SPIN-FOAM DYNAMICS OF by Atousa Chaharsough Shirazi

A Dissertation Submitted to the Faculty of The Charles E. Schmidt College of Science In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

Florida Atlantic University Boca Raton, FL December 2015 Copyright 2015 by Atousa Chaharsough Shirazi

ii

ACKNOWLEDGEMENTS

I am sincerely grateful and indebted to my advisor Dr. Jonathan Engle for his dedication, help and guidance through all my PhD years. I appreciate everything you taught me. Thank you for being patient. Besides my advisor, I would like to acknowledge the rest of my dissertation committee members for their valuable insight in preparing this manuscript . To all my professors at Florida Atlantic University, thank you for sharing your expertise with me. A special thank you to Dr. Warner Miller whose invaluable and crucial support during these years helped me to achieve my full potential. My heartfelt gratitute goes to my family. Words can not express my gratitute to my late father who inspired and encouraged me to become a scientist. Thank you for all the support, love and confidence you gave me. To my beloved mother whose ongoing love and support through all these years made this journey possible for me. Thank you for your patience during the years I was away from home pursuing my passion. Finally, I would like to thank my friends for being beside me through the good times and the bad. Thank you for all the fun times and memories that we made together during these years. You are definitely one of the significant parts of this journey.

iv ABSTRACT

Author: AtousaChaharsoughShirazi Title: Spin-foamDynamicsofGeneralRelativity Institution: Florida AtlanticUniversity Dissertation Advisor: Dr. Jonathan S. Engle Degree: Doctor of Philosophy Year: 2015

In this dissertation the dynamics of general relativity is studied via the spin- foam approach to quantum . Spin-foams are a proposal to compute a transition amplitude from a triangulated space-time for the evolution of quantum 3d geometry via path integral. Any path integral formulation of a quantum theory has two important parts, the measure factor and a phase part. The correct measure factor is obtained by careful canonical analysis at the continuum level. The basic variables in the Plebanski-Holst formulation of gravity from which spin-foam is derived are a Lorentz connection and a Lorentz-algebra valued two-form, called the Plebanski two-form. However, in the final spin-foam sum, one usually sums over only spins and intertwiners, which label eigenstates of the Plebanski two-form alone. The spin-foam sum is therefore a discretized version of a Plebanski- Holst path integral in which only the Plebanski two-form appears, and in which the connection degrees of freedom have been integrated out. Calculating the measure factor for Plebanksi Holst formulation without the connection degrees of freedom is one of the aims of this dissertation. This analysis is at the continuum level and in order to be implemented in spin-foams one needs to properly discretize and quantize

v this measure factor. The correct phase is determined by semi-classical behavior. In asymptotic analysis of the Engle-Pereira-Rovelli-Livine spin-foam model, due to the inclusion of more than the usual gravitational sector, more than the usual Regge term appears in the asymptotics of the vertex amplitude. As a consequence, solutions to the classical equations of motion of GR fail to dominate in the semi-classical limit. One solution to this problem has been proposed in which one quantum mechanically imposes re- striction to a single gravitational sector, yielding what has been called the “proper” spin-foam model. However, this revised model of , like any proposal for a theory of quantum gravity, must pass certain tests. In the regime of small curva- ture, one expects a given model of quantum gravity to reproduce the predictions of the linearized theory. As a consistency check we calculate the two-point function predicted by the Lorentzian proper vertex and examine its semiclassical limit.

vi SPIN-FOAM DYNAMICS OF GENERAL RELATIVITY

1 Introduction ...... 1

2 Background ...... 7 2.1 Plebanski-Holstformulationofgravity ...... 7 2.2 Discrete Geometry ...... 9 2.2.1 Bivector geometry ...... 10 2.3 Quantization ...... 12 2.4 EPRL model ...... 14 2.4.1 Imposingtheconstraints ...... 14 2.4.2 Boundary states ...... 15 2.4.3 The amplitude ...... 16 2.4.4 Asymptotics...... 16 2.5 “Proper” vertex amplitude ...... 18

3 “Proper” Graviton propagator ...... 21 3.1 Introduction ...... 21 3.2 Gravitonpropagatorinspin-foamformalism ...... 22 3.2.1 The metric operator ...... 23 3.3 EPRL graviton propagator ...... 24 3.4 Graviton propagator for the proper vertex ...... 27 3.4.1 Asymptoticlimit ...... 29 3.5 Hessianofthepropervertex: Calculation ...... 32

3.5.1 Derivatives of S˜EPRL with respect to ⌘ and one other variable 33 ˜o 3.5.2 Derivatives of S⇧ ...... 35 vii 3.5.3 Hessian ...... 36 3.6 Graviton propagator and its 0limit...... 36 ! 3.7 Coecient in the asymptotics of the proper vertex ...... 37 3.8 Discussion ...... 40

4 Canonical measure factor ...... 43 4.1 Introduction ...... 43 4.2 Reduced phase-space method ...... 44 4.3 Plebanski-Holstformulation ...... 46 4.4 Purely geometric path integral ...... 47 4.4.1 Integrating out the connection ...... 48 4.4.2 The determinant of a ...... 49 4.4.3 Finalcontinuumpathintegral ...... 50 4.5 ADMpathintegral ...... 51 4.5.1 Integrating out the momentum ...... 51 4.5.2 The determinant of A ...... 52 4.5.3 Space-time covariant variables ...... 53 4.5.4 Comparison with the result of the last section ...... 54 4.6 Results and conclusion ...... 55

Appendices ...... 57 ACoherentstates...... 58 A.1 SU(2)coherentstates...... 58 A.2 Extrinsiccoherentstates ...... 61 BIntegrationofapathintegralwithquadraticaction...... 63 C Equivalence of gauge-fixed and non-gauge-fixed path integrals . . . . 65 C.3 The argument ...... 65

Bibliography ...... 71

viii CHAPTER 1 INTRODUCTION

One of the main open problems in theoretical is to unify general relativ- ity (GR) and (QM). General relativity is valid for large scales and high energies and quantum mechanics governs physics at planck scale l Planck ⇠ 35 10 m.Bothofthesetheorieshavestrongempiricalsupportattheirregimeof validity. At some extreme cases where the space-time region is small and at high energies like black holes and the big bang, these two theories overlap. Therefore, having a consistent theory of “quantum gravity”(QG)thatreproducesgeneralrela- tivity and quantum mechanics in appropriate limit is of high importance. However, these two theories are not compatible and the problem is rooted in the fact that in GR, gravitational field modifies the notion of space-time. In quantum mechanics, the causal structure is defined by a fix background space-time and any dynamical field is quantized on this background. On the other hand, GR is a background independent theory meaning that there is no background needed for the field to live in. Gravi- tational field defines geometry and metric is a smooth dynamical field. Therefore, the metric is quantized in a quantum theory of GR and the standard QM approach which requires a fixed smooth background fails in this case. Di↵erent approaches to quantum theory of general relativity tackle the problems di↵erently. Among the most popular approaches are , (LQG), spin-foams, causal dynamical triangulation and . Each of these approaches have their pros and cons but so far there is no experiment favoring any of these approaches. In this dissertation for the reasons given below we choose the spin-foam approach which is a path integral formulation of loop quantum gravity, to study the dynamics 1 of general relativity. Loop quantum gravity [1–5] is a canonical quantization of general relativity that preserves the background independence principle from GR. In GR background inde- pendence is realized as di↵eomorphism invariance of the action. After quantization space-time is expected to have quantum properties. Since interaction with gravita- tional field is through geometry, the quantities involved in the interaction are geomet- ric quantities such as area, volume and length. Therefore, in LQG area and volume become operators and the eigenstates of these operators form a basis of the of states of our quantum system. One of the key results of LQG is that the area and volume operators have discrete eigenvalues [2,6]. Therefore at small scales geometry is discretized and space is built out of quanta. This discreteness is the result of compactness of the group SU(2) which is the local gauge group. The consequence of background independence is that the quanta of gravitational field does not live on aspace-timebuttheybuildspace-time. Spin-foams are a proposal for defining the dynamics of LQG via path integral [7– 10]. The spin-foam program has a di↵erent starting point from LQG, but nevertheless uses the loop quantum states which are eigenstates of the 3d boundary geometry as the boundary states and computes a transition amplitude between those states, yielding an evolution of quantum 3d geometry. It started in 3d with the work of Ponzano and Regge [11] by realizing that GR in 3d is a topological field theory (with no local degrees of freedom) and a quantum theory of GR can be obtained from a triangulated manifold by a topological (triangulation independent) state-sum. Since 4d GR has local degrees of freedom, the extension of the state-sum method to 4d is not trivial. However, the Plebanski formulation [12–15] of 4d gravity is in the form of a BF theory [16], which is a topological field theory, plus a so-called simplicity constraint. Using this, Barrett and Crane [17] extended the work of Ponzano and Regge to 4d and derived a quantum version of the simplicity constraint. The quantum constraint

2 becomes a restriction on the representations (of the local gauge group) summed over. Despite its remarkable success, the Barrett-Crane model fails to give correct tenso- rial structure for the graviton propagator in the low-energy limit [18] and its boundary states do not fully match the states of LQG. The reason for these is that the con- straints are imposed too strongly and in [19,20] a new way of imposing the constraint is suggested in 4d Euclidean case that fixes the problems with the Barett-Crane model. Later the new model was extended to the Lorentzian theory [21]. The classical continuum action that the new spin-foam models are derived from is the Plebanski-Holst action [19,22–25]. This action is obtained from the Plebanski action by adding an extra topological term that doesn’t change the equations of

