string axion inflation via random matrix theory · cosmological frontiers in fundamental physics...
TRANSCRIPT
Liam McAllisterCornell
APC/PI/Solvay Workshop
Cosmological Frontiers in Fundamental Physics
APC, June 16, 2011
Nishant Agarwal, Rachel Bean, L.M., Gang Xu, 1103.2775
Sohang Gandhi, L.M., Stefan Sjörs, 1106.0002
Based on:
I. Background and motivation: random potentials from
Planck-suppressed interactions
II. Computing the inflaton potential by perturbing around a
local Calabi-Yau geometry
III. Statistical predictions from an ensemble of models
• Inflation provides a beautiful causal mechanism to
generate the observed CMB anisotropies and distribution
of large-scale structure.
• Inflation is sensitive to Planck-scale physics: Planck-
suppressed operators generically make critical
contributions to the dynamics.
• This provides a remarkable opportunity to probe aspects of
the ultraviolet completion of gravity through observation.
• To make meaningful use of this connection, we can:
– Compute these Planck-suppressed contributions to the inflaton
action in string theory
– Search for characteristic properties: “what kind of inflation is natural
in string theory?” (contrast “can I realize inflation of type X in string
theory”).
0
For small inflaton excursions, , one must control
corrections with .
For large inflaton excursions, , one must control
an infinite series of corrections, with arbitrarily large Δ.
In general, the contributions of interest arise from
integrating out massive fields that are part of the
ultraviolet completion of gravity.
In string theory, the contributing massive fields notably
include stabilized moduli.
So we should consider the spectrum and couplings of
moduli in string theory.
• String compactifications typically include many moduli. If
massless, these are very problematic.
• Significant progress in the past decade: in flux
compactifications, many moduli obtain masses.
want: light inflaton, Planck-mass moduli
MP
Mφ
Mmoduli
H
• String compactifications typically include many moduli. If
massless, these are very problematic.
• Significant progress in the past decade: in flux
compactifications, many moduli obtain masses.
want: light inflaton, Planck-mass moduli
MP
Mφ
Mmoduli
unnatural
H
• String compactifications typically include many moduli. If
massless, these are very problematic.
• Significant progress in the past decade: in flux
compactifications, many moduli obtain masses.
hard to achieve
want: light inflaton, Planck-mass moduli
MP
Mφ
Mmoduli
unnatural
H
• String compactifications typically include many moduli. If
massless, these are very problematic.
• Significant progress in the past decade: in flux
compactifications, many moduli obtain masses.
• But, these masses are finite!
hard to achieve
want: light inflaton, Planck-mass moduli get: most masses near H
MP
Mφ
Mmoduli
unnatural
MP
H
Mφ
Mmoduli
H
• Thus, string theory strongly motivates scenarios involving
multiple light fields whose potentials are controlled by
Planck-suppressed contributions.
• With some effort we can compute the form of the inflaton
action, but computing the Wilson coefficients is hopeless in
general. At best we can make statistical statements.
• The resulting models generically involve a number of light
fields in a complicated potential.
• Key question: when inflation arises in this context, what are
its characteristic properties? What can we learn without
knowing the details of the Wilson coefficients?
• Suppose we construct an ensemble of NxN matrices by
drawing the entries of each matrix from a given
distribution W .
• For large N, we can predict the spectrum of eigenvalues
to excellent accuracy, and the result is independent of W.
• Suppose we construct an ensemble of NxN matrices by
drawing the entries of each matrix from a given
distribution W .
• For large N, we can predict the spectrum of eigenvalues
to excellent accuracy, and the result is independent of W.
2 2
2
1( ) 4
2
e.g., suppose A is an NxN matrix whose entries Aij
are drawn from a Gaussian distribution with mean
zero and variance 2/N.
Let M = A + AT .
Then the spectrum of M is
• the distribution W need not be Gaussian
• the mean of W need not be small
• the entries can have some correlations
General result:
If M is a matrix whose constituent entries are drawn from any
distribution with appropriately bounded moments, then in the
large N limit the spectrum of M approaches that of a matrix
whose constituent entries have a Gaussian distribution with
mean zero.
This sort of universality is a strong motivation for seeking
simple inflationary dynamics in a suitable large N limit.
Z.D.Bai
mass
pro
babili
ty
300N
In N-flation, one considers hundreds of axions whose collective
excitation drives inflation.
Each individual axion mass depends on many details of the stabilized
compactification, but the spectrum of masses is simple!
Easther & L.M., 2005
Komatsu et al., 2010
Dimopoulos, Kachru, McGreevy, Wacker, 2005
Arguably simplest model: Gaussian random landscape.
– Have used weak-coupling (integer) dimensions
– Wilson coefficients can be drawn from some
distribution, e.g. a Gaussian.
– The special case J=2 was covered by RMT.
• At strong coupling, operator dimensions can change
substantially. Should we restrict to integers? Does it
matter?
• Also, we do not have a solid prior for the distribution – does
it matter whether it is Gaussian?
• Are all possible terms included with equal weight?
Our approach:
• Work in a string theory construction where we can explicitly
compute the form of the potential.
• „Experimentally‟ check for dependence on the distribution.
• Although N will not be very large, we will find universal
behavior.
String theory motivates considering many light fields
governed by a complicated potential induced by Planck-
suppressed couplings.
We will study a specific example of this sort and search for
simplifications when the number of terms in the potential
is large.
First step: identify a model where we can obtain the
structure of the potential (but not the Wilson coefficients).
Sohang Gandhi, L.M., Stefan Sjörs, 1106.0002
• In principle we‟d like to choose a compact model with
explicitly stabilized moduli, and directly compute the
effective theory up to (say) Δ=8 contributions.
