A Census of Chemically Peculiar Stars in Stellar Associations

Abstract

The pre-main-sequence evolution of the chemically peculiar (CP) stars on the upper main sequence is still a vast mystery and not well understood. Our analysis of young associations and open clusters aims to find (very) young CP stars to try to put a lower boundary on the age of such objects. Using three catalogues of open clusters and associations, we determined membership probabilities using HDBSCAN. The hot stars from this selection were submitted to synthetic $\Delta a$ photometry, spectral, and light curve classification to determine which ones are CP stars and candidates. Subsequently, we used spectral energy distribution fitting and emission line analysis to check for possible PMS CP stars. The results were compared to the literature. We detected 971 CP stars and candidates in 217 clusters and associations. A relatively large fraction, ∼10% of those, show characteristics of PMS CP stars. This significantly expands the known list of candidate PMS CP stars, bringing us closer to solving the mystery of their origin.

1. Introduction

Among the spectral types B, A, and F, we can find stars that show peculiarities in their spectra. These chemically peculiar (CP) stars are a well-established phenomenon since their discovery almost 130 years ago [1].

Classically, one makes the distinction between the four types defined by [2] with some additions made by [3].

First, CP1 (Am/Fm) stars show enhanced lines of iron and other metals, whereas calcium, for example, is underabundant. The fact that many of those stars are found in binary systems [4] and thus, are rotating rather slowly ($\upsilon sini$$<100$  km ${\mathrm{s}}^{-1}$) leads to the so-called chemical separation, where some elements get driven outwards by radiation and others start to settle down.

The second subgroup is CP2 or Bp/Ap or magnetic (mCP) stars. With (strong) magnetic fields up to tens of kG [5], they exhibit variability in their spectra and photometric time series. This is explained by the Oblique Rotator Theory [6]. Currently, we are unable to explain how such strong magnetic fields can exist in those stars, given that convection is only present in their core. Nevertheless, there have been attempts made by [7], who argue that the magnetic field is somehow frozen into the star, and by [8], where the authors suggest a process of the core going from radiative to convective states during the pre-main-sequence phase.

In the group of CP3 (HgMn) stars, we find overabundances of mercury and/or manganese. Ref. [9] explains these peculiarities with similar processes as in CP1 stars, suggesting a common evolution of these two objects.

The fourth group, defined by [2], can be seen as an extension of CP4 groups to hotter (early to mid-B-type) stars. They show similar characteristics in their variability [10]. In contrast to CP2 stars, however, the CP2 stars show primarily helium peculiarities in the form of either over- (He-strong) or underabundances (He-Weak).

In addition to [2], a new group of CP stars has been discovered: the Lambda Bootis (Lam Boo) stars. These are stars with significant underabundances of almost all heavier elements. The anomalies can be explained by diffusion and mass-loss theories [11]. Another proposed theory by [12] proposes an interplay of accretion and diffusion, where the star passes through a metal-poor interstellar cloud. However, [13] found little to no evidence for this mechanism.

A poorly researched question is the pre-main-sequence (PMS) evolution of those stars. There have been a few CP stars with characteristics of PMS stars reported in the literature [14].

Stars in their PMS phase are characterised by accretion onto the protostar while it contracts and heats up. Commonly, two subgroups are mentioned in the literature that differ by the mass range of the protostar. Lower masses (up to ∼$2{\mathrm{M}}_{\odot }$) are called T Tauri stars, while the ones above this threshold are put into the category of Herbig Ae/Be stars. Ref. [15] gives an extensive overview of how those objects are characterised and how they behave. In short, one can observe them by looking for emission lines (mostly hydrogen) as a sign of the accretion phase in those stars and via infrared excess in their spectral energy distribution (SED).

On the other hand, we have open clusters (OCs) and stellar associations. These objects are invaluable tools for various applications in astronomy, particularly in stellar astrophysics. Astronomers use them to accurately determine stellar ages using various methods, such as isochrone fitting, kinematic studies, and the lithium depletion boundary (LDB) technique, among others. With the advent of more precise and large datasets from the $Gaia$ mission [16,17], we now can use those astrometric measurements to accurately determine which stars belong to clusters and which do not. Recent examples working on open clusters using this dataset are [18,19].

Stellar associations are regions of low stellar density (${\rho }_{★}<0.1{\mathrm{M}}_{\odot }p{c}^{-3}$, [20,21]) which are categorised into three groups [22,23]:

• OB associations: Mostly including hot young stars of predominantly O and B type stars (e.g., the Perseus OB1 association);
• T associations: Dominated by low-mass T Tauri stars in their formation stage (e.g., Chamaeleon 1);
• R associations: Stellar groups containing reflection nebulae (e.g., Monocerotis R2).

The ages of stellar associations range from a few Myr to ∼${10}^2$ Myr [21] and thus are well-suited to study stars right after their formation and arrival on the main sequence [24]. The formation of associations is believed to have been due to one of two scenarios [24]: The first one states that the stars are formed where they are observed as part of a hierarchical star formation process, where over-densities in the interstellar medium can form at any scale, anywhere. The other scenario proposes that stars are formed in clusters that undergo processes such as UV radiation and stellar winds, leading to the cluster no longer being in virial equilibrium and resulting in its dispersal.

This paper aims to look for CP stars in open clusters and stellar associations with a special interest in detecting CP candidates that are still in their PMS phase. In Section 2, we describe the catalogues and data used, and the quality criteria employed to filter these datasets. Section 3 describes the transformation of the astrometry for the clustering in Section 4. The latter also describes the process of determining membership in the aggregations, and the process of classifying stars based on their CP nature. Section 5 provides an overview of how PMS stars are detected, and the last two sections, Section 7 and Section 8, present the results, a discussion, and an outlook on possible future work in this field.

2. Data

2.1. List of Clusters and Associations

Three catalogues of stellar clusters and associations were taken as a list for potential targets:

1.

Stars with Photometrically Young $Gaia$ Luminosities Around the Solar System [25];

2.

The catalogue of open clusters and moving groups by [19];

3.

A comprehensive study about substructures in the Sco-Cen region by [26].

2.2. Used Surveys

The analysis of the clusters, associations and stars in this list was performed using the following surveys:

• $Gaia$ Data Release 3 [16,17]: With its high precision in astrometry, especially parallax (distance) and proper motion, it is ideally suited to search for open clusters and associations (e.g., [27]). All three catalogues described above utilise this dataset. A new feature in this latest data release is the inclusion of low-resolution spectra generated by the blue and red photometers [28,29]. The spectra, ranging from 300 to 1100 nm (Figure 1), are great for synthetic photometry [30] and have been proven to be effective in detecting CP stars using synthetic $\Delta a$ photometry [31].
• TIC v 8.2 [32]: The updated source list for the Transiting Exoplanet Survey Satellite (TESS). The spacecraft was launched in 2018 and observes the sky in sectors of ${24}^{\circ }\times {96}^{\circ }$ with four apertures of 10 cm [33]. Each sector is observed for approximately a month in two cadences: a short cadence (2 min, SC) and a long cadence (30 min, LC). As the name states, its original goal is to observe possible planet-hosting stars. However, the time series data created by the telescope is an excellent source for studying variable stars. The instrument has one drawback: the pixel size is relatively large (21"× 21"), making the possibility of blending in more crowded fields a big problem (e.g., [34]).
• Data Release 9 (http://www.lamost.org/dr9/, accessed on 5 April 2024) of the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST, [35,36]): Data from this Multi-Object spectrograph based in China have a spectral resolution of $R\sim$ 1000–1500 and thus are well-suited for spectral classification. This has led to the discovery of numerous CP stars in the past [37,38]. However, due to its limited movement, it can observe the sky only in the northern part, i.e., between $\delta \in [-{10}^{\circ },+{90}^{\circ }]$.

