Population-based identification of H α-excess sources in the Gaia DR2 and IPHAS catalogues

MNRAS 505, 1135–1152 (2021) https://doi.org/10.1093/mnras/stab1258 Advance Access publication 2021 May 21 Population-based identification of H α-excess sources in the Gaia DR2 and IPHAS catalogues M. Fratta,1,2‹S. Scaringi ,1,2J. E. Drew ,3M. Mongui´ o ,4,5C. Knigge,6T. J. Maccarone,2 J. M. C. Court ,2K. A. Iłkiewicz ,1,2A. F. Pala,7P. Gandhi 6and B. G¨ ansicke 8 1Centre for Extragalactic Astronomy, Department of Physics, University of Durham, South Road, Durham DH1 3LE, UK 2Department of Physics and Astronomy, Texas Tech University, Lubbock, TX 79409-1051, USA 3Department of Physics and Astronomy, Faculty of Maths and Physical Sciences, University College London, Gower Street, London WC1E 6BT, UK 4Institut d’Estudis Espacials de Catalunya, Universitat de Barcelona (ICC-UB), Mart´ ı i Franqu` es 1, E-08028, Spain 5Universitat polit` ecnica de Catalunya, Departament de F´ ısica, c/Esteve Terrades 5, E-08860 Castelldefels, Spain 6School of Physics and Astronomy, University of Southampton, University Road, Southampton SO17 1BJ, UK 7European Southern Observatory, Karl Schwarzschild Strasse 2, Garching bei Munchen, D-85748, Germany 8Astronomy and Astrophysics Group, Department of Physics, University of Warwick, Gibbet Hill Road, Coventry CV4 7AL, UK Accepted 2021 April 27. Received 2021 April 23; in original form 2021 February 10 ABSTRACT We present a catalogue of point-like H α-excess sources in the Northern Galactic Plane. Our catalogue is created using a new technique that leverages astrometric and photomeric information from Gaia to select H α-bright outliers in the INT Photometric HαSurvey of the Northern Galactic Plane (IPHAS), across the colour–absolute magnitude diagram. To mitigate the selection biases due to stellar population mixing and to extinction, the investigated objects are first partitioned with respect to their positions in the Gaia colour–absolute magnitude space, and Galactic coordinates space, respectively. The selection is then performed on both partition types independently. Two significance parameters are assigned to each target, one for each partition type. These represent a quantitative degree of confidence that the given source is a reliable H α-excess candidate, with reference to the other objects in the corresponding partition. Our catalogue provides two flags for each source, both indicating the significance level of the H α-excess. By analysing their intensity in the H αnarrow band, 28 496 objects out of 7474 835 are identified as H α-excess candidates with a significance higher than 3. The completeness fraction of the H αoutliers selection is between 3 and 5 per cent. The suggested 5σconservative cut yields a purity fraction of 81.9 per cent. Key words: techniques: photometric – catalogues – stars: emission-line, Be – Hertzsprung–Russell and colour–magnitude – novae, cataclysmic variables. 1 INTRODUCTION Hαemission can be observed from both extended sources, such as nebulosities associated with either star-forming regions and/or stellar remnants, and from point-like sources, with no associated extended emission. These latter objects can fall into different source-types and can span various evolutionary stages of stellar populations. The many classes of H αemitting point-like sources include (but are not limited to) a wide range of young stellar-objects (YSOs), classical Be stars, compact planetary nebulae, luminous blue variables (LBVs), hypergiants, Wolf–Rayet stars, and rapidly rotating stars. Furthermore, many interacting binary systems exhibit H αin emission due to accretion (e.g. cataclysmic variables, CVs; symbiotic stars, SySt; or binary systems in which the accreting compact object is a black hole or a neutron star). The H αemitting population is heterogeneous and challenging to identify. Because of this, samples of these objects are plagued by selection biases, which, in turn, prevent stellar evolution models from being adequately tested. E-mail: [email protected] Large, wide-field, high-angular-resolution H αimaging surveys provide the basis to discover and characterize H α-excess sources. Among the previous surveys targeting the ionized diffuse interstellar medium (ISM) that have aimed to increase the sample of known Hαsources, we can include, for instance, the H αobservations of the Large and Small Magellanic Clouds (Davies, Elliott & Meaburn 1976). However, this survey only observed small patches of the sky, and the limiting magnitude was quite stringent. On the other hand, the Virginia Tech H αand [S II] Imaging Survey of the Northern Sky (VTSS, Dennison, Simonetti & Topasna 1999) and the Southern HαSky Survey Atlas (SHASSA; Gaustad et al. 2001), covered wider areas of sky, but they suffered from relatively poor angular resolution. Among the imaging surveys that focused on point sources, Kohoutek & Wehmeyer (1999) obtained a list of ∼4000 point-like Hαemitters located in the northern Galactic plane (|b|≤10◦). Parker et al. (2005), with their Anglo-Australian Observatory/UK Schmidt Telescope (AAO/UKST) SuperCOSMOS H αSurvey (SHS), inspected an area of ∼4000 deg2in the Southern Milky Way, plus an additional ∼700 deg2area around the Magellanic Clouds. The Isaac Newton Telescope (INT) Photometric H αSurvey of the Northern Galactic Plane (IPHAS; Drew et al. 2005) provides photometry with the 2 broad-band rand ifilters, as well as with C 2021 The Author(s). Published by Oxford University Press on behalf of Royal Astronomical Society. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. Downloaded from https://academic.oup.com/mnras/article/505/1/1135/6279691 by CSIC - Instituto De Ganaderia De Montana user on 16 February 2022 1136 M. Fratta et al. the narrow-band H αfilter (see also Drew et al. 2005 and Irwin & Lewis 2001). Witham et al. (2008) used the IPHAS pre-publication photometric measurements (without a uniform calibration) to identify candidate H αemission-line sources. Their method is based on producing two-colour diagrams (TCDs) for each IPHAS field, using r−Hαand r−i, respectively, as vertical and horizontal axes. Each Wide Field Camera (WFC) pointing covers an area of 0.22 deg2in the sky. H αline excess source candidates are then selected by iteratively fitting the stellar locus and retaining positive outliers in r−Hα.This procedure is performed within pre-defined magnitude ranges to try and mitigate the effect of extinction, which can become substantial when looking through the Galactic plane (Sale et al. 2014). Using their conservative method, Witham et al. (2008) identified in total 4853 H αemitting candidates. Only a small fraction of these candidates could be confirmed through a comparison with previously developed narrow-line emitters catalogues. A spectroscopic followup (presented in Raddi et al. 2013) was then performed on 370 outliers with r<18, and 97 per cent of them did show H αemission lines. More recently, Mongui´ o et