Mon.Not.R.Astron.Soc.000,000–000(0000) Printed25January2012 (MNLATEXstylefilev2.2) i The structure of H in galactic disks: Simulations vs observations David M. Acreman1(cid:63), Clare L. Dobbs1,2,3, Christopher M. Brunt1, Kevin A. Douglas4 2 1 School of Physics, University of Exeter, Stocker Road, Exeter EX4 4QL. 1 2 Max-Planck-Institut fu¨r extraterrestrische Physik, Giessenbachstraße, D-85748 Garching, Germany 0 3 Universitats-Sternwarte Mu¨nchen, Scheinerstraße 1, D-81679 Mu¨nchen, Germany 2 4 Arecibo Observatory/NAIC, HC 3 Box 53995, Arecibo, PR 00612 USA n a J 25January2012 4 2 ABSTRACT ] A We generate synthetic Hi Galactic plane surveys from spiral galaxy simulations G which include stellar feedback processes. Compared to a model without feedback we find an increased scale height of Hi emission (in better agreement with observations) . h and more realistic spatial structure (including supernova blown bubbles). The syn- p theticdatashowHiself-absorptionwithamorphologysimilartothatseeninobserva- o- tions. The density and temperature of the material responsible for Hi self-absorption r is consistent with observationally determined values, and is found to be only weakly t s dependent on absorption strength and star formation efficiency. a [ Key words: methods: numerical – surveys – ISM: atoms – ISM: structure 1 v 0 7 1 INTRODUCTION realistic observables, such synthetic observations also allow 9 features to be related more directly to underlying physical 4 Hi emission, from the 21-cm line of atomic hydrogen, is processesthanispossiblewithrealobservations.Hencesyn- . widely used as a tracer of large scale Galactic structure 1 theticobservationscanbeusedtotestphysicalmechanisms (Kalberla&Kerp2009)butalsorevealssmallerscalestruc- 0 proposed to explain features seen in observations. ture in the interstellar medium. This smaller scale struc- 2 1 ture includes shells formed by stellar feedback processes, The transition of hydrogen gas from the atomic phase : which inject kinetic and thermal energy, along with heavy tothemolecularphaseisakeysteptomolecularcloud,and v elements. Such feedback is instrumental in determining the ultimatelystarformation.Inorderforthistransitiontooc- i X evolution of a galaxy and its effects can be seen in our own curthegasmustbecomecooleranddenserthantypicalISM Galaxy (Heiles 1979, 1984; Taylor et al. 2003) and also in material.Thepassageofmaterialthroughtheshocksassoci- r a nearbyexternalgalaxies(Fukuietal.2009;Bagetakosetal. ated with the density waves in a grand design spiral galaxy 2011). Hence feedback processes are seen to play an im- resultsinsubstantialcompression(Roberts1972),whichal- portant role in the formation of smaller scale structure and lows the gas to cool efficiently down to low temperatures itsappearanceinHiemission.Numericalsimulationsofthe (Cowie1981;Berginetal.2004;Dobbsetal.2008;Kimetal. interstellarmedium(ISM)indiskgalaxiesarebecomingin- 2008). The high density, low temperature material in a spi- creasingly sophisticated and can now include the effects of ralshockprovidesfavourableconditionsfortheformationof stellar feedback processes (Wada 2008; Shetty & Ostriker molecular clouds (although molecular clouds can also form 2008;Dobbsetal.2011b;Tasker&Tan2011).Atthesame in the absence of spiral shocks). Whilst CO traces molec- time advances are being made in the generation of syn- ular clouds, cold Hi with a low molecular fraction is more theticobservationsfromsimulations(Hennebelleetal.2007; difficult to detect. It can however be detected as Hi self- Douglas et al. 2010; Parkin 2011, and references therein). absorption (HISA). HISA is observed when colder, denser These synthetic observables allow a direct comparison be- foreground material is located along the same line of sight, tween simulation results and real observations, and as the andatthesamelineofsightvelocity,aswarmerbackground modelsbecomeincreasinglysophisticatedwecanexpectthe material (Li & Goldsmith 2003; Gibson et al. 2005). Given syntheticobservationstocomparemorefavourablywithreal theserequirements,usingHISAisnotasuitablemethodto observations.Aswellastestingwhethermodelscanproduce obtain a complete map of cold Hi. However HISA can pro- videinformationincaseswhereothertracersofstarforming clouds (such as the J=1–0 rotational transition of the CO (cid:63) [email protected] molecule)arenoteffective(Douglas&Taylor2007),forex- (cid:13)c 0000RAS 2 David M. Acreman et al ample shortly after passing through a shock (Bergin et al. ergy, where gas is assumed to have formed stars, although 2004),orinphotondominatedregionswherethereislessef- sinkorstarparticlesarenotintroduced.Forfeedbacktooc- fectiveself-shieldingofCOagainsttheinterstellarradiation curthedensityofagasparticleinaconvergingflowmustex- fieldcomparedtoH (Kaufmanetal.1999;Allenetal.2004; ceed1000cm−3,whilstthesurroundingregionofgas(within 2 Shetty et al. 2011). Thus although HISA is found to corre- a radius of ∼ 20 pc) must be gravitationally bound and in late with emission from molecular species (Goldsmith & Li an energetically favourable state (Dobbs et al. 2011a). The 2005;Kavarsetal.2005)thecorrespondencebetweenHISA number of supernovae is estimated from the mass of gas and tracers of star forming clouds is not straightforward. in the computed region, adopting a certain star formation Syntheticobservationsprovideavaluablewayofexamining efficiency and a Salpeter initial mass function (IMF). For tracers such as HISA in a more controlled environment. each feedback event, the total amount of energy deposited In this paper we present synthetic Hi Galactic plane is given by equation 1 from Dobbs et al. (2011a), which surveys derived from simulations of galaxies which include computes the number of massive stars expected to form, stellar feedback and compare the Hi emission and HISA and assumes each supernova injects 1051 erg of energy. The structure with similar features found in observations. We numberofstarsformed(andthereforetheamountofenergy begin in Section 2 by outlining the method used to make deposited)dependsonastarformationefficiencyparameter, thesyntheticHisurveys.InSection3weassesstheeffectof (cid:15),whichrepresentsthefractionofthemolecularmassineach feedbackontheobservablepropertiesoftheHigasbycom- boundregionassumedtobeturnedintostars.Thestarfor- paring our synthetic data with observational data and with mation efficiency is an absolute value and does not depend aprevioussyntheticsurveywhichdidnotincludefeedback. onthefree-falltime.Theenergyisdistributedaccordingto The distribution of HISA in our synthetic observations is a snowplough solution, which describes the pressure-driven discussed in Section 4.1 and the properties of the material phase of a supernova after the blast wave (Woltjer 1972; producing HISA are investigated in Section 4.2. We finish Chevalier 1974; McKee & Ostriker 1977) and is deposited by presenting our conclusions in Section 5. as2/3kineticenergy,and1/3thermalenergy.Thefeedback is assumed to be instantaneous, and although we consider supernova explosions, the energy could account for numer- ous feedback processes, such as stellar winds, radiation and 2 METHOD supernovae. Our synthetic observations are generated from an SPH Model data at a simulated time of 250 Myr are used, (smoothed particle hydrodynamics) simulation of a spiral at which time the simulated galaxy has reached a state of galaxy. The variable spatial resolution of the SPH method quasi-equilibrium. In Model C approximately one third of permits a large range of density and spatial scales to be re- the ISM is in each of the cold (< 150K), unstable (150– solved. This enables a simulation with a domain covering 5000K)andwarm(>5000K)phases.Withahigherstarfor- a whole galaxy, which can still resolve individual molecular mation efficiency there is more material in the warm phase clouds. SPH is a Lagrangian method which means that the and less material in the cold phase. The thermal evolution materialassociatedwithanSPHparticleretainsitsidentity of the gas is modelled according to the thermodynamics of asthesimulationprogresses.Consequentlywecantrackma- Dobbs et al. (2008). The gas exhibits a maximum temper- terial as it passes thorough a spiral arm and determine its ature of 2×106K. A lower limit of 20K is imposed prior fate at a later time. A synthetic Galactic plane survey was to the temperature update in the model. This limit pre- previouslypresentedbyDouglasetal.(2010)(hereafterPa- ventsmaterialfrombecomingtoocoldforthesimulationto per1)usinganSPHmodelwithoutfeedbackprocesses.The treat accurately but allows some cooling below 20K to oc- newresultspresentedhereincludestellarfeedbackprocesses cur. The simulation follows the evolution of molecular gas, in the SPH model, which was proposed as an important although only a small fraction ((cid:46) 10 percent) is molecular. mechanism to resolve differences between Paper 1 results Energy is injected as soon as the criteria for feedback are and observations. met,whichresultsindensegasbeingrapidlydisrupted.In- The SPH simulations used for this work are models C cludingadelaybeforeinjectingenergyfromfeedbackwould andDfromDobbsetal.