MAGNETI C FIELDS IN NORMA L GALAXIESAND PROSPECTS FOR THE SQUARE KILOMETER ARRAY

R. BECK

Max-Planck-Institutfur RadioastronomieAuf dem Hugel69

53121BonnGermany

E-mail: rbeck@mpifr-bonn.mpg.de

Grand-design, flocculent and even irregular galaxieshost interstellar magnetic fields with a wellorderedspiral structure. In grand-designgalaxiesthe fieldsarealignedparallel to theoptical spiralarms, but the strongest regular fields are found in interarmregions, sometimesforming “magneticspiral arms” betweenthe optical ones. Magneticfields in the optical spiral armsare strong butirregular, dueto tangling by processesrelatedto star formation. Faradayrotation of thepolarizationvectorsshows patternswhich supporttheexistenceof coherentlarge-scale fields in galactic disks,signaturesof dynamoaction. In barredgalaxies the magnetic field seemsto follow the gasflowwithin the bar. However, the location of the shock front in the magnetic field deviatesfrom thatexpectedfrom hydrodynamicalmodels. Within (andinterior to) thecircumnuclearring the field isagain of spiral shapewhich leadsto magnetic stresses, possibly drivinggasinflow towardstheactivenucleus. TheSquareKilometer Array should be able to revealthe wealth of magnetic structuresingalaxies.

1 Intr oduction

Linearlypolarizedradiocontinuumemissionis apowerful tool to studythestrengthand structureofinterstellarmagneticfields in galaxies.Cosmic-ray electronsspirallingaroundthefield linesemitsynchrotronradiation,thedominantcontribution to radiocontinuumemissionat centimeteranddecimeter wavelengths.If cosmic-rayelectronssuffer strongenergy lossesvia synchrotronemission,synchrotronintensitydependsmainly on theproductionrateof cosmic rays[1]. In suchacasethesynchrotronradiospectrumshowsasteepeningwith decreasingwavelengthwhich,at centimeterwavelengths,isobserved only in a few galaxies,e.g. NGC 2276with its exceptionallystrongmagneticfield [2]. In asample of 74spiral galaxiesno tendency of thesynchrotronspectralindex � to steepenwith increasingmagneticfield strengthwasfound[3]. The averagevalue is � � � � � � � � � � �

, with astandarddeviation of 0.13 [4]. In conclusion,most galacticspectraarenot significantlyaffectedbysynchrotron(or inverseCompton)losses,andsynchrotronintensitydependson thetotal fieldcomponent� � in theplane ofthesky with at leastthepower ( � � ). Thedependenceis stronger( � � � � � � ) if equipartitionbetweencosmic-ray and magneticfield energy densitiesholds.Equipartitionis probablyvalid onscalesof a few kiloparsecs,but may beviolatedon smaller scales[5] and instronglyinteractinggalaxies[2].

A mapof thetotal radiointensity(Fig. 1) isamap of thetotal interstellar magneticfieldsilluminatedbycosmic-ray electrons, polarizedintensity(Fig. 2a) revealstheresolvedregular field. Small-scalevariationsin radiointensityaremainly dueto variationsin field strength.Thevisibility of fields islimited by thediffusionlengthof cosmic-rayelectronsof a few kiloparsecsfrom their birthplacesinstar-forming regions. Thusmagneticfields maystill bestrongfar away from radio-brightregions.Using Faradayrotationmeasuresof backgroundsources,regularfields canbedetecteduntil largedistancesfrom a galaxy’s centerif ahaloof ionizedgasexists. In M31, for example, theregularfield isstronguntil at least25 kpc radius[6].

M.P. van Haarlem (editor)Perspectiveson RadioAstronomy – Sciencewith Large AntennaArraysNetherlandsFoundationfor Researchin Astronomy – 1999

M83 6.2cm Total Intensity + B-vectors (VLA+Effelsberg)

DE

CL

INA

TIO

N (

J200

0)

RIGHT ASCENSION (J2000)13 37 20 15 10 05 00 36 55 50 45 40

-29 48

49

50

51

52

53

54

55

56

Figure1: Total radio emission of M83 and � –vectorsof polarizedemission at � � � � cm (VLA, � � � � synthesizedbeam),combinedwith the extendedemission observed with the Effelsberg 100-mtelescope( � � � � resolution) andsuperimposed onto an optical imageof D. Malin (AAO). The length of � –vectors is proportional to polarizedintensity. Faradayrotationhasnotbeencorrectedasit is below � � ¦ (Beck,Ehle,Sukumar& Allen, in prep.)

2 Depolarization and Faraday rotation

Synchrotronemissionis highly linearly polarized,intrinsically70–75%in a completelyregularmagneticfield. Theobservable degreeof polarizationin galaxiesis reducedby Faraday(wavelength-dependent) depolarizationin magnetizedplasma clouds,by geometrical(wavelength-independent) depolarizationdueto variationsof themagneticfield orientationacrossthetelescopebeamandalong theline of sight, and bya contribution of unpolarizedthermal emission(onaverage10–20%atcentimeterwavelengths,up to 50%locally). Typical fractionalpolarizationsingalaxiesarelessthana few percentin centralregionsandspiral arms,20–40%in betweenthespiralarmsandin outerregions.Thus,polarizedradiointensitiesare weak,and only thelargesttelescopesaresufficiently sensitive to detectthem. TheEffelsberg 100-msingle-dishtelescopeprovidesanangularresolutionof � � � �

at � � � � cmand� � � � at � � cm(Fig. 5). Synthesistelescopes(VLA, ATCA, WSRT) offer

higherangularresolutionbut misslarge-scalestructuresin extendedobjectslike nearbygalaxies.Missingflux densityin StokesQ and U mapsleadsto wrongpolarizationangles.Combinationofsingle-dishandsynthesisdatain all Stokesparametersis required(Figs.1, 2a and 4).

