Physical Cosmology Notes

Phil Bull, November 22, 2019

Contents
1. Expanding universe 5
1.1. What is cosmology? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2. Brief history of the Universe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3. The state of the Universe today . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4. Olbers’ paradox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5. Historical discovery of the expanding universe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.6. Expansion and redshift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.7. Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.8. Expansion and the Hot Big Bang model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2. Geometry and distance 12

2.1. Recession velocity and Hubble’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2. Parallax distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3. Proper vs comoving coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4. Measuring distances: the space-time metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5. Friedmann-Lemaı̂tre-Robertson-Walker (FLRW) metric . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.6. Geometry of space: open, closed, and flat universes . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.7. Space-time metric with curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.8. Useful unit conversions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3. Friedmann equation 18
3.1. The Friedmann equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2. Hubble parameter and expansion rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3. Critical density and curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.4. Change in energy density as space expands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.5. Matter-only solution to the Friedmann equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.6. Interchangeability of time, redshift, and scale factor . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.7. Age of the Universe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.8. Matter, curvature, and the fate of the Universe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.9. Newtonian derivation of the Friedmann equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

4. Distances and horizons 26

4.1. Cosmological distances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2. Distance travelled by a light ray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.3. Luminosity distance and standard candles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.4. Distance ladder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.5. Angular diameter distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.6. Cosmological horizon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

5. Cosmic acceleration 30
5.1. Conservation equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.2. Equation of state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.3. Cosmic acceleration and deceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.4. Deceleration parameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.5. Properties of the cosmological constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.6. Cosmological constant solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.7. Age and Hubble horizon in an exponentially-expanding space-time . . . . . . . . . . . . . . . . . . . 33
5.8. The Cosmological Constant problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

1
 5.9. The fate of our Universe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

6. Cosmic Microwave Background Radiation 37

6.1. What was the early universe like? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
6.2. Formation of the Cosmic Microwave Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
6.3. The surface of last scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
6.4. Recombination and decoupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
6.5. Recombination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
6.6. Decoupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
6.7. Blackbody spectrum of the CMB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

7. Cosmic Microwave Background Anisotropies 43

7.1. CMB anisotropies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
7.2. Physical processes that cause anisotropies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
7.3. Baryon acoustic oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
7.4. Diffusion damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
7.5. Secondary anisotropies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
7.6. Spherical harmonics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
7.7. Power spectrum of the CMB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
7.8. Features in the CMB power spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
7.9. Dependence on cosmological parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

8. Inflation 56
8.1. How special is our Universe? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
8.2. The horizon problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
8.3. The flatness problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
8.4. The (magnetic) monopole problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
8.5. The inflationary mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
8.6. Cosmological Klein-Gordon equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
8.7. Scalar field dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
8.8. Slow-roll approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
8.9. Quantum fluctuations and the primordial power spectrum . . . . . . . . . . . . . . . . . . . . . . . . 63
8.10. Reheating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

9. Big Bang Nucleosynthesis 65

9.1. Temperature vs redshift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
9.2. Thermal history of the early Universe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
9.3. Neutron decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
9.4. Nucleosynthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
9.5. Helium fraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
9.6. Observational evidence for Big Bang Nucleosynthesis . . . . . . . . . . . . . . . . . . . . . . . . . . 66

10. Dark matter 67

10.1. Observational evidence for dark matter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
10.2. Types of dark matter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
10.3. Warm vs cold dark matter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
10.4. Hierarchical structure formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
10.5. Dark matter halos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
10.6. Virial theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
10.7. Spherical collapse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
10.8. Press-Schechter formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
10.9. Cores vs cusps: Feedback mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

2
11. Structure formation 69
11.1. Cosmic web . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
11.2. Perturbation theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
11.3. Growth of matter fluctuations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
11.4. Fourier transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
11.5. Matter power spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
11.6. Correlation function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
11.7. Peculiar velocities and redshift-space distortions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
11.8. Fingers of God effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

12. Observational cosmology 70

12.1. Galaxies as a biased tracer of large-scale structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
12.2. Baryon acoustic oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
12.3. Spectroscopic vs photometric galaxy surveys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
12.4. Strong gravitational lensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
12.5. Weak gravitational lensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
12.6. Epoch of reionisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
12.7. 21cm intensity mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
12.8. Radio interferometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

3
About this module
Problem sheets and tutorials
• There are 10 problem sheets, which will be released each week after the first lecture of the week (on Tuesday).
Seven of them are assessed as your coursework, and should be handed in by Wednesday at 4pm the following
week, to the School Office on Floor 1.
• Each problem sheet includes some maths practice questions, some general exercises to help build your under-
standing, and either a practice exam question, or an assessed coursework question. The assessed questions are
graded, and contribute 10% of the final grade for the module. Solutions will be released for each problem sheet
at the tutorial sessions, and posted online the following week.
• Please attempt all of the questions! The problem sheets are designed so that if you can do all of the general
exercises, you should have everything you need to complete the practice exam question or assessed question.
• There are two tutorial sessions per week, at 11–12pm on Tuesday (Laws: 2.07 (1)) and 11–12pm on Wednesday
(ENG: 2.16 (2)). They are run by myself and Rebeca Martinez Carrillho. You are free to attend either one, but it
is important that you attend every week! We will help with all of the questions except the assessed ones, which
you’re expected to do yourself. You are also encouraged to work in groups with other students to figure out the
general questions, but you must attempt the assessed questions on your own. I have a zero tolerance policy for
copying/cheating on assessed coursework questions!

Mid-term exam
There will be a mid-term exam on Tuesday of Week 6 (TBC!), just before Reading Week. This will be worth 10%
of the total marks for the module. If you can do all of the questions on the problem sheets up to Week 5, you should
be well-prepared for the mid-term. (Students who are unable to attend the mid-term due to extenuating circumstances
should discuss re-sits with me.)

Final exam
There will be a final exam in the January exam period, worth 80% of the total marks for the module. A revision session
will be held in Week 12. If you can do all of the questions in the problem sheets, you will be well-prepared for the
exam. The exam questions will not be based directly on these problems though, so make sure you understand the
material (rather than just practising the problems)!

Past exam papers

Past exam papers will be made available, but please not that the module content has changed since previous years. In
particular, you should expect the exam to be more difficult! (This is why it’s strongly advisable to attend the tutorials,
do all the problems etc...)

Office hours and extra support

My office hours are from 1–2pm on Tuesdays and 3–4pm on Thursdays, in room 501 of the G O Jones building. Please
let me know if you’re intending to come so I can keep an eye out for you! The office hours are a good time to ask about
material you don’t get or have some confusion about. Rebeca and I can also discuss problems and answer questions
during the tutorials.
Students on the MSc course who have a timetable clash with the Thursday lecture can catch up on Q-Review. Note
that these lectures do contain essential material, so it’s important that you do catch up!

4
1. Expanding universe
In this section you will learn some of the basic facts about cosmology, including a brief history of the Universe and the
fact that space is expanding. As part of this, you will learn some of the history behind the discovery of the expansion
of the Universe, and some of the key arguments in support of this observation.

Reading for this topic:

– An Introduction to Modern Cosmology (A. Liddle), Chapter 1: Brief History of Cosmological Ideas.
– An Introduction to Modern Cosmology (A. Liddle), Chapter 2: Observational Overview

1.1. What is cosmology?

Cosmology is the study of the whole Universe – how it formed, how it evolves with time, what its basic properties and
constituents are, and how it is structured on the largest distance scales.
Cosmologists study the Universe through a combination of astronomical observations, mathematical theories, and
computer simulations. The field has only been recognised as a scientific discipline for about 100 years – before then it
was the domain of philosophers and theologians!
This module is about the fundamental facts and findings of cosmology, and the physical theories that we have
developed to explain them.

Figure 1: Illustration showing different periods of cosmic history, along with roughly what time they occurred since
the Big Bang. A more detailed description is given here. (Adapted from figure by Planck / C. Carreau)

5
1.2. Brief history of the Universe
One of the most fundamental discoveries in cosmology is the fact that the Universe as a whole evolves with time.
Cosmic history can be divided into several epochs, depending on what the dominant physical processes were at the
time. The transitions between these epoch are often very interesting physically, as they tend to imprint particular
observable signatures.
The figure above illustrates the various phases of the Universe’s history. A brief explanation of some of them is
given below.
• Big Bang – The very beginning of the Universe (t = 0s).
• Inflation epoch – Brief period in the early Universe when space expanded exponentially quickly (t ≈ 10−30 s).
• Nucleosynthesis – Protons and neutrons formed into the first nuclei (t ≈ 3 min).
• Recombination – Nuclei and free electrons combined to form the first neutral atoms; the Universe became
transparent to light (t ≈ 380, 000 years).
• Dark ages – The Universe was filled with neutral gas, and there were no luminous sources giving off light
(t . 200 million years).
• Cosmic Dawn – The first stars and galaxies formed. Their light began to reionise the neutral gas in the Universe
(t . 1 billion years).
• Structure formation – Galaxies began to form into large structures, forming a cosmic web of galaxy clusters,
voids, and filaments. The galaxies themselves changed with time, as different stellar populations inside them
evolved (t & 1 billion years).

1.3. The state of the Universe today

It is only in the last couple of decades that cosmologists have been able to measure the basic properties of the Universe
with reasonable precision. We now know how much matter there is to better than 1% accuracy. We also know how
fast the Universe is expanding with around percent-level accuracy too (the current best measurement is 67.36 ± 0.54
km/s/Mpc). As we’ll see later, this lets us infer the age of Universe, which is currently best estimated to be 13.797 ±
0.023 Gyr.

Figure 2: Chart showing the fraction of the cosmic energy density today taken up by different types of matter and radi-
ation. These values are either measured from astronomical observations or inferred from a combination of observations
and physical models. N.B. The chart on the right shows the fractions that belong in the very narrow white wedge in the
plot on the left. (Wikipedia)

The chart above shows our current best estimates of the composition of the Universe today (i.e. not its composition
far into the past, which we’ll see was quite different!). There are several different species of matter, radiation, and other
types of energy, some of which are more abundant than others.

6
1.4. Olbers’ paradox
Why is the night sky dark? We see plenty of stars, especially if we look through a telescope, but most of the sky is still
dark. This should not be the case if the Universe is infinitely large and infinitely old though – every direction you look
in would, eventually, end up on a star. As a result, the whole sky would be bright. The statement of this problem is
called Olbers’ paradox.
Stars that are further away appear fainter though. How bright should we expect the sky to be if most of the stars are
at great distances? Imagine that the Universe is filled with a uniform density of stars, n, on average. For simplicity, let’s
also assume that they all have identical luminosity, L. Each star will be observed on Earth with a flux f = L/(4πr2 ),
where r is the distance of the star from Earth. If we draw a spherical shell of width dr at a distance r around Earth, we
obtain a total number of stars N = 4πr2 ndr in that shell. The total flux of starlight from each shell is therefore
L
dftot = 4πr2 ndr = nLdr. (1)
4πr2
As you can see, it doesn’t depend on distance! In fact, each shell contributes the same amount to the total flux received
at Earth – while individual stars get fainter as the distance increases, the size of each shell (and therefore the number
of stars per shell) increases by the same factor, and so the two effects cancel. If we integrate the flux over shells at all
radii, we then get
Z Z ∞ Z ∞
ftot = dftot = n L dr = n L dr = ∞. (2)
r=0 r=0

Needless to say, this is not what we observe, and so there must be something wrong with our assumptions. One thing
we assumed is that all of the flux from all of the stars would reach us on Earth. This is not true if we consider that stars
have a small, but non-zero, angular size, and are themselves opaque. Each star would block the light from any stars
behind it, and so along each line of sight we would see only the light from the nearest star in that direction. This is an
improvement – the brightness of the sky is no longer infinite – but it does not solve the paradox. Since every line of
sight ends at a star, we would still have a uniformly bright night’s sky about as bright as the surface of the Sun! (Prove
this in the first problem sheet!).
Couldn’t the darkness that we see be caused by intervening clouds of gas and dust that absorb the light? We
certainly do see dust clouds in the sky, such as the Coalsack nebula, that block out the light from the stars behind them.
But in an infinitely old Universe, the clouds would absorb so much energy over time that they would heat up and glow
as brightly as the stars behind them!
The solution – obvious to us now – is that the Universe must either be finite in age, or expanding, or both, such that
starlight has not had chance to reach Earth along every line of sight, even if it was filled uniformly with stars. Since the
speed of light is finite, there is a maximum distance light could have travelled since the start of time. If the Universe
is expanding, stars very far away would be expanding away from us faster than the light can travel towards us, and so
their light would never reach us. These are both types of cosmic horizon, which we will learn more about in Week 4.

Figure 3: A picture of the Coalsack nebula. A more detailed description is given here. (Credit: ESO.)

7
1.5. Historical discovery of the expanding universe
Cosmology did not exist as a scientific field until after Einstein developed his theory of General Relativity. While some
scientists did try to come up with ‘world models’ before then, they were often highly speculative. Even the poet Edgar
Allen Poe had a stab at it, and came surprisingly close to what we now know to be the right answer. Wikipedia has a
timeline of cosmological theories that explains a lot of the historical ideas.
Few of the pre-20th Century thinkers had good scientific justifications or observational support for their theories
however. Below is a brief timeline of the scientific discovery of the expansion of the Universe:

1908 Henrietta Swan Leavitt (astronomer and ‘computer’) discovered a relation between the pulsation period and
absolute luminosity of Cepheid variable stars. This made it possible to measure the distance to very remote
astronomical objects.

1913 Vesto Slipher (an astronomer) first measured the radial velocities of ‘spiral nebulae’. It was not yet known that
they were in fact galaxies separate from our own. By 1917 he had found that most spiral nebulae seemed to be
moving away from Earth at quite high velocities.

1916 Albert Einstein published his General Theory of Relativity, including the Field Equations that connect the geom-
etry of space-time to its matter/energy content.

1917 Willem de Sitter (an astronomer) found the first cosmological solution to the Field Equations – an empty ‘vac-
uum’ universe that seemed to expand.

1917 Einstein did not like this. He found another solution, called the Einstein static universe, that contained matter
and had no expansion, but required a new ‘cosmological constant’ to be added into his equations to keep things
from expanding. It was unstable, and Einstein himself thought it was ugly!

1922 Alexander Friedmann (a meteorologist) derived the Friedmann equations from Einstein’s Field Equations. These
showed how the Universe expands or contracts in the presence of a perfectly homogeneous fluid. Einstein himself
reviewed Friedmann’s calculations, but didn’t realise their significance!

1924 Edwin Hubble (astronomer) observed Cepheid variables in spiral nebulae. He concluded that the nebulae were
far too distant to be part of the Milky Way; they were galaxies separate from our own.

1927 Georges Lemaı̂tre (a Catholic priest and astronomer) derived a mathematical relation to explain how galaxies
would seem to be travelling away from us in an expanding universe. His work was also not widely appreciated.

1929 Hubble plotted radial velocity data compiled by Milton Humason (his assistant) and Vesto Slipher against the
distances to galaxies he had measured using Cepheid variables. He found that the further the galaxies were
from us, the faster they seemed to be moving away. This was soon interpreted as evidence that the Universe is
expanding. His estimate of the expansion rate was too large by a factor of seven however!

Further reading: Who discovered the expanding universe? (H. Kragh & R. W. Smith); Henrietta Leavitt – Cel-
ebrating the Forgotten Astronomer (AAVSO); The Contribution of V. M. Slipher to the Discovery of the Expanding
Universe (C. O’Raifeartaigh); Edwin Hubble: Redshift increases with distance (Wikipedia)

1.6. Expansion and redshift

What does it mean for space to be expanding? We are used to thinking in a Newtonian way, where space itself is a fixed
‘stage’ on which particles and other objects move around. It seems much more natural to think of particles moving
away from each other – expanding out from some location – than to think of the space itself changing. And yet that
is what general relativity tells us is happening – and indeed must be happening. The expansion of the Universe is a
relativistic effect, that we can only approximate using Newtonian physics.
Where is the centre of the Universe? The Universe does not have a centre! Every point is expanding away from
every other point. Similarly, the Big Bang is not an explosion. Explosions travel out from a point, while the Big Bang
happened everywhere at the same time.
One way to understand what’s going on is by using the raisin loaf model as an analogy. This asks us to imagine
a very large (actually, infinite!) lump of raisin loaf dough being baked in an oven. The raisins are embedded in the

8
Figure 4: The ‘raisin loaf’ model provides an analogy for how the expansion of space is experienced by observers
within the expanding space. A more detailed description is given here. (Credit: OpenStax Astronomy.)

loaf, in the same way that galaxies are ‘embedded’ in space. As time goes by, the dough expands homogeneously and
isotropically, i.e. at an equal rate everywhere, and in all directions. We can then ask: What happens to the distance
between raisins (galaxies) as this happens? And how fast does each raisin appear to be moving away from the others?
For the first question, this analogy neatly shows that the distances should increase proportionally as the dough
(space) expands. If raisins A and B are twice as far apart as raisins A and C at the start, they will remain twice as far
apart as the dough expands.
For the second question, we can pick any raisin as a reference point and consider how the other raisins appear to
be moving relative to it. We find that they are all moving away, and that the speed at which they are moving away
(receding) is proportional to their distance. So, if we use raisin A as our reference, we will see raisin B moving away
at twice the speed of raisin C (since B is twice the distance from A). This is because there is twice as much dough
between A and B as between A and C, and so there is twice as much expansion (therefore the distance increased twice
as much). It’s exactly the same for galaxies embedded in an expanding space.
The raisin loaf model also gives some insight into why we don’t need there to be a centre to the expansion. We
could have chosen any raisin as our reference point, and would have found exactly the same behaviour. In the real
Universe, all of space is expanding at the same rate in every direction, and so all galaxies are receding away from all
other galaxies.
What expands when space expands? In principle, everything – the separation between every point in space
and every other point increases with time. So certainly, the distance between two distant galaxies increases as space
expands. But do the galaxies themselves expand? Or the stars within them? Do we expand?
First, we can work out how much an object of a certain length should be expanding. The current expansion
rate of the Universe has been measured to be around 70 km/s/Mpc. That means every megaparsec of space expands
by 70 km every second. For a 1.8m tall human, the corresponding expansion is 70km/s/Mpc × 1.8m × (3.086 ×
1022 )−1 Mpc/m ≈ 4.1 × 10−18 m/s. For comparison, a Hydrogen atom is around 10−10 m across, so clearly this is a
tiny amount! Over a lifetime of 80 years, this would amount to a total change in height of only 10 nm.
A galaxy, however, is considerably larger than a human. If we take a typical spiral galaxy to be around 40 kpc in
diameter, we get an expansion of 2.8 km/s. This seems small compared to the size of the galaxy, but would actually be
observable with sensitive astronomical instrumentation! We don’t see this expansion however.
The reason is that galaxies are gravitationally bound. Because they have so much mass concentrated in a relatively
small volume, the force of their own gravity overcomes the cosmological expansion, and so they do not increase in size
as the Universe expands. Similarly, humans are bound together by electromagnetic interactions between the atoms and
molecules in our bodies that are vastly stronger than the cosmological expansion.
Further reading: The expanding universe (OpenStax Astronomy)

1.7. Spectra
One of the most important tools in astronomy is spectroscopy. Atoms and molecules emit and absorb electromagnetic
radiation with distinctive lines in their spectra, corresponding to differences in energy levels of their electrons. By

9
measuring spectra from distant sources, we can learn which atoms and molecules are present, and their abundance.

Figure 5: Spectrum of a spiral galaxy, with some spectral lines labelled. (Credit: Steward Observatory / R. Kennicutt)

As discussed above, the expansion of space causes the wavelength of light to increase. The greater the distance the
light has travelled, the more stretching of space it will have experienced, and so the greater the change in wavelength.
Emission and absorption lines will therefore be shifted to the redder (longer wavelength) end of the spectrum. By
measuring the new wavelength of the lines, and comparing with what we know to be their ‘rest-frame’ wavelength
(the wavelength when they were emitted), we can calculate how much space must have expanded since the light was
emitted. This stretching factor is called the redshift (usually denoted by z), and can be calculated using

λobs /λemit = (1 + z). (3)

Spectra can be shifted for other reasons besides the stretching of space. Another is the Doppler shift, due to a relative
velocity between the emitter and observer. The shift is given by
v
λobs /λemit = 1 + , (4)
c
where a positive velocity is defined as moving away from the observer. The spectral shifts we measure for astronomical
objects is typically a combination of these two effects.

