A Kinematically Cold Structure of Candidate Young OB Stars Toward The Anticenter

TL;DR

This study identifies a kinematically cold, young OB star structure in the Galaxy's outskirts, likely caused by a satellite-induced wobble, providing insights into galactic disk perturbations.

Contribution

First detection of a young, cold OB star structure in the Galaxy's outer regions, suggesting a satellite-induced disk wobble rather than a spiral arm or warp.

Findings

The structure contains ~300 OB candidates between 12-15 kpc from the Galactic center.

It exhibits a mean upward vertical motion of 5.7 km/s and outward radial motion of 8-16 km/s.

Velocity dispersion confirms the young age of the structure.

Abstract

We combine GALEX and Gaia DR2 catalogs to track star formation in the outskirts of our Galaxy. Using photometry, proper motions and parallaxes we identify a structure of ~ 300 OB-type candidates located between 12 and 15 kpc from the Galactic center that are kinematically cold. The structure is located between l = 120 and 200 degrees, above the plane up to ~700 pc and below the plane to ~ 1 kpc. The bulk motion is disk-like; however we measure a mean upward vertical motion of 5.7 +-0.4 km/s, and a mean outward radial motion of between 8 and 16 km/s. The velocity dispersion along the least dispersed of its proper-motion axes (perpendicular to the Galactic disk) is 6.0 +-0.3 km/s confirming the young age of this structure. While spatially encompassing the outer spiral arm of the Galaxy, this structure is not a spiral arm. Its explanation as the Milky-Way warp is equally unsatisfactory. The structure's vertical extent, mean kinematics and asymmetry with respect to the plane indicate that its origin is more akin to a wobble generated by a massive satellite perturbing the Galaxy's disk. The mean stellar ages in this outer structure indicate the event took place some 200 Myr ago.

Tables

Table 1. Table 1: Kinematical Results for the OB candidates sample

Run	Configuration	Distance	$U_s$	$V_s$	$W_s$	${\sigma }_{{\mu }_l}$	${\sigma }_{{\mu }_b}$	$N$
		Source/Corrections	(km s-1)	(km s-1)	(km s-1)	(mas yr-1)	(mas yr-1)
A	baseline1	spiral arm	$-{14.8}_{-2.2}^{+2.2}$	${2.5}_{-0.7}^{+0.7}$	${5.7}_{-0.4}^{+0.4}$	${0.316}_{-0.012}^{+0.012}$	${0.254}_{-0.010}^{+0.010}$	396
B	$R_0=8.5$ kpc	spiral arm	$-{13.3}_{-2.1}^{+2.2}$	${2.5}_{-0.6}^{+0.7}$	${5.7}_{-0.4}^{+0.4}$	${0.317}_{-0.012}^{+0.012}$	${0.254}_{-0.009}^{+0.010}$	396
C	${\Theta }_0=220$ km s-1	spiral arm	$-{11.2}_{-2.1}^{+2.2}$	${2.5}_{-0.7}^{+0.7}$	${5.8}_{-0.4}^{+0.4}$	${0.319}_{-0.012}^{+0.012}$	${0.254}_{-0.009}^{+0.010}$	396
D	${\Theta }_0=240$ km s-1	spiral arm	$-{16.4}_{-2.2}^{+2.1}$	${2.4}_{-0.7}^{+0.7}$	${5.7}_{-0.4}^{+0.4}$	${0.316}_{-0.012}^{+0.013}$	${0.254}_{-0.009}^{+0.010}$	396
E	$d\Theta /dR=0.0$	spiral arm	$-{12.9}_{-2.1}^{+2.2}$	$-{5.4}_{-0.7}^{+0.7}$	${5.7}_{-0.4}^{+0.4}$	${0.322}_{-0.011}^{+0.013}$	${0.255}_{-0.010}^{+0.011}$	396
F	baseline	parallax + 0.03	$-{14.2}_{-2.2}^{+2.2}$	${4.1}_{-0.7}^{+0.7}$	${5.9}_{-0.3}^{+0.3}$	${0.326}_{-0.012}^{+0.012}$	${0.253}_{-0.009}^{+0.010}$	396
G	baseline	parallax + 0.05	$-{14.0}_{-1.7}^{+1.7}$	${4.7}_{-0.6}^{+0.6}$	${6.1}_{-0.3}^{+0.3}$	${0.322}_{-0.012}^{+0.012}$	${0.253}_{-0.010}^{+0.010}$	396
H	baseline	parallax + 0.03 ${ϵ}_{\pi }\le 0.05$	$-{16.0}_{-2.4}^{+2.4}$	${4.6}_{-0.8}^{+0.8}$	${5.7}_{-0.4}^{+0.4}$	${0.337}_{-0.013}^{+0.015}$	${0.263}_{-0.011}^{+0.012}$	304
I	baseline	parallax + 0.05 ${ϵ}_{\pi }\le 0.05$	$-{15.6}_{-2.2}^{+2.2}$	${5.3}_{-0.7}^{+0.8}$	${5.9}_{-0.3}^{+0.4}$	${0.334}_{-0.014}^{+0.015}$	${0.263}_{-0.012}^{+0.012}$	304
J	with $V_{LOS}$ constraint	parallax + 0.03	$-{11.1}_{-1.3}^{+1.3}$	${3.2}_{-0.5}^{+0.5}$	${5.8}_{-0.3}^{+0.3}$	${0.326}_{-0.012}^{+0.013}$	${0.253}_{-0.009}^{+0.010}$	396
K	with $V_{LOS}$ constraint	parallax + 0.03 ${ϵ}_{\pi }\le 0.05$	$-{8.5}_{-1.6}^{+1.5}$	${2.5}_{-0.6}^{+0.6}$	${5.3}_{-0.4}^{+0.4}$	${0.339}_{-0.014}^{+0.014}$	${0.265}_{-0.011}^{+0.013}$	304
Above Galactic plane sample
L	baseline	parallax + 0.03	$-{8.8}_{-3.1}^{+3.2}$	${2.3}_{-1.1}^{+1.1}$	${5.1}_{-0.4}^{+0.4}$	${0.330}_{-0.015}^{+0.015}$	${0.258}_{-0.012}^{+0.013}$	263
Below Galactic plane sample
M	baseline	parallax + 0.03	$-{5.4}_{-6.0}^{+6.1}$	${3.3}_{-1.1}^{+1.1}$	${8.6}_{-0.9}^{+1.0}$	${0.317}_{-0.019}^{+0.022}$	${0.235}_{-0.016}^{+0.017}$	133
1 baseline: ${\Theta }_0=233.6$ km s-1, $d\Theta /dR=-1.34$ km s-1kpc-1, $R_0$ = 8.122 kpc, $(u_{\odot },v_{\odot },w_{\odot })=(11.1,12.2,7.1)$ km s-1.

