EDP Sciences logo
Subscriber Authentication Point
Search
Menu
Home All issues Volume 695 (March 2025) A&A, 695 (2025) A226 Full HTML
Open Access
Issue
A&A
Volume 695, March 2025
Article Number A226
Number of page(s) 17
Section Catalogs and data
DOI https://doi.org/10.1051/0004-6361/202452286
Published online 20 March 2025

© The Authors 2025

Licence Creative CommonsOpen Access article, published by EDP Sciences, under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

This article is published in open access under the Subscribe to Open model. Subscribe to A&A to support open access publication.

1 Introduction

The termination of hydrogen burning in a stellar core marks the end of a star’s main sequence (MS) phase. As the core contracts and temperatures rise, hydrogen shell burning starts, causing the star to expand rapidly. This expansion leads to the star becoming progressively larger and cooler, transitioning into the yellow giant (YG) and yellow supergiant (YSG) phases. However, this evolutionary stage is brief and difficult to observe, creating a ‘gap’ between the MS and the red giant branch (RGB) in the Hertzsprung-Russell diagram, known as the Hertzsprung gap. Early surveys of YSGs in the Milky Way, Magellanic Clouds, and Andromeda revealed significant discrepancies between observations and stellar evolution models, particularly in estimating the duration of this phase (Massey et al. 2000; Drout et al. 2009; Neugent et al. 2010; Drout et al. 2012).

Adding further complexity, most massive stars are found in binary or multiple systems, with 70% of O-type stars and 50% of B-type stars having companions (e.g. Moe & Di Stefano 2017 and references therein). The rapid expansion of YSGs in such systems often triggers interactions like Case B mass transfer and substantial mass loss (Marchant & Bodensteiner 2024). This interaction makes the YSG phase particularly valuable for studying binary evolution. In cases where a YSG is part of a binary system, unstable mass transfer can lead to a common envelope phase (Paczynski 1976). The transfer of angular momentum of the binary into the envelope can result in its partial (or total) ejection, producing a rare astrophysical transient known as a luminous red nova (LRN). While YSGs have been linked to other rare transients, such as core-collapse supernovae (SNe) (Smartt 2015) and failed SNe (Georgy 2012; Neustadt et al. 2021), they are most notably identified as progenitors of LRNe (MacLeod et al. 2017; Blagorodnova et al. 2017, 2021; Cai et al. 2022).

Luminous red novae are optical and infrared transients with luminosities between those of novae and SNe, and they evolve over several weeks to months. Their peak brightness correlates with the progenitor’s mass (Kochanek et al. 2014; Blagorodnova et al. 2021), enabling us to investigate binary evolution across a wide range of stellar masses. Unlike optical LRNe, infrared LRNe can result from common envelope ejections in stars more evolved than YSGs (MacLeod et al. 2022). These stars experience significant mass loss, causing them to become obscured in optical wavelengths while remaining detectable in the infrared. The best-studied LRN to date, V1309 Sco, was discovered in the Milky Way. Extensive photometric data up to seven years before the onset of the transient revealed its progenitor, a contact binary system with a quickly decaying period that may have evolved via the merger of its components (Tylenda et al. 2011). Interestingly, the light curve of V1309 Sco exhibited a slow, steady brightening beginning approximately five years prior to the merger. Similar precursor emissions have been observed in extragalactic LRNe (Kankare et al. 2015; Blagorodnova et al. 2017, 2020; Pastorello et al. 2019, 2021a), though their greater distances and faintness have made it difficult to achieve equally detailed sampling. It remains an open question as to whether this precursor phase reflects continuous or episodic mass loss in the years leading up to the major outburst, potentially signalling the early stages of a binary system merger. Long-term observations of LRN progenitor stars are therefore crucial for testing their binary origins and gaining valuable insights into pre-merger mass loss episodes in these systems.

The primary challenge in detecting precursors is their modest absolute magnitude, which means most extragalactic LRNe are only identified during their brightest outbursts. The lack of deep archival data further complicates efforts to detect precursor brightening. To overcome this, a comprehensive catalogue of potential progenitors in nearby galaxies (i.e. a large and deep survey) is essential. While early discoveries of LRNe were limited to the Milky Way and M31 (e.g. V4332 Sgr and V838 Mon; Martini et al. 1999; Tylenda & Soker 2006; M31 LRN-2015, Williams et al. 2015; Blagorodnova et al. 2020), the advent of large synoptic surveys, such as the Asteroid Terrestrial-impact Last Alert System (ATLAS, Tonry et al. 2018), the Zwicky Transient Facility (ZTF; Bellm et al. 2019), MeerLICHT (Bloemen et al. 2016), and BlackGEM (Groot et al. 2019, 2024), has paved the way for the systematic detection of LRNe throughout the Local Group and beyond (e.g. Karambelkar et al. 2023). The upcoming Vera C. Rubin Observatory, with its Legacy Survey of Space and Time (LSST), promises to revolutionise transient astronomy, providing an unprecedented opportunity to detect faint LRNe and LRN precursors across vast distances. With LSST expected to generate 10 million transient alerts daily, the primary challenge will be identifying LRN precursors amidst this vast amount of data unless a targeted strategy is developed to detect them.

We present a strategy for identifying LRN precursors by cataloguing YSG candidates within 20 Mpc, leveraging archival Hubble Space Telescope (HST) photometry. Our goal is to identify potential precursor candidates by cross-matching future transient alerts with this catalogue, offering a new approach to studying the processes driving LRN precursor brightening and triggering mechanisms. An initial study provided a similar census of YGs and YSGs in the Milky Way (Addison et al. 2022). Using Gaia Data Release (DR) 2 and Early Data Release 3 (EDR3) sources (Gaia Collaboration 2018, 2021), they modelled the distribution of a sample of Milky Way stars in the colour–magnitude (MG vs BP – RP) diagram. Statistical modelling allowed them to select the Hertzsprung gap as the least-populated region of the diagram. As a result, they identified 21 candidates exhibiting signs of a steady increase in brightness. Most of these candidates showed Balmer lines in emission and an infrared excess in their spectral energy distributions, further evidence of an accreting system with dust production. Interestingly, one of them turned out to be the precursor of a type-I X-ray burst (Reig et al. 2022).

While the census of Galactic YSG candidates is crucial, the rate of LRNe is comparable to that of core-collapse SNe (Karambelkar et al. 2023), suggesting that the largest population of LRN precursors may be found beyond our Galaxy. To address this, we applied a similar methodology to catalogue extragalactic YSGs using archival HST photometry of galaxies within 20 Mpc. If a future transient alert matches an object in this catalogue, it could be flagged as a potential precursor and prioritised for follow-up observations. This approach offers a practical strategy to for sifting through the vast number of LSST alerts and conducting the first in-depth study of the processes driving LRN precursor brightening.

The paper is structured as follows: In Sect. 2 we define the data and methods used to select the samples and remove contaminants. Section 3 presents a statistical description of the selected sample and the different sources of contamination. We crossmatch this catalogue with LRN progenitors from the literature, the Transient Name Server1 (TNS), and time-domain surveys in Sect. 4, showcasing some individual precursor candidates. These results are discussed in Sect. 5 and compared to previous searches for YSGs and LRN progenitors. We conclude and give some guidance on how to use this catalogue in Sect. 6.

2 Methods

2.1 Selection criteria

To identify progenitors of transients produced by massive stars, we needed a YSG sample complete down to 8 M⊙ . This completeness is key for future population studies. According to MESA (Modules for Experiments in Stellar Astrophysics; Paxton et al. 2011) stellar models, an 8 M⊙ star crossing the Hertzsprung gap has an absolute magnitude of MF814W ∼ –5. HST can detect such a star up to 20 Mpc away with a 1-hour exposure (mlim,F814W ~ 27.5). Shorter typical exposures (e.g. 600 s) only detect YSGs down to 16 M⊙ at this distance. Thus, our sample is limited to galaxies within 20 Mpc with deep HST exposures in several bands, allowing us to conduct a census of stars in the Local Universe up to the Virgo cluster, and ensuring a minimum completeness limit of 16 M⊙ to cover the Hertzsprung gap.

To get a complete sample of nearby galaxies, we used the Heraklion Extragalactic Catalogue (HECATE; Kovlakas et al. 2021), which is based on the HyperLEDA database (Paturel et al. 2003), but supplemented with robust distance estimates, galaxy mass, metallicity and star formation rate. The base sample, containing 1883 galaxies closer than 20 Mpc, is cross-matched with HST observations available on the Mikulski Archive for Space Telescopes (MAST), to find galaxies with overlapping deep (texp > 300 s) exposures in two optical filters2. Only galaxies having at least 2% of their area covered by HST are considered, discarding de facto the Magellanic Clouds3. The list of selected filters is detailed in Table 1, together with the number of galaxies selected for each filter. We selected the F475W, F555 W, F606W, and F814W filters because they have responses close to the Johnson–Cousins V (F475W, F555W, and F606W) and I (F814W) filters, allowing us to study V against V – I colour–magnitude diagrams (CMDs). The resulting HST-observed sample contains 575 galaxies. The distributions of distance and angular sizes are shown in the top panel of Fig. 1.

Table 1

HST filters and galaxy selection.

