, M.D. Gorski , Department of Physics & Astronomy, Institute for Earth and Space Exploration, University of Western Ontario, 1151 Richmond Street, London, ON N6A 3K7, Canada International Centre for Radio Astronomy Research, Curtin University, GPO Box U1987, Perth, WA 6845, Australia Department of Space, Earth & Environment, Astronomy and Plasma Physics, Chalmers University of Technology 412 96 Gothenburg, Sweden
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
Precise and accurate measurements of distances to Galactic X-ray binaries (XRBs) reduce uncertainties in the determination ofXRB physical parameters. We have cross-matched the XRB catalogues of Liu et al. (2006, 2007) to the results of
Gaia
DataRelease 2. We identify 86 X-ray binaries with a
Gaia candidate counterpart, of which 32 are low-mass X-ray binaries (LMXBs)and 54 are high-mass X-ray binaries (HMXBs). Distances to
Gaia candidate counterparts are, on average, consistent with thosemeasured by
Hipparcos and radio parallaxes. When compared to distances measured by
Gaia candidate counterparts, distancesmeasured using Type I X-ray bursts are systematically larger, suggesting that these bursts reach only 50% of the Eddington limit.However, these results are strongly dependent on the prior assumptions used for estimating distance from the
Gaia parallaxmeasurements. Comparing positions of
Gaia candidate counterparts for XRBs in our sample to positions of spiral arms in theMilky Way, we find that HMXBs exhibit mild preference for being closer to spiral arms; LMXBs exhibit mild preference forbeing closer to inter-arm regions. LMXBs do not exhibit any preference for leading or trailing their closest spiral arm. HMXBsexhibit a mild preference for trailing their closest spiral arm. The lack of a strong correlation between HMXBs and spiral armsmay be explained by star formation occurring closer to the midpoint of the arms, or a time delay between star formation andHMXB formation manifesting as a spatial separation between HMXBs and the spiral arm where they formed.
Key words:
X-rays: binaries – X-rays: bursts – Galaxy: structure – parallaxes
X-ray binaries (XRBs) are rare systems comprised of a main-sequence star in a close binary orbit with a neutron star (NS) orblack hole (BH). The accretion of material from the main-sequencecompanion onto the compact object results in X-ray emission whichdominates much of the point source population of the X-ray sky.Aside from the type of accretor, XRBs are principally categorizedbased on the mass of the companion. Binaries where the compactobject accretes from the wind of a star >
10 M (cid:12) are classified ashigh-mass X-ray binaries (HMXBs), while those that accrete from theRoche lobe overflow of a < (cid:12) companion are known as low-massX-ray binaries (LMXBs; van Paradijs 1998; Casares et al. 2017).There are a handful of XRBs where the companion is of intermediatemass 1 − (cid:12) , but they are rare compared to the other two types ofsystem. It is expected that many primordial intermediate-mass X-raybinaries have evolved to LMXBs in the present day through masstransfer (Podsiadlowski & Rappaport 2000).XRBs are interesting extraterrestrial laboratories that permit thetesting of our understanding of physical processes under extremesof gravity, rotation rate, pressure, temperature, and magnetic fieldstrength. In addition, a number of interesting astrophysical phenom- ★ E-mail:[email protected] ena can be studied through XRBs, such as wind physics, neutronstar equation of state, and high-energy radiative processes. Asidefrom their value to these astrophysical questions, XRBs can alsoprovide independent constraints on their formation environment onlarger scales (Lehmer et al. 2010; Boroson et al. 2011; Zhang et al.2012; Tremmel et al. 2013). LMXBs can act as independent trac-ers of stellar mass, since low-mass stars comprise the bulk of thestellar mass in a population (Gilfanov 2004). Additionally, LMXBsare preferentially found in areas of high stellar density, such as theglobular clusters of the Galaxy and in the direction of the Galac-tic centre, likely due to their formation by dynamical mechanisms(Clark 1975; Pooley et al. 2003; Muno et al. 2005; Verbunt & Lewin2006; Degenaar et al. 2012). By contrast, the high-mass companionsof HMXBs are short-lived, so they are useful for tracing recent starformation in a long-term galactic evolution context. Observations ofnearby galaxies have suggested that the star formation rate (SFR)of a galaxy scales with both the number of HMXBs and their col-lective X-ray luminosity, albeit with a moderate dispersion (Grimmet al. 2003; Mineo et al. 2012). Finally, XRBs are one of the fewways to observe the high mass end of the initial mass function inan evolved population, since isolated neutron stars and black holesare challenging to observe and study (Verbunt & Hut 1987; Verbunt2003; Dabringhausen et al. 2012). © a r X i v : . [ a s t r o - ph . H E ] F e b R. M. Arnason et al.
Although field Milky Way XRBs can often be easier to study becauseof their close proximity (compared to XRBs in globular clusters orother galaxies), investigating the relationship between XRBs andgalaxy parameters for the Milky Way is complicated. Our locationwithin the disc of the Milky Way means that lines of sight whereXRBs are expected to be more abundant tend to be heavily extincted.XRBs tend to have a spatial distribution that is distinct from or-dinary stars belonging to the same parent stellar population becausethe supernova that forms the compact object in an XRB system canimpart a velocity kick to the system, often known as a “natal” kick.This velocity kick has two effects: it gives the XRB system a pe-culiar velocity relative to galactic rotation, and it can substantiallydisplace the system (depending on XRB type) from the star form-ing region where its progenitor formed (González Hernández et al.2005; Dhawan et al. 2007). Repetto et al. (2012) investigated hownatal kicks at the birth of black hole LMXBs are necessary to explaintheir observed distribution in the Milky Way, particularly the pres-ence of LMXBs at significant (1 kpc) distances above the disc. Theyfound that these kicks tend to be similar to those found for neutronstars, a property which has been interpreted as a consequence of theasymmetry of the supernova explosion (Janka 2013).Naively, we expect that if HMXBs are correlated with star forma-tion on a global scale, they should have a spatial correlation with thesites of star formation in the spiral arms. The shape and extent of theMilky Way’s spiral arms is not easy to resolve compared to externalgalaxies observed face-on. Positions of the spiral arms themselvesare typically inferred through the fitting of analytical models to anensemble of observational tracers, including CO maps, Hii regions,pulsars, masers, stellar kinematics, and dust emission (Vallée 2014).To date, investigations of the correlation between HMXBs and thespiral arms have been done using only two proxies of the spiral arms.Bodaghee et al. (2012) measured spatial cross-correlation betweenOB associations and HMXBs, finding that they have a characteristicoffset of 0 . ± . . ± .
05 kpc and 1 . ± . . A principal reason for desiring accurate distances to XRBs in theMilky Way is that many of these XRBs can be studied in detail. Withthe exception of XRBs located in the direction of the Galactic centre,in the Milky Way the population of XRBs can be studied to fainterX-ray luminosities, and identifications of a unique optical counter-part are more straightforward. Since individual XRBs are most easilystudied in the Milky Way, our understanding of individual XRBs inother galaxies and their parameters as an ensemble population areaffected by studies of nearby XRBs. Measuring the distance to in-dividual XRBs accurately is important because the uncertainty on a number of desired properties in an XRB system can be limited by theuncertainty on distance. For example, measurements of distance canaffect the inferred size of the accretor (i.e., neutron star radius), in-ferred mass of either component of the system (either the companionmass or the mass of the accreting neutron star/black hole), inferredmass transfer rate, and other relevant accretion physics due to theinferred luminosity (Galloway et al. 2003; Jonker & Nelemans 2004;Nättilä et al. 2017; Steiner et al. 2018).The principal difficulty in measuring distances to XRBs is that theylack a universal property or characteristic that would allow them tobe used as a standard candle. XRBs are also extremely rare com-pared to ordinary stars, meaning that population-based methods ofdetermining distances to objects, such as main sequence fitting of astar cluster, cannot be used on XRB populations. Although one canuse the main sequence of ordinary stars in a cluster to determinethe distance to XRBs in that cluster, the rarity of XRBs means thatconstructing an “XRB main sequence" is untenable. X-ray emissionfrom the accretor which irradiates the companion may modify theexpected emission at longer wavelengths, causing an excess in thebluer filters of the visible domain (Phillips et al. 1999; Muñoz-Dariaset al. 2005; Linares et al. 2018; Bozzo et al. 2018). Failing to accountfor these effects on the expected optical emission of an XRB may leadto incorrect estimates of distance from photometric methods. Theseeffects are themselves modified by the mass transfer rate, accretiongeometry, orbital phase, and accretion state of the system, meaningthat they can change with time and may require simultaneous multi-wavelength observations for distances to be usefully constrained.A number of techniques have been used to constrain distance mea-surements of Milky Way XRBs. The most common of these is tomeasure a photometric distance by assuming that the emission isdominated by the companion at longer wavelengths. In general, thismethod is subject to substantial uncertainties, not only due to thecontribution of the accretor, but also due to uncertainties in spec-tral classification and calibrating the absolute magnitude (Reig &Fabregat 2015). A small number of XRBs have had their distancesdetermined via radio parallax or the proper motion of a launchedjet (Hjellming & Johnston 1981; Bradshaw et al. 1999; Miller-Joneset al. 2009). This form of measurement provides relatively tight con-straints on distance, but is possible only for objects that are sufficientlyradio-bright and moderately nearby.An X-ray specific method of measuring distances is to use theobserved flux from Type I X-ray bursts. These bursts occur when asufficient amount of accreted material, mostly hydrogen, accumulateson the surface of a neutron star to trigger a thermonuclear runawaythat produces a characteristic burst (Lewin et al. 1993). The burstis specifically the result of nuclear burning on the neutron star. Asubset of these bursts have steady hydrogen burning followed by ig-nition of a helium layer beneath the hydrogen layer on the surface.The ignition of this helium layer produces a burst that is sufficient tolift the photosphere off the surface of the neutron star. These burstsare known as photospheric radius expansion (PRE) bursts, and theluminosity of the X-ray burst is expected to be at the Eddington lu-minosity during the expansion and contraction of the photosphere(Kuulkers et al. 2003). Since the Eddington limit is fixed for a par-ticular accretor mass (and gas composition/opacity), this means thatthe mass, radius, and distance of a neutron star can be constrained bycomparing the observed flux to the modelled Eddington luminosityfor that object. (Strohmayer & Bildsten 2006; Bhattacharyya 2010).The use of X-ray bursts to infer distance was suggested not long afterthe detection of such bursts by early X-ray satellites. This relationhas been calibrated using X-ray bursts observed in Galactic globularclusters (van Paradijs 1978, 1981; Verbunt et al. 1984) and applied
MNRAS000
MNRAS000 , 1–16 (2021) aia distances to Galactic X-ray binaries to several Galactic XRBs that exhibit either PRE or PRE-like bursts(Basinska et al. 1984; Galloway et al. 2003; Jonker et al. 2004).Evaluations of this method have shown that uncertainties around themodelling assumptions in this method can result in uncertainties indistance, neutron star mass, and neutron star radius (Galloway et al.2008b).With the exception of Type I X-ray bursts, most of the distance-determination techniques require the identification of an opti-cal/infrared counterpart to the X-ray source. Identification of a coun-terpart requires high spatial resolution and accurate determination ofX-ray position. Existing catalogues of XRBs include sources whichhave not been re-detected since their discovery prior to the era of highangular resolution telescopes, and as such have poorly-determinedpositions that could have many candidate counterparts. The presenceof interstellar extinction along particular lines of sight can interferewith the identification of optical counterparts for many XRB sources.Aside from studies of individual objects using telescopes such as the Hubble Space Telescope , the principal existing parallax survey ofobjects in the Milky Way was conducted by the
Hipparcos satellite(Perryman et al. 1997).
