A&A617,A58(2018) Astronomy https://doi.org/10.1051/0004-6361/201832582 & (cid:13)c ESO2018 Astrophysics Feedback from reorienting AGN jets I. Jet–ICM coupling, cavity properties and global energetics(cid:63) S. Cielo1,A. Babul2,3,4,V. Antonuccio-Delogu5,J. Silk1,4,6,7,8,andM. Volonteri1,4 1 SorbonneUniversités,UPMCUniv.Paris6etCNRS,UMR7095,IAP,Paris,98bisbdArago,75014Paris,France e-mail:[email protected]; [email protected]; [email protected] 2 UniversityofVictoria,3800FinnertyRoad,Victoria,BCV8P5C2,Canada 3 InstituteofComputationalScience,CentreforTheoreticalAstrophysicsandCosmology,UniversityofZurich, Winterthurerstrasse190,8057Zurich,Switzerland 4 Institutd’AstrophysiquedeParis,98bisbdArago,75014Paris,France 5 INAF/IstitutoNazionalediAstrofisica-CataniaAstrophysicalObservatory,ViaS.Sofia78,95126Catania,Italy 6 AIM–Paris-Saclay,CEA/DSM/IRFU,CNRS,UnivParis7,91191Gif-sur-Yvette,France 7 DepartmentofPhysicsandAstronomy,TheJohnsHopkinsUniversity,Baltimore,MD21218,USA 8 BIPAC,UniversityofOxford,1KebleRoad,OxfordOX13RH,UK Received3January2018/Accepted7June2018 ABSTRACT Aims.Wetesttheeffectsofre-orientingjetsfromanactivegalacticnucleus(AGN)ontheintraclustermediuminagalaxycluster environmentwithshortcentralcoolingtime.Weinvestigateboththeappearanceandthepropertiesoftheresultingcavities,andthe efficiencyofthejetsinprovidingnear-isotropicheatingtothecoolingclustercore. Methods.WeusenumericalsimulationstoexplorefourmodelsofAGNjetsoverseveralactive/inactivecycles.Wekeepthejetpower anddurationfixedacrossthemodels,varyingonlythejetre-orientationangleprescription.Wetrackthetotalenergyoftheintracluster medium(ICM)intheclustercoreovertime,andthefractionofthejetenergytransferredtotheICM.Wepayparticularattentionto wheretheenergyisdeposited.WealsogeneratesyntheticX-rayimagesofthesimulatedclusterandcomparethemqualitativelyto actualobservations. Results.Jetswhosere-orientationisminimal((cid:46)20◦)typicallyproduceconicalstructuresofinterconnectedcavities,withtheopening angleoftheconesbeing∼15−20◦,extendingto∼300kpcfromtheclustercentre.Suchjetstransferabout60%oftheirenergytothe ICM,yettheyarenotveryefficientatheatingtheclustercore,andevenlessefficientatheatingitisotropically,becausethejetenergy isdepositedfurtherout.Jetsthatre-orientateby(cid:38)20◦generallyproducemultiplepairsofdetachedcavities.Althoughsmaller,these cavitiesareinflatedwithinthecentral50kpcandaremoreisotropicallydistributed,resultinginmoreeffectiveheatingofthecore. Suchjets,overhundredsofmillionsofyears,candepositupto80%oftheirenergypreciselywhereitisrequired.Consequently,these modelscometheclosestintermsofapproachingaheating/coolingbalanceandmitigatingrunawaycoolingoftheclustercoreeven thoughallmodelshaveidenticaljetpower/durationprofiles.Additionally,thecorrespondingsyntheticX-rayimagesexhibitstructures andfeaturescloselyresemblingthoseseeninrealcool-coreclusters. Keywords. galaxies:clusters:intraclustermedium–galaxies:jets–X-rays:galaxies:clusters–methods:numerical 1. Introduction 1995;Binney&Tabor1995;Ciotti&Ostriker2001;Babuletal. 2002;McCarthyetal.2008),andwhiletodaythereisbroadcon- Despiteearlyclaimsthatclustersofgalaxiesarestraightforward sensus that this is indeed what is happening (for a review, see systems to model, steadily improving observations as well as McNamara & Nulsen 2007, 2012; Fabian 2012; Soker 2016), decade-long theoretical and computational efforts indicate that there are a number of critical details associated with this jet- theyareanythingbut.Thereis,asofyet,noclearconsensuson heatingpicture(commonlyreferredtoasradio-modeAGNfeed- howtheremarkablediversityofobservedclustercoreproperties, back)thathaveyettobeproperlyunderstood;ofthese,twopar- rangingfrom“strongcoolcore”to“extremenon-coolcore”(i.e. ticularlystandout. central cooling times ranging from one to two hundred million Thefirstconcernstheoriginofthegaswhoseaccretiononto years up to several gigayears) and everything in between, has the SMBH powers the jet: Is it due to hot/Bondi accretion or emerged. In the case of cool core groups and clusters, power- a drizzle of cold clouds condensing out of the ambient gas in ful jets from central supermassive black holes (SMBHs) have theclustercoresandfree-fallingontothecentralactivegalactic long been suspected of injecting the required energy into the nucleus (AGN)? For a detailed discussion, we refer the reader intracluster medium (ICM) to compensate for radiative losses to Prasad et al. (2015, 2017) and references therein. Here, we and maintain global stability (see for instance Rephaeli & Silk simplysummarisethecurrentstateofaffairsbynotingthatthat several different lines of observational evidence seem to col- lectively favour the “cold rain” model and theoretical studies (cid:63) The movies associated to Figs. 1–4 are available at indicatethatthecoldcloudsareexpectedtonaturallyforminthe https://www.aanda.org/ presence of AGN-induced turbulence. The latter calls attention A58,page1of22 OpenAccessarticle,publishedbyEDPSciences,underthetermsoftheCreativeCommonsAttributionLicense(http://creativecommons.org/licenses/by/4.0), whichpermitsunrestricteduse,distribution,andreproductioninanymedium,providedtheoriginalworkisproperlycited. A&A617,A58(2018) toabroadersetofrelatedquestionsregardingtheimpactofAGN repletewiththeaccumulateddetritusofcoolgasdrawnout feedbackongasaccretionprocesses;forexample,doestheAGN ofthecentralgalaxiesbyAGNjetsandbubbles(c.f.Saxton simplyexpelthegasaroundtheSMBH,leadingtotheshutdown et al. 2001; Brüggen 2003; Revaz et al. 2008; Pope et al. of AGN activity for a period of time, or does it promote other 2010; Duan & Guo 2018 and references therein), besides mechanismsofaccretionontotheAGNthatalsocontributetoits being locally thermally unstable and susceptible to in situ self-regulation?Inarecentstudy,Cieloetal.