How AGN and SNe feedback affect mass transport and black hole growth in high redshift galaxies
By using cosmological hydrodynamical simulations we study the effect of supernova (SN) and active galactic nuclei (AGN) feedback on the mass transport of gas on to galactic nuclei and the black hole (BH) growth down to redshift . We study the BH growth in relation with the mass transport processes associated with gravity and pressure torques, and how they are modified by feedback. Cosmological gas funelled through cold flows reaches the galactic outer region close to free-fall. Then torques associated to pressure triggered by gas turbulent motions produced in the circum-galactic medium by shocks and explosions from SNe are the main source of mass transport beyond the central 100 pc. Due to high concentrations of mass in the central galactic region, gravitational torques tend to be more important at high redshift. The combined effect of almost free-falling material and both gravity and pressure torques produces a mass accretion rate of order M/yr at pc scales. In the absence of SN feedback, AGN feedback alone does not affect significantly either star formation or BH growth until the BH reaches a sufficiently high mass of M to self-regulate. SN feedback alone, instead, decreases both stellar and BH growth. Finally, SN and AGN feedback in tandem efficiently quench the BH growth, while star formation remains at the levels set by SN feedback alone due to the small final BH mass, few M. SNe create a more rarefied and hot environment where energy injection from the central AGN can accelerate the gas further.
Subject headings:galaxies: formation — large-scale structure of the universe — stars: formation — turbulence.
There is dynamical evidence for the existence of super massive BHs in the center of many nearby galaxies (Ferrarese & Ford, 2005) with masses in the range M suggesting that BHs, formed at the first evolutionary stages of our Universe, now are living in the galactic centers around us including our galaxy (Ghez et al., 2005).
Scaling relations connect the mass of BHs in local galaxies with their host galaxy properties, such as, e.g., the galactic bulge mass (e.g. Magorrian et al., 1998; Gültekin et al., 2009), and the bulge stars velocity dispersion (e.g. Ferrarese & Merritt, 2000; Tremaine et al., 2002). Such relations suggest a co-evolution between the BH and its host galaxy (see for instance Dressler, 1989; Kormendy & Richstone, 1995; Magorrian et al., 1998; Gebhardt et al., 2000), with a recent work suggesting that such a co-evolution can be trigger in galactic bulges with a critical mass above M in the early stages of the galactic evolution (Park et al., 2016).
Detection of very bright quasars at redshift with luminosities above L implies the existence of BH with masses of the order M when our Universe was about Gyr old (Fan et al., 2001; Willott, 2007; Mortlock et al., 2011; De Rosa et al., 2014), i.e. BHs should have formed very early in the history of our Universe and grow rapidly in order to attain such high masses after Gyr (Dubois et al., 2012; Di Matteo et al., 2012, 2016). To understand such a rapid early evolution is one of the main challenges of the current galaxy formation theories (Volonteri, 2010; Haiman, 2013).
In previous work, Prieto & Escala (2016) (hereafter PE16) have studied the mass transport (MT) process in high redshift galactic discs focusing on the effect of SN feedback without taking into account AGN activity on such objects. They show that SN feedback is able to affect dramatically the BH growth in the first galaxies at high redshift (see also Dubois et al., 2015; Habouzit et al., 2016; Bower et al., 2016).
As in PE16, we study the evolution of a high redshift galaxy. In particular, we compute internal dynamical properties of the system to study the effect of AGN activity in the MT process in galactic discs at redshift . Outflows generated by AGN activity can have spatial extensions of at least kpc and can reach velocities of the order of km/s (e.g. Humphrey et al., 2010; Nesvadba et al., 2011; Arribas et al., 2014; Genzel et al., 2014). Such powerful outflows should be capable of affecting the host galaxy evolution as well as the BH growth. We explore the individual effect of SN and AGN in the galaxy evolution and, furthermore, we study how super-Eddington accretion may affect the system.
The paper is organized as follows. Section §2 contains the numerical details of our simulations. In section §3 we present our results based on the gas dynamic analysis of our simulations and its effect the on the BH growth. In section §4 we discuss them, and present our main conclusions.
2. Methodology and Numerical Simulation Details
The simulations analysed in this paper are an extension of the work presented in PE16 with a few modifications. Therefore, we briefly recall the physical ingredients of the numerical experiments.
The simulations were performed with the cosmological N-body hydrodynamical code RAMSES (Teyssier, 2002). The code uses adaptive mesh refinement, and solves the Euler equations with a second-order Godunov method and MUSCL scheme using a MinMod total variation diminishing scheme to reconstruct the cell centered values at cell interfaces.
Cosmological initial conditions were generated with the MPGRAFIC code (Prunet et al., 2008) inside a cMpc side box. Cosmological parameters were taken from Planck Collaboration (2013) with total matter density , dark energy density , baryon matter density , reduced Hubble parameter , amplitude of the power spectrum at scale of and spectral index .
We have selected a M DM halo at to be re-simulated at high resolution from to . The high-resolution particles have mass M (which corresponds to an effective resolution of DM particles inside the box). Such a mass allows us to resolve our final halo with DM particles. In order to resolve all the interesting regions, we allowed refinements inside the Lagrangian patch associated with a sphere of radius around the selected DM halo at (here is the DM halo virial radius, defined as the radius associated to a spherical overdensity 200 times that of the mean matter density of the Universe at the corresponding redshift). The Lagrangian volume (the mask) is defined by an additional scalar advected passively with the flow throughout the simulation. At the beginning the passive scalar has a value equal to 1 inside the mask and 0 outside. In regions where this passive scalar is larger than we allow refinement in a cell if any of these conditions is met: i) its total mass is more than 8 times that of the initial mass resolution, ii) the Jeans length is resolved by less than 4 cells (Truelove et al., 1997), iii) if the relative pressure variation between cells is larger than a factor of 2. Following these criteria, we reach a maximum co-moving spatial resolution of cpc and a proper spatial resolution of pc at redshift , whereas the coarse grid inside the mask has a resolution of ckpc. The gravitational force resolution is throughout the simulation.
