betaFeedback

WorksPhysics and Astronomy

Mass-loss Rates of 74 Bow Shock OB Stars: No Evidence for Bi-stability Behavior

Abstract

Accurate measurements of wind-driven mass-loss rates are essential for understanding the evolution and fate of massive stars. Established techniques for measuring M˙\dot{M} suffer from large dispersions and become insensitive in the regime of weak-winded late-O to early B stars. We employ a promising new technique to measure stellar mass loss for \numstars~early type stars that produce infrared stellar bow shock nebulae. We use updated stellar parameters, improved parallactic distances and proper motions, and modern dust models to characterize the bow shock nebulae and obtain new measurements of mass-loss rates. \mdot ranges from 2\times10^{-9}~\moperyear for weak-winded late-O and early B stars at \log\left(\lstar/L_\odot\right)\approx4.5 to 4\times10^{-6}~\moperyear for the most luminous stars in our sample at \log\left(\lstar/L_\odot\right)\approx6.3. Stellar space velocities relative to their local standard of rest are insufficient to accurately reflect the relative star-ISM velocities for the majority of our sample, which have speeds much less than canonical runaway thresholds, suggesting the pervasive influence of bulk interstellar flows in forming bow shock nebulae. Adopting a minimum relative velocity of 20~\kms is physically motivated and yields mass-loss rates in excellent agreement with the latest theoretical prescriptions while requiring no adjustable parameters. We find no enhancements in \mdot for stars cooler than 25~kK, in agreement with modern mass-loss recipes that do not include bi-stability effects.

Introduction

Massive stars—stars with masses ranging from 8≲M∗≲120 M⊙8 \lesssim M_\ast\lesssim120\ M_\odot—are the sites of nucleosynthesis and the main sources of mechanical and radiative feedback in their host galaxies. Although the stellar initial mass function dictates that early type stars be far fewer in number than their low-mass counterparts,

massive stars dominate the radiative output of their host galaxies. Due to their high luminosities, massive stars also host radiation-driven stellar winds. Although massive stars have other channels through which their mass is altered throughout their lifetimes (binary interactions being foremost among them), mass loss through stellar winds acts continually to reduce the star’s mass and, when integrated over the star’s lifetime, can reduce the star’s mass by as much as one fourth ((J. I. Castor et al. 1975). After the main- and post main-sequence stages of evolution, massive stars evolve into either neutron stars or black holes depending on the stellar mass at the moment of core collapse. Systems of degenerate objects are of particular importance in the age of gravitational wave astronomy [30, 78, 47, 2, 3 b].

Several techniques for measuring mass-loss rates have been developed since the discovery of massive star stellar outflows [79, 83, 89]. Recent works combine multiple methods to obtain measurements of M˙\dot{M} while taking into account complicating factors [65, 91]. One common method makes use of the P-Cygni profiles in the UV metal-resonance transitions [33, 45, 63]. Two wind properties can be extracted from UV techniques—terminal wind speed v∞v_{\infty} and mass-loss rate M˙\dot{M}. Deriving mass-loss rates M˙\dot{M} from UV resonance line technique requires modeling of parameters that are not easily observable, such as the clumping factor in the winds. Another challenge with UV techniques is that under typical conditions, UV resonance lines are saturated in absorption. In such cases, the derived mass-loss rate can only serve as a lower limit. Since the optical depth of the line is proportional to the density of the wind ρ\rho, the UV method is classified as a ρ\rho diagnostic—a method in which the derived mass-loss rate scales linearly with the density of the wind.

Another technique to measure mass-loss rate utilizes the Hα\alpha recombination lines in the wind to measure mass-loss rates [60, 58, 80, 64]. The benefit of Hα\alpha methods is that unlike the UV resonance line technique, there is no uncertainty about the ionization of hydrogen for hot stars, unlike metal species which exhibit several ionization states in different fractions. Since the Hα\alpha-derived mass-loss rates scale as ρ2\rho^{2}, and hot stars are known to exhibit small-scale density inhomogeneities in their winds (known as micro-clumping, [29, 82, 25, 73]), it is common for Hα\alpha methods to yield large over-estimates in M˙\dot{M}.

stars, unlike metal species which exhibit several ionization states in different fractions. Since the Hα\alpha-derived mass-loss rates scale as ρ2\rho^{2}, and hot stars are known to exhibit small-scale density inhomogeneities in their winds (known as micro-clumping, [29, 82, 25, 73]), it is common for Hα\alpha methods to yield large over-estimates in M˙\dot{M}. In their study of 40 Galactic O stars, A. W. Fullerton et al. [31] compared ρ\rho- and ρ2\rho^{2}-derived measurements of M˙\dot{M} and found a large discrepancy between the two methods. This discrepancy seemed to suggest microclumping was more significant than previously thought, and Hα\alpha methods would have to be revised significantly to account for these effects. However, once thought to be independent of clumping effects, UV methods may also suffer complications. Clumps that are optically thin in Hα\alpha may be optically thick at some frequencies in the UV [72]—an effect known as “macroclumping.” Accounting for macroclumping in the winds can help resolve the discrepancy between Hα\alpha and UV measurements [92]. As of yet, the effects of wind micro- and macroclumping on mass loss and spectral transitions are still being investigated and a fully consistent study is still needed [89].

A new technique to measure mass-loss rates that is physically independent of existing methods uses stellar bow shock nebulae [49, 50], hereafter K18 and K19, respectively). Stellar bow shock nebulae are arc-shaped features formed when dust grains in the interstellar medium are shocked and heated by the winds of a high-velocity, early type star. The mass-loss rate of the central star can be measured given information on the properties of the ambient interstellar medium and morphology of the resulting bow shock. T. R. Gull & S. Sofia [34] were the first to propose this technique in their work on the newly discovered shock-like nebulae around a T Orionis K0e variable star, LL Ori, and an O9.5 V runaway, ζ Oph, from optical images. Balancing ram pressure from the incident stellar wind on the bulk motion of the surrounding material, they found that the mass-loss rate of the central star could be calculated from

M˙=4πR02v∗2ρavw(1)\dot{M} = \frac{4\pi R_0^2 v_*^2 \rho_a}{v_w} \tag*{(1)}

with R0R_0 representing the “standoff distance” between the star and the apex of the bow shock, v∗v_* being the velocity of the star with respect to its surrounding ISM, ρa\rho_a signifying the ambient density of the pre-shock ISM, and vwv_w denoting the speed of the outflowing stellar wind. Using estimates on these parameters for both bow shock nebulae, T. R. Gull & S. Sofia [34] derived independent measures of mass-loss rates for both stars. In the case of ζ Oph, their value of 2.3×10−8 M⊙ yr−12.3 \times10^{-8}\ M_\odot\ \mathrm{yr}^{-1} agrees well with the contemporary (D. C. Abbott [5], 2.20×10−8 M⊙ yr−12.20 \times10^{-8}\ M_\odot\ \mathrm{yr}^{-1}) and modern (R. Björklund et al. [13], 1.91×10−8 M⊙ yr−11.91 \times10^{-8}\ M_\odot\ \mathrm{yr}^{-1}) theoretical expectations. Using bow shocks to measure mass-loss rates would remain an untapped concept for several decades until a reformulation from K18 allowed this technique to be extended to the growing collection of IR wavelength bow shock nebulae ([71, 77, 36, 51, 75, 76, 53] (D. van Buren et al. 1995)). K18 refined measurements of mass-loss rates by inferring interstellar number densities from archival infrared Herschel Space Observatory images (most targets selected from Hi-GAL survey S. Molinari et al. [69]). K18 applied this method to measure the mass-loss rates for 20 OB stars that supported infrared bow shock nebulae. With this preliminary sample, they found that bow shock-derived mass-loss rates

were situated between Hα measurements (uncorrected for clumping) and UV measurements. After a promising demonstration of the bow shock method, K19 expanded the bow shock sample to 70 OB stars. They found mass-loss rates ranging from ∼10−7\sim10^{-7} to 10−9 M⊙ yr−110^{-9}\ M_\odot\ \mathrm{yr}^{-1} with good agreement between results from the bow shock method and UV measurements from F. Martins et al. [65] for early type stars (Teff>35 kKT_{\mathrm{eff}} > 35\ \mathrm{kK}), but showed disagreement for late-O stars with bow shock measurements 1.0–1.5 orders of magnitude greater than UV measurements. They also found that Hα measurements were greater than bowhshock measurements for late-O stars, but noted that many of the Hα measurements were upper limits on account of the insensitivity of the Hα method in this regime of weak-winded stars. Interestingly, they found that stars on the cool side of the bi-stability jump (i.e., with effective temperatures Teff≲25 kKT_{\mathrm{eff}} \lesssim25\ \mathrm{kK}) showed evidence for the predicted enhancement, with M˙\dot{M} orders of magnitude greater than the J. S. Vink et al. [90] prescription, which included bi-stability enhancements. It is worth noting that their analysis was based on red wavelength optical spectra, which limited the precision on determinations of stellar effective temperature. For stars hotter than 25 kK, their results were ∼0.4\sim0.4 dex below the J. S. Vink et al. [90] predictions, showing that, on average, bow shock-derived mass-loss rates are a factor of 3 lower than the commonly used prescription.

H. A. Kobulnicky & W. T. Chick [48] (hereafter K22) studied the kinematics of stars at the center of bow shock nebulae. They found that although bow shock stars had a slightly higher median 2D space velocity than the control sample of O stars selected from the Galactic O-Star Catalog (GOSC, [62, 84], 14.6 km s−1\mathrm{s}^{-1} versus 11.3 km s−1\mathrm{s}^{-1}, respectively), bow shock stars had an almost identical runaway fraction as the control sample (24% versus 22%). Said another way, bow shock stars are not exclusively runaway stars. In fact, the majority of bow shock stars have space velocities less than 25 km s−1^{-1} (76%). It is also true that not all runaway stars are bow shock stars. In their search of bow shock nebulae around runaway stars, C. S. Peri et al. [76] found that only 10–15% of runaway stars exhibit IR bow shocks. K18 proposed the interstellar density was the determining factor in forming an IR bow shock rather than the central star’s spectral type or velocity with respect to its surroundings. They found an empirical lower limit on the ambient interstellar density of na≳5 cm−3n_a \gtrsim5\ \mathrm{cm}^{-3}. Following this result, K22 identified two distinct populations of bow shock nebulae: fast-moving stars in which the morphological symmetry axis of the bow shock is parallel to the star’s velocity vector (31%) and slow-moving stars with an isotropic distribution of alignments between the bow shock’s morphology and peculiar motion (69%). They also noted that stars in the slow-moving, randomly aligned group tended to face H II regions or bright-rimmed clouds visible at 8 μ\mum—the “in situ” bow shocks, coined by M. S. Povich et al. [77]. This led the authors to conclude that local flows played a role in forming the bow shock in these cases, while the aligned group arose from a scenario in which the star’s runaway velocity generated the supersonic motion needed to form a bow shock.

The opacity of the star’s atmosphere due to metal-resonance transitions is thought to play a significant role in the rate at which mass is expelled from the star through these stellar winds. Therefore, some works have predicated that recombination from one ion to another can produce dramatic increases in mass-loss rate M˙\dot{M}—the bi-stability jump. Simulations of J. S. Vink et al. [90] predict a bi-stability jump at stellar effective temperature Teff≃25T_{\mathrm{eff}} \simeq25 kK due to the recombination of Fe IV to Fe III, and minor bi-stability jumps at 15 and 35 kK, which the authors attributed to the recombination of Fe III to Fe II and C IV to C III, respectively. Observational works have thus set out to find evidence of this predicted bi-stability jumps, focusing primarily on the bi-stability jump at Teff≃25T_{\mathrm{eff}} \simeq25 kK. P. Benaglia et al. [11], for example, used radio-continuum techniques to study M˙\dot{M} for 12 late-O to early B supergiants and found some evidence of enhancements due to bi-stability. The previous bow shock work of K19 also reported increases M˙\dot{M} for stars cooler than 25 kK. However, recent works find no increase in mass-loss rates for stars on the cool side of the bi-stability gap ([23, 7]).

In this work, we measure the mass-loss rates for 74 stars that lie at the center of bow shock nebulae in the same manner as K18/K19, with some important improvements. The Gaia Data Release 3 (Gaia DR3, Gaia Collaboration et al. [32]) provides newer parallaxes and proper motions for bow shock-driving stars, yielding more accurate measurements of distances and peculiar space velocities. In addition, N. Patten et al. [74] used blue-optical spectra to obtain more precise measurements of stellar temperature TeffT_{\mathrm{eff}}, surface gravity gg, and projected-rotational velocity vsin⁡iv \sin i for a large sample of bow shock-driving stars. This improvement enables more accurate estimates of terminal wind speeds v∞v_{\infty}, another factor in mass-loss-rate calculations. Finally, A. S. Whisnant et al. [93] used citizen science data to tabulate the stand-off distances R0R_0 for a subset of stars in the H. A. Kobulnicky et al. [53] and T. Jayasinghe et al. [46] bow shock catalogs, from which this work selects targets. These combined improvements enable more robust estimates for a comparable number of early-type stars as K19.

In § 2 we describe our target selection. In § 3 we present the parameters needed to measure M˙\dot{M} using the bow shock method and how we obtain each parameter. In § 4 we present updated measurements of M˙\dot{M} and compare our values to different theoretical prescriptions. In § 5 we summarize the main findings of this work and discuss the implications of our results on the predicted bi-stability effect and what our results suggest about stars that form bow shocks and their velocities relative to their surroundings/

Target Selection

We selected stars with previously measured stellar parameters from the spectroscopic study P25, which selected a subset of bright, northern targets with reliable parallax measurements from the K16/J19 catalogs. We then limited these 103 P25 OB stars to only include stars with 70 μ\mum images from Herschel Space Observatory (HSO) surveys. This cut yielded 71 bow shock stars. We also searched the literature for K16/J19 stars with HSO data that also have existing spectroscopic references in the literature. This led us to search the large-scale spectroscopic O- and B-star project IACOB (A. de Burgos et al. [24]) and a study on the Carina nebula (R. J. Hanes et al. [37]). We then fitted the SED of these stars using the literature measurements on fundamental stellar parameters TeffT_{\mathrm{eff}} and log⁡g\log g. This yielded three additional stars1, raising the total number to 74.