(2011b)whichincludestellarfeed- enable longer for molecular gas to form. backandself-gravityofthegas,whichwerenotincludedin The SPH data provide three dimensional distributions the simulation of Dobbs (2008) used in Paper 1. In all the ofHidensity,temperatureandvelocitywhichareusedtoset simulations the gas is assumed to orbit in a fixed potential, up an adaptive mesh refinement (AMR) grid for the torus composed of halo, disc and four armed spiral components. radiative transfer code (Harries 2000). torus carries out The resulting rotation curve is comparable to that of the non-parallel ray traces (one per image pixel and velocity MilkyWay,whichallowsstructuresinthesimulationstobe channel) to solve the radiative transfer equation and gener- comparedtoobservedstructuresinourownGalaxy.Forex- atethree-dimensionalspectraldatacubesof21cmHiemis- amplethesimulationshavespiralarmswhicharesimilarto sion in Galactic longitude-latitude-velocity co-ordinates. those in our Galaxy which allows us to generate synthetic The AMR grid is generated by splitting the grid cells imagesanalogoustoobservationsofthenearbyPerseusarm so that there is never more than one SPH particle per cell. inourGalaxy.Thetotalsurfacedensity(includingHelium) Mapping temperature, density and velocity onto the AMR is constant with values of 8 M(cid:12)pc−2 for models including grid is carried out as described in Paper 1 and Acreman feedback and 10 M(cid:12)pc−2 for the model without feedback et al. (2010) using the method of Rundle et al. (2010). The used in Paper 1. This is comparable to the surface density observer is placed within the model galaxy, at a location in the solar neighbourhood (Wolfire et al. 2003). analogous to that of the Sun within our own Galaxy, and Stellar feedback is inserted as kinetic and thermal en- the synthetic survey is generated as if it were the Galactic (cid:13)c 0000RAS,MNRAS000,000–000 The structure of Hi in galactic disks 3 second quadrant (90◦ < l < 180◦). The ray tracing uses have small scale filamentary structure, which is not seen in thedensitysub-samplingmethodofRundleetal.(2010)to thesyntheticobservations,however,wedonotexpecttore- linearlyinterpolatedensityvalueswithinacellontheAMR produce this structure at the current SPH resolution. The grid. Density sub-sampling was not required in Paper 1 as resolutionoftheSPHsimulationisgovernedbythesmooth- thespatialresolutionoftheSPHsimulationandAMRgrid inglengthsoftheparticles1.Thedensitythresholdforfeed- were finer (the computational demands of adding feedback back to occur is at a number density of 1000cm−3 which processes require that a lower spatial resolution is used). correspondstoasmoothinglengthof5.6pc(withaparticle The data cubes have a velocity resolution of 0.5 km/s and massof2500M ).Atadistanceof2.5kpc(typicalofmate- (cid:12) anangularresolutionof1arcmininlongitudeandlatitude. rialseeninFig.1)thiscorrespondstoanangularsizeof0.13 Each cell on the AMR grid is assigned a thermal line degrees(or7.8arcmin).Structureonthisscaleisseeninthe width,whichdoesnotincludeanyadditionalturbulentcom- synthetic data (e.g. at l=120 in Fig 1(c)) but is spherical, ponent to account for unresolved structure. This avoids rather than filamentary. addinganad-hocparameterbutwilltendtounder-estimate the line width if unresolved structure makes a significant 3.2 Vertical distribution of Hi emission contribution to the velocity dispersion (Hennebelle et al. 2007). The velocity dispersion of the gas and clouds in the In order to allow a quantitative comparison of the vertical simulations are discussed in Dobbs et al. (2011b). distribution of Hi emission, longitudinally averaged profiles ofbrightnesstemperatureagainstlatitudewereextractedfor the longitude range 126–144 degrees, in the same velocity 3 EFFECT OF FEEDBACK ON HI channels shown in Fig. 1. Emission seen in this longitude STRUCTURE andvelocityrangeisfromspiralarmmaterialatadistance ofapproximately2.5kpcfromtheobserver.Atthisdistance 3.1 Latitude-longitude structure material at a latitude of 1 degree is 44 pc out of the mid- Figure 1 shows latitude-longitude plots of Hi emission for plane. simulations with and without feedback, and also for the The spiral arms in the simulations have a narrow ve- CanadianGalacticPlaneSurvey(CGPS)observations(Tay- locity width, compared to the observed Perseus arm, and lor et al. 2003). The CGPS data are from a velocity chan- this results in emission from the simulated spiral arms nel which contains emission from Perseus arm material. In being distributed over fewer velocity channels but with thesyntheticobservationstheobserverhasbeenpositioned increased brightness temperatures (the structure of the so that at similar velocities we see a spiral arm analogous armsinlongitude-velocityspaceisdiscussedfurtherinSec- to the Perseus arm. Fig. 1(a) shows synthetic observations tion 3.3). Consequently the synthetic surveys have higher fromasimulationwithoutfeedbackfromPaper1(hereafter brightnesstemperaturesthantheCGPSdataandtheshape referred to as NoFeedback). Figure 1(b) is from Run C of of the raw profiles cannot easily be compared. To facilitate Dobbsetal.