1.8. Expansion and the Hot Big Bang model

While the fact that the Universe is expanding was established in the late 1920’s, there were several competing theories
for how the expansion was happening and what this meant for the age of the Universe. One theory, the hot Big Bang
model, stated that the Universe began in a hot and very dense state (mathematically, taking the form of a singularity),
with its contents cooling rapidly as space expanded. This would mean that the Universe had finite age (a beginning, the
Big Bang), and would look different (hotter, denser) in the past than the present (cooler, less dense).
Another theory, the Steady State model, said that the Universe was infinitely old, but was continually expanding
at the same rate. As the expansion happened, new matter was being created throughout space at a slow but steady rate.
Enough matter should be created to keep the density of the Universe constant in time, cancelling out the diluting effect
of the expansion. The properties of this Universe would not change with time; looking back into the past, the Universe
would seem very similar to the Universe we see today.
Other models existed, such as cyclic models that allowed the Universe to expand from a Big Bang-like singularity,
then collapse back in on themselves, and then ‘bounce’ back into a new expanding phase. This cycle would be repeated,
possibly forever. Each phase of expansion would look like it came from a hot Big Bang until the Universe started to
collapse again.

10
 All of these theories were seriously entertained by scientists until around the late 1950’s and early 1960’s, when
very distant objects called quasars were discovered. (We now know quasars to be the very bright, active nuclei of
galaxies, associated with supermassive black holes.) The observations showed that the abundance of quasars changed
significantly the further away (and therefore the further back in time) you look. This matched the predictions of the hot
Big Bang model (“the Universe was denser in the past”), and did not support the Steady State theory (“the Universe
has always looked very similar and doesn’t change with time”). Further evidence came from the discovery in 1964 of
the Cosmic Microwave Background – remnant radiation from a very hot previous phase of cosmic history. This was a
key prediction of the hot Big Bang model, but not the Steady State model.
Further reading: Steady-state model (Wikipedia)

Understanding:

X What is cosmology?
X What are the major epochs in the Universe’s history?
X What physical processes are important during each of these epochs?
X What does it mean for space to be expanding?
X What does it mean for a universe to be homogeneous and isotropic?
X What is Olbers’ paradox, and how is it resolved?
X How does the expansion of space affect the frequency of EM radiation?
X How are spectra used to measure redshift?
X What is the evidence for the hot Big Bang model?

11
2. Geometry and distance
In the last section we learned about the expansion of space, and how observers perceive that expansion. In this sec-
tion, we will develop a mathematical understanding of the expansion, and how it affects how we define and measure
distances. We will use the language of General Relativity to define some of the relevant quantities, for example by
defining a mathematical object called the spacetime metric that tells us how to measure distances even when space is
expanding. We will also learn about the possible geometries of the Universe – space itself can curve, as well as expand.
We will examine the different ways it is allowed to curve.

Reading for this topic:

– An Introduction to Modern Cosmology (A. Liddle), Chapters 2, 4, and 5

– An Introduction to Modern Cosmology (A. Liddle), Advanced Topic 1: General Relativistic Cosmology

2.1. Recession velocity and Hubble’s Law

Recall from the previous section that the expansion of space (if it is expanding homogeneously and isotropically) causes
all distances to increase proportionally, i.e. by the same factor. As we saw in the raisin loaf model, this means that
objects that are further away are expanding away from us faster than objects that are nearby – the increase in their
distance per unit time is proportional to their current distance.
The rate of increase in distance due to the expansion can be interpreted as a velocity that we call the recession
velocity. Objects at larger distances have larger recession velocities – an observation that we call Hubble’s Law:

v = H0 d. (5)

Here, v is the recession velocity and d is the distance. The recession velocity is proportional to the distance, as
we expected from thinking about the raisin loaf model. The constant of proportionality, H0 , is called the Hubble
parameter.1 It has units of inverse-time, but is most commonly quoted in a more convenient mixture of units: km/s/Mpc
(speed per unit distance).
This relation was discovered observationally by Hubble – he plotted Slipher and Humason’s measurements of the
recession velocity of galaxies against his own measurements of distances from Cepheid variables (using the distance
measurement technique of Swan Leavitt), and saw a linear trend between them. Prior to this, Lemaı̂tre had discovered
the relation theoretically, by calculating how distances would change in an expanding space. The Hubble Law is
therefore sometimes called the Hubble-Lemaı̂tre Law instead.
Note that if the Universe is expanding, the distance of an object from us is always increasing, and so the recession
velocity is a positive quantity. The radial velocities we measure from spectra can sometimes be negative however, if
the galaxies have a peculiar velocity towards us that counter-acts the expansion and causes a big enough Doppler shift.
These galaxies are seen with a blueshift. (Of course the peculiar velocity could also be pointing in the direction away
from us, in which case the galaxy would seem to have an even bigger redshift.)
Since the recession velocity increases with distance, it gets very large for very distant objects. First of all, this
means that every distant galaxy has a redshift, and never a blueshift – at small distances, where the recession velocity is
small, it is easy for galaxies to have a large enough peculiar velocity to cancel it out and appear to be moving towards
us. At large distances, the recession velocity due to the expansion of space always wins over the Doppler shift.
We can see from the Hubble Law that as the distance, d, becomes larger and larger, the velocity also gets larger and
larger – with no limit! In fact, when d > c/H0 , this implies that the recession velocity v > c! This is perfectly fine –
plenty of galaxies have been found at such large distances. Remember that the recession velocity is not a real velocity
– it’s just a way of rewriting the redshift in a different way, by analogy with the Doppler shift. It is not really a Doppler
shift however, so not really a velocity, and so doesn’t need to follow the same rules as actual velocities (such as being
less than the speed of light).
Further reading: Hubble’s Law (HyperPhysics); Hubble’s Law (Khan Academy)

1
It is not actually constant, as we will see in later sections.

12
Figure 6: The ‘discovery plot’ for the expansion of the Universe. It shows distance (x axis) plotted against velocity (y
axis) for a number of nearby galaxies. (Credit: E. Hubble)

Taylor expansions
Almost any function can be written as a polynomial with an infinite number of terms,
∞
X
f (x) = bn (x − a)n , (6)
n=0

where bn are a set of coefficients that are to be found, and a is some reference point that we are free to choose. If
the infinite series converges rapidly (i.e. the terms in the polynomial become smaller and smaller as n increases),
we can accurately approximate the function using only the first few terms in the series.
This is the logic behind a Taylor series expansion. By taking the first few terms in an infinite series expansion
around a suitable reference point (let’s call it a), we can often get a very good and simple approximation to a
function that is valid within some region around the reference point.
To calculate the value of each coefficient, we can take nth-order partial derivatives, i.e. ∂ n f (x)/∂xn , and then
set the argument to the reference value, x = a. This isolates the coefficient for the nth term, up to a numerical
factor: lower-order terms (< n) are removed by differentiation, while higher-order terms (> n) cancel when
x = a. (You can prove this to yourself by repeatedly differentiating the expression above and evaluating at
x = a to see which terms remain.)
Using this insight, we can write a general expression for a Taylor series expansion around a point a as

1 ∂ 2 f  1 ∂ 3 f 
  
1 ∂f  2
f (x) ≈ f (a) + (x − a) + (x − a) + (x − a)3 + . . . (7)
1! ∂x a 2! ∂x2 a 3! ∂x3 a

Further reading: Taylor & Maclaurin polynomials (Khan Academy)

2.2. Parallax distance

The distance measurements used by Hubble were made by using Cepheid variable stars as “standard candles”. Another
way of measuring astronomical distances is by observing the parallax. This is the maximum change in the angular

13
position of an object on the sky over the course of the year; as the Earth orbits the Sun, our viewing angle changes very
slightly, leading to a tiny shift in the angular position of an object. The further away the object is, the less its angle will
change from one side of the Earth’s orbit to another.
The parallax (change in angle) is most conveniently expressed in units of arcseconds. The measured parallax in
arcseconds is related to the distance in parsecs by:
 −1
d ∆θ
= . (8)
pc arcsec

So, an object that has a parallax of 0.5 arcsec is at a distance of 2 pc, while an object with a parallax of 1 µas (1 micro
arcsec = 10−6 arcsec) is at a distance of 1 Mpc.
Parallax distances are only useful for relatively nearby objects, since the parallax angles that we would need to
measure become really tiny for objects at cosmological distances. Measuring a parallax of 1 µas is very challenging,
but can be done.
Further reading: Parallax (Wikipedia); Stellar distance using parallax (Khan Academy)

2.3. Proper vs comoving coordinates

Measuring distances in an expanding Universe is complicated, since space itself is constantly changing. We can choose
a set of spatial coordinates to make this a bit more straightforward though.
First, imagine looking at a ‘snapshot’ of the Universe, frozen at a fixed time. What would the distances between
the galaxies be? Those distances are called the proper distances, and increase as space expands. Importantly, we can’t
directly measure proper distances! How could we? Even if we did have a ruler big enough, we wouldn’t be able to
tell when one end had reached the galaxy we were trying to measure the distance to; we would have to wait for signals
from the end of the ruler to travel back to us at the speed of light (or slower). By the time those signals made it back,
the Universe would have expanded more, and the ruler wouldn’t be touching the galaxy any more! (We will see how
the proper distance is related to quantities we can measure later on.)
Proper coordinates are quite inconvenient, as the distances between galaxies and other distant objects are constantly
changing, simply because the Universe is expanding. By defining coordinates that also expand, we can factor out the
expansion. This is shown in the diagram below – in this coordinate system, the galaxies always stay in the same place
on the coordinate grid, even though the (proper) distance between them is increasing.
These coordinates, with the expansion factored out, are called comoving coordinates. Proper coordinates are related
to comoving coordinates by a factor of a(t), like so:

~xp = a(t)~x, (9)

where ~xp is a vector in proper coordinates and ~x is the same vector in comoving coordinates.
Galaxies and other objects can move within a comoving coordinate system, but this movement will not be due to
the expansion (since that has been factored out). Instead, it will be due to their peculiar velocity, if they have one.

2.4. Measuring distances: the space-time metric

In a 3D Euclidean space, the line element – the infinitesimal unit along a straight line that connects two points – is given
simply by dr2 = dx2 + dy 2 + dz 2 , where dx, dy, and dz are the (infinitesimal) distance intervals in each dimension.
In 4 dimensions, we can define an analogous line element along a curve separating two points (or events) in space
and time. There are some important differences when a time dimension is added though. In the Minkowski space-time
of special relativity, the line element becomes

ds2 = −c2 dt2 + dx2 + dy 2 + dz 2 = −c2 dt2 + dr2 . (10)

The time component is scaled by the speed of light, and contributes negatively to the space-time interval. A more
detailed explanation is beyond the scope of this course, but light travels on paths where the space-time interval is
always ds2 = 0, while normal matter travels on paths where ds2 < 0.

14
Figure 7: Relationship between comoving and proper coordinates. The galaxies stay at the same comoving coordinates
as space expands (i.e. they stay fixed to the same points on the comoving grid). The proper distance between them
increases however (the grid itself expands, so all proper distances increase). (Credit: E. Bertschinger)

It is useful to define a quantity called the space-time metric. This describes how intervals in the individual dimen-
sions contribute to the total space-time interval ds. The metric is a special type of 4 × 4 matrix called a tensor, and for
a Minkowski space-time can be written as
 2 
−c
 1 
gab =  . (11)
 1 
1
This can be used to define the line element like so:
3 X
X 3
ds2 = gab dxa dxb , (12)
a=0 b=0

where indices a and b run over the 4 space-time dimensions, and we have defined a space-time vector (4-vector) as
xa = (t, x, y, z).

Einstein summation convention

Lots of calculations in special and general relativity involve summing over one or more sets of indices on 4-
vectors and tensors (as in the expression for ds2 above). This gets very cumbersome for anything other than
simple calculations.
To help with this, we can introduce some new notation called the Einstein summation convention. When using
this convention, if you see a pair of repeated indices, the rule is that you’re supposed to sum over them. For
example, the example above can be rewritten as

ds2 = gab xa xb . (13)

Index a is repeated, which means we should do a sum over all values of a, from 0..3. Likewise for b. Sometimes
people use indices i, j, k to denote only the 3D (spatial) part of a 4-vector, i.e. components 1..3. So, writing
xi xi would denote the sum x2 + y 2 + z 2 . Latin letters early in the alphabet (e.g. a, b, c) or Greek letters (e.g.
µ, ν) normally denote 4D indices however.
There is also significance to whether the indices are raised (e.g. xa ) or lowered (e.g. xa ). This has to do with
whether the 4-vector is a covariant or contravariant vector. You can convert between the two by multiplying by
the metric, e.g. xa = gab xb .
Further reading: Einstein notation (Wikipedia); Covariant and contravariant vectors (Wikipedia); Covariant
and contravariant vectors (animation).

15
2.5. Friedmann-Lemaı̂tre-Robertson-Walker (FLRW) metric
The metric for a homogeneous and isotropic expanding spacetime is
 2 
−c
 a2 
gab =  2
. (14)
 a 
a 2

This is called the Friedmann-Lemaı̂tre-Robertson-Walker (FLRW) metric, sometimes just called the FRW metric
(which is a bit mean to Lemaı̂tre). It is named after the people who came up with the corresponding solutions to
Einstein’s equations, and the people who wrote it down in this particular mathematical form.
Writing this as a line element, we obtain
ds2 = −c2 dt2 + a2 (t)dl2 , (15)
where dl is the spatial part of the line element.
It is important to note that the scale factor, a(t), is the same in all spatial directions. In other words, space stretches
the same amount in every direction – the expansion is isotropic. What’s more, note how the scale factor only depends
on time, but not position. The stretching of space happens by the same amount in every location – the expansion is
homogeneous.
So far we have been working in Cartesian coordinates (x, y, z) for the spatial part of the metric, but there is no
reason why we need to. It is often more convenient to work in spherical polar coordinates (r, θ, φ) for example. The
line element in spherical polars becomes
ds2 = −c2 dt2 + a2 (t) dr2 + r2 dθ2 + r2 sin2 θdφ2 .


(16)
As another point of notation, we will often drop the explicit dependence on time of quantities such as the scale factor,
so we can write a = a(t).

2.6. Geometry of space: open, closed, and flat universes

We are used to doing geometry in a Euclidean space; parallel lines never meet, and the angles of a triangle add up to
180◦ . This is not true in the other types of geometry however! In a positively-curved space, the angles of a triangle add
up to greater than 180◦ , and parallel lines always meet at some point! Conversely, in hyperbolic spaces, the angles of a
triangle add up to less than 180◦ and parallel lines diverge.
How should we visualise these different types of geometry? It’s easiest to think about the 2D analogues of these
spaces embedded in the 3D space that we’re used to. The 2D analogue of 3D Euclidean space is a flat plane, extending
to infinity in the x and y directions. You can just imagine an infinite sheet of paper. The 2D analogue of a hypersphere
is the surface of a sphere (not the sphere itself!). You can imagine a globe for this one. The 2D analogue of a hyperbolic
space is the surface of a saddle. This one is a bit more difficult, as you have to imagine the saddle extending off the
infinity in all directions.
Importantly, all of these spaces are still homogeneous and isotropic. It doesn’t matter where you are on a hyper-
sphere for example; space still looks like same in every direction, and has the same properties at every point.

2.7. Space-time metric with curvature

With spatial curvature, the FLRW metric becomes
dr2
 
2 2 2 2 2 2 2 2 2
ds = −c dt + a (t) + r dθ + r sin θdφ . (17)
1 − kr2
Note how k has dimensions of (distance)−2 to make the units work out. Depending on the sign of k, we arrive at
the three different possibilities for the curvature of space:
• k = 0: Flat, or Euclidean.
• k > 0: Positive curvature, i.e. a closed universe. The 3 space dimensions have the shape of a hypersphere.
• k < 0: Negative curvature, i.e. an open universe. The 3 space dimensions have a hyperbolic (saddle-like) shape.

16
Figure 8: Illustrations of different type of curvature. Top is a closed surface, middle is an open surface, and bottom is
a flat surface. These illustrations show curved 2D surfaces embedded in a 3D space; when we talk about the curvature
of the Universe, we are talking about a 3D ‘surface’ embedded in a 4D space. (Credit: Wikipedia)

2.8. Useful unit conversions

• 1 pc = 3.086 × 1016 m

• 1 yr = 3.154 × 107 s (≈ π × 107 s)

• c/(100 km/s/Mpc) = 2997.9 Mpc (≈ 3000 Mpc)

Understanding:

X What are the definitions of ‘recession velocity’ and ‘Hubble’s Law’?

X What are proper and comoving coordinates?
X What is the FLRW metric and the corresponding line element for an expanding universe?
X How can a Taylor series be used to approximate a function?
X What does it mean for space to be curved?
X What are the different types of curved spaces, and what are their properties?

17
3. Friedmann equation
In this section, you will learn about the most important equation in cosmology – the Friedmann equation, which de-
scribes how the expansion rate of the Universe depends on its matter content and geometry. You will learn two different
ways of writing the Friedmann equation (one with units, one that is dimensionless), and how to solve the equation in
a few different cases, so that you can calculate the age of the universe and its future fate. This section also includes an
intuitive, non-relativistic derivation of the Friedmann equation based on Newtonian gravity.

Reading for this topic:

– An Introduction to Modern Cosmology (A. Liddle), Chapters 3, 5, 6, and 8.

– An Introduction to Modern Cosmology (A. Liddle), Advanced Topic 1: General Relativistic Cosmology.

3.1. The Friedmann equation

In previous sections we defined the scale factor, a(t), which describes the factor by which the Universe has expanded
at any given time. But how can we determine the scale factor and its time evolution in a given universe?
General Relativity gives us the tools to derive an evolution equation for the scale factor. GR links the geometry of
the Universe and how it is changing to the matter/energy content of the Universe. Different types and configurations of
matter and energy cause the geometry to respond and involve in different ways.
For a homogeneous and isotropic space-time, the equation that describes this link between geometry and energy
content is the Friedmann Equation:
 2
ȧ 8πGρ kc2 Λc2
= − 2 + (18)
a 3 a 3
We will derive this later on, but for now let’s define each of the terms. First, we see on the left a ratio of the time
derivative of the scale factor, ȧ = da/dt, to the scale factor itself. On the right, the first term is the energy density, ρ,
scaled by a factor; the second term depends on the curvature of space, k; and the third term involves a cosmological
constant, Λ, which we will study in more detail later.
Note that only the factors of a = a(t) and ρ = ρ(t) depend on time in this equation. The curvature, k, and
cosmological constant, Λ, are fixed, as are the physical constants G and c.
Importantly, the density ρ here is an energy density, not a mass density. It’s very useful to use units where c = 1
though, so in practice we don’t need to distinguish (e.g. since E = mc2 can be written E = m in these units). This
density is the sum of the densities of all of the matter and radiation fields in the Universe (we will break this into its
separate pieces later).
It is sometimes convenient to interpret the cosmological constant term as an energy density also, by defining
Λc2 8πGρΛ
= . (19)
3 3
We will discuss this interpretation in Week 4 – this simple rewriting of this term turns out to be related to one of the
most fundamental problems in all of physics!

3.2. Hubble parameter and expansion rate

We previously met the Hubble parameter, H0 , in the context of the Hubble Law. It tells us the rate of expansion of the
Universe today. We can relate this to the scale factor and its time derivative like so:
H0 = (ȧ/a)t=t0 . (20)
That is, the Hubble parameter is the time derivative of the scale factor divided by the scale factor, all evaluated at t0
(today). We can use a very similar definition to find the expansion rate at any given time:
H(t) ≡ ȧ/a. (21)
As you can see, the left hand side of the Friedmann equation is nothing but the expansion rate squared! So the expansion
rate depends on the matter/energy density of the Universe, its curvature, and the size of the cosmological constant.

18
3.3. Critical density and curvature
In the previous section, we discussed a way of classifying universes according to how space is curved. The sign of the
curvature parameter, k, determines whether the universe is open, closed, or flat.
We can relate the curvature to the density and expansion rate, using a quantity called the critical density, ρcrit . For
a given expansion rate, a universe that has a lower density than ρcrit will be open, while one with a greater density than
ρcrit will be closed. A universe at exactly ρcrit will be flat.
This can be understood as a prediction of General Relativity. In GR, as John Wheeler famously said, “spacetime
tells matter how to move; matter tells spacetime how to curve”. The greater the density of matter/energy in our
spacetime, the more spacetime will curve around it. So, universes with a high density have positive curvature, k > 0,
i.e. there is enough matter that they curve back in on themselves (closed universe). Universes with a low density have
negative curvature, k < 0 (open universe). And universes where the density is ‘just right’, i.e. exactly balanced to equal
ρcrit , are flat (k = 0).
The critical density of the Universe today, ρcrit,0 = ρcrit (t0 ), is a useful reference value in cosmology. If we ignore
the cosmological constant (or write it as a mass/energy density and add it into ρ), the Friedmann equation today in a
flat (critical density) universe becomes
8πG
H02 = ρcrit,0 . (22)
3
We can rearrange this to get an expression that allows us to calculate the critical density for a given expansion rate,

3H02
ρcrit,0 ≡ . (23)
8πG
Let’s see how we can use the critical density as a reference value. Instead of working with all of the different dimension-
ful prefactors for each term in the Friedmann equation, we can rewrite the whole equation in a (mostly) dimensionless
way by dividing through by the critical density:
H 2 (t) 8πGρ/3 kc2 /a2
= − (24)
H02 8πGρcrit,0 /3 H02
ρ kc2
= − 2 2 (25)
ρcrit,0 a H0

Note how we chose to divide the curvature term by H02 instead of 8πGρcrit,0 /3 because it results in a neater expression.
We have also continued to treat the Λ term as an energy density, ρΛ , so we can hide it away inside the ρ term. Next,
let’s evaluate this expression at t = t0 . We get:

H 2 (t0 ) ρ(t0 ) kc2

= 1 = − . (26)
H02 ρcrit,0 H02

Let’s define the fractional energy density,

ρ
Ω= . (27)
ρcrit,0

Using this definition, we can write the equation from above as

kc2
Ω0 − = 1. (28)
H02

It then makes sense to define an analogous parameter for the curvature,

kc2
Ωk ≡ 1 − Ω0 = − . (29)
H02

In other words, Ωk is the fractional difference between the total energy density and the critical density. Note how Ωk is
positive when k is negative and vice versa.