Table 2. Table 2: Kinematical Results for the GUMS sample

Solution type	$U_s$	$V_s$	$W_s$	${\sigma }_{{\mu }_l}$	${\sigma }_{{\mu }_b}$	$N$
	(km s-1)	(km s-1)	(km s-1)	(mas yr-1)	(mas yr-1)
no $V_{LOS}$ constraint	$-{3.5}_{-4.7}^{+5.0}$	$-{28.9}_{-2.2}^{+2.1}$	$-{1.9}_{-1.0}^{+1.1}$	${1.181}_{-0.048}^{+0.049}$	${0.832}_{-0.033}^{+0.034}$	331
with $V_{LOS}$ constraint	${0.6}_{-2.0}^{+1.9}$	$-{30.5}_{-1.4}^{+1.4}$	$-{2.0}_{-1.0}^{+1.0}$	${1.176}_{-0.049}^{+0.042}$	${0.830}_{-0.033}^{+0.040}$	331

Table 3. Table 3: List of OB candidates

Gaia Id	GALEX Id	long.	lat.	$\pi$	${\mu }_l$	${\mu }_b$	${ϵ}_{\pi }$	${ϵ}_{{\mu }_l}$	${ϵ}_{{\mu }_b}$	$G_{B0}$	$G_{R0}$	$FU{V}_0$	$NU{V}_0$	${ϵ}_{FU{V}_0}$	${ϵ}_{NU{V}_0}$	Sep.	$E_{B-V}$
				(mas)	(mas yr-1)		(mas)	(mas yr-1)								(“)
3322616700534957824	6381084043318920709	200.3039	-9.6691	0.043	0.417	-0.037	0.061	0.100	0.085	14.457	14.784	14.650	14.477	0.101	0.060	0.183	0.540
3321918987392153088	6381048937329986770	200.8701	-8.6551	0.113	0.156	0.123	0.051	0.079	0.079	14.303	14.504	14.800	14.709	0.138	0.076	0.392	0.486

Equations

$${\mu }_l=(V_1\cos l-U_1\sin l)/(Kd)$$

$${\mu }_b=(W_1\cos b-(U_1\cos l+V_1\sin l)\sin b)/(Kd)$$

$$V_{LOS}=W_1\sin b+(U_1\cos l+V_1\sin l)\cos b)$$

$$U_1=U_s\cos \beta +(V_s+\Theta (R))\sin \beta -u_{\odot }$$

$$V_1=-U_s\sin \beta +(V_s+\Theta (R))\cos \beta -v_{\odot }-{\Theta }_0$$

$$W_1=W_s-w_{\odot }$$

$$\ln L=-\frac{1}{2}\sum _i\left(\frac{({\mu }_{l,i}-{\mu }_{l,i}^{model}{)}^2}{({\sigma }_{{\mu }_l}^2+{ϵ}_{{\mu }_l,i}^2)}+\frac{({\mu }_{b,i}-{\mu }_{b,i}^{model}{)}^2}{({\sigma }_{{\mu }_b}^2+{ϵ}_{{\mu }_b,i}^2)}\right)$$

Full text

While spatially encompassing the outer spiral arm of the Galaxy, this structure is not a spiral arm. Its explanation as the Milky-Way warp is equally unsatisfactory. The structure’s vertical extent, mean kinematics and asymmetry with respect to the plane indicate that its origin is more akin to a wobble generated by a massive satellite perturbing the Galaxy’s disk. The mean stellar ages in this outer structure indicate the event took place some 200 Myr ago.

Galaxy: disk, kinematics and dynamics — catalogs — surveys

††journal: ApJ††facilities: GALEX, Gaia, LAMOST

1 Introduction

Discovery of young stars in regions of low gas density — such as the outskirts of our Galaxy or that of the Magellanic Clouds, as well as the Leading Arm of the Magellanic Stream or the Stream itself — compels one to explain how such star formation occurs. Presumably, the role of major dynamical interactions is critical in triggering such episodes. For instance, the interaction of the Clouds with the Milky Way is still far from understood in its complex hydrodynamical and gravitational aspects (Pardy et al., 2018; Tepper-Garcia et al., 2019; Fox et al., 2019), while toward the anticenter there is evidence of the interaction of the Sagittarius dwarf galaxy with the Milky Way’s disk as revealed by the phase-space structure of disk stars (Antoja et al., 2018; Tian et al., 2018; Cheng et al., 2019; Bland-Hawthorn et al., 2019; Laporte et al., 2019). Searching for young stars in such regions can thus provide important observational constraints on both the interaction and the star-forming process.

With this motivation in mind, we conducted a pilot search of such candidates using all-sky surveys in the UV, optical, and IR, in combination with proper-motion measures, when available (Casetti-Dinescu et al., 2012). That study provided a selection procedure and candidates for spectroscopic follow-up in the Leading Arm of the Magellenic Stream and in the outskirts of the Large Magellanic Cloud (LMC). Subsequent spectroscopic studies hinted at the presence of such stars in the Leading Arm and in the extended disk of the LMC (Casetti-Dinescu et al., 2014; Moni Bidin et al., 2017; Zhang et al., 2017). Critically lacking at that time were sufficient numbers of precise proper motions, needed to confirm the implied coherent structures of young main-sequence stars. With the release of Gaia data release 2 (DR2) (Gaia Collaboration et al., 2018a), young stars in the Leading Arm were shown to be runaway Milky-Way disk stars (Zhang et al., 2019), while some stars in the outskirts of the LMC’s disk were confirmed as having formed in situ (Casetti-Dinescu et al., 2018). Here we present the results of a new search for young, OB-type star candidates using the Galaxy Evolution Explorer GR6/7 (GALEX) (Bianchi et al., 2017) and Gaia DR2 data.

Using a methodology that combines photometry, proper motions and parallaxes, we isolate candidate OB-type stars that are kinematically cold and spatially correlated. Doing so, we identify a structure that resides between $l=\sim 120\arcdeg-200\arcdeg$ near the Galactic plane. While partially overlapping with the outer spiral arm (Reid et al., 2014), our structure is unlikely to be part of it. We characterize it spatially and kinematically, and conclude that the most likely interpretation of this newly-found structure is as a wobble of the disk induced by the passage of a massive satellite through it.

Over the entire sky, the only other significant structures revealed by our detection criteria — geared to select young, kinematically cold stars at a given distance — are the Magellanic Clouds.

2 Sample Selection

2.1 Rationale

Our selection strategy is guided by the science objectives: finding extremely blue objects over as large an area as possible. Furthermore, we seek kinematically cold samples of stars — an indication they were born “locally” — with low mean proper motions meaning they are also distant. The combination of GALEX with Gaia DR2 provides the (nearly) full-sky measures needed to accomplish this, i.e., UV photometry from GALEX and optical photometry, positions and proper motions from Gaia.

Note that we do not start by trimming Gaia DR2 using parameters such as e.g., astrometric-excess-noise, photo-BP-RP-excess-factor as do many analyses that make use ofGaia DR2 data. Instead we choose to check these parameters at the end of the selection process. Also, we do not work with individual 3D velocities in our selection, as distance errors propagate into these and can blur the kinematical coldness that we seek, especially for distant objects. We will thus rely on proper motions directly to detect kinematically cold samples.