al. (2020) developed the IGAPS (INT Galactic Plane Survey) catalogue, that includes ∼295 million objects. Of these, 53 234 833 (18 per cent) unblended sources with r< 19.5 mag were tested for H α-excess. IGAPS consists of a crossmatch between IPHAS and UVEX (the UV-Excess survey of the Northern Galactic Plane; Groot et al. 2009). With the use of the 2.5-m INT, the latter survey provides photometric measurements for the sources included in a 10◦×185◦sky area, centred on the Galactic equator. More specifically, it provides U,g,andr intensities, with a limiting magnitude of 21–22 mag. The g,r,and imagnitudes in IGAPS were calibrated with reference to the ‘PanSTARRS photometric reference ladder’ (Magnier et al. 2013), while the H αnarrow-band calibration was based on the methods described in Glazebrook et al. (1994). In the context of IPHAS, the main metric for H α-excess is r− Hα. However, this colour-index is not quite constant for stars without emission lines, but varies as a function of the spectral type. Without first confining distinct populations, the measured H αexcess of a star in the IPHAS TCD cannot have a consistent relation with the net emission equivalent width, and candidates can remain lost in the main stellar locus. For this reason, population-based H α-excess selections generally produce more complete results. An example of such study is presented in Mohr-Smith et al. (2017): These authors performed their selection of H α-excess candidates on a set of previously identified O and early B stars, across the Carina Arm. Their goal was an assessment of the relative frequency of the classical Be (CBe) phenomenon in the VST Photometric H αSurvey of the Southern Galactic plane and Bulge (VPHAS +;Drewetal.2014)fieldofview. Without any knowledge of the distances, and using only IPHAS measurements, degeneracies may exist in associating a particular object with a specific stellar population. Because of this, the emissionline candidate lists of Witham et al. (2008) and Mongui´ o et al. (2020) are necessarily conservative and incomplete. In our work, H αline excess candidates are identified from IPHAS survey by using two independent and complementary methods: (a) selecting H α-excess sources relative to nearby sources in the calibrated Gaia colour– absolute magnitude diagram (CAMD), and (b) selecting H α-excess sources relative to groups of objects that occupy nearby positions in the sky. It is relevant to stress the fact that the objects that are labelled as H αline excess candidates in this study are not necessarily H αemitters; the only conclusion that can be reached through this selection process is that their H αintensity is higher than that associated with objects they are compared to. The input catalogue used in this work to identify H α-excess sources is that of Scaringi et al. (2018) (hereafter Gaia/IPHAS catalogue), which is the result of a positional sub-arcsec cross-match between the sources in the Gaia and IPHAS DR2 fields of view. When performing the cross-match, Scaringi et al. (2018) took into account the proper motions provided by Gaia in order to rewind the positions of the objects back to the IPHAS DR2 observation epoch. This catalogue contains a list of approximately 8 million sources, all found in the Northern Galactic plane. In Section 2, a more detailed description of the input catalogue is provided. Section 3 consists of an explanation of our selection process. The results obtained by our algorithm are presented in Section 4. In Section 5, these results are discussed. Section 6 presents two possible science cases. In Section 7 we draw our conclusions. 2 THE INPUT CATALOGUE The targets in the Gaia/IPHAS catalogue occupy an area of the sky included between |b|≤5◦and 29◦≤l≤215◦, and are mostly found within a distance radius of ∼1.5 kpc from us. These distances are calculated directly as the inverse of Gaia parallax measurements, with the caveat that they satisfy the parallax over error > 5 criterion (Scaringi et al. 2018; median parallax uncertainties as well as the systematic parallax offset are discussed in Lindegren et al. 2018). The choice of inferring the distances via parallax inversion is justified by Scaringi et al. (2018) with the introduction of two parameters that quantify the goodness of Gaia astrometric fit and the false-positive rate: fcand fFP, respectively. To compute fc,they binned the targets according to their Gaia G-band magnitudes; fc corresponds to the percentile assigned to each object in the bin, with respect to the χ2of the astrometric fit. On the other hand, fFP reflects the presence of spurious negative parallaxes in Gaia measurements, due to poor astrometric fits. To obtain fFP, Scaringi et al. (2018)first produced a mirror sample of their catalogue, including only objects with negative parallaxes, with ‘parallax over error < −5’. They thus binned the objects in the catalogue (including the mirror sample) with respect to their G-band measurements, and further with respect to the χ2of their astrometric fit. They thus define fFP as the fraction of objects from the mirror sample (false positives) in each bin. To obtain the absolute magnitude for the Gaia G band (MG), Scaringi et al. (2018) used the distances calculated with the parallaxinversion method. Despite the precautions taken, this approximation contributes to the uncertainties on MG. However, the effects on MGintroduced by the use of parallax-inversion method instead of probabilistic methods (Astraatmadja & Bailer-Jones 2016) to obtain the distances are generally negligible. In fact, 97.2 per cent of the objects in our meta-catalogue fall in the |δMG|≤0.1 mag range, δMGbeing the difference between the G-band absolute magnitudes obtained with the parallax-inversion defined distances and with probabilistically defined distances.1 Besides the errors on Gaia photometric measurements and the effects connected to the parallax-inversion defined distances, the location of the sources in the CAMD is also affected by the different extinctions that alter their colours. All these uncertainties may be the causes of stellar population mixing. Our approach to overcome this obstacle is presented in the last paragraph of Section 3.1.1. 