2.3. Data Retrieval and Quality Assessments

The data for the clusters and associations were downloaded in a region around the astrometric parameters of each object. This means all stars within the parallax range $[{\varpi }_{min},{\varpi }_{max}]$ of the cluster, an angular radius r around its core and within the boundaries of the proper motions $[{\mu }_{\alpha ,min},{\mu }_{\alpha ,max}]$ and $[{\mu }_{\delta ,min},{\mu }_{\delta ,max}]$, were retrieved from the $Gaia$ archive using a python script, and after a recommended zero-point correction on the parallax [39], they were stored in a .csv file for each region on the sky. Generally, a few key quality criteria were taken during this step.

First, for the lists from [25,26], a quality cut on the parallax and its error was taken, leaving only sources with $\varpi /{\sigma }_{\varpi }>10$. In the case of [19], this value was lowered to $\varpi /{\sigma }_{\varpi }>5$ in order to include fainter sources with less accurate astrometry due to the large distances of clusters in this catalogue. Additionally, Ref. [19] gives a parameter called “astrometric signal-to-noise ratio” of which they argue that a value above five means that the cluster is likely to be real. Accordingly, all clusters below this threshold were removed, along with the 133 globular clusters included in their catalogue.

For all the datasets, we took a cut in the re-normalised unit weight error (RUWE), removing poor astrometric solutions and binaries, leaving only stars with $\mathrm{RUWE}<1.4$ in the final dataset [40].

After these astrometric quality cuts, some additional quality criteria on the photometric measurements from $Gaia$ had to be taken into account. It is recommended to remove all sources with $G_{BP}>20.3$ mag due to a systematic brightening effect in those faint regions [41]. The same paper also recommends a cut on the so-called “corrected flux-excess factor” $C^{*}$ that is defined as

$$C^{*}=C-f(G_{BP}-G_{RP})$$

(1)

where C is the original flux excess factor defined as the combined flux in the blue and red photometers divided by the total flux in G:

$$C={\displaystyle \frac{F{G}_{BP}+F_{G_{RP}}}{F_G}}$$

(2)

See [41] for details on this correction. We used a cut of $3\sigma$ on the corrected flux excess factor $C^{*}$.

3. Astrometric Transformations

After all the quality criteria described above were imposed, some astrometric transformations had to be taken into account for the planned clustering on the datasets.

At first, the parallax and galactic coordinates l, b were transformed into 3D Euclidean distances X, Y, Z using a standard transformation:

$$\left(\begin{array}{c}X\\ Y\\ Z\end{array}\right)=R\left(\begin{array}{c}coslcosb\\ sinlcosb\\ sinb\end{array}\right).$$

(3)

where R is simply calculated by taking the inverse of the parallax in mas to get to a distance of pc:

$$R={\displaystyle \frac{1000}{\varpi }}$$

(4)

The next step was to transform the proper motions in the International Celestial Reference System (ICRS) frame ${\mu }_{{\alpha }_{★}}(={\mu }_{\alpha }·cos\delta )$, ${\mu }_{\delta }$ to galactic proper motions ${\mu }_{l_{★}}$, ${\mu }_b$. This was done utilising the vectors and matrices described by the $Gaia$ DR3 documentation (https://gea.esac.esa.int/archive/documentation/GDR3/Data_processing/chap_cu3ast/sec_cu3ast_intro/ssec_cu3ast_intro_tansforms.html, accessed on 28 February 2023).

Finally, the proper motions in the galactic frame were transformed into transverse velocities (in km/s) via

$$\left(\begin{array}{c}{v}_{T_{l_{★}}}\\ v_{T_b}\end{array}\right)=4.74·R\left(\begin{array}{c}{\mu }_{l_{★}}\\ {\mu }_b\end{array}\right)$$

(5)

This finally resulted in a five-dimensional dataset ($X,Y,Z,v_{T_{l_{★}}},v_{T_b}$) that we then used for the subsequent membership analysis of the clusters and associations.

4. Methods

4.1. Clustering

Determining the membership of stars in open clusters and associations is not always a straightforward task, and depending on the data and methods used, one can obtain vastly different results in membership numbers and probabilities. Commonly used methods are Gaussian Mixture Models (GMMs, e.g., [42]), Unsupervised Photometric Membership Assignment in Stellar Clusters (UPMASK, [43]) or, more recently, Significant Mode Analysis (SiGMA, [26]). Other methods try to use a combination of photometry, astrometry, and spectroscopy to determine which stars are members of young associations (e.g., [44]).

To determine which stars of our list were likely members of open clusters and associations, we used the HDBSCAN algorithm [45] that was also used by [19,25].

HDBSCAN is a clustering algorithm with several possible hyperparameters in its implementation. The most important ones are the following:

• min_cluster_size: This is somewhat self-explanatory; it describes how many points should be at a minimum in a cluster. Larger values will result in smaller substructures that get lumped together in a bigger cluster.
• min_samples: This describes how dense a cluster should be. Larger values will result in only the densest cores being considered to be clusters.
• cluster_selection_method: There are two choices for this parameter:

  -

  eom: This stands for “excess of mass” and searches for over-densities in the n-dimensional parameter space, looking for clusters in accordance with the other parameters. It generally favours larger and fewer clusters.

  -

  leaf: This lets the user recover finer and smaller clusters in the dataset.

• metric: This chooses the distance metric that the algorithm uses to calculate the distance between the points in the input set. Commonly used choices are euclidean, manhattan or malahanobis.

Of course, HDBSCAN is not perfect and has some shortcomings. According to [19], it tends to be a bit overconfident, meaning that the algorithm assigns more data points to the cluster than there actually are. However, they used an all-sky approach, whereas this work only clusters the stars in the regions around the clusters and associations. This should result in at least the cores of the aggregations to be recovered.

We randomly selected 30% of the data in each cluster region to look for the best hyperparameter combination of the values seen in

Table 1. Hyperparameter range searched.

Parameter	Range of Values
min_cluster_size	[5, 50] in steps of 5
min_samples	[3, 5, 10, 20, 30, 40]
cluster_selection_method	[eom, leaf]
metric	[euclidean, manhattan]
algorithm	[best]

. The different combinations were evaluated using a silhouette score, from which the highest, i.e., the best parameter combination, was chosen for the final clustering on the set. An example of the clustering result can be seen in Figure 2. In total, the clustering resulted in a list of roughly 462.5 thousand stars that have a high ($p>0.5$) probability of being members of the clusters and associations in question after removing stars that are below the main sequence (i.e., White Dwarfs). From the 430 thousand stars in 2022 clusters and moving groups that resulted from the clustering in and around the catalogue of [19], around 200 k also belong to the initial catalogue of [19]. This is a recovery rate of 15%, a result of the difference in the clustering method: Ref. [19] used an all-sky method, whereas we used the regions around the individual clusters and aggregations. For the catalogue from [25] we recovered a similar fraction (13% or ∼12,500 stars in 26 associations) due to the same reasons of a different approach to the clustering. This discrepancy in membership, of course, may affect the number/fraction of CP stars detected in these groups.