Hipparcos provides parallax for only ∼ sources, and has a fairly shallow limiting magnitude of 12. A handfulof nearby XRBs have had their distances determined via Hipparcos parallax (see, for example, Chevalier & Ilovaisky 1998).
Hipparcos data provides reliable measurements of distance within a few hun-dred parsecs of the Sun, which excludes (based on estimates using theother distance methods described above) the overwhelming majorityof XRBs known in the Milky Way.
Gaia
DR2 as a Probe of XRB Distances
The successor to
Hipparcos , the
Gaia satellite, was launched in 2013and aims to have full five-parameter measurements (position, propermotion, parallaxes) for ∼ Gaia results and an early release of a thirdversion (Gaia Collaboration et al. 2016b, 2018, 2020).
Gaia datarelease 2 (DR2), released in April 2018 and based on the first 22months of data taken, contains over 1.3 billion sources which havefull five-parameter measurements, an improvement of five ordersof magnitude of
Hipparcos for parallax measurements. Dependingon the required uncertainties,
Gaia
DR2 contains measurements forobjects to a limiting 𝐺 magnitude of 17–21. So far, Gaia
DR2 hasalready provided a wealth of information for studying populations inand nearby the Milky Way that deviate from the expected dynamicsof ordinary stars in the Milky Way. For example, measurements ofcandidate hypervelocity stars using
Gaia
DR2 have shown that manyof them are in fact bound to the Milky Way, but at least one objecthas an origin in the direction of the Magellanic Clouds, suggestingthe presence of a supermassive black hole in the Large MagellanicCloud (Boubert et al. 2018; Erkal et al. 2019). Gandhi et al. (2019)searched for
Gaia
DR2 candidate counterparts for Galactic black holetransients, finding that distances from
Gaia counterparts generallyagreed with prior distance estimates. Notably, they found that theblack hole BW Cir has a
Gaia distance of ∼ . ± . Gaia to measure dis-tances to XRBs and assess the accuracy of pre-
Gaia distance mea-surements. We include not only binaries with black holes/black holecandidates but also neutron star/neutron star candidate binaries and those with no clear identification of accretor type. Given that XRBsare expected to deviate from the Milky Way’s stellar distribution insubtle to dramatic ways,
Gaia
DR2 offers a unique chance to createa sample of XRBs whose distances are determined by a uniformmethod, as compared with the heterogeneous mix of methods usedfor XRB distance determination whose accuracies, systematics, andmodel dependencies may vary greatly. It also offers an opportunity tocalibrate alternative methods of measuring distance for use in the gen-eral case where parallax measurements are not available.
Gaia
DR2measurements are subject to several known systematic effects, in-cluding centroid wobble caused by unresolved stellar companions(Belokurov et al. 2020) and variation in the parallax zero-point withsource colour and spatial location (Lindegren et al. 2018; Arenouet al. 2018). However the widespread use of
Gaia data means thatthese systematics have been investigated and characterized by manydifferent groups (e.g., Chan & Bovy 2020, summarize many determi-nations of the zero-point offset). Uncertainties and systematics are,in general, more poorly understood for the one-off distance measure-ments available in the literature for many XRBs.
Cross-matching XRBs to
Gaia requires input catalogue(s) of knownXRBs and XRB candidates. To date, the most comprehensive cata-logues of XRBs in the Milky Way are the catalogues of high-mass andlow-mass XRBs by Liu et al. (2006, 2007). In general, properties ofthese XRBs (including positional uncertainties) are compiled usingthe best/most recent (at the time of catalogue creation) observationsof these objects. These catalogues are assembled from published ob-servations taken with a variety of X-ray telescopes, including
Uhuru , Einstein , ROSAT , RXTE , Chandra , and
XMM-Newton . As such, thespecific X-ray energies sampled, sensitivities, and coverage of thesecatalogues are non-uniform. Since the most recent updates to thesecatalogues were in 2006 and 2007, they do not include a numberof Galactic XRB candidates discovered since then. However, an ad-vantage of these catalogues is that many of these objects have beenstudied in detail, especially those with identified counterparts. Thisimplies that the expected number of non-XRB contaminants shouldbe low.
In order to assemble a sample of XRBs for
Gaia counterpart match-ing, we combine the Liu et al. (2006) and Liu et al. (2007) cataloguesof Galactic HMXBs and LMXBs. Although the most recent revisionof these catalogues is now over a decade old, they still represent themost complete sample in the literature. In total, these catalogues con-tain 301 XRBs or XRB candidates. We have removed two objectsfrom the Liu catalogues: 1H 0556+286, and 1H 1255-567 (Mu-2Cru), on the basis that they appear to have been misclassified asHMXBs and are in fact ordinary stars (Berghoefer et al. 1996; Tor-rejón & Orr 2001). The majority of the objects have positional ac-curacies (equivalent 90 percent confidence) ∼ (cid:48)(cid:48) or better, typicallythrough identification of an optical counterpart or high-resolution X-ray observation. However, a number of the candidate objects in thesecatalogues have poorly determined positions, especially those thathave not been re-observed since the beginning of the Chandra era.We assume that long-wavelength counterparts identified in the cata-logues are true counterparts to the LMXB/HMXB or LMXB/HMXBcandidates. In order to feasibly attempt to identify
Gaia counterparts,
MNRAS , 1–16 (2021)
R. M. Arnason et al. we select only objects whose positional accuracy is quoted in the cat-alogues as better than < (cid:48)(cid:48) , which provides a sample of 220 XRBs,of which 136 are LMXBs and 84 are HMXBs. Distances to XRBs are estimated using many different methods anda goal of this work is to evaluate the quality of these methods (seealso Jonker & Nelemans 2004; Thévenin et al. 2017). The Liu et al.(2006) and Liu et al. (2007) catalogues provide distance estimatesin the notes to the main catalogue files. We include the previousdistances and original references, as well as an indication of thedistance measurement method in Tables 1 and 2, for LMXBs andHMXBs matched to
Gaia sources, respectively. In some cases, onlya distance range is quoted in the Liu et al. catalogues and we givethe centre of this range. In cases where an upper or lower limitwas given, we quote that number as the distance. Fifteen of theXRBs with
Gaia candidate counterparts (subsection 2.4) had noprevious distance measurement as of the Liu catalogues. The resultsof our literature search for these objects are discussed in Appendix A;we found published distances for ten of these fifteen objects. Wealso updated distances for fifteen additional objects that had morerecently-published distances than those given in the Liu catalogues.The majority of the objects in our sample have distances mea-sured through photometry of the companion, using measured appar-ent magnitude and extinction with an assumed absolute magnitudebased on modelling. Many XRBs with a neutron star have had theirdistance measured using Type I X-ray bursts. Aside from these cat-egories, there are also a handful of objects with
Hipparcos or radioparallaxes, and a variety of other methods for individual objects. Weuse the following labels for different distance methods: • phot: photometric distance using apparent magnitude, extinc-tion, and assumed absolute magnitude of companion • SEDfit: broad-band SED is fit to an assumed model of thecompanion star/accretion disc with distance as a fitted parameter • 𝐴 𝑉 : distance measured using extinction models/Galactic col-umn density • jetPM: distance measured using jet proper motion • cluster: distance is assumed to be that of an associated clus-ter/OB association • burst: X-ray burst is used as a standard candle to obtain distance • VLBAPLX/VLBIPLX: parallax measured using radio interfer-ometery • Kin: distance inferred from the kinematics of associated Hii re-gions • HipPLX: distance measured using parallax from the
Hipparcos satellite • unknown: no previous distance measurementIn the literature, uncertainties on distances to XRBs are reportedin different ways, including making approximations with no quoteduncertainties. As such, in this work we do not attempt to track theuncertainties associated with previous measurements, except for ahandful of cases. In particular, we expect that distances from a radioparallax (VLBI or VLBA) measurement should be more precisethan those from Gaia
DR2, and
Gaia
DR2 distances should agreewith parallaxes measured with the
Hipparcos satellite. In generalwe expect that the distance to
Gaia candidate counterparts is morereliable and that the
Gaia
DR2 methodology and systematics are,when taken as a whole, better understood than for the heterogeneousensemble of other methods.
