Discovery of a Hypervelocity L Subdwarf at the Star/Brown Dwarf Mass Limit

Abstract

We report the discovery of a high velocity, very low-mass star or brown dwarf whose kinematics suggest it is unbound to the Milky Way. CWISE J124909.08+362116.0 was identified by citizen scientists in the Backyard Worlds: Planet 9 program as a high proper motion ($\mu$ $=$ 0$\farcs$9/yr) faint red source. Moderate resolution spectroscopy with Keck/NIRES reveals it to be a metal-poor early L subdwarf with a large radial velocity ($-$103$\pm$10 km/s), and its estimated distance of 125$\pm$8 pc yields a speed of 456$\pm$27 km/s in the Galactic rest frame, near the local escape velocity for the Milky Way. We explore several potential scenarios for the origin of this source, including ejection from the Galactic center $\gtrsim$3 Gyr in the past, survival as the mass donor companion to an exploded white dwarf. acceleration through a three-body interaction with a black hole binary in a globular cluster, and accretion from a Milky Way satellite system. CWISE J1249+3621 is the first hypervelocity very low mass star or brown dwarf to be found, and the nearest of all such systems. It may represent a broader population of very high velocity, low-mass objects that have undergone extreme accelerations.

1 Introduction

The majority of stars in the neighborhood of the Sun have low relative velocities ($v$ $\approx$ 10-30 km/s) reflecting their common origin in star forming clusters concentrated in the plane of the Milky Way. A rare subset of nearby stars have much higher velocities ($v_{\rm tan}$ $\gtrsim$ 400 km/s; $<$0.3% of stars within 1 kpc; Favia et al. 2015). These stars may originate from the Milky Way’s ancient halo population, or underwent strong dynamical interactions with compact objects such as the Milky Way’s central supermassive black hole (Hills, 1988) of compact binaries in dense clusters (Yu & Tremaine, 2003; Fragione & Gualandris, 2019), or may be the survivors of the supernova explosion of a binary companion (Blaauw, 1961; Portegies Zwart, 2000). The fastest “hypervelocity” stars are unbound to the Milky Way’s gravitational potential and may even have extragalactic origins (Abadi et al., 2009; Piffl et al., 2011). These rare objects trace extreme interactions that may be explored through their trajectories, velocity distributions, and atmospheric properties (Brown, 2015).

The Gaia mission (Gaia Collaboration et al., 2021) has greatly expanded our sample of high-velocity stars by providing 5D (position, parallax, and proper motion) or 6D (plus radial velocity or RV) coordinates for billions of stars out to kiloparsec distances. These measurements, combined with detailed chemical abundances from RAVE (Steinmetz et al., 2006), LAMOST (Cui et al., 2012), APOGEE (Majewski et al., 2017), and other spectral surveys have enabled the discovery and characterization of over a dozen hypervelocity stars (e.g., Du et al. 2019; Quispe-Huaynasi et al. 2022; Liao et al. 2023; Scholz 2024) that originate from environments as diverse as the Galactic center, globular clusters, or satellite systems. Current studies focus on deep optical measurements of rare and distant stars, and primarily sample main sequence and red giant stars over a limited range of mass (0.7 M_⊙ $\lesssim$ M $\lesssim$ 2 M_⊙) and age ($\lesssim$10 Gyr for high-velocity OBAFG stars) which may limit our ability to probe compositions and origins.

The citizen science project Backyard Worlds: Planet 9 (BYW; Kuchner et al. 2017) takes advantage of multi-epoch infrared photometry and astrometry from the Wide-field Infrared Survey Explorer (WISE; Wright et al. 2010) and its extended NEOWISE mission (Mainzer et al., 2014) to search for faint, infrared moving sources identified by a community of citizen scientists. BYW is ideally designed to find low-mass, high velocity stars and brown dwarfs, including local low-temperature metal-poor subdwarfs—the L, T, and Y subdwarfs—from the thick disk and halo populations (Schneider et al., 2020; Meisner et al., 2020, 2021; Kirkpatrick et al., 2021a; Brooks et al., 2022; Burgasser et al., 2024). In this Letter, we report the discovery of a nearby, metal-poor L subdwarf, CWISE J124909.08+362116.0 (hereafter J1249+3621) whose speed may exceed the local escape velocity of the Milky Way, making it the first low-mass hypervelocity star and the nearest such system to the Sun.

