TOPICS IN COSMOLOGY: ISLAND UNIVERSES, COSMOLOGICAL PERTURBATIONS AND DARK ENERGY
by SOURISH DUTTA
Submitted in partial fulfillment of the requirements
for the degree Doctor of Philosophy
Department of Physics
CASE WESTERN RESERVE UNIVERSITY
August 2007 CASE WESTERN RESERVE UNIVERSITY
SCHOOL OF GRADUATE STUDIES
We hereby approve the dissertation of
______
candidate for the Ph.D. degree *.
(signed)______(chair of the committee)
______
______
______
______
______
(date) ______
*We also certify that written approval has been obtained for any proprietary material contained therein. To the people who have believed in me. Contents
Dedication iv
List of Tables viii
List of Figures ix
Abstract xiv
1 The Standard Cosmology 1 1.1 Observational Motivations for the Hot Big Bang Model ...... 1 1.1.1 Homogeneity and Isotropy ...... 1 1.1.2 Cosmic Expansion ...... 2 1.1.3 Cosmic Microwave Background ...... 3 1.2 The Robertson-Walker Metric and Comoving Co-ordinates ...... 6 1.3 Distance Measures in an FRW Universe ...... 11 1.3.1 Proper Distance ...... 12 1.3.2 Luminosity Distance ...... 14 1.3.3 Angular Diameter Distance ...... 16 1.4 The Friedmann Equation ...... 18 1.5 Model Universes ...... 21 1.5.1 An Empty Universe ...... 22 1.5.2 Generalized Flat One-Component Models ...... 22 1.5.3 A Cosmological Constant Dominated Universe ...... 24 1.5.4 de Sitter space ...... 26 1.5.5 Flat Matter Dominated Universe ...... 27 1.5.6 Curved Matter Dominated Universe ...... 28 1.5.7 Flat Radiation Dominated Universe ...... 30 1.5.8 Matter Radiation Equality ...... 32 1.6 Gravitational Lensing ...... 34 1.7 The Composition of the Universe ...... 39
v 1.7.1 Measuring the Total Matter Content ...... 39 1.7.2 Ordinary and Dark Matter ...... 42 1.7.3 Supernovae and Cosmic Acceleration ...... 45 1.7.4 Cosmic Microwave Background Anisotropies ...... 49 1.8 Shortcomings of the Standard Cosmology ...... 58 1.8.1 The Flatness Problem ...... 58 1.8.2 The Entropy Problem ...... 59 1.8.3 The Horizon Problem ...... 59 1.8.4 The Monopole Problem ...... 60 1.9 Inflation ...... 62 1.9.1 Inflation and the Problems of the Standard Cosmology . . . . 62 1.9.2 The Dynamics of Inflation ...... 64 1.9.3 The Slow Roll Parameters ...... 66 1.9.4 Models of Inflation ...... 67 1.9.5 Issues with Inflation ...... 69
2 Island Cosmology 76 2.1 Introduction ...... 76 2.2 The Model ...... 77 2.3 NEC violations in de Sitter space ...... 80 2.4 Extent and duration of NEC violation ...... 82 2.5 Likelihood – the role of the observer ...... 87 2.6 The NEC violating field ...... 89 2.7 Assumptions ...... 91 2.8 Conclusions ...... 94
3 Perturbation Spectra 98 3.1 Introduction ...... 98 3.2 The Theory of Cosmological Perturbations ...... 98 3.2.1 The metric perturbations ...... 99 3.2.2 Gauge Issues in Cosmology ...... 101 3.2.3 The Comoving Curvature Perturbation ...... 103 3.2.4 The Power Spectrum ...... 104 3.2.5 The Equations of Motion ...... 106 3.2.6 Quantum Theory of Cosmological Perturbations ...... 108 3.3 Difficulties in computing perturbations from Island Cosmology . . . . 112 3.3.1 Classical vs Quantum Fields ...... 112 3.3.2 Backreaction on spacetime ...... 113 3.4 Perturbations in a Spectator Field ...... 117 3.5 A Classical Treatment of Island Cosmology ...... 120
vi 3.5.1 Introduction ...... 120 3.5.2 The de Sitter phase ...... 121 3.5.3 The Phantom Phase ...... 122 3.5.4 The Radiation Dominated FRW Phase ...... 125 3.5.5 Calculational Strategy ...... 125 3.6 Spectrum from a Classical Treatment ...... 127 3.6.1 vk in the de Sitter and FRW phases ...... 127 3.6.2 vk in the Phantom Phase ...... 128 3.6.3 Calculation of Unknown Constants ...... 135 3.6.4 Determination of the Scalar Power Spectrum ...... 138 3.6.5 Determination of the Tensor Power Spectrum ...... 138 3.7 Conclusion ...... 140
4 Dark Energy Voids 144 4.1 Introduction ...... 144 4.2 What is causing the acceleration? ...... 145 4.3 The model ...... 147 4.3.1 Linearization ...... 150 4.3.2 Potential ...... 153 4.3.3 Initial conditions ...... 153 4.4 Results ...... 155 4.4.1 Density contrasts ...... 155 4.4.2 Equation of state ...... 160 4.5 Discussion ...... 163 4.5.1 Void formation ...... 163 4.5.2 Generality ...... 165 4.6 Conclusions ...... 166
Bibliography 172
vii List of Tables
3.1 Scalar-Vector-Tensor decomposition of metric perturbations . . . . . 100
viii List of Figures
1.1 The spectrum of the CMB, as seen by COBE. http://lambda.gsfc. nasa.gov/product/cobe/firas_image.cfm ...... 4 1.2 Gravitational lensing geometry. The lens distorts the ”true” angles β (that would have been seen without any lensing effects) into the angles θ. Source [15] ...... 34 1.3 A flow chart showing the classifications of Supernovae http://rsd-www. nrl.navy.mil/7212/montes/snetax.html...... 46 1.4 top panel: A Hubble diagram made from data from both the Supernova Cosmology Project and the High-z Supernova Search Team taken from [13].bottom panel The residual of the distances relative to a ΩM = 0.3, ΩΛ = 0.7 Universe...... 50
1.5 Best fit regions in the (ΩM ,ΩΛ) plane for data from both the Super- nova Cosmology Project and the High-z Supernova Search Team. The agreement of the two experiments is remarkable. Source [13] . . . . . 51 1.6 The Dipole Anisotropy in the CMB as seen by COBE http://map. gsfc.nasa.gov/m_uni/uni_101Flucts.html...... 52 1.7 The CMB spectrum once the dipole is subtracted. http://map.gsfc. nasa.gov/m_uni/uni_101Flucts.html...... 53 1.8 The WMAP three-year power spectrum (in black) together with data from other recent experiments measuring the CMB angular power spec- trum. Taken from [44] ...... 56 1.9 A plot of length scale vs (logarithmic) scale factor in an FRW cosmol- ogy. The blue line shows the evolution of the Hubble scale. The red lines show the evolution of physical scales. Since the Hubble length evolves faster than the physical scales, sub-Horizon modes have never been in causal contact prior to Horizon entry...... 61 1.10 Physical scales entering the horizon at the time of last scattering have been in causal contact before...... 63
ix 2.1 Sketch of the behavior of the Hubble length scale with conformal time, η, in the Island model, and the evolution of fluctuation modes. At early times, inflation is driven by the presently observed dark energy, assumed to be a cosmological constant. As the cosmological constant is very small, the Hubble length scale is very large – of order the present horizon size. Exponential inflation in some horizon volume ends not due to the decay of the vacuum energy as in inflationary scenarios but due to a quantum fluctuation in the time interval (ηi, ηf ) that violates the null energy condition (NEC). The NEC violating quantum fluctu- ation causes the Hubble length scale to decrease. After the fluctuation is over, the universe enters radiation dominated FRW expansion, and the Hubble length scale grows with time. The physical wavelength of a −1 quantum fluctuation mode starts out less than HΛ at some early time ηi. The mode exits the cosmological horizon during the NEC violating fluctuation (ηexit) and then re-enters the horizon at some later epoch (ηentry) during the FRW epoch (The modes are drawn as straight lines for illustrative purposes only, they actually grow in proportion to the scale factor)...... 78
2.2 We show a classical de Sitter spacetime for conformal time η < ηP , that transitions to a faster expanding classical de Sitter spacetime for η > ηQ. The inverse Hubble size is shown by the white region. A bundle of ingoing null rays originating at point a is convergent initially but becomes divergent in the superhorizon region at point b. This can only occur if the NEC is violated in the region η ∈ (ηP , ηQ). In the quantum domain, a classical picture of spacetime may not be valid and this is made explicit by the question marks...... 83
2.3 A spacetime diagram similar to that in Fig. 2.2 but one in which the NEC violation occurs over a sub-horizon region (shaded region in the diagram). Now the null ray bundle from a to b goes from being converg- ing (within the horizon) to diverging (outside the horizon). However, it does not encounter any NEC violation along its path, and this is not possible as can be seen from the Raychaudhuri equation. Since the ingoing null rays are convergent as far out as the point P , the size of the quantum domain has to extend out to at least the inverse Hubble size of the initial de Sitter space. Therefore the NEC violating patch has to extend beyond the initial horizon...... 85
x 4.1 The evolution of the DDE overdensity δφ at the center of the matter perturbation, r = 0, with redshift (1+z).The scale of the perturbation is σHi = 0.01, and the mass is m/H0 = 1. Initially homogenous, the DDE develops an underdensity at late times in response to the matter perturbation...... 156 4.2 Same as Figure 4.1, with the y-axis on a logarithmic scale. The DDE tends to cluster initially, but eventually forms a void. The kink in the plot signifies the change-over from positive to negative perturbation. . 156
4.3 Logarithmic profiles of the matter density contrast log10 |δm| and the dark energy density contrast log10 |δφ|, at three different redshifts. Solid lines denote the DDE profiles, and dotted lines denote the matter profiles.158 4.4 DDE density contrast δφ at the center of the matter perturbation r = 0 as a function of the redshift (1 + z) for fixed mass m/H0 = 1 and different initial matter perturbations’ widths. The larger the initial matter perturbation, the stronger is the void. The curves of σHi = 0.01 (dashed) and 0.1 (solid) almost overlap. The figure zooms on late times, z < 3...... 159 4.5 Same as Figure 4.4 with the y-axis on a logarithmic scale. The shorter scales start evolving at later times than the longer scales, but their evolution is faster. The curves of σHi = 0.01 (dashed) and 0.1 (solid) almost overlap...... 159 4.6 The DDE density contrast δφ at the center of the matter perturbation, r = 0, against redshift (1+z) for σHi = .01 and three different masses. The figure zooms on late times, z < 7...... 161 4.7 Same as Figure 4.6 with the y-axis on a logarithmic scale. The pertur- bation’s is extremely sensitive to the mass scale...... 161 4.8 Plot of w1 vs r for three different redshifts...... 162 4.9 Plot of % deviation in w vs r at three different redshifts...... 162 4.10 Percentage variation of the local Hubble parameter at three different redshifts ...... 163 4.11 Profile of δφ¨ at three different redshifts ...... 164 4.12 The equivalent of figure 4.3, with the double exponential potential, equation (4.33)...... 166 4.13 The equivalent of figure 4.8, with the double exponential potential, equation (4.33)...... 167
xi Acknowledgements
I wish to acknowledge all the wonderful people in my life, in particular: My mother
Sharmila, who among many, many other things was my first and best teacher of physics - who always took the time to answer all my science-related questions when
I was little, and fostered in me an inalienable spirit of inquiry; my father Shyamal, who again among many other things, was my first and best teacher of writing, and who taught me, through the exemplary life that he led, the virtues of perseverance, dedication, sincerity and integrity; my wonderful little sister Arundhati, a multi- faceted, multi-talented and superluminary individual, an over-achieving powerhouse perennially brimming with potential, who is a constant source of joy and pride for me; my advisor Tanmay, who always saw the best in me, who was always available to help me with whatever I needed, who always did everything in his power to bring me the best opportunities, and who is that rare combination of exceptionally brilliant physicist and sterling human being; Irit, who instilled in me the confidence to trust my ability to do research, and who helped me discover my taste for gravitational physics and numerical analysis; Harsh, whose depth of understanding and gift for exposition will always remain an inspiration for me; Punam, whose warmth and kindness have made my years in Cleveland exceedingly pleasant and have left me
xii with some truly delightful memories; and finally, my love Maia who is one of the greatest blessings in my life, who epitomizes love and compassion and sincerity and gentleness, who every day inspires me to be a better human being, and without whose constant encouragement (and persistent prodding!) this thesis would never have been completed.
xiii Topics in Cosmology: Island Universes, Cosmological Perturbations and Dark
Energy
Abstract
by
SOURISH DUTTA
This thesis is a report on the research I have done over the last three years. I begin by reviewing the Standard Big Bang cosmology and the theory of inflation. I then describe Island Cosmology, an alternative theory of cosmic origin, which is based on a different hypothesis, and can create the observed universe without requiring an inflationary phase. After fully describing this model, I move on to the subject of com- puting perturbation spectra. The theory of cosmological perturbations, as it applies to calculating the perturbation spectrum from inflation, is reviewed next. Computing the perturbation spectrum in Island Cosmology involves several challenges, which I then discuss and also review progress made so far in computing the spectrum. In the
final part of this thesis I describe my research in gravitational collapse in the presence of a dynamical dark energy component (DDE). I review the very interesting result that in linear regime, the collapsing matter induces the formation of DDE voids, or regions of underdensity. I describe the physics behind the formation of these voids, as well as possible observational consequences.
xiv Chapter 1
The Standard Cosmology
1.1 Observational Motivations for the Hot Big Bang
Model
The Standard Cosmology, or the “Hot Big Bang Model” is based on three pillars of observational evidence, namely, homogeneity and isotropy, Hubble expansion and the cosmic microwave background. I briefly explain each of these below:
1.1.1 Homogeneity and Isotropy
Also known as the Cosmological Principle, this is the assumption that there is neither a special location in the Universe (homogeneity) nor a special direction (isotropy).
Studies of large scale structure show that the Universe is considerably heterogeneous up to the scale of galactic superclusters, filaments and great voids. However, on scales larger than 100 Mpc the Universe appears to be homogeneous and isotropic (see, for
1 example, [1] for a comprehensive discussion of the evidence in support of homogeneity and isotropy). 1
1.1.2 Cosmic Expansion
It is found that the light from most galaxies is redshifted, and the degree of redshift is proportional to the distance of the galaxy. First discovered by Edwin Hubble in
1929, this relationship is known as the famous ”Hubble Law”:
H z = 0 r (1.1) c where z and r stand for the redshift and distance of a galaxy respectively. H0 is called the Hubble constant, which is actually a slowly varying function of time, and the subscript 0 denotes its value at the present time. In what follows, we work in units such that the speed of light c = 1.
The Hubble law clearly indicates that the Universe is expanding, or in other words, the observed redshifts of galaxies are actually Doppler shifts. The distance between any two objects (which are not interacting gravitationally or otherwise) grows in proportion to a function of time known as the scale factor a(t), which sets the scale of the geometry of space. As I explain in 1.2, I will follow the convention of taking the scale factor to be dimensionless.
1Interestingly, recent analysis of the Wilkinson Microwave Anisotropy Probe (WMAP) [2] seems to indicate that the quadrupole and the octopole of the Cosmic Microwave Background seem to have a strong correlation with the orientation of the ecliptic plane and its motion, a result that seems at odds with the Copernican Principle. However, I will not discuss this highly interesting result any further in this report.
2 Using the classical non-relativistic form of the Doppler shift z = v (where v is the radial velocity of the source with respect to us, the observers) one can re-express
Hubble’s law as follows:
v = H0r (1.2)
Obviously, owing to the non-relativistic approximation, this form of the Hubble law is valid only for small scales where the expansion velocity is small.
Also, it is easy to show [1] that the redshift and the scale factor are related as
1 (1 + z) ∝ (1.3) a
1.1.3 Cosmic Microwave Background
The Universe is currently immersed in a bath of radiation, at a temperature of T0 = 2.725 ± 0.001K which peaks in the microwave range at a frequency of 160.2 GHz, corresponding to a wavelength of 1.9mm [3]. This background of microwave radiation, called the Cosmic Microwave Background, has a spectrum that is blackbody within
3.4 × 10−8 erg cm−2 s−1 sr−1 over the frequency range from 2 to 20 cm−1 [4]. Figure
1.1 shows the spectrum of the CMB as measured by the FIRAS instrument on the
COBE. The data matches the curve almost exactly, with error bars being smaller than the width of the curve itself.
The CMB is one of the strongest confirmations of the Hot Big Bang model. In fact the Hot Big Bang model provides a natural explanation for the CMB, as a relic of an immensely hot and dense phase in the history of the Universe. Very briefly, the
3 Figure 1.1: The spectrum of the CMB, as seen by COBE. http://lambda.gsfc. nasa.gov/product/cobe/firas_image.cfm
4 argument is as follows:
The mere fact that the CMB is blackbody radiation strongly indicates that it must be a relic of a phase where the constituents of the Universe were in thermal equilibrium. Since the Universe is not in equilibrium now, the assumption of a prior equilibrium state is the only way one can explain the blackbody nature of the CMB.
During this early equilibrium phase, very frequent Thompson scattering between the photons and the baryons and electrons caused the mean free paths of the photons to be small - making the Universe opaque. As the Universe expanded, the plasma cooled adiabatically. Approximately 380,000 years after the Big Bang (at z = 1088), the temperature of the plasma dropped to 3000K, and electrons and protons could combine to form neutral Hydrogen atoms, or in other words, the baryonic component of the Universe went from ionized to neutral. This event is called recombination.
Shortly after this, the rate of expansion of the Universe exceeded the rate of interaction between the photons and the electrons (this is called decoupling). After decoupling, the photons started free-streaming through the Universe - redshifting and cooling with the expansion of the Universe - and today make up the CMB.
The drop in temperature by a factor of 1100 of the radiation background from
3000K when the photons started free-streaming (at z = 1088) to 2.73K today, follows directly from the adiabatic expansion of the Universe, and it can be shown that
T (t) ∝ a(t)−1 (1.4)
(In § 1.5.7 I will formally derive this relationship and show that it is not exact, but is a good approximation). In other words, the factor of 1100 drop in the temperature
5 of the radiation background is accompanied by a 1100-fold increase in the cosmic
scale factor.
Perhaps the most striking feature of the cosmic microwave background is its very
high degree of isotropy (the anisotropy is of the order of 10−5). Incidentally, this
isotropy is also a measure of the homogeneity of the Universe at the epoch of last
scattering. I will discuss the CMB anisotropy in far greater detail in § 1.7.4.
Taken together, the above observations lead to the conclusion that the Universe started out as an immensely hot and dense plasma which has been expanding and cooling ever since. This is the crux of the Hot Big Bang Model. In the rest of this chapter I will provide a brief review of this model, highlighting aspects of it which are relevant to my research. This chapter is based on the following excellent books and reviews which have helped me immensely in developing my understanding of cosmology [1, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14]
1.2 The Robertson-Walker Metric and Comoving
Co-ordinates
Guided by the homogeneity and isotropy of our Universe we seek a solution that is spatially homogeneous and isotropic, and evolves in time. The Friedman-Robertson-
Walker (FRW) metric fits this description perfectly, as it describes a Universe meeting the requirements of homogeneity and isotropy, and where lengths can arbitrarily expand (or contract) with time.
The general form of the metric, in terms of a spherical polar coordinate system,
6 is the following:
ds2 = dt2 − a(t)2 dr02 + S(r0)2dΩ2 (1.5)
Here r0 is the radial co-ordinate, θ and φ are the polar and azimuthal angles, and t denotes time. a(t) is the scale factor introduced in 1.1 but in this particular form of the metric it has dimensions of length. Also dΩ2 = dθ2 + r02dφ2 and the function
S(r0) is determined by the curvature of space:
R sin(r0/R) positive curvature 0 S(r ) = r0 flat space (1.6) R sinh(r0/R) negative curvature
Here, R = a(t0) is the scale factor at the present epoch. Note that S(r) only determines the curvature of space locally and not globally, which is what one would expect from a local theory such as General Relativity. The global topology need not be the same as the local topology. For example, a locally
flat topology might be part of either an infinite plane, or a 3-torus.
Making a transformation to a different radial coordinate r = S(r0), and rescaling the scale factor by R to make it dimensionless, the FRW metric takes the following form:
dr2 ds2 = dt2 − a2(t) + r2dΩ2 (1.7) 1 − kr2 where k denotes the spatial curvature of the Universe.
For the rest of this chapter I will let r0 denote the radial coordinate used in (1.5)
7 and r denote the radial coordinate in (1.7).
The rate at which the scale factor changes determines the speed at which the
Universe expands (or contracts). This is measured by the Hubble parameter H(t) defined by a˙ H(t) = (1.8) a
Using the approximate distance redshift relationship (1.2), one could in principle measure the present value of the Hubble parameter, if the distances to galaxies were known accurately. However, the value of the Hubble parameter is still subject to some uncertainty characterized by the parameter h, and is usually expressed as follows:
h H = 100h kms−1Mpc−1 ' Mpc−1 (1.9) 0 3000
Observations suggest that h lies between 0.5 and 0.8, and the best present day
+.04 measurement is .73−.03 [3]. The inverse of the Hubble parameter sets a scale for the age of the Universe (the Hubble time) , as well as its spatial width (the Hub- ble distance). The present Hubble time and Hubble distance are 9.78h−1Gyr and
2998h−1Mpc respectively [9].
The second derivative of the scale factor measures the rate at which the Universe accelerates. This is usually characterized by the quantity:
aa¨ q = − (1.10) 0 a˙ 2 t=t0
Instead of the function S(r), curvature is now signified by the parameter k, which takes on values of (−1, 0, +1) for spaces of constant positive, zero or negative spatial
8 curvature. k also sets the curvature scale rcurv of the Universe which is defined as:
−1/2 rcurv = a(t)|k| (1.11)
This is the physical distance at which effects of curvature start becoming signifi- cant. From (1.7), it is clear that when the physical distance rcurv, spacetime looks flat.
It is sometimes convenient to use conformal time (η) , which is defined by dη = dt/a(t). Re-writing the FRW metric in terms of conformal time, it is easy to see that this metric is conformally flat, that is, it has the same form as the Minkowski line element, up to a conformal factor2:
dr2 ds2 = a2(η) dη2 − + r2dθ2 + r2 sin2 θdφ2 (1.12) 1 − kr2
There are two sets of transformations [6] which leave this metric invariant. These
are:
1. The rigid rotations
0i i j x = Rjx (1.13)
where xi are Cartesian coordinates and R is a rotation matrix, and
2. The quasi-translations which translate the origin to the point defined by the
vector a n h x.aio x0 = x + a 1 − kx21/2 − 1 − 1 − ka21/2 (1.14) a2
2The conformal flatness of this metric can also be rigorously tested by computing its Weyl cur- vature, which turns out to be zero.
9 The second transformation implies that any point whose coordinates are fixed with respect to the origin, can equally serve as the origin, and that the Universe will look the same from that point as it does from the origin. In other words, in an FRW spacetime, spatial homogeneity is guaranteed by isotropy at one location.
From the equation of motion for a particle in a gravitational field [6],(where τ is the proper time)
d2xλ dxµ dxν + Γλ = 0 (1.15) dτ 2 µν dτ dτ
µ , and from the fact that the Christoffel symbol Γtt vanishes for the FRW metric, one can deduce that the trajectories of fixed x are geodesics. Typical galaxies fol- low geodesics, i.e., they are in free fall, and hence in this coordinate system they will have fixed coordinates. Hence this set of coordinates, aptly named comoving coordinates have the added benefit that typical galaxies have fixed spatial loca- tions. Observers at rest in comoving co-ordinates are called comoving observers.
Only comoving observers find the Universe isotropic. The fact that we see a dipole anisotropy in the CMB is because the motion of the earth is not comoving.
