A catalogue of low-mass X-ray binaries in the Galaxy: from the INTEGRAL to the Gaia era

Low-mass X-ray binaries (LMXBs) are high-energy sources that require multi-wavelength follow up campaigns to be fully characterized. New transients associated to LMXBs are regularly discovered, and previously known systems are often revisited by astronomers to constrain their intrinsic parameters. All of this information compiled into a catalogue may build up to a useful tool for subsequent studies on LMXBs and their population. We provide an update on past LMXB catalogues dating back 16 years and propose to the community a database on Galactic LMXBs with the most complete manually curated set of parameters and their original references. On top of a fixed version accessible through Vizier, we propose to host the catalogue independently on our GitHub collaboration, side-by-side with our previous catalogue on high-mass X-ray binaries. The database will be regularly updated based on new publications and community inputs. We build a working base by cross-matching previous LMXB catalogues and supplementing them with lists of hard X-ray sources detected in the past 20 years. We compile information from Simbad on LMXBs as a starting point for a thorough, manual search in the literature to retrieve important parameters that characterize LMXBs. We retrieve newly detected LMXBs and candidates directly from literature searches. Counterparts to these LMXBs are compiled from hard X-rays to infrared and radio domains. Every piece of information presented on the LMXBs is curated and backed by accurate references. We present a catalogue of 339 Galactic LMXBs listing their coordinates, companion star spectral type, systemic radial velocity, component masses and compact object nature, the presence of type I X-ray bursts as well as orbital data. Coordinates and identifiers of counterparts at various wavelengths are given, including 140 LMXBs detected in {\it Gaia} DR3.


Introduction 1
More than 60 years have passed since the identification of the 2 first extrasolar X-ray source (Giacconi et al. 1962).This partic-3 ular source, Scorpius X-1, belongs to the category of low-mass 4 X-ray binaries (LMXB), which are composed of a compact ob-5 ject -either a black hole (BH) or a neutron star (NS) -that ac-6 cretes material from a low-mass companion star (M 1 M ).X-rays down to infrared wavelengths.The formation of relativis-13 tic jets commonly occurs during specific phases of activity in 14 microquasars (i.e.BH LMXBs).However, these X-ray systems 15 are but a single phase in the whole evolution of binary stars, 16 from their formation up to their endpoint as compact binaries An online version of the catalogue is publicly available at https: //binary-revolution.github.io/LMXBwebcat,and the database in the associated GitHub repository will be continuously updated based on community inputs.and gravitational wave sources (see a recent review by Tauris & van den Heuvel 2023).
As a follow-up to our previous work on building an updated catalogue of Galactic high-mass X-ray binaries (HMXBs, Fortin et al. 2023b, now containing more than 160 sources in HMXBwebcat 1 ), we present our latest catalogue of 339 LMXBs in the Milky Way here.Information available on LMXBs in the literature suffers from the same caveat as HMXBs: as these sources are inherently difficult to observe and require multi-wavelength and time series campaigns to be properly constrained, most of the discoveries on LMXBs end up spread over several years and can come from many different teams of researchers across the world.This is why the catalogue presented in this paper is centred on two important concepts.Firstly, for each parameter listed, we provide the original reference where it was derived; this results in a compilation of more than 600 unique, curated references.Secondly, in parallel to being hosted in Vizier in a fixed version, we host it independently on a dedicated website, LMXBwebcat, on which the database is able to receive updates through our GitHub collaboration 2 as new observations are common to both Liu et al. (2007) and Bird et al. (2016) using positional and identifier cross-matching performed with TOP-CAT (Taylor 2005).We then queried Simbad for all the sources containing the type (or subtype) LXB; this returns a list of 674 sources, most of which are extragalactic.These are usually grouped in small clusters around their host galaxy; hence, we used the same spatial filtering method as that presented in Fortin et al. (2023b), where we check each source for the presence of close neighbours.However, contrary to HMXBs, LMXBs are known to be present in globular clusters, and these structures may hold up to ∼10 LMXBs very close together (as is the case in NGC 6440, for instance).Hence, we used the list of globular clusters available in Vasiliev & Baumgardt (2021) to check the likelihood of a group of LMXBs being part of a Galactic globular cluster, before filtering them out if they have neighbours closer than 10 .This way, we retrieved 251 Galactic LMXB candidates from Simbad.
Eventually, after cross-matching the Simbad LXBs with the sample of high energy sources from Liu et al. (2007) and Bird et al. (2016), we obtain a working base of 573 sources.We note that a significant amount (216) were listed in the INTEGRAL catalogue as unidentified, and most of them will be discarded through a thorough manual process of checking new information published on these sources in the literature.

