Astronomy & Astrophysics manuscript no.

aa ©ESO 2023
December 15, 2023

Sustained super-Eddington accretion in high-redshift quasars

Alessandro Lupi1,2,3,⋆ , Giada Quadri3 , Marta Volonteri4 , Monica Colpi3, 2 , and John A. Regan5

1
Dipartimento di Scienza e Alta Tecnologia, Università degli Studi dell’Insubria, via Valleggio 11, I-22100, Como, Italy
2
INFN, Sezione di Milano-Bicocca, Piazza della Scienza 3, I-20126 Milano, Italy
3
Dipartimento di Fisica “G. Occhialini”, Università degli Studi di Milano-Bicocca, Piazza della Scienza 3, I-20126 Milano, Italy
4
Institut d’Astrophysique de Paris, UMR 7095, CNRS and Sorbonne Université, 98 bis boulevard Arago, 75014 Paris, France
5
Centre for Astrophysics and Space Sciences Maynooth, Department of Theoretical Physics, Maynooth University, Maynooth,
Ireland

Received December 15, 2023; accepted

arXiv:2312.08422v1 [astro-ph.GA] 13 Dec 2023

ABSTRACT

Observations of z ≳ 6 quasars provide information on the early phases of the most massive black holes (MBHs) and galaxies. Current
observations, able to trace both gas and stellar properties, show that most MBHs at high redshift seem overmassive compared to the
local population, in line with the elliptical galaxy population, or even above, thus implying a very rapid growth of these objects. To
assess the physical conditions for such a rapid growth and explain the existence of a population of already mature MBHs when the
Universe was less than 1 Gyr old, we here explore whether episodes of accretion above the Eddington limit can occur across cosmic
epochs. By employing state-of-the-art high-resolution cosmological zoom-in simulations of a z ∼ 7 quasar, where different accretion
regimes are consistently included, together with their associated radiative and kinetic feedback, we show that super-Eddington phases
can be sustained for relatively long time-scales (tens of Myr) and discuss how they affect the growth of MBHs. We also show, by
means of a semi-analytic evolution, that the MBH spin remains relatively low during super-Eddington phases, and this would result
in a lower feedback efficiency, hence a potentially faster growth that might then explain the overmassiveness of high-redshift MBHs.
Key words. black holes: evolution – galaxies: formation – galaxies: evolution – galaxies: high redshift

1. Introduction native possibility, which is also suggested by recent results (e.g.

Davies et al. 2019; Yang et al. 2023), is that these MBHs grew
Massive black holes (MBHs) are ubiquitous in the Universe and rapidly because of intermittent phases of super-Eddington accre-
inhabit the centre of massive galaxies up to redshift z ≳ 6 (Fan tion (Madau et al. 2014; Volonteri et al. 2015; Lupi et al. 2016;
et al. 2006; Mortlock et al. 2011; Bañados et al. 2018; Fan et al. Pezzulli et al. 2016).
2023; Maiolino et al. 2023, e.g.), with masses up to 109−10 M⊙ . The plausibility of super-Eddington accretion has been
Observationally, they are commonly identified via gas accretion, demonstrated in the local Universe in tidal disruption events (Lin
thanks to the conversion of gravitational energy into radiation, et al. 2017) and ultra-luminous X-ray sources (Bachetti et al.
which makes them shine as active galactic nuclei (AGN), some- 2014), and was initially proposed in the context of non-spherical
times producing also powerful collimated jets. accretion flows within the ‘slim-disc’ model (Abramowicz et al.
According to Soltan’s argument (Soltan 1982) and to the evo- 1988). Recent simulations of accretion discs show that the slim-
lution of the AGN luminosity function, MBHs gained most of disc solution, in which the radiation is trapped in the innermost
their mass via radiatively efficient accretion, hence they should regions of the accretion disc and advected within the fluid, mak-
have formed from lower-mass ‘seeds’. In this respect, observa- ing the disc moderately luminous, is instead characterised by
tions of high-redshift MBHs pose tight constraints on the min- powerful radiatively driven outflows and jets escaping through
imum initial seed MBH mass, which has to be of the order of a central funnel perpendicular to the disc itself (Sadowski
˛ &
Mseed ≳ 104 M⊙ , if we assume the growth to occur via radia- Narayan 2016; Sadowski
˛ et al. 2016; Jiang et al. 2019). To date,
tively efficient accretion with ηrad ∼ 0.1, at the Eddington limit. the impact of these jets on the interstellar medium of a galaxy
In order to explain these results, several groups have focussed host and on the MBH growth itself has only been considered in
on the initial mass of heavy seed MBHs, which has to be par- idealised setups as isolated galaxies (Massonneau et al. 2023b),
ticularly high to be able to explain the most massive objects circum-nuclear discs (Sassano et al. 2023), and atomic cooling
observed (Fan et al. 2023), hence requiring very peculiar con- haloes just after the formation of a heavy seed MBH (Regan et al.
ditions (e.g. Latif et al. 2015; Schauer et al. 2017; Lupi et al. 2019).
2021; Latif et al. 2022). In order to explain these results, sev- In this work, we move forward and study the MBH evolu-
eral groups have focussed on studying “heavy” seed MBHs, with tion in the quasar host originally studied in Lupi et al. (2019,
high initial mass to ease explaining the most massive objects 2022) properly accounting for super-Eddington accretion phases
observed (Fan et al. 2023). The formation of the most massive in a full cosmological context. The simulation adopts the same
seeds requires however peculiar conditions (e.g. Latif et al. 2015; initial conditions of the high-redshift quasar originally studied in
Schauer et al. 2017; Lupi et al. 2021; Latif et al. 2022). An alter- Lupi et al. (2019, 2022, hereafter Paper I and II). This is the IV of
a series of papers addressing properties of high-redshift quasar
⋆
[email protected] hosts and their MBHs. In Paper I, we presented and discussed
Article number, page 1 of 7
 A&A proofs: manuscript no. aa