19
3.4. Change in energy density as space expands
We saw in Section 1 that there are many different types of matter and radiation that contribute to the total energy density
of the universe. As you might expect, their density typically decreases as the universe expands – because distances
stretch according to the scale factor, a(t), volumes increase by a factor a3 .
The way that the density changes depends on the type of energy we are dealing with. The simplest to understand
is regular matter, like atoms, stars, galaxies and so on. In the later stages of cosmic history, matter is mostly conserved
(neither created nor destroyed), and so the total mass of the different types of matter is constant. As the matter expands
with the expansion of space, however, the volume that it takes up increases, like a3 . The density of matter therefore
scales as

ρm ∝ 1/a3 . (30)

The density of cold dark matter is also expected to scale in the same way.
Radiation, like photons and highly relativistic particles, behaves differently. Its energy density is governed by its
temperature, T , scaling as ρr ∝ T 4 . Taking a typical photon of temperature T , its energy is related to its wavelength
by kB T ≈ hc/λ. We know how wavelength scales with the expansion of the universe (λ ∝ (1 + z) ∝ 1/a), and so we
can see that the energy density of radiation should scale as

ρr ∝ 1/a4 . (31)

There are other types of energy density that scale differently to matter and radiation, which we will discuss in later
sections. But for now, matter and radiation, plus curvature and the cosmological constant, cover most of the possibilities
that we will be interested in.
We can define a fractional energy density for any type of matter or energy that we like. For example, the fractional
energy density of matter today is defined as Ωm = ρm /ρcrit,0 . Since we have normalised a = 1 at t = t0 , we can write

ρm ∝ Ωm a−3 . (32)

Similarly, for radiation,

ρr ∝ Ωr a−4 . (33)

Note how the matter and radiation energy density, ρm and ρr , depend on a(t), which depends on t – their energy
densities are time-dependent quantities. The fractional energy densities Ωm and Ωr are not time-dependent however –
they are both evaluated at t = t0 , and so are constant.
Recall the definition of Ω0 as being the total fraction of energy density in all types of matter and radiation at t = t0 .
We can break this up into a sum of matter and radiation parts, plus a cosmological constant, plus the curvature term, to
obtain

Ωm + Ωr + Ωk + ΩΛ = 1. (34)

This lets us rewrite the Friedmann equation in a simpler form, with only a single dimensionful parameter, H0 :

H 2 (a) = H02 Ωm a−3 + Ωr a−4 + Ωk a−2 + ΩΛ .

(35)

This is a very useful form for the Friedmann equation, as we only need to remember the fractional densities of matter,
radiation etc. You should remember this form of the equation!

3.5. Matter-only solution to the Friedmann equations

Now we are in a position to solve the Friedmann equation for different compositions of the universe, to see how it
affects the scale factor, a(t). We’ll start with a simple one, involving a flat universe that only contains matter.
First, we write down the Friedmann equation with only a matter component.
 2
ȧ
= H02 Ωm a−3 . (36)
a

20
Since this universe only has matter, Ωm = 1. Taking the square root of both sides of the equation,
1 da 3
= ±H0 a− 2 , (37)
a dt
we can then shuffle all terms that depend on a to one side to obtain
3
a 2 da 1
H0 dt = ± = ±a 2 da. (38)
a
We now have everything we need to solve the equation – the left-hand side depends only on t, and the right-hand side
only depends on a, so all we have to do is integrate both sides with the appropriate limits. Logically, we can start our
integration at the Big Bang, where the scale factor is 0. What is the corresponding time coordinate of the Big Bang?
Actually, we are free to choose – as long as we are consistent, we can choose whatever zero point for the time we like.
We can even label the Big Bang as being at a negative time, so ‘today’ is at t = 0. For simplicity, we’re going to choose
the origin of the time coordinate, t = 0, to be at the Big Bang though. Our other integration limits will be the value of
the scale factor a at some time t, so we can write
Z t Z a
1
H0 dt0 = ± (a0 ) 2 da0 , (39)
0 0

where we have added primes to the integration variables to avoid confusion with the integration limits. Both integrals
are straightforward; we obtain
2 3 a
h it  
H0 t = ± a 2 . (40)
0 3 0

Inserting the integration limits, and taking the positive branch of the solution, we obtain
2 3
H0 t = a 2 (41)
3
 2
3H0 t 3
=⇒ a = . (42)
2
This is the solution for the scale factor in a matter-dominated universe.

3.6. Interchangeability of time, redshift, and scale factor

In a universe that is expanding, the scale factor a always increases with time. Since a(t) is monotonically increasing,
the inverse relation t(a) is also monotonically increasing, and so there is a unique value of a for every t and vice versa.
This means that we can use the scale factor as an alternative time coordinate if we like.
Since the relation a(z) = 1/(1 + z) is also monotonic, this means that we can use the redshift as a time coordinate
too (where increasing z implies that we are looking further back in time).
Not all cosmological models continue expanding forever, and so we can’t always use a and z as alternative time
coordinates. But since most of the universes we will be studying are continuously expanding, we will often find this
property useful.

3.7. Age of the Universe

Given a cosmological model, we can use the Friedmann equation to calculate the age of the Universe (the time between
the Big Bang and today). We start by taking the square root of the Friedmann equation to obtain
1 da
= ±H(a). (43)
a dt
Rearranging and then integrating, we get
Z 1 Z t0
da da
= dt =⇒ = dt (44)
aH 0 aH 0
Z 1
t0 da
=⇒ [t]0 = t0 = . (45)
0 aH

21
Note the integration limits. The scale factor integral runs from a = 0 (the Big Bang) to a = 1 (the scale factor today).
The time integral runs from the corresponding times: t = 0 for the Big Bang, and t = t0 for today. We have also
chosen the positive sign from the square root. Now all we need is to plug in a model for the Hubble function, H(a),
and evaluate the final integral.
3
For a purely matter-dominated Universe, H(a) = H0 a− 2 . The integral becomes
Z 1 Z 1
da 1 1 2 3 2
3 = a 2 da = [a 2 ]10 = (46)
0 a × H0 a 2 − H0 0 3H0 3H0

3.8. Matter, curvature, and the fate of the Universe

We discussed above how universes with higher or lower energy densities than the critical density will be closed (pos-
itive curvature) or open (negative curvature). Let’s now study a universe containing only matter and some amount of
curvature to see how this affects the scale factor.
The Friedmann equation in this kind of universe can be written as

H 2 (a) = H02 Ωm a−3 + Ωk a−2 .

(47)

Taking the square root, cancelling some factors of a, and rearranging, we obtain
da
H0 dt = ± p (48)
a Ωm a + Ωk a−2
−3

da
= ±p . (49)
Ωm a−1 + Ωk
A solution to the integral of the right-hand side is not looking particularly obvious as it is currently written. But we can
rearrange further to obtain
 1
1 a 2
H0 dt = ± √ da. (50)
Ωm 1 + (Ωk /Ωm )a
Now we need to use some intuition! Integrals of this form often have trigonometric functions as their solutions. We
must also be careful of the sign of Ωk , since that can change the sign in the denominator and therefore the solution to
the integral. While there is an analytic solution to this equation, it is very ugly! It is also difficult to invert, so we can’t
write down the function a(t), which is what we really want.
A more elegant solution can be obtained by considering a parametric solution, where we rewrite the integral in
terms of a proxy parameter that we can then separately relate to t. The conformal time happens to be a useful parameter
for this purpose. Recall that the conformal time, τ , is defined through the relation dt = adτ . We now have two separate
integrals to do:
 1
1 1 2
H0 dτ = ± √ 2
da (51)
Ωm a + (Ωk /Ωm )a
dt = adτ. (52)
Here are the solutions. For a closed universe (k > 0):
a(τ ) ∝ (1 − cos τ ) (53)
t(τ ) ∝ (τ − sin τ ). (54)
For an open universe (k < 0):
a(τ ) ∝ (cosh τ − 1) (55)
t(τ ) ∝ (sinh τ − τ ). (56)
You can verify that these are solutions to the Friedmann equation by differentiating them.
Let’s inspect these solutions. First, let’s note that conformal time is a parameter that we can continue to increase,
and then see what happens to the scale factor and time coordinate as we do so. It is most informative to then plot a vs
t, ignoring the fact that they are both really functions of τ .

22
 The plot below shows the typical behaviour of the solutions a(t) for open, closed, and flat universes. The open
universe continues to expand, and never stops expanding. The flat universe also continues to expand, but eventually
(asymptotically) stops expanding. The closed universe is the most interesting – it expands up to a point, and then starts
to recollapse! This is called a ‘Big Crunch’ and is sometimes described as being like the Big Bang but in reverse. It
would be very spectacular if it happened to our Universe!
Note that even a small amount of positive or negative curvature can completely change the future fate of the
universe. A matter-dominated universe that is only slightly above the critical density would have a radically different
outlook to one that is only slightly under. What is the fate of our Universe? Are we destined for a Big Crunch, or a
lonely eternal expansion as all other galaxies progressively recede from view? Our current best measurement of the
curvature is Ωk . ±10−2 . In other words, it’s very close to flat, but might be non-zero. And if it is non-zero, we don’t
know if it’s positive or negative!
All three types of universe are decelerating – the expansion is slowing down with time, as you can see from the
changing gradients in the plot. We will see in the next section how adding a cosmological constant changes everything
by causing the expansion to accelerate, changing the fate of the universe completely.

1.4 Closed
Open
1.2 Flat
1.0

0.8
a(t)

0.6

0.4

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
t [arb. units]
Figure 9: Solutions for a(t) for universes with matter plus three different types of curvature.

3.9. Newtonian derivation of the Friedmann equation

The Friedmann equation is a general relativistic result, but it turns out that we can get to the same expression by thinking
about the homogeneous and isotropic expansion of matter with Newtonian equations of motion.
First, consider a thin shell around an arbitrary point (it can be any point) in a space filled with a homogeneous
distribution of matter. Label this point with a radial coordinate r = 0 and place the shell at a radius of r = R. Now,
what are the forces acting on this shell?
We know from Gauss’ Law in Newtonian mechanics that the matter outside the shell has no effect on it – considering
concentric shells of progressively larger radius cancel out, the forces from opposite points on those shells cancel each
other out. Only matter inside the shell in question exerts a net force. This force is the same as if the matter inside the
shell was concentrated into a point mass at the centre.
The total mass within the shell is
4
M (< R) = πR3 ρ, (57)
3

23
for a density ρ. The force acting on a small segment of the shell (mass m) is then

GmM (< R)
F =− . (58)
R2
We can calculate the acceleration of this small segment. To avoid confusion with the scale factor, let’s write the
acceleration as the second derivative of the radial position (R),

d2 R GM (< R)
2
=− . (59)
dt R2
So, each segment of the shell should be accelerating towards the centre.
Now, we can integrate this equation to find the velocity of the shell and so forth. The right-hand size depends only
implicitly on t, since R is a function of t, which makes it a bit fiddly to do the integral. There is a neat trick we can use
to help, though. First, multiply both sides by dR/dt:

dR d2 R GM (< R) dR
=− . (60)
dt dt2 R2 dt
Then, integrate:

dR d2 R
Z Z
GM (< R) dR
2
dt = − dt. (61)
dt dt R2 dt
The right-hand side can be integrated as follows:
Z Z
GM (< R) dR GM (< R) GM (< R)
− = − dR = + . (62)
R2 dt R2 R

There is a subtlety with this result – we have treated M (< R) as if it is a constant. That’s because it is constant if we
think carefully about what it means. Before, when we calculated M (< R), we just wanted to know how much mass
was within a particular radius (at fixed time). Now, we are tracking the mass within a particular shell of matter. As the
shell expands outwards, so does all of the material within it – the radius increases, the density decreases, but the mass
enclosed stays the same. In other words, the quantity R(t) that we are interested in is the radius of a shell containing a
constant mass. Other definitions of the radius would involve also tracking a flux of matter into or out of the shell; our
definition is unique in that it only tracks the same blob of matter at all times, no matter how much it collapses/expands.
The left-hand side can be integrated as follows. If we first write:

dR d2 R
Z Z  
dR d dR
dt = dt, (63)
dt dt2 dt dt dt

we can make the replacement u = dR/dt, to obtain

Z Z
du 1
u dt = udu = u2 + C. (64)
dt 2
Substituting the definition for u back in and equating to the result for the right-hand side, we obtain
 2
1 dR GM (< R)
= + C. (65)
2 dt R

This is an ‘energy equation’; the LHS looks like a kinetic energy term (proportional to ‘velocity’ squared), while the
RHS looks like the gravitational potential energy per unit mass. With this interpretation, we can identify the integration
constant C as the total energy of the system.
Now, substituting in the expression for the mass enclosed within the shell, we obtain
2
4π R3 ρ

1 dR
=G + C. (66)
2 dt 3 R

24
We can evaluate R with respect to some reference radius, R0 , at time t0 . Making the substitution R(t) = a(t)R0
(where a(t0 ) = 1) and rearranging, we obtain
 2
1 da 8πG C
= ρ(t) + 2 2 . (67)
a dt 3 a r0

We can now identify the total energy C with the curvature parameter multiplied by an overall distance/time-scale,
−kr02 /c2 . The rest of the equation is already in the same form of the Friedmann equation, with ρ(t) ∝ R−3 ∝ a−3 as
expected.

Understanding:

X What is the Friedmann equation?

X What does each term in the Friedmann equation mean, and what are the relevant units?
X What is the critical density and how can it be used to rewrite the Friedmann equation?
X How do the densities of matter and radiation depend on the scale factor?
X How can you solve the Friedmann equation to find the scale factor as a function of time?
X How can you solve the Friedmann equation to find the age of the Universe?
X How does curvature affect the time evolution of the scale factor?

25
4. Distances and horizons
In this section, you will learn about the comoving distance that light travels from an object observed at some at redshift;
how to calculate physically-meaningful distances in cosmology based on standard luminosities and sizes; how these
distances are defined with respect to the comoving distance; and the concept of a cosmological horizon.

Reading for this topic:

– An Introduction to Modern Cosmology (A. Liddle), Chapters 3, 6, and 7.

– An Introduction to Modern Cosmology (A. Liddle), Advanced Topic 2: Distances and Luminosities.

4.1. Cosmological distances

We have previously discussed how defining distances in cosmology is a fraught process, because space itself is con-
stantly changing (expanding), and because the distances are so large compared with how long it takes light to travel
across them. We discussed in Section 2 how the FLRW metric can be used to convert from comoving distances to
proper distances, and what the interpretation of those distances was. Neither was directly observable, however; they
were both mathematical constructs that are useful for keeping track of things in our calculations.
In this section, we will learn about a few different types of distance that are much more closely related to things we
can actually observe. All of them will turn out to be defined with reference to light rays that have travelled from distant
objects.

4.2. Distance travelled by a light ray

All of our calculations in this section depend on the distance travelled by a light ray. Consider a photon emitted from
a galaxy at some redshift, z. It is eventually observed by us, the observer, at z = 0. We know that the photon must
have been emitted a long time ago; it takes light a long time to travel across the vast distances in our Universe, and that
distance is also increasing, due to the cosmic expansion.
One way of measuring the distance travelled by the photon is to keep track of the comoving distance it has traversed
between being emitted and being detected by us. We call this distance r(z). It is uniquely defined for any given redshift;
for a redshift z, we only see the photons that have travelled a comoving distance of r(z), i.e. photons emitted by galaxies
at a comoving coordinate of r away from us. Photons emitted at the same time by galaxies that are further away in
comoving coordinates than r have not had chance to reach us yet, while those emitted at the same time by galaxies
closer to us would have already reached Earth a long time ago (so we missed them). Hence, for a given redshift, we
only see photons that were emitted from a comoving distance r(z) away from us.
Recall that the space-time distance travelled by light rays is ds = 0. We can use the line element for an FLRW
metric (assuming a flat universe), subject to this condition, to write down a relation between comoving coordinate of a
photon and another variable (in this case, time):

ds2 = −c2 dt2 + a2 dr2 = 0 (68)

=⇒ cdt = ±adr. (69)

Using the Friedmann equation, we can write da/dt = ±aH and substitute for dt, giving
da
dr = ±c (70)
a2 H Z
r a
c da0
Z
=⇒ r(a) = dr0 = − , (71)
0 1 (a0 )2 H(a0 )
where we have chosen the negative sign so that the distance comes out positive (this make sense; the light ray is
travelling from a coordinate of r to reach us at a coordinate 0, so the order of the integration limits should be flipped
unless a minus sign is used).

26
4.3. Luminosity distance and standard candles
The comoving distance travelled by a photon isn’t something that we can measure directly – there are no properties of
the photon that we can measure that would tell us how far it has moved since it was emitted. There are closely-related
quantities that are measurable, though.
Consider light emitted from a distant object that has luminosity, L. Let’s also assume that the light is emitted
isotropically from the object. The light will therefore travel out from the object in a spherical shell. As the size of the
shell increases, the light will be diluted, making the object appear fainter and fainter from greater and greater distances.
What we can measure is the flux of light coming from the object, f , as we see it from a large distance away. The flux
and luminosity are related to each other by the simple relation
L
f= , (72)
4πd2L
where dL is the radius of the shell that corresponds to how much the light has been diluted.
We call dL (z) the luminosity distance to an object observed at redshift z. It is the distance we would infer if we
knew the intrinsic luminosity of the source, measured its flux, and used the simple relation above. The derivation is
beyond the scope of this module, but it can be shown that

dL (z) = (1 + z) r(z), (73)

i.e. the luminosity distance is just the comoving distance travelled by photons, scaled by an additional factor of (1 + z)
(this is related to the fact that the energy, and therefore flux, of the photons is also diluted by the expansion of space).
Note that the redshift z here is the true cosmological redshift, and does not include peculiar velocities.
How can we measure the luminosity distance to an object at redshift z? Well, we can measure z and the flux directly
using a telescope. The last remaining ingredient is the intrinsic luminosity, L. Certain types of astronomical objects
have luminosities that we can figure out by inspecting some of their other properties. These are called standard candles.
One such type of object are the Cepheid variable stars studied by Henrietta Swan-Leavitt; these exhibit a relationship
between the period at which they pulsate vs their intrinsic luminosity that can be calibrated using measurements of
very close-by Cepheids. If we then measure the period of pulsation of very far away Cepheids by making repeated
measurements of the flux over time with a telescope, and then use the period-luminosity relation to derive their intrinsic
luminosity L, we can then infer dL .

4.4. Distance ladder

Other types of standard candle include Type Ia supernovae, which have very predictable luminosities when their pro-
genitor stars explode. The luminosity can be measured from properties such as the colour of the supernova, and how
long it takes for its light to fade. Type Ia supernovae have the advantage of being extremely bright, so we can see them
from extremely large distances (even thousands of megaparsecs away). Cepheid stars are much fainter, however, and
can only be detected in relatively nearby galaxies.
Supernovae are quite rare however, and so we don’t have many examples from the local universe, where we would
be able to calibrate the relations that tell us their luminosity. Instead, we have to construct a distance ladder. The idea
is that we start with fainter, nearby types of standard candle (like Cepheids) that we can calibrate quite accurately. We
then look for another type of standard candle that is brighter (but generally rarer) that we can see at larger distances. The
game is to find galaxies that contain more than one type of standard candle, so we can use the already well-calibrated
one to work out the calibration for the other. To calibrate the Type Ia supernovae, a distance ladder with several ‘rungs’
is needed!
Further reading: Cosmic distance ladder (Wikipedia).

