Defence talk

Research Statement - Aniket Agrawal

The primary topic of my thesis was non-Gaussianity. In the first part I worked on the non-Gaussianity
of PGWs. The second part is on NG in LSS. Since time is so short, I will only talk about the first part
relating to NG of PGWs.
Let me now describe the work I did for PGWs. I want to consider tensor fluctuations of the FRW
metric during inflation. So I write

ds2 = −dt2 + a2 (t)[δij + hij ]dxi dxj (1)

hij here are transverse traceless dof so they are GWs. I have also assumed de-Sitter expansion through-
out, so that H is constant. By using this, and writing the Einstein equation, we get

2hij = Sij (2)

this equation of motion for these GWs. Now generically, the solution to this equation can be expressed
as the sum of a homogeneous component, and an inhomogeneous component

hij = hH I
ij + hij . (3)

In the absence of any source term, we only have the homogeneous component. This is the component
that sourced by quantum fluctuations of the metric. At first order these obey a linear eq of motion,

2hij = 0 . (4)

By expanding hij in a Fourier series, and imposing commutation relations on hij we can show that the
GWs from vacuum fluctuations have a power spectrum that looks like this
!2
L/R H 2
Ph (k) = , (5)
MP k3

and that they are parity invariant, i.e. the power spectrum of both left- and right-handed GWs is the
same. However, GR is a non-linear theory so we can also expand this eq of motion to second order. We
only need to go up to second order to get the tree-level bispectrum of these GWs.
When we do that, we get an equation that looks like this.

2hij = O(h2 ) . (6)

We can solve this as follows : we expand the field into a perturbation series,

ψ̂kp (τ ) = ψ̂1p (τ, k) + ψ̂2p (τ, k) + . . . . (7)

such that the first term obeys a linear eom second obeys a second-order eom and so on. Then the second
order term is given in terms of the source function using the Greens function as
Z 0
R (2)
ψ̂2 (τ, k) = dη Gψ (τ, η, k)Ŝpq (η, k)eLpq (k) . (8)
−∞

1
Then, at tree-level the bispectrum of GWs is given as
D E D E
ψ̂ R (τ, k1 )ψ̂ R (τ, k2 )ψ̂ R (τ, k3 ) = ψ̂1R (τ, k1 )ψ̂1R (τ, k2 )ψ̂2R (τ, k3 )
D E D E
R R R R R R
+ ψ̂1 (τ, k1 )ψ̂2 (τ, k2 )ψ̂1 (τ, k3 ) + ψ̂2 (τ, k1 )ψ̂1 (τ, k2 )ψ̂1 (τ, k3 ) . (9)

For vacuum fluctuations, we find that the bispectrum of 3 RH GWs peaks in the squeezed limit.
What is that? The bispectrum is characterised using the triangle...
We can characterise the degree of NG using a parameter analogous to fNL, which is defined as the
ratio Bh (k, k, k)/Ph2 (k). For 3 RH GWs this ratio has the value 3.6. Thus, even vacuum fluctuations
are not exactly Gaussian.
The source term on the RHS can also have non-zero transverse traceless components. Let me describe
one case in which it is non-zero, which concerns one of the two main parts of my thesis. So suppose we
have standard GR, " #
Z
√ 2
M P
S = dτ d3 x −g R , (10)
2
we have an inflaton field, which has all the general features required for inflation so it rolls down its
potential slowly and generates scalar perturbations.
" #
Z
√ 2
MP
S = dτ d3 x −g R + Lφ . (11)
2

Now lets say we also have an axion in the Universe that does not couple to the inflaton except at
the background level and it has an energy density much smaller than that of the inflaton. In this case
it does not participate in inflation directly.
" #
Z
√ 2
M P
S = dτ d3 x −g R + Lφ + 1/2(∂χ)2 + U (χ) . (12)
2

This axion then naturally couples to SU(2) gauge fields.

" #
Z
√ 2
M P
S = dτ d3 x −g R + Lφ + 1/2(∂χ)2 + U (χ) + 1/4F 2 + λ/4f χF F̃ . (13)
2

We also require that these gauge fields have a small energy density compared to the inflaton so that
they also do not disturb the bkg evolution.
As the Universe inflates, the axion rolls down its potential. Because of this coupling to the gauge
fields, as it rolls down, some of its KE is lost to the gauge fields. The SU(2) gauge fields have a tensor
dof, which also gets excited by the KE of the rolling axion. This tensor dof can then linearly source
GWs.
Let me illustrate this on a graph. X axis here shows the conformal time, scaled by the wavenumber.
Right side corresponds to beginning of inflation while far left corresponds to end of inflation. x = 1
denotes horizon crossing for the mode with wavenumber k. The tensor dof starts off as a vacuum
fluctuation deep inside the horizon and gets exponentially amplified a little before it crosses the horizon.
After horizon crossing, its amplitude decays as 1/a3 , and becomes constant after that. The GW is
sourced by this tensor gauge field, so it peaks a little after horizon crossing, and then stays constant on
super-horizon scales. This can then later re-enter the horizon and seed the B-mode polarisation of the
CMB.

