HOLOGRAPHY OF DE SITTER SPACE AND DISORDERED SYSTEMS
A DISSERTATION SUBMITTED TO THE DEPARTMENT OF PHYSICS AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
George Konstantinidis Coss May 2015
© 2015 by George Konstantinidis Coss. All Rights Reserved. Re-distributed by Stanford University under license with the author.
This work is licensed under a Creative Commons Attribution- Noncommercial 3.0 United States License. http://creativecommons.org/licenses/by-nc/3.0/us/
This dissertation is online at: http://purl.stanford.edu/zp606yx6742
ii I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.
Stephen Shenker, Primary Adviser
I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.
Sean Hartnoll
I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.
Eva Silverstein
Approved for the Stanford University Committee on Graduate Studies. Patricia J. Gumport, Vice Provost for Graduate Education
This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file in University Archives.
iii Abstract
This dissertation is devoted to the study of models that capture the intricate dy- namics of two different physical setups: exponentially expanding universes and dis- ordered systems. We explore late time divergences in the perturbative corrections of wavefunctions of interacting light fields on a fixed de Sitter background. The di- vergences are holographically interpreted as shifts in the conformal weights of dual CFT operators. We then compute functional determinants in a Euclidean CFT for various non-perturbative deformations. According to the dS/CFT correspondence, these functional determinants calculate the late time Hartle-Hawking wavefunctional of asymptotically de Sitter space in higher spin gravity as a function of the profile of the fields in the bulk. Numerical experiments suggest that upon fixing the average of the bulk scalar profile, the wavefunction becomes normalizable in all the other (infi- nite) directions of the deformations we study. For disordered systems, we investigate the extent to which quiver quantum mechanics models encode the complex dynamics of multicentered black holes in string theory. In a certain limit of the quiver system we display the emergence of a conformal symmetry, which mimics the emergence of conformal symmetry in the near-horizon geometries of extremal black holes. Finally, we take a Newtonian multiparticle limit of the quiver system away from the confor- mal regime. We study the dynamics of the system numerically to look for signs of ergodicity breaking, cages, and transitions to chaos.
iv Acknowledgements
I would like to first thank Dionysios Anninos. Dio, thank you for being patient, sup- portive, inspirational and fun, and thank you for teaching me so many great lessons about life and physics, but most of all thank you for believing in me from day one. This dissertation would not have existed without the support of my parents Con- stantine and Sofi, my brother Lazaros, and my grandparents Alice, Calliope, George, and Lazaros. I am very grateful to my thesis advisor Stephen Shenker for his sound and valuable advice in helping me find my way and graduate, as well as funding my research. I thank Sean Hartnoll for his unwavering support and for making me feel I can always count on him. I am also grateful to other faculty who provided guidance and funding including Savas Dimopoulos and Eva Silverstein. I thank my collaborators, Tarek Anous, Paul de Lange, Frederik Denef, Daniel Freedman, Raghu Mahajan and Edgar Shaghoulian. For all their help along the way, I thank Abdul- lah Al Akeel, Konstantinos Askitopoulos, Taylor Barrella, Kurt Barry, Nathan Ben- jamin, Leslie Brian, Simon Foreman, Abe Fornis, Daniel Fragiadakis, Andrea Fuentes, Kostas Gounaras, Xinlu Huang, George Karakonstantakis, Myrto Karapiperi, Tina Kechakis, Panagiotis Kontogiannis, Stefanos Koulandrou, Dimitris Koutsogiorgas, Lampros Lamprou, Petros Lekakis, Dafni Leon, Anna Liaroutsou, Chelsea Liekhus- Schmaltz, Emily Livermore, Jonathan Maltz, Nikolas Manetas, Cecilia Michel, the Nakas family, Chaya Nanavati, Giannis Pagkalos, Ashley Perko, Giannis Platis, Elise Potts, Djordje Radicevic, Michalis Rossides, Verina Samoli, Lauren Smith, Douglas Stanford, Athina Taka, Filippos Theodorakis, Basak Tulga, Antonis Tzouanakos, Christina Varvia, Dimitris Vrodamitis, Tim Weiser, and Vivian Wong.
v Contents
Abstract iv
Acknowledgementsv
1 Introduction1
2 Wavefunctions in de Sitter space4 2.1 Introduction...... 4 2.2 The Schr¨odingerequation in a fixed de Sitter background...... 7 2.2.1 Bunch-Davies wavefunction...... 8 2.2.2 A Euclidean AdS approach...... 9
2.2.3 Interaction corrections to ZAdS ...... 12 2.3 Self-interacting scalars in four-dimenions...... 13 2.3.1 Tree level contributions for the massless theory...... 13 2.3.2 Loop correction to the two-point function...... 15 2.3.3 Tree level contributions to the conformally coupled case.... 17 2.3.4 Comments for general massive fields...... 19 2.4 Gauge fields and gravity in four-dimensions...... 21 2.4.1 SU(N) gauge fields...... 21 2.4.2 Scalar QED...... 23 2.4.3 Gravity...... 24 2.5 3d CFTs and (A)dS/CFT...... 26
2.6 dS2 via Euclidean AdS2 ...... 29 2.6.1 Tree level corrections for the massless theory...... 30
vi 2.6.2 Loop corrections for the massless theory...... 31 2.7 Outlook...... 33
3 Higher Spin dS/CFT 35 3.1 Introduction...... 35 3.2 Wavefunctionals and the free Sp(N) model...... 40 3.2.1 Wavefunctional...... 41 3.2.2 Functional determinant for radial deformations...... 42 3.3 Simple examples of radial deformations...... 43 3.3.1 Single Gaussian...... 44 3.3.2 Gaussian Ring...... 45 3.3.3 Double Gaussian...... 47 3.4 Radial deformations of flat R3 and pinching limits...... 48 3.4.1 Balloon Geometry...... 48 3.4.2 Conformal Flatness of Balloon Geometry...... 50 3.4.3 Wavefunctions and balloon geometries...... 51 3.5 Spherical harmonics and a conjecture...... 52 3.5.1 Three-sphere harmonics...... 52 3.6 Three-sphere squashed and massed...... 57 3.6.1 Squashed and massed...... 57 3.7 Basis change, critical Sp(N) model, double trace deformations.... 60 3.7.1 Perturbative analysis on R3 ...... 61 3.7.2 Large N saddles for uniform S3 profiles...... 62 3.7.3 Double trace deformations as convolutions...... 64 3 3.7.4 Euclidean AdS4 with an S boundary...... 65 3.8 Extensions of higher spin de Sitter holography?...... 66
3.8.1 AdS4 ...... 66
3.8.2 dS4 ...... 67
4 Conformal Quivers 70 4.1 Introduction...... 70 4.2 General Framework: Quiver quantum mechanics...... 74
vii 4.2.1 Full quiver theory...... 74 4.2.2 Some properties of the ground states...... 77 4.3 Coulomb branch and a scaling theory...... 78 4.3.1 Supersymmetric configurations...... 80 4.3.2 Scaling Theory...... 81 4.4 Conformal Quivers: emergence of SL(2, R)...... 83 4.4.1 Conditions for an SL(2, R) invariant action...... 84 4.4.2 Two particles...... 85 4.4.3 SL(2, R) symmetry of the full three particle scaling theory.. 86 4.4.4 Wavefunctions in the scaling theory...... 89 4.5 Melting Molecules...... 92 4.5.1 Two nodes - melting bound states...... 93 4.5.2 Three nodes - unstable scaling solutions...... 98 4.6 Outlook...... 100 4.6.1 Random Hamiltonians and emergent SL(2, R)?...... 100 4.6.2 Holographic considerations...... 101 4.6.3 Possible extensions...... 102
5 Supergoop Dynamics 105 5.1 Introduction...... 105 5.1.1 String glasses...... 105 5.1.2 Supergoop...... 107 5.1.3 Dynamics...... 109 5.2 General Framework...... 111 5.2.1 Supersymmetric multiparticles...... 111 5.2.2 Classical features and multicentered black holes...... 113 5.2.3 Three particles...... 115 5.2.4 Regime of validity...... 116 5.3 Classical Phase Space...... 117 5.3.1 Two particles are integrable...... 117 5.3.2 Three particles are chaotic...... 118
viii 5.4 Euler-Jacobi Ground States...... 119 5.4.1 Euler-Jacobi three body problem...... 119 5.4.2 Classical Ground States...... 120 5.4.3 Quantum Ground States...... 122 5.5 Euler-Jacobi Dynamics: classical integrability...... 125 5.5.1 Setup and coordinate systems...... 126 5.6 Beyond Euler-Jacobi: the stringy double pendulum...... 128 5.6.1 Collinear dynamics...... 129 5.6.2 Poincar´eSections...... 130 5.7 Trapping...... 133 5.7.1 Setup and Energetics...... 133 5.7.2 A trap...... 134 5.7.3 Topology of the potential landscape...... 135 5.8 Holography of Chaotic Trajectories?...... 135
A De Sitter Wavefunction 139 A.1 A quantum mechanical toy model...... 139 A.1.1 Path integral perturbation theory...... 140 A.2 The bulk-to-bulk propagator...... 143
B Higher Spin de Sitter 146 B.1 Conformal Transformation from round S3 to flat R3 ...... 146 B.1.1 Coordinate transformation and Weyl rescaling...... 146 B.1.2 Numerical Error...... 147 B.1.3 Balloon Geometries...... 147 B.2 Review of the squashed sphere...... 148 B.3 Perturbative Bunch-Davies modes for m2`2 = +2...... 149
B.3.1 Continuation from Euclidean AdS4 ...... 150 B.3.2 Wavefunctional...... 151 B.4 Wavefunctionals for bulk gauge fields...... 152
ix C Conformal Quivers 155 C.1 Notation...... 155 C.2 Superpotential corrections of the Coulomb branch...... 156 C.3 Finite D-terms...... 157 C.3.1 Non-linear D-term solutions for two-node quivers...... 158 C.4 Three node Coulomb branch...... 160 C.4.1 Second order Lagrangian for three-node quiver...... 161 C.4.2 Coulomb branch potential...... 166 C.5 Thermal determinant...... 167
D Supergoop Dynamics 169 D.1 Two-Body Problem...... 169
Bibliography 172
x List of Tables
xi List of Figures
2.1 Witten diagrams for the order λ contributions to ZAdS[ϕ~k, zc] in the theory with φ4 interaction...... 13
3.1 Examples of the radial deformations (3.3.1) on the left, (3.3.2) in the middle and (3.3.3) on the right. We have suppressed the polar coordi- nate θ of the S2 but kept the azimuthal direction...... 44 2 3.2 Plot of |ΨHH (λ, A)| for N = 2 for the Gaussian profile (3.3.1) with
λ = 1 using lmax = 45. The solid blue line is an interpolation of the numerically determined points (shown in red). The wavefunction grows and oscillates in the negative A direction...... 45 2 3.3 Left: Density plot of |ΨHH (λ, a, A)| for N = 2 for the profile (3.3.2) as
a function of A (vertical) and λ (horizontal) for a = 5 using lmax = 45. Again, the wavefunction grows and oscillates in the negative A and 2 positive λ directions. Right: Plot of |ΨHH (λ, a, A)| for the profile (3.3.2) as a function of λ for A = −0.022 and a = 5...... 46 2 3.4 Left: Density plot of |ΨHH (λi, ai,Ai)| for N = 2 for the double Gaus-
sian profile (3.3.3) as a function of A1 (x-axis) and A2 (y-axis) for
a1 = 0, a2 = 5, λ1 = λ2 = 1 using lmax = 45. The wavefunction
grows and oscillates for negative A1 and A2. Right: Density plot of 2 |ΨHH (λi, ai,Ai)| for the double Gaussian profile (3.3.3) as a function
of λ1 (x-axis) and λ2 (y-axis) with A1 = −1, A2 = −1/100, a1 = 0 and
a2 = 5...... 47 3.5 The “balloon” deformation of R3, defined by (3.4.1) and (3.4.3), rep- resented schematically for positive ζ...... 49
xii 3.6 Plot of g(x) (left) and dg(x)/dx (right) as obtained by numerically solving equation 3.4.4( ζ∗ = 1741.51, ζ = 9ζ∗/10, γ = 1.36612, a = −95, λ = 30)...... 50 2 3.7 Density plot of the |ΨHH (ζ, m)| for N = 2 as a function of the pinch- ing parameter ζ (vertical axis) and an overall mass deformation m
(horizontal axis) using lmax = 45, ζ∗ = 1741.51, γ = 1.36612, a = −95 and λ = 30...... 53 2 ∗ 3.8 Left: |ΨHH (ζ, m)| for N = 2 as a function of m for ζ = −ζ /2 (red dots), ζ = 0 (blue squares), ζ = ζ∗/2 (green diamonds) and ζ = 9ζ∗/10
(black triangles) using lmax = 45, ζ∗ = 1741.51, γ = 1.36612, a = −95 and λ = 30. Right: Same as left but for different plot range...... 53 2 3.9 Left: Plot of |ΨHH (c1)| for N = 2 for the first harmonic mapped to 3 2 R given in (3.5.2) using lmax = 45. Right: Plot of log |ΨHH (c1)| for
N = 2 using lmax = 45...... 55
3.10 Plot of log |ΨHH (ck)| for N = 2 for the first nine spherical harmonics 3 mapped to R given in (3.5.2) using lmax = 45. Notice that only the
zeroth harmonic is non-normalizable in the negative c0 direction.... 55 2 3.11 |ΨHH (A)| (left) and log|ΨHH (A)| (right) for N = 2 as a function of A, the overall size of the radial deformation in (3.5.4) which is constructed
to be orthogonal to the zero mode of the three-sphere (lmax = 45)... 56 2 3.12 Plot of |ΨHH [mS3 ]| given by expression (3.6.1) for N = 2...... 58 2 3.13 Left: Density plot of |ΨHH [ρ, mS3 ]| for N = 2. The fainter peak centered at the origin reproduces perturbation theory in the empty de
Sitter vacuum. Horizontal lines are lines of constant mS3 and vertical 2 lines are lines of constant ρ. Right: Plot of |ΨHH [ρ, −2.25]| for N = 2. Notice that it peaks away from ρ = 0...... 59 2 3.14 Left: Density plot of |ΨHH [ρ, mS3 ]| for N = 2 for a slightly larger range. Notice the local de Sitter maximum visible in figure 3.13 is al-
ready too faint to be seen. Horizontal lines are lines of constant mS3
and vertical lines are lines of constant ρ. Right: Plot of 2 log |ΨHH [ρ, −4.5]| for N = 2...... 60
xiii 3.15 Plots of the Nth root of |ΨHH [˜σ, Σ0]|, |ΨHH [˜σ, Σ1]|, |ΨHH [˜σ, Σ2]| at large f from left to right. Notice that the higher saddles dominate nearσ ˜ = 0 but fall off faster for largeσ ˜...... 63 th O(N) 3.16 Plot of the N root of the finite part of Zcrit [σA] for a uniform
source σA over the whole three-sphere. We have normalized such that O(N) Zcrit [σA] = 1 at σA = 0...... 66
4.1 A 3-node quiver diagram which captures the field content of the La-
grangian L = LV + LC + LW , each piece of which is given in (4.2.1), (4.2.2), and (4.2.5). This quiver admits a closed loop if κ1, κ2 > 0 and κ3 < 0...... 75 ˜ 4.2 Examples of H + a K eigenfunctions. Left: Plot of ψλ(x) for n ∈ ˜ (0.2, 5.2) in unit increments. Right: Plot of ψλ(x) for n ∈ (−5.7, −0.7) in unit increments...... 91 4.3 Thermal effective potentials of a two node quiver (θ = −1 and µ = 1). As the temperature is increased the system explores various thermal configurations of stable and metastable minima. From top left to bot- tom right the system is of type1 → 2 → 3 → 2 → 4a → 5...... 96 4.4 Thermal effective potentials of a two node quiver (for θ = −1 and µ = 1). Left: An example of phase type 4b. Right: A case where the potential of the supersymmetric minimum decreases as the tempera- ture is increased. A similar observation was made for supersymmetric bound states in [208]...... 97 4.5 Thermal effective potentials of a two node quiver (θ = −1 and µ = 1). Asκ ˜ is increased we note that the first minimum disappears...... 99 4.6 Thermal potential along the scaling direction |qi| = λ κi for κ1 = κ2 = −κ3 = 1 at T = 0 (left) and T 6= 0 (right)...... 100 4.7 2-loop Feynman diagrams contributing to the δxδx term of the effective Lagrangian. Solid lines represent φ, while dotted lines correspond to ψ propagators...... 103
xiv 4.8 A schematic representation of a system in a mixed Higgs-Coulomb branch. The long arrows represent very massive strings. Note that
there is a closed loop connecting Γ1,Γ2 and Γ3...... 104
5.1 Examples of ground states for 100 electric Γe = (0, 1) plus 100 magnetic
Γm = (1, 0) particles...... 114
5.2 Left: Classical moduli space M in the δ1 − δ2 plane for θ3 6= 0. The nature of M for the different regions is shown in figure 5.3. Right:
Classical moduli space for M with θ3 = 0. In regions i and ii the centers at z = a and z = −a are enclosed respectively...... 122
5.3 Classical moduli space in the δ1 − δ2 plane for θ3 6= 0. The order of the figures left to right starting at the top are the regions in figure 5.2... 123 5.4 Three node quiver with a closed loop (left) and without a closed loop (right)...... 125 5.5 The Euler-Jacobi flower. The red balls represent the fixed background centers and the blue line represents the classical trajectory of the probe. In this case, the trajectory precesses around only one of the fixed centers.126 5.6 Examples of closed phase space trajectories in the integrable probe regime. The plots show slices of phase space in the Cartesian coordi- nate system...... 128 5.7 Examples of open phase space trajectories in the chaotic regime. The plots show slices of phase space in the Cartesian coordinate system.. 129
xv 5.8 Poincar´esections of collinear setup with κ31 = 10, κ32 = −10, κ21 =
−10 and θ3 = −1, θ2 = 1 and energies E = {0.10, 0.20, 0.23, 0.26, 0.27, 0.30}. Note that the κ’s form a closed loop. The horizontal axis represents the position of particle 3 while the vertical axis represents its conjugate momentum. Any given plot is produced by varying the initial positions and momenta of the two probes subject to a fixed total energy. The
pair (x3(t), p3(t)) is plotted every time the resulting trajectory of parti-
cle 2 crosses some fiducial point (x2(t) = xc) with positive momentum
(p2(t) > 0), i.e. roughly every time particle 2 completes a full cycle as it oscillates back and forth. In the quasi-integrable regime, different initial conditions correspond to different contours. The first Poincar´e section shows a quasi-integrable behavior with two fixed points corre- sponding to the two low energy normal modes...... 132 5.9 The first row represents a low energy probe, which remains stuck in a subset of the phase space for seemingly arbitrarily long times. The second row represents an intermediate energy probe which illustrates the non-uniform escapes that occur from the low energy trapping be- havior. We see that for a while it remains trapped in some subset of phase space, after which it escapes and gets stuck in some other subset of phase space. The final row represents a high energy probe which uniformly explores the molecule. The associated plots represent the percentage of the molecule explored as a function of the integration time, up to 15000 time steps in increments of 1500. The tapering off of the high energy probe is simply due to saturating the entire molecule. Below these points the increase is very uniform. The initial energy increases from 50% of the escape energy in the first row to 60% of the escape energy in the third row. These percentages, however, are very dependent on the parameters (e.g. κ, θ, etc.) in the problem...... 136
xvi 5.10 These contour plots show equipotential surfaces in the plane of a 2D molecule consisting of one hundred centers. From left to right, we have chosen κ = 1, κ = 1.5, κ = 3.5, and in all cases θ = −10. We observe that as the magnitude of κ increases, the minima (blue region), which initially lied near each center, are collectively expelled, forming an overall minimum that surrounds the molecule as a whole. For κ = 1, the trajectory remains close to the plane of the molecule and has been superimposed on the left contour plot (transparent white line). The axes label the x and y positions of the probe particle...... 137
2 B.1 Left: Comparison of |ΨHH (mS3 )| for N = 2 as obtained by calculating
ZCFT [mS3 ] analytically (blue line) and given in equation (3.6.1), and by numerically evaluating the functional determinant using the Dunne-
Kirsten regularization method with lmax = 45 (red dots). We have 2 normalized such that |ΨHH (mS3 )| = 1 at mS3 = 0. Right: Plot of the
percentage error as a function of the numerical cutoff (at mS3 = −2.2). 147
C.1 Example of Feynman diagram contributing to DiDj from the superpo- tential...... 157
C.2 Plot of D0 for µ = −θ = κ = 1. Notice that the solution is real for all values of |q|...... 159
C.3 Left: Plot of <[D1] (blue) and <[D2] (violet). Right: Plot of =[D1]
(blue) and =[D2] (violet). Both plots are for µ = −θ = κ = 1. Notice that when complex, the solutions form a conjugate pair...... 159
C.4 Left: Plot of V evaluated on D0 for µ = −θ = κ = 1. Right: Plot of V evaluated on the perturbative D solution in violet, compared with
the full non-perturabative D0 in blue...... 160 2 C.5 1-loop Feynman diagrams contributing to the D1 term of the effective Lagrangian...... 162
C.6 1-loop Feynman diagram contributing to the D1D2 term of the effective Lagrangian...... 163
xvii C.7 1-loop Feynman diagrams contributing to the δx1δy2 term of the effec- tive Lagrangian...... 163
C.8 1-loop Feynman diagrams contributing to the δx1δx2 term of the effec- tive Lagrangian...... 164 a b C.9 1-loop Feynman diagrams contributing to the δx1 δx1 term of the effective Lagrangian...... 165
D.1 Scatering angle...... 171
xviii Chapter 1
Introduction
Many experimental phenomena at very small scales are accurately described theoreti- cally using quantum field theory. As the name suggests, this theory assumes the world is composed of fluctuating fields which are quantized. Two simple examples are the electron and photon fields with their associated quanta. However, attempts to quan- tize the important gravitational field are obstructed by the non-renormalizability of the Einstein-Hilbert action. Renormalizability is an indication of our ability to use a theory to make predictions at smaller scales and higher energies. Therefore, the non- renormalizability of gravity implies that, at sufficiently high energies, we are unable to make any theoretical predictions. Nevertheless, the quantization of the gravitational field is a pressing problem. A consistent quantum theory of gravity will hopefully shed light on poorly under- stood experimental phenomena and theoretical paradoxes. These include the physics of black holes, dark energy, and the beginning and ultimate fate of our universe. The difficulties in quantizing gravity indicate that, at higher energies, new physics is required to provide a complete theoretical description. String theory is a possible completion. More technically speaking, string theory was initially understood as perturbative excitations of strings. Recently, a non-perturbative definition of string theory was provided in anti-de Sitter space by postulating its equivalence to a conformal field 5 theory of lower dimension. One example relates string theory on AdS5 × S to N = 4
1 CHAPTER 1. INTRODUCTION 2
super-Yang Mills theory, a supersymmetric, conformal gauge theory. The anti-de Sitter space/conformal field theory correspondence and the more general principle of relating theories of gravity to lower dimensional gauge theories appear under different names including AdS/CFT correspondence, holography, gauge/gravity duality and holographic duality. This dissertation will be focused on using the holographic principle to gain a better theoretical handle of a model of an expanding universe called de Sitter space, and disordered/glassy systems. For the former, our primary tool will be the conjectured de Sitter/conformal field theory (dS/CFT) duality, which is inspired by AdS/CFT. For disordered systems, we will use the fact that certain quantum mechanics models can capture aspects of the complex dynamics of bound black hole configurations. Experiments indicate that our universe experienced an exponential expansion at cosmologically early times called inflation, and is also currently undergoing an accel- erated expansion. De Sitter space is the maximally symmetric geometry satisfying Einstein’s equations with a positive cosmological constant and plays a significant role in the theory of inflationary cosmology. Recent astronomical observations indicate our universe is entering a new asymptotically de Sitter phase, with a small positive value for the cosmological constant. Therefore, de Sitter space is more relevant to the evolution of our cosmos as it describes an expanding universe with positive cosmologi- cal constant, as opposed to anti-de Sitter space which requires a negative cosmological constant. In this context, one important problem is to understand the dynamics of quantum fields on the time-dependent de Sitter background. In Chapter2, we study a variety of fields in de Sitter space, and investigate the late-time evolution of their wave- functions. We take a new approach at perturbatively analyzing well-known infrared divergences of light scalar fields, and use holography to argue against the existence of such divergences in pure Einstein theory. In Chapter3, we explore the validity of a conjectured holographic duality, be- tween a Euclidean conformal field theory and a higher spin theory of gravity, called dS/CFT. In this case, the partition function of the CFT, which we find by numerically computing a functional determinant, is thought to be the same as the wavefunction CHAPTER 1. INTRODUCTION 3
of the gravity theory. Beyond the, global, wavefunction approach, one may want to use holography to understand an individual static patch observer in de Sitter space. The worldline data of a static patch observer is characterized by the emergence of an SL(2, R) symme- try. Interestingly, the static patch of four-dimensional de Sitter space is conformally 2 equivalent to AdS2 × S , whose SL(2, R) does not seem to reside within a larger structure containing a Virasoro algebra. It is unclear and worth investigating what a holographic dual may look like in cases when there is an SL(2, R) symmetry without a full Virasoro.