4.5. Angular diameter distance

Instead of using standard candles, we can also use standard rulers, i.e. objects that we know the intrinsic size of. We
can measure the angular sizes of objects like galaxies, again by taking images of them with telescopes. If we know
their intrinsic (proper/physical) size d, we can then infer a distance, dA , using the definition
d
θ≈ , (74)
dA

27
where we have used the small-angle approximation, tan θ ≈ θ, since most astronomical objects subtend a very small
angle on the sky. The distance dA is called the angular diameter distance.
The angular diameter distance is also related to the comoving distance, but in a different way to the luminosity
distance:
r(z)
dA (z) = . (75)
(1 + z)

We can now relate the angular diameter distance to the luminosity distance, finding that

dL (z) = (1 + z)2 dA (z), (76)

where the redshift z is the true cosmological redshift, and not the ‘observed’ redshift that includes peculiar velocities.
The angular diameter distance has a particularly counter-intuitive property. At small to intermediate redshifts, it
increases with redshift, which means that objects of fixed physical size subtend smaller and smaller angles as they get
further and further away. But for sufficiently large redshifts, the angular diameter distance begins to get smaller again!
So, in an expanding universe, objects that are very far away (in terms of their comoving distance from us) can actually
have a larger angular size than objects that are closer! This is just a projection effect caused by the expansion of the
Universe, but it’s still quite surprising.

4.6. Cosmological horizon

In the previous subsections, we learned about several different ways to define cosmological distances. Some, like the
angular diameter distance, were defined according to how we make observations of distant objects. Others, like the
comoving distance, were mathematical constructs that were useful for keeping track of coordinates.
Another useful type of distance is a horizon. It defines how far a particle or wave could have possibly travelled
since it was emitted. Or, in other words, it is the maximum radius out to which some physical process could have
influenced other events.
Horizons are very useful in cosmology. By measuring the size of a horizon, we can figure out how long a physical
process must have been going on for the horizon to have reached that size. Different physical processes have different
horizon sizes, and so by observing the Universe on distance scales longer and shorter than a given horizon, we can
disentangle the different physical effects that must have been going on to cause the structures that we see.
A good example of a horizon is the particle horizon, rH . This is a very fundamental distance in cosmology – it is
the furthest (comoving) distance a particle (travelling at the speed of light or less) could have possibly travelled since
the Big Bang. No physical process can have any effect over distances greater than the particle horizon; we say that
regions of the Universe separated by a distance greater than the particle horizon are causally disconnected, since there
is no way the two regions could ever have been in contact via a causal physical process.
To calculate the particle horizon, we need to calculate the comoving distance that could have (in principle) been
covered by a light ray since the Big Bang. Using the definition of the comoving distance travelled by light, r(z), we
can write
c da0
Z 1
rH = r(z → ∞) = 0 2 0
, (77)
0 (a ) H(a )

where we have integrated from our current value of the scale factor (a = 1) to the scale factor at the Big Bang (a = 0).
This is the size of our particle horizon today.
How big was the particle horizon at some time in the past? If we look at an object at a redshift z from us, their
particle horizon is

c da0
Z a
rH (z) = 0 2 0
, (78)
0 (a ) H(a )

where we have integrated back from their scale factor, a = 1/(1 + z), until the Big Bang at a = 0.
Another type of horizon is the Hubble radius, rHR . This is the approximate size of the observable Universe at a
given time. We can define this using the Hubble Law, v = H0 d (which, remember, is only an approximation at low

28
redshift, so this result will also be approximate). If we set the recession velocity vrec = c, we obtain d = c/H0 .
Replacing d with a comoving distance, d = arHR , and generalising to the expansion rate at any scale factor, H(a),
c
rHR ≈ . (79)
aH
This quantity is not strictly a horizon – we can see light that has travelled from further comoving distances than this –
but it does correspond to the approximate distance over which causal processes can operate at any given time. In other
words, only objects with a comoving separation r < rHR can have any significant physical interaction between them.
Further reading: Cosmological horizon (Wikipedia).

Understanding:

X What is the comoving distance, r(z)?

X How is the luminosity distance defined?
X How is the angular diameter distance defined?
X How are the luminosity distance and angular diameter distance related?
X How can the different types of distance be measured?
X What is the particle horizon? How can it be calculated?
X What is the Hubble radius or Hubble horizon?

29
5. Cosmic acceleration
In this section, you will learn how the expansion of the Universe can be accelerating or decelerating, depending on what
kind of matter it contains; how to measure acceleration using the deceleration parameter; how a universe dominated
by the cosmological constant behaves; how the acceleration caused by a cosmological constant changes the size of the
cosmological horizon; and what the likely fate of our own Universe is.

Reading for this topic:

– An Introduction to Modern Cosmology (A. Liddle), Chapters 3, 6, and 7.

5.1. Conservation equation

As we saw in previous sections, the density of different types of matter/energy scales with the scale factor in different
ways; (non-relativistic) matter scales as ρm ∝ a−3 and (relativistic) radiation scales as ρr ∝ a−4 for example. We
also saw that universe predominantly filled with different types of matter/energy expanded in different ways; the mix
of matter, radiation, and curvature could quite radically change how the scale factor, a(t), changes with time.
There is a useful equation that relates the time evolution of the density of a particular type of matter/energy to
the time evolution of the scale factor. This is called the conservation equation, as it is derived from the stress-energy
conservation equation of General Relativity. It can be written as:
ȧ  p
ρ̇ = −3 ρ+ 2 . (80)
a c
Note how this equation depends on the expansion rate (ȧ/a = H), as well as the density ρ and relativistic pressure p.
Importantly, this equation holds separately for each type of matter/energy (as long as they are not interacting with each
other). So, in a universe that contains both matter and radiation, we can write down a separate conservation equation
for ρm and ρr , for example, with the same expansion rate, but different densities and pressures.
What do we mean by relativistic pressure? This is not quite the same as thermal pressure. First of all, we normally
think of pressure as exerting an outward force (e.g. to keep stars from collapsing in on themselves), and so would
expect it to counteract gravitational attraction somehow. This is what the thermal pressure does, but it’s not really what
the relativistic pressure in this equation describes. In fact, something with lots of relativistic pressure (p > 0) has a
stronger gravitational pull! This is because pressure increases the amount of energy in a region, which causes a stronger
space-time curvature (and so greater gravitational attraction).
Second, the thermal pressure of most regular (baryonic) matter and dark matter is very small compared to its rest
mass, and so the density ρ dominates. The pressure can become larger in extreme systems such as the dense interiors
of very massive stars, but this actually makes them more likely to collapse in on themselves, not less! Again, this is
because relativistic pressure increases the gravitational attraction.
Instead, we can think of the relativistic pressure as a separate quantity that is characteristic of a particular type of
matter/energy, and determined by its equation of state.

5.2. Equation of state

Different types of matter/energy have different characteristic amounts of relativistic pressure compared to their energy
density. The relative amounts of pressure and density are encoded in the equation of state parameter w for that particular
type of matter energy,

p = wρc2 . (81)

We can work out the equation of state for matter and radiation by substituting this relation into the conservation equation
and solving for w,
ȧ ȧ
ρ̇ = −3 (ρ + wρ) = −3 ρ (1 + w) . (82)
a a

30
For matter, we can plug-in ρm = ρm,0 a−3 to obtain

ȧ
−3ρm,0 a−4 ȧ = −3 ρm,0 a−3 (1 + w) , (83)
a
where we have used dρm /dt = (dρm /da)(da/dt) = ȧ(dρm /da). Cancelling common factors from both sides, we get

1 = (1 + w) =⇒ w = 0, (84)

so the equation of state for matter is w = 0. This is what we expected – non-relativistic matter should have negligible
relativistic pressure, p ≈ 0, which implies w = 0. Repeating a similar exercise for radiation, we find w = 1/3.
More generally, we can rearrange the conservation equation to put all the factors of a on one side and ρ on the
other:
ρ̇ ȧ
= −3 (1 + w) . (85)
ρ a
Now, if we integrate with respect to time:
Z Z
1 dρ 1 da
dt = −3 (1 + w) dt (86)
ρ dt a dt
Z Z
dρ da
=⇒ = −3 (1 + w) . (87)
ρ a
We can solve both sides of the equation once we know the equation of state. Note that the equation of state can, in
general, depend on scale factor too, although for matter and radiation it is constant.

5.3. Cosmic acceleration and deceleration

The equation of state of different types of matter and energy is interesting because it determines how their densities
scale with scale factor. It also helps determine whether the expansion of the universe is accelerating or decelerating.
We have used the Friedmann equation and the expansion rate H = ȧ/a as a measure of whether the Universe is
expanding. Very simply, if ȧ > 0, it’s expanding. In the Big Crunch example, we saw that some types of universe can
collapse too (ȧ < 0).
The next question to ask is whether the expansion is getting faster with time (accelerating), or slowing down
(decelerating); in other words, whether the scale factor has a positive second derivative, ä > 0 (accelerating) or a
negative second derivative, ä < 0 (decelerating). Note that ä > 0 does not imply that the expansion rate H is getting
bigger with time! Recall that H = ȧ/a. While it’s true that ȧ gets bigger with time in an accelerating Universe, a is
also getting bigger!
General Relativity provides us with another equation that we can use to calculate the acceleration/deceleration of
the cosmic expansion. It is called the Raychaudhuri equation,
 2
ä ȧ kc2 8πG
2 + + 2 − Λc2 = − 2 p. (88)
a a a c

Note how it depends on the relativistic pressure, p, that we saw in the conservation equation, as well as some other
terms (e.g. curvature, cosmological constant) that are familiar from the Friedmann equation. In fact, the Raychaudhuri
equation can be derived by combining the Friedmann and conservation equations.

5.4. Deceleration parameter

Before anyone had measured the parameters of our cosmological model very precisely (in the 1960’s, 70’s, and 80’s),
astronomers tried to at least get some sense for how the scale factor was evolving with time given the imprecise data
they had available. Instead of working with different FLRW models, they used a Taylor expansion of the scale factor
instead, which allowed them to fit a very simple model with only a couple of free parameters to data such as the
luminosity distance (measured as a function of redshift).

31
 Recall the definition of a Taylor expansion from Section 2. Expanding around the scale factor today, a(t0 ) = 1, we
can write a Taylor expansion of the scale factor in increasing powers of t − t0 . For some reason, astronomers at the
time decided to add to this expansion by introducing factors of H0 , to obtain
1
a(t) ≈ 1 + H0 (t − t0 ) − q0 H02 (t − t0 )2 + . . . (89)
2
Recall that the dots above the factors of a denote time derivatives, e.g. ȧ = da/dt. The coefficient of the linear term
in this Taylor expansion is just the Hubble constant, H0 , which we have already seen is equal to the first derivative of
the scale factor, divided by a, and evaluated at t = t0 . The next term involves H02 , and another coefficient, q0 . This is
called the deceleration parameter, and is used to determine whether the expansion of the universe is slowing down or
speeding up. Also for historical reasons, astronomers decided that the deceleration parameter would be defined with a
minus sign in front of it, so q0 > 0 means that the universe is decelerating, while q0 < 0 means that it is accelerating.
The deceleration parameter can also be defined as a function of scale factor or redshift,
!
 a 2 ä Ḣ
q(a) = − =− 1+ 2 . (90)
ȧ a H

Matter- and radiation-dominated universes are always decelerating, regardless of whether they are open, closed, or flat,
and so they have a positive deceleration parameter, q > 0. But other, more exotic, types of matter/energy can cause the
expansion to accelerate, as we will see in the next subsection.
Historically, the deceleration parameter was important in the discovery that the expansion of our Universe is actually
accelerating (a discovery that won the Nobel Prize in 2011). In the 1990s, it was becoming apparent that the Universe
had much less mass than would be required for it to be flat (Ωm ≈ 0.3, instead of Ωm ≈ 1). This could have just
meant that Ωk ≈ 0.7, i.e. that we live in an open universe. This type of universe would still be decelerating, however.
Measurements of the luminosity distance, made with distant Type Ia supernovae, were used to accurately measure the
deceleration parameter for the first time, showing that q0 < 0 in our universe – it’s accelerating! Combined with the
observation that the matter density is around Ωm ≈ 0.3, this led to the conclusion that around 70% of the energy density
of our Universe is made up of a cosmological constant, Λ – or perhaps something even more unusual.

5.5. Properties of the cosmological constant

The cosmological constant, Λ, is an unusual contribution to the total energy density of the Universe. First of all, it stays
constant with time! It does not dilute as the Universe expands, as was the case with matter and radiation. If we use this
information to solve the conservation equation (by setting ρ = const., which implies ρ̇ = 0), we find that the equation
of state for the cosmological constant is w = −1. Since p = wρc2 , this implies that the cosmological constant has a
negative relativistic pressure! This is quite unusual, but the upshot is that the cosmological constant causes a kind of
gravitational repulsion rather than attraction, causing the expansion of the universe to speed up rather than slow down.
The density of the cosmological constant has remained constant in time, while other components like matter and
radiation have decreased in density with time. It therefore stands to reason that, back in the past, these other components
must have had much higher densities than the cosmological constant. Since radiation- and matter-dominated universes
are decelerating (see above), this implies that the deceleration parameter must have been positive in the past, but
has become negative now that the cosmological constant is the dominant form of energy density. We say that the
accelerating expansion of the Universe is a late-time phenomenon, that has only started occurring in the Universe
‘recently’ (that is, within the past few billion years...)

5.6. Cosmological constant solution

Now let’s consider a flat universe that only has a cosmological constant, ΩΛ = 1. This is called a de Sitter space-time.
The Friedmann equation reduces to
 2
ȧ
= H02 ΩΛ = H02 , (91)
a
since ΩΛ = 1. Rearranging, we get
da
H0 dt = ± . (92)
a

32
Integrating from some reference value of the scale factor (a = a∗ at t = t∗ ), we find
Z t p Z a
da
H0 ΩΛ dt = ± (93)
t∗ a∗ a
=⇒ H0 (t − t∗ ) = ± (log a − log a∗ ) (94)
= ± log(a/a∗ ). (95)

Exponentiating both sides and rearranging once more, we get

a(t) = a∗ e±H0 (t−t∗ ) . (96)

The appropriate choice of sign in the exponent to get an expanding universe is the positive sign. Also, if we want to
have a = 1 at t = t0 , we can set a∗ = 1 and t∗ = t0 , so we have the solution

a(t) = eH0 (t−t0 ) . (97)

This is an exponential expansion of space, and gives rise to some interesting properties. We have already seen from the
Friedmann equation that the expansion rate, H, is constant in time. If we write the cosmological constant in the form
of an energy density, ρΛ , we can see that this is constant too. So, as the universe expands, the energy density doesn’t
dilute (as with matter or radiation) – it stays the same! This seems quite unintuitive – as more space is created by the
expansion, more energy is created too.
Doesn’t it violate the law of energy conservation is the universe continually ‘creating’ energy? See the ‘Energy
conservation’ box below for more discussion of energy conservation – it turns out that energy is not conserved in an
expanding universe (but a more general quantity that involves energy is conserved). We will study this curious result in
more detail in the next sub-section.

5.7. Age and Hubble horizon in an exponentially-expanding space-time

Another curious feature arises when we try to work out the age of this universe. If we set a = 0 (denoting the Big
Bang) and solve for t = tBB , we get H0 (tBB − t0 ) = log 0 = −∞. So the age of the universe, t0 − tBB → ∞! A
universe with only a cosmological constant (i.e. not even a tiny amount of any other kind of matter or energy) has no
Big Bang – it is an eternal universe that has been expanding forever and will continue to expand forever.
For our next curious feature, let’s work out the Hubble horizon, rHR = c(aH)−1 . We obtain
c c −H0 (t−t0 )
rHR (t) = = e . (98)
eH0 (t−t0 ) H0 H0
The Hubble horizon is getting smaller with time! That is, as the universe expands, physical processes are only able to
operate across smaller and smaller comoving distances. Distant parts of the universe are progressively falling out of
contact with each other. This is the opposite of what happens in (e.g.) matter-only and radiation-only models, where
the Hubble horizon continues to expand.

5.8. The Cosmological Constant problem

Where does the energy for the cosmological constant come from? Strictly, the cosmological constant is just a term that
crops up in the Friedmann equation when you derive it mathematically, sort of like an integration constant. You could
think of it as an inherent property of space-time, in the same way that curvature is an inherent property of space. It is
zero in some types of universe and non-zero in others.
Another common interpretation of the cosmological constant is that it represents the vacuum energy or zero-point
energy of the universe. Recall the harmonic oscillator example from quantum mechanics, which has energy levels
En = ~ω(n + 21 ). Its zero-point energy, at the lowest energy level (n = 0), is non-zero (E0 = ~ω/2). This is an
example of a system that has a vacuum energy – even when it is ‘empty’ (i.e. in its ground state), there is still some
energy there.
Our Universe isn’t described by the harmonic oscillator, but it is described by the Standard Model of Particle
Physics, which includes electromagnetism and the strong and weak nuclear forces. These forces are described within
the framework of Quantum Field Theory (QFT), which allows every time of field to have a possibly non-zero vacuum

33
energy. Each field has some kind of particle or particles associated with it, so there is an electron field, a photon field
etc, and each of these can have some vacuum energy (which may be positive or even negative). If we add all of the
vacuum energies from QFT together, we arrive at some number that we can interpret as the total vacuum energy of the
universe, ρvac .
What happens when we do this for our Universe? We get an absolutely huge number! The vacuum energy should
be so large that atoms and nuclei could never have formed, as the expansion rate in the early Universe would have
been too high. What’s more, we have measured the apparent vacuum energy density in our Universe using distant
supernovae and the luminosity distance-redshift relation. The measured value is around 60 orders of magnitude smaller
than what we expect from QFT!
This discrepancy between the expected vacuum energy and the observed value of Λ is called the Cosmological
Constant problem. It can be understood as a fine-tuning problem, where a parameter of our model must be tuned to a
very, very specific number in order for things to work out. Theoretical physicists tend to hate fine tuning – it suggests
that we don’t understanding something about how the parameters that describe our model are actually being chosen
(if they were being chosen randomly, how could they have possibly ended up with such specific values!). To see this
fine-tuning problem for the cosmological constant, consider that the observed cosmological constant is actually the sum
of the inherent value of Λ that space-time has (the ‘integration constant’ from GR, also called the ‘bare’ cosmological
constant), and the QFT vacuum energy,
8πG
Λobs = Λbare + ρvac = Λbare + Λvac . (99)
c2
Both have the same effect on the Universe, and so we can only ever measure their sum, Λobs , which we know to be
small. We can also calculate ρvac from QFT, which we find to be very large. This implies that the fractional difference
between the two terms is
|Λbare | − |Λvac |
≈ 10−60 , (100)
|Λvac |
i.e. that they are almost exactly the same, except for a tiny, tiny difference. Why would the bare cosmological constant
have anything to do with the QFT vacuum energy? Why would they be so similar? This puzzle is perhaps the biggest
in all of fundamental physics.

5.9. The fate of our Universe

The shrinking-horizon property of exponentially expanding solutions will turn out to be important when we discuss
cosmic inflation in later sections. It’s also important for understanding the fate of our own Universe, which has a large
fraction of its energy density in Λ (ΩΛ ≈ 0.7). As our Universe continues to expand, the matter and radiation energy
density will continue to dilute away, while the cosmological constant will remain the same. Eventually, the energy
density of the cosmological constant will completely dominate over all other forms of matter and energy, and we will
find a solution that is very, very close to the Λ-only (de Sitter) model that we have been studying. The Λ term even
dominates over the curvature term, so even if our Universe did have a non-zero positive or negative curvature, it would
still approach the de Sitter solution if we waited for a long enough time.
As the comoving Hubble horizon continues to shrink in the Λ-only model, we will eventually find ourselves cut
off from even the closest galaxies in the Hubble flow. Any light that could still reach us from distant galaxies would
have been redshifted by an extreme amount. Eventually, our galaxy will have exhausted all of its raw materials for
star formation, and the last stars will die. It is a very lonely, and seemingly inevitable fate! At least it will take tens
of trillions of years to get to that point – our Sun only has about 4.5 billion years of fuel left, so would have died
long before that point (and taken Earth with it). Even further into the future, it’s possible that any remaining matter
(e.g. neutron stars) could disintegrate due to proton decay, and black holes would eventually evaporate due to Hawking
radiation (about 10100 years into the future). The Universe would be left filled with a very low-density, low-energy, and
homogeneous background of photons and other stable fundamental particles that would hardly ever interact with one
another. Thinking about the thermodynamics of the Universe at this time, the entropy would be extremely high, and
there would be no free energy for any physical or chemical processes to occur. Apart from the continuing expansion
of space, the Universe would essentially be frozen. This future fate of the Universe is therefore often called the Big
Freeze, or alternatively the Heat Death of the Universe, because of the absolute thermodynamic equilibrium that
would be reached.