2.2 Catalog Matching

Gaia DR2 positions were updated with proper motions to an approximate mean epoch of 2008 for GALEX. The entire GALEX All Sky Survey (AIS) catalog with field radius $<0.55\arcdeg$ was then matched by positions with Gaia DR2, using a $3\arcsec$ tolerance match, and keeping the nearest match in cases where multiple matches were found. The resulting cross-match list includes 29,861,012 objects. In Figure 1 we show the distribution of the separations between GALEX and Gaia DR2; the peak is at $\sim 0.\arcsec 55$ with a long tail to larger values, justifying our $3\arcsec$ matching radius.

2.3 Reddening Correction

GALEX magnitudes are corrected for reddening using the prescription given by Bianchi et al. (2017), namely $A_{FUV}=8.06\times E(B-V)$ and $A_{NUV}=7.95\times E(B-V)$, where $E(B-V)$ is from Schlegel et al. (1998) (hereafter, SFD98). ForGaia DR2 magnitudes, we adopt the reddening procedure described in Gaia Collaboration et al. (2018d) with $E(B-V)$ values on the Schlafly & Finkbeiner (2011) scale. This is an iterative procedure, with coefficients determined in Gaia Collaboration et al. (2018d). In Figure 2, we show the color correction in three color indices as a function of reddening $E(B-V)$ (on the SFD98 scale). The color $(FUV-NUV)$ shows the smallest variation in correction with reddening, a fact already noted by Bianchi et al. (2017), and makes this color well-suited to be among the selection criteria. Other colors that combine the UV with the optical domain, such as e.g., $(NUV-G_{B})$, although being good discriminants for various stellar populations, are extremely sensitive to the reddening. Therefore, any errors in the reddening or in the correction itself will convey large errors to the color indices, thus prohibiting their use in exploration of regions with large reddening. For this reason, we prefer to use the $(FUV-NUV)$ color, even if the need for $FUV$ measures substantially limits our sample. Specifically, GALEX has $FUV$ magnitudes only for a few percent of the entire catalog at nominal colors $(G_{B}-G_{R})>0$. However, at the blue end, where $(G_{B}-G_{R})\leq 0$, roughly 20% of the objects have $FUV$ magnitudes.

2.4 Empirical Definition of the Color Domain using Various Stellar and Extragalactic Objects

Our combined GALEX/Gaia DR2 catalog was matched with various lists of specific objects as classified from spectroscopy or variability studies. Matching was done either by position or by Gaia DR2 identifier if the list in question had previously been matched with Gaia DR2. The two extragalactic lists used here are the Large Astrometric Quasar Catalog 4 (LAQC4) (Gattano et al., 2018), and the AGN catalog based on WISE photometry (Secrest, N. J., et al., 2015) as previously matched with DR2 (Gaia Collaboration et al., 2018a). Other lists used are RR Lyrae stars as classified by Gaia DR2 (Clementini et al., 2019) and WISE (Gavrilchenko et al., 2014), and white dwarfs and OB stars. The white-dwarf sample is from Gentile Fussilo et al. (2015) where we used only those objects with SDSS and BOSS spectra. The OB samples are from Maiz Apellanitz et al. (2016, 2019) and Liu et al. (2019). The former sample of O-type stars is at low latitudes, and thus only a handful of stars are matched with our catalog. From the latter study we have included only the OB main-sequence class stars. Finally, for completeness, we also include a subdwarf candidate sample derived from Gaia DR2 photometry and astrometry by Geier et al. (2019). In Figure 3 we show $(FUV-NUV)_{0}$ as a function of $(G_{B}-G_{R})_{0}$ for these various samples.

The Gaia DR2 RR Lyrae sample shows a subset of objects at $(G_{B}-G_{R})_{0}\sim 1.3$. We have checked a few of these objects by hand and found them to be galaxies present in the LEDA catalog (Paturel et al., 2005). In our GALEX/Gaia DR2 merged catalog matched with the Gaia DR2 RR Lyrae sample, only about 1% of such purported RR Lyrae stars have these unusual colors; we consider these misclassified RR Lyrae stars. The WISE RR Lyrae sample does not show such a population. This sample is shallower than the Gaia DR2 one, with a limiting magnitude of $G_{R}\sim 15.5$.

2.5 First Selection: High-quality Measurements

Precise proper motions are required in our analysis to trace kinematically cold substructures. Therefore, we first trim the merged GALEX/Gaia DR2 catalog by proper-motion uncertainty: specifically, we retain only objects with proper-motion uncertainty $\sqrt{(\epsilon_{\mu_{\alpha}}^{2}+\epsilon_{\mu_{\delta}}^{2})}\leq 0.2$111in what follows, $\mu_{\alpha}$ represents $\mu_{\alpha}cos~{}\delta$, and $\mu_{l}$ represents $\mu_{l}cos~{}b$ mas yr*-1*. Here, $\epsilon_{\mu_{\alpha}}$ and $\epsilon_{\mu_{\delta}}$ represent individual proper-motion uncertainties from Gaia DR2. The value of 0.2 mas yr*-1* represents approximately 10 km s*-1* at a distance of 10 kpc. Thus, this is a velocity-error limit for distant objects that we adopt in order to search for cold kinematical structures. This trimming is effectively a cut at a faint magnitude limit, retaining objects in Gaia DR2 with a good SNR, both photometrically and astrometrically. This cut corresponds roughly to limiting magnitudes of $G\sim 18$, $G_{B}\sim 18.3$, $G_{R}\sim 17.5$. We also trim in magnitude uncertainties in $FUV$ and $NUV$. Inspecting the distribution of the estimated uncertainties with magnitude, as illustrated in Figure 4, we adopt the following limits: $\epsilon_{FUV}\leq 0.2$ and $\epsilon_{NUV}\leq 0.1$. This cut effectively imposes magnitude limits of $FUV\sim 21.3$ and $NUV\sim 21.0$. These preliminary cuts, in proper-motion and magnitude uncertainties, yield a sample of 4,837,190 objects.

2.6 Second Selection: Blue Candidates

Our selection of blue candidates is based on cuts in both colors shown in Fig. 3. The first cut is for $(G_{B}-G_{R})_{0}\leq 0.0$. According to the models in the simulation GaiaSimu Universe Model Snapshot (Robin et al., 2012), this limit corresponds to objects with $T_{eff}\geq 9000$K. While this effective temperature limit corresponds to early A-type stars, the empirical plots shown in Fig. 3 indicate that OB-type stars are well represented by this limit. The discrepancy may be in part due to the uncertain reddening correction in regions of large extinction. To further clean our sample we trim in UV colors as well. We consider two limits: a very blue sample at $(FUV-NUV)_{0}\leq 0.23$ which corresponds to stellar types earlier than B8V (Vennes et al., 2011) or $T_{eff}\geq 12,500$K, and another one at $(FUV-NUV)_{0}\leq 1.0$, which is more in line with the $(G_{B}-G_{R})_{0}\leq 0.0$ cut, i.e., for $T_{eff}\sim 9000$ K. In Fig. 3 these limits define regions that are predominantly populated by OB-type main sequence, subdwarfs and white dwarfs with very little contamination from other objects.

