Proceedings of the II International Conference on Fundamental Interactions

Pedra Azul, Brazil, June 2004

arXiv:gr-qc/0409061v3 9 Feb 2005

INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN

FOAMS

ALEJANDRO PEREZ∗
Institute for Gravitational Physics and Geometry, Penn State University, 104 Davey Lab
University Park, PA 16802, USA
and
Centre de Physique Théorique CNRS
Luminy Case 907 F-13288 Marseille cedex 9, France†

These notes are a didactic overview of the non perturbative and background independent
approach to a quantum theory of gravity known as loop quantum gravity. The definition
of real connection variables for general relativity, used as a starting point in the program,
is described in a simple manner. The main ideas leading to the definition of the quantum
theory are naturally introduced and the basic mathematics involved is described. The
main predictions of the theory such as the discovery of Planck scale discreteness of
geometry and the computation of black hole entropy are reviewed. The quantization
and solution of the constraints is explained by drawing analogies with simpler systems.
Difficulties associated with the quantization of the scalar constraint are discussed.
In a second part of the notes, the basic ideas behind the spin foam approach are
presented in detail for the simple solvable case of 2+1 gravity. Some results and ideas
for four dimensional spin foams are reviewed.

Keywords: Quantum Gravity; Loop variables; Non-perturbative methods; Path Integrals;

Yang-Mills theory.

1. Introduction: why non perturbative quantum gravity?

The remarkable experimental success of the standard model in the description of
fundamental interactions is the greatest achievement of (relativistic) quantum field
theory. Standard quantum field theory provides an accomplished unification be-
tween the principles of quantum mechanics and special relativity. The standard
model is a very useful example of a quantum field theory on the fixed background
geometry of Minkowski spacetime. As such, the standard model can only be re-
garded as an approximation of the description of fundamental interactions valid
when the gravitational field is negligible. The standard model is indeed a very good
approximation describing particle physics in the lab and in a variety of astrophysical
situations because of the weakness of the gravitational force at the scales of interest
(see for instance the lecture by Rogério Rosenfeld1 ). Using the techniques of quan-
tum field theory on curved spacetimes2 one might hope to extend the applicability
∗ Penn State University, 104 Davey Lab, University Park, PA 16802, USA.
† State completely without abbreviations, the affiliation and mailing address, including country.

1
2 Alejandro Perez

of the standard model to situations where a non trivial, but weak, gravitational
field is present. These situations are thought to be those where the spacetime cur-
vature is small in comparison with the Planck scale, although a clear justification
for its regime of validity in strong gravitational fields seems only possible when a
full theory of quantum gravity is available. Having said this, there is a number of
important physical situations where we do not have any tools to answer even the
simplest questions. In particular classical general relativity predicts the existence of
singularities in physically realistic situations such as those dealing with black hole
physics and cosmology. Near spacetime singularities the classical description of the
gravitational degrees of freedom simply breaks down. Questions related to the fate
of singularities in black holes or in cosmological situations as well as those related
with apparent information paradoxes are some of the reasons why we need a theory
of quantum gravity. This new theoretical framework—yet to be put forward—aims
at a consistent description unifying or, perhaps more appropriately, underlying the
principles of general relativity and quantum mechanics.
The gravitational interaction is fundamentally different from all the other known
forces. The main lesson of general relativity is that the degrees of freedom of the
gravitational field are encoded in the geometry of spacetime. The spacetime geom-
etry is fully dynamical: in gravitational physics the notion of absolute space on top
of which ‘things happen’ ceases to make sense. The gravitational field defines the
geometry on top of which its own degrees of freedom and those of matter fields
propagate. This is clear from the perspective of the initial value formulation of gen-
eral relativity, where, given suitable initial conditions on a 3-dimensional manifold,
Einstein’s equations determine the dynamics that ultimately allows for the recon-
struction of the spacetime geometry with all the matter fields propagating on it. A
spacetime notion can only be recovered a posteriori once the complete dynamics of
the coupled geometry-matter system is worked out. Matter affects the dynamics of
the gravitational field and is affected by it through the non trivial geometry that
the latter defines a . General relativity is not a theory of fields moving on a curved
background geometry; general relativity is a theory of fields moving on top of each
other4 .
In classical physics general relativity is not just a successful description of the
nature of the gravitational interaction. As a result of implementing the principles
of general covariance, general relativity provides the basic framework to assessing
the physical world by cutting all ties to concepts of absolute space. It represents
the result of the long-line of developments that go all the way back to the thought
experiments of Galileo about the relativity of the motion, to the arguments of
Mach about the nature of space and time, and finally to the magnificent conceptual
synthesis of Einstein’s: the world is relational. There is no well defined notion of
absolute space and it only makes sense to describe physical entities in relation to

a See for instance Theorem 10.2.2 in 3 .

 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 3

other physical entities. This conceptual viewpoint is fully represented by the way
matter and geometry play together in general relativity. The full consequences of
this in quantum physics are yet to be unveiledb .
When we analyze a physical situation in pre-general-relativistic physics we sep-
arate what we call the system from the relations to other objects that we call the
reference frame. The spacetime geometry that we describe with the aid of coordi-
nates and a metric is a mathematical idealization of what in practice we measure
using rods and clocks. Any meaningful statement about the physics of the system
is a statement about the relation of some degrees of freedom in the system with
those of what we call the frame. The key point is that when the gravitational field is
trivial there exist a preferred set of physical systems whose dynamics is very simple.
These systems are inertial observers; for instance one can think of them as given by
a spacial grid of clocks synchronized by the exchange of light signals. These physical
objects provide the definition of inertial coordinates and their mutual relations can
be described by a Minkowski metric. As a result in pre-general relativistic physics
we tend to forget about rods and clocks that define (inertial) frames and we talk
about time t and position x and distances measure using the flat metric ηab c . How-
ever, what we are really doing is comparing the degrees of freedom of our system
with those of a space grid of world-lines of physical systems called inertial observers.
From this perspective, statements in special relativity are in fact diffeomorphism in-
variant. The physics from the point of view of an experimentalist—dealing with the
system itself, clocks, rods, and their mutual relations—is completely independent of
coordinates. In general relativity this property of the world is confronted head on:
only relational (coordinate-independent, or diffeomorphism invariant) statements
are meaningful (see discussion about the hole argument in 4 ). There are no sim-
ple family of observers to define physical coordinates as in the flat case, so we use
arbitrary labels (coordinates) and require the physics to be independent of them.
In this sense, the principles of general relativity state some basic truth about the
nature of the classical world. The far reaching consequences of this in the quantum
realm are certainly not yet well understood.d However, it is very difficult to imagine
that a notion of absolute space would be saved in the next step of development
of our understanding of fundamental physics. Trying to build a theory of quantum
gravity based on a notion of background geometry would be, from this perspective,
reminiscent of the efforts by contemporaries of Copernicus of describing planetary
motion in terms of the geocentric framework.

b For a fascinating account of the conceptual subtleties of general relativity see Rovelli’s book4 .
c In principle we first use rods and clocks to realize that in the situation of interests (e.g. an
experiment at CERN) the gravitational field is trivial and then we just encode this information in
a fix background geometry: Minkowski spacetime.
d A reformulation of quantum mechanics from a relational perspective has been introduced

by Rovelli5 and further investigated in the context of conceptual puzzles of quantum

mechanics6,7,8,9,10 . It is also possible that a radical change in the paradigms of quantum me-
chanics is necessary in the conceptual unification of gravity and the quantum11,12,13,14 .
4 Alejandro Perez

1.1. Perturbative quantum gravity

Let us make some observations about the problems of standard perturbative quan-
tum gravity. In doing so we will revisit the general discussion above, in a special
situation. In standard perturbative approaches to quantum gravity one attempts to
describe the gravitational interaction using the same techniques applied to the defi-
nition of the standard model. As these techniques require a notion of non dynamical
background one (arbitrarily) separates the degrees of freedom of the gravitational
field in terms of a background geometry ηab for a, b = 1 · · · 4—fixed once and for
all—and dynamical metric fluctuations hab . Explicitly, one writes the spacetime
metric as
gab = ηab + hab . (1)
Notice that the previous separation of degrees of freedom has no intrinsic meaning
in general relativity. In other words, for a generic space time metric gab we can write

gab = ηab + hab = η̃ab + h̃ab , (2)

where ηab and η̃ab can lead to different background light-cone structures of the un-
derlying spacetime (M, gab ); equally natural choices of flat background metrics lead
to different Minkowski metrics in this sense. This is quite dangerous if we want to
give any physical meaning to the background, e.g., the light cone structures of the
two ‘natural’ backgrounds will be generally different providing different notions of
causality! Equation (1) is used in the classical theory in very special situations when
one considers perturbations of a given background ηab . In quantum gravity one has
to deal with arbitrary superpositions of spacetimes; the above splitting can at best
be meaningful for very special semi-classical states ‘peaked’, so to say, around the
classical geometry ηab with small fluctuations. It is very difficult to imagine how
such a splitting can be useful in considering general states with arbitrary quan-
tum excitations at all scales. Specially because of the dual role of the gravitational
field that simultaneously describes the geometry and its own dynamical degrees of
freedom. More explicitly, in the standard background dependent quantization the
existence of a fixed background geometry is fundamental in the definition of the
theory. For instance, one expects fields at space-like separated points to commute
alluding to standard causality considerations. Even when this is certainly justified
in the range of applicability of the standard model, in a background dependent
quantization of gravity one would be using the causal structure provided by the un-
physical background ηab . Yet we know that the notion of causality the world really
follows is that of the full gab (see Footnote a). This difficulty has been raised several
times (see for instance3 ). Equation (1) could be meaningful in special situations
dealing with semi-classical issues, but it does not seem to be of much use if one
wants to describe the fundamental degrees of freedom of quantum gravity.
If we ignore all these issues and try to setup a naive perturbative quantization
of gravity we find that the theory is non renormalizable. This can be expected from
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 5

Fig. 1. The larger cone represents the light-cone at a point according to the ad hoc background
ηab . The smaller cones are a cartoon representation of the fluctuations of the true gravitational
field represented by gab .

dimensional analysis as the quantity playing the role of the coupling constant turns
out to be the Planck length ℓp . The non renormalizability of perturbative gravity
is often explained through an analogy with the (non-renormalizable) Fermi’s four
fermion effective description of the weak interaction15 . Fermi’s four fermions theory
is known to be an effective description of the (renormalizable) Weinberg-Salam
theory. The non renormalizable UV behavior of Fermi’s four fermion interaction is
a consequence of neglecting the degrees of freedom of the exchanged massive gauge
bosons which are otherwise disguised as the dimension-full coupling ΛFermi ≈ 1/m2W
at momentum transfer much lower than the mass of the W particle (q 2 << m2W ).
A similar view is applied to gravity to promote the search of a more fundamental
theory which is renormalizable or finite (in the perturbative sense) and reduces
to general relativity at low energies. From this perspective it is argued that the
quantization of general relativity is a hopeless attempt to quantizing a theory that
does not contain the fundamental degrees of freedom.
These arguments, based on background dependent concepts, seem at the very
least questionable in the case of gravity. Although one should expect the notion of a
background geometry to be useful in certain semi-classical situations, the assump-
tion that such structure exists all the way down to the Planck scale is inconsistent
with what we know about gravity and quantum mechanics. General considerations
indicate that standard notions of space and time are expected to fail near the Planck
scale ℓp e . From this viewpoint the non renormalizability of perturbative quantum

e For instance a typical example is to use a photon to measure distance. The energy of the photon

in our lab frame is given by Eγ = hc/λ. We put the photon in a cavity and produce a standing
wave measuring the dimensions of the cavity in units of the wavelength. The best possible precision
is attained when the Schwarzschild radius corresponding to energy of the photon is of the order
of its wavelength. Beyond that the photon can collapse to form a black hole around some of the
maxima of the standing 4 3
p wave. This happens for a value λc for which λc ≈ GEγ /c = hG/(λc c ).
The solution is λc ≈ hG/c3 which is Planck length.
6 Alejandro Perez

gravity is indicative of the inconsistency of the separation of degrees of freedom

in (1). The nature of spacetime is expected to be very different from the classical
notion in quantum gravity. The treatment that uses (1) as the starting point is
assuming a well defined notion of background geometry at all scales which directly
contradicts these considerations.
In fact one could read the issues of divergences in an alternative way. The ac-
cepted jargon is that in a renormalizable theory the microscopic physics does not
affect low energy processes as its effects can be recast in simple redefinition of a
finite set of parameters, i.e., renormalization. In the case of gravity this miracle does
not happen. All correlation functions blow up and one would need to add an infinite
number of corrections to the original Lagrangian when integrating out microscopic
physical degrees of freedom. Now imagine for the moment that a non perturba-
tive quantization of general relativity was available. If we now want to recover a
‘low energy effective’ description of a physical situation —where (low energy) phys-
ical considerations single out a preferred background—we would expect the need
of increasingly higher derivative correction terms to compensate the fact the ‘high
energy’f degrees of freedom know nothing about that low energy background. In
this sense the objections raised above to the violent and un-natural splitting (1) are
related to the issue of non renormalizability of general relativity. The nature of the
gravitational interaction is telling us that the standard paradigm of renormalization
(based on the notion of a fixed background) is no longer applicable. A reformulation
of quantum field theory is necessary to cope with background independence.
It is possible that new degrees of freedom would become important at more
fundamental scales. After all, that is the story of the path that lead to the standard
model where higher energy experiments has often lead to the discovery of new in-
teractions. It is also possible that including these degrees of freedom might be very
important for the consistency of the theory of quantum gravity. However, there is
constraint that seem hardly avoidable: if we want to get a quantum theory that
reproduces gravity in the semi-classical limit we should have a background inde-
pendent formalism. In loop quantum gravity one stresses this viewpoint. The hope
is that learning how to define quantum field theory in the absence of a background
is a key ingredient in a recipe for quantum gravity. Loop quantum gravity uses the
quantization of general relativity as a playground to achieve this goal.
I would like to finish this subsection with a quote of Weinberg’s 1980 paper that
would serve as introduction for the next section:

“It is possible that this problem—the non renormalizability of general relativity—

f The notion of energy is observer dependent in special relativity. In the background independent

context is not even defined. We use the terminology ‘high (low) energy’ in this introduction to refer
to the fundamental (semi-classical) degrees of freedom in quantum gravity. However, the reader
should not take this too literally. In fact if one tries to identify the fundamental degrees of freedom
of quantum gravity with a preferred observer conflict with observation seem inevitable16,17,18 .
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 7

has arisen because the usual flat-space formalism of quantum field theory simply
cannot be applied to gravitation. After all, gravitation is a very special phenomenon,
involving as it does the very topology of space and time”

All these considerations make the case for a background independent approach
to quantum gravity. The challenge is to define quantum field theory in the absence
of any preestablished notion of distance: quantum field theory without a metric.

1.2. Loop quantum gravity

Loop quantum gravity is an attempt to define a quantization of gravity paying spe-
cial attention to the conceptual lessons of general relativity (the reader is encouraged
to read Rovelli’s book: Quantum Gravity4 as well as the recent review by Ashtekar
and Lewandowski19. For a detailed account of the mathematical techniques in-
volved in the construction of the theory see Thiemann’s book20). The theory is
explicitly formulated in a background independent, and therefore, non perturba-
tive fashion. The theory is based on the Hamiltonian (or canonical) quantization
of general relativity in terms of variables that are different from the standard met-
ric variables. In terms of these variables general relativity is cast into the form of a
background independent SU (2) gauge theory partly analogous to SU (2) Yang-Mills
theory. The main prediction of loop quantum gravity (LQG) is the discreteness21,22
of the spectrum of geometrical operators such as area and volume. The discreteness
becomes important at the Planck scale while the spectrum of geometric operators
crowds very rapidly at ‘low energy scales’ (large geometries). This property of the
spectrum of geometric operators is consistent with the smooth spacetime picture of
classical general relativity.
Thus it is not surprising that perturbative approaches would lead to inconsisten-
cies. In splitting the gravitational field degrees of freedom as in (1) one is assuming
the existence of a background geometry which is smooth all the way down to the
Planck scale. As we consider contributions from ‘higher energies’, this assumption
is increasingly inconsistent with the fundamental structure discovered in the non
perturbative treatment.
The theory is formulated in four dimensions. Matter can be coupled to the
gravitational degrees of freedom and the inclusion of matter has been studied in
great detail. However, throughout these lectures we will study the pure gravita-
tional sector in order simplify the presentation. On the super-symmetric extension
of canonical loop quantum gravity see 23,24,25,26,27 . A spin foam model for 3d
super-gravity has been defined by Livine and Oeckl28.
Despite the achievements of LQG there remain important issues to be addressed.
There are difficulties of a rather technical nature related to the complete charac-
terization of dynamics and the quantization of the so-called scalar constraint. We
will review these once we have introduced the basic formalism. The most important
question, however, remains open: whether the semi-classical limit of the LQG is
8 Alejandro Perez

consistent with general relativity and the Standard Model.

Before starting with the more technical material let us say a few words regard-
ing unification. In loop quantum gravity matter is essentially added by coupling
general relativity with it at the classical level and then performing the (background
independent) canonical quantization. At this stage of development of the approach
it is not clear if there is some restriction in the kind of interactions allowed or if
degrees of freedom corresponding to matter could arise in some natural way directly
from those of geometry. However, there is a unification which is addressed head on
by LQG: the need to find a common framework in which to describe quantum field
theories in the absence of an underlying spacetime geometry. The implications of
this are far from being fully developed.

2. Hamiltonian formulation of general relativity: the new variables

In this section we introduce the variables that are used in the definition of loop
quantum gravity. We will present the SU (2) Ashtekar-Barbero variables by starting
with conventional ADM variables, introducing triads at the canonical level, and
finally performing the well known canonical transformation that leads to the new
variables. This is, in my opinion, the simplest way to get to the new variables.
For a derivation that emphasizes the covariant four dimensional character of these
variables see 19 .

2.1. Canonical analysis in ADM variables

The action of general relativity in metric variables is given by the Einstein-Hilbert
action
Z
1 √
I[gµν ] = dx4 −gR, (3)
2κ
where κ = 8πG/c3 = 8πℓ2p /~, g is the determinant of the metric gab and R is the
Ricci scalar. The details of the Hamiltonian formulation of this action in terms
of ADM variables can be found in3 . One introduces a foliation of spacetime in
terms of space-like three dimensional surfaces Σ. For simplicity we assume Σ has
no boundaries. The ten components of the spacetime metric are replaced by the six
components of the induced Riemannian metric qab of Σ plus the three components
of the shift vector Na and the lapse function N . In terms of these variables, after
performing the standard Legendre transformation, the action of general relativity
becomes
Z Z
1
I[qab , π ab , Na , N ] = dt dx3 [π ab q̇ab
2κ Σ
1
+2Nb ∇(3) a (q −1/2 π ab ) + N (q 1/2 [R(3) − q −1 πcd π cd + q −1 π 2 ])], (4)
2
where π ab are the momenta canonically conjugate to the space metric qab , π =
π ab qab , ∇(3) a is the covariant derivative compatible with the metric qab , q is the
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 9

determinant of the space metric and R(3) is the Ricci tensor of qab . The momenta
π ab are related to the extrinsic curvature Kab g of Σ by
π ab = q −1/2 (K ab − Kq ab ) (5)
where K = Kab q ab and indices are raised with q ab . Variations with respect to the
lapse and shift produce the four constraint equations:
−V b (qab , π ab ) = 2∇(3) a (q −1/2 π ab ) = 0, (6)
and
−S(qab , π ab ) = (q 1/2 [R(3) − q −1 πcd π cd + 1/2q −1 π 2 ]) = 0. (7)
V b (qab , π ab ) is the so-called vector constraint and S(qab , π ab ) is the scalar constraint.
With this notation the action can be written as
Z Z
1  
I[qab , π ab , Na , N ] = dt dx3 π ab q̇ab − Nb V b (qab , π ab ) − N S(qab , π ab ) ,
2κ Σ
(8)
ab b ab
where we identify the Hamiltonian density H(qab , π , Na , N ) = Nb V (qab , π ) +
N S(qab , π ab ). The Hamiltonian is a linear combination of (first class) constraints,
i.e., it vanishes identically on solutions of the equations of motion. This is a generic
property of generally covariant systems. The symplectic structure can be read off
the previous equations, namely
 ab a b
 ab
π (x), qcd (y) = 2κ δ(c δd) δ(x, y), π (x), π cd (y) = {qab (x), qcd (y)} = 0 (9)
There are six configuration variables qab and four constraint equations (6) and (7)
which implies the two physical degrees of freedom of gravity h .

