Powerful quasar outflow in a massive disc galaxy at z∼5
MMon. Not. R. Astron. Soc. , 000–000 (0000) Printed 15 October 2018 (MN L A TEX style file v2.2)
Powerful quasar outflow in a massive disc galaxy at z ∼ Michael Curtis ? , Debora Sijacki Institute of Astronomy and Kavli Institute for Cosmology, University of Cambridge, Madingley Road, Cambridge CB
HA, UK
15 October 2018
ABSTRACT
There is growing observational evidence of high-redshift quasars launching energetic,fast outflows, but the effects that these have on their host galaxies is poorly under-stood. We employ the moving-mesh code
AREPO to study the feedback from a quasarthat has grown to ∼ M (cid:12) by z ∼ Key words: methods: numerical – black hole physics – cosmology: theory – cosmol-ogy: galaxy formation
High redshift quasars, thought to be powered by accre-tion processes on to supermassive black holes (Lynden-Bell1969), are unique probes not only of black hole growth inthe early Universe but also of galaxy formation in extremeenvironments. Assembly of black holes with masses in excessof 10 M (cid:12) in less than 1 Gyr of cosmic time (Mortlock et al.2011) requires copious amounts of low angular momentumgas to be channelled into the innermost regions of a galaxyto sustain high accretion rates comparable to or even in ex-cess of the Eddington limit. Some of the most promisingsites for this physical process to take place are very massiveproto-clusters embedded within collapsing large-scale over-densities and surrounded by a rich web of filaments (e.g.Springel et al. 2005b; Sijacki et al. 2009; Di Matteo et al.2012; Dubois et al. 2012; Costa et al. 2014). Even if a verysmall fraction of the quasar luminosity couples to the sur-rounding matter this can lead to powerful outflows, whichhave indeed been recently observed in distant quasars (e.g.Maiolino et al. 2012; Cicone et al. 2015).A crucial question that then arises is how these out-flows interact with the host galaxy and affect its gas content,star formation rate and morphology. The quasar-driven out-flows are sufficiently energetic that they may heat and un-bind a large fraction of the gas in a galaxy (Silk & Rees ? E-mail: [email protected]
We use the finite volume moving mesh code
AREPO (Springel 2010) which adopts the TreePM approach for grav-ity and a Voronoi mesh to discretise the fluid. We selectone of the most massive haloes from the Millennium simu- c (cid:13) a r X i v : . [ a s t r o - ph . GA ] J a n Curtis & Sijacki
Simulation type M (M (cid:12) ) M DM , (M (cid:12) ) M star , (M (cid:12) ) M gas , (M (cid:12) ) R (kpc) SFR (M (cid:12) yr − )Default 8 . × . × . × . ×
108 96.5Refinement 8 . × . × . × . ×
108 397
Table 1.
Properties of the halo hosting the largest supermassive black hole at z = 4 .
