Draft version January 6, 2025

Typeset using LATEX twocolumn style in AASTeX631

Photometric Determination of Unresolved Main-sequence Binaries in the Pleiades: Binary Fraction

and Mass Ratio Distribution
Rongrong Liu ,1, 2 Zhengyi Shao ,1, 3 and Lu Li 1

1 Shanghai Astronomical Observatory, Chinese Academy of Sciences,

80 Nandan Road, Shanghai 200030, People’s Republic of China
2 School of Astronomy and Space Sciences, University of Chinese Academy of Sciences,

No. 19A Yuquan Road, Beijing 100049, People’s Republic of China

3 Key Lab for Astrophysics, Shanghai 200234, People’s Republic of China
arXiv:2501.01617v1 [astro-ph.SR] 3 Jan 2025

ABSTRACT
Accurate determination of binary fractions (fb ) and mass ratio (q) distributions is crucial for under-
standing the dynamical evolution of open clusters. We present an improved multiband fitting technique
to enhance the analysis of binary properties. This approach enables an accurate photometric deter-
mination of fb and q distribution in a cluster. The detectable mass ratio can be down to the qlim ,
limited by the minimum stellar mass in theoretical models. First, we derived an empirical model for
magnitudes of Gaia DR3 and 2MASS bands that match the photometry of single stars in the Pleiades.
We then performed a multiband fitting for each cluster member, deriving the probability density func-
tion (PDF) of its primary mass (M1 ) and q in the Bayesian framework. 1154 main-sequence (MS)
single stars or unresolved MS+MS binaries are identified as members of the Pleiades. By stacking
their PDFs, we conducted a detailed analysis of binary properties of the cluster. We found the fb of
this sample is 0.34 ± 0.02. The q distribution exhibits a three-segment power-law profile: an initial
increase, followed by a decrease, and then another increase. This distribution can be interpreted as a
fiducial power-law profile with an exponent of -1.0 that is determined in the range of 0.3 < q < 0.8, but
with a deficiency of binaries at lower q and an excess at higher q. The variations of fb and q with M1
reveal a complex binary distribution within the Pleiades, which might be attributed to a combination
of primordial binary formation mechanisms, dynamical interactions, and the observational limit of
photometric binaries imposed by qlim (M1 ).

Keywords: Open star clusters (1160) —Binary stars (154) — Mass ratio (1012) — Bayesian statistics
(1900)

1. INTRODUCTION The Pleiades cluster, one of the nearest OCs, contains

The member stars of an open cluster (OC) formed in over a thousand main-sequence stars, making it an ideal
the same molecular cloud and thus constitute a single target for detecting and analyzing binaries. As early as
stellar population (SSP) (Lada & Lada 2003). Young eighty years ago, Smith & Struve (1944) identified two
open clusters typically contain a high proportion of pri- Pleiades members as spectroscopic binaries. Over the
mordial binary (Offner et al. 2023). With frequent past forty years, the fb of the Pleiades has been dis-
three-body encounters, these binaries may dissolve or cussed by numerous studies. But significantly divergent
have their companions exchanged with other single stars, results have been yielded, from 13% to 73%, depend-
leading to changes in the statistical properties of bina- ing on different stellar mass ranges and spatial regions
ries, such as the binary fraction (fb ) and the mass ratio (Stauffer 1982; Mermilliod et al. 1992; Steele & Jameson
(q) distribution (Heggie 1975; Marks et al. 2011; Geller 1995; Bouvier et al. 1997; Martı́n et al. 2003; Pinfield
et al. 2013a,b, 2015). Therefore, these binary proper- et al. 2003; Lodieu et al. 2007; Torres 2020; Almeida
ties, as important as the stellar mass function, provide et al. 2023; Malofeeva et al. 2023).
critical constraints on the models of star formation and Determining the mass ratio of a binary is another chal-
dynamical evolution in OCs. lenge. Traditionally, q is defined as the dynamical mass
ratio derived from radial velocity variations using spec-
2

troscopic observations. Unfortunately, this method is Regardless of the technique employed, detecting of low
expensive and usually requires decades to gather suffi- mass-ratio binaries remains extremely challenging. Nev-
cient high-precision data. For instance, despite 39 years ertheless, this problem is very important for two reasons.
of data accumulation, Torres et al. (2021) identified 48 First, it is essential for a comprehensive understanding
binaries within a spectroscopic sample of 289 Pleiades of stellar mass distribution. Second, low mass-ratio bi-
members, of which only 21 had their dynamical mass naries are crucial for studying dynamical effects due to
ratios determined. Moreover, this method is less effec- their smaller binding energies, which make them more
tive for binaries with lower mass ratios, longer orbital sensitive to dynamical interactions.
periods, or higher orbital inclinations, which will result To address this problem, two key issues should be con-
in detection biases. sidered in photometric methods. One is the systematic
In contrast, using multiband photometric data to in- deviation between theoretical model magnitudes and ob-
vestigate those unresolved binaries is more efficient due servational data, which can introduce biases in the mea-
to the extensive dataset and the convenience of estab- surement of binary mass ratios. Such deviations may
lishing a complete magnitude-limited sample. For clus- be due to either the inaccuracies in models or the im-
ter members, since their distances, ages, metallicities, perfect calibrations of observations. Despite substantial
and dust extinctions are identical to those of the clus- improvements in both theoretical models and observa-
ter, the magnitudes of a cluster member in all bands tional techniques in recent years (Chen et al. 2014; Tang
are entirely dependent on its mass, or the masses of its et al. 2014; Chen et al. 2015; Fu et al. 2016, 2018; Chen
binary components. et al. 2019; Weiler et al. 2018; Evans et al. 2018; Riello
Several recent studies have investigated binaries in the et al. 2021; Montegriffo et al. 2023), minor discrepancies
Pleiades using Gaia’s photometric data on the color- remain and may affect on mass-ratio estimates. In stud-
magnitude diagram (CMD) (Niu et al. 2020; Jadhav ies of OCs, a common method to reduce these biases is to
et al. 2021; Cordoni et al. 2023; Donada et al. 2023). construct the so-called empirical isochrone. For exam-
Despite the high-precision photometric data of Gaia, ple, one can measure the color differences between the
detecting binaries with q < 0.5 remains difficult with observational main sequence and the model predictions
optical data alone. Malofeeva et al. (2022) and Mal- on the CMD, and correct the model colors (Milone et al.
ofeeva et al. (2023) improved the detection by combin- 2012; Fritzewski et al. 2019; Li et al. 2020). If multiple
ing photometric data from optical to mid-infrared bands bands are included, discrepancies of multiple colors will
on a pseudo-color diagram, which can identify binaries be measured. Then, the magnitudes of each band can
including low mass-ratio ones. However, their studies be adjusted to a reference band, followed by multiband
were limited to stars within a mass range from 0.5M⊙ to fitting (Thompson et al. 2021). An alternate approach
1.8M⊙ . This limitation is caused by the potential bias is introducing error floors for individual observational
in measurements of q, due to discrepancies between ob- magnitudes in the fitting. These additional errors should
servational magnitudes and theoretical models at lower be large enough to cover discrepancies between models
masses, as well as the difficulty of distinguishing binaries and observations, as well as the zero-point differences
from single stars at higher masses. between photometric surveys (Childs et al. 2024).
The most straightforward method of employing multi- The second issue arises from the lower limit of stel-
band photometric data is taking a multiband magnitude lar mass (∼ 0.075M⊙ ). Companions with masses below
fitting, which is broadly equivalent to the spectral en- this threshold, such as brown dwarfs or giant planets,
ergy distribution (SED) fitting. Some studies have used do not contribute a detectable increment in luminos-
this technique to detect binaries in OCs by combining ity. This factor leads to a lower limit on the mass ratio
optical and infrared photometry. For example, Thomp- of photometric binaries, which depends on the mass of
son et al. (2021) employed BINOCS to fit optical to the primary star. Therefore, when discussing photomet-
mid-infrared magnitudes for stars in eight OCs. They ric binaries, it is crucial to specify the ranges of stellar
accurately identified binaries with q ≳ 0.3 within these masses and mass ratios. However, observational errors
clusters after correcting the model isochrone of SSP. often make it difficult to distinguish low mass-ratio bi-
Similarly, Childs et al. (2024) applied BASE-9, which naries from single stars, resulting in a higher detection
is also a multiband fitting procedure, with optical to limit of q. For instance, analyses based only on optical
near-infrared (NIR) data to derive the binary fraction photometry usually adopt mass ratio limits between 0.5
and mass-ratio distribution for six OCs. Their results and 0.7 (Jadhav et al. 2021; Cordoni et al. 2023; Don-
were notably accurate for binaries with q ≳ 0.4. ada et al. 2023). Combining optical and infrared data
can reduce this limit to q ∼ 0.2 to 0.4 (Thompson et al.
 3