Our simulations include optically thin (no self-shielding) gas cooling following the Sutherland & Dopita (1993) model down to temperature K with a contribution from metals assuming a primordial composition of the various heavy elements. Below this temperature, the gas can cool down to K due to metal lines cooling (Dalgarno & MacCray, 1972). We started the simulations with an initial metallicity of Z (Powell et al., 2011). A uniform UV background is activated at , following Haardt & Madau (1996).
We adopted a star formation number density threshold of H cm with a star formation efficiency (e.g. Rasera & Teyssier, 2006; Dubois & Teyssier, 2008). When a cell reaches the conditions for star formation, star particles can be spawned following a Poisson distribution with a mass resolution of . In order to ensure numerical stability we do not allow to convert more than of the gas into stars inside a cell in one time step.
After 10 Myr the most massive stars explode as SN releasing a specific energy of erg/M, returning 10 per cent of the stellar particle mass back into the gas and with a yield of inside a sphere of . As in PE16 we used the delayed cooling implementation of SN feedback (Teyssier et al., 2013). It means that, where SNe explode, if the gas “non-thermal” energy (stored in a separate passive variable) is above an energy threshold , gas cooling is turned off in order to take into account the unresolved chaotic turbulent energy source of the explosions. This non-thermal energy component dissipates energy at its own rate with characteristic time . In this work (as in PE16) we use Myr and the energy threshold is the one associated to a turbulent velocity dispersion km/s through the equation , with the gas density. Such a velocity dispersion is appropriate for our spatial resolution (see Dubois et al., 2015; Prieto & Escala, 2016, for details).
In order to follow the evolution of the central BH in the simulations, we introduced a sink particle (Bleuler & Teyssier, 2014) when the main DM halo has a mass M at redshift . The BH seed mass is M. Such a BH mass is in the range of masses associated with direct collapse scenario (e.g. Oh & Haiman, 2002; Lodato & Natarajan, 2006; Begelman et al., 2006, 2008; Agarwal et al., 2012; Latif et al., 2013, 2014; Choi et al., 2015). Only this one single BH is allowed to form in the simulation.
In order to compute the mass accretion rate onto the BH we use the standard Bondi-Hoyle (Bondi, 1952) implementation of the accretion rate, . In all our simulations expect for one, we cap the accretion rate at the Eddington luminosity. For more details on the numerical implementation of the processes described above, see PE16.
In order to avoid spurious oscillations in the sink position we have introduced a drag force acting on the BH particle (Biernacki et al., in prep.). The drag force comes from a new source of momentum acting on the sink. Such a momentum variation is proportional the cell gas-sink particle relative velocity. Furthermore, it depends on the non accreted mass above the Eddington limit: when the BH accretion rate is below the Eddington limit the drag force is null, it works only if the Bondi BH mass accretion rate is above the Eddington limit. In this sense it can be interpreted as a force associated to the Eddington pressure around the BH particle. In the simulation where accretion is not capped at the Eddington limit there is no drag force. However, even in this extreme case the sink particle stays at the center of the galaxy.
We have also included AGN feedback from the BH. AGN feedback is modeled with thermal energy input (Teyssier et al., 2011; Dubois et al., 2012). The rate of energy deposited by the BH inside the injection radius is
In the above expression, is the radiative efficiency for a standard thin accretion disc (Shakura & Sunyaev, 1973) and is the fraction of this energy coupled to the gas in order to reproduce the local BH-galaxy mass relation (Dubois et al., 2012). As explained in Booth & Schaye (2009), in order to avoid gas over-cooling the AGN energy is not released instantaneously every time step but it is accumulated until the surrounding gas temperature can be increased by K. The time between the energy injection events depends on the BH mass, mean gas density, and the mass accretion rate. It is kyr in our simulations.
In this work we will show results from four simulations:
SNe run: it includes star formation, SN feedback, modified Bondi-Hoyle-Lyttleton (BHL) accretion rate onto sinks without AGN feedback. This case was extensively analyzed in PE16 and here is used to compare it with our AGN simulations.
SNeAGN run: same as SNe plus AGN feedback.
NoSNeAGN run: same as SNeAGN but without SN feedback, and
SNeAGNnoEdd run: same as SNeAGN but without capping at the Eddington limit.
Figure 1 shows the gas number density projection for the four simulations at the end of the experiments. Each panel present different features depending on the different feedback recipes as will be explained in the following sections.
3.1. Mass transport on large scales
Because in a cosmological context at high redshift we can not study the small-galactic scale phenomena without taking into account the effects of the large scale structure, here we study the behavior of mass accretion from few hundreds pc up to , with the DM halo virial radius.
Fig. 2 shows the mass accretion rate, computed taking into account all the gas mass crossing a spherical shell at a given radius centred at the sink cell position:
where is the gas mass density and is the total radial gas velocity.
The left column of Fig. 2 shows the total gas mass accretion rate for our four simulations as a function of radius. The right column shows instead the mass accretion rate associated with low gas density , with the critical density of the Universe. For the redshift range shown in the figure, cm. This range of densities is below the mean baryon density in cold flows in our simulations, which is cm. The vertical lines mark the DM virial radius at each sampled redshift.