9. We list the total virial mass, as well as the massin dark matter, stars and gas together with the virial radius and the star formation rate integrated over the whole halo. lation (Springel et al. 2005b) and resimulate it at high res-olution, including baryons as in Sijacki et al. (2009); Costaet al. (2014). The spatial and mass resolutions of our sim-ulation are: (cid:15) grav = 1 kpc (comoving), m DM = 8 × M (cid:12) and m gas = 10 M (cid:12) , where m gas is the mean cell mass atthe beginning of the simulation. We include primordial gascooling and heating and a UV background as in Sijacki et al.(2009) and a sub-grid model for star formation and associ-ated feedback (Springel & Hernquist 2003). No galactic out-flows are explicitly included. We seed 10 M (cid:12) h − black holesin haloes with 10 M (cid:12) h − which then grow at the Bondirate, capped at the Eddington limit, and through black holemergers. Feedback consists of thermal energy which is in-jected into the surrounding gas cells within the black holesmoothing length and weighted by mass. In both simula-tions this region contains 64 times the mean gas cell mass.For full details of the black hole model see Springel et al.(2005a); Sijacki et al. (2009); Curtis & Sijacki (2015).We present the results of two otherwise identical simula-tions, one of which adopts the super-Lagrangian refinementmethod detailed in Curtis & Sijacki (2015), which allowsus to increase the resolution of our simulations around theblack holes. To do this, we split and merge cells over anadaptive region defined by the smoothing length of eachblack hole (which is typically around 1 kpc for the largestblack hole), forcing the innermost cells to be the size of theBondi radius of each black hole. For z (cid:46) ∼ . (cid:12) . The region of refinement itselfis typically ∼ After the first black holes are seeded at z ∼
15, a subsetof simulated black holes begin growing rapidly for z (cid:46) (cid:12) yr − at z = 7. Theblack hole growth and host galaxy properties in the simula-tions with and without the super-Lagrangian refinement are very similar during this phase. The differences begin to oc-cur during the end of the Eddington growth. At this point,the black hole has reached a sufficiently high mass to cut offits own accretion flow, which for the most massive black holein our simulations happens at z ∼
7. We find that the natureof this initial shut off and the manner in which the feedbackcontinues to regulate the black hole growth after this phaseis important in determining the galaxy properties, and wefocus our subsequent analysis on this later period. It is worthnoting that in the simulation with super-Lagrangian refine-ment the black hole growth is somewhat more efficient for z (cid:54) . z ∼
5, reaching 1 . × M (cid:12) . The bolo-metric luminosity of the quasar in this period varies between10 and 10 erg s − .In Fig. 1 we show gas temperature maps centred onthe largest halo in our simulation (detailed halo proper-ties are listed in Table 1). Here, the rich web of cold fila-ments is feeding the halo with gas from large scales, followingthe distribution of the dark matter. As the filaments reachthe virial radius the surrounding gas rises in temperature,from ∼ K to much higher temperatures of around a few10 K. This is mostly the result of the gas virially shock-ing, in line with the classical analytical models of galaxyformation (Rees & Ostriker 1977; Silk 1977; White & Frenk1991), whereby the cooling radius of the in-falling gas ismuch smaller than the virial radius, allowing for a quasi-hydrostatic atmosphere to form in the halo. In addition tothis, feedback from the central black hole heats the gas to ahigh temperature of a few 10 K. In line with previous simu-lations (see e.g. Kereˇs et al. 2005; Ocvirk et al. 2008; Nelsonet al. 2013) the filaments are not totally disrupted at thevirial radius and the cold filamentary accretion continuesdown to smaller scales (Di Matteo et al. 2012; Dubois et al.2012; Costa et al. 2014). While the filaments are largely dis-rupted roughly half-way through the halo, the residual ac-cretion of the cold gas persists across the successively smallerscales shown in Fig. 1, connecting large scale filaments withthe central galaxy at the bottom of the potential well.The bottom right-hand panel of Fig. 1 shows the gastemperature of a slice through the centre of the halo. Thegas that has reached this far into the halo has circularizedinto a massive cold gaseous disc, which forms not long afterthe end of the Eddington limited phase of accretion, and isin place by z ∼
