Gravity, Geometry and the Quantum
Abhay Ashtekar1,2
arXiv:gr-qc/0605011v2 14 Jul 2006
Institute for Gravitational Physics and Geometry
Physics Department, Penn State, University Park, PA 16802-6300
2 Institute for Theoretical Physics, University of Utrecht,
Princetonplein5, 3584 CC Utrecht, The Netherlands
After a brief introduction, basic ideas of the quantum Riemannian geometry underlying loop quantum gravity are summarized. To illustrate physical ramifications
of quantum geometry, the framework is then applied to homogeneous isotropic cosmology. Quantum geometry effects are shown to replace the big bang by a big
bounce. Thus, quantum physics does not stop at the big-bang singularity. Rather
there is a pre-big-bang branch joined to the current post-big-bang branch by a ‘quantum bridge’. Furthermore, thanks to the background independence of loop quantum
gravity, evolution is deterministic across the bridge.
I.
INTRODUCTION
General relativity and quantum theory are among the greatest intellectual achievements
of the 20th century. Each of them has profoundly altered the conceptual fabric that underlies
our understanding of the physical world. Furthermore, each has been successful in describing
the physical phenomena in its own domain to an astonishing degree of accuracy. And
yet, they offer us strikingly different pictures of physical reality. Our past experience in
physics tells us that these two pictures must be approximations, special cases that arise
as appropriate limits of a single, universal theory. That theory must therefore represent a
synthesis of general relativity and quantum mechanics. This would be the quantum theory
of gravity that we invoke when faced with phenomena, such as the big bang and the final
state of black holes, where the worlds of general relativity and quantum mechanics must
unavoidably meet.
Remarkably, the necessity of a quantum theory of gravity was pointed out by Einstein
already in 1916. In a paper in the Preussische Akademie Sitzungsberichte he wrote:
Nevertheless, due to the inneratomic movement of electrons, atoms would have
to radiate not only electromagnetic but also gravitational energy, if only in tiny
amounts. As this is hardly true in Nature, it appears that quantum theory would
have to modify not only Maxwellian electrodynamics but also the new theory of
gravitation.
Ninety years later, our understanding of the physical world is vastly richer but a fully
satisfactory unification of general relativity with quantum physics still eludes us. Indeed,
the problem has now moved to the center-stage of fundamental physics. (For a brief historical
account of the evolution of ideas see, e.g., [1].)
2
A key reason why the issue is still open is the lack of experimental data with direct bearing
on quantum gravity. As a result, research is necessarily driven by theoretical insights on
what the key issues are and what will ‘take care of itself’ once this core is understood.
As a consequence, there are distinct starting points which seem natural. Such diversity
is not unique to this problem. However, for other fundamental forces we have had clearcut experiments to weed-out ideas which, in spite of their theoretical appeal, fail to be
realized in Nature. We do not have this luxury in quantum gravity. But then, in absence
of strong experimental constraints, one would expect a rich variety of internally consistent
theories. Why is it then that we do not have a single one? The reason, I believe, lies the
deep conceptual difference between the description of gravity in general relativity and that
of non-gravitational forces in other fundamental theories. In those theories, space-time is
given a priori, serving as an inert background, a stage on which the drama of evolution
unfolds. General relativity, on the other hand, is not only a theory of gravity, it is also a
theory of space-time structure. Indeed, in general relativity, gravity is encoded in the very
geometry of space-time. Therefore, a quantum theory of gravity has to simultaneously bring
together gravity, geometry and the quantum. This is a band new adventure and our past
experience with other forces can not serve as a reliable guide.
Loop quantum gravity (LQG) is an approach that attempts to face this challenge squarely
(for details, see, e.g., [2, 3, 4]). Recall that Riemannian geometry provides the appropriate
mathematical language to formulate the physical, kinematical notions as well as the final
dynamical equations of any classical theory of relativistic gravity. This role is now assumed
by quantum Riemannian geometry. Thus, in LQG both matter and geometry are quantum
mechanical ‘from birth’.
In the classical domain, general relativity stands out as the best available theory of
gravity. Therefore, it is natural to ask: Does quantum general relativity, coupled to suitable
matter (or supergravity, its supersymmetric generalization) exist as consistent theories nonperturbatively? In particle physics circles the answer is often assumed to be in the negative,
not because there is concrete evidence which rules out this possibility, but because of the
analogy to the theory of weak interactions. There, one first had a 4-point interaction model
due to Fermi which works quite well at low energies but which fails to be renormalizable.
Progress occurred not by looking for non-perturbative formulations of the Fermi model but
by replacing the model by the Glashow-Salam-Weinberg renormalizable theory of electroweak interactions, in which the 4-point interaction is replaced by W ± and Z propagators.
It is often assumed that perturbative non-renormalizability of quantum general relativity
points in a similar direction. However this argument overlooks a crucial and qualitatively
new element of general relativity. Perturbative treatments pre-suppose that space-time is a
smooth continuum at all scales of interest to physics under consideration. This assumption
is safe for weak interactions. In the gravitational case, on the other hand, the scale of
interest is the Planck length and there is no physical basis to pre-suppose that the continuum
approximation should be valid down to that scale. The failure of the standard perturbative
treatments may largely be due to this grossly incorrect assumption and a non-perturbative
treatment which correctly incorporates the physical micro-structure of geometry may well
be free of these inconsistencies.
Are there any situations, outside LQG, where such physical expectations are borne out by
3
detailed mathematics? The answer is in the affirmative. There exist quantum field theories
(such as the Gross-Neveu model in three dimensions) in which the standard perturbation
expansion is not renormalizable although the theory is exactly soluble! Failure of the standard perturbation expansion can occur because one insists on perturbing around the trivial,
Gaussian point rather than the more physical, non-trivial fixed point of the renormalization
group flow. Interestingly, thanks to the recent work by Lauscher, Reuter, Percacci, Perini
and others, there is now growing evidence that situation may be similar with general relativity (see [5] and references therein). Impressive calculations have shown that pure Einstein
theory may also admit a non-trivial fixed point. Furthermore, the requirement that the fixed
point should continue to exist in presence of matter constrains the couplings in physically
interesting ways [6].