Note— (1) Stellar bow shock nebulae identifier from the K16 or addendum J19 catalogs; (2) Common alternative identifier, when available; (3) Right Ascension (J2000) in HH:MM:SS; (4) Declination (J2000) in DD:MM:SS; (5) Stellar effective temperature in kK; (6) Uncertainty in stellar effective temperature in kK; (7) Log surface gravity in cgs units; (8) Uncertainty in log surface gravity in cgs units; (9) Stellar radius in solar radii; (10) Uncertainty in stellar radius in solar radii; (11) Stellar luminosity in solar luminosity units; (12) Uncertainty in stellar luminosity in solar luminosity units; (13) Stellar mass in solar masses; (14) Uncertainty in stellar mass in solar masses; (15) binarity flag (from P25) with SB1 corresponding to a single-lined spectroscopic binary with significant radial velocity variations, SB2 indicating a double-line spectroscopic binary, R signifying a Gaia DR3 RUWE value > 1.4, and EB indicating an eclipsing binary.

Table 1 lists the bow shock stars studied in this work. Column (1) presents the bow shock identifier from the K16 or the J19 catalogs. The star κ\kappa Cas does not appear in the K16/J19 catalogs but was included in the sample of P25 because the central star is a blue supergiant and lies at the center of a bow shock nebula. Column (2) lists the common name for the bow shock star, when the bow shock star is a part of other catalogs. Columns (3) and (4) list the Right Ascension and Declination in J2000 coordinates. Columns (5) and (6) list the temperature and uncertainty in kK. Columns (7) and (8) list the log of the surface gravity and the uncertainty in cgs units. Columns (9) and (10) list the stellar radius and uncertainty in solar units. Columns (11) and (12) present the log of the stellar luminosity and the uncertainty in the log of stellar luminosity in solar luminosity units. Columns (13) and (14) list the stellar mass and uncertainty in solar units. Column (15) lists the binarity indicator from P25, with SB1 indicating a single-lined spectroscopic binary that has significant radial velocity variation, SB2 indicating a double-lined spectroscopic binary, R signifying a Gaia DR3 RUWE value >1.4> 1.4, and EB indicating an eclipsing binary.

Our full list of targets spans galactic longitudes ℓ=0.1∘\ell= 0.1^\circ–359.1∘359.1^\circ and galactic latitudes b=−1.0∘b = -1.0^\circ–23.6∘23.6^\circ with a median latitude of 2.9∘2.9^\circ. Most bow shock stars are within the ±1∘\pm1^\circ of the galactic plane with the only outlier ζ\zeta Oph (BS013⋆_\star) at a galactic latitude +23.6∘+23.6^\circ. The distances range from 135 pc for the closest star (ζ\zeta Oph, BS013⋆_\star) to 11.5 kpc for the furthest star (BS303⋆_\star) with a median distance of 2.02 kpc. The target stars span temperatures 20.0–46.8 kK, with a median temperature of 27.3 kK and log⁡g\log g 2.69–4.64, with a median log⁡g\log g of 3.83.

Measuring

In K18, the authors showed that Equation (1) can be rewritten as:

M˙ (M⊙ yr−1)=1.67×10−28×[R0 (arcsec)]2D (kpc)[Va (km s−1)]2Iν (Jy sr−1)Vw (km s−1) ℓ (arcsec) jν (Jy cm2 sr−1nucleon−1)(2)\dot{M}\, (M_\odot\ {\rm yr}^{-1}) = 1.67 \times10^{-28} \times\frac{[R_0\ (\mathrm{arcsec})]^2 D\ (\mathrm{kpc}) [V_a\ (\mathrm{km\ s}^{-1})]^2 I_\nu\ (\mathrm{Jy\ sr}^{-1})}{V_w\ (\mathrm{km\ s}^{-1})\,\ell\ (\mathrm{arcsec})\,j_\nu\ (\mathrm{Jy\ cm^2\ sr}^{-1}\mathrm{nucleon}^{-1})} \tag*{(2)}

where R0R_0 is the standoff distance between the star and the apex of the nebula, DD is the distance to the bow shock star, VaV_a is the velocity of the star with respect to its surroundings, IνI_\nu is the specific intensity of the emitting dust in the bow shock nebula, VwV_w is the terminal velocity of the stellar wind, ℓ\ell is the chord length across the bow shock arc, and jνj_\nu is the dust emissivity per nucleon of grains at the apex of the nebula. The expression is linear for the quantities DD, IνI_\nu, VwV_w, ℓ\ell, and jνj_\nu, while R0R_0 and VaV_a take a squared dependence.

Figure 1 displays a three-color representation of the bow shock nebula surrounding G353.4162+00.4482 (bow shock candidate star BS692⋆_\star from K16). Red and green depict archival HSOHSO 160 and 70 μm\mu\mathrm{m} images and blue represents Spitzer Space Telescope (SST) 24 μm\mu\mathrm{m} image data. We use the Multiband Imaging Photometer GALactic plane ([35, 35]) 24 μm\mu\mathrm{m} or Widefield Infrared Survey Explorer ([94 WISE]) 22 μm\mu\mathrm{m} survey data throughout this work to assess the bow shocks’ location and morphology. The bow shock is most evident at 24 μm\mu\mathrm{m} and is visible at 70 μm\mu\mathrm{m} but not at 160 μm\mu\mathrm{m}. The bow shock is the dominant feature at 24 μm\mu\mathrm{m}, but at 70 μm\mu\mathrm{m}, both background and foreground nebulosity are visible. We obtain the chord length as the length across the bow shock arc perpendicular to the standoff distance vector. A white ×\times marks the location of the driving OB star, which is not directly visible in the images. The yellow vectors illustrate the morphological parameters R0R_0 and ℓ\ell. To obtain R0R_0, we adopt the standoff distances measured by the citizen science project described in A. S. Wisniewski et al. (2025). They found overall good agreement between their crowd-sourced measurements for R0R_0 compared to single-measurement previous values from K16. For the subset of stellar bow shock nebulae with “small” standoff distances (R0≲10′′R_0 \lesssim10^{\prime\prime}) as measured in K16, A. S. Whisnant et al. [93] reported their updated measurements of R0R_0 were systematically larger by ∼1. ⁣′′5\sim1.\!^{\prime\prime}5. We judge that the multiple measurements for R0R_0 obtained in A. S. Whisnant et al. [93] are superior to the subjective measurements made by a single scientist in K16, so we adopt values from the former, when available. We take the measurements of the projected R0R_0 for unknown viewing angles to approximate the standoff distance at 0∘0^\circ since simulations have shown that the peak emission in the infrared is invariant for a wide range of viewing angles (D. M. Acreman et al. [6]).

Three color image of object BS692⋆

Figure 1. Three color image of object BS692⋆ from K16 with red/green/blue representing 160 and 70 μ\mum HSO and 24 μ\mum SST data respectively. The × marks the location of the driving star and the vectors depict the morphological parameters R0R_0 and ℓ\ell.

Another required parameter is the velocity of the star relative to the surrounding ISM, VaV_a. K19 and K22 estimated this quantity using proper motion/parallax data from Gaia DR2 and EDR3, respectively. However, proper motions must be corrected for solar motion relative to the Local Stand of Rest (LSR) and the adoption of a Galactic circular velocity as a function of Galactocentric distance (i.e., a galactic rotation curve). K22 performed this analysis on bow shock stars with available proper motions and parallaxes as well as on a control sample of non-bow shock GOSC stars. M. Carretero-Castrillo et al. (2023, hereafter C23) performed a similar runaway analysis of Galactic O stars using Gaia DR3 proper-motion data with some differences in their methodology compared to K22. C23 applied the zero-point correction to the Gaia parallaxes (described in [61]) while K22 did not. C23 also used a different rotation curve and solar motion with respect to the LSR to compute the peculiar 2D space velocity.

Figure 2 plots the ratio of the C23 to the K22 peculiar 2D space velocities for the overlapping subsample of 23 GOSC stars analyzed in both works as a function of the K22 values. The horizontal dashed black line shows the median ratio of 0.92. The grey band encompasses the RMS of 0.12. Overall, there is good agreement between the two independent determinations of velocity. Because of a difference in methodology, C23 reported slightly higher uncertainties than K22. The error bars in Figure 2 represent the K22 and C23 uncertainties added in quadrature. Low-velocity stars on the left side of the plot have a typical uncertainty of 5 km s−1^{-1}, which corresponds to 20%, while the fastest stars at 90 km s−1^{-1} have the same nominal uncertainty, amounting to only 7%. The differences between K22 and C23 do not correlate with galactic longitude as might be expected if systematic differences in the galactic rotation curve were responsible. The difference between the K22 and C23 values could arise from systematics such as the choice of mean solar motion with respect to the LSR or the application of the zero-point offset correction. Nevertheless, the K22 and C23 velocities correlate with each other and are in agreement within the reported uncertainties.

Figure 2. The ratio of the C23 to the K22 peculiar space velocities as a function of K22 peculiar space velocity for the 23 GOSC stars studied in both works. The dashed line and the shaded region represent the median and RMS of the ratio.

In Equation (2), we take VWV_{\mathrm{W}} to be the terminal wind speed of the stellar outflows v∞v_{\infty}. The terminal wind speed is measurable from the P-Cygni profiles of UV metal-resonance lines, which all OB stars exhibit. M. R. Mokiem et al. [68] tabulated measurements from UV-based spectroscopy of 49 Galactic O- and early B-type stars with effective temperatures ranging from 25–47 kK (Table A.1 in their Appendix). They reported terminal wind speeds ranging from 1500 km s−1^{-1} for the coolest stars to 3200 km s−1^{-1} for the hottest stars. Since UV spectra do not exist for our sample of bow shock stars2, K18/K19 used the results of M. R. Mokiem et al. [68] to estimate the terminal wind speed for their sample, with TeffT_{\mathrm{eff}} approximated from rough spectral types in the literature. In this work, we adopt the v∞v_{\infty}–TeffT_{\mathrm{eff}} calibration of C. Hawcroft et al. [38], which used 66 Galactic OB stars spanning 15 kK – 45 kK from the UV spectroscopic study of I. D. Howarth & R. K. Prinja [45] with updated stellar parameters provided by G. Holgado et al. [44]. The C. Hawcroft et al. [38] relation shows little-to-no dependence on luminosity class—no v∞v_{\infty} dependence on log⁡g\log g. Their relation shows an RMS of 277 km s−1^{-1}. We add this RMS in quadrature to the error produced by the uncertainty in temperature measured in P25.

For distance DD we adopt the inverse parallactic distance provided by Gaia DR3 without any zero-point correction given the large distances to these stars.

We measure the specific intensity at 70 μ\mum IνI_{\nu} from archival HSO 70 μ\mum UNIMAPB level 2.5 images from the Photodetector Array Camera and Spectrometer (PACS) photometer. In cases where there is structure in the bow shock or confounding background emission, we find it necessary to smooth the images by 1–3 pixels to increase the signal-to-noise ratio. In most cases, the peak specific intensity at 70 μ\mum coincides with the peak specific intensity at 24 μ\mum. However in some instances where the field has contaminating background and foreground emission, the peak location at 70 μ\mum is unclear. In such cases, we assume the two peaks are coincident and use the peak location at 24 μ\mum to measure the peak at 70 μ\mum.3

Using adjacent regions on either side of the peak along the line of R0R_0, we estimate the background flux by fitting a linear baseline. We subtract this estimate of the background from the peak brightness to measure the background-subtracted specific intensity IνI_{\nu}. We adopt an uncertainty on IνI_{\nu} of 20% based on the typical fluctuations in the background signal.

Figure 3 (left panel) displays a three-color representation of the bow shock nebula surrounding BS386⋆ in K16 in the same color mapping as in Figure 1. The right panel displays the specific intensity of the bow shock nebula along the R0R_0 vector, relative to the stellar source at 0, with the left axis corresponding to the 22 μ\mum data in units of W4 DN and the right axis corresponding to the 70 μ\mum data in units Jy pix−2^{-2}. The 22 μ\mum peak brightness is 208 DN, and the 70 μ\mum peak brightness is 1.25×10−21.25 \times10^{-2} Jy pix−2^{-2} for this nebula. The peak at 70 μ\mum is coincident with the peak at 22 μ\mum.

Figure 3. Left: Three color image of bow shock BS386 (surrounding the stellar X-ray source Vela X-1) with red/green/blue representing 160 and 70 μ\mum HSO and 22 μ\mum WISE data, respectively. Right: The brightness of the nebula along R0R_0, relative to the stellar source at 0, in 22 μ\mum (blue) and 70 μ\mum (green).

Most of the parameters in Equation (2) are readily measurable from optical and infrared data and require few assumptions. The major exception to this is jνj_{\nu}, which must rely on an adopted set of theoretical properties for dust grains. We adopt the B. T. Draine et al. [27] models to compute jνj_{\nu}. They modeled the infrared dust emission as a function of different radiation field intensities UU, heating stellar populations, and dust compositions. UU is defined as

U=γ⋆u⋆umMMP,(3)U = \gamma_{\star} \frac{u_{\star}}{u_{\mathrm{mMMP}}}, \tag*{(3)}

where γ⋆\gamma_{\star} is a small model-dependent factor of order 1–5, and umMMPu_{\mathrm{mMMP}} is the modified mean interstellar radiation field (ISRF) used by D21 (umMMP83=3.1293×10−2u_{\mathrm{mMMP83}} = 3.1293 \times10^{-2} erg s−1^{-1} cm−2^{-2}/c), which in turn is based on the mean ISRF energy density calculated by J. S. Mathis et al. [66]. We compute u∗u_\ast—the energy density at the apex of the infrared nebula—from

u∗=R∗2σTeff4R02c,(4)u_\ast= \frac{R_\ast^2 \sigma T_{\mathrm{eff}}^4}{R_0^2 c}, \tag*{(4)}

where R∗R_\ast is the radius of the star, TeffT_{\mathrm{eff}} is the effective temperature of the star, and R0R_0 is the standoff distance. D21 calculated the infrared emission spectra for three different irradiating spectral energy distributions: young stellar populations of different ages, an updated ISRF, and M31 bulge stars. They used the BPASS (J. J. Eldridge et al. [28]; E. R. Stanway & J. J. Eldridge [85]) and BC03 (G. Bruzual & S. Charlot [16]) stellar evolution models to provide the input spectral energy distributions for the young star populations. They considered several different dust compositions that included different PAH ionization distributions (high, standard, and low) and particle size distributions (large, standard, and small). We adopt their emissivities calculated from the BPASS models at solar metallicity and age 10 Myr to best reflect the stars at the center of bow shock nebulae. We also select the models with standard grain size and ionization distributions. We interpolate between models to obtain jνj_\nu as a function of UU. Our approach differs from the one adopted in K19 in that we use the model jνj_\nu created specifically for young stellar populations, so we do not require the ad-hoc correction suggested by W. J. Henney & S. J. Arthur [39].