2.2 HST data retrieval

To retrieve HST sources, we used two databases comprising third-party catalogues in which source extraction has already been performed. The third release of the Hubble Source Catalog (HSCv3; Whitmore et al. 2016) includes Advanced Camera for Surveys (ACS) and Wide-Field Camera 3 (WFC3) data that were public as of 2017 October 1. As such, it contains data for 449 of our galaxies observed before this date. HSCv3 data were retrieved using MAST queries with the CasJobs API4. For 118 additional galaxies, whose observation requirements are met only after 2017, we retrieved their source catalogues from MAST. At the time of the writing, eight galaxies could not be retrieved in these databases, due to proprietary exposures or missing source catalogues. As an example, the southern region of M31, which has sparse coverage in HSCv3, is included in the MAST retrieval. This leads to uniform coverage of the inner galaxy, as shown in Fig. 2.

The MAST third party catalogues were produced at the time of the data reduction of the corresponding observations, using DAOPHOT (Stetson 1987), SExtractor (Bertin & Arnouts 1996), or the Hubble Advanced Products pipeline (Tran et al. 2020). For both types of catalogues, magnitudes are retrieved in the large aperture photometry (MagAper2). We made sure that MAST catalogues are reliable and have well-calibrated photometry, by comparing their sky coverage and their CMD with HSCv3 data for two test galaxies (NGC 45 and ESO 209-9). In all tested filters, the bias and standard deviation between HSCv3 and MAST photometry is <0.05 mag and <0.2 mag, respectively. Once downloaded, all catalogues of a given galaxy are merged and cross-matched to obtain a single ‘master’ catalogue, using astropy v5.3.3 (Astropy Collaboration 2013)5. To discard compact star clusters, cosmic rays and sources affected by confusion, we discarded sources with concentration indices (CIs) <0.85 or >1.5 or magnitude errors >0.1.

thumbnail Fig. 1

Distribution of some properties of the sample of 575 HST-observed galaxies. Top: semi-major axis, a, and distance from HECATE. Bottom: metallicity and Galactic extinction (see the main text for details).
thumbnail Fig. 2

Density of HST sources under study for the galaxy M31, as retrieved through the HSCv3 and MAST databases. The background image is from the Digitized Sky Survey 2 (Lasker et al. 1996).
Table 2

MIST parameter grid.

2.3 CMD analysis

2.3.1 Stellar evolution tracks

To analyse the CMD of galaxies, a physical reference is needed, such as isochrones or stellar evolution tracks in the observer’s frame. However, the shape of an observed CMD heavily depends on intrinsic and extrinsic parameters. Intrinsic parameters include the galaxy star formation history and its metallicity, while extrinsic parameters correspond to measurement uncertainties and the extinction along the line of sight, commonly modelled with the extinction parameter Aλ. Using the synthetic photometry of stellar evolution tracks provided by MIST (MESA Isochrones & Stellar Tracks6; Dotter 2016; Choi et al. 2016), we generated tracks for a grid of zero-age main sequence (ZAMS) stellar masses, extinctions, and metallicities, summarised in Table 2. The ranges of extinctions and metallic- ities correspond to the ranges of estimated Galactic extinctions and metallicities of galaxies in our sample, and the mass range encompasses most of known LRNe progenitors (e.g. MacLeod et al. 2022). The upper limit of the mass range does not restrict the YSG selection; instead, it marks the point beyond which stellar masses become poorly estimated, as the Hertzsprung gap can no longer be consistently defined as a continuous post-MS phase. It is important to note that our selection will contain YSG regardless of their stellar evolution phase, either post-MS or post-red supergiant. While some YSG have been identified as post-red supergiant in recent studies (e.g. Humphreys et al. 2023), we cannot distinguish between them in our catalogue. However, post-red supergiant YSGs typically show only a fast, low-amplitude variability; therefore, we expect a stable behaviour in future LSST observations.

To estimate extinctions along the line of sight for each HECATE galaxy, we used the 2D dust map of Delchambre et al. (2023) released as part of Gaia DR3. Metallicities and galaxy masses are provided in HECATE for only 15% (276) and 62% (1159) of our base sample, respectively. To estimate metallicities, we therefore relied on the absolute B magnitude – oxygen abundance relations of Pilyugin et al. (2004) (Eqs. (12) and (15) for spiral and irregular galaxies, respectively7): 12 + [O/H] = min(5.8 - 0.139 Babs, 6.93 - 0.079 Babs). Gas-phase oxygen abundances were converted to metallicity using the relation Z = [M/H] = 12 + [O/H] – 8.69, where 8.69 is the solar oxygen abundance given in Asplund et al. (2009). The resulting distributions of Galactic extinction and metallicity are shown in Fig. 1 (bottom panel). In the following, all CMDs are in the Vega magnitude system, aperture-corrected (using the table8 recommended by the HSCv3 documentation) and de-reddened of Galactic extinction.

thumbnail Fig. 3

Example of CMD analysis: extinction-corrected CMD ofIC 1613 sources with MIST tracks for a metallicity of Z=–1.2 and an extinction of AV=0.2. Hertzsprung gap candidates are selected above the blue lines. The dotted contour shows the Gaussian mixture model representing the data.

2.3.2 Hertzsprung gap definition

The Hertzsprung gap of each galaxy is identified using the MIST tracks at different masses. Specifically, we define this gap as the time interval between two points, the post-MS bright turnoff (the point when the V -band luminosity first reaches 95% of its post-MS bright turnoff value), and the pre-RGB faint turnoff, as illustrated in Fig. 3 by the left-hand and right-hand solid blue lines. To account for the internal extinction of the galaxy and retrieve extinguished YSG candidates, we artificially applied a 0.4 mag extinction to the red-side selection cut, using a relative visibility value of R(V) = 3.1. Although this additional extinction is arbitrary, it ensures the retrieval of most extinguished candidates, as discussed in Sect. 2.4. YSG candidates are searched between these lines and above the MIST track of a reference ZAMS star mass, which ranges from 3 to 16 M⊙ depending on the galaxy. To ensure the completeness of the selected samples at this reference mass, it was chosen to be the lowest mass for which the faint end of the Hertzsprung gap is well observed (i.e. significantly brighter than the sensitivity of HST exposures). Consequently, we excluded 102 galaxies for which the available HST data are too shallow to accurately detect YSGs with masses of 16 M⊙ or below.

2.3.3 Statistical modelling

Due to the significant uncertainties in the estimate of distance, metallicity, Galactic extinction, HST photometry of faint sources, and stellar evolution models, a mismatch between the true locus of the observed Hertzsprung gap and its MIST estimate is inevitable for some galaxies. In particular, some galaxies have part of their MS and/or RGB intersecting the selected area of the CMD. To exclude these densely populated regions, we adopted a similar approach as the YSG selection method of Addison et al. (2022). The CMD of each galaxy is statistically modelled as a mixture of Gaussian components (Gaussian mixture). The number of components is determined based on the logarithm of the number of sources, balancing the need to avoid overfitting in sparse CMDs while allowing for the fitting of complex structures in dense CMDs. To better retrieve the tails of the distribution (i.e. CMD features such as the RGB), the faintest and most populated regions of dense CMDs, as estimated by kernel density estimation, were artificially depopulated by a factor of up to 20. This step only removes sources in faint regions and not the YSG selection area.

To exclude the dense regions of the CMD, sources located within the Gaussian mixture contour of a reference likelihood are discarded. This contour intersects the Hertzsprung gap of the reference MIST track, with the likelihood value optimised through visual inspection. The resulting contour for IC 1613 is shown by the dotted line in Fig. 3, where the reference MIST track corresponds to a 3 M⊙ star. The RGB is well outside the YSG selection area. A representative subset of the CMDs analysed in this study is shown in Appendix A. This statistical modelling requires a minimum of 100 sources to provide a reliable CMD match. Therefore, we discarded 96 galaxies, leaving 369 galaxies in our sample.

2.4 Sample cleaning

2.4.1 Removing known contaminants

Different types of contaminants may be included at this stage of the selection. First, the line-of-sight chance alignment of Milky Way stars, notably yellow dwarfs and white dwarfs, with the galaxy under consideration can cause them to appear within the Hertzsprung gap of its CMD. Likewise, background objects such as quasars can also appear in this locus. Second, the extinction affecting MS stars inside each galaxy can displace them from the MS to the Hertzsprung gap in the observed CMD. Additionally, objects other than YSG also naturally appear in the Hertzsprung gap, such as Cepheids and other variables within the instability strip, luminous blue variables (LBVs), or unresolved globular clusters (e.g. Kraft 1966; Justham et al. 2014; Mora et al. 2007).

As an initial cleaning process, we cross-matched between our sample and the SIMBAD9 and Gaia DR3 (Gaia Collaboration 2023b) catalogues using a matching radius of 0.5″. SIM- BAD quasars, globular clusters, and classical Cepheids were discarded. Additional quasars were found and excluded using the Milliquas (Flesch 2021) and Gaia Extragalactic (Gaia Collaboration 2023a) catalogues. However, we flagged known LBVs, YSGs, and red supergiants (RSGs) for further consideration within the sample. Sources exhibiting significant Gaia proper motion (PM∕PMerr > 4) are identified as Galactic and discarded. Although other types of contaminants cannot be directly excluded, their presence can be quantified statistically.