Beyond the virial radius all our simulations show a mass accretion rate M in agreement with PE16. This mass accretion rate is associated to gas almost free falling on to the DM halo. At these radii the inflowing material is dominated by gas with densities cm. Those are characteristic densities in cooling flows. These quantity shows peaks ( few M) associated to both gas clumps and DM mini haloes inside the virial radius. From the right column, we see how feedback suppresses low density ( cm) gas accretion inside the inner kpc. Because of the feedback heating, only the densest gas is able to flow into the galactic central region. At the end of the simulations, gas with densities cm can reach the outer galactic region () but not beyond, because it is heated and expelled from the galaxy. When we include material with density below 10 cm, it can penetrate the galaxy, but it cannot reach the central few pc. Only the densest gas can reach the galactic central region: material with density below 100 cm is able to reach the central few 100 pc region.
The AGN simulations show that at the low density material can not penetrate inside the virial radius. By this time, AGN activity is capable to evaporate the diffuse material. At higher redshift the diffuse gas can reach smaller radii but the accreted material is dominated by high density gas. In particular, our NoSNeAGN simulation has the deepest low density gas penetration at . In this case due to the no SN heating and the low BH mass (few M), AGN feedback is unable to alter significantly the low density kpc scale accretion. At lower redshift the BH experiences a prolonged Eddington limited accretion rate (as discussed in the following) that increases its mass by a factor of a few, injecting in turn energy in the interstellar and circum-galactic media. Such energy injection clearly affects the low density gas accretion rate at , as we can see from the figure: the low density gas barely reaches the virial radius.
We note that the expression in eq. 2 includes mass inflows associated to inward radial motions triggered by local thermal fluctuations. In order to asses how important are the local thermal fluctuations to the mass accretion rate, we compute the radial velocity dispersion in the no feedback simulation of PE16. Then, for each simulation presented in this work, we compute the mass accretion rate including all the material with radial velocity
i.e. we excluded local thermal fluctuations. Following this procedure we can see that at virial radius scales the local velocity fluctuations are no more than of the total accreted matter.
The total mass accretion rate in our four simulations has similar values at kpc scales fluctuating between M and few tens of M, showing that the effect of the AGN activity has not a notable impact on kpc scales in these small high redshift galaxies, due to the small BH mass. Such accretion rates of order few tens of M at the outer galactic edge are associated to free fall material reaching the central DM halo region from large scales, as discussed in PE16.
3.2. Mass transport in the disc
The tight relation between the large scale dynamic and the small scale evolution in the galaxy formation context (e.g. Pichon et al., 2011; Dubois et al., 2012; Prieto et al., 2015; Danovich et al., 2015) is a robust motivation to study the MT process at different scales in the first galaxies. In the following, we analyze the MT process on kpc () scales.
3.2.1 Torques in the disc
After to flow at almost free fall from scales above till the galactic edge (at ) triggered by gravity and channelized through DM filaments around the DM halo, the angular momentum (AM) re-distribution in the galaxy will produce MT, allowing the gas to flow in and to reach the center of the system. Due to the nature of the system at study, the sources of AM variations are related to gravitational forces and pressure gradients, namely:
The gravitational, , and the pressure gradient, , torques will act as source of AM transport in the galactic system and provide clues about the MT process in high redshift galaxies.
Fig. 3 shows the ratio as a function of radius at different redshifts for our four simulations, with . The data is smoothed over pc in radius. All our simulations show that pressure gradient torques tend to dominate above 100 pc, with a clear domination at almost all radii at lower redshift. This is the consequence of two processes: shocks due to the large scale in-falling gas onto the DM halo central region and shocks due to SNe and AGN feedback. There are, however, some regions of gravity domination above 100 pc. The pressure gradient domination is clearer at smaller radii. Only at high redshift gravity dominates the center of the system. In the case without SN feedback the gravitational torque has a contribution at smaller radii ( few tens pc) compared with the SN feedback simulations. This is due to the higher mass concentration in this simulation. In summary, the combined effect of gravity and pressure gradients re-distributes AM in the disk and triggers MT that feeds the central BH in these systems. We note that due to the mass difference (above one order of magnitude in DM halo) our results does not completely agree with the ones presented in Danovich et al. (2015). These authors show that the gravitational torque dominates on the galactic disc. As mentioned above such a difference arises due to the bigger mass of their studied systems at lower redshift.
3.2.2 Mass accretion rate in the disc
As shown in the previous section, high redshift galaxies receive almost free falling material from large scale and experience torques associated with gravitational forces and pressure gradients which trigger MT in the galaxy. In this context, it is relevant to study and quantify the mass accretion rate in the galactic disc due to such phenomena and on to the central BH.
Fig. 4 shows the radial gas mass accretion rate in the disc as a function of radius inside at different redshifts for our four runs. We defined the mass accretion in the disc as:
We first compute the gas AM vector inside a radius , and take it as the vertical axis of the cylindrical coordinate system of reference. The cylindrical radial coordinate is defined in the disc plane perpendicular to the AM vector, and both the gas surface density and the radial velocity are cylindrical shell averages in the direction, up to a height such that the of the baryonic mass is enclosed.
All simulations show large fluctuations in the mass accretion rate, attesting to the dynamical conditions in these environments: large-scale gas inflows and SN feedback shock the gas creating a turbulent environment. Accretion rates fluctuate between M and M with an average accretion rate of the order M. The accretion rate profiles show a number of gaps associated to SN outflows and gas clumps crossing and leaving the system.
The lack of SN heating in the NoSNeAGN simulation causes smoother accretion rate profiles compared with all the SN feedback simulations. Furthermore, in simulations with AGN feedback gas does not easily reach the central galactic region, pc, i.e. AGN feedback efficiently ejects gas from the galactic center as we will discuss in the next section.