(2017)foundthat cloud/filamentformation(cf.Cowieetal.1980;Hattorietal. jet-induced gas circulation (backflows) can funnel as much as 1995;Heckmanetal.1989;Pizzolato&Soker2005b,2010; 1M yr−1totheinnermostparsecs(seealsoAntonuccio-Delogu Nipoti & Binney 2004; Maller & Bullock 2004; McDonald (cid:12) &Silk2010). et al. 2010; Hobbs et al. 2011; Sharma et al. 2012; Gaspari The second issue concerns the coupling between the jets etal.2013,2017;Li&Bryan2014;Voitetal.2017;Prasad and the ICM: how does a central SMBH, powering apparently etal.2015,2017,andreferencestherein). narrowbipolaroutflows,successfullymanagetoheatthegasin (2) The second class of models invoke “ICM weather”, that the cluster cores in a near-isotropic fashion? In this paper, we is, wakes, bulk velocities, and turbulence on scales of a focusonthislatterissue. fewkiloparsecsorlarger,inducedeitherbymergersororbi- Establishing precisely how bipolar AGN jets interact with ting substructure (Soker & Bisker 2006; Heinz et al. 2006; and heat the ICM in the cluster cores to prevent cooling catas- Morsonyetal.2010;Mendygraletal.2012). Onepotential trophes has proven to be an especially vexing problem. The issuewiththisclassofmodelistheneedforsubstantialve- issuehasbeenthesubjectofnumerousstudiesdatingbacktothe locityshearacrosstheclustercore.Inasimulationstudyof early2000s(c.f.Reynoldsetal.2001,2002;Ommaetal.2004; 63 clusters by Lau et al. (2017), only a small fraction meet Omma&Binney2004).Inwhatwasthefirstsystematicattempt thisbar. toaddressthisproblemusingaseriesofhigh-resolution,three- (3) The third class invokes occasional changes in the orienta- dimensional hydrodynamic simulations, Vernaleo & Reynolds tionofthespinaxisoftheSMBHsthatarepoweringthejets (2006),whoconsideredthestandardmodelwherethedirection (and hence, the re-orientation of the jet axis). This can oc- of the jets is fixed, found that such jets only managed to delay cur as a result of precession and slewing (or tilting) of the theonsetofcatastrophiccooling,notpreventit(seealsoO’Neill blackholespinaxis,particularlyincombinationwithinter- &Jones2010).Theprimaryreasonforthefailurewasthatonce mittentjetactivity,andspinflips(cf.Merritt&Ekers2002; theinitialjethaddrilledthroughtheclustercoreandexcavated Pizzolato&Soker2005a;Gittietal.2006;Dunnetal.2006; alow-densitychannel,allsubsequentjetstookadvantageofthis Lodato & Pringle 2006; Campanelli et al. 2007; Sternberg channeltoflowfreelyoutofthecore,carryingtheirenergywith &Soker2008;Kesdenetal.2010;Falceta-Gonçalvesetal. them.Theauthorsconcludedthatsomeadditionalcomplexityis 2010; Merritt & Vasiliev 2012; Babul et al. 2013; Gerosa requiredtoensuremoreeffectiveheatingoftheclustercoresby etal.2015;Franchinietal.2016;Nawazetal.2016).From AGNjets. amacroscopic,clustercorescaleperspective,alloftheseare Detailed X-ray and radio observations of individual cool similarinthattheygiverisetojetswhoseorientationchanges core groups and clusters (hereafter collectively referred to as stochastically.Interestingly,arecentmulti-wavelengthstudy cool core clusters or CCC) offer intriguing hints about how (O’Sullivan et al. 2012) of the core of z=0.442 cool-core naturehasaddressedtheisotropyprobleminrealsystems.Many galaxyclusterCL09104+4109,andtheTypeIIquasi-stellar of the CCCs show evidence of multiple generations of active object (QSO) B0910+410 at its centre, offers tantalising and relic jets, radio lobes, and X-ray cavities whose angu- support for this scenario. B0910+410 is one of only two lar positions/directions in the sky are misaligned with respect z<0.5 QSOs at the centre of a galaxy cluster and it seems to each other. Since these by-products of AGN jets trace a to have switched from being a radio AGN to a QSO about nearly isotropic angular distribution about the cluster centre, 200Myr ago in response to a significant inflow of gas, and Babul et al. (2013) argue that the associated heating should do appears to be transitioning back to a radio AGN, with the the same. Moreover, since the observations indicate directional bud of a new jet clearly misaligned with respect to the old changesontime-scalesrangingfromafewtoafewtensofmil- large-scalerelicjet. lionsofyears–whichistypicallyshorterthanthecorecooling Inthispaper,thefirstofaseries,weusenumericalsimulations time–thejetsoughttobeabletoheatandmaintainthecorein toinvestigatethelatterclassofmodelswithintheframeworkof atleastaglobalequilibriumconfiguration. acool-coregalaxyclusterwithaninitialcentralcoolingtimeof There are three distinct categories of models proposed to 150 Myr.Weattempttocapturethebasicfeatureofthedifferent accountfortheobservedmisalignmentofsuccessivegenerations models within this category via stochastically re-orienting jets. ofjet-lobe-cavityfeatures: Weinvestigateexplicitlytheextenttowhichsuchjetscoupleto (1) The first invokes jets interacting with, and being de- the cooling ICM in the cluster core, the efficacy of these mod- flected by, dense clouds and filaments in the ICM (c.f. els in affecting isotropic heating in the cluster cores, and more de Gouveia Dal Pino 1999; Mendoza & Longair 2001; broadly, the impact of such jets on the thermal and dynamical Saxton et al. 2005; Prasad et al. 2018). This scenario has evolutionofthehotdiffuseICM.Weexplorethreedifferentpre- not been explored much because historically the ICM scriptionsforjetre-orientation(alsorunningacasewithoutjets was assumed to be largely homogeneous. However, there andajettedbutnon-reorientingone,forreference),andcompare may be cause to revisit this model. Observations show thepredictionsforthepropertiesofthebubblesandthestability that the central galaxies in cool core clusters are typ- ofthecoolcoreineachcase. ically surrounded by extended filamentary warm-cool InSect.2,wedescribethemodelweuseforourre-orienting gas nebulae (Hatch et al. 2007; Cavagnolo et al. 2008; jetsinaCCC,andtheset-upofoursimulations.Sections3–5are Wilman et al. 2009; McDonald et al. 2010 – see also devoted to a detailed but qualitative description of the physics Heckman et al. 1989; Crawford et al. 1999). Further sup- and appearance of the cavities and all the visible structures, port comes from theoretical and simulation studies, which obtained by comparing realistic X-ray images from the strongly indicate that the ICM in this region ought to be simulationswiththephysicalstateofthegas.Wealsodiscussthe A58,page2of22 S.Cieloetal.:FeedbackfromreorientingAGNjets.I. roleofprojectioneffects.Section6containsaquantitativeanal- cluster. Guided by the mass-concentration relation by Newman ysis of the bubbles’ properties and the heating that re-orienting etal.