Mon. Not. R. Astron. Soc. 000, 1–16 (0000) Printed 3 August 2020 (MN LATEX style file v2.

2)

Lyα Blobs from Cold Streams Undergoing

Kelvin-Helmholtz Instabilities

Nir Mandelker1,2⋆, Frank C. van den Bosch1, Daisuke Nagai1,3, Avishai Dekel4,
Yuval Birnboim4, Han Aung3
arXiv:2003.01724v2 [astro-ph.CO] 30 Jul 2020

1 Department of Astronomy, Yale University, PO Box 208101, New Haven, CT, USA;
2 HeidelbergerInstitut für Theoretische Studien, Schloss-Wolfsbrunnenweg 35, 69118 Heidelberg, Germany;
3 Department of Physics, Yale University, New Haven, CT 06520, USA;
4 Centre for Astrophysics and Planetary Science, Racah Institute of Physics, The Hebrew University, Jerusalem 91904, Israel

ABSTRACT
We present an analytic toy model for the radiation produced by the interaction
between cold streams thought to feed massive halos at high redshift and their hot
CGM. We begin by deriving cosmologically motivated parameters for the streams as
they enter the halo virial radius, Rv , as a function of halo mass and redshift. For
1012 M⊙ halos at z = 2, we find the stream density to be nH,s ∼ (0.1 − 5) × 10−2 cm−3 ,
a factor of δ ∼ (30 − 300) times denser than the hot CGM, while stream radii are in
the range Rs ∼ (0.03 − 0.50)Rv . As streams accelerate towards the halo centre, they
become denser and narrower. The stream-CGM interaction induces Kelvin-Helmholtz
Instability (KHI), which leads to entrainment of CGM mass by the stream and to
stream deceleration by momentum conservation. Assuming the entrainment rates
derived by Mandelker et al. (2020) in the absence of gravity can be applied locally at
each halocentric radius, we derive equations of motion for the stream in the halo. Using
these, we derive the net acceleration, mass growth, and energy dissipation induced by
the stream-CGM interaction, as a function of halo mass and redshift, for different
CGM density profiles. For the range of model parameters considered, we find that the
interaction induces dissipation luminosities Ldiss > 1042 erg s−1 within ∼ < 0.6R of
v
12
halos with Mv > 10 M⊙ at z = 2. The emission scales with halo mass and redshift
approximately as ∝ Mv (1 + z)2. The magnitude and spatial extent of the emission are
consistent with observed Lyα blobs, though better treatment of the UV background
and self-shielding is needed to solidify this conclusion.
Key words: cosmology — galaxies: evolution — galaxies: formation —
hydrodynamics — instabilities

1 INTRODUCTION Lyα emission, such as a starburst or AGN (Saito et al.

2006; Nilsson et al. 2006; Yang et al. 2009), though in
Hundreds of extended Lyα sources, known as Lyα some cases there is evidence for an obscured central
blobs (LABs) have been observed at redshifts z > source (Matsuda et al. 2007; Geach et al. 2007, 2009;
2 (Steidel et al. 2000, 2011; Palunas et al. 2004; Prescott et al. 2008). LABs appear different from more
Matsuda et al. 2004, 2006, 2011; Nilsson et al. 2006; recently detected giant Lyα nebulae (Cantalupo et al.
Saito et al. 2006; Smith & Jarvis 2007; Yang et al. 2014; Hennawi et al. 2015; Borisova et al. 2016;
2009, 2010). These have luminosities of LLyα ∼ Martin et al. 2016; Arrigoni Battaia et al. 2018, 2019).
1042 − 1044 erg s−1 , though many fainter sources exist The latter are typically larger, extending > 100 kpc,
as well (Rauch et al. 2008; Steidel et al. 2011). Their beyond the expected virial radius of their host dark
spatial extents are typically several tens of kpc, matter halo, and associated with obvious AGN.
occasionally reaching up to ∼ 100 kpc. Often there
is no obvious central source capable of powering the The power source of observed LABs remains
unclear. Some have speculated that all LABs are
powered by a (potentially obscured) central AGN or
⋆ E-mail: [email protected] starburst, either by photoionization or by ejection

c 0000 RAS
2 Mandelker et al.
of superwinds into the halo (Haiman & Rees 2001; emphasize, though, that no such shocks have been
Ohyama et al. 2003; Mori, Umemura & Ferrara 2004; explicitly identified in the cosmological simulations.
Weidinger, Møller & Fynbo 2004; Weidinger et al. Goerdt et al. (2010) presented a similar toy
2005; Wilman et al. 2005; Laursen & Sommer-Larsen model for LABs resulting from gravitational heating
2007; Geach et al. 2009). Another possibility is that of cold streams. Using AMR cosmological zoom-in
cooling radiation of gas accreting onto the halo and/or simulations of ∼ 1012 M⊙ halos at z ∼ 2
onto the central galaxy fuels the LABs (Fardal et al. (Ceverino, Dekel & Bournaud 2010), they confirmed
2001; Birnboim & Dekel 2003; Furlanetto et al. 2005; their model and found that cold streams resemble
Nilsson et al. 2006; Smith & Jarvis 2007; LABs in terms of luminosity, morphology, and extent.
Dijkstra, Haiman & Spaans 2006; Dijkstra & Loeb However, they also did not identify the mechanism by
2009; Goerdt et al. 2010; Faucher-Giguère et al. 2010; which the gravitational energy released by falling down
Matsuda et al. 2011). the potential well was converted into radiation.
Faucher-Giguère et al. (2010) analysed SPH
The cooling radiation scenario is particularly cosmological zoom-in simulations, with comparable
intriguing in the context of the cold-stream model of
mass and redshift to those analysed by Goerdt et al.
galaxy formation, whereby massive galaxies at high (2010). They found that the Lyα luminosity produced
redshifts are fed by narrow streams of dense gas by cooling radiation in cold streams was one to two
which trace cosmic web filaments (Dekel & Birnboim
orders of magnitude lower than in luminous LABs. They
2006; Dekel et al. 2009). Owing to their high densities argued that the main difference between their results
and short cooling times, the gas in these streams and those of Goerdt et al. (2010) was in the treatment
is not expected to shock at the virial radius.
of self-shielding of dense gas from the UV background.
Rather, the streams maintain a temperature of The simulations used in Goerdt et al. (2010) assumed
> 104 K and are thought to penetrate the hot
Ts ∼ that gas with density n > 0.1 cm−3 was self shielded,
circumgalactic medium (CGM) and reach the central
though in their estimates of the Lyα emission they
galaxy in roughly a halo crossing time. Such cold assumed that even lower density gas was in collisional
streams are ubiquitous in cosmological simulations ionization equilibrium. Faucher-Giguère et al. (2010),
(e.g. Kereš et al. 2005; Ocvirk, Pichon & Teyssier 2008;
based on radiative transfer calculations, assumed gas
Dekel et al. 2009; Ceverino, Dekel & Bournaud 2010; with n > 0.01 cm−3 to be self-shielded. This lowered
Faucher-Giguère, Kereš & Ma the temperature of stream gas due to decreased UV
2011; van de Voort et al. 2011), where they are found
heating, and thus lowered the overall Lyα luminosity.
to supply the halo with gas at rates comparable
to the predicted cosmological accretion rate, with a However, Faucher-Giguère et al. (2010)
significant fraction of the gas reaching the central also acknowledged that it is plausible that differences in
galaxy (Dekel et al. 2009, 2013). Many observed LABs the hydrodynamic method, SPH vs AMR, contributed
and giant Lyα nebulae have filamentary morphologies, to the difference in the predicted emission, especially
with spatial and kinematic properties consistent with given the low resolution in the streams in both studies.
predictions for cold streams (Nilsson et al. 2006; The resolution in most state-of-the-art simulations is
Saito et al. 2006; Smith & Jarvis 2007; Matsuda et al. adaptive, such that the effective mass of each resolution
2011; Cantalupo et al. 2014; Martin et al. 2014a,b, element is fixed. The spatial resolution thus becomes
2019; Borisova et al. 2016; Fumagalli et al. 2017; very poor, typically ∼ kpc scales, in the low density
Leclercq et al. 2017; Arrigoni Battaia et al. CGM near the virial radius (e.g. Nelson et al. 2016).
2018). Absorption line studies of the CGM around While this may be enough to resolve the largest and
high-redshift massive galaxies are also suggestive of most diffuse streams (see §3 below), dense streams can
dense, cold, inspiralling gas streams (Fumagalli et al. be as narrow as a few kpc (§3; Padnos et al. 2018;
2011; Goerdt et al. 2012; van de Voort et al. 2012; Mandelker et al. 2018) and are therefore unresolved
Bouché et al. 2013, 2016; Prochaska, Lau & Hennawi in cosmological simulations. Moreover, the different
2014). physical processes and complex subgrid models
employed by different cosmological simulations make
Dijkstra & Loeb (2009) presented an analytic toy it difficult to gain a physical understanding of
model for Lyα radiation from cold streams. They stream evolution and generalize results, which are
found that under reasonable assumptions, motivated by found to be sensitive to the numerical approach.
cosmological simulations of the time, cooling radiation For example, early simulations using the moving
from cold streams could account for all the observed mesh code AREPO (Springel 2010; Vogelsberger et al.
LABs. In their model, the Lyα emission is powered 2012) found that streams heat-up and dissolve
at ∼ > 0.5R
by the gravitational energy lost as the stream flows v (Nelson et al. 2013), while comparable
down the potential well of the dark matter halo. They Eulerian AMR (Ceverino, Dekel & Bournaud 2010;
found that if at least ∼ 20% of this energy went into Danovich et al. 2015) and Lagrangian
heating the stream and was subsequently radiated away, SPH (Kereš et al. 2005; Faucher-Giguère et al. 2010)
the resulting emission would resemble LABs. However, simulations found them to remain cold and collimated
no clear mechanism for tapping into this energy was until ∼ 0.25Rv .
identified, and the model instead simply assumed that To overcome these issues, several recent works have
this could occur through a series of weak shocks. We studied the evolution

