Notes
Notes
Notes
In GR, the quantities xi might be metric coefficients, densities, velocities, etc. GR pertur-
bation theory can be tricky for two reasons.
(1) The equations are very complicated. This is a problem of practice, not of principle.
(2) In a perturbed universe, the “natural” coordinate choice is no longer obvious. Different
coordinate choices have different equations. Results for physically measurable quantities
should be the same, but the descriptions can look quite different.
For scales l cH −1 and velocities v c, one can usually get by with Newtonian linear
perturbation theory.
In an inertial frame, the equations governing the density ρ(r, t) and velocity u(r, t) of an
ideal fluid are:
∂ρ ~ r · (ρu) = 0 (mass conservation)
Continuity equation : +∇
∂t r
∂u ~ r )u = −∇
~ r Φ (momentum conservation)
Euler equation : + (u · ∇
∂t r
Poisson equation : ∇2r Φ = 4πGρ.
Transformation of variables
We would like to have these equations in comoving coordinates x = r/a(t) with “peculiar”
˙ − ȧx = aẋ.
velocity v = u − (ȧ/a)r = (ax)
We want to use a quantity that will be small when perturbations are small, so define
the dimensionless density contrast δ(x, t) by ρ = ρb (t) [1 + δ(x, t)], ρb = background
density ∝ 1/a3 .
51
A873: Cosmology Course Notes
By the chain rule, the time derivative of a function f at fixed time t and comoving position
x = r/a is
∂f ∂f ∂r ~ ∂f ȧ r ~
= + · ∇r f = + · a ∇r f .
∂t x ∂t r ∂t x ∂t r a a
Rearranging terms yields the expression for the time derivative of f at fixed t and r:
∂f ∂f ȧ ~ ~ ≡ a∇~ r.
= − x · ∇f, ∇
∂t r ∂t x a
∂δ 1~
Continuity : + ∇ · [(1 + δ)v] = 0
∂t a
∂v 1 ~ + ȧ v = − 1 ∇φ
~
Euler : + (v · ∇)v
∂t a a a
2
Poisson : ∇2 φ = 4πGρb a2 δ with φ ≡ Φ − πGρb a2 x2 .
3
The one qualitatively new feature is the “friction” term ȧa v in the Euler equation. It
~ vanishes.
drags v to 0 if ∇φ
This is just the “kinematic redshift” that we encountered long ago; it reflects the use of
expanding, non-inertial coordinates.
It will have a crucial consequence: density contrasts in an expanding universe grow as
power laws of time (for Ωm = 1) instead of exponentials.
So far, the equations are general, within the Newtonian limit and the assumption that
~
pressure gradients are negligible, ∇p/ρ ~
∇φ.
The linear approximation is obtained by keeping only the terms that are first order in δ
or v:
∂δ 1~ ∂v ȧ 1~
+ ∇ · v = 0, + v + ∇φ = 0.
∂t a ∂t a a
Eliminate v by taking time derivative of first equation, (1/a) × divergence of second, sub-
tracting, and substituting using the continuity and Poisson equations.
Result:
∂ 2δ ȧ ∂δ
2
+2 = 4πGρb δ.
∂t a ∂t
This is a second-order differential equation for δ(x, t) with a growing and decaying mode
solution:
δ(x, t) = A(x)D1 (t) + B(x)D2 (t).
52
A873: Cosmology Course Notes
We can rewrite the equation for δ as an equation for the growth factor
3
D̈ + 2H(z)Ḋ − Ωm,0 H02 (1 + z)3 D = 0.
2
If we assume that density fluctuations appear at some very early time, then at a much
later time we care only about the growing mode D1 (t), which we can just call D(t). Thus,
D(t)
δ(x, t) = δ(x, ti ) , D(t) ∝ a(t) ∝ t2/3 for Ωm = 1.
D(ti )
The condition for a perturbation to be growing mode is δ̇ = δHf (Ωm ); growing mode per-
turbations are those for which the velocity divergence correctly reinforces the gravitational
growth.
53
A873: Cosmology Course Notes
Fourier description
Each δk = Ak eiθk is a complex number, the Fourier amplitude. (Beware that different
authors choose different conventions on where to put the 2π’s.)
If the phases θk are uncorrelated, then the field δ(x) is Gaussian. This implies that the
1-point probability distribution function of δ is Gaussian,
2
/2σ 2
P (δ) = (2πσ 2 )−1/2 e−δ ,
which tells the rms fluctuation as a function of scale (whether or not the field is Gaussian).
In linear theory, each Fourier mode evolves independently, δk (t) ∝ D(t).
