New Coronae and Stellar Associations Revealed by a Clustering Analysis of the Solar Neighborhood

We used a subset of all Gaia EDR3 entries within approximately 200 pc of the Sun with good-quality 6D kinematics to identify nearby streams of stars. To achieve this, we selected stars with a Gaia EDR3 parallax above 5 mas with Renormalised Unit Weight Errors unit weight errors (RUWE ⁵ ) > 3 to exclude astrometric solutions strongly affected by unresolved companions or other systematic problems such as cross-matching errors. We have found that combining such an RUWE < 3 criterion with additional quality criteria on parallax and proper motions was necessary to avoid an imprint of the distant Galactic plane into high-parallax (i.e., nearby) Gaia EDR3 samples, because the crowded Galactic plane is prone to cross-matching errors that can mimic large parallax values. However, we have found that additional proper motion or parallax quality cuts were not necessary when working with the sample of Gaia EDR3 stars with radial velocity measurements, because a simple RUWE < 3 quality cut applied on stars with parallaxes 5 mas or above resulted in high-quality parallax values with signal-to-noise values above 15, and similarly high-quality proper motion measurements with signal-to-noise values above 6. We note that the RUWE selection cut will bias our search against binary stars (most of which have RUWE > 1.4; e.g., see Stassun & Torres 2021), but will also greatly reduce sample contamination due to low-quality parallax or proper motion measurements.

We have calculated the individual XYZ Galactic positions and UVW-space velocities of our sample with a Monte Carlo simulation. This XYZUVW frame of reference is centered on the solar position and motion, where the positive X/U axes point toward the Galactic center, and the Y/V axes are in the direction of the rotational motion of the Galaxy, such that XYZ and UVW form right-handed coordinate systems. For each star in our sample, we have taken 100 artificial measurements of proper motion, radial velocities, and parallaxes centered on the Gaia EDR3 values with standard deviations equal to their respective measurement errors. We have then transformed each of these artificial measurements on the XYZUVW frame of reference, and taken the median value along each direction as the most likely XYZUVW values for each individual star. We have also calculated approximate measurement errors by taking the standard deviations of the resulting distributions along each dimension, and further rejected any star with resulting errors above 3 pc or 3 km s⁻¹ in the individual XYZUVW dimensions. We elected to apply these rigid selection cuts to reduce systematic and measurement errors because the HDBSCAN algorithm, described in Section 2.2, does not account for measurement errors by construction.

We note that the Monte Carlo simulation used here would be equivalent to using the flat-prior distance estimations of Bailer-Jones et al. (2021) in the specific step where parallaxes are transformed to distances, but our method also similarly deals with nonlinear effects in the transformation of measurement densities from radial velocities and proper motions to the XYZUVW-space. While non-flat Bayesian priors may be useful to calculate the distance probability density for more distant stars, our sample is constituted of stars within 200 pc of the Sun with parallaxes at a high signal-to-noise ratio (above 15), meaning that the effects of a non-flat Bayesian prior would be negligible, especially after having rejected stars with a projected measurement density above 3 pc or 3 km s⁻¹ in XYZUVW-space.

We have selected a volume of 200 pc around the Sun for this search because previous surveys have not efficiently recovered loose associations within 100 pc in a systematic way (see Section 3.6 for more details), and the 100–200 pc region will be useful to identify extensions of known large-scale structures such as those recovered by Kounkel & Covey (2019) based on their spatial overlap. We also note that considering a larger volume would require significant computer memory depending on the chosen clustering algorithm, with little benefits compared to previous clustering surveys.

The general properties of our input sample are displayed in Figure 1. The Galactic XYZ-positions of our sample stars have typical measurement errors of 0.1–0.5 pc, and the UVW-space velocities have typical measurement errors of 0.1–1.0 km s⁻¹, with no significant differences in the distribution of measurement errors along the three spatial or kinematic axes. These typical measurement errors are in the range of typical velocity dispersions in star-forming regions and moving groups. Based on these combined selection criteria (parallax above 5 mas, RUWE < 3, XYZ and UVW errors below 3 pc or 3 km s⁻¹), we obtained an input sample of 303,540 Gaia EDR3 entries.