Our aim being more distant structures, we also discard bright objects, retaining stars with $G_{R0}\geq 12.0$. In Figure 5 we show distributions of the bluest sample in Galactic coordinates, proper motions (transformed to Galactic coordinates), and also plot proper motions and parallaxes versus longitude and latitude. A total of 11,187 candidates are in the bluest sample ($(FUV-NUV)_{0}\leq 0.23$). A similar plot for the $(FUV-NUV)_{0}\leq 1.0$ sample is shown in Figure 6. There are 33,082 objects in this sample.

Inspecting Fig. 5, we see two kinematically cold clumps in proper-motion space: one at $(\mu_{l},\mu_{b})\sim(-0.9,1.6)$ mas yr*-1*, the other at $(\mu_{l},\mu_{b})\sim(0.5,0.0)$ mas yr*-1*. The first corresponds to the Magellanic Clouds, the presence of which is also seen in the plots of $\mu_{l}$ and $\mu_{b}$ vs $l$ and $b$ at $(l,b)\sim(300\arcdeg,-40\arcdeg)$. The second proper-motion clump is elongated along $\mu_{l}$ and shows a strong variation in $\mu_{l}$ with $l$. It is located at $l\sim 120\arcdeg-210\arcdeg$ and within $|b|\leq 15\arcdeg$. We will refer to this region in Galactic coordinates as our Region of Interest (ROI). In the bottom, left plot of Fig. 5, the parallax distribution also shows two prominent clumps: one at $l\sim 300\arcdeg$ and $\pi\sim 0$ mas, corresponding to the Magellanic Clouds, and another at $l~{}\sim 120\arcdeg-210\arcdeg$ and $\pi\sim 0.2$ mas, corresponding to our ROI.

Fig. 6 displays the sample with $(FUV-NUV)_{0}\leq 1.0$ and also shows the same two proper-motion clumps. Here the clump at $(\mu_{l},\mu_{b})\sim(0.5,0.0)$ mas yr*-1* is more extended. Likewise, the parallax distribution in the ROI is also more extended than in the bluest sample. We conclude this sample is likely more contaminated with foreground populations. In what follows, we will focus on the bluest sample.

Thus, over the entire sky covered by the merged GALEX/Gaia DR2 catalog, and within the faint magnitude limit of $G_{R}\leq 17.5$, we have identified two structures populated by very blue stars that are also kinematically cold and with low mean proper motion. The blue color hints at young ages, and this is definitely the case for the Magellanic Clouds. The cold kinematics together with low mean bulk proper motion hints at large distances. However, it is possible to have more nearby structures with intrinsically low velocity dispersion and with a systemic motion not too different from that of the sun. Therefore, we will explore the distances of these candidates as indicated by their Gaia DR2 parallaxes.

2.7 Third Selection: Distant Objects

To further refine our search for distant structures, we eliminate foreground objects such as white dwarfs and subdwarfs. To do so, we plot the parallax as a function of $G_{R0}$ magnitude as shown in Figure 7. A preliminary trimming is done by hand, selecting only the sequence with low average parallax that varies slowly with magnitude. This sample is then fit with a linear function and the entire sample is trimmed within $0.2$ mas of the fitted line. The resulting sample of objects retained is highlighted in blue in Fig. 7. It consists of 4999 objects. Within this sample we focus on the region at low galactic latitude, our ROI at $120\arcdeg\leq l\leq 210\arcdeg$ and $|b|\leq 15\arcdeg$. The sample contains 664 objects inside the ROI.

2.8 Fourth Selection: Proper-motion trimming in the ROI

Focusing solely on the ROI candidates, we attempt to isolate the cold component originally seen in the proper-motion vector point diagram. We utilize the run of $\mu_{b}$ as a function of parallax, as this component shows less scattered than $\mu_{l}$ and is not complicated by any variation with Galactic latitude or longitude. The relationship is shown in Figure 8. Proper motions are tight at low parallax, however starting at parallax $\sim 0.25$ mas they appear to scatter somewhat, losing the “coldness” property we seek. For this reason, we implement a parallax cut, keeping only those objects with parallaxes $\leq 0.25$ mas. Afterward, we further isolate the sample by trimming by proper motions, using an iterative procedure. We first plot $\mu_{b}$ as a function of $l$, fit with a constant and discard objects outside $2.5\sigma$ from the fit. Next, we plot $\mu_{l}$ as a function of $l$, fit with a second-order polynomial and discard objects outside $2.5\sigma$ from the fit. This process is repeated two more times, leaving a sample of 396 objects.

At this point we consider some of the Gaia catalog parameters often used to cull out poor quality data. We check the values of the astrometric-excess-noise and find only four objects with values slightly larger than those prescribed in Gaia Collaboration et al. (2018b) at given $G$ magnitude. We choose not to discard these objects since the excess noise in the astrometric solution may be due to the presence of a companion; young OB-type stars are highly likely to be binaries. From 396 objects, 332 have astrometric-excess-noise$~{}=0$. We also check the photo-BP-RP-excess-factor of our selected objects and compare with the recommended relation photo-BP-RP-excess-factor$~{}\leq 1.3+0.06\times(G_{B}-G_{R})^{2}$ (Gaia Collaboration et al., 2018b). Only one object has a value slightly larger than the recommended limit, specifically 1.365 versus 1.342. We choose not to discard this object. The median UV photometric errors are $\epsilon_{NUV}=0.044$ and $\epsilon_{FUV}=0.076$ mag. The distribution of the relative parallax error $\epsilon_{\pi}/\pi$ peaks at 0.16 and has a long tail toward higher values. This reflects the fact that our sample includes distant stars, where Gaia DR2 parallaxes are rather uncertain. We must therefore proceed very carefully when deriving distances from the parallaxes.

3 Placing the Newly Found Structure into the Milky-Way Context

3.1 Comparison with Gaia Universe Model Snapshot

To better understand our sample in the context of the Milky Way, we make use of the Gaia Universe model snapshot version 10 (GUMS) presented in Robin et al. (2012). This model includes a parametrization of the Galactic disk warp and flare, and two nearby spiral arms. However, no distant spiral arms are present. We run one simulation of the model within our ROI and $G_{R}\leq 19.0$. Stars are then selected spatially, in a manner that mimics the area coverage of our merged catalog. This is accomplished by discarding any GUMS object that is more than 300 arcsec from its nearest counterpart in the GALEX/Gaia DR2 merged catalog. This ensures that the model data follow exactly the spatial selection imposed by the GALEX observations within the ROI. Using the absorption $A_{V}$ in the model, we convert it to the Schlafly & Finkbeiner (2011) scale and then apply the same reddening-correction scheme as for our observations (see 2.3) to obtain de-reddened magnitudes. We then apply the same $(G_{B}-G_{R})_{0}$ color cut used for the observations. As $FUV,NUV$ bands are not present in GUMS, we use the effective temperature to mimic the $(FUV-NUV)_{0}\leq 0.23$ cut applied to the observations. Specifically, we select objects with $T_{eff}\geq 12,500$ K (see 2.6). This is followed by the $G_{R0}\geq 12.0$ bright magnitude cut, as well as parallax versus magnitude and maximum parallax cuts, all identical to what was applied to the observed sample (2.6). The cut in proper-motion uncertainty applied to the observational sample effectively introduced a “fuzzy” faint magnitude limit of $G_{R0}\sim 15.5$. This we mimic in the GUMS data by imposing a linear probability distribution and random number generator to exclude stars over a limited range in $G_{R0}$, specifically from $14.25\leq G_{R0}\leq 16.25$. All stars fainter than 16.25 are excluded. (The initial GUMS data set had been retained down to $G_{R0}=19.0$.) We purposely do not perform the proper-motions cuts to the model data, as we suspect significant differences in the proper-motion distributions, between model and observations. Instead, we will compare with the proper-motion distribution of our observed candidate sample before the proper-motion cuts were made. Note that for the model data, the “measured” proper motions and parallaxes are free of measuring errors.

In Figure 9 we show the Galactic-coordinate and proper-motion distributions for the model and observations. Spatially the observations show a distinct edge at $b\sim 8\arcdeg$, while the model extends in $b$ to the boundary of the ROI. Likewise, in $l$ the observations lack candidates between $l=163\arcdeg$ and $140\arcdeg$ below the plane while the model does not. The proper-motion panels show that the observations have a much tighter distribution than does the model, in spite of the measuring errors present in the observations and not in the model.

In Figure 10 we show the proper motions versus $l$ and $b$, for model and observations. The larger scatter of the model compared to the observations is once again apparent in all plots of Fig. 10. The trend of $\mu_{l}$ with $l$ is similar for model and observations, however the model’s proper motions are shifted by $\sim 1$ mas yr*-1* to more positive values at all longitudes. The run of $\mu_{b}$ versus $l$ also appears shifted toward negative proper motions in the model compared to the observations. The $\mu_{b}$ offset is relatively easy to interpret, conceptually. At these low latitudes, the average $\mu_{b}$ of disk stars will represent the reflex solar motion perpendicular to the disk. The model has an average $\mu_{b}=-0.29\pm 0.07$ mas yr*-1*, while the observations have an average of $\mu_{b}=-0.01\pm 0.02$ mas yr*-1* (for 484 objects with no proper-motion trimming). This points to a systemic upward motion of our sample of OB-type candidates. We will discuss this further in Section 4. Regarding $\mu_{l}$, the apparent offset between model and observations may be due to the specific values of solar motion and LSR rotation adopted by the GUMS model. GUMS uses the Schonrich et al. (2010) solar peculiar motion, and an LSR rotation $\Theta_{0}=226$ km s*-1*. More recent work (e.g., Mróz et al., 2019) indicates values somewhat higher, of the order of $\Theta_{0}\sim 230$ km s*-1*. Alternatively, the offset could arise from streaming motion of the OB candidate stars in the azimuthal and radial directions, similar to those found by other recent studies for A-type and OB-type stars (Harris et al., 2019; Cheng et al., 2019).

For our OB-candidate sample we obtain line-of-sight (LOS) velocities from the LAMOST DR4 v2 (Zhao et al., 2012) survey. Only a fraction have LOS velocity measures; 166 out of the 484 candidates not trimmed by proper motion. We plot these as a function of $l$, together with the model data, in Figure 11. In this case, the agreement between model predictions and observations is good.

From this, we conclude that our sample of OB-type candidates has roughly disk-like motion when compared with a generic model of the Galaxy. However some discrepancies are present in the proper motions, in terms of small offsets and the notably tight proper-motion dispersion of the observations compared to the model.

3.2 The Newly Found Structure Compared with Observations of the Outer Spiral Arm

We show the spatial distribution of our blue, kinematically cold and distant sample in Figure 12 top and middle panels. Here, we consider the sample trimmed of proper-motion outliers (see 2.8). The top panel’s gray background indicates the area covered by the merged GALEX/Gaia DR2 catalog. The background of the middle panel shows the SFD98 map of reddening, represented as a color scale. The bottom panel shows the run of $E(B-V)$ at the location of each of our candidates as a function of longitude. We display the SFD98 values here, but remind the reader that our reddening correction used the updated Schlafly & Finkbeiner (2011) scale (see 2.3). Although our candidates reside in regions of high reddening, it is not the case that they reside exclusively in these regions. That is, there are certainly regions covered by the GALEX/Gaia DR2 where reddening is high but where no such candidates are found. This supports our search strategy as genuinely finding blue OB-type stars rather than finding artifacts of reddening correction. The newly found structure predominantly resides at low latitudes, $|b|\leq 10\arcdeg$. Its outermost extent is better defined above the plane, where the structure fades abruptly at $b\sim 8\arcdeg$. Given the area coverage of the catalog, the presence of the structure across the inner few degrees of the plane is unknown. Limits in longitude are harder to infer due to the area coverage of the merged catalog. Nonetheless, below the plane there seems to be a lower limit of $l\sim 163\arcdeg$, while above the plane the apparent upper limit is $l\sim 170\arcdeg$.

Let us compare our sample of OB candidates with other tracers in the outer arm: the high mass star-forming regions (HMSFR) from (Reid et al., 2014; Quiroga-Nuñez et al., 2019), the molecular clouds from Du et al. (2016), and the open clusters from Molina Lera et al. (2018). In Figure 13 we present these tracers’ distribution in Galactic coordinates as well as their proper motions, $V_{LSR}$, and parallax values as functions of longitude.

LOS velocities for our candidates are from LAMOST DR4 (see 3.1) and have been transformed to the LSR reference frame using the Schonrich et al. (2010) Solar peculiar motion. The two open clusters from Molina Lera et al. (2018) are Meyer 2 and BDSD 42. According to their location on the sky and distance moduli, Molina Lera et al. (2018) suggest that these two clusters are part of the outer arm. The clusters do not have measurements of LOS velocities, nor Gaia DR2-based proper motions according to the recent Gaia-DR2 open cluster catalog by Cantat-Gaudin et al. (2018).

Fig. 13 top panel shows that all of the tracers, excepting our OB candidates, are at very low latitudes and therefore do not (or can not) map any vertical extent of the arm. The run of $\mu_{l}$ and $\mu_{b}$ with longitude shows good agreement between our sample and the four HMSFRs. The LOS velocities expressed with respect to the local standard of rest, $V_{LSR}$, as a function of longitude shows good agreement between the HMSFRs and the molecular clouds. Our candidates display a large scatter, but an overall trend that agrees with the tighter trend shown by the other two tracers. The large scatter may be due to the high incidence of tight binaries for these early-type stars (Sana et al., 2012). Finally, the bottom panel of Fig. 13 shows the distribution of parallaxes compared to those of the HMSFRs as measured and compiled in Reid et al. (2014) and in Quiroga-Nuñez et al. (2019), and to the two open clusters. Parallaxes for the open clusters were derived from the distance moduli determined by Molina Lera et al. (2018). Overall, there is reasonable agreement, although to this point we have not attempted to correct the Gaia DR2 parallaxes for the known offset (Gaia Collaboration et al., 2018b). (We do explore such a correction in Sect. 4.) Overall, our OB candidates appear to conform with the paradigm of the outer arm.

4 Kinematical Analysis

We further explore the nature of the newly found structure by comparing its properties with a mapping of the outer arm combined with a kinematical prediction of the Galaxy’s rotation curve. To this end, we adopt the Reid et al. (2014) spatial description of this arm, their Table 2. The arm has a pitch angle $\psi=13.8\arcdeg\pm 3.3\arcdeg$, a reference radius $R_{ref}=13.0\pm 0.3$ kpc, and a Gaussian width of $0.63\pm 0.18$ kpc. For the rotation curve of the Galaxy we use the recent determination by Mróz et al. (2019) based on $\sim 770$ Cepheids. Thus, $\Theta_{0}=233.6\pm 2.8$ km s*-1*, with a small gradient $d\Theta/dR=-1.34\pm 0.21$ km s*-1* kpc*-1*. This value of the circular rotation velocity at the sun’s location is consistent with the review value of $\Theta_{0}=238\pm 15$ km s*-1* given in Bland-Hawthorn and Gerhard (2016). The sun is located at $R_{0}=8.122\pm 0.031$ kpc (Abuter et al., 2018)222A more recent estimate by the same group (Abuter et al., 2019) is $R_{0}=8.178\pm 0.022$ kpc. This value, as well as the 2018 value are within the extensively reviewed value of $R_{0}=8.2\pm 0.1$ kpc presented in Bland-Hawthorn and Gerhard (2016). We explore a value as large as 8.5 kpc in our tests., and its peculiar motion is $(u_{\odot},v_{\odot},w_{\odot})=(11.1\pm 1.3,12.2\pm 2.1,7.1\pm 0.7)$ km s*-1* (Schonrich et al., 2010). While we list these quantities with their estimated uncertainties, in our analysis we will adopt them as constants; we will show later that varying these parameters has little impact on the results of the kinematical analysis, unless the variations are substantially larger than their formal uncertainties.

We follow the formalism described in Mróz et al. (2019) and Reid et al. (2014) (see also Fig. 8 in Reid et al. (2009). The proper motions and LOS velocity of each star can be expressed as:

[TABLE]

[TABLE]

[TABLE]

where $d$ is the heliocentric distance in kpc, and $K$ = 4.74 km s*-1* kpc*-1* per mas yr*-1*. The expressions $U_{1},V_{1}$ and $W_{1}$ are given by:

[TABLE]

[TABLE]

[TABLE]

where $\beta$ is the angle between the sun and the source as viewed from the Galactic center, and $\Theta(R)=\Theta_{0}+\frac{d\Theta}{dR}(R-R_{0})$. $U_{s},V_{s}$ and $W_{s}$ are the “non-circular” or “streaming” velocity components of each star, in a cylindrical reference frame: $U_{s}$ points toward the galactic center (at the location of the star), $V_{s}$ is along Galactic rotation and $W_{s}$ is perpendicular to the Galactic plane, positive toward the North Galactic Pole.

We solve for the mean $U_{s},V_{s},W_{s}$ of our entire sample. In other words, we will determine the systemic streaming motion of our sample as a departure from the underlying circular disk motion. Simultaneously, we also solve for the intrinsic proper-motion dispersion of our sample $\sigma_{\mu_{l}}$, $\sigma_{\mu_{b}}$.

The likelihood function we use is:

[TABLE]

where $\epsilon_{\mu_{l}},\epsilon_{\mu_{b}}$ are individual Gaia proper-motion errors, and the summation is over our sample of 396 OB candidates. The best-fit parameters are found by maximizing the likelihood function, with uncertainties derived using the Markov chain Monte Carlo technique described in Foreman-Mackey et al. (2013). Uncertainties represent $68\%$ confidence range of marginalized posterior distributions. To begin with, we use only the proper motions to constrain the model, and not the LOS velocities (which are available only for a subsample of our stars). For distances we test two different assumptions. In the first we adopt distances from the outer spiral arm description of Reid et al. (2014). Specifically, for each object with a given longitude we determine an in-plane distance as if it were in the spiral arm, then use the latitude to de-project that distance above the plane ($d=d_{in-plane}/\cos~{}b$). For in-plane distances we randomly draw a value from the 0.63 kpc half-width of this arm model. No uncertainties in distances are considered in this method which we will refer to as “spiral arm distance”. Under the second assumption, we adopt distances derived from the Gaia DR2 parallax measures. A pre-correction is made for the systematic offset that has been found in the Gaia parallax zero point and we explore two values for this offset, 0.03 and 0.05 mas (Gaia Collaboration et al., 2018b; Schonrich et al., 2019). For each of these values we test two separate methods for handling spurious measures. In the first, if parallaxes are negative or zero we adopt a large fixed distance of 1000 kpc. In the second, we eliminate those stars with parallax uncertainties larger than 0.05 mas, which has the effect of discarding all objects with negative parallaxes. We will refer to these as tests using “parallax distances”.

Our maximum-likelihood results are summarized in Table 1. The first column indicates the run. The second column, “Configuration” indicates what, if any, modification to the baseline model parameters is assumed. The third column indicates what assumption was used concerning stellar distances. The subsequent columns show the resulting best-fit values for the kinematical parameters. In the first five runs we present results with the spiral-arm distances: baseline plus four trials with $R_{0},\Theta_{0},d\Theta/dR$ being modified. The low limit of $\Theta_{0}$ was motivated by the value used in the GUMS model (see Section 3.1). Among these tests, only $V_{s}$ changes significantly ($\sim 8\sigma$) from its baseline value, and this is due to the change in the rotation velocity gradient. $R_{0}$ and $\Theta_{0}$ have little influence on the results. On the whole, the results indicate a streaming outward motion and a streaming vertical, upward motion for our sample. We have also performed a few tests changing the solar peculiar motion. In an attempt to force the best-fit streaming velocities close to zero, we had to adopt $u_{\odot}=20$ km s*-1*, a value unreasonably large. We therefore discard modifications of the solar peculiar motion as an explanation for the streaming velocities of our sample.

Changing the assumption regarding the distances by using Gaia parallaxes (run F) leads to the same conclusion: a net outward radial motion and a net vertical motion, with values agreeing to within errors of those obtained under the spiral-arm distance assumption. Different parallax zero-point corrections, as well as discarding objects with large parallax errors (runs G through I), still give consistent results with the spiral-arm runs for $U_{s}$ and $W_{s}$. These results are surprising, especially along the radial direction, since no LOS velocities were used in the fit. To check whether the information from $V_{LOS}$ could change this outcome, we provide another test where we incorporate the velocities in the maximum-likelihood process. We do so in the following way. We plot $V_{LOS}$ as a function of longitude for our sample and fit it with a line, which is a reasonably good approximation (see Fig. 11). From the slope, intercept and standard deviation of this fit we generate an artificial $V_{LOS}$ value for every star in the sample, drawn from a normal distribution described by the linear fit to the observed $V_{LOS}$ values. These $V_{LOS}$ values are then used as input together with the proper motions in the maximum likelihood procedure. We use as individual velocity errors $3$ km s*-1*, the median LAMOST formal error of the sample. Results from this fit are listed in Tab.  1 for parallax distances of the entire sample, and of the sample with parallax error $\leq 0.05$ mas, (runs J and K). While $W_{s}$ is pretty much unchanged, $U_{s}$ changes to lower absolute values, but these are still significantly different from zero.

An unstated assumption up to now has been that the stars seen above the plane and below the plane are from the same structure. To explore this assumption we split the sample by latitude obtaining an above- and a below-the-Galactic-plane sample. The solutions for these two separate samples are the last runs shown in Tab. 1. Both show a net, significant upward streaming motion, with the sample below the plane displaying a larger value than that of the above sample. The radial outward motion is still present at lower magnitudes and with larger uncertainties, especially in the below sample. The poorer constraint of $U_{s}$ is due to both a smaller sample size and to a more limited longitude range of each of the above and below samples when compared to the entire sample. The separation by latitude also forces the samples to encompass slightly different longitude ranges. The $U_{s}$ solution is particularly susceptible to the longitude range covered, hence the more uncertain results. Nevertheless, $U_{s}$ still indicates streaming, radially outward motion.

Finally, we test our fitting procedure using the GUMS simulation data set (Section 3.1), trimmed accordingly to mimic our observed sample. Recall that in Section 3.1, we did not trim the selected GUMS sample in proper motions. Here, we do so in order to discard outliers in proper-motion space: we keep objects with $-5.0<\mu_{l}<5.0$ and $-3.0<\mu_{b}<3.0$ mas yr*-1*. We perform two fits, first with no $V_{LOS}$ input and then with its input. Formal measuring errors for proper motions and LOS velocities are zero in the simulation, but in the fit we have set these to very small values to avoid zero divisions. Results are listed in Table 2. We note that the GUMS model has slightly different Galaxy parameter values, such as $R_{0}=8.5$ kpc, $\Theta_{0}=226$ km s*-1*, and a specific asymmetric drift for various stellar populations (Robin et al., 2012). The solar peculiar motion is that from Schonrich et al. (2010), the same as the one used in our kinematical analysis. The results of the fits indicate that $U_{s}$ and $W_{s}$ for this sample are not significantly different from zero, while $V_{s}$ is. Thus, there is no streaming motion in either radial or vertical direction, while along the azimuthal direction the lagging is likely due to the values adopted by the GUMS model for the asymmetric drift. As already seen in Section 3.1 and in Fig. 10, the intrinsic proper-motion dispersions derived here from the fit are a factor of 3 to 4 times larger in the GUMS simulation compared to the observations.

We will adopt as our final solution the one using observed parallaxes with 0.03 mas correction offset, and keeping objects with parallax errors $\leq 0.05$ mas (i.e., run H in Tab. 1). Note that the uncertainties listed in this Table do not include a contribution from the uncertainty in the parallax. To estimate this, we repeat this run using parallaxes drawn from a Gaussian distribution with $\sigma$ equal to the Gaia parallax uncertainty estimate. From a set of 200 repeats, we measure the additional scatter in the fitted parameters due to parallax errors. This scatter we add in quadrature to the errors listed in Tab. 1. With this done, our final values are: $(U_{s},V_{s},W_{s})=(-16.0\pm 2.5,4.6\pm 0.9,5.7\pm 0.4)$ km s*-1* and $(\mu_{l},\mu_{b})=(0.337\pm 0.058,0.263\pm 0.012)$ mas yr*-1*. $V_{s}$ is sensitive to certain input parameters, including the rotation velocity gradient and the specific solar motion, so its uncertainty is probably underestimated. Nonetheless, it is close to zero, and no case for a robust streaming motion along this direction can be made. $U_{s}$ shows a preference for outward streaming motion; however its formal uncertainty is still probably underestimated. Specifically, $U_{s}$ and $V_{s}$ are strongly correlated: if $V_{s}$ is closer to zero or becomes negative, then $U_{s}$ will decrease its absolute value as run “E” in Tab. 1 indicates. Furthermore, inclusion of LOS velocities also changes $U_{s}$ toward lower absolute value. Based on this, we estimate the range of $U_{s}$ is probably between $\sim-16$ and -8 km s*-1*. $W_{s}$ has the most robust value at $5.7\pm 0.4$ km s*-1*, clearly indicating an upward streaming motion.

We now estimate the velocity dispersion indicated by our data. We do so for the latitude proper-motion dispersion, consistently the lower of the two components. The average distance to our sample differs slightly between the above and below sample. Using only objects with $\epsilon_{\pi}\leq 0.05$ mas the mean parallax for the sample above the plane is $<\pi_{above}>=0.203\pm 0.003$ mas and for the sample below the plane it is $<\pi_{below}>=0.159\pm 0.008$ mas. This includes a correction of 0.03 mas added to the Gaia parallaxes (see above). Using the proper-motion dispersions from Tab. 1 runs L and M respectively, we obtain velocity dispersions of $\sigma_{W}^{above}=6.0\pm 0.3$ km s*-1*, and $\sigma_{W}^{below}=7.0\pm 0.6$ km s*-1*. The two values are consistent with each other at a $1.5\sigma$ level.

5 Origin of the Young and Kinematically Cold Structure

The colors and distances of our candidate OB stars implies an age range of a few million to about a couple hundred million years old for the structure (see e.g., Fig. 13 in Bland-Hawthorn et al. (2019)). The cold kinematics also suggests a young age. As such, ones first hypothesis might be membership to the outer spiral arm, a place in the outskirts of the Galaxy where it is known that star formation takes place. However, the vertical extent of our structure, as well as small but significant departures from the MW disk motion do not favor such an origin. Indeed, the outer arm as described by Reid et al. (2014) does not appear to have a significant vertical motion.

Vertical motions of the amplitude found here were reported before; see for instance the recent work of Cheng et al. (2019) and Poggio et al. (2018). Cheng et al. (2019) who studied a sample of $\sim 12,000$ OB LAMOST stars demonstrate in their Fig. 3 that the mean $V_{z}$ (equivalent to our $W$) is positive and of the order of 7 km s*-1* in the direction of the anticenter. However their radial velocity $V_{R}$ (equivalent to our $-U_{s}$) is more ambiguous in this region of the Galaxy, ranging from positive to negative values between longitudes of $150\arcdeg$ and $180\arcdeg$. They too conclude that this kinematical feature does not align with the outer spiral arm. Poggio et al. (2018) use Gaia DR2 to study photometrically-selected large samples of upper main sequence stars and giants. Their kinematical maps, which sample well quadrants 2 and 3 of the Galaxy, indicate an upward motion of amplitude $\sim 5$ km s*-1* in both stellar populations. They conclude that this motion is due to the Galactic warp: since the direction sampled is near the line of nodes of the warp, the vertical motion is the largest. To this end, we adopt the recent warp model from Chen et al. (2019) based on $\sim 1300$ Cepheids to check whether our structure can be explained by the warp. We calculate the projection on the sky of the warp at the distance of the outer spiral arm. This projection is shown with a continuous (slanted) line in Fig. 12, top panel. It is clear that the warp cannot explain the sky spatial distribution of our sample: our objects are much farther from the plane than the predictions of the warp model.