2 Identification and Spectral Observations

J1249+3621 was identified by citizen scientists Tom Bickle, Martin Kabatnik, and Austin Rothermich in multi-epoch unWISE images (Lang, 2014; Meisner et al., 2018; Schlafly et al., 2019) on the BYW citizen science portal. Its $W2$ magnitude and $J-W2$ color (Table 2) suggest an early-type L dwarf at an estimated distance of $\approx$100 pc (Kirkpatrick et al., 2021b). Combining astrometry from PanSTARRS (Chambers et al., 2016) and the UKIDSS Hemisphere Survey (UHS; Dye et al. 2018) yields a proper motion of $\mu$ = 884$\pm$5 mas yr^-1, suggesting a tangential velocity of $v_{\rm tan}$ $\approx$ 420 km s^-1 and making it a high priority target for spectroscopic followup.

Table 1: Properties of J1249+3621

Property	Value	Reference
$\alpha_{J2000}$	12h49m09$\fs$08	1
$\delta_{J2000}$	+36$\arcdeg$21$\arcmin$16$\arcsec$	1
$\mu_{\alpha}\cos\delta$	344$\pm$5 mas/yr	2,3
$\mu_{\delta}$	$-$814$\pm$5 mas/yr	2,3
$i$ (AB)	21.48$\pm$0.15 mag	2
$z$ (AB)	20.01$\pm$0.06 mag	2
$y$ (AB)	19.13$\pm$0.05 mag	2
$J$ (Vega)	17.10$\pm$0.03 mag	3
$K$ (Vega)	16.46$\pm$0.04 mag	3
$W1$ (Vega)	15.92$\pm$0.04 mag	1
$W2$ (Vega)	15.59$\pm$0.07 mag	1
SpT	sdL1	4
$d_{est}$	125$\pm$8 pc	4
$v_{\rm tan}$	524$\pm$33 km/s	4
$T_{\rm eff}$	1715 K to 2320 K	4,5,6
$\log{g}$	4.4 to 5.1 (cm/s2)	4,5,6
[M/H]	$-$1.4 to $-$0.5	4,5,6
[$\alpha$/Fe]	+0.25$\pm$0.07	4,5
Est. Mass	0.082${}^{+0.002}_{-0.003}$ M⊙	4,7
$RV$	$-$103$\pm$10 km/s	4
$U_{LSR}$	449$\pm$28 km/s	4
$V_{LSR}$	$-$292$\pm$19 km/s	4
$W_{LSR}$	$-$15$\pm$11 km/s	4
$v_{GRF}$	456$\pm$27 km/s	4

References. — (1) CatWISE2020 (Marocco et al., 2021) at astrometric epoch 2015 May 28 (UT); (2) PanSTARRS (Chambers et al., 2016); (3) UKIRT Hemisphere Survey (Dye et al., 2018); (4) This paper; (5) Alvarado et al. (2024); (6) Mukherjee et al. (2024); (7) Gerasimov et al. (2024)

J1249+3621 was observed on 30 January 2024 (UT) in clear and windy conditions with the Near-Infrared Echellette Spectrometer (NIRES; Wilson et al. 2004) on the Keck II 10m telescope, a cross-dispersed spectrograph that provides $\lambda/\Delta\lambda$ $\approx$ 2700 spectra over 0.9–2.45 $\mu$m. We obtained six exposures of 300 s each at an average airmass of 1.06 with the slit aligned with the parallactic angle. Exposures were made in an ABBA pattern, nodding 10$\arcsec$ along the slit for background subtraction. We also observed the A0 V star HD 108140 ($V$ = 9.35) at a similar airmass, and dome flat lamp exposures at the start of the night for pixel response calibration. Data were reduced using a modified version of the Spextool package (Cushing et al., 2004), following the procedure of Vacca et al. (2003) for flux calibration and telluric absorption correction.

