Astronomy&Astrophysicsmanuscriptno.starburst˙astroph c ESO2009 (cid:13) January23,2009 Cosmic Rays VI Starburst Galaxies at multiwavelengths JuliaK.Becker1,2,⋆,PeterL.Biermann3,4,5,6,7,JensDreyer2,andTanjaM.Kneiske2,8 9 0 1 Go¨teborgsUniversitet,Institutionenfo¨rFysik,41296Go¨teborg,Sweden 0 2 TechnischeUniversita¨tDortmund,Institutfu¨rPhysik,D-44221Dortmund,Germany 2 3 MPIforRadioastronomy,AufdemHu¨gel69,D-53121Bonn,Germany 4 Dept.ofPhys.&Astron.,Univ.ofBonn,Germany n 5 Dept.ofPhys.&Astron.,Univ.ofAlabama,Tuscaloosa,AL,USA a J 6 Dept.ofPhys.&Astron.,Univ.ofAlabama,Huntsville,AL,USA 7 Inst.Nucl.Phys.FZ,KarlsruheInst.ofTechn.(KIT),Karlsruhe,Germany 3 8 UniversityofHamburg,Institutfu¨rExperimentalphysik,Hamburg,Germany 1 January23,2009 ] E ABSTRACT H . Context.Starburstgalaxiesshowadirectcorrelationbetweenradioandfar-infraredemission.Hightargetdensitiesandahighrate h of supernova explosions imply the possibility of accelerating hadronic cosmic rays and producing decay products from hadronic p interactions,likehigh-energyneutrinosandphotons. - o Aims.Weproposeanexplanationforthefar-infrared/radiocorrelationofgalaxiesintermsoftheenergybalanceoftheinterstellar r mediumanddeterminethefluxfromhigh-energyphotonsandneutrinosfromstarburstgalaxies. t Methods.Wepresentacatalogofthe127brighteststarburstgalaxieswithredshiftsofz<0.03.Inordertoinvestigatethecorrelation s between radio- and far-infraredemission, we apply theleaky box approximation. Further, wederive photon- and neutrino spectra a [ fromproton-protoninteractionsinsupernovaremnants(SNRs).Here,weassumethatafractionoftheSNR’senergyistransferredto theaccelerationofcosmicrays.WealsoinvestigatethepossibilityofdetectingGammaRayBurstsfromnearbystarburstgalaxies, 1 usingthecatalogdefinedhere. v Results.Weshowthattheradioemissionisonlyweaklydependentonthemagneticfield.Itturnsoutthattheintensityoftheradio 5 signal is directly proportional to the number of supernova explosions, which scales with the far-infrared luminosity. In addition, 7 wefindthathigh-energyphotons fromproton-proton interactionsinSNRsinstarburstscanmakeupseveral percentofthediffuse 7 gamma-raybackground.Theneutrinofluxfromthesamesourceshasamaximumenergyof 105 GeV.Neutrinoscan,ontheother 1 hand,canbeobservedifaGammaRayBursthappensinanearbystarburst.About0.03GRB∼speryearareexpectedtooccurinthe . entirecatalog.Thetruenumberisexpectedtobeevenhigher,sinceweonlyincludethebrightestsources.Thenumberofeventsper 1 burstinIceCubevariesbetweenaboutoneeventandmorethan1000events.ThisprovidesgoodprospectsforIceCubetodetecta 0 significantevent,sincethebackgroundforaGRBsearchisclosetozero. 9 0 Keywords.Galaxies:starburst–Infrared:galaxies–Radiocontinuum:galaxies–Catalogs–cosmicrays–Neutrinos : v i X 1. Introduction Galaxiesalsohaveabundantfar-infrared(FIR)emission,which r is dueto dust. Thisdustis heatedby stars, oftenmostly young a Radio emission from galaxies is usually dominated by syn- stars. As was noted from the mid-eighties, this thermal dust chrotron emission from a population of non-thermal, energetic emissioncorrelatesratherwellwiththenon-thermalradioemis- electrons in a magnetic field which permeates most of the in- sion. The correlation in its most simple form is just a propor- terstellar medium. This radio emission is often spatially struc- tionalitybetweenfar-infraredandnon-thermalemission.Asref- tured, such as in the starburst galaxy M82, showing individual erence wavelengths, 60µm and 100µm are used for the far- compact sources, which can be interpreted as fairly young su- infrared. Frequencies between 1.4GHz and 5GHz are used as pernova remnants (Kronbergetal., 1985; Kronberg&Sramek, typicalfortheradioregime. 1985; Barteletal., 1987). The origin of these energetic elec- Manyattemptshavebeenmadetounderstandtheproportional- trons,apartofthecosmicrays,isthusexpectedtobetheyoung itybetweenradioandfar-infrared.Onthebasisofrathersimple supernova remnants, see Baade&Zwicky (1934); Shklovskii modeling of galactic evolution, a strong correlation is actually (1953)andforaextensivereviewBerezinskyetal.(1990).Thus expected,sincebothsupernovaremnantsandthedominantheat- the radio emission is a key to interpret the physics of cosmic ingbyultravioletlightfrommassivestarsderivesfromthesame rays, and conversely, any attempt to understand cosmic rays stellarpopulation(Biermann,1976;Biermann&Fricke,1977). should also try to understand the properties of the radio emis- Using such models the far-infrared luminosity of NGC2146 sion. had been predicted by Kronberg&Biermann (1981) and veri- fied subsequentlyby IRAS observations(Moshiretal., 1990a). ⋆ Correspondingauthor.Contact:[email protected],phone: The correlation as seen in the data was first clearly stated by +46-31-7723190 deJongetal.(1985),andsubsequentlydiscussedatsomelength 2 JuliaK.Beckeretal.:CosmicRaysVI by many authors (Bicayetal., 1989; Wunderlichetal., 1987; FIR luminosity vs. distance Wunderlich&Klein, 1988, 1991; Condonetal., 1991). It still ] 34 s defiesaclearexplanation. g/ An extensive attempt to interpret the correlation was made er 33 [ bLyiseVnfo¨ellkdeatnadl.,c1o9ll9a6b)o.rTathoersele(Vctor¨olkn,sa1r9e8b9e;liXevueedttaol.l,os1e9a9l4lbe,na-; )micron32 0 ergy in this model and therefore, the correlation is calorimet- L6 ric.However,thiswouldpredictanactualsteepeningofthera- g( 31 o dioemissionbetweenthereferencefrequencies,aneffectwhich L is not seen in the data. The solution of a spatial mixture of 30 No cuts cutoffs would still allow for a locally loss-dominated scenario. z<0.03, S >= 4 Jy ,S >= 20 mJy However,first spatially resolvedobservationsof M 33 indicate 29 60 micron 1.4 GHz 4 Jy flux that the local star-forming regions typically have flat spectra, too (Tabatabaeietal., 2007a,b). It is therefore likely that star 2-08.05 0 0.05 0.1 0.15 0.2 0.25 0.3 formingregionsaregenerallyinjectiondominated. Luminosity distance [Gpc] Here, we proposea simple model,basedupona particularpic- tureoftheenergybalanceintheinterstellarmedium.Thismodel Fig.1. 60µ luminosity - distance diagram. Crosses indicate all alsousessomesimple assumptions,aswewillemphasize.The 309pre-selectedstarburstgalaxies,squaresshowthoseremain- model is local, and so automatically allows for starbursts and ingafterthecutsS >20mJyandS >4Jyandz<0.03. 