2.1 Additional data quality constraints This work focuses on the subset of targets from the Gaia/IPHAS catalogue that pass strict quality criteria, in order to minimize the inclusion of spurious cross-matches. Some quality control selection 1The absolute value of the maximum difference is |δMG,max|=0.36 mag. MNRAS 505, 1135–1152 (2021) Downloaded from https://academic.oup.com/mnras/article/505/1/1135/6279691 by CSIC - Instituto De Ganaderia De Montana user on 16 February 2022 Gaia/IPHAS catalogue of H α-excess sources 1137 cuts have already been applied during the compilation of the Gaia/IPHAS catalogue, which are mostly aimed at retaining only those sources with good Gaia parallax measurements and good IPHAS photometry. Additional cuts are applied here, in order to (i) remove sources with low-quality astrometric fits and/or high false-positive probabilities (see Section 2); (ii) remove targets close to t he saturation limit of IPHAS; (iii) only retain targets for which we have a valid measurement in each band of interest (r,i,Hα,MG,GBP,andGRP). The following cuts are thus applied: (i) retain sources that satisfy both fc<0.98 and fFP ≤0.02 (as suggested in Scaringi et al. 2018); (ii) retain sources with r≥13 mag, i≥12 mag, and H α≥ 12.5 mag; (iii) retain only sources with measurements in all r,i,Hα,MG, GBP,andGRP bands. These cuts yield 7474 835 sources out of the original 7927 224. Fig. 1shows the Gaia CAMD (i.e. the Gaia Hertzsprung–Russell diagram, HRD) and IPHAS two-colour diagram with the targets that pass the additional quality cuts. 3 SELECTING H α-EXCESS SOURCE CANDIDATES The aim of this work is to identify H α-excess candidates in a vast sample of objects. This task is achieved by selecting ‘positive outliers’ in the r−Hαversus r−itwo-colour space. To mitigate the selection biases due to stellar population mixing and to Galactic extinction, the sources in the master-catalogue are first partitioned with respect to their positions in the Gaia CAMD and Galactic coordinates space, respectively. The two r−Hαoutliers selections, performed on the CAMD-based and on the coordinates-based (or also ‘positional’-based) partitions, are independent and complementary. The selection strategy performed on the coordinates-based partitions hinges on the one applied by Witham et al. (2008). We point out that our CAMD-based selection can still be improved, since some populations may overlap in the colour–absolute magnitude space. It is worth pointing out that other techniques using more novel machine learning approaches could be employed for the selection of H α-excess sources. Our choice of a more rational approach is based on the relative simplicity of the algorithm, which allows to locate exactly in which partition a specific source has been selected from. Furthermore, the approach used here allows us to examine and understand the underlying population used to infer the H α-excess significance values. The separation of the sources in the two parameter spaces is described in Section 3.1, whilst the proper selection of H α-excess sources is discussed in Section 3.2. 3.1 Partitioning algorithms 3.1.1 CAMD-based partitions Using the calibrated Gaia CAMD shown in the top panel in Fig. 1, subsets (i.e. the partitions) are defined such that they (a) contain a large enough number of sources (500) to be able to statistically identify outliers and (b) are small enough in the colour–absolute magnitude space to make the underlying source population as homogeneous as possible. To balance these two requisites, an iterative method is applied. Figure 1. Positions in the Gaia MGversus GBP −GRP CAMD (top panel) and IPHAS r−Hαversus r−iTCD (bottom panel) of the sources in the Gaia/IPHAS catalogue (Scaringi et al. 2018). The grey dots represent the objects that satisfy the quality constraints described in Section 2.1, while the targets that do not pass this first selection are displayed with the black dots. The red and the green lines in the top panel represent respectively the synthetic zero age main-sequence (ZAMS) track (Bressan et al. 2012)and the synthetic white dwarfs track (Carrasco et al. 2014). The red line and the yellow line in the bottom panel (both taken from Drew et al. 2005) depict, respectively, the synthetic ZAMS track for zero reddening and the synthetic red giant (RG) track, in this parameter space. First, a fine grid of 840 ×840 equally spaced ‘elemental’ cells is generated, covering the whole CAMD (each elemental cell with dimensions lx∼0.007 mag and ly∼0.024 mag, respectively). No elemental cells contain enough objects to be considered a partition. The side lengths of the grid cells are then increased to the next integer divisor of 840, in units of lxand ly, respectively. The second iteration produces 420 ×420 cells, 4852 of which satisfy the criteria to become partitions (these belong to the densest regions of the CAMD). These partitions are labelled according to the order by which they are generated during the current iteration (left to right, top to bottom), from 0 to 4851. The iterations carry on for all the integer divisors of 840, between 2 and 60.2At the end of the iterative procedure, 204 459 objects are still left without a partition assignment. These ‘leftovers’ are assigned to the closest partition. 9181 CAMD −partitions result from this process, with a maximum density of 1514 sources per 2Caveat: each partition must not be completely surrounded by another one; furthermore, all the elemental cells within each partition must be contiguous. MNRAS 505, 1135–1152 (2021) Downloaded from https://academic.oup.com/mnras/article/505/1/1135/6279691 by CSIC - Instituto De Ganaderia De Montana user on 16 February 2022 1138 M. Fratta et al. Figure 2. Map of the partitions in the Gaia CAMD. The colour code refers to the order in which the partitions were created (no partitions are assigned to the light blue area). The area covered by the single partition increases where the density of sources decreases. The red line represents the synthetic ZAMS track (Bressan et al. 2012), while the green line depicts synthetic white dwarfs track (Carrasco et al. 2014). The yellow star points to partition 7331, while the diamond refers to partition 0, which are discussed in Section 3.2. partition. The map of the resulting partitions in the CAMD is shown in Fig. 2 To account for the uncertainty on the positions of the objects in the CAMD, this partitioning process is repeated upon change of the side lengths of the elemental cells. As an example, a 20 per cent increase of the side lengths of the elemental cells produces a 0.8 per cent variation in the number of selected outliers, meaning that our selection is independent (to a reasonable extent) on the size of the elemental cells. 3.1.2 Coordinates-based partitions For this different partitioning algorithm, an evenly spaced grid in the bversus lspace is created. The size of each cell is 1.205 × 1.004 deg2, i.e. about five times bigger than the ‘cell size’ used by Witham et al. (2008) (which performed their selection on an IPHAS field-by-field basis), and is chosen so that all the cells are either empty, or contain at least 500 objects. This procedure results in 1674 positional −partitions, with a maximum density of 12 604, and a minimum density of 546 objects per partition. 