HII Regions Around Runaway Hot Stars Jonathan Mackey Argelander-Institute for Astronomie (Aifa), Bonn

HII regions around runaway hot stars Jonathan Mackey Argelander-Institute for Astronomie (AIfA), Bonn.

StarBench Workshop, Exeter, 8-11 April 2013 Collaborators: Norbert Langer (AIfA, Bonn, Germany), Vasilii Gvaramadze (Sternberg Astronomical Institute, Lomonosov Mocow State University), Dominique Meyer (AIfA, Bonn, Germany). Outline

• Motivation for our work - understanding massive stars and their feedback effects. • Introduction to runaway stars and bow shocks. • Bow shock and H II region of Zeta Ophiuchi. • Simulations of H II regions around runaway stars. • H II regions around stars with varying velocity. Massive Star Evolution • Massive star evolution is not very well understood (e.g. Langer 2012). • Metallicity, binarity, rotation, and magnetic ﬁelds are all important. • Some mass-loss rates are uncertain at order-of-magnitude levels. SN 1987a in the LMC • Assigning masses to nearby evolved massive stars is very difﬁcult. • CSM studies can help to constrain mass-loss rates, and hence post-MS evolution. Sher 25 in NGC 3603 Massive Star Feedback • Massive stars drive galaxy evolution. • Ionising radiation, stellar winds, supernova feedback, enrichment. • Feedback over lifetime of stars only modelled for idealised situations. • Supernova/GRB interaction with CSM also not explored in detail. • BUT, a star cluster in a molecular cloud is a very complex environment. • A runaway star in diffuse ISM is a much simpler system to model... may provide cleaner constraints. NGC 602 in the SMC Runaway Stars / Exiles •About 15-25% of O stars are isolated (Gies,1987). •Most of these are classical runaways (V>30km/s) or lower velocity exiles (e.g. Gvaramadze+,2012). •Ejected from star clusters, or slung from binaries when the primary explodes. •Usually isolated, and in the diffuse ISM. •CSM reaches a “steady” state in a “short” time. •CSM shaped mostly by stellar wind and RCW 158, from Gvaramadze & ionising photons. Bomans (2008,A&A,490,1071)

Gies, 1987, ApJS, 64, 545. Gvaramadze, Weidener, et al.,2012, MNRAS, 424, 3037. What Can We Learn?

•If star is not evolving rapidly, bow shock size and shape constrain wind and ISM properties. •H II region also constrains ISM properties. ! •If star is evolving, bow shock may not be in steady state, and CSM is more complicated. •Combines historical variations in stellar wind (and ionising photon luminosity) with hydrodynamical evolution. MS Runaway Star: Zeta Oph Gvaramadze, Langer & Mackey (2012,MNRAS,427,L50) Zeta Ophiuchi

• O 9.5Vnn runaway star (Morgan, Code & Whitford 1955; Lesh 1968). • Distance d≈112 pc (van Leeuwen 2007). • Space Velocity in ISM v*=26.5 km/s (for this distance). • rapidly rotating (v sin i =400 km/s, Howarth & Smith, 2001) and He enriched, so secondary of a binary system with supernova? • Widely varying estimates of mass-loss rate, Mdot. −8 −1 • Mdot(obs) ≈ 2.2×10 M⊙yr (Gull & Soﬁa, 1979), −9 −1 Mdot(UV) ≈ 1.58×10 M⊙yr (Marcolino+,2009), −7 −1 Mdot(Hα) ≈ 1.43×10 M⊙yr (Mokiem+,2005), −7 −1 Mdot(theory) ≈ 1.29×10 M⊙yr (Vink+,2000), −8 −1 Mdot(theory) ≈ 1.3×10 M⊙yr (Lucy,2010). Example: Zeta Oph

• Main Sequence O9.5V runaway star. • H II region has R=10pc or 5o (top left Hα image). • Central region is bright in IRAS far-IR data (top right) • Zoom in shows bow shock with R~0.16pc (Spitzer 24um image)

Gvaramadze, Langer, Mackey (2012,MNRAS,427,L50); NASA/JPL-Caltech. Hα image from Southern Hemispheric Hα Sky Survey Atlas (SHASSA; Gaustad et al. 2001) Infrared Astronomical Satellite (IRAS) 60 μm image of the same ﬁeld. Infrared Astronomical Satellite (IRAS) 60 μm image of the same ﬁeld. Spitzer Space Telescope 24 μm image (PI: A. Noriega-Crespo) of the bow shock.