2021; Malofeeva et al. 2022, 2023; Childs et al. 2024). and parallax) from Gaia DR3, we employed a two-
Nonetheless, some low mass-ratio binaries remain chal- component mixture model to determine the overall kine-
lenging to be identified definitively. matic parameters of the cluster, and simultaneously cal-
In this work, we aim to apply novel strategies to ad- culate the kinematic membership probability (Pk ) for
dress these two issues and investigate the stellar masses each sample star (Shao et al. 2024, in preparation). In
of Pleiades members, including the establishment of em- this process, the observational errors of proper motions
pirical photometric models, multiband fitting, and a and parallax are taken into account in order to obtain
study of unresolved main-sequence binaries in detail. the intrinsic dispersions in the space of these observables
The rest of the paper is organized as follows. In and the more rigorous statistical estimation of member-
Section 2, we introduce the observational data of the ship probabilities for individual stars. The astrometric
Pleiades, including the selection of kinematic members, data and Pk values of sample stars are listed in Table 1.
the Gaia and 2MASS photometric data, and the fun- Thanks to Gaia’s high-precision proper motions and par-
damental physical parameters of the cluster. In Sec- allaxes, a clear distinction between cluster members and
tion 3, we first introduce the method for correcting dis- field stars can be shown in the Pleiades region. There
crepancies between observational magnitudes and the are 97.6% sample stars having Pk > 0.9 or Pk < 0.1.
PARSEC model, presenting the empirical model of each Only a small fraction of objects remain ambiguous. The
adopted band for the Pleiades. Then, we describe the effectiveness index 1 of this kinematic membership deter-
method and the result of multiband fitting and present mination is Ek = 0.97, which is very close to a complete
the binary probability for each cluster member. In Sec- separation of members and field stars.
tion 4, we first describe the method for analyzing fb and A total of 1390 stars having Pk ≥ 0.5 were chosen for
q distribution by stacking probability density functions subsequent analysis, with an average Pk value of 0.978.
(PDFs) of all cluster members. Then, we discuss fb These members exhibit a high concentration in both co-
and q distribution in the Pleiades and their correlations ordinate and proper motion spaces, as shown in panels
with the primary mass. Finally, Section 5 summarizes (a) and (b) of Figure 1 respectively. Panel (c) demon-
this paper. strates that the parallaxes of the members also follow a
clustered distribution, distinctly different from that of
2. OBSERVATIONAL DATA AND PHOTOMETRIC field stars, which is in agreement with our expectations
MODEL OF THE PLEIADES CLUSTER for a cluster region (see Equations (2) and (3) in Shao
This section presents the observational data and pho- & Li (2019)).
tometric model used in this work, including kinematic Table 2 compares our membership sample with sev-
members (Section 2.1), photometric data (Section 2.2), eral previous works also based on the Gaia astrome-
the fundamental physical parameters of the Pleiades try (Lodieu et al. 2019; Cantat-Gaudin & Anders 2020;
and the corresponding fiducial photometric model (Sec- Hunt & Reffert 2023; Chulkov 2024). The comparison
tion 2.3). is limited to the same ranges of G-band magnitude, co-
ordination, and parallax. Overall, our sample overlaps
2.1. Sample and Kinematic Members well with the previous samples. Lodieu et al. (2019) and
Hunt & Reffert (2023) have obtained significantly more
The Pleiades cluster is centered on α = 56.601◦ ,
members than ours; however, all of these unmatched
δ = 24.114◦ , with a distance ranging from 130 to 140 pc
members exhibit large proper motion dispersions and
(Southworth et al. 2005; Melis et al. 2014) and a half-
have low membership probabilities in their clustering
number radius of ∼ 1.37◦ (approximately 3pc). Firstly,
algorithms.
sample stars were restricted to G < 19 mag and lo-
Usually, for distant clusters, the distance of a member
cated within a radius of 8.9◦ from the Pleiades cen-
star can be approximated by the value of the cluster, as
ter. For this flux-limited subset, ∼ 98.9% of Gaia DR3
the distance variations among members are negligible.
sources possess a 5-parameter astrometric solution. We
However, since the Pleiades is remarkably close to us,
further restricted the sample to parallaxes between 6.1
its member stars exhibit noticeable distance variations,
and 8.6 mas. This resulted in an initial sample of 6041
stars. Notably, this sample has not been constrained
1
by the renormalized unit weight error (RUWE) value,
P P P
The index E = 1 − N · i Pi (1 − Pi )/[ i Pi · i (1 − Pi )], where
because numerous unresolved binaries are among stars i = 1, ..., N for individual stars. E = 0 indicates that the two
components are totally mixed and cannot be separated anymore,
with RUWE>1.4. whereas E = 1 implies that the components are fully distinguish-
Subsequently, within the five-dimensional astromet- able (Shao & Zhao 1996).
ric data space (equatorial coordinates, proper motions,
4

Table 1. Observational Data and Kinematic Membership Probabilities of Stars in the Region of the Pleiades.

SourceId RA DEC µα µδ ϖ RUWE Pk DM

(deg) (deg) (mas/yr) (mas/yr) (mas) (mag)
(1) (2) (3) (4) (5) (6) (7) (8) (9)
68051390279853824 53.002 23.775 21.20 ± 0.02 -43.69 ± 0.02 7.28 ± 0.02 1.069 0.993 5.689
70461554129560448 57.052 26.584 20.98 ± 0.03 -46.56 ± 0.02 7.63 ± 0.02 1.029 0.997 5.589
71377000637253504 54.539 27.586 18.97 ± 0.09 -43.57 ± 0.07 6.92 ± 0.08 5.014 0.798 5.769
68587161680620288 55.496 25.329 20.10 ± 0.04 -44.18 ± 0.03 7.38 ± 0.03 2.026 1.000 5.660
64401389632616960 55.348 22.032 19.50 ± 0.09 -44.14 ± 0.07 7.35 ± 0.07 0.983 0.999 5.665

GBP G GRP J H Ks
(mag) (mag) (mag) (mag) (mag) (mag)
(10) (11) (12) (13) (14) (15)
14.099 ± 0.005 13.306 ± 0.003 12.435 ± 0.005 11.33 ± 0.02 10.69 ± 0.02 10.52 ± 0.02
13.949 ± 0.005 13.179 ± 0.003 12.313 ± 0.005 11.15 ± 0.02 10.49 ± 0.02 10.36 ± 0.02
14.154 ± 0.003 13.341 ± 0.003 12.430 ± 0.004 11.24 ± 0.02 10.63 ± 0.02 10.47 ± 0.02
14.373 ± 0.005 13.344 ± 0.003 12.326 ± 0.005 10.98 ± 0.02 10.30 ± 0.03 10.09 ± 0.02
18.864 ± 0.032 16.889 ± 0.003 15.564 ± 0.005 13.56 ± 0.02 12.86 ± 0.03 12.54 ± 0.03

Note—The first six columns list the Gaia DR3 source ID and astrometric data. The seventh column presents the RUWE which
is commonly larger for binaries than single stars. The eighth column presents the kinematic membership probability, derived
from the Gaia astrometric data. The ninth column shows the modified distance modulus (see Appendix A). Columns (10)-(15)
provide photometric data from Gaia DR3 and 2MASS. (The complete version of this table is available in a machine-readable
format.)

Figure 1. Distributions of equatorial coordinates (a), proper motions (b), and parallaxes (c) for stars in the Pleiades region.
The crosses denote field stars with kinematic membership probability Pk < 0.5. The gray points represent kinematic member
stars with Pk ≥ 0.5. The black circle in panel (a) shows the half-number radius R50 = 1.37◦ of the kinematic members.

corresponding to about 0.05 mag in distance modulus The photometric data used in this paper include GBP ,
(DM ), or about 3 pc in distance. Therefore, for the sub- G, GRP from Gaia DR3 (Gaia Collaboration et al.
sequent photometric fitting, we applied individual DM 2016, 2023; Babusiaux et al. 2023), and J, H, Ks from
for each star, derived from the modified Gaia parallax 2MASS (Skrutskie et al. 2006). This extensive cover-
as described in Appendix A. The modified DM values age from optical to NIR bands improves the detection of
are listed in column (9) of Table 1. low mass-ratio binaries compared to using optical bands
alone. For Gaia BP and RP bands, the spatial resolu-
tion is 3.5 by 2.1 arcsec (De Angeli et al. 2023), while for
2.2. Photometric Data
 5

Table 2. Comparison of Pleiades Members with Other Works.

NTW NOW Nmatch G Radius Parallax Source Reference

(mag) (deg) (mas)
365 389 362 < 15 <2 6.1 - 8.6 Gaia DR3 Chulkov (2024)
1123 1061 1027 < 18 <4 6.9 - 7.9 Gaia DR2 Cantat-Gaudin & Anders (2020)
1390 1616 1313 < 19 < 8.9 6.1 - 8.6 Gaia DR2 Lodieu et al. (2019)
1390 1563 1353 < 19 < 8.9 6.1 - 8.6 Gaia DR3 Hunt & Reffert (2023)

2MASS bands, the effective resolution is approximately outliers appearing as discrete dots in panel (c) are those
5 arcsec 2 . Therefore, the photometry of individual ob- affected by the blending of their neighboring sources
jects might be confused by their close neighbors. In in the 2MASS bands, which leads to redder G − Ks
order to eliminate this influence, we reject 153 cluster colors. Others may be contaminated field stars. The
members having nearby Gaia sources within 5 arcsec ra- different distribution patterns across these three CMDs
dius. Of remaining 1237 kinematic members, 1231 have demonstrate the necessity of combining optical and NIR
magnitudes for all six bands, representing 99.5% of the photometry to identify non-cluster or particular clus-
total. Six of the remainder have the Gaia photometric ter members, even though the kinematic membership
data alone. Detailed photometric data for each star are method is highly effective. These stars, including field
also provided in Table 1. stars, MS+WD binaries and multiple stellar systems,
Typical photometric errors are about 0.003∼0.005 will be subsequently excluded by the multiband fitting
mag for Gaia and 0.02∼0.03 mag for 2MASS. However, quality (see Section 3.2.3 for details). In this paper, we
the photometric uncertainties of Gaia are reported to focus only on MS+MS binaries of the Pleiades.
be approximately 0.01 mag (Riello et al. 2021), which
is significantly larger than the typical errors from the 2.3. Fundamental Parameters of the Pleiades and Its
catalog. Using unreasonably high-precision photometry Fiducial theoretical Model
for some individual bands can lead to local convergence In order to construct the empirical photometric model
or bias in the multiband fitting. Therefore, we set an for the Pleiades, we need a well-defined fiducial theoret-
error floor of 0.01 mag for Gaia magnitudes. Thus, for ical model depending on the fundamental physical pa-
each band of Gaia, the photometric error of each star is rameters of this cluster. In this work, we adopt the
2
adjusted to err = (errobs + 0.012 )1/2 . This adjustment PARSEC1.2s model (Bressan et al. 2012) due to its ex-
has an additional benefit of reducing the large difference tensive range of age and metallicity. The model’s mini-
between Gaia and 2MASS errors, thereby relatively in- mum stellar mass3 is 0.09M⊙ , enabling it to detect stars
creasing the weight of 2MASS data in the fitting process, down to such a low mass.
making it more balanced with the optical photometry. By using the Gaia photometric data, we applied the
Figure 2 illustrates the distribution of sample stars MiMO algorithm (Li & Shao 2022) to fit the cluster’s
on CMDs using Gaia, 2MASS, and one kind of combi- parameters. This algorithm constructs a mixture model
nation of them. These CMDs clearly reveal the main of field stars and cluster members (both singles and bi-
sequence of single stars and the binary distribution of naries) to match the observed stellar number density dis-
cluster members. Faint members exhibit greater scat- tribution on the CMD. Only stars brighter than G = 16
ter due to their larger errors. Some objects between the mag were included in the fitting, because stars fainter
isochrones of binary and triple stars may be the mul- than this magnitude may exhibit significant GBP or GRP
tiple stellar systems in the Pleiades, which is not the deviations from their true values (Riello et al. 2021), and
focus of this paper. Outlier members to the lower left of the model may also have considerable deviations in the
the main sequence in panel (a) may represent member low-mass range (Chen et al. 2019). In practice, we as-
binaries consisting of a main-sequence star and a white sumed that the distance modulus of the Pleiades is fixed
dwarf (MS+WD), which will be investigated in a sepa- at DMc = 5.65 mag (corresponding to the average mod-
rate study due to their unique research interest. Most ified parallax of kinematic members: ϖc = 7.41 mas),