3.2 Detrending and identification of outliers. The r−Hαversus r−iTCDsareusedtoidentifyHαline excess sources from every partition. First the main stellar population locus is found in each partition by iteratively fitting a line to the data, and applying Chauvenet’s criterion. The latter consists of calculating a threshold3beyond which only outliers are expected to be found. The outliers are removed from the data at the end of each iteration. In theory, in order to tackle the population-mixing issue, the fit should be forced to the upper branch in the TCD (as done by Witham et al. 3Chauvenet’s threshold depends on the root mean square of the distribution and on the number of objects that constitute such distribution. Figure 3. Graphical depiction of the detrending and outliers selection processes performed on CAMD-partition 7331. The top panel shows the corresponding non-detrended IPHAS TCD, in which only one linear trend is clearly visible. The blue line depicts the best-fitting linear model. It was subtracted from the data points to obtain the detrended r−Hαparameter (bottom left-hand panel). The red dots represent the positive outliers of the distribution. The bottom right-hand panel shows the detrended r−Hα distribution of this partition. The Gaussian behaviour of the underlying population is well described by the best-fitting model (the black solid line). The red arrows point to the three outliers of the distribution. The position of the CAMD-partition 7331 in the Gaia CAMDisshowninFig.2. 2008) by removing only the negative outliers. However, for most of the partitions, the resulting best-fitting line does not deviate sensibly from the model obtained by the direct application of the unmodified Chauvenet’s criterion. Once the stellar locus has been located, it is used as a baseline to identify the outliers: each TCD is detrended by subtracting the corresponding linear model from the data. A second iterative application of Chauvenet’s criterion on the detrended TCD enables us to isolate the outliers, and hence to calculate the standard deviation (rms) of the remaining sources. The objects that satisfy the following relation are selected as H α-excess candidates, from either the CAMD-based and/or the positional-based partitions: σ=y (δy)2+(mfit ×δx)2+rms2≥3.(1) Here, ycorresponds to the (r−Hα)detrended intensity, δyis the instrumental uncertainty on this value, δxis the instrumental error on the r−iintensity, and mfit is the slope of the best-fitting line. Thus defined, σ,orsignificance, represents the confidence that each source is an outlier of the corresponding distribution. Since the partitioning process is implemented in two different parameter spaces, two significances are assigned to each source: CAMD − significance (σCAMD)andPOS −significance (σPOS). Objects that satisfy relation (1), either from the CAMD-based and/or from the positional-based selection, will henceforth be referred to as ‘3σ outliers’. Fig. 3provides a graphical depiction of the detrending (top panel) and selection (bottom left-hand panel) processes relative to the CAMD-partition 7331, as an example of a well-behaved partition. We point out that Chauvenet’s criterion assumes an underlying Gaussian population, while a non-negligible amount of our partitions seems to deviate from this (mainly due to population mixing). However, the application of Chauvenet’s criterion on non-Gaussian MNRAS 505, 1135–1152 (2021) Downloaded from https://academic.oup.com/mnras/article/505/1/1135/6279691 by CSIC - Instituto De Ganaderia De Montana user on 16 February 2022 Gaia/IPHAS catalogue of H α-excess sources 1139 Figure 4. Graphical depiction of the detrending and outliers selection processes performed on CAMD-partition 0. As it stands out clearly, the Gaussian model is not a good fit to the underlying population. The position of partition 0 in the Gaia CAMD is shown in Fig. 2. Figure 5. Graphical depiction of the detrending and outliers selection processes performed on positional-partition 154. The red line in the top panel represents the synthetic ZAMS track, for zero reddening (Drew et al. 2005). partitions provides a more robust H α-excess outlier selection, since the standard deviation of these partitions is overestimated. As can be noticed from the bottom right-hand panel of Fig. 3, the detrended r−Hαdistribution relative to the CAMD-partition 7331 constitutes a good example of Gaussian underlying population. On the other hand, Fig. 4presents an example of partition (CAMD-partition 0) in which the underlying distribution deviates from a standard Gaussian distribution. As a reference, Fig. 5presents the TCDs (before and after the detrending process) and the histogram of the detrended r−Hαvalues relative to positional-partition 154. Two trends are identifiable from the TCDs: the top one represents the locus in which unreddened MS stars lie, while the bottom trend corresponds to the reddened RG track. These two trends reflect in the bimodality recognizable in the histogram in the bottom right-hand panel. This effect does not alter the number of outliers selected from partitions Figure 6. The top panel shows the layout of the r−Hα3σoutliers in the IPHAS TCD, while their position in the Gaia CAMD is presented in the bottom panel. The red dots represent all the 3σoutliers identified by either our CAMD-based or positional-based selection, while the blue dots represent the subset of 5084 outliers selected from both the partition types. The intensity of both these colours scales inversely with the density of objects. that present it, since the second Gaussian population is always redder than the main one. Our algorithm selects both positive and negative outliers; however, since our goal is to identify the H α-excess candidates, the term ‘outliers’ will henceforth refer to only the positive ones. 4 RESULTS Our selection identifies 28 496 r−Hα3σoutliers (0.4 per cent of the total data set) above the previously identified stellar loci. More specifically, 25 030 outliers are selected from the CAMD-partitions and 8550 from the positional-partitions. In Fig. 6, the locations in the IPHAS TCD and Gaia CAMD of these outliers are presented. It appears particularly noticeable in the top panel that many of these candidates would have not stood out as outliers, if the chosen statistical analysis had been applied directly in the two-colour domain. From the bottom panel, mainly four regions of the CAMD with a particularly high density of outliers can be highlighted: the white dwarfs (WDs) track, the M-dwarfs area, the YSOs region, and the region where reddened early MS stars of spectral type B or A sit. Fig. 7displays the map of the fraction of outliers per CAMDpartition. Mainly two regions with a relatively high fraction of MNRAS 505, 1135–1152 (2021) Downloaded from https://academic.oup.com/mnras/article/505/1/1135/6279691 by CSIC - Instituto De Ganaderia De Montana user on 16 February 2022 1140 M. Fratta et al. Figure 7. Fraction of outliers per CAMD-partition. outliers stand out. These regions are: the area commonly associated with YSOs, i.e. to the right of the ZAMS (the red line) and centred at around MG=7.5 mag, and the region where reddened, bright, B or A spectral types stars lie (i.e. the top area in the CAMD, centred around GBP −GRP =1.2 mag). Further information about the statistical composition of the H α-excess candidates that occupy these areas of