We searched for counterparts to our XRB sample by cross-matchingwith the
Gaia
DR2 public release. Initially, we collected all po-tential counterparts with a tolerance of < (cid:48)(cid:48) and then refined thematches to only include counterparts whose angular separation wasless than the quoted positional uncertainty for each individual ob-ject. As per the catalog description, any object that does not have aquoted positional uncertainty is assumed to be accurate to ∼ (cid:48)(cid:48) orbetter (Liu et al. 2006, 2007). We have chosen the conservative caseof a ∼ (cid:48)(cid:48) positional uncertainty for these objects. In the case thatan object had asymmetric positional uncertainties in right ascensionversus declination, we conservatively chose the maximum of thesetwo. With this refinement, 99 XRBs from the Liu catalogues have atleast one candidate Gaia counterpart. In total, we find 126 potentialcounterparts for the Liu XRBs. Most objects have only one counter-part, while a handful (those with more poorly determined positionalaccuracy) return two or more potential counterparts.We further refined our sample of potential XRB counterparts byconsidering the probability that each
Gaia source is aligned withthe position of the XRB by chance alone. To estimate probability ofour X-ray sources matching a random
Gaia source, we picked 5000random coordinates within 0.1 deg of each X-ray source and cross-matched these random realisations against the Gaia catalog to identifythe closest real Gaia source to each random pair of coordinates andmeasured the angular distance between each random pair of coordi-nates and the closest real
Gaia source to that pair. The distribution ofthese distances in the vicinity of each X-ray source (for which theserandom samples were generated and crossmatched against Gaia) isdirectly proportional to the probability of chance overlap between asource in the Gaia catalog and any random pair of coordinates (in-formed partially by the density of Gaia catalog in the vicinity of eachX-ray source). We approximated the probability of a random matchby the fraction of random points which are located within a distanceof a
Gaia source equal to the separation between the X-ray sourceand the candidate
Gaia counterpart. After removing the counterpartswith a probability of chance overlap greater than 10%, we obtain 88
Gaia candidate counterparts to the Liu XRB sample, most of whichhave reported parallaxes. A complete list of Liu et al. (2006, 2007)catalog sources that were excluded from the final sample and thestep at which they were excluded is found in the online supportinginformation. At this level, only two objects have more than one po-tential
Gaia counterpart: AX J1639.0-4642 and SAX J1711.6-3808.Each of these objects has one potential counterpart with a parallax,and one without. In the case of AX J1639.0-4642, the counterpartwith parallax is the more probable and we retain that parallax for ouranalysis. The opposite is true for SAX J1711.6-3808; as for other ob-jects where the
Gaia counterpart does not have a parallax, it does notfeature in our further analysis. We searched the literature for morerecent distance determinations and any changes to the XRB-typeclassification for all of the 88 matches. We found no strong evidenceto reclassify individual source types but did find a few additionalpublished distances (see Appendix A).Before proceeding, we consider the potential biases of our samplecompared to the unmatched sample. HMXBs have more luminousmain-sequence components and unsurprisingly are more likely tohave a counterpart than LMXBs: there are 187 LMXBs and 114HMXBs in the Liu catalogues, but we find only 33 and 55
Gaia can- Except for 2S 0053+604 (see Appendix A), the counterparts without par-allaxes are faint (18 . < 𝐺 < .
0) and the lack of parallax measurement isconsistent with the distributions given by Gaia Collaboration et al. (2018).MNRAS , 1–16 (2021) aia distances to Galactic X-ray binaries didate counterparts to LMXBs and HMXBs, respectively (these num-bers each decrease by one after removing the extra candidate coun-terparts as described above). Our counterpart matching is also moresensitive to objects that are away from the Galactic centre and awayfrom the Galactic plane - the fraction of objects in the Liu cataloguethat have a Gaia candidate counterpart is higher in those directions.
MNRAS , 1–16 (2021) R . M . A r na s on e t a l . Table 1.
Properties of
Gaia candidate counterparts to Galactic LMXBs
Names RA DEC P interloper
Gaia DR2 ID 𝜃 sep m G , mean GOF d
Gaia d prev d prev Type d prev
Ref" mag kpc kpc
GRO J0422+32/V518 Per 04 21 42.790 +32 54 27.10 0.0100 172650748928103552 0.86 . ± . · · · . ± . · · · · · · · · · · · ·
4U 0614+091/V1055 Ori 06 17 07.400 +09 08 13.60 0.0051 3328832132393159296 0.65 . ± . . + . − . . ± . . + . − . . ± . · · · . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . · · · . ± . . + . − . . ± . -0.0 . + . − . . ± . · · · · · · · · · · · ·
4U 1724-307/Ter 2 17 27 33.300 -30 48 07.00 0.0347 4058208396397618688 0.34 . ± . . + . − . ± . . + . − . ± . . + . − . ± . . + . − . ± . · · · . ± . . + . − . 𝑉 Cadolle Bel et al. (2007)4U 1755-33/V4134 Sgr 17 58 40.000 -33 48 27.00 0.0277 4042473487415175168 0.35 . ± . · · · . ± . . + . − . . ± . -0.1 . + . − . . ± . . + . − . . ± . · · · . ± . . + . − . . ± . . + . − . . ± . . + . − . 𝜃 sep indicates the separation between the candidate Gaia counterpart and the quoted position of the XRB in Liu et al. (2006, 2007). GOF is the
Gaia
DR2goodness-of-fit statistic astrometric_gof_al . M N R A S , ( ) a i ad i s t an ce s t o G a l a c ti c X - r a y b i na r i e s Table 2.
Properties of
Gaia candidate counterparts to Galactic HMXBs
Names RA DEC P interloper
Gaia DR2 ID 𝜃 sep m G , mean GOF d
Gaia d prev d prev Type d prev
Ref" mag kpc kpc
2S 0053+604/gamma Ca 00 56 42.50 +60 43 00.0 0.0041 426558460877467776 0.66 . ± . · · · . ± . . + . − . . ± . . + . − . . ± . -1.1 . + . − . . ± . . + . − . · · · V 0332+53/BQ Cam 03 34 59.90 +53 10 24.0 0.0025 444752973131169664 0.71 . ± . . + . − . . ± . -0.0 . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . 𝑉 Megier et al. (2009)1H 1253-761/HD 109857 12 39 14.60 -75 22 14.0 0.0003 5837600152935767680 0.16 . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . . ± . . + . − . · · · Grillo et al. (1992)IGR J16318-4848/*1 16 31 48.31 -48 49 00.7 0.0002 5940777877435137024 0.04 . ± . . + . − . . ± . . + . − . · · · · · · · · · IGR J16465-4507/- 16 46 35.26 -45 07 04.5 0.0008 5943246345430928512 0.11 . ± . . + . − . . ± . . + . − . . ± . . + . − . 𝑉 Megier et al. (2009)XTE J1739-302/- 17 39 11.58 -30 20 37.6 0.0039 4056922105185686784 0.4 . ± . . + . − . ± . . + . − . . ± . · · · . ± . -2.6 . + . − . . ± . . + . − . 𝑉 Bamba et al. (2001)XTE J1901+014/star 19 01 39.90 +01 26 39.2 0.0045 4268294763113217152 0.6 . ± . . + . − . · · · · · · · · · XTE J1906+09/star 19 04 47.48 +09 02 41.8 0.0014 4310649149314811776 0.23 . ± . . + . − . 𝑉 Marsden et al. (1998)3A 1909+048/SS 433 19 11 49.60 +04 58 58.0 0.0102 4293406612283985024 0.56 . ± . . + . − . . ± . . + . − . 𝑉 Wen et al. (2000)IGR J19140+0951/- 19 14 04.20 +09 52 58.3 0.0056 4309253392325650176 0.41 . ± . . + . − . . ± . . + . − . · · · · · · · · · XTE J1946+274/*A 19 45 39.30 +27 21 55.4 0.0422 2028089540103670144 0.76 . ± . . + . − . ± . -0.0 . + . − . . ± . . + . − . . ± . . + . − . 𝑉 Wilson et al. (2002)RX J2030.5+4751/SAO 49725 20 30 30.80 +47 51 51.0 0.0125 2083644392294059520 0.54 . ± . . + . − . . ± . -0.3 . + . − . · · · Reig et al. (2005)1H 2202+501/BD +49 3718 22 01 38.20 +50 10 05.0 0.0029 1979911002134040960 0.37 . ± . . + . − . . ± . . + . − . 𝜃 sep indicates the separation between the candidate Gaia counterpart and the quoted position of the XRB in Liu et al. (2006, 2007). GOF is the
Gaia
DR2 goodness-of-fit statistic astrometric_gof_al . M N R A S , ( ) R. M. Arnason et al.
To obtain the distance for each counterpart, we match the
Gaia sourceID to the catalogue of Bailer-Jones et al. (2018), which uses aBayesian method to infer distances. In this work, we quote distanceuncertainties as the 1 𝜎 bounds on the posterior probability densityfunction for distance. In general, this function is asymmetric aboutthe peak value, so we have asymmetric error bars. The prior of thisBayesian method models the Galactic stellar density as an exponen-tial disc, so the particular distance prior assumed for each object de-pends on that object’s position in Galactic coordinates. For ordinarystars, information such as line-of-sight extinction, measured 𝑇 eff , andmagnitude/colours in the Gaia filters can provide additional distanceconstraints. However, for XRBs we prefer the position-plus-parallax-only method used by Bailer-Jones et al. (2018), since modelling theexpected value of the additional other parameters in an XRB systemis more complex than for an individual star or ordinary binary.Since LMXBs do not follow the same spatial distribution as thestellar distribution assumed by Bailer-Jones et al. (2018), we mustcautiously interpret the distances to LMXBs (Grimm et al. 2002)(seesubsection 3.1). For example, an exponential disc model would prefersmaller distances for objects along lines of sight that are out ofthe plane of the Milky Way. However, this may not be optimal forLMXBs, given that they can be displaced from the stellar distributionby supernova kicks. Of the matched XRBs, 76 of the Liu cataloguecounterparts have a parallax: 24 of these counterparts are associatedwith LMXBs, while 52 are associated with HMXBs. As the GOFvalues in Tables 1 and 2 show, the astrometric goodness-of-fit is poorfor many of these sources. The
Gaia
DR2 documentation suggeststhat | GOF | > 𝜎 ). A reason for the high GOF and largeexcess noise in these systems could be the orbital motion.Several of the matched objects have a negative measured parallax.In this case, the distance we obtain is dominated by the assumptionsof the prior (see discussion in Luri et al. 2018 and Hogg 2018). Weplot the positions of the Gaia candidate counterparts to the XRBs fora face-on projection of the Milky Way in Figures 1 and 2.
