Issue |
A&A
Volume 690, October 2024
|
|
---|---|---|
Article Number | A178 | |
Number of page(s) | 13 | |
Section | Galactic structure, stellar clusters and populations | |
DOI | https://doi.org/10.1051/0004-6361/202450256 | |
Published online | 08 October 2024 |
The spectroscopic binary fraction of the young stellar cluster M17★
1
Max-Planck-Institut für Astronomie,
Königstuhl 17,
D-69117
Heidelberg,
Germany
2
Anton Pannekoek Institute for Astronomy, University of Amsterdam,
Science Park 904,
1098
XH
Amsterdam,
The Netherlands
3
Instituut voor Sterrenkunde, KU Leuven,
Celestijnenlaan 200D bus 2401,
3001
Leuven,
Belgium
★★ Corresponding author; ramirez@mpia.de
Received:
5
April
2024
Accepted:
14
August
2024
Context. Significant progress has been made toward understanding the formation of massive (M > 8 M⊙) binaries in close orbits (with periods of less than a month). Some of the observational studies leading to this progress are the detection of a very low velocity dispersion among the massive stars in the young region M17 and the measurement of a positive trend of velocity dispersion with age in Galactic clusters. The velocity dispersion observed in M17 could be explained either by the lack of binaries among the stars in this region, which implies the highly unlikely scenario of a different formation mechanism for M17 than for other Galactic regions, or by larger binary separations than typically observed, but with a binary fraction similar to other young Galactic clusters. The latter implies that, over time, the binary components migrate toward each other. This is in agreement with the finding that the radial velocity dispersion of young Galactic clusters correlates positively with their age.
Aims. We aim to determine the origin of the strikingly low velocity dispersion by determining the observed and intrinsic binary fraction of massive stars in M17 through multi-epoch spectroscopy.
Methods. We performed a multi-epoch spectroscopic survey consisting of three epochs separated by days and months, respectively. We complemented this survey with existing data covering timescales of years. We determined the radial velocity of each star at each epoch by fitting the stellar absorption profiles. The velocity shifts between epochs were used to determine whether a close companion is present.
Results. We determined an observed binary fraction of 27% and an intrinsic binary fraction of 87%, consistent with that of other Galactic clusters. We conclude that the low velocity dispersion is due to a large separation among the young massive binaries in M17. Our result is in agreement with a migration scenario in which massive stars are born in binaries or higher-order systems at large separation and harden within the first million years of evolution. Such an inward migration may either be driven by interaction with a remnant accretion disk or with other young stellar objects present in the system, or by dynamical interactions within the cluster. Our results imply that possibly both dynamical interactions and binary evolution are key processes in the formation of gravitational wave sources.
Key words: binaries: close / binaries: spectroscopic / stars: early-type / stars: formation / stars: pre-main sequence / open clusters and associations: general
© The Authors 2024
Open 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.
Open Access funding provided by Max Planck Society.
1 Introduction
Multiplicity is a fundamental aspect of the process of massive star formation and evolution up until the formation of gravitational wave progenitors (e.g., Blaauw 1991; Zinnecker & Yorke 2007; Sana et al. 2012; Langer 2012; de Mink et al. 2013; Tan et al. 2014; Offner et al. 2023). The majority of massive stars (M > 8 M⊙) are part of binary or higher-order multiple systems (Sana et al. 2014; Bordier et al. 2022), and about 40–50% of these systems have an orbital period of less than a month. This is true in the Milky Way (e.g. Sana et al. 2012, 2013; Dunstall et al. 2015; Moe & Di Stefano 2015b; Barbá et al. 2017; Moe & Di Stefano 2017; Banyard et al. 2022) as well as in the lower-metallicity environment of the Large Magellanic Cloud (LMC) (Almeida et al. 2017; Villaseñor et al. 2021). Surveys of ~2–5 Myr old massive clusters reveal intrinsic spectroscopic OB binary fractions of fbin > 0.5 and show that multiple companions are very common (e.g. Penny et al. 1993; Hillwig et al. 2006; Sana et al. 2008, 2009, 2011; Ritchie et al. 2009; Bosch et al. 2009; Mason et al. 2009; Kobulnicky et al. 2014). The orbital periods of these spectroscopically selected samples vary from about 1 day up to ~10 years, with a preference toward close binaries with periods of less than one month (less than ~1 au in separation). These are efficiently detected by spectroscopy through periodic Doppler shifts of the photospheric lines in their spectra (e.g., Sana et al. 2013; Dunstall et al. 2015; Kobulnicky et al. 2014).
The current understanding is that about 70% of all O-type stars are so close that they are expected to interact with their companion, about half of them before leaving the main sequence (Sana et al. 2012). About 25% may even merge with a companion, offering a route toward the formation of the most massive stars (e.g. Bonnell & Bate 2005) and to the formation of magnetic fields (Schneider et al. 2019; Frost et al. 2024). Furthermore, massive binaries produce a variety of exotic products later in their evolution such as X-ray binaries, rare types of supernovae (Ibc, IIn), gamma-ray bursts, and, eventually, gravitational wave sources (Smith et al. 2011; Pols 1997; de Mink & Mandel 2016; Abbott et al. 2017).
Ramírez-Tannus et al. (2017) and Sana et al. (2017) measured a distinctively low RV dispersion of 5.6±0.2 km s−1 in a sample of 11 pre- and near-main-sequence stars in the young (0.65±0.25 Myr) and nearest giant H II region M17 (Stoop et al. 2024). The observed very low velocity dispersion could be explained by two scenarios: (1) the lack of binaries among the stars in M17, or (2) a binary fraction similar to those observed in other, somewhat older Galactic clusters, but larger binary separations. In order to determine which scenario is causing the low velocity dispersion in M17, we carried out a multi-epoch spectroscopic campaign of a larger number of stars to determine the binary fraction of the massive stars in this region.
The targets in the sample used by Sana et al. (2017) (hereafter the 2017 sample) were selected based on the brightness of the K-band, and on the presence of emission lines in their K-band spectra presented by Hanson et al. (1997). The 2017 sample contains seven stars with infrared (IR) excess longward of 3 μm, five of which also have a hot gaseous inner disk identified by double-peaked emission lines and CO bandhead emission (Ramírez-Tannus et al. 2017). The 2017 sample selection favors objects with a disk, which might potentially be a selection bias against binaries, since disk removal might be sped up under the influence of a companion star (Jensen et al. 1996; Kraus et al. 2012; Cheetham et al. 2015). Therefore, in this study, we changed the strategy of source selection to contain the most massive stars in M17 from the study by Broos et al. (2013). Additionally, we included the stars from the previous study so as to reinvestigate the strikingly low velocity dispersion and assess possible biases.
In this paper, we first discuss the observed sample and radialvelocity (RV) measurements (Section 2). We continue with the RV dispersion from a single-epoch analysis, and the observed and intrinsic binary fractions that follow from a multi-epoch analysis (Section 3). The results are discussed in Section 4. Finally, our conclusions are presented in Section 5.
2 Observations and data reduction
A sample of 27 targets was chosen such that it includes the most massive stars in M17 (from Broos et al. 2013) as well as several pre-main-sequence stars (see Hanson et al. 1997; Ramírez-Tannus et al. 2017). They all reside in the central cluster of M17 (NGC 6618). The brightness of the objects ranges from ~13 to ~18 mag in the V band and from ~7 to ~13 mag in the K band; this large range is mainly due to the high and variable extinction toward each of the sources in M17.
