Structure

 forma,on  

Newtonian perturbed fluid equations

Clustering and Collapse

Structure formation starts from the primordial power spectrum, and is driven by
gravity, which is described by General Relativity.

We can consider two regimes of structure formation:

Linear evolution (clustering)

While the density contrast is still small, δ << 1, the harmonic modes evolve
independently of each other, with a system of uncoupled equations.

Scales in this regime are called linear.

Linear scales follow the expansion (they are expanding with the comoving
background) but they expand with a slower rate (their density decreases with a
slower rate than a-3) à they are slowly forming over-dense structures à they
cluster, and their spatial distribution becomes increasingly correlated.
Non-linear evolution (collapse)

When the density contrast of a certain scale gets large, δ ~ 1, that scale becomes
non-linear.

More precisely, it is considered that the transition happens when the growing
dimensionless matter power spectrum reaches Δ2 (k) = 1

That scale decouples from the expansion (i.e. stops following the expansion) and
starts to decrease in proper size. The density constrast starts increasing very fast,
resulting in a gravitationally bound structure à there is a collapse and structure is
formed (in reality linear evolution does not really produces structure, just slowly
growing values of the density contrast).

Linearized perturbed Einstein equations are no longer valid to compute the non-
linear evolution.

Different approaches are needed à higher-order perturbation theory

(renormalization methods, ‘Feynman diagrams’), the Halo model approximation,
N-body simulations.
Newtonian Treatment

However, gravity is well described by Newtonian physics if

- scales are sub-Hubble,

i.e., much smaller than the scale of curvature of the spacetime

- the fluid is non-relativistic (in terms of velocity) v << c

- the fluid is non-relativistic (in terms of matter) p << ρ

Remember: H(a) is a decreasing function.

Since the comoving Hubble radius grows (and a comoving scale does not grow)
all scales gradually enter in the Hubble radius (becoming ‘Newtonian’) à
the universe becomes ‘less relativistic’ with time à the Newtonian description is
more accurate for the late universe (assuming no dark energy)
Under these conditions, the Newtonian treatment of a cosmological fluid
accurately describes structure formation.

In this description the evolution of the density contrast (and also of the potential
and the peculiar velocity) is fully described by 3 equations (2 conservation
equations + 1 constraint equation):

- continuity equation
- Euler equation
- Poisson equation

This set of equations has similar information than the set of 3 equations
used to describe the evolution of the homogeneous universe:

- Energy density conservation à a continuity equation

- Friedmann eq. à the zero-order equivalent to Poisson equation.

- Raychaudhuri eq. à a second-order equation of motion (like Euler eq.).

Note that even though it is originally a constraint equation (Einstein eq)
and not a conservation equation, they are related in the homogeneous case
where only 2 independent equations are needed.
Newtonian perturbed fluid equations

Continuity equation

The continuity equation is the equation of conservation of mass.

Extending it to the relativistic framework, it becomes an equation for the
conservation of energy.

This equation tells us there is no creation of energy à at a certain location of a fluid

within a fixed volume, energy (or mass) may only change in time because it may
flow to another location (within the volume).

So the local fluid flow, determined by its velocity, is responsible for a change
of density.

So besides expansion (that automatically induces a change of density), there is

another way to make the density change: the peculiar velocity à it is an
inhomogeneous contribution to the evolution of density, which was not
present in the energy conservation equation in the homogeneous universe.
In physical coordinates (r) the equation is:

u is the flow velocity of the fluid

Now, if the fluid flows in an evolving background (as it is the case in cosmology),
the volume changes in time à the physical coordinate ‘r’ changes with time,

and it has a dependence on both space and time. It is a “coordinate

that moves”.

So the total time derivative dρ/dt also includes spatial derivatives:

This is called a convective derivative or a material derivative.

Notice that it is also similar to a covariant derivative in General Relativity, where

the covariant derivative is equal to a partial derivative plus the terms that involves
the connections (that represents the change of the reference frame - in that case
not due to expansion but due to curvature - ).

