This action might not be possible to undo. Are you sure you want to continue?
Contents
1 Introduction: The Harmonic Oscillator 5
2 The WKB Method 8
2.1 Some HamiltonJacobi preliminaries . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 The WKB approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3 Symplectic Manifolds 17
3.1 Symplectic structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Cotangent bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3 Mechanics on manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4 Quantization in Cotangent Bundles 36
4.1 Prequantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2 The Maslov correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.3 Phase functions and lagrangian submanifolds . . . . . . . . . . . . . . . . . . 46
4.4 WKB quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5 The Symplectic Category 64
5.1 Symplectic reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.2 The symplectic category . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.3 Symplectic manifolds and mechanics . . . . . . . . . . . . . . . . . . . . . . 79
6 Fourier Integral Operators 83
6.1 Compositions of semiclassical states . . . . . . . . . . . . . . . . . . . . . . 83
6.2 WKB quantization and compositions . . . . . . . . . . . . . . . . . . . . . . 87
7 Geometric Quantization 93
7.1 Prequantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
7.2 Polarizations and the metaplectic correction . . . . . . . . . . . . . . . . . . 98
7.3 Quantization of semiclassical states . . . . . . . . . . . . . . . . . . . . . . . 109
8 Algebraic Quantization 111
8.1 Poisson algebras and Poisson manifolds . . . . . . . . . . . . . . . . . . . . . 111
8.2 Deformation quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
8.3 Symplectic groupoids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
A Densities 119
B The method of stationary phase 121
C
ˇ
Cech cohomology 124
D Principal T
bundles 125
2
Preface
These notes are based on a course entitled “Symplectic geometry and geometric quantization”
taught by Alan Weinstein at the University of California, Berkeley, in the fall semester of
1992 and again at the Centre Emile Borel (Institut Henri Poincar´e) in the spring semester
of 1994. The only prerequisite for the course (and for these notes) was a knowledge of
the basic notions from the theory of diﬀerentiable manifolds (diﬀerential forms, vector ﬁelds,
transversality, etc.). The aim of the course was to give students an introduction to the ideas of
microlocal analysis and the related symplectic geometry, with an emphasis on the role which
these ideas play in formalizing the transition between the mathematics of classical dynamics
(hamiltonian ﬂows on symplectic manifolds) and that of quantum mechanics (unitary ﬂows
on Hilbert spaces).
There already exist many books on the subjects treated here, but most of them provide
too much detail for the reader who just wants to ﬁnd out what the subject is about. These
notes are meant to function as a guide to the literature; we refer to other sources for many
details which are omitted here, and which can be bypassed on a ﬁrst reading.
The pamphlet [63] is in some sense a precursor to these notes. On the other hand, a much
more complete reference on the subject, written at about the same time, is [28]. An earlier
work, one of the ﬁrst to treat the connections between classical and quantum mechanics from
a geometric viewpoint, is [41]. The book [29] treats further topics in symplectic geometry
and mechanics, with special attention to the role of symmetry groups, a topic pretty much
ignored in the present notes. For more extensive treatment of the PDE aspects of the subject,
we refer to [43] for a physicsoriented presentation and to the notes [21] and the treatises
[32], [46], and [56]. For “geometric quantization”, one may consult [35], [53], [54], [60] or
[71]. For classical mechanics and symplectic geometry, we suggest [1], [2], [6], [8], [25], [38],
[59]. Finally, two basic references on quantum mechanics itself are [13] and [20].
Although symplectic geometry is like any ﬁeld of mathematics in having its deﬁnitions,
theorems, etc., it is also a special way of looking at a very broad part of mathematics and
its applications. For many “symplecticians”, it is almost a religion. A previous paper by
one of us [64] referred to “the symplectic creed”.
1
In these notes, we show how symplectic
geometry arises from the study of semiclassical solutions to the Schr¨odinger equation, and in
turn provides a geometric foundation for the further analysis of this and other formulations
of quantum mechanics.
These notes are still not in ﬁnal form, but they have already beneﬁtted from the comments
1
We like the following quotation from [4] very much:
In recent years, symplectic and contact geometries have encroached on all areas of mathemat
ics. As each skylark must display its comb, so every branch of mathematics must ﬁnally display
symplectisation. In mathematics there exist operations on diﬀerent levels: functions acting on
numbers, operators acting on functions, functors acting on operators, and so on. Symplec
tisation belongs to the small set of highest level operations, acting not on details (functions,
operators, functors), but on all the mathematics at once. Although some such highest level
operations are presently known (for example, algebraisation, Bourbakisation, complexiﬁcation,
superisation, symplectisation) there is as yet no axiomatic theory describing them.
3
and suggestions of many readers, especially Maurice Garay, Jim Morehead, and Dmitry
Roytenberg. We welcome further comments. We would like to thank the Centre Emile Borel
and the Isaac Newton Institute for their hospitality. During the preparation of these notes,
S.B. was supported by NSF graduate and postdoctoral fellowships in mathematics. A.W.
was partially supported by NSF Grants DMS9001089 and 9301089.
4
1 Introduction: The Harmonic Oscillator
In these notes, we will take a “spiral” approach toward the quantization problem, beginning
with a very concrete example and its proposed solution, and then returning to the same
kind of problem at progressively higher levels of generality. Speciﬁcally, we will start with
the harmonic oscillator as described classically in the phase plane R
2
and work toward the
problem of quantizing arbitrary symplectic manifolds. The latter problem has taken on a
new interest in view of recent work by Witten and others in the area of topological quantum
ﬁeld theory (see for example [7]).
The classical picture
The harmonic oscillator in 1 dimension is described by Newton’s diﬀerential equation:
m¨ x = −kx.
By a standard procedure, we can convert this secondorder ordinary diﬀerential equation
into a system of two ﬁrstorder equations. Introducing the “phase plane” R
2
with position
and momentum coordinates (q, p), we set
q = x p = m˙ x,
so that Newton’s equation becomes the pair of equations:
˙ q =
p
m
˙ p = −kq.
If we now introduce the hamiltonian function H: R
2
→R representing the sum of kinetic
and potential energies,
H(q, p) =
p
2
2m
+
kq
2
2
then we ﬁnd
˙ q =
∂H
∂p
˙ p = −
∂H
∂q
These simple equations, which can describe a wide variety of classical mechanical systems
with appropriate choices of the function H, are called Hamilton’s equations.
2
Hamilton’s
equations deﬁne a ﬂow on the phase plane representing the timeevolution of the classical
system at hand; solution curves in the case of the harmonic oscillator are ellipses centered
at the origin, and points in the phase plane move clockwise around each ellipse.
We note two qualitative features of the hamiltonian description of a system:
1. The derivative of H along a solution curve is
dH
dt
=
∂H
∂q
˙ q +
∂H
∂p
˙ p = −˙ p ˙ q + ˙ q ˙ p = 0,
2
If we had chosen ˙ x rather than m˙ x as the second coordinate of our phase plane, we would not have
arrived at this universal form of the equations.
5
i.e., the value of H is constant along integral curves of the hamiltonian vector ﬁeld.
Since H represents the total energy of the system, this property of the ﬂow is interpreted
as the law of conservation of energy.
2. The divergence of the hamiltonian vector ﬁeld X
H
= ( ˙ q, ˙ p) = (
∂H
∂p
, −
∂H
∂q
) is
∇ X
H
=
∂
2
H
∂q ∂p
−
∂
2
H
∂p ∂q
= 0.
Thus the vector ﬁeld X
H
is divergencefree, and its ﬂow preserves area in the phase
plane.
The description of classical hamiltonian mechanics just given is tied to a particular coordinate
system. We shall see in Chapter 3 that the use of diﬀerential forms leads to a coordinate
free description and generalization of the hamiltonian viewpoint in the context of symplectic
geometry.
The quantum mechanical picture
In quantum mechanics, the motion of the harmonic oscillator is described by a complex
valued wave function ψ(x, t) satisfying the 1dimensional Schr¨ odinger equation:
i
∂ψ
∂t
= −
2
2m
∂
2
ψ
∂x
2
+
k
2
x
2
ψ.
Here, Planck’s constant has the dimensions of action (energy time). Interpreting the
right hand side of this equation as the result of applying to the wave function ψ the operator
ˆ
H
def
= −
2
2m
∂
2
∂x
2
+
k
2
m
x
2,
where m
x
2 is the operator of multiplication by x
2
, we may rewrite the Schr¨odinger equation
as
i
∂ψ
∂t
=
ˆ
Hψ.
A solution ψ of this equation does not represent a classical trajectory; instead, if ψ is
normalized, i.e.
R
ψ
∗
ψ = 1,
then its squarenorm
ρ(x, t) = [ψ(x, t)[
2
is interpreted as a probability density for observing the oscillator at the position x at time
t. The wave function ψ(x, t) itself may be viewed alternatively as a tdependent function of
x, or as a path in the function space C
∞
(R, C). From the latter point of view, Schr¨odinger’s
equation deﬁnes a vector ﬁeld on C
∞
(R, C) representing the time evolution of the quantum
system; a wave function satisfying Schr¨odinger’s equation then corresponds to an integral
curve of the associated ﬂow.
Like Hamilton’s equations in classical mechanics, the Schr¨odinger equation is a general
form for the quantum mechanical description of a large class of systems.
6
Quantization and the classical limit
The central aim of these notes is to give a geometric interpretation of relationships between
the fundamental equations of classical and quantum mechanics. Based on the present dis
cussion of the harmonic oscillator, one tenuous connection can be drawn as follows. To the
classical position and momentum observables q, p we associate the diﬀerential operators
q → ˆ q = m
x
p → ˆ p = −i
∂
∂x
.
The classical hamiltonian H(q, p) = p
2
/2m+kq
2
/2 then corresponds naturally to the operator
ˆ
H. As soon as we wish to “quantize” a more complicated energy function, such as (1+q
2
)p
2
,
we run in to the problem that the operators ˆ q and ˆ p do not commute with one another, so
that we are forced to choose between (1 + ˆ q
2
)ˆ p
2
and ˆ p
2
(1 + ˆ q
2
), among a number of other
possibilities. The diﬀerence between these choices turns out to become small when → 0.
But how can a constant approach zero?
Besides the problem of “quantization of equations,” we will also treat that of “quan
tization of solutions.” That is, we would like to establish that, for systems which are in
some sense macroscopic, the classical motions described by solutions of Hamilton’s equa
tions lead to approximate solutions of Schr¨odinger’s equation. Establishing this relation
between classical and quantum mechanics is important, not only in verifying that the theo
ries are consistent with the fact that we “see” classical behavior in systems which are “really”
governed by quantum mechanics, but also as a tool for developing approximate solutions to
the quantum equations of motion.
What is the meaning of “macroscopic” in mathematical terms? It turns out that good
approximate solutions of Schr¨odinger’s equation can be generated from classical information
when is small enough. But how can a constant with physical dimensions be small?
Although there remain some unsettled issues connected with the question, “How can
become small?” the answer is essentially the following. For any particular mechanical
system, there are usually characteristic distances, masses, velocities, . . . from which a unit
of action appropriate to the system can be derived, and the classical limit is applicable
when divided by this unit is much less than 1. In these notes, we will often regard
mathematically as a formal parameter or a variable rather than as a ﬁxed number.
7
2 The WKB Method
A basic technique for obtaining approximate solutions to the Schr¨odinger equation from
classical motions is called the WKB method, after Wentzel, Kramers, and Brillouin. (Other
names, including Liouville, Green, and Jeﬀreys are sometimes attached to this method.
References [13] and [47] contain a discussion of some of its history. Also see [5], where the
method is traced all the way back to 1817. For convenience, nevertheless, we will still refer
to the method as WKB.) A good part of what is now called microlocal analysis can be
understood as the extension of the basic WKB idea to more precise approximations and
more general situations, so the following discussion is absolutely central to these notes.
2.1 Some HamiltonJacobi preliminaries
In this section we will carry out the ﬁrst step in the WKB method to obtain an approximate
solution to the “stationary state” eigenvalue problem arising from the Schr¨odinger equation.
The geometric interpretation of this technique will lead to a correspondence between classical
and quantum mechanics which goes beyond the one described in Chapter 1.
Consider a 1dimensional system with hamiltonian
H(q, p) =
p
2
2m
+ V (q),
where V (x) is a potential (for example the potential kx
2
/2 for the harmonic oscillator).
Hamilton’s equations now become
˙ q =
p
m
˙ p = −V
t
(q).
For ﬁxed ∈ R
+
, Schr¨odinger’s equation assumes the form
i
∂ψ
∂t
=
ˆ
Hψ,
where
ˆ
H = −
2
2m
∂
2
∂x
2
+ m
V
is the Schr¨ odinger operator.
As a ﬁrst step toward solving the Schr¨odinger equation, we look for stationary states,
i.e. solutions of the form
ψ(x, t) = ϕ(x) e
−iωt
,
socalled because as time evolves, these solutions keep the same form (up to multiplication by
a complex scalar of norm 1). Substituting this expression for ψ in the Schr¨odinger equation,
we obtain
ω ϕ(x) e
−iωt
= (
ˆ
Hϕ)(x) e
−iωt
.
Eliminating the factor e
−iωt
above, we arrive at the timeindependent Schr¨ odinger equa
tion:
(
ˆ
H −E) ϕ = 0,
8
where E = ω. This equation means that ϕ is to be an eigenfunction of the linear diﬀerential
operator
ˆ
H; the eigenvalue E represents the energy of the system, which has a “deﬁnite value”
in this state.
Suppose for the moment that the potential V is constant, in which case the force −V
t
(x)
is zero, and so we are dealing with a free particle. Trying a solution of the form ϕ(x) = e
ixξ
for some constant ξ, we ﬁnd that
(
ˆ
H −E) ϕ = 0 ⇔ (ξ)
2
= 2m(E −V ).
For V < E, the (real) value of ξ is thus determined up to a choice of sign, and one has an
abundance of exact solutions of the Schr¨odinger equation which are oscillatory and bounded.
Such a wave function is not squareintegrable and as such is said to be “unnormalizable”; it
represents a particle which is equally likely to be anywhere in space, but which has a deﬁnite
momentum (since it is an eigenfunction of the momentum operator ˆ p).
3
When E < V , the
constant ξ is imaginary, and there are only real exponential solutions, which are unbounded
and admit no physical interpretation.