the main evolution of the target galaxy and its central MBH, fo- gives σturb = ||∇v||/5. Relative to Lupi et al. (2019, L19
cussing on the stellar and gas tracers (total gas and [CII] emis- hereon), we have updated our star formation efficiency em-
sion), and found that super-Eddington phases were measured in ploying the Padoan et al. (2012) model, as in Lupi & Bovino
the simulation, even though accretion was capped at the Edding- (2020).
ton limit. In paper II, we extended the analysis focussing on the – Stellar mechanical feedback: we redetermined the scalings
dynamics and morphology of the main galaxy as a function of in Martizzi et al. (2015) to improve the agreement with their
redshift. In Paper III (Lupi et al. in prep.) we will discuss the results, i.e. we properly account for the initial fraction of
evolution of the entire MBH population forming during the sim- kinetic and thermal energy during the Sedov-Taylor phase
ulation, and in Paper V (Quadri et al. in prep.) we will focus in fkin ∼ 0.28.
detail on the impact of the super-Eddington regime on the galaxy – Stellar radiative feedback: instead of the cost-effective ap-
host and on the properties of quasar outflows. proximated radiation transport of L19, we include here on-
The paper is organised as follows: in Section 2, we recap the the-fly radiation transport as in Lupi et al. (2020), with the
setup of the simulation and describe the improvements relative reduced speed of light set to cred = 1000 km s−1 , which is
to the previous works; in Section 3, we present our results; in large enough compared to the gas motions to ensure consis-
Section 4, we draw our conclusions. tent results. Since, in addition, we follow X-ray chemistry,
we now follow radiation in 11 photobins ranging from 0.7
2. Numerical setup eV up to 10 keV, where two bins are used to cover soft (0.2-
2 keV) and hard (2-10 keV) X-rays.
The simulation follows the evolution of a massive halo (Mhalo ∼
3×1012 M⊙ at z = 6) expected to represent a quasar host (Di Mat-
teo et al. 2017). The initial conditions were accurately created to 2.1. MBH accretion/feedback and dynamics
match the expected halo mass (Di Matteo et al. 2017; Tenneti In addition to the changes above, for this work we have devised
et al. 2018) and the galaxy overdensity significance (Uchiyama a novel set of prescriptions for MBH growth and dynamics, that
et al. 2018; Mignoli et al. 2020) via music (Hahn & Abel 2013), we discuss in detail in the following. MBH seeding, instead, is
adopting the Planck Collaboration et al. (2016) cosmological identical to that in L19, and occurs in galaxies with a stellar mass
parameters, with Ωm = 0.3089, ΩΛ = 0.6911, Ωb = 0.0489, > 108 M⊙ not yet hosting a MBH, galaxies that are identified
σ8 = 0.8159, ns = 0.9667, and H0 = 67.74 km s−1 Mpc−1 , through an on-the-fly Friends-of-Friends algorithm (see L19 for
with no contribution from radiation and curvature. From a parent details).
dark-matter-only simulation, we recursively zoomed-in on a La-
grangian volume extending up to 2.5 virial radii of the target halo
following the approach by Fiacconi et al. (2017) to exclude any 2.1.1. MBH dynamics
contamination by low-resolution dark matter particles within the
virial radius. In most cosmological simulations, the mass and spatial resolu-
The simulation has been run with gizmo (Hopkins 2015), tion of seed MBHs is not sufficient to accurately resolve the dy-
descendant of Gadget3 (Springel et al. 2008) and Gadget2 namical friction bringing MBHs to the centre of galaxies, and
(Springel 2005) in its meshless-finite-mass mode. The spatial often also the interaction with other particles, leading to spuri-
resolution of the simulation was set to 40, 10, and 2.5 pc h−1 ous scattering of the MBHs. For this reason, most simulations
for dark matter, stars, and MBHs, respectively, whereas fully include an ad-hoc MBH pinning procedure (Di Matteo et al.
adaptive softening was assumed for the gas component, down 2005; Schaye et al. 2015; Barai et al. 2018) which moves the
to a minimum of ∼ 5 pc. The mass resolution was ∼ 104 M⊙ for MBH to the potential minimum inside its kernel at every time-
baryons and ∼ 105 M⊙ for dark matter. step. Although effective, this procedure can produce unphysical
Our simulation was performed with state-of-the-art sub-grid behaviours like superluminal motions or artificial suppression
prescriptions that allowed us to follow in detail non-equilibrium of the MBH wandering. A more physically motivated prescrip-
chemistry of primordial species, star formation and stellar feed- tion artificially corrects the dynamics accounting for the unre-
back, as well as MBH seeding, accretion, and feedback. Com- solved dynamical friction effect (Dubois et al. 2013; Tremmel
pared to the previous works of the series, here we have slightly et al. 2015), which however requires a sufficiently large mass ra-
revised and improved many of the sub-resolution prescriptions, tio between the MBH and the other tracers (gas, stars, and dark
which we describe as follows: matter) (Tremmel et al. 2015; Pfister et al. 2019). In Lupi et al.
(2019), we ensured a reasonable dynamical evolution by seed-
– Gas thermodynamics and chemistry: we further extended ing the MBH with an already large mass of MBH = 106 M⊙ . In
our chemical network to include high-ionization states of this work, instead, we opted for decoupling the MBH mass in a
several important species that are commonly observed in physical mass (used for accretion) and a dynamical mass (for the
quasar hosts, namely C[I-IV], O[I-VI], N[I-V], and Fe[I- dynamics) (Anglés-Alcázar et al. 2017), which evolve together
II], also accounting for their contribution to the gas low- as soon as the first reaches the second. The physical mass was set
temperature cooling. Since MBHs are commonly surrounded to MBH = 105 M⊙ , and the dynamical mass to a value ten times
by a hot corona emitting in X-rays, we also incorporated de- larger, that ensured a better dynamical evolution. From a physi-
tailed X-ray chemistry calculations in our network, account- cal point of view, this initially higher dynamical MBH mass can
ing for the impact of Compton heating by the AGN (assum- be considered as an unresolved stellar envelope (a nuclear stel-
ing T Compton = 3.23 × 106 K and T Compton = 8.41 × 107 K lar cluster) surrounding the MBH, which is commonly found in
for soft and hard X-rays respectively),that will be extendedly many galaxies in the local Universe (see Neumayer et al. 2020,
discussed in Lupi et al. in preparation. for a review). From a numerical point of view, however, it is
– Star formation: we have slightly revised our estimate for the simply used to avoid the spurious scattering by other particles,
turbulent support of the gas as in Hopkins et al. (2013), ac- which despite the very high resolution adopted in this simulation
counting for the particle distribution inside the kernel, which are still more massive than individual stars.