Now, in the expanding universe, the velocity in the expression of the total
derivative, concerning the change of the physical position vector with time,
is the velocity of the expansion (the background velocity)

(no peculiar velocity involved here in the coordinates change),

where x represents the (x,y,z) comoving coordinates, the “truly spatial”

coordinates, i.e., independent of time

x=rx/a, y=ry/a, z=rz/a

Notice also that since the expansion makes the physical sizes to increase, while
the comoving sizes are constant, this is equivalent to consider that the
physical orthonormal reference frame shrinks (such that the corresponding
sizes increase), i.e.,

So the total time derivative of the density is,

includes local contribution only comoving contribution

(perturbation, peculiar, inhomogeneous) (background, homogeneous)
This deals with the first term of the continuity equation, dρ/dt .

Considering now the second term, the flow velocity of the fluid,

this may also be decomposed in local and comoving contributions à the

sum of the expansion velocity with the peculiar velocity v.

Remember: peculiar velocity is the inhomogeneous contribution to the velocity à it means that
the scale (the perturbation) does not follow exactly the expansion à the perturbed region may
expand slower than the mean universe.
Inserting the total velocity u and the comoving coordinates in the second term of
the continuity equation, we get the comoving continuity equation:

(see homework)

(in this derivation notice that the

divergence of the comoving
vector (x,y,z) is 3)

Notice that after all these steps the comoving continuity equation looks like:

the original equation (classical, non-expanding, with partial time

derivative and local peculiar velocity flow)

plus an extra term that shows the contribution of the expansion to the
evolution of density (i.e. the relativistic correction).
Let us now insert the density contrast:

we will have terms in the equation involving only background quantities

and other terms involving perturbations. We can then separate the equation in
zero-order and first-order “sub-equations”.

Let us also neglect the terms that are not linear in the perturbed quantities (they
are small <<1, when δ < 1). In this case the two sub-equations are:
(see homework)

the zero order equation - the background counterpart -

is the well-known continuity equation of the
homogeneous Universe (this is the reason why the
energy conservation equation was called the continuity
equation, even though apparently it did not look like the
continuity equation of a fluid)

the first-order equation - the comoving, perturbed and

linearized continuity equation - this one looks like the
continuity equation of a fluid
Poisson equation

Poisson eq. connects the gravitational potential with the matter density.

In physical coordinates (r) the equation is:

In comoving coordinates (x=r/a) it is:

Once again, we can insert the density contrast and separate the equation in zero and
first-order parts.

We may also define a homogeneous and a perturbed potential:

With this definition, the equation is immediately separated.

The first order equation is the

comoving, perturbed and
linearized Poisson equation:
and the zero-order equation,
the background counterpart is just:

We do not see immediately that this is the Friedmann equation (which is the
zero order counterpart of Poisson equation).

But the Friedmann equation tells us that

and so
3 (isotropy, x, y, z, equivalent)

Integrating it, we get

i.e., the “potential of the universe” is given by the square of the “velocity of the
universe”, which is consistent with the dimensional result of a potential being a
velocity square (ex: virial theorem).
Euler equation

The Euler eq. tells us how the velocity field changes in time given a gravitational
potential.

It is the equation of motion of the fluid. In physical coordinates (r) the equation is:

source

Again, like in the continuity equation, this is a total time derivative, i.e., the
acceleration of the fluid is not just the change of the velocity field at a given
position but it also depends on the expansion.

Inserting the total derivative (i.e., the comoving coordinates),

and the velocity perturbations through
we get,

It is a vectorial equation à
there is one equation per
component of v
The second term is the only zero-order one.

The zero-order equation is then:

Note that inserting here the background Poisson equation (integrated once with
respect to x, to get the gradient and not the Laplacian), this equation becomes:

Inserting here the Friedmann equation, to replace , we recover the

Raychadhuri equation (or the other way around substituting with the Raychadhuri
equation, we recover the Friedmann equation)

à the background counterpart of the the Euler equation is then a

combination of the Friedmann and the Raychaudhuri equations.

Indeed in the homogeneous universe, we saw that it is possible to combine those

two equations to obtain an energy conservation equation.
The first order Euler equation is written directly from the remaining terms:

Considering only linear terms in the perturbations, we get the comoving,

perturbed and linearized Euler equation:
Density contrast evolution equation

In summary, the 3 resulting (first-order) perturbed linearized equations are:

(we will follow the standard notation of using ϕ for the potential perturbation
instead of δϕ)

For the matter component

and the Poisson equation can also be written as

Now, combining the:

time derivative of the continuity equation with the

divergence of the Euler equation and
inserting Poisson equation,

we get the evolution equation for the density contrast: (see homework)

(linearized)

Let us now look at some properties of this equation.

