Mon. Not. R. Astron. Soc. 400, 1109–1120 (2009) doi:10.1111/j.1365-2966.2009.15533.x
Lyα blobs as an observational signature of cold accretion streams into
galaxies
Mark Dijkstra and Abraham Loeb
Astronomy Department, Harvard University, 60 Garden Street, Cambridge, MA 02138, USA
Accepted 2009 August 11. Received 2009 August 11; in original form 2009 April 15
ABSTRACT
Recent hydrodynamic simulations of galaxy formation reveal streams of cold (T ∼ 104 K)
gas flowing into the centres of dark matter haloes as massive as 1012−13.5 M

at redshifts z ∼
1–3. In this paper, we show that if 20 per cent of the gravitational binding energy of the
gas is radiated away then the simulated cold flows are spatially extended Lyα sources with
luminosities, Lyα linewidths and number densities that are comparable to those of observed
Lyα blobs. Furthermore, the filamentary structure of the cold flows can explain the wide
range of observed Lyα blob morphologies. Since the most massive haloes form in dense
environments, the association of Lyα blobs with overdense regions arise naturally. We argue
that Lyα blobs – even those which are clearly associated with starburst galaxies or quasars –
provide direct observational support for the cold accretion mode of galaxies. We discuss
various testable predictions of this association.
Key words: line: formation – cooling flows – galaxies: formation – galaxies: haloes – inter-
galactic medium – cosmology: theory.
1 INTRODUCTION
The physical origin of spatially extended Lyα sources, also known
as Lyα blobs (LABs), is still enatic. LABs have been associated
with cooling radiation, in which gas that collapses inside a host
dark matter halo releases a significant fraction of its gravitational
binding energy in Lyα line emission (Haiman, Spaans & Quataert
2000; Fardal et al. 2001; Birnboim & Dekel 2003; Dijkstra, Haiman
& Spaans 2006, also see Katz & Gunn 1991). Other mechanisms
that have been invoked to explain the origin of LABs include the
following: photoionization of cold (T ∼ 104 K), dense, spatially ex-
tended gas by obscured quasars (Haiman & Rees 2001; Geach et al.
2009), population III stars (Jimenez & Haiman 2006) or spatially
extended inverse Compton X-ray emission (Scharf et al. 2003); the
compression of ambient gas by superwinds into dense, cold Lyα
emitting shells (e.g. Taniguchi & Shioya 2000; Mori, Umemura &
Ferrara 2004); star formation that is triggered by relativistic jets
(Rees 1989) or some combination of photoionization and cooling
(Furlanetto et al. 2005).
Since their discovery by Steidel et al. (2000; also see Keel et al.
1999), several tens of new LABs have been found (e.g. Matsuda
et al. 2004; Dey et al. 2005; Saito et al. 2006; Smith et al. 2009).
These are typically associated with massive haloes that reside in
dense parts of the Universe (Steidel et al. 2000; Matsuda et al.
2004, 2006). Multiwavelength studies of LABs have revealed a
clear association of the brighter blobs with regular Lyman break
E-mail: mdijkstr@cfa.harvard.edu
galaxies (LBGs; e.g. Matsuda et al. 2004), sub-millimeter (sub-
mm) and infrared (IR) sources which imply star formation rates
of ∼103 M

yr−1 (Chapman et al. 2001; Geach et al. 2005, 2007;
Matsuda et al. 2007), or with unobscured (Bunker et al. 2003;
Weidinger, Møller & Fynbo 2004) and obscured quasars
(Basu-Zych & Scharf 2004; Geach et al. 2007; Smith et al. 2009).
However, in other blobs this association has been ruled out, which
has led to the conclusion that cooling radiation by cold accreting
gas may have been observed (Matsuda et al. 2006; Nilsson et al.
2006; Smith & Jarvis 2007; Saito et al. 2008; Smith et al. 2008).
We propose that the sources that are associated with the LABs
may have little to do in powering the spatially extended Lyα emis-
sion. This proposal is physically motivated by recent hydrodynam-
ical simulations of galaxy formation, which show that baryons as-
semble into galaxies through a two-phase medium which contains
filamentary streams of cold (T ∼ 104 K) gas embedded within
a hot gaseous halo (e.g. Keresˇ et al. 2005; Dekel & Birnboim
2006; Agertz, Teyssier & Moore 2009; Dekel et al. 2009; Keresˇ
et al. 2009; Ocvirk, Pichon & Teyssier 2008). The cold flows con-
tain ∼5–50 per cent of the total gas content (Birnboim, Dekel &
Neistein 2007; Keresˇ et al. 2009) in haloes with masses in the range
Mhalo ∼ 1012–1013.5 M