The � –vectorsof linearly polarizedemissionjust indicateanisotropyof themagneticfield distributionin theemissionregion. Imaginethatamagneticfield without any regularstructure(anisotropicrandomfield) is compressedin onedimensionby ashock.Emissionfrom theresultinganisotropicfield islinearly polarizedwith ordered� -vectors,but thefield is incoherent, i.e. it reversesits directionfrequentlywithin thetelescopebeam. Faradayrotation measures ( � ) are essentialto distinguishbetweencoherentandincoherentfields. Thesign of � gives thedirectionof thefield componentalongtheline of sight.

250

NGC6946 6.2cm Polarized Intensity + B-vectors (VLA+Effelsberg)

DE

CL

INA

TIO

N (

B19

50)

RIGHT ASCENSION (B1950)20 34 30 15 00 33 45 30 15

60 04

02

00

59 58

56

54

NGC6946 18cm Polarized Intensity HPBW=15" (VLA)

DE

CL

INA

TIO

N (

B19

50)

RIGHT ASCENSION (B1950)20 34 30 15 00 33 45 30 15

60 04

02

00

59 58

56

54

0 50 100 150 MicroJy/beam

Figure2: 2a (left) Polarizedradio emission of NGC 6946at � � � � cm (VLA, � � � � synthesizedbeam),combinedwith the extendedemissionobserved with theEffelsberg 100-mtelescope( � � � � resolution)andsuperimposedontoanopticalimagefrom thePOSS.Thelengthof � –vectorsisproportionalto polarizedintensity (from [12])2b (ri ght) Polarizedradio emission of NGC 6946at � � ! � � cm (VLA, � � � � synthesizedbeam).Comparison withFig. 2arevealsregionsof Faradaydepolarization,especially in thesouthernandwestern partsof thegalaxy (Beck,in prep.)

At centimeterwavelengthsFaradayrotationanglesof thepolarizationvectorsvary with � " . Typicalinterstellarrotationmeasuresof # 50 $ % & ' ( " leadto � � ¦ rotationat � � cm and * ¦ at � * cm. Belowabout � * cm Faradayrotationis so small thatthe � –vectors(ie. theobserved + –vectorsrotatedby , � ¦ )directly tracetheorientationof theregularfield in thesky plane.To detectsmall rotationmeasures,observationsat thelargestpossiblewavelengthin theFaraday-thinregime arerequired,e.g. � - � * cm.Such systems are availableat theEffelsberg, ATCA and WSRT telescopes,but not yet at theVLA.

At decimeter wavelengthsFaradaydepolarizationsignificantlyaffectsthepolarizedradioemission[7,8]. DifferentialFaradayrotationalong theline of sight leadsto zeropolarizedintensity(“Faradayshadows”) at certainwavelengthswheretheobserved rotationof thepolarizationanglereachesmultiples of , � ¦ . Hence,filamentsof vanishingpolarizedintensity, accompaniedby , � ¦ jumpsin polarizationangleacrossthefilament,do not indicateregionswith vanishingelectrondensityorvanishingfieldstrength.

Polarizationsurveys in our Galaxy have revealedvariousfeaturesin theforegroundFaradayscreen[9–11]. This opensanew window to study thesmall-scalefield structures.However, truefeatureshave to bedistinguishedfrom Faradayshadows by mappingin several nearbyfrequencychannels.For example,somesmall featureswith zeropolarizedintensityin NGC 6946at � � � cm(Fig. 2b) disappearat � � �

cm andhenceareonly “shadows”.

Randomfields cause“Faradaydispersion”,and only polarizedemissionfrom an upperpart of thediskor thehalo isdetected.Largeregionsof agalaxycanbedepolarizedcompletely (Fig. 2b). HereFaradayrotationanglesdo nolongervary linearly with � " and � is no longerproportionalto theregularfieldstrength.Theobserved � may even changeits signwithout a reversalin field direction[7].Nevertheless,polarizationdataat longwavelengthscontainvaluableinformations: Observationsatseveral nearbywavelengthstracedifferentlayersof thegalaxyandthusallow galactic tomography.(Note,however, thatthethicknessof theobservable layerincreaseswith frequency.)

251

3 Magnetic field strengths

The averagestrengthof thetotal . � � / andtheresolvedregularfield . � 0 � / componentsin theplaneof thesky canbederived from thetotal and polarizedradiosynchrotronintensity, respectively, ifenergy-densityequipartitionbetweencosmic raysandmagneticfields or minimum total energy densityis assumed. Thestandardminimum-energy formulaegenerallyuseafixedintegrationinterval in radiofrequencyto determine thetotal energy densityof cosmic-ray electrons.Thisproceduremakesitdifficult to compareminimum-energy fieldstrengthsbetweengalaxiesbecauseafixed frequencyinterval correspondsto differentelectronenergy intervals, dependingon thefield strengthitself. Wheninsteadafixed integration interval in energy is used,theminimum-energy andenergy equipartitionestimatesgive similar valuesfor . � � � � � / , where � is thesynchrotronspectralindex (typically # � � , ).

The resultingestimateof . � � � � � / 1 2 3 � � � 4 is larger thanthemeanfield . � � / if thefieldstrengthvariesalongthepathlength.If, on theotherhand,thefieldhasavolume filling factor 5 of smaller than1, theequipartitionestimateis smaller thanthefield strengthin thefilamentsby a factor 5 1 2 3 � � � 4 [13].Nothing isknown about 5 yet.