6. Detailed galaxy properties are listed inTable 2. Note that the axis of rotation of the disc is largelyperpendicular to the orientation of the main filaments. Par-allel to the rotational axis of the disc is the hot, outflowinggas that has been launched by the powerful feedback fromthe central supermassive black hole. This is entirely hydro-dynamically driven - the black hole injects thermal energyinto the gas in the very central region which then expandsand rises in the comparatively cool medium of the disc and c (cid:13)000
6. Detailed galaxy properties are listed inTable 2. Note that the axis of rotation of the disc is largelyperpendicular to the orientation of the main filaments. Par-allel to the rotational axis of the disc is the hot, outflowinggas that has been launched by the powerful feedback fromthe central supermassive black hole. This is entirely hydro-dynamically driven - the black hole injects thermal energyinto the gas in the very central region which then expandsand rises in the comparatively cool medium of the disc and c (cid:13)000 , 000–000 uasar outflow in a disc galaxy at z ∼ − −
10 0 10 20 x [kpc] − − z [ k p c ] z = . . . . . . . . . P r o j e c t e d T [ K ] − −
20 0 20 40 x [kpc] − − z [ k p c ] z = . . . . . . . . P r o j e c t e d T [ K ] -10 -5 0 5 10 x [kpc] -10-50510 z [ k p c ] z = . . . . . . . . T [ K ] − −
100 0 100 200 x [kpc] − − z [ k p c ] z = . . . . . . . . . P r o j e c t e d T [ K ] − −
10 0 10 20 x [kpc] − − z [ k p c ] z = . . . . . . . . . P r o j e c t e d T [ K ] − −
100 0 100 200 x [kpc] − − z [ k p c ] z = . . . . . . . . . P r o j e c t e d T [ K ] -10 -5 0 5 10 x [kpc] -10-50510 z [ k p c ] z = . . . . . . . . T [ K ] − −
20 0 20 40 x [kpc] − − z [ k p c ] z = . . . . . . . . P r o j e c t e d T [ K ] − −
100 0 100 200 x [kpc] − − z [ k p c ] z = . . . . . . . . . P r o j e c t e d T [ K ] -10 -5 0 5 10 x [kpc] -10-50510 z [ k p c ] z = . . . . . . . . T [ K ] − −
20 0 20 40 x [kpc] − − z [ k p c ] z = . . . . . . . . P r o j e c t e d T [ K ] − −
10 0 10 20 x [kpc] − − z [ k p c ] z = . . . . . . . . . P r o j e c t e d T [ K ] − −
10 0 10 20 x [kpc] − − z [ k p c ] z = . . . . . . . . . P r o j e c t e d T [ K ] − −
100 0 100 200 x [kpc] − − z [ k p c ] z = . . . . . . . . . P r o j e c t e d T [ K ] − −
20 0 20 40 x [kpc] − − z [ k p c ] z = . . . . . . . . P r o j e c t e d T [ K ] -10 -5 0 5 10 x [kpc] -10-50510 z [ k p c ] z = . . . . . . . . T [ K ] Figure 1.
The temperature distribution of the gas surrounding the largest supermassive black hole at z = 5 .
3. In the top row we showthe projected temperature, mass-averaged over a slice 50 kpc thick, whilst in the bottom left panel the projection is averaged over 25 kpc.The velocity field of the gas is over-plotted with arrows, with the mean size of the arrows representing a velocity of 250 km s − . In thebottom right-hand panel we show the gas temperature of a slice through the centre of the galaxy. Here we can see a cold gaseous discwhose net angular momentum axis is aligned with that of a hot, fast-moving bipolar outflow, launched by the central black hole. surrounding warm gas. The peak velocity of the outflow,2800 km s − , occurs at 5 kpc from the black hole, as the cen-tre of the outflow accelerates slightly, before dropping off.The mass averaged velocity of the outflowing gas, however,is fastest in the centre of the outflow and at its base wherethe pressure gradient is strongest.The cold gaseous disc shown here is not present in oursimulation without super-Lagrangian refinement - a simi-lar structure forms at the same time, i.e. at z ∼
7, butin the case without refinement, the cold gas is completelyoverwhelmed by the feedback from the black hole. This isparticularly interesting given that, in the refinement simula-tion, the black hole grows slightly faster and, as such, morecumulative feedback energy has been injected into the gas by z ∼
5. This underlines the importance of resolving the innerregion of the outflowing gas - if the resolution is insufficient(both spatially or temporally) then the feedback will blowaway the cold gas before the hot quasar-driven wind is ableto rise out of the galaxy and subsequent predictions aboutthe morphology of the host galaxy will likely be incorrect.