Motivated by these observations, and the additional importance of AdS2 as the near-horizon geometry of extremal black holes, in Chapter4 we consider the extent to which a quiver quantum mechanics system holographically captures the dynamics of AdS2 geometries and extremal black holes. More specifically, in a certain limit of the model, we show the emergence of an SL(2, R) symmetry. In Chapter5, we will consider multiparticle configurations of the quiver quantum mechanics system mentioned above, in order to study the low-energy dynamics of extremal black hole bound states. Stable bound states of extremal black holes have long relaxation times and a seemingly complicated free energy landscape. These general characteristics are notably shared by many disordered systems. Besides being poorly understood, disordered/glassy systems display many interest- ing phenomena. For example, when supercooled to form a glass, liquids often display diverging viscosities. Additionally, these systems are believed to break ergodicity. The question we will investigate is whether the Newtonian limit of a multiparticle quiver quantum mechanics system captures the complex dynamics of a disordered system, or at least some aspects of it. The work presented is based on the papers [1–4]. Chapter 2
Wavefunctions in de Sitter space
2.1 Introduction
The geometry of the inflationary epoch of our early universe was approximately de Sitter [5–9], and our universe is currently entering a de Sitter phase once again. It is thus of physical relevance to examine how to deal with quantum effects in a de Sitter universe. Such issues have been studied heavily in the past. The technical aspects of most calculations have involved the in-in/Schwinger-Keldysh formalism which is reviewed in [10], and focus on computing field correlations at a fixed time. Indeed, in the context of quantum cosmology we are interested in correlations of quantum fields at a given time rather than scattering amplitudes—which condition on events both in the far past as well as in the far future. A complementary approach is to build a perturbation theory for solutions of the Schr¨odingerequation itself. Knowledge of the wavefunction allows us to consider ex- pectation values of a broad collection of observables, which in turn permits a richer characterization of the state [11]. Thus, an understanding of the wavefunction and its time evolution is of interest. Although generally complicated, there is one particular solution of the Schr¨odingerequation in a fixed de Sitter background which exhibits a simplifying structure. This solution is the Bunch-Davies/Hartle-Hawking wavefunc- tion ΨBD [12–15], and its form strongly resembles that of the partition function in a Euclidean AdS background upon analytic continuation of the de Sitter length and
4 CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 5
conformal time. This observation led to the elegant proposal of a close connection between dS and Euclidean AdS perturbation theory in [16] (see also [17,18]). It is our goal in this chapter to exploit the connection between dS and AdS to develop a more systematic perturbative framework for the construction of this wave- function. We do this by considering a series of examples. The perturbative frame- work in an AdS spacetime has been extensively studied in the past [19–21] and is our primary calculational tool. Our examples involve self-interacting scalar fields, both massless and massive, as well as gauge fields and gravitons. We recast many of the standard issues involving infrared effects of massless fields1 in the language of the wavefunction. Many of these infrared effects exhibit correlations that grow logarithmically in the scale factor, as time proceeds and we display how such effects appear in the wavefunction itself. It is worth noting that though most calculations of ΨBD involve taking a late time limit, our approach requires no such limit and we construct ΨBD perturbatively for any arbitrary time. For massless scalar fields, the finite time dependence of the wavefunction at tree level is captured by the exponen- tial integral function Ei(z), whose small argument behavior contains the logarithmic contributions. An interesting difference between the approach described here and the in-in for- malism is that the two approaches use different propagators. For a massless scalar in 2 Euclidean AdS4, we use the Green’s function:
1 0 G (z, z0; k) = − (1 − kz)(1 + kz0)ek(z−z ) AdS 2k3L2 2kzc 0 e (1 − kz )(1 + kz)(1 + kz ) 0 − c e−k(z+z ) , (2.1.1) (1 + kzc) valid for z < z0. (For z0 < z, one simply exchanges the two variables.) The math- ematical purpose of the second term is to enforce the Dirichlet boundary condition at the cutoff zc. It is perhaps more significant physically that the sum of the two terms is finite as k → 0. Thus, loop integrals using (2.1.1) do not produce infrared
1See [22–38] for an incomplete list of references on the topic of infrared issues in de Sitter space. 2The Euclidean AdS metric is ds2 = L2(dz2 + d~x2)/z2 and we work in momentum space. CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 6
divergences at small k. The Green’s functions considered in the in-in formalism [37] are obtained by the continuation to dS4 of the first term in square brackets. Its real part gives
1 G (η, η0; k) = (1 + k2ηη0) cos[k(η − η0)] + k(η − η0) sin[k(η − η0)] , (2.1.2) C 2k3`2 which is singular as k → 0. One of the main motivations of our approach is to connect our results with the idea [16,39,40] that Ψ BD (at late times) is holographically computed by the partition function of a conformal field theory. If this correspondence, known as the dS/CFT correspondence, is indeed true,3 infrared effects in de Sitter spacetime should be related to quantities in the putative conformal field theory itself. This could lead to a better understanding of possible non-perturbative effects. Moreover, in analogy with how the radial coordinate in AdS is related to some (as of yet elusive) cutoff scale in the dual CFT [46–49], it is expected that the scale factor itself is connected to a cutoff scale in the CFT dual to de Sitter space [50–54]. Our calculations may help elucidate such a notion. Of further note, having a better understanding of ΨBD at finite times allows us to compute quantum expectation values of fields within a single cosmological horizon, rather than metaobservables inaccessible to physical detectors.4 We begin in section 2.2 by explaining how solutions to the Schr¨odingerequa- tion can be captured by a Wick rotation to Euclidean time, hence establishing the connection between de Sitter and anti-de Sitter calculations. We then proceed in section 2.3 to examine a self-interacting scalar field with φ4 interactions in a fixed four-dimensional de Sitter background, whose contributions to the wave function con- tain terms that depend logarithmically on the conformal time η. In section 2.4 we discuss the case of gauge fields and gravitons. We argue that, to all orders in the tree-level approximation, no logarithms are present for a pure Einstein theory with a
3Recently several concrete realizations of this proposal have emerged [41–43] for theories of four- dimensional de Sitter space involving towers of interacting massless higher spin fields. Aspects of de Sitter holography are reviewed in [44, 45]. 4A complementary approach would be to compute quantities directly in the static patch of de Sitter [59, 60]. CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 7
positive cosmological constant. We discuss our results in the context of holography in section 2.5. Finally, in section 2.6 we go to two-dimensional de Sitter space in order to compute loop effects for a cubic self-interacting massless scalar. In appendix A.1 we set up a quantum mechanical toy model where the mathematics of our calculations is exhibited in a simple context.
2.2 The Schr¨odingerequation in a fixed de Sitter background
The main emphasis of this section is to show that our method perturbatively solves the functional Schr¨odingerequation for a scalar field in the Bunch-Davies state. We will first provide the exact solution for a free field, and then show how the result can be obtained by continuation from Euclidean AdS as in [16]. We then treat interactions perturbatively.
We use conformal coordinates for dS(d+1),
`2 ds2 = −dη2 + d~x2 , ~x ∈ d , η ∈ (−∞, 0) . (2.2.1) η2 R
For simplicity we consider a self-interacting scalar but analogous equations will also hold for other types of fields. The action is:
(d−1) Z Z 2 ` dη 2 2 ` V (φ(η, ~x) SL = d~x (d−1) (∂ηφ(η, ~x)) − (∂~xφ(η, ~x)) − 2 . (2.2.2) 2 Rd |η| η
We specify the potential later, but we envisage the structure of a mass term plus φn interactions. It is convenient to take advantage of the symmetries of Rd and work in momentum space. Thus, we define:
Z ~ dk i~k·~x φ(η, ~x) = d e φ~k(η) . (2.2.3) Rd (2π) CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 8
Henceforth we denote the magnitude of the momentum by k ≡ |~k|. Upon defining the canonical momenta π~k = −iδ/δφ~k conjugate to φ~k, we can write the Schr¨odinger equation governing wavefunctions Ψ[ϕ~k, η] in a fixed dSd+1 background:
2 !! X 1 |η|(d−1) `(d−1) k2 ` π π + ϕ ϕ + V˜ (ϕ ) Ψ[ϕ , η] 2 `(d−1) ~k −~k |η|(d−1) 2 ~k −~k |η| ~k ~k ~k∈Rd
= i ∂ηΨ[ϕ~k, η] . (2.2.4)
The variable ϕ~k is the momentum mode φ~k evaluated at the time η where Ψ is ˜ evaluated. The potential V (φ~k) is the Fourier transform of the original V (φ(η, ~x)); it has the structure of a convolution in ~k-space.
2.2.1 Bunch-Davies wavefunction
In principle, we can construct solutions to (2.2.4) by considering Feynman path inte- grals over the field φ. We are particularly interested in the solution which obeys the Bunch-Davies boundary conditions. This state is defined by the the path integral:
Z Y iS[φ~k] ΨBD[ϕ~k, ηc] = Dφ~k e , (2.2.5) ~k∈Rd
ikη in which we integrate over fields that satisfy φ~k ∼ e in the kη → −∞ limit and
φ~k(ηc) = ϕ~k at some fixed time η = ηc. The natural generalization of this state to include fluctuating geometry at compact slicing is given by the Hartle-Hawking wavefunction. The boundary conditions resemble those defined in the path integral construction of the ground state of a harmonic oscillator. As usual, physical expectation values are given by integrating over the wavefunc- tion squared. For example, the n-point function of ϕ~k, all at coincident time ηc, is: R Q dϕ |Ψ [ϕ , η ]|2 ϕ . . . ϕ ~k ~k BD ~k c ~k1 ~kn hϕ~ . . . ϕ~ i = . (2.2.6) k1 kn R Q 2 ~k dϕ~k |ΨBD[ϕ~k, ηc]|
As a simple example we can consider the free massless field in a fixed dS4 back- ground. In this case we can obtain ΨBD as the exact solution of the Schr¨odinger CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 9
equation (2.2.4):
1/4 Y 2k3 i `2 k2 e−ikηc/2 ΨBD[ϕ~k, ηc] = exp ϕ~k ϕ−~k p . (2.2.7) π 2 ηc(1 − ikηc) (1 − ikηc) ~k
Although ηc can be considered to be an arbitrary point in the time evolution of the state, we are ultimately interested in the late time structure of the wave function.
At late times, i.e. small negative ηc we find:
2 Z ~ 2 ` dk ik 3 log ΨBD[ϕ~k, ηc] = 3 − k ϕ~k ϕ−~k + .... (2.2.8) 2 (2π) ηc
Notice that the small ηc divergence appears as a phase of the wavefunction rather than its absolute value, i.e. it plays no role in the expectation values of the field ϕ~k.
The late time expectation value of ϕ~k ϕ−~k is given by:
1 hϕ ϕ i = , (2.2.9) ~k −~k 2 `2 k3 which diverges for small k. The divergence stems from the fact that ΨBD is non- normalizable for the ~k = 0 mode.
2.2.2 A Euclidean AdS approach
When computing the ground state wavefunction of the harmonic oscillator from the path integral, one wick rotates time and considers a Euclidean path integral with boundary condition in the infinite past. Similarly, for the dS wavefunction, we can continue to Euclidean time z = −iη and consider a Euclidean path integral. Now, the path integral is over configurations that decay in the infinite Euclidean past, defined here as the limit z → ∞. If in addition we continue L = −i`, we see that the calculation becomes that of constructing the Euclidean partition function in a fixed
Euclidean AdS(d+1) background:
L2 ds2 = dz2 + d~x2 , ~x ∈ d , z ∈ (0, ∞) . (2.2.10) z2 R CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 10
In other words we have that: ΨBD[ϕ~k, ηc] = ZAdS[ϕ~k, izc] (with L = −i`) at least in the context of perturbation theory in a fixed (A)dS background. The Euclidean path integral calculation incorporates, in principle at least, both classical and quantum effects. Let us ignore quantum effects temporarily and discuss how AdS/CFT works at the classical level. To be concrete, consider a massive scalar with quartic self-interaction. The classical action is
(d−1) Z Z ∞ 2 2 2 L dz 2 2 m L 2 λ L 4 S = d~x (d−1) (∂zφ) + (∂~xφ) + 2 φ + 2 φ . (2.2.11) 2 d z z 4z R zc
One seeks a solution φ(z, ~x) of the classical equation of motion that satisfies the boundary condition φ(z, ~x) → ϕ(~x) as z → zc. The cutoff zc is needed to obtain correct results for correlation functions in the dual CFT. The classical solution is then substituted back in the action to form the on-shell action Scl[ϕ(~x)] which is a functional of the boundary data. In the classical approximation the partition function
−Scl[ϕ(~x)] is the exponential of the on-shell action i.e. ZAdS = e , and n-point correlation functions of the CFT operator dual to the bulk field φ are obtained by taking n variational derivatives with respect to the sources ϕ(~x). Let us now perform the Euclidean version of the calculation that gives the result (2.2.7). For this purpose we ignore the quartic term in (2.2.11). In ~k-space, we wish to solve the previously mentioned boundary value problem5 captured by the classical equation of motion:
2 2 2 2 2 2 z ∂z − (d − 1)z∂z − (k z + m L ) φ~k(z) = 0 , φ~k(z = zc) = ϕ~k . (2.2.12)
The exponentially damped solution of the ODE involves the modified Bessel function
Kν(kz), and the solution of the boundary value problem can be neatly written as
d/2 √ z Kν(kz) 1 2 2 2 φ~k(z) ≡ K(z; k)ϕ~k = d/2 ϕ~k ν = d + 4m L . (2.2.13) zc Kν(kzc) 2
5There are many useful reviews of the AdS/CFT correspondence, including [55–57]. The present boundary value problem is discussed in Sec. 23.10 of [58]. CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 11
This equation defines the important bulk-to-boundary propagator K(z, k). We follow the procedure outlined above and substitute the solution (2.2.13) into the action (2.2.11). After partial integration the on-shell action reduces to the surface term at z = zc:
1 Z d~k L(d−1) S [ϕ ] = − K(z; k) ∂ K(z; k) ϕ ϕ at z = z . (2.2.14) cl ~k 2 (2π)d z z ~k −~k c
Let’s restrict to the case of a massless scalar in AdS4 which is the case d = 3, ν = 3/2 of the discussion above. The Bessel function simplifies greatly for half-odd-integer index, and the bulk-to-boundary propagator becomes:
(1 + kz)e−kz K(z; k) = −kz . (2.2.15) (1 + kzc)e c
The on-shell action then becomes:
Z ~ 2 2 1 dk L k zc Scl[ϕ~k] = 3 ϕ~k ϕ−~k . (2.2.16) 2 (2π) zc 1 + kzc
To discuss the AdS/CFT interpretation we need to take the small zc limit, which gives: Z ~ 2 1 dk L 2 3 2 3 Scl → 3 k zc − k zc + O(zc ) ϕ~kϕ−~k . (2.2.17) 2 (2π) zc 2 The first term is singular as zc → 0, but the factor k ϕ~kϕ−~k is local in ~x-space. In fact it contributes a contact term δ(~x−~y) in the ~x-space correlation function. Such contact terms are scheme-dependent in CFT calculations and normally not observable. The remaining finite term has the non-local factor k3. It’s Fourier transform gives the 6 observable part of the 2-point correlator, hO3(x)O3(y)i ∼ 1/|~x − ~y| which is the power law form of an operator of scale dimension ∆ = 3. In AdS/CFT a bulk scalar 2 of mass m is dual to a scalar operator O∆ of conformal dimension ∆ = (d/2 + ν). It is more pertinent to discuss the relation between the Lorentzian and Euclidean signature results. In the free Lorentzian theory we can write ΨBD = exp(iSL) . Then upon continuation L → −i`, z → −iη, zc → −iηc, the Euclidean on-shell action CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 12
(2.2.16) and its Lorentzian counterpart are related by
− SE ≡ −Scl → iSL . (2.2.18)
This is the expected relation for field theories related by Wick rotation. Henceforth, the Euclidean signature AdS/CFT correspondence will be our primary method of computation. In this way we will be using a well developed and well tested formalism. After completion of a Euclidean computation, we will continue to de Sitter space and interpret the results as contributions to the late time wave function ΨBD.
2.2.3 Interaction corrections to ZAdS
We now consider the effect of interactions in the bulk action, such as, for example, the φ4 term in (2.2.11). We treat the quantum fluctuations using a background
field expansion φ = φcl + δφ. The classical field satisfies the non-linear classical equation of motion with Dirichlet boundary condition limz→zc φcl(~x,z) = ϕ(~x), while the fluctuation δφ vanishes at the cutoff. The partition function is then: Z − Scl −S[δφ,φcl] ZAdS[ϕ~k, zc] = e Dδφ e . (2.2.19)
Exact solutions of the non-linear classical equation are beyond reach, but the reason- ably efficient perturbative formalism of Witten diagrams leads to series expansions in the coupling constant. For example, φcl = φ0 + λφ1 + ..., where φ0 solves the free equation of motion coming from the quadratic piece of the action and the full Dirichlet boundary condition.6
Witten diagrams without loops contribute to Scl, while those with internal loops appear in the perturbative development of the fluctuation path integral. The basic building blocks of Witten diagrams are the bulk-to-bulk Green’s function G(z, w; k) and the bulk-to-boundary propagator K(z; k). It is significant that G satisfies the
Dirichlet condition at the cutoff, i.e. G(zc, w; k) = G(z, zc; k) = 0. In Appendix
6In Appendix A we present and develop a quantum mechanical toy model. This model is instruc- tive because perturbative computations are quite feasible and their structure is closely analogous to those in our field theories. CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 13
B, these propagators are explicitly constructed for the main cases of interest in this chapter.
Witten diagrams for the order λ contributions to ZAdS[ϕ~k, zc] in the theory with φ4 interaction are depicted in Fig. 2.1.
Fig. 2.1: Witten diagrams for the order λ contributions to ZAdS[ϕ~k, zc] in the theory with φ4 interaction.
2.3 Self-interacting scalars in four-dimenions
4 We now discuss several contributions to ΨBD from interactions, mostly in φ theory.
As previously mentioned, we carry out calculations in Euclidean AdS4, then continue to dS4 by taking z = −iη, zc = −iηc and L = −i` . We use the metric (2.2.10) and action (2.2.11) in d = 3.