the CAMD can be obtained through a cross-match with SIMBAD data base (Wenger et al. 2000). Out of 981 outliers that occupy the former overpopulated region, 608 are classified as YSOs (or candidates), or T-Tauri stars; 154 of them are classified as emissionline objects, while 146 simply as ‘Star’.4Among the outliers included in the latter group of interest, 131 find a classification in SIMBAD: 55 of them are identified Be stars (or candidates), 34 are emission-line stars, 24 are classified as ‘Star’, and 10 are red giant branch stars. In Fig. 8, the map of the fraction of outliers per positional-partition is shown. To rule out systematic effects, an analysis of the relationship between the size of each partition and the fraction of outliers within it was performed; no such correlation was found. The distribution of this ratio in the Galactic coordinate space is consistent with being homogeneous, with no significant trend in either direction. None the less, some areas in the bversus ldiagram with a relatively high density of H α-excess candidates can be highlighted. These might correspond, say, to regions with a high rate of star formation, such as molecular clouds, or to open clusters. Two examples are the known open clusters IC1396 (centred at l=99. ◦30 , b=03. ◦74 ; Kharchenko et al. 2013) and NGC 2264 (centred at l=202. ◦94, b=02. ◦30; Dias et al. 2014; Kuhn et al. 2019; Barentsen et al. 2013), which are easily identifiable in Fig. 8. The latter star-forming region has been the subject of previous studies, such as the one presented by Barentsen et al. (2013). They applied the method of the Bayesian inference to identify 115 accreting objects in NGC 2264. Positional-partition 1289 (highlighted by the yellow circle in Fig. 8), corresponds to the sky area in which NGC 2264 is located, and is in fact the positional-partition with the highest fraction of H α-excess candidates: out of 2826 objects, our algorithm selects 71 outliers (2.5 per cent). Nine positional-partitions centred around this partition are showninFig.9. The apparently empty areas in the sky are due to the 4We point out that the generic ‘Star’ label in SIMBAD refers to objects that have been identified, but for which there is not enough information for a more specific classification. Figure 8. Fraction of outliers per positional-partition. Positional-partition 1289 is highlighted by the yellow circle. Figure 9. Graphical depiction of the nine positional-partitions centred around positional-partition 1289 in the Galactic coordinates space. This latter partition is the one with the highest fraction of H α-excess candidates (the red dots). quality cuts applied when compiling IPHAS DR2; because of these cuts, IPHAS DR2 provides photometric measurements for sources covering 92 per cent of its footprint (Barentsen et al. 2014). Ideally, the concept of ‘outliers of a distribution’ would be nonarbitrary. However, realistically speaking the definition of ‘outlier’ is strongly dependent on the chosen threshold. This can be mitigated by the choice of different confidence levels during the selection process. One possibility consists in considering as outliers all the objects that are selected using Chauvenet’s criterion: this would be ideal if all partitions were to display Gaussian distributions. On the other hand, one can choose to select as outliers all the objects that satisfy σ≥3 (equation 1); this constitutes a more relaxed threshold, when compared to Chauvenet’s one. We point out that by setting this threshold, a certain amount of falsepositives in our selection is to be expected. However, this amount is not easily quantifiable, since the distributions are not always MNRAS 505, 1135–1152 (2021) Downloaded from https://academic.oup.com/mnras/article/505/1/1135/6279691 by CSIC - Instituto De Ganaderia De Montana user on 16 February 2022 Gaia/IPHAS catalogue of H α-excess sources 1141 Gaussian, and also they are not equally populated. The suggested threshold is the 5σone (σ≥5), as a compromise to reduce falsepositives fraction, while retaining a robust candidate selection. In fact, all the candidates selected using a 5σthreshold would have been included using Chauvenet’s criterion as well. By applying the 5σcut, 6774 outliers (0.09 per cent of the complete data set, 23.8 per cent if compared to the 3σsample) are identified: 6455 from the CAMD-partitions and 2209 from the positionalpartitions. For all the sources in the master-catalogue, the two flagCAMD and flagPOS specifications are evaluated. These entries can assume a value of 0 (if the significance is lower than 3), 1 (if the significance is greater than or equal to 3, but smaller than 5), or 2 (if the significance is equal to or greater than 5). A very similar classification of the significance levels was previously adopted by Witham et al. (2008) and Mongui´ o et al. (2020). Our results are presented in a meta-catalogue, the first 10 rows ofwhichareshowninTable1. The full set of metrics computed during the catalogue generation is also published. In this full version, the necessary pieces of information to trace back each source to the corresponding CAMD-based and positional-based partitions are provided, as well as the detrending model information for each TCD. Our hope is that this additional information will aid future users of the catalogue to further tune the selection of H α-excess sources to suit a specific task. 5 DISCUSSION In Fig. 10, the positions of our 3σoutliers in the CAMD (left-hand column) and TCD (right-hand column) are shown. The most evident differences between the selections applied on the CAMD-partitions and on the positional-partitions are: as follows (i) The CAMD-based selection is less efficient, compared to the positional-based one, in identifying outliers along the WD track. This effect stands out clearly from the comparisons of both the TCDs and the CAMDs. It is due to the constraints set when partitioning the Gaia CAMD: the partitions in the least populated regions of the CAMD have to be large enough in size to contain at least 500 sources each. For this reason, our algorithm creates only three large partitions in the WD track and surrounding area of the CAMD. This results in a limited amount of detected outliers. (ii) The CAMD-based selection identifies more H α-excess candidates in the region of the CAMD where the reddened B and A types emission line stars are, if compared to the positional-based algorithm. (iii) Most M-dwarf H α-excess candidates are identified mainly through the positional-based selection. We believe that this is related to the fact that M-dwarf stars have an intrinsically more intense r −HαIPHAS colour, compared to other MS stars. In contrast with the CAMD-partitions, in the positional-partitions these objects are blended with other populations, and hence they stand out as outliers. (iv) The CAMD-based selection is more efficient at identifying YSOs of various types. This can be observed in the top CAMD in Fig. 10 