3.3. BH evolution
We have shown that after to flow almost radially at free fall from scales above and reach the galactic outer region, gravitational and pressure gradient torques can produce a substantial mass accretion rate of the order M/yr at distances few 10 pc from the center of the system in high redshift galaxies. We now move to explore the BH mass growth and how dynamical features of the system depend on SN and AGN feedback.
3.3.1 BH accretion rate
We show the BH accretion rate as a function of redshift in Fig. 5. As already shown by PE16, the SN explosions have a clear effect on the central BH mass accretion rate ejecting gas out of system and then decreasing the amount of material that can feed the BH throughout the simulation (see also Dubois et al., 2015). This is especially important at , when the halo mass is M and the stellar mass few M. Under such conditions SN explosions can accelerate a fraction of the central galactic gas beyond the local escape velocity depleting of gas the BH neighborhood. The SN run is characterized by intermittent Eddington limited accretion episodes with an average and a final BH mass (see table 1).
AGN and SN feedback in tandem (SNeAGN) reduce significantly the mass accretion rate, a factor overall, with and a final BH mass . Beside the SN energy injection, the local effect of the AGN activity is able to eject the already low amount of gas from the galactic center, reducing dramatically the mass accretion rate. Notwithstanding, the central BH has a number of Eddington limited episodes.
The NoSNeAGN run shows an interesting behavior. Early on the BH accretes close to the Eddington limit as in the no SN feedback case of PE16 (see the gray solid line in Fig. 5). This behavior continues to , when the now sufficiently massive BH is capable to accelerate the surrounding gas beyond the central escape velocity. The lack of SN heating allows the gas to easily reach the galactic center and pile up to feed the BH. At the same time the accumulation of gas deepens the potential well and increases the central escape velocity, as shown in Fig. 6, making it difficult for gas affected by AGN feedback to leave the central region. At , the BH mass has grown enough that AGN feedback becomes sufficiently strong to quench BH accretion (see the lower-left panel in Fig. 5). Overall, the mean accretion rate, , and, final BH mass, , are in between the cases with only SN or both SN and AGN feedback.
The case without Eddington limit, but with both types of feedback included, SNeAGNnoEdd is characterized by an early peak of super-Eddington accretion at . This early burst, in a tiny galaxy, causes catastrophic feedback that essentially shuts off accretion until . It is important to note that feedback has not been modified to take into account the lower radiative efficiency expected in super-Eddington accretion discs (see Volonteri et al., 2015, and references therein; a simulation following such a scenario is work in progress and will be part of an upcoming paper). A lower radiative efficiency decreases the injected energy from feedback (see Eq. 1) and would be less disruptive on its surroundings (Lupi et al., 2016). In this case the average is with a final mass .
Table 1 shows for all our simulations, for the whole duration of the simulation, as well as only at and to highlight the effects described above. In all simulations including SN feedback, the high redshift () is lower than the low redshift () . Between redshift 6 and 10 the system has become more massive and SN feedback alone cannot unbind the gas. In the simulation without SN feedback, but with AGN feedback, the mean Eddington ratio is higher at higher redshift (): AGN feedback affects the BH accretion rate only when the BH has become sufficiently massive to drive powerful outflows and clear its immediate surroundings.
3.3.2 Dynamical conditions
These simulations show how important outflows are on the dynamics of the central region of high-redshift galaxies. Depending on their power they can change the mass distribution around the BHs and affect the dynamical conditions by changing the local escape velocity. Fig. 6 shows the escape speed for our four simulations at different radii as a function of redshift. The escape speed from a given radius is defined by with the total mass (BH, gas, stars and DM) inside a radius . The NoSNeAGN simulation has the largest escape speed at the injection radius (solid green line, this is the velocity needed for the gas to leave the region from which the BH accretes, not for gas to leave the galaxy or the halo) and at (dashed green line, this can be considered the velocity needed to leave the galaxy). Due to the lack of SN heating, cold dense gas can pile up in the galaxy and in the BH vicinity increasing the escape velocity. Such conditions clearly favor a high accretion rate onto the BH until it reaches a sufficiently high mass for its AGN activity to impact its surroundings.
In contrast, the SNeAGNnoEdd simulation has the lowest escape speed above at the injection radius. In this case the strong AGN feedback associated to the early bursts at super-Eddington rates expels the gas around the BH, creating a very low density environment with a low escape speed (see bottom right panel of Fig. 5), which favors mass depletion in the BH vicinity, and keeps the accretion to minimal levels. At the escape speed increases at similar (but still lower) values compared with the other runs. At this redshift the main DM halo merges with a galaxy which has not been affected by AGN activity, increasing the enclosed central mass and consequently the escape velocity.
The SNe and SNeAGN runs show similar escape speeds at . Their escape speeds have clearer differences at the injection radius. They fluctuate due to the SN and BH energy injection which in both cases heat up and reduce the amount of available gas in the galactic center. Despite the similar escape speed in these two simulations, their BH growth is rather different (see Fig. 5). In these simulations the mass inside is dominated by the stellar component, while accretion depends on gas content around the BH (as we will discuss later), which shows rapid and large fluctuations early on for simulations including SN feedback (Dubois et al., 2015). In the SNeAGN simulation, beside the SN mass outflows, AGN feedback is also at work, reducing substantially the BH growth rate compared with the BH mass accretion rate in the SN only simulation.