The basic idea at this stage of the WKB method is that, if V varies with x, then ξ should
vary with x as well; a more general solution candidate is then
ϕ(x) = e
iS(x)/
,
for some realvalued function S known as a phase function. This proposed form of the
solution is the simplest version of the WKB ansatz, and in this case we have
(
ˆ
H −E) ϕ =
¸
(S
t
(x))
2
2m
+ (V −E) −
i
2m
S
tt
(x)
e
iS(x)/
.
Since we will consider to be “small”, our ﬁrstorder approximation attempt will ignore the
last term in brackets; to kill the other two terms, we require that S satisfy the eikonal or
HamiltonJacobi equation:
H(x, S
t
(x)) =
(S
t
(x))
2
2m
+ V (x) = E,
i.e.
S
t
(x) = ±
2m(E −V (x)).
To understand the phase function S geometrically, we consider the classical phase
4
plane
R
2
· T
∗
R with coordinates (q, p). The diﬀerential dS = S
t
dx can be viewed as a mapping
dS: R →T
∗
R, where as usual we set p = S
t
. Then S satisﬁes the HamiltonJacobi equation
if and only if the image of dS lies in the level manifold H
−1
(E). This observation establishes
a fundamental link between classical and quantum mechanics:
When the image of dS lies in a level manifold of the classical hamiltonian, the
function S may be taken as the phase function of a ﬁrstorder approximate solu
tion of Schr¨odinger’s equation.
3
See [55] for a grouptheoretic interpretation of such states.
4
These two uses of the term “phase” seem to be unrelated!
9
The preceding discussion generalizes easily to higher dimensions. In R
n
, the Schr¨odinger
operator corresponding to the classical hamiltonian
H(q, p) =
¸
p
2
i
2m
+ V (q)
is
ˆ
H = −
2
2m
∆ + m
V
,
where ∆ denotes the ordinary Laplacian. As before, if we consider a WKB ansatz of the
form ϕ = e
iS/
, then
(
ˆ
H −E) ϕ =
¸
∇S
2
2m
+ (V −E) −
i
2m
∆S
e
iS/
will be O() provided that S satisﬁes the HamiltonJacobi equation:
H
x
1
, . . . , x
n
,
∂S
∂x
1
, . . . ,
∂S
∂x
n
=
∇S(x)
2
2m
+ V (x) = E.
Since ϕ is of order zero in , while (
ˆ
H − E) ϕ = O(), the ansatz ϕ again constitutes a
ﬁrstorder approximate solution to the timeindependent Schr¨odinger equation.
We will call a phase function S : R
n
→R admissible if it satisﬁes the HamiltonJacobi
equation. The image L = im(dS) of the diﬀerential of an admissible phase function S is
characterized by three geometric properties:
1. L is an ndimensional submanifold of H
−1
(E).
2. The pullback to L of the form α
n
=
¸
j
p
j
dq
j
on R
2n
is exact.
3. The restriction of the canonical projection π : T
∗
R
n
→ R
n
to L induces a diﬀeomor
phism L · R
n
. In other words, L is projectable.
While many of the basic constructions of microlocal analysis are motivated by operations
on these projectable submanifolds of T
∗
R
n
· R
2n
, applications of the theory require us to
extend the constructions to more general ndimensional submanifolds of R
2n
satisfying only
a weakened version of condition (2) above, in which “exact” is replaced by “closed”. Such
submanifolds are called lagrangian.
For example, the level sets for the 1dimensional harmonic oscillator are lagrangian sub
manifolds in the phase plane. A regular level curve of the hamiltonian is an ellipse L. Since
L is 1dimensional, the pullback to L of the form p dq is closed, but the integral of p dq
around the curve equals the enclosed nonzero area, so its pullback to L is not exact. It is
also clear that the curve fails to project diﬀeomorphically onto R. From the classical stand
point, the behavior of an oscillator is nevertheless completely described by its trajectory,
suggesting that in general the state of a system should be represented by the submanifold
L (projectable or not) rather than by the phase function S. This idea, which we will clarify
later, is the starting point of the geometrical approach to microlocal analysis.
10
For now, we want to note an important relationship between lagrangian submanifolds of
R
2n
and hamiltonian ﬂows. Recall that to a function H : R
2n
→ R, Hamilton’s equations
associate the vector ﬁeld
X
H
= ˙ q
∂
∂q
+ ˙ p
∂
∂p
=
¸
j
∂H
∂p
j
∂
∂q
j
−
∂H
∂q
j
∂
∂p
j
.
A simple computation shows that X
H
and the form α
n
are related by the equation
X
H
dα
n
= −dH,
i.e.
dα
n
(X
H
, v) = −dH(v)
for every tangent vector v. If L is a lagrangian submanifold of a level set of H, then TL
lies in the kernel of dH at all points of L, or, in other words, the 2form dα
n
vanishes on
the subspace of T
p
R
2n
generated by T
p
L and X
H
(p) for each p ∈ L. The restriction of dα
n
to the tangent space T
p
R
2n
of R
2n
at any point p deﬁnes a nondegenerate, skewsymmetric
bilinear form, and thus, as we will see in the next chapter, subspaces of T
p
R
2n
on which dα
n
vanishes can be at most ndimensional. These remarks imply that X
H
is tangent to L, and
we have the following result.
HamiltonJacobi theorem . A function H: R
2n
→R is locally constant on a lagrangian
submanifold L ⊂ R
2n
if and only if the hamiltonian vector ﬁeld X
H
is tangent to L.
If the lagrangian submanifold L is locally closed, this theorem implies that L is invariant
under the ﬂow of X
H
.
2.2 The WKB approximation
Returning to our WKB ansatz for a stationarystate solution of the Schr¨odinger equation,
we recall that if S: R
n
→R is an admissible phase function, then ϕ(x) = e
iS(x)/
satisﬁes
(
ˆ
H −E) ϕ = O().
Up to terms of order , in other words, ϕ is an eigenfunction of
ˆ
H with eigenvalue E.
There is no way to improve the order of approximation simply by making a better choice
of S. It is also clear on physical grounds that our ansatz for ϕ is too restrictive because
it satisﬁes [ϕ(x)[ = 1 for all x. In quantum mechanics, the quantity [ϕ(x)[
2
represents the
probability of the particle being at the position x, and there is no reason for this to be
constant; in fact, it is at least intuitively plausible that a particle is more likely to be found
where it moves more slowly, i.e., where its potential energy is higher. We may therefore hope
to ﬁnd a better approximate solution by multiplying ϕ by an “amplitude function” a
ϕ(x) = e
iS(x)/
a(x).
11
If S is again an admissible phase function, we now obtain:
(
ˆ
H −E) ϕ = −
1
2m
¸
i
a∆S + 2
¸
j
∂a
∂x
j
∂S
∂x
j
+
2
∆a
¸
e
iS/
.
If a is chosen to kill the coeﬃcient of on the right, then ϕ will be an eigenfunction of
ˆ
H
modulo terms of order O(
2
). This condition on a is known as the homogeneous transport
equation:
a∆S + 2
¸
j
∂a
∂x
j
∂S
∂x
j
= 0.
If S is an admissible phase function, and a is an amplitude which satisﬁes the homogeneous
transport equation, then the secondorder solution ϕ = e
iS/
a is called the semiclassical
approximation.
Example 2.1 In 1 dimension, the homogeneous transport equation amounts to
aS
tt
+ 2a
t
S
t
= 0.
Solving this equation directly, we obtain
a
2
S
tt
+ 2aa
t
S
t
= (a
2
S
t
)
t
= 0
⇒ a =
c
√
S
t
for some constant c. Since S is assumed to satisfy the HamiltonJacobi equation, we have
S
t
=
2m(E −V ), and thus
a =
c
[4(E −V )]
1
4
.
If E > V (x) for all x ∈ R, then a is a smooth solution to the homogeneous transport equation.
Notice that a = [ϕ[ is largest where V is largest, as our physical reasoning predicted.
Since the expression above for a does not depend explicitly on the phase function S, we
might naively attempt to use the same formula when im(dS) is replaced by a nonprojectable
lagrangian submanifold of H
−1
(E). Consider, for example, the unbounded potential V (x) =
x
2
in the case of the harmonic oscillator. For [x[ <
√
E, the function a is still welldeﬁned
up to a multiplicative constant. At [x[ =
√
E, however, a has (asymptotic) singularities;
observe that these occur precisely at the projected image of those points of L where the
projection itself becomes singular. Outside the interval [x[ ≤
√
E, the function a assumes
complex values.
´
To generate better approximate solutions to the eigenfunction problem, we can extend
the procedure above by adding to the original amplitude a = a
0
certain appropriately chosen
functions of higher order in . Consider the next approximation
ϕ = e
iS/
(a
0
+ a
1
).
12
Assuming that e
iS/
a
0
is a semiclassical approximate solution, we obtain:
(
ˆ
H −E) ϕ = −
1
2m
¸
i
2
a
1
∆S + 2
¸
j
∂a
1
∂x
j
∂S
∂x
j
−i∆a
0
+
3
∆a
1
¸
e
iS(x)/
.
Evidently, ϕ will be a solution of the timeindependent Schr¨odinger equation modulo terms
of order O(
3
) provided that a
1
satisﬁes the inhomogeneous transport equation
a
1
∆S + 2
¸
j
∂a
1
∂x
j
∂S
∂x
j
= i∆a
0
.
In general, a solution to the eigenfunction problem modulo terms of order O(
n
) is given
by a WKB ansatz of the form
ϕ = e
iS/
(a
0
+ a
1
+ + a
n
n
),
where S satisﬁes the HamiltonJacobi equation, a
0
satisﬁes the homogeneous transport equa
tion, and for each k > 0, the function a
k
satisﬁes the inhomogeneous transport equation:
a
k
∆S + 2
¸
j
∂a
k
∂x
j
∂S
∂x
j
= i∆a
k−1
.
This situation can be described in the terminology of asymptotic series as follows. By an
dependent function f
on R
n
we will mean a function f : R
n
R
+
→ C, where is
viewed as a parameter ranging in R
+
. Such a function is said to be represented by a
formal asymptotic expansion of the form
¸
∞
k=0
a
k
k
, where each coeﬃcient a
k
is a smooth
complexvalued function on R
n
, if, for each K ∈ Z
+
, the diﬀerence
f
−
K
¸
k=0
a
k
k
is O(
K+1
) locally uniformly in x. When f
admits such an expansion, its coeﬃcients a
k
are
uniquely determined. It is obvious that any dependent function which extends smoothly
to = 0 is represented by an asymptotic series, and a theorem of E.Borel (see [28, p.28])
tells us that, conversely, any asymptotic series can be “summed” to yield such a function.
The principal part of an asymptotic series
¸
∞
k=0
a
k
k
is deﬁned as its ﬁrst term which is
not identically zero as a function of x. The order of a is the index of its principal part.
If we consider as equivalent any two dependent functions whose diﬀerence is O(
∞
), i.e.
O(
k
) for all k, then each asymptotic series determines a unique equivalence class. A WKB
“solution” to the eigenfunction problem
ˆ
Hϕ = Eϕ is then an equivalence class of functions
of the form
ϕ = e
iS/
a,
where S is an admissible phase function and a is an dependent function represented by a
formal asymptotic series
a ∼
∞
¸
k=p
a
k
k
13
with the property that its principal part a
p
satisﬁes the homogeneous transport equation
a
p
∆S + 2
¸
j
∂a
p
∂x
j
∂S
∂x
j
= 0,
and for k > p, the a
k
satisfy the recursive transport equations:
a
k
∆S + 2
¸
j
∂a
k
∂x
j
∂S
∂x
j
= i∆a
k−1
.
This means that the dependent function ϕ (or any function equivalent to it) satisﬁes the
Schr¨odinger equation up to terms of order O(
∞
).
Geometry of the transport equation
In Section 2.1, we saw that a ﬁrstorder WKB approximate solution ϕ = e
iS/
to the time
independent Schr¨odinger equation depended on the choice of an admissible phase function,
i.e., a function S satisfying the HamiltonJacobi equation H(x,
∂S
∂x
) = E. The generalized or
geometric version of such a solution was a lagrangian submanifold of the level set H
−1
(E).
We now wish to interpret and generalize in a similar way the semiclassical approximation
with its amplitude included.
To begin, suppose that a is a function on R
n
which satisﬁes the homogeneous transport
equation:
a∆S + 2
¸
j
∂a
∂x
j
∂S
∂x
j
= 0.
After multiplying both sides of this equation by a, we can rewrite it as:
¸
j
∂
∂x
j
a
2
∂S
∂x
j
= 0,
which means that the divergence of the vector ﬁeld a
2
∇S is zero. Rather than considering
the transport equation as a condition on the vector ﬁeld a
2
∇S (on R
n
) per se, we can lift all
of this activity to the lagrangian submanifold L = im(dS). Notice ﬁrst that the restriction
to L of the hamiltonian vector ﬁeld associated to H(q, p) =
¸
p
2
i
/2 +V (q) is
X
H
[
L
=
¸
j
∂S
∂x
j
∂
∂q
j
−
∂V
∂q
j
∂
∂p
j
.
The projection X
(x)
H
of X
H
[
L
onto R
n
(the (x) reminds us of the coordinate x on R
n
) there
fore coincides with ∇S, and so the homogeneous transport equation says that a
2
X
(x)
H
is
divergencefree for the canonical density [dx[ = [dx
1
∧ ∧ dx
n
[ on R
n
. But it is better to
reformulate this condition as:
L
X
(x)
H
(a
2
[dx[) = 0;
14
that is, we transfer the factor of a
2
from the vector ﬁeld X
(x)
H
= ∇S to the density [dx[. Since
X
H
is tangent to L by the HamiltonJacobi theorem, and since the Lie derivative is invariant
under diﬀeomorphism, this equation is satisﬁed if and only if the pullback of a
2
[dx[ to L
via the projection π is invariant under the ﬂow of X
H
.
This observation, together with the fact that it is the square of a which appears in the
density π
∗
(a
2
[dx[), suggests that a solution of the homogeneous transport equation should
be represented geometrically by a halfdensity on L, invariant by X
H
. (See Appendix A for
a discussion of densities of fractional order.)
In other words, a (geometric) semiclassical state should be deﬁned as a lagrangian sub
manifold L of R
2n
equipped with a halfdensity a. Such a state is stationary when L lies in
a level set of the classical hamiltonian and a is invariant under its ﬂow.