as the cluster of H α-excess candidates found to the right of the MS track, and centred at around MG∼7.5 mag. These systems would be difficult to identify with a positional-based partition, unless they display strong H αemission. The position of the sources in the CAMD constitutes an indication of the stellar population they most likely belong to. A cross-match between our outliers sample and the SIMBAD data base (Wenger et al. 2000) provides a further statistical representation of the Table 1. The table shows the first 10 rows of our meta-catalogue. SourceID RA Dec. Distance rerieiHαe H αGBP GRP MGFlag Flag (Gaia DR2) (◦)( ◦) (kpc) (mag) (mag) (mag) (mag) (mag) (mag) (mag) (mag) (mag) CAMD POS 429950213735495552 0.011 55 62.530 21 4.911 91 13.217 0.001 12.919 0.002 12.963 0.001 13.418 12.938 −0.196 0 1 430049096757310848 0.040 97 62.806 30 0.245 88 17.815 0.017 17.868 0.026 17.663 0.018 17.612 17.808 10.816 0 1 432344808313438336 0.042 68 66.348 93 0.725 53 18.881 0.016 17.163 0.014 17.851 0.019 20.173 16.946 9.068 1 1 432167168466312448 0.050 05 65.177 39 0.801 85 18.114 0.009 16.901 0.012 17.378 0.013 19.016 16.783 8.374 0 1 429950179375766144 0.069 64 62.530 50 0.442 58 13.017 0.001 12.419 0.001 12.593 0.001 13.477 12.461 4.819 2 1 429952687636601728 0.093 28 62.647 48 0.326 89 13.639 0.001 13.100 0.002 13.276 0.002 14.408 13.114 6.252 0 1 429521606051123712 0.119 78 61.693 65 2.611 61 16.222 0.003 15.723 0.006 15.738 0.006 16.619 15.392 4.000 1 2 432168650239229696 0.121 83 65.206 98 0.500 34 13.080 0.001 12.362 0.001 12.617 0.001 13.615 12.402 4.591 2 0 429915231214176768 0.124 47 62.172 43 1.189 93 17.926 0.010 16.656 0.011 17.262 0.015 18.942 16.643 7.368 1 0 430048723107383808 0.129 39 62.771 23 1.201 74 15.160 0.004 14.507 0.003 14.679 0.003 15.498 14.192 4.520 2 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Notes. For each source, the following entries are provided: the Gaia DR2 SourceID; the Gaia DR2 barycentric equatorial coordinates at epoch 2015.5; the distance (calculated in Scaringi et al. 2018); the IPHAS DR2 photometric measurements with corresponding errors; the Gaia DR2 photometric measurements; and the two labels flagCAMD and flagPOS. These two latter parameters express how likely each source is to be an outlier, within its corresponding r−Hαdistribution (either in the CAMD-partitions and/or in the positional-partitions). MNRAS 505, 1135–1152 (2021) Downloaded from https://academic.oup.com/mnras/article/505/1/1135/6279691 by CSIC - Instituto De Ganaderia De Montana user on 16 February 2022 1142 M. Fratta et al. Figure 10. The top row presents the r−Hα3σoutliers (the red dots) identified from CAMD-partitions, while the ones obtained from positional-partitions are shown in the bottom row. populations our H α-excess candidates belong to. A cross-matching radius of 1 arcsec yields 1825 matches. Of them, 822 are classified as YSOs (or candidate YSOs) or T-Tauri stars, 376 sources are classified with the generic epithet of ‘Star’, 233 are emission-line stars, 113 are classified as Be stars (or candidates), 44 as Orion Variable stars, 13 as WDs (plus 43 WD candidates), and 8 are known CVs (or candidates). The WDs included in our list of outliers and in SIMBAD are further classified, according to their spectral type: six of them are DB white dwarfs, four are DA type, two DC type, one DAB type, and one DBA type. The non-DA type WDs are identified as H α-excess candidates by our algorithm because they do not present the strong absorption lines, typical of DA type WDs. The left-hand column in Fig. 11 shows the rmagnitude distributions of our 5σoutliers. The bin size for these histograms is 0.2 mag. As can be noticed, both the distributions are bimodal; the peak around the 13th magnitude for the CAMD-outliers, as well as the one around r=13.5 mag for the positional-outliers, is to be partially imputed to an observational bias. The secondary mode for the 5σCAMD outliers is 16.90 mag, and it is very close to the mode of the rintensity of the whole data set (which is r∼16.60 mag). On the other hand, the most frequent rintensity for the 5σPOS outliers is 18.15 mag (i.e. more than 1.5 magnitude fainter than the mode of the whole data set), confirming the fact that, generally speaking, our positionalselection is more efficient in identifying fainter outliers than the CAMD-selection. The blue areas in the histograms represent the most populated bins around r=13 mag (for the CAMD-outliers) and r= 13.5 mag (for the positional-outliers), while the red areas indicate the three most populated bins around the respective secondary modes. In the right-hand column of the same figure, the CAMD densities of the sources belonging to these coloured regions are shown. According to their positions in the CAMD, the brightest outliers in both the distributions are active B or A types stars. The CAMDoutliers belonging to the red bins mainly cluster in the region of the CAMD associated with YSOs. Also the positional-outliers with an r magnitude close to the secondary mode mainly occupy the region of the CAMD associated with YSOs; however, some of them lie on the WD track, some in between the WD track and the MS (making them good CV candidates), and some can be found on the M-dwarfs region. In the following subsections, our results are compared with previous similar studies. More specifically, they are cross-matched with the catalogues developed by Witham et al. (2008) and by Mongui´ o et al. (2020) (IGAPS). Moreover, a further validation of our selection, based on the visual inspection of LAMOST DR5 spectra (Yao et al. 2019) is presented. Since accreting compact objects often show X-ray emission, a fraction of our H α-excess candidates are expected to be found in X-ray surveys as well. Therefore, in the last subsection, the quantitative results of the cross-matches with three Xray surveys are briefly discussed. These surveys are: the ROSAT AllSky Survey Faint Source Catalogue (‘faint-ROSAT’ hereafter; Voges et al. 2000), the ROSAT All-Sky Survey Bright Source Catalogue (‘bright-ROSAT’ hereafter; Voges et al. 1999), and the Chandra Source Catalogue (‘CSC’ hereafter; Evans et al. 2010). MNRAS 505, 1135–1152 (2021) Downloaded from https://academic.oup.com/mnras/article/505/1/1135/6279691 by CSIC - Instituto De Ganaderia De Montana user on 16 February 2022 Gaia/IPHAS catalogue of H α-excess sources 1143 Figure 11. rmagnitude distributions (left-hand column) of the 5σCAMD outliers (top row) and of the 5σPOS outliers (bottom row), with a bin width of 0.2 mag. Both the distributions are bimodal; the blue areas in the histograms point to the bins around the brightest of the two modes, respectively, while the red areas in the histograms display the three most populated bins around the secondary modes. In the right-hand column, the density in the Gaia CAMD of the objects that occupy the areas around the modes in the corresponding histogram are shown. 