quantomorphisms and quantized energy levels for...
TRANSCRIPT
Quantomorphisms and Quantized Energy Levels for Metaplectic-cQuantization
by
Jennifer Vaughan
A thesis submitted in conformity with the requirementsfor the degree of Doctor of PhilosophyGraduate Department of Mathematics
University of Toronto
c© Copyright 2016 by Jennifer Vaughan
Abstract
Quantomorphisms and Quantized Energy Levels for Metaplectic-c Quantization
Jennifer Vaughan
Doctor of Philosophy
Graduate Department of Mathematics
University of Toronto
2016
Metaplectic-c quantization was developed by Robinson and Rawnsley as an alternative to the
classical Kostant-Souriau quantization procedure with half-form correction. This thesis extends
certain properties of Kostant-Souriau quantization to the metaplectic-c context. We show that
the Kostant-Souriau results are replicated or improved upon with metaplectic-c quantization.
We consider two topics: quantomorphisms and quantized energy levels. If a symplectic man-
ifold admits a Kostant-Souriau prequantization circle bundle, then its Poisson algebra is realized
as the space of infinitesimal quantomorphisms of that circle bundle. We present a definition
for a metaplectic-c quantomorphism, and prove that the space of infinitesimal metaplectic-c
quantomorphisms exhibits all of the same properties that are seen in the Kostant-Souriau case.
Next, given a metaplectic-c prequantized symplectic manifold (M,ω) and a function H ∈
C∞(M), we propose a condition under which E, a regular value of H, is a quantized energy
level for the system (M,ω,H). We prove that our definition is dynamically invariant: if two
functions on M share a regular level set, then the quantization condition over that level set
is identical for both functions. We calculate the quantized energy levels for the n-dimensional
harmonic oscillator and the hydrogen atom, and obtain the quantum mechanical predictions in
both cases. Lastly, we generalize the quantization condition to a level set of a family of Poisson-
commuting functions, and show that in the special case of a completely integrable system, it
reduces to a Bohr-Sommerfeld condition.
ii
Contents
1 Introduction 1
2 Background 4
2.1 Symplectic manifolds and bundles . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 Hamiltonian vector fields and the Poisson algebra . . . . . . . . . . . . . . 5
2.1.2 Circle bundles and connection one-forms . . . . . . . . . . . . . . . . . . . 5
2.1.3 Complex line bundles and connections . . . . . . . . . . . . . . . . . . . . 6
2.1.4 Holonomy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Kostant-Souriau quantization and the half-form correction . . . . . . . . . . . . . 7
2.2.1 Prequantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.2 Quantization and the half-form correction . . . . . . . . . . . . . . . . . . 9
2.3 Metaplectic-c quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3.1 Definitions on vector spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3.2 Definitions over a symplectic manifold . . . . . . . . . . . . . . . . . . . . 12
3 Metaplectic-c Quantomorphisms 15
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2 Kostant-Souriau quantomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2.1 Infinitesimal quantomorphisms of a prequantization circle bundle . . . . . 16
3.2.2 The Lie algebra isomorphism . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2.3 An operator representation of C∞(M) . . . . . . . . . . . . . . . . . . . . 20
3.3 Metaplectic-c Quantomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3.1 Infinitesimal metaplectic-c quantomorphisms . . . . . . . . . . . . . . . . 22
iii
3.3.2 The Lie algebra isomorphism . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3.3 An operator representation of C∞(M) . . . . . . . . . . . . . . . . . . . . 28
3.4 Two diffeomorphisms that are not quantomorphisms . . . . . . . . . . . . . . . . 29
4 Quantized Energy Levels and Dynamical Invariance 33
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2 Quantized Energy Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2.1 Vector subspaces and quotients . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2.2 Quotients and reduction over a manifold . . . . . . . . . . . . . . . . . . . 36
4.3 Dynamical Invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3.1 Invariance Under a Diffeomorphism . . . . . . . . . . . . . . . . . . . . . 40
4.3.2 Regular Level Set of Multiple Functions . . . . . . . . . . . . . . . . . . . 41
4.4 The Harmonic Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.4.1 Initial Choices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.4.2 From Cartesian to Symplectic Polar Coordinates . . . . . . . . . . . . . . 48
4.4.3 Quantized Energy Levels of the Harmonic Oscillator . . . . . . . . . . . . 54
4.4.4 Example of Dynamical Invariance . . . . . . . . . . . . . . . . . . . . . . 57
4.5 Comparison with Kostant-Souriau Quantization . . . . . . . . . . . . . . . . . . . 60
4.5.1 The Quantized Energy Condition and Its Properties . . . . . . . . . . . . 60
4.5.2 Lack of Half-Shift in the Harmonic Oscillator . . . . . . . . . . . . . . . . 62
5 The Hydrogen Atom 64
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.1.1 Choices for Subsequent Calculations . . . . . . . . . . . . . . . . . . . . . 65
5.2 Quantizing the Hydrogen Atom . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.2.1 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.2.2 Metaplectic-c Prequantization of (M,ω) . . . . . . . . . . . . . . . . . . . 67
5.3 Ligon-Schaaf Regularization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.3.1 TS3 and the Ligon-Schaaf Map . . . . . . . . . . . . . . . . . . . . . . . . 70
5.3.2 Metaplectic-c Prequantization for TS3 . . . . . . . . . . . . . . . . . . . . 71
5.3.3 Lifting the Ligon-Schaaf Map . . . . . . . . . . . . . . . . . . . . . . . . . 74
iv
5.3.4 Rescaling the Delaunay Hamiltonian . . . . . . . . . . . . . . . . . . . . . 76
5.4 Quantization of a Free Particle on S3 . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.4.1 Orbits of ξK and Local Coordinates for TS3 . . . . . . . . . . . . . . . . . 77
5.4.2 Change of variables over C . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.4.3 Restrictions to C ⊂ TS3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.4.4 Quantized Energy Levels for (TS3, ω3,K) . . . . . . . . . . . . . . . . . . 85
6 Generalization of the Quantized Energy Condition 90
6.1 The quantization condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.2 Generalized dynamical invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6.3 Completely integrable systems and Bohr-Sommerfeld conditions . . . . . . . . . . 96
6.3.1 Kostant-Souriau quantization and Bohr-Sommerfeld conditions . . . . . . 96
6.3.2 Quantized energy condition for k = n . . . . . . . . . . . . . . . . . . . . 97
6.3.3 The n-dimensional harmonic oscillator . . . . . . . . . . . . . . . . . . . . 97
6.3.4 The 2-dimensional hydrogen atom . . . . . . . . . . . . . . . . . . . . . . 100
v
Chapter 1
Introduction
In physics, a classical mechanical system is described by a phase space, which contains position
and momentum coordinates for the component particles. The physically observable quantities
for the system are functions on phase space. To form the quantum mechanical version of such
a system, we construct a Hilbert space of quantum states, and we identify observables with
self-adjoint operators on the Hilbert space.
Geometric quantization is a branch of symplectic geometry that generalizes the concept
of quantization to abstract symplectic manifolds. Given a symplectic manifold (M,ω) and a
function H ∈ C∞(M), the corresponding Hamiltonian vector field ξH on M has integral curves
that satisfy Hamilton’s equations in local canonical coordinates. Thus we can view (M,ω)
as a classical phase space, and H as a Hamiltonian energy function. To quantize (M,ω), we
require a recipe for constructing a suitable Hilbert space of quantum states, and a Lie algebra
isomorphism from C∞(M), or a subalgebra of it, to operators on those states.
The original formulation of geometric quantization is due to Kostant [12] and Souriau [20].
Their quantization procedure requires that (M,ω) admit a prequantization circle bundle and
a metaplectic structure. In Chapter 2, after we review the necessary elements of symplectic
geometry and principal bundles, we present the Kostant-Souriau quantization procedure with
half-form correction.
The metaplectic-c group is a circle extension of the symplectic group. Metaplectic-c quan-
tization, which was developed by Robinson and Rawnsley [15], is a variant of Kostant-Souriau
quantization in which the prequantization circle bundle and metaplectic structure are replaced
1
Chapter 1. Introduction 2
by a single object called a metaplectic-c prequantization. Robinson and Rawnsley proved that
metaplectic-c quantization can be applied to all systems that admit metaplectic quantizations,
and to some where the Kostant-Souriau process fails. In the final section of Chapter 2, we give
a detailed description of a metaplectic-c prequantization and its properties.
The objective of this document is to explore some of the ways in which metaplectic-c quan-
tization replicates or improves upon known results for Kostant-Souriau quantization. We begin
in Chapter 3 by examining the notion of a quantomorphism: that is, an isomorphism of pre-
quantization bundles that preserves all of their structures. In the context of a Kostant-Souriau
prequantization circle bundle for (M,ω), it is known that the space of infinitesimal quantomor-
phisms is isomorphic, as a Lie algebra, to C∞(M). We formulate a definition of a metaplectic-c
quantomorphism, and prove that the space of infinitesimal metaplectic-c quantomorphisms is
again isomorphic to C∞(M).
The remainder of the document focuses on the concept of a quantized energy level. In
quantum mechanics, the energy spectrum for a spatially confined particle is discrete: only
certain energy levels are permitted. Various interpretations of this phenomenon can be found
in the literature for different forms of geometric quantization. Given (M,ω) and a function
H ∈ C∞(M), we propose a new definition for a quantized energy level E of H in the case where
(M,ω) admits a metaplectic-c prequantization. Our definition is evaluated over a regular level
set of H, and it does not require either symplectic reduction or a choice of polarization.
We present our definition in Chapter 4, and show that its properties compare favourably
with others that have been studied. Our main result, Theorem 4.3.5, states that if two functions
H1, H2 ∈ C∞(M) are such that H−11 (E1) = H−1
2 (E2) for regular values E1, E2, then E1 is a
quantized energy level for (M,ω,H1) if and only if E2 is a quantized energy level for (M,ω,H2).
That is, our condition measures a geometric property of the level set, and does not depend on
the dynamics of a specific choice of H. As such, we refer to Theorem 4.3.5 as the dynamical
invariance theorem. Also in Chapter 4, we demonstrate a computational technique for lifting
a local change of coordinates on M to the level of metaplectic-c prequantizations, and use this
technique to evaluate the quantized energy levels for the n-dimensional harmonic oscillator.
In Chapter 5, we calculate the quantized energy levels of the hydrogen atom, using the
physical model that is identical to the Kepler problem. This calculation is more technically
Chapter 1. Introduction 3
demanding than that for the harmonic oscillator because the hydrogen atom has fewer symme-
tries. We use the Ligon-Schaaf regularization map to transport the problem to TS3, and apply
a further transformation to relate the negative quantized energy levels of the hydrogen atom to
those of a free particle on S3. The latter step depends on the dynamical invariance property.
We show that the metaplectic-c quantized energy levels agree with the physical prediction from
quantum mechanics.
Lastly, in Chapter 6, we show that our quantized energy condition generalizes from one
function on M to a family of Poisson-commuting functions. After proving a generalized version
of the dynamical invariance theorem, we consider the special case of a completely integrable
system, where the size of the Poisson-commuting family is maximal. In that case, the quan-
tized energy condition simplifies to a Bohr-Sommerfeld condition. Thus our quantized energy
condition provides a framework that encompasses both the quantized energy levels of a single
function H and the Bohr-Sommerfeld leaves of the real polarization generated by a completely
integrable system.
Significant portions of this document have already been published. Chapter 3 appears in
[21], Chapter 4 appears in [22], and Chapter 5 appears in [23]. Portions of Chapter 2 and this
introduction are amalgams of material from all three papers. Minor truncations and edits have
been performed for the sake of internal consistency and to eliminate redundancy. Additional
background material has been added to Chapter 2, and the content of Chapter 6 does not
appear elsewhere.
Chapter 2
Background
Our starting point is a symplectic manifold (M2n, ω): that is, a 2n-dimensional manifold
equipped with a closed, nondegenerate two-form. This object plays the role of a phase space
in classical mechanics. Classical observables are elements of C∞(M), the smooth functions on
M . One objective of geometric quantization is to build a Hermitian line bundle over M and
a Lie algebra homomorphism from C∞(M) to operators on sections of the bundle, thereby
reproducing the transition from classical to quantum mechanics in which observables become
operators on a Hilbert space of quantum states. In Section 2.1, we review the basic facts that
we will require concerning symplectic manifolds, circle bundles and complex line bundles, and
connections.
In Section 2.2, we describe the Kostant-Souriau quantization procedure. We begin with the
prequantization stage, in which (M,ω) is required to admit a prequantization line bundle (L,∇).
Then we introduce a metaplectic structure and a choice of polarization F , and sketch how to
construct the complex line bundle of half-forms ∧1/2F . We conclude with the quantization
stage, in which the functions in C∞(M) whose flows preserve the polarization are mapped to
operators on polarized sections of L⊗ ∧1/2F .
Finally, in Section 2.3, we summarize the construction of a metaplectic-c prequantization,
as developed by Robinson and Rawnsley [15]. A metaplectic-c prequantization is an object that
plays the role of a prequantization line bundle and a metaplectic structure simultaneously. As
we will see in later chapters, certain features of Kostant-Souriau quantization with half-form
correction can be reproduced by a metaplectic-c prequantization alone, without requiring a
4
Chapter 2. Background 5
choice of polarization. Since our work focuses on the prequantization stage of Robinson and
Rawnsley’s procedure, we omit a detailed discussion of the quantization stage. Some elements
of it are presented in Section 3.3.3 in the context of metaplectic-c quantomorphisms.
Some global remarks concerning notation: for any vector field ξ, the Lie derivative with
respect to ξ is written Lξ. The space of smooth vector fields on a manifold P is denoted by
X (P ). Given a smooth map F : P → M and a vector field ξ ∈ X (P ), we write F∗ξ for the
pushforward of ξ only if the result is a well-defined vector field on M . If P is a bundle over
M , Γ(P ) denotes the space of smooth sections of P , where the base is always taken to be the
symplectic manifold M . Planck’s constant will only appear in the form ~.
2.1 Symplectic manifolds and bundles
2.1.1 Hamiltonian vector fields and the Poisson algebra
Let (M,ω) be a symplectic manifold. Given f ∈ C∞(M), define its Hamiltonian vector field
ξf ∈ X (M) by
ξfyω = df.
Define the Poisson bracket on C∞(M) by
f, g = ξfg = −ω(ξf , ξg), ∀f, g ∈ C∞(M).
A standard calculation establishes that for all f, g ∈ C∞(M),
[ξf , ξg] = ξf,g.
2.1.2 Circle bundles and connection one-forms
Let Yp−→ M be a right principal U(1) bundle over M . For any θ ∈ u(1), the Lie algebra of
U(1), let ∂θ be the vector field on Y given by
∂θ(y) =d
dt
∣∣∣∣t=0
y · exp(tθ), ∀y ∈ Y.
Chapter 2. Background 6
The vector field ∂θ is called the vector field generated by the infinitesimal action of θ ∈ u(1).
A u(1)-valued one-form γ on Y is called a connection one-form if γ is invariant under the
right principal action, and for all θ ∈ u(1), γ(∂θ) = θ. If Y is equipped with a connection
one-form γ, then there is a two-form $ on M , called the curvature of γ, such that dγ = p∗$.
For any ξ ∈ X (M), let ξ be the lift of ξ to Y that is horizontal with respect to γ. That
is, p∗ξ = ξ and γ(ξ) = 0. For any θ ∈ u(1), note that p∗∂θ = 0, which implies that p∗[ξ, ∂θ] =
[p∗ξ, p∗∂θ] = 0 and γ([ξ, ∂θ]) = −(p∗$)(ξ, ∂θ) = 0. Therefore [ξ, ∂θ] = 0 for all θ.
2.1.3 Complex line bundles and connections
Now let L→M be a complex line bundle over M . A connection ∇ on L allows us to take the
derivative of a section s of L in the direction of a vector field ξ on M . It can be viewed as a
map from sections of L to L-valued one-forms on M in the sense that, if s is a section of L and
ξ is a vector field on M , then at every point m ∈ M , ∇s acts on the vector ξ(m) to yield a
value in Lm, which we interpret as the derivative of s in the direction of that vector. Formally,
a connection on L is a map
∇ : Γ(L)→ Γ(L)⊗ Ω1(M)
such that, for all r, s ∈ Γ(L), all ξ ∈ X (M), and all f ∈ C∞(M),
∇ξ(r + s) = ∇ξr +∇ξs,
∇ξ(fs) = f∇ξs+ (ξf)s.
If (Y, γ)p−→M is a circle bundle with connection one-form over M , then it has an associated
complex line bundle with connection (L,∇) over M . The complex line bundle L is given by
L = Y ×U(1) C. We write an element of L as an equivalence class [y, z] with y ∈ Y and z ∈ C.
The connection ∇ on L is constructed from the connection one-form γ through the following
process.
Given any s ∈ Γ(L), define the map s : Y → C so that [y, s(y)] = s(p(y)) for all y ∈ Y .
Then s has the equivariance property
s(y · λ) = λ−1s(y), ∀ y ∈ Y, λ ∈ U(1).
Chapter 2. Background 7
Conversely, any map s : Y → C with the above equivariance property can be used to construct
a section s of L by setting s(m) = [y, s(y)] for all m ∈M and any y ∈ Y such that p(y) = m.
Let ξ ∈ X (M) be given, and let ξ be its horizontal lift to Y . If s : Y → C is an equivariant
map, then so is ξs. This follows from the fact that [ξ, ∂θ] = 0 for all θ ∈ u(1). Define the
connection ∇ on L so that for any ξ ∈ X (M) and s ∈ Γ(L), ∇ξs is the section of L that
satisfies
∇ξs = ξs.
2.1.4 Holonomy
Let (Y, γ) be a circle bundle with connection one-form over M . Let u : R → M be a path in
M such that u(t) = u(t + 1) for all t, and let C ⊂ M be the image of u in M . At every point
y ∈ Yu(t), the tangent vector u(t) lifts horizontally to a tangent vector ˜u(t) ∈ TyY , resulting in
a vector field on Y |C .
Fix a starting point y0 ∈ Yu(0). Let u(t) be the integral curve for the lifted vector field such
that u(0) = y0. Then u(t) must be a lift of u(t). In particular, since u(0) = u(1), we have
u(1) ∈ Yu(0). The resulting map y0 7→ u(1) is an automorphism of the fiber Yu(0), called the
holonomy1 of γ over the closed curve u(t). If u(1) = y0, then γ is said to have trivial holonomy
over u(t).
The concept of holonomy can be translated to the corresponding line bundle with connection
(L,∇). Let u(t) be a lift of u(t) to L. The image of u(t) can be thought of as a section s of L,
defined only over C. The lift u(t) is called horizontal if ∇u(t)s = 0 for all t. Given a starting
point u(0) in the fiber Lu(0), the condition of being horizontal uniquely determines the rest of
the lift. As before, u(1) ∈ Lu(0), so we obtain a map u(0) 7→ u(1) from the fiber Lu(0) to itself,
called the holonomy of ∇ over u(t).
2.2 Kostant-Souriau quantization and the half-form correction
Geometric quantization in its original form was developed in the 1960’s by Kostant [12] and
Souriau [20]. The half-form correction was added by Blattner, Kostant and Sternberg [2]. The
1This material is standard, but we are particularly indebted to the clarity of exposition in [11]
Chapter 2. Background 8
following overview is based on the detailed treatments that can be found in [10], [18], and [24].
2.2.1 Prequantization
Let (M,ω) be a symplectic manifold. In the prequantization stage, we require (M,ω) to admit
a prequantization circle bundle.
Definition 2.2.1. A prequantization circle bundle for (M,ω) is a right principal U(1)
bundle Yp−→M , together with a connection one-form γ on Y satisfying dγ = 1
i~p∗ω.
If (M,ω) admits a prequantization circle bundle, then the associated complex line bundle
with connection (L,∇) is called a prequantization line bundle. Note that if we view L as
Y ×U(1) C, then the usual Hermitian inner product on C can be used to define a Hermitian
structure on L, and that Hermitian structure is compatible with the connection∇. The manifold
(M,ω) admits a prequantization circle bundle (equivalently, a prequantization line bundle) if
and only if the cohomology class[
1hω]∈ H2(M,R) is integral.
Assume that (M,ω) admits a prequantization circle bundle (Y, γ), with corresponding pre-
quantization line bundle (L,∇). One of the goals of the prequantization process is to produce
a representation r : C∞(M) → End Γ(L). To be consistent with quantum mechanics in the
case of a physically realizable system, the map r is required to satisfy the following axioms,
which are based on an analysis by Dirac [5] on the relationship between classical and quantum
mechanical observables.
(1) r(1) is the identity map on Γ(L),
(2) for all f, g ∈ C∞(M), [r(f), r(g)] = i~r(f, g) (up to sign convention).
In the context of Kostant-Souriau prequantization, a suitable map is given by
r(f) = i~∇ξf + f, ∀f ∈ C∞(M).
The fact that this map satisfies condition (2) can be verified by direct computation; as we will
show in Section 3.2.3, it can also be viewed as a consequence of the Lie algebra isomorphism
between C∞(M) and the space of infinitesimal quantomorphisms of (Y, γ).
Chapter 2. Background 9
2.2.2 Quantization and the half-form correction
Recall that geometric quantization is motivated by the transition from classical to quantum
mechanics. The phase space for a classical system consists of all possible combinations of
position and momentum coordinates. However, the wave functions for the corresponding quan-
tum system depend only on the position coordinates (or, more generally, on a complete set of
commuting observables). This observation motivates the introduction of a structure called a
polarization.
Assume that (M,ω) admits a prequantization line bundle (L,∇). A polarization2 F is an
involutive Lagrangian subbundle of the complexified tangent bundle TMC such that dim(Fm ∩
Fm) is constant over all m ∈ M . Given a choice of F , we can restrict our attention to those
sections of L that are flat in the directions parallel to F .
Let a choice of F be fixed. The canonical bundle for F is the top exterior power of the
annihilator of F in (T ∗M)C, denoted by KF . The half-form bundle, ∧1/2F , is a choice of
square root of KF , when such a square root exists. A half-form is a section of ∧1/2F over M .
To guarantee that the half-form bundle ∧1/2F exists, we require that (M,ω) admit a meta-
plectic structure, which is a lifting of the structure group for (M,ω) from the symplectic group
to its unique connected double cover, the metaplectic group. In order for (M,ω) to admit
a metaplectic structure, the first Chern class c1(TM) ∈ H2(M,Z) must be even: that is,
12c1(TM) ∈ H2(M,Z). If this condition is satisfied, then the equivalence classes of metaplectic
structures for (M,ω) are in one-to-one correspondence with H1(M,Z2).
Assume that (M,ω) admits a metaplectic structure. Then the metaplectic structure can be
used to construct a square root of KF . From KF , the half-form bundle ∧1/2F inherits a partial
connection ∇ζ , defined for all ζ ∈ X (M) such that ζ is parallel to the polarization F . It also
inherits a partial Lie derivative Lξ, defined for all ξ ∈ X (M) such that the flow of ξ preserves
F .
Let C∞F (M) denote the subalgebra of C∞(M) consisting of functions whose Hamiltonian
flows preserve the polarization F . The Kostant-Souriau representation with half-form correc-
2Other technical constraints, such as positivity, may be imposed on F , but since we will not be performingany explicit computations with polarizations, we will not address these details.
Chapter 2. Background 10
tion, rF : C∞F (M)→ End Γ(L⊗ ∧1/2F ), is defined by
rF (f) = r(f)⊗ I + I ⊗ Lξf , ∀f ∈ C∞F (M).
This corrected map is a Lie algebra homomorphism.
A section s⊗ ν ∈ Γ(L⊗∧1/2F ) is called polarized if ∇ζs = 0 and ∇ζν = 0 for all ζ ∈ X (M)
such that ζ is parallel to F . Let ΓF (L ⊗ ∧1/2F ) denote the space of polarized sections. The
operators rF (f) restrict to operators on ΓF (L⊗ ∧1/2F ). The polarized sections with compact
support form the pre-Hilbert space of quantum states, where the Hermitian inner product arises
from the Hermitian structure on L, and a pairing from half-forms to densities. This completes
the construction of the algebra representation and the Hilbert space as suggested by quantum
mechanics.
2.3 Metaplectic-c quantization
The Kostant-Souriau quantization procedure imposes two conditions on (M,ω): namely, that it
admit a prequantization circle bundle and a metaplectic structure. Each of these requirements
comes with a cohomological condition to be satisfied. The motivation behind metaplectic-c
quantization is to replace the prequantization circle bundle and metaplectic structure with a
single bundle that can play both roles simultaneously.
The material in this section is a summary of results originally presented by Robinson and
Rawnsley [15]. More detail, including proofs of the properties that we state, can be found in
[15]. A similar structure, called a spin-c prequantization, was studied in [4, 7, 8, 9].
2.3.1 Definitions on vector spaces
Let V be an n-dimensional complex vector space with Hermitian inner product 〈·, ·〉. If we view
V as a 2n-dimensional real vector space, then the action of the scalar i ∈ C becomes the real
automorphism J : V → V . Define the real bilinear form Ω on V by
Ω(v, w) = Im 〈v, w〉 , ∀v, w ∈ V.
Chapter 2. Background 11
Then (V,Ω) is a 2n-dimensional symplectic vector space. The Hermitian and symplectic struc-
tures are compatible in the sense that
〈v, w〉 = Ω(Jv,w) + iΩ(v, w), ∀v, w ∈ V.
The symplectic group Sp(V ) is the group of real automorphisms g : V → V such that
Ω(gv, gw) = Ω(v, w) for all v, w ∈ V . Since the fundamental group for Sp(V ) is Z, Sp(V ) has
a unique double cover called the metaplectic group, denoted by Mp(V ). Let Mp(V )σ−→ Sp(V )
denote the covering map. The metaplectic-c group Mpc(V ) is defined to be
Mpc(V ) = Mp(V )×Z2 U(1),
where Z2 ⊂ U(1) is the usual subgroup 1,−1, and where Z2 ⊂ Mp(V ) consists of the two
preimages of I ∈ Sp(V ) under the covering map σ.
By construction, Mpc(V ) contains U(1) and Mp(V ) as subgroups. The inclusion of each
subgroup into Mpc(V ) yields a short exact sequence and a group homomorphism on Mpc(V ).
One such sequence is
1→ U(1)→ Mpc(V )σ−→ Sp(V )→ 1,
where the group homomorphism σ is called the projection map, and its restriction to Mp(V ) ⊂
Mpc(V ) is the double covering map. The other is
1→ Mp(V )→ Mpc(V )η−→ U(1)→ 1,
where the group homomorphism η is called the determinant map, and it has the property that
for any λ ∈ U(1) ⊂ Mpc(V ), η(λ) = λ2. At the level of Lie algebras, the map σ∗ ⊕ 12η∗ yields
the identification
mpc(V ) = sp(V )⊕ u(1).
Given g ∈ Sp(V ), let
Cg =1
2(g − JgJ).
Chapter 2. Background 12
By construction, Cg commutes with J , so it is a complex linear transformation of V . The
determinant of Cg as a complex transformation is written DetCCg. It can be shown that Cg is
always invertible, so DetCCg is a nonzero complex number.
A useful embedding of Mpc(V ) into Sp(V )×C \ 0 can be defined as follows. An element
a ∈ Mpc(V ) is mapped to the pair (g, µ) ∈ Sp(V )×C\0, where g ∈ Sp(V ) satisfies σ(a) = g,
and µ ∈ C \ 0 satisfies η(a) = µ2DetCCg. The latter condition defines µ up to a sign, and
we adopt the convention that the parameters of I ∈ Mpc(V ) are (I, 1). This choice uniquely
determines the sign of µ for all a ∈ Mpc(V ). Following [15], we refer to (g, µ) as the parameters
of a.
We note two properties of this parametrization, both of which will be useful in later chapters.
First, if a ∈ Mp(V ) ⊂ Mpc(V ), then η(a) = 1, and so the parameters of a are (g, µ) where
σ(a) = g and µ2DetCCg = 1. Second, if λ ∈ U(1) ⊂ Mpc(V ), then the parameters of λ are
(I, λ), and for any other element a ∈ Mpc(V ) with parameters (g, µ), the parameters of the
product aλ are (g, µλ).
2.3.2 Definitions over a symplectic manifold
Let (M,ω) be a 2n-dimensional symplectic manifold, and assume that a 2n-dimensional model
symplectic vector space (V,Ω) with complex structure J has been fixed. We think of the
symplectic frame bundle for (M,ω) as being modeled on Sp(V ), in the following sense.
Definition 2.3.1. The symplectic frame bundle Sp(M,ω)ρ−→M is a right principal Sp(V )
bundle over M given by
Sp(M,ω)m = b : V → TmM : b is a symplectic linear isomorphism , ∀m ∈M.
The group Sp(V ) acts on the fibers by precomposition.
In the previous section, we defined a metaplectic structure for (M,ω) to be a lifting of the
structure group for (M,ω) from Sp(V ) to Mp(V ). Analogously, a metaplectic-c structure for
(M,ω) is a lifting of the structure group to Mpc(V ).
Definition 2.3.2. A metaplectic-c structure for (M,ω) consists of a right principal Mpc(V )
Chapter 2. Background 13
bundle PΠ−→M and a map P
Σ−→ Sp(M,ω) such that
Σ(q · a) = Σ(q) · σ(a), ∀q ∈ P, ∀a ∈ Mpc(V ),
and such that the following diagram commutes.
PΣ //
Π
Sp(M,ω)
ρyy
(M,ω)
To complete the construction of a metaplectic-c prequantization, we equip the metapletic-
c structure with a u(1)-valued one-form that is analogous to the connection one-form on a
prequantization circle bundle.
Definition 2.3.3. A metaplectic-c prequantization of (M,ω) is a metaplectic-c structure
(P,Σ) for (M,ω), together with a u(1)-valued one-form γ on P such that:
(1) γ is invariant under the principal Mpc(V ) action;
(2) for any α ∈ mpc(V ), if ∂α is the vector field on P generated by the infinitesimal action of
α, then γ(∂α) = 12η∗α;
(3) dγ = 1i~Π∗ω.
Note that if (P,Σ, γ) is a metaplectic-c prequantization for (M,ω), then (P, γ) is a circle
bundle with connection one-form over Sp(M,ω). The circle that acts on the fibers of P is the
center U(1) ⊂ Mpc(V ). We will view P as a bundle over M or a bundle over Sp(M,ω) as the
circumstance demands.
For any α ∈ mpc(V ), we can write α = κ ⊕ τ under the identification of mpc(V ) with
sp(V )⊕ u(1). Then γ(∂α) = τ , and naturality of the exponential map and equivariance of the
map Σ with respect to σ ensure that Σ∗∂α = ∂κ, where ∂κ is the vector field on Sp(M,ω)
generated by the infinitesimal action of κ ∈ sp(V ). In fact, ∂α is uniquely specified by the two
conditions γ(∂α) = τ and Σ∗∂α = ∂κ, an observation that will be useful in Chapter 3.
There is a cohomology constraint that must be satisfied in order for (M,ω) to admit a
metaplectic-c prequantization, and that condition is very closely related to the ones we saw for
Chapter 2. Background 14
the prequantization circle bundle and metaplectic structure. Specifically, (M,ω) is metaplectic-
c prequantizable if and only if the cohomology class[
1hω]
+ 12c1(TM) ∈ H2(M,R) is integral.
Thus, if (M,ω) admits a prequantization circle bundle and a metaplectic structure, then it is
metaplectic-c prequantizable, but the converse is not true.
Two metaplectic-c prequantizations for (M,ω) are considered equivalent if there is a dif-
feomorphism between them that preserves the one-forms and commutes with the respective
maps to Sp(M,ω). If (M,ω) is metaplectic-c prequantizable, then the set of equivalence classes
of metaplectic-c prequantizations for (M,ω) is in one-to-one correspondence with the locally
constant cohomology group H1(M,U(1)), which also parametrizes the circle bundles with flat
connection over (M,ω). In particular, if H1(M,U(1)) is trivial, then the metaplectic-c pre-
quantization of (M,ω) is unique up to isomorphism.
Chapter 3
Metaplectic-c Quantomorphisms
3.1 Introduction
Recall that a prequantization circle bundle for a symplectic manifold (M,ω) consists of a circle
bundle Y →M and a connection one-form γ on Y such that dγ = 1i~ω. Souriau [20] defined a
quantomorphism between two prequantization circle bundles (Y1, γ1)→ (M1, ω1) and (Y2, γ2)→
(M2, ω2) to be a diffeomorphism K : Y1 → Y2 such that K∗γ2 = γ1. This condition implies
that K is equivariant with respect to the principal circle actions. Souriau then defined the
infinitesimal quantomorphisms of a prequantization circle bundle (Y, γ) to be the vector fields
on Y whose flows are quantomorphisms. Kostant [12] proved that the space of infinitesimal
quantomorphisms, which we denote Q(Y, γ), is isomorphic to the Poisson algebra C∞(M). In
Section 3.2, we present an explicit construction of the isomorphism from Q(Y, γ) to C∞(M).
The objective of the rest of the chapter is to adapt the concept of an infinitesimal quantomor-
phism to the case where (M,ω) admits a metaplectic-c prequantization (P,Σ, γ). In Section 3.3,
we define a metaplectic-c quantomorphism, which is a diffeomorphism of metaplectic-c prequan-
tizations that preserves all of their structures. Our definition is based on Souriau’s, but includes
a condition that is unique to the metaplectic-c context. We then use the metaplectic-c quanto-
morphisms to define Q(P,Σ, γ), the space of infinitesimal metaplectic-c quantomorphisms. We
show that every property that was proved for Q(Y, γ) has a parallel for Q(P,Σ, γ). In particu-
lar, Q(P,Σ, γ) is isomorphic to the Poisson algebra C∞(M). The construction in Section 3.2 is
used as a model for the proofs in Section 3.3. We indicate when the calculations are analogous,
15
Chapter 3. Metaplectic-c Quantomorphisms 16
and when the metaplectic-c case requires additional steps.
3.2 Kostant-Souriau quantomorphisms
In this section, we construct a Lie algebra isomorphism from C∞(M) to the space of infinitesimal
quantomorphisms. As we have already noted, the fact that these algebras are isomorphic was
originally stated by Kostant [12] in the context of line bundles with connection. His proof can
be reconstructed from several propositions across Sections 2 – 4 of [12]. Kostant’s isomorphism
is also stated by Sniatycki [18], but much of the proof is left as an exercise. We are not aware of
a source in the literature for a self-contained proof that uses the language of principal bundles,
and this is one of our reasons for performing an explicit construction here.
The other goal of this section is to motivate the analogous constructions for a metaplectic-
c prequantization, which will be the subject of Section 3.3. Each result that we present for
Kostant-Souriau prequantization will have a parallel in the metaplectic-c case. When the proofs
are identical, we will simply refer back to the work shown here, thereby allowing Section 3.3 to
focus on those features that are unique to metaplectic-c structures.
3.2.1 Infinitesimal quantomorphisms of a prequantization circle bundle
Definition 3.2.1. Let (Y1, γ1)p1−→ (M1, ω1) and (Y2, γ2)
p2−→ (M2, ω2) be prequantization circle
bundles for two symplectic manifolds. A diffeomorphism K : Y1 → Y2 is called a quantomor-
phism if K∗γ2 = γ1.
Let K : Y1 → Y2 be a quantomorphism. Notice that for any θ ∈ u(1), the vector field ∂θ on
Y1 is completely specified by the conditions γ1(∂θ) = θ and dγ1(∂θ) = 0, and the same is true
on Y2. Since K∗γ2 = γ1, we see that K∗∂θ = ∂θ for all θ, and so K is equivariant with respect
to the principal circle actions.
Definition 3.2.2. Let (Y, γ)p−→ (M,ω) be a prequantization circle bundle. An infinitesimal
quantomorphism of (Y, γ) is a vector field ζ ∈ X (Y ) whose flow φt on Y is a quantomorphism
from its domain to its range for each t. The space of infinitesimal quantomorphisms of (Y, γ)
is denoted by Q(Y, γ).
Chapter 3. Metaplectic-c Quantomorphisms 17
Let ζ ∈ X (Y ) have flow φt. The connection form γ is preserved by φt if and only if Lζγ = 0.
Therefore the space of infinitesimal quantomorphisms of (Y, γ) is
Q(Y, γ) = ζ ∈ X (Y ) |Lζγ = 0.
If K : Y1 → Y2 is a quantomorphism, then it induces a diffeomorphism (in fact, a symplec-
tomorphism) K ′ : M1 →M2 such that the following diagram commutes.
Y1K //
p1
Y2
p2
M1K′ //M2
This implies that for any ζ ∈ Q(Y, γ) with flow φt, there is a flow φ′t on M that satisfies
p φt = φ′t p. If ζ ′ is the vector field on M with flow φ′t, then p∗ζ = ζ ′. In other words,
elements of Q(Y, γ) descend via p∗ to well-defined vector fields on M .
3.2.2 The Lie algebra isomorphism
Let (Y, γ)p−→ (M,ω) be a prequantization circle bundle. We will now present an explicit
construction of a Lie algebra isomorphism from C∞(M) to Q(Y, γ). Recall that the vector field
∂2πi on Y satisfies γ(∂2πi) = 2πi ∈ u(1) and p∗∂2πi = 0.
Lemma 3.2.3. For all f, g ∈ C∞(M),
[ξf , ξg] = ξf,g −1
2π~p∗f, g∂2πi.
Proof. It suffices to show that
p∗[ξf , ξg] = p∗
(ξf,g −
1
2π~p∗f, g∂2πi
)and
γ([ξf , ξg]) = γ
(ξf,g −
1
2π~p∗f, g∂2πi
)
First, we have
p∗
(ξf,g −
1
2π~p∗f, g∂2πi
)= ξf,g = [ξf , ξg].
Chapter 3. Metaplectic-c Quantomorphisms 18
Since p∗ξf = ξf and p∗ξg = ξg, it follows that p∗[ξf , ξg] = [ξf , ξg]. Thus the first equation is
verified.
Next, note that
γ
(ξf,g −
1
2π~p∗f, g∂2πi
)=
1
i~p∗f, g,
and
γ([ξf , ξg]) = − 1
i~(p∗ω)(ξf , ξg) =
1
i~p∗f, g.
Therefore the second equation is also verified.
Lemma 3.2.4. The map E : C∞(M)→ X (Y ) given by
E(f) = ξf +1
2π~p∗f∂2πi, ∀f ∈ C∞(M)
is a Lie algebra homomorphism.
Proof. Let f, g ∈ C∞(M) be arbitrary. We need to show that
ξf,g +1
2π~p∗f, g∂2πi =
[ξf +
1
2π~p∗f∂2πi, ξg +
1
2π~p∗g∂2πi
].
Using Lemma 3.2.3, the left-hand side becomes
[ξf , ξg] + 21
2π~p∗f, g∂2πi.
Expanding the right-hand side yields
[ξf , ξg] +
[ξf ,
1
2π~p∗g∂2πi
]+
[1
2π~p∗f∂2πi, ξg
]+
[1
2π~p∗f∂2πi,
1
2π~p∗g∂2πi
].
The fourth term vanishes because ∂θ(p∗f) = ∂θ(p
∗g) = 0 for any θ ∈ u(1). To evaluate the
third term, recall that that [∂θ, ξ] = 0 for any θ ∈ u(1) and ξ ∈ X (M). Therefore [∂2πi, ξg] = 0,
so this term reduces to
− 1
2π~(ξgp
∗f)∂2πi =1
2π~p∗f, g∂2πi.
Chapter 3. Metaplectic-c Quantomorphisms 19
By the same argument, the second term also reduces to
1
2π~p∗f, g∂2πi.
Combining these results, we find that the right-hand side of the desired equation is
[ξf , ξg] + 21
i~p∗f, g∂2πi,
which equals the left-hand side.
Lemma 3.2.5. For all f ∈ C∞(M), E(f) ∈ Q(Y, γ).
Proof. We need to show that LE(f)γ = 0. We calculate
LE(f)γ = E(f)y dγ + d(E(f)y γ) =1
i~p∗(ξfyω)− 1
i~p∗df = 0.
So far, we have shown that E : C∞(M)→ Q(Y, γ) is a Lie algebra homomorphism. We will
now construct a map F : Q(Y, γ) → C∞(M), and show that E and F are inverses. This will
complete the proof that C∞(M) and Q(Y, γ) are isomorphic.
Let ζ ∈ Q(Y, γ) be arbitrary. Then Lζγ = ζy dγ+d(γ(ζ)) = 0. This implies that ∂θy (ζy dγ+
d(γ(ζ))) = 0 for any θ ∈ u(1). Since dγ(ζ, ∂θ) = 1i~(p∗ω)(ζ, ∂θ) = 0, it follows that ∂θy d(γ(ζ)) =
L∂θγ(ζ) = 0. We can therefore define the map F : Q(Y, γ)→ C∞(M) so that
− 1
i~p∗F (ζ) = γ(ζ), ∀ζ ∈ Q(Y, γ).
Theorem 3.2.6. The map E : C∞(M) → Q(Y, γ) is a Lie algebra isomorphism with inverse
F .
Proof. Let f ∈ C∞(M) and ζ ∈ Q(Y, γ) be arbitrary. We will show that F (E(f)) = f and
E(F (ζ)) = ζ. Using the definitions of E and F , we have
− 1
i~p∗F (E(f)) = γ(E(f)) = γ
(ξf +
1
2π~p∗f∂2πi
)= − 1
i~p∗f.
Chapter 3. Metaplectic-c Quantomorphisms 20
This implies that F (E(f)) = f .
To show that E(F (ζ)) = ζ, it suffices to show that γ(E(F (ζ))) = γ(ζ) and p∗E(F (ζ)) = p∗ζ.
By definition,
E(F (ζ)) = ξF (ζ) +1
2π~p∗F (ζ)∂2πi = ξF (ζ) +
1
2πiγ(ζ)∂2πi.
It is immediate that γ(E(F (ζ))) = γ(ζ), and that p∗E(F (ζ)) = ξF (ζ). Observe that
ζy p∗ω = i~ζy dγ = −i~d(γ(ζ)) = p∗(dF (ζ)),
having used Lζγ = 0. Therefore (p∗ζ)yω = dF (ζ), which implies that p∗ζ = ξF (ζ). Thus
p∗E(F (ζ)) = p∗ζ. This concludes the proof that E(F (ζ)) = ζ.
Since E and F are inverses, and we know from Lemma 3.2.5 that E : C∞(M)→ Q(Y, γ) is
a Lie algebra homomorphism, it follows that E and F are the desired Lie algebra isomorphisms.
The primary goal of Section 3.3 is to duplicate the above construction for the infinitesi-
mal quantomorphisms of a metaplectic-c prequantization. However, before moving on to the
metaplectic-c case, we will show how the map E can be used to represent the elements of
C∞(M) as operators on the space of sections of the prequantization line bundle for (M,ω).
This result will also have an analogue in the metaplectic-c case, which we will discuss in Section
3.3.3.
3.2.3 An operator representation of C∞(M)
Let (L,∇) be the complex line bundle with connection associated to (Y, γ). Recall the associ-
ation between a section s of L and an equivariant function s : Y → C from Section 2.1.2. We
note the following properties.
• For any f ∈ C∞(M) and s ∈ Γ(L), the equivariant function corresponding to the section
fs is f s = p∗fs.
• The vector field ∂2πi is generated by the infinitesimal action of 2πi ∈ u(1) on Y . Thus,
Chapter 3. Metaplectic-c Quantomorphisms 21
for all y ∈ Y ,
(∂2πis)(y) =d
dt
∣∣∣∣t=0
s(y · e2πit) = −2πis(y).
Further recall that the Kostant-Souriau representation map r : C∞(M) → End Γ(L) is
defined by
r(f)s =(i~∇ξf + f
)s, ∀f ∈ C∞(M), s ∈ Γ(L).
Using the preceding observations, we see that
r(f)s =(i~ξf + p∗f
)s =
(i~ξf −
1
2πip∗f∂2πi
)s = i~E(f)s.
Since we proved in Lemma 3.2.4 that E(f, g) = [E(f), E(g)] for all f, g ∈ C∞(M), the
following is immediate.
Theorem 3.2.7. The map r : C∞(M)→ End Γ(L) satisfies
[r(f), r(g)] = i~r(f, g), ∀f, g ∈ C∞(M).
Thus the same map that provides the isomorphism from C∞(M) to Q(Y, γ) also yields the
usual Kostant-Souriau representation of C∞(M) as a space of operators on Γ(L). We will see
a similar result in the case of metaplectic-c prequantization.
3.3 Metaplectic-c Quantomorphisms
Having reviewed the properties of infinitesimal quantomorphisms in Kostant-Souriau prequan-
tization, we will now explore their parallels in metaplectic-c prequantization. In Section 3.3.1,
we develop our definition for a metaplectic-c quantomorphism, and use it to define an infinites-
imal metaplectic-c quantomorphism. The remainder of the chapter is dedicated to proving the
metaplectic-c analogues of the results presented in Section 3.2.
Suppose (M,ω) admits a metaplectic-c prequantization (P,Σ, γ). The space of infinitesimal
quantomorphisms of (P,Σ, γ) consists of those vector fields on P whose flows preserve all of the
structures on (P,Σ, γ). Note that one of these structures is the map PΣ−→ Sp(M,ω), which
Chapter 3. Metaplectic-c Quantomorphisms 22
does not have a direct analogue in the Kostant-Souriau case. We will show how to incorporate
this additional piece of information in the next section.
3.3.1 Infinitesimal metaplectic-c quantomorphisms
As in Section 3.2.1, we begin by developing the idea of a quantomorphism between metaplectic-c
prequantizations. Let (P1,Σ1, γ1)Σ1−→ Sp(M1, ω1)
ρ1−→ (M1, ω1) and (P2,Σ2, γ2)Σ2−→ Sp(M2, ω2)
ρ2−→ (M2, ω2) be metaplectic-c prequantizations for two symplectic manifolds, and let Πj = ρj
Σj for j = 1, 2. Let K : P1 → P2 be a diffeomorphism. We will determine the conditions that K
must satisfy in order for it to preserve all of the structures of the metaplectic-c prequantizations.
First, by analogy with the Kostant-Souriau definition, assume that K satisfies K∗γ2 = γ1.
Fix m ∈M1, and consider the fiber P1m. For any q ∈ P1m, notice that
TqP1m = ξ ∈ TqP1 |Π1∗ξ = 0 = ker dγ1q.
The same property holds for a fiber of P2 over a point in M2. By assumption, K∗ is an
isomorphism from ker dγ1q to ker dγ2K(q) for all q ∈ P1. Therefore Π2 is constant on K(P1m).
Moreover, since K is a diffeomorphism, we can make the analogous argument with K−1, and
conclude that K(P1m) is in fact a fiber of P2 over M2, and every fiber of P2 is the image of a
fiber of P1. Thus K induces a diffeomorphism K ′′ : M1 →M2 such that the following diagram
commutes.
P1K //
Π1
P2
Π2
M1
K′′ //M2
Lemma 3.3.1. The map K ′′ : M1 →M2 is a symplectomorphism.
Proof. It suffices to show that K ′′∗ω2 = ω1. Using the properties of K, γ1 and γ2, we calculate
Π∗1(K ′′∗ω2) = (K ′′ Π1)∗ω2 = (Π2 K)∗ω2 = K∗(i~dγ2) = i~dγ1 = Π∗1ω1.
Therefore K ′′∗ω2 = ω1, as required.
Chapter 3. Metaplectic-c Quantomorphisms 23
Assume that a model symplectic vector space (V,Ω) has been fixed. Recall from Definition
2.3.1 that an element b ∈ Sp(M1, ω1)m is a map b : V → TmM1 such that b∗ω1m = Ω. Since K ′′
is a symplectomorphism, the composition K ′′∗ b : V → TK′′(m)M2 satisfies (K ′′∗ b)∗ω2K′′(m) = Ω,
which implies that K ′′∗ b ∈ Sp(M2, ω2)K′′(m). Let K ′′ : Sp(M1, ω1)→ Sp(M2, ω2) be the lift of
K ′′ given by
K ′′(b) = K ′′∗ b, ∀b ∈ Sp(M1, ω1).
Then K ′′ is a diffeomorphism, and it is equivariant with respect to the principal Sp(V ) actions.
Thus, if we assume that K∗γ2 = γ1, we obtain the diffeomorphisms K ′′ : M1 → M2 and
K ′′ : Sp(M1, ω1) → Sp(M2, ω2), where both K and K ′′ are lifts of K ′′. However, K is not
necessarily a lift of K ′′. Indeed, there might not be any map K ′ : Sp(M1, ω1)→ Sp(M2, ω2) of
which K is a lift. A map K for which there is no corresponding K ′ is constructed in Section
3.4, Example 3.4.1. In Section 3.2.1, we showed that a diffeomorphism of prequantization circle
bundles that preserves the connection forms must be equivariant with respect to the principal
circle actions. By contrast, Example 3.4.1 demonstrates that it is possible for K to preserve
the prequantization one-forms without being equivariant with respect to the principal Mpc(V )
actions.
Suppose we make the additional assumption that K(q · a) = K(q) · a for all q ∈ P1 and
a ∈ Mpc(V ). Then K induces a diffeomorphism K ′ : Sp(M1, ω1) → Sp(M2, ω2) that satisfies
K ′Σ1 = Σ2K. Combining this with the map K ′′ : M1 →M2 yields the following commutative
diagram.
P1K //
Σ1
P2
Σ2
Sp(M1, ω1)
K′ //
ρ1
Sp(M2, ω2)
ρ2
M1
K′′ //M2
We now have two maps, K ′ and K ′′, which are diffeomorphisms from Sp(M1, ω1) to Sp(M2, ω2).
By construction, ρ2 K ′ = ρ2 K ′′, and both K ′ and K ′′ are equivariant with respect to the
principal Sp(V ) actions. However, it is still possible for K ′ and K ′′ to be different. A map K
for which K ′ 6= K ′′ is given in Example 3.4.2.
Chapter 3. Metaplectic-c Quantomorphisms 24
As will be shown in Section 3.3.2, this potential discrepancy between K ′ and K ′′ must be
prevented in order to construct the desired isomorphism between C∞(M) and the infinitesimal
quantomorphisms. We therefore propose the following definition.
Definition 3.3.2. The diffeomorphism K : P1 → P2 is a metaplectic-c quantomorphism
if
(1) K∗γ2 = γ1,
(2) the induced diffeomorphism K ′′ : M1 →M2 satisfies K ′′ Σ1 = Σ2 K.
Let K : P1 → P2 be a metaplectic-c quantomorphism. Given our concept of a quantomor-
phism as a diffeomorphism that preserves all of the structures of a metaplectic-c prequantization,
we would expect that K is equivariant with respect to the Mpc(V ) actions. Equivariance is a
consequence of the definition, as we now show.
Let α ∈ mpc(V ) be arbitrary, and write α = κ⊕ τ under the identification of mpc(V ) with
sp(V ) ⊕ u(1). The vector field ∂α on P1 is completely specified by the conditions γ1(∂α) = τ
and Σ1∗∂α = ∂κ, and the same is true on P2. Notice that
γ2(K∗∂α) = γ1(∂α) = τ,
and
Σ2∗K∗∂α = K ′′∗Σ1∗∂α = K ′′∗∂κ = ∂κ,
where the final equality follows from the fact that K ′′ is equivariant with respect to Sp(V ).
Thus K∗∂α = ∂α for all α ∈ mpc(V ), which implies that K is equivariant with respect to
Mpc(V ), as desired.
Now consider a single metaplectic-c prequantized space (P,Σ, γ)Σ−→ Sp(M,ω)
ρ−→ (M,ω)
with Π = ρ Σ.
Definition 3.3.3. A vector field ζ ∈ X (P ) is an infinitesimal metaplectic-c quantomor-
phism if its flow φt is a metaplectic-c quantomorphism from its domain to its range for each
t.
Chapter 3. Metaplectic-c Quantomorphisms 25
Let ζ ∈ X (P ) have flow φt. Property (1) of a quantomorphism holds for φt if and only if
Lζγ = 0. If we assume that φt satisfies property (1), then there is a flow φ′′t on M such that
Π φt = φ′′t Π. The vector field that it generates on M is Π∗ζ.
Lemma 3.3.1 shows that φ′′t is a family of symplectomorphisms. Therefore we can lift φ′′t
to a flow on Sp(M,ω), denoted by φ′′t, where φ′′t(b) = (φ′′t)∗ b for all b ∈ Sp(M,ω). Let the
vector field on Sp(M,ω) that has flow φ′′t be Π∗ζ. Then property (2) of a quantomorphism
holds for φt if and only if Σ∗ζ is a well-defined vector field on Sp(M,ω) and Σ∗ζ = Π∗ζ.
We conclude that the space of infinitesimal metaplectic-c quantomorphisms of (P,Σ, γ) is
Q(P,Σ, γ) = ζ ∈ X (P )∣∣∣ Lζγ = 0 and Σ∗ζ = Π∗ζ,
where it is understood that the condition Σ∗ζ = Π∗ζ can only be satisfied if Σ∗ζ is well defined.
In the next section, we will construct a Lie algebra isomorphism from C∞(M) to Q(P,Σ, γ).
3.3.2 The Lie algebra isomorphism
We begin with a procedure, given by Robinson and Rawnsley in Section 7 of [15], for lifting
a Hamiltonian vector field on M to Sp(M,ω) and then to P . These steps will be used in
constructing the isomorphism E : C∞(M)→ Q(P,Σ, γ).
Fix f ∈ C∞(M), and let its Hamiltonian vector field ξf have flow ϕt on M . We know that
ϕt∗ preserves ω because Lξfω = 0. Let ϕt be the lift of ϕt to Sp(M,ω) given by
ϕt(b) = ϕt∗ b, ∀b ∈ Sp(M,ω),
and let the vector field on Sp(M,ω) with flow ϕt be ξf . We have ρ∗ξf = ξf by construction.
Also, ϕt commutes with the right principal Sp(V ) action on Sp(M,ω), so [ξf , ∂κ] = 0 for all
κ ∈ sp(V ). Now let ξf be the lift of ξf to P that is horizontal with respect to γ. Then Σ∗ξf = ξf
and γ(ξf ) = 0. A summary of the key properties of ξf , ξf and ξf is below.
Chapter 3. Metaplectic-c Quantomorphisms 26
(P, γ)
Σ
ξf γ(ξf ) = 0, Σ∗ξf = ξf , Π∗ξf = ξf
Sp(M,ω)
ρ
ξf [ξf , ∂κ] = 0 ∀κ ∈ sp(V ), ρ∗ξf = ξf
(M,ω) ξf
In Section 3.2.2, we made use of the vector field ∂2πi on Y . The corresponding object in
this context is the vector field ∂2πi on P , where 2πi ∈ u(1) ⊂ mpc(V ). This vector field satisfies
γ(∂2πi) = 2πi and Σ∗(∂2πi) = 0.
Lemma 3.3.4. For all f, g ∈ C∞(M),
[ξf , ξg] = ξf,g −1
2π~Π∗f, g∂2πi.
Proof. It suffices to show that
Σ∗[ξf , ξg] = Σ∗
(ξf,g −
1
2π~Π∗f, g∂2πi
)and
γ([ξf , ξg]) = γ
(ξf,g −
1
2π~Π∗f, g∂2πi
).
A calculation establishes that ξf,g = [ξf , ξg]. The rest of the proof proceeds identically to that
of Lemma 3.2.3.
Lemma 3.3.5. The map E : C∞(M)→ X (P ) given by
E(f) = ξf +1
2π~Π∗f∂2πi, ∀f ∈ C∞(M)
is a Lie algebra homomorphism.
Proof. Precisely analogous to Lemma 3.2.4.
Lemma 3.3.6. For all f ∈ C∞(M), E(f) ∈ Q(P,Σ, γ).
Proof. We need to show that LE(f)γ = 0 and Σ∗E(f) = Π∗E(f). The verification that LE(f)γ =
0 is the same as that in Lemma 3.2.5. Note that Σ∗E(f) = ξf and Π∗E(f) = ξf , so Π∗E(f) =
ξf = Σ∗E(f). Thus the necessary conditions are satisfied, and E(f) ∈ Q(P,Σ, γ).
Chapter 3. Metaplectic-c Quantomorphisms 27
As before, we will construct an inverse for E, and conclude that E is a Lie algebra iso-
morphism. Let ζ ∈ Q(P,Σ, γ) and α ∈ mpc(V ) be arbitrary. Using an identical argument to
the one that precedes Theorem 3.2.6, the fact that ∂αy (Lζγ) = 0 implies that L∂αγ(ζ) = 0.
Therefore we can define F : Q(P,Σ, γ)→ C∞(M) so that
− 1
i~Π∗F (ζ) = γ(ζ), ∀ζ ∈ Q(P,Σ, γ).
Theorem 3.3.7. The map E : C∞(M)→ Q(P,Σ, γ) is a Lie algebra isomorphism with inverse
F .
Proof. Let f ∈ C∞(M) and ζ ∈ Q(P,Σ, γ) be arbitrary. From the definitions of E and F ,
F (E(f)) satisfies
− 1
i~Π∗F (E(f)) = γ(E(f)) = γ
(ξf +
1
2π~Π∗f∂2πi
)= − 1
i~Π∗f.
Thus F (E(f)) = f .
Next, we claim that γ(E(F (ζ))) = γ(ζ) and Σ∗E(F (ζ)) = Σ∗ζ. Observe that
E(F (ζ)) = ξF (ζ) +1
2π~Π∗F (ζ)∂2πi = ξF (ζ) +
1
2πiγ(ζ)∂2πi.
It is immediate that γ(E(F (ζ))) = γ(ζ) and Σ∗E(F (ζ)) = ξF (ζ). From the definition of
Q(P,Σ, γ), we know that Σ∗ζ = Π∗ζ. It remains to show that Π∗ζ = ξF (ζ). We calculate
ζyΠ∗ω = ζy i~dγ = −i~d(γ(ζ)) = Π∗dF (ζ).
This demonstrates that (Π∗ζ)yω = dF (ζ), which implies that Π∗ζ = ξF (ζ) as needed. Thus we
have shown that E(F (ζ)) = ζ, and this completes the proof that E and F are inverses.
If the definition of Q(P,Σ, γ) did not include the condition that Σ∗ζ = Π∗ζ, this proof
would fail in the final step. We would be able to show that Σ∗E(F (ζ)) = ξF (ζ) = Π∗ζ, but
this vector field would not necessarily equal Σ∗ζ, and so F would not be the inverse of E. This
explains why property (2) of a metaplectic-c quantomorphism is necessary in order to obtain a
subalgebra of X (P ) that is isomorphic to C∞(M).
Chapter 3. Metaplectic-c Quantomorphisms 28
3.3.3 An operator representation of C∞(M)
In [15], Robinson and Rawnsley construct an infinite-dimensional Hilbert space E ′(V ) of holo-
morphic functions on V ∼= Cn, on which the group Mpc(V ) acts via the metaplectic repre-
sentation. They then define the bundle of symplectic spinors for the prequantized system
(P,Σ, γ)Π−→ (M,ω) to be
E ′(P ) = P ×Mpc(V ) E ′(V ).
We omit the details of the metaplectic representation here; the only fact we need is that the
subgroup U(1) ⊂ Mpc(V ) acts on elements of E ′(V ) by scalar multiplication. We write an
element of E ′(P ) as an equivalence class [q, ψ] with q ∈ P and ψ ∈ E ′(V ).
Section 7 of [15] contains the following construction. Let s ∈ Γ(E ′(P )) be given, and define
the map s : P → E ′(V ) so that [q, s(q)] = s(Π(q)) for all q ∈ P . This map s satisfies the
equivariance condition
s(q · a) = a−1s(q), ∀q ∈ P, a ∈ Mpc(V ),
where the action on the right-hand side is that of the metaplectic representation. Conversely,
if s : P → E ′(V ) is any map with the equivariance property above, it can be used to define
a section s ∈ Γ(E ′(P )) by setting s(m) = [q, s(q)] for each m ∈ M and any q ∈ P such that
Π(q) = m.
Let f ∈ C∞(M) be arbitrary, and recall the lifting ξf → ξf → ξf of ξf to P . A standard
calculation establishes that [ξf , ∂α] = 0 for all α ∈ mpc(V ). Thus, if s : P → E ′(V ) is an
equivariant map, then so is ξf s. Define the map D : C∞(M)→ End Γ(E ′(P )) such that for all
f ∈ C∞(M) and s ∈ Γ(E ′(P )), Dfs is the section of E ′(P ) that satisfies
Dfs = ξf s.
Further, define δ : C∞(M)→ End Γ(E ′(P )) by
δfs = Dfs+1
i~fs, ∀f ∈ C∞(M), s ∈ Γ(E ′(P )).
Chapter 3. Metaplectic-c Quantomorphisms 29
Theorem 7.8 of [15] states that δ is a Lie algebra homomorphism.
We see that the construction of D precisely parallels the construction of the connection ∇
on the prequantization line bundle L associated to a prequantization circle bundle (Y, γ). As
in Section 3.2.3, we make two observations.
• For any s ∈ Γ(E ′(P )) and f ∈ C∞(M), f s = Π∗fs.
• For any equivariant map s : P → E ′(V ), ∂2πis = −2πis.
Therefore
δfs =
(ξf +
1
2π~Π∗f∂2πi
)s = E(f)s.
The fact that δ is a Lie algebra homomorphism then follows immediately from Lemma 3.3.5.
This construction would apply equally well to any associated bundle where the subgroup U(1) ⊂
Mpc(V ) acts on the fiber by scalar multiplication.
3.4 Two diffeomorphisms that are not quantomorphisms
A metaplectic-c quantomorphismK between two metaplectic-c prequantizations (P1,Σ1, γ1)Σ1−→
Sp(M1, ω1)ρ1−→ (M1, ω1) and (P2,Σ2, γ2)
Σ2−→ Sp(M2, ω2)ρ2−→ (M2, ω2) is a diffeomorphism
K : P1 → P2 such that
(1) K∗γ2 = γ1,
(2) the induced diffeomorphism K ′′ : M1 →M2 satisfies K ′′ Σ1 = Σ2 K,
where K ′′ : Sp(M1, ω1)→ Sp(M2, ω2) is the lift of K ′′ given by
K ′′(b) = K ′′∗ b, ∀b ∈ Sp(M1, ω1).
We claimed that condition (1) is insufficient to guarantee that K is the lift of some map
K ′ : Sp(M1, ω1)→ Sp(M2, ω2). In particular, a diffeomorphism K that only satisfies condition
(1) might not be equivariant with respect to the principal Mpc(V ) actions. We further claimed
that K might be equivariant and satisfy condition (1), yet fail to satisfy condition (2). We will
now construct examples to support these claims.
Chapter 3. Metaplectic-c Quantomorphisms 30
Let M = R2 \ (0, 0) with Cartesian coordinates (p, q) and polar coordinates (r, θ). Equip
M with the symplectic form ω = dp∧ dq = rdr ∧ dθ, and observe that the one-form β = 12r
2dθ
satisfies dβ = ω. Let V = R2 with basis x, y and symplectic form Ω = x∗ ∧ y∗, and consider
the global trivialization of the tangent bundle TM such that for all m ∈M , TmM is identified
with V by mapping x → ∂∂p
∣∣∣m
and y → ∂∂q
∣∣∣m
. Identify Sp(M,ω) with M × Sp(V ) using this
trivialization.
Let P = M ×Mpc(V ), and define the map Σ : P → Sp(M,ω) by Σ(m, a) = (m,σ(a)) for all
m ∈M and a ∈ Mpc(V ). Let ϑ0 be the trivial connection on the product bundle M ×Mpc(V ),
and let γ = 1i~β + 1
2η∗ϑ0. Then (P, γ) is a metaplectic-c prequantization of (M,ω). In both of
the examples below, we will give a diffeomorphism K : P → P .
To facilitate the construction in Example 3.4.1, we introduce a more explicit representation
for elements of the metaplectic-c group. By definition, Mpc(V ) = Mp(V ) ×Z2 U(1). The
restriction of the projection map Mpc(V )σ−→ Sp(V ) to Mp(V ) yields the double covering
Mp(V )σ−→ Sp(V ). Write an element of Mpc(V ) as an equivalence class [h, e2πit] with h ∈
Mp(V ) and t ∈ R. In terms of this parametrization, the projection map is given by
σ[h, e2πit] = σ(h),
and the determinant map Mpc(V )η−→ U(1) is given by
η[h, e2πit] = (e2πit)2.
Example 3.4.1. We will define a diffeomorphism K : P → P that preserves γ, but that does
not descend through Σ to a well-defined map on Sp(M,ω).
Let µ : R → Mp(V ) be any smooth nonconstant path such that µ(t + 1) = µ(t) for all
t ∈ R. Note that the composition σ µ : R → Sp(V ) is also nonconstant. Now define
F : Mp(V )× U(1)→ Mp(V )× U(1) by
F (h, e2πit) = (hµ(2t), e2πit), ∀h ∈ Mp(V ), t ∈ R.
This map is a diffeomorphism of Mp(V )×U(1), and it descends to a diffeomorphism of Mpc(V ),
Chapter 3. Metaplectic-c Quantomorphisms 31
which we also denote F . For any [h, e2πit] ∈ Mpc(V ), observe that
η(F [h, e2πit]) = η[hµ(2t), e2πit] = (e2πit)2 = η[h, e2πit].
This implies that for any α ∈ mpc(V ), 12η∗F∗α = 1
2η∗α.
Define the diffeomorphism K : P → P by K(m, a) = (m,F (a)) for all m ∈ M and a ∈
Mpc(V ). Since K is the identity on M , it preserves β. From the property of F shown above,
K also preserves 12η∗ϑ0, and thus it preserves γ. Fix (m, g) ∈ Sp(M,ω), and let h ∈ Mp(V ) be
such that σ(h) = g. Then the fiber of P over (m, g) is P(m,g) = (m, [h, e2πit]) | t ∈ R. Notice
that
Σ K(m, [h, e2πit])) = Σ(m, [hµ(2t), e2πit]) = (m, gσ(µ(2t))),
which is not constant with respect to t. Thus K(P(m,g)) is not contained within a single fiber
of P over Sp(M,ω), which shows that there is no map K ′ : Sp(M,ω) → Sp(M,ω) such that
K ′ Σ = Σ K.
If K : P → P is equivariant with respect to Mpc(V ), then it induces a diffeomorphism
K ′ : Sp(M,ω) → Sp(M,ω) that satisfies K ′ Σ = Σ K. This map and K ′′ are both lifts of
K ′′ : M →M , but they might not be the same map.
Example 3.4.2. We will define a diffeomorphismK : P → P that preserves γ and is equivariant
with respect to Mpc(V ), but where K ′ 6= K ′′.
Let Tλ : M → M be the map that rotates M about the origin by the angle λ, where λ is
not an integer multiple of 2π. Define K : P → P by
K(m, a) = (Tλ(m), a), ∀m ∈M, a ∈ Mpc(V ).
Then K∗γ = γ, and K(q ·a) = K(q)·a for all q ∈ P and a ∈ Mpc(V ). The map K ′ : Sp(M,ω)→
Sp(M,ω) is given by
K ′(m, g) = (Tλ(m), g), ∀m ∈M, g ∈ Sp(V ),
Chapter 3. Metaplectic-c Quantomorphisms 32
and the map K ′′ : M →M is simply Tλ. If we let Tλ also denote the automorphism of V given
by rotation about the origin by λ, then under our chosen identification of TM with M ×V , we
have
K ′′∗ (m, v) = (Tλ(m), Tλ(v)), ∀m ∈M, v ∈ V.
Therefore K ′′ : Sp(M,ω)→ Sp(M,ω) is given by
K ′′(m, g) = (Tλ(m), Tλ g), ∀m ∈M, g ∈ Sp(V ).
Hence K ′ 6= K ′′.
Chapter 4
Quantized Energy Levels and
Dynamical Invariance
4.1 Introduction
It is well known from physics that the quantum mechanical versions of systems such as the har-
monic oscillator and the hydrogen atom are restricted to discrete energy levels. More generally,
suppose (M,ω) is a quantizable symplectic manifold and H : M → R is a function on M . If
we think of H as a Hamiltonian energy function, then we can ask what it means for a regular
value E of H to be a quantized energy level for the system (M,ω,H).
One possible answer to this question involves the construction of the symplectic reduction.
The orbits of the Hamiltonian vector field ξH partition the level set H−1(E). If ME , the space
of orbits, is a manifold, then ω induces a symplectic form ωE on ME , and (ME , ωE) is the
symplectic reduction of (M,ω) at E. The values of E for which this new symplectic manifold
is quantizable can be taken to be the quantized energy levels of the system. This definition has
been applied to the hydrogen atom in the context of Kostant-Souriau quantization [16], and
to the harmonic oscillator in the context of metaplectic-c quantization [15]. In both cases, the
physically predicted energy spectrum is obtained. However, this definition is limited to cases
where the space of orbits is a manifold. It would be more convenient to have a quantized energy
condition that is evaluated over the original manifold, rather than the quotient.
33
Chapter 4. Quantized Energy Levels and Dynamical Invariance 34
In this chapter, we propose a definition for a quantized energy level of (M,ω,H) in the case
where (M,ω) admits a metaplectic-c prequantization. Given one additional assumption, which
we describe in Section 4.2.2, if E is a quantized energy level of (M,ω,H), then the symplectic
reduction (ME , ωE) is metaplectic-c quantizable, provided that it is a manifold. The quantized
energy condition is evaluated on a bundle over the level set H−1(E).
This scenario was first studied by Robinson [14], and our work is strongly motivated by his
results. However, we choose a more flexible condition than the one Robinson considered. As we
will show, if H1 and H2 are two functions on (M,ω) with regular values E1 and E2, respectively,
such that H−11 (E1) = H−1
2 (E2), then under our definition, E1 is a quantized energy level of
(M,ω,H1) if and only if E2 is a quantized energy level of (M,ω,H2). In other words, the
quantization condition only depends on the geometry of the level set, and not on the dynamics
of a particular Hamiltonian. This is not true of the definition considered in [14].
Section 4.2 summarizes Robinson’s results concerning symplectic reduction, and concludes
with our proposed definition for a quantized energy level of a metaplectic-c quantizable system.
The proof of the dynamical invariance of our definition is given in Section 4.3. In Section 4.4,
we consider the example of the harmonic oscillator. This example serves two purposes. First,
we demonstrate a computational technique in which a local change of coordinates on M is lifted
to the metaplectic-c prequantization. Second, we state an example of a function on M that has
exactly one level set in common with the harmonic oscillator, and show that the quantization
condition on that level set is identical for both functions.
Finally, in Section 4.5, we adapt the quantized energy definition to apply to a Kostant-
Souriau quantizable symplectic manifold. We show that the Kostant-Souriau and metaplectic-c
definitions have almost identical properties, but the metaplectic-c version is the only one that
yields the correct quantized energy levels for the harmonic oscillator.
4.2 Quantized Energy Levels
Given a symplectic manifold (M,ω) and a smooth function H : M → R, the symplectic
reduction of (M,ω) at the regular value E of H is constructed by taking the quotient of the
level set H−1(E) by the orbits of the Hamiltonian vector field ξH . Suppose (M,ω) admits
Chapter 4. Quantized Energy Levels and Dynamical Invariance 35
a metaplectic-c prequantization (P,Σ, γ). As discussed in Section 3.3.2, there is a natural
lift of ξH to a vector field ξH on Sp(M,ω); ξH can then be lifted to a vector field ξH on P
that is horizontal with respect to γ. Robinson [14] constructed a certain associated bundle
PS → H−1(E), on which ξH induces a vector field, and he examined the conditions under
which the quotient of PS by the induced orbits of ξH yields a metaplectic-c prequantization for
the symplectic reduction of (M,ω) at E.
In this section, we review the key results from [14], then state our alternative definition for
a quantized energy level of the system (M,ω,H). Our definition is evaluated on the bundle PS ,
and ensures that the symplectic reduction acquires a metaplectic-c prequantization under the
same conditions as Robinson’s definition. Further, as we will show in Section 4, our definition
depends only on the geometry of the level set H−1(E), and not on the specific choice of H.
This is an improvement over the definition from [14].
4.2.1 Vector subspaces and quotients
Assume that the model vector space V has been chosen as in Section 2.3.1. Let W ⊂ V be a
real subspace of codimension 1. Define the symplectic orthogonal of W to be
W⊥ = v ∈ V : Ω(v,W ) = 0 .
Then W⊥ is a one-dimensional subspace of W . The form Ω induces a symplectic form on the
quotient space W/W⊥. This new symplectic vector space has symplectic group Sp(W/W⊥) and
metaplectic-c group Mpc(W/W⊥). The complex structure J on V induces a complex structure
on W/W⊥ in such a way that
W/W⊥ ∼= (W⊥ ⊕ JW⊥)⊥, (4.2.1)
where the isomorphism respects both the symplectic structures and the complex structures.
We write an element of W/W⊥ as an equivalence class [w] for some w ∈W .
Let Sp(V ;W ) be the subgroup of Sp(V ) consisting of those symplectic automorphisms that
Chapter 4. Quantized Energy Levels and Dynamical Invariance 36
preserve W :
Sp(V ;W ) = g ∈ Sp(V ) : gW = W .
Then elements of Sp(V ;W ) also preserve W⊥, and there is a group homomorphism
Sp(V ;W )ν−→ Sp(W/W⊥),
defined by
(νg)[w] = [gw], ∀g ∈ Sp(V ;W ), ∀w ∈W.
Let Mpc(V ;W ) ⊂ Mpc(V ) be the preimage of Sp(V ;W ) under the projection map σ.
Robinson and Rawnsley [15] constructed a group homomorphism Mpc(V ;W )ν−→ Mpc(W/W⊥)
such that the diagram below commutes.
Mpc(V ) ⊃ Mpc(V ;W )ν //
σ
Mpc(W/W⊥)
σ
Sp(V ) ⊃ Sp(V ;W )ν // Sp(W/W⊥)
The map ν has the following property. Given a ∈ Mpc(V ;W ), let χ(a) = DetR(σ(a)|W⊥).
Then χ is a real-valued character on Mpc(V ;W ), and it satisfies
η ν(a) = η(a)signχ(a), ∀a ∈ Mpc(V ;W ).
Since signχ(a) is identically 1 on a neighborhood of I, we see that η∗ ν∗ = η∗ as maps between
Lie algebras.
4.2.2 Quotients and reduction over a manifold
Let (M,ω) be a symplectic manifold that admits a metaplectic-c prequantization (P,Σ, γ). The
definitions from Section 2.3.2 yield the three-level structure
(P, γ)Σ−→ Sp(M,ω)
ρ−→ (M,ω), ρ Σ = Π.
Let H : M → R be a smooth function on M . We define the Hamiltonian vector field ξH
Chapter 4. Quantized Energy Levels and Dynamical Invariance 37
using the convention that
ξHyω = dH.
Let the flow of ξH be φt, and note that φt is a symplectomorphism from its domain to its range
for each t. The lift of φt to a flow φt on Sp(M,ω) is given by
φt(b) = φt∗ b, ∀b ∈ Sp(M,ω).
Let ξH be the vector field on Sp(M,ω) whose flow is φt. Define the vector field ξH on P to be
the lift of ξH that is horizontal with respect to the one-form γ, and let the flow of ξH be φt.
We now restrict our attention to a particular level set. Let E ∈ R be a regular value of H,
and let S = H−1(E). The null foliation TS⊥ is defined fiberwise over S by
TsS⊥ = ζ ∈ TsM : ω(ζ, TsS) = 0 , ∀s ∈ S.
Then TsS⊥ is a one-dimensional subspace of the tangent space TsS, and TsS
⊥ = span ξH(s).
The constructions that follow are taken from [14]. As in Section 4.2.1, fix a subspace W ⊂ V
of codimension 1. By definition, an element b ∈ Sp(M,ω)m is a symplectic linear isomorphism
b : V → TmM . Let Sp(M,ω;S) be the subset of Sp(M,ω)|S given by
Sp(M,ω;S)s = b ∈ Sp(M,ω)s : bW = TsS , ∀s ∈ S.
Then Sp(M,ω;S) is a principal Sp(V ;W ) bundle over S. Note that any b ∈ Sp(M,ω;S)s maps
W⊥ to TsS⊥.
Viewing P as a circle bundle over Sp(M,ω), let PS be the result of restricting P to
Sp(M,ω;S). Then PS is a principal Mpc(V ;W ) bundle over S. Let γS be the pullback of
γ to PS . After all of these restrictions, we have constructed the new three-level structure
(PS , γS)→ Sp(M,ω;S)→ S.
For each s ∈ S, TsS/TsS⊥ is a symplectic vector space with symplectic structure induced
Chapter 4. Quantized Energy Levels and Dynamical Invariance 38
by ωs. Let Sp(TS/TS⊥) be the symplectic frame bundle for TS/TS⊥, modeled on W/W⊥:
Sp(TS/TS⊥)s =b′ : W/W⊥ → TsS/TsS
⊥ : b′ is a symplectic linear isomorphism, ∀s ∈ S.
Recall the group homomorphism ν : Sp(V ;W ) → Sp(W/W⊥) from Section 4.2.1. The bundle
associated to Sp(M,ω;S) by the map ν can be naturally identified with Sp(TS/TS⊥).
Next, let PS → S be the bundle associated to PS → S by the group homomorphism
ν : Mpc(V ;W ) → Mpc(W/W⊥). Then PS is a principal Mpc(W/W⊥) bundle over S, and a
circle bundle over Sp(TS/TS⊥). The one-form γS induces a connection one-form γS on PS .
This completes the construction of another three-level structure,
(PS , γS)→ Sp(TS/TS⊥)→ S.
Now let us consider the actions of φt, φt and φt on all of these bundles. First, it is clear that
φt preserves S, since ξH(s) ∈ TsS at each s ∈ S. Therefore φt restricts to a flow on Sp(M,ω;S),
and so φt restricts to a flow on PS . One can verify that φt and φt are equivariant with respect
to the principal Sp(V ) and Mpc(V ) actions, respectively, which implies that φt induces a flow
on the associated bundle Sp(TS/TS⊥), and φt induces a flow on PS . We let φt and φt also
denote the respective flows induced on Sp(TS/TS⊥) and PS .
Suppose the null foliation is fibrating, so that the quotient of S by the orbits of ξH is a
manifold. Let the quotient manifold be ME , and let ωE be the symplectic form on ME induced
by ω. As discussed in [14], if the quotient of Sp(TS/TS⊥) by φt is well defined, then it is
naturally isomorphic to the symplectic frame bundle Sp(ME , ωE). To ensure that this quotient
is well defined, it is sufficient to require that if s ∈ S is fixed by φt, then φt∗ is the identity on
TsS. The culmination of all of these steps is shown below.
(P, γ)
Σ
(PS , γS)incl.oo
ν // (PS , γS)
Sp(M,ω)
ρ
Sp(M,ω;S)incl.oo
ν // Sp(TS/TS⊥)
/φt // Sp(ME , ωE)
(M,ω) S
incl.oo = // S/φt // (ME , ωE)
Chapter 4. Quantized Energy Levels and Dynamical Invariance 39
The obvious way to complete the picture is to factor (PS , γS) by φt. Robinson [14] addressed
the question of when this yields a metaplectic-c prequantization for (ME , ωE) with the following
theorem.
Theorem 4.2.1. (Robinson, 1990) Assume that the symplectic reduction (ME , ωE) is a man-
ifold, and that its symplectic frame bundle Sp(ME , ωE) can be identified with the quotient of
Sp(TS/TS⊥) by the induced flow φt. If γS has trivial holonomy over all of the closed orbits of
φt on Sp(TS/TS⊥), then the quotient of (PS , γS) by φt is a metaplectic-c prequantization for
(ME , ωE).
To satisfy the holonomy condition in Theorem 4.2.1, it is sufficient – though not necessary
– to show that γS has trivial holonomy over all closed orbits of φt on Sp(M,ω;S). This is the
quantization condition that was explored in the remainder of [14]. However, we argue that a
more robust condition arises when we evaluate the holonomy over orbits in Sp(TS/TS⊥). As
such, we propose the following definition.
Definition 4.2.2. When γS has trivial holonomy over all closed orbits of φt on Sp(TS/TS⊥),
we say that E is a quantized energy level of the system (M,ω,H).
In those cases where E is a quantized energy level and (ME , ωE) exists, Theorem 4.2.1
gives a sufficient condition for (ME , ωE) to admit a metaplectic-c prequantization. However,
Definition 4.2.2 can also be applied to systems where the symplectic reduction does not exist.
For the remainder of the chapter, we do not make any assumptions about (ME , ωE).
The next section contains the proof of the dynamical invariance property. We will show that
if S is a regular level set of two functions H1 and H2 on M , then S corresponds to a quantized
energy level for (M,ω,H1) if and only if it does so for (M,ω,H2).
4.3 Dynamical Invariance
We continue to use all of the definitions established in Section 4.2. To prove that the quantized
energy condition is dynamically invariant, we proceed in two stages. In Section 4.3.1, we prove
Theorem 4.3.2, which states that if E is a quantized energy level for the system (M,ω,H),
then for any diffeomorphism f : R → R, the value f(E) is a quantized energy level for the
Chapter 4. Quantized Energy Levels and Dynamical Invariance 40
system (M,ω, f H). While this is a less general statement, it allows us to establish some
useful preliminary observations. Then, in Section 4.3.2, we prove Theorem 4.3.5, which states
that if the functions H1, H2 : M → R have regular values E1 and E2, respectively, such that
H−11 (E1) = H−1
2 (E2), then E1 is a quantized energy level for (M,ω,H1) if and only if E2 is a
quantized energy level for (M,ω,H2).
4.3.1 Invariance Under a Diffeomorphism
Let H : M → R be given, and fix a regular value E of H. Let f : R→ R be a diffeomorphism,
and let S = H−1(E) = (f H)−1(f(E)).
Denote the Hamiltonian vector field for H by ξH , and let it have flow φt on M . Then ξH
lifts to ξH on Sp(M,ω) and to ξH on P , with flows φt and φt, respectively, as in Section 4.2.2.
Similarly, denote the Hamiltonian vector field for f H by ξfH , and let it have flow ρt on M .
By the same process, we obtain the lifts ξfH on Sp(M,ω) and ξfH on P , with flows ρt and ρt,
respectively.
Observe that for any m ∈M ,
d(f H)|m =df(H)
dH
∣∣∣∣H(m)
dH|m,
which implies that
ξfH(m) =df(H)
dH
∣∣∣∣H(m)
ξH(m).
Since S is a level set of H,df(H)
dH
∣∣∣∣H(m)
is constant over S. Let c ∈ R be that constant value
on S. Then ξfH = cξH everywhere on S, which implies that ρt = φct everywhere on S.
Recall that Sp(TS/TS⊥) can be naturally identified with the bundle associated to Sp(M,ω;S)
by the group homomorphism ν : Sp(V ;W ) → Sp(W/W⊥). We write an element of W/W⊥ as
an equivalence class [w] for some w ∈ W , and we write an element of TsS/TsS⊥ as an equiva-
lence class [ζ] for some ζ ∈ TsS. As discussed in Section 4.2.2, the lifted flow φt on Sp(M,ω)
maps Sp(M,ω;S) to itself and induces a flow φt on Sp(TS/TS⊥). More explicitly, for any
b′ ∈ Sp(TS/TS⊥), we define φt(b′) by choosing b ∈ Sp(M,ω;S) such that [bw] = b′[w] for all
Chapter 4. Quantized Energy Levels and Dynamical Invariance 41
w ∈W , and setting
φt(b′)[w] = [φt(b)w], ∀w ∈W.
These remarks apply equally well to ρt.
Lemma 4.3.1. ρt = φct on Sp(TS/TS⊥).
Proof. Fix s ∈ S and b′ ∈ Sp(TS/TS⊥)s. Let b ∈ Sp(M,ω;S)s be such that for all w ∈ W ,
[bw] = b′[w]. Since ρt = φct on S, we have ρt∗|TS = φct∗ |TS. Therefore, for any w ∈W ,
ρt(b)w = ρt∗|s(bw) = φct∗ |s(bw) = φct(b)w,
where we used the fact that bw ∈ TsS. Using the definitions of φt and ρt on Sp(TS/TS⊥), it
follows that
ρt(b′)[w] = [ρt(b)w] = [φct(b)w] = φct(b′)[w].
Thus ρt = φct on Sp(TS/TS⊥).
Lemma 4.3.1 implies that ξfH = cξH on Sp(TS/TS⊥). Since ξfH is just a constant multiple
of ξH , it is clear that the flows of both vector fields have the same orbits. Thus, if γS has trivial
holonomy over all closed orbits of φt on Sp(TS/TS⊥), then it also has trivial holonomy over
all closed orbits of ρt. The following is now immediate.
Theorem 4.3.2. If E is a quantized energy level for (M,ω,H) and f : R→ R is a diffeomor-
phism, then f(E) is a quantized energy level for (M,ω, f H).
4.3.2 Regular Level Set of Multiple Functions
Let H1, H2 : M → R be two smooth functions, and assume that there are E1, E2 ∈ R such that
Ej is a regular value of Hj for j = 1, 2, and H−11 (E1) = H−1
2 (E2). Let this shared level set be
S. Let the Hamiltonian vector fields ξH1 and ξH2 on M have flows φt and ρt, respectively. The
induced flows φt and ρt on Sp(TS/TS⊥) are defined as in Section 4.3.1.
The key observation in this scenario is that TS and TS⊥ do not depend on H1 or H2. For
all s ∈ S, we have
TsS⊥ = span ξH1(s) = span ξH2(s) ,
Chapter 4. Quantized Energy Levels and Dynamical Invariance 42
which implies that there is a map c : S → R such that ξH2(s) = c(s)ξH1(s) for all s ∈ S. The
map c need not be constant on S, and ξH1 and ξH2 need not be parallel away from S.
In Lemma 4.3.1, we were able to establish a relationship between ρt and φt. In this case,
we will not be able to relate ρt and φt directly, but must instead work in terms of their time
derivatives.
For any s ∈ S, ρt∗|s and φt∗|s preserve TS. Therefore, for all ζ ∈ TsS, the tangent vectors
ddt
∣∣t=0
(ρt∗|sζ
)and d
dt
∣∣t=0
(φt∗|sζ
)are elements of TζTS. Note that the vector space TsS can
be naturally identified with its tangent space TζTsS, which is a subspace of TζTS. Thus
ξH1(s) ∈ TsS can also be viewed as an element of TζTS. This identification is used in the
statement of the following result.
Lemma 4.3.3. For all s ∈ S and all ζ ∈ TsS,
d
dt
∣∣∣∣t=0
(ρt∗|sζ
)= c(s)
d
dt
∣∣∣∣t=0
(φt∗|sζ
)+ (ζc)ξH1(s).
Proof. We ignore M , and treat S as a (2n−1)-dimensional manifold. The vector fields ξH1 and
ξH2 on S have flows φt and ρt, respectively, and satisfy ξH2 = cξH1 .
Fix s ∈ S, and let U be a coordinate neighborhood for s with coordinatesX = (X1, . . . , X2n−1).
We write φt = (φt1, . . . , φt2n−1) and ρt = (ρt1, . . . , ρ
t2n−1) with respect to the local coordinates.
Then φt∗|s becomes a (2n− 1)× (2n− 1) matrix with the (j, k)th entry given by
(φt∗|s
)jk
=∂φtj(X)
∂Xk
∣∣∣∣∣X=s
,
and similarly for ρt∗|s. Also, since ξH2 = cξH1 , we have
d
dt
∣∣∣∣t=0
ρtj = cd
dt
∣∣∣∣t=0
φtj , j = 1, . . . , 2n− 1,
everywhere on U .
Let ζ ∈ TsS be arbitrary. Then ζ =∑2n−1
k=1 ak∂
∂Xkfor some coefficients a1, . . . , a2n−1 ∈ R,
and (φt∗|sζ
)j
=2n−1∑k=1
ak∂φtj(X)
∂Xk
∣∣∣∣∣X=s
= ζφtj .
Chapter 4. Quantized Energy Levels and Dynamical Invariance 43
The same process establishes that (ρt∗|sζ)j = ζρtj . When we take the time derivative of the
latter equation, we find
d
dt
∣∣∣∣t=0
(ρt∗|sζ
)j
= ζ
(d
dt
∣∣∣∣t=0
ρtj
)= ζ
(cd
dt
∣∣∣∣t=0
φtj
).
Now we apply the Leibniz rule and recall that ddt
∣∣t=0
φtj(s) = ξH1(s)j . The result is
d
dt
∣∣∣∣t=0
(ρt∗|sζ
)j
= c(s)d
dt
∣∣∣∣t=0
(φt∗|sζ)j + (ζc)ξH1(s)j . (4.3.1)
The coordinates on U naturally generate coordinates for TS|U , which in turn provide a
basis for TζTS. In terms of that basis,
d
dt
∣∣∣∣t=0
(ρt∗|sζ
)=
(ξH2(s)1, . . . , ξH2(s)2n−1,
d
dt
∣∣∣∣t=0
(ρt∗|sζ
)1, . . . ,
d
dt
∣∣∣∣t=0
(ρt∗|sζ
)2n−1
),
with the analogous expression for ddt
∣∣t=0
(φt∗|sζ
), and
ξH1(s) = (0, . . . , 0, ξH1(s)1, . . . , ξH1(s)2n−1)
as an element of TζTS. Substituting Equation (4.3.1) into the expression for ddt
∣∣t=0
(ρt∗|sζ
)shows that
d
dt
∣∣∣∣t=0
(ρt∗|sζ
)= c(s)
d
dt
∣∣∣∣t=0
(φt∗|sζ
)+ (ζc)ξH1(s),
as desired.
Lemma 4.3.4. For all s ∈ S and all b′ ∈ Sp(TS/TS⊥)s,
ξH2(b′) = c(s)ξH1(b′).
Proof. Within the model vector space V , fix a symplectic basis (x1, . . . , xn, y1, . . . , yn). Let
W = span x1, . . . , xn, y2, . . . , yn ,
Chapter 4. Quantized Energy Levels and Dynamical Invariance 44
so that
W⊥ = span x1 and W/W⊥ = span [x2], . . . , [xn], [y2], . . . , [yn] .
For convenience, let (z1, . . . , z2n) = (x1, . . . , xn, y2, . . . , yn, y1). Then
W = span z1, . . . , z2n−1 , W⊥ = span z1 , W/W⊥ = span [z2], . . . , [z2n−1] .
For this proof, we will not view an element of the symplectic frame bundle as a map, but as
a basis for a tangent space. Explicitly, over m ∈M , we identify b ∈ Sp(M,ω)m with the ordered
2n-tuple (ζ1, . . . , ζ2n) ∈ (TmM)2n, where bzj = ζj for j = 1, . . . , 2n. Similarly, over s ∈ S, we
identify b′ ∈ Sp(TS/TS⊥)s with the ordered (2n− 2)-tuple ([ζ2], . . . , [ζ2n]) ∈(TsS/TsS
⊥)2n−2,
where b′[zj ] = [ζj ] for j = 2, . . . , 2n− 1.
Let s ∈ S and b ∈ Sp(M,ω;S)s be given. Let bzj = ζj ∈ TsM for j = 1, . . . , 2n. Then
ζj ∈ TsS for j = 1, . . . , 2n − 1 and ζ1 ∈ TsS⊥. The flow ρt on Sp(M,ω;S), evaluated at b, is
given by
ρt(b) = ρt∗|m b =(ρt∗|mζ1, . . . , ρ
t∗|mζ2n
),
and the analogous expression holds for φt.
When we descend to Sp(TS/TS⊥), the image of b is the element b′ = ([ζ2], . . . , [ζ2n−1]) ∈
(TsS/TsS⊥)2n−2. The induced flow ρt on Sp(TS/TS⊥), evaluated at b′, is
ρt(b′) =([ρt∗|sζ2], . . . , [ρt∗|sζ2n−1]
).
As t varies, this expression describes a curve in (TS/TS⊥)2n−2, through b′. The tangent vector
to this curve at b′ is
ξH2(b′) =d
dt
∣∣∣∣t=0
ρt(b′) =
(d
dt
∣∣∣∣t=0
[ρt∗|sζ2], . . . ,d
dt
∣∣∣∣t=0
[ρt∗|sζ2n−1]
)∈ T[ζ2](TS/TS
⊥)× . . .× T[ζ2n−1](TS/TS⊥).
The pushforward of the quotient map TS → TS/TS⊥, based at ζj , is a linear surjection
TζjTS → T[ζj ](TS/TS⊥) whose kernel is TζjTsS
⊥. If we identify TζjTsS⊥ with TsS
⊥, then we
Chapter 4. Quantized Energy Levels and Dynamical Invariance 45
find a natural isomorphism between T[ζj ](TS/TS⊥) and TζjTS/TsS
⊥. Applying that isomor-
phism to each component of ξH2(b′) yields
ξH2(b′) =
([d
dt
∣∣∣∣t=0
(ρt∗|sζ2)
], . . . ,
[d
dt
∣∣∣∣t=0
(ρt∗|sζ2n−1)
])∈ Tζ2TS/TsS⊥ × . . .× Tζ2n−1TS/TsS
⊥.
Since ζj ∈ TsS for j = 2, . . . , 2n− 1, we can use Lemma 4.3.3:
d
dt
∣∣∣∣t=0
(ρt∗|sζj
)= c(s)
d
dt
∣∣∣∣t=0
(φt∗|sζj
)+ (ζjc)ξH1(s),
where each term in the above equation is viewed as an element of TζjTS. Upon descending to
TζjTS/TsS⊥, the multiple of ξH1(s) vanishes and we find that
[d
dt
∣∣∣∣t=0
(ρt∗|sζj
)]= c(s)
[d
dt
∣∣∣∣t=0
(φt∗|sζj
)].
Therefore
ξH2(b′) =
(c(s)
[d
dt
∣∣∣∣t=0
(φt∗|sζ2)
], . . . , c(s)
[d
dt
∣∣∣∣t=0
(φt∗|sζ2n−1)
])= c(s)ξH1(b′).
This completes the proof.
Thus ξH1 and ξH2 are parallel on Sp(TS/TS⊥), although the multiplicative factor that
relates them is not necessarily constant. This implies that φt and ρt have identical orbits in
Sp(TS/TS⊥), and if γS has trivial holonomy over the closed orbits of φt, then the same must
also be true for ρt. We can now conclude the result that was the objective of this section.
Theorem 4.3.5. If H1, H2 : M → R are smooth functions such that H−11 (E1) = H−1
2 (E2) for
regular values Ej of Hj , j = 1, 2, then E1 is a quantized energy level for (M,ω,H1) if and only
if E2 is a quantized energy level for (M,ω,H2).
Chapter 4. Quantized Energy Levels and Dynamical Invariance 46
4.4 The Harmonic Oscillator
In this section, we apply our quantized energy condition to the n-dimensional harmonic oscilla-
tor. The metaplectic-c quantization of this system has already been examined in [14] and [15],
so it will come as no surprise that we obtain the correct quantized energy levels:
EN = ~(N +
n
2
), N ∈ Z.
Our first objective in studying this example is to present a useful computational technique: we
will locally change coordinates from Cartesian to symplectic polar on the base manifold, then
show how to lift this change to the symplectic frame bundle and prequantization bundle. Once
symplectic polar coordinates have been established, we will turn to our second objective, which
is to construct an explicit example that illustrates the principal of dynamical invariance.
4.4.1 Initial Choices
Let (x1, . . . , xn, y1, . . . , yn) be a basis for the model vector space V such that Ω =∑n
j=1 x∗j ∧ y∗j .
All elements of V will be written as ordered 2n-tuples with respect to this basis. Assume
that V is identified with Cn by mapping the point (a1, . . . , an, b1, . . . , bn) ∈ V to the point
(b1 + ia1, . . . , bn + ian) ∈ Cn. Then the complex structure J on V is given in matrix form by
J =
0 I
−I 0
, where I is the n× n identity matrix.
A note concerning notation: given expressions aj and bj , j = 1, . . . , n, we let (aj)1≤j≤n
represent the n × n diagonal matrix diag(a1, . . . , an), and we let
aj
bj
1≤j≤n
represent the
2n× 1 column vector (a1, . . . , an, b1, . . . , bn)T .
Let M = R2n, with Cartesian coordinates (p1, . . . , pn, q1, . . . , qn) and symplectic form ω =∑nj=1 dpj ∧ dqj . The energy function for the harmonic oscillator is H = 1
2
∑nj=1
(p2j + q2
j
),
which has Hamiltonian vector field
ξH =
n∑j=1
(qj
∂
∂pj− pj
∂
∂qj
).
Chapter 4. Quantized Energy Levels and Dynamical Invariance 47
Let this vector field have flow φt on M .
There is a global trivialization of TM given by identifying xj 7→ ∂∂pj
∣∣∣m
, yj 7→ ∂∂qj
∣∣∣m
at every
point m ∈ M , which induces a global trivialization of Sp(M,ω). Let P be the trivial bundle
M ×Mpc(V ), with bundle projection map PΠ−→M . Define the map P
Σ−→ Sp(M,ω) by
Σ(m, a) = (m,σ(a)), ∀m ∈M, ∀a ∈ Mpc(V ),
where the ordered pair on the right-hand side is written with respect to the global trivialization
stated above. Then (P,Σ) is a metaplectic-c structure for (M,ω).
On M , define the one-form β by
β =1
2
n∑j=1
(pjdqj − qjdpj),
so that dβ = ω. Let ϑ0 be the trivial connection on the product bundle M × Mpc(V ), and
define the one-form γ on P by
γ =1
i~Π∗β +
1
2η∗ϑ0.
Then (P,Σ, γ) is a metaplectic-c prequantization for (M,ω). In fact, it is the unique metaplectic-
c prequantization up to isomorphism, since M is contractible.
Fix E > 0, a regular value of H, and let S = H−1(E). Let m0 ∈ S be given. Observe that
H and ω are invariant under following transformations: (1) rotation of the pjqj-plane about
the origin by any angle, for any j, and (2) simultaneous rotations of the pjpk- and qjqk-planes
about the origin by the same angle, for any j 6= k. Thus, after suitable rotations, we can
assume without loss of generality that m0 = (p01, . . . , p0n, 0, . . . , 0) in Cartesian coordinates,
where p0j 6= 0 for all j. It is easily established that
φt(m0) =
(cos t)1≤j≤n (sin t)1≤j≤n
(− sin t)1≤j≤n (cos t)1≤j≤n
p0j
0
1≤j≤n
.
This expression describes a periodic orbit with period 2π. Let C be the orbit of ξH through m0:
C =φt(m0) : t ∈ R
.
Chapter 4. Quantized Energy Levels and Dynamical Invariance 48
4.4.2 From Cartesian to Symplectic Polar Coordinates
On the Manifold
By symplectic polar coordinates, we mean the local coordinates (s1, . . . , sn, θ1, . . . , θn) given by
sj =1
2
(p2j + q2
j
), θj = tan−1
(qjpj
), j = 1, . . . , n, (4.4.1)
whenever these expressions are defined. The polar angles θj are all defined modulo 2π. For
later reference, the inverse coordinate transformations are
pj =√
2sj cos θj , qj =√
2sj sin θj , j = 1, . . . , n. (4.4.2)
Let
U =
(p1, . . . , pn, q1, . . . , qn) ∈M : p2j + q2
j > 0, j = 1, . . . , n,
so that symplectic polar coordinates and the corresponding vector fields
∂∂s1, . . . , ∂
∂sn, ∂∂θ1
, . . . ,
∂∂θn
are defined everywhere on U . Observe that the orbit C is contained in U . We will construct
a local trivialization for Sp(M,ω) over U , and a local trivialization for P over C.
When we convert to symplectic polar coordinates on U , we find
ω =n∑j=1
dsj ∧ dθj ,
which implies that for all m ∈ U ,
∂∂s1
∣∣∣m, . . . , ∂
∂sn
∣∣∣m, ∂∂θ1
∣∣∣m, . . . , ∂
∂θn
∣∣∣m
is a symplectic basis
for TmM . Further,
β =n∑j=1
sjdθj , H =n∑j=1
sj , ξH = −n∑j=1
∂
∂θj, (4.4.3)
and m0 = (s01, . . . , s0n, 0, . . . , 0), where s0j = 12p
20j , j = 1, . . . , n. The orbit C has the form
C = (s01, . . . , s0n, τ, . . . , τ) : τ ∈ R/2πZ .
The goal is to lift this change of coordinates to the symplectic frame bundle, then to the
metaplectic-c prequantization. To facilitate this process, we introduce the following notation.
Chapter 4. Quantized Energy Levels and Dynamical Invariance 49
Let Φc : U → R2n be the Cartesian coordinate map, and let Uc = Φc(U). Similarly, let
Φp : U → Rn × (R/2πZ)n be the symplectic polar coordinate map, and let Up = Φp(U). Let F
denote the transition map Φp Φ−1c . Then we have the following commutative diagram.
UΦc
zz
Φp
''Uc ⊂ R2n F // Up ⊂ Rn × (R/2πZ)n
This will serve as a model for the changes of coordinates on Sp(M,ω) and P .
On the Symplectic Frame Bundle
Recall that an element of the fiber Sp(M,ω)m is a symplectic isomorphism from V to TmM .
Over U , we define the sections bc and bp of Sp(M,ω) as follows. At each m ∈ U , they are given
by
bc(m) : V → TmM such that
xj
yj
1≤j≤n
7→
∂∂pj
∣∣∣m
∂∂qj
∣∣∣m
1≤j≤n
,
bp(m) : V → TmM such that
xj
yj
1≤j≤n
7→
∂∂sj
∣∣∣m
∂∂θj
∣∣∣m
1≤j≤n
.
Using these sections, we define two maps Φc, Φp on Sp(M,ω) over U :
Φc : Sp(M,ω)|U → Uc × Sp(V ) such that Φc(bc(m) · g) = (Φc(m), g), ∀m ∈ U, ∀g ∈ Sp(V ),
Φp : Sp(M,ω)|U → Up × Sp(V ) such that Φp(bp(m) · g) = (Φp(m), g), ∀m ∈ U, ∀g ∈ Sp(V ).
In other words, the section bc determines the Cartesian trivialization of Sp(M,ω) over U , and bp
determines the symplectic polar trivialization. Lifting the change of coordinates to Sp(M,ω)|U
is equivalent to finding a map F : Uc × Sp(V )→ Up × Sp(V ) that is a lift of F : Uc → Up and
such that the following commutes.
Sp(M,ω)|UΦc
ww
Φp
''Uc × Sp(V )
F // Up × Sp(V )
Chapter 4. Quantized Energy Levels and Dynamical Invariance 50
If such a map F exists, then it must satisfy
F Φc(bc) = Φp(bc),
and indeed, once this condition is satisfied, then the value of F everywhere else will be de-
termined by the Sp(V ) group action and the fact that F is a lift of F . Let us then evaluate
Φp(bc).
For all m ∈ U , let
G(m) =
(∂pj∂sj
∣∣∣m
)1≤j≤n
(∂qj∂sj
∣∣∣m
)1≤j≤n(
∂pj∂θj
∣∣∣m
)1≤j≤n
(∂qj∂θj
∣∣∣m
)1≤j≤n
,
so that
G(m)
∂∂pj
∣∣∣m
∂∂qj
∣∣∣m
1≤j≤n
=
∂∂sj
∣∣∣m
∂∂θj
∣∣∣m
1≤j≤n
.
Then G(m) is a symplectic matrix, and can be viewed as an element of Sp(V ). From the
definitions of bc(m) and the group action, we see that
bc(m) : G(m)
xj
yj
1≤j≤n
7→ G(m)
∂∂pj
∣∣∣m
∂∂qj
∣∣∣m
1≤j≤n
=
∂∂sj
∣∣∣m
∂∂θj
∣∣∣m
1≤j≤n
.
Thus
Φp(bc(m)) = (Φp(m), G(m)),
and more generally,
Φp(bc(m) · g) = (Φp(m), G(m)g), ∀g ∈ Sp(V ).
Hence we define the map F by
F (Φc(m), g) = (Φp(m), G(m)g), ∀m ∈ U, ∀g ∈ Sp(V ).
Chapter 4. Quantized Energy Levels and Dynamical Invariance 51
Observe, using Equation (4.4.2), that the entries of G(m) are singly defined with respect
to the angles θj , so G(m) is singly defined as m traverses the curve C, or any other closed
path through U . This demonstrates that we can change coordinates from Cartesian to polar
everywhere on Sp(M,ω)|U .
On the Metaplectic-c Prequantization
When we lift the change of variables to P , we restrict our attention further from U to C. Let
Cc = Φc(C) and let Cp = Φp(C). To emphasize that all of our calculations take place over the
closed curve C, we let m(τ) = (s01, . . . , s0n, τ, . . . , τ) ∈ C for all τ ∈ R/2πZ, and we abbreviate
G(m(τ)) by G(τ). Then G(τ) is a closed loop through Sp(V ) with period 2π.
Recall that P = M ×Mpc(V ), and that Σ : P → Sp(M,ω) is given by Σ(m, a) = (m,σ(a))
for all (m, a) ∈ P , where the right-hand side is written with respect to the Cartesian trivial-
ization. In other words, P was constructed to be consistent with Cartesian coordinates. To
formalize this property, we define
Φc : P |C → Cc ×Mpc(V ) such that Φc(m(τ), a) = (Φc(m(τ)), a), ∀(m(τ), a) ∈ P |C .
Then Φc is compatible with Φc in the sense that the following diagram commutes.
P |C Σ //
Φc
Sp(M,ω)|C
Φc
Cc ×Mpc(V )σ // Cc × Sp(V )
In the bottom line, σ maps Cc ×Mpc(V ) to Cc × Sp(V ) by acting on the second component.
Our goal is to construct a map Φp : P |C → Cp ×Mpc(V ) that is compatible with Φp in the
same sense. To do so, we will find a map F such that the diagram below commutes, then set
Φp = F Φc.
Chapter 4. Quantized Energy Levels and Dynamical Invariance 52
P |CΦc
yy %%
Σ
,,Cc ×Mpc(V )
F //
σ
,,
Cp ×Mpc(V )σ
,,
Sp(M,ω)|CΦc
ww
Φp
''Cc × Sp(V )
F // Cp × Sp(V )
Since F must be a lift of F and therefore of F , we assume that F takes the form
F (Φc(m(τ)), a) = (Φp(m(τ)), G(τ)a), ∀m(τ) ∈ C, ∀a ∈ Mpc(V ),
where G(τ) ∈ Mpc(V ). From the condition that
σ F Φc = F σ Φc,
it follows that we must have σ(G(τ)) = G(τ) for all τ ∈ R/2πZ. That is, G(τ) must be a lift of
G(τ) to Mpc(V ). More specifically, in order for F to be singly defined, G(τ) must be a closed
loop in Mpc(V ) with period 2π.
There is another consideration: the effect of this change of coordinates on the one-form γ.
Recall that γ is defined on P by
γ =1
i~Π∗β +
1
2η∗ϑ0,
where ϑ0 is the trivial connection on the product bundle M ×Mpc(V ). When we change to
symplectic polar coordinates, we would like γ to retain the same form, where β can be written
in polar coordinates as in Equation (4.4.3), and where ϑ0 now represents the trivial connection
on the product bundle Cp×Mpc(V ). Since ker(η) = Mp(V ), we can accomplish this by requiring
that the path G(τ) lie within Mp(V ).
Recall the parametrization of Mpc(V ) that was described in Section 2.3.1. If G(τ) is
a lift of G(τ) to Mp(V ), then the parameters of G(τ) have the form (G(τ), µ(τ)) where
µ(τ)2DetCCG(τ) = 1. To determine µ(τ), we must examine the matrix G(τ) ∈ Sp(V ) explicitly.
Chapter 4. Quantized Energy Levels and Dynamical Invariance 53
Using Equation (4.4.2), we find
G(τ) =
(
1√2s0j
cos τ
)1≤j≤n
(1√2s0j
sin τ
)1≤j≤n(
−√
2s0j sin τ)
1≤j≤n(√
2s0j cos τ)
1≤j≤n
.
Now we calculate
CG(τ) =1
2(G(τ)− JG(τ)J)
using the complex structure noted in Section 4.4.1, then convert to an n × n complex matrix.
The result is the diagonal matrix
CG(τ) =1
2
((√2s0j +
1√2s0j
)eiτ
)1≤j≤n
,
which has complex determinant
DetCCG(τ) =n∏j=1
1
2
(√2s0j +
1√2s0j
)eiτ = Keinτ ,
where K is a positive real value that is constant over C. Thus, a lift of G(τ) to Mp(V ) is
parametrized by
G(τ) 7→(G(τ),
1√Ke−inτ/2
). (4.4.4)
Notice that (G(τ + 2π), µ(τ + 2π)) = (G(τ), µ(τ)e−inπ). When n is even, we get a closed
loop through Mp(V ) with period 2π, which was the optimal outcome. When n is odd, however,
traversing C once multiplies µ(τ) by −1. This shows that we cannot always lift G(τ) to a closed
path through Mp(V ). Nevertheless, we choose the path G(τ) as defined in Equation (4.4.4)
because it preserves the form of γ. If we let εn ∈ Mpc(V ) be the element whose parameters are
(I, e−iπn), then G(τ + 2π) = εnG(τ).
To prevent F from being multi-valued, let C = m(τ) ∈ C : τ ∈ (0, 2π), and let Cc and
Cp be the images of C in Cartesian and symplectic polar coordinates, respectively. Then let
F : Cc ×Mpc(V )→ Cp ×Mpc(V ) be given by
F (Φc(m(τ)), a) = (Φp(m(τ)), G(τ)a), ∀(m(τ), a) ∈ P |C ,
Chapter 4. Quantized Energy Levels and Dynamical Invariance 54
and identify P |C with Cp × Mpc(V ) under the map Φp = F Φc. This gives us symplectic
polar coordinates for P over C, and we will manually adjust for the fact that closing the loop
multiplies G(τ) by εn. On Cp ×Mpc(V ), γ takes the form
γ =1
i~
n∑j=1
s0jdθj +1
2η∗ϑ0,
where ϑ0 is the trivial connection on the product bundle, and where the values s0j are constant
over Cp.
4.4.3 Quantized Energy Levels of the Harmonic Oscillator
Having established the change of coordinates, we will now drop the explicit use of Φp, F and
their lifts. Elements of U will be written with respect to the symplectic polar coordinates
(s1, . . . , sn, θ1, . . . , θn), which for convenience we abbreviate by Xk, k = 1, . . . , 2n. Using the
identifications xj 7→ ∂∂sj
∣∣∣m, yj 7→ ∂
∂θj
∣∣∣m
for all m ∈ U , j = 1, . . . , n, we write
Sp(M,ω)|U = U × Sp(V ) and P |C = C ×Mpc(V ).
At the level of tangent spaces, we have
T(m,I) Sp(M,ω) = TmM × sp(V ), ∀m ∈ U,
and
T(m,I)P = TmM ×mpc(V ) = TmM × (sp(V )⊕ u(1)) , ∀m ∈ C.
Recall that on U , ξH = −∑n
j=1∂∂θj
. Therefore TmS⊥ = span
∑nj=1
∂∂θj
∣∣∣m
at each m ∈ U .
Within the model vector space V , let
W⊥ = span y1 + . . .+ yn .
Then
W = span x1 − x2, . . . , xn−1 − xn, y1, . . . , yn ,
Chapter 4. Quantized Energy Levels and Dynamical Invariance 55
and
W/W⊥ = span [x1 − x2], . . . , [xn−1 − xn], [y1 − y2], . . . , [yn−1 − yn] .
The local trivialization of Sp(M,ω)|U induces the local trivializations Sp(M,ω;S)|U∩S = U ∩
S × Sp(V ;W ) and Sp(TS/TS⊥)|U∩S = U ∩ S × Sp(W/W⊥).
By definition, PS is the bundle associated to PS by the group homomorphism ν : Mpc(V )→
Mpc(W/W⊥). Over C, we have the local trivializations P |C = C × Mpc(V ), PS |C = C ×
Mpc(V ;W ) and PS |C = C ×Mpc(W/W⊥), where the associated bundle map ν : PS |C → PS |Cacts by
ν(m, a) = (m, ν(a)), ∀(m, a) ∈ PS |C .
Using properties of ν noted in Section 4.2.1, we have ν(εn) = εn ∈ Mpc(W/W⊥) for all n.
Therefore, within PS , the same factor of εn is required to close the loop from C to C. Further,
since η∗ ν∗ = η∗, the one-form γS induced on PS does not change form under the map ν:
γS |C =1
i~
n∑j=1
s0jdθj +1
2η∗ϑ0,
where ϑ0 is now the trivial connection on the product bundle C ×Mpc(W/W⊥). Lastly, the
relevant tangent spaces are
T(m,I) Sp(TS/TS⊥) = TmM × sp(W/W⊥), ∀m ∈ U,
and
T(m,I)PS = TmM × (sp(W/W⊥)⊕ u(1)), ∀m ∈ C.
Recall that we chose the initial point m0 = (s01, . . . , s0n, 0, . . . , 0), and that C is the orbit of
ξH through m0. We will show that the orbit of ξH through (m0, I) is closed in Sp(TS/TS⊥).
We first need to lift φt to the flow φt on Sp(M,ω). By definition, for any m ∈ U , φt(m, I) =
(φt(m), φt∗|m). This implies that
ξH(m, I) =d
dt
∣∣∣∣t=0
φt(m, I) =
(ξH(m),
d
dt
∣∣∣∣t=0
φt∗|m).
Chapter 4. Quantized Energy Levels and Dynamical Invariance 56
We write φt = (φt1, . . . , φt2n) with respect to the symplectic polar coordinates. Then φt∗|m is
a 2n× 2n matrix, which we interpret as an element of Sp(V ), and its components are given by
(φt∗)jk =∂φtj∂Xk
∣∣∣m
. Noting that ddt
∣∣t=0
φtj = (ξH)j , we compute
(d
dt
∣∣∣∣t=0
φt∗|m)jk
=∂
∂Xk
∣∣∣∣m
(ξH)j . (4.4.5)
But ξH = −∑n
j=1∂∂θj
on U , which has constant components, so ddt
∣∣t=0
φt∗ is identically 0. Thus
ξH(m, I) = (ξH(m), 0), ∀m ∈ U.
In particular, since the sp(V ) component of ξH is constant over the orbit C, we can find the
flow for ξH on Sp(M,ω) through (m0, I) by exponentiating: it is simply
φt(m0, I) = (φt(m0), I).
This is clearly a closed orbit with period 2π. The induced flow on the bundle Sp(TS/TS⊥) also
takes the form
φt(m0, I) = (φt(m0), I).
Now we lift ξH to PS , horizontally with respect to γS . Over the curve C, we calculate that
ξHyβ = −∑n
j=1 s0j = −E. Then, for any m ∈ C, the horizontal lift of ξH to T(m,I)PS is
ξH(m, I) =
(ξH(m), 0⊕ E
i~
).
The mpc(V ) component is constant over C, so the flow is again calculated by exponentiating:
φt(m0, I) =
(φt(m0), exp
(0⊕ Et
i~
))= (φt(m0), eEt/i~),
where eEt/i~ ∈ U(1) ⊂ Mpc(W/W⊥).
The quantized energy levels are those values of E for which this orbit closes. However, we
must remember that one circuit about C introduces an extra factor of εn. Referring to the
properties of the parametrization stated in Section 2.3.1, we see that φ0(m0, I) = φ2π(m0, I)
Chapter 4. Quantized Energy Levels and Dynamical Invariance 57
if and only if e2πE/i~ = e−inπ, or in other words, when 2πEi~ + inπ = −2πiN for some N ∈ Z.
Upon rearranging, we find that the quantized energy levels must take the form
E = ~(N +
n
2
), N ∈ Z.
Thus the quantization condition correctly reproduces the expected energy levels of the harmonic
oscillator.
4.4.4 Example of Dynamical Invariance
Fix n = 2. It is convenient to use the definitions sj = 12(p2
j + q2j ) and ∂
∂θj= pj
∂∂qj− qj ∂
∂pjevery-
where on M , for j = 1, 2. We will point out when a statement only holds on the neighborhood
U .
Let k ∈ R be a positive constant, and consider the two functions H1, H2 : M → R given by
H1 = s1 + s2 − k, H2 = (s1 + s2 − k)(s1 + 2s2 + 1).
The function H1 is just the energy function for the harmonic oscillator, shifted by k: its
quantized energy levels are
EN = ~N − k, N ∈ Z.
Notice that H−11 (0) = H−1
2 (0). Let this shared level set be S. The energy E = 0 is a quantized
energy for the system (M,ω,H1) if and only if k~ ∈ Z.
The two Hamiltonian vector fields are
ξH1 = − ∂
∂θ1− ∂
∂θ2,
ξH2 = (s1 + 2s2 + 1)ξH1 −H1
(∂
∂θ1+ 2
∂
∂θ2
)= −(2s1 + 3s2 − k + 1)
∂
∂θ1− (3s1 + 4s2 − 2k + 1)
∂
∂θ2.
Since H1 = 0 on S, we see that ξH2 = (s1 + 2s2 + 1)ξH1 everywhere on S. Thus the vector fields
are parallel on S, as expected, and so they share the same orbits in S. However, ξH1 and ξH2
are not parallel away from S. Let φt be the flow of ξH1 , and let ρt be the flow of ξH2 .
Chapter 4. Quantized Energy Levels and Dynamical Invariance 58
Consider the initial point m0 ∈ S ∩ U , where m0 = (s01, s02, 0, 0) with s01 + s02 = k and
s01, s02 6= 0. The orbit of both φt and ρt through m0 is
C = (s01, s02, τ, τ) : τ ∈ R/2πZ .
From Section 4.4.3, we know that
ξH1(m, I) = (ξH1(m), 0), ∀m ∈ C,
and therefore
φt(m0, I) = (φt(m0), I).
By the same calculation, ξH2(m, I) = (ξH2(m), ddt
∣∣t=0
ρt∗|m) for m ∈ C. Applying Equation
(4.4.5) to the components of ξH2 yields
d
dt
∣∣∣∣t=0
ρt∗|m =
0 0 0 0
0 0 0 0
−2 −3 0 0
−3 −4 0 0
,
which we interpret as an element of sp(V ). Let this matrix be denoted by κ. Then
ξH2(m, I) = (ξH2(m), κ),
and since the Lie algebra component is constant over C, we obtain the flow on Sp(M,ω) through
(m0, I) by exponentiating:
ρt(m0, I) = (ρt(m0), exp(tκ)).
Chapter 4. Quantized Energy Levels and Dynamical Invariance 59
A calculation establishes that
exp(tκ) =
1 0 0 0
0 1 0 0
−2t −3t 1 0
−3t −4t 0 1
,
which is clearly not periodic. Thus ρt has no closed orbits on Sp(M,ω) over S ∩ U .
Now let us transfer to Sp(TS/TS⊥). If we apply the definitions and identifications laid out at
the beginning of Section 4.4.3, now setting n = 2, then we find W⊥ = span y1 + y2, W/W⊥ =
span [x1 − x2] , [y1 − y2], and the identification of Sp(M,ω)|C with C × Sp(V ) induces an
identification of Sp(TS/TS⊥)|C with C × Sp(W/W⊥). Notice that
exp(tκ)(x1 − x2) = x1 − x2 + t(y1 + y2),
exp(tκ)(y1 − y2) = y1 − y2.
Therefore the path through Sp(W/W⊥) induced by exp(tκ) is
ν(exp(tκ)) =
1 0
0 1
.
Thus, on Sp(TS/TS⊥)|C ,
ρt(m0, I) =(ρt(m0), I
),
which coincides with φt(m0, I).
The above calculation omits certain cases: namely, if the starting point m0 has s01 = 0
or s02 = 0. We can no longer eliminate such cases by performing a rotation, because H2 is
not symmetric with respect to s1 and s2. If m0 /∈ U , then we have to modify our approach.
For example, if s01 = 0, then we retain Cartesian coordinates for the p1q1-plane and convert
to symplectic polar on the p2q2-plane. The calculation is more complicated, but the result is
similar: over C, we find that ξH2 = (ξH2 , κ) for some constant value of κ ∈ sp(V ). The path
ρt(m0, I) = (ρt(m0), exp(tκ)) does not close in Sp(M,ω), but the induced path in Sp(TS/TS⊥)
Chapter 4. Quantized Energy Levels and Dynamical Invariance 60
coincides with φt(m0, I). The same pattern holds if we take s02 = 0.
Thus, if the quantized energy condition were stated in terms of the holonomy of γS over
closed orbits in Sp(M,ω;S), as it was in [14], then the value E = 0 would satisfy the condition
vacuously for the system (M,ω,H2), regardless of the value of k. It is only by descending
to Sp(TS/TS⊥) that we recover the quantization condition k~ ∈ Z. Hence our definition of a
quantized energy level is dynamically invariant, while that in [14] is not.
4.5 Comparison with Kostant-Souriau Quantization
4.5.1 The Quantized Energy Condition and Its Properties
The quantized energy condition in Definition 4.2.2 can be easily adapted to Kostant-Souriau
prequantization. Indeed, the Kostant-Souriau case is simpler, since it does not involve the
symplectic frame bundle.
Assume that (M,ω) admits a prequantization circle bundle (Y, γ), and let H : M → R
be a smooth function. We now mimic the constructions in Section 4.2.2, using Y in place of
P . Denote the Hamiltonian vector field corresponding to H by ξH as before, and let ξH be
the lift of ξH to Y that is horizontal with respect to γ. Fix E, a regular value of H, and let
S = H−1(E) ⊂M . Let Y S be the restriction of Y to S, and let γS be the pullback of γ to Y S .
(Y, γ)
(Y S , γS)incl.oo
(M,ω) S
incl.oo
Since there is no group homomorphism with which to construct an associated bundle, the
construction stops here and we give the definition of a quantized energy level using (Y S , γS).
Definition 4.5.1. If the connection one-form γS has trivial holonomy over all closed orbits of
the Hamiltonian vector field ξH on S, then E is a Kostant-Souriau (KS) quantized energy
level for the system (M,ω,H).
It is straightforward to establish that this definition has a dynamical invariance property.
Suppose H1, H2 : M → R are smooth functions such that H−11 (E1) = H−1
2 (E2) = S for regular
Chapter 4. Quantized Energy Levels and Dynamical Invariance 61
values E1 and E2. We argued in Section 4.3.2 that ξH1 and ξH2 are parallel on S, which implies
that they have the same orbits. Therefore γS has trivial holonomy over the orbits of one if and
only if it has trivial holonomy over the orbits of the other. The KS version of the dynamical
invariance theorem is immediate.
Theorem 4.5.2. If H1, H2 : M → R are smooth functions such that H−11 (E1) = H−1
2 (E2) for
regular values Ej of Hj , j = 1, 2, then E1 is a KS quantized energy level for (M,ω,H1) if and
only if E2 is a KS quantized energy level for (M,ω,H2).
In Section 4.2.2, we examined the case in which the symplectic reduction of (M,ω) at
E is a manifold. Theorem 4.2.1, due to Robinson [14], gives the conditions under which the
quantization condition is sufficient to imply that the symplectic reduction admits a metaplectic-
c prequantization. A similar theorem can be given in the context of prequantization circle
bundles.
First, we state a general result, which was used in [14] to prove Theorem 4.2.1. Let S
be an arbitrary manifold, and suppose that (Z, δ) is a principal circle bundle with connection
one-form over S. Let the curvature of δ be $. Suppose that F is a foliation of S whose leaf
space SF is a smooth manifold. Denote the leaf projection map by Sπ−→ SF . If δ has trivial
holonomy over all of the leaves of F , then Z can be factored to produce a well-defined circle
bundle ZF → SF . Further, δ descends to a connection one-form δF on ZF , and the curvature
$F of δF satisfies π∗$F = $.
Now apply this result to the circle bundle (Y S , γS)→ S, where the foliation is given by the
orbits of the vector field ξH on the level set S, and the symplectic reduction (ME , ωE) is its
leaf space.
Theorem 4.5.3. Suppose that the symplectic reduction (ME , ωE) for (M,ω) at E is a manifold.
If γS has trivial holonomy over all closed orbits of ξH on S, then the quotient of (Y S , γS) by
the orbits of ξH is a prequantization circle bundle for (ME , ωE).
Thus the Kostant-Souriau version of the quantized energy condition is sufficient to ensure
that the symplectic reduction admits a prequantization circle bundle, whenever the symplectic
reduction is a manifold. This is a slight improvement over the metaplectic-c result, since it does
not depend on a quotient of the symplectic frame bundle being well defined.
Chapter 4. Quantized Energy Levels and Dynamical Invariance 62
4.5.2 Lack of Half-Shift in the Harmonic Oscillator
In this section, we will determine the KS quantized energy levels of the n-dimensional harmonic
oscillator. The calculation will be significantly simpler than that in Section 4.4, but it will yield
the wrong answer.
Let M = R2n, with Cartesian coordinates (p1, . . . , pn, q1, . . . , qn) and symplectic form ω =∑nj=1 dpj ∧ dqj . The energy function and corresponding Hamiltonian vector field for the har-
monic oscillator are
H =1
2
n∑j=1
(p2j + q2
j ), ξH =
n∑j=1
(qj
∂
∂pj− pj
∂
∂qj
).
Let the flow of ξH on M be φt. We know from Section 4.4.1 that all of the orbits of ξH are
circles, and that φt+2π(m) = φt(m) for all m ∈M .
Let Y = M × U(1), with projection map YΠ−→M . We define
β =1
2
n∑j=1
(pjdqj − qjdpj)
on M , and let γ = 1i~Π∗β + ϑ0 on Y , where ϑ0 is the trivial connection on the product bundle
M × U(1). Then (Y, γ) is a prequantization circle bundle for (M,ω), and it is unique up to
isomorphism.
Let E > 0 be arbitrary, and let S = H−1(E). At any point s ∈ S, we calculate that
ξHyβ = −E. Therefore the lifted vector field ξH at the point (s, I) ∈ Y is
ξH(s, I) =
(ξH(s),
E
i~
),
where we identify the tangent space T(s,I)Y with TsM × u(1). Note that the u(1) component
is constant over all of S, so in particular it is constant over an orbit. Let the flow of ξH be φt.
By exponentiation, we find that
φt(s, I) =(φt(s), eEt/i~
).
Chapter 4. Quantized Energy Levels and Dynamical Invariance 63
From this, it follows that the holonomy of γS over the orbits in S is trivial if and only if E = N~
for some N ∈ N. Thus the KS quantized energy levels are inconsistent with the predictions of
quantum mechanics when the dimension n is odd.
This shortcoming in the Kostant-Souriau prequantization of the harmonic oscillator is well
known, and the standard solution is to proceed from prequantization to quantization while in-
troducing the half-form correction, as described in Section 2.2.2. In this context, the quantized
energy levels for the system (M,ω,H) can be taken to be the eigenvalues of the operator cor-
responding to the energy function H. When the quantization recipe is applied to the harmonic
oscillator, the presence of the half-form bundle adds the n2 shift to the energy eigenvalues. How-
ever, this correction comes at the cost of introducing a choice of polarization, and the quantized
energy definition can no longer be evaluated over a single level set of H.
By comparison, when we use the metaplectic-c formulation of a quantized energy level,
we find that the correct harmonic oscillator energies are encoded in the geometry of the level
sets of the energy function. The result is dynamically invariant, independent of polarization,
and consistent with physical prediction. This example illustrates the benefits of metaplectic-c
quantization and our quantized energy definition.
Chapter 5
The Hydrogen Atom
5.1 Introduction
The quantized energy levels of the hydrogen atom have been calculated for various physical
models, using various flavors of geometric quantization. Notable examples include:
• Simms [16], who used the observation that the space of orbits corresponding to a fixed
negative energy is isomorphic to S2 × S2, and determined those energies for which the
reduced manifold admits a prequantization circle bundle;
• Sniatycki [19], who looked at the 2-dimensional relativistic Kepler problem and computed
a Bohr-Sommerfeld condition for the completely integrable system given by considering
the energy and angular momentum functions simultaneously;
• Duval, Elhadad and Tuynman [6], who took the phase space to include the spins of the
electron and proton, then chose a polarization and determined the Kostant-Souriau quan-
tized operator corresponding to the energy function with fine and hyperfine interaction
terms.
These examples exist at one of two possible extremes. On one hand, the quantized energy
condition can be evaluated over the symplectic reduction of a particular level set of the energy
function, as in [16]. This definition only looks at one energy level at a time, but it requires
constructing the symplectic reduction and establishing that the result is a smooth manifold. On
64
Chapter 5. The Hydrogen Atom 65
the other hand, the quantized energy levels can be determined from properties of the quantized
system as a whole, as in [19] or [6]. These approaches are characterized by requiring a polariza-
tion, or the equivalent information – recall that the Hamiltonian vector fields corresponding to
the Poisson-commuting functions in a completely integrable system generate a real polarization
– and the calculation is not restricted to a single level set of the energy function in question.
Our proposed definition for a quantized energy level acts as a middle ground between these
two extremes. The objective of this chapter is to apply the metaplectic-c quantized energy
condition to the hydrogen atom, using the physical model that is equivalent to the Kepler
problem. We will show that the quantized energy levels are in agreement with the quantum
mechanical prediction.
In Section 5.2, we set up our model of the hydrogen atom and construct a metaplectic-c
prequantization for its phase space. Section 5.3 presents the Ligon-Schaaf regularization map,
which is a symplectomorphism from the negative-energy domain of the hydrogen atom to an
open submanifold of TS3. We show how to relate the energy function for the hydrogen atom to
that of a free particle on S3. Finally, in Section 5.4, we determine the quantized energy levels
of a free particle on S3, and use these to determine the quantized energy levels for the hydrogen
atom.
We will apply the constructions that were described in Section 4.2. In addition, we will
make repeated use of the notation and conventions that are described below.
5.1.1 Choices for Subsequent Calculations
In the sections that follow, we will need model symplectic vector spaces of several different
dimensions. Let us fix some standardized choices.
For any n ∈ N, let (Vn,Ωn) be a 2n-dimensional symplectic vector space. Let (v1, . . . , vn,
w1, . . . , wn) be a symplectic basis for Vn, and write all elements of Vn as ordered 2n-tuples with
respect to this basis. The symplectic form can be written in terms of the dual basis as
Ωn =
n∑j=1
v∗j ∧ w∗j .
Assume that each real vector (a1, . . . , an, b1, . . . , bn) ∈ Vn is identified with the complex vector
Chapter 5. The Hydrogen Atom 66
(b1 + ia1 . . . , bn+ ian) ∈ Cn. The resulting complex structure J on Vn is written in matrix form
as J =
0 I
−I 0
, where I is the n× n identity matrix.
When we require a subspace of Vn of codimension 1, we choose
Wn = span v1, . . . , vn, w1, . . . , wn−1 .
Then
W⊥n = span vn and Wn/W⊥n = span [v1], . . . , [vn−1], [w1], . . . , [wn−1] .
Using equation 4.2.1, it is immediate that Wn/W⊥n is isomorphic to Vn−1 as a symplectic vector
space and a complex vector space. The commutative diagram from Section 4.2.1 containing the
group homomorphisms ν and ν can be rewritten as follows.
Mpc(Vn) ⊃ Mpc(Vn;Wn)ν //
σ
Mpc(Vn−1)
σ
Sp(Vn) ⊃ Sp(Vn;Wn)
ν //// Sp(Vn−1)
5.2 Quantizing the Hydrogen Atom
5.2.1 Setup
Let R3 represent R3 with the origin removed, and let M = T R3 = R3 × R3. We use Cartesian
coordinates q = (q1, q2, q3) on R3 and p = (p1, p2, p3) on R3. Equip M with the symplectic form
ω =∑3
j=1 dqj ∧ dpj .
We consider the model of the hydrogen atom that is equivalent to the Kepler problem.
Assume that a proton is fixed at the origin in R3, and that an electron of mass me interacts with
it via the electrostatic force, which obeys an inverse-square law with constant of proportionality
k. Then M is the phase space for the motion of the electron, where q and p represent its position
and momentum, respectively. The Hamiltonian energy function is
H =1
2me|p|2 − k
|q|,
Chapter 5. The Hydrogen Atom 67
and the corresponding Hamiltonian vector field is
ξH =3∑j=1
(1
mepj
∂
∂qj− k
|q|3qj
∂
∂pj
).
Let ξH have flow φt on M .
The solutions to the Kepler problem are well known [1, 3]. The angular momentum vector
L = q×p is a constant of the motion, as is the eccentricity vector e = 1mek
p×L− q|q| . In position
space, the orbits corresponding to a given energy E ∈ R are conic sections with eccentricity
|e| =
√1 +
2E|L|2mek2
, (5.2.1)
having the origin as a focus.
In particular, suppose E < 0. Then the value of |L| lies in the interval
[0,√−mek2
2E
]. If
|L| > 0, then the orbit is an ellipse, with |L| =√−mek2
2E being the special case of a circle. All
elliptical orbits with energy E have period 2πΛ , where Λ =
√− 8E3
mek2. If |L| = 0, however, then
e = 1 and the orbit is a line segment. Physically, this represents the case where the electron
begins from rest and collapses on a straight-line trajectory into the proton. Such motion is
not periodic, which implies that the level set H−1(E) contains orbits of ξH that are not closed.
Further, the collapse occurs in finite time, meaning that the vector field ξH is not complete.
The objective of this chapter is to determine the quantized energy levels for (M,ω,H). In
the next section, we construct a metaplectic-c prequantization for (M,ω), and formulate our
approach for performing the quantized energy calculation.
5.2.2 Metaplectic-c Prequantization of (M,ω)
We choose the model symplectic vector space V3, as described in Section 5.1.1. The tangent
bundle TM can be identified with M × V3 with respect to the global trivialization
vj 7→∂
∂qj
∣∣∣∣m
, wj 7→∂
∂pj
∣∣∣∣m
, ∀m ∈M, j = 1, 2, 3.
This yields an identification of the symplectic frame bundle Sp(M,ω) with M × Sp(V3).
Chapter 5. The Hydrogen Atom 68
Let P = M ×Mpc(V3), with bundle projection map PΠ−→ M . Define the map Σ : P →
Sp(M,ω) by
Σ(m, a) = (m,σ(a)), ∀m ∈M,∀a ∈ Mpc(V3),
where the right-hand side is written with respect to the global trivialization above. Let
β =3∑j=1
(qjdpj + d(qjpj)) (5.2.2)
on M , so that dβ = ω. The reason for this choice of β will be made clear in Section 5.3.1. Let
γ be the u(1)-valued one-form on P given by
γ =1
i~Π∗β +
1
2η∗ϑ0,
where ϑ0 is the trivial connection on the product bundle. Then (P,Σ, γ) is a metaplectic-c
prequantization for (M,ω), and it is unique up to isomorphism.
The quantized energy levels of (M,ω,H) are those regular values E of H such that the
holonomy of γS is trivial over all closed orbits of ξH on Sp(TS/TS⊥), where S = H−1(E). If
E ≥ 0, then the quantization condition can be evaluated immediately. From equation 5.2.1 for
the eccentricity of the orbit, we see that if E = 0, then the orbits are parabolas in position
space, and if E > 0, then they are hyperbolas. In these cases, ξH has no closed orbits in S,
which implies that ξH cannot have any closed orbits in Sp(TS/TS⊥). Therefore the holonomy
condition is satisfied vacuously, and all nonnegative energy levels are quantized energy levels.
This is consistent with the physical prediction from quantum mechanics: a particle that is not
spatially confined has a continuous energy spectrum.
It remains to consider the orbits corresponding to negative energy. Let
N = m ∈M : H(m) < 0 .
Then N is an open, simply connected submanifold of M . By restriction, the symplectic form
on M induces one on N , and the metaplectic-c prequantization for M induces one for N . We
use the same symbols to denote the restricted objects: (N,ω) is a symplectic manifold, and
Chapter 5. The Hydrogen Atom 69
(P,Σ, γ) is its metaplectic-c prequantization. Since N is simply connected, (P,Σ, γ) is unique
up to isomorphism.
Let E < 0 be fixed, and let S = H−1(E) ⊂ N . Through the process described in Section
4.2.2, we obtain the three-level structures
(PS , γS)→ Sp(N,ω;S)→ S
and
(PS , γS)→ Sp(TS/TS⊥)→ S.
Lift ξH to ξH on Sp(N,ω;S), and let it descend to Sp(TS/TS⊥).
To evaluate the quantization condition using these bundles, we would have to determine
the closed orbits of ξH on Sp(TS/TS⊥), then lift ξH to ξH on PS , horizontally with respect
to γS , and ensure that every lift of a closed orbit in Sp(TS/TS⊥) is closed in PS . However,
this procedure is computationally prohibitive in all but the special case of the circular orbit.
Instead, we will use Ligon-Schaaf regularization to transform the quantized energy calculation
on (N,ω) into one on an open submanifold of TS3. This is the subject of Section 5.3.
5.3 Ligon-Schaaf Regularization
Let T+S3 represent the result of removing the zero section from TS3. In Section 5.2.1, we noted
that the vector field ξH is not complete. The Ligon-Schaaf map is a symplectomorphism from
(N,ω) to an open submanifold of T+S3, having the property that ξH is mapped to a vector
field that is complete on T+S3. This map was presented in [13], and an in-depth discussion of
it can be found in [3].
In this section, we will state the Ligon-Schaaf map and list its relevant properties. Then we
will construct a metaplectic-c prequantization for TS3, and show how to lift the Ligon-Schaaf
map to the level of symplectic frame bundles, and of metaplectic-c prequantizations. Using the
lifted maps, we will be able to relate the quantized energy levels of the hydrogen atom to those
of a free particle on S3.
Our conventions and notation largely follow those in [3]. We state results without proof;
Chapter 5. The Hydrogen Atom 70
much more detail can be found in [3].
5.3.1 TS3 and the Ligon-Schaaf Map
Consider TR4 = R4 × R4 with Cartesian coordinates (x, y), where x = (x1, x2, x3, x4) and
y = (y1, y2, y3, y4). Let TR4 have symplectic form
ω4 =4∑j=1
dxj ∧ dyj ,
and let
β4 = −4∑j=1
yjdxj (5.3.1)
on TR4, so that dβ4 = ω4. The submanifold TS3 is given by
TS3 =
(x, y) ∈ TR4 : |x|2 = 1, x · y = 0⊂ TR4,
where we take the usual Euclidean inner product on R4. From now on, we abbreviate TR4 by
Z.
By treating (TS3, ω3) as a level set of the constraint functions c1(x, y) = 12(|x|2 − 1) and
c2(x, y) = x · y on (Z, ω4), it can be shown that the restriction of ω4 to TS3 yields a symplectic
form on TS3. Let (TS3, ω3) be the resulting symplectic manifold. Let β3 be the restriction of
β4 to TS3, so that dβ3 = ω3.
Let T+S3 represent TS3 with the zero section removed:
T+S3 =
(x, y) ∈ TS3 : |y| > 0.
Define the map D : T+S3 → R by
D(x, y) = −mek2
2|y|2, ∀(x, y) ∈ T+S3.
Chapter 5. The Hydrogen Atom 71
As shown in [3], the Hamiltonian vector field for D on T+S3 is
ξD =
4∑j=1
(mek
2
|y|4yj
∂
∂xj− mek
2
|y|2xj
∂
∂yj
),
and its flow is
ψtD(x, y) =
cos mek2
|y|3 t1|y| sin
mek2
|y|3 t
−|y| sin mek2
|y|3 t cos mek2
|y|3 t
x
y
.
The map D is called the Delaunay Hamiltonian, and ξD is the Delaunay vector field. Since the
orbits of ξD must preserve |y|, it is clear from the form of ψtD that ξD is complete on T+S3,
and every orbit is closed.
Now let S3n represent S3 with the north pole (0, 0, 0, 1) removed. The Ligon-Schaaf map
LS : N → T+S3n is given by
LS(q, p) = (A sinϕ+B cosϕ,−νA cosϕ+ νB sinϕ),
where
ν =
√− mek2
2H(q, p), ϕ =
1
ν(q · p),
A =
(q
|q|− 1
mek(q · p)p, 1
ν(q · p)
), B =
(1
ν|q|p, 1
mek|p|2|q| − 1
).
This map has the following properties:
(1) LS is a diffeomorphism between N and T+S3n;
(2) LS∗β3 =∑3
j=1 (qjdpj + d(qjpj)) = β (recall equation 5.2.2);
(3) H = D LS.
Assertions (1) and (2) imply that LS is a symplectomorphism; from that and (3), it follows
that LS∗ξH = ξD.
5.3.2 Metaplectic-c Prequantization for TS3
We begin by constructing a metaplectic-c prequantization for (Z = TR4, ω4), and let it induce a
metaplectic-c prequantization for (TS3, ω3). For (Z, ω4), we proceed in precisely the same way
Chapter 5. The Hydrogen Atom 72
as we constructed (P,Σ, γ) for (M,ω) in Section 5.2.2. Choose the model vector space V4, and
identify the symplectic frame bundle Sp(Z, ω4) with Z × Sp(V4) using the global trivialization
for the tangent bundle given by
vj 7→∂
∂xj
∣∣∣∣(x,y)
, wj 7→∂
∂yj
∣∣∣∣(x,y)
, ∀(x, y) ∈ Z, j = 1, . . . , 4.
Let Q4 = Z × Mpc(V4) with bundle projection map Q4Π4−→ Z. Define the map Γ4 : Q4 →
Sp(TR4, ω4) by
Γ4(x, y, a) = (x, y, σ(a)), ∀(x, y) ∈ Z,∀a ∈ Mpc(V4).
Lastly, define the u(1)-valued one-form δ4 on Q4 by
δ4 =1
i~Π∗4β4 +
1
2η∗ϑ0,
where ϑ0 is the trivial connection on Q4, and where β4 was defined in equation 5.3.1. Then
(Q4,Γ4, δ4) is the unique metaplectic-c prequantization for (Z, ω4) up to isomorphism.
In order to construct the metaplectic-c prequantization of (TS3, ω3) induced by (Q4,Γ4, δ4),
we proceed by the process of symplectic reduction. Let R : Z → R be given by
R(x, y) =1
2|x|2, ∀(x, y) ∈ Z.
The Hamiltonian vector field for R on Z is
ξR = −4∑j=1
xj∂
∂yj,
and its flow on Z is
ψtR(x, y) = (x, y − tx), ∀(x, y) ∈ Z.
It is straightforward to show that TS3 can be identified with the space of orbits of ξR on the
level set A = R−1(12). The orbit projection map p : A→ TS3 is
p(x, y) = (x, y − (x · y)x), ∀(x, y) ∈ A.
Chapter 5. The Hydrogen Atom 73
Let i : A → Z be the inclusion map. Recall that ω3 is the symplectic form on TS3 obtained
by restricting ω4. A calculation shows that p∗ω3 = i∗ω4. Therefore (TS3, ω3) is the symplectic
reduction of (Z, ω4) at the level set corresponding to R = 12 .
Choose the model subspace W4 as described in Section 5.1.1, and construct the three-level
structure
(Q4A, δ4A)Γ4A−→ Sp(TA/TA⊥)→ A.
Then Sp(TA/TA⊥) is a principal Sp(V3) bundle over A, and Q4A is a principal Mpc(V3) bundle
over A.
Let the symplectic frame bundle Sp(TS3, ω3) be modeled on V3, so that Sp(TS3, ω3) is
a principal Sp(V3) bundle over TS3. Then Sp(TA/TA⊥) and Sp(TS3, ω3) have isomorphic
fibers. If we were to complete the symplectic reduction process, we would identify Sp(TS3, ω3)
with a quotient of Sp(TA/TA⊥). However, since TS3 ⊂ A, it is more convenient to identify
Sp(TS3, ω3) with a subbundle of Sp(TA/TA⊥).
For each z ∈ A, p∗|z appears in the short exact sequence
0→ TzA⊥ → TzA
p∗|z−→ Tp(z)TS3 → 0.
Note that the projection map p is the identity on TS3 ⊂ A. It follows that for all z ∈
TS3, p∗|z : TzA/TzA⊥ → TzTS
3 is a symplectic isomorphism. Therefore Sp(TS3, ω3) and
Sp(TA/TA⊥)|TS3 are isomorphic as principal Sp(V3) bundles over TS3. Using that isomor-
phism, we view Sp(TS3, ω3) as a subbundle of Sp(TA/TA⊥).
Let (Q,Γ, δ) be the result of restricting (Q4A,Γ4A, δ4A) to Sp(TS3, ω3) ⊂ Sp(TA/TA⊥).
Then (Q,Γ, δ) is a metaplectic-c prequantization for (TS3, ω3), and it is unique up to isomor-
phism. We will also use the notation (Q,Γ, δ) to denote the metaplectic-c prequantizations for
T+S3 and T+S3n obtained by restriction.
Chapter 5. The Hydrogen Atom 74
5.3.3 Lifting the Ligon-Schaaf Map
Recall that the Ligon-Schaaf map LS : N → T+S3n is a symplectomorphism. Define the map
LS : Sp(N,ω)→ Sp(T+S3n, ω3) by
LS(b) = LS∗ b, ∀b ∈ Sp(N,ω),
and observe that this is an isomorphism of principal Sp(V3) bundles.
In Section 5.2.2, we used the global trivialization
vj 7→∂
∂qj
∣∣∣∣m
, wj 7→∂
∂pj
∣∣∣∣m
, ∀m ∈ N, j = 1, 2, 3,
to identify Sp(N,ω) with N × Sp(V3). From the map LS, we see that Sp(T+S3n, ω3) can be
identified with T+S3n × Sp(V3) with respect to the global trivialization
vj 7→ LS∗∂
∂qj
∣∣∣∣LS(m)
, wj 7→ LS∗∂
∂pj
∣∣∣∣LS(m)
, ∀LS(m) ∈ T+S3n, j = 1, 2, 3.
In terms of these trivializations for Sp(N,ω) and Sp(T+S3n, ω3), LS is given simply by
LS(m, g) = (LS(m), g), ∀m ∈ N, ∀g ∈ Sp(V3). (5.3.2)
As we have seen before, once we have a global trivialization for the symplectic frame bundle,
it is straightforward to construct a metaplectic-c prequantization. Let Q′ = T+S3n ×Mpc(V3)
with bundle projection map Q′Π′−→ T+S3
n. Define the map Γ′ : T+S3n → Sp(T+S3
n, ω3) by
Γ′(z, a) = (z, σ(a)), ∀z ∈ T+S3n, ∀a ∈ Mpc(V3).
Let
δ′ =1
i~Π′∗β3 +
1
2η∗ϑ0,
where β3 is the restriction to T+S3n of the one-form defined in equation 5.3.1, and where ϑ0 is
the trivial connection on Q′. Then (Q′,Γ′, δ′) is a metaplectic-c prequantization for (T+S3n, ω3).
Chapter 5. The Hydrogen Atom 75
Recall that P = M ×Mpc(V3). Define LS : P → Q′ by
LS(m, a) = (LS(m), a).
This is clearly an isomorphism of principal Mpc(V3) bundles. Since LS∗β3 = β, we have
LS∗δ′ = γ. Lastly, it follows from equation 5.3.2 that Γ′ LS = LS Σ. Therefore LS is an
isomorphism of metaplectic-c prequantizations.
All of these observations combine to yield the following commutative diagram.
(P, γ)
Σ
LS // (Q′, δ′)
Γ′
Sp(N,ω)
LS // Sp(T+S3n, ω3)
(N,ω)
LS // (T+S3n, ω3)
Each of the maps LS, LS, and LS is an isomorphism. From these isomorphisms and the fact
that D LS = H, it follows that E < 0 is a quantized energy level of (N,ω,H) if and only if
it is a quantized energy level of (T+S3n, ω3, D).
Recall that we constructed the metaplectic-c prequantization (Q,Γ, δ) for (T+S3n, ω3) in the
previous section. Since T+S3n is simply connected, (Q′,Γ′, δ′) must be isomorphic to (Q,Γ, δ).
Therefore we can calculate the quantized energy levels for the system (T+S3n, ω3, D) using
(Q,Γ, δ).
In fact, we claim that E < 0 is a quantized energy level of (N,ω,H) if and only if it is
a quantized energy level of (T+S3, ω3, D). When we replace the north pole in S3, we acquire
more closed orbits: namely, those with x-components that pass through (0, 0, 0, 1). However,
due to the rotational symmetry of the system (T+S3, ω3, D) and the one-form β3, an orbit that
passes through (0, 0, 0, 1) can always be transformed into one that does not, without altering
the holonomy condition. Thus the quantized energy levels of (T+S3n, ω3, D) and (T+S3, ω3, D)
are identical.
Chapter 5. The Hydrogen Atom 76
5.3.4 Rescaling the Delaunay Hamiltonian
So far, we have shown that the negative quantized energy levels of the hydrogen atom are the
same as the quantized energy levels of the Delaunay Hamiltonian on T+S3. In this section, we
will make one final transformation that relates these energies to the quantized energy levels of
a free particle on S3.
Let K : TS3 → R be given by
K(x, y) =1
2|y|2, ∀(x, y) ∈ TS3.
In the context of classical mechanics, this energy function describes a free particle on S3. It is
shown in [3] that the corresponding Hamiltonian vector field on TS3 is
ξK =4∑j=1
(yj
∂
∂xj− |y|2xj
∂
∂yj
),
and the flow of this vector field is
ψtK(x, y) =
cos |y|t 1|y| sin |y|t
−|y| sin |y|t cos |y|t
x
y
, ∀(x, y) ∈ TS3 such that |y| > 0. (5.3.3)
If |y| = 0, then the particle is stationary, and the flow is simply ψtK(x, 0) = (x, 0). Clearly
K = 0 on the zero section of TS3, and K > 0 on T+S3.
Recall that the range of the Delaunay Hamiltonian D is R<0. Let f : R>0 → R<0 be given
by
f(z) = −mek2
4z, ∀z ∈ R>0. (5.3.4)
Note that f is a diffeomorphism, and D = f K on T+S3. Using the dynamical invariance
property of quantized energy levels, we see that E is a quantized energy level of (T+S3, ω3, D)
if and only if f−1(E) is a positive quantized energy level of (TS3, ω3,K). It remains to find the
positive quantized energy levels for a free particle on S3.
Chapter 5. The Hydrogen Atom 77
5.4 Quantization of a Free Particle on S3
5.4.1 Orbits of ξK and Local Coordinates for TS3
Let E > 0 be arbitrary, and let S = K−1(E) ⊂ TS3. From the flow of ξK in equation 5.3.3, it
is apparent that all orbits of ξK in S are closed, with period 2π√2E . Let z0 = (x0, y0) ∈ S be an
arbitrary initial point, and let C ⊂ S be the the orbit of ξK through z0:
C =ψtK(z0) : t ∈ R
⊂ TS3,
We can use the rotational symmetry of the system (TS3, ω3,K) to make some simplifying
assumptions about C. If we view x0 and y0 as two perpendicular vectors in R4, then there is
some rotation about the origin in R4 that carries them both to the x3x4-plane. By performing
this rotation in x-space and y-space, we can assume without loss of generality that x0 and
y0 take the form x0 = (0, 0, x30, x40) and y0 = (0, 0, y30, y40), where |x|2 = x230 + x2
40 = 1,
x · y = x30y30 + x40y40 = 0, and |y|2 = y230 + y2
40 = 2E . The orbit C then lies in x3x4y3y4-space.
Thus far, we have treated TS3 as a submanifold of Z = TR4, using the coordinates (x, y)
on Z to describe points in TS3. Now we make a local change of coordinates on Z that will yield
symplectic coordinates for TS3 on a neighborhood that contains C. Specifically, we introduce
4-dimensional spherical coordinates and their conjugate momenta. Let U ⊂ Z be the open set
U =
(x, y) ∈ TR4 : x23 + x2
4 > 0.
On U , let the new spatial coordinates be (a, b, c, r), where
a = arctan
√x2
2 + x23 + x2
4
x1,
b = arctan
√x2
3 + x24
x2,
c = arctanx4
x3,
r =√x2
1 + x22 + x2
3 + x24.
The angles a and b are defined modulo π, and the angle c is defined modulo 2π. Let ρ1 =
Chapter 5. The Hydrogen Atom 78
√x2
2 + x23 + x2
4 and ρ2 =√x2
3 + x24. The conjugate momenta corresponding to the spherical
coordinates are
pa =x1
ρ1(x2y2 + x3y3 + x4y4)− ρ1y1,
pb =x2
ρ2(x3y3 + x4y4)− ρ2y2,
pc =x3y4 − x4y3,
pr =1
r(x1y1 + x2y2 + x3y3 + x4y4).
Later, we will need the inverse transformations, which are
x1 = r cos a,
x2 = r sin a cos b,
x3 = r sin a sin b cos c,
x4 = r sin a sin b sin c,
(5.4.1)
y1 = pr cos a− par
sin a,
y2 = pr sin a cos b+par
cos a cos b− pbr sin a
sin b
y3 = pr sin a sin b cos c+par
cos a sin b cos c+pb
r sin acos b cos c− pc
r sin a sin bsin c,
y4 = pr sin a sin b sin c+par
cos a sin b cos c+pb
r sin acos b sin c+
pcr sin a sin b
cos c.
(5.4.2)
For convenience, we let aj , j = 1, . . . , 4, range over a, b, c, r.
On U , one can verify that
β4 = −4∑j=1
yjdxj = −4∑j=1
pajdaj , (5.4.3)
and so
ω4 =4∑j=1
dxj ∧ dyj =4∑j=1
daj ∧ dpj .
The submanifold TS3 is characterized by the constant values r = 1 and pr = 0, which implies
Chapter 5. The Hydrogen Atom 79
that the restrictions to TS3 ∩ U of β4 and ω4 are
β3 = −3∑j=1
pajdaj , ω3 =3∑j=1
daj ∧ dpaj .
Thus (a, b, c, pa, pb, pc) are symplectic coordinates for TS3 ∩U . On this neighborhood, the map
K takes the form
K =1
2
(p2a +
p2b
sin2 a+
p2c
sin2 a sin2 b
).
Several times now, once we had a set of symplectic coordinates such as (a, b, c, r, pa, pb, pc, pr),
we used the trivialization of the symplectic frame bundle given by the coordinate vector fields to
construct a metaplectic-c prequantization. We could apply this same procedure to U ; however,
since U is not simply connected, it is not necessarily the case that the metaplectic-c prequanti-
zation so constructed would be isomorphic to the result of restricting (Q4,Γ4, δ4) to U . Instead,
we must show how the local change of variables from Cartesian to spherical coordinates can be
lifted from Z to Q4. This is the subject of the next section.
5.4.2 Change of variables over C
Recall that the integral curve of ξK through the initial point z0 = (x0, y0) is
ψtK(x0, y0) =
cos√
2Et 1√2E sin
√2Et
−√
2E sin√
2Et cos√
2Et
x0
y0
,
and its image C is a closed curve lying in x3x4y3y4-space. Since C ⊂ U , the points in C can
be rewritten in spherical coordinates. Upon converting, we find that any point in C satisfies
pa = pb = 0 and a = b = π2 . Further, pc is a constant value over C satisfying p2
c = 2E . Since
C ⊂ TS3, we also have r = 1 and pr = 0. Therefore, in spherical coordinates, the orbit takes
the form
C =(π
2,π
2, c, 1, 0, 0, pc, 0
): c ∈ R/2πZ
.
Let z(c) represent the point(π2 ,
π2 , c, 1, 0, 0, pc
)∈ C.
The change of coordinates from Cartesian to spherical must now be lifted to the symplectic
Chapter 5. The Hydrogen Atom 80
frame bundle and the metaplectic-c prequantization for (Z, ω4). The change of coordinates on
Sp(Z, ω4) will take place over U , and that on Q4 will take place over C. We follow the same
steps that we applied to the harmonic oscillator in Section 4.4.2.
On the neighborhood U , we have two different coordinate maps: the Cartesian map Φc :
U → R8, and the spherical map Φs : U → R4× (R/πZ)2×R/2πZ×R. The change of variables
on U is simply the transition map F = Φs Φ−1c . Let Φc(U) = Uc and Φs(U) = Us. Then each
of the following maps is a diffeomorphism.
UΦc
~~
Φs
Uc
F // Us
Let bc be the section of Sp(Z, ω4) over U given by
bc(z) : V4 → TzZ such that vj 7→∂
∂xj
∣∣∣∣z
, wj 7→∂
∂yj
∣∣∣∣z
, ∀z ∈ U, j = 1, . . . , 4.
That is, bc is the section that defines the trivialization of Sp(Z, ω4)|U with respect to Cartesian
coordinates. Let the map Φc : Sp(Z, ω)|U → Uc × Sp(V4) be given by
Φc(bc(z) · g) = (Φc(z), g), ∀z ∈ U, ∀g ∈ Sp(V4).
Similarly, let bs be the section of Sp(Z, ω4)|U given by
bs(z) : V4 → TzZ such that vj 7→∂
∂aj
∣∣∣∣z
, wj 7→∂
∂paj
∣∣∣∣z
, ∀z ∈ U, j = 1, . . . , 4,
and define the map Φs : Sp(Z, ω)|U → Us × Sp(V4) by
Φs(bs(z) · g) = (Φs(z), g), ∀z ∈ U, ∀g ∈ Sp(V4).
To perform the change of coordinates on the level of the symplectic frame bundle, we must
lift F to a map F : Uc × Sp(V4) → Us × Sp(V4) in such a way that the following diagram
commutes.
Chapter 5. The Hydrogen Atom 81
Sp(Z, ω)|UΦc
ww
Φs
''Uc × Sp(V4)
F // Us × Sp(V4)
In equations 5.4.1 and 5.4.2, we gave explicit formulas for xj and yj in terms of ak and pak .
At each z ∈ U , let G(z) be the 8×8 matrix consisting of the partial derivatives of the Cartesian
coordinates with respect to the spherical ones:
G(z) =
∂xk∂aj
∣∣∣z
∂yk∂aj
∣∣∣z
∂xk∂paj
∣∣∣z
∂yk∂paj
∣∣∣z
1≤j,k≤4
.
Then
G(z)
∂∂xk
∣∣∣z
∂∂yk
∣∣∣z
1≤k≤4
=
∂∂aj
∣∣∣z
∂∂paj
∣∣∣z
1≤j≤4
, ∀z ∈ U.
Since the Cartesian and spherical coordinate vectors are both symplectic bases for TzZ, G(z)
is a symplectic matrix and can be treated as an element of Sp(V4). By an identical argument
to that in Section 4.4.2, the desired map F is given by
F (Φc(z), g) = (Φs(z), G(z)g), ∀z ∈ U, ∀g ∈ Sp(V4).
The final lift to the metaplectic-c prequantization will take place over C. Let Cc and Cs be
the images of C under Φc and Φs, respectively. Recall that points in C are denoted by z(c) with
c ∈ R/2πZ. We write G(c) for G(z(c)). The components of G(c) are single-valued with respect
to c, so G(c) is a closed path through Sp(V4).
It is clear from the construction of Q4 how to define a local trivialization with respect to the
Cartesian coordinates. Let Φc : Q4|C → Cc ×Mpc(V4) be given by Φc(z(c), a) = (Φc(z(c)), a)
for all z(c) ∈ C and a ∈ Mpc(V4). Then Φc and Φc make the following diagram commute.
Q4|CΓ4 //
Φc
Sp(Z, ω4)|C
Φc
Cc ×Mpc(V4)σ // Cc × Sp(V4)
We require a local trivialization of Q4 with respect to the spherical coordinates that has the
analogous relationship to Φs. To find the appropriate map Φs : Q4|C → Cs ×Mpc(V4), we will
Chapter 5. The Hydrogen Atom 82
construct a map F that satisfies the diagram shown below, then define Φs = F Φc.
Q4|CΦc
xx &&
Γ4
,,Cc ×Mpc(V4)
F //
σ
,,
Cs ×Mpc(V4)σ
,,
Sp(Z, ω4)|CΦc
ww
Φs
''Cc × Sp(V4)
F // Cs × Sp(V4)
As before, if F exists, then it has the form
F (Φc(z(c)), a) = (Φs(z(c)), G(c)a), ∀z(c) ∈ C,∀a ∈ Mpc(V4),
where G(c) ∈ Mpc(V4) and σ(G(c)) = G(c) for all c. To preserve the form of δ4, we will lift the
path G(c) to a path in Mp(V4), then check whether G(c) is single-valued with respect to c.
If G(c) ∈ Mp(V4), then its parameters take the form (G(c), µ(c)), where µ(c)2DetCCG(c) = 1.
To determine µ(c), we must calculate G(c). Using the expressions in equations 5.4.1 and
5.4.2, we compute the partial derivatives that form the matrix G, and evaluate at the point
z(c) =(π2 ,
π2 , c, 1, 0, 0, pc, 0
)∈ Cs. Using the abbreviations S(c) = sin c and C(c) = cos c, the
result is
G(c) =
−1 0 0 0 0 0 0 0
0 −1 0 0 0 0 0 0
0 0 −S(c) C(c) 0 0 −pcC(c) −pcS(c)
0 0 C(c) S(c) 0 0 pcS(c) −pcC(c)
0 0 0 0 −1 0 0 0
0 0 0 0 0 −1 0 0
0 0 0 0 0 0 −S(c) C(c)
0 0 0 0 0 0 C(c) S(c)
.
Next, we evaluate CG(c) = 12(G(c)− JG(c)J), using the matrix form for J noted in Section
Chapter 5. The Hydrogen Atom 83
5.1.1, and convert it to a 4× 4 complex matrix. We find that
CG(c) =
−1 0 0 0
0 −1 0 0
0 0 −S(c)− i2pcC(c) C(c)− i
2pcS(c)
0 0 C(c) + i2pcS(c) S(c)− i
2pcC(c)
,
which has complex determinant
DetCCG(c) = −1− 1
4p2c .
This value is real and constant over C. Thus we can define G(c) to be the element of Mp(V4) ⊂
Mpc(V4) with parameters(G(c), i
(1 + 1
4p2c
)−1/2)
, for all c ∈ R/2πZ, whereupon G(c) is the
desired closed path through Mp(V4).
Having determined G(c), we now define the map F : Cc ×Mpc(V4)→ Cs ×Mpc(V4) by
F (Φc(z(c)), a) = (Φs(z(c)), G(c)a) ∀z(c) ∈ C, ∀a ∈ Mpc(V4).
The local trivialization of Q4|C that is compatible with spherical coordinates comes about by
setting Φs = F Φc. The one-form δ4 on Q4 induces a one-form δ4s on Cs×Mpc(V4) that takes
the form
δ4s =1
i~Π∗β4 +
1
2η∗ϑ0,
where β4 is written in spherical coordinates as in equation 5.4.3, and where ϑ0 is now the trivial
connection on Cs ×Mpc(V4).
5.4.3 Restrictions to C ⊂ TS3
From now on, we no longer write the maps Φs, Φs and Φs explicitly, but write elements of Z,
Sp(Z, ω4)|U andQ4|C with respect to spherical coordinates and the local spherical trivializations.
Further, we treat TS3 as a six-dimensional manifold with local coordinates (a, b, c, pa, pb, pc) on
TS3 ∩ U , and with symplectic form ω3 =∑3
j=1 daj ∧ dpaj .
Recall from Section 5.3.2 that we defined the map R(x, y) = 12 |x|
2 on Z, and set A = R−1(12).
Chapter 5. The Hydrogen Atom 84
We argued that the symplectic frame bundle Sp(TS3, ω3) is isomorphic to Sp(TA/TA⊥)|TS3 ,
and that the metaplectic-c prequantization (Q,Γ, δ) for (TS3, ω3) is obtained by restricting
(Q4A,Γ4A, δ4A) to Sp(TA/TA⊥)|TS3 .
In spherical coordinates, we have
R(z) =1
2r2, ∀z ∈ Z ∩ U,
which has Hamiltonian vector field
ξR = r∂
∂pr.
It is immediate that at each point z ∈ A ∩ U ,
TzA = span
∂
∂a,∂
∂b,∂
∂c,∂
∂pa,∂
∂pb,∂
∂pc,∂
∂pr
∣∣∣∣z
and TzA⊥ = span
∂
∂pr
∣∣∣∣z
,
which implies that
TzA/TzA⊥ = span
[∂
∂a
],
[∂
∂b
],
[∂
∂c
],
[∂
∂pa
],
[∂
∂pb
],
[∂
∂pc
]∣∣∣∣z
.
We see that over A∩U , the local trivialization of Sp(Z, ω4)|U with respect to spherical coordi-
nates induces a local trivialization
Sp(Z, ω4;A)|A∩U = A ∩ U × Sp(V4;W4).
The group homomorphism ν : Sp(V4;W4)→ Sp(V3) then induces a local trivialization
Sp(TA/TA⊥)|A∩U = A ∩ U × Sp(V3).
Restricting further to TS3 ∩ U ⊂ A ∩ U , we find that
Sp(TS3, ω3)|TS3∩U = TS3 ∩ U × Sp(V3),
Chapter 5. The Hydrogen Atom 85
with the local trivialization given by
vj 7→∂
∂aj
∣∣∣∣z
, wj 7→∂
∂paj
∣∣∣∣z
, ∀z ∈ TS3 ∩ U, j = 1, 2, 3.
On the level of metaplectic-c bundles, we can write
QA4 |C = C ×Mpc(V4;W4)
by restricting Q4|C to Sp(Z, ω;A)|C . Then the group homomorphism ν : Mpc(V4;W4) →
Mpc(V3) induces the local trivialization
Q4A|C = Q|C = C ×Mpc(V3).
The prequantization one-form on Q is
δ =1
i~Π∗β3 +
1
2η∗ϑ0,
where ϑ0 is the trivial connection on C ×Mpc(V3), and where
β3 = −3∑j=1
pjdaj = −pcdc
over C.
5.4.4 Quantized Energy Levels for (TS3, ω3, K)
In the local coordinates (a, b, c, pa, pb, pc), the map K : TS3 → R is given by
K =1
2
(p2a +
p2b
sin2 a+
p2c
sin2 a sin2 b
),
and its Hamiltonian vector field is
ξK = pa∂
∂a+
pb
sin2 a
∂
∂b+
pc
sin2 a sin2 b
∂
∂c+
(p2b
sin3 acos a+
p2c cos a
sin3 a sin2 b
)∂
∂pa+
p2c cos b
sin2 a sin3 b
∂
∂pb.
Chapter 5. The Hydrogen Atom 86
Recall that we fixed E > 0 and defined S = K−1(E), and that z0 is a point in S. The
closed curve C is the orbit of ξK through z0. We will show that the orbit of ξK through
(z0, I) ∈ Sp(TS/TS⊥) is closed, and then we will determine the values of E for which the orbit
of ξK through (z0, I) ∈ QS is also closed.
The lift of the flow ψtK to ψtK on Sp(TS3, ω3) is given by
ψtK(z, I) = (z, ψtK∗|z), ∀z ∈ TS3 ∩ U,
which implies that the lifted vector field ξK is
ξK(z, I) =
(ξK(z),
d
dt
∣∣∣∣t=0
ψtK∗|z), ∀z ∈ TS3 ∩ U.
As a 6× 6 matrix, ddt
∣∣t=0
ψtK∗|z can be interpreted as an element of the Lie algebra sp(V3), and
its components are (d
dt
∣∣∣∣t=0
ψt∗|z)jk
=∂(ξK)j∂Xk
∣∣∣∣X=z
,
where Xk ranges over (a, b, c, pa, pb, pc). In particular, when we evaluate this matrix of partial
derivatives at z(c) ∈ C, we obtain
d
dt
∣∣∣∣t=0
ψtK∗|z(c) =
0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1
−p2c 0 0 0 0 0
0 −p2c 0 0 0 0
0 0 0 0 0 0
.
This value is constant over C; we denote it by κ. Thus
ξK(z(c), I) = (ξK(z(c)), κ), ∀z(c) ∈ C,
Chapter 5. The Hydrogen Atom 87
which implies that the flow of ξK through (z0, I) ∈ Sp(TS3, ω3) is
ψtK(z0, I) = (ψtK(z0), exp(tκ)).
Let λ =√p2c =√
2E . A calculation shows that
exp(tκ) =
cosλt 0 0 1λ sinλt 0 0
0 cosλt 0 0 1λ sinλt 0
0 0 1 0 0 0
−λ sinλt 0 0 cosλt 0 0
0 −λ sinλt 0 0 cosλt 0
0 0 0 0 0 1
.
Thus the orbit through (x0, I) is closed, with period 2πλ .
Over C, ξK reduces to
ξK = pc∂
∂c.
Therefore, for all z(c) ∈ C, we have
Tz(c)S⊥ = span
∂
∂c
∣∣∣∣z(c)
, Tz(c)S = span
∂
∂a,∂
∂b,∂
∂c,∂
∂pa,∂
∂pb
∣∣∣∣z(c)
,
and
Tz(c)S/Tz(c)S⊥ = span
[∂
∂a
],
[∂
∂b
],
[∂
∂pa
],
[∂
∂pb
]∣∣∣∣z(c)
.
We identify Sp(TS/TS⊥)|C with C × Sp(V2) in the obvious way.
Upon descending to Sp(TS/TS⊥)|C , the induced flow takes the form
ψtK(z0, I) = (ψtK(z0), ν(exp(tκ))),
Chapter 5. The Hydrogen Atom 88
where ν is the group homomorphism Sp(V3;W3)ν−→ Sp(V2). We calculate that
ν(exp(tκ)) =
cosλt 0 1λ sinλt 0
0 cosλt 0 1λ sinλt
−λ sinλt 0 cosλt 0
0 −λ sinλt 0 cosλt
.
The corresponding vector field on Sp(TS/TS⊥)|C is
ξK(z(c), I) = (ξK(z(c)), κ), ∀z(c) ∈ C,
where
κ =d
dt
∣∣∣∣t=0
ν(exp(tκ)) =
0 0 1 0
0 0 0 1
−λ2 0 0 0
0 −λ2 0 0
∈ sp(V2).
The local trivializations lift to the level of metaplectic-c bundles, so thatQS |C = C×Mpc(V2).
The induced one-form δS on QS takes the form
δS = − 1
i~pcdc+
1
2η∗ϑ0
over C. Therefore the horizontal lift of ξK to QS |C is
ξK(z(c), I) =
(ξK(z(c)), κ⊕ λ2
i~
), ∀z(c) ∈ C.
Since the mpc(V2) component is constant, the orbit of ξK through (z0, I) ∈ QS is
ψtK(z0, I) =
(ψtK(z0), exp
(tκ⊕ λ2t
i~
)).
We can write the Mpc(V2) component as exp(tκ ⊕ 0)eλ2t/i~, where exp(tκ ⊕ 0) ∈ Mp(V2) ⊂
Mpc(V2) and eλ2t/i~ ∈ U(1) ⊂ Mpc(V2).
The parameters of exp(tκ⊕0) have the form (exp tκ, µ(t)) where µ(t)2DetCCexp(tκ) = 1. As
Chapter 5. The Hydrogen Atom 89
a 2× 2 complex matrix,
Cexp(tκ) =
cosλt+ i2
(λ+ 1
λ
)sinλt 0
0 cosλt+ i2
(λ+ 1
λ
)sinλt
,
which has determinant
DetCCexp(tκ) =
(cosλt+
i
2
(λ+
1
λ
)sinλt
)2
.
This value circles the origin twice as t ranges from 0 to 2π/λ. Therefore the parameters of
exp(tκ⊕ 0) are (exp(tκ),
(cosλt+
i
2
(λ+
1
λ
)sinλt
)−1),
which describes a closed orbit in Mp(V2) with period 2π/λ.
Thus the orbit in QS is closed if and only if the U(1) term is also a closed orbit with period
2π/λ. We require eλ2t/i~ = 1 when t = 2π/λ, or equivalently, λ2
i~2πλ = −2πin for some n ∈ Z.
Rearranging and recalling that λ2 = p2c = 2E results in E = 1
2n2~2 for some n ∈ Z.
We now apply the argument made in Section 5.3.4. We are only concerned with the strictly
positive quantized energies ofK, because those correspond to quantized energies of the Delaunay
Hamiltonian D on T+S3. Thus we can ignore the energy level corresponding to n = 0, and we
can also dismiss n < 0 as redundant. This leaves us with
En =1
2n2~2, n ∈ N.
The positive value En is a quantized energy level of K if and only if f(En) is a quantized energy
level of D, where the diffeomorphism f is given in equation 5.3.4, and the quantized energy
levels of D are precisely the negative quantized energy levels of the hydrogen atom. At last, we
find that the negative quantized energy levels of the hydrogen atom are
En = f(En) = −mek2
2n2~2, n ∈ N,
which is exactly the quantum mechanical prediction for our model of the hydrogen atom.
Chapter 6
Generalization of the Quantized
Energy Condition
The quantized energy level definition presented in Chapter 4 has a natural generalization from
a single function to a family of k Poisson-commuting functions. In this chapter, we will present
the generalized definition, and prove the corresponding generalized version of the dynamical
invariance theorem. When the family of functions is of maximal size, we will show that the
quantization condition reduces to a Bohr-Sommerfeld condition.
A comment concerning notation: in this chapter, superscripts j, l range over elements of the
Poisson-commuting family, while subscripts a, c range over components with respect to local
coordinates.
6.1 The quantization condition
Let (M,ω) be a 2n-dimensional manifold, and let H1, . . . ,Hk be a family of Poisson-commuting
functions, where 1 ≤ k ≤ n. Then H = (H1, . . . ,Hk) is a map from M to Rk. Let E =
(E1, . . . , Ek) be a regular value of H, and let S = H−1(E).
Let the Hamiltonian vector field for the function Hj be denoted by ξj . Note that if a regular
value E of H exists, then the Hamiltonian vector fields ξ1, . . . , ξk are linearly independent
everywhere on S. Further, for any s ∈ S, TsS⊥ = span
ξ1(s), . . . , ξk(s)
⊂ TsS, which shows
that TsS is a coisotropic subspace of TsM .
90
Chapter 6. Generalization of the Quantized Energy Condition 91
Let (V,Ω) be a 2n-dimensional model symplectic vector space, and let W ⊂ V be a
coisotropic subspace of codimension k. Then W/W⊥ acquires a symplectic structure from
Ω. As in Section 4.2.1, we let Sp(V ;W ) be the subgroup of Sp(V ) that preserves W , and
Mpc(V ;W ) be the preimage of Sp(V ;W ) in Mpc(V ). Robinson and Rawnsley constructed a
lift of the natural group homomorphism Sp(V ;W )ν−→ Sp(W/W⊥) to the level of metaplectic-c
groups for any coisotropic subspace W [15]. Therefore we have the same commuting diagram
as we did in Section 4.2.1.
Mpc(V ) ⊃ Mpc(V ;W )ν //
σ
Mpc(W/W⊥)
σ
Sp(V ) ⊃ Sp(V ;W )ν // Sp(W/W⊥)
Assume that (M,ω) admits a metaplectic-c prequantization (P,Σ, γ). By exactly the same
process as in Section 4.2.2, we construct the following three-level structures.
(P, γ)
Σ
(PS , γS)incl.oo
ν // (PS , γS)
Sp(M,ω)
ρ
Sp(M,ω;S)incl.oo
ν // Sp(TS/TS⊥)
(M,ω) S
incl.oo = // S
The second column is obtained by restriction, and the third column is obtained by taking
associated bundles. The Hamiltonian vector fields ξ1, . . . , ξk lift to ξ1, . . . , ξk on the level of
symplectic frame bundles, and ξ1, . . . , ξk on the level of metaplectic-c bundles.
With these constructions established, we can now state the generalized version of our quan-
tized energy condition.
Definition 6.1.1. Let E = (E1, . . . , Ek) ∈ Rk be a regular value for the familyH = (H1, . . . ,Hk)
of Poisson commuting functions on M . Suppose that the one-form γS has trivial holonomy over
any closed path u(t) in Sp(TS/TS⊥) such that every tangent vector u(t) is in spanξ1, . . . , ξk
.
Then the value E is a quantized energy level for the system (M,ω,H).
In Section 6.2, we state and prove a generalized version of the dynamical invariance theorem.
The proof follows essentially the same steps as in Section 4.3.2, and as such, we omit some of the
details of the calculations. Then, in Section 6.3, we examine the special case of a completely
Chapter 6. Generalization of the Quantized Energy Condition 92
integrable system. When k = n, we will see that the quantization condition reduces to a
Bohr-Sommerfeld condition.
6.2 Generalized dynamical invariance
Let H = (H1, . . . ,Hk) and L = (L1, . . . , Lk) be two families of linearly independent Poisson
commuting functions on M . Suppose E = (E1, . . . , Ek) and F = (F 1, . . . , F k) are regular values
of H and L, respectively, such that H−1(E) = L−1(F ). We will show that the quantization
condition is identical for H and for L.
Let ξ1, . . . , ξk be the Hamiltonian vector fields corresponding to H1, . . . ,Hk, and let them
have flows φ1t, . . . , φkt. Similarly, let η1, . . . , ηk be the Hamiltonian vector fields corresponding
to L1, . . . , Lk, and let these vector fields have flows ρ1t, . . . , ρkt.
Let S = H−1(E) = L−1(F ). Then S is a 2n − k-dimensional submanifold of M . For all
s ∈ S, TsS⊥ is a k-dimensional subspace of TsS, and
TsS⊥ = span
ξ1(s), . . . , ξk(s)
= span
η1(s), . . . , ηk(s)
.
The above relationship implies that there is a function C : S → Rk×k such that
ηj(s) =k∑l=1
Cjl(s)ξl(s), ∀s ∈ S.
For any 1 ≤ j ≤ k, ξj(s) and ηj(s) are elements of TsS, which is naturally identified with
TζTsS for any ζ ∈ TsS. Since TζTsS ⊂ TζTS, we can think of ξj(s) or ηj(s) as an element of
TζTS. The following lemma is stated in terms of that identification.
Lemma 6.2.1. For all s ∈ S and all ζ ∈ TsS,
d
dt
∣∣∣∣t=0
(ρjt∗ |sζ
)=
k∑l=1
(Cjl(s)
d
dt
∣∣∣∣t=0
(φlt∗ |sζ
)+(ζCjl
)ξl(s)
),
for any 1 ≤ j ≤ k.
Proof. We work on the 2n−k-dimensional manifold S. Let U ⊂ S be a coordinate neighborhood
Chapter 6. Generalization of the Quantized Energy Condition 93
for s ∈ S, and let Z = (Z1, . . . , Z2n−k) be local coordinates on U . With respect to these
coordinates, we write φjt = (φjt1 , . . . , φjt2n−k) near s, and likewise for ρjt.
Since ηj =∑k
l=1Cjlξl on S, we have ηjb =
∑kl=1C
jlξlb on U for each 1 ≤ b ≤ 2n− k, and so
d
dt
∣∣∣∣t=0
ρjtb =k∑l=1
Cjld
dt
∣∣∣∣t=0
φltb
on U .
The pushforward φjt∗ |s is a 2n−k×2n−k matrix with entries given by(φjt∗ |s
)ac
= ∂φjta∂Zc
∣∣∣Z=s
.
For any ζ ∈ TsS,(φjt∗ |sζ
)a
= ζφjta . The same relationship holds for ρtj∗ |sζ. Taking the time
derivative yields
d
dt
∣∣∣∣t=0
(ρjt∗ |sζ
)a
=
k∑l=1
(Cjl(s)
d
dt
∣∣∣∣t=0
(φlt∗ |sζ
)a
+(ζCjl
)ξla(s)
), (6.2.1)
for 1 ≤ j ≤ k and 1 ≤ a ≤ 2n− k.
Given the coordinates Z on U , we get natural coordinates for TS|U , and thence coordinates
for TζTS. In terms of these,
d
dt
∣∣∣∣t=0
(ρjt∗ |sζ
)=
(ηj1(s), . . . , ηj2n−k(s),
d
dt
∣∣∣∣t=0
(ρjt∗ |sζ
)1, . . . ,
d
dt
∣∣∣∣t=0
(ρjt∗ |sζ
)2n−k
),
d
dt
∣∣∣∣t=0
(φjt∗ |sζ
)=
(ξj1(s), . . . , ξj2n−k(s),
d
dt
∣∣∣∣t=0
(φjt∗ |sζ
)1, . . . ,
d
dt
∣∣∣∣t=0
(φjt∗ |sζ
)2n−k
),
ξl(s) =(
0, . . . , 0, ξl(s)1, . . . , ξl(s)2n−k
).
If we apply Equation (6.2.1) in the first line and compare with the second two lines, recalling
that ηj(s) =∑k
l=1Cjl(s)ξl(s), we conclude that
d
dt
∣∣∣∣t=0
(ρjt∗ |sζ
)=
k∑l=1
(Cjl(s)
d
dt
∣∣∣∣t=0
(φlt∗ |sζ
)+(ζCjl
)ξl(s)
).
Now we use Lemma 6.2.1 to establish a relationship between the vector fields η1, . . . ηk and
ξ1, . . . , ξk on Sp(TS/TS⊥).
Chapter 6. Generalization of the Quantized Energy Condition 94
Lemma 6.2.2. For all s ∈ S and all b′ ∈ Sp(TS/TS⊥)s,
ηj(b′) =k∑l=1
Cjl(s)ξl(b′), 1 ≤ j ≤ k.
Proof. For V , choose a symplectic basis x1, . . . , xn, y1, . . . , yn. Let
W = span x1, . . . , xn, yk+1, . . . , yn .
Consequently,
W⊥ = span x1, . . . , xk , W/W⊥ = span [xk+1], . . . , [xn], [yk+1], . . . , [yn] .
For convenience, let
(z1, . . . , zn) = (x1, . . . , xn, yk+1, . . . , yn, y1, . . . , yk)
so that
W = span z1, . . . , z2n−k , W⊥ = span z1, . . . , zk , W/W⊥ = span [zk+1], . . . , [z2n−k] .
For any s ∈ S and any b ∈ Sp(M,ω)s, identify b with the basis (ζ1, . . . , ζ2n) ∈ (TsM)2n
where bzc = ζc for 1 ≤ c ≤ 2n. Note that ζc ∈ TsS for all 1 ≤ c ≤ 2n− k and ζc ∈ TsS⊥ for all
1 ≤ c ≤ k. Similarly, identify b′ ∈ Sp(TS/TS⊥)s with ([ζk+1], . . . , [ζ2n−k]) ∈(TsS/TsS
⊥)2n−2k,
where b′[zc] = [ζc] for k + 1 ≤ c ≤ 2n− k.
The lifted flow ρjt on Sp(M,ω) acts on b by
ρjt(b) = ρjt∗ |s b =(ρjt∗ |sζ1, . . . ρ
jt∗ |sζ2n
).
This element b descends to the element b′ ∈ Sp(TS/TS⊥) that is represented by ([ζk+1], . . . , [ζ2n−k]) ∈(TsS/TsS
⊥)2n−2k. The induced flow ρjt on Sp(TS/TS⊥) acts on b′ by
ρjt(b′) =([ρjt∗ |sζk+1], . . . , [ρjt∗ |sζ2n−k]
),
Chapter 6. Generalization of the Quantized Energy Condition 95
which is a path through (TS/TS⊥)2n−2k. If we take the time derivative, we get
ηj(b′) =d
dt
∣∣∣∣t=0
ρjt(b′) =
(d
dt
∣∣∣∣t=0
[ρjt∗ |sζk+1], . . . ,d
dt
∣∣∣∣t=0
[ρjt∗ |sζ2n−k]
)∈ T[ζk+1](TS/TS
⊥)× . . .× T[ζ2n−k](TS/TS⊥).
The pushforward of the projection map TS → TS/TS⊥, based at ζc, is a linear surjection
TζcTS → T[ζc](TS/TS⊥), with kernel equal to TζcTsS
⊥. Identifying this tangent space with
the vector space TsS⊥ yields a natural identification between T[ζc](TS/TS
⊥) and TζcTS/TsS⊥.
Using this, we get
ηj(b′) =
[(d
dt
∣∣∣∣t=0
ρjt∗ |sζk+1
), . . . ,
[d
dt
∣∣∣∣t=0
ρjt∗ |sζ2n−k
]]∈ Tζk+1
TS/TsS⊥ × . . .× Tζ2n−kTS/TsS
⊥.
Now, since ζc ∈ TsS for all k + 1 ≤ c ≤ 2n− k, Lemma 6.2.1 applies and we have
d
dt
∣∣∣∣t=0
ρjt∗ |sζc =
k∑l=1
(Cjl(s)
d
dt
∣∣∣∣t=0
φlt∗ |sζc + (ζcCjl)ξl(s)
)
for all c. Recall that ξl(s) ∈ TsS⊥ for all 1 ≤ l ≤ k. Therefore, upon taking equivalence classes
with respect to the quotient by TsS⊥, the terms of the form (ζcC
jl)ξl(s) all vanish and we are
left with [d
dt
∣∣∣∣t=0
ρjt∗ |sζc]
=
[k∑l=1
Cjl(s)d
dt
∣∣∣∣t=0
φjt∗ |sζc
], k + 1 ≤ c ≤ 2n− k.
It follows that
ηj(b′) =
k∑l=1
Cjl(s)ξl(b′),
as desired.
We have established that each of the Hamiltonian vector fields η1, . . . , ηk on Sp(TS/TS⊥)
is a linear combination of the vector fields ξ1, . . . , ξk. Therefore, if there is a closed curve in
Sp(TS/TS⊥) whose tangent at every point is in the subspace spanned byη1, . . . , ηk
, then that
tangent is also in the subspace spanned byξ1, . . . , ξk
, and vice versa. From our definition of
Chapter 6. Generalization of the Quantized Energy Condition 96
a quantized energy level, the following is immediate.
Theorem 6.2.3. If H = (H1, . . . ,Hk) and L = (L1, . . . , Lk) are two families of Poisson-
commuting functions on M such that H−1(E) = L−1(F ) for some regular values E of H and
F of L, then E is a quantized energy level for (M,ω,H) if and only if F is a quantized energy
level for (M,ω,L).
6.3 Completely integrable systems and Bohr-Sommerfeld con-
ditions
In this section, we consider the special case in which k = n. As we will show, when the size
of the Poisson-commuting family is maximal, the quantization condition simplifies to a Bohr-
Sommerfeld condition. First, we briefly review Bohr-Sommerfeld conditions in the context of
Kostant-Souriau quantization. Our summary is based on the more detailed treatments available
in [11, 18, 19, 24].
6.3.1 Kostant-Souriau quantization and Bohr-Sommerfeld conditions
A completely integrable system on the 2n-dimensional manifold (M,ω) is a family of n Poisson-
commuting functions H = (H1, . . . ,Hn) that are linearly independent almost everywhere. Ex-
cluding the singular points, the Hamiltonian vector fields ξ1, . . . , ξn form a real polarization:
that is, an involutive, Lagrangian subbundle of the tangent bundle TM . A leaf of the polariza-
tion is a Lagrangian submanifold of M .
Assume that (M,ω) admits a Kostant-Souriau prequantization line bundle (L,∇). A leaf
of a real polarization F is called Bohr-Sommerfeld provided there exists a global, nonvanishing
section of L over S that is horizontal in the directions of F , with respect to the connection
∇. Equivalently, a leaf S is Bohr-Sommerfeld if the connection has trivial holonomy over every
closed curve in S. Since the connection is flat over a Lagrangian submanifold, homotopic curves
have equal holonomy. Therefore it suffices to check the holonomy of ∇ over the generators of
π1(S).
Now assume that (M,ω) also admits a metaplectic structure, which implies that we can
construct the half-form bundle ∧1/2F . Then we modify the Bohr-Sommerfeld condition so that
Chapter 6. Generalization of the Quantized Energy Condition 97
the Bohr-Sommerfeld leaves are those over which there is a horizontal section of L⊗∧1/2F . This
corrected Bohr-Sommerfeld condition is necessary in order to obtain the quantum mechanical
energy levels for the harmonic oscillator [19]. In the next sections, we will show that our
quantization condition reduces to a Bohr-Sommerfeld condition when k = n, and specifically
to one that replicates the half-form correction for the harmonic oscillator.
6.3.2 Quantized energy condition for k = n
Let H = (H1, . . . ,Hn) be a family of n Poisson-commuting functions on M . As usual, let
E = (E1, . . . , En) be a regular value of H, and let S = H−1(E). The crucial observation is
that for all s ∈ S, TsS⊥ = TsS = span
ξ1(s), . . . , ξn(s)
, which implies that Sp(TS/TS⊥) has
trivial fiber and can be identified with S. The bundle (PS , γS) is a principal circle bundle with
connection one-form over Sp(TS/TS⊥), and can now be thought of as a principal circle bundle
with connection one-form over S. In particular, dγS = 0 since S ⊂M is Lagrangian.
Per our definition, E is a quantized energy level of (M,ω,H) if γS has trivial holonomy
over all closed paths in Sp(TS/TS⊥) whose tangents are in the span of the Hamiltonian vector
fields ξ1, . . . , ξn. But now Sp(TS/TS⊥) can be identified with S, and ξ1, . . . , ξn span all of TS.
Thus E is a quantized energy level of (M,ω,H) if γS has trivial holonomy over all closed paths
in S. This is a Bohr-Sommerfeld condition, as described in the previous section.
Our primary examples in previous chapters have been the harmonic oscillator and the hy-
drogen atom. In both cases, the Hamiltonian energy function can be made part of a completely
integrable system. In the remainder of this chapter, we revisit each system and apply the
generalized quantization condition to a regular level set of the completely integrable system.
6.3.3 The n-dimensional harmonic oscillator
Let M = R2n, with Cartesian coordinates (p1, . . . , pn, q1, . . . , qn) and symplectic polar coordi-
nates (s1, . . . , sn, θ1, . . . , θn). We use all of the same definitions that were established in Sections
4.4.1 and 4.4.2. In particular, the metaplectic-c prequantization for (M,ω) is (P,Σ, γ), where
P = M × Mpc(V ), Σ : P → Sp(M,ω) is defined with respect to the global trivialization of
Chapter 6. Generalization of the Quantized Energy Condition 98
Sp(M,ω) given by the symplectic frame(
∂∂p1
, . . . , ∂∂pn
, ∂∂q1, . . . , ∂
∂qn
), and
γ =1
i~Π∗β +
1
2η∗ϑ0,
where ϑ0 is the trivial connection on P and
β =1
2
n∑j=1
(pjdqj − qjdpj)
on M .
Let H = (H1, . . . ,Hn), where
Hj =1
2(p2j + q2
j ) = sj
for each 1 ≤ j ≤ n. In Cartesian coordinates, the corresponding Hamiltonian vector fields are
ξj = qj∂
∂pj− pj
∂
∂qj.
These vector fields clearly Poisson commute, and they are linearly independent unless qj = pj =
0 for some j. Thus a regular value of the family takes the form E = (E1, . . . , En) ∈ Rk, where
Ej > 0 for all j. Assume such a regular value E has been fixed, and consider the regular level
set S = H−1(E). Then S is an n-torus.
We know from Section 4.4.2 that we can change coordinates from Cartesian to symplectic
polar over the open subset of U ⊂M where sj > 0 for all j. In terms of these coordinates, we
have the Hamiltonian vector fields ξj = − ∂∂θj
for all j. Let φjt be the flow for the vector field
ξj . It is clear that all orbits in the level set S are closed, with period 2π.
Points in S take the form (E1, . . . , En, θ1, . . . , θn). If we fix a starting point, then the orbits
of the Hamiltonian vector fields ξj through that point span the fundamental group π1(S).
Therefore we need to show that γS has trivial holonomy over each of these orbits.
Our procedure is exactly the same as that in Sections 4.4.2 and 4.4.3: we locally change vari-
ables from Cartesian to symplectic polar, lift that change of variables to the level of metaplectic-c
structures, and construct local trivializations for all of the relevant bundles over an orbit C. The
Chapter 6. Generalization of the Quantized Energy Condition 99
only difference is that the curve C is now the orbit for ξj . We omit the details of the calculations
when they are repetitions of those given in Section 4.4.
Without loss of generality, fix the starting point m0 = (E1, . . . , En, 0, . . . , 0) ∈ S. Let the
orbit for ξj through m0 be C. Then a point in C takes the form
m(τ) = (E1, . . . , En, 0, . . . , τ, . . . , 0),
where τ ∈ R/2πZ is the jth angle coordinate.
For all m ∈ U , we construct the matrix of partial derivatives G(m) that corresponds to
the change of variables, just as in Section 4.4.2, and then we restrict it to lie over C. Let
G(τ) = G(m(τ)). In order to lift the change of variables to the level of metaplectic-c structures,
we calculate
DetCCG(τ) =1
2
(√2Ej +
1√2Ej
)eiτ ,
which we denote by Keiτ where K is a positive real constant. Then a lift of G(τ) to Mp(V )
has parameters
G(τ) 7→(G(τ),
1√Ke−iτ/2
).
The matrix G(τ) is single-valued with respect to τ , but e−iτ/2 is not. As τ ranges from 0
to 2π, G(τ) ranges from (I, 1) to (I,−1). We use G(τ) to perform the change of variables over
the subset C, then manually correct for the change in sign of the µ parameter when we close
the loop.
Let (x1, . . . , xn, y1, . . . , yn) be the usual symplectic basis for V . Let W = span y1, . . . , yn,
so W⊥ = W and W/W⊥ = 0. Using the local trivializations over U given by xj 7→ ∂∂sj
and
yj 7→ ∂∂θj
, we obtain the identifications Sp(M,ω)|U = U×Sp(V ), Sp(M,ω;S)|U = U×Sp(V ;W ),
and Sp(TS/TS⊥)|U = U × I = U .
Over C, the lifted change of variables gives us the identifications P |C = C ×Mpc(V ), PS |C =
C ×Mpc(V ;W ), and PS |C = C × U(1). The one-form γS takes the form
γS |C =1
i~
n∑j=1
Ejdθj +1
2η∗ϑ0.
Chapter 6. Generalization of the Quantized Energy Condition 100
Since Sp(TS/TS⊥)|U is identified with U , the lift of ξj(m(τ)) to Sp(TS/TS⊥) is simply
ξj(m(τ), I) = ξj(m(τ)).
Then the horizontal lift to PS with respect to γS is
ξj(m(τ), I) =
(ξj(m(τ)),− 1
i~Ej
).
Since the Lie algebra component is constant over the orbit, we obtain the lifted flow by expo-
nentiating:
φjt(m0, I) =(φjt(m0), e−Ejt/i~
).
On S, φjt has period 2π. For the U(1) component, we introduce a negative sign upon closing
the loop from C to C. Therefore the quantization condition is
2πiEj~
= 2πi
(Nj +
1
2
)
for some Nj ∈ N ∪ 0, which implies that
Ej = ~(Nj +
1
2
), Nj ∈ N ∪ 0 .
Note that the Hamiltonian energy function for the n-dimensional harmonic oscillator is just
H1 + . . .+Hn. Using the generalized dynamical invariance property, we see immediately that
if we describe S as a level set of the completely integrable system (H1 + . . .+Hn, H2, . . . ,Hn),
the quantized energy levels for H1 + · · ·+Hn are E1 + . . .+En = ~(N1 + 1
2 + . . .+Nn + 12
)=
~(N + n
2
), for some N ∈ N ∪ 0. This result agrees with the analysis from Section 4.4 using
just a single function, and also with the quantum mechanical calculation.
6.3.4 The 2-dimensional hydrogen atom
For simplicity, we restrict our attention to the two-dimensional version of the hydrogen atom
(equivalently, the Kepler problem). Our initial definitions are special cases of those that appear
in Section 5.2.
Chapter 6. Generalization of the Quantized Energy Condition 101
Let M = R2 × R2 with Cartesian coordinates (q1, q2, p1, p2) and symplectic form ω =∑2j=1 dqj ∧ dpj . As before, the energy function for the hydrogen atom is
H =1
2me|p|2 − k
|q|,
where k,me > 0. Let (P,Σ, γ) be the metaplectic-c prequantization for M , where P = M ×
Mpc(V4), Σ : P → Sp(M,ω) is defined with respect to the global trivialization of Sp(M,ω)
given by the Cartesian coordinate vector fields(
∂∂q1, ∂∂q2, ∂∂p1
, ∂∂p2
), and
γ =1
i~Π∗β +
1
2η∗ϑ0,
where ϑ0 is the trivial connection on P and
β =
2∑j=1
(qjdpj + d(qjpj))
on M .
Let (r, θ) be polar coordinates for R2, with conjugate momenta (pr, pθ) on R2. The change
of coordinates is given by
r =√q2
1 + q22,
θ = tan−1
(q2
q1
),
pr =1
r(q1p1 + q2p2) ,
pθ =q1p2 − q2p1,
and the inverse transformation is
q1 =r cos θ,
q2 =r sin θ,
p1 =pr cos θ − 1
rpθ sin θ,
p2 =pr sin θ +1
rpθ cos θ.
Chapter 6. Generalization of the Quantized Energy Condition 102
Note that polar coordinates are defined everywhere on M , so we could have defined Σ : P →
Sp(M,ω) in terms of the global trivialization for Sp(M,ω) given by the symplectic frame(∂∂ ,
∂∂θ ,
∂∂pr
, ∂∂pθ
), and let γ be as given above, with β written in polar coordinates. However,
since (M,ω) is not simply connected, it admits multiple nonisomorphic metaplectic-c prequan-
tizations. It is not obvious that converting from Cartesian to polar has no effect on the form of
γ. For the sake of comparison with our results in Chapter 5, we begin with the metaplectic-c
prequantization that is defined in terms of Cartesian coordinates, and convert.
At any point m ∈M , the matrix of partial derivatives representing the transformation from
Cartesian to polar coordinates is
G(m) =
∂q1∂r
∂q2∂r
∂p1∂r
∂p2∂r
∂q1∂θ
∂q2∂θ
∂p1∂θ
∂p2∂θ
∂q1∂pr
∂q2∂pr
∂p1∂pr
, ∂p2∂pr
∂q1∂pθ
∂q2∂pθ
∂p1∂pθ
∂p2∂pθ
=
cos θ sin θ 1r2pθ sin θ − 1
r2pθ cos θ
−r sin θ r cos θ −pr sin θ − 1rpθ cos θ pr cos θ − 1
rpθ sin θ
0 0 cos θ sin θ
0 0 −1r sin θ 1
r cos θ
.
To determine a lift to Mp(V4), we calculate CG(m), convert to a complex matrix, and take the
determinant. The result is
DetCCG(m) = r +1
r+
1
4r3p2θ.
While this is not necessarily constant, it is real and positive everywhere on M . Therefore, over
any closed orbit in M , we can consistently take√
DetCCG(m). Thus a lift of G(m) to Mp(V4)
is given by
G(m) =
(G(m),
1√DetCCG(m)
),
which is single-valued over any closed orbit in M . This demonstrates that we can, in fact,
convert to polar coordinates without affecting the form of γ.
From now on, we use the global trivializations of P and Sp(M,ω) given by the symplectic
Chapter 6. Generalization of the Quantized Energy Condition 103
frame(∂∂r ,
∂∂θ ,
∂∂pr
, ∂∂pθ
). The one-form γ is given by
γ =1
i~Π∗β +
1
2ϑ0,
where
β = 2rdpr + prdr − pθdθ
on M .
In terms of the polar coordinates,
H =1
2me
(p2r +
1
r2p2θ
)− k
r,
and ω = dpr ∧ dr + dpθ ∧ dθ. A calculation establishes that
ξH =
(k
r2− 1
mer3p2θ
)∂
∂pr− 1
mepr∂
∂r− 1
mer2pθ∂
∂θ.
Now consider pθ. Physically, pθ is the angular momentum of the system. Let E < 0 be
a fixed negative energy value. As discussed in Section 5.2.1, the angular momentum can take
values in
[0,√−mek2
2E
]. The value pθ = 0 represents a degenerate elliptical orbit in which the
electron collapses into the proton in finite time. The value pθ =√−mek2
2E represents a circular
orbit.
Observe that
ξpθ = − ∂
∂θ,
from which we see that H, pθ = 0. Therefore the system (H, pθ) Poisson commutes. Further,
ξH and ξpθ are linearly independent unless pr = 0 and kr2− 1
mer3p2θ = 0. These two conditions
imply that r = − k2E and p2
θ = −mek2
2E : that is, ξH and ξpθ are linearly independent everywhere
except on the circular orbit.
Let (E,L) ∈ R2 be such that E < 0 and 0 < L <√−mek2
2E . Then (E,L) is a regular value
of the family (H, pθ). Let S be the corresponding level set. Then, from Section 5.2.1, all orbits
of ξH in S are closed with period 2πΛ , where Λ =
√− 8E3
mek2. All orbits of ξpθ in S are closed with
period 2π.
Chapter 6. Generalization of the Quantized Energy Condition 104
Given a starting point m0 ∈ S, the orbits of ξH and ξpθ generate π1(S). Since S is a
Lagrangian submanifold of M , we have TS⊥ = TS, and so Sp(TS/TS⊥) is naturally identified
with S. We must therefore evaluate the holonomy of γS over the orbits ξH and ξpθ through m0.
Let ξH have flow φtH , and let pθ have flow φtpθ on M .
First, we consider ξpθ . Let m(τ) be a point on the orbit of ξpθ through m0, where τ ∈ [0, 2π]
and m(0) = m0. On Sp(TS/TS⊥), we simply have
ξpθ(m(τ), I) = ξpθ(m(τ)).
Using the expression for β in polar coordinates, we calculate that the horizontal lift to PS is
ξpθ(m(τ), I) =
(ξpθ(m(τ)),− L
i~
).
Upon exponentiating, we obtain the integral curve through (m0, I):
φtpθ(m0, I) =(φtpθ , e
−Lt/i~).
Since the orbit in S has period 2π, one sees immediately that the quantization condition for
the angular momentum is
L = N~, N ∈ Z.
This is consistent with the Bohr model of the hydrogen atom, in which angular momentum is
assumed to be quantized in units of ~.
Now we apply the same process to ξH . Let m(τ) be a point on the orbit of ξH through m0,
where now τ ∈[0, 2π
Λ
]and m(0) = m0. Since
ξHyβ =2k
r− 1
me
(p2r +
1
r2p2θ
)= −2E,
we see that the lift to PS is
ξH(m(τ), I) =
(ξH(m(τ)),
2E
i~
).
Chapter 6. Generalization of the Quantized Energy Condition 105
Then the integral curve through m0 is
φtH(m0, I) =(φtH(m0), e2Et/i~
).
Using the fact that the period of the orbit in S is 2πΛ where Λ =
√− 8E3
mek2, we find that the
quantization condition for the energy is
E = − mek2
2~2N2, N ∈ N.
This agrees with our result from Chapter 5.
Bibliography
[1] V.I. Arnold, Mathematical Methods of Classical Mechanics, Second Edition, trans. K. Vogt-
mann and A. Weinstein, Springer-Verlag New York, 1989.
[2] R.J. Blattner, Quantization and representation theory, Proc. Sym. Pure Math. vol. 26,
1973, pp. 147-165.
[3] R. Cushman and L. Bates, Global Aspects of Classical Integrable Systems, Birkhauser
Verlag, Basel, 1997.
[4] A. Cannas da Silva, Y. Karshon, and S. Tolman, Quantization of presymplectic manifolds
and circle actions, Trans. Amer. Math. Soc. 352 (2000), no. 2, pp. 525-552.
[5] Dirac P.A.M., The Fundamental Equations of Quantum Mechanics, Proc. Roy. Soc. Lond.
A Vol. 109 (1925), 642–53.
[6] C. Duval, J. Elhadad and G.M. Tuynman, Hyperfine interaction in a classical hydrogen
atom and geometric quantization, J. Geom. Phys. 3 no. 3, 1986, pp. 401-420.
[7] S. Fuchs, Additivity of spinc-quantization under cutting, Trans. Amer. Math. Soc. 361
(2009), no. 10, pp. 5345-5376.
[8] V. Ginzburg, V. Guillemin, and Y. Karshon, Moment maps, cobordisms, and Hamiltonian
group actions, American Mathematical Society, Providence, R.I., 2002.
[9] M. Grossberg and Y. Karshon, Equivariant Index and the Moment Map for Completely
Integrable Torus Actions, Adv. Math. 133 (1998), no. 2, pp. 185-223.
106
Bibliography 107
[10] V. Guillemin and S. Sternberg, Geometric Asymptotics, American Mathematical Society,
Providence, R.I., 1977.
[11] M. Hamilton, Singular Bohr-Sommerfeld Leaves and Geometric Quantization (doctoral
dissertation), 2005.
[12] Kostant B., Quantization and unitary representations, in Lectures in Modern Analysis
and Applications III, Editor C.T. Taam, Lecture Notes in Mathematics Vol. 170, Springer
Berlin Heidelberg, 1970, 87–208.
[13] T. Ligon and M. Schaaf, On the global symmetry of the classical Kepler problem, Rep.
Math. Phys. 9, 1976, pp. 281-300.
[14] P.L. Robinson, Mpc structures and energy surfaces, Quart. J. Math. 41 (1990), no. 3,
pp. 325-334.
[15] Robinson P.L., Rawnsley J.H., The metaplectic representation, Mpc-structures and geo-
metric quantization, Memoirs of the A.M.S. Vol. 81, no. 410 (1989).
[16] D.J. Simms, Bohr-Sommerfeld orbits and quantizable symplectic manifolds, Math. Proc.
Camb. Phil. Soc. 73 (1973), no. 3, pp. 489-491.
[17] D.J. Simms and N.M.J. Woodhouse, Lectures on Geometric Quantization, Springer-Verlag,
Berlin, 1976.
[18] J. Sniatycki, Geometric Quantization and Quantum Mechanics, Springer-Verlag New York,
1980.
[19] J. Sniatycki, Bohr-Sommerfeld Conditions in Geometric Quantization, Rep. Math. Phys.
7 no. 2, 1975, pp. 303-311.
[20] Souriau J.-M., Structure of Dynamical Systems: A Symplectic View Of Physics, trans-
lated by C.H. Cushman-de Vries, translation editors R.H. Cushman and G.M. Tuynman,
Birkhauser, Boston, 1997.
[21] J. Vaughan, “Metaplectic-c Quantomorphisms,” Symmetry, Integrability, and Geometry:
Methods and Applications 11 (025), 2015, 16 pages. arXiv:1410:5529.
Bibliography 108
[22] J. Vaughan, “Dynamical Invariance of a New Metaplectic-c Quantization Condition,” 2015.
arXiv:1507.06720.
[23] J. Vaughan, “Metaplectic-c Quantized Energy Levels of the Hydrogen Atom,” 2015.
arXiv:1510:03115.
[24] Woodhouse N.M.J., Geometric Quantization, Second Edition, Clarendon Press, Oxford,
1991.