Example 2.2 Recall that in the case of the 1dimensional harmonic oscillator, stationary
classical states are simply those lagrangian submanifolds of R
2
which coincide with the
regular level sets of the classical hamiltonian H(q, p) = (p
2
+ kq
2
)/2. There is a unique (up
to a constant) invariant volume element for the hamiltonian ﬂow of H on each level curve
H. Any such level curve L, together with a square root of this volume element, constitutes
a semiclassical stationary state for the harmonic oscillator.
´
Notice that while a solution to the homogeneous transport equation in the case of the 1
dimensional harmonic oscillator was necessarily singular (see Example 2.1), the semiclassical
state described in the preceding example is a perfectly smooth object everywhere on the
lagrangian submanifold L. The singularities arise only when we try to transfer the half
density from L down to conﬁguration space. Another substantial advantage of the geometric
interpretation of the semiclassical approximation is that the concept of an invariant half
density depends only on the hamiltonian vector ﬁeld X
H
and not on the function S, so it
makes sense on any lagrangian submanifold of R
2n
lying in a level set of H.
This discussion leads us to another change of viewpoint, namely that the quantum states
themselves should be represented, not by functions, but by halfdensities on conﬁguration
space R
n
, i.e. elements of the intrinsic Hilbert space H
R
n (see Appendix A). Stationary states
are then eigenvectors of the Schr¨odinger operator
ˆ
H, which is deﬁned on the space of smooth
halfdensities in terms of the old Schr¨odinger operator on functions, which we will denote
momentarily as
ˆ
H
fun
, by the equation
ˆ
H(a[dx[
1/2
) = (
ˆ
H
fun
a)[dx[
1/2
.
From this new point of view, we can express the result of our analysis as follows:
If S is an admissible phase function and a is a halfdensity on L = im(dS) which
is invariant under the ﬂow of the classical hamiltonian, then e
iS/
(dS)
∗
a is a
secondorder approximate solution to the timeindependent Schr¨odinger equation.
15
In summary, we have noted the following correspondences between classical and quantum
mechanics:
Object Classical version Quantum version
basic space R
2n
H
R
n
state lagrangian submanifold of R
2n
with halfdensity on R
n
halfdensity
timeevolution Hamilton’s equations Schr¨odinger equation
generator of evolution function H on R
2n
operator
ˆ
H on smooth
halfdensities
stationary state lagrangian submanifold in level set eigenvector of
ˆ
H
of H with invariant halfdensity
Proceeding further, we could attempt to interpret a solution of the recursive system of
inhomogeneous transport equations on R
n
as an asymptotic halfdensity on L in order to
arrive at a geometric picture of a complete WKB solution to the Schr¨odinger equation. This,
however, involves some additional diﬃculties, notably the lack of a geometric interpretation
of the inhomogeneous transport equations, which lie beyond the scope of these notes. Instead,
we will focus on two aspects of the semiclassical approximation. First, we will extend the
geometric picture presented above to systems with more general phase spaces. This will
require the concept of symplectic manifold, which is introduced in the following chapter.
Second, we will “quantize” semiclassical states in these symplectic manifolds. Speciﬁcally,
we will try to construct a space of quantum states corresponding to a general classical
phase space. Then we will try to construct asymptotic quantum states corresponding to
halfdensities on lagrangian submanifolds. In particular, we will start with an invariant
halfdensity on a (possibly nonprojectable) lagrangian submanifold of R
2n
and attempt to
use this data to construct an explicit semiclassical approximate solution to Schr¨odinger’s
equation on R
n
.
16
3 Symplectic Manifolds
In this chapter, we will introduce the notion of a symplectic structure on a manifold, moti
vated for the most part by the situation in R
2n
. While some discussion will be devoted to
certain general properties of symplectic manifolds, our main goal at this point is to develop
the tools needed to extend the hamiltonian viewpoint to phase spaces associated to general
ﬁnitedimensional conﬁguration spaces, i.e. to cotangent bundles. More general symplectic
manifolds will reappear as the focus of more sophisticated quantization programs in later
chapters. We refer to [6, 29, 63] for thorough discussions of the topics in this chapter.
3.1 Symplectic structures
In Section 2.1, a lagrangian submanifold of R
2n
was deﬁned as an ndimensional submanifold
L ⊂ R
2n
on which the exterior derivative of the form α
n
=
¸
p
i
dq
i
vanishes; to a function
H: R
2n
→R, we saw that Hamilton’s equations associate a vector ﬁeld X
H
on R
2n
satisfying
X
H
dα
n
= −dH.
Finally, our proof of the HamiltonJacobi theorem relied on the nondegeneracy of the 2
form dα
n
. These points already indicate the central role played by the form dα
n
in the
study of hamiltonian systems in R
2n
; the correct generalization of the hamiltonian picture to
arbitrary conﬁguration spaces relies similarly on the use of 2forms with certain additional
properties. In this section, we ﬁrst study such forms pointwise, collecting pertinent facts
about nondegenerate, skewsymmetric bilinear forms. We then turn to the deﬁnition of
symplectic manifolds.
Linear symplectic structures
Suppose that V is a real, mdimensional vector space. A bilinear form ω: V V →R gives
rise to a linear map
˜ ω: V →V
∗
deﬁned by contraction:
˜ ω(x)(y) = ω(x, y).
The ωorthogonal to a subspace W ⊂ V is deﬁned as
W
⊥
= ¦x ∈ V : W ⊂ ker ˜ ω(x)¦.
If ˜ ω is an isomorphism, or in other words if V
⊥
= ¦0¦, then the form ω is said to be nonde
generate; if in addition ω is skewsymmetric, then ω is called a linear symplectic structure on
V . A linear endomorphism of a symplectic vector space (V, ω) which preserves the form ω
is called a linear symplectic transformation, and the group of all such transformations
is denoted by Sp(V ).
17
Example 3.1 If E is any real ndimensional vector space with dual E
∗
, then a linear sym
plectic structure on V = E ⊕E
∗
is given by
ω((x, λ), (x
t
, λ
t
)) = λ
t
(x) −λ(x
t
).
With respect to a basis ¦x
i
¦ of E and a dual basis ¦λ
i
¦ of E
∗
, the form ω is represented by
the matrix
ω =
0 I
−I 0
.
It follows that if a linear operator on V is given by the real 2n 2n matrix
T =
A B
C D
,
then T is symplectic provided that A
t
C, B
t
D are symmetric, and A
t
D −C
t
B = I. Note in
particular that these conditions are satisﬁed if A ∈ GL(E), D = (A
t
)
−1
, and B = C = 0,
and so GL(E) is isomorphic to a subgroup Gl(E) of Sp(V ). More generally, if K : E → F
is an isomorphism, then the association
(x, λ) →(Kx, (K
−1
)
∗
λ)
deﬁnes a linear symplectomorphism between E ⊕E
∗
and F ⊕F
∗
equipped with these linear
symplectic structures.
´
Since the determinant of a skewsymmetric mm matrix is zero if m is odd, the existence
of a linear symplectic structure on a vector space V implies that V is necessarily even
dimensional and therefore admits a complex structure, i.e. a linear endomorphism J such
that J
2
= −I. A complex structure is said to be compatible with a symplectic structure on
V if
ω(Jx, Jy) = ω(x, y)
and
ω(x, Jx) > 0
for all x, y ∈ V . In other words, J is compatible with ω (we also call it ωcompatible)
if J : V → V is a linear symplectomorphism and g
J
(, ) = ω(, J) deﬁnes a symmetric,
positivedeﬁnite bilinear form on V .
Theorem 3.2 Every symplectic vector space (V, ω) admits a compatible complex structure.
Proof. Let ' , ` be a symmetric, positivedeﬁnite inner product on V , so that ω is represented
by an invertible skewadjoint operator K: V →V ; i.e.
ω(x, y) = 'Kx, y`.
18
The operator K admits a polar decomposition K = RJ, where R =
√
KK
t
is positive
deﬁnite symmetric, J = R
−1
K is orthogonal, and RJ = JR. From the skewsymmetry of K
it follows that J
t
= −J, and so J
2
= −JJ
t
= −id; i.e., J is a complex structure on V .
To see that J is ωcompatible, ﬁrst note that
ω(Jx, Jy) = 'KJx, Jy` = 'JKx, Jy` = 'Kx, y` = ω(x, y).
Also,
ω(x, Jx) = 'Kx, Jx` = 'JRx, Jx` = 'Rx, x` > 0,
since R and ' , ` are positivedeﬁnite.
2
Corollary 3.3 The collection . of ωcompatible complex structures on a symplectic vector
space (V, ω) is contractible.
Proof. The association J → g
J
described above deﬁnes a continuous map from . into the
space { of symmetric, positivedeﬁnite bilinear forms on V . By the uniqueness of the polar
decomposition, it follows that the map which assigns to a form ' , ` the complex structure J
constructed in the preceding proof is continuous, and the composition . →{ →. of these
maps equals the identity on .. Since { is contractible, this implies the corollary.
2
If J is ωcompatible, a hermitian structure on V is deﬁned by
', ` = g
J
(, ) + iω(, ).
As is easily checked, a linear transformation T ∈ GL(V ) which preserves any two of the
structures ω, g
J
, J on V preserves the third and therefore preserves the hermitian structure.
In terms of the automorphism groups Sp(V ), GL(V, J), O(V ), and U(V ) of ω, J, g, and ', `,
this means that the intersection of any two of Sp(V ), GL(V, J), O(V ) equals U(V ).
To determine the Lie algebra sp(V) of Sp(V ), consider a 1parameter family of maps e
tA
associated to some linear map A: V →V . For any v, w ∈ V , we have
d
dt
t=0
ω(e
tA
v, e
tA
w) = ω(Av, w) + ω(v, Aw),
and so A ∈ sp(V) if and only if the linear map ˜ ω ◦ A: V →V
∗
is selfadjoint. Consequently,
dim(V ) = 2k implies dim(sp(V)) = dim(Sp(V)) = k(2k + 1).
Distinguished subspaces
The ωorthogonal to a subspace W of a symplectic vector space (V, ω) is called the symplectic
orthogonal to W. From the nondegeneracy of the symplectic form, it follows that
W
⊥⊥
= W and dimW
⊥
= dimV −dimW
19
for any subspace W ⊂ V . Also,
(A + B)
⊥
= A
⊥
∩ B
⊥
and (A ∩ B)
⊥
= A
⊥
+ B
⊥
for any pair of subspaces A, B of V . In particular, B
⊥
⊂ A
⊥
whenever A ⊂ B.
Note that the symplectic orthogonal W
⊥
might not be an algebraic complement to W.
For instance, if dimW = 1, the skewsymmetry of ω implies that W ⊂ W
⊥
. More generally,
any subspace contained in its orthogonal will be called isotropic. Dually, we note that if
codimW = 1, then W
⊥
is 1dimensional, hence isotropic, and W
⊥
⊂ W
⊥⊥
= W. In
general, spaces W satisfying the condition W
⊥
⊂ W are called coisotropic or involutive.
Finally, if W is selforthogonal, i.e. W
⊥
= W, then the dimension relation above implies that
dimW =
1
2
dimV . Any selforthogonal subspace is simultaneously isotropic and coisotropic,
and is called lagrangian.
According to these deﬁnitions, a subspace W ⊂ V is isotropic if the restriction of the
symplectic form to W is identically zero. At the other extreme, the restriction of ω to certain
subspaces Z ⊂ V may again be nondegenerate; this is equivalent to saying that Z∩Z
⊥
= ¦0¦
or Z + Z
⊥
= V . Such subspaces are called symplectic.
Example 3.4 In E⊕E
∗
with its usual symplectic structure, both E and E
∗
are lagrangian
subspaces. It also follows from the deﬁnition of this structure that the graph of a linear map
B: E →E
∗
is a lagrangian subspace of E ⊕E
∗
if and only if B is selfadjoint.
If (V, ω) is a symplectic vector space, we denote by V ⊕V the vector space V ⊕V equipped
with the symplectic structure ω ⊕ −ω. If T : V → V is a linear symplectic map, then the
graph of T is a lagrangian subspace of V ⊕V .
The kernel of a nonzero covector α ∈ V
∗
is a codimension1 coisotropic subspace ker α of
V whose symplectic orthogonal (ker α)
⊥
is the distinguished 1dimensional subspace of ker α
spanned by ˜ ω
−1
(α).
´
Example 3.5 Suppose that (V, ω) is a 2ndimensional symplectic vector space and W ⊂ V
is any isotropic subspace with dim(W) < n. Since 2n = dim(W) + dim(W
⊥
), there exists a
nonzero vector w ∈ W
⊥
`W. The subspace W
t
of V spanned by W∪¦w¦ is then isotropic and
dim(W
t
) = dim(W)+1. From this observation it follows that for every isotropic subspace W
of a (ﬁnitedimensional) symplectic vector space V which is not lagrangian, there exists an
isotropic subspace W
t
of V which properly contains W. Beginning with any 1dimensional
subspace of V , we can apply this remark inductively to conclude that every ﬁnitedimensional
symplectic vector space contains a lagrangian subspace.
´
Various subspaces of a symplectic vector space are related as follows.
Lemma 3.6 If L is a lagrangian subspace of a symplectic vector space V , and A ⊂ V is an
arbitrary subspace, then:
20
1. L ⊂ A if and only if A
⊥
⊂ L.
2. L is transverse to A if and only if L ∩ A
⊥
= ¦0¦.
Proof. Statement (1) follows from the properties of the operation
⊥
and the equation L = L
⊥
.
Similarly, L + A = V if and only if (L + A)
⊥
= L ∩ A
⊥
= ¦0¦, proving statement (2).
2
Example 3.7 Note that statement (1) of Lemma 3.6 implies that if L ⊂ A, then A
⊥
⊂ A,
and so A is a coisotropic subspace. Conversely, if A is coisotropic, then A
⊥
is isotropic,
and Example 3.5 implies that there is a lagrangian subspace L with A
⊥
⊂ L. Passing to
orthogonals, we have L ⊂ A. Thus, a subspace C ⊂ V is coisotropic if and only if it contains
a lagrangian subspace.