In order to optimize the binary detection, we designed the observing campaign such that each star is targeted three times: the first two epochs separated by at least 48 hours and the third one four weeks after. Adopting a flat mass-ratio distribution (Sana et al. 2012), this approach provides binary detection rates of 70% for periods up to 6 months, and better than 90% for periods up to 10 days (Sana et al. 2013), and it allows us to obtain statistical constraints on their period distribution through the RV amplitude. However, due to bad weather conditions, the proposed observing program could not be completed in full; therefore, the number of observations varies from one to five epochs per star including the 2017 X-shooter sample from Ramírez-Tannus et al. (2017) (see Table 1).
Sample overview.
2.1 Data reduction
The 2019 VLT/X-shooter (Vernet et al. 2011) observations were taken between May and September under good weather conditions with seeing ranging between 0.5″ and 1.2″ and clear skies. The slit widths used were 1″, 0.9″, and 0.6″, corresponding to a spectral resolution of 5100, 8800, and 8100 for the UVB, VIS, and NIR arms, respectively. All spectra were taken in nodding mode. For some stars, additional epochs from archival observations were added to the analysis. The log of observations is available in electronic form at the CDS.
The spectra were reduced using the ESO-pipeline esorex 3.3.5 (Modigliani et al. 2010) with the exception of the 2012 observation of B331, for which we used the reduced spectra from Ramírez-Tannus et al. (2017). The flux calibration was done with spectrophotometric standards from the ESO database. We used Molecfit 1.5.9 to correct for the telluric lines (Smette et al. 2015; Kausch et al. 2015). For B243, B268, B275, and B337, for which the nebular contamination is significant, the nebular subtraction was performed by reducing each nodding position in staring mode and fitting the non-sky subtracted nebular lines. The fitted nebular spectrum was then removed from the science spectrum in the manner described in van Gelder et al. (2020); Derkink et al. (2024).
2.2 Stellar classification
For most of the targets, we used the spectral classification of Ramírez-Tannus et al. (2017) and Derkink et al. (2024). For the remaining stars, the spectral type was determined in this paper by comparing the stacked spectra with the LAMOST survey (Liu et al. 2019), Galactic O-Star Catalog (GOSC, Sota et al. 2011), and Gray’s Atlas (Gray 2000). All spectral types are reported in Table 1. Targets 326 and B86 are identified as double-lined spectroscopic binaries (SB2; see Section 3). More details on the spectral classification can be found in Appendix A.
2.3 Sample
As we are interested in the binary fraction of the massive stars in M17, we removed from the sample B228, B230, B269, B290, B293, and B336, which are cool stars (A-, F-, or G-type) and probably in the foreground of M17 based on their Gaia distance measurements. This leaves us with a sample of 21 OB stars with one to five epochs of observation. Of these 21 stars, eight stars show an IR excess in their spectral energy distribution (SED; Backs et al. 2024). Seven of these stars with disks are part of the 2017 sample. The masses of our targets range from 3 to 30 M⊙ (Backs et al. 2024). Six stars (B181, B213, B234, B272, B331, and CEN55) only have one observation, preventing us from calculating RV variations between spectra. This leaves us with a multi-epoch sample size of 15 stars. The single-epoch spectra can still be used to determine the RV dispersion of the sample (see Sec. 3.1).
2.4 Radial-velocity measurements
We measured the RV of the stars in each epoch (including the five stars with only one epoch) following the procedure described by Sana et al. (2013). We developed a python code (RVFITTER1) that allows one to fit a given set of spectral lines using Gaussian, Lorenzian, or Voigt profiles. The outcome of RVFITTER is two velocity measurements: i) the RV resulting from fitting all lines individually, allowing each line per epoch and star to have a different line depth and width (__without_constraints__), where the RV is given by the median of the measurements for all lines in a given epoch and the error corresponds to the standard deviation. The second output (__with_constraints__) is the RV, assuming that all lines at a given epoch share the same RV and that a given line fit has a constant amplitude and standard deviation, as was assumed in Sana et al. (2013). In the latter case, the error was estimated from the covariance matrix. For this paper, we have adopted the velocities given by the method __with_constraints__ and used Gaussian profiles for the fit, with the exception of B181, B243, B268, B337, and B293, for which we only fit hydrogen lines to determine RV, and therefore used the Lorenzian2 profile __with_constraints__. The results of adopting Gaussian and Lorenzian profiles agree within the 1σ errors. The lines used per star are given in Appendix B. We also measured the RVs of lines originating from the interstellar medium (ISM) to check and correct the RVs of the stars to account for differences in wavelength calibration between the epochs. The typical error on the RV measurements is about 0.2 km s−1 and the differences between the RVs of the ISM lines are on the order of 1–3 km s−1, except for B268 and B337, which have differences of ~8 km s−1. Source 326 is a similar mass SB2 with overlapping absorption features. Therefore, the RVs could not be measured for this star. B86 is also an SB2, but the mass and luminosity difference between both stars is large enough that a few single absorption lines of one of the stars could be used to obtain RV measurements.
3 Multiplicity analysis
In this section, we discuss the analysis of the multiplicity of the stars in our sample. In Section 3.1, we present the RV dispersion (σ1D) of the newly obtained sample as if we had single-epoch observations in order to compare our results to the work presented in Sana et al. (2017) and Ramírez-Tannus et al. (2021). In Section 3.2, we establish the spectroscopic binary fraction based on the RV variations (SB1) and/or the SB2 nature of the spectra, and Section 3.3 establishes the intrinsic binary fraction in M17.
3.1 Radial-velocity dispersion
Sana et al. (2017) and Ramírez-Tannus et al. (2021) performed Monte Carlo simulations in order to find the minimum orbital period (Pcutoff) and binary fraction (fbin) that best reproduced the observed σ1D measured in clusters with only single-epoch observations. In order to compare our newly obtained multi-epoch data to the single-epoch data represented in Ramírez-Tannus et al. (2021), we followed the procedure described by the authors. In short, we computed the most probable σ1D for each cluster by randomly drawing one epoch per star and computing the RV dispersion, σ1D, of the sample cluster as (1)
and RVi is the RV of a star at a certain epoch given by RVstar,i − RVISM,i (RVstar and RVISM are given in electronic form at the CDS), ⟨υRV⟩ is the weighted mean RV (given in Table 2), and is the weight for each measurement. We repeated this procedure 10000 times and then calculated the most probable σ1D as the 50th percentile. The corresponding uncertainty is taken as half of the difference between the 16th and 84th percentiles of the resulting σ1D distributions. The error on the distribution is slightly higher than those presented in Ramírez-Tannus et al. (2021). This difference is due to the fact that in this paper we calculated the error in σ1D based on the percentiles and not as the standard deviation of a Gaussian fit to the σ1D distribution. The latter was to avoid making the assumption that the σ1D distributions are Gaussian. Using the spectra of Sana et al. (2017), but with the analysis presented here, we find a σ1D of 7.4 ± 0.6 km s−1. This is slightly higher than reported in 2017 (5.6 ± 0.2 km s−1), but still very low in comparison to slightly older Galactic clusters. The difference is due to the new reduction, normalization, and clipping of the data.
With the extended sample presented in this paper, we obtain a σ1D for M17 of 8.9 ± 0.5 km s−1, which is higher than the 5.6 ± 0.2 km s−1 reported for the 2017 sample. The difference is due to the inclusion of three stars with significant RV variations (see Section 4). However, these values remain much lower than the than the observed σ1D of ~30 km s−1 in other young (~2–5 Myr) clusters with significant short-period binary populations.