2nd term

The second term is a friction term à it works against the growth.

Its coefficient is 2H(a),

showing that the background expansion works against structure

formation à it is called the Hubble drag

3rd term

The third term has a fixed sign (it is always negative), so the solution is
monotonous.

- a decreasing or increasing exponential, or power-law (depending on

the relation between the coefficients of the second and third terms) -

it is not an oscillating solution à the density contrast may grow

The cosmological parameters (and in particular the parameters of the
homogeneous universe) are important for the efficiency of structure
formation.

For example:
A larger amount of matter Ωm à larger coefficient of the 3rd term à solution with
faster growth à more matter favours structure formation
(the physical reason behind it is that a higher Ωm creates a ‘deeper potential well’
through Poisson equation, which favours clustering).

Notice that the mean matter density Ωm and the density contrast δ are independent
quantities. Unlike the density, there can be other cosmological inhomogeneous
fields where the amplitude of clustering is lower if Ωm is larger (it is the case of the
temperature anisotropies of CMB).
Now, the density contrast is also a function of space.

The equation we derived only tells us about the time evolution of the density
contrast.

Can its spatial variation be separated from the time evolution?

We need to Fourier transform δ(x,t) and study the time evolution of all δk
scales.

Fourier transforming the 3 perturbed quantities that appear in the equations (density
contrast, peculiar velocity and potential) the 3 (non-linearized) equations become:

Continuity

Euler

Poisson
Note that the evolution of a scale δk depends on a sum over all other scales
k’ à there is mode coupling.

But we also see that mode coupling is higher-order in the perturbed quantities.
It goes to zero if we linearize equations:

Continuity
These equations are almost
identical to the linear equations in
real space.
Euler
The only difference is that spatial
derivatives are now replaced by
Poisson multiplications with the scale k.

This means that in the linear regime, the evolution of the modes is independent.

Each scale has its own evolving equation, with no influence from the other scales.

This simplifies a lot the problem à the equations are the same for all scales (in the
Newtonian approach, i.e., not valid for relativistic large super-Hubble scales).
What about other vector and tensor perturbations - are they relevant for the
evolution of the density contrast?

Remember that the wavenumber k is a 3D vector. However, from the isotropy (in the
statistical cosmological principle), we just need to consider the modulus of k.
In other words, the only relevant direction is the one set by the flow of the
clustering matter (the 1D longitudinal k mode). The transversal modes (i.e. the
ones orthogonal to the direction of the flow v) are suppressed

So the internal product v.k in the continuity equation implies that only the longitudinal
component of the velocity field contributes to the growth, i.e., if v.k = 0 à δ would not
grow à the orthogonal component of the velocity field produces no growth of
perturbations.
(this just means that if the matter flow does not go to the center of the potential, there
is no growth of density contrast).

In addition if we impose irrotational and adiabatic initial conditions for the fluid
perturbations in the energy-momentum tensor à the velocity perturbation is a
pure potential flow (representing radial collapsing directions, i.e., parallel with the flow)
à the velocity field is a gradient field, it is described by a scalar

à Any of these two arguments show that the vector part of v does not impact the
growth of δ.
Moreover, vector metric perturbations decay à also no vector contributions from
the metric perturbations to the growth of structure.

Finally, notice no tensor contribution appear in the perturbed fluid equations.

All these reasons show that:

metric tensor perturbations,

metric vector perturbations

matter vector perturbations

are not relevant here à structure formation is a scalar process.

 Now, combining the three linearized equations in Fourier space, we get the
single equation that describes the evolution of all modes in linear sub-Hubble
regimes.

This equation looks exactly the same as the one in real space (and later on we will
drop the k index) :

Note that even though the equation is the same for all scales, this does not mean
that all scales will have the exact same evolution, since there will be differences
because of:

- domain of validity of the Newtonian approach (size of the scale)

- cosmological component considered (dark matter, baryonic matter, radiation)

- epoch of the Universe (matter, radiation or dark energy dominated regimes)

We can now start computing the evolution of the density contrast of the various
components of the cosmological fluid, and in the various epochs.

The various epochs and components in the Universe

Remember that the mean densities (zero order) of the various components of the
cosmological fluid evolve differently with time à the Universe is dominated by
one component at a time, defining different epochs.