We consider whether the choice of model makes a difference on dust emissivity jνj_\nu at 70 μ\mum by plotting in Figure 4 a normalized jνj_\nu versus U/γ∗U/\gamma_\ast for a subset set of dust emission models from D21 and B. T. Draine & A. Li [26]. Each set of models is normalized to the dust emissivity from the D21 BPASS model at age 10 Myr and solar metallicity. The figure shows the effects of different heating efficiencies of dust grains for different age stellar populations. The yellow track with filled triangles denotes the D21 BC03 model at 10 Myr and solar metallicity. It lies near unity at every irradiation intensity, indicating excellent agreement between the D21 BPASS and D21 BC03 models at the same age and metallicity. The other three models (DL07 ISRF—purple circles, D21 ISRF—pink up side down triangles, and D21 M31 bulge—cyan +'s) have ratios substantially less than 1.0 for some regimes (with ratios ranging from 0.2–0.8) showing emissivities of grains heated by old stellar populations depart significantly from emissivities of grains heated by young stellar populations. The DL07 ISRF and D21 ISRF models generally show good agreement with each other and are factors of 1.5–3 lower than the D21 BPASS model, with larger deviations on the lower U/γ⋆U/\gamma_\star end. This factor of 1.5–3 difference represents the ad-hoc correction K19 applied to the DL07 ISRF dust emissivities. We expect that the D21 BPASS model—tailored specifically for irradiation by young stellar populations—to better reflect the infrared dust properties of bow shock nebulae around early type stars. Accordingly, we adopt the D21 BPASS model emissivities as most suitable for use in this work.

Normalized dust emissivities at 70 µm for several different D21 and DL07 models

Figure 4. Normalized dust emissivities at 70 µm for several different D21 and DL07 models as a function of irradiation parameter U/γ⋆U/\gamma_\star.

We also consider differences that could arise in our choice of ages for the irradiating stellar populations. Figure 5 plots the dust emissivity calculated at 70 µm heated by single-age D21 BPASS models, for isochrone ages 3, 10, 100 and 300 Myr, versus radiation intensity UU. The calculated dust emissivities are normalized to the 10 Myr model. All emissivities agree to within ∼5%\sim5\% over the range of 3≤log⁡U≤53 \le\log U \le5. Relatively larger deviations are exhibited for the 100 and 300 Myr models on the low-UU end, with differences exceeding 12%. We find that while the choice of age matters in obtaining jνj_\nu, differences are negligible between the 3 and 10 Myr models. Since the central stars to bow shock nebulae are almost exclusively OB stars ([19 P25]), the 3 and 10 Myr comparisons are most pertinent to this study, so we adopt the 10 Myr model. The other model free parameters—grain size and ionization distributions—affect the dust emissivity to much smaller degrees.

Figure 5. Normalized dust emissivities at 70 µm for the D21 BC03 models at solar metallicity versus irradiation parameter UU for different stellar population ages.

Note—(1) Stellar bow shock nebulae identifier from the K16/J19 catalogs; (2) Projected stand-off distance from A. S. Whisnant et al. [93] and K16; (3) The uncertainty in the measured stand-off distance; (4) The parallactic distance to the bow shock central star; (5) The propagated uncertainty in the parallactic distance; (6) The peculiar 2D space velocity of the bowshock central stars from K22; (7) The uncertainty in the computed peculiar 2D space velocity; (8) The background-subtracted specific intensity of the bow shock arc at 70 µm; (9) The logarithm of the irradiation parameter computed from Equations (3) and (4); (10) The interpolated terminal wind speed from the C. Hawcroft et al. [38] relation; (11) The chord length across the bow shock arc; (12) The calculated dust-emissivity per nucleon of the dust grains at the apex of the bow shock arc; (13) The calculated interstellar density from Equation 5 of K18; (14) bowshock morphology indicator; (15) PAH indicator; (16) Confidence flag, assessed from the flags in Columns (13), (14), and (15).

Table 2 presents all of the measured parameters needed to measure mass-loss rate M˙\dot{M}. Column (1) presents the bow shock identifier from the K16/J19 catalogs. Columns (2) and (3) list the projected standoff distance R0R_0 and its uncertainty in arcseconds, sourced from A. S. Whisnant et al. [93], when available, or K16. A. S. Whisnant et al. [93] provided the uncertainty in their citizen-science measurements on R0R_0 which we list in Column (3). For bow shocks with no citizen-science R0R_0 measurements, we assume a 10% uncertainty. Columns (4) and (5) present the Gaia DR3 parallactic distance and uncertainty in pc. Columns (6) and (7) present the peculiar 2D space velocity and its uncertainty in km s−1^{-1} from K22. Column (8) lists the measured background-subtracted specific intensity of the bow shock peak at 70 μ\mum in MJy sr−1^{-1}. It is worth noting that the HSO PACS photometer has two imaging modes resulting in maps with 3.′′2 or 1.′′6 pixels. Converting fluxes to standard intensity units requires proper recognition of the imaging mode used. Most of the 70 μ\mum images were acquired in the parallel mode (i.e., a pixel size of 3.′′2) but some notable targets4 were acquired with the 1.′′6 pixel mode.5 Column (9) lists the logarithm of the irradiation parameter UU, calculated from Equations (3) and (4) (with γ⋆=4.58\gamma_\star= 4.58 from D21). Column (10) lists the terminal wind speed in km s−1^{-1}. Column (11) lists the measured chord length ℓ\ell in arcseconds. Column (12) lists the log of the interpolated dust emissivity in 70 μ\mum for dust grains at the apex of the nebula jνj_\nu in Jy cm2^2 sr−1^{-1} nucleon−1^{-1} from the D21 BPASS models at 10 Myr and solar metallicity. Column (13) lists the calculated ambient interstellar density computed from na=Iν/(4jν)n_a = I_\nu/(4j_\nu)6 as in K18. Due to the difference in pixel sizes and background subtraction, our calculated densities are lower than those found in K18, by at least a factor of 4. Furthermore, K18 claimed 70 μ\mum detections in some cases where we determine only upper limit. We find nan_a ranges from 0.4–392 cm−3^{-3} and has a median value of 2.2 cm−3^{-3}. This is markedly lower than the median nan_a of 16 cm−3^{-3} reported by K18. The lower values of nan_a in this study challenge one of the findings of K18, in which the authors concluded that ambient densities of at least na≳5n_a \gtrsim5 cm−3^{-3} were required to form bow shocks.

IDR0R_0 (′′)σR0\sigma_{R_0} (′′)DD (pc)σD\sigma_D (pc)V2DV_{2D} (km s−1^{-1})σV2D\sigma_{V_{2D}} (km s−1^{-1})IνI_\nu (MJy sr−1^{-1})log⁡U\log U (dex)VwV_w (km s−1^{-1})ℓ\ell (′′)log⁡jν\log j_\nu (Jy cm2^2 sr−1^{-1} nucleon−1^{-1})nan_a (cm−3^{-3})Morph.PAHConf.
(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)
BS00127.774.6128511441926153.66217446.10-11.1912.20YYH
BS00711.131.5712172511116603.4997611.20-11.29392.00YYL
BS0103.100.312462101192815.07153841.80-10.560.48NNL
BS013299.0029.9013512233973.921979352.00-11.063.90YYH
BS01626.541.093069186521263.63149620.70-11.215.33YNH
BS01911.660.8931981742622994.65119912.80-10.736.55YNH
BS0268.600.763270154821894.5416069.35-10.786.15YNH
BS0277.540.90180249622133.879986.68-11.0835.90YNL
BS0436.800.68224822262614.4414448.82-10.823.38YNH
BS0512.500.251418453152806.3621722.80-10.11283.00YYL
BS05628.002.8051449843232283.56142917.20-11.240.93YNH
BS06122.441.5520781571027694.23182017.70-10.9128.70NNL
BS0633.500.351758565122805.34187036.70-10.4616.90YYL
BS0656.102.28181456315294.068624.69-10.99102.00YNL
BS0678.470.991861664120904.7519848.14-10.69112.00YYL
BS15012.101.7861310828104.96126337.90-10.6023.30NYH
BS20010.130.94452345234131004.56126411.10-10.771.94YNL
BS2552.000.201272251921495.237414.11-10.5015.10NNH
BS28811.761.062911108112723.81147829.50-11.111.82YNH
BS3006.900.6920534741434.45167532.30-10.810.71YNH
BS3024.900.49245215972374.3294024.40-10.870.76YNH
BS3037.800.781145524415252544.247637.61-10.910.84NNH
BS3054.900.491922329955825.42211921.70-10.436.24NNH
BS30613.673.4121806371484.01147934.70-11.021.10YNL
BS3145.400.544525297112594.36140820.00-10.860.78YNH
BS32035.503.55317916672813.68153130.90-11.182.11NYH
BS32110.501.05474825311113.76185412.70-11.140.43NYH
BS32219.501.95491624931503.64203120.30-11.211.35YNH
BS32425.702.575325303302273.50200134.90-11.280.46YNH
BS32512.901.29131838122244.5090016.30-10.791.14YNH
ID (1)R0R_0 (′′) (2)σR0\sigma_{R_0} (′′) (3)DD (pc) (4)σD\sigma_D (pc) (5)V2DV_{2D} (km s−1^{-1}) (6)σV2D\sigma_{V_{2D}} (km s−1^{-1}) (7)IνI_\nu (MJy sr−1^{-1}) (8)log⁡U\log U (dex) (9)VwV_w (km s−1^{-1}) (10)ℓ\ell (′′) (11)log⁡jν\log j_\nu (Jy cm2^2 sr−1^{-1} nucleon−1^{-1}) (12)nan_a (cm−3^{-3}) (13)Morph. (14)PAH (15)Conf. (16)
BS33025.502.55171152911214.2898553.00-10.891.73YNH
HBS33145.604.5616913942984.10224335.00-10.972.61YNH
BS33228.202.821732425112203.78198124.80-11.1364.30NNL
BS33515.601.562012491221384.27118418.80-10.904.80NNH
BS33970.007.001582457733844.21335354.30-10.926.22NYL
BS34015.901.59147345192414.26133737.60-10.900.97YYL
BS34120.202.02177048112904.48207625.20-10.802.13YNH
BS344193.0019.301717364526983.563470193.00-11.246.18YYH
BS35023.002.302316433102273.77117045.10-11.140.60YNH
BS35330.503.054645233674343.60254526.90-11.230.77YNH
BS35615.601.56279197102223.93137513.60-11.061.10YNH
BS35721.802.18273827282243.75116633.60-11.150.62YNH
BS3588.900.89868921203.9979410.80-11.033.82YNH
BS36191.709.174564267292192.76106380.40-11.740.48YNH
BS36214.701.473894174272333.70125620.80-11.171.00YYH
BS36529.902.99105819112233.90143937.00-11.071.15YYH
BS36736.003.6021005982553.67201340.20-11.191.67YYH
BS36814.001.40198162811994.34199932.30-10.863.81YYH
BS36973.007.30196119393503.33250572.20-11.381.43YYH
BS381236.0023.601382100113382.922769141.00-11.631.41YNH
BS38431.003.103421142422283.59230826.10-11.237.24YYH
BS38653.005.302015624821633.88120289.70-11.081.82YYH
BS4077.800.78252111210218904.93204618.80-10.6227.60YYH
BS4097.400.7429212061927784.73214813.30-10.7016.60YYH
BS4145.800.58235081711084.80142310.30-10.673.47YYH
BS63432.611.7116044592753.92193441.70-11.062.15YNH
BS6406.950.681561651224994.94148620.80-10.6110.50YNH
BS66716.471.431003362926434.71172428.50-10.7119.10YNH
BS6736.000.6038823072131504.1714607.07-10.947.99YNH
BS6879.290.7126931522031314.7714678.85-10.684.37YNH
BS69232.082.0917946217115104.05244952.70-11.0026.70YYH
BS70015.801.583868310163664.20145813.80-10.931.74YNH
IDR0R_0 (″)σR0\sigma_{R_0} (″)DD (pc)σD\sigma_D (pc)V2DV_{2D} (km s−1^{-1})σV2D\sigma_{V_{2D}} (km s−1^{-1})IνI_\nu (MJy sr−1^{-1})log⁡U\log U (dex)VwV_w (km s−1^{-1})ℓ\ell (″)log⁡jν\log j_\nu (Jy cm2^2 sr−1^{-1} nucleon−1^{-1})nan_a (cm−3^{-3})Morph.PAHConf.
BS70526.521.04234910024315703.74146526.40-11.1560.10YNH
H2G0063185-006662234.503.4527052218210604.10114624.90-10.9724.70YYH
H2G0065094-006134411.951.16198392536984.02154810.30-11.0158.70YYH
H2G0118310-00629909.700.9718206752954.3310457.15-10.879.00YNH
H2G0592097+001251917.653.72170810042904.91137515.20-10.622.44NNL
H2G0600615-001834020.132.0220905881373.73174417.90-11.162.37YNH
H2G0770694+019356819.601.9614299542623.99177039.90-11.031.93YNH
H2G1045666+012808516.301.63223888421264.42226129.40-10.832.16NYH
κ\kappa Cas158.0015.80939144235353.741028115.00-11.150.76YYH
H2G3525610+000188713.511.41185566163404.21142429.50-10.921.03YNH
H2G3533026+000809556.634.90179256241563.91273677.20-11.060.78NYL

Table 2. Mass-loss parameters for bow shock stars. Table 2 continued.