ζ Oph and the weak-wind problem 3

−7 −1 47 −1 1.29 10 M⊙ yr (for stellar parameters derived by Mar- 3.63 10ζ Ophs ,and andR theSt =9 weak-wind.6pc,onehasfromequation(3) problem 3 × × 26 −2 colino et al. 2009), which is comparable to M˙ Hα 1.43 that M˙ obsv∞ 2.1 10 gcms .Thankstotheexcellent −7 −1−7 −1 ≈ × 47 −1 ≈ × ˙ 10 M1.29⊙ yr10 ,inferredbyMokiemetal.(2005)fromsynthe-M⊙ yr (for stellar parameters derived by Mar- 3.63consensus10 s on,andv∞RStof=9ζ Oph.6pc,onehasfromequation(3) (see Section 2), one finds M of × × 26 −2 −8 −1 sizingcolino the H etα al.line 2009), using which the FASTWIND is comparable code. to M˙ H Aα value1.43 two thatthisM˙ obs starv∞ separately,2.1 10 Mgcms˙ obs =2.Thankstotheexcellent.2 10 M⊙ yr .Thisis −7 −1 −9 ≈ −1 × ≈ × × ˙ orders10 of magnitudeM⊙ yr ,inferredbyMokiemetal.(2005)fromsynthe- lower, M˙ UV 1.58 10 M⊙ yr ,was consensus14 times on v∞ largerof ζ thanOph (seeM˙ UV Section,7timeslowerthan 2), one finds M Mof˙ Hα,and ≈ × ≈ ˙ −8 −1 inferredsizing by the Marcolino Hα line using et al. the (2009) FASTWIND from CMFGEN code. A value model two thiscomparable star separately, to M˙MLucyobs.Thediscrepancybetween=2.2 10 M⊙ yr .ThisisM˙ UV and ˙ −9 −1 ˙ × ˙ orders of magnitude lower, MivUV 1.58 10 M⊙ yr ,was 14˙ times larger than MUV,7timeslowerthanMHα,and fits to the UV doublet of C λλ≈1548,× 1551, while Lucy ≈Mobs would be˙ even more severe if the wind˙ is clumpy. For inferred by Marcolino˙ et al. (2009)−8 from CMFGEN−1 model comparable to˙ MLucy.ThediscrepancybetweenMUV and ζ Oph and the weak-wind problem 3 (2010) predicted MLucy 1.30 iv10 M⊙ yr by using the ˙instance, MUV should be reduced by a factor of three ( √f) fits to the UV doublet≈ of C× λλ1548, 1551, while Lucy Mobs would be even more severe if the wind is clumpy. For ∼ updated moving reversing˙ layer theory−8 of Lucy−1 & Solomon if the volume˙ filling factor, f,inthewindis 0.1(Marcolino (2010) predicted MLucy 1.30 10 M⊙ yr −7by using the−1 instance, MUV should be reduced by a factor of three∼ ( √f)47 −1 (1970). ≈ × 1.29 10 M⊙ yr (for stellaret al. parameters 2009), so that derivedM˙ obs by40 Mar-M˙ UV. 3.63∼ 10 s ,andRSt =9.6pc,onehasfromequation(3) updated moving reversing layer theory of× Lucy & Solomon if the volume filling factor, f,inthewindis 0.1(Marcolino× 26 −2 ˙ ≈ ∼ ˙ −3 The(1970). fast rotation of ζ Oph makescolino its spectral et al. 2009), classifi- whichet is al. comparable 2009), so that toM˙ obsMHα40M˙ UV1.43. that Mobsv∞ 2.1 10 gcms .Thankstotheexcellent −7 −1 Using equations (2) and (3), one finds n 3.6cm , ≈ ≈ × ≈−3 ≈ × ˙ cation somewhatThe fast ambiguous. rotation of ζ AlthoughOph makes10 the itsM widely spectral⊙ yr accepted,inferredbyMokiemetal.(2005)fromsynthe- classifi- whichUsing agrees equations well (2) with and (3), estimates one finds ofn theconsensus electron3.6cm , number on v∞ of ζ Oph (see Section 2), one finds M of cation somewhat ambiguous. Although the widely accepted −3 ≈ ˙ −8 −1 spectral type is O9.5 V, some other classificationssizing the Hα wereline sug- using thewhichdensity FASTWIND agrees within well with Sh code. 2-27 estimates A of value3 of. two0cm the electron(basedthis number star on radio separately, ob- Mobs =2.2 10 M⊙ yr .Thisis spectral type is O9.5 V, some other classifications were sug- density˙ within Sh 2-27−9 of 3.0cm≈−1 −3 (based on radio ob- ˙ × ˙ gested as well. For instance, Conti &orders Leep (1974) of magnitude classified lower, Mservations,UV 1.58 Hα 10surfaceM≈⊙ brightness,yr ,was and studies14 of times interstellar larger than MUV,7timeslowerthanMHα,and this stargested as O9 as well. V(e), For while instance, Herrero Conti et al. & Leep (1992) (1974) prefer classified O9 