Let me conclude this introduction with an important caveat. Suppose one manages
to establish that non-perturbative quantum general relativity (or, supergravity) does exist
as a mathematically consistent theory. Still, there is no a priori reason to assume that
the result would be the ‘final’ theory of all known physics. In particular, as is the case
with classical general relativity, while requirements of background independence and general
covariance do restrict the form of interactions between gravity and matter fields and among
matter fields themselves, the theory would not have a built-in principle which determines
these interactions. Put differently, such a theory would not be a satisfactory candidate
for unification of all known forces. However, just as general relativity has had powerful
implications in spite of this limitation in the classical domain, quantum general relativity
should have qualitatively new predictions, pushing further the existing frontiers of physics.
In section III we will see an illustration of this possibility.
II.
QUANTUM RIEMANNIAN GEOMETRY
In a recent short review [1] I have provided a semi-qualitative description of the quantum
Riemannian geometry. To complement that discussion, in this section I will provide a concise
but more mathematical summary.
The starting point of LQG is a Hamiltonian formulation of general relativity based on
spin connections [7]. Since all other basic forces of nature are also described by theories of
connections, this formulation naturally leads to an unification of all four fundamental forces
at a kinematical level. Specifically, the phase space of general relativity is the same as that
of a Yang-Mills theory. The difference lies in dynamics: whereas in the standard Yang-Mills
theory the Minkowski metric features prominently in the definition of the Hamiltonian, there
are no background fields whatsoever once gravity is switched on.
Let us focus on the gravitational sector of the theory. Then, the phase space Γgrav consists
of canonically conjugate pairs (Aia , Piab ), where Aia is a connection on a 3-manifold M and
Piab a 2-form, both of which take values in the Lie-algebra su(2). The connection A enables
one to parallel transport chiral spinors (such as the left handed fermions of the standard
electro-weak model) along curves in M. Its curvature is directly related to the electric and
4
magnetic parts of the space-time Riemann tensor. The dual Pia of Piab plays a double role.1
Being the momentum canonically conjugate to A, it is analogous to the Yang-Mills electric
field. In addition, Eia := 8πGγPia , has the interpretation of an orthonormal triad of frame
field (with density weight 1) on M, where γ is the ‘Barbero-Immirzi parameter’ representing
a quantization ambiguity. Each triad Eia determines a positive definite ‘spatial’ 3-metric qab ,
and hence the Riemannian geometry of M. This dual role of P is a reflection of the fact
that now SU(2) is the (double cover of the) group of rotations of the orthonormal spatial
triads on M itself rather than of rotations in an ‘internal’ space associated with M.
To pass to quantum theory, one first constructs an algebra of ‘elementary’ functions on
Γgrav (analogous to the phase space functions x and p in the case of a particle) which are to
have unambiguous operator analogs. The holonomies
R
he (A) := P exp − e A
(2.1)
associated with a (piecewise analytic) curve/edge e on M is a (SU(2)-valued) configuration
function on Γgrav . Similarly, given a (piecewise analytic) 2-surface S on M, and a su(2)valued (test) function f on M,
R
PS,f := S Tr (f P)
(2.2)
is a momentum-function on Γgrav , where Tr is over the su(2) indices. The symplectic structure on Γgrav enables one to calculate the Poisson brackets {he , PS,f }. The result is a linear
combination of holonomies and can be written as a Lie derivative,
{he , PS,f } = LXS,f he ,
(2.3)
where XS,f is a derivation on the ring generated by holonomy functions, and can therefore
be regarded as a vector field on the configuration space A of connections. This is a familiar
situation in classical mechanics of systems whose configuration space is a manifold. Functions
he and vector fields XS,f generate a Lie algebra. As in quantum mechanics on manifolds, the
first step is to promote this algebra to a quantum algebra by demanding that the commutator
be given by i~ times the Lie bracket. The result is a ⋆-algebra a, analogous to the algebra
generated by operators exp iλx̂ and p̂ in quantum mechanics. By exponentiating also the
momentum operators P̂S,f one obtains W, the analog of the quantum mechanical Weyl
algebra generated by exp iλx̂ and exp iµp̂.
The main task is to obtain the appropriate representation of these algebras. In that
representation, quantum Riemannian geometry can be probed through the momentum operators P̂S,f , which stem from classical orthonormal triads. As in quantum mechanics on
manifolds or simple field theories in flat space, it is convenient to divide the task into two
parts. In the first, one focuses on the algebra C generated by the configuration operators ĥe
and finds all its representations, and in the second one considers the momentum operators
P̂S,f to restrict the freedom.
C is called the holonomy algebra. It is naturally endowed with the structure of an Abelian
⋆
C algebra (with identity), whence one can apply the powerful machinery made available
1
Pia is a vector density, defined via 3
R
M
Pi[ab ωc]i =
R
M
Pia ωai for any 1-form ω on M .