2.4.2 Quantifying foreground contamination

To estimate the line-of-sight foreground contamination towards each galaxy, we utilised the TRILEGAL (TRIdimensional modeL of thE GALaxy; Girardi et al. 2005) simulation, which models the Milky Way’s stellar content, incorporating the Sun’s position and various Galactic components. It includes the thin disc, thick disc, halo, and bulge, along with their star formation histories. This simulation, notably used by Dal Tio et al. (2022) to simulate the Milky Way as observed by the upcoming LSST survey, and accessible through an API from NOIRLab10, simulates the photometry and stellar parameters of the observed Galactic population along designated lines of sight. Patches of 0.25 deg2 were queried for each galaxy in our sample. Given that LSST’s coverage is mostly restricted to the Southern Hemisphere (Dec<0), some galaxies lie outside the simulated regions. For cases where LSST covers the symmetric position relative to the Sun-Galactic poles plane, we used this position instead, assuming similar results due to the symmetric nature of TRILEGAL components. This planar symmetry was verified using test coordinates through the TRILEGAL web interface11. For the 30% of galaxies (106 in total) still uncovered, we utilised the TRILEGAL web interface with the same parameters as Dal Tio et al. (2022). Notably, the halo profile in the web interface differs from the LSST simulation, employing a r–1/4 profile versus Dal Tio et al. (2022)’s r–1/2.5 profile. We adjusted for this by approximating the profile with r–1/4, Ω=0.0025 M⊙ pc−3 and rh=5700 pc, based on comparable star counts and magnitude distributions.

TRILEGAL sources matching the footprint of the selected HST exposures and exhibiting synthetic HST photometry within the specific Hertzsprung gap selection region are identified as contaminants. Figure 4 illustrates this method for M83, showing the CMD of TRILEGAL sources overlapping with HST exposures and overlaying the observed YSG candidates. The results of this foreground contamination analysis are presented in Sect. 3.2.

thumbnail Fig. 4

Illustration of the estimation of the foreground contamination rate using the TRILEGAL Milky Way simulations. The CMD shows TRILEGAL sources overlapping with the footprint of M83 HSC sources. Highlighted in red are the YSG candidates identified within this sample. TRILEGAL sources falling within the black polygon, circled in red, are identified as foreground contaminants. They represent 3.1% of the M83 YSG candidates.

2.4.3 Extinguished main sequence stars

Although we accounted for the Galactic extinction along the line of sight for each galaxy using the appropriate MIST model, internal extinction inside each galaxy is not modelled at this stage. Consequently, a significant fraction of Hertzsprung gap sources may actually be extinguished MS stars. To quantify this, we assigned each candidate a probability of being an extinguished MS star. Assuming a normal distribution for extinction, we measured extinction scatter by evaluating the standard deviation of AV in a thin region of the CMD, chosen to be the RGB, as illustrated in Fig. 5. This standard deviation is used to model the additional extinction in the selection region, resulting in a probability map for each CMD (Fig. 5, bottom panel). This probability is naturally highest at the blue edge of the selection region, which separates the MS region from the Hertzsprung gap. By averaging these probabilities on a per-galaxy basis, we can determine the contamination rate due to extinction. The distribution of AV scatters estimated for all selected galaxies is shown in Fig. 6. Besides providing contamination probabilities, it also suggests that the 0.4 mag extinction applied to the red edge of the selection region is sufficient to recover most of the extinguished YSG (Sect. 2.3.2). As a result, some RSGs are included in the sample, as discussed in Sect. 5.1 for the case of M31. However, obscured YSGs and RSGs have also been identified as progenitors of certain infrared transients thought to share similarities with LRNe (Jencson et al. 2019).

thumbnail Fig. 5

Estimation of IC 1613 internal extinction. Top: density plot of the RGB of IC 1613 in the magnitude vs. relative AV plane. The x-axis zero point corresponds to the mean of the distribution. The top curve shows the kernel density estimation of AV values, with a standard deviation of ± 0.21 (dotted lines). Bottom: resulting map showing the probability of a YSG being an extinguished MS star in the selection region of YSG sources in the IC 1613 galaxy.

3 Results

3.1 Selected candidates

Our analysis identified a total of 154 494 YSG candidates before cleaning, a portion of which are listed in Table 3. The cumulative distribution of their distances is depicted in Fig. 7, indicating that about 25% of these candidates are located in galaxies at approximately 0.8 Mpc. These candidates are actually located in M31 (40 566 candidates) and M33 (11 603), the two galaxies with most candidates. The three next galaxies with the most YSG candidates are M101 (8885), NGC 300 (7525), and NGC 253 (4000).

As a result of the CMD analysis, we identified YSG candidates in 353 galaxies. The HST data provenance, available blue/green filter, coverage, and the parameters used for fitting the CMD as described in Sect. 2.3 are provided for each galaxy in Table 4. This table also details the number of candidates for each galaxy and the estimated contribution from different contaminants.

YSG candidates exhibit apparent magnitudes in the F814W band ranging from 16 to 25, with a peak at approximately 22.2 (Fig. 8). Single-star ZAMS stellar masses, luminosities, and temperatures for these candidates were estimated by interpolation using the MIST stellar evolutionary tracks within the Hertzsprung gap. The distribution of stellar masses, shown in Fig. 8, appears bimodal: a lower-mass peak, around approximately 5 M⊙, primarily corresponds to candidates in M31, M33 and NGC 253, and a higher-mass peak, around approximately 10 M⊙, corresponds to more distant candidates. This bimodality is further illustrated in Fig. 9, showing the tight correlation between interpolated stellar masses and F814W absolute magnitude.

thumbnail Fig. 6

Distribution of the standard deviations of extinction for our galaxy sample.
Table 3

YSG candidates (extract).

Table 4

Galaxies with more than 400 YSG candidates, sorted by distance (extract).

thumbnail Fig. 7

Cumulative distribution of distances for YSG candidates before and after cleaning, and the estimated number of extinguished MS stars.

3.2 Contamination fraction

To evaluate the contamination fraction from internal extinction and foreground sources, we considered various types of contaminants. Foreground contamination was quantified using the TRILEGAL star counts. According to the method detailed in Sect. 2.4, approximately 5% of the sources in our YSG sample are expected to be foreground contaminants (including sources having high Gaia proper motion). The foreground contamination mainlyaffects galaxies at low Galactic latitudes (|b| < 25°); for 75% of the galaxies, fewer than 10% of their YSG candidates are foreground contaminants, as shown in Fig. 10 (right panel). As illustrated in this figure, galaxies closer to the Galactic plane or towards the Galactic centre exhibit higher contamination fractions. After using Gaia proper motions to clean the sample, removing 6623 sources (4.3% of the sample), we therefore estimate the remaining foreground contamination to be ≲1%.

Additionally, the probability of each YSG candidate being an extinguished MS star was computed using the modelled internal extinction (Sect. 2.4). The mean probability for all sources in our YSG catalogue is 0.25, with the fraction of contaminants remaining constant across all distances (Fig. 7). As expected, this probability is significantly influenced by the temperature of the source, and the filter used in the CMD analysis, as depicted in Fig. 11.

Finally, contaminants such as quasars, Cepheids and LBVs were identified and removed using Gaia and SIMBAD databases. These known contaminants represent a minority of the sample across all distances (Fig. 7). These sources are typically brighter in both observed and absolute magnitudes, as shown in Fig. 12. The clean sample of YSG contains 146 502 sources.

thumbnail Fig. 8

Distributions of observed F814W magnitude and stellar masses of YSG candidates.
thumbnail Fig. 9

MIST-inferred ZAMS stellar masses of YSG candidates as a function of their absolute F814W magnitude.
thumbnail Fig. 10

Foreground contamination rates per galaxy. Left: foreground contamination rate as a function of Galactic latitude derived from the TRILEGAL Milky Way simulations. The colour encodes the Galactic longitude. To maintain readability, only galaxies with a minimum of 100 YSG candidates are included. Right: empirical cumulative distribution function (ECDF) of the foreground contamination rate. For 75% of the galaxies, fewer than 10% of their YSG candidates are foreground contaminants.
thumbnail Fig. 11

Probability of YSG candidates being an extinguished MS star as a function of their MIST temperature. Three example galaxies are shown, illustrating the impact of the HST blue/green filter used to analyse the CMD.
thumbnail Fig. 12

Fraction of identified contaminants using SIMBAD and Gaia catalogues as a function of observed and absolute F814W magnitudes.
Table 5

Retrieval status of known LRN progenitors in the YSG sample.

4 Applications

4.1 Completeness for known LRN progenitors

Many extragalactic LRNe discovered during the last decade had pre-outburst images taken several years before the transient (e.g. MacLeod et al. 2022 and references therein). For eight of them, listed in Table 5, their host galaxy is in our sample, allowing us to assess the rate of retrieved progenitors in our sample of YSG candidates. Three out of eight progenitors are missing in our sample. For AT2015dl, the transient is located outside the footprint of the HST observations. For NGC4490-OT, although the progenitor was measured and studied by Smith et al. (2016), its apparent magnitude of mF606W = 23.58±0.24 was too faint to be retrieved as an HSCv3 source (the faintest retrieved sources in a 30 arcsec radius circle around the progenitor position were 23.2 mag). For AT2019zhd, the progenitor had a low luminosity (MF555W = 0.21 ± 0.14; Pastorello et al. 2021a), placing it in the most populated region of the Hertzsprung gap. Consequently, it could not be selected after the Gaussian mixture cut. Moreover, its proximity to the edge of the field of view prevents this source from being retrieved as a MAST-catalogued source, and either way its magnitude error would exclude it from our sample. Considering these results, our sample appears to represent a relatively comprehensive selection of LRN progenitors, with a ∼70% completeness level in regions covered by HST.