Several expected results are evident in the Galactic distributionsof the XRB sample. First, as shown in Figure 1, HMXBs appearto trace out the nearby (i.e., within 5–8 kpc) arms of the Galaxy.Since HMXB luminosity is correlated with star formation rate instar-forming galaxies (Grimm et al. 2003; Mineo et al. 2012), andspiral arms are the primary sites of star formation, it is reasonableto infer that they should be spatially close to spiral arms. Figure 2shows that the LMXBs are preferentially found in the direction towardthe Galactic centre. A Rayleigh test rejects the null hypothesis thatthe Galactic longitudes of the LMXBs are uniformly distributed at 𝑝 = . × − (and a Kuiper two-sample test rejects the hypothesisthat the Galactic longitude distributions of the LMXBs and HMXBsare drawn from the same distribution at 𝑝 = . Figure 1.
Face-on distribution of
Gaia counterparts for Liu HMXBs. Thespiral arms are modelled using the symmetric spiral arm model of Vallée(2008). Interarm regions are modelled as the symmetric arm model phaseshifted by 45 degrees. Error bars for distance/parallax represent the 1 𝜎 un-certainties. The sun is located at the red star in the middle upper portion ofthe Figure. Figure 2.
Face-on distribution of
Gaia counterparts for Liu LMXBs. The spi-ral arms are modelled using the symmetric spiral arm model of Vallée (2008).Interarm regions are modelled as the symmetric arm model phase shifted by45 degrees. Error bars for distance/parallax represent the 1 𝜎 uncertainties.The sun is located at the red star in the middle upper portion of the Figure.MNRAS000
Gaia counterparts for Liu LMXBs. The spi-ral arms are modelled using the symmetric spiral arm model of Vallée (2008).Interarm regions are modelled as the symmetric arm model phase shifted by45 degrees. Error bars for distance/parallax represent the 1 𝜎 uncertainties.The sun is located at the red star in the middle upper portion of the Figure.MNRAS000 , 1–16 (2021) aia distances to Galactic X-ray binaries Table 3.
Statistical significance of previous distance measurements ofLMXBs comparing to
Gaia
DR2 distances using Bailer-Jones priors.Method mean 95%CI p-valueBurst − . [− . , − . ] . [− . , . ] − . [− . , . ] − . [− . , . ] We find
Gaia parallax measurements for less than one third of thecombined LMXB/HMXB catalogue; in general, parallax measure-ments will not be available for Galactic XRBs. Hence it is usefulto use objects with parallax measurement as a diagnostic for otherdistance methods. We show a comparison of previous distance mea-surements with those derived from the
Gaia candidate counterpartsusing Bland-Altman plots in Figures 3 and 4. These two figures showthe ratio of the difference between previous measured distances and
Gaia
DR2 distances (this work) to the average of the measurementsversus the average of the measurements. To estimate the error bars inthese plots, we only used the uncertainties for our
Gaia counterparts,but omit the uncertainties on previous measurements given the dif-ficulties in comparing methods, instruments, and the fact that manydistances are assumed rather than measured directly. A useful com-ponent of this work is to tabulate distance methods for XRBs with
Gaia candidate counterparts, since in the Liu catalogue, distances arereported without specifying the methodology or whether distancesare measured or assumed.Most previous distance-measurement methods produce distancesconsistent with
Gaia , within their uncertainties where available.
Gaia
DR2 measurements agree well with objects whose parallax hasbeen previously measured either by
Hipparcos or radio interferome-tery (VLBI/VLBA). Radio interferometery parallaxes are expected tobe significantly more accurate than
Gaia ; therefore, comparing withradio parallaxes verifies the assumptions of the
Gaia prior, at least forthe distance ranges and directions where objects with radio parallaxare available. In addition, the mean and median differences betweendistances previously measured photometrically and
Gaia
DR2 dis-tances are consistent with zero.LMXB distances measured using Type I X-ray bursts do show ev-idence of a trend with a plausible physical interpretation. As shownin the left panel of Figure 3, distances measured using Type I X-raybursts are systematically larger than those measured via
Gaia candi-date counterparts. This figure shows the ratio of difference betweenpreviously measured distance and
Gaia distance to the average ofpreviously measured distance and
Gaia distance. The mean of thisquantity over all the methods excluding the Type I X-ray bursts is − . ± .
32, consistent with the null hypothesis of no difference(zero mean) with a 𝑝 -value of 0 .
17. However, the mean for the TypeI X-ray bursts is − . ± .
29 with a 𝑝 -value of 0 .
01, which meanswe can reject the null hypothesis of no systematic difference be-tween measured distances from
Gaia and distances from Type I X-raybursts. The statistical significances of comparing other methods usedin measuring LMXB distances with respect to Gaia distances are re-ported in Table 3. In this table, means and 95% confidence-intervalsof differences are reported, and the p-values are calculated for thenull-hypothesis of zero means. According to this table, distancesmeasured using Type I X-ray bursts are the only method which showa systematic difference with new measured distances in this work.These results would seem to suggest that Type I X-ray bursts are intrinsically less luminous than predicted by modelling. This agreeswith previous results on systematic biases in distance determinationvia Type I X-ray bursts. Galloway et al. (2008b) demonstrated thatthe choice to assume that the touchdown flux (the flux measuredwhen the expanded photosphere of the neutron star touches downback onto its surface) is either at the Eddington luminosity or sub-Eddington may introduce large systematic uncertainties to distancemeasurements of X-ray bursting XRBs. Studies of bursting sourcesusing the Rossi X-ray Timing Explorer have indicated that a numberof these sources are significantly sub-Eddington in their peak fluxes(e.g., Galloway et al. 2008a).As mentioned in subsection 2.4, LMXBs do not follow the samespatial distribution as the stellar distribution assumed by Bailer-Joneset al. (2018). To investigate the effect of the priors on LMXB dis-tances, we also measured the
Gaia
DR2 distances using the priordeveloped by Atri et al. (2019), which considers the distribution ofLMXBs in the Milky Way based on the work of (Grimm et al. 2002).The right panel of Figure 3 shows the result of our distance com-parisons using this prior. The most noticeable effect of using theAtri et al. (2019) prior is an increase in distances for most of theLMXBs. As a result, distances measured using Type I X-ray burstsare systematically smaller than those measured via
Gaia candidatecounterparts. This is in contradiction with the suggestive results ofour analysis using the Bailer-Jones et al. (2018) prior, which indicatedthat distances based on type I X-ray bursts are overestimated, and thusthat bursts only reach 0.5 L
Eddington . As a model-independent check,we have also done the same analysis without using any prior. In thisscenario, the Type I X-ray burst distances are consistent with the nullhypothesis of no difference with the
Gaia distances, with a 𝑝 -valueof 0 .
19. This discrepancy highlights the importance of priors whenusing
Gaia data for sources with large uncertainties.Individual objects with particularly large discrepancies betweenpreviously-published and
Gaia candidate counterpart distances arediscussed in Appendix B.
To investigate the relationship between XRBs and Galactic structure,we compare the XRB distributions to a model of the spiral arms ofthe Milky Way. Pettitt et al. (2020) show evidence for spiral structuretraced by young stars, though it is not clear what the precise phys-ical properties of the spiral structure are. Gorski & Barmby (2020)suggest a four arm spiral structure traced by maser-bearing evolvedstars. We use the symmetric arm model of Vallée (2008). This modelis analytically defined: the precise shape, symmetries, structure, andextent of the spiral arms of the Galaxy are nontrivial to determine dueto our location within the Milky Way. This symmetric model is fittedto agree with a variety of observations, including dust, HI gas, COgas, and maps of stellar velocities. This model defines the midpointof four identical arms phase shifted by 90 ◦ . We further define themidpoint of interarm regions by shifting the existing arms by 45 ◦ .For each XRB, we compute three properties:(i) the two-dimensional distance to the nearest spiral arm for aface-on projection(ii) whether the XRB is leading or trailing its closest spiral arm(iii) whether the XRB is closer to the midpoint of a spiral arm orthe midpoint of an interarm regionGiven that many of the uncertainties for the distances quite large,counts of these quantities depend strongly on the posterior distribu-tion function of the distances. In order to assess how much thesequantities change, we create 10,000 realizations of the distance for MNRAS , 1–16 (2021) R. M. Arnason et al.
Figure 3.
Difference between literature distance measurements for Liu catalogue LMXBs and the distances obtained in this work, versus the average of thesedistances. Error bars are lower limits because the uncertainties in the previous distance measurements were omitted. The horizontal yellow lines show the meanand 95% confidence interval of the sources with previously measured distances using Type I X-ray burst. Left panel:
Gaia
DR2 distances calculated using theprior of Bailer-Jones et al. (2018). Right panel:
Gaia
DR2 distances calculated using the prior of Atri et al. (2019). Type I X-ray burst distances show a systematicoverestimate with respect to the Gaia distances using the Bailer-Jones prior, and a systematic underestimate with respect to the Gaia distances that use the Atriet al. priors. The comparison using the prior of Bailer-Jones et al. (2018) implies that X-ray bursts (assuming they are PRE bursts) only reach 0.5 L
Eddington . Figure 4.
