The Dark Ages of the

Universe
Simon Glover
 Lecture 2:
Recombination
• At high z, Tgas is large and the Universe is highly ionized

• As the Universe expands, Tgas falls

• Consequently, we eventually reach a point at which

recombination wins out over ionization

• Once we reach this point, the Universe quickly undergoes

a transition from mostly ionized to mostly neutral

• This is the epoch of recombination. In this lecture, we

will look in detail at how this takes place and what the
consequences are for the Universe at lower redshifts
• Suppose the gas is in thermodynamic equilibrium and
that we can neglect any interaction with the radiation
other than Compton scattering

• If the chemical timescale is much smaller than the Hubble

timescale, then density, temperature barely change in
time it takes for chemistry to reach equilibrium

• At any given redshift, can nd equilibrium state by

assuming constant density, pressure and solving for
chemical composition that minimizes:

Here, U is the internal energy and S is the entropy of the

system
fi
• This quantity is known as the Helmholtz free energy

• Suppose our system has Ne free electrons, Np protons

and NH hydrogen atoms. Then, we can write the
Helmholtz free energy as

• Zc is the canonical partition function

• Now take the natural logarithm of Zc and use Stirling’s
approximation:

• We obtain:

• Since Ne = Np and NH = NB - Ne, where NB is the baryon

number, we can eliminate Np and NH and write this as:
 • Di erentiating with respect to Ne yields:

• It follows that F will have a minimum if:

• Therefore, in equilibrium at constant density and

temperature, we have:
ff
• If we now substitute in the values of the various partition
functions (see any good stat. physics textbook), we
obtain the Saha equation:

• Here, is the ionization potential of hydrogen, nB is the

number density of baryons and x = ne/nB is the fractional
ionization
𝛘
Glover et al (2014)
 • In reality, recombination proceeds more slowly than
suggested by the Saha equation.

• Saha eq. assumes that the radiation eld has no impact,

merely acts as a heat bath

• At rst sight, this seems like a reasonable assumption: at

z ~ 1500, the CMB temperature is ~4000 K and so the
number density of photons with E > 13.6 eV is very small
fi
fi
• If hydrogen recombines directly to the ground state, it is
unlikely to be ionized again by the CMB

• BUT: recombination to the ground state produces a

photon with energy > 13.6 eV

• At z ~ 1500, xH ~ 0.03 and hence nH ~ 20 cm-3

• Mean free path of these photons is 1/(σ0 nH) ~ 0.003 pc

• Photons produced by ground-state recombination are

absorbed “on the spot”

• Ground-state recombination leads to no net change in

fractional ionization
 Energy (Rydberg)

Atkins (2005)
• Recombination into energy levels with n > 1 does not
produce ionizing photons

• However, if H atom remains in excited state, much easier

for CMB to ionize it

• E.g. if n=2, need only 3.4 eV photon, rather than 13.6 eV

n=2 n=1
• Successful recombination only possible if atom leaves
excited state before being ionized by CMB

• BUT: what happens to photons produced by radiative

decay to the ground state?

• Consider e.g. simpli ed model of H atom that includes

only 1s, 2s, 2p levels plus continuum

• Decay from 2p to 1s produces a 10.2 eV Lyman-α photon

• Absorption cross-section for these photons is very large,

so most are re-absorbed
fi
• How do we get from n=2 to n=1 without exciting another
H atom from n=1 to n=2?

• Two possibilities:

- Two photon decay from 2s to 1s (forbidden, slow)

- Redshifting of Lyman-α photons

• If Lyman-α photon emitted far in red wing of line, cosmic

expansion may cause it to lose enough energy to leave
the line before it can be reabsorbed
• Overall e ect is to slow down recombination

• Full details of how we can calculate the slow down are

given in the lecture notes

• By comparing redshifting rate, two-photon rate, can

determine which process dominates

• Find that initially, when x ~ 1, redshifting of photons out of

the line dominates, but that two-photon decay quickly
dominates once x drops
ff
Glover et al (2014)
• Delayed recombination explains di erence from Saha eq.
prediction around z ~ 1000

• At lower redshifts, another issue becomes important

• Recombination timescale: trec = 1 / (krec ne) ~ x(z)-1 (1+z)-3

• Expansion timescale: tH = 1/H ~ (1+z)-3/2

• Initially, recombination much faster than expansion. However,

as x and z decrease, eventually reach a point where tH ~ trec

• At lower redshifts, expansion is faster than recombination.

