Two-Point Functions in (Loop) Quantum Cosmology: Classical and Quantum Gravity November 2010
Two-Point Functions in (Loop) Quantum Cosmology: Classical and Quantum Gravity November 2010
Two-Point Functions in (Loop) Quantum Cosmology: Classical and Quantum Gravity November 2010
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/47817679
CITATIONS READS
20 41
3 authors:
Daniele Oriti
Max Planck Institute for Gravitational Physics (
100 PUBLICATIONS 3,114 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Daniele Oriti on 09 July 2017.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
AEI-2010-162
DAMTP-2010-91
Gianluca Calcagni
Max Planck Institute for Gravitational Physics (Albert Einstein Institute), Am
Muhlenberg 1, D-14476 Golm, Germany
E-mail: [email protected]
Steffen Gielen
DAMTP, Centre for Mathematical Sciences, Wilberforce Road, Cambridge CB3
0WA, U.K.
Max Planck Institute for Gravitational Physics (Albert Einstein Institute), Am
Muhlenberg 1, D-14476 Golm, Germany
E-mail: [email protected]
Daniele Oriti
Max Planck Institute for Gravitational Physics (Albert Einstein Institute), Am
Muhlenberg 1, D-14476 Golm, Germany
E-mail: [email protected]
1. Introduction
One of the major problems of modern theoretical physics is the resolution of singularities
in the structure of classical spacetime. In particular, the big bang problem in General
Relativity has been inspiring a steady trend of research aiming at the removal of the
initial singularity via quantum effects, i.e., by replacing classical General Relativity with
a quantum theory of spacetime and gravity. The formal development of a background-
independent theory of quantum gravity is no easy task and some physical insight is
gained by looking at a mini-superspace toy model where the gravitational degrees of
freedom are reduced to a minimum. On this FriedmannRobertsonWalker (FRW)
background, one can write the gravity and matter Hamiltonian and promote it to an
operator on a Hilbert space of states associated with the probability for the universe
to be in a given phase-space configuration. The ensuing quantum theory can then be
studied by exact or perturbative techniques and the fate of the big bang singularity
examined. While traditional quantum cosmology corresponded to a symmetry-reduced
version of the canonical quantum gravity in ADM variables [1], a new scheme for
quantum cosmology, called loop quantum cosmology (LQC) [2], has been developed
drawing inspiration from loop quantum gravity (LQG), an approach to the canonical
quantization of gravity based on connection variables [3, 4]. While the traditional
WheelerDeWitt quantization generally fails to solve the problem, loop quantum
cosmology does avoid the singularity by replacing it with a bounce at a finite curvature
and matter density. Unfortunately, the robustness of the results of LQC depend on
the yet unclear connection between the model and the full theory of loop quantum
gravity, itself still under active development and far from being complete. In particular,
concerning the full theory, most recent activities towards a complete definition of the
quantum dynamics have focused on a covariant or path integral version of the theory,
going under the name of spin foam models [5, 6]. Just as LQG states are graphs labelled
by group elements or group representations, spin foam histories are 2-complexes labelled
by the same type of data. Recently, a path-integral formulation has been proposed also
for LQC [7, 8]. The main virtues of this reformulation are the following: a) obviously,
this alternative definition of the quantum dynamics could be advantageous from the
purely technical point of view as well as offer new conceptual insights; b) it recasts
the dynamics of LQC in a form that formally resembles the spin foam formulation of
the full theory, in particular it is expressed in terms of a vertex expansion (see below
and [7, 8]), and could be used as testing ground for techniques later to be applied to
spin foam models; c) it could exemplify issues and conceptual points that one would
eventually have to address in the spin foam formulation of quantum gravity, and suggest
solutions to the same. On this last point, several interesting questions and suggestions
already came out of the work in [7, 8] and from follow-up work (e.g., [9]), concerning
the possibility of defining different types of two-point functions in LQC, which maybe
encode different causal properties and dynamical information. In fact, this point had
been raised earlier in the quantum gravity literature [10, 11] and later in the spin foam
Two-point functions in (loop) quantum cosmology 3
in the presence of a massless scalar field, as we will show, there exist physically equivalent
representations of the two-point functions, as well as an identical catalogue of them,
obtained in roughly the same manner, and depending on the same choices. These
results hopefully settle some recent debate about the origin of different definitions for
the inner product in cosmology, and their mutual relations. Such issues are also discussed
from the perspective of building a third quantized or multiverse theory, where the role
of different sectors and the choice of inner product are drastically reinterpreted. We
stress that our findings hold independently of the form of the gravitational part of the
Hamiltonian constraint: While they apply rigorously to loop quantum cosmology where
the gravitational configuration variable defined below takes discrete values, they can
be extended at least formally to the standard WheelerDeWitt equation, upon the
replacement of summations over by integrations.
2.1. Kinematics
Focussing on the gravitational variables first, after a choice of orthonormal frame {0 eai }
and dual {0 eia } compatible with 0 qab one may parametrize the AshtekarBarbero su(2)
connection Aia and the densitized triad Eia by
p
c p det 0 q 0 a
Aia = 1/3 0 eia , Eia = 2/3
ei , (1)
V0 V0
where powers of V0 , the 3-volume of the fiducial cell as measured by 0 qab , have been
introduced to make the coordinates (c, p) invariant under rescalings of 0 qab . The
coordinates are canonically conjugate variables,
8G
{c, p} = , (2)
3
where is the BarberoImmirzi parameter, and the only non-trivial dynamics of the
dimensionally reduced system is encoded in the Hamiltonian constraint of first-order
general relativity.
Here and in the following, spatial tangent indices are denoted by a, b, c and internal su(2) indices are
denoted by i, j, k.