3 Analysis

3.1 Classification and Atmosphere Parameters

Figure 1 compares a smoothed version of our spectrum to near-infrared spectral standards from the SpeX Prism Library Analysis Toolkit (SPLAT; Burgasser & Splat Development Team 2017). J1249+3621 exhibits the characteristic features of L-type dwarfs, with strong H₂O absorption at 1.4 $\mu$m and 1.9 $\mu$m, and FeH, Na I, and K I absorption in the 1.0–1.3 $\mu$m region. Its NIR spectral slope is distinctly bluer than normal L dwarf spectra, and its 2.3 $\mu$m CO band is highly suppressed indicating enhanced H₂ collision-induced absorption (CIA) in a low-metallicity, low-temperature atmosphere (Linsky, 1969; Burgasser et al., 2003). Indeed, the spectrum of J1249+3621 best matches that of the L subdwarf 2MASS J17561080+2815238 (sdL1; Kirkpatrick et al. 2010) but is somewhat bluer, while not as blue as the extreme L subdwarf WISE J043535.80+211509.2 (esdL1; Luhman & Sheppard 2014; Zhang et al. 2017). We classify J1249+3621 as sdL1 based on comparison to a broad range of dwarf and subdwarf spectra. This classification, WISE $W1W2$ and UHS $JK$ Vega magnitudes, and the spectral type/absolute magnitude relations for L sudwarfs from Gonzales et al. (2018) and Zhang et al. (2019) allow us to estimate a spectrophotometric distance of 125$\pm$8 pc for J1249+3621, implying $v_{\rm tan}$ = 524$\pm$33 km/s.

To further evaluate its physical properties, we compared the smoothed spectrum of J1249+3621 to the Sonora Elf Owl (Mukherjee et al., 2024) and Spectral ANalog of Dwarfs (SAND; Alvarado et al. 2024) atmosphere models. These models encompass the temperatures (1500 K $\lesssim$ $T_{\rm eff}$ $\lesssim$ 2400 K) and subsolar metallicities ([M/H] $\lesssim$ $-$0.5) of L subdwarfs, and contain up-to-date opacities and treatments for condensation and disequilibrium chemistry. We used a Metropolis-Hastings Markov Chain Monte Carlo (MCMC) fitting algorithm (Metropolis et al., 1953; Hastings, 1970) to fit the models to the apparent spectral flux densities of J1249+3621, following the procedure described in Burgasser et al. (2024). Our best fit models and parameters are shown in Figure 1. The Elf Owl and SAND grids yield distinctly different effective temperatures ($T_{\rm eff}$ = 2260$\pm$60 K versus 1785$\pm$70 K) and metallicities ([M/H] = $-$0.63$\pm$0.10 versus $-$1.28$\pm$0.10), but similar surface gravities (logg = 4.66$\pm$0.11 versus 4.88$\pm$0.24). We note that the Elf Owl grid does not extend to [M/H] $<$ $-$1 and does not include a prescription for condensate cloud formation, although the latter may be less important in metal-poor L subdwarf atmospheres (Burgasser et al., 2007; Gonzales et al., 2021). The best-fit Elf Owl model is a marginally better fit to the observed spectrum, and its scaled surface fluxes are in good agreement with the spectrophotometric distance estimate ($d$ = 102 pc) assuming a radius of 0.08 R_⊙. The model fit discrepancies could be resolved by a direct distance measurement. Nevertheless, this analysis confirms our interpretation of J1249+3621 as a low-temperature, metal-poor object. We note that the SAND models suggest significant alpha enrichment ([$\alpha$/Fe] = +0.25$\pm$0.07) and the Elf Owl models marginal C/O enrichment (C/O = 0.71$\pm$0.16), potential clues to the origin of this source.