The spatial scale is λ = 2π/k, so even if small scale modes have become nonlinear, 4πk 3 δk >
∼
1, large scale modes may still follow linear theory. (This statement is not obviously true,
but it holds in most circumstances.)
On scales in the linear regime, the shape of the power spectrum is preserved, and its
amplitude grows ∝ D 2 (t).
54
A873: Cosmology Course Notes
We are often interested in the properties of the density field after it has been smoothed,
by convolution with a window function WR (r) of characteristic radius R.
Common choices for W (r) include a top hat,
h ri
W (r) = Θ 1 −
R
and a Gaussian 2
/2R2
W (r) = e−r .
and Z
f (kR) =
W d3 rWR (r)e−2πik·r
For a power spectrum P (k) ∝ k n with n > −3, σ 2 (R) ∝ R−(3+n) increases towards smaller
scales and is dominated by waves with k ∼ 1/(2πR).
These results hold generally, not just in linear theory, provided that one uses the non-linear
P (k) rather than the linear P (k). However, if one starts with Gaussian fluctuations, the
probability distribution P (δ) becomes non-Gaussian in the non-linear regime.
55
A873: Cosmology Course Notes
A scale that exits the Hubble radius during inflation with fluctuation amplitude δH reenters
the Hubble radius during the radiation or matter-dominated era with approximately the
same amplitude.
If the fluctuations are scale-invariant, then this amplitude δH is constant, but it applies at
a different time for each scale.
To understand the implications of this, it is useful to blow up the upper right of the “last-in
first-out” inflation diagram (see figures at end of section).
Consider fluctuations that reenter the Hubble radius during the matter-dominated era,
when D1 (a) ∝ a.
What does δH = const. imply for the power spectrum P (k, t) at a fixed time, where k
represents the comoving wavenumber 2π/λ?
Scale-invariance implies
4πk 3 P (k, tc ) ≈ δH
2
≈ const.,
where tc is the time when the comoving scale λ “crosses” (is equal to) the Hubble radius.
The condition for crossing the Hubble radius is:
a2 (t) 2
δH a2 (t) 2 2
P (k, t) = P (k, tc ) = ∝ δH a k.
a2 (tc ) 4πk 3 a2 (tc )
Bottom line: scale-invariant fluctuations produce P (k, t) ∝ k on scales that reenter the
Hubble radius during the matter dominated era.
56
A873: Cosmology Course Notes
(in practice, there is logarithmic growth, hence an additional factor of log k.)
The transition between scales that enter in the radiation dominated regime and scales
that enter in the matter dominated regime is a slow one, so the change from P (k) ∝ k to
P (k) ∝ k −3 is gradual.
The linear theory CDM power spectrum is written
In practice, this scale is slightly modified by the gravitational effects of the baryon com-
ponent.
Since inflation may not produce perfectly scale-invariant fluctuations, the spectrum is often
written
P (k, t) ∝ k n T 2 (k)D12 (t),
where n = 1 corresponds to scale-invariant inflationary fluctuations and n 6= 1 is called a
“tilted” spectrum.
Typical inflation models predict n = 0.9 − 1, but larger tilts (and n > 1) are possible.
The effective value of n could also change with scale; this is called “running of the spectral
index.” It is expected to be small in typical inflation models.
A feature in the inflaton potential can also produce “broken” scale invariance, with a sharp
feature in the power spectrum.
57
A873: Cosmology Course Notes
If the dark matter is mildly relativistic at the time that a scale reenters the Hubble radius,
then it will stream preferentially out of overdense regions and into underdense regions,
ironing out fluctuations on that scale.
If the dark matter consisted of neutrinos with rest mass ∼ 10 − 30 eV, then they would
be mildly relativistic at teq (essentially by definition, since their number densities are
similar to those of photons), and fluctuations on scales smaller than cteq would be strongly
suppressed, radically changing T (k).
This gives a “hot dark matter” power spectrum. A pure hot dark matter model fails
because the absence of small scale fluctuations means that small objects are unable to
form at high redshift.
If the dark matter particle has a mass ∼ 1 keV, then primordial fluctuations are suppressed
on sub-galactic scales but not on galactic scales.
Galaxies can still form fairly early, but they are not preceded by generations of smaller
collapsed structures.
This “warm” dark matter model is still viable and offers some attractive features.
However, the “cold” dark matter model, in which particles are massive enough that fluctu-
ations are not erased on any cosmologically interesting scale, seems more natural, because
the particle mass does not have to be fine-tuned and lies more in the range expected from
particle physics models.
Baryon wiggles
58
A873: Cosmology Course Notes
59