34
 Can anything save our Universe from this chilly (and very boring) fate? Well, it could be worse! If our Universe
is actually filled with a Dark Energy field, and not a cosmological constant, there could be other fates in store. One is
called a Big Rip, where the expansion becomes faster than exponential (this happens for a cosmological fluid with an
equation of state w < −1). This would eventually cause everything in the Universe to be ripped apart – even galaxies
and atoms, which aren’t in the Hubble flow. Depending on the value of w, this could even happen as soon as tens of
billions of years in the future.
Another possibility is false vacuum decay. It could be that the apparent ground state of our Universe is not truly
the lowest energy state – i.e. a false vacuum state – and that the matter/energy fields could spontaneously tunnel into a
nearby, lower energy state after enough time. Measurements of the Higgs field from the LHC suggest that it could be
in a metastable state, meaning that it could spontaneously find a lower-energy vacuum at some point in the future, but
this is far from certain. The false vacuum decay would be a very violent event, with a wave of the new vacuum state
ripping through our Universe at the speed of light and leaving completely reconfigured matter and energy fields in its
wake. The effect on QFT fields could be such that chemistry, nuclear bonding etc. would become impossible and all of
the matter would disintegrate instantly. Yikes.
Another, slightly more cheering, possibility is that the Big Freeze could eventually cycle through into the birth of
a new universe. This idea, called conformal cyclic cosmology, has been championed by Roger Penrose. It relies on
noticing that the metric of the Universe after the Big Freeze can be conformally transformed into the metric that we
would expect to have at very early times in our own Universe (during cosmic inflation; see later sections). The old
universe is in some sense mathematically equivalent a new universe, and so perhaps all of the same physical processes
would kick in again and generate matter and radiation again? It is a speculative idea, but has testable predictions.
Further reading: Future of an expanding universe (Wikipedia); False vacuum decay (Wikipedia); Vacuum decay:
the ultimate catastrophe (Cosmos Magazine); Conformal cylic cosmology (Wikipedia); Last Contact (short story about
the Big Rip by Steven Baxter)

Energy conservation
The Cosmological Constant seems to violate energy conservation! As the Universe expands, the total volume
gets bigger, but the energy density of the CC stays the same, implying that the total energy in the Universe
continually increases. This happens even in a universe with only Λ (no matter, radiation, or curvature).
In fact, energy conservation is violated in many cosmological models, even ones without Λ. As first proven
by Emmy Noether, conservation laws come from underlying symmetries in the laws of physics. Different
symmetries give rise to different conservation laws; for example, rotational symmetry of a system typically
gives rise to the conservation of angular momentum in that system.
The symmetry that gives rise to energy conservation is time-translation invariance. If the system follows laws
that do not change with time, energy will be conserved. This is not the case for most cosmological models
however, which are obviously changing with time due to the expansion (or contraction) of space.
Recall that, in relativity, we can transform between different space and time coordinate systems. There is nothing
special about any given space or time coordinate. Space and time together do follow certain laws however
(according to the principle of general covariance). We find that there are conservation laws associated with
this generalised symmetry, such as the conservation of stress-energy. The stress-energy tensor, Tab describes a
generalised form of energy and momentum (and other things), and follows the conservation law ∇a Tab = 0.
Energy density is the time-time component of the stress-energy tensor, T00 . Since this can be transformed to a
different quantity under a coordinate transformation, it’s clear that energy generally won’t be conserved in this
picture. The covariant quantity – the stress-energy tensor – is conserved though.

35
Understanding:

X What is the conservation equation?

X How can it be used to work out how the density of a type of matter/energy depends on scale factor?
X What is the Raychaudhuri equation?
X How does the equation of state relate energy density and relativistic pressure?
X What is the equation of state for matter, radiation, and a cosmological constant?
X What does it mean to have accelerating expansion?
X How is the deceleration parameter defined?
X What is the Cosmological Constant?
X What is the Cosmological Constant problem?
X How does the Hubble horizon behave during exponential expansion?

36
6. Cosmic Microwave Background Radiation
In this section you will learn about the Cosmic Microwave Background radiation, and how it formed when the hot,
dense ionised gas that initially filled the Universe cooled down and became transparent and neutral. You will also learn
about the frequency spectrum of the CMB, and the fact that it is very close to a perfect blackbody with a very low
temperature.

Reading for this topic:

– An Introduction to Modern Cosmology (A. Liddle), Chapters 10 and 12.

6.1. What was the early universe like?

Recall that in the hot Big Bang model, the early universe is denser and hotter than the late universe. The further back in
time we look (i.e. the higher the redshift), the higher the average temperature must have been. It stands to reason that,
if we look back far enough in time, the Universe must have been hot enough to ionise neutral atoms. In fact, above a
certain temperature, it must have been so hot on average that neutral atoms couldn’t have existed – all of the normal
(baryonic) matter in the Universe would have been ionised.
What are the implications of this? First of all, ionised gas is very good at scattering photons, especially if it is quite
dense (as it would have been in the early Universe). The mean free path of photons (the distance a photon can travel,
on average, before being scattered by an electron in an ionised gas), is given by
1
lmfp ≈ , (101)
ne σ T
where ne is the number density of free electrons, and σT is the scattering cross-section of electrons as seen by photons.
Since the energy density (and also number density) of regular matter scales like ρm ∝ a−3 ∝ (1 + z)3 , the mean
free path will get smaller at higher redshifts. When it becomes small enough, the probability of a photon travelling
through the ionised gas without being scattered becomes very low – the gas becomes opaque to light. This is indeed
what the early Universe was like – so densely filled with ionised gas that photons could not travel through it at all.
There are a couple of other important facts about the early universe that we need to understand:

1. Most of the normal baryonic matter in the early Universe was hydrogen produced soon after the Big Bang;

2. Processes soon after the Big Bang also produced large numbers of photons.2

So, the early Universe was hot, dense, and contained large numbers of photons. Neutral hydrogen has a binding energy
of 13.6 eV, so photons with an energy greater than this can easily ionise it. This is indeed what happened during most
of the early period of cosmic history; there were so many energetic photons around that if a neutral hydrogen atom
managed to form, it would rapidly be ionised again by a photon. The early universe was filled with hydrogen, but it
simply couldn’t exist in a neutral form for more than a fraction of a second before being ionised again.

6.2. Formation of the Cosmic Microwave Background

Even further into the past, the Universe was of course even hotter and denser, and so it must have been opaque to
light since the Big Bang. There must, therefore, have been a time when the Universe first became cool enough (and
the density became low enough) that photons could finally travel freely – a time when the Universe first became
transparent. The photons emitted or scattered at around this time would have been able to travel through the Universe
virtually unimpeded since then; the mean free path would have become too large for there to be a significant chance of
them scattering off any matter. Could we see those first unimpeded photons today?
The answer is yes! These photons give rise to a phenomenon called the Cosmic Microwave Background radiation,
or CMB. The CMB has a number of important properties. One is that it is observed at microwave wavelengths! We
will see shortly that the CMB photons were originally emitted primarily at visible/infrared wavelengths, but since then
2
In fact the energy density of the early universe was dominated by radiation more than matter, ρr  ρm , although by the time of recombina-
tion matter had become the dominant component.

37
the Universe has expanded greatly and the photons have been significantly redshifted. The approximate redshifting
factor happens to be ∼ 1100 today, and so photons that were a few hundred nanometers in wavelength when emitted
(i.e. visible light) would now be almost a millimetre in wavelength as seen today – in other words, at microwave
frequencies.
Another important property of the CMB is that it was emitted from everywhere in the Universe at almost exactly
the same time. Since our Universe is very close to being homogeneous and isotropic, the mean free path of photons
will have dropped below the threshold required to make the Universe transparent at about the same time everywhere.
The resulting photons will have travelled in all directions, from everywhere in space; the Universe is therefore filled
with them! It is a background of photons that exists everywhere throughout space.
Further reading: Cosmic Microwave Background Explained [PBS/YouTube].

6.3. The surface of last scattering

Recall the definition of the comoving distance travelled by light from Section 4. The CMB photons we see on Earth
today are the ones that have travelled from regions of the Universe at a comoving distance of r(z ≈ 1100) away from
us. Photons emitted from regions that are closer to us already travelled past us a long time ago, while photons emitted
from further away are yet to arrive. They were all emitted at the same time though.
If we were to draw the regions that we are currently receiving CMB photons from, we would trace out a spherical
surface around us at a constant comoving distance r = r(z ≈ 1100). This surface is called the last-scattering surface.
We can draw a different last-scattering surface around every point in the Universe; different observers in the Universe
will all see a slightly different CMB, emitted from different regions in space. It always corresponds to photons that
were emitted at the same time though – when the Universe first became transparent.

6.4. Recombination and decoupling

For the Universe to be opaque, the mean free path of photons has to be short enough that practically every photon
will scatter off an electron/ion with a high probability after only a short time. As discussed above, the mean free path
depends primarily on the number density of free electrons. For enough free electrons to be available, we need the
Universe to be hot enough for gas to be ionised. We also need the Universe to be dense enough that there is a high
probability of photons interacting with electrons after travelling only a short distance.
These two conditions are somewhat separate. We could imagine an early universe that was very dense but much
colder, so that little of the gas was ionised. This would be transparent to some forms of EM radiation. We can also
imagine a hot early universe that had a very low density of normal matter, so that ions/electrons were quite rare even
though everything is ionised. In our Universe, the time when the Universe became cool enough for neutral atoms to
form was also around the same time that its density dropped enough to allow photons to propagate freely though.
The time when the first neutral atoms formed is called recombination. It is poorly named, as this was the first time
that ions and electrons had ever been combined – there is no “re” about it. Before recombination, a large fraction of the
hydrogen in the Universe was ionised, and so there were many free electrons (and protons).
The time when photons were first able to travel freely, without being scattered by electrons, is called decoupling.
Before this, we say that photons were ‘coupled’ to electrons by EM scattering processes.

6.5. Recombination
Forming neutral atoms was surprisingly hard in the early Universe! Consider the reaction

p + e−  H + γ. (102)

Ionised hydrogen is just a proton, p, and for every proton there will be an electron e− if the Universe has no net charge.
These can bind together to form a neutral hydrogen atom H and a photon γ. The opposite can also happen, where a
photon come in and ionises a neutral hydrogen atom, leaving behind a proton and electron.
For recombination to happen, this process has to proceed very efficiently from left to right much more frequently
than the opposite ionisation process from right to left. As the Universe expands and cools, there is less energy available
to ionise any neutral hydrogen that forms, and so the recombination process does begin to happen more efficiently.
Unfortunately, this process also emits a photon with energy equal to the binding energy, so every time we create a
neutral hydrogen atom, we release a photon that is capable of ionising another neutral hydrogen atom! In the dense

38
early Universe, these photons soon find another hydrogen atom to ionise. The result is that almost all of the hydrogen
stays ionised, even when the temperature drops well below the ionisation energy.
The relative abundance of protons, electrons, and neutral hydrogen atoms can be calculated using the Saha equation.
This is an approximate equation that relates the occupancy of two states to the energy difference between them. It
assumes that a thermal equilibrium exists (essentially, that switching between states can happen very rapidly and
efficiently), which is a good approximation in the early Universe.
The two states that we are interested in as far as recombination is concerned are the left- and right-hand sides of
the reaction above. One is the ‘ionised’ state, with separate protons and electrons, while the other is the ‘neutral’ state,
consisting only of neutral hydrogen (and photons). Working in terms of the number densities of each type of particle,
we can write
 3  
np ne me kB T 2 E∞
≈ exp − , (103)
nH 2π~2 kB T
where np , ne , nH are the number densities of the particles, me is the electron mass, T is the temperature, and E∞ is
the energy required to ionise hydrogen from its ground state, E∞ = 13.6 eV.
The right-hand side of this equation might remind you of the Maxwell-Boltzmann distribution, which describes the
distribution of velocities of a thermal gas of massive particles. It is closely related. If a gas of particles (in this case,
free electrons) is in thermal equilibrium at temperature T , the particles will tend to have a distribution of energies given
by this form of equation. The threshold energy E∞ determines whether the electrons have enough energy to remain
free, or whether there is a good chance of them being absorbed to form a neutral atom.

Figure 10: Plot showing the left-hand side of the Saha equation (y axis) as a function of temperature, T (blue line). The
dashed black line shows the y value where xe = 0.5. This intercepts the Saha equation where Trec ≈ 4000 K. This is a
good rough estimate of the temperature of recombination.

We can solve this equation to get a rough estimate of when recombination must have happened. First, we note that
the Universe seems to be neutral on average, so there must be roughly one electron for every proton. This lets us set
ne ≈ np . We also need to know the total number density of hydrogen (neutral or ionised) in the Universe, which is
measured to be nH + np ≈ 1.6 (1 + z)3 m−3 (hydrogen is a type of matter, so its density scales like a−3 = (1 + z)3 ).
Recall that the temperature itself scales with redshift like T ∝ (1 + z).
If we now define the ratio of the electron number density to the total hydrogen number density as
ne
xe = , (104)
np + nH

39
we can see that xe → 0 when the hydrogen is completely neutral, while xe → 1 when it is completely ionised. If we
rewrite the Saha equation in terms of this variable, we get
3
x2e
  
1 me kB T 2 E∞
≈ exp − , (105)
1 − xe np + nH 2π~2 kB T

A reasonable guess for when recombination happens is when around half the hydrogen is ionised and half is neutral,
so xe ≈ 0.5. We can solve the Saha equation graphically (i.e. by plotting it) to find the temperature at which the RHS
is equal to the LHS with xe = 0.5. The result is shown in Fig. 10: xe = 0.5 when T ≈ 4000 K. This corresponds to a
recombination redshift of zrec ≈ 1450, which is not too far away from the value of zrec ≈ 1090 obtained from a more
precise calculation.
A temperature of 4000 K corresponds to an energy of approximately 0.35 eV – much lower than the ionisation
energy of hydrogen (13.6 eV). This is how much extra the Universe had to cool in order for neutral hydrogen to form
without energetic photons rapidly reionising it again.

6.6. Decoupling
We have seen how the Universe needs to cool down significantly before neutral atoms can exist. Now let’s figure out
when the Universe would have become transparent to light.
The condition for light to be able to travel freely in the Universe is that its mean free path must be large. Large
compared to what? Well, ideally, each photon would interact with matter at most once as it travelled across the
Universe. Otherwise, it would potentially be scattered multiple times, and the Universe would again be opaque out to
some distance.
As a rough guide, the mean free path of photons should therefore exceed the Hubble radius, rHR ∝ (aH)−1 , at a
particular time if we want the Universe to be transparent. Recall that this is the approximate maximum distance over
which physical processes can operate at a given time in cosmic history. If the mean free path is larger than rHR , then a
typical photon will never scatter off an electron as it travels through space.
We can find when this should happen by simply equating the expression for the mean free path (see above) with the
expression for the Hubble radius (see previous sections). If we’re careful about remembering whether these quantities
are in proper or comoving coordinates, we arrive at the following expression that defines when decoupling happens:

lmfp c
& , (106)
a aH(a)

where we have written the comoving mean free path on the left-hand side. Substituting the definition of the mean free
path from above, we obtain
1 c
& =⇒ H(a) & ne σT c, (107)
ne σ T a aH(a)

which implies that decoupling happens at a scale factor a = adec twhen

H(adec ) ' ne (adec )σT c. (108)

3
We know that decoupling happens in the matter-dominated era of cosmic history, when ρm  ρr , so H(a) ∝ a− 2 at
this time. We also know from above that the number density of electrons should scale like matter, ne ∝ a−3 , while the
Universe remains strongly ionised. We could therefore solve this equation for adec to get an estimate of the redshift at
which decoupling happened.
A complication is that ne will drop even more rapidly when recombination starts to happen, as electrons become
bound into neutral hydrogen atoms and xe → 0. In turn, this will cause the mean free path to rapidly increase. As a
result, decoupling must therefore happen at a very similar time to recombination, as recombination will rapidly drive
ne → 0, therefore satisfying the condition for decoupling to happen from above for practically any value of the Hubble
radius.

40
6.7. Blackbody spectrum of the CMB
The ionised hydrogen gas and radiation were very close to being in thermal equilibrium before recombination and
decoupling happened. We saw that this gave a particular distribution of electron energies in the Saha equation. It also
gives the radiation a very particular energy distribution, known as a blackbody or Planck distribution, which has the
form
2hν 3 1
Bν (T ) = 2 hν . (109)
c e kB T − 1

When last scattering happened, the Universe was so close to being in thermal equilibrium that the resulting radiation
– the CMB – was emitted with an almost perfect blackbody spectrum. Measurements of this spectrum made by the
COBE space mission are shown in the plot below. The measurement is so precise that the error bars have been blown
up by a factor of 400 just so that you can see them, but even so the CMB fits the expected blackbody curve almost
exactly.
The measured temperature of the CMB today is T = 2.725 K. By scaling this temperature with redshift, we can
figure out what the temperature of the Universe must have been at any redshift, T (z) = T0 (1 + z), where T0 = 2.725
K. We can also use this to work out the energy density of radiation today, since ρr ∝ T 4 .

Figure 11: Frequency spectrum of the cosmic microwave background, measured by the COBE mission. It is extremely
close to a blackbody spectrum with temperature T = 2.725 K; the error bars on the measurement have been drawn
400× bigger than they really are just to be visible! (Credit: COBE FIRAS)

Figure 12: Colours corresponding to blackbody radiation of different temperatures. (Credit: Wikimedia)

41
Understanding:

X What is the Cosmic Microwave Background (CMB)?

X What was the Universe like before the CMB formed?
X What is the mean free path of a photon in an ionised medium?
X What are decoupling and recombination?
X What roles did decoupling and recombination play in the formation of the CMB?
X What is the last scattering surface of the CMB? How is it defined for different observers?
X What is a blackbody frequency spectrum?
X How does the temperature of a blackbody spectrum vary with redshift?
X At approximately what redshift did the CMB form?

42
7. Cosmic Microwave Background Anisotropies
In this section you will learn about the small fluctuations (temperature anisotropies) that we observe in the Cosmic
Microwave Background radiation. You will see how the CMB anisotropies can be analysed using the spherical har-
monic expansion and a statistical tool called the power spectrum, and how different features in the power spectrum are
attributable to different physical effects.

Reading for this topic:

– An Introduction to Modern Cosmology (A. Liddle), Chapters 10 and 12.

– An Introduction to Modern Cosmology (A. Liddle), Advanced Topic 5.4 (CMB Anisotropies).

7.1. CMB anisotropies

The early Universe was extremely homogeneous – typical fluctuations in the radiation and matter energy density were
around 1 part in 105 or so. This homogeneity was caused by two main processes. The first was inflation, a process that
occurred in the first fractions of a second of cosmic history, and which we will learn about in the following section.
Inflation set the ‘initial conditions’ of our Universe, smoothing out any inhomogeneities that may have been there
before, and leaving behind only tiny fluctuations by the time it finished (again, in the first fractions of a second).
The second was thermalisation, the tendency for any temperature and pressure differences to average out because
the early Universe was in a state very close to thermal equilibrium. As we learned in the previous section, the mean free
path of photons was very low before last-scattering; photons could typically travel only a very short distance before
being scattered off an electron. On average, this tends to make larger fluctuations smaller – slightly hotter regions of
the Universe would tend to lose energy, while slightly colder regions would gain energy. This process resulted in the
almost-perfect blackbody radiation that we now see as the CMB.
The fact is that there were small fluctuations in the early Universe however. Different physical processes shaped
these fluctuations; some processes worked to erase them (like thermalisation); some generated them (like inflation);
and some allowed existing fluctuations to grow (like gravitational attraction).

Figure 13: Temperature anisotropies of the CMB, shown in a Mollweide projection. (Credit: Planck Collaboration)

In fact, there were many physical processes at work in the early Universe. Each of them had a distinctive effect on
the fluctuations in the energy density of radiation, baryons, and dark matter, which we are now able to observe because
they caused fluctuations in the CMB temperature. These fluctuations are called anisotropies, because we see them as

43
deviations from the average CMB temperature that are different in different directions. Without these fluctuations, the
CMB would be perfectly isotropic, i.e. perfectly uniform across the whole sky. By observing these fluctuations, we are
in some sense getting a ‘snapshot’ of the Universe as it was very early in its history, around 380,000 years after the Big
Bang (when last-scattering occurred).
A map of these fluctuations is shown in the figure above. This map was observed by the Planck satellite, an ESA
mission that mapped the CMB temperature very precisely at microwave frequencies. The figure you see is a projection
of the whole sky (a sphere) onto an egg-like shape. This type of projection is called a Mollweide projection, and is also
used for making maps of the Earth for example, as in the figure below. Note how it distorts the shapes of objects (like
continents). It preserves the area of the objects though, so the continents in this map are at least the correct size. We
use the Mollweide projection when we make maps of the CMB precisely because of this area-preserving property –
the size of the fluctuations is an important signature of the physical processes that caused the anisotropies, as we shall
soon see.

Figure 14: A Mollweide projection of the surface of the Earth. (Credit: Strube/Wikipedia)

Also note the temperature scale of the CMB map. The mean temperature (T = 2.725 K) has been subtracted from
this map, leaving behind fluctuations that are at most a few hundred µK (micro-Kelvin) in size. A quick calculation
gives us the typical size of the fluctuations as a fraction of the mean temperature:
∆T 100 µK
∼ ≈ 4 × 10−5 . (110)
T 2.725 K
This is pretty tiny! To measure the fluctuations, we need a device that can tell apart the temperatures in a radiation field
to 1 part in 105 or better. A lot of the technology to make these measurements possible was developed right here in the
Physics Department at Queen Mary in the 1960’s and 70’s, back when it was called Queen Mary College. The research
group was eventually spun out into a company called QMC Instruments, which is now based in Cardiff and still plays
an important role in developing sensitive EM radiation detectors for use in astronomy.

