© ESO, 2012

1. Introduction

Many astrophysical objects perturb the interstellar medium (ISM) and produce different kinds of observable structures. Early-type stars with large peculiar velocities (termed runaway stars) are one example of the perturbing agent. They are responsible for the generation of stellar bow shocks.

Runaway stars have been studied for several decades. There are currently two proposed mechanisms for the origin of their high velocities (Hoogerwerf et al. 2000). One is the binary supernova scenario (BSS, Zwicky 1957; Blauuw 1961) and the other is the dynamical ejection scenario (DES, Poveda et al. 1967; Gies & Bolton 1986). In the BSS, the runaway star that was originally part of a binary system, acquires a high speed when its companion explodes as a supernova. In the context of the DES, the runaway star achieves its velocity thanks to the dynamical interaction with one or more stars. In some cases, the trajectory of the stars and the original systems can be reconstructed, and the mechanism that has kicked the star can be identified. The gathered evidence has not always enabled strong conclusions to be made (see for instance the dissenting findings by Comerón & Pasquali 2007; Gvaramadze & Bomans 2008). Studies such as Moffat et al. (1998) of runaway stars suggest that the BSS is slightly more likely, and that those stars with significant peculiar supersonic motion relative to the ambient ISM, tend to form bow shocks in the direction of the motion.

Abundant studies of individual or smaller groups of runaway stars can be found. However, compilations of large number of high velocity stars are scarce. A good example is the Galactic O-Star Catalog (GOSC, Maíz Apellániz et al. 2004), which contains the physical properties of about 40 of these stars. Tetzlaff et al. (2010) carried out an extensive kinematical and probability study, that allowed us to identify thousands of runaway stars. The authors used the Hipparcos catalogue (Perryman et al. 1997) and built a database of  ~2500 objects.

The effects of runaway stars on the environment and the interaction between those stars and the ISM have been widely documented. A fundamental contribution can be found in Van Buren & McCray (1988), who presented a list of 15 bow shock-like candidates, found while studying Galactic HII regions detected in IRAS (InfraRed Astronomical Satellite) data. Their study triggered an extended search for stellar bow-shock candidates, which resulted in almost 60 similar sources of which around 20 were actually candidates (Van Buren et al. 1995; Noriega-Crespo et al. 1997: NC97). Some of the candidates were studied by Brown & Bomans (2005), using the Hα all-sky survey (SHASSA/VTSS). They searched images of 37 objects of the list of Van Buren et al. (1995) and detected 8 bow shocks. They also calculated environmental parameters and in all the cases they found consistency with the features of the warm ionized medium.

In the past few years, searches of relatively small sky regions have revealed more stellar bow shocks. Povich et al. (2008) reported the discovery of six stellar bow shocks in the star-forming regions M 17 and RCW 49 from Spitzer GLIMPSE (Galactic Legacy Infrared Mid-Plane Survey Extraordinaire) images. By combining 2MASS (Two Micron All-Sky Survey), Spitzer, MSX (Midcourse Space eXperiment), and IRAS data, they obtained the SEDs of the bow shocks and stars associated with them. Other 10 bow-shock candidates were found by Kobulnicky et al. (2010) using mid-IR images from the Spitzer Space Telescope Cygnus X Legacy Survey. Arnal et al. (2011) exposed a study of the surrounding emission related to the star HD 192281, a member of the OB association Cygnus OB8. They analyzed neutral hydrogen, radio continuum, molecular gas (CO) emission and infrared (IR) data, and derived several parameters of the medium. They concluded that HD 192281 seems to generate a stellar bow shock and it is possible that triggered star formation is underway. Bow shocks have been seen around some massive X-ray binaries, such as Vela X-1 (Kaper et al. 1997) and 4U 1907+09 (Gvaramadze et al. 2001a).

Gvaramadze et al. (2011b) carried out a search for OB stars running away from the young stellar cluster NGC 6357. They discovered seven bow shocks, and thoroughly discussed the scenario for these structures to develop.