In Fig. 2, we show the gas mass distribution as a func-tion of its velocity for both simulations. Our intent here isto investigate what difference the use of refinement has onthe kinematic properties of the gas, even at distances muchlarger than those at which we increase the resolution, i.e. at ∼ z = 4 .
9, butthe results are very similar from z = 7 . × M (cid:12) within25kpc, its momentum flux is 5 −
40 Lc − , while the outflowrate is ∼
700 M (cid:12) yr − (if we consider speeds higher than400 km s − this number drops to ∼
300 M (cid:12) yr − ).In the right-hand panel of Fig. 2 we show a similar plot,but for the inflowing gas. There is a significant difference inthe mass distribution at low radii - the non-refinement simu-lation exhibits little inflow across all velocity bins, whilst the c (cid:13) , 000–000 Curtis & Sijacki v rad , outflow [km/s] M [ M (cid:12) ] v rad , inflow [km/s] M [ M (cid:12) ] No Refinement: 0 kpc ≤ r ≤ ≤ r ≤
15 kpcNo Refinement: 20 kpc ≤ r ≤
25 kpcRefinement: 0 kpc ≤ r ≤ ≤ r ≤
15 kpcRefinement: 20 kpc ≤ r ≤
25 kpc
Figure 2.
The velocity distribution of the gas for simulations with refinement (blue) and without (red) at z = 5 .
3. In the left-handpanel, we show the mass of outflowing gas as a function of speed while in the right-hand panel, we plot the mass of gas with inflowingradial orbits. In both, we show the relevant quantity summed over three spherical shells.
Disc R visual R mass / M ( < R visual ) SFR( < R visual )(kpc) (kpc) (10 M (cid:12) ) (M (cid:12) yr − )Gas 3.5 2.05 3.53 332Stellar 3.5 0.86 27.5 - Table 2.
Galaxy properties at z = 4 . refinement simulation shows a much larger mass of inflow-ing gas at lower velocities. This difference is still sizeable at10 kpc out from the black hole, an order of magnitude out-side of our refinement region. By the time we reach 20 kpc,the two simulations agree well. The difference, in both cases,is a reflection of the simulation resolution and hence the spa-tial distribution of the outflow. With refinement, the outflowis tightly collimated with a small opening angle (as the cellsare not enforced to maintain approximately constant mass),and the gas rises vertically out of the galaxy without inter-acting much with the cold disc of accreting gas. Withoutrefinement the hot gas expands, leading to a very coarseeffective resolution and it interacts with the inflowing gas,reducing both the inflow and outflow at small radii. We now focus on the morphological properties of the quasarhost galaxy in our simulation with refinement. The blackhole sits at the centre of an exponential stellar disc, witha prominent and kinematically distinct central bulge com-ponent. The M BH − M bulge ratio for our galaxy is 0 . . × M (cid:12) at z = 4 should on average host a cen-tral galaxy with a stellar mass of ∼ M (cid:12) , indicating thatwhile we overproduce around twice as many stars as pre-dicted by the mean relation we are within the 1 σ scatter,which has an upper bound of 2 . × M (cid:12) . We do not in-clude strong stellar feedback in our simulations in the form of energetic supernova-driven outflows, which may likely bringour simulated disc stellar mass closer to the mean relation.In the left-hand panel of Fig. 3 we show the rotationcurves of different components, with stellar bulge dominat-ing in the centre, followed by the stellar disc at the inter-mediate range of radii and finally by dark matter outside ofthe galaxy. In the right-hand panel we plot the tangentialvelocity of the gas divided by the sound speed of the gasand the radial velocity dispersion of the gas, respectively.The tangential velocity of the stars divided by their radialvelocity dispersion is shown as well, indicating much largervelocity dispersion support. The gas disc is aligned with thestellar component and covers a similar radial extent. Thesurface density profile is however not well modelled by anexponential, especially in the inner region. In Fig. 4 we showthe projected line of sight velocity of the gas centred on thegalaxy. There is a clear signature of the rotating gas discboth when viewed face-on and edge-on with typical veloci-ties of 200 km s − and 700 km s − , respectively. Recent observations have started to shed light on the proper-ties of quasar feedback at high redshifts and several studieshave found evidence of large-scale outflows (e.g Maiolinoet al. 2012; Cicone et al. 2015). However, there are cur-rently very few observed examples which can constrain indetail both the quasar outflow and the host galaxy prop-erties, especially at high redshift. A notable exception, butin the local Universe, is Mrk 231, a ULIRG galaxy with a5 × erg s − quasar at its core. The inferred star forma-tion rate of the galaxy of 200M (cid:12) yr − , its regular rotationpattern together with the outflow rate of 700M (cid:12) yr − andvelocity of 750km s − (Feruglio et al. 2010) agrees very wellwith our results, indicating that our simulated system couldbe a high redshift counterpart of Mrk 231. Recent ALMAobservations by Carniani et al. (2013) have found a quasar- SMG pair at z ∼ .