Adopting generous parameter ranges of 1715 K $\lesssim$ $T_{\rm eff}$ $\lesssim$ 2320 K and $-$1.4 $\leq$ [M/H] $\leq$ $-$0.5, and assuming an age $\tau$ $\geq$ 5 Gyr, the SAND evolutionary models (SANDee; Gerasimov et al. 2024) predict a mass of 0.082${}^{+0.002}_{-0.003}$ M_⊙, placing this source marginally above the metallicity-dependent Hydrogen Burning Minimum Mass (HBMM $\approx$ 0.080 M_⊙ for [M/H] = $-$1). The relatively narrow uncertainty range indicated by the evolutionary models is due to the steep decline in temperatures below the HBMM for old low-temperature sources, and does not account for potential systematic biases (e.g., non-solar abundance patterns). We conclude that J1249+3621 is likely a low-mass, metal-poor star, with a 10% probability of being a high-mass brown dwarf.

Figure 1: (Left): Keck/NIRES spectrum of J1249+3621 smoothed to a resolution of $\lambda/\Delta\lambda$ = 150 (black lines), compared to the L dwarf templates 2MASSW J2130446-084520 (L1, blue dot-dashed line; data from Bardalez Gagliuffi et al. 2014), 2MASS J17561080+2815238 (sdL1, magenta line; data from Kirkpatrick et al. 2010), and WISE J043535.80+211509.2 (esdL1, red dashed line; data from Luhman & Sheppard 2014). The spectrum of J1249+3621 is normalized at 1.3 $\mu$m and the comparison spectra normalized to maximize agreement in the 1.0–1.3 $\mu$m range. The dwarf and subdwarf comparisons are offset for ease of comparison. Major spectral features are labeled, as are regions of strong telluric absorption at 1.35–1.5 $\mu$m and 1.75–1.85 $\mu$m (vertical grey bands). (Right) Comparison of the smoothed spectrum of J1249+3621 (black lines) to best fit models (magenta lines) from the Sonora Elf Owl (top; Mukherjee et al. 2024) and SAND (bottom; Alvarado et al. 2024) atmosphere grids. Spectra are scaled to apparent fluxes using the UHS $J$-band magnitude of J1249+3621, and the models scaled to minimize $\chi^{2}_{r}$. Model parameters are listed in the figure captions. Difference spectra (observed minus computed; black line) are compared to the $\pm$1$\sigma$ uncertainty of the observed flux densities (grey band) in the bottom panels.

3.2 Radial Velocity and Kinematics

At full resolution, the Keck/NIRES data permit assessment of the radial velocity of J1249+3621, which we approached using a forward modeling technique described in Burgasser et al. (2024). In brief, we fit the extracted NIRES spectrum without telluric correction to a high resolution stellar atmosphere model, $M(\lambda)$, from Allard et al. (2012, BT-Settl) and an empirical telluric absorption template, $T(\lambda)$, from Livingston & Wallace (1991) using the parameterized data model

$$D(\lambda+\delta_{\lambda})=C(\lambda)\times\left[\left(M(\lambda^{*})\otimes\kappa_{R}(v\sin{i})\right){\times}T^{\alpha}(\lambda)\right]\otimes\kappa_{G}(v_{b})+\delta_{f}.$$ (1)