Figure 7 shows the gas, stars and DM enclosed mass as a function of redshift for all simulations at three different distances from the BH position, namely , and . In all these simulations baryons dominate the enclosed mass in the BH vicinity, with a major contribution from the stellar component as mentioned above. This figure can explain why the SNeAGN experiment has a lower BH accretion rate compared with the only SN feedback run despite their escape speed being similar: the gas content around the BH in the AGN feedback simulation is lower than the one around the BH in the only SNe simulation. Only the no-Eddington limited experiment shows that DM dominates the central galactic region at high redshift, . In this particular case, the extreme BH feedback is strong enough to expel the central galactic gas and inhibit star formation.
Fig. 8 shows the average mass weighted gas outflow speed normalized by the escape speed, , as a function of redshift for our four simulations at two different radii, namely and . The data is smoothed over Myr. The dashed gray line shows . At the injection radius the NoSNeAGN feedback simulation (green line) creates outflows with speeds well below the escape speed due to the high mass concentration (see Fig. 7) around the BH as mentioned above. At larger radii all the SN feedback simulations have an approximately similar behavior, showing that the stronger effects can be seen in the BH vicinity.
The run without Eddington limit (cyan line) produces nuclear outflows with speeds well above the escape speed at ejecting most of the gas around the BH and quenching efficiently BH growth and central star formation. The combined effect of gas ejection and reduced star formation decreases the central mass concentration, reducing the escape velocity. Under such conditions, feedback can easily unbind the gas, and only the merger at increases the central mass at levels comparable with the other SN simulation runs.
The SNe and SNeAGN simulations have a similar behavior in the innermost region of the galaxy. These simulations have a highly fluctuating behavior with periods of bound and unbound central gas which is reflected in the BH mass accretion rate fig. 5, however the nature of the gas that is accelerated in both cases is different as we will see now.
3.4. Velocity distribution in the galactic gas
Fig. 9 shows the average mass inside radial outflow velocity bins for gas inside colored by the ratio . The mass in each velocity bin is averaged over a time interval Myr and they are shown for three different redshift interval, namely , and . The cooling time scale is defined as where is the gas internal energy and is the gas cooling function, that is both a function of the gas temperature and the gas metallicity. The radial time scale is defined as where r is the radial coordinate and is the gas outflow radial velocity with respect the BH motion. The origin of the system is set at the BH position. The left column of Fig. 9 shows the hot, K, gas and the right column shows the cold, K, gas. The black solid line marks the 1D velocity dispersion of the NoSNe simulation averaged over . This velocity dispersion gives us an idea about an outflow turbulent velocity triggered only by gravitational processes. In general, the feedback simulations presented here develop outflows with .
When we look at the hot gas at high redshift, , the NoSNeAGN run produces slower hot gas compared with all the SN feedback simulations. In this case the AGN activity can not accelerate the hot galactic gas above . In other words, without the SN heating the hot gas can not be accelerated above this velocity during the Eddington-limited growth phase of the BH. SN feedback can create a low density rarefied hot environment where energy injection from SNe (and AGN feedback for the SNeAGN run) can easily accelerate the gas. In contrast, without SN feedback, the heated gas is surrounded by high density gas and it is much more difficult to accelerate it at higher velocities. We note that this fast hot gas is no more than a few percent of the total enclosed gas mass. Note that this result differs quantitatively, but not qualitatively, from Costa et al. (2015) because they simulate a more massive halo of at powering a luminous quasar, while the halo in our simulations is of much lower mass and AGN feedback remains weak, and its effect limited.
In the cold gas phase, clear differences exist between our experiments. Without AGN feedback, cold gas cannot be accelerated above , as shown also by Costa et al. (2015). The inclusion of AGN feedback can accelerate cold gas to higher velocities, up to , for the case without SNe and up to , for the case with both SN and AGN feedback. This confirms, in a different regime, that combined effect of supernovae and AGN feedback boost each other and is what accelerates the cold gas: the AGN accelerates further the gas than has been already accelerated by SNe. Finally, note that the large time scale ratios for the NoSNeAGN are caused by the lack of metal enrichment.
In the redshift interval the picture remains similar, but the now more massive BHs produce a more efficient AGN feedback, especially on the cold gas. This efficiency increases with time, and at the NoSNeAGN simulation shows hot gas outflows up to , but not beyond, in contrast with all the cases including SN feedback, which go as high as . In the cold gas phase, SNe alone become completely ineffective at accelerating the cold gas, and it is only in the presence of an AGN that cold gas can be affected in the now massive galaxy (see the discussion in Dubois et al. 2015). Additionally, in the SNe only case, for the cold gas at , while time ratios of order are present in the simulations including AGN feedback.
The cold gas in the no-Eddington limit simulation (not shown) can surpass the during super-Eddington accretion episodes at redshift above ten. Below the outflows can reach during super-Eddington episodes. Those episodes, however, are very short lived, and are followed by long quiescent periods, therefore the outflow velocity, averaged over hundreds of Myr, remains low.
Note that inside the mass of hot gas is much lower than the mass of cold gas in our simulations. At hot gas is not more than few percent of the total gas mass inside of the virial radius, increasing to a few per cent at . In the final redshift range, , all the simulations including SN feedback reach values in the range . The NoSNeAGN feedback remains at of hot gas and is the case with the lowest hot gas fraction also in the previous redshift intervals.
3.4.1 BH and stellar mass evolution
The discussion on the BH accretion rate and gas velocities presented in the previous sections now allows us to understand the BH and stellar mass evolution in this high-redshift galaxy. Fig. 10 shows the BH and galaxy stellar mass evolution as a function of redshift. The SNe simulation (black) reaches the highest BH mass at the end of the experiment (except for the run with no feedback at all described in PE16 and reported here for comparison). SN feedback can accelerate the gas because of pressure gradients associated to temperature differences of the order K (depending on the gas density) in the galactic disc.
