Projection Operator Formalism for Quantum Constraints

Material Information

Projection Operator Formalism for Quantum Constraints
Little, Jeffrey Scott
Place of Publication:
[Gainesville, Fla.]
University of Florida
Publication Date:
Physical Description:
1 online resource (122 p.)

Thesis/Dissertation Information

Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Committee Chair:
Klauder, John R.
Committee Members:
Muttalib, Khandker A.
Detweiler, Steven L.
Yelton, John M.
Robinson, Paul L.
Shabanov, Sergei
Graduation Date:


Subjects / Keywords:
Algebra ( jstor )
Coordinate systems ( jstor )
Equations of motion ( jstor )
Hilbert spaces ( jstor )
Mathematical vectors ( jstor )
Mathematics ( jstor )
Momentum ( jstor )
Physics ( jstor )
Quantum field theory ( jstor )
Quantum mechanics ( jstor )
Physics -- Dissertations, Academic -- UF
constraint, projection, quantum
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Physics thesis, Ph.D.


Motivated by several theoretical issues surrounding quantum gravity, a course of study has been implemented to gain insight into the quantization of constrained systems utilizing the Projection Operator Formalism. Throughout this dissertation we will address several models and techniques used in an attempt to illuminate the subject. We also attempt to illustrate the utility of the Projection Operator Formalism in dealing with any type of quantum constraint. Quantum gravity is made more difficult in part by its constraint structure. The constraints are classically first-class; however, upon quantization they become partially second-class. To study such behavior, we will focus on a simple problem with finitely many degrees of freedom and will demonstrate how the Projection Operator Formalism is well suited to deal with this type of constraint. Typically, when one discusses constraints, one imposes regularity conditions on these constraints. We introduce the 'new' classification of constraints called 'highly irregular' constraints, due to the fact these constraints contain both regular and irregular solutions. Quantization of irregular constraints is normally not considered; however, using the Projection Operator Formalism we provide a satisfactory quantization. It is noteworthy that irregular constraints change the observable aspects of a theory as compared to strictly regular constraints. More specifically, we will attempt to use the tools of the Projection Operator Formalism to study another gravitationally inspired model, namely the Ashtekar-Horowitz-Boulware model. We will also offer a comparison of the results obtained from the Projection Operator Formalism with that of the Refined Algebraic Quantization scheme. Finally, we will use the Projection Operator Method to discuss time-dependent quantum constraints. In doing so, we will develop the formalism and study a few key time-dependent models to help us obtain a larger picture on how to deal with reparameterization invariant theories such as General Relativity. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis (Ph.D.)--University of Florida, 2007.
Adviser: Klauder, John R.
Statement of Responsibility:
by Jeffrey Scott Little.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Little, Jeffrey Scott. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
LD1780 2007 ( lcc )


This item has the following downloads:

Full Text



ACKNOW LEDGMENTS .................................

A B ST R A C T . . . . . . . . . .


1 INTRODUCTION ..................................

1.1 Philosophy .................. ................
1.2 Outline of the Remaining C'!i lpters ......................


2.1 Classical Picture . . . . . . ..... .
2.1.1 Geometric Pl a round .........................
2.1.2 Constraints Appear ...........................
2.1.3 Another Geometric Interlude ................. .....
2.1.4 O bservables . . . . . . . .
2.2 Q uantization . . . . . . . . .
2.2.1 Canonical Quantization Program .. ...............
2.2.2 What About Constraints? The Dirac Method .. .........

3 OTHER M ETHODS .. . ...........................

3.1 Faddeev-Popov Method .. ................
3.1.1 Yet Another Geometric Interlude from the Constraint
3.1.2 Basic Description .. ................
3.1.3 Comments and Criticisms .. ............
3.2 Refined Algebraic Quantization .......
3.2.1 Basic Outline of Procedure .. ............
3.2.2 Comments and Criticisms .. ............
3.3 Master Constraint Program .. ...............
3.3.1 Classical Description .. ...............
3.3.2 Quantization . . . . . .
3.3.3 MCP Constraint Example .. ............
3.3.4 Comments and Criticisms .. ............
3.4 Conclusions . . . . . . .


4.1 Method and Motivation .. ................
4.1.1 Squaring the Constraints .. .............
4.1.2 Classical Consideration .. .............. Quantum Consideration .. ......... Projection Operator Justification ......


where H is the Hamiltonian and {., .} are the classical Poisson brackets. The Poisson

brackets are defined by the following:

{f,g} f g f (212)
Oq Oi pi i Oqi'

where f, g e C2(M). The Poisson brackets have the following properties for any f, g, and

hE C2 (M):

{f,g} = -{g, f}, (2 13)

{f,gh} = {f,g}h+ {f,h}g, (2-14)

{{f,g},h} + {{h, f},g} +{{g,h}, f} 0. (2-15)

Equation (2-13) illustrates that the Poisson brackets are antisymmetric in respect

to its arguments. Equation (2-14) serves as the connection of the Poisson bracket

with point-wise multiplication of the functions over the phase space. Finally, (2-15)

demonstrates that the Poisson bracket obeys the famous Jacobi identity. With these three

properties it is possible to show that the classical functions over phase space form a Lie

algebra with respect to the Poisson bracket.[14]5

The closed, non-degenerate, symplectic two-form is defined by the Poisson brackets of

the dynamical variables,

S- {q,pj} = 6>. (2-16)

The symplectic form is a crucial element when we move from a classical discussion of a

system to its quantum analogue.

5 Technically, a Lie algebra would only require the first and the third properties. One
must also show that the functions in C" form a vector space, in order to be classified an

the classical reduced phase space, by noting that the super-selection sectors are related

to the classical singularities in the classical phase space. This dependence on the classical

regime to determine quantum behavior is rather disturbing. This dependency should be

reversed. The main difference in the RAQ procedure and the Projection Operator method

is when 6(h) is taken to zero. In the Projection Operator method the limit is taken after

the evaluation of the matrix elements, while the RAQ method requires the limit to be

taken before the evaluation. As we have demonstrated in this chapter, these two methods

are related, however their implementation is different and the results obtained in this

particular type of model are not the same.

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



Jeffrey Scott Little

December 2007

('C! i': John Klauder
Major: Physics

Motivated by several theoretical issues surrounding quantum gravity, a course of

study has been implemented to gain insight into the quantization of constrained systems

utilizing the Projection Operator Formalism. Throughout this dissertation we will address

several models and techniques used in an attempt to illuminate the subject. We also

attempt to illustrate the utility of the Projection Operator Formalism in dealing with any

type of quantum constraint.

Quantum gravity is made more difficult in part by its constraint structure. The

constraints are classically first-class; however, upon quantization they become partially

second-class. To study such behavior, we will focus on a simple problem with finitely

many degrees of freedom and will demonstrate how the Projection Operator Formalism is

well suited to deal with this type of constraint.

Typically, when one discusses constraints, one imposes regularity conditions on these

constraints. We introduce the "r i, classification of constraints called "highly .-i,, II., l,,

constraints, due to the fact these constraints contain both regular and irregular solutions.

Quantization of irregular constraints is normally not considered; however, using the

Projection Operator Formalism we provide a satisfactory quantization. It is noteworthy

that irregular constraints change the observable aspects of a theory as compared to

strictly regular constraints. More specifically, we will attempt to use the tools of the

Projection Operator Formalism to study another gravitationally inspired model, namely

Let us examine the extended-coherent state overlap of an equivalent form of the

projection operator [13]

E T e ""d (8-42)

where f({) is some function chosen to insure (8-42) converges absolutely and T is the time

ordered product.

(p", q", p, t"Ep',q',p', t') = (p", q"\ (p, t"p', t') 0 p', q
SN N-1
= lim(p", q" p Ptk+1lEt) tk(t) 0 l~k q') (8 43)
kO k=o
We can generalize the result from [54] namely

lim 6-+ 0(p",q", p", t" E(-6 < -(P, Q + Pt) < 6)p', q', p, t')

S(p", q1" 0, 0|e-' "+(Q))/2-(t"-t')e-(p+(PQ))/2 p' q', 0, 0) (8-44)

to include time dependent Hamiltonians.

(p", q", p, t" IE(- < R(P, Q, T) + Pt) < )p', q', p', t')

T(p", q", 0, 0e- "+l(P,Q),T) 2e ( 1), dte (p'+ (P,Q,T))/2 lp, q', 0, ) (8 45)

As in (8-44), we observe the variables pt and t are not needed to span the reduced

Hilbert space, therefore we can integrate p"' and p' without altering the physics.

Therefore, the most general statement we can make about a system with time-

dependent constraint is encapsulated in the following

lim Te-i N 1), -(t)dtEN_le-ij(N 2) N-2 e i "I E, (8-46)


E,= Te- -f( tDR(A) (8-47)


The primary goal of this chapter is to introduce regularity conditions on constraints,

as well as present a i, --" classification of constraints called highly .:,. i'l.n constraints

and also illustrate techniques used to deal with quantum versions of these constraints. The

basis of this chapter comes from [9] and [10].

5.1 Classification

In constrained dynamics one typically places regularity conditions on the constraint

to insure linear independence. If we consider A classical constraints, ,a, a {1, .., A}

the regularity condition can be stated in terms of the rank of the Jacobian matrix of the

constraints [16]

Rank( A, (5-1)
f(p",q,) r '
where n {1, ..., M}, 2M is the dimensionality of phase space, and F is the constraint

hypersurface (0a 0). If this condition fails, then the constraint (or set of constraints) is

called irregular [16]. Irregular constraints can appear in following form

a, (5-2)

where Qa is a regular constraint and r is an exponent r > 1 In the literature [16]

the measure of irregularity is based on the order of the zero on the constraint surface.

For example, (5-2) is an rth order irregular constraint. We should note that while the

constraints Oa and or are equivalent (i.e. the constraints generate the same constraint

hypersurface), the dynamics and set of observables associated with each given system are

not necessarily equivalent.

The term highly .:, ,-/;,. l, constraint refers to a constraint function that involves

both regular and irregular constraints or two or more constraints of varying order [9]. For

Once this calculation is performed for all values of n, consistent with (6-10), then we

can write the reproducing kernel for the physical Hilbert space in the following manner

/C = max mCnm (6-15)

Similarly, the physical Hilbert space can be written as

'Hphys= ( m. (6 16)
n m

The support of the reproducing kernel is only in the classically allowed regions. This

implies there is no tunneling into classically forbidden regions as reported by Boulware


6.5 Super-selection Sectors?

Before determining whether or not the physical Hilbert space calculated in the

preceding section contains super-selection sectors, let us first divert the question and

discuss what is formally meant by super-selection sectors. Suppose a physical Hilbert

space is given by the following,

9phys e iN (6-17)

which is to w that the physical Hilbert space is given by the direct sum of individual

Hilbert spaces. The physical Hilbert space is said to decompose into super-selection sectors

if for any two states |1), 102) that belong to two different sectors 7-H and Hj, respectively,

and for any observable 0 in Aobs, where Aob8 the *-algebra of all observables, the following


physK(i |0\|)phys = 0. (6-18)

In (6-18), O denotes a generic self-adjoint operator in the unreduced Hilbert space.

In previous works using the RAQ procedure [31], super-selection sectors arose because

each sector had a different degree of divergence. Since O is a self-adjoint operator in the

unreduced space, (6-18) is forced to vanish to avoid a contradiction from the varying

degrees of divergency [46]. As we have shown in the proceeding section the physical

Cm(', pp, y', py,; x, p,; py,) constant
2 sin(1/SQ()( y')) Di( ,-p'2')/2+/ /2
27(y y')
x exp[-(Yi- y)2/2 iYm(py py) (Y- y')2/2]

x *(ik-(y' + ip, y~; pi)e(i-7(y + ip, y2; pi)

x exp[-n2] exp[in((x' x) + (-p p)],


where pi = exp[-27r2]. Following the prescription set forth in Section 2, we perform the

required similarity transformation to extract the leading 6 dependency of the reproducing


Sm() 2 sin(1/Sn(5) (Y y')) ei(pxz-'_p')/2+p2/2+p /2
27(y y')
x exp[-(Y y)2/2- ity(p' py) (Y- ')2/2]

x *(ix(y' + ip, y; pi)(i7 x(y + ipy y; pi)

x exp[-n2] exp[in((x' x) + i(-p' -p)],


The limit 6 -- 0 can now be taken in a suitable manner to determine the reduced

reproducing kernel for this portion of the physical Hilbert space [13] which then reads

'mnn constant

Sei( -p'.,x')/2+-(p')2/2+(p,)2/2
x Zmexp[-(y.- y)2/2 iy(p- py) (y, y')2/2]

x e*(i-(y' + ipy yT; pl)6(iw(Y + iy Ym; P1)

x exp[-n2] exp[2in7((x' x) + i(p' px))]

for each value of m. Each of these reduced reproducing kernel Hilbert spaces is isomorphic
to a one-dimensional Hilbert space (i.e. 7i Q C). We continue the procedure for each

whole number value of n until the maximum allowed value (of n) is reached.


where v" represents an arbitrary function. It is clear that these transformations, actually
define an equivalence relation which implies that the set of gauge orbits can be identified

as a quotient space [14]. If one defines a set of surface forming vectors, the gauge orbits
would correspond with the null vectors. [14]
To avoid this rather complicated situation of the quotient, it is often si, -.- -1. I that
one must impose a gauge choice to eliminate the redundancy. A gauge choice (Xa) has the
following property

Xa(p, q) 0, (3-3)

where a E {1,..., A} [14]. We must also choose such a function that intersects the gauge
orbits once and only once. A word of warning-One can guarantee this is the case locally;
however, it may not be guaranteed globally, (i.e. for the entire constraint surface). This
is known as the Gribov problem [16]. However, we are considering the ideal case for this

With this mathematical description established, we can now properly discuss the
Fadeev-Popov procedure. [27]
3.1.2 Basic Description

This method requires us to depart from the canonical quantization scheme as
described in ('! Ilpter 2. The philosophy of this method is to reduce the classical theory
first, and quantize second, which is yet another departure from the Dirac procedure from
C'!i lpter 2. For a first-class system, the formal path integral is given by

Jf (i/h) J H(p,q)Ab]dpq (3 4)

To solve the constraint problem in this framework, we assert the constraints are satisfied
classically within the functional integral by imposing a 6-functional of the constraints.
Since the resulting integral may be divergent, we suppress this possibility, by requiring a
choice of an auxiliary condition called a gauge fixing term of the form x (p, q) = 0, a

{1,.., J}. With this choice we have lost canonical covariance, which can be restored

One should also check the quantum equivalent to the consistency equation (2.1.2). In

essence this means we must consider

[Ka(P, Q),H(P,Q)] |)p = 0, (2-36)

[Ka(P, Q), Q b(P,Q)] 1 )p 0, (2-37)

where H(P, Q) is the unconstrained, self-adjoint Hamiltonian operator. Once again,

we are faced with a possible deficiency of the Dirac procedure. In general, we cannot

attest to the validity of these equations, but if we restrict our arguments to considering

only closed, first-class systems (2-36) and (2-37) will hold true. In the case of closed,

first-class systems, the Poisson brackets transforms into the commutator brackets, which

are expressed in the following form:

[ a(P, Q),H(P,Q)] = ihh Kb(P,Q), (2-38)

[ta(P, Q), b(P, Q)] hcbc(P, Q). (2-39)

If Equation (2-36) or (2-37) fails and the classical system is classified as first class, the

quantum system is said to have an anomaly. We will examine such a system in C'!i lpter

4. Furthermore, we find that our definition for the physical Hilbert space may be vacuous

when considering classically open, first-class or second-class systems since there may not

be a zero in the spectrum [13]. Dirac attempts to remedy the problem of second-class

constraints by redefining the Poisson bracket [7]. Therefore, the standard approach in the

Dirac procedure prefers closed, first-class systems. We will return to a discussion of the

Dirac bracket in ('!i lpter 8.

Another deficiency to note is the fact the Dirac procedure does not offer a definition

of the inner product of the physical Hilbert space. This, along with some of the other

deficiencies that are illustrated in this chapter, will serve as the primary motivation for

the discussion of the more modern methods to quantize constraint systems discussed in


8.1 Classical Consideration

8.1.1 Basic Model

Our discussion will begin considering a classical regular system with a single degree

of freedom, whose canonical variables are named p and q. Such a system can generally be

described by the action functional.

I / (pq- H(p,q))dt (8-1)

where = dq/dt and H(p, q) is the Hamiltonian of the system. The evolution of the

system is obtained by varying the functional with respect to the dynamical variables, this

reads as,

4 = a (82)
O aH (8-3)

subject to the suitable boundary conditions. As stated above, this system is purely

dynamical, however, it is well known any action can be converted into an equivalent action

that is a reparameterization invariant. [48] Let us begin this conversion by promoting

the dependent parameter t to a dynamical variable. This is appealing from a relativist's

perspective because the spatial and temporary coordinates are treated symmetrically. We

also must introduce the formal momentum pt conjugate to t. The integration variable in

(8 1) is now replaced by a new independent parameter 7-, which corresponds to proper

time or a more general function of time 1 We can express the reparameterization

1 Proper time is the time seen by an observer in the rest frame of a system. [6] However,
for our purposes we can consider the Lagrange multiplier are not strictly increasing see

formally with the aid of the Faddeev-Popov determinant, det({xb, 0c}) By determining

a particular gauge fixing term, the hope is to integrate overall gauge orbits. The ensuing

path integral becomes

I ) -H(p,'q))dt bI jXb b} det({xb, c0})DpDq. (3-5)

Expression (3-4) serves as a motivation to the introduction of (3-5), but they are not

to be viewed as equivalent statements. The result of (3-5) could then be expressed as a

path integral over the reduced phase space,

J exp{ [pq* H*(p* q*)]dtDpDq (3-6)

where p* and q* are reduced phase coordinates and H*(p*, q*) is the Hamiltonian of the

reduced phase space. Since we have satisfied the constraints classically, we are no longer

confident that our formal path integral is defined over Euclidean space. This presents

a dilemma since the formal path integral is ill-defined over non-Euclidean spaces [13].

As with the Dirac Procedure, the Faddeev method can be modified to accommodate

second-class constraints [28].

3.1.3 Comments and Criticisms

While the Faddeev-Popov procedure has yielded some of the most fruitful results in

physics [1], it is not without its flaws. One of the most glaring flaws is the fact that one

must first reduce the classical theory and then quantize it. The universe3 is quantum

mechanical; therefore, there may be some quantum mechanical correction to the classical

theory. Let us consider the following simple model to illustrate this fact. Consider the

3 At least up to the GUT energy scales [3]. We are not so bold to -iv that quantum
mechanics may be superceded by a more complete description of nature. Of course, we
assume however quantum mechanics is the proper route to look at nature until more
evidence is discovered.

convergent integral namely,

IE(, ab < 6(h)2) liM sin[(6(h)2 + (t2 t)] t) (4 12)
C--o+ -_^ 7r7

where the conditionally convergent integral is defined by the following [40];

-i sin[(6(h)2+ ()7] if II< 6,
lim / ed7
J 7 '0 if XI > 6.

Equation (4-12) is true assuming that the constraints are not explicitly dependent on

time. We examine that case in ('! lpters 7 and 8. The matrix Mab is the most general

case, but for our current purposes, we are free to select Mab ab, which would

yield the desired form of squaring the constraints. Having illustrated the motivation

behind considering squaring the constraints, we will divert our attentions to some of the

mathematical tools that are required in implementing the POF.

4.2 Tools of the Projection Operator Formalism

4.2.1 Coherent States

As one may recall from C'! Ilpter 2, one of the limitations of the Dirac procedure

is the lack of assurance of a normalizable state. To address this concern using the

projection operator method, let us consider the coherent state as a suitable Hilbert

space representation. Let P' and Qj denote the standard Heisenberg self-adjoint operators

that obey the commutation relation

[QP,P] ibiln. (4-13)

The Weyl (canonical) coherent state may be defined as

Ip, q) = e-( '' P e(/ '' 0) (4-14)

for a finite number of degrees of freedom, (p, q) E R2N, and the states are strongly

continuous in the labels (p, q), |0) is some fiducial vector often taken to be the ground

for all j, and thus both sets {Js} and {Li} are number conserving. We will revisit
this particular technique in C'! ipter 9. This conservation implies that we can study

the fulfillment of both sets of constraints in each of the number-operator subspaces

independently of one another. We observe that the subspace for which N = 0 consists

of just a single state, and this state is an eigenvector of each Ji as well as each Li,

i {1, 2, 3}, all with eigenvalue zero. In the interest of simplicity in this paper, we restrict

our attention to the lowest nontrivial subspace in which the constraints Ji = 0 are satisfied
on a non-vanishing subspace. In particular, we confine our attention to a subspace of the

entire Hilbert space corresponding to an eigenvalue of the total number operator of two.

Note that the subspace of interest is six-dimensional and that it is spanned by the six

vectors given by the two representatives

1, 1,0) a a 0), etc., (4-46)

12,0,0) (1/v) (a{)2 0), etc., (4-47)

where as usual 10) (= 0, 0, 0)) denotes the no particle state for which aj|0) = 0 for all j.

The first non-empty subspace that produces a non-trival result is the 2-particle

subspace. With the additional simplification that 7 /3/2, we can express the eigenevector

that corresponds to least eigenvalue in this 6-dimensional subspace as,

OL) 1
OL) ( (12 d|2, 0,0) + d' 10,2, 0) + 00,0,2), (4-48)
1 + d2+ dlI

where d = 1 23 + O(32) and d' = 1 + O(02). The projection operator of this subspace

is constructed as the following;

E2 =OL) (OL| (4-49)







operator that projects onto the intersection of the two sets,

lim (E(B)E(B2))n = E(BI n B2). (4-2)
n- 0oo

If B1 n B2 = 0, then E(0) = 0. We will use this property in ('!C ipters 5, 6 and 8.

Reverting to the Dirac prescription of the physical Hilbert space it is defined as

NFp {- ))p )p = 0, Va} Q-{ker4} (4-3)

where Ka is the quantum analogue to the classical constraint a,, and a E {1,..., A}. In an

ideal situation3 (4 3) is equivalent to the following

Np {kert a} {ker ac,}. (4-4)

The fact that (4-4) will not ah--iv-i lead to a non-trival result, is a clue on how to arrive

at the true answer. Assuming that a~K, is self-adjoint acting on a Hilbert space, we can

use the following result from spectral theory to obtain our desired projection operator, E.

Namely, the operator 4Kaa can be written in the following representation [39]

a@a= J AdE (4-5)

where dE is the so-called projection valued measure [33] on the spectrum of 4a4a, which

was denoted by A which contained a spectral range of 0 to oc4 The projection operator

that was used in (4-1) can be introduced based on the result of (4-5)

E(aa < 6(h)2) = AdE (4-6)

3 Which is generally not the case for reasons mentioned in ('!, plter 3 and later in this

4 This range may or may have not included 0.

constraint subspace as

p(p, q) = IE /,K (p, q; p, qn), N < o0. (5-11)

The inner product for these vectors is given by
(Qqj)p > n,rn-I

where rT is also an element of the dense set of vectors. The completion of these vectors

in the sense of Cn 1.r! sequences with the relevant inner product will yield the physical

Hilbert space.

Without explicitly calculating the reproducing kernel, we will consider the following

highly irregular quantum constraint

+ = Q2(1- Q), (5-13)

where Q acts as a multiplication operator. Clearly this constraint vanishes when Q = 0

and Q = 1. Assuming, 0 < 6 < 1, the projection operator for this constraint can be

written in the following form

E(-6 < K < 6) = E(-6 < Q2 < 6) + E(-6 < (1- Q) < 6). (5-14)

Since the zeros of this operator fall in the continuum, it is clear from the previous

discussion we cannot take the limit 6 -- 0 in its present naked form. The reproducing

kernel can be expressed as the following

Co (p ', q'E(-6 < Q2 < 6)p,q) + (p',q'IE(-6 < Q 1 < 6)p,q). (5-15)

By construction these projection operators E(-6 < Q2 < 6) and E(-6 < Q 1 < 6)

project onto orthogonal spaces. To leading order in 6(h) the reproducing kernel can be

where the essential change resides in the definition of the variables li. For some smooth,
non-vanishing function, f, we define (note: q = q1, etc.)

1- f(pi,p2, qq29, (4 38)

for all i, and choose for further study the particular example for which

f(Pi, 2, q, q2) a- + (P3/A)(p + q1) + (7/ )(p + q2). (4-39)

The symbol h is a fixed constant equal in value to the physical value of Planck's constant
h, namely 1.06 x 10-27 erg-sec. When the classical limit is called for, and thus Planck's
constant h -- 0, we emphasize that h retains its original numerical value. The reason for
such a small divisor is to emphasize the quantum corrections; different divisors can be
considered by rescaling 3 and 7. We recognize, in this simple case, that we could absorb
the factor f by a redefinition of the Lagrange multipliers in (4-37). In more complicated
systems (e.g., gravity) this simplification is either extremely difficult or perhaps even
impossible. Therefore, as a further analog, we retain f as a part of la. A straight forward
analysis leads to

{li, lj} = { fj, fjj }

f2{ji,jj} + f{ji, f},jj + f{f, jjji + {f, f}jijj

fijklk + {jj, f}lj + {f, jj}lt

f ijklk + Ciab[--qaOf /aqb + PbOf /ll' ]l

jab[- qaOf /aqb +pbOf /lj'']i .

Since f > 0, our modified set of constraints is classified as open, first-class.

eliminate the many technical details surrounding the mathematical machinery -,'' We

will motivate this quantum constraint program in much the same manner the original

authors did, [33]. Particularly, we will describe the classical analog and then discuss the

quantization of the classical theory.

3.3.1 Classical Description

Given a phase space M and a set of constraints functions Cj(p, q)jil, where I is some

countable index set and (p, q) E R2N and 2N is the dimensionality of the phase space, the

master constraint replaces this set of constraints with a single expression, which is sum of

the square of the constraint operators in a strictly positive semi-definite form, as shown in

the following expressions:

M CJ(pq)gJkCk(pq) (3 17)
where gi is chosen to be positive definite.o1 We will attempt to justify this act of adding

the squares of constraint functions in the next chapter, as well as discuss some of the

potential pitfalls of this procedure. The set of constraint equations Cj = 0 for all j E I

has now been reduced to a single equation, M = 0. Despite being a great simplification

another difficulty immediately presents itself. Namely, how can we recognize observables in

the theory? As we noted in C'i plter 2, observables are functions over the phase space that

commute weakly with all the constraint functions

{O, C (p,q)} 0 (3-18)

9 For a technical account of this program we refer the reader to the seminal works on
this program namely [33] and [34]

10 We should also note we can make a further modification on (3-17) if the constraints
are actually fields. If this is the case, we must smear them over some set of test functions.
For more details on this procedure see [33].


"... ,, ;,,,,/ all ,ql.-I -, problems in pi.;;-. in the p',.-f.lund issues of principle that

confront us i.,.~n no difficulties are more central than those associated with the concept of

time ..." John Archibald Wheeler

Time is a crucial element to any dynamical system; it is the evolution parameter

of such a system. The nature of time is an extremely popular topic covered by many

physicists, as well as, philosophers [22]. While the physical (or meta-physical) nature of

time is outside the main focus of this dissertation, time-dependency in quantum mechanics

offers us an interesting caveat to explore and study.

In the methodologies developed and discussed in the previous chapters, the primary

goal was to solve quantum mechanical time-independent constraints. The exclusion of time

was made primarily out of simplification. In most of the literature about constraints [12]

the topic of time-dependent constraints is either briefly covered or it is not covered at all.

However, it is clear that for a more complete discussion of constraint dynamics we must

also include constraints that are explicitly dependent on time. Time dependence can enter

a dynamical system through the Hamilitonian, constraints, or in the most general case

or combination of the two. The inclusion of explicitly time-dependent constraints offers

not only an interesting academic exercise but also gives physicists the tools required to

examine more physical theories than those that previously could be discussed.

This is not the first occasion on which the projection operator has been used to deal

with the case of time-dependent constraints. In [47], Klauder derived an expression for

evolution operator of time-dependent constraint. The construction of this expression was

based on modifying the expression for the time-independent case. Although this expression

seemed to be correct the author chose not to pursue this subject matter further.

Primarily, he made this choice because the formula did not reduce to a simpler operator

expression. While the projection operator will be the primary mode of exploration

hold true, we categorize this type of constraint as first class. We can make a further

refinement of this class by considering the nature of the structure coefficients, hb and

cab. If the coefficients are constants, then the constraints are closed first class. If they are

functions over phase space, then the constraints are open first class. As one can deduce

from (), the Lagrange multipliers are undetermined by the equations of motion and thus

can be arbitrarily chosen, a phenomenon called "choosing a g ,, ; Therefore, first-class

systems are said to be gauge systems. Based upon this definition of a first-class system,

we can assert that once the dynamics are restricted to the constraint surface, initially,

they will alv--,v- remain on the surface. Well known examples of first-class systems include

Yang-Mills theories and General Relativity, with the former being closed and the latter

being open.

If det{( Ob} / 0 the constraints are classified as second class [16]. No longer

having the availability of the preceding criteria of (2-24) and (2-25), it follows that the

Lagrange multipliers are determined by the equations of motion so that (2-22) is satisfied.

The Lagrange multipliers force the dynamics to remain on the constraint surface for a

second-class system. Namely the Lagrange multipliers can be determined by the following


A b _[{a(p, ),Ob(p, q)}1{0(p, q), H(p, q)}. (2-26)

2.1.3 Another Geometric Interlude

When the dynamics are restricted to a sub-manifold in the phase space, some of the

mathematical structures present in the entire phase space are no longer present. Most

notably is the symplectic 2-form w. One could imagine looking at a particular coordinate

patch of the sub-manifold and determining the symplectic form for that particular patch.

However, if we attempt to repeat this process for the entire sub-manifold, we would find

that there exist some regions in which the 2-form is degenerate. Often this degenerate

are both regular and irregular. In this general setting we can classify the AHB constraint

as a highly irregular constraint.

6.3 Quantum Dynamics

We now proceed to canonically quantize the system (6-1). We will assume our chosen

canonical coordinates are Cartesian ones suitable for quantization [20]. We then promote

the canonical dynamical variables (x, y,px,Py) to a set of irreducible self-adjoint operators

(X, Y, PF, Py). Conjugate pairs corresponding to compact, periodic spatial components

will not obey the standard Heisenberg-Weyl relationship [45] because the eigenvalues of

the conjugate momentum operators are not continuous but discrete. Continuing with the

canonical quantization procedure, we promote the constraint to a suitable function of

self-adjoint operators

C C <= P- R(Y). (63)

Note, there is no ordering ambiguity for this operator. We assume the constraint operator

is a self-adjoint operator in the unconstrained Hilbert space. We can now implement the

quantum constraint using the projection operator method.

6.4 The Physical Hilbert Space via the Reproducing Kernel

The projection operator for the Ashtekar-Horowitz-Boulware model is chosen to be

E(C2 < 62) = E(-6

Since the function R(y) is a continuous function, we must introduce an appropriate set of

bras and kets to deal with the subtleties described in section 2.

6.4.1 The Torus T2

Before constructing the model with the configuration space of a torus we must

determine the correct coherent states to use. We wish to use the coherent states not only

for computational ease, but also to determine the classical limit, which will be addressed

later in this chapter. The torus is the Cartesian product of 2 circles. It follows that the

the Ashtekar-Horowitz-Boulware model. We will also offer a comparison of the results

obtained from the Projection Operator Formalism with that of the Refined Algebraic

Quantization scheme.

Finally, we will use the Projection Operator Method to discuss time-dependent

quantum constraints. In doing so, we will develop the formalism and study a few

key time-dependent models to help us obtain a larger picture on how to deal with

reparameterization invariant theories such as General Relativity.