Difference between previous distance measurements for Liu cat-alogue HMXBs and the distances obtained in this work, versus the averageof these distances. Previous distances are obtained from the literature refer-ence given in Table 2, while the distance in this work is the distance to the
Gaia candidate counterpart for each HMXB. each object using the posterior distribution function defined in Bailer-Jones et al. (2018), and compute the three quantities above for eachobject in each iteration.After computing whether each object is closer to an arm or inter-arm region, whether it is leading or trailing the nearest spiral arm,and the distance to the nearest spiral arm, we calculate the frac-tion of objects leading/trailing and fraction of objects close to anarm/interarm for each of the 10,000 runs. Under this construction,since we have effectively partitioned the galaxy into two equally-sizedregions (closer to arm/closer to interarm, leading/trailing the near-est spiral arm), we expect the following for the distribution of thesefractions: If the distribution of LMXBs/HMXBs fractions peaks at avalue greater than 0.5 for a particular structure (arm/interarm/leadingedge/trailing edge), then we interpret that LMXBs/HMXBs as beingcorrelated with that structure. Conversely, if the distribution peaksat a value less than 0.5, we interpret LMXBs/HMXBs as being anti-correlated with that structure. If the distribution peaks at 0.5, weinterpret LMXBs/HMXBs as being uncorrelated with that structure.We treat the uncorrelated case as the null hypothesis for LMXBs andHMXBs individually.In each run, we exclude from the fraction any object that lies ata distance of less than 3.1 kpc from the Galactic centre, classifyingthem separately as bulge sources. We choose 3.1 kpc because it isgiven as the half-length of the bar superimposed on the cartographicplots of Vallée (2008), and it is noted therein that it becomes difficultto separate the beginnings of the spiral arms from the bar itself atapproximately this distance. In each run, on average two HMXBs andfive LMXBs were classified as bulge sources. The resulting fractionsand their uncertainty distributions are plotted in Figure 5.Across the simulation, LMXBs and HMXBs both appear to ex-hibit a roughly normal distribution in both fractions, though in both
MNRAS000
MNRAS000 , 1–16 (2021) aia distances to Galactic X-ray binaries Figure 5.
Distributions of population fraction correlated with spiral arms and leading edges for 10,000 realizations of the XRBs with
Gaia candidate counterparts.The vertical line marks a fraction of 0.5, where the populations would be interpreted as being uncorrelated with the structure.
Figure 6.
Cumulative distribution of XRB distances to the nearest spiralarm. We mark the characteristic clustering scales of HMXBs against OBassociations and star-forming complexes measured by Bodaghee et al. (2012)and Coleiro & Chaty (2013) for reference using the vertical grey regions. Wealso plot the 0.1, 0.5, 0.9, and 0.99 quantiles of the HMXB distribution forcomparison. the leading/trailing or arm/interarm case, the LMXB distributionpossesses a larger spread. To compare these measurements to eachother and to the null hypothesis (that they are uncorrelated witharms/interarms and leading/trailing spiral arms), we tested these un-certainty distributions for normality. Since the interarm/trailing frac-tion is complementary to the arm/leading fraction, we consider onlythe arm/leading fractions. None of the four distributions is consid-ered normal by the D’Agostini 𝐾 test or the Anderson-Darling test at 𝑝 = .
05. Only the LMXB leading fraction is considered normal bythe Jarque-Bera test for 𝑝 = .
05. Since the distributions are not trulynormal, we report the fraction measurements and their uncertaintyin two ways: first using the standard deviation as the 1 𝜎 uncertainty,and then reporting the the 95th/5th quantiles as the uncertainty.Our measurements of the fraction of HMXBs/LMXBs that arecorrelated with spiral arms/inter-arm regions yields the followingresults: • Fraction of HMXBs that are closer to a spiral arm: 0 . ± . 𝜎 ), 0 . + . − . at the 95th and 5th quantiles • Fraction of LMXBs that are closer to a spiral arm: 0 . ± . 𝜎 ), 0 . + . − . at the 95th and 5th quantiles • Fraction of HMXBs that are leading the nearest spiral arm:0 . ± .
05 (at 1 𝜎 ), 0 . + . − . at the 95th and 5th quantiles • Fraction of LMXBs that are leading the nearest spiral arm:0 . ± .
10 (at 1 𝜎 ), 0 . + . − . at the 95th and 5th quantilesWe cannot reject the null hypothesis that HMXBs or LMXBs arespatially uncorrelated with spiral arms, at even 1 𝜎 , since the un-certainties overlap with 𝐹 fraction = .
5. We cannot reject the nullhypothesis for either HMXBs or LMXBs exhibiting no preferenceleading or trailing their nearest spiral arm. The LMXB and HMXBfractions also overlap with each other at the 1 𝜎 level. LMXBs ex- MNRAS , 1–16 (2021) R. M. Arnason et al. hibit a mild preference for being found in interarm regions, whileHMXBs show only a mild preference for being found in the spiralarms. LMXBs appear to be uncorrelated with leading or trailing theirspiral arm, while at low significance the HMXBs appear to prefertrailing their nearest spiral arm.In the context of Galactic structure, previous work has shown thatHMXBs trace SFR on Galactic scales (Grimm et al. 2003), so it isreasonable to expect they should trace it on resolved scales in somefashion and should exhibit a distinct spatial correlation. Naively it canbe assumed that star formation should happen at the leading edge ofa spiral arm where the gas accumulates (see Koda et al. 2012 for M51as an illustrative example of star formation and its relation to spiralarm structure). Taking these assumptions together, HMXBs shouldbe found at the leading edge of spiral arms, and should exhibit a strongpreference for spiral arms versus interarm regions. However, we findonly a mild preference for spiral arms: the distribution of fractionsfor the simulation peaks at 54% of the HMXBs being closer to anarm than an interarm region, but the wings of the distribution includethe uncorrelated and anti-correlated cases.The lack of strong preference for HMXBs being closely associatedwith spiral arms could have a number of possible implications: • Star formation does not occur at the leading edge of spiral arms. • The time delay between star formation and HMXB accretionstarting manifests itself as a spatial separation between the spiral armand HMXBs due to the pattern speed of spiral arms. • HMXB natal kicks may be larger than expected. • Our sample is not large enough and does not have sufficientlysmall distance uncertainties as an ensemble to measure the correla-tion we expect from first principles.Our HMXB sample comprises only ∼
50 objects, and the uncertaintiesare still substantial. As such, though we can rule out a very strongspatial correlation or anti-correlation between HMXBs and spiralarms (using the Gaia DR2 data specifically), we cannot use ourresult to distinguish between the scenarios listed above. Since weare unable to reject the null hypothesis that HMXBs are uncorrelatedwith spiral arms, our result is consistent with Bodaghee et al. (2012)’sanalysis, which found that HMXBs are not spatially correlated withspiral arms. The scale at which the HMXB/SFR correlation breaksdown (if at all) is not well-constrained. In nearby galaxies, the X-raysources are typically studied by considering the integrated propertiesof the entire population (for example, X-ray luminosity function) andcomparing to global parameters of the galaxy. Correlating XRBswith galactic structure is challenging since galaxies that are closeenough to resolve on the desired scales require many fields in orderto encompass the entire galaxy. In addition, contamination fromX-ray sources in front of or behind the galaxy creates additionaldifficulties. Swartz et al. (2003) investigated the relationship betweenthe spiral arms of M81 and its X-ray source population, finding strongcorrelation between spiral arm position and X-ray source density.They note that brighter sources tend to be closer to spiral arms,attributable to the brightest and shortest-lived HMXBs being closedownstream from their spiral arms. More recently, Kuntz et al. (2016)performed a deep