With this newly calculated σ1D and updated age estimates for NGC 6611 from Stoop et al. (2023), NGC 6231 from van der Meij et al. (2021), and M17 from Stoop et al. (2024), we can update the σ1D versus age relation first introduced in Ramírez-Tannus et al. (2021). To calculate σ1D for the clusters with more than one epoch of observations, we followed the procedure described in Ramírez-Tannus et al. (2021). We drew RV values for each star at a random epoch from a Gaussian centered on the reported RV and with a width equal to the measurement uncertainty. We repeated this procedure 105 times and then calculated the most probable σ1D and its standard deviation. The updated relation is shown in Figure 1 and the data is presented in Table 2: we include the clusters studied by Sana et al. (2012), those by Ramírez-Tannus et al. (2020), R136 in the LMC from Hénault-Brunet et al. (2012), Westerlund 2 (Wd2) from Zeidler et al. (2018), and Trumpler 14 from Kiminki & Smith (2018). With the updated data, after performing an orthogonal distance regression, we find a positive correlation between σ1D and the age of the clusters. The Pearson coefficient is 0.8, which indicates a strong correlation, and is higher than that found by Ramírez-Tannus et al. (2021). We note that, with the exception of NGC 6611, the stars in 0 to 2 Myr old clusters span a lower mass range than the more massive clusters (see Table 2). This could partially explain the low σ1D, as the binary properties vary with primary mass (e.g. Moe & Di Stefano 2017). Nevertheless, restricting the samples of M17 and Wd2 to stars with M > 15 M⊙ does not change our conclusions.
3.2 Observed spectroscopic binary fraction
In order to determine the spectroscopic binary fraction, we performed a statistical test described in Sana et al. (2013). A star is detected as a binary if the following two criteria are fulfilled simultaneously: a) two measurements deviate significantly from each other, and b) those same measurements differ by a minimum threshold, C. The latter condition is to avoid RV variations caused by pulsations or photospheric variability. To make this concrete, a star is a binary if it meets the following criteria: (2)
where RV and σRV correspond to the RV measured and its corresponding one sigma error at a given epoch. As was stated in Sana et al. (2013), the confidence threshold of 4.0 gives a false positive rate of 1/1000 given the sampling of our measurements. The value of C = 20 km s−1 corresponds to the maximum RV variation that can be caused by stellar variability (e.g. Ritchie et al. 2009).
Following this method, we detect three sources (B86, B150, and B253) with significant RV variations in our sample with multi-epoch observations. Including source 326, which is an SB2, we find a total of four binaries in our sample of 15 stars with multi-epoch spectroscopy. The latter results in an observed spectroscopic binary fraction fbin = 0.27 ± 0.11, where the 1σ error comes from the size of the sample. Adopting C values of 15 and 10 km s−1 would result in an observed spectroscopic binary fraction of our sample of fbin = 0.33 ± 0.12 and fbin = 0.40 ± 0.13, respectively, the difference being that lower C values are more favorable to detect longer period binaries.
Including the two SB2s in the center of M17 (CEN1a and CEN1b; Bordier et al. 2022, see Sec. 4.3), we obtain an observed binary fraction fbin = 0.35 ± 0.08. A recent study from Stoop et al. (2024) shows that about 30% of the O stars in the original population in the center of M17 have been dynamically ejected. In total, they find 13 OB runaway stars. From single-epoch spectroscopy, two of those runaways are identified as SB2s. Assuming that the rest of their sample consists of single stars, including the SB2 runaway stars results in an observed binary fraction fbin = 0.21 ± 0.08 (eight binaries out of 30 stars). We note that this value corresponds to a lower limit of the observed fbin, as it is possible that other runaway stars are also binaries. Multi-epoch spectroscopy is needed in order to better constrain the observed fbin including the runaways.
Fig. 1 Observed RV dispersion σ1D as a function of age of a number of Galactic clusters and R136 in the LMC. The dashed black line shows the positive correlation of σ1D with age (see also Ramírez-Tannus et al. 2021). The gray-shaded region indicates the 1σ uncertainty of the fit. Blue points indicate the clusters included in Sana et al. (2012), green points show Wd2, Trumpler 14, and R136, and orange points M17 and the young clusters in Ramírez-Tannus et al. (2020). |
3.3 Intrinsic binary fraction
The binary fraction reported in Section 3.2 corresponds to a minimum estimate of the real binary fraction of the OB stars in M17. The intrinsic fraction depends on the sensitivity of our observations in terms of time sampling and RV measurement accuracy as well as on the orbital properties and orbital inclinations of the stars in M17. In order to provide constraints on the intrinsic binary fraction of our sample, we took a Monte-Carlo approach. The methodology is based on that described in Sana et al. (2013).
In this approach, we simulated multi-epoch RV measurements of stars in clusters and determined the largest change in RV (ΔRV) for the individual stars. The distribution of ΔRV values depends on the time sampling, which was chosen to be identical to the observations, and the intrinsic binarity of the population and orbital properties of the binaries. We determined the agreement between the simulated and the observed ΔRV distributions using a Kolmogorov-Smirnov (KS) test. We varied fbin and Pcutoff considered in the simulations to find a best-fitting simulated distribution. Below, we list the distributions considered for the other orbital parameters.
For the primary masses, we adopted the masses derived in Section 2.2, and created random systems, drawing from distributions based on the binary properties derived by Sana et al. (2012) for Galactic young clusters. The mass-ratio distribution is uniform with 0.1 < q < 1. The probability density function of the period is described as pdf(P) ∝ (log P)−0.5, with the period in days and log Pcutoff < log P < 3.5. The eccentricity distribution depends on the period of the binary system: for P < 4 days, we assumed circular orbits; for periods between 4 and 6 days, the eccentricities were sampled from pdf(e) ∝ e−0.5, with 0 ≤ e < 0.5; and for periods longer than 6 days the same distribution was used, but with 0 ≤ e < 0.9. We varied the binary fraction, fbin, from 0.1 to 1 in steps of 0.01 and the Pcutoff from 1.4 to 3500 days in steps of 1 day.
Figure 2 shows the resulting p-values for the KS test in the above-described procedure for different combinations of fbin and Pcutoff. The figure shows the general trend that our observations can be drawn from parent populations containing wider orbits as the binary fraction increases. We find that our observed sample has the highest probability of having been drawn from a parent population with a binary fraction of 87% and a minimum period of 50 days. The size of the errors on the RV measurements has a significant influence on the results. Increasing the RV errors by a few km s−1 has the consequence that populations with lower intrinsic binary fractions have a higher probability of being drawn from the same parent distribution as our observations. The contour in Figure 2 shows p-values from the KS-test of 0.1. Only populations outside that contour can be rejected as being drawn from the same parent population as our sample.
Fig. 2 p-values (color bar) for the KS-test obtained from comparing the ΔRV from our observations to that of simulated populations with different combinations of fbin and Pcutoff. The white line indicates the 0.1 contour outside of which we can reject the null hypothesis that the simulated populations are drawn from the same parent population as our observations. The red dot at fbin = 87% and Pcutoff = 50 days shows the population that best represents our observations. |
4 Discussion
4.1 Radial-velocity dispersion of M17
In order to explore possible biases in the original target selection, in this section we explore the effect of measuring σ1D on different subsamples of our M17 dataset. Figure 3 shows the σ1D distribution calculated in Section 3.1 for our sample of young stars in M17, with the exception of source 326 (SB2), for which we could not measure RV (N = 20), and for two sets of subsamples; the stars added in this work (new stars; N = 9) versus the original sample (N = 11; Sana et al. 2017; Ramírez-Tannus et al. 2021) and the stars with IR excess (N = 8) versus those with no IR-excess (N = 12; Backs et al. 2024). On the one hand, the σ1D obtained for the stars with (solid blue line) and without (dashed blue line) disks agree within the errors. Nevertheless, the shapes of the distributions differ significantly. The latter is because of the presence of the binary B150 in the disk sample; this star shows significant RV variations that cause the distribution to broaden. Another star with significant ΔRVs is present in the sample without disks, but in this case the sample size is larger, so the effect of including it is not as significant as in the sample with disks. That the σ1D in both subsamples is similar suggests that in M17, which does not host many close binaries, the lifetime of disks is not affected due to binary interactions. On the other hand, the distributions showing the original 2017 sample (dashed orange line) versus the newly added stars (solid orange line) show a significant difference in σ1D, where the original sample has a lower σ1D. Nevertheless, these two subsamples are consistent with each other, considering their 1.5 sigma uncertainty. If the difference were real, this would imply that the original sample selected by Ramírez-Tannus et al. (2017) lacks close binaries with respect to the sample in this paper (including the binaries B86, B150, and B253). However, given the newly included binaries in both the distributions with and without disks, we do not think that the reason for the difference in the original versus new sample is due to a bias in the 2017 sample selection, which was tailored to target young objects with circumstellar disks.