7.2. Physical processes that cause anisotropies

As mentioned above, many different physical processes contribute to cause the temperature anisotropies that we see in
the CMB. What are these processes, and how do they cause the anisotropies?
Any process that is capable of changing the temperature of the CMB radiation field locally will cause anisotropies.
Processes such as cosmological redshift are global, in the sense that they affect everywhere in the Universe in the same
way at the same time. As such, while the cosmological redshift does change the temperature of the CMB, it cannot
cause anisotropies because it doesn’t change the temperature by different amounts in different places/directions.
What do we mean by ‘local’? We mean that the process can occur more or less strongly in some regions than
in others. The ‘strength’ of the effect will depend on some underlying physics, to be discussed shortly. The size of
the region affected will also depend on the underlying physical process. This is a really important point – different

44
physical processes act over different distances. So, if we map out the CMB anisotropies as a function of their size, we
can potentially identify which processes were at work, and how strong those processes were. We can build a picture of
what physics was like in the early Universe!
What can locally change the temperature of the CMB radiation field? Recall that before last scattering, photons and
baryons were tightly coupled together by Thomson scattering (scattering of photons off electrons in the highly-ionised
gas). As a result, we would expect any changes in the baryon distribution to very rapidly get picked up by the photon
distribution and vice versa. The coupling between the two is so strong that we refer to the photons and baryons as a
‘coupled photon-baryon fluid’ before last scattering. If we can cause fluctuations in this fluid somehow, they will be
converted into temperature anisotropies very efficiently.
The easiest way to do this is to change the energy density of the baryons locally. If a region has baryons with
a slightly larger energy density, some of that excess energy will be transmitted into the photons, hence increasing
their temperature (recall that ρr ∝ T 4 ). So, by increasing the density of baryons in a region, we can increase the
CMB temperature. This can be achieved through gravitational collapse – if a region starts off with a slightly higher
matter/energy density than average, gravity will tend to attract more matter/energy towards it, thus increasing its density
further. Likewise, regions that are slightly lower density than the average will tend to decrease further in density as
matter is attracted away from them by their surroundings.
This works just as well for the dark matter too; fluctuations in the dark matter density are also enhanced by gravi-
tational collapse. Since there is about 5 times as much dark matter as baryonic matter in our Universe, the dark matter
fluctuations can be quite important – baryons tend to fall into the potential wells caused by the collapsing dark mat-
ter, thus enhancing the fluctuations in the baryon distribution too. This is of course then transmitted to the photon
distribution, and so fluctuations in both dark matter and baryons cause temperature anisotropies.
The process of gravitational collapse requires matter to move into some regions and out from others. There are
therefore flows of matter in the early Universe, with a range of velocities. CMB photons that are scattered off baryons
that are moving with respect to everything else receive a small Doppler shift, changing their frequency (and therefore
energy) by a small amount depending on the direction of their relative motion. Region of the last-scattering surface
that were moving towards us when last scattering happened will have their photons slightly blueshifted, while regions
moving away from us at that time will be slightly redshifted. If we recall that temperature depends on redshift as
T = T0 (1 + z), we can see that a similar relation must also hold for the Doppler shift. Writing this as a fractional
change in the temperature, we obtain
∆T v
=− . (111)
T c
Regions with a velocity away from us (v > 0) are redshifted, which reduces the temperature – hence the minus sign.
Another type of redshift is gravitational redshift. Photons travelling out from a gravitational potential well lose
energy in the process, and so are redshifted slightly, reducing their temperature. Likewise, photons falling into a
potential well gain energy and so are blueshifted. This effect is called the Sachs-Wolfe effect, and affects photons that
were inside a potential well when last scattering occurred.
To recap:

• Fluctuations in the energy density of photons, baryons, and dark matter are enhanced by gravitational attraction.
Increased density ‘heats up’ the photon temperature.

• The motion of matter flowing into/out of over-dense/under-dense regions causes a Doppler shift. This redshifts/
blueshifts the photons, also affecting their temperature.

• Photons travelling into/out of gravitational wells are blue/redshifted, which also changes their temperature.

7.3. Baryon acoustic oscillations

The fact that the photons and baryons are so tightly coupled, while also collapsing under gravity, gives rise to an
interesting phenomenon.
We saw above that gravitational collapse makes over-dense regions even denser. The compression of the baryons
heats them up, which also heats up the photons within that region. Both photons and baryons exert an outward pressure
– radiation pressure and thermal pressure respectively – that counteracts the gravitational collapse.

45
 A similar situation can be found in stars that are forming. The stars collapse under gravity, causing the proto-stellar
material to heat up. In this case, nuclear fusion starts, and generates a very large outward pressure that prevents the star
from collapsing further. The star eventually enters hydrostatic equilibrium – the outward pressure exactly balances the
gravitational collapse and the size of the star stabilises.
This does not happen in the early Universe though. First of all, the densities are not high enough for nuclear
fusion to start. The energy density of the photon-baryon fluid is only increased by adiabatic compression, and there
is no additional source of energy to increase the pressure as the fluid is compressed. In fact, the photon-baryon fluid
responds to the compression with a significant increase in outward pressure that overcomes the gravitational attraction
and causes the region to expand outwards. This expansion cannot continue for long due to the inward pressure of
the surrounding photon-baryon fluid though, and so gravitational collapse soon takes over again. The compression
and expansion cycles continue, setting up oscillations of the photon-baryon fluid. The fluid does not stabilise into
hydrostatic equilibrium – it continues to oscillate until decoupling suddenly prevents the photons and baryons from
strongly interacting with each other any more.
These oscillations propagate as sound waves in the photon-baryon fluid, hence the name ‘baryon acoustic oscil-
lations’ (BAO). They are almost like standing waves, in that they have wavelengths that are close to multiples of the
sound horizon – the maximum distance a sound wave could have travelled since the Big Bang. The sound horizon is
smaller than the Hubble horizon and particle horizon because the speed of sound in the photon-baryon fluid is smaller
than the speed of light in vacuum. The sound speed (squared) can be calculated as
ṗγ
c2s = , (112)
ρ̇γ
where pγ and ργ are the pressure and density of the photons respectively. Since we know how the pressure and density
of photons vary with time (they are types of radiation), we can work out that c2s ' c2 /3. The sound horizon can be
worked out by integrating the sound speed over time, and for our Universe is found to have a value of rs ' 150 Mpc
(comoving) at z ≈ 1090.
If the BAO were a standing wave, the sound horizon would give us the wavelength of the fundamental mode. There
are also harmonic modes, with wavelengths that are integer fractions of the sound horizon. The Universe is constantly
expanding though, so the size of the fundamental mode should be constantly increasing – the actual acoustic waves
therefore never quite manage to resonate across the sound horizon. The photons also cause a small amount of dissipation
of the waves (more on this later), which also damps the resonance. As a result, the waves do not have a single, well-
defined wavelength (plus harmonics) – they are smeared out a bit, and so take a broader range of wavelengths around
the acoustic horizon size.
Nevertheless, this leaves a preferred scale in the photon-baryon fluid. The typical distance between two over-dense
clumps of baryons will tend to be some multiple of the wavelength of the sound waves – there are more baryons around
the peak of a wave, and fewer around its trough. Recall that higher density regions of baryons tend to be hotter and
therefore give rise to positive temperature fluctuations. The BAOs therefore cause temperature anisotropies across the
last scattering, with sizes of order the sound horizon (or harmonics of this). As we will see shortly, these sound waves
leave a very distinctive statistical pattern in the CMB anisotropies.
What happens when decoupling occurs, and the photons and baryons are no longer strongly coupled together?
The photons immediately stream away, reducing the outward pressure and preventing any further oscillations from
occurring. The acoustic waves stall, left in whatever configuration that they were in immediately before decoupling.
The pattern of peaks and troughs is frozen in to the baryon distribution. Regions with a high density of baryons can
subsequently collapse gravitationally, but the distance between peaks in the baryon density left by the sound waves will
not change. This preferred distance scale is called the BAO scale, and is left imprinted in the distribution of baryons
(and dark matter), unchanged even today.
This brings us to an important subtlety that we haven’t mentioned yet, which is the role of dark matter. While dark
matter is not strongly coupled to the photons during the pre-decoupling era, it does still interact with the photons and
baryons gravitationally. In our Universe, the dark matter density is around 5 times larger than the baryon density, so
if there are over-densities in the dark matter, the baryons will tend to fall towards them. Over-densities in the baryons
can also drag the dark matter towards them a bit, although this is a less prominent effect. The dark matter therefore
responds to the waves in the baryon distribution more slowly, acting as a sort of counterweight to the oscillations.
When decoupling happens, the dark matter continues to be gravitationally attracted towards the baryon over-
densities, and eventually catches up with them. This is called the drag epoch, as the baryons are dragging the dark
matter towards them (and vice versa; the baryons will be attracted back towards the dark matter too). The sound horizon

46
at decoupling is therefore not quite the final size of the BAO waves that we observe today; instead we see a slightly
modified wavelength, due to the brief period of baryon/dark matter dragging. This doesn’t last for very long however,
and only changes the wavelength by a couple of Mpc. The redshift at which the dragging completes is only slightly
smaller than decoupling, zdrag ≈ 1060, and the final BAO feature has a comoving size of rBAO ≈ 147 Mpc.
A final subtlety: the acoustic oscillations are happening everywhere in the Universe at the same time. As a result,
there are many, many overlapping sound waves being generated across the Universe at the same time, with similar
wavelengths. So, we don’t see individual waves in the distribution of baryons when we observe the CMB anisotropies;
we see a superposition of many such waves, all jumbled up on top of each other. A similar effect can be obtained by
dropping many stones of the same size in a pond. The waves they generate will all be broadly similar in wavelength,
but the fact that many of them have been dropped at different positions results in a complicated superposition of many
waves on the surface of the pond. If we look at the CMB anisotropies, there is a similar effect; we don’t see the BAO
as individual waves, but when we analyse the anisotropies statistically, we see that there is a preferred separation of
higher-temperature (over-dense) regions that is around the BAO scale.
Does this sound complicated? It is! The theory behind baryon acoustic oscillations was only worked out in the late
nineties, and the conclusive measurements of the phenomenon were made in the early 2000s. The fact that we see them
is a beautiful confirmation that we understand a lot of what’s going on in the early Universe though.
To recap:

• Baryons and photons are tightly coupled before last scattering.

• Gravitational attraction compresses the baryon-photon fluid, heating it up. This generates outward thermal/radi-
ation pressure, which pushes it out again.

• This sets up oscillations,√with a typical wavelength given by the sound horizon (which can be calculated from
the sound speed, cs ≈ c/ 3).

• After decoupling, the sound waves in the baryons are frozen. They leave an imprint in the CMB temperature
anisotropies, as well as a preferred distance scale in the matter distribution that we can observe even today.

• The dark matter is also affected by these sound waves, and is dragged along with the baryons by gravitational
attraction.

7.4. Diffusion damping

While the mean free path of the photons is small before decoupling, it is not completely negligible – they do travel
a short distance before scattering. This allows the photons to diffuse out from hotter, higher-density regions into
surrounding cooler, lower-density regions, as long as those regions are sufficiently nearby. A combination of the photon
pressure and gravitational attraction drags the baryons along with the photons too, causing the baryons to diffuse also.
Dark matter is pulled along with the photons and baryons too. This effect smooths out the temperature anisotropies
over distances smaller than the photon mean free path, erasing the anisotropies generated by other processes. It is called
diffusion damping or Silk damping.
The mean free path slowly increases as the Universe expands, so the diffusion affects larger and larger regions
as time goes by. The diffusion process takes time, however; regions that have been smaller than the photon mean
free path for only a short time will not have suffered from as much diffusion, so the anisotropies will only be mildly
damped, while those that have been smaller than the photon mean free path for a long time will have been almost
entirely smoothed out. This causes a scale-dependent damping of the CMB anisotropies – the smaller distance/angular
scales you look at, the stronger the damping effect. At sufficiently small angular scales, there are no primary CMB
anisotropies at all, although secondary anisotropies (see the next section) can be generated on small scales long after
last scattering. On sufficiently large angular scales, larger than the mean free path of photons immediately before
decoupling, there is no diffusion damping effect.

7.5. Secondary anisotropies

All of the anisotropies that we’ve discussed so far are caused by physical processes happening at the time of last
scattering, on the surface of last scattering. They are known as primary anisotropies.

47
 The temperature of the CMB photons can also be changed as they travel through space after last-scattering, causing
secondary anisotropies. These tend to be much weaker than the primary anisotropies, as after last-scattering, CMB
photons are not confined to one local region, but travel rapidly through the Universe. As such, they typically traverse
a number of regions with different local physical conditions before reaching the observer. Since some regions will
increase the temperature and others will decrease it, travelling through many regions means that the net effect will be
the sum of many positive and negative contributions, which tend to cancel each other out on average. Still, some small
secondary effects are certainly observable.
The figure below shows a map of gravitational lensing of the CMB, as observed by the Planck satellite. As CMB
photons travel through the Universe, their paths are bent, very slightly, by regions of higher mass density that they
pass close to. The fact that light is bent by mass is a key prediction of general relativity, and this map is a wonderful
confirmation of this prediction on the scale of the entire Universe! What you are seeing is essentially a map of all of
the mass between our surface of last scattering and Earth, averaged into a single number from each line of sight (i.e. in
each direction on the sky). Several CMB experiments have made this measurement by making a very high-resolution
map of the CMB, and then figuring out which direction the CMB photons would have come from if they hadn’t been
lensed. It’s an incredible measurement that has only been made for the first time within the past 10 years.

Figure 15: A Mollweide projection of the CMB lensing potential as measured by the Planck satellite. This is essentially
a map of the projected mass density throughout the entire Universe, from last-scattering until today. The grey regions
are regions that couldn’t be measured because our own galaxy is in the way. (Credit: ESA/Planck)

Another type of secondary anisotropy is the integrated Sachs-Wolfe (ISW) effect. This is similar to the Sachs-
Wolfe effect, in that it is caused by the gravitational redshifting of CMB photons. This time, the photons are passing
through gravitational potential wells on their way to the observer, instead of at the surface of last scattering. As they
fall into potential wells they gain energy and are blueshifted, but this is exactly cancelled by the redshift as they travel
back out of the well. On average, then, there is no net blue- or redshifting caused by potentials that the photons travel
through after being emitted.
This is only true if the potentials do not change with time however. If a potential gets deeper in the time it takes
a photon to travel through it, the photon will experience a steeper slope on the way out than it did on the way in.
The blueshift experienced when travelling into the potential (when it was shallower) will therefore be less than the
redshift experienced when leaving the potential (now it has grown deeper). In our Universe, potential wells remain
almost constant until quite late times, and so the ISW effect doesn’t happen until the photons have travelled a long
distance. When dark energy starts to dominate the cosmic energy budget at late times, however (when ρΛ > ρm ), it
causes the potential wells to decay (get shallower)! This adds a small but non-negligible ISW contribution to the CMB
anisotropies that we see. The fact that the ISW effect can be measured from the CMB is strong evidence that dark
energy is real – without it, there would be no ISW secondary anisotropy to observe.

48
 A final type of secondary anisotropy is called a spectral distortion. All of the anisotropies that we have discussed
so far simply change the temperature of the CMB photons by a small amount, but still leave their energy distribution as
an almost perfect blackbody (Planck) distribution. Spectral distortions, on the other hand, also induce deviations in the
photon energy distribution away from a perfect blackbody. This can be achieved by giving energy to the photons from
matter that is not in thermal equilibrium with them.
The biggest source of such energy is the extremely hot but low-density gas that exists inside galaxy clusters. This
gas has been heated up to millions of Kelvin during the formation of the galaxy clusters, and is so hot that it glows at X-
ray wavelengths! As CMB photons travel through this gas, they occasionally scatter off a high-energy electron, gaining
some of its energy and thus departing from the blackbody energy distribution that they had previously. The physical
process at work is called Compton scattering, and gives rise to different types of spectral distortion called the thermal
and kinetic Sunyaev Zel’dovich effects. By measuring the size of these effects, we can learn about the distribution
of galaxy clusters between us and the last-scattering surface, and how hot gas comes to be distributed throughout the
Universe long after recombination.

Figure 16: A massive galaxy cluster, seen via the CMB temperature anisotropy caused by the Sunyaev-Zel’dovich
effect (left) and in X-ray emission (right). Particular hot-spots in X-ray emission are shown using white contours in
both panels. (Credit: ESA/Planck)

7.6. Spherical harmonics

We can’t get much information by simply staring at a map of the CMB. As discussed above, the really important
information comes from figuring out how the temperature anisotropies depend on the size of the region being observed,
as this tells us which physical processes were at work at the time the anisotropies were generated. Some physical
processes operate across larger distances than others, and some have a bigger effect than others.
To see this, what we would really like is some measure of the amount of anisotropy as a function of angular scale,
i.e. the size of the observed region. The mathematical tool that allows us to do this is called the spherical harmonic
transform, and it is very similar in principle to a Fourier transform. Instead of viewing a map of the CMB, we can use
the spherical harmonic transform to break up the map into a sum of many waves on the sky, with a range of different
‘wavelengths’. The amplitude of each wave component tells us how much of the anisotropy comes from angular scales
corresponding to the wavelength of that wave. Long waves (on the sky) correspond to large angular scales, while short
waves correspond to small angular scales.
Mathematically, we can write the spherical harmonic expansion of the temperature anisotropies as a sum,
∆T XX
(n̂) = a`m Y`m (n̂). (113)
T ml

This says that we can work out the size of the fractional temperature anisotropy, ∆T /T , in any direction on the sky, n̂,
by summing over all of the functions Y`m , where each function has some amplitude a`m . The ` and m label different
modes, or wavenumbers. There are two wavenumbers because we are performing this expansion on the sky, which is a
2D surface (so we need 2D coordinates).

49
Figure 17: Spherical harmonic modes. Each row (‘order’) corresponds to an ` mode, with 2` + 1 m-modes per row
(Credit: M. Adjeiwaah et al.)

Recall that Fourier transforms work in a similar way. A function f (x) can be written as a sum over Fourier modes,
X
f (x) = fn e2πinx/L (114)
n

where kn = 2πn/L is the Fourier wavenumber, and fn is the amplitude of each Fourier mode. Larger values of n
correspond to larger Fourier wavenumbers, which have smaller wavelengths. We can write any reasonable continuous
1D function as a sum over Fourier modes. Similarly, we can write any continuous 2D field on a sphere (like the sky) as
a sum over spherical harmonic modes. Just as a function is uniquely specified by a set of Fourier coefficients fn , a 2D
field on a sphere is uniquely specified by a set of spherical harmonic coefficients, a`m .
The figure above shows the spherical harmonic functions Y`m for the first few values of ` and m. For a given value
of `, only integer values of m between −` and +` are allowed. For each ` mode, there are 2` + 1 different m modes.
Let’s pay particular attention to the first two rows of the figure. The first row is ` = 0. This is called the monopole
mode, and there is only one m-mode (m = 0). This mode is perfectly uniform over the entire sky. Its interpretation is
as the average over the whole sky – if you take any map of the sky and average over all angles, you will get the value
a00 , corresponding to (`, m) = (0, 0).
The next row corresponds to the dipole. This corresponds to an anisotropy pattern on the sky that is positive in one
direction and negative in the opposite direction. If this were a Fourier mode, we would have fit a single wavelength
across the whole sky – one peak and one trough. In general, this wave could be ‘pointing’ in any direction, correspond
to the three Cartesian coordinates, x, y, z, or some combination of them. This is why there are three m-modes for
` = 1; each of a1,−1 , a1,0 , a1,+1 tells us how much of this mode is pointing in each of the directions x, y, and z.
It gets harder to visualise, but we can extend this mathematical pattern to higher values of `. You can see that the
m = 0 mode always ‘points’ in the Cartesian z direction for any value of `. You can also see that higher values of `
mean that more ‘wavelengths’ of the wave can be fit across the sky. Larger ` implies shorter wavelengths, and therefore
smaller angular scales on the sky. The approximate relation between wavenumber ` and angular scale ∆θ is
π
∆θ ' , (115)
`
where angles are measured in radians. An angle of 1 degree corresponds to l ≈ 180, while 10 degrees is ` ≈ 18.
To calculate the values of the spherical harmonic coefficients, we can take a map of the sky (i.e. a value of T in
every direction n̂), multiply it by the complex conjugate of the spherical harmonic function, and then integrate over the

50
whole sky:
Z
∗
a`m = T (n̂)Y`m (n̂)dΩ. (116)

This result follows from the fact that the spherical harmonics form an orthonormal basis.