c 0000 RAS, MNRAS 000, 1–16

 Lyα Blobs from Cold Streams 3
of cold streams using analytic models and idealized, stream and the background, δ = ρs /ρb . We assume
high-resolution simulations. Initial work focusing on that the stream and the background are in pressure
pure hydro instabilities found that sufficiently narrow equilibrium, and therefore Tb /µb = δ Ts /µs , where
streams would be disrupted by Kelvin-Helmholtz µb ∼ µs ∼ 0.6 are the mean molecular weights in the
instabilities (KHI) within the CGM, prior to reaching background and stream respectively. For µs = µb , the
the central galaxy (Mandelker et al. 2016; Padnos et al. Mach number with respect to the sound speed in the
2018; Mandelker et al. 2019). However, subsequent stream is Ms = Vs /cs = δ 1/2 Mb .
work including either self-gravity (Aung et al. 2019), The shearing motion between the stream and the
magnetic fields (Berlok & Pfrommer 2019), or radiative background induces KHI, which leads to the formation
cooling (Mandelker et al. 2020; hereafter M20), found of a turbulent mixing region surrounding the stream2 .
that each of these effects stabilizes the stream against The typical density and temperature in this region are3
disruption. (Begelman & Fabian 1990; Gronke & Oh 2018)
Besides their importance for studying stream 1/2
evolution and survival, the aforementioned studies ρmix ∼ (ρb ρs ) = δ −1/2 ρs , (1)
identify a self-consistent dissipation mechanism acting
on the stream, induced by its interaction with the hot 1/2
Tmix ∼ (Tb Ts ) = δ 1/2 Ts . (2)
CGM. In particular, M20, the only such study thus far
to include radiative cooling, found that the formation The cooling time in the mixing region is thus
of a turbulent mixing zone between the stream and kB Tmix
the halo following the onset of KHI, resulted in halo tcool, mix = , (3)
(γ − 1)nmix Λ(Tmix )
gas cooling and condensing onto the stream. This led
to the loss of kinetic energy from the stream and where γ is the adiabatic index of the gas, kB is
thermal energy from the background. M20 found that Boltzmann’s constant, nmix is the particle number
roughly half of this energy was radiated by gas with density in the mixing region, and Λ(Tmix ) is the cooling
temperatures T < 5 × 104 K, and would thus mostly be function evaluated at Tmix .
emitted as Lyα. However, the results of M20 were based In the non-radiative case (Padnos et al. 2018;
on analysis of an infinite stream,1 with constant density Mandelker et al. 2019), the turbulent mixing layer
and cross section, and without external acceleration. expands into both the stream and the background and
None of these assumptions are expected to hold for its width is well approximated by
realistic streams. Motivated by these results, we present
in this paper a toy model generalising the results of h(t) = αVs t, (4)
M20 to account for the effect of a halo potential on the where the dimensionless growth rate, α, is (Dimotakis
stream. Using this model, we make predictions for the 1991)
total radiation produced by streams with cosmologically 

2
motivated properties within dark matter halos. α ≃ 0.21 × 0.8exp −3Mtot + 0.2 , (5)
The remainder of this paper is organized as follows. with Mtot = Vs /(cs + cb ). This approximation is an
In §2 we review the main conclusions of M20 regarding excellent fit for 2d planar slabs and also for 3d cylinders
KHI in radiatively cooling streams and the associated
energy dissipation rates. In §3 we derive cosmologically
2 On a fundamental level, the mixing seen in the numerical
motivated scaling relations for stream properties as a
function of halo mass and redshift. In §4 we present simulations referenced here is artificial, since the viscous scale
our toy model for stream evolution in dark matter is not resolved. The mixing between the two phases is likely
driven by turbulence that, by construction, cascades down to
halos, and make predictions for the total amount of
the viscosity length which is the same as the diffusion length.
radiation that may result from the instability. In §5 we The diffusion scale depends on the Coloumb interaction length of
discuss limitations of our model, and we summarize our an electron (its mean free path before Thompson scattering). At
conclusions in §6. ρ ∼ 10−26 gr cm−3 and T ∼ 105 K, this is approximately 10−3 pc.
The assumption is that in reality, the fluids will eventually mix
at this scale, and that the dissipation rate (through cooling)
2 THEORETICAL FRAMEWORK is conserved throughout the inertial range, explaining why the
simulations are converged in terms of the total dissipation rate
In this section, we summarize the main results of M20 (Gronke & Oh 2018, 2020; Fielding et al. 2020; M20).
3 In practice, the mixing region contains a broad distribution of
regarding KHI in radiatively cooling streams, and the
densities and temperatures rather than a charecteristic value (e.g.
associated energy dissipation rates. The system we M20, figure 10). However, studies have shown that the evolution
consider is a cylindrical stream with radius Rs , density of the mixing region can be well described by considering a
ρs , and temperature Ts , flowing with velocity Vs through fluid with ρmix and Tmix as defined above (Gronke & Oh 2018,
a static background (Vb = 0) with density ρb and 2020; Ji, Oh & Masterson 2019; M20), though other studies
temperature Tb . We define the Mach number of the have explored different characteristic densities and temperatures
flow with respect to the sound speed in the background, (Li et al. 2020; Hobbs & Feldmann 2020). It has also been shown
Mb = Vs /cb , and the density contrast between the that when the mixing region is well resolved, it maintains
pressure equilibrium with the unmixed phases even as it cools
(Fielding et al. 2020), consistent with the definitions of ρmix and
1 In practice, a stream in a periodic box. Tmix .

c 0000 RAS, MNRAS 000, 1–16

4 Mandelker et al.
< R . We thus obtain the timescale for
so long as h ∼ cooling time at T = 1.5Ts , which is roughly the minimal
s
the mixing layer to grow to the size of the stream, i.e. cooling time. However, M20 note that any temperature
h = Rs , in the range ∼ (1.2 − 2)Ts works equally well. The
Rs density is given by assuming pressure equilibrium.
tshear = . (6) As more mass is entrained in the flow, conservation
αVs
of momentum causes the stream to decelerate. The
For δ ∼ 100 and Mb ∼ 1, we get α ∼ 0.05 and Vs ∼ 10cs . velocity of the stream as a function of time is well
In this case, tshear is comparable to the stream sound approximated by6
crossing time,
2Rs Vs,0
tsc = . (7) Vs (t) = . (11)
cs 1 + t/tent

M20 found that a key parameter for determining This reduces the kinetic energy per-unit-length
stream evolution is the ratio tcool, mix /tshear 4 . If tshear < associated with bulk laminar flow, which is well fit by
tcool, mix , the evolution proceeds similarly to the Ek,0
non-radiative case studied by Mandelker et al. (2019), Ek (t) = . (12)
and the stream is eventually disrupted by KHI. 1 + t/tent
However, if tcool, mix < tshear , then background gas In addition to the stream losing kinetic energy, the
entrained in the mixing layer cools and condenses onto background gas entrained by the stream loses thermal
the stream before it is disrupted by KHI. The stream energy, at a rate per-unit-length of
thus remains cold, dense and collimated, and is not
disrupted by hydrodynamic instabilities. Rather, the ṁc2b 9m0 c2b
stream mass actually increases with time, as it entrains Ėth,b = ṁ(es − eb ) ≃ − =− , (13)
γ(γ − 1) 10tent
more and more gas from its hotter surroundings.
Similar conclusions were reached in recent studies where eb = P/[(γ − 1)ρb ] is the thermal energy per unit
of high-velocity clouds in a hot CGM environment mass of the background fluid which is larger than that
(Gronke & Oh 2018, 2020). The condition tcool, mix = in the cold component by a factor δ ≫ 1, γ = 5/3 is
tshear leads to a critical stream radius, the adiabatic index of the gas, and c2b = γP/ρb is the
3/2 Ts,4 adiabatic sound speed in the background. In the final
Rs,crit ≃ 0.3 kpc α0.1 δ100 Mb , (8) equation we have used eq. (9) to approximate ṁ.
ns,0.01 Λmix,−22.5
M20 found that the cooling radiation emitted
where Ts,4 = Ts /104 K, ns,0.01 = nH,s /0.01 cm−3 , per-unit-length by the stream is very well approximated
Λmix,−22.5 = Λ(Tmix )/10−22.5 erg s−1 cm3 , δ100 = by
δ/100, and α0.1 = α/0.1.
The ratio tcool, mix /tshear = (Rs /Rs,crit)−1 , so 5
Ldiss ≃ Ėk + Ėth . (14)
streams with Rs > Rs,crit grow in mass rather than 3
dissolve. M20 derived an approximate expression for the
entrainment rate of hot gas onto the stream. The cold The factor 5/3 accompanying Ėth accounts for the fact
mass-per-unit-length (hereafter line-mass) as a function that the pressure of the background gas just outside the
of time is given by5 mixing layer remains roughly constant, and therefore
  the condensation of background gas onto the stream is
t well approximated as an isobaric cooling flow. In this
m(t) = m0 1 + , (9) case, the emitted radiation is given by the difference in
tent
enthalpy, rather than energy, between the initial and
where m0 = πRs2 ρs is the initial stream line-mass, and final states (Fabian 1994). In eq. (14) we have ignored
we have introduced the entrainment timescale, any net heating of the stream, which is found to be
 1/4 very small (M20). For stream temperatures of order
δ tcool,s Ts ∼ 104 K and background temperatures of order
tent = tsc , (10)
2 tsc Tb ∼ 106 K, roughly half the luminosity is emitted at
where tcool,s is the cooling time at the stream temperatures T < 5 × 104 K, and is thus expected to
temperature and density. In practice, M20 assumed contribute substantially to Lyα radiation (M20).
the stream to be in thermal equilibrium with a UV
background, so the net cooling time at Ts is formally
infinite. Therefore, tcool,s is replaced by tcool,1.5Ts , the
6 When t
cool, mix < tshear , the entrained background material
very rapidly mixes with the stream material (M20, figures 3 and
4 There is some controversy in the literature over whether the 12). In this case, the whole stream moves at roughly Vs , save for
relevant cooling time is tcool, mix or tcool, hot , i.e. the cooling a very narrow region near the outer edge of the mixing region
time in the hot phase (Li et al. 2020). We here adopt tcool, mix , where the velocity quickly drops to 0 towards the background.
When tcool, mix ∼ > t
following Gronke & Oh (2018, 2020), and M20. shear , the velocity distribution is much wider,
5 See Gronke & Oh (2020) for a similar expression for the case and there can be a strong velocity gradient between the stream
of spherical cold clouds. axis and its interface.