Two-point functions in (loop) quantum cosmology 5
2.2. Dynamics
To obtain the space of physical states, one is now left with solving the Hamiltonian
constraint which can be written in the form
C (, ) + 2 (, ) ,
(7)
g
where is a wavefunction on configuration space and only acts on Hkin . In LQC,
is a difference operator which typically takes the form (e.g., [7])
(, ) := A()( + 4l0 , ) + B()(, ) + C()( 4l0 , ) , (8)
k The algebra of continuous functions on this space is equivalent to the almost periodic functions on
R as introduced in [17].
Two-point functions in (loop) quantum cosmology 6
where A, B, C are functions (dependent on the details of the quantization scheme) and
l0 is an elementary length unit, usually defined by the square root of the area gap
encountered in loop quantum gravity (i.e., the Planck length up to a numerical factor).
In WheelerDeWitt theory is a differential operator acting on L2 (R). The form of
the operator in either theory will not be important in the following, except for the
g
fact that in LQC there is an intervals worth of superselection sectors in Hkin ; namely,
preserves all subspaces spanned by {|0 + 4nl0 i|n Z} for some 0 . We may restrict to
one of these subspaces, i.e., assume that wavefunctions only have support on a discrete
lattice which we take to be 4l0 Z to include the physically most interesting point = 0.
This restriction picks out a separable subspace of the originally non-separable Hilbert
space. Our analysis will naturally apply to any with this property in the setting of
LQC, and to any differential operator in WheelerDeWitt theory in the formal sense
that we will explain below.
Moreover, the space of states solving equation (7) is reducible in terms of two
sectors of solutions satisfying, respectively, (p ) (, ) = 0, resulting from the
fact that p is a Dirac observable for the system governed by the above dynamics. This
will allow us to perform an explicit splitting of the space of physical states into positive-
and negative-frequency states.
Alternatively, one may remove the absolute value in the definition (10), to obtain the
indefinite normalization
hp |p i = 2p (p p ) . (13)
For our purposes, the volume eigenstates |i can equivalently be understood as
playing the role of the spatial coordinates x (or of the spatial momenta p, as our
manipulations are performed always in the basis). We will make use of this analogy
to extend what is known about two-point functions for a relativistic particle to the case
at hand, in particular their composition laws as analysed in [20, 21]. We note that the
inclusion of a scalar field simplifies the formalism due to the existence of an internal
time variable, but it is also important in order to give an operational meaning to a
quantum system describing the universe as a whole, in particular in this case where
gravity is approximated by a single degree of freedom.
As we will show in detail, entirely as in the particle case, the definition of different
two-point functions is fully characterized by:
The canonical representation chosen for the physical quantum states in
configuration space (more precisely, in the space). Typically, one has three
choices. The first is a non-relativistic representationP in terms of states
|, inr |, ; +inr + |, ; inr , (14)
where
Z
dp
|, ; inr = p eip (p ) |, p i . (15)
2|p |
These states are orthogonal at equal values of in the physical inner product
and, choosing the positive-definite normalization (12) the suitable positive-definite
resolution of the identity in momentum space, they satisfy the completeness relation
X
1= |, ; +inr h, ; +|nr + |, ; inr h, ; |nr . (16)
This expression and the normalization factors in equation (15) depend on the details
of the normalization of kinematical states in Fourier representation. One can always
make all these definitions mutually consistent.
The second possibility is a relativistic representation in terms of states
|, i |, ; +i + |, ; i , (17)
where
Z
dp ip
|, ; i = e (p ) |, p i , (18)
2|p |
P The name is inherited from the particle case, where it refers to the invariance properties of the
states under SO(1, 1) transformations. In the LQC case, due to the discrete nature of the basis, the
transformation properties under the same group are trickier at the quantum level. We maintain the
nomenclature nevertheless.
Two-point functions in (loop) quantum cosmology 8
which are not orthogonal at fixed . Choosing the indefinite normalization (13) and
the corresponding indefinite resolution of the identity in momentum space, these
states satisfy the completeness relation
X
1= |, ; +i h, ; +| + |, ; i h, ; | . (19)
It is immediate to see [20] that the corresponding inner product, in the relativistic
representation, is not positive definite (it is positive for the positive-frequency
sector, and negative for the negative-frequency one). This would suggest that
the corresponding quantum canonical theory is not well-defined. A similar issue
is present also in the full (formal) quantum geometrodynamics in WheelerDeWitt
form [19, 22]. One can obtain a positive definite inner product in the relativistic
representation, and in the presence of both sectors of solutions, by choosing the
normalization (13) instead of (12), so that the decomposition (19) is replaced by
X
1= |, ; +i h, ; +| |, ; i h, ; | . (20)
However, we will show that such non-standard resolution of the identity affects
the composition properties of the two-point function defining the physical inner
product.
Whether one restricts consideration to either positive- or negative-frequency sectors
of the Hilbert space, or works in the full Hilbert space. As we have just seen,
the inclusion of both sectors has important consequences already at the canonical
level. However, the presence of a complete set of Dirac observables induces a
superselection rule which allows one to pick only one type of modes. In particular,
taking the positive-frequency sector the three resolutions of the identity (16), (19)
and (20) collapse into two positive-definite expressions, one nonrelativistic,
1=
X
|, ; +inr h, ; +|nr , (21)
and one relativistic:
1=
X
|, ; +i h, ; +| . (22)
The two representations are different in many respect, but physically equivalent, as
they are related by a unitary transformation [19].+
The integration range in proper time (lapse) in the group averaging representation
of the two-point function, or, equivalently, the class of histories summed over in the
sum-over-histories formulation of the same. Integrals over the full real line give,
generically, solutions of the constraint equation, and thus true inner products for
the canonical theory; integrals of the positive semi-axis only give Green functions
(propagators) for the same constraint equation (i.e., solutions of the constraint
equation in the presence of a delta source).
+
This can be seen by interpreting the states in each sector as distributions over an auxiliary Hilbert
space S, and noticing that the definitions (15) and (18) involve different spaces S.
Two-point functions in (loop) quantum cosmology 9
We now proceed to define the various two-point functions for (loop) quantum
cosmology, depending on the above choices, and to give, for each of them, their covariant
(spin foam) representation in terms of sum-over-histories. This will also clarify their
mutual relations, and the origin of their differences. Having done so, we proceed to
derive the composition laws that the different two-point functions satisfy, and discuss
their physical implications.
Since one has a constrained system, the first quantity of interest is the physical inner
product between states, which is usually defined, following Dirac [23], starting from
generic (kinematical) states, labelled by unrelated and p , by inserting a projector onto
those states annihilated by the constraints. Such an inner product defines a solution for
the Hamiltonian constraint operator. It is the analogue of a transition amplitude for
systems with an external time variable.
A simple (but, in LQC, apparently unnoticed) fact is the following. For the relativistic
particle, the analogue of (23) is the Hadamard function, hence the suggestive notation
GH . On the other hand, the analogue of (24), where a factor of the frequency is included,
leads to a non-relativistic two-point function, namely the positive- or negative-frequency
NewtonWigner function if one restricts to positive or negative frequencies, or a sum
of the two if one includes both. We will show that both expressions (23) and (24)
admit exactly the same interpretation in quantum cosmology and obey the expected
composition laws. In particular, equation (24) is the sum of positive- and negative-
frequency NewtonWigner functions.
Two-point functions in (loop) quantum cosmology 10
We will see that the inclusion of the factor 2|p | into (24) changes the character of
the inner product thus defined from relativistic to non-relativistic. In particular, we will
show that the non-relativistic two-point function (27) satisfies a different composition
law than the Hadamard function (23). Therefore, both the relativistic particle analogy
and the composition laws satisfied by the different two-point functions are consistent
with equation (29). However, starting from the two distinct expressions for the Newton
Wigner function G NW , one can proceed as in [7, 8] and derive distinct vertex expansions
for the inner product in cosmology, which we will report below. This functional difference
may come as a surprise [7, 8], but is what one could expect immediately when looking
at the two expressions (24) and (27), one involving the operator in the exponentand
the other its square root H. Here, equation (29) shows the existence of a unique
NewtonWigner function in both the group averaging and deparametrized framework,
thus settling the issue.
N
X X N
X 1 NX
M 1 N
X 2 1
where [8]
A(f , M 1 , . . . , 1 , i ; ) = f M 1 . . . 2 1 1 i (39)
p n 1 p
eiwm wm
Y 1 k X
Qp .
(nk 1)! wk wk m=1 j=1 (wm wm wj wj )
k=1 j6=m
Here,
k l hk ||l i (40)
are the matrix elements of , and in the second line the p distinct values appearing in
(i , 1 , . . . , M 1 , f ) are denoted by w1 , . . . , wp with multiplicities ni so that n1 + . . . +
np = M + 1. It is this step that can only be done formally if the variable has a
continuous range. Our analysis, which is rigorous for LQC, will be understood to hold
for WheelerDeWitt theory in this formal sense.
In order to give the complete vertex expansion of the two-point functions, one has
to insert the above expression in equations (31) and (33), perform the integration over
, and finally the one over p .
This is, of course, the continuum limit of the discretization of the -interval, which is the counterpart
of the continuum limit of the (proper) time interval for the relativistic particle. In the spin foam analogy,
it would correspond, in the full theory, to the infinite-refinement limit of the triangulation on which the
spin foam is based. Defining this limit in the full spin foam case, or in simplicial quantum gravity, is of
course a highly non-trivial task. Equally non-trivial would be to show that this limit can equivalently
written as a kind of sum over discretizations of the same topology (here, the line interval). In the
(L)QC case this is much easier, and in fact the existence of a vertex expansion for the two-point
function, defined in the same limit, is a proof of such equivalence, at least for this simple case.
Two-point functions in (loop) quantum cosmology 13
p
nk 1 X p
cos( wm wm )
1
p Qp . (41)
wk wk m=1
wm wm j=1 (wm wm wj wj )
j6=m
The reader will observe the appearance of a cosine typical of the Hadamard two-point
function.
Integrating over in (33), one notices that the delta distributions (p2 wm wm )
appearing under the p integral are replaced by equation (30) with wm wm , so
that restricting to positive or negative p corresponds to picking out positive- and
negative-frequency solutions to the square root of the constraint (7) as one does for
the NewtonWigner function of the relativistic particle [20].
Integrating over both positive and negative p one finds [7]
X X
GNW (f , f ; i , i ) = f M 1 . . . 2 1 1 i
M =0 M 1 ,...,1
m 6=m+1
p p
nk 1 X p
2 cos( wm wm )
Y 1
Qp . (42)
k=1
(nk 1)! wk wk m=1 j=1 (wm wm wj wj )
j6=m
Once more, one recognizes the cosine factor typical of symmetrized (over frequency
sectors) solutions of the Hamiltonian constraint, and the different functional dependence
on the energy H resulting from the non-relativistic representation of the canonical
states.
Restricting to positive p gives
X X
G+ (
NW f , f ; i , i ) = f M 1 . . . 2 1 1 i
M =0 M 1 ,...,1
m 6=m+1
p p
nk 1 X
i
e wm wm
Y 1
Qp . (43)
k=1
(nk 1)! wk wk m=1 j=1 (wm wm wj wj )
j6=m
This procedure provides, in the simplified setting of loop quantum cosmology, an explicit
spin-foam type expansion of the physical inner product GH or GNW , thus strengthening
(at least on a formal level and in this ultra-minimalistic model) the relation between
canonical and covariant approaches.
From the expression for the deparametrized inner product (27), it is immediate
that it can be expressed in the form
X X
+
GNW (f , f ; i , i ) = A(f , M 1 , . . . , 1 , i ; )
M 1 ,...,1 ,H
M =0
m 6=m+1
Two-point functions in (loop) quantum cosmology 14
X X
= Hf M 1 . . . H2 1 H1 i
M =0 M 1 ,...,1
m 6=m+1
p p
nk 1 X
eiHwm wm
Y 1
Qp , (44)
k=1
(nk 1)! Hwk wk m=1 j=1 (Hwm wm Hwj wj )
j6=m
where
Hk l hk | |l i . (45)
As noticed in [7], group averaging and the deparametrized framework lead to distinct
vertex expansions. However, as we have already discussed above, the two expansions
only correspond to different perturbative expressions of the same two-point function.
This can be recognized working at the non-perturbative level, i.e., before operating the
vertex expansion, or else (if one is able to) after this has been performed, by re-summing
the whole series.
so that in the gravitational context one only integrates over positive lapse. This suggests
to define the correspondent quantity in quantum cosmology as
Z+
iGF (f , f ; i , i ) d AH (; )A (f , i ; ) , (47)
0
where AH is given in equation (32) (no factor 2|p |). Let us stress that the restriction in
the Schwinger parameter , the (L)QC analogue of the proper time for the relativistic
particle [20] and of the lapse for quantum gravity [10] is not a restriction in any time
variable, e.g. , and it is thus fully consistent with (symmetry-reduced) background
independence [11]. As is usual for the Feynman propagator, one has to choose a contour
Two-point functions in (loop) quantum cosmology 15
in the complex p plane since now the integrand has poles. We follow the causal +i
prescription:
p nk 1
X X Y 1
iGF (f , f ; i , i ) = f M 1 . . . 2 1 1 i
M =0 M 1 ,...,1 k=1
(nk 1)! wk wk
m 6=m+1
+
p Z ip
X dp i e .
2
Qp (48)
m=1
2 p w m w m + i j=1 ( w m w m w w
j j
)
j6=m
p
nk 1 X
wm wm
ei () + ei wm wm ()
.
2 wm wm pj=1 (wm wm wj wj )
p
wk wk
Q
m=1
j6=m
(49)
Because of the open contour (or equivalently, because of the restriction to the positive
semi-axis in the integration) the two-point function GF is not a solution of the
Hamiltonian constraint equation; thus it does not define an inner product for the
canonical theory. However, it is a proper Green function (propagator) for the same
equation, i.e., it defines a transition amplitude taking into account the relative ordering
in the time variables labelling the states (and thus defining a background independent
notion of in and out).
We may also determine Wightman functions G as in [20] by choosing a closed
p
contour around only one of the poles p = wm wm . This obviously corresponds to
restricting to positive- or negative-frequency sectors. One finds:
p
X X Y 1
G (f , f ; i , i ) = f M 1 . . . 2 1 1 i
M =0 M 1 ,...,1
k=1
(nk 1)!
m 6=m+1
p
nk 1 X
ei wm wm
p Qp . (50)
wk wk m=1
2 w m w m j=1 ( w m w m w j w j
)
j6=m
The choice of a closed contour also implies that the two Wightman functions, like the
Hadamard two-point function, satisfy the Hamiltonian constraint for (loop) quantum
cosmology, and thus define appropriate physical inner products for the two sectors
corresponding to positive- and negative-frequency, in the relativistic representation.
One can easily show that the two-point functions we have defined satisfy the
relations familiar from the relativistic particle case:
GH (f , f ; i , i ) = G+ (f , f ; i , i ) + G (f , f ; i , i ) , (51)
iGF (f , f ; i , i ) = G+ (f , f ; i , i )(f i ) + G (f , f ; i , i )(i f ) . (52)
Two-point functions in (loop) quantum cosmology 16
p
nk 1 X p
sin wm wm
, (54)
wm wm pj=1 (wm wm wj wj )
p
wk wk
Q
m=1
j6=m
where the products of differences of matrix elements in equations (41) and (49) have
disappeared. The integrand now is local, inasmuch as it features a product over
amplitudes of elements forming the particular history of the Universe one considers.
Since, in contrast to [9], we have included the scalar field , one is left with an
p
integration over p where one picks up residues at the poles p = wi wi , i = 1, . . . , p,
p
which are of order ni . It is not too difficult to see that the residue at p = + wm wm
is
" p #
1 nm 1 1 i eip (p wm wm )nm
(nm 1)! pnm 1 2
QM 2
k=0 (p k k )
p wm wm
nm 1 ip
1 1 ie
=
(nm 1)! pnm 1 2 (p + wm wm )nm pj=1 (p2 wj wj )nj p wm wm
p Q
j6=m
p nk 1
Y 1 1
=
k=1
(nk 1)! wk wk (nm 1)!
k6=m
nm 1 ip
1 ie
nm 1
p Qp
p 2 (p + wm wm ) n m 2
j=1 (p wj wj ) p wm wm
j6=m
p nk 1
i wm wm
Y 1 1 p ie
= , (57)
(nk 1)! wk wk 2 2 wm wm pj=1 (wm wm wj wj )
Q
k=1 j6=m
p
with a similar expression for p = wm wm , so that one reproduces the non-local
result (49). Therefore, while providing an interesting new viewpoint on the inner product
for the purely gravitational sector, the findings of [9] do not modify our analysis: In
the presence of free scalar matter, quantum cosmological two-point functions are always
non-local, in the sense specified above.
4. Composition laws
Having computed the analogue for (loop) quantum cosmology of all two-point functions
known for the relativistic particle, we are now in a position to verify the composition
laws they satisfy, and to interpret them in the cosmological setting. Such composition
laws are in fact crucial for using these two-point functions as canonical inner products
(or transition amplitudes). We will derive these laws using their sum-over-histories
(or vertex expansion) formulation. It goes without saying that the results match the
expectations coming from the particle analogy [20].
Two-point functions in (loop) quantum cosmology 18
This rule incorporates the requirement that any path from (x , t ) to (x , t ) that only
goes forward in time will pass through precisely one x at a specified intermediate time
t < t < t . Note that this identity is, as we have anticipated above, independent of the
form of H (or, equivalently, of ).
Using the representation (33) of the physical inner product in the timeless
framework, a composition law of the form (59) would amount to
Z+ XZ Z
+ +
?
d ANW (f i ; )A (f , i ; ) = d d ANW (f ; )ANW ( i ; )
A (f , ; )A (, i ; ) . (60)
From the definition A (f , i ; ) hf |ei |i i, it is again immediate that
X X
A (f , ; )A (, i ; ) = hf |ei |ih|ei |i i = A (f , i ; + ) . (61)
Clearly, this composition rule extends to the case of WheelerDeWitt theory where
has continuous range. The calculations in this section can be extended to the continuous
case if one accepts the summation over M to be of formal nature.
Further evaluating the right-hand side of (60), we have
Z+ Z+
r.h.s. of (60) = d d ANW (f ; )ANW ( i ; )A (f , i ; + )
dp dp
X X Z
= f M 1 . . . 2 1 1 i d d (4|p p |)
M =0 M 1 ,...,1 2 2
m 6=m+1
p
ip2 i p 2 ip (f ) ip (i )
Y 1
e e e e
(nk 1)!
k=1
p
nk 1 X i(+ )wm wm
e
Qp , (62)
wk wk m=1 j=1 (wm wm wj wj )
j6=m
Two-point functions in (loop) quantum cosmology 19
X X
f M 1 . . . 2 1 1 i
M =0 M 1 ,...,1
m 6=m+1
p p
nk 1 X p p
4 cos[ wm wm (f )] cos[ wm wm ( i )]
Y 1
Qp , (63)
k=1
(nk 1)! wk wk m=1 j=1 (wm wm wj wj )
j6=m
nk 1 Xp
ei wm wm (f ) ei wm wm (i )
Qp
wk wk m=1 j=1 (wm wm wj wj )
j6=m
= G+
NW (f , f ; i , i ) , (64)
confirming that G+NW as given in (43) does obey the non-relativistic composition law
(59). The same result will hold for the negative-frequency function G
NW . On the other
hand, the composition law is spoiled for GNW due to the interference of positive- and
negative-frequency sectors.
p p
nk 1 X
ei(+ )wm wm
1Y
f M 1 . . . 2 1 1 i Qp
k=1
(nk 1)! wk wk m=1 j=1 (wm wm wj wj )
j6=m
p
X X Y 1
= f M 1 . . . 2 1 1 i
M =0 M 1 ,...,1 k=1
(nk 1)!
m 6=m+1
nk 1 Xp
i wm wm i wm wm
e e
p Qp
wk wk m=1
2i w m w m j=1 ( w m w m w j w j
)
j6=m
X X
= f M 1 . . . 2 1 1 i
M =0 M 1 ,...,1
m 6=m+1
p p
nk 1 X p
sin( wm wm )
Y 1
wm wm pj=1 (wm wm wj wj )
p
(nk 1)! wk wk
Q
k=1 m=1
j6=m
= Gc (f , f ; i , i ) , (68)
where Gc is the causal two-point function (54). This is precisely the composition law
one has for the Hadamard function of the relativistic particle [20],
Z
Gc (x |x ) = d GH (x |x) GH (x|x ) . (69)
The reason why the causal two-point function Gc and not GH appears on the left-hand
side is again interference between positive- and negative-frequency sectors.
We conclude that the two definitions used for the inner product in cosmology, the
deparametrized definition (27) and the timeless definition (23), indeed satisfy different
composition laws, one being non-relativistic (after singling out positive or negative
frequencies) and the other relativistic and involving the causal two-point function (54).
Z+
X X dp dp ip (f ) ip (i )
= i f M 1 . . . 2 1 1 i e e
M =0 M 1 ,...,1 2 2
m 6=m+1
p nk 1
Y 1
ip + ip
k=1
(nk 1)! wk wk
p 2 1 2 1
X (p wm wm + i) (p wm wm + i)
Qp
m=1 j=1 ( wm wm wj wj )
j6=m
p nk 1
1X X Y 1
= . . . 2 1 1 i
2 M =0 M 1 ,...,1 f M 1 k=1
(nk 1)! wk wk
m 6=m+1
p
wm wm i wm wm
X ei (f )( i ) e (i )( f )
p Qp
m=1
wm wm j=1 (wm wm wj wj )
j6=m
(
iGF (f , f ; i , i ) , i < < f ,
= (72)
iGF (f , f ; i , i ) , f < < i ,
in agreement with equation (70).
Choosing the closed contour used to define Wightman functions G in (50), one
also confirms
Z
G (x |x ) = i d G (x |x) G (x|x ) .
(73)
As we mentioned, the causal two-point function does not have a vertex expansion, but
from its definition (53) one sees that it correctly obeys the same relativistic composition
law.
Two-point functions in (loop) quantum cosmology 22
5. Discussion
In this paper, we have catalogued the possible two-point functions for (loop) quantum
cosmology, for a single homogeneous universe in the presence of a free massless scalar
field, relying heavily on the analogy of this system with the free relativistic particle. For
each of them we have constructed a covariant sum-over-histories representation, in the
form of a vertex expansion in the sense of [7]. Finally, we have derived the composition
laws that such two-point functions satisfy. Our results hold for homogeneous and
isotropic LQC in a flat universe as well as for the traditional WheelerDeWitt quantum
cosmology, although only in a more formal sense (due to the continuous nature of volume
eigenvalues). Our results hopefully clarify (or at least provide a basis for clarifying)
some issues that have been recently debated in the LQC literature.
The features that characterize the different two-point functions are:
a) the sector of the canonical Hilbert space they refer to (positive- or negative-
frequency, or both),
b) the relativistic or non-relativistic representation chosen for the quantum states,
c) their composition law,
d) their nature as canonical inner products or propagators (Greens functions).
If one restricts to a single sector of solutions, say positive-frequency, then both
the relativistic (positive) Wightman function (50) and the non-relativistic (positive)
NewtonWigner function (27) provide a satisfactory and physically equivalent definition
of the canonical inner product. In fact, they are positive definite and satisfy an
appropriate composition law (of relativistic and non-relativistic type, respectively). The
Feynman propagator (49), reduced in this case to the retarded propagator, provides
instead the Green function that propagates solutions of the constraint equation into
other solutions. Notice that all these functions are complex-valued, and transform into
their complex conjugate under switch of the ordering of their two arguments.
The choice between a relativistic and a non-relativistic representation, when
available, is then purely a matter of practical convenience, and it depends on the physical
issue one is set to address.
The NewtonWigner function G+ NW , for example, was studied in connection to
the problem of localization for a quantum mechanical particle. Its associated inner
product, in fact, allows one to consider wavefunctions whose modulus squared gives
the probability to find a particle in a volume about a certain spatial point [26]. This
result, however, depends on the choice of a frame because the NewtonWigner function
Indeed, the same vertex expansion could be derived for the relativistic particle in flat space, again
with a similar purely formal nature.
Two-point functions in (loop) quantum cosmology 23
are allowed to fluctuate. In fact, when considering relativistic quantum particles, one
notices that conservation of the particle number is incompatible with Lorentz symmetry
and the field theory formalism is naturally invoked. Then, the relativistic scalar
product for the field is not only well-defined but also uniquely so. Even though the
cosmological or quantum gravity analogue of this symmetry argument is unclear to us,
at present, this would suggest that also in the quantum cosmology case, as soon as more
gravitational and matter degrees of freedom enter the picture (including anisotropies and
inhomogeneities), and thus one moves closer to the full theory, the natural formalism is
that of a field theory on (symmetry-reduced) superspace, a third quantization [30]. This
had been considered also in the full theory [22, 29]. This formalism would accommodate
also another possible generalization of the framework considered in this paper, that is
the possibility of topology change. This entails the creation/annihilation of (possibly
homogeneous and isotropic) universes, quanta of the superspace field naturally carrying
both orientations (in analogy with the particle/antiparticle distinction). This framework
would represent a rather important generalization of the quantum cosmology setting
(possibly useful in addressing issues like the origin of the cosmological constant [30, 31]).
Any further discussion of this possible generalization would go beyond the scope of this
paper. However, it is natural to draw the lesson of the relativistic many-particle system
and ask ourselves the fate of the cosmological two-point functions in a multiverse picture.
In this case, the single-particle/single-universe two-point functions do not need to satisfy
any positive definiteness requirement (since they do not play the role of inner products),
the non-relativistic representation of canonical states would simply not be available, and
we would expect the Feynman propagator to be elected over the other choices by the
requirements of causal propagation (positive proper time/lapse, ordering of arguments)
and correct composition law. The causal two-point function would also play the useful
role of the mean value of the commutator of (super-)field operators in the Fock vacuum.
Thus it would encode correlations on superspace, rather than being the propagator of
particle/single-universe states. Work on such third quantized (loop) quantum cosmology
is in progress [32].
To conclude, from our results in the symmetry-reduced setting let us collect some
lessons and suggestions for quantum gravity and for the group field theory (GFT)/spin
foam approach in particular. Clearly, the step from symmetry-reduced models to full
theory is huge, both technically and conceptually. Therefore, any insight we try to
import from the former to the latter should be taken cum grano salis and treated only
as a suggestion to be explored further.
First, we have seen how important the choice of relativistic versus non-relativistic
representation is for the quantum theory. In the covariant setting, the analogue of
the spin foam constructions, this is reflected in the form of boundary states appearing
in the sum-over-histories and in the composition law chosen for the same. We have
pointed out that the non-relativistic formalism, although physically equivalent to
the relativistic one, is only available in very special cases, and thus a relativistic
Two-point functions in (loop) quantum cosmology 25
representation is to be preferred. At the covariant level, this implies that one should
try to construct relativistic transition amplitudes/inner products which compose by
means of a relativistic composition law. The relativistic nature is characterized by the
presence of embedding information, i.e., of the normal to the canonical surface, in
both the definition of the decomposition of the identity and the composition law for
transition amplitudes/inner products. This is required for the correct implementation
of covariance properties as well as for the correct composition of boundary data in
the two-point functions. It is natural to conjecture that something similar happens
in the full LQG theory and in the related spin foam models, and that the needed
embedding information is to be represented by similar extrinsic data, in particular by
the normal vector to the canonical hypersurface, in turn characterizing the extrinsic,
i.e., boost component of the gravitational connection. A generalization of LQG states
to include such data has been proposed [33, 34, 35, 36] and has proven crucial also in
covariant spin foam constructions [37, 38, 39, 40, 41]. So our results seem to support
such generalization.
Second, we have pointed out that the presence of matter fields always leads to a
non-local vertex expansion. This non-locality is defined in contrast to the factorized
form, in terms of single simplex contributions, that usual spin foam models take [5, 28].
In turn, this local form translates naturally in (or can be seen as a direct consequence
of) the underlying GFT formulation of the same models [48]. There are several attitudes
one can take with respect to this fact: a) One could expect a similar loss of locality in
spin foam models coupled to matter, and thus the impossibility of a corresponding GFT
formulation; b) one can consider this non-locality as an artefact of the symmetry-reduced
context (also considering that symmetry reduction is itself a rather non-local procedure)
or of the specific type of matter considered. The second attitude is based also on the fact
that the locality of spin foam models and GFTs has a clear spacetime interpretation, in
that spin foam vertices or their dual simplices represent regions or events in a discrete
spacetime, while the vertex expansion in cosmology lacks such interpretation exactly
because of the prior symmetry reduction. The available examples of spin foam models
and GFTs for gravity coupled to matter [42, 43, 45, 44, 46, 47] seem to support the
second possibility, although the issue should be considered as still open.
Third, we have stressed that the crucial difference in constructing and
understanding two-point functions in quantum cosmology is whether one restricts to
positive or negative-frequency solutions or keeps both, and that the analogue of this
frequency condition in full quantum gravity would be the spacetime orientation. In spin
foam models and GFTs, one has to specify orientation data for all the discrete elements
of the spacetime substratum used in the quantum theory, i.e., edges and faces of the spin
foam 2-complex, or simplices and sub-simplices of the dual triangulation. Moreover,
opposite orientation data for all these elements have to be included, for consistency
of composition properties. This seems to suggest that no analogue of the frequency
restriction is available in the full theory (as in the case of general inhomogeneous or
anisotropic cosmologies, or for generic matter fields). Of course, one can still consider
Two-point functions in (loop) quantum cosmology 26
amplitudes that either are symmetric under spacetime orientation (parity or time)
reversal (in analogy with the Hadamard function) or transform into their own complex
conjugate under reversal (in analogy with Feynman or causal two-point functions).
Known spin foam models [5], based on constraining quantum BF theory, seem to be
naturally of the symmetric type. On the other hand, constructions of non-symmetric
(complex) and causally restricted transition amplitudes for LQG states have been
considered in the spin foam framework [12, 13, 14], and could be useful to investigate
this issue further. In fact, quantum BF theory itself gives rise to symmetric (real)
transition amplitudes that can be consistently interpreted as covariant definitions of
the inner product of topological gravity. In contrast, causally restricted models are
associated with true propagators or transition amplitudes that have an in-built notion
of in and out states consistent with background independence (inasmuch as this notion
is encoded in orientation data only and no time variable is needed). The situation in
the full theory seems thus to be at odds with the results in the quantum cosmology
setting: while the latter would suggest that a symmetric transition amplitude that
is both positive definite and includes both sectors of solutions would fail to compose
properly, this seems not to be the case for BF spin foam models. On the other hand,
one could be tempted to attribute this situation to the triviality, in terms of dynamical
degrees of freedom, of BF theory itself. However, a similar orientation independence is
present in the BarrettCrane model which, if coming from a BF-type construction, is
not a topological model. Similarly puzzling is the situation, concerning this orientation
business, from the GFT perspective. GFTs are third quantized frameworks [48] where
the fundamental field is defined on simplicial superspace, the space of geometries for a
single simplex (a building block of a discrete space), and in which creation/annihilation
of simplices and topology change are naturally described. Hence, one would expect that
the corresponding Feynman amplitudes be naturally complex and interpreted as built
out of propagators, rather than solutions, to the canonical constraints. This is, again,
not the case for current GFT models, that generate the symmetric spin foam amplitudes
we have just discussed. On the other hand, it would be the case for any effective GFT
dynamics around non-trivial background configurations [49, 50, 51], which naturally
translates into non-trivial (and singular) free propagators, analogous to the Feynman
two-point function of ordinary quantum field theory.
Much remains to be understood about these issues, and the simplified quantum
cosmology setting could be a very useful ground for clarifying them, leading to a better
development and understanding of a full quantum theory of gravity.
Acknowledgements
This work is funded by the A. Von Humboldt Stiftung, through a Sofja Kovalevskaja
Prize, which is gratefully acknowledged. SG also acknowledges support from EPSRC,
Trinity College, Cambridge, and the Max Planck Society.
That is, amplitudes obtained by restricting the range of proper time/lapse integration.
Two-point functions in (loop) quantum cosmology 27
References
[1] Wiltshire D L 1996 An introduction to quantum cosmology Cosmology: The Physics of the
Universe ed B Robson et al (Singapore: World Scientific) (Preprint arXiv:gr-qc/0101003).
[2] Bojowald M 2008 Loop quantum cosmology Living Rev. Relativity 11 4
[http://www.livingreviews.org/lrr-2008-4]
[3] Thiemann T 2007 Modern canonical quantum General Relativity (Cambridge: Cambridge
University Press)
[4] Rovelli C 2006 Quantum Gravity (Cambridge: Cambridge University Press)
[5] Perez A 2003 Spin foam models for quantum gravity Class. Quantum Gravity 20 R43-R104
(Preprint gr-qc/0301113)
[6] Oriti D 2003 Spin foam models of quantum spacetime (Ph.D. thesis, University of Cambridge)
(Preprint gr-qc/0311066)
[7] Ashtekar A, Campiglia M and Henderson A 2009 Loop quantum cosmology and spin foams Phys.
Lett. B 681 347-52 (Preprint 0909.4221)
[8] Ashtekar A, Campiglia M and Henderson A 2010 Casting loop quantum cosmology in the spin
foam paradigm Class. Quantum Gravity 27 135020 (Preprint 1001.5147)
[9] Henderson A, Rovelli C, Vidotto F and Wilson-Ewing E 2010 Local spinfoam expansion in loop
quantum cosmology Preprint 1010.0502
[10] Teitelboim C 1982 Quantum mechanics of the gravitational field Phys. Rev. D 25 3159-79
[11] Halliwell J J and Hartle J B 1991 Wave functions constructed from an invariant sum over histories
satisfy constraints Phys. Rev. D 43 1170-94
[12] Livine E R and Oriti D 2003 Implementing causality in the spin foam quantum geometry Nucl.
Phys. B 663 231-79 (Preprint gr-qc/0210064)
[13] Oriti D 2005 The Feynman propagator for spin foam quantum gravity Phys. Rev. Lett. 94 111301
(Preprint gr-qc/0410134)
[14] Oriti D and Tlas T 2006 Causality and matter propagation in 3d spin foam quantum gravity Phys.
Rev. D 74 104021 (Preprint gr-qc/0608116)
[15] Hartle J B and Kuchar K V 1986 Path integrals in parametrized theories: The Free relativistic
particle Phys. Rev. D 34 2323-31
[16] Baierlein R F, Sharp D H and Wheeler J A 1962 Three-dimensional geometry as carrier of
information about time Phys. Rev. 126 1864-5
[17] Bohr H 1925 Zur Theorie der fast periodischen Funktionen I. Eine Verallgemeinerung der Theorie
der Fourierreihen Acta Mathematica 45 29-127
[18] Ashtekar A, Corichi A and Singh P 2008 Robustness of key features of loop quantum cosmology
Phys. Rev. D 77 024046 (Preprint 0710.3565)
[19] Ashtekar A, Pawlowski T and Singh P 2006 Quantum nature of the big bang: An analytical and
numerical investigation. I Phys. Rev. D 73 124038 (Preprint gr-qc/0604013)
[20] Halliwell J J and Ortiz M E 1993 Sum over histories origin of the composition laws of relativistic
quantum mechanics Phys. Rev. D 48 748-68 (Preprint gr-qc/9211004)
[21] Halliwell J J and Thorwart J 2001 Decoherent histories analysis of the relativistic particle Phys.
Rev. D 64 (2001) 124018 (Preprint gr-qc/0106095)
[22] Kuchar K V 1991 Time and interpretations of quantum gravity Winnipeg 1991, Proceedings,
General relativity and relativistic astrophysics 211-314
[23] Dirac P A M 1964 Lectures on Quantum Mechanics (New York: Yeshiva University, Phys. Dept.)
[24] Giulini D 2000 Group averaging and refined algebraic quantization Nucl. Phys. Proc. Suppl. 88
385-8 (Preprint gr-qc/0003040)
[25] Newton T D and Wigner E P 1949 Localized states for elementary systems Rev. Mod. Phys. 21
400-6
[26] Ruijsenaars S N M 1981 On Newton-Wigner localization and superluminal propagation speeds
Annals Phys. 137 33-43
Two-point functions in (loop) quantum cosmology 28
[27] Redmount I H and Suen W-M 1993 Path integration in relativistic quantum mechanics Int. J.
Mod. Phys. A 8 1629-36 (Preprint gr-qc/9210019)
[28] Baez J C 2000 An Introduction to spin foam models of quantum gravity and BF theory Lect. Notes
Phys. 543 25-94 (Preprint gr-qc/9905087)
[29] Isham C 1992 Canonical quantum gravity and the problem of time GIFT Seminar on Theoretical
Physics, Salamanca, Spain 157-287 (Preprint gr-qc/9210011)
[30] Giddings S B and Strominger A 1989 Baby universes, third quantization and the cosmological
constant Nucl. Phys. B 321 481-508; McGuigan M 1988 Third quantization and the Wheeler-
DeWitt equation Phys. Rev. D 38 3031-51
[31] Banks T 1988 Prolegomena to a theory of bifurcating universesNucl. Phys. B 309 493-512
[32] Calcagni G, Gielen S and Oriti D 2010 in preparation
[33] Alexandrov S and Livine E R 2003 SU (2) loop quantum gravity seen from covariant theory Phys.
Rev. D 67 044009 (Preprint gr-qc/0209105)
[34] Livine E R 2002 Projected spin networks for Lorentz connection: Linking spin foams and loop
gravity Class. Quantum Gravity 19 5525-42 (Preprint gr-qc/0207084)
[35] Alexandrov S 2010 The new vertices and canonical quantization Phys. Rev. D 82 024024 (Preprint
1004.2260)
[36] Dupuis M and Livine E R 2010 Lifting SU (2) spin networks to projected spin networks Phys. Rev.
D 82 064044 (Preprint 1008.4093)
[37] Freidel L and Krasnov K 2008 A new spin foam model for 4d gravity Class. Quantum Gravity 25
125018 (Preprint 0708.1595)
[38] Livine E R and Speziale S 2008 Solving the simplicity constraints for spinfoam quantum gravity
Europhys. Lett. 81 50004 (Preprint 0708.1915)
[39] Gielen S and Oriti D 2010 Classical general relativity as BF-Plebanski theory with linear
constraints Class. Quantum Gravity 27 185017 (Preprint 1004.5371)
[40] Baratin A and Oriti D 2010 Group field theory with non-commutative metric variables Phys. Rev.
Lett. to appear (Preprint 1002.4723)
[41] Baratin A and Oriti D 2010 The Barrett-Crane model revisited to appear
[42] Oriti D and Pfeiffer H 2002 A spin foam model for pure gauge theory coupled to quantum gravity
Phys. Rev. D 66 124010 (Preprint gr-qc/0207041)
[43] Speziale S 2007 Coupling gauge theory to spinfoam 3D quantum gravity Class. Quantum Gravity
24 5139-59 (Preprint 0706.1534)
[44] Fairbairn W 2007 Fermions in three-dimensional spinfoam quantum gravity Gen. Relativity
Gravitation 39 427-76 (Preprint gr-qc/0609040)
[45] Louapre D and Freidel L 2004 Ponzano-Regge model revisited: I. Gauge fixing, observables and
interacting spinning particles Class. Quantum Gravity 21 5685-726 (Preprint hep-th/0401076)
[46] Oriti D and Ryan J 2006 Group field theory formulation of 3D quantum gravity coupled to matter
fields Class. Quantum Gravity 23 6543-75 (Preprint gr-qc/0602010)
[47] Dowdall R J 2009 Wilson loops, geometric operators and fermions in 3d group field theory 2009
Preprint 0911.2391; Dowdall R J and Fairbairn W J 2010 Observables in 3d spinfoam quantum
gravity with fermions Gen. Rel. Grav. to appear Preprint 1003.1847
[48] Oriti D 2009 The group field theory approach to quantum gravity Approaches to quantum gravity
ed D Oriti (Cambridge: Cambridge University Press) 310-31 (Preprint gr-qc/0607032); Oriti D
2008 Group field theory as a quantum field theory of simplicial geometry Quantum gravity ed B
Fauser et al 101-26 (Preprint gr-qc/0512103)
[49] Fairbairn W J and Livine E R 2007 3D spinfoam quantum gravity: matter as a phase of the group
field theory Class. Quantum Gravity 24 5277-97 (Preprint gr-qc/0702125)
[50] Girelli F, Livine E R and Oriti D 2010 Four-dimensional deformed special relativity from group
field theories Phys. Rev. D 81 024015 (Preprint 0903.3475)
[51] Oriti D and Sindoni L 2010 Towards classical geometrodynamics from group field theory
hydrodynamics Preprint 1010.5149