5.1 Comparison with Witham’s catalogue Comparing the emitters’ list in Witham et al. (2008) (which counts 4853 objects) with the full master-catalogue developed by Scaringi et al. (2018) (after the application of the quality cuts described in Section 2.1), 1213 common sources (25.0 per cent of Witham et al. 2008 outliers sample) are found. The cross-match is performed by using a radius of 1 arcsec; however, the number of matches does not change significantly if this parameter is increased up to a generous 5 arcsec. Although the remaining 75 per cent of Witham’s outliers can be found in Gaia DR2 archive, their astrometric/photometric measurements did not satisfy the quality constraints applied by Scaringi et al. (2018) when producing the Gaia/IPHAS catalogue. Out of the 1213 common targets, 1115 (91.9 per cent) are identified as outliers by our algorithm as well, with a significance (either σCAMD and/or σPOS) equal to or higher than 3. By comparing the common sources with our CAMD-based outliers, 1053 of common outliers (94.4 per cent) are recovered, whilst the positional-based selection finds 1054 (94.5 per cent) of them (i.e. 992 common outliers are identified by both our selection criteria). A subset of 933 out of 1115 common objects is characterized by σCAMD ≥5 and/or σPOS ≥5; 893 of them have a σCAMD ≥5, while 671 of them have σPOS ≥5 (hence 631 Witham’s emitters are found by both our selection criteria, with a significance equal to or greater than 5). In the two panels of Fig. 12, the positions in the Gaia CAMD of the objects resulting from the cross-match with Witham’s catalogue are presented. More specifically, the top panel shows the matches between Witham’s list and our 3σCAMD-outliers, while the bottom panel shows an analogous diagram for our positional-outliers. As previously stated in this work, the CAMD-based selection is more efficient in recovering bright objects, such as the ones that lie on the active, reddened, B or A type stars are tracked. On the other hand, the cross-match with our positional-based selection yields more matches in the M-dwarf region of the CAMD, and on the WD track. By checking the differences in the photometric measurements in Witham et al. (2008) (minus the uncertainties) and in IPHAS DR2 (plus the uncertainties), some variable objects can be spotted. However, we point out that this apparent variability might be due to the different calibrations.5Of the 98 3σobjects missing in our list of outliers, 42 showed a stronger r−Hαemission at the epoch of Witham’s study, if compared to IPHAS DR2. This decrease in the r− Hαintensity might be the reason why those particular sources were identified as H αemitting candidates in Witham et al. (2008), but not by the methods implemented in our study. Fig. 13 shows this apparent r−Hαintensity drop. 5As previously mentioned, the study of Witham et al. (2008) was performed on IPHAS pre-publication measurements, while our selection algorithm is applied to IPHAS DR2 calibrated data. MNRAS 505, 1135–1152 (2021) Downloaded from https://academic.oup.com/mnras/article/505/1/1135/6279691 by CSIC - Instituto De Ganaderia De Montana user on 16 February 2022 1150 M. Fratta et al. Figure 21. The left-hand panel presents the positions in the Gaia CAMD of our r−Hα3σoutliers that are included in the surveys bright-ROSAT (Voges et al. 1999, the blue points), faint-ROSAT (Voges et al. 2000, the red dots), and CSC (Evans et al. 2010, the black dots), for which a SIMBAD classification is available. The right-hand panel shows the positions of the same objects in the IPHAS TCD. Figure 22. Position in Gaia CAMD of our 3σHαoutliers that show X-Ray emission. The blue dots represent the objects that are already classified in SIMBAD, while the red and black ones are unclassified (or simply classified as ‘star’ or ‘X-Ray emitting source’). the average value for the ith bin) are selected as variables. With this method, 22 199 variable sources are identified. According to our Hα-excess selection, 2243 of them are also H α-excess sources. The top panel of Fig. 23 shows the versus G-band magnitude diagram: The blue dots represent the variable objects, whilst the red ones represent the variable H α-excess sources. The bottom panel in the same figure shows the location of the variable H α-excess sources in Gaia CAMD. While most of them cluster in the region associated with YSOs, many red dots are found in the region of the CAMD where B or A type stars lie, and between the MS and the WD tracks. A combination of X-Ray emission and variability makes our accreting WD candidates identification more robust. The bottom panel in Fig. 23 presents the position in the CAMD of the 41 variable objects that show X-Ray emission and are not yet classified in SIMBAD. The five black objects in Fig. 22 are included among them. 7 CONCLUSIONS In this study, a new method for selecting H α-excess candidates from a vast photometric survey is presented. Our analysis is performed on the Gaia/IPHAS catalogue, produced by Scaringi et al. (2018). It comprises targets included in the |b|≤5◦and 29◦≤l≤215◦ranges, within a radius of ∼1.5 kpc. Gaia photometric measurements and parallaxes play a key role in the development of our selection: by locating the sources in the Gaia CAMD, it is possible to associate them to a stellar population. In order to minimize the effects due to stellar population mixing, the targets are partitioned in the Gaia CAMD; to mitigate the effects of extinction, they are further (and independently) partitioned in the Galactic coordinate space. For each partition, the main locus in the IPHAS TCD, and subsequently the r−Hαoutliers, are found by applying the iterative Chauvenet’s criterion twice. The H α-excess candidates are thus defined as the sources that satisfy the criterion in equation (1). This process leads to the identification a new set of H α-excess candidates in the Northern Galactic plane. In fact, the partition of the sources in two different parameter spaces enables the identification of Hαline candidates that would be otherwise hidden among different stellar populations. More specifically, 28 496 H α-excess candidates (0.4 per cent of the total data set) are identified, with either σCAMD ≥ 3 and/or σPOS ≥3. However, a 5σcut is suggested, as it constitutes a solid agreement between completeness and conservativeness. By applying this latter cut, 6774 objects (23.8 per cent of the 28 500 3σ outliers) are identified as H α-excess sources. The visual inspection of the available LAMOST DR5 spectra of our 3σoutliers shows that 48.9 per cent of them exhibit a clear H αemission line. This purity fraction does not improve significantly if constraining the outliers to the 5σsubset: 49.7 per cent of them are confirmed H α emitters by the available LAMOST spectra. These apparently low percentages are explained by the fact that our algorithm identifies Hα-excess sources, rather than H αemitters. This is also consistent with the spectral confirmation rate being systematically higher