Themeanmagneticfield strengthfor thesample of 74 spiral galaxiesis . � � / � , 6 G (0.9 nT) with astandarddeviation of * 6 G [14]. In nearbygalaxiestheaveragetotal fieldstrengthsin thegalacticplane(correctedfor inclination)rangebetween . � / # 7 6 8 in M33 [15] and # � � 6 8 in M51 [16]. In spiralarms thetotal field strengthscanreach# � � 6 8 locally, like in NGC 6946[17], M51andM83.Interactinggalaxieshosteven strongermagneticfields,probablylargerthantheequipartitionvalues[2].Thestrongestfield within anormal galaxyfoundsofar is that in thecircumnuclearring of NGC 1097with � # 7 � 6 8 [18]. Thestrengthsof theresolvedregularfields � 0 aretypically 1–5 6 8 locally, but# � * 6 8 in aninterarmregion of NGC 6946(Sect.6); thesearealwayslower limits dueto thelimitedangularresolution.

4 Magnetic fields and gasclouds

Comparisonof themapsof thetotal radioemissionof M51 [19] andthetotal (cold + warm) dustemission[20] revealsasurprisinglycloseconnection.To understandits origin it is crucial to considerthestrong(probablydominant) influenceof thefieldstrengthon radiointensity. Magneticfields areobviously anchoredin gascloudswhich are tracedby thedust.Remarkably, onedustlanecrosses theeasternspiral arm of M51,andso doesthetotal field. Furthermore, thetotal radioandfar infraredluminositiesof galaxiesare tightly correlated.The correlationcanbeexplained,globally andlocallyandwith thecorrectslope,by aclosecouplingof magneticfields to gasclouds[3,5]. The radio–far-infaredcorrelationwithin M31 indicatesthatthecoupling isvalid evenfor themore diffusegasmixed with cool dust[5]. Thedetailedcomparisonbetweenthetotal synchrotronintensityandthecoolgas( 9 : � 9 " ) in aspiralarm of M31 confirmeda couplingof themagneticfield to thegas[21]. Thetotal field strength � is high in spiralarmsbecausethegasdensityis highestthere.

The correlationbetweengasand regular fields is lessobvious. Long,prominentdustlanesareoftenconnectedto featuresof regularfields,e.g. in M83(Fig. 1), in theanomalousarmof NGC 3627[22]andin flocculentgalaxieslike NGC 4414(Fig. 3). On theotherhand,regularfields are sometimesobserved in interarmregionswith very little gasor dust(Sect.6).

5 Magnetic field structur e

Grand-designspiralgalaxiesare shapedby densitywaves,but therole of magneticfields isyetunknown. Strongshocksshouldcompressthemagneticfield andincreasethedegreeof radiopolarizationon theinneredgesof thespiralarms. Radio observations,however, show a larger variety ofphenomenae.

The total radiointensityshows thetotal (=regular+random) field, thepolarizedradiointensitytheresolvedregularfield only. Thestrongesttotal and regularfields in M51arefoundat thepositionsof theprominentdustlaneson theinneredgesof theopticalspiralarms [19], asexpectedfrom compression

252

NGC4414 3.5cm Polarized Intensity + B-vectors (VLA)

DE

CL

INA

TIO

N (

B19

50)

RIGHT ASCENSION (B1950)12 24 04 02 00 23 58 56 54

31 31 00

30 30

00

29 30

00

28 30

Figure3: Polarizedradio emissionof theflocculent galaxy NGC 4414at � ; � � cm(VLA, � � � � synthesizedbeam),superimposed onto an optical H < imageobtainedby C. Horellou. The length of � –vectors is proportional topolarizedintensity (Soidaet al., in prep.)

by densitywaves,but theregularfields extendfar into theinterarmregions. In M83(Fig. 1) thetotalandpolarizedemissionpeakon theinneredgeof thenorthernopticalarm; in thesouthernarmthetotalemissionshows noshift with respectto theopticalarm, while thepolarizedemissionis stronglyshiftedinto theinterarmregion betweenthesouthernarmand thebar. NGC 1566[23] andM81 [24] showalmost no signsof field compression;their strongestregularfields occurin interarm regions, while thetotal field is still highestin theopticalspiralarms. Field tanglingin thespiralarms,e.g. duetoincreasedturbulentmotionsof gascloudsandsupernova shockfronts,may explain this result[25]. Insome cases,however, theinterarmfieldsareconcentratedin “magneticarms” which cannotbeexplainedby thelack of field tangling(seeSect.6).

Radio polarizationobservationsshow thatthe � –vectorsof theregularfields largely follow theopticalspiralstructurein M51 [16,19],M81 [24], M83 (Fig. 1) and NGC 1566[23], thoughgenerallyoffsetfrom theopticalarms. In thedensity-wave picturethemagneticfield is frozeninto thegascloudsandistransportedby thegasflow. Thusthefield orientationshouldreflectthestreaming linesof thegas,notthestructureof thespiralwave itself. Thepitch anglesof thestreaming linesaresmall in interarmregionsandlarger in spiral arms (thoughstill smaller thanthatof thespiralwave). However, theobserved pitch anglesof theregularfield arelargerthanthoseof thestreaming linesalmost everywhere.In theinterarmregionsof NGC 6946thefield pitch angleis # � � ¦ [26], while thegasflow is almostazimuthal. Hence,theregularmagneticfield is not frozeninto thegasflow, but probablymodified byturbulentdiffusion[27] and/orshapedby dynamo action(Sect.7).

Regularspiralmagneticfields with strengthssimilar to thosein grand-designgalaxieshave beendetectedin flocculentgalaxies(Fig. 3) and evenin irregulargalaxies(Fig. 4). Themeandegreeofpolarization(correctedfor differentspatialresolutions)is similar betweengrand-designandflocculentgalaxies[28]. Apparently, densitywaves have arelatively small effect on thefieldstructure.