2 https://irsa.ipac.caltech.edu/data/2MASS/docs/releases/ 3 The stellar mass used through in this paper is the initial mass,
allsky/doc/sec2 2a.html whereas the corresponding present mass can be inferred according
to the cluster’s age.
6

Figure 2. Distributions of stars in the Pleiades region on CMDs of Gaia (a), 2MASS (b), and combined optical and NIR bands
(c). The crosses denote stars with kinematic membership probability Pk < 0.5. The gray points show stars with Pk ≥ 0.5.
The star symbols mark the stars with neighbors in 5 arcsec. The error bars represent original observational errors. The
black solid and dashed lines represent isochrones for q = 0 and 1, respectively, from the PARSEC 1.2s model of [Fe/H]=0.1,
log(Age)/yr = 8.026 and AV = 0.135 mag. The dot-dash lines represent isochrons for triples with equal masses. The horizontal
line marks G = 16 mag. Note that three stars have been excluded from the plot due to invalid errors.

the metallicity is fixed at [Fe/H]=0.1 (Fu et al. 2022), cantly redder than the observation, leading to a large
and the binary mass ratio distribution follows a power amount of low-mass binaries may be misclassified as
law with exponent γq = 0. Based on these assump- single stars. Small discrepancies exist throughout the
tions, the best-fit values for the cluster age and extinc- rest of the higher mass range. They will also influence
tion are log(Age/yr) = 8.026 and AV = 0.135 mag from the determination of mass ratios to some extent. This
the MiMO. Then, we downloaded the model magnitudes mismatch highlights the necessity of correcting the dis-
with these fixed or fitted parameters from the website of crepancies between observational data and the fiducial
the PARSEC, and defined it as the fiducial photometric model, which will be discussed in the following section.
model for the Pleiades. It offers 476 sets of model mag-
3. MULTIBAND FITTING METHODOLOGY AND
nitudes as functions of stellar mass from 0.09 to 5.2 M⊙ .
RESULTS FOR THE PLEIADES MEMBERS
Among these, 112 sets in the range of 0.09 to 4.9 M⊙
were used in multiband fitting for stellar mass and mass Taking advantage of the large fraction of single stars in
ratio of cluster members (see Table 3 and Section 3.2.2). the cluster that perfectly conform to the main sequence
It should be noted that the choice of theoretical model of a SSP, we can quantify the discrepancy between the
and the accuracy of cluster parameters will not have a model and the observational data in terms of the stellar
substantial impact on the subsequent construction of the mass-magnitude relation. This allows us to correct the
empirical model. This is because the fiducial model ob- theoretical model and derive an empirical model for the
tained from the fitting will ultimately be adjusted to the cluster (Section 3.1). Subsequently, based on this em-
real observational data of the cluster, whereas minor dif- pirical model, we utilize the multiband fitting within a
ferences in models and parameters could be eliminated Bayesian framework to derive the posterior probability
during the adjustment. density distribution (PDF) of mass and mass ratio for
The fiducial model is represented by solid black lines single stars, or binaries (Section 3.2).
in Figure 2. It reveals that neither the optical nor 3.1. Correction of Photometric Model and
the NIR model magnitudes can perfectly match the Construction of Empirical Isochrone
data. Particularly for stars with masses below 0.5M⊙
3.1.1. Correction Strategy
(G ≳ 16 mag), the optical model photometry is signifi-
 7

In studies of photometric binaries in clusters, empiri- 3.1.2. Empirical Model of the Pleiades’ Photometry
cal models can be employed as a more practical substi- As shown in Figure 3, for stars with 0.6M⊙ < Ms <
tute for theoretical ones. For instance, in works based 1M⊙ , the magnitude corrections ∆m(Ms ) are relatively
on CMD, a typical procedure is to hold the magnitude small across all bands. While larger corrections are ob-
constant and adjust the model in the colors, to match served in both lower and higher mass ranges. For ex-
the observed main sequence of single stars, i.e. using ample, in the low-mass range, the maximum corrections
a m − ∆C relation to calibrate the model isochrone, are approximately 0.03 mag for G and GRP , 0.07 mag
where ∆C is the difference between the observed and for GBP and up to 0.5 mag for, J, H and Ks .
model colors (Fritzewski et al. 2019; Milone et al. 2012; Table 3 presents the original PARSEC model mag-
Li et al. 2020). nitudes alongside their corresponding correction val-
In the present work, we propose an optimized correc- ues, for stellar masses ranging from 0.09 to 4.9M⊙ .
tion method aiming to directly establish relationships The empirical model is then derived as Memp (Ms ) =
between the mass of single stars (Ms ) and magnitude Mmodel (Ms ) + ∆m(Ms ) for each band. It is worth
corrections (∆m) in various bands, rather than m − ∆C noting that this calibration is made-to-measure for the
relations. Subsequently, we adjust the theoretical model Pleiades cluster with its fundamental parameters. That
to an empirical model for stellar magnitudes, namely, a means the magnitude corrections not only depend on
set of Memp (Ms ) functions of corresponding bands. The stellar mass but also exhibit slight variations with the
steps involved are as follows. adopted metallicity, extinction, and age of the clus-
We first assume that all cluster members are single ter. Therefore, when applying these correction values
stars, and used the fiducial model for a multiband fitting to other clusters with different cluster parameters, mi-
of the stellar mass for each member. For the ith target, nor deviations should be anticipated.
we employed the standard χ2 fit to produce its best-fit
mass (Ms ) and corresponding model magnitudes in each 3.2. Multiband Fitting for Stellar Mass and Mass
band. Subsequently, we calculated the discrepancy be- Ratio of Cluster Members
tween the best-fit model and the observed magnitudes,
We employed the empirical model constructed from
defined as δm(i) = mobs (i) − Mmodel (i) − DMc . Then,
Gaia and 2MASS magnitudes to perform multiband fit-
δm of all stars with their stellar masses can be obtained
ting on members of the Pleiades. This enabled us to
as plotted in panels of Figure 3 for each band.
determine their stellar masses, including those of sin-
Ideally, if both the observational and model magni-
gle stars and the primary and secondary components
tudes were exactly correct, and each cluster member
of binaries. Our fitting procedure, conducted within a
were a single star, then all δm values would equal 0
Bayesian inference framework, introduced an optimized
mag. In the case of the theoretical model not aligning
form of the model magnitudes of an individual object
perfectly with the observational data, single stars will
in the cluster. The posterior PDF of fitting parame-
appear around a curve that deviates from the horizon-
ters was then derived for each member, allowing us to
tal line on the δm − Ms plane but remain highly con-
calculate their best estimates and uncertainties. Addi-
centrated. High mass-ratio binaries in the cluster and
tionally, we computed the probability of each member
some contaminated field stars, which do not fit well with
being a binary, denoted as Pb .
the single-star model, will distribute in a more scattered
region. 3.2.1. Model Magnitudes of Cluster Members and the
Finally, we employed a new robust regression method Lower Mass Limitation
based on the standard Gaussian Process (GP) and It-
Let us assume that each cluster member, which is an
erative Trimming (ITGP) algorithm (Li et al. 2021) to
unresolved object, possesses a companion, and define a
determine the ridgeline on the δm − Ms plane, where
photometric binary to be an object with detectable lu-
the main-sequence single stars should converge. ITGP is
minosity contribution from its secondary star. In this
specially designed to rule out outliers through an itera-
manner, we can adopt a unified stellar magnitude ex-
tive process, and has been successfully applied to model
pression of the object. For any given band, the absolute
the main sequence on CMDs of open clusters (Li et al.
magnitude in the empirical model can be expressed as
2020). The ridgelines, shown as solid lines in Figure 3
follows:
represent the ∆m(Ms ) functions for each band. They
were extrapolated to the lower mass limit of the PAR-

−0.4Memp (M1 )
−2.5 log[10
 + 10−0.4Memp (qM1 ) ]
SEC model (0.09M⊙ ) to obtain complete ones.

′
Memp = qM1 ≥ Mlim , (1)


qM1 < Mlim ,

Memp (M1 ),
8

Figure 3. δm − Ms planes and ridgelines of six photometric bands for the Pleiades. Ms represents the mass derived from the
single-star Multiband fitting. δm represents the magnitude difference between the observational data and the best-fit model in
each band. Each error bar corresponds to an individual star, with the length indicating the observational error. The solid black
line shows the ridgeline that represents the Ms − ∆m relation. The dashed black line marks the position where the magnitude
difference equals zero.