W. J. Henney & S. J. Arthur [40] provided a taxonomy of arc-like, infrared nebulae based on the coupling between gas and dust and the optical depth of the shell. They classified nebulae in which the gas and dust are coupled with low optical depths as wind-supported bow shocks, and they found that this scenario arises around stars with a broad range of stellar parameters and interstellar ambient densities na<100n_a < 100 cm−3^{-3}. W. J. Henney & S. J. Arthur [41] selected 23 sources from H. A. Kobulnicky et al. [52] and found all but one met the criteria of wind-supported bow shocks, confirming most arc-like nebulae are indeed bow shocks. Since we also select a subset of stars from K16, we expect most of our sources are wind-supported bow shocks as well. We note that four sources in Table 2 have na>100n_a > 100 cm−3^{-3} (i.e., BS007, BS051, BS065, and BS067), so we consider these may not be wind-supported bow shocks. Hence, any measurements of M˙\dot{M} for these sources may not be reliable.7 Some of the bow shock candidates cataloged in K16 may be interstellar bubbles, misidentified as bow shocks due to viewing angle effects (e.g. E. Churchwell et al. [20]). Additionally, episodic mass-loss events for evolved stars can produce nebulae that are not steady-state shock fronts formed from stellar winds. In light of this, it is unavoidable that some fraction of the nebulae cataloged in K16 and J19 are not formed from the stellar winds of a hot star. We aim to distinguish stellar bow shocks from morphologically similar phenomena on the basis of three criteria. First, we verify that our computed ambient densities (Column (13) of Table 2) are consistent with wind-supported bow shock nebulae, and not other classes if arc-like infrared nebulae as discussed in W. J. Henney & S. J. Arthur [40], by flagging as suspect objects with na>100 cm−3n_a > 100\ \mathrm{cm}^{-3}. Second, we ensure the morphology of the bow shock—from SST 24 μ\mum/WISE 22 μ\mum images—is consistent with a steady-state stellar bow shock and discount ring-like and non-arc-like structures. We tabulate our results of morphology re-examination in Column (14) of Table 2, with “Y” signifying arcuate structure and “N” signifying non-arcuate structure. Third, we search for indications of bubbles by searching for PAH emission along the edge of the nebula—consistent with the photo-dissociation front of a bubble—by examining SST Infrared Array Camera 8.0 μ\mum(Capak; Peter [17])/WISE W3 12 μ\mum images. We note, however, that these passbands also contain warm dust emission. We tabulate our results of PAH indicators in Column (15) of Table 2, with “Y” signifying possible PAH emission and “N” signifying no PAH emission. We consider all three of these criteria for all objects, and assign a confidence indicator, either high (H) or low (L), based on these analyses. We tabulate the confidence indicator in Column (16) of Table 2.

Figure 6 displays infrared images of four selected bow shock nebulae that demonstrate the conditions for the morphology and PAH criteria. All panels in Figure 6 display images from SST with red/green/blue representing 24/8.0/4.5 μ\mum data. The top row presents a bow shock nebula that satisfies our morphology re-examination, and one that fails. BS016 (upper left) is well-defined in 24 μ\mum, and has a clear, arcuate structure. There is no contaminating background emission at 24 μ\mum, so we assign a “Y” for this bow shock’s morphology indicator. BS332 (upper right), on the other hand, has a more dubious structure. There is a poorly defined bow shock arc surrounding the star, and there is also significant foreground/background signal in the 24 μ\mum data. The dubious morphology raises the possibility that this nebula does not represent a stellar-wind bow shock, so we assign an “N” flag for morphology. The bottom row of Figure 6 presents a bow shock where there is no PAH emission (BS341) and one where there is possible PAH emission (BS340), indicated by the 8.0 μ\mum data (green). BS341 (bottom left) shows a clear bow shock structure in 24 μ\mum and no PAH signal coincident with the arc, so we assign an “N” flag for the PAH indicator for this nebula. Conversely, BS340 (bottom right) shows a clear bow shock structure and 8.0 μ\mum emission along the arc, giving it an orange/red appearance, so we assign a “Y” flag for its PAH indicator.

Four images of bow shock nebulae

Figure 6. Four images of bow shock nebulae images from SST with red/green/blue representing 24/8.0/4.5 μm\mu\mathrm{m} data. BS016 is a high-confidence nebula because of its morphology, BS332 is low confidence because of irregular morphology, 341 is high confidence because of its morphology and no apparent PAH emission along the shock, and 340 is low confidence because of the strong PAH emission along the bow shock arc.

Figure 7 displays a four panel plot of derived parameters versus stellar effective temperature TeffT_{\mathrm{eff}}. Temperatures ranges from 19.8 kK on the cool end to 46.8 kK for the hottest star with median temperature 27.3 kK. The top panel plots the logarithm of the dimensionless irra... diation parameter UU. log⁡U\log U ranges from 2.8 to 6.4 with a median of 3.8. log⁡U\log U does not correlate with stellar effective temperature, contrary to what might be expected given the greater radiant luminosity of hot stars. The lack of correlation in the top panel can be explained by the correlation in the second panel, which plots the standoff distance R0R_0 in parsecs versus TeffT_{\mathrm{eff}}. R0R_0 in parsecs ranges from 0.01–2.03 pc with median 0.18 pc. The second panel shows that hotter stars have larger standoff distances, which will reduce the irradiation parameter UU at the apex of the bow shock. The larger standoff distances for hotter stars can be understood as a greater momentum flux of the wind. We note that there are several outliers in the second panel and label them. The third panel from the top plots the logarithm of the specific inten-

Figure 7. Distributions of the irradiation energy parameter log⁡U\log U, dust emissivity at 70 μ\mum log⁡jν\log j_{\nu}, specific intensity at 70 μ\mum log⁡Iν\log I_{\nu}, and stellar radius in solar radii, versus stellar effective temperature TeffT_{\rm eff} in kK.

sity at 70 μ\mum IνI_{\nu} in Jy sr−1^{-1}. Smaller points denote bowshock nebulae that were not detectable above the background at 70 μ\mum, which we treat as upper limits on IνI_{\nu}. We find no upper limit cases for stars hotter than 35 kK. IνI_{\nu} ranges from 107.010^{7.0}–109.710^{9.7} Jy sr−1^{-1} with a median value of 108.010^{8.0} Jy sr−1^{-1}. Specific intensity shows no correlation with effective temperature, unsurprisingly, for the same reason as the lack of correlation in the top panel. The bottom panel plots stellar radius versus temperature. The stellar radii range from 3.6–68.4 R⊙R_{\odot} with a median value of 11.5 R⊙R_{\odot}. The data show a correlation between the two parameters, with a bifurcation below Teff≈27T_{\rm eff} \approx27 kK. The upper branch corresponds to evolved stars while the lower branch corresponds to main-sequence stars. At hotter temperatures the trends converge, signaling that the hottest stars are main-sequence stars.

Figure 8 is a five-panel plot of derived parameters versus parallactic distance DD to consider whether there are any biases with distance. Excluding one object at 11.5 kpc, distances range from 130 pc to 5.3 kpc with a median distance of 2.1 kpc. The top panel plots the projected standoff distance R0R_0 in parsecs, which ranges from 0.01–2.03 pc, and has a median of 0.18 pc. The second panel plots the projected standoff distance R0R_0 in arcseconds versus distance. Excluding three anomalously large bow shock nebulae with projected standoff distances exceeding 100″, R0R_0 ranges from 2″.0–91″.7 with a median of 15″.8. Most bow shocks were identified in K16 at 24 μ\mum from SST images, which have a nominal resolution of 6″.0, and we expect bow shock nebulae with standoff distances R0≲3′′R_0 \lesssim3'' to be systematically underrepresented in the K16 master sample. We plot this minimum resolvable projected standoff distance as a horizontal dashed black line in the second panel. Consequently, the median bow shock nebula with an R0=0.18R_0 = 0.18 pc should be resolvable out to 12 kpc, but the smallest ones (R0=0.01R_0 = 0.01 pc) would become missing beyond 1.2 kpc. The third panel plots the chord length of the bow shock nebulae. Excluding large bow shock nebulae with ℓ>100′′\ell> 100'', we find chord lengths ranging from 2″.8–90″ with median of 24″.9. In a similar fashion as the second panel, we plot the minimum measurable chord length due to the resolution of SST at 24 μ\mum as a horizontal dashed black line. The fourth panel plots the logarithm of the specific intensity at 70 μ\mum in Jy sr−1^{-1}. Smaller points represent nondetections (which we treat as upper limits on IνI_{\nu}) and larger points denote detections above the background. log⁡Iν\log I_{\nu} ranges from 7.1–9.7 with a median of 8.0. At distances D<3.5D < 3.5 kpc the full range of specific intensities is present. Nearly 2/3 of these nearby bow shocks with log⁡Iν>8.5\log I_{\nu} > 8.5 are upper limits. However, beyond 3.5 kpc, there is a paucity of sources with log⁡Iν>8.5\log I_{\nu} > 8.5. We cannot attribute this lack of high IνI_{\nu}, distant sources to confusing foreground and background contamination, since the most-distant sources are mostly confirmed detections. We also cannot attribute this lack to beam dilution as these nebulae exceed the angular resolution of both SST (6″) and HSO (5.5″)—by factors of several—having an angular size of 20″, consistent with the rest of the sample. The absence of high IνI_{\nu} distant targets in our sample is apparent, although do not have an explanation. The bottom panel plots the logarithm of stellar luminosity L∗L_{\ast} in solar luminosity units. log⁡L∗\log L_{\ast} ranges from 3.4–6.3, with median 5.0. The lack of low-luminosity stars beyond 2 kpc can readily be explained by the combination of distance and extinction, which act to push these stars beyond our spectroscopy limit of Gaia BP≲15B_P \lesssim15 mag. An extended bow shock study would need more-sensitive blue-optical spectra to include these fainter bow shock-driving stars.

Figure 8. Distributions of the projected standoff distance R0R_0 in both parsecs and arcseconds, chord length ℓ\ell in arcseconds, log surface brightness IνI_{\nu} in Jy sr−1^{-1}, and log stellar luminosity L∗L_{\ast} in solar luminosity units versus the parallactic distance DD, in kpc.

for bow shock stars

Table 3 presents the results of our computations of mass-loss rates for 74 bow shock stars. Column (1) lists the bow shock identifier, either from the K16 or J19 catalogs. Column (2) lists the stellar mass of the bow shock central star in solar masses. Column (3) lists the logarithm of stellar luminosity of the bow shock central star in solar luminosity units. Column (4) presents the 2D peculiar space velocity of the bow shock central star in km s−1^{-1} from K22 and calculated from proper-motion data. Column (5) presents the logarithm of the calculated mass-loss rate of the bow shock central star using the parameters listed in Table 2 and Equation (2). Column (6) presents the corrected mass-loss rate M˙20\dot{M}_{20}—assuming VaV_a takes a minimum value of 20 km s−1^{-1}—which is expounded upon later in this section. Column (7) presents the random uncertainty in the calculated mass-loss rates in columns (5) and (6) using standard error propagation and given uncertainties of parameters in Equation (2). Columns (8) and (9) list the predicted mass-loss rate of each tabulated star using the analytical formulae of J. S. Vink et al. [90] ( eq:24 and 25, V01 hereafter) and R. Björklund et al. [13] (eq:7, B23 hereafter).