Borrowing an analogy from Weinberg [6], comoving co-ordinates can be thought of as a set of coordinates axes painted on the surface of a balloon. Regardless of whether the balloon expands or contracts, any given point on the balloon will have the same location with respect to the origin of these coordinates, since the axes expand or contract as well.
Any velocities with respect to comoving observers are called peculiar velocities and it can be shown that unless supported by interactions, peculiar velocities “red-
10 shift”, or decrease in proportion to the scale factor a(t). Furthermore, it is also easy to prove that the wavelength of light in an FRW Universe redshifts as well. Both of these points are described in detail in [1].
One can also understand comoving coordinates in the language of differential geometry as follows [15]. Homogeneity and isotropy require that our spacetime can be sliced into spacelike hypersurfaces which are maximally symmetric. This can be mathematically represented as R1 × Σ, where R1 is a one-dimensional Euclidean
manifold representing the time direction and Σ is a spherically symmetric three-
manifold. Let us define a set of coordinates (u1, u2, u3) and a maximally symmetric
three-metric γij(u) on Σ, such that the line element on Σ can be written as
2 i j dσ = γij(u)du du
If the spacetime metric can also be written in the form
ds2 = dt2 − a2(t)dσ2 (1.16)
which is free of cross-terms dtdui and the coefficient of dt2 is independent of time,
then these coordinates (ui, t) are called comoving coordinates.
1.3 Distance Measures in an FRW Universe
The distance between two points in an expanding Universe is a subtle issue. An
idealized distance measure is the proper distance. However, as I explain below the
proper distance is not something one can actually measure.
11 Very short distances, such as distances to objects within the solar system, can be measured by bouncing radar signals of these objects. However, radar echoes become too weak to detect for distances on the order of 10 AU. Within our galaxy, distances can be determined using trigonometric parallax. The principle behind this that the apparent position of an astronomical object will shift if one changes the point of observation. Using this method the Hipparcos Satellite has determined distances to
2.5 million stars in our galaxy [16]. However, on cosmological scales (≈ 100) Mpc, trigonometric parallaxes become too small to measure.
On scales larger than our galaxy, two distance measures are available, called an- gular diameter distance or the luminosity distance. However, both of these can be shown to be approximately equal to the proper distance for small redshifts.
Below, I briefly review these measures below.
1.3.1 Proper Distance
0 Consider an object located at the point E (re(or re), θ, φ) and an observer viewing
0 this object from the point O (r0(or r0), θ, φ). The proper distance of E measured from
O at time t0 can be defined to be the length of a spatial geodesic between the two points when the scale factor is fixed at the value a(t0). From the FRW metrics (1.5) and (1.7), one can see that this would be given by
Z r0 dr 0 0 dprop(t0) = a(t0) √ = a(t0)(r − r ) (1.17) 2 0 e re 1 − kr
The proper distance can also be intuitively understood as follows: imagine a line of comoving observers between E and O each separated by co-ordinate distance
12 0 δr(or δr ). Suppose at the present time t0, each observer measures the physical dis- tance to the next observer. The distance δs(t) measured by any observer is given
by
a(t0) 0 δs(t0) = √ δr = a(t0)δr 1 − kr2
If one adds up these infinitesimal sub-distances, it is clear that one gets the above
expressions for proper distance (1.17).
Now suppose E sends a light signal to O. The signal is emitted at t = te and
received at t = t0. Since light follows null geodesics, from (1.7) we can rewrite the expression for the proper distance as follows
Z t0 dt0 dprop(t0) = a(t0) 0 (1.18) te a(t )
The time difference (t0 − te) is called the lookback time. The expression (1.18) would have been practically useful if the exact functional
form of the scale factor were known, as well as the lookback time. In practice the
best we can do is approximately determine the scale factor in the form of a Taylor
expansion about the present time t0
1 a(t) = a(t ) + a0(t )(t − t ) + a00(t )(t − t )2 + ... 0 0 0 2 0 0
Plugging this into (1.18) and retaining the two lowest order terms in the lookback time, one gets H d (t ) ' (t − t ) + 0 (t − t )2 (1.19) prop 0 0 e 2 0 e
Using (1.3), one can express the lookback time in terms of the redshift, and hence
13 write an expression for the proper distance of an object at a redshift z:
1 1 + q0 2 dprop(t0, z) = z − z (1.20) H0 2
1.3.2 Luminosity Distance
The luminosity distance of a celestial object is a distance derived by comparing its observed luminosity to its absolute luminosity, which is assumed to be known.
Suppose we have a ”standard candle” (an object of known luminosity) with a measured flux f. The luminosity distance is defined as
L 1/2 d = (1.21) L 4πf
In a static Euclidean Universe, the above quantity would give the proper distance to the object. However, the fact that the Universe need not be Euclidean introduces
2 complications:
1. Spatial curvature: According to (1.5), the curvature of space changes the proper
0 0 0 area of a sphere centered at the object and with a radius of ρ = |re − r0| from
2 02 2 0 2 4πa(t0) ρ to 4πa(t0) S(ρ ) . For positive curvature the area of a sphere is greater than the flat space case, and for negative curvature the area is lesser.
2. Cosmic expansion: The expansion of the Universe causes photons to redshift,
i.e., if a photon of wavelength λe emitted at time te is received now t0 and
14 measured to have a wavelength λ0, then we must have
a(t0) λ0 = λe = (1 + z)λe a(te)
which causes a corresponding shift in the energy of the detected photon:
E E = e (1.22) 0 1 + z
The expansion also causes photons emitted at time intervals δte to be received
at the detector at intervals δt0, where
a(t0) δt0 = δt0 = (1 + z)δte (1.23) a(te)
The net result of (1.22) and (1.23) is that the net flux of power received at the detector is
L L f = 2 0 2 2 = 2 2 2 (1.24) 4πa (t0)S(ρ ) (1 + z) 4πa (t0)|re − r0| (1 + z)
From equations (1.21) and (1.24) we get the following expressions for the luminosity distance:
dL = a(t0)|re − r0|(1 + z) = dprop(1 + z)
Using the expression connecting the proper distance to the redshift (1.20), one can
15 express the luminosity distance in terms of the redshift:
1 (1 − q0) 2 dL ≈ z + z (1.25) H0 2
Hence in the limit that z → 0
z dL ≈ dprop ≈ (1.26) H0
1.3.3 Angular Diameter Distance
The angular diameter distance is a distance measure determined by computing the observed angular diameter of an object to its known diameter.
Suppose we have a ”standard yardstick” of known proper length l, subtending an angle δθ at O. The angular diameter distance dA is defined by
l d = (1.27) A δθ
As in the previous case, this would give the proper distance to the object only if the
Universe is static and Euclidean.
To find how this expression changes in curved and expanding space, consider a set of comoving coordinates whose origin is at O, according to which the two ends of the yardstick are located at (r0, θ, φ) and (r0, θ + δθ, φ).
From the FRW metric (1.5), the length of the object can be written as
0 l = δs = a(te)S(r )δθ = a(te)rδθ
16 which gives, according to (1.27)
a(t )S(r0) a(t )r d d = 0 = 0 = prop (1.28) A 1 + z 1 + z 1 + z
Again, using (1.20), the angular diameter distance for flat space can be re-expressed as in terms of the redshift as
1 (3 + q0) 2 dA ≈ z − z (1.29) H0 2
Which implies, as for the luminosity distance, the angular diameter distance is practically identical to the proper distance for low redshifts.
z dA ≈ dprop ≈ (1.30) H0
For large redshifts, the angular diameter distance and the luminosity distance are generally different from each other and differ from the proper distance. The specifics of how they differ depends upon the particular model of matter-energy used.
The use of standard yardsticks and angular diameter distances in determining cosmological parameters is complicated by several practical difficulties. First of all, assigning an angle δθ to galaxies and clusters is difficult since these objects rarely have smooth luminosity distributions. Secondly neither galaxies nor clusters are static and isolated. Galaxies tend to grow with time as they merge with other galaxies. Clusters grow with time as well as more and more galaxies fall into them. Since angular diameter distances are defined such that they would have been proper distances in a static Euclidean Universe, they are useful in gravitational lensing calculations (see §
17 1.6).
Luminosity distances are somewhat more promising as cosmological probes and have been used extensively. Data from Type IA Supernovae, in combination from the
Cepheid data from the HST Key project, have played a key role in determining the
Hubble parameter.
Also, data from Supernovae and CMB anisotropies (see section 1.7) have allowed for the geometry of spacetime to be determined - they have shown that our Universe appears to be spatially flat and dominated by vacuum energy.
1.4 The Friedmann Equation
The dynamical evolution of the FRW spacetime is governed by the Einstein equations, which follow directly from the action SEH + SM
1 Z √ S = − d4x −g (R + 2Λ) (1.31) EH κ Z X 4 √ SM = d x −gLfields (1.32) all fields
SEH is the familiar Einstein-Hilbert action for general relativity and SM denotes the sum of the Lagrangian densities of all the fields. κ = 8πG, where G is Newton’s
2 √ constant and G = 1/mPl, the inverse Planck mass squared. −g is the determinant of the metric and Λ is a constant (which will later be identified as the cosmological constant).
Varying the action with respect to the metric gµν one obtains the Einstein equa-
18 tions 1 R − Rg = κT + Λg (1.33) µν 2 µν µν µν
Here Rµν and R stand for the Ricci tensor and scalar respectively, and Tµν stands for the energy-momentum tensor for all fields present. It is conventional to put the Λ
term on the right hand side, which allows it to be interpreted as a source of energy-
momentum of vacuum. The energy density associated with it is given by ρΛ = Λ/κ.
The simplest possible choice for Tµν which is in accordance with the symmetries of the FRW metric is that of a perfect fluid with pressure p and density ρ.
Tµν = −pgµν + (p + ρ) uµuν (1.34)
Plugging (1.34) into the Einstein equations (1.33) one obtains the Friedman-
Lemaitre equations
a˙ 2 κρ k Λ H2 = = − + (1.35) a 3 a2 3 a¨ Λ κ = − (ρ + 3p) (1.36) a 3 6
(1.35) is called the Friedman equation and (1.36) is called the acceleration equa- tion.
It is sometimes convenient to write the acceleration equation in terms of H as
follows: k H˙ = −4πG (p + ρ) + (1.37) a2
One more useful equation can be obtained, either by combining (1.35) and (1.36),
19 µν or from the energy conservation equation T;µ , or simply from the First Law of Ther- modynamics:
ρ˙ = −3H (p + ρ) (1.38)
The Friedman equation can be recast as
K = Ω − 1 (1.39) H2a2 tot
P where Ωtot = Ωi and Ωi = ρi/ρc.Ωi is the density parameter of the species i, i a dimensionless measure of density, expressed as a multiple of the critical density
2 −26 2 −3 ρc, defined as ρc = 3H /κ = 1.88 ∗ 10 h kg m [3]. Ωtot is the total density parameter, which is equal to the sum of the contributions of all the different fields P present in the model, as well as the vacuum: Ωtot = Ωi,. i
Combining (1.39) (at t = t0) and (1.35) we can get one more useful form of the Friedman equation: 2 H X Ωi,0 1 − Ωtot,0 = + (1.40) H2 a(t)3(1+wi) a(t)2 0 i An equation connecting the pressure p and density ρ of the fluid is called the
equation of state:
p = wρ (1.41)
where w is the equation of state parameter. For pressureless matter w = 0, for
radiation w = 1/3 and for a cosmological constant w = −1/3.
20 Equation (1.39) makes it clear that Ωtot determines the curvature of the Universe:
k = +1 ⇒ Ω > 1
k = 0 ⇒ Ω = 1 (1.42)
k = −1 ⇒ Ω < 1
1.5 Model Universes
If the Universe has only one (spatially homogeneous) component, equations (1.35),(1.38) and (1.41) can be solved with appropriate boundary conditions to completely deter- mine the pressure p(t), density ρ(t) and the scale factor a(t). Assuming a single component is obviously a oversimplification since the Universe is clearly a mixture of several types of matter/energy. However, it is still useful to study such solutions as some of these components have individually dominated during different phases of the Hot Big Bang evolution. For example, after nucleosynthesis the Universe passed through a radiation-dominated phase, followed by a matter-dominated phase and currently seems to be in a vacuum-energy dominated state.
In this section I will review a few simple one and two-component models (which can be analytically solved) and which help us understand different epochs in the history of the Universe.
21 1.5.1 An Empty Universe
Let us start with the simplest solution - a Universe which has no matter or energy
in it (i.e. ρ = 0 and Λ = 0). From equation (1.35) we notice that there are two
possibilities:
1. A Universe with H(t) = k = 0, which is a static flat spacetime.
2. A Universe with H(t) = t−1 and k = −1, an empty negatively curved universe
also called a Milne Universe.
A purely empty Universe is not very interesting from a physical point of view, but a Universe with a very small matter density can be approximated by it. However, observations clearly rule out this possibility for our Universe.
1.5.2 Generalized Flat One-Component Models
Now consider the general equation of state (1.41) with a constant w. From (1.38),
one obtains the relationship
ρ(t) ∝ a−3(1+w) (1.43)
Note that at early times, when a is small, if w > −1/3 the density term in the
Friedmann equation (1.35) will dominate the curvature term k/a2. This implies that
curvature can be neglected at early times, but can dominate at later times since it
redshifts slower than matter or radiation, unless of course the Universe is already
vacuum-dominated by then.
In the discussion below, we will neglect spatial curvature and deduce the properties
of a one-component Universe with a general equation of state parameter w.
22 For a spatially flat Universe consisting of only a single component, one can inte- grate (1.35) to obtain (for w 6= −1)
2/3(1+w) a(t) = (t/t0) (1.44)
where the present time t0 (the age of the Universe) is given by
1 r 4 t0 = (1.45) (1 + w) 3κρ0
The energy density ρ scales as
t −2 ρ(t) = ρ0 (1.46) t0
The age of the Universe is related to the Hubble time by the relationship
2 t = H−1 (1.47) 0 3(1 + w) 0 implying that the Universe is younger or older than the Hubble time depending on whether w > −1/3 or w < 1/3.
The proper distance to an object from which light was emitted at a time t = te can be easily calculated from (1.18) to be (when w 6= −1/3)
" 1+3w # 3+3w 3(1 + w) te dprop = t0 1 − 1 + 3w t0 1 2 = 1 − (1 + z)−(1+3w)/2 (1.48) H0 (1 + 3w)
23 For a given point O in a spacetime, the particle horizon distance (or simply horizon distance for short) dh can be defined as the farthest point in causal contact with O. An observer at O has no information about points more distant than the horizon distance because light has not had the time to reach O.
The distance to the horizon can be found from (1.48) in the limit z → ∞. Clearly, whether the horizon distance is finite or not depends on w. For w > −1/3 (which includes flat matter and radiation-dominated universes) the horizon distance is finite and is given by 1 2 dh = (1.49) H0 1 + 3w
On the other hand, if w < −1/3 (which includes a flat cosmological constant dominated Universe), the horizon distance is infinite.
1.5.3 A Cosmological Constant Dominated Universe
We now consider a Universe dominated by a cosmological constant, which is a species with w = −1.
If the cosmological constant Λ > 0, then from the Friedmann equation (1.35) we get
q cosh Λ (t − t ) positive curvature (k = +1) 3 0 a(t) q = exp Λ (t − t ) flat space(k = 0) (1.50) a(t ) 3 0 0 q Λ sinh 3 (t − t0) negative curvature (k = −1)
While apparently different, these solutions all describe the same spacetime, de-
24 Sitter space, in different coordinates.
If, on the other hand, the cosmological constant is negative, Λ < 0, then according
to the Friedmann equation the spatial curvature has to be negative as well. This is
called Anti-de Sitter space, and the scale factor behaves as
r ! a (t) Λ = sin − t (1.51) a(t0) 3
In the rest of this subsection we will study the properties of the flat space solution in (1.50), which describes a Universe dominated by a cosmological constant.
First note that according to (1.38), the energy density of the w = −1 species, ρΛ is a constant. If Λ is interpreted as vacuum energy, then the constancy of the energy
density can be physically understood as the constant creation and annihilation of
virtual particle-antiparticle pairs.
The Friedmann equation (1.35) implies that
a˙ κΛ = ≡ H (1.52) a 3 0
which implies that the Hubble parameter is a constant (H = H0), and that space expands exponentially.
The proper distance at (t = t0) to an object with redshift z can be calculated from (1.18) to be
dprop(t0) = z/H0 (1.53)
The proper distance to the same object at (t = te) was obviously shorter by a factor
25 of (1 + z), i.e., 1 z dprop(t0) = (1.54) H0 1 + z
−1 In the limit z → ∞, dprop(t0) → ∞ but dprop(t0) → H0 . This implies that highly redshifted objects z >> 1 at actually located far beyond the Hubble distance at the time they are observed. This is simply a consequence of the existence of the MAS
(introduced in § 1.35): once an object moves outside the MAS (which coincides with the Hubble distance for an FRW cosmology) it can no longer be observed.
1.5.4 de Sitter space
The spacetime described above is called de Sitter space ([17, 18]). While the (r, θ, φ, t) coordinates together with the assumption of spatial flatness make it easy to visualize the properties of this spacetime, in general, these coordinates do not span the entirety of the space, and flatness is also not required.
A broader (albeit coordinate dependent) definition of de Sitter space is a maxi- mally symmetric spacetime of positive curvature (k 6= 0). It might be visualized as a
4-dimensional hyperboloid
− u2 + w2 + x2 + y2 + z2 = α2 (1.55) which is embedded in a five-dimensional space having the metric
ds2 = −du2 + dw2 + dx2 + dy2 + dz2 (1.56)
Let a set of 4-dimensional coordinates (χ, θ, φ, t) be actually induced on the hyper-
26 boloid through the relations
u = α sinh(t/α)
w = α cosh(t/α) cos(r)
x = α cosh(t/α) sin(r) cos(θ) (1.57)
y = α cosh(t/α) sin(r) sin(θ) cos(φ)
z = α cosh(t/α) sin(r) sin(θ) sin(φ)
The induced metric on the hyperboloid therefore becomes
ds2 = −dt2 + α2 cosh2(t/α) dχ2 + sin2 (χ) dθ2 + sin2 (θ) dφ2 (1.58)
This metric clearly describes a spatial 3-sphere whose dimensions are governed by the t-dependent ”scale factor” cosh2(t/α). As t progresses from −∞ to +∞, the
sphere first contracts, reaches a minimum radius at t = 0, and then expands. It
can be shown that this set of coordinates spans the entire space [19]. Clearly, the
coordinates in (1.50) are incomplete in the past.
We emphasize that this physical picture is purely an artifact of the choice of
coordinates, and there can be alternative descriptions which are equally valid.
1.5.5 Flat Matter Dominated Universe
The properties of such a Universe can be deduced directly from § 1.5.2 by setting
w = 0.
27 −3 2/3 The matter density and the scale factor evolve respectively as ρm ∝ a a(t) ∝ t . The proper distance to a galaxy of redshift z is given by
2 1 dprop(t0) = 1 − √ (1.59) H0 1 + z
implying a finite horizon size equal to 2/H0.
The age of the Universe is given by t0 = 2/(3H0).
1.5.6 Curved Matter Dominated Universe
A Universe filled with matter received considerable attention in the mid-twentieth
century. The fate of this Universe depends crucially on the total matter density Ωtot, as can be seen from the Friedmann equation (1.40):
2 H Ωtot,0 1 − Ωtot,0 2 = 3 + 2 (1.60) H0 a(t) a(t)
Since the Hubble term is squared, it is clear that the scale factor can have either a contracting (˙a < 0) or an expanding (˙a > 0) solution, and that the contracting
solution is the time reversal of the expanding solution.
First consider a positively curved Universe Ωtot > 1. From (1.60) it is clear that if the Universe starts out expanding from a small a (the ”Big Bang”), it will keep
expanding till it reaches a maximum scale factor aturn, and then start contracting till the scale factor vanishes and the matter density diverges (the ”Big Crunch”) . The
scale factor at turnaround (˙a = 0) can be easily deduced to be
28 Ωtot,0 aturn = (1.61) Ωtot,0 − 1
For a negatively curved Universe Ωtot > 1, (1.60) implies that if the Universe is
expanding at t = t0, it will continue to expand forever. Initially, the curvature term will be negligible and the Universe will behave like a flat matter dominated Universe
(subsection 1.5.5). Later the curvature term will dominate and the Universe will
behave like the Milne Universe (subsection 1.5.1).
Interestingly, equation (1.60) does have analytical solutions for both the positive
and negative curvature cases, which allow for the above conclusions to be deduced
rigorously (see for example [1]).
For Ωtot > 1 the solution for the scale factor a(t) is given parametrically as
1 Ω a(θ) = tot,0 (1 − cos θ) 2 (Ωtot,0 − 1) 1 Ωtot,0 t (θ) = 3/2 (θ − sin θ) (1.62) 2H0 (Ωtot,0 − 1)
where θ runs from 0 to 2π. The above expression allows one to calculate the time duration τ between the ”Big Bang” and the ’Big Crunch”:
π Ωtot,0 τ = 3/2 (1.63) H0 (Ωtot,0 − 1)
For the negative curvature case (Ωtot < 1) the solution for the scale factor a(t) is given by
29 1 Ωtot,0 a (η) = 3/2 (cosh η − 1) 2 (Ωtot,0 − 1) 1 Ωtot,0 t (η) = 3/2 (sinh η − η) (1.64) 2 (Ωtot,0 − 1)
1.5.7 Flat Radiation Dominated Universe
The Friedmann equation (1.35) makes it clear that the early Universe was a hot and dense gas of relativistic particles. In addition, particles which make up matter now must have once been relativistic when the Universe was sufficiently hot, and thereby contributed to radiation. The properties of such a Universe can be deduced by setting w = 1/3 in the equations in § 1.5.2.
The density of relativistic particles scales as ρ ∝ a−4 and the scale factor grows as a(t) ∝ t1/2. The extra factor of a difference from the matter dominated case can be physically understood as coming from the energy loss from redshift.
A galaxy at redshift z has a proper distance of
1 z dprop(t0) = (1.65) H0 (1 + z)
Taking the limit z → ∞ we note that there exists a finite horizon of size 1/H0.
Thermodynamics of a radiation-dominated Universe
During the radiation dominated phase photons and other relativistic species are in
thermal equilibrium at the same temperature with zero chemical potential. The
30 energies, number densities and pressures of all species can be readily computed from
thermodynamic formulae (see [1]). The total energy density can be written as
2 " 4 4# 2 π X Ti 7 X Ti π ρ = g + g T 4 ≡ g (T )T 4 (1.66) R 30 i T 8 i T 30 ∗ i=bosons i=fermions
Here only relativistic bosonic and fermionic species are being summed over, since
the contributions of non-relativistic species are exponentially smaller. We assume
that Ti ' T . The value of g∗ depends upon the particle physics model being used.
In the standard model of particle physics g∗ = 106.75 and can be as high as several hundred in more complicated models [9].
Using (1.66), and the time dependence of the energy density and scale factor as
described above, one can find an expression linking the age of the Universe to its
temperature during the epoch of radiation domination:
1/2 90 −2 t = 2 T (1.67) 4π κg∗ (T )
Or, expressing t in seconds and T in MeV, one can rewrite the above as
t 2.4 T 1/2 = 1/2 (1.68) sec g∗ (T ) MeV
From the first and second laws of thermodynamics, it is easy to show that (see [1]) under conditions of thermal equilibrium, the entropy per comoving volume S, given by
31 a3(p + ρ) S = (1.69) T
is a constant. Defining the entropy density s as s = S/a3 = (p + ρ) /T . Since the entropy density is dominated by relativistic particles, it can also be expressed in a form similar to (1.66) as
2π2 s = g T 3 (1.70) 45 ∗S
T where g∗S = g∗ , but is numerically close to g∗ since all species are at the same Ti temperature.
3 3 The fact that g∗Sa T remains constant as the Universe expands shows that as long
1 as the number of degrees of relativistic species g∗S stays the same, T ∝ a . g∗S changes when entropy is produced, e.g.through a first-order phase transition or the decoupling of a species, but even then the amount of entropy produced is small compared to the total entropy, and hence the inverse relationship between temperature and scale factor is a good approximation.