Table 3. The Fiducial Photometric Model and Its Correction ∆m(Ms ) of the Gaia and 2MASS Bands for the Pleiades.

Ms GBP G GRP J H Ks ∆GBP ∆G ∆GRP ∆J ∆H ∆Ks

(M⊙ ) (mag) (mag) (mag) (mag) (mag) (mag) (mag) (mag) (mag) (mag) (mag) (mag)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
0.0900 16.667 13.883 12.400 9.635 9.043 8.719 -0.680 -0.030 0.037 0.508 0.504 0.436
0.0924 16.620 13.835 12.352 9.587 8.995 8.671 -0.664 -0.030 0.036 0.503 0.497 0.433
0.1000 16.469 13.682 12.199 9.434 8.842 8.519 -0.616 -0.031 0.034 0.487 0.476 0.424
0.1080 16.328 13.539 12.056 9.292 8.698 8.376 -0.569 -0.031 0.032 0.471 0.455 0.416
0.1196 15.990 13.280 11.811 9.106 8.512 8.192 -0.505 -0.033 0.028 0.450 0.427 0.403
0.1200 15.973 13.269 11.801 9.100 8.507 8.187 -0.503 -0.033 0.028 0.449 0.426 0.403
0.1366 15.369 12.871 11.441 8.895 8.300 7.987 -0.409 -0.036 0.014 0.413 0.385 0.376
0.1400 15.258 12.795 11.372 8.854 8.259 7.947 -0.390 -0.037 0.012 0.404 0.377 0.369
0.1593 14.756 12.445 11.052 8.654 8.057 7.752 -0.289 -0.040 0.005 0.355 0.331 0.328
0.1600 14.738 12.433 11.040 8.647 8.050 7.745 -0.286 -0.040 0.004 0.353 0.329 0.326

Note—Columns (1)-(7) are the initial stellar mass (Ms ) and its corresponding magnitudes of six photometric bands from the
PARSEC1.2s with [Fe/H]=0.1, log(Age)/yr = 8.026 and AV = 0.135 mag. Columns (8)-(13) show the corresponding correction
values for each band. The empirical model magnitudes are expressed as Memp (Ms ) = Mmodel (Ms ) + ∆m(Ms ). The complete
version of this table is available in a machine-readable format.
 9

Appendix A and Table 2 for details). Then we have two

parameters, the primary mass M1 and the mass ratio q
need to be fit.
In the multiband fitting, a formal χ2 is written as:
X X (mobs,j − memp,j )2
χ2 = χ2j = , (3)
j j
errj2

where j denotes six observational bands, including GBP ,

G, GRP , J, H, and Ks , mobs,j represents observed mag-
nitude in each band, errj indicates the observational er-
ror involving error floors of 0.01mag for the Gaia data,
and memp,j refers to the empirical model magnitude ex-
pressed by Equations (1) and (2).
In the Bayesian Framework, we often transfer the χ2
to the likelihood function, which is a probability for ob-
taining the observational magnitudes at given M1 and
q and is defined as:
Figure 4. Lower limit of the detectable mass ratio (qlim ) L = exp(−χ2 /2) (4)
varying with primary mass. The blue line denotes the thresh-
old where the secondary mass is 0.09M⊙ . Stars with mass
It is certain that the maximum likelihood is correspond-
ratios above this line are binary stars, and those below are
considered “single stars”. ing to the least χ2 point. Besides, the advantage of us-
ing the Bayesian Inference is that it not only focuses on
the best-fit points, but also produces an entire posterior
where Memp (M1 ) and Memp (qM1 ) represent the abso-
PDF in the fitting-parameter space.
lute magnitudes of the primary and secondary stars from
For priors of fitting parameters, we set M1 follows a
the empirical model, respectively. Then the modeled ap-
log-uniform distribution ranging from 0.09 to 4.9M⊙ ,
parent magnitude of this object is
and q follows a uniform distribution between 0 and 1.
′
memp = Memp + DM (2) We employed the Nautilus (Lange 2023) package, which
uses an importance-sampling technique, to generate the
with DM to be its distance modulus. posterior PDF for each star in the parameter space.
Mlim denotes the lowest stellar mass in the model. During the sampling of M1 and q, the value of Memp
Our empirical model is modified from the PARSEC is linearly interpolated between model values listed in
1.2s model, which has a lower mass limit of 0.09M⊙ , Table 3.
slightly higher than the well-known stellar mass limit With the posterior PDF derived from the sampling
of ∼ 0.075M⊙ . Equation (1) implies an assumption process, we can further quantify the probability of each
that companion stars with masses below 0.09M⊙ are star being a binary system. Specifically, using the poste-
too faint to contribute a detectable luminosity incre- rior PDF of the ith object in the q − M1 plane, denoted
ment. So, they are considered, or defined as single stars as pi (M1 , q), we can determine the binary probability
in this work. of this target, Pb (i), as the proportion of the PDF that
The secondary mass threshold results in a lower limit lies above qlim (illustrated by the blue line in Figure 4):
on detectable binary mass ratios, which is a function of Z Z
the primary mass, qlim (M1 ) = Mlim /M1 , as shown by pi (M1 , q)dM1 dq
the blue line in Figure 4. This line separates the q − M1 Pb (i) = Z
M1 q>qlim
Z . (5)
plane into two distinct regions. Stars above this line are pi (M1 , q)dM1 dq
classified as binaries, whereas those below are considered M1 q
as single stars, regardless of their q values.
3.2.3. Fitting Result of the Pleiades
3.2.2. Bayesian Inference After removing 153 stars with neighbors in 5 arcsec,
For each cluster member, its age, metallicity, and ex- we performed multiband fitting on the remaining 1237
tinction are fixed at the cluster’s values, and the dis- kinematic members. The typical values of fitted param-
tance modulus is fixed at its own modified DM (see eters can be characterized by the best-fit values, where
10

Table 4. Results of Multiband Fitting for Stars with Pk ≥ 0.5

SourceId M1 q M1,16 M1,50 M1,84 q16 q50 q84 Pb flag of exclusion

(M⊙ ) (M⊙ ) (M⊙ ) (M⊙ )
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
68051390279853824 0.656 0.11 0.655 0.656 0.657 0.02 0.07 0.12 0.000 ...
70461554129560448 0.662 0.14 0.661 0.662 0.663 0.04 0.13 0.18 0.495 ...
71377000637253504 0.659 0.38 0.658 0.659 0.661 0.35 0.38 0.41 1.000 ...
68587161680620288 0.571 1.00 0.574 0.579 0.586 0.95 0.97 0.99 1.000 ...
64401389632616960 ... ... ... ... ... ... ... ... ... Ks
51761575759616000 ... ... ... ... ... ... ... ... ... GBP
66167411464494848 ... ... ... ... ... ... ... ... ... close neighbor

Note—The SourceId is the identifier provided by Gaia DR3. Columns (2) and (3) list the best-fit values of M1 and q. Columns
(4)-(9) provide the [16%, 50%, 84%] quantiles of these two parameters. The probability of a star being a binary (Pb ) is derived
from the probability density function (PDF). The “flag of exclusion” indicates the criteria used for excluding field stars and
non-main-sequence members. (The complete version of this table is available in a machine-readable format.)

the maximum posterior probability is located, or the with only 2-parameter solutions, which account for just
median value of each parameter from its corresponding 1% of the total sample.
marginal PDF. The uncertainties can be estimated as Figure 5 presents posterior PDFs for three represen-
half of the [16%, 84%] interval of the marginal PDFs. tative objects: a single star (panel (a)), a binary (panel
The fitting results show high precision, especially for (c)), and an intermediate case (panel (b)). The plus
M1 . For stars with Pb > 0.5, most of M1 uncertain- symbols indicate the best-fit points. In the single star
ties are smaller than 0.01M⊙ . The precision of M1 part (below the qlim in panels (a) and (b)), the PDF is
may be even higher for single stars. This is mainly due approximately parallel to the q coordinate. It indicates
to the high precision of photometric observations. For that there is no constraint on q, which is a natural con-
instance, the value of dM/dG is only about 0.15M⊙ sequence of the form of Equation (1). One should also
mag−1 for a star of one solar mass, while the typical notice that the best-fit value of q could be randomly as-
error of G is 0.01mag. So the precision of M1 will be signed for a single star, since we really do not know the
better than 0.01M⊙ for single stars. The joint usage of mass of the potentially existing “dark” component. In
six magnitudes will further improve the precision. Un- the binary parts of the PDFs (above the qlim in pan-
certainties of q for most stars with large Pb are smaller els (b) and (c)), a significant degeneracy can be found
than 0.1. between q and M1 , indicating that during the fitting
These fitting results, together with the binary proba- process, the reduction in the primary mass needs to be
bilities are listed in Table 4. For total 1237 fitted objects, compensated by a larger secondary star. Panel (b) is a
83 were further excluded due to their poor fits, charac- good example to show the advantage of the Bayesian in-
terized by χ2j > 25 in any band, indicating significant ference with the posterior PDF. If we only consider the
discrepancies between the empirical model and observa- best fit, this object would be classified as a binary. But
tion. Detailed exclusion reasons are also provided in Ta- in fact, it has approximately 40 percent to be a single
ble 4, column “flag of exclusion”. These excluded stars star, which is revealed by its PDF.
may be MS+WD binaries or multiple-star systems in Figure 6 shows the distribution of Pb for the final
the cluster, or possibly some contaminated background main-sequence sample, indicating that most members
stars. In brief, the multiband fitting can be regarded as are concentrated in the regions of Pb < 0.1 or Pb > 0.9.
an independent approach to diagnose the main-sequence The effectiveness index1 of Pb is E = 0.88, representing
cluster members, which can rule out those stars that do an excellent separation. This result demonstrates that
not satisfy the criteria for a single MS star or a MS+MS combining Gaia and 2MASS photometry can effectively
binary within the Pleiades. Then the remaining sample distinguish binaries from single stars in the Pleiades.
of 1154 main-sequence members will be analyzed in de- Figure 7 shows the distribution of the final 1154 clus-
tail for the binaries. Notably, this is not a complete sam- ter members on the CMD, with their Pb values color-
ple of multiple systems in the Pleiades, since it only con- coded. Most stars exhibit Pb close to either 0 or 1,
tains the MS+MS binaries, and does not include stars with only a few sources showing intermediate values.
 11