7, where host galaxies exhibit a rota-tionally supported geometry with a range of velocities anddisc sizes similar to our findings, albeit with no detectedquasar outflow. Moreover, ALMA observations of five lumi-nous quasars at z ∼ c (cid:13)000
7, where host galaxies exhibit a rota-tionally supported geometry with a range of velocities anddisc sizes similar to our findings, albeit with no detectedquasar outflow. Moreover, ALMA observations of five lumi-nous quasars at z ∼ c (cid:13)000 , 000–000 uasar outflow in a disc galaxy at z ∼ r [kpc] v [ k m / s ] z = TotalDark MatterGasStarsStars (bulge) r [kpc] R o t a t i o n a l S u pp o r t v t , gas /c s , gas v t , gas /σ r , gas v t , stars /σ r , stars Figure 3.
Left: rotation curves of different components, based on the enclosed mass. Right: the tangential speed of the gas divided bythe sound speed (red), and by the enclosed radial velocity dispersion of the gas (blue). The tangential speed of the stars divided by theenclosed radial velocity dispersion of the stars is shown in green. Lines denote the mean and the shaded area is 1 σ scatter. − − − x [kpc] − − − z [ k p c ] z = − − − − P r o j e c t e d v [ k m / s ] − − − x [kpc] − − − y [ k p c ] z = − − − P r o j e c t e d v [ k m / s ] − − − x [kpc] − − − y [ k p c ] z = − − − P r o j e c t e d v [ k m / s ] − − − x [kpc] − − − z [ k p c ] z = − − − − P r o j e c t e d v [ k m / s ] Figure 4.
Gas velocity maps of the central galaxy in a face-on (left) and edge-on (right) projection. The projection depth is 13 kpc.While the rotational signature is weak for the face-on view, it is clearly detectable for the edge-on view with speeds up to ∼
700 km s − . black hole masses and host galaxy dynamical masses brack-eting our results, find evidence of velocity gradients indica-tive of rotationally supported discs. It is worth noting how-ever that observed quasar host galaxies are more likely tobe viewed face-on, and this selection bias needs to be takeninto account when comparing to our simulation results.The results presented in this Letter indicate that a morecareful treatment of the black hole feedback injection canlead to feedback having less of an impact on galaxy proper-ties (see also Dubois et al. 2015; Feng et al. 2015), at leastconcerning the central galaxy over limited periods of time.However, the injected energy can have a longer term impact- it will heat the gas outside of the central galaxy in thecircum-galactic and inter-galactic medium. This will affectthe subsequent gas cooling on to the galaxy which, if there issufficient energy, may ultimately lead to the starvation andquenching of star formation.In light of this, it will be increasingly important in fu-ture work to attempt to model the ISM, ideally down toparsec scales and below. High density clumps within theISM will be more resistant to destruction by the centralengine, and this will further complicate the picture of how the feedback energy couples to the surrounding medium andquenches star formation. Future ALMA and JWST obser-vations will be crucial in this regard, as they will allow de-tailed comparison with the theoretical models and also pro-vide larger statistical samples to pin down the morphologicalvariety of galaxies hosting powerful quasar outflows. ACKNOWLEDGEMENTS
We thank Roberto Maiolino, Martin Haehnelt and EwaldPuchwein for many helpful suggestions on the manuscript.MC is supported by the STFC and DS acknowledges sup-port by the ERC Starting Grant 638707 “Black holes andtheir host galaxies: co-evolution across cosmic time”. Thiswork was performed on: DiRAC Darwin Supercomputer(University of Cambridge HPCS; Higher Education Fund-ing Council for England and STFC); DiRAC Complexitysystem (University of Leicester IT Services; BIS NationalE-Infrastructure grant ST/K000373/1 and STFC DiRACgrant ST/K0003259/1); the COSMA Data Centric system(Durham University; BIS National E-infrastructure grantST/K00042X/1, STFC grant ST/K00087X/1, DiRAC Op- c (cid:13) , 000–000 Curtis & Sijacki erations grant ST/K003267/1 and Durham University).DiRAC is part of the National E-Infrastructure.