Article number, page 2 of 7

 Alessandro Lupi et al.: Super-Eddington accretion in quasar hosts

Unlike Tremmel et al. (2017) and Ma et al. (2021), in this disc (MAD, Narayan et al. 2003), and Φ = ϕ/ϕMAD . The
work the unresolved dynamical friction is implemented follow- mass loading of the jet βjet ≡ Ṁjet / ṀBH in our simulation is
ing the more accurate model by Pfister et al. (2019), where determined assuming energy conservation and a jet velocity
the dark matter and stellar distributions are accounted for sep- vw = 0.1c = 3 × 104 km s−1 , which gives
arately (thus allowing for different distribution functions), and !2
also the high-velocity part of the distribution function is consid- Ṁjet c
ered. Moreover, we also included gas-driven dynamical friction βjet = = 2ηjet ≈ 83.4Φ2 (4)
ṀBH c2 vw
as described in Tanaka & Haiman (2009), following the prescrip-
tion in Escala et al. (2004). – 2.5 × 10−3 < λ < 1 (sub-Eddington radiatively efficient ac-
cretion): this is the regime occurring in typical AGN, where
2.1.2. MBH accretion and feedback the disc is geometrically thin and optically thick, and can
be understood in terms of the Shakura and Sunyaev solu-
One of the main novelties of this work is the inclusion of three tion (Shakura & Sunyaev 1973). In this case, we assume the
accretion regimes covering the entire range of accretion rates. MBH feedback to be in the form of radiation (again assum-
We classify them in terms of the Eddington ratio λ = ṀBH / ṀEdd , ing a composite black-body + corona spectrum) and bipo-
where ṀBH is the accretion rate, ṀEdd = 16 LEdd /c2 (Madau et al. lar line-driven winds whose mass loading is given by mo-
2014), LEdd is the Eddington luminosity, and c is the speed of mentum conservation during the matter-radiation interaction
light. In principle, the radiative efficiency depends on the MBH (Choi et al. 2012; Anglés-Alcázar et al. 2017), i.e.
spin which in the current simulation is not evolved over time. We
set the spin magnitude to a constant value of a = 0.7, which gives LBH c
the commonly adopted radiative efficiency in the standard sub- βw = = ηrad ≈ 1, (5)
ṀBH vw c vw
Eddington radiatively efficient accretion regime (ηrad = 0.103),
and we further assume that the spin is parallel to the angular where LBH = ηrad ṀBH c2 is the accretion luminosity. The
momentum of the gas within the MBH kernel lgas,MBH . While resulting energy coupling efficiency for the wind, which is
accretion is still modelled using the Bondi-Hoyle-Lyttleton pre- launched assuming a semi-aperture of 45 deg (Sala et al.
scription, the resulting MBH feedback has been implemented in 2021), can be estimated in
radiative and kinetic forms.
Radiative feedback is injected as in Lupi et al. (2020), as- Ṁw v2w Ṁw v2w βv2w vw
suming a composite black body plus X-ray corona spectrum ϵ= = = = = 0.05 (6)
2LBH 2ηrad ṀBH c2 2ηrad c2 2c
for the MBH, with a bolometric luminosity defined as Lbol =
ηrad ṀBH c2 , with ηrad depending on the accretion regime (see be- – λ > 1 (super-Eddington accretion): in this regime, we as-
low). The fraction of Lbol associated to the X-ray corona is de- sume radiative+kinetic MBH feedback, where the radiative
termined according to Duras et al. (2020) as: component is determined according to the slim-disc solution,
 !18.78  with the radiative efficiency obtained by Lupi et al. (2016):
Lbol log(Lbol /L⊙ )
= 12.76 1 +
−1
 