1.5.8 Matter Radiation Equality
The previous subsections have shown that density of radiation scales as a−4 and the density of matter scales as a−3. Initially radiation is dominant, but since it redshifts faster than matter, the latter comes to dominate at later times. This implies that the
Universe passes through a phase where matter and radiation have the same density.
This epoch in the history of the Universe is called the epoch of matter radiation
32 equality. In what follows, the subscript “eq” denotes the value of a quantity at the
time of equality.
The redshift at matter-radiation equality (zeq) is given by
Ωm,0 3 1 + zeq = ' 3 × 10 (1.71) Ωr,0
The temperature at matter-radiation equality is given by
3 Teq = T0 (1 + zeq) ' 9 × 10 K (1.72)
The Friedmann equation becomes
κρ a 3 a 4 H2 = eq eq + eq (1.73) 3 a a
This equation can be solved for the time behavior of the scale factor in terms of conformal time η [9]
a(η) √ η √ η 2 = (2 2 − 2) + 1 − 2 2 + 2 (1.74) aeq ηeq ηeq
where √ s (2 2 − 2) 3 ηeq = (1.75) aeq κρeq
For η ηeq the first term in (1.74) dominates, giving the radiation-dominated
2 behavior a ∝ η . When η ηeq the second term dominates, giving the matter domi- nated behavior a ∝ η. Equation (1.74) shows that in the case of a Universe consisting
of both matter and radiation, the transition from radiation to matter domination is
33 Figure 1.2: Gravitational lensing geometry. The lens distorts the ”true” angles β (that would have been seen without any lensing effects) into the angles θ. Source [15] smooth.
1.6 Gravitational Lensing
This is the phenomenon where the gravitational distortion of spacetime by a massive body like a galaxy causes the latter to act as a lens, creating a variety of different types of images of objects behind it. In mathematically treating gravitational lenses, one usually makes the assumption that they are thin-lens systems, where the bending of light occurs at a single distance, as shown in Fig. 1.2. The angles are thought of as 2-dimensional vectors on the sky.α ˆ is called the deflection angle, and ~α ≡ β~ − θ~ is called the reduced deflection angle.
Since lensing occurs in an expanding and possibly curved Universe, one needs to
34 choose distance measures carefully. It is clear that angular diameter distances are the most appropriate choice here, since the way these distances are defined, ratios such as dLS/dS (where both are angular diameter distances) are not going to change either with expansion or curvature. All distances used in gravitational lensing are therefore angular diameter distances.
From the geometry of Fig. 1.2, one can write down the lens equation for a gravi- tational lens:
d β~ = θ~ − LS αˆ (1.76) dS
Consider the simplest case of a circularly symmetric lens of mass M. We know that the deflection angle for a photon traveling through a gravitational potential Φ is given by
Z ~ αˆ = 2 ∇⊥Φds (1.77)
It is easy to show that equation (1.76) reduces to
d 4GM β = θ − LS (1.78) dLdS θ
If the source is on-axis (β = 0), then the above equation has a unique solution:
r 4GMdLS θE = (1.79) dLdS
This implies that the images form a ring (the Einstein ring) around the source
35 with radius θE, which is also called the Einstein angle. The Einstein angle sets a scale for gravitational lensing and can be used for a rough estimate of the amount of
lensing by a given object. Measuring Einstein angles of lensing galaxies allows one
to determine masses of central parts of galaxies very accurately, and generally these
measurements are in good agreement with other independent observations.
If the source lies off-axis (β 6= 0), then one can solve (1.78) to get two image
locations, one inside and one outside the Einstein radius.
1 q θ = β ± β2 + 4θ2 (1.80) ± 2 E
In the case of lenses which cannot be treated as point masses, it is customary to
define a lensing potential by integrating the gravitational potential of the source over
past-directed geodesic paths emanating from the observer:
d Z ψ θ~ = 2 LS Φds (1.81) dLdS
The reduced lensing angle is now given by
~ ~α = ∇θψ (1.82)
Lensing can cause two effects: convergence and shear, which can both be defined from the lensing potential.
The convergence ζ, which measures the degree of focusing, is given by:
36 1 ζ = ∇2ψ (1.83) 2 θ
The shear γ which measures the shape distortion, and is given (in terms of carte-
sian components) by:
1 γ2 = (∂ ψ − ∂ ψ)2 + (∂ ψ)2 (1.84) 4 xx yy xy
The lens transforms an area element given by β~ into one described by θ~. This
transformation can be characterized by the magnification µ, which is the determinant
of the magnification tensor µij: ∂θi µ = (1.85) ij ∂βj
The magnification µ, can be expressed as a combination of the convergence and
shear as [15]
1 µ = (1.86) (1 − ζ)2 − γ2
Gravitational lensing effects are classified into three categories:
Strong lensing
This is the strongest gravitational lensing effect and occurs when the source lies within the Einstein radius of the lens. The lensing might cause an Einstein ring or (more likely) a series of tangential or radial arcs. From a given set of images, it might also be possible to reconstruct the lensing potential (a classic example of this is in [9801193]).
Also, if the source is time-varying, the multiple images will vary in time too, but
37 not exactly together, since light takes different times to reach us in forming different images. The time lags can be used, in principle, to measure the Hubble parameter.
However, the success of this method depends on the accuracy with which one can reconstruct the lensing potential.
Weak lensing
If the source is located more than an Einstein distance from the lens, the images will have very small amounts of shear and magnification. Here one uses a statistical approach: by studying a collection of galaxies (which might have been lensed), one looks for a statistical overall shear or magnification imposed on the distribution (which would have bee absent for an unlensed distribution of randomly oriented galaxies).
For instance one could search for correlated distortions in shapes. These statistical measures of the lensing effect are then used to map the lensing potential. This tool has been very useful in measuring the dark matter haloes in which galaxies are embedded.
Micro lensing
This is the temporary one-time brightening of a source, possibly due to the passage of a MACHO (massive compact halo object) through the line of sight to the source.
The brightening is a consequence of the magnification caused by the lens. A micro- lensing event is easy to distinguish from a fluctuation in the source itself, because the micro-lensing light curve can be precisely determined using general relativity, and should be identical in all frequency bands. The MACHO collaboration has found 450 microlensing events from studying 50.2 million lightcurves in the Large Magellanic
Cloud [20].
38 1.7 The Composition of the Universe
In § 1.4, I introduced the concept of the density parameter Ω, which measures the densities of different components of the Universe. Of crucial importance to cosmology is the determination of the total density parameter Ωtot, and the density parameters of all the individual components. After several decade of measurements, these numbers have been determined.
It turns out that we live in a spatially flat Universe consisting of three major components whose estimated density parameters are given below:
1. Ordinary (baryonic) matter (' 4%)
2. Dark Matter (' 26%)
3. Dark Energy (' 70%)
This is called the Λ-cold dark matter (ΛCDM) cosmology and seems to be supported by evidence from a variety of different sources. In what follows I will briefly review the observational evidence that we have for this model.
1.7.1 Measuring the Total Matter Content
Galaxy clusters, the largest collapsed structures known, prove to be an invaluable tool in placing constraints on ΩM . They are considered to have arisen from the collapse of initial perturbations about a tenth of the comoving horizon size. As a result their evolution is dominated by gravitational dynamics in the linear or weakly non-linear regime, where initial conditions have not yet been erased by gas dynamics.
39 The masses of clusters can be determined using three independent techniques, which agree with each other up to ∼ 1 Mpc (see [21] and references therein):
1. Velocity Dispersion: Under assumptions of hydrostatic equilibrium, the distri-
bution of velocities of galaxies in a cluster can be used to determine the cluster
mass.
2. ICM Temperature: The mass of the cluster can also be traced from the tem-
perature of the intercluster gas.
3. Lensing: If the cluster acts as a gravitational lens, the lensing effects can be
used to determine its mass.
Clusters can be used to estimate ΩM in several different ways, which I briefly mention below.
Mass to Light ratio
The traditional method is to calculate the typical mass to light ratio (M/L) for rich clusters, and then integrate it over the entire observed luminosity density of the
Universe. It turns out that a median mass-to-light ratio of 300 ± 100h is observed for most clusters, almost independently of the individual characteristics of the clusters such as their luminosities and velocity dispersions [21]. Studies of this sort suggest that we live in a low density Universe with ΩM ∼ 0.2−0.3 [22]. The above method has been sharpened through the use of sophisticated techniques to measure the masses of clusters, such as gravitational lensing [23] and temperature profiles [24], but the conclusion remains the same.
40 Baryon Fraction
Another method is to use the baryon density instead of the luminosity density [25].
The idea here is to estimate the fraction of baryons fICM within the intercluster mediums of galaxy clusters, and then to put an upper limit to Ωb using the relation
ΩM = Ωb/fICM (1.87)
Here, of course, we are assuming that the entire Universe has the same ratio of baryons to total matter as do the clusters. This assumption is not unreasonable since on very large scales there is no reason to expect baryons to be significantly segregated from dark matter. fICM has been measured by studying X-ray data from galaxies [26] as well as through the Sunyaev-Zeldovich effect [27], and both indicate that the mass density of the Universe is low (Ω ' 0.2).
Cluster Abundance and Evolution
Constraints have also been placed on ΩM from observations of cluster abundances, combined with studies of evolution of rich massive clusters with redshift. The basic idea here is that the growth of high mass structures depends on cosmological param- eters, in particular ΩM and σ8 (which is the root-mean-square mass fluctuation on 8h−1 Mpc scale and measures the bias in the distribution of mass vs light). It turns out that if ΩM is taken to be large (ΩM = 1) then fluctuations start growing late and produce a strong evolution in recent times z > 1. The opposite is expected to happen in lower density models where fluctuations freeze out much earlier. Observations seem to support the latter scenario we see very little evolution in recent times. The matter
41 density of the Universe is inferred to be ΩM ' 0.2 − 0.3 [28, 29, 30].
Gravitational Lensing Studies
Strong lensing provides constraints on ΩM and ΩΛ, and show that we live in a low- density Universe (Ω < 0.4) regardless of the value of ΩΛ. Weak lensing studies using several cluster lenses shows that Ωm ' 0.3 with a significant contribution from Dark Matter [14].
Galaxy Correlation functions
Analyzing the bias parameter of galaxies b1 (which quantifies the strength of cluster- ing of the galaxies relative to the mass in the Universe) and the redshift correlation function β from the 2Df Galaxy Redshift Survey, one can get an independent mea-
0.6 surement of the mass density from the relationβ = Ωm /b1 which gives a value of
ΩM = 0.27 ± 0.06 [31]. This is completely independent of, yet in good agreement with, all the other measurements of ΩM described above.
1.7.2 Ordinary and Dark Matter
There is considerable evidence that there exist two types of matter - the ordinary matter we are familiar with, which is mostly made of baryons. The density parameter of baryonic matter is Ωb = 0.04 ± 0.02 [3]. This estimate comes from Big Bang nucleosynthesis, CMB power spectrum and direct counting of baryons.
Dark Matter is the other component is dark matter, which appears to be almost perfectly invisible except for its gravitational interaction. I now briefly review some
42 of the observational evidence for the existence of a dark matter component.
Rotation curves of spiral galaxies provide strong evidence in favor of the existence
of dark matter (the first such evidence came from the work of Zwicky in studying a
group of galaxies in the coma cluster [32]). For virialized galaxies, the mass up to a
given distance r, M(r) is related to the rotational velocity at that distance v by the relation
M (r) ∝ v2r/G (1.88)
The rotational velocities can be measured from the HI 21 cm lines. Away from the
luminous part of the galaxy, if there is no more matter, then equation (1.88) indicates
that the rotation velocities should fall off with r. However, observations show that
they are flat, suggesting that the mass distribution goes as M(r) ∝ r beyond the
point where light ceases, in some unseen form [14].
Another piece of evidence comes from studying the distribution of X-ray emitting
hot gas in galaxies, calculating the amount of mass necessary to bind this gas, and
comparing it to the amount of visible matter in the galaxy. The latter has fallen short
of the required amount in several galaxies, indicating the presence of invisible forms
of matter. One example is the elliptical galaxy M-87 where the shortfall is ∼ 1% [14].
The latest evidence of Dark Matter comes from X-ray and weak lensing obser-
vations of the merging cluster system 1E0657-556. The collision shifted the X-ray
plasma (containing 90% of the baryons) in both the clusters away from the galaxies
(which only contain 10% of the baryons). This separation makes it possible to de-
termine which of these (i.e. plasma or galaxies) is predominantly responsible for the
43 lensing. Observations show that the lensing surface potential is more in agreement with galaxies and not with the X-ray plasma [33]. In [33], it is argued that these observations are a clear indicator of an invisible matter form regardless of the form of the gravitational force law. However, alternative explanations for these observations have also been put forward, based on modified gravity theories [34, 35].
Dark Matter is not just suggested by observations - it has at least one interesting theoretical motivation - dark matter seems to be crucial to the theory of structure formation. The early Universe was radiation dominated, a fluid of baryons and pho- tons coupled together. Gravitational clumping of the baryons was not possible at this time. As the fluid cooled, the baryons and photons decoupled at 1 + z = 1100.
Assuming that perturbations are adiabatic, both the photon fluid and the baryon
fluid will inherit the curvature perturbations entering the horizon at that time, which also show up as anisotropies in the CMB background and have been measured (by
COBE and WMAP) to have an amplitude ∼ 10−5. With such a small initial am- plitude we would expect the baryon overdensity to grow in proportion to the scale factor, i.e., in accordance with linear perturbation theory. As a result, by today these perturbations will have grown by a factor of 1100 , and hence we can expect typical structures today to have an overdensity of ∼ 0.1, which is far too low compared to what is observed (typical galaxies have over densities of 105. However, if we assume the existence of dark matter, then by the time of decoupling the dark matter will have formed potential wells into which the baryons can fall and hence immediately commence non-linear growth.
44 1.7.3 Supernovae and Cosmic Acceleration
Supernovae
Supernovae are spectacular stellar explosions, caused either by the collapse of a stellar core, or through runaway nuclear fusion in a white dwarf. Supernovae can release up to 1031 ergs of energy. Only a small fraction of which is released as visible light, but even that is typically sufficient to temporarily outshine the entire host galaxy. The explosion shoots out stellar material into the inter stellar medium (ISM) at speeds as high as a tenth of the speed of light, setting up a shock wave in the ISM.
The classification scheme of supernovae is based on observation. The main cate- gories are Type I and Type II depending on the absence or presence of Hydrogen lines in their spectra respectively. A detailed classification scheme is shown in Fig. 1.3.
However from an astrophysical point of view, it is found that the Type II are similar to the Type Ib and Ic, in that they all originate from the collapse of the stellar core.
The Type II are believed to be formed when the cores of massive stars (> 8 solar masses) collapse to form black holes or white dwarfs. Type Ib and Ic are also formed from core-collapse after the hydrogen-rich (Ib) or helium-rich (Ic) outer mantles are blown away by stellar winds.
The Type Ia supernova have a different astrophysical origin. They are formed from binary systems where one of the stars is a white dwarf. As the latter accretes more and more mass from the companion star, it eventually crosses the Chandrasekhar limit [36] and collapses until its density triggers nuclear fusion reactions throughout the star, causing the star to explode.
Since the Chandrasekhar limit is a universal phenomenon, one would expect the
45 Early Spectra: No Hydrogen / Hydrogen
SN I SN II Si/ No Si ~3 mos. spectra He dominant/H dominant
SN Ia He poor/He rich 1985A 1989B SN Ic SN Ib SN IIb “Normal” SNII 1983I 1983N 1993J 1983V 1984L 1987K Light Curve decay after maximum: Linear / Plateau Core collapse. Believed to originate Most (NOT all) from deflagration or H is removed during detonation of an evolution by tidal stripping. accreting white dwarf. SN IIL SN IIP Core Collapse. 1980K 1987A Outer Layers stripped 1979C 1988A by winds (Wolf-Rayet Stars) 1969L or binary interactions Ib: H mantle removed Core Collapse of Theory Ic: H & He removed a massive progenitor with plenty of H .
Figure 1.3: A flow chart showing the classifications of Supernovae http://rsd-www. nrl.navy.mil/7212/montes/snetax.html.
46 luminosities of Type Ia’s to be identical, i.e., Type Ia’s are perfect candidates for standard candles. In practice, while a large majority of the true type Ia SNe have very similar light curve shapes, spectral time series and absolute magnitudes, there does exist a scatter in their peak luminosities (see [13] and references therein).
It was first noted by Phillips [37] that a strong correlation existed between the rate at which a Type Ia SNe’s rate declines (measured by the parameter ∆M15 defined as the reduction in brightness over 15 days following maximum light) and its absolute magnitude. Eliminating this correlation can reduce the dispersion in the maximum luminosities. Several techniques have been developed to do this.
Hamui [38] used this correlation between maximum luminosity and ∆M15 to reduce the scatter in the Hubble diagram for a sample of 30 SN1a from the CTSS search. A more sophisticated technique called the multi-color light curve shape method (MLCS) was developed by [39]. Other popular methods include the “stretch” method [40, 41] and the Color-Magnitude Intercept Calibration (CMAGIC) method [42]. All of these methods allow supernovae distances to be determined to a precision of 6% after correcting for photometric uncertainties and peculiar velocities [13] making them very useful standard candles.
Supernovae Magnitudes
The apparent magnitude m of a light source is defined in relation to a reference
−8 −2 flux fx = 2.53 × 10 watt m as
m ≡ −2.5 log10 (f/fx) (1.89)
47 The absolute magnitude M of a light source is similarly defined in relation to
a reference luminosity of Lx = 78.7 solar luminosities
M ≡ −2.5 log10 (L/Lx) (1.90)
(Lx is the luminosity of an object that produces a flux of fx when it is placed at a luminosity distance dL = 10 parsec. The physical interpretation of the absolute mag- nitude is therefore clear: it is the apparent magnitude a source placed at a luminosity distance of dL.) The distance modulus, defined as m − M, is a convenient measure of the lumi- nosity distance in terms of the apparent and absolute magnitudes. From equations
(1.89) and (1.90)
d m − M = 5 log L + 25 (1.91) 10 1 Mpc
From equation (1.25), we can connect the distance modulus to the redshift and the acceleration parameter:
H m − M ≈ 43.17 − 5 log 0 + 5 log z + 1.086 (1 − q ) z (1.92) 10 70km s−1 Mpc−1 10 0
Hence it is clear that if the apparent flux and the absolute flux (or equivalently, m and M) are known for a population of standard candles, the parameters H0 and
q0 can be determined from a plot of the distance modulus m − M vs z.
48 Supernovae results
Fig. 1.4 shows such a Hubble plot incorporating data from both the Supernova
Cosmology Project [41] and the High-z Supernova search team [43]. The data are
compared to three model Universes as shown. Both sets of datapoints show that
distant supernovae are fainter than one would expect from the Hubble expansion
alone, and hence the Universe is accelerating. The upper panel clearly reveals the
validity of the Hubble law for small redshifts (z 1). The lower panel subtracts out
a negatively curved matter-only Universe from the data and clearly reveals that at
high redshift (z ≥ .5) the supernova are about ∼ .25 mag fainter than what one can
expect from a pure Hubble flow in a matter-only Universe.
The supernovae data allow for a wide range of both possibilities for ΩM and ΩΛ (see Fig.1.5), but in general are better fit by cosmological constant dominated models. If we assume that we know either one of ΩM or ΩΛ, we can establish tight constraints on the other one. For instance, choosing ΩM = 0.3 based on, for instance, the large scale structure evidence described in § 1.7.1, the supernovae results imply that ΩΛ = .7
1.7.4 Cosmic Microwave Background Anisotropies
In § 1.1.3, I had introduced the CMB. I now proceed to discuss the wealth of infor- mation that lies in its anisotropies.
The CMB has a dipole anisotropy (see Fig. 1.6), meaning that in one half of the sky the CMB blackbody spectrum is blue-shifted and in the other half it is redshifted.
This is a consequence of the earth not being a comoving frame, and having a relative velocity in a frame in which the CMB is isotropic. The COBE satellite is orbiting the
49 44 High-redshift (z > 0.15) SNe: High-Z SN Search Team 42 Supernova Cosmology Project
40 Low-redshift (z < 0.15) SNe: CfA & other SN follow-up 38 Calan/Tololo SN Search ΩΜ=0.3, ΩΛ=0.7 36 ΩΜ=0.3, ΩΛ=0.0 Distance Modulus (m-M) 34 ΩΜ=1.0, ΩΛ=0.0
1.0 =0.0 Λ Ω 0.5 =0.3, Μ
Ω 0.0
-0.5 (m-M) - 0.01 0.10 1.00 z
Figure 1.4: top panel: A Hubble diagram made from data from both the Supernova Cosmology Project and the High-z Supernova Search Team taken from [13].bottom panel The residual of the distances relative to a ΩM = 0.3, ΩΛ = 0.7 Universe.
50 3
No Big Bang SCP
2 HZSNS
1 accelerating decelerating
expands forever 0
(cosmological constant) lapses eventually vacuum energy density recol
closed
flat -1 open
0 1 2 3 mass density
Figure 1.5: Best fit regions in the (ΩM ,ΩΛ) plane for data from both the Supernova Cosmology Project and the High-z Supernova Search Team. The agreement of the two experiments is remarkable. Source [13]
51 Figure 1.6: The Dipole Anisotropy in the CMB as seen by COBE http://map.gsfc. nasa.gov/m_uni/uni_101Flucts.html. earth at ∼ 8 km s−1, the earth is orbiting the sun at 30 km s−1, the sun is orbiting the galactic center at ∼ 220 km s−1, and our galaxy is orbiting the center of mass of the Local Group at ∼ 80 km s−1. Finally, the Local Group is accelerated towards the
Virgo cluster, which in turn is accelerated towards the Hydra-Centaurus supercluster, as a result of which the Local group is headed in the direction of Hydra at a speed of
.2% the speed of light [5].
After the dipole is subtracted (see Fig. 1.7), the remaining temperature fluctua- tions are extremely small. In fact in Fig. 1.7 the temperature of the hot regions (red) exceeds the temperature of the cold regions (blue) by .0002K.
Let the temperature at any point in the sky be denoted by the function T (θ, φ), where θ and φ are spherical polar angles on the sky. The dimensionless temperature
fluctuation at a given point in the sky is defined as
52 Figure 1.7: The CMB spectrum once the dipole is subtracted. http://map.gsfc. nasa.gov/m_uni/uni_101Flucts.html.
δT T (θ, φ) − hT i (θ, φ) = (1.93) T hT i
Results from COBE, WMAP and a variety of other experiments shows that
−5 (δT/T )rms ∼ 10 . Since these fluctuations are defined on the surface of a sphere centered on the observer, it is convenient to expand the function (δT/T )(θ, φ) in the
spherical harmonics:
δT X (θ, φ) = a Y (θ, φ) (1.94) T l,m l,m l,m and
∗ halmal0m0 i = δl,l0 δm,m0 Cl (1.95)
where the Kronecker delta’s spring from the assumption that the mechanism gener-
53 ating the anisotropies was isotropic. The information regarding the anisotropies is
encoded in the correlation function C (θ) between two points on the last scattering surface in directionsn ˆ andn ˆ0 separated by an angle θ:
δT δT C (θ) = (ˆn) (ˆn0) (1.96) T T
Using the properties of spherical harmonics, the correlation function can be writ-
ten as an expansion in the Legendre polynomials as follows:
1 X C (θ) = (2l + 1) C P (cos θ) (1.97) 4π l l l
The Cl’s can be interpreted as a measure of temperature fluctuations on angular scales 180◦/l. On small sections of the sky where spatial curvature can be neglected, the Spherical Harmonic expansion becomes identical to a Fourier expansion in two dimensions, with l as the Fourier wavenumber [8]. For small angular separations
R 2 2 l 1, the correlation function in Fourier space is d lCl/ (2π ), using which, the power spectrum of temperature fluctuations is conventionally written as
l(l + 1) ∆2 ≡ C T 2 (1.98) T 2π l
The right hand side is now clearly the power per logarithmic interval in l for l 1.