Fig. 3 σ1D for our full sample and for subsamples thereof. The full sample is shown with the solid black line. The orange lines show the 11 stars from Sana et al. (2017) (dashed line) and the nine new stars added to the sample in this paper (solid line). The blue lines show the sample divided in stars with (solid line) and without disks (dashed line). |
4.2 Spatial correlation
Figure 4 shows the position of our O and B stars with respect to the center of NGC 6618 (18:20:29.53, −16:10:35.56; Stoop et al. 2024). The colors of the markers indicate the maximum RV difference, ΔRV, between any two epochs, with the larger ΔRVs corresponding to the close binaries. The location of the very high-mass multiple system CEN lab coincides with the location of the center of the cluster. There is no evident correlation with the location of the stars displaying the largest ΔRV. This suggests that close binarity in M17 is not dependent on the distance to the cluster center, and indicates that our result about the lack of close binaries in M17 is independent of the position of the selected targets.
4.3 Transversal velocity dispersion
The radial velocities measured in this paper represent the motion of individual stars along the line of sight and are usually dominated by binary motions (or intrinsic variability such as pulsations). In contrast, the proper motions reflect the velocity of the systems inside the cluster and are a measurement of their movement due to the cluster dynamics. With the aim of comparing these two quantities, in this section we derive the 1D velocity dispersion of the clusters in Figure 1 based on the proper motion of the cluster members (σ1D,RA, σ1D,Dec). To find the cluster members, we used the cluster catalog with Gaia DR3 by Hunt & Reffert (2023). We excluded R136, because being in the LMC it is not included in their catalog. To calculate σ1D,RA, σ1D,DEC, we used Equation (1) and the distance listed in Hunt & Reffert (2023) for the clusters with more than 100 listed members. For the three clusters in which the catalog did not have enough stars, M17, IC 2944, and Wd2, we adopted distances from Stoop et al. (2024), Zhang (2023), Cantat-Gaudin et al. (2018), respectively. Table 2 lists σ1D,RA, σ1D,Dec and the adopted distance for each of the clusters. For all clusters, these values range from 0.4 to 2.3 km s−1 and do not show a relation with the age of the cluster. This is in agreement with the migration scenario, as we expect the binaries to harden, and therefore σ1D to increase with age, but we do not expect the dispersion of the cluster due to cluster dynamics (σ1D,RA and σ1D,Dec) to vary significantly within the explored time frame (0 to ~5 Myr). Additionally, it shows that typical cluster motions alone cannot explain the measured σ1D and implies that there must be a contribution from binary motions.
The latter underlines the need to be careful when using σ1D to derive virial masses of star clusters. For the young clusters studied here, the line-of-sight velocity dispersion is dominated by binary motions rather than by random motions of the individual systems; therefore, the cluster mass is much less than implied by the virial mass (proportional to σ1D2). Cluster masses of star clusters in virial equilibrium can be constrained reliably using the dispersion of the transverse velocity component.
Fig. 4 Location of our targets with respect to the center of M17. The dashed circle shows the size of the cluster (~2′; Stoop et al. 2024), centered on the position of CEN lab (black cross). The color bar represents the maximum RV difference between any two epochs for the sources with multi-epoch spectra. The sources with disks are marked with white edges and the SB2 326 is shown with a white star. The position of the sources with single-epoch observations is shown with an orange triangle. The figure shows that there are no indications that the location of the shortest-period binaries is dependent on distance to the cluster center. |
4.4 Eccentricity
Moe & Di Stefano (2015b) measured eccentricities of a sample of young B-type binaries and show that the youngest binaries (<1 Myr) in the sample have eccentricities of ~0.6. It could be argued that a sample with many binaries in eccentric orbits would result in a lower σ1D, since the stars spend most of their time near apastron.
In Appendix C, we explore the effect of changing the eccentricity on the properties of the simulated populations. First, we simulated single-epoch observations of populations with all properties, as is described in Section 3.3, and we varied the power-law index of the eccentricity sampling from −0.5 (corresponding to the Sana et al. (2012) empirical value) to 3 (a strong preference for eccentric orbits). Populations following the Sana et al. (2012) empirical properties (for ~2–5 Myr old clusters) have σ1D of around 40 km s−1. The effect of having preferentially eccentric orbits, even in the most extreme case in which we allow all orbits to be eccentric, reduces the σ1D of the simulated populations by at most 6 km s−1. This is not enough to explain the low σ1D observed in M17. Second, we simulated a sample for which multi-epoch observations are available with the same properties as ours. Assuming fbin = 0.7 and Pcutoff = 1.4 days (Sana et al. 2012) and having a strong preference for eccentric orbits has the effect of increasing the predicted ΔRVs, which is the opposite of what we observe in M17, where the ΔRVs are lower than those in ~2 to 5 Myr old clusters. We conclude that more eccentric orbits can explain neither the low σ1D nor the observed ΔRVs of M17. In the best case scenario, the effect of high eccentricities is minor compared to the effect of altering fbin and Pcutoff of the sample.
4.5 Timescale for binary hardening
Assuming a universal binary fraction of fbin = 0.7 (Sana et al. 2012), we can estimate the Pcutoff that best represents the measured σ1D for each cluster following the procedure described in Ramírez-Tannus et al. (2021). A more accurate synthetic RV distribution could be obtained by including a variable binary fraction that depends on the mass of the primary star. However, due to the relatively small size of most of our samples, we refrained from including this and assumed a single binary fraction for all stars.
The best-fitting distribution to explain the σ1D = 8.9 ± 0.5 km s−1 observed for M17 corresponds to Pcutoff = 415 days, and distributions with Pcutoff < 85 and < 30 days can be rejected with 68% and 95% confidence, respectively. Adopting an intrinsic binary fraction for the young stars in M17 of 87%, as was derived in Section 3.3, leads to a best-fit distribution with Pcutoff = 780 days, and distributions with Pcutoff < 188 and < 73 days can be rejected with 68% and 95% confidence, respectively. The upper limit for Pcutoff derived for ~2 to 5 Myr old clusters is a few to a few tens of days; an order of magnitude lower than that derived for M17.
These statistical constraints still allow for a few systems with shorter periods to exist. In fact, systems with periods <10 days have been detected in young clusters; for example, some B-type stars in the Moe & Di Stefano (2015a) sample have companions with P = 3.0–8.5 days and extreme mass ratios. Additionally, Kiminki & Smith (2018) find a binary with P = 9.4 days located on the outskirts of Tr 14 (HDE 303312).