(Dekel et al. 2009). As we will show in
this paper, these cold streams are probable sources of spatially ex-
tended Lyα emission. Under very reasonable assumptions, we find
that the cold accretion model ‘predicts’ the existence of spatially
extended Lyα sources with properties reminiscent of the observed
LABs around massive haloes (Mhalo  1012 M

).
The outline of this paper is as follows: we describe our model in
Section 2. In Section 3, we present our results, before discussing
C© 2009 The Authors. Journal compilation C© 2009 RAS

1110 M. Dijkstra and A. Loeb
our work in Section 4 and presenting our conclusions in Section 5.
The cosmological parameter values used throughout our discussion
are (m, , b, h, σ 8) = (0.27, 0.73, 0.046, 0.70, 0.82) (Komatsu
et al. 2009), and we denote the primordial helium abundance by
mass as Y He = 0.24 (e.g. Izotov, Thuan & Lipovetsky 1997).
2 THE MODEL
Our model is based on the scenario for cold accretion that has
emerged from recent simulations (Keresˇ et al. 2009). ‘Hot’ gas that
is in hydrostatic equilibrium with the dark matter potential well at
the virial temperature of the dark matter halo is in pressure equi-
librium with the cold gas (also see Fall & Rees 1985; Rees 1989),
which makes up a fraction f cold of the total gas mass of the halo,
Mgas. We further assume that the gas mass fraction, f gas, is half of
the universal value. Hence, the total gas mass Mgas ≡ f gasMhalo =
0.5b/MMhalo = 0.08Mhalo. This accounts for a substantial frac-
tion of baryons that may be locked up in (dwarf) galaxies that reside
within the halo of interest. The assumed gas mass fraction f gas =
0.08 is in good agreement with the value derived for groups of galax-
ies at z < 1 (e.g. Giodini et al. 2009). As the cold streams navigate
to the centre of their host dark matter halo, they are progressively
heated by the release of a fraction of their gravitational potential
energy through weak shocks (Rees & Ostriker 1977; Haiman &
Rees 2001). This heating of the cold gas is balanced by radiative
cooling , mostly through the Lyα emission line. Next, we describe
the model in more detail.
(i) We use the gas density profile for hot gas that is in hydrostatic
equilibrium with an NFW (Navarro, Frenk & White 1997) dark
matter potential well with a concentration parameter c = 3.8 (Gao
et al. 2008) at the virial temperature of the halo as derived by Makino
et al. (1998, their equations 8 and 11). This profile is very similar
to an isothermal β-model, which – according to the simulations –
provides an accurate description of the hot gas component (Keresˇ
et al. 2009). In Section 4.6, we show that our final results are robust
against variations in the assumed density profile.
(ii) The hot gas reaches the virial temperature, T h(r) = T vir =
1.9 × 106 K (Mhalo/1012 M