Fractional ionization freezes out, undergoes little further
decrease with redshift
ff
• Value of fractional ionization at low redshift depends on
value of freeze-out redshift

• Can easily show that this leads to a value:

• For the standard ΛCDM model, this gives x ~ 0.0002

• Universe is mostly, but not entirely, neutral following
recombination

• Residual ionization plays an important role in the

chemistry of the early Universe

• Given ionization history, can then solve for electron

scattering optical depth:
• Optical depth is very small at z << 1000, but increases
rapidly as we look back to larger redshifts

• Any particular photon will have some redshift zls at which

it last underwent scattering

• Probability distribution of these redshifts is known as the

visibility function:
Fractional ionization

Visibility function

Chluba & Sunyaev (2006)

• Visibility function approximately Gaussian, centred at a
redshift z ~ 1100, σz ~ 80

• Most CMB photons we observe today come from this

narrow range of redshifts

• Known as the last-scattering surface (even though it is a

volume, not a surface!)

• CMB observations primarily probe the Universe when

0.005 < x < 0.2 — Universe prior to recombination is
hidden
 CMB spectral distortion

• Photons produced during recombination distort the

spectral shape of the CMB

• Current CMB energy density ⍴CMB ~ 0.25 eV cm-3

• Scales as (1+z)4, so at z ~ 1000, ⍴CMB ~ 2.5 x 1011 eV cm-3

• Hydrogen number density at same redshift ~ 200 cm-3

• Recombination releases 13.6 eV per H atom

• Therefore expect fractional distortion of around ~ 10-8

• This seems absolutely tiny — for reference, CMB limit on
distortion near peak of CMB spectrum is ~ 10-5

• However, energy release concentrated at speci c

frequencies

• Photons escaping from e.g. Lyman-α line are far in tail of

CMB spectrum where there are few CMB photons

• Distortion spectrum is therefore potentially measurable by

future CMB experiments (e.g. PIXIE, PRISM and their
successors)
fi
Glover et al (2014), based on Chluba & Sunyaev (2006)
Glover et al (2014), based on Chluba & Sunyaev (2006)
 Pre-galactic chemistry

• Prior to recombination, Universe is highly ionized, hence

chemical state is very simple — no atoms or molecules

• After recombination, chemical complexity increases

• Despite small number of elements available, number of

molecules, molecular ions etc. that can form is large

• For example, Glover & Savin (2009) construct a model with

~400 reactions between 30 species

• Fortunately, most of this is irrelevant for our purposes, and

we can focus on much smaller subset
 • Why do we care about primordial chemistry at all?

• Simple answer: radiative cooling!

• Permitted atomic transitions (e.g. Lyman-α) only provide

e ective cooling down to temperatures T ~ 104 K

• In local ISM, lower temperature cooling provided by

forbidden lines of heavy elements (e.g. C, O, Si), by
rotational/vibrational lines of heavy molecules (e.g. CO)
and by dust
ff
Solar metallicity cooling curves

x=

Dalgarno & McCray (1972)

• None of these low temperature coolants present in metal-
free gas

• To cool below 104 K, need to rely on molecules and

molecular ions that can be formed in metal-free gas

• A variety of these are possible (e.g. H3+, HeH+, LiH), but

most important is molecular hydrogen, H2
• H2 is the most abundant molecule in the Universe

• At high z, this is 104 — 105 times more abundant than

next most important species

• Understanding formation and destruction of high-redshift

molecular hydrogen is vitally important
• How can we form a molecule?

• If we collide two reactants together (e.g. two H atoms),

initial “reaction complex” has too much energy to remain
bound

• To form a bound product, need to remove some of this

energy

• Simple way to do this is to emit a photon, i.e.

A + B → AB + photon

• This process is called radiative association

• We can estimate the rate at which a chemical reaction
takes place with the following simple formula:

Collision rate of reactants ⨉ Probability of reaction

• In the case of H2 formation by radiative association,

reactants are both neutral atoms:

H + H → H2 + photon

• Collision cross-section for two neutral, unpolarized

reactants is approximately just physical cross-section

• Collision rate coe cient hence ~ 10-11 cm3 s-1

ffi
• Collisions between H atoms occur fairly frequently. What
is the chance that a given collision results in a radiative
association?

• Again, we can estimate this:

Reaction probability = Radiative decay rate ⨉ Collision time

• The collision time is very short: given H atom velocity of

~ 1 km/s (T ~ few hundred K), H atom size ~ 10-10 m, we
have:

tcoll ~ 10-10 / 103 = 10-13 seconds

• What about radiative decay rate?

• For a strong, permitted transition with an energy ~ H2

binding energy, could be as large as A ~108 s-1

• Even in this case, most collisions do not result in reaction

• In case of H2, situation is much less favourable

• H2 has no permanent dipole moment, therefore

stabilizing transitions are forbidden, hence slow

• Reaction probability for H2 formation via radiative

association is tiny
Forrey (2016)
 • At the temperatures of interest for us, H2 formation rate
coe cient is kra ~ 10-28 cm3 s-1

• H2 formation timescale ~ 1 / (kra nH)

• This is ~ 1600 Gyr at z ~ 1000, increasing as (1+z)-3 at

lower redshifts, for gas at mean cosmic density

• Conclusion: radiative association never produces

signi cant amount of H2
ffi
fi
• In local ISM, most H2 forms on surface of dust grains. In
primordial gas, this is obviously not possible