In order to get an idea of the timescale of binary hardening, we fit the function to the data in Figure 5, where P0 is the minimum period at the moment of binary formation and t0 corresponds to the e-folding time. We assumed a value for P0 = 105 days, which is the period corresponding to a pair of 10 M⊙ stars with a separation of ~100 au (Ramírez-Tannus et al. 2021) and c = 1.4 days, which is the cutoff period of ~2–5 Myr old Galactic clusters (Sana et al. 2012). We obtain a typical e-folding time of Myr. Sana et al. (2012) finds the minimum period for ~2–5 Myr old clusters to be 1.4 days. To harden an orbit from 415 days to 1.4 days, ln(415/1.4) ≈ 6 e-folding times are needed. With t0 = 0.15 Myr the typical hardening timescale is ~0.85 Myr. For a system with two 10 M⊙ stars, this would imply a hardening rate of 3.4 au Myr−1. Adopting a Pcutoff of 780 days would lead to a hardening timescale of 1 Myr and a hardening rate of 4.4 au Myr−1. Assuming Pcutoff of 85 and 188 days would lead to hardening rates of 1.6 au Myr−1 and 2.1 au Myr−1, respectively. These rates are lower than the values of 4–7.4 au Myr−1 reported by Ramírez-Tannus et al. (2021); the reason for the difference is the significantly lower σ1D measured in the 2017 sample, which leads to a much higher Pcutoff for the best-fit distribution. Nevertheless, the hardening rates remain significant and require efficient dynamical processes.
Fig. 5 Pcutoff as a function of age of young clusters with fbin= 0.70. The uncertainties correspond to the distributions that represent the 16th to 84th percentile of the σ1D distribution. The solid line and shaded region show the fit to the data and its 1σ uncertainty. |
4.6 Migration scenario
Both our single- and multi-epoch analyses for a sample of 20 and 15 massive stars, respectively, in the young star-forming region M17 point to a lack of short-period binaries relative to older clusters (~2–5 Myr), where close binaries with periods of days are common. From the multi-epoch observations, we derive a binary fraction that, within uncertainties, is consistent with the universal value of ~70% in massive clusters of ages up to ~107 yr. These results support the migration scenario for close-binary formation proposed by Sana et al. (2017) and Ramírez-Tannus et al. (2021). We mention that if massive stars are preferentially in wider orbits during cluster formation this aids in the production of runaway stars, as the larger effective cross section of wide binary systems favors dynamical interactions (e.g., Fujii & Portegies Zwart 2011; Stoop et al. 2024).
As for which physical process may be responsible for binary hardening, mechanisms that have been proposed include interaction with a circumbinary disk, the Eccentric Kozai-Lidov (EKL) effect in triple systems (Sana et al. 2017), dynamical interactions among stellar systems in the cluster environment, and tidal circularization and orbital decay (Moe & Di Stefano 2015b). We discuss each of these possibilities below. Before doing so, we mention that dynamical disruption of coplanar (hierarchical) triples that initially fragmented within the disk in combination with energy dissipation occurring within the disk, may potentially harden the inner binary. However, this process is expected to be efficient in earlier pre-main-sequence or proto-stellar phases (Moe & Kratter 2018) and disk fragmentation is expected to produce predominantly low-mass companions (Oliva & Kuiper 2020), which are more difficult to detect spectroscopically.
Star-disk interaction: For two reasons, we anticipate that interaction with a circumbinary disk is not an efficient mechanism for the removal of orbital angular momentum, and therefore hardening, of the binary system (see also, e.g., Wei et al. 2024). First, 13 out of 21 sources in our sample do not show evidence for the presence of a disk, while the anticipated time period of binary hardening of ~106 yr is still ahead. This implies that in most cases disk removal happens prior to binary hardening, and may also suggest that the typical disk removal timescale is short compared to the binary hardening time. Second, from ALMA millimeter observations the total disk masses of B243, B268, B275, and B331 are estimated to be no higher than 9 × 10−3 M⊙ (Poorta et al., in prep.). Radiation thermodynamical modeling of hydrogen emission lines of B243 and B331 yield disk masses on the order of 10−3−10−4 M⊙ inside 20 au (Backs et al. 2023). It seems unlikely that these remnant disks are capable of storing the orbital angular momentum that is to be removed from the binary in the overall hardening process. The latter is in agreement with Tokovinin & Moe (2020), who show that disk migration must occur during the embedded protostellar phase, which is much shorter than the age of our systems (Tokovinin & Moe 2020).
Eccentric Kozai–Lidov mechanism: From interferometric observations of a sample of six young O-type stars in M17 with VLT/GRAVITY, Bordier et al. (2022) find a companion fraction of 2.3 ± 0.6; in other words, that each of the sources likely has (at least) two stellar companions. Rose et al. (2019) investigated the potential for orbit hardening through the EKL-mechanism. Their findings indicate that, starting with a cutoff period of ~9 months, proposed by Sana et al. (2017), the EKL-mechanism falls short in replicating the population of short-period binaries observed by Sana et al. (2012) for ~2–5 Myr old clusters. With our new observations, we discard distributions with cutoff periods ≲ 85 days with 68% confidence. New simulations are needed to test whether the EKL-mechanism can build up a sufficient population of short-period binaries within ~106 yr when the initial period is ~3 months.
Dynamical interactions: Ramírez-Tannus et al. (2021) state that due to the low stellar densities observed in the studied clusters, stellar encounters are expected to be infrequent. Consequently, mechanisms like binary-binary or single-binary interactions (Bate et al. 2002; Bate 2009) were not expected to play a substantial role in the formation of close binaries. Nevertheless, the evidence for many runaway stars in M17 (Stoop et al. 2024) and the fact that we expect several wide binaries in this cluster could indicate that these interactions are more important than thought before. If that were the case, dynamical interactions would be a hardening mechanism, resulting in runaways (e.g. disrupted binaries; Parker & Goodwin 2012) and close binaries.
Tidal circularization and orbital decay: Moe & Di Stefano (2015b) report that the highly eccentric young binaries in their sample nearly fill their Roche lobes at periastron. These young sources tidally circularize and decay toward shorter periods on timescales of a few million years. These two effects together could explain how young stars that are born in wide orbits harden in the first million years of evolution to represent the properties observed in ~2–5 Myr clusters. In order to test this scenario, observations of additional epochs are needed in order to find the orbital solutions of the systems and to test if there is a trend of eccentricity with cluster age.
Models of massive-star populations including binary stars predict that only a small percentage of the binaries will interact within the first 3 Myr of evolution. In most cases in which interaction does occur, the stars will merge to produce a blue straggler (Schneider et al. 2015). Our findings about the hardening of binaries during the first million years, unless specifically concerned with early blue-straggler formation, are not expected to influence predictions made by binary-population synthesis models. Therefore, starting with the initial conditions described in Sana et al. (2012) remains a valid approach if one is interested in the outcome of binary evolution after the first few million years.