III. servations,≈ Hα surface× brightness, and studies of≈ interstellar ˙ ˙ this star as O9 V(e), while Herrero etinferred al. (1992) by prefer Marcolino O9 III. et al.absorption (2009) from lines; CMFGEN Gull & Sofia model 1979 and referencescomparable therein) to MLucy.ThediscrepancybetweenMUV and Like Marcolino et al. (2009), we adopt the spectral type of absorption lines; Gull & Sofia 1979 and references therein) Like Marcolino et al. (2009), we adoptfits the to spectral the UV type doublet of ofand C iv theλλ average1548, 1551, line of while sight Lucy electron andM˙ obs neutralwould hydro- be even more severe if the wind is clumpy. For 47 −1 and the average line of sight electron and neutral− hydro- O9.5 V, so that S(0) = 3.63 10 47s −1(Martins, Schaerer −8 −1 −3 3 O9.5 V, so that S(0) = 3.63 10 (2010)s (Martins, predicted SchaererM˙ Lucy gen1gen.30 densities densities10 towardsM towards⊙ yrζ Ophbyζ Oph of using4 of.0cm the4.0cm(Howkinstance, &(Howk Savage &M˙ SavageUV should be reduced by a factor of three ( √f) × × ≈ × ≈ ≈ ∼ &Hillier2005b).Thebasicparametersof&Hillier2005b).Thebasicparametersofupdatedζζ OphOph moving are are sum- sum- reversing1999;1999; layer scaled scaled theory to d to=112 ofd Lucy=112 pc). pc). & Solomon if the volume filling factor, f,inthewindis 0.1(Marcolino marizedmarized in Table in Table 2. To 2. To this this table table we we added added our our estimate of of ˙ ˙ ˙ ˙ ∼ ˙ (1970). NoteNote that thatMobsMwasobs derivedwas derived for d =112pc,whileMar- for d =112pc,whileMar-et al. 2009), so that Mobs 40MUV. the mass-lossthe mass-loss rate, rate,M˙ M,basedontheobservedparametersobs,basedontheobservedparameters obs colinocolino et et al. al. (2009) (2009) adopted adopted a somewhat a somewhat larger distance larger distance ≈ −3 of the bow shock and theii H ii region associatedThe fast with rotationζ Oph of ζ Oph makes its spectral classifi- Using equations (2) and (3), one finds n 3.6cm , of the bow shock and the H region associated with ζ Oph ofof 146 146 pc. pc. An An even even larger larger value, value,d =222pc,wassug-d =222pc,wassug- ≈ (see next Section). cation somewhat ambiguous. Although the widely accepted which agrees well with estimates of the electron number (see next Section). gestedgested by by Megier Megier et al. et (2009), al. (2009), whose whose estimate estimate is based is on based on spectral type is O9.5 V, somethe empirical other classifications relationship between were thesug- strengthdensity of the in- within Sh 2-27 of 3.0cm−3 (based on radio ob- the empiricalii relationship between the strength of the in- ≈ gested as well. For instance,terstellar Conti Ca & Leeplines and (1974) the distances classified to early-typeservations, stars. Hα surface brightness, and studies of interstellar terstellar Ca ii lines and the distances0.6 to early-type stars. 3MASS-LOSSRATEOFζ OPH Since v∗ scales with the distance as d (v∗ 31.2and this star as O9 V(e), while Herrero−1 et al. (1992) prefer O9 III. 0≈.6absorption lines; Gull & Sofia 1979 and references therein) 3MASS-LOSSRATEOFζ OPH 41Since.7kmsv∗ scalesfor d =146and222pc,respectively),one with the distance as d (v∗ 31.2and The structure of an H ii region createdLike Marcolino by a moving et star al. (2009), we adopt−˙ 1 the spectral1.7 type of and the≈ average line of sight electron and neutral hydro- finds41.7kms that 47Mobs−1for d =146and222pc,respectively),one,sothatthelargerthedistance depends on the stellarii velocity relativeO9.5 V, to the so that local ISM.S(0) = 3.63 10 s ∝(Martins,1.7 Schaerer˙ ˙ −3 The structure of an H region created by a moving star thefinds larger that the discrepancyM˙ obs d