• On Earth, pure atomic hydrogen reacts easily (and

dangerously!) to form H2 via the reaction:

H + H + H → H2 + H

• This has a very small rate coe cient, but scales with
number density n as n3

• It is important at terrestrial densities and during collapse

of gas to form metal-free stars, but not in di use, pre-
galactic gas
ffi
ff
• H2 formation in pre-galactic metal-free gas dominated by
two sets of ion-neutral reactions

• H- pathway:

H + e- → H- + photon

H- + H → H2 + e-

• H2+ pathway:

H + H+ → H2+ + photon

H2+ + H → H2 + H+
• Behaviour of both sets of reactions very similar

• First step is slow, rate-limiting

• Second step is very fast

• H2 stabilized in second step because electron/proton

carries away excess energy as kinetic energy — no need
to emit a photon

• Electrons/protons recycled, not consumed

• Generally, H- pathway dominates, since H- formation is

faster than H2+ formation
• Once H2 forms in pre-galactic gas, it is di cult to destroy

• H2 binding energy ~ 4.5 eV, so need a temperature of

1000s of K to collisionally dissociate — this is important
close to z ~ 1000, but not at z << 1000

• What about photodissociation?

• Direct photodissociation to continuum of electronic

ground state requires ~ 4.5 eV, so many photons at z ~
1000 that can do this

• However, process is forbidden, hence very slow

ffi
• H2 photodissociation generally dominated by two-step
Solomon process

Pak et al (2003)
• Ground state H2 absorbs a UV photon, excited to Lyman
(B) or Werner (C) electronic state

• ~15% of the time, radiative decay from this state ends up

in the vibrational continuum of the ground state

• Hence get ~0.15 photodissociations per photon absorbed

• For H2 in vibrational ground state, minimum energy

required to reach Lyman state is ~11.2 eV

• UV photons capable of exciting H2 to Lyman or Werner

states commonly referred to as Lyman-Werner photons
• Number density of Lyman-Werner photons is negligible in
post-recombination Universe, because of their high energy

• H2 photodissociation therefore completely unimportant at

z << 1000 prior to formation of rst stars

• So why doesn’t the entire IGM become molecular?

fi
• Even at z ~ 1000, timescale to convert substantial fraction
of H to H2 is longer than Hubble time

• This holds even if we assume that all H-, H2+ reacts to

form H2

• In reality, H2 formation strongly suppressed at these

redshifts by CMB photodissociation of H-, H2+
Galli & Palla (2013)
• Can identify ve main periods in formation history of H2

• Phase I (z > 800):

- Abundant electrons catalyze H-, H2+ formation, despite

rapid photodissociation by CMB

- Rapid growth in H2 fraction as Universe recombines,

but maximum fractional abundance still tiny
fi
• Phase II (800 > z > 300):

- Decreasing fractional ionization reduces H-, H2+ formation

rates, but CMB photodissociation remains e ective

- H-, H2+ abundances drop, H2 formation stalls

• Phase III (300 > z > 100):

- CMB unable to photodissociate H2+ (Ebind ~ 2.65 eV)

- H2+ abundance rises substantially, provoking brief burst

of H2 formation

- Process once again freezes out as cosmological density

drops
ff
 • Phase IV (100 > z > 30ish):

- CMB no longer able to destroy H- (Ebind ~ 0.755 ev)

- Further burst of H2 formation, eventually freezing out at

fractional abundance ~ 10-6

- Presence of non-thermal recombination photons limits

e ectiveness compared to case with purely thermal CMB

• Phase V (z < 30ish):

- Formation of rst stars and galaxies. Radiative feedback

provides new source of UV and softer photons and
strongly in uences later evolution of IGM
ff
fl
fi
 Summary

• Recombination occurs over range of redshifts from z ~

1000 (H+) to z ~ 6000 (He++)

• Saha equation provides rough guide to behaviour but

incorrect in detail, for two important reasons:
- Neglects e ect of photoionization of excited H, He etc.
by CMB photons, which delays recombination
- Assumes equilibrium can always be reached — this
becomes impossible at z < 800 (“freeze out”)
ff
• Because of freeze-out, Universe doesn’t become
completely neutral — residual ionization, x ~ 10-4

• This residual ionization plays an important role in the

chemistry of the high redshift Universe

• Details of recombination important because they a ect

the CMB visibility function — ‘precision’ cosmology
requires high precision treatment

• Photons produced during recombination may one day be

visible as spectral distortion of CMB

ff
• Primordial chemistry important because most signi cant
source of gas cooling below 104 K in metal-free gas comes
from molecules (esp. H2)

• Direct formation of H2 extremely slow at pre-galactic

densities

• H2 forms indirectly, via H- or H2+ ions as intermediates

• H2 formation at high z strongly regulated by CMB

photodissociation of these ions

• Final H2 fraction at z < 100 small, xH2 ~ 10-6

fi