5
by the Gel’fand theory. This theory tells us that C determines a unique compact, Hausdorff
space Ā such that the C ⋆ algebra of all continuous functions on A is naturally isomorphic
to C. Ā is called the Gel’fand spectrum of C. It has been shown to consist of ‘generalized
connections’ Ā defined as follows: Ā assigns to any oriented edge e in M an element Ā(e)
of SU(2) (a ‘holonomy’) such that Ā(e−1 ) = [Ā(e)]−1 ; and, if the end point of e1 is the
starting point of e2 , then Ā(e1 ◦ e2 ) = Ā(e1 ) · Ā(e2 ). Clearly, every smooth connection A
is a generalized connection. In fact, the space A of smooth connections has been shown to
be dense in Ā (with respect to the natural Gel’fand topology thereon). But Ā has many
more ‘distributional elements’. The Gel’fand theory guarantees that every representation of
the C ⋆ algebra C is a direct sum of representations of the following type: The underlying
Hilbert space is H = L2 (Ā, dµ) for some measure µ on Ā and (regarded as functions on Ā)
elements of C act by multiplication. Since there are many inequivalent measures on Ā, there
is a multitude of representations of C. A key question is how many of them can be extended
to representations of the full algebra a (or W) without having to introduce any ‘background
fields’ which would compromise diffeomorphism covariance. Quite surprisingly, the requirement that the representation be cyclic with respect to a state which is invariant under the
action of the group of (piecewise analytic) diffeomorphisms on M singles out a unique irreducible representation. This result was recently established for a by Lewandowski, Okolów,
Sahlmann and Thiemann [8], and for W by Fleischhack [9]. It is the quantum geometry
analog to the seminal results by Segal and others that characterized the Fock vacuum in
Minkowskian field theories. However, while that result assumes not only Poincaré invariance
but also specific (namely free) dynamics, it is striking that the present uniqueness theorems
make no such restriction on dynamics. The requirement of diffeomorphism invariance is
surprisingly strong and makes the ‘background independent’ quantum geometry framework
surprisingly tight.
This unique representation was in fact introduced already in the mid-nineties [10, 11, 12]
and has been extensively used in LQG since then. The underlying Hilbert space is given by
H = L2 (Ā, dµo ) where µo is a diffeomorphism invariant, faithful, regular Borel measure on
Ā, constructed from the normalized Haar measure on SU(2). Typical quantum states can
be visualized as follows. Fix: (i) a graph α on M, and, (ii) a smooth function ψ on [SU(2)]n .
Then, the function
Ψγ (Ā) := ψ(Ā(e1 ), . . . Ā(en ))
(2.4)
on Ā is an element of H. Such states are said to be cylindrical with respect to the graph α
and their space is denoted by Cylα . These are ‘typical states’ in the sense that Cyl := ∪α Cylα
is dense in H. Finally, as ensured by the Gel’fand theory, the holonomy (or configuration)
operators ĥe act just by multiplication. The momentum operators P̂S,f act as Lie-derivatives:
P̂S,f Ψ = −i~ LXS,f Ψ.
Given any graph α in M, and a labelling of each of its edges by a non-trivial irreducible
representation of SU(2) (i.e., by a non-zero half integer j), one can construct a finite dimensional Hilbert space Hα,~j which can be thought of as the state space of a spin system ‘living
on’ the graph α. The full Hilbert space admits a simple decomposition: H = ⊕α,~j Hα,~j .
This is called the spin-network decomposition [13, 14]. The geometric operators discussed
in Rovelli’s talk leave each Hα,~j invariant. Therefore, the availability of this decomposition
6
greatly simplifies the task of analyzing their properties [2, 15, 16].
Key features of this representation which distinguish it from, say, the standard Fock representation of the quantum Maxwell field are the following. While the Fock representation
of photons makes a crucial use of the background Minkowski metric, the above construction
is manifestly ‘background independent’. Second, the connection itself is not represented
as an operator (valued distribution). Holonomy operators, on the other hand, are welldefined. Finally, and most importantly, the Hilbert space H and the associated holonomy
and (smeared) triad operators only provide a kinematical framework —the quantum analog of the phase space. Thus, while elements of the Fock space represent physical states
of photons, elements of H are not the physical states of LQG. Rather, like the classical
phase space, the kinematic setup provides a home for formulating quantum dynamics. In
the Hamiltonian framework, the dynamical content of any background independent theory
is contained in its constraints. In quantum theory, the Hilbert space H and the holonomy
and (smeared) triad operators thereon provide the necessary tools to write down quantum
constraint operators. The physical states are solutions to these quantum constraints. Thus,
to complete the program, one has to: i) obtain the expressions of the quantum constraints;
ii) solve the constraint equations; iii) construct the physical Hilbert space from the solutions
(e.g. by the group averaging procedure); and iv) extract physics from these physical sectors
(e.g., by analyzing the expectation values, fluctuations of and correlations between Dirac observables). While strategies have been developed —particularly through Thiemann’s ‘Master
constraint program’ [17]— to complete these steps, important open issues remain in the full
theory. However, as section III illustrates, the program has been completed in mini and
midi superspace models, leading to surprising insights and answers to some long-standing
questions.
A more detailed treatment of quantum geometry along the lines presented here can be
found in, e.g., [2].
III.
APPLICATION: HOMOGENEOUS ISOTROPIC COSMOLOGY
As emphasized in Sec. I, a central feature of general relativity is its encoding of the
gravitational field in the Riemannian geometry of space-time. This encoding is directly responsible for the most dramatic ramifications of the theory: the big-bang, black holes and
gravitational waves. However, it also leads one to the conclusion that space-time itself must
end and classical physics must come to a halt at the big-bang and black hole singularities. A central question is whether the situation improves when gravity is treated quantum
mechanically. Analysis of models within LQG strongly suggests that the answer is in the
affirmative. Because of space limitation, I will restrict myself to the big bang singularity
and that too only in the simplest setting of homogeneous, isotropic cosmology.
Let us begin with a short list of long-standing questions that any satisfactory quantum
gravity theory is expected to answer:
• How close to the Big Bang does a smooth space-time of general relativity make
sense? In particular, can one show from first principles that this approximation
is valid at the onset of inflation?
7
• Is the Big-Bang singularity naturally resolved by quantum gravity? Or, is
some external input such as a new principle or a boundary condition at the Big
Bang essential?
• Is the quantum evolution across the ‘singularity’ deterministic? Since one needs
a fully non-perturbative framework to answer this question in the affirmative, in
the Pre-Big-Bang [18] and Ekpyrotic/Cyclic [19, 20] scenarios, for example, so
far the answer is in the negative.