Zoom In Zoom Out Reset image size

We used the Python scipy implementation of the accelerated version of HDBSCAN (McInnes & Healy 2017), which can take as an input a set of N-dimensional coordinates (with coherent spatial units) for M objects, and offers several metrics to compute the mutual distances between individual objects (Euclidean distances are used by default). However, HDBSCAN can also take as an input an N × M matrix of pairwise distances that were precomputed with any custom metric not offered in the library. Although it would be possible to precompute distances in XYZUVW-space that marginalize over the unknown heliocentric radial velocities of most Gaia EDR3 entries, we found that such an approach required an unmanageable amount of computer memory even with much more restricted sample volumes. The distance metrics offered in the HDBSCAN library all respect the triangle inequality, allowing HDBSCAN to avoid computing most of the pairwise distances between all M objects, and the resulting computational requirements scale approximately with ${ \mathcal O }(M\mathrm{log}M)$ instead of ${ \mathcal O }({M}^{2})$. However, distance metrics that rely on marginalized heliocentric radial velocities do not respect the triangle inequality—because the best-matching radial velocity between a given pair of stars AB may be different from the best-matching radial velocity when testing a different pair of stars AC. As a consequence, they cannot be implemented with an ${ \mathcal O }(M\mathrm{log}M)$ scaling and require a precomputation of a large matrix of pairwise distances.

We have therefore elected to use HDBSCAN with the default Euclidean distance metric; however, we have taken an approach similar to Kerr et al. (2021) where the mutual distances in kinematic dimensions are given more importance than those in spatial dimensions because we expect coeval groups of stars to have formed with a narrow UVW distribution (typically 1–3 km s⁻¹; Zuckerman & Song 2004), but with much larger spatial extents over time—up to ≈ 15 pc in Z, and hundreds of parsecs in XY, in some cases (e.g., see Kounkel & Covey 2019; Meingast et al. 2019 and Meingast et al. 2021). Furthermore, the XYZ spatial positions have a different physical dimension than UVW velocities, and therefore it becomes important to consider a transformation factor that bears physical units for consistency. We have thus constructed a 6 × M array of $\left[X,Y,Z,c\cdot U,c\cdot V,c\cdot W\right]$ coordinates where c = 12 pc km⁻¹ s⁻¹. This value is larger than the c = 6 pc km⁻¹ s⁻¹ used by Kerr et al. (2021) because we aim to consider potentially older groups similar to Kounkel & Covey (2019), whereas Kerr et al. (2021) specifically investigated a sample of < 50 Myr-old stars. In fact, the ≈200 young associations, open clusters, and tidal tails within 500 pc of the Sun with ages ≲1 Gyr (Gagné et al. 2018c; Tarricq et al. 2020; Cantat-Gaudin et al. 2021) have typical spatial-to-kinematic ratios of their characteristic axes (defined as the cube root of their 1σ spatial or kinematic volume) in the range 1–14, and 90% of them have a ratio below 12. Those with lower ratios are usually the denser cores of open clusters, and those with larger ratios are either loose moving groups or open cluster coronae. Choosing a value of c = 12 will therefore ensure that our clustering step is able to recover associations that are spatially as large as currently known moving groups and open cluster coronae.

HDBSCAN offers two clustering approaches: "Excess of Mass" (EOM), which directs how it separates subclusters within larger clusters of stars. In the EOM approach, clusters that are most stable under changes in detection threshold are preferred, whereas the leaf method will tend to split such clusters farther apart when subclusters are detected. We elected to use the leaf clustering method to obtain a sample of individual substructures, even though some of them may be related to each other. For example, one might expect that extended structures like the Sco-Cen OB association would be preserved as a final cluster under the EOM clustering method, whereas the leaf method would further identify its subpopulations (e.g., Upper Scorpius, Lower Centaurus Crux and Upper Centaurus Lupus; see Blaauw 1946; de Zeeuw et al. 1999 and Pecaut & Mamajek 2016 for more detail), which have similar but distinct ages. Our choice of using the "leaf" clustering method will provide us with a finer sampling of individual structures, but we may find upon further study that some of these structures are related to each other through a larger past star-forming complex, for example. We used the default values for the and α parameters. ⁶ We have chosen a minimum cluster size of N = 10 with HDBSCAN to start investigating the most significant missing structures of the solar neighborhood in this work. We have avoided values of N < 10 to reduce the number of false clusters that may correspond to random collections of stars; choosing a much larger value, on the other hand, would prevent us from recovering some of the sparser young associations in the solar neighborhood, especially because we only consider stars in the temperature and G-band magnitude ranges that benefit from Gaia DR2 radial velocities. Currently known young associations within 100 pc of the Sun have between 2 and 200 members with Gaia DR2 radial velocity measurements, and 80% of these associations have at least 10 such members (some exceptions are discussed in Section 3.2).