In our Galaxy several field reversalsbetweenthespiral arms,on kpc scales,have beendetectedfrompulsarrotationmeasures[29,30]. Polarizationobservationsof some externalgalaxieshave sufficiently

253

NGC4449 6.2cm Polarized Intensity + B-vectors (VLA+Effelsberg)

DE

CL

INA

TIO

N (

J200

0)

RIGHT ASCENSION (J2000)12 28 30 25 20 15 10 05 00 27 55 50

44 09

08

07

06

05

04

03

02

Figure4: Polarizedradio emission of the irregular galaxy NGC 4449at � � � � cm (VLA, � = � � synthesizedbeam),combinedwith the extendedemission observed with the Effelsberg 100-mtelescope( � � � � resolution), and super-imposedonto an optical H < imageobtainedby D. Bomans. Thelength of � –vectorsis proportionalto polarizedintensity (from [31])

high spatialresolution,but similar reversalshave not yet beendetectedin themapsof rotationmeasures,e.g. within themain emission“ring” of M31 (Fig. 5). For othergalaxieslike M51,M83andNGC 6946theevidenceagainstreversalsisweaker but still significant.Field reversalsmayoccurpreferablyin galaxieswith lessorganizedspiralstructure.Anotherexplanationis that pulsarRMs in theGalaxy tracethefield neartheGalacticplanewhile RMs in externalgalaxiesshow theaverageregularfield alongthepathlengththroughthe“ thick disk” (seeSect.8).

Thereis increasingobservational evidencethatmagneticfields are importantfor theformationof spiralarms. Thestreaming velocity anddirectionof gascloudsandtheir collision ratescanbemodified.Furthermore,magneticfields will alsoinfluencethestar formation ratesin spiralarms. Magneticfieldsareessentialfor theonsetof starformationas they allow to removeangularmomentumfrom theprotostellarcloud duringits collapse.

6 Magnetic spiral arms

Longarmsof polarizedemissionwerediscovered in IC 342[32,33]. Observationsof anothergas-richgalaxy, NGC 6946[12], revealedasurprisingly regulardistribution of polarizedintensitywith two“magneticarms” locatedin interarm regions,without any associationwith cool gasor stars,runningparallelto theadjacentopticalspiralarms (Fig. 2a).Thesemagneticarms do not fill theentireinterarmspaceslike thepolarizedemissionin M81,but are only # � � �

–1000pc wide.Thefields in themagneticarmsmust be almost totally aligned, and thepeakstrengthof theregularfield is # � * 6 8 . Magneticarmshave alsobeenfoundin M83 (Fig. 1,southof thebar)andin NGC 2997[34].

Themagneticarms cannotbeartifactsof depolarization.Firstly, their degreeof polarizationisexceptionallyhigh (up to 50%). Secondly, they look quitesimilar at � � cm and � * cm. Thirdly, they arealsovisible as peaksin total emission,which excludestheir existencesolely dueto awindow ingeometricaldepolarization(small field tangling).

254

M31 6.3cm Total Intensity + Magnetic Field (Effelsberg)

DE

CL

INA

TIO

N (

B19

50)

RIGHT ASCENSION (B1950)00 45 44 43 42 41 40 39

42 00

41 55

50

45

40

35

Figure5: Total intensity of M31 at � � � ; cm(Effelsberg100-mtelescope,smoothedto ; � resolution),andmagneticfield orientationsat � � resolution. Faradayrotation hasbeencorrectedwith help of Effelsberg data at � � � � � cmwith � � resolution. Thelength of vectorsis proportionalto polarizedintensity (Berkhuijsen,Beck& Hoernes, inprep.)

Westill donot understandhow magneticarmsaregenerated.It wasproposedthatthey couldbemanifestationsof slow MHD waveswhich maypropagatein a rigidly rotatingdisk, with themaxima infield strengthphase-shiftedagainstthosein gasdensity[35]. However, all galaxieswith magneticarmsrotatedifferentiallybeyond 1–2kpc from thecentre.Some correlationexistsbetweenthemagneticarmsandinterarmgasfeaturesgeneratedin numericalmodelsof perturbedgalacticdisks[36]. However,suchmodelsneglect theeffectof magneticfields. In dynamo models,usingthereasonableassumptionthatthedynamo numberis largerbetweentheopticalarms thanin thearms [37], magneticarms evolvebetweentheopticalarms in adifferentially rotatingdisk [26,38,39]. However, theback-reactionof thefield ontothegashasnot beenconsideredyet in present-daydynamo models.In themagneticarms theenergy densityof thefieldmay exceedthatof thelarge-scalegasmotion and thusdistort thegasflow.

7 Faraday rotation and dynamos

Regularmagneticfields could in principle be shapedby gasflowsanddensitywaves.Faradayrotationmeasures( � ) are essentialto distinguishbetweencoherentandincoherentfields. In anincoherentfield theRMsarerandomandshow no large-scalestructure.Observation of RM coherencyon a largescale,like in M31 [6,40], NGC 6946[41] andNGC 2997[34], meansthatthefield wascoherentalreadybeforecompression,andhencetheremustbe anotherphysicalmechanism(dynamo orprimordial origin) to generatesuchan orderedfield. Therole of densitywaveswould thenbe restrictedto thealignmentof thelarge-scalecoherentfield with thespiral arms.

Thestrongestevidencefor dynamo actioncomesfrom M31 (Fig. 5). Radio observationsof M31revealeda20kpc-sizedtorusof magneticfields alignedin a single direction[40]. Only thedynamo isableto generateaunidirectionalfieldof suchdimensions.RMs from polarizedbackgroundsourcesconfirmedthis picture[6]. Theregularfield existsalsointerior to theprominent “ ring” andextendsoutto at least25 kpc radius.

