ELSEVIER Nuclear Physics B 486 (1997) 390-412

Alternative actions for quantum gravity and

intrinsic rigidity of space-time
J. Ambjorn a, G.K. Savvidy a'l, K.G. Savvidy b
a The Niels Bohr Institute, Blegdamsvej 17, DK-2100 Copenhagen O, Denmark
b Princeton University, Department of Physics, P.O. Box 708, Princeton, NJ 08544, USA

Received 27 June 1996; revised 10 October 1996; accepted 20 November 1996

Abstract
We generalize the Regge action of simplicial quantum gravity by ascribing deficit angles to the
vertices of four-dimensional simplicial manifolds. The new terms suppress vertices with deficit
angles different from zero and introduce in this way so-called intrinsic rigidity in simplicial
quantum gravity. The concept of generalized deficit angles appear in a natural way in the Steiner-
Weyl expansion formula for parallel manifolds and is related to higher order curvature terms. We
discuss the concept of rigidity in quantum gravity and its relation to the so-called goni-hedric
principle. This principle allows us to find a large class of integral invariants defined on simplicial
manifolds of various dimensions. These invariants are natural candidates for discretized actions
for higher dimensional membranes.

PACS: 02.10.Rn; 02.40.Ky; 04.20.Cv; 04.60.+n

Keywords: Simplicial gravity; Integral invariants; Rigidity of space-time; Membranes

1. Introduction

String theory and theories beyond string theory have increased the interest for quantum
gravity and physics at the Planck scale. One problem in these theories is the lack o f a
non-perturbative definition. Although string duality seems to open tantalizing possibilities
for extracting non-perturbative physics from different perturbative sectors of the theory,
it is still not clear that we will be able to address the full physical content at the
Planck scale without a non-perturbative definition of string theory or its generalization.

t Permanent address: Yerevan Physics Institute, 375036 Yerevan, Armenia.

0550-3213/97/$17.00 Copyright (~) 1997 Elsevier Science B.V. All rights reserved
PII S0550-32 13 (96)00660-8
 J. AmbjCrn et al./Nuclear Physics B 486 (1997) 390-412 391

In particular, it is not clear how much we will be able to say about the aspects related
to the quantum gravity sector of the theory.
It is therefore possible that quantum gravity, or at least important aspects of quantum
gravity, may be described entirely within the framework of a non-perturbative field
theory. A first step in this direction is a non-perturbative definition of the field theory
we will denote quantum gravity. This is a non-trivial task because the continuum theory
has to be invariant under reparametrizations. Lattice gauge theory is an example of a
successful non-perturbative regularization of a quantum field theory with a continuous
internal symmetry. The regularization breaks Euclidean invariance (which is restored
in the scaling limit), but maintains from a certain point of view the concept of local
internal invariance. In the case of gravity the situation is more difficult since we deal
with local symmetries of space-time itself, and any lattice regularization will break this
symmetry.
In classical gravity a very natural discretization was suggested by Regge [ l] by
the restriction to piecewise linear manifolds, and he showed how the Einstein-Hilbert
action had a natural geometric representation on the class of piecewise linear manifolds
and could be expressed entirely in terms of intrinsic invariants of the piecewise linear
manifolds. In this way one achieved a coordinate free description of this class of
manifolds and their actions, where the dynamical variables were the link length. The
use of Regge calculus as a prescription for quantum gravity is less straightforward,
since there is not a one-to-one correspondence between the piecewise linear metric and
the length assigned to the links (for a discussion, see Ref. [2] and also Ref. [3] and
references therein). However, it is very encouraging that a variant of Regge calculus,
known as dynamical triangulation [4-6], works perfect in two-dimensional quantum
gravity. In this formalism one fixes the link length of the piecewise linear manifold, and
the assignment of a metric depends only on the connectivity of the triangulation. The
summation over triangulations with different connectivity takes the role of integration
over inequivalent metrics and the action assigned to the manifold is calculated by Regge's
prescription for a piecewise linear manifold. In this formalism the scaling limit agrees
with the known continuum results of two-dimensional quantum gravity [7,8], i.e. one
gets a reparametrization invariant theory in the scaling limit.
While it is easy to generalize the definition of the two-dimensional discrete model
to both three [9] and four dimensions [10] (see also Ref. [l l] for an earlier slightly
different formulation), the models can presently only be analyzed by numerical methods
[ 12]. Computer simulations show that the discretized models have two phases: For large
values of the bare gravitational coupling constant the typical universes are crumpled with
very large Hausdorff dimension (dH ~ cx~). For small values of the bare gravitational
coupling constant the universes are elongated with small Hausdorff dimensions (dH
2). However, they are separated by the phase transition point where it might be possible
to define a continuum theory with Hausdorff dimension four. At present we have no
continuum theory of Euclidean gravity in dimensions larger than two with which we can
compare, but is seems natural to hope for universes where the dimension in the infrared
limit is four.
 392 J. AmbjCrnet al./Nuclear Physics B 486 (1997) 390-412

Whether one uses the formalism of dynamical triangulations or the original formalism
of Regge with fixed connectivity and variable link length, it might well be that the
simplest actions which have been used so far at the discretized level are insufficient
in producing a real four-dimensional world in the infrared limit. This motivates the
search for alternative actions which can be defined in a natural way on piecewise linear
manifolds and in particular includes high derivative terms. In this article we would like
to advocate a geometrical way of constructing of actions for quantum gravity which
includes both higher derivative terms and Regge's term and which might provide what
we denote as rigidity of space-time. Generally speaking, we understand by rigidity the
property that the dynamical dimension of the system (e.g. in the form of the Hausdorff
dimension dr4) coincides with the kinematic dimension d

dH = d . (1)