c 0000 RAS, MNRAS 000, 1–16

 Lyα Blobs from Cold Streams 5
3 COSMOLOGICALLY MOTIVATED STREAM with Θs ∼ 0.5 − 27 . The sound speed in the stream is
PROPERTIES s
γkB Ts −1/2
cs = ≃ 18.5 km s−1 Θ1/2
s µs,0.6 , (20)
In Padnos et al. (2018), section 5.1, we estimated Mb , δ, µs m p
and Rs /Rv for streams near the virial radius of a halo
supporting a virial shock. We revisit this calculation with µs,0.6 = µs /0.6. If the stream is strongly
here, both to correct an error in the previous estimate of self-shielded and predominantly neutral, then µs,0.6 ∼
Rs /Rv , and because in the context of radiatively cooling 2. Assuming the stream and the hot halo are in pressure
streams we require additional parameters, such as the equilibrium, the density contrast is given by
stream density, which were not addressed previously. 2/3
δ = ρs /ρh = Th /Ts ≃ 100 M12 (1 + z)3 (Θh /Θs ) . (21)
For an alternative derivation of stream properties
in halos without a virial shock, see Mandelker et al. Taking into account uncertainties in Θh and Θs , density
(2018). contrasts of δ ∼ (30 − 300) are reasonable for a Mv ∼
1012 M⊙ halo at z ∼ 2.
The stream velocity as it enters the halo is
proportional to the virial velocity, V0 = ηVv , with The stream density upon entering the halo virial
< η < 1 (Goerdt & Ceverino 2015). The maximal
0.5 ∼ radius, ρ0 , is given by the density contrast, δ, and the
∼ √
value η is likely to obtain is 2 if the stream enters density of the hot halo gas at Rv . The latter is given by
the halo at the escape velocity. The virial velocity and ρh (Rv ) = ∆(Rv )ρh ≃ ∆(Rv )fh fb 18π 2 ρu (z), (22)
> 1 are given by (Dekel et al. 2013)
radius at z ∼
where ρh is the mean density of the hot component in
1/3 1/2
the halo, ρu (z) is the mean density of the universe at
Vv ≃ 200 km s−1 M12 (1 + z)3 , (15) redshift z, fb ≃ 0.17 is the universal baryon fraction,
and fh denotes the fraction of the baryonic mass within
the halo which is in the hot gas component. Simulations
suggest this is 0.3 ∼< f < 0.4 for halo masses M
1/3 h ∼ 12 ∼
Rv ≃ 100 kpc M12 (1 + z)−1
3 , (16) > 2 (Roca-Fàbrega et al. 2019),
(0.5 − 2) at redshift z ∼
while observations of M12 ∼ (1 − 10) halos at
with M12 = Mv /1012 M⊙ and (1 + z)3 = (1 + z)/3. z∼>
0.1 suggest fh ∼ 0.2 − 0.3 (Singh et al. 2018).
The halo temperature at Rv is proportional to the virial We define fh,0.3 = fh /0.3. ∆(Rv ) is the ratio of the
temperature, Th = Θh Tv , with Θh ∼ > 3/8 given by the
density at Rv to the mean density in the halo, ρv =
jump conditions at the virial shock (Dekel & Birnboim 3Mv /(4πRv3 ). For a singular isothermal sphere ∆(Rv ) =
2006), though it may be larger due to additional heating 1/3, while for an NFW halo with concentration c =
by feedback from the central or satellite galaxies, or (5, 10, 20), ∆(Rv ) ∼ (0.24, 0.19, 0.14) respectively. We
when the virial shock propagates outwards. The virial write ∆1/6 = ∆(Rv )/(1/6). Assuming cosmological
temperature is given by
parameters Ωm = 0.3 and H0 = 70 km/s/Mpc, we
2/3
obtain for the stream density at Rv
Tv ≃ 1.5 × 106 K M12 (1 + z)3 . (17)
ρ0 ≃ 1.1 × 10−26 gr cm−3 (1 + z)33 δ100 feh , (23)
The sound speed in the halo is where δ100 = δ/100, and feh = ∆1/6 fh,0.3 ∼ (1 − 3).
s For a Hydrogen mass fraction of X = 0.76 this yields a
γkB Th 1/3 1/2 1/2 Hydrogen number density in the stream at Rv of
ch = ≃ 185 km s−1 M12 (1+z)3 Θh , (18)
µh m p nH,0 ≃ 5.1 × 10−3 cm−3 (1 + z)33 δ100 feh . (24)
−3
where we have assumed that µh ∼ 0.6, appropriate for We hereafter write ns,0.01 = nH,0 /(0.01 cm ).
fully ionized gas with roughly primordial composition. The stream radius at Rv can be constrained using
The stream Mach number upon entering the halo is thus the mass flux through the stream,

−1/2
Ṁs = πRs2 ρ0 V0 = fs fb Ṁv , (25)
Mb = V0 /ch ≃ 1.1 η Θh . (19)
where Ṁv is the total mass accretion rate onto the halo
Taking uncertainties in η and Θh into account, we have virial radius, and fs is the fraction of baryonic accretion
< M < 2.25.
that 0.75 ∼ b ∼
along the gas stream. Cosmological simulations suggest

If the stream is in thermal equilibrium with the 7 We note that much lower values of Θ could be possible if
UV background at z ∼ 2, its temperature is Ts ∼ s
the stream were dense enough and/or metal-rich enough to allow
1.5 × 104 (M20), with a mild dependence on density
cooling to much lower temperatures in less than a halo crossing
and metallicity. In practice, the temperature can be time. While this may be the case at z > 5, it is unlikely to occur
lower if the stream is self-shielded, or higher if the at redshifts z < 4 unless the metallicity in the stream is ∼ 0.1Z⊙
stream is highly turbulent before entering the halo, (Mandelker et al. 2018). This is very high compared to values
with turbulent dissipation timescales comparable to the found in cosmological simulations and observed in both LABs
cooling time. We therefore assume Ts = Θs 1.5 × 104 K, and Lyman limit systems at these redshifts.

c 0000 RAS, MNRAS 000, 1–16

6 Mandelker et al.
10000 1 1

1000 0.1

100 0.01 0.1

10 10-3

1 -4 0.01
10
11 12 13 14 11 12 13 14 11 12 13 14
10 10 10 10 10 10 10 10 10 10 10 10

Figure 1. Stream properties upon entering the halo virial radius, according to our model. We show, as a function of halo mass on the
x axis, the density contrast between the stream and the background, δ (eq. 21, left), the stream Hydrogen number density, nH,0 (eq. 24,
centre), and the ratio of stream radius to halo virial radius, Rs /Rv (eq. 28, right). Green, blue, and red lines represent redshifts z = 1,
2, and 4 respectively. The thin lines spanned by error bars represent√the range of stream properties obtained by varying the model
parameters within the ranges Θh ∈ (3/8, 1.0), Θs ∈ (0.5, 2.0), η ∈ (0.5, 2), feh ∈ (1.0, 3.0), and se ∈ (0.3 − 3.0) (see text for details). The
thick lines represent our fiducial model, where all the above parameters have values of 1.0. Note that Rs /Rv has no redshift dependence
in our model.

fs ∼ (0.2 − 0.5) with a typical value of fs = 1/3 (i.e. collisional ionization equilibrium (CIE), this is true for
three significant streams; Danovich et al. 2012). In the < <
Ts ∼ 3 × 104 K (e.g. Goerdt et al. 2010), i.e. Θs ∼ 2. Of
Einstein de Sitter (EdS) regime (valid at z > 1), the course, streams may still be susceptible to local sources
accretion onto the virial radius is well approximated by of UV radiation, such as star-formation in satellite
Dekel et al. (2013) galaxies located along the stream, or starbursts/AGN
5/2 activity in the central galaxy.
Ṁv /Mv ≃ 0.47 Gyr−1 s (1 + z)3 , (26)
To summarize, given the halo mass and redshift,
with halo-to-halo variance encapsulated by the the precise temperatures of the stream and halo gas, Θs
normalization s ∼ (0.5 − 2). Combining eqs. (15), (23), and Θh respectively, determine the density contrast, δ.
(25), and (26), we obtain The stream Mach number, Mb is then interchangeable
 1/2 with η, the ratio of stream to virial velocity. The stream
se
1/3
Rs ≃ 16 kpc M12 (1 + z)3
−1/2 −1/2
δ100 , (27) density is then determined by feh , the normalization
η feh of the hot halo gas density near Rv . Finally, the
with se = sfs /(1/3) ∼ (0.3 − 3). Together with eq. (16) ratio of stream radius to virial radius is set by se, the
we have normalization of the gas accretion rate along the stream.
 1/2 For η ∼ 1, se ∼ (0.3−3) and feh ∼ (1−3), a stream at z =
Rs 1/2 −1/2 se 2 has Rsv ∼ (0.09−0.50), (0.05−0.28), and (0.03−0.16)
≃ 0.16 (1 + z)3 δ100 . (28)
Rv η feh for δ = 30, 100, and 300 respectively. The corresponding
stream densities are ns,0.01 ∼ (0.15 − 0.45), (0.5 − 1.5),
Note that this is larger by a factor ∼ 2 than the and (1.5 − 4.5), with larger values of nH,0 corresponding
corresponding equation (68) in Padnos et al. (2018), to smaller values of Rsv . For a given δ, neither property
which seems to be due to a typo as it is consistent depends on halo mass.
with their equation (67). Using eq. (21) for δ, we see
that Rs /Rv ∝ Mv
−1/3
and is independent of redshift. Figure 1 shows the full plausible range of δ, nH,0 ,
More massive halos are thus fed by relatively narrower and Rsv as a function of halo mass at redshifts z = 1,
streams compared to their virial radii. We hereafter 2, and 4. As motivated above, we allow the model
substitute Rsv ≡ Rs /Rv . parameters to vary in the range Θh ∈ (3/8, 1.0), Θs ∈
√
(0.5, 2.0), η ∈ (0.5, 2), feh ∈ (1.0, 3.0), and se ∈ (0.3 −
The Hydrogen column density perpendicular to the
3.0), while in our fiducial model all these parameters
stream axis is NH,0 = nH,0 Rs . Using eqs. (24) and (27)
have values of 1.0. As noted above, inserting eq. (21)
this yields
into eq. (28) results in Rsv depending only on halo mass
!1/2
and not on redshift, as seen in the right-hand panel.
20 1/3 5/2 1/2 sefeh
−2
NH,0 ≃ 3 × 10 cm M12 (1 + z)3 δ100 . In Fig. 1, we extend our model down to halo masses
η of Mv = 1011 M⊙ . While such halos are below the
(29) critical mass for forming a stable accretion shock at Rv
For 1012 M⊙ halos at z ∼ 2, this is large enough for (Birnboim & Dekel 2003), high-resolution simulations
the streams to be largely self-shielded against the UV suggest that stellar feedback from the central galaxy can
background if the neutral fraction is xHI ∼ > 10−3 . In result in a quasi-stable hot CGM in halos of this mass

c 0000 RAS, MNRAS 000, 1–16

 Lyα Blobs from Cold Streams 7
as well (Fielding et al. 2017). In lower mass halos, a
hot CGM is not expected invalidating the fundamental
assumption of our model, pressure equilibrium between
cold streams and a quasi-stable hot atmosphere. At
the high-mass end, we show results up to Mv =
1014 M⊙ . Streams penetrating such massive halos may
be pre-heated in the IGM prior to entering the halo
(Dekel & Birnboim 2006; Birnboim, Padnos & Zinger
2016), particularly at z ∼ 1. This would result in very
large values of Θs not considered here, so these results
shoule be treated with caution.