2
 The interesting question now is is the amplitude of these GWs larger or smaller than vacuum GWs?
To answer this, we need to know what the amplitude of these GWs depend upon? The amplitude should
clearly depend on how much the gauge field is amplified. This depends on the strength of the coupling
between the gauge fields, encoded in this mQ parameter. Next, these gauge fields need to generate GWs,
and so the amplitude must depend on the efficiency of this coupling. Because the two are coupled to
each other only via gravity, this efficiency depends on the energy density fraction of the gauge fields,
labelled as B . By choosing these two parameters then, we can control the amplitude of these sourced
GWs. It turns out that their amplitude can be much larger than that of the vacuum fluctuations. So,
when we see primordial B-modes, they could have arisen from sourced GWs. In fact, these parameters
can also be used to control the scale dependence of the power spectrum. If we assume that both mQ
and epsB are constant in time, as I do in the thesis, the power spectrum is actually scale-independent
and is given by

PhR = B |F (mQ )|2 /π 2 (H/MP )2 (14)

where F is a complicated function of mQ whose form is roughly given as ∼ e1.8mQ .
Next we consider the NG of these GWs. 3 different interactions produce a non-zero GW bispectrum
in this model. Before discussing them in detail, let us consider their relative sizes.
Remember that the gauge field is exponentially amplified...
So, as a first estimate we see that the size should be proportional to the no of t-lines we have in
each diagram. This means that we expect diagram 1 to be the most dominant, and diagram 2 and 3
to be exponentially smaller. However, we also saw that the gauge field and the metric perturbation
are coupled to each other only via gravity, so that the cross here should depend on the energy density
fraction of the gauge field. For spectator gauge fields that is in fact quite small and we find that diagram
1 and 2 actually have similar amplitudes. Let us now look at the individual diagrams in detail.
The first diagram arises from self-interactions of the gauge field. Such an interaction is absent in an
Abelian theory, so this is a distinct signature of our model. The non-zero bispectrum originates from
the non-linear sourcing of the second order gauge field by first order gauge fields, which are Gaussian.
This diagram peaks at r2 = r3 ∼ 0.6, that is when two long wavelength modes source a short wavelength
mode. This bispectrum is also exactly zero in the folded limit as before.
This diagram arises from the non-linear sourcing of the metric perturbation by first order gauge
fields. As before the bispectrum is zero in the folded limit for this diagram too. What is interesting is
that this one peaks in the equilateral limit. The peak in the equilateral limit arises simply because here
the metric peturbation is directly sourced by the gauge fields, which both reach a peak close to horizon
crossing and decay afterwards. If the sourced metric perturbation has a smaller wavelength, it reaches
a peak but subsequently decays inside the horizon. On the other hand, if it has a longer wavelength,
the gauge fields do not attain their peak value when the metric perturbation crosses the horizon and
thus cannot source it efficiently.
For the third diagram, the non-zero bispectrum arises from interactions between the first order
metric perturbation and the first order gauge field. One of the first things to note about this diagram
is its extremely small amplitude. We see that for these plots, it is almost 7 orders of magnitude smaller
than the other two, as we already expected. Here also, we have a peak in the equilateral limit, owing
to a similar argument as for the previous diagram. In any case, this diagram is totally swamped by the
other two, so I will not consider it for the rest of the talk.
The non-Gaussianity parameter we used earlier, Bh /Ph2 , for this model depends inversely on the
energy density fraction of the gauge field. This is similar to the curvaton mechanism. We can plot
this ratio normalised by the energy density fraction as a function of the coupling strength mQ . Not
surprisingly, the non-Gaussianity is a monotonically increasing function of the coupling strength. What
is interesting is the exponential dependence of non-Gaussianity on this coupling strength. This is again

3
a reflection of the exponential amplification of the gauge fields. In particular, it implies that the GWs
sourced in this model, can be highly non-Gaussian. Thus, we can use the GW bispectrum to deduce
the origin of primordial B-modes.