REFERENCES
Bourne M. A., Nayakshin S., Hobbs A., 2014, MNRAS,441, 3055Carniani S. et al., 2013, A&A, 559, A29Cicone C. et al., 2015, A&A, 574, A14Costa T., Sijacki D., Haehnelt M. G., 2014, MNRAS, 444,2355Curtis M., Sijacki D., 2015, MNRAS, 454, 3445Debuhr J., Quataert E., Ma C.-P., 2012, MNRAS, 420,2221Di Matteo T., Khandai N., DeGraf C., Feng Y., CroftR. A. C., Lopez J., Springel V., 2012, ApJ, 745, L29Dubois Y., Pichon C., Haehnelt M., Kimm T., Slyz A.,Devriendt J., Pogosyan D., 2012, MNRAS, 423, 3616Dubois Y., Volonteri M., Silk J., Devriendt J., Slyz A.,Teyssier R., 2015, MNRAS, 452, 1502Fabian A. C., 1999, MNRAS, 308, L39Feng Y., Di Matteo T., Croft R., Tenneti A., Bird S.,Battaglia N., Wilkins S., 2015, ApJ, 808, L17Feruglio C., Maiolino R., Piconcelli E., Menci N., AusselH., Lamastra A., Fiore F., 2010, A&A, 518, L155Kereˇs D., Katz N., Weinberg D. H., Dav´e R., 2005, MN-RAS, 363, 2King A., 2005, ApJ, 635, L121Kormendy J., Ho L. C., 2013, ARA&A, 51, 511Lynden-Bell D., 1969, Nature, 223, 690Maiolino R. et al., 2012, MNRAS, 425, L66Mortlock D. J. et al., 2011, Nature, 474, 616Moster B. P., Naab T., White S. D. M., 2013, MNRAS,428, 3121Nelson D., Vogelsberger M., Genel S., Sijacki D., Kereˇs D.,Springel V., Hernquist L., 2013, MNRAS, 429, 3353Ocvirk P., Pichon C., Teyssier R., 2008, MNRAS, 390, 1326Rees M. J., Ostriker J. P., 1977, MNRAS, 179, 541Roos O., Juneau S., Bournaud F., Gabor J. M., 2015, ApJ,800, 19Sijacki D., Springel V., Haehnelt M. G., 2009, MNRAS,400, 100Silk J., 1977, ApJ, 211, 638Silk J., 2013, ApJ, 772, 112Silk J., Rees M. J., 1998, A&A, 331, L1Springel V., 2010, MNRAS, 401, 791Springel V., Di Matteo T., Hernquist L., 2005a, MNRAS,361, 776Springel V., Hernquist L., 2003, MNRAS, 339, 289Springel V. et al., 2005b, Nature, 435, 629Wang R. et al., 2013, ApJ, 773, 44White S. D. M., Frenk C. S., 1991, ApJ, 379, 52 c (cid:13)000