4.2 Progenitors of past and ongoing transients: Cross-match with TNS

Our YSG sample can be used to find possible progenitors for past transients and further analyse them. To this end, we crossmatched our YSG candidates to the TNS public objects as of 2024 August 31, with a match radius of 0.6″. The resulting list of 24 transients is detailed in Table 6. About half of these transients are already classified and some are well studied in the literature (e.g. SN2016bau: Aryan et al. 2021; SN2024ggi: Pessi et al. 2024; Jacobson-Galán et al. 2024). For those, our HST matches may provide useful pre-outburst luminosity levels. The TNS report of AT2019ejn mentions multiple discoveries of this transient, suggesting it may be an LBV. For three other transients (AT2020aaqy, AT2021tmm, and AT2021ytf), the F814W magnitude is brighter than the TNS discovery magnitude, making them likely to also be variable stars. For nine other transients, the type was not straightforwardly identified in TNS and they deserve a detailed analysis of their light curves and HST progenitors. Therefore, we inspected HST colour images for every candidate, using the Hubble Legacy Archive (HLA) website12. Cutouts of 10 arcsec side-to-side are displayed in Fig. 13 and the position of the transient is marked by a circle.

To inspect the temporal evolution of the candidates, we obtained forced photometry for the ZTF public data using the online service ZTF Forced Photometry Service (Masci et al. 2023). For every candidate, we retrieved all the data starting from the beginning of the public survey in March 2018 (around MJD 58178). We generally used the difference imaging flux obtained during the first year of operations (or a period with low residual flux) as a baseline to calibrate the difference imaging magnitudes. We applied the quality cuts recommended in the documentation and we imposed forcediffimchisq (reduced chi-square in point-spread-function fit) <1.3. Data points with S/N>3 were considered as detections and the upper limits are reported with a 5σ threshold. To increase the S/N for these faint objects, we binned the fluxes using a 15-day bin size. The resulting light curves are shown in Fig. 14. In the following, we report the results of a detailed analysis aimed at identifying potential LRN precursors.

AT2018mmb: This transient is located at 0.55″ from a faint (mF555W = 22.78, mF814W = 21.68) YSG in our sample, inside a stellar cluster. However, it is in a crowded region, surrounded by many sources that may cause source confusion (CI = 1.4), and a significantly brighter source (mF555W = 19.07, mF814W = 19.38) is located at just 0.37″ from the transient. The position of this bright source in the CMD is close the locus of the MS. Although this neighbour is the most likely progenitor of the transient, its recent brightening in 2024 (reaching an absolute magnitude of –9) makes it an interesting candidate for follow-up.

AT2019krl: This transient matches a YSG candidate with a separation below 0.1″ and a stellar mass ∼10.3 M⊙. Image inspection shows that the YSG is devoid of any bright source in its immediate vicinity. AT2019krl was spectroscopically classified as an SN IIn or LBV outburst by The Astronomer’s Telegram, No. 12913. However, the absolute magnitude peak at −8.5 is several magnitudes fainter than most SNe IIn. It was detected by HST at an absolute magnitude of −6, translating to a ZAMS stellar mass of ∼10 M⊙, less massive than most LBV. Still, their measured Hα line velocity (~2000 km/s) rules out a LRN-related mechanism.

AT2020adbp: This source matches a YSG in M31 observed by HST in January 2023, with mF814W = 18.7. This observation occurred after the transient report on TNS, explaining the bright HST magnitude. There is no pre-brightening image available in the HST archive, which precludes the study of the progenitor. Source confusion is unlikely for this source having CI = 1.1 in both F814W and F475W filters. follow-up of this source is encouraged.

AT2021ahsv: This transient matches a YSG candidate located at 0.19″ with M∗ ∼ 12 M⊙. Itis brighter than every close- by source. Standing in the middle of the Hertzsprung gap in the CMD, this candidate progenitor shows an interesting light curve with a slow brightening. follow-up of this source is encouraged, as it has shown a steady brightening since 2020.

AT2022llt: This transient matches a >20 M⊙ YSG with a separation of 0.06″. Although ZTF forced photometry does not reveal any long-term behaviour, the ATLAS light curve shows a slow brightening of the source in recent years (Fig. 16, top middle panel). Image inspection shows that source confusion is unlikely for this source (also supported by its CI=1.05). follow-up of this source is encouraged.

AT2023azz: This transient is located at 0.51″ from a ~9.6 M⊙ YSG candidate in M101, according to its TNS coordinates. The ZTF coordinates, taking advantage of a large number of epochs, are even more precise and point to a separation of 0.2″. Image inspection reveals a faint, red source close to a brighter one (at 0.75″). However, confusion may affect the photometry of this source, given the average CI of 1.4. Besides, its location on the CMD makes it also compatible with a RSG. follow-up of this object is still encouraged.

AT2023wgz: This transient has a massive (>20 M⊙) YSG counterpart at a 0.58″ separation. Its photometry may be moderately affected by source confusion, with a CI of 1.4 in both F814W and F555W filters. Otherwise, it is not surrounded by any source of similar or greater luminosity. Its light curve shows only a few data points, with no detection since 2021.

AT2023wot: This source matches a ∼4 M⊙ YSG in M31 with a 0.58″ separation. Although it is on the faint end of our M31 selection, image inspection shows that it is not surrounded by any source of similar or greater luminosity, ruling out source confusion (CI=1.13). The Astronomer’s Telegram, No. 16319 suggest it to be a nova (Hornoch et al. 2023), which is in agreement with the absolute magnitude at peak of the forced photometry light curve (~–5.3; Fig. 14). Such a rapid outburst also supports a nova-like behaviour.

AT2024ikg: This recent transient is found to match a bright (mF555W = 21.73), massive (∼16M⊙) YSG candidate located at 0.33” in M101. Querying the HSCv3 detailed catalogue reveals that it was observed in January 2003 and October 2013, the only repeated filters being F814W and F 435W. The source brightened from 22.06 to 21.65 (F814W) and 22.62 to 21.35 (F435W) in this interval. Inspection of the image reveals a bright RSG located just 0.25″ away from the YSG, but at fainter magnitude (mF555W = 23.05). Besides, the CI in both F606W and F814W is measured between 0.95 and 1.2 in both epochs, making source confusion unlikely. follow-up of this object is encouraged.

Table 6

YSG candidates that match TNS objects.

thumbnail Fig. 13

HST cutouts of the 12 candidates resulting from our TNS cross-match. Images are 10″ side-to-side, with a circle of 1″ radius pinpointing the location of the YSG source. Adapted from HLA colour composites.

4.3 Other precursor candidates

Luminous red nova precursors are expected to rise by only a few magnitudes in several years (e.g. Blagorodnova et al. 2020). In this context, they may not meet the criteria to be reported to TNS (criteria may vary depending on the collaboration) until long after their detectability has begun. To retrieve them, we created a Lasair13 (Smith et al. 2019) watch list to match our YSG catalogue to all ZTF transient alerts with a matching radius of 0.6″, and a Lasair filter to keep only objects with at least 2 positive detections (ncandgp ≥ 2).

To exclude variable stars, we inspected the ZTF light curve of each of the resulting 67 sources using the AlerCE broker (Förster et al. 2021), discarding all sources showing magnitudes in the past at levels brighter or similar to present levels. We used the ATLAS forced photometry tool14 on difference images to query candidate precursors and confirm the brightening trend on longer timescales. At this stage, we obtained nine precursor candidates consistently brightening over the last few years, including four objects previously analysed in Sect. 4.2 (AT2020adbp: ZTF20abhtvor, AT2021ahsv: ZTF20abbetli, AT2023azz: ZTF23aaazair, and AT2024ikg: ZTF24aalfiak).

To obtain photometry for southern candidates, we crossmatched YSG candidates to MeerLICHT and BlackGEM transients that have at least two detections with a reasonable signal-to-noise ratio (S/N > 6.5) and a high probability of being real (rather than bogus), class_real > 0.8. The matching radius used was 1″. For BlackGEM and MeerLICHT, we obtained 10 and 33 matches, respectively. The large majority of them are variable stars with no clear brightening trend. For three of them, however, a brightening is identified and confirmed with ATLAS forced photometry: MLT28037995, in NGC 300, MLT15613547, in NGC 55, and MLT17180523 in NGC 253, actually corresponding to AT2022llt. The complete list of 12 (ZTF+MeerLICHT+BlackGEM) precursor candidates is detailed in Table 7. Their HST cutouts are presented in Fig. 15 and their multi-survey light curves are shown in Fig. 16. We fitted a slope to each light curve, in order to quantify the brightening trend. Slope values range between 2.1 × 10−4 and 1.2 × 10−3 mag day−1, with an average error of 1.3 × 10−4 mag day−1. In comparison, the brightening of the precursors of M31- LRN2015 and M101 OT2015-1 were respectively 3 mags over 2 years and 1.5 mags over 6 years (Blagorodnova et al. 2017, 2020), corresponding to slopes of ~4 × 10−3 mag day−1 and ~7 × 10−4 mag day−1, respectively.

thumbnail Fig. 14

ZTF forced photometry light curves of the nine candidates resulting from our TNS cross-match. The blue areas indicate the period that was used to set the baseline flux at the location of these transients.
Table 7

Precursor candidates identified in this study.