4 STREAM EVOLUTION AND ENERGY

DISSIPATION IN DARK MATTER HALOS
Here we outline our toy model for the evolution of
streams in dark matter halos as they make their
way towards the central galaxy. This is an extension
of similar models presented in Padnos et al. (2018)
(Appendix F) and Mandelker et al. (2019) (section
5.3), now accounting for radiative cooling and energy
dissipation, as well as self-consistently accounting for Figure 2. As in Fig. 1, but showing the ratio of stream radius to
mass growth and deceleration of the stream in the the critical radius for cooling to dominate over KHI, Rs /Rs,crit
equations of motion. We begin in §4.1 by addressing (eq. 31). Streams with Rs > Rs,crit are likely to survive their
stream survival within dark matter halos. In §4.2 we journey through the CGM towards the central galaxy while
discuss how stream properties scale with halocentric remaining cold and coherent, while those with Rs < Rs,crit
radius. In §4.3 we derive equations of motion for the (grey shaded region) are susceptible to disruption by KHI. We
assume the same parameter ranges as in Fig. 1, and evaluate
stream as it penetrates the halo, and in §4.4 we evaluate
α0.1 for each parameter combination using eq. (5). We further
the dissipation and resulting luminosity induced by the assume Λmix,−22.5 ∈ (0.5, 2.0), with a fiducial value of 1.0. While
stream-halo interaction. In §4.5 and Fig. 3 we present Rs /Rs,crit spans almost 4 orders of magnitude at each halo mass
example solutions of our model. and redshift, there is a trend for the ratio to decrease towards
larger halo masses and lower redshifts, making such streams more
susceptible to disruption.
4.1 Stream Survival in Halos

We begin by addressing the survival of cold streams

in hot halos, which is necessary for the validity of our Haardt & Madau (1996) UV background. In this case,
model. This issue was addressed by M20, who found for δ = 300, se = 0.3, and feh = 1, we obtain Rs /Rs,crit ∼
that cold streams are likely to survive the journey 1.3 for M12 = (1 + z)3 = Mb = η = Θs = α0.1 = 1. This
>
from the virial radius to the central galaxy, but can suggests that for halos with Mv ∼ 1012 M⊙ at z ∼ 2,
be strengthened here using our more accurate estimate Rs > Rs,crit in virtually all cases.
of Rs (eq. 27). Based on the results of M20, summarized Inserting eq. (21) into eq. (31), we find that
in §2, streams with Rs < Rs,crit (eq. 8), i.e. where −1/3 3/2
tcool, mix > tshear , will be disrupted by KHI similar to the Rs /Rs,crit ∝ M12 (1 + z)3 . The ratio thus decreases
non-radiative case described in Mandelker et al. (2019), in more massive halos at lower redshift, making streams
while streams with Rs > Rs,crit will survive. Inserting in such halos more susceptible to disruption. This
eq. (24) and Ts,4 = 1.5Θs into eq. (8), is qualitatively consistent with the notion that cold
streams do not penetrate very massive haloes at z ∼ < 2

1/2 α0.1 Mb Θs (e.g. Dekel & Birnboim 2006; van de Voort et al. 2011).
Rs,crit ≃ 0.9 kpc (1 + z)−3
3 δ100 . (30) In Fig. 2 we show the ratio Rs /Rs,crit as a function of
e
fh Λmix,−22.5
halo mass for z = 1, 2, and 4, for the same range of
Together with eq. (27), we have parameters considered in Fig. 1. For each set of values
for Mb and δ, we used eq. (5) to evaluate α0.1 . We
Rs 1/3 5/2 −1 considered Λmix,−22.5 ∈ (0.5, 2.0), with a fiducial value
≃ 18 M12 (1 + z)3 δ100
Rs,crit of 1.0, though evaluating the cooling rate in the mixing
(31) layer for each set of parameters does not qualitatively
1/2
se1/2 feh Λmix,−22.5 change the picture for metallicity values ∼ < 0.1Z .
⊙
× . While the fiducial models always yield Rs /Rs,crit >
η 1/2 α0.1 Mb Θs
1, there is a small range of parameters which yield
This is ∼ > 1, even in the extreme case where the
Rs /Rs,crit < 1, particularly at z = 1. Furthermore,
metallicity in the mixing layer is 0, so Λmix,−22.5 ∼ 0.4 for Rs /Rs,crit < 10 the dominance of cooling over KHI
for Tmix ∼ 1.5×105 K, nmix ∼ 5×10−4 cm−3 and a z = 2 becomes somewhat marginal (M20) and streams may

c 0000 RAS, MNRAS 000, 1–16

8 Mandelker et al.
heat up. This occurs for the fiducial model at z = 1 for of the stream
Mv ∼> 5 × 1012 M .  β/2  1/2
⊙
r m(r)
The above discussion is valid as the stream enters rs (r) = Rs , (35)
Rv m0
the halo at Rv , where eqs. (24) and (27) for the
stream density and radius are applicable. As discussed As we will see in §4.5 below, the stream line mass
in §4.2 below, the stream density increases at smaller increases by less than a factor of ∼ 2 from Rv to 0.1Rv .
halocentric radii. Ignoring variations in the stream Thus, as the stream approaches the center of the halo
temperature and velocity within the halo, which are it gets denser and narrower.
expected to be small as justified below, we have Rs ∝ Under our isothermal assumption, the sound speed
−1/2
ns while Rs,crit ∝ n−1
s (eq. 8). The ratio Rs /Rs,crit in both the stream and the background remain constant
is thus expected to increase further as the stream throughout the halo,
penetrates the halo. We conclude that the stream will
not be disrupted by KHI in the halo, and the cold gas cs (r) = cs = const, ch (r) = ch = const. (36)
mass will increase as the stream makes its way towards The stream sound crossing time thus scales as
the central galaxy.  β/2  1/2
r m(r)
tsc (r) = tsc (Rv ) , (37)
Rv m0
4.2 Scaling of Properties Within the Halo
with
In this section we describe how different stream Rs
properties scale with halocentric radius as the stream tsc (Rv ) = 2 = 2Rsv δ 1/2 Mb tv , (38)
cs
penetrates the halo. We assume the stream to be on
a radial trajectory towards the halo centre and in where
local pressure equilibrium at every halocentric radius Rv −3/2
r. Based on the results of M20, who showed that tv = = 0.5 Gyr η −1 (1 + z)3 , (39)
V0
variations to the stream temperature induced by the
instabilities are very small at all times, we assume is the halo crossing time of the stream. The cooling time,
the stream to be isothermal throughout the halo. We in both the stream and the mixing region, scales as
assume the hot CGM to be isothermal with a power-law  −1  β
density profile of the form ρh ∝ r−β . Outside of ∼ nH,s r
tcool (r) = tcool (Rv ) = tcool (Rv ) .
0.1Rv , this assumption is consistent with cosmological nH,0 Rv
simulations of ∼ 1012 M⊙ halos at z = 2 with (40)
β ∼ 2 (van de Voort & Schaye 2012), high resolution Combining eqs. (10), (37), and (40), the entrainment
simulations of isolated halos with M12 ∼ 0.1 − 1 time at radius r is
with β ∼ (1.5 − 2) (Fielding et al. 2017), and stacked  5β/8  3/8
> 0.1 with
r m(r)
observations of halos with M12 ∼ 1 − 10 at z ∼ tent (r) = t0 , (41)
β ∼ 1.2 (Singh et al. 2018). Analytic models for density Rv m0
profiles of the hot CGM in galaxy clusters predict slopes with
of β ∼ 3 near the outskirts of the halo, similar to 1/4
the underlying NFW halo profile (Komatsu & Seljak t0 ≡ tent (Rv ) ≃ 0.2Rsv δ 3/2 Mb τcool tv , (42)
2001). Based on these considerations, in what follows where τcool = [tcool,1.5Ts (Rv )/tsc (Rv )]/0.002. This
we consider values of β = (1 − 3). normalization is motivated by Table 1 in M20,
The stream line-mass is given by extrapolated to nH,0 = 0.01 cm−3 and Rs = 12 kpc,
> >
which are typical values for Mv ∼ 1012 M⊙ and z ∼ 2
m(r) = πrs2 (r)ρs (r). (32) (Fig. 1).
Upon entering the halo, prior to any mass entrainment,
we have
4.3 Equations of Motion
2/3
m0 = πRs2 ρ0 ≃ 1010 M⊙ kpc−1 M12 (1 + z)−2
3
(33) In M20 we derived equations describing the mass
×R2sv ns,0.01 , entrainment and subsequent stream deceleration
induced by the KHI when no external forces are present,
where Rs = rs (Rv ) and we have used eq. (16). which are summarized in §2. We here assume that the
If both the stream and the halo are isothermal and same mass entrainment equation can be applied locally
in local pressure equilibrium, the density contrast, δ, at each radius r within the halo. Note that this is
remains constant throughout the halo. In this case, the different than the assumption made in Padnos et al.
stream density obeys (2018) and Mandelker et al. (2019). In those papers,
we assumed that KHI introduces an effective drag force
 −β
r with a local deceleration rate identical to the case with
nH,s (r) = nH,0 . (34) no external forces, i.e. the time derivative of eq. (11).
Rv
The net acceleration was then given by the sum of
Combining eqs. (32) and (34) we obtain for the radius the gravitational induced acceleration and this local