Retrieving binary parameters and new LMXBs
While pieces of information are available on the LMXBs and candidates compiled so far (mainly from the data catalogued in Liu et al. 2007, Ritter &Kolb 2011, andSimbad), our goal for this catalogue is to perform a thorough investigation on each of these sources to, firstly, identify which are interesting LMXB candidates, and, secondly, retrieve useful parameters as well as the original studies that derived them (this includes double checking references already compiled by former catalogues).We find that the later point is especially important, as some parameters listed might have been derived using different methods for each source; this information should be readily available for authors of further studies that revisit these sources.Thus, during the individual search for information on each of these sources, we only used already available data as a starting point for the manual search, and we systematically re-verified already available references.We compile new and updated data whenever they are available, confirm the already measured parameters by retrieving the original studies, and we search for older publications to ensure we have compiled every piece of information possible for each binary.Hence, none of the parameters or references we list in the catalogue come from an automated, unsupervised query on Simbad.We did use the resources in works that already compiled some useful data on LMXBs, such as spectral types, orbital periods and masses in Yungelson & Lasota (2008); a list of type I X-ray bursts seen be RXTE in Galloway et al. (2008) and its extension to BeppoSAX and INTEGRAL MINBAR (Galloway et al. 2020); orbital periods in Asai et al. (2022); the catalogue of stellar-mass black holes in X-ray binaries BlackCAT (Corral-Santana et al. 2016); or the latest catalogue of Ultra-Compact Xray Binaries UltraCompCAT (Armas Padilla et al. 2023, which is also a dynamic database9 ).
Because LMXBs can be highly variable sources, we settled on a list of parameters to look for that we believe are representative of the systems at all times.These are the spectral type of the companion star, the nature of the accreting compact object, masses of the binary components, the orbital parameters, the radial velocity of the centre of mass, the presence of a spin period and its time derivative, and the detection of type-I X-ray bursts.
We tried not to log any 'assumed' values for any parameterssuch as when NSs are assumed to be 1.4 M in orbital solutions or when the LMXBs are being located at 8 kpc by default if they are near the Galactic centre-unless they are motivated or indirectly constrained by other measurements.
Lastly, we performed 'blind' searches within the literature for newly detected LMXBs and candidates in papers ranging from 2016 to 2023 that may not appear in either of the catalogues of high-energy sources we used or may not yet be logged as LMXBs in Simbad.This was mainly done through queries of keywords such as binary, transient, and X-rays in the NASA ADS abstract service.Most new candidates discovered this way come from Astronomer's Telegrams.
We note that at this point, we have made the data in the catalogue as reliable as possible; this kind of curated data-mining is, however, bound to contain small errors or oversights.This is why the catalogue is hosted on an independent website and, as our previous catalogue already did (Fortin et al. 2023b), it will receive gradual updates and new releases as new information is either found by ourselves or put forward by the community.