1 motion. The coupling constant for this extra term introduces a new parameter R+ called the Barbero- [26, 27]. One of the most used spin- 2 foam models is the Engle-Pereira-Rovelli-Livine (EPRL) model [25] that extends the prior models [19, 20] to the case with finite Barbero-Immirzi parameter [26, 27] for both Euclidean and Lorentzian theories. The remarkable results of this model are that the space of boundary states matches the one from LQG for all values of and the area spectrum turns out to be discrete and the same as LQG area spectrum in both Euclidean and Lorentzian cases. In the path integral formulation of quantum mechanics, the transition amplitude between two points is obtained by restricting the integration in the partition function:

iScl(x) = (x)µ(x)e ~ Z D Z to all the possible paths between these points. The integrand of the path integral has two important parts: a phase part given by the exponential of i times the classical action Scl,andameasurefactorµ(x). The phase part,determinestheclassicalbehaviorusingthestationaryphase method [28]. In the classical limit, ~ 0, the phase oscillates fast and destruc- ! tive phase interference happens except in the vicinity of the stationary points of the 3 action x0 where Scl(x0) = 0, and the transition amplitude becomes

k µ(x )eiScl(x0m) Z⇠ 0m m=0 X where the sum is over the critical points m =1,...,k.Therefore,theformofthe phase part in terms of the classical action ensures that the solutions to the classical equations of motion dominate the path integral in the classical limit. The measure factor, µ(x)isdeterminedbycanonicalanalysisandisimportantfor the path integral to be equivalent to the corresponding canonical quantum theory. In order to establish this equivalence the measure factor for a constrained theory must be the Liouville measure on the reduced phase-space where all the first-class constraints are gauge-fixed [29,30]. This measure can then be extended to the full phase-space. In spin-foams the transition amplitudes are defined between loop quantum states that are eigenstates of the 3d geometry. In LQG eigenvalues of the 3-geometry have adiscretestructureandarelabeledbyspinnetworks.Therefore,inspin-foamsthe integral over paths is replaced by a discrete sum over histories of spin networks

= A() Z X where A is the contributed amplitude from each of these histories. In this dissertation, in studying the dynamics of general relativity via spin-foam approach we have paid attention to both the measure factor and the phase part. As mentioned above the correct measure factor is obtained by careful canonical analysis at the continuum level. For the Plebanski-Holst formulation of gravity from which spin-foams are derived this measure factor is calculated in [31]. The basic variables in this formulation are a Lorentz connection and a Lorentz-algebra valued two-form, called the Plebanski two-form. However, in the final spin-foam sum, one usually sums over only spins and intertwiners, which label eigenstates of the Plebanski two-form alone. The spin-foam sum is therefore a discretized version of a Plebanski- Holst path integral in which only the Plebanski two-form appears, and in which the 4 connection degrees of freedom have been integrated out. Calculating the measure factor for Plebanksi Holst formulation without the connection degrees of freedom is one of the aims of this dissertation and it is done in chapter 4 and the results are reported in [32]. This analysis is at the continuum level and in order to be implemented in spin-foams one needs to properly discretize and quantize this measure factor. On the other hand, the correct phase is determined by requiring correct behavior in the classical limit. The asymptotics of the EPRL vertex amplitude in the Euclidean and Lorentzian case has been extensively studied in [33, 34] to check that the phase part has the correct semiclassical limit. One diculty with this and all spin-foam models before it is that, due to the inclusion of more than the usual gravitational sector [35], more than one term appears in the asymptotics of the vertex amplitude. As a consequence, solutions to the classical equations of motion of GR fail to dominate in the semi-classical limit [36,37]. One solution to this problem has been proposed in which one quantum mechanically imposes restriction to a single gravitational sector, yielding what has been called the ‘proper’ spin-foam model of quantum gravity [36,37]. However, this revised model of quantum gravity, like any proposal for a theory of quantum gravity, must pass certain tests. For example, in cases where the space- time curvature is small, one expects linearized quantum gravity to be correct. As aconsequence,inthisregime,oneexpectsagivenmodelofquantumgravityto reproduce the predictions of the linearized theory. One of the aims of this dissertation is to calculate the graviton two-point function predicted by the ‘proper’ vertex and examine its semiclassical limit to see if it matches the propagator calculated from the linearized theory. This dissertation is organized as follows: In chapter 2, the necessary preliminaries and background for the spin-foams formalism is given and the Lorentzian EPRL model and its asymptotics are reviewed . In chapter 3, as a consistency check of the ‘proper’

5 spinfoam model we calculate the gravitational two-point function predicted by the ‘proper’ spin-foam vertex to the lowest order in the vertex expansion to verify that it reproduces the predictions of the linearized theory in the appropriate limit. The analysis is restricted to the Lorentzian signature, as this is the physically relevant one. In chapter 4, in order to find the proper measure factor for spin-foams we calculate the measure factor for Plebanski-Holst formulation of gravity with the connection degrees of freedom integrated out. To confirm the results this analysis is done in two independent ways. The final results are summarized and discussed in chapter 5.

6 CHAPTER 2 BACKGROUND

2.1 PLEBANSKI-HOLST FORMULATION OF GRAVITY

Consider space-time as a 4d manifold .Thespace-timemanifoldindices,are M denoted here by lower case Greek letters ↵,, µ, ⌫, 0, 1, 2, 3 and ‘internal’ ··· 2{ } indices by I,J,K, 0, 1, 2, 3 which are raised and lowered with a fixed ‘internal’ ···2{ } metric ⌘ := diag(s, 1, 1, 1). s := +1 and s := 1 corresponds to Euclidean and IJ Lorentzian gravity respectively. The starting point for spin-foams is not General Relativity but BF theory [16] which is a topological field theory whose analysis is simpler than GR. Specifically, the Plebanski formulation of GR can be written in the form of a BF theory plus a constraint. The BF action is defined by:

S [B,!]= B F IJ[!](2.1) BF IJ ^ ZM where BIJ = 1 BIJdxµ dx⌫ is a two-form with values in the Lie algebra of so(⌘)1, 2 µ⌫ ^ IJ IJ IJ µ F is the curvature two-form of the so(⌘)-valued connection ! = !µ dx and is defined by: F IJ = d ! = d!IJ + !I !KJ ! K ^ where d and d! are exterior derivative and covariant exterior derivative respectively.

IJ Consider BIJ and ! as independent variables. The Euler-Lagrange equations of motion for this action, obtained by varying these variables, are (1) No curvature:

IJ F =0and(2)d!B=0.

1For Euclidean case s = 1, so(⌘)=so(4) and for the Lorentzian case where s = 1, so(⌘)= so(3, 1). 7 IJ In order to reproduce GR’s action the Bµ⌫ field is constrained to satisfy the sim- plicity constraint

s C := ✏ BIJBKL ✏ ✏↵✏ BIJBKL 0(2.2) µ⌫⇢ IJKL µ⌫ ⇢ 4! µ⌫⇢ IJKL ↵ ⇡

↵ where ✏IJKL denotes the internal Levi-Civita tensor with ✏0123 =1,and✏µ⌫⇢, ✏ denote the Levi-Civita tensor densities of weight 1and1,respectively.Thiscon- IJ straint has 20 independent components per point and restricts Bµ⌫ to belong to one of the following five sectors [4]. The first, is the degenerate sector, in which

µ⌫⇢ IJ KL IJ ✏ ✏IJKLBµ⌫ B⇢ =0.TheotherfoursectorseachcorrespondtoBµ⌫ taking one of the following four forms

1 eI eJ (I ) BIJ = ± ^ ± 8⇡G 8 eK eL = 1 ✏IJ eK eL (II ) < ±⇤ ^ ± 2 KL ^ ± for some non-degenerate tetrad: eI . Here is the exterior product or wedge product µ ^ and is the Hodge dual operator with respect to the internal metric. However, just ⇤ aspecialcombinationofthesectors(II+) and (II ), namely the Einstein-Hilbert sector [35,38], gives the usual Einstein-Hilbert action for General Relativity. This will be relevant later on in this thesis. It is convenient to use units where 8⇡G =1, which we will do from now on. There is another term ( B ) F IJ,thatcanbeaddedtotheactionthatpreserves ⇤ IJ ^ the symmetries of the theory and doesn’t change the equations of motions. The total action now reads 1 S (B + B ) F IJ (2.3) BF ⌘ IJ ⇤ IJ ^ Z IJ 1 IJ KL 1 + where ( B) := 2 ✏ KLB , is the coupling constant and R is called the ⇤ () 2 Barbero-Immirzi parameter [26, 27]. Define, BIJ := (B + 1 B)IJ.Theaction(2.3) ⇤ constrained to the simplicity constraint (2.2) is called Plebanski-Holst action and is the starting point for spin-foams.

8 Restriction to the (II )sectorsreducesthisactiontotheHolst action of General ± Relativity [39]

1 1 s S (B + B ) F IJ = ( e e + eK eL) F IJ S . (2.4) BF ⌘ IJ ⇤ IJ ^ ±2 ⇤ I ^ J 2 ^ ^ ⌘± Holst Z Z The equations of motion derived by varying this action are

() d BIJ =0 d BIJ =0 d eI =0 ! , ! , ! and ✏ eI F JK =0 IJKL ^ the first is the torsion-free condition on !,andthesecondisequivalenttotheEinstein equations. For finite non-trivial Immirzi parameter, the sector (I )alsogivesGR, ± s but the values of the e↵ective Newton constant and immirzi parameter are G and —di↵erentthansection(II ), where they are simply G and . ±

2.2 DISCRETE GEOMETRY

Spin-foams compute a transition amplitude for evolution of quantum 3-geometries from a triangulated space-time manifold .Specifically,weconsiderasimplicial M decomposition of the manifold which we assume to be oriented. For the rest M of this chapter we restrict to the Lorentzian manifold which is more physical and for simplicity the calculation is restricted to the case of just one 4-simplex .The generalization of the results to the case of multiple simplices can be done by using group field theory techniques [40–42].

ALorentziangeometric4-simplexisaconvexhulloffivepointsinR3,1 called vertices. If all the vertices do not lie on a hyperplane the 4-simplex is called non- degenerate. The 4-simplex has five tetrahedra ⌧a on the boundary @ where a = 0, 1,...,4 and the tetrahedra are labeled by the point that the tetrahedron does not include. The numbering of the vertices imposes an orientation. Each tetraheron has 9 four triangles so there are twenty triangles associated to the boundary of a 4-simplex. But the boundary is closed and each triangle is shared between two tetrahedra. Thus, in total there will be ten triangles, tab,ontheboundarywheretheindicesdenotethat the triangle is shared between tetrahedron ⌧a and tetrahedron ⌧b.

2.2.1 Bivector geometry

Suppose V and W are two vectors in R3,1.Asimplebivectorcanbeconsideredasan antisymmetric tensor that can be expressed as a wedge product of two one-forms

V W = V W W V. ^ ⌦ ⌦

GR restricts the BIJ fields to be of the form eI eJ .ForeachchoiceofI and J, the ⇤ ^ bivector BIJ is simple.

Each oriented triangle t in R3,1 defines a simple bivector

1 X = e e . t 2 t01 ^ t02 by the exterior product of its two edges e = x x , e = x x where t01 t0 t1 t02 t0 t2 3,1 xt0 ,xt1 ,xt2 R are position coordinates for its vertices. Equivalently this bivector 2 can be defined by the 4d outward normals Na, Nb to the tetrahedra sharing it

N () N () X ():=X = A(t ) a ^ b ab tab ab ⇤ N () N () | a ^ b | where A(tab)istheareaofthetriangletab.In[17,33,34]thesetofbivectorsXab as defined above are chosen to describe the geometry of a given 4-simplex. Here we choose B = X to be our set of bivectors to be consistent with [19,25]. According ab ⇤ ab to bivector geometry theorem, a set of ten bivectors satisfy certain properties com- pletely define the geometry of a 4-simplex. Bivector geometry theorem [17, 34]: Each geometric 4-simplex determines a set of bivectors satisfying the bivector geometry constraints stated below and each set of bivectors satisfying these conditions determines a geometric 4-simplex unique up to 10 parallel translation and inversion through the origin. Namely, there is a parameter µ = 1andageometric4-simplex unique up to translation and inversion, such ± that Bab()=µBab. Bivector geometry constraints: 1) Orientation condition: The bivector changes sign if the orientation of the triangle is changed, B = B . ab ab 2) Closure: The sum of the four bivectors corresponding to the faces of each tetrahe- dron is zero: a B =0. 8 ab b:b=a X6 3) Diagonal simplicity: Each bivector must define a geometric plane, i.e., is simple,

B B =0. (2.5) ab ^ ab

4) Cross simplicity: Each set of bivectors belonging to the same tetrahedron must span a 3D hyperplane

(3,1) IJ a, Na R , such that NaI ( Bab )=0, b = a. (2.6) 8 9 2 ⇤ 8 6

5) Tetrahedron non-degeneracy: For three triangles meeting at a vertex e of the ath tetrahedron (bcde): Tr(B [B ,B ]) =0. ab ac ad 6 6) Nondegeneracy: The assignment of bivectors is non-degenerate. This means that for six triangles sharing a common vertex, the six bivectors are linearly independent. I refer to [17] for the proof of this theorem. The Lorentzian geometrical 4-simplex is the classical geometry that we start with. Therefore, a correct spinfoam theory should be peaked on such classical geometries. In the asymptotic analysis of the spinfoam vertex amplitude that will be discussed in section 2.4.4 it becomes apparent how the Lorentzian geometrical 4-simplex emerges in the semiclassical limit.

11 2.3 QUANTIZATION

In the formulation of quantum theory which we use, a Hilbert space of states is asso- ciated to the boundary of a 2-complex ⇤ that is dual to a simplicial decomposition

2 of the 4d manifold .Theverticesv,edges e and faces f of ⇤ are dual to M 4-simplices ,tetrahedra⌧ and triangles t respectively. The points where edges inter- sect the boundary are called nodes n and the lines where faces intersect the boundary are called links l, which together form the graph which is the boundary of the

2-complex ⇤. The calculations in this dissertation is restricted to one 4-simplex, which has five tetrahedra and ten triangles on the boundary @,sothatthegraph has five nodes and ten links. Before quantization one needs to properly discretize the continuum variables of the theory. As was seen in section 2.1 the basic variables in Plebanski-Holst formulation

IJ are the two-form B and the curvature two-form FIJ. In quantum theory the groups are replaced by their universal coverings. Therefore, we replace SO(3, 1) by SL(2, C).

IJ Now, the B fields are so(3, 1) ⇠= sl(2, C)valuedtwo-formsandcanbeintegratedon triangles to give an element of so(3, 1) ⇠= sl(2, C),

IJ IJ Btab := B . Ztab The only assumption in the integration above is that the B field is constant on the surface of integration, i.e., each triangle. In the discrete theory the continuum BIJ

IJ fields are replaced by their discrete version Btab which are elements of sl(2, C)and therefore can be identified with the generators of the group SL(2, C). The faces of ⇤ are dual to triangles of the 4-simplex . Therefore, by assigning an algebra element BIJ := BIJ to a face f dual to t the boundary variables of a 4-simplex will be a fab tab ab ab set of ten algebra elements B . { f } The curvature F is now replaced by holonomies Ue SL(2, C)alongedgese.On 2 2The reason for the font change is so that the edges e are not confused with tetrads e.

12 the boundary the holonomies Ul are defined along the links. The relation between this discrete variable and the continuous sl(2, C)connection,!IJ,is

! Ul = Pe l , R where P stands for the path-ordered and the connection one-form is integrated along the link l.Sothediscreteboundaryvariablesareasetoften(Bf ,Ul). Before imposing the simplicity constraints in the quantum theory it is helpful to do a change of variables. From (2.3) the variable conjugate to Ul reads

1 J = B + B . f f ⇤ f ✓ ◆ Inverting this equation gives

2 1 B = J J . f 2 s f ⇤ f ✓ ◆✓ ◆ The diagonal simplicity constraint (2.5) and the cross simplicity constraint (2.6) in terms of new variables respectively become:

1 2 B .B = J .J 1+s s J .J =0, (2.7) ⇤ f f ⇤ f f 2 f f ✓ ◆ 1 N ( B )IJ = N ( J )IJ s J IJ =0 (2.8) I ⇤ f I ⇤ f f ✓ ◆ IJ where B.B = B BIJ. Also, the closure for Bf implies the closure for Jf . Note that imposing (2.6) selects sector (II ), i.e., B = e e and it eliminates sector (I ). ± ±⇤ ^ ± These constraints need to be imposed in the quantum theory. The quantization steps then consist in (1.) associating a Hilbert space of states to the boundary graph , (2.) imposing the constraints and (3.) computing a transition amplitude for these boundary states. In the next section we will review one of the most used spin-foam models namely Engle-Pereira-Rovelli-Livine (EPRL) model [25].

13 2.4 EPRL MODEL

2.4.1 Imposing the constraints

EPRL model [25] treats the simplicity constraints by first imposing them properly in afixedSO(3, 1) gauge, and then projecting on the gauge invariant spaces.

0 Lets fix NI = I .Thegeneralcasewillberecoveredbygaugeinvariance.Inthe Lorentzian case this choice restricts all tetrahedra to be spacelike. With this choice (2.8) becomes 1 1 1 Cj = ✏j J kl s J 0j = Lj s Kj 0(2.9) f 2 kl f f f f ⇡

j 0j j 1 j kl where ✏ kl := ✏ kl,andLf := 2 ✏ klJf are the generators of the SU(2) subgroup that j 0j leaves NI invariant, and Kf := Jf are the generators of the corresponding boosts.

Since we are considering the Lorentzian case, Jf are the generators of SL(2, C)so C = J .J =2(L2 + sK2)andC = J .J =4sL.K are the casimir and pseudo- 1 f f 2 ⇤ f f casimir operators of sl(2, C). In terms of these operators, the constraint (2.7) can be written as s 2s C 1+ C =0 (2.10) 2 2 1 ✓ ◆ which is a first class constraint and in the EPRL model it is imposed strongly. On the other hand, the equation (2.9) is a second class constraint. In the EPRL model the “master” constraint technique [4,43] is used to implement this constraint:

s 2 M := (Ci)2 = Li Ki 0 f ⇡ i i X X ✓ ◆ that leads to

2 C2 =4L . (2.11)

The Hilbert space of L2 functions on SL(2, C)canbeexpandedintoirreducible unitary representations of SL(2, C). The principal series of irreducible representations of SL(2, C)arelabelledbypositiverealnumbersp and non-negative half-integers k.

Each SL(2, C) irreducible representation p,k can be decomposed into a direct sum H 14 of SU(2) irreducibles which are labeled by a non-negative integer or half-integer Hj j. The casimir operators for the representation (p, k), are given by

C = p2 k2 +1 1

C2 = pk.

Equations (2.10) and (2.11) impose p = k and k = j [25] , where j labels the sub- spaces diagonalizing L2.Therefore,impositionofthesimplicityconstraintsbecomes arestrictionontherepresentationssummedoverinthespin-foamsum.Specifically, states p, k; j, m in satisfying these constraints have the form j,j; j, m .These | i Hp,k | i states are in one-to-one correspondence with the states in the irreducible representa- tions of SU(2). Lets define a map as I

: I Hj 7! Hj,j

j, m j,j; j, m . | i 7! | i The map encodes the constraints (2.10) and (2.11). I

2.4.2 Boundary states

The boundary states which we will use in our calculations are chosen to be coherent states of SL(2, C). These states can be obtained by the map from certain SU(2) I coherent states (see appendix A). To define these coherent states, we use the real-

1 ization of j and p,k in terms of homogeneous functions on CP [44]. The SU(2) H H coherent states corresponding to spinor ⇠ and spin j are then defined as:

j dj ¯ 2j C⇠ (z)= ⇠,z r ⇡ h i where the Hermitian inner product is defined by

w, z =¯w z +¯w z h i 0 0 1 1

15 and d =2j +1. Using the explicit form of the map in terms of this realization of j I j and p,k, we find the SL(2, C)boundarystatestobe[34], H H j (j,j) j dj 1+ij j ¯ 2j C⇠ (z) = C⇠ (z)= z,z ⇠,z . I r ⇡ h i h i

2.4.3 The amplitude

The vertex amplitude for these boundary states for a single 4-simplex then is defined by [34]:

Av =( 1) (g0) dga ab 5 P SL(2,C) a Z Y Ya

jab 1 jab ab = ↵( C ,g gb C ) P I ⇠ab a I ⇠ba is defined for each triangle (ab). Here ↵ is the invariant bilinear form satisfying ↵( ,)=( 1)2j↵(, ). Then we can rewrite the amplitude as:

SEPRL Av = c( 1) (g0) dga dµzab e (2.12) 5 1 10 SL(2,C) a ! (CP ) Z Y Z Ya

2 2 J⇠ab,Zab Zba,⇠ba Zba,Zba SEPRL = jab log h i h i + ijab log h i (2.13) Zab,Zab Zba,Zba Zab,Zab Xa

2.4.4 Asymptotics

According to the correspondence principle quantum systems must reproduce classical physics in the limit of large quantum numbers. Therefore, the asymptotics of the

3 Here we are using [45] convention for defining zab. By a change of variable inz ˜ab = Jzab we get the convention used in [38]. 16 vertex amplitude (2.12) in the limit of large spins jab must be peaked on the discrete geometry described in section 2.2. To assure that all the spins grow at the same rate the asymptotics is studied by first rescaling j by j then taking the limit . ab ab !1

The EPRL action (2.13) is linear in jab.Thereforethevertexamplide(2.12)isan integral of the general form:

n S(x) n Re (S(x)) iIm (S(x)) Av = d xf(x)e = d xf(x)e e Z Z By using the extended stationary phase method [28], in the limit the dominant !1 terms come from critical points x0 of the action. Critical points are the stationary points of the action Stot(x0)=0forwhichtherealpartoftheactionisfurthermore maximum. Since the real part of (2.13) satisfies Re(S ) 0, for critical points we tot  have Re (Stot(x0)) = 0. If there are more than one critical point present, the large spin limit of the integral is approximated by the sum of individual contribituions from each critical point

2⇡ n 1 S(x ) A = ( ) 2 f(x )e i . v i i det( H) X The asymptotics of the EPRL vertexp amplitude has been studied by [33] in the Euclidean case and in [34] for the Lorentzian case. Here we state the results for the Lorentzian case which we are discussing here: Given a set of non-degenerate boundary data, in the limit ,andforp = j , !1 ab ab kab = jab, 1) if the bounday state is the Regge state [33, 34] of the boundary geometry of a Lorentzian 4-simplex

12 1 L L Av ( 1) N+ exp i SRegge + N exp i SRegge ⇠ ✓ ◆ " a

17 4-simplex E

12 1 E E E E Av ( 1) N+ exp i SRegge + N exp i SRegge (2.15) ⇠ " ! !# ✓ ◆ Xa

1 12 A N v ⇠ V ✓ ◆ 4) for a set of boundary data that is neither a non-degenerate Lorentzian 4-geometry nor admits a vector geometry solution, the amplitude is suppressed for large

K A = o( ) non-negative integer K v 8

E E where N+,N ,N+ ,N and NV are independent of . In the first two cases two terms appear in the asymptotics. The first one is exp(iSRegge)andcorrespondstothecorrectReggeactionofgravityforadiscrete geometry, the other one corresponds to the negative of the Regge action, which cor- reponds to the opposite orientation. Together they result in the cosine of the Regge action. The appearance of the second term prevents the correct classical equation of motions from dominating in the classical limit when more than one 4-simplex is

present. In this case the result is a product over simplices cos(SRegge)whichis neither of the form exp(iSRegge)noroftheformcos(SRegge). Therefore,Q in the case of multiple simplices the action is non-Regge like which results in a wrong classical limit. In order to eliminate the second term, the “proper” vertex amplitude was introduced in [36–38] which will be discussed in the next section.

2.5 “PROPER” VERTEX AMPLITUDE

According to the Bivector geometry theorem stated in section 2.2.1 a set of bivectors

Bab satisfying bivector geometry constraints determines a geometric 4-simplex up 18 to parallel translation and inversion via B ()=µB where µ = 1. It was shown ab ab ± that the second term in the asymptotics of the vertex amplitude disappears if one restrict to a particular combination of sectors (II+) and (II )whichgivesµ =1 called the Einstein-Hilbert sector [36–38]. This sign is a multiplication of two factors µ = !⌫. The first is the sign of the orientation ✏ BIJ BKL determined by BIJ IJKL ^

↵ IJ KL !(Bµ⌫):=sgn(✏ ✏IJKLB↵B ) where ✏↵ is the fixed orientation on . The second, ⌫(B )is+1whenBIJ is in M µ⌫ µ⌫ the Plebanski sector (II+), 1ifitisinthePlebanskisector(II )and⌫(B )=0 µ⌫ otherwise. In order to have a well-defined operator in quantum theory to implement the projection to the Einstein-Hilbert sector, this operator must be written in terms of the boundary data and group variables. The projection µ =1isdonebyintroducing aprojector

i ⇧ba[gab]=⇧(0, ) ba[g]tr igbagba† L . 1 h ⇣ ⌘ i 1 i Here gba = gb ga, L is the rotation generator in the spin jab representation of SU(2) and

i j k l m n ba =sgn ✏ijknbcnbdnbe✏lmnnacnadnae where c, d, e = 0,...,4 a, b and { } { }\{ }

i 1 i n [g]= tr( g g† ). ba 2 ba ba

The EPRL vertex amplitude restricted to the Einstein-Hilbert sector is now called “proper” vertex amplitude and for a general boundary state is given by:

(+) 1 Av =( 1) (g0) dga ↵( ab,ga gb ⇧ba( gab ) ba). (2.16) 5 I I { } SL(2,C) a ! Z Y Ya

19 Using Thm. 6 in [38] we can rewrite this amplitude for coherent states as:

(+) jab 1 jab A =( 1) (g ) dg ↵( ⇧ ( g )C ),g g C ). v 0 a ab ab ⇠ab a b ⇠ba 5 I { } I SL(2,C) a ! Z Y Ya

⇧abC⇠ab = dµ˜⌘ˆab C⌘ˆab (C⌘ˆab , ⇧abC⇠ab )(2.17) Z yielding

(+) jab 1 jab jab jab A =( 1) (g ) dg dµ˜ ↵( C ,g g C )(C , ⇧ ( g )C ) v 0 a ⌘ˆab ⌘ˆab a b ⇠ba ⌘ˆab ab ⇠ab 5 I I { } SL(2,C) a ! Z Y Ya

djab tation of SL(2, C)anddµ˜⌘ˆab = ⌦⌘ˆab . Here⌘ ˆ are normalized spinors⌘ ˆ = ⌘/ ⌘ , ⇡ k k so that ⌦ =⌦/ ⌘ 4.Rewritingeachinnerproductintermsofanintegralovera ⌘ˆ ⌘ k k spinor zab as before, we obtain the integral representation

A(+) =( 1) (g ) dg dµ˜ dµ eSprop . (2.18) v 0 a ⌘ˆab zab a ! Z Y Z Ya

Sprop = SEPRL + S⇧ where the first term is a modified EPRL action

2 2 ab J⌘ˆab,Zab Zba,⇠ba Zba,Zba SEPRL = SEPRL =: jab log h i h i + ijab log h i Zab,Zab Zba,Zba Zab,Zab Xa

S = Sab =: log(C , ⇧ ( g )C ) . ⇧ ⇧ ⌘ˆab ab { ab} ⇠ab jab Xa