During the investigation of the outskirts of Cygnus OB2, Comerón & Pasquali (2007) discovered a bow-shaped structure close to the high mass runaway star BD +43° 3654. From spectroscopic data, they re-classified the star and derived its stellar mass, ranking the star as one of the three most massive runaways known in the Milky Way. Stellar proper motions as well as MSX observations confirmed that the bow shock could be produced by the star, in turn probably a former member of the mentioned association. In a subsequent publication, Benaglia et al. (2010) carried out a study of the bow shock produced by BD +43° 3654, and evaluated whether the structure could give rise to high energy emission. By means of dedicated VLA observations, they calculated the spectral index in the region of the bow shock and measured values compatible with a non-thermal radiation origin. Assuming that there is a population of relativistic particles, they estimated the SEDs generated by different radiative processes.

In spite of these studies, numerous questions remain to be answered. How common are stellar bow shocks? In which conditions are they produced and detected? Questions such as how the spectral type, velocity, stellar wind, coordinates of the star, and other parameters influence the formation of a bow shock will be more clearly addressed as more examples are analyzed.

In this paper, we introduce the Extensive BOw Shock Survey (E-BOSS) generated by means of on-line available data from different missions. We used the first-release Wide-field Infrared Survey Explorer (WISE) images, as a tool to detect new structures and to contribute to what is known about those already studied.

In the next section, we briefly discuss the theoretical background and analyze the conditions that could increase the probability of detecting a stellar bow shock. Section 3 describes the procedure followed to obtain the sample in which we search for bow-shock candidates. In Sect. 4, we present the bow shocks found and characterize them. Section 5 shows the statistics performed, and Sect. 6 has a corresponding discussion. In the last section, we comment on the immediate prospects.

2. Theoretical considerations

Runaway stars move through the ISM with velocities that overcome the field stars velocities and the sound of speed in the ISM. They shock the ISM and produce the so-called bow shocks that sweep up the ISM matter into thin, dense shells (Wilkin 1996). The theoretical study of bow shocks has been developed not only from an analytical point of view but also with numerical simulations.

Wilkin (1996) derived analytical exact solutions for stellar wind bow shocks in the thin-shell limit, stressing the importance of the conserved momentum within the shell. He developed a simple method to reproduce the shape of the shell, mass column, and velocity of the shocked gas throughout the shell. Later, Wilkin (2000) improved the model studying the modifications of the bow shocks in two cases where: (i) a star moves supersonically with respect to an ambient medium with a density gradient perpendicular to the stellar velocity, and (ii) a star with a mis-aligned, axisymmetric wind moves in a uniform medium. He found that the region of the stand-off point (where ambient and wind pressures balance each other) is tilted in both cases. In that way, the star does not lie in the line that divides the bow shock into two halves.

Dgani et al. (1996a,b) peformed a stability analysis of thin isothermal bow shocks. In the first paper, they showed that the bow shocks produced by stars with fast stellar winds are more stable than those generated by slow winds. In the second article, the authors then investigated non-linear instabilities and run numerical simulations to solve the problem. They proposed that the slower the wind, the highest the instability. They also applied the model to the star α Cam and concluded that the clumps observed might be explained by the instabilities.

Comerón & Kaper (1998) conducted a semi-analytical study of bow shocks produced by OB runaway stars and derived expressions that suggest different results from those obtained with the assumption of instantaneous cooling of the shocked gas (Wilkin 1996). They ran numerical simulations that reveal a wealth of details in the formation, structure, and evolution of the bow shocks. These features strongly depend on the conditions of the medium and star. The bow shocks can either form or not; and if they do form, they can be either stable, unstable, or layered.

To complete the study of bow shocks and improve the models, they need to be observed and analyzed. In principle, owing to the presence of the shocks that the runaway stars produce in the ISM, the dust is heated and in that way re-radiates at infrared wavelengths. As we describe in other sections, bow shocks have been observed at infrared (e.g. van Buren et al. 1995, NC97), optical (e.g. Brown & Bomans 2005), and – in a few cases – radio wavelengths, and will eventually be detected at high energy waves (Benaglia et al. 2010).

3. The making of the E-BOSS sample

We searched for stellar bow shocks in various different databases described later. We account for a number of criteria that help us to identify new stellar bow shocks.