Here, $C(\lambda)$ is a fifth-order polynomial continuum correction, $\kappa_{R}(v\sin{i})$ is a rotational broadening profile for projected velocity $v\sin{i}$, $\alpha$ is a scaling exponent for the strength of the telluric absorption, $\kappa_{G}({v_{b}})$ is a Gaussian instrumental broadening profile parameterized by velocity width $v_{b}$, and $\delta_{\lambda}$ and $\delta_{f}$ represent small offsets to the wavelength scale and normalized flux density to account for residual calibration errors. We used a $T_{\rm eff}$ = 2000 K, $\log{g}$ = 5.0, solar metallicity model evaluated at a shifted wavelength $\lambda^{*}=\lambda\left(1+\frac{RV+V_{bary}}{c}\right)$, accounting for the unknown RV and known barycentric motion $v_{bary}$ = 16.82 km s^-1 at the time of observation. After identifying an optimized set of parameters using the Nelder-Mead algorithm (Nelder, 1965) with a fixed $v\sin{i}$ = 50 km s^-1, we used a Metropolis-Hastings MCMC algorithm to map the parameter uncertainty space for the five remaining free parameters: RV, $\alpha$, $v_{b}$, $\delta_{\lambda}$ and $\delta_{f}$. Fits were conducted in two wavelength regions that contain both stellar and telluric absorption features: 1.10–1.19 $\mu$m which contains Na I (1.138, 1.140 $\mu$m) and K I (1.169, 1.177 $\mu$m) stellar lines and a telluric complex over 1.11–1.15 $\mu$m, and 1.235–1.28 $\mu$m which contains K I stellar lines (1.243, 1.252 $\mu$m) and a telluric feature at 1.269 $\mu$m. We avoided the CO band at 2.3 $\mu$m, commonly used for RV forward modeling of L dwarfs (cf. Blake et al. 2010; Konopacky et al. 2010; Burgasser et al. 2015), due to the suppression of this feature by H₂ absorption. Figure 2 displays the best fit models for these regions, which yield consistent values for RV ($-$92${}^{+13}_{-14}$ km/s and $-$114${}^{+13}_{-14}$ km/s). We adopt a mean RV = $-$103$\pm$10 km/s for J1249+3621.

Figure 2: Forward modeling of the Keck/NIRES spectrum of J1249+3621 in the 1.10–1.19 $\mu$m (left), and 1.235–1.28 $\mu$m (right) spectral regions, both of which contain stellar and telluric absorption features. Each panel displays from top to bottom: the telluric spectrum in green, the stellar model in blue, the combined model in magenta overlaid on the observed spectrum in black, and the difference spectrum (data-model) in black overlaid on the $\pm$1$\sigma$ uncertainty band in grey. The reduced $\chi^{2}$ of each fit is indicated in the text to the right of the plot.

Combining the measured RV, proper motion, and position of J1249+3621 with its estimated distance, we computed $UVW$ velocities in the Local Standard of Rest (LSR). The velocities (Table 2) indicate a slightly retrograde motion relative to Galactic rotation ($V_{LSR}$ = $-$292$\pm$19 km/s), with a trajectory directed radially inward ($U_{LSR}$ = 449$\pm$28 km/s) and constrained to the Galactic disk ($W_{LSR}$ = $-$15$\pm$11 km s^-1). Assuming a local Galactic circular velocity of $v_{circ}$ = 220 km/s, the velocity of J1249+3621 translates into a Galactic rest frame speed of $v_{GRF}$ = 456$\pm$27 km/s, or 0.47$\pm$0.03 kpc/Myr. The median speed is just below the Galactic escape velocity at the Solar radius, with current estimates ranging from 521${}^{+46}_{-30}$ km/s (Williams et al., 2017, 1.6$\sigma$ above) to 580$\pm$63 km/s (Monari et al., 2018, 1.8$\sigma$ above). Given the uncertainties in the inferred velocities and potential models, we find that J1249+3621 has a significant probability of being unbound to the Milky Way.