Finding an unambiguous chain of counterparts
As we previously argued in Fortin et al. (2023b), we believe that part of what consists of a secure identification as an LMXB or candidate LMXB is having an unambiguous list of counterparts, ideally from hard X-ray to infrared.Having this positional information is also a necessary tool for astronomers to prepare new observations in follow-up campaigns.The automated search for counterparts to each LMXB is not trivial, but it is still facilitated by positional data already present in Simbad and by the stored identifiers, which also contain information about the coordinates of the source and their accuracy.Our goal here is to perform a cone search on all the catalogues listed in Table 1 in order to retrieve a proper chain of counterparts from the high energies down to the infrared wavelengths.For this, we need an initial set of coordinates to query in these catalogues.
These starting coordinates are either the ones listed in Simbad, or the ones that we parsed from the list of identifiers available in Simbad, whichever is most accurate.For instance, the candidate LMXB IGR J17480−2446 has coordinates available in Simbad, but without any information on their accuracy.It does have another identifier, CXOGlb J174804.8−244648, which we can parse into a set of coordinates with an accuracy below the arcsecond scale as it comes from Chandra observations.This is a quick method that is capable of automatically retrieving accurate positional data from Simbad that may not necessarily be available in catalogues, but that was, for instance, derived in Astronomer's Telegrams.In this particular case, this allowed us to instantly find the Chandra counterpart of IGR J17480−2446 (2CXO J174804.8−244649) with a cone search in the Chandra CSC 2 database (Evans et al. 2019).This positional cross-match is still prone to false positives, as we have to carefully look within the catalogues on a sufficiently large area so that we do not miss any potential counterparts.This is why we also performed a recursive search within the produced list of counterparts, from the poorest to the most accurate catalogues (XMM-Newton, Chandra, 2MASS, Gaia).We note that for LMXBs closely grouped within globular clusters, this method is sometimes unable to automatically separate different chains of counterparts, especially for hard X-rays for which the 212 astrometric precision is not accurate enough to separate closely 213 grouped LMXBs in clusters.Thus, for LMXBs found in clus-214 ters, we manually checked the consistency of their counterparts 215 to ensure each have a unique, precise localisation.When there is no cluster association, the 'Other distance' col-244 umn may also contain distance estimates from a number of other 245 methods, which are available in the given references.Lastly, we 246 indicate when one or several type I X-ray bursts were observed 247 by providing a reference.
In Table A.2 10 , we present the orbital parameters: orbital pe-249 riod, eccentricity, and semi-amplitude of the donor's radial ve-250 locity.We also give details on the mass of each binary compo-251

Gaia candidate counterparts
The only parameter flag that slightly differs from the others is the one attributed to the Gaia distances in Table A.1.Because LMXBs are generally very faint in the optical and nIR, and because there are so many late-type stars in the field, finding the true optical or nIR counterpart to an LMXB can be challenging even in the case where sub-arcsecond X-ray localisation is available, as interlopers can be present along with the true counterpart within the X-ray error circles.Hence, we propose candidate Gaia counterparts to LMXBs based on astrometric cross-matching as well as historical associations in the few cases where both deep and high-resolution optical or nIR imaging has been performed.
The Gaia distance flag does not refer to the reliability of the distance estimation; instead, it represents how the Gaia counterpart was associated with the LMXB.Gaia counterparts that we found compatible with either an XMM-Newton or Chandra detection (e.g. a sub-arcsecond scale association) and that do not have any other neighbour in the Gaia catalogue closer than 1 are deemed secure, and as such we do not flag them.Gaia counterparts that were found only based on a Swift detection (e.g. an arcsecond scale association) are flagged as uncertain ("1"/ †).The Gaia counterparts flagged as "2"/ † † are not necessarily unreliable, but they are not based on an association with an accurate soft X-ray detection.Instead, they come from a historical association with an optical counterpart.For instance, Sco X-1 has a Gaia counterpart but no soft X-ray position available in the Swift, Chandra, or XMM catalogues because of its tremendous X-ray brightness.
These Gaia candidate counterparts can be subject to change in further iterations of the database as new X-ray positional constraints or deeper optical and nIR imaging becomes available on LMXBs.There are ways to further determine the level of confidence in the Gaia counterparts we found, such as comparing their optical magnitudes to previously published survey data.Because LMXBs can be highly variable, we chose to perform a check between the distances inferred by Gaia parallax and the distances inferred by other means ('other distance', such as spectral energy distribution fitting or photospheric radius expansion in bursting sources).After removing the LMXBs that lie within clusters (since all of their distance information come from Gaia data), in Figure 1 we plot the Gaia distances versus the other distances from the literature.A linear fitting excluding the data points with lower or upper limits returns a proportional coefficient of 1.03 and a systematic of 50 pc, which is satisfactory since we are comparing two completely independent sets of measurements.We note that this sub-sample is representative of the whole catalogue concerning the Gaia flags: we count four flag-0, a single flag-1, and six flag-2 Gaia counterparts in the fitted sample.This may indicate that our flagging system is a bit too strict, with the caveat of a low number of sources to work with.