c 0000 RAS, MNRAS 000, 1–16

 Lyα Blobs from Cold Streams 9
deceleration. As will become evident below, these two mass profile within the halo, including the entrained
assumptions lead to different equations of motion, and background mass,
we find the current assumption to be more physical.
Differentiating eq. (9) yields dµ ητ
= − 1/2 5β/8 3/8 , (52)
dx y x µ
m0
ṁ = . (43)
tent (r) with the boundary condition µ(x = 1) = 1.
Using eq. (42), the characteristic time scale in these
The stream is subject to gravitational acceleration two equations is given by
by the external halo. The external force acting on
the stream per unit length, including the entrained τ ≃ 5R−1 −3/2
Mb−1 τcool .
−1/4
sv δ (53)
material, is
For a 1012 M⊙ halo at z = 2, with our fiducial values
dΦ
Fext = −m(r) , (44) of δ ∼ 100 (eq. 21) and Rsv ∼ 0.16 (eq. 28), τ ∼ 0.03
dr for Mb ∼ τcool ∼ 1. We thus have t0 ≫ tv (eq. 50),
where Φ(r) is the gravitational potential of the dark so the virial crossing time is much less than the mass
matter halo at r. For an NFW halo with concentration entrainment time at Rv . In this case, the stream mass
parameter c, is not expected to increase significantly before reaching
the central galaxy. This is consistent with the notion of
ln (1 + cx) constant mass flux along the stream discussed above,
Φ = −Vv2 , (45)
fc x and is also consistent with model solutions shown in
where §4.5 below.
r
x≡ , (46)
Rv
4.4 Energy Dissipation and Radiation
and
c In the case with no external forces studied in M20,
fc ≡ ln (1 + c) − . (47) the energy dissipation is given by the dissipation of
1+c
kinetic and thermal energy (§2). In this case, however,
We define P = m(r)V (r) as the total momentum we must consider dissipation of mechanical (kinetic
per unit length of stream including the entrained plus potential) and thermal energy. We begin with the
material. Using eq. (43) and the fact that V̇ = mechanical energy of our system. By “our system”, we
V (r)dV /dr, we have here mean the initial stream plus all the mass that
  will become entrained in the stream before it reaches
1 m0 1 dV 2 the central galaxy. Accounting for this extra mass from
Ṗ = m(r) V (r) + , (48)
tent (r) m(r) 2 dr the outset is important, since it has initial potential
energy before becoming entrained in the stream. The
The equation of motion for the stream is then mechanical energy per unit length of the system at
Ṗ = Fext . (49) radius r is thus
Emech (r) = 0.5m(r)V (r)2 + m(r)Φ(r)
Defining dimensionless variables
Z r (54)
V2 Φ m Rv tv dm
y ≡ 2, φ ≡ 2, µ ≡ , τ≡ = , (50) + Φ(r̃)dr̃,
Vv Vv m0 V0 t0 t0 Rint dr̃
eqs. (41) and (44)-(50) can be combined to yield an where Rint is the innermost radius we are considering.
equation for the velocity profile within the halo8 , The final term represents the potential energy of
material not yet entrained by the stream at r, but which
dy 2 η τ y 1/2 dφ will become entrained before the stream reaches Rint .
= 11/8 5β/8 − 2 , (51)
dx µ x dx Note that dm/dr < 0.
with the boundary condition that y(x = 1) = η 2 . We Using Ė = V dE/dr together with eqs. (46),
have used the sign convention that the inward velocity (50)-(52) and (54), we obtain the dissipation rate of
is negative, such that V /Vv = −y 1/2 . If there were no mechanical energy per unit length
KHI induced mass entrainment, then t0 → ∞, τ = 0,
m0 Vv2 y
and eq. (51) represents gravitational free-fall. Ėmech (r) = − . (55)
2t0 x5β/8 µ3/8
Using the relation ṁ = V (r)dm/dr, eqs. (43) and
(50) can be combined to yield an equation for the stream Note that there is no explicit dependence on the
potential, except through the solution to eq. (51) for
y(x). With no KHI induced mass entrainment, t0 → ∞
8 Had we followed the different, less physically motivated,
and Ėmech = 0 as expected.
assumption of Padnos et al. (2018) and Mandelker et al. (2019)
described above, eq. (51) would be replaced by dy/dx = The dissipation rate of thermal energy is governed
2τ y x−5β/8 − 2dφ/dx. by the entrainment rate of background mass onto the

c 0000 RAS, MNRAS 000, 1–16

10 Mandelker et al.
stream, which we have assumed to be locally the same for a 1012 M⊙ halo at z = 2 with concentration c =
as in M20. Therefore, we can use eq. (13) to obtain 10. Cyan, red, and blue curves show solutions for
δ = 30, 100, and 300 respectively. Different line styles
9m0 V02 m0 Vv2 1.6Θh show different values of ns,0.01 and Rsv , obtained from
Ėth (r) = − ≃ − , (56)
10Mb2 tent (r) 2t0 x5β/8 µ3/8 eqs. (24) and (28) using [feh , e
s] = [3, 0.3] (dotted lines),
1/2
with V0 = ηVv and η Mb−1 ∼ 0.9Θh from eq. (19). [3, 3] (solid lines), and [1, 3] (dashed lines) with the
given values of δ and z. The corresponding values of
Generalizing eq. (14), the total radiation emitted Rsv and ns,0.01 are listed in the legend, and are near
between radius r and Rv is the upper and lower bounds of the parameter ranges
Z Rv shown in Fig. 1 for Mv ∼ 1012 M⊙ and z ∼ 2. These
5
Ldiss (> r) = Ėmech (r̃) + Ėth (r̃) dr̃. (57) three models represent a narrow and dense stream, a
r 3
wide and dense stream, and a wide and dilute stream
Using eqs. (15), (39), (42), (33), and (57), the respectively. All models assume Θh = η = 1, and
normalization of the total radiation emitted between the initial stream temperature is determined by δ. We
radius r and Rv is evaluate the cooling time τcool in eq. (42) assuming a
metallicity of Zs = 0.03Z⊙ , though our results are not
m0 Rv Vv2 5/3 −1/2 strongly affected by this choice so long as Zs ∼ < 0.1Z .
≃ 2 × 1040 erg s−1 M12 (1 + z)3 ⊙
2t0 (58) For these parameters, the stream line mass is m0 ∼
−3/2 1/2 −1/4
1010 M⊙ kpc−1 R2sv ns,0.01 (eq. 33). The virial radius and
×Rsv,0.16 ns,0.01 δ100 Θh τcool , velocity are Vv ∼ 200 km s−1 and Rv ∼ 100 kpc. For
with Rsv,0.16 = Rsv /0.16. Together with eqs. (21), (24), typical values of Rsv ∼ 0.16 and ns,0.01 ∼ 1.5 (Fig. 3),
and (28), we obtain that the normalization scales as the total stream mass is thus Mstream ∼ m0 Rv ∼
Mv (1 + z)2 . In practice, the emitted radiation, Ldiss , 3 × 1010 M⊙ . Assuming all the cold gas in the CGM
deviates from this scaling due to the dependence of µ is contained in three such streams, our model predicts
and y in eqs. (55) and (56) on halo mass and redshift. a cold CGM gas mass of Mcold ∼ 1011 M⊙ in halos
of Mv ∼ 1012 M⊙ at redshift z ∼ 2. This is within
We stress that eq. (57) represents the bolometric a factor of < 2 from recent observational estimates
luminosity emitted from a single stream as a result of of the cold gas content in the CGM of halos with
the interaction between the stream and the ambient similar masses and redshifts, which host giant Lyman-α
hot CGM. As a typical halo is fed by ∼ 3 streams nebulae (Pezzulli & Cantalupo 2019).
(e.g. Dekel et al. 2009), the bolometric luminosity
emitted from the halo as a result of stream-CGM The velocity profiles show that in all cases the
interaction will be ∼ 3 times larger. Roughly half of stream accelerates towards the halo centre, with the
this luminosity is expected to be emitted from gas stream velocity only slightly lower than the free-fall
with T ∼ (1 − 5) × 104 K (M20), and ∼ > half of this velocity had there been no mass entrainment or KHI,
will be emitted in Lyα through collisional excitation shown by the solid black line. The deceleration with
(e.g. Katz, Weinberg & Hernquist 1996; Fardal et al. respect to free-fall is greater for smaller values of δ,
2001; Goerdt et al. 2010). We thus expect the total and for narrower and denser streams. However, for
Lyα luminosity emitted from the halo as a result the range of parameters explored here, the stream
of stream-CGM interaction to be comparable to the velocity at r = 0.1Rv is at least ∼ 0.8 times the
estimate given by eq. (57). However, excluded from this free-fall velocity. Overall, the velocity increases by a
estimate is Lyα emission resulting from fluorescence of factor of ∼ (2 − 2.5) from Rv to 0.1Rv . Note that
UV background photons. This can increase the total our current model predicts significantly less deceleration
Lyα luminosity by factors of up to a few, depending than Mandelker et al. (2019). This is not due to the
on the details of the stream density structure and current consideration of radiative cooling effects, but
self-shielding (M20). We ignore this contribution for rather to the different assumption made in deriving our
now, postponing it’s treatment to future work which equation of motion (eq. 51), as discussed in §4.3.
will incorporate realistic self-shielding in simulations. As expected from the velocity profiles, the entrained
We therefore treat eq. (57) as an approximate lower mass is larger for narrower, denser streams, with smaller
limit to the total Lyα emission from halos fed by cold density contrasts. For the range of parameters shown,
the stream mass increases by ∼ < 30% from R to 0.1R
streams. v v
for this halo mass and redshift.
The luminosity profiles show that the emission is
4.5 Example Solutions highest for low values of δ, in line with the expectation
4.5.1 β=2 from eq. (58), where the luminosity increases with
decreasing δ and increasing Rsv and ns,0.01 . Intuitively,
We here present solutions of our toy model for a fiducial the increased emission with stream radius is due to
CGM density slope of β = 2, addressing β dependence the larger stream surface area and resulting increased
below. Figure 3 presents the radial profiles of velocity interaction with background gas, while the increased
(left), line-mass (centre), and luminosity (right) of emission with stream density is due to the decreased
a single stream, resulting from the stream-CGM cooling time at higher densities. The increased emission
interaction according to our model. We show results with lower density contrast stems from the entrainment