• If the singularity is resolved, what is on the ‘other side’ ? Is there just a
‘quantum foam’, far removed from any classical space-time, or, is there another
large, classical universe?
For many years, these and related issues had been generally relegated to the ‘wish list’
of what one would like the future, satisfactory quantum gravity theory to eventually address. However, Since LQG is a background independent, non-perturbative approach, it
is well-suited to address them. Indeed, starting with the seminal work of Bojowald some
five years ago [21], notable progress has been made in the context of symmetry reduced,
minisuperspaces. In this section I will summarize the state of the art, emphasizing recent
developments. For a comprehensive review of the older work see, e.g., [22].
Consider the spatially homogeneous, isotropic, k=0 cosmologies with a massless scalar
field. It is instructive to focus on this model because every of its classical solutions has a
singularity. There are two possibilities: In one the universe starts out at the big bang and
expands, and in the other it contracts into a big crunch. The question is if this unavoidable
classical singularity is naturally tamed by quantum effects. This issue can be analyzed in
the geometrodynamical framework used in older quantum cosmology. Unfortunately, the
answer turns out to be in the negative. For example, if one begins with a semi-classical
state representing an expanding classical universe at late times and evolves it back via the
Wheeler DeWitt equation, one finds that it just follows the classical trajectory into the big
bang singularity [25, 26].
In loop quantum cosmology (LQC), the situation is very different [24, 25, 26]. This may
seem surprising at first. For, the system has only a finite number of degrees of freedom
and von Neumann’s theorem assures us that, under appropriate assumptions, the resulting
quantum mechanics is unique. The only remaining freedom is factor-ordering and this is
generally insufficient to lead to qualitatively different predictions. However, for reasons we
will now explain, LQC does turn out to be qualitatively different from the Wheeler-DeWitt
theory [23].
Because of spatial homogeneity and isotropy, one can fix a fiducial (flat) triad oeai and its
dual co-triad oωai . The SU(2) gravitational spin connection Aia used in LQG has only one
component c which furthermore depends only on time; Aia = c oωai . Similarly, the triad Eia
(of density weight 1) has a single component p; Eia = p (det oω) oeai . p is related to the scale
factor a via a2 = |p|. However, p is not restricted to be positive; under p → −p the metric
remains unchanged but the spatial triad flips the orientation. The pair (c, p) is ‘canonically
conjugate’ in the sense that the only non-zero Poisson bracket is given by:
{c, p} =
8πGγ
,
3
(3.1)
8
where as before γ is the Barbero-Immirzi parameter.
Since a precise quantum mechanical framework was not available for full geometrodynamics, in the Wheeler-DeWitt quantum cosmology one focused just on the reduced model,
without the benefit of guidance from the full theory. A major difference in LQC is that
although the symmetry reduced theory has only a finite number of degrees of freedom,
quantization is carried out by closely mimicking the procedure used in full LQG, outlined
in section II. Key differences between LQC and the older Wheeler-DeWitt theory can be
traced back to this fact.
Recall that in full LQG diffeomorphism invariance leads one to a specific kinematical
framework in which there are operators ĥe representing holonomies and P̂S,f representing
(smeared) momenta but there is no operator(-valued distribution) representing the connection A itself [8, 9]. In the cosmological model now under consideration, it is sufficient
to evaluate holonomies along segments µ oeai of straight lines determined by the fiducial
triad oeai . These holonomies turn out almost periodic functions of c, i.e. are of the form
N(µ) (c) := exp iµ(c/2). (The N(µ) are the LQC analogs of the spin-network functions of
LQG.) In the quantum theory, then, we are led to a representation in which operators N̂(µ)
and p̂ are well-defined, but there is no operator corresponding to the connection component
c. This seems surprising because our experience with quantum mechanics suggests that one
should be able to obtain the operator analog of c by differentiating N̂(µ) with respect to the
parameter µ. However, in the representation of the basic quantum algebra that descends to
LQC from full LQG, although the N̂(µ) provide a 1-parameter group of unitary transformations, the group fails to be weakly continuous in µ. Therefore one can not differentiate and
obtain the operator analog of c.
In quantum mechanics, this would be analogous to having well-defined (Weyl) operators
corresponding to the classical functions exp iµx but no operator x̂ corresponding to x itself. This violates one of the assumptions of the von-Neumann uniqueness theorem. New
representations then become available which are inequivalent to the standard Schrödinger
one. In quantum mechanics, these representations are not of direct physical interest because we need the operator x̂. In LQC, on the other hand, full LQG naturally leads us to
a new representation, i.e., to new quantum mechanics. This theory is inequivalent to the
Wheeler-DeWitt type theory already at a kinematical level. In particular, just as we are led
to complete the space A of smooth connections to the space Ā of generalized connections
in LQG, in LQC we are led to consider the Bohr compactification R̄Bohr of the ‘c-axis’. The
gravitational Hilbert space is now L2 (R̄Bohr , dµBohr ), rather than the standard L2 (R, dµ)
used in the Wheeler-DeWitt theory [23] where dµBohr is the LQC analog of the measure
dµo selected by the uniqueness results [8, 9] in full LQG. While in the semi-classical regime
LQC is well approximated by the Wheeler-DeWitt theory, important differences manifest
themselves at the Planck scale. These are the hallmarks of quantum geometry [2, 22].