(2011b)andhasfeedbackincludedwith5%star aquantitativecomparisonthesyntheticprofileswerescaled formation efficiency (hereafter referred to as Feedback5), by a constant scaling factor chosen to minimise the RMS and Fig. 1(c) is from Run D of Dobbs et al. (2011b) and differencefromtheCGPSdata.Theexpressionusedtofind has feedback included with 10% star formation efficiency the best value of the scaling factor f was (hereafter referred to as Feedback10). The star formation (cid:118) efficiency referred to here is the (cid:15) parameter described in (cid:117)(cid:117)1 (cid:88)N RMS=(cid:116) (C −fS)2 (1) Section 2. N i j TheCGPSdatashowabroadregionofemissionwhich, i=1 inplaces,extendsbeyond±2degreesoutsidethemid-plane. where N is the number of points in the CGPS profile, C i IncontrasttheNoFeedbackrunhasemissionwhichismuch is the brightness temperature in bin i of the CGPS profile, more confined to the mid-plane, with bright ridges of emis- andS isthebrightnesstemperatureinbinjofthesynthetic j sion at ±1 degree which are not seen in the CGPS data. profile, where bin j and bin i correspond in latitude. Every Withoutfeedbackthegasinthemodelgalaxyisoverlycon- point in the brightness temperature profile was then scaled fined to the mid-plane, resulting in large mid-plane opti- bythevalueoff whichminimisedtheexpressioninEqn.1. cal depths. In the NoFeedback model the column density As the profiles are asymmetric the fit was repeated with is highest within ±1 degree of the mid-plane but the accu- thelatitudeaxisinverted.Thescalingfactorsandresultant mulation of cold, dense material results in excessively high RMSdifferencesfromtheCGPSprofileareshowninTable1. absorptionandcorrespondinglylowlevelsofHiemission(in The scaled profiles and the unmodified CGPS profile are theopticallythicklimitthebrightnesstemperaturewillsat- plottedinFig.2(a)andinFig.2(b)(withareversedlatitude urate at the spin temperature). When feedback is included axis in the second plot). (Fig.1(b)and1(c))thegasinthemodelgalaxyismuchless The fit for both the feedback runs is significantly bet- confinedtothemid-planeandthebrightridgesofemission, ter when the latitude axis is reversed, as the skewness of seeninFig.1(a)atabout1degreeaboveandbelowthemid- plane,arenotpresent.TheHiemissionnowextendsfurther 1 Thesmoothinglengthhisgivenby out of the mid-plane, in better agreement with the CGPS observations. (cid:18)m(cid:19)1/3 h=η FeedbackresultsinsignificantholesinHiemission,with ρ Feedback10inparticularhavingmuchlesscontiguousemis- wheremistheparticlemass,ρistheparticledensity,andη=1.2 sion than the CGPS observations. The CGPS observations (Price&Monaghan2007). (cid:13)c 0000RAS,MNRAS000,000–000 4 David M. Acreman et al 200.0 s) K) ee Velocity= -46.75 km/s e ( e (degr 321 150.0 peratur ud 0 100.0 m atit 1 s te c l 2 es Galacti 3160 150 G1a4l0actic longitude (degre1e3s0) 120 110 50.0 Brightn 0.0 (a) NoFeedback Velocity= -36.25 km/s 6 200.0 s) K) ee 4 e ( gr 150.0 ur e (de 2 perat ud 0 100.0 m atit s te c l 2 es Galacti 4 50.0 Brightn 0.0 160 150 140 130 120 110 Galactic longitude (degrees) (b) Feedback5 Velocity= -36.25 km/s 6 200.0 s) K) ee 4 e ( gr 150.0 ur e (de 2 perat ud 0 100.0 m atit s te c l 2 es Galacti 4 50.0 Brightn 0.0 160 150 140 130 120 110 Galactic longitude (degrees) (c) Feedback10 200.0 s) Velocity= -46.75 km/s K) ee e ( e (degr 42 150.0 peratur ud 100.0 m atit 0 s te c l es Galacti 2160 150 140 130 120 110 50.0 Brightn 0.0 Galactic longitude (degrees) (d) CGPS Figure1.BrightnesstemperatureofHiemission,inGalacticlatitudeandlongitudeco-ordinates,fromthePerseusarmintheCGPSand Perseusarmanaloguesinthesyntheticdata.Thelongitudecoverageisthesameineachcasebutthelatitudecoveragevariesaccording totheextentofHiemission(syntheticdata)orsurveycoverage(CGPS).Fig.1(a)isfromamodelgalaxywithoutfeedback,Fig.1(b)is fromamodelgalaxywithfeedbackand5percentstarformationefficiency,Fig.1(c)isfromamodelgalaxywithfeedbackand10per centstarformationefficiency,andFig.1(d)isfromtheCanadianGalacticPlaneSurvey. (cid:13)c 0000RAS,MNRAS000,000–000 The structure of Hi in galactic disks 5 120 120 100 100 K) K) ure ( 80 ure ( 80 at at er er p p m 60 m 60 e e s t s t s s e e htn 40 htn 40 g g Bri Bri 20 20 0 0 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Latitude (degrees) Latitude (degrees) (a) Latitudeaxisnotinverted (b) Latitudeaxisinverted Figure 2.ProfilesofHibrightnesstemperaturewithlatitude.ThesolidlineisfromtheCGPS,thedashedlineisfromtheNoFeedback model, the dotted line is from Feedback5 and the dot-dashed line is from Feedback10. The normalisation of the synthetic profiles has beenscaledtofittheCGPSdata,sothattheshapeoftheprofilescanbereadilycompared.InFig.2(b)thelatitudeaxisisinvertedso thatthesyntheticprofilesbettermatchskewnessoftheCGPSprofiles. TheamountofenergyinjectedintotheISMinthesim- Table 