Dynamos are promising candidatesto generatecoherentfields,evenin galaxieswithout densitywaves.The linearmean-field dynamo [13,42] generatesmagneticfield modeswhich have spiralstructuredueto their azimuthal andradialfieldcomponents.The pitch angleof thefield spiraldependson thedynamo number, not on thepitch angleof thegasspiral. Thefield structureisdescribedby modesofdifferentazimuthal andverticalsymmetry; in generalasuperpositionof modesis generated.Alarge-scalepatternin mapsof Faradayrotationmeasures(RMs) revealsthedominanceof asingledynamo mode[43]. A single-periodicazimuthal RM variation (with aphaseequalto thepitch angleofthespiral structure)indicatesadominatingaxisymmetric dynamo mode(ASS, > � �

), asin

255

M31 [6,40] andIC 342[32]. Double-periodicazimuthal RM variationsindicateadominatingbisymmetric dynamo mode(BSS, > � � ) if their phasesvary with radialdistanceasexpected[43], as isthecasefor M81 [24] andpossiblyM33 [15]. TheinteractinggalaxyM51 is aspecialcase.Analyzingall available polarizationangledata,thefield in M51canbedescribedasmixed modes(MSS), withaxisymmetric andbisymmetric componentshaving aboutequalweightsin thedisk, togetherwith ahorizontalaxisymmetric halofield with oppositedirection[44]. Themagneticarms of NGC 6946maybetheresultof asuperpositionof theASSand thequadrisymmetric ( > � �

) modes,while theBSSmodeis suppressedby thetwo-armedspiralstructureof thegas[26]. In many othergalaxiesthedataarestill insufficient to allow afirm conclusionwhetherthelarge-scalepatternof theregularfield is evenmore complicatedor thedistribution of thermal gasis non-axisymmetric sothattheRMsaredistorted.

Thesimilarity of pitch anglesbetweenthedynamo-wave and thedensity-wave spiral is not self-evidentandindicatestheexistenceof some interactionbetweenthem. Futuredynamo modelshave to includedensitywaves and theback-reactionof thefield.

By comparingthesignsof theRM distribution and thevelocity field, inwardand outwarddirectionsoftheradialcomponentof thespiral magneticfield canbedistinguished.Surprisingly, all known ASSfields (M31, IC 342,NGC 253)andtheMSS field in NGC 6946point inwards. Dynamo actiondoesnotpreferonedirection.This indicatessome asymmetry in theinitial seedfield andexcludessmall-scaleseedfields [41], possiblyacosmologically relevant result.A largerdatabaseis needed.

8 Edge-ongalaxiesand radio halos

NGC 891,NGC 5907andNGC 7331andotheredge-ongalaxiespossessthick radio diskswith # � kpcscaleheights.In thesegalaxiestheobserved fieldorientationsaremainly parallel to thedisk [45].NGC 4565hasthemost regularplane-parallelfield [46]. Bright, extendedradiohalos(with scaleheightsof severalkpc) are rare.NGC 253is theedge-ongalaxywith thebrightestandlargesthaloobserved so far [47]. Theirregularappearanceof theNGC 253halois mainly dueto thelowersensitivity comparedwith themapof NGC 4631.The regularmagneticfield in thedisk of NGC 253isalsopredominantly parallelto theplane[48] which may bedueto strongdynamo actioneven closetothecentre.Someradiospurswith vertical field linesemergefrom theouterdisk.

NGC 4631,NGC 4666andM82 arehalo galaxieswith dominatingverticalfield components[49–52].Magneticspursin thesehalosare connectedto star-forming regionsin thedisk. Thefield is probablydraggedoutby thestrong,inhomogenousgalacticwind. Evidencewasfoundfor adirectdependenceofthehalo extenton thelevel of energy inputfrom theunderlyingdisk [53]. Themagneticfield linesintheNGC 4631halohave adipolar structure(Fig. 6) in theinnerdisk wheredifferentialrotationisweaksothatthedipolar (antisymmetric) dynamo modecanevolve. A few regionswith field orientationsparallelto thedisk arevisible in the(differentiallyrotating)outerdisk. Mapsof rotationmeasuresarerequiredto testthedipolar model.

In theplanesof edge-onspiral galaxiestheobserved polarizedemissionis weakdueto depolarizationeffects.Faradaydepolarizationaloneis insufficient to explain thelow degreesof polarizationneartheplaneof NGC 4631[50]. Thefield structurein theplaneismostly turbulentdueto star-formingprocesses,causingdepolarizationalong theline of sightaswell asacrossthetelescopebeam[45]. Thedegreeof polarizationat high frequenciesincreaseswith increasingdistancefrom theplanebecausethestar-forming activity andthusthefield turbulencedecrease.

9 Barr ed galaxies

Gasand starsin barredgalaxiesmove in highly noncircularorbits.Gasstreamlinesarestronglydeflectedin thebar region alongshockfronts,behindwhich thegasiscompressedin a fastshearingflow [54]. As thegasin thebarregion rotatesfasterthanthebar, compressionregionstracedby massivedustlanesdevelop along theedgeof thebar that is leadingwith respectto thegalaxy’s rotation.Gasinflow along thecompressionregion may fuel starburstactivity in adensering nearthegalacticcentre,althoughit is not clearhow thegascangetrid of its angularmomentumbeforefalling into theactive

256

NGC4631 22cm Total Intensity + Magnetic Field HPBW=70" (VLA)

DE

CL

INA

TIO

N (

B19

50)

RIGHT ASCENSION (B1950)12 40 15 00 39 45 30 15 00

32 54

52

50

48

46

44

Figure6: Total radioemission and � –vectorsof polarizedemission of NGC 4631at � � � cm (VLA, ? � � � syn-thesizedbeam).The � –vectorshave beencorrectedfor Faradayrotation, their length is proportionalto polarizedintensity (Krause et al., in prep.)

nucleus.Theeffectsof magneticfields on gasflows in barredgalaxieshave not yet beenaddressedinmodels.

