The Physics Behind the Cosmic Microwave Background

Without question, the source of the most precise information about the universe as a
whole and about its early state is the cosmic microwave background (CMB). This background
is incredibly smooth, with a temperature that varies by a typical fractional amplitude of
only ∼ 10−5 . Nonetheless, the fluctuations themselves have now been observed so well that
they have strongly supported the hot Big Bang model, provided evidence for the early rapid
expansion of the universe called inflation, and along with other observations provide evidence
in support of dark energy and dark matter. In this lecture we will discuss the physics behind
the CMB, and in the next will discuss the implications and what awaits future instruments.

Basics

First, let us discuss some zeroth order aspects of the CMB: its existence, its spectrum,
and its smoothness.
The first prediction of background radiation was made by George Gamow in the late
1940s. He reasoned that if the universe was once much hotter and denser than it is now,
then it would be optically thick, meaning that a typical photon would scatter many times
on a trip across the universe. This would also lead to occasional absorption and reemission,
meaning that the radiation and matter would be in thermodynamic equilibrium. At this
time, therefore, the radiation would have a blackbody spectrum.
As we discussed in an early lecture, redshifting preserves a blackbody spectrum, simply
decreasing the temperature as T ∝ a−1 ∝ (1 + z). Therefore, when the radiation and
matter decoupled, the radiation would be left to stream across the universe to us. The
current temperature of this background radiation is TCMB = 2.73 K, and the energy in this
background is greater than the energy in all other photons in the universe combined.
An interesting point about this background is that it is isotropic: the temperature is the
same in all directions, to roughly a part in 105 . At first glance this may seem unremarkable;
isn’t it just what we expect from the cosmological principle? Further thought, however,
reveals a puzzle. Recall that in a matter-dominated universe the scale factor goes with time
as a ∝ t2/3 , and in a radiation-dominated universe goes as a ∝ t1/2 . This means that the
region of the universe in causal contact with us (i.e., that could be physically affected by
things moving at light speed or slower) is constantly increasing. In turn, this implies that
at the time that the CMB set on its way to us, at redshift z ∼ 1000, only small patches
of what we see could have been in causal contact. How, then, could they have known to
coordinate their temperatures to such a degree? The best current answer turns out to be
inflation, which we shall discuss in a later lecture.
 Radiation-matter decoupling

Now, however, it is instructive to estimate the redshift from which the CMB streams. We
will start as always with a simple estimate, then proceed to a more sophisticated approach
that is an example of how to deal with potentially complicated equations.
(a) To make a first simple guess as to when the universe became transparent, let’s do the
following. The most important interaction of light with matter at that stage was Thomson
scattering. The “cross section” for scattering (i.e., the effective area of an electron for
scattering by a photon) was then

σ = 6.65 × 10−25 cm2 . (1)

The universe is mostly ionized at the present time (that is, most electrons are free rather than
bound in atoms). The number density of electrons at the current time is n0 ≈ 2×10−7 cm−3 .
At a redshift z, the number density is n(z) = n0 (1 + z)3 . The age of the universe was
t(z) = 1.4 × 1010 yr(1 + z)−3/2 (actually, it’s somewhat different than this, because of the
dominance of dark energy in the last half of the age of the universe), hence the radius of a
causally connected part of the universe was

R(z) = ct(z) ≈ 1.4 × 1028 (1 + z)−3/2 cm . (2)

The mean free path of a photon (i.e., the typical distance one would travel) is

L(z) = 1/[σn(z)] . (3)

You should find that at large redshift, L(z) < R(z), so that a photon scatters before it crosses
the universe, whereas at smaller redshift, L(z) > R(z), so that a typical photon crosses the
universe without scattering.
Using these assumptions, what is the redshift when L(z) = R(z)?
Answer:
(a) We find that L(z) = 1/(nσ) = 7.5 × 1030 (1 + z)−3 cm. Equating L(z) and R(z) gives
(1 + z)3/2 = 540, or 1 + z ≈ 66. As we know, this is much too low, because it assumes almost
complete ionization. That’s why we need to do things in a more sophisticated way.
(b) Now let’s do things more carefully, distinguishing between cross section and opacity. For
simplicity, we will pretend that the universe was pure hydrogen instead of about 25% helium
by mass. The Saha equation then tells us that the fractional ionization y ≡ ne /n (i.e., the
number density of free electrons divided by the total number density) is given by
y2 4 × 10−9 3/2 −1.579×105 /T
= T e , (4)
1−y ρ
where ρ is the total mass density (including protons) in g cm−3 and T is the temperature in
K: T = 2.725(1 + z).
With these assumptions, what is the redshift at which L(z) = R(z)?