Chandra survey of M51. This study also finds thatX-ray sources are concentrated in spiral arms, though the distancesto spiral arm midpoints are not presented. Both studies also founda non-trivial population of supernova remnants contributing to thetotal X-ray source population.In contrast to HMXBs, we expect that LMXBs should exhibit nostrong preference for spiral arms; they represent (collectively) anolder population that is also more strongly perturbed by the strengthof its SN kicks (Grimm et al. 2002). Since LMXBs can be much older, it is not expected or required that they are still near the spiral armthat formed their progenitor – there may have been multiple Galacticrotations since the LMXB itself formed. Additionally, LMXBs’ highvelocity kicks mean they can be substantially displaced from the star-forming region where they initially formed. This process is alreadyrequired to explain the presence of LMXBs at high Galactic latitudeswhere they would not be expected to form a priori due to the lowstellar density (see, for example, Repetto et al. 2012). Consequently,LMXBs as a population should be uncorrelated with spiral arms sincetheir distribution would be unperturbed by either the presence orabsence of spiral arms. This makes our result, which shows LMXBsanti-correlated with spiral arms (though at low significance), difficultto explain.We also computed the distribution of distances to the nearest spi-ral arm across all the simulations, shown in Figure 6, in order tocompare with previous works that measured the distances to OB as-sociations and star-forming complexes for HMXBs (Bodaghee et al.2012; Coleiro & Chaty 2013). In these works, clustering distancesbetween HMXBs and SFCs/OB associations were inferred from thecritical points of the cumulative distribution of the distances to thenearest SFC/OB association. As discussed in Section 1.1, distancesto OB associations and SF complexes are distinct from distances tothe spiral arms themselves, and as such we might not expect HMXBsto have the same clustering distance to the spiral arm. The distribu-tion of HMXB distances to the nearest spiral arm that we measuredoes not show a strong preference for the clustering sizes measuredfor OB associations or SF complexes in previous works, though wenote that the Vallée (2008) model does not fit spiral arms to either ofthese structures. The 0.1, 0.5, 0.9, and 0.99 quantiles of the HMXBdistribution to be at 127, 570, 1296, and 2340 pc, respectively. Forthe LMXB distribution, the 0.1, 0.5, 0.9, and 0.99 quantiles of theLMXB distribution are at 130, 610, 1090, and 1780 pc, respectively.Given the substantial width of these distributions, it is difficultto determine a characteristic separation from the spiral arms. Theinterarm separation of a few kpc as set by the symmetric arms modelmeans that, by construction, it is difficult to have an XRB more thana few kpc away from a spiral arm in face-on projection. Further, wehave chosen to model the galaxy using a symmetric model fitted toobservables in the Milky Way, which is a simple albeit potentiallyunrealistic choice. The primary advantage of this model is that itpermits us to easily define inter-arm and arm regions for analysisof the locations of XRBs. In reality, the number of arms and thesymmetry (e.g., are the four arms symmetric with each other orare there major/minor axes?) of these arms in the Milky Way isdifficult to characterize (see Vallée 2017 and references therein),and discussion exists about which tracers to use and how far toproject the model based on nearby observables. Future attempts tocharacterize the relationship between the Galaxy’s spiral arms andits XRB population would be improved by the use of a model thatrelaxes the symmetry constraint. • We have assembled the largest sample of Galactic X-ray binarieswhose distances have all been measured using the same method, andhence have the same systematics and uniform presumed biases. • Comparing XRB distances measured by
Gaia (using the Bailer-Jones prior) to previous methods shows that measuring distancesusing Type I X-ray bursts appears to systematically overestimate dis-tance. This suggests that assumptions about X-ray bursts, namely thatbursting neutron stars consistently reach the Eddington luminosity,
MNRAS000
MNRAS000 , 1–16 (2021) aia distances to Galactic X-ray binaries may need to be modified to use X-ray bursts as a distance estimator.This effect is prior-dependent, as choosing a different prior, such asthe one in Atri et al. (2019), can cause burst distances to be system-atically lower than those from Gaia
DR2. • We have compared the positions of XRBs to the locations ofthe midpoints of spiral arms in the Milky Way. Galactic HMXBsin our sample show only a modest preference for being spatiallyco-located with spiral arms versus interarm regions, and show onlya modest preference for being on the leading edge of spiral arms.This suggests that the delay time between star formation and HMXBformation/accretion beginning manifests itself observationally as aspatial separation between HMXBs and spiral arms due to the patternspeed of spiral arm rotation. Other possible explanations for thiseffect are scattering due to natal HMXB kicks or the possibility ofstar formation occurring closer to the midpoint of the arm than theleading edge. • We further find that HMXB distances to the nearest spiral armdo not show a strong preference for the clustering sizes previouslyobserved for OB associations or SF complexes. • We find that LMXBs are very weakly anti-correlated with spiralarms. This disagrees with the expectation that LMXBs should beuncorrelated with spiral arms, though we note that the significanceof this result is low.A main source of uncertainty in our analysis is the low num-ber of XRBs with
Gaia counterparts. Further releases of
Gaia willhopefully yield additional
Gaia candidate counterparts for GalacticXRBs, particularly for the intrinsically optically fainter LMXBs. Forobjects with identified
Gaia candidate counterparts, smaller distanceuncertainties are expected from the improved baseline in DR3 andsubsequent releases. The small sample size from the Liu cataloguesis another limitation of our analysis. The
Chandra
Source Catalog(Evans et al. 2010) provides an excellent foundation for studying theGalactic X-ray sky in the
Chandra era, but at present it has not beendata-mined to make a Milky Way-specific catalogue as a potentialsuccessor to the Liu catalogues. Our knowledge of the Galactic XRBsource population can be improved through future all-sky surveys,such as with the newly-launched eROSITA mission (Merloni et al.2012). This mission, designed as a successor to the ROSAT mis-sion, will survey the sky at approximately 20 times the sensitivityof ROSAT in soft X-rays (0.5–2.0 keV), while providing the firstimaging survey of the sky in hard X-rays (2–10 keV). The on-axisangular resolution of this telescope is expected to be comparable tothat of
XMM-Newton . An improved all-sky survey will allow us tofind
Gaia counterparts to an X-ray catalogue that is more up-to-dateand is has more uniform systematics, enhancing our understandingof how XRB positions correlate with Galactic structure.
ACKNOWLEDGEMENTS
We thank the referee, P. Gandhi, for constructive and helpful re-ports. This work has made use of data from the European SpaceAgency (ESA) mission
Gaia ( ), processed by the Gaia
Data Processing and Analysis Consor-tium (DPAC, ). Funding for the DPAC has been provided by nationalinstitutions, in particular the institutions participating in the
Gaia
Multilateral Agreement. R. M. A. acknowledges support from anNSERC CGS-D scholarship. P. B. acknowledges support from anNSERC Discovery Grant and the hospitality of the Rotman Institutefor Philosophy. We thank E. Cackett, S. Gallagher, and T.A.A. Sigutfor helpful discussions.
DATA AVAILABILITY
The data underlying this article are available in the article and in itsonline supplementary material.
REFERENCES
Arenou F., et al., 2018, A&A, 616, A17Atri P., et al., 2019, MNRAS, 489, 3116Augusteijn T., van der Hooft F., de Jong J. A., van Kerkwijk M. H., vanParadijs J., 1998, A&A, 332, 561Bahramian A., Heinke C. O., Degenaar N., Chomiuk L., Wijnands R., StraderJ., Ho W. C. G., Pooley D., 2015, MNRAS, 452, 3475Bailer-Jones C. A. L., Rybizki J., Fouesneau M., Mantelet G., Andrae R.,2018, AJ, 156, 58Bamba A., Yokogawa J., Ueno M., Koyama K., Yamauchi S., 2001, PASJ,53, 1179Basinska E. M., Lewin W. H. G., Sztajno M., Cominsky L. R., Marshall F. J.,1984, ApJ, 281, 337Bassa C. G., Jonker P. G., in’t Zand J. J. M., Verbunt F., 2006, A&A, 446,L17Baumgartner W. H., Tueller J., Markwardt C. B., Skinner G. K., BarthelmyS., Mushotzky R. F., Evans P. A., Gehrels N., 2013, ApJS, 207, 19Belokurov V., et al., 2020, MNRAS, 496, 1922Berghoefer T. W., Schmitt J. H. M. M., Cassinelli J. P., 1996, Astronomy andAstrophysics Supplement Series, 118, 481Bhattacharyya S., 2010, Advances in Space Research, 45, 949Blay P., Negueruela I., Reig P., Coe M. J., Corbet R. H. D., Fabregat J.,Tarasov A. E., 2006, A&A, 446, 1095Bodaghee A., Tomsick J. A., Rodriguez J., James J. B., 2012, ApJ, 744, 108Boroson B., Kim D.-W., Fabbiano G., 2011, ApJ, 729, 12Boubert D., Guillochon J., Hawkins K., Ginsburg I., Evans N. W., Strader J.,2018, MNRAS, 479, 2789Bozzo E., et al., 2018, A&A, 613, A22Bradshaw C. F., Fomalont E. B., Geldzahler B. J., 1999, ApJ, 512, L121Brandt S., Castro-Tirado A. J., Lund N., Dremin V., Lapshov I., Sunyaev R.,1992, A&A, 262, L15Brown A. G. A., Blaauw A., Hoogerwerf R., de Bruijne J. H. J., de ZeeuwP. T., 1999, in Lada C. J., Kylafis N. D., eds, NATO Advanced ScienceInstitutes (ASI) Series C Vol. 540, NATO Advanced Science Institutes(ASI) Series C. p. 411 ( arXiv:astro-ph/9902234 )Butters O. W., Norton A. J., Mukai K., Tomsick J. A., 2011, A&A, 526, A77Cadolle Bel M., et al., 2007, ApJ, 659, 549Cantrell A. G., et al., 2010, ApJ, 710, 1127Casares J., Ribó M., Ribas I., Paredes J. M., Martí J., Herrero A., 2005,MNRAS, 364, 899Casares J., Jonker P. G., Israelian G., 2017, in Alsabti A. W., Murdin P.,eds, , Handbook of Supernovae. Springer International Publishing AG,p. 1499, doi:10.1007/978-3-319-21846-5_111Chakrabarty D., Roche P., 1997, ApJ, 489, 254Chan V. C., Bovy J., 2020, MNRAS, 493, 4367Chernyakova M., Lutovinov A., Rodríguez J., Revnivtsev M., 2005, MNRAS,364, 455Chevalier C., Ilovaisky S. A., 1998, A&A, 330, 201Clark G. W., 1975, ApJ, 199, L143Clark G. W., 2004, ApJ, 610, 956Coe M. J., Everall C., Norton A. J., Roche P., Unger S. J., Fabregat J., RegleroV., Grunsfeld J. M., 1993, MNRAS, 261, 599Coe M. J., et al., 1994, MNRAS, 270, L57Coleiro A., Chaty S., 2013, ApJ, 764, 185Coley J. B., Corbet R. H. D., Krimm H. A., 2015, ApJ, 808, 140Corbet R. H. D., Mason K. O., 1984, A&A, 131, 385Corral-Santana J. M., Casares J., Muñoz-Darias T., Bauer F. E., Martínez-PaisI. G., Russell D. M., 2016, A&A, 587, A61Cowley A. P., Schmidtke P. C., 1990, AJ, 99, 678Dabringhausen J., Kroupa P., Pflamm-Altenburg J., Mieske S., 2012, ApJ,747, 72 MNRAS , 1–16 (2021) R. M. Arnason et al.