(Another reason for adopting this convention is because the right hand side of (1.98)
is a constant for perturbations generated via the Sachs Wolfe Effect).
How accurately the Cl’s can be measured is limited by cosmic variance. This is the notion that our sky is only one realization of an ensemble, and hence there are
54 only 2l + 1 m-samples of the power in each multipole moment [8] causing an error of:
r 2 ∆C = C (1.99) l 2l + 1 l
Cosmological models generally predict the behavior of Cl vs l. Figure 1.8 shows such a plot, together with data from recent experiments measuring the CMB power spectrum.
The different sources of temperature fluctuations in the CMB are the following:
1. Dipole: These are caused by the earth’s peculiar velocity (already discussed
above).
2. Sachs Wolf: These are caused by fluctuations in the gravitational potential
on the last scattering surface, which in turn are created when superhorizon
metric perturbations enter the horizon. These fluctuations dominate on scales
comparable to the Horizon size at last scattering [45].
3. Intrinsic: These are caused by fluctuations in the photon-baryon plasma.
4. Doppler: These are caused by the non-zero velocity of the plasma at recom-
bination (the last scattering surface) leading to Doppler shifts in the frequency
(and hence the temperature) of the CMB.
The first peak in Fig. 1.8, located at an angle of 1◦ is crucially important. This peak corresponds to the horizon size at last scattering and can be understood as fol- lows. Prior to recombination, the baryons and photons are tightly coupled and photon pressure can provide a restoring force to baryon clustering. As a result coherent os- cillations are set up in the baryon-photon plasma. The rise in temperature associated
55 Angular Scale 90° 2° 0.5° 0.2° 6000
WMAP Acbar 5000 Boomerang CBI VSA 4000 ] 2 K + [ /
/2 3000 l +1)C l ( l 2000
1000
0 10 100 5001000 1500 Multipole moment l
Figure 1.8: The WMAP three-year power spectrum (in black) together with data from other recent experiments measuring the CMB angular power spectrum. Taken from [44]
56 with the compression of the photons acts as damping force (the Silk damping, [46]), and hence the system as a whole can be modelled as a damped harmonic oscillator
[47]. Now, heuristically, we can expect that oscillation modes which are larger than the horizon will not have had enough time to evolve appreciably till last scattering.
For subhorizon modes, the ones which are “caught at the maxima or minima of their oscillation at recombination” [47] will show up as peaks in the Cl vs l plot. Owing to the effect of photon damping, we can expect the modes much smaller than the hori- zon, which have had considerable time to equilibriate, to make smaller peaks than the Horizon sized mode. We therefore expect the largest peak in the Cl vs l plot to correspond to the angular size of the Horizon at the time of last scattering.
Since we know the physical size of the Horizon at last scattering, we can deduce the location of this peak for different spatial geometries. For a flat matter-dominated
Universe, the peak should lie at l ' 220, which is exactly what is observed experi- mentally.
A careful analysis of the entire CMB spectrum allows us to constrain almost all cosmological parameters. However, there are degeneracies in the parameters which allow constraints to be set only when priors are assumed. For example, if one assumes a flat vacuum energy dominated Universe, the WMAP data constrains Ωtot and h. If one assumes curvature, there is a degeneracy between Ωm, h and the curvature, and one of these has to be assumed in order to constrain the others [48].
The combined data from large scale structure, supernovae and the CMB seem to consistently bear out the flat ΛCDM cosmology.
57 1.8 Shortcomings of the Standard Cosmology
For all its successes, the Standard Hot Big Bang model suffers from one major flaw:
several observational features of the Universe, though not logically inconsistent with
the Hot Big Bang, still require remarkable initial conditions. I now briefly review
some of these instances in this section, and in the next section show how the theory
of inflation comes to the rescue.
1.8.1 The Flatness Problem
The Friedmann equation, in the form (1.39) indicates that the difference between
2 Ωtot and 1 increases with time. It increases in proportion to a for a flat radiation- dominated Universe (see § 1.5.7) and in proportion to a for a flat matter-dominated
Universe (see § 1.5.5). If the Universe starts out flat, it stays flat forever. However, if it does not start out flat, then in order to be spatially flat today, then |Ωtot − 1| has to be extraordinarily close to, but not exactly, 0 in the past, a very unseemly initial condition requiring severe fine-tuning.
Peebles and Dicke [49] have pointed out that at t = 1 second, at the beginning of
−15 nucleosynthesis, Ωtot must have equalled 1to an accuracy of 10 . If we assume that General Relativity is valid all the way up to the Planck scale, then from (1.39) (and
−64 assuming radiation domination) one can estimate that |Ωtot − 1|tPl ≈ O [10 ] Another way of looking at the flatness problem is to ask how the Universe got to be so old. Only the highly specialized initial conditions described in the previous paragraph would cause the Universe to survive in its present state for this long. Any other initial conditions would have lead to a curved Universe that recollapses very
58 quickly, or an open Universe that cools to below 3K within one second [9].
1.8.2 The Entropy Problem
This problem concerns the constancy of the entropy per comoving volume S described in § 1.5.7. Using equations (1.39), (1.69) and (1.70) one can show that at the Planck
Time, the entropy within our causal horizon is ≈ O [1060].
Such an incredibly large number demands an explanation.
1.8.3 The Horizon Problem
The crux of this problem lies in the remarkable isotropy of the CMB, even among regions of the sky that have never been in causal contact. In other words, different parts of the sky, which cannot have had any microphysical interaction, seem to have the same temperature. The following simple calculation serves to illustrate this point very clearly:
The size of our visible patch today is roughly equal to the inverse Hubble length
−1 H0 . Let LLS be the radius of this volume at last scattering. Clearly,
−1 a(tLS) −1 T0 LLS = H0 = H0 (1.100) a(t0) TLS
(the subscripts 0 and LS stand for today and last scattering respectively). On the other hand, the Hubble length at last scattering (which defines the causal volume at last scattering), is given by (assuming matter domination since last scattering)
59 −3/2 −1 −1 TLS HLS = H0 (1.101) T0
Taking the ratio of the two volumes (our visible patch and the causal volume) we
find that 3 LLS T0 6 −3 = ≈ 10 (1.102) HLS TLS
This implies that the volume occupied by our visible patch at the time of last scattering contained ∼ 106 causally disconnected regions. The horizon problem is nicely illustrated Fig. 1.9, which shows that for an FRW cosmology, physical scales entering the horizon at a certain time have never been inside the horizon before, simply because the Horizon evolves faster than the scales.
Further, not only do causally disconnected regions in the sky have the same tem- perature, they also seem to have the same degree of temperature anisotropy ∼ 10−5
This leads to the question of whether the large-scale inhomogeneity that is observed
must also be attributed to initial conditions.
1.8.4 The Monopole Problem
All Grand Unified Theories (GUT’s) predict the existence of magnetic monopoles, i.e.,
massive particles carrying a net magnetic charge. In fact Preskill [50] argues that if
the GUT phase transition were second order (or weakly first order), the production of
monopoles would be so copious that they would come to dominate the energy density
of the Universe before the time of Helium synthesis. The standard cosmology pro-
vides no explanation for the observed absence of magnetic monopoles in the Universe
60 Figure 1.9: A plot of length scale vs (logarithmic) scale factor in an FRW cosmology. The blue line shows the evolution of the Hubble scale. The red lines show the evolution of physical scales. Since the Hubble length evolves faster than the physical scales, sub-Horizon modes have never been in causal contact prior to Horizon entry.
61 (especially given the age of the Universe determined from other sources).
1.9 Inflation
1.9.1 Inflation and the Problems of the Standard Cosmology
The theory of inflation [51] dramatically resolves all these problems, simply by propos- ing that for a very brief period in its history, the Universe underwent a period of rapid expansion. Formally, an inflationary phase is one during which the scale factor accel- erates (¨a > 0) [9]. In most models of inflation the scale factor grows exponentially.
In order to not disturb the successes of the Hot Big Bang, inflation must occur before nucleosynthesis. After inflation ends, all the energy responsible for inflation is released as thermal energy through a non-adiabatic process which increases the entropy of the Universe by a very large factor. This process is dubbed reheating.
Equation (1.36) implies that an inflationary phase can only occur if the over- all pressure of the Universe is negative to the extent that p < −ρ/3. In the next subsection I describe how this is made possible from the particle-physics perspective.
It is not hard to see how inflation resolves the problems mentioned in the previous section (see, for example, [1] for a detailed discussion). The resolution of the horizon problem can be seen from Fig. 1.10 which plots the Hubble scale vs time and shows the evolution of modes. In this scenario, it is clear that superhorizon modes entering the Horizon now were once inside the Horizon.
The fact that modes which were superhorizon at the time of recombination were once inside the horizon provides a causal mechanism for the generation of large scale
62 −1 H length scale FRW
k’
k
−1 H inf
t tt t iH0
Figure 1.10: Physical scales entering the horizon at the time of last scattering have been in causal contact before. structure. Metric perturbations in the inflationary spacetime (probably generated from quantum fluctuations in the inflaton) are stretched to superhorizon scales during inflation. When these modes re-enter the Horizon at the time of last scattering, they leave an imprint on the matter content of the Universe at that time. However,
Trodden and Vachaspati [52] have argued that if inflation was preceded by a pre- inflationary stage subject to some reasonable characteristics (i.e., General Relativity being valid, the null energy condition being satisfied, and spacetime topology being trivial), then homogeneity on super-Hubble scales is required as an initial condition.
Models in which inflation originates at the Planck epoch may not be subject to this
63 condition.
The flatness problem is resolved since the inflationary phase drives Ωtot very close to 1 rather than away from it as can be seen from (1.39), especially if the growth of the scale factor is exponential.
Monopoles and other relics can be removed simply by assuming that inflation occurs before (or during) their production. The exponential expansion of spacetime makes their density sufficiently low in order for them to be undetectable later.
The entropy problem is resolved by assuming that the reheating process is non- adiabatic, leading to the creation of a large amount of entropy.
1.9.2 The Dynamics of Inflation
In the previous subsection I show that a brief period of exponential expansion in the early history of the Universe is extremely beneficial in resolving the problems of the
Standard Cosmology. However, the question remains as to how such an inflationary epoch can come about.
As I have shown in § 1.5.3, the Friedmann equation admits an inflationary solution if the equation of state parameter w = −1, implying p = −ρ. A substance with negative pressure is hard to imagine in our everyday existence, scalar fields (which have never been observed experimentally yet) can easily have negative pressure, as I show below.
A scalar field φ is typically described by a Lagrangian:
µ L = ∂ (φ)∂µ(φ) − V (φ) (1.103)
64 and a stress tensor given by:
Tµν = ∂µ(φ)∂ν(φ) − gµνL (1.104) gµν represent the components of the metric, and V (φ) represents the potential. If the field is a quantum field, then φ is actually a quantum operator. The scalar field responsible for inflation is usually called the inflaton. The equation of motion of the
field is the Klein-Gordon equation:
1 √ ∂V √ ∂ −ggµν∂ φ + = 0 (1.105) −g µ ν ∂φ
The energy density and pressure of the field are given by
1 1 ρ (φ) = T 0 = φ˙2 + (∇φ)2 + V (φ) (1.106) 0 2 2 1 1 p (φ) = −T i = φ˙2 − (∇φ)2 − V (φ) (1.107) 0 2 6
In inflationary cosmology, it is customary to split the inflaton field into a homogenous background and a time and space dependent perturbation.
φ (~x, t) = φ0 (t) + δφ (x, t) (1.108)
φ0 is the expectation value of the inflaton field and can be treated as a classical field, and δφ represents quantum fluctuations about the homogenous background. In what follows, I will suppress the subscript 0 in denoting the homogenous part of the field.
65 The equation of motion of φ (t) and δφ (x, t) can be derived from (1.105):
φ¨ + 3Hφ˙ + V 0 (φ) = 0 (1.109)
δφ00 + 3Hδφ0 − ∇2δφ0 + V 00δφ − 4φ0Φ0 + 2V 0Φ = 0 (1.110)
To first order, the equations (1.106) and (1.107) now indicate that if the potential energy of the field dominates its kinetic energy, e.g., V (φ) φ˙2, then p ' −ρ will hold, and the Universe will inflate exponentially. In the next subsection, we will investigate this condition in greater detail.
1.9.3 The Slow Roll Parameters
In order to ensure an inflationary phase, the usual assumption is that the field is rolling slowly down its potential. To make this possible, one uses the approximations:
1. φ˙2 V (φ), and
2. φ¨ 3Hφ˙
For these approximations to be valid, it is necessary for two conditions to hold
1
|η| 1 (1.111)
66 These are called the slow-roll conditions. The slow roll parameters are defined as
[9]
H˙ φ˙2 1 V 0 2 ≡ − = 4πG = (1.112) H2 H2 16πG V 1 V 00 η ≡ (1.113) 8πG V
quantifies the change in the Hubble rate during inflation. Inflation ends when ≥ 1.
1.9.4 Models of Inflation
An enormous number of models have been proposed to satisfy this condition in in-
genious ways, using one or more scalar fields (see [53] for a comprehensive review of
inflationary models). The models usually specify the potentials of the scalar field(s)
responsible for inflation. A typical potential for single-field inflation can be charac-
terized [10] by two parameters A and B and a function f as follows:
φ V (φ) = A4f (1.114) B
where A measures the vacuum energy density during inflation and B measures the
change in the field value during inflation. The function f specifies the model.
Based on the above characterization, inflationary models can be (very broadly)
classified into three categories [54]:
1. Large Field Models
2. Small Field Models
67 3. Hybrid Models
Large Field Models
These are models in which the scalar field is displaced very far away from the minimum
of its potential, typically to values of several times the Planck mass. The field is
assumed to be in this state as a result of having emerged from a quantum-gravity
state where the energy density was on the order of the Planck energy. The potential
energy of the field is therefore very high V (φ) ∼ mPl, leading to a high value of the Hubble parameter, and the consequent heavy Hubble damping causes the field to roll
slowly. Inflation ends when the field starts rolling fast, and this can be shown to
occur when the field is of the order of the Planck mass. This kind of inflation is called
chaotic inflation [55]. Examples of large field models are polynomial potentials
V (φ) = A4 (φ/B)p and exponential potentials V (φ) = A4 exp (φ/B).
Small Field Models
This is a class of models in which the field is initially near the origin and is evolving away from an unstable equilibrium at the origin toward a nonzero vacuum expectation value of hφi = 0. Such potentials arise naturally from spontaneous symmetry break- ing, such as the class of models referred to as “new” inflation ([56, 57]) and “natural” inflation [58]. Inflation ends when the field reaches the minimum and starts oscil- lating. A generic form of potentials of this kind are V (φ) = A4 [1 − (φ/B)p], which
can be regarded as a lowest order Taylor expansion (since the field is small) of an
arbitrary potential about the origin.
68 Hybrid Models
These models typically involve more than one scalar field. One field responsible for
inflation rolls towards a non-zero vacuum expectation value. An instability in the
other field ends the inflationary stage. A typical potential for the field responsible
for inflation is V (φ) = A4 [1 + (φ/B)p]. The presence of the second field introduces another free parameter in the model. Examples of hybrid models can be found in
[59, 60, 61].
Other Models
While the above classification is useful in comparing the predictions of different models with CMB observations, it is certainly not exhaustive. There are plenty of models which do not neatly fit into the above scheme, such as models involving brane inflation
[62], which does not require a scalar field at all. Other examples are the logarithmic potentials V (φ) = ln (φ) typically used in supersymmetric inflation (see [63] for a review), or the potentials of the type V (φ) = φ−p used, for instance, in intermediate
inflation [64].
1.9.5 Issues with Inflation
While inflation is a very elegant solution to a number of problems of the standard
cosmology, it has some problems of its own, some of which I list below. An extensive
discussion can be found in [65].
1. First of all, being a class of models rather than a model, the inflationary
paradigm allows for considerable freedom in choosing the inflaton potential,
69 the parameters of which can be tuned to match any observations. This leaves
unresolved several questions such as the start of inflation itself - why the inflaton
started out at the top of the hill.
2. The “solution” to the flatness problem described above is not entirely satisfac-
tory. Open inflationary models [66] are designed to make the Universe end up
positively curved.
3. As I show in Chapter 3, the power spectrum generated from inflation depends
on the slow roll parameters, and is hence easily tunable. By varying these
parameters, one can make inflationary models agree with scale-invariant, or
either slightly red or blue-shifted spectra. Also, for scalar perturbations, there
is no minimum predicted amplitude - their amplitude depends entirely on the
model.
4. Trodden and Vachaspati [52] have shown that inflationary models based on cer-
tain mild conditions (the classical Einstein equations, the null energy conditions,
and trivial topology), require the initial inflationary patch to be homogenous.
Hence inflation does not quite resolve the homogeneity problem by itself.
5. Finally, there is no evidence of the existence of a field suitable to be an inflaton.
As a result of all the above, it is useful to investigate cosmological models which do not fall within the inflationary paradigm. In the next two chapters I describe a cosmological model which relies on a different hypothesis, and examine its viability.
70 Works Cited
[1] E.W. Kolb and M.S. Turner. The Early Universe. Addison Wesley, 1990.
[2] Dominik J. Schwarz et al. Is the low-l microwave background cosmic? Phys. Rev. Lett., 93:221301, 2004. astro-ph/0403353.
[3] W.M. Yao et al. Review of Particle Physics. Journal of Physics G, 33:1+, 2006.
[4] John C. Mather et al. Measurement of the Cosmic Microwave Background spec- trum by the COBE FIRAS instrument. Astrophys. J., 420:439–444, 1994.
[5] B. Ryden. Introduction to Cosmology. Addison Wesley, 1995.
[6] S. Weinberg. Gravitation and Cosmology: Principles and Applications of the General Theory of Relativity. Wiley, 1972.
[7] Mark Trodden and Sean M. Carroll. TASI lectures: Introduction to cosmology. 2004. astro-ph/0401547.
[8] Wayne Hu and Scott Dodelson. Cosmic microwave background anisotropies. Ann. Rev. Astron. Astrophys., 40:171–216, 2002. astro-ph/0110414.
[9] D.H. Lyth and A.R. Liddle. Cosmological Inflation and Large Scale Structure. Cambridge University Press, 2000.
[10] Antonio Riotto. Inflation and the theory of cosmological perturbations. 2002. hep-ph/0210162.
[11] David Langlois. Inflation, quantum fluctuations and cosmological perturbations. 2004. hep-th/0405053.
[12] Robert H. Brandenberger. Lectures on the theory of cosmological perturbations. Lect. Notes Phys., 646:127–167, 2004. hep-th/0306071.
[13] S. Perlmutter and B.P. Schmidt. Measuring Cosmology with Supernovae. In K. Weiler, editor, Supernovae and Gamma-Ray Bursters, volume 598 of Lec- ture Notes in Physics, Berlin Springer Verlag, pages 195–217, 2003. astro- ph/0303428.
[14] Keith A. Olive. Dark matter. 2003. astro-ph/0301505.
[15] S.M Carroll. Spacetime and Geometry: An Introduction to General Relativity. Addison Wesley, 2004.
71 [16] E. Hog et al. The Tycho-2 catalogue of the 2.5 million brightest stars. Astron. Astrophys., 355:L27–L30, 2000.
[17] W. de Sitter. On the relativity of inertia. Remarks concerning Einsteins latest hypothesis. Proceedings of the Royal Academy of Sciences Amsterdam, 19:1217– 1225, 1916.
[18] W. de Sitter. On the Curvature of Space. Koninklijke Akademie van Weten- schappen, 1917.
[19] S.W. Hawking and G.F.R. Ellis. The Large Scale Structure of Space-Time. Cam- bridge University Press.
[20] Christian L. Thomas et al. Galactic bulge microlensing events from the macho collaboration. Astrophys. J., 631:906–934, 2005. astro-ph/0410341.
[21] N.A. Bahcall. Cosmology with clusters of galaxies. Nuclear Physics B Proceedings Supplements, 138:16–20, 2005.
[22] Neta A. Bahcall et al. Where is the dark matter? Astrophys. J., 447:L81, 1995. astro-ph/9506041.
[23] Ian Smail et al. Gravitational lensing of distant field galaxies by rich clusters: Ii. – cluster mass distributions. Mon. Not. Roy. Astron. Soc., 273:277–294, 1995. astro-ph/9402049.
[24] Aaron D. Lewis et al. X-ray mass estimates at z ∼ 0.3 for the cnoc cluster sample. Astrophys. J., 517:587–608, 1999. astro-ph/9901062.
[25] Simon D. M. White et al. The baryon content of galaxy clusters: A challenge to cosmological orthodoxy. Nature, 366:429–433, 1993.
[26] Joseph J. Mohr et al. Properties of the intracluster medium in an ensemble of nearby galaxy clusters. Astrophys. J., 517:627, 1999. astro-ph/9901281.
[27] J. E. Carlstrom et al. Imaging the Sunyaev-Zel’dovich Effect. 1999. astro- ph/9905255.
[28] Neta A. Bahcall et al. Constraining Ω with Cluster Evolution. Astrophys. J., 485:L53, 1997. astro-ph/9706018.
[29] R. G. Carlberg et al. Redshift evolution of galaxy cluster densities. 1996. astro- ph/9612169.
72 [30] Neta A. Bahcall and Xiao-hui Fan. The Most Massive Distant Clusters: Deter- mining Ω and σ. Astrophys. J., 504:1, 1998. astro-ph/9803277.
[31] Licia Verde et al. The 2dF Galaxy Redshift Survey: The bias of galaxies and the density of the Universe. Mon. Not. Roy. Astron. Soc., 335:432, 2002. astro- ph/0112161.
[32] F. Zwicky. On the Masses of Nebulae and of Clusters of Nebulae. Astrophys. J., 86:217, 1937.
[33] Douglas Clowe et al. Catching a bullet: direct evidence for the existence of dark matter. 2006. astro-ph/0611496.
[34] J. W. Moffat. Gravitational lensing in modified gravity and the lensing of merging clusters without dark matter. 2006. astro-ph/0608675.
[35] Garry W. Angus, Benoit Famaey, and HongSheng Zhao. Can mond take a bullet? analytical comparisons of three versions of mond beyond spherical symmetry. Mon. Not. Roy. Astron. Soc., 371:138, 2006. astro-ph/0606216.
[36] S. Chandrasekhar. The highly collapsed configurations of a stellar mass. Mon. Not. Roy. Astron. Soc., 91:456–466, 1931.
[37] M.M. Phillips. The absolute magnitudes of Type IA supernovae. Astro- phys.J.Lett, 413:L105–L108, 1993.
[38] Mario Hamuy et al. The absolute luminosities of the calan/tololo type ia super- novae. Astron. J., 112:2391, 1996. astro-ph/9609059.
[39] Adam G. Riess et al. A Precise distance indicator: Type Ia supernova multicolor light curve shapes. Astrophys. J., 473:88, 1996. astro-ph/9604143.
[40] S. Perlmutter et al. Measurements of the Cosmological Parameters Ω and Λ from the First Seven Supernovae at Z >= 0.35. Astrophys. J., 483:565, 1997. astro-ph/9608192.
[41] S. Perlmutter et al. Measurements of Ω and Λ from 42 High-Redshift Supernovae. Astrophys. J., 517:565–586, 1999. astro-ph/9812133.
[42] Lifan Wang et al. Multi-Color Light Curves of Type Ia Supernovae on the Color- Magnitude Diagram: a Novel Step Toward More Precise Distance and Extinction Estimates. Astrophys. J., 590:944–970, 2003. astro-ph/0302341.