(2013),whichtakesintoaccountthepresenceofacentral jets are able to provide to the core, while Sect. 7 relates these BCG(seetheirSect.10.1),andtherecentanalysisoftheVirgo propertiestoobservablelarge-scaleinflowsandoutflows,tothe clusterobservationsbySimionescuetal.(2017),wesetourclus- temperatureandstabilityofthecoolcore,andtotheenergybal- terconcentrationparametertoc =10.Thehalochoicedefines 200 anceofX-raygas.InSect.8,werelateourfindingstothoseof oursimulationbox.Inordertoencompassthehalor ,weuse 200 previous simulation studies and discuss how our results would acubicboxof4Mpcside. change by varying the inactivity duration during a jet cycle. In Having defined our dark matter halo, we next add to it a Sect. 9, we draw our conclusions on how re-orienting jets im- spherically symmetric hot gas component whose radial profile pact the shape of the X-ray cavities, the ICM as a whole, and issubjecttothefollowingthreeconstraints. thehalocore.Afollowingpaperfeaturingthesamesimulations willbededicatedtoananalysisofenergygenerationandtrans- Initialentropyprofile. For the starting entropy profile of the port, differentiating the effects of the different physical mecha- ICM in our simulations, we adopt the functional form nisms(radiativecooling,shock-heating,advectiveorconvective ln(S(r))= ln(S0)+αln(r/rc),whereS(r)≡kBT(r)/ne(r)2/3, transport,mixing,turbulentdissipation,etc.),andhowthesede- kB is the Boltzmann constant, T is the gas temperature, ne terminethehalos’gaseousprofiles.Inthefollowing,wereferto is the gas electron density, calculated by assuming fully thisworkasPaperII. ionized plasma with the given metallicity, and the power- law index α=1.1 when r is larger than the core radius r , and α=0 otherwise. This simple functional form has c 2. Modelsandnumericalimplementations previously been used to describe the observed diversity of entropy profiles across the cool-core/non-cool core spec- We test the reorienting jets model using a total of five numeri- trum (Cavagnolo et al. 2009) as well as starting configura- calsimulations.Thesesimulationswererunusingthehydrody- tionsintheoreticalandnumericalstudies(Babuletal.2002; namical, adaptive mesh refinement (AMR) code FLASH v4.2 McCarthyetal.2008;Prasadetal.2015).Wechooseacore (Fryxell et al. 2000), adopting a modified setup described in radius of r =12kpc and a core entropy S =12keVcm2, Cielo et al. (2017). In our computational setup, FLASH solves c 0 which results in a core with a cooling time of 150Myr and the non-relativistic Euler equations for an ideal gas, with spe- qualifiesoursimulatedclusterasacool-coresystem.Weap- cific heat ratio γ=5/3, initially placed in hydrostatic equilib- preciate that recent studies show that groups and clusters riumwithinagravitationalpotentialwell.Thegravityactingon withshortcentralcoolingtimesdonothaveisentropiccores, thegasisthatduetothegasitselfaswellasastatic,spherically and instead exhibit an r2/3 profile (Panagoulia et al. 2014; symmetric,darkmatterhalo(seebelowfordetails). O’Sullivanetal.2017;Babyketal.2018).Sincewedonot The metallicity of the gas is set to [Fe/H]=−0.1 through- couplethejetactivitytothestateoftheICM(cf.Sect.2.2), out and over the course of the simulation, the gas is subject thedetailedstructureoftheinnerentropyprofilehasnobear- to radiative cooling following the same prescriptions as used ing on our primary objective, which is to investigate the by Cielo et al. (2017). Specifically, we use the cooling func- efficacy of the re-orientating jets at affecting near-isotropic tionofSutherland&Dopita(1993),extendedtohigherplasma heating in the core region. For this, we only require that temperatures (i.e. ∼1010K) as described in Appendix B of the core cooling time is shorter than the simulation run Antonuccio-Delogu & Silk (2008) to allow for a proper treat- time. mentofgasinthejetbeamsandthecavities. Initialhydrostaticequilibrium. We require the gas to be in hy- We allow the AMR in FLASH to refine up to level 10 drostatic equilibrium (HSE) within the dark matter poten- (i.e. it refines at most ten times), if density and temperature tial.TheHSEequationwiththechosenentropyprofileisnot gradients require so1. At each refinement operation, a block of analytically integrable, so we numerically integrate it sepa- interest is split in two along every spatial dimension. In ad- ratelyandimportthetabulatedprofileintoFLASH(thespa- dition, all the blocks are further divided into eight computa- tial sampling of the integration is chosen equal to the grid tional cells along each dimension, giving a resolution element usedinthesimulations). of4Mpc/(8×210)(cid:39)488pc.Refinementcanbetriggered,upto Profilenormalization. Inordertogettherighthot-gas-to-dark- maximum level, anywhere in the simulation box, without geo- matterratioforthegivenhalomass,wenormalizetheprofile metricalrestrictions;forinstance,inrefiningwedonotprivilege sothattheratioofhotgastodarkmattermassattheradius thecentralregionoverthejet-inflatedcavities. r (cid:39)1Mpcissetto60%ofthecosmicvalue,inagreement 500 with the observational results shown in Liang et al. (2016) 2.1. Initialconditions forVirgo-masssystems. Theresultinggastemperatureisbetweenoneandafewkiloelec- All simulations feature the same initial conditions meant to re- tronvoltsthroughout. produce a CCC comparable in mass to the Virgo Cluster. We assume a flat background cosmology corresponding to Ω h2= m 0.1574,Ωbh2=0.0224and H0=0.7(Komatsuetal.2011),and 2.2. Reorientingjets:parametersandimplementation useastatic,sphericallysymmetricgravitationalpotentialforan NFW halo (M (cid:39)4.2×1014M , r (cid:39)1.7Mpc) to define our Our implementation of the jet source terms is essentially the 200 (cid:12) 200 same as in Cielo et al. (2017). There, the bipolar jets were in- troduced as source terms within a rectangular prism consisting ofeightcentralcells(fourcellsperbeam)whoselongaxiswas 1 TherefinementcriterionusedisthesameasinCieloetal.