1. Scaling factors applied to latitude profiles from syn- thetic data and the corresponding RMS differences from the ulations is directly proportional to the star formation rate. CGPSprofile. The star formation rate (SFR) in the Feedback5 and Feed- back10modelsis∼0.7M /year,whichisconsistentwiththe (cid:12) Simulation Scalefactor RMS Reversedaxis lower limit of the observationally determined Galactic star difference(K) axis? formation rate of Robitaille & Whitney (2010) but approx- imately a factor 2 less than the SFR of Murray & Rahman NoFeedback 0.82 28.6 No (2010). We note that SFRs determined from both obser- Feedback5 0.535 8.86 No vations and models are sensitive to assumptions about the Feedback10 0.762 8.20 No slope and limits of the IMF (Calzetti et al. 2009). Addi- NoFeedback 0.821 28.5 Yes tionallywecanonlyinsertfeedbackonaresolvablescalein Feedback5 0.539 3.39 Yes the simulations, and cannot take into account all the sub- Feedback10 0.767 3.51 Yes resolution processes which may dissipate or radiate energy. Hence we cannot definitively say how well energy injection in the models is similar to that in the Milky Way. However the profile better matches the skewness of the CGPS data. the effect of feedback has clearly been to make the models Both Feedback5 and Feedback10 fit the CGPS data well, more realistic in this regard. provided the normalisation of the profile is scaled, and it is not possible to infer from these results whether 5% or 10% efficiencyisabetterfittothedata.ConverselytheNoFeed- 3.3 Longitude-velocity structure back model does not fit the CGPS data well as the profile showsalargediparoundthemid-planegivingaqualitatively Longitude-velocity plots in the Galactic mid-plane are different profile to that observed. shown in Fig. 3. These plots show local material around Although the star formation efficiency is different in v=0 and Perseus arm material at more negative velocities Feedback5 and Feedback10, the rate of star formation is (between−20and−60km/s).Theinclusionoffeedbackre- comparable at 250 Myr, as the process is largely self- sultsinmuchmorestructureinthespiralarmsinlongitude- regulating(asshowninfig.11ofDobbsetal.2011a).Conse- velocityspace.InFeedback5thePerseusarmmaterialiscon- quentlytheenergyinputfromstellarfeedbackissimilarfor tiguous but structured, whereas in Feedback10 the Perseus Feedback5andFeedback10(asthesupernovarateispropor- arm material appears less contiguous. The local material is tionaltothestarformationrate)andweexpecttheimpact highlydisruptedbyfeedbackbutasthismaterialiscloseto on the vertical distribution of Hi to be similar. Conversely the observer the angular size of the SPH smoothing length a synthetic survey from a simulation with twice the surface is large and the material is poorly resolved. density(runM10fromDobbsetal.2011a,notshownhere) The CGPS data show material at more negative veloc- has a larger scale height than Feedback5 and Feedback10. itiesthanthesimulations(beyond−60km/s)butwewould This is consistent with the higher star formation rate (ap- not expect to see material in this velocity range in the syn- proximatelyfourtimeshigher)andanincreasedlevelofstel- thetic data because the model galaxies do not have a suffi- lar feedback. cientlylargeradialextenttorepresentthismaterial.Forgas (cid:13)c 0000RAS,MNRAS000,000–000 6 David M. Acreman et al 200.0 100 K) ocity (km/s) 864000 110500..00 s temperature ( Vel 20 nes 50.0 ht 0 Brig 20 0.0 160 150 140 130 120 110 Galactic longitude (degrees) (a) NoFeedback 120 200.0 100 K) ocity (km/s) 864000 110500..00 s temperature ( Vel 20 nes 50.0 ht 0 Brig 20 0.0 160 150 140 130 120 110 Galactic longitude (degrees) (b) Feedback5 120 200.0 100 K) ocity (km/s) 864000 110500..00 s temperature ( Vel 20 nes 50.0 ht 0 Brig 20 0.0 160 150 140 130 120 110 Galactic longitude (degrees) (c) Feedback10 200.0 100 e (K) ocity (km/s) 864000 110500..00 s temperatur Vel 20 50.0 nes ht g 0 Bri 0.0 160 150 140 130 120 110 Galactic longitude (degrees) (d) CGPS Figure3.BrightnesstemperatureofHiemission,inGalacticlongitude-velocityco-ordinatesintheGalacticmid-plane.Fig.3(a)isfrom a model galaxy without feedback, Fig. 3(b) is from a model galaxy with feedback and 5 per cent star formation efficiency, Fig. 3(c) is fromamodelgalaxywithfeedbackand10percentstarformationefficiency,andFig.3(d)isfromtheCanadianGalacticPlaneSurvey. (cid:13)c 0000RAS,MNRAS000,000–000 The structure of Hi in galactic disks 7 in axisymmetric circular rotation the line of sight velocity 4 HI SELF-ABSORPTION (from equation 1 of Kalberla & Kerp (2009)) is When generating synthetic observations the calculation of v(R,z)=(cid:20)R(cid:12)Θ(R,z)−Θ (cid:21)sinlcosb (2) Hiintensitycanbesplitintoseparateemittingandabsorb- R (cid:12) ingcomponents.Thisallowsustoproduceadatacubecon- taining only the absorption component, in order to identify where v(R,z) is the line of sight velocity at point (R,z), where HISA is present. Furthermore, each cell of the AMR Θ(R,z) is the tangential velocity at (R,z), Θ is the ob- (cid:12) grid is identified as either a net source of emission or a net server’s tangential velocity, R is the distance of the ob- (cid:12) source of absorption by calculating the change in intensity server from the Galactic centre, and l and b are Galactic duetothegridcell,normalisedbythecolumndensityofthe latitudeandlongitude.Inoursyntheticsurveystheobserver cell.TheSPHparticlesarethenassociatedwiththechange is located at R = 7.1 kpc within a model galaxy with an (cid:12) inemissionfromtheAMRcellinwhichtheyreside,thuswe outerextentof10kpc,andatboththeselocationsthetan- are able to determine which SPH particles are responsible gential velocity is close to 220 km/s. Hence for material in for HISA. Compared to observers we are in the privileged the Galactic plane (b = 0) the line of sight velocity at the position of being able to unambiguously identify absorbing outer edge of the model galaxy is components in the data cube and furthermore we can iden- v(R,z)=−63.8sinl km/s (3) tifyabsorptionwithspecificSPHparticles.InSection4.1we usetheabsorption-onlydatacubestoexaminethedistribu- Atl=110themaximumextentinvelocityspaceisexpected tionofHISAontheplaneofthesky,theninSection4.2we to be −60 km/s decreasing to −22 km/s at l=160. This is use the identification of absorbing SPH particles to investi- inagreementwiththemostextremenegativevelocitiesseen gate the properties of the material which causes HISA. in the synthetic data in Fig. 3. 3.4 Expanding shells 4.1 HISA distribution Several shells of material are seen in the synthetic obser- The velocity integrated absorption component is plotted in vations (see Fig. 4 for an example from the Feedback5 Fig.5fortheNoFeedback,Feedback5andFeedback10mod- model),similartoshellsseeninHiobservations(Heiles1979, els. Equivalent plots from the CGPS data are shown in fig- 1984; Hu 1981; McClure-Griffiths et al. 2002; Ehlerova´ & ure 1 of Gibson et al. (2005). Palouˇs2005).Suchstructuresareexpectedasaconsequence In the NoFeedback model (Fig. 5(a)) there is a broad of energy feedback from massive stars and the feature in band of strong HISA, due to an over concentration of Hi Fig. 4 is associated with a feedback event which occurred in the mid-plane. HISA in the NoFeedback case is confined 1.2 Myr earlier2. The feedback event comprised 20 super- to ±1 degree from the mid-plane, whereas observed HISA novae, an atypically energetic event, which injected a total has a greater vertical extent. The HISA morphology in the of 2×1052 erg of energy into the ISM. Feedback5modelismorerealistic,withextended,lowinten- In longitude-latitude space (Fig. 4(a)) an expanding sity HISA surrounding knots of stronger absorption. Both shell appears as a variable radius ring with a maximum ra- themorphologyandverticalextentoftheFeedback5model diusatthecentralvelocityofthematerial.Theradiusofthe are more similar to the CGPS observations of Gibson et al. shell decreases in velocity channels away from the central (2005).AlthoughtheFeedback5HISAhaslesssubstructure value and terminates with two “caps” of emission from the in the strong HISA complexes this is to be expected, given frontandbackofthebubble.Bycalculatingtheaveragesur- thesmallspatialscaleoftheobservedstructuresrelativeto face brightness in concentric annuli about the centre of the themodelresolution.IntheFeedback10modeltheHISAhas shell, the structure can be plotted in velocity-radius space. a much reduced diffuse component, compared to the other A plot of this type, made using the kshell tool from the models. Gibson et al. (2005) find nearly ubiquitous weak karmasoftwarepackage3,isshowninFig.4(b).Invelocity- HISA and in this regard the Feedback10 model does not radius space the structure is an arch shape which confirms match the observations. Models with higher star formation that this structure is indeed expanding. efficiencieshavelessmaterialinthecoldphase,asshownin Figure4(b)showsthattheshellhasanexpansionveloc- fig. 4 of Dobbs et al. (2011a), so Feedback10 is expected to ityofapproximately10km/sandamaximumangularradius have less HISA than Feedback5 as there is less cold atomic of approximately 2 degrees, which corresponds to 80 pc at hydrogen.WeconcludethattheFeedback5modelhasamore 2.3kpc(thedistancefromtheobserver’sposition).Theshell realistic HISA morphology than either of the other models, radiusandexpansionvelocityaresmallerthantypicalvalues albeit with limited spatial resolution. fromthesamplesofHeiles(1979)andHeiles(1984),however Heiles(1984)notesabiastowardsselectinglargershellsand 4.2 Properties of material responsible for HISA our shell is more typical of the shells found by Hu (1981). Thesizeofourshellisalsoconsistentwiththesmallershells Figure 6 shows a histogram of number densities (Fig. 6(a)) inthesampleofMcClure-Griffithsetal.