73 [43] Adam G. Riess et al. Observational Evidence from Supernovae for an Acceler- ating Universe and a Cosmological Constant. Astron. J., 116:1009, 1998. astro- ph/9805201. [44] G. Hinshaw et al. Three-year Wilkinson Microwave Anisotropy Probe (WMAP) observations: Temperature analysis. Astrophys. J. Supp., 170:288–334, 2007. astro-ph/0603451. [45] R. K. Sachs and A. M. Wolfe. Perturbations of a cosmological model and angular variations of the microwave background. Astrophys. J., 147:73–90, 1967. [46] Joseph Silk. Cosmic black body radiation and galaxy formation. Astrophys. J., 151:459–471, 1968. [47] Wayne Hu et al. The Physics of microwave background anisotropies. 1995. astro-ph/9504057. [48] D. N. Spergel et al. Wilkinson Microwave Anisotropy Probe (WMAP) three year results: Implications for cosmology. 2006. astro-ph/0603449. [49] R H Dicke and P J E Peebles. General Relativity: An Einstein Centenary Survey, 1979. [50] John Preskill. Cosmological production of superheavy magnetic monopoles. Phys. Rev. Lett., 43:1365, 1979. [51] Alan H. Guth. The Inflationary Universe: A Possible Solution to the Horizon and Flatness Problems. Phys. Rev., D23:347–356, 1981. [52] T. Vachaspati and M. Trodden. Causality and cosmic inflation. Phys. Rev., D61:023502, 2000. gr-qc/9811037. [53] David H. Lyth and Antonio Riotto. Particle physics models of inflation and the cosmological density perturbation. Phys. Rept., 314:1–146, 1999. hep- ph/9807278. [54] Scott Dodelson, William H. Kinney, and Edward W. Kolb. Cosmic microwave background measurements can discriminate among inflation models. Phys. Rev., D56:3207–3215, 1997. astro-ph/9702166. [55] Andrei D. Linde. Chaotic inflation. Phys. Lett., B129:177–181, 1983. [56] Andreas Albrecht and Paul J. Steinhardt. Cosmology for grand unified theories with radiatively induced symmetry breaking. Phys. Rev. Lett., 48:1220–1223, 1982.
74 [57] Andrei D. Linde. A new inflationary universe scenario: A possible solution of the horizon, flatness, homogeneity, isotropy and primordial monopole problems. Phys. Lett., B108:389–393, 1982.
[58] Katherine Freese, Joshua A. Frieman, and Angela V. Olinto. Natural inflation with pseudo - nambu-goldstone bosons. Phys. Rev. Lett., 65:3233–3236, 1990.
[59] Andrei D. Linde. Axions in inflationary cosmology. Phys. Lett., B259:38–47, 1991.
[60] Andrei D. Linde. Hybrid inflation. Phys. Rev., D49:748–754, 1994. astro- ph/9307002.
[61] Edmund J. Copeland, Andrew R. Liddle, David H. Lyth, Ewan D. Stewart, and David Wands. False vacuum inflation with einstein gravity. Phys. Rev., D49:6410–6433, 1994. astro-ph/9401011.
[62] G. R. Dvali and S. H. Henry Tye. Brane inflation. Phys. Lett., B450:72–82, 1999. hep-ph/9812483.
[63] Andrei D. Linde. Particle physics and inflationary cosmology. 2005. hep- th/0503203.
[64] John D. Barrow and Andrew R. Liddle. Perturbation spectra from intermediate inflation. Phys. Rev., D47:5219–5223, 1993. astro-ph/9303011.
[65] Arthur Lue, Glenn D. Starkman, and Tanmay Vachaspati. A post-wmap per- spective on inflation. 2003. astro-ph/0303268.
[66] Martin Bucher, Alfred S. Goldhaber, and Neil Turok. An open universe from inflation. Phys. Rev., D52:3314–3337, 1995. hep-ph/9411206.
75 Chapter 2
Island Cosmology
2.1 Introduction
As I discussed in the previous chapter, the theory of inflation, together with the standard Big Bang Cosmology, is the most widely accepted description of the origin of our cosmos. At the core of the inflationary paradigm lies the hypothesis that an inflaton exists in Nature.
In this chapter, I describe an alternative cosmological model which I have proposed together with Tanmay Vachaspati called “Island Cosmology” [1]. Island Cosmology relies on a very different hypothesis than inflation.
In a nutshell, Island Cosmology works as follows: the universe is initially assumed to be filled with cosmological constant of the currently observed value but is otherwise empty. In this eternal or semi-eternal de Sitter spacetime, local quantum fluctuations violate the null energy condition (NEC) and create islands of matter, one of which is our Universe. With further cosmic evolution the island disappears and the local
76 spacetime returns to its initial cosmological-constant dominated state.
2.2 The Model
I begin with a short overview of Island Cosmology, describing each of its different stages.
1. In the beginning the universe is inflating due to the observed dark energy that
1 −1 we assume is a cosmological constant (Λ) . The de Sitter horizon size (HΛ ) is
−1 comparable to our present horizon, (H0 ). We call this spacetime the Λ-sea.
2. A quantum fluctuation of some field (e.g. scalar field, photon) in a horizon-size
volume (which we label as an I-region) in the expanding phase of de Sitter
spacetime drives the Hubble constant to a large value. Even as the Hubble
length scale is decreasing the universe continues to expand. From the accel-
eration equation (1.37), we see that this can occur only when (p + ρ) < 0, or
µν NµNνT < 0, or in other words, the Null Energy Condition (NEC) is vio- lated. Later I will show, following Refs. [2, 3, 4], that quantum field theoretic
fluctuations allow for this possibility.
3. After the NEC violating fluctuation (which is assumed to be of a very short
temporal duration - see § 2.4) is over, the Hubble constant within the I-region is
large, and the I-region gets filled with classical radiation, as the NEC violating
field decays into relativistic particles. Thereafter, the I-region evolves as a
1We make no attempt to address why Λ is so small compared to the Planck scale.
77 Figure 2.1: Sketch of the behavior of the Hubble length scale with conformal time, η, in the Island model, and the evolution of fluctuation modes. At early times, inflation is driven by the presently observed dark energy, assumed to be a cosmological constant. As the cosmological constant is very small, the Hubble length scale is very large – of order the present horizon size. Exponential inflation in some horizon volume ends not due to the decay of the vacuum energy as in inflationary scenarios but due to a quantum fluctuation in the time interval (ηi, ηf ) that violates the null energy condition (NEC). The NEC violating quantum fluctuation causes the Hubble length scale to decrease. After the fluctuation is over, the universe enters radiation dominated FRW expansion, and the Hubble length scale grows with time. The physical wavelength −1 of a quantum fluctuation mode starts out less than HΛ at some early time ηi. The mode exits the cosmological horizon during the NEC violating fluctuation (ηexit) and then re-enters the horizon at some later epoch (ηentry) during the FRW epoch (The modes are drawn as straight lines for illustrative purposes only, they actually grow in proportion to the scale factor).
78 radiation-dominated FRW Universe, as described in § 1.5.7, and eventually
forms our observed Universe. We call a Universe created in this manner an
Island Universe.
4. With further evolution, the Island Universe dilutes and eventually the I-region
is again dominated by the cosmological constant and the spacetime returns to
its normal inflating state.
Island Cosmology does include elements of earlier work such as eternal inflation models, steady state models and ekpyrotic models. While eternal inflation models [5,
6], especially Garriga and Vilenkin’s “recycling universe” [7], (see also the discussion in [8]) use NEC violating quantum fluctuations in the inflaton field to drive the Hubble length scale to smaller values, in Island cosmology, these quantum fluctuations can occur in any quantum field and have to be large. In both inflationary cosmology and our case, the quantum fluctuation needs to violate the NEC. Furthermore, in both cases the back-reaction of the fluctuation is assumed to lead to a faster rate of cosmological expansion. In the language of [9], the evolution we are considering is one of the “miraculous” trajectories that go directly from a dead de Sitter region of spacetime to a region that is “macroscopically indistinguishable from our universe”
(MIFOU). Eventually the trajectory leaves the MIFOU region and returns to the dead de Sitter region. In Steady State Cosmology [10, 11], matter is sporadically produced by “minibangs” in a hypothetical C-field. The explosive events in Island
Cosmology, on the other hand, are quantum field theoretic in origin and seed the matter content of an entire Universe. The decreasing Hubble scale is also a feature of the ekpyrotic cosmological model [12]. However, in that model, the motivation
79 for the decrease lies in extra-dimensional brane-world physics and results in a period
of contraction of our three dimensional universe. Island Cosmology does not involve
any brane-world physics, and has no contracting phase, as the Universe continues to
expand even while the Hubble scale drops.
Having summarized the main idea, I now discuss each step of this model in greater
detail.
2.3 NEC violations in de Sitter space
In the first stage of the model, we assume that the Universe is filled with cosmological
constant. As I have explained in the previous chapter, observations are consistent
with some form of dark energy, the simplest explanation of which is a cosmological
constant. The model does not necessarily have to start out with a singularity, and
nor does spacetime need to be created out of nothing as in quantum cosmology. All
that we need is an expanding de Sitter background, and this could be the expanding
phase of a classical de Sitter spacetime with no beginning and no end. The scale
factor of the universe at this stage is given by:
1 HΛt a(t) = a0e ≡ − (2.1) HΛη
where η ∈ (−∞, 0) is the conformal time.
In de Sitter spacetime, as well as any other spacetime, there are fluctuations of
the energy-momentum tensor, Tµν, of quantum fields. This is simply a consequence of the fact that the vacuum, |0i, is an eigenstate of the Hamiltonian but not of the
80 ˆ energy-momentum density operator, Tµν. In short-hand notation:
ˆ X † † † Tµν|0i = [(... )alak + (... )al ak] |0i X = [(... )|0i + (... )|2; k, li] (2.2)
where, the ellipses within parenthesis denote various combinations of mode functions
† and their derivatives; ak, al are creation and annihilation operators and |2; k, li is a two particle state. The final expression is not proportional to |0i, implying that ˆ the vacuum is not an eigenstate of Tµν and there will be fluctuations of the energy- momentum tensor in de Sitter space.
It has been shown [2, 3, 4] that quantum field theory of a light scalar field in the
Bunch-Davies vacuum [13] in de Sitter space leads to violations of the NEC. I now
briefly summarize the general arguments behind this conclusion.
The first step is to construct a “smeared NEC operator”
Z √ ˆren 4 µ ν ˆren OW ≡ d x −g W (x; R,T ) N N Tµν (2.3)
where W (x; R,T ) is a smearing function on a length scale R and time scale T . The vector N µ is chosen to be null, and the superscript ren denotes that the operator has been suitably renormalized. As shown in [2] the smeared operator will be not be proportional to the vacuum state either, and will fluctuate. The scale of the
fluctuations can be estimated on dimensional grounds:
2 ˆren 2 8 Orms ≡ h0|(OW ) |0i ∼ HΛ (2.4)
81 −1 ˆ in the special case when R = T = HΛ . Since, in de Sittter space, h0|Tµν|0i ∝ gµν, we also have:
ˆren h0|OW |0i = 0 (2.5)
ˆren Therefore the fluctuations of OW are both positive and negative. Assuming a sym- metric distribution, we come to the conclusion that quantum fluctuations of a scalar
field violate the NEC with 50% probability. Exactly the same arguments can be applied to quantum fluctuations of a massless gauge field such as the photon.
Note that the above calculation does not give us the probability distribution of the violation amplitude, for which we would have to calculate the actual probabil-
ˆren ity distribution for the operator OW . However, by continuity we can expect that large amplitude NEC violations will also occur with some diminished but non-zero probability.
2.4 Extent and duration of NEC violation
What is the spatial and temporal extent of these NEC-violating fluctuations? Such
fluctuations can occur on all spatial and temporal scales, but based on causality and predictability, I now argue that only fluctuations of a large (horizon-sized) spatial extent and small temporal duration are relevant to creating islands of matter. Smaller
fluctuations are irrelevant because the spacetime is likely to respond only locally before returning to its original state.
Consider the spacetime diagram of Fig. 2.2. In that diagram we show an initial de Sitter space that later has a patch in which the space is again de Sitter though
82 + η
η b b η Q Q ? ? η ? ? ? P P η a a
r
Figure 2.2: We show a classical de Sitter spacetime for conformal time η < ηP , that transitions to a faster expanding classical de Sitter spacetime for η > ηQ. The inverse Hubble size is shown by the white region. A bundle of ingoing null rays originating at point a is convergent initially but becomes divergent in the superhorizon region at point b. This can only occur if the NEC is violated in the region η ∈ (ηP , ηQ). In the quantum domain, a classical picture of spacetime may not be valid and this is made explicit by the question marks.
83 −1 with a larger expansion rate. Hence the initial Hubble length scale Hi is larger than
−1 the final Hubble length scale Hf . Therefore there are ingoing null rays that are within the horizon initially that propagate and are eventually outside the horizon.
An example of such a null ray is the line from a to b. At point a a bundle of such rays will be converging whereas at point b the bundle will be diverging. It can be demonstrated from the Raychaudhuri equation (provided some mild conditions are satisfied, such as general relativity being valid, and spacetime topology being trivial) that the transition from convergence to divergence of a bundle of null rays can only occur if there is a NEC violation somewhere along the null ray (see [14]).
−1 Now if the NEC violation only occurred on a scale smaller than Hi , one could imagine a null ray that would never enter the NEC violating region and yet go from being converging to diverging (see Fig. 2.3). This would clearly be inconsistent with the Raychaudhuri equation.
Furthermore, after the energy condition violations are over, the faster expanding region would have to either instantly revert to the ambient expansion rate, or some spacetime feature, such as a singularity, would have to occur to prevent a null ray from entering the faster-expanding region from the slower-expanding region. Addi- tional boundary conditions would have to be imposed on the singularity to restore predictability. An example of such a process can be found in Ref. [15] in connection with topological inflation [16, 17].
Another way of understanding the loss of predictability is the following. Whenever a faster expanding universe is created, it must be connected by a wormhole to the ambient slower expanding region. The wormhole can be kept open if the energy
84 η
η b b Q η Q ? ? η ? P P η a a
r
Figure 2.3: A spacetime diagram similar to that in Fig. 2.2 but one in which the NEC violation occurs over a sub-horizon region (shaded region in the diagram). Now the null ray bundle from a to b goes from being converging (within the horizon) to diverging (outside the horizon). However, it does not encounter any NEC violation along its path, and this is not possible as can be seen from the Raychaudhuri equation. Since the ingoing null rays are convergent as far out as the point P , the size of the quantum domain has to extend out to at least the inverse Hubble size of the initial de Sitter space. Therefore the NEC violating patch has to extend beyond the initial horizon.
85 conditions are violated [18]. But, if the wormhole neck is small, as soon as the energy condition violations are over, it must collapse and pinch off into a singularity. Signals from the singularity can propagate into the faster expanding universe destroying predictability. However, if the neck of the wormhole is larger than the horizon size of the ambient universe, the ambient expansion can hold up the wormhole and the neck does not collapse even after the NEC violation is over.
Our argument that NEC violations on scales larger than the horizon are needed to produce a faster expanding universe is consistent with earlier work [19] showing that it is not possible to produce a universe in a laboratory without an initial singularity
(also see [20]). Subsequent discussion of this problem in the quantum context [21,
22, 23], however, showed that a universe may tunnel from nothing without an initial singularity, just as in quantum cosmology [24, 25]. Such a tunneling event, however, is irrelevant to Island Cosmology, as the newly created universe is causally disconnected from the ambient Λ-sea. Without an inflaton, the process would therefore produce only a second Λ-sea.
Based on the above arguments, and on the results of the earlier investigations cited, we conclude that to get a faster expanding region that lasts beyond the duration of the quantum fluctuation and remains predictable, the spatial extent of the NEC
−1 violating fluctuation must be larger than Hi :
−1 R > Hi (2.6) where R is the spatial extent of the fluctuation and shows up as the spatial smearing ˆ scale in the calculation of Orms.
86 We also argue that the temporal scale of the fluctuation has to be small. This is ˆ because, an explicit evaluation [2] shows that Orms is proportional to inverse powers of the temporal smearing scale and diverges as the smearing time scale T → 0. Hence the briefer the fluctuation, the stronger it can be, as we might also expect from an application of the Heisenberg time-energy uncertainty relation. Therefore we take the time scale of the NEC violation to be vanishingly small:
T → 0 (2.7)
2.5 Likelihood – the role of the observer
In the preceding sections I have described the nature of the fluctuation. I now turn to the question of how likely are fluctuations of the kind described.
In Sec. 2.4 I have pointed out that the NEC-violating fluctuations need to have two requirements to be cosmologically relevant - they need to have a superhorizon spatial extent and must be of vanishningly small duration. There is one more require- ment that is absolutely crucial - the fluctuations must have the correct amplitude in order to have sufficient energy density to lead to our observed Universe. Clearly, only if the temperature produced is high enough and the end point of the NEC violat- ing fluctuation is a thermal state with all the different forms of matter in thermal equilibrium, further evolution of the island will simply follow the standard big bang cosmology.
Admittedly the three requirements of large spatial extent, small temporal extent and large amplitude make these fluctuations rare. However, since spacetime is eternal
87 in this model we can wait indefinitely for such a fluctuation to occur.
The probability of fluctuations in the Λ-sea that can lead to an inflating cos- mology versus those that produce an FRW universe have been considered by several researchers [9, 26]. In particular, Dyson et al. [9] estimate probabilities based on a
“causal patch” picture, which assumes that the physics beyond the de Sitter horizon is irrelevant to the physics within, and that the latter should be regarded as the com- plete physics of the Universe. Based on this picture, the authors of [9] conclude that it is much more probable to directly create a universe like ours than to arrive at our present state via inflation. Albrecht and Sorbo [26] have argued that the conclusion rests crucially on the causal patch picture, and provide a different calculation leading to the conclusion that inflationary cosmology is favored. Both the above calculations assume the existence of fields that are suitable for inflation. However, Island Cos- mology does not rely on the hypothesis of the inflaton, and so the comparison of the likelihood of inflation versus no inflation is moot.
The monopole overabundance problem can be resolved in Island Cosmology in a manner similar to that proposed in Ref. [27], by assuming that the temperature required for magnetic monopoles production is higher than that required for matter- genesis. If the temperature at the beginning of the FRW phase is below that needed for monopole formation but above the matter-genesis temperature then there will be no cosmological magnetic monopole problem.
Another important question is where we are located on the island. Are we close to the edge of the island (“beach”)? In that case we would observe anisotropies in the
CMB since in some directions we would see the Λ-sea while in others we would see
88 inland. However, the island is very large (by a factor a0/af ) compared to our present
−1 horizon, HΛ . If we assume a uniform probability for our location on the island, our
−1 distance from the Λ-sea will be an O(1) fraction of HΛ a0/af . Since a0/af is of order
Tmg/T0 – the ratio of the matter-genesis temperature to the present temperature – we are most likely to be sufficiently inland so as not to observe any anisotropy in the
CMB.
Whereas inflationary models crucially rely on the existence of a suitable scalar
field (inflaton), I have so far not specified the quantum field that causes the NEC violating fluctuation. I now turn to this issue.
2.6 The NEC violating field
The phantom energy that is assumed to describe the effects of the NEC violating quantum fluctuation, by definition, satisfies ρ + p < 0. In addition, the assumption that the backreaction is given by Eq. (3.38), requires ρ > 0. Hence we need a quantum field that can give NEC violating fluctuations while still having positive energy density. In other words, the energy density should be positive but the pressure should be sufficiently negative so that the NEC is violated.
First consider a scalar field, φ, with potential V (φ). The energy density and pressure are:
1 1 ρˆ = φ˙2 + (∇φ)2 + V (φ) 2 2 1 1 pˆ = φ˙2 − (∇φ)2 − V (φ) (2.8) 2 6
89 where the hats on ρ and p emphasize that these are quantum operators Therefore:
(∇φ)2 ρˆ +p ˆ = φ˙2 + (2.9) 3
The operatorsρ ˆ andρ ˆ +p ˆ are not proportional to each other and fluctuations in one do not have to be correlated with fluctuations of the other. The energy density in a region can be positive while the NEC is violated. Therefore a scalar field, even if V (φ) = 0, can provide suitable NEC violating fluctuations.
The particle physics in the very early stages of the model is described by low energy particle physics that we know so well. At present we do not have any experimental evidence for a scalar field. One field that we know of today is the electromagnetic field.
Could the electromagnetic field give rise to a suitable NEC violating fluctuation?
For the electromagnetic field we have:
1 ρˆ = (E2 + B2) 2 1 1 pˆ = (E2 + B2) = ρˆ (2.10) 6 3
So nowρ ˆ andp ˆ are not independent operators and
4 ρˆ +p ˆ = ρˆ (2.11) 3
From this relationship between the operators, it is clear that the only electromagnetic
fluctuation that can violate the NEC also has negative energy density. This means that even though the electromagnetic field can violate the NEC, it does not satisfy
90 the positive energy density condition needed in the working hypothesis to find the backreaction. (Our working hypothesis for the backreaction is described in § 3.3.2).
It may be possible that the electromagnetic field will still be found to be suitable once we know better how to handle the backreaction problem. Then perhaps we will not need to rely on the working hypothesis that requires positive energy density.
There is a possible loophole in our discussion of the electromagnetic field. The equation of statep ˆ =ρ/ ˆ 3 follows from the conformal invariance of the electromagnetic
ˆµ field Tµ = 0. However, we know that quantum effects in curved spacetime give rise to ˆµ a conformal anomaly and the trace hTµ i is not precisely zero. So we can expect that the equation of statep ˆ =ρ/ ˆ 3 is also anomalous. Whether this anomaly can allow for
NEC violations with positive energy density is not clear to us.
Note that it is not necessary for the NEC violation to originate from a fluctuation of a massless or light field. The arguments of Sec. 2.3 are very general and apply to ˆ massive fields as well. Though, for a massive field, Orms will be further suppressed by exponential factors whose exponent depends on powers of HΛ/m. While the likelihood of a suitable NEC violating fluctuation from a very massive field is much smaller compared to that of a light or massless field, the massive field fluctuations are clearly more important if the light field doesn’t even exist! The discussion in the previous section of the likelihood still applies.
2.7 Assumptions
Island cosmology involves several assumptions that I have pointed out above but now summarize and discuss.
91 Our first assumption is that the dark energy is a cosmological constant. This is consistent with observations and moreover is the simplest explanation of the Hubble acceleration. We assume that the cosmological constant provides us with a back- ground de Sitter spacetime that is eternal2. As de Sitter spacetime also has a con- tracting phase, the singularity theorems of Ref. [29] are evaded.
The second assumption is that there is a scalar field in the model responsible for the NEC violation. It would have been more satisfactory if the electromagnetic field could have played this role but we have shown (up to the loophole of the conformal anomaly) that the conformal invariance of the electromagnetic field prevents NEC violations with positive energy density. It is possible that with a better understanding of the backreaction of quantum energy-momentum fluctuations on the spacetime, the electromagnetic field might still provide suitable NEC violations (see Sec. 3.3.2).
The basic formalism of quantum field theory in curved spacetime clearly leads to
NEC violations and this is not an assumption. (Though one could reasonably question the applicability of quantum field theory on systems with horizons.) Then there seems little doubt that there should exist large amplitude NEC violations, though occurring much more infrequently than the small amplitude violations. The idea that NEC violating fluctuations could have played an important cosmological role is also to be found in the “eternal inflation” scenario [30]. Indeed, the current scenario may also be viewed as an eternal inflation scenario – since the universe is eternally inflating due to a cosmological constant! While we may not be able to test the idea of cosmological
NEC violating fluctuations, we can certainly test quantum fluctuations with and
2For a discussion of the timescale on which the spacetime can remain de Sitter, see Ref. [28].
92 without horizons in laboratory experiments [31, 32, 33, 34].
The third assumption we have made has to do with NEC violations in regions of small spatial extent. Based on work done on the possibility of creating a universe in a laboratory, topological inflation, and wormholes, we have argued for the conjecture that small scale violations of NEC can only give rise to universes that are affected by signals originating at a singularity. Hence predictability is lost in such universes.
Our assumption is that even if we did know how to handle the spacetime singularities affecting these universes, they would turn out to be unsuitable for matter genesis.