Suppose that V is a symplectic vector space with an isotropic subspace I and a lagrangian
subspace L such that I ∩ L = 0. If W ⊂ L is any complementary subspace to I
⊥
∩ L, then
I + L ⊂ W + I
⊥
, and so W
⊥
∩ I ⊂ I
⊥
∩ L. Thus, W
⊥
∩ I ⊂ I ∩ L = 0. Since I
⊥
∩ W = 0
by our choice of W, it follows that I + W is a symplectic subspace of V .
´
A pair L, L
t
of transverse lagrangian subspaces of V is said to deﬁne a lagrangian
splitting of V . In this case, the map ˜ ω deﬁnes an isomorphism L
t
· L
∗
, which in turn
gives rise to a linear symplectomorphism between V and L⊕L
∗
equipped with its canonical
symplectic structure (see Example 3.1). If J is a ωcompatible complex structure on V and
L ⊂ V a lagrangian subspace, then L, JL is a lagrangian splitting. By Example 3.5, every
symplectic vector space contains a lagrangian subspace, and since every ndimensional vector
space is isomorphic to R
n
, the preceding remarks prove the following linear “normal form”
result:
Theorem 3.8 Every 2ndimensional symplectic vector space is linearly symplectomorphic
to (R
2n
, ω
n
).
Theorem 3.2 also implies the following useful result.
Lemma 3.9 Suppose that V is a symplectic vector space with a ωcompatible complex struc
ture J and let T
ε
: V →V be given by T
ε
(x) = x + εJx.
1. If L, L
t
are any lagrangian subspaces of V , then L
ε
= T
ε
(L) is a lagrangian subspace
transverse to L
t
for small ε > 0.
2. For any two lagrangian subspaces L, L
t
of V , there is a lagrangian subspace L
tt
trans
verse to both L
t
and L.
Proof. It is easy to check that T
ε
is a conformal linear symplectic map, i.e. an isomorphism
of V satisfying ω(T
ε
x, T
ε
y) = (1 + ε
2
) ω(x, y). Thus, L
ε
is a lagrangian subspace for all
ε > 0. Using the innerproduct g
J
on V induced by J, we can choose orthonormal bases
21
¦v
i
¦, ¦w
i
¦ of L
t
and L, respectively, so that for i = 1, , k, the vectors v
i
= w
i
span L∩L
t
.
Then ¦w
i
+ εJw
i
¦ form a basis of L
ε
, and L
t
, L
ε
are transverse precisely when the matrix
M = ¦ω(v
i
, w
j
+ εJw
j
)¦ = ¦ω(v
i
, w
j
) + ε g
J
(v
i
, w
j
)¦ is nonsingular. Our choice of bases
implies that
M =
ε id 0
0 A + ε B
where A = ¦ω(v
i
, w
j
)¦
n
i,j=k+1
and B is some (n−k)(n−k) matrix. Setting I = span¦v
i
¦
n
i=k+1
and W = span¦w
i
¦
n
i=k+1
, we can apply Example 3.7 to conclude that A is nonsingular, and
assertion (1) follows.
To prove (2), observe that for small ε > 0, the lagrangian subspace L
ε
is transverse to
L, L
t
by (1).
2
In fact, the statement of preceding lemma can be improved as follows. Let ¦L
i
¦ be a count
able family of lagrangian subspaces, and let A
i
, B
i
be the matrices obtained with respect to L
as in the proof above. For each i, the function t →det(A
i
+tB
i
) is a nonzero polynomial and
therefore has ﬁnitely many zeros. Consequently, the lagrangian subspace T
t
(L) is transverse
to all L
i
for almost every t ∈ R.
The lagrangian grassmannian
The collection of all unoriented lagrangian subspaces of a 2ndimensional symplectic vector
space V is called the lagrangian grassmannian L(V ) of V . A natural action of the group
Sp(V ) on L(V ), denoted : Sp(V ) L(V ) →L(V ) is deﬁned by (T, L) =
L
(T) = T(L).
Lemma 3.10 The unitary group associated to an ωcompatible complex structure J on V
acts transitively on L(V ).
Proof. For arbitrary L
1
, L
2
∈ L(V ), an orthogonal transformation L
1
→ L
2
induces a
symplectic transformation L
1
⊕L
∗
1
→L
2
⊕L
∗
2
in the manner of Example 3.1, which in turn
gives rise to a unitary transformation L
1
⊕JL
1
→L
2
⊕JL
2
mapping L
1
onto L
2
.
2
The stabilizer of L ∈ L(V ) under the U(V )action is evidently the orthogonal subgroup of
Gl(L) deﬁned with respect to the innerproduct and splitting L⊕JL of V induced by J (see
Example 3.1). Thus, a (noncanonical) identiﬁcation of the lagrangian grassmannian with
the homogeneous space U(n)/O(n) is obtained from the map
U(V )
L
→L(V ).
The choice of J also deﬁnes a complex determinant U(V )
det
2
J
→ S
1
, which induces a ﬁbration
L(V ) → S
1
with 1connected ﬁber SU(n)/SO(n), giving an isomorphism of fundamental
groups
π
1
(L(V )) · π
1
(S
1
) · Z.
22
This isomorphism does not depend on the choices of J and L made above. Independence of J
follows from the fact that . is connected (Corollary 3.3). On the other hand, connectedness
of the unitary group together with Lemma 3.10 gives independence of L.
Passing to homology and dualizing, we obtain a natural homomorphism
H
1
(S
1
; Z) →H
1
(L(V ); Z).
The image of the canonical generator of H
1
(S
1
; Z) under this map is called the universal
Maslov class, µ
V
. The result of the following example will be useful when we extend our
discussion of the Maslov class from vector spaces to vector bundles.
Example 3.11 If (V, ω) is any symplectic vector space with ωcompatible complex structure
J and lagrangian subspace L, then a check of the preceding deﬁnitions shows that
m
L
((T, L
t
)) = m
L
(L
t
) det
2
J
(T)
for any T ∈ U(V ) and L
t
∈ L(V ). (Recall that : Sp(V )L(V ) →L(V ) denotes the natural
action of Sp(V ) on L(V )).
Now consider any topological space M. If f
1
, f
2
: M →L(V ) are continuous maps, then
the deﬁnition of the universal Maslov class shows that (f
∗
1
− f
∗
2
)µ
V
equals the pullback of
the canonical generator of H
1
(S
1
; R) by the map
(m
L
◦ f
1
)(m
L
◦ f
2
)
−1
.
(Here we use the fact that when S
1
is identiﬁed with the unit complex numbers, the mul
tiplication map S
1
S
1
→ S
1
induces the diagonal map H
1
(S
1
) → H
1
(S
1
) ⊕ H
1
(S
1
) ·
H
1
(S
1
S
1
) on cohomology). If T : M → Sp(V ) is any map, we set (T f
i
) = (T, f
i
).
Since Sp(V ) deformation retracts onto U(V ), it follows that T is homotopic to a map
T
t
: M →U(V ), and so ((T f
1
)
∗
−(T f
2
)
∗
)µ
V
is obtained via pullback by
(m
L
◦ (T
t
f
1
))(m
L
◦ (T
t
f
2
))
−1
.
From the ﬁrst paragraph, it follows that this product equals (m
L
◦f
1
)(m
L
◦f
2
)
−1
, from which
we conclude that
(f
∗
1
−f
∗
2
)µ
V
= ((T f
1
)
∗
−(T f
2
)
∗
)µ
V
.
´
Symplectic manifolds
To motivate the deﬁnition of a symplectic manifold, we ﬁrst recall some features of the
diﬀerential form −dα
n
= ω
n
=
¸
n
j=1
dq
j
∧ dp
j
which appeared in our earlier discussion.
First, we note that
¸
n
j=1
dq
j
∧dp
j
deﬁnes a linear symplectic structure on the tangent space
of R
2n
at each point. In fact:
ω
n
∂
∂q
j
,
∂
∂p
k
= δ
jk
ω
n
∂
∂q
j
,
∂
∂q
k
= 0 ω
n
∂
∂p
j
,
∂
∂p
k
= 0
23
and so
˜ ω
n
∂
∂q
j
= dp
j
˜ ω
n
∂
∂p
j
= −dq
j
,
from which it is clear that ˜ ω
n
is invertible.
Next, we recall that the hamiltonian vector ﬁeld associated via Hamilton’s equations to
H : R
2n
→R satisﬁes
X
H
ω
n
= dH,
or in other words,
X
H
= ˜ ω
−1
n
(dH),
so we see that the symplectic form ω
n
is all that we need to obtain X
H
from H. This
description of the hamiltonian vector ﬁeld leads immediately to the following two invariance
results. First note that by the skewsymmetry of ω
n
,
L
X
H
H = X
H
H = ω
n
(X
H
, X
H
) = 0,
implying that X
H
is tangent to the level sets of H. This again reﬂects the fact that the ﬂow
of X
H
preserves energy. Since ω
n
is closed, we also have by Cartan’s formula (see [1])
L
X
H
ω
n
= d(X
H
ω
n
) + X
H
dω
n
= d
2
H = 0.
This equation implies that the ﬂow of X
H
preserves the form ω
n
and therefore generalizes our
earlier remark that the hamiltonian vector ﬁeld associated to the 1dimensional harmonic
oscillator is divergencefree.
We now see what is needed to do hamiltonian mechanics on manifolds. A 2form ω on a
manifold P is a smooth family of bilinear forms on the tangent spaces of P. By assuming that
each of these bilinear forms is nondegenerate, we guarantee that the equation X
H
= ˜ ω
−1
(dH)
deﬁnes a hamiltonian vector ﬁeld uniquely for any H. Computing the Lie derivative of H
with respect to X
H
L
X
H
H = X
H
H = ˜ ω(X
H
)(X
H
) = ω(X
H
, X
H
) = 0,
we see that the conservation of energy follows from the skewsymmetry of the form ω.
Finally, invariance of ω under the hamiltonian ﬂow is satisﬁed if
L
X
H
ω = d(X
H
ω) + X
H
dω = 0.
Here, the term d(X
H
ω) = d
2
H is automatically zero; to guarantee the vanishing of the
second term, we impose the condition that ω be closed.
Thus we make the following deﬁnition:
Deﬁnition 3.12 A symplectic structure on a manifold P is a closed, nondegenerate
2form ω on P.
24
The condition that ω be nondegenerate means that ˜ ω deﬁnes an isomorphism of vector
bundles TP → T
∗
P, or equivalently, that the top exterior power of ω is a volume form on
P, or ﬁnally, that ω deﬁnes a linear symplectic structure on each tangent space of P.
An immediate example of a symplectic manifold is furnished by R
2n
with its standard
structure ω
n
=
¸
n
j=1
dq
j
∧ dp
j
(a diﬀerential form with constant coeﬃcients and not just
a single bilinear form). Darboux’s theorem (Section 4.3) will tell us that this is the local
model for the general case. In the next section, we will see that the cotangent bundle of any
smooth manifold carries a natural symplectic structure.
Generalizing our earlier discussion of distinguished subspaces of a symplectic vector space,
we call a submanifold C ⊂ P (co)isotropic provided that each tangent space T
p
C of C is a
(co)isotropic subspace of T
p
P. When C is coisotropic, the subspaces (
p
= (T
p
C)
⊥
comprise
a subbundle (TC)
⊥
of TC known as the characteristic distribution of C. It is integrable
because ω is closed. Of particular interest in our discussion will be lagrangian submanifolds
of P, which are (co)isotropic submanifolds of dimension
1
2
dim(P). More generally, if L is a
smooth manifold of dimension
1
2
dim(P) and ι : L → P is an immersion such that ι
∗
ω = 0,
we will call the pair (L, ι) a lagrangian immersion.
Example 3.13 If C ⊂ P is a hypersurface, then C is a coisotropic submanifold. A simple
check of deﬁnitions shows that if H: P →R is a smooth function having C as a regular level
set, then the hamiltonian vector ﬁeld X
H
is tangent to the characteristic foliation of C.
If (L, ι) is a lagrangian immersion whose image is contained in C, then Lemma 3.6 implies
that for each p ∈ L, the characteristic subspace (
ι(p)
⊂ T
ι(p)
C is contained in ι
∗
T
p
L, and thus
X
H
induces a smooth, nonsingular vector ﬁeld X
H,ι
on L. In view of the remarks above, this
assertion generalizes the HamiltonJacobi theorem (see the end of Section 2.1) to arbitrary
symplectic manifolds and lagrangian immersions.
´
New symplectic manifolds can be manufactured from known examples by dualizing and by
taking products. The symplectic dual of a manifold (P, ω) consists of the same underlying
manifold endowed with the symplectic structure −ω. Evidently P and its dual P share the
same (co)isotropic submanifolds. Given two symplectic manifolds (P
1
, ω
1
) and (P
2
, ω
2
), their
product P
1
P
2
admits a symplectic structure given by the sum ω
1
⊕ω
2
. More explicitly, this
form is the sum of the pullbacks of ω
1
and ω
2
by the projections of P
1
P
2
to P
1
and P
2
. As
is easily veriﬁed, the product of (co)isotropic submanifolds of P
1
and P
2
is a (co)isotropic
submanifold of P
1
P
2
.
A symplectomorphism from (P
1
, ω
1
) to (P
2
, ω
2
) is a smooth diﬀeomorphism f : P
1
→
P
2
compatible with the symplectic structures: f
∗
ω
2
= ω
1
. A useful connection among duals,
products, and symplectomorphisms is provided by the following lemma.
Lemma 3.14 A diﬀeomorphism f : P
1
→ P
2
between symplectic manifolds is a symplecto
morphism if and only if its graph is a lagrangian submanifold of the product P
2
P
1
.
The collection Aut(P, ω) of symplectomorphisms of P becomes an inﬁnitedimensional Lie
group when endowed with the C
∞
topology (see [49]). In this case, the corresponding Lie
25
algebra is the space χ(P, ω) of smooth vector ﬁelds X on P satisfying
L
X
ω = 0.
Since L
X
ω = d(X ω), the association X →X ω deﬁnes an isomorphism between χ(P, ω)
and the space of closed 1forms on P; those X which map to exact 1forms are simply the
hamiltonian vector ﬁelds on P. The elements of χ(P, ω) are called locally hamiltonian
vector ﬁelds or symplectic vector ﬁelds.