E(-6 < J sin(t) + 3 cos(-t) < 6)
2 2

S E(iL < 62(h))

+ Et(-6 < J sin( t) + J3cos( t) < 6 /L262) (9-10)
2 2

where Et represents the explicit time-dependent nature of the second operator on the

right-hand side of equation (9-10). Having discovered the full nature of the projection

operator let us digress a bit to briefly discuss simplifications to equation (8-46), with the

given description of the model. As we can easily observe, the Hamiltonian H commutes

strongly with the constraint Q(p, q), it follows that;

[En, H] = 0. (9-11)

This equation holds for all time-slices and therefore all n. Equation (8-46) then reduces in

the following manner;

(.|e i /6 E 1 ... E) )

(. {I-|/TEN- 1N-1,0 I) (9-12)

where E, is defined in equation (8-47). As we can determine from (9-12), the infinite

product of projection operators will merge into one projection operator. In turn this will

project onto the set which is the intersection of the initial projected space (i.e. J = 0)

and the final (i.e. J = 0). This operator will of course project onto the sub-space that

carries the trivial representation of this algebra, which is the time-independent portion of

the projection operator. The conclusion that we can draw from (9-12) is that the physical

Hilbert space for this model is not trivial. While this model may not be conclusive proof

that the physical Hilbert space for a general time-dependent constraint is not trivial. It

with the help of various generalizations of this model assist in answering the full query.

state of a harmonic oscillator9 The additional requirement that (4-14) truly are coherent

states is that they possess a resolution of unity [41]:

= p,q)(p, q (4-15)

We will offer (4-15) as an accepted truth without proof [41].

These coherent states also offer a connection to the classical limit of quantum

operators. This property is known as the I. Il: correspondence principle". [41] We

exploit, and also state more carefully, this property of coherent states in a subsequent


The coherent states are convenient because they form an overcomplete basis of the

Hilbert space. Using this representation, we can express a dense set of vectors in the

functional Hilbert space in terms of the coherent state overlap,

{(p,q)= (p, q ) aa(p, q p, q, N < oo, (4-16)
where a, E C. The inner product of such vectors can be expressed as the following,

(QqT) a*M.(pq.Pn,.n1pm, q), (4-17)

where rT is an element of the dense set. The completion of such a set of vectors leads to

the unconstrained Hilbert space, which leads us to the topic of the Reproducing Kernel

Hilbert space.

4.2.2 Reproducing Kernel Hilbert Spaces

Reproducing Kernel Hilbert space is well established, yet it is an under utilized

mathematical technique to describe functional Hilbert spaces. If a reproducing kernel can

be defined, it will completely define the space. One such example of reproducing kernel

9 However, one could select another fiducial vector depending on the situation.

The quantization of this model proceeds much like that of the case with Ji, namely we
promote 1i to a suitable self-adjoint quantum operator Li

SH L, = aJ + (j3/2h) [(P12 + Q2) Ji + J, (P2 + Q2)1

(y/2h) [(p2 + Q2) Ji + Ji (p2 + Q2)], (4 40)
+ (7/2) [(P22 P22 (4-40)

where a + p + 7 1. In dealing with the quantum theory, we drop the distinction between
h and h. A quick calculation shows that any other factor-ordering of the definition of
(4-40) will yield an equivalent result. The commutation of the Li yields a surprising
result, namely

[Li, Lk] Cijk(ihFLk h2 (Okll(aka + alak)))

2iii((QaQl + aP1l)F +... (4-41)

where F = a + (p/h)(P2 + Q) + (/)(P22 + Q2). The second, third, fourth, etc. terms
in (4-41) represent the anomaly in the quantum theory. This anomaly corresponds to a
transmutation to a partially second class system. With this being noted we will continue
the quantum analysis of the system.
Let us introduce conventional annihilation and creation operators represented by

a, (Qj + iPj)/ 2, (4-42)

a = (Qj iPj)/2 (4-43)

If we define

N = aai + ata2 + aa3 (4-44)

as the total number operator, it is evident that

[Jj, N] =0, [Lj,N]= 0, (4-45)

The coherent states, that will be again useful in this analysis, satisfy lp, q, pt, t) = P, q) 0

lpt, t). In order to determine our projection operator E we need to determine a set of
normalized vector states I|) to minimize the following relationship

(|l|(P2 + (Q cT)2 + (P c)) 2 o h (9-19)

Using the logic emploiv -1 by Klauder in [13], the state that minimized (9-19) would be

the following state Ict, c) 0 10, t) Using the theory of Weyl operators we can construct a

representation of the desired projection E, given by

|c,ct) 0 |0,t)(0 (c,ct =t -iA(Pt) -i(Q-cT)-(P-c)(-A2 /4dAd .(9-20)

We should observe that in this case f e(-"2 2-2)/4 dA 2, rather than one as would

be the case for normalized measure. The consequences of this projection operator can

immediately be discerned following the procedure described in [13].

9.3 Conclusions

In this chapter, we successfully dealt with two systems that have explicit time

dependence. It is hoped that techniques that we developed in the past two chapters will

help us have a better understanding on how to deal with quantization of theories that

are reparameterization invariant, such as General Relativity. The methods discussed and

developed in the last two chapters can be extended to include field theories.

It is well known that a field theory that is not reparameterization invariant can

be transmuted into one reparameterization invariant by a similar technique to that

emploiv, 1 in ('! i lter 8 [55]. Namely this can be accomplished by changing the space-time

coordinates (x" where pe {1..., N}, and N is the number of space time dimensions) to

functions over space-time,which in essence introduces N scalar fields to the field theory,

xP" y- (x). (9-21)

function /C2, which leads to

/C2(p, q";p', q) u lim/C2(p ;p,q')
C- [(2-q")2+(2-q')2]/2-i(p"-p) (5-48)

This procedure leads to a new function k2, which, provided it is still continuous

which it is -leads to a reduced reproducing kernel and thereby also to a new reproducing

kernel Hilbert space '72 Generally, the dimensionality of the space as well as the

definition of the inner product are different for the new reproducing kernel Hilbert space;

however, one inl,. ,;- has the standard inner product definition that is appropriate for any

reproducing kernel Hilbert space [42]. In the present case, it follows that /2C defines a

one-dimensional Hilbert space 712 Note that even though the coordinate value for the

constrained coordinate Q is now set at Q = 2 -as is clear from the special dependence of

/12(p", q"; p', q') on p" and p' the range of the values q" and q' is still the whole real line.
The only remnant that q" and q' have of their physical significance is that kC2(p", q"; p, q')

peaks at q" = q' = 2. It is noteworthy that an example of this type of irregular constraint

was considered previously by [40].

A similar procedure is carried out for the remaining components in the original

reproducing kernel. Let us next consider

IC(p", q"; p', q') (p", qE(- < < < 6)|p', q')
1 il f (x--q")2/2-i(p"-p')x6-(x-q')2/2ld

where 61 6/2. To leading order

C, (p", q"; p', q') -= e-(q"+q')/2 i (5-49)
V7T 6 p"( p')

which is a function of positive type. It is noteworthy to note that this constraint is of

regular type. [14]

0-particle subspace3

T pi
E(-6 < J sin( t) + j3 cos(tt) < 6) = 10, 0, 0) (0, 0, 01 (9-7)
2 2

1-particle subspace

E (-6 < Ji sin(t) +3 cos( t) < 6)
2 2
S (sintl|, 0,0) + cos t|0,0,1))((1,0, 0 sint + (0,0, 1| cost) (9-8)

2-particle subspace

E2(-6 < J sin(t) +j3cos( t) < 6)
2 2

S(12, 0, 0) + |0, 2, 0) + |0, 0, 2))((2, 0, 0| + (0, 2, 0| + (0, 0, 21)
+ -((sin(t) + 1)12,0, 0) + 10,2,0)+ 1 + cos(t) 10, 0, 2))
3 2 2
x (sin( t)+ 1)(2,0,01 + (0,2,01 + (1 +cos(t))(0, 0, 2) (9-9)
2 2

The construction of the higher numbered projection operators continues in a similar

fashion. The key observation at this juncture is that the projection operator decomposes

into a time-dependent part and a time-independent portion. The time-independent part is

associated with the Casmir operator from the full su(2) algebra, while the time-dependent

portion is attributed to the remainder of the constraint modulo contribution from EsiL

0. Therfore we can write the full projection operator as the following:

3 The superscript on the left-hand side of the equations designates the number subspace


The primary objective of this chapter is to examine three distinct methods to deal

with quantum constraints. These methods are the Faddeev-Popov procedure, the Refined

Algebraic Quantization Program, and the Master Constraint Program, each of which has

its own distinct strengths and weaknesses. During this chapter we will use the notation

that is standard in literature, while also noting deficiencies of the methods in order to offer

more motivation for the study of the projection operator method, which is the topic of

('! ipter 4.

3.1 Faddeev-Popov Method

3.1.1 Yet Another Geometric Interlude from the Constraint Sub-Manifold

Before discussing the Faddeev-Popov method, it is important to discuss some

properties of the constraint sub-manifold that were neglected in the previous chapter.

Let us begin with the following classical action,

I = dt(pj H(p, q) Aa), (3-1)

where j E {1,..., N}, a E {1,..., A}, and A" are Lagrange mulipliers which enforce the

constraints Qa1 We will consider the situation when the constraints that are present in a

classical system are closed, first-class. The constraint submanifold can also be identified

as the space of gauge orbits [14]. A gauge orbit is defined by the following: consider that

F defines a particular physical configuration,2 and the gauge orbit of F consists of all

gauge-equivalent configurations to F. [12] A gauge transformation is defined as

6,F F= v{F,Oa}, (3-2)

1 We are still under the assumption from C'!i lpter 2 that the constraints are regular
(which will be defined in ('! Ilpter 5) and complete.
2 One could consider F to be a dynamical variable.

which can be accomplished as long as the conjugate momentum (2-5) is invertible in terms

of q. This condition is satisfied by the Hessian condition

det / 0. (2-8)

We will return to the case when (2-8) fails shortly. Until then we will discuss the

Hamiltonian formalism in more detail. For a more complete account see [17] and [15]. At

this point we will no longer mention the cotangent bundle, but rather we will note that

this space is symplectomorphic to the more familiar space, phase space AM, i.e.,

T*Q .M (2-9)

The geometric framework of the Hamiltonian framework is a rich and beautiful

subject. However, for the sake of brevity we will only recount the most crucial elements to

the development of the constraint picture. For a more complete description, we point the

readers to [17], [15], and [18].

2.1.1 Geometric Playground

The natural geometric framework for Hamiltonian dynamics is a 2 n dimensional,

real, symplectic manifold called phase space, where n is the number of degrees of freedom

for a system.4 The coordinates on the manifold are determined by the equations of

motion. As is well known, the equations of motion for an unconstrained Hamiltonian (2-7)

are given by

q = {,H}, (2-10)

pj {pj, H}, (2-11)

4 This definition of the phase-space manifold is true when the system has a finite
number of degrees of freedom. Extra care must be taken in the definition when dealing
with a case with an infinite number of degrees of freedom.

corresponding to a one-dimensional Hilbert space R73. For /C3 we are led to

3(p", q"; p', q') = (p", q" IE(2 < Q < 3)lp', q')
1 4
J e-(x-q")2/2-i(p"-p')xz-(x-q')2/2dx, (5-53)

In this case, no 6 appears and no infinite rescaling is needed, so we may simply choose

3 (p, q"; p',q') C3(p", q"; p', q'). (5-54)

Although we do not have an explicit analytic expression for /3 we do have a well-defined

integral representation in (5-53). Furthermore, it follows that 73 is infinite dimensional.

Finally, we define the reproducing kernel for the physical Hilbert space as

IC(p", q"; p', q') ,I1(p", q"; p', q') (5-55)

In turn, the physical Hilbert space 7'ip is defined as the reproducing kernel Hilbert space

i uniquely determined by the reproducing kernel /(p", q"; p', q').

Observe, by our procedure, all elements of the reduced classical phase space (R x Z)

are represented on an equivalent basis in K This feature has been designed so that

the classical limit of the expressions within 'ip correspond to all aspects of the reduced

classical phase space. We will now turn to a discussion of observables of this model.

5.3 Observables

Let us restrict our discussions of observables to those that are self-adjoint operators 0

in the unconstrained Hilbert space. We also limit to constraints that are both classically

and quantum mechanically first class2 We first discuss the situation in the case of a

regularized (6 > 0) enforcement of the constraint [('!i plter 4]. In this case, all physical

2 These arguments can be extended to include both quantum mechanical second-class
and first class with an anomaly. The exclusion is made for the sake of the model under

that satisifies the following condition: For every Q01, 2 E then qr(01) is a solution of the
constraint equation,

(Ct(rlT0))[2 = ( )[C ] =0. (3-15)

In addition to (3-15) the rigging map (3-14) must also satisfy the following two

conditions, which are true for every 01, 02 E :

1. The rigging map Tr is real and positive semi-definite

(Tl/)[021] (92)[]11,*

(r101)[Oi] > 0.

2. The rigging map inteterwines with the representations of the obervable algebra

O(TI01) I(01),

where O E Aobs-

Once the rigging map has been determined, the vectors rl that span the solution

space are C 1 r,, completed with respect to the following inner-product

(011u| 2) (01)[02], (3-16)

for every 01, 02 E ) and (.|-) is the inner product of the auxiliary Hilbert space. Thus,

we define the physical Hilbert space derived by the techniques of the RAQ. We will

revisit the Refined Algebraic Quantization program in C'i plter 6 in the context of the

Ashtekar-Horowitz-Boulware model [32].

3.2.2 Comments and Criticisms

While the RAQ does resolve some of the ambiguities of the Dirac procedure, the

resolution is not without cost. One of the prices that we must 1p is that we must

also have an additional mathematical structure on particular subspaces on the Hilbert

space. Namely, we require that the invariant subspace must also be equipped with a

topology that is finer than the one inherited by the auxiliary Hilbert space. As is well

The constraint will clearly commute with the Hamiltonian, H = (p + p + p) + (q +

q2 + q3) therefore this truly is a first-class system. For the case of this analysis we will
restrict the allowable values of t to a compact subspace of R namely, [0, 1]. It is easy to

observe that the constraint surface initially is defined by j3 = 0 but evolves in a smooth

fashion into the vanishing loci of ji. Moving to to quantum analog of this system2 the

issue surrounding the alternating constraints is potentially very interesting since J3 and

Jl are examples of incompatible observables, therefore it is impossible to diagonalize them

simultaneously. Utilizing the technique in which we established the physical Hilbert space

for the Casmir operator of su(2) in C'! lpter 4, Section 3.2, we will use a similar technique

to analyze this model. As before, let us introduce conventional annihilation and creation

operators given by

a (Q + iPj)/ (9-3)

a = Qj- iPj)/v2h. (9-4)

If we define the number operator

N = aa + a a2 + aa3 (9-5)

it is obvious that

[J, N] 0 =[J3, N]. (9-6)

Based on this conservation, we can study the fulfillment of the time-dependent constraints

in each of the number-operator subspaces independently of one another. Based on this

information we can proceed with the following analysis:

2 Since this model is similar to the example discussed in C'! lpter 4, we will forgo the
formal arguments of the quantization scheme.


The primary goal of this chapter is to introduce and motivate the Projection Operator

Formalism (POF). The projection operator method is a relatively new procedure for

dealing with quantum constraints [13] [26].The ]lhi ... l.hi,- of this formalism is to first

quantize the entire theory, and then reduce the quantum theory by using the constraints.

We will attempt to illustrate how the POF attempts to remedy some of the deficiencies of

the methods discussed in C'! lpter 3. In the final section of the chapter we will examine

three constraint models. The first is a constraint that has a zero in the continuum,

whereas the second and the third are models that were examined in [8]. They help

illustrate the power of the projection operator formalism in dealing with all classifications

of constraints. In this chapter it is understood that h = 1 unless stated elsewise.

4.1 Method and Motivation

Following the Dirac procedure's initial footsteps, we canonically quantize the

unconstrained classical theory as described in the preceding section. We then deviate

from the Dirac method by introducing a projection operator, E, which takes vectors from

the unconstrained Hilbert space to the constraint subspace (i.e. the physical Hilbert space

or even better the regularized1 physical Hilbert space) [13]

-Hp =- E (4-1)

We require E to be Hermitian which satisfies the relation E2 = E (idempotent), these are

basic properties of a projection operator. More precisely, suppose that B1 and B2 denote

measurable2 sets on the Hilbert space. The product of two operators yields a projection

1 We will explain this more clearly a little later in the chapter.
2 Borel measurable [33]

where S,(6) > 0 for all n which leads to

K = Sn(6)/,. (5-19)

The rescaled K serves as the reproducing kernel for the Hilbert space H. Although the

inner product of R and H) are different the set of functions are identical. The goal of this

little exercise is of course to take a suitable limit 6 -- 0 to yield a function that can serve

as a reproducing kernel for the physical Hilbert space. At this point, we can take such a


K = lim C, (5-20)
'Hphys -H (5-21)

where K is the reduced reproducing kernel for the physical Hilbert space 'Hphys. Having

discussed the basic theory behind this classification of constraint, in the next section

we will consider a simple but robust toy model that demonstrates the strength of the

Projecion Operator Formalism to deal with these kinds of constraints.

5.2 Toy Model

The model we choose to study involves just one configuration variable q, -oo < q <

oo, and its conjugate variable p. The classical action is taken to be

I f[p- AR(q)]dt, (5-22)

where A is a Lagrange multiplier designed to enforce the single constraint

R(q)= 0. (5-23)

constraints which satisfy the following A equations:

1 0
02 0

OA = 0.


In order to determine the constraint subspace in the phase space, all A equations must be

satisfied simultaneously. This set of equations would be at least classically equivalent to

the following set:


Finally, if we add all of the preceding equations together, we arrive at the conclusion that

S-= 0o


is equivalent to the set of A equations (4-7). As we stated before, this is classically

equivalent, but are we certain that this will be justified quantum mechanically? Quantum Consideration

When moving to operators, the next point of concern is whether or not the procedure

of summing the squares of operators is well defined. Since we are assuming (I is

4.2 Tools of the Projection Operator Formalism .
4.2.1 Coherent States .. ...........
4.2.2 Reproducing Kernel Hilbert Spaces .
4.3 Constraint Examples .. ............
4.3.1 Constraint with a Zero in the Continuous
4.3.2 Closed, First-Class Constrant .. ....
4.3.3 Open, First Class constraint .......
4.4 Conclusions . . . . .



Classification .. ..........
Toy M odel .. ............
Observables .. ...........
Observation and Conclusions .


6.1 Introduction . . . . .
6.2 Classical Theory .. ..............
6.3 Quantum Dynamics .. .............
6.4 The Physical Hilbert Space via the Reproducing
6.4.1 The Torus T2 ........
6.5 Super-selection Sectors? .. ..........
6.6 Classical Lim it . .. .. .. .. .. ... .
6.7 Refined Algebraic Quantization Approach .
6.8 Commentary and Discussion .. .........


7 PROBLEM W ITH TIME .. . ........................


8.1 Classical Consideration . ...
8.1.1 Basic M odel . ......
8.1.2 Commentary and Discussion .
8.2 Quantum Considerations . ..
8.2.1 Gitman and Tyutin Prescription
Constraints . ......
8.2.2 Canonical Quantization . .
8.2.3 D irac . . . .
8.2.4 Projection Operator Formalism


Time-Dependent Quantum Const
Observations and Comparisons .

for Time-Dependent

raints .........
. ..


9 TIME-DEPENDENT MODELS .............. . . ...

9.1 First-Class Constraint ........... . . .
9.2 Second Class Constraint .......... . . .





1.2 Outline of the Remaining Chapters

C'!I ipters 2, 3 and 4 serve as the background for the dissertation. The main topic

of discussion in C'! iplter 2 is the introduction of constraint dynamics as well as the

description of the Dirac procedure to deal with quantum constraints. The primary goal

of ('!C lpter 3 is to introduce the reader to three other alternative programs to deal with

the problem of quantum constraints. These methods are the Fadeev-Popov procedure,

the Refined Algebraic Quantization Program, and the Master Constraint Program,

each of which has its own distinct strengths and weaknesses. The goal of Chapter 4 is

to examine the projection operator formalism. In this chapter we will also exam three

distinct constraint examples in this formalism. One of the constraint models is a system

where the constraints are classically first class; however, upon quantization they become

partially second class, similar to the constraints of gravity. This particular model served

as the basis of [8]. Whenever encountered in the dissertation, repeated indices are to be


The primary goal of C'!i lpter 5 is to introduce a i, v.-" classification of constraints

called highly ., ,. glr, constraints and also illustrate techniques used to deal with

quantum versions of these constraints. The basis of this chapter comes from [9] and

[10]. Using the techniques gained from C'!i lpter 5, in C'!i lpter 6 we offer a complete

discussion of the quantization of the Ashtekar-Horowitz-Boulware Model [11]. The

Ashtekar-Horowitz-Boulware model is a mathematical model also inspired by the

constraints of gravity. This chapter is based on the results obtained in [9]. We also

compare the results obtained by the Refined Algebraic Quantization program with the

Projection Operator formalism. This comparison leads to the conclusion that the two

methods are not compatible dealing with all constraints.

The remaining chapters are devoted to the topic of time-dependent quantum

constraints. Until now, the methods used to delve into this topic [12] have been unsatisfactory

due to the fact that these methods avoid solving for the quantum constraints. The aim

[33] B. Dittrich, T. Thiemann, "Testing the Master Constraint Programme for Loop
Quantum Gravity I. General Framework," Class. Quant. Gray. 23, (2006), 1025.

[34] T. Thiemann, "The Phoenix Project: Master Constraint Programme for Loop
Quantum Gravity," Class. Quant. Gray. 23, (2006), 2211.

[35] B. Dittrich, T. Thiemann, "Testing the Master Constraint Programme for Loop
Quantum Gravity II. Finite Dimensional Systems," in Class. Quant. Gray. 23,
(2006), 1067.

[36] B. Dittrich, T. Thiemann, "Testing the Master Constraint Programme for Loop
Quantum Gravity III. SL(2,R) Models," in Class. Quant. Gray., 23, (2006) 1089.

[37] B. Dittrich, T. Thiemann, "Testing the Master Constraint Programme for Loop
Quantum Gravity V. Interacting Field Theories, in Class. Quant. Gray. 23, (2006),

[38] R. Shankar, Principles of Quantum Mechanics, Plenum Press, 2nd ed., New York,
New York, 1994.

[39] J. Klauder, "Path Integrals, and Classical and Quantum Constraints" http:
//arxiv. org/abs/quant-ph/0507222 (2005).

[40] J. Klauder, "Coherent State Quantization of Constraint Systems," Ann. Phys. 245
(1997) 419.

[41] J. R. Klauder, B. Skagerstam, Coherent States, World Scientific Publishing,
Singapore, 1985.

[42] N. Aronszajn, Proc. C.,n,,1, l1.:,: Phil. Soc., 39 (1943), 133; Trans. Amer. Math
Soc. 68 337 (1950); H. Meschkowski, Hilbertsche Raum mit Kernfunktion,'
Springer-V i1 :- Berlin, 1962

[43] H. Araki, "Hamiltonian Formalism and the Canonical Commutation Relations in
Quantum Field Theory," J. Math. Phys 1 (1960), 492.

[44] D. G. Boulware, "Comment on 'On the Canonical Approach to Quantum Gravity', "
Phys. Rev. D 28 (1983), 414.

[45] J. A Gonzlez, M. A. del Olmo, "Coherent States on the Circle," J. Phys. A: Math.
Gen. 31 (1998),8841.

[46] J. Louko, "Group Averaging,Positive Definiteness and Duperselection Sectors," J.
Phys. Conf. Ser. 33 (2006), 142.

[47] J. Klauder, Universal Procedure for Enforcing Quantum Constraints ", Nucl.
Phys. B547 (1999), 397.

[48] G. Fulop, D. M. Gitman, I.V. Tyutin. "Reparametrization Invariance as Gauge
Symmetry," Int. J. Theor. Phys. 38 (1999) 1941.



In the preceding chapter, we have developed an approach to contend with explicit

time-dependency in constraints within the projection operator formalism. Despite this

development, some looming questions persist. The primary purpose of this chapter is to

elucidate these unresolved questions by considering some simple quantum mechanical

models. One of the most pressing questions is whether or not the physical Hilbert space

of a time-dependent constraint is trivial 1 As shown in the previous chapter, (8-46), the

I. ,!, Ii i, i operator" for time-dependent constraints can be written as an infinite product

of projection operators. However, as we will illustrate in our first model, even with the

requirement of a stringent polarization of the states from the total Hilbert space, the

physical Hilbert space is non-trivial. The second model is designed to demonstrate how a

second-class system should be considered within this context.

9.1 First-Class Constraint

We deviate from the prescription described in the previous chapter by not pursuing

the reparameterization invariant form of the model discussed briefly. The primary

motivation of this model is stated above. Let us begin with the simple 3 degree-of-freedom

classical extended Hamiltonian.

1 12 )
HE (P +P + ) + (q + q + q32 + (t)(ji sin(j-t)+ j3 costt) (9-1)
2 2 2 2

where jl = q2p3 q3p2 and j3 = p2ql q2p and A(t) is the Lagrange multiplier that

enforces the single first-class constraint

0(p, q) = j sin( t) + jcos(t). (9-2)

1 Which is to that the physical Hilbert space contains only the zero vector.

1 Which is to I that the physical Hilbert space contains only the zero vector.

This unitary operator acts to translate the operator P, in the following manner

UxPxU = P h. (6-22)

As observed in [9], the observable part of an operator can ah--iv- be expressed as

OE = EOE, (6-23)

where 0 is a self-adjoint operator in the unconstrained Hilbert space.

The observable part of the Hermitian combination of U, and Ux is

W, = UEUE = EUIEULE. (6-24)

By observation, we note

W E(-6 < Px R(Y) < 6)E(-6 < (P h)2 R(Y) < 6). (6-25)

These projection operators are acting on mutually orthogonal subspaces; therefore, the

operator is identically zero. This result informs us that this is a gauge dependent question

which is consistent with the classical picture. Recall from Section 3 the x dynamical

variable is gauge independent only when px 0. Quantum mechanically, we have

posed the question to find a pl.y-ical" wave function that has support on both a gauge

independent sector and gauge dependent sector. This is impossible.

If we were to examine the same query for the corresponding Hermitian combination

of the in,!, operator for the Y coordinate, we would obtain the unit operator. The

classical limit of this operator is again in complete agreement with the classical theory.

As we have previously observed the classical dynamical variable y is alv--i gauge


Now we consider the following quotient to establish the classical limit of the Y i,!.

operator" Uy
(x,py,;py,p, IEU E, p,;y, py,)
(X, P ,; Y,py, I|E Ix, ,; Y,py,)

where 6(h) is a regularization parameter. We require the projection operator to possess

the properties of all projection operators namely Hermitian Et = E and idempotent

E2 = E. The relation (8-37), implies that the operator projects onto the spectral

interval [0, 62(h)]. The projection operator formalism allows us to deal with all constraints

simultaneously and to place all types of constraints on equal footing.

8.2.5 Time-Dependent Quantum Constraints

The projection operator of the time-dependent quantum constraints, follows the same

form -ii--.- -1. 1 by Klauder in [13], namely,

,, 2+)sin[62(h) + ]1A
E lim lim jr dAe- a (8 38)
L--oo &-O++ -L

To obtain further insight it will be convenient to use the canonical coherent states of the

unconstrained Hilbert space.

qpt, t) = exp[ia (p, q, pt, t)]e-i PjiQ3 e-itPt T I rT) (8-39)

where ITr) is a normalized fiducial vector in R-. These coherent states admit a resolution of

unity given as
/ _,dpjdy dtdpt
1= P qiPt,t) ,qPt,t dp dtd (8-40)
27 27
where the domain of integration is the entire extended phase space. The overlap of these

vectors are given by the following:

p- t'j I pt, t) !\, q (p', t' pt, t)

exp{- [| -p2 1 -q2 p'2 -2 pq }

x exp( 1 Pt2 t2 + tl ttl} (841)
4h t 2ht

Expression (8-41) defines a positive definite functional which can be chosen as the

reproducing kernel and used to define a reproducing kernel Hilbert space R.

ICj(Q;; N) (NN exp{f /2v o 2cos }dcos

sinh v/22 (4 3)

/*2 -2 2 22
z.*252 (zu,'* 2)2
N"N[1 + + + ...] (4-32)
3! 5!

From (4-32) we can deduce that the physical Hilbert space for every even particle sector is

one-dimensional. The Hilbert space found using this method is unitarily equivalent to the

one determined by the Master Constraint Program.

4.3.3 Open, First Class constraint

The next constraint model's inspiration is that of gravity (General Relativity). It is

well known that the constraint algebra of gravity,

{Ha(x),Hb(Y)} = 6x,y)Hb(x) b(x,y)Ha(x), (4-34)

{Ha(x),H(y)} = 6,a(x,y)H(x), (4-35)

{H(x),H(y)} = 6,(x,y)gab(x)Hb(x), (4-36)

is classically first-class; however, upon quantization the constraints transmute to partially

second class. [8] The analysis of the model we are about to examine served as the primary

motivation behind [8]. This model is also a type of constraint that that we considered

in C'!I pter 3, in terms of the Fadeev-Popov procedure. The action for our choice of the

modified system is very similar in form, i.e.,

I2 / (pa- Abl) dt (4 37)


Jeffrey Scott Little was born Janurary 10, 1980, in the small eastern Kentucky town

of Pikeville. He is the eldest of three children and the only son of Jeff and Linda Little.

His strong interest in science was evident at a very early age. When Scott was 14, he

discovered his passion for quantum physics and read nearly everything he could find on the

subject. Scott graduated from Shelby Valley High School in 1998, as class valedictorian.

After high school, Scott set off to matriculate at Western Kentucky University in Bowling

Green, Kentucky. While at Western he became extremely interested in the study of formal

mathematics. During his senior year he entertained the idea of attending graduate school

in mathematics; however, he realized that it would be possible to pursue both passions

through a physics career. Scott graduated from Western Kentucky with a double bachelors

degree in physics and mathematics in the Spring of 2002.

After finishing his undergraduate career he accepted an Alumni Fellowship from the

University of Florida to continue his studies of physics. Though moving from a relatively

small department at Western to the much larger Physics Department at Florida was

initially daunting, Scott overcame his fears and succeeded in his course work. In the

Spring of 2004, Scott became a student of John Klauder. Dr. Klauder allowed Scott to not

only study physics but also allowed him to stay connected to the formal mathematics that

he had grown fond of during his stay at Western. Under Dr. Klauder's tutelage, Scott was

able to research and publish three papers on quantum constraints.

In June of 2006, at the age of 26, Scott married the love of his life Megan (Carty)

Little. Scott obtained his Ph.D. in Physics in the Fall of 2007. Scott and Megan currently

reside in Louisville, KY, where Scott is continuing to research a wide variety of theoretical

problems and is an instructor at the University of Louisville.

will now turn our attention to discussion of the quantum analysis of the afformentioned


8.2 Quantum Considerations

8.2.1 Gitman and Tyutin Prescription for Time-Dependent Second-Class

Before proceeding with the discussion of the projection operator formalism,

we will briefly describe the method Gitman and Tyutin prescribed for dealing with

time-dependent second-class constraints. [12] For simplicity, we will limit this discussion

to include bosonic variables, however, one could extend any of the following arguments

to include fermionic degrees of freedom as well. Also, for convenience we will use the

notation used by the original authors, namely, = (q,p) which can explicitly depend on

time, as well as, {., *}D(6) represents the Dirac bracket with respect to a set of second-class

constraints Qa(q], t). The Dirac brackets are defined in the following manner

{f, 9}D(6) {f, 9} {f, }Ca{ bg9} (8-24)

where {., .} is the Poisson brackets, Qa is a constraint, and Cab is an invertible matrix [12].

Whenever encountered the Dirac bracket is taken assumed to be defined for the extended

space, (I; t, pt), as described in the preceding section.