5 Discussion

5.1 Comparison of our YSG candidates to spectroscopic samples

The literature provides a comprehensive representation of the brightest stars in M31 and M33, as well as their nature. To assess the validity of our method and further investigate the completeness and reliability of our YSG catalogue, we compared our sample to the spectroscopic YSG and RSG samples in M31 and M33 from Drout et al. (2009), Drout et al. (2012), Gordon et al. (2016), and Massey et al. (2016). In M31, within the CMD region used to select YSGs, Drout et al. (2009) identified 120 probable YSGs and 2772 foreground dwarfs. This identification was achieved by comparing the radial velocity obtained from their spectra with the expected radial velocity at each position in M31, considering its peculiar velocity and rotation curve. Similarly, in M33, they identified 135 YSGs and 781 dwarfs (Drout et al. 2012). In a redder and less luminous region of the CMD, they identified 204 probable RSGs and 204 foreground dwarfs. The studies by Gordon et al. (2016) and Massey et al. (2016) refined these samples by identifying spectral types and members through their spectroscopic campaigns, including the extensive Local Group Galaxy Survey (LGGS; >1800 spectra).

Conversely, we used Gaia proper motions to eliminate foreground contaminants. To evaluate the effectiveness of this method, we cross-matched the samples from Drout et al. (2009, 2012) and Gordon et al. (2016) with Gaia DR3. For a minority of classified dwarfs, proper motion data are unavailable, constituting an anomaly. These stars are removed from the sample. The distribution of proper motion S/N is shown in Fig. 17, categorised by the type of star they identified. The vast majority (98.5%) of foreground sources are correctly eliminated using Gaia’s proper motions. Additionally, 96.9% of the eliminated sources are indeed foreground stars. Gaia proper motions prove to be an effective method for detecting foreground stars, comparable to spectroscopy or radial velocity methods.

Comparing the samples from Drout et al. (2009, 2012) and Gordon et al. (2016) to our sample, we obtain 931 associations (separation of less than 0.6″) for classified stars. The foreground contamination in this magnitude range is substantial, with approximately 89% of selected stars being foreground stars. After our cleaning, 96% of them are eliminated, and only 4% of M31’s YSGs are removed. Therefore, the completeness of our sample is preserved, leaving only 23% residual contamination in this bright sample (typically V < 19). At fainter magnitudes, contamination is less significant. In M33, our method performs equally good. The pre-cleaning contamination is 56% (31 out of 56), and none of these 56 stars remain after filtering out high proper motions.

Red supergiants represent another type of contaminant in our sample. This is a direct consequence of the selected regions in the CMD, which include a portion of the RGB to avoid missing YSGs affected by extinction. In literature samples, the selection region for RSGs is at a lower magnitude than for YSGs in the CMD, complicating the estimation of the fraction of RSG contaminants. At V < 18.5, however, our selection includes 44 YSGs and 14 RSGs, resulting in a contamination rate of 24%.

In addition to using proper motions, variability is an independent criterion that helps in distinguishing between YSGs and foreground stars. For example, massive YSGs have been known to vary erratically compared to low-mass dwarfs. The bottom panel of Fig. 17 shows the distribution of the Gaia-measured dGmag, the difference between brightest and faintest G-band magnitude observed during the course of the survey. Interestingly, stars classified as foreground but with a low PM exhibit variability on average 6 times greater than those with proper motion data (average dGmag=0.2 instead of 0.033; Fig. 17). This suggests a misclassification of these stars, which may be actual members of M31. To test this hypothesis, we modelled the distribution of dGmag as a sum of two distributions: one for high-PM sources identified as foreground and one for low-PM sources identified as YSGs and RSGs. We could then estimate the contribution of YSGs and RSGs within low-PM foreground-classified stars. The fraction of these candidate members is 68 ± 11%. This suggests that the residual contamination within low-PM stars with V < 18.5 is actually ~8% and not 23%. Furthermore, if our sample were to be used to identify progenitors of variable or transient events, the expected value of dGmag would be relatively large, further reducing the probability of an association with a foreground source. On the other hand, although we expect most background active galactic nuclei to be excluded in our sample selection (Sect. 2.4), some unresolved active galactic nuclei undetected in X-rays may still produce some alerts in future deep transient surveys.

Finally, we can quantify the completeness of our YSG catalogue in different magnitude bins. Using for reference the samples of YSG candidates in M31 from Drout et al. (2009) and (Gordon et al. 2016, see the corresponding CMD in our Fig. 18) and applying the same proper motion cut as done in our study, we obtain the completeness estimates shown in Fig. 19. While the brightest stars are not always detected in third-party HST catalogues (and surveys other than HST are well adapted to recover those), the fraction of recovered YSGs among recovered HST sources consistently remains around 60%. The lost fraction is primarily due to sources with high CI or located in the dotted- line rectangle shown in Fig. 18. The latter occurs because our conservative threshold, designed to exclude the MS and some extinguished MS stars, also omits YSGs in the bluer part of the CMD.

thumbnail Fig. 15

HST cutouts of precursor candidates. Images are 10″ side-to- side with a 1″ radius circle centred on the HST position.
thumbnail Fig. 16

ATLAS forced photometry light curves of the best 12 precursor candidates in ZTF, MeerLICHT, and BlackGEM. ATLAS photometry is represented by orange and cyan circles (o and c bands), ZTF photometry by red and green diamonds (r and g bands), and MeerLICHT/BlackGEM photometry by black and brown triangles (q and i bands). ATLAS light curves were re-binned to 60-day bins, other surveys to 15-day bins.
thumbnail Fig. 17

Gaia properties of YSG candidates in M31 and M33. Top: distribution of the Gaia DR3 proper motion signal-to-noise for M31 and M33 stars classified in the literature. Bottom: distribution of the variability indicator dGmag for M31 stars. The stars classified as foreground stars but with a low proper motion are shown in blue.
thumbnail Fig. 18

CMD of M31 stars classified in the literature, using the photometry of Massey et al. (2016). Smaller markers without contours are stars present in the HST footprint but not in HST catalogues, smaller markers with grey contours are stars in HST catalogues but not in our YSG selection, and larger markers with black contours are the selected YSG candidates. The dotted lines highlight a region encompassing most of the HST-detected YSGs that are absent from our sample.

5.2 Impact of photometric uncertainties and source confusion

Source confusion is commonplace in crowded fields and can significantly bias the photometry of HST sources, leading to incorrect star classifications. This happens when two stars, such as a MS star and a red giant, are in close angular proximity. Their combined light can be mistaken for a single star, such as a YSG, due to the blended colour and luminosity. This effect is another source of contamination affecting our YSG catalogue.

One method for quantifying source confusion is using the CI, defined as the difference in magnitude between two apertures, typically a smaller aperture of 0.05″ and a larger aperture of 0.15″ (for ACS and WFC3/UVIS), normalised such that its distribution peaks at 1. Objects with CI values around 1.0 are likely to be stars, while those with significantly higher CI values are likely to be extended sources or the confusion of several stars. Figure 20 presents the distribution of YSG candidates in the CI – mF814W plane. At bright magnitudes, the CIis close to 1, showing that sources are point-like and source confusion is not significant. The faint sources are more numerous and thus denser on the sky, especially in distant galaxies. Therefore, they are more affected by source confusion and can have CI values close to 2. Such sources are currently excluded from our sample, which is restricted to CI values < 1.5, thereby reducing its completeness. A future release of the catalogue, incorporating DOLPHOT photometry (Dolphin 2016), is anticipated to address this issue more effectively.

Similarly, photometric uncertainties can bias the selection of YSGs in HST observations. These uncertainties arise from various sources, such as photon noise, background subtraction errors, and instrumental effects. When dealing with YSG candidates, identified based on their position in the CMD, even small errors in photometry can shift stars into or out of the YSG region. For example, in the NGC 4455 galaxy, an edge-on spiral galaxy located 7.3 Mpc away, observations were conducted using the F814W filter for 1030 seconds. These parameters are close to the median values of our sample of galaxies with YSG candidates. When the CMD is regenerated using magnitude values and errors modelled as a normal distribution, 9% of YSG candidates shift out of the selection region. Conversely, a similar number of previously unselected stars are shifted into the YSG selection region. Likewise, the variability of YSG can cause them to move out of the selection region; however, this variability primarily occurs over short timescales and within specific regions of the Hertzsprung-Russell diagram, such as the instability strip (e.g. Evans 1993).

In this context, our conservative selection region helps maintain a cleaner sample, although it may exclude some legitimate YSGs. Other sources of errors, such as uncertainties in stellar models, errors in extinction, distance, or metallicity of the host, may bias the selection region to a more significant extent.

thumbnail Fig. 19

Fraction of M31 YSG candidates from the literature recovered as HST sources and HST-selected YSG candidates, as a function of magnitude.

5.3 Search for LRN progenitors and precursors with LSST

In a recent paper, Strotjohann et al. (2024) explored the potential for identifying progenitor stars of core-collapse SNe using data from ground-based wide-field surveys such as ZTF and LSST. Due to challenges like crowding and atmospheric blurring, identifying these progenitor stars in pre-explosion images is difficult. They thus suggest combining numerous pre- and post-SN images to detect the disappearance of progenitor stars. As a proof of concept, the authors implemented this approach using ZTF data. Despite analysing hundreds of images and achieving limiting magnitudes of approximately 23 in the g and r bands, no progenitor stars or long-lived outbursts were detected for 29 SNe within a redshift of z ≤ 0.01. The sensitivity limits achieved were several magnitudes less than those in previously detected progenitors.