7.7. Power spectrum of the CMB

To see how big the anisotropies are as a function of angular scale, we can calculate a quantity called the power spectrum.
The power spectrum is a statistical quantity related to the variance of the temperature anisotropies.
Why do we need to take the variance? It’s because the anisotropies themselves are random but correlated. The
initial conditions of the Universe are random (which we’ll discuss when we learn about inflation), which leads to the
anisotropies having a high level of randomness too – we can’t predict where a particular temperature fluctuation will
occur, for example, or exactly how big it will be. The physical processes at work in the early Universe correlate the
anisotropies though, as they interact and so can come to share some of the same properties. Because they are correlated,
if we know the size of one anisotropy, we can make a reasonable estimate of the size of another, neighbouring, one.
The upshot of all this is that the anisotropies are random, but we can predict some things about them – particularly
their typical size as a function of angular scale. This is exactly what the variance (and therefore the power spectrum)
tells us. Note that the mean of the anisotropies is zero, by construction – we are measuring them as fluctuations around
the average CMB temperature, which we measure separately.
We normally measure the power spectrum as an average of the variance of the m-modes,
+`
1 X
C` = |a`m |2 . (117)
2` + 1
m=−`

The square of the coefficients, |a`m |2 , gives the variance of each spherical harmonic mode. The factor 2` + 1 is the
number of m-modes per `-mode, and so dividing by this factor gives the average. The reason we average over the
m-modes is because the CMB is expected to be statistically isotropic. Since we believe our Universe to be almost
homogeneous, this means that the temperature fluctuations on the surface of last scattering should have been very
similar everywhere. As we now observe them in the CMB radiation, the fluctuations should therefore have no preferred
direction; they should not be larger on one side of the Universe compared to the other, for example. As a result of
this statistical isotropy, we expect all of the m-modes to have the same statistical distribution (i.e. no dependence on
direction), and therefore the same variance. This is why we can average them together, which we do in order to reduce
the statistical uncertainty on the measurement.
The figure below shows the power spectrum of CMB temperature fluctuations measured by the Planck satellite.
The x-axis shows the spherical harmonic wavenumber, `. Recall that larger values of ` mean smaller angular scales.

7.8. Features in the CMB power spectrum

• Acoustic peak – There are actually several acoustic peaks (see baryon acoustic oscillations below), but the first
one is the biggest. It is observed at an angular scale of about 1◦ , and corresponds to the sound horizon of the
photon-baryon fluid at the time of last scattering. This is essentially the fundamental mode of the sound waves in
the photon-baryon fluid. It’s important because it is the biggest, and therefore sets the typical angular size of the
temperature fluctuations – if you look at the CMB map above, the typical size of the fluctuations that are most
easily seen by eye is about 1◦ .
The first acoustic peak is a very useful standard ruler. We can work out the size of the sound horizon at last
scattering if we know some basic properties of the photon-baryon fluid, such as the relative densities of photons
and baryons. The observed angular size of the acoustic peak can then be used to infer the angular diameter
distance to last scattering, i.e. dA (z ≈ 1090) = rs /∆θpeak , where rs is the size of the sound horizon.
This measurement is very important in establishing the geometry of the Universe; if it was open or closed, the
typical size of the anisotropies would be different, and so the acoustic peak would be at a different angular scale.
The fact that it’s observed at around a degree, when combined with other observations, tells us that our Universe
must be very close to flat.

51
Figure 18: CMB power spectrum measured by the Planck satellite. The spherical harmonic multipole ` is shown along
the bottom x-axis, with corresponding angular scale shown along the top x-axis. (Credit: Planck Collaboration)

• Baryon acoustic oscillations – The sound waves in the photon-baryon fluid are also present on scales smaller
than the sound horizon, and so we see a series of other peaks in the power spectrum extending to smaller scales.
These are preferentially seen at distance scales that are a multiple of the sound horizon – i.e. harmonics of the
fundamental acoustic mode. You can think of them as the modes that exist as standing waves in the photon-
baryon fluid.
The peaks are not narrow, and the power spectrum does not go to zero in between them, as we might expect for
harmonic modes of a musical instrument, for example. The peaks are broader because of damping, and because
the size of the Universe is changing as the photon-baryon fluid oscillates (due to cosmic expansion). This makes
the harmonics less well defined.
The power spectrum does not go to zero in between the peaks because of the Doppler shift. The peaks themselves
are caused by the acoustic waves in the density of the photon-baryon fluid (higher density means higher CMB
photon temperature and so on). Recall that the velocity of the fluid also causes a temperature change however. If
you think back to the simple harmonic oscillator, you may remember that the velocity of the oscillator is highest
when the displacement of the oscillator is lowest (e.g. a pendulum travels fastest when it is perfectly vertical).
The same is true here – the velocity is largest in between the density fluctuations of the photon-baryon fluid. This
means that there are temperature fluctuations caused by the Doppler shift with wavelengths exactly in between
the wavelengths of the baryon density fluctuations. This tends to fill in the power spectrum between the acoustic
oscillation peaks, although the effect is not as large because the Doppler shift produces a smaller temperature
anisotropy than the density fluctuations.
As you can see, a lot of different physics is at work in producing the BAO peaks, and so it stands to reason
that we can learn about a lot of physics by measuring them! These peaks do indeed tell us a lot about the early
Universe, including the expansion rate at last scattering, and the relative amounts of dark matter, photons, and
baryons. In particular, if matter was made purely of baryons (i.e. if there was no cold dark matter), these peaks
would be much bigger.

• Damping tail – The amplitude of the CMB power spectrum decays away as we look on smaller and smaller
scales (larger values of `). This is primarily due to the diffusion damping effect, which is stronger for smaller
angular scales. Experiments have actually been able to measure the power spectrum out to around the ninth

52
 BAO peak, at an ` of around 3000 or so. Beyond this, the primary CMB anisotropies are obscured by secondary
anisotropies like the Sunyaev-Zel’dovich effect, which preferentially show up on smaller scales.

• Sachs-Wolfe plateau – The flat, low part of the power spectrum, visible at ` . 50, is called the Sachs-Wolfe
plateau. This is caused by gravitational redshifts at the surface of last scattering. Inflation left fluctuations in
the gravitational potential on all distance scales – there are about as many small potential wells as large ones for
example – so the Sachs-Wolfe plateau therefore extends across most of the range in ` in the figure. It’s just that
it’s easiest to see at low values of `, where other effects don’t matter as much, or the anisotropies haven’t been
damped away.

• Integrated Sachs-Wolfe feature – The slight uptick in the power spectrum as we go to very low ` is caused
by the integrated Sachs-Wolfe (ISW) effect. This is similar to the Sachs-Wolfe effect in that it is caused by
gravitational redshifts – photons gaining or losing energy (and therefore changing temperature) as they pass
through potential wells. Unlike the Sachs-Wolfe effect, however, this is not due to the potential wells that the
photons were emitted from, but the ones that they pass through as they travel from the last scattering surface
to the observer. The ISW effect only occurs in the recent, dark energy-dominated phase of cosmic history, as
before then the potential wells were almost constant (neither growing nor decaying). The fact that we see an ISW
feature at all is therefore evidence for the existence of dark energy, or something like it.

• Amplitude – The overall amplitude of the power spectrum describes how big the CMB fluctuations are in general.
Increasing the amplitude would make all of the values on the y-axis bigger for example. It is set by the factor
As e−τ , where As is the amplitude of the initial ‘primordial’ power spectrum of fluctuations left by inflation, and
τ the optical depth to last scattering.
The primordial amplitude As gives us a starting point for the size of the fluctuations – the bigger the initial
amplitude (just after inflation at t ≈ 0), the bigger the amplitude at last scattering (t ≈ 380, 000 yr).
The optical depth determines how many CMB photons have been scattered (e.g. by intervening free electrons)
in the ∼ 13.8 Gyr between last scattering and the time we observe them, today. The more free electrons there
are along the line of sight to the CMB, the larger the optical depth, and so the larger the probability of CMB
photons being scattered. Scattering smooths out the CMB fluctuations, reducing the observed amplitude of the
power spectrum. It does this uniformly for all angular scales however, so this effect doesn’t change the shape of
the power spectrum.

• Tilt – Inflation left behind a distribution of shallow gravitational potential wells on all distance scales (i.e. of all
widths). These were the seeds of fluctuations in the photon-baryon fluid, which eventually became the CMB tem-
perature anisotropies. The initial distribution of potential wells is described by the primordial power spectrum,
which we will learn about in the section on Inflation. This power spectrum tells us about the size distribution of
the potential wells – how many large vs. small wells were left after inflation ended for example.
In our Universe, the distribution is slightly ‘tilted’, in that there are slightly more wide potential wells than nar-
row ones. Since these potential wells cause the Sachs-Wolfe effect, and also affected how the photon-baryon
fluctuations formed, the CMB power spectrum also gains a slight tilt in it (so the power spectrum is slightly
larger at low ` than high `, all other things being equal). It’s hard to see by eye, but it is necessary to take the tilt
into account when we make detailed predictions of the power spectrum.

Further reading: How the CMB shows us that dark matter exists (Ethan Siegel).

7.9. Dependence on cosmological parameters

The CMB power spectrum is one of our most precise observational probes. CMB measurements are now routinely
used to constrain the cosmological parameters that describe our Universe with accuracies of 1% or even better.
The way this works is to make the most precise measurements of the power spectrum as possible using a CMB
experiment. We then compare the measurements to theoretical calculations of the power spectrum for a range of
different values of the main cosmological parameters. The figure below shows how the CMB power spectrum varies
when several cosmological parameters are varied. By finding the theoretical prediction that best matches the observed
power spectrum, we can figure out what the cosmological parameters corresponding to the real Universe are. The most

53
accurate measurements of these parameters so far come from the Planck CMB satellite. For the parameters that we
have come across previously in this course, Planck has found the following values:

H0 = 67.27 ± 0.60 km/s/Mpc (118)

Ωm = 0.3166 ± 0.0084 (119)
ΩΛ = 0.6834 ± 0.0084 (120)
rs = 144.39 ± 0.30 Mpc (121)
zeq = 3407 ± 31. (122)

These values were derived under the assumption that the Universe is spatially flat. If we allow the curvature to vary
too, Planck finds a value of Ωk = 0.001 ± 0.002. This is consistent with the Universe being flat.

Figure 19: The CMB power spectrum for a range of different cosmological parameter values (only one parameter is
changed at a time; the other parameters are fixed). This figure shows how sensitive the power spectrum is to the values
of the cosmological parameters. Taken from CMB Parameter Sensitivity (Credit: W. Hu).

Further reading: CMB map simulator (interactive tool)

54
Understanding:

X What are CMB anisotropies?

X What physical effects cause CMB anisotropies?
X What are the baryon acoustic oscillations, how do they form, and what do they look like in the CMB
power spectrum?
X What is the approximate comoving size of the BAO feature?
X What is diffusion damping?
X What are the Sachs-Wolfe and integrated Sachs-Wolfe effects?
X What are spherical harmonics, and why are they used to analyse the CMB?
X What patterns do spherical harmonics make on the sky?
X What is the relationship between spherical harmonic mode, `, and angle on the sky, θ?
X What is a power spectrum? Why do we use it to study the CMB anisotropies?
X What does the CMB power spectrum look like, and where do its various features come from?
X What is the acoustic peak and how can it be used to measure the distance to the CMB?

55
8. Inflation
In this section you will learn about inflation, a period very early in the Universe’s history where space is thought to have
expanded very rapidly. Through the mechanism of accelerating exponential expansion, inflation can explain a number
of puzzling properties of the Universe, including why it is observed to be so close to flat and so close to homogeneous.
We will study simple models for the inflationary mechanism, based on a dynamical scalar field called the inflaton, and
will develop some useful mathematical results that allow us to predict how inflation affects observable properties of our
Universe. We’ll also see how inflation generates the ‘seeds’ of galaxies and other large-scale structure that we see today.

Reading for this topic:

– An Introduction to Modern Cosmology (A. Liddle), Chapter 13.

8.1. How special is our Universe?

The theory of inflation is a fundamental ingredient of the standard model of cosmology. While it describes a very brief
and very remote period of cosmic history, tiny fractions of a second after the ‘Big Bang’, it is essential to make scientific
sense of the state that we have found the Universe to be in for much of its subsequent history. In particular, inflation
provides us with a physical mechanism to solve a handful of problems that would otherwise make our Universe seem
extremely, almost impossibly, improbable.
We will learn about these problems in the following three sections, but all of them ask a similar kind of question –
why does our Universe seem to be in such a special configuration, when so many other possible universes seem much
more likely to have occurred? There are actually quite a few ways in which our Universe is special:

• It is very close to being spatially flat (and was even flatter in the past);

• It is quite old, even though it could have easily started out so dense that it would recollapse immediately;

• It has a cosmological constant that is just the right size to be observable today, without having blown the whole
Universe apart a long time ago;

• It is filled with normal matter, and doesn’t have an equal amount of anti-matter that would have annihilated it all;

• It is very close to being homogeneous and isotropic everywhere, even when comparing regions that have never
been in causal contact;

• It isn’t filled with weird high-energy particle relics that would frequently rip through planets, stars, galaxies etc.
and destroy them;

• It is hospitable enough that at least one part of it has hosted complex living things for several billion years.

Why does our Universe have these properties? In other words, why does it appear to be so finely-tuned to be flat/s-
mooth/old/hospitable to life? This is an absolutely huge question that straddles cosmology, fundamental physics, and
even philosophy.
It turns out that inflation can give us compelling answers to at least a few of these questions without needing the
Universe to have randomly begun in a very, very, very particular and special initial state. In other words, it provides a
physical mechanism for some important properties of the Universe to generically end up in the ‘special’ configurations
we see them in, without needing the initial conditions of the Universe themselves to be special.

8.2. The horizon problem

The CMB shows that the Universe is very smooth on very large distance scales. This is surprising because parts of
our last-scattering surface that are widely enough separated should never have been in causal contact! Signals from
one side of our last scattering surface haven’t had time to reach the other side, even at the speed of light, so how could
the two regions have possibly come so very close to being in thermal equilibrium with one another? What are the
odds that two completely independent regions of the Universe would randomly happen to have almost the exact same

56
temperature, density, and expansion rate? What are the odds that every region of the Universe that we can see on our
last scattering surface had randomly ended up to be almost the same, to within one part in about 100,000? This sounds
fishy – as if the initial conditions of our Universe have been fine-tuned so that every region has very similar properties
from the start (subsequently evolving independently, but producing almost the same results since the initial conditions
were so similar).
This strange occurrence is called the horizon problem, and can be put on a firmer mathematical footing by calcu-
lating which regions could and couldn’t have been in causal contact since the Big Bang.
Since the CMB is our ‘smoking gun’ for this problem, let’s start by calculating the Hubble radius at the time of
last-scattering:
c
rHR (aLS ) = ≈ 234 Mpc, (123)
aLS H(aLS )
where we have used the expansion rate at last-scattering inferred from CMB observations (HLS ≈ 1.4×106 km/s/Mpc)
and zLS ≈ 1090. This comoving distance is larger than the comoving sound horizon, rs ≈ 150 Mpc, as expected – the
acoustic peaks in the CMB were formed due to physical processes before last-scattering, and so should be within the
Hubble radius.
Using the angular diameter distance to the CMB, we can also calculate the angular scale corresponding to the
Hubble radius at this redshift. The angular diameter distance3 can be calculated by observing that the first acoustic
peak arises at `s ≈ 220, so
rs π
∆θs = = =⇒ dA = 10.5 Gpc. (124)
dA `s
The corresponding angular scale for the Hubble radius at last-scattering is therefore
π dA
`HR ≈ ≈ 140. (125)
rHR
If regions of our last scattering surface separated by distances greater than rHR were truly independent, we would not
expect the CMB anisotropies to be correlated on angular scales larger than this (` . 140). The CMB power spectrum
would be flat here, without any features, and the map of CMB anisotropies would not have any coherent structures on
larger angular scales than this (except for secondary anisotropies). In other words, if we filtered out all of the structures
on smaller angular scales, we would expect to be left with a purely uniform random, uncorrelated distribution of
temperature anisotropies (also known as ‘white noise’).

Figure 20: The CMB power spectrum as measured by the Planck satellite. (Credit: ESA/Planck)

Our measurements of the CMB map and power spectrum show that this isn’t the case however; the power spectrum
still has features at ` < 140 (see above), and the filtered CMB map does not look like random noise.
3
Since rs is in comoving units, dA will also be in comoving units in this calculation, which is why it comes out to be such a large number.
To get dA in proper units, which we normally use for this quantity, just multiply by a.

57
8.3. The flatness problem
Another curious observation is that we find space to be so close to flat today (|Ωk | . 5 × 10−3 according to Planck). It
could have started off with any sort of curvature, depending on how much mass/energy the Universe started with. Why
should we find the total density of mass/energy to be so close to the critical density today? Again, it seems as if the
Universe has been fine-tuned to have just enough mass/energy to keep it almost spatially flat.
This observation gets even more curious if we consider how close to flat the Universe must have been in the past.
Recall that the curvature term in the Friedmann scales like Ωk a−2 , and so grows with time. For the curvature to be
small today, it must have been even smaller in the past. In fact, it turns out that the initial spatial curvature of the
Universe, fractions of a second after the Big Bang, must have been absolutely tiny in order to produce the observational
bounds of . 1% on Ωk that we see today.
To see this, let’s rewrite the curvature term as the fractional difference between the total energy density of the
Universe and the critical density at a given time. Recall that Ωtot = ρtot (t0 )/ρcr,0 = 1 − Ωk , all evaluated at t0 (today).
Using the Friedmann equation, we can write the total energy density as a fraction of the critical density at any time t.
First, divide both sides of the Friedmann equation by H 2 :

H2 8πGρtot (t) kc2

= − . (126)
H2 3H 2 a2 H 2
The left-hand side is just 1, while the first term on the right-hand side can be simplified by noticing that the critical
density as a function of time is

3H 2
ρcr (t) = . (127)
8πG
If we borrow the fractional density (‘Ω’) notation to define Ωtot (t) = ρtot /ρcr (t), we can write the Friedmann equation
as
kc2 H02 Ωk H02 Ωk
1 = Ωtot (t) − = Ωtot (t) + =⇒ 1 − Ωtot (t) = . (128)
a2 H 2 (aH)2 (aH)2

(Note that we have folded the dark energy density into the total energy density as usual.)
Consider a Universe with a small deviation from flatness today, Ωk = 10−2 . We can use the expression above to
calculate the fractional deviation from flatness at any time in the past. For example, at decoupling, adec = 1/(1 +
1090) = 9.17 × 10−4 , and the Universe was matter-dominated, so H 2 ≈ H02 Ωm a−3 . Plugging this in, we obtain

Ωk H02 Ωk −5
1 − Ωtot (adec ) = 2 2 −3 = Ω adec ≈ 3 × 10 , (129)
adec H0 Ωm adec m

where we took Ωm ≈ 0.3. So, a 1% deviation from flatness today requires that the Universe was only 0.003% away
from being perfectly flat at decoupling! Let’s look even further back, at the time when the Universe had cooled enough
for the first bound hadrons to form (Thad ≈ 1010 K, corresponding to energies of around 1 MeV). This happened at
a redshift4 of z ≈ 3.7 × 109 , when the Universe was strongly radiation-dominated, so H 2 ≈ H02 Ωr a−4 . Plugging in
ahad = 1/(1 + z) ≈ 3 × 10−10 and Ωr ≈ 10−5 , we obtain

Ωk 2
1 − Ωtot (ahad ) = a ≈ 10−16 . (130)
Ωr had
This is a tiny number, and we haven’t even gone back as far in time as physics reliably allows us to – the further back
we look, the worse the problem gets! For the Universe to be only 1% flat today, it must have been almost perfectly flat
at very early times. What could have caused the early Universe to be so almost perfectly flat so that it still looks close
to flat today? Even if the number above had been 10−15 instead of 10−16 , the value of Ωk would be ten times larger and
space would be very noticeably non-Euclidean. It seems that the (almost-)flatness of space also requires a very large
fine-tuning of the initial state of the Universe.
4
You can work out the redshift corresponding to this temperature from T = T0 (1 + z), where T0 ≈ 2.725 K for our Universe.