As shown in the bottom left panel of Fig. 5, due to the high mass concentration around the BH (and its high escape velocity) the BH in the NoSNeAGN case has the fastest growth at high redshift amongst all our simulations, comparable with the no feedback case of PE16. As shown in table 1, the BH grows at the Eddington rate at , until it reaches M. At this stage AGN feedback is capable of accelerating the surrounding gas beyond the central escape velocity, reducing dramatically the BH growth rate. The BH ends the simulation with a mass of M.
At the beginning of the evolution, the SNeAGNnoEdd BH shows a large mass jump, increasing almost by three times its initial value due to short bursts of super-Eddington accretion accompanied by strong feedback. Afterward, as discussed in the previous sections, the central region is devoid of gas and BH growth is stunted until a merger at that replenishes the gas supply and the BH experiences a high super-Eddington accretion () episode increasing its mass by almost one order of magnitude. By this time, however, AGN feedback is not as disruptive anymore in the denser and more massive nucleus and the BH continues its evolution with irregular mass accretion rate and with more super-Eddington episodes.
The SNeAGN feedback has the slowest mass growth and then it reaches the lowest mass at the end of the simulation ( M). In this case, beside the SNe outflows, AGN activity can create outflows associated to temperature gradients of K, much larger than the ones produced by SNe. This result shows that the combined effect of SNe and AGN feedback works together to quench efficiently the BH growth.
At all the simulations including AGN feedback (SNeAGN, NoSNeAGN and SNeAGNnoEdd) show a sudden change in the BH mass. At this redshift the system experiences a merger, bringing fresh high density gas to the central galactic region to fuel the BH. Fig. 11 shows two episodes of rapid BH mass growth for the SNeAGN simulation, one at and other one at (mentioned above). In this two examples it is possible to associate a merger event with a jump in the BH mass evolution (see Dubois et al., 2015).
Regarding the stellar mass evolution, the simulation with no SNe and only AGN feedback, NoSNeAGN, produces the galaxy with the highest stellar mass at the end of the simulation, comparable but lower to the NoSNe run in PE16. The final stellar mass is M. As expected, the lack of fast outflows allows high gas concentration, and therefore an efficient star formation throughout the simulation. In contrast, all the cases including SN feedback finish the simulation with a lower stellar mass, of the order of M, showing that the BH mass is not large enough in these galaxies to quench star formation.
All the SN feedback simulations have a similar stellar masses at , M. Above this redshift it is possible to see that the stellar masses differ. In other words, it seems that below M SNe and AGN feedback produce a larger scatter in the stellar mass content at similar DM halo mass.
4. Discussion and Conclusions
We have run cosmological zoom-in simulations of high redshift galaxies in order to study the effect of SN and AGN feedback on the mass transport in these objects and their consequences on the central BH growth.
As in PE16 we find that the mass accretion rate beyond the virial radius is of the order of M. These high mass accretion rates are associated to material in almost free-fall going onto the central DM halo region. In these systems, feedback from SNe and AGN is able to suppress the low density mass accretion at the galactic edge. Material with densities associated with cooling flows, i.e. cm can not penetrate inside the central kpc. Only high density cm gas is able to reach the inner few 100 pc. The simulations with only SNe, and the simulation with SNe plus AGN feedback are those for which the suppression is the highest, in contrast to the AGN only experiment where gas can reach the galactic center more easily.
In these turbulent galactic environments, torques acting on the gas have two sources: gravity force associated to an inhomogeneous mass distribution and pressure gradients associated to circum-galactic shocks driven by cosmic infall, or SN and AGN feedback. Such torques are required to re-distribute the gas AM from the external edge of the galaxy to the central regions, at pc scales, and trigger a radial mass accretion rate of M .
All the SN feedback simulations produce a lower at compared with their at , showing that SN feedback can quench efficiently BH growth in small galaxies at high redshift until a critical mass is reached (Dubois et al., 2015). Furthermore, it is possible to see that after merger events, the BH has a significant growth. During such events a large amount of gas reaches the BH, feeding it with fresh gas.
Our four simulations show very different BH mass accretion histories depending on the physical ingredients we included. Although SN feedback alone is able to alter dramatically the BH accretion rate (Dubois et al., 2015; Habouzit et al., 2016, and PE16), the SN and AGN feedback in tandem are the most efficient to quench the BH growth. Due to the high gas density around the BH at high redshift, the simulation without SN feedback can grow near the Eddington limit until it reaches the , and self-regulates by its AGN activity.
Regarding stellar mass growth, AGN feedback alone is also unable to regulate SF, although the BH-stellar mass ratio is large, in fact for the noSNeAGN case, the ratio is well above at almost all times. SN feedback, in these small galaxies, seems to be much more effective at affecting SF.
Our simulations show that most of the central gas inside of the halo virial radius is cold ( K). The hot ( K) gas is no more than a few percent above redshift . Below this redshift, all the simulations including SN feedback increase the amount of hot gas reaching fractions . In contrast, the no SN feedback simulation keeps its hot gas fraction very low over its entire evolution.
The simulation without SN feedback shows that it is not possible to accelerate the hot gas beyond . In contrast all the SN feedback simulations can easily surpass this value, showing that SN heating creates a low density rarefied gas where the SN and AGN feedback accelerate the gas at velocities as high as .
The cold gas phase in the innermost galactic region of the SN feedback simulation can not be accelerated beyond (Costa et al., 2015) whereas with AGN plus SN feedback the cold gas can reaches up to , showing that the combined effect of SN and AGN heating creates the strongest outflows of cold gas and consequently is the most efficient combination to quench the BH and galaxy growth at high redshift.