(2017),i.e. aligned along one of the axes of the simulation grid. The only FLASH’sdefaultrefinementstrategybasedonLöhner’serrorestimator (seeFLASHusermanual,orLöhner1987).Weapplythecriterionto differencehereisthatweallowtheinclinationofthejetaxiswith bothdensityandtemperature,andsetthisparameterto0.8forrefine- respect to the simulation grid to vary. This, in turn, means that mentand0.6forde-refinement. the number of injection cells also varies with the jets’ inclina- A58,page3of22 A&A617,A58(2018) tion. We parametrize the jets’ orientation using standard spher- Table1.Orientationangles(θ,ϕ)ofalljetsineachrun. ical coordinates θ (angle between the z axis of the grid and the jetaxis)andϕ(angleofthepositivedirectionofthe xaxiswith Jet ton 0000 0030 2030 0090 thejetaxisprojectiononthez=0plane).Whenjetsareinjected # Myr deg deg deg deg atanangle,theinjectioncellsandthemomentumdirectionare 1 0 (0,0) (0,0) (0,0) (0,0) changedtofollowthatorientation. 2 42 (0,0) (12,0) (25,0) (75,0) We present four simulation runs with jets and a control run 3 84 (0,0) (15,173) (44,167) (36,96) with no jets. In the jetted runs, we do not couple the trigger- 4 126 (0,0) (28,150) (69,160) (18,160) ing of the jets or their power to the state of the ICM. All four 5 168 (0,0) (41,161) (95,136) (72,177) jetted runs are identical except for the prescription for their re- 6 210 (0,0) (65,138) (73,137) (82,16) orientation angles, meaning that all jets have the same power, 7 252 – – – (75,173) density,internalenergyandinjectionbaseradius. Following Cielo et al. (2014), the jets’ density ρ is jet set equal to 1/100 of the central halo gas density (see also hundred million years duration of the simulations covers com- Perucho et al. 2014; Guo 2016). Further, we fix the jet kinetic fortablytheinitialcentralcoolingtimeofthehaloof150Myr,so powerto Pjet=1045ergs−1,inagreementwithmeasuresofme- thatintheno-jetrunalargecoolingflowdevelops.Thenumber chanicalluminositiesfromX-raycavitiesingalaxyclusters(e.g. of jet events we covered is in principle not free from statistical Hlavacek-Larrondoetal.2012)andapproximatelyequal tothe under-sampling;however,avisualinspectionoftherunsreveals totalradiativelossesofthehalogas. that the intended solid angle coverage is achieved in all cases. Besides their kinetic power Pjet, the jets also have inter- For instance, in run 0090, cavities never substantially overlap, nal energy, the flux of which, Ujet, can be simply computed sothatjetsareindeedaffectingaportionofsolidanglethatisas from the jet Power Pjet and the jet’s internal Mach number largeaspossible;onthecontrary,inrun0030,thejetsquiteoften (cid:112) M :=v / γp /ρ ,whichwesetto3: endupinthetrailofthepreviouscavity,asweseeinSect.4. jet jet jet jet P 1 2P 9 U = jet = jet =0.2P . (1) jet M2 γ(γ−1) 9 10 jet 3. Cavities:physicalbackgroundandmethod jet Our refinement criterion keeps the jet/cavity system maximally Thetotalenergyfluxisthereforeaconstant1.2×1045ergs−1 refined at all times, therefore providing detailed insight on the whenever the jets are on, and zero otherwise. All jets follow physics of the bubbles, including velocity and turbulent struc- the same on/off schedule: each jet event lasts 40Myr, followed ture,riseandexpansionintheexternalgas,andtheimplications by a 2Myr quiescent period, during which the jet source terms fortheenergeticsofthecool-corehalo. are switched off. After that, another jet event starts, generally We are also able to pair this physical view with synthetic along a different direction. Overall, this means that feedback is observations, via the production of realistic X-ray emissiv- active with constant power for ∼95% of the time. Our choice ity maps (see Sect. 3.2 for details), in order to provide of 40/2Myr for the timing of the jet on/off cycle is guided by direct comparison of the individual features from X-ray ob- thetypicaltimescalebetweenmisalignedjeteventsobservedin servations of galaxy clusters. The morphology of real X-ray galaxyclusters,ascataloguedbyBabuletal.(2013),aswellas cavitiesisoftenmorecomplexthanacollectionofbubblepairs thecharacteristicjetalignmenttimescaleintheirpreferredphys- (e.g.Zhuravlevaetal.2016),asseveralotherphysicalprocesses icalmodel. are at work. Some are due to the bubbles themselves: shocks, In this work involving constant-power periodic jets, we bothweakandstrong,(Nusseretal.2006),aswellasthegenera- mainly explore the parameter space of reorientation angles, for tionanddissipationofpressurewaves(Sternberg&Soker2009; which we adopt the following prescription: the jet polar angles Fabianetal.2017andreferencestherein)duringtheinflationof (θ, ϕ) are chosen at random (from a spherical distribution) in a the bubbles; wakes, ripples and ICM motions excited by buoy- givenintervalofangulardistancewithrespecttothepreviousjet antlyrisingbubbles(Churazovetal.2002;Nusseretal.2006), axis; we use this interval to label our simulation runs. For ex- liftingofandsubsequentmixingwithlow-entropygas(Brüggen ample,inrun2030,theaxisofanyjetwillformanangle(cho- 2003; Pope et al. 2010), and so on. Other processes are due to senatrandom)between20◦ and30◦ relativetothepreviousjet the cluster environment: shocks, cold fronts, streams and other axis.Therunswepresentare0000(i.e.jetsalongafixeddirec- transient structures associated with mergers (Poole et al. 2006 tion),0030,2030,and0090(giventhebipolarnatureofthejets, andreferencestherein),tailsofdiffuseionizedgasram-pressure 0090meansthenewdirectionischosentotallyatrandomonthe stripped from cluster/group galaxies (e.g. Boselli et al. 2016), sphere,withnoconstraints). and so on. Many of these processes can be isolated and stud- Table 1 lists all orientation angles (polar, azimuthal) or (θ, iedsimplybyprocessingtheX-rayimages(seeChurazovetal. ϕ)ofeachjet(indegrees,roundedtothenearestinteger)forall 2016), while simulations can provide insights about the origins simulation runs. Random numbers are drawn from a spherical ofthevariousfeatures. distribution;theneachdirectionischosensothatitsangulardis- tancewithrespecttothepreviousjet,expressedagainindegrees, 3.1. Structuregeneratedbyasinglejet lieswithintheintervalthatlabelstherun. Wenotethatthefirstjetisalwaysalongthez-axis,whilethe Several numerical studies (e.g. Vernaleo & Reynolds 2006; secondliesalwaysinthey=0planeforvisualizationsimplicity; Sutherland&Bicknell2007;Cieloetal.2014)haveinvestigated asthehalo’sinitialconditionsaresphericallysymmetric,thisis the evolution of AGN jets