for our positional-outliers than for our CAMD-outliers. However, by retaining only the outliers that are at least half magnitude fainter than IPHAS saturation limits, 67.7 per cent of our 3σoutliers – and 81.9 per cent of our 5σoutliers – are spectroscopically confirmed as reliable H α-excess sources. This latter selection cuts are therefore suggested. Our 3σselection identifies between 3 and 5 per cent of the H α emitters in the Northern Galactic plane. Despite this constituting an improvement with respect to previous similar studies, it also suggests that our knowledge of the H αemitters in the Galaxy is still far from being complete. MNRAS 505, 1135–1152 (2021) Downloaded from https://academic.oup.com/mnras/article/505/1/1135/6279691 by CSIC - Instituto De Ganaderia De Montana user on 16 February 2022 Gaia/IPHAS catalogue of H α-excess sources 1151 Figure 23. Top panel: variable objects (the blue dots) and variable H α outliers (the red dots) in our meta-catalogue. Bottom panel: position in the CAMD of our variable H α-excess sources (the red dots). The black triangles represent the variable H αoutliers that also show X-Ray emission, and are not classified in SIMBAD. The results of our analysis are presented in our meta-catalogue of the Gaia/IPHAS H α-excess sources. This includes all the 7474 835 objects in the master-catalogue, and for each of them, the following specifications are provided: the Gaia DR2 SourceID, the equatorial coordinates, the distance, IPHAS DR2 and Gaia DR2 photometric measurements, as well as the two flagCAMD and flagPOS labels that specify the confidence level that the source is an H α-excess candidate. Moreover, a full-version of our catalogue is available, in which the whole set of metrics obtained during the H α-excess selection process is added. A cross-match with SIMBAD (Wenger et al. 2000) shows that 6.4 per cent of our 3σoutliers have been previously identified. However, if followed up spectroscopically, our list of outliers can be used to enhance the census of identified H αemitting point-like sources, such as CVs or SySts. This constitutes a profitable starting point to address, for instance, the problem of the difference between observed and predicted CVs spatial density in the Galactic plane (de Kool 1992; Kolb 1993). Although Belloni et al. (2020) seem to have found a promising way to overcome this impasse, their conclusions are still to be confirmed (Pala et al. 2020). Moreover, the identification of new H αemitting sources can foster population studies, which, by definition, need a vast amount of objects to be performed. In addition, newly classified sources can provide further pieces of the puzzle for a better understanding about the possible evolutionary models of the stellar population they belong to. With the arrival of Gaia early Data Release 3 (Gaia eDR3), our intention is to apply our analysis on the list of objects resulting from the cross-match between Gaia eDR3 and IGAPS. This will provide more up to date results, compared to the ones in our current metacatalogue. ACKNOWLEDGEMENTS The cross-matches, as well as some of the figures used for this paper, were produced with the use of the astronomy-oriented software ‘TOPCAT’ (Taylor 2005). MM acknowledges the support by the Spanish Ministry of Science, Innovation and University (MICIU/FEDER, UE) through grant RTI2018-095076-B-C21, and the Institute of Cosmos Sciences University of Barcelona (ICCUB, Unidad de Excelencia ‘Mar´ ıa de Maeztu’) through grant CEX2019000918-M. DATA AVAILABILITY Both the light and the full versions of our meta-catalogue can be found in VizieR as ‘The Gaia/IPHAS catalogue of H α-excess sources’. REFERENCES Abrahams E. S., Bloom J. S., Mowlavi N., Szkody P., Rix H.-W., Ventura J.-P., Brink T. G., Filippenko A. V., 2020, preprint (arXiv:2011.12253) Astraatmadja T. L., Bailer-Jones C. A. L., 2016, ApJ, 833, 119 Barentsen G., Vink J. S., Drew J. E., Sale S. E., 2013, MNRAS, 429, 1981 Barentsen G. et al., 2014, MNRAS, 444, 3230 Belloni D., Schreiber M. R., Pala A. F., G¨ ansicke B. T., Zorotovic M., Rodrigues C. V., 2020, MNRAS, 491, 5717 Bressan A., Marigo P., Girardi L., Salasnich B., Dal Cero C., Rubele S., Nanni A., 2012, MNRAS, 427, 127 Carrasco J. M., Catal´ an S., Jordi C., Tremblay P.-E., Napiwotzki R., Luri X., Robin A. C., Kowalski P. M., 2014, A&A, 565, A11 Davies R. D., Elliott K. H., Meaburn J., 1976, Mem. R. Astron. Soc., 81, 89 de Kool M., 1992, A&A, 261, 188 Dennison B., Simonetti J. H., Topasna G. A., 1999, Am. Astron. Soc. Meeting Abstr. 195, 53.09 Dias W. S., Monteiro H., Caetano T. C., L´ epine J. R. D., Assafin M., Oliveira A. F., 2014, A&A, 564, A79 Drew J. E. et al., 2005, MNRAS, 362, 753 Drew J. E. et al., 2014, MNRAS, 440, 2036 Evans I. N. et al., 2010, ApJS, 189, 37 Farnhill H. J., Drew J. E., Barentsen G., Gonz´ alez-Solares E. A., 2016, MNRAS, 457, 642 Gaustad J. E., McCullough P. R., Rosing W., Van Buren D., 2001, PASP, 113, 1326 Glazebrook K., Peacock J. A., Collins C. A., Miller L., 1994, MNRAS, 266, 65 Groot P. J. et al., 2009, MNRAS, 399, 323 Irwin M., Lewis J., 2001, New Astron. Rev., 45, 105 Kharchenko N. V., Piskunov A. E., Schilbach E., R¨ oser S., Scholz R. D., 2013, A&A, 558, A53 Kohoutek L., Wehmeyer R., 1999, A&AS, 134, 255 Kolb U., 1993, A&A, 271, 149 Kuhn M. A., Hillenbrand L. A., Sills A., Feigelson E. D., Getman K. V., 2019, ApJ, 870, 32 Lindegren L. et al., 2018, A&A, 616, A2 Magnier E. A. et al., 2013, ApJS, 205, 20 Mohr-Smith M. et al., 2017, MNRAS, 465, 1807 Mongui´ o M. et al., 2020, A&A, 638, A18 Pala A. F. et al., 2020, MNRAS, 494, 3799 Parker Q. A. et al., 2005, MNRAS, 362, 689 Raddi R. et al., 2013, MNRAS, 430, 2169 MNRAS 505, 1135–1152 (2021) Downloaded from https://academic.oup.com/mnras/article/505/1/1135/6279691 by CSIC - Instituto De Ganaderia De Montana user on 16 February 2022 1152 M. Fratta et al. Sale S. E. et al., 2014, MNRAS, 443, 2907 Scaringi S. et al., 2018, MNRAS, 481, 3357 Skrutskie M. F. et al., 2006, AJ, 131, 1163 Taylor M. B., 2005, in Shopbell P., Britton M., Ebert R., eds, ASP Conf. Ser. Vol. 347, Astronomical Data Analysis Software and Systems XIV. Astron. Soc. Pac., San Francisco, p. 29 Voges W. et al., 1999, A&A, 349, 389 Voges W. et al., 2000, IAU Circ., 7432, 3 Wenger M. et al., 2000, A&AS, 143, 9 Witham A. R., Knigge C., Drew J. E., Greimel R., Steeghs D., G¨ ansicke B. T., Groot P. J., Mampaso A., 2008, MNRAS, 384, 1277 Wramdemark S., 1981, A&AS, 43, 103 Yao S. et al., 2019, ApJS, 240, 6 SUPPORTING INFORMATION Supplementary data are available at https://drive.google.com/file/ d/1eHDYa HltSPQjKjPaay7T sM7YMlqqM2/view?usp=sharing online. Please note: Oxford University Press is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article. This paper has been typeset from a T EX/L A T EX file prepared by the author. MNRAS 505, 1135–1152 (2021) Downloaded from https://academic.oup.com/mnras/article/505/1/1135/6279691 by CSIC - Instituto De Ganaderia De Montana user on 16 February 2022