1.4GHz 60µ gradientsindiskgalaxies,whileupholdingthecorrelation.The ThedashedlineshowsthesensitivityforS >4Jy. 60µ modelleadstosomespecificpredictionswhichcanbechecked withfurtherdata.Acatalogofstarburstgalaxiesispresentedto performfirstchecks. Onlylocalsourcesareconsidered,sinceouraimistoinvestigate Theoutlineofthispaperisasfollows:Wefirstdefinevariables, theclosestsources.Thecatalogpresentedhereconsistsofatotal used throughout the paper, in Section 2. In Section 3, a cata- of127starburstgalaxies.Thisisasub-samplefromalargersam- logof127nearby,brightstarburstgalaxiesispresented,includ- pleof309starburstgalaxies,applyingcutsatbothFIRandradio ing far-infrared,radioand X-ray data. The correlationbetween wavelengthstoensureacomplete,localsample.Thesecutsare FIR and radio emission is outlined together with the difficulty discussedinthefollowingparagraphs.Differenttestswereper- ofexplainingit.InSection4,wepresentamodelexplainingthe formedinordertoverifythattheconsideredgalaxiesareindeed FIR-radiocorrelation.Thepossibleemissionofcosmicraysand starbursts,aspresentedinthefollowingparagraphs.Inorderto secondariesproducedinproton-protonandproton-photoninter- removecontaminationfromSeyfertgalaxies,we onlyuse high actionsis discussedin Section5. In particular,we examinethe ratiosofFIRtoradiofluxdensity,i.e.S /S >30.Totest 60µ 1.4GHz possibilityofdetectingsecondariesfromsupernovaremnants,as ifthesampleconsistsofstarburstgalaxiesasopposedtoregular wellascosmicraysandsecondariesfromGammaRayBurstsin galaxies,wecheckthatthecorrelationbetweenradiopowerand starbursts.Finally,implicationsarediscussedinSection6. FIR luminosityis a direct proportionality.Apartfrom that, our maincriterionforthecatalogisthatthesourcesarecloserthan z < 0.03, i.e. located in the supergalactic plane, and that they 2. Definitions havebothradio andIRdetections.The latter givesusinforma- Inthefollowingsections,theelectromagneticspectraatdifferent tionabouttheratiooftheIRtoradiosignal,whichwerequireto wavelengthsareusedinordertoinvestigatethestarburstnature belargerthan30.Thisensuresa highIRcomponentcompared ofthecatalogsources.Table1givesasummaryofthedifferent to the radio part, i.e. that the sources are indeed starbursts and parametersused.Concerningspectralpower-lawfitsbetweenthe notSeyfertgalaxies.Further,weapplysensitivitycuts,weonly wavelengths,weusetheconvention includesourceswithafluxdensity> 4Jyat60µmandaradio fluxdensityat1.4GHzlargerthan20mJy.Figures1and2show ν α the60µmresp.1.4GHzluminosityofstarburstgalaxiesversus S =S , (1) 0· ν ! their luminosity distance. The dotted lines represent the sensi- 0 tivity for 4 Jy, resp. 20 mJy. Crosses represent all 309 sources with S as the flux per area and frequency interval, in units of we selected in the beginning, squares show those 127 sources Jy=10−26W/m2/Hz. Here, S0 is the flux at a reference fre- remainingafter the cuts at S1.4GHz > 20 mJy and S60µ > 4 Jy, quencyν0. as well as z < 0.03. We apply those cuts in order to ensure a complete,localsampleinbothFIRandradiowavelengths. Since the sources are closer than z = 0.03, many of the star- 3. Localstarbursts:asample bursts are located in the supergalactic plane. Their spatial dis- Inthissection,wepresentasampleoflocalstarburstgalaxies1. tributionshouldthereforebeaflatcylinderwithafurthermore ThedataoftheindividualsourcesarepresentedinappendixA. sphericalcomponent,forthose sourcesnotin the supergalactic plane. We therefore expect that the number of sources with a 1 The data have been collected from the references (Heeschen&Wade, 1964; Whiteoak, 1970; Sramek, Douglasetal., 1996; Condonetal., 1996, 1998; Whiteetal., 2000; 1975; Sramek&Tovmassian, 1976; Disney&Wall, 1977; Condonetal., 2002; Sandersetal., 2003; Stricklandetal., 2004; Dressel&Condon, 1978; Ku¨hretal., 1981; Condonetal., 1983; Vollmeretal., 2004; Suraceetal., 2004; Bravo-Alfaroetal., 2004; Wright&Otrupcek, 1990; Condon, 1983; Beichmanetal., 1988; Leroyetal., 2005a; Nagaretal., 2005; Ottetal., 2005; Leroyetal., Soiferetal., 1989; Moshiretal., 1990b; Beckeretal., 1991; 2005b; Ionoetal., 2005; Tajeretal., 2005; Tengetal., 2005; Fabbianoetal., 1992; White&Becker, 1992; Brinkmannetal., Guainazzietal.,2005;Baan&Klo¨ckner,2006;Tu¨llmannetal.,2006; 1994; Knapp, 1994; Wrightetal., 1994; Griffithetal., 1994, 1995; Gallimoreetal., 2006; Lisenfeldetal., 2007; Rosa-Gonza´lezetal., Beckeretal., 1995; Rigopoulouetal., 1996; Wrightetal., 1996; 2007;Shuetal.,2007) JuliaK.Beckeretal.:CosmicRaysVI 3 parameter symbol units Electronspectrum ––psericmonadryary ddNNee//ddEEee ∝∝EEee−−ααpeseercim GGeeVV−−11ss−−11ssrr−−11ccmm−−22 Electronenergy E keV e espectralindex –primary αprim e –secondary αsec e Protonspectrum dNp/dEp =Ap·(Ep/Emax)−αp· GeV−1s−1sr−1cm−2 exp( E /E ) · − p max Protonenergy E GeV p pspectralindex α p pcutoffenergy E GeV max Neutrinospectrum dNν/dEν =Aν·(Eν/GeV)−αν GeV−1s−1sr−1cm−2 Neutrinoenergy E GeV ν νspectralindex α ν normalizationfactor A GeV 1s 1sr 1cm 2 ν − − − − Radiofluxdensity –at1.4GHz,2.4GHz, 2.7GHz&5GHz S ,S , S , S mJy 1.4GHz 2.4GHz 2.7GHz 5GHz IRfluxdensity,IRAS –at12µm, 25µm, 60µm&100µm S ,S ,S &S Jy 12µ 25µ 60µ 100µ IRfluxdensity,2MASS Jy –at1.25µm,1.65µm&2.17µm S , S ,S Jy 1.25µm 1.65µm 2.17µm X-rayfluxdensity,ROSAT –btw.[0.1-4.5],[0.1-2.3]or[0.2-2.0]keV S nJy ROSAT spectralindex –btw1.4GHz&5GHz α –btw1.4GHz&60µm α rir –btw60µm& 1keV α ∼ xir Table1.Summaryofparametersusedinthispaper. Radio luminosity (cid:10) 1.4GHz vs. distance NN>(S> Sv)s .v IsR. IfRlu fxl u(cid:10) 6x0 (cid:10) 6m0i croµnm w withit hc uctust sa papplpielided s] 32 S) erg/ 31 N(>102 fit N(>S)=N0(S+S0)-β [ ) Hz 30 G 4 L1. 29 g( 10 o 28 L No cuts 27 z<0.03, S >= 4 Jy ,S >= 20 mJy 60 micron 1.4 GHz 1 26 20 mJy flux 25 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 -1 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 Luminosity distance [Gpc] Log(S )[Jy] 60µm Fig.2. 1.4 GHz luminosity - distance diagram. Same notation Fig.3.logN logS representationofthecatalog.AnS 1.2 fit − − − as Fig. 1. The dashed line shows the sensitivity for S > matchesthedatanicely,withaturnoveratS =10.56Jy. 1.4GHz 0 20mJy. Here,N , S andβarefitparameters.Usinganerrorof √N,the 0 0 parametersaredeterminedto fluxdensitylargerthanS, N(> S),shouldfollowabehaviorof N0 = 3155 1297.9 ± Sde−r1,−whSi−le1.5a.sAphpeurirceaSld−1is−trbiebhuativoinorreissuelxtspeinctaendSfor1.a5 flbaethcayvliionr-. S0 = (10.56±3.78)Jy − β = 1.2 0.2. − Figure3showsthelogarithmicnumberofsourcesaboveanFIR ± fluxdensityS60µ.Wefitthedatawiththefollowingfunction: The behavior N(>S) S 1.2 0.2 matches the expectation that − ± thefunctionshouldlie∼betweenS 1.0andS 1.5.Inthefollowing − − paragraphs,wewillinvestigatefurtherwhethertheclassification N(>S)= N (S +S ) β (2) ofthe127sourcesasstarburstsisjustified. 