We note that our results should depend on the BH feedback prescription. In particular a more realistic approach based on a quasar and radio-jet-like mode as in Dubois et al. (2012) would produce a lower effect on the surrounding gas increasing the BH growing rate and reducing the gas outflows velocity. A simulation including a jet like BH feedback will be presented in an up-coming paper.
- Abel et al. (2002) Abel T., Bryan G. L., & Norman M. L. 2002, Science, 295, 93
- Agarwal et al. (2012) Agarwal B., Khochfar S., Johnson J. L., Neistein E., Dalla Vecchia C., & Livio M., 2012, MNRAS, 425, 2854
- Arribas et al. (2014) Arribas S., Colina L., Bellocchi E., Maiolino R., VillarMartín M., 2014, A&A, 568, A14
- Bandara et al. (2009) Bandara K., Crampton D., Simard L., 2009, ApJ, 704, 1135
- Begelman et al. (2006) Begelman M. C. Volonteri, M. & Rees M. J. 2006, MNRAS, 370, 289
- Begelman et al. (2008) Begelman M. C. Rossi E. M. & Armitage P. J. 2008, MNRAS, 387, 1649
- Bleuler & Teyssier (2014) Bleuler A. & Teyssier R., 2014, MNRAS, 445, 4015
- Bondi (1952) Bondi H., 1952, MNRAS, 112, 195
- Booth & Schaye (2009) Booth C. M. & Schaye J., 2009, MNRAS, 398, 53
- Bower et al. (2016) Bower, R., Schaye, J., Frenk, C. S., et al. 2016, arXiv:1607.07445
- Bromm & Larson (2004) Bromm V., & Larson R. B. 2004, ARA&A, 42, 79
- Choi et al. (2015) Choi J., Shlosman I. & Begelman M. C., MNRAS, 450, 4411
- Clark et al. (2011) Clark P. C., Glover S. C. O., Smith R. J.. Greif T. H., Klessen R. S., Bromm V., 2011, Science, 331, 1040
- Costa et al. (2015) Costa, T., Sijacki, D., & Haehnelt, M. G. 2015, MNRAS, 448, L30
- Danovich et al. (2015) Danovich M., Dekel A., Hahn O., Ceverino D. & Primack J., 2015, MNRAS, 449, 2087
- Dalgarno & MacCray (1972) Dalgarno, A. & McCray, R. A., 1972, ARA&A, 10, 375
- De Rosa et al. (2014) De Rosa G. et al., 2014, ApJ, 790, 145
- Devecchi & Volonteri (2009) Devecchi B. & Volonteri M., 2009, ApJ, 694, 302
- Devecchi et al. (2012) Devecchi B., Volonteri M., Rossi E. M., Colpi M., Portegies Zwart S., 2012, MNRAS, 421, 1465
- Di Matteo et al. (2012) Di Matteo, T., Khandai, N., DeGraf, C., et al., 2012, ApJL, 745, L29
- Di Matteo et al. (2016) Di Matteo, T., Croft, R. A. C., Feng, Y., Waters, D., & Wilkins, S., 2016, arXiv:1606.08871
- Dressler (1989) Dressler A., 1989, in IAU Symposium, Vol. 134, Active Galactic Nuclei, Osterbrock D. E., Miller J. S., eds., p. 217
- Dubois & Teyssier (2008) Dubois Y. & Teyssier R., 2008, A&A, 477, 79
- Dubois et al. (2012) Dubois, Y., Devriendt, J., Slyz, A., & Teyssier, R. 2012, MNRAS, 420, 2662
- Dubois et al. (2012) Dubois Y., Pichon C., Haehnelt M., Kimm T., Slyz A., Devriendt J., & Pogosyan D., 2012, MNRAS, 423, 3616
- Dubois et al. (2015) Dubois Y., Volonteri M., Silk J., Devriendt J., Slyz A. & Teyssier R., 2015, MNRAS, 452, 1502
- Fan et al. (2001) Fan X., et al., 2001, AJ, 121, 54
- Ferrarese & Merritt (2000) Ferrarese L. & Merritt D., 2000, ApJ, 539, 9
- Ferrarese & Ford (2005) Ferrarese L. & Ford H., 2005, SSRv, 116, 523
- Gammie (2001) Gammie Ch. F., 2001, ApJ, 553, 174
- Gebhardt et al. (2000) Gebhardt K. et al., 2000, ApJ, 539, L13
- Genzel et al. (2014) Genzel R. et al., 2014, ApJ, 796, 7
- Ghez et al. (2005) Ghez A. M., Salim S., Hornstein S. D., Tanner A., Lu J. R., Morris M., Becklin E. E. & Duchêne G., 2005, ApJ, 620, 744
- Greif et al. (2011) Greif T. H., Springel V., White S. D. M., Glover S. C. O., Clark P. C., Smith R. J., Klessen R. S. & Bromm V., 2011, ApJ, 737, 75
- Gültekin et al. (2009) Gültekin K., Richstone D. O., Gebhardt K., Lauer T.R., Tremaine S., Aller M. C., Bender R., Dressler A., Faber S. M., Filippenko A. V., Green R., Ho L. C., Kormendy J., Magorrian J., Pinkney J. & Siopis C., 2009, ApJ, 698, 198
- Haardt & Madau (1996) Haardt F. & Madau P., 1996, ApJ, 461, 20
- Habouzit et al. (2016) Habouzit, M., Volonteri, M., & Dubois, Y. 2016, arXiv:1605.09394
- Haiman (2013) Haiman, Z., 2013, ASSL, 396, 293