2.3.1 Tree level contributions for the massless theory
First we focus on the massless case m2L2 = 0. The relevant bulk-to-boundary prop- agator is given in (2.2.15). The tree-level contribution to ΨBD (left of Fig. 2.1) is captured by the integral:7
4 λ L4 Z ∞ Y dz − K(z; k ) = 8 z4 i zc i=1 4 2 2 2 3 kΣzc 3 λL kΣ + kΣ zc + kΣ (3kπ − kΣ) zc + 3kP zc − e kΣ kΣ3 zc Ei(−kΣzc) − 3 , (2.3.1) 8 3 kΣ zc (1 + k1zc)(1 + k2zc)(1 + k3zc)(1 + k4zc)
7 R ∞ −t The Ei(z) function is defined as Ei(z) = − −z dt e /t. It has a branch cut along the positive real axis of z ∈ C. We are primarily interested in this function along the negative real axis and the negative imaginary axis, both away from the origin. CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 14
where we have defined the following quantities:
4 4 4 4 X X Y X 3 kΣ ≡ ki , kπ ≡ kikj , kP ≡ ki , kΣ3 ≡ ki . (2.3.2) i=1 i=1 i=1 i=1 j=i+1
We expand the result for small values of zc and analyze the divergent structure. In −3 2 −1 −2 summary, we find terms of order ∼ zc as well as ∼ k zc divergence but no ∼ zc 3 term. Furthermore, we find a ∼ k log k zc term.
Upon analytic continuation to dS4 the power law divergences become phases of the wavefunction. On the other hand, the logarithmic term contributes to the absolute value of ΨBD[ϕ~k, ηc] at small ηc:
log |ΨBD[ϕ~k, ηc]| = 4 Z ~ ~ ~ λ ` dk1 dk2 dk3 (kΣ3 log(−kΣ ηc) + ...) ϕ~ ϕ~ ϕ~ ϕ~ , (2.3.3) 24 (2π)3 (2π)3 (2π)3 k1 k2 k3 k4
P ~ where i ki = 0 due to momentum conservation. Thus we encounter contributions to |ΨBD[ϕ~k, ηc]| that grow logarithmically in the late time limit, |ηc| → 0. In fact, at late enough times the correction is no longer a small contribution compared to the λ = 0 pieces, and all subleading corrections will also begin to compete. In this way one recasts several of the infrared issues encountered when studying massless fields in the in-in/Schwinger-Keldysh formalism [10, 24]; now from the viewpoint of the wavefunction. Similar logarithmic terms are present at tree level in a cubic self-interacting mass- less theory, and their effect was noted in the context of non-Gaussian contributions to inflationary correlators in [64]. In this case one finds the late time correction:
4 Z ~ ~ λ ` dk1 dk2 log |ΨBD[ϕ~ , ηc]| = − (kΣ3 log(−kΣ ηc) + ...) ϕ~ ϕ~ ϕ~ , (2.3.4) k 6 (2π)3 (2π)3 k1 k2 k3
P ~ where i ki = 0 , and kΣ, kΣ3 are defined as in (2.3.2). In the case of slow roll inflation, these infrared effects are suppressed by the small slow roll parameters [16]. CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 15
As was mentioned in the introduction, the dS/CFT proposal connects ΨBD to the partition function of a conformal field theory. Here, one envisions some theory in de Sitter space that contains such light scalars in its spectrum, including the graviton (dual to the stress tensor of the CFT) and so on. In section 2.5 we will explore this connection and in particular, discuss a possible holographic interpretation of such divergences based on recent analyses of 3d CFT’s in momentum space [62,63].
2.3.2 Loop correction to the two-point function
It is of interest to understand the late time structure of loop corrections in the φ4 model. We will calculate the diagram on the right in Fig. 2.1, which corresponds to the following integral:
4 Z ∞ Z 3 λ L dz 2 d~p Iloop(k, zc) = − 4 K(z; k) 3 G(z, z; p) . (2.3.5) 4 zc z (2π)
To render the ~p-integral finite we must impose an ultraviolet cutoff. Recall that ~p is a coordinate momentum, such that the physical (proper) momentum at a given z is given by ~pph = z ~p/L . We impose a hard cutoff on ~pph, such that the ultraviolet cutoff of ~p is z-dependent, i.e. |~pUV | = |ΛUV L|/z . A large ΛUV expansion reveals 8 terms that diverge quadratically and logarithmcally in ΛUV :
3 λ L2 −|Λ L|2 + 2 log |Λ L| . (2.3.6) 8(2π)2 UV UV
To cancel the quadratic divergence, we can add a local counterterm:
Z dz d~k 3 λ L2 δ φ (z) φ (z) , δ = |Λ L|2 . (2.3.7) z4 (2π)3 ~k −~k 8(2π)2 UV
8It is worth comparing the divergence structure in (2.3.6) to a coincident point expansion of 1 3 the SO(4, 1) invariant Green’s function: G(u) ∼ L2 (2/u) F (3, 2, ; 4; −2/u). The argument u = [(z − z0)2 + (~x − ~x0)2]/2zz0 is an SO(4, 1) invariant variable. Near u = 0, we write z = z0 and 0 1 2 2 ~x = ~x + ~. In this limit G(u) ∼ L2 [−z / + 2 ln(/z) + ...] . This is precisely of the form (2.3.6), although the divergence is cut off by the physical length ~xph,UV = ~L/z. What we are suggesting is that the physical cutoff is a de Sitter invariant cutoff. CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 16
Upon addition of the counterterm, the ~p-integral can be performed analytically ren- dering an expression containing the Ei(z) function that is only logarithmically diver- gent in |ΛUV L|. The remaining z-integral is complicated, but we are mainly interested in its small zc behavior, which we can extract. We find the following terms divergent in the small zc expansion (to leading order in ΛUV ):
2 2 λ L 1 3k 3 − 2 log|ΛUV L| − 3 + + k log(zc k) , (2.3.8) 2(2π) zc 2zc
The logarithmic term contributes to the absolute value of the wavefunction upon analytic continuation to dS4:
`2 Z d~k λ log |Ψ [ϕ , η ]| = − k3 1 − log |Λ `| log(−η k) + ... ϕ ϕ . BD ~k c 2 (2π)3 (2π)2 UV c ~k −~k (2.3.9)
Notice that at late times, the width of the |ΨBD[ϕ~k, ηc]| for a fixed k mode narrows, which is physically sensible as the quartic part of the potential dominates compared to the kinetic term. To order λ, the “cosmological two-point correlation function” can be obtained from this wave function (including the contribution from (2.3.3)) via the general expression (2.2.6). The result closely resembles the late time two-point function computed, for example, in [37]. Notice that there is no need to impose an infrared cutoff when considering loop corrections of the wavefunction itself. As a final note, we could have also considered a slightly different subtraction where our counterterm also removes the logarithic divergence in |ΛUV L|. Evaluation of the integrals proceeds in a similar fashion leading to the following result upon continuation to dS4:
`2 Z d~k λ log |Ψ [ϕ ]| = − k3 1 + a k3 log(−η k) + ... ϕ ϕ , BD ~k 2 (2π)3 1 (2π)2 c ~k −~k (2.3.10) where a1 = −(−5 + 4γE + 4 log 2)/4 ≈ −0.02 . The result is now independent of the ultraviolet cutoff altogether. CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 17
2.3.3 Tree level contributions to the conformally coupled case
We now analyze a conformally coupled scalar in a fixed Euclidean AdS4 background with m2L2 = −2. This case is of particular interest as it arises in the context of higher spin Vassiliev (anti)-de Sitter theories [65, 66]. The bulk-to-boundary propagator simplifies to: z K(z; k) = ek(zc−z) , (2.3.11) zc and the free quadratic on-shell classical action is given by:
L2 Z d~k k 1 Scl = 3 ϕ~k ϕ−~k 2 − 3 . (2.3.12) 2 R3 (2π) zc zc
n For the sake of generality, we consider a self-interaction of the form λnφ(~x,z) /2n n with n = 3, 4,... For such a theory, the order λn tree level (ϕ~k) contribution requires computing integrals of the form:
Z ∞ dz z n 1 kΣ(zc−z) kΣ zc In(ki, zc) = 4 e = 3 e E(4−n)(kΣ zc) , (2.3.13) zc z zc zc
9 where En(z) is the exponential integral function and kΣ ≡ k1 + k2 + ... + kn. Ex- panding the integral reveals that logarithms will only occur in the small zc expansion for the case n = 3. For n = 3 we find the following small zc expansion:
γE + log(kΣzc) kΣ(−1 + γE + log(kΣzc)) In=3(ki, zc) = − 3 − 2 zc zc 2 kΣ(−3 + 2γE + 2log(kΣzc)) 1 3 − − kΣ(−11 + 6γE + 6log(kΣzc)) . (2.3.14) 4zc 36
9 R ∞ −zt n The function En(z) = 1 dte /t for z ∈ C. It has a branch cut along the negative real axis. We are mostly interested in this function along the positive real axis and positive imaginary axis, both away from the origin. CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 18
When we continue to dS4 by taking zc = −iηc and L = −i`, the leading contribution 3 to the order (ϕ~k) piece of the wavefunction at small ηc is given by:
λ `4 Z d~k d~k 1 π log Ψ(3) = − 3 1 2 ϕ ϕ ϕ −i γ − i log(−k η ) + , BD 3 3 ~k1 ~k2 ~k3 3 E Σ c 6 (2π) (2π) ηc 2 (2.3.15) ~ ~ ~ with k1 + k2 = −k3 due to momentum conservation (see [61] for related calculations). 3 We see that the absolute value of the wavefunction receives a ∼ 1/ηc divergent piece which is momentum independent (such that it becomes a contact term in position space). Interestingly, the cubic self-interaction of the conformally coupled scalar is absent in the classical Vasiliev equations [67,68]. As another example, consider the quartic coupling which is conformal in four- dimensions. We find: 1 In=4(ki, zc) = 4 . (2.3.16) zc kΣ
Upon continuing to dS4 this gives a momentum-dependent contribution to the real part of the exponent of the wavefunction, but none to the phase. One can also consider loop corrections analogous to those computed in section 2.3.2. As an example we consider the one loop correction to the two-point function in the φ4 theory. The relevant Green function is given by:
wz G(z, w; k) = e−k(w+z) e2kz − e2kzc , z < w , (2.3.17) 2kL2 and similarly for z > w. The relevant integral is (2.3.5), though in this case there 2 is only a quadratic divergence ΛUV to be cancelled. A small zc expansion of the regulated integral reveals the following contribution to the wavefunction:
2 Z ~ ` dk k 3λ4 log |ΨBD[ϕ~k, ηc]| = − 3 2 1 − 2 log(−kηc) + ... ϕ~k ϕ−~k . (2.3.18) 2 (2π) ηc 4(2π)
Once again, we see that the wavefunction becomes narrower as time proceeds which is physically sensible. CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 19
2.3.4 Comments for general massive fields
We discuss the non-interacting case (with λ = 0) but non-zero mass. The solutions of the Klein-Gordon equation are given by:
3/2 r z Kν(kz) 9 2 2 φ~k(z) = ϕ~k , ν ≡ + m L . (2.3.19) zc Kν(kzc) 4
Once again we have imposed that the solution vanishes at z → ∞. We are interested in the regime ν ∈ [0, 3/2], since this range corresponds to light non-tachyonic scalars q 9 2 2 in dS4 upon analytic continuation (such that ν = 4 − m ` ). Heavy particles in dS4 have pure imaginary ν. The on-shell action is found to be:
2 Z L d~k (3 − 2ν) k K(ν−1)(kzc) Scl = − 3 ϕ~k ϕ−~k 3 − 2 . (2.3.20) 2 (2π) 2 zc zc Kν(kzc)
For generic values of ν we can expand the action at small zc and find:
2 Z ~ 2ν ! L dk 1 (3 − 2ν) 2π csc(πν) kzc Scl = − 3 ϕ~k ϕ−~k 3 − 2 + ... . (2.3.21) 2 (2π) zc 2 Γ(ν) 2
2ν The above diverges at small zc, in the region ν ∈ (0, 3/2), even for the ∼ k piece. In the context of AdS/CFT the boundary data for a scalar field with ∆−d = ν−d/2 6= 0 d−∆ must be “renormalized” via ϕ~k → zc ϕ~k to achieve finite correlation functions as the cutoff zc → 0. (See Sec. 23.10 of [58] for a discussion.) In the conformally coupled case discussed in section 2.3.3, ∆ = 2 and d = 3 such that ϕ~k → zc ϕ~k . This 2 3 renormalization absorbs the 1/zc divergence. Upon continuing to dS4 the ∼ 1/zc term 2(ν−3/2) in (2.3.20) becomes a phase and we find a factor (i|ηc|) which has a growing real part as |ηc| → 0.
A small zc expansion in the ν = 0 case reveals ∼ log kzc terms in addition to 3 the 1/zc divergence. Only the logarithmic term contributes to the absolute value of ΨBD[ϕ~k, η] upon continuing zc = −iηc and L = −i`. In addition, we have that for ν = 1 there are also logarithmic terms in the small zc expansion. These become phases upon continuing zc = −iηc. CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 20
Tree-level diagrams
Once again, we can ask whether the presence of logarithmic contributions to the late time wavefunction occur for more general values of ν. The general bulk-to-boundary propagator is: z 3/2 K (kz) K(z; k) = ν . (2.3.22) zc Kν(kzc) n Consider again self-interactions of the simple form λnφ(~x,η) /2n. The tree level integrals of interest are:
4 Z ∞ n L λn dz Y I(ν)(z , k ) = K(z; k ) . (2.3.23) n c i 2n z4 i zc i=1
For generic ν, we will find that at small zc the non-local piece in momentum will be n(ν−3/2) accompanied by a divergent factor zc . Upon continuation to dS4, the local 3 pieces which go as 1/zc or 1/zc will become phases of the wavefunction. On the other n(ν−3/2) hand zc will not contribute a pure phase to the wavefunction. However, upon n computing a physical expectation value of (ϕ~k) by integrating over the tree level 2 n(3/2−ν) |ΨBD| one finds that it decays as |ηc| at late times. That the correlations decay in time for massive fields makes physical sense, since the particles dilute due to the expansion of space, and is consistent with a theorem of Weinberg [22].
On the other hand, an examination of the small z behavior of the Bessel Kν(kz) function:
(kz)ν Γ(−ν) (kz)2 (kz)−ν Γ(ν) (kz)2 K (kz) = 1 + + ... + 1 + + ... ν 21+ν 2(1 + ν) 21−ν 2(1 − ν) (2.3.24) reveals that logarithmic terms can only occur of special values of ν. They can only appear when the integrand of (2.3.23) contains terms that go as 1/z in its small z expansion, which integrate to a logarithm. For n = 3, we have already discussed the massless ν = 3/2 and conformally coupled ν = 1/2 cases at tree level, as well as the ν = 0 and ν = 1 cases at the free level. For general n, ν = 3/2 will still give rise to logarithmic contributions, as will ν = (3/2−3/n), where the logarithmic contributions P 6/n are of the form ∼ log( i ki zc)/zc and ν = (3/2 − 1/n), where the logarithmic CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 21
P contributions are of the form ∼ log( i ki zc)/zc (in the range ν ∈ [0, 3/2]). In the latter case, the logarithmic contribution always appears as a phase upon analytic continuation to dS, not so for the former case. It would be interesting if these are the only values of ν that give logarithms to higher order in perturbation theory.
2.4 Gauge fields and gravity in four-dimensions
In this section we consider classical contributions to the de Sitter wave function for massless gauge fields and gravitons in a fixed dS4 background. Compared with scalars treated in earlier sections, there are significant changes in the structure of the wave functions because of the gauge symmetry.
2.4.1 SU(N) gauge fields
In the non-Abelian case, the Yang-Mills action is given by:
4 Z Z ∞ L dz a a µρ νσ a a abc b c SYM = d~x 4 Tr FµνFρσg g ,Fµν = ∂[µAν] + g f AµAν , (2.4.1) 4 3 z R zc where a = 1,...,N 2 − 1 is the adjoint index and f abc are the SU(N) structure constants. This action is conformally invariant at the classical level. This means that there will be no singular terms in 1/zc in AdS vertex integrals and thus no terms in the de Sitter wave function that are logarithmically sensitive to ηc. The reason for this is that the two inverse metrics in the action (2.4.1) soften the vertex integrals by a factor of z4 and cancel the 1/z4 from the metric determinant.
We perform calculations in AdS4 in the Az = 0 gauge, so only transverse spatial components of the gauge potential remain. In ~k-space these components are given by
~ ~ ~ ~ ∗ ~ Ai(z, k) = K(z; k)ci(k), kici(k) = 0, ci(k) = ci(−k) (2.4.2) K(z; k) = e−k(z−zc) . (2.4.3)
The bulk-to-boundary propagator is so simple because Ai obeys the same linearized equation as in flat space. Although usually not written explicitly, the transverse CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 22
2 projector Πij = δij − kikj/k is understood to be applied to spatial vector modes. As our first calculation, we obtain the contribution of the free gauge field. Metric factors cancel and we have the gauge-fixed action
Z Z ∞ 1 2 1 2 S = d~x (∂zAi) + (Fij) . (2.4.4) zc 2 4
After partial integration, as in Sec. 2.2.2, the on-shell action reduces to the surface term in ~k-space:
1 Z d~k S = − K(z, k)∂ K(z, k)c (~k)c (−~k) (2.4.5) 2 (2π)3 z i i 1 Z d~k = k(δ − k k /k2)c (~k)c (−~k) . (2.4.6) 2 (2π)3 ij i j i j
The result contains the ~k-space correlator of two conserved currents in the boundary
3d CFT. This structure, which contains no dependence on zc may be compared with its analogue in (2.3.11) for the conformally coupled scalar. The bulk fields φ and
Ai are both dual to CFT operators with ∆ = 2. There is only partial cancellation 2 of metric factors for the scalar, so the singular factor 1/zc remains. As discussed in Sec. 2.3.4, this factor can be absorbed by renormalization of sources in AdS, but it gives a late-time power law singularity in |ΨBD|. Next consider the tree-level three-point function. The relevant integral is straight- forward and gives a result with no dependence on zc, namely:
T g f abc ijk (2.4.7) k1 + k2 + k3 where Tijk is the same antisymmetric tensor that appears in flat space, namely:
~ ~ ~ ~ ~ ~ Tklm = (k1)lδkm − (k1)mδkl + (k2)mδkl − (k2)kδlm + (k3)kδlm − (k3)lδkm . (2.4.8)
Comparing (2.4.7) to the three point function of the conformally coupled scalar in
(2.3.14) we note the absence of logarithmic terms depending on zc. CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 23
2.4.2 Scalar QED
Consider now a massive charged scalar field coupled to a U(1) gauge field, with Euclidean action:
Z Z ∞ 4 dz αβ ∗ 2 ∗ SSQED = L d~x 4 g (∂α + iAα)φ(∂β − iAβ)φ + m φ φ . (2.4.9) 3 z R zc
Properties of this theory were also considered in [32]. Transverse modes in ~k-space thus have the cubic interaction:
Z ~ ~ Z ∞ 2 dk1 dk2 dz ~ ~ ~ ∗ Sint = L Ai(z, k3)(k1 − k2)i φ~ (z) φ (z) , (2.4.10) 3 3 2 k1 ~k2 (2π) (2π) zc z
~ ~ ~ where momentum conservation requires k3 = −k1 −k2 . Again, a transverse projector is understood to be applied to spatial vector modes. Using this interaction vertex, we find the following contribution to the partition function:
Z ~ ~ 2 dk1 dk2 ˜ ~ ~ ~ ˜ ~ ~ L ϕ~ ϕ~ Ai(k3)(k1 − k2)i Im2L2 , Ai(k) ≡ Ai(zc, k) . (2.4.11) (2π)3 (2π)3 k1 k2
For a scalar field of mass m2L2 and bulk-to-boundary propagator K(z; k) the radial integral is Z ∞ dz −k(z−zc) 2 2 Im L = 2 e K(z; k1)K(z; k2) . (2.4.12) zc z We compare the two cases of massless and conformally coupled (m2L2 = −2) scalars with bulk-to-boundary propagators:
−kz (1 + kz)e z −k(z−zc) 2 2 2 2 Km L =0(z; k) = −kz ,Km L =−2(z; k) = e . (2.4.13) (1 + kzc)e c zc
Our motivation is to explore the appearance of log(kzc) terms in the 3-point function. For the massless case we find:
1 k1k2 (k1+k2+k3)zc + + e k3Ei[−(k1 + k2 + k3)zc] zc k1+k2+k3 Im2L2=0 = (2.4.14) (1 + k1zc)(1 + k2zc) CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 24
whose series expansion reveals a logarithmic term from the Ei(z)-function. For m2L2 = −2, the integral is elementary and gives:
Z ∞ e−(k1+k2+k3)(z−zc) 1 2 2 Im L =−2 = dz 2 = 2 . (2.4.15) zc zc (k1 + k2 + k3)zc
As in the case of the conformally coupled self-interacting scalar, we can absorb the 2 1/zc divergence into a renormalization of the boundary data.
After all is said and done, we find a ∼ log(kzc) term in the 3-point function of the massless scalar but not in the conformally coupled case. The “practical” reason for the absence is the cancellation of the ∼ 1/z factors in the integrand of (2.4.15) due to the softer behavior of the scalar bulk-to-boundary propagators. It would be interesting to study the loop corrections to the wave function for these theories.