0 0 − · 4 JuliaK.Beckeretal.:CosmicRaysVI Radio Luminosity vs. FIR Luminosity (z>0.03, S60µm/SRadio>30, S60µm>=4Jy, SRadio>=20mJy) radiofluxdensityat1.4GHz,S1.4GHz: ) 32 ] z H S s/ 31 s := 60µ . (6) g/ 60µ/1.4GHz S r 1.4GHz e 30 [ z H G 1.4 29 For Seyfert galaxies, this ratio is about s60µ/1.4GHz ∼10, while P it is significantly higher in the case of starburst galaxies, ( og 28 s60µ/1.4GHz 300. The histogram of the ratio between the FIR L flux density∼at 60µm and the radio flux density at 1.4GHz is 27 showninFig.5.All127sourceshavearatioofs60µ/1.4GHz >30, whichconfirmsthatthesourcesarenotlikelytobeSeyferts. 26 41 42 43 44 45 46 47 Log(L [erg/s]) FIR Flux ratios 60µm and Radio(1.4 GHz-5 GHz)(z>=0.03, fraction>=30, S60µm>=4 Jy, S1.4GHz>=20mJy) Fig.4.RadiopowerP1.4GHzatν=1.4GHzversusFIRluminos- nt ityL .Adirectproportionality,P L isfound. u 16 FIR 1.4GHz FIR o ∝ C 14 3.1.FIRluminosityversusRadiopower 12 10 Lookingatawelldefinedsampleofgalaxies,itturnsoutthatthe correlationbetweenradioandfar-infrared(FIR)emissionisnot 8 linear, i.e., that the radio luminosity is proportional to the far- 6 infraredluminositytothepower1.30 0.03(Xuetal.,1994b). ± 4 AsXuandcollaboratorsnote,thefar-infraredemissionhastwo heating sources, stars that do explode later as supernova rem- 2 nants,andalsostars,thatwillneverexplodeassupernovae.This 0 0.5 1 1.5 2 2.5 3 3.5 second population of stars needs to be corrected for, and their Log(S / S ) contribution to the dust heating needs to be eliminated. This 60µm 1.4GHz then leads to a corrected far-infrared luminosity, which is di- Fig.5. Ratio of the flux density at 60µm and at 1.4 GHz. All rectlyproportionaltotheradioluminosity(Xuetal.,1994b). sources in the sample have ratios larger than 30, which indi- Theproportionalityholdsalongadiskinagalaxy,evenforfairly catesahighstarformationrate.Themedianisaround100.This shortlivedphaseslikeastarburst,suchasinM82,andthusre- matches previous investigations, e.g. Biermannetal. (1985), quiresclearlylocalphysics,withashortreadjustmenttimescale. who find a mean value of 250 at higher radio frequencies, Thisposesaseveredifficultyforanyproposaltoexplainthera- ν=5GHz. dio/FIRcorrelation. The FIRluminosityin the rangeof 60µmand100µmis given as(Xuetal.,1994b): L :=4πd2 F . (3) FIR l · FIR 3.3.RadiotoInfraredandX-raytoInfraredspectralindices Here,d istheluminositydistanceoftheindividualsourcesand l A further criterion of distinguishing regular galaxies and F :=1.26 10 14 2.58 S60µ + S100µ Wm 2 (4) Seyferts is their spectral index from X-ray to IR (XIR) FIR · − ·" · Jy ! Jy !# − and from radio to IR (RIR). The diagram of the XIR (1 keV to 60 µm) versus RIR (5 GHz to 60 µm) index of istheFIRfluxdensityatEarthasdefinedinHelouetal.(1988). the sources in shown in Fig. 6. Derived from figure 3 in Thenormalizationfactorcomesfromthe frequencyintegration (Rodriguez-Pascualetal., 1993), starburst galaxies have spec- andfromthe conversionofJy to W/m2/Hz.InFig.4, theloga- tral indices scattering around (RIR,XIR) (0.6, 1.9), starburst rithmoftheradiopowerat1.4GHz,P1.4GHzversusthelogarithm Seyfert-Igalaxiesshow(RIR,XIR)Sy I (0.48, ∼1.2),Se−yfert- oftheFIRluminosityLFIR isshownforourcatalog.Thecircles II galaxies have (RIR,XIR)Sy II −(0.∼47, 1.6−) and quasars show the single sources and the solid line is a fit through the are located at (RIR,XIR) − (∼0.28, 1.−1). The values for quasar data.Thefityieldsacorrelationof the RIR and XIR indices of∼starburs−t galaxies given by Chinietal. (1989) are slightly higher, which matches the sam- P L 1.0. (5) 1.4GHz ∝ FIR ple examined here: Chinietal. (1989) give a RIR index of This demonstratesthat short stellar lifetimes dominatethe cor- 0.82 and a XIR index of 1.66. We find average values of − relationinoursample,andsothisisstronglysupportingourhy- (RIR, XIR)=(0.82, 1.77) which is compatible with the ex- − pothesis,thatthemajorityofoursamplegalaxiesarestarbursts. pectedresult. Still, we do not have X-ray data for all the sources, so there may still be some contamination from both Seyferts and regu- 3.2.Infraredtoradiofluxdensityratio lar galaxiesinthe sample.As we onlyused catalogswherethe Generally,regulargalaxiesaredistinguishedfromactivegalax- sources have previously been identified as starbursts, this con- iesbytheirratiooftheFIRfluxdensityat60µm,S ,andthe taminationshouldbesmall. 60µ JuliaK.Beckeretal.:CosmicRaysVI 5 RIR Index vs. XIR Index Ithadbeenobservedearlythatthethreemaincomponentsofthe x1.2 interstellarmedium,thegas,thecosmicrays,andthemagnetic e field have very similar energy densities, or pressures. Since all d Seyfert I Galaxies R In 1 SQeSyOfesrt II Galaxies tahtraepepdreorxivimeafrteomeqvueipryardtiitffioenreinmtpplhiyessitchaaltptrhoectehsrseees,titmoekesecpaltehseomf RI0.8 Mean of this sample changeare also all threethe same. Thisisthe basic premiseof This sample the following argument, and it will lead naturally to an under- 0.6 standingofthefar-infraredradiocorrelation.Sointhisspecific sense it is a calorimetric argument similar to Vo¨lk (1989), al- 0.4 thoughweapproachtheproblemisasomewhatdifferentway. The following line of reasoning is visibly influenced by 0.2 earlier papers, like Biermann (1950); Biermann&Schlu¨ter (1951); Cox (1972); Cox&Smith (1974); McKee&Ostriker -02.2 -2 -1.8 -1.6 -1.4 -1.2 -1 -0.8 (1977); Beuermannetal. (1985); Kronbergetal. (1985); XIR Index Kronberg&Sramek (1985); Snowdenetal. (1997); Hunteretal. (1997); Becketal. (2003); Hanaszetal. (2004, 2006),andofcourseothersmentionedinduecourse. Fig.6.Radio-to-IRspectralindexversusX-ray-to-IRspectralin- First we wish to establish the concepts which we use, for easy dex.Thecrossesrepresentthose48sourcesinourcatalogwith reference,andthenapplythemtotheproblemhere. radio, FIR and X-ray measurements. The blue triangle shows theaverageofthevalues.Theopencircleshowstheaveragelo- cationofSeyfert-Igalaxies,theopensquarerepresentsaverage 4.1.Thethreemaincomponentsoftheinterstellarmedium Seyfert-II galaxies and the star indicates QSOs. The last three valuesaretakenfromRodriguez-Pascualetal.