Quantum geometry

with intrinsic local causality

Fotini Markopoulou and Lee Smolin
arXiv:gr-qc/9712067v1 16 Dec 1997

Center for Gravitational Physics and Geometry

Department of Physics
The Pennsylvania State University
University Park, PA, USA 16802

December 2, 1997

ABSTRACT
The space of states and operators for a large class of background in-
dependent theories of quantum spacetime dynamics is defined. The SU (2)
spin networks of quantum general relativity are replaced by labelled compact
two-dimensional surfaces. The space of states of the theory is the direct sum
of the spaces of invariant tensors of a quantum group Gq over all compact
(finite genus) oriented 2-surfaces. The dynamics is background independent
and locally causal. The dynamics constructs histories with discrete features
of spacetime geometry such as causal structure and multifingered time. For
SU (2) the theory satisfies the Bekenstein bound and the holographic hy-
pothesis is recast in this formalism.

email addresses: [email protected], [email protected]

1
1 Introduction
In this article we describe a new class of quantum geometries that is to
be used in developing a theory of quantum gravity. These are natural ex-
tensions of the spin network states that have been shown to comprise the
non-perturbative state space of quantum general relativity[1, 2]1 . In this for-
mulation the labeled graphs on which spin networks are based are replaced
by 2-manifolds and the invariant tensors of a quantum group Gq associated
to them. The motivations for this generalization comes partly from results
of non-perturbative quantum gravity and string theory.
This formulation has a kinematical part and a dynamical part. The
kinematical part, which is described in the next four sections, generalizes
the spin network states in two ways. The first is that the SU (2) group
of the spin network states of quantum general relativity is replaced by an
arbitrary quantum group Gq . Within the framework of non-perturbative,
diffeomorphism invariant quantum field theories, this is the natural way to
extend the degrees of the freedom of the theory to include gauge fields[6] and
supersymmetry [7]. The quantum deformation is motivated from physics by
three considerations. First, in quantum general relativity the introduction
of a cosmological constant Λ requires a quantum deformation of SU (2) with
q = e2π/k+2 defined by[8, 9, 10]
6π
k= . (1)
G2 Λ
Second, the truncation in the number of representations with q at a root of
unity improves the formulation of the dynamics by making the sums involved
in the path integral less divergent. It also introduces new symmetries in the
theory which are not present in the classical case when q → 1. These are
analogous to the duality symmetries of perturbative string theory. As we
argue below, this may play a role in the interpretation of the theory.
The second sense in which our proposal extends the spin network states
of quantum general relativity is that the states are defined intrinsically,
without the use of a background manifold. In quantum general relativity
the spin network states are diffeomorphism classes of embeddings of graph
in a fixed three-manifold Σ [1, 2]. We go beyond this to a purely algebraic
definition of the state space which depends on no prior specification of a
manifold.
1
For recent reviews see [3, 4]. Spin networks were originally introduced by Penrose [5]
as a model of quantum geometry.

2
 One result of non-perturbative quantum gravity has been the discov-
ery that geometrical quantities, including area[11, 2], volume[11, 2, 12] and
length[13] have discrete spectra. This is true before the introduction of
dynamics or matter couplings and signals that the combination of diffeo-
morphism invariance and quantum theory requires that quantum geometry
be essentially discrete. At the same time, the application of these tech-
niques to the hamiltonian constraint of general relativity [14, 15, 16] leads
to a theory without a good continuum limit[17, 18]. Given this, it seems
more natural to construct the theory purely from algebra and combinatorics
and let continuum notions arise in the classical limit of the theory.
We may note that the dualities of string theory suggest that one and
the same physical situation may sometimes be described in two different
ways, which differ in the topology and manifold structures of the underlying
manifolds[19]. Other results show that in string theory there are continuous
phase transitions whose semiclassical description involves abrupt changes
of the topology of the underlying manifold[20]. These suggest that the
fundamental, non-perturbative, description should not be based on fixed
topological manifolds.
But without background manifolds the theory cannot be formulated in
terms of the embeddings of surfaces or membranes. The alternative is to
construct the states and operators that are to represent quantum geometry
algebraically, using only combinatorics and representation theory. This is
the main goal of this paper. In [21] and [22] results are presented consistent
with the hypothesis [4] that the resulting extension of the spin networks
formalism may serve as framework for non-perturbative string theory.2
A theory formulated without reference to any background manifold still
requires dynamics and that dynamics should have built into it some notion
of local causality. Below, in section 7, we show that this can be achieved
by an extension of a formulation of spin network dynamics proposed earlier
by one of us[23]. The dynamics is based on discrete histories M, which
are combinatorial structure which have two properties shared by classical
spacetime:

1. Each history M contains a finite set E of elements that may be called

2
In fact the basic idea of the present formulation is rather like the idea behind the
transition from quantum field theory to perturbative string theory. Just as Feynman
diagrams are replaced by string worldsheets, the present generalization of quantum general
relativity extends spin network states to 2-dimensional surfaces and the states of field
theories defined on the surfaces.

3
 “events”. This set of events is a partially ordered set. We thus have
the finite element analogue of the points of a Lorentzian spacetime.

2. Each history M contains a large number of connected sets of causally

unrelated events, which may be called “quantum spacelike surfaces”.
Each spacelike surface is also a quantum state. Thus, the theory has a
discrete analogue of the many-fingered time of general relativity, which
means that a discrete analogue of spacetime diffeomorphism invariance
is built in.

Each history is then given an amplitude which is a product of factors

each associated to a local transition in the quantum geometry. Causality
and locality impose restrictions on the choice of these amplitudes which
are discussed below. The issue of the choice of dynamics and the related
question of the continuum limit is discussed in [24, 22].
In the next three sections we introduce the space of states that we pro-
pose extends spin networks and describe useful decompositions of them
which are based on 3- and 4-punctured spheres. Section 5 introduces an
algebra of operators that act on the states and the interpretations of some
of them, which yields a picture of quantum geometry, is the subject of sec-
tion 6. The dynamics of the theory is described in section 7, while section
8 discusses coarse-grained observables and entropy and their relationship to
the holographic hypothesis and Bekenstein bound. The conclusion is largely
devoted to describing ongoing work that will be reported in other papers.

2 The space of states

The space of states that we investigate here is both the extension of SU (2)
spin networks to a quantum group Gq and of the spin network states of
canonical quantum gravity to the non-embedded case.
Given a quantum group Gq and a compact oriented 2-surface S, let VG S
q
be the space of Gq invariant tensors on S.3 We then define the space of
states HGq of Gq quantum gravity to be
M
S
H Gq = VG q
(2)
S
3
Equivalently this is the space of conformal blocks of the WZW theory corresponding
to level k on S [25, 26, 27] or the space of states of Gq Chern-Simons theory on S, seen
as a spatial slice of some 3-manifold[28].

4
where the sum is over all compact 2-surfaces of finite genus. Each VG S is
q
finite dimensional when q is at a root of unity. HGq is equipped with the
natural inner product (see (5) below) and is a Hilbert space.
The sense in which these states may be considered to constitute an exten-
sion of the spin network states of quantum general relativity will be discussed
shortly, but we note that this is not a new notion. It is known that the quan-
tum deformation of spin networks requires that their edges be enlarged to
ribbons or tubes[28, 26, 27]. This is to allow dependence of the states on
twistings of the edges, necessary for the q-deformed case[28, 26, 30, 29]. In
the next sections we investigate properties of these states that are important
for their physical interpretation.

3 Trinion decomposition: basis states

We begin by reviewing some of the properties of the state spaces VG S that
q
we will need to discuss their role in representing the states of quantum
gravity4 . For the purposes of describing the states and operators on HGq it
will be very useful to understand the behavior of the states in VG S under
q
decompositions of the surface S into a union of punctured spheres. We begin
by discussing the decomposition of a genus g surface S into 3-punctured
spheres, or trinions. Given a surface S we may choose a maximal set of
non-intersecting elements of π 1 [S], which we shall call circles, cα . Cutting S
along the circles cα decomposes it into a set of N trinions, BI3 , I = 1, ..., N .
The trinions are joined on their punctures so that each circle cα corresponds
to the punctures on two trinions. (See Figure 1a.) This may be done in
several different ways. (See Figure 1b.)
A trinion decomposition will be called non-degenerate if no two trinions
meet at more than two circles (see Figure 2).
Associated to each trinion decomposition of S is a class of bases of VG S ,
q
which is constructed as follows. Gq has a list of irreducible representations,
which we will label by jα . (For the q taken at a root of unity, which we
will assume, this is a finite list.) Each of the three circles of a trinion BI3
may be labeled by a representation jα , α = 1, 2, 3. For each choice of the
representations jα there is a linear space VjI1 j2 j3 of intertwiners µI . The
intertwiners are the maps

µI : j1 ⊗ j2 ⊗ j3 −→ 1. (3)
4 S
Complete characterizations of VG q
may be found in [25, 26].

5
 cα
3
BI
(a) (b)

Figure 1: (a) A genus 4 surface cut into six trinions BI3 by circles cα . (b)
The same surface in a different trinion decomposition.

Figure 2: This trinion decomposition is degenerate because the two trinions

have two circles in common.

6
A choice of a set of jα on the punctures of a trinion is called consistent if
the corresponding VjI1 j2 j3 has strictly positive dimension.
The space of states VG S (subset of H ) associated with the surface S
q Gq
is constructed by taking direct products of all the constituent spaces VjI1 j2 j3
and summing over the representations,
XO
S I
VG q
= V{j}I
(4)
jα I

where I labels an arbitrary trinion BI3 in S with labels {j}I .

A generic state in VG S will be denoted |S, Ψi. A basis in V S is then
q Gq
constructed as follows. We choose an orthogonal basis of intertwiners in the
space V{j}I
I
of each of the trinions, denoted µρI . A basis of states in VG
S is
q
then given by a choice of jα on each circle cα in S and a choice of a basis
element µρI on each trinion. These basis states are denoted |S, jα , µρI i.
Given a trinion decomposition of every finite genus 2-surface S, the states
|S, jα , µρI i provide an orthonormal basis for the state space HGq . The inner
product on HGq is given by
Y Y
hS, jα , µρI |S ′ , jα′ , νIτ i = δSS ′ δjα jα′ hµρI |νIτ iI (5)
α I

where the same trinion decomposition is assumed for the two states when
ρ
S∼ = S ′ and hµI |νIτ iI is the inner product in the space of intertwiners V I on
the I-th trinion.
Note that, given a particular trinion decomposition of S, the states in
the basis |S, jα , µρI i may be thought of as generalized combinatorial trivalent
spin networks. (See Figure 3 but note that for q 6= 1 these are quantum
spin networks[30].). The edges eα of the corresponding graph Γ are labeled
with the same representations jα as the corresponding circles cα , while the
trivalent nodes vI associated to the trinions are labeled by the intertwiners
µI . Because of this association we sometimes call the basis states |S, jα , µρI i
tubular spin networks.
The assignment of a graph Γ to the surface S depends on the choice of the
trinion decomposition and the same is thus true of the basis |S, jα , µρI i. If we
choose a different trinion decomposition of S, based on a different maximal
G
set of non-intersecting elements of π 1 [S], we have a different basis for VS q .
The recoupling identities of the representation theory of Gq [30] then provide
the change of basis formulas. Alternatively, they may be computed using
the modular transformations of the corresponding rational conformal field
theory as in [25, 26].

7
Figure 3: A trinion decomposition of a genus 5 surface reduced to a spin
network graph.

We may note that when q → 1 the spaces VGS become infinite dimensional

as there are an infinite number of representations jα . Then, the Moore-

Seiberg operators are no longer well defined unitary operators. Thus, in
the limit q → 1 the states in HGq are the usual combinatorial spin network
states of SU (2).

4 Decomposition in 4-punctured spheres

Just as the trinion decomposition is related to an extension of trivalent
spin networks, we can associate an extension of 4-valent spin networks to
the bases of states in V S that come from decomposing S into 4-punctured
spheres (from now on we drop the suffix Gq of VG S ). To accomplish this we
q
pick a (non-maximal) set of non-intersecting circles cα on S that decompose
it into 4-punctured spheres BI4 . As before we can label these circles with
representations jα .
It will also be useful to work with general n-punctured spheres. In gen-
eral, a 2-sphere with n punctures, denoted BIn , is labelled by representations
j1 , ..., jn of the group Gq . Given BIn and the labels j1 , ..., jn , there is a linear
space of intertwiners, VjI1 ,...,jn , consisting of the invariant maps
µI : j1 ⊗ ... ⊗ jn −→ 1. (6)
As in the 3-punctured case, the dimension of VjI1 ,...,jn is required to be non-
zero otherwise the choice of j1 , ..., jn is inconsistent and not allowed. Now,
given any decomposition of S into n-punctured spheres along a set of circles
cα , we have representation of the states in V S in terms of triples |S, jα , µI i.
The formulas (4) and (5) still hold.

8
Figure 4: The tubular 4-simplex P, a genus 6 surface decomposed to 5
4-punctured spheres.

Returning to the decompositions in terms of 4-punctured spheres, we

may note that such decompositions may be made, at least locally, in a
surface S by grouping the trinions in some trinion decomposition into pairs.
Finally, as in the case of trinions, we call a decomposition of a surface S
into 4-punctured spheres non-degenerate if no two 4-punctured spheres share
more than one puncture.

4.1 The tubular 4-simplex

In fact, a genus g surface always has a non-degenerate decomposition into
4-punctured spheres for g ≥ 6. It is easy to see that the smallest number of
4-punctured spheres that can fit together non-degenerately is 5. These make
up a genus 6 surface which may be thought of as a tubular generalization
of the 4-simplex (see [23]), as every 4-punctured sphere BI4 , I = 1, ..., 5
is connected to every other one once. (See Figure 4). This surface plays a
special role in the dynamics. We shall call it P and refer to as the generating
surface. Together with ten fixed circles cIJ connecting BI4 and BJ4 that
decompose it into five such 4-punctured spheres P will be called the tubular
4-simplex. Its q → 1 limit is a 4-valent graph with 4-valent nodes vI for
each 4-punctured sphere BI4 of S and an edge eIJ for for each circle cIJ .
Labeling the circles cIJ by representations jα and the BI4 by a basis µρI
B4
in the corresponding spaces of intertwiners V{jIα } we have a basis of states
|P, jα , µρI i. Each of these is a coloring of the 4-simplex.

9
4.2 Tubular evolution moves
Consider a non-degenerate decomposition of a surface S into n 4-punctured
spheres,
n
K
S= BI4 , (7)
I=1
where ⊙ denotes the gluing of a pair of punctures with the same labels.
Given S, there is a set of local moves each of which yields another surface
S ′ expressed as a non-degenerate composition of 4-punctured spheres S ′ =
Jm 4 5
I=1 BI where in general m 6= n.
To define these moves let us now put forward some notation. An ele-
mentary local region, L, is a set of n ≤ 4 4-punctured 2-spheres BI4 ,
n
K
L= BI n ≤ 4, (8)
I=1

each pair of which is connected by exactly one tube. L, therefore, is a

2-surface with 4 or 6 punctures, their number given by

number of free punctures = 4n − n(n − 1). (9)

n has to be at most 4 for L to have any free punctures. For n = 1, 2 the

genus (not counting punctures) of L is 0, for n = 3 it is 1 and for n = 4 it
is 3.
Given these definitions, a local move is the following. Given a decompo-
J
sition S = nI=1 BI4 , remove an elementary local region L in it and replace
it with a new one L′ that has the same number of punctures and same labels
on the punctures. (See Fig. 5).
The topology of the new local region is determined by requiring that L
and L′ can be composed along their common punctures to form the gener-
ating surface, P (see Figure(6)),

L′ ⊙ L = PL′ ⊙L . (10)

Namely, L′ is the complement of L in PL′ ⊙L . 6 We call such a substitu-

tion a tubular evolution move. (See Figs. 5, 6 and 7). The result is a new
5
These moves are a generalization of the Pachner moves from combinatorial topology
[31] that played an important role in the evolution of spin networks in [23].
6
In terms of the vector space representations of L and L′ , equation (10) is the tensor
 ′
D
product of the vector space of L with the dual vector space of L′ , V L ⊗ V L .

10
 Figure 5: An elementary substitution move.

L L
L

Figure 6: Left: The substitution move seen as a three manifold that defines
a cobordism from L to L′ . Right: Joining the L and L′ together makes a
generating surface P.

11
 Figure 7: The four elementary substitution moves, 1 ↔ 4 and 2 ↔ 3.

2-manifold S ′ which has a decomposition into 4-punctured spheres that fall

into two sets, those in S − L and those in L′ .
It is clear that if the original decomposition of S into 4-punctured spheres
is non-degenerate then so is the new one. No two 4-punctured spheres in
S − L share more than one connection because the original decomposition is
non-degenerate. The same is true for the 4-punctured spheres in L′ because
every elementary local region is non-degenerate. The non-degeneracy of L′
implies that there can be at most one connection between any sphere in L′
and one in S − L. 7
There are four kinds of tubular evolution moves, depending on the num-
ber of 4-punctured spheres in the old and new elementary regions L and
L′ . As in the case of the Pachner moves used in [23], these are denoted the
1 → 4, 4 → 1, 2 → 3 and 3 → 2 moves. In terms of the corresponding
surfaces one can see from Figure 7 that these result in a change of genus by
+3, −3, +1 and −1 respectively. This means that starting with the tubular
4-simplex P which has genus 6, one can make r successive 2 → 3 moves to
reach a surface of any genus g = 6 + r. Therefore, each surface with genus
g ≥ 6 has a non-degenerate decomposition into 4-punctured spheres.
Note also that if B 4 is a 4-punctured sphere with labels j1 , j2 , j3 , j4 it
may be decomposed along a circle c1 into two trinions B13 and B23 . If we call
the label on c1 by l we have
4 X B3 B3
VjB1 j2 j3 j4 = Vj11j2 l ⊗ Vlj32j4 . (11)
l
7
This is an extension of the basic fact that the Pachner moves applied on a P L
triangulation takes preserves non-degeneracy of triangulations, i.e. when no two tetrahedra
share more than one face.

12
Thus, we see that there are many trinion decompositions of a surface S that
are subdivisions of a decomposition of S into 4-punctured spheres. In terms
of the analogy to spin networks this corresponds to what has been called
decomposing a 4-valent node of a spin network in terms of two trivalent
nodes and an internal, or “virtual”, edge. In the present context all of these
are connected by elements of the modular group [25, 26].
Clearly, a given surface S has more than one inequivalent decompositions
into 4-punctures spheres. As an example, consider the tubular 4-simplex of
Figure 4. The relationship between these different compositions correspond
to transformations between two bases in which the roles of the representa-
tions and the intertwiners are exchanged. This has interesting consequences
for quantum geometry that we will discuss below, when we describe how the
geometrical interpretation of the theory is constructed.
We will use the tubular evolution rules to define the dynamics of the
theory. But first we have to define operators on HGq that implement them.

5 Tube operators
We now turn to the operators on the space of states HGq . The Moore-
Seiberg [25] operators are a set of unitary operators that act inside each V S .
However, if our theory is to be a generalization of spin networks there must
′
be operators that take us from states in one V S to states in another V S on a
different surface S ′ . We will see here that several useful sets of operators can
be constructed, which will play a role in the interpretation and dynamics of
the theory. They are analogous to the loop operators whose algebra defines
the loop representation of general relativity [32]. Here, because the states
are defined without any reference to a background manifold, the operators
are defined relationally, in terms of decompositions of the surfaces S into
pieces.
Let Υ denote a genus g compact oriented 2-surface with n ≥ 1 punctures
jk (k = 1, .., n). Given a compact S let rI denote the maps

rI : Υ → S (12)

taking Υ homomorphically to a component of S. In general there will be a

set of such maps; they are distinguished by the index I.
For each I the map picks out a set of n non-intersecting circles cIk , k =
1, ..., n in S. Cutting S on these circles decomposes it into the two pieces

13
rI (Υ) and (S − rI (Υ)). The space of intertwiners V S decomposes as
X (S−r (Υ))
VS = VjΥ1 ...jn ⊗ Vj1 ...jnI . (13)
k

A state |S, Ψi ∈ V S then decomposes to a sum over the representations

(S−r (Υ))
j1 , ..., jn of the product of a state in VjΥk and a state in Vjk I ,
X
|S, Ψi = |Υ, jk , Ψ1 i ⊗ | (S − rI (Υ)) , jk , Ψ2 i. (14)
k

Using this decomposition we then define three classes of operators. The

first two are block diagonal in the decomposition (2), while the third changes
the topology of the surface S.

5.1 Surface operators

Let F ({jk }) be a symmetric function of n representation labels {jk }. Given
such a function and a 2-surface Υ with n punctures, there is an hermitian
operator FbΥ that acts in HGq as follows. On the spaces VjΥ1 ...jn , FbΥ is the
diagonal operator equal to F ({jk }). Then on a general state
XX
FbΥ |S, Ψi = F ({jk })|Υ, jk , Ψ1 i ⊗ | (S − rI (Υ)) , jk , Ψ2 i. (15)
I k

This operator looks for the instances of the submanifold Υ in each surface
S and, in each state, measures a property of the boundary separating Υ from
the remainder S − rI (Υ) given by the function F ({jα }). The punctured sur-
face Υ can be thought of as the algebraic representation of a 3-dimensional
region R in the quantum geometry that a state in V S represents. The sur-
face operators thus measure properties of boundaries of regions in space.
This interpretation will be developed in the next section, where we will see
that an example in the case of SU (2) is given by the area operator obtained
in quantum general relativity [11, 2].

5.2 Bulk operators

Once a region R of an abstract quantum geometry has been identified by a
map rI : Υ → S we can also try to measure bulk properties of that region.
These will be eigenvalues of operators that act on the space VjΥ1 ...jn and
depend on the topology of Υ and hence on dimVjΥ1...jn . To define such an

14
 bj ...j on V Υ
operator let us choose, for every j1 ...jn an operator B 1 n j1 ...jn . The
bΥ
corresponding bulk operator Bj1 ...jn is defined on the state space HGq as
XX
BbjΥ1 ...jn |S, Ψi = b
B|Υ, jk , Ψ1 i ⊗ | (S − rI (Υ)) , jk , Ψ2 i. (16)
I k

Examples of bulk operators are the volume operators which we will de-
scribe in the next section.

5.3 Substitution operators

In the last section we defined the tubular evolution moves. These are exam-
ples of a large class of substitution operations that take us from one manifold
S to a different manifold S ′ by cutting out a piece, Υ1 of S and replacing
it with a different manifold Υ2 with the same boundary. The tubular evo-
lution moves are examples of these. For such substitutions we can define
′
linear operators that act on HGq and take states from V S to those in V S .
Start with two punctured surfaces, Υ1 and Υ2 , each with an ordered set
of n punctures, with labels j1 , ..., jn . They can be represented by vector
spaces VjΥ11...jn and VjΥ12...jn . Note that in general dimVjΥ11...jn 6= dimVjΥ12...jn .
Given two vector spaces, we have the  space of linear maps from the first
 to
the second, denoted hom V Υ1 , V Υ2 . A particular cb ∈ hom V Υ1 , V Υ2 acts
on a state, |Υ1 , jk , Ψ1 i ∈ VjΥ11...jn as

cb|Υ1 , jk , Ψ1 i = |Υ2 , jk , Ψ2 i ∈ VjΥ12...jn , (17)

giving a state in VjΥ12...jn . Given any such cb we construct a substitution

operator CbΥ1 ,Υ2 ,ĉ , defined by
XX h i
CbΥ1 ,Υ2 ,bc |S, Ψi = | (S − rI (Υ1 )) , jk , Ψ1 i ⊗ cb|Υ1 , jk , Ψ1 i . (18)
I k

The action of CbΥ1 ,Υ2 ,bc is pictured in Fig. 5.

Note that we may also glue Υ1 and Υ2 along their identical boundaries
as in the right hand figure of Fig. 6.

6 Geometrical interpretations
So far, we have defined states in HGq in terms of labelled 2-dimensional
manifolds. We shall now interpret them in terms of observables related to

15
3-dimensional space. These arise as natural extensions of the observables of
quantum general relativity: the area and volume operators.
The subtlety is that here there is no background manifold. All of the
properties of space, including its topological and metric properties, must be
coded into the states. In the absence of any background manifold to provide
surfaces and regions, geometrical observables are constructed relationally,
from information coded into the states.
Let us begin with the space of states V S associated to a given 2-surface
S. A microscopical geometrical interpretation of these states exists for every
decomposition of S into a set of n-punctured 2-spheres, BIn , with n ≥ 3,
joined on a set of circles, cα . Let us consider a basis of states which is
(partially) determined by definite values for the representations jα for these
Bn
circles. This state is of the form, |S, jα , µI i with intertwiners µI ∈ Vj1I...jn
for each of the punctured spheres.
The geometrical interpretation is constructed as follows. Associated to
each BIn is a region RI . These regions have three kinds of properties:

• Surface properties: A surface property of a region RI is a function

F (j1 , ..., jn ) of the labels on the punctures of the corresponding BIn .
Surface properties are measured by surface operators (15).

• Bulk properties: A bulk property of a region RI is measured a

Bn
hermitian operator Bb (16) in the space of intertwiners Vj1I...jn .

• Shared properties: Two regions RI and RJ may have shared prop-

erties if they have a set of common punctures with labels, say, j1 , ..., jk .
If this set is non-empty, then j1 , ..., jk is the common boundary of
RI and RJ . A shared property of RI and RJ is then a function
G(j1 , ..., jk ).

In the SU (2) case we may import the kinematical structure from quan-
tum general relativity found in [11, 2, 1] to give us examples of each kind of
observable:

• The areap of RI is a surface property. It is given by F (j1 , ..., jn ) =

P
lP2 l nα=1 jα (jα + 1).

• The volume of the interior of BIn is an example of a bulk property.

As we know from [2, 12] the volume operator is a hermitian operator

16
 Bn
Vb [j1 , ..., jn ] that acts in the space or intertwiners Vj1I,...,jn 8

• The area also gives an example of a shared property. If BIn and BJm
share a set of k spins j1 , ..., jk then the area of the common boundary of
2 Pk p
RI and RJ is given by lP l α=1 jα (jα + 1) summed over the common
punctures of the two regions.

Note that, given a division of S into punctured spheres, we may simul-

taneously diagonalize all of the area and volume operators on the corre-
sponding regions RI . Thus, a common eigenstate |S, jα , µI i may be called
a microscopic quantum geometry. It is a set of regions together with i) an
area for the boundary of each one, ii) an area for each common boundary,
such that the area of each is the sum of its common boundaries with the
others and iii) a volume for each region.
For a general Gq , we expect a generalized microscopic quantum geome-
try to be the maximal set of simultaneous eigenvalues of surface and bulk
observables for a decomposition of S into punctured spheres. There is also a
notion of a coarse-grained quantum geometry. We will discuss this in section
8.

6.1 Duality between edges and intertwiners

The reader may have noticed that the geometrical interpretations available
to the states in V S are not determined by S. There is a geometrical interpre-
tation for every way of dividing S into punctured spheres. We regard this
freedom as an intrinsic and attractive feature of the generalization from spin
network states to the space of states HGq . For example, in the 4-valent spin
networks two kinds of edges appear: real edges connecting the nodes and
“virtual” edges that may be used to label the intertwiners of the 4-valent
nodes. Thus, in the usual spin network formalism they play different roles.
For example, consider the tubular 4-simplex P and the two different
decompositions into 4-punctured spheres illustrated in Figure 4. These may
be described in terms of two sets of circles cα and cα′ , as shown in Figure 8.
(The full set (cα , cα′ ) make up a maximal set of non-intersecting circles on
P and define a trinion decomposition of P.) This decomposition represents
states |P, j(cα ), µI )i. If we now read the decomposition with a different set
8
Here we take the definition given in [2] that does not require any assumptions about
structure not present in our case such as linear relations among tangent vectors at the
nodes.

17
 cα
cα
cα
cα cα
cα
cα cα
cα

Figure 8: A choice of decomposition into 4-punctured spheres where cα are

representations and c′α intertwiners.

of circles, including c′α , separating the five 4-punctured spheres, we obtain

a different set of basis states |P, j(c′α ), µI ′ i. This shows that the distinction
between spins and intertwiners in this formalism is dependent on the choice
of n-punctured spheres. Therefore, so is the geometrical interpretation.

7 Causal evolution
We now discuss the evolution of the states in HGq . The dynamics of the
theory will be based on the evolution moves defined in section 4.2. By
composing the moves we produce sequences of states that we call histories.
These discrete histories share three characteristics of Lorentzian spacetimes.
i) There is a set of events which is a discrete partially ordered set with no
closed causal loops. This is a discrete analogue of a Lorentzian spacetime.
ii) There are connected sets of causally unrelated events, the combinatorial
analogues of spacelike surfaces. iii) A history can be decomposed in many
ways into sequences of spacelike surfaces, leading to a discrete analogue of
many fingered time.

7.1 The evolution operator

The evolution of states is generated by an operator that implements the
evolution moves described in section 4.2. This will be a substitution operator
of the form defined in section 5.3. To do this let ρ = 1, 2, 3, 4 correspond
to the four kinds of Pachner moves 1 → 4, 2 → 3, 3 → 2, 4 → 1. Then take
Lρ to be the elementary local region consisting of ρ 4-punctured spheres, so
that L′ρ , the complement of Lρ in P consists of 5 − ρ 4-punctured spheres.
We will call Lρ the past set and L′ρ the future set of the P associated with

18
the ρ → 5 − ρ move (See Figure 7.).
For ρ = 1 and 4 the Lρ and L′ρ each have 4 punctures, which are labeled
by representations jγ , γ = 1, ..., 4. For ρ = 2, 3 there are six punctures and
γ = 1, ..., 6. For each ρ and sets of 4 or 6 representations jγ we may choose
L L′
operators cbρ,jγ ∈ hom(Vjγρ , Vjγρ ). The ρ’th move is then implemented by the
substitution operator
b ρ |S, Ψi = Cb
H cρ |S, Ψi
Lρ ,L′ρ ,b
nρ
XX
= | (S − rI (Lρ )) , jγ , Ψi ⊗ [cbρ |Lρ , jγ , Ψi] . (19)
I k=1

The total evolution operator is then given by

X
b =
H bρ
H (20)
ρ

To see how these act, let us start with an initial state |S, Ψi and act on
it with one of the H ρ . If S is large enough, there will be numerous regions
in it homeomorphic to Lρ . To each of them there is a map rI : Lρ → S.
For each I we then cut from S the region rI (Lρ ) and replace it by L′ρ . This
results each time in a new 2-surface which we call Sρ,I . The result of the
application of H b ρ is then a superposition of the states given by the action
(19). The exact map from the old states to the new states is given by the
linear maps cbρ . (We suppress the dependence of cbρ on the representations
jγ .)
The operator H is hermitian when each of the cbρ are appropriately cho-
sen. In this case a formal unitary evolution operator may be written down
as
b
Ub = eıHt (21)
where t is a parameter having nothing to do with the physical time (it
just scales the operators cbρ .) The amplitude for an initial state |initiali =
|Sinitial , Ψinitial i to evolve to a final state |f inali = |Sf inal , Ψf inal i is formally
given by
A[|initiali → |f inali] = hf inal|Ub |initiali. (22)

7.2 Amplitudes for causal evolution and a discrete path in-

tegral
n  
b
By decomposing the action of Ub at each order n of the action of H in
terms of 4-punctured spheres produced by the evolution moves, the am-

19
plitude (22) can be given in terms of a sum over a set of histories, M =
{|1i, |2i, |3i....} in which each |I + 1i results from the previous |Ii by the
application of one of the four moves. The theory gives an amplitude to each
transition from an initial state to one of its successor states. The amplitude
is given by
AL→L′ = hL′ , jk′ , µ′I |cbρ |L, jk , µI i, (23)
i
where |L, jk , µI is a trinion basis state for the initial elementary local region
to be cut out and |L′ , jk′ , µ′I > is a basis state on the elementary local region
that replaces it.
Consider now an (N − 1)-step history M = {|1i, |2i, ..., |N i}. Each
transition is a generalized evolution move which has an amplitude AI given
by (23) for the transition from |Ii to |I + 1i, I = {1, ..., N − 1}. The
amplitude of the history M is then given by
Y
A[M] = AI . (24)
I

Let us then have two states, |initiali and |f inali. There is an infinite number
of histories M such that the first state is equal to |initiali and the last state
is equal to |f inali. By analogy to the simplical case we may denote this
as ∂M = |initiali ∪ |f inali. The transition amplitude to evolve to |f inali
given |initiali is then,
X
A[|initiali → |f inali] = A[M]. (25)
M|∂M=|initiali∪|f inali

As this is an infinite sum one may first compute the amplitude for
|initiali to evolve to |f inali in N steps. This is given by
X
AN [|initiali → |f inali] = A[M], (26)
M||1i=|initiali,|Ni=|f inali

i.e., the sum over (N − 1)-step histories that take the initial to the final
state. However, note that while the full amplitude (22) is formally unitary
by construction the same is not the case for the N step amplitude (26).

7.3 The causal structure

We now show that each history M has defined on it a discrete causal struc-
ture as a result of its construction from the evolution moves. Each his-
tory consists of N states |Ii which are elements of HGq . Furthermore, the

20
states |Ii come as labeled spin-tubes. Each one has a set of descriptions in
terms of generalized areas and volumes because of its decompositions into
n-punctured spheres. Each history may be thought of as consisting of a
succession of quantum 3-geometries. Besides the representations and inter-
twiners, there is another structure defined on the histories: each history M
is a causal set, whose structure is determined as follows.
Each history M is also a set of genus-6 elementary spin-tubes Pi . Each
Pi is divided into two parts Li and L′i corresponding to the elementary
local regions that were removed and inserted. The 4-punctured spheres in
Li are the past set of Pi . The remaining 4-punctured spheres, which are
in the complement L′i are the future set of Pi . Now, consider a particular
4-punctured sphere s in the future set L′i in some Pi . Let us assume that s
has been acted on by at least one generalized evolution move Pj for j > i.
Then s also belongs to the past subset Lj of Pj . If now s′ is a 4-punctured
sphere in the future subset L′j of Pj , we will say that s′ is to the immediate
causal future of s.
Now, consider a sequence of r 4-punctured spheres si , i = 1, ..., r, such
that for each si , i < r either i) si+1 is to the immediate causal future of si ,
or ii) there is some |Ii ≡ |SI , ΨI i ∈ M such that si and si+1 are both in
the surface SI and si ∩ si+1 6= 0. (This, is either each 4-punctured sphere in
the sequence is to the immediate causal future of its predecessor, or it and
its predecessor overlap in a single surface associated with a state |Ii in the
history.) When this is the case we will say that sr is to the causal future of
s1 , sr > s1 .
It is clear that the relation > is transitive and that given two 4-punctured
spheres s1 and s2 , s1 > s2 > s1 is never the case. Thus, the 4-punctured
spheres in each history M constitute a causal set, which is defined in [33] to
be a partially ordered set with no closed causal loops which is locally finite.
The latter means that given any s1 and s2 the set contained in the causal
past of s2 and the future of s1 is finite. As argued in [33, 34] a discrete
set that has on it a causal structure is a candidate for a discrete model of
spacetime.
The 4-punctured spheres of a history M, defined by the evolution moves
that construct it, are then the events of M. We will call the set of events
E. By construction, E is a causal set. It differs from the causal set of Sorkin
and collaborators[33] in that there is additional structure, associated to a
notion of space.
Each history M may be foliated by a number of sets of causally unrelated
events of M that we will call the spacelike slices Γ. A spacelike slice of M

21
is a subset {sa } of E glued together according to the following rules:
1. No two sa in Γ may be causally related.

2. Two events sa and sb in Γ may be glued together if there is a state

|Ii ∈ M in which they are glued along some circle.

3. The set Γ is maximal in that no sa may be added to it without violating

these conditions.
Associated with Γ is a state |Γi ∈ HGq given by |SΓ , j, µa i. Here the in-
tertwiners µa are fixed because sa ∈ Γ are given. Similarly, each circle cab
along which two adjacent 4-punctured spheres sa and sb are glued is in fact
a circle labeled by a fixed representation j. Hence the labels on the state
|Γi = |SΓ , j, µa i are uniquely determined by the history M.
The N original states {|1i, |2i, ..., |Ii, ..., |N i} are spacelike slices accord-
ing to this definition. But there are many more sequences which may
be constructed given the history M = {|1i, |2i, ..., |Ii, ..., |N i} that have
|1i as the initial state and N i as final. We call the set of such states
WM . One may in general select other sequences of elements of WM , e.g.
M′ = {|1i, |2′ i, ..., |I ′ i, ..., |N i}, that have the property that every event in
E is a 4-punctured sphere in a decomposition of at least one |I ′ i. As far as
the local geometry and causal structure are concerned these are equivalent
descriptions of the history M. Thus, this quantum theory has a discrete
analogue of multi-fingered time.
Thus, a discrete history M combines discrete analogues of both the
canonical picture of quantum gravity and the spacetime causal structure. It
is the marriage of both kinds of structure within a completely discrete ap-
proach to quantum gravity that we believe gives this approach its particular
power.

7.4 Connection with spin foam and membranes

In a number of recent papers, [39, 40, 41, 42, 43] a concept Baez calls “spin-
foam” has been introduced. These are networks of colored 2-surfaces embed-
ded in a four-dimensional spacetime whose slices by three-manifolds are spin
networks. Gupta[47] has shown that the spin foam can be given a Lorentzian
formulation by the addition of a causal structure and that that formulations
is in a particular sense dual to the formulation of [23]. There is an analogous
spacetime foam structure associated with the histories M, although it has
not been so far investigated. It can be constructed by noting that each of

22
the evolution moves may be seen as three-dimensional cobordisms between
the two surfaces Lρ and L′ρ (See Fig. 6). The resulting three-manifolds
may be joined together to construct a three-dimensional timelike combina-
torial manifold associated to each history M. This is a non-perturbative,
background independent membrane.

8 Coarse graining, entropy and the holographic

hypothesis
Before closing we make some comments about coarse graining and entropy
that will enable us to comment also on the relationship of our proposal to
the holographic hypothesis[35, 36] and the Bekenstein bound[37].
The basic idea is that in addition to the fine grained observables discussed
previously there are coarse grained observables that describe statistical in-
formation about the states defined in section 2. There are two kinds of
course grainings which are relevant. In the first we retain information about
the topology of the surface S while in the second we retain only information
that can be measured by observers at the boundaries of the regions.
Before describing these we may note that the existence of coarse grained
observables in itself means that the theory genuinely has local observables
that are not determined by the values of the coarse grained observables.

8.1 Coarse graining by topology

We can coarse grain the information in a state |S, Ψi by forgetting the in-
formation about the state Ψ ∈ V S and retaining only statistical information
about the surface S. This results in a density matrix which is constructed
by tracing over the representations jα and intertwiners µρI . To each surface
S is then associated a density matrix which is ρS = PS , the projection op-
erator onto V S . There is an entropy associated with this coarse graining.
Associated to each surface S is an entropy S[S] = ln(dimV S ).
As the dynamics changes the topology an entropy change can be asso-
ciated with the evolution operators defined in the last section. This makes
possible a thermodynamic treatment of the evolution, which will be de-
scribed elsewhere.

23
8.2 Coarse graining by regions
Rather than coarse graining by the topology of S we can coarse grain by
splitting space into regions and measuring statistical information about each
region. To do this we must take into account what we learned from our
discussion of geometrical interpretations, which is that as the topology and
geometry of space are defined from the states, the splitting of space into
regions must be defined intrinsically in terms of the states. We then define
a coarse grained quantum geometry as a coarse grained interpretation of a
quantum state |S, {j}, {µ}i. Let us then consider a decomposition of S into
a set of regions Ri along mi circles cγ . Each piece consists of a component
of S we will call Wi . Each Wi is a punctured surface, punctured by the mi
labels jγ on the circles cγ .
To each region we will also associate a punctured S 2 , with mi punctures
with the same labels as the Si . Coarse graining will mean that for each
region Ri we forget the details of the topology of the component Wi . This
means that all observables concerning the region must be representable as
operators in the space of intertwiners on the associated punctured S 2 . There
are then two spaces of intertwiners which are relevant, VjWγ i and VjSγ . Coarse
2

graining consists of replacing a microscopic state, which is a vector in VjWγ i

2
with a density matrix in VjSγ .
In correspondence with the different notions of properties we may define
a coarse grained surface property of the region Wi to be a function of the
labels j1 , ..., jim and a coarse grained bulk property to be an operator in
2
VjSγ . Finally, two regions may share properties when the corresponding
Wi ’s are glued along punctures. Moreover, given a full set of labelings on
2
the punctured surface Wi we have a state in VjSγ by considering the W i as
a framed spin network embedded in the interior of the surface Si in R3 .
A coarse grained description of the quantum geometry is then given by a
2
density matrix in the spaces VjSγ that corresponds to each of the regions Ri .
It corresponds to what observers may measure about the world, assuming
they can only measure on boundaries.

8.3 Connection with the holographic hypothesis and Beken-

stein bound
The possibility of describing coarse grained properties in this way also sug-
gests a formulation of the holographic[36, 35] hypothesis that is entirely
non-perturbative and background independent. This arises in the case that

24
we split the universe into two regions, and assume that we can only make
measurments in one of them.
Let us introduce a splitting of a surface S along a set of p non-intersecting
elements of π 1 [S], which we will call the cγ , γ = 1, ..., p. The two halves
may be called S + and S − ; the cα are in each case their ends. Let us
further consider a basis of states in which there are definite representations
jγ defined on the surfaces.
In the absence of a background manifold we will simply represent the
splitting by a p-punctured S 2 , labeled by the jγ . Each half S ± then has
± ±
on it a space of intertwiners VjSγ . An element VjSγ defines what we will
call a quantum geometry with boundary. Given a quantum geometry, i.e. a
state in a VS , there are many ways to split it into two halves, giving two
quantum geometries with boundaries. The splitting of the world into two
parts constitutes a simple coarse graining of it.
Now consider an observer who lives in one a half, S + , who is for some
reason unable to measure any information about the topology or state of S
in the other half S − . This might, for example, arise if the causal structure
(which we have shown makes sense at this, non-perturbative background
independent level) does not enable him or her to receive any information
from the other half. In this case the observer effectively lives in a quantum
+
geometry with boundary defined by the half VjSγ .
What information can the observer have about the physics of the other
−
half VjSγ ? All they can measure is correlations between measurements they
may make at the p ends. This means that the possible states they may dis-
tinguish by their measurements are given exactly by the space of conformal
blocks on the p-punctured S 2 associated with their boundary. This is the
2
space VjSγ which we described before.
To summarize, the following may be considered a non-perturbative for-
mulation of the holographic hypothesis: When an observer is unable to mea-
sure information corresponding to the interior of a region of a quantum ge-
ometry, because of the presence of a causal horizon, or for any other reason,
the information accessible to them by measuring observables at the bound-
2
ary of that region is represented by a finite dimensional space of states VjSγ
for some p-punctured S 2 .
This has several further consequences. First, in the SU (2) case it is
known that[8]
 2
 c A[jγ ]
ln dim[VjSγ ] ≤ 2 (27)
4 lP lanck

25
 √
for large numbers of punctures, where c = 8ln(2)/ 3. Here A[jγ ] is the area
q
P
operator of quantum general relativity [11, 2] with eigenvalues γ lP2 lanck jγ (jγ + 1).
Thus, the Bekenstein bound[37] is automatically satisfied 9 .
In the case of a general Gq we do not know which observable corresponds
to the area. It may be any surface property, which means it must be an
additive function A of the casimers of Gq . The Bekenstein bound gives us a
constraint on that definition, which is that
 2

A[jγ ] < 4lP2 lanck ln dim[VjSγ ] . (28)

We may note that the Bekenstein bound (28), together with certain
other assumptions is, as Jacobson has shown [48], equivalent to the Einstein
equations. Jacobson’s argument in [48] can be interpreted to imply that
any finite theory of quantum gravity that has a classical limit such that a)
the relationship (28) is satisfied on every horizon which exists by virtue of
an observer being accelerated and b) quantum fields behave as conventional
free fields in the limit of low curvatures, then the field equations of general
relativity are true to leading order in curvatures as a consequence of the
ordinary laws of thermodyanics[48]. This suggests that statistical assump-
tions about the dynamics, together with (28) may be sufficient to derive the
classical limit of the theory.

9 Conclusion
The general framework introduced here becomes a theory with two inputs:
a group or algebra Gq , and a choice of the dynamical operators H b ρ that
define the evolution. The main question that must be investigated is how
these operators are to be chosen. Good choices should lead to a theory with
a good continuum limit which reproduces classical general relativity with
matter fields. This is currently being investigated in several directions.
1. The algebra of the tube operators introduced here should be worked
out. It will be interesting to see if there is a set of local operators
that generate the algebra and if they are related to the loop algebra
of quantum gravity [32] in the q → 1 limit.
9
We may note that the constant c is not equal to one. This is not surprising given that
the quantity lP lanck in the area formula is given by the bare Newton’s constant. Unless
the theory has a continuum limit the macroscopic, renormalized Newton’s constant which
plays a role in black hole thermodynamics cannot be defined. This result suggests then
predicts that in those cases Gren = cGbare .

26
2. It appears possible to choose the evolution operator cbρ to agree with
the dynamics generated by the lorentzian hamiltonian constraint of
Thiemann[16]. A path integral representation of Thiemann’s lorentzian
constraint, along the lines of [40], may then be possible.
It seems that the evolution generated by Thiemann’s constraints is
ultralocal[17, 18]. However we may note that the evolution generated
by the 1 → 4 and 4 → 1 moves are ultralocal in the sense that they
do not lead to long-range propagation. As suggested already in the
Euclidean context in [40] it follows that the other moves are necessary
in order to have long-range propagation.

3. More generally, the relation of the causal theory to the euclidean path
integral approaches [39, 40, 41, 42, 43] should be investigated. In this
direction, Gupta in [47] has formulated a causal spin foam.

4. All of the above involve so far only the SU (2) spin networks. The
extension to other groups is important. The SO(8) case is of special
interest because of its connection to supersymmetry and triality. It
is currently under investigation with Asok. The general case of a
supergroup should be investigated.

5. Two connections with string theory have been investigated. In [21], we

take Gq to be the projective group of the circle. Its representations are
parametrised by relatively prime pairs of integers (p, q). The states in
this case turn out to be combinatorial (p, q) string networks [49] whose
dynamics is a simple case of section 7. Second, in [22] perturbations
of the SU (2) theory have been studied which are given by a (1+1)-
dimensional system with couplings determined by cbρ . When the full
theory has a good continuum limit, the action for the 1 + 1 system is
given to leading order by the Nambu action of bosonic string theory.
An argument may be given that if, for some choice of Gq and cbρ the
induced 1 + 1 dimensional theory is a consistent perturbative string
theory then the continuum limit of the non-perturbative theory exists.

6. In [24] we argued that the existence of a continuum limit can be seen

as a critical phenomenon which is analogous to directed percolation.
To investigate this we have invented a set of simple models that have
dynamical causal structure of the type described here[50]. Further,
these models are discrete dynamical systems since evolution proceeds
by discrete local steps. This leads to proposals for the evaluation of the

27
 path integrals proposed here which are discussed in [51]. Finally, given
the remarks in the previous section, one may use statistical mechanics
to make general statements about the evolution of the states based on
the entropy S[S].

ACKNOWLEDGEMENTS
We are grateful to Chris Isham for suggestions on constructing the space of
states of the theory and how to accommodate the action of the evolution
operators. While writing this paper we consulted papers of Louis Crane
on conformal field theory[26] and were struck by the resonance of the kine-
matical framework of section 2 with ideas sketched there. We would like to
thank him and John Baez, Sameer Gupta, and Carlo Rovelli for comments
on a draft of this paper. We also thank them and Stuart Kauffman and
Mike Reisenberger for conversations and encouragement. This work was
supported by NSF grant PHY-9514240 to The Pennsylvania State Univer-
sity and a NASA grant to the Santa Fe Institute.

References
[1] C. Rovelli and L. Smolin, “Spin networks and quantum gravity” gr-
qc/9505006, Physical Review D 52 (1995) 5743-5759.

[2] C. Rovelli and L. Smolin Discreteness of area and volume in quantum

gravity Nuclear Physics B 442 (1995) 593. Erratum: Nucl. Phys. B 456
(1995) 734.

[3] C. Rovelli,Loop quantum gravity, r-qc/9710008.

[4] L. Smolin, The future of spin networks gr-qc/9702030.

[5] R Penrose: in Quantum theory and beyond ed T Bastin, Cambridge U

Press 1971; in Advances in Twistor Theory, ed. L. P. Hughston and
R. S. Ward, (Pitman,1979) p. 301; in Combinatorial Mathematics and
its Application (ed. D. J. A. Welsh) (Academic Press,1971); Theory of
quantized directions unpublished manuscript.

[6] See, for example, T. Thiemann, QSD V : Quantum Gravity as the

Natural Regulator of Matter Quantum Field Theories , gr-qc/9705019,

28
 Kinematical Hilbert Spaces for Fermionic and Higgs Quantum Field
Theories , gr-qc/9705021.

[7] D. Armand-Ugon, R. Gambini, O. Obregon, J. Pullin, Towards a loop

representation for quantum canonical supergravity, hep-th/9508036,
Nucl.Phys. B460 (1996) 615. See also, T. Jaocobson, New variables
for canonical supergravity Class. Quant. Grav. 5 (1988) 923;H. Kunit-
omo and T. Sano The Ashtekar formulation for canonical N=2 su-
pergravity, Prog. Theor. Phys. suppl. (1993) 31; Takashi Sano and J.
Shiraishi, The Non-perturbative Canonical Quantization of the N=1
Supergravity, Nucl. Phys. B410 (1993) 423, hep-th/9211104; The
Ashtekar Formalism and WKB Wave Functions of N=1,2 Supergrav-
ities, hep-th/9211103; K. Ezawa, ASHTEKAR’S FORMULATION
FOR N=1, N=2 SUPERGRAVITIES AS CONSTRAINED BF THEO-
RIES, Prog.Theor.Phys.95:863-882,1996, hep-th/9511047.T. Kadoyoshi
and S. Nojiri, N=3 AND N=4 TWO FORM SUPERGRAVITIES,
Mod.Phys.Lett.A12:1165-1174,1997, hep-th/9703149; L. F. Urrutia To-
wards a loop representation of connection theories defined over a super-
lie algebra hep-th/9609010.

[8] L. Smolin, Linking topological quantum field theory and nonperturbative

quantum gravity gr-qc/9505028, J. Math. Phys. 36 (1995) 6417.

[9] S. Major and L. Smolin, Quantum deformation of quantum gravity gr-

qc/9512020, Nucl. Phys. B 473 (1996) 267.

[10] R. Borissov, S. Major and L. Smolin, The geometry of quantum spin

networks gr-qc/9512043, Class. and Quant. Grav. 12 (1996) 3183.

[11] L Smolin: in Quantum Gravity and Cosmology, eds J Pérez-Mercader

et al, World Scientific, Singapore 1992.

[12] R. Loll, Nucl. Phys. B444 (1995) 619; B460 (1996) 143; R. DePietri
and C. Rovelli, Geometry eigenvalues and scalar product from recou-
pling theory in loop quantum gravity, gr-qc/9602023, Phys.Rev. D54
(1996) 2664; Simonetta Frittelli, Luis Lehner, Carlo Rovelli, The com-
plete spectrum of the area from recoupling theory in loop quantum grav-
ity gr-qc/9608043; R. Borissov, Ph.D. thesis, Temple, (1996).

[13] T. Thiemann, A length operator for canonical quantum gravity, gr-

qc/9606092.

29
[14] C Rovelli L Smolin, The physical hamiltonian for non-perturbative
quantum gravity, Phys Rev Lett 72 (1994) 446; L. Smolin, Time, mea-
surement and information loss in quantum cosmology in Directions in
general relativity, volume 2: Papers in honor of Dieter Brill, edited by
B. L. Hu and T. A. Jacobson, Cambridge University Press, 1993.

[15] R. Borissov, Graphical Evolution of Spin Network States, gr-

qc/9606013, Phys.Rev. D55 (1997) 6099-6111.

[16] T. Thiemann, Quantum spin dynamics I, Harvard preprint (1996), gr-

qc/9606089; Quantum spin dynamics II, Harvard preprint (1996), gr-
qc/9606090.

[17] L. Smolin, The classical limit and the form of the hamiltonian constraint
in non-pertubative quantum gravity CGPG preprint, gr-qc/9609034.

[18] R. Gambini, J. Lewandowski, D. Marolf, J. Pullin, On the consistency of

the constraint algebra in spin network quantum gravity, gr-qc/9710018.

[19] C. M. Hull and P. K. Townsend, Unity of superstring dualities, Nucl.

Phys. B348 (1995) 109; E. Witten, String theory in various dimensions,
Nucl. Phys. B443 (1995) 85.

[20] See, for example, P. S. Aspinwall, B. R. Greene, D. R. Morrison Mul-

tiple Mirror Manifolds and Topology Change in String Theory, hep-
th/9301043, Phys. Lett. B303 (1993) 249-259; Calabi-Yau Moduli Space,
Mirror Manifolds and Spacetime Topology Change in String Theory,
hep-th/9309097, Nucl. Phys. B420 (1994) 184-242; B. R. Greene, D. R.
Morrison, A. Strominger, Black Hole Condensation and the Unification
of String Vacua, hep-th/9504145, Nucl. Phys. B451 (1995) 109.

[21] F. Markopoulou and L. Smolin Non-perturbative dynamics of algebraic

(p,q) string networks preprint, December 1997.

[22] L. Smolin Strings as perturbations of evolving spin networks, prepring,

Dec. 1997.

[23] F. Markopoulou, Dual formulation of spin network evolution gr-

qc/9704013, CGPG preprint (1997).

[24] F. Markopoulou and L. Smolin Causal evolution of spin networks gr-

qc/9702025. CGPG preprint (1997), Nuclear Physics B, in press.

30
[25] G. Moore and N. Seiberg, Classical and quantum conformal field theo-
ries Comun. Math. Phys. 123 (1988) 177.

[26] L. Crane, Commun. Math. Phys. 135 (1991) 615; Phys. Lett. B 259
(1991) 243.

[27] E. Verlinde, Fusion rules and modular transformations in 2D confor-

mal field theory Nucl. Phys. B 300 (1988) 360; G. Moore and N. Yu.
Reshetikhin, A comment on quantum group symmetry in conformal field
theory, Nucl. Phys. B 328 (1989) 557;

[28] E. Witten, Quantum field theory and the Jones Polynomial Commun.
Math. Phys. 121 (1989) 351

[29] N. Yu. Reshetikhin and V. G. Turaev, Ribbon graphs and their invari-
ants derived from quantum groups, Commun. Math. Phys. 127 (1990)
1; Invariants of 3-manifolds via link polynomials and quantum groups,
Invent. Math. 103 (1991) 547; V Turaev O Viro: Topology 31 (1992)
865.

[30] L. Kauffman and S. Lins Tempereley-Lieb Recoupling Theory and In-

variants of 3-Manifolds Princeton University Press, 1994, and references
therein.

[31] V. Pachner, Europ. J. Combinatorics, 12 (1991) 129.

[32] C Rovelli L Smolin: Phys Rev Lett 61 (1988) 1155; Nucl Phys B133,
80 (1990).

[33] L. Bombelli, J. Lee, D. Meyer and R. D. Sorkin, Spacetime as a causal

set Phys. Rev. Lett. 59 (1987) 521.

[34] G. ’t Hooft, Quantum gravity: a fundamental problem and some radical

ideas. Cargèse Summer School Lectures 1978. Publ. “Recent Develop-
ments in Gravitation”. Cargèse 1978. Ed. by M. Lévy and S. Deser.
Plenum, New York/London, 323; Quantization of Space and Time
in 3 and in 4 Space-time Dimensions, Lectures held at the NATO
Advanced Study Institute on “Quantum Fields and Quantum Space
Time”, Cargèse, July 22 – August 3, 1996. gr-qc/9608037; The scatter-
ing matrix approach for the quantum black hole: an overview. J. Mod.
Phys. A11 (1996) pp. 4623-4688. gr-qc/9607022.

31
[35] G. ’tHooft, Dimensional reduction in quantum gravity gr-qc/9310006.

[36] L. Susskind, The world as a hologram, hep-th/9409089, J. Math. Phys.

(1995) ; Strings, black holes and Lorentz contractions hep-th/9308139;
Phys. Rev. Lett. 71 (1993) 2367; Phys. Rev. D 49 (1994) 6606; D 50
(1994) 2700; L. Susskind and P. Griffin, Partons and black holes hep-
th/9410306;

[37] J. D. Bekenstein, Lett. Nuovo Cim 11 (1974) 467.

[38] L. Smolin, The Bekenstein bound, topological quantum field theory and
pluralistic quantum cosmology, gr-qc/9508064.

[39] M. Reisenberger, A lattice worldsheet sum for 4-d Euclidean general

relativity , gr-qc/9711052.

[40] M. Reisenberger and C. Rovelli, “Sum over Surfaces” form of Loop

Quantum Gravity, gr-qc/9612035

[41] L. Crane, A Proposal for the Quantum Theory of Gravity gr-

qc/9704057.

[42] J. Barrett and L. Crane, Relativistic spin networks and quantum grav-
ity, gr-qc/9709028 ; L. Crane, On the interpretation of relativistic spin
networks and the balanced state sum, gr-qc/9710108.

[43] J. Baez, Spin foam models, gr-qc/9709052.

[44] A. Barbieri Quantum tetrahedra and simplicial spin networks gr-

qc/9707010.

[45] H. Ooguri, Topological Lattice Models in Four Dimensions, hep-

th/9205090, Mod. Phys. Lett. A7 (1992) 2799-2810.

[46] L. Crane and D. Yetter, On algebraic structures implicit in topological

quantum field theories, Kansas preprint, (1994); in Quantum Topology
(World Scientific, 1993) p. 120; L. Crane and I. B. Frenkel, J. Math.
Phys. 35 (1994) 5136-54;

[47] S. Gupta, preprint in preparation.

[48] T. Jacobson, Thermodynamics of Spacetime: The Einstein Equation of

State, Phys.Rev.Lett. 75 (1995) 1260-1263, gr-qc/9504004.

32
[49] A. Sen String networks, hep-th/9711130.

[50] F. Markopoulou and L. Smolin, Models of evolving spin networks

preprint in preparation.

[51] S. Kauffman and L. Smolin, Evolving spin networks as discrete dynam-

ical systems, preprint in preparation.

[52] See, for example, J. Hartle, Spacetime Quantum Mechanics and the
Quantum Mechanics of Spacetime, gr-qc/9304006; M. Gell-Mann and J.
Hartle, Strong Decoherence, gr-qc/9509054; C. J. Isham, Topos Theory
and Consistent Histories: The Internal Logic of the Set of all Consis-
tent Sets, gr-qc/9607069, Int.J.Theor.Phys. 36 (1997) 785-814; Quan-
tum logic and decohering histories, quant-ph/9506028.

33