2.4.3 Gravity
It is a well known result that classical solutions in pure Einstein gravity with a positive cosmological constant Λ = +3/`2 have a uniform late time (small η) expansion. In four spacetime dimensions, this is given by [70]:
ds2 dη2 1 = − + g(0) + η2g(2) + η3g(3) + ... dxidxj , |η| 1 . (2.4.16) `2 η2 η2 ij ij ij
The independent data in this expansion is the conformal class
(0) (3) ω(~x) (0) (3) gij , gij ∼ e gij , gij . (2.4.17)
(3) The Einstein equations impose that gij is transverse and traceless with respect to (0) the boundary three-metric gij . Two of the phase space degrees of freedom reside (0) (3) in gij and the other two in gij . The Einstein equations also require that the term linear in η inside the parenthesis is absent. If g(0) and g(3) are appropriately related, the above solution will obey the Bunch-Davies boundary condition (this will require (3) gij to be complex). If Λ < 0 there is an analogous expansions of the same structure known as the CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 25
Fefferman-Graham expansion [69]. The on-shell action for such solutions satisfying the Bunch-Davies boundary condition (i.e. that the three-metric vanishes at large z in EAdS) has been studied extensively [71]. Indeed, the on-shell classical action is given at some fixed z = zc by:
Z Z ∞ Z √ 3 √ 1 i Sgr = 2 d~x dz g − 2 d~x hKi , (2.4.18) 8πGL M zc 8πGL ∂M where hij is the induced metric on the fixed zc slice and Kij is the extrinsic curvature,
1 α K = L α g (z, ~x) , n = (z,~0) . (2.4.19) ij 2 n ij
The second term in (2.4.16), known as the Gibbons-Hawking term, is required for a well defined variational principle. For the first term we have used the on-shell condition R = −12/L2. We can evaluate the classical action (2.4.18) on the classical solutions obeying the
Euclidean AdS4 analogue of (2.4.16):
ds2 dz2 1 = + g(0) + z2g(2) + z3g(3) + ... dxidxj , z 1 . (2.4.20) L2 z2 z2 ij ij ij and expand in small zc. The expansion of the on-shell classical action contains only 3 2 divergences of the form 1/zc and 1/zc at small zc, but no 1/zc or log zc divergences [72]. The absence of a ∼ 1/z term in (2.4.20) is crucial for the logs to be absent in the 10 small zc expansion of the on-shell classical action. The divergent terms amount 11 to pure phases in ΨBD[gij, η] upon analytic continuation. (See [75] for a related discussion.) The important point is that there are no logarithmic divergences for small zc, which translates to the statement that the Bunch-Davies wavefunction exhibits 3 no ∼ k log(−ηc k) growth at tree level. 10Note that in odd space-time dimensions, there is a piece of the Fefferman-Graham expansion which contributes logarithmic terms to the phase of the wavefunction as well as local terms to its absolute value [16]. 11 2 2 2 2 i j There is a slight subtlety in assuming that the full solution ds /` = −dη /η + gij(~x,η)dx dx allowing for the expansion (2.4.16) can indeed by analytically continued to z = −iη at the non- (0) linear level. For small enough deviations away from the flat metric gij = δij the bulk-to-bulk and bulk-to-boundary propagators allow for such a continuation. CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 26
Thus the Fefferman-Graham expansion for dS4 seems to explain the absence of logarithms in the gravitational 3-point functions calculated, for example, in [73, 74]. This is in stark contrast to the case of the massless scalar.
2.5 3d CFTs and (A)dS/CFT
The dS/CFT correspondence proposes that the Bunch-Davies (or Hartle-Hawking) wavefunction of dS4 at late times is computed by the partition function of a three- dimensional Euclidean conformal field theory. It is closely related to the Euclidean AdS/CFT proposal, as we have tried to make clear above. In the AdS/CFT context the small zc cutoff is identified with a cutoff in the dual theory. This is due to the manifestation of the dilatation symmetry as the (z, ~x) → λ(z, ~x) isometry in the bulk.
For instance, bulk terms that diverge as inverse powers of zc (with even powers of k) are interpreted as local terms in the dual theory. On the other hand, the tree level zc-dependent logarithmic terms, such as those in the small zc expansion of (2.3.1), are not local in position space and yet seem to depend on the cutoff. One may ask whether they have an interpretation from the viewpoint of a putative CFT dual. Recent analyses of CFT correlation functions in momentum space [62, 63] give a suggestive answer. Recall that the symmetries of CFTs have associated Ward identities, governing correlation functions. For concreteness we specifically consider the Ward identities, expressed in momentum space, constrainging the three point functions of a scalar operator O with weight ∆. The Ward identity for the dilatation symmetry is given by:
3 ! X 6 + (pj∂j − ∆) hO(~p1)O(~p2)O(~p3)i = 0 , (2.5.1) i=1 whereas for the special conformal transformations we have:
3 X 4 − 2∆ (~p )α ∂2 + ∂ hO(~p )O(~p )O(~p )i = 0 . (2.5.2) i i p i 1 2 3 i=1 i CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 27
The Latin index labels a particular momentum insertion, O(~pi), whereas the Greek index labels a specific Euclidean component of ~pi. We have also removed the δ(~p1 +
~p2 + ~p3) conservation rule from the correlator. The solution to the above equations is most conveniently expressed as an integral over an auxiliary coordinate:
hO(~p1)O(~p2)O(~p3)i = Z ∞ ∆−3/2 1/2 c3 (p1p2p3) dz z K∆−3/2(z p1) K∆−3/2(z p2) K∆−3/2(z p3) , (2.5.3) 0 where Kν(z) is the modified Bessel function of the second kind. The above integral should look familiar; we have indeed encountered it in our previous analysis of massive fields. As previously noted, from the bulk point of view the conformal weight of a 2 2 p 2 2 scalar of mass m L in AdS4 is ∆ = 3/2 + ν where ν ≡ 9/4 + m L . The case
ν = 3/2, i.e. a massless scalar field in AdS4, corresponds to a marginal operator with ∆ = 3. Thus, the auxiliary variable z can be precisely identified with an AdS bulk coordinate, and the modified Bessel functions can be thought of as bulk-to-boundary propagators (2.3.22). The integral (2.5.3) is of course divergent for general ∆ near z = 0. Motivated by our bulk analysis, we chose a slightly different cutoff procedure from [62], where we instead cut the integral off at some small z = zc. Now from a CFT analysis, we find the appearance of logarithmic contributions, which will generally be cutoff dependent, to the three-point function of a scalar operator (this observation remains true even in the cutoff prescription chosen in [62]). Because these logarithmic terms contain a dependence on the cutoff scale zc, they are consequently referred to as anomalies in [62]. They may be present in the theory non-perturbatively. Thus, from the holographic point of view, terms logarithmic in zc that are associated to anomalies in the 3d CFT, such as those in the three-point function, will be present to all orders rather than part of a resummeable series. Let us note that we also observed such logarithms in higher point functions at special values of ν (e.g. ν = 3/2), where the analogous general CFT analysis is more cumbersome. In a similar fashion, the tree level logarithms we discussed for the Bunch-Davies wavefunction in section 2.3.4 (for ν = 0 or ν = 1) are related to a divergence in the Fourier transform of the two-point function of a weight ∆ = 3/2 or CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 28
∆ = 5/2 scalar operator [62,76]. It is also possible, however, that the appearance of these logarithms are the result of small shifts in the conformal weights of certain operators in the 3d CFT. For instance, imagine that loop corrections (such as 1/N corrections in a large N CFT) shift ∆(0) by an order ∼ 1/N amount, i.e. ∆ = ∆(0) +α/N +O(1/N 2) with α ∼ O(1). The two point function in momentum space will then have a large N expansion:
(0) 2α k2(∆−3/2) = k2(∆ −3/2) 1 + log k + ... . (2.5.4) N
2(∆−3) From the bulk AdS4 point of view, we must include the factors zc to obtain the zc-dependent bulk partition function, as we discussed in section 2.3.4, such that the expansion becomes:
2(∆−3/2) 2(∆(0)−3/2) (zck) (zc k) 2α 3 = 3 1 + log(zck) + ... . (2.5.5) zc zc N
As we already noted, we can extrapolate the perturbative results in AdS to those in dS by continuing zc = −iηc and L = −i`. From the point of view of a putative dual CFT of dS, the zc = −iηc continuation corresponds to an analytic continuation of the cutoff itself. Though unusual from the point of view of field theory, it may be interesting to consider general properties of field theories with such imaginary cutoffs. Notice that the expansion (2.5.5) now contains ∼ log(−kηc) pieces which 12 are resummed to a power law behavior in ηc. With this interpretation we might view (2.3.10) as a small negative shift in the weight ∆ = 3 of the relevant operator dual to the bulk massless field, such that it becomes slightly relevant. On the other hand, the fact that the three-point function of a marginal scalar operator contains an anomalous logarithm suggests that the wavefunction has a non-trivial time evolution. In the case we consider, where it is due to a cubic self-interaction of a massless scalar
12A concrete realization occurs in the conjectured duality between the three-dimensional Sp(N) critical model [3,41] and the minimal higher spin theory in dS4. The bulk scalar has a classical mass m2`2 = +2 and is dual to a spin zero operator whose conformal weight is ∆ = 2 at N = ∞, but receives 1/N corrections [77] (related by N → −N to those of the critical O(N) model). Similar corrections will also occur for the extended dS/CFT proposals in [42, 43]. CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 29
(see (2.3.4)), this might have been expected given that we are not perturbing about a stable minimum of the bulk scalar potential. However, this anomalous logarithm may disappear should we correct the propagators to reflect the negative shift in weight.
The CFT stress tensor operator Tij also weight ∆ = 3 and is thus also a marginal operator. In (A)dS/CFT it is dual to the bulk graviton. Absence of a Weyl anomaly in three dimensional CFTs can be expressed as the following property of the CFT partition function: ω(x) ZCFT [gij] = ZCFT [e gij] , (2.5.6) where ω(x) is a smooth function and we are removing local counterterms. The above implies that correlation functions of the stress tensor, given by variational derivatives with respect to gij, cannot depend on the Weyl factor of gij (in the absence of any other sources) and in particular cannot depend on the logarithm of the cutoff. This strongly suggests, if we are to take the picture of dS/CFT seriously, that late time log ηc contributions to the wavefunction ΨBD[gij], such as the one describing pure Einstein theory, are absent to all orders in perturbation theory. This agrees with several computations of the cubic contribution [73, 74], as well as our general tree level argument in section 2.4.3, which are all devoid of such logarithmic terms. These observations, however, do not preclude the possibility of ΨBD[gij, η] peaking far from the de Sitter vacuum.
2.6 dS2 via Euclidean AdS2
We now proceed to study several perturbative corrections of the Bunch-Davies wave- function about a fixed dS2 (planar) background. We consider the massless scalar field in Euclidean AdS2 whose action is:
1 Z Z ∞ L2 m2 L2 λ S = d~x dz (∂ φ(~x,z))2 + (∂ φ(~x,z))2 + φ(~x,z)2 + φ(~x,z)3 . E 2 z ~x z2 3 z2 R zc (2.6.1) The reason for reducing to two-spacetime dimensions is that the integrals needed 2 for order λ calculations are far simpler that for AdS4, although their mathematical CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 30
structure and the physical issues are quite similar. We will focus on cubic interactions.
2.6.1 Tree level corrections for the massless theory
3 The simplest contribution to consider is the order λ (ϕ~k) contribution. For massless fields, the bulk-to-boundary K(z; k) and bulk-to-boundary G(z, w; k) propagators are given in appendix A.2. This correction is a tree level diagram involving three bulk- to-boundary propagators. In order to calculate it, we must evaluate the integral:
L2 λ Z ∞ dz L2 λ 1 zc kΣ − 2 K(z; k1)K(z; k2)K(z; k3) = − + e kΣ Ei(−zc kΣ) , 6 zc z 6 zc (2.6.2) where kΣ ≡ k1 + k2 + k3. In the limit of small zc we find a ∼ log(zc kΣ) contribution.
Continuing to dS2 by taking L = −i` and zc = −iηc, the Bunch-Davies wavefunction at late times to order λ is given by:
log ΨBD = ! Z d~k k `2λ Z d~k i 1 − ϕ ϕ + 2 ϕ ϕ ϕ + k (γ + log(−η k ) , (2.6.3) ~k1 −~k1 ~k1 ~k2 ~k3 Σ E c Σ 2π 2 6 2π ηc
~ ~ ~ where we must impose k3 = −k1 −k2 due to momentum conservation. Once again, we note that the absolute value of the Bunch-Davies wave function receives a logarithmic contribution. 2 4 At order λ we have a ∼ (ϕ~k) contribution to the wave function which also involves an integration over the bulk-to-bulk propagator. The integrals can also performed to obtain a result that behaves (schematically) in the small zc limit as 2 2 ∼ λ k log kzc . The integral we need is:
2 4 Z λ L dz dw 2 2 G (z, w, ~q) K z; k1 K z; k2 K w; k3 K w; k4 , (2.6.4) 8 D z w CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 31
2 where the domain of integration is D = [zc, ∞] . In the small zc limit, we find:
2 4 λ L 1 1 2 1 2 − (s + q) (log[(s + q)zc]) − (t + q) (log[(t + q)zc]) + ... , (2.6.5) 8 zc 2 2
~ ~ ~ ~ ~ ~ ~ ~ where s ≡ |k1| + |k2| and t ≡ |k3| + |k4| and q ≡ |~q| = |k1 + k2| = |k3 + k4| (note that s, t > q by the triangle inequality).
2.6.2 Loop corrections for the massless theory
A tadpole diagram contributes to the wavefunction at order λ. The relevant integral is given by: L2 λ Z ∞ dz Z d~p 2 K(z, k = 0) G(z, z; p) . (2.6.6) 2 zc z 2π Note that K(z, k = 0) ≡ 1. To render the integral finite we impose a physical ultra- violet cutoff, which becomes a z-dependent cutoff pUV = ΛUV L/z for the coordinate momentum over which we are integrating. We can add a counterterm to the action of the form: Z Z ∞ dz S = δ L2 d~x φ(~x,z) . (2.6.7) ct z2 R zc
The constant δ can be selected to cancel the logarithmic divergence in ΛUV rendering the following result for the full integral (2.6.6):
L2 λ −1 + γ + log 2 E . (2.6.8) 4π zc
Upon continuation to dS2 this contributes only to the phase of the wavefunction. 2 2 At order λ we have two distinct loop corrections to the ∼ (ϕ~k) term. One involves attaching a tadpole to the tree level propagator whose ultraviolet divergence can be treated as above. The relevant integral is given by:
L4 λ2 Z dw dz Z d~p Itadpole(k, zc; L) = 2 2 G(w, w; p) G(z, w; 0) K(z; k) K(z; k) . 4 D w z 2π (2.6.9)
At small zc the above integral contains a finite term plus a logarithmic piece in zc. CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 32
The result is:
Itadpole(k, zc; L) = 4 2 2 L λ −12 + π + 6 (γE + log 2) (γE + log 2) 2 − k (logkzc) + ... . (2.6.10) 2π 24zc 4 where the subleading pieces are at most logarithmic in zc. The other order λ2 contribution comes from a ‘sunset’ diagram, which is ultraviolet finite in two-dimensions and thus requires no regularization. It involves an integral of the form:
4 2 Z Z L λ dz dw d~p ~ Isunset(k, zc; L) = 2 2 G(z, w; p)G(z, w; |~p + k|) K(z; k) K(w; k) . 4 D z w R 2π (2.6.11) For ~k = 0, the above integral can be performed analytically and we find:
L4 λ2 π2 − 8 Isunset(0, zc; L) = . (2.6.12) 4 zc 8π
We were not able to perform the full integral analytically, however a numerical eval- uation reveals the following small zc expansion:
L4 λ2 I (k, z ; L) − I (0, z ; L) = (a k log kz + a k + ...) , (2.6.13) sunset c sunset c 2 1 c 2
2 with a1 ≈ +0.261 ... and a2 ≈ +0.58 ... The (ϕ~k) piece of the late time (absolute value of the) wavefunction to order λ2 is then:
log |ΨBD[ϕ~k, ηc]| = Z d~k k − + I (k, −iη ; −i`) + I (k, −iη ; −i`) ϕ ϕ . (2.6.14) 2π 2 sunset c tadpole c ~k −~k
2 Thus we see that at loop level there are logarithmic corrections to the (ϕ~k) piece of the wavefunction. For the sunset diagram, the loop correction required no ultraviolet cancelation and so the logarithmic term present in the result is free of any potential CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 33
scheme dependence.
2.7 Outlook
In this chapter we have explored the late time structure of ΨBD in a de Sitter back- ground, by computing its quantum corrections employing a perturbative framework heavily used in the AdS/CFT literature. We have identified several types of behavior, including the logarithmic growth in conformal time. Logarithmic growth commonly appears in the correlators computed in the in-in formalism. Furthermore, we have connected the late time properties of ΨBD to certain anomalies and shifts in confor- mal weights of a CFT putatively dual to a bulk de Sitter theory containing the types of fields and interactions we studied. There are several interesting avenues left to explore.
• Graviton loops: One would like to firmly establish the absence (or presence) of logarithmic growth for pieces of the wavefunction that depend on the metric only, both for a pure Einstein theory and more general theories of gravity, such as those with higher derivative terms.
• Higher spin holography: We found that cubic interactions for conformally cou-
pled scalars lead to an additional local cubic piece of ΨBD that was intricately re- lated to a logarithmic phase. Such scalars are present in the higher spin Vasiliev theory, but the cubic scalar coupling is absent at the classical level [67,68]. At
loop level, however, there may be a contribution to the cubic piece of |ΨBD|, which can be computed in the dual CFT. The presence of such additional lo-
cal contributions may give interesting new contributions to ΨBD for large field values. Similar considerations may also interesting for the alternate boundary condition dual to a ∆ = 1 scalar operator in the CFT.
• Resummation: We discussed a possible interpretation of the logarithmic growths as pieces of a series corresponding to a small shift in the conformal weight ∆ of an operator in the dual CFT. For a massless scalar with φ4 self-interactions, we saw that such a shift would cause the dual operator to be marginally relevant, CHAPTER 2. WAVEFUNCTIONS IN DE SITTER SPACE 34
∆ < 3, rather marginally irrelevant. It would be interesting to relate this picture of resummation to other proposals involving dynamical renormalization group methods (see for example the review [24]).
• Stochastic inflation: It would be interesting to relate our calculations and in- terpretations to the framework of stochastic inflation [29] which proposes a non-perturbative approach for interacting fields in a fixed de Sitter background. Another approach to study strongly coupled (conformal) field theories in a fixed de Sitter background is using the AdS/CFT correspondence where AdS has a de Sitter boundary metric, on which the CFT resides (see for example [79]). Chapter 3
Higher Spin dS/CFT
3.1 Introduction
As we have seen in the previous chapter, a natural object to consider when studying an asymptotically (approximately) de Sitter spacetime, such as the universe during the inflationary era, is the wavefunction [12] as a function of small fluctuations of the bulk fields. For a free massless scalar φ (such as the inflaton in slow roll inflation) in the Bunch-Davies vacuum state1 |Ei [14, 15, 82–88] in a fixed four-dimensional de Sitter background:
`2 ds2 = −dη2 + d~x2 , η ∈ (−∞, 0) , (3.1.1) η2 we found the late time Hartle-Hawking Gaussian wavefunctional at η = ηc → 0 [16]:
2 Z 3 ` d k 3 lim |ΨHH [ϕ(~x), ηc] | ∼ exp − 3 k ϕ~k ϕ−~k , (3.1.2) ηc→0 2 (2π) where ϕ~k are the Fourier components of the late time profile ϕ(~x). Such a Gaussian wavefunction gives rise to the scale-invariant fluctuations of the cosmic background radiation. Understanding the behavior of such a wavefunction for a large range of
1It might be more appropriate to call it the Bunch-Davies-Hartle-Hawking-Euclidean- Schomblond-Spindel-Chernikov-Tagirov-Mottola-Allen-Sasaki-Tanaka-Yamamoto-Critchley- Dowker-Candelas-Raine-Boerner-Duerr vacuum state.