• 1.

  We examined the surroundings of early-type runaway stars, asthese have high velocities and strong winds that sweep upinterstellar matter and also high luminosities that contribute tothe dust heating.

• 2.

  We selected nearby stars (d < 3 kpc), which thus have brighter bow shocks.

We generated the initial sample in two different ways. We considered first bow shock candidates that had previously been detected in the IRAS data (NC97). Secondly, we carried out a systematic search around runaway stars that could produce bow shocks, using the catalogue of Tetzlaff et al. (2010). The next section describes the process in more detail. In the rest of the paper, we divide the E-BOSS sample into two groups, groups 1 and 2.

Fig. 1 WISE images of the bow shock candidates around 6 OB stars. Color mapping: blue = 3.4 microns; green = 12.1 microns; red = 22.2 microns. The color scales are in data numbers (DNs); the WISE data is not calibrated in their surface brightness. The vectors indicate the direction of the star proper motion: the thicker one represents the one derived from Hipparcos data by van Leeuwen (2007); the thinner one is the same but corrected for the ISM motion caused by Galactic rotation. The vectors length are not scaled with the original values.

Fig. 2 Same as in Fig. 1, for another six OB stars.

Fig. 3 Same as in Figs. 1 and 2, for another six OB stars.

Fig. 4 Same as in Figs. 1–3, for another six OB stars.

Fig. 5 WISE images of bow shock candidates as in Figs. 1–4, for HIP 97796 and 114990. MSX images for BD +43° 3654 and HIP 101186. Color mapping for MSX: blue = 8.3 microns; green = 12.1 microns; red = 21.3 microns.

3.1. Group 1

For this group, we took the bow shock candidates database of NC97. These authors searched for bow shocks and other features at IR wavelengths from the HiReS-IRAS maps. Table 1 lists the 56 objects that are OB stars from NC97, hereafter called group 1 (the WR stars HD 50896 and HD 192163, from the original list, were excluded).

3.2. Group 2

The second group was extracted from the catalogue of Tetzlaff et al. (2010). From a sample of 7663 young stars observed by Hipparcos (Perryman et al. 1997), the authors built a catalogue of 2546 runaway stars candidates. Their study lists stellar names, velocities, spectral types, ages, and masses. From this last list, we found 244 stars of spectral type O to B2. We list this 244 stars in Table 2; they comprise what we call hereafter group 2. A total of 17 stars are common to both groups.

3.3. Information collation

We used the NASA/IPAC Infrared Science Archive¹ to gather relevant infrared and sub-millimeter missions data (IRAS, MSX, WISE, Spitzer, etc.). In practice, data from MSX and WISE² have proven to be the most useful. In particular, WISE has discovered various bow shocks. Although the published WISE data covers half of the sky, in effect, it covers more than two thirds of both group 1 and group 2 samples. When further data becomes available, it will be added to a subsequent release of the E-BOSS sample.

In addition to the infrared data, we searched for bow shock emission at Hα, using the Virginia Tech. Spectral Survey (VTSS, Dennison et al. 1997) and the Southern Hemispheric Hα Sky Survey Atlas (SHASSA, Gaustad et al. 2001). These surveys cover most of the sky with a spatial resolution of 0.8′ for SHASSA and 1.6′ for VTSS. We did not find any convincing bow shock candidates for either our group 1 or group 2 sample. This is in contrast to Brown & Bomans (2005), who found 8 possible bow shock candidates, starting with a target list of 37 stars from van Buren et al. (1995). We note that the stars are often located in complex Hα emission regions making identification of a bow shock feature difficult. In some cases, we did detect possible Hα emission from a bow shock but which is not coincident with a clear bow shock detected by WISE (for example, HD 48099 and HD 149757). In these cases, we relied on the infrared detection.