propagating through an unperturbed justachoiceofreferenceframe. ICM. The evolution follows a characteristic trajectory that also Therundurationforthejettedsimulationsrangesfrom227 provides a fitting description of the first jet event in our simu- to277Myr;allsimulationsterminatedduringthesixthjetevent, lations. Below, we briefly summarise the main features of this except run 0090, which reached the seventh jet event. The few trajectoryasitunfoldsinourruns. A58,page4of22 S.Cieloetal.:FeedbackfromreorientingAGNjets.I. Theinitialinteractionbetweenajetandtheambientgascre- ates a ∼1010K hot spot (HS; usually no wider than 1 or 2kpc) andabow-shockpropagatingforseveraltensofkiloparsecs.The twobow-shocksencloseanellipsoidalcocoon,filledwithsparse, hot,andturbulentgas.Internally,theexpandingcocoonsupports large-scalegascirculationthatresultsinbackflows,whileglob- allyitbehaveslikeanalmostuniformoverpressurizedbubble. Near the HS, the shocked gas collects in two very hot cav- ities, which in a few million years evolve into lobe-like struc- tures,typicalofclassical radiogalaxies.Thegasinthe “lobes” is denser and hotter than in the rest of the cocoon; Cielo et al. (2014) name this the lobe phase. The jets then switch off: the bow-shockslosetheirdriveandslowdown,firstbecomingtran- sonicandeventuallysubsonic,whilethelobegasdetachesfrom thecentre,forminghotlow-densitybubbles. The rising bubbles retain their inner velocity structure, a vortex-ring-like bubble-wide circulation, as expected for light, supersonicjets(seeGuo2015).Themotionofthebubble-ICM boundary due to vortices inside the bubbles and the backflow of the ICM around the bubble excite sound waves (Sternberg & Soker 2009). As the bubbles ascend and move into regions wheretheambientgaspressureislower,theyexpandandcool. Adiabatic cooling dominates over radiative cooling because of thebubble’slowdensity. The second and subsequent generations of jets that follow propagatethroughanalreadyperturbedICM,sotheirevolution isstronglyimpactedbyencounterswithstructuresgeneratedby earlier jets. For example, AGN jets possess high velocities but lowinertia:denseolderbow-shockfrontscanactlikewalls,de- flecting the jet beams and exciting oblique shocks or “ripples” in dense regions of the surrounding gas; channels and cavities carvedbypreviousjetsactaslow-resistanceconduitsforsubse- quentjetflows. 3.2. Method:physicalpropertiesofthegas versusX-ray maps Figures1–4showthreedifferentpanelsforasinglesnapshotof eachjettedrun,presentedfromtheleasttothemostisotropicjet distribution,thatis,0000,0030,2030and0090. By comparing those figures, we observe the different pre- dictionsforlocation,visualaspect,andphysicalstateofthehot bubbles in the cluster’s gaseous halo. All images still present a veryhighdegreeofcentralsymmetry;thisisduetotheabsence ofsubstructureorasymmetryinourinitialhalo,ortheabsenceof acentralgalaxy(sinceacentralclumpyISMmayinduceasym- metry in radio jets, as shown by Gaibler et al. 2011). This is not necessarily true in real CCC cavities, nonetheless most of the complex features we observe in our figures can be directly compared with the observations and are indicative of the vari- ousphysicaljet–jetandjet–ICMinteractionstakingplaceinthe centralfewhundredkiloparsecs.Onscalesofafewmegaparces, cosmologicalaccretionaswellastheinteractionoftheICMwith infallingsubstructurecangiverisetoshocksandpressurewaves (see, e.g. Poole et al. 2006; Storm et al. 2015) but these can be easilydistinguishedfromfeaturesoriginatedbyAGNfeedback. Fig.1.Run0000:3Dtemperature,focusingoncoolcoreandjetmate- In the following, we describe the content of each panel in rial,lineofsightparalleltothex-axis(top);X-ray,projectionalongthe Figs.1–4,fromtoptobottom. x-axis(centre),pressuresliceinthex=0plane(bottom).Size:400kpc, Three-dimensional rendering of the gas temperature. linearscale.Labelsmarkcavities(C),bow-shocks(B)andripples(R), numberedfromoldesttoyoungest.Seetextandassociatedmovie. Obtained with a ray-casting technique. The line of sight is along the X-axis. The X-ray gas background has been made transparent (as indicated by the opacity annotation next to the – the halo’s cool core, extending up to a few tens of kilopar- colour key in each plot) in order to highlight jets and cavities. secandpresentingthelowesttemperaturesinthesimulation Fromcoldesttohottest,wecanrecognise: (aboutandbelow107K); A58,page5of22 A&A617,A58(2018) Fig.2.AsinFig.1,butforrun0030.Wenotehowcavitiesaregrouped Fig.3.AsinFig.1,butforrun2030.Thistimebow-shocksandcavities together,asoftentwoormorejetsendupinflatingthesamebubble.See aremostlydistinctfromoneanother,butsubjecttoprojectioneffects. textandassociatedmovie. Seetextandassociatedmovie. – thebow-shocks,around108K(althoughtheycoolrelatively The temperature of the cavity gas shown in the three- rapidly to the background gas temperature) are generally dimensional (3D) renderings is a much more reliable proxy of morevisibleonlyaroundthemostrecentjets; their age than their volume or their projected distance from – the jet-inflated bubbles, ranging from a few ×107 to the cluster centre. In the presence of multiple bubbles, the ∼2×108K(theyounger,thehotter); combinationofprojectioneffectsandunknownrelativeorienta- – the latest jet beams/Hot Spots, as hot as a few ×1010K, de- tionsofthejetsthatgaverisetothemmakesageestimatesbased pendingonthejetage. onvolumeanddistancehighlyuncertain. A58,page6of22 S.Cieloetal.:FeedbackfromreorientingAGNjets.I. plasma and its temperature (or energy, should the emission be non-thermal,asinthecaseofpuresynchrotron)holds.Radioor C1 hard X-ray observations (see Sect. 5) ought to be able to effec- tivelyconstrainthere-orientationhistory. Synthetic X-ray maps. The middle panel of each figure C4 C3 presents synthetic soft-X-ray observations, in which jets and bubblesappearasvoids,providingaviewcomplementarytothe temperature.Inordertomaketheseimagesasrealisticaspossi- C2 ble,wegeneratethembyprocessingoursimulationoutputwith thepyXSYMsoftware(basedontheworkbyBiffietal.2013). C5 The software computes thermal and line emission from the hot gasintheX-rayband,thenitgeneratesandpropagatesthecor- responding individual photons through the simulation