Catalogue statistics and uses
We present a list of 339 Galactic LMXBs and candidates in this new catalogue, as represented in Figure 2. We note that contrary to the previous catalogue of Liu et al. (2007), we do include quiescent LMXB candidates (qLMXBs hereafter); 25 are present in the current version of our catalogue.Hence, after subtracting these qLMXBs to compare our total sample to Liu et al. (2007), we provide a 67% increase in the total number of LMXBs and   In our previous catalogue dedicated to HMXBs, almost three 391 quarters of the HMXBs were detected by Gaia and were of suf-392 ficient astrometric quality to invert the distance from the paral-393 lax; in this LMXB catalogue, we note that only about 40% of 394 the sources (140) have a Gaia DR3 counterpart, and only 98 of 395 them have distance information available in Bailer- Jones et al. 396 (2021).This is explained by the fact that optical counterparts of 397 LMXBs are fainter on average, and also because these are older 398 astrophysical objects that had time to migrate out of the Galac-399 tic plane towards the bulge of the Milky Way, which is too far 400 away for any source to have a reliable Gaia parallax, even when 401 a counterpart is detected.This still allows us to localise ∼30% 402 of our LMXBs within the Galaxy thanks to Gaia as shown in 403    ing the writing of this paper, a similar LMXB catalogue was released by Avakyan et al. (2023), where the authors list 349 Galactic LMXBs.We have similar approaches, as the authors also propose to update their database based on community inputs.We did find, however, notable differences: first and foremost, we do not cite previous catalogues as references for the parameters in our catalogue; instead, we look for the original studies cited in Liu et al. (2007) or Ritter & Kolb (2003) and update them when necessary.This results in a compilation of 570 unique references for the parameters we compile in the present catalogue.Secondly, we carefully review any reference provided by Simbad, as we have found many instances where they do not point to the original study deriving the cited parameter.For example, several companion spectral types are given in Yungelson & Lasota (2008), which is a paper discussing evolutionary models for BH LMXBs that happens to list these spectral types (and some other parameters such as orbital periods or mass ratios) coming from already published data.Thirdly, we are slightly more strict concerning the inclusion of LMXB candidates, as we do not list X-ray sources for which there is no particular indication of them being at least candidate LMXBs.For this reason, we did not include the following in our catalogue: AX J1824.5−2451,[BSP2003] 24, Swift J174038.1−273712,CXOGC J174538.0−290022,XMM J174544−2913.0,SWIFT J174553.7−290347,CXOGC J174622.2−290634,Swift J175233.9−290952, and NGC 6752 CX19.Thanks to our minute investigation of each LMXB candidate, we were also able to exclude duplicates where one or several counterparts of the same source were listed as independent LMXBs.This was the case for KS 1741−293, IGR J17353−3539, TYC 6824−713−1, [ZGV2011] 9, and 3U 1728−16.Lastly, we provide corrected identifications for some sources that may have been wrongly reported to be LMXBs, such as the HMXB Cir X-1 (Jonker et al. 2007b), the cataclysmic variable OGLE BLG-ELL-12042 (Gomez et al. 2021), and some which are less likely to be LMXBs ([PLV2002] CX10 is a candidate CV in Pooley et al. 2002).
It is of prime importance for a catalogue to carefully list references.Firstly, this provides at least some insurance to the users that the data were manually curated as, for now, there are no machine learning algorithms that outperform manual data mining in the literature as we present here.Secondly, we also find that it is important to acknowledge all the observational work done on LMXBs and X-ray binaries in general since, as the current catalogue shows, there are still many missing parameters on Xray binaries that will only be derived thanks to regular proposals for follow-up observations by teams of astronomers.Lastly, the main motivator for such catalogues is to untangle the huge amount of information spread throughout the literature; citing previous catalogues as a source for data goes against this endeavour of making data more accessible to the community.