Consider a classical Hamiltonian system with a set of second-class constraints Qa(I, t)

and with a Hamiltonian H(p, q, t). The Dirac brackets [7] are used to avoid having to solve

the constraints. Therefore, the evolution of the canonical variables is given by

d ri' = {, H + pL)D(p) Qa(T, t = 0.(8-25)

The quantization of the classical system, follows in the Schrodinger picture, in

which the canonical variables Tr are assigned to operators rTs that satisfy the equi-time

commutation relations;

[s' T1, D(6), La(l ) 0. (826)

9.2 Second Class Constraint

The second and final model we will consider in this chapter is inspired by the work

of J. Antonio Garcia, J. David Vergara and Luis F. Urrutia [49]. In this work, the

authors extend the BRST-BFV method [14], to deal with non-stationary systems (i.e.

time-dependent systems). For this dissertation, we chose not to discuss the BRST-BFV

method, however for a description of the method see [14] and [49].

The model used to illustrate the author's technique was a two-dimensional rotor

with a time-dependent radius. However, for this discussion we abate the model in [49], by

reducing the number of degrees of freedom from 3 to 2, as well as, setting the Hamiltonian

equal to zero to emphasize the constraints. Consider the following time-dependent classical


1 = ct, (9-13)

2 = p -c, (9-14)

03 Pt, (9-15)

where c is a positive constant and q, p, pt, are the canonical position, its corresponding

conjugate momentum and conjugate momentum corresponding to the temporal coordinate.

Based on the the Poisson bracket of the constraints, this constraint system is a second-class


The canonical quantization of this model is straight forward. We simply follow the

same procedure as stated in the preceding chapter, which implies that we promote all of

the canonical coordinates (p, q; pt, t) to irreducible self-adjoint operators(P, Q; Pt, T). We

promote the constraints to self-adjoint operators as indicated by the following:

p1 I 1 = Q cT, (916)

2 2 = P c, (9-17)

'3 3 = Pt (9-18)

9.3 Conclusions . . . . . . . . 111

10 CONCLUSIONS AND OUTLOOK. ................ ... .. .. 113

10.1 Summary ................. ............... 113
10.2 Ending on a Personal Note .... ........... ...... .. 115



REFERENCES .................. ................ .. .. 118

BIOGRAPHICAL SKETCH .................. ............. 122

However, if the constraint's spectrum contains a zero in the continuum then the

projection operator vanishes as 6 0 [13], which is unacceptable. To overcome this

obstacle, this limit must be evaluated as a rescaled form limit. To accomplish this, we

will need to introduce suitable bras and kets in the unconstrained Hilbert space. For this

discussion it will be convenient to choose canonical coherent states (1p, q)) to fulfill this

choice. We regard the following expression as the rescaled form

S(6)(p', q' IEp, q), (5-8)

where S(6) is the appropriate coefficient needed to extract the leading contribution of

(p',q'| Ep, q), for 0 < 6 < 1. For example, if (p',q' E|l' q) oc 6 to leading order, then
S(6) oc 6-1, for small 6. The limit 6 -- 0 can now be taken in a suitable' fashion. The

expression (5-8) is a function of positive semi-definite type and this means that it meets

the following criteria

limrn j azS()(p,q E,':,qi) >0, (5-9)
6O j' Cla aIS(6) (pj, qj E0(5

for all finite N, arbitrary complex numbers {ac} and coherent state labels {pi, qi}. A

consequence of the previous statement is that (5-8) can lead to a reduced reproducing

kernel for the physical Hilbert space

IC(p', q';p, q) limS(6)(p',q' Elp, q). (5-10)

The reproducing kernel completely defines the physical Hilbert space [13]. The

reproducing kernel makes it possible to express a dense set of vectors in the functional

1 non-trivial

extended, in a natural manner, the symplectic two-form

S= u +dt A dpt (8-16)

where w is the symplectic form of the original parameterized space ([17]), defined in
C!i lpter 2. The Poisson bracket {., .} which is defined by the symplectic form, should be

understood unless otherwise specified to be that of the extended space. The equivalent

action can be written in the following manner:

I' = 2dr[pj* + p* (pt + H(p, q, t)) A'] (8-17)

where A" = A(r)A which is merely a redefinition of the Lagrange multiplier. Notice once

again the canonical Hamiltonian vanishes. As in the previous section the dynamics of the

system arises from implementing the constraints.

A{qJ,pt + H} + {qJ, 0} (8-18)
A{pd, p + H} + A {py, ,} (8-19)
e (p,q,t) = 0 for all a { 1,..,A (8-20)

=d A{Qpt + HJ+ Ab0, b} M 0 (8-21)

8.1.2 Commentary and Discussion

Inspired by reparameterization invariant theories [48] along with other models

proposed by other authors [49], [50], we have arrived at a starting position to deal with

time-dependent constraints. This was done by changing the dimension of the entire

unconstrained phase-space from R2N to R2N+2, which was accomplished by promoting t

to a dynamical variable, and introducing its conjugate momentum pt. The consequence of

example, let us consider the following two constraints:

1 = q(1 q)2, (5-3)

S q- 3)(q 4)3. (54)

The first constraint is regular at q = 0 and irregular at q = 1 of order 2. The second

coinstraint is irregular at both q = 3, of order 2, and at q = 4, of order 3. Both of

these constraints are representative of the class of highly irregular constraints. Since the

dynamics as well as observability [9] of a given system are potentially not the same for

regular and irregular constraints, careful consideration must be observed when quantizing

such systems. The projection operator formalism [13] seems to provide an appropriate

framework to deal with systems with irregular constraints [9].

The usual form of the projection operator is given by

E(Za+d < 62(h)), (5-5)

where EYa( is the sum of the squares of the constraint operators and 6(h) is a small

regularization factor. The projection operator is then used to extract a subspace of the

unconstrained Hilbert space, R. If ZEK has a discrete isolated 0 then 6 can be chosen

to be an extremely small number. However, if ZEK has a 0 in the continuum, we can

not choose an appropriate 6 to select the proper subspace. We will discuss this distinct

possibility shortly. In the limit as 6 -- 0 if appropriate, this subspace becomes the Physical

Hilbert space,

limEl ) = I)hys, (5-6)

lim E- Ph-pys. (5-7)

coherent states for the unconstrained Hilbert space can be written as the direct product of

2 coherent states on different circles.

Coherent states on a circle can be generated by coherent states of a line with the use

of the Weil-Berezin-Zak (WBZ) transformation [45]. We shall use X and Y to denote the

characteristic lengths of the x and y coordinates, respectively. The WBZ transform, T, is a

unitary map from L2(IR) to L2(S1 x S1*), where S1* is the dual to S1. The transformation

is given by the following

(TQ)(x, k) -E ,.ifXk (x nX) (6-5)

where ip E L2(R), x E S1, and k E S1* or stated otherwise k E [0, 2). We project a

corresponding fiber of L2(S1 x S1*) onto L2(S1) by fixing a value of k. Using the standard

canonical coherent states in L2(R), it has been shown the coherent states on a circle have

the following form (h = 1)

1 1 1 X
T (I') = 7/4exp( p(x + ip)) exp(- (x + ip- x')2)0(i (x + i- x'- ik);pl),
kx/) F1/4 2 2 2
S(x' x,p,k) (6-6)

where pi = exp(-X2) and

((z) = ,,p",22in, (6-7)

|p| < 1, is the Jacobi theta function. These states are not normalized [45]. For each
value of k these states satisfy the minimal axioms of generalized coherent states; i.e., a

continuous labeling of the states where the label set has a topology isomorphic to IR2 and

a resolution of unity [40].

We can express the coherent states on T2 as the following,

|x, px, k1; y, py, k,)= \x, px, k) 0 y, py, k,), (6-8)


"It is very important that we do not all follow the same fashion... Its necessary to increase

the amount of variety .... the only n,. to do it is to implore ;,. c; few guys to take a risk

." -Richard Feynman

The Standard Model is the archetype of the kind of success physics has had in

describing the physical universe. The theory provides an explanation of the interactions

of matter with the electroweak and strong forces on a fundamental level. The way these

forces enter into the theory is based on Yang-Mills theory, a generalization of Maxwell's

theory of electromagnetism. As is the case with electromagnetism, the equations of motion

for a Yang-Mills field contain constraints that reduce the number of degrees of freedom [1].

This is a key characteristic of constrained systems. The process of converting a classical

theory to a quantum theory is made more difficult by the presence of these constraints.

Commonly used techniques to deal with these systems have been inadequate in providing

a description of the low momentum behavior of the strong force, which is associated with

the mass-gap conjecture [2].

General relativity, like the Standard Model, is another example of a constrained

system. The quantization of gravity has presented theoretical physics a cornucopia of

problems to solve for the past 50 years. To answer these deep theoretical questions,

physicists have employ, ,1 several and seemingly conflicting viewpoints. These perspectives,

range from Superstrings [3], the main goal is the unification of all forces in one quantum

mechanical description, Loop Quantum Gravity [4], in which the main objective of this is

a consistent background independent description of quantum gravity, to Causal Sets [5]

in which the approach preassumes that space-time is discretized, and the Affine Quantum

Gravity Program [6], in which the aim of this approach is to solve quantum constraint

problems with the Projection operator formalism. There are several deep underlying

theoretical issues surrounding the quantization of gravity, one of which is that gravity


Obtaining a Ph.D. in any field is never a complete individual effort, I owe many

thanks to many people that helped me pursue this life-long goal. First of all, I would like

to thank my advisor John Klauder, for giving me the chance to study quantization of

constraints systems. I am indebted to him for all the patience, time, and encouragement

that he has afforded me over the years that I spent under his tutelage at the University

of Florida. His passion for the course of study has helped me obtain a broader view

of physics, as well as a more developed palate for various topics in physics. In fact I

am grateful to the entire physics department for allowing me the opportunity to study

theoretical physics. I gratefully acknowledge the Alumni Fellowship Association, which

allowed me to attend the University without an overwhelming teaching responsibility.

I would also like to thank my grandparents, Ruby, Granville, Hazel, and Veral, who

instilled in me from an extremely early age that I could accomplish anything if I set my

mind to it. Thanks go to my dad, Jeff, who gave me a sense of scientific curiosity and

to my mom, Linda, who was my wonderful math instructor from fractions to calculus,

not to mention all of their love and support, and to my sisters, Lisa and Sierra, whose

constant encouragement aided me through my early college and graduate career. I am

assuredly indebted to my dear Aunt Brenda, who carefully edited several chapters of this

manuscript, even though she is not a physics person. I would also like to thank my wife's

family for all of their support these past 2 years.

Thanks go to my friends at the University of Florida Larry, Ethan, Wayne, Ian,

Lester, Jen, and Garret, whose compassion and conversations about a wide variety of

topics are unmeasurable.

Saving the best for last, I thank the love of my life, my darling wife, Megan. Without

whom I would have never completed this dissertation. I thank her for all of the love,

support, and encouragement that she has given me; she is the source of my inspiration to

achieve, more than I ever dreamed could be achieved.

We rescale this function differently so that

k (p", q"; p', q')


and then take the limit 6 -- 0 leading to

l (p", q"; p', q')

lim kCi(p", q"; p', q')
e-'q"2+q']/2, (5-51)

a continuous function of positive type that characterizes the one-dimensional Hilbert space


Our procedure of scaling the separate parts of the original reproducing kernel by

qualitatively different factors (i.e., 61 and 62) has not appeared previously in the projection

operator formalism. This difference in scaling is motivated by the goal of having each and

every element of the reduced classical phase space represented on an equal basis in the

quantum theory. It is only by this procedure that we can hope that the classical limit of

expressions associated with the physical Hilbert space can faithfully recover the physics in

the classical constrained phase space. Scaling of Ci and 2C by finitely different factors has

been addressed in [9].

Let us continue to examine the remaining CK, 3 < n < 4. For KC4 we have

4 (p", q"; p', q')

(p", q"IE(-6 < 8(Q- 3)3 < 6)p',q'
1 f3+ ) p
e -(x-q")2 /2-i(p"-p')x -(x-q')2/2 dx
V/T J3-,63

where 63 [6/8]1 3. The now familiar procedure leads to

", q"; ', q) = e-[(3-q")2+(3-q') ]/2-2i(p"-p')



throughout this project, the expression derived in [47] will not be the starting point for

our investigation.

We will, however, advocate the use of the reparameterization invariant description to

discuss systems with time-dependent constraints. We should mention that this starting

point is not a new approach to deal with time-dependent constraints. We will alter past

efforts on this topic by exploring the "non-local" point of view. The phrase "non-local"

point of view was coined by Gitman [48] when describing a physical system in which

one assumes a reparameterization invariant form of a theory. However, it is well known,

that if an action is a reparameterization invariant then the Hamiltonian vanishes on

the constraint surface1 Physics described in a reparameterization invariant form

is not dependent on the frame of reference [48]. We should mention that using the

reparameterization invariant approach is not new, however, the implementation of this

symmetry with the projection operator would seem to be new.

In the next two chapters we will discuss some of the facets of the problem with

time-dependent constraints. In ('!i lpter 8 we will motivate and develop the techniques in

which one can study constraints with an explicit time-dependent feature. We will also give

a brief introduction to an alternative to the projection operator, which is the approach

used by Gitman, [12] and compare and contrast the two approaches. The primary goal

of ('! Ilpter 9 is to implement the formalism developed in C'!i lpter 8, in a few examples of

time-dependent constraints.

1 See Appendix for this result.

In the Schr6dinger picture it is stated vectors evolve in time, where the time evolution is

generated by a unitary operator. Operators in this picture are stationary, which implies

the operators are time-independent. However, in this system the canonical operators r]s

carry over an explicit time-dependence from their classical analogues, therefore these

operators evolve in time. This is a departure from the traditional Schr6dinger picture. In a

later work, the authors recognize this distinction by calling this picture "rule" exist..

At this juncture we realize this current picture is unable to illustrate the full time

evolution of the system. In order to fully obtain the time evolution we will move to a

unitarily equivalent picture, the Heisenberg picture. In the Heisenberg picture of quantum

mechanics, the state vectors remains fixed while the operators evolve in time. [38] In the

Heisenberg picture the operators rIH are related to the operators rTs by IH U= -1'IsU,

where U is the time evolution operator. The operator U is related to the Hamiltonian Hs

by the differential equation,
0U -i
SHs (8-27)
at h
We can evaluate the total time derivative of TlH by the following

dUlH d(U-1UsU)
dt dt
U-1-( [ Hs, Hs] + {, pt}D())U I 4 ~Us (8-29)

Equation (8-29) establishes the connection between the quantum equations of motion and

the classical equations of motion namely,

= d {, H + pt)D() 'H (8-30)

In the most general setting, the above described evolution is not considered "unitary",

because in general no "Il Ii!illin ,,, exists whose commutator would result in the total

derivative. The principal agent for this non-unitary character is the second term in the

left-hand side of equation (8-29), which is time variation of ls. Therefore, the dynamics

are evolving, as well as the constraint surface.

and DR(A) is the weak measure defined in C'i ipter 4. This result agrees with the result

obtained in [47]. This of course is assuming that the constraints are continuous in t. This

statement is also applicable if ,a are second-class constraints.

8.2.6 Observations and Comparisons

Despite the fact that the Projection Operator Formalism and the approach used by

Gitman [12] start on very similar grounds, the approaches end on very different grounds.

The Gitman approach advocates the use of Dirac Brackets, which is a method used to

avoid solving for second-class systems, while the Projection Operator Formalism treats all

constraints on equal theoretical footing. In the proceeding chapter we will examine two

different constraint models with the aid of the projection operator formalism.

where 6(h)2 is a regularization5 parameter. As it is often emphasized by Klauder, 6(h)2

is only a small parameter, not a Dirac 6-functional [40]. Equation (4-6) projects onto a

subspace of the Hilbert space with a spectral measure of o a from 0 to 6(h)2. The true

physical Hilbert space (4-1) is determined when the limit as 6(h)2 -- 0 will be taken6

in an appropriate7 manner. In the following subsection we will offer functional form of

4-6. We will now turn our attention to further motivating the process of squaring the


4.1.1 Squaring the Constraints

Much like the Master Constraint Program (\ICP), the Projection Operator

Formalism (POF) also relies on the summing of the square of constraints to replace a

set of constraints {a}0 ,, with a single term. Unlike the MCP, the POF offers further

justification for only dealing with the sum of the squares instead of appealing to simplicity

arguments. By simplicity arguments we mean, why stop at a second-order polynomial

expression of the constraints, why not consider fourth-order or higher? The authors of

[34] only mention that second order was chosen because it is the simplest expression.

Instead, we attempt to offer some mathematical arguments that indicate that the sum of

the squares of the quantum mechanical constraints is sufficient.

4.1.2 Classical Consideration

Before moving to the quantum mechanical description of the constraint story, it is

important for us to be able to motivate the tale classically first. Let us consider a set of A

5 Hence a regularized physical Hilbert space!!

6 Or taken depending on the type of constraint. [13]

7 More on this later.

We can conclude based on our assumption that A / 0, a.e., that the statement

following must be true,

= H


O. D

Therefore, in all reparameterization invariant theories the Hamiltonian vanishes.



Scheme. This comparison has left us with the conclusion that these methods are different

and incompatible with each other. To the author's knowledge, this is the first time

the Projection Operator Formalism has been used when the configuration space has a

non-trival topology.

The remaining chapters were devoted to the topic of time-dependent quantum

constraints. We developed the formalism in which the topic can be approached in the

context of the Projection Operator. This was accomplished by extending the classical

phase space of the time-dependent system, thereby elevating the time parameter to a

dynamical variable. In the same chapter, we compared the Projection Operator Formalism

to the approach that was first discussed in[12]. While these methods start from the

same point (i.e. an extended phase space) the conclusions reached are very different. In

the following chapter we were successful in analyzing two examples of time-dependent


The story of classical and quantum constraints that we have presented within this

dissertation is by no means a complete account. In fact it is impossible to give a complete

account of any research. By researching we merely point the direction to new research,

leading to new questions to ask and to attempt to answer. The topics in physics are

alv -i.- bigger than the individual physicist. However, this is the beauty of the subject,

that things we leave unresolved can be picked up in the future generations.

There are several unresolved issues left from this dissertation that can be addressed

by the author or future researchers. These include but are not limited to, "How do the

methods of the Projection Operator generalize to a full quantum field theory?", "What

lessons learned from the simple models that we analyzed in this dissertaion can be applied

in more realistic theories such as Quantum Gravity?", "Can we use the formalism obtained

in C'!i lpter 8 to examine more realistic theories?"


[1] S. Weinberg, The Quantum Theory of Fields Vol. 2, Cambridge University Press,
Cambridge, 1998.

[2] A. Jaffe and E. Witten, "Quantum Yang-Mills Theory"
-mbilbao/pdffiles/yangmills.pdf, 04/19/2007.

[3] P. Deligne, et al., Quantum Fields and Strings: A Course for Mathematicians Vol.
1-2, American Mathematical Society, Providence, 1999.

[4] A. Ashtekar, J. Lewandowski, "Background Independent Quantum Gravity: A
Status Report, Class. Quant. Gray., 21 (2004) R53 .

[5] J. Henson, "The Causal Set Approach to Quantum Gravity,"
abs/gr-qc/0601121 (2006).

[6] J. Klauder, "Affine Quantum Gravity," Int. J. Mod. Phys. D12 (2003) 1769.

[7] P.A.M. Dirac, Lectures on Quantum Mechanics, Belfer Graduate School of Science,
Yeshiva University, New York, 1964.

[8] J. S. Little, J. Klauder, Elementary Model of Constraint Quantization with an
Anomaly ," Phys.Rev. D, D71 (2005) 085014.

[9] J. Klauder, J. S. Little, "Highly Irregular Quantum Constraints ," Class. Quant.
Gray, 23 (2006) 3641.

[10] J. S. Little, "The Projection Operator Method and the Ashtekar-Horowitz-Boulware
Model", (2006).

[11] A. Ashtekar, G. T. Horowitz, "On the Canonical Approach to Quantum Gravity", in
Phys. Rev. D, 26 (1982), 3342.

[12] D.M. Gitman, I.V. Tyutin, Quantization of Fields with Constraints, Springer-V, 1 1
Berlin 1990.

[13] J. Klauder, "Quantization of Constrained Systems," Lect. Notes Phys., 572, (2001)

[14] M. Henneaux, C. Teitelboim, Quantization of Gauge S, -/ i'- Princeton University
Press, Princeton, NJ, 1992.

[15] J. Jose, E. Saletan, Classical D.i iii. A Contemprary Approach, Cambridge
University Press 1998.

[16] J. Govaerts, Hamiltonian Quantisation and Constrained D.i.,..nn- Leuven
University Press, Belgium, 1991.

[17] V. I. Arnold, Mathematical Methods of Classical Mechanics, Singer V i1 .- 3rd ed.
New York, New York (1989).

[18] L. Castellani, "Symmetries in Constrained Hamiltonian systems," Ann. Phys, 143,
(1982), 357.

[19] A. A Deriglazov, K. E. Evdokimov. "Local Symmetries in the Hamiltonian
Framework. 1. Hamiltonian Form of the Symmetries and the Noether identities,"
Int. J. Mod. Phys. A15 (2000) 4045.

[20] P.A.M Dirac, The Principles of Quantum Mechanics, 4th ed., Oxford Science
Publications, 1998.

[21] J. Klauder, \. l i cal Quantization" http://arxiv. org/abs/quant-ph/9804009

[22] J. Simon, C'l,'.ri, Without Time Relationalism and Field Quantization, Dissertation,
Universitat Regensburg, Naturwissenschaftliche Fakultat II Physik, 2004.

[23] N.P. Landsman, Mathematical Topics Between Classical and Quantum Mechanics
Springer-Verlag New York, 1998.

[50] D. Giulini, "That Strange Procedure Called Quantisation",
abs/quantph/0304202 (2003).

[25] A. Ashtekar, R. S. Tate, "An Algebraic Extension of Dirac Quantization:
Examples," (1994).

[26] S.V. Shabanov, "Geometry of the Physical Phase Space in Quantum Gauge
Models," Ph;,: Reports 326 1 (2000), hep-th/0002043.

[27] L.D. Faddeev, "Feynman Integral for Singular Lagrangians," in Theor. Math. Phys.
1, (1969), 1.

[28] P. S(li in.,vic, "Path Integral Quantization of Field Theories with Second-Class
Constraints," Ann. Phys. 100, (1976), 227.

[29] Adriano Di Giacomo, "Confinement of Color: Open Problems and Perspectives,"
http: //arxiv. org/abs/hep-lat/0602011, (2006).

[30] D. Giulini, D. Marolf, "On the Generality of Refined Algebraic Quantization," Class.
Quant. Gray., 16, (1999), 247.

[31] D. Marolf, "Refined Algebraic Quantization: Systems with a Single Constraint,"
http: //arxiv. org/abs/gr-qc/9508015, (1995).

[32] J. Louko, A. Molgado, "Superselection sectors in the Ashtekar-Horowitz-Boulware
Model", Class. Quantum Grav. 22, (2005) 4007.

3.2.1 Basic Outline of Procedure

The prescription that RAQ follows begins with the basic treatment of canonical

quantization that was described in C'! lpter 2, although, in general, this prescription

generally relaxes the Cartesian coordinate requirement. First, one must represent the

constraints Ci as self-adjoint operators (or their exponential action, as Unitary operators)

that act on an auxiliary Hilbert space,Hanx, which in turn is the prerequisite linear space.

Since, in general, the constraints, C, have continuous spectrum, it follows that the solutions

of the constraints could be generalized vectors. Thus, we will consider a dense, subspace

of Haux (4 C 'Ha,) which can be equipped with a topology finer than that of the regular

auxiliary Hilbert space. The distributional solutions of the constraints are contained in the

algebraic dual V*, (space of all linear maps -- C). The topology of V* is that of point

wise convergence, which is to -v a sequence f, E 4* converges to f E 4* if and only if

fQn() f(0) for all Q GE 4. This concept, as we mentioned in the introduction of the

section, is based on the theory of Ri.-.-. -l Hilbert spaces. The subspace 4 is chosen based

upon the condition that it is left invariant by the constraints Ci or the exponentiated

action, or that it can be determined based on physical choices. [31] Another technical

requirement is that for every A E Aobs,(i.e. the algebra6 of observables), which commutes

with Ci, A as well as its adjoint At, are defined on 4 and map 4 to itself. We will attempt

to describe this particular requirement briefly.

The final stage of the RAQ procedure entails constructing an anti-linear map called a

rigging map,

I : 4 ---*, (3-14)

6 This is a *-algebra on the Haux.

C'!i ipter 3, as well as, the motivation behind the development of the projection operator

formalism [13] in C'!i ipter 4.

We can quickly quantize this system by promoting the p's and q's as irreducible self

adjoint operators. As noted before [1, M1] = 0. However, as mentioned before, 0 is not in

the spectrum of M since the least eigenvalue of M is Thiemann [33] offered a means

to rectify this by -.- -ii,.-; a modification of the quantum master constraint by the


M = M AI (3-25)

where A = inf{spectrum(M) and I is the identity operator on the auxiliary or a

kinematical Hilbert space. According to Thiemann, [[33]], (3-25) will still have the same

classical limit of the master constraint because A oc h. In general, if a system contains a

constraint that is classically an open, first-class constraint, like gravity or the system that

we will discuss in C'!i lpter 4, where A would be proportional to h2.

Assuming the operator M is a densely defined self-adjoint operator, we can now

proceed with the quantization by first addressing the auxiliary Hilbert space. Using the

fact that M is a self-adjoint operator with a positive semi-definite spectrum, the auxiliary

Hilbert space can be written as the following direct integral [34];

Hagx = dJ(x)Ha,(x) (3-26)

where d4(x) is the spectral measure [33] of the master constraint operator (3-25). Each

addend contribution to the sum, 'H'f(x), in (3-26) is a separable Hilbert space with the

inner product induced by the auxiliary Hilbert space, Haux.

Using this particular construction we are now able to address the task of solving

the quantum master constraint equation M = 0. By the mathematical description of

the auxiliary Hilbert space (3-26), it follows that the action of M on -H5,x(x) is simply

multiplication of x. We can solve the quantum master constraint equation by identifying

the physical Hilbert space by the following

9-phys (0). (3-27)

known, this choice of the invariant subspace can lead to non-physical results, such as,

super-selection sectors7 in a variety of constraint models where these structures are not

motivated.[32] For this primary reason the RAQ procedure has difficultys when dealing

with constraints that have zeroes in the continuum. The RAQ method also has difficulty

when the constraint algebra produces a quantum anomaly, as well as cases in which an

infinite number of constraints are present. This leads us to the third and final method that

we will discuss in this chapter the Master Constraint Program.

3.3 Master Constraint Program

The third and final method that is discussed in this chapter is known as the Master

Constraint Program (\!CP). Like the RAQ, MCP follows the mantra of the Dirac

procedure in that one must quantize first and reduce second. We notice the same basic

philosophy in the next chapter when we discuss the Projection operator formalism. The

Master Constraint Program was developed by Thiemann, et al. [33], in an attempt to

overcome situations in which the RAQ procedure fails. These I !liures" include, but are

not limited to, cases in which an infinite number of constraints are present, as well as

when the structure functions are not constants, but rather are functions over the phase

space. This program also attempts to eliminate other ambiguity from the RAQ procedure,

namely the requirement of additional input into the physical theory. As mentioned before,

this additional input is a dense and invariant subspace which is equipped with a finer

topology than that of the Hilbert space in which it is embedded. [34] During this section,

since we only intend to give a heuristic account of the Master Constraint Program, we will

7 See chapter 6 for further details.

8 When we ;- difficulty, we mean conventional approaches such as group averaging
techniques fail. Extra mathematical constructs must be implemented. [31]

The first term in the determinant is zero by the 6 functional of the O's. The second term is

an N x N matrix multiplied by a scalar f, and therefore becomes.

f e() J adt b Xb ntfdet xb, ODpDq. (313)
7j Htfn

We observe that all the factors of f completely cancel. As one can see, the Faddeev

method is insensitive to the definition of f, as long as it be non-zero. Hence, this method

considers the Q,'s and Q,'s as identical constraints. We will examine a similar model in

C'!I Ipter 4 [8], which demonstrates that in order to understand the entire theory, one must

also consider the quantum mechanical corrections.

Another difficulty in this method derives from the selection of gauge choice X,. As we

noted in the previous subsection, the choice of gauge is only guaranteed locally. In more

complicated gauge theories, such as Yang-Mills it is well known [29] that there does not

exist a gauge choice that slices the gauge orbits once and only once, a fact which limits the

effectiveness to probe the non-perturbative regime of these gauge theories. [26]

3.2 Refined Algebraic Quantization

The Refined Algebraic Quantization Program (RAQ) is in stark contrast to

the Fadeev-Popov method mentioned in the previous section. RAQ attempts to

quantize the entire classical theory first including the constraints, then attempts to

impose the quantum constraints in order to determine a Physical Hilbert space. In

this respect, the RAQ attempts to extend and resolve some of the ambiguities of the

Dirac Procedure namely, "How is the the inner product imposed on the physical Hilbert

space?" and "Which linear space do the linear constraints act on?". [30] Refined Algebraic

Quantization comes in two main varieties, Group Averaging and a more rigorous version

that is based on the theory of 1i.-.-. Hilbert spaces. In this chapter we will focus on the

former rather than the latter because most experts will agree that there does not exist a

group averaging technique for all constraints in this formalism. See [30] and [31] for more

complete discussions on the failures of group averaging.

Unlike the previous case considered in the subsection, 6 can be taken to zero to yield a
non-trivial result. We will still utilize the reproducing kernel to discover the functional

nature of the physical Hilbert space. Before proceeding it will be convenient to introduce
the equivalent form of the canonical coherent states;

I = e(Q-iP)- (Q+iP) 0) (4 30)

where z'- (q+ ip)/V2h and denotes the standard Euclidean dot product [41].
A coherent state path integral can also be used to calculate the matrix elements of
the projector as shown in previous works [13]. Let us begin with a preliminary equation,

(,",i Te-(i1) f A' Jdtl 5,,')

M Jexp{(i/h) ((pa&' q a))/2 A"ja)dt}DpDq

N"N' exp{//* e-(i/h)o ', )

where M, N", and N' are normalization factors, (q+ ip)/V2h, j is a 3 x 3 matrix
representation of the rotation algebra, T denotes time ordering, and J is a suitable
functional of {A0Q)}.
Following [13], we could integrate over X with respect to a suitable measure R(A) to
create the desired projection operator. However, it is equivalent and simpler to proceed as

ICe(r;) (e-a E(2 _< h2) e)

where d) is the normalized Haar measure of S(3). Consequentlyd,

where dp(0) is the normalized Haar measure of SO(3). Consequently,

following classical action
I J/dt(p AaO), (3 7)

where j E {1,...,N} and a E {1,...,N}. This system is purely constraint4 The
definition of the constraints are

a = f(p, q)a(p, q), (3-8)

where f is a non-vanishing function over the phase space and {(a} defines a closed-first
class constraint, i.e.

{Oa, b} ,.' (3-9)

where Cab is a constant. It is clear while (3-8) defines the same constraint sub-manifold as
the case in which ,a are constraint, the constraints ,a are an open, first-class -' 111 .
We realize that the arbitrary function can be classically absorbed into the definition of the
Lagrange multipliers; however, we are ignoring this to emphasize the quantum mechanical
behavior. The Fadeev-Popov method for this model begins with the following expression:

J (i/f) '-J a]dlpDqDA, (3-10)

is replaced with the gauge-fixed expression

Se(i/) f t b det{b, f}DpDq, (3 11)

where xb(p, q) is some appropriate gauge choice. A simple identity leads to

(/hp) fadt r xb 6b} det({xb, f}c + { b, O}f)DpDq. (3-12)

4 Gravity is such a system [8]
5 This system is similar to the one discussed in [8], which we will return to in the next

3.4 Conclusions

In this chapter, we have examined three distinct constraint quantization programs.

All of these programs have their distinct advantages and disadvantages depending on the

particular constraint under consideration. In the following chapter we will examine the

projection operator method and examine the tools of that formalism and how it attempts

to overcome the difficulties of the preceding methods.