The detection of radio emission and the measurement of a non-thermal radio spectral index coincident with the location of the bow-shock candidate related to the star BD  + 43° 3654 encouraged us to check for low frequency radio emission around the E-BOSS members. We used the postage stamp server for NRAO/VLA Sky Survey (NVSS, Condon et al. 1998), which returns radio images of the sky in various formats. Most of the radio maps contain only point sources, which are not particularly related to the IR features. However, three E-BOSS objects, apart from BD +43° 3654, correspond interesting comma-shaped radio sources at 1.4 GHz. A thorough investigation based on dedicated radio observations towards these candidates, HIP 88652, HIP 38430, and HIP 11891 is under way (Peri et al., in prep.).

4. Results

4.1. Group 1

We searched for WISE and MSX data toward the 56 group-1 targets, at all available wavelengths, with a field size of 1 sq degree. About 55% of the targets have MSX data, and 70% have WISE data. Table 1 gives the Hipparcos number, alternative names, and the spectral type of the stars as in NC97, or updated from GOSC (Maíz Apellániz et al. 2004) whenever possible. Columns 4 and 5 list the results obtained with MSX and WISE. We identified several types of structures around the stars that are represented with different symbols in Table 1. The categories are: a point source at the position of the star, diffuse emission in the vicinity of the star, no emission, a bow-shaped emission feature, a bubble candidate, and extended emission in the field. A “?” symbol implies doubtful feature. The last column reproduces the results of NC97. Two special cases are noted in the table. Cappa et al. (2008) studied the environs of HD 92206, and identified an extended HII region. The star HD 36862 is very close to HD 36861 (O8III); they both belong to the rich cluster λ Ori (Bouy et al. 2009), are not single isolated objects, and have a large spatial velocity. In both cases, a bow shock might be hidden and we discarded these two objects from our sample.

We found 18 bow-shock candidates (BS-C) out of the 56 stars of group 1 (Table 1). Of the 18 BS-C, 3 of them were detected with MSX (one is the case of BD +43° 3654) and the other 15 with WISE. Out of a total of 18 BS-C, 14 were also classified as BS-C by NC97 (see Table 1). A total of 4 new BS-C were found here, related to HIP 24575, HIP 38430, HIP 77391, and HIP 114990. We believe that the different results between group 1 stars and NC97 are due to the high resolution and sensitivity of the more recent data.

4.2. Group 2

We list the 244 O-B2 runaway candidates extracted from Tetzlaff et al. (2010) in Table 2. We searched the WISE data to identify BS-C. The table is divided into two parts: the upper part contains those stars for which WISE data were released, and the lower portion where they were not.

We found a total of 17 BS-C, marked with bold font in the top part of Table 2; seven of them had been identified in group 1.

4.3. The E-BOSS sample

The 28 objects of the E-BOSS sample are shown in Figs. 1 to 5. In Table 3, the star name is given in the first column, the group membership in Col. 2, and Galactic coordinates in Cols. 3 and 4. Spectral types are in Col. 5. The distances, in Col. 6, were taken from several sources, namely Megier et al. (2009), Mason et al. (1998), Schilbach & Röser (2008), Hanson (2003), and Thorburn et al. (2003), or derived from Hipparcos parallaxes (van Leeuwen 2007). The wind terminal velocities v_∞ are either from Howarth et al. (1997) or derived using Table 3 of Prinja et al. (1990). To compute the stellar mass-loss rates Ṁ, we used the routine described by Vink et al. (2001), and stellar parameters derived from standard models (e.g. Martins et al. 2005). The stellar tangential velocities v_tg were taken from Tetzlaff et al. (2010), or derived from proper motions (van Leeuwen 2007, see last two columns of Table 3). Radial velocities v_r are from the Second Catalogue of Radial Velocities with Astrometric Data (Kharchenko et al. 2007).

In Figs. 1 to 5, we show the WISE and MSX images of the 28 BS-C. Superimposed on the images, we have plotted with arrows two directions, representing the proper motion of the star derived by van Leeuwen (2007), and that of the corrected proper motion after taking into account the Galactic rotation of the ISM at the location of the star (as done for example in Comerón & Pasquali 2007; Moffat et al. 1998, 1999).

For each BS-C, we measured geometrical parameters, such as the spatial extent l, the width w, and the distance from the star to the midpoint of the bow shock structure R (Table 4, Cols. 2 to 7).