5 Conclusions
With the goal of testing the early main-sequence migration scenario for the formation of massive binary stars, we performed a RV analysis of a population of 21 young stars in the ~0.7 Myr old giant HII region M17. Our results can be summarized as follows:
We find that both single-epoch and multi-epoch analyses with typical sample sizes of around ten stars lead to the conclusion that there is a lack of short-period binaries in the young region M17 relative to ~2–5 Myr old clusters;
The derived RV dispersion for the extended sample presented in this paper is σ1D = 8.9 ± 0.5 km s−1. This is higher than the σ1D = 5.6 ± 0.2 km s−1 derived for 11 stars by Sana et al. (2017). The difference is mainly due to the inclusion of three single-line spectroscopic binaries (SB1; B150, B253, and B86) in the new sample. However, the new, somewhat higher value observed for the young stars in M17 remains strikingly low in comparison with the ~30 km s−1 observed in older clusters with a similar mass in the Milky Way;
The σ1D obtained for samples containing stars with and without disks agree within the errors. This implies that the low σ1D reported by Sana et al. (2017) is not due to the initial sample selection that aimed to target stars with IR-excess (Ramírez-Tannus et al. 2017);
We derive a spectroscopic observed binary fraction for M17 of fbin = 27%. Correcting for observational biases, we estimate an intrinsic binary fraction of 87%, consistent with the universal value (Sana et al. 2012);
Assuming fbin = 70%, the most likely Pcutoff to explain the observed σ1D in M17 is 415 days. We reject distributions with Pcutoff < 85(30) days with 68(95)% confidence, respectively. The derived value of fbin = 87% leads to a best-fit distribution with Pcutoff = 780 days and distributions with Pcutoff < 188(73) days can be rejected with 68(95)% confidence;
The transverse velocity dispersion of stellar systems in star clusters, derived from their proper motions (σ1DRA, σ1D,Dec), is not affected by binary motions and only reflects random motions. We find that it does not vary with age and ranges between 0.4 and 2.3 km s−1. We stress the importance of using the tangential velocity dispersion, rather than relying on the RV dispersion dominated by binary motions, when estimating the masses of star clusters thought to be in virial equilibrium. Using the latter would lead to an overestimate of the star cluster mass.
We confirm that three to 30 M⊙ stars in the young massive cluster M17 fit in a trend of binary hardening on the pre- and early main-sequence first presented by Ramírez-Tannus et al. (2021). This early migration process shows that early binary evolution together with possibly dynamical interactions are key processes in the formation of gravitational wave sources. Though our findings are statistically significant, the validity of the trend deserves further scrutiny. Larger and unbiased samples of young massive stars are needed to address the dependence of the underlying distributions on primary mass (see Moe & Di Stefano 2017). Additional epoch observations will allow us to find the exact periods, mass ranges, and eccentricities of young systems like M17, to be able to fully constrain close binary formation theories. The more precise the trend of hardening with time is established, the better the trend itself can be used to age-date young clusters from relatively simple spectroscopic datasets.
Data availability
Tables containing the log of observations and radial velocity measurements are available 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/690/A178
Acknowledgements
We would like to thank Max Moe for his valuable suggestions which have improved the quality of this manuscript. We thank Tim Michael Heinz Wolf for the assistance developing RVFitter, the code used to measure radial velocities, and Stefano Rinaldi and Morgan Fouesneau for valuable discussions about the statistical approach of this paper. MCRT acknowledges support by the German Aerospace Center (DLR) and the Federal Ministry for Economic Affairs and Energy (BMWi) through program 50OR2314 ‘Physics and Chemistry of Planet-forming disks in extreme environments’. The research leading to these results has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement numbers 772225: MULTIPLES). This work is based on observations collected at the European Organization for Astronomical Research in the Southern Hemisphere under ESO programs 60.A-9404(A), 085.D-0741, 089.C-0874(A), 091.C-0934(B) and 103.D-0099. This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France. This research made use of Astropy (Astropy Collaboration 2013, 2018, 2022), NumPy (Harris et al. 2020), and Matplotlib (Hunter 2007).
Appendix A Spectral classification
A.1 OB stars
B86-B9 III
This star has relatively strong Mg II absorption at 4481 Å compared to the neighboring He i at 4471 Å line, which indicates that B86 is a B8-B9 star. Since the spectrum shows Fe II 4233 Å and considering the ratio of Mg II to the bordering He I line, the star is classified as B9. The luminosity class is III since the Si II lines at 4128-30 Å are of similar strength compared to He I 4026 Å.
B93 - B2 V
B93 shows no He II lines, but only He I lines in the spectrum, where He 14144 Å is stronger than He 14121 Å, indicating early B, but not as hot as B0 or B1. The Si III 4552 Å ratio with Si II 4128 Å and with Mg II 4481 Å are similar to that of spectral type B2. This is confirmed by the presence of O II 4070-76 Å. Since N II 3995 Å is not detected, the luminosity class is V.
B181 - B0 V
We detect one strong He II line in this star at 4686 Å and some He I lines. This corresponds to spectral type B0, as we do not detect He II 4541 Å as is expected for late O-type stars. The He II line is stronger than He I 4711 Å, indicating luminosity class V.
B205 - B2 V
The stacked spectrum of B205 has strong He i lines and no He II lines, which indicates that the star must be an early B type star but cooler than B0. Based on the ratio of He I 4142 Å and He I 4144 Å (which is 1:2) and the detection of Si iii at 4552, 4568 and 4575 Å, the star is classified as B2. Since the hydrogen lines are relatively broad and OII is not detected at 4350 Å, the luminosity class is V.
B213 - B3 III-V
The spectrum shows strong He I lines compared to HI, but no He II lines. The Mg II at 4481 Å is weak compared to He I at 4471 Å and Si IV 4089 Å is not detected, suggesting a B3 classification. The spectrum shows relatively broad HI and He I lines, but is too noisy to distinguish between luminosity class III and V.
B234 - B3 III-V
The spectrum of B234 is similar to B213 and we arrive at the same spectral classification.
B272 - B7 III
B272 is of late-B type since the spectrum shows a relatively strong Mg II feature at 4481 Å. However, He I at 4471 Å is stronger than Mg II in the ratio 3:2. This indicates a spectral type B7. The luminosity class III is based on the similar strength of the Si II lines at 4128-30 Å and He I 4026 Å.
CEN55 - B2 III
The spectrum shows similar characteristics as B93 and B205 that have a spectral type B2. CEN55 shows N II 3995 Å in the spectrum corresponding to luminosity class III.
326 - O6 III + B1 V
Visual inspection of the spectrum of 326 reveals that it is a double-lined spectroscopic binary (SB2). The last epoch shows a relatively large spread between the absorption lines in a few He I lines and H I. This epoch is used to give a tentative indication of spectral type.
One of the stars shows strong He II lines and exhibits characteristics of an O6 star; He II at 4200 Å is of similar strength as He I at 4026 Å and N III 4634-40-42 Å is in emission. The luminosity class is III inferred from a similar strength of He II 4541 Å and He I 4471 Å.
The second star does not seem to have He II lines. Its Mg II 4481 Å is relatively weak compared to He I 4471 Å where the latter is also weaker than in the O6 III star. He I 4144 Å is similarly strong as He I 4121, which is stronger than Si II 4128-30 Å, therefore this star is an early B type. The spectrum shows O II 4070-76 Å, Si III at 4552, 4568 and 4575 Å and Si IV 4089 Å, which is found for B1 V stars.
A.2 Cooler stars
The cooler stars in the sample were given a general classification using distinct features in Gray (2000). Strong hydrogen lines indicate an A-type star, where for F-type stars the Ca II K-line grows stronger and the G-band starts to appear. The G-band is strongest in the G-type stars in the sample.
Fig. A.1 Spectra of the stars at the wavelengths used for classification. The stars are listed from hottest (top) to coolest (bottom). B337 and B331 are not included due the high extinction in this wavelength range. |
Appendix B Lines used to measure radial velocity per star
Spectral lines used per star to measure the RV.
Appendix C Eccentricity
In this section we focus in the effect of varying the eccentricity e on the observed σ1D of clusters. We follow the approach described in Sana et al. (2017). For the basic properties of the population we adopt the values of Sana et al. (2012) for Galactic young clusters. The eccentricity distribution depends on the period of the binary. For P < 4 days circular orbits are assumed, for periods between 4 and 6 days, the eccentricities are sampled from pdf(e) ∝ e−0.5, with 0 ≤ e < 0.5; for periods longer than 6 days the same distribution is used, but with 0 ≤ e < 0.9. The left panel of Figure C.1 shows the effect of varying the power index of the eccentricity, γ, from −0.5 (corresponding to Sana et al. (2012) empirical value) to 3 (strong preference for eccentric orbits) in steps of 0.35 on the σ1D of different populations. The σ1D value varies from ~41 km s−1 to ~40 km s−1. In the right panel we show the effect of changing the eccentricity power as below, but removing the eccentricity dependence on the binary period. In this extreme case the eccentricity varies from ~40 km s−1 to ~34 km s−1, in none of the cases the effect of eccentricity is enough to account for the low σ1D observed in M17.