Conversely, the study projects that LSST, over its 10-year survey, could detect around 50 RSG progenitors and several yellow and blue supergiants. It estimates that progenitors of Type Ic SNe would be detectable if they are brighter than –4.0 magnitudes in the LSST i band, respectively. Given their similar spectral type (A5) compared to YSGs, we assume a similar performance for YSG detection: we therefore expect that 77% of our YSG sample will be re-detected by LSST, for sources having MF814W < –4 (or M* > 6 M⊙; Fig. 9). This will provide exquisite variability constraints for sources of r < 23, allowing us to identify new LRN precursors, given LSST repeated observations and the long timescales of these events. Such early precursor discoveries will be essential to schedule precursor spectroscopy, allowing us to probe the line profiles, outflow structure and density (and therefore mass loss), ionisation structure, and possible shocks in the system (e.g. Pejcha et al. 2016; Molnar et al. 2017; Blagorodnova et al. 2021).

Furthermore, LSST can expand the progenitor sample for the 1235 galaxies within 20 Mpc and at Dec < 30° that are not covered by deep multi-band HST exposures. In particular, it will cover more than 1200 massive galaxies (MB < –14 to select starforming hosts), where LRNe are most likely15, and and whose YSGs comprise 96% of our sample. Assuming the same distance distribution of YSG candidates as the one in this study, this represents a 130% increase in our M* > 6 M⊙ sample. It is important to note that some progenitors detected only a few years before the LRN are likely already undergoing mass transfer and can be considered precursors (this is the case of e.g. AT2021blu, imaged both 15 years and 2 years prior to the transient, showing a 1-mag brightening between these two epochs Pastorello et al. 2023). Ongoing large-scale, deep surveys such as Euclid (Laureijs et al. 2011) are also of significant interest for mapping stellar populations and identifying potential transient progenitors in nearby galaxies (Bonanos et al. 2024).

The rate of LRNe in the luminosity range -16 ≤ Mr ≤ –11 mag was recently constrained to 7.8−3.7+6.5×10−5 Mpc−3 yr−1$7.8_{ - 3.7}^{ + 6.5} \times {10^{ - 5}}\,{\rm{Mp}}{{\rm{c}}^{ - 3}}{\rm{y}}{{\rm{r}}^{ - 1}}$ (Karambelkar et al. 2023), comparable to the core-collapse SN rate (Perley et al. 2020). However, LRNe are about three magnitudes fainter, and Karambelkar et al. (2023) suggest a luminosity function in the form dN/dL ∝ L−2.5. At magnitude r < 18.5, we thus expect LSST to detect about 50 LRNe (i.e. 1.7% of the 3300 expected bright core-collapse SNe; Strotjohann et al. 2024) over its 10-year survey.

Strotjohann et al. (2024) also postulate how LSST will detect more than a thousand pre-SN outbursts, depending on their brightness and duration. In the case of LRNe precursors, a system brightening from –1 to –4 mag in absolute r-band magnitude (similar to the precursor of M31-LRN2015) would be detected in a single LSST exposure up to a distance of 5 Mpc (not taking into account the effects of the luminous background from the galaxy). This distance limit becomes 20 Mpc when using the final survey LSST r-band sensitivity. Considering the host galaxy’s brightness, late-time precursors or those from massive stars, with Mr < –6 (Blagorodnova et al. 2020, Figs. 14 and 16) would have a 6% probability of detection by LSST (Strotjohann et al. 2024). If half of LRNe have such a precursor, given their volumetric rate, one can expect about 100 new precursor detections in the era of LSST. The rate, luminosity function, and timing of LRN precursors will be measurable using this large dataset. This might contribute to revealing their intrinsic mechanisms.

Overall, the cadence of LSST, its multi-band coverage, along with the depth of the survey, allows for the detection of progenitors, years-long faint precursors and variability patterns that precede LRN events. Systematic cataloguing and data mining techniques will be crucial in identifying these specific observational signatures within LSST vast datasets. An approach based on the systematic use of archival data and the prediction of future variability using various light-curve analysis techniques is under study (Tranin et al., in prep.).

thumbnail Fig. 20

HST photometric CI of YSG candidates as a function of their F814W magnitude.

6 Conclusion

In this study we used HST imaging of nearby galaxies to find possible LRN progenitors and precursors, making it possible to predict their outburst and rapidly identify new transients matching the position of a candidate YSG. We retrieved the catalogues of HST sources and their photometry for 369 galaxies with distances of less than 20 Mpc using different public databases. After building CMDs for each galaxy, we selected the Hertzsprung gap stars using MIST stellar evolution tracks, coupled with a statistical representation of the CMD. Foreground contaminants were mostly removed using Gaia proper motions, and the remaining foreground contaminants were quantified using the TRILEGAL simulations of the Milky Way stellar content. The previous spectroscopic identification of foreground stars within the Hertzsprung gap of M31 and M33 showed excellent agreement with the results of our method. Additionally, we constrained the number of contaminants resulting from internal extinction to less than 20% of the sample and quantified this for each source. The use of MIST stellar evolution tracks and a meticulous filtering process to exclude contaminants proved crucial in accurately identifying candidates.

Our study identified 146 502 YSG candidates in 353 galaxies, a significant increase over previous research. The resulting sample of candidates was cross-matched with the TNS and the ongoing surveys ZTF, BlackGEM, and MeerLICHT. Candidates exhibiting outstanding variability were identified and analysed. In particular, we identified 12 precursor candidates based on their consistent brightening over the past few years. Their spectroscopic follow-up and identification will be the subject of upcoming work.

The YSG catalogue resulting from this study and the pipeline have been publicly released. The Python scripts used to retrieve and analyse HST data are available at the following address: https://github.com/htranin/LRNsearch. The insights gained from this catalogue can inform future models of stellar evolution and enhance our ability to predict and study rare transient events. This work advances our capabilities of understanding YSGs and their role as progenitors and precursors of LRNe, filling a gap in previous studies. LSST will be a game changer in the quest for LRNe and their progenitors and precursors: we estimate that the 10-year survey will more than double the number of detected extragalactic YSGs within 20 Mpc and provide excellent variability constraints for sources of magnitude r < 23. Notably, based on the rate of previous extragalactic LRNe, we expect about 100 LRN precursors to be discovered over the course of LSST, and about 50 bright r < 18.5 LRNe. Future research should focus on continuous monitoring of brightening YSG candidates to capture and analyse transient episodes as they occur. We emphasise the importance of closely monitoring future transients that have YSG progenitors to ensure the identification and study of LRN events and other rare transients.

Data availability

The full versions of Tables 3 and 4 are available in electronic form at the CDS via anonymous ftp to cdsarc.cds.unistra.fr (130.79.128.5) or via https://cdsarc.cds.unistra.fr/viz-bin/cat/J/A+A/695/A226

Acknowledgements

H. T. and N. B. acknowledge to be funded by the European Union (ERC, CET-3PO, 101042610). Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. PJG is supported by NRF SARChI grant 111692. We thank Zeljko Ivezic and the anonymous referee for their valuable comments, which significantly enhanced the quality of this paper. We thank Rick White and Bernie Shiao for their assistance in retrieving the HSCv3 data. We acknowledge the extensive use of the MAST database to conduct this study. Based on observations with the MeerLICHT telescope. MeerLICHT is built and run by a consortium consisting of Radboud University, the University of Cape Town, the South African Astronomical Observatory, the University of Oxford, the University of Manchester and the University of Amsterdam. MeerLICHT is hosted by SAAO. Based on observations with the BlackGEM telescopes. BlackGEM is built and run by a consortium consisting of Radboud University, the Netherlands Research School for Astronomy (NOVA), and KU Leuven with additional support from Armagh Observatory and Planetarium, Durham University, Hamburg Observatory, Hebrew University, Las Cumbres Observatory, Tel Aviv University, Texas Tech University, Technical University of Denmark, University of California Davis, the University of Barcelona, the University of Manchester, University of Potsdam, the University of Valparaiso, the University of Warwick, and Weizmann Institute of science. BlackGEM is hosted and supported by ESO. Supported by the National Science Foundation under Grants No. AST-1440341 and AST-2034437 and a collaboration including current partners Caltech, IPAC, the Oskar Klein Center at Stockholm University, the University of Maryland, University of California, Berkeley, the University of Wisconsin at Milwaukee, University of Warwick, Ruhr University, Cornell University, Northwestern University and Drexel University. Operations are conducted by COO, IPAC, and UW. Based on observations made with the NASA/ESA Hubble Space Telescope, and obtained from the Hubble Legacy Archive, which is a collaboration between the Space Telescope Science Institute (STScI/NASA), the Space Telescope European Coordinating Facility (ST-ECF/ESAC/ESA) and the Canadian Astronomy Data Centre (CADC/NRC/CSA).

Appendix A Example of CMDs

Figure A.1 presents a representative subset of CMDs analysed in this study. The selection area of YSG candidates is shown by solid blue lines.

thumbnail Fig. A.1

Representative sample of the CMDs analysed in this study. The blue MIST track corresponds to the reference stellar mass at which YSG observation completeness is ensured.