(1993).Notethat Thethreemaincomponentsarethegas,thecosmicraysandthe individualgalaxiesscatteraroundthegivenvalues(Chinietal., magneticfield.Atleastthegasandthemagneticfieldisclearly 1989). spatiallyhighlyinhomogeneous(Becketal.,2003): The gas has a number of components, molecular clouds, neu- tralHydrogenclouds,diffuseneutralHydrogen,diffuseionized 4. TheinterstellarmediumandtheFIR/radio Hydrogen, HII regions, stellar wind bubbles, supernova rem- nantswithX-rayemittingshells,atunnelnetworkofconnected correlation oldersupernovaremnants(Cox&Smith,1974),athickhotdisk The interstellar medium connects the formation of stars, the (Beuermannetal., 1985; Snowdenetal., 1997; Kronbergetal., explosion of supernova remnants, the regularization and en- 2007), and a wind (Breitschwerdt, 2008; Everettetal., 2008). hancement of the magnetic field, and the transport of cosmic Thetunnelnetworkprobablyconnectstothehotthickdisk.The rays. In order to understand the observation that the thermal windisprobablyfedfromthehottestregionsofthetunnelnet- hot dust emission from a galaxy is simply proportional to the work, in a fashion perhaps similar to the Solar wind being fed non-thermal radio emission from relativistic cosmic ray elec- fromcoronalholes(Stepanianetal.,2008,e.g.). trons,weneedtounderstandtheinterstellarmedium,oratleast Themagneticfieldispermeatingalmosteverything,andalsohas get close enough to a comprehension, that we can understand a thick disk. The field permeates the clouds, and is often con- thisamazingcorrelation(deJongetal.,1985;Wunderlichetal., finedvisiblybytheclouds(Appenzeller,1974,e.g.).Themag- 1987;Wunderlich&Klein,1988,1991;Condonetal.,1991). netic field is strongly perturbed by HII regions, and supernova Obviously, since the very massive stars power the far-infrared explosions.Themagneticfieldistransportedoutofthediskby emission through a large fraction of the ultra-violet emission, thewind. whichisabsorbedbydust,andtheninsupernovaexplosionspro- The cosmic rays, produced by supernova explosions in shock- ducetheenergeticcosmicrayelectrons,whichemittheobserved waves (Baade&Zwicky, 1934; Fermi, 1949; Drury, 1983; radioemission,thereshouldbeacorrelation.Usingsimpleini- Berezinskyetal., 1990) cannot easily be repelled by anything, tial models for stellar population evolution, this rather naive andsogothroughallclouds,andalltheneutralandionizedgas. earlypicturesuccessfullypredictedtheapproximatefar-infrared In bulk they cannot travel faster then the Alfve´n speed, since emissionalreadyin1977(Biermann,1976;Biermann&Fricke, otherwisetheywouldexcitewavesintheplasma,scatteringthe 1977; Kronberg&Biermann, 1981). However, the tightness of particles,effectivelyreducingtheirbulkvelocity. the correlation was neither anticipated nor predicted. The cor- Themagneticfieldisinstrumentaltoallowcosmicrayaccelera- relation was finally discovered by deJongetal. (1985). Since tioninshocks,andperhapsthroughoutthemedium. the dust emission just measures the total output in ultra-violet The cosmic rays in turn drive the dynamo mechanism to en- by youngstars, it seems obvious, that in the limit of much ab- hance an existing magnetic field and give it spatial coherence sorption,thefar-infraredemissionwouldjustbeproportionalto (Parker, 1969, 1992; Ferrie`re, 1996; Ferrie`re&Schmitt, 2000; the star formation rate, given a general initial mass function. Hanaszetal.,2004,2006;Otmianowska-Mazuretal.,2008).In However, the radio emission is approximately proportional to the classical Biermann-battery mechanism (Biermann, 1950; theproductofthecosmicrayelectrondensityandthemagnetic Biermann&Schlu¨ter, 1951), only a rotating star with surfaces fieldenergydensity,andthereforeitisnotreallyobviousatall, of density and pressure in non-coincidence is required to pro- that integrating along a vertical column through the disk of a duceaseedfield,which,however,isgenerallyweak;thedynamo galaxythesetwoemissioncomponentsshouldbebasicallypro- mechanism in stars can strongly enhance magnetic fields, and portional. throughwindsejectthem (Bisnovatyi-Koganetal., 1973, e.g.): Modern descriptions of starburst galaxies such as M82 or Thiswouldconstituteaveryirregular,butpotentiallyrelatively NGC2146 are in Dopitaetal. (2005, 2006a,b); Grovesetal. strong source of magnetic fields in galaxies;in such a case the (2008). dynamomechanismonaGalacticscaleisrequiredmoretoreg- 6 JuliaK.Beckeretal.:CosmicRaysVI ularizethefieldratherthantostrengthenit.Inthemechanismof 4.2.Supernovaexplosions Lucek&Bell(2000);Bell(2004,2005)theshockwavescandi- FordidacticsimplicitywebasicallyadopttheapproachofSedov rectlyenhancethemagneticfield,usinganexistingpopulationof (1958); Cox (1972), but use more modern cooling approxima- cosmicrays.Asnotedalready,thecosmicrayscoupleeffectively tions, and allow for much lower environmental densities, but tothegas.Anisotropiesinthephasespacedistributionofparti- otherwiserescaletheirequations. cles,alwayspresentinshockwaves,alsocanproducenewmag- Anexplosionrunsintotheinterstellarmedium,andexpandsinto netic fields on small scales (Weibel, 1959; Bykov&Toptygin, the tenuous gas, which surrounds the clouds, and extends far 2005). Galactic magnetic fields have also been reviewed by above and below the central layer of cool clouds. We consider Becketal.(1996);Kulsrud(1999);Kulsrud&Zweibel(2008). theexpansionintothesurroundinglowdensitymediumandask, Thermodynamicallyboththecosmicraysandthemagneticfield whentheexpansionrunsintothecoolinglimit: canbethoughtofasarelativisticgas,withalmostzeronetmass The first question is what density should be used: density. The galaxy has a wind (Westmeieretal., 2005; Breitschwerdt, 2008; Everettetal., 2008; Gresseletal., Thereforetheensembleofcosmicraysandmagneticfieldcon- 2008; Otmianowska-Mazuretal., 2008) and it is getting fed stitutealightfluidpushingagainsttheheavyfluidofthenormal fromthetunnelnetworkofCox&Smith(1974)probably.This gas,andgivenenoughenergydensity,thesetwocomponentses- implies thatthe Alfve´nspeed mustapproachthe escape speed, capeviaaninstability(Parker,1965;Kowaletal.,2003,2006). for a cosmic ray driven magnetic wind. Other galaxies also Therefore all three components are strongly coupled, and the