Without this assumption, we should also be considering such universes as possible homes.
The fourth assumption is that the final state of the fluctuation is a thermal state.
All the different energy components are also assumed to be in thermal equilibrium.
We have then assumed that the critical temperature needed for observers to exist is the temperature at which matter-genesis occurs. One could relax this assumption but one would need an adequate characterization of the most likely state to be able to calculate cosmological observables (e.g.spectrum of density fluctuations).
This brings us to the part of the model where we argue that even if the large amplitude fluctuations are infrequent, they are the only ones that are relevant for observational cosmology. This is quite similar to the arguments given in the context of eternal inflationary cosmology where thermalized regions are relatively rare but these are the only habitable ones. It also occurs in chaotic inflation [35], where closed universes of all sizes and shapes are produced but only a few are large and homogeneous enough to develop into the present universe. So this part of Island
93 Cosmology is no weaker (and harder to quantify) than other cosmological models.
2.8 Conclusions
To conclude, we have investigated a new cosmological model, which we call “island cosmology”, where large NEC violating quantum fluctuations (“upheavals”) in a cos- mological constant de Sitter universe create islands of matter. In island cosmology, spacetime may be non-singular and eternal3. and an inflationary stage is not neces- sary.
Island cosmology is attractive because it is a minimalistic model. It uses currently observed features of the Universe as its ingredients and combines them with well- established results from quantum field theory to account for the Universe that we live in now. However, the crucial test for any cosmological model comes from the spectrum of density perturbations that the model predicts. The next chapter is devoted to exploring this question in detail.
3The essential point is to have an expanding de Sitter phase; whether this is part of an eternal de Sitter spacetime or originates at a big bang makes no difference.
94 Works Cited
[1] Sourish Dutta and Tanmay Vachaspati. Islands in the Lambda-sea. Phys. Rev., D71:083507, 2005. astro-ph/0501396.
[2] A. Guth et al. Energy Conditions in Eternal Inflation unpublished.
[3] Tanmay Vachaspati. Eternal inflation and energy conditions in de Sitter space- time. 2003. astro-ph/0305439.
[4] Serge Winitzki. Null energy condition violations in eternal inflation. 2001. gr- qc/0111109.
[5] A.D. Linde. Phys. Lett., B175:395, 1985.
[6] A. Vilenkin. Phys. Rev., D27:2848, 1983.
[7] Jaume Garriga and Alexander Vilenkin. Recycling universe. Phys. Rev., D57:2230–2244, 1998. astro-ph/9707292.
[8] Sean M. Carroll and Jennifer Chen. Spontaneous inflation and the origin of the arrow of time. 2004. hep-th/0410270.
[9] Lisa Dyson et al. Disturbing implications of a cosmological constant. JHEP, 10:011, 2002. hep-th/0208013.
[10] H. Bondi and T. Gold. The steady-state theory of the expanding universe. Mon. Not. Roy. Astron. Soc., 108:252, 1948.
[11] F. Hoyle. A new model for the expanding universe. Mon. Not. Roy. Astron. Soc., 108:372, 1948.
[12] J. Khoury et al. The ekpyrotic universe: Colliding branes and the origin of the hot big bang. Phys. Rev., D64:123522, 2001. hep-th/0103239.
[13] T.S Bunch and P.C.W. Davies. Quantum field theory in de sitter space: Renor- malization by point-splitting. Proc. Roy. Soc. Lond., 360:117, 1978.
[14] T. Vachaspati and M. Trodden. Causality and cosmic inflation. Phys. Rev., D61:023502, 2000. gr-qc/9811037.
[15] A. Borde et al. Creation and structure of baby universes in monopole collisions. Phys. Rev., D59:043513, 1999. gr-qc/9808069.
95 [16] Andrei D. Linde. Monopoles as big as a universe. Phys. Lett., B327:208–213, 1994. astro-ph/9402031.
[17] Alexander Vilenkin. Topological inflation. Phys. Rev. Lett., 72:3137–3140, 1994. hep-th/9402085.
[18] M. S. Morris and K. S. Thorne. Wormholes in space-time and their use for interstellar travel: A tool for teaching general relativity. Am. J. Phys., 56:395– 412, 1988.
[19] Edward Farhi and Alan H. Guth. An obstacle to creating a universe in the laboratory. Phys. Lett., B183:149, 1987.
[20] Andrei D. Linde. Hard art of the universe creation (stochastic approach to tunneling and baby universe formation). Nucl. Phys., B372:421–442, 1992. hep- th/9110037.
[21] E. Farhi et al. Is it possible to create a universe in the laboratory by quantum tunneling? Nucl. Phys., B339:417–490, 1990.
[22] W. Fischler et al. Quantization of False Vacuum Bubbles: A Hamiltonian Treat- ment of False Vacuum Bubbles. Phys. Rev., D42:4042–4055, 1990.
[23] Willy Fischler, Daniel Morgan, and Joseph Polchinski. Quantum nucleation of false vacuum bubbles. Phys. Rev., D41:2638, 1990.
[24] J. B. Hartle and S. W. Hawking. Wave Function of the Universe. Phys. Rev., D28:2960–2975, 1983.
[25] Alexander Vilenkin. Creation of Universes from Nothing. Phys. Lett., B117:25, 1982.
[26] Andreas Albrecht and Lorenzo Sorbo. Can the universe afford inflation? Phys. Rev., D70:063528, 2004. hep-th/0405270.
[27] G. R. Dvali, Alejandra Melfo, and Goran Senjanovic. Is there a monopole prob- lem? Phys. Rev. Lett., 75:4559–4562, 1995. hep-ph/9507230.
[28] Naureen Goheer, Matthew Kleban, and Leonard Susskind. The trouble with de sitter space. JHEP, 07:056, 2003. hep-th/0212209.
[29] Arvind Borde, Alan H. Guth, and Alexander Vilenkin. Inflationary space- times are incomplete in past directions. Phys. Rev. Lett., 90:151301, 2003. gr- qc/0110012.
96 [30] Arvind Borde and Alexander Vilenkin. Violation of the weak energy condition in inflating spacetimes. Phys. Rev., D56:717–723, 1997. gr-qc/9702019.
[31] Uwe R. Fischer and Ralf Schutzhold. Quantum simulation of cosmic inflation in two-component bose-einstein condensates. Phys. Rev., A70:063615, 2004. cond- mat/0406470.
[32] T. A. Jacobson and G. E. Volovik. Effective spacetime and hawking radiation from moving domain wall in thin film of he-3-a. Pisma Zh. Eksp. Teor. Fiz., 68:833–838, 1998. gr-qc/9811014.
[33] W. G. Unruh. Experimental black hole evaporation. Phys. Rev. Lett., 46:1351– 1353, 1981.
[34] Tanmay Vachaspati. Cosmic Problems for Condensed Matter Experiment. 2004. cond-mat/0404480.
[35] Andrei D. Linde. Chaotic inflation. Phys. Lett., B129:177–181, 1983.
97 Chapter 3
Perturbation Spectra
3.1 Introduction
This chapter explores the issue of determining the nature of perturbations generated by Island Cosmology. I first review the theory of cosmological perturbations, and demonstrate how the theory predicts a scale invariant spectrum for inflation. I then describe the challenges involved in computing the spectrum generated in Island Cos- mology. Finally I review my research [1] in calculating this spectrum based on some classical assumptions.
3.2 The Theory of Cosmological Perturbations
The goal of this theory is to find the relationship between matter and metric pertur- bations by expanding the Einstein equations to linear order about the background metric. The following discussion is based on the books and reviews ([2, 3, 4, 5, 6]). In
98 what follows, Greek letters denote spacetime indices, Roman numerals denote spatial
indices, and primes denote derivatives with respect to conformal time (except when
noted otherwise).
3.2.1 The metric perturbations
The most general linear perturbation in the spatially flat FRW metric (see § 1.2 for
an introduction to the FRW metric) can be written as
2 2 2 i i j ds = a (1 + 2A) dη − 2Bidx dη − (δij + hij) dx dx (3.1)
where i and j stand for spatial indices. The space and time perturbations shown above can be regarded as three-dimensional tensors whose indices can be raised or lowered by the spatial metric δij. Since general vector and tensor fields are often a mixture of “pure” scalar, vector and tensor modes (based on their transformations
under spatial rotations), it is instructive to decompose the perturbations in (3.1)
accordingly.
A spatial vector field Bi can be decomposed into a longitudinal part and a
transverse part, where the longitudinal part is curl-free and the transverse part is
divergence-free:
¯i ¯i Bi = ∇iB + B , ∇iB = 0 (3.2)
The longitudinal part is written as the divergence of a scalar B, and the transverse part B¯i has two vector degrees of freedom (the index i can take on three values, but one degree of freedom is lost from the second relation in (3.2).
99 Variable Scalar Vector Tensor
AA ¯ Bi B Bi (2) ¯ ¯ij hij C, E Ei (2) E (2)
Total 4 4 2 Table 3.1: Scalar-Vector-Tensor decomposition of metric perturbations
Similarly, any symmetric tensor can be built from two scalar, two vector and two
transverse and traceless tensors as follows:
¯ ¯ hij = 2Cδij + 2∇i∇jE + 2∇(iEj) + Eij (3.3)
¯ ¯ij ¯ij where the Eij are transverse ∇iE = 0 and traceless δijE tensors. The perturbations in (3.1) therefore break up into 4 scalar, 4 vector, and 2 tensor modes, as summarized in Table 3.2.1.
At linear order, the scalar, vector and tensor modes can be treated independently.
The tensor fluctuations do not couple (at linear order) to matter perturbations, and vector perturbations redshift away in an expanding background. As a result it is only the scalar perturbations that are most important in understanding the formation of structure.
100 3.2.2 Gauge Issues in Cosmology
One must be particularly cautious about gauge artefacts (spurious gauge modes
which do not correspond to physical degrees of freedom) in the theory of cosmological
perturbations because of the way perturbations are defined. A perturbation in a given
physical quantity (e.g. the Ricci scalar) at a given point is the difference between the value of the quantity in the real physical spacetime, and the value it assumes at the
same point in an unperturbed spacetime (see, for example [2]). Clearly, one needs a
map to transform between “corresponding” points on the two different spacetimes.
A particular choice of such a map is called a gauge choice. Fixing a gauge amounts to threading the spacetime into lines of fixed x, and slicing the spacetime into hypersurfaces of fixed t.
It is not hard to see that some of the metric perturbation degrees of freedom in table 3.2.1 are spurious gauge modes. This can be seen by studying the effect of coordinate transformations on the metric. Consider the coordinate transformation
xµ → x˜µ = xµ + ξµ (3.4)
Decomposing the spatial part of ξµ as in (3.2) into the gradient of a scalar ξ and a transverse piece ξi, it is clear that there are four physical degrees of freedom (2 scalar and 2 vector) associated with this transformation. Clearly, the tensor modes are not affected by this transformation, and are automatically gauge-invariant.
One can now apply this transformation to the metric (3.1), and (keeping in mind that the line-element is left invariant by this transformation), one can derive trans-
101 formation rules for the scalar perturbation variables
a0 A˜ = A − ξ0 − ξ00 a B˜ = B + ξ0 − ξ0 a0 1 C˜ = C + ξ0 − ∇2ξ a 3 E˜ = E − ξ (3.5)
Also from (3.4) it is easy to deduce that a perturbation in a scalar quantity f
(such as a field or an energy density) transforms as
δf˜ = δf − f 0ξ0 (3.6)
There are two standard approaches to dealing with the ambiguity of having more gauge modes than physical modes. One of them is to work with gauge invariant variables, which are combinations of perturbation variables, and which correspond to physical degrees of freedom and are hence unaffected by gauge transformations.
One set of gauge invariant variables introduced by Bardeen [7] is the following:
1 Φ = A + [(B − E0) a]0 (3.7) a a0 Ψ = C − (B − E0) (3.8) a
Φ and Ψ are commonly known as the Bardeen potentials.
A different approach is to fix a gauge, for which there is a variety of gauge choices
available [3]. A very common choice is the longitudinal or conformal Newtonian
102 gauge, in which B = E = 0. An advantage of longitudinal gauge is that the Bardeen potentials Φ and Ψ are numerically equal to the metric perturbations A and C in this gauge. The slicing and the threading are clearly orthogonal.
Another commonly used gauge is the synchronous gauge, which corresponds to the choice A = B = 0, so that the only non-zero perturbations are D and E. The advantage of this gauge is that the threading consists of geodesics, making this gauge a natural choice if one wishes to work in comoving co-ordinates. Synchronous gauge is a popular choice for numerical computations and is used in the CMBFAST1 code.
[8] provides a detailed review and comparison of both of these gauges.
3.2.3 The Comoving Curvature Perturbation
The curvature perturbation ψ is defined in terms of R(3), the intrinsic spatial curvature on hypersurfaces of constant conformal time η, for a flat Universe, through the relationship [3] 4 4 1 R(3) = ∇2ψ = ∇2 D + E (3.9) a2 a2 3
From the connection with the metric perturbation variables, it is clear that ψ is gauge dependent.
In a generic gauge, let the perturbation in some scalar quantity φ be δφ.
Now suppose we transform to a synchronous gauge, where the threading consists of geodesics and the slicing is comoving. Observers in these comoving coordinates will find the φ to be isotropic, i.e., δφcom = 0.
1http://cfa-www.harvard.edu/~mzaldarr/CMBFAST/cmbfast.html
103 But the value of the perturbation in the two gauges is linked via (3.6):
0 δφcom = δφ − φ δη (3.10) implying that the time step δη necessary to switch from a generic gauge to a comoving gauge is δη = δφ/φ0.
From the gauge transformation laws 3.5, we conclude that the change in ψ is as follows: δφ R ≡ ψ ≡ ψ + Hδη = ψ + H (3.11) com φ0
The above quantity, the curvature perturbation on comoving hypersurfaces R, or the comoving curvature perturbation is a crucial cosmological observable. It is particularly useful in inflationary cosmology as it is gauge invariant (by construction) and roughly constant on superhorizon scales [9], and can therefore be used to relate scales leaving the horizon during inflation to scales re-entering the horizon at last scattering [2].
In the classical treatment of Island Cosmology that follows (§ 3.6.2), the relation- ship between R and the Mukhanov variable (explained in § 3.2.6) is used to compute the latter.
3.2.4 The Power Spectrum
Consider a quantum field g(x, t) expanded in Fourier modes gk(t) as follows:
Z d3k g (x, t) = eik·xg (t) (3.12) (2π)3/2 k
104 The power spectrum of the field g(x, t), Pg is defined through the relationship
2π2 h0| g∗ g |0i = δ3 (k − k ) P (3.13) k1 k2 1 2 k3 g
where |0i denotes the vacuum state.
The variance of the field g(x, t) from equations (3.12) and (3.13) works out to
Z dk h0| g2(x, t) |0i = P (k) (3.14) k g
However, using Parseval’s theorem, the variance of g(x, t) can be written as
Z 3 2 d k 2 h0| g (x, t) |0i = |gk (t)| (3.15) (2π)3
Comparing (3.14) and (3.15), we get another very useful definition of the power
spectrum k3 P (k) = |g (t)|2 (3.16) g 2π2 k
The power spectrum measures the variance of the field or “power” per unit logarithmic
wavenumber interval. A scale-invariant spectrum Pg (k) = const is one in which the power per unit logarithmic wavenumber interval is independent of the wavenumber.
The departure of a spectrum from scale-invariance (or the spectral tilt) is mea- sured by the power spectral index ns which is usually defined by the relation
d ln P n − 1 = g (3.17) s d ln k
105 3.2.5 The Equations of Motion
To derive the linearized equations describing the evolution of the perturbations, one
can either work with gauge-invariant perturbation variables, or one can fix a gauge.
For simplicity, we choose the latter option and work in the longitudinal gauge (see
i § 3.2.2). We also assume that there is no anisotropic stress, e.g., δTj = 0, which at once implies, from the constraint equation, that the two perturbation variables are
equal: A = C ≡ Φ. The assumption of no anisotropic stress is appropriate for perfect
fluids and scalar field matter. The field is linearized as in (1.108).
We therefore have two fluctuating variables: the metric perturbations represented
by Φ and the field perturbations by δφ. The equations governing the evolution of the perturbations are the following: (the subscript 0 in the homogenous part of the field
0 ∂V is dropped, and V (φ) ≡ ∂φ ):
∇2Φ − 3HΦ0 − H0 + 2H2 Φ = 4πG φ0δφ0 + V 0a2δφ
Φ0 + HΦ = 4πGφ0δφ (3.18)
Φ00 + 3HΦ0 + H0 + 2H2 Φ = 4πG φ0δφ0 + V 0a2δφ
Combining these equations, and eliminating δφ one gets a second order equation governing the metric perturbation Φ:
φ00 φ00 Φ00 + 2 H − − ∇2Φ + 2 H0 − H Φ = 0 (3.19) φ0 φ0
The fact that the perturbations are expressible in terms of a single fluctuating variable
indicates that the system has one true dynamical degree of freedom.
106 The form of (3.19) provides some insight into the physics of the evolution of the perturbations. The last term represents the gravitational force causing the instability, the Laplacian term represents the restoring force due to pressure, and the second term represents Hubble friction. For superhorizon modes k < H there is no pressure force, and the perturbations freeze out, while the growth of their amplitude is governed by the other two forces. For subhorizon modes, the pressure force leads to damped oscillations.
(3.19) can actually be solved asymptotically in the short and long wavelength approximations [2]. Introducing the new variable
a u = Φ (3.20) φ0 one can reduce (3.19) to the form
θ00 u00 − ∇2u − u = 0, θ = H/ (aφ0) (3.21) θ
Solving this in the asymptotic limits, one can obtain explicit expressions for Φ and
δφ (see [2]) for details).
A different way of manipulating (3.19) is as follows. Defining a variable ζ
2 H−1φ˙ + φ ζ ≡ Φ + (3.22) 3 1 + w where w = p/ρ, then (3.19) can be expressed in a very simple form
ζH˙ (1 + w) = O ∇2Φ (3.23)
107 Since the right side of (3.23) is negligible for superhorizon modes, (3.23) implies conservation of ζ on large scales. The variable ζ happens to be a gauge-invariant quantity which can be physically interpreted as the curvature perturbation on slices of constant energy density. It is approximately equal to the comoving curvature perturbation R (see § 3.2.3) on large scales [3].
The constancy of ζ on large scales allows one to use (3.22) to relate initial and
final values of the metric perturbation. For example, if the amplitude of Φ is known at the time of horizon exit during inflation, one can use (3.22) to deduce the amplitude of Φ at Horizon entry during last scattering.
Finally, note that the constraint equations link Φ to the amplitude of density perturbations, and hence the amplitude of large angle temperature anisotropies in the CMB via the Sachs Wolfe effect [10]:
δT (e) 1 = Φ(x ) (3.24) T 3 ls
−1 where e signifies the direction of observation, and xls = 2H0 e.
3.2.6 Quantum Theory of Cosmological Perturbations
The inhomogeneities in the inflaton are the result of quantum fluctuations, so in order to fully understand their origin and evolution, a quantum mechanical treatment is es- sential. The smallness of the fluctuations allows one to sidestep the problem of metric backreactions by using the semiclassical approximation, namely, that the spacetime responds only to the expectation value of the energy-momentum tensor. Given this hypothesis (which is well-motivated by the smallness of the CMB anisotropies), and
108 given that the system has only one true dynamical degree of freedom (as demon- strated in the previous subsection), the quantum mechanical theory of cosmological perturbations reduces to the quantum theory of a single free scalar field with a time dependent mass. I now provide a very brief summary of the steps in this analysis.
1. The starting point is the Einstein-Hilbert action (1.31), and a single scalar field.
2. The metric is perturbed using the longitudinal gauge prescription, and the field
is split as in (1.108).
3. The action is then expanded up to second order in the perturbative variables.
4. A gauge-invariant variable v, which is a combination of metric and matter
perturbations, is identified, such that the action takes on a canonical form in v:
1 Z z00 S(2) = d4x v02 − (∇φ)2 + v2 (3.25) 2 z
The variable v, called the Mukhanov variable plays a crucial role in the theory of cosmological perturbations. v and z are defined as follows:
φ v = a δφ + Φ (3.26) H aφ0 z = (3.27) H
From equation (3.11), we see that v R = (3.28) z
109 The action (3.25) immediately dictates the following equation of motion for the
Mukhanov variable (in Fourier space):
z00 v00 + k2 − v = 0 (3.29) k z k
The next step is to solve this equation. If we assume that the slow roll conditions
hold (see § 1.9.3), then φ0 and H evolve much slower than the scale factor, and (3.27) implies that z00 a00 ≈ (3.30) z a
If we also assume that the spacetime evolves as de Sitter (with the scale factor given
by (2.1)) i.e., we take → 0, then equation (3.29) turns into the familiar equation of
a massless scalar field in de Sitter space:
a00 v00 + k2 − v = 0 (3.31) k a k
The exact solution of Eq. (3.31) with the boundary condition that small wave-
length modes go over into Minkowski space modes is:
e−ikη i vk = √ 1 − (3.32) 2k kη
∗ The other independent mode is vk. These are the mode functions for the Bunch- Davies vacuum [11]. A derivation of these mode functions using inverse scattering
technology can be found in [12]. The Bunch Davies modes make for a good choice of
initial conditions also because they are known to be a local attractor in the space of
110 initial conditions in an expanding background [13].
Now using (3.16), and in the limit of superhorizon modes k |η| 1, the power spectrum of the comoving curvature perturbation (due to the modes exiting the Hori- zon during inflation) turns out to be
1 H4 PR = (3.33) 2 ˙2 4π φ k=aH
This is the well-known scale-invariant, or Harrison-Zeldovich spectrum, which is one of the most successful predictions of inflation, and seems to be bourne out by
CMB observations. Note that our simplified analysis (where we effectively took the slow-roll parameters to be zero) hides the fact that H and φ˙ slowly evolve during inflation, and so a small scale-dependence is implicit in the ratio H2/φ˙. In fact it
2 ˙ 2 ˙ is better to replace H /φ by Hk /φk, where the subscript k signifies the value of a quantity at the time the mode k exists the horizon.
In terms of the first slow roll parameter (defined in (1.112)) the power spectrum can be written as
1 V PR = 2 4 (3.34) 24π mPl
This equation shows how CMB observations of the amplitude of the perturbations can be used to constrain the scale of the inflation potential.
We take the scale-dependence into account in calculating the spectral tilt from
(3.17)
111 d ln P d ln H4 d ln φ˙2 n − 1 = R = k − k = 2η − 6 (3.35) s d ln k d ln k d ln k where we have used the manipulation d ln k = d ln (aH) ' d ln a. The form of the spectral index in terms of the slow roll parameters (3.35) agrees with the standard result from inflation.
3.3 Difficulties in computing perturbations from
Island Cosmology
3.3.1 Classical vs Quantum Fields
Unlike most models of inflation, however, computing the perturbation spectrum in
Island Cosmology is extremely difficult. First of all, in Island Cosmology, both the background field φ0 and the perturbation δφ are quantum operators. To calculate density fluctuations due to δφ, one needs a suitable model for the evolution of φ0 during the NEC violating fluctuation. This evolution is quantum and not described as a solution to some classical equation of motion. The closest related problems that have been addressed in the literature are the production of particles during the quantum creation of the universe and the fluctuations of a vacuum bubble that has itself been produced in a tunneling event [14, 15, 16]. These analyses rely on the existence of an instanton describing the tunneling event. In our case, the NEC violation is not described by an instanton; instead it is described by the most probable fluctuation leading to matter-genesis. Hence the existing techniques do not apply directly and
112 new techniques are needed.
3.3.2 Backreaction on spacetime
Another major difficulty in computing the perturbation spectrum is the lack of an adequate characterization of the backreaction of the metric. Hence the best one can do is assume some hypothesis for the backreaction. I explain this in detail below and also discuss our working hypothesis for the backreaction on the spacetime.
In this model, we have quantum fields in a classical spacetime. On the surface, this seems to be an ideally suited for the machinery developed in the theory of semiclassical relativity:
ˆren Gµν = 8πGhTµν i (3.36) where hi denotes the expectation value in some specified quantum state. In other words, semiclassical relativity assumes that the spacetime responds only to the expec- tation value of the energy-momentum tensor - fluctuations in the energy-momentum tensor about the mean are unimportant. However, in studying Island Cosmology,
ˆren we need to determine the effects of fluctuations of Tµν . So it is essential that we go beyond semiclassical relativity to be able to treat the backreaction of the NEC violating fluctuations on the spacetime. For small fluctuations, one could envisage expanding the metric around a fixed background and quantizing the metric fluctua- tions. Such an attempt has been made in Refs. [17] though not in the context of NEC violations. The perturbative scheme can not however hope to capture the physics of large fluctuations of the kind we are interested in.
Since a rigorous treatment of the backreaction is not possible, we shall adopt a
113 “working hypothesis” in which the NEC violating fluctuation behaves like “phantom
energy” i.e. a classical perfect fluid with equation of state w ≡ p/ρ < −1 where
ρ > 0. Furthermore, in the sudden approximation discussed in the previous section,
the phantom energy exists only for a vanishingly small time period. Hence the energy
content of the universe has the following time dependence:
ρ = Λ , w = −1 , η < ηf 1 ρ = ρ , w = + , η > η (3.37) FRW 3 f
where ηf denotes the instant at which the NEC violating fluctuation occurs and
ρFRW denotes the energy density after the NEC violating fluctuation is over and this is assumed to be dominantly in the form of radiation. The initial condition for the
FRW phase is: ρFRW(ηf ) = ρmg, the radiation density required for matter-genesis. With this working hypothesis for the energy content of the local universe, the backreaction on the spacetime is given by:
8πG H2 = ρ (3.38) 3
where, as usual, H =a/a ˙ and a(t) is the local scale factor. For η < ηf , ρ is a constant and H = HΛ which is constant. In a vanishingly small interval around
ηf , ρ increases rapidly due to the NEC violating fluctuation and this means that H also increases correspondingly. This implies that the scale factor grows faster than exponentially during the NEC violating fluctuation, yielding a vanishingly short period of “super-inflation”. After the NEC violating fluctuation is over, the region is
114 filled with radiation energy density and the FRW epoch starts. We summarize the
behavior of the Hubble scale, H, as follows:
H = HΛ , η < ηf
H = HFRW , η > ηf (3.39)
with the initial condition HFRW(ηf ) = Hmg where Hmg is the Hubble constant at the epoch of matter-genesis.
In writing Eq. (3.38), we are assuming that spacelike surfaces of constant ρ are flat.
This is a consequence of our choice of the Bunch Davies modes as initial conditions
which assume a spatially flat spacetime slicing. That is why we have not included
the spatial curvature term, k/a2. This also consistent with our working hypothesis
since the initial state (η < ηf ) is flat and the universe expands even faster during the phantom energy stage which is entirely classical.
In the oft studied example where a scalar field tunnels through to a different
value, it is known that the surfaces of constant field have negative spatial curvature.
The energy density in the field is purely due to potential energy and so the surfaces
of constant field are also surfaces of constant energy density. Hence the tunneling
event produces an open universe with negative spatial curvature [18]. The scenario
in this model is different from the tunneling scenario because there is no instanton
that describes the NEC violating fluctuation. It can be shown explicitly that the
tunneling process preserves de Sitter invariance [15] (though see the caveat mentioned
in Footnote 33 in [19]) and this symmetry implies hyperbolic spatial slicings (i.e. open universe slicings) of the spacetime. In our case we know that NEC violations only
115 occur if de Sitter invariance is broken. This can be seen by considering
ˆren ˆren hTµν Tλσ i (3.40)
If we demand that this be a tensor respecting the de Sitter symmetry, then it must be expressible in terms of the metric tensor since this is the only tensor available to us.
µ ν ˆren 2 However, then, when we contract with null vectors to get h(N N Tµν ) i, the result
α β will be zero since gαβN N = 0, and there will be no NEC violating fluctuations. This “working hypothesis” is probably the weakest assumption in our analysis.
However, we cannot do any better at the moment because the backreaction of quan- tum fluctuations on the spacetime requires that we consider a quantum theory of gravity as well. The backreaction problem also occurs in eternal inflationary cosmol- ogy where a similar working hypothesis is used.
A rigorous calculation perturbation spectrum from Island Cosmology, particularly addressing the backreaction issue using quantum cosmology, string theory, or loop gravity is an open question for future work. In the rest of this chapter I describe some efforts to compute the perturbations using assumptions which alleviate these difficulties. The next section calculates the spectrum of perturbations generated in a field uncoupled to the NEC violating field, assuming that the NEC violation is practically instantaneous. In the section after that I investigate the scenario of a phantom field mimicking the NEC violation.
116 3.4 Perturbations in a Spectator Field
In this section, we consider a light field other than the NEC-violating field φ, and not interacting directly with φ (a “spectator” field), and find its power spectrum based on the assumption that the NEC violating fluctuation is instantaneous at the conformal time ηf .
Let us denote such a field generically by χ, its eigenmodes by χk(η) exp(ik · x). From the discussion on cosmological perturbations earlier in this chapter, one can conclude that this field will have a spectrum of perturbations given by:
k3 v 2 P (k, η) = k (3.41) χ 2π2 a
After the quantum fluctuation is over, During the radiation dominated FRW epoch, i.e. for η > ηf , we have s t a(t) = af (3.42) tf where tf is the cosmic time at which the NEC violating fluctuation occurs, and af ≡ a(tf ). (Note that the Hubble parameter is discontinuous at ηf in the sudden approximation but the scale factor is continuous.) In terms of the conformal time one
finds:
2 a(η) = af + af Hf (η − ηf ) (3.43)
Clearly a00 = 0. Therefore, (from (3.31)):
−ikτ +ikτ vk = αke + βke , τ ≡ η − ηf > 0 (3.44)
117 Next we need to solve Eq. (3.31) at η = ηf . This step is non-trivial since a is
0 00 continuous at ηf but a is discontinuous. Hence a has a delta function contribution. Using Eqs. (2.1) and (3.43) we find:
a00 2 = Θ(η − η) + a ∆Hδ(η − η ) (3.45) a η2 f f f
where Θ(·) is the Heaviside function and
∆H ≡ Hf − HΛ ≈ Hf (3.46)
Integrating Eq. (3.31) in an infinitesimal interval around ηf , we find the junction conditions:
vk(ηf +) = vk(ηf −)
0 0 vk(ηf +) = vk(ηf −) + af ∆Hvk(ηf ) (3.47) where the last term is due to the δ−function piece in a00/a. We can now find the coefficients αk and βk by inserting the de Sitter and FRW mode functions and their derivatives at η = ηf in the junction conditions. This gives:
1 i α = v + v0 + a H v k 2 kf− k kf− f f kf− 1 i β = v − v0 + a H v (3.48) k 2 kf− k kf− f f kf−
0 where vkf− ≡ vk(ηf −) and similarly for the (conformal) time derivative vk.
118 We are interested in the long wavelength fluctuations for which k|ηf | → 0. Then
0 the dominant contributions come from the vkf− and af Hf vkf− terms in Eq. (3.48)
2 and are of order 1/(kηf ) . However, the af Hf vkf− term is much larger than the vkf− term because Hf >> HΛ. (Recall from Eq. (2.1) that af ηf = −1/HΛ.) Therefore
1 1 H √ f αk ≈ + 2 2 2k (kηf ) HΛ 1 1 H √ f βk ≈ − 2 (3.49) 2 2k (kηf ) HΛ
Therefore −i 1 H √ f vk(η) ≈ 2 sin(kη) (3.50) 2k (kηf ) HΛ
Using Eqs. (3.43) and (3.50) in (3.41), together with η >> ηf gives:
1 1 sin(kη)2 Pχ(k, ηk) ≈ 2 4 2 4 (3.51) 4π af HΛηf kη
Making use of Eq. (2.1), af ηf = −1/HΛ, and taking the limit kη → 0, we finally get:
H2 P (k, η ) ≈ Λ (3.52) χ k 4π2
Since the result does not depend on k, the spectrum of χ fluctuations is scale invariant, as in the inflationary case [20], with amplitude set by the cosmological constant.
(Corrections to scale-invariance, of course, can be expected from the fluctuation not being exactly instantaneous).
As discussed earlier in this section, the result in Eq. (3.52) applies to all very light or massless fields other than, and not interacting directly with, the NEC violating
119 field. In particular, in the context of the gravitational wave power spectrum, the
perturbation of the metric is equivalent to χ/mP where mP is the Planck mass.
2 Hence the power in gravitational waves is proportional to (HΛ/mP ) and is very tiny.
3.5 A Classical Treatment of Island Cosmology
3.5.1 Introduction
In this section we present a classical computation of the perturbations generated
in Island Cosmology by assuming that the NEC-violating field behaves as a classical
phantom field for the duration of the fluctuation. Using an exactly-solvable potential,
we show that the model generates a scale-invariant spectrum of scalar perturbations,
as well as a scale-invariant spectrum of gravitational waves. The scalar perturbations
can have sufficient amplitude to seed cosmological structure, while the gravitational
waves have a vastly diminished amplitude.
We set our notation as follows: let us choose our cosmic time coordinate t such that
the phantom phase (described in Section 3.5.3) begins at t = ti = 0. In this paper we find it more convenient to work in conformal time (η) where dt = a(η) dη and η ranges between (−∞, 0) as t goes from −∞ to +∞. We set our conformal time coordinates such that the period of phantom cosmology lasts between η = ηi = −1/HΛ and
η = ηf . The above choices of ti and ηi are arbitrary and for convenience. Primes denote derivatives with respect to conformal time, and dots denote derivatives with
respect to cosmic time. We adopt the convention that the suffix i denotes the value
of a quantity at η = ηi and the suffix f denotes the value at η = ηf .
120 I now briefly recapitulate the different stages of Island Cosmology, with special emphasis on the assumptions relevant to the present analysis.
3.5.2 The de Sitter phase
This phase represents the initial state of the Universe, before the onset of the phantom behavior. In this phase, we assume that the Universe is de Sitter space inflating due to the observed dark energy, which we assume is a cosmological constant. The Hubble parameter (H) has the same value that it has today, which we call HΛ . As discussed in the previous chapter, this expanding de Sitter background can be part of a classical de Sitter spacetime with no beginning and no end, with early contraction and then expansion. We will only consider the expanding phase of the de Sitter spacetime in the following discussion.
We assume that the matter content of the Universe is a classical scalar field (φ) having a Lagrangian L given by:
µ L = λ∂ (φ)∂µ(φ) − V (φ) (3.53) and a stress tensor given by:
Tµν = λ∂µ(φ)∂ν(φ) − gµνL (3.54) gµν represent the components of the metric, and V (φ) represents the potential. For the sake of generality, we have inserted the constant λ which determines the sign of the kinetic term. Obviously, λ = +1 for an ordinary classical scalar field, and λ = −1
121 for a phantom field.
The equation of motion of the field is the Klien Gordon equation:
1 √ ∂V λ√ ∂ −ggµν∂ φ + = 0 (3.55) −g µ ν ∂φ
The scale factor a(t) and Hubble value H(t) during this period can be written as follows:
For (−∞ < t ≤ ti = 0):
a(t) = eHΛt (3.56) a˙(t) H(t) = = H (3.57) a(t) Λ
In terms of conformal time, for (−∞ < η ≤ ηi):
1 a(η) = − (3.58) ηHΛ a0(η) 1 H(η) = = − (3.59) a(η) η
3.5.3 The Phantom Phase
In this phase, lasting between the times ηi and ηf , the Universe undergoes an NEC- violating quantum fluctuation. We model this phase by assuming that the matter content of the Universe behaves like a phantom field φp for the duration of the fluc- tuation.
122 We assume that this hypothetical phantom field φp is classical, i.e., its Lagrangian and stress tensor are given by Eq. (3.53) and Eq. (3.54), and it satisfies the Klien
Gordon equation Eq. (3.55).
Using the above equations, one can readily determine the pressure p and energy
density ρ of the field. These work out to:
φ02 p = 1 P T = λ p − a2(η)V (φ ) (3.60) 3 ii 2 p φ02 ρ = T = λ p + a2(η)V (φ ) (3.61) 00 2 p
2 Clearly, p + ρ = λφp indicating, as one would expect, that the NEC is violated during this period if our matter field is phantom (λ = −1).
−1 −1 Also during this phase, the Hubble horizon size drops from HΛ to Hf . We assume that this drop is linear in cosmic time, ending at time tf . The validity of this assumption, as well as its implications on the matter content of the Universe are discussed later. Thus,
−1 −1 H (t) = HΛ − αt (for 0 ≤ t ≤ tf ) (3.62)
Here α is a dimensionless parameter measuring the rate at which the horizon size changes during the phantom phase. We assume that the quantum fluctuation is very abrupt, and hence α is very large.
Using the definition of conformal time, it is easy to deduce that during this phase
123 (ηi ≤ η ≤ ηf ),
− 1 a(η) = [−α − HΛ(1 + α)η] 1+α (3.63) a0(η) H H(η) = = Λ (3.64) a(η) [−α − HΛ(1 + α)η] H(η) H(η) = = H [a(η)]α (3.65) a(η) Λ
We also need to address the nature of the back-reaction of matter on geometry.
Let us make the working hypothesis that the back-reaction is fully described by the
Friedmann equation.
The actual time duration of this phase can be calculated by demanding continu- ity of the Hubble value at t = tf (or equivalently, η = ηf ). Thus we require (see Eq. (3.65)),
H(ηf ) = Hf (3.66)
Solving the above equation for ηf , we obtain
" 1 # 1+ α 1 HΛ ηf = − − α (3.67) HΛ(1 + α) Hf
We assume that Hf HΛ and since α → ∞, the first term in the square brackets can be ignored leaving us
α 1 ηf ' − − O (3.68) HΛ(1 + α) (1 + α)Hf
From this we can compute the duration of the phantom phase in conformal time (∆η),
124 as follows:
∆η = ηf − ηi 1 1 = (3.69) HΛ α + 1
Again, since α → ∞, this is a vanishingly small interval.
3.5.4 The Radiation Dominated FRW Phase
−1 In this epoch we have a volume of space of Hubble length Hf filled with classical radiation. Rapid interactions thermalize the radiation, after which this volume follows a standard FRW evolution.
The scale factor in this radiation-dominated epoch can therefore be written as
s t a(t) = af (3.70) tf
In terms of the conformal time, this reduces to
2 a(η) = af + af Hf (η − ηf ) (3.71)
3.5.5 Calculational Strategy
We make two key assumptions to facilitate our calculation, which we discuss below.
These are:
1. During the NEC-violating explosive event the energy content of the Universe
125 behaves as a phantom field.
2. During the NEC violation, the drop in the Hubble scale is linear in cosmic time.
Assumption (1) is an attempt to model the behavior of the matter field during the NEC-violating event. To calculate density fluctuations due to fluctuations in the
NEC-violating field, one needs a suitable model for the evolution of the field itself during the NEC-violating fluctuation. This evolution is quantum and not described as a solution to some classical equation of motion. For the purpose of this calculation, we have made the simplifying assumption that the matter field behaves in the same manner as a classical object that would also violate the NEC and produce the same effect on the spacetime. Of course this purely classical treatment cannot substitute for a rigorous quantum mechanical treatment of the NEC-violation, but we hope that it captures the essential elements of the physics involved.
Assumption (2) can be justified considering that the drop in the Hubble length need not be linear throughout the explosive event, but only during the window of time δt that it takes for the scales observed today to leave the horizon. Since the
−1 fluctuation itself is very short lived, by expanding H (t) in a Taylor series about ti, the drop in H−1(t) over δt can be well approximated to be linear.
We now turn our attention to computing the spectrum of perturbations that would be generated in this cosmological model. Our plan of action is the following:
1. Working in k (momentum or wavenumber) space, we first find expressions for a
gauge-invariant variable vk(η) , representing the true degrees of freedom of the system in all the three stages of the model.
126 2. The unknown coefficients that arise in the above expressions are then deter-
0 mined by demanding continuity of vk(η) and its time derivative vk(η) at tran-
sition times ηi and ηf .
3. As discussed earlier, the adiabatic density perturbation responsible for structure
in the Universe is conveniently characterized by the curvature perturbation R
seen by comoving observers. Once we fully determine vk(η) in the radiation dominated phase, we obtain the co-moving curvature perturbation spectrum
from the relations (3.28) and (3.16).
3.6 Spectrum from a Classical Treatment
We next find expressions for vk in the three stages of the model. The unknown coefficients that appear in these expressions will be determined through a matching
process in § 3.6.3, and then the scalar and tensor power spectra are computed in §
3.6.4 and § 3.6.5 respectively.
3.6.1 vk in the de Sitter and FRW phases
For a scalar field in de Sitter space, the variable vk(η) satisfies (3.31) with the solution (3.32)
During the FRW phase after the NEC-violating fluctuation, from the discussion in §
3.4 we know that vk satisfies
ikτ −ikτ vk(τ) = αke + βke (3.72)
127 where αk and βk are constants of integration and τ = η − ηf > 0
3.6.2 vk in the Phantom Phase
In this case we will calculate vk starting from first principles.
Matter and metric perturbations
The first step is to perturb the matter and metric. Working in longitudinal gauge
and assuming no anisotropic stress, the scalar metric perturbations are written as:
1 + 2Φ 0 0 0 0 −1 + 2Φ 0 0 2 gµν = a (3.73) 0 0 −1 + 2Φ 0 0 0 0 −1 + 2Φ
Given our choice of gauge, the metric perturbation Φ(η, x) coincides with the gauge- invariant Bardeen potential (see, for example [2]).
The phantom matter field φp(η, x) is perturbed as follows:
φp(η, x) = φ(η) + δφ(x, η) (3.74)
128 Evolution of the perturbations
To find time evolution of the perturbations, we use the perturbed Einstein equations up to first order. The i-i component of the zero-th order equations reads:
a0 2 a00 φ02 − Λa2 + − 2 = 8πG λ − a2V (φ) (3.75) a a 2 while the 0-0 component reads:
a0 2 φ02 Λa2 + 3 = 8πG λ + a2V (φ) (3.76) a 2
Adding these equations, one obtains the familiar relationship:
H2 − H0 = 4πGλφ02 (3.77)
The i-i, 0-0 and 0-i components of the first order equations are respectively:
a0 2 a00 a0 Λa2Φ − 2Φ + 4Φ + 3 Φ0 + Φ00 = a a a 1 1 8πG a2V (φ)Φ − a2V δφ + λφ0δφ0 − λΦφ02 (3.78) 2 φ 2 a0 Λa2Φ − 3 Φ0 + ∇2Φ = a 1 1 8πG a2V (φ)Φ + a2V δφ + λφ0δφ0 (3.79) 2 φ 2 a0 Φ0 + Φ = 4πGλφ0δφ (3.80) a
(where Vφ = ∂V/∂φ).
129 Perturbing the Klein-Gordon equation Eq. (1.105) using Eq. (3.74) yields, at zero-
th order:
00 0 2 λ [φ + 2Hφ ] = −a Vφ (3.81)
Using Eq. (3.77), Eq. (3.78), Eq. (3.79), Eq. (3.80) and Eq. (3.81), one obtains the
equation of motion of Φ(η, x):
φ00 φ00 Φ00 − ∇2Φ + 2 H − Φ0 + 2 H0 − H Φ = 0 (3.82) φ0 φ0
Applying the Fourier transform:
Z d3k Φ(x, η) = Φ (η) eik.x (3.83) (2π)3/2 k
we obtain: φ00 φ00 Φ00 + 2 H − Φ0 + k2 + 2 H0 − H Φ = 0 (3.84) k φ0 k φ0 k
Note that Eq. (3.84) is independent of λ, indicating that it has the same form for a phantom field as it would for a normal field. This is a surprising result, since all the equations used to derive Eq. (3.84) are λ dependent. Physically, this result implies that the evolution of the metric perturbation is insensitive to whether the matter content of the Universe is normal or phantom.
To solve Eq. (3.84), we need to determine the dynamics of the phantom field φ(η), which is in turn determined by the potential V (φ). We choose a particular form of the potential which allows for a solution in closed form:
130 (3 + α) 2αH V (φ ) = K2 exp Λ φ (3.85) p 2α K p
Here K is a constant that has the dimensions of mass squared, and sets the scale of the potential.
For this potential, it is easy to verify that the exact form of φ(η) which satisfies
Eq. (3.81) (with of course λ = −1 as is the case in the phantom phase) is:
K φ(η) = − ln [−α − HΛ(1 + α)η] (3.86) HΛ(1 + α)
Further, to facilitate the back-reaction as discussed in Section 3.5.3, and satisfy our ansatz given by Eq. (3.62) we must require that our field satisfies the Friedmann equation Eq. (3.76). The result of this is to fix the value of K:
r 3α K = H (3.87) Λ 4πG
The matter field φ(η) has the interesting property that
φ00(η) = (1 + α) H(η) (3.88) φ0(η)
Using Eq. (3.88), Eq. (3.84) reduces to
00 0 2 2 Φk − 2HαΦk + k − 2H α Φk = 0 (3.89)
131 This is a familiar second order differential equation of the form:
00 0 Φk + P (η)Φk + Q(η)Φk = 0 (3.90) with
P (η) = −2H(η)α
Q(η) = k2 − 2H2(η)α which has the solution (see e.g.[21])
− 1 R P (η) dη α Φk(η) = e 2 χ(η) = a χ(η) (3.91) where χ(η) satisfies the differential equation
1 1 χ00(η) + Q − P 0 − P 2 χ(η) = 0 (3.92) 2 4
In our case, this reduces to
2 00 2 αHΛ χ (η) + k + 2 χ(η) = 0 (3.93) [α + HΛ(1 + α)η]
At this point it is convenient to temporarily switch to a new time variable x defined by
x = α + HΛ(1 + α)η (3.94)
Note that x = −1 when η = ηi and from Eq. (3.68), x = O [HΛ/Hf ] or x ' 0 when
132 η = ηf . With x as the time variable, Eq. (3.93) takes the form
d2χ(x) 1 1 + m2 − p2 − χ(x) = 0 (3.95) dx2 4 x2
where m and p are defined by
k m = (3.96) HΛ(1 + α) 1 α 1/2 p = − (3.97) 4 (α + 1)2
The solution to Eq. (3.95) can be written in terms of Bessel functions as
√ χ(x) = mx [AmJp(mx) + BmYp(mx)] (3.98)
where Am, Bm are constants and Jp and Yp denote Bessel functions of the first and second kind of order p respectively.
Putting together Eq. (3.98) and Eq. (3.91), we conclude that the time evolution of Φk is fully described by the equation
α√ Φk(η) = a mx [AmJp(mx) + BmYp(mx)] (3.99)
where m, x and p are defined by Eq. (3.96), Eq. (3.94) and Eq. (3.97) respectively,
and α is defined in Eq. (3.62).
133 Calculating the Mukhanov Variable
We are first going to compute Rk and z and then use Eq. (3.28) to compute vk.
Calculation of Rk: Noticing that for this space, we have
H0 − 1 = α (3.100) H2 and using equations Eq. (3.80) and Eq. (3.77), we can eliminate the δφ in Eq. (3.11),
0 giving us an expression for Rk involving Φk and Φk as the only first order variables:
1 R = Φ − [Φ0 + HΦ ] (3.101) k k αH k k
(Again, the absence of λ indicates that the expression for Rk is insensitive to whether the field is real or phantom.)
Calculation of z: In the definition of z Eq. (3.27) we eliminate φ0 using Eq. (3.77), and introduce α using Eq. (3.100) to get
r α z = a −4πGλ r 2α = am (3.102) Pl −λ
p where mPl is the reduced Planck mass defined by mPl = 1/8πG, where G is New- ton’s gravitational constant.
134 Final expression for vk: Combining Eq. (3.28), Eq. (3.101) and Eq. (3.102), we
0 can express vk entirely in terms of Φk and Φk as follows:
r α 1 v = a Φ − (Φ0 + HΦ ) (3.103) k −4πGλ k αH k k
3.6.3 Calculation of Unknown Constants
The above calculations produced four unknown constants Am, Bm, in Eq. (3.98) and
αk and βk, in Eq. (3.72). These constants can be determined by demanding continuity
of vk and its time derivative at the two transition times ηi and ηf .
To determine Am and Bm, we perform the above matching process at η = ηi =
1 − , or in terms of x, (from Eq. (3.94)), at x = xi = −1 . In other words, we need HΛ to simultaneously solve the equations
vk(de Sitter)|(η=−1/HΛ) = vk(phantom)|(x=−1)
0 0 vk(de Sitter)|(η=−1/HΛ) = vk(phantom)|(x=−1) (3.104)
to find the unknowns Am and Bm.
The expression obtained for vk in the phantom phase (by substituting the values of the above constants) is fairly complicated. However, since we are only interested in
the super-horizon modes, we can make the approximation that k (or m) → 0, and use
the appropriate asymptotic forms of the Bessel functions Jp(mx) and Yp(mx). Also,
1 since α is large, from Eq. (3.97), p ' 2 . With these simplifications the expression for
135 vk (in the phantom phase) reduces to:
ik/HΛ (1+α) i e a HΛ(−1 + xα) vk(x) ' − √ 2k3/2(1 + α) +O k−1/2 (3.105) i eik/HΛ a(1+α)H2 v0 (x) ' − √ Λ + O k−1/2 (3.106) k 2k3/2x
Now having fully determined the form of vk(η) during the phantom phase, we can
determine the coefficients αk and βk in Eq. (3.72) to determine vk(η) in the final FRW phase. In particular, we need to solve simultaneously the equations:
vk(phantom)|(x'0) = vk(FRW)|(η=ηf )
0 0 vk(phantom)|(x'0) = vk(FRW)|(η=ηf ) (3.107)
For brevity, let us call the leading term in k in the expression for vk (Eq. (3.105)) at
0 η = ηf (or x ' 0) as l1 and the leading term in k in the expression for vk (Eq. (3.106))
at x ' 0 as l2. Thus we have
ik/HΛ (1+α) i e a HΛ l1 = √ (3.108) 2k3/2(1 + α) and,
ik/HΛ (1+α) 2 i e a HΛ l2 = − √ 3/2 2k xf (2+2α) i ei/kHΛ a H2 = √ f Λ (3.109) 2k3/2
136 −(1+α) Where in the last manipulation we have used the result that −xf = af , which follows from the form of the scale factor during the phantom phase (Eq. (3.63)) and
the definition of x (Eq. (3.94)). Eq. (3.107) now implies that
1 l2 αk = 2 l1 + ik
1 l2 βk = 2 l1 − ik (3.110)
Using Eq. (3.65) at (η = ηf ) to find
α af HΛ = Hf (3.111) we note from equations Eq. (3.109) that,
l1 1 1 = 1+α = 1 (3.112) l2 (1 + α)af HΛ af Hf (1 + α)
since both α and Hf are large. Eq. (3.110) now reduces to
1 l α ' 2 k 2 ik 1 l β ' − 2 (3.113) k 2 ik
Hence the form of vk(η) in the phantom phase becomes:
l l v (η) ' 2 eikτ − 2 e−ikτ k 2ik 2ik sin(kτ) = l τ (3.114) 2 kτ
137 3.6.4 Determination of the Scalar Power Spectrum
Now we are in a position to determine the power spectrum of the co-moving curvature perturbation in the FRW space using Eq. (3.16) which gives
3 2 k sin(kτ) τ PR = 2 l2 (3.115) 2π kτ zFRW(η)
2 Making the approximation aFRW(η) = af Hf τ from Eq. (3.71) since at the time a mode re-enters the horizon, η ηf , using the relation Eq. (3.111), and taking the limit k → 0, we finally obtain
3 −ik/H 2 k i e Λ PR ' √ Hf (3.116) 2 3/2 2π 2 2k mPl H 2 = f (3.117) 4πmPl
Hence we find that our model produces a (nearly) scale invariant spectrum of cosmo-
14 logical perturbations, with amplitude set by Hf /mPl. If we assume Hf ∼ 10 GeV (approximately GUT scale), then the power spectrum matches the COBE DMR ob- servations [22] of CMB temperature fluctuations of order 10−5. In other words, the perturbation spectrum can have an amplitude sufficiently large to seed the cosmolog- ical structure that we see today.
3.6.5 Determination of the Tensor Power Spectrum
We know from the theory of cosmological perturbations (see, for example, [5]) that gravitational waves are essentially equivalent to two massless scalar fields (for each
138 polarization) up to a renormalization factor of 2/mPl. Hence we can write the tensor power spectrum PRT as 4 PRT = 2 2 PRψ (3.118) mPl where the first factor on the right comes from the two polarization states, the sec- ond represents the renormalization mentioned above, and PRψ is the spectrum of perturbations of a massless scalar field ψ other than, and not interacting with, the
NEC-violating field. The ψ-field perturbations can be computed using essentially the same machinery as above, with a small difference in the final step:
1. We solve Eq. (3.84) with ψ representing the matter field.
2. We find expressions for vk in the three stages of the model (up to constants of integration).
0 3. We demand the continuity of vk and vk at ηi and ηf to fully specify the expression
for vk in the FRW region (that is, evaluate the undetermined constants obtained
in the previous step), and from this, compute the perturbation spectrum PRψ using the relation 3 2 k vk(η) PRψ(η) = 2 (3.119) 2π a(η) FRW
The final expression for the power spectrum of tensor perturbations turns out to be !2 2 HΛ HΛ PRT = 8 3 ' 8 (3.120) af πmPl πmPl
The last step follows because it is easy to show (by Taylor expanding a(η) about the point η = ηi) that the scale factor hardly changes from its initial value of 1 during
139 the NEC-violating event. The result agrees (up to O(1) numerical factors) with the corresponding result obtained in (3.52).
This calculation is independent of the nature of the NEC-violating phantom field.
It is valid as long as the spacetime responds to the NEC-violation with a sharp drop in the Hubble parameter.
2 Eq. (3.120) indicates that power in gravitational waves is proportional to (HΛ/mPl) = 10−122. This clearly precludes any possible detection of these gravitational waves, ei- ther directly, or in the CMB polarization.
A similar scenario was investigated by Y.S Piao [23], and while he obtains a scale invariant spectrum of scalar perturbations, the tensor perturbation spectrum in his calculation turns out to be blue-shifted. The source of the discrepancy could lie in his assumption that the canonical relationship satisfied by vk in the case of the tensor spectrum has a time-dependent mass given by a00(η)/a(η) (Eq. (12) in [23]) during the phantom phase. In our approach we derive vk from first principles and find that the time dependent mass can have a more complicated form.
3.7 Conclusion
In this chapter, I have briefly reviewed the theory of cosmological perturbations and demonstrated how inflationary models lead to an approximately scale invariant spec- trum.
I have then discussed the subtleties and issues inherent in attempting to compute the perturbation spectrum in Island Cosmology. The spectrum of density perturba- tions due to the NEC violating field itself is still an open problem. Determining this
140 spectrum will be crucial to determining if the model agrees with observations. For
2 example, if the scale of fluctuations in this field is still set by (HΛ/mP ) then the fluctuations are too small to seed the structure that we know and the island will be
2 a desert. On the other hand, if the scale is set by (Hf /mP ) then there is a chance that island cosmology can be a viable model. In that case, quantum NEC violations provide a definite mechanism by which regions that are “macroscopically indistin- guishable from our universe” can be produced from the dead de Sitter sea. Assuming that the fluctuation is instantaneous, I show that fields other than the NEC violating
field are likely to have a scale invariant spectrum of perturbations.
The third part of this chapter reviews a classical treatment of Island Cosmology, which attempts to alleviate some of the difficulties mentioned earlier through classical assumptions regarding the behaviour of matter and spacetime during the quantum
fluctuation. These calculations yield an adiabatic spectrum of scale-invariant pertur- bations, whose amplitude is determined by the value of the Hubble constant at the end of the NEC-violating fluctuation. If we assume the latter to be approximately
GUT scale, the perturbation spectrum turns out to have an amplitude sufficient to seed the cosmological structure seen today. We also obtain a scale-invariant spectrum of gravitational waves of amplitude set by (HΛ/mPl).
141 Works Cited
[1] Sourish Dutta. A classical treatment of island cosmology. Phys. Rev., D73:063524, 2006. astro-ph/0511120.
[2] V.F. Mukhanov et al. Phys. Rept., 215:203, 1992.
[3] D.H. Lyth and A.R. Liddle. Cosmological Inflation and Large Scale Structure. Cambridge University Press, 2000.
[4] Robert H. Brandenberger. Lectures on the theory of cosmological perturbations. Lect. Notes Phys., 646:127–167, 2004. hep-th/0306071.
[5] David Langlois. Inflation, quantum fluctuations and cosmological perturbations. 2004. hep-th/0405053.
[6] Antonio Riotto. Inflation and the theory of cosmological perturbations. 2002. hep-ph/0210162.
[7] James M. Bardeen. Gauge Invariant Cosmological Perturbations. Phys. Rev., D22:1882–1905, 1980.
[8] Chung-Pei Ma and Edmund Bertschinger. Cosmological perturbation theory in the synchronous and conformal Newtonian gauges. Astrophys. J., 455:7–25, 1995. astro-ph/9506072.
[9] David Wands, Karim A. Malik, David H. Lyth, and Andrew R. Liddle. A new approach to the evolution of cosmological perturbations on large scales. Phys. Rev., D62:043527, 2000. astro-ph/0003278.
[10] R. K. Sachs and A. M. Wolfe. Perturbations of a cosmological model and angular variations of the microwave background. Astrophys. J., 147:73–90, 1967.
[11] T.S Bunch and P.C.W. Davies. Quantum field theory in de sitter space: Renor- malization by point-splitting. Proc. Roy. Soc. Lond., 360:117, 1978.
[12] Sourish Dutta and Tanmay Vachaspati. Islands in the Lambda-sea. Phys. Rev., D71:083507, 2005. astro-ph/0501396.
[13] Robert H. Brandenberger and Christopher T. Hill. Energy density fluctuations in de sitter space. Phys. Lett., B179:30, 1986.
[14] V. A. Rubakov. Particle creation during vacuum decay. Nucl. Phys., B245:481, 1984.
142 [15] Tanmay Vachaspati and Alexander Vilenkin. Quantum state of a nucleating bubble. Phys. Rev., D43:3846–3855, 1991.
[16] Jaume Garriga and Alexander Vilenkin. Quantum fluctuations on domain walls, strings and vacuum bubbles. Phys. Rev., D45:3469–3486, 1992.
[17] N. C. Tsamis and R. P. Woodard. The quantum gravitational back-reaction on inflation. Annals Phys., 253:1–54, 1997. hep-ph/9602316.
[18] Martin Bucher, Alfred S. Goldhaber, and Neil Turok. An open universe from inflation. Phys. Rev., D52:3314–3337, 1995. hep-ph/9411206.
[19] Alexander Vilenkin. Approaches to quantum cosmology. Phys. Rev., D50:2581– 2594, 1994. gr-qc/9403010.
[20] Viatcheslav F. Mukhanov and G. V. Chibisov. Quantum Fluctuation and Non- singular Universe. (In Russian). JETP Lett., 33:532–535, 1981.
[21] P.M. Morse and H. Feshbach. Methods of Theoretical Physics, Part I. McGraw- Hill, 1953.
[22] John C. Mather et al. Measurement of the Cosmic Microwave Background spec- trum by the COBE FIRAS instrument. Astrophys. J., 420:439–444, 1994.
[23] Yun-Song Piao. Is the island universe model consistent with observations? Phys. Rev., D72:103513, 2005. astro-ph/0506072.
143 Chapter 4
Dark Energy Voids
4.1 Introduction
The question of why the Universe accelerates has puzzled cosmologists for over a decade. In the framework of Einstein’s General Relativity, there are two possible explanations: a cosmological constant, or a dynamical dark energy component (DDE).
Determining which of these is correct (assuming, of course, that General Relativity can be trusted on cosmological scales) is one of the most crucial problems in physics, with far reaching implications particularly in particle physics.
In this chapter, I report on my work attempting to find ways of distinguishing a cosmological constant from a dynamical component via gravitational clustering. A cosmological constant, by definition, will remain unaffected in the vicinity of gravita- tionally collapsing matter. The behavior of a field, on the other hand, is not so clear.
As I describe below, we find that a scalar field tends to form voids in the presence of collapsing matter, a behavior that could have interesting observational implications.
144 4.2 What is causing the acceleration?
As I have described in Chapter 1 there is plenty of evidence that the Universe is
accelerating [1, 2, 3, 4, 5]. Nonetheless, the source of the accelerated expansion is as
elusive as ever.
If one assumes that Einstein’s general relativity can be trusted on cosmological
scales, then the only way one can have an accelerating Universe is by hypothesizing the
existence of some form of Dark Energy, which could either be a cosmological constant
or a dynamical field. While the cosmological constant has a fixed ratio of pressure to
energy density, w = p/ρ = −1, dark energy due to a dynamical component such as a scalar field will in general have a varying equation of state (EOS), w(z). Observing a
deviation from −1 or a time-evolution in the EOS will be decisive evidence in favor of
the existence of DDE. However, there are known degeneracies [6] which make this task
extremely difficult, unless the deviation from a cosmological constant is strong. The
current observational limit on the EOS of dark energy is roughly w ≈ −1 ± 0.1 at the
1σ level [1, 2], which is consistent with a cosmological constant. Future experiments hold out the possibility of narrowing this limit by maybe a factor of 10. For recent reviews, see [7, 8].
The effect of the cosmological constant and DDE on the expansion rate can be identical, and there is a need for probes that go beyond the background to distinguish between the two. An example of such a probe is structure formation. There are numerous works exploring the formation of structure in the presence of homogeneous dark energy [9, 10, 11, 12, 13, 14, 15, 16, 17]. An exciting and somewhat controversial possible difference between the cosmological constant and DDE is their clustering
145 behavior. While the cosmological constant is exactly homogeneous on all scales, DDE is expected to be not perfectly homogeneous [18], and the implications of this on the
CMB are well known [19]. However, it is usually assumed that the clustering of DDE is negligible on scales less than 100 Mpc. Whether small perturbations in DDE can be neglected is debatable, and a deeper understanding of the DDE inhomogeneous dynamics is clearly needed.
Several recent works have explored the consequences of DDE clustering on scales shorter than 100 Mpc. Some have adopted a phenomenological approach, param- eterizing the clustering degree of DDE [20, 21, 22, 23, 24]. These works point out potential observables of DDE clustering, justifying further investigation. Works which attempt a more fundamental treatment are mostly in the context of coupled dark en- ergy [25, 26, 27, 28], or other non-trivial models of DDE [29, 30, 31, 32, 33, 34], as clustering is most probable in such theories. However, less attention has been given to the clustering in simpler models of DDE.
In this chapter, I describe my research (with Irit Maor) into exploring the in- homogeneous behavior of DDE. Our approach is straightforward: starting with a gravitational action which includes matter and DDE, we numerically follow the lin- ear evolution of spherical perturbations of matter and the DDE response to these perturbations. For the sake of simplicity, our model for the DDE is a light scalar
field, which is not explicitly coupled to the matter density. As the only coupling between the DDE and the matter is gravitational, our results are conservative in the sense that any model more complicated can be expected to show stronger DDE perturbations than shown here, simply because of the additional coupling beyond
146 gravity.
The striking feature that emerges from our calculation is that in the vicinity of collapsing matter, the DDE develops a spatial profile and tends to form voids. The mechanism that allows the void to form is that although initially the field’s evolution is friction dominated due to the cosmic expansion, the collapse of matter slows down the local expansion. This allows the field to locally roll down and lose energy, creating the void. The presence of the matter perturbation is necessary to trigger this mechanism.
The plan of this chapter is as follows: in § 4.3 I describe our model in detail.
In § 4.4 I present our results. Discussions and conclusions are in § 4.5 and § 4.6 respectively.
4.3 The model
We are interested in spherical perturbations around a flat FRW universe. The most general line element in comoving coordinates is then
ds2 = dt2 − U(t, r)dr2 − V(t, r) dθ2 + sin2 θdϕ2 , (4.1) where U(t, r) and V(t, r) are general functions [35].
We take a cosmic mix of non-relativistic matter and a DDE component as the energy source. The matter component is described by a perfect and pressureless
fluid, with an energy-momentum tensor given by
Tµν(m) = diag (ρ, 0, 0, 0) , (4.2)
147 where ρ is the energy density of matter.
We model the DDE with a classical scalar field φ with a Lagrangian L given by
1 L = (∂ φ)2 − V (φ) , (4.3) 2 µ
and an energy-momentum tensor given by
Tµν(φ) = ∂µφ∂νφ − gµνL . (4.4)
The EOS of the DDE w is defined as
p w = φ , (4.5) ρφ
with the energy density ρφ and the pressure pφ are read off the energy momentum
ij tensor, T00(φ) and −g Tij(φ)/3 respectively.
It is convenient to rewrite Einstein’s equations in the following way,
1 R = K T − g T α . (4.6) µν µν 2 µν α
where Rµν is the Reimann tensor, and K = 8πG. As there is no explicit interaction between the matter and the DDE, energy conservation applies to each separately,
µν µν ∇µT (m) = 0 , ∇µT (φ) = 0 . (4.7)
148 The time-evolution of the system is given by the following equations (where dots denote time-derivatives and primes denote derivatives with respect to the radial co-
0 dV ordinate, except for V (φ) ≡ dφ ):
1 U¨ 1 U˙ V˙ 1 U˙ 2 1 1 U 0 V0 1 V02 V00 + − + + − 2 U 2 U V 4 U 2 U 2 U V 2 V2 V 1 1 −K ρ + V (φ) + φ02 = 0 (4.8) 2 U 1 V¨ 1 U˙ V˙ 1 1 1 U 0 V0 V00 + + + − 2 V 4 U V V 2U 2 U V V 1 −K ρ + V (φ) = 0 (4.9) 2 ! V˙ 1 U˙ ρ˙ + + ρ = 0 (4.10) V 2 U ! V˙ 1 U˙ φ¨ + + φ˙ + V 0(φ) V 2 U 1 V0 1 1 U 0 − − φ00 − φ0 = 0 . (4.11) U V U 2 U
These are subject to the following constraint equations:
1 1 U˙ V˙ 1 V˙ 2 1 1 U 0 V0 1 V02 V00 + + + + − V 2 U V 4 V2 U 2 U V 4 V2 V 1 1 −K ρ + V (φ) + φ˙2 + φ02 = 0 (4.12) 2 2U 1 U˙ V0 1 V˙ V0 V˙ 0 + − − Kφφ˙ 0 = 0 . (4.13) 2 U V 2 V V V
149 4.3.1 Linearization
We now proceed to separate our variables to a homogeneous background and a time and space-dependent perturbation, which we will then linearize. Working in the synchronous gauge [36], we redefine the metric functions U and V as follows:
U(t, r) = a(t)2e2ζ(t,r)
V(t, r) = r2a(t)2e2ψ(t,r) .
(4.14)
Here a(t) is the scale factor of the spatially homogenous and flat background, and
ζ(t, r) and ψ(t, r) are the deviations. We introduce a perturbation around a homoge- neous background also in the matter and the DDE,
ρ(t, r) = ρ(t) + δρ(t, r)
φ(t, r) = φ(t) + δφ(t, r)
V (φ + δφ) = V (φ) + δV (φ, δφ) .
The zeroth order of equations (4.8) - (4.13) gives
1 3H2 − K ρ + V + φ˙2 = 0 (4.15) 2 1 H˙ + 3H2 − K ρ + V = 0 (4.16) 2 ρ˙ + 3Hρ = 0 (4.17)
φ¨ + 3Hφ˙ + V 0 = 0 , (4.18)
150 where H =a/a ˙ is the Hubble function.
To linear order, the evolution equations (4.8)-(4.11) give
2 ζ0 2ψ0 ζ¨ + 4Hζ˙ + 2Hψ˙ + − − ψ00 a2 r r 1 −K δρ + δV = 0 (4.19) 2 1 2ζ 2ψ ζ0 4ψ0 ψ¨ + 5Hψ˙ + Hζ˙ + − + − − ψ00 a2 r2 r2 r r 1 −K δρ + δV = 0 (4.20) 2 δρ˙ + 3Hδρ + ρ ζ˙ + 2ψ˙ = 0 (4.21)
δφ¨ + 3Hδφ˙ + δV 0 1 2 + ζ˙ + 2ψ˙ φ˙ − δφ00 + δφ0 = 0 , (4.22) a2 r and the constraint equations (4.12)-(4.13) reduce to
2 ζ ψ 2ζ0 6ψ0 2Hζ˙ + 4Hψ˙ + − + − − ψ00 a2 r2 r2 r r −K δρ + δV + φδ˙ φ˙ = 0 (4.23) 2 ψ˙ + ζ˙ − Kφδφ˙ 0 + 2ψ˙ 0 = 0 . (4.24) r
Combining equations (4.19), (4.20) and (4.23) gives
ζ¨ + 2ψ¨ + 2H ζ˙ + 2ψ˙ +K δρ − δV + 2φδ˙ φ˙ = 0 . (4.25)
151 The only combination which is relevant to the equations of motion (4.21) and
(4.22) is χ ≡ ζ˙ + 2ψ˙. Comparing (4.17) and (4.21) it is clear that χ can be thought of
as 3δH, and therefore characterizes the spatial profile of the Hubble function. At the
cost of losing some information about the metric, we can reduce the number of our
variables and equations from 4 to 3 by solving for χ instead of for ζ and ψ. Equations
(4.21), (4.22) and (4.25) yield
δρ˙ + 3Hδρ + ρχ = 0 (4.26) 1 δφ¨ + 3Hδφ˙ + δV 0 + χφ˙ − ∇2 (δφ) = 0 (4.27) a2 1 χ˙ + 2Hχ + K δρ − δV + 2φδ˙ φ˙ = 0 . (4.28) 2
Finally, by Fourier-transforming δρ(t, r), δφ(t, r) and χ(t, r) into δρk(t, k), δφk(t, k)
and χk(t, k) respectively, equations (4.26)-(4.28) can be written as a set of ordinary differential equations:
δρ˙k + 3Hδρk + ρχk = 0 (4.29) k2 δφ¨ + 3Hδφ˙ + V 00 + δφ + φχ˙ = 0 (4.30) k k a2 k k 1 χ˙ + 2Hχ + K δρ − V 0δφ + 2φ˙ δφ˙ = 0 , (4.31) k k 2 k k k k where we have used the fact that to linear order, δV = V 0δφ and δV 0 = V 00δφ.
152 4.3.2 Potential
Observationally distinguishing between various potentials of DDE is a formidable task
[37, 38, 39], and a careful analysis of the growth of structure in various potentials might prove a useful tool. Our present goal though is to trace generic properties of
DDE. Accordingly, we choose to work with a simple mass potential,
1 V (φ) = m2φ2 . (4.32) 2
We take the mass scale comparable to the present Hubble scale, mφ/H0 ∼ 1. The light mass assures a slow roll behavior, which will provide accelerated cosmic expansion.
Unless noted otherwise, the figures presented here refer to this mass potential.
In order to verify the generality of our results, we repeated the analysis for a more complicated potential - the double exponential [40],
√ √ Kαφ Kβφ V (φ) = V0 e + e , (4.33) with α = 20.1 and β = 0.5. As we later show, the resulting behavior for the two potentials is qualitatively the same.
4.3.3 Initial conditions
We want to study how the DDE reacts to the clustering of matter. Thus our initial conditions are of perturbed matter and homogeneous DDE. The matter perturbation
153 is taken as a spherical Gaussian,