c 0000 RAS, MNRAS 000, 1–16

 Lyα Blobs from Cold Streams 11
2.6 1.4
2.4 1.35
42
10
2.2 1.3
2 1.25
1.8 1.2 41
10
1.6 1.15
1.4 1.1
40
10
1.2 1.05
1 1
0.8 0.95 39
10
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Figure 3. Solutions of our toy model for stream evolution, in an NFW halo with virial mass Mv = 1012 M⊙ and concentration c = 10
at redshift z = 2, with a CGM density slope of β = 2. Cyan, red, and blue lines show solutions for δ = 30, 100, and 300 respectively.
Different line styles represent different values of the stream density, ns,0.01 = n0 /(0.01 cm−3 ), and the stream radius normalized by the
virial radius, Rsv = Rs /Rv , as indicated in the legend. For each δ, these correspond to values of the model parameters feh = (1 − 3) and
se = (0.3−3), and are near the upper and lower bounds expected for these parameters for this halo mass and redshift (Fig. 1). Dotted lines
represent dense and narrow streams, solid lines represent dense and wide streams, and dashed lines represent dilute and wide streams.
All models assume η = Θh = 1. In each panel, the x axis is the halocentric radius normalized to the halo virial radius. On the left, we
show the stream velocity normalized to the virial velocity, with the black line representing the free-fall velocity profile if there were no
KHI or mass entrainment, Vff . All cases undergo net acceleration, though the deceleration with respect to free-fall increases with lower
δ, narrower and more dilute streams. However, the stream velocity at 0.1Rv is at least ∼ 0.8Vff for the range of parameters considered
here. In the centre, we show the total line-mass, including the background mass entrained in the stream. Lower values of δ, as well as
narrower and more dilute streams, entrain more background mass. At 0.1Rv , the stream can increase its line-mass by up to ∼ 35%. On
the right, we show the luminosity induced by (mechanic plus thermal) energy dissipation between each radius r and Rv . KHI induced
dissipation can emit ∼ (1041 − 1042 )erg s−1 within the halo, with ∼ 90% of the dissipation occurring in the inner ∼ < 0.6R .
v

2.7 2 1045

2.6 44
1.8 10
2.5
1.6 1043
2.4
1042
2.3 1.4

1041
2.2 1.2

2.1 1040
1
2 1039
1011 1012 1013 1014 1011 1012 1013 1014 1011 1012 1013 1014

Figure 4. Stream properties at 0.1Rv according to our model, as a function of halo mass and redshift. We show the stream velocity
(left), the line-mass (centre), and the total luminosity induced by the stream-CGM interaction (right). Different colours and line-styles
are as in Figs. 1 and 2. The range of model parameters at Mv = 1012 M⊙ and z = 2 is identical to those used in Fig. 3. For different
values of Mv and z, δ, ns,0.01 , and Rsv were then scaled according to eqs. (21), (24), and (28) respectively. Streams in more massive
halos entrain less mass from the CGM compared to their initial mass, and thus decelerate less compared to the free-fall velocity. For
Mv ∼ 1011 M⊙ , the stream mass can potentially double, while for Mv ∼ 1014 M⊙ it increases by ∼ < 5% at most, roughly independent of

redshift. Streams always net-accelerate towards the halo centre, with the velocity at 0.1Rv at least ∼ 70% (90%) of the free fall velocity
for 1011 (1014 )M⊙ halos. The emitted luminosity is a strong function of both halo mass and redshift (eq. 58). Luminosities greater than
1043 erg s−1 are possible in halos with Mv ∼ > 1013 , 5 × 1012 , and 1012 M
⊙ at z = 1, 2, and 4 respectively.

time being shorter for lower values of density contrast is in the range ∼ (1041 − 1042 ) erg s−1 , with 50% (90%)
(eq. 42), and is consistent with more mass entrainment of the luminosity emitted within ∼ 0.3Rv (0.6Rv ).
and deceleration. For the range of parameters shown
here, the total luminosity produced by a single stream in The results for halo concentrations of c = 5
a 1012 M⊙ halo at z = 2 with CGM density slope β = 2 and 20 are extremely similar. The only noticeable
difference is that the free-fall velocity increases with

c 0000 RAS, MNRAS 000, 1–16

12 Mandelker et al.
reached in halos with Mv /M⊙ ∼ > 1013 , 4 × 1012 , and
12
2.8 10 at z = 1, 2, and 4 respectively, while Ldiss >
1042 (1044 ) erg s−1 can be achieved in halos ∼ 5 times
less (more) massive. As discussed at the end of §4.4,
2.6 we expect this to be comparable to the total Lyα
luminosity produced by the stream-CGM interaction,
2.4 and a rough lower limit to the total Lyα luminosity
emitted by all sources in the halo. As described in
§1, observed LABs range from a few times 1042 to
∼ 1044 erg s−1 , and extend on the sky to ∼ > 50 kpc or
2.2
> > 12
more, ∼ 0.5Rv for a ∼ 10 M⊙ halo at z ∼ > 2. Our

model thus suggests that the interaction between cold

2 streams and a hot CGM is certainly a viable source for
low-to-intermediate luminosity LABs, and potentially
for high-luminosity LABs in ∼ > 1013 M halos at z > 3.
1.8 ⊙ ∼
Figure 5 shows the ratio of the mechanical energy
dissipated from Rv to 0.1Rv to the thermal energy
1.6 dissipated in the same interval, for the same parameters
11 12 13 14
10 10 10 10 used in Fig. 4. The ratio is nearly independent of
redshift and depends only weakly on halo mass. For
1011 M⊙ halos, the system dissipates ∼ (1.6 − 2.7) times
Figure 5. Ratio of dissipated mechanical to thermal energy from more mechanical than thermal energy. For 1014 M⊙
Rv to 0.1Rv , as a function of halo mass and redshift. Line styles halos, the lost mechanical energy is ∼ (2.5 − 2.75) times
and colours are the same as in Fig. 4, as are the model parameters greater than the lost thermal energy. In M20, we found
used. Roughly twice as much mechanical energy as thermal energy that at t ≪ tent , ∆Ek /∆Eth ∼ 0.5Mb2 . Substituting
is lost at all redshifts, though the ratio decreases slightly towards ∆Emech for ∆Ek , our current results are consistent with
lower halo masses. this, since most of the energy is dissipated in the inner
halo where the stream velocity is a factor of ∼ 2 larger
than the initial velocity, and thus Mb ∼ 2 (Fig. 3).
halo concentration. However, the deceleration with
respect to free-fall, the mass entrainment, and the
luminosity do not depend strongly on c. 4.5.2 β dependence
Figure 4 shows stream properties at 0.1Rv Figure 6 shows the stream properties at 0.1Rv predicted
predicted by our model, as a function of halo mass and by our model as a function of halo mass for β = 1
redshift. We show the stream velocity normalized by the (green), 2 (blue) and 3 (red). We show only the results
virial velocity (left), the stream line-mass normalized for redshift z = 2 (blue lines in Fig. 4), and use
by the initial line-mass (centre), and the total emitted the same values of c, η, δ, ns,0.01 , and Rsv used in
luminosity (right). At each halo mass and redshift, Fig. 4. Since ns,0.01 normalizes the density at Rv , larger
we explore a range of values for δ, ns,0.01 , and Rsv . values of β imply larger gas densities in the inner halo.
These were normalized for a 1012 M⊙ at z = 2 to the Thus, as β increases, the stream entrains more mass
same values used in Fig. 3, and then scaled to different and decelearates more with respect to free-fall. For
halo masses and redshifts following eqs. (21), (24), and Mv = 1011 M⊙ , the stream line-mass can grow from
(28). We thus solve nine models for each value of Mv Rv to 0.1Rv by factors of up to 1.5 and 4 for β = 1
and z, all of which assume Θh = η = c = 1 and and 3, respectively, while the corresponding velocities at
β = 2 as in Fig. 3, and plot the resulting range of 0.1Rv can be as low as 0.87 and 0.55 times the free-fall
solutions, as in Figs. 1 and 2. The thick lines represent a value. For a 1014 M⊙ halo, the stream mass increases by
model which has (δ, Rsv , ns,0.01 ) = (100, 0.16, 1.50) for < >
∼ 5%, and the velocity is ∼ 95% of the free-fall value
M12 = (1 + z)3 = 1 (solid red lines in Fig. 3). for β = (1 − 3). These results are rather insensitive to
As the halo mass increases, the stream entrains less redshift, as seen in Fig. 4 for β = 2.
mass and its velocity approaches the free-fall velocity. The total luminosity emitted at r > 0.1Rv increases
For Mv = 1011 M⊙ , the stream line-mass can grow by with β, as a result of the increased mass entrainment
up to a factor of 2 from Rv to 0.1Rv , while the velocity and deceleration. While we only show results for z = 2,
at 0.1Rv can be as low as 0.7 times the free-fall value. the scaling with redshift for all β is similar to that
For a 1014 M⊙ halo, the stream mass increases by ∼ < 5%,
>
seen for β = 2 in Fig. 4, and follows from eq. (58).
and the velocity is ∼ 95% of the free-fall value. This is For β = 3, luminosities of Ldiss ∼ > 1043 erg s−1 can be
primarily due to the increase of δ with Mv , resulting in >
reached in halos with Mv /M⊙ ∼ 5 × 1012 , 2 × 1012 ,
a longer entrainment timescale (eq. 42), and is roughly and 5 × 1011 at z = 1, 2, and 4 respectively, while
independent of redshift. Ldiss > 1042 (1044 ) erg s−1 can be achieved in halos
The total luminosity emitted at r > 0.1Rv increases ∼ 5 times less (more) massive. For β = 1, these
with both halo mass and redshift, in accordance with threshold masses increase by a factor of ∼ 4. The radial
eq. (58). Luminosities of Ldiss ∼ > 1043 erg s−1 can be profiles of the emission become steeper with increasing

c 0000 RAS, MNRAS 000, 1–16

 Lyα Blobs from Cold Streams 13
3 45
10
2.6
1044
2.4 2.5
43
10
2.2
2 42
10
2
1041
1.5
1.8
1040
1.6
1
39
10
11 12 13 14 11 12 13 14 11 12 13 14
10 10 10 10 10 10 10 10 10 10 10 10

Figure 6. Similar to Fig. 4, we show as a function of halo mass the stream velocity (left), the line-mass (centre), and the total luminosity
induced by the stream-CGM interaction (right) at 0.1Rv . The redshift is kept fixed at z = 2, while green, blue, and red lines represent
β = 1, 2, and 3 respectively. All other model parameters are as in Fig. 4, as are the meanings of the different linestyles. As β increases,
the inner CGM becomes denser, leading to increased mass entrainment, deceleration, and emission from streams. The trends with halo
mass and redshift are qualitatively similar for all β = (1 − 3).