showevidenceforwinds(Chyz˙yetal.,2000a,b;Chyz˙y&Beck, dataconfirmanapproximateenergyequipartitionbetweencos- 2004;Chyz˙yetal.,2006,2007): micraysandthemagneticfield,andthesumofthesetwocom- ponentsequalinenergydensitytothegas. B V = 400km/s. (7) A 4πρ ≃ The one given parameter is the total energy input, integrated across all spatial inhomogeneities, since the energy supply is With B 3µGauss, this implies a density of about given by the stars, in the form of winds, explosions, and radi- n = 3 10≈4cm 3. − − ation. Since the inhomogeneitiesare extreme,especially in the · Thetimetostartcoolingis density,itis importantto use spatiallyintegratedenergydensi- tiesforreferenceasmuchaspossible.Sowenotethattheenergy E 2/11 density of the magnetic irregularitiesintegratedoverall spatial τc = 5 106yrs 51 (rΛ 21n 3.5)−5/11, (8) scalesisalsoinapproximateequipartition(Becketal.,1996). · n 3.5! − − − The energy input can be estimated from the explosions of su- where E is the energy of the explosion in units of 1051erg, 51 pernovaeto about1 supernovaof 1051 erg,every100years, so n is the tenuous density in units of 3 10 4cm 3, Λ is 3.5 − − 21 atL = 3 1041erg/s;theuncertaintyinthisisaboutafactor th−e cooling coefficient in units of 10 21erg· cm3s 1, and−r is a kin − − · of 3. Some fraction of this energy goes into cosmic rays. This compaction parameter of order unity. This low density reflects fractioncouldbelarge(Drury,1983). thefindingthatthetenuousmediumisofverylowdensity,and due to substructure may on volume average be of even lower Other energy input can be estimated from the infrared emis- densitythansuggestedbytheX-raydata(Snowdenetal.,1997; sion (Cox&Mezger, 1989): About 1/3 to 1/4 of the total stel- Everettetal., 2008), of order 3 10 3 cm 3; but we do use the lar radiationis absorbedby dust and reradiatedin the infrared, temperatureof105 K,nearthem·ax−imum−,andalso closetothe beyond a wavelength of 25 µ, about 1010 L . Of this, about stable region(Field, 1965). However,the coolingsuggestedby 2 109L =8 1042 erg/siscomingfromyoun⊙gstarformingre- gi·ons. ⊙ · the X-rayspectrumisan integralovertheentire evolution,and so only sensitive to the earliest part of the evolution. As soon The energy density of magnetic fields can be estimated to as two or more supernovaremnantsoverlap,and start building be about 1.6 10 12dyn/cm2 (Becketal., 1996; Everettetal., anetwork(Cox&Smith,1974),thenthetemperatureevolution − · 2008),theenergydensityofcosmicraysisaboutthesame,and will be different, giving again higher temperatures, consistent theirsumisaboutequaltothegaspressure,of4 10 12dyn/cm2. withobservations. − · Using a scale height of full width of 3 kpc, and a radius of 10 Theradiusatthatstageis kpc,we obtaina crudeestimate of the energycontent.Thisre- quires for magnetic fields and cosmic rays together an average E 3/11 supplyofenergyofabout4·1041erg/s,usingthetimescaleob- Rc = 8·102pc n 531.5! (rΛ−21n−3.5)−2/11 (9) tained from cosmic rays (see below). Such numbersare uncer- − tainbyprobablyafactorof2.Everettetal.(2008)alsoestimate andthetemperaturethenisinitially therequiredwind-powertoabout4 1041erg/s,aGalacticwind drivenbycosmicrays;theydiscuss·otherestimates. E 2/11 Td = 1.7 105K 51 (rΛ n )6/11. (10) 21 3.5 Itis interestingtonote thatto withinthe uncertaintiesall these · n 3.5! − − − power estimates (supernovae, wind power, magnetic field and This corresponds to an injection scale of turbulence. cosmicrayreplenishment)agreebetterthantheirrespectiveer- Interestingly, the time scale is of the same order of magni- rorestimates. tude to what we derive from cosmic ray transport, and the Thereforethetimescalestoreplenishanyoneofthecomponents length scale is not far from the scale height of the hot disk must also be approximately be the same. We do have the real (Snowdenetal., 1997), demonstrating qualitative consistency. numberfromradioactiveisotopesofcosmicraysinteracting,and WenotethatSnowdenetal.(1997)gaveamuchhighertemper- thenumberisabout10millionyears(Brunetti&Codino,2000, ature,ofabout4 106K, with a densityof 3 10 3 cm 3. Also, − − · · e.g.).Thisisthenthetimescaleforallkeyprocesses. Everettetal.(2008)suggestahighertemperature.However,the JuliaK.Beckeretal.:CosmicRaysVI 7 luminosity of that phase is a very small fraction of the entire Using the approachof Cox (1972) with the environmentof the dissipationintheISM. tenuous hot phase of the interstellar medium (Snowdenetal., As Field (1965) shows the cooling is stable if the temperature 1997; Everettetal., 2008) the cooling stage of an expanding dependenceofthecoolingfunctionΛis sufficientlystrongand shellofasupernovaremnantmightleadtosuchaconfiguration, its doublelogarithmicderivativepositive,or in the presence of ofaverythinshellatlargedistances,withstrongmagneticfields. heating larger than 2. This is the cooling phase we are inter- Insuchapicturethisstagewouldencompassmostofthesuper- ested in. This is the case attemperaturesbelow andnear about nova’senergydissipation,andsosimilarconsiderationsmayap- 105K. plytotheinterpretationoftheX-raydata(Snowdenetal.,1997; ∼ Datasuggestthatthebreak-upofsupernovaremnantshellshas Everettetal.,2008). beenobservedinthestarburstgalaxyM82(Barteletal.,1987). However, in that case it is not clear whether we are observing 4.4.Turbulence thebreak-upofawind-shellproducedinasnow-ploweffectby the stellar wind prior to the supernovaexplosion,or the break- Turbulence is an ubiquitous phenomenon, and also is a key upofthesnow-plowofthenormalsupernovaexplodingintothe ingredient in the interstellar medium (see reviews by Rickett interstellarmedium. (1977);Goldsteinetal.