f2−10 keV ≡  (1) " #
L2−10 keV 12.15 A(a) 0.985 0.015
ηrad = + , (7)
16λ 1/λ + B(a) 1/λ + C(a)
From this expression, we determine the soft X-ray fraction
f0.2−2 keV assuming a power-law spectrum with slope -1.7 for the where A(a), B(a), and C(a) are spin-dependent coefficients
corona (Regan et al. 2019), and finally obtain the residual black that in our case assume values 1.915, 0.795, and 0.017 re-
body component as spectively. Although radiation is expected to be trapped in
the innermost regions of the slim disc, in this work we
fBB = 1 − f0.2−2 keV − f2−10 keV . (2) neglect any change in the spectral shape of the radiation
Kinetic feedback, instead, is implemented for the different spectrum, leaving this exploration for a future study. The
regimes as1 kinetic efficiency in this case is computed according to
Tchekhovskoy & Giannios (2015) and Sa̧dowski et al. (2016)
– λ < 2.5 × 10−3 (ADAF regime, Yuan & Narayan 2014): in for a magnetically-driven jet, i.e.
this regime, the disc is optically thin and geometrically thick,
due to inefficient cooling. Ions and electrons decouple, re- ηjet = 1.3a2 Φ2 ≈ 0.637Φ2 , (8)
sulting in a two-temperature accretion flow where the radia-
tive efficiency decays roughly as ηrad ∝ λ2/3 (Xie & Yuan corresponding to a mass loading βjet ≈ 127.4Φ2 .
2012). In this regime, a jet is launched along lgas,MBH and it
forms a cylinder with the base defined by the MBH kernel A crucial parameter affecting the jet efficiency is the limiting
with an efficiency determined as (Blandford & Znajek 1977; magnetic field relative to the MAD limit. In this work, we con-
Tchekhovskoy & Giannios 2015): sider two cases, one at the MAD limit (Φ = 1) and one at half the
limit (Φ = 0.5; HMAD hereon) (see Sa̧dowski et al. 2016, for a
a ϕ
!2 discussion). At every accretion event, we estimate the accretion
ηjet ≈ 2.5 √ ≈ 0.417Φ2 , (3) rate on the MBH from its 96 nearest gas neighbours, we com-
(1 + 1 − a ) 2 2 ϕMAD
pare it with the Eddington limit, and then determine ηrad and β
where ϕ is the maximum magnetic flux in the disc, ϕMAD for the corresponding regime. An important aspect that must be
is a critical value corresponding to a magnetically-arrested considered is that, for very high accretion rates, the large mass-
loading factor might yield a total gas mass affected by the MBH
1
The final normalizations of the equations in this section correspond accretion/feedback process over a time-step ∆t which exceeds
to the spin magnitude assumed in our simulation a = 0.7. the available mass in the MBH kernel. In these cases, we follow
Article number, page 3 of 7
 A&A proofs: manuscript no. aa