References

  1. Addison, H., Blagorodnova, N., Groot, P. J., et al. 2022, MNRAS, 517, 1884 [NASA ADS] [CrossRef] [Google Scholar]
  2. Aryan, A., Pandey, S. B., Zheng, W., et al. 2021, MNRAS, 505, 2530 [NASA ADS] [CrossRef] [Google Scholar]
  3. Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P. 2009, ARA&A, 47, 481 [NASA ADS] [CrossRef] [Google Scholar]
  4. Astropy Collaboration (Robitaille, T. P., et al.) 2013, A&A, 558, A33 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  5. Bellm, E. C., Kulkarni, S. R., Graham, M. J., et al. 2019, PASP, 131, 018002 [Google Scholar]
  6. Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  7. Blagorodnova, N., Kotak, R., Polshaw, J., et al. 2017, ApJ, 834, 107 [NASA ADS] [CrossRef] [Google Scholar]
  8. Blagorodnova, N., Karambelkar, V., Adams, S. M., et al. 2020, MNRAS, 496, 5503 [CrossRef] [Google Scholar]
  9. Blagorodnova, N., Klencki, J., Pejcha, O., et al. 2021, A&A, 653, A134 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  10. Bloemen, S., Groot, P., Woudt, P., et al. 2016, SPIE Conf. Ser., 9906, 990664 [NASA ADS] [Google Scholar]
  11. Bonanos, A. Z., Tramper, F., de Wit, S., et al. 2024, A&A, 686, A77 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  12. Cai, Y. Z., Pastorello, A., Fraser, M., et al. 2022, A&A, 667, A4 [Google Scholar]
  13. Choi, J., Dotter, A., Conroy, C., et al. 2016, ApJ, 823, 102 [Google Scholar]
  14. Dal Tio, P., Pastorelli, G., Mazzi, A., et al. 2022, ApJS, 262, 22 [NASA ADS] [CrossRef] [Google Scholar]
  15. Delchambre, L., Bailer-Jones, C. A. L., Bellas-Velidis, I., et al. 2023, A&A, 674, A31 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  16. Dolphin, A. 2016, DOLPHOT: Stellar photometry, Astrophysics Source Code Library [record ascl:1608.013] [Google Scholar]
  17. Dotter, A. 2016, ApJS, 222, 8 [Google Scholar]
  18. Drout, M. R., Massey, P., Meynet, G., Tokarz, S., & Caldwell, N. 2009, ApJ, 703, 441 [NASA ADS] [CrossRef] [Google Scholar]
  19. Drout, M. R., Massey, P., & Meynet, G. 2012, ApJ, 750, 97 [NASA ADS] [CrossRef] [Google Scholar]
  20. Evans, N. R. 1993, AJ, 105, 1956 [Google Scholar]
  21. Flesch, E. W. 2021, arXiv e-prints [arXiv:2105.12985] [Google Scholar]
  22. Förster, F., Cabrera-Vives, G., Castillo-Navarrete, E., et al. 2021, AJ, 161, 242 [CrossRef] [Google Scholar]
  23. Gaia Collaboration (Brown, A. G. A., et al.) 2018, A&A, 616, A1 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  24. Gaia Collaboration (Brown, A. G. A., et al.) 2021, A&A, 649, A1 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  25. Gaia Collaboration (Bailer-Jones, C. A. L., et al.) 2023a, A&A, 674, A41 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  26. Gaia Collaboration (Vallenari, A., et al.) 2023b, A&A, 674, A1 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  27. Georgy, C. 2012, A&A, 538, L8 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  28. Ginsburg, A., Sipocz, B. M., Brasseur, C. E., et al. 2019, AJ, 157, 98 [NASA ADS] [CrossRef] [Google Scholar]
  29. Girardi, L., Groenewegen, M. A. T., Hatziminaoglou, E., & da Costa, L. 2005, A&A, 436, 895 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  30. Gordon, M. S., Humphreys, R. M., & Jones, T. J. 2016, ApJ, 825, 50 [NASA ADS] [CrossRef] [Google Scholar]
  31. Groot, P., Bloemen, S., & Jonker, P. 2019, in The La Silla Observatory – From the Inauguration to the Future, 33 [Google Scholar]
  32. Groot, P. J., Bloemen, S., Vreeswijk, P. M., et al. 2024, PASP, 136, 115003 [NASA ADS] [CrossRef] [Google Scholar]
  33. Hornoch, K., Shafter, A. W., Kucakova, H., Yosuf, I., & Luo, A. 2023, The Astronomer’s Telegram, 16319, 1 [Google Scholar]
  34. Humphreys, R. M., Jones, T. J., & Martin, J. C. 2023, AJ, 166, 50 [NASA ADS] [CrossRef] [Google Scholar]
  35. Jacobson-Galán, W. V., Davis, K. W., Kilpatrick, C. D., et al. 2024, ApJ, 972, 177 [CrossRef] [Google Scholar]
  36. Jencson, J. E., Kasliwal, M. M., Adams, S. M., et al. 2019, ApJ, 886, 40 [Google Scholar]
  37. Justham, S., Podsiadlowski, P., & Vink, J. S. 2014, ApJ, 796, 121 [NASA ADS] [CrossRef] [Google Scholar]
  38. Kankare, E., Kotak, R., Pastorello, A., et al. 2015, A&A, 581, L4 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  39. Karambelkar, V. R., Kasliwal, M. M., Blagorodnova, N., et al. 2023, ApJ, 948, 137 [NASA ADS] [CrossRef] [Google Scholar]
  40. Kochanek, C. S., Adams, S. M., & Belczynski, K. 2014, MNRAS, 443, 1319 [NASA ADS] [CrossRef] [Google Scholar]
  41. Kovlakas, K., Zezas, A., Andrews, J. J., et al. 2021, MNRAS, 506, 1896 [NASA ADS] [CrossRef] [Google Scholar]
  42. Kraft, R. P. 1966, ApJ, 144, 1008 [Google Scholar]
  43. Lasker, B. M., Doggett, J., McLean, B., et al. 1996, in Astronomical Society of the Pacific Conference Series, 101, Astronomical Data Analysis Software and Systems V, eds. G. H. Jacoby & J. Barnes, 88 [NASA ADS] [Google Scholar]
  44. Laureijs, R., Amiaux, J., Arduini, S., et al. 2011, arXiv e-prints [arXiv:1110.3193] [Google Scholar]
  45. MacLeod, M., De, K., & Loeb, A. 2022, ApJ, 937, 96 [NASA ADS] [CrossRef] [Google Scholar]
  46. MacLeod, M., Macias, P., Ramirez-Ruiz, E., et al. 2017, ApJ, 835, 282 [NASA ADS] [CrossRef] [Google Scholar]
  47. Marchant, P., & Bodensteiner, J. 2024, ARA&A, 62, 21 [NASA ADS] [CrossRef] [Google Scholar]
  48. Martini, P., Wagner, R. M., Tomaney, A., et al. 1999, AJ, 118, 1034 [NASA ADS] [CrossRef] [Google Scholar]
  49. Masci, F. J., Laher, R. R., Rusholme, B., et al. 2023, arXiv e-prints [arXiv:2305.16279] [Google Scholar]
  50. Massey, P., Waterhouse, E., & DeGioia-Eastwood, K. 2000, AJ, 119, 2214 [NASA ADS] [CrossRef] [Google Scholar]
  51. Massey, P., Neugent, K. F., & Smart, B. M. 2016, AJ, 152, 62 [NASA ADS] [CrossRef] [Google Scholar]
  52. Moe, M., & Di Stefano, R. 2017, ApJS, 230, 15 [Google Scholar]
  53. Molnar, L. A., Van Noord, D. M., Kinemuchi, K., et al. 2017, ApJ, 840, 1 [Google Scholar]
  54. Mora, M. D., Larsen, S. S., & Kissler-Patig, M. 2007, A&A, 464, 495 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  55. Neugent, K. F., Massey, P., Skiff, B., et al. 2010, ApJ, 719, 1784 [NASA ADS] [CrossRef] [Google Scholar]
  56. Neugent, K. F., Massey, P., Skiff, B., & Meynet, G. 2012, ApJ, 749, 177 [NASA ADS] [CrossRef] [Google Scholar]
  57. Neustadt, J. M. M., Kochanek, C. S., Stanek, K. Z., et al. 2021, MNRAS, 508, 516 [NASA ADS] [CrossRef] [Google Scholar]
  58. O’Grady, A. J. G., Drout, M. R., Neugent, K. F., et al. 2024, ApJ, 975, 29 [Google Scholar]
  59. Paczynski, B. 1976, in Structure and Evolution of Close Binary Systems, 73, eds. P. Eggleton, S. Mitton, & J. Whelan, 75 [NASA ADS] [CrossRef] [Google Scholar]
  60. Pastorello, A., Mason, E., Taubenberger, S., et al. 2019, A&A, 630, A75 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  61. Pastorello, A., Fraser, M., Valerin, G., et al. 2021a, A&A, 646, A119 [EDP Sciences] [Google Scholar]
  62. Pastorello, A., Valerin, G., Fraser, M., et al. 2021b, A&A, 647, A93 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  63. Pastorello, A., Valerin, G., Fraser, M., et al. 2023, A&A, 671, A158 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  64. Paturel, G., Petit, C., Prugniel, P., et al. 2003, A&A, 412, 45 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  65. Paxton, B., Bildsten, L., Dotter, A., et al. 2011, ApJS, 192, 3 [Google Scholar]
  66. Pejcha, O., Metzger, B. D., & Tomida, K. 2016, MNRAS, 461, 2527 [NASA ADS] [CrossRef] [Google Scholar]
  67. Perley, D. A., Fremling, C., Sollerman, J., et al. 2020, ApJ, 904, 35 [NASA ADS] [CrossRef] [Google Scholar]
  68. Pessi, T., Cartier, R., Hueichapan, E., et al. 2024, A&A, 688, L28 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  69. Pilyugin, L. S., Vílchez, J. M., & Contini, T. 2004, A&A, 425, 849 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  70. Reig, P., Tzouvanou, A., & Pantoulas, V. 2022, ATel, 15612, 1 [Google Scholar]
  71. Smartt, S. J. 2015, PASA, 32, e016 [NASA ADS] [CrossRef] [Google Scholar]
  72. Smith, N., Andrews, J. E., Van Dyk, S. D., et al. 2016, MNRAS, 458, 950 [Google Scholar]
  73. Smith, K. W., Williams, R. D., Young, D. R., et al. 2019, RNAAS, 3, 26 [NASA ADS] [Google Scholar]
  74. Stetson, P. B. 1987, PASP, 99, 191 [Google Scholar]
  75. Strotjohann, N. L., Ofek, E. O., Gal-Yam, A., et al. 2024, ApJ, 960, 72 [NASA ADS] [CrossRef] [Google Scholar]
  76. Tonry, J. L., Denneau, L., Heinze, A. N., et al. 2018, PASP, 130, 064505 [Google Scholar]
  77. Tran, H. D., Burger, M., & Hack, W. 2020, in Astronomical Society of the Pacific Conference Series, 527, Astronomical Data Analysis Software and Systems XXIX, eds. R. Pizzo, E. R. Deul, J. D. Mol, J. de Plaa, & H. Verkouter, 587 [Google Scholar]
  78. Tylenda, R., & Soker, N. 2006, A&A, 451, 223 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  79. Tylenda, R., Hajduk, M., Kaminski, T., et al. 2011, A&A, 528, A114 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  80. Whitmore, B. C., Allam, S. S., Budavári, T., et al. 2016, AJ, 151, 134 [NASA ADS] [CrossRef] [Google Scholar]
  81. Williams, S. C., Darnley, M. J., Bode, M. F., & Steele, I. A. 2015, ApJ, 805, L18 [Google Scholar]
2