35 CHAPTER 3. HIGHER SPIN DS/CFT 36
values for its arguments, which include the metric, and with the inclusion of quantum corrections is a basic problem in quantum cosmology. The perturbative Bunch-Davies wavefunction (3.1.1) was noticed to be a simple analytic continuation [16] of the partition function of a free massless field in a fixed Euclidean anti-de Sitter space. These observations, coupled with the correspondence between anti-de Sitter space and conformal field theory, motivate the proposal that at late times (or large spatial volume) ΨHH is computed by a statistical (and hence Euclidean) conformal field theory, in what has come to be known as the dS/CFT conjecture [16, 39, 40].2 In its weakest form dS/CFT conjectures that the Taylor coefficients of the logarithm of the late time Hartle-Hawking wavefunctional expanded # about the empty de Sitter vacuum at large N ∼ (`/`p) are the correlation functions of such a non-unitary CFT. Namely, at some late time cutoff η = ηc → 0 we have:
log ΨHH [ϕ(~x), ηc] = ∞ X 1 Z Z d3x ... d3x ϕ(~x ) . . . ϕ(~x ) hO(~x ) ... O(~x )i . (3.1.3) n! 1 n 1 n 1 n CFT n=1
The correlators hO(~x1) ... O(~xn)iCFT , where the operator O has been rescaled by an appropriate ηc dependent factor, compute late time bulk correlation functions with future boundary conditions [127]. The bulk late time profiles ϕ(~x) are taken to be infinitesimal such that ΨHH [ϕ(~x), ηc] is merely a generating function of late time correlators about ϕ(~x) = 0. In its strongest form, the claim is that the CFT is a non-perturbative definition of ΨHH for finite deviations away from the pure de Sitter vacuum and at finite N. Particularly, ΨHH is computed by the partition function of the putative CFT with sources turned on. The single-trace operators are dual to the bulk de Sitter fields. Abstractly speaking, if we could write down a complete basis B (which may include information about topology, geometry and matter) for the Hilbert space of our theory, ΨHH would be computing the overlap of the Hartle- Hawking state |Ei with a particular state |βi ∈ B. One could also consider computing
2See [45] for a discussion of several aspects of de Sitter space. Other proposals include [11,13,59, 89–102]. CHAPTER 3. HIGHER SPIN DS/CFT 37
the partition function with sources for more general multi-trace operators turned on. As we shall see this computes the overlap of |Ei with states which are not sharp eigenstates of the field operator or its conjugate momentum. More dramatically, if there are single-trace operators which are irrelevant they may correspond to an exit from the de Sitter phase (see for instance [52]). de Sitter space arises as a non-linear classical solution to four-dimensional higher spin gravity [65,66,103–106]. This theory has a tower of light particles with increasing spin, including a spinless bulk scalar with mass m2`2 = +2 and a spin-two graviton. The scalar potential has a minimum about the pure de Sitter solution and the kinetic terms of the higher spin particles carry the right signs and are canonical. Hence the de Sitter vacuum is perturbatively stable and free of tachyons and ghosts in this theory. Beyond perturbation theory, the late time Hartle-Hawking wavefunctional of asymptotically de Sitter space in higher spin gravity (at least when the topology at I+ is sufficiently simple) is conjectured to be computed by the partition function of the 2 Sp(N) model [107] (see also [108,109]), with N ∼ (`/`p) . This model comes in two flavors, either a free theory of N anti-commuting scalar fields transforming as vectors under the Sp(N) symmetry or as a critical theory obtained from the Sp(N) model by a double trace deformation [110]. The partition function of the critical model (at least at large N) as a functional of the sources of the single-trace operators computes the wavefunction in the ordinary field basis. On the other hand, the free theory computes the wavefunction in a slightly modified basis. The quantum mechanical analogue of this basis is given by eigenstates of the Hermitian operatorς ˆ = (βxˆ − αpˆ) with α, β ∈ R. Given the wavefunction in the coordinate basis ψ(x) we can compute in the ςˆ-basis by performing the transform:
i βx2 Z − −ςx iβx2 Z 1 α 2 1 − iςx ψ(ς) = √ dx e ψ(x) , ψ(x) = √ e 2α dς e α ψ(ς). 2πα 2πα (3.1.4) Normalizability in thex ˆ-basis implies normalizability in theς ˆ-basis and vice-versa. However, a node-less ψ(x) will not necessarily give a node-less ψ(ς). The free Sp(N) model computes the late time Hartle-Hawking wavefunction of CHAPTER 3. HIGHER SPIN DS/CFT 38
the bulk scalar in the eigenbasis of the late time operator [78]:
√ 2 ˆ 3 i δ N ηc σˆ = φ − ηc πˆφ , πˆφ ≡ −p , (3.1.5) det gij δφ
ˆ where φ is the bulk scalar field operator andπ ˆφ is the field momentum density op- 2 erator. We have taken |ηc| 1 as a late time cutoff and the combination gij/ηc represents the spatial metric at this time in Fefferman-Graham gauge. It is some- what remarkable that there exists a basis for which the wavefunctional is computed by a free dual theory given that in the ordinary field basis it is computed by a strongly coupled theory. The partition function of the free Sp(N) theory on an R3 topology is an explicit resummation of the correlation functions and there are no non-perturbative phenomena. The remaining coordinates of the wavefunctional, which are late time profiles of the higher spin fields including the bulk graviton, are computed in the ordinary field basis for both the free and critical models. The wavefunctionals for the bulk scalar in the two different bases are related by a functional version of the transform (3.1.4). As shown in [78], in the large N limit performing the transform amounts to finding the saddles of a functional equation. One of these saddles (not necessarily the dominant one) gives the field basis wave- function whose perturbative expansion agrees with that computed perturbatively in the bulk about the pure de Sitter solution. In most of what follows, we will perform computations in theσ ˆ-basis, which amounts to the calculation of a functional deter- minant. This is a considerably hard object to compute and we will limit ourselves to situations where we have conditioned all bulk fields except the metric gij and the scalar φ to have vanishing late time profiles. We should emphasize that this amounts to a very sharp conditioning and ultimately the situation could be altered by allowing higher spin deformations.3 The main body of this chapter is dedicated to the study of the partition function of the Sp(N) model for a large class of SO(3) preserving deformations of the bulk scalar and graviton. That is to say we study the late time wavefunction for bulk
3It might be worth mentioning that non-linear bulk solutions exist in higher spin gravity for which only the bulk metric and scalar are turned on [111]. CHAPTER 3. HIGHER SPIN DS/CFT 39
graviton and scalar configurations with late time profile on an R3 topology:
ds2 = dr2 + f(r)2 r2 dΩ2 , φ = φ(r) . (3.1.6)
2 2 where dΩ2 is the round metric on an S whose SO(3) symmetry is the one preserved. We also study the analogous problem on an S3 topology, which amounts to a simple conformal transformation of the R3 case, where the late time profiles take the form:
ds2 = dψ2 + f(ψ)2 sin2 ψ dΩ2 , φ = φ(ψ) . (3.1.7)
This allows us to examine the behavior of the wavefunction of higher spin de Sit- ter space for inhomogeneous and anisotropic deformations which extend the rather uniform and homogenous deformations studied in [78]. In [78] it was found that the wavefunction in theσ ˆ-basis diverges as a function of uniform mode of the bulk scalar on the round metric on S3. One of the questions we would like to understand is whether such non-normalizabilities persist for less ‘global’ late time configurations such as the above SO(3) deformations. The above metric deformations are confor- mally equivalent to the flat/round metric on R3/S3. We find particularly striking numerical evidence, discussed in section 3.5, that upon fixing the uniform mode of the bulk scalar on S3 (in the conformal frame where it is endowed with the standard metric) all other directions of a general SO(3) late time deformation are normalizable. We have also analyzed a different geometric deformation, this time homogeneous but anisotropic, which does not keep the metric in the same conformal class. This is a squashing deformation α of the round metric on S3 expressed as an S1 fiber over an S2: 1 1 ds2 = dθ2 + cos2 θdφ2 + (dψ + sin θdφ)2 , (3.1.8) 4 1 + α along with a uniform late time profile for the bulk scalar. When α = 0 the metric (3.1.8) reduces to the standard metric on S3. Once again, so long as the scalar profile is kept fixed we find that the wavefunction is bounded in the α direction. We begin by briefly reviewing the Sp(N) model in section 3.2. In section 3.3, using technology developed in [112], we compute the wavefunction (in theσ ˆ-basis) CHAPTER 3. HIGHER SPIN DS/CFT 40
for some Gaussian radial deformations of the bulk scalar. This amounts to computing a functional determinant of a scalar field with a radially dependent mass term. In section 3.4 we compute the wavefunction for a radial deformation of the geometry, in the presence of a radial mass deformation, that takes it from the flat metric on R3 to a 2 2 2 2 2 general form ds = dr +r f(r) dΩ2 . In section 3.5 we compute the wavefunction for several harmonics on the three-sphere and linear combinations thereof and note that the wavefunction seems to diverge only when the zero harmonic becomes large and negative. We then discuss the behavior of the wavefunction on a squashed sphere with a uniform profile of the late time bulk scalar in section 3.6. In section 3.7 we make some general remarks about double trace deformations. We end by speculating on possible extensions of higher spin holography in section 3.8. Most of our calculations can carry over to the O(N) model and its AdS4 dual in higher spin gravity [113–116].
3.2 Wavefunctionals and the free Sp(N) model
We wish to study the Hartle-Hawking wavefunctional of an asymptotically de Sitter higher spin gravity for deformations of the bulk scalar and graviton away from the pure de Sitter solution. We will restrict the topology of space to be R3 or S3 and allow only deformations that decay sufficiently fast at infinity. In this section, we remind the reader of the Sp(N) theory and discuss how to compute its functional determinant for certain SO(3) invariant radial deformations. One motivation to do so is to understand the behavior of the wavefunction of higher spin de Sitter space for mass deformations that are more ‘localized’ than those studied in [78] which were uniform over the entire S3. It is also worth noting that computing analogous pieces of the Hartle-Hawking wavefunction for a simple toy model of Einstein gravity coupled to a scalar field with a simple potential would require significant numerical work even in the classical limit, let alone at finite N. One would have to find a complex solution of the Euclidean equations of motion that caps off smoothly in the interior and has the prescribed boundary values at large volume, and compute its on-shell action. CHAPTER 3. HIGHER SPIN DS/CFT 41
3.2.1 Wavefunctional
Recall that the action of the free Sp(N) model on a curved background gij with a i (0) A B source, m(x ), turned on for the J = ΩABχ χ ≡ χ · χ operator (dual to the bulk scalar) is given by:
1 Z √ R[g] S = d3x g Ω ∂ χA∂ χBgij + χAχB + m(xi)χAχB , 2 AB i j 8 {A, B} = 1, 2,...,N, (3.2.1) where N should be even. The fields χA are anti-commuting scalars that transform as Sp(N) vectors and ΩAB is the symplectic form. Notice that due to the presence of the conformal coupling, the action is invariant under a local Weyl transformation 2W (xi) i of the metric: gij → e gij, so long as we also rescale the source as m(x ) → e−2W (xi)m(xi). From the bulk point of view this amounts to performing a coordinate transformation η = e−W (xi)η, as can be seen by studying the Starobinski-Fefferman- Graham expansion [69,70,117] near η = 0:
ds2 dη2 1 = − + g (xi)dxidxj + . . . , φ = η ν(xi) + η2 µ(xi) + .... (3.2.2) `2 η2 η2 ij √ Notice that µ(xi) ≡ Nm(xi) is not the coefficient of the slowest falling power of η for bulk scalar. At the linearized level, the ν(xi) profile is dual to the vev of the J (0) operator in the presence of an infinitesimal m(xi) source. Computing the late time Hartle-Hawking wavefunctional in theσ ˆ-basis with σ = m(xi) amounts to computing the partition function of the Sp(N) theory with finite sources turned on. Given that it is a Gaussian theory, we can integrate out the anti-commuting χA fields and find:
N/2 i i 2 R[g] i lim ΨHH gij, m(x ), ηc = Zfree[gij, m(x )] = det −∇g + + m(x ) , ηc→0 8 (3.2.3) CHAPTER 3. HIGHER SPIN DS/CFT 42
where: N Z i Y A −S[χA,m(xi)] Zfree[gij, m(x )] ≡ Dχ e . (3.2.4) A=1
2W (xi) In the case of a metric gij = e δij that is conformally equivalent to the flat metric on R3, it is convenient to compute the functional determinant in the conformal frame where gij is the flat metric. This amounts to rescaling the source to:
mˆ (xi) = e2W (xi)m(xi) . (3.2.5)
We will use this fact in section 3.4.
3.2.2 Functional determinant for radial deformations
We have seen that for conformally flat metrics our problem reduces to computing a functional determinant: 2 i det −∇ 3 +m ˆ (x ) , (3.2.6) R where ∇2 is the Laplacian of the round metric on 3, namely ds2 = dr2 + r2dΩ2. R3 R 2 The above object is badly divergent unless we regulate it somehow. We will regulate it using a heat kernel or zeta function approach, both of which give the same answer. In fact, this precise problem has been studied by Dunne and Kirsten in [112] for functionsm ˆ (xi) which only depend on the radial coordinate, i.e.m ˆ (xi) =m ˆ (r), and which vanish sufficiently fast at infinity. It was shown that the zeta function regulated determinant is given by the following sum:
∞ R ∞ det [−∇2 + µ2 +m ˆ (r)] X dr r mˆ (r) log = (2l + 1) log T (l)(∞) − 0 . det [−∇2 + µ2] 2l + 1 l=0 (3.2.7) In the above, the factor (2l + 1) originates from the degeneracy of eigenfunctions on a two-sphere and T (l)(r) solves the equation:
2 d (l) (1 + l) I3/2+l(µr) d (l) (l) − 2 T (r) − 2 + µ T (r) +m ˆ (r)T (r) = 0 , (3.2.8) dr r I1/2+l(µr) dr CHAPTER 3. HIGHER SPIN DS/CFT 43
with boundary conditions T (l)(0) = 1 and dT (l)(0)/dr = 0. The parameter µ2 ∈ R is a constant mass parameter that we will set to zero. The derivation of the above formula employs the Gelfand-Yaglom theorem [118], which expresses the regulated functional determinant of a one-dimensional Schr¨odingeroperator in terms of a single boundary value problem. The problem of computing the logarithm of a ratio of functional determinants for purely radial operators reduces to an infinite number of Gelfand-Yaglom problems, one for each l, whose solutions need to be summed (this is the first piece on the right hand side of (3.2.7)) and regularized (this is the second piece on the right hand side of (3.2.7)). The applicability of the formula requiresm ˆ (r) to vanish faster than r−2 at infinity, and these are the only types of deformations for which we will compute the wavefunction in the latter sections. When implementing the above formula we must sum up to a certain cutoff l = lmax which we take to be lmax = 45. A discussion of how the error decreases with lmax is given in appendix B.1.
3.3 Simple examples of radial deformations
The purpose of this section is to exploit the general formula (3.2.7) for a simple set of radial functions. By studying ΨHH as a functional ofm ˆ (r) we can identify some qualitative features already observed in [78], such as regions where the wavefunction oscillates and grows exponentially, as well as some new ones. Furthermore, we can study its dependence on more detailed features of the localized deformation. The zeroes of the wavefunction in theσ ˆ-basis occur only when the effective potential V (r) = l(l + 1)/r2 +m ˆ (r) of the differential operator −∇2 +m ˆ (r) is negative for eff R3 some range of r. If the effective potential were positive for all r it could not have vanishing eigenvalues, and hence the wavefunction could not vanish. Thus, we expect all oscillations of the wavefunction in theσ ˆ-basis to occur in directions wherem ˆ (r) is negative for some range of r. Assessing the magnitude of the wavefunction as a functional ofm ˆ (r) is a more complicated task. We observe that the wavefunctional acquires increasingly high local R ∞ maxima between its oscillations only in regions where the quantity Imˆ ≡ 0 dr r mˆ (r) appearing in (3.2.7) becomes large and negative. CHAPTER 3. HIGHER SPIN DS/CFT 44
Fig. 3.1: Examples of the radial deformations (3.3.1) on the left, (3.3.2) in the middle and (3.3.3) on the right. We have suppressed the polar coordinate θ of the S2 but kept the azimuthal direction.
It is important to note that because we are working with the flat metric on R3, which has no scale, our functional determinants will have an associated scaling sym- metry given by r → r/λ andm ˆ (r) → mˆ (r/λ)/λ2. We should thus fix the scaling when studying the functional determinant/wavefunction.
3.3.1 Single Gaussian
We first considerm ˆ (r) to be given by a general single Gaussian profile:
e−r2/λ2 Z ∞ A mˆ (r) = A 2 , dr r mˆ (r) = . (3.3.1) λ 0 2
See the left panel of figure 3.1 for an illustration of this deformation. Using equation
(3.2.7) we can explore |ΨHH (λ, A)|, where from now on whenever we write ΨHH it is implied as a late time wavefunctional. Using the scaling relation r → r/λ˜ andm ˆ (r) → mˆ (r/λ˜)/λ˜2 we can set λ = 1. In figure 3.2 we show a plot of the functional determinant. We immediately notice the same qualitative feature that was present for the constant mass deformation on a round S3 (displayed later on in figure 3.12). Namely, it oscillates and grows exponentially in the negative A direction. This is somewhat expected since our deformation is qualitatively similar to the mass deformation on the flat metric on R3 one gets by the conformal transformation of a constant mass on S3 (see appendix B.1). In particular, all oscillations occur for A < 0 CHAPTER 3. HIGHER SPIN DS/CFT 45
30
25
20
15
10
5
-20 -15 -10 -5 0
2 Fig. 3.2: Plot of |ΨHH (λ, A)| for N = 2 for the Gaussian profile (3.3.1) with λ = 1 using lmax = 45. The solid blue line is an interpolation of the numerically determined points (shown in red). The wavefunction grows and oscillates in the negative A direction.
and the magnitude of the local maxima increases for increasing |A| for fixed λ.
3.3.2 Gaussian Ring
We can also study the functional determinant of a profile of the type:
mˆ (r) = A e−(r−a)2/λ2 r2 , ∞ Z A λ h 2 2 √ haii dr r mˆ (r) = 2e−a /λ λ a2 + λ2 + a π 2a2 + 3λ2 1 + Erf , 0 4 λ (3.3.2) √ which describes a Gaussian-like ring peaked around r ∼ 1/2 a + a2 + 4λ2. Erf[x] denotes the error function. See the middle panel of figure 3.1 for an illustration of this deformation. The factor of r2 is included to ensure that the profile is continuously differentiable near the origin. Again, using the scaling relation r → r/λ˜ andm ˆ (r) → mˆ (r/λ˜)/λ˜2 we either fix the value of λ, |a| or |A|. We show an example in figure 3.3 where we have fixed the value of a. CHAPTER 3. HIGHER SPIN DS/CFT 46
3.0
2.5
2.0
1.5
1.0
0.5
0.0 0.5 1.0 1.5
2 Fig. 3.3: Left: Density plot of |ΨHH (λ, a, A)| for N = 2 for the profile (3.3.2) as a function of A (vertical) and λ (horizontal) for a = 5 using lmax = 45. Again, the wavefunction grows and oscillates in the negative A and positive λ directions. Right: 2 Plot of |ΨHH (λ, a, A)| for the profile (3.3.2) as a function of λ for A = −0.022 and a = 5. CHAPTER 3. HIGHER SPIN DS/CFT 47
2 Fig. 3.4: Left: Density plot of |ΨHH (λi, ai,Ai)| for N = 2 for the double Gaussian profile (3.3.3) as a function of A1 (x-axis) and A2 (y-axis) for a1 = 0, a2 = 5, λ1 = λ2 = 1 using lmax = 45. The wavefunction grows and oscillates for negative 2 A1 and A2. Right: Density plot of |ΨHH (λi, ai,Ai)| for the double Gaussian profile (3.3.3) as a function of λ1 (x-axis) and λ2 (y-axis) with A1 = −1, A2 = −1/100, a1 = 0 and a2 = 5.
3.3.3 Double Gaussian
As a third example we consider a double Gaussian profile:
2 2 2 2 2 −(r−a1) /λ −(r−a2) /λ mˆ (r) = r A1 e 1 + A2 e 2 . (3.3.3)
See the right panel of figure 3.1 for an illustration of this deformation. An example of 2 |ΨHH (λi, ai,Ai)| with a1 = 0 is shown in figure 3.4. Once again we observe a pattern of maxima encircled by regions where the wavefunction squared vanishes identically.
Furthermore, the wavefunction grows for increasingly negative values of A1 and A2. CHAPTER 3. HIGHER SPIN DS/CFT 48
3.4 Radial deformations of flat R3 and pinching limits
In this section, we introduce and study a class of SO(3) preserving deformations of the flat metric on R3. We show that they are conformally equivalent to the flat metric on R3. Thus, the wavefunction can only depend on such deformations of the metric if we also turn on a radial mass mb(r). Turning on such a mass, we can then perform the symmetry transformation discussed at the end of section 3.2.1 to get a mass deformationm ˆ (r) on the flat metric on R3. We will pick a functional form that is related by the symmetry transformations of 3.2.1 to a constant mass deformation on a deformed three-sphere geometry (see appendix B.1 for details). Depending on the sign of a parameter, the deformed three-sphere will look either like a peanut or an inverse peanut, i.e. like bulbous pears inverted relative to one another and conjoined on their fatter ends. The partition function that we compute can then also be understood as the answer for the partition function on this deformed three-sphere with a constant mass deformation. The pinching limit will be when the waist of the peanut-shaped geometry vanishes. We emphasize that how we perceive these deformations of the late time metric depends greatly on what we decide are natural constant time slices, since there always exists a conformal frame where the late time metric is the flat one. Ideally, it would be useful to analyze a qualitatively similar geometric deformation that would take the original geometry outside its conformal class, but we must restrict to the former case in this section since we will be constrained by considering SO(3) preserving defor- mations. Section 3.6 will go beyond this restriction by considering a new conformal class.
3.4.1 Balloon Geometry
Consider the following class of SO(3) preserving metrics defined on R3:
2 2 2 2 2 2 2 2 2 ds = dr + r fζ (r) dΩ2 , dΩ2 ≡ dθ + sin θdφ , (3.4.1) CHAPTER 3. HIGHER SPIN DS/CFT 49
Fig. 3.5: The “balloon” deformation of R3, defined by (3.4.1) and (3.4.3), represented schematically for positive ζ.
∗ with r ∈ [0, ∞). Consider a family of smooth functions fζ (r) with ζ ≤ ζ for positive ∗ ζ that tend to unity both at large r and near r = 0. We require that fζ (r) vanishes at some r = r∗ for the critical value ζ = ζ∗. We furthermore impose that:
2 d 2 2 lim r fζ (r) = 2 . (3.4.2) ζ→ζ∗ 2 dr r=r∗
For positive ζ, the geometries described by (3.4.1) can be pictured as a two-sphere whose size at some finite ζ < ζ∗ grows, then shrinks and subsequently grows again. 3 If fζ (r) = 1 the geometry is of course nothing more than the flat metric on R . As we approach ζ = ζ∗, the size of the two-sphere tends to vanish at r = r∗ eventually pinching the geometry into a warped three-sphere and a slightly deformed metric on R3. The choice (3.4.2) ensures there are no conical singularities at the pinching point. It would be of interest in and of itself to study geometries with conical singularities (see for example [119]). As a concrete example, we will take:
1 2 2 f (r)2 = 1 − ζ r2 + (γr)4 e−(r−a) /λ . (3.4.3) ζ ζ∗
The parameters a and λ are chosen and γ is tuned to obey the condition (3.4.2). Though ζ∗ is not an independent parameter it is useful to isolate in the expression. A schematic representation of this deformation, for positive ζ, is presented in figure CHAPTER 3. HIGHER SPIN DS/CFT 50
40 0.15
30 0.10 20 0.05 10
100 200 300 400 100 200 300 400 Fig. 3.6: Plot of g(x) (left) and dg(x)/dx (right) as obtained by numerically solving equation 3.4.4( ζ∗ = 1741.51, ζ = 9ζ∗/10, γ = 1.36612, a = −95, λ = 30).