Degenaar N., Wijnands R., Cackett E. M., Homan J., in’t Zand J. J. M.,Kuulkers E., Maccarone T. J., van der Klis M., 2012, A&A, 545, A49Dhawan V., Mirabel I. F., Ribó M., Rodrigues I., 2007, ApJ, 668, 430Erkal D., Boubert D., Gualandris A., Evans N. W., Antonini F., 2019, MN-RAS, 483, 2007Evans I. N., et al., 2010, ApJS, 189, 37Filliatre P., Chaty S., 2004, ApJ, 616, 469Foellmi C., Depagne E., Dall T. H., Mirabel I. F., 2006, A&A, 457, 249Foight D. R., Güver T., Özel F., Slane P. O., 2016, ApJ, 826, 66Fortin F., Chaty S., Sander A., 2020, ApJ, 894, 86Gaia Collaboration et al., 2016a, A&A, 595, A1Gaia Collaboration et al., 2016b, A&A, 595, A2Gaia Collaboration et al., 2018, A&A, 616, A1Gaia Collaboration Brown A. G. A., Vallenari A., Prusti T., de Bruijne J.H. J., Babusiaux C., Biermann M., 2020, A&A, in pressGalloway D. K., Psaltis D., Chakrabarty D., Muno M. P., 2003, ApJ, 590, 999Galloway D. K., Muno M. P., Hartman J. M., Psaltis D., Chakrabarty D.,2008a, ApJS, 179, 360Galloway D. K., Özel F., Psaltis D., 2008b, MNRAS, 387, 268Gandhi P., Rao A., Johnson M. A. C., Paice J. A., Maccarone T. J., 2019,MNRAS, 485, 2642Gelino D. M., 2001, PhD thesis, Center for Astrophysics and Space Sciences,University of California, San DiegoGelino D. M., Harrison T. E., 2003, ApJ, 599, 1254Gilfanov M., 2004, MNRAS, 349, 146González Hernández J. I., Rebolo R., Peñarrubia J., Casares J., Israelian G.,2005, A&A, 435, 1185Gorski M. D., Barmby P., 2020, MNRAS, 495, 726Green G. M., Schlafly E., Zucker C., Speagle J. S., Finkbeiner D., 2019, ApJ,887, 93Grillo F., Sciortino S., Micela G., Vaiana G. S., Harnden F. R. J., 1992, ApJS,81, 795Grimm H.-J., Gilfanov M., Sunyaev R., 2002, A&A, 391, 923Grimm H.-J., Gilfanov M., Sunyaev R., 2003, MNRAS, 339, 793Grindlay J. E., Petro L. D., McClintock J. E., 1984, ApJ, 276, 621Hakala P., Ramsay G., Muhli P., Charles P., Hannikainen D., Mukai K., VilhuO., 2005, MNRAS, 356, 1133Hjellming R. M., Johnston K. J., 1981, ApJ, 246, L141Hjellming R. M., Rupen M. P., 1995, Nature, 375, 464Hogg D. W., 2018, preprint, ( arXiv:1804.07766 )Hutchings J. B., Cowley A. P., Crampton D., van Paradijs J., White N. E.,1979, ApJ, 229, 1079Ilovaisky S. A., Chevalier C., Motch C., 1982, A&A, 114, L7Janka H.-T., 2013, MNRAS, 434, 1355Janot-Pacheco E., Ilovaisky S. A., Chevalier C., 1981, A&A, 99, 274Jonker P. G., Nelemans G., 2004, MNRAS, 354, 355Jonker P. G., Galloway D. K., McClintock J. E., Buxton M., Garcia M.,Murray S., 2004, MNRAS, 354, 666Kaaret P., Piraino S., Halpern J., Eracleous M., 1999, ApJ, 523, 197Kaluzienski L. J., Holt S. S., Swank J. H., 1980, ApJ, 241, 779Kawai N., Suzuki M., 2005, The Astronomer’s Telegram, 534, 1Koda J., et al., 2012, ApJ, 761, 41Krimm H. A., et al., 2013, ApJS, 209, 14Krivonos R., Tsygankov S., Revnivtsev M., Grebenev S., Churazov E., Sun-yaev R., 2010, A&A, 523, A61Krzeminski W., 1974, ApJ, 192, L135Kuntz K. D., Long K. S., Kilgard R. E., 2016, ApJ, 827, 46Kuulkers E., den Hartog P. R., in’t Zand J. J. M., Verbunt F. W. M., HarrisW. E., Cocchi M., 2003, A&A, 399, 663La Parola V., Cusumano G., Romano P., Segreto A., Vercellone S., ChincariniG., 2010, MNRAS, 405, L66Leahy D. A., 2002, A&A, 391, 219Lehmer B. D., Alexander D. M., Bauer F. E., Brandt W. N., Goulding A. D.,Jenkins L. P., Ptak A., Roberts T. P., 2010, ApJ, 724, 559Lewin W. H. G., van Paradijs J., Taam R. E., 1993, Space Sci. Rev., 62, 223Lin D., Webb N. A., Barret D., 2012, ApJ, 756, 27Linares M., Shahbaz T., Casares J., 2018, ApJ, 859, 54Lindegren L., et al., 2018, A&A, 616, A2 Liu Q. Z., van Paradijs J., van den Heuvel E. P. J., 2006, A&A, 455, 1165Liu Q. Z., van Paradijs J., van den Heuvel E. P. J., 2007, A&A, 469, 807Luri X., et al., 2018, A&A, 616, A9Lyuty V. M., Za˘itseva G. V., 2000, Astronomy Letters, 26, 9MacDonald R. K. D., et al., 2014, ApJ, 784, 2Marsden D., Gruber D. E., Heindl W. A., Pelling M. R., Rothschild R. E.,1998, ApJ, 502, L129Masetti N., et al., 2002, A&A, 382, 104Masetti N., et al., 2006a, A&A, 449, 1139Masetti N., Orlandini M., Palazzi E., Amati L., Frontera F., 2006b, A&A,453, 295Masetti N., et al., 2006c, A&A, 455, 11Masetti N., et al., 2009, A&A, 495, 121Mason K. O., Cordova F. A., 1982, ApJ, 262, 253Massey P., Johnson K. E., Degioia-Eastwood K., 1995, ApJ, 454, 151McBride V. A., et al., 2006, A&A, 451, 267McClintock J. E., Remillard R. A., Margon B., 1981, ApJ, 243, 900Megier A., Strobel A., Galazutdinov G. A., Krełowski J., 2009, A&A, 507,833Merloni A., et al., 2012, preprint, ( arXiv:1209.3114 )Miller-Jones J. C. A., Jonker P. G., Dhawan V., Brisken W., Rupen M. P.,Nelemans G., Gallo E., 2009, ApJ, 706, L230Mineo S., Gilfanov M., Sunyaev R., 2012, MNRAS, 419, 2095Motch C., Haberl F., Dennerl K., Pakull M., Janot-Pacheco E., 1997, A&A,323, 853Muñoz-Darias T., Casares J., Martínez-Pais I. G., 2005, ApJ, 635, 502Muno M. P., Chakrabarty D., Galloway D. K., Savov P., 2001, ApJ, 553, L157Muno M. P., Pfahl E., Baganoff F. K., Brandt W. N., Ghez A., Lu J., MorrisM. R., 2005, ApJ, 622, L113Nättilä J., Miller M. C., Steiner A. W., Kajava J. J. E., Suleimanov V. F.,Poutanen J., 2017, A&A, 608, A31Negueruela I., Roche P., Buckley D. A. H., Chakrabarty D., Coe M. J.,Fabregat J., Reig P., 1996, A&A, 315, 160Negueruela I., Roche P., Fabregat J., Coe M. J., 1999, MNRAS, 307, 695Negueruela I., Smith D. M., Harrison T. E., Torrejón J. M., 2006, ApJ, 638,982Parkes G. E., Murdin P. G., Mason K. O., 1980, MNRAS, 190, 537Pellizza L. J., Chaty S., Negueruela I., 2006, A&A, 455, 653Perryman M. A. C., et al., 1997, A&A, 323, L49Pettitt A. R., Ragan S. E., Smith M. C., 2020, MNRAS, 491, 2162Phillips S. N., Shahbaz T., Podsiadlowski P., 1999, MNRAS, 304, 839Podsiadlowski P., Rappaport S., 2000, ApJ, 529, 946Polcaro V. F., et al., 1990, A&A, 231, 354Pooley D., et al., 2003, ApJ, 591, L131Prišegen M., 2019, A&A, 621, A37Reid M. J., McClintock J. E., Narayan R., Gou L., Remillard R. A., OroszJ. A., 2011a, ApJ, 742, 83Reid M. J., McClintock J. E., Narayan R., Gou L., Remillard R. A., OroszJ. A., 2011b, ApJ, 742, 83Reig P., Fabregat J., 2015, A&A, 574, A33Reig P., Chakrabarty D., Coe M. J., Fabregat J., Negueruela I., Prince T. A.,Roche P., Steele I. A., 1996, A&A, 311, 879Reig P., Negueruela I., Papamastorakis G., Manousakis A., Kougentakis T.,2005, A&A, 440, 637Reig P., Zezas A., Gkouvelis L., 2010, A&A, 522, A107Repetto S., Davies M. B., Sigurdsson S., 2012, MNRAS, 425, 2799Reynolds A. P., Bell S. A., Hilditch R. W., 1992, MNRAS, 256, 631Reynolds A. P., Quaintrell H., Still M. D., Roche P., Chakrabarty D., LevineS. E., 1997, MNRAS, 288, 43Sadakane K., Hirata R., Jugaku J., Kondo Y., Matsuoka M., Tanaka Y.,Hammerschlag-Hensberge G., 1985, ApJ, 288, 284Samus’ N. N., Kazarovets E. V., Durlevich O. V., Kireeva N. N., PastukhovaE. N., 2017, Astronomy Reports, 61, 80Smith D. M., 2004, The Astronomer’s Telegram, 338, 1Steele I. A., Negueruela I., Coe M. J., Roche P., 1998, MNRAS, 297, L5Steiner A. W., Heinke C. O., Bogdanov S., Li C. K., Ho W. C. G., BahramianA., Han S., 2018, MNRAS, 476, 421MNRAS , 1–16 (2021) aia distances to Galactic X-ray binaries Stevens J. B., Reig P., Coe M. J., Buckley D. A. H., Fabregat J., Steele I. A.,1997, MNRAS, 288, 988Strohmayer T., Bildsten L., 2006, in Lewin W. H. G., van der Klis M., eds,Compact stellar X-ray sources. Cambridge Astrophysics Series. Cam-bridge University Press, pp 113–156, doi:10.2277/0521826594Swartz D. A., Ghosh K. K., McCollough M. L., Pannuti T. G., Tennant A. F.,Wu K., 2003, ApJS, 144, 213Tetarenko B. E., Sivakoff G. R., Heinke C. O., Gladstone J. C., 2016, ApJS,222, 15Thévenin F., Falanga M., Kuo C. Y., Pietrzyński G., Yamaguchi M., 2017,Space Sci. Rev., 212, 1787Tomsick J. A., Gelino D. M., Kaaret P., 2005, ApJ, 635, 1233Torrejón J. M., Orr A., 2001, A&A, 377, 148Torrejón J. M., Negueruela I., Smith D. M., Harrison T. E., 2010, A&A, 510,A61Tremmel M., et al., 2013, ApJ, 766, 19Tsygankov S. S., Lutovinov A. A., 2005, Astronomy Letters, 31, 88Vallée J. P., 2008, AJ, 135, 1301Vallée J. P., 2014, AJ, 148, 5Vallée J. P., 2017, The Astronomical Review, 13, 113Verbunt F., 2003, in Piotto G., Meylan G., Djorgovski S. G., RielloM., eds, Astronomical Society of the Pacific Conference SeriesVol. 