9Hphys = cl{span {R(r)Yoo(0, Q)}}.

The inner product of this physical Hilbert space is inherited from the L2(R+, r2dr). Thus

completes the quantization of this simple constraint model in the Master Constraint

Program 14 We will return to this model in the next chapter, when we will discuss it in

the context of the projection operator formalism.

3.3.4 Comments and Criticisms

Despite the many successes [35], [36], and [37] that the program has had in resolving

several of the ambiguities associated with the Refined Algebraic Quantization, it still

may not be the perfect choice to use for all constraints. If the constraint's spectrum

contains a zero in the continuum, then particular care, in the form of rather cumbersome

mathematical machinery, must be used. Not that this yields an incorrect result, however

it almost appears to be extraneous to the material. This is somewhat of a biased opinion

because as we will see in the next chapter the projection operator formalism's answer to

this seems more satisfactory. Again though, we emphasize the fact that the results have

been shown to be equivalent to the results found in [13]. Another possible criticism of this

program is not a criticism of the program, but instead, a criticism of its implementation.

That is to -iv that authors tend to di-plivi a heavy reliance on the classical analysis of

groups to solve constraints [35]. The main critique of this point comes from the fact that

most of the work done with groups such as sl(2, C) neglect the zero representation [41],

which should be the representation corresponding to physical Hilbert space. However, just

as the authors pointed out in [35], this particular constraint is not physically realizable,

and therefore not subject to experiment.

14 Actually, it is not the very end of this discussion. We must also include a discussion
of the quantum observables in the theory. We will simply point the reader to [35] for a
discourse on that topic.


invariant form of the action integral in the following manner.

I' = (tt* + pq* A() [pt + H(p, q)])d (8-4)

where (.)* denotes the derivative with respect to r. The price paid in promoting t to a

dynamical variable is that the Hamiltonian vanishes weakly in the extended phase space.

We have identified the primary unexpressable velocity t* as the Lagrange multiplier that

enforces the (first-class) constraint pt + H(p, q) = 0. Therefore, we have turned a theory

that was a dynamical system into one that is purely gauge. Effectively, we have recast

the original theory in such a manner that it can be related in any temporary reference

frame[48]. The equations of motion of (8-4) are as follows:

dq A H dp H(8
7- AO (8-5)
dr p dr 6q
dt dpt
dt A (), 0 (8-6)
dr dr
pt + H(p,q) = 0 (8-7)

As the equation of motion appear above q, momentum p, and the physical time t,

measured are gauge dependent quantities. However, pt is gauge independent and

therefore an observable quantity in this theory. By identifying the usual time as the

gauge dependent quantity

dt = A(r)dr (8-8)

we can quickly reduce the preceding equations of motion (8-4) to the familiar parameterized


H (8-9)
OH (810)

The dynamics of this system arises from imposing the constraints.

physical Hilbert space on a wide variety of constraint systems. This program has been

"tested" in systems that included, but are not limited to, simple quantum mechanical

constraints that form a non-compact algebra (like sl(2, R)) [36] to a fully interacting

quantum field theory [37]. Therefore, in order to proceed with the quantization of the

MCP, one must first promote the master constraint (3-17) to a self-adjoint operator that

acts on an auxiliary Hilbert space

M fMf. (3-23)

The main difference, at this point in the discussion, between the auxiliary Hilbert space of

MCP and that of RAQ, is that MCP requires the Hilbert space to be separable.

At first glance it may appear that one has eliminated the possible quantum anomaly

because a commutator of any operator with the same operator is zero, [Af, f] = 0. While

this is a true statement, the quantum anomaly has only been reformulated in another

manner, videlicet the spectrum of fM may not contain zero. An example that illustrates

this point more clearly is as follows: Consider a classical system with a classical phases

space, R, with two constraints,

C1i pi

C2 ql.

Using the classification system we described in C!i lpter 2, we can identify this system as

a second-class system. We should note that this system will not have a quantum anomaly

as defined by C!i lpter 2 and [8]. However, it will provide a distinct property that we are

attempting to illustrate, which is the quantum master constraint need not posses a 0 in

the spectrum. The corresponding master constraint of this system can be written as the


M = +q2 (3 24)

a 2(()- ;ep(- h)2
exp[iy (exp(- )) (6-26)
2r e( (p, h); exp(--2h))
As h -- 0 this expression becomes

exp ,2'], (6-27)

where y is subject to the condition R(y) = p2. While this expression is imaginary, we can

extract from it the classical reduced phase space coordinate y.

Now we direct our attention to the expectation value of the physical conjugate

momentum, P,

(x, p1; y,py EP. E I p; y,p)
(x,px;y,p. |ElI' pxy,py)

p i +x (6 28)
-- z dy' dx'(x, px; y,pyIx', y') } (x', y'x, py; Y,py). (6-28)
/C J J Ox'
We implement the constraints by integrating over the appropriate intervals as

described in Section 5. We can continue this calculation in a similar manner to that which

is performed in [9].

(x,p;y,py\EP.,E\, px;y,py) '(4p; exp(- 2/(h)))
= pX + (6-29)
(x,pX; yp, E P,; y, ;py,) 20( (h); exp(-T2/(h)))'
O'(z; p) 2i > pfe" eel2inz. (6-30)
As h approaches 0, the second term vanishes which can be seen in the definition of the

Jacobi theta function (6-7) [45], thus recovering this aspect of the classical theory from its

quantum analog. Using the same technique, we can also calculate the classical limit of the

expectation value of the Py operator. The projection operator formalism is well suited to

not only properly impose quantum constraints, but also allow one to return to the proper

classical theory in the limit h -- 0.

of positive type the function

(p", q" Ep', q')


serves as a reproducing kernel for a reproducing kernel Hilbert space, a functional

representation by continuous functions on the original phase space (IR x R), of the

regularized (by 6 > 0) physical Hilbert space. Our goal is to take a suitable limit 6 -- 0 so

as to yield a function that can serve as a reproducing kernel for the true physical Hilbert

space for the present problem.

Clearly the limit 6 -- 0 of the given expression vanishes and that is an unacceptable

result. Suppose we assume 0 < 6 < 1, e.g., 6 = 10-000. Then it is clear (even for a much

larger 6 as well!), for the example at hand, that

E(-6 < R(Q) < 6)

SE(-6 < 2Q < 6)

+ E(-6 < (8(2- Q)3/2 < ) + E(2 < Q < 3)

+ E(-6 <8(Q 3)3 < 6)

- Ei +E2 +E3 +E4,

where E,, 1 < n < 4, corresponds to the terms in the line above in order. By construction,

for very small 6, it follows that these projection operators obey

ELT, 6, E:


i.e., they project onto mutually orthogonal subspaces. In like manner the reproducing

kernel decomposes into

C(p", q"; p', q')

(p", q" IE(-6 R(Q) < 6)lp', q)

S ,/ (p"l, q"; p' q'),





C (p", q"1; p, q') (p", q" E Ipl, q').

stationary points. The result, which is well-known, is the Euler-Lagrange equations 2

6S = 0 (2-3)
d BL A L
t 0 (2 4)
dt 9dq dq0

While this formulation is useful in determining a wide variety of physical quantities

[15], it is not as sensitive to particular features of a given classical theory as we need.3

In order to illuminate these features we must first pass to an equivalent formulation of

classical mechanics, namely the Hamiltonian formulation. In making the transition from

Lagrangian to the Hamiltonian, we must first identify the conjugate momentum,

pa = (2-5)

This can be recognized as the fiber derivative from the tangent bundle of Q to the

cotangent bundle of Q [15] (otherwise known as the tangent bundle's dual)

a : TQ T* Q. (2-6)

The next step in the procedure is to identify the Hamiltonian, which follows from a

Legendre transformation of the Lagrangian

H =paq- L (p,q) (2-7)

2 This result can be generalized if L is a functional of (qa, q, g,... qa ) where k is a
finite number. The variation, which is determined by the functional derivative, of this
6L i(t L i" L itL
equation is given by 6, (t- t) + (t- t') + + 6(k) (t- t') L [14]. Integrating this
equation with respect to t' and setting the result to zero will yield the stationary points of
the corresponding action.

3 Obviously, if we express 2-4 in terms of a second order differential equation, the subtle
point we are about to make becomes clearer. [16]


_11illi, mi,,.l. .,l study and research are very suggestive of mountaineering. Wi,,i i'/ made

several efforts before he climbed the Matterhorn in the 1860's even then it cost the life of

four of his ',,I/;/ Now, however, ,i,.; tourist can be hauled up for a small cost, and perhaps

does not appreciate the diff;. ;,ll; of the c,:':,,rl1 ascent. So in mathematics, it i,,ini be

found hard to realize the great initial difficulty of i,,.;. :,.j a little step which now seems so

natural and obvious, and it i,,ii; not be surprising if such a step has been found and lost

again." Louis Joel Mordell (1888-1972; Three Lectures on Fermat's Last Theorem, p.4)

10.1 Summary

In C'!i pters 2, 3 and 4 we gave modest account of the background information needed

for the remaining chapters of the dissertation. In chapter 4, we successfully analyzed a

constraint that mimicked the aspect of the gravitational constraint, that it was classically

a first-class system; however, upon quantization it became a partially second-class system.

In the same chapter we also analyzed a closed, first-class quantum system, as well as a

first-class system with a zero in the continuum.

C'! lpter 5 introduces the classification of constraints called "highly irreg 11l I

constraints. During this chapter, we described a general procedure to solve the quantum

analog to the "highly irreg 11 o constraints utilizing the Projection Operator Formalism.

We also successfully analyzed a simple example of this type of constraint using only the

Projection Operator Formalism.

In C'! lpter 6, we used the mathematical tools established in C'i lpter 5 to give a

complete account of the quantization of the Ashtekar-Horowitz-Boulware model.[11]

This model was inspired by the Hamilitonian constraint of General Relativity to answer

whether or not there could be quantum mechanical tunneling into classically forbidden

regions of phase space. During the course of this chapter we compare the results obtained

by the Projection Operator Method with that of the Refined Algebraic Quantization

required only in the neighborhood of the zero set {q : R(q) = 0}; however, with an

eye toward the Ashtekar Horowitz Boulware model, discussed in Chapter 6, we choose

R(q) E C1 for all q.)

Our discussion will cover a wide class of R functions, and for convenience of

explanation we shall focus on one specific example; generalization to other examples is

immediate. The example we have in mind is given by

R(q) = q(q 2)3/20(2 q) + (q 3)0(q 3), (5-31)

1, x > 0
0(x) (5-32)
0, x < 0.

For this example, the zero set is given by

Z q = 0,q 2,q 3, and 2 < q< 3}; (5-33)

only for q = 0 is R'(q) / 0. (Although physically motivated models would typically

not include intervals in the zero set of R, we do so to illustrate the versatility of our


In summary, the phase space for the unconstrained classical system is parameterized

by (p, q) E R x R, and the phase space for the constrained system is parameterized by

the points (p, q) E R x Z. This latter space consists of several one-dimensional lines and

a two-dimensional strip. From the standpoint of this elementary example all elements of

R x Z are equally significant.

We now turn to the quantization of this elementary example following the precepts of

the projection operator formalism [13]. In this approach one quantizes first and reduces

second. The ultimate reduction leads to a physical Hilbert space appropriate to the

constrained system.

this change of space2 is that the new Hamiltonian vanishes, and we have one additional

constraint. However, constraints with explicit time dependence now pose the same

mathematical structure of time-independent constraints in the Dirac procedure [12] with

the aid of the extended symplectic form. We should also note that we have refrained

from introducing a temporal gauge fixing term such as a chronological fixing gauge in our

action. This is a point of divergence from the previous authors on the subject. As is well

known, a gauge-fixing term has the potential to introduce topological obstructions that

can cause difficulty in the analysis of the quantum system. This technique of introducing

a gauge is used in quantization schemes such as Faddeev-Popov [27] which advocates

reduction before quantization. Since one of the main philosophies of the projection

operator formalism is to quantize the entire dynamical space and reduce second (i.e.

eliminate the redundant variables), there is no need to introduce such a term in the

action. Dirac observables are phase-space functions that commute weakly with all of the

constraints. An extensive amount of literature has been devoted to the task of identifying

observables in systems such as General Relativity and other generally covariant systems

[51]. If o is a classical observable in a system with time-dependent constraint then the

following must be true:
S- A{o,p, + H} + A{o, ,} O 0 (8-23)

where {., .} are understood to be the Poisson brackets for the extended space. Therefore,

o is a constant of motion on the constraint surface in the extended phase-space, which

implies that an observable is independent of a choice of reference frame or gauge. Since we

will not make any further use of the concept of an observable in the discussion of systems

with time-dependent constraints, we will defer this discussion to a future project. We

2 An additional requirement of a global Cartesian coordinate system must be imposed
when we proceed to the quantization of the described system ala Dirac

where A is a Lagrange multiplier corresponding to the constraint C. The configuration

space of the AHB model is C = T2 S1 x S1. The constraint has the following form

C p R(y), (6-2)

where the function R(y) E C1(S1) is assumed to be positive somewhere. When the

constraint equation is satisfied the classical solutions are limited to the regions of the

configuration space where R(y) > 0. The constraint region in the 4 dimensional phase

space will involve a proper subset of configuration space. Note that the Hamiltonian

equals zero in this model to emphasize the role of the constraint.

The dynamics of this system are given by the following 5 equations of motion.

x = -2Ap1, y = 0,
Px = 0, P, = A- ,
p R(y) = 0.

From these equations of motion, we can make some statements on some observability

properties of this theory. The dynamical variable x is gauge dependent for all px except

for px = 0. The conjugate momentum of y also appears to be gauge dependent if the

constraints are regular around a given set of y that satisfies the constraint equation in the

phase space.

If yo satisfies the constraint equation and R(v) o = 0, then the constraint is an

example of irregular constraint about y yo, whereas if () y=o / 0, then the constraint

is regular about yo. If there are multiple solutions to the constraint equation p R(y) = 0

then we may have a condition where combinations of regular constraints and irregular

constraints, this is the characteristic of a "highly irreg- 11i constraint. For the most

general analysis, we can then assume that the constraint equation contains solutions that

where x, y e S1 and kx e Sl*,and ky E S'*. For simplification we will choose X, Y = 27.

We will make a further simplification by choosing a value for kx and ky. We justify such

a choice by noting that the spectrum of the momentum operator is shifted from the

expected value by k [45], effectively, we can set the new ground state at k. Therefore, we

can safely choose zero for both kx and ky. Thus we will make the following notational


X, P Y, py,)o= |X,P, )o 0 \,py)o, (6-9)

The construction of the reproducing kernel is based on properties of the constraint

operator as well as the coherent states (6-9). The constraint operator and the compactness

of x restrict the spectrum of its conjugate momentum P, and thereby of R(Y). Allowed

values of y are determined by the following equation

o(n R(Y)ln)o o (n (n)2 o = n2 n Z. (6-10)

where In)o is the orthonormal basis for L2(S1). We will proceed with the quantization

of this model by implementing the method discussed in Section 2 for each n sector of

the theory. Since we are not choosing a particular R(y), we will only be discussing the

physical Hilbert space in general. We consider the following two types of solutions to the

constraint equation.

I.) (Point Solutions) The solution y = ym is a point value solution to the equation

(6-10) for a given value of n. The index m corresponds to multiple values of the y that

satisfies the equation for a given value of n.

II.) (Interval Solutions) The solutions y = ym, satisfy the equation (6-10) for all

elements in an interval I(m'). This classification of solutions also includes a countable

union of disjoint intervals. Although physically motivated models exclude such constraint

solutions, we include them to illustrate the versatility of our approach

{ym'} = {m' IR(ym.)= (n)2 VYm' E Im'}. (6-11)

[49] A. Garcia D. Vergara, L. F. Urrutia, "BRST-BFV Method for Nonstationary
Systems," Pb.i; ..: i Review D 51 (1995), 5806.

[50] C. Rovelli, "Partial observables," Phys. Rev. D 65 (2002), 124013.

[51] B. Dittric, "Partial and Complete Observables for Hamiltonian Constrained
Systems," http: //arxiv. org/abs/gr-qc/0411013, (2004).

[52] W. Pauli. "Die allgemeinen Prinzipien der Wellenmechanik," in Handbuch der
Physik 1 Springer, Berlin, 1926.

[53] J. Klauder, "Attractions of Affine Quantum Gravity," http://arxiv. org/abs/
gr-qc/0411055, (2004).

[54] J. Klauder, "Ultralocal Fields and Their Relevance for Reparametrization Invariant
Quantum Field Theory, J. Phys. A 34 (2001), 3277.

[55] S.P. Gavrilov, D.M. Gitman, "Quantization of Systems with Time-Dependent
Constraints. Example of Relativistic Particle in Plane Wave," in Class. Quant. Gray.
10 (1993), 57.

2.1.2 Constraints Appear

Now let us move to the case in which (2-8) fails6 that is

det 0. (2-17)

If (2-17) occurs, then it arises because the conjugate moment are not all independent

[14], since there exist redundant variables in the dynamical variables. In other words, there

exist relations that are associated with the definition of the conjugate momentum (2-7)

O~(p, q) 0 (2-18)

where a E {1,... A}. These relations (2-18) are known as primary constraints [14]. It is

an important to note that the primary constraints are not determined by the equations of

motion. The set of equations (2-18) define a subspace of the phase space called a primary

constraint sub-manifold, whose dimension are 2N A. Technically, we are assuming that

the constraints obey a regularity condition [14]. We will examine this regularity condition,

and instances when it fails in C'! lpter 5. Until then, we will assume and only consider

examples in which these conditions are satisfied.

We can also relate the presence of constraints by considering the Noether theorem.

When a global transformation exists that leaves the action invariant, the result is

a conserved quantity. However, when this is a local transformation, the result is a

constraint. See [19] for details.

It is clear that Hamilton's equations (2-10) and (2-11) are no longer valid if primary

constraints are present. All the dynamics should take place on the primary constraint

surface. We can achieve this by making the following modification to the Hamiltonian,

6 While Dirac may have not been the first to consider this case, his seminal work on the
topic [7] serves as the modern inspiration of the topic

unconstrained systems, we must note that, if our system contains constraints, we cannot

apply the standard quantization techniques. [12].
2.2.2 What About Constraints? The Dirac Method

An important point to make is that the procedure described in the preceding

subsection is done so without the presence of constraints. If we have constraints in a

particular classical system that we are now attempting to quantize, we may not have all

of the mathematical structures required to give us a quantization rule. To address this

important issue Dirac proposed the following procedure. Quantize the entire classical

theory first, then reduce the Hilbert space to the physically relevant -,II-p 1.. '"- The

pertinent question in this method is, "What is the quantum analog to the constraint

equation (2-18)?" In response to this query, we will use Dirac's procedure [7]. To initiate

this procedure, we begin by promoting the constraints to self-adjoint operators,

Qa(p, q) (P, Q). (2-34)

The next step is to determine the kernel of o,, known as the physical Hilbert space [25]

Fp = l|)p| )>)p =0, V a}. (2-35)

If the constraint possesses a zero in continuum of the spectrum, (i.e. suppose that the

constraint =- PI), then we immediately encounter a potential difficulty in implementing

this procedure. Based solely on the construction of the physical Hilbert space we cannot

guarantee that Ip(Ql )pl < 00.

10 The philosophy of quantize first reduce second serves as a in i' '. motivation to the rest
of the dissertation.

observable corresponding to the classical observable f. An immediate consequence of

(2-29), is the following

Q({[(pj) [Q q), Q(pj)]

S[Q(q),Q(pj)] ih6t (2-31)

where (qJ,pj) are the phase space canonical coordinates and 1 is the identity operator

on the Hilbert space Ri. We should note that there is not a definitive method by which to

pick the quantization map since the quantization map is not a homomorphism between

the two algebra. Ambiguity exists in the process because that there are some phase space

functions (e.g. p4q) that would correspond to multiple self-adjoint operators (e.g. P2Qp2

or (p4Q + QP4)/2). It has also shown by [50] there does not exist a quantization map that

can be defined for all elements from the full algebra of the classical observables. Despite

these difficulties we will proceed, noting potential problems due to these ambiguities in the

procedure as necessary.

Thus, for the remaining sections in this work we will assume that we have a

quantization map and are free to use it. The notation that we will use is as follows:

(pj, qt) represent the real c-number of phase space coordinates, while (Pj, Q,) represent the
set of irreducible, self-adjoint operators in which the canonical coordinates are mapped.

The commutator of the (Pj, QJ) follows directly (2-31)

[Qi, P1] ih6J1. (2-32)

Secondly, we promote quantizable phase space observables (f) to self-adjoint operators F:

f(pj,q') v F(Pj, Q). (2-33)

If there exists factor order ambiguity, we will appeal to experiment to select the proper

definition of F. However, as mathematically precise as the quantization procedure is for

To my wonderful wife and soul mate, Megan.


The primary goal of this chapter is to introduce the reader to the concept of

constraints in classical physics. We will discuss the quantization of these classical systems

in the framework proposed by Dirac, [7], as well as, discuss deficiencies in the method,

which will help motivate the development of the projection operator formalism in C! Ilpter


2.1 Classical Picture

A natural starting point for the discussion of constraints is from a classical perspective.

We will begin in the Lagrangian formulation of classical mechanics. In this formulation, we

begin with the action functional

I j dtL(qa, qa,t) (2-1)

where L is the Lagrangian, t is a continuous parameter (often associated with time),

q, E Q, where Q is some configuration space, a E {1, 2,...N},' denotes the derivative with

respect to t, and q, is an element of a fiber of the tangent bundle Q, T7Q. Formally, we

can write the Lagrangian functional as

L : TQ R. (2-2)

Later in the discussion we will define the configuration space, but for now it is just some

C"-manifold. The goal' of classical mechanics is to determine the equations of motion.

The equations of motion are determined by varying the action (2-1) and determining its

1 This of course is from our point of view.


What are the consequences if an action I[q(t)] is invariant under an infinitesimal

temporal transformation, which is the characteristic of a reparametrization invariant

theory, i.e.

I[q(t)] = I[q(t + c(t))]for c < 1? (A-l)

Given the infinitesimal transformation1

S t- + tc(t), (t) = c(t2) = 0, (A-2)

= t = (A-3)
S6q = e t (A 4)

By our assumption that I[q(t)] = I[q(t + e(t))] it follows that

6I = (6L)dt 0 (A-5)

J' 'LL 9L L d
t + sq + 6q)dt (A 6)
t2 L &+L C+ L d + +)dt (A-7)
at t aq 9q dt

/2 (dL+ Oq-d)dt (A-8)

Integrating by parts,
/( de 9L .de
t2 QLd )dt, (A-9)
Jt'i dt 9L dt
t2 q L) )dt. (A-10)


1 In general, we do not have to assume that > 0, we point the reader to [6], for a
discussion on this matter, however, we must insist that A / 0 almost everywhere.

We can also construct the fundamental kernel for the modified case using these results.


ICL( ii; 5/) = (I//O)(O |I) + ('11|OL)(OL I) )+... (4-50)
N"N' (dzi 2 + d'/z2*2 + z3f*2)(dzl2 + d'z 2 + z32)
= VN"N'[1 + + ... ].
2!(d2+ d'2 + 1)

One final note regarding this particular model, as with the case for other partially or fully

quantum mechanical second class constraint systems, the limit as 6 -- 0 is not taken. The

Hilbert space is determined by the space corresponding to the least eigenvalue.

4.4 Conclusions

In this chapter, we have discussed the Projection operator method to deal with

quantum constraints. We have also discussed the quantization models of 3 distinctly

different constraint models. For the remaining chapters of this dissertation, we will be

using the Projection Operator to analyze various quantum constraint situations.

where 0 E C0('M). However, after eliminating the need to deal with each constraint

separately, it is immediately apparent that

{f,M} 0 (3-19)

is not valid for just observable functions but any general function, f, over the phase

space. Thiemann amended this deficit in [33] by offering the following modification of the

identification of an observable to the previous known scheme,

{O, {O,M }}M =o 0 (3-20)

where O is a twice differentiable function. In fact with the scheme, [34], all observables in

a given theory can aslo be identified in the following manner. Suppose, using Thiemann's

notation, we let a^ denote the one-parameter group of automorphisms over the phase

space M, which is defined as the time evolution of the master constraint, it follows that

we can define the ergodic mean [33] of any 0 E C"(M)),

1 fT
O = lim dt a^(0). (3-21)
T-oo 2T J-T

lim dt eit{M}'O(p, q)
T-oo 2T J_
1 'T (t"
= lirm E odt {O, {O, {O, o ,M}...}} (3-22)
T--- 2T f m-

If we assume that we can commute the integral with the Poisson brackets, then it is

easy to see that(3-21) will satisfy (3-20). We will end the discussion with the classical

considerations of the master constraint program on that particular note and address the

issue of quantum observables later in C'! lpter 5 in the context of the Projection Operator


3.3.2 Quantization

The modus operandi of the Master Constraint program is to use well- known and

well-established theorems of self-adjoint operators in Operator theory to construct the

is non-renormalizable. Therefore traditional quantum field theory techniques [1], appear

to be useless when approaching this subject. Canonical quantization schemes of gravity

are also made difficult by the theory's constraint classification [7]. Classically, gravity's

constraints are one algebraic class, but upon quantization the constraints morph into

another type1 Conventional techniques are unsuited for this type of quantum system.

The construction of these techniques does not assume a change from one type of constraint

to another when the system is quantized. The projection operator method is well suited to

handle this situation since all constraints are treated in the same theoretical framework.

Both the Yang-Mills and quantum gravity serve as the primary motivation for this

dissertation. It is hoped that studying simpler models will eventually aid us in studying

more realistic quantum theories.

1.1 Philosophy

When faced with a particular theoretical problem, it has been our approach to follow

the preceding philosophy to obtain an appropriate physical answer. In our analysis we

have followed the time-honored principles that: (1.) Mathematics will give all possible

solutions with no regard to the physics; (2.) When the mathematics leads to a choice,

physics should be the guide in choosing the next step. We will not deviate from this

long-standing point of view in this dissertation.

We also approach problems with the point of view that a "( i'pl!, I. description

of the universe must be a quantum mechanical one. Therefore, a quantum mechanical

description of a particular model will alvi:-, supercede the classical description. This is

the primary reason that we cite the mantra, quantize first, reduce second over and over in

this dissertation.

1 This will be discussed further in C'!i pters 2 and 4

kernel defines a one dimensional Hilbert space. We should note that (4-25) is gaussian

peaked at the classical solution p' = p" = p = 2. As stated we will return to several more

examples of constraints with zeros in the continuum when we delve into this topic during

('! Ilpters 5 and 6.

4.3.2 Closed, First-Class Constrant

The next constraint system under consideration is a set of constraints that force the

angular momentum ji, i E {1, 2, 3} to vanish. With the angular momentum ji =- ijkqJpk,

the action integral we choose is

I, = (pa A bb) dt, (4-26)

where Ab denotes the Lagrange multipliers to enforce the constraints. Note that the

Hamiltonian has been chosen to be zero for simplicity, so we can focus directly on the

issues surrounding the constraints.

From the definition of the ji's, one can immediately determine the Poisson algebra,

given as usual by

{Jji, Jj} = ijk k. (4-27)

Since this bracket yields a Lie algebra, our system is clearly a closed first-class constraint

system [14].

The quantization of this model is straight forward and we promote the dynamical

variables (pj, qi) to the set of irreducible self adjoint operators (Pj, Q'), which obey the

standard Heisenberg relation. The constraint ji are promoted to self-adjoint operators Ji

ji J = CijkQjPk. (4-28)

The projection operator of these constraints (4-28) takes the form of

E(JiJi < 6(h)2). (4-29)

For simplification, we will only assume that R will only contain the first type of

solution. We point the interested reader to the previous chapter or [9] to determine

the physical Hilbert space contribution for type II solutions. The calculation of the

reproducing kernel can be decomposed into portions corresponding to each value n E Z in

the following manner:

o(x',p',; ,pyE(-6 < R(Y) P2 < 6)lx,p ,py)

J= d {o(x',;p'yp|E(P, = n)E(-6 < R(Y) P2 < x,px; y,py)o
n--.oo W, p, Y" py I E!(P

= E _--(x',p' y',py\\E(-6 < R(y) n2 6IX, P; y, )o

= f --o(x', p'y',p|E(-6 < R(Y) n2 < xp y,py)o.y

To determine the point solution contribution, we fix a value for n and proceed as

m(x ,p' y p,; x, p, ;y, py) Inconstant

= f dy' [dyo{x',py',ip[|y'}{y'|E(-6 < R(Y) n2 6)\y}{y\\x,p 9y,py,).

S 'dy ldyox',p,'II >'P|E(-6 < R(y) n2 I6)x, pyP)o6(Y Y)
rym +1/Sm(8)
= dy"(y',p,|y" y"y, p,) (x', p' in,k){n~x,p),

where 1/Sm(6) is the leading S dependency as described in Section 2. For small 1/Sm(6)

values, the integral can be approximated as follows

which is defined on all of M

HE(p, q)= H(p, q) + A (2-19)

where A" are Lagrange multipliers, that enforces the dynamics of the system to occur only

on the sub-manifold. Therefore the equations of motion are given by the following

q = {qH} + A"{q', 0}, (2-20)

j pj, H} + A'{pj, 0}, (2-21)

S= 0. (2-22)

Since the set of primary constraints must be satisfied for all t, it follows that

a = { H} + {, bAb A 0. (2-23)

where w is defined as weakly equal to, or equal to on the constraint sub-manifold. A

direct consequence of (2.1.2) is that the solutions to (2.1.2) may not be independent of

the set of primary constraints (2-18). If this is the case, we define a set of new constraints

(Xb), b E {1,... B} which also satisfies (2.1.2). This set of constraints is called secondary
constraints. We repeat the process of solving the consistency equation (2.1.2) to uncover

all the constraints. With that being said we will ahv--, assume that all constraints have

been uncovered. This statement is often referred to as the set of constraints is complete


Now assuming that the set of constraints is complete, equation (2.1.2) also serves as

the starting point of the discussion of the classification. [7] One possibility for (2-22) to be

valid on the constraint surface is to allow each Poisson bracket to be separately zero by

being proportional to a constraint. This hypothesis leads to our first classification: when

{0, H} hQbb, (2-24)

{1a, Ob = Ccab c, (2-25)

is that which is defined by the coherent states overlap (p', q'lp, q) which also defines the

inner product of the Hilbert space it defines. Using the fact that (p', q' lE|p, q) is a function

of positive type1' so it can be used as a reproducing kernel,11

C4(p',q'; p,q) (p', q'IEp, q). (4-18)

As in the case of the unconstrained Hilbert space, we can express a dense set of

vectors in the functional constraint subspace as

p (p, q) = a (p, q;Pn, n) N < oc. (4-19)

The inner product for these vectors is given by

(Q, TI)p a* i /C3, (pm., qm; pn, qn), (4 20)
m= in= 1

where r] is also an element of the dense set of vectors. Using basic properties of the

reproducing kernel and coherent states, we know that the norm defined by the inner

product of these vectors will be finite [13]. This guarantees that the norm of vectors in the

completion will also be finite. If we multiply a reproducing kernel KC by a constant, the

reproducing kernel KC still corresponds to the same functional space. This is a key point

and one that we exploit in the next section when we deal with a constraint that possesses

a zero in the continuum.

10 That is that Ek QQ1'ak(pF' 'ElK,, q) > 0, for all N < oo and arbitrary complex
numbers {aj} and label sets {pj,qj}.

1 If the projection operator is equal to unity, then we are left with the unconstrained
Hilbert space.

2007 Jeffrey Scott Little

6.7 Refined Algebraic Quantization Approach

Before proceeding with the Refined Algebraic Quantization (RAQ) of this model,
we must first impose some additional technical issues on the constraint (6-2). First we

must assume the constraint (6-2) contains a finite number of zeros and that all stationary

points (i.e. R'(y) = 0, R"(y) = 0, [i.e. the nth dervative of R with respect to y]) only to

have a finite order that no zeros of (6-2) are to be stationary points. As with the analysis

in the preceding sections we must also require that R(y) be positive at least somewhere.