¿From asample of galaxieswith strongopticalbarsobserved with theEffelsberg, VLA and ATCAtelescopes,thestrongestregularmagneticfieldsweredetectedin NGC 1097,NGC 1365,NGC 4535andNGC 7479.NGC 1097andNGC 1365arebarredgalaxiesof morphologicaltypeSBbcwith theirbarslying almostin theplaneof thesky so thatspectroscopicobservationsof theshearinggasflow arevery difficult.

Thegeneralsimilarity of the � –vectorsin NGC 1097(Fig. 7) and gasstreamlinesaroundthebar asobtainedin simulations[54] is striking. This suggeststhattheregularmagneticfield is alignedwith theshearingflow. Weobserve ridgesof enhancedtotal magneticfieldswhich coincidewith theopticaldustlanes.Theupstreamanddownstreamregionsof enhancedpolarizedemissionareseparatedby a strip ofzeropolarizedintensity, thelocationof theshockfront, wheretheobserved � –vectorschangetheirorientationabruptly (Fig. 7). This large deflectionangleleadsto geometrical depolarizationwithin thetelescopebeambecausethestrip wherethefield changesits directionis narrower thanthespatialresolutionof our observations. Our polarizationobservations imply thattheshockfront in themagneticfield is located700–900pc in front of thedustlanes,in contrastto conditionsin classicalshocks.Furthermore, the degreeof field alignment is largestupstream(with adegreeof polarizationof up to50%, right half of thebar in Fig. 7), not downstream. This indicatesthattheshockgeneratesfieldturbulence.Numerical modelsincludingmagneticfields are required.Strongdeflectionsof the� –vectorsare alsoobserved in thebarof M83 (Fig. 1) and in NGC 1672(Fig. 8).

The circumnuclearring of NGC 1097is asite of ongoingintensestar formation, with anactive nucleusin its centre.Thelocal equipartitionstrengthsof thetotal andregularmagneticfields are � # 7 � 6 Gand � 0 # @ 6 G in thering. Thefield strengthreachesits absolutemaximum wherethecompressionregion intersectswith thering. The regularfield swingsfrom alignmentalong thebar to a spiral patternnearthering [18]. In contrastto thebar, conditionsfor dynamo actionare ideal in thering. Theorientationof theinnermostfield agreeswith thatof thespiral dustfilamentsvisible in theopticalHSTimage.Magneticstressin thecircumnuclearring candrive massinflow to feedtheactive nucleusinNGC 1097.

257

NGC1097 3.5cm Polarized Intensity + B-vectors (VLA)

DE

CL

INA

TIO

N (

J200

0)

RIGHT ASCENSION (J2000)02 46 24 23 22 21 20 19 18

-30 16 15

30

45

17 00

15

30

45

18 00

Figure 7: Total emission and � -vectors of polarizedradio emission of the southernhalf of the barredgalaxyNGC 1097at � ; � � cm(VLA, � � � � synthesizedbeam),superimposedonto an optical imageof H. Arp (MPE/CerroTololo). The length of � –vectorsisproportionalto thedegreeof polarization. (from [18])

Our resultshave revealedaprincipal differencebetweenthebehavioursof magneticfields in barredandnon-barredgalaxies.In barsthemagneticfield appearsto interactstronglywith thegasflow. Innon-barredgalaxiesthefield linesareof overall spiralshapesothattheregularfieldsdo not follow thegasflow, which is typical for dynamo-generatedfields.

10 Open questions

All gas-richgalaxies,evenirregularones,hostfieldsof # � nT strength.If thefieldsdo not fill thewholeinterstellarspace,they areeven stronger. Thedynamical importanceof magneticfields ingalaxiescannotbe neglectedanymore. Increasingresolutionof theradiotelescopesand highersensitivity of thereceiver systems revealedaspectrumof features:On largestscales,extendedspiralfields of differentsymmetry modeswerefound,typical signaturesof dynamo action.How thefield canadoptasimilar pitch angleas thatof theopticalspiral remainsamystery. Thepreferredinwarddirectionof axisymmetric fields, if confirmedby futureobservations,also awaitsexplanation.Onintermediatescales,“magneticarms” betweentheopticalspiralarms (Figs.1,2a) arestill puzzlingasthey seemto bedisconnectedfrom thegas.On theotherhand,fields canalsobealignedby gasflows indensity-wave and barpotentials.Theunexpectedlocationof theshockfront in abarredgalaxy(Fig. 7)tells usthatstrongmagneticfields interactwith thegasflow, but detailsarestill unobservable. Radiohalosare theresultof magneticfields pulledoutwardsby galacticwinds.Alternatively,dynamo-generateddipolarfields (Fig. 6) may enhancethewind. Somestructuresin radiohalosshowsimilarities to thosein thesolarcorona:loops,spurs[50] andpossiblycoronalholes[17].

258

Figure 8: Total emission and � -vectorsof polarizedradio emissionof the barredgalaxy NGC 1672at � � � �

cm (ATCA,� � � � synthesizedbeam), superimposedonto a Digital Sky Survey (DSS) optical

image. The lengthof � –vectorsis proportional to the polarizedintensity (Ehle, Haynes,Beck et al.,unpublished)

11 Limitations of present-day polarization observations

Regularfields in galaxiesshow structureson kiloparsecscaleswhich canbeobserved with present-dayradiotelescopes.Typical polarizedintensitiesat � � cm are50–1006 Jy per � � � � beam(VLA Dconfiguration,smoothed).In caseof equipartitionbetweencosmic-ray and magneticfield energydensitiesandassuming a constantfield strengthalong apathlengthof 1 kpc, theseintensitiescorrespondto regularfields of 5–6 6 8 strength.The � � � � beamcanresolve only 730pcat 10Mpc distance,thusthedetailedfieldstructureremainsunexplored.With thenext higherVLA configuration( # � � � beam) thepolarizedintensitydropsto 5–106 Jy perbeamwhich cannotbedetectedwithin reasonabletime. Forexample,only thecircumnuclearring of NGC 1097is visible in polarizedemissionat � � cm and � * cmwith theVLA C-configuration.Within theopticalspiralarmspolarizedemissionis evenlower duetofield tanglingon scalesof 10–100pc. Hence,while theVLA in principleprovidessufficiently highangularresolution,its sensitivityis too low to studythedetailedfield structure.