Conclusion
The continuous endeavour of catching X-ray transients and performing multi-wavelength follow-ups has led to the discovery of many new LMXBs and helped characterise their properties tremendously in these past 16 years.To bring together all this information, we present a new catalogue of 339 Galactic LMXBs that constitutes a tool for further studies on either individual systems or their whole population.Being hosted on an independent website, LMXBwebcat, this catalogue is also ready to receive updates and corrections based on new publications or input from the community.Compared to the 187 LMXBs listed in Liu et al. (2007), the current number of known Galactic LMXBs  It is therefore essential to continue monitoring new discover-512 ies on X-ray binaries so that we are able to better grasp their 513 properties as a population, their role in the Galactic ecology 514 and how, at the endpoint of their evolution, they may become 515 gravitational-wave sources detectable by LISA (Tauris 2018).

516
We would also like to note that optical and nIR follow-ups are

7
LMXBs are soft X-ray emitters powered by mass transfer hap-8 pening through Roche-lobe overflow (see a general review on 9 accreting binaries in Chaty 2022).An accretion disc can form 10 around the compact object and can be responsible for the major-11 ity of the radiation emitted during periods of high activity from

Fig. 1 . 2 .
Fig. 1.Comparison between distances to LMXBs from the literature versus the distances of the proposed Gaia counterparts inferred from their parallax.Only LMXBs that have both information and are not located within a cluster are plotted (N=16); black arrows indicate lower or upper limits.The dashed black line is the best linear fit, excluding sources with only lower or upper limits (N=11).

Fig. 3 .
Fig. 3. Corbet diagram of 50 NS LMXBs in the current catalogue alongside the 78 NS HMXBs now available in Fortin et al. (2023b), which have both orbital and spin-period information (all flags included).Symbiotic LMXBs are indicated in green rhombuses.