between,sothatthelargerthedistanceMobs and MUVgen, namely densities towards ζ Oph of 4.0cm (Howk & Savage For supersonically moving stars the number density within ˙ × ˙ ∝ ≈ depends on the stellar velocity relative&Hillier2005b).Thebasicparametersof to the local ISM. Mtheobs larger22 45 theMUV discrepancyforζ Ophd in the are between range sum- fromM˙ 1461999;and to 222M˙ scaledUV pc., namely to d=112 pc). the H ii region is comparable to that of the ambient ISM, ≈ − ˙ ˙obs For supersonically moving stars themarized number density in Table within 2. To thisCorrespondingly,˙ table we addedMH˙α ourremains estimate larger than of Mobs by a factor while the radius of the H ii region is of the order of the Mobs 22˙ 45MUV for d in the range from 146Note to that 222 pc.M˙ obs was derived for d =112pc,whileMar- the H ii region is comparable to that of the ambient ISM,˙ of 2to4.≈ MVink− also depends˙ on the adopted distance˙ (via Strömgren radius (e.g. Tenorio Tagle,the Yorke mass-loss & Bodenheimer rate, Mobs,basedontheobservedparametersCorrespondingly,≈ MHα remains larger than Mobs by a factor ii theii stellar luminosity; Vink et al. 2000); this wouldcolino decrea etse al. (2009) adopted a somewhat larger distance while1979), the radius which is of given the by H (e.g.region Lequeuxof is the of 2005) the bow order shock of and the the Hof region2to4.˙ associatedM˙ Vink also with dependsζ Oph on the adopted distance (via (increase)≈ MVink by a factor of a few for d =112(222)pc.of 146 pc. An even larger value, d =222pc,wassug- Strömgren radius (e.g. Tenorio Tagle,(see Yorke next & Bodenheimer Section). the stellar luminosity; Vink et al. 2000); this would˙ decrease 1/3 Similarly, it follows from equation (3) thatgestedMobs by Megier et al. (2009), whose estimate is based on 1979), which is3S given(0) by (e.g. Lequeux 2005) 1/2 ˙ ∝ RSt = , (1) S(0)(increase),sothatMVinkM˙ by21 aM factor˙ UV (or of a0 few.2M˙ H for)ifd =112(222)pc.ζ Oph is 2 obs α the empirical relationship between the strength of the in- ! 4παBn " ≈ ≈ ˙ ˙ ˙ 1/3 an O9 VSimilarly, star (Conti it & follows Leep 1974), from and equationMobs (3)37M thatUV ( Mobsii 3S(0) ˙ 1/2 ≈ terstellar≈ Ca ∝lines and the distances to early-type stars. where αB is the recombination coefficient of hydrogen to all 0.4MHα)ifitsspectraltypeisO9III(Herreroetal.1992).˙ ˙ ˙ RSt = 2 , (1) S(0) ,sothatMobs 21MUV (or 0.2MHα)ifζ Oph is 0.6 ! 4παBn " −133MASS-LOSSRATEOF3 −1 Here we usedOPH the calibrations≈ of Martins et≈ al. (2005b).Since v∗ scales with the distance as d (v∗ 31.2and but the ground state ( 2.6 10 cm s for fully ionized ζ ˙ ˙ UV ≈ 4 × an O9 V star (Conti & Leep 1974), and Mobs 37M−1 ( ≈ gas of temperature of 10 K). On˙ the other hand, the possible leakage of ionizing41.7kms≈ pho- for≈d =146and222pc,respectively),one where αB is the recombination coefficient of hydrogen to all 0.4MHα)ifitsspectraltypeisO9III(Herreroetal.1992). The stellar wind of a moving starThe interacts structure with the of ISM an H ii region createdii by a moving star ˙ 1.7 −13 3 −1 tonsHere from we the used H theregion calibrations (caused by of porosity Martins of thefinds et ISM; al. (2005b). that see Mobs d ,sothatthelargerthedistance but theand ground produces state a bow ( shock2.6 ahead10 ofcm thes star.for The fully minimum ionized ∝ ≈ 4 × depends on the stellar velocityWood et relativeal. 