Anotherlimitation isgalacticdistance.Thoughradiointensityperunit beamsizeisdistance-independent, moredistantgalaxiescanbeobserved with thesame signal-to-noiseratio only aslong asthespatial resolutionis sufficient to resolve them. If thelarge-scalestructureof themagneticfield is not fully resolved,polarizedintensitydecreasesdueto geometricaldepolarizationacrossthebeam. Thecurvatureof aspiralfield is significantbeyondascaleof # � kpc, thuswe losepolarizedintensityperstandard� � � � beamfor galaxiesbeyond # @ �

Mpc distance.Again,usingthe next largerVLA configurationdoesnot helpbecausetheintensitydropsby a factorof # � �

. Hence,polarizationobservationsof distantgalaxiesare sensitivity-limited, too.

With theSquareKilometer Array thefield structurescouldbe observed of 10x moredistantgalaxieswhich are several Gyr youngerthanthenearbyobjects.Thedynamo timescaleis some � � A yr [13] sothatweaker or lessregularfieldsmay beexpectedat larger distances.Using Faradayrotationmapping,evenyoungergalaxiescanbeobserved (seebelow).

259

Faradayrotationmeasuresof backgroundsourcescantracemagneticfields to largerradii thanpolarizedintensity. Within thedisks,theratio betweeninternal � andbackground� allows todetectfieldreversalsalong theline of sight. However, thenumberof available sourcesis very low. In theM31field( # � � � squaredegrees)only 22 backgroundsourceswith polarizedflux densitieshigherthan0.3 mJywere foundat � � �

cm within 1 hourVLA on-sourceobservation time [6]. In amuch smaller field likethatof NGC 6946,thenumberof available sourcesdropsto lessthan1. Evenwithin several hoursobservation timeonly a few polarizedbackgroundsourcesaredetectable.Themuchbettersensitivity oftheSKA will allow for thefirst time asystematic RM mappingof galaxies.

Signaturesof a regularfield with abisymmetric spiralstructureweredetectedin theintervening( B � � � * , � ) galaxyin front of PKS1229–021[55]. TheSKA will allow suchstudiesof themagneticfield structurein very younggalaxieswith higherresolutionand much bettersensitivity.

12 Prospectsfor the Square Kilometer Arra yC High-resolutionpolarizationmappingat centimeterwavelengths:

– Polarizedintensitiesat differentspatialresolutionsto obtaintheturbulencespectrumof theregularfield, to be comparedwith modelsof field generationand destruction

– Detailedcomparisonbetweenmagneticfieldsand gascomponentsof varioustemperaturesanddensities:Which gascomponentsinteractwith thefield?

– Detailedcomparisonbetweenfieldsand thegasflow in star-forming regions,spiral arms,bars,centralrings: Whattanglesthefield?Whereis thefield frozeninto thegasflow, wheredoesit diffuseaway?

– Field structurein halos:Searchfor loops,streamers,coronalholesandindicationsformagneticreconnection

– Needed:Systems in theFaraday-thinregime ( D � GHz)C High-resolutionpolarizationmappingatdecimeter wavelengths:

– Field structurein theforegroundFaradayscreen

– ISM tomography

– Needed:Systems with a largenumberof bands(0.3–2GHz)C High-resolutionFaradaymapping:

– Directionof axisymmetric spiralfields

– Searchfor field reversalsin disks

– Reversalsin halos:Wind or dynamo?

– Polarizedbackgroundsources:Extentof nearbymagneticdisks

– Distantpolarizedradiogalaxies:Field structurein intervening galaxies

– Needed:Systems between2 GHz and10 GHzC High-resolutionZeemanmapping(beyondthescopeof this paper)

References

[1] H.J. Volk, Astronomy & Astrophysics218,67-70(1989)[2] E. Hummel, R. Beck,Astronomy & Astrophysics303,691-704(1995)[3] S. Niklas,R. Beck,Astronomy & Astrophysics320,54-64(1997)[4] S. Niklas,U. Klein,R. Wielebinski,Astronomy & Astrophysics322,19-28(1997)[5] P. Hoernes,E.M. Berkhuijsen,C. Xu, Astronomy & Astrophysics334,57-70(1998)

260

[6] J.L. Han,R. Beck,E.M. Berkhuijsen,Astronomy & Astrophysics335,1117-1123(1998)[7] D.D. Sokoloff, A.A. Bykov, A. Shukurov, E.M. Berkhuijsen,R. Beck,A.D. Poezd,Monthly

Noticesof theRoyal Astronomical Society 299,189-206(1998)and303,207-208(1999)(Erratum)

[8] M. Urbanik,D. Elstner, R. Beck,Astronomy & Astrophysics326,465-476(1997)[9] M.H. Wieringa,A.G. de Bruyn,D. Jansen,W.N. Brouw, KatgertP., Astronomy & Astrophysics

268,215-229(1993)[10] A.D. Gray, T.L. Landecker, P.E. Dewdney, A.R. Taylor, Nature393,660-662(1998)[11] B. Uyaniker, E. Furst,W. Reich,P. Reich,R. Wielebinski,Astronomy & AstrophysicsSuppl.