Figure 3: Projected orbit of J1249+3621 from galpy (Bovy, 2015) for 100 initial conditions sampling uncertainties in distance, proper motion, and RV (black lines). The top panels display the forward orbit over 10 Gyr projected onto the Galactic plane on wide (left) and narrow (right) distance scales. The bottom panels display the backward orbit over 150 Myr projected onto the Galactic plane (left) and XZ coordinates (right). The top left panel shows the present-day positions (magenta crosses) and forward-projected orbits (magenta lines) of 50 satellite galaxies from Fritz et al. (2018), as well as the Milky Way’s virial radius of 180 kpc (dashed blue circle; Sylos Labini et al. 2023). The top right panel shows the present-day positions of 161 globular clusters from Vasiliev (2019, magenta circles) and the closest-approach radius of $\sim$1 kpc from the Galactic center (yellow dashed circle). Globular clusters and their backward-projected orbits (magenta lines) are also shown in the bottom panels.

The Galactic orbit of J1249+3621 was generated using galpy (Bovy, 2015). We used the axisymmetric MWPotential2014 potential to integrate the trajectory of J1249+3621 forward and backward in time by up to 10 Gyr, with a finer sampling of the backward orbit up to 150 Myr. We drew 100 random initial conditions sampling the uncertainties on distance, proper motion, and radial velocity assuming independent Gaussian distributions. Figure 3 displays the forward and backward trajectories projected onto the disk plane and in cylindrical coordinates. The forward motion of J1249+3621 shows a close approach to the inner region of the Milky Way, coming within 0.94${}^{+0.28}_{-0.19}$ kpc of the Galactic center, then extending beyond the Milky Way’s virial radius of 180 kpc (Sylos Labini et al., 2023). The median model remains bound to the Milky Way on a $\sim$3 Gyr, highly eccentric orbit, but 17% of our simulated orbits are unbound over 10 Gyr. The backward orbit is approximately radial and tightly confined to the Galactic plane, converging to within 2$\arcdeg$ of 05^h22^m25^s +38$\arcdeg$37$\arcmin$00$\arcsec$ (galactic coordinate 38$\fd$6 +1$\fd$2) by 50 Myr in the past.

4 Assessing the Origins of J1249+3621

4.1 Was J1249+3621 Ejected from the Galactic Center?

J1249+3621 has a unique trajectory and speed; less than 0.002% of stars in Gaia within 200 pc of the Sun have comparable tangential velocities. While this could nevertheless represent the extreme tail of the halo velocity distribution Hawkins & Wyse (2018), we explored potential origins of J1249+3621 in the context of currently known hypervelocity stars. Its small but nonzero orbital angular momentum in the Galactic rest frame ($L_{z}$ = 572${}^{+147}_{-146}$ kpc km s^-1) and inward trajectory would seem to argue against ejection from the Galactic center through the traditional Hill’s mechanism (Hills, 1988). However, if J1249+3621 is bound to the Milky Way, which is the case for 83% of our orbit simulations, we could be observing it on a return pass after intervals of roughly 3 Gyr, with torques imparted by asymmetries in the Galactic potential; i.e., spiral structure or the inner bar (Sellwood & Binney, 2002; Daniel & Wyse, 2018). Moreover, while the majority of Galactic center stars are metal-rich (Carr et al., 2000; Ramírez et al., 2000; Cunha et al., 2007), recent studies have identified metal-poor, alpha-enhanced M giants in the nuclear region similar in nature to J1249+3621 (Schultheis et al., 2015, 2020). A small number of metal-poor hypervelocity stars have also been associated with ejection from the Galactic center (Li et al., 2012, 2023). Thus, despite its radially inward trajectory, ejection via the Hill’s mechanism is a possible origin for J1249+3621.