IDM⋆M_\star (M⊙M_\odot)log⁡(L⋆/L⊙)\log(L_\star/L_\odot) (dex)V2DV_{2D} (km s−1^{-1})log⁡M˙\log\dot{M} (M⊙M_\odot yr−1^{-1})log⁡M˙20\log\dot{M}_{20} (M⊙M_\odot yr−1^{-1})σlog⁡M˙\sigma_{\log\dot{M}} (M⊙M_\odot yr−1^{-1})log⁡M˙Vink\log\dot{M}_{\mathrm{Vink}} (M⊙M_\odot yr−1^{-1})log⁡M˙Bjorklund\log\dot{M}_{\mathrm{Bjorklund}} (M⊙M_\odot yr−1^{-1})
(1)(2)(3)(4)(5)(6)(7)(8)(9)
BS001⋆_\star53.15.1319-6.90-6.850.19-6.98-7.74
BS007⋆_\star7.83.4411-7.05-6.530.17-9.15-10.94
BS010⋆_\star13.64.4419-10.18-10.140.15-8.01-8.69
BS013⋆_\star14.94.8223-7.77-7.770.17-7.01-7.72
BS016⋆_\star34.85.195-8.23-7.020.36-6.93-7.52
BS019⋆_\star78.05.6526-7.29-7.290.13-5.54-7.03
BS026⋆_\star72.05.168-8.71-7.920.25-7.34-8.03
BS027⋆_\star9.43.866-8.62-7.580.32-8.31-10.03
BS043⋆_\star14.24.536-9.71-8.660.32-6.99-8.54
BS051⋆_\star95.05.103-9.83-8.180.31-7.38-8.19
BS056⋆_\star146.05.6432-6.86-6.860.88-5.89-7.38
BS061⋆_\star64.95.2610-7.47-6.860.21-6.93-7.64
BS063⋆_\star58.04.595-10.07-8.860.21-8.32-9.16
BS065⋆_\star9.13.823-8.88-7.230.44-8.40-10.15
BS067⋆_\star59.64.854-8.65-7.250.26-7.73-8.54
BS150⋆_\star16.14.448-9.19-8.390.27-7.30-8.90
BS200⋆_\star80.05.5734-7.43-7.430.36-5.73-7.23
BS255⋆_\star8.53.7219-9.32-9.280.15-8.61-10.41
BS288⋆_\star18.54.6311-8.76-8.240.20-7.82-8.47
BS300⋆_\star17.04.474-10.87-9.470.25-7.98-8.71
BS302⋆_\star12.94.227-10.24-9.330.28-7.71-9.40
BS303⋆_\star85.05.9052-6.62-6.620.88-5.14-6.68
BS305⋆_\star93.05.089-9.67-8.980.50-7.43-8.24
BS306⋆_\star19.14.727-9.49-8.580.26-7.62-8.25
BS314⋆_\star31.04.8511-9.40-8.880.20-6.74-8.28
BS320⋆_\star54.55.487-8.07-7.150.28-6.53-7.11
BS321⋆_\star65.14.901-11.24-8.640.88-7.72-8.51
BS322⋆_\star74.05.313-9.26-7.610.32-6.81-7.55
BS324⋆_\star114.05.4930-7.41-7.410.14-6.69-7.42
BS325⋆_\star18.74.6712-9.28-8.830.19-6.94-8.57
BS330⋆_\star30.95.259-8.57-7.870.16-5.94-7.49
BS331⋆_\star89.04.844-8.95-7.560.45-7.89-8.75
BS332⋆_\star15.84.845-7.71-6.510.21-6.99-7.70
BS335⋆_\star21.04.9212-8.24-7.800.19-6.41-7.96
BS339⋆_\star89.06.0077-5.87-5.870.13-5.23-5.73
BS340⋆_\star17.44.6819-8.84-8.800.15-6.79-8.34
BS341⋆_\star91.05.2911-8.80-8.280.20-6.96-7.72
BS344⋆_\star104.06.3145-5.40-5.400.13-4.64-5.06
BS350⋆_\star22.14.9010-8.84-8.240.23-6.48-8.04
BS353⋆_\star73.05.6167-6.56-6.560.14-6.02-6.70
BS356⋆_\star19.34.8610-8.82-8.220.21-6.45-7.96
BS357⋆_\star23.84.998-8.91-8.120.25-6.32-7.86
BS358⋆_\star6.83.422-10.94-8.940.45-9.15-10.97
BS361⋆_\star48.65.7829-6.18-6.180.14-5.00-6.47
BS362⋆_\star59.45.0227-7.72-7.720.14-6.77-8.35
IDM⋆M_\star (M⊙M_\odot)log⁡(L⋆/L⊙)\log(L_\star/L_\odot) (dex)V2DV_{2D} (km s−1^{-1})log⁡M˙\log\dot{M} (M⊙M_\odot yr−1^{-1})log⁡M˙20\log\dot{M}_{20} (M⊙M_\odot yr−1^{-1})σlog⁡M˙\sigma_{\log\dot{M}} (M⊙M_\odot yr−1^{-1})log⁡M˙Vink\log\dot{M}_{\mathrm{Vink}} (M⊙M_\odot yr−1^{-1})log⁡M˙Bjorklund\log\dot{M}_{\mathrm{Bjorklund}} (M⊙M_\odot yr−1^{-1})
BS365⋆_\star16.04.5811-9.01-8.490.20-6.95-8.50
BS367⋆_\star58.95.148-8.52-7.720.25-7.07-7.82
BS368⋆_\star53.44.938-9.03-8.230.16-7.49-8.27
BS369⋆_\star68.05.349-8.02-7.330.32-6.57-7.30
BS381⋆_\star69.05.6511-7.18-6.670.27-5.86-6.50
BS384⋆_\star81.05.364-8.25-6.850.45-6.67-7.42
BS386⋆_\star39.55.5948-6.40-6.400.13-5.29-6.75
BS407⋆_\star25.85.2110-8.28-7.680.21-6.43-7.11
BS409⋆_\star50.95.1019-7.88-7.840.16-7.02-7.78
BS414⋆_\star15.74.677-9.65-8.740.18-6.74-8.27
BS634⋆_\star65.95.069-8.61-7.910.22-7.33-8.09
BS640⋆_\star71.04.7412-8.92-8.480.19-8.34-9.07
BS667⋆_\star55.84.8529-7.60-7.600.13-7.82-8.57
BS673⋆_\star14.44.5721-7.88-7.880.18-7.81-8.43
BS687⋆_\star78.05.2820-8.13-8.130.17-7.21-7.83
BS692⋆_\star98.05.2817-6.98-6.840.12-6.93-7.70
BS700⋆_\star84.05.6216-7.94-7.750.21-6.51-7.06
BS705⋆_\star25.55.0424-6.04-6.040.14-7.11-7.69
2G0063185-0066622⋆_\star68.05.768-6.92-6.120.25-5.23-6.70
2G0065094-0061344⋆_\star15.14.455-8.27-7.070.54-8.05-8.74
2G0118310-0062990⋆_\star15.64.505-9.17-7.970.37-7.19-8.82
2G0592097+0012519⋆_\star145.05.624-9.59-8.190.48-5.94-7.44
2G0600615-0018340⋆_\star20.34.698-8.81-8.020.16-7.57-8.29
2G0770694+0193568⋆_\star19.24.604-9.86-8.470.45-7.73-8.47
2G1045666+0128085⋆_\star111.05.254-9.69-8.290.45-7.12-7.91
κ\kappa Cas60.05.7823-7.06-7.060.23-5.15-6.63
2G3525610+0001887⋆_\star21.24.7516-8.93-8.740.21-6.74-8.28
2G3533026+0008095⋆_\star67.05.6424-7.77-7.770.12-5.88-6.53

Table 3. M˙\dot{M} for bow shock stars. Table 3 continued

Figure 9 plots the logarithm of the mass-loss rates, in M⊙ yr−1M_\odot\,\mathrm{yr}^{-1}, versus the logarithm of stellar luminosity, in solar luminosity units, for 74 bow shock stars. Following the nomenclature of K16, we classify targets according to their apparent local environment. Blue circles represent isolated bow shock nebula (I, 41), red diamonds denote bow shock nebulae that face H II regions (FH, 13), and green ×’s represent bow shock nebulae that either face bright-rimmed clouds or lie within H II regions (FB/H, 20). These latter two flags are grouped together given the ambiguity in classification for a bow shock nebula on the edge of an H II region which faces a photodissociation front. K22 found that for runaway bow shock stars (V2D≥25 km s−1V_{2D} \ge25\ \mathrm{km}\ \mathrm{s}^{-1}) the morphological position angle of the IR nebula is highly aligned with the star’s peculiar motion vector, while for non-runaway stars (V2D<25 km s−1V_{2D} < 25\ \mathrm{km}\ \mathrm{s}^{-1}) the difference between these two vectors (ΔPA\Delta\mathrm{PA} in the notation of K22) is isotropic. The authors concluded that in the non-runaway cases—the majority of bow shock stars (76%)—bulk ISM flows played a more prominent role in forming the bow shock than the star’s space velocity. This can lead to an under-measurement of M˙\dot{M} since the bow shock technique assumes the star’s peculiar 2D space velocity is equivalent to the star’s velocity relative to the ambient material (VaV_a in Equation (2)). This is corroborated by J. van den Eijnden & L. Kaper (2026), who found that ignoring the effect of bulk ISM motions for stars with low space velocities (less than about 15 km s−1\mathrm{km}\ \mathrm{s}^{-1}) leads to a systematic underestimate in M˙\dot{M}. Based on these findings, we adopt hollow symbols to mark bow shock stars with space velocities V2D<20 km s−1V_{2D} < 20\ \mathrm{km}\ \mathrm{s}^{-1} (55 of 74 targets).

Figure 9. Mass-loss rates of bow shock stars plotted versus stellar luminosity. Symbols denote local environment as in the legend. Hollow points denote slow-moving stars. Large points denote high-confidence measurements. Orange lines signify prescriptions from J. Krtička (2014) and J. Krtička & J. Kubát (2017). The magenta track marks the models from R. Björklund et al. (2023). The black line marks the best-fit trend for measurements of M˙\dot{M} for stars with log⁡L⋆>4.5\log L_{\star} > 4.5.

Downward arrows mark upper limits for cases where no 70 μm\mu\mathrm{m} emission is detected from the bow shock above the background (30 targets). The blue star signifies the measurement for the runaway star ζ Oph. Orange lines signify theoretical prescriptions for B-main-sequence and O-dwarf, giant, and supergiant stars (J. Krtička [55]; J. Krtička & J. Kubát [56]). Magenta lines signify the solar-metallicity model calculations of M˙\dot{M} for dwarf, giant, and supergiant stars from R. Björklund et al. [12], tabulated in Table A.1 in their Appendix. The semi-transparent black line marks the best-fit trend to the data for targets with log⁡L∗>4.5\log L_* > 4.5. Text labels mark some outliers discussed in the text below. The point at the lower right illustrates the median random uncertainties in M˙\dot{M} and L∗L_*.

The points in Figure 9 show a strong correlation between mass-loss rate and luminosity. The slope of the best-fit trend has the same slope as the models but is offset downward by 0.8–1.0 dex. The upper envelope of the data—mostly consisting of large solid points indicating high-confidence measurements moving at substantial velocities—agrees well with both the R. Björklund et al. [12] and the J. Krtička & J. Kubát [56]/J. Krtička [55] predictions. The hollow points scatter below these models by 1–2 dex. Measurements of M˙\dot{M} for each of the three classes of environments are consistent with the distribution of the full sample, indicating that measurements of M˙\dot{M} are not significantly impacted by environment. Below log⁡L∗<5\log L_* < 5, the sample is increasingly dominated by upper limits. The upper-limits follow the same correlation as the measurements with detections, although with a higher dispersion. Low-confidence objects (small points) exhibit the same trend as high-confidence nebulae (large symbols) albeit with a larger dispersion. Measurements for M˙\dot{M} have an RMS of 0.85 dex about the best-fit line. The prototypical bow shock star ζ\zeta Ophiuchi (BS013⋆_\star) lies in the middle of the plot (blue star). ζ\zeta Oph is situated along the top of the distribution, helping define the upper envelope. It shows excellent agreement with the J. Krtička & J. Kubát (2017) and R. Björklund et al. (2021) models. We calculate a mass-loss rate of M˙=1.7×10−8 M⊙ yr−1\dot{M} = 1.7 \times10^{-8}\ M_{\odot}\ \mathrm{yr}^{-1} which is 40% higher than the measurement of K19. Our measurement for M˙\dot{M} is within 10% of the B23 analytical prescription8. The new measurement and B23 prediction are both lower than the V01 of 9.77×10−8 M⊙ yr−19.77 \times10^{-8}\ M_{\odot}\ \mathrm{yr}^{-1}. Measurements of M˙\dot{M} for ζ\zeta Oph vary greatly in the literature, and our measurement lies in the middle of other

values. On the low end, W. L. F. Marcolino et al. [63] measured a mass-loss rate of 1.58×10−9 M⊙ yr−11.58 \times10^{-9}\ M_{\odot}\ \mathrm{yr}^{-1}, which is a factor of 10 lower than our value. X-ray data yielded a similar value of 1.50×10−9 M⊙ yr−11.50 \times10^{-9}\ M_{\odot}\ \mathrm{yr}^{-1} (D. H. Cohen et al. [21]). Towards the higher end, the radio-continuum analysis of M. M. Rubio-Díez et al. [81] yielded an upper limit of M˙<1.8×10−7 M⊙ yr−1\dot{M} < 1.8 \times10^{-7}\ M_{\odot}\ \mathrm{yr}^{-1}, a factor of 10 higher than our value.

BS344⋆ (BD+43 3654) is a well-studied runaway star (V2D=45 km s−1V_{2D} = 45\ \mathrm{km}\ \mathrm{s}^{-1}) and is the most-luminous star in our sample (log⁡L∗=6.3\log L_{*} = 6.3, log⁡M˙=−5.4\log\dot{M} = -5.4). We measure a mass-loss rate of M˙=4.0×10−6 M⊙ yr−1\dot{M} = 4.0 \times10^{-6}\ M_{\odot}\ \mathrm{yr}^{-1}. This is a factor of five lower than K19 on account of the reduced 70 μm\mu\mathrm{m} specific intensity from rigorous background subtraction. BS344⋆ is more luminous than any of the plotted models but shows excellent agreement with an extrapolation of either the R. Björklund et al. [12] or J. Krtička & J. Kubát [56] predictions. Our measurement is a factor of seven lower than the V01 prediction and is a factor of two lower than the B23 prediction.

BS386⋆ (Vela X-1) is a runaway compact object+B1 supergiant binary system (log⁡L∗=5.6\log L_{*} = 5.6, log⁡M˙=−7.0\log\dot{M} = -7.0). We measure a mass-loss rate of M˙=1.0×10−7 M⊙ yr−1\dot{M} = 1.0 \times10^{-7}\ M_{\odot}\ \mathrm{yr}^{-1}, which agrees well with both the plotted J. Krtička & J. Kubát [56] and R. Björklund et al. [12] predictions. Our value is a factor of 50 times lower than the V01, and a factor of 2 lower than the B23 analytical formulae. In their review of Vela X-1, P. Kretschmar et al. [54] provided an overview of the different measurements of mass loss by stellar winds for the donor star of this system as well as the method used in these studies (see their Table 7). X-ray- and UV-based measurements yield a range of mass-loss rates, taking values from 2.7×10−7 M⊙ yr−12.7 \times10^{-7}\ M_{\odot}\ \mathrm{yr}^{-1} on the low end, up to <5.3×10−6 M⊙ yr−1< 5.3 \times10^{-6}\ M_{\odot}\ \mathrm{yr}^{-1} on the high end. Our measurement is consistent with these figures and lies on lower end of published values.

The measurement of M˙\dot{M} for BS007⋆ (log⁡L∗=3.4\log L_{*} = 3.4, log⁡M˙=−7.1\log\dot{M} = -7.1) stands out as conspicuously high for its low luminosity. There is an apparent bright source of 70 μm\mu\mathrm{m} emission at the location of the 24 μm\mu\mathrm{m} peak. However, it has the lowest 24/70 flux ratio of any bow shock measured in Table 1 in H. A. Kobulnicky et al. [52], indicating that there may be 70 μm\mu\mathrm{m} contamination along the same sight line. BS007⋆ lies within 1° of the Galactic plane and 3° from Galactic center. In such a crowded field, it is likely that there are overlapping foreground or background structures that contaminate the 70 μm\mu\mathrm{m} signal, inflating our measurement of M˙\dot{M}. This high 70 μm\mu\mathrm{m} intensity leads to an anomalously high ambient density (na=392 cm−3n_a = 392\ \mathrm{cm}^{-3}, the highest in our sample), suggesting that the nebula may not represent a stellar-wind bow shock, which casts doubt on our measurement of M˙\dot{M} using the bow shock method.