)2/3 [(1 + z)/4], and co-exists in pres-
sure equilibrium with the cold gas. Pressure equilibrium between
the hot (h) and cold (c) gas implies that nc(r)T c(r) = nh(r)T h(r), in
which n(r) and T(r) denote the number density of particles and the
gas temperature, respectively. We obtain T c(r) under the assump-
tion that cooling balances the heating of the cold flows that occurs
as they navigate into the centre of the dark matter halo.
We parametrize the gravitational heating rate by assuming that
a fraction f grav(f hot) of the change in the gravitational potential
energy of each gas element along its trajectory is converted into
heat in the cold (hot) gas. The remaining fraction (1 − f grav −
f hot) is converted into additional kinetic energy of the gas element.
Throughout the paper, we assume that the transfer of energy into
the hot gas is negligible, i.e. f hot = 0. This assumption appears
reasonable given that the majority of the cold gas mass is in smooth
continuous streams (Keresˇ et al. 2009; Dekel et al. 2009), as opposed
to discrete clouds which are likely to heat the hot gas (e.g. Dekel
& Birnboim 2008, and references therein). Under this assumption,
which is discussed in more detail in Section 4.6.2, f grav = 1 corre-
sponds to infall at a constant velocity, while f grav = 0 corresponds
to free-fall.
For the sake of simplicity, we adopt the conservative working
assumption of a constant non-zero f grav inside the virial radius (rvir)
of galaxy haloes and f grav = 0 outside. The most recent smoothed
particle hydrodynamical (SPH) simulations (Keresˇ et al. 2009) in-
dicate that the cold flows propagate inwards at an approximately
constant speed, implying f grav ∼ 1; however, Adaptive Mesh Re-
finement (AMR) simulations (Ocvirk et al. 2008; Dekel et al. 2009)
indicate that the cold gas accelerates throughout its motion and that
therefore f grav may be smaller.1 Future simulations might be able to
refine our working assumption by resolving the precise dynamics
and heating of the cold flows. Hence, the heating rate per particle is
given by
H (r) = fgrav × GM(< r)μmp
r2
v(r) + Hγ (r), (1)
where μ is the mean molecular weight per particle in the cold flow
(in units of mp) and v(r) denotes the infall velocity which is given
as
v2(r) = 2v2(rvir) + 2(1 − fgrav − fhot)
∫ r
rvir
ds
GM(< s)
s2
, (2)
and we assume that v(rvir) =
√
2vcirc (with our results not being
sensitive to this choice). The term Hγ (r) denotes the heating rate
due to absorption of ionizing radiation. At the typical densities in
the cold flows (nc  0.1 cm−3), the gas is self-shielded from external
ionizing radiation, and hence Hγ (r) = 0. It is possible, however,
that galaxies embedded within the cold flow may photoionize some
surrounding region which would locally boost H(r) (Furlanetto et al.
2005). In any case, ignoring this extra photoheating term only makes
our predicted Lyα luminosities from the cold flows smaller, and
therefore makes our results more conservative. As was mentioned
above, we assume that f hot = 0.
We obtain an equilibrium temperature T c(r) at radius r by equat-
ing H(r) to the cooling rate per particle in the cold flow, which is
given by
(r, T , xH I) = 1
nc
[nenH ICcool(Tc) + nenH IIRcool(Tc)] . (3)
Here, the first term in the square brackets denotes the cooling rate (in
erg s−1 cm−3) due to collisional excitation of H atoms by electrons.
The second term denotes the cooling rate due to recombination
events of free electrons and protons (other cooling processes are
negligibly small for the typical gas temperature in the cold flow).
The rate coefficients Ccool(T c) and Rcool(T c) were taken from Hui &
Gnedin (1997, hereafter HG97). Finally, the ionization state of the
gas determines its cooling rate. Under the assumption that the gas
is self-shielding (which is justified later), we obtain a one-to-one
relation between T and xH I through xH I = αrec,B(T c)/[αrec,B(T c) +
C ion(T c)]. Here, C ion(T c) denotes the collisional ionization rate
coefficient and αrec,B(T c) denotes the case B recombination coeffi-
cient (as the cold neutral gas is optically thick in all Lyman-series
lines and case B applies; the related coefficients were taken from
HG97). At the temperatures of interest, helium is neutral inside
the cold gas and free electrons are only supplied by hydrogen, i.e.
nH I(r) = xH I(r)nc(r)/[1 + (Y He/4)] and nH II(r) = ne(r) = [1 −
xH I(r)]nc(r)/[1 + (Y He/4)].
In practice, we assume that T c = 104 K and compute nc assuming
pressure equilibrium. Since temperature determines the ionization
state of the gas in self-shielded gas, we obtain a cooling rate which
1 Our discussion focuses on the extended Lyα emission, and not on the
compact core of the galaxy where the cold streams are finally brought to
rest. Note that dust may suppress the Lyα luminosity from the core, but is
less likely to affect the cold streams which carry metal-poor intergalactic
gas.
C© 2009 The Authors. Journal compilation C© 2009 RAS, MNRAS 400, 1109–1120