This is easily seen by computing the time-evolution of the mean density, from the
zero-order continuity equation
Matter (dark matter and baryonic matter)

is defined by having pressure p=0, i.e., w=0 (equation-of-state)

Inserting p=0 in the continuity equation, the solution is easy to find:

ρ0 is an integration constant, i.e., an initial condition, the density at a0, and as we know,
the initial conditions define the model parameters.

If we choose to define the initial condition today, at a0 = 1, then ρ0 is the density today,
which defines the well-known matter density cosmological parameter,

Note that the parameter Ωm, not only parameterizes the matter density today, but
also depends on the Hubble constant. So the values of Ωm in two Universes
with the same matter densities but different expansion rates, would be
different.

For this reason, it is useful to introduce the physical matter density parameter,
defined as ωm = Ωm h2
Radiation

is a cosmological component consisting on relativistic particles, with v=c,

and so eventual pressure perturbations propagate with speed of sound equal to the
speed of light.

However this does not imply p = ρ, but p=ρ/3, i.e., w=1/3, because the flux of
photons hitting a given surface may be spread over the 3D space, and so the net
pressure over a surface is on average 1/3.

Note that in a 5D space-time (with 4 spatial directions), the radiation equation of

state would be w=1/4

Note that, even in 4D space-time, if the Universe was not isotropic, the pressure
would not be isotropically distributed, and the mean pressure could be different
from 1/3.

Anyway, inserting p=ρ/3 in the continuity equation, the solution is easy to find:

where ρ0 defines the radiation density parameter

Note that the energy density of a radiation fluid dilutes faster than a matter fluid,
the extra a-1 factor explains the extra dilution due to the redshift of the photon
frequency.

Notice also that if we consider that the Universe is a black-body with temperature T
and energy density ρr, the energy is related to the temperature as ρr ~ T4 (Stefan-
Boltzmann law) à the temperature of the Universe decreases linearly with a,
and this is why the temperature is used as a “time scale” in the early universe (as the
redshift is also used).
Dark energy

is a component that produces accelerated expansion.

From the second Friedmann equation (Raychadhuri equation)

we see that the acceleration is positive if w < -1/3

Notice that in GR pressure is a source of gravity (just like mass or energy density), a
negative pressure contributes to a repulsion, and may accelerate the expansion.

There are many models of dark energy. The simplest and most used one is the
cosmological constant, where w = -1

Inserting p = - ρ in the continuity equation, we find the solution,

ρ = constant

i.e. the dark energy cosmological constant does not dilute as the Universe expands.
The evolution rate of the densities is not all the information needed to
determine which is the dominating component, since the amplitudes of the
densities depend on the initial conditions.

For example, consider the three following ΛCDM models:

i) ii) iii)
Λ

m
r

ln a ln a ln a

In model i) all components start with the same amplitude à this model is always
dominated by dark energy

In model iii) radiation never dominates

The concordance model looks like model ii)

we saw that from the measurement of the CMB temperature à Ωr ~ 0.00008,
and also that the various cosmological probes point to Ωm ~ 0.3 and ΩΛ ~ 0.7
Concordance model: given these amplitudes at a=1, we can search for valid times
in the history of the Universe (with a>0 and a<1) where the densities were identical:

ρr (aeq) = ρm (aeq) à aeq = Ωr / Ωm ~ 0.00027 à zeq ~ 3500

ρΛ (at) = ρr (at) à at = (Ωr / ΩΛ)1/4 ~ 0.1 à zt ~ 8

ρΛ (aΛ) = ρm (aΛ) à aΛ = (Ωm / ΩΛ)1/3 ~ 0.75 à zΛ ~ 0.3

We see that the concordance Universe starts to be radiation dominated, until the
equality redshift z ~ 3500, where it starts to be matter dominated. At z ~8 the
decreasing radiation density reaches the value of the cosmological constant, but this
does not define another regime since matter continues to dominate (at z ~8, Ωr = ΩΛ
~ 0.7 and Ωm ~270). Finally, from z ~0.3 until today and into the future, the universe
is dark energy dominated à this defines the three epochs of the concordance
Universe.

(Note that the dark energy transition redshift does not mean that the Universe expansion
started to accelerate at z = 0.3. It is just the redshift where dark energy and matter have the
same mean densities).