4.2. Extinction Correction and Final Target Selection

This list, however, is still not sufficient to sort out all the stars that are not in question of being chemically peculiar. Since we wanted to look at stars with $T_{\mathrm{eff}}>6500$ K (or hotter than spectral type around mid-F), the cooler stars had to be removed from the list. Although we also wanted to look at PMS CP stars, we opted for this boundary, since there has been no evidence for PMS CP stars in T Tauri stars that we could find in the literature. There might be other classes of peculiar stars in the lower temperature region, predominately barium stars. These, however, are not in the focus of the current work. All previously discovered PMS CP stars belong to the class of Herbig Be/Ae stars; see [46] and references therein. For this purpose, we took the StarHorse21 catalogue from [47]. The authors of this paper compiled measurements from $Gaia$ EDR3, the Two-Micron All-Sky Survey (2MASS, [48]), the Panoramic Survey Telescope and Rapid Response System (Pan-STARRS1, [49]), and the Wide-Field Infrared Survey Explorer All-Sky Release (AllWISE, [50]) to derive extinction values $A_V$ at $\lambda =5420$ Å, and stellar parameters such as $T_{\mathrm{eff}}$, $logg$, $M/{\mathrm{M}}_{\odot }$, and $XYZ$-positions with respect to the galactic center. Approximately 333 thousand stars of the clustered sample are included in this catalogue. Out of those, around a third, 105,270 stars, have a temperature higher than the threshold as mentioned above. See Figure 3 for the CMD of this final target list. We want to emphasise the point that the choice of another catalogue of stellar parameters might result in a different number of stars selected. See [46] for a discussion of the influence of different catalogues on this matter.

4.3. $\Delta a$ Photometry

The first method we used for detecting possible CP stars was $\Delta a$ photometry. This powerful method, developed by [51], makes use of the flux depression in stellar spectra around $\lambda 5200$. First reported by [52] in the spectrum of HD 221568, this depression has been proven to indicate the presence of a strong magnetic field in mCP stars. The most critical element affected here is iron, although there are also contributions of silicon, chromium, and others [53]. The method itself is relatively simple. One takes three filters, $g_1$ to the left (shorter $\lambda$) of the depression, $g_2$ right at the depression, and the Strömgren y filter (see Figure 4) at the right (longer $\lambda$) of the flux depression. Then the magnitudes in each filter are measured or calculated synthetically. These magnitudes can then be converted into an index a that is defined as

$$a=g_2-{\displaystyle \frac{g_1+y}{2}}$$

(6)

This value a is then compared to the value of a (seemingly) normal star that has a value $a_0$ to create $\Delta a$:

$$\Delta a=a-a_0(g_1-y)$$

(7)

The colour term $(g_1-y)$ corrects for temperature differences in the stars, and the function $a_0(g_1-y)$ is then called the "normality line". All stars that are a certain distance (typically, one uses 95% or $3\sigma$ confidence intervals) above this line are considered to be magnetic stars with a significant flux depression.

Before the synthetic photometry, the list of the clustered stars was matched to the catalogue of the $Gaia$ BP/RP (XP in short) spectra via the $Gaia$ DR3 IDs, of which 64 thousand stars had a spectrum. For each of those, the signal-to-noise ratio (S/N) was calculated using the quick way, described in [28]. Basically, since the spectra are downloaded as a set of coefficients for each passband, one can estimate the S/N by the $L_2$ norm of the coefficient vector divided by the $L_2$ norm of the vector of errors on the coefficients:

$$S/N_{XP}={\displaystyle \frac{{\parallel \mathtt{xp}\_\mathtt{coefficients}\parallel }_2}{{\parallel \mathtt{xp}\_\mathtt{coefficient}\_\mathtt{errors}\parallel }_2}}$$

(8)

The final estimation was done by just taking the mean of the S/N in each passband. All stars with an $S/N>100$ were considered for the subsequent analysis. This left 26,814 spectra to be analysed using synthetic $\Delta a$ photometry.

The photometry was done in the same way as described in [31]. At first, the spectra that initially came in flux units of W${\mathrm{m}}^{-2}$n${\mathrm{m}}^{-1}$ were normalised at a common wavelength of $\lambda =4020$ Å. Additionally, they were interpolated using a third-order polynomial technique to match the resolution of the used filters, which have $\Delta \lambda =1$ Å.

Actual astronomical magnitudes in the AB system [54], for example, for a given filter curve $f\left(\lambda \right)$ and a spectrum $s\left(\lambda \right)$ are calculated using the following quantity [55]:

$$m_f=-2.5·log{\displaystyle \frac{\int f\left(\lambda \right)s\left(\lambda \right)\lambda d\lambda }{\int f\left(\lambda \right)(c/\lambda )d\lambda }}-56.10$$

(9)

However, since the interest here only lies in the difference of (instrumental) magnitudes, one can take the much more uncomplicated approach where the filter and the spectrum are just multiplied and the values are then summed up:

$$m_f=\sum _{i}f\left({\lambda }_i\right)s\left({\lambda }_i\right)$$

(10)

After we calculated the synthetic magnitudes this way, the normality line of the form $a_0=c_1+c_2·(g_1-y)$, i.e., the coefficients $c_1,c_2$ needed to be found. For this purpose, we used the Python 3.10 package PyMC [56]. This uses the theorem of Bayes

$$P\left(\widehat{\theta }\right|\widehat{X},\widehat{y})={\displaystyle \frac{P\left(\widehat{y}\right|\widehat{X},\widehat{\theta })P\left(\widehat{\theta }\right)}{P\left(\widehat{y}\right|\widehat{X})}}$$

(11)

with the input vector $\widehat{X}=(g_1-y)$ and $\widehat{y}=a$ to infer the vector of coefficients $\widehat{\theta }$ via a Monte Carlo process. The resulting normality line (see also Figure 5) had the following form:

$$a_0=0.454\left(3\right)(g_1-y)-2.815\left(29\right)$$

(12)

Additionally, PyMC calculated the standard deviation $\sigma$ of this fit line, and it gave a value of $\sigma \approx 0.433$. All stars with a distance of more than $3\sigma \approx 1.298$ above $a_0$ are considered to be candidate mCP stars, which amounted to 341 individual stars.

4.4. Spectral Classification

Another method to find CP stars is the classification of stellar spectra. The $Gaia$ DR3 IDs of the 105 thousand candidate stars on the upper main sequence were submitted to the ninth data release of LAMOST (http://www.lamost.org/dr9/, accessed on 8 April 2024), which contained 4143 spectra of this set. Removing all spectra with S/N lower than 50 in the g-band and, in the case of stars with multiple spectra, keeping the ones with higher S/N left us with 1440 spectra to classify. Those were classified using the MKCLASS code [57]. This program takes an input spectrum and compares it to a standard library to derive the final spectral type after some iterations. The results are comparable to those made by humans, with only minor deviations in the temperature and luminosity classes. It is also well-suited to classify CP stars as done by [37], for example.