domain. TheprojectionisalongtheX-axis. Forthesemaps,wechooseanX-raybandof[0.5, 7.0]keV, B4 then set all the sources at redshift z=0.02 and collect the pho- tonsfor1Msfroma6000cm2 telescopearea.Thesevaluesare chosentomatchthespecificsofarealisticobservationwiththe Chandratelescope,exceptforthetelescopearea,whichforour mock observations is aboutten times larger than the one of the C1 ACIS-IdetectoronChandra.Typicalbackgroundcounts(about B1 0.76 photonscm−2s−1) are added, but turn out to comprise less than1%ofthetotalsignal.Galacticabsorptionisalsoincluded with a Tuebingen-Boulder model (see Wilms et al. 2000). The C4 C3 generatedphotonsarethencollectedonthesimulateddetector; B2-B3 spectraandimagescanbeobtainedatthisstage.Onceweobtain C2 the raw images, we apply a standard unsharp mask filter, as is sometimesdoneintheliterature,toemphasizestructureofaspe- B5 cificsize;thisfilteringoperationisnotincludedinpyXSIM.The filterdoesnotpreservethephotoncountineachspaxel;however, herewearemostlyconcernedwiththevisibilityandappearance B4 of the cavities rather than flux measurements. In these initial X-rayimages,wedonotsimulateanyspecificdetectorresponse, but just collect all generated photons. As a by-product, the im- ageretainssomevisualimprintoftheoriginalgrid.Allimages havearatherbrightcoreandaconsequentlyreducedcontrastin theperipheralregions,dueatohighcentralcoolingluminosity (as in run 0000) or to the latest bow-shocks (as in 0090). The C1 youngestbubblesaredistinguishablemostofthetime,although theycansometimesbeconcealedbyolderbubblesoroutshone B1 byabrightbow-shock. Pressure slices. Finally, in the bottom panels of Figs. 1– C4 4 we show central plane slices of the gas pressure (expressed ininternalunitsinordertoavoidnumericalroundingerrors;our unitcorrespondstoabout3.9×10−15Pascal).Whilethefirsttwo C5 panelsareprojectionsalongthexdirection,theseslicesarecon- B5 tainedintheplanex=0,theviewingdirectionbeingythistime, in order to provide a different point of view. Only the structure inthecentralplane(inwhichthefirsttwojetbeamslie)arevis- B4 B2-B3 ible. The gas pressure clearly shows the waves and shocks that leaveimprintsintheX-raygasandallowustotracktheirorigin2. The full time-evolution movies of the pressure slices, provided as additional material to this paper, are very instructive in this respect. Labels. To facilitate discussion to follow, the various struc- Fig.4.AsinFig.1,butforrun0090.Thistime,thebow-shocksarethe tures in the panels in each of the figures are annotated and brightest features, and one can see many cavities around the distance labelled. We use the letters “B”, “C” and “R” to denote bow- (projected)of100kpc.Seetextandassociatedmovie. shocks, cavities (both lobes and bubbles) and ripples, respec- tively.By“ripples”,wearereferringtothosecomplexesofweak shockfrontsappearinginsomeofourX-raymaps,mostlynear Admittedly, there are instances where young jet material the core and in the vicinity of the youngest bow-shocks. All ends up in old bubbles. However, even in this case, it is possi- bletodiscernplasmaofdifferenttemperatureswithinthecavity 2 Howeveradditionalworkisrequiredtodistinguishweakshocksfrom and consequently, a qualitative relation between the age of the pressurewaves,aswewillshowinPaperII. A58,page7of22 A&A617,A58(2018) recognised features are numbered sequentially from oldest to thesamecavity;thereforeweobservethefrequentformationof youngest; features associated with the same jet event, based on composite bubbles, such as the features labelled as C1–C2 and ouranalysisofthetime-evolutionmovie,areassignedthesame C3–C4. number.Thenumbersdonotnecessarilyrefertothejetnumbers Finally,wenotethattheconicalstructureofinterconnected listedinTable1,asweonlylabelwhatisvisibleinthelastpanel, cavitiesthatemergesinrun0030isagenericfeatureofconfig- meaning that some jets may be skipped. Bilaterally symmetric urations where the jets generally change direction by small an- structuresareonlylabelledononeside. gles, regardless of the physical process responsible. Composite cavity structure appears, for example, in the jet simulations of Mendygral et al. (2012), in which the bulk flows in the ICM 4. Results:physicsofcavitiesandtheirappearance cause the jets to deflect by small angles, as well as in simula- tionsofYang&Reynolds(2016),inwhichintermittentjetspre- 4.1. Run0000 cessaboutafixedaxis.Inbothofthesecases,thejetseitherend Run 0000 is shown in Fig. 1. Here all jets keep inflating the upintersectingandpushinginto,ornewlyformingcavitiesend same cavity (labelled as C1): the 2Myr interval between two up breaking through and expanding into, pre-existing cavities successivejeteventsissufficientlyshortthatthereisnotenough excavatedduringearlierjetcycles. time for the channel carved out by the first jet to collapse, and Intermsoftemperature,wecanobservetwopopulationsof allfollowingjetsfollowthispathoflowresistancetowardsC1. plasmaofdifferentages,andonlythemostrecentbubblescon- Thelatterkeepsmovingfurtherawayfromthecentreandgrows tain young jet plasma. The bleeding of one cavity into another larger but never completely detaches from the centre to form a impactstheirsizeevolution,whichinturnwillplayhavocwith proper bubble because the channel is repeatedly refreshed by thepowerestimationsbasedonthismeasure.Theexternalshape a new jet beam. These connections are visible in the X-ray as of older cavities in X-ray matches rather well the correspond- ratherlargejet“chimneys”aroundthejetbeams. ingshapesinthetemperatureview,butprovidesnoinformation Most of the energy of the jets reaches C1, yet in the chim- about its internal structure. Overall, the X-ray map of this run neys, several weak shocks take place. These include both self- resemblesthenumerousX-rayimagesofrealgalaxygroupsand collimation shocks of the beam, and internal reflections of the clusters thatshow a single pairof prominent cavities(cf. Abell latter on the chimney walls. These shocks leave ripples in the 25975;McNamaraetal.2001). X-raygas,suchastheonesmarkedasR3andR4.Inotherwords: All cavities form secondary bow-shocks, but these are afreshjetencountersadensemedium,thechimneywalls,with weaker and fainter than the one associated with the very first