Answer:
(b) The Saha equation is moderately complicated, so we need to be able to approach it
carefully. Suppose that the ionization fraction at the epoch of transparency is y. We then
find a redshift that is given by (1 + z)3/2 = 540/y, or 1 + z ≈ 66y −2/3 . We note that
ρ = 2 × 10−7 (1 + z)3 × 1.7 × 10−24 g cm−3 , where the second factor is the mass of a proton,
meaning that we can write the Saha equation as
y2 4
= 5.3 × 1022 (1 + z)−3/2 e−5.794×10 /(1+z) , (5)
1−y
where we have also used T = 2.725(1 + z). Substituting 1 + z = 66y −2/3 gives
y 2/3
≈ 1020 e−878y . (6)
1−y
We need to solve this for y. It looks ghastly; how do we do it? The key is in the exponential.
It is clear that if y is anywhere close to unity, the exponential factor will be vanishingly
small, so we will not get close to the answer. Therefore, y must be much less than unity. As
a result 1 − y is close to 1, so we get
2/3
y ≈ 1020 e−878y , (7)
where y ¿ 1. This is still a transcendental equation, but since the right hand side varies
wildly with different choices of y, we can converge fairly rapidly on the solution. For example,
y = 10−3 gives a right hand side of 1.5 × 1016 , which is much too large. y = 10−2 gives a right
hand side of 425, which is better but still too big. y = 3 × 10−2 gives 1.5 × 10−17 (!!), which
is much too small. Eventually, we find that y = 0.01373 does the trick, giving 1 + z = 1151.
The actual answer is 1 + z = 1089. What is the cause of the discrepancy?
The main problem is that in fact the ionization fraction is not quite what the Saha
equation would say, because the ionization is not in equilibrium. When an electron and
a proton combine to form hydrogen, a photon is emitted that has an outstanding chance
to ionize another hydrogen atom. This would leave no net change in the ionization. Other
processes are necessary, e.g., that the photon redshift enough before absorption that it cannot
ionize the atom, or that two-stage recombination happens (i.e., the electron and proton first
form an excited state of hydrogen, then radiate to the ground state; neither of those photons
could ionize hydrogen in its ground state). Consideration of the rate equations means that
there is residual ionization left as the universe expands. This higher ionization fraction
pushes the redshift of transparency to a lower value than it would otherwise have.
 By the way, a secondary contributor is that, given the recent rapid expansion of the
universe, the density at a given cosmic time is slightly different than we assumed. By itself,
however, this would only change the redshift to about 1140 instead of 1150.

Acoustic Oscillations

Typically, in space no one can hear you scream. However, if you scream loudly enough
then even the relatively low density in space will react with acoustic oscillations. In fact,
quantum theory predicts that very early in the universe fluctuations would have introduced
disturbances at all scales. Because of this generic prediction, we expect that the universe is
“ringing” at various special frequencies. In turn, this implies special angular scales at which
ripples in the CMB will be especially prominent. The precise angular scales of those ripples,
and their relative amplitudes, contain the information for which the CMB is justifiably
famous.
To understand this, let us consider the basics of acoustic oscillations. If a particular
region has an excess of pressure, then it will expand or propagate. If there are no boundaries,
then any frequency will do; this is the case when we talk in open air, for example. If instead
there are boundaries, then special wavelengths will resonate, leading to higher amplitudes
than others. This is the idea behind musical instruments. For example, a violin string is
clamped at both ends, so wavelengths that fit an integer number of times are most easily
excited and thus give a particular tune (determined by finger placement). Therefore, the
“fundamental” (fitting just once) and the overtones (fitting two, three, four, ..., times) are
evident. See Figure 1 for a mechanical analogy.

In the case of the universe as a whole, it is thought that fluctuations at all frequencies
(at roughly the same amplitude) were produced at the same moment. Therefore, there
were compressions and rarefactions on all scales, and thus also temperature variations on all
scales. A very low frequency would have managed only a small part of a cycle by the time the
universe became transparent, so the amplitude of variation would not be especially large. A
very high frequency would have gone through many oscillations, and also would not be high
amplitude (particularly because photons would tend to leak out of small, high-frequency
regions). However, a frequency such that maximum compression was reached just as the
universe became transparent would have an extra-high amplitude. So would double that
frequency, for which maximum compression and then maximum rarefaction occurred, the
latter just as the universe became transparent. This first overtone, however, is expected to
have a lower amplitude than the fundamental because gravity and pressure gradients would
then be working at cross purposes, as opposed to together at the fundamental. Extra strong
oscillations would in fact be expected at all harmonics of the fundamental, with amplitudes
that tend to diminish at high harmonic numbers because photons can more easily stream
out of smaller regions (and thus smooth out fluctuations in temperature).
As we will discuss in the next lecture, the analysis of these acoustic peaks has brought
an unprecedented level of precision to inferences about the basic cosmological parameters.
The reason this is so is that, fundamentally, physics at the CMB epoch and before (back to
a few microseconds after the Big Bang, or even earlier) is simple and well-understood. For
example, recall that at redshifts z ∼ 1000, the universe is to an excellent approximation flat,
with no cosmological constant. In addition, this is long before the era of structure formation,
so that all perturbations are linear and thus easy to treat. It is this firm grasp of the physics
that makes the CMB such a reliable tool. In the next lecture, we will discuss what we have
learned from it.

Intuition Builder

Suppose that matter consisted entirely of baryons, rather than being

mainly dark matter that has negligible interaction with radiation. What
qualitative differences might one see in terms of CMB fluctuations?
Fig. 1.— Acoustic oscillations in the early universe. When baryons are pushed to-
gether by gravity, they exert a pressure gradient that pushes them apart. This leads to
periodic compression and rarefaction. For an animated GIF version of this figure, see
http://background.uchicago.edu/∼whu/intermediate/plane.gif