296, New Horizons in Globular Cluster Astronomy. p. 245( arXiv:astro-ph/0210057 )Verbunt F., Hut P., 1987, in Helfand D. J., Huang J.-H., eds, IAU SymposiumVol. 125, The Origin and Evolution of Neutron Stars. p. 187Verbunt F., Lewin W. H. G., 2006, in Lewin W. H. G., van der Klis M., eds,Compact stellar X-ray sources. Cambridge Astrophysics Series. Cam-bridge University Press, pp 341–379, doi:10.2277/0521826594Verbunt F., van Paradijs J., Elson R., 1984, MNRAS, 210, 899Wachter S., Smale A. P., 1998, ApJ, 496, L21Wen L., Remillard R. A., Bradt H. V., 2000, ApJ, 532, 1119Wenger M., et al., 2000, A&AS, 143, 9Wilson C. A., Finger M. H., Coe M. J., Laycock S., Fabregat J., 2002, ApJ,570, 287Wilson C. A., Finger M. H., Coe M. J., Negueruela I., 2003, ApJ, 584, 996Zhang Z., Gilfanov M., Bogdán Á., 2012, A&A, 546, A36in’t Zand J. J. M., et al., 2002, A&A, 389, L43in’t Zand J. J. M., Cumming A., van der Sluys M. V., Verbunt F., Pols O. R.,2005, A&A, 441, 675van Paradijs J., 1978, Nature, 274, 650van Paradijs J., 1981, A&A, 101, 174van Paradijs J., 1998, in Buccheri R., van Paradijs J., Alpar A., eds, NATOAdvanced Science Institutes (ASI) Series C Vol. 515, NATO AdvancedScience Institutes (ASI) Series C. p. 279 ( arXiv:astro-ph/9802177 )van Paradijs J., White N., 1995, ApJ, 447, L33
APPENDIX A: UPDATED DISTANCES ANDCLASSIFICATIONS OF XRBS WITH
Gaia
COUNTERPARTS
For the 88
Gaia candidate counterparts to the Liu XRB sample(see §2.3), we searched the literature for more recently-publisheddistances and compilations (Tetarenko et al. 2016; Corral-Santanaet al. 2016; Wenger et al. 2000) for updates to classifications.Two Liu catalogue objects have controversial classifications butdo not figure in our analysis because their
Gaia
DR2 counterpartshave no parallax. SIMBAD notes that the nature of 2S 0053+604 ( 𝛾 Cas) as an X-ray binary is controversial (for a summary, see Prišegen2019). Although the star itself is in
Gaia
DR2, its bright magnitude( 𝐺 = .
82) means that its observations require special processingexpected in a later data release (Gaia Collaboration et al. 2018). Theobject designated by Liu et al. (2006) as Swift J061223.0+701243.9is claimed by SIMBAD to have incorrect nomenclature. As far as wecan tell, this object is real and correctly designated by Liu but it is not particularly well-studied, with no published distance estimate. Themost recent analysis is by Butters et al. (2011) who conclude thatSwift J061223.0+701243.9 is probably an intermediate polar, but anX-ray binary nature cannot be ruled out.Six objects listed by Liu et al. (2006) as LMXBs are classifiedby SIMBAD as HMXBs. For three of these (1A 0620-00, GS 1124-684, GS 2023+338) the reference for the HMXB classification isTetarenko et al. (2016); however that catalogue does not give explicitLMXB/HMXB classifications. SIMBAD lists 3A 1516-569 (Cir X-1) and 3A 1954+319 as being classified as LMXBs by Baumgartneret al. (2013) and as HMXBs by Samus’ et al. (2017) and Krivonoset al. (2010) respectively. Neither of the latter two sources gives areference or justification for the HMXB classification. SIMBAD listsGRO J1655-40 as being classified as HMXB by Lin et al. (2012) andLMXB by Krimm et al. (2013). However, Lin et al. (2012) did notclassify sources as high- or low-mass XRBs, and the classificationsin Krimm et al. (2013) are cited as originating from the literatureor SIMBAD itself. With no strong reasons to reclassify these sixobjects, we retain them in our list of LMXBs.We were able to find published distance estimates for ten objectsthat had no distance estimates listed by Liu et al. (2006, 2007). Fifteenadditional objects in our sample had distance determinations morerecent than those listed by Liu et al. (2006, 2007). We tabulate thesein Table A1 and use them in our analysis in subsection 3.1.Two objects in our sample have controversial distances: Cyg X-1and GRO J1655-40. The discrepancy between radio parallax distance(from Reid et al. 2011b) and optical parallax from
Gaia of Cyg X-1is peculiar, as this system is one of the closest and brightest X-ray binaries (both in radio and optical). This apparent tension islikely caused by impact of the radio jet on the radio parallax (Miller-Jones et al., in prep). The
Gaia distance is more consistent with thatreported in the Liu catalogue (2.14 kpc; Massey et al. 1995). Foellmiet al. (2006) challenged the accepted distance to GRO J1655-40 of3.2 kpc, finding a distance of 1.7 kpc. Despite their strong claim,these authors show in their table 1 that the uncertainty in spectralclass allows the upper limit on distance to be as high as 3–4 kpc(e.g., if the companion is F7ii ). Interestingly, the
Gaia counterpartparallax is consistent with the larger distance. However, the locationof the source makes distance calculation based on parallax stronglydependent on the prior model: there is a large discrepancy betweenthe distance based on Bailer-Jones prior ( ∼ ∼ − APPENDIX B: DISTANCE DISCREPANCIES
In this section we discuss five objects with large discrepancies be-tween distances gathered from the literature and measured from
Gaia
DR2 with the Bailer-Jones et al. (2018) prior. Here wedefine ‘large’ as | 𝑑 Gaia − 𝑑 prev |/( . × ( 𝑑 Gaia + 𝑑 prev )) > Gaia low-to-high range. We report the
Gaia
DR2 astrometric_gof_al value as GOF. This quantity is expected tofollow a normal distribution with zero mean and unit standard devi-ation; hence absolute values (cid:38)
4U 2129+47/V1727 Cyg : Gaia distance 1.75 kpc, GOF 1.92. TheLiu et al. (2007) distance for this object is from the work of Cow-ley & Schmidtke (1990), who derive a distance of 6.3 kpc to theoptical companion. Those authors mention that it is unclear that thecompanion and XRB are a true physical association, and that pre-vious distance estimates to the XRB system generally give smaller
MNRAS , 1–16 (2021) R. M. Arnason et al.
Table A1.
XRBs with
Gaia candidate counterparts and newer publisheddistancesName Liu dist type d prev (newer) typekpc kpcLMXBsGRO J0422+32 · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · ·
Gaia
DR2 distances published after the Liu et al.catalogues. distances (e.g., 2.2 kpc; McClintock et al. 1981). We conclude thatthe
Gaia distance is consistent with these earlier estimates.
IGR J16318-4848 : Gaia distance 5.22 kpc, GOF 32.4. The Liuet al. (2006) distance for this object is from the work of Filliatre &Chaty (2004) who give a range of distances between 0.9 and 6.2 kpc,derived from SED fitting. A more recent work (Fortin et al. 2020)determines a distance from
Gaia matching and derives the samedistance as our work. We conclude that the
Gaia distance, althoughimprecise, is consistent with the broad range in the previous estimate.
IGR J16465-4507 : Gaia distance 2.70 kpc, GOF 9.0. The Liuet al. (2006) distance for this object is from the work of Smith (2004)who give an estimated distance of 12.5 kpc based on photometryof the companion. The discussion of this object by La Parola et al.(2010) explains that the optical companion is highly absorbed; Opti-cal spectroscopic studies also provide additional evidence reaffirmingthe optical counterpart. The tension between the
Gaia and previousdistance estimates remains unresolved.
XTE J1906+09 : Gaia distance 2.77 kpc, GOF 6.8. The Liu et al.(2006) distance distance for this object is from the work of Marsdenet al. (1998) who give an estimate distance of 10 kpc based on neu-tral hydrogen absorption. However, 3D dust maps in this directions(Green et al. 2019) indicate that 𝐸 ( 𝑔 − 𝑟 ) ≤ .
2, which would sug-gest that the Galactic hydrogen column density in this direction is ≤ × cm − (Bahramian et al. 2015; Foight et al. 2016). Thuswe conclude that the Gaia distance is likely more reliable for thisobject.
KS 1947+300 : Gaia distance 3.1 kpc, GOF 0.0. The Liu et al.(2006) distance distance for this object is from the work of Tsygankov & Lutovinov (2005) who give an estimate distance of 9.5 kpc basedon its X-ray pulsation properties. While the
Gaia fit appears good, itis important to note that the measured parallax is insignificant whenuncertainties are considered.
This paper has been typeset from a TEX/L A TEX file prepared by the author.MNRAS000