For the classifications, we used three standard libraries:

• libnor36;
• libr18;
• libr18_225.

Out of the 1440 spectra, 608 individual stars showed CP characteristics in their spectra for at least one of these libraries. This high fraction of roughly 42% shows that either an unusual amount of stars in young open clusters and associations are CP stars or that some stars have been misclassified by one or more standard libraries of MKCLASS. Especially, libr18_225 has a high recovery rate of CP1 stars, and thus, one has to be careful with the classifications.

Of course, one has to be cautious with the spectra of lower quality, although they should still result in reasonable estimations of the spectral type.

4.5. Light Curve Analysis

A volume-limited subset of the stars with $T_{\mathrm{eff}}$$>6500$ K was also checked for light curves from TESS. This was limited in volume to the list of [25] because of the large pixel size of TESS (Section 2) and to minimise blending as much as possible. The light curves of 1022 stars were downloaded using the Python package eleanor [58]. For the process of retrieving the time series data, an aperture with a radius of one pixel was used (Figure 6). The resulting data were cut by 3$\sigma$ to remove outliers and then converted from flux differences to magnitude differences via Pogson’s equation

$$\Delta mag=-2.5·{log}_{10}\left({\displaystyle \frac{F}{F_0}}\right)$$

(13)

where $F_0$ is the mean flux of the observations.

After this process, we used the astropy [59,60,61] implementation of the Lomb–Scargle periodogram [62,63] that follows the algorithm described in [64]. Additionally, one can calculate the false-alarm probability (FAP) using the methods from [65] to check if the found periods are significant and thus correct. We used a value of $log\left(\mathrm{FAP}\right)\le -3$ for a period to be significant. This process was done in two frequency regions:

• $0.001\mathrm{c}/\mathrm{d}\le f\le 5\mathrm{c}/\mathrm{d}$;
• $f\ge 5\mathrm{c}/\mathrm{d}$.

We then classified the light curves based on their frequency spectrum and phase-folded light curve (Figure 7 and Figure 8). Generally, we followed the criteria given in [34] while analysing the light curves with the types of variability taken from the Variable Star Index (VSX, [66]) and the notation from the General Catalogue of Variable Stars (GCVS, [67]). The most prominent variability classes in our sample are pulsators such as Gamma Doradus (GDOR) or Delta Scuti stars (DSCT), Rotating stars (ROT) and their subgroup of $\alpha$ Canum Venaticorum (ACV) stars, and stars with variability of unknown origin or unclear signal (VAR). A general scheme we used can be seen in

Table 2. Basic classification scheme for the visual classification of variability types. This is a short overview of the most prominent classes in the sample, a more detailed description can be found in [34].

Type	Features in Light Curve	Features in Periodogram
GDOR	(ir)regular, beating, sharp curves	two or more peaks $<5$ c/d
DSCT	beating, short periods	two or more peaks $>5$ c/d
ROT	sinusoid variation, repeating pattern with stable features	Peaks at harmonics of rotational frequency
ACV	rotation with “double waves” due to chemical spots	Same as ROT but clear CP2 stars from LC and/or spectrum
VAR	variability present, no clear features	one or more peaks with unknown origin

.

Classification of light curves presents several challenges. These range from unclear signals in the periodograms, poor data, ambiguity in classification due to a lack of more detailed information about the star in question, and blending effects. Accordingly, there is a non-negligible chance that light curves will be classified incorrectly. See e.g., [34] for a more detailed discussion on that matter.

However, 512 (50%) out of the 1022 light curves showed apparent variability and could be classified. The result can be seen in Figure 9. A total of 22 stars had light curves with attributes from CP stars, i.e., a so-called “double wave” from the Oblique Rotator [6]. The classes in question are ACV (after $\alpha$ CVn, also Figure 8) and SXARI (after SX Arietis), which are essentially hotter analogues of the ACV variables.

Altogether, 971 individual CP stars and candidates in 217 open clusters and associations were detected from at least one of those three methods.

5. Pre-Main-Sequence Stars

Finally, from all the detected CP stars, we wanted to assess how many of those are still in their PMS or accretion phase. To achieve this, one can employ two primary methods: fitting the spectral energy distribution (SED) or searching for emission in the Balmer lines of the spectrum.

5.1. Spectral Energy Distributions

The SEDs of the stars in question were created using the Virtual Observatory SED Analyser (VOSA, http://svo2.cab.inta-csic.es/theory/vosa/, accessed on 2 October 2024, [68]). This tool essentially searches for available distances, extinction values, and photometry. Then it fits an SED to the data and also allows one to determine accurate parameters such as radius, mass, and luminosity of the star in question, depending on the accuracy of the input parameters.

We took the coordinates of the 971 stars in question and submitted them to VOSA. One could also include other parameters, such as magnitudes or fluxes in one or more filters, distances, and extinction values ($A_V$). However, to avoid inhomogeneities in one or more of these values, we stuck with the coordinates. An SED with infrared excess is seen in Figure 10. Out of the whole sample of 971 CP stars and candidates, 92 show IR excess (

Table A4. List of the CP Stars and candidates that show IR excess.

Gaia DR3 ID	Filter Where Excess Starts
182621906350208384	WISE/WISE.W4
183323227266520704	WISE/WISE.W4
183337177320364928	WISE/WISE.W1
1871158238309294208	WISE/WISE.W3
187964540726357632	WISE/WISE.W1
2007069351654097664	WISE/WISE.W1
2070344734000894592	WISE/WISE.W3
2075877648313310848	WISE/WISE.W1
2075877678366383744	WISE/WISE.W4
2076066351994581120	WISE/WISE.W4
2085246793048249344	WISE/WISE.W3
216590390373910912	Spitzer/IRAC.I4
216617641943232000	Spitzer/MIPS.24mu
251761469543936512	WISE/WISE.W3
253455576447809280	WISE/WISE.W4
3017409624342379520	WISE/WISE.W3
3019790754200699136	WISE/WISE.W1
3019875485316186752	WISE/WISE.W1
3125739694655670912	UKIRT/UKIDSS.K
3209504544905800576	WISE/WISE.W4
3210561587897894528	WISE/WISE.W4
3304135250100217600	WISE/WISE.W2
3330884473923251968	WISE/WISE.W4
3342955019950652928	WISE/WISE.W4
3355226737949262336	WISE/WISE.W4
3355241409558154368	WISE/WISE.W4
3370708239624841344	WISE/WISE.W3
3372437530897057920	WISE/WISE.W4
3372466874113370240	WISE/WISE.W4
3373851365410096896	WISE/WISE.W3
3373860500801601408	WISE/WISE.W3
3377260401211784448	WISE/WISE.W3
3377541841828979456	WISE/WISE.W4
3383565585001209728	WISE/WISE.W3
3383771812150964096	WISE/WISE.W3
3394744216639326208	WISE/WISE.W3
3444662212044370944	WISE/WISE.W3
3445205168927696512	WISE/WISE.W4
3455831880089411072	WISE/WISE.W3
3455880567838460544	WISE/WISE.W3
435065038743702528	WISE/WISE.W1
4469325071096213504	WISE/WISE.W3
458406742891002496	WISE/WISE.W1
473267192293775360	WISE/WISE.W1
48058592395097856	WISE/WISE.W3
1824774481337240576	WISE/WISE.W1
1981253265315002880	WISE/WISE.W3
2015462885981225984	WISE/WISE.W3
2028730486687781248	WISE/WISE.W4
2055676974020815488	WISE/WISE.W1
2058950666823409152	WISE/WISE.W4
2164328273542938880	UKIRT/UKIDSS.K
2178496541693778688	WISE/WISE.W3
2930678291716664192	WISE/WISE.W1
2931333566285206912	WISE/WISE.W1
2931478907978887424	WISE/WISE.W1
3051142842935888512	WISE/WISE.W4
3124537309971018752	WISE/WISE.W1
3125525015011266304	UKIRT/UKIDSS.K
3126379163747692672	WISE/WISE.W3
3126387135207429376	WISE/WISE.W3
3126748977611495936	WISE/WISE.W1
3128813203316883456	WISE/WISE.W3
3129234045689550976	WISE/WISE.W1
3324289470039249408	WISE/WISE.W1
3352222215048693504	WISE/WISE.W3
3355188117603280640	WISE/WISE.W1
4064497976508146688	WISE/WISE.W2
428891315316060160	WISE/WISE.W1
435067993681192320	UKIRT/UKIDSS.K
464638293759397632	WISE/WISE.W1
465827037629837696	WISE/WISE.W4
5254330709292306560	WISE/WISE.W1
5254332770876859008	WISE/WISE.W4
5310688514207181824	WISE/WISE.W1
5333062854296532480	WISE/WISE.W4
5335935942551982080	WISE/WISE.W2
5337166162957067648	WISE/WISE.W1
5337376891236690560	WISE/WISE.W3
5338495403810748416	AKARI/FIS.WIDE-S
5338502202699086080	WISE/WISE.W1
5339377246496587392	WISE/WISE.W1
5339411846756340736	WISE/WISE.W1
5339894291828433024	WISE/WISE.W1
5339935695306646016	WISE/WISE.W3
5339942189297580416	WISE/WISE.W3
5351221941647302912	WISE/WISE.W1
5351241973377009280	WISE/WISE.W1
5352004244161147136	WISE/WISE.W1
546642482291623168	WISE/WISE.W3
5518730442165840768	WISE/WISE.W1
5520118060206928896	WISE/WISE.W1
5547527132743054336	WISE/WISE.W1
5547599769226387840	WISE/WISE.W3
5547978963304967296	WISE/WISE.W1
5619904272346555520	WISE/WISE.W3
5715306662901791232	WISE/WISE.W2
5865056126594144128	WISE/WISE.W2
6054029567979605632	WISE/WISE.W1
6054273556477806976	WISE/WISE.W1
6054329940813382016	WISE/WISE.W2
6054868770221377920	Spitzer/IRAC.I4
6054918488763944832	WISE/WISE.W2
6056528822281714816	WISE/WISE.W1

).

5.2. Emission Lines

The other method we used to detect stars in their PMS phase is the detection of emission lines of hydrogen, especially H$\alpha$. Ref. [70] made an approach similar to the $\Delta a$ photometry to detect emission lines. They took one wide and one narrow synthetic (Gaussian) filter with a Full-Width at Half-Maximum (FWHM) of 125 Å  and 35 Å, respectively. The filters are both centred at 6555 Å  (Figure 11). The filters were then folded with the $Gaia$ XP spectra of a sample of known Herbig stars, and then compared the synthetic magnitudes to gain an index:

$$H\alpha ={\displaystyle \frac{H{\alpha }_{\mathrm{wide}}}{H{\alpha }_{\mathrm{narrow}}}}$$

(14)

This is similar to the $\beta$-index from Stömgren photometry [71] used to determine effective temperatures of stars.

If this value is below a certain threshold, one can assume the star exhibits emission in H$\alpha$ and is, thus, likely a PMS object.

We took this approach and used the LAMOST spectra to determine the spectral types of the stars. They were treated in the same way as the XP spectra for the synthetic photometry, i.e., normalised at $\lambda 4020$ and interpolated to a resolution of 1 Å. We also had to modify the filters, shifting the central wavelength to 6564 Å  to get the best results. Then, the magnitudes were calculated in the same way as in Section 4.3.

Additionally, the spectra were checked with the Python package specutils [72] to search for emission lines in the region around the H$\alpha$ line. The combined product of these two processes can be seen in Figure 12. From all the stars with emission lines, twelve are also (candidate) CP stars (

Table A5. CP candidates that show emission lines. The columns are (1) Gaia DR3 ID, (2) LAMOST ID, (3) right ascension $\alpha$ in degrees, (4) declination $\delta$ in degrees, (5) parallax $\varpi$ and parallax error ${\sigma }_{\varpi }$, (6) the colour $\left(g{1}_{-}y\right)$ from the synthetic photometry, (7) the H$\alpha$ index, (8)–(10) the spectral types derived from MKCLASS.

Gaia DR3 ID	LAMOST	$\mathit{\alpha }$	$\mathit{\delta }$	$\mathit{\varpi }\pm {\mathit{\sigma }}_{\mathit{\varpi }}$	$({\mathit{g}}_1-\mathit{y})$	H$\mathit{\alpha }$	libr18	libnor36	libr18_225
183576114939066752	J052117.64+345221.5	80.324	34.873	0.325 ± 0.03	9.529	1.903	kB9hF0mK0 Eu	B9.5 III	kB9.5hF0mG2 Eu
3131334967590689536	J063207.38+045455.8	98.031	4.916	0.649 ± 0.017	11.703	1.729	F0 mB5 V Lam Boo	B7 V	kB8hF0mK0 Eu
182636268718056576	J052815.47+342525.6	82.064	34.424	0.489 ± 0.02	9.835	1.486	B4 IV-V	B4 V	kB7hF2mF9
2059202695503955584	J200754.27+361314.1	301.976	36.221	0.316 ± 0.022	9.638	1.406	kA2hA3mA6	A1 IV	kA2hA3mA7
2007069351654097664	J225100.01+570219.6	342.75	57.039	0.297 ± 0.011	9.616	1.326	kB9.5hF0mG2 Eu	B9 II-III	kA0hF0mG8 Eu
182694783351830912	J052405.96+340124.2	81.025	34.023	0.411 ± 0.016	9.041	1.668	B9 III	B9.5 II-III	kA0hF1mG2
3343027622077611392	J055508.01+115959.2	88.783	12	0.639 ± 0.025	9.763	1.851	F0 mB8 V Lam Boo	B9 IV	F0 mA1 V metal-weak
182635620182511744	J052805.13+342130.3	82.021	34.358	0.461 ± 0.015	9.466	1.871	B6 V	B7 V	kB9hF1mK0 Eu
1979250470521255424	J214103.58+503109.8	325.265	50.519	0.241 ± 0.019	9.172	1.691	B9.5 II-III	A0 II-III	kA0hF0mA5 Eu
3131335040608985216	J063213.98+045514.2	98.058	4.921	0.671 ± 0.017	12.859	1.57	B8 IV Si	emission-line?	F1 V Fe-0.8
182659976937528320	J052819.83+342720.1	82.083	34.456	0.803 ± 0.016	10.12	1.526	kA2hF5mK0	A2 V	kA2hA3mK0 Eu
3372466255638764928	J062601.84+194004.7	96.508	19.668	0.275 ± 0.018	9.576	1.723	kB9hF0mK0 Eu	B9 III	kB9.5hF0mK0 Eu

).

6. Comparison with Literature

To verify whether we had identified any previously known CP stars, we also performed cross-matching with catalogues from the literature. The catalogues in question are as follows:

1.

Ref. [73]: A collection of CP stars from the literature. The authors compiled a list of 8205 stars that have been previously flagged as CP stars. However, no assessment of the correctness of the classifications was done, so some of them have been confirmed, while others have been proven to be “normal” stars. A total of 156 stars in the target list of stars hotter than 6500K are contained in this catalogue, of which 9 could be detected in this work.

2.

Ref. [74]: The authors of this paper searched 200,000 stars from LAMOST DR5 for CP features using a random forest classifier together with an additional manual classification. They publish a list of 9372 CP1 and 1132 CP2 stars. Subsequently, they investigated on the spatial distribution of the stars. This catalogue contains 27 objects, which could also be recovered using my analysis.

3.

Ref. [37]: This catalogue of 1002 CP stars was conceived using spectra from LAMOST DR4 and a modified version of MKCLASS. Additionally, they checked for the flux depression using $\Delta a$ photometry. The authors checked for the evolutionary status on the main sequence, mass, and spatial distribution. Four stars from the list of possible targets in this work can be found in their catalogue, half of which I could detect.

4.

Ref. [75]: This catalogue uses spectra from the Apache Point Observatory Galactic Evolution Experiment (APOGEE, [76]) to look for HgMn (CP3) stars. A total of 260 new CP3 stars were catalogued by their work, of which I could recover seven.

5.

Ref. [38]: Here, the authors used an XGBOOST algorithm and manual classification on LAMOST DR8. They arrived at a list of 17,986 CP1 and 2708 CP2 stars and included astrophysical parameters ($T_{\mathrm{eff}}$, $logg$, $[Fe/H]$) of the stars. In total, 15 out of the 17 stars that are also in my target list could be rediscovered by my methods.

6.

Ref. [77]: The authors used MKCLASS and an ensemble regression model with spectra from LAMOST DR10. Out of the 27,543 CP stars and candidates, they list ∼11,614 new Am stars and candidates as well as 4978 new Ap stars and candidates. They also include stellar parameters ($T_{\mathrm{eff}}$, $logg$, $[Fe/H]$) in their catalogue. I could recover 59 stars from their list.

Whenever possible, we used the $Gaia$ DR3 IDs to avoid mismatching, and in all the other cases, a search radius of 1 arcsecond was used to minimise the risk of false identification.

We note the low recovery rates when cross-matching to the catalogues. The reasons for this are not clear. Effects arising from far distances can likely be ruled out since all the catalogues and surveys mentioned above also include far and faint sources. A possible explanation may be due to the low coverage of the densest parts of the galactic plane by large spectroscopic surveys and the TESS mission. We are also aware that a part of our CP stars and candidates may have been misclassified by either the automatic or manual classification of the spectra and light curves in our sample or some of the catalogues using machine learning algorithms to classify their spectra.

7. Results

7.1. $\Delta a$ Photometry

Looking at the flux depression in stars with $Gaia$ XP spectra, 341 could be detected using $\Delta a$ photometry. Three of the detected stars additionally have a spectral type from LAMOST, all being CP2 stars (

Table 3. Stars that have a positive $\Delta a$ value and a spectral type from LAMOST.

$\mathit{Gaia}$ DR3	LAMOST	$\mathit{\alpha }$	$\mathit{\delta }$	libnor36	libr18	libr18_225	$\Delta \mathit{a}$
253428535333057920	J042756.29+442830.8	66.985	44.475	kB9.5hA1mA2 Si	kB9.5hA2mA5 SrSi	kA0hF0mA6 SiEu	1.519
253436747311206528	J042906.36+442429.2	67.277	44.408	A0 II-III Si	A0 II-III Si	F1 V Fe-5.9	2.014
3344712726726459520	J061909.56+143120.1	94.790	14.522	kA1hA2mA5 Si	A0 III-IV SrSi	kA1hA8mA8 SiEu	1.357

). See

Table A2. CP stars detected via $\Delta a$ photometry with their basic astrometric properties. The columns are as follows: (1) Source: the Gaia DR3 ID, (2) $\alpha$: the right ascension in degrees, (3) $\beta$: the declination in degrees, (4) $\varpi$: the parallax and its error ${\sigma }_{\varpi }$ in mas, (5) System: The cluster or association the star belongs to, (6) The colour $(g_1-y)$ from the synthetic photometry, (7) $\Delta a$: the $\Delta a$ value from the photometry. The stars are sorted by their right ascension $\alpha$. Again, the first 30 entries are shown.

Source	$\mathit{\alpha }$	$\mathit{\delta }$	$\mathit{\varpi }\pm {\mathit{\sigma }}_{\mathit{\varpi }}$	System	$({\mathit{g}}_1-\mathit{y})$	$\Delta \mathit{a}$
5436414262899903872	144.758	−36.896	0.227 ± 0.044	HSC 2138	15.122	1.707
2930678291716664192	110.647	−18.713	0.251 ± 0.042	FSR 1252	11.584	2.208
461829660022360832	50.986	59.24	0.953 ± 0.015	Theia 1722	10.614	1.777
5960713736908382336	265.155	−40.215	0.673 ± 0.039	Trumpler 29	11.36	1.423
5881393391868603264	226.247	−55.643	0.523 ± 0.014	NGC 5823	9.197	2.32
5881395625251644288	226.379	−55.536	0.566 ± 0.013	NGC 5823	9.957	2.022
5881396037568526336	226.393	−55.515	0.526 ± 0.013	NGC 5823	9.775	1.319
5881345799292540288	226.371	−55.671	0.542 ± 0.013	NGC 5823	9.324	1.371
5881345837987362304	226.387	−55.669	0.593 ± 0.013	NGC 5823	9.113	1.46
5881348414936867584	226.425	−55.581	0.526 ± 0.017	NGC 5823	9.096	1.444
200085728707316096	73.536	40.35	0.932 ± 0.024	HSC 1284	12.231	1.547
4145604723045931136	273.091	−16.472	0.539 ± 0.021	OC 0025	9.862	1.503
5716196751916693760	114.074	−20.58	0.362 ± 0.029	NGC 2421	11.456	1.89
5716196958080774144	114.015	−20.578	0.344 ± 0.043	NGC 2421	11.495	1.698
5715446026000379264	114.086	−20.613	0.405 ± 0.038	NGC 2421	10.768	1.602
436077895107052928	45.559	47.929	0.394 ± 0.038	UBC 1246	11.529	1.429
436088684064785664	45.472	48.101	0.275 ± 0.043	UBC 1246	11.257	1.591
1870942218639565696	312.984	37.477	0.797 ± 0.017	Roslund 7	12.453	1.41
6071466104392664064	186.029	−58.125	0.389 ± 0.024	NGC 4337	11.567	1.583
5520069849191410304	124.238	−44.843	0.055 ± 0.037	OC 0473	10.916	1.468
…	…	…	…	…	…	…

for the first entries of the list of spectral classifications. The entire list will be available online. A colour–magnitude diagram (CMD) can be seen in Figure A2.

7.2. Spectral Classification

From the LAMOST spectra, we detected 608 CP stars in three standard libraries. One example spectrum of a newly detected CP2 star is shown in Figure 13. Additionally, a few new Lambda Bootis stars were detected. An example is in Figure A1. Note that not all spectra were of good quality, so there is a chance that some of them have been classified incorrectly. A CMD of these stars can be seen in Figure A3. The beginning of this catalogue is presented in

Table A1. Stars that were classified as peculiar in at least one library from MKCLASS. The columns are (1) Gaia DR3 ID, (2) LAMOST ID, (3) right ascension $\alpha$ in degrees, (4) declination $\delta$ in degrees, (5) parallax $\varpi$ and parallax error ${\sigma }_{\varpi }$, (6) open cluster/association the star is in, (7) classification from libnor36, (8) classification from libr18, and (9) classification from libr18_225. Only the first 30 entries are shown.

Source	LAMOST ID	$\mathit{\alpha }$	$\mathit{\delta }$	$\mathit{\varpi }\pm {\mathit{\sigma }}_{\mathit{\varpi }}$	System	libnor36	libr18	libr18_225
422486316486081280	J000940.81+575109.5	2.42	57.853	0.288 ± 0.021	UBC 1195	B9 V	B9 V	kA0hA3mG8
422850976388400896	J001115.46+575859.3	2.814	57.983	0.295 ± 0.013	UBC 1195	kB9hF0mG8 Eu	F0 mA3 IV-V metal-weak	kB9.5hF0mG8 Eu
419025466196922752	J002705.69+524904.1	6.774	52.818	0.566 ± 0.019	Theia 2936	kA8hA8mF4	kA7hA8mF5	kA7hA8mF4
423487627981131648	J010004.63+552553.8	15.019	55.432	0.767 ± 0.015	AT 21	A6 IV-V (Sr)	A7 V	A7 V
423486318006267008	J010036.07+552153.7	15.15	55.365	0.835 ± 0.012	AT 21	kA2hA3mA6 Si	A1 IV-V SrSi	kA2hA6mA7 SrSi
522537579650553856	J010816.40+613405.5	17.068	61.568	0.854 ± 0.023	NGC 381	F5 V (Sr)	F5 V (Sr)	F5 V (Sr)
426344193550526080	J011006.42+602835.7	17.527	60.477	0.361 ± 0.018	UBC 413	kB9.5hF0mA5 Eu	kB9.5hA9mA7 Eu	kA0hA9mA7 Eu
410923920922247168	J011920.73+551700.2	19.836	55.283	1.662 ± 0.015	Theia 714	A5 III Sr	kA6hA9mF0	A9 V
510674433300466816	J012534.95+610827.0	21.396	61.141	0.383 ± 0.016	COIN-Gaia 31	kB9.5hF0mK0 Eu	B9.5 III	kA0hF0mK0 Eu
506120805889910272	J015402.86+575513.7	28.512	57.92	0.419 ± 0.017	UBC 1577	kA1hF5mK2	A0 III-IV	A8 mA1 V metal-weak
504941442233879296	J015618.23+565650.4	29.076	56.947	0.455 ± 0.021	UBC 1577	B9 III-IV Si	B9.5 III-IV Si	F0 mA0 V metal-weak
504940239643050752	J015627.82+565457.6	29.116	56.916	0.41 ± 0.019	UBC 1577	kA2hF5mK0	A1 V	kA3hA3mK0
504942473026002432	J015707.89+565841.0	29.283	56.978	0.447 ± 0.017	UBC 1577	A2 V	A2 V	A2 V Eu
504868904521760768	J015803.44+563443.4	29.514	56.579	0.452 ± 0.014	UBC 1577	A1 V	kA1hA2mA4	kA2hA3mG2
504535791156850176	J015815.54+553501.6	29.565	55.584	0.763 ± 0.017	NGC 744	A1 V	A0 V	kA1hA3mK0
504523765248458624	J015831.44+553159.6	29.631	55.533	0.725 ± 0.02	NGC 744	kA6hA8mF0	kA6hA8mF1	kA6hA8mF0
505262671425427712	J015954.67+570441.4	29.978	57.078	0.448 ± 0.023	UBC 1577	A1 V	DZ	kA2hA3mK0
505103272298128896	J020240.75+571130.1	30.67	57.192	0.406 ± 0.019	UBC 1577	kB9.5hF4mK2	kB9.5hA1mA2	kB9.5hF1mK0
505086371610011904	J020337.90+570205.8	30.908	57.035	0.427 ± 0.016	UBC 1577	A3 mA0 V Lam Boo	A0 IV-V	A3 mA0 V Lam Boo
505036412548787072	J020523.60+565926.4	31.348	56.991	0.412 ± 0.015	UBC 1577	F0 mB8 V Lam Boo	F0 mA3 IV-V metal-weak	kB9hF0mG2 Eu
458378499188409856	J021851.84+571255.0	34.716	57.215	0.38 ± 0.014	NGC 869	kB9hF0mA1 Eu	F0 V Fe-0.9	F0 mA2 V metal-weak
458455121404670720	J022212.78+571021.1	35.553	57.173	0.404 ± 0.02	NGC 884	A0 IV	A0 IV-V	kA0hA3mK1
458406742891002496	J022246.66+570448.8	35.694	57.08	0.383 ± 0.019	NGC 884	kB9hA8mK0 Eu	B9.5 IV	kA0hA8mK0 Eu
450639616650222720	J023611.58+502045.3	39.048	50.346	1.377 ± 0.019	COIN-Gaia 8	A1 V	A3 mA0 metal weak	kA2hA3mK0
450567632995957888	J023759.19+495427.9	39.497	49.908	1.385 ± 0.02	COIN-Gaia 8	kA2hF5mK0	A2 V	kA3hF5mK0
436028657602103168	J030153.84+474751.2	45.474	47.798	0.414 ± 0.019	UBC 1246	kB9.5hF1mG2 Eu	B9 III	kA0hF1mG2 Eu
435058750911681280	J031436.47+471151.0	48.652	47.197	0.292 ± 0.018	NGC 1245	F0 V Eu	A9 V (Sr)	A9 V Eu
435065038743702528	J031439.44+471619.9	48.664	47.272	0.316 ± 0.02	NGC 1245	?	A9 V	F0 V Eu
435064660786649088	J031445.87+471246.9	48.691	47.213	0.292 ± 0.017	NGC 1245	A8 IV-Vn	kA7hA9mF1	kA7hA8mF0
435064798225554688	J031449.77+471438.3	48.707	47.244	0.316 ± 0.016	NGC 1245	A4 IV Eu	A7 V Eu	A6 IV Eu
…	…	…	…	…	…	…	…	…

; the complete list will also be available online.

7.3. Photometric Time Series

Out of the approximately 1000 stars that had light curves available from TESS, 20 showed variability connected to CP stars, namely ACV or SXARI variability. The periods range from approximately one-third of a day to five days, which aligns with previous findings on these variables. A CMD of the variable CP stars can be seen in Figure A4 and the list with basic astrometry, periods, and FAP can be found in

Table A3. Stars with light curves that show CP properties. The columns are (1) Gaia DR3 ID, (2) right ascension $\alpha$ in degrees, (3) declination $\delta$ in degrees, (4) parallax $\varpi$ and parallax error ${\sigma }_{\varpi }$, (5) the name of the association the star is in, designation from [25], (6) variable star type according to the VSX, (7) the FAP of the period, (8) the period of the variability.

Source	TIC	$\mathit{\alpha }$	$\mathit{\delta }$	$\mathit{\varpi }\pm {\mathit{\sigma }}_{\mathit{\varpi }}$	System	Type	$\mathit{log}\left(\mathit{FAP}\right)$	$\mathit{P}\left(\mathit{d}\right)$
3314527696567187584	17560109	67.283	18.455	$4.531\pm 0.020$	Taur Ori 4	ACV	$-\infty$	1.017
3238478566084270208	397061497	75.275	3.717	$3.479\pm 0.048$	Taur Ori 4	SXARI	$-\infty$	2.42
3207058750010997632	306345958	78.520	−7.917	$3.728\pm 0.022$	Greater Ori	ACV	−128.252	4.028
3213882697128481792	4070682	80.103	−3.510	$3.181\pm 0.020$	Greater Ori	SXARI	$-\infty$	1.043
3346259567786672384	247675468	86.855	12.648	$3.351\pm 0.037$	Taur Ori 4	ACV	$-\infty$	0.952
3346656590266363648	247763365	87.357	13.922	$3.654\pm 0.030$	Taur Ori 4	ACV	$-\infty$	0.382
3330219445485837184	436332628	91.617	10.750	$5.555\pm 0.041$	Taur Ori 2	SXARI	$-\infty$	0.362
3104244792087711360	42884620	97.045	−4.899	$3.184\pm 0.045$	Mon SW	SXARI	$-\infty$	1.173
5514764057046837248	388354822	124.907	−50.382	$3.128\pm 0.016$	Vela CG4	ACV	$-\infty$	1.439
5515989290962253184	82893704	125.065	−47.969	$3.400\pm 0.020$	Vela CG4	ACV	−300.311	2.501
5322386214086050176	89943694	127.570	−50.941	$3.082\pm 0.011$	Vela CG4	ACV	$-\infty$	1.463
5424236729248685824	77105310	141.327	−43.624	$5.625\pm 0.012$	IC 2391	ACV	$-\infty$	1.471
5232258082033470720	397732137	159.330	−69.363	$4.652\pm 0.022$	Cha	ACV	−222.997	1.584
5239734932919731456	461125565	159.573	−65.042	$6.606\pm 0.030$	IC 2391	SXARI	$-\infty$	0.549
5239736204230183680	464907721	159.878	−65.083	$6.652\pm 0.012$	IC 2391	ACV	$-\infty$	0.86
4523888885382756224	50325659	278.215	17.303	$3.179\pm 0.019$	Cerberus	SXARI	$-\infty$	2.99
2262275415015759872	229793060	284.719	69.531	$5.115\pm 0.037$	Cep Far North	ACV	$-\infty$	1.126
2254485134616928128	229954300	288.681	65.273	$4.858\pm 0.031$	Cep Far North	SXARI	$-\infty$	1.306
2301907242919520640	236003103	305.141	82.848	$4.238\pm 0.022$	Cep Far North	ACV	−306.611	1.323
2190139274622512128	429331590	313.699	57.611	$3.012\pm 0.021$	Cep Cyg	ACV	−220.353	1.417

.

7.4. PMS Stars

From the analysis of the CP candidates in terms of their evolutionary status, we found that twelve stars show emission lines in H$\alpha$. Additionally, 92 stars showed infrared excess in their SED, another sign of an early evolutionary status. See Table A4 for the stars and the start of the IR excess as well as Table A5 for the stars with H$\alpha$ emission.

7.5. Binarity

Binarity might play a significant role in the CP phenomenon, as already described in Section 1. Basically, the CP stars rotate much more slowly than the “normal” stars in the same spectral region. To reduce the angular momentum, binary systems are most suitable. The general binarity rate of all B- and A-type stars is about 50% [78]. For the CP stars, there is no recent study of the binarity rate available. However, looking into the corresponding literature, we find a value of 50% [4,79], even for the CP1 stars [80]. The $Gaia$ DR3 catalogue contains 1.5 × ${10}^9$ sources with a full, single-star astrometric solution, but potentially also contains a large number of binary systems. The presence of an unresolved companion can affect the single-star solution [81] and thus the parallax and proper motions, most important to derive the membership probability to a stellar association, for example. Undetected binary systems can be traced by an elevated RUWE value [82,83]. As described in Section 2.3, we took a cut in the RUWE parameter of 1.4, respectively. This should already exclude the binary systems for which the $Gaia$ single-star solution is not valid. Our sample comprises 20 objects with a flag of multiple source identifiers. Those are good candidates for being binary systems. But it can also mean that there are other close members of the stellar association. Finally, we also checked for known wide and astrometric binaries discovered by $Gaia$ [84,85]. In total, we found thirteen matches or slightly above 1%. We can, therefore, conclude that our sample consists mainly of single stars.

8. Summary and Outlook

Using different methods ($\Delta a$ photometry, spectral, and light curve classification), we detected 971 CP stars in 217 open clusters and associations; most of those CP candidates have not been named in the literature before. The membership probabilities of the stars in the respective stellar aggregation were determined using HDBSCAN. Additionally, we used SED-fitting and emission line photometry to determine which CP stars are likely to be in their PMS phase.

We found that 12 CP stars show emission in H$\alpha$, whereas 92 CP stars show infrared excess. Note that the excess is relatively weak in most cases, suggesting that the stars are already in the late PMS evolution stage and close to the Zero-Age Main Sequence (ZAMS).

The stars with emission lines all belong to the category of CP2 (mCP) stars, agreeing with the suggestion that the magnetic field is at least partially responsible for the occurrence of CP stars [14]. However, this is still a tiny sample and a much more detailed investigation is needed to confirm or disprove these ideas.

Nevertheless, the present list of about 100 new PMS CP candidates vastly expands the list of known PMS sources with CP properties, which will be a stepping stone for future research on the formation of these objects, particularly the origin and early evolution of the magnetic fields in stars on the upper main sequence and the onset of the diffusion processes in these objects. A recent study by [86] already discusses the occurence of CP stars in open clusters and also made an analysis of the evolutionary status. They discuss a decrease in CP stars in the evolution from the ZAMS to the terminal-age main sequence (TAMS). We did not analyse the present sample in this way due to the relatively large number of unconfirmed CP candidates. This, however, is surely an opportunity for future research with higher quality of spectral data from future telescope such as the Extremely Large Telescope which is currently under construction.