Following the program described in ('!i lpter 3 we must first choose an auxiliary

Hilbert space, 'Hax,. The auxiliary Hilbert space of choice is the Hilbert space of

square-integrable complex functions over the configuration space. The canonical
inner-product is given by the following;

(01, 2) aux f ]= lf 11. i11'(x,y,)2(x, y), (6- 31)

where (.)* denotes complex conjugation. The classical constraint is promoted to an

operator that acts on the auxiliary Hilbert space,

C 02 R(Y), (6-32)

where R(Y) acts as a multiplication operator namely, R(Y)O(x, y) = R(y)((x, y) for all

0(x, y) cE tax. The operator, C is an essentially self-adjoint operator on Haux, therefore
the operator will exponentiate to the one parameter unitary operator via Stone's theorem,

U(t)= e-itc t e R. (6-33)

Keeping in line with the RAQ program, we must now choose a test space 4 C aux-. In
this model the convenient choice is the set of functions of the form;

f(x, y) = _--, f(y), (6-34)

The classical equations of motion for our simple system are given by

S= 0, (5-24)

S= -R'(q), (5-25)

R(q) = 0, (5-26)

with solutions

q(t) q= q=(0), (5-27)

p(t) = -R'(q) A(t')dt' + p(0), (5-28)

where qi is a root of R(q) = 0. If R'(qi) = 0 then the solution becomes

q(t) = qi q(0), p(t) p(0). (5-29)

The function A(t) is not fixed by the equations of motion, which is normal for

first-class constrained systems. To explicitly exhibit a solution to the classical equations

of motion it is generally necessary to specify the function A(t), and this constitutes a

choice of gauge. Gauge dependent quantities are defined to be unobservable, while gauge

independent quantities are declared to be observable. In the present example, if R'(qi) / 0,

then p(t) is gauge dependent, while if R'(qi) = 0, p(t) is, in fact, gauge independent. This

behavior -i .-.-. -I that the momentum p in the subset of the reduced classical phase space

for which {q : R(q) = 0, R'(q) / 0} is unobservable, while the momentum p in the subset

of the reduced classical phase space for which {q : R(q) = 0, R'(q) = 0} is observable.

We discuss this point further below. The reduced classical phase space is given by R x Z,


Z = {q : R(q) = 0}. (5-30)

Clearly, for the classical theory to be well defined, it is sufficient for R(q) E C1,

namely that R(q) and R'(q) are both continuous. (Strictly speaking this continuity is

Full Text








ObtainingaPh.D.inanyeldisneveracompleteindividualeort,Iowemanythankstomanypeoplethathelpedmepursuethislife-longgoal.Firstofall,IwouldliketothankmyadvisorJohnKlauder,forgivingmethechancetostudyquantizationofconstraintssystems.Iamindebtedtohimforallthepatience,time,andencouragementthathehasaordedmeovertheyearsthatIspentunderhistutelageattheUniversityofFlorida.Hispassionforthecourseofstudyhashelpedmeobtainabroaderviewofphysics,aswellasamoredevelopedpalateforvarioustopicsinphysics.InfactIamgratefultotheentirephysicsdepartmentforallowingmetheopportunitytostudytheoreticalphysics.IgratefullyacknowledgetheAlumniFellowshipAssociation,whichallowedmetoattendtheUniversitywithoutanoverwhelmingteachingresponsibility.Iwouldalsoliketothankmygrandparents,Ruby,Granville,Hazel,andVeral,whoinstilledinmefromanextremelyearlyagethatIcouldaccomplishanythingifIsetmymindtoit.Thanksgotomydad,Je,whogavemeasenseofscienticcuriosityandtomymom,Linda,whowasmywonderfulmathinstructorfromfractionstocalculus,nottomentionalloftheirloveandsupport,andtomysisters,LisaandSierra,whoseconstantencouragementaidedmethroughmyearlycollegeandgraduatecareer.IamassuredlyindebtedtomydearAuntBrenda,whocarefullyeditedseveralchaptersofthismanuscript,eventhoughsheisnotaphysicsperson.Iwouldalsoliketothankmywife'sfamilyforalloftheirsupportthesepast2years.ThanksgotomyfriendsattheUniversityofFloridaLarry,Ethan,Wayne,Ian,Lester,Jen,andGarret,whosecompassionandconversationsaboutawidevarietyoftopicsareunmeasurable.Savingthebestforlast,Ithanktheloveofmylife,mydarlingwife,Megan.WithoutwhomIwouldhavenevercompletedthisdissertation.Ithankherforallofthelove,support,andencouragementthatshehasgivenme;sheisthesourceofmyinspirationtoachieve,morethanIeverdreamedcouldbeachieved. 4


page ACKNOWLEDGMENTS ................................. 4 ABSTRACT ........................................ 8 CHAPTER 1INTRODUCTION .................................. 10 1.1Philosophy .................................... 11 1.2OutlineoftheRemainingChapters ...................... 12 2CONSTRAINTSANDTHEDIRACPROCEDURE ............... 14 2.1ClassicalPicture ................................ 14 2.1.1GeometricPlayground ......................... 16 2.1.2ConstraintsAppear ........................... 18 2.1.3AnotherGeometricInterlude ...................... 20 2.1.4Observables ............................... 21 2.2Quantization .................................. 22 2.2.1CanonicalQuantizationProgram ................... 22 2.2.2WhatAboutConstraints?TheDiracMethod ............ 24 3OTHERMETHODS ................................. 27 3.1Faddeev-PopovMethod ............................ 27 3.1.1YetAnotherGeometricInterludefromtheConstraintSub-Manifold 27 3.1.2BasicDescription ............................ 28 3.1.3CommentsandCriticisms ....................... 29 3.2RenedAlgebraicQuantization ........................ 31 3.2.1BasicOutlineofProcedure ....................... 32 3.2.2CommentsandCriticisms ....................... 33 3.3MasterConstraintProgram .......................... 34 3.3.1ClassicalDescription .......................... 35 3.3.2Quantization ............................... 36 3.3.3MCPConstraintExample ....................... 39 3.3.4CommentsandCriticisms ....................... 41 3.4Conclusions ................................... 42 4PROJECTIONOPERATORFORMALISM .................... 43 4.1MethodandMotivation ............................ 43 4.1.1SquaringtheConstraints ........................ 45 4.1.2ClassicalConsideration ......................... 45 ................... 46 .............. 47 5


................. 48 4.2.1CoherentStates ............................. 48 4.2.2ReproducingKernelHilbertSpaces .................. 49 4.3ConstraintExamples .............................. 51 4.3.1ConstraintwithaZerointheContinuousSpectrum ......... 51 4.3.2Closed,First-ClassConstrant ..................... 52 4.3.3Open,FirstClassconstraint ...................... 54 4.4Conclusions ................................... 58 5HIGHLYIRREGULARCONSTRAINTS ...................... 59 5.1Classication .................................. 59 5.2ToyModel .................................... 64 5.3Observables ................................... 72 5.4ObservationandConclusions .......................... 75 6ASHTEKAR-HOROWITZ-BOULWAREMODEL ................. 76 6.1Introduction ................................... 76 6.2ClassicalTheory ................................ 77 6.3QuantumDynamics ............................... 79 6.4ThePhysicalHilbertSpaceviatheReproducingKernel ........... 79 6.4.1TheTorusT2 79 6.5Super-selectionSectors? ............................ 84 6.6ClassicalLimit ................................. 85 6.7RenedAlgebraicQuantizationApproach .................. 88 6.8CommentaryandDiscussion .......................... 90 7PROBLEMWITHTIME .............................. 92 8TIMEDEPENDENTCONSTRAINTS ....................... 94 8.1ClassicalConsideration ............................. 94 8.1.1BasicModel ............................... 94 8.1.2CommentaryandDiscussion ...................... 97 8.2QuantumConsiderations ............................ 99 8.2.1GitmanandTyutinPrescriptionforTime-DependentSecond-ClassConstraints ............................... 99 8.2.2CanonicalQuantization ......................... 101 8.2.3Dirac ................................... 102 8.2.4ProjectionOperatorFormalism .................... 102 8.2.5Time-DependentQuantumConstraints ................ 103 8.2.6ObservationsandComparisons ..................... 105 9TIME-DEPENDENTMODELS ........................... 106 9.1First-ClassConstraint ............................. 106 9.2SecondClassConstraint ............................ 110 6


................................... 111 10CONCLUSIONSANDOUTLOOK ......................... 113 10.1Summary .................................... 113 10.2EndingonaPersonalNote ........................... 115 APPENDIX AREPARAMETERIZATIONINVARIANTTHEORIES .............. 116 REFERENCES ....................................... 118 BIOGRAPHICALSKETCH ................................ 122 7


Motivatedbyseveraltheoreticalissuessurroundingquantumgravity,acourseofstudyhasbeenimplementedtogaininsightintothequantizationofconstrainedsystemsutilizingtheProjectionOperatorFormalism.Throughoutthisdissertationwewilladdressseveralmodelsandtechniquesusedinanattempttoilluminatethesubject.WealsoattempttoillustratetheutilityoftheProjectionOperatorFormalismindealingwithanytypeofquantumconstraint. Quantumgravityismademoredicultinpartbyitsconstraintstructure.Theconstraintsareclassicallyrst-class;however,uponquantizationtheybecomepartiallysecond-class.Tostudysuchbehavior,wewillfocusonasimpleproblemwithnitelymanydegreesoffreedomandwilldemonstratehowtheProjectionOperatorFormalismiswellsuitedtodealwiththistypeofconstraint. Typically,whenonediscussesconstraints,oneimposesregularityconditionsontheseconstraints.Weintroducethe\new"classicationofconstraintscalled\highlyirregular"constraints,duetothefacttheseconstraintscontainbothregularandirregularsolutions.Quantizationofirregularconstraintsisnormallynotconsidered;however,usingtheProjectionOperatorFormalismweprovideasatisfactoryquantization.Itisnoteworthythatirregularconstraintschangetheobservableaspectsofatheoryascomparedtostrictlyregularconstraints.Morespecically,wewillattempttousethetoolsoftheProjectionOperatorFormalismtostudyanothergravitationallyinspiredmodel,namely 8


Finally,wewillusetheProjectionOperatorMethodtodiscusstime-dependentquantumconstraints.Indoingso,wewilldeveloptheformalismandstudyafewkeytime-dependentmodelstohelpusobtainalargerpictureonhowtodealwithreparameterizationinvarianttheoriessuchasGeneralRelativity. 9


\Itisveryimportantthatwedonotallfollowthesamefashion...Itsnecessarytoincreasetheamountofvariety....theonlywaytodoitistoimploreyoufewguystotakearisk..."-RichardFeynman 1 ].Thisisakeycharacteristicofconstrainedsystems.Theprocessofconvertingaclassicaltheorytoaquantumtheoryismademoredicultbythepresenceoftheseconstraints.Commonlyusedtechniquestodealwiththesesystemshavebeeninadequateinprovidingadescriptionofthelowmomentumbehaviorofthestrongforce,whichisassociatedwiththemass-gapconjecture[ 2 ]. Generalrelativity,liketheStandardModel,isanotherexampleofaconstrainedsystem.Thequantizationofgravityhaspresentedtheoreticalphysicsacornucopiaofproblemstosolveforthepast50years.Toanswerthesedeeptheoreticalquestions,physicistshaveemployedseveralandseeminglyconictingviewpoints.Theseperspectives,rangefromSuperstrings[ 3 ],themaingoalistheunicationofallforcesinonequantummechanicaldescription,LoopQuantumGravity[ 4 ],inwhichthemainobjectiveofthisisaconsistentbackgroundindependentdescriptionofquantumgravity,toCausalSets[ 5 ]inwhichtheapproachpreassumesthatspace-timeisdiscretized,andtheAneQuantumGravityProgram[ 6 ],inwhichtheaimofthisapproachistosolvequantumconstraintproblemswiththeProjectionoperatorformalism.Thereareseveraldeepunderlyingtheoreticalissuessurroundingthequantizationofgravity,oneofwhichisthatgravity 10


1 ],appeartobeuselesswhenapproachingthissubject.Canonicalquantizationschemesofgravityarealsomadedicultbythetheory'sconstraintclassication[ 7 ].Classically,gravity'sconstraintsareonealgebraicclass,butuponquantizationtheconstraintsmorphintoanothertype Wealsoapproachproblemswiththepointofviewthata\complete"descriptionoftheuniversemustbeaquantummechanicalone.Therefore,aquantummechanicaldescriptionofaparticularmodelwillalwayssupercedetheclassicaldescription.Thisistheprimaryreasonthatwecitethemantra,quantizerst,reducesecondoverandoverinthisdissertation. 11


8 ].Wheneverencounteredinthedissertation,repeatedindicesaretobesummed. TheprimarygoalofChapter5istointroducea\new"classicationofconstraintscalledhighlyirregularconstraintsandalsoillustratetechniquesusedtodealwithquantumversionsoftheseconstraints.Thebasisofthischaptercomesfrom[ 9 ]and[ 10 ].UsingthetechniquesgainedfromChapter5,inChapter6weoeracompletediscussionofthequantizationoftheAshtekar-Horowitz-BoulwareModel[ 11 ].TheAshtekar-Horowitz-Boulwaremodelisamathematicalmodelalsoinspiredbytheconstraintsofgravity.Thischapterisbasedontheresultsobtainedin[ 9 ].WealsocomparetheresultsobtainedbytheRenedAlgebraicQuantizationprogramwiththeProjectionOperatorformalism.Thiscomparisonleadstotheconclusionthatthetwomethodsarenotcompatibledealingwithallconstraints. Theremainingchaptersaredevotedtothetopicoftime-dependentquantumconstraints.Untilnow,themethodsusedtodelveintothistopic[ 12 ]havebeenunsatisfactoryduetothefactthatthesemethodsavoidsolvingforthequantumconstraints.Theaim 12


12 ]withtheProjectionOperator.InChapter9,wewillexaminetwotime-dependentconstraints,onerst-classandonesecond-class.Wewillconcludewithabriefsummaryandapossiblelookforwardtofutureresearchproblems. 13


Theprimarygoalofthischapteristointroducethereadertotheconceptofconstraintsinclassicalphysics.WewilldiscussthequantizationoftheseclassicalsystemsintheframeworkproposedbyDirac,[ 7 ],aswellas,discussdecienciesinthemethod,whichwillhelpmotivatethedevelopmentoftheprojectionoperatorformalisminChapter4. whereListheLagrangian,tisacontinuousparameter(oftenassociatedwithtime),qa2Q,whereQissomecongurationspace,a2f1;2;:::Ng,_denotesthederivativewithrespecttot,and_qaisanelementofaberofthetangentbundleQ,TQ.Formally,wecanwritetheLagrangianfunctionalas Laterinthediscussionwewilldenethecongurationspace,butfornowitisjustsomeC1-manifold.Thegoal 2{1 )anddeterminingits 14


(2{3) dt@L @_qa@L @qa=0 (2{4) Whilethisformulationisusefulindeterminingawidevarietyofphysicalquantities[ 15 ],itisnotassensitivetoparticularfeaturesofagivenclassicaltheoryasweneed. @_qa:(2{5) ThiscanberecognizedastheberderivativefromthetangentbundleofQtothecotangentbundleofQ[ 15 ](otherwiseknownasthetangentbundle'sdual) ThenextstepintheprocedureistoidentifytheHamiltonian,whichfollowsfromaLegendretransformationoftheLagrangian qa=(tt0)@L @qa+_(tt0)@L @_qa++(k)(tt0)@L @qka[ 14 ].Integratingthisequationwithrespecttot0andsettingtheresulttozerowillyieldthestationarypointsofthecorrespondingaction.3 2{4 intermsofasecondorderdierentialequation,thesubtlepointweareabouttomakebecomesclearer.[ 16 ] 15


2{5 )isinvertibleintermsof_qa.ThisconditionissatisedbytheHessiancondition det@2L @_qa@_qb6=0:(2{8) Wewillreturntothecasewhen( 2{8 )failsshortly.UntilthenwewilldiscusstheHamiltonianformalisminmoredetail.Foramorecompleteaccountsee[ 17 ]and[ 15 ].Atthispointwewillnolongermentionthecotangentbundle,butratherwewillnotethatthisspaceissymplectomorphictothemorefamiliarspace,phasespaceM,i.e., ThegeometricframeworkoftheHamiltonianframeworkisarichandbeautifulsubject.However,forthesakeofbrevitywewillonlyrecountthemostcrucialelementstothedevelopmentoftheconstraintpicture.Foramorecompletedescription,wepointthereadersto[ 17 ],[ 15 ],and[ 18 ]. 2{7 )aregivenby _qi=fqi;Hg; _pj=fpj;Hg; 16


@qi@g @pi@f @pi@g @qi;(2{12) wheref;g2C2(M).ThePoissonbracketshavethefollowingpropertiesforanyf;g,andh2C2(M): Equation( 2{13 )illustratesthatthePoissonbracketsareantisymmetricinrespecttoitsarguments.Equation( 2{14 )servesastheconnectionofthePoissonbracketwithpoint-wisemultiplicationofthefunctionsoverthephasespace.Finally,( 2{15 )demonstratesthatthePoissonbracketobeysthefamousJacobiidentity.WiththesethreepropertiesitispossibletoshowthattheclassicalfunctionsoverphasespaceformaLiealgebrawithrespecttothePoissonbracket.[ 14 ] Thesymplecticformisacrucialelementwhenwemovefromaclassicaldiscussionofasystemtoitsquantumanalogue. 17


2{8 )fails det@2L @_qa_qb=0:(2{17) If( 2{17 )occurs,thenitarisesbecausetheconjugatemomentaarenotallindependent[ 14 ],sincethereexistredundantvariablesinthedynamicalvariables.Inotherwords,thereexistrelationsthatareassociatedwiththedenitionoftheconjugatemomemtum( 2{7 ) wherea2f1;:::Ag.Theserelations( 2{18 )areknownasprimaryconstraints[ 14 ].Itisanimportanttonotethattheprimaryconstraintsarenotdeterminedbytheequationsofmotion.Thesetofequations( 2{18 )deneasubspaceofthephasespacecalledaprimaryconstraintsub-manifold,whosedimensionare2NA.Technically,weareassumingthattheconstraintsobeyaregularitycondition[ 14 ].Wewillexaminethisregularitycondition,andinstanceswhenitfailsinChapter5.Untilthen,wewillassumeandonlyconsiderexamplesinwhichtheseconditionsaresatised. WecanalsorelatethepresenceofconstraintsbyconsideringtheNoethertheorem.Whenaglobaltransformationexiststhatleavestheactioninvariant,theresultisaconservedquantity.However,whenthisisalocaltransformation,theresultisaconstraint.See[ 19 ]fordetails. ItisclearthatHamilton'sequations( 2{10 )and( 2{11 )arenolongervalidifprimaryconstraintsarepresent.Allthedynamicsshouldtakeplaceontheprimaryconstraintsurface.WecanachievethisbymakingthefollowingmodicationtotheHamiltonian, 7 ]servesasthemoderninspirationofthetopic 18


whereaareLagrangemultipliers,thatenforcesthedynamicsofthesystemtooccuronlyonthesub-manifold.Thereforetheequationsofmotionaregivenbythefollowing _qi=fqi;Hg+afqi;ag; _pj=fpj;Hg+afpj;ag; Sincethesetofprimaryconstraintsmustbesatisedforallt,itfollowsthat _a=fa;Hg+fa;bgb0:(2{23) whereisdenedasweaklyequalto,orequaltoontheconstraintsub-manifold.Adirectconsequenceof( 2.1.2 )isthatthesolutionsto( 2.1.2 )maynotbeindependentofthesetofprimaryconstraints( 2{18 ).Ifthisisthecase,wedeneasetofnewconstraints(b),b2f1;:::Bgwhichalsosatises( 2.1.2 ).Thissetofconstraintsiscalledsecondaryconstraints.Werepeattheprocessofsolvingtheconsistencyequation( 2.1.2 )touncoveralltheconstraints.Withthatbeingsaidwewillalwaysassumethatallconstraintshavebeenuncovered.Thisstatementisoftenreferredtoasthesetofconstraintsiscomplete[ 14 ]. Nowassumingthatthesetofconstraintsiscomplete,equation( 2.1.2 )alsoservesasthestartingpointofthediscussionoftheclassication.[ 7 ]Onepossibilityfor( 2{22 )tobevalidontheconstraintsurfaceistoalloweachPoissonbrackettobeseparatelyzerobybeingproportionaltoaconstraint.Thishypothesisleadstoourrstclassication:when 19


Ifdetfa;bg6=0,theconstraintsareclassiedassecondclass[ 16 ].Nolongerhavingtheavailabilityoftheprecedingcriteriaof( 2{24 )and( 2{25 ),itfollowsthattheLagrangemultipliersaredeterminedbytheequationsofmotionsothat( 2{22 )issatised.TheLagrangemultipliersforcethedynamicstoremainontheconstraintsurfaceforasecond-classsystem.NamelytheLagrangemultiplierscanbedeterminedbythefollowingequations 20


14 ]foramoredetailedaccountofthepre-symplecticform.Amoreindepthdiscussionofthegeometryofconstraintsurfacewilloccurinthefollowingchapter. whereo2C1(M).Inthecontextofrst-classconstraints( 2{27 )isasucientconditionthatguaranteesagaugeinvariantfunction[ 14 ].Ifoisanobservable,itisclearbythedenitionofanobservablefunction( 2{27 )thato+aaisalsoanobservable.UsingthisobservationitispossibletopartitionthesetoffunctionsC1(M)byvirtueofthisequivalencerelationintoobservablefunctionsandnon-observablefunctions.Intakingthediscussionfurther,ifweweretoconsiderthevectorspaceofC1(M)equippedwiththePoissonbrackets,whichdenesaLiealgebra,alongwithpointwisemultiplication,wecanthenidentifythefunctionsthatvanishontheconstraintsub-manifold(i.e.theconstraints)astheidealNinC1(M).WecanclassifyalgebraofobservablefuctionsasthequotientalgebraC1(M)=N.[ 14 ]Thisidenticationofthealgebraicstructureoftheobservables 14 ] 21


2.2.1CanonicalQuantizationProgram 20 ].Thegoalofthecanonicalquantizationprocedureistondarulethatassociatesphasespacefunctionswithself-adjointoperators.Knowingthegoaloftheprogram,let'sbegintheimplementation.First,withwemustinsistthatthecoordinatesofphasespacebeatcoordinates,whichimpliestheremustbeaglobalCartesiancoordinatepatchfortheentirephasespacemanifold. 22 ]and[ 23 ]. (2{28) (2{29) (2{30) whereSA(H)isthesetofself-adjointoperatorsactingontheHilbertspaceHandf;g2C1(M).Theself-adjointoperatorQ(f)shouldberecognizedasthequantum 20 ],andmorerecentlybyKlauder[ 21 ].AccordingtoKlauder[ 21 ],thismetricstructurecomesintheformofa\shadow"metricwhichisproportionalto~.9 22


2{29 ),isthefollowing where(qj;pj)arethephasespacecanonicalcoordinatesand1istheidentityoperatorontheHilbertspaceH.Weshouldnotethatthereisnotadenitivemethodbywhichtopickthequantizationmapsincethequantizationmapisnotahomomorphismbetweenthetwoalgebra.Ambiguityexistsintheprocessbecausethattherearesomephasespacefunctions(e.g.p4q)thatwouldcorrespondtomultipleself-adjointoperators(e.g.P2QP2or(P4Q+QP4)=2).Ithasalsoshownby[ 50 ]theredoesnotexistaquantizationmapthatcanbedenedforallelementsfromthefullalgebraoftheclassicalobservables.Despitethesedicultieswewillproceed,notingpotentialproblemsduetotheseambiguitiesintheprocedureasnecessary. Thus,fortheremainingsectionsinthisworkwewillassumethatwehaveaquantizationmapandarefreetouseit.Thenotationthatwewilluseisasfollows:(pj;qi)representtherealc-numberofphasespacecoordinates,while(Pj;Qi)representthesetofirreducible,self-adjointoperatorsinwhichthecanonicalcoordinatesaremapped.Thecommutatorofthe(Pj;Qj)followsdirectly( 2{31 ) [Qi;Pj]=i~ij1:(2{32) Secondly,wepromotequantizablephasespaceobservables(f)toself-adjointoperatorsF: Ifthereexistsfactororderambiguity,wewillappealtoexperimenttoselecttheproperdenitionofF.However,asmathematicallypreciseasthequantizationprocedureisfor 23


12 ]. 2{18 )?"Inresponsetothisquery,wewilluseDirac'sprocedure[ 7 ].Toinitiatethisprocedure,webeginbypromotingtheconstraintstoself-adjointoperators, Thenextstepistodeterminethekernelofa,knownasthephysicalHilbertspace[ 25 ] Iftheconstraintpossessesazeroincontinuumofthespectrum,(i.e.supposethattheconstraint=P1),thenweimmediatelyencounterapotentialdicultyinimplementingthisprocedure.BasedsolelyontheconstructionofthephysicalHilbertspacewecannotguaranteethatjPhjiPj<1. 24


2.1.2 ).Inessencethismeanswemustconsider [a(P;Q);H(P;Q)]jiP=0; [a(P;Q);b(P;Q)]jiP=0; whereH(P;Q)istheunconstrained,self-adjointHamiltonianoperator.Onceagain,wearefacedwithapossibledeciencyoftheDiracprocedure.Ingeneral,wecannotattesttothevalidityoftheseequations,butifwerestrictourargumentstoconsideringonlyclosed,rst-classsystems( 2{36 )and( 2{37 )willholdtrue.Inthecaseofclosed,rst-classsystems,thePoissonbracketstransformsintothecommutatorbrackets,whichareexpressedinthefollowingform: [a(P;Q);H(P;Q)]=i~hbab(P;Q); [a(P;Q);b(P;Q)]=i~ccabc(P;Q): IfEquation( 2{36 )or( 2{37 )failsandtheclassicalsystemisclassiedasrstclass,thequantumsystemissaidtohaveananomaly.WewillexaminesuchasysteminChapter4.Furthermore,wendthatourdenitionforthephysicalHilbertspacemaybevacuouswhenconsideringclassicallyopen,rst-classorsecond-classsystemssincetheremaynotbeazerointhespectrum[ 13 ].Diracattemptstoremedytheproblemofsecond-classconstraintsbyredeningthePoissonbracket[ 7 ].Therefore,thestandardapproachintheDiracprocedureprefersclosed,rst-classsystems.WewillreturntoadiscussionoftheDiracbracketinChapter8. AnotherdeciencytonoteisthefacttheDiracproceduredoesnotoeradenitionoftheinnerproductofthephysicalHilbertspace.This,alongwithsomeoftheotherdecienciesthatareillustratedinthischapter,willserveastheprimarymotivationforthediscussionofthemoremodernmethodstoquantizeconstraintsystemsdiscussedin 25


13 ]inChapter4. 26


Theprimaryobjectiveofthischapteristoexaminethreedistinctmethodstodealwithquantumconstraints.ThesemethodsaretheFaddeev-Popovprocedure,theRenedAlgebraicQuantizationProgram,andtheMasterConstraintProgram,eachofwhichhasitsowndistinctstrengthsandweaknesses.Duringthischapterwewillusethenotationthatisstandardinliterature,whilealsonotingdecienciesofthemethodsinordertooermoremotivationforthestudyoftheprojectionoperatormethod,whichisthetopicofChapter4. 3.1.1YetAnotherGeometricInterludefromtheConstraintSub-Manifold wherej2f1;:::;Ng,a2f1;:::;Ag,andaareLagrangemuliplierswhichenforcetheconstraintsa 14 ].Agaugeorbitisdenedbythefollowing:considerthatFdenesaparticularphysicalconguration, 12 ]Agaugetransformationisdenedas 27


14 ].Ifonedenesasetofsurfaceformingvectors,thegaugeorbitswouldcorrespondwiththenullvectors.[ 14 ] Toavoidthisrathercomplicatedsituationofthequotient,itisoftensuggestedthatonemustimposeagaugechoicetoeliminatetheredundancy.Agaugechoice(a)hasthefollowingproperty wherea2f1;:::;Ag[ 14 ].Wemustalsochoosesuchafunctionthatintersectsthegaugeorbitsonceandonlyonce.Awordofwarning{Onecanguaranteethisisthecaselocally;however,itmaynotbeguaranteedglobally,(i.e.fortheentireconstraintsurface).ThisisknownastheGribovproblem[ 16 ].However,weareconsideringtheidealcaseforthisdiscussion. Withthismathematicaldescriptionestablished,wecannowproperlydiscusstheFadeev-Popovprocedure.[ 27 ] Tosolvetheconstraintprobleminthisframework,weasserttheconstraintsaresatisedclassicallywithinthefunctionalintegralbyimposinga-functionaloftheconstraints.Sincetheresultingintegralmaybedivergent,wesuppressthispossibility,byrequiringachoiceofanauxiliaryconditioncalledagaugexingtermoftheforma(p;q)=0;a2f1;::;Jg.Withthischoicewehavelostcanonicalcovariance,whichcanberestored 28


Expression( 3{4 )servesasamotivationtotheintroductionof( 3{5 ),buttheyarenottobeviewedasequivalentstatements.Theresultof( 3{5 )couldthenbeexpressedasapathintegraloverthereducedphasespace, wherepandqarereducedphasecoordinatesandH(p;q)istheHamiltonianofthereducedphasespace.Sincewehavesatisedtheconstraintsclassically,wearenolongercondentthatourformalpathintegralisdenedoverEuclideanspace.Thispresentsadilemmasincetheformalpathintegralisill-denedovernon-Euclideanspaces[ 13 ].AswiththeDiracProcedure,theFaddeevmethodcanbemodiedtoaccommodatesecond-classconstraints[ 28 ]. 1 ],itisnotwithoutitsaws.Oneofthemostglaringawsisthefactthatonemustrstreducetheclassicaltheoryandthenquantizeit.Theuniverse 3 ].Wearenotsoboldtosaythatquantummechanicsmaybesupercededbyamorecompletedescriptionofnature.Ofcourse,weassumehoweverquantummechanicsistheproperroutetolookatnatureuntilmoreevidenceisdiscovered. 29


wherej2f1;:::;Nganda2f1;:::;Ng.Thissystemispurelyconstraint af(p;q)a(p;q);(3{8) wherefisanon-vanishingfunctionoverthephasespaceandfagAa=1denesaclosed-rstclassconstraint,i.e. whereccabisaconstant.Itisclearwhile( 3{8 )denesthesameconstraintsub-manifoldasthecaseinwhichaareconstraint,theconstraintsaareanopen,rst-classsystem isreplacedwiththegauge-xedexpression whereb(p;q)issomeappropriategaugechoice.Asimpleidentityleadsto 8 ]5 8 ],whichwewillreturntointhenextchapter. 30


Weobservethatallthefactorsoffcompletelycancel.Asonecansee,theFaddeevmethodisinsensitivetothedenitionoff,aslongasitbenon-zero.Hence,thismethodconsidersthea'sanda'sasidenticalconstraints.WewillexamineasimilarmodelinChapter4[ 8 ],whichdemonstratesthatinordertounderstandtheentiretheory,onemustalsoconsiderthequantummechanicalcorrections. Anotherdicultyinthismethodderivesfromtheselectionofgaugechoicea.Aswenotedintheprevioussubsection,thechoiceofgaugeisonlyguaranteedlocally.Inmorecomplicatedgaugetheories,suchasYang-Millsitiswellknown[ 29 ]thattheredoesnotexistagaugechoicethatslicesthegaugeorbitsonceandonlyonce,afactwhichlimitstheeectivenesstoprobethenon-perturbativeregimeofthesegaugetheories.[ 26 ] 30 ]RenedAlgebraicQuantizationcomesintwomainvarieties,GroupAveragingandamorerigorousversionthatisbasedonthetheoryofriggedHilbertspaces.Inthischapterwewillfocusontheformerratherthanthelatterbecausemostexpertswillagreethattheredoesnotexistagroupaveragingtechniqueforallconstraintsinthisformalism.See[ 30 ]and[ 31 ]formorecompletediscussionsonthefailuresofgroupaveraging. 31


31 ]AnothertechnicalrequirementisthatforeveryA2Aobs,(i.e.thealgebra ThenalstageoftheRAQprocedureentailsconstructingananti-linearmapcalledariggingmap, 32


(Ci(1))[2]=(1)[Ci2]=0:(3{15) Inadditionto( 3{15 ),theriggingmap( 3{14 )mustalsosatisfythefollowingtwoconditions,whicharetrueforevery1;22:1:Theriggingmapisrealandpositivesemi-denite(1)[2]=(2)[1];(1)[1]0:2:TheriggingmapinteterwineswiththerepresentationsoftheobervablealgebraO(1)=(O1); Oncetheriggingmaphasbeendetermined,thevectorsthatspanthesolutionspaceareCauchycompletedwithrespecttothefollowinginner-product forevery1;22andhjiistheinnerproductoftheauxiliaryHilbertspace.Thus,wedenethephysicalHilbertspacederivedbythetechniquesoftheRAQ.WewillrevisittheRenedAlgebraicQuantizationprograminChapter6inthecontextoftheAshtekar-Horowitz-Boulwaremodel[ 32 ]. 33


32 ]ForthisprimaryreasontheRAQprocedurehasdiculty 33 ],inanattempttoovercomesituationsinwhichtheRAQprocedurefails.These\failures"include,butarenotlimitedto,casesinwhichaninnitenumberofconstraintsarepresent,aswellaswhenthestructurefunctionsarenotconstants,butratherarefunctionsoverthephasespace.ThisprogramalsoattemptstoeliminateotherambiguityfromtheRAQprocedure,namelytherequirementofadditionalinputintothephysicaltheory.Asmentionedbefore,thisadditionalinputisadenseandinvariantsubspacewhichisequippedwithanertopologythanthatoftheHilbertspaceinwhichitisembedded.[ 34 ]Duringthissection,sinceweonlyintendtogiveaheuristicacountoftheMasterConstraintProgram,wewill 31 ] 34


33 ].Particularly,wewilldescribetheclassicalanalogandthendiscussthequantizationoftheclassicaltheory. 2Xj;k2ICj(p;q)gjkCk(p;q)(3{17) wheregijischosentobepositivedenite. 33 ]and[ 34 ]10 3{17 )iftheconstraintsareactuallyelds.Ifthisisthecase,wemustsmearthemoversomesetoftestfunctions.Formoredetailsonthisproceduresee[ 33 ]. 35


isnotvalidforjustobservablefunctionsbutanygeneralfunction,f,overthephasespace.Thiemannamendedthisdecitin[ 33 ]byoeringthefollowingmodicationoftheidenticationofanobservabletothepreviousknownscheme, whereOisatwicedierentiablefunction.Infactwiththescheme,[ 34 ],allobservablesinagiventheorycanaslobeidentiedinthefollowingmanner.Suppose,usingThiemann'snotation,weletMtdenotetheone-parametergroupofautomorphismsoverthephasespaceM,whichisdenedasthetimeevolutionofthemasterconstraint,itfollowsthatwecandenetheergodicmean[ 33 ]ofanyO2C1(M), O=limT!11 2TZTTdtMt(O): =limT!11 2TZTTdteitfM;gO(p;q)limT!11 2TZTT1m=0dt(t)m IfweassumethatwecancommutetheintegralwiththePoissonbrackets,thenitiseasytoseethat( 3{21 )willsatisfy( 3{20 ).WewillendthediscussionwiththeclassicalconsiderationsofthemasterconstraintprogramonthatparticularnoteandaddresstheissueofquantumobservableslaterinChapter5inthecontextoftheProjectionOperatorFormalism. 36


36 ]toafullyinteractingquantumeldtheory[ 37 ].Therefore,inordertoproceedwiththequantizationoftheMCP,onemustrstpromotethemasterconstraint( 3{17 )toaself-adjointoperatorthatactsonanauxiliaryHilbertspace Themaindierence,atthispointinthediscussion,betweentheauxiliaryHilbertspaceofMCPandthatofRAQ,isthatMCPrequirestheHilbertspacetobeseparable. Atrstglanceitmayappearthatonehaseliminatedthepossiblequantumanomalybecauseacommutatorofanyoperatorwiththesameoperatoriszero,[^M;^M]=0.Whilethisisatruestatement,thequantumanomalyhasonlybeenreformulatedinanothermanner,videlicetthespectrumof^Mmaynotcontainzero.Anexamplethatillustratesthispointmoreclearlyisasfollows:Consideraclassicalsystemwithaclassicalphasesspace,R,withtwoconstraints,C1=p1C2=q1: 8 ].However,itwillprovideadistinctpropertythatweareattemptingtoillustrate,whichisthequantummasterconstraintneednotpossesa0inthespectrum.Thecorrespondingmasterconstraintofthissystemcanbewrittenasthefollowing, 37


33 ]oeredameanstorectifythisbysuggestingamodicationofthequantummasterconstraintbythefollowing: ^~M=^MI(3{25) where=inffspectrum(^M)andIistheidentityoperatorontheauxiliaryorakinematicalHilbertspace.AccordingtoThiemann,[[ 33 ]],( 3{25 )willstillhavethesameclassicallimitofthemasterconstraintbecause/~.Ingeneral,ifasystemcontainsaconstraintthatisclassicallyanopen,rst-classconstraint,likegravityorthesystemthatwewilldisscussinChapter4,wherewouldbeproportionalto~2. Assumingtheoperator^~Misadenselydenedself-adjointoperator,wecannowproceedwiththequantizationbyrstaddressingtheauxiliaryHilbertspace.Usingthefactthat^~Misaself-adjointoperatorwithapositivesemi-denitespectrum,theauxiliaryHilbertspacecanbewrittenasthefollowingdirectintegral[ 34 ]; whered(x)isthespectralmeasure[ 33 ]ofthemasterconstraintoperator( 3{25 ).Eachaddendcontributiontothesum,Haux(x),in( 3{26 )isaseparableHilbertspacewiththeinnerproductinducedbytheauxiliaryHilbertspace,Haux. Usingthisparticularconstructionwearenowabletoaddressthetaskofsolvingthequantummasterconstraintequation^~M=0.BythemathematicaldescriptionoftheauxiliaryHilbertspace( 3{26 ),itfollowsthattheactionof^~MonHaux(x)issimplymultiplicationofx.WecansolvethequantummasterconstraintequationbyidentifyingthephysicalHilbertspacebythefollowing 38


wherei2f1;2;3gqiandpiarethecanonicalpositionandcanonicalmomentumrespectively 3{28 )formaclosedLieAlgebrathatwecanidentifywiththealgebraofso(3): Theclassicalmasterconstraint( 3{17 ),correspondstotheCasmiroperatorofthegroup Thequantizationofthismodelisstraightforward.TheauxiliaryHilbertspaceisthesetofallsquareintegrablefunctionsoverR3,alsoknownasL2(R3).Thecanonicalpositionqiispromotedtoaself-adjointoperatorQiinwhichtheactionontheauxiliaryHilbertspaceismultiplicationbyqi.Theconjugatemomentumpiispromotedtoaself-adjointoperatorPiinwhichtheactiononL2(R3)isdierentiation,Pi=i~@ @qi.The 35 ].12 39


3{30 )ispromotedtoacorrespondingself-adjointoperator ^M=^Jyi^Ji:(3{31) Sincethereisnoorderingambiguityin( 3{31 ),onedoesnotneedtosubtracttoobtainazerointhespectrumof( 3{31 ).Usingthetechniquesthatarewellknowninquantummechanics[ 38 ],wewillusesphericalcoordinatestodeterminetheeigenvaluesofthequantummasterconstraint( 3{31 ) sin2@2 sin@ @(sin@ @));(3{32) where2[0;]and2[0;2).Theeigenvaluesandeigenfunctionsof( 3{32 )arewellknown[ 38 ].Theeigenvaluesof( 3{32 )are~2l(l+1)wherel2NandtheeigenfunctionsarethesphercialharmonicfunctionsYlm(;),wherelml.[ 38 ]AgeneralizedeigenfunctioncouldbewrittenasthetheproductofageneralelementinL2(R+;r2dr),whichwewilldenotebyR(r)withthesphericalharmonicfunctionsYlm(;).UsingthisfactwecanproceedtothenextstepoftheMCP,whichistorewritetheauxiliaryHilbertspaceintermsofadirectintegraldecomposition wherecldonotestheclosureoftheset,andspanisthelinearspanofthesetofvectorsdenedinthebrackets.ThephysicalHilbertspaceandtheinducedinnerproductcomedirectlyfromselectingthesubspacethatcorrespondswiththel=0eigenvalue. 38 ]. 40


TheinnerproductofthisphysicalHilbertspaceisinheritedfromtheL2(R+;r2dr).ThuscompletesthequantizationofthissimpleconstraintmodelintheMasterConstraintProgram 35 ],[ 36 ],and[ 37 ]thattheprogramhashadinresolvingseveraloftheambiguitiesassociatedwiththeRenedAlgebraicQuantization,itstillmaynotbetheperfectchoicetouseforallconstraints.Iftheconstraint'sspectrumcontainsazerointhecontinuum,thenparticularcare,intheformofrathercumbersomemathematicalmachinery,mustbeused.Notthatthisyieldsanincorrectresult,howeveritalmostappearstobeextraneoustothematerial.Thisissomewhatofabiasedopinionbecauseaswewillseeinthenextchaptertheprojectionoperatorformalism'sanswertothisseemsmoresatisfactory.Againthough,weemphasizethefactthattheresultshavebeenshowntobeequivalenttotheresultsfoundin[ 13 ].Anotherpossiblecriticismofthisprogramisnotacriticismoftheprogram,butinstead,acriticismofitsimplemetation.Thatistosaythatauthorstendtodisplayaheavyrelianceontheclassicalanalysisofgroupstosolveconstraints[ 35 ].Themaincritiqueofthispointcomesfromthefactthatmostoftheworkdonewithgroupssuchassl(2;C)neglectthezerorepresentation[ 41 ],whichshouldbetherepresentationcorrespondingtophysicalHilbertspace.However,justastheauthorspointedoutin[ 35 ],thisparticularconstraintisnotphysicallyrealizable,andthereforenotsubjecttoexperiment. 35 ]foradiscourseonthattopic. 41




TheprimarygoalofthischapteristointroduceandmotivatetheProjectionOperatorFormalism(POF).Theprojectionoperatormethodisarelativelynewprocedurefordealingwithquantumconstraints[ 13 ][ 26 ].Thephilosophyofthisformalismistorstquantizetheentiretheory,andthenreducethequantumtheorybyusingtheconstraints.WewillattempttoillustratehowthePOFattemptstoremedysomeofthedecienciesofthemethodsdiscussedinChapter3.Inthenalsectionofthechapterwewillexaminethreeconstraintmodels.Therstisaconstraintthathasazerointhecontinuum,whereasthesecondandthethirdaremodelsthatwereexaminedin[ 8 ].Theyhelpillustratethepoweroftheprojectionoperatorformalismindealingwithallclassicationsofconstraints.Inthischapteritisunderstoodthat~=1unlessstatedelsewise. 13 ] WerequireEtobeHermitianwhichsatisestherelationE2=E(idempotent),thesearebasicpropertiesofaprojectionoperator.Moreprecisely,supposethatB1andB2denotemeasurable 33 ] 43


limn!1(E(B1)E(B2))n=E(B1\B2):(4{2) IfB1\B2=;,thenE(;)=0.WewillusethispropertyinChapters5,6and8. RevertingtotheDiracprescriptionofthephysicalHilbertspaceitisdenedas whereaisthequantumanaloguetotheclassicalconstrainta,anda2f1;:::;Ag.Inanidealsituation 4{3 )isequivalenttothefollowing Thefactthat( 4{4 )willnotalwaysleadtoanon-trivalresult,isaclueonhowtoarriveatthetrueanswer.Assumingthataaisself-adjointactingonaHilbertspace,wecanusethefollowingresultfromspectraltheorytoobtainourdesiredprojectionoperator,E.Namely,theoperatoraacanbewritteninthefollowingrepresentation[ 39 ] aa=Z10dE(4{5) wheredEistheso-calledprojectionvaluedmeasure[ 33 ]onthespectrumofaa,whichwasdenotedbywhichcontainedaspectralrangeof0to1 4{1 )canbeintroducedbasedontheresultof( 4{5 ) 44


40 ].Equation( 4{6 )projectsontoasubspaceoftheHilbertspacewithaspectralmeasureofaafrom0to(~)2.ThetruephysicalHilbertspace( 4{1 )isdeterminedwhenthelimitas(~)2!0willbetaken 4{6 .Wewillnowturnourattentiontofurthermotivatingtheprocessofsquaringtheconstraint. 34 ]onlymentionthatsecondorderwaschosenbecauseitisthesimplestexpression.Instead,weattempttooersomemathematicalargumentsthatindicatethatthesumofthesquaresofthequantummechanicalconstraintsissucient. 13 ]7 45


Inordertodeterminetheconstraintsubspaceinthephasespace,allAequationsmustbesatisedsimultaneously.Thissetofequationswouldbeatleastclassicallyequivalenttothefollowingset: Finally,ifweaddalloftheprecedingequationstogether,wearriveattheconclusionthat Aa2a=0(4{9) isequivalenttothesetofAequations( 4{7 ).Aswestatedbefore,thisisclassicallyequivalent,butarewecertainthatthiswillbejustiedquantummechanically? 46


40 ]. Theformalpathintegralformoftheprojectionoperatoristhefollowingexpression. whereTisthetime-orderedproductandR()istheformalmeasureoverthec-numberLagrangemultipliersf(t)g.Asshownin[ 40 ]theprojectionoperator( 4{10 ),isconstructedintwomainsteps.Thetimeintervalisdenedasapositiverealvalueequaltot2t1.TherststepistoconstructaGaussianmeasurethatwouldcausetheodd-momentsoftheLagrangemultiplierstovanish(i.e.Ra(t)D(t)=0),whilekeepingtheevenmoments(i.eRa(t)b(t0)D=2 0Mab,where(0)isasmallparametercorrespondingtoatimestep,Mabisapositivematrix,andisareal,positiveintegrationparameter.) whereNistheformalnormalizationofthepathintegral( 4{11 ).Thenalstepistointegrate( 4{11 )over.Toaccomplishthisfeatwewillintroduceaconditionally 33 ]. 47


wheretheconditionallyconvergentintegralisdenedbythefollowing[ 40 ];lim!0+Zeixsin[((~)2+)] Equation( 4{12 )istrueassumingthattheconstraintsarenotexplicitlydependentontime.WeexaminethatcaseinChapters7and8.ThematrixMabisthemostgeneralcase,butforourcurrentpurposes,wearefreetoselectMab=ab,whichwouldyieldthedesiredformofsquaringtheconstraints.Havingillustratedthemotivationbehindconsideringsquaringtheconstraints,wewilldivertourattentionstosomeofthemathematicaltoolsthatarerequiredinimplementingthePOF. 4.2.1CoherentStates [Qi;Pj]=iji1H:(4{13) TheWeyl(canonical)coherentstatemaybedenedas foranitenumberofdegreesoffreedom,(p;q)2R2N,andthestatesarestronglycontinuousinthelabels(p;q),j0iissomeducialvectoroftentakentobetheground 48


4{14 )trulyarecoherentstatesisthattheypossessaresolutionofunity[ 41 ]: Wewilloer( 4{15 )asanacceptedtruthwithoutproof[ 41 ]. Thesecoherentstatesalsooeraconnectiontotheclassicallimitofquantumoperators.Thispropertyisknownasthe\weakcorrespondenceprinciple".[ 41 ]Weexploit,andalsostatemorecarefully,thispropertyofcoherentstatesinasubsequentchapter. ThecoherentstatesareconvenientbecausetheyformanovercompletebasisoftheHilbertspace.Usingthisrepresentation,wecanexpressadensesetofvectorsinthefunctionalHilbertspaceintermsofthecoherentstateoverlap, wheren2C.Theinnerproductofsuchvectorscanbeexpressedasthefollowing, (;)=N;MXn;m=1nmhpn;qnjpm0;qm0i;(4{17) whereisanelementofthedenseset.ThecompletionofsuchasetofvectorsleadstotheunconstrainedHilbertspace,whichleadsustothetopicoftheReproducingKernelHilbertspace. 49


AsinthecaseoftheunconstrainedHilbertspace,wecanexpressadensesetofvectorsinthefunctionalconstraintsubspaceas Theinnerproductforthesevectorsisgivenby (;)P=NXm=1NXn=1nnK(pm;qm;pn;qn);(4{20) whereisalsoanelementofthedensesetofvectors.Usingbasicpropertiesofthereproducingkernelandcoherentstates,weknowthatthenormdenedbytheinnerproductofthesevectorswillbenite[ 13 ].Thisguaranteesthatthenormofvectorsinthecompletionwillalsobenite.IfwemultiplyareproducingkernelKbyaconstant,thereproducingkernelKstillcorrespondstothesamefunctionalspace.Thisisakeypointandonethatweexploitinthenextsectionwhenwedealwithaconstraintthatpossessesazerointhecontinuum. 50


4.3.1ConstraintwithaZerointheContinuousSpectrum 1=P2;(4{21) where1istheonlyconstraintpresent.Itisclearthatthespectrumof1isthereallineandthatitpossessesazerointhatcontinuum.Thisimpliesthatlim!0E0,whichisanunacceptableresult.Toresolvethisquandaryoneshouldlookattheprojectionoperatoroverlapwithasetofcoherentstatesoftheformof( 4{16 ),whichfollows UsingtheTaylorseriesbyexpanding( 4{22 )asafunctionof,utilizingtheresolutionofunityandfunctionalcoherentstatesoverlapwendthat, Assuggestedin[ 13 ],wewillmultiply( 4{22 )by1 2toextractthegermofthereproducingkernel.Wemustemphasizethatthisisstillthesamefunctionalspacedescribedby( 4{22 ).Thefunctionalformofthereproducingkernelisexpressedbythefollowing: 1 2hp00;q00jE(P2)jp0;q0i=Z+2+2dke(kp00)2=2+ik(q00q0)(kp0)2=2(4{24) Atthisjuncturewewillnotdiscusshowtoevaluatethisintegral,sincethetopicisdiscussedindepthinsubsequentchapters.However,wewillstatetheresult: lim!0K=e1 2((p002)2+(p02)2)+2i(q00q0)(4{25) Acharacteristicofthisreducedreproducingkernel( 4{25 )isthatitdoesnotdenethesamefunctionalspaceastheunreducedreproducingkernel( 4{22 ).Thisreproducing 51


4{25 )isgaussianpeakedattheclassicalsolutionp0=p00=p=2.AsstatedwewillreturntoseveralmoreexamplesofconstraintswithzerosinthecontinuumwhenwedelveintothistopicduringChapters5and6. wherebdenotestheLagrangemultiplierstoenforcetheconstraints.NotethattheHamiltonianhasbeenchosentobezeroforsimplicity,sowecanfocusdirectlyontheissuessurroundingtheconstraints. Fromthedenitionoftheji's,onecanimmediatelydeterminethePoissonalgebra,givenasusualby SincethisbracketyieldsaLiealgebra,oursystemisclearlyaclosedrst-classconstraintsystem[ 14 ]. Thequantizationofthismodelisstraightforwardandwepromotethedynamicalvariables(pj;qi)tothesetofirreducibleselfadjointoperators(Pj;Qi),whichobeythestandardHeisenbergrelation.Theconstraintjiarepromotedtoself-adjointoperatorsJi Theprojectionoperatoroftheseconstraints( 4{28 )takestheformof 52


2~z(~Qi~P)1 2~z(~Q+i~P)j0i(4{30) where~z(~q+i~p)=p 41 ]. Acoherentstatepathintegralcanalsobeusedtocalculatethematrixelementsoftheprojectorasshowninpreviousworks[ 13 ].Letusbeginwithapreliminaryequation,namely, Following[ 13 ],wecouldintegrateover~withrespecttoasuitablemeasureR(~)tocreatethedesiredprojectionoperator.However,itisequivalentandsimplertoproceedasfollows,KJ(~z00;~z0)h~z00jE(J2~2)j~z0i=Zh~z00je(i=~)~~Jj~z0gd(); 53


=N00N0[1+~z002~z02 (4{32) (4{33) From( 4{32 )wecandeducethatthephysicalHilbertspaceforeveryevenparticlesectorisone-dimensional.TheHilbertspacefoundusingthismethodisunitarilyequivalenttotheonedeterminedbytheMasterConstraintProgram. isclassicallyrst-class;however,uponquantizationtheconstraintstransmutetopartiallysecondclass.[ 8 ]Theanalysisofthemodelweareabouttoexamineservedastheprimarymotivationbehind[ 8 ].ThismodelisalsoatypeofconstraintthatthatweconsideredinChapter3,intermsoftheFadeev-Popovprocedure.Theactionforourchoiceofthemodiedsystemisverysimilarinform,i.e., 54


foralli,andchooseforfurtherstudytheparticularexampleforwhich Thesymbol~~isaxedconstantequalinvaluetothephysicalvalueofPlanck'sconstant~,namely1:061027erg-sec.Whentheclassicallimitiscalledfor,andthusPlanck'sconstant~!0,weemphasizethat~~retainsitsoriginalnumericalvalue.Thereasonforsuchasmalldivisoristoemphasizethequantumcorrections;dierentdivisorscanbeconsideredbyrescalingand.Werecognize,inthissimplecase,thatwecouldabsorbthefactorfbyaredenitionoftheLagrangemultipliersin( 4{37 ).Inmorecomplicatedsystems(e.g.,gravity)thissimplicationiseitherextremelydicultorperhapsevenimpossible.Therefore,asafurtheranalog,weretainfasapartofla.Astraightforwardanalysisleadstofli;ljg=ffji;fjjg=f2fji;jjg+ffji;fgjj+fff;jjgji+ff;fgjijj=fijklk+fji;fglj+ff;jjgli=fijklk+iab[qa@f=@qb+pb@f=@pa]ljjab[qa@f=@qb+pb@f=@pa]li: 55


where++=1.Indealingwiththequantumtheory,wedropthedistinctionbetween~~and~.Aquickcalculationshowsthatanyotherfactor-orderingofthedenitionof( 4{40 )willyieldanequivalentresult.ThecommutationoftheLiyieldsasurprisingresult,namely [Li;Lk]=ijk(i~FLk~2(k1l(ayka1+ay1ak)))2iia1Ji((QaQ1+PaP1)F+::: whereF=+(=~)(P21+Q21)+(=~)(P22+Q22).Thesecond,third,fourth,etc.termsin( 4{41 )representtheanomalyinthequantumtheory.Thisanomalycorrespondstoatransmutationtoapartiallysecondclasssystem.Withthisbeingnotedwewillcontinuethequantumanalysisofthesystem. Letusintroduceconventionalannihilationandcreationoperatorsrepresentedby Ifwedene asthetotalnumberoperator,itisevidentthat [Jj;N]=0;[Lj;N]=0; 56


whereasusualj0i(=j0;0;0i)denotesthenoparticlestateforwhichajj0i=0forallj. Therstnon-emptysubspacethatproducesanon-trivalresultisthe2-particlesubspace.Withtheadditionalsimplicationthat=2,wecanexpresstheeigenevectorthatcorrespondstoleasteigenvalueinthis6-dimensionalsubspaceas, whered=12+O(2)andd0=1+O(2).Theprojectionoperatorofthissubspaceisconstructedasthefollowing; 57


=N00N0[1+(dz1002+d0z2002+z3002)(dz102+d0z202+z302) 2!(d2+d02+1)+:::]: Onenalnoteregardingthisparticularmodel,aswiththecaseforotherpartiallyorfullyquantummechanicalsecondclassconstraintsystems,thelimitas!0isnottaken.TheHilbertspaceisdeterminedbythespacecorrespondingtotheleasteigenvalue. 58


Theprimarygoalofthischapteristointroduceregularityconditionsonconstraints,aswellaspresenta\new"classicationofconstraintscalledhighlyirregularconstraintsandalsoillustratetechniquesusedtodealwithquantumversionsoftheseconstraints.Thebasisofthischaptercomesfrom[ 9 ]and[ 10 ]. 16 ] Rank@a wheren2f1;:::;Mg,2Misthedimensionalityofphasespace,andistheconstrainthypersurface(a=0).Ifthisconditionfails,thentheconstraint(orsetofconstraints)iscalledirregular[ 16 ].Irregularconstraintscanappearinfollowingform whereaisaregularconstraintandrisanexponentr>1.Intheliterature[ 16 ]themeasureofirregularityisbasedontheorderofthezeroontheconstraintsurface.Forexample,( 5{2 )isanrthorderirregularconstraint.Weshouldnotethatwhiletheconstraintsaandraareequivalent(i.e.theconstraintsgeneratethesameconstrainthypersurface),thedynamicsandsetofobservablesassociatedwitheachgivensystemarenotnecessarilyequivalent. Thetermhighlyirregularconstraintreferstoaconstraintfunctionthatinvolvesbothregularandirregularconstraintsortwoormoreconstraintsofvaryingorder[ 9 ].For 59


Therstconstraintisregularatq=0andirregularatq=1oforder2.Thesecondcoinstraintisirregularatbothq=3,oforder2,andatq=4,oforder3.Bothoftheseconstraintsarerepresentativeoftheclassofhighlyirregularconstraints.Sincethedynamicsaswellasobservability[ 9 ]ofagivensystemarepotentiallynotthesameforregularandirregularconstraints,carefulconsiderationmustbeobservedwhenquantizingsuchsystems.Theprojectionoperatorformalism[ 13 ]seemstoprovideanappropriateframeworktodealwithsystemswithirregularconstraints[ 9 ]. Theusualformoftheprojectionoperatorisgivenby wherea2aisthesumofthesquaresoftheconstraintoperatorsand(~)isasmallregularizationfactor.TheprojectionoperatoristhenusedtoextractasubspaceoftheunconstrainedHilbertspace,H.If2ahasadiscreteisolated0thencanbechosentobeanextremelysmallnumber.However,if2ahasa0inthecontinuum,wecannotchooseanappropriatetoselectthepropersubspace.Wewilldiscussthisdistinctpossibilityshortly.Inthelimitas!0ifappropriate,thissubspacebecomesthePhysicalHilbertspace, lim!0EjijiPhys; lim!0EHHPhys: 60


13 ],whichisunacceptable.Toovercomethisobstacle,thislimitmustbeevaluatedasarescaledformlimit.Toaccomplishthis,wewillneedtointroducesuitablebrasandketsintheunconstrainedHilbertspace.Forthisdiscussionitwillbeconvenienttochoosecanonicalcoherentstates(jp;qi)tofulllthischoice.Weregardthefollowingexpressionastherescaledform whereS()istheappropriatecoecientneededtoextracttheleadingcontributionofhp0;q0jEjp;qi,for0<1.Forexample,ifhp0;q0jEjp;qi/toleadingorder,thenS()/1,forsmall.Thelimit!0cannowbetakeninasuitable 5{8 )isafunctionofpositivesemi-denitetypeandthismeansthatitmeetsthefollowingcriteria lim!0Nj;l=1jlS()hpj;qjjEjpl;qli0;(5{9) forallniteN,arbitrarycomplexnumbersflgandcoherentstatelabelsfpl;qlg.Aconsequenceofthepreviousstatementisthat( 5{8 )canleadtoareducedreproducingkernelforthephysicalHilbertspace ThereproducingkernelcompletelydenesthephysicalHilbertspace[ 13 ].Thereproducingkernelmakesitpossibletoexpressadensesetofvectorsinthefunctional 61


Theinnerproductforthesevectorsisgivenby (;)P=N;MXn;m=1nmK(pn;qn;pm;qm);(5{12) whereisalsoanelementofthedensesetofvectors.ThecompletionofthesevectorsinthesenseofCauchysequenceswiththerelevantinnerproductwillyieldthephysicalHilbertspace. Withoutexplicitlycalculatingthereproducingkernel,wewillconsiderthefollowinghighlyirregularquantumconstraint =Q2(1Q);(5{13) whereQactsasamultiplicationoperator.ClearlythisconstraintvanisheswhenQ=0andQ=1.Assuming,0<1,theprojectionoperatorforthisconstraintcanbewritteninthefollowingform Sincethezerosofthisoperatorfallinthecontinuum,itisclearfromthepreviousdiscussionwecannottakethelimit!0initspresentnakedform.Thereproducingkernelcanbeexpressedasthefollowing ByconstructiontheseprojectionoperatorsE(

whereKQ=0andKQ=1areleadingordercontributionstothereproducingkernelsaroundthetwosolutionstotheconstraintequation.See[ 9 ]forfurtherdetails.Unlikeexpression( 5{8 )theredoesnotexistasingleS()toextracttheleadingorderdependencyfortheentireHilbertspace.Toaddressthisdicultywewillconsiderthefollowingargument[ 9 ]. OurpreviousexamplehadconstraintsolutionsaroundQ=0andQ=1,wewillnowaddressthisinamoregeneralsetting.Webeginbydeterminingthereproducingkernelforeachsolutionintheconstraintequation.RecallthatthesumofreproducingkernelswillproduceadirectsumofthecorrespondingreproducingkernelHilbertspacesifthespacesaremutuallyorthogonal.Thiswillbethecaseforhighlyirregularconstraints.SoletKrepresentthe(>0)reproducingkernelforthereproducingkernelHilbertspaceH K=Nn=1Kn;(5{17) whereKnisthedeterminedreproducingkernelforeachuniquesolutionoftheconstraint.TheHilbertspacegeneratedhasthefollowingform, whereHncorrespondstotheKnforeachn.However,wehavenottakenthelimitas!0,andsincetheleadingorderdependencyispotentiallydierentforeachreproducingkernelKn,theredoesnotexistasingleS()thatcanbeusedtoextracttheleadingordercontributionofeachreproducingkernel.Toaccomplishthistaskwedenea(similarity)transformationS,


^K=Nn=1Sn()Kn:(5{19) Therescaled^KservesasthereproducingkernelfortheHilbertspace^H.AlthoughtheinnerproductofHand^Haredierentthesetoffunctionsareidentical.Thegoalofthislittleexerciseisofcoursetotakeasuitablelimit!0toyieldafunctionthatcanserveasareproducingkernelforthephysicalHilbertspace.Atthispoint,wecantakesuchalimit. ~Klim!0^K; where~KisthereducedreproducingkernelforthephysicalHilbertspaceHphys.Havingdiscussedthebasictheorybehindthisclassicationofconstraint,inthenextsectionwewillconsiderasimplebutrobusttoymodelthatdemonstratesthestrengthoftheProjecionOperatorFormalismtodealwiththesekindsofconstraints. whereisaLagrangemultiplierdesignedtoenforcethesingleconstraint 64


_q=0; _p=R0(q); withsolutions whereqlisarootofR(q)=0.IfR0(ql)=0thenthesolutionbecomes Thefunction(t)isnotxedbytheequationsofmotion,whichisnormalforrst-classconstrainedsystems.Toexplicitlyexhibitasolutiontotheclassicalequationsofmotionitisgenerallynecessarytospecifythefunction(t),andthisconstitutesachoiceofgauge.Gaugedependentquantitiesaredenedtobeunobservable,whilegaugeindependentquantitiesaredeclaredtobeobservable.Inthepresentexample,ifR0(ql)6=0,thenp(t)isgaugedependent,whileifR0(ql)=0,p(t)is,infact,gaugeindependent.Thisbehaviorsuggeststhatthemomentumpinthesubsetofthereducedclassicalphasespaceforwhichfq:R(q)=0;R0(q)6=0gisunobservable,whilethemomentumpinthesubsetofthereducedclassicalphasespaceforwhichfq:R(q)=0;R0(q)=0gisobservable.Wediscussthispointfurtherbelow.ThereducedclassicalphasespaceisgivenbyRZ,where Clearly,fortheclassicaltheorytobewelldened,itissucientforR(q)2C1,namelythatR(q)andR0(q)arebothcontinuous.(Strictlyspeakingthiscontinuityis 65


OurdiscussionwillcoverawideclassofRfunctions,andforconvenienceofexplanationweshallfocusononespecicexample;generalizationtootherexamplesisimmediate.Theexamplewehaveinmindisgivenby where Forthisexample,thezerosetisgivenby onlyforq=0isR0(q)6=0.(AlthoughphysicallymotivatedmodelswouldtypicallynotincludeintervalsinthezerosetofR,wedosotoillustratetheversatilityofourapproach.) Insummary,thephasespacefortheunconstrainedclassicalsystemisparameterizedby(p;q)2RR,andthephasespacefortheconstrainedsystemisparameterizedbythepoints(p;q)2RZ.Thislatterspaceconsistsofseveralone-dimensionallinesandatwo-dimensionalstrip.FromthestandpointofthiselementaryexampleallelementsofRZareequallysignicant. Wenowturntothequantizationofthiselementaryexamplefollowingthepreceptsoftheprojectionoperatorformalism[ 13 ].Inthisapproachonequantizesrstandreducessecond.TheultimatereductionleadstoaphysicalHilbertspaceappropriatetotheconstrainedsystem. 66


[Q;P]=i1(5{34) inunitswhere~=1.(Whenweeventuallyexaminetheclassicallimit,weshallrestoretheparameter~tovariousexpressionsasneeded.)Theprojectionoperatorofinterestisgivenby where>0isatemporaryregularizationparameterthatwilleventuallybesenttozeroinasuitablemanner.Sincethelimit!0willultimatelybetakenasaformlimit,weneedtointroducesuitablebrasandketsinthisoriginal,unconstrainedHilbertspace.Forthatpurposewewillagainchoosecanonicalcoherentstatesdened,forthepresentdiscussion,by Asusual,wechoosej0itosatisfy(Q+iP)j0i=0;namely,j0iisthenormalizedgroundstateofanharmonicoscillatorwithunitfrequency.Thusweareledtoconsiderthecomplexfunction whichiscontinuous(actuallyC1)inthecoherentstatelabelsanduniformlyboundedbyunitysinceE=Ey=E21. Itisimportanttoremarkthatthefunction( 5{37 )isafunctionofpositivetype,acriterionthatmeans Nj;k=1jkhpj;qjjEjpk;qki0(5{38) forallN<1andarbitrarycomplexnumbersfjgandlabelsetsfpj;qjg;thispropertyholdsbecauseEisaprojectionoperator.Asaconsequenceofbeingacontinuousfunction 67


servesasareproducingkernelforareproducingkernelHilbertspace,afunctionalrepresentationbycontinuousfunctionsontheoriginalphasespace(RR),oftheregularized(by>0)physicalHilbertspace.Ourgoalistotakeasuitablelimit!0soastoyieldafunctionthatcanserveasareproducingkernelforthetruephysicalHilbertspaceforthepresentproblem. Clearlythelimit!0ofthegivenexpressionvanishesandthatisanunacceptableresult.Supposeweassume0<1,e.g.,=101000.Thenitisclear(evenforamuchlargeraswell!),fortheexampleathand,that whereEn;1n4,correspondstothetermsinthelineaboveinorder.Byconstruction,forverysmall,itfollowsthattheseprojectionoperatorsobey i.e.,theyprojectontomutuallyorthogonalsubspaces.Inlikemannerthereproducingkerneldecomposesinto where 68


SinceEnEm=nmEnitfollows,fromthecompletenessofthecoherentstates,that ThisequationimpliesthattheHn,1n4,form4mutuallydisjoint(sub)HilbertspaceswithinL2(R2).Forthepresentexamplewith>0,eachHnisinnitedimensional. LetusrstconsiderK2(p00;q00;p0;q0)hp00;q00jE(<8(2Q)3=2<)jp0;q0i=1 ThisfunctionisalreadyofpositivetypeandiscorrecttoO(2)[i.e.,toO(104000)!]. Asdiscussedfrequentlybefore[ 13 ],wecanextractthe\germ"fromthisreproducingkernelbyrstscalingitbyafactorofO(2=3),sayby=(22),priortotakingthelimit!0.Consequently,werstdeneanewreproducingkernel ^K2(p00;q00;p0;q0)=p WeremarkthatthespaceoffunctionsthatmakeupthereproducingkernelHilbertspace^H2(generatedby^K2)isidenticaltothespaceoffunctionsthatmakeupthereproducingkernelHilbertspaceH2(generatedbyK2).Next,wetakethelimitas!0ofthe 69


~K2(p00;q00;p0;q0)lim!0^K2(p00;q00;p0;q0)=e[(2q00)2+(2q0)2]=2i(p00p): Thisprocedureleadstoanewfunction~K2,which,provideditisstillcontinuous{whichitis{leadstoareducedreproducingkernelandtherebyalsotoanewreproducingkernelHilbertspace~H2.Generally,thedimensionalityofthespaceaswellasthedenitionoftheinnerproductaredierentforthenewreproducingkernelHilbertspace;however,onealwayshasthestandardinnerproductdenitionthatisappropriateforanyreproducingkernelHilbertspace[ 42 ].Inthepresentcase,itfollowsthat~K2denesaone-dimensionalHilbertspace~H2.NotethateventhoughthecoordinatevaluefortheconstrainedcoordinateQisnowsetatQ=2{asisclearfromthespecialdependenceof~K2(p00;q00;p0;q0)onp00andp0{therangeofthevaluesq00andq0isstillthewholerealline.Theonlyremnantthatq00andq0haveoftheirphysicalsignicanceisthat~K2(p00;q00;p0;q0)peaksatq00=q0=2.Itisnoteworthythatanexampleofthistypeofirregularconstraintwasconsideredpreviouslyby[ 40 ]. Asimilarprocedureiscarriedoutfortheremainingcomponentsintheoriginalreproducingkernel.LetusnextconsiderK1(p00;q00;p0;q0)=hp00;q00jE(<2Q<)jp0;q0i=1 whichisafunctionofpositivetype.Itisnoteworthytonotethatthisconstraintisofregulartype.[ 14 ] 70


^K1(p00;q00;p0;q0)p andthentakethelimit!0leadingto ~K1(p00;q00;p0;q0)lim!0^K1(p00;q00;p0;q0)=e[q002+q02]=2; acontinuousfunctionofpositivetypethatcharacterizestheone-dimensionalHilbertspace~H1. Ourprocedureofscalingtheseparatepartsoftheoriginalreproducingkernelbyqualitativelydierentfactors(i.e.,1and2)hasnotappearedpreviouslyintheprojectionoperatorformalism.Thisdierenceinscalingismotivatedbythegoalofhavingeachandeveryelementofthereducedclassicalphasespacerepresentedonanequalbasisinthequantumtheory.ItisonlybythisprocedurethatwecanhopethattheclassicallimitofexpressionsassociatedwiththephysicalHilbertspacecanfaithfullyrecoverthephysicsintheclassicalconstrainedphasespace.Scalingof~K1and~K2bynitelydierentfactorshasbeenaddressedin[ 9 ]. LetuscontinuetoexaminetheremainingKn,3n4.ForK4wehaveK4(p00;q00;p0;q0)=hp00;q00jE(<8(Q3)3<)jp0;q0i=1 ~K3(p00;q00;p0;q0)=e[(3q00)2+(3q0)2]=22i(p00p0)(5{52) 71


Inthiscase,noappearsandnoinniterescalingisneeded,sowemaysimplychoose ~K3(p00;q00;p0;q0)=K3(p00;q00;p0;q0):(5{54) Althoughwedonothaveanexplicitanalyticexpressionfor~K3,wedohaveawell-denedintegralrepresentationin( 5{53 ).Furthermore,itfollowsthat~H3isinnitedimensional. Finally,wedenethereproducingkernelforthephysicalHilbertspaceas ~K(p00;q00;p0;q0)4n=1~Kn(p00;q00;p0;q0):(5{55) Inturn,thephysicalHilbertspaceHPisdenedasthereproducingkernelHilbertspace~Huniquelydeterminedbythereproducingkernel~K(p00;q00;p0;q0). Observe,byourprocedure,allelementsofthereducedclassicalphasespace(RZ)arerepresentedonanequivalentbasisin~K.ThisfeaturehasbeendesignedsothattheclassicallimitoftheexpressionswithinHPcorrespondtoallaspectsofthereducedclassicalphasespace.Wewillnowturntoadiscussionofobservablesofthismodel. 72


[E;O]=0:(5{56) WenotethateventhoughEisafunctionofthesquaresoftheconstraintswearenotboundtothesameconcernsThiemanfacedintheMasterConstraintprogram.Thestartingpointsofthediscussionofobservablesinthetwoframeworksaredierent.TheMasterConstraintProgramsuersfromthefactthatmultiplicationoffunctionsinaclassicalspaceiscommutative,hencetheadditionalrequirementofanobservablefunction.However,intheProjectionOperatorFormalismwebegininthequantumregimewherethemultiplicationofself-adjointoperatorsmaynotbecommutativetherefore( 5{56 )issucient.WecantakeageneraloperatorG(P;Q)intheunconstraintedHilbertspaceanddene asitsobservablecomponentsinceclearly[E;GE(P;Q)]=0.Infact,everyobservablecanbeexpressedintheprecedingform( 5{57 ).Theequation( 5{56 )isvalidfor>0.Howerveraslongas>0wehaveyettocapturethetruephysicalHilbertspaceofagiventheory.Thereforethelimit!0mustbetakeninasuitablefashiontodiscussobservables.Ifquantumconstraint lim!0[E;O]=0![;O]jiPhys=0:(5{58) 73


whereoistheclassicalanalogofO.Weconsider( 5{59 )tobeaweakequationbecauseitneedstovanishontheconstrainthypersurface.Itisobviousthatif( 5{58 )istruethenOisgaugeindependentinthephysicalHilbertspace.IntheHeisenbergpicturetheevolutionoftheoperatorisgivenby _OjiPhys=i whereHEistheobservablepartoftheHamiltonianintheformof( 5{57 ).ThereforeallobservableswillstayinthephysicalHilbertspaceastheyevolvewithtime.Thesametypestatementcanalsobesaidintheclassicalworld.However,inourparticularmodel,thelimit!0mustbetakenasaformlimitbecausehasazerointhecontinuousspectrum.Observablesintheseinstancesmustbehandledatthelevelofthereproducingkernel.RecallfromtheprevioussectionthatthephysicalHilbertspaceisisomorphictoaninnitedirectsumofcomplexnumbers.Inthisrealizationtheprojectionoperatoristheunitoperator,andthereforetheobservablescorrespondtogeneralsymmetricmatrices. Wewillnowdirectourattentiontoacalcuationofthecoherentstatematrixelementofthephysicalconjugatemomentumatthelevelofthereproducingkernel.Specically,werstnotethathp00;q00jPEjp0;q0ihp00;q00jEPEjp0;qi=ZdxZdx0hp00;q00jEjxihxjPjx0ihx0jEjp0;q0i=i~ZdxZdx0hp00;q00jEjxi0(xx0)hx0jEjp0;q0i=i~Zdxhp00;q00jEjxid dxhxjEjp0;q0i:


Similarly,itfollowsthat UsingasimilartechniqueofAraki[ 43 ],wenowdeterminethedesiredmatrixelementsbyadding( 5{61 )and( 5{62 ),anddividingbytwoleadsto, 2(( 561 )+( 562 )) (5{63) =1 2nZIndxhp00;q00jEjjxi[p00+p0+i(q00q0)]hxjEjp0;q0i =[p00+p0+i(q00q0)] 2~K(p00;q00;p0;q0): Finally,ifwesochoose,weallowonlythegaugeindependentmatrixelementsbyhandselectingtheportionsofthereproducingkernelthatcorrespondtotheirregularconstraints, 2~K0(p00;q00;p0;q0)(5{66) where~K0isthereducedreproducingkernelexceptforthecomponentcorrespondingto~K1.Inchapter6,wewillfurtherdiscusstheconceptofobservablesandtheclassicallimitofquantummechanicalobservablesintheAshtekar-Horowitz-Boulwaremodel. 75


TheprimarymotivationofthischapteristoanalyzetheAshtekar-Horowitz-BoulwaremodelutilizingtheProjectionOperatorFormalism.WewillalsocomparetheresultobtainedbythePOFapproachwiththatobtainedviamethodsoftheRenedAlgebraicQuantizationprogram.Thebasisofthischaptercanbefoundin[ 10 ]. 11 ]wasformulatedtomimicaparticularpropertyoftheHamiltonianconstraintofGeneralRelativity.InthissimplemodeltheconstraintoftheHamiltoniansystemwassuchthattheclassicalconstraintsubspacedidnotprojectdowntoallofthecongurationspace.UsingthemethodsdescribedbyDirac[ 7 ],theconstraintofthissimplequantummechanicalsystemwasimposed.Itwasarguedthatbyrequiringtheadditionalconditionofnormalizationoftheconstraintsolutions,thereisquantummechanicaltunnelingintoclassicallyforbiddenregions.Thismodelwasoriginallyformulatedwiththecongurationspaceofasphere[ 11 ]. Later,Boulwaremodiedtheconstraintproblembynotingthecurvatureofthecongurationspaceplaysnoroleintheanalysisandalteredthecongurationspacetoatorus-acompactyetgloballyatcongurationspace[ 44 ].Inthequantizationofthemodiedmodel,theadditionalrequirementoftheself-adjointpropertywasimposedonthecanonicalmomentum.Usingthisadditionalcriterion,itwasshownthatnotunnelingwouldoccurintotheclassicallyforbiddenregionsforthephysicalstates. Recently,LoukoandMolgudoinvestigatedthismodelusingtechniquesoftherenedalgebraicquantizationprogram(RAQ)todetermineitsphysicalHilbertspacestructure[ 32 ].Themethodstheyemployedledtotheexistenceofsuper-selectionsectorsinthephysicalHilbertspace.ThebasicformalismofRAQisunabletodetermineariggingmapforaconstraintthathasbothregularandirregularsolutions.Modicationsweremadeindenitionoftheriggingmaptoaccommodateforthevarietyofsolutions,(i.e.rsolutionsof 76


Usingtheprojectionoperatorformalism[ 13 ],weareabletoascertainthephysicalHilbertspaceoftheAshtekar-Horowitz-Boulware(AHB)modelwithtechniqueswhichwefeelareclosertotheessenceoftheDiracprocedure[ 7 ]thanthosein[ 32 ].ThephysicalHilbertspaceofthismodelisshownnottodecomposeintosuper-selectionsectors.Weareinclinedtotakethepointofviewthatsuper-selectionsectorsarebasedonphysicalprinciplesnotpuremathematics.Theapproachinwhichweultimatelyemployisasimilaritytransformation.Physicsisinvariantundersimilaritytransformations.Weshouldalsonotethetwomethods(ProjectionOperatorvs.RAQ)arenotequivalent.Wewereabletogeneralizetoaclassoffunctions(i.e.functionsthathaveintervalsolutionstoconstraintequation)thatthepreviouswork[ 32 ]cannotanalyzewithoutfurthermodications.Thepreviouswork[ 9 ]servesasaguideforthispresentendeavor. Thischapterisorganizedasfollows:Section2providesabriefintroductiontotheclassicalAHBmodel.Section3presentsthecanonicalquantizationofthemodel.Section4dealswithconstructingthephysicalHilbertspaceusingtheprojectionoperatorformalism.Section5dealswithdeningsuper-selectionsectorsanddeterminingwhetherornotthePhysicalHilbertspaceobtainedinSection5containssuper-selectionsectors.Section6dealswiththeclassicallimitoftheconstrainedquantumtheoryandestablishesthattheclassicallimitistheclassicaltheoryoftheoriginalmodel.Section7containsanaccountoftheRAQapproachtothismodel. 77


wherethefunctionR(y)2C1(S1)isassumedtobepositivesomewhere.WhentheconstraintequationissatisedtheclassicalsolutionsarelimitedtotheregionsofthecongurationspacewhereR(y)0.Theconstraintregioninthe4dimensionalphasespacewillinvolveapropersubsetofcongurationspace.NotethattheHamiltonianequalszerointhismodeltoemphasizetheroleoftheconstraint. Thedynamicsofthissystemaregivenbythefollowing5equationsofmotion. _x=2px;_y=0;_px=0;_py=dR(y) Ify0satisestheconstraintequationanddR(y) 78


6{1 ).WewillassumeourchosencanonicalcoordinatesareCartesianonessuitableforquantization[ 20 ].Wethenpromotethecanonicaldynamicalvariables(x;y;px;py)toasetofirreducibleself-adjointoperators(X;Y;Px;Py).Conjugatepairscorrespondingtocompact,periodicspatialcomponentswillnotobeythestandardHeisenberg-Weylrelationship[ 45 ]becausetheeigenvaluesoftheconjugatemomentumoperatorsarenotcontinuousbutdiscrete.Continuingwiththecanonicalquantizationprocedure,wepromotetheconstrainttoasuitablefunctionofself-adjointoperators Note,thereisnoorderingambiguityforthisoperator.Weassumetheconstraintoperatorisaself-adjointoperatorintheunconstrainedHilbertspace.Wecannowimplementthequantumconstraintusingtheprojectionoperatormethod. SincethefunctionR(y)isacontinuousfunction,wemustintroduceanappropriatesetofbrasandketstodealwiththesubtletiesdescribedinsection2. 79


CoherentstatesonacirclecanbegeneratedbycoherentstatesofalinewiththeuseoftheWeil-Berezin-Zak(WBZ)transformation[ 45 ].WeshalluseXandYtodenotethecharacteristiclengthsofthexandycoordinates,respectively.TheWBZtransform,T,isaunitarymapfromL2(R)toL2(S1S1),whereS1isthedualtoS1.Thetransformationisgivenbythefollowing (T)(x;k)n2ZeinXk(xnX)(6{5) where2L2(R),x2S1,andk2S1orstatedotherwisek2[0;2 2p(x+ip))exp(1 2(x+ipx0)2)(iX where1=exp(X2 (z)=n2Zn2e2inz;(6{7)jj<1,istheJacobithetafunction.Thesestatesarenotnormalized[ 45 ].Foreachvalueofkthesestatessatisfytheminimalaxiomsofgeneralizedcoherentstates;i.e.,acontinuouslabelingofthestateswherethelabelsethasatopologyisomorphictoR2andaresolutionofunity[ 40 ]. WecanexpressthecoherentstatesonT2asthefollowing, 80


45 ],eectively,wecansetthenewgroundstateatk.Therefore,wecansafelychoosezeroforbothkxandky.Thuswewillmakethefollowingnotationalchange Theconstructionofthereproducingkernelisbasedonpropertiesoftheconstraintoperatoraswellasthecoherentstates( 6{9 ).TheconstraintoperatorandthecompactnessofxrestrictthespectrumofitsconjugatemomentumPxandtherebyofR(Y).Allowedvaluesofyaredeterminedbythefollowingequation wherejni0istheorthonormalbasisforL2(S1).WewillproceedwiththequantizationofthismodelbyimplementingthemethoddiscussedinSection2foreachnsectorofthetheory.SincewearenotchoosingaparticularR(y),wewillonlybediscussingthephysicalHilbertspaceingeneral.Weconsiderthefollowingtwotypesofsolutionstotheconstraintequation. I.)(PointSolutions)Thesolutiony=ymisapointvaluesolutiontotheequation( 6{10 )foragivenvalueofn.Theindexmcorrespondstomultiplevaluesoftheythatsatisestheequationforagivenvalueofn. II.)(IntervalSolutions)Thesolutionsy=ym0satisfytheequation( 6{10 )forallelementsinanintervalI(m0).Thisclassicationofsolutionsalsoincludesacountableunionofdisjointintervals.Althoughphysicallymotivatedmodelsexcludesuchconstraintsolutions,weincludethemtoillustratetheversatilityofourapproach 81


9 ]todeterminethephysicalHilbertspacecontributionfortypeIIsolutions.Thecalculationofthereproducingkernelcanbedecomposedintoportionscorrespondingtoeachvaluen2Zinthefollowingmanner: 82


2(yy0)ei(pxxp0xx0)=2+p02=2+p2x=2xexp[(ymy)2=2iym(p0ypy)(ymy0)2=2](i(y0+ip0yymn;1)(i(y+ipyymn;1)exp[n2]exp[in((x0x)+i(p0xpx)]; where1=exp[22].FollowingtheprescriptionsetforthinSection2,weperformtherequiredsimilaritytransformationtoextracttheleadingdependencyofthereproducingkernel. ^Kmn=constant=Sm()2sin(1=Sm()(yy0)) 2(yy0)ei(pxxp0xx0)=2+p02x=2+p2x=2exp[(ymy)2=2iym(p0ypy)(ymy0)2=2](i(y0+ip0yymn;1)(i(y+ipyymn;1)exp[n2]exp[in((x0x)+i(p0xpx)]; Thelimit!0cannowbetakeninasuitablemannertodeterminethereducedreproducingkernelforthisportionofthephysicalHilbertspace[ 13 ]whichthenreads ~Kmn=constant=1 (6{14) foreachvalueofm.EachofthesereducedreproducingkernelHilbertspacesisisomorphictoaone-dimensionalHilbertspace(i.e.~HC).Wecontinuetheprocedureforeachwholenumbervalueofnuntilthemaximumallowedvalue(ofn)isreached. 83


6{10 ),thenwecanwritethereproducingkernelforthephysicalHilbertspaceinthefollowingmanner ~K=nmaxnm~Knm(6{15) Similarly,thephysicalHilbertspacecanbewrittenas Thesupportofthereproducingkernelisonlyintheclassicallyallowedregions.ThisimpliesthereisnotunnelingintoclassicallyforbiddenregionsasreportedbyBoulware[ 44 ]. whichistosaythatthephysicalHilbertspaceisgivenbythedirectsumofindividualHilbertspaces.ThephysicalHilbertspaceissaidtodecomposeintosuper-selectionsectorsifforanytwostatesj1i;j2ithatbelongtotwodierentsectorsHiandHj,respectively,andforanyobservableOinAobs,whereAobsthe*-algebraofallobservables,thefollowingholds, In( 6{18 ),Odenotesagenericself-adjointoperatorintheunreducedHilbertspace.InpreviousworksusingtheRAQprocedure[ 31 ],super-selectionsectorsarosebecauseeachsectorhadadierentdegreeofdivergence.SinceOisaself-adjointoperatorintheunreducedspace,( 6{18 )isforcedtovanishtoavoidacontradictionfromthevaryingdegreesofdivergency[ 46 ].Aswehaveshowninthepreceedingsectionthephysical 84


6{18 )willonlyholdifandonlyiftheoperatorisproportionaltotheprojectionoperator.Ingeneral( 6{18 ),doesnothold,thereforethephyicalHilbertspace( 6{17 )doesnotdecomposeintosuper-selectionsectors.WewillnowdiscusstheclassicallimitofthequantizedAHBmodel. wherejp;qiarecanonicalcoherentstates.ThisprovidestheconnectionbetweenanoperatorO(P;Q)andanassociatedfunctionontheclassicalphasespacemanifold.Inthelimit,~!0,wendthisfunctionreducestotheclassicalfunctionthatcorrespondstotheweakcorrespondenceofquantumoperator.ThisstatementcaneasilybeseenifOisapolynomial,however,thisconditionisnotnecessary.Thisresultcanbegeneralizedtoanynumberofphasespacevariablesaswillbedemonstratedbelow. Beforeevaluatingtheclassicallimitofthemodel,wemustdiscussthefundamentaldierencebetweenquantummechanicsonacompactcongurationspaceandthatofanunboundedspace.Theconjugatemomentumoperator(Px)hasadiscretespectrumifthecongurationspaceiscompact.Thereforethestandardcanonicalcommutationrelation [X;Px]=i~;(6{20) isinappropriate.Toalleviatethisproblemweconsiderthe\angle"operator[ 45 ] 85


Asobservedin[ 9 ],theobservablepartofanoperatorcanalwaysbeexpressedas whereOisaself-adjointoperatorintheunconstrainedHilbertspace. TheobservablepartoftheHermitiancombinationofUxandUyxis Byobservation,wenote Theseprojectionoperatorsareactingonmutuallyorthogonalsubspaces;therefore,theoperatorisidenticallyzero.Thisresultinformsusthatthisisagaugedependentquestionwhichisconsistentwiththeclassicalpicture.RecallfromSection3thexdynamicalvariableisgaugeindependentonlywhenpx=0.Quantummechanically,wehaveposedthequestiontonda\physical"wavefunctionthathassupportonbothagaugeindependentsectorandgaugedependentsector.Thisisimpossible. IfweweretoexaminethesamequeryforthecorrespondingHermitiancombinationofthe\angle"operatorfortheYcoordinate,wewouldobtaintheunitoperator.Theclassicallimitofthisoperatorisagainincompleteagreementwiththeclassicaltheory.Aswehavepreviouslyobservedtheclassicaldynamicalvariableyisalwaysgaugeindependent. NowweconsiderthefollowingquotienttoestablishtheclassicallimitoftheY\angleoperator"Uyhx;px;y;py;jEUyEjx;px;y;pyi hx;px;;y;py;jEjx;px;;y;py;i


( As~!0thisexpressionbecomes exp[i2y];(6{27) whereyissubjecttotheconditionR(y)=p2x.Whilethisexpressionisimaginary,wecanextractfromittheclassicalreducedphasespacecoordinatey. Nowwedirectourattentiontotheexpectationvalueofthephysicalconjugatemomentum,Px hx;px;y;pyjEjx;px;y;pyi @x0hx0;y0jx;px;y;pyi:(6{28) WeimplementtheconstraintsbyintegratingovertheappropriateintervalsasdescribedinSection5.Wecancontinuethiscalculationinasimilarmannertothatwhichisperformedin[ 9 ]. hx;px;y;pyjEjx;px;;y;py;i=px+0( 2( where 0(z;)=2i1Xn=nn2einze2inz:(6{30) As~approaches0,thesecondtermvanisheswhichcanbeseeninthedenitionoftheJacobithetafunction( 6{7 )[ 45 ],thusrecoveringthisaspectoftheclassicaltheoryfromitsquantumanalog.Usingthesametechnique,wecanalsocalculatetheclassicallimitoftheexpectationvalueofthePyoperator.Theprojectionoperatorformalismiswellsuitedtonotonlyproperlyimposequantumconstraints,butalsoallowonetoreturntotheproperclassicaltheoryinthelimit~!0. 87


6{2 ).Firstwemustassumetheconstraint( 6{2 )containsanitenumberofzerosandthatallstationarypoints(i.e.R0(y)=0,Rn(y)=0,[i.e.thenthdervativeofRwithrespecttoy])onlytohaveaniteorderthatnozerosof( 6{2 )aretobestationarypoints.AswiththeanalysisintheprecedingsectionswemustalsorequirethatR(y)bepositiveatleastsomewhere. FollowingtheprogramdescribedinChapter3wemustrstchooseanauxiliaryHilbertspace,Haux.TheauxiliaryHilbertspaceofchoiceistheHilbertspaceofsquare-integrablecomplexfunctionsoverthecongurationspace.Thecanonicalinner-productisgivenbythefollowing; (1;2)aux=ZZdxdy1(x;y)2(x;y);(6{31) where()denotescomplexconjugation.TheclassicalconstraintispromotedtoanoperatorthatactsontheauxiliaryHilbertspace, ^C=@2 whereR(Y)actsasamultiplicationoperatornamely,R(Y)(x;y)=R(y)(x;y)forall(x;y)2Haux.Theoperator,^Cisanessentiallyself-adjointoperatoronHaux,thereforetheoperatorwillexponentiatetotheoneparameterunitaryoperatorviaStone'stheorem, KeepinginlinewiththeRAQprogram,wemustnowchooseatestspaceHaux.Inthismodeltheconvenientchoiceisthesetoffunctionsoftheform; 88


6{33 )isasfollows, thereforebythiscalculationisinvariantundertheactionofU(t).Onefurthercommentmustbemadebeforeproceedingwiththerestoftheprocedure,ifO2AobsthenOcommuteswithU(t)andisdenselydenedin.ThenalphaseoftheRAQprocedureistodeterminetheanti-linearriggingmapviathegroupaveragemap, orequivalentlywecandiscussthemapthroughthematrixelements[ 30 ] AtthispointadeviationfromthestandardRAQapproachisrequired.[ 46 ]Since( 6{37 )isnotabsolutelyconvergent,thisisduetothefactthegaugegroupgeneratedbyU(t)isanon-compactgroup.Formally,itwasestablished[ 32 ]thattheriggingmapcouldbewrittenasthefollowingequation, orequivalently, whereymjsaresolutionsto 89


6{38 )and( 6{39 )arethedeltafunctionsforRandS1,respectively.Assumingthat( 6{40 )hassolutionsthenitwasshownin[ 32 ]that( 6{39 )doessatisfytheaxiomsoftheriggingmap.AkeycomponentofthevericationoftheaforementionedaxiomsisthatinducesarepresentationofAobsonthephysicalHilbertspace.Thiscanbestatedintermsofthematrixelements forall1;22andA2Aobs.Itcanbeshown[ 32 ]thattherepresentationofAobsonHRAQisirreducibleandistransitive. IntheprecedingdiscussionthephysicalHilbertspacedidnotdecomposeintosuper-selectionsectors.Theadventofsuper-selectionsectorsappearstobeadirectresultofrelaxingtheconditiontoallowforsolutionsof( 6{2 )toincludestationarypoints.Withthisrelaxedcondition,afurthermodicationoftheriggingmapisrequiredtoavoiddivergencesin( 6{39 ).Thisisaccomplishedbyreplacingthedenominatorwithfractionalpowersofhigherderivatives,whichdependsontheorderofthestationarypoint.Thereplacementofthedenominatorcanbethoughtofasarenormalizationoftheaveragingprocedure.Eachoftheserenormalizedriggingmapscanbeshown[ 32 ]tocarryatransitiverepresentationofAobs.ThetotalHilbertspaceHtotRAQcanberegardedasthedirectsumofindividualHilbertspaces.TherepresentationofAobsalsodecomposesintotherepresentationofthesummands.Whichinturnimpliesthepresenceofsuper-selectionsectorsinHtotRAQ. 6{39 )hasdenitiveconnectiontothereducedphasespacemethodforquantization.Theauthorsof[ 32 ]alsocommentonthecloseconnectionto 90




Timeisacrucialelementtoanydynamicalsystem;itistheevolutionparameterofsuchasystem.Thenatureoftimeisanextremelypopulartopiccoveredbymanyphysicists,aswellas,philosophers[ 22 ].Whilethephysical(ormeta-physical)natureoftimeisoutsidethemainfocusofthisdissertation,time-dependencyinquantummechanicsoersusaninterestingcaveattoexploreandstudy. Inthemethodologiesdevelopedanddiscussedinthepreviouschapters,theprimarygoalwastosolvequantummechanicaltime-independentconstraints.Theexclusionoftimewasmadeprimarilyoutofsimplication.Inmostoftheliteratureaboutconstraints[ 12 ]thetopicoftime-dependentconstraintsiseitherbrieycoveredoritisnotcoveredatall.However,itisclearthatforamorecompletediscussionofconstraintdynamicswemustalsoincludeconstraintsthatareexplicitlydependentontime.TimedependencecanenteradynamicalsystemthroughtheHamilitonian,constraints,orinthemostgeneralcaseorcombinationofthetwo.Theinclusionofexplicitlytime-dependentconstraintsoersnotonlyaninterestingacademicexercisebutalsogivesphysiciststhetoolsrequiredtoexaminemorephysicaltheoriesthanthosethatpreviouslycouldbediscussed. Thisisnottherstoccasiononwhichtheprojectionoperatorhasbeenusedtodealwiththecaseoftime-dependentconstraints.In[ 47 ],Klauderderivedanexpressionforevolutionoperatoroftime-dependentconstraint.Theconstructionofthisexpressionwasbasedonmodifyingtheexpressionforthetime-independentcase.Althoughthisexpressionseemedtobecorrecttheauthorchosenottopursuethissubjectmatterfurther.Primarily,hemadethischoicebecausetheformuladidnotreducetoasimpleroperatorexpression.Whiletheprojectionoperatorwillbetheprimarymodeofexploration 92


47 ]willnotbethestartingpointforourinvestigation. Wewill,however,advocatetheuseofthereparameterizationinvariantdescriptiontodiscusssystemswithtime-dependentconstraints.Weshouldmentionthatthisstartingpointisnotanewapproachtodealwithtime-dependentconstraints.Wewillalterpasteortsonthistopicbyexploringthe\non-local"pointofview.Thephrase\non-local"pointofviewwascoinedbyGitman[ 48 ]whendescribingaphysicalsysteminwhichoneassumesareparameterizationinvariantformofatheory.However,itiswellknown,thatifanactionisareparameterizationinvariantthentheHamiltonianvanishesontheconstraintsurface 48 ].Weshouldmentionthatusingthereparameterizationinvariantapproachisnotnew,however,theimplementationofthissymmetrywiththeprojectionoperatorwouldseemtobenew. Inthenexttwochapterswewilldiscusssomeofthefacetsoftheproblemwithtime-dependentconstraints.InChapter8wewillmotivateanddevelopthetechniquesinwhichonecanstudyconstraintswithanexplicittime-dependentfeature.Wewillalsogiveabriefintroductiontoanalternativetotheprojectionoperator,whichistheapproachusedbyGitman,[ 12 ]andcompareandcontrastthetwoapproaches.TheprimarygoalofChapter9istoimplementtheformalismdevelopedinChapter8,inafewexamplesoftime-dependentconstraints. 93


8.1.1BasicModel where_q=dq=dtandH(p;q)istheHamiltonianofthesystem.Theevolutionofthesystemisobtainedbyvaryingthefunctionalwithrespecttothedynamicalvariables,thisreadsas, _q=@H @p; _p=@H @q; subjecttothesuitableboundaryconditions.Asstatedabove,thissystemispurelydynamical,however,itiswellknownanyactioncanbeconvertedintoanequivalentactionthatisareparameterizationinvariant.[ 48 ]Letusbeginthisconversionbypromotingthedependentparameterttoadynamicalvariable.Thisisappealingfromarelativist'sperspectivebecausethespatialandtemporarycoordinatesaretreatedsymmetrically.Wealsomustintroducetheformalmomentumptconjugatetot.Theintegrationvariablein( 8{1 )isnowreplacedbyanewindependentparameter,whichcorrespondstopropertimeoramoregeneralfunctionoftime 6 ]However,forourpurposeswecanconsidertheLagrangemultiplierarenotstrictlyincreasingseeAppendix 94


where()denotesthederivativewithrespectto.ThepricepaidinpromotingttoadynamicalvariableisthattheHamiltonianvanishesweaklyintheextendedphasespace.WehaveidentiedtheprimaryunexpressablevelocitytastheLagrangemultiplierthatenforcesthe(rst-class)constraintpt+H(p;q)=0.Therefore,wehaveturnedatheorythatwasadynamicalsystemintoonethatispurelygauge.Eectively,wehaverecasttheoriginaltheoryinsuchamannerthatitcanberelatedinanytemporaryreferenceframe[ 48 ].Theequationsofmotionof( 8{4 )areasfollows: d=@H @p;dp d=@H @q d=();dpt (8{6) (8{7) Astheequationofmotionappearaboveq,momentump,andthephysicaltimet,measuredaregaugedependentquantities.However,ptisgaugeindependentandthereforeanobservablequantityinthistheory.Byidentifyingtheusualtimeasthegaugedependentquantity wecanquicklyreducetheprecedingequationsofmotion( 8{4 )tothefamiliarparameterizedform. _q=@H @p; _p=@H @q; Thedynamicsofthissystemarisesfromimposingtheconstraints. 95


wherej2f1;:::;Nganda2f1;:::;Ag.ThedynamicsthissystemisgivenbythefamiliarHamiltonian'sequations Atthispoint,theequationsofmotionareidenticaltothosethatappearinthetimeindependentcase.Thedistinctionappearswhenweforcethedynamicstolieontheconstraintsurface(i.e.subspaceinthephasespacedenedbya=0)foralltimet. where0implies( 8{15 )vanishesontheconstraintsurface.Forsimplicity,wewillassumethatthesetofconstraintsarecompletetoallorders(e.g.secondary,tertiary,etc.)ofconstraintshavebeenuncoveredusingtheDiracprocedure[ 14 ].ThedistinctionbetweenrstandsecondclassconstraintsismadebasedonthealgebraofthePoissonbrackets.However,weforgothisdistinctionforthemomentforthesakeofgenerality. Thecostofexplicittimedependenceintheconstraintsin( 8{14 )isthepresenceofthepartialtimederivativeinequation( 8{15 ).Despitetheadditionalterminequation( 8{15 ),wecanmaintaintheusualstructureoftime-independentconstraints,byfollowingtheproceduredescribedinthepreceedingsubsection.Asbeforewewillpromotettoadynamicalvariable,andintroduceitsformalconjugatemomentumpt.Byintroducingadditionaldynamicalvariablesandconjugatepairstothephase-space,wealsohave 96


=!+dt^dpt(8{16) where!isthesymplecticformoftheoriginalparameterizedspace([ 17 ]),denedinChapter2.ThePoissonbracketf;gwhichisdenedbythesymplecticform,shouldbeunderstoodunlessotherwisespeciedtobethatoftheextendedspace.Theequivalentactioncanbewritteninthefollowingmanner: where~a=()awhichismerelyaredenitionoftheLagrangemultiplier.NoticeonceagainthecanonicalHamiltonianvanishes.Asintheprevioussectionthedynamicsofthesystemarisesfromimplementingtheconstraints. (8{21) (8{22) 48 ]alongwithothermodelsproposedbyotherauthors[ 49 ],[ 50 ],wehavearrivedatastartingpositiontodealwithtime-dependentconstraints.Thiswasdonebychangingthedimensionoftheentireunconstrainedphase-spacefromR2NtoR2N+2,whichwasaccomplishedbypromotingttoadynamicalvariable,andintroducingitsconjugatemomentumpt.Theconsequenceof 97


12 ]withtheaidoftheextendedsymplecticform.Weshouldalsonotethatwehaverefrainedfromintroducingatemporalgaugexingtermsuchasachronologicalxinggaugeinouraction.Thisisapointofdivergencefromthepreviousauthorsonthesubject.Asiswellknown,agauge-xingtermhasthepotentialtointroducetopologicalobstructionsthatcancausedicultyintheanalysisofthequantumsystem.ThistechniqueofintroducingagaugeisusedinquantizationschemessuchasFaddeev-Popov[ 27 ]whichadvocatesreductionbeforequantization.Sinceoneofthemainphilosophiesoftheprojectionoperatorformalismistoquantizetheentiredynamicalspaceandreducesecond(i.e.eliminatetheredundantvariables),thereisnoneedtointroducesuchatermintheaction.Diracobservablesarephase-spacefunctionsthatcommuteweaklywithalloftheconstraints.AnextensiveamountofliteraturehasbeendevotedtothetaskofidentifyingobservablesinsystemssuchasGeneralRelativityandothergenerallycovariantsystems[ 51 ].Ifoisaclassicalobservableinasystemwithtime-dependentconstraintthenthefollowingmustbetrue: d=fo;pt+Hg+~afo;ag0(8{23) wheref;gareunderstoodtobethePoissonbracketsfortheextendedspace.Therefore,oisaconstantofmotionontheconstraintsurfaceintheextendedphase-space,whichimpliesthatanobservableisindependentofachoiceofreferenceframeorgauge.Sincewewillnotmakeanyfurtheruseoftheconceptofanobservableinthediscussionofsystemswithtime-dependentconstraints,wewilldeferthisdiscussiontoafutureproject.We 98


8.2.1GitmanandTyutinPrescriptionforTime-DependentSecond-ClassConstraints 12 ]Forsimplicity,wewilllimitthisdiscussiontoincludebosonicvariables,however,onecouldextendanyofthefollowingargumentstoincludefermionicdegreesoffreedomaswell.Also,forconveniencewewillusethenotationusedbytheoriginalauthors,namely,=(q;p)whichcanexplicitlydependontime,aswellas,f;gD()representstheDiracbracketwithrespecttoasetofsecond-classconstraintsa(;t).TheDiracbracketsaredenedinthefollowingmanner wheref;gisthePoissonbrackets,aisaconstraint,andCabisaninvertiblematrix[ 12 ].WheneverencounteredtheDiracbracketistakenassumedtobedenedfortheextendedspace,(;t;pt),asdescribedintheprecedingsection. ConsideraclassicalHamiltoniansystemwithasetofsecond-classconstraintsa(;t)andwithaHamiltonianH(p;q;t).TheDiracbrackets[ 7 ]areusedtoavoidhavingtosolvetheconstraints.Therefore,theevolutionofthecanonicalvariablesisgivenby dt=_=f;H+ptgD()a(;t)=0:(8{25) Thequantizationoftheclassicalsystem,followsintheSchrodingerpicture,inwhichthecanonicalvariablesareassignedtooperatorsSthatsatisfytheequi-timecommutationrelations; [S;0S]=i 99

PAGE 100

Atthisjuncturewerealizethiscurrentpictureisunabletoillustratethefulltimeevolutionofthesystem.Inordertofullyobtainthetimeevolutionwewillmovetoaunitarilyequivalentpicture,theHeisenbergpicture.IntheHeisenbergpictureofquantummechanics,thestatevectorsremainsxedwhiletheoperatorsevolveintime.[ 38 ]IntheHeisenbergpicturetheoperatorsHarerelatedtotheoperatorsSbyH=U1SU,whereUisthetimeevolutionoperator.TheoperatorUisrelatedtotheHamiltonianHSbythedierentialequation, @t=i WecanevaluatethetotaltimederivativeofHbythefollowing =U1(i Equation( 8{29 )establishestheconnectionbetweenthequantumequationsofmotionandtheclassicalequationsofmotionnamely, Inthemostgeneralsetting,theabovedescribedevolutionisnotconsidered\unitary",becauseingeneralno\Hamiltonian"existswhosecommutatorwouldresultinthetotalderivative.Theprincipalagentforthisnon-unitarycharacteristhesecondtermintheleft-handsideofequation( 8{29 ),whichistimevariationofS.Therefore,thedynamicsareevolving,aswellastheconstraintsurface. 100

PAGE 101

20 ].Followingtheconventionalprogramwe\promote"thephase-space(qj;pj;q0=t;p0=pt)coordinatestoirreducible,self-adjointoperators(Qj;Pj;Q0=T;P0=Pt).Thenon-vanishingcommutationbracketfollowsthestructureoftheclassicalextendedPoissonbracketi.e. [Q;P]=~ Apossibleobjectionthatthereadermayhaveistoquestiontheself-adjointnatureoftheToperator.IfthespectrumofTisequaltotheentirerealline,asexpected,thiswouldimplythatthespectrumofPtwouldalsobeunbounded.However,asiswellknown,ifweidentifyPtwiththeenergyEthenPtmustbeboundedfrombelow,whichwouldimplyTisnotaself-adjointoperator.[ 52 ].This,however,assumesthatPthasidentiedorforcedtobecomethenegativeHamiltonian,whichisaconstraint.Thecontradictionisavertedbecausewehavenotimposedtheconstraints,onlyquantizedtheentireclassicalsystem. Wefollowthebeliefthatabstractoperatorformulationofquantummechanicsisfundamental,aswellascorrect[ 53 ].Therefore,preceedingwiththecanonicalquantizationoftheclassicaltheory,bypromotingtheclassicalconstraintstoself-adjointfunctionsoftheirreducibleoperators. (8{32) Onepossibleobjectionatthisjunctureisthereexistmanywaysofquantizingagivenclassicalsystem.Whilethisiscertainlytrue,wewillassertthatwecanappealto 101

PAGE 102

ajiphys=0 (8{34) 0jiphys=0 (8{35) foralla+1constraints.However,thisprocedurewillonlyworkforaselecttypeofrst-classconstraints.Infact,ifoneadheresstrictlytotheDiracprocedure( 8{35 )willresultinatrivialsolutionsincetheconstraint0islinearinPtwhichimpliesthatitsspectrumwillcontainazerointhecontinuum,therebycausingthephysicalHilbertspacetobecomprisedofonlythe0element,whichisundesirableandunacceptable.Wewillthereforeappealtotheuseoftheprojectionoperatorformalism[ 13 ]tocircumventthesepossibledilemmas. 13 ]deviatesfromtheDiracmethodbyintroducingaprojectionoperatorE,whichtakesvectorsfromtheunconstrainedHilbertspace(H)totheconstraintsubspace(i.e.thephysicalHilbertspaceorevenbettertheregularizedphysicalHilbertspacewhichwillbedescribedshortly.) Thegeneralformoftheprojectionoperatoristhefollowing: 102

PAGE 103

8{37 ),impliesthattheoperatorprojectsontothespectralinterval[0;2(~)].Theprojectionoperatorformalismallowsustodealwithallconstraintssimultaneouslyandtoplacealltypesofconstraintsonequalfooting. 13 ],namely, (8{38) ToobtainfurtherinsightitwillbeconvenienttousethecanonicalcoherentstatesoftheunconstrainedHilbertspace. wherejiisanormalizedducialvectorinH.Thesecoherentstatesadmitaresolutionofunitygivenas wherethedomainofintegrationistheentireextendedphasespace.Theoverlapofthesevectorsaregivenbythefollowing: 4~[jp0pj2jq0qj2+i 4~[jp0tptj2jt0tj2+i Expression( 8{41 )denesapositivedenitefunctionalwhichcanbechosenasthereproducingkernelandusedtodeneareproducingkernelHilbertspaceH. 103

PAGE 104

13 ] wheref()issomefunctionchosentoinsure( 8{42 )convergesabsolutelyandTisthetimeorderedproduct. Wecangeneralizetheresultfrom[ 54 ]namelylim!0hp00;q00;p00t;t00jE(
PAGE 105

47 ].Thisofcourseisassumingthattheconstraintsarecontinuousint.Thisstatementisalsoapplicableifaaresecond-classconstraints. 12 ]startonverysimilargrounds,theapproachesendonverydierentgrounds.TheGitmanapproachadvocatestheuseofDiracBrackets,whichisamethodusedtoavoidsolvingforsecond-classsystems,whiletheProjectionOperatorFormalismtreatsallconstraintsonequaltheoreticalfooting.Intheproceedingchapterwewillexaminetwodierentconstraintmodelswiththeaidoftheprojectionoperatorformalism. 105

PAGE 106

Intheprecedingchapter,wehavedevelopedanapproachtocontendwithexplicittime-dependencyinconstraintswithintheprojectionoperatorformalism.Despitethisdevelopment,someloomingquestionspersist.Theprimarypurposeofthischapteristoelucidatetheseunresolvedquestionsbyconsideringsomesimplequantummechanicalmodels.OneofthemostpressingquestionsiswhetherornotthephysicalHilbertspaceofatime-dependentconstraintistrivial 8{46 ),the\evolutionoperator"fortime-dependentconstraintscanbewrittenasaninniteproductofprojectionoperators.However,aswewillillustrateinourrstmodel,evenwiththerequirementofastringentpolarizationofthestatesfromthetotalHilbertspace,thephysicalHilbertspaceisnon-trivial.Thesecondmodelisdesignedtodemonstratehowasecond-classsystemshouldbeconsideredwithinthiscontext. 2(p21+p22+p23)+1 2(q21+q22+q23)+(t)(j1sin( wherej1=q2p3q3p2andj3=p2q1q2p1and(t)istheLagrangemultiplierthatenforcesthesinglerst-classconstraint 106

PAGE 107

2(p21+p22+p23)+1 2(q21+q22+q23)thereforethistrulyisarst-classsystem.ForthecaseofthisanalysiswewillrestricttheallowablevaluesofttoacompactsubspaceofRnamely,[0;1].Itiseasytoobservethattheconstraintsurfaceinitiallyisdenedbyj3=0butevolvesinasmoothfashionintothevanishinglociofj1.Movingtotoquantumanalogofthissystem Ifwedenethenumberoperator itisobviousthat [J1;N]=0=[J3;N]:(9{6) Basedonthisconservation,wecanstudythefulllmentofthetime-dependentconstraintsineachofthenumber-operatorsubspacesindependentlyofoneanother.Basedonthisinformationwecanproceedwiththefollowinganalysis: 107

PAGE 108

PAGE 109

PAGE 110

49 ].Inthiswork,theauthorsextendtheBRST-BFVmethod[ 14 ],todealwithnon-stationarysystems(i.e.time-dependentsystems).Forthisdissertation,wechosenottodiscusstheBRST-BFVmethod,howeverforadescriptionofthemethodsee[ 14 ]and[ 49 ]. Themodelusedtoillustratetheauthor'stechniquewasatwo-dimensionalrotorwithatime-dependentradius.However,forthisdiscussionweabatethemodelin[ 49 ],byreducingthenumberofdegreesoffreedomfrom3to2,aswellas,settingtheHamiltonianequaltozerotoemphasizetheconstraints.Considerthefollowingtime-dependentclassicalconstraints: wherecisapositiveconstantandq,p,pt,arethecanonicalposition,itscorrespondingconjugatemomentumandconjugatemomentumcorrespondingtothetemporalcoordinate.BasedonthethePoissonbracketoftheconstraints,thisconstraintsystemisasecond-classsystem. Thecanonicalquantizationofthismodelisstraightforward.Wesimplyfollowthesameprocedureasstatedintheprecedingchapter,whichimpliesthatwepromoteallofthecanonicalcoordinates(p;q;pt;t)toirreducibleself-adjointoperators(P;Q;Pt;T).Wepromotetheconstraintstoself-adjointoperatorsasindicatedbythefollowing: 110

PAGE 111

UsingthelogicemployedbyKlauderin[ 13 ],thestatethatminimized( 9{19 )wouldbethefollowingstatejct;cij0;tiUsingthetheoryofWeyloperatorswecanconstructarepresentationofthedesiredprojectionE,givenby WeshouldobservethatinthiscaseRe(222)=4ddd 13 ]. ItiswellknownthataeldtheorythatisnotreparameterizationinvariantcanbetransmutedintoonereparameterizationinvariantbyasimilartechniquetothatemployedinChapter8[ 55 ].Namelythiscanbeaccomplishedbychangingthespace-timecoordinates(xwhere2f1:::;Ng,andNisthenumberofspacetimedimensions)tofunctionsoverspace-time,whichinessenceintroducesNscalareldstotheeldtheory, 111

PAGE 112


PAGE 113

Chapter5introducestheclassicationofconstraintscalled"highlyirregular"constraints.Duringthischapter,wedescribedageneralproceduretosolvethequantumanalogtothe"highlyirregular"constraintsutilizingtheProjectionOperatorFormalism.WealsosuccessfullyanalyzedasimpleexampleofthistypeofconstraintusingonlytheProjectionOperatorFormalism. InChapter6,weusedthemathematicaltoolsestablishedinChapter5togiveacompleteaccountofthequantizationoftheAshtekar-Horowitz-Boulwaremodel.[ 11 ]ThismodelwasinspiredbytheHamilitonianconstraintofGeneralRelativitytoanswerwhetherornottherecouldbequantummechanicaltunnelingintoclassicallyforbiddenregionsofphasespace.DuringthecourseofthischapterwecomparetheresultsobtainedbytheProjectionOperatorMethodwiththatoftheRenedAlgebraicQuantization 113

PAGE 114

Theremainingchaptersweredevotedtothetopicoftime-dependentquantumconstraints.WedevelopedtheformalisminwhichthetopiccanbeapproachedinthecontextoftheProjectionOperator.Thiswasaccomplishedbyextendingtheclassicalphasespaceofthetime-dependentsystem,therebyelevatingthetimeparametertoadynamicalvariable.Inthesamechapter,wecomparedtheProjectionOperatorFormalismtotheapproachthatwasrstdiscussedin[ 12 ].Whilethesemethodsstartfromthesamepoint(i.e.anextendedphasespace)theconclusionsreachedareverydierent.Inthefollowingchapterweweresucessfulinanalyzingtwoexamplesoftime-dependentconstraints. Thestoryofclassicalandquantumconstraintsthatwehavepresentedwithinthisdissertationisbynomeansacompleteaccount.Infactitisimpossibletogiveacompleteaccountofanyresearch.Byresearchingwemerelypointthedirectiontonewresearch,leadingtonewquestionstoaskandtoattempttoanswer.Thetopicsinphysicsarealwaysbiggerthantheindividualphysicist.However,thisisthebeautyofthesubject,thatthingsweleaveunresolvedcanbepickedupinthefuturegenerations. Thereareseveralunresolvedissuesleftfromthisdissertationthatcanbeaddressedbytheauthororfutureresearchers.Theseincludebutarenotlimitedto,\HowdothemethodsoftheProjectionOperatorgeneralizetoafullquantumeldtheory?",\WhatlessonslearnedfromthesimplemodelsthatweanalyzedinthisdissertaioncanbeappliedinmorerealistictheoriessuchasQuantumGravity?",\CanweusetheformalismobtainedinChapter8toexaminemorerealistictheories?" 114

PAGE 115


PAGE 116

WhataretheconsequencesifanactionI[q(t)]isinvariantunderaninnitesmaltemporaltransformation,whichisthecharacteristicofareparametrizationinvarianttheory,i.e. Giventheinnitesimaltransformation dt=_q ByourassumptionthatI[q(t)]=I[q(t+(t))]itfollowsthat (A{5) =Zt2t1(@L @tt+@L @qq+@L @_q_q)dt =Zt2t1(@L @tt+@L @q_q+@L @_qd dt_q)dt =Zt2t1(dL dt+@L @_q_qd dt)dt Integratingbyparts;=Zt2t1(Ld dt@L @_q_qd dt)dt; =Zt2t1((@L @_q_qL)d dt)dt: (A{11) dt>0,wepointthereaderto[ 6 ],foradiscussiononthismatter,however,wemustinsistthatd dt6=0almosteverywhere. 116

PAGE 117

dt6=0,a.e.,thatthestatementfollowingmustbetrue, @_q_qL=0a.e. (A{12) Therefore,inallreparameterizationinvarianttheoriestheHamiltonianvanishes. 117

PAGE 118

[1] S.Weinberg,TheQuantumTheoryofFieldsVol.2,CambridgeUniversityPress,Cambridge,1998. [2] A.JaeandE.Witten,\QuantumYang-MillsTheory" [3] P.Deligne,etal.,QuantumFieldsandStrings:ACourseforMathematiciansVol.1-2,AmericanMathematicalSociety,Providence,1999. [4] A.Ashtekar,J.Lewandowski,\BackgroundIndependentQuantumGravity:AStatusReport,"Class.Quant.Grav.,21(2004)R53. [5] J.Henson,\TheCausalSetApproachtoQuantumGravity," [6] J.Klauder,\AneQuantumGravity,"Int.J.Mod.Phys.,D12(2003)1769. [7] P.A.M.Dirac,LecturesonQuantumMechanics,BelferGraduateSchoolofScience,YeshivaUniversity,NewYork,1964. [8] J.S.Little,J.Klauder,\ElementaryModelofConstraintQuantizationwithanAnomaly,"Phys.Rev.D,D71(2005)085014. [9] J.Klauder,J.S.Little,\HighlyIrregularQuantumConstraints,"Class.Quant.Grav,23(2006)3641. [10] J.S.Little,\TheProjectionOperatorMethodandtheAshtekar-Horowitz-BoulwareModel" [11] A.Ashtekar,G.T.Horowitz,\OntheCanonicalApproachtoQuantumGravity",inPhys.Rev.D,26(1982),3342. [12] D.M.Gitman,I.V.Tyutin,QuantizationofFieldswithConstraints,Springer-Verlag,Berlin1990. [13] J.Klauder,\QuantizationofConstrainedSystems,"Lect.NotesPhys.,572,(2001)143. [14] M.Henneaux,C.Teitelboim,QuantizationofGaugeSystems,PrincetonUniversityPress,Princeton,NJ,1992. [15] J.Jose,E.Saletan,ClassicalDynamics:AContempraryApproach,CambridgeUniversityPress1998. [16] J.Govaerts,HamiltonianQuantisationandConstrainedDynamics,LeuvenUniversityPress,Belgium,1991. 118

PAGE 119

V.I.Arnold,MathematicalMethodsofClassicalMechanics,SingerVerlag,3rded.NewYork,NewYork(1989). [18] L.Castellani,\SymmetriesinConstrainedHamiltoniansystems,"Ann.Phys,143,(1982),357. [19] A.ADeriglazov,K.E.Evdokimov.\LocalSymmetriesintheHamiltonianFramework.1.HamiltonianFormoftheSymmetriesandtheNoetheridentities,"Int.J.Mod.Phys.A15(2000)4045. [20] P.A.MDirac,ThePrinciplesofQuantumMechanics,4thed.,OxfordSciencePublications,1998. [21] J.Klauder,\MetricalQuantization" [22] J.Simon,ChangeWithoutTimeRelationalismandFieldQuantization,Dissertation,UniversitatRegensburg,NaturwissenschaftlicheFakultatII-Physik,2004. [23] N.P.Landsman,MathematicalTopicsBetweenClassicalandQuantumMechanicsSpringer-VerlagNewYork,1998. [50] D.Giulini,\ThatStrangeProcedureCalledQuantisation", [25] A.Ashtekar,R.S.Tate,\AnAlgebraicExtensionofDiracQuantization:Examples," [26] S.V.Shabanov,\GeometryofthePhysicalPhaseSpaceinQuantumGaugeModels,"PhysicsReports3261(2000),hep-th/0002043. [27] L.D.Faddeev,\FeynmanIntegralforSingularLagrangians,"inTheor.Math.Phys.1,(1969),1. [28] P.Senjanovc,\PathIntegralQuantizationofFieldTheorieswithSecond-ClassConstraints,"Ann.Phys.100,(1976),227. [29] AdrianoDiGiacomo,\ConnementofColor:OpenProblemsandPerspectives," [30] D.Giulini,D.Marolf,\OntheGeneralityofRenedAlgebraicQuantization,"Class.Quant.Grav.,16,(1999),247. [31] D.Marolf,\RenedAlgebraicQuantization:SystemswithaSingleConstraint," [32] J.Louko,A.Molgado,\SuperselectionsectorsintheAshtekar-Horowitz-BoulwareModel",Class.QuantumGrav.22,(2005)4007. 119

PAGE 120

B.Dittrich,T.Thiemann,\TestingtheMasterConstraintProgrammeforLoopQuantumGravityI.GeneralFramework,"Class.Quant.Grav.23,(2006),1025. [34] T.Thiemann,\ThePhoenixProject:MasterConstraintProgrammeforLoopQuantumGravity,"Class.Quant.Grav.23,(2006),2211. [35] B.Dittrich,T.Thiemann,\TestingtheMasterConstraintProgrammeforLoopQuantumGravityII.FiniteDimensionalSystems,"inClass.Quant.Grav.23,(2006),1067. [36] B.Dittrich,T.Thiemann,\TestingtheMasterConstraintProgrammeforLoopQuantumGravityIII.SL(2,R)Models,"inClass.Quant.Grav.,23,(2006)1089. [37] B.Dittrich,T.Thiemann,\TestingtheMasterConstraintProgrammeforLoopQuantumGravityV.InteractingFieldTheories,"inClass.Quant.Grav.23,(2006),1143. [38] R.Shankar,PrinciplesofQuantumMechanics,PlenumPress,2nded.,NewYork,NewYork,1994. [39] J.Klauder,\PathIntegrals,andClassicalandQuantumConstraints" [40] J.Klauder,\CoherentStateQuantizationofConstraintSystems,"Ann.Phys.245(1997)419. [41] J.R.Klauder,B.Skagerstam,CoherentStates,WorldScienticPublishing,Singapore,1985. [42] N.Aronszajn,Proc.CambridgePhil.Soc.,39(1943),133;Trans.Amer.MathSoc.68337(1950);H.Meschkowski,HilbertscheRaummitKernfunktion,'Springer-Verlag,Berlin,1962 [43] H.Araki,\HamiltonianFormalismandtheCanonicalCommutationRelationsinQuantumFieldTheory,"J.Math.Phys1(1960),492. [44] D.G.Boulware,\Commenton`OntheCanonicalApproachtoQuantumGravity',"Phys.Rev.D28(1983),414. [45] J.AGonzlez,M.A.delOlmo,\CoherentStatesontheCircle,"J.Phys.A:Math.Gen.31(1998),8841. [46] J.Louko,\GroupAveraging,PositiveDenitenessandDuperselectionSectors,"J.Phys.Conf.Ser.33(2006),142. [47] J.Klauder,\UniversalProcedureforEnforcingQuantumConstraints",Nucl.Phys.B547(1999),397. [48] G.Fulop,D.M.Gitman,I.V.Tyutin.\ReparametrizationInvarianceasGaugeSymmetry,"Int.J.Theor.Phys.38(1999)1941. 120

PAGE 121

A.Garcia,D.Vergara,L.F.Urrutia,\BRST-BFVMethodforNonstationarySystems,"PhysicalReviewD51(1995),5806. [50] C.Rovelli,\Partialobservables,"Phys.Rev.D65(2002),124013. [51] B.Dittric,\PartialandCompleteObservablesforHamiltonianConstrainedSystems," [52] W.Pauli.\DieallgemeinenPrinzipienderWellenmechanik,"inHandbuchderPhysik1Springer,Berlin,1926. [53] J.Klauder,\AttractionsofAneQuantumGravity," [54] J.Klauder,\UltralocalFieldsandTheirRelevanceforReparametrizationInvariantQuantumFieldTheory,"J.Phys.A34(2001),3277. [55] S.P.Gavrilov,D.M.Gitman,\QuantizationofSystemswithTime-DependentConstraints.ExampleofRelativisticParticleinPlaneWave,"inClass.Quant.Grav.10(1993),57. 121

PAGE 122

JereyScottLittlewasbornJanurary10,1980,inthesmalleasternKentuckytownofPikeville.HeistheeldestofthreechildrenandtheonlysonofJeandLindaLittle.Hisstronginterestinsciencewasevidentataveryearlyage.WhenScottwas14,hediscoveredhispassionforquantumphysicsandreadnearlyeverythinghecouldndonthesubject.ScottgraduatedfromShelbyValleyHighSchoolin1998,asclassvaledictorian.Afterhighschool,ScottsetotomatriculateatWesternKentuckyUniversityinBowlingGreen,Kentucky.WhileatWesternhebecameextremelyinterestedinthestudyofformalmathematics.Duringhissenioryearheentertainedtheideaofattendinggraduateschoolinmathematics;however,herealizedthatitwouldbepossibletopursuebothpassionsthroughaphysicscareer.ScottgraduatedfromWesternKentuckywithadoublebachelorsdegreeinphysicsandmathematicsintheSpringof2002.AfternishinghisundergraduatecareerheacceptedanAlumniFellowshipfromtheUniversityofFloridatocontinuehisstudiesofphysics.ThoughmovingfromarelativelysmalldepartmentatWesterntothemuchlargerPhysicsDepartmentatFloridawasinitiallydaunting,Scottovercamehisfearsandsucceededinhiscoursework.IntheSpringof2004,ScottbecameastudentofJohnKlauder.Dr.KlauderallowedScotttonotonlystudyphysicsbutalsoallowedhimtostayconnectedtotheformalmathematicsthathehadgrownfondofduringhisstayatWestern.UnderDr.Klauder'stutelage,Scottwasabletoresearchandpublishthreepapersonquantumconstraints.InJuneof2006,attheageof26,ScottmarriedtheloveofhislifeMegan(Carty)Little.ScottobtainedhisPh.D.inPhysicsintheFallof2007.ScottandMegancurrentlyresideinLouisville,KY,whereScottiscontinuingtoresearchawidevarietyoftheoreticalproblemsandisaninstructorattheUniversityofLouisville. 122