(2002),observedin and temperatures (Fig. 6(b)) for absorbing particles from the Southern Galactic Plane Survey. Feedback5, located in the region used to generate the syn- thetic survey. The solid line is for all particles in a cell 2 This time does not include the age of the supernova bubble with net absorption (6% of particles in region), the dashed lineisforparticlesassociatedwithabsorptionstrongerthan when the energy is inserted. This timescale (denoted t in Ap- pendix1,Dobbsetal.2011a)isoforder105 years. 1×10−23 erg/s/sr (2% of particles in region) and the dot- 3 http://www.science-software.net/karma/ ted line is for particles associated with absorption stronger (cid:13)c 0000RAS,MNRAS000,000–000 8 David M. Acreman et al Velocity= -23.5 km/s 250.0 40 150.0 35 2 200.0 30 ees) 1 e (K) 100.0e (K) Galactic latitude (degr 01 110500..00Brightness temperatur Velocity (km/s) 221055 50.0 Brightness temperatur 10 50.0 2 5 0.0 0 0.0 167 166 165 164 163 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Galactic longitude (degrees) Radius (degrees) (a) Shellinlongitude-latitudespace (b) Shellinvelocity-radiusspace Figure 4. An expanding shell of Hi emission from the Feedback5 model. The velocity-radius plot (right) shows the arch shape which is expected from an expanding shell. The angular offset is the radius of an annulus about the centre of the cell, where the centre is determinedbyeye. than 2×10−23 erg/s/sr (1% of particles in region). There there is sufficient spread in our temperature values to en- is no evidence for particles associated with stronger HISA compasstheirobservationallydeterminedtemperatures(see to be higher density or at a lower temperature, indeed the Fig. 6(b) and Fig. 6(d)). Material can drop out of the gas temperature range is 10-100K (typical of the cold neutral phaseandformsupernovaeaboveadensityof1000cm−3,so interstellar medium) for almost all the absorbing particles. thedensestandcoldestregionsofHISAproducinggas(seen ForthesehistogramsthemeanHinumberdensityis130–190 intheobservations)areatthelimitoftheSPHsimulations’ cm−3 and the mean temperature is 32–36 K. representation. The density and temperature distributions for absorb- The next point to be addressed is whether particles ing particles from Feedback10 are shown in Fig. 6(c) and which contribute to HISA are in a different phase to other Fig. 6(d) respectively. Particles associated with net absorp- particles at similar temperatures, and whether HISA pro- tion are plotted as a solid line but no other thresholds are ducing material is in pressure balance with other material. used, as the number of absorbing particles is much smaller InFig.7weplotpressureagainstnumberdensityforHISA than for Feedback5 (1% of particles are absorbing in Feed- particles(Fig.7(a))andforallotherparticleswithatemper- back10). For Feedback10 the mean Hi number density is ature below 150K (Fig. 7(b)). There are approximately five 490 cm−3 and the mean temperature is 27K. The range of timesasmanyparticlesinFig.7(b)asinFig.7(a),indicat- temperaturesanddensitiesseeninFeedback10aresimilarto ingthatthereisasignificantamountofcold,denseHiwhich those seen in Feedback5, although the HISA in Feedback10 is not observed in HISA. For particles with a number den- isfromslightlycolderanddensermaterial.Thestarforma- sityabove102cm−3 thereisnoapparentdifferencebetween tion efficiency has only a modest effect on the properties of thetwopopulations.Atlowerdensitiesthedistributionsare material seen in HISA, even though the morphology of the also similar but there is more spread (towards higher pres- HISA is significantly different. sures) in the non-HISA particles. The thermal pressure of In their study of HISA clouds in Perseus, Klaassen HISA material does not appear to be significantly different etal.