Ser. 138,31-45(1999)[12] R. Beck,P. Hoernes,Nature379,47-49(1996)[13] R. Beck,A. Brandenburg, D. Moss,A. Shukurov, D. Sokoloff, AnnualReviews Astronomy

Astrophysics34,155-206(1996)[14] S. Niklas,PhD Thesis,Universityof Bonn(1995)[15] U.R. Buczilowski, R. Beck,Astronomy & Astrophysics241,47-56(1991)[16] N. Neininger, Astronomy & Astrophysics263,30-36(1992)[17] R. Beck,Astronomy & Astrophysics251,15-26(1991)[18] R. Beck,M. Ehle,V. Shoutenkov, A. Shukurov, D. Sokoloff, Nature397,324-327(1999)[19] N. Neininger, C. Horellou,1996,in Polarimetry of theInterstellar Medium(Astr. Soc. Pacific),

Edited by W.G. Roberge & D.C.B. Whittet, p.592-597[20] D.L. Block,B.G. Elmegreen,A. Stockton,M. Sauvage,TheAstrophysicalJournal486,L95-98

(1997)[21] E.M. Berkhuijsen,E. Bajaja,R. Beck,Astronomy & Astrophysics279,359-375(1993)[22] M. Soida,M. Urbanik,R. Beck,R. Wielebinski,Astronomy & Astrophysics345,461-470

(1999)[23] M. Ehle, R. Beck,R.F. Haynes,A. Vogler, W. Pietsch,M. Elmouttie,S. Ryder, Astronomy &

Astrophysics306,73-85(1996)[24] M. Krause,R. Beck,E. Hummel, Astronomy & Astrophysics217,17-30(1989)[25] S. Sukumar, R.J. Allen, Nature340,537-539(1989)[26] R. Rohde,R. Beck,D. Elstner, Astronomy & Astrophysics(1999),in press[27] S. von Linden,K. Otmianowska-Mazur, H. Lesch,G. Skupniewicz, Astronomy & Astrophysics

333,79-91(1998)[28] J.Knapik,M. Soida,R.-J.Dettmar, R. Beck,M. Urbanik,Astronomy & Astrophysics(1999),in

press[29] A.G. Lyne,F.G. Smith, MonthlyNoticesof theRoyal Astronomical Society237,533-541(1989)[30] J.L. Han, R.N. Manchester, G.J. Qiao,Monthly Noticesof theRoyal Astronomical Society 306,

371-380(1999)[31] K.T. Chyzy, R. Beck,S. Kohle,U. Klein,M. UrbanikM., Astronomy & Astrophysics(1999),

submitted[32] M. Krause,E. Hummel, R. Beck,Astronomy & Astrophysics217,4-16(1989)[33] M. Krause,1993,in TheCosmic Dynamo (Kluwer, Dordrecht),Edited by F. Krauseet al.,

p.305-310[34] J.L. Han, R. Beck,M. Ehle,R.F. Haynes,R. Wielebinski,Astronomy & Astrophysics348,

405-417(1999)[35] Z. Fan,Y.-Q. Lou,Monthly Noticesof theRoyal Astronomical Society 291,91-109(1997)[36] P.A. Patsis,P. Grosbol,N. Hiotelis,Astronomy & Astrophysics323,762-774(1997)[37] A. Shukurov, Monthly Noticesof theRoyal Astronomical Society 299,L21-24(1998)[38] D. Moss,Monthly Noticesof theRoyal Astronomical Society 297,860-866(1998)[39] R. Rohde,D. Elstner, Astronomy & Astrophysics333,27-30(1998)[40] R. Beck,Astronomy & Astrophysics106,121-132(1982)[41] F. Krause,R. Beck,Astronomy & Astrophysics335,789-796(1998)[42] R. Wielebinski,F. Krause,TheAstronomy & AstrophysicsReviews 4, 449-485(1993)[43] M. Krause,1990,in Galactic andIntergalacticMagneticFields(Kluwer, Dordrecht),Edited by

R. Becketal., p.187-196

261

[44] E.M. Berkhuijsen,C. Horellou,M. Krause,N. Neininger, A. Poezd,A. Shukurov, D. Sokoloff,Astronomy & Astrophysics318,700-720(1997)

[45] M. Dumke,M. Krause,R. Wielebinski,U. Klein,Astronomy & Astrophysics302,691-703(1995)

[46] S. Sukumar, R.J. Allen, TheAstrophysicalJournal382,100-107(1991)[47] C.L. Carilli, M.A. Holdaway, P.T.P. Ho, C.G. dePree,TheAstrophysicalJournal399,L59-62

(1992)[48] R. Beck,C.L. Carilli, M.A. Holdaway, U. Klein, Astronomy & Astrophysics292,409-424

(1994)[49] E. Hummel, R. Beck,M. Dahlem, Astronomy & Astrophysics248,23-29(1991)[50] G. Golla, E. Hummel, Astronomy & Astrophysics284,777-792(1994)[51] M. Dahlem, M.G. Petr, M.D. Lehnert,T.M. Heckman, M. Ehle, Astronomy & Astrophysics320,

731-745(1997)[52] H.-P. Reuter, U. Klein,H. Lesch,R. Wielebinski,P.P. Kronberg, Astronomy & Astrophysics

282,724-730(1994)[53] M. Dahlem, U. Lisenfeld,G. Golla, TheAstrophysicalJournal444,119-128(1995)[54] E. Athanassoula,Monthly Noticesof theRoyal Astronomical Society 259,345-364(1992)[55] P.P. Kronberg, J.J. Perry, E.L.H. Zukowski, TheAstrophysicalJournal387,528-535(1992)

262