Fig. C.1 Effect of changing the power index of the eccentricity, γ, from −0.5 to 3. Left: Keeping the dependency of the eccentricity distribution on the period of the binary. For P < 4 days circular orbits are assumed, for periods between 4 and 6 days, the eccentricities are sampled from pdf(e) ∝ e−0.5, with 0 ≤ e < 0.5; for periods longer than 6 days the same distribution is used, but with 0 ≤ e < 0.9. Right: Allowing all orbits to be eccentric. |
References
- Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2017, Phys. Rev. Lett., 119, 161101 [Google Scholar]
- Almeida, L. A., Sana, H., Taylor, W., et al. 2017, A&A, 598, A84 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Astropy Collaboration (Robitaille, T. P., et al.) 2013, A&A, 558, A33 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Astropy Collaboration (Price-Whelan, A. M., et al.) 2018, AJ, 156, 123 [Google Scholar]
- Astropy Collaboration (Price-Whelan, A. M., et al.) 2022, ApJ, 935, 167 [NASA ADS] [CrossRef] [Google Scholar]
- Backs, F., Poorta, J., Rab, C., et al. 2023, A&A, 671, A13 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Backs, F., Brands, S. A., Ramírez-Tannus, M. C., et al. 2024, A&A, 690, A113 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Banyard, G., Sana, H., Mahy, L., et al. 2022, A&A, 658, A69 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Barbá, R. H., Gamen, R., Arias, J. I., & Morrell, N. I. 2017, in The Lives and Death-Throes of Massive Stars, 329, eds. J. J. Eldridge, J. C. Bray, L. A. S. McClelland, & L. Xiao, 89 [Google Scholar]
- Bate, M. R. 2009, MNRAS, 392, 590 [Google Scholar]
- Bate, M. R., Bonnell, I. A., & Bromm, V. 2002, MNRAS, 336, 705 [Google Scholar]
- Baume, G., Rodríguez, M. J., Corti, M. A., Carraro, G., & Panei, J. A. 2014, MNRAS, 443, 411 [NASA ADS] [CrossRef] [Google Scholar]
- Blaauw, A. 1991, in NATO Advanced Science Institutes (ASI) Series C, 342, NATO Advanced Science Institutes (ASI) Series C, eds. C. J. Lada, & N. D. Kylafis, 125 [NASA ADS] [Google Scholar]
- Bonnell, I. A., & Bate, M. R. 2005, MNRAS, 362, 915 [NASA ADS] [CrossRef] [Google Scholar]
- Bordier, E., Frost, A. J., Sana, H., et al. 2022, A&A, 663, A26 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Bosch, G., Terlevich, E., & Terlevich, R. 2009, AJ, 137, 3437 [Google Scholar]
- Broos, P. S., Getman, K. V., Povich, M. S., et al. 2013, ApJS, 209, 32 [NASA ADS] [CrossRef] [Google Scholar]
- Cantat-Gaudin, T., Jordi, C., Vallenari, A., et al. 2018, A&A, 618, A93 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Cheetham, A. C., Kraus, A. L., Ireland, M. J., et al. 2015, ApJ, 813, 83 [NASA ADS] [CrossRef] [Google Scholar]
- de Mink, S. E., & Mandel, I. 2016, MNRAS, 460, 3545 [NASA ADS] [CrossRef] [Google Scholar]
- de Mink, S. E., Langer, N., Izzard, R. G., Sana, H., & de Koter, A. 2013, ApJ, 764, 166 [Google Scholar]
- Derkink, A. R., Ramírez-Tannus, M. C., Kaper, L., et al. 2024, A&A, 681, A112 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Dunstall, P. R., Dufton, P. L., Sana, H., et al. 2015, A&A, 580, A93 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Frost, A. J., Sana, H., Mahy, L., et al. 2024, Science, 384, 214 [NASA ADS] [CrossRef] [Google Scholar]
- Fujii, M. S., & Portegies Zwart, S. 2011, Science, 334, 1380 [Google Scholar]
- Göppl, C., & Preibisch, T. 2022, A&A, 660, A11 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Gray, R. O. 2000, A digital spectral classification atlas, https://ned.ipac.caltech.edu/level5/Gray/frames.html [Google Scholar]
- Hanson, M. M., Howarth, I. D., & Conti, P. S. 1997, ApJ, 489, 698 [NASA ADS] [CrossRef] [Google Scholar]
- Harris, C. R., Millman, K. J., van der Walt, S. J., et al. 2020, Nature, 585, 357 [Google Scholar]
- Hénault-Brunet, V., Evans, C. J., Sana, H., et al. 2012, A&A, 546, A73 [CrossRef] [EDP Sciences] [Google Scholar]
- Hillwig, T. C., Gies, D. R., Bagnuolo, William G., J., et al. 2006, ApJ, 639, 1069 [NASA ADS] [CrossRef] [Google Scholar]
- Hunt, E. L., & Reffert, S. 2023, A&A, 673, A114 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Hunter, J. D. 2007, Comput. Sci. Eng., 9, 90 [Google Scholar]
- Jensen, E. L. N., Mathieu, R. D., & Fuller, G. A. 1996, ApJ, 458, 312 [Google Scholar]
- Kausch, W., Noll, S., Smette, A., et al. 2015, A&A, 576, A78 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Kiminki, M. M., & Smith, N. 2018, MNRAS, 477, 2068 [Google Scholar]
- Kobulnicky, H. A., Kiminki, D. C., Lundquist, M. J., et al. 2014, ApJS, 213, 34 [Google Scholar]
- Kraus, A. L., Ireland, M. J., Hillenbrand, L. A., & Martinache, F. 2012, ApJ, 745, 19 [Google Scholar]
- Langer, N. 2012, ARA&A, 50, 107 [CrossRef] [Google Scholar]
- Lim, B., Sung, H., Kim, J. S., Bessell, M. S., & Karimov, R. 2014, MNRAS, 438, 1451 [Google Scholar]
- Liu, Z., Cui, W., Liu, C., et al. 2019 arXiv e-prints [arXiv:1902.07607] [Google Scholar]
- Mason, B. D., Hartkopf, W. I., Gies, D. R., Henry, T. J., & Helsel, J. W. 2009, AJ, 137, 3358 [Google Scholar]
- Modigliani, A., Goldoni, P., Royer, F., et al. 2010, in Proc. SPIE, 7737, 773728 [Google Scholar]
- Moe, M., & Di Stefano, R. 2015a, ApJ, 801, 113 [NASA ADS] [CrossRef] [Google Scholar]
- Moe, M., & Di Stefano, R. 2015b, ApJ, 810, 61 [NASA ADS] [CrossRef] [Google Scholar]
- Moe, M., & Di Stefano, R. 2017, ApJS, 230, 15 [Google Scholar]
- Moe, M., & Kratter, K. M. 2018, ApJ, 854, 44 [NASA ADS] [CrossRef] [Google Scholar]
- Offner, S. S. R., Moe, M., Kratter, K. M., et al. 2023, in Astronomical Society of the Pacific Conference Series, 534, Protostars and Planets VII, eds. S. Inutsuka, Y. Aikawa, T. Muto, K. Tomida, & M. Tamura, 275 [NASA ADS] [Google Scholar]
- Oliva, G. A., & Kuiper, R. 2020, A&A, 644, A41 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Parker, R. J., & Goodwin, S. P. 2012, MNRAS, 424, 272 [NASA ADS] [CrossRef] [Google Scholar]
- Penny, L. R., Gies, D. R., Hartkopf, W. I., Mason, B. D., & Turner, N. H. 1993, PASP, 105, 588 [NASA ADS] [CrossRef] [Google Scholar]
- Pietrzyński, G., Graczyk, D., Gallenne, A., et al. 2019, Nature, 567, 200 [Google Scholar]
- Pols, O. 1997, in Astronomical Society of the Pacific Conference Series, 130, The Third Pacific Rim Conference on Recent Development on Binary Star Research, ed. K.-C. Leung, 153 [NASA ADS] [Google Scholar]