This cross-match was performed using the application programming interface (API) of MAST, available as part of the astroquery v0.4.6 package (Ginsburg et al. 2019).

3

Recent studies targeting the Magellanic Clouds using ground-based surveys and Swift-UVOT provide YSG samples of good completeness (Neugent et al. 2010, 2012; Humphreys et al. 2023; O’Grady et al. 2024).

5

A data-dependent matching radius r1 (0.3<r1arcsec<2)$(0.3 < \frac{{{r_1}}}{{{\mathop{\rm arcsec}\nolimits} }} < 2)$ is used to associate sources from different filters, and another one r2 (0.1<r2arcsec<0.3)$(0.1 < \frac{{{r_2}}}{{{\mathop{\rm arcsec}\nolimits} }} < 0.3)$ is used to eliminate duplicate sources. r2 is the local minimum of the distribution of separations in the interval [0.1,0.3] arcsec. This choice allows us to eliminate duplicate sources while keeping close neighbours, given the different levels of astrometric accuracies when using different source-extracting algorithms (SExtractor and DAOPHOT).

7

Given that only 12 galaxies in our sample are elliptical, we applied this relation accordingly.

15

Every extragalactic LRN discovered in the last decade belonged to a massive star-forming host, the least massive host being the Spider galaxy with MB = –16∙7 (Pastorello et al. 2023).

All Tables

Table 1

HST filters and galaxy selection.

In the text
Table 2

MIST parameter grid.

In the text
Table 3

YSG candidates (extract).

In the text
Table 4

Galaxies with more than 400 YSG candidates, sorted by distance (extract).

In the text
Table 5

Retrieval status of known LRN progenitors in the YSG sample.

In the text
Table 6

YSG candidates that match TNS objects.

In the text
Table 7

Precursor candidates identified in this study.

In the text

All Figures

thumbnail Fig. 1

Distribution of some properties of the sample of 575 HST-observed galaxies. Top: semi-major axis, a, and distance from HECATE. Bottom: metallicity and Galactic extinction (see the main text for details).
In the text
thumbnail Fig. 2

Density of HST sources under study for the galaxy M31, as retrieved through the HSCv3 and MAST databases. The background image is from the Digitized Sky Survey 2 (Lasker et al. 1996).
In the text
thumbnail Fig. 3

Example of CMD analysis: extinction-corrected CMD ofIC 1613 sources with MIST tracks for a metallicity of Z=–1.2 and an extinction of AV=0.2. Hertzsprung gap candidates are selected above the blue lines. The dotted contour shows the Gaussian mixture model representing the data.
In the text
thumbnail Fig. 4

Illustration of the estimation of the foreground contamination rate using the TRILEGAL Milky Way simulations. The CMD shows TRILEGAL sources overlapping with the footprint of M83 HSC sources. Highlighted in red are the YSG candidates identified within this sample. TRILEGAL sources falling within the black polygon, circled in red, are identified as foreground contaminants. They represent 3.1% of the M83 YSG candidates.
In the text
thumbnail Fig. 5

Estimation of IC 1613 internal extinction. Top: density plot of the RGB of IC 1613 in the magnitude vs. relative AV plane. The x-axis zero point corresponds to the mean of the distribution. The top curve shows the kernel density estimation of AV values, with a standard deviation of ± 0.21 (dotted lines). Bottom: resulting map showing the probability of a YSG being an extinguished MS star in the selection region of YSG sources in the IC 1613 galaxy.
In the text
thumbnail Fig. 6

Distribution of the standard deviations of extinction for our galaxy sample.
In the text
thumbnail Fig. 7

Cumulative distribution of distances for YSG candidates before and after cleaning, and the estimated number of extinguished MS stars.
In the text
thumbnail Fig. 8

Distributions of observed F814W magnitude and stellar masses of YSG candidates.
In the text
thumbnail Fig. 9

MIST-inferred ZAMS stellar masses of YSG candidates as a function of their absolute F814W magnitude.
In the text
thumbnail Fig. 10

Foreground contamination rates per galaxy. Left: foreground contamination rate as a function of Galactic latitude derived from the TRILEGAL Milky Way simulations. The colour encodes the Galactic longitude. To maintain readability, only galaxies with a minimum of 100 YSG candidates are included. Right: empirical cumulative distribution function (ECDF) of the foreground contamination rate. For 75% of the galaxies, fewer than 10% of their YSG candidates are foreground contaminants.
In the text
thumbnail Fig. 11

Probability of YSG candidates being an extinguished MS star as a function of their MIST temperature. Three example galaxies are shown, illustrating the impact of the HST blue/green filter used to analyse the CMD.
In the text
thumbnail Fig. 12

Fraction of identified contaminants using SIMBAD and Gaia catalogues as a function of observed and absolute F814W magnitudes.
In the text
thumbnail Fig. 13

HST cutouts of the 12 candidates resulting from our TNS cross-match. Images are 10″ side-to-side, with a circle of 1″ radius pinpointing the location of the YSG source. Adapted from HLA colour composites.
In the text
thumbnail Fig. 14

ZTF forced photometry light curves of the nine candidates resulting from our TNS cross-match. The blue areas indicate the period that was used to set the baseline flux at the location of these transients.
In the text
thumbnail Fig. 15

HST cutouts of precursor candidates. Images are 10″ side-to- side with a 1″ radius circle centred on the HST position.
In the text
thumbnail Fig. 16

ATLAS forced photometry light curves of the best 12 precursor candidates in ZTF, MeerLICHT, and BlackGEM. ATLAS photometry is represented by orange and cyan circles (o and c bands), ZTF photometry by red and green diamonds (r and g bands), and MeerLICHT/BlackGEM photometry by black and brown triangles (q and i bands). ATLAS light curves were re-binned to 60-day bins, other surveys to 15-day bins.
In the text
thumbnail Fig. 17

Gaia properties of YSG candidates in M31 and M33. Top: distribution of the Gaia DR3 proper motion signal-to-noise for M31 and M33 stars classified in the literature. Bottom: distribution of the variability indicator dGmag for M31 stars. The stars classified as foreground stars but with a low proper motion are shown in blue.
In the text
thumbnail Fig. 18

CMD of M31 stars classified in the literature, using the photometry of Massey et al. (2016). Smaller markers without contours are stars present in the HST footprint but not in HST catalogues, smaller markers with grey contours are stars in HST catalogues but not in our YSG selection, and larger markers with black contours are the selected YSG candidates. The dotted lines highlight a region encompassing most of the HST-detected YSGs that are absent from our sample.
In the text
thumbnail Fig. 19

Fraction of M31 YSG candidates from the literature recovered as HST sources and HST-selected YSG candidates, as a function of magnitude.
In the text
thumbnail Fig. 20

HST photometric CI of YSG candidates as a function of their F814W magnitude.
In the text
thumbnail Fig. A.1

Representative sample of the CMDs analysed in this study. The blue MIST track corresponds to the reference stellar mass at which YSG observation completeness is ensured.
In the text

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

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

Initial download of the metrics may take a while.