Runaway O-star Bow Shocks as Particle Accelerators? The Case of AE Aur revisited
Blagoy Rangelov, Thierry Montmerle, S. R. Federman, Patrick Boisse, Stefano Gabici
DDraft version September 5, 2019
Typeset using L A TEX twocolumn style in AASTeX62
Runaway O-star Bow Shocks as Particle Accelerators? The Case of AE Aur revisited
Blagoy Rangelov, Thierry Montmerle, S. R. Federman, Patrick Boiss´e, and Stefano Gabici Department of Physics, Texas State University, 601 University Drive, San Marcos, TX 78666, USA Sorbonne Universit´e, CNRS, UMR 7095, Institut d’Astrophysique de Paris, 98bis, Bd Arago, 75014 Paris, France Department of Physics & Astronomy, The University of Toledo, 2801 West Bancroft Street, Toledo, OH 43606, USA APC, Universit´e Paris Diderot, CNRS/IN2P3, CEA/IRFU, Observatoire de Paris, Sorbonne Paris Cit´e, F-75013 Paris, France
ABSTRACTWe present results of our
Chandra /ACIS observations of the field centered on the fast, runaway Ostar AE Aur and its bow shock. Previous
XMM-Newton observations revealed an X-ray “blob” near theIR arc tracing the bow shock, possibly a nonthermal source consistent with models of Inverse Comptonscattering of dust IR photons by electrons accelerated at the shock. The new, subarcsecond resolution
Chandra data, while confirming the presence of the
XMM-Newton source, clearly indicate that thelatter is neither extended nor coincident with the IR arc and strongly suggest it is a backgroundAGN. Motivated by results published for the bow shock of BD+43 ◦ . . × − erg cm − s − (3 σ ) and, in the radio range, of 2 mJy (1.4 GHz), and 0.4 mJy (5.0 GHz),are used to put constraints on model predictions for particle acceleration within the bow shock. Inthe “classical” framework of Diffusive Shock Acceleration, we find that the predicted X-ray and radioemission by the bow shock is at least two orders of magnitude below the current upper limits, consistentwith the systematic non-detections of up to 60 stellar bow shocks. The only exception so far remainsthat of BD+43 ◦ Keywords: stars: individual (AE Aur, BD+43 ◦ INTRODUCTIONThe origin of cosmic rays (CRs), and more gener-ally the investigation of the physical mechanisms ableto accelerate particles (electrons and nuclei) up to en-ergies as high as 10 eV, remains one of the most fas-cinating problems in astrophysics. On a galactic scale,the origin of CRs is widely attributed to their accel-eration by supernova shock waves, seen as supernovaremnants (SNRs) expanding supersonically in the in-terstellar medium (ISM). Nuclei are able to travel formillions of years throughout the Galaxy and the Galac-tic halo (e.g., Gaisser et al. 2016), losing their en-ergy along the way by spallation collisions in the non-relativistic regime. This is deduced from various sec-ondary/primary abundance ratios at Earth, and, above ∼ π -decay induced γ -ray emission, clearlyseen in particular along the Galactic plane (e.g., Stronget al. 2007). On the other hand, electrons generally losetheir energy on much smaller distance scales by radia-tive emission (bremsstrahlung, synchrotron), detectable with radio telescopes usually in the cm range, or bycollisions with ambient photons (Inverse Compton; IC),boosting their energies up to X-rays or even γ -rays.So, from the observational point of view, one mustdistinguish between “direct” clues to investigate the ac-celeration processes (e.g., radio observations), and “in-direct” ones, i.e., those necessitating a target to be re-vealed such as molecular clouds for relativistic protons(Abdo et al. 2010; Acciari et al. 2009; Katsuta et al.2012; Katagiri et al. 2016; Tavani et al. 2010), or radia-tion from interstellar dust for electrons.In this context, by far the most popular accelerationmechanism is the so-called “Diffusive Shock Accelera-tion” (DSA) (e.g., Drury 1983). This mechanism, invarious forms, has been successfully tested (at the costof adjusting some parameters) in a variety of environ-ments, either isolated SNR (for instance in relation withhistorical supernovae, e.g., SN1006), middle-aged SNRin regions of star formation, or even colliding winds inmassive binary systems (Petruk et al. 2017; Cardillo a r X i v : . [ a s t r o - ph . H E ] S e p Rangelov et al.
Figure 1.
Exposure corrected
Chandra
ACIS-I merged im-age of AE Aur in the 0.3–7 keV energy band smoothed with a5 (cid:48)(cid:48) kernel. The two brightest sources are AGNs (see text); thesource at the center is our target, AE Aur, and the source tothe north-east corresponds to the XMM “LS blob” detectedby L´opez-Santiago et al. (2012). North is up and East is tothe left. et al. 2017; Uchiyama et al. 2010; Gabici et al. 2009;Malkov et al. 2011; Torres et al. 2010).In this paper, we want to investigate further anothersituation where cosmic-ray acceleration (here, electrons)could potentially take place: that of high-velocity, mas-sive stars (called “runaways”) traveling supersonicallythrough a dense enough ISM. In such conditions, thesestars are preceded by a more or less paraboloid-shaped“bow shock”, resulting from the collision between thestellar wind and the ambient ISM (e.g., van Buren &McCray 1988; Peri et al. 2012). In fact, direct detectionof electrons accelerated by the bow shock of a very mas-sive star, by way of their synchrotron radio emission, wasfirst obtained by Benaglia et al. (2010) for BD+43 ◦ XMM-Newton archivaldata to study one of the fastest known runaway starsassociated with a bow shock, the late-O (O9.5V) starAE Aur. This star is travelling at v (cid:63) = 150 km s − in amoderately dense ISM (Tetzlaff et al. 2011) and has been Figure 2.
AE Aur and its bow-shock in
Spitzer µ m band(this band is preferred to other shorter wavelength ones be-cause it is less contaminated by emission from AE Aur andless affected by saturation). The white region shows the bow-shock extraction region used for X-ray analysis (see text fordetails). The “LS blob” and AE Aur are shown with + and × signs, respectively. The contour avoids the “blob” regionto exclude its X-ray flux. ejected from the Orion nebula a few millions years ago(Hoogerwerf et al. 2001); it is located at a distance of530 pc and its mass-loss rate is ˙ M = 2 × − M (cid:12) yr − .A ∼
40 ks exposure (bin size 4 (cid:48)(cid:48) ) revealed a seeminglyextended X-ray “blob” spatially correlated with an arc-shaped mid-IR dust feature seen with the
Wide-field In-frared Survey Explorer ( WISE ; the spatial resolution is6 . (cid:48)(cid:48) µ m: Cutri & et al. 2012), although not withits apex. Their analysis of the X-ray spectrum (1-8 keV)favored non-thermal emission (photon index Γ ∼ − . XMM-Newton archival data, and of the higherangular resolution of
Spitzer images obtained by Franceet al. (2007) compared to
WISE (IRAC resolution of 2arcsec at 8 µ m), showed that the “LS blob” of X-rayemission near AE Aur is not spatially coincident withthe bow shock. In addition, their analysis of five otherrunaway stars yielded only upper limits to their bow-shock X-ray fluxes.In view of these conflicting results, we obtained several Chandra /ACIS images (17 (cid:48) × (cid:48) FOV) having a muchbetter angular resolution ( ∼ . (cid:48)(cid:48) handra observations of AE Aur XMM-Newton , to map out the area around AE Aur indetail and its bow shock. The resulting merged imageis shown in Figure 1.The outline of the paper is as follows. We first presentan analysis of our observations, including other sourcesin the field to help characterize the nature of the “LSblob” (Section 2). In Section 3, we find very clear ev-idence that the
XMM-Newton “LS blob” is actually afaint point source with no counterpart at other wave-lengths and unrelated to the bow shock, but having adefinitely hard, non-thermal spectrum. Further analy-sis of the global
Chandra image suggests that the “LSblob” is likely a background AGN. Moreover, we findno diffuse emission from the area delineated by the IRarc, neither in the
Chandra
X-ray image, nor in archivalEVLA radio data. In Section 4, we put AE Aur in thecontext of other well-studied runaway stars with bowshocks and discuss the resulting constraints on theoreti-cal models of particle acceleration by stellar bow shocks(Section 5). Finally, we summarize our main conclusionsin the last Section ( § OBSERVATIONS AND DATA REDUCTION2.1.
Chandra X-ray Observatory
We carried out a campaign to observe the region ofAE Aur with
Chandra . The program (PI: Rangelov)was split into five observations totaling 140.53 ks expo-sure over the period of two months. The first data setwas taken on 2016 December 16 (ObsID 19943; 14.88 ks),followed by 2016 December 17 (ObsID 19445; 44.49 ks),2017 January 3 (ObsID 19979; 26.72 ks), 2017 January4 (ObsID 19941; 26.72 ks), and 2017 January 6 (Ob-sID 19951; 27.72 ks). All data were taken with theACIS-I instrument operated in “VeryFaint” Timed ex-posure mode. We processed the data using the
Chan-dra
Interactive Analysis of Observations (CIAO ) soft-ware version 4.8 and Chandra
Calibration Database ver-sion 4.7.2. The data have been restricted to the en-ergy range between 0.3 and 7 keV and filtered in threeenergy bands, 0.3–1.2 keV (soft), 1.2–2 keV (medium),and 2–7 keV (hard). We used the CIAO’s Mexican-hatwavelet source detection routine wavdetect (Freemanet al. 2002) to create source lists. Wavelets of 1.4, 2,4, 8, and 16 pixels and a detection threshold of 10 − were used, which typically results in one spurious de-tection per million pixels. In order to find fainter point http://cxc.cfa.harvard.edu/ciao/ sources, all five datasets were merged prior to running wavdetect . We detected a total of 114 X-ray sources inthe merged data. The srcflux CIAO tool was then runindividually on each observation (using the coordinatesfound by wavdetect ).In the following, we adopt the designation CAX- nn forthese sources, where “C” stands for Chandra and “A”for AE Aur; nn is the rank when sources are ordered byincreasing right ascension. An analysis of sources in thewhole Chandra image is presented below ( § XMM-Newton Telescope
We used archival data from
XMM-Newton (PI Dami-ani, ObsID 0206360101) with total exposure time of58.9 ks. This is the same dataset as that analyzed byLS12. We re-processed the data ourselves using the
XMM-Newton
Science Analysis System (SAS ) softwareversion 17.0.0 and followed the standard source extrac-tion procedures from the SAS Data Analysis Threads to extract spectra from MOS1, MOS2 and PN (see Sec-tion 3 for details).2.3. Multi-Wavelength Data Analysis
We searched multi-wavelength (MW) catalogs for po-tential counterparts to all X-ray sources and compiled aset of MW parameters for each of the sources with coun-terparts. We collected optical measurements from theUSNO-B (Monet et al. 2003) catalog, the near-infrared(NIR) from the
Two Micron All-Sky Survey (2MASS;Skrutskie et al. 2006), and the IR from
WISE (Cutri& et al. 2012). MW sources were considered potentialcounterparts if they were located within the error radiusof each X-ray source. We used a 95% confidence levelpositional uncertainty of 1 . (cid:48)(cid:48) (cid:48) off-axis X-ray source with 15 counts (see Equation 12 in Kim etal. 2007). The X-ray properties and magnitudes for theMW counterparts (up to 11 MW parameters) are thenused to classify these sources with our machine-learningpipeline (see the Appendix in Hare et al. 2016 for de-tails).To help understand the large-scale topography of theISM in the direction of our Chandra field, Figure 2 showsthe 24 µ m Spitzer/MIPS image covering a ∼ ◦ × ◦ areacentered on AE Aur within the merged Chandra image(Figure 1). AE Aur corresponds to CAX-72 while the Standard CIAO procedures found at http://cxc.harvard.edu/ciao/threads/wavdetect merged/ were followed to merge the data.We used an exposure-time-weighted average PSF map in the cal-culation of the merged PSF. Rangelov et al.
Figure 3.
CO mapping of a ∼ . (cid:48) × . (cid:48) area coveringAE Aur and its bow shock (Pierre Gratier, private commu-nication). The yellow dotted line encompasses the IR bowshock, as seen in the Spitzer image (Figure 2), excising the“blob” X-ray source CAX-76. The red arrow indicates theproper motion of the star, which is actually located behindthe CO clumps, explaining its relatively high extinction (seetext for details.)
Chandra counterpart to the
XMM-Newton “LS blob” isCAX-76. BOW SHOCK RESULTS AND ANALYSISBefore discussing the properties of CAX-76 and pre-senting a global analysis of other sources detected in our
Chandra image, we summarize the information availableon the interstellar medium in the field. This materialinduces significant absorption which must be taken intoaccount when fitting
Chandra spectra, especially at lowenergies.3.1.
The interstellar medium in the field of AE Aur H I and H column densities in front of AE Aur wereinferred from UV spectra and amounts to N (H I ) =2.7 × cm − and N (H ) = 6.4 × cm − (Boiss´e etal. 2005), leading to a total H column density, N H = N (H I ) + 2 × N (H ) = 4 . × cm − , appropriate forcomputing AE Aur’s X-ray absorption.Figure 3 shows the ∼ . (cid:48) × . (cid:48) CO mapping at 4 (cid:48)(cid:48) resolution by Gratier et al. (2014), revealing two tinymolecular clumps along the AE Aur line of sight. How-ever, no CO emission has been detected toward CAX-76and the area delineated by the IR arc is also essentiallydevoid of CO emission. Thus, for a source lying nearAE Aur in space, the appropriate N H value should bearound 2.7 × cm − because there is little molecularmaterial in the vicinity. For extragalactic sources, the total Galactic N (H I )value in this direction should be adopted, N (H I ) =5.8 × cm − , as provided by the HI4PI survey (HI4PICollaboration et al. 2016). This value is in fact a lowerlimit since additional absorption, internal to the sources,might be present. The HI4PI survey indicates that spa-tial variations across the Chandra field are very limited( N (H I ) ranges from 5.6 up to 6.0 × cm − ), thus theabove N (H I ) value should hold throughout the field.CO mapping at 22 (cid:48)(cid:48) resolution over a significant portionof the Chandra field (Gratier et al., in preparation) indi-cates that molecular gas is present in the form of denseclumps but that their surface covering factor is no largerthan a few percent; thus, for most background sources,X-ray absorption can be computed on the basis of H I data alone (i.e., N H ≈ N (H I )).3.2. Properties and nature of source CAX-76
We detect a weak source, CAX-76, consistent withthe location of the original “LS blob” reported byLS12. The source has the following right ascension (RA)and declination (DEC), RA = 5:16:20.45 and DEC =+34:19:05.18. Its offsets with respect to AE Aur are∆RA = 34 . (cid:48)(cid:48) . (cid:48)(cid:48) Chandra resolution, implying an ex-tent no larger than about 0.5 arcsec. The source has nocounterpart in the MW catalogs. We have investigatedoptical/NIR/IR images (e.g.,
DSS , , Spitzer ) tosee whether its X-ray emission could be attributed toa background, uncatalogued (flaring) star or AGN, butwe find no counterpart either. The closest point-likestellar source, 2MASS 05162332+3420290 (
JHK magni-tudes 15.863, 15.428 and 15.208, respectively), is 15 . (cid:48)(cid:48) Chandra observation andused CIAO’s task combine spectra to combine all fivespectra into one. In order to improve the S/N ratio, wealso extracted the
XMM-Newton spectrum in a similarfashion to LS12, using an annular background regionaround AE Aur, with the inner and outer radii of theannulus placed in such a way to tightly encompass the“LS blob” (which was of course excluded from the back-ground extraction). LS12 used only the PN spectrum fortheir analysis because they report that the MOS spec-tra are not constraining enough. We have checked thatthe inclusion of the MOS spectra does not change ourconclusions, but we analyze further the
XMM-Newton
PN data only, to maintain a more direct comparison tothe LS12 results. handra observations of AE Aur Table 1.
X-ray model fit parameters for the “blob” source (CAX-76)Data Model N Ha kT b Γ c χ DOFModels with fixed N H Chandra
APEC 2.7 64 – 24.92 18
Chandra
PL 2.7 – 1 . ± . XMM-Newton
APEC 2.7 6 ± XMM-Newton
PL 2.7 – 1 . ± . Chandra + XMM-Newton
APEC 2.7 26 – 37.27 32
Chandra + XMM-Newton
PL 2.7 – 1 . ± . H Chandra
APEC 3 64 – 24.89 17
Chandra
PL 10 − – 0 . ± . XMM-Newton
APEC 8 ± . ± . XMM-Newton
PL 9 ± ± Chandra + XMM-Newton
APEC 1.9 55 – 37.13 31
Chandra + XMM-Newton
PL 2 – 1 . ± . a Hydrogen column density in units of 10 cm − for phabs model. b Temperature for APEC model in units of keV. c Photon index for PL model.
Note —Parameters without listed uncertainties indicate models where given pa-rameter reaches the lower/upper parameter boundary during the fit.
Due to the limited number of counts at low energies( (cid:47)
Chandra and
XMM-Newton
X-ray spectra. We have attempted to fit the data withabsorbed PL ( A ( E ) = KE − Γ , where Γ is the photonindex and K is the normalization) and APEC (emissionspectrum from collisionally-ionized diffuse gas) modelsin XSPEC (Arnaud 1996). For each model and dataset, we first fixed N H = N H (AE Aur), then took it asa free fitting parameter. Results are presented in Ta-ble 1. We do not provide uncertainties for parametersthat reach the lower/upper parameter boundary duringthe XSPEC fit. In these cases, the fit provides unreal-istic parameters, such as very large kT values for theAPEC models.Our
XMM-Newton -only formal PL fit yields Γ =1 . ± . N H fixed to N H (AE Aur). This is different(although consistent within 1 σ ) from the value reportedby LS12, Γ = 2 . ± .
6. While we tried to reproduce the
XMM-Newton
CAX-76 extraction procedures used byLS12, we may have taken a slightly different backgroundregion, which is subject to contamination by photonsfrom AE Aur itself. This affects more heavily the lower energy X-ray photons (AE Aur completely disappearsabove (cid:39) . ± . XMM-Newton reprocessed value, Γ = 1 . ± . Chandra data,and Γ = 1 . ± . XMM-Newton and
Chandra spectra. Therefore, we concur with LS12in that there is little doubt that the CAX-76 spectrum,although difficult to fit, is hard and cannot be thermal.Note that, contrary to
XMM-Newton , there is no
Chan-dra data below 1 keV. This will affect the spectral fits,especially the derived extinction. As discussed below(Section 3.4), the analysis of other X-ray sources in the
Chandra field provides further evidence that CAX-76is most likely a background AGN. The various valuesof Γ found for CAX-76 can be compared with the onesfound for AGNs in the 2 −
10 keV interval, which lie inthe range Γ ∼ . − . Rangelov et al.
Figure 4.
Chandra and
XMM-Newton
X-ray spectra ofCAX-76 shown in red and black, respectively. A simple ab-sorbed PL (Γ = 1 .
3) model is used for illustration (see textand Table 1 for details).
Zooming in: X-rays from the bow shock?
We find no evidence of extended X-ray emission inthe vicinity of AE Aur associated with its bow shock.Using the exposure corrected X-ray and
Spitzer µ mimages (which do not show saturation), we created an ad hoc extraction region covering the apex of the bowshock (shown in Figure 2 and Figure 3), designed insuch a way as to avoid contamination from point sources(specifically CAX-76 and CAX-72) and chip gaps. Wemeasured a count rate of 1 . × − cts s − (0.3–7 keVband) in this 1771 arcsec extraction region. For com-parison, we obtained the upper limit to the bow-shockflux by sampling the background count rates from 12different regions in the vicinity of AE Aur and by calcu-lating the standard deviation from the mean value. Thebackground regions had the same shape and size as thebow-shock region. The average count rate and standarddeviation are (1 . ± . × − cts s − . We considerthe standard deviation as a conservative 1 σ upper limit,corresponding to a 3 σ upper limit of 3 . × − cts s − inthe 0.3–7 keV band. Upper limits on fluxes can then beestimated using the Chandra
PIMMS tool . Assumingthat the background emission can be modeled by a PLspectrum with Γ = 1 . N H = 2 . × cm − (moleculargas is present in front of AE Aur, but not towards theIR arc), the upper limits to the absorbed (unabsorbed) http://cxc.harvard.edu/toolkit/pimms.jsp Hard Color S o f t C o l o r Figure 5.
Hardness ratio diagram for all sources based onall five observations. The two colors are created using thefollowing equations
Soft Color = (M-S)/T and
Hard Color= (H-M)/T , where S, M, H and T are the counts in thesoft (0.5–1.2 keV), medium (1.2–2 keV), hard (2–7 keV) andbroad (0.5–7 keV) bands, respectively. Error bars are shownfor sources with more than 40 counts in all observations. fluxes and luminosities are 6( ± × − erg cm − s − and 2( ± × erg s − (for d = 530 pc), respectively.3.4. Global Chandra/ACIS image analysis
Without any preconceived ideas about the nature ofthe sources (even when there were obvious counterpartson MW images, which was rare; more below), we firstran our MW classification tool on all X-ray sources inthe
Chandra field (see Figure 1). The automated pro-cedure produced eight classifications with a confidencelevel higher than 70%, including three “AGNs” and five“stars”. Note that the calculated confidence levels foreach type do not include uncertainties associated withthe X-ray flux determination, which can be substantialfor faint X-ray sources; they also do not include thepossibility of assigning false MW counterparts to the X-ray sources. Table 2 shows the MW parameters (whenavailable) for ObsID 19445. We do not list class designa-tion and classification with confidence levels below 70%because we do not deem those results accurate enough(the accuracy is lower usually due to the lack of MWcounterparts). (Only the ten classified sources, togetherwith CAX-72 and CAX-76, are displayed for brevity,while the entire table is available in electronic format.)To help with the investigation of the nature of the X-ray sources, we also made hardness ratio diagrams (seeFigures 5 and 6) constructed as follows:
Soft Color =(M-S)/T and
Hard Color = (H-M)/T , where S, M, H handra observations of AE Aur Table 2.
X-ray sources (ObsID 19445)CAX a F X b I c J d W1 e Class Probability12 78.9413882 34.2290756 3396.7 235 ± . . ± . . ± . f . ± . g . ± . . ± . h . ± .
485 79.1070493 34.3624738 18.4 0 . ± . . ± . . ± . a Background subtracted net counts for ObsID 19445 only. b X-ray flux in the 0.5–7 keV band in units of 10 − erg cm − s − for ObsID 19445. c I magnitude from USNO-B1. d J magnitude from 2MASS. e W1 magnitude from WISE . f This source is unidentified, but has a radio counterpart, EVS-2. See Table 3. g This source is AE Aur. h This source is the “LS blob”.
Note — Only sources with confident classification and the “LS blob” are listed here. The available multi-wavelength parameters are used to classify these sources with our machine-learning pipeline (see theAppendix in Hare et al. 2016 for details), which produces a classification (“Class”) and correspondingclassification confidence (“Probability”). This table is available in its entirety in electronic format. and T are the counts in all five observations in the soft(0.5–1.2 keV), medium (1.2–2 keV), hard (2–7 keV) andbroad (0.5–7 keV) bands, respectively. Figure 6 showstwo absorbed models: PL with Γ in the [0.4, 4] in-terval with a step of ∆Γ = 0 .
4, and bremsstrahlungwith kT in the [0.1, 10.0] (keV) interval in logarith-mic scale with ∆ log ( kT ) = 0 .
33. Both models have N H = 2 . × cm − , the value obtained for AE Aur.These figures show that one cannot distinguish betweenPL models with Γ ≥ kT ≥ Evidence for radio emission from the bow shock?
The presence of energetic electrons associated with abow shock might also be revealed by non-thermal ra-dio emission, like that detected near the massive O star,BD+43 ◦ software are displayedin Figure 7 together with the Chandra
X-ray image.Detailed inspection of these images near the positionof AE Aur reveals no hint of extended emission at theposition of the IR arc, in particular around the apex.Using the same procedure as described above for the X-ray emission (Section 3.3), we obtain 3 σ upper limits of1 . × − mJy/arcsec and 1 . × − mJy/arcsec onthe brightness of the arc at 1.4 and 5 GHz, respectively,corresponding to upper limits for the flux emitted withinthe 1771 arcsec area of 2 mJy and 0.4 mJy. A num-ber of point sources are clearly detected in the L and Cbands: their flux values together with the correspond- https://casa.nrao.edu/ Rangelov et al.
Table 3.
Selected radio sources in the C- and L-band archival EVLA fields centered on AE AurEVS- a RA DEC L flux (mJy) b C flux (mJy) c α d IdentificationEVS-1 78.9412408 34.2294106 3.19 CAX-12 (AGN)EVS-2 79.0314167 34.2807639 1.47 0.93 − .
36 CAX-51 e EVS-3 79.0705606 34.3124611 3.24 0.81 − .
08 unidentifiedEVS-4 79.1143333 34.2407917 13.1 1.60 − .
63 unidentifiedEVS-5 79.1932599 34.2359209 2.27 CAX-104 (AGN) a “EVS” = EVLA source (see Figure 7). b L band = tuned at 1.52 GHz (config. C: beamsize 11”). c C band = tuned at 5.5 GHz (config. B: beamsize 1 . d α = spectral index from L frequency to C frequency. e EVS-2 coincides with CAX-51 to better than 0 . Hard Color S o f t C o l o r Figure 6.
Shown are absorbed PL (blue line and dots)with Γ in the range of 0.4–4 (∆Γ = 0 .
4) and absorbedbremsstrahlung (red line and triangles) with kT in the rangeof 0.1–10 (keV) in logarithmic steps (∆ log ( kT ) = 0 . N H = 2 . × cm − . For clarity, threeof the model parameters (0.1, 0.77 and 10) are listed next tothe symbols for the BR model, while only one (0.4) for thePL model. Only X-ray sources with more than 40 counts (inall observations) are plotted with black dots. ing spectral indices ( α ) and identifications are given inTable 3.As shown in the composite Chandra -EVLA Figure 7,two strong EVLA sources are detected in the L band(but not in the C band because of the reduced FOV):EVS-1 and EVS-5. Both show typical X-ray spectra of AGNs. Another EVLA source, EVS-3, is present nearAE Aur but, with an RA offset of − (cid:48)(cid:48) , it is clearly dis-tinct from the star. AE Aur (= CAX-72) is undetectedin either band, which is consistent with the 3 σ upperlimit of 0.36 mJy previously obtained with the VLA at6 cm (4860 MHz) by Bieging et al. (1989). We also giveflux values in Table 3 for two other sources detectedin both bands, EVS-2 (= CAX-51) and EVS-4. Notethat EVS-4 appears close to CAX-82 in Figure 7 but isnot coincident with it. These three EVLA sources haveno other counterpart. Given their steep spectral index,EVS-2 and EVS-3 could be distant non-thermal radiogalaxies.The region of AE Aur was also observed at 1.4 GHzand 45 (cid:48)(cid:48) resolution as part of the NRAO VLA Sky Survey (NVSS; Condon, & Kaplan 1998), which allowed us tosearch for radio counterparts of our X-ray sources in theNVSS point source catalogue. Only two sources, CAX-12 (= EVS-1) and CAX-104 (= EVS-5), have NVSScounterparts, which further supports our findings thatthese are likely AGN (we checked that the fluxes givenin the NVSS catalogue for the EVS-3 source and theradio counterparts to CAX-12 and CAX-104 are consis-tent with those measured in the EVLA L-band image).The spectra of both X-ray sources can be satisfactorilyfitted using absorbed PL models with N H ≈ cm − ,in agreement with the constraints on N H discussed inSect.3.1 and the presence of some internal absorption.In Figure 8, we display the spectra of CAX-72 (AE Aur;thermal), CAX-12 (AGN; non-thermal) and CAX-76(the “LS blob”). The spectrum of CAX-76 looks hard,even AGN-like, but the X-ray data alone (no MW coun- handra observations of AE Aur Figure 7.
X-ray (
Chandra ) vs. Radio (EVLA) images centered on AE Aur (16 (cid:48) × (cid:48) ). The EVLA C-band field and beamare much smaller than for the L-band because of the different EVLA configurations (C and B respectively). For comparison wehave indicated the equivalent Chandra “beam” of ∼ . (cid:48)(cid:48) Chandra
PSF is notGaussian). The radio sources discussed in the text are designated “EVS-1” to “EVS-5” (“EVS” standing for “EVLA Source”).Identifications (and non-identifications) are indicated in italics. The position of AE Aur is indicated by a white cross, and thebow shock is outlined by the yellow dotted contour. The faint X-ray LS “blob” is also indicated on the
Chandra image. Furtherdetails are given in Table 3 and in the text ( § Figure 8.
X-ray spectra of three sources : “blob” (= CAX-76; grouped by 5 counts per bin), AE Aur (= CAX-72;40 counts per bin), and one radio-detected AGN (EVS-1 =CAX-12; 100 counts per bin). terpart was found) is not sufficient to fully characterizethe true nature of this source. DISCUSSION4.1.
AE Aur in context
As mentioned in the Introduction, AE Aur is one ofthe fastest runaway stars ( v (cid:63) = 140 km s − ) showing abow shock. This is the main reason why it was consid-ered as a prime candidate to test particle accelerationby stellar bow shocks via their X-ray emission, but aswe confirm in this paper, no diffuse, arc-shaped X-ray emission is seen when observed at high spatial resolu-tion with Chandra , nor is there any indication of radioemission being present from EVLA observations.However, other parameters, intrinsic to the star, maybe important, from the point of view of momentumtransfer rate from the shock to particles: the terminalwind velocity ( v ∞ ), and the mass-loss rate ( ˙ M ). In ad-dition, 2D hydrodynamic models by Green et al. (2019)show that the density of ISM and the orientation ofthe shock with respect to the observer are important.While, for massive stars, the terminal velocities are al-ways comparable ( v ∞ ≈ − − , which is ofthe same order as the escape velocity [ v ∞ = 2 − × v esc ;see Groenewegen et al. 1989]), ˙ M may be very different,depending primarily on the spectral type (i.e., the UVradiation field), but even within the same spectral type(case of “weak wind” stars; Shenar et al. 2017). For AEAur, v ∞ = 1200 km s − , and ˙ M = 2 × − M (cid:12) yr − .In the above context, two more runaway bow shocks(from the list of Peri et al. 2012) have been recentlystudied in some detail, associated with one late O star(like AE Aur), and one early O star:1) ζ Oph (O9.5Vnn; d = 200 pc; v (cid:63) = 24 km s − ): v ∞ = 1500 km s − , ˙ M = 2 × − M (cid:12) / yr − , one orderof magnitude smaller than AE Aur, in spite of havingthe same spectral type; and moving much slower;2) BD+43 ◦ d = 1 . v (cid:63) = 14 km s − ): v ∞ = 2300 km s − , ˙ M = 6 . × − M (cid:12) / yr − , over oneorder of magnitude higher than AE Aur, and havinga much earlier spectral type (higher mass, much moreluminous); twice as slow as ζ Oph, and 10 times slowerthan AE Aur.0
Rangelov et al.
Observational evidence for particle acceleration?
The only direct indication so far of electron acceler-ation by a bow shock is the detection of resolved, ex-tended radio synchrotron emission coinciding with thatof BD+43 ◦ ◦ ζ Oph (which doesnot show radio emission), were not detected in X-rays,respectively, by
Suzaku (Schulz et al. 2014) or
XMM-Newton and by
Chandra or Suzaku (Toal´a et al. 2016).Extending the sample to six more runaway O stars inthe
XMM-Newton archive did not produce more X-raydetections (Toal´a et al. 2017), nor did a study of a largesample of 60 IR-bright galactic stellar bow shocks by
Chandra (Binder et al. 2019). On the other hand, themodel for BD+43 ◦ Fermi theIC dust γ -ray emission would be undetectable. Using asimilar model it would be detectable for ζ Oph, becauseit is much closer (del Valle & Romero 2012; in spite ofits smaller ˙ M and v (cid:63) ). However, this was not confirmedby a very sensitive Fermi data analysis, where 27 bowshocks from Peri et al. (2012) were analyzed by Schulzet al. (2014), with no detection and with upper limits ∼ THEORETICAL CONSTRAINTS ON PARTICLEACCELERATION5.1.
Expected X-ray emission from the bow shock
The available mechanical energy in bow shock sys-tems is dominated by the kinetic luminosity of the stellarwind, which for AE Aur is equal to: L w = 12 ˙ M v ∞ ∼ erg / s . (1)A fraction η e of such energy can be converted intonon-thermal electrons through DSA operating at thewind termination shock (Benaglia et al. 2010), and theaccelerated electrons can in turn emit non-thermal ra- diation from radio frequencies up to the X-ray domainand possibly beyond (e.g. del Palacio et al. 2018 andreferences therein). According to DSA theory, electronsare accelerated at the shock and injected in the sys-tem at a rate Q e ( E e ) = Q ( E e /m e c ) − , where the nor-malisation constant Q can be computed by imposing (cid:82) d E e Q e ( E e ) E e = η e L w , which gives Q = η e L w / ( m e c ) ln( E max /m e c ) . (2)Here, E max is the maximum energy of the electrons ac-celerated at the shock and m e c the rest mass energy ofthe electron.Electrons accelerated at the wind termination shockcan produce non-thermal X-ray photons either as theresult of IC scattering soft ambient photons, or via syn-chrotron emission in the magnetic field compressed (andpossibly even amplified) at the shock. Let us considerfirst the IC scattering channel. The two most promi-nent photon targets to be considered are the radiationcoming from the dust heated by the bow shock andthat coming from the star. These two thermal radiationfields are characterized by temperatures of T d ∼
100 K(France et al. 2007) and T (cid:63) ∼ . × K (Martins etal. 2015), which in turn correspond to typical photonenergies ( ∼ . kT , k is the Boltzmann constant) equalto (cid:15) d ∼ × − eV and (cid:15) (cid:63) ∼ E e , of the am-bient photon (cid:15) , and of the upscattered photon E X arerelated as E X ∼ γ (cid:15) , where γ = E e /m e c is the electronLorentz factor. This implies that the electrons emittingIC photons in the X-ray band ( E X ≈ E e ≈ − τ c neededto advect electrons away from the system (see e.g., Fig-ure 1 in Pereira et al. 2016). Under these circum-stances, the equilibrium spectrum of electrons is sim-ply N e ( E e ) ∼ Q e ( E e ) τ c , and the IC luminosity of thesystem L X can be estimated as: E X L X ( E X ) ∼ N e ( E e ) P IC ( E e ) d E e d E X E X (3)where P IC ≡ d E e / d t is the power emitted by an electronof energy E e in the form of IC photons. Note that P IC scales as the total energy density of soft ambient photons(dust emission plus radiation from the star). The ICflux observed at the Earth can then be expressed in a handra observations of AE Aur τ IC ≡ E e /P IC ( E e ): E X F X ( E X ) ∼ η e (cid:18) L w πd (cid:19) τ c τ IC ( E e ) (cid:20) ln (cid:18) E max m e c (cid:19)(cid:21) − (4)where d ∼
530 pc is the distance to the AE Aur bowshock system.The expected IC flux from AE Aur depends then onthree poorly known quantities: the electron accelerationefficiency η e , the ratio between the characteristic advec-tion and IC time scales τ c /τ IC , and the maximum energyof the accelerated electrons E max . In order to choose anappropriate value for η e we can proceed with an anal-ogy with supernova remnant shocks, which are charac-terized by shock velocities and Mach numbers similarto those of wind termination shocks in bow shock sys-tems. Supernova remnant shocks are believed to accel-erate mainly cosmic ray protons (with an accelerationefficiency of ≈ γ -ray domain) constrain the accelerationefficiency of electrons to values of the order of η e ≈ − or less (Cristofari et al. 2013; H. E. S. S. Collaborationet al. 2018). Any realistic estimate of the ratio τ c /τ IC should rely on detailed modeling, especially for whatconcerns the exact relative spatial distribution of accel-erated electrons and soft ambient photons (coming bothfrom the runaway star and from the dust heated by thebow shock). Even though this is not a straightforwardtask, several works seem to converge toward values ofthe order of τ c /τ IC ≈ − − − for the relevant elec-tron energies (see e.g., Pereira et al. 2016; del Palacio etal. 2018 for a modeling of AE Aur and other bow shocksystems). In particular, Pereira et al. (2016) claim thatfor the AE Aur system the energy density of the am-bient radiation is largely dominated by dust emission,and found a value for the ratio τ c /τ IC roughly equal to10 − . Finally, values of E max of the order of (cid:46) ≈
10. Given these fiducial values we can nowestimate the IC flux in the X-ray band as E X F X ( E X ) (cid:46) × − (cid:16) η e − (cid:17) (cid:18) τ c /τ IC − (cid:19)(cid:20) ln( E max /m e c )10 (cid:21) − erg / cm / s (5)which is indeed an upper limit since it is based on themost optimistic value for η e .Our estimate of the IC X-ray emission is consistentwith the upper limit reported in Sec. 3.3, but is in dis- agreement with previous and more optimistic estimatesreported in the literature (e.g., L´opez-Santiago et al.2012; Pereira et al. 2016; del Palacio et al. 2018). Themain reason for that is the different choice of the param-eter η e , which in these previous works was assumed tobe equal to η e ≈ .
1, which implicitly implies that windtermination shocks were assumed to be much more ef-fective than SNR shocks in accelerating electrons. Withthis respect, our estimate is thus more conservative: weassumed that wind termination shocks behave as SNRones.To conclude, we stress that a similar reasoning canbe adopted (referring to AE Aur) to interpret the ra-dio upper limits. For the fiducial value of the magneticfield of ∼ µ G, electrons of energy (cid:38) τ s for such electrons isof the order of several megayears. An estimate of theexpected radio emission from the bow shock can thenbe obtained by substituting τ IC with τ s in Eq. 5 andby noticing that τ c /τ s ≈ − ... − , which is about2 orders of magnitude smaller than τ c /τ IC . However,this is almost exactly compensated by the fact that theradio upper limits are much more stringent than thosederived from X-ray observations. Therefore we can con-clude that also the non-detection of the bow shock ofAE Aur in radio waves is consistent with the assump-tion that such shocks behave as SNR ones.Within this framework, the detection of radio syn-chrotron emission from the bow shock of the runawaystar BD+43 ◦ . × − M (cid:12) /yr (Peri et al. 2012) up to1 . × − M (cid:12) /yr (Kobulnicky et al. 2010), i.e. a fac-tor of ≈ ...
800 larger than that estimated for AE Aur.Also, the estimated velocity of the wind of BD+43 ◦ ≈ ... × times largerthan that of AE Aur, which under certain conditionsmight explain the enhanced synchrotron radio emissionfrom BD+43 ◦ . We note that bow shocks associated with jets from youngstars (e.g., Herbig-Haro objects) can accelerate particles and giverise to detectable radio synchrotron emission under certain condi-tions (e.g., Padovani et al. 2015; Anglada et al. 2018). Two recentexamples are the HH80-81 complex of aligned knots (Rodr´ıguez-Kamenetzky et al. 2019), or the jet from a massive star in theG035.02+0.35 star-forming region (Sanna et al. 2019). However,there are major differences with runaway bow shocks: in par-ticular, in the cases mentioned, the magnetic field must be high( B ∼ . − Rangelov et al.
Maximum electron energy
The estimate of E max deserves some further discus-sion. Although it has little impact on the estimate ofthe IC X-ray emission (it enters Eq. 4 as a logarithm),it might dramatically affect the estimate of the syn-chrotron X-ray emission. Electrons of energy E e gy-rating around a magnetic field of strength B radiatesynchrotron photons of energy: E sX ≈ . (cid:18) B µ G (cid:19) (cid:18) E e
10 TeV (cid:19) keV (6)so the question arises whether magnetic fields of the or-der of hundreds of µ G could be found at wind termina-tion shocks, and/or whether electrons can be acceleratedthere beyond 10 TeV.del Palacio et al. (2018) noticed that such large valuesof the magnetic field at the shock are unlikely to be ofstellar origin, but require some form of in situ ampli-fication mechanism. One way to estimate the value ofthe amplified magnetic field is to persist with the anal-ogy with SNR shocks. In SNRs the magnetic field canbe significantly amplified at the shock due to plasmainstabilities connected with the acceleration of cosmicrays (see e.g., Bell et al. 2013). X-ray observations of anumber of SNR shocks showed that a small fraction ξ B (about few percent) of the shock ram pressure can beconverted into magnetic field (V¨olk et al. 2005). If weassume that the position of the wind termination shockroughly coincides with the standoff radius R s , i.e. thedistance from the star where the ram pressure of thewind ˙ M v ∞ / πR s balances that of the ambient medium (cid:37) ISM v (cid:63) (Wilkin 1996), the value of the magnetic fieldcan be estimated as B π = ξ B (cid:37) ISM v (cid:63) (7)which gives: B ≈ (cid:18) ξ B . (cid:19) / µ G , (8)where (cid:37) ISM is the mass density of the interstellarmedium at the location of the bow shock. In orderto obtain a conservative value for the magnetic field,we adopted a value of the ambient gas density of ∼ − , as reported in Peri et al. (2012). A larger valueof ∼
20 cm − was found by Gratier et al. (2014), and the base of the jets rather than being present in the ambient ISM;on the other hand, the successful VLA observations of HH objectsare much more sensitive ( ∼ µ Jy/beam) than in the archivaldata for the AE Aur bow shock region ( ∼ this would increase the estimate of B by a moderatefactor of (cid:112) / ∼ .