- Ramírez-Tannus, M. C., Kaper, L., de Koter, A., et al. 2017, A&A, 604, A78 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Ramírez-Tannus, M. C., Poorta, J., Bik, A., et al. 2020, A&A, 633, A155 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Ramírez-Tannus, M. C., Backs, F., de Koter, A., et al. 2021, A&A, 645, L10 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Ritchie, B. W., Clark, J. S., Negueruela, I., & Crowther, P. A. 2009, A&A, 507, 1585 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Rochau, B., Brandner, W., Stolte, A., et al. 2011, MNRAS, 418, 949 [NASA ADS] [CrossRef] [Google Scholar]
- Rose, S. C., Naoz, S., & Geller, A. M. 2019, MNRAS, 488, 2480 [Google Scholar]
- Sana, H., Gosset, E., Nazé, Y., Rauw, G., & Linder, N. 2008, MNRAS, 386, 447 [Google Scholar]
- Sana, H., Gosset, E., & Evans, C. J. 2009, MNRAS, 400, 1479 [NASA ADS] [CrossRef] [Google Scholar]
- Sana, H., Momany, Y., Gieles, M., et al. 2010, A&A, 515, A26 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Sana, H., James, G., & Gosset, E. 2011, MNRAS, 416, 817 [Google Scholar]
- Sana, H., de Mink, S. E., de Koter, A., et al. 2012, Science, 337, 444 [Google Scholar]
- Sana, H., de Koter, A., de Mink, S. E., et al. 2013, A&A, 550, A107 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Sana, H., Le Bouquin, J. B., Lacour, S., et al. 2014, ApJS, 215, 15 [Google Scholar]
- Sana, H., Ramírez-Tannus, M. C., de Koter, A., et al. 2017, A&A, 599, L9 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Schneider, F. R. N., Izzard, R. G., Langer, N., & de Mink, S. E. 2015, ApJ, 805, 20 [Google Scholar]
- Schneider, F. R. N., Ohlmann, S. T., Podsiadlowski, P., et al. 2019, Nature, 574, 211 [Google Scholar]
- Smette, A., Sana, H., Noll, S., et al. 2015, A&A, 576, A77 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Smith, N., Li, W., Filippenko, A. V., & Chornock, R. 2011, MNRAS, 412, 1522 [NASA ADS] [CrossRef] [Google Scholar]
- Sota, A., Maíz Apellániz, J., Walborn, N. R., et al. 2011, ApJS, 193, 24 [Google Scholar]
- Stoop, M., Kaper, L., de Koter, A., et al. 2023, A&A, 670, A108 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Stoop, M., Derkink, A., Kaper, L., et al. 2024, A&A, 681, A21 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Sung, H., Bessell, M. S., Chun, M.-Y., et al. 2017, ApJS, 230, 3 [NASA ADS] [CrossRef] [Google Scholar]
- Tan, J. C., Beltrán, M. T., Caselli, P., et al. 2014, Protostars and Planets VI, 149 [Google Scholar]
- Tokovinin, A., & Moe, M. 2020, MNRAS, 491, 5158 [NASA ADS] [CrossRef] [Google Scholar]
- van der Meij, V., Guo, D., Kaper, L., & Renzo, M. 2021, A&A, 655, A31 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- van Gelder, M. L., Kaper, L., Japelj, J., et al. 2020, A&A, 636, A54 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Vernet, J., Dekker, H., D’Odorico, S., et al. 2011, A&A, 536, A105 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Villaseñor, J. I., Taylor, W. D., Evans, C. J., et al. 2021, MNRAS, 507, 5348 [CrossRef] [Google Scholar]
- Wei, D., Schneider, F. R. N., Podsiadlowski, P., et al. 2024, A&A, 688, A87 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
- Zeidler, P., Sabbi, E., Nota, A., et al. 2018, AJ, 156, 211 [Google Scholar]
- Zhang, M. 2023, ApJS, 265, 59 [NASA ADS] [CrossRef] [Google Scholar]
- Zinnecker, H., & Yorke, H. W. 2007, ARA&A, 45, 481 [Google Scholar]
All Tables
All Figures
Fig. 1 Observed RV dispersion σ1D as a function of age of a number of Galactic clusters and R136 in the LMC. The dashed black line shows the positive correlation of σ1D with age (see also Ramírez-Tannus et al. 2021). The gray-shaded region indicates the 1σ uncertainty of the fit. Blue points indicate the clusters included in Sana et al. (2012), green points show Wd2, Trumpler 14, and R136, and orange points M17 and the young clusters in Ramírez-Tannus et al. (2020). |
|
In the text |
Fig. 2 p-values (color bar) for the KS-test obtained from comparing the ΔRV from our observations to that of simulated populations with different combinations of fbin and Pcutoff. The white line indicates the 0.1 contour outside of which we can reject the null hypothesis that the simulated populations are drawn from the same parent population as our observations. The red dot at fbin = 87% and Pcutoff = 50 days shows the population that best represents our observations. |
|
In the text |
Fig. 3 σ1D for our full sample and for subsamples thereof. The full sample is shown with the solid black line. The orange lines show the 11 stars from Sana et al. (2017) (dashed line) and the nine new stars added to the sample in this paper (solid line). The blue lines show the sample divided in stars with (solid line) and without disks (dashed line). |
|
In the text |
Fig. 4 Location of our targets with respect to the center of M17. The dashed circle shows the size of the cluster (~2′; Stoop et al. 2024), centered on the position of CEN lab (black cross). The color bar represents the maximum RV difference between any two epochs for the sources with multi-epoch spectra. The sources with disks are marked with white edges and the SB2 326 is shown with a white star. The position of the sources with single-epoch observations is shown with an orange triangle. The figure shows that there are no indications that the location of the shortest-period binaries is dependent on distance to the cluster center. |
|
In the text |
Fig. 5 Pcutoff as a function of age of young clusters with fbin= 0.70. The uncertainties correspond to the distributions that represent the 16th to 84th percentile of the σ1D distribution. The solid line and shaded region show the fit to the data and its 1σ uncertainty. |
|
In the text |
Fig. A.1 Spectra of the stars at the wavelengths used for classification. The stars are listed from hottest (top) to coolest (bottom). B337 and B331 are not included due the high extinction in this wavelength range. |
|
In the text |
Fig. C.1 Effect of changing the power index of the eccentricity, γ, from −0.5 to 3. Left: Keeping the dependency of the eccentricity distribution on the period of the binary. For P < 4 days circular orbits are assumed, for periods between 4 and 6 days, the eccentricities are sampled from pdf(e) ∝ e−0.5, with 0 ≤ e < 0.5; for periods longer than 6 days the same distribution is used, but with 0 ≤ e < 0.9. Right: Allowing all orbits to be eccentric. |
|
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.