58
8.4. The (magnetic) monopole problem
In the early 1980s, a lot of theoretical work revolved around trying to find a theory that unified all of the fundamental
forces of nature. At sufficiently high energies, the EM and weak nuclear forces merge into one ‘electroweak’ force.
This was a very pleasing feature to physicists at the time, as it suggested that there is a deeper symmetry to the Universe
than the various symmetries suggested by the four fundamental forces that we experience at lower energies. A lot of
serious effort went into trying to find a Grand Unified Theory (GUT), which would also unify the electroweak force
with the strong nuclear force.
A common prediction of theories that were put forward as candidates for the GUT is that magnetic monopoles
should be generated in large quantities in the early Universe. Normal magnets are dipoles, with a north and south pole,
and no net magnetic charge. Monopoles are only either a north or a south pole however, and so can have a net positive
or negative magnetic charge. Magnetic monopoles have never been observed in nature.
Why would magnetic monopoles form? They are a class of object called a topological defect. As the Universe
cooled from its initially high-energy state, where the three forces were unified, it would eventually undergo spontaneous
symmetry breaking (SSB), where the forces became separate again. Spontaneous symmetry breaking results in a phase
transition in the quantum fields that pervade the Universe, and topological defects are associated with these phase
transitions. You can think of them as boundaries between regions of the Universe that froze into different vacuum
states after the phase transition (see the discussion of vacuum energy in Section 5). A common analogy is to think
about the boundary between two parts of a crystal that have begun forming separately and eventually grow together;
the crystals on each side of the boundary are unlikely to be aligned, and so a distinct boundary between the two different
crystal phases gets locked in between them and cannot easily be changed. In fact, the mathematical description of phase
transitions and topological defects in cosmology has a lot of similarities to the way similar phenomena are described in
solid state physics! This field – applying solid-state theory to cosmology – was pioneered by Tom Kibble at Imperial
College London, amongst others.
Topological defects have very strange properties, and are completely unlike any normal form of matter or energy
that we are used to. They can come in a small number of possible shapes: monopoles are point-like (0-dimensional)
objects, cosmic strings are like lines (1-D objects), and domain walls are like surfaces (2-D objects). Each type of
object describes a boundary between different types of regions, and each type of object has a different equation of
state; w = −1/3 for cosmic strings and w = −2/3 for domain walls, for example.
They would have a number of interesting observational consequences if we could observe them, including causing
very strong lensing artefacts in the CMB. These have been searched for extensively, and we don’t see them. According
to the GUT theories, we would expect at least a handful of topological defects within every Hubble radius. Our
observations are consistent with there being none. This is the monopole problem – despite topological defects being
a clear, and indeed almost inevitable, consequence of higher-energy theories like GUTs, why do we not see any in
nature? They should be very noticeable!
Further reading: Magnetic monopoles in Grand Unified Theories (Wikipedia); Cosmic strings (Wikipedia).

8.5. The inflationary mechanism

Cue the invention of cosmic inflation, a theory of the early universe that solves these problems in a compelling way.
So compelling, in fact, that since it was first proposed by Alan Guth and others in the early 1980s, many thousands of
scientific papers have been written about it. It even has the philosophers arguing about its implications for the nature
of reality.
What is inflation, and how does it solve these problems? The idea is reasonably simple: instead of the Universe
being filled with radiation at very early times (around 10−36 sec after the Big Bang), what if there was a new force of
nature that took over instead? This is similar to what you might expect from a Grand Unified Theory, where the EM,
weak, and strong nuclear forces are all combined into a single force. There are various options for how such a force
might behave, but a simple and theoretically-compelling option is to model it as a scalar field.
Recall that the electric and magnetic fields are vector fields, which have a direction and magnitude at every point in
space. A scalar field is more like the electric potential, in the sense that it only has a magnitude at every point in space.
This is about as simple as it gets for a force of nature. Fundamental force fields are associated with force-carrying
particles, and the hypothetical early universe scalar field is no exception. Just as the Higgs boson (force-carrying
particle) is associated with the Higgs field (a scalar field), the inflaton is the particle associated with the early universe
scalar field.

59
 The inflaton has two important properties that allow it to solve the horizon, flatness, and monopole problems.
First, it has an equation of state of w ≈ −1, which means that it causes accelerated cosmic expansion, much like the
cosmological constant. Recall from Section 5 that accelerating expansion causes the Hubble radius to shrink instead of
grow. This will turn out to be the most important property of the inflaton. This kind of expansion also causes the scale
factor to increase almost exponentially, rapidly ‘inflating’ the size of the Universe (hence the name).
The second property of inflation is that, unlike the cosmological constant, the inflaton stops causing acceleration
after a while. If it simply had w = −1 forever, it would dominate the energy density of the Universe forever, and no
radiation- or matter-dominated periods would ever arise. The inflaton eventually decays in a process called reheating
however, converting its energy into matter and radiation.
How does a shrinking Hubble radius and exponential expansion help to solve the various problems?
• Horizon problem: In the inflationary paradigm, everything that we see today (including every region of our
last scattering surface) was once contained in a very small patch of the Universe that was in thermal equilibrium
before inflation began. Inflation then caused a brief but intense period of exponential expansion that rapidly
increased the size of this patch by many orders of magnitude (typically a factor of e50 or more). This expansion
was ‘faster than light’, in the sense that the space between objects expanded faster than light could travel between
them, effectively removing them from being in causal contact with one another. (There is no speed limit on how
fast space can expand, so this doesn’t violate special relativity.)
This solves the horizon problem. The reason that different regions of our last-scattering surface look so similar
despite being outside one another’s Hubble radius is that they were actually previously in causal contact. Inflation
then rapidly expanded them apart from one another, placing them outside the Hubble radius (i.e. the Hubble
radius effectively shrank during inflation).
• Flatness problem: During exponential expansion, H ≈ const., so |1 − Ωtot (a)| ∝ a−2 . During inflation, the
fractional difference between the total matter density and the critical density therefore goes down, causing the
Universe to get flatter. Since the scale factor increases very rapidly during exponential expansion, a very large
flattening effect can build up in just a short time. For a factor of e50 increase in the scale factor during inflation,
|1 − Ωtot | can be reduced by a factor of e100 ≈ 1043 . Inflation therefore almost inevitably sets the Universe to
be very close to flat, as we observe today.
• Monopole problem: The solution to the monopole problem is more or less the same as the solution of the horizon
problem – while many monopoles and other topological defects may have been produced before inflation, they
are expanded outside our Hubble radius during inflation, and so we can’t see them any more. Inflation predicts
that at most a handful would be observable within our Hubble radius today, and likely fewer than that. Our
observations are consistent with there being none within our Hubble radius.
Note that after inflation, when normal radiation- and then matter-dominated expansion happened, the Hubble radius
started to grow again. Slowly but surely, regions of the Universe that had been pushed outside our Hubble radius by
inflation re-entered it, and so came back into causal contact with us. The structures and correlations that we see on
very large angular scales in the CMB today are relics of cosmic history that have lain untouched since the very first
moments after the Big Bang, when the Universe was less than 10−30 seconds old.
Further reading: Inflationary universe: A possible solution to the horizon and flatness problems (A. Guth)

What is an e-fold?
Inflation causes the scale factor to increase by a very large amount in a very short time. To keep track of this
when discussing possible theoretical models of inflation, cosmologists often measure the expansion factor in
e-folds.
An e-fold is just an increase in the scale factor of one power of e. Ten e-folds would therefore be a factor of
e10 ≈ 2.2 × 104 . Inflation must typically last for at least 50 e-folds to solve the horizon and flatness problems.
It could have gone on for much longer than this however!
If the number of e-folds was very large, our initial patch of the Universe could have been blown up to a size
many, many times larger than the Hubble radius today. This is one reason why we think that the actual Universe
is much larger than the observable Universe.

60
8.6. Cosmological Klein-Gordon equation
For most models of inflation, the evolution of the inflaton field can be described using the cosmological Klein-Gordon
(KG) equation,

dV
φ̈ + 3H φ̇ + = 0. (131)
dφ
This is the equation of motion for a homogeneous scalar field, φ, in an expanding universe. Recall that dots are our
shorthand notation for derivatives with respect to cosmic time, t. The first term therefore describes the ‘acceleration’
of the scalar field, i.e. the rate at which its ‘speed’ φ̇ is changing. The second term is the damping term, and depends
on the expansion rate, H. The third term depends on the gradient of the potential that the scalar field inhabits, V (φ).
Different models of inflation have different potentials, that will give rise to different behaviours of φ(t) when the scalar
field is allowed to evolve according to the KG equation.
Note that the KG and Friedmann equations are coupled to one another, as the scalar field appears as a source of
energy density in the Friedmann equation. The scalar field alters the solution for a(t), which also alters the solution for
φ(t) (through the factor of H in the second term).
The KG equation is a second-order ODE, and can be solved once initial conditions for φ(ti ) = φi and φ̇(ti ) = φ̇i
are given, and a form for the potential V (φ) has been chosen.
In the early Universe, H was very large, and so the second term tends to take over initially, allowing us to approxi-
mate the KG equation as:

3H φ̇ ≈ 0. (132)

The effect of this term is to drive the ‘speed’ φ̇ to zero, hence why it is called the damping term. The initial speed φ̇i
therefore doesn’t matter so much. This kind of behaviour is called an attractor solution, as it doesn’t matter where you
start – after a while, the equations always draw you into the same place (i.e. where φ̇ ≈ 0).
Once φ̇ ≈ 0, the other terms can become important however. We will study them in more detail when we discuss
slow-roll inflation. For now, the important thing to realise is that the shape of the potential, V (φ), is very important in
determining what happens to the scalar field next. The typical analogy is to think of the scalar field as a ball rolling up
and down hills (the potential). The value of the scalar field, φ, is analogous to the position of the ball in the potential.
Steeper potentials will allow the inflaton to lose or gain more energy as it rolls up or down them respectively.
Shallower potentials will result in a more gentle evolution however. The question is what kind of behaviour is needed
for the inflaton to behave in a way that can solve the various problems discussed above.

8.7. Scalar field dynamics

Continuing with the analogy of the inflaton as a ball rolling around in a landscape described by the potential, V (φ),
we can define the kinetic energy of the inflaton to be φ̇2 /2. This should look familiar from Newtonian dynamics – it’s
simply the speed of the field squared, divided by two (we don’t need to worry about a ‘mass’ of the field, as we can
always redefine some units to make the mass equal to 1). The potential energy of the field is just given by the value of
the potential, V (φ).
We can use these definitions of kinetic and potential energy to understand the expressions for the energy density
and relativistic pressure of the scalar field,

φ̇2
ρφ = + V (φ) (133)
2
φ̇2
pφ = − V (φ), (134)
2
where we have used units where c = 1. The energy density is just the sum of the two different types of energy, while
the pressure is their difference (recall that relativistic pressure is different from thermal pressure). We can then write
the equation of state for the scalar field as
φ̇2
pφ 2 − V (φ)
wφ = = . (135)
ρφ φ̇2
2 + V (φ)

61
Let’s consider two limits of the equation of state – one where the scalar field is dominated by its potential energy
(|V |  φ̇2 ), and one where it is dominated by its kinetic energy (|V |  φ̇2 ). In the first case, we see that wφ ≈
−V /V = −1, which you may remember as the equation of state of a cosmological constant. This means that a
scalar field that is potential-dominated will behave in the same way as a cosmological constant and therefore cause
accelerating expansion! This is exactly what we need to solve our horizon, flatness, and monopole problems – by
having the Universe expand exponentially, we can shrink the Hubble radius, therefore removing the pesky monopoles
and any inhomogeneous regions from view. Accelerating expansion will also cause gravitational potentials to decay
(recall the ISW effect from the last section), smoothing out inhomogeneities in the Universe in the process.
The second case leads to wφ ≈ φ̇2 /φ̇2 = +1, which is quite an unusual equation of state. In this case, the scalar
field does not cause accelerated expansion, but it doesn’t behave like matter (w = 0) or radiation (w = 1/3) either.
In fact, it will tend to make the Universe collapse in on itself, which is obviously a situation that should be avoided if
we’re trying to model the early Universe.

8.8. Slow-roll approximation

As shown above, we want to avoid the inflaton getting too much kinetic energy, and would quite like to end up with
an accelerating expansion with wφ ≈ −1 instead. We are therefore interested in building inflationary models that can
produce this behaviour, especially if it can be achieved without too much fine-tuning. How can we do this?
The key is to construct a potential that keeps the inflaton from gaining too much kinetic energy, while still allowing
it to evolve. The former condition is trivially satisfied if we trap the inflaton at the bottom of a steep potential that keeps
the field stuck with exactly φ̇ = 0. In this case there is no way for inflation to end, however – the field would stay
stuck in the same configuration forever, without being able to move, and so the Universe would continue to expand at
an accelerating rate forever. This does not fit in with what we observe in our Universe. A better option is to put the
scalar field in a shallow potential and then allow it to slowly roll down the slope.
This leads to an important approximation called the slow-roll approximation. When the field is slowly-rolling, we
can neglect a handful of terms in the Friedmann and KG equations that makes them easier to solve.
Under the slow-roll approximation, we require that the kinetic energy of the scalar field is significantly smaller
than its potential energy, φ̇2 /2  |V (φ)|. If we apply this approximation to the energy density of the field, we obtain
ρφ ≈ V (φ). Plugging this into the Friedmann equation (for a flat Universe containing only a scalar field), we obtain
8πG 8πG
H2 = ρφ ≈ V. (136)
3 3
−2
In natural units, the prefactor can be rewritten in terms of the Planck mass, 8πG = Mpl , if needed. Continuing with
the slow-roll approximation, the KG equation becomes
dV
3H φ̇ + ≈ 0. (137)
dφ
Note that we have neglected φ̈; the slow-roll approximation says that φ̇ is small compared to V , which means that the
time derivative of φ̇ must be small also. The term involving 3H φ̇ is not neglected however, as while φ̇ may be small, the
prefactor
√ 3H is actually very large (the expansion rate H is very large in the early universe, and is actually proportional
to V as we can see from the Friedmann equation).
With a bit of rearranging, we can now turn the Friedmann and KG equations into a single integral equation. First,
we can use the Friedmann equation to write da/dt = aH. Rewriting dφ/dt = (dφ/da)(da/dt) = aHdφ/da, the KG
equation becomes
dφ dV
3aH 2 =− . (138)
da dφ
Substituting the Friedmann equation for H 2 , and moving all of the factors of a to the right-hand side, we obtain
8πG V da
− dφ = . (139)
(dV /dφ) a
If we know what V (φ) is, we can evaluate the left-hand side, perform an integral of both sides, and thus find an
expression for a(φ). This can be inverted to find φ(a).
Note that we could also solve the Friedmann equation under this approximation. But, since we know that wφ ≈ −1
in the slow-roll approximation, we already know that the solution we obtain should be very close to a(t) ∝ eHt , where
H ≈ const.

62
8.9. Quantum fluctuations and the primordial power spectrum
We have seen that inflation flattens out and homogenises the Universe, leaving it very close to a perfect FLRW universe
(i.e. one that is perfectly homogeneous and isotropic). If this was the end of the story, our Universe would be even
smoother than we see it today, with far fewer structures and even smaller anisotropies in the CMB.
It turns out that inflation also generates small fluctuations in the cosmic energy density though. These are seen
as small fluctuations in the gravitational potential that give rise to the Sachs-Wolfe effect in the CMB. They also act
as the seeds of the fluctuations in the baryon/photon/dark matter density that become acoustic oscillations and other
inhomogeneities. Ultimately, these fluctuations grow via gravitational collapse to become galaxies and galaxy clusters.
These original seed fluctuations are called primordial energy density perturbations.
The small fluctuations from inflation are generated by quantum fluctuations in the inflaton field itself. At the very
high energies present in the early Universe, it is no surprise that some quantum phenomena might arise, and so we
often describe the inflaton as a ‘semi-classical field’, that has some non-quantum and some quantum properties. The
mildly quantum nature of the inflaton means that there is some uncertainty in when inflation ends, for example – the
end of inflation corresponds to when the temperature of the Universe dips below a certain value, but the Heisenberg
Uncertainty Principle tells us that we can’t know both the energy and time of an event perfectly precisely. The slight
variations in when inflation ended therefore led to some parts of the Universe inflating slightly more than others, leading
to small fluctuations in the energy density.
These quantum fluctuations are generated for the entire duration of inflation. As they are generated, they too
are expanded outside the Hubble radius, where they cannot evolve in time (gravitational collapse is a causal process,
so cannot affect fluctuations larger than the horizon). The fluctuations are therefore frozen in to the energy density
distribution on very large scales, and do not change or evolve. After inflation ends, when the Hubble radius starts to
grow again, the fluctuations re-enter the horizon, and can start growing due to gravitational collapse etc. This means
that the fluctuations we see on the very large angular scales in the CMB are the ones that have most recently re-entered
the horizon. They are frozen relics from the very first moments of cosmic history!
Inflation makes quite specific predictions for the size of the primordial perturbations. Since they are being generated
throughout inflation, they respond to any changes in the physical behaviour of the inflaton field. Most inflationary
models predict that the inflaton slowly rolls at a more or less constant rate however, with H ≈ const., so all of the
fluctuations that are generated have more or less the same general statistical properties. The time at which the fluctuation
is generated determines the eventual distance scale on which it will be observed; fluctuations generated early on during
inflation will be expanded the most, and so will appear on very large scales, while fluctuations generated towards the
end of inflation will be expanded less and so will be seen on smaller scales. The typical size of the fluctuations is almost
constant however – fluctuations on large scales will have almost the same variance as fluctuations on small scales.
Putting this all together, we find that most inflationary models predict an almost scale-invariant power spectrum
for the primordial perturbations. Recall that a power spectrum is a measure of the variance of a fluctuation field as a
function of distance scale. Scale-invariant means that the fluctuations have the same variance on all distance scales, so
their typical size on small scales is the same as on large scales. The inflaton is not perfectly slowly rolling (i.e. φ̇ 6= 0),
so the inflaton does change its kinetic energy slightly with time, and so the fluctuations generated around the start of
inflation are slightly different to the ones generated at the end. The power spectrum is therefore slightly tilted, with
slightly lower variance on small scales than on large scales. We normally write the primordial power spectrum as P (k),
the variance as a function of Fourier mode k. Most inflationary models predict
P (k) ∝ k ns −1 , (140)
where ns is called the tilt or spectral index of the primordial power spectrum. We can measure ns from the CMB power
spectrum, and find a value of around ns ≈ 0.96 – very close to scale invariant, but not exactly. This is in important
prediction of inflation.
Another prediction of inflation is that the fluctuations should follow an almost Gaussian statistical distribution. We
can measure the statistical distribution of the CMB temperature fluctuations and use this to figure out if the primordial
fluctuations were Gaussian or not. So far, all of our observations suggest that they are – another small hint that inflation
really happened in the early Universe.

8.10. Reheating
An important feature of inflation is that after a short time (typically 10−32 seconds or so) it ends. This can happen when
the inflaton gains enough kinetic energy, for example, which makes the equation of state less negative (wφ > −1) so

63
the expansion of the Universe stops accelerating. When building theoretical models of inflation, and common way to
get it to end is to add a sudden dip in the potential. After slowly rolling fro a while, the inflaton falls into the dip and
rolls up and down inside it, sort of like a harmonic oscillator trapped in a potential well. These oscillations cause the
inflaton to decay, converting the energy density locked in the inflaton field into high-energy particles and radiation.
The process of converting the inflaton into normal radiation/matter is called reheating. After a short while, the inflaton
completely decays away and essentially vanishes, never affecting the Universe again – it has done its job of setting up
the initial conditions of the Universe in exactly the right way.

What really happened at the Big Bang?

What we call the Big Bang – the point at time t = 0, when the Universe was infinitely dense, and all of space
was crushed into a singularity – is just an extrapolation of our solutions to the Friedmann equation as far back
in time as we can go. Most cosmologists don’t think there was actually a real, physical singularity like this!
Instead, we think that there was possibly a phase of the Universe’s history immediately before inflation started.
We don’t know how long this pre-inflationary phase lasted, or even if it makes sense to talk about time passing
during this phase! The Universe would have been so hot and dense at this time that energies, lengths, and times
would be around the Planck scale, where our theories of gravity and quantum mechanics break down. Perhaps
theories of quantum gravity, like string theory, will give us sensible solutions to what could have happened
during this time. It seems unlikely that we can ever access this epoch observationally however, since inflation
does such a good job of smoothing out any structure that existed before. Perhaps part of the Universe’s history
is forever hidden from our view.
Talk about a pre-inflationary phase does lead us to the idea of a multiverse however. Perhaps there is a much
larger reality out there that contains many Universes like our own, some of which have not started expanding
yet, and some of which have been expanding for a long time or have even started to contract and have crunched
back in on themselves. Perhaps this multiverse contains more than the 3+1 dimensions of space and time that
we observe in our Universe. Perhaps, sometimes, universes within this multiverse collide with one another and
leave faint marks or scars on each other that can be observed? It’s impossible to say much about any of these
scenarios yet, as we don’t have any observational evidence for any of them, or even a fully consistent theory
that we can do calculations with.
Regardless, we have abundant evidence that something very much like a Big Bang happened – the Universe
began to rapidly grow from an extremely hot, dense state, and then continued expanding. It just probably didn’t
come from a mathematical singularity.

Understanding:

X What are the horizon, flatness, and monopole problems?

X What is the inflationary mechanism?
X How does inflation solve these problems?
X What is the inflaton and what are its properties?
X What is an e-fold?
X What does the cosmological Klein-Gordon equation describe?
X How are the density and pressure of the inflaton related to its kinetic energy and potential?
X What is the slow-roll approximation?
X How does inflation generate primordial fluctuations?

64