4.2 Is J1249+3621 the Surviving Companion of a Type Ia Supernova?

An alternative disk origin for J1249+3621 is as the tight binary companion to an accreting white dwarf that exceeded the 1.4 M_⊙ Chandrasekhar limit (Chandrasekhar, 1931, 1935), underwent a thermonuclear explosion, and released J1249+3621 at high speed (Blaauw, 1961; Portegies Zwart, 2000; Shen et al., 2018). Several accreting short-period white dwarf-brown dwarf pairs (polars and cataclysmic variables) are known with white dwarf masses extending up to 0.94 M_⊙ (Longstaff et al. 2019 and references therein). If J1249+3621 was the initially more massive donor in such a system, its subsequent ejection speed would exceed 690$\left(\frac{P}{1~{}hr}\right)^{-1/3}$ km s^-1 based on orbital motion alone, where $P$ is the orbit period at detonation. Periods of 1–1.5 hr are sufficient for Roche lobe overflow for donors near the HBMM, based on theory and as observed for low-mass cataclysmic variables (e.g., Kolb & Baraffe 1999; Littlefair et al. 2006). This scaling law yields ejection speeds of 550–700 km s^-1, on par with J1249+3621’s LSR speed of 534 km/s. Contributions from supernova shockwaves and mass stripping could drive ejection velocities even higher (Pan et al., 2012; Rau & Pan, 2022).

As the ejection direction in this scenario is isotropic, the probability of an individual source such as J1249+3621 passing by the Sun is very low. However, the overall higher density of stars in the Galactic plane makes it more likely that we would see a closely-passing ejectee with a trajectory confined to the plane. We note that there are no known supernova remnants in the projected past position of J1249+3621, but as remnants dissipate and merge with the interstellar medium within $\lesssim$1 Myr (Leahy & Williams, 2017) this a relatively weak constraint on the time since this source could have been ejected.

4.3 Was J1249+3621 Ejected from a Globular Cluster?

Another possible origin for J1249+3621 is dynamical ejection from a globular cluster (GC). The top-heavy present-day mass functions of these clusters is evidence of efficient tidal dispersion of low-mass objects (Fregeau et al., 2002), and large velocity kicks exceeding the 10–100 km/s escape velocities of GCs are a natural outcome of strong three- and four-body dynamical interactions that are commonplace in these systems (Leonard, 1991). Such kicks are amplified by interactions with black hole binaries (Cabrera & Rodriguez, 2023), which are now understood to dominate the dynamics of the centers of most GCs (Kremer et al., 2020a). The characteristic kick velocity imparted to a star of mass $m$ during a strong encounter with a binary with components of equal mass $M$ and semi-major axis $a$ is roughly:

$$v_{\rm kick}\approx\sqrt{\frac{GM^{2}}{ma}}\approx 600\Bigg{(}\frac{M}{20\,M_{\odot}}\Bigg{)}\Bigg{(}\frac{m}{0.1\,M_{\odot}}\Bigg{)}^{-1/2}\Bigg{(}\frac{a}{10\,\rm{au}}\Bigg{)}^{-1/2}\,\rm{km/s}.$$ (2)

Thus, for $a\lesssim 10\,$au and/or compact object masses $M\gtrsim 20\,M_{\odot}$, sufficiently high velocities can be achieved.

To test this mechanism, we searched for low-mass high velocity stars in the CMC Cluster Catalog (Kremer et al., 2020b), a public suite of $N$-body simulations intended to serve as a proxy for the Galactic GC population (Harris, 1996). These simulations adopt a Kroupa (2001) initial mass function ranging from $0.08-150\,M_{\odot}$, and therefore include stars of mass comparable to J1249+3621. At times $t>8\,$Gyr, typical of the ages of Milky Way GCs, we identify roughly $4,000$ low-mass stars (M $<$ 0.2 M_⊙) ejected with velocities of at least $100\,$km/s, corresponding to a rate of roughly 1 Myr^-1 across the Milky Way. Of these, six stars are ejected with $V>500\,$km/s, corresponding to a rate of roughly 2 Gyr^-1. all six hypervelocity subdwarfs awere ejected via dynamical encounters with stellar-mass black hole binaries with properties similar to the scales in Eqn. 2. Hence, simulations support the scenario of GC ejection as a means of generating very high velocity, very low mass objects, albeit as exceedingly rare events. The isotropic distribution of ejections further reduces the probability of this scenario, with the same caveat that the higher concentration of GCs in the Galactic plane makes detection of closely-passing ejectees on planar orbits more likely.