Figure 5. Probability density functions (PDFs) of three members. The contour plots in panels (a), (b), (c) illustrate the PDFs
of stars with different binary probabilities: Pb = 0, 0.6, 1, respectively. The SourceIds of Gaia DR3 are marked in each panel.
Each contour corresponds to the 1σ, 2σ, 3σ (39.3%, 86.5% and 98.9%) confidence levels. The blue lines denote the lower limit
of detectable mass ratios.
of -0.10. It indicates a systematic discrepancy between
the two techniques. Similar conditions have been re-
ported in other studies (Cohen et al. 2020; Thompson
et al. 2021; Childs et al. 2024).
Additionally, it is interesting to mention that, since we
have not used the RUWE value to constrain our kine-
matic member, our photometric diagnoses of binary re-
sults can be used in reverse to characterize the RUWE.
For 685 single stars (Pb < 0.1), all but two of them have
RUWE<1.4. While for 331 binaries (Pb > 0.9), 91 of
them having RUWE>1.4. That means, using RUWE
value as a criterion may only pick up one-third of bina-
ries.
Figure 6. Distribution of binary probability (Pb ) of main- 4. BINARY FRACTION AND MASS RATIO
sequence members.
DISTRIBUTION
In this section, we generate combined probability den-
However, a lot of intermediate Pb values are found for
sity functions (PDFs) for all sample stars using a “stack-
massive stars, which implies that, by using current pho-
ing” method (Section 4.1). These PDFs are then em-
tometric data, it is still difficult to definitively determine
ployed to analyze the binary fraction (Section 4.2) and
whether or not they have a low-mass companion.
its correlation with the primary mass (Section 4.3). We
We identified 17 common objects in our final sample
also investigate the mass-ratio distribution (Section 4.4)
and the dynamical mass-ratio subsample of Torres et al.
and its dependence on the primary mass (Section 4.5).
(2021). These stars are confirmed as photometric bi-
naries in our work with Pb ∼ 1. Figure 8 shows the 4.1. Combined PDF of Primary Mass and Mass Ratio
comparison of their q values, revealing a significant cor-
In general, the best-fit or median values of M1 and
relation between these two independent measurements.
q for each member star in Table 4 can be used for the
This validates the reliability of our method for binary
statistical analysis of binaries in the cluster. However,
detection and the accuracy of the mass-ratio measure-
from the Bayesian perspective, this method lacks rigor
ments. The binary with the lowest q value in our results
and may be susceptible to increased statistical fluctu-
shows the largest discrepancy in q between two q values,
ations, particularly when the number of a sample or
but this divergence falls within the estimated uncertain-
sub-sample is limited. To address this issue, we pro-
ties. Moreover, the q values obtained from our photo-
pose a novel approach. As shown in the left panel of
metric method are systematically smaller than those ac-
Figure 9, we stacked the normalized PDFs for all sam-
quired from double-line spectroscopy, with a mean offset
ple stars in the q − M1 plane to establish a combined
12

Figure 7. Distributions of binary probability (Pb ) for Pleiades kinematic members on CMDs of Gaia, 2MASS, and the combined
data. The star symbols mark the stars with neighbors in 5 arcsec. The gray crosses show the stars excluded after multiband
fitting. The colored points represent the final 1154 members, with color indicating Pb . All photometric magnitudes have been
adjusted to absolute magnitudes using the DM in Table 1. The dotted lines mark Ms = 0.5, 1 and 2M⊙ from bottom to top.
The black solid and dashed lines represent the empirical isochrones for q = 0 and 1, respectively. The dot-dash lines represent
isochrons for triple stars with equal masses.

that of individual stars, pi . The P (M1 , q) satisfies the

following equation:
Z Z
P (M1 , q)dM1 dq = N, (6)
M1 q

where N = 1154 represents the number of photometric

members in our final sample. The P (M1 , q) essentially
acts as a complete “mapping” from the photometric data
to the stellar mass distribution of the cluster, which en-
compasses all information about the single, primary, and
secondary stars.
In our sample, the lower luminosity limit is G = 19
mag, corresponding to a single star mass of 0.11M⊙ .
Consequently, the following binary analysis is limited to
M1 > 0.11M⊙ , which is to the right of the blue solid
vertical line in Figure 9. The total number of stars in
this region, derived from integrating the PDF, is n =
Figure 8. Comparison of mass ratio from this work with 1145.2. Note that this value is a fractional number from
double-line binaries (Torres et al. 2021). All these stars have the integration rather than a direct counting. We use n
Pb ≈ 1. The black line indicates y=x, and the gray line to distinguish it from the simple count of stars, N .
shows an offset of -0.10. When we use P (M1 , q) to investigate the binary frac-
tion and its relation with primary mass and mass ratio,
PDF of the main-sequence stars in the cluster, denoted it is flexible to calculate various statistical parameters
as P (M1 , q), where the capital P represents probabil- by using integration instead of counting in all cases. For
ity density for the whole sample, distinguishing it from example, the total number of binaries in our sample with
 13

Figure 9. Combined probability density function (PDF) of primary mass and mass ratio for our sample. The left panel shows
the direct result, with grayscale representing probability density. The dashed blue line indicates the position of the binary star
with a secondary mass of 0.09 M⊙ . Binaries are located above this line, and single stars are below it. The solid blue line marks
M1 = 0.11. The stars to the left of this line are incomplete and are thus excluded from the statistical analysis. The right panel
presents a smoothed version of the left panel, displaying only the region for binary analysis, with color indicating probability
density.

primary masses higher than 0.11M⊙ is Numerous studies have investigated fb of the Pleiades,
Z Z with the range of mass and mass ratio varying depending
nb = P (M1 , q)dM1 dq. (7) on observations and methodologies. To compare with
M1 >0.11M⊙ q≥qlim previous studies, we calculate fb separately within the
Although nb is the result of integration, it is still sub- same mass and mass-ratio ranges of them. These re-
ject to the “law of small numbers”, so its uncertainty sults are detailed in Table 5 and visually compared in
1/2
can be roughly estimated by nb based on the Poisson Figure 10. Depending on the type of observational data
fluctuation. Thus, the total binary fraction of our sam- used, these studies can be divided into three categories.
ple is fb = nb /n, with a corresponding uncertainty of The first one is the spectroscopic binaries by using ra-
1/2 dial velocity variations (Torres et al. 2021). They found
nb /n. Similarly, for other subsamples, such as when
fb = 0.25 ± 0.03 for periods shorter than 10000 days,
restricting the range of M1 or q, we can calculate the
in a G magnitude range of 3.7 − 15 mag, corresponding
corresponding number of binaries, binary fraction, and
to a mass range of approximately 0.5 − 4.9M⊙ for the
their associated errors. Consequently, it will be easy to
Pleiades. This fraction is notably smaller than ours (see
obtain nb or fb as functions of M1 and q.
the blue point in Figure 10), which can be explained by
4.2. Binary Fraction the fact that the ∆RV method is less effective in de-
tecting long-period, low mass-ratio binaries, or binaries
This work covers a primary mass range from 0.11M⊙
with high orbital inclinations (face-on).
to 4.9M⊙ for Pleiades members, representing the broad-
The second one is the photometric binaries by using
est mass range analyzed for binaries in this cluster to
optical data only. It includes several methods: sepa-
date. By using Equation (7), we determine the total
rating regions of single and binary stars on CMDs with
binary fraction of our sample fb = 0.34 ± 0.02.
14

Table 5. Comparison of Binary Fraction with Other Works.

This Work Previously Published Works

fb fb Mass range (M⊙ ) qmin Data Reference
0.53 ± 0.04 0.25 ± 0.03 0.5 - 4.9 - RV Torres et al. (2021)
0.12 ± 0.02 0.14 ± 0.02 0.4 - 3.6 0.6 GBP , G, GRP Jadhav et al. (2021)
0.12 ± 0.02 0.059 ± 0.040 0.3 - 0.9 0.6 GBP , G, GRP Cordoni et al. (2023)
0.11 ± 0.01 0.078 ± 0.009 0.180 - 2.295 0.6 GBP , G, GRP Donada et al. (2023)
0.22 ± 0.04 0.20 ± 0.03 0.6 - 1.0 0.5 GBP , G, GRP Niu et al. (2020)
0.54 ± 0.04 0.41 ± 0.04 0.57 - 3.75 - GBP , G, GRP Niu et al. (2020)
0.34 ± 0.02 > 0.19 All (?) - GBP , G, GRP Almeida et al. (2023)
0.50 ± 0.04 0.70 ± 0.14 0.5 - 1.8 - GBP , H, Ks , W1 , W2 Malofeeva et al. (2022)
0.50 ± 0.04 0.73 ± 0.03 0.5 - 1.8 - GBP , H, Ks , W1 , W2 Malofeeva et al. (2023)

Note—Row one compares our results with studies based on radial velocity variations using spectroscopic observations. Rows
two through seven compare our findings with studies based on Gaia optical photometry data. Rows eight and nine compare
our results with studies combining Gaia’s optical data with 2MASS and WISE infrared photometry data on a pseudo-colors
diagram. The (-) denotes no restriction on the mass ratio range; hence, we use the total binary fraction for comparison.