2.2. Toward the new variables: the triad formulation

Now we do a very simple change of variables. The idea is to use a triad (a set of
three 1-forms defining a frame at each point in Σ) in terms of which the metric qab
becomes
qab = eia ejb δij , (10)
where i, j = 1, 2, 3. Using these variables we introduce the densitized triad
1 abc
Eia := ǫ ǫijk ejb ekc . (11)
2
Using this definition, the inverse metric q ab can be related to the densitized triad
as follows
qq ab = Eia Ejb δ ij . (12)

g The extrinsic curvature is given by Kab = 21 Ln qab where na is the unit normal to Σ.
h This counting of physical degrees of freedom is correct because the constraints are first class29 .
10 Alejandro Perez

We also define
1
Kai := p Kab Ejb δ ij . (13)
det(E)
A simple exercise shows that one can write the canonical term in (8) as

π ab q̇ab = −πab q̇ ab = 2Eia K̇ai (14)

and that the constraints V a (qab , π ab ) and S(qab , π ab ) can respectively be written
as V a (Eia , Kai ) and S(Eia , Kai ). Therefore we can rewrite (8) in terms of the new
variables. However, the new variables are certainly redundant, in fact we are using
the nine Eia to describe the six components of q ab . The redundancy has a clear ge-
ometrical interpretation: the extra three degrees of freedom in the triad correspond
to our ability to choose different local frames eia by local SO(3) rotations acting in
the internal indices i = 1, 2, 3. There must then be an additional constraint in terms
of the new variables that makes this redundancy manifest. The missing constraint
comes from (13): we overlooked the fact that Kab = Kba or simply that K[ab] = 0.
By inverting the definitions (11) and (13) in order to write Kab in terms of Eia and
Kai one can show that the condition K[ab] = 0 reduces to

Gi (Eja , Kaj ) := ǫijk E aj Kak = 0. (15)

Therefore we must include this additional constraint to (8) if we want to use the
new triad variables. With all this the action of general relativity becomes

I[Eja , Kaj , Na , N, N j ] =
Z Z h i
1
dt dx3 Eia K̇ai − Nb V b (Eja , Kaj ) − N S(Eja , Kaj ) − N i Gi (Eja , Kaj ) ,(16)
κ Σ

where the explicit form of the constraints in terms of triad variables can be worked
out from the definitions above. The reader is encouraged to do this exercise but
it is not going to be essential to understanding what follows (expressions for the
constraints can be found in reference30). The symplectic structure now becomes
 a  a 
Ej (x), Kbi (y) = κ δba δji δ(x, y), Ej (x), Eib (y) = Kaj (x), Kbi (y) = 0 (17)

The counting of physical degrees of freedom can be done as before.

2.3. New variables: the Ashtekar-Barbero connection variables

The densitized triad (11) transforms in the vector representation of SO(3) under
redefinition of the triad (10). Consequently, so does its conjugate momentum Kai
(see equation (13)). There is a natural so(3)-connection that defines the notion of
covariant derivative compatible with the triad. This connection is the so-called spin
connection Γia and is characterized as the solution of Cartan’s structure equations

∂[a eib] + ǫi jk Γj[a ekb] = 0 (18)

 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 11

The solution to the previous equation can be written explicitly in terms of the triad
components
1  
Γia = − ǫij k ebj ∂[a ekb] + δ kl δms ecl em
a ∂b e s
c , (19)
2
where eai is the inverse triad (eai eja = δij ). We can obtain an explicit function of the
densitized triad—Γia (Ejb )—inverting (11) from where

1 ǫabc ǫijk Ejb Ekc sgn(det(E)) Eia

eia = p and eai = p . (20)
2 |det(E)| |det(E)|
The spin connection is an so(3) connection that transforms in the standard inho-
mogeneous way under local SO(3) transformations. The Ashtekar-Barbero31,32,33
variables are defined by the introduction of a new connection Aia given by
Aia = Γia + γKai , (21)
where γ is any non vanishing real number called the Immirzi parameter34 . The
new variable is also an so(3) connection as adding a quantity that transforms as a
vector to a connection gives a new connection. The remarkable fact about this new
variable is that it is in fact conjugate to Eai . More precisely the Poisson brackets of
the new variables are
 a  a 
Ej (x), Aib (y) = κ γδba δji δ(x, y), Ej (x), Eib (y) = Aja (x), Aib (y) = 0. (22)

All the previous equations follow trivially from (17) except for Aja (x), Aib (y) = 0
which requires more calculations. The reader is invited to check it.
Using the connection variables the action becomes
I[Eja , Aja , Na , N, N j ] =
Z Z h i
1
dt dx3 Eia Ȧai − N b Vb (Eja , Aja ) − N S(Eja , Aja ) − N i Gi (Eja , Aja ) , (23)
κ Σ
where the constraints are explicitly given by:
Vb (Eja , Aja ) = Eja Fab − (1 + γ 2 )Kai Gi (24)

Eia Ejb  ij k j

S(Eja , Aja ) = p ǫ k Fab − 2(1 + γ 2 )K[a
i
Kb] (25)
det(E)

Gi (Eja , Aja ) = Da Eia , (26)

where Fab = ∂a Aib − ∂b Aia + ǫi jk Aja Akb is the curvature of the connection Aia and
Da Eia = ∂a Eia + ǫij k Aja Eka is the covariant divergence of the densitized triad. We
have seven (first class) constraints for the 18 phase space variables (Aia , Ejb ). In ad-
dition to imposing conditions among the canonical variables, first class constraints
are generating functionals of (infinitesimal) gauge transformations. From the 18-
dimensional phase space of general relativity we end up with 11 fields necessary
12 Alejandro Perez

to coordinatize the constraint surface on which the above seven conditions hold.
On that 11-dimensional constraint surface, the above constraint generate a seven-
parameter-family of gauge transformations. The reduce phase space is four dimen-
sional and therefore the resulting number of physical degrees of freedom is two, as
expected.
The constraint (26) coincides with the standard Gauss law of Yang-Mills theory
~ ·E
(e.g. ∇ ~ = 0 in electromagnetism). In fact if we ignore (24) and (25) the phase
space variables (Aia , Ejb ) together with the Gauss law (26) characterize the physical
phase space of an SU (2)i Yang-Mills (YM) theory. The gauge field is given by
the connection Aia and its conjugate momentum is the electric field Ejb . Yang-Mills
theory is a theory defined on a background spacetime geometry. Dynamics in such a
theory is described by a non vanishing Hamiltonian—the Hamiltonian density of YM
theory being H = Eai Eia + Bai Bia . General relativity is a generally covariant theory
and coordinate time plays no physical role. The Hamiltonian is a linear combination
of constraints.j Dynamics is encoded in the constraint equations (24),(25), and (26).
In this sense we can regard general relativity in the new variables as a background
independent relative of SU (2) Yang-Mills theory. We will see in the sequel that
the close similarity between these theories will allow for the implementation of
techniques that are very natural in the context of YM theory.

2.3.1. Gauge transformations

Now let us analyze the structure of the gauge transformations generated by the
constraints (24),(25), and (26). From the previous paragraph it should not be sur-
prising that the Gauss law (26) generates local SU (2) transformations as in the case
of YM theory. Explicitly, if we define the smeared version of (26) as
Z Z
G(α) = dx3 αi Gi (Aia , Eia ) = dx3 αi Da Eia , (27)
Σ Σ

a direct calculation implies


δG Aia = Aia , G(α) = −Da αi and δG Eia = {Eia , G(α)} = [E, α]i . (28)

i The constraint structure does not distinguish SO(3) from SU (2) as both groups have the same

Lie algebra. From now on we choose to work with the more fundamental (universal covering) group
SU (2). In fact this choice is physically motivated as SU (2) is the gauge group if we want to include
fermionic matter35,36,37 .
j In the physics of the standard model we are used to identifying the coordinate t with the physical

time of a suitable family of observers. In the general covariant context of gravitational physics
the coordinate time t plays the role of a label with no physical relevance. One can arbitrarily
change the way we coordinatize spacetime without affecting the physics. This redundancy in the
description of the physics (gauge symmetry) induces the appearance of constraints in the canonical
formulation. The constraints in turn are the generating functions of these gauge symmetries. The
Hamiltonian generates evolution in coordinate time t but because redefinition of t is pure gauge,
the Hamiltonian is a constraint itself, i.e. H = 0 on shell38,29 . More on this in the next section.
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 13

If we write Aa = Aia τi ∈ su(2) and E a = Eia τ i ∈ su(2), where τi are generators of

SU (2), we can write the finite version of the previous transformation
A′a = gAa g −1 + g∂a g −1 and E a′ = gE a g −1 , (29)
which is the standard way the connection and the electric field transform under
gauge transformations in YM theory.
The vector constraint (24) generates three dimensional diffeomorphisms of Σ.
This is clear from the action of the smeared constraint
Z
a
V (N ) = dx3 N a Va (Aia , Eia ) (30)
Σ
on the canonical variables

δV Aia = Aia , V (N a ) = LN Aia and δV Eia = {Eia , V (N a )} = LN Eia , (31)
where LN denotes the Lie derivative in the N a direction. The exponentiation of
these infinitesimal transformations leads to the action of finite diffeomorphisms on
Σ.
Finally, the scalar constraint (25) generates coordinate time evolution (up to
space diffeomorphisms and local SU (2) transformations). The total Hamiltonian
H[α, N a , N ] of general relativity can be written as
H(α, N a , N ) = G(α) + V (N a ) + S(N ), (32)
where
Z
S(N ) = dx3 N S(Aia , Eia ). (33)
Σ
Hamilton’s equations of motion are therefore
   
Ȧia = Aia , H(α, N a , N ) = Aia , S(N ) + Aia , G(α) + Aia , V (N a ) , (34)
and
Ėia = {Eia , H(α, N a , N )} = {Eia , S(N )} + {Eia , G(α)} + {Eia , V (N a )} . (35)
The previous equations define the action of S(N ) up to infinitesimal SU (2) and
diffeomorphism transformations given by the last two terms and the values of α
and N a respectively. In general relativity coordinate time evolution does not have
any physical meaning. It is analogous to a U (1) gauge transformation in QED.

2.3.2. Constraints algebra

Here we simply present the structure of the constraint algebra of general relativity
in the new variables.
{G(α), G(β)} = G([α, β]), (36)
where α = αi τi ∈ su(2), β = β i τi ∈ su(2) and [α, β] is the commutator in su(2).
{G(α), V (N a )} = −G(LN α). (37)
14 Alejandro Perez

{G(α), S(N )} = 0. (38)

{V (N a ), V (M a )} = V ([N, M ]a ), (39)

where [N, M ]a = N b ∂b M a − M b ∂b N a is the vector field commutator.

{S(N ), V (N a )} = −S(LN N ). (40)

Finally

{S(N ), S(M )} = V (S a ) + terms proportional to the Gauss constraint, (41)

where for simplicity we are ignoring the terms proportional to the Gauss law (the
complete expression can be found in 19 ) and
Eia Ejb δ ij
Sa = (N ∂b M − M ∂b N ). (42)
|detE|
Notice that instead of structure constants, the r.h.s. of (41) is written in terms of
field dependent structure functions. For this reason it is said that the constraint
algebra does not close in the BRS sense.

2.3.3. Ashtekar variables

The connection variables introduced in this section do not have a simple relationship
with four dimensional fields. In particular the connection (21) cannot be obtained
as the pullback to Σ of a spacetime connection39 . Another observation is that the
constraints (24) and (25) dramatically simplify when γ 2 = −1. Explicitly, for γ = i
we have

VbSD = Eja Fab (43)

Eia Ejb
S SD = p ǫij k Fab
k
(44)
det(E)

GSD
i = Da Eia , (45)

where SD stands for self dual; a notation that will become clear below. Notice that
with γ = i the connection (21) is complex30 (i.e. Aa ∈ sl(2, C)). To recover real
general relativity these variables must be supplemented with the so-called reality
condition that follows from (21), namely

Aia + Āia = Γia (E). (46)

 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 15

In addition to the simplification of the constraints, the connection obtained for this
choice of the Immirzi parameter is simply related to a spacetime connection. More
precisely, it can be shown that Aa is the pullback of ωµ+IJ (I, J = 1, · · · 4) where
1 IJ i
ωµ+IJ = (ωµ − ǫIJKL ωµKL ) (47)
2 2
is the self dual part of a Lorentz connection ωµIJ . The gauge group—generated by
the (complexified) Gauss constraint—is in this case SL(2, C).
Loop quantum gravity was initially formulated in terms of these variables. How-
ever, there are technical difficulties in defining the quantum theory when the con-
nection is valued in the Lie algebra of a non compact group. Progress has been
achieved constructing the quantum theory in terms of the real variables introduced
in Section 2.3.

2.4. Geometric interpretation of the new variables

The geometric interpretation of the connection Aia , defined in (21), is standard. The
connection provides a definition of parallel transport of SU (2) spinors on the space
manifold Σ. The natural object is the SU (2) element defining parallel transport
along a path e ⊂ Σ also called holonomy denoted he [A], or more explicitly
Z
he [A] = P exp − A, (48)
e

where P denotes a path-order-exponential (more details in the next section).

The densitized triad—or electric field—Eia also has a simple geometrical mean-
ing. Eia encodes the full background independent Riemannian geometry of Σ as
is clear from (12). Therefore, any geometrical quantity in space can be written as
a functional of Eia . One of the simplest is the area AS [Eia ] of a surface S ⊂ Σ
whose expression we derive in what follows. Given a two dimensional surface in
S ⊂ Σ—with normal
∂xb ∂xc
na = ǫabc (49)
∂σ 1 ∂σ 2
where σ 1 and σ 2 are local coordinates on S—its area is given by
Z √
ab
AS [q ] = h dσ 1 dσ 2 , (50)
S

where h = det(hab ) is the determinant of the metric hab = qab − n−2 na nb induced
on S by q ab . From equation (12) it follows that det(q ab ) = det(Eia ). Let us contract
(12) with na nb , namely
qq ab na nb = Eia Ejb δ ij na nb . (51)
Now observe that q nn = q ab na nb is the nn-matrix element of the inverse of qab .
Through the well known formula for components of the inverse matrix we have that
16 Alejandro Perez

det(qab − n−2 na nb ) h
q nn = = . (52)
det(qab ) q

But qab − n−2 na nb is precisely the induced metric hab . Replacing q nn back into (51)
we conclude that

h = Eia Ejb δ ij na nb . (53)

Finally we can write the area of S as an explicit functional of Eia :

Z q
AS [Eia ] = Eia Ejb δ ij na nb dσ 1 dσ 2 . (54)
S

This simple expression for the area of a surface will be very important in the quan-
tum theory.

3. The Dirac program: the non perturbative quantization of GR

The Dirac program38,29 applied to the quantization of generally covariant systems
consists of the following stepsk :

(i) Find a representation of the phase space variables of the theory as operators in
an auxiliary or kinematical Hilbert space Hkin satisfying the standard commu-
tation relations, i.e., { , } → −i/~[ , ].
(ii) Promote the constraints to (self-adjoint) operators in Hkin . In the case of grav-
ity we must quantize the seven constraints Gi (A, E), Va (A, E), and S(A, E).
(iii) Characterize the space of solutions of the constraints and define the correspond-
ing inner product that defines a notion of physical probability. This defines the
so-called physical Hilbert space Hphys .
(iv) Find a (complete) set of gauge invariant observables, i.e., operators commuting
with the constraints. They represent the questions that can be addressed in the
generally covariant quantum theory.

3.1. A simple example: the reparametrized particle

Before going into the details of the definition of LQG we will present the general
ideas behind the quantization of generally covariant systems using the simplest pos-
sible example: a non relativistic particle. A non relativistic particle can be treated

k There is another way to canonically quantize a theory with constraints that is also developed

by Dirac. In this other formulation one solves the constraints at the classical level to identify
the physical or reduced phase space to finally quantize the theory by finding a representation
of the algebra of physical observables in the physical Hilbert space Hphys . In the case of four
dimensional gravity this alternative seem intractable due to the difficulty in identifying the true
degrees of freedom of general relativity.
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 17

in a generally covariant manner by introducing a non physical parameter t and pro-

moting the physical time T to a canonical variable. For a particle in one dimension
the standard action
Z  
dX
S(pX , X) = dT pX − H(pX , X) (55)
dT
can be replaced by the reparametrization invariant action
Z h i
Srep (PT , PX , T, X, N ) = dt pT Ṫ + pX Ẋ − N (pT + H(pX , X)) , (56)

where the dot denotes derivative with respect to t. The previous action is invariant
under the redefinition t → t′ = f (t). The variable N is a Lagrange multiplier
imposing the scalar constraint

C = pT + H(pX , X) = 0. (57)

Notice the formal similarity with the action of general relativity (8) and (16). As in
general relativity the Hamiltonian of the system Hrep = N (pT + H(pX , X)) is zero
on shell. It is easy to see that on the constraint surface defined by (57) Srep reduces
to the standard S, and thus the new action leads to the same classical solutions. The
constraint C is a generating function of infinitesimal t-reparametrizations (analog to
diffeomorphisms in GR). This system is the simplest example of generally covariant
system.
Let us proceed and analyze the quantization of this action according to the rules
above.

(i) We first define an auxiliary or kinematical Hilbert space Hkin . In this case we
can simply take Hkin = L2 (R2 ). Explicitly we use (kinematic) wave functions
of ψ(X, T ) and define the inner product
Z
< φ, ψ >= dXdT φ(X, T )ψ(X, T ). (58)

We next promote the phase space variables to self adjoint operators satisfying
the appropriate commutation relations. In this case the standard choice is that
Xb and Tb act simply by multiplications and pbX = −i~∂/∂X and pbT = −i~∂/∂T
(ii) The constraint becomes—this step is highly non trivial in a field theory due to
regularization issues:
2
b = −i~ ∂ − ~2 ∂ + V (X).
C (59)
∂T ∂X 2
b >= 0 is nothing else than the familiar
Notice that the constraint equation C|ψ
Schroedinger equation.
(iii) The solutions of the quantum constraint are in this case the solutions of
Schroedinger equation. As it is evident from the general form of Schroedinger
equation we can characterize the set of solutions by specifying the initial wave
18 Alejandro Perez

form at some T = T0 , ψ(X) = ψ(X, T = T0 ). The physical Hilbert space is

therefore the standard Hphys = L2 (R) with physical inner product
Z
< φ, ψ >p = dX φ(X)ψ(X). (60)

The solutions of Schroedinger equation are not normalizable in Hkin (they are
not square-integrable with respect to (58) due to the time dependence impose
by the Schroedinger equation). This is a generic property of the solutions of the
constraint when the constraint has continuous spectrum (think of the eigen-
states of Pb for instance).
(iv) Observables in this setting are easy to find. We are looking for phase space
functions commuting with the constraint. For simplicity assume for the moment
that we are dealing with a free particle, i.e., C = pT + p2X /(2m). We have in
this case the following two independent observables:

O b − pbX (Tb − T0 )
b1 = X and b2 = pbX ,
O (61)
m
where T0 is just a c-number. These are just the values of X and P at T = T0 .
In the general case where V (X) 6= 0 the explicit form of these observables as
functions of the phase space variables will depend on the specific interaction.
Notice that in Hphys , as defined above, the observables reduce to position O1 =
X and momentum O2 = pX as in standard quantum mechanics.

We have just reproduced standard quantum mechanics by quantizing the

reparametrization invariant formulation. As advertised in the general framework
of the Dirac program applied to generally covariant systems, the full dynamics is
contained in the quantum constraints—here the Schroedinger equation.