Regan et al. (2019) and assume that the estimated accretion only dashed line. We also report for comparison the growth assuming
occurs for a fraction of time accretion at constant Eddington rate as black dot-dashed lines.
( ) The thick curve starts rising at t = 500 Myr, when the MBH ef-
Mngbs
facc = min 1, , (9) fectively starts growing in the simulation, whereas the thin curve
(1 + β)∆MBH starts at t = 350 Myr, immediately after seeding. The three stages
just discussed can be noticed also here in this Figure. The time
where ∆MBH = ṀBH ∆t. At this point, we randomly flag enough at which the MBH starts growing is slightly different between
gas particles around the MBH to guarantee ∆M = (1 + β) ṀBH ∆t, the two simulations, because of the small differences appeared
where part of the mass is accreted and part is kicked away in when the simulations have been restarted from the parent one,
a kinetic wind/jet. In the case of Mdyn,MBH > MMBH , we sim- that built up over time changing the MBH history. Apart from
ply select particles to be kicked away in an outflow, in order this, and the plateaus in the growth history when the accretion
to ensure mass conservation. In order to prevent the outflowing rate drops, the growth exhibits a similar behaviour, i.e. an initial
gas particles to propagate over large distances without interact- rapid growth well above the Eddington limit, which lasts longer
ing (over a single timestep they might be end up well outside in the HMAD case due to the lower MBH feedback efficiency,
the galaxy), we update the timestep ∆t of the kicked particles followed by a self-regulation phase when the MBH settles at the
to ∆t′ = min{∆t, CCFL ∆x/vw }, where ∆x is the effective gas cell Eddington limit or at a fraction of it.
size (Hopkins 2015) and CCFL is the Courant factor.
Despite the quite large efficiency of the jet feedback during
super-Eddington phases, the HMAD MBH is able to grow by al-
3. Results most three orders of magnitude in less than 100 Myr, whereas
the MAD one stays one order of magnitude below over the same
With the model just described, we have run the same initial con- time interval. Another important aspect to keep in mind when
ditions of Lupi et al. (2019) down to z ∼ 9.7, when the MBH looking at this evolution is that, when the MBH in the HMAD
has already formed near the centre of its host, but the galaxy has case starts to self-regulate its growth, the MBH mass is so large
still a low enough mass for supernovae to stunt the MBH growth that super-Eddington phases begin to require extremely large in-
completely. At this point, we split the simulation in two equiv- flow rates in the galaxy nucleus, which are less and less likely
alent runs, MAD and HMAD, which only differ by the mag- as the galaxy evolves and the gas fraction diminishes. In the
netic flux parameter Φ. The simulations have been run down to Eddington-limited case, an initially more massive MBH or an
z ∼ 7.5, in order to follow the early growth of the MBH seeds in earlier start of accretion are needed to ensure a growth in mass
the galaxy host, whose evolution we are going to discuss in this comparable to the two cases we simulated here.
section. Since this rapid growth may have important implications for
In Fig. 1, we report the accretion rate on the MBH in the the MBH-galaxy correlations, we report in Fig. 3 the evolution
two cases, as directly obtained from the simulation (red dots), of the MBH together with the galaxy host stellar mass for the two
and averaged over a 1 Myr timescale (green solid line). As a cases, using the same colour scheme of Fig. 2. We also show for
comparison, we also report as a blue dashed line the Eddington comparison the observations of low-redshift AGN by Reines &
accretion rate at every step of the evolution. The instantaneous Volonteri (2015) as grey dots, elliptical galaxies by Kormendy
ṀMBH exhibits very large excursions, which are associated to & Ho (2013) as grey stars, and high-redshift observations by
the intermittency between large inflows and the MBH feedback ALMA as green squares (Neeleman et al. 2021) and JWST as
self-regulation. red diamonds (Yue et al. 2023; Stone et al. 2023a,b; Harikane
On average, however, we can divide the evolution in three et al. 2023; Maiolino et al. 2023). We also report as magenta
main stages (see also Lupi et al. 2019). (i) In the early stages crosses the ‘little red dots’ observed by Greene et al. (2023),
after MBH seeding (t ≲ 530 Myr), supernova feedback strongly with the stellar masses estimated according to JWST photometry
perturbs the gas in the galaxy, preventing the MBH (which is only and including ALMA photometry by Labbe et al. (2023),
also offset from the galaxy centre) from efficiently accreting. (ii) where the magenta shaded area connects the two mass estimates
When the MBH settles in the centre of the galaxy and the po- for each source.2 We see that, in both cases, the MBH starts at
tential well becomes deep enough to sustain ingent gas inflows the lower end of the distribution and rapidly grows, but only the
to the centre (M⋆ ≃ 1010 M⊙ ), ṀMBH rapidly grows, easily ex- HMAD case is able to reach the upper limit of the distribution,
ceeding Eddington by a factor of a few (MAD) or a few tens thanks to a less effective MBH feedback. Then, when the MBH
(HMAD). At this stage, the HMAD simulation shows an almost starts to self-regulate, even the growth of the HMAD MBH slows
unimpeded super-Eddington growth for about 60 Myr, followed down, moving toward the region occupied by giant ellipticals
by (iii) a decrease of the accretion rate to values between a frac- (see L19 for a discussion).
tion and 100% Eddington, corresponding to the self-regulation
stage. In both cases, there are sudden drops in the accretion rates, Finally, note that the super-Eddington phase was able to
around z = 8.3 and z = 9 for the HMAD and MAD cases re- start only after the weakening of the impact of supernova feed-
spectively. These rapid variations are associated to the potential back on the galaxy, in agreement with other simulations (Dubois
misalignment of the gas in the galaxy nuclear region and the et al. 2014; Anglés-Alcázar et al. 2017; Lupi et al. 2019),
galactic disc plane, which reflects in winds/jets directly hitting hence preventing the simulated MBHs to actually reach the re-
the disc instead of escaping perpendicular to it. In these phases, gion occupied by high-redshift overmassive MBHs, especially
the gas is expelled from the galaxy centre, and falls back on those observed in low-mass galaxies (e.g. Maiolino et al. 2023).
a timescale comparable with the free-fall time of the gas, i.e. This might be a potential issue in explaining the overmassive
5 Myr ≲ tff ≲ 50 Myr for a typical gas density between 1 and
100 cm−3 . 2
Note that the emission in this case is dominated by the AGN and
The evolution just discussed reflects in the mass growth re- dust, hence the stellar mass estimates are only indicative, but likely
ported in Fig. 2 for both simulations, where the MAD case is smaller than 1010 M⊙ because the Little Red Dots are selected to be
reported as an orange solid line and the HMAD one as a blue point sources.