β, reflecting the steepening of the CGM density profile. Gronke & Oh 2020), it is unclear whether this gas will
As β increases from 1 to 3, the radius containing be strongly self-shielded. We therefore speculate that
50% (90%) of the total emission outside 0.1Rv decreases a significant fraction of the emission will indeed be in
from ∼ 0.4Rv (0.8Rv ) to ∼ 0.2Rv (0.45Rv ). Lyα. We will address this issue in future work, where
we will include self-shielding in our simulations.
Related to the previous point, and as also
5 LIMITATIONS OF THE CURRENT MODEL highlighted at the end of §4.4, we have only computed
a lower limit to the luminosity, produced purely by
It is encouraging that our model, based on the the stream-CGM interaction. As shown by M20 (see
dissipation mechanisms described in detail in M20, can also Goerdt et al. 2010; Faucher-Giguère et al. 2010),
produce spatially extended emission with luminosities fluorescent radiation caused by the UV background
comparable to observed LABs in halos of relevant mass can contribute significantly to the total luminosity. The
and redshift. Our current model presents a significant magnitude of this contribution is unclear and depends
improvement over those presented in Padnos et al. on the level of self-shielding as discussed above, but this
(2018) and Mandelker et al. (2019). Furthermore, does imply that the Lyα emission from cold streams
unlike the analytic models presented in Dijkstra & Loeb may be even larger than predicted here. We will consider
(2009) and Goerdt et al. (2010), our model invokes this along with self-shielding in future work.
a well-defined dissipation mechanism to produce the The radial profiles of the emission predicted by our
radiation. However, it is still very simplified and limited model are in reasonable agreement with that found for
in a number of ways, which we discuss here. the cooling radiation from the CGM of a ∼ 1012 M⊙
Firstly, we have only addressed the bolometric halo at z ∼ 3, found in a cosmological simulation with
luminosity produced by the instability, not specifically full radiative transfer (Trebitsch et al. 2016). However,
the Lyα emissivity. As highlighted at the end of §2 as described in that paper, the radial profiles of the
and §4.4, the simulations presented in M20 found that observed emission may be much flatter due to scattering
roughly half of the radiation is emitted from gas with of the radiation from the inner halo to larger radii,
temperatures T ∼ (1.5 − 5) × 104 K, and is thus likely which also leads to polarized emission as observed
to be dominated by Lyα. When accounting for the fact (Trebitsch et al. 2016). Therefore, we cannot address
that halos typically have ∼ 3 streams, we estimated that the observed emission profiles until accounting for
the single-stream luminosities computed in our model radiative transfer effects, which we defer to future work.
are comparable to the total Lyα luminosity produced In §4.2, we made a number
by stream-CGM interactions in the halo. However, the of simplifying assumptions when evaluating the scaling
simulations of M20 did not include self-shielding from of stream properties with halocentric radius. First,
the UV background. While this is unlikely to alter the the assumption that the hot CGM is isothermal
total dissipated energy or bolometric luminosity, it may with a power-law density profile is clearly an
alter the temperature distribution of the emitting gas, oversimplification. This was motivated by cosmological
and thus the Lyα contribution (Faucher-Giguère et al. and isolated simulations and low-z observations
2010). However, since most of the emission comes (van de Voort & Schaye 2012; Fielding et al. 2017;
from the turbulent mixing zone surrounding the stream Singh et al. 2018), and makes the model analytically
rather than from the stream interior (M20, see also tractable. As the properties of the high-z CGM are not

c 0000 RAS, MNRAS 000, 1–16

14 Mandelker et al.
observationally constrained, we are confident that the paper (Mandelker et al. 2020), we showed how the
range of profiles we considered likely brackets realistic interaction between these cold streams and the hot
values for the resulting luminosities. However, when CGM leads to energy dissipation and radiation through
better constraints for the high-z CGM are available, the combination of Kelvin-Helmholtz instability (KHI)
the model should be refined to reflect these. and radiative cooling. Here, we expanded upon these
Furthermore, we assumed the streams to be results by modeling the effect of the dark-matter halo
on purely radial orbits towards the halo centres. potential with a toy model. In our model, the dark
Cosmological simulations suggest that streams have matter halo has three main effects on stream evolution:
<
typical impact parameters of ∼ 0.3Rv , which plays a (i) It accelerates the stream towards the halo centre,
large role in the growth of angular momentum in disc which counteracts the KHI induced deceleration, and
galaxies (e.g. Danovich et al. 2015). It has also been shortens the amount of time a given stream element
suggested that this impact parameter can make the spends flowing towards the central galaxy.
streams unstable to Rayleigh-Taylor instabilities in the (ii) It induces a density profile in both the stream and
inner halo (Kereš & Hernquist 2009). This is likely to the CGM gas, such that the gas density and emissivity
affect our results near the inner halo, which should both grow towards the halo centre.
be addressed in future work incorporating non-radial (iii) It focuses the stream towards the halo centre,
orbits. However, it seems likely that this will increase reducing the surface area of the stream, and hence the
the cooling emission, particularly near pericentre where total emitted radiation, closer to the halo centre.
streams are likely to interact and lead to strong shocks
(e.g. Mandelker et al. 2018). Another complication in Combining this model with cosmologically
the inner halo is the interaction of cold streams with motivated boundary conditions for the values of stream
outflowing gas from the galaxies. This changes the parameters at Rv as a function of halo mass and
density and pressure of the background gas in the inner redshift, we estimated the luminosity emitted by a
halo, affecting the KHI, and is also likely to lead to typical cold stream as a result of KHI, as it penetrates
strong shocks in the streams, which may disrupt the the halo towards the central galaxy. We found both the
streams in the inner halo altogether. These shocks are magnitude and spatial extent of this to be comparable
likely to increase the cooling radiation in the inner to those of observed Lyα blobs (LABs), namely Ldiss >
halo, though they are unlikely to affect our results at 1042 erg s−1 emitted within ∼ < 0.6R
v of halos with
r∼ > (0.2 − 0.3)R . Modeling the detailed interaction > 2. This supports previous
v Mv > 1012 M⊙ at z ∼
between inflows and outflows is beyond the scope of claims that LABs may constitute direct observational
this paper, and is left for future work. evidence of cold streams (Dijkstra & Loeb 2009;
Finally, while our analytic toy model has allowed Goerdt et al. 2010), and expands upon those works
us to gain insight into the evolution of cold streams by identifying, for the first time, a self-consistent
in dark matter halos, there may be additional physical dissipation mechanism powering the emission. We
effects for which we did not account, such as the tidal predict that the LAB luminosity increases with halo
field of the halo, which may alter the evolution of mass slightly super-linearly (Fig. 4). This is similar to
KHI. This may invalidate the main assumption, which the scaling predicted by the model of Dijkstra & Loeb
was that the KHI induced mass entrainment derived (2009), which was found to be consistent with observed
by M20 could be applied locally at each halocentric LAB luminosity functions and clustering. While our
radius. Furthermore, M20 did not account for additional model makes a number of simplifying assumptions,
physics, such as self-gravity, magnetic fields, thermal it is encouraging that the dissipation induced by
conduction, or cosmic rays9, or for initially non-linear KHI seems to produce emission in the right ball-park
perturbations in the streams caused by, e.g., subhalos or for explaining LABs, both in terms of luminosity
mergers along the streams. All of these will be studied and spatial extent. Future work should continue to
in detail in future work both analytically and using explore these issues using more refined analytic models
idealized simulations, as outlined in M20. Nevertheless, and idealized simulations that account for additional
we hope that the model presented here will serve as a physics.
useful benchmark against which future simulations that
include an explicit halo potential can be compared.
ACKNOWLEDGMENTS
We thank the
6 SUMMARY AND CONCLUSIONS referee, Andreas Burkert, for his thoughtful comments
Massive halos of Mv ∼ > 1012 M > and constructive report which helped improve the
⊙ at redshift z ∼ 2
are understood to be fed by cold streams with quality and clarity of this manuscript. We thank Nicolas
> 104 K. These streams flow along
temperatures Ts ∼ Cornuault, Drummond Fielding, Max Gronke, Joseph
cosmic web filaments and penetrate the hot CGM F. Hennawi, Suoqing Ji, Neal Katz, S. Peng Oh,
of these halos, with Th ∼>
106 K. In our previous X. Prochaska, Santi Roca-Fabrega, and Chuck Steidel
for helpful discussions. NM and FCvdB acknowledge
support from the Klauss Tschira Foundation through
9 See section 6 in M20 for a discussion of the potential influence the HITS Yale Program in Astropysics (HYPA).
of these effects on their results FCvdB received additional support from the National