3.5.
3.4.2 Conformal Flatness of Balloon Geometry
It is important to note that the geometry (3.4.1) is conformally flat. This can be shown in a straightforward fashion. Consider a coordinate transformation r = g(x). It immediately follows that if the following ordinary differential equation:
dg(x) x = g(x)f (g(x)) , (3.4.4) dx ζ has a smooth solution for g(x) whose derivative is positive for all x > 0 then our metric becomes: dg(x)2 ds2 = dx2 + x2dΩ2 . (3.4.5) dx 2 Though we cannot solve the non-linear o.d.e analytically, we can easily evaluate it nu- merically and confirm for several cases that g(x) satisfies the necessary requirements. Hence, our metric (3.4.1) is indeed conformally equivalent to the flat metric on R3. In figure 3.6 we give a numerical example of this. This result is already of some interest even for the case of ordinary Einstein grav- ity. It informs us that upon conditioning that all other fields vanish at late times, the absolute value of the late time Hartle-Hawking wavefunction, |ΨHH [gij]|, is indepen- dent of any radial SO(3) preserving deformation of the late time metric. Indeed, from CHAPTER 3. HIGHER SPIN DS/CFT 51
the bulk perspective a smooth conformal transformation of the late time three-metric can be induced by a time diffeomorphism that preserves the Starobinski-Fefferman- Graham form. From a holographic perspective, the late time wavefunction is com- puted by the partition function of a three-dimensional conformal field theory and thus only depends on the conformal metric (recall there are no conformal anomalies in three dimensions).
3.4.3 Wavefunctions and balloon geometries
We now examine what happens to the functional determinant as we vary the waist parameter ζ for an example. We will also turn on a mass that would correspond to a uniform mass m on the deformed three-sphere (as discussed further in appendix B.1.3), which upon the conformal transformation discussed there becomes:
2 2 m (r) = m . (3.4.6) b 1 + r2
This is the mass deformation on the balloon geometry. The final deformationm ˆ (x), to be used in the Dunne-Kirsten formula, is obtained by performing a conformal 3 2 2 2 2 rescaling of the balloon geometry to the flat metric on R : ds = dx + x dΩ2. This requires a conformal rescaling of mb(r) to:
dg(x)2 2 2 mˆ (x) = m . (3.4.7) dx 1 + r(x)2
Thus, we will study the functional determinant as a function of m and ζ. In figures 3.7 and 3.8 we display our numerical results. As expected, at m = 0, nothing changes as we vary ζ since the balloon geometries are conformally flat. However, when we turn on m 6= 0 the wavefunction becomes sensitive to changes in ζ. Interestingly, decreasing the girth of the throat while keeping everything else fixed is favored, at least near m = 0. Thus, though supressed exponentially with respect to the local maximum of the wavefunction at m = 0, the wavefunction does not vanish in the pinching limit. It is tempting to speculate that such pieces of the wavefunction might CHAPTER 3. HIGHER SPIN DS/CFT 52
be connected to the fragmentation picture of [120].
3.5 Spherical harmonics and a conjecture
In this section, we present numerical evidence that when mapping the problem back to the three-sphere (using the discussion in appendix B.1), all profiles give a normalizable wavefunction upon fixing their average value over the whole three-sphere. Thus, it is conceivable that the only divergence of the wavefunction occurs precisely for large and negative values of a uniform profile over the whole three-sphere [78]; a single direction in an infinite dimensional configuration space! For instance, as we shall show below, by mapping the Gaussian profile (3.3.1) to the three-sphere and removing the zero mode from its expansion in terms of three-sphere harmonics, the resultant profile produces a normalizable wavefunction as a function of its amplitude.
3.5.1 Three-sphere harmonics
We now study some examples of SO(3) invariant deformations which correspond to harmonics of the three-sphere conformally mapped back to R3. These harmonics are the eigenfunctions of the Klein-Gordon operator on the three-sphere with metric 2 2 2 2 th 2 ds = dψ + sin ψ dΩ2. The k harmonic (independent of the S coordinates) is given by:
Fk(ψ) = ck csc(ψ) sin[ (1 + k)ψ ] , k = 0, 1, 2,.... (3.5.1)
As explained in appendix B.1, to evaluate the partition function for this deformation using the method of Dunne and Kirsten we must first perform the coordinate trans- formation ψ(r) = 2 cot−1 r−1, and then scale the deformation by the inverse of the conformal factor that maps the three-sphere metric to the metric on R3. The final radial deformation which is to be used in the Dunne-Kirsten formula is:
2 sin[ 2(1 + k)tan−1r ] mˆ (r) = c . (3.5.2) k k (r + r3) CHAPTER 3. HIGHER SPIN DS/CFT 53
2 Fig. 3.7: Density plot of the |ΨHH (ζ, m)| for N = 2 as a function of the pinching parameter ζ (vertical axis) and an overall mass deformation m (horizontal axis) using lmax = 45, ζ∗ = 1741.51, γ = 1.36612, a = −95 and λ = 30.
20 ì 1.0ìæòà ìæòà ìæòà òìò à ìòà òìàìæà æ ìæòà æ æ æ ìà òà ò 0.8 ìæ 15 ò ìòà ò æ ò ìà æ ìà à æ ò ì ò æ ì ìà æ ò æ à à 0.6 à ì ò ò ò ìæ ì ìà 10 ò à æ æ òà æ à ìò ì æ à ò ìæ 0.4 æ ò ò æ à ò ìà ì à ì æ ì ì à ò ò òæ 5 à ìà ò æ à æ æ ò ìàæ 0.2 ìà à ò ìæ ì æ òà ò ì òìàæ ì ò ì ò òæà æàìæòàìæòàìæòà à àæ òìàæ ìò ìæòàìæòàìò ìæòàìæòàìæòàìæòà æ æ òìòìòìòàìæòàìæòàìæòàìæòàìæòàìæòàìæòàìæàæàæ ìæòàìæòàìæòàìæòà ìæòàìæòàìæòàìæòàìò ìòà -5 -4 -3 -2 -1 1 -1.0 -0.5 0.5 1.0
2 ∗ Fig. 3.8: Left: |ΨHH (ζ, m)| for N = 2 as a function of m for ζ = −ζ /2 (red dots), ζ = 0 (blue squares), ζ = ζ∗/2 (green diamonds) and ζ = 9ζ∗/10 (black triangles) using lmax = 45, ζ∗ = 1741.51, γ = 1.36612, a = −95 and λ = 30. Right: Same as left but for different plot range. CHAPTER 3. HIGHER SPIN DS/CFT 54
Taking k = 0 corresponds to the zero mode which has been previously studied in [78] and was found to be oscillatory and divergent as the coefficient ck goes to large negative values. In figures 3.9 and 3.10 we plot the partition functions for higher harmonics as a function of the coefficient ck. We notice that they are all well-behaved and normalizable, at least in the range we have explored. This motivates us to consider deformations which are linear combinations of spherical harmonics. Having looked at a deformation that is the linear combination of the zero mode with the first harmonic, as well as a deformation that is the linear combination of the first harmonic with the second harmonic, we notice that the partition function is not divergent so long as the coefficient of the zero mode is kept fixed. Postponing a more systematic study for the future, here we simply consider a modified version of the single Gaussian deformation given in (3.3.1). Previously, we had found that as the overall coefficient of the profile becomes large and negative, the partition function diverges. The new profile we will study is obtained by mapping the Gaussian profile to the three-sphere, subtracting off its zero mode, and mapping it back to R3. Notice that the Gaussian profile mapped to the three-sphere is con- structed from an infinite number of harmonics, and here we are subtracting the piece that seems problematic from our analysis of harmonics and finite linear combinations thereof. Since F0(ψ) = c0 is simply a constant, the condition to be met is:
Z π r(ψ)2 + 12 dψ mˆ (r(ψ)) sin2 ψ = 0 , (3.5.3) 0 2
ψ −r2 2 2 where r(ψ) = tan( 2 ) andm ˆ (r) = Ae − 4a1/(r + 1) . In this case the integral can be done explicitly and we can solve for the coefficient a1 analytically. The final form of the single Gaussian radial deformation orthogonal to the zero mode of the three-sphere becomes: √ 2 8A (1 − e π Erfc[1]) mˆ (r) = A e−r − √ . (3.5.4) π (1 + r2)2
We plot the functional determinant as a function of A in figure 3.11. Interestingly, the partition function is once again well-behaved for large values of A. An analogous CHAPTER 3. HIGHER SPIN DS/CFT 55
1.0 -10 -5 5 10 0.8
0.6 -5
0.4 -10 0.2
-10 -5 5 10 -15
2 3 Fig. 3.9: Left: Plot of |ΨHH (c1)| for N = 2 for the first harmonic mapped to R given 2 in (3.5.2) using lmax = 45. Right: Plot of log |ΨHH (c1)| for N = 2 using lmax = 45.
-40 -20 20 40 -40 -20 20 40 -40 -20 20 40 -50 -50 -100
-100 -100 -200
-150 -150 -300
-40 -20 20 40 -40 -20 20 40 -40 -20 20 40 -20 -50 -40 -50 -60 -100 -80 -100 -100 -150 -120 -150 -140 -200
-40 -20 20 40 -40 -20 20 40 -40 -20 20 40
-50 -50 -50
-100 -100 -100
-150 -150 -150 -200 -200
Fig. 3.10: Plot of log |ΨHH (ck)| for N = 2 for the first nine spherical harmonics 3 mapped to R given in (3.5.2) using lmax = 45. Notice that only the zeroth harmonic is non-normalizable in the negative c0 direction. CHAPTER 3. HIGHER SPIN DS/CFT 56
1.0 -100 -50 50 100 0.8 -5
0.6 -10
0.4 -15 - 0.2 20 -25 -100 -50 50 100
2 Fig. 3.11: |ΨHH (A)| (left) and log|ΨHH (A)| (right) for N = 2 as a function of A, the overall size of the radial deformation in (3.5.4) which is constructed to be orthogonal to the zero mode of the three-sphere (lmax = 45).
analysis for the balloon geometries, where we fix the zero-mode on the conformally related three-sphere, also results in the boundedness of the partition function in the ζ direction (ζ defined in (3.4.3)). The above results motivate the conjecture:
The partition function of any SO(3) symmetric “radial” deformation for which the three-sphere zero mode harmonic is fixed is bounded.
Notice that the three-sphere is chosen as the geometry in the conformal class on which to fix the uniform profile. For example, in the case of the peanut geometries one can show that fixing the uniform profile on the peanut is not sufficient to ensure that the partition function is bounded (though fixing some non-uniform profile would suffice). This is simply because keeping the uniform mass profile fixed to some neg- ative value while taking ζ large and negative, which corresponds to the conformally related sphere getting fatter at the waist, implies that the uniform profile on the sphere is getting large and negative. Consistent with the rest of our observations, we see that the partition function is unbounded in this direction. Furthermore, the next section will provide evidence for a more general conjecture that would extend beyond the conformal class of the sphere. CHAPTER 3. HIGHER SPIN DS/CFT 57
3.6 Three-sphere squashed and massed
In this section, we would like to briefly revisit and extend some of the observations of [78] for a constant mass deformation on an S3. There, it was observed that in theσ ˆ-basis the wavefunction on an S3 with a uniform mass deformation oscillated and diverged at large negative mS3 . We explore in this section a new direction which is the squashing parameter of the round metric on the three-sphere (in the presence of a non-zero uniform scalar profile) and its effect on the zeroes and maxima. Unlike the previous SO(3) preserving deformations which were inhomogeneous, this SO(3) × U(1) preserving deformation is homogeneous yet anisotropic. Furthermore, the squashed three-sphere is not conformal to the ordinary round three-sphere. In fact, squashed spheres with different values of the squashing parameter belong to distinct conformal classes. Part of our motivation is to provide further evidence that the zeroes of the wave- function are extended and that the local maxima of the wavefunction (other than the pure de Sitter one) will no longer necessarily peak about homogeneous and isotropic geometries. Additionally, and in the same spirit as the observations made in section 3.5, we find that upon fixing the value of the uniform profile over the whole squashed three-sphere the wavefunction is normalizable in the squashing direction.
3.6.1 Squashed and massed
Consider turning on a constant mass mS3 for the free Sp(N) model on the round metric on an S3 whose radius a is fixed to one unless otherwise specified. The partition function is given by [78]:
m Z S3 √ √ −1 1 3ζ(3) π π 3 N log Zfree[mS ] = log 4 − 2 − dσ 1 − 4σ cot 1 − 4σ . 16 π 8 0 2 (3.6.1)
The analyticity of Zfree[mS3 ] in the complex mS3 -plane is ensured by the uniform convergence of the (regularized) infinite product of analytic eigenvalues which defines it. For physical applications we restrict to real mS3 and note that this implies that 0 the Taylor series expanded about any mS3 = mS3 converges to the partition function CHAPTER 3. HIGHER SPIN DS/CFT 58
20
15
10
5
-3 -2 -1 1 2 3
2 Fig. 3.12: Plot of |ΨHH [mS3 ]| given by expression (3.6.1) for N = 2. above. Furthermore, the partition function has a zero if and only if one of the product eigenvalues in the functional determinant vanishes, which can only happen for mS3 < 0. As shown in figure 3.12, this wavefunction grows exponentially and oscillates in 4 the negative mS3 direction. It is worth noting that the mS3 dependent part of the phase of the wavefunction vanishes. The real part of log Zfree/N goes as |mS3 | for 3/2 large negative mS3 and −(mS3 ) for large positive mS3 .
It is worth studying what happens to the zeroes and local maxima of Zfree in the presence of an additional deformation. A computationally convenient deformation is to squash the round metric on S3 into that of a squashed sphere, which is a homoge- neous yet anisotropic geometry. In this sense, this deformation is complementary to the inhomogeneous deformations we have been studying so far. We review the metric and eigenvalues of the squashed sphere with squashing parameter ρ in appendix B.2. Our method of regularization is a straightforward extension of heat kernel techniques used in section 3.2 of [78], and details can be found therein. In figure 3.13 we present a plot of the wavefunction as a function of the mass mS3 and the squashing parameter ρ (the round metric on S3 occurs at ρ = 0).
4Divergences of the Hartle-Hawking wavefunctional have been discussed in other circumstances such as Einstein gravity coupled to a scalar field with a quadratic scalar potential and vanishing cosmological constant [121] or the wavefunction of dS3 on a toroidal boundary [122]. Their physical interpretation remains to be understood. For example, it might be a result of very sharp conditioning or an indication of an instability. CHAPTER 3. HIGHER SPIN DS/CFT 59
20
15
10
5
-2 -1 1 2 3
2 Fig. 3.13: Left: Density plot of |ΨHH [ρ, mS3 ]| for N = 2. The fainter peak centered at the origin reproduces perturbation theory in the empty de Sitter vacuum. Horizontal lines are lines of constant mS3 and vertical lines are lines of constant ρ. Right: Plot 2 of |ΨHH [ρ, −2.25]| for N = 2. Notice that it peaks away from ρ = 0.
We find that the local maxima are in general pushed away from ρ = 0 and the zeroes of the wavefunction are extended to enclose the local maxima. The fact that the zeroes are extended (codimension one in the (ρ, mS3 ) plane) is perhaps not sur- prising given that the zeroes in figure 3.12 arise from ΨHH [mS3 ] (which is purely real) changing sign. This feature should not disappear, at least for small perturbations to the equation which determines ΨHH [mS3 ]. Furthermore, the maxima of figure 3.12 are pushed even higher when squashing is allowed. This can be seen in the right hand side of figure 3.13, where one notices that the second local maximum of ΨHH [mS3 ] at ρ = 0 is pushed further up to some ρ > 0. This observation strongly suggests that away from the origin, and under such extreme conditioning of the late time profiles of the bulk fields, the wavefunction peaks in regions where the metric and perhaps even the higher spin fields are highly excited. On the other hand, the perturbative de Sitter saddle centered at the origin remains a true local maximum. CHAPTER 3. HIGHER SPIN DS/CFT 60
-1 1 2 3
-20
-40
-60
2 Fig. 3.14: Left: Density plot of |ΨHH [ρ, mS3 ]| for N = 2 for a slightly larger range. Notice the local de Sitter maximum visible in figure 3.13 is already too faint to be seen. Horizontal lines are lines of constant mS3 and vertical lines are lines of constant ρ. Right: Plot of 2 log |ΨHH [ρ, −4.5]| for N = 2.
3.7 Basis change, critical Sp(N) model, double trace deformations
In this section we further discuss the transformation to the field basis and clarify the role of double trace deformations. We would like to comment on the transform to the √ bulk field basis, which gives us the wavefunctional as a functional of ν ≡ Nσ˜ (with ν defined in (3.2.2) andσ ˜ being related the source of the single-trace operator dual to the bulk scalar), in the large N limit. As discussed in [78], one has to consider the free theory deformed by a relevant double trace operator f(χ · χ)2/(8N). We also keep a source −ifσ˜ turned on for the single-trace χ · χ operator. The parameter f ∈ C has units of energy. Performing a Hubbard-Stratonovich transformation by introducing an auxiliary scalar field σ we CHAPTER 3. HIGHER SPIN DS/CFT 61
find:
1 Z √ Nσ2 S(f) = d3x g Ω ∂ χA∂ χBgij − ifσχ˜ AχB − − σ Ω χAχB . 2 AB i j f AB (3.7.1) Integrating out the χA fields, the partition function becomes:
√ Z Z 2 (f) − Nf R d3x g σ˜2 3 √ σ Z [˜σ] = e 2 Dσ exp N d x g + iσσ˜ Z [σ] , (3.7.2) crit 2f free where we have transformed variables σ → (σ − ifσ˜). Having interpreted Zfree[σ] = hσ|Ei = ΨHH [σ], we see that the above expression is a Fourier type transform of the σˆ-basis wavefunction. In order for this to become the actual basis changing transform (3.1.4) to the eigenbasis of the field operator, the constant f must be taken to infinity. The f → ∞ limit (where we are keeping the size of the three-sphere fixed) corresponds to sending the ultraviolet cutoff (of the infrared fixed point theory) to infinity, in the same sense as [123], or roughly speaking it corresponds to the late time limit in the bulk.5 We begin by performing a perturbative analysis for infinitesimal deformations of the free Sp(N) theory at large N on an R3.
3.7.1 Perturbative analysis on R3
For the sake of simplicity, we will put the theory on the flat metric on R3, akin to studying perturbations in a small piece of I+.6 For reasons that will be clear
5Another way to think about this is keeping f fixed and taking the large size limit of the three- sphere that the CFT lives on. This is because the dimensionless quantity is |f|a where a is the size of the sphere. Indeed at late times the three-sphere grows large. 6The parallel story in anti-de Sitter space has been studied extensively [80, 124–126]. From the 2 2 2 2 2 bulk perspective in planar AdS4: ds = `A(dz + d~x )/z , at least perturbatively about the empty AdS4 vacuum, the finite fA ∈ R double trace deformed theory computes correlation functions of the bulk scalar quantized with mixed boundary conditions. Near the boundary z → 0 of AdS4, the 2 2 2 bulk scalar with mass m `A = −2 behaves as φ(z, ~x) ∼ α(~x)z + β(~x)z with α(~x) = fAβ(~x). This boundary condition is different from the conformally invariant one which sets either α(~x) or β(~x) to zero, corresponding to the free or critical O(N) models respectively. In de Sitter space we would consider a wavefunction of a scalar of mass m2`2 = +2 computed by imposing future boundary 2 conditions [16,17,127] φ(η, ~x) ∼ α(~x)η + β(~x)η (with α(~x) = fDβ(~x)) and Bunch-Davies conditions φ ∼ eikη for k|η| 1. At the level of perturbation theory this is computed by continuing the CHAPTER 3. HIGHER SPIN DS/CFT 62
momentarily we choose f = i|f| to be pure imaginary. We are interested in setting up a perturbative expansion about the σ ∼ 0 Gaussian peak of Zfree[σ]. From (3.7.2) we can compute the two point function of O ≡ χ · χ at large N in the double trace deformed Sp(N) theory on an R3. We do this by taking two variational derivatives of the logarithm of (3.7.2) with respect toσ ˜(xi) and evaluating atσ ˜(xi) = 0:
|f|2G(k) N hO O i = N ,G(k) ≡ hO O i = − , (3.7.3) ~k −~k f N + 2i|f|G(k) ~k −~k f=0 k where k is the magnitude of the momentum. For k |f| this becomes the two-point function of the free Sp(N) model, whereas for k |f| this becomes the two-point function of the critical Sp(N) model. Expanding for large |f| we find:
N|f| N hO O i = −i − k + ... (3.7.4) ~k −~k f 2 4
Notice that the real part of the two-point function (3.7.4) is negative. Recalling (3.1.3), this is the Gaussian suppression of the Hartle-Hawking wavefunction near the de Sitter vacuum. Also, the local momentum independent term has become a phase for pure imaginary f. We can compare this to the bulk Hartle-Hawking wavefunction for a free m2`2 = +2 scalar in planar coordinates computed in (B.3.13) −1 of the appendix. This allows us to (roughly) identify |f| with the late time cutoff |ηc| at large |f|. In a similar way we can compute the rest of the perturbative correlators of the critical Sp(N) theory [110]. Beyond such perturbative analyses, we must resort to a saddle point approxima- tion which we now proceed to.
