The Feynman propagator for spin foam quantum gravity

Daniele Oriti
Department of Applied Mathematics and Theoretical Physics,
University of Cambridge, Wilberforce Road,
Cambridge CB3 0WA, England, EU
e.mail: [email protected]
(Dated: February 7, 2008)
We link the notion causality with the orientation of the 2-complex on which spin foam models
are based. We show that all current spin foam models are orientation-independent, pointing out
the mathematical structure behind this independence. Using the technology of evolution kernels
for quantum fields/particles on Lie groups/homogeneous spaces, we construct a generalised version
of spin foam models, introducing an extra proper time variable and prove that different ranges of
integration for this variable lead to different classes of spin foam models: the usual ones, interpreted
as the quantum gravity analogue of the Hadamard function of QFT or as a covariant definition of
arXiv:gr-qc/0410134v3 19 Apr 2005

the inner product between quantum gravity states; and a new class of causal models, corresponding
to the quantum gravity analogue of the Feynman propagator in QFT, non-trivial function of the
orientation data, and implying a notion of “timeless ordering”.

I. INTRODUCTION man propagator and the Hadamard function, is exactly

the way they encode (or fail to encode) causality restric-
Spin foam models [1, 2] are a new more rigorous im- tions, i.e. the fact that one of their arguments is in the
plementation of the path integral or sum-over-histories causal future of the other, or in other words the order
approach to quantum gravity, in which spacetime is re- between these arguments. This order-dependence can be
placed by a combinatorial 2-complex, and the histories imposed in each of the histories summed over in a clear
of the gravitational field are characterized by data taken way using a proper time formulation [3, 4]: starting from
from the representation theory of the local gauge group of the same proper time dependent expression g(x1 , x2 , T ),
gravity, i.e. the Lorentz group, and no familiar notions of the Hadamard function is obtained by integrating over
metric on differentiable manifolds are used. The main ob- both positive and negative values of the proper time; the
ject that such a sum-over-histories formulation defines is resulting amplitude is then a-causal, and does not reg-
the partition function of the theory, and from this, allow- ister any ordering between its arguments; the Feynman
ing the underlying spacetime 2-complex to have bound- propagator is instead obtained restricting this integra-
aries, quantum amplitudes functions of the boundary tion to positive proper times only, and this ordering is
data, that, when a canonical interpretation is available, precisely what makes it causal. What about quantum
should be interpreted as transition amplitudes between gravity? Here no external time coordinate is available,
quantum gravity states. The partition function in these so no definition of the Feynman propagator in terms of
models, for a given 2-complex σ, has the general struc- time-ordered products is possible; however, as we said,
ture: the crucial property that distinguishes the two Green
XY Y Y functions of the relativistic particle is the symmetry un-
Z(σ) = Af (ρf ) Ae (ρf |e ) Av (ρf |v ), (1) der a change in the orientation of the history connecting
{ρf } f e v the initial and final states and, when a proper time for-
mulation is available, this is what the signs of the proper
where the amplitudes for faces f , edges e and vertices time amounts to. Therefore we see a way out of the men-
v of the 2-complex are all functions of the representa- tioned difficulty in the quantum gravity case. Indeed it
tions ρf of the Lorentz group associated to the faces of has been proposed [4] that a similar situation as in the
the complex. This work deals, in the spin foam setting, particle case arises in quantum gravity when formulated
with two related issues common, but not restricted, to in the path integral formalism; the distinction between
any sum-over-histories formulation of quantum gravity: the different transition amplitudes is again obtained by
1) given a definition of transition amplitudes for quan- choosing different range of integration in the proper time
tum gravity states, which kind of transition amplitude formulation of the theory; using a canonical decomposi-
does it provide? 2) is it possible to identify a notion of tion of the metric variables, the role of proper time is
causality and to implement some causality conditions ex- played by the lapse function; again an unrestricted range
plicitly in quantum gravity, in spite of the necessary (be- of integration leads to the (quantum gravity analogue of
cause of diffeomorphism invariance) absence of any time the) Hadamard function, while a restriction to positive
parameter in the quantities one wants to compute? The lapses leads to the (analogue of the) Feynman propagator
two issues are related because it is known [3] that in the [4]. The problem we address in this work is how to con-
case of free relativistic particle and quantum field the- struct the analogue quantities in the spin foam context,
ory the main difference between the different transition i.e. in a purely combinatorial, algebraic and group the-
amplitudes one can construct, in particular the Feyn-
 2

oretic way, in absence of any smooth manifold structure condition on the values of the orientation variables, that
and any metric field. basically follows from Stokes’s theorem, is then [7, 8]:
∀v ǫf |v = αe1 |v αe2 |v µv , where e1 and e2 label the two
links that belong to the boundary of the face f and touch
II. CAUSALITY AS ORIENTATION the vertex v. This orientation structure is the quantum
seed of the classical causal structure, in a Lorentzian con-
text.
In spin foam models spacetime is replaced by a combi-
natorial 2-complex, i.e. by a collection of vertices, links
and faces, with extra data assigning more geometric in-
III. ORIENTATION-INDEPENDENCE OF
formation to it. What can be the analogue of causal rela- CURRENT SPIN FOAM MODELS
tions in such a context? Consider just the first layer of the
spin foam 2-complex, i.e. only vertices and links connect-
ing them, therefore forming a graph. If we add to it ori- We have now to see how the orientation data intro-
entation data, i.e. arrows on the links, we obtain an ori- duced in the previous section enter in current spin foam
ented (or directed) graph, a set of oriented links connect- models. Our analysis holds for the Ponzano-Regge [6]
ing a set of vertices. Now the vertices can be interpreted and all Barrett-Crane-like models [1, 2] based on ordi-
as a set of fundamental spacetime events and the oriented nary Lie groups and homogeneous spaces, with no ex-
links are then the causal relations between them. We can plicit quantum group structure. Physically these can
assign an orientation variables αe|v = ±1 to each link, be interpreted as models of quantum gravity without
with respect to each vertex v it connects, and µ = ±1 cosmological constant. We use a ’first order’ formula-
to each vertex. The spacetime interpretation of these tion of spin foam models [7], in terms of both group
variables is that of indicating whether the vertex is a fu- variables (or variables with values in an homogeneous
ture pointing or past pointing contribution to the overall space) and representation variables. Let us anticipate
spacetime diagram, and whether the link is ingoing or the result before justifying it: all current spin foam mod-
outgoing with respect to each vertex. A consistency con- els do not depend, in their amplitudes, on the orien-
dition for the assignment of orientation data to the graph tation of the underlying 2-complex. The way this is
is that when a link e connects two vertices it has the achieved is quite simple: in the expression for the am-
opposite orientation in the two: αe|v1 µv1 = −αe|v2 µv2 . plitudes for spin foams the terms that can be understood
We also assign an extra orientation variable to each face as contributions from opposite orientations are summed
ǫf = ±1. This structure is basically that of a poset or simmetrically thus erasing the dependence on the ori-
causal set [5], but the set of vertices-events endowed with entation data. We deal here explicitely only with the
the ordering relation now fails to satisfy in general any n-dimensional models for n ≥ 4 based on the homo-
reflexivity condition, in other words, our causal relations geneous spaces SO(n − 1, 1)/SO(n − 1) ≃ H n−1 and
allow for closed timelike loops. Also, while the combina- SO(n)/SO(n − 1) ≃ S n−1 for brevity, but the analysis
torial structure above is general, its causal interpretation can be extended with the same results for all the other
makes sense only in a Lorentzian context; therefore the spin foam models [9]. The Barrett-Crane-like models in
issue we will be confronting in the following is the gen- 4 and higher dimensions [1] take the form:
eral one of constructing spin foam models that take into   !
account appropriately the orientation of the underlying YZ YYZ Y
2-complex, i.e. of orientation-dependent transition am- Z= dρf  dxe Af (ρf )
f v e∈v He f
plitudes for quantum gravity. In a Lorentzian context
these will have the interpretation of causal amplitudes
Y Y
Ae (ρf ∈e ) Av (xe∈v , ρf ∈v )
or of quantum gravity analogues of the Feynman prop- e v
agator. The 2-complexes used in spin foam models are
not generic: they are topologically dual to simplicial n- where of course the precise combinatorics varies accord-
dimensional complexes: to each vertex corresponds a n- ing to the dimension, but in any case: ρf are the unitary
simplex, to each link a (n-1)-dimensional simplex, to each irreps of the local gauge group of gravity (SO(n − 1, 1)
face an (n-2)-dimensional simplex. The orientation data or SO(n)), and these (class I) unitary representations are
we assigned have then a clear geometric interpretation labelled by either a half-integer j in the Riemannian case,
in this simplicial picture: the µv variable for a vertex or by a positive real parameter ρ in the Lorentzian, He
takes the values ±1 according to whether the n-simplex is the homogeneous space to which the vectors xe , inter-
dual to it is isomorphic to a n-simplex in Minkowski (Eu- preted as normals to (n-1)-simplices in each n-simplex,
clidean) space or the isomorphism holds for the opposite belong. The vertex amplitudes factorize in terms of face
orientation; the variable αe|v = ±1 indicate whether the contributions, and it is at this level that the information
normal to the (n-1)-simplex dual to the link e is ingoing about the orientation of the 2-complex is erased:
or outgoing with respect to the n-simplex dual to v, and Y
the variables ǫf also characterize the orientation of the Av (xe∈v , ρf ∈v ) = Af ∈v (θf , ρf ) =
(n-2)-simplex dual to the face f . A second consistency f ∈v
 3
Y 
= Wfµv =+1 (ϑf , ρf ) + Wfµv =−1 (ϑf , ρf ) choice of values for the orientation data, dropping the
f ∈v sum over µv or ǫf . Hovewer, we will shortly see that a
much more natural and elegant construction exists. Also,
where the amplitudes depend on the xe vectors only the expressions 2 are reminescent of the decomposition of
through the invariant distances θf = cosh(h)−1 (xe1∈f · Hadamard functions into Wightman functions. The new
xe2∈f ), and in the last step we have traded the orienta- construction we are proposing shows that this analogy is
tion data on the faces for those on the vertices by us- indeed exact.
ing the second consistency relation, indicating as ϑf the
oriented angle αe1 αe2 θf . This orientation independent
structure is pretty general (independent on signature or IV. SPIN NETWORK FEYNMANOLOGY AND
dimension), and such are the mathematics behind this PARTICLES ON LIE GROUPS/HOMOGENEOUS
and the orientation d ependent functions W. In fact the SPACES
amplitudes Af ∈v are constructed out of representation
ρ
functions of the relevant Lie group, Af ∈v = D00f (θf ), i.e. The amplitudes assigned to spin foam faces, edges
they are given by zonal spherical functions. In turn all and vertices in the Barrett-Crane-type models is given
these representation functions can be uniquely expressed by the evaluation of simple spin networks and was de-
in terms of sums of so-called “representation functions scribed in [13] in analogy to the evaluation of Feynman
ρ
of the 2nd kind”Eklf (g) [10], solutions of the same set diagrams: assign a variable valued in the relevant ho-
of equations [11] with different boundary conditions [11], mogeneous space to each vertex of the given spin net-
ρ
and it is precisely these representation functions of the work, assign a zonal spherical function D00f (θf ) to each
2nd kind that originate the orientation-dependent am- line and sum over all the possible values of the vari-
plitudes Wf ∈v [9]. The relation between representation ables on the vertices to get the final amplitude. In
functions of the 1st and 2nd kind is best understood in this prescription the zonal spherical function is treated
terms of their analytic expressions in terms of associated as a kind of propagator, and this is what it really is,
Legendre functions of the 1st and 2nd kind [9, 11]: as it turns out. Consider a scalar field φ(g) with mass
n−3
m living on the Lie group/homogeneous space G, with
n−1


ρ 2 2 Γ 2
3−n
each point on it labelled by g; consider its free evolution
Af ∈v = D00 (θf ) = Pσ 2 (θ) =
(sin(h)θ)
n−3
2
parametrised by a proper time coordinate s; the equa-
n−3
n−1

n−3
tion of motion in proper time is: (i∂s + ∆)φ(g, s) =
2 2 Γ 2 e−iπ 2 0 with ∆ being the Laplace-Beltrami operator on G.
= n−3 n−3
n−1


×
(sin(h)θ) 2 Γ σ + 2 2 Γ −σ −
cos πσ The dynamics is completely captured by the evolution
h n−3 n−3 i
ρ −ρ
(heat) kernel K(g, g ′ , s) = K(gg−1 , s) [14, 15], in the
× Q−σ−1
2
(θ) − Qσ 2 (θ) = E00 (θ) + E00 (θ) = sense that given the initial condition ψ(g0 , 0), we have:
R
ψ(g, s) = dg0 K(g, g0 , s)ψ(g0 , 0). The various phys-
= Wfµv =+1 (ϑf , ρf ) + Wfµv =−1 (ϑf , ρf )
ical propagators are obtained from the evolution ker-
nel according to how the proper time is integrated out;
where the σ = 2j + (n − 3)/2 in the n-dimensional Rie-
the Hadamard function R +∞is obtained via the expression:
mannian case, and σ = iρ−1/2 in the Lorentzian case. A 2

similar relation can be obtained [9] for the 3-dimensional H(g, g , m ) = −i −∞ dsK(g, g , s)e−im s , while re-
′ 2 ′

Ponzano-Regge case, and for the other Lorentzian mod- stricting the range of integration to positive proper times
els. This orientation independence leads to interpret- only gives the Feynman propagator GF (g, g ′ , m2 ) =
R +∞ 2
ing the current spin foam models as a-causal transition −i 0 dsK(g, g ′ , s)e−im s , where the usual Feynman
amplitudes, i.e. as the quantum gravity analogue of prescription m2 → m2 −iǫ is assumed for convergence. It
the Hadamard function; when a canonical formulation turns out [9] that the functions entering the expressions
is available, they can be equivalently thought of as defin- for the quantum amplitudes of all current spin foam mod-
ing the physical inner product between quantum gravity els correspond to the Hadamard 2-point functions for a
states, invariant under spacetime diffeomorphisms, or as scalar field on the relevant Lie group/homogeneous space
a covariant definition of the matrix elements of the pro- with m2 = −C(ρf ) where C(ρf ) is the Casimir eigenvalue
jector operator onto physical states [12]. Given the uni- of the representation labelling the face of the 2-complex
versal structure outlined above, we are lead to look for (therefore the link of the spin network whose evaluation
a universal way of modifying current spin foam models gives the amplitude for the vertex):
and to a new, again universal, definition of orientation-
dependent or causal spin foam models [9]; these would i q ρ
H(θf , m2 ) = ∆ρf ∈v D00f ∈v (θf ) =
correspond to causal transition amplitudes and to a quan- 2π
Z +∞
tum gravity analogue of the Feynman propagator of field f
= −i dsf |v KM (ϑf , Tv )e+iC(ρf )Tv ,
theory. The brute force procedure is obvious and was −∞
performed in the 4-dimensional Lorentzian case in [7]:
simply restrict the spin foam amplitudes switching from with the “oriented proper time”being T = µv s, ∆ρ is
D functions to E functions corresponding to a consistent the dimension of the representation ρ in the Rieman-
 4

nian case, or in the Lorentzian case (where the uni- In this last case, a regularization prescription for conver-
tary representations are infinite dimensional) the contri- gence is implicit for each variable (so that the expression
bution of the representation to the Plancherel measure, has to be understood in the complex domain): ρf →
and the Casimir eigenvalues are C(ρ) = 2j(2j + n − 2) ρf + iǫ, ϑf → ϑf + iδ. The explicit form of the evolution
with j half-integer in the Riemannian case, and C(ρ) = kernel K differs of course in the different models and af-
+ρ2 + (n − 2/2)2 with ρ positive real in the Lorentzian fects the exact form of the amplitudes [9], that however
case based on the timelike hyperboloid. Again a similar all share the general structure here presented[25]. Let
formula holds for the Ponzano-Regge models and for the us discuss some properties of these new kind of transition
models based on the (n-1)-dimensional DeSitter space. amplitudes. First, all these models can be recast in the
It is clear that the result is independent of the value of form of quantum causal histories models [7, 17]; therefore
the various orientation data. On the other hand simply they define quantum amplitudes for causal sets, if the un-
imposing the above restriction in the proper time integra- derlying oriented graph does not contain closed timelike
tion (that amount indeed to a causality restriction for the loops, and with the additional restriction of the vertices
particle evolution [4]), one obtains the expression for the being (n + 1)-valent in n dimensions. Second, while the
Feynman propagator that one needs to define orientation un-oriented models can be related to the classical Regge
dependent or causal spin foam models, and its expression action only in a asymptotic limit, when their vertex am-
in terms of representation functions of the 2nd kind, con- plitudes result in being proportional to the cosine of it
firms their interpretation as Wightman functions: [8, 18, 19], here the connection with the Regge action is
Z +∞ manifest without any approximation; for example, in the
2 +iC(ρf )Tf |v 4-dimensional case the relevant evolution
i ϑkernel has the
G(ϑf , m , µv ) = −i dsf |v KM (ϑf , Tf |v )e = 1 ϑ
2
−iT
0 form [14]: K(ϑ, T ) = (4πiT )3/2 sinh ϑ e 4T and the
causal vertex amplitude takes the form:
h i
i
p ρf −ρf
= 2π ∆ρf θ(µv )E00 (ϑf ) + θ(−µv )E00 (ϑf ) =
Y Z +∞ 
i
∆ρf [θ(µv )W + (ϑf , ρf ) + θ(−µv )W − (ϑf , ρf )] .
p
= 2π ACv = −2πiρ f ds f |v K(ϑf , T f |v )e iC(ρf )Tf |v
=
f ∈v 0
Similarly for the other spin foam models [9]. This is  
a non-trivial function of the orientation data, with the Y 1 P
= −  ei f ∈v µv ρf ϑf
usual “time ordering”being replaced by a “timeless or- ρf sinh ϑf
dering”! f

and therefore the product over vertex amplitudes in the

spin foam model gives
V. A PROPER TIME FORMALISM AND THE  
FEYNMAN PROPAGATOR FOR SPIN FOAM Y YY 1 P P
 ei f ∈v ρf v µv ϑf ∈v =
QUANTUM GRAVITY ACv =

v
ρ sinh ϑf
v f ∈v f
Taking seriously this particle analogy and the associ-  
ated spin network feynmanology, and having also in mind
YY 1  eiSR (ρf ,ϑf ) ,
=
the group field theory approach [1], we can generalise v
ρ f sinh ϑf
f ∈v
the current formulation of spin foam models to include
a proper time variable, and obtaining a general expres- i.e. the exponential of the Regge action (in first or-
sion from which both orientation independent and causal der formalism [7]) for simplicial gravity, with an addi-
models can be derived, simply changing the integration tional contribution to the overall measure, as one would
contour in proper time. The expression for the models
dealt with above looks as follows:
YYZ YXYYZ
Z= dsf |v dxe A(ρf , xe , Tf |v ) [25] The technology related to quantum particles on homogeneous
C He spaces could have been used to generalise and then modify also
v f |v f ρf v e|v
Q Q the face and edge amplitudes. The reason why we have not done
A(ρf , xe , Tf |v ) = f Af (ρf ) e Ae (ρf |e ) so is twofold: on the one hand the usual interpretation of these
Q Q 
−1  contributions to the spin foam models is that of a contribution
K(ϑf |v , Tf |v )eiC(ρf )Tf |v
p
v f |v −2πi ∆ρf . to the overall measure, therefore the implementation of causality
may be needed only at the vertex level, that is instead supposed
to encode the dynamics of the theory; on the other hand, the
A similar formula holds for all the other models. This form of the edge amplitudes in some version of the 4-d Barrett-
generalised expression encompasses both the usual un- Crane model is understood as arising directly from the form of
oriented models and the new causal ones; indeed the first boundary spin network states [16], and we prefer, at this stage, to
are obtained by choosing the extended range of integra- keep that structure without modification. We believe, however,
that alternative formulations of the models and possible defini-
tion C = (−∞, +∞) for the T variable (and this erases tions of modified spin network states, maybe in order to induce
the dependence on µv and αe , while the quantum grav- an orientatation-dependence in their structure, deserve further
ity Feynman propagator is obtained with C = (0, +∞). analysis.
 5

expect from a sum-over-histories formulation of quan- inner product between quantum gravity states; the other
tum gravity based on a simplicial discretization. The is a new class of models and corresponds to the quantum
same is easily shown in the 4d Riemannian case and in gravity analogue of the Feynman propagator in QFT, i.e.
the 3-dimensional case, while the proof for the higher a causal transition amplitude, non-trivial function of the
dimensional models is less straightforward [9]. Third, orientation data, that implies a notion of “timeless or-
the causal/oriented models seem to solve the problem, dering”; 5) we have shown how the causal model is man-
present in all BF-type formulations of quantum grav- ifestly related to simplicial gravity in the 4d Lorentzian
ity [20, 21], of different isomorphic sectors related by a case. All these results hold true in full generality, for
change of orientation all summed over in the path inte- the type of spin foam models considered, i.e. regardless
gral quantization and thus interfering, leading to a dis- of the spacetime dimension and signature. These results
crepancy with the straightforward quantization of GR, open quite a few lines of possible further research. We list
and affecting the computation of geometric quantitites some of them: 1) a group field theory formulation of gen-
in spin foam models [20]; in the causal models it seems eralised and oriented models (work on this is in progress);
instead that a restriction to the GR sector is achieved. 2) an investigation of the possibility of defining a notion
of “positive and negative energy”sectors, also at the level
of spin network states, based on orientation/causal prop-
VI. CONCLUSIONS erties, in a timeless framework; 3) the study of the role
and significance of the proper time formalism, and of the
Let us summarise our results: 1) we have linked the possible link between this group theoretic proper time
notion of causality in a Lorentzian context with the ori- parameter and the lapse function of canonical quantum
entation of the 2-complex on which spin foam models gravity; 4) an analysis of the relationship between the
are based, and have identified the relevant data charac- spin foam approach and other approaches to quantum
terizing this orientation; 2) we have shown that all cur- gravity, starting from the new causal models, that seem
rent spin foam models are orientation-independent; 3) to be the at the point of convergence of simplicial quan-
using the technology of evolution kernels for quantum tum gravity [22], dynamical triangulations (when a sum
fields/particles on Lie groups/homogeneous spaces, we over 2-complexes is implemented) [23] and causal sets,
have constructed a generalised version of spin foam mod- in addition to canonical loop quantum gravity [24]; 5)
els, introducing an extra variable with the interpretation the new models can be interpreted as defining the ma-
of proper time; 4) we have proven that different ranges trix elements of an “evolution operator”, the basis for
of integration for this proper time variable lead to differ- any “scattering calculation”, whose property could be
ent classes of spin foam models: one corresponds to the studied, also in order to understand whether a notion
usual ones, interpreted as the quantum gravity analogue of unitary evolution is feasible in Quantum Gravity, in
of the Hadamard function of QFT or, equivalenty, in a absence of an external time coordinate, but with a clear
canonical interpretation, as a covariant definition of the notion of causality.

[1] D. Oriti, Rept. Prog. Phys. 64, 1489 (2001), [13] L. Freidel, K. Krasnov, J. Math. Phys. 41, 1681 (2000),
gr-qc/0106091; hep-th/9903192;
[2] A. Perez, Class. Quant. Grav. 20, R43 (2003), [14] R. Camporesi, Phys. Rept. 196, 1 (1990);
gr-qc/0301113; [15] N. Krausz, M.S. Marinov, J. Math. Phys. 41, 5180 (2000),
[3] J.J. Halliwell, M.E. Ortiz, Phys. Rev. D48, 748 (1993), quant-ph/9709050;
gr-qc/9211004; [16] D. Oriti, Phys. Lett. B 532, 363 (2002), gr-qc/0201077;
[4] C. Teitelboim, Phys. Rev. D 25, 3159 (1982) [17] F. Markopoulou, Class. Quant. Grav. 17, 2059 (2000),
[5] R. Sorkin, gr-qc/0309009; hep-th/9904009;
[6] L. Freidel, Nucl. Phys. Proc. Suppl. 88, 237 (2000), [18] L. Freidel, D. Louapre, Class. Quant. Grav. 20, 1267
gr-qc/0102098; (2003), hep-th/0209134;
[7] E.R. Livine and D. Oriti. , Nucl. Phys. B663, 231 (2003); [19] J. C. Baez, J. D. Christensen, G. Egan, Class. Quant.
gr-qc/0210064; Grav. 19, 6489 (2002), gr-qc/0208010;
[8] J. W. Barrett, C. M. Steele, Class. Quant. Grav. 20, 1341 [20] L. Freidel, K. Krasnov, Class. Quant. Grav. 16, 351
(2003), gr-qc/0209023; (1999), hep-th/9804185;
[9] D. Oriti, in preparation; [21] R. De Pietri, L. Freidel, Class. Quant. Grav. 16, 2187
[10] W. Ruhl, The Lorentz Group and Harmonic Analysis, (1999), gr-qc/9804071;
W.A. Benjamin Inc. (1970); [22] R. M. Williams, Int. J. Mod. Phys. B 6, 2097 (1992),
[11] N.Y. Vilenkin, A.U. Klimyk, Representations of Lie [23] J. Ambjorn, J. Jurkiewicz, R. Loll, Phys. Rev. Lett. 85,
Groups and Special Functions, Vol.2, Dordrecht: Kluwer 924 (2000), hep-th/0002050;
(1993); [24] C. Rovelli, Quantum Gravity, Cambridge University
[12] C. Rovelli, Phys. Rev. D59, 104015 (1999), Press (2004).
gr-qc/9806121;