c 0000 RAS, MNRAS 000, 1–16

 Lyα Blobs from Cold Streams 15
Aeronautics and Space Administration through grant Fardal M. A., Katz N., Gardner J. P., Hernquist L.,
No. 17-ATP17-0028Grant Nos. 17-ATP17-0028 and Weinberg D. H., Davé R., 2001, ApJ, 562, 605
19-ATP19-0059 issued as part of the Astrophysics Faucher-Giguère C.-A., Kereš D., Dijkstra M.,
Theory Program. DN acknowledges support by Hernquist L., Zaldarriaga M., 2010, ApJ, 725, 633
National Science Foundation grant AST-1412768 and Faucher-Giguère C.-A., Kereš D., Ma C.-P., 2011,
the hospitality at the Aspen Center for Physics, MNRAS, 417, 2982
which is supported by National Science Foundation Fielding D., Quataert E., McCourt M., Thompson
grant PHY-1607611. AD is partly supported by T. A., 2017, MNRAS, 466, 3810
the grants NSF AST-1405962, BSF 2014-273, GIF Fielding D. B., Ostriker E. C., Bryan G. L., Jermyn
I-1341-303.7/2016, and DIP STE1869/2-1 GE625/17-1. A. S., 2020, ApJ, 894, L24
YB acknowledges ISF grant 1059/14. This work is Fumagalli M. et al., 2017, MNRAS, 471, 3686
supported in part by the facilities and staff of the Yale Fumagalli M., Prochaska J. X., Kasen D., Dekel A.,
Center for Research Computing. Ceverino D., Primack J. R., 2011, MNRAS, 418, 1796
Furlanetto S. R., Schaye J., Springel V., Hernquist L.,
2005, ApJ, 622, 7
Geach J. E. et al., 2009, ApJ, 700, 1
DATA AVAILABILITY Geach J. E., Smail I., Chapman S. C., Alexand er
D. M., Blain A. W., Stott J. P., Ivison R. J., 2007,
The data underlying this article will be shared on
ApJ, 655, L9
reasonable request to the corresponding author.
Goerdt T., Ceverino D., 2015, MNRAS, 450, 3359
Goerdt T., Dekel A., Sternberg A., Ceverino D.,
Teyssier R., Primack J. R., 2010, MNRAS, 407, 613
REFERENCES
Goerdt T., Dekel A., Sternberg A., Gnat O., Ceverino
D., 2012, MNRAS, 424, 2292
Arrigoni Battaia F., Hennawi J. F., Prochaska J. X., Gronke M., Oh S. P., 2018, MNRAS, 480, L111
Oñorbe J., Farina E. P., Cantalupo S., Lusso E., 2019, Gronke M., Oh S. P., 2020, MNRAS, 492, 1970
MNRAS, 482, 3162 Haardt F., Madau P., 1996, ApJ, 461, 20
Arrigoni Battaia F., Prochaska J. X., Hennawi J. F., Haiman Z., Rees M. J., 2001, ApJ, 556, 87
Obreja A., Buck T., Cantalupo S., Dutton A. A., Hennawi J. F., Prochaska J. X., Cantalupo S.,
Macciò A. V., 2018, MNRAS, 473, 3907 Arrigoni-Battaia F., 2015, Science, 348, 779
Aung H., Mandelker N., Nagai D., Dekel A., Birnboim Hobbs A., Feldmann R., 2020, arXiv e-prints,
Y., 2019, MNRAS, 490, 181 arXiv:2001.06012
Begelman M. C., Fabian A. C., 1990, MNRAS, 244, Ji S., Oh S. P., Masterson P., 2019, MNRAS, 487, 737
26P Katz N., Weinberg D. H., Hernquist L., 1996, ApJS,
Berlok T., Pfrommer C., 2019, arXiv e-prints, 105, 19
arXiv:1904.02167 Kereš D., Katz N., Weinberg D. H., Davé R., 2005,
Birnboim Y., Dekel A., 2003, MNRAS, 345, 349 MNRAS, 363, 2
Birnboim Y., Padnos D., Zinger E., 2016, ApJ, 832, Kereš D., Hernquist L., 2009, ApJ, 700, L1
L4 Komatsu E., Seljak U., 2001, MNRAS, 327, 1353
Borisova E. et al., 2016, ApJ, 831, 39 Laursen P., Sommer-Larsen J., 2007, ApJ, 657, L69
Bouché N. et al., 2016, ApJ, 820, 121 Leclercq F. et al., 2017, A&A, 608, A8
Bouché N., Murphy M. T., Kacprzak G. G., Péroux Li Z., Hopkins P. F., Squire J., Hummels C., 2020,
C., Contini T., Martin C. L., Dessauges-Zavadsky M., MNRAS, 492, 1841
2013, Science, 341, 50 Mandelker N., Nagai D., Aung H., Dekel A., Birnboim
Cantalupo S., Arrigoni-Battaia F., Prochaska J. X., Y., van den Bosch F. C., 2020, MNRAS, 494, 2641
Hennawi J. F., Madau P., 2014, Nature, 506, 63 Mandelker N., Nagai D., Aung H., Dekel A., Padnos
Ceverino D., Dekel A., Bournaud F., 2010, MNRAS, D., Birnboim Y., 2019, MNRAS, 484, 1100
404, 2151 Mandelker N., Padnos D., Dekel A., Birnboim Y.,
Danovich M., Dekel A., Hahn O., Ceverino D., Primack Burkert A., Krumholz M. R., Steinberg E., 2016,
J., 2015, MNRAS, 449, 2087 MNRAS, 463, 3921
Danovich M., Dekel A., Hahn O., Teyssier R., 2012, Mandelker N., van Dokkum P. G., Brodie J. P., van
MNRAS, 422, 1732 den Bosch F. C., Ceverino D., 2018, ApJ, 861, 148
Dekel A., Birnboim Y., 2006, MNRAS, 368, 2 Martin D. C., Chang D., Matuszewski M., Morrissey
Dekel A. et al., 2009, Nature, 457, 451 P., Rahman S., Moore A., Steidel C. C., 2014a, ApJ,
Dekel A., Zolotov A., Tweed D., Cacciato M., Ceverino 786, 106
D., Primack J. R., 2013, MNRAS, 435, 999 Martin D. C., Chang D., Matuszewski M., Morrissey
Dijkstra M., Haiman Z., Spaans M., 2006, ApJ, 649, P., Rahman S., Moore A., Steidel C. C., Matsuda Y.,
14 2014b, ApJ, 786, 107
Dijkstra M., Loeb A., 2009, MNRAS, 400, 1109 Martin D. C., Matuszewski M., Morrissey P., Neill
Dimotakis P. E., 1991, Turbulent free shear layer J. D., Moore A., Steidel C. C., Trainor R., 2016, ApJ,
mixing and combustion. Tech. rep. 824, L5
Fabian A. C., 1994, ARA&A, 32, 277 Martin D. C. et al., 2019, Nature Astronomy, 372

c 0000 RAS, MNRAS 000, 1–16

16 Mandelker et al.
Matsuda Y., Iono D., Ohta K., Yamada T., Kawabe
R., Hayashino T., Peck A. B., Petitpas G. R., 2007,
ApJ, 667, 667
Matsuda Y. et al., 2004, AJ, 128, 569
Matsuda Y., Yamada T., Hayashino T., Yamauchi R.,
Nakamura Y., 2006, ApJ, 640, L123
Matsuda Y. et al., 2011, MNRAS, 410, L13
Mori M., Umemura M., Ferrara A., 2004, ApJ, 613,
L97
Nelson D., Genel S., Pillepich A., Vogelsberger M.,
Springel V., Hernquist L., 2016, MNRAS, 460, 2881
Nelson D., Vogelsberger M., Genel S., Sijacki D., Kereš
D., Springel V., Hernquist L., 2013, MNRAS, 429,
3353
Nilsson K. K., Fynbo J. P. U., Møller P.,
Sommer-Larsen J., Ledoux C., 2006, A&A, 452, L23
Ocvirk P., Pichon C., Teyssier R., 2008, MNRAS, 390,
1326
Ohyama Y. et al., 2003, ApJ, 591, L9
Padnos D., Mandelker N., Birnboim Y., Dekel A.,
Krumholz M. R., Steinberg E., 2018, ArXiv e-prints
Palunas P., Teplitz H. I., Francis P. J., Williger G. M.,
Woodgate B. E., 2004, ApJ, 602, 545
Pezzulli G., Cantalupo S., 2019, MNRAS, 486, 1489
Prescott M. K. M., Kashikawa N., Dey A., Matsuda
Y., 2008, ApJ, 678, L77
Prochaska J. X., Lau M. W., Hennawi J. F., 2014, ApJ,
796, 140
Rauch M. et al., 2008, ApJ, 681, 856
Roca-Fàbrega S. et al., 2019, MNRAS, 484, 3625
Saito T., Shimasaku K., Okamura S., Ouchi M.,
Akiyama M., Yoshida M., 2006, ApJ, 648, 54
Singh P., Majumdar S., Nath B. B., Silk J., 2018,
MNRAS, 478, 2909
Smith D. J. B., Jarvis M. J., 2007, MNRAS, 378, L49
Springel V., 2010, MNRAS, 401, 791
Steidel C. C., Adelberger K. L., Shapley A. E., Pettini
M., Dickinson M., Giavalisco M., 2000, ApJ, 532, 170
Steidel C. C., Bogosavljević M., Shapley A. E.,
Kollmeier J. A., Reddy N. A., Erb D. K., Pettini M.,
2011, ApJ, 736, 160
Trebitsch M., Verhamme A., Blaizot J., Rosdahl J.,
2016, A&A, 593, A122
van de Voort F., Schaye J., 2012, MNRAS, 423, 2991
van de Voort F., Schaye J., Altay G., Theuns T., 2012,
MNRAS, 421, 2809
van de Voort F., Schaye J., Booth C. M., Haas M. R.,
Dalla Vecchia C., 2011, MNRAS, 414, 2458
Vogelsberger M., Sijacki D., Kereš D., Springel V.,
Hernquist L., 2012, MNRAS, 425, 3024
Weidinger M., Møller P., Fynbo J. P. U., 2004, Nature,
430, 999
Weidinger M., Møller P., Fynbo J. P. U., Thomsen B.,
2005, A&A, 436, 825
Wilman R. J., Gerssen J., Bower R. G., Morris S. L.,
Bacon R., de Zeeuw P. T., Davies R. L., 2005, Nature,
436, 227
Yang Y., Zabludoff A., Eisenstein D., Davé R., 2010,
ApJ, 719, 1654
Yang Y., Zabludoff A., Tremonti C., Eisenstein D.,
Davé R., 2009, ApJ, 693, 1579

c 0000 RAS, MNRAS 000, 1–16