K16/J19 IDCommon IdentifierR.A. (2000) (HH:MM:SS)Decl. (2000) (°:′:″)TT (kK)σT\sigma_T (kK)log⁡g\log g (dex)σlog⁡g\sigma_{\log g} (dex)R⋆R_\star (R⊙R_\odot)σR⋆\sigma_{R_\star} (R⊙R_\odot)log⁡(L⋆/L⊙)\log(L_\star/L_\odot) (dex)σlog⁡(L⋆/L⊙)\sigma_{\log(L_\star/L_\odot)}M⋆M_\star (M⊙M_\odot)σM⋆\sigma_{M_\star} (M⊙M_\odot)binary
BS001⋆CD-29 1400417:48:07.68−29:07:55.634.12.03.830.2510.90.785.130.0853.126.0SB2
BS007⋆…17:58:30.65−26:09:49.322.31.54.440.183.570.23.440.107.80.84…
BS010⋆HD 31493718:00:17.47−25:14:14.627.81.53.810.168.030.514.440.0813.61.2…
BS013⋆ζ Oph16:37:09.55−10:34:01.232.10.53.70.068.460.474.820.0614.90.9…
BS016⋆…18:00:55.22−22:57:37.127.41.93.360.1816.83.25.190.2534.89.1…
BS017⋆…18:03:33.49−22:34:38.325.32.14.040.322.58.35.340.2487.027.0SB2
BS019⋆…18:05:11.18−21:04:43.324.52.03.480.2531.013.05.650.3078.024.0…
BS026⋆…18:11:59.40−19:36:55.428.51.03.90.1615.290.995.160.0772.024.0SB2
BS027⋆TYC 6272-1457-118:12:13.03−19:35:24.022.51.03.960.135.461.23.860.269.41.8…
BS043⋆TYC 6269-2270-118:16:25.25−17:05:20.826.91.13.740.138.250.774.530.1014.21.1…
BS051⋆NGC 6618 25818:20:22.70−16:08:34.134.01.04.410.2110.91.75.100.1395.030.0SB2
BS056⋆…18:17:17.18−15:29:25.426.82.03.710.2428.86.05.640.23146.099.0SB2
BS061⋆HD 16818318:18:58.70−13:59:28.330.60.63.830.1215.31.45.260.0864.917.0EB
BS063⋆BD-13 493418:19:05.57−13:54:50.431.11.14.210.217.080.424.590.0658.039.0SB2
BS065⋆BD-13 493718:19:20.04−13:54:21.621.20.73.720.096.220.313.820.069.10.7…
BS067⋆NGC 6611 58418:18:23.64−13:36:28.132.21.24.150.218.540.54.850.0759.615.0…
BS150⋆TYC 5118-279-118:49:25.06−02:21:09.725.11.13.970.138.420.424.440.0716.13.0R
BS200⋆…19:02:08.863:30:47.225.11.83.50.2532.89.45.570.3180.023.0…
BS255⋆TYC 1054-952-119:19:00.8213:42:59.020.00.83.710.16.170.293.720.078.50.64…
BS288⋆HD 35012319:32:44.3319:58:40.827.21.33.880.138.790.54.630.0818.53.7…
BS300⋆HD 34476519:44:20.0623:52:45.129.20.84.060.126.640.284.470.0517.02.7…
BS302⋆LS II +23 4719:45:47.5224:06:00.022.01.03.790.148.540.674.220.0912.91.8R
BS303⋆…19:46:06.1724:11:13.920.21.62.70.1868.422.05.900.3185.021.0SB1
BS305⋆HD33893619:46:22.6824:37:48.033.51.04.590.1510.21.55.080.1593.025.0SB1; SB2
BS306⋆BD+24 388319:47:21.2824:37:48.027.23.03.580.2312.41.54.720.1431.01.1SB2
BS314⋆HD 33896119:50:22.4127:27:25.526.51.43.690.1512.41.54.850.1533.01.0SB2
BS320⋆HD 19161120:09:26.0636:29:19.727.81.13.090.1323.92.45.480.1154.520.0…
BS321⋆…20:14:33.2936:29:49.930.91.84.240.749.360.744.900.0965.131.0SB1
BS322⋆…20:14:51.1036:35:18.232.71.03.980.1713.960.955.310.0774.030.0SB2
BS324⋆LS II +38 1920:13:28.9238:31:31.632.41.04.010.1517.31.35.490.08114.049.0R
BS325⋆HD 19430320:23:35.7136:55:44.021.60.83.290.115.350.824.670.0718.72.8
K16/J19 IDCommon IdentifierR.A. (2000) (HH:MM:SS)Decl. (2000) (◦:′:″)TT (kK)σT\sigma_T (kK)log⁡g\log gσlog⁡g\sigma_{\log g}R⋆R_\star (R⊙R_\odot)σR⋆\sigma_{R_\star} (R⊙R_\odot)log⁡(L⋆/L⊙)\log(L_\star/L_\odot) (dex)σlog⁡(L⋆/L⊙)\sigma_{\log(L_\star/L_\odot)} (dex)M⋆M_\star (M⊙M_\odot)σM⋆\sigma_{M_\star} (M⊙M_\odot)binary
(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)
BS330⋆HD 22915920:22:54.0239:12:28.822.41.02.940.1228.41.45.250.0730.95.5...
BS331⋆LS II +39 5320:27:17.5739:44:32.634.71.03.920.1316.90.434.840.0589.01.1SB1; SB2
BS332⋆...20:26:24.8940:01:41.232.20.83.950.138.640.434.840.0515.81.1SB1; SB2
BS335⋆TYC 3156-1106-120:28:15.3840:44:04.624.41.23.320.1416.990.774.920.0521.02.2...
BS339⋆...20:36:04.5140:56:13.245.64.54.280.3316.15.06.000.3689.024.0SB2
BS340⋆HD 19522920:28:30.2442:00:35.325.81.23.420.1211.240.564.680.0717.42.1...
BS341⋆...20:34:28.9441:56:16.833.10.84.320.1613.42.85.290.2191.026.0SB1; SB2
BS344⋆BD+43 365420:33:36.0743:59:07.446.80.93.7840.0521.690.876.310.04104.012.0...
BS350⋆...21:22:55.7050:24:25.224.21.63.160.1716.54.54.900.2322.16.3SB1; R
BS353⋆...22:16:32.9359:23:30.837.71.33.980.1214.790.995.610.0773.021.0SB2
BS356⋆HD 24001522:39:17.7159:01:00.526.21.03.250.113.750.694.860.0719.32.3SB1
BS357⋆HD 24001622:39:22.5459:00:25.924.21.03.240.118.12.04.990.1223.84.8SB1; R
BS358⋆TYC 4264-1036-122:34:06.5060:58:50.220.50.73.830.084.270.163.420.056.80.52...
BS361⋆TYC 4278-522-122:51:39.7261:08:51.023.21.82.690.1744.33.05.780.0748.66.8SB1
BS362⋆...22:55:45.460:24:40.325.12.13.340.2115.21.15.020.1159.413.0SB2
BS365⋆HD 261900:30:28.3265:16:19.626.81.23.540.139.290.454.580.0716.02.0...
BS367⋆TYC 4046-1614-102:17:53.2161:11:12.832.50.93.950.1511.60.545.140.0558.912.0...
BS368⋆V* KM Cas02:29:30.4661:29:44.232.31.03.910.179.260.54.930.0653.429.0SB2
BS369⋆BD+60 58602:54:10.6860:39:03.637.31.04.220.111.181.25.340.1068.021.0...
BS381⋆HD 5466207:09:20.26-10:20:47.839.91.24.030.113.91.35.650.0969.018.0...
BS384⋆CD-35 441508:16:32.09-35:38:52.835.41.54.260.2212.530.775.360.0781.053.0SB2
BS386⋆V* GP Vel09:02:06.86-40:33:16.924.51.12.80.136.41.85.590.2239.56.7EB
BS407⋆HD 9324910:44:43.87-59:21:25.232.80.53.50.0912.580.745.210.0625.83.8...
BS409⋆HD 9302710:43:17.90-60:08:03.533.81.03.950.110.41.25.100.1150.924.0...
BS414⋆V* V738 Car10:48:46.55-60:35:40.639.91.04.030.110.41.25.100.1150.924.0...
BS634⋆HD 15275616:57:15.12-42:43:41.227.30.74.180.1610.130.654.740.0571.023.0SB2; R
BS640⋆HD 3266516:58:24.05-42:44:20.529.61.54.280.110.20.654.740.0571.023.0EB
BS667⋆V* V1012 Sco17:15:22.32-38:12:46.827.10.74.050.110.051.14.850.0955.831.0...
BS673⋆...17:20:38.23-38:01:48.727.11.84.010.299.11.84.570.2114.41.5...
BS687⋆...17:28:59.78-36:06:54.736.82.14.190.3519.69.95.280.4878.059.0SB2
BS692⋆...17:27:11.23-34:14:43.236.81.14.160.1410.620.525.280.0598.021.0SB2
BS700⋆...17:33:41.88-31:16:32.127.01.93.820.2525.011.05.620.3584.025.0SB2
BS705⋆CD-29 1392517:45:08.21-29:56:45.227.11.53.450.1614.80.95.040.0925.55.0...
K16/J19 IDCommon IdentifierR.A. (2000) (HH:MM:SS)Decl. (2000) (°:′:″)TT (kK)σT\sigma_T (kK)log⁡g\log gσlog⁡g\sigma_{\log g}R⋆R_\star (R⊙R_\odot)σR⋆\sigma_{R_\star} (R⊙R_\odot)log⁡(L⋆/L⊙)\log(L_\star/L_\odot) (dex)σlog⁡(L⋆/L⊙)\sigma_{\log(L_\star/L_\odot)} (dex)M⋆M_\star (M⊙M_\odot)σM⋆\sigma_{M_\star} (M⊙M_\odot)binary
(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)
2G0063185-0066622⋆...18:02:29.23-23:49:34.024.01.63.040.1943.011.05.760.2868.021.0SB2
2G0065094-0061344⋆...18:02:41.88-23:38:02.027.91.73.710.187.350.494.450.1015.12.5SB2
2G0118310-0062990⋆HD 16696518:13:51.38-18:59:22.223.00.83.430.111.510.594.500.0715.62.1...
2G0592097+0012519⋆...19:41:43.0823:16:01.926.22.03.690.2429.510.05.620.36145.0110.0SB2; R
2G0600615-0018340⋆HD 34487819:44:44.0623:51:07.929.80.73.960.128.180.374.690.0520.34.1...
2G0770694+0193568⋆HD 19342720:18:45.1739:24:22.330.10.64.030.097.240.554.600.0719.23.8R
2G1045666+0128085⋆BD+57 251322:22:02.5758:43:09.534.90.54.640.1111.420.515.250.13111.013.0SB2
κ Casκ Cas00:33:00.0062:55:54.522.81.22.720.1150.19.25.780.2260.017.0...
2G3525610+0001887⋆CD-35 1156117:26:36.10-35:11:32.326.71.73.770.1910.580.714.750.1021.24.2SB2
2G3533026+0008095⋆HD 31988117:28:21.67-34:32:30.539.61.83.970.1413.860.735.640.0767.018.0SB2
2G3585210+0089852⋆...17:29:12.94-31:32:03.534.70.54.440.1311.140.755.210.0690.022.0SB2

Table 1. Spectral and stellar parameters of program stars

BS705⋆ (log⁡L∗=5.0\log L_{*} = 5.0, log⁡M˙=−6.0\log\dot{M} = -6.0) also appears as an outlier on Figure 9. The bow shock nebula shows no detectable signal above the background at 70 μm\mu\mathrm{m}, meaning the derived mass-loss rate should be regarded as an upper limit. The star lies within a degree of Galactic center and similar to BS007⋆, there is likely foreground and background contaminating emission at 70 μm\mu\mathrm{m}. We measure a specific intensity at 70 μm\mu\mathrm{m} IνI_{\nu} of 1.57×109 Jy sr−11.57 \times10^{9}\ \mathrm{Jy}\ \mathrm{sr}^{-1}, among the brightest in our sample. As a consequence, it has among the lowest F24/F70F_{24}/F_{70} flux ratios (H. A. Kobulnicky et al. [52]) and highest ambient densities of na=60.1 cm−3n_a = 60.1\ \mathrm{cm}^{-3}—however, still within the expected range of stellar bow shock nebulae (na<100 cm−3n_a < 100\ \mathrm{cm}^{-3}). We attribute background/foreground contamination as the source of the high M˙\dot{M} but find no reason to discount this measurement otherwise. As an upper limit, this measurement is consistent with the theoretical predictions.

The majority of objects in the sample (55 of 74) have space velocities less than 20 km s−120\ \mathrm{km\,s^{-1}} (open symbols in Figure 9; we will call these the “low-velocity stars”). These group towards the lower portion of the plot and lie an order of magnitude below the upper envelope of measurements. Compared to the models, the hollow symbols lie as much as 1–2 orders of magnitude below the theoretical predictions. Since ISM flows of unconstrained velocity are likely the mechanism for forming these bow shocks (as found in K22), we regard our calculated mass-loss rates—which rely upon space velocities computed from stellar proper-motion data—as lower limits. Low-velocity stars with 70 μm70\ \mu\mathrm{m} upper limits seem to follow the same trend as the other low-velocity stars. Evidently, the impact of having 70 μm70\ \mu\mathrm{m} non-detections is less important than the impact of having anomalously low VaV_a. This can be understood by reference to Equation (2). The former parameter enters only linearly while VaV_a takes a squared dependence, making it the largest source of uncertainty (indeed, systematic uncertainty!) in computing M˙\dot{M}.

J. van den Eijnden & L. Kaper (2026) found that mass-loss rates calculated from low-velocity stars using the bow shock technique are systematically underestimated when ISM bulk flows are ignored. To investigate the potential reasons for the offset observed for low-velocity stars, we plot in Figure 10 mass-loss rate versus the star’s 2D peculiar space velocity in the same manner as Figure 10 in J. van den Eijnden & L. Kaper (2026). The dashed vertical line denotes the threshold for runaway stars of 25 km s−125\ \mathrm{km\,s^{-1}} as used in K22, and the black line shows the best-fit trend for the low-velocity stars. The runaway stars on the right side of the panel show minimal-or-no correlation between mass-loss rate and V2DV_{2D} while the low-velocity stars show a positive correlation. This correlation between M˙\dot{M} and V2DV_{2D} on the left-hand side is unexpected and unphysical, unless perhaps lower-mass stars have lower space velocities. Said another way, this trend can be explained if low-mass OB stars are preferentially produced with low space-velocities. J. van den Eijnden & L. Kaper (2026) raise this as a possibility, and we investigate this question in the following figure.

Figure 10. Logarithm of mass-loss rate, in solar masses per year, versus peculiar 2D space velocity, in km s−1\mathrm{km\,s^{-1}}. The dashed vertical line shows the runaway velocity of 25 km s−125\ \mathrm{km\,s^{-1}} and the black line shows the best-fit linear relation for the low-velocity stars.

Figure 11 plots the peculiar 2D space velocity, in km s−1^{-1}, versus stellar mass, in solar masses. We find no correlation between space velocity and stellar mass. Runaway stars span the entire range of masses in the sample and have masses as low as 15 M⊙M_\odot. We find space velocity is independent of spectral type. Therefore, the underlying cause as to why low-velocity stars have low mass-loss rates cannot simply be attributed to the kinematic properties of these stars. V2DV_{2D} being independent of stellar mass strongly suggests that we must consider the effects of ISM bulk flows for the slowest stars in our sample, per the suggestion of J. van den Eijnden & L. Kaper (2026).