Article number, page 4 of 7

 Alessandro Lupi et al.: Super-Eddington accretion in quasar hosts

Fig. 1. Evolution of the MBH accretion rate in the HMAD (left-hand panel) and MAD (right-hand panel) simulations. The red dots correspond to
the instantaneous accretion rate, i.e. at every time-step of the simulations, whereas the green solid line corresponds to the accretion rate averaged
over 1 Myr. The blue dashed line represents the instantaneous Eddington accretion rate at the MBH mass.

z
12 11 10 9 8 1010 Reines+15
108 Kormendy+13
MAD L19
HMAD 109 MAD
Edd HMAD
High-z (JWST)
MBH (M )

107 Edd (t0 = 350)Myr

108
MBH (M )

High-z (ALMA,Mdyn )
Greene+23
7
10
106

106

105 105
350 400 450 500 550 600 650 107 108 109 1010 1011 1012
t (Myr) M? (M )

Fig. 2. Evolution of the MBH mass in the MAD (shown as an orange Fig. 3. Evolution of the MBH mass in the MAD (shown as an orange
solid line) and HMAD (shown as a blue dashed line) simulations. For solid line) and HMAD (shown as a blue dashed line) simulations rel-
comparison, we also show two black dot-dashed lines corresponding ative to the galaxy stellar mass, compared with the local AGN obser-
to the MBH growth for a constant Eddington accretion, the thick one vations by Reines & Volonteri (2015) (grey dots), local ellipticals by
starting at t = 500 Myr (consistent with the HMAD case) and the thin Kormendy & Ho (2013), and high-redshift (z >= 6) observations by
one starting at t = 350 Myr. ALMA (Neeleman et al. 2021) and JWST (Yue et al. 2023; Stone et al.
2023a,b; Harikane et al. 2023; Maiolino et al. 2023). We also show the
‘little red dots’ by Greene et al. (2023) as magenta crosses, with the
MBHs, unless an even weaker super-Eddington feedback or stellar masses estimated including or neglecting ALMA photometry re-
more favourable inflow conditions occur around these MBHs. spectively, connected by magenta shaded areas (see Labbe et al. 2023,
for details).

4. Discussion and Conclusions

that super-Eddington phases might be sustained over time-scales
In this work, we have performed high resolution cosmological of a few tens of Myr in massive systems where large gas inflows
zoom-in simulations of a massive halo at high redshift, expected are frequent, systems that are quite different from those inves-
to be a quasar host. The simulations include for the first time in tigated in Regan et al. (2019) and Massonneau et al. (2023b).
a cosmological simulation a sub-resolution model for the three However, some important caveats need to be mentioned:
main accretion regimes, from ADAF up to super-Eddington. The
model accounts for MBH accretion-powered feedback in the
form of radiation and kinetic winds/jets, with an efficiency de- – The resolution of the simulation is not high enough to prop-
termined consistently for the different cases. The results we have erly resolve the Bondi-Hoyle-Lyttleton radius of the MBH,
found here, that will be complemented by a more thorough anal- hence potentially affecting the estimated accretion rate. Note,
ysis of the MBH-galaxy interplay and the properties of the MBH nonetheless, that a higher accretion rate reflects in a stronger
outflows in a forthcoming paper (Quadri et al. in prep.), show feedback, which has the potential to clear out the central re-
Article number, page 5 of 7
 A&A proofs: manuscript no. aa

z z
9.5 9 8.5 8 9.5 9 8.5 8
1.0 1.0
a a
0.8 φ/φMAD 0.8 φ/φMAD

0.6 0.6
q

q
0.4 0.4

0.2 0.2

0.0 0.0
500 520 540 560 580 600 620 640 660 500 520 540 560 580 600 620 640 660
t (Myr) t (Myr)

Fig. 4. Evolution of the MBH spin (blue solid line) and MAD-ness (orange dashed line) in the HMAD (left-hand panel) and MAD (right-hand
panel) simulations, as modelled in Ricarte et al. (2023).