Figure 4 .Fig. 5 .
Fig. 5. Distribution of soft X-ray luminosities of Galactic LMXBs seen by Swift and with a distance determination (N=89).
the literature and the distances to globular clusters, we reach the 405 number of 208 LMXBs with a distance estimation.We note that 406 regardless of the method used for distance estimation (parallax, 407 globular cluster association, or X-ray modelling), we advise the 408 users to remain cautious of the determined distance values when 409 considering individual systems as they may be subject to various 410 biases, such as poor parallax quality, wrong association within 411 a globular cluster, foreground star, and so on.When considering 412 the whole sample, these biases should at least partly cancel out, 413 but there are bound to be outliers.414 To show the potential of the data we aggregated in this cata-415 logue, we constructed the distribution of X-ray luminosities of 416 the Galactic LMXBs using the 2SXPS Swift database (Evans 417 et al. 2020).We combined the unabsorbed 0.3-10 keV Apec 418 flux (labelled FAU0 in 2SXPS) with either the Gaia distances 419 or, when not available, the 'Other distance' determination in the 420 LMXB catalogue.In the case where the LMXBs are associated 421 with globular clusters, we used the distance to the cluster.This 422 luminosity distribution, shown in Figure 5, is the first step to ob-423 taining an estimation of the total X-ray budget of LMXBs in the 424 Milky Way, and it is only presented here as an example of how 425 this catalogue can be used.To go further, users may want to con-426 sider other methods of flux measurements from the 2SXPS cat-427 alogue, combine fluxes from other observatories such as XMM-428 Newton (Webb et al. 2020) or Chandra (Evans et al. 2019), and 429 of course discuss the impact of the behaviour of LMXBs from 430 quiescence to outburst on the average flux values available in the 431 aforementioned catalogues.
of our work is the retrieval of informa-434 tion about INTEGRAL sources marked as unidentified in Bird 435 et al. (2016).We kept five of them as candidate LMXBs, which 436 appear in our catalogue; in Table B.1, we also provide a separate 437 list for the 44 INTEGRAL sources for which we found an iden-438 tification in the literature that is not present in Simbad.Most are background active galactic nuclei (AGNs) or cataclysmic vari-440 ables (CVs).The purpose of this table is to provide easily di-441 gestible data for the CDS to update references in Simbad.

443
After the publication of our previous HMXB catalogue and dur-444 has almost doubled, and given the upcoming observational land-508 scape dedicated to high energies and transient sky astronomy 509 (i.e.eROSITA, LSST, or SVOM), these numbers are likely to 510 keep growing further. 511

517
still much needed to complete the set of observables (such as 518 spectral types or orbital solutions from radial velocities) avail-519 able on LMXBs.We hope this catalogue can also be a tool to 520 identify which parameter needs constraints and to facilitate the 521 identification of ideal targets for astronomers to revisit. of compact objects".SC is grateful to the CNES (Centre National d'Études 525 Spatiales) for the funding of MINE (Multi-wavelength INTEGRAL Network).526 FG is a CONICET researcher and acknowledges support from PIP 0113 and 527 PIBAA 1275 (CONICET).This work made use of NASA's Astrophysics Data 528 System (ADS) web services, and of the services associated to the Centre de 529 Données Astronomiques de Strasbourg (CDS) Simbad and Vizier.This work 530 has made use of data from the European Space Agency (ESA) mission Gaia

Table 1 .
Queried catalogues for the counterpart search.The 'Compact' column provides information on the nature of the 231 accreting compact object, and 'Spectype' refers to the spectral 232 type of the donor star.When available, we list the distance in-233 ferred from Gaia EDR3 parallaxes by Bailer-Jones et al. (2021).234Wealsocross-matchedour LMXBs with the list of globular clus-235 ters fromBaumgardt & Vasiliev (2021), which contains informa-236 tion about their distances inferred from various means, including 237 Gaia parallaxes.Hence, in the case of a spatial association with 238 a globular cluster we provide the name of the cluster and the in-239 ferred distance in the 'Other distance' column.In the text, we 240 only keep the largest error bound for Gaia distances, and values 241 are voluntarily rounded for the sake of readability; the unaltered 242 numbers are available in the electronic versions of the catalogue.243 217The general characteristics of Galactic LMXBs are presented in 218 TableA.110.The full catalogue provides their Simbad identifier 219 (Main ID field from Simbad) and a short list of the most used 220 IDs we found in the literature.To build this list, we queried each 221 known ID from Simbad in ADS and retrieved how many papers 222 used them in their title, abstract, or full text.We ranked them 223 from most popular to least popular, and thus provide identifiers 224 that should reflect the community's preferred naming conven-225 tion.In this paper, we list the LMXBs under their most pop-226 ular identifier, except for a few sources for which no popular 227 ID can be recovered in ADS; in this case, we use their Simbad 228 identifier.The online catalogue provides the full list of identi-229 fiers known by Simbad to facilitate queries in LMXBwebcat.230