Figure 11. Peculiar 2D space velocity, in km s−1\mathrm{km\,s^{-1}}, versus stellar mass, in solar masses.

We investigate the correlation in M˙\dot{M} and V2DV_{2D} by adopting several lower-velocity limits. For each lower-velocity limit, each star’s velocity was corrected to a minimum value and M˙\dot{M} was recomputed for each case. We then computed the slope of the correlation between M˙\dot{M} and V2DV_{2D} to investigate how the slope of the trend line changes with varying V2DV_{2D} floors. Figure 12 plots the slope and uncertainty of the best-fit line between M˙\dot{M} and V2DV_{2D} (for low-velocity stars) versus the adopted minimum-velocity value. We find the correlation between M˙\dot{M} and V2DV_{2D} disappears when a minimum velocity of 20 km s−1^{-1} is adopted. This is in good agreement with the findings of J. van den Eijnden & L. Kaper (2026), who found bulk flows of 15 km s−1^{-1} must be responsible for producing bow shocks around low-velocity stars. Based on the findings of J. van den Eijnden & L. Kaper (2026) and the evidence of Figure 12, we are physically motivated to adopt a velocity floor of Va=20V_a = 20 km s−1^{-1} for stars with space velocities less than this value. This lower limit is also consistent with numerical simulations of bow shocks (F. Comeron & L. Kaper [22]; D. M. A. Meyer et al. [67]; D. M. Acreman et al. [6]), which find that relative velocities of 20 km s−1^{-1} is representative of the slowest bow shock-forming stars.

Figure 12. The slope between M˙\dot{M} and V2DV_{2D} for low-velocity stars (as in Figure 10) for different velocity floors. Error bars show the uncertainty in the computed slopes and the dashed-horizontal line shows zero slope.

Figure 13 plots the logarithm of the corrected mass-loss rate M˙20\dot{M}_{20}—adopting a minimum relative velocity of 20 km s−1^{-1}—versus the logarithm of stellar luminosity, in solar luminosity units, as in Figure 9. Hollow points again denote stars with low measured space velocities, but now these are assumed to have Va=20V_a = 20 km s−1^{-1}. These stars have shifted upward substantially relative to their positions in Figure 9 and have joined the trend defined by the high-confidence, high-velocity stars. As before, the points show a strong correlation between M˙\dot{M} and L⋆L_\star. The grey best-fit line shows excellent agreement with the predictions and is coincident to the R. Björklund et al. [12] models, at least for the lower-luminosity stars. This is remarkable because there are no free parameters in the mass-loss rate calculation. Even at the high-luminosity end, the trend line is only offset downward from the predictions by about 0.3 dex. The dispersion of the data from the mean relation is 0.58 dex, which is 0.27 dex smaller than the RMS in Figure 9. Lowvelocity stars with assumed Va=20 km s−1V_a = 20\ \mathrm{km\ s}^{-1} show a similar degree of scatter from the mean relation as the high-velocity stars (0.60 dex versus 0.56 dex). Low-velocity stars with no detectable 70 μm\mu\mathrm{m} signal (i.e., hollow points with downward-facing arrows) now scatter both above and below the models, as opposed to beneath the models as in Figure 9. This would seem to indicate that the systematic downward shift of low-velocity stars in Figure 9 has been properly corrected here, but the correction may be insufficient for some individual cases.

Figure 13. Mass-loss rates corrected to a minimum velocity of Va=20 km s−1V_a = 20\ \mathrm{km\ s}^{-1} versus stellar luminosity. Points are colored in the same way as in Figure 9. The grey track signifies best-fit trend line for measurement of M˙\dot{M} for stars above log⁡L∗=4.5\log L_* = 4.5. Hollow points have shifted upwards compared to Figure 9. With this correction, our measurements show good agreement with the R. Björklund et al. [12] models.

BS687⋆ now appears as an outlier below the theoretical predictions and the general trend in the other data near the center of the plot. The star has a peculiar 2D space velocity of 20 km s−1\mathrm{km\ s}^{-1}, which is just high enough to not be affected by our VaV_a correction. Since the measurement of M˙\dot{M} is anomalously low for its luminosity, and its VaV_a is just above the threshold for the correction, it is likely that this star also experiences an ISM bulk flow which causes the total relative velocity between the star and its surroundings VaV_a to exceed the correction value of 20 km s−1^{-1}. This is consistent with outflows from H II regions ([14]). This highlights the potential problems of applying a constant minimum velocity correction, as ISM bulk flows can exceed the correction we apply. Ideally, we would apply a correction for ISM bulk flows on a case-by-case basis, but we must apply a minimum average correction owing to the infeasibility of measuring ISM bulk flows for every star.

The bi-stability jump

We search for signatures of enhanced M˙\dot{M} due to bi-stability behavior near 25 kK by plotting in Figure 14 the difference between corrected mass-loss rates (M˙20\dot{M}_{20}) and predictions from the analytical formulae of V01 (which includes bi-stability enhancements, left panel) and of B23 (which does not include bi-stability enhancements, right panel) versus stellar effective temperature in kK. The gray shaded region in both panels shows where the predicted bi-stability behavior should appear in the 22.5–27.5 kK range. The horizontal dashed line marks a difference of zero.

Difference between corrected mass-loss rates and model predictions versus stellar effective temperature

Figure 14. The difference between corrected mass-loss rates M˙20\dot{M}_{20} and the V01 prediction (left panel) and the difference between corrected mass-loss rates M˙20\dot{M}_{20} and the B23 prediction (right panel) versus stellar effective temperature. Our corrected measurements of M˙\dot{M} show better agreement with the B23 models over the V01 prediction, suggesting that M˙\dot{M} measured from the bow shock technique does not show signs of bi-stability enhancements.

We find the B23 formula matches our corrected measurements of M˙\dot{M} better than the V01 prescription. The points in the left panel are shifted—on average—0.81 dex below the dashed line, meaning our measurements are systematically lower than what Equations 24 and 25 from V01 predict. In comparison, points in the right panel are offset 0.01 dex above the B23 prediction, which is small in comparison to the median uncertainty in M˙20\dot{M}_{20}. Many observational studies have suggested a reduction in the V01 relation by a factor of 2–3 ([70, 86, 15]), and our results are in agreement with this finding. The points in both panels have a dispersion of 0.91 dex about the mean, which is 4 times greater than the median statistical uncertainty in our corrected M˙\dot{M}. Points in the left-hand panel also show a systematic deviation from the V01 relation, with cooler stars lying farther below the prediction than hotter stars, while no such trend is evident in the right-hand panel. Our data show no effect of bi-stability enhancements, consistent with the findings of recent studies ([7] (A. de Burgos et al. 2024a)). The four outliers at the upper left in the right-hand panel are upper limits and therefore, do not constitute evidence of enhanced mass loss from bi-stability behavior.

Interestingly, K19 claimed evidence of bi-stability in the previous application of the bow shock technique in their study of 70 bow shock stars, with stars on the cool side of the bi-stability gap having mass-loss rates 1–4 dex higher than the V01 prescription. This result was surprising, as V01 already includes bi-stability effects, and these findings suggested that bi-stability enhancements may have been more significant than predicted since the bow shock technique was able to probe less-luminous stars which would otherwise be inaccessible by other techniques. Focusing on the coolest stars in our sample, our findings run counter to those of K19.

To investigate this discrepancy, we turn our attention to stars cooler than 25 kK that are in common in both this work and the sample of K19, which yields 4 stars—BS007⋆, BS357⋆, BS361⋆, and BS362⋆. We use nearly identical values for all parameters in Equation (2), except for IνI_\nu. For all stars, our values of IνI_\nu are at least a factor of four lower, and as much as a factor of six lower. This can largely be attributed to the erroneous pixel size used in K19. We attribute the rest of this difference to our more rigorous methodology by which we fit a profile to the 70 μm\mu\mathrm{m} data and subtract an estimate of the background. Both differences inflated M˙\dot{M} for these cool stars in K19, and likely the other stars cooler than 25 kK in that work, giving a false impression of bi-stability enhancements. In some cases, we determine objects are not detected at 70 μ\mum, in disagreement with K19. Our corrected measurements of M˙\dot{M} agree well with the analytical formulae of B23, and we take this as evidence of the lack of bi-stability enhancements.

Enhancement in M˙\dot{M} for rapid rotators?

Figure 15 plots the difference between the corrected mass-loss rates of bow shock stars, M˙20\dot{M}_{20}, and the B23 prediction (which was computed with a rotation rate 40% the critical rotational velocity) versus the projected rotational velocity vsin⁡iv \sin i. Points are colored the same as in Figures 9 and 13. We exclude bow shock stars that we classify as double-lined spectroscopic binaries in P25 as well as bow shock upper-limit cases, which leaves 23 stars. The points show a uniform dispersion about the horizontal dashed line, with a median difference from the B23 prescription of -0.12 dex—which is smaller than the

Difference between corrected mass-loss rates and the B23 prediction versus projected rotational velocity

Figure 15. The difference between the corrected mass-loss rates M˙20\dot{M}_{20} and the B23 prediction versus projected rotational velocity. There is no correlation, suggesting that any effects from rotation on mass loss are small compared to the uncertainties.

median uncertainty in M˙20\dot{M}_{20}—and a standard deviation of 0.25 dex. ζ\zeta Oph, the fastest rotator in our sample at vsin⁡i=360v \sin i = 360 km s−1^{-1}, is an illustrative case and shows excellent agreement with the B23 prediction. ζ Oph—which is near its critical rotational velocity (see Figure 13 of P25)—should show the most dramatic increase in mass-loss rate if rotation is a significant factor in determining mass loss. From the limited number of rapidly rotating stars available, we conclude that any effects of rotational velocity on M˙\dot{M} are minimal compared to the random uncertainties in measuring M˙\dot{M} using the bow shock technique.

Discussion & Conclusions

We have shown that the bow shock method is capable of measuring mass-loss rates for stars with luminosities 4.5≲log⁡(L∗/L⊙)≲6.34.5 \lesssim\log(L_\ast/L_\odot) \lesssim6.3. This includes stars less luminous than the weak-winded star ζ Ophiuchi. Our sample contains both dwarf and evolved stars. At log⁡(L∗/L⊙)≲4.5\log(L_\ast/L_\odot) \lesssim4.5, there are very few objects that can be studied, but even in this regime the bow shock technique is promising, since late-O and early B stars are largely inaccessible by other techniques of measuring mass loss. Many of these low-luminosity stars have M˙\dot{M} upper limits, and a limiting factor for this method is the detectability of the bow shock above the 70 μ\mum background. Overlapping foreground and background emission at 70 μ\mum dilutes the signal from the bow shock nebula and inflates mass-loss rates above predictions in some cases. However, taken as a whole, these upper-limit cases are close to the overall trend. This may signify that some upper limits are nearly detections, but the signals are still not strong enough to distinguish over the background.

Our measurements of M˙\dot{M} lie below the analytical mass-loss prescriptions of V01 by 0.83 dex, but they show excellent agreement with the predictions of B23—with a median offset of +0.01 dex. We do not find evidence of bistability effects, which runs counter to the previous bow shock work of K19, but is in line with other recent observational studies (A. de Burgos et al. [23]; T. Alkousa et al. [7]). Overall, we find a reduction in mass-loss rates from the previous bow shock studies of K18/K19 due to the use of the correct pixel size and more-rigorous background subtraction when measuring IνI_\nu.

Previous works have shown that most stars that form bow shock nebulae are not runaways and have suggested that interstellar flows play a role in forming the bow shock. Mass-loss rates computed by taking the peculiar-space velocities as the star-ISM relative velocity show an unexpected and unphysical systematic correlation between M˙\dot{M} and VaV_a for slow-moving stars. This motivated the introduction of a floor on VaV_a of 20 km s−1^{-1}—representative of the true star-ISM velocity—which is corroborated by the work of J. van den Eijnden & L. Kaper (2026). This is a large but physically motivated correction (e.g., Figure 11). When this correction is applied, we find it produces a relation with a smaller RMS (0.58 dex versus 0.85 dex) about the trend and a relation that is in better agreement with models. This correction to VaV_a is necessary for the use of this method of obtaining M˙\dot{M}. The uncertainty in VaV_a is the dominant source of error in M˙\dot{M} but nevertheless, this method is successful in producing a tight relationship between mass-loss rate M˙\dot{M} and luminosity L∗L_\ast, with the same slope and offset as models. With the recognition of the need for the lower limit on VaV_a, the bow shock method has become a successful technique for measuring M˙\dot{M} with striking agreement to modern models. The dispersion in this relation is small compared to radio-recombination, Hα, and UV-based studies. For example, the dispersions in M˙\dot{M} as found in I. D. Howarth & R. K. Prinja [45] were about a full dex. Recent compilations yield measurements of M˙\dot{M} with dispersions of about 0.5 dex (A. de Burgos et al. [23]), similar to this work.

The results for objects within H II regions, facing H II regions, and facing bright-rimmed clouds, are consistent with measurements from isolated nebulae. We conclude that this technique is robust for bow shocks in any of these environments. This makes the bow shock technique applicable to a wide range of scenarios in which stellar bow shock nebulae can form and does not require knowledge about the immediate environment. This work establishes the bow shock technique as a promising and robust method with no adjustable parameters—with the exception of the physically motivated lower limit on VaV_a. The primary limitation on broad use of this method is the limited number of 70 μm\mu\mathrm{m} detections.

Acknowledgments

This work has been supported by the National Science Foundation through grants AST-2108348, AST-1852289, and AST-2108349. We thank Stan Owocki, whose contribution to the field of stellar outflows is a great inspiration to us (Sta 2026). We also thank Daniel Maschmann for his coding expertise. We thank Dr. Matthew Povich for helpful comments on early versions of this paper.

This work is based, in part, on observations made with the Spitzer Space Telescope, which was operated by the Jet Propulsion Laboratory, California Institute of Technology under a contract with NASA. Support for this work was provided by NASA through an award issued by JPL/Caltech.