3.2. The program of loop quantum gravity

A formal description of the implementation of Dirac’s program in the case of gravity
is presented in what follows. In the next section we will start a more detailed review
of each of these steps.

(i) In order to define the kinematical Hilbert space of general relativity in terms
of the new variables we will choose the polarization where the connection is re-
garded as the configuration variable. The kinematical Hilbert space consists of
a suitable set of functionals of the connection ψ[A] which are square integrable
with respect to a suitable (gauge invariant and diffeomorphism invariant) mea-
sure dµAL [A] (called Ashtekar-Lewandowski measure40). The kinematical inner
product is given by
Z
< ψ, φ >= µAL [ψφ] = dµAL [A] ψ[A]φ[A]. (62)

In the next section we give the precise definition of Hkin .

 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 19

(ii) Both the Gauss constraint and the diffeomorphism constraint have a natu-
ral (unitary) action on states on Hkin . For that reason the quantization (and
subsequent solution) is rather straightforward. The simplicity of these six-out-
of-seven constraints is a special merit of the use of connection variables as will
become transparent in the sequel.
The scalar constraint (25) does not have a simple geometric interpretation.
In addition it is highly non linear which anticipates the standard UV problems
that plague quantum field theory in the definition of products of fields (operator
valued distributions) at a same point. Nevertheless, well defined versions of the
scalar constraint have been constructed. The fact that these rigorously defined
(free of infinities) operators exist is again intimately related to the kind of
variables used for the quantization and some other special miracles occurring
due to the background independent nature of the approach. We emphasize that
the theory is free of divergences.
(iii) Quantum Einstein’s equations can be formally expressed now as:

G \
b i (A, E)|Ψ >:= D a
a Ei |Ψ >= 0

Vba (A, E)|Ψ >:= Eia\i (A)|Ψ >= 0,

Fab
√ \ ij
−1
b
S(A, E)|Ψ >:= [ detE Eia Ejb Fab (A) + · · · ]|Ψ >= 0. (63)

As mentioned above, the space of solutions of the first six equations is well
understood. The space of solutions of quantum scalar constraint remains an
open issue in LQG. For some mathematically consistent definitions of Sb the
characterization of the solutions is well understood19 . The definition of the
physical inner product is still an open issue. We will introduce the spin foam
approach in Section 5 as a device for extracting solutions of the constraints
producing at the same time a definition of the physical inner product in LQG.
The spin foam approach also aims at the resolution of some difficulties appearing
in the quantization of the scalar constraint that will be discussed in Section
4.3.2. It is yet not clear, however, whether these consistent theories reproduce
general relativity in the semi-classical limit.
(iv) Already in classical gravity the construction of gauge independent quantities
is a subtle issue. At the present stage of the approach physical observables are
explicitly known only in some special cases. Understanding the set of physical
observables is however intimately related with the problem of characterizing
the solutions of the scalar constraint described before. We will illustrate this by
discussing simple examples of quasi-local Dirac observable in Section 4.3.2. For
a vast discussion about this issue we refer the reader to Rovelli’s book4 .
20 Alejandro Perez

4. Loop quantum gravity

4.1. Definition of the kinematical Hilbert space
4.1.1. The choice of variables: Motivation
As mentioned in the first item of the program formally described in the last sec-
tion, we need to define the vector space of functionals of the connection and a
notion of inner product to provide it with a Hilbert space structure of Hkin . As
emphasized in Section 2.4, a natural quantity associated with a connection consists
of the holonomy along a path (48). We now give a more precise definition of it:
Given a one dimensional oriented path e : [0, 1] ⊂ R → Σ sending the parameter
s ∈ [0, 1] → xµ (s), the holonomy he [A] ∈ SU (2) is denoted
Z
he [A] = P exp − A , (64)
e

where P denotes a path-ordered-exponential. More precisely, given the unique so-

lution he [A, s] of the ordinary differential equation
d
he [A, s] + ẋµ (s)Aµ he [A, s] = 0 (65)
ds
with the boundary condition he [A, 0] = 1, the holonomy along the path e is defined
as

he [A] = he [A, 1]. (66)

The previous differential equation has the form of a time dependent Schroedinger
equation, thus its solution can be formally written in terms of the familiar series
expansion
1 Zs1 sZn−1
∞ Z
X
he [A] = ds1 ds2 · · · dsn ẋµ1 (s1 ) · · · ẋµn (sn ) Aµ1 (s1 ) · · · Aµn (sn ), (67)
n=0 0 0 0

which is what the path ordered exponential denotes in (64). Let us list some im-
portant properties of the holonomy:

(i) The definition of he [A] is independent of the parametrization of the path e.

(ii) The holonomy is a representation of the groupoid of oriented paths. Namely,
the holonomy of a path given by a single point is the identity, given two oriented
paths e1 and e2 such that the end point of e1 coincides with the starting point
of e2 so that we can define e = e1 e2 in the standard fashion, then we have

he [A] = he1 [A]he2 [A], (68)

where the multiplication on the right is the SU (2) multiplication. We also have
that

he−1 [A] = h−1

e [A]. (69)
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 21

(iii) The holonomy has a very simple behavior under gauge transformations. It is
easy to check from (29) that under a gauge transformation generated by the
Gauss constraint, the holonomy transforms as

h′e [A] = g(x(0)) he [A] g −1 (x(1)). (70)

(iv) The holonomy transforms in a very simple way under the action of diffeo-
morphisms (transformations generated by the vector constraint (24)). Given
φ ∈ Diff(Σ) we have

he [φ∗ A] = hφ−1 (e) [A], (71)

where φ∗ A denotes the action of φ on the connection. In other words, trans-

forming the connection with a diffeomorphism is equivalent to simply ‘moving’
the path with φ−1 .

Geometrically the holonomy he [A] is a functional of the connection that provides

a rule for the parallel transport of SU (2) spinors along the path e. If we think of it
as a functional of the path e it is clear that it captures all the information of the field
Aia . In addition it has very simple behavior under the transformations generated
by six of the constraints (24),(25), and (26). For these reasons the holonomy is a
natural choice of basic functional of the connection.

4.1.2. The algebra of basic (kinematic) observables

Shifting the emphasis from connections to holonomies leads to the concept of gen-
eralized connections. A generalized connection is an assignment of he ∈ SU (2) to
any path e ⊂ Σ. In other words the fundamental observable is taken to be the
holonomy itself and not its relationship (64) to a smooth connection. The algebra
of kinematical observables is defined to be the algebra of the so-called cylindrical
functions of generalized connections denoted Cyl. The latter algebra can be written
as the union of the set of functions of generalized connections defined on graphs
γ ⊂ Σ, namely

Cyl = ∪γ Cylγ , (72)

where Cylγ is defined as follows.

A graph γ is defined as a collection of paths e ⊂ Σ (e stands for edge) meeting
at most at their endpoints. Given a graph γ ⊂ Σ we denote by Ne the number of
paths or edges that it contains. An element ψγ,f ∈ Cylγ is labelled by a graph γ
and a smooth function f : SU (2)Ne → C, and it is given by a functional of the
connection defined as

ψγ,f [A] := f (he1 [A], he2 [A], · · · heNe [A]), (73)

where ei for i = 1, · · · Ne are the edges of the corresponding graph γ. The symbol
∪γ in (72) denotes the union of Cylγ for all graphs in Σ. This is the algebra of basic
22 Alejandro Perez

observables upon which we will base the definition of the kinematical Hilbert space
Hkin .
Before going into the construction of the representation of Cyl that defines Hkin ,
it might be useful to give a few examples of cylindrical functions. For obvious reason,
SU (2) gauge invariant functions of the connection will be of particular interest in
the sequel. The simplest of such functions is the Wilson loop: given a closed loop γ
the Wilson loop is given by the trace of the holonomy around the loop, namely
Wγ [A] := Tr[hγ [A]]. (74)
Equation (70) and the invariance of the trace implies the Wγ [A] is gauge invariant.
The Wilson loop Wγ [A] is an element of Cylγ ⊂ Cyl according to the previous
definition. The graph consists of a single closed edge (e = γ) and an example is
shown in Figure 2. Notice, however, that we can also define Wγ [A] as
Wγ [A] := Tr[he1 [A]he2 [A]], (75)
using (68) in which case Wγ [A] ∈ Cylγ ′ (γ ′ is illustrated in the center of Figure 2).
Moreover, we can also think of Wγ [A] ∈ Cylγ ′′ (γ ′′ is shown on the right of Figure 2)
as a function of he1 [A] and he2 [A] and he3 [A] (with trivial dependence on the third
argument). There are many ways to represent an element of Cyl as a cylindrical
function on a graph. This flexibility in choosing the graph will be important in the
definition of Hkin and its inner product in the following section.

111
000
000
111
111
000 111
000
000
111 000
111
e1
e e1
1
0
e2 11
00
11
00
0
1 11
00
1
0
e2 1
0
0
1
00
11
1
0 0
1 11
00
0
1 00
11
e3
111
000
000
111

Fig. 2. An example of three different graphs on which the Wilson loop function (74) can be
defined. The distinction must be physically irrelevant.

Let us come back to our examples. There is a simple generalization of the previ-
ous gauge invariant function. Given an arbitrary representation matrix M of SU (2),
then clearly WγM [A] = Tr[M (he [A])] is a gauge invariant cylindrical function. Uni-
j
tary irreducible representation matrices of spin j will be denoted by Πmm′ for
−j ≤ m, m′ ≤ j. The cylindrical function
j
Wγj [A] := Tr[Π (he [A])] (76)

is the simplest example of spin network function41,42,43,44 . This function is repre-

sented as on the left of Figure 3.
A more sophisticated spin network function can be associated to the graph on
the right of Figure 2. We take different representation matrices of spins 1, 1/2 and
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 23

11
00 111
000
11
00 111
000
11
00
1
00
11
j
j
1/2 1
0 k 11
00
1
0 11
00
1
0 00
11 11
00 11
00
1
0 00
11 11
00 00
11
11
00
00
11
11
00 1/2 11
00
l
11
00 11
00

Fig. 3. Examples of spin networks.

1/2 evaluate them on the holonomy along e1 , e2 , and e3 respectively. We define

1 1/2 1/2
ij AC BD
Θ1,1/2,1/2
e1 ∪e2 ∪e3 [A] =Π (he1 [A]) Π (he2 [A])AB Π (he3 [A])CD σi σj , (77)

where i, j = 1, 2, 3 are vector indices, A, B, C, D = 1, 2 are spinor indices, sum over

repeated indices is understood and σiAC are Pauli matrices. It is easy to check that
Θ1,1/2,1/2
e1 ∪e2 ∪e3 [A] is gauge invariant. This is because the Pauli matrices are invariant
tensors in the tensor product of representations 1 ⊗ 1/2 ⊗ 1/2 which is where gauge
transformations act on the nodes of the graph e1 ∪ e2 ∪ e3 . Such spin network
function is illustrated on the middle of Figure 3. We can generalize this to arbitrary
representations. Given an invariant tensor ι ∈ j ⊗ k ⊗ l the cylindrical function

Θj,k,l
e1 ∪e2 ∪e3
[A] =
j k l
m m m n n n
Π (he1 [A])m1 n1 Π (he2 [A])m2 n2 Π (he3 [A])m3 n3 ι 1 2 3 ι 1 2 3 (78)

is gauge invariant by construction. The example is shown on the right of Figure 3.

p
k s

Fig. 4. Schematic representation of the construction of a spin network. To each node we associate
an invariant vector in the tensor product of irreducible representations converging at the node. In
this case we take ιn1 n2 n3 n4 ∈ j ⊗ k ⊗ p ⊗ s, and the relevant piece of spin network function is
j k p s
Π (he1 [A])m1 n1 Π (he2 [A])m2 n2 Π (he3 [A])m3 n3 Π (he4 [A])m4 n4 ιn1 n2 n3 n4 .

One can generalize the construction of these examples to the definition of spin
networks on arbitrary graphs γ ⊂ Σ. The general construction is analogous to the
one in the previous examples. One labels the set of edges e ⊂ γ with spins {je }.
To each node n ⊂ γ one assigns an invariant tensor, also called an intertwiner,
ιn in the tensor product of representations labelling the edges converging at the
24 Alejandro Perez

corresponding node (see Figure 4). The spin network function is defined
O O je
sγ,{je },{ιn } [A] = ιn Π (he [A]) , (79)
n⊂γ e⊂γ

where the indices of representation matrices and invariant tensors is left implicit in
order to simplify the notation. An example is shown in Figures 4 and 5.
Intertwiners in the tensor product of an arbitrary number of irreducible repre-
sentations can be expressed in terms of basic intertwiners between three irreducible
representations. In the case of SU (2) the latter are uniquely defined up to a nor-
malization; they are simply related to Clebsh-Gordon coefficients—Inv[j1 ⊗ j2 ⊗ j3 ]
is either trivial or one dimensional according to the standard rules for addition of
angular momentum. The construction is illustrated on the left of Figure 5, on the
right we show an explicit example of spin network with the nodes decomposed in
terms of three valent intertwiners..

k 5
2
j
2 1 3
s 3 2 2 1
2 2 2
i
1 1
2

1 1 1
1 1 1

1
p 1

Fig. 5. On the left: any invariant vector can be decomposed in terms of the (unique up to
normalization) three valent ones. At each three node the standard rules of addition of angular
momentum must be satisfied for a non vanishing intertwiner to exist. On the right: an example of
spin network with the explicit decomposition of intertwiners.

So far we have introduced the algebra of functionals of (generalized) connections

Cyl. Spin networks where presented here as special examples of elements of Cyl,
which in addition are SU (2) gauge invariant. In the Subsection 4.1.5 we will show
how spin network functions define a complete basis of Hkin . But in order do that
we must define Hkin .

4.1.3. The Ashtekar-Lewandowski representation of Cyl

The cylindrical functions introduced in the previous section are the candidates for
states in Hkin . In this section we define Hkin and provide a representation of the
algebra Cyl. Essentially what we need is the notion of a measure in the space of
generalized connections in order to give a meaning to the formal expression (62)
and thus obtain a definition of the kinematical inner product. In order to do that
we introduce a positive normalized state (state in the algebraic QFT sense) µAL on
the (C ⋆ -algebra) Cyl as follows. Given a cylindrical function ψγ,f [A] ∈ Cyl (as in
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 25

(73)) µAL (ψγ,f ) is defined as

Z Y
µAL (ψγ,f ) = dhe f (he1 , he2 , · · · heNe ), (80)
e⊂γ

where he ∈ SU (2) and dh is the (normalized) Haar measure of SU (2) l . The measure
µAL is clearly normalized as µAL (1) = 1 and positive
Z Y
µAL (ψγ,f ψγ,f ) = dhe f (he1 , he2 , · · · heNe )f (he1 , he2 , · · · heNe ) ≥ 0. (81)
e⊂γ

Using the properties of µAL we introduce the inner product

< ψγ,f , ψγ ′ ,g >: = µAL (ψγ,f ψγ ′ ,g ) =

Z Y
= dhe f (he1 , · · · heNe )g(he1 , · · · heNe ), (82)
e⊂Γγγ ′

where we use Dirac notation and the cylindrical functions become wave function-
als of the connection corresponding to kinematical states ψγ ′ ,g [A] =< A, ψγ ′ ,g >=
g(he1 , · · · heNe ), and Γγγ ′ is any graph such that both γ ⊂ Γγγ ′ and γ ′ ⊂ Γγγ ′ .
The state µAL is called the Ashtekar-Lewandowski measure40 . The previous equa-
tion is the rigorous definition of (62). The measure µAL —through the GNS
construction45—gives a faithful representation of the algebra of cylindrical func-
tions (i.e., (81) is zero if and only if ψγ,f [A] = 0). The kinematical Hilbert space
Hkin is the Cauchy completion of the space of cylindrical functions Cyl in the
Ashtekar-Lewandowski measure. In other words, in addition to cylindrical func-
tions we add to Hkin the limits of all the Cauchy convergent sequences in the µAL
norm. The operators depending only on the connection act simply by multiplication
in the Ashtekar-Lewandowski representation. This completes the definition of the
kinematical Hilbert space Hkin .

4.1.4. An orthonormal basis of Hkin .

In this section we would like to introduce a very simple basis of Hkin as a prelim-
inary step in the construction of a basis of the Hilbert space of solutions of the
Gauss constraint Hkin
G
At this stage this is a simple consequence of the Peter-Weyl
46
theorem . The Peter-Weyl theorem can be viewed as a generalization of Fourier
theorem for functions on S 1 . It states that, given a function f ∈ L2 [SU (2)], it can

l The Haar measure of SU (2) is defined by the following properties:

Z
dg = 1, and dg = d(αg) = d(gα) = dg −1 ∀α ∈ SU (2).
SU (2)
26 Alejandro Perez

be expressed as a sum over unitary irreducible representations of SU (2), namely m

Xp ′ j
f (g) = 2j + 1 fjmm Πmm′ (g), (83)
j

where
p Z
′ j
fjmm = 2j + 1 dg Πm′ m (g −1 )f (g), (84)
SU(2)

and dg is the Haar measure of SU (2). This defines the harmonic analysis on SU (2).
The completeness relation
X j j X j
δ(gh−1 ) = (2j + 1) Πmm′ (g) Πm′ m (h−1 ) = (2j + 1)Tr[Π (gh−1 )], (85)
j j

follows. The previous equations imply the orthogonality relation for unitary repre-
sentations of SU (2)
Z
′
dg φjm′ m φjq′ q = δjj ′ δmq δm′ q′ , (86)
SU(2)

where we have introduce the normalized representation matrices φjmn :=

√ j
2j + 1 Πmn for convenience. Given an arbitrary cylindrical function ψγ,f [A] ∈ Cyl
we can use the Peter-Weyl theorem and write

ψγ,f [A] = f (he1 [A], he2 [A], · · ·heNe [A]) =

X m ···m ,n ···n j
1 Ne 1 Ne j1
= fj1 ···jNe
φm1 n1(he1 [A]) · · · φmNNe e nNe(heNe [A]), (87)
j1 ···jNe

1 Ne 1 m ···m
Ne ,n ···n
where according to (84) fj1 ···jNe
is just given by the kinematical inner
product of the cylindrical function with the tensor product of irreducible represen-
tations, namely (82)
m ···m ,n1 ···nNe j
1
fj1 ···jNe
Ne
=< φjm1 1 n1 · · · φmNNe e nNe , ψγ,f >, (88)

where <, > is the kinematical inner product introduced in (82). We have thus
proved that the product of components of (normalized) irreducible representations
QNe ji
i=1 φmi ni [hei ] associated with the Ne edges e ⊂ γ (for all values of the spins j
and −j ≤ m, n ≤ j and for any graph γ) is a complete orthonormal basis of Hkin !

m The
P
link with the U (1) case is direct: for f ∈ L2 [U (1)] we have f (θ) =R n fn exp(inθ), where
exp(inθ) are unitary irreducible representations of U (1) and fn = (2π) −1 dθexp(−inθ)f (θ). The
measure (2π)−1 dθ is the Haar measure of U (1).
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 27

4.1.5. Solutions of the Gauss constraint: Hkin

G
and spin network states
We are now interested in the solutions of the quantum Gauss constraint; the first
three of quantum Einstein’s equations. These solutions are characterized by the
states in Hkin that are SU (2) gauge invariant. These solutions define a new Hilbert
space that we call Hkin
G
. We leave the subindex kin to keep in mind that there are
still constraints to be solved on the way to Hphys . In previous sections we already
introduced spin network states as natural SU (2) gauge invariant functionals of the
connection42,43,44,47 . Now we will show how these are in fact a complete set of
orthogonal solutions of the Gauss constraint, i.e., a basis of Hkin
G
.
The action of the Gauss constraint is easily represented in Hkin . At this stage
it is simpler to represent finite SU (2) transformations on elements of Hkin (from
which the infinitesimal ones can be easily inferred) using (70). Denoting UG [g] the
operator generating a local g(x) ∈ SU (2) transformation then its action can be
defined directly on the elements of the basis of Hkin defined above, thus
UG [g]φjmn [he ] = φjmn [gs he gt−1 ], (89)
where gs is the value of g(x) at the source point of the edge e and gt the value
of g(x) at the target. From the previous equation one can infer the action on an
arbitrary basis element, namely
Ne
Y Ne
Y
UG [g] φjmi i ni [hei ] = φjmi i ni [gsi hei gt−1
i
]. (90)
i=1 i=1

Now by definition of the scalar product (82) and due to the invariance of the Haar
measure (see Footnote m) the reader can easily prove that UG [g] is a unitary oper-
ator. From the definition it also follows that
UG [g2 ]UG [g1 ] = UG [g1 g2 ]. (91)
The projection operator onto the set of states that are solutions of the Gauss con-
straint can be obtained by group averaging techniques. We can denote the projector
PG by
Z
PG = D[g] UG [g], (92)

where the previous expression denotes a formal integration over all SU (2) trans-
formations. Its rigorous definition is given by its action on elements of Cyl. From
equation (89) the operator UG [g] acts on ψγ,g ∈ Cyl at the end points of the edges
e ⊂ γ, and therefore, so does PG . The action of PG on a given (cylindrical) state
ψγ,f ∈ Hkin can therefore be factorized as follows:
Y
PG ψγ,f = PGn ψγ,f , (93)
n⊂γ

where PGn acts non trivially only at the node n ⊂ γ. In this way we can define the
action of PG by focusing our attention to a single node n ⊂ γ. For concreteness let
28 Alejandro Perez

us concentrate on the action of PG on an element of ψγ,f ∈ Hkin defined on the

graph illustrated in Figure 4. The state ψγ,f ∈ Hkin admits an expansion in terms
of the basis states as in (87). In particular we concentrate on the action of PG at the
four valent node thereto emphasized, let’s call it n0 ⊂ γ. In order to do that we can
factor out of (87) the (normalized) representation components φjmn corresponding
to that particular node and write
ψγ,f [A] =
X
m1 ···m4 ,n1 ···n4
= φjm1 1 n1(he1 [A]) · · ·φjm4 4 n4(he4 [A]) × [REST]j1 ···j4
[A], (94)
j1 ···j4
m ···m ,n ···n
where [REST]j1 1···j4 4 1 4 [A] denotes what is left in this factorization, and e1 to
e4 are the four edges converging at n0 (see Figure 4). We can define the meaning of
(92) by giving the action of PGn0 on φjm1 1 n1(he1 [A]) · · ·φjm4 4 n4(he4 [A]) as the action on
a general state can be naturally extended from there using (93). Thus we define
Z
PG φm1 n1(he1 [A]) · · ·φm4 n4(he4 [A]) = dgφjm1 1 n1(ghe1 [A]) · · ·φjm4 4 n4(ghe4 [A]), (95)
n0 j1 j4

where dg is the Haar measure of SU (2). Using the fact that

j
φjmn
1
(gh[A]) =Πmq (g) φjqn(h[A]) (96)
the action of PGn0 can be written as
PGn0 φjm1 1 n1(he1 [A]) · · ·φjm4 4 n4(he4 [A]) =
n0
= Pm 1 ···m4 ,q1 ···q4
φjq11 n1(he1 [A]) · · ·φjq44 n4(he4 [A]), (97)
where
Z
j1 j4
n0
Pm 1 ···m4 ,q1 ···q4
= dg Πm1 q1 (g) · · · Πm4 q4 (g). (98)

If we denote Vj1 ···j4 the vector space where the representation j1 ⊗ · · · ⊗ j4 act, then
previous equation defines a map P n0 : Vj1 ···j4 → Vj1 ···j4 . Using the properties of
the Haar measure given in Footnote m one can show that the map P n0 is indeed a
projection (i.e., P n0 P n0 = P n0 ). Moreover, we also have
j1 j4
n0
Pm 1 ···m4 ,q1 ···q4
Πq1 n1 (g) · · · Πq4 n4 (g) =
j1 j4
n0
=Πm1 q1 (g) · · · Πm4 q4 (g)Pqn10···q4 ,n1 ···n4 = Pm 1 ···m4 ,n1 ···n4
, (99)
i.e. P n0 is right and left invariant. This implies that P n0 : Vj1 ···j4 → Inv[Vj1 ···j4 ],
i.e., the projection from Vj1 ···j4 onto the (SU (2)) invariant component of the finite
dimensional vector space. We can choose an orthogonal set of invariant vectors
ια
m1 ···m4 (where α labels the elements), in other words an orthonormal basis for
Inv[Vj1 ···j4 ] and write
X
n0
Pm 1 ···m4 ,n1 ···n4
= ια α∗
m1 ···m4 ιm1 ···m4 , (100)
α
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 29

where ∗ denotes the dual basis element. In the same way the action PG on a node
n ⊂ γ of arbitrary valence κ is governed by the corresponding P n given generally
by
X
n
Pm 1 ···mκ ,n1 ···nκ
= ια ακ ∗
m1 ···mκ ιm1 ···mκ .
κ
(101)
ακ

According to the tree decomposition of intertwiners in terms of three valent invariant

vectors described around Figure 5, ακ is a (κ − 3)-uple of spins.
Any solution of the Gauss constraint can be written as PG ψ for ψ ∈ Hkin . Equa-
tion (93) plus the obvious generalization of (97) for arbitrary nodes implies that the
result of the action of PG on elements of Hkin can be written as a linear combina-
tion of products of representation matrices φjmn contracted with intertwiners, i.e.
spin network states as introduced as examples of elements of Cyl in (79). Spin net-
work states therefore form a complete basis of the Hilbert space of solutions of the
quantum Gauss law Hkin G
!

4.2. Geometric operators: quantization of the triad

We have introduced the set of basic configuration observables as the algebra of
cylindrical functions of the generalized connections, and have defined the kinemati-
cal Hilbert space through the Ashtekar-Lewandowski representation. By considering
finite gauge transformations we avoided quantizing the Gauss constraint in the pre-
vious section, avoiding for simplicity, and for the moment, the quantization of the
triad field Eia present in (26). In this section we will quantize the triad field and
will define a set of geometrical operators that lead to the main physical prediction
of LQG: discreteness of geometry eigenvalues.
The triad Eia naturally induces a two form with values in the Lie algebra of
SU (2), namely, Eia ǫabc . In the quantum theory Eia becomes an operator valued
distribution. In other words we expect integrals of the triad field with suitable test
functions to be well defined self adjoint operators in Hkin . The two form naturally
associated to Eia suggests that the smearing should be defined on two dimensional
surfaces:
Z a b Z a b
b α] = dσ 1 dσ 2 ∂x ∂x αi E
E[S, ca ǫabc = −i~κγ dσ 1 dσ 2 ∂x ∂x αi δ ǫabc ,
i
∂σ 1 ∂σ 2 ∂σ 1 ∂σ 2 δAic
S S
(102)
where αi is a smearing function with values on the Lie algebra of SU (2). The
previous expression corresponds to the natural generalization of the notion of electric
flux operator in electromagnetism. In order to study the action of (102) in Hkin we
notice that
 Z 
δ δ d k
h e [A] = P exp ds ẋ (s)A τ
d k =
δAic δAic
Z
= ds ẋc (s)δ (3) (x(s) − x)he1 [A]τi he2 [A], (103)
30 Alejandro Perez

where he1 [A] and he2 [A] are the holonomy along the two new edges defined by the
point at which the triad acts. Therefore
b α]he [A] =
E[S,
Z
∂xa ∂xb ∂xc
= −i8πℓ2p γ dσ 1 dσ 2 dσ 3 1 2 ǫabc δ (3) (x(σ), x(s))αi he1 [A]τi he2 [A]. (104)
∂σ ∂σ ∂s
Finally using the definition of the delta distribution we obtain a very simple expres-
sion for the action of the flux operator on the holonomy integrating the previous
expression. In the cases of interest the result is:

e2

b α]he [A] = −i8πℓ2pγαi he1 [A]τi he2 [A]

E[S, , (105)
e1

and
e1

b α]he [A] = i8πℓ2 γαi he1 [A]τi he2 [A]

E[S, , (106)
p

e2

b α]he1 [A] = 0 when e is tangential to S or e ∩ S = 0.

and E[S,
From its action on the holonomy, and using basic SU (2) representation theory,
b α] to spin network states and therefore to
one can easily obtain the action of E[S,
b α] is self-adjoint. The
any state in Hkin . Using (82) one can also verify that E[S,
b α] for all
reader can also verify that it is SU (2) gauge covariant. The operators E[S,
surfaces S and all smearing functions α contain all the information of the quantum
Riemannian geometry of Σ. In terms of the operators E[S, b α] we can construct any
geometric operator.

4.2.1. Quantization of the Gauss constraint

We have already imposed the Gauss constraint in Section 4.1.5 by direct construc-
tion of the Hilbert space of SU (2) gauge invariant states Hkin
G
with a natural com-
plete basis given by the spin network states. Here we show that an important identity
for flux operators follows from gauge invariance. Given a spin network node n where
Nn edges converge, take a sphere S of radius ǫ (defined in some local coordinates)
centered at the corresponding node (see Figure 6). The identity follows from the
fact that for a gauge invariant node
Nn
X
bS (α)|ψ >=
lim E bei (α)|ψ >= 0,
E (107)
ǫ→0
i=1
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 31

where Nn is the number of edges at the node, and Eei (α) is the flux operator through
a piece of the sphere that is punctured by only the edge ei . If we partition the sphere

Fig. 6. The Gauss constraint imposes the net flux of non Abelian electric field be zero around
nodes.

in pieces that are punctured by only one edge, using (105) at each edge one notices
that (107) produces the first order term in an infinitesimal gauge transformation
gα = 1 − αi τi ∈ SU (2) at the corresponding node: the operator acting in (107)
is indeed the quantum Gauss constraint action on the given node! Because the
node is gauge invariant the action of the such operator vanishes identically. The
total quantum flux of non-Abelian electric field must vanish according the Gauss
constraint.

4.2.2. Quantization of the area

The simplest of the geometric operators is the area of a two-dimensional
surface21,22,48 S ⊂ Σ which classically depends on the triad Eia as in (54). If
we introduce a decomposition of S in two-cells, we can write the integral defining
the area as the limit of a Riemann sum, namely
AS = lim AN
S (108)
N →∞

where the Riemann sum can be expressed as

N p
X
AN
S = Ei (SI )E i (SI ) (109)
I=1

where N is the number of cells, and Ei (SI ) corresponds to the flux of Eia through
the I-th cell. The reader is invited to check that the previous limit does in fact
define the area of S in classical geometry. The previous expression for the area sets
the path to the definition of the corresponding quantum operator as it is written in
terms of the flux operators that we defined in the previous section. The quantum
area operator then simply becomes
bS = lim A
A bN
S, (110)
N →∞
32 Alejandro Perez

where we simply replace the classical Ei (SI ) by E bi (SI ) according to (102). The
bi (SI )E
important action to study is that of E b (SI ) which on the holonomy along a
i

path that crosses SI only once is

bi (SI )E
E b i (SI )he [A] = i2 (8πℓ2 γ)2 he1 [A]τi τ i he2 [A] = (8πℓ2 γ)2 (3/4)he [A], (111)
p p

where we have used that τi = iσi /2. The action of the square of the flux through the
cell SI is diagonal on such holonomy! Using the definition of the unitary irreducible
representation of SU (2) it follows that
bi (SI )E
E bi (SI )φj (he [A])mn = (8πℓ2p γ)2 (j(j + 1))φj (he [A])mn , (112)
when the edge is that of an arbitrary spin network state. The remaining important
case is when a spin network node is on SI . A careful analysis shows that the action
is still diagonal in this case48 . Notice that the cellular decomposition is chosen so
that in the limit N → ∞ each SI is punctured at most at a single point by either
an edge (the case studied here) or a node.

m
s
l
u
sI j

Fig. 7. On the left: the regularization of (110) is defined so that in the limit in which the two cells
are shrunk they are punctured by at most one edge. Center: The simplest eigenstate of the area
of S is illustrated, a two cell at finite regulator is emphasized. On the right: a generic eigenstate
of the area.

It is then a straightforward exercise to show that the action of the area operator
is diagonalized by the spin network states. Spin network states are the eigenstates
of the quantum area operator! We have
p
bS |ψ >= 8πℓ2 γ j(j + 1)|ψ >,
A (113)
p

for a single puncture and more generally

Xq
bS |ψ >= 8πℓ2 γ
A jp (jp + 1)|ψ > . (114)
p
p

The eigenvalues when nodes lay on S are also know in closed form. We do not ana-
lyze that case here for lack of space; however, the eigenvalues can be computed in a
direct manner using the tools that have been given here. The reader is encouraged
to try although the full answer can be found explicitly in the literature48 . Notice
that the spectrum of the area operator depends on the value of the Immirzi param-
eter γ (introduced in (21)). This is a general property of geometric operators. The
spectrum of the area operator is clearly discrete.
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 33

4.2.3. Quantization of the volume

The volume of a three dimensional region B ⊂ Σ is classically given by
Z
√ 3
VB = q d x, (115)
B
and it can be expressed in terms of the triad operator. In fact using (12) we conclude
that
1
q = |det(E)| = ǫabc Eia Ejb Ejc ǫijk . (116)
3!
Therefore,
Z s
1
VB = ǫabc Eia Ejb Ejc ǫijk d3 x (117)
B 3!
Following the similar regularization techniques as in the case of the area operator
we write the previous integral as the limit of Riemann sums defined in terms of a
decomposition of B in terms of three-cells (think of a cubic lattice for concreteness),
then we quantize the regularized version21,22,49,50 using the fundamental flux op-
erators of the previous section associated to infinitesimal two cells. Explicitly,
VbB = lim VbBN (118)
N →∞
where
s
N
X 1
VbBN = bi (S a )E
ǫabc E I
bj (S b )E
I
bk (S c ) ǫijk
I (119)
3!
I
a
The way the surfaces SI are chosen for a given three-cell I is illustrated on the

s 2I

s1I
s3I

Fig. 8. Regularization of the volume operator. The infinitesimal two cells on which Ei (SIa ) are
defined. The cellular decomposition is shrunk in the limiting process so that spin network nodes
are contained in only one three cell.

left of Figure 8. The limit N → ∞ is taken by keeping spin network nodes inside a
single three-cell.
Let us finish this section by mentioning some general properties of the volume
operator.
34 Alejandro Perez

(i) There are at least two consistent quantization of the volume operator. One
is sensitive to the differential structure at the node50 while the other is fully
combinatorial22. The quantization of the scalar constraint first introduced by
Thiemann51 uses the version of volume operators that is sensitive to the dif-
ferential structure at the node.
(ii) Three valent nodes are annihilated by the volume operator. This is a simple
consequence of the Gauss constraint. The identity (107) implies that for a three
valent node one can write one of the flux operators in (119) as a linear com-
bination of the other two. The ǫabc in (119) makes the action of the operator
equal to zero.
(iii) The action of the volume operator vanishes on nodes whose edges lie on a plane,
i.e., planar nodes.
(iv) The spectrum of the volume is discrete. The eigenvalue problems is how-
ever more involved and an explicit closed formula is known only in special
cases49,52,53,54 . Recently new manipulations have lead to simplifications in
the spectral analysis of the volume operator55.

4.2.4. Geometric interpretation of spin network states

The properties of the area and volume operator provide a very simple geometrical
interpretation of spin network states. Edges of spin networks carry quanta of area
while volume quantum numbers can be used to label nodes (plus some additional
label to resolve degeneracy when needed). This interpretation is fully background
independent and its combinatorial character implies that the geometric information
stored in a quantum state of geometry is intrinsically diffeomorphism invariant. We
can visualize a spin network as a polymer like excitation of space geometry consisting
of volume excitations around spin network nodes connected by spin network links
representing area excitations of the (dual) surfaces separating nodes. The picture is
in complete agreement with the general direction in which a background indepen-
dent formulation would be set up. The embedding or the coordinate system that we
choose to draw the spin network graph does not carry any physical information. All
the information about the degrees of freedom of geometry (hence the gravitational
field) is contained in the combinatorial aspects of the graph (what is connected
to what) and in the discrete quantum numbers labelling area quanta (spin labels
of edges) and volume quanta (linear combinations of intertwiners at nodes). The
fact that spin networks up to their embedding are physically relevant in LQG will
become more clear in the next section, when we solve the diffeomorphism constraint.

4.2.5. Continuum geometry

After the discovery that geometric operators have discrete spectra an obvious ques-
tion is whether the discreteness is compatible with the smooth geometry picture of
the classical theory. One can in fact check that the spectrum crowds very rapidly
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 35

when one gets to larger geometries as the spacing between eigenvalues decreases
exponentially for large eigenvalues.

4.3. Solutions of the diffeomorphism constraint: HD

kin and abstract
spin networks
In Section 4.1.5 we solved the three quantum Einstein’s equations defining the
Hilbert space of SU (2) gauge invariant states given by Hkin G
⊂ Hkin . Now we turn
our attention to the next three quantum constraint equations, the vector constraint
(24). The technique that we apply to find the Hilbert space of diffeomorphism
invariant states Hkin
D
is analogous to the one used to obtain HkinG
. However, because
the orbits of the diffeomorphisms are not compact, diffeomorphism invariant states
are not contained in the original Hkin . In relation to Hkin , they have to be regarded
as distributional states40 .
A simple example will allow us to introduce the basic ideas. Consider an ordinary
particle quantum system defined on a cylinder. Assume the kinematical Hilbert
space is given by square integrable functions ψ(θ, z) or Hkin = L2 (S 1 × R). In
addition assume one has to solve the following constraint equations:

pbθ ψ = 0 and pbz ψ = 0. (120)

The first constraint generates rotations around the z-axis, i.e. it has compact orbits,
and, in this sense, is the analog of the Gauss constraint in LQG. The solutions of
the first constraint are wave functions invariant under rotations around z, those
that do not depend on θ. Therefore they are contained in the original Hilbert space
Hkin because, as the orbits of pθ are compact, square integrable functions ψ(z) (i.e.,
independent of θ) exist. The second constraint has non compact orbits and in this
sense is the analog of the diffeomorphism constraint in LQG. The solution of the
second constraint are functions that do not depend on z. They cannot be contained
in Hkin , as functions ψ(θ) cannot be in Hkin = L2 (S 1 ×R). However given a suitable
dense subset of Φ ⊂ Hkin of test functions, e.g. functions of compact support, then
any solution of the latter constraint can be given a meaning as a distribution. For
instance a solution of both constraints does not depend neither on θ nor on z. Its
wave function corresponds to a constant (ψ0 |θ, z >= c . The solution (ψ0 | is clearly
not in Hkin . We use a rounded brackets in the notation to recall this fact. (ψ0 | is
in Φ⋆ , the topological dual of Φ, i.e., the set of linear functionals from Φ to C. Its
action on any arbitrary function of compact support |α >∈ Φ ⊂ Hkin is given by
Z
(ψ0 |α >= dz dθ c α(θ, z), (121)

which is well defined. The action of (ψ0 | extracts the gauge invariant information
from the non gauge invariant state |α >. As vector spaces we have the relation Φ ⊂
Hkin ⊂ Φ⋆ , usually called the Gelfand triple. In the case of LQG diffeomorphism
invariant states are in the dual of the cylindrical functions Cyl. The Gelfand triple of
36 Alejandro Perez

interest is Cyl ⊂ Hkin ⊂ Cyl⋆ . Diffeomorphism invariant states have a well defined
meaning as linear forms in Cyl⋆ .
Let us now apply the same idea to define diffeomorphism invariant states. Dif-
feomorphism transformations are easily represented in Hkin . We denote UD [φ] the
operator representing the action of a diffeomorphism φ ∈ Dif f (Σ). Its action can
be defined directly on the dense subset of cylindrical functions Cyl ⊂ Hkin . Given
ψγ,f ∈ Cyl as in (73) we have

UD [φ]ψγ,f [A] = ψφ−1 γ,f [A], (122)

which naturally follows from (71). Diffeomorphisms act on elements of Cyl (such as
spin networks) by simply modifying the underlying graph in the obvious manner.
Notice that UD [φ] is unitary according to the definition (82).
Notice that because (122) is not weakly continuous there is no well defined
notion of self-adjoint generator of infinitesimal diffeomorphisms in Hkin n . In other
words the unitary operator that implements a diffeomorphism transformation is well
defined but there is no corresponding self adjoint operator whose exponentiation
leads to UD [φ]. Therefore, the diffeomorphism constraint cannot be quantized in
the Ashtekar-Lewandowski representation. This is not really a problem as UD [φ] is
all we need in the quantum theory to look for diffeomorphism invariant states. In
LQG one replaces the second set of formal equations in (63) by the well defined
equivalent requirement

UD [φ]ψ = ψ (123)

for (distributional states) ψ ∈ Cyl⋆ .

We are now ready to explicitly write the solutions, namely
X X
([ψγ,f ]| = < ψγ,f |UD [φ] = < ψφγ,f |, (124)
φ∈Dif f (Σ) φ∈Dif f (Σ)

were the sum is over all diffeomorphisms. The brackets in ([ψγ,f ]| denote that the
distributional state depends only on the equivalence class [ψγ,f ] under diffeomor-
phisms. Clearly we have ([ψγ,f ]|UD [α] = ([ψγ,f ]| for any α ∈ Dif f (Σ). Now we need
to check that ([ψγ,f ]| ∈ Cyl⋆ so that the huge sum in (124) gives a finite result when
applied to an element |ψγ ′ ,g >∈ Cyl, i.e., it is a well defined linear form. That this
is the case follows from (82) (see the remark in Footnote n) as in

([ψγ,f ]|ψγ ′ ,g > (125)

only a finite number of terms from (124) contribute. In fact for spin networks with
no discrete symmetries there is only one non trivial contribution.
The action of ([ψγ,f ]| is diffeomorphism invariant, namely

([ψγ,f ]|ψγ ′ ,g >= ([ψγ,f ]|UD [φ]ψγ ′ ,g > (126)

n This is because for any φ ∈ Dif f (Σ) the state UD [φ]ψ is orthogonal to ψ for a generic ψ ∈ Hkin .
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 37

The inner product < , >dif f needed to promote the set of diffeomorphism invariant
states to the Hilbert space Hkin
D
is defined as
< [ψγ,f ]|[ψγ ′ ,g ] >dif f = ([ψγ,f ]|ψγ ′ ,g > (127)
Due to (126) the previous equation is well defined among diffeomorphism equiva-
lence classes of states under the action of diffeomorphisms and hence this is denoted
by the brackets on both sides.

4.3.1. The quantization of the scalar constraint

In this section we sketch the regularization and definition of the last constraint to
be quantized and solved: the scalar constraint. The smeared version of the classical
scalar constraint is
Z
Eia Ejb  ij k j

S(N ) = dx3 N p ǫ k Fab − 2(1 + γ 2 )K[a
i
Kb] =
det(E)
Σ
= S (N ) − 2(1 + γ 2 ) T (N ),
E
(128)
where Kai =γ −1
(Aia − Γia ),
and in the second line we have introduced a convenient
separation of the constraint into what is called the Euclidean contribution S E (N )
given by
Z
Eia Ejb ij k
S E (N ) = dx3 N p ǫ k Fab , (129)
det(E)
Σ

and the extra piece

Z
Eia Ejb j
T (N ) = dx3 N p i
K[a Kb] . (130)
det(E)
Σ

The terms in the constraint look in fact very complicated. On the one hand they
are highly non linear which anticipates difficulties in the quantization related to
regularization issues and potential UV divergences, factor ordering ambiguities, etc.
For instance the factor 1/det(E) looks quite complicated at first sight as does the
spin connection Γia in the expression of Kai (recall its definition (19) and (20) in
terms of the basic triad variables).
The crucial simplification of the apparently intractable problem came from the
ideas of Thiemann56 . He observed that if one introduces the phase space functional

Z
K̄ := Kai Eia , (131)
Σ

then the following series of identities hold:

1  i
Kai = γ −1 (Aia − Γia ) = Aa , K̄ , (132)
κγ
38 Alejandro Perez

1  E
K̄ = S (1), V , (133)
γ 3/2
Rp
where V = det(E) is the volume of Σ, and finally
EbEc 4  k
p i j ǫijk ǫabc = Aa , V . (134)
det(E) κγ
With all this we can write the terms in the scalar constraint by means of Poisson
brackets among quantities that are simple enough to consider their quantization.
The Euclidean constraint can be written as
Z
 j
S (N ) = dx3 N ǫabc δij Fab
E i
Ac , V , (135)
Σ

while the term T (N ) becomes

Z
N    
T (N ) = dx3 2 3 ǫabc ǫijk Aia , S E (1), V {Ajb , S E (1), V } Akc , V . (136)
κ γ
Σ

The new form suggests that we can quantize the constraint by promoting the ar-
gument of the Poisson brackets to operators and the Poisson brackets them self
to commutators in the standard way. One needs the volume operator V , whose
quantization was already discussed, and the quantization of the connection and
curvature. We present here the basic idea behind the quantization of these. For a
precise treatment the reader is encouraged to read Thiemann’s original work51,57
and book20. Given an infinitesimal loop αab on the ab-plane with coordinate area
ǫ2 , the curvature tensor can be regularized observing that
hαab [A] − h−1 2 i 4
αab [A] = ǫ Fab τi + O(ǫ ). (137)
 i
Similarly the Poisson bracket Aa , V is regularized as
 i
h−1
ea [A] {hea [A], V } = ǫ Aa , V + O(ǫ ),
2
(138)
where ea is a path along the a-coordinate of coordinate length ǫ. With this we can
write
S E (N ) =
X
= lim NI ǫ3 ǫabc Tr [Fab (A) {Ac , V }] =
ǫ→0
I
X h  i
= lim NI ǫabc Tr (hαIab [A] − h−1
αI [A])h −1
eI [A] h eI [A], V
c
, (139)
ǫ→0 ab c
I

where in the first equality we have replaced the integral (135) by a Riemann sum
over cells of coordinate volume ǫ3 and in the second line we have written it in
terms of holonomies. Notice that the explicit dependence on the cell size ǫ has
disappeared in the last line. The cells are labelled with the index I in analogy to
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 39

e1
e1 α 31
e3
e3 α 23
e2 e2 α 12

Fig. 9. A tetrahedral infinitesimal cell adapted to a four valent spin network node. On the right:
the three non trivial contributions of the cell to the regularized expression (140).

the regularization used for the area and volume in previous sections. The loop αIab
is an infinitesimal closed loop of coordinate area ǫ2 in the ab-plane associated to the
I-th cell, while the edge eIa is the corresponding edge of coordinate length ǫ dual
to the ab-plane (see Figure 9 for a cartoon of the regularization first introduced by
Thiemann). The idea now is to promote this regulated expression to an operator by
quantizing the ingredients of the formula: notice that we already know how to do
that as the expression involves the holonomies and the volume, both well defined
operators in Hkin . The quantum constraint can formally be written as

X h n oi
SbE (N ) = lim NI ǫabc Tr (b
hαIab [A] − b
h−1
αI
[A])b
he−1
I [A]
b
heIc [A], Vb . (140)
ǫ→0 ab c
I

Now in order to have a rigorous definition of SbE (N ) one needs to show that the
previous limit exists in the appropriate Hilbert space.
It is useful to describe some of the qualitative features of the argument of the
limit in (140) which we refer to as the regulated quantum scalar constraint and we
denote Sbǫ (N ). It is easy to see that the regulated quantum scalar constraint acts
only on spin network nodes, namely

X
Sbǫ (N )ψγ,f = Nn Sbǫn ψγ,f , (141)
nγ

where Sbǫn acts only on the node n ⊂ γ and Nn is the value of the lapse N (x) at
the node. This is a simple consequence of the very same property of the volume
operator (118). Due to the action of the infinitesimal loop operators representing
the regularized curvature, the scalar constraint modifies spin networks by creating
new links around nodes whose amplitudes depend on the details of the action of
the volume operator, the local spin labels and other local features at the nodes. If
we concentrate on the Euclidean constraint, for simplicity, its action on four valent
nodes can be written as
40 Alejandro Perez

k k

l l
P p
Sbǫn j = op Sjklm,opq j q
+
o
m m
k k
q
p l p l
P q P
+ op Sjlmk,opq j + op Sjmkl,opq j o (142),
o
m m

where q = 1/2 in the case of (140) (we will see in the sequel that q can be any arbi-
trary spin: this is one of the quantization ambiguities in the theory), and Sjklm,opq
are coefficients that are independent of ǫ and can be computed explicitly from (140).
Now we analyze the removal of the regulator. Since the only dependence of ǫ
is in the position of the extra link in the resulting spin network states, the limit
ǫ → 0 can be defined in the Hilbert space of diffeomorphism invariant states Hkin D
.
The key property is that in the diffeomorphism invariant context the position of
the new link in (142) is irrelevant. Therefore, given a diffeomorphism invariant state
([φ]| ∈ Hkin
D
∈ Cyl⋆ , as defined as in (124), the quantity (φ|Sbǫ (N )|ψ > is well defined
and independent of ǫ. In other words the limit
b )|ψ >= lim (φ|Sbǫ (N )|ψ >
(φ|S(N (143)
ǫ→0

exists trivially for any given ψ ∈ Hkin . A careful analysis on how this limit is defined
can be found in20 .
An important property of the definition of the quantum scalar constraint is that
the new edges added (or annihilated) are of a very special character. For instance,
not only do the new nodes in (142) carry zero volume but also they are invisible
to the action of the quantum scalar constraint. The reason for that is that the
new three-valent nodes are planar and therefore the action of the commutator of
the holonomy with the volume operator in (140) vanishes identically (recall the
properties of the volume operator at the end of Section 4.2). For this reason it is
useful to refer to these edges as exceptional edges.
This property of Thiemann’s constraint is indeed very important for the consis-
tency of the quantization. The non trivial consistency condition on the quantiza-
tion of the scalar constraint corresponds to the quantum version of (41). The correct
commutator algebra is satisfied in the sense that for diffeomorphism invariant states
(φ| ∈ Hkin
D
(defined as in (124))
b ), S(M
(φ|[S(N b )]ψ >= 0, (144)
for any ψ ∈ HkinG
. The l.h.s. vanishes due to the special property of exceptional
edges as can be checked by direct calculation. Notice that r.h.s. of (41) is expected
to annihilate elements of Hkin
D
—at least for the appropriate factor ordering— so
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 41

that the previous equation is in agreement with (41) and the quantization is said to
be anomaly-free58 . All known59,60,61 consistent quantizations satisfy the Abelian
property (144) in Hkin D
. We will come back to this issue in the next Subsection.
Notice that (134) is the co-triad eia according to (20); this opens the way for
the quantization of the metric qab = eia ejb δij that is necessary for the inclusion of
matter. A well defined quantization of the scalar constraint including Yang-Mills
fields, scalar fields and fermions has been put forward by Thiemann62 .

4.3.2. Solutions of the scalar constraint, physical observables, difficulties

Here we briefly explore some of the generic consequences of the theory constructed so
far. The successful definition of the quantum scalar constraint operator including the
cases with realistic matter couplings is a remarkable achievement of loop quantum
gravity. There are however some issues that we would like to emphasize here.
There is a large degree of ambiguity on the definition of the quantum scalar
constraint. The nature of solutions or the dynamics seems to depend critically on
these ambiguities. For instance it is possible to arrive at a completely consistent
quantization by essentially replacing the holonomies in (140)—defined in the fun-
damental representation of SU (2)—by the corresponding quantities evaluated on
an arbitrary representation59 . In the applications of the theory to simple systems
such as in loop quantum cosmology this is known to have an important physical
effect63 . Ambiguities are also present in the way in which the paths defining the
holonomies that regularize the connection Aia and the curvature Fab i
(A) in (140)
are chosen. See for instance Section C in19 for an alternative to Thiemann’s pre-
scription and a discussion of the degree of ambiguity involved. There are factor
ordering ambiguities as well, which is evident from (140). Therefore instead of a
single theory we have infinitely many theories which are mathematically consistent.
A yet unresolved issue is whether any of these theories is rich enough to reproduce
general relativity in the classical continuum limit.

m
p
n k
j

q
l o

Fig. 10. Solutions of the scalar constraint as dressed (diff-invariant) spin networks.

All the known consistent quantizations of the scalar constraint satisfy property
(144). Quantizations that satisfy this property seem to share a property that is often
42 Alejandro Perez

referred to as ultra-locality64. This property can be illustrated best in terms of the

kind of solutions of the scalar constraint. We will keep the discussion here as general
as possible; therefore, we will study generic features of formal solutions. We should
point out that exact solutions are known for specific quantizations of the scalar
constraint. For instance, an algorithm for constructing the general solution of the
quantum scalar constraint is described in detail by Ashtekar and Lewandowski19
in terms of the quantum operator S(N b ) introduced therein (see also 65 ). These
solutions satisfy the property described below for generic formal solutions.
Quantization of the scalar constraint satisfying (144) act on spin network nodes
by adding (and/or annihilatingo ) exceptional edges. As explained above these ex-
ceptional edges are characterized by being invisible to subsequent actions of the
constraint. These exceptional edges are added (or destroyed) in the local vicinity of
nodes. For that reason, solutions of the scalar constraint can be labelled by graphs
with ‘dressed’ nodes as the one illustrated in Figure 10. The shadowed spheres
denote certain (generally infinite) linear combinations of spin networks with excep-
tional edges, diagrammatically

Ω
Ο
=

=α +β + ··· + γ + ···+ δ + · · · (145),

where we have simplified the notation by dropping the spin labels and the coefficients
α, β · · · δ depend on the details of the definition of the scalar constraint and the spin
labelling of the corresponding spin networks. The Greek letter Ω denotes the possible
set of quantum numbers labelling independent solutions at the dressed nodes.
Notice that the generic structure of these formal solutions—which is shared by
the exact solutions of explicit quantizations—is on the one hand very appealing.
The solutions of all the constraints of LQG seem quite simple: they reduce to sim-
ple algebraic relations to be satisfied by the coefficients of (145). In some explicit
cases19, these relations even reduced to finite dimensional matrix operations.
The structure of solutions also suggest the possibility of defining a large variety
of Dirac (i.e. physical) observables. For instance, the fact that generic solutions can
be characterized by dressed spin-network states as in Figure 10 implies that the
spin labelling the links joining different dressed nodes are indeed quantum num-
bers of Dirac observables (the operator corresponding to these quantum numbers
evidently commute with the action of the constraint). There are infinitely many
Dirac (quasi-local) observables that one can construct for a given quantization of

o Theversion of quantum scalar constraint whose action is depicted in (142) is not self adjoint.
One can introduce self adjoint definitions which contain a term that creates exceptional edges and
another one that destroys them.
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 43

the scalar constraint satisfying (144)p . However, at present it is not very clear what
the physical interpretation of these observables would be.
On the other hand, the ultra-local character of the solutions has raised some
concerns of whether the quantum theories of the Thiemann type can reproduce the
physics of gravity in the classical limit64 . In order to illustrate this point let us go
back to the classical theory. Since at the classical level we can invert the transfor-
mations that lead to new variables in Section 2 let us work with the constraints in
ADM variables. We simplify the discussion by considering the York-Lichnerowicz
conformal decompositions for initial data. For the discussion here we can specialize
to the time-symmetric case Kab = 0 and we take the ansatz qab = ψ 4 qab 0
for some
0
given qab defined up to a conformal factor. The vector constraint is in this case
identically zero and the only non trivial constraint is the scalar constraint that can
be shown to be
1
∆0 ψ − R(q 0 ) = 0, (146)
8
where ∆0 is the covariant Laplacian defined with respect to qab 0
. The point is that
the previous equation is manifestly elliptic which is a general property of the scalar
constraint written in this form67 . This means that if we give the value of ψ on a
sphere, the scalar constraint (146) will determine the value of ψ inside. The scalar
constraint in general relativity imposes a condition among unphysical degrees of
freedom (represented by ψ in this case) that “propagate” along the initial value
surfaceq . The only point in writing the initial value problem in this way is to em-
phasize this property of the scalar constraint.
Coming back to the quantum theory one can construct semi-classical states by
taking linear combinations of kinematical spin-network states in order to approx-
imate some classical geometry qab 68,69 . In that context one could define a sphere
with some given semi-classical area. Now because the scalar constraint acts only in
the immediate vicinity of nodes and does not change the value of the spins of the
edges that connect different dressed nodes it is not clear how the elliptic character
of the classical scalar constraint would be recovered in this semi-classical context. In
other words, how is it that quantizations of the scalar constraint that are ultra-local
in the sense above can impose conditions restricting unphysical degrees of freedom
in the interior of the sphere once boundary conditions defining the geometry of the

p The intuitive idea here is presented in terms of a given solution of the constraint based on a given
family of graphs: the dressed spin-networks. One should keep in mind that defining a quantum
operators representing a Dirac observable implies defining its action on the whole Hphys . A P simple
example of Dirac operator whose action is defined far all solutions is for instance OD = e je ,
i.e., the sum of the spins connecting dressed nodes66 .
q With the time-symmetric ansatz K
ab = 0 for the five degrees of freedom in the conformal metric
0 only one local physical degree of freedom remains after factoring out the action of the con-
qab
straints as generating functions of gauge transformations. In this context our truncation leads to
only one free data per space-point. Despite these restrictions there are many radiating spacetimes
and other interesting solutions in this sector.
44 Alejandro Perez

sphere are given? Because we still know very little about the semi-classical limit
these concerns should be taken as open issues that deserve a more precise analysis.
Motivated by these concerns different avenues of research have been explored
with the hope of finding alternatives and some guiding principles that would lead
to a clearer understanding of the physics behind the quantum scalar constraint. For
instance, the previous concerns have lead to the exploration of the formalism of
consistent discretizations presented in these lectures by Rodolfo Gambini70,71,72 .
The spin foam approach has been motivated to a large extent by the hope of solving
the issue of ambiguities and ultra-locality from a covariant perspective as well as by
the search of a systematic definition of the physical scalar product. An alternative
strategy to the quantization of the scalar constraint and the construction of the
physical inner product that in essence circumvents the anomaly freeness condition
(144) has been recently proposed by Thiemann73 .

4.4. An important application: computation of black hole entropy

In the actual talks we described the main ideas behind the computation of black hole
entropy in LQG. The philosophy was to quantize a sector of the theory containing
an isolated horizon74 and then to count the number of physical states N compatible
with a given macroscopic area a0 of the horizon. The entropy S of the black hole
is defined by S = ln(N ). The counting can be made exactly when a0 >> ℓ2p . The
result is
γ0 a0 a0
S= 2
+ O(ln( 2 )), (147)
γ 4ℓp ℓp
where the real number γ0 = 0.2375... follows from the counting 75,76 . The value of
the Immirzi parameter comes from the fact that γ appears as a pre-factor in the
spectrum of the area operator. It is important to emphasize that the computation
of S is independent of the details of the quantization of the scalar constraint. Semi-
classical considerations lead to S = a0 /(4ℓ2p ), the computation above can be used
to fix the value of the Immirzi parameter, namely
γ = γ0 . (148)
The above computation can be performed for any black hole of the Kerr-Newman
family and the result is consistent with the chosen value of γ.
The reader interested in the details of this calculation is encouraged to study
the original papers77,78,79,80,81 , the resent review by Ashtekar and Lewandowski,
or the book of Rovelli4. Attention is drawn to the resent results of Domagala and
Lewandowski75 and Meissner76 .

5. Spin Foams: the path integral representation of the dynamics in

loop quantum gravity
The spin foam approach was motivated by the need to shed new light on the issue of
the dynamics of loop quantum gravity by attempting the construction of the path
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 45

integral representation of the theory. In this section we will introduce the main ideas
behind the approach by considering simpler systems in some detail. For a broader
overview see the review articles82,83 and references therein.
The solutions of the scalar constraint can be characterized by the definition of
the generalized projection operator P from the kinematical Hilbert space Hkin into
the kernel of the scalar constraint Hphys . Formally one can write P as
 
Y Z Z
P =“ b
δ(S(x))” = D[N ] exp i N (x)S(x) [ . (149)
x⊂ Σ Σ

A formal argument84,85 shows that P can also be defined in a manifestly covariant

manner as a regularization of the formal path integral of general relativity (in 2d
gravity this is shown in86 , here we will show that this is the case in 3d gravity). In
first order variables it becomes
Z
P = D[e] D[A] µ[A, e] exp [iSGR (e, A)] (150)

where e is the tetrad field, A is the spacetime connection, and µ[A, e] denotes the
appropriate measure.
In both cases, P characterizes the space of solutions of quantum Einstein equa-
tions as for any arbitrary state |φ >∈ Hkin then P |φ > is a (formal) solution of
(63). Moreover, the matrix elements of P define the physical inner product (< , >p )
providing the vector space of solutions of (63) with the Hilbert space structure that
defines Hphys . Explicitly
< s, s′ >p :=< P s, s′ >,
for s, s′ ∈ Hkin .
When these matrix element are computed in the spin network basis (see Section
4.1.5), they can be expressed as a sum over amplitudes of ‘spin network histories’:
spin foams (Figure 11). The latter are naturally given by foam-like combinatorial
structures whose basic elements carry quantum numbers of geometry (see Section
4.2). A spin foam history87, from the state |s > to the state |s′ >, is denoted
by a pair (Fs→s′ , {j}) where Fs→s′ is the 2-complex with boundary given by the
graphs of the spin network states |s′ > and |s > respectively, and {j} is the set of
spin quantum numbers labelling its edges (denoted e ⊂ Fs→s′ ) and faces (denoted
f ⊂ Fs→s′ ). Vertices are denoted v ⊂ Fs→s′ . The physical inner product can be
expressed as a sum over spin foam amplitudes
< s′ , s >p =< P s′ , s >=
P P Q Q Q
N (Fs→s′ ) f ⊂Fs→s′ Af (jf ) e⊂Fs→s′ Ae (je ) v⊂Fs→s′ Av (jv ),
Fs→s′ {j}
(151)
where N (Fs→s′ ) is a (possible) normalization factor, and Af (jf ), Ae (je ), and Av (jv )
are the 2-cell or face amplitude, the edge or 1-cell amplitude, and the 0-cell of
46 Alejandro Perez

vertex amplitude respectively. These local amplitudes depend on the spin quantum
numbers labelling neighboring cells in Fs→s′ (e.g. the vertex amplitude of the vertex
magnified in Figure 11 is Av (j, k, l, m, n, s)).

l
j
p n s
o k m
q
n
s
m
l
j
p
k l
q o j
k

l
j

Fig. 11. A spin foam as the ‘colored’ 2-complex representing the transition between three different
spin network states. A transition vertex is magnified on the right.

The underlying discreteness discovered in LQG is crucial: in the spin foam rep-
resentation, the functional integral for gravity is replaced by a sum over amplitudes
of combinatorial objects given by foam-like configurations (spin foams) as in (151).
A spin foam represents a possible history of the gravitational field and can be inter-
preted as a set of transitions through different quantum states of space. Boundary
data in the path integral are given by the polymer-like excitations (spin network
states, Figure 5) representing 3-geometry states in LQG.

6. Spin foams in 3d quantum gravity

We introduce the concept of spin foams in a more explicit way in the context of
the quantization of three dimensional Riemannian gravity. In Section 6.2 and 6.3
we will present the definition of P from the canonical and covariant view point
formally stated in the introduction by Equations (149) and (150) respectivelyr . For
other approaches to 3d quantum gravity see the book of Carlip91 .

r It is well known that the physical inner product for 3d Riemannian gravity can be defined using

group averaging techniques88 , here we review this and use the approach to introduce the spin
foam representation89 ,90 .
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 47

6.1. The classical theory

Riemannian gravity in 3 dimensions is a theory with no local degrees of freedom,
i.e., a topological theory. Its action (in the first order formalism) is given by
Z
S(e, ω) = Tr(e ∧ F (ω)), (152)
M

where M = Σ × R (for Σ an arbitrary Riemann surface), ω is an SU (2)-connection

and the triad e is an su(2)-valued 1-form. The gauge symmetries of the action are
the local SU (2) gauge transformations
δe = [e, α] , δω = dω α, (153)
where α is a su(2)-valued 0-form, and the ‘topological’ gauge transformation
δe = dω η, δω = 0, (154)
where dω denotes the covariant exterior derivative and η is a su(2)-valued 0-form.
The first invariance is manifest from the form of the action, while the second is a
consequence of the Bianchi identity, dω F (ω) = 0. The gauge symmetries are so large
that all the solutions to the equations of motion are locally pure gauge. The theory
has only global or topological degrees of freedom.
Upon the standard 2+1 decomposition, the phase space in these variables is
parametrized by the pull back to Σ of ω and e. In local coordinates one can express
them in terms of the 2-dimensional connection Aia and the triad field Ejb = ǫbc ekc ηjk
where a = 1, 2 are space coordinate indices and i, j = 1, 2, 3 are su(2) indices. The
symplectic structure is defined by
{Aia (x), Ejb (y)} = δab δ ij δ (2) (x, y). (155)
Local symmetries of the theory are generated by the first class constraints
Db Ejb = 0, i
Fab (A) = 0, (156)
which are referred to as the Gauss law and the curvature constraint respectively—
the quantization of these is the analog of (63) in 4d. This simple theory has been
quantized in various ways in the literature91 , here we will use it to introduce the
spin foam quantization.

6.2. Spin foams from the Hamiltonian formulation

The physical Hilbert space, Hphys , is defined by those ‘states’ in Hkin that are an-
nihilated by the constraints. As we discussed in Section 4.1.5, spin network states
solve the Gauss constraint—D \ a
a Ei |s >= 0—as they are manifestly SU (2) gauge
invariant. To complete the quantization one needs to characterize the space of solu-
tions of the quantum curvature constraints (Fbab
i
), and to provide it with the physical
inner product. The existence of Hphys is granted by the following

Theorem 1. There exists a normalized positive linear form P over Cyl, i.e.
48 Alejandro Perez

P (ψ ⋆ ψ) ≥ 0 for ψ ∈ Cyl and P (1) = 1, yielding (through the GNS construction45 )

the physical Hilbert space Hphys and the physical representation πp of Cyl.

The state P contains a very large Gelfand ideal (set of zero norm states)
J := {α ∈ Cyl s.t. P (α⋆ α) = 0}. In fact the physical Hilbert space Hphys := Cyl/J
corresponds to the quantization of finitely many degrees of freedom. This is ex-
pected in 3d gravity as the theory does not have local excitations (no ‘gravitons’).
The representation πp of Cyl solves the curvature constraint in the sense that for
any functional fγ [A] ∈ Cyl defined on the sub-algebra of functionals defined on
contractible graphs γ ⊂ Σ, one has that
πp [fγ ]Ψ = fγ [0]Ψ. (157)
This equation expresses the fact that ‘Fb = 0’ in Hphys (for flat connections parallel
transport is trivial around a contractible region). For s, s′ ∈ Hkin , the physical inner
product is given by
< s, s′ >p := P (s⋆ s), (158)
where the ∗-operation and the product are defined in Cyl.

Wp

ε
Σ
Fig. 12. Cellular decomposition of the space manifold Σ (a square lattice in this example), and
the infinitesimal plaquette holonomy Wp [A].

The previous equation admits a ‘sum over histories’ representation89. We shall

introduce the concept of the spin foam representation as an explicit construction of
the positive linear form P which, as in (149), is formally given by
Z Z Y
P = D[N ] exp(i Tr[N Fb (A)]) = δ[F[(A)], (159)
Σ x⊂Σ

where N (x) ∈ su(2). One can make the previous formal expression a rigorous def-
inition if one introduces a regularization. Given a partition of Σ in terms of 2-
dimensional plaquettes of coordinate area ǫ2 one has that
Z X
Tr[N F (A)] = lim ǫ2 Tr[Npi Fpi ], (160)
ǫ→0
Σ pi

where Npi and Fpi are values of N i and ǫab Fabi

[A] at some interior point of the
i ab
plaquette p and ǫ is the Levi-Civita tensor. Similarly the holonomy Wpi [A] around
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 49

the boundary of the plaquette pi (see Figure 12) is given by

Wpi [A] = 1 + ǫ2 Fpi (A) + O(ǫ2 ), (161)
j
where Fpi = τj ǫab Fab (xpi )
(τj are the generators of su(2) in the fundamental repre-
sentation). The previous two equations lead to the following definition: given s ∈ Cyl
(think of spin network state based on a graph γ) the linear form P (s) is defined as
Y Z
P (s) := lim < Ω dNpi exp(iTr[Npi Wpi ]), s > . (162)
ǫ→0
pi

where < , > is the inner product in the AL-representation and |Ω > is the ‘vacuum’
(1 ∈ Cyl) in the AL-representation. The partition is chosen so that the links of the
underlying graph γ border the plaquettes. One can easily perform the integration
over the Npi using the identity (Peter-Weyl theorem)
Z X j
dN exp(iTr[N W ]) = (2j + 1) Tr[Π (W )]. (163)
j

Using the previous equation

YX j(pi )
P (s) := lim (2j(pi ) + 1) < Ω Tr[ Π (Wpi )]), s >, (164)
ǫ→0
pi j(pi )

where j(pi ) is the spin labelling elements of the sum (163) associated to the ith
j
plaquette. Since the Tr[Π (W )] commute the ordering of plaquette-operators in the
previous product does not matter. It can be shown that the limit ǫ → 0 exists and
one can give a closed expression of P (s).
Some remarks are in order:
Remark 1: The argument of the limit in (164) satisfies the following inequalities
 
X Y
(2j(pi ) + 1) µAL  χj(pi ) (Wpi [A]) s[A]s′ [A] ≤
j(pi ) pi
 
X Y
≤C (2j(pi ) + 1) µAL  χj(pi ) (Wpi [A]) =
jpi pi
X
=C (2j + 1)2−2g , (165)
j

where we have used (82), C is a real positive constant, and the last equation follows
immediately from the definition of the Ashtekar-Lewandowski measure µAL 86 . The
convergence of the sum for genus g ≥ 2 follows directly.
Remark 2: The case of the sphere g = 0 is easy to regularize. In this case (164)
diverges simply because of a redundancy in the product of delta distributions in the
notation of (20). This is a consequence of the discrete analog of the Bianchi identity.
50 Alejandro Perez

k k
k P P
Tr[Π (Wp )]  j = j = Nj,m,k j
m m

Fig. 13. Graphical notation representing the action of one plaquette holonomy on a spin network
state. On the right is the result written in terms of the spin network basis. The amplitude Nj,m,k
can be expressed in terms of Clebsch-Gordan coefficients.

It is easy to check that eliminating a single arbitrary plaquette holonomy from the
product in (164) makes P well defined and produces the correct (one dimensional)
Hphys .
The case of the torus g = 1 is more subtle; in fact our prescription must be
modified in that case91.
Remark 3: It is immediate to see that (164) satisfies hermitian condition

< P s, s′ >= < P s′ , s >. (166)

Remark 4: The positivity condition also follows from the definition < P s, s > ≥ 0.
j(pi )
Now in the AL-representation, each Tr[ Π (Wpi )] acts by creating a closed loop
in the jpi representation at the boundary of the corresponding plaquette (Figures 13
and 15). One can introduce a (non-physical) time parameter that works simply as a
coordinate providing the means of organizing the series of actions of plaquette loop-
operators in (164); i.e., one assumes that each of the loop actions occur at different
‘times’. We have introduced an auxiliary time slicing (arbitrary parametrization).
If one inserts the AL partition of unity
X X
1= |γ, {j} >< γ, {j}|, (167)
γ⊂Σ {j}γ

where the sum is over the complete basis of spin network states {|γ, {j} >}—
based on all graphs γ ⊂ Σ and with all possible spin labelling—between each time
slice, one arrives at a sum over spin-network histories representation of P (s). More
precisely, P (s) can be expressed as a sum over amplitudes corresponding to a series
of transitions that can be viewed as the ‘time evolution’ between the ‘initial’ spin
network s and the ‘final’ ‘vacuum state’ Ω. The physical inner product between spin
networks s, and s′ is defined as

< s, s′ >p := P (s⋆ s′ ),

and can be expressed as a sum over amplitudes corresponding to transitions inter-

polating between the ‘initial’ spin network s′ and the ‘final’ spin network s (e.g.
Figures 14 and 16).
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 51

j j j j
k k k

m m m m

j j j j
k k
k k k
m m m m

j j j

k k
k j

m m m m

Fig. 14. A set of discrete transitions in the loop-to-loop physical inner product obtained by
a series of transitions as in Figure 13. On the right, the continuous spin foam representation
in the limit ǫ → 0.

k j k j k
j n
n
P
  p o
n P 1 j k m
Tr[Π (Wp )]  = = ∆n ∆j ∆k ∆m
m m o,p n o p m

. .

Fig. 15. Graphical notation representing the action of one plaquette holonomy on a spin network
vertex. The object in brackets ({}) is a 6j-symbol and ∆j := 2j + 1.

Spin network nodes evolve into edges while spin network links evolve into 2-
dimensional faces. Edges inherit the intertwiners associated to the nodes and faces
inherit the spins associated to links. Therefore, the series of transitions can be
represented by a 2-complex whose 1-cells are labelled by intertwiners and whose
2-cells are labelled by spins. The places where the action of the plaquette loop
operators create new links (Figures 15 and 16) define 0-cells or vertices. These
foam-like structures are the so-called spin foams. The spin foam amplitudes are
purely combinatorial and can be explicitly computed from the simple action of the
loop operator in the AL-representation (Section 4.1.3). A particularly simple case
arises when the spin network states s and s′ have only 3-valent nodes. Explicitly
52 Alejandro Perez

j j
4 5

X Y νf Y j
< s, s′ >p := P (s⋆ s′ ) = (2jf + 1) 2
3
, (168)
j6
{j} f ⊂Fs→s′ v⊂Fs→s′
j j2
1

where the notation is that of (151), and νf = 0 if f ∩ s 6= 0 ∧ f ∩ s′ 6= 0, νf = 1 if

f ∩ s 6= 0 ∨ f ∩ s′ 6= 0, and νf = 2 if f ∩ s = 0 ∧ f ∩ s′ = 0. The tetrahedral diagram
denotes a 6j-symbol: the amplitude obtained by means of the natural contraction
of the four intertwiners corresponding to the 1-cells converging at a vertex. More
generally, for arbitrary spin networks, the vertex amplitude corresponds to 3nj-
symbols, and < s, s′ >p takes the general form (151).

j j j

m p p
m n m
n
o o

j
k k k p n
k
j j j m o
p p p
m m m
n n n
o
o o

k k k

j j j
p
p p
m m m
n n n

o o o
k k k

Fig. 16. A set of discrete transitions representing one of the contributing histories at a fixed
value of the regulator. On the right, the continuous spin foam representation when the regulator
is removed.

6.3. Spin foams from the covariant path integral

In this section we re-derive the SF-representation of the physical scalar product of
2+1 (Riemannian)s quantum gravity directly as a regularization of the covariant

sA generalization of the construction presented here for Lorentzian 2+1 gravity has been studied
by Freidel92 .
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 53

path integral93,94 . The formal path integral for 3d gravity can be written as
Z  Z 
P = D[e]D[A] exp i Tr[e ∧ F (A)] . (169)
M

Assume M = Σ × I, where I ⊂ R is a closed (time) interval (for simplicity we ignore

boundary terms).

Fig. 17. The cellular decomposition of M = Σ × I (Σ = T 2 in this example). The illustration

shows part of the induced graph on the boundary and the detail of a tetrahedron in ∆ and a face
f ⊂ ∆⋆ in the bulk.

In order to give a meaning to the formal expression above one replaces the 3-
dimensional manifold (with boundary) M with an arbitrary cellular decomposition
∆. One also needs the notion of the associated dual 2-complex of ∆ denoted by ∆⋆ .
The dual 2-complex ∆⋆ is a combinatorial object defined by a set of vertices v ⊂ ∆⋆
(dual to 3-cells in ∆) edges e ⊂ ∆⋆ (dual to 2-cells in ∆) and faces f ⊂ ∆⋆ (dual to
1-cells in ∆). The intersection of the dual 2-complex ∆⋆ with the boundaries defines
two graphs γ1 , γ2 ⊂ Σ (see Figure 17). For simplicity we ignore the boundaries until
the end of this section. The fields e and A are discretized as follows. The su(2)-
valued 1-form field e is represented by the assignment of ef ∈ su(2) to each 1-cell
in ∆. We use the fact that faces in ∆⋆ are in one-to-one correspondence with 1-
cells in ∆ and label ef with a face subindex (Figure 17). The connection field A is
54 Alejandro Perez

ge3
ge2
ge4
f
ge1
ge5

Fig. 18. A (2-cell) face f ⊂ ∆⋆ in a cellular decomposition of the spacetime manifold M . Also
the corresponding dual 1-cell. The connection field is discretized by the assignment of the parallel
transport group-elements gei ∈ SU (2) to edges e ⊂ ∆⋆ (i = 1, · · · 5 in the face shown here).

represented by the assignment of group elements ge ∈ SU (2) to each edge in e ⊂ ∆⋆

(see Figure 18).
With all this (169) becomes the regularized version P∆ defined as
Z Y Y
P∆ = def dge exp [iTr [ef Wf ]] , (170)
f ⊂∆⋆ e⊂∆⋆

where def is the regular Lebesgue measure on R3 , dge is the Haar measure on SU (2),
and Wf denotes the holonomy around (spacetime) faces, i.e., Wf = ge1 . . . geN for N
being the number of edges bounding the corresponding face (see Figure 18). The
discretization procedure is reminiscent of the one used in standard lattice gauge
theory. The previous definition can be motivated by an analysis equivalent to the
one presented in (160).
Integrating over ef , and using (163), one obtains
XZ Y Y  
jf
1 N
P∆ = dge (2jf + 1) Tr Π (ge . . . ge ) . (171)
{j} e⊂∆⋆ f ⊂∆⋆

Now it remains to integrate over the lattice connection {ge }. If an edge e ⊂ ∆⋆

jf
bounds n faces f ⊂ ∆⋆ there will be n traces of the form Tr[Π (· · · ge · · · )] in (171)
containing ge in the argument. In order to integrate over ge we can use the following
identity
Z X
j1 j2 jn
n
Iinv := dg Π (g)⊗ Π (g) ⊗ · · · ⊗ Π (g) = Cjι1 j2 ···jn Cj∗1ιj2 ···jn , (172)
ι

n
where Iinv is the projector from the tensor product of irreducible representations
Hj1 ···jn = j1 ⊗ j2 ⊗ · · · ⊗ jn onto the invariant component Hj01 ···jn = Inv[j1 ⊗ j2 ⊗
· · · ⊗ jn ]. On the r.h.s. we have chosen an orthonormal basis of invariant vectors
(intertwiners) in Hj1 ···jn to express the projector. Notice that the assignment of
intertwiners to edges is a consequence of the integration over the connection. Using
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 55

the (172) one can write P∆ in the general SF-representation form (151)
X Y Y
P∆ = (2jf + 1) Av (jv ), (173)
{f } f ⊂∆⋆ v⊂∆⋆

where Av (ιv , jv ) is given by the appropriate trace of the intertwiners corresponding

to the edges bounded by the vertex. As in the previous section this amplitude is given
in general by an SU (2) 3N j-symbol. When ∆ is a simplicial complex all the edges
in ∆⋆ are 3-valent and vertices are 4-valent. Consequently, the vertex amplitude
is given by the contraction of the corresponding four 3-valent intertwiners, i.e., a
6j-symbol. In that case the path integral takes the (Ponzano-Regge95) form

j j
4 5

X Y Y j
3
P∆ = (2jf + 1) . (174)
j6
{j} f ⊂∆⋆ v⊂∆⋆
j j2
1

The labelling of faces that intersect the boundary naturally induces a labelling of
the edges of the graphs γ1 and γ2 induced by the discretization. Thus, the boundary
states are given by spin network states on γ1 and γ2 respectively. A careful analysis
of the boundary contribution shows that only the face amplitude is modified to
(2jf + 1)νf /2 , and that the spin-foam amplitudes are as in Equation (168).
A crucial property of the path integral in 3d gravity (and of the transition
amplitudes in general) is that it does not depend on the discretization ∆—this is
due to the absence of local degrees of freedom in 3d gravity and not expected to
hold in 4d. Given two different cellular decompositions ∆ and ∆′ one has
′
τ −n0 P∆ = τ −n0 P∆′ , (175)
P 2
where n0 is the number of 0-simplexes in ∆, and τ = j (2j + 1) . This trivial
scaling property of transition amplitudes allows for a simple definition of transition
amplitudes that are independent of the discretization96. However, notice that since
τ is given by a divergent sum the discretization independence statement is formal.
Moreover, the sum over spins in (174) is typically divergent. Divergences occur due
to infinite gauge-volume factors in the path integral corresponding to the topological
gauge freedom (154). Freidel and Louapre97 have shown how these divergences can
be avoided by gauge-fixing un-physical degrees of freedom in (170). In the case
of 3d gravity with positive cosmological constant the state-sum generalizes to the
Turaev-Viro invariant98,99,100,101 defined in terms of the quantum group SUq (2)
with q n = 1 where the representations are finitely many and thus τ < ∞. Equation
(175) is a rigorous statement in that case. No such infrared divergences appear in
the canonical treatment of the previous section.
 56 Alejandro Perez

7. Spin foams in four dimensions

7.1. SF from the canonical formulation
There is no rigorous construction of the physical inner product of LQG in four
dimensions. The spin foam representation as a device for its definition has been in-
troduced formally by Rovelli102 and Rovelli and Reisenberger103. In 4-dimensional
LQG difficulties in understanding dynamics are centered around the quantum
√ −1\ ij
scalar constraint Sb = detE Eia Ejb Fab (A) + · · · (see (63))—the vector constraint
Vba (A, E) is solved in a simple manner. The physical inner product formally becomes
 
Y Z Z
hP s, s′ idif f = b
δ[S(x)] = D[N ] < exp i N (x)S(x) b  s, s′ >dif f
x
Σ
 n
Z ∞ n
X Z
i b 
= D[N ] < N (x)S(x) s, s′ >dif f , (176)
n=0
n!
Σ
where < , >dif f denotes the inner product in the Hilbert space of solutions of the
vector constraint, and the exponential has been expanded in powers in the second
line.
From early on, it was realized that smooth loop states are naturally annihilated
b
by S(independently of any quantization ambiguity104,105 ). Consequently, Sb acts
only on spin network nodes. Generically, it does so by creating new links and nodes
modifying the underlying graph of the spin network states (Figure 19).

n
j k j k
n o
p
p o
R P
b
N (x)S(x)  = N (xn )Snop
Σ nop
m m
k
j
. .
m

Fig. 19. The action of the scalar constraint and its spin foam representation. N (xn ) is the value
of N at the node and Snop are the matrix elements of S. b

Therefore, each term in the sum (176) represents a series of transitions—given

by the local action of Sb at spin network nodes—through different spin network
states interpolating the boundary states s and s′ respectively. The action of Sb
can be visualized as an ‘interaction vertex’ in the ‘time’ evolution of the node
(Figure 19). As in the explicit 3d case, equation (176) can be expressed as sum over
‘histories’ of spin networks pictured as a system of branching surfaces described by
a 2-complex whose elements inherit the representation labels on the intermediate
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 57

states. The value of the ‘transition’ amplitudes is controlled by the matrix elements
b Therefore, although the qualitative picture is independent of quantization
of S.
ambiguities transition amplitudes are sensitive to them.
Before even considering the issue of convergence of (176), the problem with this
definition is evident: every single term in the sum is a divergent integral! Therefore,
this way of presenting spin foams has to be considered as formal until a well defined
regularization of (149) is provided. That is the goal of the spin foam approach.
Instead of dealing with an infinite number of constraints Thiemann recently
proposed to impose one single master constraint defined as
Z
S 2 (x) − q ab Va (x)Vb (x)
M = dx3 p . (177)
det q(x)
Σ

Using techniques developed by Thiemann this constraint can indeed be promoted

to a quantum operator acting on Hkin . The physical inner product is given by
ZT
′ c
< s, s >p := lim < s, dt eitM s′ > . (178)
T →∞
−T

A SF-representation of the previous expression could now be achieved by the stan-

dard skeletonization that leads to the path integral representation in quantum me-
chanics. In this context one splits the t-parameter in discrete steps and writes
c c c/N ]N .
eitM = lim [eitM /N ]N = lim [1 + itM (179)
N →∞ N →∞

The SF-representation follows from the fact that the action of the basic operator
1 + itMc/N on a spin network can be written as a linear combination of new spin
networks whose graphs and labels have been modified by the creation of new nodes
(in a way qualitatively analogous to the local action shown in Figure 19). An explicit
derivation of the physical inner product of 4d LQG along these lines is under current
investigation.

7.2. Spin foams from the covariant formulation

In four dimensions the spin foam representation of the dynamics of LQG has been
motivated by lattice discretizations of the path integral of gravity in the covari-
ant formulation42,106,107,108 . This has lead to a series of constructions which are
refereed to as spin foam models82 . These treatments are closer related to the con-
struction of Section 6.3. Here we illustrate the formulation which has captured much
interest in the literature: the Barrett-Crane model (BC model)109,110.

7.2.1. Spin foam models for gravity as constrained quantum BF theory

The BC model is one of the most extensively studied spin foam models for quantum
gravity. To introduce the main ideas involved we concentrate on the definition of
58 Alejandro Perez

the model in the Riemannian sector. The BC model can be formally viewed as
a spin foam quantization of SO(4) Plebanski’s formulation of GR111 . Plebanski’s
Riemannian action depends on an so(4) connection A, a Lie-algebra-valued 2-form
B and Lagrange multiplier fields λ and µ. Writing explicitly the Lie-algebra indices,
the action is given by
Z
 IJ 
I[B, A, λ, µ] = B ∧ FIJ (A) + λIJKL B IJ ∧ B KL + µǫIJKL λIJKL , (180)
where µ is a 4-form and λIJKL = −λJIKL = −λIJLK = λKLIJ is a tensor in the
internal space. Variation with respect to µ imposes the constraint ǫIJKL λIJKL = 0
on λIJKL . The Lagrange multiplier tensor λIJKL has then 20 independent compo-
nents. Variation with respect to λ imposes 20 algebraic equations on the 36 com-
ponents of B. The (non-degenerate) solutions to the equations obtained by varying
the multipliers λ and µ are

B IJ = ±ǫIJKL eK ∧ eL , and B IJ = ±eI ∧ eJ , (181)

in terms of the 16 remaining degrees of freedom of the tetrad field eIa .If one sub-
stitutes the first solution into the original action one obtains Palatini’s formulation
of general relativity; therefore on shell (and on the right sector) the action is that
of classical gravity.
The key idea in the definition of the model is that the path integral for the
theory corresponding to the action I[B, A, 0, 0], namely
Z  Z 
 IJ 
Ptopo = D[B]D[A] exp i B ∧ FIJ (A) (182)

can be given a meaning as a spin foam sum, (151), in terms of a simple generalization
of the construction of Section 6. In fact I[B, A, 0, 0] corresponds to a simple theory
known as BF theory that is formally very similar to 3d gravity112. The result is
independent of the chosen discretization because BF theory does not have local
degrees of freedom (just as 3d gravity).
The BC model aims at providing a definition of the path integral of gravity
pursuing a well-posed definition of the formal expression
Z  Z 
 IJKL
  IJ 
PGR = D[B]D[A] δ B → ǫ eK ∧ eL exp i B ∧ FIJ (A) , (183)

where D[B]D[A]δ(B → ǫIJKL eK ∧ eL ) means that one must restrict the sum in
(182) to those configurations of the topological theory satisfying the constraints
B = ∗(e ∧ e) for some tetrad e. The remarkable fact is that this restriction can
be implemented in a systematic way directly on the spin foam configurations that
define Ptopo 113,114 .
In Ptopo spin foams are labelled with spins corresponding to the unitary irre-
ducible representations of SO(4) (given by two spin quantum numbers (jR , jL )).
Essentially, the factor ‘δ(B → ǫIJKL eK ∧ eL )’ restricts the set of spin foam quan-
tum numbers to the so-called simple representations (for which jR = jL = j 87,115 ).
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 59

This is the ‘quantum’ version of the solution to the constraints (181). There are vari-
ous versions of this model, some versions satisfy intriguing finiteness116,117,118,119
propertiest . The simplest definition of the transition amplitudes in the BC -model
is given by
ι2 ι *2
j12 j 23 j* j *23
12

ι1 j 13 j *13
ι 3 ι 1* ι *3
X Y Y X j25 j 24 j* j 24
*
⋆ νf
P (s s) = (2jf +1) 25
,
{j} f ⊂Fs→s′ v⊂Fs→s′ ι1 ···ι5 j15 j 14 j35 j 34 j 15
*
j *14 j35
* j *34

ι5 j 45 ι4 ι*5 j*45 ι *4

(184)
where we use the notation of (168), the graphs denote 15j symbols, and ιi are half in-
tegers labelling SU (2) normalized 4-intertwinersu. No rigorous connection with the
Hilbert space picture of LQG has yet been established. The self-dual version of Ple-
banski’s action leads, through a similar construction, to Reisenberger’s model107 . A
general prescription for the definition of constrained BF theoriesv on the lattice has
been studied by Freidel and Krasnov123. Lorentzian generalizations of the Barrett-
Crane model have been defined124,125. A generalization using quantum groups was
studied by Noui and Roche126 .
2

1
111111111
000000000 3
000000000
111111111
0 111111111
000000000
000000000
111111111
000000000
111111111
000000000
111111111
000000000
111111111
000000000
111111111
000000000
111111111
5
000000000
111111111
4

Fig. 20. The dual of a 4-simplex.

The simplest amplitude in the BC model corresponds to a single 4-simplex. A

4-simplex can be viewed as the simplest triangulation of the 4-dimensional space
time given by the interior of a 3-sphere (the corresponding 2-complex is shown
in Figure 20). States of the 4-simplex are labelled by 10 spins j (labelling the 10
edges of the boundary spin network, see Figure 20) which can be shown to be
related to the area in Planck units of the 10 triangular faces that form the 4-
simplex. A first indication of the connection of the model with gravity was that

t Fora discussion of the freedom involved see120 .

u Reisenberger121 proved that the 4-simplex BC amplitude is unique up to normalization.
v Gambini and Pullin studied an alternative modification of BF theory leading to a simple model

with intriguing properties122 .

60 Alejandro Perez

the large j-asymptotics appeared to be dominated by the exponential of the Regge

action127,128 (the action derived by Regge as a discretization of GR). This estimate
was done using the stationary phase approximation to the integral129 that gives
the amplitude of a 4-simplex in the BC model. However, more detailed calculations
showed that the amplitude is dominated by configurations corresponding to degen-
erate 4-simplexes130,97,131. This seem to invalidate a simple connection to GR and
is one of the main puzzles in the model.

7.3. Spin foams as Feynman diagrams

The main problem with the models of the previous section is that they are defined
on a discretization ∆ of M and that—contrary to what happens with a topological
theory, e.g. 3d gravity (Equation 175)—the amplitudes depend on the discretiza-
tion ∆. Various possibilities to eliminate this regulator have been discussed in the
literature but no explicit results are yet known in 4d. An interesting proposal is a
discretization-independent definition of spin foam models achieved by the introduc-
tion of an auxiliary field theory living on an abstract group manifold—Spin(4)4 and
SL(2, C)4 for Riemannian and Lorentzian gravity respectively132,133,134,135,136 .
The action of the auxiliary group field theory (GFT) takes the form
Z Z
2 λ
I[φ] = φ + M (5) [φ], (185)
G4 5! G10
where M (5) [φ] is a fifth order monomial, and G is the corresponding group. In the
simplest model M (5) [φ] =
φ(g1 , g2 , g3 , g4 )φ(g4 , g5 , g6 , g7 )φ(g7 , g3 , g8 , g9 )φ(g9 , g6 , g2 , g10 )φ(g10 , g8 , g5 , g1 ).
The field φ is required to be invariant under the (simultaneous) right action of
the group on its four arguments in addition to other symmetries (not described
here for simplicity). The perturbative expansion in λ of the GFT Euclidean path
integral is given by
Z X λN
P = D[φ]e−I[φ] = A[FN ], (186)
sym[FN ]
FN

where A[FN ] corresponds to a sum of Feynman-diagram amplitudes for diagrams

with N interaction vertices, and sym[FN ] denotes the standard symmetry factor. A
remarkable property of this expansion is that A[FN ] can be expressed as a sum over
spin foam amplitudes, i.e., 2-complexes labelled by unitary irreducible representa-
tions of G. Moreover, for very simple interaction M (5) [φ], the spin foam amplitudes
are in one-to-one correspondence to those found in the models of the previous section
(e.g. the BC model). This duality is regarded as a way of providing a fully com-
binatorial definition of quantum gravity where no reference to any discretization
or even a manifold-structure is made. Transition amplitudes between spin network
states correspond to n-point functions of the field theory137 . These models have
been inspired by generalizations of matrix models applied to BF theory138,139.
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 61

Divergent transition amplitudes can arise by the contribution of ‘loop’ diagrams

as in standard QFT. In spin foams, diagrams corresponding to 2-dimensional bub-
bles are potentially divergent because spin labels can be arbitrarily high leading to
unbounded sums in (151). Such divergences do not occur in certain field theories
dual (in the sense above) to the Barrett-Crane (BC) model. However, little is known
about the convergence of the series in λ and the physical meaning of this constant.
Nevertheless, Freidel and Louapre140 have shown that the series can be re-summed
in certain models dual to lower dimensional theories. Techniques for studying the
continuum limit of these kind of theories have been proposed141,142,143,144. There
are models defined in this context admitting matter degree of freedom145,146 .

7.4. Causal spin foams

Let us finish by presenting a fundamentally different construction leading to spin
foams. Using the kinematical setting of LQG with the assumption of the existence
of a micro-local (in the sense of Planck scale) causal structure Markopoulou and
Smolin147,148,149 define a general class of (causal) spin foam models for gravity.
The elementary transition amplitude AsI →sI+1 from an initial spin network sI to
another spin network sI+1 is defined by a set of simple combinatorial rules based
on a definition of causal propagation of the information at nodes. The rules and
amplitudes have to satisfy certain causal restrictions (motivated by the standard
concepts in classical Lorentzian physics). These rules generate surface-like excita-
tions of the same kind one encounters in the previous formulations. Spin foams
FsNi →sf are labelled by the number of times, N , these elementary transitions take
place. Transition amplitudes are defined as
X
hsi , sf i = A(FsNi →sf ) (187)
N

which is of the generic form (151). The models are not related to any continuum
action. The only guiding principles in the construction are the restrictions imposed
by causality, and the requirement of the existence of a non-trivial critical behavior
that reproduces general relativity at large scales. Some indirect evidence of a possible
non trivial continuum limit has been obtained in certain versions of these models
in 1 + 1 dimensions.

8. Some final bibliographic remarks

We did not have time to discuss the applications of loop quantum gravity to cos-
mology. The interested reader is referred to review article150 and the references
therein.
We did not have the chance to mention in this lectures the important area of
research in LQG devoted to the study of the low energy limit of the theory. We
refer the reader to the general sources19,4,20 for bibliography and an overview of
results and outlook.
62 Alejandro Perez

In Section 4 we introduced a representation of the basic kinematical observables

in the kinematical Hilbert space Hkin . That was the starting point for the definition
of the theory. The reader might wonder why one emphasizes so much the Hilbert
space as a fundamental object when in the context of standard quantum field theory
it is rather the algebra of observables what plays the fundamental role. Hilbert
spaces correspond to representations of the algebra of observables which are chosen
according to the physical situation at hand. However, when extra symmetry is
present it can happen that there is no freedom and that a single representation
is selected by the additional symmetry. This is in fact the case in LQG if one
imposes the condition of diffeomorphism invariance on the state that defines the
representation of the holonomy and the flux operators 151,152,153,154,155,156,157 .

Acknowledgments
I would like to thank the organizer of the Second International Conference on Fun-
damental Interactions for their support and for a wonderful conference. Special
thanks to them also for their great hospitality and the wonderful time we had in
Pedra Azul. I thank Abhay Ashtekar, Bernd Bruegmann, Rodolfo Gambini, Jurek
Lewandowski, Marcelo Maneschy, Karim Noui and Carlo Rovelli for discussions and
to Mikhail Kagan and Kevin Vandersloot for the careful reading of the manuscript. I
thank Hoi Lai Yu for listening to an early version to these lectures and for insightful
questions. Many thanks to Olivier Piguet and Clisthenis Constantinidis. This work
has been supported by NSF grants PHY-0354932 and INT-0307569 and the Eberly
Research Funds of Penn State University.

References
1. Rogerio Rosenfeld. The standard model. Proceedings of the II International confer-
ence on fundamental interactions, Pedra Azul, Brazil, June 2004.
2. R. M. Wald. Quantum field theory in curved space-time and black hole thermody-
namics. Chicago, USA: Univ. Pr. (1994) 205 p.
3. R. M. Wald. General relativity. Chicago, Usa: Univ. Pr. ( 1984) 491p.
4. C. Rovelli. Quantum Gravity. Cambridge, UK: Univ. Pr. (2004) 480 p.
5. Carlo Rovelli. Relational quantum mechanics. quant-ph/9609002, 1995.
6. Norbert Grot, Carlo Rovelli, and Ranjeet S. Tate. Time-of-arrival in quantum me-
chanics. Phys. Rev., A54:4679, 1996.
7. Rodolfo Gambini and Rafael A. Porto. Relational time in generally covariant quan-
tum systems: Four models. Phys. Rev., D63:105014, 2001.
8. Rodolfo Gambini and Rafael A. Porto. Relational reality in relativistic quantum
mechanics. Phys. Lett., A294:129–133, 2002.
9. Rodolfo Gambini and Rafael A. Porto. Multi-local relational description of the mea-
surement process in quantum field theory. New J. Phys., 4:58, 2002.
10. Rodolfo Gambini and Rafael A. Porto. A physical distinction between a covariant
and non covariant description of the measurement process in relativistic quantum
theories. New. J. Phys., 5:105, 2003.
11. R. Penrose. Gravity and quantum mechanics. Prepared for 13th Conference on Gen-
eral Relativity and Gravitation (GR-13), Cordoba, Argentina, 29 Jun - 4 Jul 1992.
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 63

12. R. Penrose. Twistor theory, the einstein equations, and quantum mechanics. Pre-
pared for International School of Cosmology and Gravitation: 14th Course: Quantum
Gravity, Erice, Italy, 11-19 May 1995.
13. R. Penrose. Gravitational collapse of the wavefunction: An experimentally testable
proposal. Prepared for 9th Marcel Grossmann Meeting on Recent Developments in
Theoretical and Experimental General Relativity, Gravitation and Relativistic Field
Theories (MG 9), Rome, Italy, 2-9 Jul 2000.
14. R. Penrose. The role of gravity in quantum state reduction. Prepared for International
Conference on Non-Accelerator Particle Physics - ICNAPP, Bangalore, India, 2-9 Jan
1994.
15. S. Weinberg. The Quantum Theory of Fields. Cambridge University Press, 1995.
16. Robert C. Myers and Maxim Pospelov. Experimental challenges for quantum gravity.
Phys. Rev. Lett., 90:211601, 2003.
17. Alejandro Perez and Daniel Sudarsky. Comments on challenges for quantum gravity.
Phys. Rev. Lett., 91:179101, 2003.
18. John Collins, Alejandro Perez, Daniel Sudarsky, Luis Urrutia, and Hector Vucetich.
Lorentz invariance: An additional fine-tuning problem. Phys. Rev. Lett. (in press,
preprint gr-qc/0403053), 2004.
19. Abhay Ashtekar and Jerzy Lewandowski. Background independent quantum gravity:
A status report. Class. Quant. Grav., 21:R53, 2004.
20. Thomas Thiemann. Introduction to modern canonical quantum general relativity.
gr-qc/0110034, 2001.
21. L. Smolin. Recent Developments in non Perturbative Quantum Gravity; in Quantum
Gravity and Cosmology, Proceedings of the 1991 GIFT International Seminar on
Theoretical Physics. Available at hep-th/9202022. World Scientific, Singapore, 1992.
22. C. Rovelli and L. Smolin. Discretneess of the area and volume in quantum gravity.
Nucl Phys B, 442 (1995), Erratum: 456:593,734, 1995.
23. Ted Jacobson. New variables for canonical supergravity. Class. Quant. Grav., 5:923,
1988.
24. H. J. Matschull and H. Nicolai. Canonical quantum supergravity in three-dimensions.
Nucl. Phys., B411:609–646, 1994.
25. Yi Ling. Introduction to supersymmetric spin networks. J. Math. Phys., 43:154–169,
2002.
26. Yi Ling. Extending loop quantum gravity to supergravity. UMI-30-51694.
27. Yi Ling and Lee Smolin. Holographic formulation of quantum supergravity. Phys.
Rev., D63:064010, 2001.
28. Etera R. Livine and Robert Oeckl. Three-dimensional quantum supergravity and
supersymmetric spin foam models. Adv. Theor. Math. Phys., 7:951–1001, 2004.
29. M. Henneaux and C. Teitelboim. Quantization of gauge systems. Princeton, USA:
Univ. Pr. (1992) 520 p.
30. A. Ashtekar. Lectures on non perturbative canonical gravity. Word Scientific, 1991.
31. J. Fernando Barbero. From euclidean to lorentzian general relativity: The real way.
Phys. Rev., D54:1492–1499, 1996.
32. J. Fernando Barbero. Real ashtekar variables for lorentzian signature space times.
Phys. Rev., D51:5507–5510, 1995.
33. J. Fernando Barbero. A real polynomial formulation of general relativity in terms of
connections. Phys. Rev., D49:6935–6938, 1994.
34. G. Immirzi. Real and complex connections for canonical gravity. Class. Quant. Grav.,
14:L177–L181, 1997.
35. H. A. Morales-Tecotl and C. Rovelli. Loop space representation of quantum fermions
64 Alejandro Perez

and gravity. Nucl. Phys. B, 451:325–361, 1995.

36. Hugo A. Morales-Tecotl and Carlo Rovelli. Fermions in quantum gravity. Phys. Rev.
Lett., 72:3642–3645, 1994.
37. John C. Baez and Kirill V. Krasnov. Quantization of diffeomorphism-invariant the-
ories with fermions. J. Math. Phys., 39:1251–1271, 1998.
38. P. A. M. Dirac. Lectures on Quantum Mechanics. New York : Belfer Graduate School
of Science, Yeshiva University, 1964.
39. Joseph Samuel. Is barbero’s hamiltonian formulation a gauge theory of lorentzian
gravity? Class. Quant. Grav., 17:L141–L148, 2000.
40. A. Ashtekar and J. Lewandowski. Projective techniques and functional integration.
J. Math. Phys., 36:2170, 1995.
41. R. Penrose. Quantum Theory and beyond, ed T. Bastin. Cambridge University Press.
42. Michael P. Reisenberger. World sheet formulations of gauge theories and gravity.
gr-qc/941235, 1994.
43. C. Rovelli and L. Smolin. Spin networks and quantum gravity. Phys. Rev. D, 53:5743,
1995.
44. J. Baez. Spin network states in gauge theory. Adv.Math., 117:253–272, 1996.
45. R. Haag. Local quantum physics: Fields, particles, algebras. Berlin, Germany:
Springer (1992) 356 p. (Texts and monographs in physics).
46. J. Fuchs and C. Schweigert. Symmetries, lie algebras and representations: A graduate
course for physicists. Cambridge, UK: Univ. Pr. (1997) 438 p.
47. L. Smolin. The Future of Spin Networks. gr-qc/9702030.
48. A. Ashtekar and J. Lewandowski. Quantum theory of gravity i: Area operators. Class.
Quant. Grav., 14:A55–A81, 1997.
49. R. Loll. Simplifying the spectral analysis of the volume operator. Nucl.Phys. B,
500:405–420, 1997.
50. A. Ashtekar and J. Lewandowski. Quantum theory of gravity ii: Volume operators.
gr-qc/9711031.
51. T. Thieman. Anomaly-free formulation of non-perturbative, four-dimensional
lorentzian quantum gravity. Phys. Lett. B, 380:257, 1996.
52. R. Loll. Spectrum of the volume operator in quantum gravity. Nucl. Phys., B460:143–
154, 1996.
53. R. Loll. The volume operator in discretized quantum gravity. Phys. Rev. Lett.,
75:3048–3051, 1995.
54. T. Thiemann. Closed formula for the matrix elements of the volume operator in
canonical quantum gravity. J. Math. Phys., 39:3347–3371, 1998.
55. Johannes Brunnemann and Thomas Thiemann. Simplification of the spectral analysis
of the volume operator in loop quantum gravity. gr-qc/0405060, 2004.
56. T. Thiemann. Gauge field theory coherent states (gcs) : I. general properties.
Class.Quant.Grav., 18:2025, 2001.
57. T. Thiemann. Quantum spin dynamics (qsd). Class. Quant. Grav., 15:839–873, 1998.
58. T. Thiemann. Qsd iii: Quantum constraint algebra and physical scalar product in
quantum general relativity. Class. Quant. Grav., 15:1207–1247, 1998.
59. M. Gaul and C. Rovelli. A generalized hamiltonian constraint operator in loop quan-
tum gravity and its simplest euclidean matrix elements. Class.Quant.Grav., 18:1593–
1624, 2001.
60. C. Di Bartolo, R. Gambini, J. Griego, and J. Pullin. Canonical quantum gravity in the
vassiliev invariants arena: Ii. constraints, habitats and consistency of the constraint
algebra. Class.Quant.Grav., 17:3239–3264, 2000.
61. C. Di Bartolo, R. Gambini, J. Griego, and J. Pullin. Consistent canonical quanti-
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 65

zation of general relativity in the space of vassiliev knot invariants. Phys.Rev.Lett.,

84:2314–2317, 2000.
62. T. Thieman. Quantum gravity as the natural regulator of matter quantum field
theories. Class.Quant.Grav., 15:1281–1314, 1998.
63. Martin Bojowald and Kevin Vandersloot. Loop quantum cosmology, boundary pro-
posals, and inflation. Phys. Rev., D67:124023, 2003.
64. Lee Smolin. The classical limit and the form of the hamiltonian constraint in non-
perturbative quantum general relativity. gr-qc/9609034, 1996.
65. T. Thiemann. Quantum spin dynamics (qsd) ii. Class. Quant. Grav., 15:875–905,
1998.
66. Private disscussion with J. Lewandowski.
67. Gregory B. Cook. Initial data for numerical relativity. Living Rev. Rel., 3:5, 2000.
68. Abhay Ashtekar, Carlo Rovelli, and Lee Smolin. Weaving a classical geometry with
quantum threads. Phys. Rev. Lett., 69:237–240, 1992.
69. Thomas Thiemann. Gauge field theory coherent states (gcs). i: General properties.
Class. Quant. Grav., 18:2025–2064, 2001.
70. Rodolfo Gambini. Consistent discretizations and quantum gravity. Proceedings of the
II International conference on fundamental interactions, Pedra Azul, Brazil, June
2004.
71. J. Pullin R. Gambini. Canonical quantization of general relativity in discrete space-
times. gr-qc/0206055.
72. J. Pullin C. Di Bartolo, R. Gambini. Canonical quantization of constrained theories
on discrete space-time lattices. gr-qc/0206055.
73. Thomas Thiemann. The phoenix project: Master constraint programme for loop
quantum gravity. gr-qc/0305080, 2003.
74. Abhay Ashtekar and Badri Krishnan. Isolated and dynamical horizons and their
applications. gr-qc/0407042, 2004.
75. Marcin Domagala and Jerzy Lewandowski. Black hole entropy from quantum geom-
etry. gr-qc/0407051, 2004.
76. Krzysztof A. Meissner. Black hole entropy in loop quantum gravity. gr-qc/0407052,
2004.
77. C. Rovelli. Black hole entropy from loop quantum gravity. Phys Rev Lett, 14:3288,
1996.
78. Kirill V. Krasnov. On statistical mechanics of gravitational systems. Gen. Rel. Grav.,
30:53–68, 1998.
79. A. Ashtekar, J. Baez, A. Corichi, and K. Krasnov. Quantum geometry and black hole
entropy. Phys. Rev. Lett., 80:904–907, 1998.
80. A. Ashtekar, John C. Baez, and Kiriil Krasnov. Quantum geometry of isolated hori-
zons and black hole entropy. Adv. Theor. Math. Phys., 4:1–94, 2000.
81. Abhay Ashtekar, Alejandro Corichi, and Kirill Krasnov. Isolated horizons: The clas-
sical phase space. Adv. Theor. Math. Phys., 3:419–478, 2000.
82. Alejandro Perez. Spin foam models for quantum gravity. Class. Quant. Grav., 20:R43,
2003.
83. D. Oriti. Spacetime geometry from algebra: Spin foam models for non- perturbative
quantum gravity. Rept. Prog. Phys., 64:1489–1544, 2001.
84. J.B. Hartle J. Halliwell. Wave functions constructed from an invariant sum over
histories satisfy constraints. Phys. Rev., D43:1170–1194, 1991.
85. C. Rovelli M. Reisenberger. Spacetime states and covariant quantum theory.
Phys.Rev. D, 65:125016, 2002.
86. E. R. Livine, A. Perez, and C. Rovelli. 2d manifold-independent spinfoam theory.
66 Alejandro Perez

Class. Quant. Grav., 20:4425–4445, 2003.

87. J. C. Baez. Spin foam models. Class.Quant.Grav., 15:1827–1858, 1998.
88. Donald Marolf, Jose Mourao, and Thomas Thiemann. The status of diffeomorphism
superselection in euclidean 2+1 gravity. J. Math. Phys., 38:4730–4740, 1997.
89. Karim Noui and Alejandro Perez. Three dimensional loop quantum gravity: Physical
scalar product and spin foam models. gr-qc/0402110, 2004.
90. Karim Noui and Alejandro Perez. Three dimensional loop quantum gravity: Coupling
to point particles. gr-qc/0402111, 2004.
91. S. Carlip. Quantum gravity in 2+1 dimensions. Cambridge, UK: Univ. Pr. (1998)
276 p.
92. L. Freidel. A ponzano-regge model of lorentzian 3-dimensional gravity.
Nucl.Phys.Proc.Suppl., 88:237–240, 2000.
93. J. Iwasaki. A definition of the ponzano-regge quantum gravity model in terms of
surfaces. J. Math. Phys., 36:6288–6298, 1995.
94. Junichi Iwasaki. A reformulation of the ponzano-regge quantum gravity model in
terms of surfaces. gr-qc/9410010.
95. T Regge G Ponzano. Semiclassical limit of Racah Coeficients. Spectroscopy and
Group Theoretical Methods in Physics, F. Block et al (Eds), North-Holland, Ams-
terdam, 1968.
96. Jose A. Zapata. Continuum spin foam model for 3d gravity. J. Math. Phys., 43:5612–
5623, 2002.
97. Laurent Freidel and David Louapre. Diffeomorphisms and spin foam models. Nucl.
Phys., B662:279–298, 2003.
98. O. Y. Viro V. G. Turaev. Statesum invariants of 3-manifolds and quantum 6j-
symbols. Topology, 31:865–902, 1992.
99. V.G. Turaev. Quantum invariants of knots and 3-manifolds. W. de Gruyter, Berlin
; New York, 1994.
100. S.L. Lins L.H. Kauffman. Temperley-Lieb recoupling theory and invariants of 3-
manifolds. Princeton University Press, 1994, Princeton, N.J., 1994.
101. F. Girelli, R. Oeckl, and A. Perez. Spin foam diagrammatics and topological invari-
ance. Class. Quant. Grav., 19:1093–1108, 2002.
102. C. Rovelli. The projector on physical states in loop quantum gravity. Phys.Rev. D,
59:104015, 1999.
103. M. P. Reisenberger and C. Rovelli. “sum over surfaces” form of loop quantum gravity.
Phys.Rev. D, 56:3490–3508, 1997.
104. L. Smolin T. Jacobson. Nonperturbative quantum geometries. Nucl. Phys., B299:295,
1988.
105. C. Rovelli L. Smolin. Loop space representation of quantum general relativity. Nucl.
Phys. B, 331:80, 1990.
106. M. P. Reisenberger. A left-handed simplicial action for euclidean general relativity.
Class.Quant.Grav, 14:1753–1770, 1997.
107. M. P. Reisenberger. A lattice worldsheet sum for 4-d euclidean general relativity.
gr-qc/9711052.
108. J. Iwasaki. A lattice quantum gravity model with surface-like excitations in 4-
dimensional spacetime. gr-qc/0006088.
109. J. W. Barrett and L. Crane. Relativistic spin networks and quantum gravity.
J.Math.Phys., 39:3296–3302, 1998.
110. J. W. Barrett and L. Crane. A lorentzian signature model for quantum general rel-
ativity. Class.Quant.Grav., 17:3101–3118, 2000.
111. J.F. Plebanski. On the separation of einsteinian substructures. J. Math. Phys.,
 INTRODUCTION TO LOOP QUANTUM GRAVITY AND SPIN FOAMS 67

18:2511, 1977.
112. J. C. Baez. An introduction to spin foam models of quantum gravity and bf theory.
Lect.Notes Phys., 543:25–94, 2000.
113. D. Yetter L. Crane. A Categorical construction of 4-D topological quantum field the-
ories. in “Quantum Topology” L Kaufmann and R Baadhio Eds. (World Scientific,,
Singapore, 1993.
114. D.N. Yetter L. Crane, L. Kauffman. State-sum invariants of 4-manifolds. J Knot
Theor Ramifications, 6:177–234, 1997.
115. J. W. Barrett J. C. Baez. The quantum tetrahedron in 3 and 4 dimensions.
Adv.Theor.Math.Phys., 3:815–850, 1999.
116. L. Crane, A. Perez, and C. Rovelli. A finiteness proof for the lorentzian state sum
spinfoam model for quantum general relativity. gr-qc/0104057.
117. L. Crane, A. Perez, and C. Rovelli. Perturbative finiteness in spin-foam quantum
gravity. Phys. Rev. Lett., 87:181301, 2001.
118. A. Perez. Finiteness of a spinfoam model for euclidean quantum general relativity.
Nucl.Phys. B, 599:427–434, 2001.
119. A. Perez and C. Rovelli. A spin foam model without bubble divergences. Nucl.Phys.
B, 599:255–282, 2001.
120. A. Perez M. Bojowald. Spin foam quantization and anomalies. gr-qc/0303026, 2003.
121. M. P. Reisenberger. On relativistic spin network vertices. J.Math.Phys., 40:2046–
2054, 1999.
122. Rodolfo Gambini and Jorge Pullin. A finite spin-foam-based theory of three and four
dimensional quantum gravity. Phys. Rev., D66:024020, 2002.
123. K. Krasnov L. Freidel. Spin foam models and the classical action principle.
Adv.Theor.Math.Phys., 2:1183–1247, 1999.
124. A. Perez and C. Rovelli. 3+1 spinfoam model of quantum gravity with spacelike and
timelike components. Phys.Rev. D, 64:064002, 2001.
125. A. Perez and C. Rovelli. Spin foam model for lorentzian general relativity. Phys.Rev.
D, 63:041501, 2001.
126. Karim Noui and Philippe Roche. Cosmological deformation of lorentzian spin foam
models. Class. Quant. Grav., 20:3175–3214, 2003.
127. L. Crane and D.N. Yetter. On the classical limit of the balanced state sum.
gr-qc/9712087.
128. J. W. Barrett and R. M. Williams. The asymptotics of an amplitude for the 4-simplex.
Adv.Theor.Math.Phys., 3:209–215, 1999.
129. J. W. Bar-
rett. The classical evaluation of relativistic spin networks. Adv.Theor.Math.Phys.,
2:593–600, 1998.
130. G. Egan J. Baez, D. Christensen. Asymptotics of 10j symbols. gr-qc/0208010.
131. John W Barrett and Christopher M. Steele. Asymptotics of relativistic spin networks.
Class. Quant. Grav., 20:1341–1362, 2003.
132. R. De Pietri, L. Freidel, K. Krasnov, and C. Rovelli. Barrett-crane model from a
boulatov-ooguri field theory over a homogeneous space. Nucl.Phys. B, 574:785–806,
2000.
133. M. P. Reisenberger and C. Rovelli. Spacetime as a feynman diagram: the connection
formulation. Class.Quant.Grav., 18:121–140, 2001.
134. M. P. Reisenberger and C. Rovelli. Spin foams as feynman diagrams. gr-qc/0002083.
135. C. Petronio R. De Pietri. Feynman diagrams of generalized matrix models and the
associated manifolds in dimension 4. J. Math. Phys., 41:6671–6688, 2000.
136. A. Mikovic. Quantum field theory of spin networks. Class.Quant.Grav., 18:2827–2850,
68 Alejandro Perez

2001.
137. A. Perez and C. Rovelli. Observables in quantum gravity. gr-qc/0104034.
138. D. Boulatov. A model of three-dimensional lattice gravity. Mod.Phys.Lett. A, 7:1629–
1646, 1992.
139. H. Ooguri. Topological lattice models in four dimensions. Mod.Phys.Lett. A, 7:2799–
2810, 1992.
140. Laurent Freidel and David Louapre. Non-perturbative summation over 3d discrete
topologies. hep-th/0211026, 2002.
141. Fotini Markopoulou. Coarse graining in spin foam models. Class. Quant. Grav.,
20:777–800, 2003.
142. F. Markopoulou. An algebraic approach to coarse graining. hep-th/0006199.
143. Robert Oeckl. Renormalization of discrete models without background. Nucl. Phys.,
B657:107–138, 2003.
144. Robert Oeckl. Renormalization for spin foam models of quantum gravity.
gr-qc/0401087, 2004.
145. A. Mikovic. Spin foam models of matter coupled to gravity. Class.Quant.Grav.,
19:2335–2354, 2002.
146. A. Mikovic. Quantum field theory of open spin networks and new spin foam models.
gr-qc/0202026.
147. L. Smolin F. Markopoulou. Causal evolution of spin networks. Nucl.Phys. B, 508:409–
430, 1997.
148. F. Markopoulou. Dual formulation of spin network evolution. gr-qc/9704013.
149. L. Smolin F. Markopoulou. Quantum geometry with intrinsic local causality.
Phys.Rev. D, 58:084032, 1998.
150. Martin Bojowald and Hugo A. Morales-Tecotl. Cosmological applications of loop
quantum gravity. Lect. Notes Phys., 646:421–462, 2004.
151. Hanno Sahlmann and Thomas Thiemann. Irreducibility of the ashtekar-isham-
lewandowski representation. gr-qc/0303074, 2003.
152. Hanno Sahlmann and Thomas Thiemann. On the superselection theory of the weyl
algebra for diffeomorphism invariant quantum gauge theories. gr-qc/0302090, 2003.
153. Hanno Sahlmann. When do measures on the space of connections support the triad
operators of loop quantum gravity? gr-qc/0207112, 2002.
154. Hanno Sahlmann. Some comments on the representation theory of the algebra un-
derlying loop quantum gravity. gr-qc/0207111, 2002.
155. Andrzej Okolow and Jerzy Lewandowski. Diffeomorphism covariant representations
of the holonomy- flux *-algebra. Class. Quant. Grav., 20:3543–3568, 2003.
156. Andrzej Okolow and Jerzy Lewandowski. Automorphism covariant representations
of the holonomy-flux *-algebra. gr-qc/0405119, 2004.
157. J. Lewandowski, A. Okolow, H Sahlmann, and Thiemann T. Uniqueness of the dif-
feomorphism invariant state on the quantum holonomy-flux algebra. Preprint, 2004.