Example 3.15 A linear symplectic form ω on a vector space V induces a symplectic struc
ture (also denoted ω) on V via the canonical identiﬁcation of TV with V V . The symplectic
group Sp(V ) then embeds naturally in Aut(V, ω), and the Lie algebra sp(V) identiﬁes with
the subalgebra of χ(V, ω) consisting of vector ﬁelds of the form
X(v) = Av
for some A ∈ sp(V). Note that these are precisely the hamiltonian vector ﬁelds of the
homogeneous quadratic polynomials on V , i.e. functions satisfying Q(tv) = t
2
Q(v) for all
real t. Consequently, sp(V) is canonically identiﬁed with the space of such functions via the
correspondence
A ↔Q
A
(v) =
1
2
ω(Av, v).
´
Symplectic vector bundles
Since a symplectic form on a 2nmanifold P deﬁnes a smooth family of linear symplectic
forms on the ﬁbers of TP, the frame bundle of P can be reduced to a principal Sp(n) bundle
over P. More generally, any vector bundle E →B with this structure is called a symplectic
vector bundle. Two symplectic vector bundles E, F are said to be symplectomorphic if
there exists a vector bundle isomorphism E →F which preserves their symplectic structures.
Example 3.16 If F → B is any vector bundle, then the sum F ⊕ F
∗
carries a natural
symplectic vector bundle structure, deﬁned in analogy with Example 3.1.
´
With the aid of an arbitrary riemannian metric, the proof of Theorem 3.2 can be gener
alized by a ﬁberwise construction as follows.
Theorem 3.17 Every symplectic vector bundle admits a compatible complex vector bundle
structure.
Example 3.18 Despite Theorem 3.17, there exist examples of symplectic manifolds which
are not complex (the almost complex structure coming from the theorem cannot be made
integrable), and of complex manifolds which are not symplectic. (See [27] and the numerous
earlier references cited therein.) Note, however, that the K¨ahler form of any K¨ahler manifold
is a symplectic form.
26
´
A lagrangian subbundle of a symplectic vector bundle E is a subbundle L ⊂ E such
that L
x
is a lagrangian subspace of E
x
for all x ∈ B. If E admits a lagrangian subbundle,
then E is symplectomorphic to L⊕L
∗
, and the frame bundle of E admits a further reduction
to a principal GL(n) bundle over B (compare Example 3.1).
Example 3.19 If L is a lagrangian submanifold of a symplectic manifold P, then the re
stricted tangent bundle T
L
P is a symplectic vector bundle over L, and TL ⊂ T
L
P is a
lagrangian subbundle. Also note that if C ⊂ P is any submanifold such that TC contains a
lagrangian subbundle of T
C
P, then C is coisotropic (see Lemma 3.6).
´
In general, the automorphism group of a symplectic vector bundle E does not act tran
sitively on the lagrangian subbundles of E. Nevertheless, a pair of transverse lagrangian
subbundles can be related as follows.
Theorem 3.20 Let E → B be a symplectic vector bundle and suppose that L, L
t
are la
grangian subbundles such that L
x
is transverse to L
t
x
for each x ∈ B. Then there exists a
compatible complex structure J on E satisfying JL = L
t
.
Proof. Let J
0
be any compatible complex structure on E. Since L
t
and J
0
L are both
transverse to L, we can ﬁnd a symplectomorphism T : L ⊕ L
t
→ L ⊕ J
0
L which preserves
the subbundle L and maps L
t
to J
0
L. A simple check of the deﬁnition then shows that
J = T
−1
J
0
T is a compatible complex structure on E which satisﬁes JL = L
t
.
2
Example 3.21 If E is a symplectic vector bundle over M, then any pair L, L
t
of lagrangian
subbundles of E deﬁne a cohomology class µ(L, L
t
) ∈ H
1
(M; Z) as follows.
Assuming ﬁrst that E admits a symplectic trivialization f : E → M V for some
symplectic vector space V , we denote by f
L
, f
L
: M → L(V ) the maps induced by the
lagrangian subbundles f(L), f(L
t
) of M V . Then
µ(L, L
t
) = (f
∗
L
−f
∗
L
) µ
V
,
where µ
V
∈ H
1
(L(V ); Z) is the universal Maslov class. From Example 3.11 it follows that
this class is independent of the choice of trivialization f.
For nontrivial E, we note that since the symplectic group Sp(V ) is connected, it follows
that for any loop γ : S
1
→ M, the pullback bundle γ
∗
E is trivial. Thus µ(L, L
t
) is well
deﬁned by the requirement that for every smooth loop γ in M,
γ
∗
µ(L, L
t
) = µ(γ
∗
L, γ
∗
L
t
).
´
27
Example 3.22 As a particular case of Example 3.21, we consider the symplectic manifold
R
2n
· T
∗
R
n
with its standard symplectic structure. Then the tangent bundle T(R
2n
) is
a symplectic vector bundle over R
2n
with a natural “vertical” lagrangian subbundle V R
n
deﬁned as the kernel of π
∗
, where π: T
∗
R
n
→R
n
is the natural projection.
If ι : L →R
2n
is a lagrangian immersion, then the symplectic vector bundle ι
∗
T(T
∗
(R
n
))
has two lagrangian subbundles, the image L
1
of ι
∗
: TL → ι
∗
T(T
∗
(R
n
)) and L
2
= ι
∗
V R
n
.
The class µ
L,ι
= µ(L
1
, L
2
) ∈ H
1
(L; Z) is called the Maslov class of (L, ι).
A check of these deﬁnitions shows that the Maslov class of (L, ι) equals the pullback of
the universal Maslov class µ
n
∈ H
1
(L(R
2n
); Z) by the Gauss map G: L → L(R
2n
) deﬁned
by G(p) = ι
∗
T
p
L ⊂ R
2n
. (See [3] for an interpretation of the Maslov class of a loop γ in
L as an intersection index of the loop G ◦ γ with a singular subvariety in the lagrangian
grassmannian).
´
3.2 Cotangent bundles
The cotangent bundle T
∗
M of any smooth manifold M is equipped with a natural 1form,
known as the Liouville form, deﬁned by the formula
α
M
((x, b))(v) = b(π
∗
v),
where π: T
∗
M →M is the canonical projection. In local coordinates (x
1
, , x
n
) on M and
corresponding coordinates (q
1
, , q
n
, p
1
, , p
n
) on T
∗
M, the equations
q
j
(x, b) = x
j
(x) p
j
(x, b) = b
∂
∂x
j
imply that
α
M
=
n
¸
j=1
p
j
dq
j
.
Thus, −dα
M
=
¸
n
j=1
dq
j
∧ dp
j
in these coordinates, from which it follows that the form
ω
M
= −dα
M
is a symplectic structure on T
∗
M. Note that if M = R
n
, then ω
M
is just the
symplectic structure ω
n
on T
∗
R
n
· R
2n
discussed previously, and α
M
= α
n
.
Lagrangian immersions and the Liouville class
Given a lagrangian immersion ι : L → T
∗
M, we set π
L
= π ◦ ι, where π : T
∗
M → M
is the natural projection. Critical points and critical values of π
L
are called respectively
singular points and caustic points of L. Finally, (L, ι) is said to be projectable if
π
L
is a diﬀeomorphism. A nice property of the Liouville 1form is that it can be used to
parametrize the set of projectable lagrangian submanifolds. To do this, we use the notation
ι
ϕ
to denote a 1form ϕ on M when we want to think of it as a map from M to T
∗
M.
28
Lemma 3.23 Let ϕ ∈ Ω
1
(M). Then
ι
∗
ϕ
α
M
= ϕ.
Proof. Because ι
ϕ
is a section of T
∗
M, it satisﬁes π ◦ ι
ϕ
= id
M
. By the deﬁnition of α
M
, it
follows that for each v ∈ T
p
M,
ι
∗
ϕ
α
M
(p)(v) = α
M
(ι
ϕ
(p))(ι
ϕ∗
v) = 'ι
ϕ
(p), π
∗
(ι
ϕ∗
v)` = 'ι
ϕ
(p), v`.
2
For this reason, α
M
is often described as the “tautological” 1form on T
∗
M. Taking exterior
derivatives on both sides of the equation in Lemma 3.23, we get
dϕ = dι
∗
ϕ
α
M
= ι
∗
ϕ
dα
M
= −ι
∗
ϕ
ω
M
.
From this equation we see that the image of ϕ is a lagrangian submanifold of T
∗
M precisely
when the form ϕ is closed. This proves
Proposition 3.24 The relation ϕ ↔ (M, ι
ϕ
) deﬁnes a natural bijective correspondence be
tween the the vector space of closed 1forms on M and the set of projectable lagrangian
submanifolds of T
∗
M.
Generalizing our WKB terminology, we will call S : M → R a phase function for a
projectable lagrangian embedding (L, ι) ⊂ T
∗
M provided that ι(L) = dS(M). The preceding
remarks imply a simple link between phase functions and the Liouville form:
Lemma 3.25 If (L, ι) ⊂ T
∗
M is a projectable lagrangian embedding, then S : M →R is a
phase function for L if and only if d(S ◦ π
L
◦ ι) = ι
∗
α
M
.
Thus, L is the image of an exact 1form on M if and only if the restriction of the Liouville
form to L is itself exact. This motivates the following deﬁnition.
Deﬁnition 3.26 If L, M are nmanifolds and ι : L →T
∗
M is an immersion such that ι
∗
α
M
is exact, then ι is called an exact lagrangian immersion.
If ι : L → T
∗
M is an exact lagrangian immersion, then Lemma 3.25 suggests that the
primitive of ι
∗
α
M
is a sort of generalized phase function for (L, ι) which lives on the manifold
L itself. We will return to this important viewpoint in the next chapter.
Example 3.27 A simple application of Stokes’ theorem shows that an embedded circle in
the phase plane cannot be exact, although it is the image of an exact lagrangian immersion
of R.
A general class of exact lagrangian submanifolds can be identiﬁed as follows. Associated
to a smooth submanifold N ⊂ M is the submanifold
N
⊥
= ¦(x, p) ∈ T
∗
M : x ∈ N, T
x
N ⊂ ker(p)¦,
known as the conormal bundle to N. From this deﬁnition it follows easily that dimT
∗
M =
2 dimN
⊥
, while the Liouville form of T
∗
M vanishes on N
⊥
for any N.
If T is a smooth foliation of M, then the union of the conormal bundles to the leaves of T
is a smooth submanifold of T
∗
M foliated by lagrangian submanifolds and is thus coisotropic
(see Example 3.19).
29
´
Although many lagrangian immersions ι : L → T
∗
M are not exact, the form ι
∗
α
M
is
always closed, since dι
∗
α
M
= ι
∗
ω
M
= 0. The deRham cohomology class λ
L,ι
∈ H
1
(L; R)
induced by this form will play an important role in the quantization procedures of the next
chapter and is known as the Liouville class of (L, ι).
Example 3.28 To generalize the picture described in Example 3.27, we consider a smooth
manifold M, together with a submanifold N ⊂ M and a closed 1form β on N. Then
N
⊥
β
= ¦(x, p) ∈ T
∗
M : x ∈ N p[
TxN
= β(x)¦
is a lagrangian submanifold of T
∗
M whose Liouville class equals [π
∗
N
β] ∈ H
1
(N
⊥
β
; R), where
π
N
: N
⊥
β
→N is here the restriction of the natural projection π: T
∗
M →M.
´
Fiberpreserving symplectomorphisms
On each ﬁber of the projection π: T
∗
M →M, the pullback of α
M
vanishes, so the ﬁbers are
lagrangian submanifolds. Thus, the vertical bundle V M = ker π
∗
is a lagrangian subbundle
of T(T
∗
M). Since α
M
vanishes on the zero section Z
M
⊂ T
∗
M, it follows that Z
M
is
lagrangian as well, and the subbundles TZ
M
and V M deﬁne a canonical lagrangian splitting
of T(T
∗
M) over Z
M
.
A 1form β on M deﬁnes a diﬀeomorphism f
β
of T
∗
M by ﬁberwise aﬃne translation
f
β
(x, p) = (x, p + β(x)).
It is easy to see that this map satisﬁes
f
∗
β
α
M
= α
M
+ π
∗
β,
so f
β
is a symplectomorphism of T
∗
M if and only if β is closed.
Theorem 3.29 If a symplectomorphism f : T
∗
M → T
∗
M preserves each ﬁber of the pro
jection π: T
∗
M →M, then f = f
β
for a closed 1form β on M.
Proof. Fix a point (x
0
, p
0
) ∈ T
∗
M and let ψ be a closed 1form on M such that ψ(x
0
) =
(x
0
, p
0
). Since f is symplectic, the form µ = f ◦ ψ is also closed, and thus the map
h = f
−1
µ
◦ f ◦ f
ψ
is a symplectomorphism of T
∗
M which preserves ﬁbers and ﬁxes the zero section Z
M
⊂ T
∗
M.
Moreover, since the derivative Dh preserves the lagrangian splitting of T(T
∗
M) along Z
M
and equals the identity on TZ
M
, we can conclude from Example 3.1 that Dh is the identity
at all points of Z
M
. Consequently, the ﬁberderivative of f at the arbitrary point (x
0
, p
0
)
equals the identity, so f is a translation on each ﬁber. Deﬁning β(x) = f(x, 0), we have
f = f
β
.
30
2
If β is a closed 1form on M, then the ﬂow f
t
of the vector ﬁeld X
β
= −˜ ω
−1
M
(π
∗
β) is
symplectic; since V M ⊂ ker π
∗
β, the HamiltonJacobi theorem implies furthermore that the
ﬂow f
t
satisﬁes the hypotheses of Theorem 3.29.
Corollary 3.30 For any closed 1form β on M, the time1 map f = f
1
of the ﬂow of X
β
equals f
β
.
Proof. By Theorem 3.29, the assertion will follow provided that we can show that f
∗
α
M
=
α
M
+ π
∗
β. To this end, note that the deﬁnition of the Lie derivative shows that f satisﬁes
f
∗
α
M
= α
M
+
1
0
d
dt
(f
∗
t
α
M
) dt = α
M
+
1
0
f
∗
t
(L
X
β
α
M
) dt.
By Cartan’s formula for the Lie derivative, we have
L
X
β
α
M
= d(X
β
α
M
) −X
β
ω
M
= π
∗
β,
the latter equality following from the fact that X
β
⊂ V M ⊂ ker α
M
and dα
M
= −ω
M
.
Another application of Cartan’s formula, combined with the assumption that β is closed
shows that
L
X
β
π
∗
β = 0,
and so f
∗
t
π
∗
β = π
∗
β for all t. Inserting these computations into the expression for f
∗
α
M
above, we obtain
f
∗
α
M
= α
M
+ π
∗
β.