The ISM ambient density in the vicinity of the star n_ISM can be estimated using the expression that gives the so-called stagnation radius R₀ (see Wilkin 1996, for definition and details)

where the ambient medium density is ρ_a = μ   n_ISM and v_∗ is the spatial stellar velocity. We estimated the volume density of the ISM in H atoms at the bow shock position assuming R₀ ~ R, a mass per H atom μ = 2.3 × 10^-24 g, and the helium fractional abundance Y = 0.1. The values obtained for n_ISM are given in column 8 of Table 4, and should be interpreted with caution. In many cases, the width of the bow shock is substantial compared to R, which adds an uncertainty to the R₀ values used in the last equation.

There are additional factors that might affect the values of R₀ and n_ISM, such as errors in the mass-loss rate due to clumping, in addition to the potential source of errors in the parameters used, as described above.

5. Statistics

In Figs. 6 and 7, we present the (l,b) distribution of group 1 and group 2 stars, showing those with and without bow shocks. There does not seem to be any preferential location for stars with bow shocks. We note that there are no bow shocks at high latitudes, but that the small number of stars there means we cannot say anything conclusive.

Fig. 6 Distribution on the (l,   b) plane of group-1 stars.

Fig. 7 Distribution on the (l,   b) plane of group-2 stars.

Figures 8 and 9 show the occurrence of bow shocks as a function of spectral type for each group. We might have expected more detections of bow shocks from the more massive, earlier-type, or fastest stars, which is not seen. The number of stars in each subtype grows with later spectral types, probably reflecting that there is no strong bias in the Group 2 stellar sample.

Figures 10 and 11 show that bow shocks were detected around both the lower mass stars and the youngest stars, but the very small numbers in many of the bins prevents us to draw any clear conclusion.

6. Discussion

The E-BOSS sample introduced here constitutes the most substantial sample of stellar bow shocks and should provide a reliable basis to more detailed studies of the structure and formation of bow shocks.

For group 1, the availability of MSX and WISE images has enabled us to improve previous results, which relied on the IRAS data. Some structures were identified as bow shocks, others rejected, and in some cases new bow shocks were revealed. One interesting example is the feature around HD 34078 (HIP 24575). The IRAS data, discussed in NC97, revealed an excess of emission at 60 μm, but no discernible bow shock. With the WISE bands, two structures can be clearly seen (Fig. 2). There are filaments around the star mainly at longer wavelengths (red = 22.2 microns), and a typical bow-shaped feature becomes visible in the longer bands (red = 22.2 microns and green = 12.1 microns) in the direction of the stellar motion. Another example from group 1 is the object related to the B5 III star HD 22928 (HIP 17358). The combination of low mass-loss rate and wind terminal velocity, with a high stellar velocity leads to the formation of a bow shock very close to the star. This situation results in a very large derived value of the ISM density (Table 4). The stellar motion is dominated by the tangential velocity, and is a nearby star. These help us to more clearly resolve the bow shock.

The strategy we adopted here for group 2, beginning with the early-type runaways, which are most likely to form bow shocks, has been successful in identifying a significant number of new bow shocks. These results support the hypothesis that some runaway stars (around 10%) produce detectable bow shocks. There are many possible reasons why a detectable bow shock is not formed, such as a low-density ISM, an extremely high stellar velocity, or a low mass-loss rate, the inclination of the stellar velocity vector with respect to the plane of the sky, and confusion with strong field sources.

The current release of WISE data covers around 57 per cent of the sky, but covers a slightly larger fraction of the Galactic plane. For our group 2 sample (244 stars), around 67 per cent of the stars were covered in the first WISE data release (164 stars). A total of 17 of these were found to have bow shocks (a detection rate of  ~10 per cent). On the basis of these statistics, we would expect to find around 8 additional bow shock candidates in the remaining 80 group 2 sources.

Fig. 8 Distribution of group-1 stars by spectral types.

Fig. 9 Distribution of group-2 stars by spectral types.