(1995)).Keyconceptstoturbulencethe- oryhavebeenintroducedbyPrandtl(1925);Karman&Howarth (1938); Kolmogorov(1941a,b,c); Obukhov(1941); Heisenberg 4.3.Magneticinhomogeneities (1948);Kraichnan(1965), andhavebeenreviewedbySagdeev Inthemagneticfielddatainourgalaxy(Becketal.,2003)there (1979).Onekeyargumentwhichwewishtouse,istheconcept isalreadystrongevidenceforsmallscalesubstructure,sincedif- oftheturbulentcascade.Theretheenergyoftheturbulenceisin- ferentmeasuresofthe magneticfield yieldverydifferentnum- jectedintothegasatsomelargewavelength,andcascadesdown bers: linear measures such as Faraday Rotation Measures indi- throughwavenumberspace,tothesmallwavelengthswherethe cate much lower strengths of the magnetic field than quadratic energy is dissipated. In many examples this leads in a three- measuressuchassynchrotronemission.Thisistypicalforsmall dimensionalisotropicmodeltotheKolmogorovcascade,which scalesubstructure(Leeetal.,2003;deAvillez&Breitschwerdt, canbedescribedinalocalapproximationbythefollowingdiffu- 2004,2007), whereforagiventotalenergycontenthighinten- sionequationinwavenumberspace(McIvor,1977;Achterberg, sitysheetscanholdalltheenergyforasmallvolumefraction.In 1979): suchapicturelinearmeasuresgiveamuchsmallernumberthan quadraticmeasures,asiswellknownfrommathematicallyiso- d I(k) 1 ∂ k4 ∂ I(k) morphicargumentsinthermalemission.Ofcourseweshouldbe = Aδ(k k ) (14) comparingtheproperintegrals,alsoinvolvingthespatialdistri- dt 4πk2 − k2 ∂k 3τk ∂k 4πk2!! − o butionofthermalelectrondensityandcosmicrayelectronden- HereI(k)istheenergydensityoftheturbulenceperwavenumber sity (see, e.g., Bowyeretal. (1995)). We ignore all this in our k,andpervolumeelement,andτ isthetimescaleofdiffusion, simpleexercise. k whichcanbewrittenas Wecanquantifythisbyintegratingalongalongthincylinderof unitlength.We refertothemagneticfieldas B ,whenitisho- 0 1 mogeneous,andfortheinhomogeneouscasethemagneticfield τ = . (15) k 1/2 is B overmost of the length, and enhancedby a factor 1/x in k γ I(k)k/ρ 1 eff a region of length x: This then gives for the integrated energy (cid:16) (cid:17) density Here ρ is the matter density, and γeff is an effective adiabatic constantfor the turbulentenergy.The turbulence has a source- 1 term,herelimitedtoasinglewavenumberk .Theturbulencedif- B2 +B2 (1 x) = B2. (11) o 1· x 1· − 0 fusionequationbasicallysaysthattheturbulencemovesthrough wavenumberspace with noadditionalsourceor sink,as a con- Keeping the integrated energy content B2 constant, the linear 0 stantenergycurrentin wavenumberphasespace (Kolmogorov, measureofthemagneticfieldisgivenby 1941a,b,c).Thesolutionstothisdiffusionequationcanbewrit- 1 tenas B x+B (1 x) = B (2 x). (12) 1 1 1 · x · · − · − Combiningthetwoexpressionsgives I(k) k2 for k k and (16) o ∼ ≤ I(k) k 5/3 for k k . x ∼ − ≥ o (2 x) (13) r1 x · − ThislatterbehavioriscommonlyreferredtoastheKolmogorov − fortheratioof linearmeasureversusquadraticmeasure.Inthe cascade,andisfoundubiquitouslyinnature. limitofsmallxthisisjust √x.Theobservationssuggestthatthis ratioisoforder1/5(Becketal.,2003),andsox=0.04byorder 4.5.Thecoolingoftheinterstellarmedium ofmagnitude.Thisimpliesthatmostofthemagneticenergyis containedinshellsofavolumeafewpercent,possiblyaslowas The interstellarmediumhasa numberof phases, whichappear 1percent.Sincethelinearmeasureisproportionaltothebending tobeinapproximatepressureequilibrium.Concentratingonthe of ultra high energy cosmic rays, this implies that the bending phaseofthehighesttemperature,thehighestspeedofsignalling is reduced by a factor between 5 and 10 over what we might (beitsoundwaves,orAlfve´nwaves,orotherwavemodes),we reasonably expect otherwise. Obviously, in realistic situations notethatitstemperatureisintherangewherethetypicalgaseous much of this effect will be smoothed out, and so perhaps even emission is detectedin the X-rayregime.The coolingcurveof moreextremesituationsmayberequired. suchagashasbeenextensivelydiscussedbymany(Cox,1972; 8 JuliaK.Beckeretal.:CosmicRaysVI Sutherland&Dopita,1993;Dopita&Sutherland,2003,e.g.).It the ratio Boron to Carbon already give information as to the shows the following features, starting at low temperature, and energy dependence of the leakage time, and such an analy- considering the cooling coefficient Λ(T) with n the interstellar sis gives an energy dependence as E 0.6 0.1 (Engelmannetal., − ± densityinparticlescm 3 1990,e.g).Wehavearguedelsewherealready(Biermann,1995; − Wiebel-Soothetal.,1995,1998),thatthisreflectstheenergyde- n2Λ(T)erg/cm3/sec. (17) pendence of the amount of target material seen for spallation. Mostofthetargetinteractionwithheavynucleiamongthecos- This cooling curve Λ(T) rises from near 104K to a lo- mic rays happens near the most massive stars, the Wolf-Rayet cal peak near 3 105K. The level of cooling along this local peak is g≈iven· by Λ 10 21ergs/cm 3/s, dropping to stars,withtheturbulenceexcitedbythecosmicraysthemselves, Λ 10 22ergs/cm 3/s nea≈r 1−06K, and−towards a mini- giving an energy dependence of the spallation secondaries of mu≈m ne−ar Λ 3 −10 23erg/c≈m 3/s in the range 3 106K E−5/9(Biermann,1998;Biermannetal.,2001;Biermann,2006), to 108K. T≈he ·peak− is due t−o many edges an≈d li·nes in consistentwith the data,which give E−0.54 (Ptuskin, 1999). On ≈ theotherhand,mostofthegammaemissionfromπ-zerodecay the soft X-ray range. At higher temperatures the continuum arisesfromtheinteractionamongthemorenumeroussupergiant emission begins to dominate and that emission is then given stars, the red supergiants, for which we suggest that the turbu- by Λ 1.4 10 27T1/2erg/cm 3/s. The sharp cutoff to low temper≃atures·nea−r 104K is du−e to beginning recombination lencearisesfrominstabilities. and thus a smaller density of free electrons to interact with. A contemporary discussion including the effects of strong 4.7.Theradioemission departures from ionization equilibrium has been given by In order to calculate the radio emission from the total number Schmutzler&Tscharnuter(1993);Breitschwerdt&Schmutzler of relativistic electronsin a columnwe first have to proceedto (1994). workoutthecosmicraylosstimeforelectrons,secondthecool- ingtimeforthetenuoushotmedium,putthemequal,andthird 4.6.Theleakyboxapproximation calculatetheradioemissionpersupernovaeventfromacolumn inthedisk. ConsideracolumnperpendiculartoagalacticdiskofheightH, Themaximumenergyintheelementaldistributionandspectrum andthenumberofcosmicrayparticlesinitasafunctionofpar- ofthecosmicraysisinprotonsneartheirrestmassenergy,since ticleenergyandtimeN(E,t);thenwehavethebalanceequation their spectrum is steeper than 2 in energy.This means that the d N(E,t) energy which has to be used in the expression for the leakage N(E,t) + = Q(E). (18) timeisfixed.So,theleakagetimeisgivenby dt τ(E) whereτ(E)istheescapetimescale.Thesignofthesecondterm H2I k τ o o . (24) ispositive,since theprocessdescribedisa loss.Ina stationary CR ∼ B5/3r2/3E1/3 statewethenobtainreadily o The maximum energy content of the population in the protron N(E) = Q(E)τ(E). (19) spectrumisherenearE =1GeV,sothat Thecharacteristictimeoflosscanbewrittenas H2I k τ (E =1GeV) o o . (25) H2 CR ∼ B5/3r2/3 τ(E) . (20) o ∼ κ inthiscase. Here,κisthediffusioncoefficientofcosmicrayparticles,which Thecoolingtimeisgivenby canbewrittenapproximatelyinthequasi-linearapproximation as nk T B2/8π 1 κ ∼ 13rgc BI2(k/)8kπ, (21) τcool ∼ n2ΛB(T) ∼ B2/8π/2kBT⋆ 2 Λ(T⋆) (26) T2 1 (cid:0) 1 (cid:1) wherewewritefortheturbulentenergydensity ⋆ ∼ Λ(T ) B2 ∼ B2 ⋆ I(k)k I k (k/k ) 2/3 I k B 2/3r 2/3E2/3, (22) ∼ o o o − ∼ o o − o− intheapproximationthatΛ(T⋆)/T⋆2 isaconstant,andassuming usingheretheassumptionofaKolmogorovspectrum.Here,r equipartitionagain.Thisisreasonableinthedissipationstageof g 1/k E/BistheLarmorradiusofaparticleofenergyE unde∼r supernova remnants when Λ approaches a double-logarithmic cons∼ideration, which gyrates in a magnetic field of strength B. derivativeof2,belowatemperatureof105K.Also,inthattem- Thebasicradiusr 1/k correspondstotheinjectionscaleof peraturerangethecoolingtimeisaminimum,foragivenenergy o o theturbulence.This∼thenleadstoadependenceofthediffusion density.Wenotedabovethatthedissipationstagereachesthose coefficientonthevariousparameters temperatures. Puttingthetwotimescalesequalthenyieldstherelation κ E1/3B5/3r2/3I 1k 1. (23) ∼ o o− o− H2I k 1 o o . (27) Therefore,weadoptthepointofviewthattheleakagetimescale B5/3r2/3 ∼ B2 τ(E)isproportionalto(relativistic)energyE 1/3,andsothatthe o − equilibrium density of energetic particles N(E) is proportional Now,wewishtoconsiderelectrons,whichemitradioemission toE−1/3aswell. atacertainfrequencyν,whichgivestheconditionthat There is a difficulty with this argument, which can be solved: The secondary to primary ratio of cosmic ray nuclei such as E ν1/2B 1/2 (28) − ∼ JuliaK.Beckeretal.:CosmicRaysVI 9 andso,makinguseofthe frequencyandmagneticfielddepen- 4.8.Whereisthelimittothediffusionlimit? denceofEqu.(26)andEqu.(28),thediffusiontimeforelectrons Thelineofreasoningseemstobeadeusexmachinainthesense whichemitatνisgivenby thatweseemtobealwaysinthediffusionlimit,independentof τCR,e ∼ H2E−1/3B−5/3ro−2/3Ioko ∼ B−11/6ν−1/6 (29) whether we have starburst galaxieswith relatively strong mag- TheemissionisfromaspectrumofelectronsofE 2.42 0.04atin- neticfields,ornormalgalaxieslikeourown. − ± jection(Biermann&Strom, 1993),andso the totalradioemis- Theequationsaboveimplythat sionS persupernovaeventinacolumninadiskcanbewritten ν as (2k T)28π 6πm c τ = B < e = τ . (31) Sν B1.71±0.02ESN,CR,eν−0.71±0.02B−11/6ν−1/6 (30) diff Λ B2 σTγeB2 syn ∼ B 0.12 0.02ν 0.88 0.02 ∼ − ± − ± We note first that both sides depend on the magnetic field forthecasethatallsupernovaremnantsproducethesamenum- strength squared, so the argument is independent on the mag- berofrelativisticelectrons.Here,weuseanintegrationalonga neticfield:Insertingnumberswefind verticalcolumntoobtainthetotalnumberofenergeticelectrons per supernovaexplosion.Thisis reasonablesince we are using (T )2 4 107 the adiabatic phase of supernovaremnant evolution, where the 5 < · , (32) Λ γ energyisconserved.Thisargumentleadstosuchaweakdepen- 21 e − dence on the strength of the magnetic field, that the resulting whereγ istheLorentzfactorofthemainlyemittingelectrons. e offsetissmallerthantheerrorsinthedata. This time, however, we need to ask what density contrast the Thereforetheradioemissionisonlyveryweaklydependenton mainradioemittingsubstructureshave,sincethelefthandside the magnetic field, independent of all other parameters, and is ofthe equationusesgrandtotalvolumeaverages,andthe right directly proportionalto the number of supernovae per time in- hand samples just those regions, where B2 is especially high. terval,andsoalsototheluminosityofmassivestars. Alreadysimplearguments,asshownabove,demonstratethatthe Thecosmicrayelectrondatashowthataboveafew10ofGeV averageofB2isquiteabithigherthantheaverageofBsquared. the spectrum is already in the loss limit, so steeper by unity Wecanassumehere,thatsuchafactormightbeoforder30,or (Kardashev, 1962). Below that energythe direct data are com- evenhigher,bringingthelimitingenergyoftheelectronsdown promisedbytheSolarwindmodulation,butfromtheradioemis- toγ <106,orpossiblyevenmuchless.Wehavealreadyargued e sionofnormalgalaxieswedoknowthattheelectronspectrumis earlierthatforthecoolingtobemaximal,thetemperaturehasto closelyinagreementwithwhatweobtainforprotons,atsome- beclosetoorbelow105K. what higher energy, allowing us to conclude that the electron Thisallowsustounderstandperhaps,whygalaxiesare usually spectrum corresponds to the diffusion limit, when the leakage inthediffusionlimit,sometimesofcourse,inastarburst,inthe outof theGalaxyis faster thanthesynchrotronloss. Using the injectionlimit:Inthatcasetheoverallspectrumcorrespondsto expressionsforthetwotimescaleswecancheck: injection. There was just not enough time to achieve diffusive Radioemissionat5GHzcorrespondstoanelectronenergyof8 approximateequilibrium. GeV,andasynchrotronlosstimeof4 107yr. · Aswenoteelsewhere,galaxiesandstarburstgalaxiesaresome- timessoyoungastobeintheinjectionlimit,orsooldastobe 4.9.Implications in the loss limit, rather than the leakage or diffusion limit that Firstofall,thistheorydoesjustwhatonewouldnaivelyexpect, we consider here. For any reasonably young age of a starburst relatethemassivestarswhichheatthedustthroughtheirultravi- there is always some low radio frequency ν , below which we 1 oletlightdirectlywiththesubsequentsupernovae.Allthetheory are still in the injection stage, below which the age of the star- does,isworkoutthenon-linearityinherentinsynchrotronemis- burst is longer than the diffusion time scale. So below that the sion, and shows them to introduce negligible dependencies on radiofluxdensityislowerthanthediffusionlimitderivedabove, variousparameters. andtheradiofluxdensityhasthespectrumofinjection,follow- ingBiermann&Strom(1993)S ν 0.71,andsorelativetothe Starburst galaxies as well as quiescent galaxies equally obey ν − diffusionlimittheradioemissiona∼tfrequencyν < ν isweaker thecorrelationbetweentheradioandthefar-infraredemission. 