2
Using Theorem 3.29, we can furthermore classify all ﬁberpreserving symplectomorphisms
from T
∗
M to T
∗
N.
Corollary 3.31 Any ﬁberpreserving symplectomorphism F : T
∗
M → T
∗
N can be realized
as the composition of a ﬁbertranslation in T
∗
M with the cotangent lift of a diﬀeomorphism
N →M.
Proof. By composing F with a ﬁbertranslation in T
∗
M we may assume that F maps the
zero section of T
∗
M to that of T
∗
N. The restriction of F
−1
to the zero sections then induces
a diﬀeomorphism f : N → M such that the composition F ◦ (f
−1
)
∗
is a ﬁberpreserving
symplectomorphism of T
∗
N which ﬁxes the zero section. From the preceding theorem, we
conclude that F = f
∗
.
2
31
The Schwartz transform
If M, N are smooth manifolds, then the map S
M,N
: T
∗
M T
∗
N → T
∗
(M N) deﬁned in
local coordinates by
((x, ξ), (y, η)) →(x, y, −ξ, η)
is a symplectomorphism which we will call the Schwartz transform.
5
An elementary, but
fundamental property of this mapping can be described as follows.
Proposition 3.32 If M, N are smooth manifolds, then the Schwartz transform S
M,N
satis
ﬁes
(S
M,N
)
∗
α
MN
= α
M
⊕−α
N
.
In particular, S
M,N
induces a diﬀeomorphism of zero sections
Z
M
Z
N
· Z
MN
and an isomorphism of vertical bundles
V M ⊕V N · V (M N).
Using the Schwartz transform, we associate to any symplectomorphism F : T
∗
M →T
∗
N
the lagrangian embedding ι
F
: T
∗
M →T
∗
(M N) deﬁned as the composition of S
M,N
with
the graph Γ
F
: T
∗
M →T
∗
M T
∗
N.
Example 3.33 By Corollary 3.31, a ﬁberpreserving symplectomorphism F : T
∗
M →T
∗
N
equals the composition of ﬁberwise translation by a closed 1form β on M with the cotangent
lift of a diﬀeomorphism g : N →M. A computation shows that if Γ ⊂ M N is the graph
of g and p : Γ → M is the natural projection, then the image of the composition of the
lagrangian embedding (T
∗
M, ι
F
) with the Schwartz transform S
M,N
equals the submanifold
Γ
⊥
p
∗
β
⊂ T
∗
(M N) deﬁned in Example 3.28. In particular, if F is the cotangent lift of g,
then the image of (T
∗
M, ι
F
) equals the conormal bundle of Γ.
´
Finally, we note that multiplying the cotangent vectors in T
∗
M by −1 deﬁnes a sym
plectomorphism T
∗
M → T
∗
M which can be combined with the Schwartz transform S
M,N
to arrive at the usual symplectomorphism T
∗
M T
∗
N · T
∗
(M N). Thus in the special
case of cotangent bundles, dualizing and taking products leads to nothing new.
3.3 Mechanics on manifolds
With the techniques of symplectic geometry at our disposal, we are ready to extend our
discussion of mechanics to more general conﬁguration spaces. Our description begins with a
comparison of the classical and quantum viewpoints, largely parallelling the earlier material
on the 1dimensional harmonic oscillator given in the introduction. We then turn to the
semiclassical approximation and its geometric counterpart in this new context, setting the
stage for the quantization problem in the next chapter.
5
The name comes from the relation of this construction to the Schwartz kernels of operators (see Sec
tion 6.2).
32
The classical picture
The hamiltonian description of classical motions in a conﬁguration space M begins with the
classical phase space T
∗
M. A riemannian metric g = (g
ij
) on M induces an inner product
on the ﬁbers of the cotangent bundle T
∗
M, and a “kinetic energy” function which in local
coordinates (q, p) is given by
k
M
(q, p) =
1
2
¸
i,j
g
ij
(q) p
i
p
j
,
where g
ij
is the inverse matrix to g
ij
. Regular level sets of k
M
are sphere bundles over M.
The hamiltonian ﬂow associated to k
M
is called the cogeodesic ﬂow due to its relation
with the riemannian structure of M described in the following theorem.
Theorem 3.34 If M is a riemannian manifold, then integral curves of the cogeodesic ﬂow
project via π to geodesics on M.
A proof of this theorem can be given in local coordinates by using Hamilton’s equations
˙ q
i
=
∂H
∂p
i
=
¸
j
g
ij
p
j
˙ p
i
= −
∂H
∂q
i
= −
1
2
¸
u,v
∂g
uv
∂q
i
p
u
p
v
to derive the geodesic equation
¨ q
k
+
¸
i,j
Γ
k
ij
˙ q
i
˙ q
j
= 0.
For details, see [36].
In physical terms, Theorem 3.34 states that a free particle on a manifold must move along
a geodesic. A smooth, realvalued potential V : M →R induces the hamiltonian function
H(q, p) = k
M
(q, p) + V (q)
on T
∗
M. Integral curves of the hamiltonian ﬂow of H then project to classical trajectories
of a particle on M subject to the potential V .
The quantum mechanical picture
For the time being, we will assume that the Schr¨odinger operator on a riemannian manifold
M with potential function V is deﬁned in analogy with the ﬂat case of R
n
with its standard
metric. That is, we ﬁrst deﬁne the operator on the function space C
∞
(M) by
ˆ
H = −
2
2m
∆ + m
V
,
where ∆ denotes the LaplaceBeltrami operator. As before,
ˆ
H induces a (densely deﬁned)
operator
ˆ
H on the intrinsic Hilbert space H
M
of M by the equation
ˆ
H(a[dx[
1/2
) = (
ˆ
Ha)[dx[
1/2
,
33
where [dx[ is the natural density associated to the metric on M, and the timeindependent
Schr¨odinger equation on M assumes the familiar form
(
ˆ
H −E)ϕ = 0.
The advantage of this viewpoint is that both the classical state space T
∗
M and the quantum
state space H
M
are objects intrinsically associated to the underlying diﬀerential manifold
M. The dynamics on both objects are determined by the choice of metric on M.
The semiclassical approximation
The basic WKB technique for constructing semiclassical solutions to the Schr¨odinger equa
tion on M proceeds as in Section 2.2. Speciﬁcally, a halfdensity of the form e
iS/h
a is a
secondorder approximate solution of the eigenvector problem (
ˆ
H −E) ϕ = 0 provided that
the phase function S: M →R satisﬁes the HamiltonJacobi equation
H ◦ ι
dS
= E,
and the halfdensity a satisﬁes the homogeneous transport equation, which assumes the
coordinatefree form
a∆S + 2L
∇S
a = 0.
We can formulate this construction abstractly by considering ﬁrst a projectable, exact
lagrangian embedding ι : L → T
∗
M. By deﬁnition, this means that π
L
= π ◦ ι is a diﬀeo
morphism, where π: T
∗
M →M is the natural projection, and Lemma 3.25 implies that for
any primitive φ: L →R of ι
∗
α
M
, the composition S = φ ◦ π
−1
L
is a phase function for (L, ι).
Now if H : T
∗
M → R is any smooth function, then (L, ι) satisﬁes the HamiltonJacobi
equation provided that E is a regular value of H and
H ◦ ι = E.
In this case, the embedding ι and hamiltonian vector ﬁeld X
H
of H induce a nonsingular
vector ﬁeld X
H,ι
on L (see Example 3.13). If a is a halfdensity on L, then the requirement
that (π
−1
L
)
∗
a satisfy the homogeneous transport equation on M becomes
L
X
H,ι
a = 0.
If these conditions are satisﬁed, then the halfdensity e
iφ/
a on L can be quantized (i.e.
pulledback) to yield a secondorder approximate solution (π
−1
L
)
∗
e
iφ/
a to the Schr¨odinger
equation on M, as above.
This interpretation of the WKB approximation leads us to consider a semiclassical state
as a quadruple (L, ι, φ, a) comprised of a projectable, exact lagrangian embedding ι : L →
T
∗
M, a “generalized” phase function φ: L → R satisfying dφ = ι
∗
α
M
, and a halfdensity a
on L. Our correspondence table now assumes the form
34
Object Classical version Quantum version
basic space T
∗
M H
M
state (L, ι, φ, a) as above on M
timeevolution Hamilton’s equations Schr¨odinger equation
generator of evolution function H on T
∗
M operator
ˆ
H on H
M
stationary state state (L, ι, φ, a) such that eigenvector of
ˆ
H
H ◦ ι = E and L
X
H,ι
a = 0
Since the HamiltonJacobi and transport equations above make sense for any triple (L, ι, a)
consisting of an arbitrary lagrangian immersion ι : L →T
∗
M and a halfdensity a on L, it is
tempting to regard (L, ι, a) as a further generalization of the concept of semiclassical state in
T
∗
M, in which we drop the conditions of projectability and exactness. Our goal in the next
chapter will be to determine when and how such “geometric” semiclassical states can be
used to construct “analytical” semiclassical approximate solutions to the timeindependent
Schr¨odinger equation.
35
4 Quantization in Cotangent Bundles
This chapter deals with the problem of constructing semiclassical approximate solutions to
the timeindependent Schr¨odinger equation from the data contained in a “geometric” semi
classical state. The starting point will be a lagrangian immersion ι : L →T
∗
M whose image
in contained in a regular level set of the classical hamiltonian H associated to some metric
and potential on the conﬁguration space M. Given a halfdensity a on L which is invariant
under the ﬂow induced by H, our goal is to use this data in order to construct, in a more or
less systematic way, an approximate solution to Schr¨odinger’s equation on M.
This process will be referred to as quantization. (It is, of course, only one of many
operations which go by this name.) As we shall see, the question of whether or not a given
semiclassical state (L, ι, a) can be quantized is answered largely in terms of the geometry of
the lagrangian immersion (L, ι). For this reason, we will often speak loosely of “quantizing
lagrangian submanifolds” or “quantizable lagrangian submanifolds.”
4.1 Prequantization
The simplest quantization procedure consists of pullingback halfdensities from projectable
lagrangian embeddings in T
∗
M. As seen in Chapter 3, a triple (L, ι, a) for which ι : L →T
∗
M
is a projectable exact lagrangian embedding is quantized by choosing a primitive φ of ι
∗
α
M
and forming the halfdensity
I
(L, ι, a)
def
= (π
−1
L
)
∗
e
iφ/
a
on M. The choice of φ is unique only up to an additive constant, which leads to an ambiguity
in the overall phase of I
(L, ι, a). (This ambiguity was overcome in Chapter 3 by including
the choice of φ in the deﬁnition of a semiclassical state).
To generalize this procedure, suppose now that ι : L → T
∗
M is a projectable, but not
necessarily exact lagrangian embedding. Since the 1form ι
∗
α
M
on L is closed, it is locally
exact by the Poincar´e lemma, and so we can choose a good cover ¦L
j
¦ of L (see Appendix C)
and functions φ
j
: L
j
→ R such that dφ
j
= ι
∗
α
M
[
L
j
. Given a halfdensity a on L, we set
a
j
= a[
L
j
and deﬁne a halfdensity on π
L
(L
j
) by quantizing (L
j
, ι[
L
j
, φ
j
, a
j
) in the sense
above:
I
j
= (π
−1
L
j
)
∗
e
iφ
j
/
a
j
.
To quantize (L, ι, a), we must piece together the I
j
to form a welldeﬁned global halfdensity
I
(L, ι, a) on M. This is possible for arbitrary a provided that the functions φ
j
can be chosen
so that the oscillatory coeﬃcients e
iφ
j
/
agree where their domains overlap; that is, we must
have
φ
j
−φ
k
∈ Z
def
= 2π Z
on each L
j
∩ L
k
. According to the discussion in Appendix C, this is precisely the condition
that the Liouville class λ
L,ι
be integral. At this point, however, we are forced to confront
once more the nature of . If it is a formal variable, the notion of integrality is meaningless.
If denotes a number ranging over an interval of the real numbers, then the Liouville class
will be 2π times an integral class for all only if it is zero, i.e. only for an exact lagrangian
36
submanifold. Since we are speciﬁcally trying to go beyond the exact case, this interpretation
is not acceptable either. Instead, we will make the following compromise.
Deﬁnition 4.1 A projectable lagrangian submanifold (L, ι) ⊂ T
∗
M is quantizable if its
Liouville class λ
L,ι
is integral for some ∈ R
+
. The values of for which this condition
holds will be called admissible for (L, ι).
If L is quantizable but not exact, the set of all admissible forms a sequence converging to
zero, consisting of the numbers
0
/k, where
0
is the largest such number, and k runs over
the positive integers. If L is exact, all > 0 are admissible.
Example 4.2 Let M = S
1
and consider the closed 1form β = p dθ on S
1
, for p ∈ R. The
cohomology class represented by this form is integral provided that
p
S
1
dθ ∈ Z
,
and so the admissible values of are the numbers ¦p/k : k ∈ Z
+
¦. Geometrically this means
that all horizontal circles in the cylinder T
∗
S
1
are quantizable in the sense deﬁned above,
but with diﬀering sets of admissible .
The situation changes as soon as we consider the closed 1form τ = a dθ
1
+ b dθ
2
on the
torus S
1
S
1
. If τ is integral for some ∈ R
+
, then a/ and b/ are both integers,
meaning that a/b is a rational number if b = 0. In general, the condition that closed 1form
β on a manifold M be quantizable is equivalent to the requirement that the ratio of any
two nonzero periods of β be rational. (A period of a closed 1form β on M is any number
obtained by integrating β around some closed loop in M). Thus, the class of projectable
quantizable lagrangian submanifolds of T
∗
M is rather limited whenever dim(H
1
(M; R)) > 1.
´
The simple quantization technique described above does not generalize immediately to
nonprojectable immersed lagrangian submanifolds (L, ι) ⊂ T
∗
M, since π
L
cannot be used to
pushforward halfdensities from L to M. For the time being, however, we will focus on the
set of regular points of π
L
in order to pass from halfdensities on L to halfdensities deﬁned
near noncaustic points of L. Regardless of how this is to be carried out, it is desirable that
quantization be linear with respect to halfdensities:
I
(L, ι, sa
1
+ a
2
) = s I
(L, ι, a
1
) + I
(L, ι, a
2
).
Certainly this condition holds for the procedure we have been using in the projectable case,
and it is reasonable to adopt as a general rule. One consequence is
I
(L, ι, 0) = 0.
If a semiclassical state is represented by the union of a disjoint pair (L
1
, ι
1
, a
1
), (L
2
, ι
2
, a
2
)
of lagrangian submanifolds carrying halfdensities, then by linearity we should have
I
(L
1
∪ L
2
, ι
1
∪ ι
2
, (a
1
, a
2
)) = I
(L
1
∪ L
2
, ι
1
∪ ι
2
, (a
1
, 0)) + I
(L
1
∪ L
2
, ι
1
∪ ι
2
, (0, a
2
))
= I
(L
1
, ι
1
, a
1
) + I
(L
2
, ι
2
, a
2
).
37
Now consider an arbitrary immersed lagrangian submanifold (L, ι) ⊂ T
∗
M and half
density a on L. If p ∈ π
L
(L) is noncaustic and π
L
is proper, then there is a contractible
neighborhood U ⊂ M of p for which π
−1
L
(U) consists of ﬁnitely many disjoint open subsets
L
j
⊂ L such that each (L
j
, ι[
L
j
) is a projectable lagrangian submanifold of T
∗
U. Choosing a
generalized phase function φ
j
: L
j
→R for each L
j
, we note that by the preceding remarks,
the quantization of (L, ι, a) should look something like
¸
j
(π
−1
L
j
)
∗
e
iφ
j
/
a
on U. As before, the requirement that φ
j
be a generalized phase function for L
j
determines
each φ
j
only up to an additive constant, and so the meaning of the preceding sum is ambigu
ous. To quantize halfdensities on L consistently, we must therefore decide how to specify
the relative phases of the oscillatory coeﬃcients e
iφ
j
/
.
If L is exact, then we may use any function φ: L →R satisfying dφ = ι
∗
α
M
to ﬁx phases,
as in the following example.
Example 4.3 For L = R, the lagrangian embedding ι : L → T
∗
R given by ι(x) = (x
2
, x)
has a singular point at x = 0, and we denote by L
+
, L
−
the right and left projectable
components of L, respectively (these correspond to the upper and lower components of the
parabola ι(L)). A phase function φ: L →R for (L, ι) is given by
φ(x) = 2x
3
/3.
If a = B(x) [dx[
1/2
is any halfdensity on L, then since
(π
−1
L
)
∗
∂
∂q
= ±2
−1
q
−1/2
∂
∂x
,
the transformation rule for halfdensities implies
(π
−1
L
+
)
∗
a = 2
−1/2
q
−1/4
B(q
1/2
) [dq[
1/2
(π
−1
L
−
)
∗
a = 2
−1/2
q
−1/4
B(−q
1/2
) [dq[
1/2
.
Thus, the prequantization of (L, ι, a) is given for q > 0 by
I
(L, ι, a)(q) =
e
2iq
3/2
/3
B(q
1/2
) + e
−2iq
3/2
/3
B(−q
1/2
)
2
−1/2
q
−1/4
[dq[
1/2
.
The parabola ι(L) lies in the regular level set H
−1
(0) of the hamiltonian for a constant
force ﬁeld
H(q, p) =
1
2
(p
2
−q),
and it is easy to check that the induced vector ﬁeld X
H,ι
on L equals X
H,ι
= (1/2) ∂/∂x.
Thus, a halfdensity a on L is invariant under the ﬂow of X
H,ι
if and only if a = B[dx[
1/2
for some B ∈ R. From the expression above, we obtain
I
(L, ι, a) =
e
2iq
3/2
/3
+ e
−2iq
3/2
/3
2
−1/2
q
1/4
B[dq[
1/2
38
as a semiclassical approximate solution to the Schr¨odinger equation
−
2
2
∂
2
ψ
∂x
2
−
x
2
ψ = Eψ.
Unfortunately, this solution blows up at q = 0 (and is not deﬁned for q < 0), so we have more
work ahead of us. In particular, we have no way as yet to check that letting φ be continuous
at 0 as a function on L is the right way to assure that we have a good approximation in the
immediate vicinity of q = 0. In fact, we will see later that this is the wrong choice!
´
Since exactness is only used to insure that the function e
iφ/
is welldeﬁned on L, we can treat
certain nonexact cases in a similar way. For this purpose, we make the following provisional
deﬁnition, generalizing Deﬁnition 4.1.
Deﬁnition 4.4 An immersed lagrangian submanfold (L, ι) ⊂ T
∗
M is said to be prequan
tizable if its Liouville class λ
L,ι
is integral for some ∈ R
+
. The values of for which
this condition holds will again be called admissible for (L, ι).
If is admissible for some prequantizable lagrangian immersion (L, ι), then there exists a
good cover ¦V
j
¦ of the manifold L and functions φ
j
: V
j
→ R such that dφ
j
= ι
∗
α
M
[
V
j
and
φ
j
−φ
k
∈ Z
on each V
j
∩ V
k
. Consequently, the φ
j
describe a single function φ: L →T
=
R/Z
which satisﬁes dφ = ι
∗
α
M
and which deﬁnes a global oscillatory function e
iφ/
on L. If
a is a halfdensity on L, we can now quantize (L, ι, φ, a) by summing the pullbacks of e
iφ/
a
to M. In the previous notation, the value of I
(L, ι, φ, a) on U is deﬁned as
¸
j
(π
−1
V
j
)
∗
e
iφ/
a.
Example 4.5 Consider the 1dimensional harmonic oscillator with hamiltonian
H(q, p) =
1
2
(q
2
+ p
2
).
According to Deﬁnition 4.4, a number ∈ R
+
is admissible for the level set H
−1
(E) provided
that
H
−1
(E)
α
1
= 2πE ∈ Z
.
Thus, energy levels of the 1dimensional harmonic oscillator corresponding to level sets for
which a particular value of is admissible are given by E = n. As one can read in any
textbook on quantum mechanics, the actual quantum energy levels are E = (n + 1/2).
The additional 1/2 can be explained geometrically in terms of the nonprojectability of the
classical energy level curves, as we shall soon see.
´
39
Prequantum bundles and contact manifolds
Prequantizability can be described geometrically in terms of principal T
bundles with con
nection over T
∗
M. It is customary to make the following deﬁnition.
Deﬁnition 4.6 For ﬁxed ∈ R
+
, the prequantum T
bundle associated to a cotangent
bundle (T
∗
M, ω
M
) consists of the trivial principal bundle Q
M,
= T
∗
M T
together with
the connection 1form ϕ = −π
∗
α
M
+ dσ.
Here, σ denotes the multiplevalued linear variable in T
and π: Q
M,
→T
∗
M is the bundle
projection. If ι : L → T
∗
M is any lagrangian immersion, then the curvature of the induced
connection on ι
∗
Q
M,
coincides with ι
∗
ω
M
and therefore vanishes. The holonomy of this
connection is represented by the modZ
reduction of the Liouville class λ
L,ι
.
From the prequantum standpoint, the basic geometric object representing a classical
state is therefore a quadruple (L, ι, a, φ) consisting of a lagrangian immersion ι : L →T
∗
M,
a halfdensity a on L, and a parallel lift φ of L to the T
bundle ι
∗
Q
M,
. If we associate
to Q
M,
the prequantum line bundle c
M,
by means of the representation x → e
−ix/
of T
in U(1), then φ induces a parallel section of ι
∗
c
M,
corresponding to the (inverse of the)
oscillatory function e
iφ/
on L which appeared in the preceding section. This remark proves
the following geometric characterization of prequantizability.
Theorem 4.7 An immersed lagrangian submanifold (L, ι) ⊂ T
∗
M is prequantizable if and
only if there exists a nonzero parallel section over L of the line bundle ι
∗
c
M,
for some > 0.
Prequantum T
bundles and parallel lifts of lagrangian submanfolds constitute our ﬁrst
examples of the fundamental objects of contact geometry, the odddimensional counter
part of symplectic geometry. Digressing brieﬂy from quantization, we assemble here a few
facts about contact manifolds.
Deﬁnition 4.8 A contact form on a (2n+1)dimensional manifold Q is a 1form ϕ such
that ϕ ∧ (dϕ)
n
vanishes nowhere on Q. A manifold endowed with a contact form is called a
strict contact manifold.
To interpret the condition on ϕ we note ﬁrst that the kernel of any nowhere vanishing 1form
ϕ deﬁnes a 2ndimensional distribution in Q. If ξ, η are (local) vector ﬁelds lying in this
distribution, we have
dϕ(ξ, η) = ξ ϕ(η) −η ϕ(ξ) −ϕ([ξ, η]) = −ϕ([ξ, η]).
This says that the distribution is integrable (in the sense of Frobenius) iﬀ dϕ is zero on
ker(ϕ). The condition ϕ ∧ (dϕ)
n
= 0, on the other hand, means that the kernel of dϕ is
1dimensional and everywhere transverse to ker(ϕ). Consequently, dϕ is a linear symplectic
form on ker(ϕ), and the “largest” integral submanifolds of ker(ϕ) are ndimensional (i.e.
ker(ϕ) is “maximally nonintegrable”).
Deﬁnition 4.9 A legendrian submanifold of a 2n+1dimensional strict contact manifold
(Q, ϕ) is an ndimensional integral submanifold for ϕ.
40
If ϕ is a contact form and f a nowherevanishing function on Q, then f ϕ is again a contact
form which is said to be equivalent to ϕ, since they have the same legendrian submanifolds.
This leads to the following deﬁnition.
Deﬁnition 4.10 A contact structure on a manifold M is a codimension one subbundle
E ⊂ TM which is locally deﬁned by contact forms, and a manifold endowed with such a
structure is called simply a contact manifold.
When the quotient TM/E is trivial, E is the kernel of a globally deﬁned contact form. Such
a contact structure is called coorientable.
Our basic example of a strict contact manifold is furnished by a prequantum T
bundle
Q
M,
over a cotangent bundle T
∗
M; the image of a parallel lift of a lagrangian immersion
ι : L → T
∗
M to Q
M,
is an immersed legendrian submanifold of Q
M,
. Although we will
not pursue the idea in these notes, a possible generalization of the quantization procedure
described above might therefore begin by reinterpreting geometric semiclassical states as
triples (R, , a) consisting of a legendrian immersion (R, ) in a contact manifold together
with a halfdensity a on R.
4.2 The Maslov correction
It turns out that the naive quantization procedure of the preceding section is incorrect,
since it ignores a certain structure which arises from the relation of L to the ﬁbers of the
projection T
∗
M
π
→M. This factor will be incorporated into our quantization procedure using
a procedure due to Maslov. To begin, we will illustrate this idea in the case of lagrangian
submanifolds of the phase plane.
Given a lagrangian immersion ι : L → T
∗
R · R
2
, we denote by π
p
the composition of ι
with the projection of R
2
onto the paxis. If π
p
is a diﬀeomorphism, then (L, ι) is said to be
pprojectable, in which case there exists an “alternate” generalized phase function τ : L →R
satisfying dτ = ι
∗
(−q dp), obtained by thinking of R
2
as the cotangent bundle of pspace.
A simple example of an embedded lagrangian submanifold of T
∗
R which does not project
diﬀeomorphically onto the qaxis is a vertical line, or ﬁber, of the form ι(x) = (q
0
, x) for x ∈
R. Since the wave function corresponding to a constant halfdensity a on L should correspond
to a probability distribution describing the position of a particle at q
0
with completely
indeterminate momentum, it should be a delta function supported at q
0
. Following an idea
of Maslov, we analyze this situation by pretending that p is position and q momentum, and
then quantizing to obtain a function on pspace. Using the phase function τ(x) = −q
0
x on
L, we obtain
(π
−1
p
)
∗
e
iτ/
[dx[
1/2
= e
−iq
0
p/
q
0
[dp[
1/2
.
The result is exactly the asymptotic Fourier transform (see Appendix B) of the delta function
which we guessed above!
In its simplest form, Maslov’s technique is to suppose that (L, ι) ⊂ T
∗
R is pprojectable,
so that dτ = ι
∗
(−q dp) for some phase function τ on L. If a is a halfdensity on L, we deﬁne
a function B on pspace by the equation
B[dp[
1/2
= (π
−1
p
)
∗
e
iτ/
a.
41
The Maslov quantization of (L, ι, τ, a) is then given by the halfdensity
J
(L, ι, τ, a)
def
= T
−1
(B) [dq[
1/2
on qspace, where T
denotes the asymptotic Fourier transform. To relate this procedure to
our earlier quantization by pullback, we must compare the results in the case of lagrangian
submanifolds L ⊂ R
2
which are biprojectable, i.e. projectable in both the q and pdirections.
For simplicity, we begin with the example of linear lagrangian subspaces.
Example 4.11 For real k = 0, consider the lagrangian embedding ι : R → T
∗
R given by
ι(x) = (x, kx). Generalized phase functions on (L, ι) for the forms p dq and −q dp are given
by φ(x) = kx
2
/2 and τ(x) = −kx
2
/2, respectively. If a is a constant halfdensity on L, then
the transformation rule for halfdensities implies
(π
−1
L
)
∗
a = A[dq[
1/2
(π
−1
p
)
∗
a = [k[
−1/2
A[dp[
1/2
for a real constant A determined by a. Quantization by pullback therefore gives
I
(L, ι, φ, a) = e
ikq
2
/2
A[dq[
1/2
.
On the other hand, we have (π
p
)
∗
τ(p) = −p
2
/2k, and a computation
6
shows that
T
−1
((π
−1
p
)
∗
e
iτ/
)(q) = [k[
1/2
e
−iπsgn(k)/4
e
ikq
2
/2
,
and so Maslov’s technique yields
J
(L, ι, τ, a) = e
−iπsgn(k)/4
I
(L, ι, φ, a).
Thus, the halfdensity obtained from (L, ι, a) by Maslov quantization diﬀers from the simple
pullback by a constant phase shift.
´
To establish a similar correspondence in somewhat greater generality, consider an arbi
trary biprojectable lagrangian embedding (L, ι) ⊂ R
2
with phase functions φ and τ corre
sponding to p dq and −q dp, respectively. For simplicity, we will assume that the additive
constants in φ and τ are chosen so that
φ = τ + ι
∗
(qp).