The application of HDBSCAN with this particular configuration to our sample described in Section 2.1 yielded a total of 241 clusters, which are listed in Table 1 with their individual lists of Gaia EDR3 source identifications. The median XY Galactic positions of all Crius groups are shown in Figure 2, which shows how they fill the gap within ≈100 pc from the Sun where previous clustering studies were less efficient due to projection effects. Following the spirit of Kounkel & Covey (2019) in their nomenclature of Theia groups, we hereafter refer to our resulting 241 putative ensembles of co-moving stars as Crius 1 to Crius 241. ⁷

Zoom In Zoom Out Reset image size

Table 1. Individual Gaia EDR3 Entries Recovered as Clusters by HDBSCAN

Crius	Gaia EDR3	Right Ascension	Declination	${\mu }_{\alpha }\cos \delta$	μδ	Parallax	Radial Velocity
Group ID	Source ID	(hh:mm:ss.sss)	dd:mm:ss.ss	(mas yr−1)	(mas yr−1)	(mas)	(km s−1)
1	2491729335519097600	01:58:24.266	−05:09:53.51	246.99 ± 0.04	147.53 ± 0.04	14.09 ± 0.04	51.5 ± 0.4
1	83929987451343104	02:56:05.817	18:53:12.75	109.72 ± 0.02	19.47 ± 0.01	6.66 ± 0.02	88.9 ± 0.8
1	459661938487952000	02:56:35.793	55:26:06.82	727.63 ± 0.02	−455.23 ± 0.02	49.32 ± 0.02	76.2 ± 0.3
1	57366301921437952	03:38:35.335	20:09:45.29	95.44 ± 0.02	16.35 ± 0.01	8.41 ± 0.02	102 ± 1
1	3441580727627952512	05:45:22.715	27:38:13.04	−5.22 ± 0.02	−31.27 ± 0.01	15.63 ± 0.02	108 ± 1
1	3813898626334257664	11:21:30.469	05:17:40.32	−275.02 ± 0.02	60.96 ± 0.01	11.41 ± 0.01	12.8 ± 0.5
1	3813323276811078656	11:32:39.157	05:13:43.76	−301.36 ± 0.02	72.59 ± 0.01	12.03 ± 0.02	15.6 ± 0.4
1	3930922058356114432	12:47:13.039	14:42:16.43	−354.72 ± 0.02	149.31 ± 0.02	16.50 ± 0.02	−24 ± 1
1	6197890597023337344	15:00:38.072	−38:41:07.83	−210.94 ± 0.02	−92.50 ± 0.01	14.09 ± 0.02	−86 ± 1
1	4595150639153952000	17:27:35.020	27:01:38.22	−9.09 ± 0.01	370.94 ± 0.02	20.34 ± 0.01	−72.1 ± 0.3

Note. Only a portion of the full table is shown here. The complete table is available as a machine-readable table.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

We determined whether each of the Crius groups identified in the previous section corresponds to a known association, moving group, or open cluster by comparing their list of Gaia EDR3 names with membership lists of known associations in the solar neighborhood. This set of memberships builds on those compiled by Malo et al. (2013), Riedel et al. (2017), and Gagné et al. (2018c) and gathers more than 450 publications that includes but is not limited to membership lists on the Strasbourg Astronomical Data Center (Ochsenbein et al. 2000). The full list, including those not discussed in this work, will be presented in the form of an online database in a future publication (J. Gagné et al. 2022, in preparation), and includes all known associations and open clusters discussed in the literature that we could identify with an average distance within 400 pc of the Sun, including those that were discovered under more than one name, or are currently thought to be an unphysical collection of stars. These include recent surveys such as Kerr et al. (2021), Liu et al. (2021), and Tarricq et al. (2021). Every instance where a number of Crius Gaia EDR3 entries matches members of a known association was investigated in both UVW-space velocities and XYZ Galactic positions using the 3D data visualization software Partiview (Levy 2003) to determine whether the overlap between the known association and the Crius group is significant.