Returning to the detected BS-C, several bow shocks in both group 1 and group 2 show a complex layered structure, such as HD 30614 (HIP 22783), HD 42933 (HIP 29276), and HD 15629 (HIP 11891). In many cases, but not all, the structures are aligned with the stellar velocity. An example of a misaligned bow shock candidate is HD 36512 (HIP 25923). We include this object in the sample because at this stage we lack any precise knowledge of the true stellar velocity.

HD 48099 (HIP 32076) is maybe the most classic example of a bow shock, and this object will be the subject of a future detailed study that will include modeling of the infrared emission in the formation of a bow shock (del Valle et al., in prep.).

With this sample, we will be able to investigate the IR luminosity and dust temperature and compare the scaling of these quantities with stellar parameters (van Buren & McCray 1998), and also detectability (Stevens et al., in prep.).

The current sample is insufficiently large to distinguish properties according to stellar luminosity class, binarity status, or particular stellar classes, such as Of-type stars. However, when new data are released, the situation will improve.

7. Summary and prospects

We have discovered a significant number of bow shock candidates around early-type runaway stars providing a higher quality sample of bow shocks, the E-BOSS sample. For the set, we have determined a number of parameters of these objects. We have found no strong trends concerning the frequency of bow shocks with stellar mass, position, age, velocity, and spectral type.

In terms of future work, extending our study to later spectral types will allow us to systematically search for and investigate radiative bow shocks (see for example Gáspár et al. 2008). The extensive Tetzlaff et al. (2010) database will also serve as a starting point for that study. We will also incorporate in forthcoming E-BOSS versions other runaway databases together with individual objects from the literature.

Detailed studies of individual objects will help us to more clearly understand the stellar winds of the bow-shock producer stars, the medium in which they travel, and the stellar history, among other things. We plan to continue our studies of individual objects in the E-BOSS sample, as well as statistical studies of the global sample.

Fig. 10 Distribution of group-2 stars by stellar mass (5 M⊙ binning).

Fig. 11 Distribution of group-2 stars by age (5 Myr binning).

Acknowledgments

C. S. Peri and P.B. are supported by the ANPCyT PICT-2007/00848. P.B. also acknowledges support from CONICET PIP 0078 and UNLP G093 projects, and thanks the School of Physics and Astronomy of the University of Birmingham for kind hospitality. This publication makes uses of the NASA/IPAC Infrared Science Archive, which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration, the SIMBAD database, operated at CDS, Strasbourg, France, and data products from the Wide-field Infrared Survey Explorer, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, funded by the National Aeronautics and Space Administration. C. S. Peri is grateful to M. V. del Valle for a discussion on theoretical issues. We also thank an unknown A&A referee for the comments and suggestions that have improved the article.

References

All Tables

All Figures

Fig. 1 WISE images of the bow shock candidates around 6 OB stars. Color mapping: blue = 3.4 microns; green = 12.1 microns; red = 22.2 microns. The color scales are in data numbers (DNs); the WISE data is not calibrated in their surface brightness. The vectors indicate the direction of the star proper motion: the thicker one represents the one derived from Hipparcos data by van Leeuwen (2007); the thinner one is the same but corrected for the ISM motion caused by Galactic rotation. The vectors length are not scaled with the original values.

In the text

	Fig. 2 Same as in Fig. 1, for another six OB stars.
In the text

	Fig. 3 Same as in Figs. 1 and 2, for another six OB stars.
In the text

	Fig. 4 Same as in Figs. 1–3, for another six OB stars.
In the text

	Fig. 5 WISE images of bow shock candidates as in Figs. 1–4, for HIP 97796 and 114990. MSX images for BD +43° 3654 and HIP 101186. Color mapping for MSX: blue = 8.3 microns; green = 12.1 microns; red = 21.3 microns.
In the text

	Fig. 6 Distribution on the (l,   b) plane of group-1 stars.
In the text

	Fig. 7 Distribution on the (l,   b) plane of group-2 stars.
In the text

	Fig. 8 Distribution of group-1 stars by spectral types.
In the text

	Fig. 9 Distribution of group-2 stars by spectral types.
In the text

	Fig. 10 Distribution of group-2 stars by stellar mass (5 M⊙ binning).
In the text

	Fig. 11 Distribution of group-2 stars by age (5 Myr binning).
In the text