The new representation also leads to a qualitative difference in the structure of the Hamiltonian constraint operator: the gravitational part of the constraint is a difference operator, rather than a differential operator as in the Wheeler-DeWitt theory. The derivation
[23, 25, 26] can be summarized briefly
follows. In the classical theory, the gravitational
R 3 asijk
k
part of the constraint is given by d x ǫ e−1 Eia Ejb Fab k where e = | det E|1/2 and Fab
the
9
curvature of the connection Aia . The part ǫijk e−1 Eia Ejb of this operator involving triads can
be quantized [21, 23] using a standard procedure introduced by Thiemann in the full theory
[4]. However, since there is no operator corresponding to the connection itself, one has to
k
express Fab
as a limit of the holonomy around a loop divided by the area enclosed by the
loop, as the area shrinks to zero. Now, quantum geometry tells us that the area operator has
a minimum non-zero eigenvalue, ∆, and in the quantum theory it is natural to shrink the
loop only till it attains this minimum. There are two ways to implement this idea in detail
(see [23, 25, 26]). In both cases, it is the existence of the ‘area gap’ ∆ that leads one to a
difference equation. So far, most of the LQC literature has used the first method [23, 25]. In
the resulting theory, the classical big-bang is replaced with a quantum bounce with a number of desirable features. However, it also has one serious drawback: at the bounce, matter
density can be low even for physically reasonable choices of quantum states (for details, see
[25]); i.e. the theory predicts certain departures from classical general relativity even in
the low curvature regime. The second and more recently discovered method [26] cures this
problem while retaining the physically appealing features of the first and, furthermore, has
a more direct motivation. Due to space limitation, I will confine myself only to the second
method.
Let us represent states as functions Ψ(v, φ), where φ is the scalar field and the dimensionless real number v represents geometry. Specifically, |v| is the eigenvalue of the operator
V̂ representing volume (essentially the cube of the scale factor):
p √
8πγ 3
3 3 3
3
√
V̂ |vi = K (
) 2 |v| ℓPl |vi where K =
(3.2)
6
2 2
Then, the LQC Hamiltonian constraint assumes the form:
∂φ2 Ψ(v, φ) = [B(v)]−1 C + (v) Ψ(v + 4, φ) + C o (v) Ψ(v, φ) + C − (v) Ψ(v − 4, φ)
=: −Θ Ψ(v, φ)
(3.3)
where the coefficients C ± (v), C o (V ) and B(v) are given by:
3πKG
|v + 2| |v + 1| − |v + 3|
8
C − (v) = C + (v − 4) and C o (v) = −C + (v) − C − (v)
3
3
3
1/3
1/3
B(v) =
.
K |v| |v + 1| − |v − 1|
2
C + (v) =
(3.4)
Now, in each classical solution, φ is a globally monotonic function of time and can therefore be taken as the dynamical variable representing an internal clock. In quantum theory
there is no space-time metric, even on-shell. But since the quantum constraint (3.3) dictates
how Ψ(v, φ) ‘evolves’ as φ changes, it is convenient to regard the argument φ in Ψ(v, φ) as
emergent time and v as the physical degree of freedom. A complete set of Dirac observables
is then provided by the constant of motion p̂φ and operators v̂|φo determining the value of v
at the ‘instant’ φ = φo .
10
Physical states are the (suitably regular) solutions to Eq (3.3). The map Π̂ defined
by Π̂ Ψ(v, φ) = Ψ(−v, φ) corresponds just to the flip of orientation of the spatial triad
(under which geometry remains unchanged); Π̂ is thus a large gauge transformation on the
space of solutions to Eq. (3.3). One is therefore led to divide physical states into sectors,
each providing an irreducible, unitary representation of this gauge symmetry. Physical
considerations [25, 26] imply that we should consider the symmetric sector, with eigenvalue
+1 of Π̂.
To endow this space with the structure of a Hilbert space, one can proceed along one of
two paths. In the first, one defines the action of the Dirac observables on the space of suitably
regular solutions to the constraints and selects the inner product by demanding that these
operators be self-adjoint [27]. A more systematic procedure is the ‘group averaging method’
[28]. The technical implementation [25, 26] of both these procedures is greatly simplified by
the fact that the difference operator Θ on the right side of (3.3) is independent of φ and can
be shown to be self-adjoint and positive definite on the Hilbert space L2 (R̄Bohr , B(v)dµBohr ).
The final result can be summarized as follows. Since Θ is a difference operator, the
physical Hilbert space Hphy has sectors Hǫ which are superselected; Hphy = ⊕ǫ Hǫ with
ǫ ∈ (0, 2). The overall predictions are insensitive to the choice of a specific sector (for
details, see [25, 26]). States Ψ(v, φ) in Hǫ are symmetric under the orientation inversion Π̂
and have support on points v = |ǫ| + 4n where n is an integer. Wave functions Ψ(v, φ) in a
generic sector solve (3.3) and are of positive frequency with respect to the ‘internal time’ φ:
they satisfy the ‘positive frequency’ square root
√
−i∂φ Ψ = Θ Ψ .
(3.5)
of Eq (3.3). The physical inner product is given by:
X
hΨ1 | Ψ2 i =
B(v) Ψ̄1 (v, φo )Ψ2 (v, φo)
(3.6)
v∈{|ǫ|+4n}
and is ‘conserved’, i.e., is independent of the ‘instant’ φo chosen in its evaluation. On these
states, the Dirac observables act in the expected fashion:
p̂φ Ψ = −i~∂φ Ψ
√
v̂|φo Ψ(v, φ) = ei
Θ(φ−φo )
v Ψ(v, φo)
(3.7)
To construct semi-classical states and for numerical simulations, it is convenient to express
physical states as linear combinations of the eigenstates of p̂φ and Θ. To carry out this step, it
is convenient to consider the Wheeler-DeWitt theory first. Let us begin with the observation
that, for v ≫ 1, there is a precise sense [26] in which the difference operator Θ approaches
the Wheeler DeWitt differential operator Θ, given by
(3.8)
ΘΨ(v, φ) = 12πG v∂v v∂v Ψ(v, φ)
Thus, if one ignores the quantum geometry effects, Eq (3.3) reduces to the Wheeler-DeWitt
equation
(3.9)
∂φ2 Ψ = −Θ Ψ.