• This would require a tour de force combining , gs
corrections with unprecedented understanding of the
moduli potential.
• Idea: study the effective theory of a local region of a
stabilized compactification, and use 10d locality to
simplify the calculation.
• Key simplification: in type IIB string theory, the best-
understood methods of moduli stabilization yield 10d
solutions that can be formulated in systematic expansion
around conformally Calabi-Yau backgrounds.
• Now we want to incorporate corrections that encode the
effects of moduli stabilization.
• So we need to find new solutions in perturbation theory
around the conformally Calabi-Yau background.
• The small parameter is the weak breaking of the ISD
condition, equivalent to the smallness of the moduli
potential in string units.
• We find a remarkable simplification!
Consider k scalar functions of a variable r,
and a background
Gandhi, L.M., Sjörs, 1106.0002
The type IIB equations of motion expanded
around any conformally Calabi-Yau
background are triangular, to any order!
(SUSY is not required.)
So in principle we can solve iteratively.
Borokhov and Gubser, 2002
Bena, Graña, Halmagyi 2009 et seq.
Dymarsky 2011
Gandhi, L.M., Sjörs, 1106.0002
Triangularity applied to ODEs for linearized
perturbations of Klebanov-Strassler:
The type IIB equations of motion expanded
around any conformally Calabi-Yau
background are triangular, to any order!
(SUSY is not required.)
So in principle we can solve iteratively.
Background: a noncompact warped CY cone with ISD fluxes, e.g. KS.
Goal: finite, warped, non-SUSY region attached to compact bulk.
We obtain separable solutions: we give all Green‟s functions in terms of
angular harmonics on the Sasaki-Einstein base. For T1,1 the harmonics
are known in closed form.
The same Green‟s functions apply at any order, so we can obtain a
solution to any desired order by a purely algebraic procedure.
cf. Ceresole, Dall‟Agata, D‟Auria 1999
• Now we use these tools to study the
particular local geometry that is relevant
for D3-brane inflation.
CY orientifold, with
fluxes and nonperturbative W
(KKLT 2003)
warped throat
(e.g. Klebanov-Strassler)
anti-D3-brane
D3-brane
Dvali&Tye 1998
Dvali,Shafi,Solganik 2001
Burgess,Majumdar,Nolte,Quevedo,Rajesh,Zhang 2001
Kachru, Kallosh, Linde, Maldacena, L.M., Trivedi, 2003
CY orientifold, with
fluxes and nonperturbative W
(KKLT 2003)
warped throat
(e.g. Klebanov-Strassler)
anti-D3-brane
D3-brane
Note: the inflaton potential is NOT just the Coulomb
term! Planck-suppressed effects from moduli
stabilization are dominant!
• We found the supergravity solution corresponding to the
most general bulk contributions and read off the structure of
the potential.
Baumann, Dymarsky, Kachru, Klebanov, L.M., 2008, 2009, 2010
Gandhi, L.M., Sjörs, 2011
We will keep the 333 terms with Δ<7.8.
We draw the coefficients ci from a Gaussian distribution,
generating an ensemble of potentials. (We checked that the
choice of distribution is not significant.)
We draw a potential from the ensemble, choose a random
initial condition (with appropriately bounded kinetic energy),
and find the full six-field evolution of the homogeneous
background. Repeat 7 x 107 times.
We stare at the results and try to identify “robust observables”,
i.e. quantities that depend very weakly on the details of the
distribution.
Typical successful trajectories. The angles evolve initially
and then settle into an angular minimum.
cf. Spinflation, Easson et al., 2007
Total number of e-folds, Ne
P(N
e)
x 1
05
model parameters measure
Freivogel, Kleban, Rodriguez Martinez and Susskind, 2005
model parameters measure
• When multiple fields are active during inflation,
computing the perturbations is subtle.
– Bending trajectories induce conversion of
isocurvature to curvature outside the horizon
– Prospect of interesting non-Gaussianity
• As a first step, we can identify trajectories that
bend so little that a single-field treatment is
applicable. Use
• Worthwhile to do the full multifield analysis!
Groot Nibbelink and van Tent, 2001
Gordon, Wands, Bassett and Maartens, 2001
Total number of e-folds
• Overshooting the inflection point is rather rare; one can start far
above it without difficulty.
• We find attractor behavior in the angular directions.
• Characteristic of successful examples: a funnel towards an inflection
point, which need not be parallel to the radial direction.
• We find 60+ e-folds of inflation approximately once in 103 trials.
• Restricting (for now) to cases that are effectively single-field when
the CMB is produced, we find models consistent with all data
approximately once in 105 trials, provided we set the overall scale in
the right range.
• No tensors: r < 10-12.
• DBI inflation did not occur by chance.
• The probability of N e-folds is proportional to N-3.
• String theory strongly motivates considering inflation
models with multiple light fields whose potential is
controlled by Planck-suppressed contributions.
• We studied a particular case with N=6 fields (D3-brane
inflation).
• To compute the structure of the inflaton potential, we
developed a toolkit for finding solutions in perturbation
theory around conformally Calabi-Yau backgrounds.
• Drawing the Wilson coefficients from various
distributions, we constructed ensembles of potentials.
• We then studied the typical behavior in these
ensembles.
• Idea: when there are many fields and many terms in the
potential, central limit behavior will take over, yielding a
sort of universality.
• Predictivity comes at the cost of statistical insensitivity to
ultraviolet physics.
• We found scaling behavior: the probability of Ne e-folds of
inflation is a power law, Ne-3 .
• We did not yet study multifield effects in the
perturbations, but there is plausibly a rich story involving
bending trajectories and transients.
• To do: extension to more general, higher-dimensional
systems.