Subsequently, all Crius groups for which no member matched a known association were also investigated in UVW-space velocities and XYZ Galactic positions using Partiview, considering all plausible matching groups in the literature for which bulk UVW-space velocities fell within a radius of ≈10 km s⁻¹ of the median UVW-space velocity of Crius groups, and for which bulk XYZ Galactic positions fell within a radius of ≈50 pc in of the median XYZ Galactic position of Crius groups.

These two methods allowed us to identify several Crius groups that likely correspond to known associations, listed in Table 4 and discussed in Section 3.1. In most cases, the overlap was mostly complete, meaning that we have recovered the majority of the known members in the substructure, and the full spatial overlap of the Crius group is similar to that of the known young association.

In a few cases, a Crius group was recovered that only spans a fraction of the full spatial distribution of a known young association; these cases are discussed further in Section 3.2. In other instances, we have recovered significant spatial extensions of known associations that share the same UVW-space velocities within 3 km s⁻¹. A number of these extensions simply correspond to the forefront of known loose associations toward nearby distances or extensions in the tangential direction that prevented previous clustering algorithms from identifying them because they are based on 2D tangential velocities rather than 3D space velocities (as discussed in Section 3.4). However, a number of other Crius groups seem to correspond to the coronae of less-well-studied open clusters such as Group 23 of Oh et al. (2017, sometimes also named Oh 23) and Volans–Carina (Gagné et al. 2018a), in line with the recent discoveries of a large number of similar structures around known open clusters (see, e.g., Meingast et al. 2021). These new candidate coronae are listed in Table 5 and discussed further in Section 3.5. The remaining Crius groups that do not seem to correspond to known structures are listed in Table 6 and discussed further in Section 3.3.

Table 5. Newly Recognized Coronae

Name	Age	Number of Stars in		Distances (pc)		Other
	(Myr)	Core	Corona	Core	Corona	Names
Volans–Carina	${89}_{-6}^{+7}$	48	93	87	26–124	Oh 30 b , Theia 424, Crius 221
Oh 59 b	≈ 160	33	376	160	82–249	Theia 209, Theia 133 a , Crius 226
Oh 51 b	≈ 90	11	170	161	93–431	Theia 218, Crius 172
Platais 3	≈ 180	85	241	179	108–518	Theia 787, Crius 156
RSG 2	≈ 125	60	276	198	123–329	Oh 16 b , Theia 214, Crius 189, Crius 217
Alessi 9	≈ 280	85	418	208	119–762	Theia 598, Crius 171
UPK 612	≈ 100	60	276	216	154–476	Theia 216
Oh 23 b	≈ 300	17	347	218	96–493	Theia 599, Crius 190, Crius 202

Notes. Ages are from Gagné et al. (2018a; Volans–Carina), Kounkel & Covey (2019; Oh 23, Oh 51, and Oh 59), Tarricq et al. (2012; Platais 3), Cantat-Gaudin et al. (2020; Alessi 9), and Röser et al. (2016; RSG 2).

^a Only a small part of Theia 133 is likely associated with the Oh 59 corona; the rest is mainly associated with the corona of the α Per open cluster (e.g., see Meingast et al. 2021). Further studies based on Gaia DR3 will help to properly disentangle the coronae of α Per and Oh 59 given their similar kinematics. ^b Oh et al. (2017) groups.

Download table as:  ASCIITypeset image

There are a few relatively well-studied nearby moving groups that were notably not recovered in our clustering analysis of the solar neighborhood. These groups are discussed further below.