3.7.2 Large N saddles for uniform S3 profiles
We now put the theory on the round metric on S3. At large N, we can evaluate (3.7.2) by solving the saddle point equation (for σ = σ(˜σ) andσ ˜ uniform over the
Euclidean AdS4 partition function by z → −iη, `A → i` and fA → ifD (where fD ∈ R for the ‘normalizable’ profile of the scalar field to be real). CHAPTER 3. HIGHER SPIN DS/CFT 63
1.0 2.0 8
0.8 1.5 6 0.6 1.0 4 0.4 0.5 2 0.2
-2 -1 1 2 -2 -1 1 2 -2 -1 1 2
Fig. 3.15: Plots of the Nth root of |ΨHH [˜σ, Σ0]|, |ΨHH [˜σ, Σ1]|, |ΨHH [˜σ, Σ2]| at large f from left to right. Notice that the higher saddles dominate nearσ ˜ = 0 but fall off faster for largeσ ˜.
whole three-sphere):
16πσ √ π √ + 16πiσ˜ = 1 − 4σ cot 1 − 4σ . (3.7.5) f 2
For a given solution Σi of (3.7.5), we can evaluate Zcrit[˜σ, Σi]. For example, there is a solution Σ0 where σ ∼ 0 whenσ ˜ ∼ 0 (with f → ∞). It is the piece of the wavefunctional evaluated from this saddle nearσ ˜ = 0 that reproduces the dS invariant perturbation theory in the Bunch-Davies state (about the pure de Sitter vacuum).
In figure 3.15 we plot Zcrit for the first few Σi at large f = i|f|. These haveσ ˜ = 0 √ π near the subsequent zeroes of cot 2 1 − 4σ in (3.7.5). Notice that for all large N saddles Zcrit[˜σ] is peaked atσ ˜ = 0 but the saddles coming from the more negative σ peaks contribute more nearσ ˜ = 0. Away from the large N limit we must compute the integral in (3.7.2) without resorting to a saddle point approximation. If we restrict to uniform σ andσ ˜, we note that as we increase f = i|f| more and more of the growing negative σ peaks in
Zfree[σ] contribute to the integral before it is cutoff by the rapid oscillations due to the iNσ2/|f| piece. One can check that the integral grows for large and negativeσ ˜ upon fixing |f|. Some of the saddles in the large N limit should correspond to classical (complex) bulk solutions with a uniform late time profile of the scalar on the round metric of the three-sphere. Some of these solutions labelled by a continuous parameter, which only involve the bulk metric and scalar, were found by Sezgin and Sundell in [111]. CHAPTER 3. HIGHER SPIN DS/CFT 64
It is worth noting that though it may sound confusing that there are bulk solutions that have only the metric and scalar turned on (since all higher spin fields interact on an equal footing), this is natural from the CFT since turning on an SO(4) symmetric source for J (0) = χ · χ need not source the traceless higher spin currents due to symmetry reasons. The metric is non-vanishing since in effect we have also turned on a source for it by having the round metric on S3 at the boundary. At finite N, all these saddles mix quantum mechanically. Each ΨHH [˜σ, Σi] comes with a phase, so one should be careful when summing contributions from different saddles. Finally, it would be extremely interesting to understand the Lorentzian cosmologies associated to the wavefunction using the ideas developed in [128, 129] (see also [130]). In order to have a classical cosmology (or an ensemble of such cosmologies) we must ensure that the wavefunction takes a WKB form with a phase oscillating much more rapidly than its absolute value. At least in the large N limit, and for large |f|, this is ensured by the first term in (3.7.2).
3.7.3 Double trace deformations as convolutions
We can also consider keeping f finite and real. This defines a double-trace deformed field theory, in and of its own right, whose partition function Z(f)[σ0] can be computed in the large N limit. We also keep a uniform (on S3) source σ0 turned on for the single-trace χ · χ operator. This partition function is no longer computing overlaps between the Bunch-Davies vacuum and some late time field configuration which is ˆ an eigenstate of the field operator φ. Given that Zfree[σ] = hσ|Ei = ΨHH [σ], we see that Z(f)[σ0] is computing instead a convolution of the wavefunction in theσ ˆ-basis:7
" # Z Z (σ0 − σ)2 Z(f)[σ0] = Dσ exp N dΩ Ψ [σ] . (3.7.6) 3 2f HH
7Though we will not do so here, we can study higher multitrace deformations of the Sp(N) theory and arrive at similar expressions. Thus, the fact that multi-trace operators are irrelevant seems far less threatening than having low-spin, single-trace operators which are irrelevant and correspond to bulk tachyons. CHAPTER 3. HIGHER SPIN DS/CFT 65
We can also view Z(f)[σ0] as computing the overlap of the Hartle-Hawking state with the state: " # Z Z (σ0 − σ)2 |fi ≡ Dσ exp N dΩ |σi . (3.7.7) 3 2f
Notice that though the integral itself is convergent for finite f, the resulting function Z(f)[σ0] will grow exponentially at large negative σ0. One could also consider more generally a complex valued f ∈ C which would correspond to a kind of windowed Fourier transform.
3 3.7.4 Euclidean AdS4 with an S boundary
3 The situation can be contrasted with the case of Euclidean AdS4 (with an S bound- O(N) 3 ary) [126]. The partition function Zcrit [σA] of the critical O(N) model on an S , dual to Euclidean AdS4 in higher spin gravity, is obtained again from the free O(N) model by a double trace deformation in the limit fA → +∞:
Z Z 2 O(N) NfA R 2 σ O(N) 2 dΩ3σA Zcrit [σA] = lim e Dσ exp N dΩ3 − σAσ Zfree [σ] , fA→∞ 2fA
fA ∈ R . (3.7.8)
The first term in front of the integral is local in the limit fA → +∞ and we can remove it by adding a counterterm. To ensure convergence of the integral we must choose an appropriate contour, which in this case is given by σ running along the O(N) imaginary axis (see for example [78]). Zfree [σ] is the partition function of the free O(N) model and is related to the free Sp(N) partition function by N → −N. Note O(N) that Zfree has poles precisely at the values where the wavefunction in theσ ˆ-basis (3.6.1) vanishes. At large N the integral (3.7.8) can be evaluated by a saddle point O(N) approximation. In figure 3.16 we display Zcrit [σA] for the case of a uniform source 3 fAσA over the whole S . In principle, this plot should be reproducible by computing the regularized on-shell Vasiliev action on the asymptotically Euclidean AdS4 Sezgin- Sundell solution [111]. CHAPTER 3. HIGHER SPIN DS/CFT 66
1.0
0.8
0.6
0.4
0.2
-0.4 -0.2 0.2
th O(N) Fig. 3.16: Plot of the N root of the finite part of Zcrit [σA] for a uniform source σA O(N) over the whole three-sphere. We have normalized such that Zcrit [σA] = 1 at σA = 0.
3.8 Extensions of higher spin de Sitter hologra- phy?
So far, our discussion was restricted to the minimal bosonic higher spin theory. A natural question that arises, particularly given possible interpretational issues of the wavefunction such as its (non)-normalizability, is whether this theory is part of a larger framework. We briefly discuss possible extensions of higher spin de Sitter holography, inspired by the analogous situation in anti-de Sitter space.
3.8.1 AdS4
There exist parity violating deformations of the bulk equations of motion which de- form the original Vasiliev equations (in anti-de Sitter space) to a one-parameter fam- ily [65, 66, 104]. It was proposed that the dual description is given by coupling the theory to a Chern-Simons theory with level k, at least for simple enough topologies. The new parameter is given by the ’t Hooft coupling λ = N/k which is small when the dual is higher spin gravity. Such a theory was shown in the infinite N limit with fixed small λ [131–133] to have a spectrum of single-trace operators which is precisely that of the free U(N) model, namely a tower of higher spin currents which are conserved CHAPTER 3. HIGHER SPIN DS/CFT 67
√ (up to O(1/ N) corrections), in accordance with a bulk higher spin theory. As discussed in [132, 134] in the context of anti-de Sitter higher spin gravity, one can also endow the bulk higher spin fields with U(M) Chan-Paton factors, such that the fields all lie in the adjoint representation of the U(M). In the dual theory, this corresponds to adding a U(M) flavor symmetry to the U(N) model. If the flavor symmetry is weakly gauged, which can be achieved through the procedure described in [135], then one can form single-trace operators of the form Tr (ABAB . . . AB). The fields A and B transform in the (, ) and (, ) of the U(N) × U(M) gauge group respectively. As M/N increases the ‘glue’ between each TrAB (which are dual to higher spin fields in the bulk) becomes stronger. From the bulk point of view, it was suggested that the higher spin fields, now endowed with additional U(M) interactions, form bound states with binding energy that increases as we crank up M/N. These bound states would correspond to the Tr (ABAB . . . AB) operators. An appropriately supersymmetrized version of this story [132, 134] was conjectured to connect the higher spin (supersymmetric) theory to the ABJ model [136] where such long string operators are dual to bulk strings.
3.8.2 dS4
It is convenient in our discussion for the bulk to contain a spin-one gauge field in its spectrum. We thus consider the non-minimal higher spin model with even and odd spins whose dual (at least at the level of correlation functions on R3) is a free/critical U(N) theory with N anti-commuting scalars transforming as U(N) vectors. We refer to this as the U˜(N) model. Given that the parity violating deformations of the Vasiliev equations in AdS retain the original field content and reality conditions, it seems natural that they are present for the de Sitter theory as well. Thus, one might consider that such theories are dual to parity violating extensions of the U˜(N) theory obtained by adding a level k Chern-Simons term to the free anti-commuting complex scalars. The Lagrangian CHAPTER 3. HIGHER SPIN DS/CFT 68
of this theory is:
ik Z 1 Z S = − d3x Aa∂ Aa + f abcAaAbAc + d3x ∇ χA∇iχ¯A¯ . (3.8.1) CSM 4π ijk i i k 3 ijk i j k i
a A The fields Ai are (possibly complexified) U(N) gauge fields, the χ fields are anti- commuting complex scalars transforming in the fundamental of the U(N) gauge sym- metry, and the ∇i derivative is covariant with respect to U(N) gauge transformations. Though classically this theory is conformally invariant, this need not be the case when we include loops, as the β-function might be non-vanishing. The ordinary U(N) Chern-Simons theory coupled to a vector of charged scalars has two exactly marginal deformations at infinite N as in [131, 137, 138]. These are the ’t Hooft A A¯ 3 coupling λ ≡ N/k and the coupling constant λ6 of the triple trace interaction (χ χ¯ ) . Though the Chern-Simons-U˜(N) theory (3.8.1) is non-unitary, it is conceivable that it is also has a vanishing β-function at large N [139]. This theory would also have an unchanged spectrum of single-trace operators at large N and small λ by the same arguments as those in [131,140]. 8 One can also consider adding U(M) Chan-Paton factors to the bulk higher spin de Sitter theory. This merely requires tensoring the ∗ algebra with that of M × M matrices. Once again, this will not affect the reality conditions on the higher spin fields and they will all transform in the adjoint of the U(M). This corresponds to adding a U(M) flavor symmetry to the U˜(N) vector model, which can be weakly gauged (see appendix B.4 for a discussion). The single-trace operators Tr (ABAB . . . AB) have increasingly real conformal weight. From the point of view of dS/CFT this would imply that the bulk theory has a tower of tachyonic bulk fields since the conformal
8 1 At finite k, the partition function on a non-trivial topology such as M = S × Riemmg will 2 grow as [141] (see also [142]): ZCFT ∼ exp (g − 1)N log k + O(N) where g is the genus of the M. This drastically favors higher topologies if interpreted as a probability, but it is unclear whether and how one should compare topologies and what the correct normalization for ΨHH is. It is interesting that at finite k one might also encounter monopole operators. In ABJM, such operators are dual to D0-branes in the bulk. It is unclear how they should be understood in the context of de Sitter space and higher spin gravity. For instance, they have a conformal weight that goes like k, which might suggest taking k to be complex or imaginary [143] in the de Sitter case. It is also worth noting that the potentially infinite wealth of topological data at I+ might be at odds with the finiteness of de Sitter entropy [144]. In Einstein gravity adding too much topology at I+ often results in bulk singularities [145]. The issue of topology in the context of dS/CFT is further discussed in [146]. CHAPTER 3. HIGHER SPIN DS/CFT 69
p 2 2 weight of a bulk field goes as ∆± ∼ 3/2 ± 9/4 − m ` . One might suspect that these will be the continuations of the higher spin bound states previously discussed for the anti-de Sitter case. Thus we see that even though the fundamental constituents (i.e. the higher spin particles) of such an extension of higher spin de Sitter gravity are not pathological (at least at the level of perturbation theory), they may form configurations which resemble tachyonic fields in de Sitter space. It is of interest to understand whether the late time behavior of such a theory can ever be asymptotically de Sitter [139]. From the CFT point of view these are highly irrelevant operators which are not conserved currents. For there to be a late time de Sitter phase, one would require that turning on such irrelevant deformations can flow the theory to a UV fixed point. Chapter 4
Conformal Quivers
4.1 Introduction
We will now shift our attention to the study of quiver quantum mechanics. Our goal is to investigate the extent to which such models capture the complicated dynamics of specific AdS2 geometries and disordered systems. When studying the extremal limit of black hole solutions in gravity one immedi- ately encounters a rather unusual geometric feature. The geometry develops a throat which becomes infinitely deep at extremality. This can be seen in the simple example of a d-dimensional Reissner-Nordstr¨omblack hole whose metric is:
2 2 2 1 2 r dr 2 2 ds = − 2 (r − r+)(r − r−) dt + + r dΩd−2 . (4.1.1) r (r − r+)(r − r−)
The inner and outer horizon lie at r = r− and r = r+ > r−. In the extremal limit r+ tends to r− and the proper distance between some finite distance r and the horizon diverges. One can isolate the geometry parametrically near the horizon within the infinitely deep throat and find a new solution to the Einstein-Maxwell system, namely d−2 the Bertotti-Robinson geometry AdS2 × S . The AdS2 structure is not contingent upon supersymmetry but rather extremality, i.e. that the temperature of the horizon vanishes. Turning on the slightest temperature will destroy the precise AdS2 structure and bring the horizon back to a finite distance. In the extremal near horizon region,
70 CHAPTER 4. CONFORMAL QUIVERS 71
the R × SO(d − 1) isometry group is enhanced to an SL(2, R) × SO(d − 1). While the Hawking temperature of the black hole vanishes, the entropy, given by one-quarter of the size of the horizon in Planck units, is still macroscopically large. It is interesting that a full SL(2, R) symmtetry (rather than just a dilatation symmetry) emerges in the near horizon region given that we are only losing a single scale, the temperature. From a holographic point of view, the surprise stems from the fact that in quantum mechanics invariance under dilatations does not generally imply invariance under the full SL(2, R) conformal transformations [147]. We can study such extremal, and in addition supersymmetric, black holes in string theory. They are given by wrapping branes around compact dimensions such that they are pointlike in the non-compact directions. The electric and magnetic charges of these branes are (schematically) given by the number of times they wrap around various compact cycles. A simple and beautiful example [148] is given by taking type 1 4 IIA string theory compactified on a small S ×T and wrapping N4 D4-branes around 4 the T and sprinkling N0 D0-particles on top. In the large N0 and N4 limit, this gives 3 an extremal D0-D4 black hole with an AdS2 × S near horizon. Using T -duality on the S1, we can map this system to a D1-D5 system in type IIB string theory. Since the circle becomes effectively non-compact in the IIB frame, our former black 3 hole now becomes a black string. This black string has an AdS3 × S near horizon with an SL(2, R)L × SL(2, R)R isometry group. We can express the AdS3 as a Hopf fibration of the real line over an AdS2 base space. The SL(2, R) isometry of the AdS2 base space is the left (or right) moving part of the isometries of the AdS3. The real line becomes the T -dual direction and thus we see how the original AdS2 naturally
fits inside the AdS3. From the point of view of the AdS/CFT correspondence, the
AdS3 is dual to a certain two-dimensional CFT and the large black hole degeneracy is counted by the supersymmetric ground states of the CFT which only excite one of the two SL(2, R)’s. The isometry of the near horizon supersymmetric AdS2 reflects the remaining SL(2, R). There is another class of extremal black holes which arise in string theory which can be obtained by considering a Calabi-Yau compactification, with generic SU(3) holonomy, of eleven dimensional M-theory down to five non-compact dimensions. CHAPTER 4. CONFORMAL QUIVERS 72
Typically, such a Calabi-Yau will not have U(1) isometries that we can easily T -dualize along. Yet, wrapping branes around the supersymmetric cycles of the Calabi-Yau yields extremal five-dimensional black holes [149,150], which again have an SL(2, R) isometry in the near horizon. This AdS2 does not naturally fit inside an AdS3 with a two-dimensional CFT dual.1 Remarkably, attempts to count the microstates of such supersymmetric black holes have faced significant difficulties [157]. In fact, whenever the precise counting of microstates has been successful it has involved a Cardy formula [148,158,159]. Lacking the larger Virasoro structure, one may wonder where the ‘isolated’ SL(2, R) of these black holes originates and to what extent it is robust.
Perhaps an additional motivation for understanding such an ‘isolated’ AdS2 ge- ometry is the emergence of an SL(2, R) symmetry in the worldline data of the static patch of de Sitter space [45, 59]. It is interesting to note that the static patch of 2 four-dimensional de Sitter space is conformally equivalent to AdS2 × S , whose ‘iso- lated’ SL(2, R) does not seem to reside within a larger structure containing a Virasoro algebra. One particular way the SL(2, R) isometeries of the black hole manifest themselves is in the worldine dynamics of D-particles propagating in the near horizon region. We might then ask whether there are microscopic models, such as matrix quantum mechanics models with a large number of ground states, whose effective eigenvalue dynamics describe an SL(2, R) invariant multiparticle theory.2 Some insight into these issues can be provided by studying certain quiver quan- tum mechanics models, which capture the low energy dynamics of strings connecting a collection of wrapped branes [164–170] in a Calabi-Yau compactification of type IIA string theory to four-dimensions. Under certain conditions these quiver theories
1 One might also consider this AdS2 as a degenerate limit of the warped AdS3/NHEK near horizon geometry of the rotating black hole, in the limit of vanishing angular momentum. The SL(2, R) might then be a global subgroup of the full symmetries associated to the duals of such geometries [151–156]. 2The original N × N Hermitean matrix models [160] in the double scaling limit has eigenvalue dynamics described by free fermions which naturally have an SL(2, R) symmetry. Of course, such models contain only the eigenvalue degrees of freedom due to the U(N) gauge invariance that allows for a diagonalization of the matrix, and hence do not have O(N 2) degrees of freedom. These models are dual to strings propagating in two dimensions (for some reviews see [161–163]). CHAPTER 4. CONFORMAL QUIVERS 73
have an exponential number of (supersymmetric) ground states whose logarithm goes as the charge of the branes squared, which is the same scaling as the entropy of a supersymmetric black hole in N = 2 supergravity. It has been argued [166,167] that the near horizon AdS2 of these supersymmetric black holes is related to the expo- nential explosion in the number of ground states in the quiver quantum mechanics. The states in question are referred to as pure-Higgs states since they reside in the Higgs branch of the quantum mechanics, where all the branes sit on top of each other. Interestingly, going to the Coulomb branch after integrating out the massive strings stretched between the wrapped branes, leads a non-trivial potential and veloc- ity dependent forces governing the wrapped brane position degrees of freedom in the non-compact space. Moreover, whenever the Higgs branch has an exponentially large number of ground states, the Coulomb branch exhibits a family of supersymmetric scaling solutions [165, 166, 171] continuously connected to its origin. The equations determining the positions of the wrapped branes in such supersymmetric zero energy scaling solutions are reproduced in four-dimensional N = 2 supergravity [165], be- lieved to be the appropriate description of the system in the limit of a large number of wrapped branes. In this chapter we will touch upon some of these issues. We do so by discussing two aspects of the Coulomb branch of such quiver theories, particularly those describ- ing three wrapped branes containing scaling solutions.
First we derive the Coulomb branch Lagrangian of a three node quiver model (see figure 4.1), and establish the existence of a low energy scaling limit where the the- ory exhibits the full SL(2, R) symmetry of conformal quantum mechanics [172–175]. These scaling theories have velocity dependent forces, a non-trivial potential as well as a metric on configuration space. It is also worth noting that the full quiver quantum mechanics theory is itself not a conformal quantum mechanics (and most certainly not a two-dimensional conformal field theory). The emergence of a full SL(2, R) sym- metry rather than only a dilatation symmetry in the scaling limit is not guaranteed, and is reminiscent of the emergence of a full SL(2, R) in the near horizon geometry of extremal black holes. CHAPTER 4. CONFORMAL QUIVERS 74
Second, we study the behavior of the Coulomb branch upon integrating out the strings in a thermal state, rather than in their ground state. At sufficiently high tem- peratures, the Coulomb branch melts into the Higgs branch. This is reminiscent of the gravitational analogue where increasing the temperature of a black hole increases its gravitational pull, or a particle falling back into the finite temperature de Sitter horizon.
4.2 General Framework: Quiver quantum mechan- ics
In this section we discuss the quiver quantum mechanics theory and its Coulomb branch. These theories constitute the low energy, non-relativistic and weakly coupled sector of a collection of branes along the supersymmetric cycles of a Calabi-Yau three fold. The wrapped branes look pointlike in the four-dimensional non-compact Minkowski universe.
4.2.1 Full quiver theory
The N = 4 supersymmetric quiver quantum mechanics comprises the following fields: α α α α chiral multiplets Φij = {φij, ψij,Fij } and vector multiplets Xi = {Ai, xi, λi,Di}. The α α ψij are the fermionic superpartners of the φij, the λi are the fermionic superpartners α of the scalars xi, Ai is a U(1) connection, and Fij and Di are auxiliary scalar fields. α ¯ The Φij transform in the (1i, 1j) of the U(1)i × U(1)j. The index α = 1, 2,..., |κij| denotes the specific arrow connecting node i to node j (see figure 4.1). The chiral multiplets encode the low energy dynamics of strings stretched between the wrapped
D-branes of mass mi sitting at three-vector positions xi in the non-compact four- dimensions. The index i = 1, 2,...,N denotes the particular wrapped D-brane. The (i) (i) electric-magnetic charge vector, Γi = (QI ,PI ), of the wrapped branes depends on the particular cycles that they wrap, and the Zwanziger-Schwinger product of their (i) (j) (i) (j) charges are given by the κij = (PI QI − QI PI ). The κij count the number of CHAPTER 4. CONFORMAL QUIVERS 75
Fig. 4.1: A 3-node quiver diagram which captures the field content of the Lagrangian L = LV + LC + LW , each piece of which is given in (4.2.1), (4.2.2), and (4.2.5). This quiver admits a closed loop if κ1, κ2 > 0 and κ3 < 0. intersection points in the internal manifold between wrapped branes i and j. In what follows we measure everything in units of the string length ls which we have set to one.
The Lagrangian L = LV + LC + LW for the three-node quiver quantum mechanics [165] describing the low energy non-relativistic dynamics of three wrapped branes which are pointlike in the (3 + 1) non-compact dimensions is given by:
3 X µi L = q˙ i · q˙ i + DiDi + 2iλ¯iλ˙ i − θiDi , (4.2.1) V 2 i=1 and,
3 X i 2 i i i i 2 i 2 ¯i i LC = |Dtφα| − q · q + siD |φα| + |Fα| + iψαDtψα i=1 √ ¯i i i ¯i i i − siψα(σ · q )ψα + i 2 siφαλ ψα − h.c. . (4.2.2)
In (4.2.1-4.2.2) and what follows we will mostly work with the relative degrees of freedom (i.e. xij ≡ xi − xj, Dij ≡ Di − Dj, etc.) since the center of mass degrees of freedom decouple and do not play a role in our discussion. The notation we use CHAPTER 4. CONFORMAL QUIVERS 76
1 2 3 is somewhat non-standard (for example (q , q , q ) ≡ (x12, x23, x13)) and is given in appendix C.1 along with all of our conventions. We note that the relative Lagrangian is only a function of two of the three vector multiplets, since q3 = q1 + q2, D3 = 1 2 3 1 2 D + D and λ = λ + λ . The si encode the orientation of the quiver. For the majority of our discussion we choose s1 = s2 = −s3 = 1, corresponding to a closed loop like the one in figure 4.1. The reduced masses µi (which we denote in superscript notation in (4.2.1-4.2.2)) are related to the masses mi ∼ 1/gs, where gs is the string coupling constant, of the wrapped branes sitting at the xi by:
m m m m m m µ1 = 1 2 , µ2 = 2 3 , µ3 = 1 3 . (4.2.3) m1 + m2 + m3 m1 + m2 + m3 m1 + m2 + m3
The superpotential, which is allowed by gauge invariance only when the quiver has a closed loop, is given by:
X 1 2 3 W (φ) = ωαβγφαφβφγ + higher order terms , (4.2.4) α,β,γ
(where we take coefficients ωαβγ to be arbitrary) and contributes the following piece to the Lagrangian:
3 3 2 ! X ∂W (φ) i X ∂ W (φ) i j LW = F + h.c. + ψ ψ + h.c. . (4.2.5) ∂φi α i j α β i=1 α i,j=1 ∂φα∂φβ
We only consider cubic superpotentials and ignore the higher order terms. This is i i consistent so long as the φα are small, which in turn can be assured by taking the |θ | sufficiently small [166]. The theory contains a manifest SO(3) global R-symmetry. In the absence of a superpotential, the Lagrangian is diagonal in the arrow (Greek) indices and thus the theory also exhibits a U(|κ1|) × U(|κ2|) × U(|κ3|) global symmetry under which the i i φα transform as U(|κ |) vectors. The superpotential explicitly breaks this symmetry down to the U(1)1 × U(1)2 × U(1)3 gauge symmetry. The theory can be obtained by dimensionally reducing an N = 1 gauge theory in four-dimensions to the (0 + 1)-dimensional worldline theory. It can also be viewed as CHAPTER 4. CONFORMAL QUIVERS 77
the dimensional reduction of the N = 2 two-dimensional σ-models studied extensively, for example, in [176]. In order for the theory to have supersymmetric vacua we also demand that the Fayet-Iliopoulos constants sum to zero: θ1 + θ2 + θ3 = 0. In units where ~ = 1 is dimensionless, and where we choose dimensions for which [t] = 1, the dimensions of energy are automatically set to [E] = −1. We also find the following dimensional assignments: [φ] = 1/2, [D] = −2, [x] = −1, [µ] = 3, [ωαβγ] = i −3/2, [ψ] = 0, [λ] = −3/2, [F ] = −1/2 and [θ] = 1. The mass squared of the φα fields 2 2 upon integrating out the auxiliary D fields is given by Mij = (|xij| + θi/mi − θj/mj). 2 2 i For physics whose energies obey E /M 1 we can integrate out the massive φα fields and study the effective action on the Coulomb branch. Finally, notice also that the coupling ωαβγ has positive units of energy and is thus strong in the infrared 3/2 i limit since the natural dimensionless quantity is ωαβγ/E . The contribution from φα loops grows as κi and thus the effective coupling constant at low energies is given by geff ∼ gsκ. In the large geff limit, the wrapped branes backreact and the appropriate description of the system is given by four-dimensional N = 2 supergravity [165].
4.2.2 Some properties of the ground states
An immediate question about the above model regards the structure of the ground i i ˆ i i states Ψg [Φα, Q ] of the theory, satisfying: H Ψg [Φα, Q ] = 0. Though an explicit i i expression for the full Ψg [Φα, Q ] remains unknown, the degeneracy of ground states has been extensively studied [165–170]. In particular, the degeneracy of ground states localized near Qi = 0, i.e. the ground states of the Higgs branch of the theory, were shown to grow exponentially in κi when the theory contains a superpotential, the quiver admits closed loops (e.g. κ1, κ2 > 0 and κ3 < 0) and the κi obey the triangle inequality (i.e. |κ2| + |κ3| ≥ |κ1| and cyclic permutations thereof). This growth is related to the exponential explosion in the Euler characteristic of the complete intersection manifold M [166] given by imposing the constraints from the F -term 3 i i i i i (δFα L|Fα=0 = 0) onto the D-term constraints (δD L|D =0 = 0). Since κ goes as 3As an example, we can take θ1, θ2 < 0. Then the complete intersection manifold is given by 3 3 1 2 κ1−1 κ2−1 setting φα = 0, imposing the κ F -term constraints: ωαβγ φαφβ = 0, inside a CP ×CP space 1 2 1 2 2 2 1 coming from the D-term constraints: |φα| = −θ and |φα| = −(θ − θ ). The space is a product CHAPTER 4. CONFORMAL QUIVERS 78
the charge squared of the associated U(1) gauge symmetry, the number of ground states scales in the same way as the Bekenstein-Hawking entropy of the associated black hole solutions in the large geff limit. Though a complete match between the ground states of a single Abelian quiver model and the entropy of a BPS black hole in N = 2 supergravity is not known,4 the vast number of microstates makes these systems potentially useful candidate toy models to study features of extremal or near extremal black holes. i Other pieces of Ψg localized near Φα = 0, i.e. the (quantum) Coulomb branch of the theory, have also been studied [165]. Unlike the Higgs branch quiver with a closed loop, superpotential and an exponential growth in its number of ground states, it was found that the number of Coulomb branch ground states grows only polynomially in the κi. Interpreting the qi as the relative positions of wrapped branes, these ground states can be viewed as describing various multiparticle configurations, as we will soon proceed to describe in further detail. To each ground state in the quantum Coulomb branch there exists a corresponding ground state in the Higgs branch, but the converse is not true. Another way to view this statement is that whenever a given i i Ψg has non-trivial structure in the Q directions and peaks sharply about Φα = 0, i it will also have a non-trivial structure in the Φα directions and peak sharply about Qi = 0 but not vice versa.
4.3 Coulomb branch and a scaling theory
i i For large enough |q | we can integrate out the massive Φα’s (in their ground state) i from the full quiver theory (4.2.1). This can be done exactly given that the Φα appear quadratically in (4.2.1) whenever the superpotential vanishes. One finds the bosonic
k i of CP ’s since we have to identify the overall phase of the φα due to the U(1) gauge connection. 1 2 i When |κ | + |κ | − 2 ≥ |κ3|, which for large κ amounts to the κ satisfying the triangle inequality, κ1−1 κ2−1 the number of constraints become less or equal to the dimension of CP × CP , allowing for more complicated topologies for M. 4Indeed, there are several quiver diagrams with the same net charges and one might suspect that all such quivers are required to obtain the correct entropy of the supersymmetric black hole (see for example [177]). CHAPTER 4. CONFORMAL QUIVERS 79
quantum effective Coulomb branch Lagrangian (up to quadratic order in q˙ i and Di):
2 3 i 1 X X si |κ | L = G q˙ i · q˙ j + Di Di − s |κi|Ad(qi) · q˙ i − + θi Di . (4.3.1) c.b. 2 ij i 2|qi| i=1 i=1
The terms linear in q˙ i and Di follow from a non-renormalization theorem [165], whereas the quadratic piece in q˙ i is derived in appendix C.4.1. Recall that the system is only a function of q1 and q2 since q3 = q1 + q2. The three-vector Ad is the vector potential for a magnetic monopole:
−y x Ad(x) = xˆ + yˆ , (4.3.2) 2r(z ± r) 2r(z ± r) and Gij is the two-by-two metric on configuration space:
1 3 1 |κ1| 1 |κ3| 3 1 |κ3| ! µ + µ + 4 |q1|3 + 4 |q1+q2|3 µ + 4 |q1+q2|3 [Gij] = 3 2 3 . (4.3.3) 3 1 |κ | 2 3 1 |κ | 1 |κ | µ + 4 |q1+q2|3 µ + µ + 4 |q2|3 + 4 |q1+q2|3
Upon integrating out the auxiliary Di-fields, we obtain a multi-particle quantum mechanics with (bosonic) Lagrangian:
2 3 1 X X L = G q˙ i · q˙ j − s |κi|Ad(qi) · q˙ i − V (qi) . (4.3.4) c.b. 2 ij i i=1 i=1
That the quantum effective Coulomb branch theory has a non-trivial potential V (qi) should be contrasted with other supersymmetric cases such as interacting D0-branes or the D0-D4 system [178] where the potential vanishes and the non-trivial structure of the Coulomb branch comes from the moduli space metric. The potential V (qi) is also somewhat involved and is given in appendix C.4.2. CHAPTER 4. CONFORMAL QUIVERS 80
4.3.1 Supersymmetric configurations
The supersymmetric configurations of the Coulomb branch consist of time indepen- dent solutions which solve the equations V (qi) = 0. For (4.3.1), this amounts to:
s |κi| s |κ3| i + 3 + 2θi = 0 , i = 1, 2 . (4.3.5) |qi| |q3|
In appendices C.2 and C.3 we review that these supersymmetric configurations are robust against corrections of the Coulomb branch theory from the superpotential and from integrating out higher orders in the auxiliary D fields.
Bound states
There are bound state solutions [165, 179] of (4.3.5) which are triatomic (or more generally N-atomic if dealing with N wrapped branes) molecular like configurations. Of the original nine degrees of freedom, three can be removed by fixing the center of mass. Then the bound state condition (4.3.5) fixes another two-degrees of freedom. Thus, bound state solutions have a four-dimensional classical moduli space. Due to the velocity dependent terms in the Lagrangian, the flat directions in the moduli space are dynamically inaccessible at low energies – the particles resemble electrons in a magnetic field. Several dynamical features of the three particles were studied in [4,180].
Scaling solutions
There are also scaling solutions [166] of (4.3.5) which are continuously connected to the origin |qi| = 0. They occur whenever the κi form a closed loop in the quiver diagram (e.g. κ1, κ2 > 0 and κ3 < 0) and obey the same triangle inequality (|κ2| + |κ3| ≥ |κ1| and cyclic permutations thereof) that the |qi| are subjected to. These solutions can be expressed as a series:
∞ i i X n |q | = |κ | anλ , λ > 0 . (4.3.6) n=1 CHAPTER 4. CONFORMAL QUIVERS 81
The coefficient a1 = 1, while the remaining an can be obtained by systematically solving (4.3.5) in a small λ expansion, and will hence depend on θi. The moduli space of the scaling solutions is given by the three rotations as well as the scaling direction parameterized by λ. Though the angular directions in the moduli space are dynamically trapped due to velocity dependent forces, the scaling direction is not and constitutes a flat direction even dynamically. n+1 n Requiring that the series expansion converges, i.e. (an+1λ )/(anλ ) 1, leads to the condition: 1 λ . (4.3.7) θ Because of this condition on the λ’s, one should be cautious when dealing with such scaling solutions. They occur in the near coincident limit of the branes where the bifundamentals that we have integrated out become light. In order for the mass of 2 2 the bifundamentals, Mij = (|xij| + θi/mi − θj/mj), to remain large we require:
θ 1/2 |qi| . (4.3.8) µ
i i i i α Taking µ = ν µb and |q | = |qb |/ν the inequalities (4.3.7) and (4.3.8) can be satisfied i i i i in the limit ν → ∞, with µb , qb , κ and θ fixed and furthermore α ∈ (0, 1/2). Notice that (4.3.7) implies that the distances between particles in a scaling regime ∼ λκ is much less than the typical inter-particle distance of a bound state ∼ κ/θ.
4.3.2 Scaling Theory
To isolate the physics of the Coulomb branch in the scaling regime we take an infrared limit of the Lagrangian (4.3.1), pushing the qi near the origin and dilating the clock t. In particular, we would like the ∼ κ/|qi|3 part of the metric in configuration space to dominate over the ∼ µ piece leading to:
κ1/3 |qi| . (4.3.9) µ CHAPTER 4. CONFORMAL QUIVERS 82
Additionally we must satisfy the inequalities (4.3.7) and (4.3.8). Again taking µi = i i i α α i i i i ν µb , q = qb /ν and in addition t = ν bt with fixed µb , qb , κ and θ , we can also satisfy (4.3.9) in the limit ν → ∞, so long as we also ensure α ∈ (1/3, 1/2).5 The rescaling of t is required to maintain a finite action in the scaling limit. In type √ II string compactifications µ ∼ 1/lP ∼ v/gsls, where lP is the four-dimensional Planck length and v is the volume of the Calabi-Yau in string units [165], therefore the ν → ∞ limit corresponds to a parametrically small string coupling. Furthermore the scaling throat deepens as we increase the mass of the wrapped branes. The rescaling above amounts simply to setting the θi and µi to zero in (4.3.1) and i i i replacing q and t with qb and bt. We call the remaining Lagrangian with vanishing θ and µi the scaling theory. Notice from equation (C.4.18) that the potential V (qi) in i i this limit becomes a homogeneous function of order one, i.e. V (ν qb ) = νV (qb ) and the linear in velocity term is retained.
Consistency of the small Di expansion
One can infer from the supersymmetry variations [165] that supersymmetric configu- rations have vanishing Di. Furthermore, as we have already noted, we have performed an expansion in small Di fields prior to integrating them out (see appendix C.3 for more details) in order to obtain the Coulomb branch. Thus, in order for a scaling theory to exist and be consistent with a small Di expansion, it must be the case that zero energy scaling configurations exist. Had we considered a two-particle theory, where no such scaling solutions can exist, taking a small q limit would be inconsistent with the small D expansion. That is because the dimensionless small quantity in our perturbation series is actually D/|q|2 and the supersymmetric configuration D = 0 occurs at |q| = −κ/2θ. Expanding the non-linear D equation (C.3.1) (see appendix C.3) in powers of ≡ D/|q|2, while imposing |q| (θ/µ)1/2, one finds the following consistency condition:
θ 1 κ 1 1 = + 1 − + O(2) . (4.3.10) µ |q|2 2µ |q|3 2
5Another limit one might imagine is given by: |qi| (κ/µ)1/3, |qi| κ/θ. In this case the metric on configuration space remains flat while the potential scales like ∼ 1/|q|2. CHAPTER 4. CONFORMAL QUIVERS 83
Indeed, the first term on the right hand side is negligible by construction, since we took |q| (θ/µ)1/2 to keep the strings massive. Smallness of the second term in the equation would require |q|3 κ/µ, in contradiction with the condition (4.3.9) required to isolate the scaling theory.
We now consider the three-node case for si that admit a closed loop in the quiver. 2 2 2 For small 2 ≡ D /|q | (taking all masses the same and θ1 = θ2 = θ and again imposing |q2| (θ/µ)1/2) the equation of motion of the auxiliary D2-field (in the form of (C.3.2)) is given by:
3θ |κ2| 1 = + δ − + O(2) (4.3.11) 2 2µ|q2|2 2µ|q2|3 2 2 2
|q2| |κ1| |κ3| where δ ≡ 1 − 2 |κ2| |q1| + |q3| measures how close the configuration is to the scaling solution. For sufficiently small δ ∼ 1 we can consistently satisfy (4.3.11) in addition to imposing the scaling inequalities (4.3.9). In this sense the scaling theory is in fact a theory of the deep infrared configura- tions residing parametrically near the zero energy scaling solutions. This is consistent with the dilation of time required to obtain the scaling theory.
4.4 Conformal Quivers: emergence of SL(2, R)
In this section we uncover that the bosonic scaling theory action, i.e. (4.3.4) with the θi and µi set to zero, has an SL(2, R) symmetry. This is the symmetry group of conformal quantum mechanics [175]. The group SL(2, R) is generated by a Hamilto- nian H, a dilatation operator D and a special conformal transformation K, with the following Lie algebra:
[H,D] = −2iH , [H,K] = −iD , [K,D] = 2iK . (4.4.1)
As we have already mentioned, the full SL(2, R) symmetry is not guaranteed by the existence of time translations and dilatations alone [181]. This is suggested by the fact that the H and D operators form a closed subalgebra of the full SL(2, R). The CHAPTER 4. CONFORMAL QUIVERS 84
presence of a full SL(2, R) is actually quite remarkable, particularly given the specific form of the scaling theory Lagrangian which has velocity dependent forces and a non-trivial potential. As discussed in the previous section, it is not true that, for any finite κi, µi and θi, the Coulomb branch can be described precisely by the scaling theory action. There will always be small corrections that break its manifest scaling symmetry: qi → γ qi, and t → t/γ. This is comforting, given that the full quiver theory has a finite number of ground states—yet a conformal quantum mechanics has a diverging number of arbitrarily low-energy states, with a density of states that behaves as dE/E. The corrections serve as a cutoff for the infrared divergence in the number of states, such that the Coulomb branch can fit consistently inside the full quiver theory (related discussions can be found in [167,182]).
4.4.1 Conditions for an SL(2, R) invariant action
The conditions under which an action will be SL(2, R) invariant (up to possible surface terms) have been studied extensively in [147,181,183]. Showing that a general theory with bosonic Lagrangian describing N degrees of freedom :
1 L = q˙i G q˙j − A q˙i − V (q) , i = 1, 2,...,N (4.4.2) 2 ij i has an SL(2, R) symmetry is equivalent to finding a solution to the following equations [183]:
2 ∇(iZj) = Gij , (4.4.3) i −Z ∂iV = V, (4.4.4)
2 Zi = ∂if , (4.4.5) j Z Fji = 0 ,Fij ≡ ∂[iAj] . (4.4.6)
Equations (4.4.3) and (4.4.4) ensure the existence of a dilatation symmetry. In par- ticular, equation (4.4.3) implies that the metric on configuration space allows for a CHAPTER 4. CONFORMAL QUIVERS 85
conformal Killing vector field (also referred to as a homothetic vector field). Equa- tions (4.4.5) and (4.4.6) ensure the that the action remains invariant under special conformal transformations, where f is an arbitrary function of the qi. Indices are raised and lowered with the metric Gij.
Interestingly, equation (4.4.5) imposes that the conformal Killing form Zi of the metric be exact, which is generically not the case. Hence the existence of a dilatation symmetry does not necessarily imply the symmetry of the full conformal group. Once a solution to (4.4.3-4.4.6) is found, the three conserved quantities are then given by [183]:
1 Q = tn+1 q˙i G q˙j − (n + 1)tnZi G q˙j + tn+1V (q) + F , n = −1, 0, 1 (4.4.7) n 2 ij ij n where F−1 = F0 = 0 and F1 = f. The charge Q−1 is the Hamiltonian, whereas Q0 and Q1 are related to dilatations and special conformal transformations, respectively. These three charges generate the SL(2, R) algebra (4.4.1) (up to factors of i) under the Poisson bracket. In what follows, we find Zi and f for the scaling theory described in section 4.3.2.
4.4.2 Two particles
As a warm up, we can study a simple model consisting of a two node quiver with an equal number of arrows going to and from each node, with a total number κ > 0 arrows altogether. This is the theory describing, for example, the low energy dynamics of a wrapped D4-D0 brane system (see [178]). The bosonic Lagrangian for the relative position q = (qx, qy, qz) on the Coulomb branch, in the scaling limit, is given by:
κ q˙ 2 L = . (4.4.8) 2 |q|3
The above Lagrangian is also the non-relativistic limit of one describing a BPS particle 2 in an AdS2 × S background. The wordline theory has three degrees of freedom and −3 a diagonal metric on configuration space: gij = κ |q| δij. In addition to the Hamiltonian H, the above theory has a dilatation operator D and special conformal CHAPTER 4. CONFORMAL QUIVERS 86
generator K given by: