44 tayangan

Diunggah oleh ArsLexii

- Physics2012.pdf
- Energy Universe /Energetic Physics/
- Quantum Math Makes Human Irrationality More Sensible
- The Unified Field and the Illusion of Time
- DAriano_FQXi2013-It From Bit
- Bohm_meaning+information
- Copy of Nd Less Technical Introduction to the...
- 1607.07356
- Witness Gravity’s Quantum Side in the Lab (Nature Paper)
- FrontMatter_Giancoli7e_Manual.pdf
- On Deformed Heisenberg Algebras and Gravity’s Rainbow
- What is Time
- A New Approach to the Hard Problem - 07.14 - Final - KI - TSC 2014
- The Einstein-myth and the Crisis in Modern Physics
- Quantum theories new
- Why Physics Needs Quantum Foundations
- Quantum in Ed
- The DKP oscillator with a linear interaction in the cosmic string space-time
- Introduction to Cosmology, Physics, Philosophy
- Feynman Integral on Lie Groups

Anda di halaman 1dari 95

GRAVITY

A Thesis

Submitted to the Graduate Faculty of the

Louisiana State University and

Agricultural and Mechanical College

in partial fulllment of the

requirements for the degree of

Master of Science

in

The Department of Physics and Astronomy

by

Muxin Han

B.S., Beijing Normal University, Beijing, 2005

May 2007

Acknowledgements

First of all, I am grateful to Dr. Jorge Pullin for his advise and many corrections of this thesis,

and to Dr. Jonathan Dowling for all his kind help in these two years. I also would like to thank

Dr. Hwang Lee for his kind support and being a member of my committee.

I would like to thank all the people who have discussed issues with me concerning the

subject in the thesis. They are: Dr. Abhay Ashtekar, Dr. Lai-Him Chan, Dr. Weiming Huang,

Dr. Jerzy Lewandowski, Dr. Yongge Ma, Dr. Andrzej Okolow, Dr. Jorge Pullin, Dr. Carlo

Rovelli, Dr. Thomas Thiemann, Dr. Dmitry Uskov, Dr. Robert M. Wald, and Dr. Hongbao

Zhang.

This work is supported by the assistantship of LSU, the Horace Hearne Institute for Theo-

retical Physics at LSU, and funding from Advanced Research and Development Activity.

ii

Table of Contents

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 Motivation of Quantum Gravity . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Purpose of Loop Quantum Gravity . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Classical Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.1 Lagrangian Formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.2 Hamiltonian Formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

3 Foundations of Loop Quantum Gravity . . . . . . . . . . . . . . . . . . . . . . . 11

3.1 General Programme for Algebraic Quantization . . . . . . . . . . . . . . . . . 11

3.2 Quantum Conguration Space . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3.3 Cylindrical Functions on Quantum Conguration Space . . . . . . . . . . . . . 20

3.4 Loop Quantum Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.5 Spin-network Decomposition of Kinematical Hilbert Space . . . . . . . . . . . 25

3.6 Quantum Riemannian Geometry . . . . . . . . . . . . . . . . . . . . . . . . . 31

4 Implementation of Quantum Constraints . . . . . . . . . . . . . . . . . . . . . . 34

4.1 Solutions of Quantum Gaussian Constraint . . . . . . . . . . . . . . . . . . . . 34

4.2 Solutions of Quantum Dieomorphism Constraint . . . . . . . . . . . . . . . . 35

4.3 Hamiltonian Constraint Operator . . . . . . . . . . . . . . . . . . . . . . . . . 37

4.4 Master Constraint Programme . . . . . . . . . . . . . . . . . . . . . . . . . . 44

5 Quantum Matter Field on a Quantum Background . . . . . . . . . . . . . . . . . 51

5.1 Polymer-like Representation of a Scalar Field . . . . . . . . . . . . . . . . . . 51

5.2 Dieomorphism Invariant Hamiltonian of a Scalar Field . . . . . . . . . . . . 54

5.3 Hamiltonian Constraint Equation for the Coupled System . . . . . . . . . . . . 61

5.4 Master Constraint for the Coupled System . . . . . . . . . . . . . . . . . . . . 63

6 The Semiclassical Limit of Quantum Dynamics . . . . . . . . . . . . . . . . . . . 68

6.1 The Construction of Coherent States . . . . . . . . . . . . . . . . . . . . . . . 68

6.2 Algebraic Quantum Gravity Approach . . . . . . . . . . . . . . . . . . . . . . 74

7 Conclusion and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

Vita . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

iii

Abstract

In the last 20 years, loop quantum gravity, a background independent approach to unify general

relativity and quantum mechanics, has been widely investigated. The aim of loop quantum

gravity is to construct a mathematically rigorous, background independent, nonperturbative

quantum theory for the Lorentzian gravitational eld on a four-dimensional manifold. In this

approach, the principles of quantum mechanics are combined with those of general relativity

naturally. Such a combination provides us a picture of quantum Riemannian geometry, which

is discrete at a fundamental scale. In the investigation of quantum dynamics, the classical

expressions of constraints are quantized as operators. The quantum evolution is contained in

the solutions of the quantum constraint equations. On the other hand, the semi-classical analysis

has to be carried out in order to test the semiclassical limit of the quantum dynamics.

In this thesis, the structure of the dynamical theory in loop quantum gravity is presented ped-

agogically. The outline is as follows: rst we review the classical formalism of general relativity

as a dynamical theory of connections. Then the kinematical Ashtekar-Isham-Lewandowski rep-

resentation is introduced as a foundation of loop quantum gravity. We discuss the construction

of a Hamiltonian constraint operator and the master constraint programme, for both the cases of

pure gravity and matter eld coupling. Finally, some strategies are discussed concerning testing

the semiclassical limit of the quantum dynamics.

iv

1 Introduction

1.1 Motivation of Quantum Gravity

The current view of physics is that there exist four fundamental interactions: strong interaction,

weak interaction, electromagnetic interaction and gravitational interaction. The description for

the former three kinds of forces is quantized in the well-known standard model. The interactions

are transmitted via the exchange of particles. However, the last kind of interaction, gravitational

interaction, is described by Einsteins theory of general relativity, which is a classical theory

which describes the gravitational eld as a smooth metric tensor eld on a manifold, i.e., a 4-

dimensional spacetime geometry. There is no and hence no quantum structure of spacetime.

Thus there is a fundamental inconsistency in our current description of the whole physical

world. Physicists widely accept the assumption that our world is quantized at fundamental level.

So all interactions should be brought into the framework of quantum mechanics fundamentally.

As a result, the gravitational eld should also have quantum structure.

Throughout the last century, our understanding of nature has considerably improved from

macroscale to microscale, including the phenomena at molecule, atom, sub-atom, and elemen-

tary particle scale. The standard model of particle physics agrees with all present experimental

tests in laboratory (see e.g. [158]). However, because unimaginably large amount of energy

would be needed, no experimental tests exist for processes that happen near the Planck scale

p

(G/c

3

)

1/2

10

33

cm and t

p

(G/c

5

)

1/2

10

43

s, which are viewed as the most

fundamental scales. The Planck scale arises naturally in attempts to formulate a quantum the-

ory of gravity, since

p

and t

p

are unique combinations of speed of light c, Planck constant ,

and gravitational constant G, which have the dimensions of length and time respectively. The

dimensional arguments suggest that at Planck scale the smooth structure of spacetime should

break down, and therefore the well-known quantum eld theory is invalid since it depends on

a xed smooth background spacetime. Hence we believe that physicists should go beyond the

successful standard model to explore the new physics near Planck scale, which is, perhaps, a

quantum eld theory without a background spacetime, and this quantum eld theory should

include the quantum theory of gravity. Moreover, current theoretical physics is thirsting for a

quantum theory of gravity to solve at least the following fundamental diculties.

Classical Gravity - Quantum Matter Inconsistency

The equation relating matter and the gravitational eld is the famous Einstein eld equa-

tion:

R

[g]

1

2

R[g]g

= T

[g], (1)

where the left hand side of the equation concerns spacetime geometry which has classical

smooth structure, while the right hand side concerns also matter eld which is funda-

mentally quantum mechanical in standard model. In quantum eld theory the energy-

momentum tensor of matter eld should be an operator-valued tensor

T

. One possible

way to keep classical geometry consistent with quantum matter is to replace T

[g] by

the expectation value <

T

xed spacetime. However, in the solution of this equation the background g

has to be

changed due to the non-vanishing of <

T

1

into the denition of the vacuum expectation value etc. The result of the iterations does

not converge in general [70]. On the other hand, some other arguments show that such a

semiclassical treatment may violate the principle of superposition in quantum mechanics

[55]. This inconsistency motivates us to quantize the background geometry to arrive at an

operator formula also on the left hand side of Eq.(1).

Singularities in General Relativity

Einsteins theory of General Relativity is considered as one of the most elegant theories

in the 20th century. Many experimental tests conrm the theory in the classical domain

[159]. However, Penrose and Hawking proved that singularities are inevitable in general

spacetimes with matter satisfying certain conditions in, by now well known, singularity

theorems (for a summary, see [91][156]). Thus general relativity as a classical theory

breaks down in certain regions of spacetime in a generic way. One naturally expects that,

in extra strong gravitational eld domains near the singularities, the gravitational theory

would probably be replaced by an unknown quantum theory of gravity.

Innities in Quantum Field Theory

It is well known that there are innity problems in quantum eld theory in Minkowski

spacetime. In curved spacetime, the problem of divergences is even more complicated,

since the renormalization process in curved spacetime is ambiguous, the expectation value

of stress tensor can be xed up to some local curvature terms, and it also depends on a

fundamental scale of spacetime. Although much progress on the renormalization have

been made [92][157], a fundamentally satisfactory theory is still far from reaching. So it

is expected that some quantum gravity theory, playing a fundamental role at Planck scale,

could provide a natural cut-o to cure the innities in quantum eld theory. The situation

of quantum eld theory on a xed spacetime looks just like that of quantum mechanics

for particles in electromagnetic eld before the establishing of quantum electrodynamics,

where the particle mechanics (actress) is quantized but the background electromagnetic

eld (stage) is classical. The history suggests that such a semi-classical situation is only

an approximation which should be replaced by a much more fundamental and satisfactory

theory.

1.2 Purpose of Loop Quantum Gravity

The research on quantum gravity is quite active. Many quantization programmes for gravity

are being carried out (for a summary see e.g. [146]). In these dierent kinds of approaches,

Among these dierent kinds of approaches, the idea of loop quantum gravity nds its roots in

researchers from the general relativity community. It follows closely the motivations of general

relativity, and hence it is a quantum theory born with background independence. Roughly

speaking, loop quantum gravity is an attempt to construct a mathematically rigorous, non-

perturbative, background independent quantum theory of four-dimensional, Lorentzian general

relativity plus all known matter in the continuum. The project of loop quantum gravity inherits

the basic idea of Einstein that gravity is fundamentally spacetime geometry. Here one believes

2

in that the theory of quantum gravity is a quantum theory of spacetime geometry with dieo-

morphism invariance (this legacy is discussed comprehensively in Rovellis book [122]). To

carry out the quantization procedure, one rst casts general relativity into the Hamiltonian for-

malism as a dieomorphism invariant Yang-Mills gauge eld theory with a compact internal

gauge group. Thus the construction of loop quantum gravity can also be applied to all back-

ground independent gauge eld theories. One can therefore claim that the theory can also be

called as a background independent quantum gauge eld theory.

All classical elds theories, other than the gravitational eld, are dened on a xed space-

time, which provides a foundation to the perturbative Fock space quantization. However general

relativity is only dened on a manifold and hence is a background independent classical eld

theory, since gravity itself is the background. So the situation for gravity is much dierent

from other elds by construction [122], namely gravity is not only the background stage, but

also the dynamical actress. Such a double character for gravity leads to many diculties in the

understanding of general relativity and its quantization, since we cannot follow the strategy in

ordinary quantum theory of matter elds. However, an amazing result in loop quantum gravity

is that the background independent programme can even help us to avoid the diculties in or-

dinary quantum eld theory. In perturbative quantum eld theory in curved spacetime, the de-

nition of some basic physical quantities, such as the expectation value of energy-momentum, is

ambiguous and it is dicult to calculate the back-reaction of quantum elds to the background

spacetime [157]. One could speculate that the diculty is related to the fact that the present

formulation of quantum eld theories is background dependent. For instance, the vacuum state

of a quantum eld is closely related to spacetime structure, which plays an essential role in the

description of quantum eld theory in curved spacetime and their renormalization procedures.

However, if the quantization programme is by construction background independent and non-

perturbative, it may be possible to solve the problems fundamentally. In loop quantum gravity

there is no assumption of a priori background metric at all and the gravitational eld and matter

elds are coupled and uctuating naturally with respect to each other on a common manifold.

In the following sections, we will review pedagogically the basic construction of a com-

pletely new, background independent quantum eld theory, which is completely dierent from

the known quantum eld theory. For completeness and accuracy, we will use detailed math-

ematical terminology. However, for simplicity, we will skip the complicated proofs of many

important statements. One may nd the missing details in the references cited. Thus our review

will not be comprehensive. We refer to Ref.[146] and [90] for a more detailed exploration,

Refs. [20] and [148] for more advanced topics. It turns out that in the framework of loop

quantum gravity all theoretical inconsistencies introduced in the previous section are likely to

be cured. More precisely, one will see that there is no UV divergence in quantum elds of

matter if they are coupled with gravity in the background independent approach. Also recent

works show that the singularities in general relativity can be smeared out in symmetry-reduced

models [43][101][48]. The crucial point is that gravity and matter are coupled and consistently

quantized non-perturbatively so that the problems of classical gravity and quantum matter in-

consistency disappear.

3

2 Classical Framework

2.1 Lagrangian Formalism

In order to canonically quantize classical gravity, a Hamiltonian analysis has to be performed

to obtain a canonical formalism of the classical theory suitable to be represented on a Hilbert

space. A well known canonical formalism of general relativity is the ADM formalism (ge-

ometrodynamics) derived from the EinsteinHilbert action[156][97], which has been problem-

atic to cast into a quantum theory rigorously. Another well-known action of general relativity

is the Palatini formalism, where the tetrad and the connection are regarded as independent dy-

namical variables. However, the dynamics of the Palatini action has similar diculties at the

time of quantization as the dynamics derived from the EinsteinHilbert action [4][87]. In 1986,

Ashtekar presented a formalism of true connection dynamics for general relativity with a rela-

tively simple Hamiltonian constraint, and thus opened the door to apply quantization techniques

from gauge eld theory [2][3][123]. However, a drawback of that formalism is that the canon-

ical variables are complex, which need the implementation of complicated reality conditions

if one is to represent real general relativity. Moreover, the quantization based on the complex

connection could not be carried out rigorously, since the internal gauge group is noncompact.

In 1995, Barbero modied the Ashtekar new variables to give a system of real canonical vari-

ables for dynamical theory of connections [34]. Then Holst constructed a generalized Palatini

action to support Barberos real connection dynamics [93]. Although there is a free parameter

(Barbero-Immirzi parameter ) in the generalized Palatini action and the Hamiltonian constraint

is more complicated than the Ashtekar one, the generalized Palatini Hamiltonian with the real

connections is widely used by loop theorists for the quantization programme. All the following

analysis is based on the generalized Palatini formalism.

Consider a 4-manifold, M, on which the basic dynamical variables in the generalized Pala-

tini framework are a tetrad e

I

and an so(1, 3)-valued connection

I J

free), where the capital Latin indices I, J, ... refer to the internal S O(1, 3) group and the Greek

indices , , ... denote spacetime indices. A tensor with both spacetime indices and internal

indices is named as a generalized tensor. The internal space is equipped with a Minkowskian

metric

I J

(of signature , +, +, +) xed once for all, such that the spacetime metric reads:

g

=

I J

e

I

e

J

.

The generalized Palatini action in which we are interested is given by [20]:

S

p

[e

K

,

I J

] =

1

2

_

M

d

4

x(e)e

I

e

J

(

I J

+

1

2

I J

KL

KL

), (2)

where e is the square root of the determinant of the metric g

,

I J

KL

is the internal Levi-Civita

symbol, is the real Barbero-Immirzi parameter, and the so(1, 3)-valued curvature 2-form

I J

of the connection

I J

reads:

I J

:= 21

[

I J

]

=

I J

I J

+

IK

J

K

,

here 1

I J

acting on both

spacetime and internal indices. Note that the generalized Palatini action returns to the Palatini

4

action when

1

1

= i.

Moreover, besides spacetime dieomorphism transformations, the action is also invariant under

internal S O(1, 3) rotations:

(e, ) (e

) = (b

1

e, b

1

b + b

1

db),

for any S O(1, 3) valued function b on M. The gravitational eld equations are obtained by

varying this action with respect to e

I

and

I J

connection

I J

. One has

I J

= (d

I J

)

+

IK

J

K

+

IK

J

K

= 21

[

I J

]

by the denition of covariant generalized derivative 1

. Note that

I J

is a Lorentz covariant

generalized tensor eld since it is the dierence between two Lorentz connections [107][104].

One thus obtains

S

p

=

1

2

_

M

d

4

x(e)e

I

e

J

(

I J

+

1

2

I J

KL

KL

)

=

1

_

M

(

I J

+

1

2

I J

KL

KL

)1

[(e)e

I

e

J

],

where we have used the fact that 1

of weight +1 and

neglected the surface term. Then it gives the equation of motion:

1

[(e)e

I

e

J

] =

1

4

1

I JKL

e

K

e

L

] = 0,

where

is the spacetime Levi-Civita symbol. This equation leads to the torsion-free Cartans

rst equation:

1

[

e

I

]

= 0,

which means that the connection

I J

compatible with the tetrad e

I

. As a result, the second term in the action (2) can be calculated

as:

(e)e

I

e

I JKL

KL

=

,

which is exactly vanishing, because of the symmetric properties of Riemann tensor. So the

generalized Palatini action returns to the Palatini action, which will certainly give the Einstein

eld equation.

2.2 Hamiltonian Formalism

To carry out the Hamiltonian analysis of action (2), suppose the spacetime M is topologically

R for some 3-dimensional compact manifold without boundary. We introduce a foliation

5

parameterized by a smooth function t and a time-evolution vector eld t

such that t

(dt)

= 1

in M, where t

of as:

t

= Nn

+ N

, (3)

here N is called the lapse function and N

is dened as n

I

n

I

. It is convenient to carry out a partial gauge xing, i.e., x a internal

constant vector eld n

I

with

I J

n

I

n

J

= 1. Note that the gauge xing puts no restriction on

the real dynamics

1

. Then the internal vector space V is 3+1 decomposed with a 3-dimensional

subspace W orthogonal to n

I

, which will be the internal space on . With respect to the internal

normal n

I

and spacetime normal n

q

I

i

and q

a

respectively, where we use i, j, k, ... to denote the 3-dimensional internal space indices

and a, b, c, ... to denote the indices of space . Then an internal reduced metric

i j

and a reduced

spatial metric on , q

ab

, are obtained by these two projection maps. The two metrics are related

by:

q

ab

=

i j

e

i

a

e

j

b

, (4)

where the orthonormal co-triad on is dened by e

i

a

:= e

I

q

i

I

q

a

. Now the internal gauge group

S O(1, 3) is reduced to its subgroup S O(3) which leaves n

I

invariant. Finally, two Levi-Civita

symbols are obtained respectively as

i jk

:= q

I

i

q

J

j

q

K

k

n

L

LI JK

,

abc

:= q

a

q

b

q

c

t

,

where the internal Levi-Civita symbol

i jk

is an isomorphism of Lie algebra so(3). Using the

connection 1-form

I J

i

a

:=

1

2

q

a

q

i

I

I J

KL

n

J

KL

,

K

i

a

:= q

i

I

q

I J

n

J

,

where is a spin connection on and K will be related to the extrinsic curvature of on shell.

After the 3+1 decomposition and the Legendre transformation, action (2) can be expressed as

[93]:

S

p

=

_

R

dt

_

d

3

x[

P

a

i

[

t

A

i

a

1

tot

(A

i

a

,

P

b

j

,

i

, N, N

c

)], (5)

from which the symplectic structure on the classical phase space is obtained as

A

i

a

(x),

P

b

j

(y) :=

i

j

a

b

3

(x, y), (6)

where the conguration and conjugate momentum are dened respectively by:

A

i

a

:=

i

a

+ K

i

a

,

P

a

i

:=

1

2

abc

i jk

e

j

b

e

k

c

=

1

_

det qe

a

i

,

1

However, there are some arguments that such a gauge xing is a non-natural way to break the internal Lorentz

symmetry (see e.g. [131]).

6

here det q is the determinant of the 3-metric q

ab

on and hence det q = ()

3

det P. In the

denition of the conguration variable A

i

a

, we should emphasize that

i

a

is restricted to be the

unique torsion free so(3)-valued spin connection compatible with the triad e

a

i

. This conclusion

is obtained by solving a second class constraint in the Hamiltonian analysis [93]. In the Hamil-

tonian formalism, one starts with the elds (A

i

a

,

P

a

i

). Then neither the basic dynamical variables

nor their Poisson brackets depend on the Barbero-Immirzi parameter . Hence, for the case

of pure gravitational eld, the dynamical theories with dierent are related by a canonical

transformation. However, as we will see, the spectrum of geometric operators are modied by

dierent value of , and the non-perturbative calculation of black hole entropy is compatible

with Bekenstein-Hawkings formula only for a specic value of [68]. In addition, it is argued

that the Barbero-Immerzi parameter may lead to observable eects in principle when the grav-

itational eld is coupled with Fermions [112]. In the decomposed action (5), the Hamiltonian

density 1

tot

is a linear combination of constraints:

1

tot

=

i

G

i

+ N

a

C

a

+ NC,

where

i

1

2

i

jk

jk

t

, N

a

and N are Lagrange multipliers. The three kinds of constraints in the

Hamiltonian are expressed as [20]:

G

i

= D

a

P

a

i

:=

a

P

a

i

+

k

i j

A

j

a

P

a

k

,

C

a

=

P

b

i

F

i

ab

1 +

2

K

i

a

G

i

,

C =

2

2

_

det q

P

a

i

P

b

j

[

i j

k

F

k

ab

2(1 +

2

)K

i

[a

K

j

b]

]

+ (1 +

2

)

a

(

P

a

i

_

det q

)G

i

, (7)

where the conguration variable A

i

a

performs as a so(3)-valued connection on and F

i

ab

is the

so(3)-valued curvature 2-form of A

i

a

with the well-known expression:

F

i

ab

:= 2D

[a

A

i

b]

=

a

A

i

b

b

A

i

a

+

i

jk

A

j

a

A

k

b

.

In any dynamical system with constraints, the constraint analysis is essentially important be-

cause they reect the gauge invariance of the system. From the above three kinds of constraints

of general relativity, one can know the gauge invariance of the theory. The Gauss constraint

G

i

= 0 has crucial importance in formulating the general relativity into a dynamical theory of

connections. The corresponding smeared constraint function, g() :=

_

d

3

x

i

(x)G

i

(x), gener-

ates a transformation on the phase space as:

A

i

a

(x), g() = D

a

i

(x)

P

a

i

(x), g() =

k

i j

j

(x)

P

a

k

(x),

which are just the innitesimal versions of the following gauge transformation for the so(3)-

valued connection 1-form A and internal rotation for the so(3)-valued densitized vector eld

P

respectively:

(A

a

,

P

b

) (g

1

A

a

g + g

1

(dg)

a

, g

1

P

b

g).

7

To display the meaning of the vector constraint C

a

= 0, one may consider the smeared constraint

function:

+(

N) :=

_

d

3

x(N

a

P

b

i

F

i

ab

(N

a

A

i

a

)G

i

).

It generates the innitesimal spatial dieomorphism by the vector eld N

a

on as:

A

i

a

(x), +(

N) = [

N

A

i

a

(x),

P

a

i

(x), +(

N) = [

P

a

i

(x).

The smeared scalar constraint is weakly equivalent to the following function, which is re-

expressed for quantization purpose as

S(N) :=

_

d

3

xN(x)

C(x)

=

2

2

_

d

3

xN

P

a

i

P

b

j

_

det q

[

i j

k

F

k

ab

2(1 +

2

)K

i

[a

K

j

b]

]. (8)

It generates the innitesimal time evolution o . The constraints algebra, i.e., the Poisson

brackets between these constraints, play a crucial role in the quantization programme. It can be

shown that the constraints algebra of (7) has the following form:

g(), g(

) = g([,

]),

g(), +(

N) = g([

N

),

g(), S(N) = 0,

+(

N), +(

) = +([

N,

N

]),

+(

N

M),

S(N), S(M) = +((N

b

M M

b

N)q

ab

)

g((N

b

M M

b

N)q

ab

A

a

))

(1 +

2

)g(

[

P

a

a

N,

P

b

b

M]

det q

), (9)

where det qq

ab

=

2

P

a

i

P

b

j

i j

. Hence the constraints algebra is closed under the Poisson

brackets, i.e., the constraints are all rst class. Note that the evolution of constraints is consistent

since the Hamiltonian H =

_

d

3

x1

tot

is a linear combination of the constraints functions. The

evolution equations of the basic canonical pair read

[

t

A

i

a

= A

i

a

, H, [

t

P

a

i

=

P

a

i

, H.

Together with the three constraint equations, they are completely equivalent to the Einstein

eld equations. Thus general relativity is cast as a dynamical theory of connections with a

compact structure group. Before nishing the discussion of this section, several issues should

be emphasized.

8

Canonical Transformation Viewpoint

The above construction can be reformulated in the language of canonical transformations,

since the phase space of connection dynamics is the same as that of triad geometrodynam-

ics. In the triad formalism the basic conjugate pair consists of densitized triad

E

a

i

=

P

a

i

and extrinsic curvature K

i

a

. The Hamiltonian and constraints read

1

tot

=

i

G

i

+ N

a

C

a

+ NC

G

i

=

k

i j

K

j

a

E

a

k

, (10)

C

a

=

E

b

j

[a

K

j

b]

, (11)

C =

1

_

det q

[

1

2

det qR +

E

[a

i

E

b]

j

K

i

a

K

j

b

], (12)

where

a

is the S O(3) generalized derivative operator compatible with triad e

a

i

and R is

the scalar curvature with respect to it. Since

E

a

i

is a vector density of weight one, we have

E

a

i

=

a

E

a

i

+

k

i j

j

a

E

a

k

= 0.

One can construct the desired Gauss law by

G

i

:=

1

E

a

i

+ G

i

,

=

a

P

a

i

+

k

i j

(

j

a

+ K

j

a

)

P

a

k

,

which is weakly zero by construction. This motivates us to dene the connection A

a

i

=

i

a

+K

i

a

. Moreover, the transformation from the pair (

E

a

i

, K

j

b

) to (

P

a

i

, A

j

b

) can be proved to

be a canonical transformation [34][146], i.e., the Poisson algebra of the basic dynamical

variables is preserved under the transformation:

E

a

i

P

a

i

=

E

a

i

/

K

j

b

A

j

b

=

j

b

+ K

j

b

,

as

P

a

i

(x), A

j

b

(y) =

E

a

i

(x), K

j

b

(y) =

a

b

j

i

(x, y),

A

i

a

(x), A

j

b

(y) = K

i

a

(x), K

j

b

(y) = 0,

P

a

i

(x),

P

b

j

(y) =

E

a

i

(x),

E

b

j

(y) = 0.

The Preparation for Quantization

The advantage of a dynamical theory of connections is that it is convenient to be quan-

tized in a background independent fashion. In the following procedure of quantization,

the quantum algebra of the elementary observables will be generated by holonomy, i.e.,

connection smeared on a curve, and electric ux, i.e., a densitized triad smeared on a 2-

surface. So no information about a background geometry is needed to dene the quantum

9

algebra. In the remainder of the thesis, in order to incorporate also spinors, we will en-

large the internal gauge group to be S U(2). This does not damage the prior constructions

because the Lie algebra of S U(2) is the same as that of S O(3). Due to the well-known

nice properties of compact Lie group S U(2), such as the Haar measure and Peter-Weyl

theorem, one can obtain the background independent representation of the quantum alge-

bra and the spin-network decomposition of the kinematic Hilbert space.

Analysis on Constraint Algebra

The classical constraint algebra (9) is an innite dimensional Poisson algebra. Unfortu-

nately, it is not a Lie algebra since the Poisson bracket between two scalar constraints has

structure function depending on dynamical variables. This causes problems when solving

the constraints quantum mechanically. On the other hand, one can see from Eq.(9) that

the algebra generated by Gauss constraints forms not only a subalgebra but also a 2-side

ideal in the full constraint algebra. Thus one can rst solve the Gauss constraints indepen-

dently. It is convenient to nd the quotient algebra with respect to the Gauss constraint

subalgebra as

+(

N), +(

) = +([

N,

N

]),

+(

N

M),

S(N), S(M) = +((N

b

M M

b

N)q

ab

),

which plays a crucial role in solving the constraints quantum mechanically. But the subal-

gebra generated by the dieomorphism constraints can not form an ideal. Hence the pro-

cedures of solving the dieomorphism constraints and solving Hamiltonian constraints

are entangled with each other. This leads to certain ambiguity in the construction of a

Hamiltonian constraint operator [134][149]. Fortunately, the Master Constraint Project

addresses the above two problems as a whole by introducing a new classical constraint al-

gebra [149]. The new algebra is a Lie algebra where the dieomorphism constraints form

a 2-side ideal. We will come back to this point in the discussion on quantum dynamics of

loop quantum gravity.

10

3 Foundations of Loop Quantum Gravity

In this chapter, we will begin to quantize the above classical dynamics of connections as a

background independent quantum eld theory. The main purpose of the chapter is to construct

a suitable kinematical Hilbert space 1

kin

for the representation of quantum observables. In the

following discussion, we formulate the construction in the language of algebraic quantum eld

theory [85]. It should be emphasized that the following constructions can be generalized to all

background independent non-perturbative gauge eld theories with compact gauge groups.

3.1 General Programme for Algebraic Quantization

In the strategy of loop quantum gravity, a canonical programme is performed to quantize gen-

eral relativity, which has been cast into a dieomorphism invariant gauge eld theory, or more

generally, a dynamical system with constraints. The following is a summary for a general pro-

cedure to quantize a dynamical system with rst class constraints [133][4].

Algebra of Classical Elementary Observables

One starts with the classical phase space (A, , ) and R (R can be countable innity

2

)

rst-class constraints C

r

(r = 1...R) such that C

r

, C

s

=

R

t=1

f

t

rs

C

t

, where f

t

rs

is generally

a function on phase space, namely, the structure function of Poisson algebra. The algebra

of classical elementary observables P is dened as:

Denition 3.1.1: The algebra of classical elementary observables P is a collection of

functions f (m), m Aon the phase space such that

(1) f (m) Pshould separate the points of A, i.e., for any m m

such that f (m) f (m

R.)

(2) f (m), f

(m) P f (m), f

(3) f (m) P

f (m) P (closed under complex conjugation).

So P forms a sub -Poisson algebra of C

R, P is gener-

ated by the conjugate pair (q, p) with q, p = 1.

Quantum Algebra of Elementary Observables

Given the algebra of classical elementary observables P, a quantumalgebra of elementary

observables can be constructed as follows. Consider the formal nite sequences of classi-

cal observable ( f

1

... f

n

) with f

k

P. Then the operations of multiplication and involution

are dened as

( f

1

, ..., f

n

) ( f

1

, ..., f

m

) := ( f

1

, ..., f

n

, f

1

, ..., f

m

),

( f

1

, .., f

n

)

:= (

f

n

, ...,

f

1

).

One can dene the formal sum of dierent sequences with dierent number of elements.

Then the general element of the newly constructed free -algebra F(P) of P, is formally

2

This includes the case of eld theory with innite many degree of freedom, since one can introduce the

expression C

n,

=

_

d

3

x

n

(x)C

(x), where

n

(x)

n=1

forms a system of basis in L

2

(, d

3

x).

11

expressed as

_

N

k=1

( f

(k)

1

, ... f

(k)

n

k

), where f

(i)

n

i

P. Consider the elements of the form (se-

quences consisting of only one element)

( f + f

) ( f ) ( f

), (z f ) z( f ), [( f ), ( f

)] i( f , f

),

where z C is a complex number, and the canonical commutation bracket is dened as

[( f ), ( f

)] := ( f ) ( f

) ( f

) ( f ).

A 2-side ideal I of F(P) can be constructed from these element, and is preserved by the

action of involution . One thus obtains

Denition 3.1.2: The quantum algebra A of elementary observables is dened to be the

quotient -algebra F(P)/I.

Note that the motivation to construct a quantum algebra of elementary observables is

to avoid the problem of operators ordering in quantum theory so that the quantum algebra

A can be represented on a Hilbert space without ordering ambiguities.

Representation of Quantum Algebra

In order to obtain a quantum theory, we need to quantize the classical observables in the

dynamical system. The, so called, quantization is nothing but a -representation map

3

from the quantum algebra of elementary observable A to the collection of linear opera-

tors [(1) on a Hilbert Space 1. At the level of quantum mechanics, the well-known

Stone-Von Neumann Theorem concludes that in quantum mechanics, there is only one

strongly continuous, irreducible, unitary representation of the Weyl algebra, up to uni-

tary equivalence (see, for example, Ref.[113]). However, the conclusion of Stone-Von

Neumann cannot be generalized to the quantum eld theory because the latter has innite

many degrees of freedom (for detail, see, for example [157]). In quantum eld theory,

a representation can be constructed by GNS(Gelfand-Naimark-Segal)-construction for a

quantum algebra of elementary observable A, which is a unital -algebra actually. The

GNS-construction for the representation of quantum algebra A is briey summarized as

follows.

Denition 3.1.3: Given a positive linear functional (a state) on A, the null space

N

:= a A(a

in A. Then a quotient map can be dened as [.]: A A/N

; a [a] := a + bb N

.

3

A map : A [(1) is a *-representation if and only if (1) there exists a dense subspace 1 of 1 contained

in

aA

[D((a)) D((a

))] where D((a)) is the domain of the operator (a) and (2) for every a, b A and C

the following conditions are satised in 1,

(a + b) = (a) + (b), (a) = (a),

(a b) = (a)(b), (a

) = (a)

.

Note that [(1) fails to be an algebra because the domains of unbounded operators cannot be the whole Hilbert

space. However, the collection of bounded operators on any Hilbert space is really a -algebra.

12

The GNS-representation for A with respect to is a -representation map:

: A

[(1

), where 1

:= (A/N

equipped well-dened inner product

< [a][b] >

1

:= (a

b)

on 1

,

where

cyclic representation, i.e.,

a A) = 1

and

is called

a cyclic vector in the representation space. In fact

and

(

(a)

a A) = 1

can be expressed as

(a) =<

(a)

>

1

.

Thus a positive linear functional on A is equivalent to a cyclic representation of A, which

is a triple (1

direct sum of cyclic representations ( for proof, see [58] ) .

So the kinematical Hilbert space 1

kin

= 1

tained by GNS-construction. In the case that there are gauge symmetries in our dy-

namical system, supposing that there is a gauge group G acting on A by automorphisms

g

: A A, g G, it is preferred to construct a gauge invariant representation of A.

So we require the positive linear functional on A to be gauge invariant, i.e.,

g

= .

Then the group G is represented on the Hilbert space 1

as:

U(g)

(a)

(

g

(a))

,

and such a representation is a unitary representation of G. In loop quantum gravity, it is

crucial to construct an internal gauge invariant and dieomorphism invariant representa-

tion for the quantum algebra of elementary observables.

Implementation of the Constraints

In the Dirac quantization programme for a system with constraints, the constraints should

be quantized as some operators in a kinematical Hilbert space 1

kin

. One then solves

them at quantum level to get a physical Hilbert space 1

phys

, that is, to nd a quantum

analogy

C

r

of the classical constraint formula C

r

and to solve the general solution of

the equation

C

r

= 0. However, there are several problems in the construction of the

constraint operator

C

r

.

(i) C

r

is in general not in P, so there is a factor ordering ambiguity in quantizing C

r

to

be an operator

C

r

.

(ii) In quantum eld theory, there are ultraviolet(UV) divergence problems in construct-

ing operators. However, the UV divergence can be avoided in the background inde-

pendent approach.

13

(iii) Sometimes, quantum anomalies may appear when there are structure functions in

the Poisson algebra. Although classically we have C

r

, C

s

=

R

t=1

f

t

rs

C

t

, r, s, t =

1, ..., R, where f

t

rs

is a function on phase space, quantum mechanically it is possible

that [

C

r

,

C

s

] i

R

t=1

f

t

rs

C

t

due to the ordering ambiguity between

f

t

rs

and

C

t

. If one

sets [

C

r

,

C

s

] =

i

2

R

t=1

(

f

t

rs

C

t

+

C

t

f

t

rs

), for satisfying

C

r

= 0, we have

[

C

r

,

C

s

] =

i

2

R

t=1

C

t

f

t

rs

=

i

2

R

t=1

[

C

t

,

f

t

rs

]. (13)

However, [

C

t

,

f

t

rs

] are not necessary to equal to zero for all r, s, t = 1...R. If

not, the problem of quantum anomaly appears and the new quantum constraints

[

C

t

,

f

t

rs

] = 0 have to be imposed on physical quantum states, since the classical

Poisson brackets C

r

, C

s

are weakly equal to zero on the constraint surface A A.

Thus too many constraints are imposed and the physical Hilbert space 1

phys

would

be too small. Therefore this is not a satisfactory solution and one needs to nd a

way to avoid the quantum anomalies.

Solving the Constraints and Physical Hilbert Space

In general the original Dirac quantization approach can not be carried out directly, since

there is usually no nontrivial 1

kin

such that

C

r

= 0. This happens when the con-

straint operator

C

r

has generalized eigenfunctions rather than eigenfunctions. One then

develops the so-called Rened Algebraic Quantization Programme, where the solutions

of the quantum constraint can be found in the algebraic dual space of a dense subset in

1

kin

(see e.g. [84]). The quantum dieomorphism constraint in loop quantum gravity is

solved in this way. Another interesting way to solve the quantum constraints is the Mas-

ter Constraint Approach proposed by Thiemann recently [149], which seems especially

suited to deal with the particular feature of the constraint algebra of general relativity. A

master constraint is dened as M :=

1

2

R

r,s=1

K

rs

C

s

C

r

for some real positive matrix K

rs

.

Classically one has M = 0 if and only if C

r

= 0 for all r = 1...R. So quantum mechan-

ically one may consider solving the Master Equation:

M = 0 to obtain the physical

Hilbert space 1

phys

instead of solving

C

r

= 0, r = 1...R. Because the master con-

straint M is classically positive, one has opportunities to implement it as a self-adjoint

operator on 1

kin

. If it is indeed the case and 1

kin

is separable, one can use the direct in-

tegral representation of 1

kin

associated with the self-adjoint operator

M to obtain 1

phys

:

1

kin

_

R

d()1

,

< >

kin

=

_

R

d() < >

1

, (14)

where is a so-called spectral measure and 1

M with

the eigenvalue . The physical Hilbert space is then formally obtained as 1

phys

= 1

=0

with the induced physical inner product < >

1

=0

4

. Now the issue of quantum anomaly

4

One need to be careful for such a formal prescription, see the later discussion of master constraint or [63].

14

is represented in terms of the size of 1

phys

and the existence of sucient numbers of

semi-classical states.

Physical Observables

We denote Aas the original unconstrained phase space, Aas the constraint surface, i.e.,

A := m AC

r

(m) = 0, r = 1...R, and

A as the reduced phase space, i.e. the space

of orbits for gauge transformations generated by all C

r

. The concept of Dirac observable

is dened as the follows.

Denition 3.1.4:

(1) A function O on A is called a weak Dirac observable if and only if the function de-

pends only on points of

A, i.e., O, C

r

A

= 0 for all r = 1...R. For the quantum version,

a self-adjoint operator

O is a weak Dirac observable if and only if the operator can be

well dened on the physical Hilbert space.

(2) A function O on A is called a strong Dirac observable if and only if O, C

r

A

= 0

for all r = 1...R. For the quantum version, a self-adjoint operator

O is a strong Dirac

observable if and only if the operator can be dened on the kinematic Hilbert space 1

kin

and [

O,

C

r

] = 0 in 1

kin

for all r = 1...R.

A physical observable is at least a weak Dirac observable. While Dirac observables have

been found explicitly in symmetry reduced models, some even with an innite number

of degrees of freedom, it seems extremely dicult to nd explicit expressions for them

in full general relativity. Moreover the Hamiltonian is a linear combination of rst-class

constraints. So there is no dynamics in the reduced phase space, and the meaning of

time evolution of the Dirac observables becomes subtle. However, using the concepts of

partial and complete observables [121][115][122], a systematic method to get Dirac ob-

servables can be developed, and the problem of time in such system with a Hamiltonian

H =

R

r=1

r

C

r

may also be solved.

Classically, let f (m) and T

r

(m)

R

r=1

be gauge non-invariant functions (partial observables)

on phase space A, such that A

sr

C

s

, T

r

is a non-degenerate matrix on A. A system

of classical weak Dirac observables (complete observables) F

f ,T

labelled by a collection

of real parameters

r

R

r=1

can be constructed as

F

f ,T

:=

n

1

n

R

(

1

T

1

)

n

1

(

R

T

R

)

n

R

n

1

! n

R

!

X

n

1

1

X

n

R

R

( f ),

where

X

r

( f ) :=

R

s=1

A

1

rs

C

s

, f

C

r

, f . It can be veried that [

X

r

,

X

s

]

A

= 0 and

F

f ,T

, C

r

A

= 0, for all r = 1...R (for details see [61] and [62]).

The partial observables T

r

(m)

R

r=1

may be regarded as clock variables, and

r

is the time

parameter for T

r

. The gauge is xed by giving a system of functions T

r

(m)

R

r=1

and

corresponding parameters

r

R

r=1

, namely, a section in A is selected by T

r

(m) =

r

for

each r, and F

f ,T

is the value of f on the section. To solve the problem of dynamics, one

assumes another set of canonical coordinates (P

1

, , P

NR

,

1

, ,

R

; Q

1

, , Q

NR

, T

1

,

15

, T

R

) by canonical transformations in the phase space (A, , ), where P

s

and

r

are

conjugate to Q

s

and T

r

respectively. After solving the complete system of constraints

C

r

(P

i

, Q

j

,

s

, T

t

) = 0

R

r=1

, the Hamiltonian H

r

with respect to the time T

r

is obtained as

H

r

:=

r

(P

i

, Q

j

, T

t

). Given a system of constants (

0

)

r

R

r=1

, for an observable f (P

i

, Q

j

)

depending only on P

i

and Q

j

, the physical dynamics is given by [61][150]:

(

r

)

=

0

F

f ,T

A

= F

0

H

r

, f ,T

A

= F

0

H

r

,T

, F

0

f ,T

A

,

where F

0

H

r

,T

is the physical Hamiltonian function generating the evolution with respect to

r

. Thus one has addressed the problem of time and dynamics as a result.

Semi-classical Analysis

An important issue in the quantization is to check whether the quantum constraint oper-

ators have correct classical limits. This has to be done by using the kinematical semi-

classical states in 1

kin

. Moreover, the physical Hilbert space 1

phys

must contain enough

semi-classical states to guarantee that the quantum theory one obtains can return to the

classical theory when 0. The semi-classical states in a Hilbert space 1 should have

the following properties.

Denition 3.1.5: Given a class of observables S which comprises a subalgebra in the

space [(1) of linear operators on the Hilbert space, a family of (pure) states

m

mA

are said to be semi-classical with respect to S if and only if:

(1) The observables in S should have correct semi-classical limit on semi-classical states

and the uctuations should be small, i.e.,

lim

0

m

( a) a(m)

a(m)

= 0,

lim

0

m

( a

2

)

m

( a)

2

m

( a)

2

= 0,

for all a S.

(2) The set of cyclic vectors

m

related to

m

via the GNS -representation (

, 1

)

is dense in 1.

Seeking for semiclassical states are one of open issues of current research in loop quantum

gravity. Recent original works focus on the construction of coherent states of loop quan-

tum gravity in analogy with the coherent states for harmonic oscillator system [142][143]

[144][145][19][15].

The above is the general programme to quantize a system with constraints. In the following sub-

section, we will apply the programme to the theory of general relativity and restrict our view to

the representation with the properties of background independence and spatial dieomorphism

invariance.

16

3.2 Quantum Conguration Space

In quantum mechanics, the kinematical Hilbert space is L

2

(R

3

, d

3

x), where the simple R

3

is the

classical conguration space of free particle which has nite degrees of freedom, and d

3

x is the

Lebesgue measure on R

3

. In quantum eld theory, it is expected that the kinematical Hilbert

space is also the L

2

space on the conguration space of the eld, which is innite dimensional,

with respect to some Borel measure naturally dened. However, it is often hard to dene a

concrete Borel measure on the classical conguration space, since the integral theory on innite

dimensional space is involved [57]. Thus the intuitive expectation should be modied, and the

concept of quantum conguration space should be introduced as a suitable enlargement of the

classical conguration space so that an innite dimensional measure, often called cylindrical

measure, can be well dened on it. The example of a scalar eld can be found in the refer-

ences [20][24]. For quantum gravity, it should be emphasized that the construction for quantum

conguration space must be background independent. Fortunately, general relativity has been

reformulated as a dynamical theory of S U(2) connections, which would be great helpful for our

further development.

The classical conguration space for gravitational eld, which is denoted by 7, is a col-

lection of the su(2)-valued connection 1-form eld smoothly distributed on . The idea of the

construction for quantum conguration is due to the concept of holonomy.

Denition 3.2.1: Given a smooth S U(2) connection eld A

i

a

and an analytic curve c with the

parameter t [0, 1] supported on a compact subset (compact support ) of , the corresponding

holonomy is dened by the solution of the parallel transport equation [104]

d

dt

A(c, t) = [A

i

a

c

a

i

]A(c, t), (15)

with the initial value A(c, 0) = 1, where c

a

is the tangent vector of the curve and

i

su(2)

constitute an orthonormal basis with respect to the Killing-Cartan metric (, ) := 2Tr(),

which satisfy [

i

,

j

] =

k

i j

k

and are xed once for all. Thus the holonomy is an element in

S U(2), which can be expressed as

A(c) = /exp (

_

1

0

[A

i

a

c

a

i

] dt), (16)

where A(c) S U(2) and / is a path-ordering operator along the curve c (see the footnote at

p382 in [104]).

The denition can be well extended to the case of piecewise analytic curves via the relation:

A(c

1

c

2

) = A(c

1

)A(c

2

), (17)

where stands for the composition of two curves. It is easy to see that a holonomy is invariant

under the re-parametrization and is covariant under changing the orientation, i.e.,

A(c

1

) = A(c)

1

. (18)

17

So one can formulate the properties of holonomy in terms of the concept of the equivalent

classes of curves.

Denition 3.2.2: Two analytic curves c and c

have the same source s(c) (beginning point ) and the same target t(c) (end point ), and the

holonomies of the two curves are equal to each other, i.e., A(c) = A(c

) A 7. A equivalent

class of analytic curves is dened to be an edge, and a piecewise analytic path is an composition

of edges.

To summarize, the holonomy is actually dened on the set / of piecewise analytic paths with

compact supports. The two properties (17) and (18) mean that each connection in 7 is a ho-

momorphism from /, which is so-called a groupoid by denition [155], to our compact gauge

group S U(2). Note that the internal gauge transformation and spatial dieomorphism act co-

variantly on a holonomy as

A(e) g(t(e))

1

A(e)g(s(e)) and A(e) A( e), (19)

for any S U(2)-valued function g(x) on and spatial dieomorphism . All above discussion

is for classical smooth connections in 7. The quantum conguration space for loop quantum

gravity can be constructed by extending the concept of holonomy, since its denition does not

depend on an extra background. One thus obtains the quantum conguration space 7 of loop

quantum gravity as the following.

Denition 3.2.3: The quantum conguration space 7is a collection of all quantum connections

A, which are algebraic homomorphism maps without any continuity assumption from the col-

lection of piecewise analytic paths with compact supports, /, on to the gauge group S U(2),

i.e., 7 := Hom(/, S U(2))

5

. Thus for any A 7and edge e in /,

A(e

1

e

2

) = A(e

1

)A(e

2

) and A(e

1

) = A(e)

1

.

The transformations of quantum connections under internal gauge transformations and dieo-

morphisms are dened by Eq.(19).

The above discussion on the smooth connections shows that the classical conguration space

7 can be understood as a subset in the quantum conguration space 7. Moreover, the Giles

theorem [82] shows precisely that a smooth connection can be recovered from its holonomies

by varying the length and location of the paths.

On the other hand, it was shown in [155][146] that the quantum conguration space 7 can

be constructed via a projective limit technique and admits a natural dened topology. To make

the discussion precise, we begin with a few denitions:

Denition 3.2.4:

5

It is easy to see that the denition of 7does not depend on the choice of local section in S U(2)-bundle, since

the internal gauge transformations leave 7invariant.

18

1. A nite set e

1

, ..., e

N

of edges is said to be independent if the edges e

i

can only intersect

each other at their sources s(e

i

) or targets t(e

i

).

2. A nite graph is a collection of a nite set e

1

, ..., e

N

of independent edges and their

vertices, i.e. their sources s(e

i

) and targets t(e

i

). We denote by E() and V() respectively

as the sets of independent edges and vertices of a given nite graph . And N

is the

number of elements in E().

3. A subgroupoid () / can be generated from by identifying V() as the set of ob-

jects and all e E() together with their inverses and nite compositions as the set of

homomorphisms. This kind of subgoupoid in / is called tame subgroupoid. () is inde-

pendent of the orientation of , so the graph can be recovered from tame subgroupoid

up to the orientations on the edges. We will also denote by N

in E() where is recovered by the tame subgroupoid .

4. [ denotes the set of all tame subgroupoids in /.

One can equip a partial order relation on [

6

, dened by

groupoid in

) in [, there

exists

) [ such that ,

, where

. Dene X

Hom(, S U(2))

as the set of all homomorphisms from the subgroupoid () to the group S U(2). Note that an

element A

e E(), so that one has a bijection : X

()

S U(2)

N

()

such that is a topological homomorphism. For any pair

projection map P

from X

to X

from

to the sub-

groupoid , and these projections satisfy the consistency condition P

= P

. Thus

a projective family X

, P

lim

(X

) is naturally obtained.

Denition 3.2.5: The projective limit lim

(X

, P

is a subset

of the direct product space X

:=

[

X

dened by

lim

(X

) := A

[

P

= A

.

Note that the projection P

.

One can equip the direct product space X

then can prove that the projective limit, lim

(X

), is a closed subset in X

Hausdor space with respect to the topology induced from X

between the projective limit and the prior constructed quantum conguration space 7. As one

might expect, there is a bijection between 7and lim

(X

) [146]:

: 7 lim

(X

);

A A

[

,

6

A partial order on [ is a relation, which is reective ( ), symmetric (

= ) and

transitive (

19

where A

means the restriction of the domain of the map A 7 = Hom(/, S U(2)). As a result,

the quantum conguration space is identied with the projective limit space and hence can be

equipped with the topology. In conclusion, the quantum conguration space 7 is constructed

to be a compact Hausdor topological space.

3.3 Cylindrical Functions on Quantum Conguration Space

Given the projective family X

, P

well dened as follows.

Denition 3.3.1: Let C(X

, two func-

tions f

C(X

) and f

C(X

by f

, if and only if P

= P

> ,

, where P

induced from P

dened to be the space of equivalent classes [ f ], i.e.,

Cyl(7) := [

C(X

)]/ .

One then can easily prove the following proposition by denition.

Proposition 3.3.1:

All continuous functions f

on X

a equivalent class [ f

for all

pull back map,

(P

)(A(e

1

), ..., A(e

N

), ..., A(e

N

)) = f

(A(e

1

), ..., A(e

N

)), (20)

where N

denotes the number of independent edges in the graph recovered from the groupoid .

On the other hand, by denition, given a cylindrical function f Cyl(7) there exists a suitable

groupoid such that f = [ f

functions f , f

map, there exists a common groupoid and f

, f

C(X

) such that f = [ f

] and f

= [ f

].

Let f , f

], and f

= [ f

operations are well dened

f + f

:= [ f

+ f

], f f

:= [ f

], z f := [z f

],

f := [

f

],

where z Cand

f denotes complex conjugate. So we construct Cyl(7) as an Abelian -algebra.

In addition, there is a unital element in the algebra because Cyl(7) contains constant functions.

Moreover, we can well dene the sup-norm for f = [ f

] by

j f j := sup

A

(A

), (21)

20

which satises the C

property j f

f j = j f j

2

. Then Cyl(7) is a unital Abelian C

-algebra, after

the completion with respect to the norm.

From the theory of C

space of continuous functions on its spectrum space via an isometric isomorphism, the so-called

Gelfand transformation (see e.g. [146]). So we have the following theorem [17][18], which

nishes this section.

Theorem 3.3.1:

(1) The space Cyl(7) has a structure of a unital Abelian C

spect to the sup-norm.

(2) Quantum conguration space 7 is the spectrum space of completed Cyl(7) such that

Cyl(7) is identical to the space C(7) of continuous functions on 7.

3.4 Loop Quantum Kinematics

In analogy with the quantization procedure of section 3.1, in this subsection we would like to

perform the background-independent construction of algebraic quantum eld theory for gen-

eral relativity. First we construct the algebra of classical observables. Taking account of the

future quantum analogs, we dene the algebra of classical observables P as the Poission -

subalgebra generated by the functions of holonomies (cylindrical functions) and the uxes of

triad elds smeared on some 2-surface. Namely, one can dene the classical algebra in analogy

with geometric quantization in nite dimensional phase space case by the so-called classical

Ashtekar-Corichi-Zapata holonomy-ux -algebra as the following [96].

Denition 3.4.1

The classical Ashtekar-Corichi-Zapata holonomy-ux -algebra is dened to be a vector space

P

ACZ

:= Cyl(7) +

C

(7), where +

C

(7) is the vector space of algebraic vector elds spanned

by the vector elds Y

f

(S ) Cyl(7), and their commutators, here the smeared ux vector

eld Y

f

(S ) is dened by acting on any cylindrical function:

Y

f

(S ) :=

_

S

abc

P

c

i

f

i

, ,

for any su(2)-valued function f

i

with compact supports on S and are cylindrical functions on

7. We equip P

ACZ

with the structure of an -Lie algebra by:

(1) Lie bracket , : P

ACZ

P

ACZ

P

ACZ

is dened by

(, Y), (

, Y

) := (Y

, [Y, Y

]),

for all (, Y), (

, Y

) P

ACZ

with ,

Cyl(7) and Y, Y

+

C

(7).

(2) Involution: p p p P

ACZ

is dened by complex conjugate of cylindrical functions

and vector elds, i.e., p := (, Y) p = (, Y) P

ACZ

, where Y := Y.

21

(3) P

ACZ

admits a natural action of Cyl(7) by

(, Y) := (

Y),

which gives P

ACZ

a module structure.

Note that the action of ux vector eld Y

f

(S ) on can be expressed explicitly on any cylindrical

function

C

1

(X

()

) via a suitable regularization[146]:

Y

f

(S )

=

_

S

abc

P

c

i

f

i

,

,

=

eE()

_

S

abc

P

c

i

f

i

, A(e)

mn

A(e)

mn

eE()

(S, e)

2

f

i

(S e)[

S e,s(e)

(A(e)

i

)

mn

S e,t(e)

(

i

A(e))

mn

]

A(e)

mn

vV()S

e at v

(S, e)

2

f

i

(v)X

(e,v)

i

,

where A(e)

mn

is the S U(2) matrix element of the holonomy along the edge e, X

(e,v)

i

is the

left(right) invariant vector eld L

(

i

)

(R

(

i

)

) of the group associated with the edge e if v is the

source(target) of edge e by denition:

L

(

i

)

(A(e)) :=

d

dt

t=0

(A(e) exp(t

i

)),

R

(

i

)

(A(e)) :=

d

dt

t=0

( exp(t

i

)A(e)),

and

(S, e) =

_

_

0, if e S = , or e lies in S ;

1, if e lies above S and e S = p;

1, if e lies below S and e S = p.

Since the surface S is oriented with normal n

a

, above means n

a

e

a

p

> 0, and below means

n

a

e

a

p

< 0, where e

a

p

is the tangent vector of e at p. And one should consider e S contained

in the set V() and some edges are written as the union of elementary edges which either lie in

S , or intersect S at their source or target. On the other hand, from the commutation relations

for the left(right) invariant vector elds, one can see that the commutators between ux vector

elds do not necessarily vanish when S S

of the non-commutativity of quantum Riemannian structures [23].

The classical Ashtekar-Corichi-Zapata holonomy-ux -algebra serves as a classical alge-

bra of elementary observables in our dynamical system of gauge elds. Then one can construct

the quantum algebra of elementary observables from P

ACZ

in analogy with Denition 3.1.2.

Denition 3.4.2[96]

22

The abstract free algebra F(P

ACZ

) of the classical -algebra is dened by the formal direct sum

of nite sequences of classical observables (p

1

, ..., p

n

) with p

k

P

ACZ

, where the operations of

multiplication and involution are dened as

(p

1

, ..., p

n

) (p

1

, ..., p

m

) := (p

1

, ..., p

n

, p

1

, ..., p

m

),

(p

1

, .., p

n

)

:= ( p

n

, ..., p

1

).

A 2-sided ideal I can be generated by the following elements,

(p + p

) (p) (p

), (zp) z(p),

[(p), (p

)] i(p, p

),

((, 0), p) ( p),

where the canonical commutation bracket is dened by

[(p), (p

)] := (p) (p

) (p

) (p).

Note that the ideal I is preserved by the involution , and the last set of generators in the ideal

I cancels the overcompleteness generated from the module structure of P

ACZ

[4].

The quantum holonomy-ux -algebra is dened by the quotient -algebra A = F(P

ACZ

)/I,

which contains the unital element 1 := ((1, 0)). Note that a sup-norm has been dened by

Eq.(21) for the Abelian sub--algebra Cyl(7) in A.

For simplicity, we denote the one element sequences (equivalence classes)

((, 0)) and

((0, Y))

Cyl(7), Y +

C

(7) in A by

and

Y respectively, where the hat denotes the equiv-

alence class with respect to the quotient. In particular, for all cylindrical functions

and ux

vector elds

Y

f

(S ),

and

Y

f

(S )

=

Y

f

(S ).

It can be seen that the free algebra F(P

ACZ

) is simplied a great deal after the quotient, and every

element of the quantum algebra A can be written as a nite linear combination of elements of

the form

1

Y

f

11

(S

11

),

2

Y

f

21

(S

21

)

Y

f

22

(S

22

),

...

k

Y

f

k1

(S

k1

)

Y

f

k2

(S

k2

) ...

Y

f

kk

(S

kk

),

...

Moreover, given a cylindrical function and a ux vector eld Y

f

(S ), one has the relation from

the commutation relation:

Y

f

(S )

= i

Y

f

(S ) +

Y

f

(S ). (22)

23

Then the kinematical Hilbert space 1

kin

can be obtained properly via the GNS-construction for

unital -algebra A in the same way as in Denition 3.1.3. By the GNS-construction, a positive

linear functional, i.e. a state

kin

, on A denes a cyclic representation (1

kin

,

kin

,

kin

) for A. In

our case of quantum holonomy-ux -algebra, the state with both Yang-Mills gauge invariance

and dieomorphism invariance is dened for any

eld Y

f

(S ) +

C

(7) as [96]:

kin

(

) :=

_

S U(2)

N

_

eE()

d

H

(A(e))

(A(e)

eE()

),

kin

( a

Y

f

(S )) := 0, a A,

where d

H

is the Haar measure on the compact group S U(2) and N

in E(). This

kin

is called Ashtekar-Isham-Lewandowski state. The null space N

kin

A with

respect to

kin

is dened as N

kin

:= a A

kin

( a

map can be dened as:

[.] : A A/N

kin

;

a [ a] := a +

b

b N

kin

.

The GNS-representation for A with respect to

kin

is a representation map:

kin

: A [(1

kin

)

such that

kin

( a

b) =

kin

( a)

kin

(

b), where 1

kin

:= (A/N

kin

) = (Cyl(7)) by straightforward

verication and the () denotes the completion with respect to the natural equipped inner product

on 1

kin

,

< [ a][

b] >

kin

:=

kin

( a

b).

To show how this inner product works, given any two cylindrical functions = [

],

=

[

< [

][

] >

kin

:=

_

X

(P

)(P

)d

, (23)

for any groupoid

. The measure d

on X

back of the product Haar measure d

N

H

on the product group S U(2)

N

bijection between X

and S U(2)

N

, where N

. In addition, a nice result shows that given such a family of measures

[

, a probability

measure is uniquely well-dened on the quantum conguration space 7 [17], such that the

kinematical Hilbert space 1

kin

coincides with the collection of the square-integrable functions

with respect to the measure on the quantum conguration space, i.e. 1

kin

= L

2

(7, d), just

as we expected at the beginning of our construction.

The representation map

kin

is dened by

kin

( a)[

b] := [ a

b], a A, and [

b] 1

kin

.

Note that

kin

( a) is an unbounded operator in general. It is easy to verify that

kin

(

Y

f

(S ))[

] = i[

Y

f

(S )]

24

via Eq.(22), which gives the canonical momentum operator. In the following context, we denote

the operator

kin

(

Y

f

(S )) by

P

f

(S ) on 1

kin

, and just denote the elements [

] in 1

kin

by for

simplicity.

Moreover, since

kin

:= 1 is a cyclic vector in 1

kin

, the positive linear functional which we

begin with can be expressed as

kin

( a) =<

kin

kin

( a)

kin

>

kin

.

Thus the Ashtekar-Isham-Lewandowski state

kin

on A is equivalent to a cyclic representation

(1

kin

,

kin

,

kin

) for A, which is the Ashtekar-Isham-Lewandowski representation for quantum

holonomy-ux -algebra of background independent gauge eld theory. One thus obtains the

kinematical representation of loop quantum gravity via the construction of algebraic quantum

eld theory. It is important to note that the Ashtekar-Isham-Lewandowski state is the unique

state on the quantum holonomy-ux -algebra A invariant under internal gauge transformations

and spatial dieomorphisms

7

, which are both automorphisms

g

and

ed that

kin

g

=

kin

and

kin

=

kin

. So these gauge transformations are represented

as unitary transformations on 1

kin

, while the cyclic vector

kin

, representing no geometry vac-

uum state, is the unique state in 1

kin

invariant under internal gauge transformations and spatial

dieomorphisms. This is a very crucial uniqueness theorem for canonical quantization of gauge

eld theory [96]:

Theorem 3.4.1: There exists exactly one Yang-Mills gauge invariant and spatial dieomor-

phism invariant state (positive linear functional) on the quantum holonomy-ux -algebra. In

other words, there exists a unique Yang-Mills gauge invariant and spatial dieomorphism in-

variant cyclic representation for the quantumholonomy-ux -algebra, which is called Ashtekar-

Isham-Lewandowski representation. Moreover, this representation is irreducible with respect to

an exponential version of the quantum holonomy-ux algebra (dened in [130]), which is anal-

ogous to the Weyl algebra.

Hence we have nished the construction of kinematical Hilbert space for background inde-

pendent gauge eld theory and represented the quantum holonomy-ux algebra on it. Then

following the general programme presented in the last subsection, we should impose the con-

straints as operators on the kinematical Hilbert space since we are dealing with a gauge system.

3.5 Spin-network Decomposition of Kinematical Hilbert Space

The kinematical Hilbert space 1

kin

for loop quantum gravity has been well dened. In this

subsection, it will be shown that 1

kin

can be decomposed into the orthogonal direct sum of 1-

dimensional subspaces and nd a basis, called spin-network basis, in the Hilbert space, which

consists of uncountably innite elements. So the kinematic Hilbert space is non-separable. In

the following, we will show the decomposition in three steps.

Spin-network Decomposition on a Single Edge

7

The proof of this conclusion depends on the compact support property of the smear functions f

i

(see [96] for

detail).

25

Given a graph consisting of only one edge e, which naturally associates with a group

S U(2) = X

(e)

, the elements of X

(e)

are the quantum connections only taking non-

trivial values on e. Then we consider the decomposition of the Hilbert space 1

(e)

=

L

2

(X

(e)

, d

(e)

) = L

2

(S U(2), d

H

), which is nothing but the space of square integrable

functions on the compact group S U(2) with the natural L

2

inner product. It is natural

to dene several operators on 1

(e)

. First, the so-called conguration operator

f (A(e))

whose operation on any in a dense domain of L

2

(S U(2), d

H

) is nothing but multipli-

cation by the function f (A(e)), i.e.,

f (A(e))(A(e)) := f (A(e))(A(e)),

where A(e) S U(2). Second, given any vector su(2), it generates left invariant vector

eld L

()

and right invariant vector eld R

()

on S U(2) by

L

()

(A(e)) :=

d

dt

t=0

(A(e) exp(t)),

R

()

(A(e)) :=

d

dt

t=0

( exp(t)A(e)),

for any function C

1

(S U(2)). Then one can dene the so-called momentum operators

on the single edge by

J

(L)

i

= iL

(

i

)

and

J

(R)

i

= iR

(

i

)

,

where the generators

i

su(2) constitute an orthonormal basis with respect to the

Killing-Cartan metric. The momentum operators have the well-known commutation re-

lation of the angular momentum operators in quantum mechanics:

[

J

(L)

i

,

J

(L)

j

] = i

k

i j

J

(L)

k

, [

J

(R)

i

,

J

(R)

j

] = i

k

i j

J

(R)

k

, [

J

(L)

i

,

J

(R)

j

] = 0.

Third, the Casimir operator on 1

e

can be expressed as

J

2

:=

i j

J

(L)

i

J

(L)

j

=

i j

J

(R)

i

J

(R)

j

. (24)

The decomposition of 1

e

= L

2

(S U(2), d

H

) is provided by the following Peter-Weyl

Theorem.

Theorem 3.5.1 [53]:

Given a compact group G, the function space L

2

(G, d

H

) can be decomposed as an or-

thogonal direct sum of nite dimensional Hilbert spaces, and the matrix elements of the

equivalence classes of nite dimensional irreducible representations of G form an orthog-

onal basis in L

2

(G, d

H

).

Note that a nite dimensional irreducible representation of G can be regarded as a matrix-

valued function on G, so the matrix elements are functions on G. Using this theorem, one

can nd the decomposition of the Hilbert space:

L

2

(S U(2), d

H

) =

j

[1

j

1

j

],

26

where j, labelling irreducible representations of S U(2), are the half integers, 1

j

denotes

the carrier space of the j-representation of dimension 2 j +1, and 1

j

is its dual space. The

basis e

j

m

e

j

n

in 1

j

1

j

maps a group element g S U(2) to a matrix

j

mn

(g), where

m, n = j, ..., j. Thus the space 1

j

1

j

is spanned by the matrix element functions

j

mn

of equivalent j-representations. Moreover, the spin-network basis can be dened.

Proposition 3.5.1 [56]

The systemof spin-network functions on 1

(e)

, consisting of matrix elements

j

mn

in nite

dimensional irreducible representations labelled by half-integers j, satises

J

2

j

mn

= j( j + 1)

j

mn

,

J

(L)

3

j

mn

= m

j

mn

,

J

(R)

3

j

mn

= n

j

mn

,

where j is called angular momentum quantum number and m, n = j, ..., j magnetic

quantum number. The normalized functions

_

2 j + 1

j

mn

form an orthonormal basis in

1

(e)

by the above theorem and

_

7

e

j

mn

d

e

=

1

2 j + 1

n

,

which is called the spin-network basis on 1

(e)

. So the Hilbert space on a single edge has

been decomposed into one dimensional subspaces.

Note that the system of operators

J

2

,

J

(R)

3

,

J

(L)

3

forms a complete set of commutable op-

erators in 1

(e)

. There is a cyclic vacuum state in the Hilbert space, which is the

( j = 0)-representation

(e)

=

j=0

= 1, representing that there is no geometry on the

edge.

Spin-network Decomposition on Finite Graph

Given a groupoid generated by a graph with N oriented edges e

i

and M vertices,

one can dene the conguration operators on the corresponding Hilbert space 1

=

L

2

(X

, d

) = L

2

(S U(2)

N

, d

N

H

) by

f (A(e

i

))(A(e

1

), ..., A(e

N

)) := f (A(e

i

))(A(e

1

), ..., A(e

N

)).

The momentum operators

J

i

(e,v)

associated with a edge e connecting a vertex v are dened

as

J

i

(e,v)

:= (1 ...

J

i

... 1),

where we set

J

i

=

J

(L)

i

if v = s(e) and

J

i

=

J

(R)

i

if v = t(e), so

J

i

(e,v)

= iX

(e,v)

i

. Note that

J

i

(e,v)

only acts nontrivially on the Hilbert space associated with the edge e. Then one

can dene a vertex operator associated with vertex v in analogy with the total angular

momentum operator via

[

J

v

]

2

:=

i j

J

v

i

J

v

j

,

27

where

J

v

i

:=

e at v

J

(e,v)

i

.

Obviously, 1

1

=

e

1

(e)

=

e

[

j

(1

e

j

1

e

j

)] =

j

[

e

(1

e

j

1

e

j

)]

=

j

[

v

(1

v=s(e)

j(s)

1

v=t(e)

j(t)

)],

where j := ( j

1

, ..., j

N

) assigns to each edge an irreducible representation of S U(2), in

the fourth step the Hilbert spaces associated with the edges are allocated to the vertexes

where these edges meet so that for each vertex v,

1

v=s(e)

j(s)

s(e)=v

1

e

j

and 1

v=t(e)

j(t)

t(e)=v

1

e

j

.

The group of gauge transformations g(v) S U(2) at each vertex is reducibly represented

on the Hilbert space 1

v=s(e)

j(s)

1

v=t(e)

j(t)

in a natural way. So this Hilbert space can be

decomposed as a direct sum of irreducible representation spaces via Clebsch-Gordon

decomposition:

1

v=s(e)

j(s)

1

v=t(e)

j(t)

=

l

1

v

j(v),l

.

As a result, 1

1

=

j

[

v

(

l

1

v

j(v),l

)] =

j

[

l

(

v

1

v

j(v),l

)]

j

[

l

1

,j,l

]. (25)

It can also be viewed as the eigenvector space decomposition of the commuting operators

[

J

v

]

2

(with eigenvalues l(l + 1)) and [

J

e

]

2

i j

J

e

i

J

e

j

. Note that l := (l

1

, ..., l

M

) assigns

to each vertex(objective) of an irreducible representation of S U(2). One may also

enlarge the set of commuting operators to further rene the decomposition of the Hilbert

space. Note that the subspace of 1

representation of gauge transformations is trivial.

Spin-network Decomposition of 1

kin

Since 1

kin

has the structure 1

kin

= (

[

1

direct sum of 1

cisely described as a theorem below.

Theorem 3.5.2:

Consider assignments j = ( j

1

, ..., j

N

) to the edges of any groupoid [ and assign-

ments l = (l

1

, ..., l

M

) to the vertices. The edge representation j is non-trivial on each

edge, and the vertex representation l is non-trivial at each spurious

8

vertex, unless it is

the base point of a close analytic loop. Let 1

8

A vertex v is spurious if it is bivalent and e e

meeting at v.

28

subspaces 1

,j,l

(assigned the above conditions) according to Eq.(25). Then 1

kin

can be

decomposed as the direct sum of the Hilbert spaces 1

, i.e.,

1

kin

=

[

1

C.

Proof:

Since the representation on each edge is non-trivial, by denition of the inner product, it

is easy to see that 1

and 1

has

at leat an edge e more than the other due to

_

7

e

j

mn

d

e

=

_

7

e

1

j

mn

d

e

= 0

for any j 0. Now consider the case of the spurious vertex. An edge e with j-

representation in a graph is assigned the Hilbert space 1

e

j

1

e

j

. Inserting a vertex v

into the edge, one obtains two edges e

1

and e

2

split by v both with j-representations,

which belong to a dierent graph. By the decomposition of the corresponding Hilbert

space,

1

e

1

j

1

e

1

j

1

e

2

j

1

e

2

j

= 1

e

1

j

(

l=0...2 j

1

v

l

) 1

e

2

j

,

the subspace for all l 0 are orthogonal to the space 1

e

j

1

e

j

, while the subspace for

l = 0 coincides with 1

e

j

1

e

j

since 1

v

l=0

= C and A(e) = A(e

1

)A(e

2

). This completes

the proof.

Since there are uncountably many graphs on , the kinematical Hilbert 1

kin

is non-

separable. We denote the spin-network basis in 1

kin

by

s

, s = ((s), j

s

, m

s

, n

s

) and

vacuum

kin

0

= 1, where

s

:=

_

eE((s))

_

2 j

e

+ 1

j

e

m

e

n

e

( j

e

0),

which form a orthonormal basis with the relation <

s

s

>

kin

=

ss

. And Cyl

(7)

Cyl(7) denotes the linear span of the spin network functions

s

for (s) = .

The spin-network basis can be used to construct the so-called spin network representation

of loop quantum gravity.

Denition 3.5.1: The spin-network representation is a vector space

1 of complex valued func-

tions

: S C; s

(s),

where S is the set of the labels s for the spin network states.

1 is equipped with the scalar

product

<

,

>:=

sS

(s)

(s)

29

between square summable functions.

The relation between the Hilbert spaces

1 and 1

kin

is claried by the following proposition

[146].

Proposition 3.5.2:

The spin-network transformation

T : 1

kin

1;

(s) :=<

s

, >

kin

is a unitary transformation with inverse

T

1

=

sS

(s)

s

.

Thus the connection representation and the spin-network representation are Fourier trans-

forms of each other, where the role of the kernel of the integral is played by the spin-network

basis. Note that, in the gauge invariant Hilbert space of loop quantum gravity which we will

dene later, the Fourier transform with respect to the gauge invariant spin-network basis is the

so-called loop transform, which leads to the unitary equivalent loop representation of the theory

[118][73][122].

To conclude this subsection, we show the explicit representation of elementary observables

on the kinematical Hilbert space 1

kin

. The action of canonical momentum operator

P

f

(S ) on

dierentiable cylindrical functions

Cyl

P

f

(S )

(A(e)

eE()

) =

vV()S

f

i

(v)[

e at v

(S, e)

J

(e,v)

i

]

(A(e)

eE()

)

=

vV()S

f

i

(v)[

J

(S,v)

i(u)

J

(S,v)

i(d)

]

(A(e)

eE()

), (26)

where

J

(S,v)

i(u)

J

(e

1

,v)

i

+ ... +

J

(e

u

,v)

i

,

J

(S,v)

i(d)

J

(e

u+1

,v)

i

+ ... +

J

(e

u+d

,v)

i

, (27)

for the edges e

1

, ..., e

u

lying above S and e

u+1

, ..., e

u+d

lying below S . And it was proved that

the operator

P

f

(S ) is essentially self-adjoint on 1

kin

[146]. On the other hand, it is obvious to

construct conguration operators by spin-network functions:

(A(e)

eE()

) :=

s

(A(e)

eE((s))

)

(A(e)

eE()

).

Note that

s

may change the graph, i.e.,

s

: Cyl

(7) Cyl

(s)

(7). So far, the elementary

operators of quantum kinematics have been well dened on 1

kin

.

30

3.6 Quantum Riemannian Geometry

The well-established quantum kinematics of loop quantum gravity is now in the same status

as Riemannian geometry before the appearance of general relativity and Einsteins equation,

giving general relativity mathematical foundation and oering living place to the Einstein equa-

tion. Instead of classical geometric quantities, such as length, area, volume etc., the quantities

in quantum geometry are operators on the kinematical Hilbert space 1

kin

, and their spectrum

serve as the possible values of the quantities in measurements. So far, the kinematical quan-

tum geometric operators constructed properly in loop quantum gravity include length operator

[141], area operator [125][21], two dierent volume operators [18][125][22],

Q operator [100],

etc.. Recently, a consistency check was proposed for the dierent regularizations of the volume

operator [77][78]. We thus will only introduce the volume operator dened by Ashtekar and

Lewandowski [22], which is shown to be correct in the consistency check.

First, we dene the area operator with respect to a 2-surface S by the elementary operators.

Given a closed 2-surface or a surface S with boundary, we can divide it into a large number N

of small area cells S

I

. Taking account of the classical expression of an area, we set the area of

the 2-surface to be the limit of the Riemannian sum

A

S

:= lim

N

[A

S

]

N

= lim

N

I=1

_

P

i

(S

I

)P

j

(S

I

)

i j

.

Then one can unambiguously obtain a quantum area operator from the canonical momentum

operators

P

i

(S ) smeared by constant functions. Given a cylindrical function

Cyl

(7)

which has second order derivatives, the action of the area operator on

by requiring that each area cell contains at most only one intersecting point v of the graph and

S as

A

S

:= lim

N

[

A

S

]

N

= lim

N

I=1

_

P

i

(S

I

)

P

j

(S

I

)

i j

.

The regulator N is easy to remove, since the result of the operation of the operator

P

i

(S

I

) does

not change when S

I

shrinks to a point. Since the renement of the partition does not aect the

result of action of [

A

S

]

N

on

A

S

, which is shown to be self-adjoint

[21], is well dened on 1

kin

and takes the explicit expression as:

A

S

= 4

2

p

vV(S )

_

(

J

(S,v)

i(u)

J

(S,v)

i(d)

)(

J

(S,v)

j(u)

J

(S,v)

j(d)

)

i j

,

where

J

(S,v)

i(u)

and

J

(S,v)

i(d)

have been dened in Eq.(27). It turns out that the nite linear combinations

of spin-network basis in 1

kin

diagonalizes

A

S

with eigenvalues given by nite sums,

a

S

= 4

2

p

v

_

2j

(u)

v

( j

(u)

v

+ 1) + 2 j

(d)

v

( j

(d)

v

+ 1) j

(u+d)

v

( j

(u+d)

v

+ 1), (28)

where j

(u)

, j

(d)

and j

(u+d)

are arbitrary half-integers subject to the standard condition

j

(u+d)

j

(u)

j

(d)

, j

(u)

j

(d)

+ 1, ..., j

(u)

+ j

(d)

. (29)

31

Hence the spectrum of the area operator is fundamentally pure discrete, while its continuum

approximation becomes excellent exponentially rapidly for large eigenvalues. However, in fun-

damental level, the area is discrete and so is the quantum geometry. One can see that the

eigenvalue of

A

S

does not vanish even in the case where only one edge intersects the surface at

a single point, whence the quantum geometry is distributional.

The form of Ashtekar and Lewandowskis volume operator was introduced for the rst time

in [18], and its detailed properties are discussed in [22]. Given a region R with a xed coordinate

system x

a

a=1,2,3

in it, one can introduce a partition of R in the following way. Divide R into

small volume cells C such that, each cell C is a cube with coordinate volume less than and

two dierent cells only share the points on their boundaries. In each cell C, we introduce three

2-surfaces s = (S

1

, S

2

, S

3

) such that x

a

is constant on the surface S

a

. We denote this partition

(C, s) as /

V

s

R

= lim

0

C

_

q

C,s

,

where

q

C,s

=

()

3

3!

i jk

abc

P

i

(S

a

)P

j

(S

b

)P

k

(S

c

).

This motivates us to dene the volume operator by naively changing P

i

(S

a

) to

P

i

(S

a

):

V

s

R

= lim

0

C

_

q

C,s

,

q

C,s

=

()

3

3!

i jk

abc

P

i

(S

a

)

P

j

(S

b

)

P

k

(S

c

).

Note that, given any cylindrical function

Cyl

to be at the intersecting points of the triples of 2-surfaces s = (S

1

, S

2

, S

3

) in corresponding

cells. Thus the limit operator will trivially exist due to the same reason in the case of the area

operator. However, the volume operator dened here depends on the choice of orientations for

the triples of surfaces s = (S

1

, S

2

, S

3

), or essentially, the choice of coordinate systems. So it

is not uniquely dened. Since, for all choice of s = (S

1

, S

2

, S

3

), the resulting operators have

correct semi-classical limit, one settles up the problem by averaging dierent operators labelled

by dierent s [22]. The process of averaging removes the freedom in dening the volume

operator up to an overall constant

0

. The resulting self-adjoint operator acts on any cylindrical

function

Cyl

(7) as

V

R

=

0

vV()

_

q

v,

,

where

q

v,

= (8

2

p

)

3

1

48

e,e

,e

at v

i jk

(e, e

, e

)

J

(e,v)

i

J

(e

,v)

j

J

(e

,v)

k

,

32

here (e, e

, e

) sgn(

abc

e

a

e

b

e

c

)

v

with e

a

as the tangent vector of edge e and

abc

as the

orientation of . The only unsatisfactory point in the present volume operator is the choice

ambiguity of

0

. However, fortunately, the most recent discussion shows that the overall un-

determined constant

0

can be xed to be

6 by the consistency check between the volume

operator and the triad operator [77][78].

33

4 Implementation of Quantum Constraints

After constructing the kinematical Hilbert space 1

kin

of loop quantum gravity, one should im-

plement the constraints on it to obtain the physical Hilbert space which encodes the complete

information of quantum dynamics of general relativity, since the Hamiltonian of general rela-

tivity is a linear combination of the constraints. Recalling the constraints (7) in the Hamiltonian

formalism and the Poission algebra (9) among them, the subalgebra generated by the Gauss

constraints g() forms a Lie algebra and a 2-sided ideal in the constraints algebra. So in this

section, we rst solve the Gaussian constraints independently of the other two kinds of con-

straints and nd the solution space 1

G

, which is constituted by internal gauge invariant quan-

tum states. Then, although the subalgebra generated by the dieomorphism constraints is not

an ideal in the constraint algebra, we still would like to solve them independently of the scalar

constraints for technical convenience. After that, the quantum operator corresponding to the

Hamiltonian constraint(scalar constraint) is dened on the kinematical Hilbert space, and we

will also discuss an alterative for the implementation of the scalar constraint, which is called

the master constraint programme by modifying the classical constraint algebra.

4.1 Solutions of Quantum Gaussian Constraint

Recall that the classical expression of Gauss constraints reads

g() =

_

d

3

x

i

D

a

P

a

i

=

_

d

3

x

P

a

i

D

a

i

P(D),

where D

a

i

=

a

i

+

i

jk

A

j

a

k

. As the situation of triad ux, the Gauss constraints can be

dened as cylindrically consistent vector elds Y

D

on 7, which act on any cylindrical function

f

Cyl

(7) by

Y

D

f

(A(e)

eE()

) := P(D), f

(A(e)

eE()

).

Then the Gauss constraint operator can be dened in analogy with the momentum operator,

which acts on f

as:

g() f

(A(e)

eE()

) := iY

D

f

(A(e)

eE()

)

=

vV()

[

i

(v)

J

v

i

] f (A(e)

eE()

),

which is the generator of internal gauge transformations on Cyl

tor is easily obtained in terms of the spin-network decomposition, which is the internal gauge

invariant Hilbert space:

1

G

=

,j

1

,j,l=0

C.

One then naturally gets the gauge invariant spin-network basis T

s

, s = ((s), j

s

, i

s

) in 1

G

via a

group averaging technique at each vertex[126][25][32](we will call T

s

spin-network state in the

following context):

T

s=(,j,i)

=

vV()

i

v

eE()

j

e

(A(e)), ( j

e

0)

34

assigning a non-trivial spin representation j on each edge and a invariant tensor i (intertwiner) on

each vertex. We denote the vector space of nite linear combinations of vacuum state and gauge

invariant spin-network states Cyl(7/g), which is dense in 1

G

. And Cyl

(7/g) Cyl(7/g)

denotes the linear span of the gauge invariant spin network functions T

s

for (s) = . All

Yang-Mills gauge invariant operators are well dened on 1

G

. However, the condition of acting

on gauge invariant states often changes the structure of the spectrum of quantum geometric

operators. For the area operator, the spectrumdepends on certain global properties of the surface

S (see [20][21] for details). For the volume operators, a non-zero spectrum arises from at least

4-valent vertices.

4.2 Solutions of Quantum Dieomorphism Constraint

Unlike the strategy in solving Gaussian constraint, one cannot dene an operator for the quan-

tum dieomorphism constraint as the innitesimal generator of nite dieomorphism trans-

formations (unitary operators since the measure is dieomorphism invariant) represented on

1

kin

. The representation of nite dieomorphisms is a family of unitary operators

U

acting on

cylindrical functions

by

:=

, (30)

for any spatial dieomorphism on . An 1-parameter subgroup

t

in the group of spatial

dieomorphisms is then represented as an 1-parameter unitary group

U

t

on 1

kin

. However,

t

is not weakly continuous, since the subspaces 1

()

and 1

(

t

)

are orthogonal to each

other no matter how small the parameter t is. So one always has

<

>

kin

<

>

kin

=<

>

kin

0, (31)

even in the limit when t goes to zero. Therefore, the innitesimal generator of

U

t

does not

exist. In the strategy to solve the dieomorphism constraint, due to the Lie algebra structure of

dieomorphism constraints subalgebra, the so-called group averaging technique is employed.

We now outline the procedure. First, given a colored graph (a graph and a cylindrical function

living on it), one can dene the group of graph symmetries GS

by

GS

:= Di f f

/TDi f f

,

where Di f f

is the

group of dieomorphisms which trivially acts on . We dene a projection map by averaging

with respect to GS

of GS

P

Di f f ,

:=

1

n

GS

,

for all cylindrical functions

()

, where n

.

Second, we average with respect to all remaining dieomorphisms which move the graph .

For each cylindrical function

Cyl

) associated to it in the

35

algebraic dual space Cyl

Cyl

(7/g)

as:

(

)[

] :=

Di f f ()/Di f f

<

U

P

Di f f ,

>

kin

.

It is well dened since, for any given graph

It is easy to verify that (

(

)[

U

] = (

)[

].

Thus we have dened a rigging map : Cyl(7/g) Cyl

Di f f

, which maps every cylindrical

function to a dieomorphism invariant one, where Cyl

Di f f

is spanned by vacuum state T

0

= 1

and rigged spin-network functions T

[s]

(T

s

), [s] = ([], j, i) associated with dieomor-

phism classes [] of graphs . Moreover a Hermitian inner product can be dened on Cyl

Di f f

by the natural action of the algebraic functional:

< (

)(

) >

Di f f

:= (

)[

].

The dieomorphism invariant Hilbert space 1

Di f f

is dened by the completion of Cyl

Di f f

with

respect to the above inner product < >

Di f f

. The dieomorphism invariant spin-network func-

tions T

[s]

form an orthonormal basis in 1

Di f f

. Finally, we have obtained the general solutions

invariant under both Yang-Mills gauge transformations and spatial dieomorphisms.

In general relativity, the problem of observables is a subtle issue due to the dieomorphism

invariance [116][119][120]. Now we discuss the operators as dieomorphism invariant ob-

servables on 1

Di f f

. We call an operator

O [(1

kin

) a strong observable if and only if

U

1

O

U

=

O, Di f f (). We call it a weak observable if and only if

O leaves 1

Di f f

invariant. Then it is easy to see that a strong observable

O must be a weak one. One notices that

a strong observable

O can rst be dened on 1

Di f f

by its dual operator

O

as

(

O

Di f f

)[] :=

Di f f

[

O],

then one gets

(

O

Di f f

)[

U

] =

Di f f

[

O

U

] =

Di f f

[

U

1

O

U

] = (

O

Di f f

)[],

for any

Di f f

1

Di f f

and 1

kin

. Hence

O

Di f f

is also dieomorphism invariant. In

addition, a strong observable also has the property of

O

) = (

O

) since,

1

kin

,

<

O

)(

) >

Di f f

= (

O

))[

] = (

)[

O

]

=

Di f f ()/Di f f

<

U

P

Di f f ,

>

kin

=

1

n

Di f f ()/Di f f

GS

<

U

>

kin

=

1

n

Di f f ()/Di f f

GS

<

U

>

kin

= < (

O

)(

) >

Di f f

.

36

Note that the Hilbert space 1

Di f f

is still non-separable if one considers the C

n

dieomorphisms

with n > 0. However, if one extends the dieomorphisms to be semi-analytic dieomotphisms,

i.e. homomorphisms that are analytic dieomorphisms up to nite isolated points (which can

be viewed as an extension of the classical concept to the quantum case), the Hilbert space 1

Di f f

would be separable [69][20].

4.3 Hamiltonian Constraint Operator

In the following, we consider the issue of scalar constraint in loop quantum gravity. One may

rst construct a Hamiltonian constraint (scalar constraint) operator in 1

kin

or 1

Di f f

, then at-

tempt to nd the physical Hilbert space 1

phys

by solving the quantum Hamiltonian constraint.

However, diculties arise here due to the special role played by the scalar constraints in the

constraint algebra (9). First, the scalar constraints do not form a Lie subalgebra. Hence the

strategy of group averaging cannot be used directly on 1

kin

for them. Second, modulo the

Gaussian constraint, there is still a structure function in the Poisson bracket between two scalar

constraints:

S(N), S(M) = +((N

b

M M

b

N)q

ab

), (32)

which raises the danger of quantum anomalyies in quantization. Moreover, the dieomorphism

constraints do not form an ideal in the quotient constraint algebra modulo the Gaussian con-

straints. This fact results in that the scalar constraint operator cannot be well dened on 1

Di f f

,

as it does not commute with the dieomorphism transformations

U

struction of 1

Di f f

does not appear very useful for the nal construction of 1

phys

, which is our

nal goal. However, one may still rst try to construct a Hamiltonian constraint operator in 1

kin

for technical convenience.

We recall the classical expression of Hamiltonian constraint:

S(N) :=

2

2

_

d

3

xN

P

a

i

P

b

j

_

det q

[

i j

k

F

k

ab

2(1 +

2

)K

i

[a

K

j

b]

]

= S

E

(N) 2(1 +

2

)7(N). (33)

The main idea of the construction is to rst express S(N) in terms of the combination of Poisson

brackets between the variables which have been represented as operators on 1

kin

, then replace

the Poisson brackets by canonical commutators between the operators. We will use the volume

functional for a region R and the extrinsic curvature functional dened by:

K :=

_

d

3

x

P

a

i

K

i

a

.

A key trick here is to consider the following classical identity of the co-triad e

i

a

(x) [134]:

e

i

a

(x) =

()

2

2

abc

i jk

P

b

j

P

c

k

_

det q

(x) =

2

A

i

a

(x), V

R

,

37

where V

R

is the volume functional for a neighborhood R containing x. And the expression of

the extrinsic curvature 1-form K

i

a

(x):

K

i

a

(x) =

1

A

i

a

(x), K.

Note that K can be expressed by a Poisson bracket between the constant-smeared Euclidean

Hamiltonian constraint and the total volume of the space :

K =

2

S

E

(1), V

. (34)

Thus one can obtain the equivalent classical expressions of S

E

(N) and 7(N) as:

S

E

(N) =

2

2

_

d

3

xN

P

a

i

P

b

j

_

det q

i j

k

F

k

ab

=

2

d

3

xN(x)

abc

Tr(F

ab

(x)A

c

(x), V

R

x

),

7(N) =

2

2

_

d

3

xN

P

a

i

P

b

j

_

det q

K

i

[a

K

j

b]

=

2

3

_

d

3

xN(x)

abc

Tr(A

a

(x), KA

b

(x), KA

c

(x), V

R

x

),

where A

a

= A

i

a

i

, F

ab

= F

i

ab

i

, Tr represents the trace of the Lie algebra matrix, and R

x

denotes an arbitrary neighborhood of x . In order to quantize the Hamiltonian constraint

as a well-dened operator on 1

kin

, one has to express the classical formula of S(N) in terms

of holonomies A(e) and other variables with clear quantum analogs. As a rst attempt [134],

this can be realized by introducing a triangulation T(), where the parameter describes how

ne the triangulation is, and the triangulation will ll out the spatial manifold when

0. Given a tetrahedron T(), we use s

i

()

i=1,2,3

to denote the three outgoing oriented

segments in with a common beginning point v() = s(s

i

()), and use a

i j

() to denote the

arc connecting the end points of s

i

() and s

j

(). Then several loops

i j

() are formed by

i j

() := s

i

() a

i j

() s

j

()

1

. Thus we have the identities:

_

s

i

()

A

a

s

a

i

(), V

R

v()

= A(s

i

())

1

A(s

i

()), V

R

v()

+ o(),

_

s

i

()

A

a

s

a

i

(), K = A(s

i

())

1

A(s

i

()), K + o(),

_

P

i j

F

ab

(x) =

1

2

A(

i j

())

1

1

2

A(

i j

()) + o(

2

),

where P

i j

is the plane with boundary

i j

. Note that the above identities are constructed by taking

account of internal gauge invariance of the nal formula of Hamiltonian constraint operator. So

we have the regularized expression of S(N) by the Riemannian sum [134]:

S

E

(N) =

2

3

2

T()

N(v())

i jk

38

Tr(A(

i j

())

1

A(s

k

())

1

A(s

k

()), V

R

v()

),

7

(N) =

2

6

4

T()

N(v())

i jk

Tr(A(s

i

())

1

A(s

i

()), KA(s

j

())

1

A(s

j

()), K

A(s

k

())

1

A(s

k

()), V

R

v()

),

S

(N) = S

E

(N) 2(1 +

2

)7

(N), (35)

such that lim

0

S

(N) = S(N). It is clear that the above regulated formula of S(N) is invari-

ant under internal gauge transformations. Since all constituents in the expression have clear

quantum analogs, one can quantize the regulated Hamiltonian constraint as an operator on 1

kin

(or 1

G

) by replacing them by the corresponding operators and Poisson brackets by canonical

commutators, i.e.,

A(e)

A(e), V

R

V

R

, ,

[ , ]

i

,

and K

K

2

i

[

S

E

(1),

V

].

Removing the regulator by 0, it turns out that one can obtain a well-dened limit operator

on 1

kin

(or 1

G

) with respect to a natural operator topology.

Now we begin to construct the Hamiltonian constraint operator in analogy with the classical

expression (57). All we should do is dene the corresponding regulated operators on dierent

1

separately, then remove the regulator so that the limit operator is dened on 1

kin

(or 1

G

)

cylindrically consistently. In the following, given a vertex and three edges intersecting at the

vertex in a graph of

Cyl

of the vertex adapted to the three edges. Then we average with respect to the triples of edges

meeting at the given vertex. Precisely speaking, one can make the triangulations T() with the

following properties [134][146].

The chosen triple of edges in the graph is embedded in a T() for all , so that the vertex

v of where the three edges meet coincides with a vertex v() in T().

For every triple of segments (e

1

, e

2

, e

3

) of such that v = s(e

1

) = s(e

2

) = s(e

3

), there is a

tetrahedra T() such that v = v() = s(s

i

()), and s

i

() e

i

, i = 1, 2, 3. We denote

such a tetrahedra as

0

e

1

,e

2

,e

3

.

For each tetrahedra

0

e

1

,e

2

,e

3

one can construct seven additional tetrahedron

e

1

,e

2

,e

3

, =

1, ..., 7, by backward analytic extensions of s

i

() so that U

e

1

,e

2

,e

3

:=

7

=0

e

1

,e

2

,e

3

is a neigh-

borhood of v.

The triangulation must be ne enough so that the neighborhoods U(v) :=

e

1

,e

2

,e

3

U

e

1

,e

2

,e

3

(v)

are disjoint for dierent vertices v and v

of the

graph , there exists a triangulation T() such that

vV()

U(v) U

.

The distance between a vertex v() and the corresponding arcs a

i j

() is described by the

parameter . For any two dierent and

, the arcs a

i j

(

) and a

i j

(

) with respect to

one vertex v() are semi-analytically dieomorphic with each other.

39

With the triangulations T(), the integral over is replaced by the Riemanian sum:

_

=

_

U

+

_

U

,

_

U

vV()

_

U(v)

+

_

U

v

U(v)

,

_

U(v)

=

1

E(v)

e

1

,e

2

,e

3

[

_

U

e

1

,e

2

,e

3

(v)

+

_

U(v)U

e

1

,e

2

,e

3

,(v)

],

where n(v) is the valence of the vertex v = s(e

1

) = s(e

2

) = s(e

3

), and E(v)

_

n(v)

3

_

de-

notes the binomial coecient which comes from the averaging with respect to the triples

of edges meeting at given vertex v. One then observes that

_

U

e

1

,e

2

,e

3

(v)

= 8

_

0

e

1

,e

2

,e

3

(v)

in the limit 0.

The triangulations for the regions

U(v) U

e

1

,e

2

,e

3

(v),

U

vV()

U(v),

U

, (36)

are arbitrary. These regions do not contribute to the construction of the operator, since the

commutator term [A(s

i

()), V

R

v()

]

Thus we nd the regulated expression of Hamiltonian constraint operator with respect to the

triangulations T() as [134]

E,

(N) =

16

3i

2

vV()

N(v)

E(v)

v()=v

i jk

Tr(

A(

i j

())

1

A(s

k

())

1

[

A(s

k

()),

V

U

v

]),

(N) =

4

2

3i

3

vV()

N(v)

E(v)

v()=v

i jk

Tr(

A(s

i

())

1

[

A(s

i

()),

K

]

A(s

j

())

1

[

A(s

j

()),

K

A(s

k

())

1

[

A(s

k

()),

V

U

v

]),

(N)

= [

S

E,

(N) 2(1 +

2

)

7

(N)]

vV()

N(v)

S

,

for any cylindrical function

Cyl

T

s

with (s) = .

40

By construction, the operation of

S

(N) on any

Cyl

combination of that of

S

v

with respect to dierent vertices of . Hence, for each > 0,

S

(N)

is a well-dened internal gauge invariant and dieomorphism covariant operator on Cyl(7/g).

The last step is to remove the regulator by taking the limit 0. However, the action of

the Hamiltonian constraint operator on

adds arcs a

i j

() with a

1

2

-representation with respect

to each v() of

9

, i.e. the action of

S

the operator does not converge with respect to the weak operator topology in 1

kin

when 0,

since dierent 1

()

with dierent graphs are mutually orthogonal. Thus one has to dene a

weaker operator topology to make the operator limit meaningful. By physical motivation and

the naturally available Hilbert space 1

Di f f

, the convergence of

S

so-called Uniform Rovelli-Smolin Topology [124], where one denes

S

only if

Di f f

[

S

Di f f

Cyl

Di f f

and Cyl(7/g). Since the value

of

Di f f

[

S

sequence converges to a nontrivial result

Di f f

[

S

0

(N)] with arbitrary xed

0

> 0. Thus we

have dened a dieomorphism covariant, densely dened, closed but non-symmetric operator,

S(N) = lim

0

S

(N) =

S

0

(N), on 1

kin

(or 1

G

) representing the Hamiltonian constraint.

Moreover, a dual Hamiltonian constraint operator

S

depending on

a specic value of

(

S

(N))[] := [

S

(N)],

for all Cyl

Di f f

Cyl

Di f f

Cyl

, one gets

(

S

(N)

Di f f

)[] =

Di f f

[

S

(N)].

which is independent of the value of .

Several remarks on the Hamiltonian constraint operator are listed in the following.

Finiteness of

S(N) on 1

kin

In ordinary quantum eld theory, the continuous quantum eld is only recovered when

one lets lattice spacing to approach zero, i.e., takes the continuous cut-o parameter to its

continuous limit. However, this will produce the well-known innities in quantum eld

theory and make the Hamiltonian operator ill-dened on the Fock space. So it seems

surprising that our operator

S(N) is still well dened, when one takes the limit 0

with respect to the Uniform Rovelli-Smolin Topology so that the triangulation goes to the

continuum. The reason behind it is that the cut-o parameter is essentially noneective

due to the dieomorphism invariance of our quantum eld theory. This is why there

is no UV divergence in the background independent quantum gauge eld theory with

dieomorphism invariance. On the other hand, from a convenient viewpoint, one may

think the Hamiltonian constraint operator as an operator dually dened on a dense domain

in 1

Di f f

. However, we will see that the dual Hamiltonian constraint operator cannot leave

1

Di f f

invariant.

9

The Hamiltonian constraint operator depends indeed on the choice of the representation j on the arcs a

i j

(),

which is known as one of the regularization ambiguities in the construction of quantum dynamics. For the simplic-

ity of the theory, one often choose the lowest label of representation j =

1

2

.

41

Implementation of Dual Quantum Constraint Algebra

One important task is to check whether the commutator algebra (quantum constraint al-

gebra) among the corresponding quantum operators of constraints both physically and

mathematically coincides with the classical constraint algebra by substituting quantum

constraint operators to classical constraint functionals and commutators to Poisson brack-

ets. Here the quantum anomaly has to be avoided in the construction of constraint opera-

tors (see the discussion for Eq.(13)). First, the subalgebra of the quantum dieomorphism

constraint algebra is free of anomaly by construction:

U

1

U

1

=

U

1 ,

which coincides with the exponentiated version of the Poisson bracket between two dif-

feomorphism constraints generating the transformations ,

Di f f ().

Second, the quantum constraint algebra between the dual Hamiltonian constraint opera-

tor S

U

on dieomorphism-invariant

states coincides with the classical Poisson algebra between +(

cylindrical function

the graph , the triangulations T( ) T() are compatible with the graph .

Then we have by denition:

( ([

S(N),

U

])

Di f f

)[

]

= ([

S

(N),

U

]

Di f f

)[

]

=

Di f f

[

S

(N)

(N)

]

=

vV()

N(v)

Di f f

[

S

] N( v)

Di f f

[

S

]

=

vV()

[N(v) N( v)]

Di f f

[

S

]

= (

S

(N

N)

Di f f

)[

]. (37)

Thus there is no anomaly. However, Eq.(37) also explains why the Hamiltonian constraint

operator

S(N) cannot leave 1

Di f f

invariant.

Third, we compute the commutator between two Hamiltonian constraint operators. No-

tice that

[

S(N),

S(M)]

vV()

[M(v)

S(N) N(v)

S(M)]

S

vV()

V(

)

[M(v)N(v

) N(v)M(v

)]

S

,

where

S(N) or

S(M), which adds the arcs

a

i j

() on , T() is the triangulation adapted to and T(

) adapted to

42

added vertices by

S

v

is planar, they will never contributes the nal result. So one has

[

S(N),

S(M)]

v,v

V(),vv

[M(v)N(v

) N(v)M(v

)]

S

=

1

2

v,v

V(),vv

[M(v)N(v

) N(v)M(v

)][

S

v

S

v

S

v

]

=

1

2

v,v

V(),vv

[M(v)N(v

) N(v)M(v

)][(

U

,v

U

v,v

)

S

v

]

,

(38)

where we have used the facts that [

S

v

,

S

] = 0 for v v

v,v

such that

S

v

=

U

,v

v

. Obviously, we have in the Uniform Rovelli-Smolin

Topology

([

S(N),

S(M)])

Di f f

= 0

for all

Di f f

Cyl

Di f f

. As we have seen in classical expression Eq.(32), the Poisson

bracket of any two Hamiltonian constraints is given by a generator of the dieomrophism

transformations. Therefore it is mathematically consistent with the classical expression

that two Hamiltonian constraint operators commute on dieomorphism invariant states,

as it is presented above. However, as it has been discussed in [72][95], the domain of

dual Hamiltonian constraint operator can be extended to a slightly larger space (habitat)

in Cyl

, whose elements are not necessary dieomorphism invariant. And it turns out that

the commutator between two Hamiltonian constraint operators continues to vanish on the

habitat, which seems to be problematic. Fortunately, the quantum operator corresponding

to the right hand side of classical Poisson bracket (32) also annihilates every state in the

habitat [72], so the quantum constraint algebra is consistent at this level. But it is not

clear that whether the quantum constraint algebra, especially the commutator between

two Hamiltonian constraint is consistent with the classical one (32) on some larger space

in Cyl

10

. On the other hand, more works

on the semi-classical analysis are also needed to test the classical limit of Eq.(38) and

commutation relation (32). The way to do it is looking for some proper semi-classical

states for calculating the classical limit of the operators. But due to the graph-changing

property of the Hamiltonian constraint operator, the semi-classical analysis for the Hamil-

tonian constraint operator and the quantum constraint algebra is still an open issue so far.

General Regularization Scheme of the Hamiltonian Constraint

In [20], a general scheme of regulation is introduced for the quantization of the Hamilto-

nian constraint, and includes Thiemanns regularization we introduced above as a specic

choice. Such a general regularization can be summarized as follows: rst, we assign a

10

However, some scholars disagree with such an argument involving the habitat and consider the habitat to be

unphysical and completely irrelevant (see, e.g. Ref.[151]).

43

partition of into cells of arbitrary shape. In every cell of the partition we dene edges

s

J

, J = 1, ..., n

s

and loops

i

, i = 1, ..., n

, where n

s

, n

We use to represent the scale of the cell Then x an arbitrary chosen representation

of S U(2). This structure is called a permissible classical regulator if the regulated

Hamiltonian constraint expression with respect to this partition has correct limit when

0.

Second, we assign the dieomorphism covariant property and let the partition adapted to

the choice of the graph. That is, given a cylindrical function

Cyl

3

(7/g), we make

the partition suciently rened that every vertex v V() is contained in exact one cell

of the partition. And if (, v) is dieomorphic to (

, v

, the quin-

tuple (, v, , (s

J

), (

i

)) is dieomorphic to the quintuple (

, v

, (s

J

), (

i

)), where

and

respectively, containing v

and v

respectively.

As a result, the Hamiltonian constraint operator in this general regularization scheme is

expressed as:

E,

(N) =

vV()

N(v)

i

2

i,J

C

iJ

Tr(([A(

i

)] [A(

1

i

)])[A(s

1

J

)][[A(s

J

)],

V

U

v

]),

(N) =

vV()

iN(v)

I,J,K

T

I JK

Tr([A(s

1

I

)][[A(s

I

)],

K][A(s

1

J

)][[A(s

J

)],

K]

[A(s

1

K

)][[A(s

K

)],

V

U

v

]),

(N)

= [

S

E,

(N) 2(1 +

2

)

7

(N)]

,

where C

iJ

and T

I JK

are xed constants independent of the value of , the values of them

are determined such that the above expressions have correct classical limits. After remov-

ing the regulator via dieomorphism invariance the same as we did above, we obtain a

well-dened dieomorphism covariant operator on 1

kin

(or 1

G

) in the sense of the Uni-

form Rovelli-Smolin Topology, or dual-dene the operator on some suitable domain in

Cyl

. Note that such a general scheme of construction exhibits that there is a great deal

of freedom in choosing the regulators, so that there are considerable ambiguities in our

quantization for seeking a proper quantum dynamics for gravity, which is also an open

issue today.

4.4 Master Constraint Programme

Although the Hamiltonian constraint operator introduced above is densely dened on 1

kin

and

dieomorphism covariant, there are still several unsettled problems which are listed below.

It is unclear whether the commutator between two Hamiltonian constraint operators re-

produces the classical Poisson bracket between two Hamiltonian constraints. Hence it is

unclear if the quantum Hamiltonian constraint produces the correct quantum dynamics

with correct classical limit [72][95].

44

The dual Hamiltonian constraint operator does not leave the Hilbert space 1

Di f f

invariant.

Thus the inner product structure of 1

Di f f

cannot be employed in the construction of

physical inner product.

Classically the collection of Hamiltonian constraints does not form a Lie algebra. So one

cannot employ group averaging strategy in solving the Hamiltonian constraint quantum

mechanically, since the strategy depends crucially on group structure.

One may see that all above issues come from the properties of the constraint algebra at classical

level. However, if one could construct an alternative classical constraint algebra, giving the

same constraint phase space, which is a Lie algebra (no structure functions), where the subalge-

bra of dieomorphism constraints forms an ideal, then the programme of solving the constraints

would be in a much better position. Such a constraint Lie algebra was rst introduced by Thie-

mann in [149]. The central idea is to introduce the master constraint:

M :=

1

2

_

d

3

x

C(x)

2

_

det q(x)

, (39)

where

C(x) is the scalar constraint in Eq.(8). One then gets the master constraint algebra:

+(

N), +(

) = +([

N,

N

]),

+(

N), M = 0,

M, M = 0.

The master constraint programme has been well tested in various examples [63][64][65]

[66][67]. In the following, we extend the dieomorphism transformations such that the Hilbert

space 1

Di f f

is separable. This separability of 1

Di f f

and the positivity and the dieomorphism

invariance of M will be working together properly and provide us with powerful functional

analytic tools in the programme to solve the constraint algebra quantum mechanically. The

regularized version of the master constraint can be expressed as

M

:=

1

2

_

d

3

y

_

d

3

x

(x y)

C(y)

_

V

U

C(x)

_

V

U

x

.

Introducing a partition / of the 3-manifold into cells C, we have an operator

H

C

acting on

any cylindrical function f

Cyl

(7/g) in 1

G

as

C

f

vV()

C

(v)

E(v)

v()=v

h

,

v

f

, (40)

via a family of state-dependent triangulations T() on as we did in the last section, where

C

(v) is the characteristic function of the cell C(v) containing a vertex v of the graph , and the

expression of

h

,

v

reads

h

,

v

=

16

3i

2

i jk

Tr(

A(

i j

())

1

A(s

k

())

1

[

A(s

k

()),

_

V

U

v

])

+2(1 +

2

)

4

2

3i

3

i jk

Tr(

A(s

i

())

1

[

A(s

i

()),

K

A(s

j

())

1

[

A(s

j

()),

_

V

U

v

]

A(s

k

())

1

[

A(s

k

()),

K

]). (41)

45

Note that

h

,

v

is similar to that involved in the regulated Hamiltonian constraint operator in

the last section, while the only dierence is that now the volume operator is replaced by its

quare-root in Eq.(41). Hence the action of

H

C

on f

adds arcs a

i j

() with 1/2-representation

with respect to each v() of . Thus, for each > 0,

H

C

is a Yang-Mills gauge invariant and

dieomorphism covariant operator dened on Cyl(7/g). The family of such operators can give

a limit operator

H

C

densely dened on 1

G

by the uniform Rovelli-Smollin topology. Then a

master constraint operator,

M, acting on any

Di f f

Cyl

Di f f

can be dened as [88]

(

M

Di f f

)[ f

] := lim

/;,

Di f f

[

C/

1

2

H

C

(

H

C

)

], (42)

for any f

H

C

(

H

C

)

is

also a nite linear combination of spin-network functions on an extended graph with the same

skeleton of , hence the value of (

M

Di f f

)[ f

Di f f

. Thus

M

Di f f

lies in

the algebraic dual of the space of cylindrical functions. Furthermore, we can show that

Mleaves

the dieomorphism invariant distributions invariant. For any dieomorphism transformation

on ,

(

U

M

Di f f

)[ f

] = lim

/;,

Di f f

[

C/

1

2

H

C

(

H

C

)

]

= lim

/;,

Di f f

[

U

C/

1

2

H

1

()

1

(C)

(

H

1

(

1

(C)

)

]

= lim

/;,

Di f f

[

C/

1

2

H

C

(

H

C

)

], (43)

where in the last step, we used the fact that the dieomorphism transformation leaves the

partition invariant in the limit / and relabel (C) to be C. So we have the result

(

U

M

Di f f

)[ f

] = (

M

Di f f

)[ f

]. (44)

So given a dieomorphism invariant spin-network state T

[s]

, the result state

MT

[s]

must be a

dieomorphism invariant element in the algebraic dual of Cyl(7/g), which means that

MT

[s]

=

[s

1

]

c

[s

1

]

T

[s

1

]

,

then

lim

/;,

0

T

[s]

[

C/

1

2

H

C

(

H

C

)

T

s

2

] =

[s

1

]

c

[s

1

]

T

[s

1

]

[T

s

2

],

where the cylindrical function

_

C/

1

2

H

C

(

H

C

)

T

s

2

is a nite linear combination of spin-network

functions on some graphs

with the same skeleton of (s

2

) up to nite number of arcs. Hence

xing the dieomorphism equivalence class [s], only for spin-networks s

2

lies in nite number

of dieomorphism equivalence class the left hand side of the last equation is non-zero. So there

46

are also only nite number of classes [s

1

] in the right hand side such that c

[s

1

]

is non-zero. As a

result,

MT

[s]

is a nite linear combination of dieomorphism invariant spin-network states and

lies in the Hilbert space of dieomorphism invariant states 1

Di f f

for any [s]. And

M is densely

dened on 1

Di f f

.

Given two dieomorphism invariant spin-network functions T

[s

1

]

and T

[s

2

]

, one can give the

matrix elements of

M as [88][89]

< T

[s

1

]

MT

[s

2

]

>

Di f f

= (

MT

[s

2

]

)[T

s

1

[s

1

]

]

= lim

/;,

C/

1

2

T

[s

2

]

[

H

C

(

H

C

)

T

s

1

[s

1

]

]

= lim

/;,

C/

1

2

1

n

(s

2

)

Di f f ()/Di f f

(s

2

)

GS

(s

2

)

<

U

T

s

2

[s

2

]

C

(

H

C

)

T

s

1

[s

1

]

>

Kin

=

s

lim

/;,

C/

1

2

1

n

(s

2

)

Di f f ()/Di f f

(s

2

)

GS

(s

2

)

<

U

T

s

2

[s

2

]

C

T

s

>

Kin

< T

s

(

H

C

)

T

s

1

[s

1

]

>

Kin

=

[s]

vV((s[s]))

1

2

lim

,

0

T

[s

2

]

[

H

v

T

s,c[s,c]

]

s,c[s,c]

< T

s

(

H

v

)

T

s

1

[s

1

]

>

Kin

, (45)

where Di f f

denotes

the graph symmetry quotient group Di f f

/TDi f f

where TDi f f

is trivial on the graph , and n

the resolution of identity trick in the fourth step. Since only a nite number of terms in the

sum over spin-networks s, cells C /, and dieomorphism transformations are non-zero

respectively, we can interchange the sums and the limit. In the fth step, we take the limit

C v and split the sum

_

s

into

_

[s]

_

s[s]

, where [s, c] denotes the dieomorphism equivalent

class associated with s. Here we also use the fact that, given (s) and (s

to a dieomorphism transformation, there is always a dieomorphism transforming the graph

associated with

H

v

T

s

(v (s)) to that of

H

T

s

(v

(s

)) with (v) = v

, hence T

[s

2

]

[

H

v

T

s[s]

]

is constant for dierent s [s].

Since the term

_

s[s]

< T

s

(

H

v

)

T

s

1

[s

1

]

>

Kin

is independent of the parameter

that by xing a arbitrary family of state-dependent triangulations T(

),

s[s]

< T

s

(

H

v

)

T

s

1

[s

1

]

>

Kin

=

< U

T

s

(

H

v

)

T

s

1

[s

1

]

>

Kin

=

<

H

v

U

T

s

T

s

1

[s

1

]

>

Kin

47

=

< U

1

(

1

(v)

T

s

T

s

1

[s

1

]

>

Kin

= T

[s

1

]

[

H

1

(

)

vV((s))

T

s

], (46)

where are the dieomorphism transformations spanning the dieomorphism equivalent class

[s]. Note that the kinematical inner product in above sum is non-vanishing if and only if ((s)))

coincides with the graph obtained from certain skeleton (s

1

) by the action of (

H

v

)

and v

V(((s))), i.e., the scale

1

(

the action coincides with the scale of certain tetrahedrons in (s) and

1

(v) is a vertex in (s).

Then we can express the matrix elements (83) as:

< T

[s

1

]

MT

[s

2

]

>

Di f f

=

[s]

vV((s[s]))

1

2

lim

,

0

T

[s

2

]

[

H

v

T

s[s]

]T

[s

1

]

[

H

v

T

s[s]

]

=

[s]

vV((s[s]))

1

2

(

H

v

T

[s

2

]

)[T

s[s]

](

H

v

T

[s

1

]

)[T

s[s]

]. (47)

From Eq.(85) and the fact that the master constraint operator

M is densely dened on 1

Di f f

,

it is obvious that

M is a positive and symmetric operator in 1

Di f f

. Therefore, the quadratic

form Q

M

associated with

M is closable [114]. The closure of Q

M

is the quadratic form of

a unique self-adjoint operator

M, called the Friedrichs extension of

M. We relabel

M to be

M, the qualitative description of the kernel of the

Hamiltonian constraint operator in Ref.[136] can be transcribed to describe the solutions to the

equation:

M

Di f f

= 0. In particular, the dieomorphism invariant cylindrical functions based

on at most 2-valent graphs are obviously normalizable solutions. In conclusion, there exists a

positive and self-adjoint operator

M on 1

Di f f

corresponding to the master constraint (75), and

zero is in the point spectrum of

M.

Note that the quantum constraint algebra can be easily checked to be anomaly free. i.e.,

[

M,

U

] = 0, [

M,

M] = 0.

which is consistent with the classical master constraint algebra in this sense. As a result, the dif-

culty of the original Hamiltonian constraint algebra can be avoided by introducing the master

constraint algebra, due to the Lie algebra structure of the latter. Since zero is in the spectrum of

phys

which is the kernel of the

master constraint operator with some suitable physical inner product, and the issue of quantum

anomaly is represented in terms of the size of 1

phys

and the existence of semi-classical states.

Note that we will see in the next section that the master constraint programme can be straight-

forwardly generalized to include matter elds [89]. We list some open problems in the master

constraint programme for further research.

Kernel of Master Constraint Operator

Since the master constraint operator

M is self-adjoint, it is a practical problem to dene

DID of 1

Di f f

:

1

Di f f

_

d()1

,

48

< >

Di f f

=

_

R

d() < >

1

,

where () is the spectral measure with respect to the master constraint operator

M. It

is expected that we can identify 1

=0

with the physical Hilbert space. However, such a

prescription is ambiguous in the case that zero is only in the continuous spectrum, loses

physical information in the case that zero is an embedded eigenvalue and unambiguous

only if zero is an isolated eigenvalue in which case however the whole machinery of the

DID is not needed at all because 1

=0

1

Di f f

and the physical inner product coincide

with the kinematical (dieromorphism invariant) one [63]. There are some improved pre-

scriptions also presented in [63] by decomposing the measure with respect to the spectrum

types before direct integral decomposition, some ambiguities can be canceled by some

physical criterion, e.g., a complete subalgebra of bounded Dirac observables should be

represented irreducibly as self-adjoint operators on the physical Hilbert space, and the

resulting physical Hilbert space should admits a sucient number of semiclassical states.

Nonetheless, due to the complicated structure of the master constraint operator, it is di-

cult anyhow to manage the spectrum analysis and direct integral decomposition. On the

other hand, for the self-adjointness of the master constraint operator and the Lie-algebra

structure of the constraint algebra, a formal group averaging strategy was introduced in

[149] as a more concrete way to get the physical Hilbert space. It is realized by a formal

rigged map

phys

:

phys

: Cyl

Di f f

phys

f

phys

( f ) :=

_

R

dt

2

< e

i

Mt

f . >

Di f f

,

where e

i

Mt

is a one parameter continuous unitary group on 1

Di f f

by the self-adjointness

of

M, and

phys

is a subset of the algebraic dual of Cyl

Di f f

. It is trivial to see that

phys

( f )

is invariant under the (dual) transformation of e

i

Mt

. Thus a inner product can be formally

dened between two algebraic functionals

phys

( f ) and

phys

( f

) in

phys

via:

<

phys

( f )

phys

( f

) >

phys

:=

phys

( f )[ f

],

=

_

R

dt

2

< e

i

Mt

f f

>

Di f f

=

_

R

dt

2

_

R

d()e

it

< f () f

() >

1

=

_

R

d()() < f () f

() >

1

= [

_

R

d()()] < f (0) f

(0) >

1

=0

,

where we have used the spectrum decomposition with respect to the self-adjoint operator

M, the operator e

i

Mt

is represented by multiplication by a number e

it

on each 1

, and

the vector valued function f () is the spectrum decomposition representation of state

f 1

Di f f

. Although we can see from the above argument that the physical inner product

49

is proportional to the inner product in the ber Hilbert space 1

=0

, unfortunately, the

factor

_

R

d()() is divergent when has pure point part, e.g. zero is in the discrete

spectrum of

M. That is one reason why we claim that the above argument is formal.

On the other hand, the group averaging strategy and the formal physical inner product

we just dened has potential relationships with path-integral formulation and spin foam

models due to the positivity of the master constraint operator

M [149], and hopefully,

we may obtain the physical transition amplitude from this physical inner product in the

future. However, the whole technique of group averaging for solving the master constraint

is still formal so far, and the rigorous calculations for it has not done yet as far as we know.

Dirac Observables

Classically, one can prove that a function O C

to the scalar constraint if and only if

O, O, M

A

= 0.

We dene O to be a strong observable with respect to the scalar constraint if and only if

O, M

A

= 0,

and to be a ultra-strong observable if and only if

O, S(N)

A

= 0.

In quantum version, an observable

O is a weak Dirac observable if and only if

O leaves

1

phys

invariant, while

Ois nowcalled a strong Dirac observable if and only if

Ocommutes

with the master constraint operator

M. Given a bounded self-adjoint operator

O dened

on 1

Di f f

, for instance, a spectral projection of some observables leaving 1

Di f f

invariant,

if the uniform limit exists, the bounded self-adjoint operator dened by group averaging

[O] := lim

T

1

2T

_

T

T

dt

U(t)

1

O

U(t)

commutes with

M and hence becomes a strong Dirac observable on the physical Hilbert

space.

Testing the Classical Limit of the Master Constraint Operator

One needs to construct spatial dieomorphism invariant semiclassical states to calculate

the expectation value and uctuation of the master constraint operator. If the results

coincide with the classical values up to corrections, one can go ahead to nish our

quantization programme with condence.

50

5 Quantum Matter Field on a Quantum Background

In ordinary quantum eld theory, the quantum eld is dened on a smooth background space-

time. However, it is expected that the smooth structure of a spacetime may break down at

Planck scale, so the present treatment of quantum eld theory is valid only in a semiclassical

sense. Thus we would like to modify the formulation of present quantum eld theory to make

it compatible with the quantum theory of gravity(spacetime) which we already established in

previous sections so as to explore the behavior of the quantum matter eld under Planck scale

and at extremely strong gravitational elds, e.g. inside the black hole or at the early age of the

universe.

In the following, an alternative quantization of scalar eld will be introduced, the advantage

of such a quantization scheme is that the quantumscalar eld doesnt depend on the background.

We will also see that the quantization technique for the previous Hamiltonian constraint can be

generalized to quantize the Hamiltonian of matter elds coupled to gravity. Then it is shown

that an operator corresponding to the Hamiltonian of the scalar eld can be well dened on

the coupled dieomorphism invariant Hilbert space. It is even positive and self-adjoint without

any divergence. Thus quantum gravity acts exactly as a natural regulator for the quantum scalar

eld in the polymer representation. Moreover, to study the whole dynamical systemof the scalar

eld coupled to gravity, a Hamiltonian constraint operator is dened in the coupled kinematical

Hilbert space. The contribution of the scalar eld to the Hamiltonian constraint can be promoted

to a positive self-adjoint operator. To avoid possible quantum anomalies and nd the physical

Hilbert space, we will also introduce the master constraint programme for the coupled system.

A self-adjoint master constraint operator is obtained in the dieomorphism invariant Hilbert

space, which assures the feasibility of the programme.

5.1 Polymer-like Representation of a Scalar Field

We begin with the total Hamiltonian of the gravity coupled with a massless real scalar eld

which is a linear combination of constraints:

1

tot

=

i

G

i

+ N

a

C

a

+ NC,

where

i

, N

a

and N are Lagrange multipliers, and the three constraints in the Hamiltonian are

expressed as [30][87]:

G

i

= D

a

P

a

i

:=

a

P

a

i

+

k

i j

A

i

a

P

a

k

, (48)

C

a

=

P

b

i

F

i

ab

A

i

a

G

i

+

a

, (49)

C =

2

2

_

det q

P

a

i

P

b

j

[

i j

k

F

k

ab

2(1 +

2

)K

i

[a

K

j

b]

]

+

1

_

det q

[

M

2

i j

P

a

i

P

b

j

(

a

)

b

+

1

2

M

2

], (50)

here the real number

M

is the coupling constant, and denotes the momentum conjugate to :

:=

[

M

N

_

det q(

N

a

a

).

51

Thus one has the elementary Poisson brackets

A

i

a

(x),

P

b

j

(y) =

a

b

i

j

(x, y),

(x), (y) = (x, y).

Note that the second term of the Hamiltonian constraint (50) is just the Hamiltonian of the real

scalar eld.

Then we look for the background independent representation for the real scalar eld coupled

to gravity, following the polymer representation of the scalar eld [27]. The classical cong-

uration space, 1, consists of all real-valued smooth functions on . Given a set of a nite

number of points X = x

1

, ..., x

N

in , a equivalence relation can be dened by: given two scalar

eld

1

,

2

1,

1

2

if and only if exp[i

i

1

(x

i

)] = exp[i

j

2

(x

j

)] for all x

i

X and all real

number

j

. Hence we obtain a bijection between 1/ and R

X

, which is N copies of the Bohr

compactication of R [132]. Since one can dene a projective family with respect to the set of

point (graph for scalar eld), thus a projective limit 1, which is a compact topological space,

is obtained as the quantum conguration space of scalar eld. Next, we denote by Cyl

X

(1) the

vector space generated by nite linear combinations of the following functions of :

T

X,

() :=

_

x

j

X

exp[i

j

(x

j

)],

where (

1

,

2

, ,

N

) are arbitrary non-zero real numbers assigned at each point. It is

obvious that Cyl

X

(1) has the structure of a -algebra. The vector space Cyl(1) of all cylindrical

functions on 1 is dened by the linear span of the linear span of T

0

= 1 and T

X,

. Completing

Cyl(1) with respect to the sup norm, one obtains a unital Abelian C*-algebra Cyl(1). Thus

one can use the GNS structure to construct its cyclic representations. A preferred positive linear

functional

0

on Cyl(1) is dened by

0

(T

X,

) =

_

1 if

j

= 0 j

0 otherwise,

which denes a dieomorphism-invariant faithful Borel measure on 1 as

_

1

d(T

X,

) =

_

1 if

j

= 0 j

0 otherwise.

(51)

Thus one obtains the Hilbert space, 1

KG

kin

which is dened by L

2

(1, d), of square integrable

functions on a compact topological measure space 1 with respect to . The inner product can

be expressed explicitly as:

< T

c

T

c

>

KG

kin

=

cc

, (52)

where the label c := (X, ) are called scalar-network.

As one might expect, the quantum conguration space 1 is just the Gelfand spectrum of

Cyl(1). More concretely, for a single point set X

0

x

0

, Cyl

X

0

(1) is the space of all almost

periodic functions on a real line R. The Gelfand spectrum of the corresponding C*-algebra

52

Cyl

X

0

(1) is the Bohr completion R

x

0

of R [27], which is a compact topological space such that

Cyl

X

0

(1) is the C*-algebra of all continuous functions on R

x

0

. Since R is densely embedded in

R

x

0

, R

x

0

can be regarded as a completion of R.

It is clear from Eq.(51) that an orthonomal basis in 1

KG

kin

is given by the scalar vacuum

T

0

= 1 and so-called scalar-network functions T

c

(), where c = (X, ) and (

1

,

2

, ,

N

)

are non-zero real numbers. So the total kinematical Hilbert space 1

kin

is the direct product of

the kinematical Hilbert space 1

GR

kin

for gravity and the kinematical Hilbert space for real scalar

eld, i.e., 1

kin

:= 1

GR

kin

1

KG

kin

. Then the spin-scalar-network state T

s,c

T

s

(A) T

c

()

Cyl

(s)

(7/g) Cyl

X(c)

(1) Cyl

(s,c)

is a gravity-scalar cylindrical function on graph (s, c)

(s) X(c). Note that generally X(c) may not coincide with the vertices of the graph (s). It is

straightforward to see that all of these functions constitutes an orthonormal basis in 1

kin

as

< T

s

(A) T

c

()T

s

(A) T

c

() >

kin

=

s

c

.

Note that none of 1

kin

, 1

GR

kin

and 1

KG

kin

is a separable Hilbert space.

Given a pair (x

0

,

0

), there is an elementary conguration for the scalar eld, the so-called

point holonomy,

U(x

0

,

0

) := exp[i

0

(x

0

)].

It corresponds to a conguration operator

U(x

0

,

0

), which acts on any cylindrical function

() Cyl

X(c)

(1) by

U(x

0

,

0

)() = U(x

0

,

0

)(). (53)

All these operators are unitary. But since the family of operators

U(x

0

, ) fails to be weakly

continuous in , there is no eld operator

(x) on 1

KG

kin

. The momentum functional smeared on

a 3-dimensional region R is expressed by

(R) :=

_

R

d

3

x(x).

The Poisson bracket between the momentum functional and a point holonomy can be easily

calculated to be

(R), U(x, ) = i

R

(x)U(x, ),

where

R

(x) is the characteristic function for the region R. So the momentum operator is dened

by the action on scalar network functions T

c=(X,)

as

(R)T

c

() := i(R), T

c

() = [

x

j

X

j

(x

j

)]T

c

().

Now we can impose the quantum constraints on 1

kin

and consider the quantum dynamics.

First, the Gauss constraint can be solved independently of 1

KG

kin

, since it only involves the

gravitational eld. It is also expected that the dieomorphism constraint can be implemented

by the group averaging strategy in the similar way as in the case of pure gravity. Given a spatial

53

dieomorphism transformation , a unitary transformation

U

space 1

kin

, which is expressed as

T

s=((s),j,i),c=(X(c),)

= T

s=(((s)),j,i),c=((X(c)),)

.

Then the dieromorphism invariant spin-scalar-network functions are dened by group averag-

ing as

T

[s,c]

:=

1

n

(s,c)

Di f f ()/Di f f

(s,c)

GS

(s,c)

T

s,c

, (54)

where Di f f

denotes

the graph symmetry quotient group Di f f

/TDi f f

where TDi f f

phisms which is trivial on the graph , and n

. Following the

standard strategy in quantization of pure gravity, an inner product can be dened on the vector

space spanned by the dieomorphism invariant spin-scalar-network functions (and the vacuum

states for gravity, scalar and both respectively) such that they form an orthonormal basis as:

< T

[s,c]

T

[s

,c

]

>

Di f f

:= T

[s,c]

[T

s

,c

[s

,c

]

] =

[s,c],[s

,c

]

. (55)

After the completion procedure, we obtain the expected Hilbert space of dieomorphism in-

variant states for the scalar eld coupled to gravity, which is denoted by 1

Di f f

.

5.2 Dieomorphism Invariant Hamiltonian of a Scalar Field

In the following discussion, we consider the quantum scalar eld on a uctuating background.

A similar idea was considered in Ref.[139], where a Hamiltonian operator with respect to a U(1)

group representation of the scalar eld is dened on a kinematical Hilbert space 1

kin

of matter

coupled to gravity. Then an eective Hamiltonian operator of the scalar eld can be constructed

as a quadratic form via

<

matter

,

H

e f f

matter

(m)

matter

>

KG

kin

:= <

grav

(m)

matter

,

H

matter

grav

(m)

matter

>

kin

, (56)

where

grav

(m) 1

GR

kin

presents a semiclassical state of gravity approximating some classical

spacetime background m where the quantum scalar eld lives. Thus the eective Hamiltonian

operator

H

e f f

matter

(m) of scalar eld contains also the information of the uctuating background

metric. In the light of this idea, we will construct a Hamiltonian operator

S

KG

for scalar eld

in the polymer-like representation. It turns out that this Hamiltonian operator can be dened

in the Hilbert space 1

Di f f

of dieomorphism invariant states for scalar eld coupled to gravity

without UV-divergence. So the quantum dynamics of the scalar eld is obtained in a dieo-

morphism invariant way, which is expected in the programme of loop quantum gravity. Thus,

here an eective Hamiltonian operator of the scalar eld could be extracted in 1

Di f f

by den-

ing <

[m]

(A, ),

S

KG

[m]

(A, ) >

Di f f

to be its expectation value on dieomorphism invariant

states () of the scalar eld, where the dieomorphism invariant semiclassical state

[m]

(A)

represents certain uctuating geometry with spatial dieomorphism invariance, and the label

54

[m] denotes the classical geometry approximated by

[m]

(A). Moreover, the quadratic proper-

ties of the scalar eld Hamiltonian will provide powerful functional analytic tools in the quan-

tization procedure, such that the self-adjointness of the Hamiltonian operator can be proved by

a theorem in functional analysis.

Then the crucial point is to dene an operator corresponding to the Hamiltonian functional

S

KG

of the scalar eld, which can be decomposed into two parts

S

KG

= S

KG,

+ S

KG,Kin

,

where

S

KG,

=

M

2

_

d

3

x

1

_

det q

i j

P

a

i

P

b

j

(

a

)

b

,

S

KG,Kin

=

1

2

M

_

d

3

x

1

_

det q

2

.

We will employ the following identities:

P

a

i

=

1

2

abc

i jk

e

j

b

e

k

c

and e

i

a

(x) =

2

A

i

a

(x), V

U

x

,

where

abc

denotes the Levi-Civita tensor tensity and V

U

x

is the volume of an arbitrary neigh-

borhood U

x

containing the point x. By using the point-splitting strategy, the regulated version

of the Hamiltonian is obtained as:

S

KG,

=

M

2

_

d

3

y

_

d

3

x

(x y)

i j

1

_

V

U

P

a

i

(x)(

a

(x))

1

_

V

U

P

b

j

(y)

b

(y)

=

32

M

4

_

d

3

y

_

d

3

x

(x y)

i j

aec

(

a

(x))Tr(

i

A

e

(x), V

3/4

U

x

A

c

(x), V

3/4

U

x

)

bf d

(

b

(y))Tr(

j

A

f

(y), V

3/4

U

y

A

d

(y), V

3/4

U

y

),

S

KG,Kin

=

1

2

M

_

d

3

x(x)

_

d

3

y(y)

_

d

3

u

det(e

i

a

(u))

(V

U

u

)

3/2

_

d

3

w

det(e

i

a

(w))

(V

U

w

)

3/2

(x y)

(u x)

(w y)

=

1

2

M

2

8

9()

6

_

d

3

x(x)

_

d

3

y(y)

_

d

3

u

abc

Tr(A

a

(u),

_

V

U

u

A

b

(u),

_

V

U

u

A

c

(u),

_

V

U

u

)

_

d

3

w

de f

Tr(A

d

(w),

_

V

U

w

A

e

(w),

_

V

U

w

A

f

(w),

_

V

U

w

)

(x y)

(u x)

(w y),

55

where we denote by A

a

A

i

a

i

,

scale such that lim

0

(x y)/

3

= (x y), and V

U

x

is the volume of the box. In order

to quantize the Hamiltonian S

KG

as a well-dened operator in the polymer-like representation,

we have to express the classical formula of the Hamiltonian in terms of elementary variables

with clear quantum analogs by introducing a triangulation T() of , where the parameter

describes how ne the triangulation is. The quantity regulated on the triangulation is required

to have correct limit when 0. Given a tetrahedron T(), we use s

i

()

i=1,2,3

to denote

the three outgoing oriented segments in with a common beginning point v() = s(s

i

()) and

use a

i j

() to denote the arcs connecting the end points of s

i

() and s

j

(). Then several loops

i j

() are formed by

i j

() := s

i

() a

i j

() s

j

()

1

. Thus we have the identities:

_

s()

dt A

a

s

a

(t), V

3/4

U

s(s())

= A(s())

1

A(s()), V

3/4

U

s(s())

+ o(),

and

_

s()

dt

a

s

a

(t) =

1

i

U(s(s()), )

1

[U(t(s()), ) U(s(s()), )] + o()

for nonzero , where s(s()) and t(s()) denote respectively the beginning and end points of

segment s() with scale associated with a tetrahedron . Regulated on the triangulation, the

scalar eld Hamiltonian reads

S

KG,

=

4

M

9

4

T()

T()

(v() v(

))

i j

lmn

1

U(v(), )

1

[U(t(s

l

()), ) U(v(), )]

Tr(

i

A(s

m

())

1

A(s

m

()), V

3/4

U

v()

A(s

n

())

1

A(s

n

()), V

3/4

U

v()

)

kpq

1

U(v(

), )

1

[U(t(s

k

(

)), ) U(v(

), )]

Tr(

j

A(s

p

(

))

1

A(s

p

(

)), V

3/4

U

v(

)

A(s

q

(

))

1

A(s

q

(

)), V

3/4

U

v(

)

),

S

KG,Kin

=

16

81

M

()

6

T()

T()

()(

T()

imn

Tr(A(s

i

(

))

1

A(s

i

(

)),

_

V

U

v(

)

A(s

m

(

))

1

A(s

m

(

)),

_

V

U

v(

)

A(s

n

(

))

1

A(s

n

(

)),

_

V

U

v(

)

)

T()

jkl

Tr(A(s

j

(

))

1

A(s

j

(

)),

_

V

U

v(

)

A(s

k

(

))

1

A(s

k

(

)),

_

V

U

v(

)

A(s

l

(

))

1

A(s

l

(

)),

_

V

U

v(

)

)

(v() v(

))

(v(

) v())

(v(

) v(

)). (57)

56

Note that the above regularization is explicitly dependent on the parameter , which will lead to

a kind of quantization ambiguity of the real scalar eld dynamics in polymer-like representation.

Introducing a partition / of the 3-manifold into cells C, we can smear the essential square

roots of S

KG,

and S

KG,Kin

in one cell C respectively and promote them as regulated operators

in 1

kin

with respect to triangulations T() depending on spin-scalar-network state T

s,c

as

W

,C

,i

T

s,c

=

vV((s,c))

C

(v)

E(v)

v()=v

h

,

,v,i

T

s,c

,

W

,C

Kin

T

s,c

=

vV((s,c))

C

(v)

E(v)

v()=v

h

,

Kin,v

T

s,c

, (58)

where

C

(v) is the characteristic function of the cell C, and

h

,

,v,i

:=

i

lmn

1

(v)

U(v, (v))

1

[

U(t(s

l

()), (v))

U(v, (v))]

Tr(

i

A(s

m

())

1

[

A(s

m

()),

V

3/4

U

v

]

A(s

n

())

1

[

A(s

n

()),

V

3/4

U

v

]),

h

,

Kin,v

:=

1

(i)

3

(v)

lmn

Tr(

A(s

l

())

1

[

A(s

l

()),

_

V

U

v

]

A(s

m

())

1

[

A(s

m

()),

_

V

U

v

]

A(s

n

())

1

[

A(s

n

()),

_

V

U

v

]). (59)

Both operators in (58) and their adjoint operators are densely dened on 1

kin

. We now give

several remarks on their properties.

Removal of regulator

It is not dicult to see that the action of the operator

W

,C

,i

on a spin-scalar-network

function T

s,c

is graph-changing. It adds nite number of vertices with representation (v)

at t(s

i

()) with distance from the vertex v. Recall that the action of the gravitational

Hamiltonian constraint operator on a spin network function is also graph-changing. As

a result, the family of operators

W

,C

,i

also fails to be weakly converged when 0.

However, due to the dieomorphism covariant properties of the triangulation, the limit

operator can be well-dened via the uniformRovelli-Smolin topology, or equivalently, the

operator can be dually dened on dieomorphism invariant states. But the dual operator

cannot leave 1

Di f f

invariant.

Quantization ambiguity

As a main dierence of the dynamics in polymer-like representation from that in U(1)

group representation [138], a continuous label appears explicitly in the expression of

(58). Hence there is an one-parameter quantization ambiguity due to the real scalar eld.

Recall that the construction of gravitational Hamiltonian constraint operator also has a

similar ambiguity due to the choice of the representations j of the edges added by its

action. A related quantization ambiguity also appears in the dynamics of loop quantum

cosmology [50].

57

Since our quantum eld theory is expected to be dieomorphism invariant, we would like to

dene the Hamiltonian operator of polymer scalar eld in the dieomorphism invariant Hilbert

space 1

Di f f

. For this purpose we x the parameter to be a non-zero constant at every point.

Then what we will do is to employ the new quantization strategy developed in Refs. [149]

and [140]. We rst construct a quadratic form in the light of a new inner product dened in

Ref.[140] on the algebraic dual 1

spin-scalar-networks T

s,c

(where the family of labels s, c includes the vacuum states for gravity,

scalar and both). Then we prove that the quadratic form is closed. Note that, although the

calculation employing this inner product is formal, it can lead to a well-dened expression

of the desired quadratic form Eq.(65). Since an arbitrary element of 1

is of the form =

_

s,c

c

s,c

< T

s,c

>

kin

, one can formally dene an inner product < >

on 1

via

< ,

>

:= <

s,c

c

s,c

< T

s,c

>

kin

,c

,c

< T

s

,c

>

kin

>

:=

s,c;s

,c

c

s,c

c

,c

< T

s,c

T

s

,c

>

kin

1

_

([s, c])([s

, c

])

=

s,c

c

s,c

c

s,c

1

([s, c])

, (60)

where the Cantor aleph denotes the cardinal of the set [s, c]. Note that we exchange the

coecients on which the complex conjugate was taken in Ref.[140], so that the inner product

<

Di f f

Di f f

>

reduces to <

Di f f

Di f f

>

Di f f

for any

Di f f

,

Di f f

1

Di f f

. Completing the

quotient with respect to the null vectors by this inner product, one gets a Hilbert space 1

. Our

purpose is to construct a quadratic form associated to some positive and symmetric operator in

analogy with the classical expression of (57). So the quadratic form should rst be given in

a positive and symmetric version. It is then natural to dene two quadratic forms on a dense

subset of 1

Di f f

1

as:

Q

KG,

(

Di f f

,

Di f f

) := lim

/

C/

64

4

M

9

4

i j

<

W

C

,i

Di f f

W

C

, j

Di f f

>

,

Q

KG,Kin

(

Di f f

,

Di f f

) := lim

/

C/

8

4

16

81

M

()

6

<

W

C

Kin

Di f f

W

C

Kin

Di f f

>

,

(61)

for any two states

Di f f

and

Di f f

which are nite linear combinations of T

[s,c]

, where the dual

limit operator

W

C

of either family of

W

,C

,i

or

W

,C

Kin

in (58) is naturally dened on dieomor-

phism invariant states as

W

C

Di f f

[T

s,c

] = lim

0

Di f f

[

W

,C

T

s,c

]. (62)

To show that the quadratic forms are well dened, we write

W

C

,i

Di f f

=

s,c

w

,i,s,c

(C) < T

s,c

>

,i,s,c

(C) = (

W

C

,i

Di f f

)[T

s,c

],

W

C

Kin

Di f f

=

s,c

w

Kin,s,c

(C) < T

s,c

>

Kin,s,c

(C) = (

W

C

Kin

Di f f

)[T

s,c

].

58

Then, by using the inner product (60) the quadratic forms in (61) become

Q

KG,

(

Di f f

,

Di f f

)

:= lim

/

C/

64

4

M

9

4

i j

s,c

w

,i,s,c

(C)w

, j,s,c

(C)

1

([s, c])

= lim

/

C/

64

4

M

9

4

i j

[s,c]

1

([s, c])

s,c[s,c]

w

,i,s,c

(C)w

, j,s,c

(C),

Q

KG,Kin

(

Di f f

,

Di f f

)

:= lim

/

C/

8

4

16

81

M

()

6

s,c

w

Kin,s,c

(C)w

Kin,s,c

(C)

1

([s, c])

= lim

/

C/

8

4

16

81

M

()

6

[s,c]

1

([s, c])

s,c[s,c]

w

Kin,s,c

(C)w

Kin,s,c

(C).

(63)

Note that, since

Di f f

is a nite linear combination of the dieomorphism invariant spin-scalar-

network basis, taking account of the operational property of

W

C

there are only a nite number

of terms in the summation

_

[s,c]

contributing to (63). Hence we can interchange

_

[s,c]

and

lim

/

_

C/

in above calculation. Moreover, for a suciently ne partition such that each cell

contains at most one vertex, the sum over cells therefore reduces to nite terms with respect to

the vertices of (s, c). So we can interchange

_

s,c[s,c]

and lim

/

_

C/

to obtain:

Q

KG,

(

Di f f

,

Di f f

)

= 64

4

M

9

4

i j

[s,c]

1

([s, c])

s,c[s,c]

lim

/

C/

w

,i,s,c

(C)w

, j,s,c

(C)

= 64

4

M

9

4

i j

[s,c]

1

([s, c])

s,c[s,c]

vV((s,c))

(

W

v

,i

Di f f

)[T

s,c

](

W

v

, j

Di f f

)[T

s,c

],

Q

KG,Kin

(

Di f f

,

Di f f

)

= 8

4

16

81

M

()

6

[s,c]

1

([s, c])

s,c[s,c]

lim

/

C/

w

Kin,s,c

(C)w

Kin,s,c

(C)

= 8

4

16

81

M

()

6

[s,c]

1

([s, c])

s,c[s,c]

vV((s,c))

(

W

v

Kin

Di f f

)[T

s,c

](

W

v

Kin

Di f f

)[T

s,c

],

(64)

where the limit / has been taken so that C v. Since given (s, c) and (s

, c

) which

are dierent up to a dieomorphism transformation, there is always a dieomorphism trans-

forming the graph associated with

W

,v

T

s,c

(v (s, c)) to that of

W

,v

T

s

,c

(v

(s

, c

)) with

(v) = v

, (

W

v

Di f f

)[T

s,c[s,c]

] is constant for dierent (s, c) [s, c], i.e., all the ([s, c]) terms

in the sum over (s, c) [s, c] are identical. Hence the nal expressions of the two quadratic

forms can be written as:

Q

KG,

(

Di f f

,

Di f f

)

59

= 64

4

M

9

4

i j

[s,c]

vV((s,c))

(

W

v

,i

Di f f

)[T

s,c[s,c]

](

W

v

, j

Di f f

)[T

s,c[s,c]

],

Q

KG,Kin

(

Di f f

,

Di f f

)

= 8

4

16

81

M

()

6

[s,c]

vV((s,c))

(

W

v

Kin

Di f f

)[T

s,c[s,c]

](

W

v

Kin

Di f f

)[T

s,c[s,c]

].

(65)

Note that both quadratic forms in (65) have nite results and hence their form domains are

dense in 1

Di f f

. Moreover, both of them are obviously positive, and the following theorem will

demonstrate their closedness.

Theorem 5.2.1: Both Q

KG,

and Q

KG,Kin

are densely dened, positive and closed quadratic

forms on 1

Di f f

, which are associated uniquely with two positive self-adjoint operators respec-

tively on 1

Di f f

such that

Q

KG,

(

Di f f

,

Di f f

) = <

Di f f

S

KG,

Di f f

>

Di f f

Q

KG,Kin

(

Di f f

,

Di f f

) = <

Di f f

S

KG,Kin

Di f f

>

Di f f

.

Therefore the Hamiltonian operator

S

KG

:=

S

KG,

+

S

KG,Kin

(66)

is positive and also have a unique self-adjoint extension.

Proof: We follow the strategy developed in Refs.[140] and [88] to prove that both Q

KG,

and

Q

KG,Kin

are closeable and uniquely induce two positive self-adjoint operators

S

KG,

and

S

KG,Kin

.

One can formally dene

S

KG,

and

S

KG,Kin

acting on dieomorphism invariant spin-scalar net-

work functions via:

S

KG,

T

[s

1

,c

1

]

:=

[s

2

,c

2

]

Q

KG,

(T

[s

2

,c

2

]

, T

[s

1

,c

1

]

)T

[s

2

,c

2

]

, (67)

S

KG,Kin

T

[s

1

,c

1

]

:=

[s

2

,c

2

]

Q

KG,Kin

(T

[s

2

,c

2

]

, T

[s

1

,c

1

]

)T

[s

2

,c

2

]

. (68)

Then we need to show that both of the above operators are densely dened on the Hilbert

space 1

Di f f

. Given a dieomorphism invariant spin-scalar network function T

[s

1

,c

1

]

, there are

only a nite number of terms T

[s

1

,c

1

]

[

W

,v

T

s,c[s,c]

] which are nonzero in the sum over equivalent

classes [s, c] in (65). On the other hand, given one spin-scalar-network function T

s,c[s,c]

, there

are also only a nite number of possible T

[s

2

,c

2

]

such that the terms T

[s

2

,c

2

]

[

W

,v

T

s,c[s,c]

] are

nonzero. As a result, only a nite number of terms survive in both sums over [s

2

, c

2

] in Eqs.

(67) and (68). Hence both

S

KG,

and

S

KG,Kin

are well dened on spin-scalar-network basis.

Then it follows from Eqs. (65), (67) and (68) that they are positive and symmetric operators

densely dened in 1

Di f f

, whose quadratic forms coincide with Q

KG,

and Q

KG,Kin

on their form

domains. Hence both Q

KG,

and Q

KG,Kin

have positive closures and uniquely induce self-adjoint

60

(Friedrichs) extensions of

S

KG,

and

S

KG,Kin

respectively [114], which we denote by

S

KG,

and

S

KG,Kin

as well. As a result, the Hamiltonian operator

S

KG

dened by Eq.(66) is also positive

and symmetric. Hence it has a unique self-adjoint (Friedrichs) extension.

We notice that, from a dierent perspective, one can construct the same Hamiltonian oper-

ator

1

KG

without introducing an inner product on 1

Using the two well-dened operators

W

,C

,i

and

W

,C

Kin

as in (58), as well as their adjoint operators

(

W

,C

,i

)

and (

W

,C

Kin

)

Di f f

corresponding to the two terms in

(57) by

(

S

KG,

Di f f

)[T

s,c

] = lim

,

0,/

Di f f

[

C/

64

4

M

9

4

i j

W

,C

,i

(

W

,C

, j

)

T

s,c

]

(

S

KG,Kin

Di f f

)[T

s,c

] = lim

,

0,/

Di f f

[

C/

8

4

16

81

M

()

6

W

,C

Kin

(

W

,C

Kin

)

T

s,c

],

(69)

for any spin-scalar-network T

s,c

. In analogy with the discussion about the master constraint

operator and Ref.[88], it can be shown that both above operators leave 1

Di f f

invariant and are

densely dened on 1

Di f f

. Moreover, the quadratic forms associated with them coincide with

the quadratic forms in (65). Thus the Hamiltonian operator

S

KG

:=

S

KG,

+

S

KG,Kin

coincides

with the one constructed in the quadratic form approach.

In summary, we have constructed a positive self-adjoint Hamiltonian operator on 1

Di f f

for

the polymer-like scalar eld, depending on a chosen parameter . Thus there is an 1-parameter

ambiguity in the construction. However, there is no UV divergence in this quantum Hamiltonian

without renormalization, since quantum gravity plays the role of a natural regulator for the

polymer-like scalar eld.

5.3 Hamiltonian Constraint Equation for the Coupled System

In this section we consider the whole dynamical system of scalar eld coupled to gravity. Re-

call that in perturbative quantum eld theory in curved spacetime, the denition of some basic

physical quantities, such as the expectation value of the energy-momentum, is ambiguous and

it is challenging dicult to calculate the back-reaction of quantum elds on the background

spacetime [157]. This is reected by the fact that the semi-classical Einstein equation,

R

[g]

1

2

R[g]g

= <

T

are known to be inconsistent and ambiguous [70][146]. One could speculate that the di-

culty is related to the fact that the usual formulation of quantum eld theories are background

dependent. Following this line of thought, if the quantization programme is by construction

non-perturbative and background independent, it may be possible to solve the problems fun-

damentally. In loop quantum gravity, there is no assumption of a priori background metric at

all. The quantum geometry and quantum matter elds are coupled and uctuating naturally

with respect to each other on a common manifold. On the other hand, there exists the time

61

problem in quantum theory of pure gravity, since all the physical states have to satisfy certain

version of quantum Wheeler-DeWitt constraint equation. However, the situation could improve

when matter eld is coupled to gravity [54][122]. In the following construction, we impose the

quantum Hamiltonian constraint on 1

kin

, and thus dene a quantum Wheeler-DeWitt constraint

equation for the scalar eld coupled to gravity. Then one can gain an insight into the problem

of time from the coupled equation, and the back-reaction of the quantum scalar eld is included

in the framework of loop quantum gravity.

We now dene an operator in 1

kin

corresponding to the scalar eld part S

KG

(N) of the total

Hamiltonian constraint functional, which can be read out from Eqs. (35) and (50) as

S

KG

(N) = S

KG,

(N) + S

KG,Kin

(N),

where

S

KG,

(N) =

M

2

_

d

3

xN

1

_

det q

i j

P

a

i

P

b

j

(

a

)

b

,

S

KG,Kin

(N) =

1

2

M

_

d

3

xN

1

_

det q

2

.

In analogy with the regularization and quantization in the previous section, the regulated version

of quantum Hamiltonian constraint

S

KG

(N) of scalar eld is expressed by taking the limit C

v:

KG

(N)T

s,c

:=

vV((s,c))

N(v)[64

4

M

9

4

i j

(

W

,v

,i

)

W

,v

, j

+ 8

4

16

81

M

()

6

(

W

,v

Kin

)

W

,v

Kin

]T

s,c

, (71)

where for any v V((s, c)), the operators

W

,v

,i

T

s,c

=

1

E(v)

v()=v

h

,

,v,i

T

s,c

,

W

,v

Kin

T

s,c

=

1

E(v)

v()=v

h

,

Kin,v

T

s,c

,

and their adjoints are all densely dened in 1

kin

. Hence the family of Hamiltonian constraint

operators (71) is also densely dened, and the regulator can be removed via the Uniform

Rovelli-Smollin topology, or equivalently the limit operator dually acts on dieomorphism in-

variant states as

(

S

KG

(N)

Di f f

)[ f ] = lim

0

Di f f

[

S

KG

(N) f ], (72)

for any f Cyl(7/g) Cyl(1). Similar to the dual of

S

GR

(N), the operator

S

KG

(N) fails to

commute with the dual of nite dieomorphism transformation operators, unless the smearing

function N(x) is a constant function over . In fact, the dual Hamiltonian constraint opera-

tor smeared by N = 1 is just the dieomorphism invariant Hamiltonian we just dened in the

62

last subsection. From Eq.(71), it is not dicult to prove that for positive N(x) the Hamilto-

nian constraint operator

S

KG

(N) of a scalar eld is positive and symmetric in 1

kin

and hence

has a unique self-adjoint extension [89]. Our construction of

S

KG

(N) is similar to that of the

Higgs eld Hamiltonian constraint in Ref.[138]. However, like the case of

S

KG

, there is a

one-parameter ambiguity in our construction of

S

KG

(N) due to the real scalar eld, which is

manifested as the continuous parameter in the expression of

h

,

,v,i

in (59). Note that now is

not required to be a constant, i.e., its value can be changed from one point to another. Thus the

total Hamiltonian constraint operator of scalar eld coupled to gravity has been obtained as

S(N) =

S

GR

(N) +

S

KG

(N). (73)

Again, there is no UV divergence in this quantum Hamiltonian constraint. Recall that, in stan-

dard quantum eld theory the UV divergence can only be cured by a renormalization procedure,

in which one has to multiply the Hamiltonian by a suitable power of the regulating parameter

. However, now has naturally disappeared from the expression of (73). So renormalization is

not needed for the polymer-like scalar eld coupled to gravity, since quantum gravity has played

the role of a natural regulator. This result heightens our condence that the issue of divergences

in quantum eld theory can be cured in the framework of loop quantum gravity.

Now we have obtained the desired matter-coupled quantum Hamiltonian constraint equation

(

S

KG

(N)

Di f f

)[ f ] = (

S

GR

(N)

Di f f

)[ f ]. (74)

Comparing it with the well-known Sch ordinger equation for a particle,

i

t

(x, t) = H( x,

i

x

)(x, t),

where (x, t) L

2

(R, dx) and t is a parameter labeling time evolution, one may take the view-

point that the matter eld constraint operator

S

KG

(N) plays the role of i

t

. Then appears as

the parameter labeling the evolution of the gravitational eld state. In the reverse viewpoint, the

gravitational eld would become the parameter labeling the evolution of the quantum matter

eld. Note that such an idea has been successfully applied in a loop quantum cosmology model

to help us to understand the quantum nature of big bang in the deep Planck regime [28][29].

5.4 Master Constraint for the Coupled System

Recall that in order to avoid possible quantum anomalies and nd the physical Hilbert space

of quantum gravity, the master constraint programme was rst introduced in the last section.

The central idea is to construct an alternative classical constraint algebra, giving the same con-

straint phase space, which is a Lie algebra (no structure functions) and where the subalgebra of

dieomorphism constraints forms an ideal. A self-adjoint master constraint operator for loop

quantum gravity is then proposed on 1

Di f f

. The master constraint programme can be gener-

alized to matter elds coupled to gravity in a straightforward way. We now take the massless

real scalar eld to demonstrate the construction of a master constraint operator according to

the same strategy as we did in the last section. By this approach one not only avoids possible

quantum anomalies which might appear in the conventional canonical quantization method, but

63

also might give a qualitative description of the physical Hilbert space for the coupled system.

We introduce the master constraint for the scalar eld coupled to gravity as

M :=

1

2

_

d

3

x

C(x)

2

_

det q(x)

, (75)

where C(x) is the Hamiltonian constraint in (50). After solving the Gaussian constraint, one

gets the master constraint algebra as a Lie algebra:

+(

N), +(

) = +([

N,

N

]),

+(

N), M = 0,

M, M = 0, (76)

where the subalgebra of dieomorphism constraints forms an ideal. So it is possible to dene a

corresponding master constraint operator on 1

Di f f

. In the following, the positivity and the dif-

feomorphism invariance of M will be working together properly and provide us with powerful

functional analytic tools in the quantization procedure.

The regulated version of the master constraint can be expressed via a point-splitting strategy

as:

M

:=

1

2

_

d

3

y

_

d

3

x

(x y)

C(y)

_

V

U

y

C(x)

_

V

U

x

. (77)

Introducing a partition / of the 3-manifold into cells C, we have an operator

H

C

acting on

any spin-scalar-network state T

s,c

via a family of state-dependent triangulation T(),

C

T

s,c

=

vV((s,c))

C

(v)

E(v)

v()=v

h

,

GR,v

T

s,c

+

vV((s,c))

C

(v)

E(v)

[64

4

M

9

4

i j

( w

,v

,i

)

w

,v

, j

+ 8

4

16

81

M

()

6

( w

,v

Kin

)

w

,v

Kin

]T

s,c

, (78)

where

h

,

GR,v

=

16

3i

2

i jk

Tr(

A(

i j

())

1

A(s

k

())

1

[

A(s

k

()),

_

V

U

v

])

+ (1 +

2

)

8

2

3i

3

i jk

Tr(

A(s

i

())

1

[

A(s

i

()),

K

A(s

j

())

1

[

A(s

j

()),

K

]

A(s

k

())

1

[

A(s

k

()),

_

V

U

v

]),

w

,v

,i

=

i

v()=v

lmn

1

U(v, )

1

[

U(t(s

l

()), )

U(v, )]

Tr(

i

A(s

m

())

1

[

A(s

m

()),

V

5/8

U

v

]

A(s

n

())

1

[

A(s

n

()),

V

5/8

U

v

]),

64

w

,v

Kin

=

1

(i)

3

v()=v

(v)

lmn

Tr(

A(s

l

())

1

[

A(s

l

()),

V

5/12

U

v

]

A(s

m

())

1

[

A(s

m

()),

V

5/12

U

v

]

A(s

n

())

1

[

A(s

n

()),

V

5/12

U

v

]). (79)

Hence the action of

H

C

on a cylindrical function f

i j

() with

1

2

-representation

and points at t(s

i

()) with representation constant with respect to each vertex v() of . Thus,

for each > 0,

H

C

is a S U(2) gauge invariant and dieomorphism covariant operator dened

on Cyl(7/g) Cyl(1). The limit operator

H

C

is densely dened on 1

Kin

by the uniform

Rovelli-Smolin topology. And the same result holds for the adjoint operator (

H

C

)

M, on 1

Di f f

can be dened by:

(

M

Di f f

)[T

s,c

] := lim

/;,

Di f f

[

C/

1

2

H

C

(

H

C

)

T

s,c

]. (80)

Since

H

C

(

H

C

)

T

s,c

is a nite linear combination of spin-scalar-network functions on an graph

with skeleton , the value of (

M

Di f f

)[T

s,c

] is nite for a given

Di f f

that is a nite linear

combination of T

[s,c]

. So

M

Di f f

is in the algebraic dual of the space of cylindrical functions.

Moreover, we can show that it is dieomorphism invariant. For any dieomorphism transfor-

mation ,

(

U

M

Di f f

)[ f

] = lim

/;,

Di f f

[

C/

1

2

H

C

(

H

C

)

]

= lim

/;,

Di f f

[

U

C/

1

2

H

1

()

1

(C)

(

H

1

(

1

(C)

)

]

= lim

/;,

Di f f

[

C/

1

2

H

C

(

H

C

)

], (81)

for any cylindrical function f

, where in the last step, we used the fact that the dieomorphism

transformation leaves the partition invariant in the limit / and relabel (C) to be C. So

we have the result

(

U

M

Di f f

)[ f

] = (

M

Di f f

)[ f

]. (82)

So given a dieomorphism invariant spin-scalar-network state T

[s,c]

, the result state

MT

[s,c]

must

be a dieomorphisminvariant element in the algebraic dual of Cyl(7/g)Cyl(1), which means

that

MT

[s,c]

=

[s

1

,c

1

]

c

[s

1

,c

1

]

T

[s

1

,c

1

]

,

then

lim

/;,

0

T

[s,c]

[

C/

1

2

H

C

(

H

C

)

T

s

2

,c

2

] =

[s

1

,c

1

]

c

[s

1

,c

1

]

T

[s

1

,c

1

]

[T

s

2

,c

2

],

65

where the cylindrical function

_

C/

1

2

H

C

(

H

C

)

T

s

2

,c

2

is a nite linear combination of spin-scalar-

network functions on some graphs

with the same skeleton of (s

2

, c

2

) up to nite number of

arcs and vertices. Hence xing the dieomorphism equivalence class [s, c], only for spin-scalar-

networks s

2

, c

2

lying in a nite number of dieomorphism equivalence class on the left hand

side of the last equation is non-zero. So there are also only nite number of classes [s

1

, c

1

] in

the right hand side such that c

[s

1

,c

1

]

is non-zero. As a result,

MT

[s,c]

is a nite linear combination

of dieomorphism invariant spin-network states so lies in the Hilbert space of dieomorphism

invariant states 1

Di f f

for any [s, c]. And

M is densely dened on 1

Di f f

.

We now compute the matrix elements of

M. Given two dieomorphism invariant spin-

scalar-network functions T

[s

1

,c

1

]

and T

[s

2

,c

2

]

, the matrix element of

M is calculated as

< T

[s

1

,c

1

]

MT

[s

2

,c

2

]

>

Di f f

= (

MT

[s

2

,c

2

]

)[T

s

1

,c

1

[s

1

,c

1

]

]

= lim

/;,

C/

1

2

T

[s

2

,c

2

]

[

H

C

(

H

C

)

T

s

1

,c

1

[s

1

,c

1

]

]

= lim

/;,

C/

1

2

1

n

(s

2

,c

2

)

Di f f /Di f f

(s

2

,c

2

)

GS

(s

2

,c

2

)

<

U

T

s

2

,c

2

[s

2

,c

2

]

C

(

H

C

)

T

s

1

,c

2

[s

1

,c

1

]

>

Kin

=

s,c

lim

/;,

C/

1

2

1

n

(s

2

,c

2

)

Di f f /Di f f

(s

2

,c

2

)

GS

(s

2

,c

2

)

<

U

T

s

2

,c

2

[s

2

,c

2

]

C

T

s,c

>

Kin

<

s,c

(

H

C

)

T

s

1

,c

1

[s

1

,c

1

]

>

Kin

=

[s,c]

vV((s,c[s,c]))

1

2

lim

,

0

T

[s

2

,c

2

]

[

H

v

T

s,c[s,c]

]

s,c[s,c]

< T

s,c

(

H

v

)

T

s

1

,c

1

[s

1

,c

1

]

>

Kin

, (83)

where Di f f

denotes

the graph symmetry quotient group Di f f

/TDi f f

where TDi f f

is trivial on the graph , and n

resolution of identity trick in the fourth step. Since only a nite number of terms in the sum

over spin-scalar-networks (s, c), cells C /, and dieomorphism transformations are non-

zero respectively, we can interchange the sums and the limit. In the fth step, we take the

limit C v and split the sum

_

s,c

into

_

[s,c]

_

s,c[s,c]

, where [s, c] denotes the dieomorphism

equivalence class associated with (s, c). Here we also use the fact that, given (s, c) and (s

, c

)

which are dierent up to a dieomorphism transformation, there is always a dieomorphism

transforming the graph associated with

H

v,(s,c)

T

s,c

(v (s, c)) to that of

H

(s

,c

)

T

s

,c

(v

(s

, c

)) with (v) = v

, hence T

[s

2

,c

2

]

[

H

v,(s,c)

T

s,c[s,c]

] is constant for dierent (s, c) [s, c].

Since the term

_

s,c[s,c]

< T

s,c

(

H

v

)

T

s

1

,c

1

[s

1

,c

1

]

>

Kin

is independent of the parameter

, one

can see that by xing a family of arbitrary state-dependent triangulations T(

),

s,c[s,c]

< T

s,c

(

H

v

)

T

s

1

,c

1

[s

1

,c

1

]

>

Kin

66

=

< U

T

s,c

(

H

v

)

T

s

1

,c

1

[s

1

,c

1

]

>

Kin

=

<

H

v

U

T

s,c

T

s

1

,c

1

[s

1

,c

1

]

>

Kin

=

< U

1

(

1

(v)

T

s,c

T

s

1

,c

1

[s

1

,c

1

]

>

Kin

= T

[s

1

,c

1

]

[

H

1

(

)

vV((s,c))

T

s,c

], (84)

where are the dieomorphismtransformations spanning the dieomorphismequivalence class

[s, c]. Note that the kinematical inner product in the above sum is non-vanishing if and only

if ((s, c))) coincides with the extended graph obtained from certain skeleton (s

1

, c

1

) by the

action of (

H

v

)

1

(

tetrahedrons added by the action coincides with the scale of certain tetrahedrons in (s, c) and

1

(v) is a vertex in (s, c). Then we can express the matrix elements (83) as:

< T

[s

1

,c

1

]

MT

[s

2

,c

2

]

>

Di f f

=

[s,c]

vV((s,c[s,c]))

1

2

lim

,

0

T

[s

2

,c

2

]

[

H

v

T

s,c[s,c]

]T

[s

1

,c

1

]

[

H

v

T

s,c[s,c]

]

=

[s,c]

vV((s,c[s,c]))

1

2

(

H

v

T

[s

2

,c

2

]

)[T

s,c[s,c]

](

H

v

T

[s

1

,c

1

]

)[T

s,c[s,c]

]. (85)

From Eq.(85) and the result that the master constraint operator

M is densely dened on 1

Di f f

,

it is obvious that

M is a positive and symmetric operator on 1

Di f f

. Hence, it is associated with

a unique self-adjoint operator

M, called the Friedrichs extension of

M. We relabel

M to be

M on 1

Di f f

corresponding to the master constraint (75). It is then possible to obtain the physical Hilbert

space of the coupled system by the direct integral decomposition of 1

Di f f

with respect to

M.

Note that the quantum constraint algebra can be easily checked to be anomaly free. Eq.(82)

assures that the master constraint operator commutes with nite dieomorphism transforma-

tions, i.e.,

[

M,

U

] = 0. (86)

Also it is obvious that the master constraint operator commutes with itself,

[

M,

M] = 0. (87)

So the quantum constraint algebra is precisely consistent with the classical constraint algebra

(76) in this sense. As a result, the diculty of the original Hamiltonian constraint algebra can

be avoided by introducing the master constraint algebra, due to the Lie algebra structure of the

latter.

67

6 The Semiclassical Limit of Quantum Dynamics

As shown in previous chapters, both the Hamiltonian constraint operator

S(N) and the master

constraint operator

M can be well dened in the framework of loop quantum gravity. However,

since the Hilbert spaces 1

kin

and 1

Di f f

, the operators

S(N) and

M are constructed in such

ways that are drastically dierent from usual quantum eld theory, one has to check whether

the constraint operators and the corresponding algebras have correct semiclassical limits with

respect to suitable semiclassical states.

6.1 The Construction of Coherent States

In order to nd the proper semiclassical states and check the classical limit of the theory, the

idea of a non-normalizable coherent state dened by a generalized Laplace operator and its

heat kernel was introduced for the rst time in [26]. Recently, kinematical coherent states were

constructed in two dierent approaches. One leads to the so-called complexier coherent states

proposed by Thiemann et al [142][143][144][145]. The other was proposed by Varadarajan

[152][153][154] and further developed by Ashtekar et al [19][15].

The complexier approach is motivated by the coherent state construction for compact Lie

groups [86]. One begins with a positive function C (complexier) on the classical phase space

and arrives at a coherent state

m

, which more possibly belongs to the dual space Cyl

rather

than 1

kin

. However, one may consider the so-called cut-o state of

m

with respect to a

nite graph as a graph-dependent coherent state in 1

kin

[146]. By construction, the coherent

state

m

is an eigenstate of an annihilation operator coming also from the complexier C and

hence has the desired semiclassical properties [143][144]. We now sketch the basic idea of

its construction. Given the Hilbert space 1 for a dynamical system with constraints and a

subalgerba of observables S in the space [(1) of linear operators on 1, the semiclassical

states with respect to S are dened in Denition 3.1.5. Kinematical coherent states

m

mA

are semiclassical states which in addition satisfy the annihilation operator property [142][146],

namely there exists a certain non-self-adjoint operator z = a + i

b with a,

b S and a certain

squeezing parameter , such that

z

m

= z(m)

m

. (88)

Note that Eq.(88) implies that the minimal uncertainty relation is saturated for the pair of ele-

ments ( a,

b), i.e.,

m

([ a

m

( a)]

2

) =

m

([

b

m

(

b)]

2

) =

1

2

m

([ a,

b]). (89)

Note also that coherent states are usually required to satisfy the additional peakedness property,

namely for any m A the overlap function <

m

,

m

> is concentrated in a phase volume

1

2

m

([ q, p]), where q is a conguration operator and p a momentum operator. So the central

element in the construction is to dene a suitable annihilation operator z in analogy with the

simplest case of harmonic oscillator. A powerful tool named as complexier is introduced in

Ref.[142] to dene a meaningful z operator which can give rise to kinematical coherent states

for a general quantum system.

68

Denition 6.1.1: Given a phase space A = T

coordinates q and momentum coordinates p, a complexier, C, is a positive smooth function on

A, such that

(1) C/ is dimensionless;

(2) lim

p

C(m)

p

= for some suitable norm on the space of the momentum;

(3) Certain complex coordinates (z(m), z(m)) of Acan be constructed from C.

Given a well-dened complexier C on phase space A, the programme for constructing co-

herent states associated with C can be carried out as the following.

Complex polarization

The condition (3) in Denition 7.3.1 implies that the complex coordinate z(m) of A can

be constructed via

z(m) :=

n=0

i

n

n!

q, C

(n)

(m), (90)

where the multiple Poisson bracket is inductively dened by q, C

(0)

= q, q, C

(n)

=

q, C

(n1)

, C. One will see that z(m) can be regarded as the classical version of an

annihilation operator.

Dening the annihilation operator

After the quantization procedure, a Hilbert space 1 = L

2

((, d) with a suitable measure

d on a suitable conguration space ( can be constructed. It is reasonable to assume that

C can be dened as a positive self-adjoint operator

C on 1. Then a corresponding oper-

ator z can be dened by transforming the Poisson brackets in Eq.(90) into commutators,

i.e.,

z :=

n=0

i

n

n!

1

(i)

n

[ q,

C]

(n)

= e

C/

qe

C/

, (91)

which is called as an annihilation operator.

Constructing coherent states

Let

q

(q) be the -distribution on ( with respect to the measure d. Since

C is assumed

to be positive and self-adjoint, the conditions (1) and (2) in Denition 7.3.1 imply that

e

C/

is a well-dened smoothening operator. So it is quite possible that the heat kernel

evolution of the -distribution, e

C/

q

(q), is a square integrable function in 1, which is

even analytic. Then one may analytically extend the variable q

in e

C/

q

(q) to complex

values z(m) and obtain a class of states

m

as

m

(q) := [e

C/

q

(q)]

q

z(m)

, (92)

such that one has

z

m

(q) := [e

C/

q

q

(q)]

q

z(m)

= [q

C/

q

(q)]

q

z(m)

= z(m)

m

(q). (93)

69

Hence

m

is automatically an eigenstate of the annihilation operator z. So it is natural to

dene coherent states

m

(q) by normalizing

m

(q).

One may check that all the coherent state properties usually required are likely to be satised

by the above complexier coherent states

m

(q) [146]. As a simple example, in the case of one-

dimensional harmonic oscillator with Hamiltonian H =

1

2

(

p

2

2m

+

1

2

m

2

q

2

), one may choose the

complexier C = p

2

/(2m). It is straightforward to check that the coherent state constructed

by the above procedure coincides with the usual harmonic oscillator coherent state up to a phase

[146]. So the complexier coherent state can be considered as a suitable generalization of the

concept of usual harmonic oscillator coherent state.

The complexifer approach can be used to construct kinematical coherent states in loop quan-

tum gravity. Given a suitable complexier C, for each analytic path e one can dene

A

C

(e) :=

n=0

i

n

n!

A(e), C

(n)

, (94)

where A(e) S U(2) is assigned to e. As the complexier C is assumed to give rise to a

positive self-adjoint operator

C on the kinematical Hilbert space 1

kin

, one further supposes that

C/T

s

=

s

T

s

, where is a so-called classicality parameter, T

s

(A)

s

form a basis in 1

kin

and are analytic in A 7. Moreover the -distribution on the quantum conguration space 7

can be formally expressed as

A

(A) =

_

s

T

s

(A

)T

s

(A). Thus by applying Eq.(92) one obtains

coherent states

A

C

(A) = (e

C/

)

A

(A)

A

A

C =

s

e

s

T

s

(A

C

)T

s

(A). (95)

However, since there are an uncountably innite number of terms in the expression (95), the

norm of

A

C

(A) would in general be divergent. So

A

C

(A) is generally not an element of 1

kin

but rather a distribution on a dense subset of 1

kin

. In order to test the semiclassical limit of

quantum geometric operators on 1

kin

, one may further consider the cut-o state of

A

C

(A)

with respect to a nite graph as a graph-dependent coherent state in 1

kin

[146]. So the

key input in the construction is to choose a suitable complexifer. There are vast possibilities

of choice. For example, a candidate complexier C is considered in Ref.[148] such that the

corresponding operator acts on cylindrical functions f

by

(

C/) f

=

1

2

(

eE()

l(e)

J

2

e

) f

, (96)

where

J

2

e

is the Casimir operator dened by Eq.(24) associated to the edge e, the positive num-

bers l(e) satisfying l(e e

) = l(e) + l(e

) and l(e

1

) = l(e) serves as a classicalization parameter.

Then it can be shown from Eq.(94) that A

C

(e) is an element of S L(2, C). So the classical inter-

pretation of the annihilation operators is simply the generalized complex S U(2) connections. It

has been shown in Refs. [143] and [144] that the cut-o state of the corresponding coherent

state,

A

C

,

(A) =

A

C

,

(A)/

A

C

,

(A), (97)

70

with

A

C

,

(A) :=

s,(s)=

e

1

2

_

eE((s))

l(e) j

e

( j

e

+1)

T

s

(A

C

)T

s

(A). (98)

has desired semiclassical properties in testing the kinematical operators (e.g. holonomy and

ux). But unfortunately, these cut-o coherent states cannot be directly used to test the semi-

classical limit of the Hamiltonian constraint operator

S(N), since

S(N) is graph-changing so that

its expectation values with respect to these cut-o states are always zero! So further work in this

approach is expected in order to overcome the diculty. Anyway, the complexier approach

provides a clean construction mechanism and manageable calculation method for semiclassical

analysis in loop quantum gravity.

We now turn to the second approach. As we have seen, loop quantum gravity is based on

quantum geometry, where the fundamental excitations are one-dimensional polymer-like. On

the other hand, low energy physics is based on quantum eld theories which are constructed

in a at spacetime continuum. The fundamental excitations of these elds are 3-dimensional,

typically representing wavy undulations on the background Minkowskian geometry. The core

strategy in this approach is then to relate the polymer excitations of quantum geometry to Fock

states used in low energy physics and to locate Minkowski Fock states in the background inde-

pendent framework. Since the quantum Maxwell eld can be constructed in both Fock repre-

sentation and polymer-like representation, one rst gains insights from the comparison between

the two representations, then generalizes the method to quantum geometry. A Laplacian oper-

ator can be dened on 1

kin

[26][19], from which one may dene a candidate coherent state

0

,

also in Cyl

kinematical operators, one considers the so-called shadow state of

0

, which is the restriction

of

0

to a given nite graph. However, the construction of shadow states is subtly dierent from

that of cut-o states.

We will only describe the simple case of the Maxwell eld to illustrate the ideas of the

construction [152][153][20]. Following the quantum geometry strategy discussed in Sec.4, the

quantum conguration space A for the polymer representation of the U(1) gauge theory can be

similarly constructed. A generalized connection A A assigns each oriented analytic edge in

an element of U(1). The space A carries a dieomorphism and gauge invariant measure

0

induced by the Haar measure on U(1), which gives rise to the Hilbert space, 1

0

:= L

2

(A, d

0

),

of polymer states. The basic operators are holonomy operators

A(e) labeled by one-dimensional

edges e, which act on cylindrical functions by multiplication, and smeared electric eld opera-

tors

E(g) for suitable test one-forms g on , which are self-adjoint. Note that, since the gauge

group U(1) is Abelian, it is more convenient to smear the electric elds in 3 dimensions [20].

The eigenstates of

E(g), so-called ux network states A

,n

, provide an orthonormal basis in 1

0

,

which are dened for any nite graph with N edges as:

A

,n

(A) := [A(e

1

)]

n

1

[A(e

2

)]

n

2

[A(e

N

)]

n

N

, (99)

where n (n

1

, , n

N

) assigns an integer n

I

to each edge e

I

. The action of

E(g) on the ux

network states reads

E(g) A

,n

= (

I

n

I

_

e

I

g)A

,n

. (100)

71

In this polymer-like representation, cylindrical functions are the nite linear combinations of

ux network states and span a dense subspace of 1

0

. Denote by Cyl the set of cylindrical

functions and by Cyl

0

Cyl

in analogy

with the case of loop quantum gravity.

The Schr odinger or Fock representation of the Maxwell eld, on the other hand, depends on

the Minkowski background metric. Here the Hilbert space is given by 1

F

= L

2

(S

, d

F

), where

S

F

is the Gaussian measure. The

basic operators are connections

A( f ) smeared in 3 dimensions with suitable vector densities

f and smeared electric elds

E(g). But

A(e) fail to be well dened. To resolve this tension

between the two representations, one proceeds as follows. Let x be the Cartesian coordinates of

a point in = R

3

. Introduce a test function by using the Euclidean background metric on R

3

,

f

r

(x) =

1

(2)

3/2

r

3

exp(x

2

/2r

2

), (101)

which approximates the Dirac delta function for small r. The Gaussian smeared form factor for

an edge e is dened as

X

a

(e,r)

(x) :=

_

e

ds f

r

(e(s) x) e

a

. (102)

Then one can dene a smeared holonomy for e by

A

(r)

(e) := exp[i

_

R

3

X

a

(e,r)

(x)A

a

(x)], (103)

where A

a

(x) is the U(1) connection one-form of the Maxwell eld on . Similarly one can

dene Gaussian smeared electric elds by

E

(r)

(g) :=

_

R

3

g

a

(x)

_

R

3

f

r

(y x)E

a

(y). (104)

In this way one obtains two Poission bracket algebras. One is formed by smeared holonomies

and electric elds with

A

(r)

(e), A

(r)

(e

) = 0 = E(g), E(g

) (105)

A

(r)

(e), E(g) = i(

_

R

3

X

a

(e,r)

g

a

) A

(r)

(e).

The other is formed by unsmeared holonomies and Gaussian smeared electric elds with

A(e), A(e

) = 0 = E

(r)

(g), E

(r)

(g

) (106)

A(e), E

(r)

(g) = i(

_

R

3

X

a

(e,r)

g

a

) A(e).

Obviously, there is an isomorphism between them,

I

r

: (A

(r)

(e), E(g)) (A(e), E

(r)

(g)). (107)

72

Using the isomorphism I

r

, one can pass back and forth between the polymer and the Fock

representations. Specically, the image of the Fock vacuum can be shown to be the following

element of Cyl

[152][153],

(V =

,n

exp[

I J

G

I J

n

I

n

J

] (A

,n

, (108)

where (A

,n

Cyl

,n

) to one and every other ux net-

work functions to zero. While the states (A

,n

do not have any knowledge of the underlying

Minkowskian geometry, this information is coded in the matrix G

I J

associated with the edges

of the graph , given by [20]

G

I J

=

_

e

I

dt e

a

I

(t)

_

e

J

dt

e

J

b

(t

)

_

d

3

x

ab

(x) [ f

r

(x e

I

(t))

1

2

f (x, e

J

(t

))], (109)

where

ab

is the at Euclidean metric and its Laplacian. Therefore, one can single out the Fock

vacuum state directly in the polymer representation by invoking Poincar e invariance without

any reference to the Fock space. Similarly, one can directly locate in Cyl

as the eigenstates of the exponentiated annihilation operators. Since Cyl

inner product, one uses the notion of shadow states to do semiclassical analysis in the polymer

representation. From Eq.(108), the action of the Fock vacuum (V on A

,n

reads

(VA

,n

) =

_

A

d

0

A

,n

, (110)

where the state V

V

(A) =

n

exp[

I J

G

I J

n

I

n

J

] A

,n

(A). (111)

Thus for any cylindrical functions

associated with ,

(V

) = (V

), (112)

where the inner product in the right hand side is taken in 1

. Hence V

shadows of (V on the graphs . The set of all shadows captures the full information in (V.

By analyzing shadows on suciently rened graphs, one can introduce criteria to test if a given

element of Cyl

represents a semi-classical state [20]. It turns out that the state (V does satisfy

this criterion and hence can be regarded as semi-classical in the polymer representation.

The mathematical and conceptual tools gained from simple models like the Maxwell elds

are currently being used to construct semiclassical states of quantum geometry. A candidate

kinematical coherent state corresponding to the Minkowski spacetime has been proposed by

Ashtekar and Lewandowki in the light of a Laplacian operator [19][20]. However, the detailed

structure of this coherent state is yet to be analyzed and there is no a priori guarantee that it is

indeed a semiclassical state.

One may nd comparisons of the two approaches from both sides [147][20]. It turns out that

Varadarajans Laplacian coherent state for the polymer Maxwell eld can also be derived from

73

Thiemanns complexier method. However, one cannot nd a complexier to get the coherent

state proposed by Ashtekar et al. for loop quantum gravity. Both approaches have their own

virtues and need further developments. The complexier approach provides a clear construc-

tion mechanism and manageable calculation method, while the Laplacian operator approach is

related closely with the well-known Fock vacuum state. One may also expect that a judicious

combination of the two approaches may lead to signicant progress in the semiclassical analysis

of loop quantum gravity.

6.2 Algebraic Quantum Gravity Approach

As we have shown in the last subsection, although Thiemanns complexier coherent state has

a clear calculable mechanism and correct semi-classical properties in testing kinematical opera-

tors, it fails to be a qualied semi-classical state for the quantum dynamics since the semiclassi-

cal limit of the Hamiltonian constraint operator

S(N) or master constraint operator

M is clearly

not correct, both

S(N) and

M are graph-changing so that their expectation values with respect

to these cut-o coherent states are always zero. So a possible way to avoid such a problem is to

dene a non-graph-changing version of Hamiltonian constraint operator or similarly, a master

constraint operator. However, such a modication is hard to make in the framework of loop

quantum gravity since the action of Hamiltonian constraint operator always adds several arcs

on certain graphs. But if the framework of loop quantum gravity is suitably modied then it

turns out that a version of non-graph-changing Hamiltonian constraint operator can be proposed

and the semi-classical analysis for the quantum dynamics can be carried out with the complex-

ier coherent states dened previously. Such a modication is recently made by Thiemann in

[79][80][81] and is called algebraic quantum gravity (AQG) approach. We describe it briey in

what follows.

Algebraic quantum gravity is a new approach to canonical quantum gravity suggested by

loop quantum gravity. But in contrast to loop quantum gravity, the quantum kinematics of al-

gebraic quantum gravity is determined by an abstract -algebra generated by a countable set

of elementary operators labeled by a single algebraic graph with countably innite number of

edges, while in loop quantum gravity the elementary operators are labeled by a collection of

embedded graphs with nite number of edges. Thus one can expect that in algebraic quantum

gravity, we lose the information of the topological and dierential structure of the manifold in

all the quantization procedure before we do semi-classical analysis. Hence the quantum the-

ory will be of course independent of the topology and dierential structure of the manifold but

based only on an algebraic graph, which only contains the information of the number of vertices

and their oriented valence.

Denition 6.2.1: An oriented algebraic graph is an abstract graph specied by its adjacency

matrix , which is an N N matrix. One of its entries

I J

stand for the number of edges that

start at vertex I and end at vertex J. The valence of the vertex I is given by v

I

=

_

J

(

I J

+

JI

).

We also use V() and E() to denote the sets of vertices and edges respectively.

In our quantization procedure, we x a specic cubic algebraic graph with a countably in-

nite number of edges N = and the valence of each vertex v

I

= 2 dim(). Such a specic

74

choice, although it detracts from the generality of the theory, is practically sucient for our use

in the semiclassical analysis.

Given the algebraic graph , we dene a quantum -algebra by associating with each edge

e an element A(e) of a compact, connected, semisimple Lie group G and an element E

j

(e) take

value in its Lie algebra g. These elements are subject to the commutation relations

[

A(e),

A(e

)] = 0,

[

E

j

(e),

A(e

)] = iQ

2

e,e

j

/2

A(e),

[

E

j

(e),

A(e

)] = iQ

2

e,e

f

jkl

E

l

(e

),

and -relations

A(e)

= [

A(e)

1

]

T

,

E

j

(e)

=

E

j

(e),

where Q stands for the coupling constant,

j

is the generators in the Lie algebra g and f

jkl

is the

structure constant of g. We denote the abstract quantum -algebra generated by above elements

and relations by A.

A natural representation of A is the innite tensor product Hilbert space 1

=

e

1

e

where

1

e

= L

2

(G, d

H

)[145], whose element is denoted by

f

e

f

e

. Two elements

f

and

f

in

1

_

e

< f

e

, f

e

>

1

e

1 converges. We denote by [ f ] the

strongly equivalence class containing

f

. It turns out that two elements in 1

are orthogonal

if they lie in dierent strongly equivalence classes. Hence the innite tensor Hilbert space 1

[ f ]

which are the closure

of strongly equivalence classes [ f ]. Furthermore, although each sector 1

[ f ]

is separable and

has a natural Fock space structure, the whole Hilbert space 1

uncountably innite number of strongly equivalence classes in it. Our basic elements in the

quantum algebra are represented on 1

in an obvious way

A(e)

f

:= [A(e) f

e

] [

e

e

f

e

],

E

j

(e)

f

:= [iQ

2

X

e

j

f

e

] [

e

e

f

e

].

As one might have expected, all these operators are densely dened and E

j

(e) is essentially

self-adjoint. Given a vertex v V(), the volume operator can be constructed by using the

operators we just dened

V

v

:=

3

p

_

1

48

e

1

e

2

e

3

=v

v

(e

1

, e

2

, e

3

)

i jk

E

i

(e

1

)

E

j

(e

2

)

E

k

(e

3

),

where the values of

v

(e

1

, e

2

, e

3

) should be assigned once for all for each vertex. When we

embed the algebraic graph into some manifold, the embedding should be consistent with the

assigned values of

v

(e

1

, e

2

, e

3

).

Then we discuss the quantum dynamics. By the regularization methods frequently used in

the last two sections, the half densitized constraints can be quantized to be composite operators

as we list below.

75

Gauss constraint

G

j

(v) :=

Q

(1/2)

v

e at v

E

j

(e);

Spatial dieomorphism constraint

D

j

(v) :=

1

E(v)

e

1

e

2

e

3

=v

v

(e

1

, e

2

, e

3

)

L(v, e

1

, e

2

)

L(v,e

1

,e

2

)

Tr(

j

[

A()

A()

1

]

A(e

3

)[

A(e

3

)

1

,

_

V

v

]);

Euclidean Hamiltonian constraint (up to an overall factor)

H

(r)

E

(v) :=

1

E(v)

e

1

e

2

e

3

=v

v

(e

1

, e

2

, e

3

)

L(v, e

1

, e

2

)

L(v,e

1

,e

2

)

Tr([

A()

A()

1

]

A(e

3

)[

A(e

3

)

1

,

V

(r)

v

]);

Lorentzian Hamiltonian constraint (up to an overall factor)

T(v) :=

1

E(v)

e

1

e

2

e

3

=v

v

(e

1

, e

2

, e

3

)

Tr(

A(e

1

)[

A(e

1

)

1

, [

H

(1)

E

,

V]]

A(e

2

)[

A(e

2

)

1

, [

A(e

3

)

1

, [

H

(1)

E

,

V]]

A(e

3

)[

A(e

3

)

1

,

_

V

v

]),

H(v) =

H

(1/2)

E

(v) +

T(v); (113)

where

V :=

_

v

V

v

,

H

(1)

E

:=

_

v

H

(1)

E

(v) and

Q

(r)

v

:=

1

E(v)

e

1

e

2

e

3

=v

v

(e

1

, e

2

, e

3

)

Tr(

A(e

1

)[

A(e

1

)

1

,

V

(r)

v

]

A(e

2

)[

A(e

2

)

1

,

V

(r)

v

]

A(e

3

)[

A(e

3

)

1

,

V

(r)

v

]).

L(v, e

1

, e

2

) denotes the set of minimal loops starting at v along e

1

and ending at v along e

1

2

. And

a loop L(v, e

1

, e

2

) is said to be minimal provided that there is no other loop within sat-

isfying the same restrictions with fewer edges traversed. Note that since we only have a single

cubic algebraic graph, the dieomorphism constraint can only be implemented by dening the

operators corresponding to dieomorphism generators because a nite dieomorphism trans-

formation is not meaningful in our algebraic treatment unless the algebraic graph is embedded

in a manifold. As a result, the (extended) master constraint can be expressed as a quadratic

combination:

M :=

vV()

[

G

j

(v)

G

j

(v) +

D

j

(v)

D

j

(v) +

H(v)

H(v)].

76

It is trivial to see that all the above operators are non-graph-changing and embedding indepen-

dent because we have only worked on a single algebraic graph so far. However, when we test

the semiclassical limit of these operators, especially the master constraint operators, we should

specify an embedding map X which map a algebraic graph to be an embedded one. With this

specic embedding, we can see the correspondence between the classical algebra of elementary

observables and the quantum -algebra. We dene the holonomy and suitably modied ux by

A(e) := A(X(e)) := /exp(

_

X(e)

A),

E

j

(e) := 2Tr[

j

_

S

e

abc

dx

a

dx

b

A(

e

(x))E

c

(x)A(

e

(x))

1

],

where S

e

is a face which intersects the edge X(e) only at an interior point p

e

of both S

e

and X(e).

We choose a system of paths

e

(x)

x

for all x S

e

, such that

e

(x) starts at s(X(e)) along X(e)

until p

e

and then runs within S

e

until x. As one might expect, the quantum-algebra we dened

previously is just consistent with the classical Poisson algebra generated by these holonomis

and uxes:

A(e), A(e

) = 0,

E

j

(e), A(e

) = Q

2

e,e

j

/2A(e),

E

j

(e), A(e

) = Q

2

e,e

f

jkl

E

l

(e

).

Then we consider the coherent states. By employing the Laplacian complexier on each

edge

C

e

:=

1

2Q

2

a

2

e

E

j

(e)E

j

(e),

the coherent state is obtained as it was in the last section:

t

e

e;(A,E)

(A)

t

e

e;g(A,E)

(A(e)) =

dim()e

t

(g(A, E)A(e)),

where

2

p

/a

2

e

represents the classical-

ization parameter. The coherent state peaks at the complexied classical phase space point

g(A, E) :=

n=0

(i)

n

n!

C

e

, A(e)

n

= exp(iE(e)/a

2

e

)A(e),

note that the parameter a

e

is specied such that E(e)/a

2

e

is dimensionless. Hence the coherent

state on the whole graph is represented by an innite tensor product state:

t

A,E

(A) :=

eE()

t

e

e;(A,E)

(A)

t

e

e;(A,E)

(A)

.

The peakness, uctuation and other semiclassical properties of these states have been checked

in [143][144] in which the most important part is that

<

A,E

A(e)

A,E

>= A(e) <

A,E

E(e)

A,E

>= E(e)

77

up to terms which vanish faster than any power of t

e

as t

e

0. And the uctuations are small.

With the semiclassical state we just constructed, the expectation value of the above (ex-

tended) master constraint operator can be calculated and its semiclassical limit can be tested. In

the following, we summarize the result of the calculation. In [80], a semiclassical calculation

for the master constraint operator is carried out based on a cubic algebraic graph. The calcula-

tion makes use of a simplifying assumption: we substitute the gauge group for gravity S U(2)

by U(1)

3

. And the result of the calculation shows that in U(1)

3

case the (extended) master

constraint operator has correct semiclassical limit

lim

t0

<

t

m

M

t

m

>= M

cubic

[m] M[m] ( 0)

where m represents a phase space point and is the lattice parameter such that the lattice be-

come continuum as 0. In addition, it is shown that the next-to-leading order terms which

contribute to the uctuation of

M are nite.

Moreover, the calculation in [81] shows that the result of the exact non-Abelian calculation

matches precisely the results of the Abelian approximation, provided that we replace the clas-

sical U(1)

3

terms h

j

e

, p

e

j

j=1,2,3

by Tr(

j

A(e)), Tr(

j

E(e))

j=1,2,3

, which means that the theory

of algebraic quantum gravity admits a semiclassical limit whose innitesimal gauge symmetry

agrees with that of general relativity.

78

7 Conclusion and Discussion

As it was shown in the previous sections, loop quantum gravity oers a conceptually clear

and mathematically rigorous approach to quantize general relativity. In this approach, we are

seeking new physics deeply below the Planck scale. In the kinematical framework, a quantum

Riemannian geometry is established and some geometrical operators, e.g. area, volume, are

well-dened, and their spectrum are shown to be discrete, which means that the structure of the

space may be discrete below the Planck scale. Such a new phenomena sheds light on quantum

eld theory, lattice gauge theory and their renormalization. Moreover, the program in the quan-

tum dynamics of loop quantum gravity represents signicant progress in the research area of

quantum gravity. Before loop quantum gravity, the quantum Wheeler-DeWitt equation was only

a formal equation and from concrete calculations. However, in the framework of loop quantum

gravity, we already have a well-dened quantum Hamiltonian constraint operator which has

an explicit action on kinematical states, so that the quantum Wheeler-DeWitt equation is well-

dened in loop quantum gravity. On the other hand, the matter eld can also be quantized in

this framework and we show that the matter Hamiltonian is free of UV-divergence and dont

need a renormalization process. Furthermore, with the coupled matter eld, a matter coupled

Hamiltonian constraint operator is obtained so that the problem of time may be solved and, such

an idea is being translated into a new understanding of the early universe in the context of loop

quantum cosmology.

Although great progress has been made, as an unnished framework, loop quantum gravity

still has many issues to be solved in the future research. To conclude this thesis, we list some of

those in the following:

First of all, we dont have the complete solutions for either Hamiltonian constraint equa-

tion or master constraint equation. Thus one cannot explicitly construct the physical

Hilbert space for loop quantum gravity. So the quantum dynamics of gravity is essen-

tially unknown so far.

To make contact with experimental results, one should know the observables in the quan-

tum theory which have to be invariant under gauge transformation. However, some of the

Dirac observables that have been constructed involve an innite number of derivatives

and extremely hard to manage [150][61][62].

The semiclassical limit of loop quantum gravity is unknown so far, although a great deal

of progress has been made in the context of algebraic quantum gravity. And in alge-

braic quantum gravity, further research work is needed to show the uctuation of master

constraint operator should be small.

As it was shown at the end of section 4.3, the regularization process for the Hamiltonian

constraint operator is ambiguous and there is a list of free parameters. Thus it is also a

research project to remove as many ambiguities as possible. And some work has been

recently done in this direction [110].

The Immirzi parameter is another free parameter in the framework of loop quantum grav-

ity, which comes in with the classical formulation. In the classical theory, dierent values

79

of the Immirzi parameter label equivalent classical theories since they are connected by

canonical transformations. However, in quantum theory, it is problematic because the

representation with dierent Immirzi parameter are not unitarily equivalent.

The construction of loop quantum gravity crucially depends on the compactness of its

gauge group S U(2), which comes from an internal partial gauge xing. And it is ar-

gued that the internal Lorentz symmetry is broken in a non-natural way [131]. So it

seems to be better to switch back to the complex Ashtekar variables which are free

of the internal gauge xing and preserve the internal Lorentz symmetry and, will also

greatly simplify the Hamiltonian constraint. However, the price is that we should work

on non-compact gauge group S L(2, C), and there is no satisfactory quantization pro-

gramme for the S L(2, C)-gravity so far, although some work has been done in this di-

rection [71][105][106].

As it was shown in section 6.2, algebraic quantum gravity provides a clear way to make

many calculations accessible and the semiclassical analysis can be carried out in this

framework. And it seems that the construction of algebraic quantum gravity admits the

non-compact internal gauge group (or non-compact reduced conguration space) since

there is only one graph in quantization process. However, in this framework, the restric-

tion of the possible representation is so loose that there even exists a possible representa-

tion in which the spectrum of geometrical operators are continuous.

The transition amplitude calculation for loop quantum gravity is accessible in the so-

called spin foam model and depends on a unclear conjecture of GFT/spinfoam duality.

On the other hand, it is still not clear how to build a path-integral formulation and connect

with spin foam models from the canonical approach, and such a connection may give the

needed support for GFT/spinfoam duality.

We have constructed the dynamics of matter quantum eld theory on a quantum back-

ground in section 5. However, it is not clear how we can make the connection with the

ordinary quantum eld theory on curved spacetime. Moreover, it is still an problem how

to nd the non-perturbative correspondence of Hadamard states in perturbative quantum

eld theory in curved spacetime, although some hints of a connection may come out in

spin foam calculations [127][42].

80

References

[1] M. Arnsdorf and S. Gupta, Loop quantum gravity on non-compact spaces, Nucl. Phys. B

577, 529 (2000).

[2] A. Ashtekar, New variables for classical and quantum gravity, Phys. Rev. Lett. 57, 2244

(1986).

[3] A. Ashtekar, New Hamitonian formulation of general relativity, Phys. Rev. D 36, 1587

(1987).

[4] A. Ashtekar, Lectures on Non-perturbative Canonical Gravity, (World Scientic, Singa-

pore, 1991).

[5] A. Ashtekar, J. C. Baez, and K. Krasnov, Quantum geometry of isolated horizon and black

hole entropy, Adv. Theor. Math. Phys. 4, 1 (2000).

[6] A. Ashtekar, C. Beetle, and J. Lewandowski, Mechanics of rotating isoloted horizons,

Phys. Rev. D 64, 044016 (2001).

[7] A. Ashtekar, C. Beetle, and J. Lewandowski, Geometry of generic isolated horizons, Class.

Quantum Grav. 19, 1195 (2002).

[8] A. Ashtekar and M. Bojowald, Black hole evaporation: Aparadigm, Class. QuantumGrav.

22, 3349 (2005).

[9] A. Ashtekar and M. Bojowald, Quantum geometry and the Schwarzschild singularity,

Class. Quantum Grav. 23, 391 (2006).

[10] A. Ashtekar, M. Bojowald, and J. Lewandowski, Mathematical structure of loop quantum

cosmology, Adv. Theor. Math. Phys. 7, 233 (2003).

[11] A. Ashtekar, L. Bombelli, and A. Corichi, Semiclassical states for constrained systems,

Phys. Rev. D 72, 025008 (2005).

[12] A. Ashtekar, J. Engle, and C. van der Broeck, Quantum horizons and black hole entropy:

Inclusion of distortion and rotation, Class. Quantum Grav. 22, L27 (2005).

[13] A. Ashtekar, J. Engle, T. Powlowski, and C. van der Broeck, Mutipole moments of isolated

horizons, Class. Quantum Grav. 21, 2549 (2004).

[14] A. Ashtekar, S. Fairhurst, and B. Krishnan, Isolated horizons: Hamiltonian evolution and

the rst law, Phys. Rev. D 62, 104025 (2000).

[15] A. Ashtekar, S. Fairhurst, and J. Willis, Quantum gravity, shadow state and quantum me-

chanics, Class. Quantum Grav. 20, 1031 (2003).

[16] A. Ashtekar and C. J. Isham, Representation of the holonomy algebras of gravity and

non-Abelian gauge theories, Class. Quantum Grav. 9, 1433 (1992).

81

[17] A. Ashtekar and J. Lewandowski, Projective techniques and functional integration, J.

Math. Phys. 36, 2170 (1995).

[18] A. Ashtekar and J. Lewandowski, Dierential geometry on the space of connections via

graphs and projective limits, J. Geom. Phys. 17, 191 (1995).

[19] A. Ashtekar and J. Lewandowski, Relation between polymer like and Fock excitations,

Class. Quantum Grav. 18, L117 (2001).

[20] A. Ashtekar and J. Lewandowski, Background independent quantum gravity: A status

report, Class. Quantum Grav. 21, R53 (2004), (preprint: gr-qc/0404018).

[21] A. Ashtekar and J. Lewandowski, Quantum theory of geometry I: Area operators, Class.

Quantum Grav. 14, A55 (1997), (preprint: gr-qc/9602046).

[22] A. Ashtekar and J. Lewandowski, Quantum theory of geometry II: Volume operators, Adv.

Theor. Math. Phys. 1, 388 (1997), (preprint: gr-qc/9711031).

[23] A. Ashtekar, A. Corich, and J. A. Zapata, Quantum theory of geometry III: Non-

commutativity of Riemannian structures, Class. Quantum Grav. 15, 2955 (1998).

[24] A. Ashtekar, J. Lewandowski, D. Marolf, J. Mour ao, and T. Thiemann, A manifest gauge-

invariant approach to quantum theories of gauge elds, (preprint: hep-th/9408108).

[25] A. Ashtekar, J. Lewandowski, D. Marolf, J. Mour ao, and T. Thiemann, Quantization of

dieomorphism invariant theories of connections with local degrees of freedom, J. Math.

Phys. 36, 6456 (1995).

[26] A. Ashtekar, J. Lewandowski, D. Marolf, J. Mour ao, and T. Thiemann, Coherent state

transforms for spaces of connections, J. Funct. Anal. 135, 519 (1996).

[27] A. Ashtekar, J. Lewandowski, and H. Sahlmann, Polymer and Fock representations for a

scalar eld, Class. Quantum Grav. 20, L11 (2003).

[28] A. Ashtekar, T. Powlowski and P. Singh, Quantum nature of the big bang, Phys. Rev. Lett.

96, 141301 (2006).

[29] A. Ashtekar, T. Powlowski and P. Singh, Quantum nature of the big bang: improved dy-

namics, Phys. Rev. D 74, 084003(2006).

[30] A. Ashtekar, J. Romano, R. Tate, New variables for gravity: Inclusion of matter, Phys.

Rev. D 40, 2572 (1989).

[31] A. Ashtekar, C. Rovelli, and L. Smolin, Weaving a classical geometry with quantum

threads, Phys. Rev. Lett. 69, 237 (1992).

[32] J. C. Baez, Spin networks and quantum gravity, in The Interface of Knots and Physics, ed.

L. Kauman (Providence, RI: American Mathematical Society, 1996).

82

[33] J. C. Baez, An introduction to spin foam models of quantum gravity and BF theory, Lect.

Notes Phys. 543, 25 (2000).

[34] J. Barbero, Real Ashtekar variables for Lorentzian signature spacetimes, Phys. Rev. D 51,

5507 (1995).

[35] J. W. Bardeen, B. Carter, and S. W. Hawking, The four laws of black hole mechanics,

Commun. Math. Phys. 31, 161 (1973).

[36] J. W. Barrett and L. Crane, Relativistic spin networks and quantum gravity, J. Math. Phys.

39, 3296 (1998).

[37] J. W. Barrett and L. Crane, A Lorentzian signature model for quantum general relativity,

Class. Quantum Grav. 17, 3101 (2000).

[38] D. Di Bartolo, R. Gambini, J. Griego, and J. Pullin, Consistent quantization of general

relativity in the space of Vassiliev knot invariants, Phys. Rev. Lett. 84, 2314 (2000).

[39] D. Di Bartolo, R. Gambini, J. Griego, and J. Pullin, Canonical quantum gravity in the

Vassiliev knot invariants arena: I. Kinematical structure, Class. Quantum Grav. 17, 3211

(2000).

[40] D. Di Bartolo, R. Gambini, J. Griego, and J. Pullin, Canonical quantum gravity in the

Vassiliev knot invariants arena: II. Constraints, habitats and consistency, Class. Quantum

Grav. 17, 3239 (2000).

[41] J. D. Bekenstein, Black holes and entropy, Phys. Rev. D 7, 2333 (1973).

[42] E. Bianchi, L. Modesto, C. Rovelli, S. Speziale, Graviton propagator in loop quantum

gravity, Class.Quantum Grav. 23, 6989, (2006).

[43] M. Bojowald, Absence of singularity in loop quantum cosmology, Phys. Rev. Lett. 86,

5227 (2001), (preprint: gr-qc/0102069).

[44] M. Bojowald, Dynamical initial conditions in quantum cosmology, Phys. Rev. Lett. 87,

121301 (2001).

[45] M. Bojowald, Ination from quantum geometry, Phys. Rev. Lett. 89, 261301 (2002).

[46] M. Bojowald, Initial conditions for a universe, Gen. Rel. Grav. 35, 1877 (2003).

[47] M. Bojowald, Spherically symmetric quantum geometry: States and basic operators,

Class. Quantum Grav. 21, 3733 (2004).

[48] M. Bojowald, Nonsingular black holes and degrees of freedom in quantum gravity, Phys.

Rev. Lett. 95, 061301 (2005).

[49] M. Bojowald, Elements of loop quantum cosmology, preprint: gr-qc/0505057.

[50] M. Bojowald, Loop quantum cosmology, Living Rev. Relativity, 8, (2005), 11.

83

[51] M. Bojowald, R. Goswami, R. Maartens, and P. Singh, A black hole mass threshold from

non-singular quantum gravitational collapse, Phys. Rev. Lett. 95, 091302 (2005).

[52] M. Bojowald and H. A. Kastrup, Symmetry reduction for quantized dieomorphism in-

variant theories of connections, Class. Quantum Grav. 17, 3009 (2000).

[53] T. Br ocker and T. Dieck, Representations of Compact Lie Group, (Springer-Verlag, 1985).

[54] J. D. Brown, K. V. Kuchar, Dust as a standard of space and time in canonical quantum

gravity, Phys. Rev. D51, 5600 (1995), (preprint: gr-qc/9409001).

[55] S. Carlip, Quantum gravity: a progress report, Rept. Prog. Phys. 64, 885 (2001).

[56] M. Carmeli and S. Malin, Theory of Spinor: An Introduction, (World Scientic, 2000).

[57] Y. Choquet-Bruhat, C. Dewitt-Morette, M. Dillard-Bleick, Analysis, Manifold, and

Physics, (North-Holland Publishing Company, 1977).

[58] J. B. Conway, A Course in Functional Analysis, Second Edition, (Springer-Verlag, 1990).

[59] L. Crane, A. Perez, and C. Rovelli, Perturbative niteness in spin-foam quantum gravity,

Phys. Rev. Lett. 87, 181301 (2001).

[60] G. Date and G. M. Hossain, Genericity of ination in isotropic loop quantum cosmology,

Phys. Rev. Lett. 94, 011301 (2005).

[61] B. Dittrich, Partial and complete observable for Hamiltonian contrained system, (preprint:

gr-qc/0411013).

[62] B. Dittrich, Partial and complete observable for canonical general relativity, (preprint: gr-

qc/0507106).

[63] B. Dittrich and T. Thiemann, Testing the master constraint programme for loop quan-

tum gravity I. General framework, Class. Quantum Grav. 23, 1025 (2006), (preprint: gr-

qc/0411138).

[64] B. Dittrich and T. Thiemann, Testing the master constraint programme for loop quantum

gravity: II. Finite dimensional systems, Class. Quantum Grav. 23, 1067 (2006), (preprint:

gr-qc/0411139).

[65] B. Dittrich and T. Thiemann, Testing the master constraint programme for loop quan-

tum gravity: III. S L(2, R) models, Class. Quantum Grav. 23, 1089 (2006), (preprint: gr-

qc/0411140).

[66] B. Dittrich and T. Thiemann, Testing the master constraint programme for loop quantum

gravity: IV. Free eld theories, Class. Quantum Grav. 23, 1121 (2006), (preprint: gr-

qc/0411141).

84

[67] B. Dittrich and T. Thiemann, Testing the master constraint programme for loop quantum

gravity: V. Interecting eld theories, Class. Quantum Grav. 23, 1143 (2006), (preprint:

gr-qc/0411142).

[68] M. Domagala and J. Lewandowski, Black hole entropy from quantum gravity, Class.

Quantum Grav. 21, 5233 (2004).

[69] W. Fairbain and C. Rovelli, Separable Hilbert space in loop quantum gravity, J. Math.

Phys. 45, 2802 (2004).

[70] E. E. Flanagan and R. M. Wald, Does backreaction enforce the averaged null energy con-

dition in semiclassical gravity, Phys. Rev. D54, 6233 (1996).

[71] L. Freidel and E. Livine, Spin networks for non-Compact groups, J. Math. Phys. 44, 1322

(2003).

[72] R. Gambini, J. Lewandowski, D. Marolf, and J. Pullin, On the consistency of the constraint

algebra in spin network quantum gravity, Int. J. Mod. Phys. D 7, 97 (1998).

[73] R. Gambini and J. Pullin, Loops, Knots, Gauge Theories and Quantum Gravity, (Cam-

bridge University Press, Cambridge, 1996).

[74] R. Gambini and J. Pullin, Canonical quantization of general relativity in discrete space-

times, Phys. Rev. Lett. 90, 021301 (2003).

[75] R. Gambini and J. Pullin, Consistent discretization and loop quantum geometry, Phys.

Rev. Lett. 94, 101302 (2005).

[76] R. Gambini and J. Pullin, Discrete spacetime, Preprint: gr-qc/0505023.

[77] K. Giesel and T. Thiemann, Consistency check on volume and triad operator quantisation

in loop quantum gravity I, (preprint: gr-qc/0507036).

[78] K. Giesel and T. Thiemann, Consistency check on volume and triad operator quantisation

in loop quantum gravity II, (preprint: gr-qc/0507037).

[79] K. Giesel and T. Thiemann, Algebraic quantum gravity (AQG) I. Conceptual setup,

preprint: gr-qc/0607099.

[80] K. Giesel and T. Thiemann, Algebraic quantum gravity (AQG) II. Semiclassical analysis,

preprint: gr-qc/0607100.

[81] K. Giesel and T. Thiemann, Algebraic quantum gravity (AQG) III. Semiclassical pertur-

bation theory, preprint: gr-qc/0607101.

[82] R. Giles, Reconstruction of gauge potentials from Wilson loops, Phys. Rev. D 24, 2160

(1981).

[83] J. Glimm and A. Jae, Quantum Physics, (Springer-Verlag, New York, 1987).

85

[84] N. Guilini and D. Marolf, On the generality of rened algebraic quantization, Class. Quan-

tum Grav. 16, 2479, (1999).

[85] R. Haag, Local Quantum Physics, 2nd ed. (Springer-Verlag, Berlin, 1996).

[86] B. C. Hall, The Segal-Bergmann coherent state transformation for compact Lie groups, J.

Funct. Anal. 122, 103 (1994).

[87] M. Han, Y. Ma, Y. Ding, and L. Qin, Hamiltonian Analysis of n-dimensional Palatini

Gravity with Matter, Mod. Phys. Lett. A20 725 (2005), (preprint: gr-qc/0503024).

[88] M. Han and Y. Ma, Master constraint operators in loop quantum gravity, Phys. Lett. B635,

225 (2006), (preprint: gr-qc/0510014).

[89] M. Han and Y. Ma, Dynamics of a scalar eld in a polymer-like representation, Class.

Quantum Grav. 23, 2741 (2006).

[90] M. Han, W. Huang and Y. Ma, Fundamental structure of loop quantum gravity, preprint:

gr-qc/0509064.

[91] S. W. Hawking, and G. F. R. Ellis, The large scale structure of space-time, (Cambridge

University Press, 1973).

[92] S. Hollands, R. M. Wald, On the renormalization group in curved spacetime, Commun.

Math. Phys. 237 123 (2003).

[93] S. Holst, Barberos Hamiltonian derived from a generalized Hilbert-Palatini action, Phys.

Rev. D 53, 5966 (1996) (preprint: gr-qc/9511026).

[94] R. K. Kaul and P. Majumdar, Logarithmic corrections to the Bekenstein-Hawking entropy,

Phys. Rev. Lett. 84, 5255 (2000).

[95] J. Lewandowski and D. Marolf, Loop constraints: a habitate and their algebra, Int. J. Mod.

Phys. D 7, 299 (1998).

[96] J. Lewandowski, A. Okolow, H. Sahlmann, T. Thiemann, Uniqueness of dieomorphism

invariant states on holonomy-ux algebras, (preprint: gr-qc/0504147).

[97] C. Liang, Introductory dierential geometry and general relativity I, II, (Beijing Normal

University Press, 2000, in Chinese).

[98] Y. Ma, Quantization of static spacetimes, Phys. Rev. D 65, 064012 (2002).

[99] Y. Ma and Y. Ling, Classical geometry from a physical state in canonical quantum gravity,

Phys. Rev. D 62, 064030 (2000).

[100] Y. Ma and Y. Ling,

Q operator for canonical quantum gravity, Phys. Rev. D 62, 104021

(2000).

86

[101] L. Modesto, Disappearance of the black hole singularity in loop quantum gravity, Phys.

Rev. D 70, 124009 (2004).

[102] L. Modesto, The Kantowski-Sachs space-time in loop quantum gravity, preprint: gr-

qc/0411032.

[103] L. Modesto, Quantum gravitational collapse, (preprint: gr-qc/0504043).

[104] M. Nakahara, Geometry, Topology, and Physics, Second Edition (Institute of Physics

Publishing, 2003).

[105] A. Okolow, Hilbert space built over connections with a non-compact structure group,

Class. Quant. Grav. 22, 1329 (2005).

[106] A. Okolow, Quantization of dieomorphism invariant theories of connections with a non-

compact structure group - an example, (preprint: gr-qc/0605138).

[107] P. Peld` an, Actions for gravity, with generalizations: A review, Class. Quantum Grav. 11,

1087 (1994), (preprint: gr-qc/9305011).

[108] A. Perez, Finiteness of a spin foam model for Euclidean quantum general relativity, Nucl.

Phys. B 599, 427, (2001).

[109] A. Perez, Spin foam models for quantum gravity, Class. Quantum Grav. 20, R43 (2003).

[110] A. Perez, On the regularization ambiguities in loop quantum gravity, Phys.Rev. D 73,

044007 (2006).

[111] A. Perez and C. Rovelli, Spin foam model for Lorentzian general relativity, Phys. Rev. D

63, 041501 (2001).

[112] A. Perez and C. Rovelli, Physical eects of the Immirzi parameter, Phys.Rev. D 73,

044013 (2006).

[113] M. Reed, and B. Simon, Methods of Modern Mathematical Phycics I: Functional Analy-

sis, (Academic Press, 1973).

[114] M. Reed and B. Simon, Methods of Modern Mathematical Physics II: Fourer Analysis,

Self-adjointness, (Academic Press, 1975), Page 177.

[115] M. Reisenberger and C. Rovelli, Spacetime states and covariant quantum theory, Phys.

Rev. D 65, 125016 (2002).

[116] C. Rovelli, What is observable in classical and quantum gravity, Class. Quantum Grav.

8, 1895 (1991).

[117] C. Rovelli, Black hole entropy from loop quantum gravity, Phys. Rev. Lett. 14, 3288

(1996).

[118] C. Rovelli, Loop quantum gravity, Living Rev. Relativity 1 (1998), 1.

87

[119] C. Rovelli, A note on the foundation of relativistic mechanics. I: Relativistic observables

and relativistic states, (preprint: gr-qc/0111037).

[120] C. Rovelli, A note on the foundation of relativistic mechanics. II: Covariant Hamiltonian

general relativity, (preprint: gr-qc/0202079).

[121] C. Rovelli, Partial observables, Phys. Rev. D 65, 124013 (2002).

[122] C. Rovelli, Quantum Gravity, (Cambridge University Press, 2004).

[123] C. Rovelli and L. Smolin, Loop representation for quantum general relativity, Nucl. Phys.

B 331, 80 (1990).

[124] C. Rovelli and L. Smolin, The physical Hamiltonian in nonperturbative quantum gravity,

Phys. Rev. Lett 72, 446 (1994).

[125] C. Rovelli and L. Smolin, Discreteness of area and volume in quantum gravity, Nucl.

Phys. B 442, 593, (1995).

[126] C. Rovelli and L. Smolin, Spin networks and quantum gravity, Phys. Rev. D 52, 5743,

(1995).

[127] C. Rovelli, Graviton propagator from background-independent quantum gravity,

Phys.Rev.Lett. 97, 151301, (2006).

[128] H. Sahlmann and T. Thiemann, Towards the QFT on curved spacetime limit of QGR: I.

A general scheme, Class. Quantum Grav. 23, 867 (2006), (preprint: gr-qc/0207030).

[129] H. Sahlmann and T. Thiemann, Towards the QFT on curved spacetime limit of QGR:

II. A concrete implementation, Class. Quantum Grav. 23, 909 (2006), (Preprint: gr-

qc/0207031).

[130] H. Sahlmann and T. Thiemann, Irreducibility of the Ashtekar-Isham-Lewandowski rep-

resentation, (preprint: gr-qc/0303074).

[131] J. Samuel, Is Barberos Hamiltonian formulation a gauge throry of Lorentzian gravity?

Class. Quantum Grav. 17, L141 (2000), (Preprint: gr-qc/0005095).

[132] S. Sternberg, Group Theory and Physics, (Cambridge University Press, 1994).

[133] lecture notes given by T. Thiemann at Beijing Normal University.

[134] T. Thiemann, Quantum spin dynamics (QSD), Class. Quantum Grav. 15, 839 (1998).

[135] T. Thiemann, Quantum spin dynamics (QSD): II. The kernel of the Wheeler - DeWitt

constraint operator, Class. Quantum Grav. 15, 875 (1998).

[136] T. Thiemann, Quantum spin dynamics (QSD): III. Quantum constraint algebra and phys-

ical scalar product in quantum general relativity, Class. Quantum Grav. 15, 1207 (1998).

88

[137] T. Thiemann, Quantum spin dynamics (QSD): IV. Euclidean quantum gravity as a model

to test Lorentzian quantum gravity, Class. Quantum Grav. 15, 1249 (1998).

[138] T. Thiemann, Quantum spin dynamics (QSD): V. Quantum gravity as the natural regula-

tor of the Hamiltonian constraint of matter quantum eld theories, Class. Quantum Grav.

15, 1281 (1998).

[139] T. Thiemann, Quantumspin dynamics (QSD): VI. QuantumPoincare algebra and a quan-

tum positivity of energy theorem for canonical quantum gravity, Class. Quantum Grav. 15,

1463 (1998).

[140] T. Thiemann, Quantum spin dynamics (QSD): VIII. The master constraint, Class. Quan-

tum Grav. 23, 2249 (2006).

[141] T. Thiemann, A length operator for canonical quantum gravity, J. Math. Phys. 39, 3372

(1998).

[142] T. Thiemann, Gauge eld theory coherent states (GCS): I. General properties, Class.

Quantum Grav. 18, 2025 (2001).

[143] T. Thiemann and O. Winkler, Gauge eld theory coherent states (GCS): II. Peakedness

properties, Class. Quantum Grav. 18, 2561 (2001).

[144] T. Thiemann and O. Winkler, Gauge eld theory coherent states (GCS): III. Ehrenfest

theorems, Class. Quantum Grav. 18, 4629 (2001).

[145] T. Thiemann and O. Winkler, Gauge eld theory coherent states (GCS): IV. Innite tensor

product and thermodynamical limit, Class. Quantum Grav. 18, 4997 (2001).

[146] T. Thiemann, Modern Canonical Quantum General Relativity, (Cambridge University

Press, in press), Draft preprint: gr-qc/0110034.

[147] T. Thiemann, Complexier coherent states for quantum general relativity, Class. Quan-

tum Grav. 23, 2063 (2006), (preprint: gr-qc/0206037).

[148] T. Thiemann, Lectures on loop quantum gravity, Lect. Notes Phys. 631 (2003) 41,

(preprint: gr-qc/0210094).

[149] T. Thiemann, The phoenix project: Master constraint programme for loop quantum grav-

ity, Class. Quantum Grav. 23, 2211 (2006), (preprint: gr-qc/0305080).

[150] T. Thiemann, Reduced phase space quantization and Dirac observables, Class. Quantum

Grav. 23, 1163 (2006), (preprint: gr-qc/0411031).

[151] T. Thiemann, Loop quantum gravity: an inside view, preprint: hep-th/0608210.

[152] M. Varadarajan, Fock representation from U(1) holonomy agebras, Phys. Rev. D 61,

104001 (2000).

89

[153] M. Varadarajan, Photons from quantized electric ux representations, Phys. Rev. D 64,

104003 (2001).

[154] M. Varadarajan, Gravitons from a loop representation of linearized gravity, Phys. Rev. D

66, 024012 (2002).

[155] J. M. Velhinho, A groupoid approach to spaces of generalized connections, J. Geom.

Phys. 41, 166 (2002).

[156] R. M. Wald, General Relativity, (The University of Chicago Press, 1984).

[157] R. M. Wald, Quantum Field Theory in Curved Spacetime and Black Hole Thermody-

namics, (The University of Chicago Press, 1994).

[158] S. Weinberg, The Quantum Theory of Fields I, II, III, (Cambridge University Press, New

York, 1995.)

[159] C. M. Will, The Confrontation between general relativity and experiments, Living Rev.

Relativity 4 (2001), 4.

90

Vita

Muxin Han was born in Beijing, Peoples Republic of China. He studied as an undergraduate

student in the Department of Physics at Beijing Normal University from 2001 to 2005, and

obtained his Bachelor of Science degree at Beijing Normal University in 2005. Muxin came

to the United States and began his graduate studies at Louisiana State University in August of

2005. His major is physics.

91

- Physics2012.pdfDiunggah olehJamie Heredge
- Energy Universe /Energetic Physics/Diunggah olehSPHERE PUBLISHING HOUSE
- Quantum Math Makes Human Irrationality More SensibleDiunggah olehoffice_gang
- The Unified Field and the Illusion of TimeDiunggah olehSky
- DAriano_FQXi2013-It From BitDiunggah olehfware28
- Bohm_meaning+informationDiunggah olehscribd_new_account
- Copy of Nd Less Technical Introduction to the...Diunggah olehAshwani Kumar
- 1607.07356Diunggah oleh程元李
- Witness Gravity’s Quantum Side in the Lab (Nature Paper)Diunggah olehadi63
- FrontMatter_Giancoli7e_Manual.pdfDiunggah olehAndi Rodik
- On Deformed Heisenberg Algebras and Gravity’s RainbowDiunggah olehmohammadzaz
- What is TimeDiunggah olehSylvia Cheung
- A New Approach to the Hard Problem - 07.14 - Final - KI - TSC 2014Diunggah olehforizsl
- The Einstein-myth and the Crisis in Modern PhysicsDiunggah olehLuis Gerardo Escandon Alcazar
- Quantum theories newDiunggah olehmaniacraghu
- Why Physics Needs Quantum FoundationsDiunggah olehahsbon
- Quantum in EdDiunggah olehryondong
- The DKP oscillator with a linear interaction in the cosmic string space-timeDiunggah olehFabiano M Andrade
- Introduction to Cosmology, Physics, PhilosophyDiunggah olehProf.-Dr. Benjamin Gal-Or
- Feynman Integral on Lie GroupsDiunggah olehCarlos Ramírez
- J. Ambjørn et al- Lorentzian 3d gravity with wormholes via matrix modelsDiunggah olehRtpom
- MindDiunggah olehscribd5533
- The Theory of Creation of EverythingDiunggah olehInternational Journal of Innovative Science and Research Technology
- NewScientist - Poster of RealityDiunggah olehoaky
- Davis, William M Physical Chemistry a Modern Introduction, Second EditionDiunggah olehAlex Alves
- WhittakerDiunggah olehEnric Toledo
- Max Planck Was a German Theoretical PhysicistDiunggah olehKen Soriano
- Jaroslav Zamastil, Jakub Benda-Quantum Mechanics and Electrodynamics-Springer International Publishing (2017)Diunggah olehJavier Berlanga
- capstoneoutlineDiunggah olehapi-397061932
- QUANTM_BDiunggah olehangelo_puccinelli

- Carlo Rovelli- Loop Quantum GravityDiunggah olehArsLexii
- Rodolfo Gambini- Lectures on loop quantum gravityDiunggah olehArsLexii
- Rainer E. Zimmermann- Loops and Knots as Topoi of Substance: Spinoza RevisitedDiunggah olehArsLexii
- Rainer E. Zimmermann- Classicity from Entangled Ensemble States of Knotted Spin Networks. A Conceptual ApproachDiunggah olehArsLexii
- Emanuele Alesci- Complete Loop Quantum Gravity Graviton PropagatorDiunggah olehArsLexii
- Sergei Alexandrov and Philippe Roche- Critical Overview of Loops and FoamsDiunggah olehArsLexii
- Rodolfo Gambini and Jorge Pullin- Quantum gravity experimental physics?Diunggah olehArsLexii
- Rodolfo Gambini and Jorge Pullin- Nonstandard optics from quantum space-timeDiunggah olehArsLexii
- Rickard M. Jonsson- Visualizing curved spacetimeDiunggah olehArsLexii
- Miguel Carrion Alvarez- Loop quantization versus Fock quantization of p-form electromagnetism on static spacetimesDiunggah olehArsLexii
- Christophe Ringeval- Cosmic Strings and Their Induced Non-Gaussianities in the Cosmic Microwave BackgroundDiunggah olehArsLexii
- William Donnelly- Entanglement Entropy in Quantum GravityDiunggah olehArsLexii
- Luisa Doplicher- Propagation kernel techniques for loop quantum gravityDiunggah olehArsLexii
- Elena Magliaro- Low energy limit in Loop Quantum GravityDiunggah olehArsLexii
- Seth A. Major- A spin network primerDiunggah olehArsLexii
- Florian Conrady- Free vacuum for loop quantum gravityDiunggah olehArsLexii
- Daniele Dominici and Samir Ferrag- Large Extra DimensionsDiunggah olehArsLexii
- Itzhak Bars and Yueh-Cheng Kuo- Interacting Two-Time Physics Field Theory With a BRST Gauge Invariant ActionDiunggah olehArsLexii
- Florian Conrady- Semiclassical Analysis of Loop Quantum GravityDiunggah olehArsLexii
- A. Mikovic- Quantum gravity vacuum and invariants of embedded spin networksDiunggah olehArsLexii
- Itzhak Bars- Super Yang-Mills Theory in 10+2 dims. Another Step Toward M-TheoryDiunggah olehArsLexii
- Eric Bergshoeff and Antoine Van Proeyen- Symmetries of string, M- and F-theoriesDiunggah olehArsLexii
- Chopin Soo and Lay Nam Chang- Superspace Dynamics and Perturbations Around "Emptiness"Diunggah olehArsLexii
- W. Chagas-Filho- Noncommutative Two Time PhysicsDiunggah olehArsLexii
- Jan Brugues et al- Newton-Hooke algebras, nonrelativistic branes, and generalized pp-wave metricsDiunggah olehArsLexii
- W. Chagas-Filho- Two Time Physics with a Minimum LengthDiunggah olehArsLexii
- P. Mitra- Quantum mechanics for two-timersDiunggah olehArsLexii
- Davoud Kamani- Ghosts in the matter formsDiunggah olehArsLexii
- W. Chagas-Filho- Gravitation with Two TimesDiunggah olehArsLexii

- NotesDiunggah olehReza Arraffi
- high school practice examDiunggah olehAlireza Kafaei
- Hatcher - The Serre Spectral SequenceDiunggah olehAdrian
- Unit B Activities and ReviewsDiunggah olehRyanFaller
- stratigys for intergrationDiunggah olehtmkzppp
- MarkovDiunggah olehsam
- ppt_presentationDiunggah olehRadixMons
- MidtermDiunggah olehJay Kay
- Lecture 3Diunggah olehMario Cacasenno
- 3.10 Solutions 17r3l2vDiunggah olehAshok Pradhan
- Non Linear Shooting Algorithm Boundary Value Problem Numerical Analysis Matlab CodeDiunggah olehshankar
- ElectromagnetismDiunggah olehEralyn Ebon
- Simulation and Analysis of power system transientsDiunggah olehdeepak_27
- Lecture 3 Polynomial Interpolation for UploadDiunggah olehRianna Cruz
- RealTime OptimizationDiunggah olehshudhai jobe
- It Lecture Notes2014Diunggah olehRasha Ziyad
- finite element analysisDiunggah olehPardha Saradhi
- Characterization of Inner Product SpacesDiunggah olehGeorge Protopapas
- Demostracion SMEH.docxDiunggah olehYinaLizbethHuarza
- Integral EquationDiunggah olehMandela Bright Quashie
- Dynamics and Controls for Robot Manipulators With Open and ClosedDiunggah oleheocv20
- solutions2Diunggah olehLeo Liang Fang
- Edexcel C2 MS Jan 2011Diunggah olehIssam Saif
- relationships in patterns -2Diunggah olehapi-288922072
- practical yield curve modelsDiunggah olehhaginile
- Logan K. Young on the Maximality of Ultra-Trivially Singular ClassesDiunggah olehLogan K. Young
- Application of Newton-Raphson Algorithm in Common Random Numbers for Finding the Optimal Solution in Simultaneous Perturbation Stochastic ApproximationDiunggah olehTI Journals Publishing
- Truss MethodDiunggah olehGerald Maginga
- Heart of AlgebraDiunggah olehKevin Dany
- Understanding the Finite-Difference Time-Domain MethodDiunggah olehDayse Bandeira