Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

PHYS30392 Cosmology

Week 10 Notes

10 Precision Cosmology from CMB Anisotropies


Recall that the blackbody temperature of the CMB varies slightly with direction. These fluctuations are known as
the CMB temperature anisotropies and are a rich source of cosmological information. Here, we will learn some of the
key physics that produce these anisotropies and how they are used to constrain cosmological parameters.

10.1 CMB Temperature Anisotropies


We write the temperature of the CMB radiation as a function of position on the sky, T (θ, φ), where (θ, φ) are
the usual (polar,azimuthal) angle variables in spherical polar co-ordinates. The mean temperature across the sky
is Z
1
µT = T (θ, φ) dΩ, (1)

where dΩ = sin θ dθ dφ is the solid angle element. Of course, the mean temperature µT = TCMB = 2.725 K.
The variance measures the mean of the square of the temperature fluctuations about their mean value, and is
calculated from
Z
2 1 2
σT = [T (θ, φ) − µT ] dΩ

Z
1 2
≡ [∆T (θ, φ)] dΩ. (2)

The r.m.s. variations relative to the mean, σT /µT ≈ 10−5 , i.e. the temperature fluctuations are very small.
We can also analyse the size of the variations on different angular scales, e.g. determine whether hot (∆T > 0) and
cold (∆T < 0) spots in the temperature distribution have a preferred angular size. To do this, we write down the
temperature field as a linear combination of the spherical harmonic functions, Y`m (θ, φ) 1 . It is conventional to work
with ∆T , rather than T , for this purpose; by construction, we have µ∆T = 0. The field can then be written as
∞ X
X +`
∆T (θ, φ) = a`m Y`m (θ, φ), (3)
`=2 m=−`

where we sum over all ` and m values. The spherical harmonic functions are defined as
s
(2` + 1) (` − m)! m
Y`m (θ, φ) = P (cos θ) eimφ , (4)
4π (` + m)! `

where P`m (cos θ) is the associated Legendre function. The integer values (`, m) specify the wave properties for each
mode in our series, defined through Y`m :
• The integer ` is known as the multipole moment (or multipole for short). For each value of `, we have 2` + 1
values of m.
• ` controls the maximum number of spots across the sky. To simplify, we focus on the equator (θ = π/2,
φ = [0, 2π]) and set m = `, its maximum value. There will be ` hot spots and ` cold spots around the equator;
the angular separation between adjacent hot and cold spots is then δφ = π/` (so a separation of 1 degree
corresponds to ` ≈ 180).
• There is only one mode (m = 0) with ` = 0. This is known as the monopole term and is analogous to the
constant term in Fourier series. The ` = 1 term is known as the dipole and produces one hot spot and one cold
spot.
1 Recall this is analogous to Fourier Series decomposition but for a 2D function defined on the surface of a sphere. Larger ` values

describe fluctuations on smaller angular scales.

1
• For cosmology, we are only interested in the ` = 2 (quadrapole) modes and above (hence the ` = 2 lower
limit on the above sum). This is because the monopole term for ∆T should be zero (since the sky-averaged
field is zero). The dipole term is due to our motion relative to the CMB (peculiar velocity), due to the local
gravitational field being anisotropic.
The spherical harmonic functions satisfy the following orthogonality relation
Z
Y`m Y`∗0 m0 dΩ = δ``0 δmm0 , (5)

where δ``0 is the Kronecker-delta function 2 . We can use this property, much like we do with Fourier Series, to find
the constant co-efficients Z

a`m = ∆T Y`m dΩ. (6)

This result can be derived as follows


Z Z XX
∗ ∗
∆T Y`m dΩ = a`0 m0 Y`0 m0 Y`m dΩ
`0 m0
XX Z

= a`0 m0 Y`0 m0 Y`m dΩ
`0 m0
XX
= a`0 m0 δ``0 δmm0
`0 m0
= a`m . (7)

Note that these co-efficients are complex (because Y`m is complex in general). However, since ∆T is real, we require
a`−m = a∗`m (and we also have that Y`−m = Y`m ∗
).
Using this decomposition, and the above properties of a`m and Y`m , we can show (see exercise) that the variance of
∆T can be written as
+`
2 1 X X
σT = |a`m |2 , (8)

` m=−`

where |a`m | = 2
a`m a∗`m . For a given `, the mean-square amplitude is given by
+`
1 X
C` = |a`m |2 . (9)
2` + 1
m=−`

This function is known as the angular power spectrum. It is a measure of the strength of the temperature fluctuations
(in this case) on different angular scales, set by the multipole number `. It is arguably the most important function
in CMB cosmology.
There is a fundamental limitation to how well we can measure C` , due to the limited number (2` + 1) of available
m modes. This is known as cosmic variance. The relative uncertainty on the measured power spectrum is given
by r
∆C` 2
= , (10)
C` 2` + 1
√ is largest for the smallest ` values (where there are fewest modes). Note when `  1, we have ∆C` /C` ∼
and
1/ `.
2δ = 1 if i = j, zero otherwise.
ij

2
10.2 Origin of the CMB Temperature Anisotropies
We will now consider the physical processes that give rise to the anisotropies in the CMB temperature field. It is
useful to consider this in the context of how the radiation field is behaving at different points in the history of the
Universe. Note we will treat this fairly schematically: a full treatment is beyond the scope of this course 3
• We assume the initial density fluctuations in all species are proportional to each other. We call such fluctuations
adiabatic. The dimensionless overdensity field for a species X is defined as
ρX − ρ̄X
δX = , (11)
ρ̄X
where ρ̄X is the spatially-averaged density of X - what we call the background density. 4 Note that δX > 0 for
regions denser than average (overdense regions) and δX < 0 for regions less dense than average (underdense
regions). For example, for baryons and photons we find (not proven) that
4
δγ = δb , (12)
3
for adiabatic fluctuations. We also have δCDM = δb i.e. the baryons and dark matter have the same overdensity
field (relative to their respective background densities). At the epochs we are dealing with here (before and
around recombination/decoupling) we will have |δ|  1 for all species i.e. the relative fluctuations are small.
• Before decoupling (z > 1100), the photons and baryons are in equilibrium and are tightly coupled via Thom-
son scattering (electrons and photons) and Coulomb scattering (protons and electrons). Fluctuations in the
gravitational potential Φ (dominated by the dark matter) will induce pressure fluctuations in the radiation
causing the photons to resist gravitational collapse. This creates acoustic (sound) waves in the photon-baryon
fluid with speed s
dP c
cs = ≈√ , (13)
dρ 3
assuming the pressure of the fluid is dominated by the radiation, P = ρc2 /3. The compressions and rarefactions
of these sound waves produce fluctuations in the radiation density. Since the radiation density ργ ∝ T 4 and
|δγ |  1, we can relate the fluctuations in density to temperature using a first order Taylor expansion
dT T (ρ̄) ρ − ρ̄
T (ρ) ≈ T (ρ̄) + (ρ − ρ̄) ≈ T (ρ̄) + . (14)
dρ ρ̄ 4 ρ̄
Using our definition of overdensity, a fluctuation in density will produce a fluctuation in temperature
∆T 1
= δγ , (15)
T 4
where we use T to mean the background temperature (i.e. T = µT = T (ρ̄)) in the denominator here and below,
for brevity.
• At decoupling (z ∼ 1100), the photons no longer interact with the (now neutral) matter and their mean free path
effectively becomes infinite, the beginning of a period known as the dark ages. As a result, the temperature field
of the radiation (largely) preserves a snapshot of the state at the time of decoupling, allowing us to effectively
observe the last scattering surface. Thus, the previous equation tells us that the CMB temperature anisotropies
should reflect the pattern of density fluctuations (due to the sound waves) at decoupling. We have to add two
other terms to make up what are known as the primary anisotropies:
1. The component of an electron’s motion along the line-of-sight at decoupling induces a Doppler effect,
leading to an additional temperature fluctuation
 
∆T 1
= v · n̂, (16)
T Doppler c
3 More details are covered in the Y4 course The Early Universe.
4 The background density depends on time only and is the value we associate with the density for a perfectly homogeneous universe.

3
where v is the electron’s velocity and n̂ is a unit vector along the line-of-sight. An electron moving away
from the observer produces a blueshift (∆T > 0).
2. Photons in overdense regions at decoupling will need to climb out of a potential well (potential fluctuation
Φ). This induces a gravitational redshift known as the (ordinary) Sachs-Wolfe effect
 
∆T Φ
= 2. (17)
T SW c
Note that an overdensity will have a negative potential fluctuation (Φ < 0) which reduces the temperature.
Together, these effects at the time of decoupling give us a prediction for the primary CMB anisotropies
∆T 1 1 Φ
= δγ + v · n̂ + 2 . (18)
T 4 c c
The density and SW terms dominate the angular power spectrum on almost all angular scales up to ` ≈ 1000,
although the Doppler term is not insignificant, particularly around ` ≈ 50 − 100.
• After decoupling (z < 1100), the photons free stream across the Universe while overdense regions of matter
(dark matter and baryons) with δ > 0 are gravitationally unstable and will eventually start collapsing to form
galaxies and large-scale structures (see later). This leads to what are known as secondary anisotropies in the
CMB and are all integrated effects, i.e. along a photon’s path from the last scattering surface to the observer.
We will discuss a few important ones here:
1. Reionisation. As the first galaxies form, they create high-energy radiation (UV and X-ray photons) that
ionise and heat the surrounding gas to a temperature T ≈ 104 K. At around redshift z ≈ 10, the combined
effect from the galaxies is to effectively reionise all gas in the Universe (X ≈ 0 → X ≈ 1), a process which
(like recombination) happens very quickly and is known as the epoch of reionisation. 5 At this point, CMB
photons are once again able to Thomson scatter off the free electrons. We quantify this effect through the
optical depth τ which is defined as Z t0
τ = σT ne c dt, (19)
tdec
where ne is the free electron density at time t. This integral gives the total probability of a scattering
event along a photon’s path (dl = c dt) from the surface of last scattering at t = tdec to the present
day at t = t0 . Note that for a universe where the baryons are hydrogen (protons) only, we can write
ne (z) = X(z) ρb (z)/mp . Assuming that ionisation is instantaneous at a redshift z = zr , we will have
ne = ρb /mp for z < zr and ne = 0 otherwise, allowing us to write the above integral from z = 0 to z = zr
only.
The effect of this scattering is to smooth out (dampen) the fluctuations on small scales. The measured
value is τ ≈ 0.06, i.e. around 6 per cent of CMB photons are scattered at low redshift. 6 Note that
τ constrains the redshift of reionisation, zr since if reionisation occurred earlier (higher redshift), the
optical depth would be larger (longer path length and higher density at higher redshift since ne ∝ (1 + z)3
when X = 1). Radio astronomy experiments are currently attempting to directly detect the epoch of
reionisation signal using observations of neutral hydrogen (21 cm line).
2. Sunyaev-Zel’dovich (SZ) effects. The gravitational collapse of regions that form groups and clusters of
galaxies produces large amounts of denser, hotter gas (Te ≈ 106 − 108 K), known as the intracluster
medium. Photon scattering from these regions is known as the SZ effects.The most important SZ effect
is known as the thermal SZ effect (tSZ), due to the thermal pressure of the electrons, Pe = ne kB Te . This
produces a CMB temperature fluctuation
  Z
∆T σT
= g(f ) y = g(f ) Pe dl, (20)
T tSZ me c2
5 We are talking here about hydrogen reionisation but also the first ionisation of helium. The second ionisation of helium is expected

to occur later, z ≈ 3, when the ionising radiation from galaxies is even stronger.
6 Note that, while Thomson scattering is once again possible, the mean density of the universe is at least a factor 106 lower than just

before recombination which significantly reduces the scattering probability.

4
where g(f ) is a frequency dependent function and y is known as the Compton-y parameter. Clusters are
typically arc-minute-sized so the SZ effect is important on small scales with ` > 1000.
3. Integrated Sachs Wolfe effect. This is like the ordinary SW effect, but is due to the fluctuations in the
gravitational potential between the last scattering surface and the observer. If the potential of a region
changes with time while the photons cross it, an additional change in temperature is produced as the
blueshift (moving to lower potential) and redshift (higher potential) effects are not equivalent
  Z
∆T 2
= Φ̇ dl. (21)
T ISW c
The ISW affects scales ` < 1000 but is much less important than the primary anisotropies.
4. Gravitational lensing. The photons are also deflected due to the gravitational potential fluctuations,
leading to magnification effects rather like a lens. This effect, also known as CMB lensing, affects all
angular scales but is larger on smaller scales.

10.3 Extracting Cosmological Information from the CMB Angular Power Spectrum
The angular power spectrum contains a number of acoustic peaks which tell us there are certain prominent separation
angles between the hot and cold spots. These peaks are related to a physical length scale known as the sound
horizon, rs , the proper distance for an acoustic wave in the photon-baryon fluid between the Big Bang and the time
of decoupling. The sound horizon is known as a standard ruler since, for a given cosmology, we can calculate its
value Z ∞
cs dz
rs (zdec ) = , (22)
1 + zdec zdec H(z)

which is just the horizon formula for light but we replace c with cs = c/ 3. Over this period in the Universe, we can
approximate the Hubble parameter using the matter and radiation terms only (ignoring curvature and dark energy),
i.e. p
H(z) = H0 Ωm (1 + z)3 + Ωr (1 + z)4 . (23)
If we define u = 1 + z, we can write the sound horizon integral as
Z ∞
c du
rs (zdec ) = √ p , (24)
H0 3Ωr (1 + zdec ) 1+zdec u2 1 + α/u
where α = Ωm /Ωr . The solution to the integral is
Z
du 2p
p =− 1 + α/u, (25)
u2 1 + α/u α
so the sound horizon is hp i
2c 1 + α/(1 + zdec ) − 1
rs (zdec ) = √ . (26)
H0 3Ωr (1 + zdec )α
We already know Ωr h2 from the temperature of the CMB, so the sound horizon depends on the value of Ωm h2 .
Inserting Ωm h2 = 0.15 and Ωr h2 = 4.2 × 10−5 , we get rs ≈ 144 kpc. Note this is a proper distance measured at
decoupling; today, such a region would expand with space to χs ≈ 160 Mpc (i.e. its comoving value).
The first acoustic peak scale (at `∗ ) gives us a measurement of the angular size of the sound horizon, θ∗ ≈ π/`∗ .
This is related to the sound horizon as θ∗ = rs (zdec )/dA (zdec ), where dA is the angular diameter distance to a source
at decoupling Z zdec
c dz
dA (zdec ) ≈ p , (27)
H0 (1 + zdec ) 0 Ωm (1 + z)3 + ΩΛ
which is approximate since we are ignoring curvature and radiation here. The measured θ∗ is consistent with a
flat universe and Ωm ≈ 0.3. Other properties of the power spectrum allow other cosmological parameters to be
determined (e.g. the relative amplitudes of the first and second peaks are sensitive to variations in baryon density
Ωb h2 ). Detailed measurements of these peaks with CMB experiments (particularly the Planck satellite) has placed
tight constraints on the ΛCDM model (including Ωm , Ωb and h).

5
10.4 Non-Examinable: Gaussian Random Fields
This optional, non-examinable section is included for those interested in more theoretical aspects of cosmology. It is
intended to provide some further insight into the statistical nature of the temperature fluctuations but is by no means
exhaustive - this is an enormous area of study!
A more rigorous way to define the power spectrum is to consider the statistical process that generated the fluctuations
in the first place. Imagine we have many copies of our universe i.e. many CMB temperature fields (this is known
in statistics as an ensemble of realisations). For each realisation (CMB sky), we have a particular value for each
complex co-efficient, a`m .
The standard model of cosmology assumes these fluctuations are quantum in origin, and were amplified to cosmo-
logical scales by inflation. Quantum fluctuations are an example of a Gaussian random field; this means the real and
imaginary (m 6= 0) parts are drawn from a Gaussian distribution with zero mean, ha`m i = 0 where hi is an average
over all realisations in our ensemble. A second, important, property of our field (not proven) is

ha`m a∗`0 m0 i = δ``0 δmm0 C` . (28)

This tells us two things: different modes are uncorrelated, so statistically independent, but the variance of each mode
(when ` = `0 , m = m0 ) is the power spectrum C` = |a`m |2 .

An important point is that the variance, C` , is a sum over the square of Gaussian-distributed co-efficients, a`m . The
2
resulting distribution of C` values (when averaging over N realisations for each C` ) is a generalised χp distribution
2
which has a mean of µ = N and variance σ = 2N . The relative uncertainty is therefore σ/µ = 2/N . When
we discussed cosmic variance, above, we were doing something similar except we were averaging over N = 2` + 1
modes within the same realisation, instead of N realisations. Since the modes are uncorrelated, it turns out that
the previous method, which is accessible through observations of only one realisation (our Universe), is an unbiased
estimate of the C` defined here but its variance is now 2 × (2` + 1). This leads to the cosmic variance formula
above.
One last, but important point. We can write each complex co-efficient in terms of its amplitude, |a`m |, and phase,
α`m , as
a`m = |a`m | exp (i α`m ) . (29)
For a given mode (`, m), the amplitude tells us the overall strength of its contribution to the temperature field (real
number) whereas the phase describes the positions of the hot/cold spots on the sky. Since the complex co-efficient
has Gaussian-distributed values with zero mean, the resulting probability distribution for these values only depends
on the amplitude, |a`m | (known as a Rayleigh distribution). This has two important consequences. Firstly, like a
Gaussian, the only parameter we need is the variance (i.e. the power spectrum) to fully specify the distribution. All
higher order statistics, such as the skewness or kurtosis of the distribution, are functions of the variance (or zero).
Secondly, the phase must be a random number, since it does not enter into the probability distribution for a`m . This
means we will only be able to extract any meaningful cosmological information from the power spectrum, which is
why it is so important.
Okay, another last point: the above assumes a Gaussian random field. An active area of cosmology is to search
for non-Gaussian signatures. Then, phase information does become important and the power spectrum is no longer
sufficient to fully specify the field. Non-Gaussianity could emerge as a result of new physics in the early universe, but
also naturally emerges when fluctuations are no longer small. This last point happens in the matter distribution when
gravity causes regions to collapse, leading to the formation of galaxies and large-scale structure (see later).

You might also like