Facilities: Gaia, Herschel (PACS), Spitzer (IRAC, MIPS), WISE

Software: astropy ([8, 9, 10])

References

  1. [2]Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2016, PhRvL, 116, 061102, doi: 10.1103/PhysRevLett.116.061102DOI
  2. [3]Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2017a, PhRvL, 119, 161101, doi: 10.1103/PhysRevLett.119.161101DOI
  3. [5]Abbott, D. C. 1978, ApJ, 225, 893, doi: 10.1086/156554DOI
  4. [6]Acreman, D. M., Stevens, I. R., & Harries, T. J. 2016, MNRAS, 456, 136, doi: 10.1093/mnras/stv2632DOI
  5. [7]Alkousa, T., Crowther, P. A., Bestenlehner, J. M., et al. 2025, A&A, 699, A314, doi: 10.1051/0004-6361/2025253799
  6. [8]Astropy Collaboration, Robitaille, T. P., Tollerud, E. J., et al. 2013, A&A, 558, A33, doi: 10.1051/0004-6361/201322068
  7. [9]Astropy Collaboration, Price-Whelan, A. M., Sipőcz, B. M., et al. 2018, AJ, 156, 123, doi: 10.3847/1538-3881/aabc4f
  8. [10]Astropy Collaboration, Price-Whelan, A. M., Lim, P. L., et al. 2022, ApJ, 935, 167, doi: 10.3847/1538-4357/ac7c74DOI
  9. [11]Benaglia, P., Vink, J. S., Martí, J., et al. 2007, A&A, 467, 1265, doi: 10.1051/0004-6361:20077139DOI
  10. [12]Björklund, R., Sundqvist, J. O., Puls, J., & Najarro, F. 2021, A&A, 648, A36, doi: 10.1051/0004-6361/202038384DOI
  11. [13]Björklund, R., Sundqvist, J. O., Singh, S. M., Puls, J., & Najarro, F. 2023, A&A, 676, A109, doi: 10.1051/0004-6361/202141948DOI
  12. [14]Bodenheimer, P., Tenorio-Tagle, G., & Yorke, H. W. 1979, ApJ, 233, 85, doi: 10.1086/157368DOI
  13. [15]Bouret, J.-C., Hillier, D. J., Lanz, T., & Fullerton, A. W. 2012, A&A, 544, A67, doi: 10.1051/0004-6361/201118594DOI
  14. [16]Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000, doi: 10.1046/j.1365-8711.2003.06897.xDOI
  15. [17]Capak; Peter. 2019, IPAC, doi: 10.26131/IRSA3 Carretero-Castrillo, M., Ribó, M., & Paredes, J. M. 2023, A&A, 679, A109, doi: 10.1051/0004-6361/202346613DOI
  16. [19]Chick, W. T., Kobulnicky, H. A., Schurhammer, D. P., et al. 2020, ApJS, 251, 29, doi: 10.3847/1538-4365/abc0e5DOI
  17. [20]Churchwell, E., Povich, M. S., Allen, D., et al. 2006, ApJ, 649, 759, doi: 10.1086/507015DOI
  18. [21]Cohen, D. H., Wollman, E. E., Leutenegger, M. A., et al. 2014, MNRAS, 439, 908, doi: 10.1093/mnras/stu008DOI
  19. [22]Comeron, F., & Kaper, L. 1998, A&A, 338, 273
  20. [23]de Burgos, A., Keszthelyi, Z., Simón-Díaz, S., & Urbaneja, M. A. 2024a, A&A, 687, L16, doi: 10.1051/0004-6361/202450301DOI
  21. [24]de Burgos, A., Simón-Díaz, S., Urbaneja, M. A., & Puls, J. 2024b, A&A, 687, A228, doi: 10.1051/0004-6361/202348808DOI
  22. [25]Dessart, L., & Owocki, S. P. 2005, A&A, 432, 281, doi: 10.1051/0004-6361:20041461DOI
  23. [26]Draine, B. T., & Li, A. 2007, ApJ, 657, 810, doi: 10.1086/511055DOI
  24. [27]Draine, B. T., Li, A., Hensley, B. S., et al. 2021, ApJ, 917, 3, doi: 10.3847/1538-4357/abff51DOI
  25. [28]Eldridge, J. J., Stanway, E. R., Xiao, L., et al. 2017, PASA, 34, e058, doi: 10.1017/pasa.2017.51DOI
  26. [29]Feldmeier, A., Puls, J., & Pauldrach, A. W. A. 1997, A&A, 322, 878
  27. [30]Fryer, C. L., Holz, D. E., & Hughes, S. A. 2002, ApJ, 565, 430, doi: 10.1086/324034DOI
  28. [31]Fullerton, A. W., Massa, D. L., & Prinja, R. K. 2006, ApJ, 637, 1025, doi: 10.1086/498560DOI
  29. [32]Gaia Collaboration, Vallenari, A., Brown, A. G. A., et al. 2023, A&A, 674, A1, doi: 10.1051/0004-6361/202243940
  30. [33]Garmany, C. D., Olson, G. L., van Steenberg, M. E., & Conti, P. S. 1981, ApJ, 250, 660, doi: 10.1086/159413DOI
  31. [34]Gull, T. R., & Sofia, S. 1979, ApJ, 230, 782, doi: 10.1086/157137DOI
  32. [35]Gutermuth & Heyer (2015). 2020, IPAC, doi: 10.26131/IRSA258DOI
  33. [36]Gvaramadze, V. V., Kroupa, P., & Pfam-Altenburg, J. 2010, A&A, 519, A33, doi: 10.1051/0004-6361/201014871DOI
  34. [37]Hanes, R. J., McSwain, M. V., & Povlich, M. S. 2018, AJ, 155, 190, doi: 10.3847/1538-3881/aab70dDOI
  35. [38]Hawcroft, C., Sana, H., Mahy, L., et al. 2024, A&A, 688, A105, doi: 10.1051/0004-6361/202245588DOI
  36. [39]Henney, W. J., & Arthur, S. J. 2019a, MNRAS, 486, 4423, doi: 10.1093/mnras/stz1130DOI
  37. [40]Henney, W. J., & Arthur, S. J. 2019b, MNRAS, 486, 3423, doi: 10.1093/mnras/stz1043DOI
  38. [41]Henney, W. J., & Arthur, S. J. 2019c, MNRAS, 489, 2142, doi: 10.1093/mnras/stz2283DOI
  39. [42]Holgado, G., Simón-Díaz, S., Herrero, A., & Barbá, R. H. 2022, A&A, 665, A150, doi: 10.1051/0004-6361/202243851DOI
  40. [43]Holgado, G., Simón-Díaz, S., Barbá, R. H., et al. 2018, A&A, 613, A65, doi: 10.1051/0004-6361/201731543DOI
  41. [44]Holgado, G., Simón-Díaz, S., Haemmerlé, L., et al. 2020, A&A, 638, A157, doi: 10.1051/0004-6361/202037699DOI
  42. [45]Howarth, I. D., & Prinja, R. K. 1989, ApJS, 69, 527, doi: 10.1086/191321DOI
  43. [46]Jayasinghe, T., Dixon, D., Povich, M. S., et al. 2019, MNRAS, 488, 1141, doi: 10.1093/mnras/stz1738DOI
  44. [47]Kalogera, V., Kim, C., Lorimer, D. R., et al. 2004, ApJL, 614, L137, doi: 10.1086/425868DOI
  45. [48]Kobulnicky, H. A., & Chick, W. T. 2022, AJ, 164, 86, doi: 10.3847/1538-3881/ac7f2bDOI
  46. [49]Kobulnicky, H. A., Chick, W. T., & Povich, M. S. 2018, ApJ, 856, 74, doi: 10.3847/1538-4357/aab3e0DOI
  47. [50]Kobulnicky, H. A., Chick, W. T., & Povich, M. S. 2019, AJ, 158, 73, doi: 10.3847/1538-3881/ab2716DOI
  48. [51]Kobulnicky, H. A., Gilbert, I. J., & Kiminki, D. C. 2010, ApJ, 710, 549, doi: 10.1088/0004-637X/710/1/549DOI
  49. [52]Kobulnicky, H. A., Schurhammer, D. P., Baldwin, D. J., et al. 2017, AJ, 154, 201, doi: 10.3847/1538-3881/aa90baDOI
  50. [53]Kobulnicky, H. A., Chick, W. T., Schurhammer, D. P., et al. 2016, ApJS, 227, 18, doi: 10.3847/0067-0049/227/2/18DOI
  51. [54]Kretschmar, P., El Mellah, I., Martínez-Núñez, S., et al. 2021, A&A, 652, A95, doi: 10.1051/0004-6361/202040272DOI
  52. [55]Krtička, J. 2014, A&A, 564, A70, doi: 10.1051/0004-6361/201321980DOI
  53. [56]Krtička, J., & Kubát, J. 2017, A&A, 606, A31, doi: 10.1051/0004-6361/201730723DOI
  54. [58]Lamers, H. J. G. L. M., & Leitherer, C. 1993, ApJ, 412, 771, doi: 10.1086/172960DOI
  55. [59]Landau, L. D., & Lifshitz, E. M. 1987, Fluid Mechanics
  56. [60]Leitherer, C. 1988, ApJ, 326, 356, doi: 10.1086/166097DOI
  57. [61]Lindegren, L., Bastian, U., Biermann, M., et al. 2021, A&A, 649, A4, doi: 10.1051/0004-6361/202039653DOI
  58. [62]Maíz Apellániz, J., Sota, A., Morrell, N. I., et al. 2013, in Massive Stars: From alpha to Omega, 198, doi: 10.48550/arXiv.1306.6417
  59. [63]Marcolino, W. L. P., Bouret, J. C., Martins, F., et al. 2009, A&A, 498, 837, doi: 10.1051/0004-6361/200811289DOI
  60. [64]Markova, N., Puls, J., Repolust, T., & Markov, H. 2004, A&A, 413, 693, doi: 10.1051/0004-6361:20031463DOI
  61. [65]Martins, F., Schaefer, D., Hillier, D. J., et al. 2005, A&A, 441, 735, doi: 10.1051/0004-6361:20052927DOI
  62. [66]Mathis, J. S., Mezger, P. G., & Panagia, N. 1983, A&A, 128, 212
  63. [67]Meyer, D. M. A., van Marle, A. J., Kuiper, R., & Kley, W. 2016, MNRAS, 459, 1146, doi: 10.1093/mnras/stw651DOI
  64. [68]Mokiem, M. R., de Koter, A., Vink, J. S., et al. 2007, A&A, 473, 603, doi: 10.1051/0004-6361:20077545DOI
  65. [69]Molinari, S., Swinyard, B., Bally, J., et al. 2010, PASP, 122, 314, doi: 10.1086/651314DOI
  66. [70]Najarro, F., Hanson, M. M., & Puls, J. 2011, A&A, 535, A32, doi: 10.1051/0004-6361/201016003DOI
  67. [71]Noriega-Crespo, A., van Buren, D., & Dgani, R. 1997, AJ, 113, 780, doi: 10.1086/118298
  68. [72]Oskinova, L. M., Hamann, W. R., & Feldmeier, A. 2007, A&A, 476, 1331, doi: 10.1051/0004-6361:20066377DOI
  69. [73]Owocki, S. P. 2015, in Astrophysics and Space Science Library, Vol. 412, Very Massive Stars in the Local Universe, ed. J. S. Vink, 113, doi: 10.1007/978-3-319-09596-7_5
  70. [74]Patten, N., Kobulnicky, H. A., Povich, M. S., et al. 2025, ApJ, 988, 183, doi: 10.3847/1538-4357/ade22fDOI
  71. [75]Peri, C. S., Benaglia, P., Brookes, D. P., Stevens, I. R., & Isequilla, N. L. 2012, A&A, 538, A108, doi: 10.1051/0004-6361/201118116DOI
  72. [76]Peri, C. S., Benaglia, P., & Isequilla, N. L. 2015, A&A, 578, A45, doi: 10.1051/0004-6361/201424676DOI
  73. [77]Povich, M. S., Benjamin, R. A., Whitney, B. A., et al. 2008, ApJ, 689, 242, doi: 10.1086/592565DOI
  74. [78]Pretorius, F. 2005, PhRvL, 95, 121101, doi: 10.1103/PhysRevLett.95.121101DOI
  75. [79]Puls, J., Vink, J. S., & Najarro, F. 2008, A&A Rv, 16, 209, doi: 10.1007/s00159-008-0015-8DOI
  76. [80]Puls, J., Kudritzki, R. P., Herrero, A., et al. 1996, A&A, 305, 171
  77. [81]Rubio-Díez, M. M., Sundqvist, J. O., Najarro, F., et al. 2022, A&A, 658, A61, doi: 10.1051/0004-6361/202040116
  78. [82]Runacres, M. C., & Owocki, S. P. 2002, A&A, 381, 1015, doi: 10.1051/0004-6361:20011526DOI
  79. [83]Smith, N. 2014, ARA&A, 52, 487, doi: 10.1146/annurev-astro-081913-040025DOI
  80. [84]Sota, A., Maiz Apellaniz, J., Walborn, N. R., et al. 2014,, VizieR On-line Data Catalog: III/274. Originally published in: 2011ApJS...193...24S; 2014ApJS...211...10S
  81. [85]Stanway, E. R., & Eldridge, J. J. 2018, MNRAS, 479, 75, doi: 10.1093/mnras/sty1533
  82. [86]Sundqvist, J. O., Puls, J., Feldmeier, A., & Owocki, S. P. 2011, A&A, 528, A64, doi: 10.1051/0004-6361/201015771DOI
  83. [89]Vink, J. S. 2022, ARA&A, 60, 203, doi: 10.1146/annurev-astro-052920-094949DOI
  84. [90]Vink, J. S., de Koter, A., & Lamers, H. J. G. L. M. 2001, A&A, 369, 574, doi: 10.1051/0004-6361:20010127DOI
  85. [91]Vink, J. S., Mehner, A., Crowther, P. A., et al. 2023, A&A, 675, A154, doi: 10.1051/0004-6361/202245650DOI
  86. [92]Šurlan, B., Hamann, W. R., Aret, A., et al. 2013, A&A, 559, A130, doi: 10.1051/0004-6361/201322390DOI
  87. [93]Whisnant, A. S., Povich, M. S., Patten, N., & Kobulnicky, H. A. 2025, AJ, 169, 337, doi: 10.3847/1538-3881/addc72
  88. [94]WISE Team. 2020, IPAC, doi: 10.2631/IRSA138

Paper details

vlnc:W_9nm6fgrc
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents