Can We Constrain or Even Rule Out Multifield Inflation Models?

Some Very Basic Background
Multifield Inflation
My Work
Future Directions and Conclusion
Can We Constrain or Even Rule Out Multifield Inflation Models?
Godfrey Leung1
ppxgl@nottingham.ac.uk
Supervisor - Prof Ed Copeland

1
School of Physics and Astronomy, University of Nottingham
2
Astronomy Centre, University of Sussex
collaboration work with - E.R.M.Tarrant2 and C.T.Byrnes2

arXiv:1206.5196,arXiv:1303.4678
22nd Jan, Heidelberg
Godfrey Leung Reheating Makes Life Harder

My Work
Overview

Inflation
Primordial observables
Non–Gaussianity
Current Constraints
Why Consider Multifield Inflation?
Motivation - why study the effect of (p)reheating on primordial observables?
Models Considered, Recap of δN formalism
Results
Slow-roll Predictions Revisited
Beyond Slow-roll
Bispectrum, fNL
Trispectrum, τNL
Scale Dependence of NonG
Consistency Relations

Some Very Basic Background Inflation
Multifield Inflation Primordial Observables
My Work NonG
Future Directions and Conclusion Current Constraints
Problems of Standard Big Bang Cosmology
Big Bang Cosmology is very successful in describing the history of our Universe,
e.g. Big Bang nucleosynthesis
However, there are some unresolved issues
For instance, horizon and flatness problems, unwanted relics from spontaneous
symmetry breaking
no explanation for primordial fluctuations that seed structure formation
requires very fine-tuned initial conditions...

My Work NonG
Inflation, a paradigm for Early Universe
It is believed that our Universe has undergone a dramatic exponential expansion

of spacetime in the very Early Universe [Guth 80]
solves the horizon and flatness problems of Standard Big Bang
Quantum fluctuations naturally give rise to primordial fluctuations as they are
stretched out of horizon
Simplest model involves a canonical scalar field slowly rolling down a flat
potential [Linde 82]
V (ϕ)
Yet it remains a phenomenological model...

My Work NonG
Primordial Inflationary Observables

To make connection with particle physics models, one has to identify the (scalar)
field(s) that plays the role of inflaton
But there are vast number of theoretical models in the market...[Encyclopaedia,
see J.Martin 13]
Important to compare model predictions to observations
Eg. Measure CMB anisotropies to high precision to study the primor-
dial fluctuations (e.g. gauge-invariant curvature perturbation ζ) and their statistics
Credits: ESA, Planck Collaboration
Spectral index, defined by dlnPζ (k)/dlnk, measures the small departures from a
scale-invariant spectrum
Tensor-to-scalar ratio, r , measures the relative strength of tensor gravitational
fluctuations to scalar fluctuations
Roughly speaking, nζ tells us something about the slope of inflaton potential,
r tells the energy scale of inflation
My Work NonG
Beyond Power Spectrum: Non-Gaussianity
NonG, a measure about departure from a Gaussian statistics in primordial

perturbations, ζ
No-go theorem: simple single–field model predicts negligible level of NonG,
fNL ∝ O() [Maldacena 05]
more complicated models however could give large detectable NonG
For example, reduced sound speed cs in DBI models (equilateral shape) and
non-Bunch-Davies(BD) initial vacuum (orthogonal shape)
could shed light about the inflation model in play [review, Chen 10]

My Work NonG
Quantifying Non–Gaussianity
Amplitude, measured by non-linearity parameters

bispectrum hζζζi, fNL
trispectrum hζζζζi, τNL and gNL
Credits: KICP, Uni of Chicago
dlnfNL (k)
scale–dependence, nfNL , nτNL and ngNL , defined by dlnk
for example

My Work NonG
Local–type NonG
In general, fNL , τNL and gNL are ki (shape) dependent

focus on local–shape, which could be generated in multifield models for
superhorizon modes
Definition, ζ = ζG + (3/5)fNL (ζG2 − < ζG2 >) + (9/24)gNL ζG3
2
τNL related to terms of the same form as fNL
peaks in the squeezed limit for fNL (k1 → 0), collapsed limit for τNL
(k1 + k2 → 0)
k2 k3 k2
k1
k3 k4 k1 .
gNL peaks in the squeezed limit (ki , kj → 0)
k1 k4 k1
k4
k3 k2
k2 k3 .

My Work NonG
Current Constraints, Planck2013

power spectrum is nearly scale invariant, with red tilt nζ = 0.9603 ± 0.0073
negligible tensor fluctuations, r < 0.12 (95%CL)
no primordial non-gaussianity, fNL consistent with O(1) in all shapes
Credits: Planck Collaboration 2013
Conclusion: Consistent with simple single field inflation

End of story for more complicated model setup like multifield inflation?
Motivation
(P)reheating
My Work
Models Considered
Why not? ”Absence of Evidence is not the Evidence of Absence”!

Motivation
(P)reheating
My Work
Models Considered
Multifield models are more natural in the particle physics setup of inflation models
For example, in string theory, other scalar degrees of freedom also become light in
general if one scalar field is made light
They offer richer phenomenology, e.g. hybrid inflation [Linde 91] and curvaton
models [Lyth 02]
Though they could give potential large nonG, fNL > O(5), they generically gives
Gaussian statistics just as single-field model
They are not ruled out!

Motivation
(P)reheating
My Work
Models Considered
Difference Between Single and Multifield Models
One significant difference between single and multifield is the presence of

isocurvature modes
isocurvature mode = perturbations in direction orthogonal to background
inflationary trajectories
curvature perturbation, ζ, is not conserved after horizon-crossing till adiabaticity
condition is reached
Adiabaticity condition
δρ ρ̇
=
δP Ṗ
or equivalently, the classical trajectories in the phase space of H converge
Thus ζ and related observables can continue to evolve

Motivation
(P)reheating
My Work
Models Considered
Motivation
Single-field inflation models are well classified and tested with Planck data
[Planck collaboration 13]
Yet a similar picture is missing for multifield models...
Technical difficulties: even for same models, different initial conditions at
horizon-crossing could lead to very different predictions
Recently, by studying N-quadratic models, multifield models could still be
predictive despite the issue of initial conditions [R.Easther et.al.13]
Yet (p)reheating is not taken into account in the analysis...

Motivation
(P)reheating
My Work
Models Considered
Why Consider (P)reheating?
(P)reheating is an important part of inflationary model building

After inflation ends, everything dilutes away except the energy density of the
inflaton φ
Energy density must be converted back to other particles including SM particles
(P)reheating occurs after slow-roll, when inflaton oscillating about its minimum
Offer rich phenomenology, e.g. anisotropic gravitational background [Rajantie
et.al. 13] and formation of solitons like oscillons [M.Amin 11]

Motivation
(P)reheating
My Work
Models Considered
Why Consider (P)reheating? (con’t)
as discussed, isocurvature mode can source curvature perturbation

to compare with observations, observables related to ζ should be evaluated at the
point where ζ is conserved
isocurvature modes may not be exhausted at the end of inflation
reheating thus could play a role in evolution of ζ
Questions:
1. Predictions evaluated during slow-roll stage reliable?
2. If observables do evolve during (p)reheating, can we use this fact to constrain physics of
(p)reheating?

Motivation
(P)reheating
My Work
Models Considered
Models Considered
We study the class of multifield models
√
Z
1 1
Sm = d 4 x −g − g µν ∂µ ϕ∂ν ϕ − g µν ∂µ χ∂ν χ − W (ϕ, χ)
2 2
In particular, we study the two-field models with canonical KE terms

We consider simple perturbative reheating and model it by adding a friction term
in EOM [Kofman 96]
φ¨I + (3H + ΓI )φ˙I + W,φI = 0
Γ is switched on after first passage through the minima after slow-roll

For simplicity, we take Γ to be constant

Motivation
(P)reheating
My Work
Models Considered
Examples
We study a number of two–field models, where minima exist in one or both field
directions
In particular, potentials with separable form
2
1. one min, quadratic exponential potential, W = W0 χ2 e −λϕ /Mp (left)
2. two min, effective N-flation model, W = W0 12 m2 χ2 + Λ4 1 − cos 2π

f ϕ (right)

Motivation
(P)reheating
My Work
Models Considered
Recap of δN Formalism
To study the evolution of ζ during reheating, we apply the δN formalism
δN formalism based on separate universe approximation
On superhorizon scale, we can treat different patches of the universe as ’separate
universes’ that evolve independently of each other, with slightly different initial
conditions [Wands et.al. astro-ph/0003278]
Under this picture, we can write ζ in terms of the difference in the number of
e–folds of expansion for different patches
X 1X
ζ = δN = N,I δϕI ∗ + N,IJ δϕI ∗ δϕJ∗ + · · ·
I
2 IJ

Motivation
(P)reheating
My Work
Models Considered
Recap of δN formalism
Written in terms of the δN derivatives, the key primordial observables fNL , nζ , r

are
8
r = P 2
,
I N,I
P
2 IJ ϕ̇∗J N,JI N,I
nζ − 1 = −2∗ + P 2
,
H∗ K N,K
P
(4) 5 IJ N,IJ N,I N,J
fNL = 2 .
6
P
2
I N,I
(4)
where fNL is the shape–independent part of fNL
local is dominated by f (4)
For canonical models, fNL NL [Vernizzi & Wands 06]

Influence of Reheating on Observables
My Work
Slow-roll Predictions
Explicit analytic formulae of δN derivatives exist for separable potentials, i.e.

For example sum-separable potential [Vernizzi & Wands 06, Elliston et.al. 12]
-First and second order δN derivatives
-Third order

My Work
Slow-roll Predictions
From analytic formulae, one can see large nonG only possible if the fields start
close to a ridge/valley at horizon-crossing
Evolution during slow-roll [Elliston et.al. 12 & Anderson et.al. 12]
4000
(6/5 fNL)2
2500 3000
2250 (6/5 fNL)
2
τNL
2000 τNL 2000 gNL
1750 g
NL
1000
1500
1250
0
1000
750 −1000
500
250 −2000
0
−3000
−250 0 10 20 30 40 50 60
0 10 20 30 40 50 60 N
N
quadratic exponential potential (left), N-flation (right)

in general, adiabaticity condition is not reached by the end of inflation (except in
N-flation for some model parameters)

My Work
Beyond Slow-Roll, Bispectrum, fNL

One min case, quadratic exponential potential (left: λ = 0.06; right: λ = 0.05,
ϕ∗ = 10−3 Mp , χ∗ = 16.0Mp )
50 100
Γχ=√10-1 Γχ=√10-1
Γχ=√10-3 Γχ=√10-3
Γχ=√10-5 Γχ=√10-5
25 50
fNL 0 fNL 0
−25 −50
−50 −100
64 65 66 67 68 69 64 65 66 67 68 69 70
N N
Two min case, effective N-flation (Λ4 = 1/4π 2 , ϕ∗ = ( 21 − 0.001)Mp and χ∗ = 16.0Mp )
40 40
Γφ=Γχ=√10-2 Γφ=√10-2 Γχ=√10-4

30 Γφ=Γχ=√10-4 30 Γφ=√10-4 Γχ=√10-2
Γφ=Γχ=0 Γφ=Γχ=0
20 20
fNL 10 fNL 10
0 0
−10 −10
−20 −20
64 65 66 67 68 69 64 65 66 67 68 69
N N
Godfrey Leung
√Reheating Makes Life Harder
My Work
Beyond Slow-Roll, Trispectrum, τNL

One min case, quadratic exponential potential (left: λ = 0.06; right: λ = 0.05)
Two min case, effective N-flation (Λ4 = 1/4π 2 , ϕ∗ = ( 21 − 0.001)Mp and χ∗ = 16.0Mp )

My Work
Other Observables, nζ and r
No generic trend for fNL and τNL , the changes are model-dependent
Similar results for nζ and r , with nζ being more robust and less sensitive to
reheating in general
quartic minima also give same qualitative results

For two minima N-flation case, observables seem only sensitive to the ratio
R ≡ Γχ /Γϕ
Trispectrum, gNL
1. sensitive to reheating as well in general
2. beyond detectable level in general even if fNL and τNL are large

My Work
Non-Separable Potential
2 2
Beyond separable form, e.g. W (χ, ϕ) = W0 (χ4 e −λϕ /Mp
+ κχ2 )
The evolution of fNL during reheating for two different κ
100 50
Γχ=√10-1 κ=1.0 Γχ=√10-1 κ=0.1
Γχ=√10-3 Γχ=√10-3
Γχ=√10-5 Γχ=√10-5
25
50
0
fNL 0 fNL
−25
−50
−50
−75
−100
60 61 62 63 64 65 66 60 61 62 63 64 65 66
N N
Same qualitative behaviour as separable potential

My Work
Scale Dependence, nfNL and nτNL
in general small < O(0.1), but could be made large for the quadratic exponential
2
model W = W0 χ2 e −λϕ /Mp in some cases
eg.

My Work
Hopeless to Constrain Multifield Models?
Hopeless?
Observables are sensitive to physics of reheating
Physics of reheating is highly model-dependent and difficult to constrain...

My Work
Hopeless to Constrain Multifield Models?
However, consistency relations exist in inflationary models, even in single field

models
Example, tensor tilt and tensor-to-scalar ratio
r = −8cs nT (1)
while different single models give different r and nT , this relation is always
satisfied
A violation automatically rules out all single-field models

My Work
Multifield Consistency Relations
There are also consistency relations in multifield models. For example
Non-vacuum dominated sum-separable potential, slow-roll limit

27
25 gNL ≈ τNL [Elliston et.al. 12]
Two-field local type models
If ζ(k) = ζkG ,ϕ + ζkG ,χ + fϕ (ζ G ,ϕ ? ζ G ,ϕ )k + gϕ (ζ G ,ϕ ? ζ G ,ϕ ? ζ G ,ϕ )k , then

3nfNL = 2nτNL [Byrnes et.al. 12]
Suyama-Yamaguchi inequality, τNL ≥ ( 65 fNL )2 . A strong violation of equality

seems rather implausible in the slow-roll limit, requires extreme fine-tuned
conditions [Tegmark et.al.12]

My Work
Consistency Relations Beyond Slow-roll
Consistency relations are much more robust to physics of reheating. For example
1. nτNL = (3/2)nfNL , two–field local type [Byrnes et.al.10]
2. (27/25)gNL ≈ τNL , non–vacuum dominated sum–separable potential [Elliston et.al.
12]
the quadratic exponential model (left) and the effective N-flation model (right)

My Work
Consistency Relations Beyond Slow-roll (con’t)
Suyama-Yamaguchi inequality, τNL ≥ ( 65 fNL )2

For large range of decay rates, Γχ and Γϕ , the equality is still mildly violated
For example, the quadratic exponential model
Consistency relations are more reliable tools to test/constrain different classes of

models

My Work
Future Directions
Recently for models with no couplings between the scalar fields, sudden decay
approximation is a good approximation [Meyers et.al. 13]
Range of allowed values is not arbitrary for wide range of reheating timescale
Apply the sudden decay approximation beyond bispectrum?
Understand in what situations consistency relations preserved after (p)reheating
and why?
Extend the sudden decay approximation to models with non-sum-separable
potentials, where there are couplings between the fields
Consider more realistic (p)reheating scenarios, e.g. Γ → Γ(φI )
Study other classes of multifield models, such as non-minimal couplings and
non-canonical KE terms

My Work
Conclusion
Take home message

Evolution during reheating is important if isocurvature modes are still present
Observables evaluated at end of inflation in general are different from that after
reheating
gNL in general would be too small to be observed in multifield models
it may still be possible to test different classes of models by studying consistency
relations between observables
More work to be done
Danke Sehr!

Can We Constrain or Even Rule Out Multifield Inflation Models?

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Can We Constrain or Even Rule Out Multifield Inflation Models?

Uploaded by

Copyright:

Available Formats

Some Very Basic Background

Can We Constrain or Even Rule Out Multifield Inflation Models?

Supervisor - Prof Ed Copeland

collaboration work with - E.R.M.Tarrant2 and C.T.Byrnes2

22nd Jan, Heidelberg

Godfrey Leung Reheating Makes Life Harder

Some Very Basic Background

Godfrey Leung Reheating Makes Life Harder

Problems of Standard Big Bang Cosmology

Godfrey Leung Reheating Makes Life Harder

Inflation, a paradigm for Early Universe

It is believed that our Universe has undergone a dramatic exponential expansion

Yet it remains a phenomenological model...

Godfrey Leung Reheating Makes Life Harder

Primordial Inflationary Observables

Credits: ESA, Planck Collaboration

Beyond Power Spectrum: Non-Gaussianity

NonG, a measure about departure from a Gaussian statistics in primordial

Godfrey Leung Reheating Makes Life Harder

Amplitude, measured by non-linearity parameters

Credits: KICP, Uni of Chicago

Godfrey Leung Reheating Makes Life Harder

In general, fNL , τNL and gNL are ki (shape) dependent

Godfrey Leung Reheating Makes Life Harder

Current Constraints, Planck2013

Credits: Planck Collaboration 2013

Conclusion: Consistent with simple single field inflation

Why Consider Multifield Inflation?

Why not? ”Absence of Evidence is not the Evidence of Absence”!

Godfrey Leung Reheating Makes Life Harder

Why Consider Multifield Inflation?

Godfrey Leung Reheating Makes Life Harder

Difference Between Single and Multifield Models

One significant difference between single and multifield is the presence of

Thus ζ and related observables can continue to evolve

Godfrey Leung Reheating Makes Life Harder

Yet (p)reheating is not taken into account in the analysis...

Godfrey Leung Reheating Makes Life Harder

Why Consider (P)reheating?

(P)reheating is an important part of inflationary model building

Godfrey Leung Reheating Makes Life Harder

Why Consider (P)reheating? (con’t)

as discussed, isocurvature mode can source curvature perturbation

Godfrey Leung Reheating Makes Life Harder

We study the class of multifield models

In particular, we study the two-field models with canonical KE terms

φ¨I + (3H + ΓI )φ˙I + W,φI = 0

Γ is switched on after first passage through the minima after slow-roll

Godfrey Leung Reheating Makes Life Harder

Godfrey Leung Reheating Makes Life Harder

Godfrey Leung Reheating Makes Life Harder

Written in terms of the δN derivatives, the key primordial observables fNL , nζ , r

Godfrey Leung Reheating Makes Life Harder

Explicit analytic formulae of δN derivatives exist for separable potentials, i.e.

Godfrey Leung Reheating Makes Life Harder

quadratic exponential potential (left), N-flation (right)

Godfrey Leung Reheating Makes Life Harder

Beyond Slow-Roll, Bispectrum, fNL

Γφ=Γχ=√10-2 Γφ=√10-2 Γχ=√10-4

Beyond Slow-Roll, Trispectrum, τNL

Godfrey Leung Reheating Makes Life Harder

Other Observables, nζ and r

quartic minima also give same qualitative results