(see green points in Figure 10). However, total binary

fractions, including lower mass-ratio binaries, are signif-
icantly lower than ours (see orange points in Figure 10).
This discrepancy supports our claim that using optical
photometric data alone may miss a substantial number
of low mass-ratio binaries, which will be more detectable
in NIR bands.
At last, Malofeeva et al. (2022) and Malofeeva et al.
(2023) combined the Gaia DR2, 2MASS, and WISE data
and used the pseudo CMD of (H − W2 ) − W1 vs W2 −
(GBP − Ks ) to analyze binaries in the Pleiades. They
divided the diagram into distinct mass ratio regions, and
reported fb = 0.70 ± 0.14 and 0.73 ± 0.03, respectively,
for the M1 range of 0.5 − 1.8M⊙ . These results are
higher than ours, fb = 0.50 ± 0.04 (see red points in
Figure 10). We propose that this discrepancy can be
attributed to two factors: the inclusion of multiple-star
systems in their sample and the contamination of single
Figure 10. Comparison of binary fraction from this work stars in the q = 0 − 0.2 interval.
with other works. The blue point indicates the comparison These comparisons demonstrate that the binary frac-
with a result using spectroscopic observations. Orange and tion in the Pleiades cluster varies significantly with the
green points represent comparisons with results using Gaia stellar mass and mass ratio ranges. Our work can re-
photometric data, with green points specifically showing high produce or explain almost all previous results. This
mass-ratio binary fraction. Red points show comparisons
suggests that detection in the complete range of mass
with results that combined Gaia, 2MASS, and WISE photo-
metric data. The black line indicates y=x. and mass ratios is extremely important in the analysis
of binaries in clusters.
model lines (Jadhav et al. 2021; Cordoni et al. 2023),
constructing a mixture model of single and binaries on 4.3. Binary Fraction as a Function of Primary Mass
CMDs (Niu et al. 2020; Donada et al. 2023), or employ- It is well established that fb correlates with primary
ing a kind of multiband fitting (Almeida et al. 2023). mass (Li et al. 2020; Offner et al. 2023), indicating that
Our comparison reveals that the fractions of high mass- stars with higher M1 are more likely to form and sustain
ratio binaries (q > 0.6 or 0.5) are consistent with ours binary systems. To explore this relation in the Pleiades,
we divided the primary mass into seven segments: [0.11,
 15

Table 6. Binary Fraction as a Function of Primary Mass for Different Mass Ratios in the Pleiades

M1 (M⊙ ) n fb fb0−0.3 fb0.3−0.8 fb0.8−1

0.11 - 4.90 1145.2 0.34 ± 0.02 0.09 ± 0.01 0.17 ± 0.01 0.07 ± 0.01
0.11 - 0.18 178.7 0.18 ± 0.03 - 0.08 ± 0.02 0.10 ± 0.02
0.18 - 0.30 345.7 0.23 ± 0.03 - 0.16 ± 0.02 0.07 ± 0.01
0.30 - 0.50 245.7 0.31 ± 0.04 0.10 ± 0.02 0.15 ± 0.02 0.06 ± 0.02
0.50 - 0.90 211.6 0.45 ± 0.05 0.15 ± 0.03 0.25 ± 0.03 0.05 ± 0.02
0.90 - 1.60 119.1 0.55 ± 0.07 0.24 ± 0.05 0.25 ± 0.05 0.06 ± 0.02
1.60 - 3.00 34.8 0.78 ± 0.15 0.51 ± 0.12 0.21 ± 0.08 0.06 ± 0.04
3.00 - 4.90 9.5 0.94 ± 0.31 0.32 ± 0.18 0.48 ± 0.23 0.13 ± 0.12

Note—n denotes the number of stars within the specified mass range. fb represents binary fraction for photometric detectable
binaries with q > qlim (M1 ). The (-) denotes a range where photometric binaries cannot be detected.

Figure 11. Binary fraction (fb ) as a function of primary mass for different mass ratios in the Pleiades. Panel (a) shows the
total binary fraction for q > qlim . Panels (b-d) represent the binary fractions for q = 0 − 0.3, 0.3 − 0.8, 0.8 − 1, respectively.
Dashed lines indicate incomplete binary fractions. The shaded regions represent fb errors.

0.18, 0.3, 0.5, 0.9, 1.6, 3.0, 4.9] M⊙ , roughly following fb remains. While for the high mass ratio, as shown in
a log-uniform dividing on M1 . panel (d), fb exhibits minimal variation with mass and
The third column of Table 6 and panel (a) of Fig- lacks a significant trend. This behavior for high mass-
ure 11 show the variation of the total binary fraction ratio binaries is consistent with the result reported by
(fb for q > qlim ) with primary mass. The fb exhibits a Cordoni et al. (2023), who found no significant correla-
monotonic increase with M1 , ranging from 0.18 in the tion between fb and primary mass of q > 0.6 binaries
lowest mass bin to 0.94 within the highest mass bin. across 72 OCs, including the Pleiades.
This trend confirms the previous conclusions that more In general, the variation of fb with primary mass can
massive stars are more likely to possess a companion. be attributed to dynamical effects within the cluster (Li
Meanwhile, our ability to detect very low mass-ratio et al. 2020). Since the binary binding energy Eb ∝ qM21 ,
companions of massive stars enhances this relationship. binaries with lower primary masses and lower mass ra-
For example, we are able to identify binaries down to tios are more easily disrupted during three-body encoun-
q < 0.03, when M1 > 3.0M⊙ . ters, leading to a decrease in fb with decreasing M1 in
Furthermore, we divided the mass ratio into three the lower mass-ratio ranges.
ranges: q = 0 − 0.3, 0.3 − 0.8, and 0.8 − 1.0 for the Alternatively, we propose another explanation for the
low, medium, and high mass ratios, and calculated fb in M1 − fb relation of photometric binaries. As shown in
these q ranges with different primary masses. They are Figure 9, the qlim value is larger for lower-mass stars,
also listed in corresponding columns of Table 6. For low meaning that many companions are too faint to be de-
and medium mass ratios, as shown in panels (b) and (c) tected photometrically. In contrast, for higher-mass
of Figure 11, the positive correlation between M1 and stars, a broader range of detectable mass ratios allows
16

for the identification of more binaries. As observed in three segments, ranging from low to high mass ratios,
the Pleiades, many low mass-ratio binaries have a mas- are as follows: γ1 = 0.8 ± 0.2, γ2 = −1.0 ± 0.3, and
sive primary star (panel (b) of Figure 11). γ3 = 3.1 ± 1.0. This modeling identifies two breakpoints
In summary, the variation in fb with primary mass is of mass ratio: qb1 = 0.32 ± 0.04, qb2 = 0.79 ± 0.05. This
influenced by both observational and dynamical effects. three-segment profile of the Pleiades resembles the mass
In the low mass-ratio range, the increase in fb with pri- ratio distribution model proposed by Moe & Di Stefano
mary mass is mainly driven by the observational limit (2017) (See Figure 2 of their paper). The difference is
of qlim . While in the medium mass-ratio range of q = that their model shows a significant excess of twin bina-
0.3 − 0.8, where the mass range of M1 = 0.3 − 4.9M⊙ ries, while the Pleiades appears a progressive excess of
is unaffected by qlim , the rise in fb with primary mass high mass-ratio binaries. If we extrapolate γ2 to the high
should be mainly influenced by dynamical effects. mass-ratio range, then the number of exceeding binaries
is estimated to be nexc
b ≈ 34.2 ± 5.8.
4.4. Binary Mass Ratio Distribution Several studies have investigated the mass ratio distri-
The mass ratio distribution function of binaries, nb (q), bution of the Pleiades, mostly focused on binaries with
is a crucial statistic in star clusters. Similar to the stellar M1 ≳ 0.5M⊙ , with varying upper mass limits. Fig-
mass function, nb (q) offers valuable insights into both ure 13 compares our results with others. The light gray
the mechanism of star formation and the dynamical pro- histograms represent the mass ratio distribution for all
cesses within a cluster. binaries as the same as in Figure 12, while the dark gray
ones are constrained to the same mass ranges of the com-
pared works. The solid line in panel (a) shows a nearly
flat mass ratio distribution obtained by Torres et al.
(2021) from the ∆RV method. Compared to our result,
there is a significant reduction in the number of low
mass-ratio binaries, likely due to the limited sensitivity
of the spectroscopic in this mass-ratio range. The solid
line in panel (b) shows the result from Niu et al. (2020),
who applied a mixture-model method on the CMD with
Gaia photometry. They employed a power-law profile
but had an index of 0.22, indicating a roughly flat dis-
tribution. Their result also shows fewer low mass-ratio
binaries than ours. This difference may arise from two
factors. One is the use of optical data alone, which may
inadequately distinguish low mass-ratio binaries from
Figure 12. Mass ratio distribution and best-fit model for single stars. Another is the absence of isochrone correc-
the Pleiades. The gray histogram and error bars show the
tion, potentially introducing bias in the measurement
mass ratio distribution. The black line represents the best-
fit model. The model is divided into three segments. Each of binary mass ratios. The solid and dashed lines in
segment follows a power law. The dark gray area represents panel (c) represent results from Malofeeva et al. (2022)
the number of exceeding high mass-ratio binaries. and Malofeeva et al. (2023), who used a pseudo-color di-
agram combining optical to infrared photometric data.
Their mass ratio distribution reveals a higher proportion
Our measurement of the secondary mass reaches the
of low mass-ratio binaries compared to high mass-ratio
lower limit of the PARSEC model at 0.09M⊙ , enabling
ones. This phenomenon is consistent with our result.
us to obtain a board mass-ratio distribution above qlim .
However, they have identified more binaries than ours,
The gray histogram in Figure 12 shows the nb (q) for
as we have already mentioned and discussed in Figure 10
all photometric binaries in our sample, with the pri-
of Section 4.2.
mary mass ranging from 0.11 to 4.9M⊙ . The pro-
In conclusion, the mass-ratio distribution of photo-
file of nb (q) can be divided into three segments. It
metric binary stars in the Pleiades shows significant
increases in the low mass-ratio range (q ≲ 0.3), sub-
complexity. Therefore, using a single power law is in-
sequently decreases in the medium mass-ratio range
adequate to accurately characterize this distribution.
(0.3 ≲ q ≲ 0.8) and then increases again in the high
mass-ratio range (q ≳ 0.8). We modeled this distribu-
tion using a three-segment power-law profile, uniformly 4.5. Mass Ratio Distribution Varying with Primary
expressed as nb (q) ∝ q γ . The power-law exponents for Mass
 17

Figure 13. Comparison of mass ratio distribution from this work with other works. The light gray histograms represent the
total mass ratio distribution across all mass ranges, whereas the dark gray ones represent the distributions within specific mass
ranges for comparison. The solid line in panel (a) corresponds to the mass ratio distribution from Torres et al. (2021). The
solid line in panel (b) reflects the distribution from Niu et al. (2020), which follows a power-law distribution with γ = 0.22. The
solid and dashed lines in panel (c) denote the mass ratio distributions reported in Malofeeva et al. (2022) and Malofeeva et al.
(2023), respectively.