Two other plausible explanations remain for our structure: first as an external system such as a satellite galaxy accreted by our Galaxy, and second, as a wobble in the disk of our Galaxy, presumably induced by an interaction/perturbation. If it were to be an external system, it is intriguing to find young stars in it: most MW satellites contain old stellar populations. Furthermore, the mean motion of this system is very similar to that of the MW disk. Thus, the second explanation, as a perturbation or wobble of the Galaxy’s disk, appears much more feasible. D’Onghia et al. (2016) perform N-body experiments to study the impact of orbiting satellites on the Galactic disk. They find that a satellite as massive as the Sagittarius dwarf galaxy can induce the disk to wobble and produce vertical displacements up to 1 kpc and vertical streaming motions as large as 15-20 km s*-1* (see their Fig. 6). In Figure 14 we show the in-plane and perpendicular-to-the-plane spatial distribution of our candidates. In particular, for the above-the-plane sample the distribution sharply ends at 500-700 pc. It ssems improbably that an external satellite would be thus confined in $Z$ but extend nearly $30\arcdeg$ in longitude (see Fig. 12, top). Rather, such a spatial distribution suggests material from the disk being displaced vertically, while nevertheless being confined by the disk’s gravity to a limiting height above the plane.

The assymmetry of the structure with respect to the Galactic plane, both spatially (i.e., the sample below the plane is more distant and more diffuse than the above-the-plane sample) and kinematically (i.e., having slightly different $W_{s}$ values, see Tab. 1), is also characteristic of a disk perturbation induced by a massive satellite (see Figs 1 and 4 in D’Onghia et al. (2016)).

The candidate OB stars could well have formed in the disk and then been displaced by a gravitational perturbation on the disk to their current locations. It takes about 50 million years for an object to move some 250 pc at a constant speed of 5 km s*-1*; integrating in a Galactic potential it takes about the same time to attain the $Z$-displacement seen in the observations for a representative orbit of a star in our sample. Since the ages of the stars are $\leq 200$ million years, this scenario is plausible. This would also imply that the perturbation occurred recently, i.e., within the order of the ages of these stars. Alternatively, gas in the disk could have been displaced to the current locations by a gravitational perturbation, and then form the stars.

We conclude that the newly found structure is most plausibly interpreted as a perturbation of the disk, with the perturber being a rather massive satellite. Neither spiral arms or the Galactic bar are likely to produce such displacements of the disk material and induce star formation at a radius of between 12 and 15 kpc from the Galactic center. Naturally, the satellite perturber that first comes to mind is the Sagittarius dwarf galaxy, which crossed the disk near the anticenter some 500 Myr ago. Sagittarius has been recently invoked to explain the phase-space structure seen in the Gaia DR2 data by Bland-Hawthorn et al. (2019); Laporte et al. (2019); Tian et al. (2018); Antoja et al. (2018). Specifically, Laporte et al. (2019) present Sagittarius - Milky Way interaction models that show that a mass of $6\times 10^{10}$ M*⊙* for the dwarf galaxy can reproduce many of the features found in Gaia DR2 kinematics, and that these features were triggered some 500-800 Myr ago. Similarly, Bland-Hawthorn et al. (2019) show that a satellite with a mass of $3\times 10^{10}$ M*⊙* and an impact parameter of 13 kpc can produce the phase-space spiral structure seen in Gaia, and they time it between 400 and 500 Myr ago. While the timing in the Bland-Hawthorn et al. (2019) analysis is gravitational-potential model dependent, they point out that a subsequent Sagittarius disk crossing will wipe out the phase-space spiral, thus favoring a younger age for this event. Furthermore, our analysis shows that the more distant sample (below the plane) is also more extended in $Z$ than the nearby sample (above the plane). This is also a characteristic of disk corrugations induced by a massive satellite as exemplified in Fig. 21 of Bland-Hawthorn et al. (2019).

Our structure qualitatively fits these scenarios, although a closer inspection of these models in light of our specific findings should be made. If stars in the structure we describe were displaced away from the plane at the time of or after their formation, then the onset of this perturbation must have been within on the order of 200 Myr, i.e., more recently than indicated by the aforementioned studies.

Spectroscopic follow-up of our OB candidates — to confirm their spectral types and ages, and to obtain LOS velocities and possibly abundances — is clearly needed in order to better understand the nature and origin of this structure. Likewise, N-body and hydrodynamic simulations tailored to the Sagittarius impact on the Galactic disk would be helpful in ascertaining the origin of this structure.

6 Summary

We combine the latest releases of the GALEX and Gaia DR2 catalogs to search for young, distant and kinematically cold stars, effectively tracking star formation in the outskirts of our Galaxy. Compared to recent Gaia DR2 studies that use individual 3D velocities that inherently include parallaxes, our analysis focuses on proper motions. These are better measured than parallaxes and are thus better suited to track cold kinematical structure. In this way we are able to push this analysis to greater distances than studies based on individual Gaia DR2 parallaxes.

We identify a structure of $\sim 300$ OB candidate stars extending from $l=120\arcdeg-200\arcdeg$ and $|b|\leq 15\arcdeg$ that shows clumping in proper motions and parallax. Traditional galactic models cannot reproduce this structure. Its mean motion is similar to the disk, however small but significant departures from the disk’s motion are measured. Specifically, the structure has a mean motion perpendicular to the disk of $5.7\pm 0.4$ km s*-1*, and a mean outward radial motion of between 8 and 16 km s*-1*. The velocity dispersion along the least dispersed of its proper-motion axes is estimated to be $6.0\pm 0.3$ km s*-1*. The structure is approximately between 12 and 15 kpc from the Galactic center and extends vertically above the plane to about 700 pc, and to about 1 kpc below the plane. While partly overlapping in properties with the outer spiral arm of the Galaxy, the structure’s vertical spatial extent and kinematics indicate it is not part of this spiral arm. The spatial and kinematical properties of this structure together with the young age of its stars suggest its origin being a perturbation of the disk induced by the passage of a massive satellite within some $\sim 200$ Myr ago.

Our list of 396 candidates (Section 2.8) together with their Galactic proper motions, parallaxes, reddening-corrected photometry and their errors, Gaia DR2 and GALEX identifiers, separation between the Gaia and the GALEX source and reddening are made available in a machine-readable format along with this paper. In Table 3 we show the header and first two lines of this list.

\movetabledown

=2cm

This work was supported by NASA ADAP grant 80NSSC18K0422. We would like to thank Elena D’Onghia for numerous illuminating conversations on this topic. GALEX (Galaxy Evolution Explorer) is a NASA Small Explorer, launched in April 2003. We gratefully acknowledge NASA’s support for construction, operation, and science analysis for the GALEX mission, developed in cooperation with the Centre National d’Etudes Spatiales (CNES) of France and the Korean Ministry of Science and Technology. This study has made use of data from the European Space Agency (ESA) mission Gaia (https://www.cosmos.esa.int/gaia), processed by the Gaia Data Processing and Analysis Consortium (DPAC, https://www.cosmos.esa.int/web/gaia/dpac/consortium). Funding for the DPAC has been provided by national institutions, in particular the institutions participating in the Gaia Multilateral Agreement. Guoshoujing Telescope (the Large Sky Area Multi-Object Fiber Spectroscopic Telescope, LAMOST) is a National Major Scientific Project built by the Chinese Academy of Sciences. Funding for the project has been provided by the National Development and Reform Commission. LAMOST is operated and managed by the National Astronomical Observatories, Chinese Academy of Sciences.