gion of the galaxy (as sometimes occurred in our simula- would likely result in a lower spin during the first accretion
tions), hindering accretion for a few tens of Myr. phases, hence a lower kinetic efficiency for the jet, even in
– The launch direction of the feedback is instantaneously the MAD case, which would enable the MBH to grow al-
aligned with the gas within the MBH kernel, without tak- most unimpeded until the MBH spin and mass become large
ing into account the actual direction of an unresolved accre- enough for feedback to start playing a role (Massonneau
tion disc. This means that, if the gas in the galaxy has set- et al. 2023a). In order to test this idea, we have computed
tled in a disc-like configuration, most of the time the kinetic the spin evolution semi-analytically, by means of the model
feedback escapes without interacting significantly with the in Ricarte et al. (2023), employing the accretion rates and
galaxy, hence without suppressing accretion for a long time. the MBH masses from our simulations. In particular, we em-
In a more realistic case, the alignment of the accretion disc ployed the 1 Myr average quantities instead of the full data
might occur on much longer time-scales, hence affecting the sample, in order to limit the effect of strong fluctuations on
impact of the feedback. our conclusions. The results are reported in Fig. 4, where the
– The results of this simulation provide a more optimistic out- blue lines correspond to the spin evolution as a function of
look on the role of super-Eddington accretion in growing time, and the orange dashed line to the disc MAD-ness as
high-redshift black holes compared to previous numerical in- defined in Ricarte et al. (2023) for fEdd > 0.03 (at lower Ed-
vestigations in a galactic/cosmological context (Regan et al. dington ratios a MAD disc is assumed):
2019; Massonneau et al. 2023b). To assess the reasons for
this difference, we measured the gas density and temperature ϕ ( fEdd / fc )α
= , (10)
around the MBH, finding that every super-Eddington burst ϕMAD 1 + ( fEdd / fc )α
was followed by an increase of the gas temperature up to
106 < T/K < 109 (above the peak of the cooling curve), where fc = 1.88 and α = 1.29.
due to the jet shocking with the gas, consistent with the re- We clearly observe that, during the super-Eddington phases,
sults in Regan et al. (2019) and Massonneau et al. (2023b). the MBH spin rapidly decreases to values around 0.3, which
Despite this strong heating, the density around the MBH did correspond to jet efficiencies about 4 times lower than those
not change dramatically, always remaining on average in the employed in the simulations, and only during the later phases
range nH ≃ 104−5 cm−3 , a value significantly larger than that when the accretion rate settles at sub-Eddington values the
in Massonneau et al. 2023b. This suggests that the combina- spin is able to grow up to the initial values. During the evo-
tion of a deep potential well, the pressure of the inflowing lution, the MAD-ness of the disc also exhibits strong varia-
gas from larger scales, and a short free-frall time (see Regan tions, with average values during the super-Eddington phases
et al. 2019 for a discussion) prevented the gas from escaping around 0.5, hence more consistent with our HMAD simula-
the galaxy nucleus. As a consequence, after each jet event, tion than the MAD one.
the accretion rate suppression we observed is short-lived and
is followed by a rapid rise as soon as the gas cooled down. Concluding, in this work we have shown that, despite the un-
Note, however, that the limited resolution of our simulation favourable conditions associated to assuming a relatively high
might still underestimate the effect of the jet feedback, by spin and a consequently strong jet feedback, MBHs in gas-rich
not properly resolving the early expansion of the shocked environments at high redshift can support long-lasting super-
gas, and further investigations are needed. Eddington accretion phases, rapidly growing in mass within their
– The magnitude of the spin in the simulations has been host galaxies. Moreover, because of this rapid growth, these
kept fixed. This has profound implications for the evolu- MBHs can easily move above the local correlations before start-
tion. First, since the jet launched during super-Eddington ing regulating themselves, potentially explaining the formation
accretion phases extracts rotational energy from the MBH of overmassive systems.
(Blandford & Znajek 1977), the spin should decrease rapidly, Acknowledgements.
thus overcoming the spin-up due to coherent accretion. This
Article number, page 6 of 7
 Alessandro Lupi et al.: Super-Eddington accretion in quasar hosts