In a continuum framework the kinematic dimension is just the dimension of the manifold
M,t. In simplicial gravity it is the largest dimension of simplexes which enter in the
triangulations.
In this context it is worth to recall the situation in two-dimensional quantum gravity.
In this case there is no non-trivial action since the Einstein-Hilbert action is just a
topological invariant and the cosmological term contains no derivatives of the metric.
Nevertheless, in the path integral we are instructed to integrate over all metrics on the
manifold M2. This is clearly a case where dynamics provides absolutely no rigidity and
the surprising result is that dH = 4. This fractal structure is present at all scales in two-
dimensional quantum gravity [ 13]. The result can be derived starting from simplicial
quantum gravity or using entirely a continuum framework.
In four-dimensional quantum gravity it is not obvious that there is a unique dH
independent of distance scale. We believe (from experiments!) that the large distance
limit of four-dimensional quantum gravity is characterized by d,q = d = 4. In this sense
four-dimensional space-time possesses a remarkable rigidity. That one should not take
this for granted starting out with formal constructions of four-dimensional manifolds is
highlighted by the results in d = 2, and it is also illustrated by the numerical simulations
of simplicial four-dimensional quantum gravity. As already mentioned, the simplest
choice of actions used so far in the simulations will for a generic choice of the bare
gravitational coupling constant lead to typical universes where dH 4: 4. In fact, if we
imitate the discussion in two-dimensional quantum gravity and only ask for dH of a
typical four-dimensional universe in a theory with no non-trivial action (dropping the
Einstein-Hilbert term in the action), the result is clear: d,q ~ ~ . This result is even
more dramatic than in the two-dimensional case.
Non-intuitive as these results might appear, they are in a way desirable results,
since they illustrate that dynamics plays a very important role in any theory of four-
dimensional quantum gravity. They also emphasize that it may be quite delicate to choose
an action in four-dimensional quantum gravity which provides us with a space-time
which has dH = 4 at large distance. The simplest minded approach is to consider terms
which at the discretized level introduce additional rigidity into quantum gravity. Such
 J. Ambjcrn et al./Nuclear Physics B 486 (1997) 390-412 393

terms will by hand introduce short distance rigidity in the theory and from a perturbative
point of view they should be irrelevant at large distances. However, we expect the fixed
point of simplicial quantum gravity to be non-perturbative and perturbative arguments
might not be valid.
In order to introduce the rigidity terms it is a necessity that they are defined in
terms of intrinsic quantities. We now outline how this can be done by a geometrical
method based on Steiner's idea of a parallel manifold [ 15-17]. Let M , - i be a smooth
( n - 1)-dimensional manifold embedded in the n-dimensional flat space E", and let
M~_ 1 denote the so-called parallel manifold defined by displacing each point P the
distance p along the outward normal at P. The hyper-volume of the parallel manifold
M,p-1 can be expanded into a polynomial of the distance p from the original manifold.
The coefficients of p~ in this expansion represent integral invariants / z k ( M , - i ) , k =
0, 1. . . . . n - 1 constructed on the (n - 1)-dimensional manifold Mn-i by means of the
intrinsic and extrinsic curvature [ 16]. For even k they have the form

M._l

where Ra,a2~l,: is the Riemann tensor, k is an even integer satisfying 0 ~< k ~< n - 1,
and

is equal to +1 or - 1 , respectively, if ,~1 . . . Ak is an even or odd permutation of ~'1 • • • ~'k,

respectively; otherwise it is zero. The summation is over all AI . . . . . ak and ~'l . . . . . ~'k,
the indices running from 1 to n - 1. In any dimension/.t2 is just the Einstein-Hilbert
action. In four dimensions/[.t 4 is the Euler characteristic,

~2 = du4 R , ~4 = die4 (R~vae - 4 R ~ + R 2) . (3)

M4 M4

This expansion makes perfect sense for a piecewise linear manifold as well, and the
volume of the parallel manifold has the same form of the polynomial in p and one can
compare these two expansions with each other and thus identify terms

/zk = ~ (volume of face) • (volume of spherical image). (4)

{ ( n - - l - k ) dim faces)

This formula uniquely defines the invariant Izk on a simplicial manifold for all values
of k and can be considered as an extension of the classical result for the integral of the
scalar curvature defined by Regge. By the notation "face" we mean vertex, edge, trian-
gle, tetrahedron and higher dimensional sub-simplexes of Mn-I and the corresponding
"spherical image" is the normal map of these faces to the unit sphere (to be defined
precisely in the next section).
An important observation emerging from the last formula is that for each k the
invariant /xk is the sum of the product of the volume orib..i,,_k of the faces (il .. • i n - k l
 394 J. AmbjCrn et al./Nuclear Physics B 486 (1997) 390-412

and the volume w~k.]i,,_~ of the corresponding normal images or spherical images of
these faces. The volume of the spherical image w~!!.i,,_k plays a very important role
in this expression for /zk because it ensures a continuity property of #k, and it can
be considered as a generalized deficit angle. Indeed, the addition of a new face to the
triangulation Mn-l does not change /zk if w!k~ 11 . . . I n - - k
= 0, i.e. if the face (il " " " in-k) is
flat. This continuity property is not fulfilled if the volume of the face o'i,...i,,_k is not
multiplied by the corresponding generalized deficit angle w~!!.i,,_k. Summarizing, the
formula for txk on a piecewise linear manifold can be written as

['l~k = V"
~ 0"il ...i,,-k . oj!tt k~
...t,,-k " (5)
(i~...i,,_k)

The generalization of the factorization property (4) of the classical integral invariants
in the form of a goni-hedric principle, allows us to construct new integral invariants on
a manifold Mn-1, such that the properties of locality and continuity are maintained for
the new invariants. In particular, the addition of new "fiat vertices" will not change the
invariants. The principle states that one should always multiply the volume of the face
by one of the geometric measures, i.e. length, area, volumes etc., associated with the
corresponding spherical image,

Z (volume of face) • (geometric measure on spherical image). (6)

{faces)

In contrast to the classical invariants/.~k we do not yet know to what type of continuum
expressions they correspond, if any, if we choose a geometric measure different from
the one used in (4), but they have the right to be considered independently as a devise
to control the geometry in a regularized version of quantum gravity.
Since the spherical images co(k) are defined by the embedding of the manifold into
Euclidean space it is not an easy task to understand why for even k the integral invariants
are nevertheless intrinsic and independent of the embedding, while for odd k they are
extrinsic. The reason is that for even k one can express w ~k) entirely in terms of internal
angles, while for odd k one has to use both internal and external angles. This can be
seen by applying the Gauss-Bonnet theorem to the spherical images [21-23,26] and
expressing w ~k) in terms of angles.
In particular, in four dimensions w ~ ) and w}4~ are the basic building blocks of intrinsic
geometry and both of them can be expressed in terms of internal angles, i.e. the standard
deficit angle at each triangle (ijk) introduced by Regge and the solid deficit angle at
each vertex (i). The first quantity regulates the flatness on the triangles while the second
one regulates the flatness of the vertices. Local rigidity is now expressed as the absence
of vertices and triangles with large deficit angles. In this way we are led to a natural
discrete representation of the gravitational action which contains terms suppressing "non-
flat" triangles and vertices. As mentioned above, such terms might be important in the
infrared regime of simplicial quantum gravity. The Einstein-Hilbert action together with
the higher derivative terms now takes the following form on piecewise linear manifolds:
 J. AmbjCrn et al./Nuclear Physics B 486 (1997) 390-412 395
1
Acti°n=-~E°'iJk"
o)(2 )
ijk ,(2)+ g2 ,~--~/{
+gl E --ijk (2)~ 2
z_~ \0)ijk ) + . . .
(ijk) (ijk) (ijk)

-t.-klE ( 1 -1 O'4)'~2,/72
, J + ~ (l
k2 ~ - 1"-~
2 227O(4'
/i . )~ + .... (7)
(0 (i)

where o'ijk is the area of the triangle (ijk), WOk

.(2) is the deficit angle associated with the
same triangle and the first term is just the Regge action of the piecewise linear manifold.
The terms associated with the coupling constants gi are functions of the deficit angles
0)(2)
L/k and in this way they correspond to higher derivative terms. The quantity 0} 4) is the
total solid angle associated with the vertex i, i.e. j~f4) = E ""ijklm'O(4)where j~(4)ijklmis the
solid angle at the vertex of the four-dimensional tetrahedron (ijklm) and the summation
is over adjacent tetrahedra. The terms associated with the ki's represent functions of the
"solid deficit angles" at each vertex i.
In the next section we shall review the Steiner-Weyl expansion method and derive
the discrete version of the classical invariants. In subsequent sections we shall construct
new extensions of these functionals in two, three, four and higher dimensions and apply
this approach to obtain discretized versions of the higher derivative terms in quantum
gravity. In a final section we discuss the future applications of the formalism.

2. Integral invariants on smooth manifolds

Let M , _ I be a compact orientable hyper-surface embedded in an Euclidean space E n

by a map P H X ~ ( P ) . At a point P of Mn-i there are n - 1 principal curvatures
Ri, i = 1 . . . . . n - 1. If dvn-1 denotes the element of the hyper-volume of Mn-l, the
integral invariants of M , _ l can be defined as a integrals over symmetric functions of
the principal radii of curvature Ri 16], [
tZk(Mn-1) =-- ... dvn-l, (8)

where k = 0, 1 . . . . . n - l and { . . . } denotes the symmetrization. In particular,/-to is the

hyper-volume,/~2 is the Hilbert-Einstein action and g , - I is the degree of the so-called
normal map,

P C M,l-1 ~ n ( P ) E S n - t ,

which maps a point P of the manifold Mn-l into the unit vector n ( P ) normal to M n - l .
Let 12 be a subset of M._1. Then n ( O ) is called the spherical image of 0 . If d0)~-1
denotes the hyper-volume element on S "-1, then

dVn-l = Rl .. • R n - l d 0 ) n - l , (9)

where the product 1 / ( R i . . . Rn-i ) is the Gauss-Kronecker curvature. This is the gener-
alization o f the famous two-dimensional formula and one can use this formula to express
the integral invariants (8) of M n - t as integrals over the spherical image of Mn-1,
396 J. Ambjern et al./Nuclear Physics B 486 (1997) 390-412

= f (R,, ... Ri,,_,_, (10)

where k = 0,1 . . . . . n - 1. It is now easy to see that /zn_~ is the degree of map
n : M , - i H S n-1 defined by the field of normals.
The integral invariants (8) and (10) appear in a natural way in the so-called Steiner-
p
Weyl expansion formula for the hyper-volume of the parallel manifold M , _ 1 [ 15,16].
P
If M , - I is embedded in E n we define a parallel manifold Mn_ 1 as a set of all points at
a distance p from M , - I , i.e. by the map P ~-* X u ( P ) + p . n u ( P ) , where n ( P ) again
denotes the outward normal at P. For p sufficiently small the hyper-volume/z0(M~_ Ip )
of the parallel manifold is equal to

P 1) = f (R1 + p ) . . . ( R n - J
/z0(Mn_ +p)dw,,_j, (11)

simply because the n - 1 principal curvatures for the parallel manifold M ,P_ I are equal
to Ri + p, i = 1,2 . . . . . n - 1. Expanding the product of the integrand we get

/zo(MP_l ) = / z 0 ( M n - l ) + p " txl ( M , - l ) + p2 . tz2( M,_l ) + . . .

+pn-l . #,-I (M,-i), (12)

thus generating the whole sequence of integral invariants which we discussed above.
It can be shown that the coefficients/x2k (M,,- l ) to the even powers of p in the expan-
sion (12) are independent of the embedding [ 16], while the coefficients/z2k+l (M,,-1)
to the odd powers of p refer explicitly to the extrinsic geometry.

3. Integral invariants on piecewise linear manifolds

The idea of a parallel manifold allows us to define the discrete versions of the integral
invariants (8) and (10) for piecewise linear manifolds and in this way generalize the
work of Regge.
First we consider two-dimensional surfaces [17]. Let D3 denote a three-dimensional
piecewise linear manifold in E 3 with a connected boundary. The boundary of D3 will
be a piecewise linear surface M2 embedded in E 3. We define angles between edges
(one simplexes) (ij) and (ik) as flij;ik and between triangles (two simplexes) (ijl) and
(ijk) as o~ot;ijk. These angles completely define the internal and external geometry of a
surface. It is easier to use the shorthand notation

/3~j;ik ~/3i for internal angles, (13)

O'ijl;i.jk ~ O{~j for external angles (14)

and if needed one can recover the whole notation, see Fig. 1.
Note that the parallel surface M e is one time differentiable even if M2 is only piecewise
linear. The area of M e contains separate parts which we shall compute. The first part,
S(M2), is equal to the area of the original surface M2, i.e. it originates from parallel
 J. Ambjcrn et al./Nuclear Physics B 486 (1997) 390-412 397

l J

k
Fig. 1. The spherical image of the vertex (i). The internal angle between edges are denoted as /~i and the
external angles between triangles as aij.

displacement of the triangles (ikj) constituting M2 the distances p . n((ijk)), where

n((ijk)) denotes the outward normal of triangle n((ijk)) in the orientable triangulation
of M2,

S(M2) = Z °'ijk " 1, (15)

(ijk)
where o'ijk is the area of the triangle (ijk) and the summation is extended over all
triangles of the surface M2. The second part of the area, pA(M2), of the parallel
surface is equal the sum of the areas of the cylinders surrounding the edges [ 17], i.e.
proportional to the displacement p, and we can write

p . A(M2) = p. Z "hiJ " (77"- Olij), (16)

(ij)
where aij is the length of the edge (ij) and aij is the dihedral angle at the edge (ij) and
the summation is over all edges. The third part of the area of M e, p2x(M2) is equal to
the sum of the areas of the spherical polygons which surround the vertices of M2, i.e.
proportional to p2, and we can write

p2 .,¥( M2) =192 . ~'~1 • (27"l'-- y;j1~i) , (17)

(i)

where Xi = 2"n'-y~ Hi is the area of the spherical polygon on a unit sphere corresponding
to a vertex (i), usually called a deficit angle, and the summation is over all vertices of Mz.
Comparing these quantities with the Steiner expansion (12) for smooth two-dimen-
sional surfaces
398 J. AmbjCrn et al./Nuclear Physics B 486 (1997) 390-412

S(Mt2) - ~ / m | R 2 d o . ) - [ - p ' f ( m l -[-R2)do)-[-p2.fdo), (18)

M2 M2 M2

we can get the natural discrete representation of the integral invariants for the two-
dimensional surface [ 25 ]

S(M2) = ~ o-0k - 1, (19)

(i;k)
A(M2) = ~ Aij . (zr - aij ),
1
(20)
(ig)

(i)

see Fig. 1.

4. Factorization and the goni-hedric principle

The important message we get from the Steiner expansion and from the formulae
above is that the integral invariants are (the sum of) products of a volume of the faces
of the surface M2 and a volume of the spherical images of these faces,

integral invariant = Z (volume of face) - (volume of spherical image). (22)

{faces}

Indeed the spherical images of the faces of M2 are

triangle (ijk) ~ point on S2, (23)

edge (ij) ~ arc on S2, (24)
vertex (i) --~ spherical polygon on S2. (25)

and the volume elements on these spherical images are

triangle (ijk) --~ 1, (26)

edge (ij) -'->('71"-- Olij), (27)

vertex (i)---} ( 2 7 r - Z f l i ) , (28)

i.e. equal to one for a point in S2 which is the image of a triangle, equal to the length
of the arc in S2 which is the image of an edge, and equal to the area of the polygon
on S2 which is the image of a vertex. They are functions of the internal and external
angles (13), and (14). The factorization or goni-hedric structure (22) of the integral
invariants ( 1 9 ) - ( 2 1 ) is transparent now.
The generalization of the factorization property (22) of the classical integral invariants
in the form of the goni-hedric principle can be used now to construct new integral
 J. Ambj¢rn et al./Nuclear Physics B 486 (1997) 390-412 399

7f -- Ol

Fig. 2. On every spherical polygonone can define the area (30), the perimeter (31) and the curvature (32).

invariants on a manifold M2. In accordance with this principle one should always
multiply the volume of the face by one of the geometric measures, associated with the
corresponding spherical image,

(volume of face) • (geometric measure on spherical image). (29)

{faces}
In the two-dimensional case considered so far the area functional S(M2) is uniquely
defined since there is no non-trivial measure associated with the point on S2 which is the
spherical image of the triangles of M2. The same is true for the linear functional A(M2),
because only the length of the arc can be associated with the spherical image (27) of
the edge (24).
Only the topological invariant (21) can lead to new expressions because the spherical
image (25) of the vertex (i) is a polygon on S2 (see Fig. 2), and the polygon has suffi-
ciently structure to allow non-trivial measures. The possible purely geometric measures
defined on a polygon on S2 are (see Fig. 2)

the area = (2¢r - ~ _ , f l i ) , (30)

the curvature = ~ (zr - fli), ( 31 )

the perimeter = ~ ( ~r - ai.i ). (32)

J
As we have seen the total area of the spherical image is proportional to the Euler
character (see (21), (28) and (30)). The total curvature of the spherical polygon
cannot be used, because it is possible by a continuous deformation of M2 to ensure
that there is zero curvature (no deficit angle) associated with a given vertex, in which
case the polygon image (25) will disappear but it will leave a non-zero curvature (31).
Such a situation is not physical interesting and we exclude it. This means that only the
perimeter (32) can be used in the construction of new invariants, and in accordance
with (29) we define

o(M2) -- E (33)
(i)
400 J. Ambjcrn et al./Nuclear Physics B 486 (1997) 390-412

Every term with dihedral angle eeij appears in both vertices (i) and (j). Therefore after
resummation one gets

O(M2) = 2Z(Tr - olij). (34)

(ij)

The trivial resummation of quantities given on a vertex into a sum over adjacent edges
will again appear in high dimensions but in the less trivial form of a dual transformation.
The invariant (34) is equal to the total length of the arcs of the spherical image of
all edges of the surface M2. This quantity is a natural definition of the total e x t r i n s i c
c u r v a t u r e of the triangulated surface. There is a simple mnemonic rule by which one can
get this integral invariant out of the formulae already available, i.e. ( 1 9 ) - ( 2 1 ) . Indeed,
if we shall consider a triangulated surface with fixed length of all edges ,~ij = a (as in
the case of dynamical triangulated surfaces), then the functional A ( M z ) is proportional
to O(Mm),

a ( m 2 ) = a . O(mm). (35)

Let us finally consider generalizations where

measure on spherical image ~ 0(measure on spherical image),

as discussed above. The obvious and simplest example is

7r - a i j ~ ]Tr - aij[.

In this way we can define a new integral invariant

~(Mm) = ~ [rr - aq], (36)

(q)
which is a natural discretized version of

M2

It is also possible to consider more general functions [ 18]

1r - a --~ O ( a ) , 0(or) = 0, (38)

where 0 ( a ) increases monotonically at both sides from the point a = ~- and near that
point has the following parametrization:

0(ce) = (,rr - a ) ~. (39)

For these more general functions (36) is replaced by

O(Mm) = Z O(aij). (40)

(;j)
The index 9 (39) is important for the convergence of the corresponding partition function
and the critical properties of the system [ 18,19].
 J. Ambjornet aL/Nuclear PhysicsB 486 (1997)390-412 401

In order to demonstrate the value of these invariants for the theory of random surfaces,
let us compute them for a sphere of radius R which is triangulated in the following way:
first we divide it in n degrees of latitude and m degrees of longitude and each square
obtained this way is divided in two triangles. With this division we get for the action
A(M2) defined by (20)

A(S2;~)=4rrR((n)l-~+(b)l-~ ) (41)

and for the action O(M2) defined by (40)

O($2; ~) = anm 1-¢ + bmnl-% (42)

where a and b are constants. We have a different behavior of the action in the limit
n,m --+ oo. If O(a) = 1 - cos o~, i.e. ¢ = 2, we get
O($2; ¢ = 2) = rr 2m + 2~r n . (43)
n m
Thus the index ~ influences in an essential way the convergence of the partition function,
and should be chosen less than one in order to have an interesting theory.

5. Three-dimensional manifolds

We consider now a four-dimensional domain D4 in E 4 bounded by the three-dimen-

sional manifold M3. The three-volume of the parallel manifold M~ has the form (12),
i.e.

V(M~) = J'RIR2R3dco3 + p" f {R1R2 + R2R3 + R3R1}do93

M3 M3

dw3, (44)
,Y

M3 M3

or in terms of the notation used so far

V(M~) = V(M3) + p . S(M3) + p2. A(M3) + p3. N(M3), (45)

which defines the volume, the area, the linear size and the topology of the domain.
We now apply the method of parallel expansion to the piecewise linear three-
dimensional manifolds to get discrete versions of the corresponding integral invariants.
Following the procedure in two dimensions we introduce angles between one, two and
three simplexes in order to define the internal and external geometry of the simplicial
manifold (see Fig. 3),

~i.j;ik ~ fli for internal angles, (46)

flijt:ijk --flij for internal angles, (47)
Cqjkl;ijkm=--aijk for external angles (48)
402 J. Ambjcrn et al./Nuclear Physics B 486 (1997) 390-412

(iplm) (ijim)

7z - J~ (ijkl)

{~jkm)
[ipmk }
7( - ~rnk
Fig. 3. The vertex of the three-dimensional simplex (shown on the upper figure) and the corresponding
spherical polyhedron on S3. The volume of this image is denoted by to~3), the area by ~ oJ~f), where the
summation is over all polygons of the given polyhedron, and finally the linear size is denoted by ~ , / 2 ijk
(2) o~ijk
(l)
and the summation is over all arcs of the polyhedron.

and we introduce the corresponding spherical images of the simplexes of M3 as

tetrahedron (ijkl) --~ point on S 3 , (49)

triangle (ijk} ~ arc on S 3, (50)
edge (ij} ~ polygon on S 3, (51)
vertex (i) ~ spherical polyhedron on S 3. (52)

Using the angles already defined on the simplexes ( 4 6 ) - ( 4 8 ) one can find the geometric
measures on the spherical images ( 4 9 ) - ( 5 2 ) ,

tetrahedron (ijkl) ~ 1, (53)

triangle (ijk) --* (zr - ce~ik), (54)
edge (ij) ~ (277"-- Z ~ i j ) , (55)

vertex (i} ~ w[3)(a,/3). (56)

 J. Ambjorn et aL /Nuclear Physics B 486 (1997) 390-412 403

Using this information it is not difficult to compute the volume of the parallel manifold.
It will contain four parts which we should identify with invariants (45) of the simplex
M3. If Uijkl denotes the volume of the tetrahedron (ijkl) and o~}3) denotes the volume in
S3 of the spherical polyhedron corresponding to vertex (i), we obtain

the volume V(M3) = ~ Uijkl" 1, (57)

(ijkl}

the area S(M3) = ~ o'/jk • (,n"- olijk) , (58)

<ii~}
the linear function A ( M 3 ) = ~ Aij. (2,n" - ~--~ 13~i) , (59)
(ij)

the topological invariant N( M3 ) = ~ 1 • w~3) ( ol, 13). (60)

<i>

All these integral invariants have the factorized form (22).

Let us now consider the possible new integral invariants along the lines already
discussed in the two-dimensional case. Again the spherical images of the two highest
dimensional simplexes, the tetrahedron and the triangle are of lowest dimension, i.e. a
point and an arc, respectively, and it is impossible to find any geometric extension for
these objects. The first non-trivial simplex is from this point of view the edge, which
has as its spherical image a polygon. For the polygon we can in addition to its area also
use its perimeter as a geometric measure, and hence we find a linear invariant different
from A(M3),

a( M3) = ~-~ ,~ij . (~'~ ('n" - olijk ) ) , (61)

<0>

in accordance with our principle (29). The spherical image of a vertex (i) is a spherical
polyhedron (52), and for this object we have the topological invariant (60), which
corresponds to the use of the three-volume w}3) as the geometric measure. However,
we can also use the arc-length or the area of the two-simplexes in the polyhedron as a
geometric measure (see Fig. 3), i.e.

~13i(7"g--Olijk) or ~ (2"n"- ~13ij)" (62)

This leads to the following new integral invariants

19(M3)=~ l . (~13i(qr--otijk)), (63)

,,~(M3)=~-~ l. (~-~(27r- ~-~13ij) ). (64)

404 J. AmbjCrn et al./Nuclear Physics B 486 (1997) 390-412

5.1. The dual representation

There is an equivalent or "dual" form of the invariants (61), (64) and (64), which
can be obtained after resummation. In (61) we can combine all edges belonging to a
given triangle (ijk) to get a sum ~i.j + ajk + l~ki = "~ijk. Hence,

A(M3) = Z Aijk " ('71" -- Ol~jk ) , (65)

(ijk)

where aijk is the angle between two neighboring tetrahedra of M3 having a common
triangle (ijk) of the perimeter hijk. This is the dual form of the integral invariant (61)
since either one can multiply the length of the edge by the total perimeter of the
polygon which is the spherical image of the edge, or one can multiply the perimeter of
the triangle by the length of the arc of the spherical image of the triangle, as in (65).
The same kind of transformations, in the form of a resummation of the quantities
given at the vertices and in this way transferring the sum to one over adjacent edges or
triangles, work for the two invariants (64) and (64). Collecting the terms belonging to
the same triangle (/jk) in (64) we can get (t~i "31- t~j -~ i~k ) ('t"1" -- O[ijk ) = 77"(77" -- Olijk ) ,
therefore

O(M3) = ~ ( 7 r -- o[ijk). (66)

(ijk)

Note that O(M3) is the total extrinsic curvature of M3. If we collect the terms belonging
to the same edge in (64) we simply get

~(M3) = ~-~ (2¢r- Z [ ~ i j ) . (67)

(ij)
~ ( M 3 ) is the total deficit angle of M3.
The goni-hedric principle (29) which we used to construct extensions can now be
formulated in dual form: either one should multiply the volume of the face by all
possible geometric measures which can be constructed on its spherical image, or one
should multiply all possible geometric measures on the face by the volume of its
spherical image.
Finally, we remark that also in the three-dimensional case one can of course introduce
more general invariants by considering functions 0 of the geometric measures on the
spherical images, as was done in the two-dimensional case.

6. Four-dimensional manifolds

The Steiner expansion for the hyper-volume g) of a smooth manifold M4 reads

J2(M~) = ~ ( M 4 ) + p . V(M4) + p2. S(M4) + p3. A(M4) + p4. x ( M 4 ) (68)

and by comparison with the same expression for a piecewise linear manifold M4 we get
the discrete versions of the above invariants
 J. AmbjCrn et al./Nuclear Physics B 486 (1997) 390-412 405

(ijklm) (ijkl) (ijk) (ij) (i)

,(1) ,(2) .(3) w}4)
1 ~ijkl ~ijk wij
Fig. 4. The sub-simplexesof the four-dimensional manifold (second line) and the corresponding spherical
images on S4 (first line). The hyper-volumeof the image is defined as w~, the volume of the image as
w~)), the area of the image as ~ /2(2)w(2) and finally the length of the image as/..., ijk ijkt
ijk ijk,

~(M4) = ~ vi2ktm" 1, (69)

(ijklm)

V(M4) = ~ Uijkl • (07" -- Olijkl), (70)

(ijkl)

S(M4)mZO'ijk'(271"--Z/?ijk), (71)
{ijk)

A(M4) = Z aij" o9~3)(a, fl), (72)

(ij)
X(M4) = Z 1. o)}4) (/?), (73)
(i)

where as before we introduce internal angles fli,/38, flOk between one, two and three
simplexes and aijkl for the external angle between two four-dimensional simplexes, as
well as the notation vijttm for the four-volume of the four-simplex (ijklrn) and the
notation o9}4) for the four-volume in S4 of the spherical image of the vertex (i) (see
Fig. 4). Note that o94 is independent of the external angle c~. This is in agreement
with the general pattern already mentioned which states that the coefficients to even
powers of p are intrinsic integral invariants, while the coefficients to the odd powers of
p contain reference to the extrinsic geometry. The whole four-volume (73) obtained by
summation over all vertices of the piecewise linear manifold M4 is proportional to The
Euler-Poincar6 character in the same way as in the two-dimensional case.
To analyze the possible new integral invariants in four and higher dimensions, we
need to introduce a universal notation for solid angles on the faces of a simplex and on
the corresponding spherical images. We shall use g2(k) for solid angle at a vertex of a
k-dimensional simplex (see Fig. 5) and og(k) for the solid angles on spherical images
(see Fig. 4). All these solid angles are functions of the previously introduced angles
denoted by a ' s and fl's with various indices, see Fig. 5.
The possible geometric measures on the spherical image of the vertex (i) are: hyper-
volume, volume, area and length [26,21-23], (see Fig. 4),

o9~4) ' ~ ogij(3) , ~ ~(2) (2) '

'llijk ogijk Z ,~(3) (1)
J lijkl o9ijkl" ( 74 )
406 J. Ambjern et al./Nuclear Physics B 486 (1997) 390-412

o o o j ~ k 1 l
i i j

O(2) 0( 3) 0(4)

Fig. 5. Solid anglesof the facesof a vertexand the geometricmeasures.

Using these measures there is one new invariant of the "area" type (like S ( M 4 ) ) ,
namely

/d,s(M4) = E O ' i j k " (EO)~j~I). (75)

(ijk)
There are two new invariants of the "length" type (i.e. like A(M4))

/d,A(M4) = Z ,A/j" E O)ijk

(ij) E J~(3)('o(l)qkt
q~t (76)

and finally three new invariants which are dimensionless, (like x(M4)),

fl'x(M4) = E l" E J ~ ) 0)~' . (77)

(i) Z ¢~(3) .(])
~ijkl wijkl
In the next section we shall perform the dual transformation of the above invariants and
address the question of their internal and external properties.

6.1. Dual representations

To get the dual form of the above invariants we combine terms in the parentheses
belonging to the same tetrahedron in the first, third and sixth of the invariants in ( 7 5 ) -
(77),

Aijkl = E ' ~. . ~fl(k l 3 )

"'u • (78)
ff2ijkl over tetrahedron(ijkl) f2ijkl

Hence, we have the following sequence of invariants constructed from w ijkl

(l) associated
with tetrahedron (ijkl):

pto,(M4) = E Orijkl ' O')(1)

•~ijkt ijkl' (79)
(~ikt) I2;jkt
 J. Arabj¢rn et aL /Nuclear Physics B 486 (1997) 390-412 407
.(1)
where wijkt = 7r-- Otijkl is the angle between two neighboring four-dimensional simplexes
having a common tetrahedron (ijkl I of volume vqkt, area tr/jkt, length Aijkl and total
internal solid angle ,Oijkt.
Combining the second and the fifth term in the parentheses in (75) and (77) belong-
ing to the same triangle

over triangle (ijk)

,(2),
we get integral invariants constructed from the ~'ijk ° associated to triangles ijk,

/-t,o, ( M 4 ) = ~ . , (81)
(/jk) 7r
(2)
where tOOk = 27r -- ~ t~ijk is the deficit angle on the triangle (ijk) which has the area
tr/jk, the perimeter ,tiyk and as usual the sum of internal angles equal to ~r. The first
invariant in this family coincides with the discrete version of the Hilbert-Einstein action
[ 1 ], the second coincides with the linear action suggested in Ref. [ 14] and the last one
is equal to the total deficit angle of the whole manifold [ 10].
The fourth invariant is simply equal to the volume of the spherical image of the edge
and associated to the linear term (72) we have two integral invariants constructed from
the three-volume of the spherical image ~o~3) of the edge </j>,

.to(M4) -~ Z { /~ij
1 } " t°~/)" (82)
(q)
To understand why some of the integral invariants introduced above are intrinsic
and are independent of the embedding one should use the Gauss-Bonnet theorem. The
Gauss-Bonnet theorem provides us with a general relation between the volume of the
spherical image of the vertex and the solid angles associated to the vertex. Indeed, the
Gauss-Bonnet theorem for the two-dimensional triangulated surface can be formulated
in the form [22-24,26]

lo !2 > + 1 n<2> = 1
47r ' 4--~ ijk -2" (83)
Hence, summing over all vertices in the triangulation we get

x(M2) = 1 Z w~2) = Z { 1 - ~1/--.,~-~$2!2)1ukJ = N o - N2/2, ( 84)

(i) (i)

where No is the number of vertices and N2 the number of triangles on the surface M2.
This discussion can be generalized to four dimensions: in every vertex of the simplicial
m a n i f o l d M4 we have [20,22-24,26,27]

3 (.0~4) "at- 1 ~ ~(2)

k "Jr-~ 1 Z/'}(4)
(2)
JZijk ~Oij "~jk~., =
1
~, (85)
408 J. Ambjcrn et al./Nuclear Physics B 486 (1997) 390-412

which shows that the hyper-volume (.O~4) on S4 can be expressed through the intrinsic
quantities. The Euler-Poincar6 character is equal to

)((M4) ~ ~
3 Z (.0}4) ~ Z
(i) (i)
{,| ~2Z n(4) },, - --Z
(ijk)
wiJ
k'(2)

= No - N 2 / 2 -4- N4 (86)

and we obtain the relation between the total deficit angle

(2) l ~ .(2)
O')t°t ~ 2~ ~ Wijk (87)
(ijk)
and the total solid deficit angle

0[4) --=Z (1 - ~ 1 ,.1~!4'"~ -- ff"£~ 1 ,..~(4) / (88)

in the form [28]

,¥(M4) =/2~4) - lo9(2)

2 tot" (89)
The Euler-Poincar6 character is

1 f 2 2
x ( M 4 ) - 128,rr2 a do4 (Ruvap - 4Ruv + R2) , (90)
M4
and comparing with the discretized version (89) of x ( M 4 ) it is not unnatural to
associate higher powers of the integral of curvature tensors with linear combinations of
J2{o4t) and w(2)tot,although we have no explicit identification of these two terms entirely
in terms of integrals of R 2, R2~, etc.
From these considerations it seems that the general discretized action in simplicial
tbur-dimensional gravity should be of the form mentioned in the Introduction,

1
acti°n=aZ°-Uk'%Jk
.(2, .~-.~(.0(2 ,
+g'z..., ijk +g2 Z
( O~j~))2 + . . .
((jk) (ijk) ((jk)

+k, Z (1 - ]--~--[2'
2 4)~2../7.
, ] + k2 Z (1 - -
~ 24 ) )d"~,
l 2 + ..., (91 )
(i) (i)
where the terms involving the square of the deficit angles introduce an intrinsic rigidity
into the simplicial manifolds. This choice of discretized action with higher derivative
terms is related, but not identical, to the terms suggested in Ref. [29].
To summarize, we expect that the general form of the action of non-perturbative
simpliciai quantum gravity can be written as

) k + ~ O(O.~j~)) + Z (p( ,f-~t4) ),

Action = ~I Z O . i j k . . ( 2o.,/j (92)
(ijk) (ijk) (i}
 J. Ambjcrn et al./Nuclear Physics B 486 (1997) 390-412 409

where the functions 0 and q0 should be chosen such that the system will develop a
non-trivial fixed point with the desired properties.

7. Higher dimensions

It is not difficult to extend these results to higher dimensional manifolds, using the
same ideas and constructions. The extension of the formulas ( 6 9 ) - ( 7 3 ) and ( 7 5 ) - ( 7 7 )
or their dual form ( 7 9 ) - ( 8 2 ) is straightforward and has the same structure as already
encountered in four and lower dimensions.
One new aspect in higher dimensions is the appearance of additional intrinsic integral
invariants which are of interest in gravity and membrane theory. The first one coincides
with the Hilbert-Einstein action,

.(2)
A(Ms) = ~ Uijkl Wijkl,
" (93)
(ijkl)

where V/jkt is the three-volume of the tetrahedron (ijkl) and ~o/jkt

(2) is the deficit angle.
The second invariant is proportional to the linear size of the manifold,

A(Ms) = ~ Aij • o.)ij(4) , (94)

(ij)

where A0 is the length of the edge and co~j4) is the hyper-volume of the corresponding
spherical image on S5. Both measures coijkl
(2) and coij
.(4) are expressible in terms of intrinsic
angles, as we have seen in the previous section, and both measures produce a sequence
of integral invariants in various dimensions. The role of the deficit angle oJijkt
(2) is well
understood because this measure leads to the Euler character of the two-dimensional
surfaces (21) and makes it possible to formulate discrete versions of classical gravity,

Action (d) = ~ u(d--2)(.O (2) , (95)

where v (d-2) denotes the volume of the (d - 2)-dimensional sub-simplex and w (2) the
deficit angle. The role of O9(4) is very similar for four and higher dimensional manifolds
where it leads to the Euler-Poincar6 character (73) and allows us to generate a new
sequence of invariants,

Action (d) = ~ u(d--4)O9 (4), (96)

which may play a role in the quantum theory of extended objects. It might be difficult
to investigate the effect of these actions by analytical means, but they seem well suited
for use in numerical simulations.
 410 J. Ambjcrn et al./Nuclear Physics B 486 (1997) 390-412

8. Discussion

We have presently no continuum theory of Euclidean quantum gravity. Let us assume

that it exists and that it can be represented by a path integral. It is natural to regularize
quantum gravity by restricting the functional integral to piecewise linear manifolds. In
dimensions d less then seven there is a one-to-one correspondence between smooth and
piecewise linear structures. In quantum gravity it is a very interesting and completely
unsolved question which structures one should include if we perform the summation
over topologies, but at least one can say that in four dimensions we will loose no
information by the restriction to piecewise linear manifolds compared to the situation
where we restrict ourselves to smooth manifolds. If we restrict ourselves to a fixed
topology we are faced with the problem of integrating over all metrics associated with
piecewise linear manitblds of the given topology. Two approaches have so far been
attempted in four dimensions. The first and oldest is a straightforward generalization of
classical Regge calculus where the link lengths are used as dynamical variables. The
second uses so-called dynamical triangulations, i.e. the connectivity of the triangulation
creates the dynamics of space-time. They both have advantages and disadvantages which
we will not discuss here. However, the phase diagrams of the two discretized theories
look quite similar and most likely they both require the addition of terms beyond the
simplest Regge term. 2 For this reason we have analyzed which terms one can add in a
"natural" way on piecewise linear manifolds. Using the Steiner construction of parallel
manifolds we were led to the goni-hedric principle for piecewise linear manifolds, and a
whole class of terms emerged, among those the original Regge term and the well known
Euler characteristic. In four dimensions these terms were functions of the two intrinsic
deficit angles which are present in four-dimensional triangulations, the triangle deficit
angle and the vertex deficit angle. The last one is new and it has only recently been used
in numerical simulations without any geometric foundation. The need for this term is
most visible in the framework of dynamical triangulations. The characteristic feature of
the so-called crumpled phase is the existence of vertices of extremely high order. Such
vertices are seemingly preferred for entropy reasons, but presumably unwanted in the
continuum limit. New terms which are functions of the vertex deficit angles will very
efficiently suppress these high order vertices. Such computer simulations are presently
performed by various groups.
The computer simulations of simplicial quantum gravity (and the analytic results in
two-dimensional quantum gravity) illustrate the somewhat ignored fact that the dynami-
cal generated dimension of space-time might differ from that of the underlying manifold.
In our real world we observe that they agree (and are four!), at least in the infrared
regime. We call it rigidity o f space-time. Simplicial quantum gravity might possess this
rigidity if the new terms proposed here are added to the action. In fact we might end

2 In the case of dynamical triangulations this is not finally clarified, but there are strong indications that the
phase transition is a first order transition if the action only contains the Regge term. In the case of Regge
calculus the transition is first order if one does not add higher curvature terms.
 J. Ambjcrn et aL /Nuclear Physics B 486 (1997) 390-412 411

up in the interesting situation that four-dimensional simplicial gravity will only have
a c o n t i n u u m l i m i t p r o v i d e d the d y n a m i c a l d i m e n s i o n o f space-time agrees with the
k i n e m a t i c d i m e n s i o n o f space-time. It seems a most important question to understand
the constraints to be i m p o s e d on the action in order that the q u a n t u m theory fulfill the
r e q u i r e m e n t o f rigidity.

Acknowledgements

G.K.S. thanks R o l f S c h n e i d e r for helpful discussions and the Niels B o h r Institute for
hospitality. J.A. a c k n o w l e d g e s the support o f the Professor Visitante Iberdrola Grant and
the hospitality at the U n i v e r s i t y o f Barcelona, where part o f this w o r k was done.

References

[1] T. Regge, Nuovo Cimento 19 (1961) 558.

[2] A. Jevicki and M. Ninomiya, Phys. Rev. D 33 (1986) 1634.
[3l H.W. Hamber, Nucl. Phys. B (Proc. Suppl.) 25A (1992) 150.
14l J. Ambjcm, B. Durhuus, J. Fr6hlich and P. Orland, Nucl. Phys. B 257 (1985) 433; B 270 (1986) 457;
B 275 (1986) 161.
[5] E David, Nucl. Phys. B 257 (1985) 543.
[6] V.A. Kazakov, I. Kostov and A.A. Migdal, Phys. Lett. B 157 (1985) 295; Nucl. Phys. B 275 (1986)
641.
[7] V. Knizhnik, A. Polyakov and A. Zamolodchikov, Mod. Phys. Lett. A 3 (1988) 819.
[8] E David, Mod. Phys. Lett. A 3 (1988) 1651;
J. Distler and H. Kawai, Nucl. Phys. B 321 (1989) 509.
[9] J. Ambj~rn, B. Durhuus and T. Jonsson, Mod. Phys. Lett. A 6 (1991) 1133;
J. Ambj~rn and S. Varsted, Phys. Lett. B 266 (1991) 285; Nucl. Phys. B 373 (1992) 557;
M.E. Agishtein and A.A. Migdal, Mod. Phys. Lett. A 6 (1991) 1863;
B. Boulatov and A. Krzywicki, Mod. Phys. Lett. A 6 (1991) 3005;
J. Ambj~rn, D.V. Boulatov, A. Krzywicki and S. Varsted, Phys. Lett. B 276 (1992) 432;
N. Sakura, Mod. Phys. Len. A 6 (1991) 2613;
N. Godfrey and M. Gross, Phys. Rev. D 43 (1991) R1749;
S. Catterall, J. Kogut and R. Renken, Phys. Lett. B 342 (1995) 53.
[ 10] J. Ambjcrn and J. Jurkiewicz, Phys. Lett. B 278 (1992) 50;
M.E. Agishtein and A.A. Migdal, Mod. Phys. Lett. A 7 (1992) 1039; Nucl. Phys. B 385 (1992) 395;
J. Ambj~rn, J. Jurkiewicz and C.E Kristjansen, Nucl. Phys. B 393 (1993) 601.
[11 ] D. Weingarten, Nucl. Phys. B 210 (1982) 229.
[12] J. Ambj~rn, J. Jurkiewicz and C.E Kristjansen, Nucl. Phys. B 393 (1993) 601;
B. Bruegmann and E. Marinari, Phys. Rev. Lett. 70 (1993) 1908;
B. Bruegmann, Phys. Rev. D 47 (1993) 3330;
S. Catterall, J. Kogut and R. Renken, Phys. Lett. B 328 (1994) 277;
B.V. DeBakker and J. Smit, Nucl. Phys. B 439 (1995) 239;
J. Ambjorn and J. Jurkiewicz, Nucl. Phys. B 451 (1995) 643;
R Bialas, Z. Burda, A. Krzywicki and B. Petersson, e-Print Archive: hep-lat/9601024.
[ 13] H. Kawai, N. Kawamoto, T. Mogami and Y. Watabiki, Phys. Lett. B 306 (1993) 19;
J. Ambjorn and Y. Watabiki, Nucl. Phys. B 445 (1995) 129.
[ 14] G.K. Savvidy and K.G. Savvidy, Interaction hierarchy. String and quantum gravity, Mod. Phys. Lett. A
11 (1996) 1379, Hep-th 9506184.
115] J. Steiner, Llber parallele Fl[ichen, Gesammelte Werke Band 2 (Berlin, 1882) S, 171-176.
1161 H. Weyl, Am. J. Math. 61 (1939) 461.
412 J. AmbjCrn et aL /Nuclear Physics B 486 (1997) 390-412

117] R.V. Ambartzumian, G.K. Savvidy, K.G. Savvidy and G.S. Sukiasian, Phys. Lett. B 275 (1992) 99.
[181 G.K. Savvidy and K.G. Savvidy, Mod. Phys. Lett. A 8 (1993) 2963; Int. J. Mod. Phys. A 8 (1993)
3993.
119] B. Durhuus and T. Jonsson. Phys. Lett. B 297 (1992) 271.
[20] C.B. Allendoerfer and A. Weil, Trans. Am. Soc. 55 (1943) 101.
[21] G. Herglotz, Abh. Math. Sem. Hansischen Univ. 15 (1943) 165.
1221 C.B. Allendoerfer, Bull. Am. Math. Soc. 54 (1948) 128.
[231 L.A. Santal6, Proc. Am. Math. Soc. I (1950) 325.
[241 L.A. Santal6, Revista Un. Mat. Argentina 20 (1962) 79.
[25] G.K. Savvidy and K.G. Savvidy, Phys. Lett. B 337 (1994) 333.
1261 R. Schneider, Geom. Dedicata 9 (1980) l 1 I.
[27] J. Cheeger, W. Miilleer and R. Schrader, Commun. Math. Phys. 92 (1984) 405.
[281 M. Rocek and R.M. Williams, Phys. Lett. B 273 (1991) 95.
[29] H.W. Hamber and R.M. Williams, Nucl. Phys. B 248 (1984) 392.