2005 and to Section the local 4) would ISM. reduce ourthe estimate larger the discrepancy between M˙ obs and M˙ UV, namely gas of temperature of 10 K). ˙ On the other hand, the possible leakage of ionizing pho- distance from the star at which the windFor pressure supersonically is balanced movingof starsMobs.Conservativelyassumingthatahalfofthephotons the number density within ˙ ˙ The stellar wind of a moving star interacts with the ISM ii ˙ Mobs 22 45MUV for d in the range from 146 to 222 pc. by the ram and the thermal pressuresthe of H theii ISM,region the stand- is comparableescapetons to from from that Sh the 2-27, of H the oneregion ambient finds (caused that ISM,Mobs byshould porosity be reduced of≈ the ISM;− see ˙ ˙ and producesoff distance, a bow is given shock by (e.g.ahead Baranov of the et star. al. 1971) The minimum byWood a factor et of al. 20051.4. and Section 4) would reduceCorrespondingly, our estimate MHα remains larger than Mobs by a factor while the radius of the H ii region≈ is of the order of the ˙ distance from the star at which the wind pressure is balanced ˙ of 2to4.MVink also depends on the adopted distance (via Constraining˙ Mass-loss1/2 rate of NoteMobs also.Conservativelyassumingthatahalfofthephotons that the possible existence of large-scale≈ inter- Mv∞ Strömgren radius (e.g. Tenorio Tagle, Yorkeii & Bodenheimer ˙ the stellar luminosity; Vink et al. 2000); this would decrease by theR0 ram= and the thermal2 pressures, of the ISM, the stand-(2) nalescape flows in from the H Sh 2-27,region one (e.g. finds caused that by photoevaporationMobs should be reduced Bow #shock4πn(µm radiusHv∗ +2kT) $ HII region radius off distance, is given by (e.g. Baranov1979), et al. which 1971) is given by (e.g.ofby molecular a Lequeux factor clumps of 2005)1 in.4. the ISM; see Section 4) also(increase) can af- M˙ Vink by a factor of a few for d =112(222)pc. fect the estimate of≈M˙ obs because R0 depends on the stellar where µ =1.4isthemeanmolecularweight,mH is the1/3 ˙ ˙ 1/2 velocityNote relative also to that the the local possible ISM. Assuming existence that of large-scaletheSimilarly, char- inter- it follows from equation (3) that Mobs mass of a hydrogenMv∞ atom, k is the Boltzmann constant,3S(0) and 1/2 ∝ 4 RSt = , nal flows in the H ii region (e.g. caused(1) by photoevaporation ˙ ˙ ˙ R0 =T ( 10 K) is2 the temperature, of the (ionized) ISM. 2 (2) acteristic velocity of the photoevaporation flowsS is(0) of orde,sothatr Mobs 21MUV (or 0.2MHα)ifζ Oph is # 4πn(µmHv∗ +2kT) $ ! 4παBn " −1 ≈ ≈ ≈ ofof the molecular sound speed clumps in the in ionized the ISM; gas, i.e. see Section10 kman s O9 4),one also V star can (Conti af- & Leep 1974), and M˙ obs 37M˙ UV ( Eliminating n between equations (1) and (2), one has ∼ findsfect that theM estimate˙ obs might of eitherM˙ bebecause larger orR smaller0 depends by a˙ factor on the stellar ≈ ≈ wheretheµ ‘observed’=1.4isthemeanmolecularweight, stellar wind momentumwhere rateαB ism theH is recombination the coefficient of hydrogenobs to all 0.4MHα)ifitsspectraltypeisO9III(Herreroetal.1992). of up to 2iftheflowvelocityisantiparallelorparallelto mass of a hydrogen atom, k is the Boltzmann constant,2 and velocity−13 relative3 −1 to the local ISM. AssumingHere that we the used char- the calibrations of Martins et al. (2005b). 25 −2 but1 the groundR0 state ( 2.6 10 ≈ cm s for fully ionized ˙ 4 ∞ the vector of the space velocity of ζ Oph. T ( 10MobsK)v is=1 the.57 temperature10 gcms of the1+ (ionized)2 ISM. ≈ 4 acteristic× velocity of the photoevaporation flows is of order × gasM of! temperature0.1pc" of 10 K). On the−1 other hand, the possible leakage of ionizing pho- ≈ % & ofFinally, the sound we note speed that in the the fast ionized rotation gas, of i.e.ζ Oph10 might km s ,one Eliminating n between equations1 (1)2 and (2),−3 2 one has 2 / The stellar/ wind of a movingmake the star stellar interacts wind anisotropic with the by ISM increasingtons∼M˙ in from the the H ii region (caused by porosity of the ISM; see v∗ S(0) RSt finds that M˙ obs might either be larger or smaller by a factor the ‘observed’ stellar wind momentum rate , (3) equatorial zone (Friend & Abbott 1986). This in turn might × 10 km s−1 ! 1048 s−1 "and! produces10 pc " a bow shockof ahead up to of the2iftheflowvelocityisantiparallelorparallelto star. The minimum Wood et al. 2005 and Section 4) would reduce our estimate % & 2 affect the geometry≈ of the bow shock and the UV line and˙ Hα 25 −2 distance1 fromR0 the star at whichthe vector the wind of the pressure space is velocity balanced of ζ Oph.of Mobs.Conservativelyassumingthatahalfofthephotons M˙ v∞ =1.57 ∗10 gcms 1+ obsZetawhere Oph has:M =R0=0.16v /cs pc,is v the*=26.5 isothermal km/s, RSt = Mach 9.62 pc number (observed), and cs = diagnostic. The Hα diagnostic would also be unreliable if the • ×1/2 byM the! ram0.1pc and" the−1 thermal pressures of the ISM, the stand- escape from Sh 2-27, one finds that M˙ obs should be reduced and(2 S(0)kT/µm = 3.6xH )1047 iss-1 thefrom isothermal spectral% type, sound M=2.5 speed& for T=10 ( 104K..9kms Oe statusFinally, of ζ Oph we is note due that to the the presence fast rotation of a circumstellar of ζ Oph might 4 1o2ff distance,≈ −−31 is2 given by (e.g. Baranov et al. 1971) for T =10 K). For2 R0 =0.16 pc, v/∗ =26.5kms /, S(0) = diskmake (see, the however, stellar Vink wind et al. anisotropic 2009). The by anisotropy increasingby a eff factorectsM˙ in of the1.4. v∗ S(0) −8 RSt−1 For v∞ =1500 km/s, Mdot = 2.2 × 10 M⊙ yr . ≈ • , (3) equatorial1/2 zone (Friend & Abbott 1986). This in turn might × 10 km s−1 ! 1048 s−1 " ! 10 pc " ˙ Note also that the possible existence of large-scale inter- ⃝c 0000% RAS, MNRAS& 000,000–000 Mv∞ affect the geometry of the bow shock and the UV line and Hα ii R0 = 2 , (2) nal flows in the H region (e.g. caused by photoevaporation 4πn(µmHv∗ +2kT) where M = v∗/cs is the isothermal Mach number# and cs = diagnostic.$ The Hα diagnostic would also beof unreliable molecular if clumps the in the ISM; see Section 4) also can af- 1/2 −1 (2kT/µmH) is the isothermal sound speed ( 10.9kms Oe status of ζ Oph is due to the presence of a circumstellar ˙ 0 4 where µ≈=1−1.4isthemeanmolecularweight,mH is the fect the estimate of Mobs because R depends on the stellar for T =10 K). For R0 =0.16 pc, v∗ =26.5kms , S(0) = disk (see, however, Vink et al. 2009). The anisotropy effects mass of a hydrogen atom, k is the Boltzmann constant, and velocity relative to the local ISM. Assuming that the char- T ( 104 K) is the temperature of the (ionized) ISM. acteristic velocity of the photoevaporation flows is of order ⃝c 0000 RAS, MNRAS 000,000–000 ≈ −1 Eliminating n between equations (1) and (2), one has of the sound speed in the ionized gas, i.e. 10 km s ,one ˙ ∼ the ‘observed’ stellar wind momentum rate finds that Mobs might either be larger or smaller by a factor 2 of up to 2iftheflowvelocityisantiparallelorparallelto 25 −2 1 R0 ≈ ˙ ∞ the vector of the space velocity of ζ Oph. Mobsv =1.57 10 gcms 1+ 2 × % M & ! 0.1pc" Finally, we note that the fast rotation of ζ Oph might 1 2 −3 2 2 / / make the stellar wind anisotropic by increasing M˙ in the v∗ S(0) RSt , (3) equatorial zone (Friend & Abbott 1986). This in turn might × 10 km s−1 ! 1048 s−1 " ! 10 pc " % & affect the geometry of the bow shock and the UV line and Hα where M = v∗/cs is the isothermal Mach number and cs = diagnostic. The Hα diagnostic would also be unreliable if the 1/2 −1 (2kT/µmH) is the isothermal sound speed ( 10.9kms Oe status of ζ Oph is due to the presence of a circumstellar 4 ≈ −1 for T =10 K). For R0 =0.16 pc, v∗ =26.5kms , S(0) = disk (see, however, Vink et al. 2009). The anisotropy effects

⃝c 0000 RAS, MNRAS 000,000–000 Conclusions - Part 1

• Possible to constrain mass-loss rate Mdot if: • the massive star is a runaway, • is hot (emits ionising photons), • has a well-defined bow shock and HII region, • and has a well-defined spectral type. −8 −1 • For Zeta Oph, we obtain Mdot = 2.2 × 10 M⊙ yr , that is: • 6-7x lower than predicted by Vink+(2000), • 7x lower than from Hα (Mokiem+,2005), and • 14x larger than from UV lines (Marcolino+,2009), but • agreeing well with predictions from Lucy (2010) and observational estimate from Gull & Sofia (1979). • Uncertainties from observational data change our result by < 2x. H II regions around runaway hot stars Numerical simulations of a hot star moving through an idealised environment. Simulations • 3D Cartesian simulations using the pion radiation-MHD code (Mackey+,2010,2011). • Star at origin; uniform ISM flowing past star with: (i) no B-field, or (ii) uniform B-field with B.v=1 (parallel), or (iii) B.v=0 (perpendicular). -3 • B-field strength 7uG. ISM density nH=2.5 cm . • Neutral gas heating and cooling based on Wolfire+(2003). • Atomic H, He cooling calculated explicitly based on ion fraction (Frank & Mellema,1994, Hummer 1994). • Photoionised gas has cooling from C,O, (Henney+2009). 3D radiation-hydro/MHD simulations • Cartesian dynamics in a 3D box. • Uniform ISM advects past star (at origin) at 26.5 km/s. • O 9.5V star in ISM: n=2.75 cm-3. • Ionising and FUV radiation included, but no stellar wind. • Models run with hydrodynamics, and aligned and perpendicular B-fields. • Calculated for solar metallicity (Z⊙). • Hot HII region expands, creating a shell at the sides (Raga+,1997). Simulations relax to a stationary Example simulation: colour scale show Hα • intensity, red contours show neutral H state, but with instability in the underdensity, and blue contours neutral H ionisation front. overdensity (units of 5e19 per contour). Hydrodynamics, 26.5 km/s, O9.5V star. Perpendicular B-field, 26.5 km/s, O9.5V star. Perpendicular B-field, Higher resolution.

Projection perpendicular to ﬁeld lines. Perpendicular B-ﬁeld, Higher resolution.

Projection parallel to ﬁeld lines. Zeta Oph HII region 3D simulations (preliminary)

• We calculate kinetic energy (K.E.) and momentum of the ISM over time (with bulk flow subtracted). • Efficiency of photon-to-K.E. conversion is low at <1% for solar metallicity. • Even still, K.E. of HII region shocks is larger than wind K.E. input even at solar metallicity. • Momentum feedback rate is 7x larger than total stellar radiation momentum, and >200x larger than wind. Observations/Conclusions •Dynamics of HII regions around moving stars is complicated. •Runaway stars are ideal laboratories for observing/modelling ionisation fronts. •There is an expanding conical shell of overdense gas behind star, could be observable in HI data. •Region directly behind star is underdense, moreso at lower metallicity (will be exacerbated by shocked wind). •Ionisation front seems unstable (cf. Garcia-Segura+,1996, Whalen & Norman,2008), forming dense knots of neutral gas. •Kinetic energy is inefficiently generated from ionisation heating (~1% or less), but can still be at a greater rate than wind mechanical feedback. •Observations of Zeta Oph unlikely to be explained with uniform ISM.