3.2.1. TW Hya Association

3.2.2. Carina-Near Moving Group

3.2.3. Argus Association

3.2.4. AB Dor Moving Group

3.2.5. Octans Association

3.2.6. Pisces Moving Group

3.2.7. Ursa Major Cluster

We have further investigated whether one of the Crius groups recovered here may correspond to a candidate extension of the Ursa Major cluster (e.g., King et al. 2003), potentially extending it toward the more distant candidate tidal tails identified by Gagné et al. (2020b) from an investigation of the Theia groups of Kounkel & Covey (2019). We found that Crius 141 (84 members) appears to match the UVW distribution of the core Ursa Major members and seems more widely distributed spatially, potentially toward the five Theia groups that may constitute the tidal tails of Ursa Major (see Figure 6). However, Crius 141 was not included in our detailed study of Crius groups because of its median absolute deviation of 16.5 pc in Z, which is slightly above our 15 pc selection criterion. In fact, Crius 141 shares four stars in common with the core list of King et al. (2003), four with their list of Ursa Major "stream" members, and two more with the candidate tidal tails of Gagné et al. (2020b). Further characterization of its members will be required to determine whether Crius 141 is related to the Ursa Major core.

Zoom In Zoom Out Reset image size

A total of 32 Crius groups passed our quality selection criteria and do not appear to correspond to a known coeval association of stars. These groups, which may correspond to new loose coeval associations, are listed in Table 6 and their median Galactic positions are shown in Figure 7. The typical number of stars in these new groups is in the range 10–28. When possible, we have assigned them with approximate age ranges based on their color–magnitude diagrams, the lower age range of which is based on the late-K or M stars compared with the empirical sequences of Gagné et al. (2021) constructed from members of nearby moving groups. The higher age range is based on a main-sequence turnoff of their OBA stars compared with the old clusters NGC 752 (1.2 Gyr; Cantat-Gaudin et al. 2020) and Ruprecht 147 (3 Gyr; Cantat-Gaudin et al. 2020). An example is displayed in Figure 5 for Crius 155. However, several of these putative new groups only have Gaia DR2 heliocentric radial velocities for their FGK members, which prevents a more accurate age-dating.

Zoom In Zoom Out Reset image size

The new candidate associations presented here still require detailed follow-up observations to determine their ages more precisely. The color–magnitude diagram positions of later-type members discovered either with a more detailed Bayesian analysis that does not require radial velocity measurements, radial velocity measurements or once Gaia DR3 is released with about 33 million heliocentric radial velocities will aid in that analysis. ⁹ Lithium measurements and rotation periods will also be important to constrain the ages of these putative new associations and further test whether their members are coeval. The identification of white dwarf members of these young associations will also be an interesting avenue to provide an independent age calibration; however, this will require an analysis that does not rely on heliocentric radial velocities given that they will not be available even in Gaia DR3.

Several groups identified by Kounkel & Covey (2019) show an elongated spatial structure, and we might expect that some of these would extend to the immediate solar neighborhood (≲100 pc); however, most of the nearest Theia groups stop short of this region, likely because of projection effects that start to impact the efficiency of clustering algorithms in 2D tangential velocity space. Conversely, the Crius groups identified here will not extend to very far distances, in part due to our 200 pc selection cut on the input sample, but also because we only consider the brighter Gaia entries that benefit from heliocentric radial velocities in Gaia DR2. We might therefore expect that some of the Crius groups discovered here are in fact a spatial extension of previously discovered Theia groups, and we have identified 17 such cases where Crius groups share similar UVW velocities with Theia groups and form a spatial extension in the same plane in the Galactic Z-coordinate. These cases are listed in Table 7 and the individual distributions of Galactic positions and space velocities of these candidate extensions are shown in the Appendix. Nine of these newly discovered extensions do not match a known spatially localized core that would make it possible to categorize them as coronae; these cases therefore simply correspond to new spatial extensions of spatially loose groups of stars.

The cases of Crius 168 and Crius 227 (displayed in the Appendix) are less straightforward because of a spatial gap between them and their corresponding Theia groups, but Gaia DR3 will likely determine without ambiguity whether they are physical extensions by bridging the gap with the radial velocities for fainter and more distant stars.

3.4.1. The MELANGE–1 Moving Group

Tofflemire et al. (2021) recently identified MELANGE–1, a new moving group at ≈100 pc from the Sun. This new association was identified following the serendipitous discovery that the young exoplanet host star HD 110082 was co-moving with a large number of other seemingly young stars. The kinematics of this group of stars, which include HD 110082, did not clearly match any previously known young association, suggesting that MELANGE–1 was likely a newly discovered moving group. Tofflemire et al. (2021) identified a total of 133 members of MELANGE–1 (including HD 110082) and combined available lithium measurements and rotation periods to determine an age of ${250}_{-70}^{+50}$ Myr for the newly discovered MELANGE–1 ensemble. They further identified that this moving group would be a compelling extension to Theia 786, given its consistent color–magnitude diagram–based age (${280}_{-60}^{+80}$ Myr), similar kinematics, and adjacent Galactic positions.