almost-zero attack angle, so the new beam gets almost totally jet, as some of the energy of the subsequent jets tends to flow reflected.Theripplesaretheonlyenergytransmittedtotheouter along the first jet channel; of all the bow-shocks, only B6 (the medium. These perturbations are the main visible difference youngest one) and, partially, B5 are visible in the X-ray image between a continuous and a pulsating beam. They propagate (theyarebothclearlyvisibleinthepressuresliceaswell). sideways, but weaken and fade in time, so their presence can The youngest jet beam shows clear signs of deflection (in beassociatedwithaspecificrecentjet.Anotherripplestructure temperature):abrighthotspotanda“plume”shape,wherethe occurs close to the boundary of the cavity, interior to the B1 post-shock beam is deflected into the nearest C5 lobe. Plume- bow-shock.Theseripplesareexcitedbyrepeatedinflationofthe likestructureslikethisonecaninducesignificant(asymmetric) cavitybysuccessivegenerationofjets,startingwiththefirstone. backflows,asobserved,forinstance,inX-shapedradiogalaxies The most peculiar aspect of run 0000 is the very large size (e.g.Robertsetal.2015).Thebeamdeflectionisalsoresponsible ofthebubbleanditscorrespondinglarge-scale(∼200−300kpc) forexcitingripplefeatures(R5andR6). bow-shock,B1.Inmanyways,thecomplexjet/giantcavity/bow- shockresemblesthoseofgalaxyclustersHydraA3 (Nulsenetal. 4.3. Run2030 2005;Wiseetal.2007)andMS0735.6+74214 (McNamaraetal. The first jet axis in run 2030 (Fig. 3) is, as in runs 0000 and 2005, 2009). We see from the temperature panel that run 0000 0030,clearlyvisibleasarisingchimneyofhotgasinbothtem- istheoneinwhichwecanobservejetmaterialthefarthestfrom peratureandX-rayemissionevenafter250Myr;however,unlike theAGN(beyondthe200kpcregionshownhere),giventheease thepreviousruns,theup-downchannelsareweakerandfading withwhichitpropagateswithintheoldcavity,dispersingallover becausetheyhavenotbeenreinforcedbysubsequentjetflows. itsvolume.Insummary,thismodelpredictsemissionfromplas- The hot bubbles in run 2030 are also much more spatially masofdifferentagesfromoneverylargebubble. spreadoutthanintheprevioustworuns.Thespatialdistribution of cavities C1–C4 in the X-ray and temperature maps may, at 4.2. Run0030 firstglance,suggestthatthesecavitiestooaresimplybranchesof amaintrunk,butthisisaprojectioneffect,asconfirmedbyob- Inrun0030(seeFig.2),onecandistinguishindividualbubbles, servingfromadifferentdirection(cf.thepressuremap).Infact, butthosearestillpartofaconnectedstructureoriginatingfrom physically connected, composite cavities are no longer present, the first jet axis, and appear as short thick branches of a main and all the bubbles can be individually identified and easily trunk.Asaconsequence,asignificantfractionoftheenergystill labelled,especiallyinthetemperatureview. flowsthroughthefirstcarvedchimney.Proofofthisisthemain The fact that the cavities are fully detached means that bow-shock,B1,almostaslargeasinrun0000. mostoftheenergyofeachjetisspentcreatingandinflatingnew Of all the features, the brightest is the core+B6 complex; bubbles closer to the halo centre, rather than inflating old, far- the cavities C6 and C5 are also visible, and C5 shows an elon- away cavities. Therefore, jets and young bubbles continuously gatedmorphology.Quiteoften,twosuccessivejetsendupwithin drive shocks near the core, keeping its internal energy higher. Thebubblesalsospendmoretimewithintheinnermost100kpc 3 see http://chandra.harvard.edu/photo/1999/0087/more/ (as C4, C5 and C6 in the shown snapshot) and are distributed 0087_comp_lg.jpg 4 seehttp://chandra.harvard.edu/photo/2006/ms0735/ 5 http://chandra.harvard.edu/photo/2015/a2597/ A58,page8of22 S.Cieloetal.:FeedbackfromreorientingAGNjets.I. over a rather large solid angle, reducing the space available for shocks and cavities makes it very difficult to unambiguously theformationofcoolingflows. establishthebubbles’orderingandenergyunlessadditionalin- An intuitive explanation of why the bubbles are now formation is available (e.g. temperature). Moreover, there is a detached from each other can be found by comparing this run separateindicationthatthesizeofthecavitiesintheX-rayview withthe0000case:thehotlobesformedbyfixed-axisjetsdefine may not be accurate: in the pressure slice, the cavity is much a (bi)conical region of jet influence (in agreement with previ- more extended than in the X-ray view. This suggests that only ousnumericalworks),whosehalf-openingangleisroughly15◦. the highest-contrast, central regions of the cavities stand out in Inrun2030,there-orientationangleisalwaysforcedtobelarger theX-rayimages(butrecallthattheviewingdirectionisdiffer- than this value. Consequently, the new jets do not interact with entbetweenX-rayandpressure). the pre-existing channels and bubbles, and instead pierce the Overall, the X-ray image is similar to the 2030 case, with surrounding ISM/CGM along a new direction. The absence of well-definedbow-shockfrontsaswellaseasilyidentifiableghost ripple-like features further confirms this hypothesis, as ripples cavitiesandnestedcocoons,showingrealisticpositionsandmor- typically arise from interactions between the jet flows and the phologies.TheverybrightcentreintheX-raymapisduetothe wallsofpreviouslyformedchannelsandcavities. cocoon/bowshockofbubblebeinginflatedbyajetalignedclose Thethreeyoungestcavities(C4–C6inFig.3)showaclearly tothelineofsight,sothatthehotspotpointsalmosttowardsthe higher plasma temperature, above 2×108K, and are also the observer. mostvisibleinX-rayview(despiteC4beingpartiallyhiddenby InX-ray,mostofthebubbles(excepttheyoungC4andC5, the brighter B5). Given that the cavities are physically distinct, whichstillretaintheirelongatedlobeshape)areroughlyspher- onewouldexpectthatinthisscenariothesizeandshapeofthe ical. The inner structure of the cavity is visible only in temper- cavities can be straightforwardly extracted from the X-ray ature; a further indication that the shape of the cavities in the observations. However, since most of the structure is concen- X-rayisnotnecessarilyagoodindicatorofthephysicalnature tratedinthecentral100kpc,projectionoverlapsarelikely(e.g. ofthecavity.Evenvortex-ring-likestructuresmayappearalmost