11
Note that the operator Θ is positive definite and self-adjoint on the Hilbert space
L2s (R, B(v)dv) where the subscript s denotes the restriction to the symmetric eigenspace
of Π and B(v) := Kv −1 is the limiting form of B(v) for large v. Its eigenfunctions ek with
eigenvalue ω 2(≥ 0) are 2-fold degenerate on this Hilbert space. Therefore, they can be
labelled by a real number k:
1
ek (v) := √ eik ln |v|
(3.10)
2π
√
where k is related to ω via ω =
12πG|k|. They form an orthonormal basis on
L2s (R, B(v)dv). A ‘general’ positive frequency solution to (3.9) can be written as
Z ∞
(3.11)
Ψ(v, φ) =
dk Ψ̃(k) ek (v)eiωφ
−∞
for suitably regular Ψ̃(k).
The complete set of eigenfunctions ek (v) of the discrete operator Θ is also labelled by a
real number k and detailed numerical simulations show that ek (v) are well-approximated
by
√
2
ek (v) for v ≫ 1. The eigenvalues ω (k) of Θ are again given by ω = 12πG|k|. Finally,
the ek (v) satisfy the standard orthonormality relations < ek |e′k >= δ(k, k ′ ). A physical state
Ψ(v, φ) can therefore be expanded as:
Z ∞
(s)
Ψ(v, φ) =
dk Ψ̃(k) ek (v) eiω(k)φ
(3.12)
−∞
√
(s)
where Ψ̃(k) is any suitably regular function of k, and ek (v) = (1/ 2)(ek (v)+ek (−v)). Thus,
as in the Wheeler-DeWitt theory, each physical state is characterized by a free function Ψ̃(k).
The difference between the two theories lies in the functional forms of the eigenfunctions
ek (v) of Θ and ek (v) of Θ. While ek (v) is well approximated by ek (v) for large v, the
differences are very significant for small v and they lead to very different dynamics.
With the physical Hilbert space and a complete set of Dirac observables at hand, we can
now construct states which are semi-classical at late times —e.g., now— and evolve them
numerically ‘backward in time’. There are three natural constructions to implement this
idea in detail, reflecting the freedom in the notion of semi-classical states. In all cases, the
main results are the same [25, 26]. Here I will report on the results obtained using the
strategy that brings out the contrast with the Wheeler DeWitt theory most sharply.
As noted before, pφ is a constant
of motion. For the semi-classical analysis, we are led
√
⋆
to choose a large value pφ (≫ G~). In the closed model, for example, this condition is
necessary to ensure that the universe can expand out to a macroscopic size. Fix a point
(v ⋆ , φo ) on the corresponding classical trajectory which starts out at the big bang and then
expands, choosing v ⋆ ≫ 1. We want to construct a state which is peaked at (v ⋆ , p⋆φ ) at a
‘late initial time’ φ = φo and follow its ‘evolution’ backward. At ‘time’ φ = φo , consider then
the function
Z ∞
(k−k⋆ )2
⋆
Ψ(v, φo) =
dk Ψ̃(k) ek (v) eiω(φo −φ ) , where Ψ̃(k) = e− 2σ2
(3.13)
−∞
12
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0
quantum
classical
-0.2
-0.4
-0.6
,φ)|
|Ψ(v
-1.2
1.5
-1
1
-0.8
0.5
-0.6
0
3
0
5*1
-0.4
4
4
10
4
10
1.5*
4
10
2.0*
4
10
2.5*
4
10
3.0*
4
10
3.5*
10
v
4.0*
1.0*
-0.2
φ
-0.8
-1
φ
-1.2
0
4
1*10
4
4
3*10
2*10
4
4*10
4
5*10
v
FIG. 1: The figure on left shows the absolute value of the wave function Ψ as a function of φ
and v. Being a physical state, Ψ is symmetric under v → −v. The figure on the right shows
the expectation values of Dirac observables v̂|φ and their dispersions. They exhibit a quantum
bounce which joins the contracting and expanding classical trajectories marked by fainter
√ lines.
In this imulation, the parameters in the initial data are: v ⋆ = 5 × 104 , p⋆φ = 5 × 103 G~ and
∆pφ /pφ = 0.0025.
√
p
where k ⋆ = −p⋆φ / 12πG~2 and φ⋆ = − 1/12πG ln(v ⋆ ) + φo . In the Wheeler-DeWitt
theory on can easily evaluate the integral in the approximation |k ∗ | ≫ 1 and calculate
mean values of the Dirac observables and their fluctuations. One finds that, as required,
the state is sharply peaked at values v ⋆ , p⋆φ . The above construction is closely related to
that of coherent states in non-relativistic quantum mechanics. The main difference is that
the observables of interest are not v and its conjugate momentum but rather v and pφ —
the momentum conjugate to ‘time’, i.e., the analog of the Hamiltonian in non-relativistic
quantum mechanics. Now, one can evolve this state backwards using the Wheeler-DeWitt
equation (3.9). It follows immediately from the form (3.11) of the general solution to (3.9)
and the fact that pφ is large that this state would remain sharply peaked at the chosen
classical trajectory and simply follow it into the big-bang singularity.
In LQC, we can use the restriction of (3.13) to points v = |ǫ| + 4n as the initial data
and evolve it backwards numerically. Now the evolution is qualitatively different (see Fig.1).
The state does remains sharply peaked at the classical trajectory till the matter density
reaches a critical value:
√
3
ρcrit =
,
(3.14)
2
16π γ 3 G2 ~
which is about 0.82 times the Planck density. However, then it bounces. Rather than
following the classical trajectory into the singularity as in the Wheeler-DeWitt theory, the
state ‘turns around’. What is perhaps most surprising is that it again becomes semi-classical
13
and follows the ‘past’ portion of a classical trajectory, again with pφ = p⋆φ , which was headed
towards the big crunch. Let us we summarize the forward evolution of the full quantum state.
In the distant past, the state is peaked on a classical, contracting pre-big-bang branch which
closely follows the evolution dictated by Friedmann equations. But when the matter density
reaches the Planck regime, quantum geometry effects become significant. Interestingly, they
make gravity repulsive, not only halting the collapse but turning it around; the quantum
state is again peaked on the classical solution now representing the post-big-bang, expanding
universe. Since this behavior is so surprising, a very large number of numerical simulations
were performed to ensure that the results are robust and not an artifact of the special choices
of initial data or of the numerical methods used to obtain the solution [25, 26].
For states which are semi-classical at late times, the numerical evolution in exact LQC
can be well-modelled by an effective, modified Friedman equation :
ȧ2
8πG h
ρ i
(3.15)
=
ρ
1
−
a2
3
ρcrit
where, as usual, a is the scale factor. In the limit ~ → 0, ρcrit diverges and we recover the
standard Friedmann equation. Thus the second term is a genuine quantum correction. Eq.
(3.15) can also be obtained analytically from (3.3) by a systematic procedure [29]. But the
approximations involved are valid only well outside the Planck domain. It is therefore surprising that the bounce predicted by the exact quantum equation (3.3) is well approximated
by a naive extrapolation of (3.15) across the Planck domain. While there is some understanding of this seemingly ‘unreasonable success’ of the effective equation (3.15), further
work is needed to fully understand this issue.
Finally let us return to the questions posed in the beginning of this section. In the model,
LQC has been able to answer all of them. One can deduce from first principles that classical
general relativity is an excellent approximation till very early times, including the onset
of inflation in standard scenarios. Yet quantum geometry effects have a profound, global
effect on evolution. In particular, the singularity is naturally resolved without any external
input and there is a classical space-time also in the pre-big-bang branch. LQC provides a
deterministic evolution which joins the two branches.
IV.
DISCUSSION
Even though there are several open issues in the formulation of full quantum dynamics in
LQG, detailed calculations in simple models have provided hints about the general structure.
It appears that the most important non-perturbative effects arise from the replacement of
i
the local curvature term Fab
by non-local holonomies. This non-locality is likely to be a
central feature of the full LQG dynamics. In the cosmological model considered in section
III, it is this replacement of curvature by holonomies that is responsible for the subtle but
crucial differences between LQC and the Wheeler-DeWitt theory.2
2
Because early presentations emphasized the difference between B(v) of LQC and B(v) = Kv −1 of the
Wheeler-DeWitt theory, there is a misconception in some circles that the difference in quantum dynamics
14
By now a number of mini-superspaces and a few midi-superspaces have been studied
in varying degrees of detail. In all cases, the classical, space-like singularities are resolved
by quantum geometry provided one treats the problem non-perturbatively. For example,
in anisotropic mini-superspaces, there is a qualitative difference between perturbative and
non-perturbative treatments: if anisotropies are treated as perturbations of a background
isotropic model, the big-bang singularity is not resolved while if one treats the whole problem
non-perturbatively, it is [30].
A qualitative picture that emerges is that the non-perturbative quantum geometry corrections are ‘repulsive’. While they are negligible under normal conditions, they dominate
when curvature approaches the Planck scale and halt the collapse that would classically
have lead to a singularity. In this respect, there is a curious similarity with the situation in
the stellar collapse where a new repulsive force comes into play when the core approaches a
critical density, halting further collapse and leading to stable white dwarfs and neutron stars.
This force, with its origin in the Fermi-Dirac statistics, is associated with the quantum nature
of matter. However, if the total mass of the star is larger than, say, 5 solar masses, classical
gravity overwhelms this force. The suggestion from LQC is that near Planck densities a new
repulsive force associated with the quantum nature of geometry may come into play which
is strong enough to prevent the formation of singularities irrespective of how large he mass
is. Since this force is negligible until one enters the Planck regime, predictions of classical
relativity on the formation of trapped surfaces, dynamical and isolated horizons [31] would
still hold. But assumptions of the standard singularity theorems would be violated. There
may be no singularities, no abrupt end to space-time where physics stops. Non-perturbative,
background independent quantum physics could continue.
The major weakness of the current status of LQG is that so far one has been able to obtain
detailed dynamical predictions only in symmetry reduced models. These results do provide
valuable hints for the full theory but a large number of ambiguities still remain there. A fascinating question is whether the singularity resolution due to quantum geometry is a rather
general feature which is largely insensitive to these ambiguities [32]. When matter satisfies
the appropriate energy conditions in general relativity, the Raychaudhuri equation captures
the attractive nature of gravity in a particularly convenient fashion, providing a central
ingredient to the singularity theorems. Is there a general equation in quantum geometry
which implies that gravity effectively becomes repulsive near generic space-like singularities,
thereby halting the classical collapse? If so, one could construct robust arguments, establishing general ‘singularity resolution theorems’ for broad classes of situations in quantum
gravity, without having to analyze models, one at a time.
is primarily due to the non-trivial ‘inverse volume’ operator of LQC. This is not correct. In deed, in the
model considered here, qualitative features of quantum dynamics, including the bounce, remain unaffected
if one replaces by hand B(v) with B(v) in the LQC evolution equation (3.3).
15
Acknowledgments: I would like to thank Martin Bojowald, Jerzy Lewandowski, and
especially Tomasz Pawlowski and Parampreet Singh for collaboration and numerous discussions. This work was supported in part by the NSF grants PHY-0354932 and PHY-0456913,
the Alexander von Humboldt Foundation, the Krammers Chair program of the University
of Utrecht and the Eberly research funds of Penn State.
[1] A. Ashtekar, Gravity and the Quantum, New J. Phys. 7 (2005) 198; arXiv:gr-qc/0410054.
[2] A. Ashtekar and J. Lewandowski, Background independent quantum gravity: A status report,
Class. Quant. Grav. 21, R53-R152 (2004), arXiv:gr-qc/0404018.
[3] C. Rovelli Quantum Gravity, (CUP, Cambridge, 2004).
[4] T. Thiemann, Introduction to Modern Canonical Quantum General Relativity (CUP, Cambridge, at press).
[5] O. Luscher and M. Reuter, Asymptotic Safety in Quantum Einstein Gravity: nonperturbative
renormalizability and fractal spacetime structure, arXiv:hep-th/0511260.
[6] R. Percacci and D. Perini, Asymptotic Safety of Gravity Coupled to Matter, Phys. Rev. D68,
044018 (2003).
[7] A. Ashtekar, New variables for classical and quantum gravity Phys. Rev. Lett. 57 2244-2247
(1986)
New Hamiltonian formulation of general relativity Phys. Rev. D36, 1587-1602 (1987).
[8] J. Lewandowski, A. Okolow, H. Sahlmann and T. Thiemann, Uniqueness of diffeomorphism
invariant states on holonomy flux algebras, arXiv:gr-qc/0504147.
[9] C. Fleishchack,
Representations of the Weyl algebra in quantum geometry,
arXiv:math-ph/0407006.
[10] A. Ashtekar and J. Lewandowski, Representation theory of analytic holonomy algebras, in
Knots and Quantum Gravity, ed J. Baez, (Oxford U. Press, Oxford, 1994)
Projective techniques and functional integration, Jour. Math. Phys. 36, 2170-2191 (1995)
A. Ashtekar and J. Lewandowski, Differential geometry on the space of connections using
projective techniques Jour. Geo. & Phys. 17, 191–230 (1995)
[11] J. C. Baez, Generalized measures in gauge theory, Lett. Math. Phys. 31, 213-223 (1994).
[12] D. Marolf and J. Mourão, On the support of the Ashtekar-Lewandowski measure, Commun.
Math. Phys. 170, 583-606 (1995).
[13] C. Rovelli and L. Smolin, Spin networks and quantum gravity Phys. Rev. D52 5743–5759
(1995).
[14] J. C. Baez, Spin networks in non-perturbative quantum gravity, in The Interface of Knots and
Physics ed Kauffman L (American Mathematical Society, Providence) pp. 167–203 (1996).
[15] C. Rovelli and L. Smolin, Discreteness of area and volume in quantum gravity Nucl. Phys.
B442 593–622 (1995) ; Erratum: Nucl. Phys. B456 753 (1995).
[16] A. Ashtekar and J. Lewandowski, Quantum theory of geometry I: Area operators, Class.
Quant. Grav. 14 A55–A81 (1997)
Quantum theory of geometry II: Volume Operators Adv. Theo. Math. Phys. 1 388–429 (1997).
16
[17] T. Thiemann, The phoenix project: Master constraint program for loop quantum gravity,
arXiv:gr-qc/0305080; T. Thiemann and B. Dittrich, Testing the master constraint program
for loop quantum gravity I. General Framework, Class. Quantum. Grav. 23 1025-1066 (2006).
[18] M. Gasperini and G. Veneziano, The pre-big bang scenario in string cosmology, Phys. Rept.
373, 1 (2003) arXiv:hep-th/0207130.
[19] J. Khoury, B. A. Ovrut, P. J. Steinhardt, N. Turok, The Ekpyrotic Universe: Colliding Branes
and the Origin of the Hot Big Bang, Phys. Rev. D64 (2001) 123522 hep-th/0103239.
[20] J. Khoury, B. . Ovrut, N. Seiberg, P. J. Steinhardt, N. Turok, From Big Crunch to Big Bang,
Phys.Rev. D65 (2002) 086007 hep-th/0108187.
[21] M. Bojowlad, Absence of singularity in loop quantum cosmology, Phys. Rev. Lett. 86, 52275230 (2001), arXiv:gr-qc/0102069, Isotropic loop quantum cosmology, Class. Quantum.
Grav. 19, 2717-2741 (2002), arXiv:gr-qc/0202077.
[22] M. Bojowald, Loop quantum cosmology, Liv. Rev. Rel. 8, 11 (2005), arXiv:gr-qc/0601085.
[23] A. Ashtekar, M. Bojowald, J. Lewandowski, Mathematical structure of loop quantum cosmology, Adv. Theo. Math. Phys. 7, 233-268 (2003), gr-qc/0304074.
[24] A. Ashtekar, T. Pawlowski and P. Singh, Quantum Nature of the Big Bang, Phys. Rev. Lett.
96, 141301 (2006) arXiv:gr-qc/0602086.
[25] A. Ashtekar, T. Pawlowski and P. Singh, Quantum Nature of the Big Bang: An analytical
and numerical Investigation, Phys. Rev. D73, 124038 (2006) arXiv:gr-qc/0604013.
[26] A. Ashtekar, T. Pawlowski and P. Singh, Quantum Nature of the Big Bang: Improved dynamics. gr-qc/0607039.
[27] A. Ashtekar, Lectures on non-perturbative canonical gravity, Notes prepared in collaboration
with R. S. Tate (World Scientific, Singapore, 1991), Chapter 10.
[28] D. Marolf, Refined algebraic quantization:
Systems with a single constraint
arXives:gr-qc/9508015;
Quantum observables and recollapsing dynamics, Class. Quant. Grav. 12 (1995) 1199-1220.
[29] J. Willis, On the low energy ramifications and a mathematical extension of loop quantum
gravity, Ph.D. Dissertation, The Pennsylvaina State University (2004); A. Ashtekar, M. Bojowald and J. Willis, Corrections to Friedmann equations induced by quantum geometry,
IGPG preprint (2004).
[30] M. Bojowald, H. H. Hernandez and H. A. Morales-Tecotl, Perturbative degrees of freedom in loop quantum gravity: Anisotropies, Class. Quant. Grav., 18 L117–L127 (2001)
arXiv:gr-qc/0511058.
[31] A. Ashtekar and B. Krishnan Isolated and dynamical horizons and their applications, Living
Rev. Rel. 10 1-78 (2004), gr-qc/0407042
[32] N. Dadhich, Singularity: Is Raychaudhuri equation required once again? to appear in: Raychaudhuri Memomrial Volume.
View publication stats