Although we have not recovered any known member of MELANGE–1 in our Crius groups, Crius 170 (10 members) shares similar space velocities and is adjacent to MELANGE–1 in Galactic positions, as shown in Figure 8. We have also independently identified Crius 170 as an extension of Theia 786 before identifying the similarity in UVW with MELANGE–1, and we also show Theia 94 in these figures because it has similar but slightly discrepant kinematics and it is located in the neighborhood of Theia 786 spatially. It is unclear at this point whether Theia 94 is related to Theia 786, Crius 170, and MELANGE–1, but it is likely to be a distinct subpopulation of a larger group at best. For now, we consider it likely that Crius 170, Theia 786, and MELANGE–1 are different spatial segments of the same coeval population of co-moving stars, and it will be interesting to investigate whether additional radial velocity measurements in Gaia DR3 will allow us to recover the full spatial extent of this population at once in a clustering analysis, and how Theia 94 fits in that picture.

Zoom In Zoom Out Reset image size

3.4.2. Older Moving Groups Associated with the Taurus–Auriga Star-forming Region

Crius 216 (18 members) includes members of both the relatively poorly characterized 118 Tau association (Mamajek 2016a) and the ${22}_{-3}^{+4}$ Myr old 32 Ori association first discovered by Mamajek (2007) and further characterized by Shvonski et al. (2010) and Bell et al. (2017). Currie et al. (2017) estimated that the age of the 118 Tau association seems older than the background Taurus–Auriga star-forming region based on its mid-M stars having Na I line strengths (Slesnick et al. 2006a, 2006b) comparable to members of the 10–16 Myr Scorpius-Centaurus complex (Pecaut & Mamajek 2016), and the lithium lines of its mid K-type stars being larger than those of typical 30–120 Myr old stars and comparable with those of members of the Cha association (3–8 Myr; Murphy et al. 2013; Dickson-Vandervelde et al. 2021). They have therefore assigned a plausible age range of 3–10 Myr for the 118 Tau association, slightly younger than the current estimates for the 32 Ori association. Kraus et al. (2017) noted that 118 Tau and 32 Ori have very similar kinematics with respect to each other, and are only separated from Taurus–Auriga members by 9 km s⁻¹ in UVW-space, indicating that these three populations might correspond to individual but contiguous star formation events that originate from a single extended molecular cloud similar to Scorpius-Centaurus. Our independent recovery of members of the 118 Tau and 32 Ori associations within Crius 216, as well as a continuous set of stars that span their spatial and kinematic XYZUVW coordinates (shown in Figure 9), strongly favors the hypothesis of Kraus et al. (2017). This hypothesis can be further tested in the near future by characterizing the ages of stars in Crius 216 with lithium measurements, as well as an isochronal age determination of its missing M-type members that will be recovered with Gaia DR3.

Zoom In Zoom Out Reset image size

It is interesting to note that HDBSCAN has not separated members of 118 Tau from 32 Ori even though we have used the "leaf" method that would encourage breaking such substructure into distinct groups. This could indicate that there are additional subpopulations of intermediate positions, ages, and space velocities distributed between the two associations, but a more detailed analysis based on Gaia DR3 that includes many more stars will be useful to assess this.

Several of the extensions of known associations identified in Section 3.4 appear to be newly recognized coronae around known open clusters and other young associations. These are listed in Table 5, and maps of their XY Galactic positions and UV-space velocities are displayed individually in the Appendix. Figure 10 shows the spatial extent of these new coronae compared with those already known in the literature. It is perhaps not surprising that additional known associations also host extended coronae given that most well-studied, nearby, and dense open clusters have now been demonstrated to host similar structures (Meingast & Alves 2019; Meingast et al. 2019; Tang et al. 2019; Röser & Schilbach 2021), which likely correspond to tidal dissipation tails around the gradually degrading denser core of gravitationally bound stars. The main reason why these coronae were not previously recognized seems to be that they are related to less-well-studied, sparse open clusters (Platais 3, Alessi 9), or recently discovered associations (Volans–Carina, Groups 23, 51, and 59 of Oh et al. 2017 and RSG 2 of Röser et al. 2016). We also note that Crius 87 (24 members), which displays a large Z median absolute deviation (29.1 pc), initially appeared to form a candidate tidal tail around the UPK 612 open cluster of Cantat-Gaudin et al. (2020), but a closer inspection of its UVW-space velocities appears inconsistent with this hypothesis. However, Theia 216 of Kounkel & Covey (2019) appears to form a previously unrecognized corona around UPK 612, with mostly consistent kinematics (Figure 18, shown in the Appendix).

Zoom In Zoom Out Reset image size

The case of Volans–Carina (${89}_{-6}^{+7}$ Myr; Gagné et al. 2018a) is particularly interesting (see Figure 11), because its candidate corona (Crius 221, 67 members) seems to reach within the immediate solar neighborhood, with a member at 22 pc from the Sun. This nearest member is the K1V-type, debris disk hosting star V419 Hya, the activity index $\mathrm{log}{R}_{\mathrm{HK}}^{{\prime} }=-4.32$ (Isaacson & Fischer 2010) of which is consistent with an age of 80–120 Myr (Mamajek & Hillenbrand 2008), and displays Ca II and X-ray emission consistent with an age of ≈ 100 Myr (Song et al. 2004). It will be particularly interesting to study its substellar population in the near future. In addition to these newly recognized coronae, we note that Crius 178 (55 members) may form a slight extension of the NGC 2451A corona toward the Sun.

Zoom In Zoom Out Reset image size

We show in Figure 12 the spreads in properties ¹⁰ for the Crius groups, compared with the Theia groups of Kounkel & Covey (2019) to better demonstrate the impact of projection effects on the ability to recover clusters of stars with coherent UVW velocities when relying on only the projected tangential velocities of their members. The Crius groups show on average tighter UVW distributions compared with Theia groups, even before our kinematic selection criteria are applied. This is probably due at least in part to the fact that Kounkel & Covey (2019) did not include Gaia DR2 heliocentric radial velocities in their analysis, and therefore any star with an available radial velocity that is in reality an interloper that just happens to share a similar 2D tangential velocity with a Theia group will significantly impact our calculation of the Theia group's UVW spread. Despite the Crius groups having small UVW dispersions, the fact that their members are spread on the sky causes them to display larger spreads in tangential velocities, on average. The distribution of the Z median absolute deviations of Crius groups are generally in line with those of Theia groups.

Zoom In Zoom Out Reset image size

We cross-matched the members of Crius groups with the Gaia DR2 source identifiers of exoplanet host stars in the NASA exoplanet archive ¹¹ and the NASA list of TESS exoplanet candidates (Stassun et al. 2019) to determine whether additional exoplanet systems had been recovered in our analysis and may potentially benefit from future age constraints based on their host associations. The results of this cross-match are listed in Table 8. We recovered 11 confirmed exoplanet systems in Crius groups that correspond to known associations or open clusters, and which memberships had previously been discussed in the literature, and five additional TESS candidate exoplanet systems in Crius groups that we have matched to known young associations, but for which memberships had not been discussed in the literature to our knowledge.

In addition to these, we have uncovered the two TESS exoplanet host stars TOI–1807 (204 ± 50 Myr) and TOI–2076 (180 ± 40 Myr) of Hedges et al. (2021) in Crius 224. These two exoplanet host stars were first noted by Hedges et al. (2021) as co-moving as well as being coeval despite their unusually large 12.5° separation on the sky. Oh et al. (2017) had also recovered these two stars as a likely binary pair based on their similar kinematics using Gaia DR1 data (in their Group 3914, with only two members). However, the large separation led Hedges et al. (2021) to hypothesize that they might be members of an ≈ 200 Myr old stream of co-moving stars , of which other members might have thus far escaped detection. This system also escaped further investigations of likely new coeval associations contained in the Oh et al. (2017) groups (e.g., see Faherty et al. 2018) because this particular group included only two stars. Nardiello et al. (2022) also investigated 24 possible stars co-moving with these two exoplanet hosts, and used lithium equivalent width measurements and TESS rotation periods to estimate an age of 300 ± 80 Myr for the group. We have recovered both stars as members of Crius 224 (28 members), and found that its members are consistent with an age of 100–700 Myr based on their color–magnitude position. This is yet another scenario similar to MELANGE–1 that is suggestive of the validity of the newly discovered groups presented here, which have likely escaped detection prior to Gaia because of their low spatial density and slightly older age, which prevented a discovery based on their X-ray or near-UV-bright members or more massive, early B-type stars.

Moreover, we have uncovered one additional confirmed radial velocity exoplanet (HD 103949b; Feng et al. 2019) in the newly discovered 100–700 Myr candidate association Crius 162 (14 members) and three additional TESS exoplanet candidates (TOI–1598b, TOI–2481b, and TOI–2133b; see Stassun et al. 2019) in three additional new candidate associations (Crius 109, ≥ 100 Myr with 16 members; Crius 121, ≥ 100 Myr with 13 members; and Crius 205, 100–700 Myr with 12 members, respectively). To our knowledge, no age constraints are currently available for HD 103949 b, TOI–1598 b, TOI–2481 b, and TOI–2133 b. It will be particularly interesting to fully characterize Crius 162, 109, 121, and 205, given that exoplanet systems with well-calibrated ages are rare and will provide valuable information about their early evolution.

Among the likely extensions of known associations discussed in Section 3.4, Crius 226 (23 total members) seems related to the α Persei open cluster, Group 59 of Oh et al. (2017), Theia 133, and Theia 209. Figure 13 shows that the kinematics of Crius 226 are consistent with a number of stars in Theia 209 and a small subset of Theia 133 members, whereas they tentatively form a bridge between the two spatially distinct structures in the XY-plane. Furthermore, a large fraction of Theia 133 members have kinematics significantly discrepant with Crius 226, and are instead clearly related to the α Persei open cluster and its corona, as noted by Kounkel & Covey (2019). It seems likely that Theia 133 is therefore composed of two populations with distinct kinematics, only one of which may be directly related to Crius 226 and Theia 209, which together may span distances in the range ≈ 100–200 pc, partially overlapping with the corona of α Persei. Kounkel & Covey (2019) estimated an age of ≈160 Myr for Theia 209 (relatively well constrained by the two B8 stars 27 Vul and bet02 Cyg, as well as 7 ≥ M5 members), slightly older than the lithium depletion boundary age of α Persei (90 ± 10; Stauffer et al. 1999). The color–magnitude diagram of Crius 226 is indicative of an age of ≈ 100 Myr or older, consistent with the age of Theia 209.

Zoom In Zoom Out Reset image size

Because of their similar kinematics and spatial overlap, we investigated whether α Persei and Group 59 of Gaia Collaboration et al. (2016) might have collided in the recent past, using the dynamical traceback approach of D. Couture et al. (2022, in preparation). ¹² Independent backward Galactic orbits were integrated for every member of α Persei and Group 59 with the galpy Python package (Bovy 2015) using the Galactic potential model I from Irrgang et al. (2013) for 100 Myr into the past.

The resulting mutual median distances over time between α Persei and Group 59 are shown in Figure 14. Although the Galactic positions of individual stars become gradually more uncertain as they are projected further in the past, the median positions of the clusters are less affected because of their large number of members. This fact is also much less important in the Z-direction, because stars in our sample are gravitationally bound to the Galactic plane. We found that the two groups came at their closest approach ≈ 39 Myr ago, while they crossed the same plane in the Z-direction twice in the past 100 Myr, ≈ 38 and ≈ 77 Myr ago. However, their respective median positions have never been closer than ≈ 260 pc, which excludes an overlap of their spatial distributions even at three times the median absolute deviation of their current membership distributions, even when including the backward projected measurement errors at this epoch (see Figure 14). We find that it is therefore unlikely that these two associations are related to each other, unless significant systematics remain in the Galactic potential of Irrgang et al. (2013b; see also Bovy 2015), or other collision events that would significantly affect the relative positions of Oh 59 and α Persei in the recent past.

Zoom In Zoom Out Reset image size