(2005)foundnumberdensitiesbetween100cm−3 and toothermaterial,althoughwenotethattheeffectofvelocity 1200 cm−3, and spin temperatures in the range 12K–24K, dispersionhasnotbeenincludedhere,whichwouldprovide inanextendedHISAcloudwhichtheytermthe“complex”. support against gravity in addition to thermal pressure. In a smaller, isolated HISA feature, which they term the Although the material responsible for HISA is identifi- “globule”,thespintemperatureswereintherange8K–22K able as the cold neutral medium, which is expected to go (no density determination was made for the globule.) The on to form stars, we can be more specific about the fate of globule is a compact structure (unresolved in the 1 arcmin theSPHparticlesresponsibleforHISA.Particleswhichare mainbeamoftheCGPS)andismuchsmallerthantheres- in giant molecular clouds (GMCs) are identified, using the olution limit of our synthetic observations (see section 3.1). clump finding algorithm of Dobbs et al. (2008), and corre- The complex region is much larger and hence is more like lated with whether the material is also observed in HISA. the HISA clouds seen in our synthetic data. The density In the Feedback5 model 4.0% of the HISA particles are of HISA producing material in Feedback5 and Feedback10 in GMCs, compared to 2.2% of material in GMCs for the is consistent with the density of the complex region found galaxy as a whole, and 70% of HISA material is involved by Klaassen et al. (2005). Although the mean temperature in a feedback event (i.e. involved in star formation) within of our HISA producing material is slightly higher than the the next 20 Myr. The fraction of HISA material in clouds, temperatures found by Klaassen et al. (2005) we note that andthefractionwhichformsstars,isnotfoundtovarywith (cid:13)c 0000RAS,MNRAS000,000–000 The structure of Hi in galactic disks 9 0.0 -50.0 es) 3 s) degre 2 -100.0 K km/ e ( 1 e ( d d u 0 -150.0 u ctic latit 12 -200.0 A amplit a S Gal 3 HI 145 140 135 130 125 120 -250.0 Galactic longitude (degrees) -300.0 (a) NoFeedback 6 0.0 4 -50.0 es) s) degre 2 -100.0 K km/ e ( e ( d d u 0 -150.0 u ctic latit 2 -200.0 A amplit a S al HI G 4 -250.0 -300.0 150 145 140 135 130 125 120 Galactic longitude (degrees) (b) Feedback5 6 0.0 4 -50.0 es) s) degre 2 -100.0 K km/ e ( e ( d d u 0 -150.0 u ctic latit 2 -200.0 A amplit a S al HI G 4 -250.0 -300.0 150 145 140 135 130 125 120 Galactic longitude (degrees) (c) Feedback10 Figure 5.Hiself-absorptionamplitude,calculatedfromthesyntheticobservations,wheretheabsorptionisintegratedoverallvelocity channels.Fig.5(a)isfromamodelgalaxywithoutfeedback,Fig.5(b)isfromamodelgalaxywithfeedbackand5percentstarformation efficiency,Fig.5(c)isfromamodelgalaxywithfeedbackand10percentstarformationefficiency (cid:13)c 0000RAS,MNRAS000,000–000 10 David M. Acreman et al 400 300 0 0 350 -1x10-23 250 -1x10-23 300 -2x10-23 -2x10-23 es 250 s 200 No. of particl 125000 o. of particle 150 N 100 100 50 50 0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 log (number density / cm-3) 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 log (Temperature/K) (a) Feedback5numberdensity (b) Feedback5temperature 70 90 80 60 70 50 s 60 e s particl 40 article 50 No. of 30 o. of p 40 20 N 30 20 10 10 0 0.5 1 1.5 2 2.5 3 3.5 4 0 log (Number density / cm-3) 0.8 1 1.2 1.4 1.6 1.8 2 2.2 log (Temperature/K) (c) Feedback10numberdensity (d) Feedback10temperature Figure6.NumberdensityandtemperaturedistributionsofSPHparticlesassociatedwithHiself-absorption(HISA).FortheFeedback5 model three different absorption thresholds are used to define HISA particles: dI < 0 (solid line), dI < −1×10−23 erg/s/sr (dashed line),anddI <2×10−23 erg/s/sr(dottedline).FortheFeedback10modeltherearefewerHISAparticlesandonlyonelineisplotted (particleswithdI<0). HISA strength. For Feedback10 we find that 2.5% of HISA Hi emission and their profiles of Hi emission with latitude material is in GMCs, compared to 0.9% in the galaxy as a match the CGPS observations well (provided the normali- whole, and 57% of HISA material goes on to be involved in sationisallowedtovary).Basedontheverticaldistribution star formation. ofHiemissionaloneitisnotpossibletodeterminewhether Feedback5 or Feedback10 is more realistic. When feedback is included more structure is seen in Hi emission, including bubbles of material associated with 5 CONCLUSIONS AND DISCUSSION supernovaevents,howeverthemodeldoesnotyethavesuf- The inclusion of feedback in the galaxy models has a sig- ficientspatialresolutiontocapturetheveryfinescalestruc- nificant effect on the derived synthetic Hi Galactic plane tureseenintheCGPSdata.Withincreasesinavailablecom- surveys. In the model without feedback (NoFeeback) gas is puting power it will be feasible to run the SPH simulations overly confined to the mid-plane, which results in excessive usingmoreparticles,givingahigherspatialresolution.This absorption. Consequently there are bands of bright emis- should result in more realistic small scale structure in the sion above and below the mid-plane (which are not seen in syntheticobservations.Howeverifthefilamentarystructure observations) and the vertical extent of the Hi emission is is due to the presence of magnetic fields then these too will too small. The two models which include feedback (Feed- need to be included before a good match with the observed back5andFeedback10)bothhavealargerverticalextentof Hi morphology on a small scale can be expected. (cid:13)c 0000RAS,MNRAS000,000–000