6. One can see from Eq. 6 that forsuch a value of the magnetic field, the contribution ofsynchrotron emission to the X-ray flux would be rele-vant only if E max significantly exceeds 10 TeV. Is sucha value achievable at the wind termination shock? Themodel presented in Pereira et al. (2016) seems to suggestthat this is not the case, and that the contribution fromsynchrotron radiation to the X-ray emission should benegligible.It has been argued by Benaglia et al. (2010) that un-der certain conditions bow shock systems could indeedaccelerate electrons to energies well beyond 1 TeV. Animportant consequence of this fact is that the IC emis-sion from such systems could be potentially detectablein gamma rays both in the GeV and TeV domain, by in-struments such as Fermi and the future ˇCerenkov Tele-scope Array. Bow shock systems such has ζ Ophiuchior BD+43 ◦ γ -ray sources (del Valle & Romero 2012; del Palacio et al.2018). However, also in this case the estimates of thegamma-ray emission were based upon the assumption ofa very efficient acceleration of electrons ( η e ≈ . η e , as suggested by theanalogy with SNR shocks, would make these systemstoo weak to be detected in γ -rays. CONCLUSIONSA number of studies examined the possible accelera-tion of electrons at stellar runaway bow shocks (resultingfrom stellar winds colliding with the surrounding denseISM), as revealed by X-rays from IC boosting of ambientIR shock dust emission. We revisited the case of the fastrunaway, late O star AE Aur, obtaining high-angularresolution X-ray observations with
Chandra /ACIS ob-servations. In short, confirming earlier findings basedon
XMM-Newton and IR observations, we find point-like X-ray emission at the location of the
XMM-Newton “blob” reported in the original paper by L´opez-Santiagoet al. (2012). However, no spatial coincidence with theIR signatures for the AE Aur bow shock occurs.In addition, to understand better the environment ofAE Aur on the sky, both at small spatial scales (sincedust emission is required for IC boosting to work), andat large scales (the area covered by the ACIS camera),we used CO and H I data to estimate the gas columndensity in this region. These data allow a reliable esti-mate of the extinction, hence a reliable correction of theX-ray spectrum of the “LS blob” at low energies. Merg-ing XMM-Newton and
Chandra data, we find that, givenits low count rate, it is not possible to find a good spec-tral fit, but the source is definitely hard and shows some handra observations of AE Aur σ upper limit of5 . . × − erg cm − s − , which can be used toput stringent theoretical constraints on DSA models (seebelow).By contrast, the only evidence so far for electron ac-celeration at stellar runaway shocks comes from the ra-dio domain, with the detection of non-thermal emissionfrom the early O star BD+43 ◦ Chandra
X-ray data, the EVLA upperlimits (3 σ ) on the flux from the AE Aur shock are 2mJy in the L band, and 0.4 mJy in the C band.By comparison, the non-thermal radio emission associ-ated with the BD+43 ◦ α ∼ − . ∼ L w , two to three orders of mag-nitude that of AE Aur: the stellar wind kinetic energy,not the runaway velocity, appears to be the dominantfactor for the detection of their bow shocks, at least inthe radio domain.More generally, the above X-ray and radio upper lim-its can be used to put theoretical constraints on DSAmodels for stellar bow shocks. Two key ingredients are: ( i ) the fraction η e of the wind kinetic energy convertedinto non-thermal electrons via DSA, important mainlyfor X-ray generation (IC boosting of IR photons); ( ii )the maximum energy E max of the accelerated electrons,important mainly for the synchrotron emission, via theambient ISM magnetic field B ; ( ii ) both are also im-portant for γ -ray emission in the GeV-TeV range. As-suming that stellar wind shocks are analogous to SNRshocks, we have taken η e ∼ − and E max significantlysmaller than 10 TeV, with B ∼ µ G. With such val-ues, we find that the X-ray, radio, and γ -ray emissionfor runaway stellar bow shocks are indeed undetectableat the current level of instrumental sensitivity, the caseof BD+43 ◦ Athena
X-ray satellite, or the ˇCerenkov Telescope Array (CTA) from the ground in the10 GeV-300 TeV domain. However, the angular resolu-tion of these instruments is relatively poor. Our studyhas shown that unsuspected, non-thermal extragalacticsources like AGNs may lie spatially close to the targets,so that a careful study of their environment on the skywill be warranted before drawing any claim for detec-tion of non-thermal emission from stellar runaway bowshocks.
Facility:
Chandra X-ray Observatory , EVLASupport for this work was provided by the NationalAeronautics and Space Administration through
Chan-dra
Award Number GO7-18131A issued by the Chan-dra X-ray Center, which is operated by the Smithso-nian Astrophysical Observatory for and on behalf ofthe National Aeronautics Space Administration undercontract NAS8-03060. This work has been financiallysupported by the Programme National Hautes Energies(PNHE). SG acknowledges support from the Observa-toire de Paris under the program “Action F´ed´eratriceCTA”. We thank Pierre Gratier for providing an appro-priate version of the CO map to prepare our Figure 3.REFERENCES
Abdo, A. A., Ackermann, M., Ajello, M., et al. 2010, ApJ,712, 459Acciari, V. A., Aliu, E., Arlen, T., et al. 2009, ApJL, 698,L133Anglada, G., Rodr´ıguez, L. F., & Carrasco-Gonz´alez, C.2018, A&A Rv, 26, 3Arnaud, K. A. 1996, Astronomical Data Analysis Softwareand Systems V, 101, 17 HI4PI Collaboration, Ben Bekhti, N., Fler, L., et al. 2016,A&A, 594, A116.Bell, A. R., Schure, K. M., Reville, B., & Giacinti, G. 2013,MNRAS, 431, 415Benaglia, P., Romero, G. E., Mart´ı, J., Peri, C. S., &Araudo, A. T. 2010, A&A, 517, L10Bieging, J.H., Abbott, D.C., & Churchwell, E.B. 1989, ApJ,340, 518Binder, B. A., Behr, P., & Povich, M. S. 2019, AJ, 157, 176 Rangelov et al.
Boiss´e, P., Le Petit, F., Rollinde, E., et al. 2005, A&A, 429,509Cardillo, M., Ursi, A., Tavani, M., et al. 2017, GRBCoordinates Network, Circular Service, No. 21689,