References Sadowski,
˛ A. & Narayan, R. 2016, MNRAS, 456, 3929
Soltan, A. 1982, MNRAS, 200, 115
Abramowicz, M. A., Czerny, B., Lasota, J. P., & Szuszkiewicz, E. 1988, ApJ, Springel, V. 2005, MNRAS, 364, 1105
332, 646 Springel, V., Wang, J., Vogelsberger, M., et al. 2008, MNRAS, 391, 1685
Anglés-Alcázar, D., Faucher-Giguère, C.-A., Quataert, E., et al. 2017, MNRAS, Stone, M. A., Lyu, J., Rieke, G. H., & Alberts, S. 2023a, ApJ, 953, 180
472, L109 Stone, M. A., Lyu, J., Rieke, G. H., Alberts, S., & Hainline, K. N. 2023b, arXiv
Bañados, E., Venemans, B. P., Mazzucchelli, C., et al. 2018, Nature, 553, 473 e-prints, arXiv:2310.18395
Bachetti, M., Harrison, F. A., Walton, D. J., et al. 2014, Nature, 514, 202 Tanaka, T. & Haiman, Z. 2009, ApJ, 696, 1798
Barai, P., Gallerani, S., Pallottini, A., et al. 2018, MNRAS, 473, 4003 Tchekhovskoy, A. & Giannios, D. 2015, MNRAS, 447, 327
Blandford, R. D. & Znajek, R. L. 1977, MNRAS, 179, 433 Tenneti, A., Di Matteo, T., Croft, R., Garcia, T., & Feng, Y. 2018, MNRAS, 474,
Choi, E., Ostriker, J. P., Naab, T., & Johansson, P. H. 2012, ApJ, 754, 125 597
Davies, F. B., Hennawi, J. F., & Eilers, A.-C. 2019, ApJ, 884, L19
Tremmel, M., Governato, F., Volonteri, M., & Quinn, T. R. 2015, MNRAS, 451,
Di Matteo, T., Croft, R. A. C., Feng, Y., Waters, D., & Wilkins, S. 2017, MNRAS,
1868
467, 4243
Tremmel, M., Karcher, M., Governato, F., et al. 2017, MNRAS, 470, 1121
Di Matteo, T., Springel, V., & Hernquist, L. 2005, Nature, 433, 604
Uchiyama, H., Toshikawa, J., Kashikawa, N., et al. 2018, PASJ, 70, S32
Dubois, Y., Gavazzi, R., Peirani, S., & Silk, J. 2013, MNRAS, 433, 3297
Dubois, Y., Pichon, C., Welker, C., et al. 2014, MNRAS, 444, 1453 Volonteri, M., Silk, J., & Dubus, G. 2015, ApJ, 804, 148
Duras, F., Bongiorno, A., Ricci, F., et al. 2020, A&A, 636, A73 Xie, F.-G. & Yuan, F. 2012, MNRAS, 427, 1580
Escala, A., Larson, R. B., Coppi, P. S., & Mardones, D. 2004, ApJ, 607, 765 Yang, J., Wang, F., Fan, X., et al. 2023, ApJ, 951, L5
Fan, X., Bañados, E., & Simcoe, R. A. 2023, ARA&A, 61, 373 Yuan, F. & Narayan, R. 2014, ARA&A, 52, 529
Fan, X., Strauss, M. A., Richards, G. T., et al. 2006, AJ, 131, 1203 Yue, M., Eilers, A.-C., Simcoe, R. A., et al. 2023, arXiv e-prints,
Fiacconi, D., Mayer, L., Madau, P., et al. 2017, MNRAS, 467, 4080 arXiv:2309.04614
Greene, J. E., Labbe, I., Goulding, A. D., et al. 2023, arXiv e-prints,
arXiv:2309.05714
Hahn, O. & Abel, T. 2013, MUSIC: MUlti-Scale Initial Conditions, Astrophysics
Source Code Library
Harikane, Y., Zhang, Y., Nakajima, K., et al. 2023, arXiv e-prints,
arXiv:2303.11946
Hopkins, P. F. 2015, MNRAS, 450, 53
Hopkins, P. F., Cox, T. J., Hernquist, L., et al. 2013, MNRAS, 430, 1901
Jiang, Y.-F., Blaes, O., Stone, J. M., & Davis, S. W. 2019, ApJ, 885, 144
Kormendy, J. & Ho, L. C. 2013, ARA&A, 51, 511
Labbe, I., Greene, J. E., Bezanson, R., et al. 2023, arXiv e-prints,
arXiv:2306.07320
Latif, M. A., Bovino, S., Grassi, T., Schleicher, D. R. G., & Spaans, M. 2015,
MNRAS, 446, 3163
Latif, M. A., Whalen, D. J., Khochfar, S., Herrington, N. P., & Woods, T. E.
2022, Nature, 607, 48
Lin, D., Guillochon, J., Komossa, S., et al. 2017, Nature Astronomy, 1, 0033
Lupi, A. & Bovino, S. 2020, MNRAS, 492, 2818
Lupi, A., Haardt, F., Dotti, M., et al. 2016, MNRAS, 456, 2993
Lupi, A., Haiman, Z., & Volonteri, M. 2021, MNRAS, 503, 5046
Lupi, A., Pallottini, A., Ferrara, A., et al. 2020, MNRAS, 496, 5160
Lupi, A., Volonteri, M., Decarli, R., Bovino, S., & Silk, J. 2022, MNRAS, 510,
5760
Lupi, A., Volonteri, M., Decarli, R., et al. 2019, MNRAS, 488, 4004
Ma, L., Hopkins, P. F., Ma, X., et al. 2021, MNRAS, 508, 1973
Madau, P., Haardt, F., & Dotti, M. 2014, ApJ, 784, L38
Maiolino, R., Scholtz, J., Curtis-Lake, E., et al. 2023, arXiv e-prints,
arXiv:2308.01230
Martizzi, D., Faucher-Giguère, C.-A., & Quataert, E. 2015, MNRAS, 450, 504
Massonneau, W., Dubois, Y., Volonteri, M., & Beckmann, R. S. 2023a, A&A,
669, A143
Massonneau, W., Volonteri, M., Dubois, Y., & Beckmann, R. S. 2023b, A&A,
670, A180
Mignoli, M., Gilli, R., Decarli, R., et al. 2020, A&A, 642, L1
Mortlock, D. J., Warren, S. J., Venemans, B. P., et al. 2011, Nature, 474, 616
Narayan, R., Igumenshchev, I. V., & Abramowicz, M. A. 2003, PASJ, 55, L69
Neeleman, M., Novak, M., Venemans, B. P., et al. 2021, ApJ, 911, 141
Neumayer, N., Seth, A., & Böker, T. 2020, A&A Rev., 28, 4
Padoan, P., Haugbølle, T., & Nordlund, Å. 2012, ApJ, 759, L27
Pezzulli, E., Valiante, R., & Schneider, R. 2016, MNRAS, 458, 3047
Pfister, H., Volonteri, M., Dubois, Y., Dotti, M., & Colpi, M. 2019, arXiv e-prints
[arXiv:1902.01297]
Planck Collaboration, Ade, P. A. R., Aghanim, N., et al. 2016, A&A, 594, A13
Regan, J. A., Downes, T. P., Volonteri, M., et al. 2019 [arXiv:1811.04953]
Reines, A. E. & Volonteri, M. 2015, ApJ, 813, 82
Ricarte, A., Narayan, R., & Curd, B. 2023, ApJ, 954, L22
Sala, L., Cenci, E., Capelo, P. R., Lupi, A., & Dotti, M. 2021, MNRAS, 500,
4788
Sassano, F., Capelo, P. R., Mayer, L., Schneider, R., & Valiante, R. 2023, MN-
RAS, 519, 1837
Sa̧dowski, A., Lasota, J.-P., Abramowicz, M. A., & Narayan, R. 2016, MNRAS,
456, 3915
Schauer, A. T. P., Regan, J., Glover, S. C. O., & Klessen, R. S. 2017, MNRAS,
471, 4878
Schaye, J., Crain, R. A., Bower, R. G., et al. 2015, MNRAS, 446, 521
Shakura, N. I. & Sunyaev, R. A. 1973, A&A, 24, 337
Sadowski,
˛ A., Lasota, J.-P., Abramowicz, M. A., & Narayan, R. 2016, MNRAS,
456, 3915

Article number, page 7 of 7