- Hawley (2000) Hawley J. F. 2000, ApJ, 528, 462
- Heger & Woosley (2002) Heger A., & Woosley S. E. 2002, ApJ, 567, 532
- Humphrey et al. (2010) Humphrey A., Villar-Martín M., Sánchez S. F., Martínez-Sansigre A., Delgado R. G., Pérez E., Tadhunter C., Pérez-Torres M. A., 2010, MNRAS, 408, L1
- Katz, Sijacki & Haehnelt (2015) Katz H., Sijacki D., Haehnelt M. G., 2015, ArXiv eprints:1502.03448
- Kormendy & Richstone (1995) Kormendy J., Richstone D., 1995, ARA&A, 33, 58
- Latif et al. (2013) Latif M. A., Schleicher D. R. G., Schmidt W. & Niemeyer J. C., MNRAS, 436, 2989
- Latif et al. (2014) Latif M. A., Bovino S., Van Borm C., Grassi T., Schleicher D. R. G. & Spaans, M., 2014, MNRAS, 443, 1979
- Lodato & Natarajan (2006) Lodato G., Natarajan P., 2006, MNRAS, 371, 1813
- Loeb & Rasio (1994) Loeb A., Rasio F. A., 1994, ApJ, 432, 52
- Lupi et al. (2014) Lupi A., Colpi M., Devecchi B., Galanti G., Volonteri M., 2014, MNRAS, 442, 3616
- Lupi et al. (2016) Lupi, A., Haardt, F., Dotti, M., et al. 2016, MNRAS, 456, 2993
- Madau & Rees (2001) Madau P., Rees M. J., 2001, ApJL, 551, L27
- Magorrian et al. (1998) Magorrian J., Tremaine S., Richstone D., Bender R., Bower G., Dressler A., Faber S. M., Gebhardt K., Green, R., Grillmair C., Kormendy J., & Lauer T., 1998, AJ, 115, 2285
- Mortlock et al. (2011) Mortlock D. J. et al., 2011, Nature, 474, 616
- Nesvadba et al. (2011) Nesvadba N. P. H., Polletta M., Lehnert M. D., Bergeron J., De Breuck C., Lagache G., Omont A., 2011, MNRAS, 415, 2359
- Oh & Haiman (2002) Oh S. P. & Haiman Z. 2002, ApJ, 569, 558
- Omukai, Schneider &Haiman (2008) Omukai K., Schneider R., Haiman Z., 2008, ApJ, 686, 801
- Park et al. (2016) Park K., Ricotti M., Natarajan P., Bogdanović T. & Wise J. H., 2016, ApJ, 818, 184
- Pichon et al. (2011) Pichon C., Pogosyan D., Kimm, T., Slyz, A., Devriendt, J. & Dubois, Y., 2011, MNRAS, 418, 2493
- Planck Collaboration (2013) Planck Collaboration, 2013, arXiv:1303.5076P
- Powell et al. (2011) Powell L. C., Slyz A. & Devriendt J., 2011, MNRAS, 414, 3671
- Prieto et al. (2015) Prieto J., Jimenez R., Haiman Z. & González R. E., 2015, MNRAS, 452, 784
- Prieto & Escala (2016) (PE16) Prieto J. & Escala A., 2016, MNRAS, 460, 4018
- Prunet et al. (2008) Prunet S., Pichon C., Aubert D., Pogosyan D., Teyssier R. & Gottloeber S., 2008, ApJs, 178, 179
- Regan & Haehnelt (2009) Regan J. A., Haehnelt M. G., 2009, MNRAS, 393, 858
- Rasera & Teyssier (2006) Rasera Y. & Teyssier R., 2006, A&A, 445, 1
- Shakura & Sunyaev (1973) Shakura N. I. & SunyaevR. A., 1973, A&A, 24, 337
- Schneider et al. (2006) Schneider R., Omukai K., Inoue A. K. & Ferrara A., 2006, MNRAS 369, 1437
- Stacy et al. (2010) Stacy A., Greif T. H. & Bromm V., 2010, MNRAS, 403, 45
- Sutherland & Dopita (1993) Sutherland R. S. & Dopita M. A., 1993, Apj Sup., 88, 253
- Tanaka & Haiman (2009) Tanaka T., Haiman Z., 2009, ApJ, 696, 1798
- Teyssier (2002) Teyssier R., 2002, A&A, 385, 337
- Teyssier et al. (2011) Teyssier R., Moore B., Martizzi D., Dubois Y. & Mayer L., 2011, MNRAS, 414, 195
- Teyssier et al. (2013) Teyssier R., Pontzen A., Dubois Y. & Read J. I., 2013, MNRAS, 429, 3068
- Tremaine et al. (2002) Tremaine S., Gebhardt K., Bender R., Bower G., Dressler A., Faber S. M., Filippenko A. V., Green R., Grillmair C., Ho L. C., Kormendy J., Lauer T. R., Magorrian J., Pinkney J. & Richstone D., 2002, ApJ, 574, 740
- Truelove et al. (1997) Truelove J. K., Klein R. I., McKee Ch. F., Holliman J. H., Howell L. H. & Greenough J. A., 1997, ApJl, 489, 179
- Volonteri, Madau & Haardt (2003) Volonteri M., Madau P., Haardt F., 2003, ApJ, 593, 661
- Volonteri et al. (2008) Volonteri M., Lodato G., Natarajan P., 2008, MNRAS, 383, 1079
- Volonteri (2010) Volonteri M., 2010, A&ARv, 18, 279
- Volonteri & Begelman (2010) Volonteri M., Begelman M. C., 2010, MNRAS, 409, 1398
- Volonteri et al. (2015) Volonteri, M., Silk, J., & Dubus, G. 2015, ApJ, 804, 148
- Willott (2007) Willott C. J. et al., 2007, AJ, 134, 2435