Figure 14. Mass ratio distributions in different mass ranges. Panels (a-f) show distributions in different mass ranges: M1 =
0.11 − 0.18, 0.18 − 0.3, 0.3 − 0.5, 0.5 − 0.9, 0.9 − 1.6, 1.6 − 4.9M⊙ . Within each panel, the dotted lines indicate the peak and
valley of the total mass ratio distribution.

By analyzing subsamples with various primary observed. First, nb (q) exhibits significant variations
masses, we aim to conduct a more detailed investigation across different mass ranges, and most of them do not
into the mass ratio distribution of binaries and its un- follow a monotonic trend. Second, as the primary mass
derlying causes. The mass binning follows the scheme increases, the position of qb1 tends to shift towards lower
presented in Section 4.3, except that the two bins of mass ratios. This strongly suggests that, at q < 0.8, the
M1 > 1.6M⊙ are combined. The segmentation points mass ratio of real binary systems, whether or not hav-
of M1 are [0.11, 0.18, 0.3, 0.5, 0.9, 1.6, 4.9] M⊙ . ing an illuminated companion, might be monotonically
Figure 14 presents the mass ratio distribution within decreasing. The observed break of photometric binaries
different primary mass ranges. Several key features are (qb1 ) attributed to the selection effects caused by the de-
18

tection limit, qlim (M1 ). Third, at lower M1 , the num- 1. By utilizing the concentrated distribution of single
ber of binaries with high-q exceeds that of binaries with stars along the main sequence, we have adjusted
medium-q, suggesting that high-q binaries are predomi- the theoretical model to match the photometric
nantly composed of low-mass stars with M1 < 0.5M⊙ . data, resulting in an empirical photometric model.
This pattern is also evident in the two-dimensional PDF This adjustment significantly reduces biases in bi-
shown in panel (b) of Figure 9. Specifically, the prob- nary mass-ratio measurements.
ability density decreases as q moves away from qlim , 2. The combined use of optical and near-infrared
reflecting the monotonic decrease in nb (q) for bina- photometric data covers the primary emission
ries with low and medium mass ratios. Additionally, a bands of stars with a wide mass range. This sig-
prominent high-density region is observed in the upper nificantly enhances the precision of mass fitting,
left of this panel, corresponding to an excess of low-M1 particularly for low-mass stars, which is crucial
and high-q binaries. for the identification and accurate quantification
The mass-ratio distribution of binaries is shaped by of low mass-ratio binaries.
various factors, and the complex profile observed in the 3. The Bayesian framework allows us to rigorously
Pleiades cluster reflects the combined impact. Besides quantify stellar masses and mass ratios using prob-
the selection effect imposed by the qlim (M1 ), the main abilistic methods, enabling our detection down
physical reasons lie in the star formation mechanisms to the theoretical lower mass limit. By stacking
and the dynamical evolution processes. The value of the probability density functions (PDFs) of indi-
γ2 = −1.0 is notably related to the stellar mass func- vidual members in a cluster, we can extend the
tion. However, this exponent index implies the absence Bayesian approach to the analysis of binary prop-
of low mass-ratio binaries compared to random pairings erties across the entire sample.
from the initial mass function (αMF ∼ −2.35). It may By applying this method to the Pleiades cluster, we
be caused by the dynamics that result in low-q binaries detected and measured all photometric main-sequence
more susceptible to disruptions in three-body encoun- binaries with mass ratios above a well-defined limit,
ters. Additionally, the excess of high-q binaries can be qlim (M1 ). The detailed results of these binary prop-
explained by two factors. First, during the formation of erties are listed below:
primordial binaries, the accretion rate of the secondary 1. The binary fraction of main-sequence members in
star is typically higher than that of the primary star the Pleiades cluster is 0.34 ± 0.02.
in the competitive accretion mechanism within binary 2. Observations show that the binary fraction in-
discs (Bate 2000; Farris et al. 2014; Matsumoto et al. creases with increasing primary mass. This trend
2019). This leads to the secondary star having a mass is particularly evident in binaries with low and in-
comparable to the primary star’s. Second, during sub- termediate mass ratios.
sequent dynamical interaction in the cluster, the lower- 3. The complete mass ratio distribution follows a
mass companions are more likely to be replaced by a three-segment power-law profile, with power-law
higher-mass single star, leading to a higher mass-ratio exponents (γ) to be 0.8 ± 0.2, −1.0 ± 0.3, and
binary. This effect is particularly pronounced in binaries 3.1 ± 1.0 for lower, median and higher mass-ratio
with low-mass primaries due to their lower binding ener- ranges that separated by two break points at q ∼
gies and the abundance of single stars with comparable 0.3 and 0.8. This pattern indicates a generally
masses available for exchange. decreasing trend, with a significant deficiency of
binaries in the low mass-ratio range and a notable
5. CONCLUSION excess in the high mass-ratio range.
4. By analyzing subsamples with different primary
We present an optimized multiband magnitude fit-
masses, we find that the proportion of high mass-
ting approach for determining the stellar mass of main-
ratio binaries decreases as the primary mass in-
sequence single stars or binary components in a cluster.
creases. This indicates that these photometric
Employing photometric data from Gaia and 2MASS,
high mass-ratio binaries are mainly contributed by
we have measured the stellar mass of members in the
binaries with lower primary masses.
Pleiades open cluster and performed a comprehensive
5. As a by-product, we found that, for the current
analysis of its main-sequence binaries, including the bi-
sample from Gaia DR3, almost all single stars have
nary fraction, mass-ratio distribution, as well as their
RUWE<1.4, while only about 1/3 of binaries are
relationships with the primary mass.
larger than this value.
The improvements and advantages of our fitting
method are as follows:
 19

These results provide a comprehensive overview of the Program of Shanghai Academic/Technology Re-
the statistical properties of photometric binaries in the search Leader. This work has made use of data
Pleiades cluster, revealing complex patterns in the dis- from the European Space Agency (ESA) mission Gaia
tribution of binary fraction and mass ratio. The com- (https://www.cosmos.esa.int/gaia), processed by the
plexity can be attributed to multiple factors, including Gaia Data Processing and Analysis Consortium (DPAC;
detection biases for low-mass binaries, binary formation https://www.cosmos.esa.int/web/gaia/dpac/consortium).
mechanisms in the star-forming stage, and the subse- Funding for the DPAC has been provided by national
quent dynamical interactions within the cluster. institutions, in particular the institutions participating
in the Gaia Multilateral Agreement. This work made
ACKNOWLEDGMENTS use of data products from the Two Micron All Sky
We sincerely thank the anonymous referee for valu- Survey, which is a joint project of the University of
able comments and suggestions. This work was Massachusetts and the Infrared Processing and Anal-
supported by the National Natural Science Founda- ysis Center/California Institute of Technology, funded
tion of China (NSFC) under grants 12273091 and by the National Aeronautics and Space Administration
U2031139; the National Key R&D Program of China and the National Science Foundation.
No. 2019YFA0405501; the science research grants from Software: PARSEC (Bressan et al. 2012), astropy
the China Manned Space Project with No. CMS-CSST- (Astropy Collaboration et al. 2013, 2018), Nautilus
2021-A08; the Science and Technology Commission (Lange 2023)
of Shanghai Municipality (Grant No. 22dz1202400);