1 by the factor (ν/ν )1/6; this is typically not far below unity. At Thereforethetheoryimpliesbynecessitythatthemagneticfield 1 thetheotherextreme,consideredbyVo¨lk(1989)thereisalsoa inastarburstregionriseswiththeoverallenergydensity. radio frequency ν , beyond which the synchrotron and inverse Galaxieswhicharesubjecttosubstantialcompressionbyanen- 2 Compton losses cut in, and become faster then the diffusion. counterwithanothergalaxy,clearlyhaveamagneticfieldwhich Thisisdirectlyvisibleintheobservedcosmicrayelectronspec- is higher than corresponds to the energy density derived from tra (Wiebel-Sooth&Biermann, 1999). So, at radio frequencies starformation,andsomaybeexpectedtohavemoreradioemis- ν>ν theradioemissionisagainweaker,thistimeby(ν/ν ) 1/3. sionthanindicatedbythegeneralradio-far-infraredcorrelation. 2 2 − Soagaintheradioemissionisslightlydown.Bothvariantsshow, This is borne out by at least one example (Hummel&Beck, thattheradioemissionisnotfarfromthediffusiveequilibrium. 1995). This concludes the demonstration of the argument. There are Furthermore,itisclearthatforveryshorttimescales,nearorder many checks on these ideas, which one can make, such as the 107years,thecorrelationcannothold,sincewethenrunintothe pressure of the interstellar medium, the X-ray luminosity, the lifetimesofthemassivestars,whichdrivetheenergybalancein possibilitytoaccountforextremegalaxiessuchasM82,thera- theinterstellarmedium. dial gradientof the far-infrared/radioratio (Bicayetal., 1989), There is likely also a lowest level of star formation activity, the thickness of the hot gaseous disk and the associated diffu- where the assumption that the hot medium is fully connected, sioncoefficientofcosmicrays,andmanyothers.Wewilldiscuss fails.Thereonemayexpectalsosubstantialdeparturesfromthe thesepointselsewhere. correlation. 10 JuliaK.Beckeretal.:CosmicRaysVI 5. Cosmicraysandtheirsecondaries Spectral indices 5 GHz < ν < 1.4 GHz (z<=0.03, S60µm/S >=30, S60µm>=4 Jy, S >=20mJy) Radio 1.4GHz As stated above,the cosmic ray intensity from starburst galax- nt ies scales with the radio and infrared emission of the sources. u16 o In this section, we discuss the emission scenarios for charged C14 cosmic rays and hadronic interactions leading to high-energy 12 pwhitohtionnsatanrdbunresutstritnhoatecmanissaicocne.leTrhateerecoasrme itcwroayssoutrocehicglhasesens- 10 loss leakage injection ergies, namely shock fronts of supernova remnants and long 8 Gamma Ray Bursts (GRBs), the latter being connected to su- pernovaIc explosions.In the first case, maximum energiesare 6 limitedtolessthan1015eVandthus,thecosmicraysfromstar- 4 bursts cannot be observed directly due to the high cosmic ray 2 backgroundinourownGalaxy.GammaRayBursts,ontheother hand,wereproposedastheoriginofcosmicraysabovethean- 0 -2 -1.5 -1 -0.5 0 0.5 kle, i.e. ECR >3 1018eV, see Vietri (1995); Waxman (1995). α · Sinceahighstarformationrateasitispresentinstarburstgalax- ies,leadstoahighratesupernovaexplosions,anenhancedrate Fig.7. Histogram of the radio spectral indices of 105 sources oflongGRBsisexpected.Thus,forclosebysources,thedistri- between1.4GHzand5GHz.Theareasbetweenthedashedlines butionofstarburstgalaxiescanbeusedtotestthehypothesisof indicatesourcesintheloss,leakageandinjectionlimit(fromthe cosmicraysfromstarbursts,asalsodiscussedinBiermannetal. left).Asaconservativeestimate,weincludethosethreesources (2008). withextremelysteepspectraaslosslimitsources.Thosesources Thedominantsourceofsecondarycosmicrayslikehigh-energy withextremelyflatspectra,totherightoftheinjectionlimitarea, photonsandneutrinosisproton-protoninteractionsindensehy- aredominatedbyabsorptionorfree-freeradiation. drogen regions and proton-photoninteractions in Gamma Ray Bursts.Proton-protoninteractionsproducepionsvia canreachdifferentvalues.Shockaccelerationofchargedparti- pp π+π π0. (33) − clesusuallyresultsinprimaryelectronspectraof → PthheoDtoehlatadrroensoicnainntceer,actions, on the other hand, yield pions via ddNEe ∝ Ee−αperim, (36) e pγ ∆+ nπ+/pπ0. (34) withαprim 2.0 2.4.Iftheelectronsescapebeforeinteracting → → e ≈ − with the ambient medium, the primary spectrum stays unmod- High-energyphotonsandneutrinosaresubsequentlyemittedin ified, αprim =αsec, referredto as the injection limit. If the elec- π resp.π0 decays: e e ± tronsare partlyscattered downto lowerenergies,the spectrum − − ofsecondariessteepenstoαsec 2.5 2.8fora primaryspec- π+ → µ+νµ →e+νeνµνµ trumwithαprim 2.0 2.4.eTh≈isisca−lledtheleakagelimit.In π− → µ−νµ →e−νeνµνµ the case of ecalor≈imetri−c sources, basically the entire energy is π0 γγ. (35) lostinthesourceandthespectrumofsecondaryelectronsisas → steepasαsec 3.2 3.4.Thisscenarioisreferredtoastheloss e ≈ − limit. 5.1.Supernovaremnants AsdiscussedinRybicki&Lightman(1979),theelectronspec- Cosmic Rays are believed to be produced in young super- tralindexαseeccorrelateswiththeindexofsynchrotronradiation nova remnants (SNRs), reaching maximum energies of around as αsec 1 1015 eVorabove,dependingontheirlocalenvironmentandon α= e − . (37) thecosmicraycomposition.Theproductionofsecondariesfrom − 2 hadronicinteractionsdependsontheproton-protonopticaldepth Thus,theobservedsynchrotronspectralindicesare intheSNRenvironment.Inthissection,wepresentamodelof 0.5 0.75 inthe injectionlimit whichsourcesareopticallythinandwhich,incontrast,aregood − →− ctrainndoisd.atesfor the productionof high-energyphotonsand neu- α=−−10..075→→−−1.12.0 iinntthhee lleoasskaligmeilti.mit (38) Thesynchrotronspectralindicesinthissamplescatterbetween 1.7<α<0.7 with a peak at α 0.7. We exclude sources 5.1.1. Opticaldepth − ∼− that may include contributions other than synchrotron radia- We use the observation of synchrotron radiation from shock- tion, i.e. those sources that have spectral indices α> 0.5. − accelerated electrons, assuming that electrons and hadrons are Here, absorption is likely to have modified the spectrum. accelerated in the same shock environment. Figure 7 shows Alternatively, spectral indices around -0.1 point to free-free the distribution of spectral indices at radio wavelengths be- radiation (Mezger&Henderson, 1967). Of the remaining 85 tween 1.4GHz and 5GHz for 105 sources in the sample with sources, 37 sources (44%) starbursts are in the injection limit, givenspectralindicesattherequiredwavelengths.Thespectrum 36sources(42%)areintheleakagelimitand12sources(14%) at these energies is produced by electron synchrotron losses. areinthelosslimit. Depending on the shape of the primary electron spectrum and The observation of a synchrotron spectrum following the in- scattering effects, the spectral index of the electron population jectedelectronspectrumhassevereimplicationsforhadronsin