There are no known globular clusters within 5$\arcdeg$ of the projected past position of J1249+3621. However, given the long timescales involved it is necessary to account for cluster motion. We used our galpy orbits to assess whether the trajectory of J1249+3621 intersected with any of the 161 GCs in the kinematic catalog of Vasiliev (2019), projecting all orbits back 150 Myr. The closest approaches by NGC 3201 and Palomar 1 are more than 4 kpc in separation, making these improbable origin sites. The low Galactic latitude of the projected past position of J1249+3621 ($b$ = +1$\fd$2) could argue for an origin from an as yet undiscovered GC hidden in the Galactic plane. Alternately, the higher concentration of GCs near the Galactic center implies that J1249+3621 could have been ejected from one of these systems, and is now making a return pass $\gtrsim$3 Gyr later. Despite these considerations, we find the GC ejection scenario less likely than the prior two scenarios.

For completeness, we note that four open clusters lie within 1$\arcdeg$ of the projected past position of J1249+3621, including the well-studied NGC 1857 system at 3.1 kpc (Herschel, 1864). However, the lower stellar densities of these systems and lack of compact binaries from massive progenitors makes dynamic ejection unlikely, and none of these clusters have significantly subsolar metallicities characteristic of J1249+3621.

4.4 Is J1249+3621 an Extragalactic Star?

As the bound orbits of J1249+3621 extend beyond the Milky Way’s virial radius, it is possible that this source could have an extragalactic origin, specifically accretion from one of the Milky Way satellites (Abadi et al., 2009; Piffl et al., 2011). Its present trajectory does not align with distant extragalactic systems such as M31 or the Magellanic Clouds. We examined the intersection of J1249+3621’s orbit with 50 Milky Way satellites from Fritz et al. (2018) using the same procedure as our GCs, integrating back to 10 Gyr. Only one system, Tucana III (Drlica-Wagner et al., 2015), comes within 5 kpc of the median orbit of J1249+3621 about 6 Gyr ago. The substantial uncertainties of the trajectories of J1249+3621 and the known dwarf satellites over Gyr timescales means we cannot strictly rule out this scenario; however, we find it the least likely of the scenarios considered here, given the trajectory of J1249+3621 being confined to the Galactic plane.

Inferring the true origin of J1249+3621 will require further investigation into its physical and atmospheric properties. A Galactic center origin requires closer examination of its orbital trajectory through refinement of its distance (by direct parallax measurement) and velocity components, as well as a more realistic, non-axisymmetric Galactic potential model. A more detailed compositional analysis would also help clarify its origin. For example, if J1249+3621 is the surviving companion of a Type Ia supernova, its atmosphere may be enriched with heavy elements, particularly nickel, depending on the degree of mass stripping by the supernova blast wave (Rau & Pan, 2022). Similarly, if J1249+3621 was ejected from the Galactic center, a globular cluster, or a satellite system, its detailed abundances may provide the chemical fingerprint of its origin. Better assessment of composition of through additional optical and infrared spectra, and improved atmosphere models exploring specific abundances (e.g., Gerasimov et al. 2022) are needed to infer a chemical-based origin.

Finally, we note that at least one other metal-poor L subdwarf, the esdL1 ULAS J231949.36+044559.5 (Zhang et al., 2018), has a high enough estimated tangential velocity (513${}^{+50}_{-46}$ km/s) to make it a promising hypervelocity candidate, although no RV has yet been reported for this source. The existence of at least one and possibly two L subdwarfs within $\sim$200 pc of the Sun with hypervelocity speeds suggests a considerably larger population of such sources could exist in the Milky Way system.