REFERENCES
Almeida, A., Monteiro, H., & Dias, W. S. 2023, MNRAS, Cordoni, G., Milone, A. P., Marino, A. F., et al. 2023,
525, 2315, doi: 10.1093/mnras/stad2291 A&A, 672, A29, doi: 10.1051/0004-6361/202245457
Astropy Collaboration, Robitaille, T. P., Tollerud, E. J., De Angeli, F., Weiler, M., Montegriffo, P., et al. 2023,
et al. 2013, A&A, 558, A33, A&A, 674, A2, doi: 10.1051/0004-6361/202243680
doi: 10.1051/0004-6361/201322068 Donada, J., Anders, F., Jordi, C., et al. 2023, A&A, 675,
Astropy Collaboration, Price-Whelan, A. M., Sipőcz, B. M., A89, doi: 10.1051/0004-6361/202245219
et al. 2018, AJ, 156, 123, doi: 10.3847/1538-3881/aabc4f Evans, D. W., Riello, M., De Angeli, F., et al. 2018, A&A,
Babusiaux, C., Fabricius, C., Khanna, S., et al. 2023, A&A, 616, A4, doi: 10.1051/0004-6361/201832756
674, A32, doi: 10.1051/0004-6361/202243790 Farris, B. D., Duffell, P., MacFadyen, A. I., & Haiman, Z.
Bate, M. R. 2000, MNRAS, 314, 33, 2014, ApJ, 783, 134, doi: 10.1088/0004-637X/783/2/134
doi: 10.1046/j.1365-8711.2000.03333.x Fritzewski, D. J., Barnes, S. A., James, D. J., et al. 2019,
Bouvier, J., Rigaut, F., & Nadeau, D. 1997, A&A, 323, 139 A&A, 622, A110, doi: 10.1051/0004-6361/201833587
Bressan, A., Marigo, P., Girardi, L., et al. 2012, MNRAS, Fu, X., Bressan, A., Marigo, P., et al. 2018, MNRAS, 476,
427, 127, doi: 10.1111/j.1365-2966.2012.21948.x 496, doi: 10.1093/mnras/sty235
Cantat-Gaudin, T., & Anders, F. 2020, A&A, 633, A99, —. 2016, IAU Focus Meeting, 29B, 144,
doi: 10.1051/0004-6361/201936691 doi: 10.1017/S1743921316004634
Chen, Y., Bressan, A., Girardi, L., et al. 2015, MNRAS, Fu, X., Bragaglia, A., Liu, C., et al. 2022, A&A, 668, A4,
452, 1068, doi: 10.1093/mnras/stv1281 doi: 10.1051/0004-6361/202243590
Chen, Y., Girardi, L., Bressan, A., et al. 2014, MNRAS, Gaia Collaboration, Prusti, T., de Bruijne, J. H. J., et al.
444, 2525, doi: 10.1093/mnras/stu1605 2016, A&A, 595, A1, doi: 10.1051/0004-6361/201629272
Chen, Y., Girardi, L., Fu, X., et al. 2019, A&A, 632, A105, Gaia Collaboration, Vallenari, A., Brown, A. G. A., et al.
doi: 10.1051/0004-6361/201936612 2023, A&A, 674, A1, doi: 10.1051/0004-6361/202243940
Childs, A. C., Geller, A. M., von Hippel, T., Motherway, Geller, A. M., de Grijs, R., Li, C., & Hurley, J. R. 2013a,
E., & Zwicker, C. 2024, ApJ, 962, 41, ApJ, 779, 30, doi: 10.1088/0004-637X/779/1/30
doi: 10.3847/1538-4357/ad18c0 —. 2015, ApJ, 805, 11, doi: 10.1088/0004-637X/805/1/11
Chulkov, D. 2024, AJ, 168, 156, Geller, A. M., Hurley, J. R., & Mathieu, R. D. 2013b, AJ,
doi: 10.3847/1538-3881/ad7025 145, 8, doi: 10.1088/0004-6256/145/1/8
Cohen, R. E., Geller, A. M., & von Hippel, T. 2020, AJ, Heggie, D. C. 1975, MNRAS, 173, 729,
159, 11, doi: 10.3847/1538-3881/ab59d7 doi: 10.1093/mnras/173.3.729
20

Hunt, E. L., & Reffert, S. 2023, A&A, 673, A114, Moe, M., & Di Stefano, R. 2017, ApJS, 230, 15,
doi: 10.1051/0004-6361/202346285 doi: 10.3847/1538-4365/aa6fb6
Jadhav, V. V., Roy, K., Joshi, N., & Subramaniam, A. Montegriffo, P., De Angeli, F., Andrae, R., et al. 2023,
2021, AJ, 162, 264, doi: 10.3847/1538-3881/ac2571 A&A, 674, A3, doi: 10.1051/0004-6361/202243880
Lada, C. J., & Lada, E. A. 2003, ARA&A, 41, 57, Niu, H., Wang, J., & Fu, J. 2020, ApJ, 903, 93,
doi: 10.1146/annurev.astro.41.011802.094844 doi: 10.3847/1538-4357/abb8d6
Lange, J. U. 2023, Monthly Notices of the Royal Offner, S. S. R., Moe, M., Kratter, K. M., et al. 2023, in
Astronomical Society, 525, 3181, Astronomical Society of the Pacific Conference Series,
doi: 10.1093/mnras/stad2441 Vol. 534, Protostars and Planets VII, ed. S. Inutsuka,
Y. Aikawa, T. Muto, K. Tomida, & M. Tamura, 275,
Li, L., & Shao, Z. 2022, ApJ, 930, 44,
doi: 10.48550/arXiv.2203.10066
doi: 10.3847/1538-4357/ac5f4f
Pinfield, D. J., Dobbie, P. D., Jameson, R. F., et al. 2003,
Li, L., Shao, Z., Li, Z.-Z., et al. 2020, ApJ, 901, 49,
MNRAS, 342, 1241,
doi: 10.3847/1538-4357/abaef3
doi: 10.1046/j.1365-8711.2003.06630.x
Li, Z.-Z., Li, L., & Shao, Z. 2021, Astronomy and
Riello, M., De Angeli, F., Evans, D. W., et al. 2021, A&A,
Computing, 36, 100483, doi: 10.1016/j.ascom.2021.100483
649, A3, doi: 10.1051/0004-6361/202039587
Lodieu, N., Dobbie, P. D., Deacon, N. R., et al. 2007,
Shao, Z., & Li, L. 2019, MNRAS, 489, 3093,
MNRAS, 380, 712, doi: 10.1111/j.1365-2966.2007.12106.x
doi: 10.1093/mnras/stz2317
Lodieu, N., Pérez-Garrido, A., Smart, R. L., & Silvotti, R.
Shao, Z., & Zhao, J. 1996, Acta Astronomica Sinica, 37, 377
2019, A&A, 628, A66, doi: 10.1051/0004-6361/201935533
Skrutskie, M. F., Cutri, R. M., Stiening, R., et al. 2006, AJ,
Malofeeva, A. A., Mikhnevich, V. O., Carraro, G., & 131, 1163, doi: 10.1086/498708
Seleznev, A. F. 2023, AJ, 165, 45, Smith, B., & Struve, O. 1944, ApJ, 100, 360,
doi: 10.3847/1538-3881/aca666 doi: 10.1086/144677
Malofeeva, A. A., Seleznev, A. F., & Carraro, G. 2022, AJ, Southworth, J., Maxted, P. F. L., & Smalley, B. 2005,
163, 113, doi: 10.3847/1538-3881/ac47a3 A&A, 429, 645, doi: 10.1051/0004-6361:20041867
Marks, M., Kroupa, P., & Oh, S. 2011, MNRAS, 417, 1684, Stauffer, J. R. 1982, AJ, 87, 1507, doi: 10.1086/113241
doi: 10.1111/j.1365-2966.2011.19257.x Steele, I. A., & Jameson, R. F. 1995, MNRAS, 272, 630,
Martı́n, E. L., Barrado y Navascués, D., Baraffe, I., Bouy, doi: 10.1093/mnras/272.3.630
H., & Dahm, S. 2003, ApJ, 594, 525, doi: 10.1086/376938 Tang, J., Bressan, A., Rosenfield, P., et al. 2014, MNRAS,
Matsumoto, T., Saigo, K., & Takakuwa, S. 2019, ApJ, 871, 445, 4287, doi: 10.1093/mnras/stu2029
36, doi: 10.3847/1538-4357/aaf6ab Thompson, B. A., Frinchaboy, P. M., Spoo, T., & Donor, J.
Melis, C., Reid, M. J., Mioduszewski, A. J., Stauffer, J. R., 2021, AJ, 161, 160, doi: 10.3847/1538-3881/abde4c
& Bower, G. C. 2014, Science, 345, 1029, Torres, G. 2020, ApJ, 901, 91,
doi: 10.1126/science.1256101 doi: 10.3847/1538-4357/abb136
Mermilliod, J. C., Rosvick, J. M., Duquennoy, A., & Torres, G., Latham, D. W., & Quinn, S. N. 2021, ApJ, 921,
Mayor, M. 1992, A&A, 265, 513 117, doi: 10.3847/1538-4357/ac1585
Weiler, M., Jordi, C., Fabricius, C., & Carrasco, J. M. 2018,
Milone, A. P., Piotto, G., Bedin, L. R., et al. 2012, A&A,
A&A, 615, A24, doi: 10.1051/0004-6361/201732489
540, A16, doi: 10.1051/0004-6361/201016384
 21

APPENDIX

A. BAYESIAN INFERENCE OF PARALLAX FOR NEARBY CLUSTER MEMBERS

For nearby star clusters, both the scale of the line-of-sight distribution and the parallax measurement uncertainties of
individual members are non-negligible. Simply adopting the average parallax of the cluster would ignore the distance
variations among individual members. Alternatively, if we directly use the observational parallax of each member, due
to the typically large uncertainties, additional scatter would be introduced into the line-of-sight distribution of the
cluster. For instance, when using Gaia parallaxes directly, one can often observe an apparent elongation of the nearby
cluster along the line of sight.
Based on the Bayesian Inference, we introduce a method to calculate the more reliable parallax for an object if it is
confirmed to be a cluster member. The posterior probability distribution function (PDF) of the modified parallax of
this object within the cluster can be expressed as:

P(ϖ|ϖo ) ∝ L(ϖo |ϖ) · π(ϖ). (A1)

The π(ϖ) represents the real ϖ distribution of cluster members. Considering that the scale of a star cluster is generally
much smaller than its distance, we can reasonably assume that the π(ϖ) follows a normal distribution,

π(ϖ) ∝ N (ϖc , σϖc ), (A2)

where ϖC is the mean parallax of the cluster, and σϖC represents the intrinsic scatter of the parallax distribution of
cluster members. This scatter can be characterized by the intrinsic dispersion of observed parallaxes or, under the
assumption of spherical symmetry, it can be estimated using the projected size of the cluster. The L(ϖo |ϖ) is the
likely probability distribution according to the parallax observation of the concerned member, which also follows a
normal distribution,
L(ϖo |ϖ) ∝ N (ϖo , σϖo ), (A3)
with ϖo and σϖo to be the observational parallax and the uncertainty of this object.
It can be rigorously proved that the P(ϖ|ϖo ) also obeys a normal distribution,

P(ϖ|ϖo ) ∝ N (ϖm , σϖm ), (A4)

with the modified parallax (ϖm ) and the uncertainty (σϖm ) are derived as:
−2 −2 −2 −2
ϖm = (ϖc σϖc
+ ϖo σϖo
)/(σϖc
+ σϖo
), (A5)

and
−2 −2 −1/2
σϖm = (σϖc
+ σϖo
) . (A6)
For the Pleiades cluster of the present work, the ϖc = 7.41 mas is the average value of our kinematic member
sample. The σϖc is adopted as the intrinsic dispersion fitted from this sample, equals 0.17 mas. This line-of-sight scale
is roughly equivalent to the projected scale of the Pleiades.