C4 and B5 in the X-ray map, or C5 and C6 in the X-ray and sphericalinX-rayimages. temperaturemaps). Thepressurepanelprovidesagoodviewoftheyoungestjet The bow-shocks (or their slowed-down remnants) are beam (within C5), with visible individual recollimation shocks detectable for longer than 100Myr in both the X-ray and pres- andterminalhotspots.Thisshowsthatshocksandjetsarestill suremapsasratherbright,sharp,cleanfronts,exceptwherethey effectivelymovingX-raygasawayfromthecentral100kpc,and interactwithpre-existingbubbles.Onecandistinguishalmostall withasolidanglecoverageofalmost4π. of them, from B2–B6 (B4 is not visible in this particular snap- The large angle misalignment between the young jet beam shotduetoitsoverlapwiththeB3feature,butitisclearatearlier andC4cavitiesseeninthepressurepanelcloselyresemblesthe times).Thehighbrightnessofthebow-shocksmayhoweveraf- jet and cavity structure seen in galaxy cluster RBS 79711. Gitti fectthemeasurementsofthegasprofile,andthereforehavetobe etal.(2006)andDoriaetal.(2012)interpretthecombinationof subtractedcarefullyfromtheimage;butthisisusuallynotanis- radio and X-ray data as suggesting that over the course of the sueifmulti-wavelengthX-raydataareavailable(e.g.Churazov three identified outbursts, the jet axis appears to have changed etal.2016). directionby∼90◦betweenoutbursts. Theimagesgeneratedfromrun2030haveanumberofsim- ilarities with deep Chandra X-ray observations of the cores of Perseus6,7byFabianetal.(2000,2011),NGC58138byRandall 5. Zoom-inX-rayandprojectioneffects et al. (2015), and especially M879,10 by Forman et al. (2005, We now present some more detailed X-ray views of the cluster 2017). The deep X-ray image of the M87 reveals a series of core, which also exemplify the projection effect-related uncer- loops and cavities that are thought to have been produced by a taintiesinthatcrowdedregion.InFig.5,weshowtwodifferent seriesofoutburstsbyaswivellingjet. projectionssidebyside,fromthe x(left)andy(right)direction, Finally, we point out that the pressure map for run 2030 for each simulation run except the trivial case of 0000, where shows good qualitative agreement with pressure disturbances verylittlestructureispresentintheinner200kpc.Thepictures seeninPerseusFabianetal.(2011). are identical in all other respects. Labels are put next to each feature,tofacilitatedirectcomparisonwithFigs.1–4. 4.4. Run0090 Each panel shows a zoom-in X-ray image of the central 200kpc.Thistime,weusepyXSIMtosimulatetheACIS-Ide- The bubbles in run 0090 (Fig. 4) are clearly detached and tector (response, point spread function, field of view) on board evolvealmostcompletelyindependently;eachjetcreatesitsown theChandraX-raytelescope.Asaside-effectofthisoperation, bow-shock, and undergoes all evolutionary stages for single jet the photon counts are now overall lower, but the imprint of the eventsdescribedinSect.3.1,virtuallyneverinteractingwithpre- simulation grid pattern is lost. The zoom-in also reveals some existing cavities. This is most clearly seen in the temperature moredetailsonthecoreandtheinnermostcavitystructure;for and pressure views while in the X-ray, projection uncertainties example,inallimages,theB5andB6featuresshowstructurein arestillsignificant,sosomecavitiesappearjoinedtogether(e.g. theirshockfronts.Mostoftheearliergenerationofcavitiesare the C3 and C4 features – and possibly C2 – can be mistaken nowmostlyinvisible. forasinglelargebubble).Thecomplexnetworkofoverlapping In run 0030, the differences between x and y are minimal 6 http://chandra.harvard.edu/photo/2000/perseus/more. (e.g.C5andC6areroughlyinthesameplaces),asthelatesttwo html jetshavesimilarinclinationwithrespecttothexandyaxis.This 7 http://chandra.si.edu/photo/2005/perseus/ is expected, as moderate re-orientation always produces struc- 8 http://chandra.harvard.edu/photo/2015/ngc5813/ turealignedwiththezaxis,andbothviewshavethesameorien- 9 http://chandra.harvard.edu/photo/2006/m87/ tationwithrespecttothat. 10 http://chandra.harvard.edu/photo/2008/m87/m87_xray. jpg 11 http://www.evlbi.org/gallery/RBS797.png A58,page9of22 A&A617,A58(2018) Fig. 5. Synthetic X-ray observations, similar to theonesappearinginFigs.2(run0030)to4(run 0090), but zooming into the innermost 200kpc and simulating the ACIS-I detector. Projections are along the x (left column) and y axes (right column).ThelabelsmatchtheonesinFigs.2–4, fordirectcomparison.Wenotetheincreasedde- tailsatthecavityboundaryandhowthedifferent perspective changes the apparent volume of the cavities. In the 2030 and 0090 cases, there is no clear axisymmetry, usedtoinferjetproperties,especiallywithalower-qualityimage so in some projections, smaller bubbles blend into the already suchastheoneinFig.4. dark background of the older ones. In run 2030, for example, We also highlight the relationship between the degree of C5andC6appeardistinctinthe xprojection(separatedbyB6), jet re-orientation and the brightness of the core: the greater the but largely overlap in the y projection. B6 is still visible as it re-orientation, the brighter the core. The core brightness is an is a young and bright feature, but it is difficult to straightfor- indicatorofshocksandhot,shockedgas.Aswehavenotedpre- wardlyassociatethisfeaturewithitscorrespondingbubble.Ifthe viously, moderate-to-strongly re-orientating jets tend to inflate overlapisonlypartial,compoundcavitiesmayappearassingle, cavitiesthataremorelocalisedtotheclustercore(thelabelsin irregularcavities(e.g. C5andC6 inrun2030/y projection,and the panels for runs 2030 and 0090 reach lower numbers), and C3andC4inrun0090/yprojection). hence, this is also where the shocks associated with the inflat- Young bubbles can also be outshone by the high brightness ingcavitiesappear.Thenetresultismoreefficientheatingofthe of their own bow-shock, if the line of sight is close enough to core gas, which has deep implications for the halo stability, as thatjet’saxis(asinrun0090,whereB5almostcompletelyhides weshowinSect.7. C5inthe xprojection).Inthelattercase,onlyatinyportionof Figure6showsaviewinaharderX-rayband,[10, 30]keV, thecavitycanbeidentified,soitsvolumeandshapecannotbe of the same central 200kpc, from the x direction. The images A58,page10of22