Next, let S(q) and T(p) be the functions deﬁned on q and pspace by pullback:
S = φ ◦ π
−1
L
T = τ ◦ π
−1
p
,
so that S and T satisfy the Legendre transform relation (see [2])
S(q) = −p(q) T
t
(p(q)) + T(p(q)),
6
See, for example, [32, Vol.1, Thm.7.6.1] .
42
where p(q) = S
t
(q). From this relation, it follows easily that
T
tt
(p(q)) = −(S
tt
(q))
−1
.
A halfdensity a on L determines functions A(q) and B(p) such that
(π
−1
L
)
∗
a = A[dq[
1/2
(π
−1
p
)
∗
a = B[dp[
1/2
and
A(q) = [S
tt
(q)[
1/2
B(p(q)).
For each q, we must now compare the Maslov halfdensity
J
(L, ι, τ, a) = (2π)
−1/2
R
e
i(pq+T(p))/
B(p) dp [dq[
1/2
with that obtained by pullback:
I
(L, ι, φ, a) = e
iS/
A[dq[
1/2
.
To this end, we set k(q) = T
tt
(p(q)) and apply the principle of stationary phase (see Ap
pendix B). The critical point of the exponent pq +T(p) occurs where q = −T
t
(p), i.e. where
p = S
t
(q) = p(q). Hence,
R
e
i(pq+T(p))/
B(p) dp = (2π)
1/2
e
−iπsgn(k)/4
e
iS(q)/
[k(q)[
−1/2
B(p(q)) + O(
3/2
).
Thus
J
(L, ι, τ, a) = e
−iπsgn(k)/4
I
(L, ι, φ, a) +O().
For biprojectable (L, ι), we therefore conclude that Maslov’s technique coincides with quan
tization by pullback up to a constant phase factor and terms of order .
The essential diﬀerence between the naive prequantization of Section 4.1 and the Maslov
quantization of a pprojectable lagrangian embedding (L, ι) which is not qprojectable lies
in the relative phase constants of the summands of I
(L, ι, a), as illustrated by the following
example.
Example 4.12 A phase function associated to −q dp for the lagrangian embedding ι(x) =
(x
2
, x) of L = R into R
2
is given by τ(x) = −x
3
/3. The Maslov quantization of a halfdensity
a = B(x) [dx[
1/2
on L is thus
J
(L, ι, τ, a) = T
−1
e
−ip
3
/3
B(p)
[dq[
1/2
,
since (π
−1
p
)
∗
a = B(p) [dp[
1/2
. For each q > 0, critical points of the function R(p) = pq −p
3
/3
occur precisely when q = p
2
, i.e. when (q, p) ∈ ι(L), and an application of the principle of
stationary phase therefore yields two terms corresponding to the upper and lower halves of
L. Speciﬁcally, we have
J
(L, ι, τ, a) =
e
−iπ/4
e
2iq
3/2
/3
B(−q
1/2
) + e
iπ/4
e
−2iq
3/2
/3
B(q
1/2
)
2
−1/2
q
−1/4
[dq[
1/2
+ O().
43
Compare this with the result of Example 4.3. The extra phase factors of e
∓iπ/4
make
J
(L, ι, τ, a) essentially diﬀerent from the prequantization of (L, ι, a). However, while the
term of order 0 of J
(L, ι, τ, a) is, like I
(L, ι, φ, a), singular at the caustic point q = 0, the
full expression for J
(L, ι, τ, a) as an integral is perfectly smooth there, at least if a has
compact support. This smoothness at caustics is a clear advantage of Maslov quantization.
´
The relative phase factor in the preceding example can be attributed to the fact that the
function T(p) = −p
3
/3 has an inﬂection point at p = 0. More precisely, since T
t
is convex,
a factor of e
iπ/2
arises in passing from the upper to the lower half of the parabola; if T
t
were
concave, the situation would be reversed.
The preceding observation leads us to assign an index to closed, immersed curves in the
phase plane. A pdependent phase function T for L ⊂ R
2
will have inﬂection points at
precisely those p for which (T
t
(p), p) is a singular point of L. Moreover, the sign of T
tt
at
nearby points depends only on L and not on the choice of T. With these remarks in mind,
suppose that (L, ι) is a closed, immersed curve in R
2
which is nondegenerate in the sense
that if T is a pdependent phase function for a subset of L, then T
t
has only nondegenerate
critical points. Under this assumption, sgn(T
tt
) changes by ±2 in the vicinity of a critical
point of T
t
, and we can assign an index to (L, ι) by summing these changes while traversing
L in a prescribed direction. The result is twice an integer known as the Maslov index m
L,ι
of (L, ι). (Compare [3]).
Example 4.13 The computation of the Maslov index can be interpreted geometrically if
we ﬁrst observe that the nondegeneracy condition requires L to remain on the same side of
a ﬁber π
−1
(q) near a singular point. Since the index only involves the sign of T
tt
, it follows
from Example 4.12 that the integer 1 should be assigned to a critical point of π
L
which is
traversed in the −p direction to the right of the ﬁber. In other words, downward motion to
the right of the ﬁber is positive, while the sign changes if either the direction of motion or
the side of the ﬁber is reversed, but not if both are.
A circle in the phase plane has a right and a left singular point, both of which are positive
according our rule when the circle is traversed counterclockwise. The Maslov index of the
circle therefore equals 2. In fact, the same is true of any closed embedded curve traversed
counterclockwise. On the other hand, a ﬁgureeight has Maslov index zero.
The reader is invited to check that if L is a circle with a ﬁxed orientation ν ∈ H
1
(L; Z),
then m
L,ι
equals 'µ
L,ι
, ν`, where µ
L,ι
is the Maslov class of (L, ι) deﬁned in Example 3.22.
´
Our goal is now to modify the prequantization procedure for arbitrary lagrangian sub
manifolds of the phase plane by incorporating the Maslov index. The basic idea is as follows.
Given an immersed lagrangian submanifold (L, ι) ⊂ R
2
, we ﬁrst choose a good cover ¦L
j
¦
of L such that the image of each L
j
under ι is either q or pprojectable and no intersection
L
j
∩ L
k
contains a critical point of π
L
. Next, we ﬁx a partition of unity ¦h
j
¦ subordinate
44
to ¦L
j
¦. To quantize a halfdensity a on L, we quantize each (L
j
, ι, a h
j
) to obtain a half
density I
j
on R either by pullback or by Maslov’s technique. As before, we would then like
to deﬁne the quantization of (L, ι, a) as the sum
I
(L, ι, a) =
¸
j
I
j
.
In order to specify the relative phases of the I
j
, and to make this deﬁnition independent of
the choice of cover ¦L
j
¦ and partition of unity, we will require that I
(L, ι, a) coincide up
to order with the usual quantization by pullback for any halfdensity a supported in a
projectable subset of L. This condition can be precisely formulated in terms of the Maslov
index and Liouville class of (L, ι) as follows.
On an open interval U of noncaustic points, each halfdensity I
j
is the sum of the pull
back of halfdensities on each component of L
j
∩π
−1
L
(U). If L
j
is quantized by pullback, this
statement is obvious; if Maslov’s technique is applied to L
j
, it follows from an application of
the principle of stationary phase as in Example 4.12. On L
j
∩ π
−1
L
(U), these halfdensities
are of the form
˜
I
j
= e
−iπs
j
/4
e
iφ
j
/
a.
Here, φ is a realvalued function on L
j
satisfying dφ
j
= ι
∗
α
1
, while s
j
are integers depending
only on the component of L
j
∩ π
−1
L
(U) in question. (More precisely, s
j
is zero if L
j
is
qprojectable and is quantized by pullback; otherwise s
j
equals sgn(T
tt
) for a suitable p
dependent phase function). For I
(L, ι, a) to be welldeﬁned, we must choose the functions
φ
j
so that
˜
I
j
=
˜
I
k
on each intersection L
j
∩L
k
regardless of the particular halfdensity a. In
other words, we require
e
i(φ
j
−φ
k
)/
e
−iπ(s
j
−s
k
)/4
= 1
at each point of L
j
∩ L
k
. Since φ
j
− φ
k
is constant on L
j
∩ L
k
, we can deﬁne a
jk
as the
(constant) value of (φ
j
−φ
k
) −π(s
j
−s
k
)/4 on L
j
∩ L
k
, so that our requirement becomes
a
jk
∈ Z
.
Evidently, this condition can be fulﬁlled on any arc of a curve in the phase plane. If L is
a circle, then this condition implies that the sum of any of the a
jk
lies in Z
, or, in other
words, that the Maslov index m
L,ι
of (L, ι) satisﬁes
π
2
m
L,ι
+
L
ι
∗
α
1
∈ Z
.
This is the simplest version of the Maslov quantization condition.
Example 4.14 Returning to the harmonic oscillator of Example 4.5, we see that the level
set H
−1
(E) satisﬁes the Maslov condition provided that for some integer n,
E = (n + 1/2).
Allowable energy levels in this case therefore correspond to the BohrSommerfeld condition,
which actually gives the precise energy levels for the quantum harmonic oscillator.
´
45
A general quantization scheme
Motivated by the simple results above, our aim in the next sections will be to develop a
systematic method for quantizing lagrangian submanifolds of cotangent bundles. The basis
of this method will again be Maslov’s technique, which relies in this context on the concept
of generalized phase functions.
Local parametrizations of an immersed lagrangian submanifold (L, ι) ⊂ T
∗
M deﬁned by
such phase functions will ﬁrst enable us to quantize a given halfdensity locally on L by
means of a slightly more general version of the (inverse) asymptotic Fourier transform. The
result will be a collection of halfdensities on M. In order for these halfdensities to piece
together appropriately, it is necessary and suﬃcient that L satisfy a general version of the
Maslov quantization condition, which we formulate in the next section using the Maslov
class.
4.3 Phase functions and lagrangian submanifolds
In this section, we generalize the concept of phase functions to nonprojectable lagrangian
submanifolds of cotangent bundles. Roughly speaking, the idea is the following. As we
saw in Chapter 3, a projectable lagrangian submanifold of T
∗
N can be locally parametrized
by the diﬀerential of a function f on U ⊂ N, viewed as a mapping df : U → T
∗
N. For
each p ∈ U, the meaning of df
p
as an element of T
∗
p
N is that for any smooth curve γ in N
satisfying γ(0) = p, we have 'df
p
, ˙ γ(0)`
def
= (f◦γ)
t
(0). To parametrize more general lagrangian
submanifolds of T
∗
N in a similar way, we can begin with a function ϕ: UR
m
→R together
with a point ˜ p = (p, v) ∈ UR
m
and attempt to deﬁne dϕ
˜ p
∈ T
∗
p
N by 'dϕ
˜ p
, ˙ γ(0)` = (ϕ◦˜ γ)
t
(0)
for any lift ˜ γ of γ to the product U R
m
such that ˜ γ(0) = ˜ p. In general, this fails, since
the value of the directional derivative (ϕ ◦ ˜ γ)
t
(0) depends on the lift ˜ γ. If, however, the
ﬁberderivative ∂ϕ/∂θ vanishes at ˜ p, the expression for dϕ
˜ p
produces a welldeﬁned element
of T
∗
p
N. As we shall see, the assumption that the map ∂ϕ/∂θ : U R
m
→R
m
is transverse
to 0 implies that the ﬁber critical set
Σ
ϕ
=
˜ p ∈ U R
m
:
∂ϕ
∂θ
= 0
is a smooth submanifold of U R
m
, and the assignment ˜ p → dϕ
˜ p
deﬁnes a lagrangian
immersion of Σ
ϕ
into T
∗
N. In general, the restriction of the projection U R
m
→U to Σ
ϕ
is noninjective, and thus the image of Σ
ϕ
is a nonprojectable lagrangian submanifold. From
the point of view of the WKB method, this generalization amounts to replacing Maslov’s
ansatz
(2π)
−n/2
R
n
e
i(/p,q)+T(p))/
a(p) dp [dq[
1/2
for the solution of Schr¨odinger’s equation by the more general form
(2π)
−m/2
R
m
e
iφ(q,θ)/
a(q, θ)[dθ[[dq[
1/2
,
46
where θ is an auxiliary variable in R
m
which may have nothing to do with the variable p dual
to q. An advantage of this generalization will be to allow a calculus which is more clearly
invariant under changes of coordinates, unlike the previous Fourier transform picture, which
requires linear structures on p and qspace.
To begin, we ﬁx some notation and terminology.
7
Let M, B be smooth manifolds, and
let p
M
: B →M be a smooth submersion. Dualizing the inclusion E = ker(p
M∗
)
ι
→TB gives
rise to an exact sequence of vector bundles over B
0 ←E
∗
ι
∗
←T
∗
B ←E
⊥
←0
where E
⊥
⊂ T
∗
B denotes the annihilator of E. The ﬁberderivative of a function φ: B →
R is the composition d
θ
φ = ι
∗
◦ dφ, and its ﬁber critical set is deﬁned as
Σ
φ
= (d
θ
φ)
−1
Z
E
∗.
(We will denote the zero section of a vector bundle F by Z
F
). The function φ is said to
be nondegenerate if its ﬁber derivative is transverse to Z
E
∗, in which case Σ
φ
is a smooth
submanifold of B. At points of Σ
φ
, the section d
θ
φ has a welldeﬁned intrinsic derivative (see
[26]), denoted ∇d
θ
φ, which induces for nondegenerate φ an exact sequence of vector bundles
over Σ
φ
0 →TΣ
φ
→T
Σ
φ
B
∇d
θ
φ
→ E
∗
[
Σ
φ
→0.
The ﬁberhessian Hφ of φ at p ∈ Σ
φ
is deﬁned as the composition ∇d
θ
φ ◦ ι : E
p
→E
∗
p
.
The bundle E
⊥
may be identiﬁed with the pullback p
∗
M
T
∗
M, giving rise to a natural
projection E
⊥
p
→ T
∗
M. On the ﬁbercritical set Σ
φ
, the diﬀerential dφ deﬁnes a section of
E
⊥
whose composition with p we denote by λ
φ
: Σ
φ
→T
∗
M.
Theorem 4.15 If φ is nondegenerate, then the map λ
φ
: Σ
φ
→T
∗
M is an exact lagrangian
immersion.
Proof. Since E
⊥
is the union of the conormal bundles of the ﬁbers of p
M
, it follows from
Example 3.19 that E
⊥
is a coisotropic submanifold of T
∗
B. Moreover, the characteristic
distribution (
⊥
of E
⊥
is tangent to the ﬁbers of the mapping E
⊥
→T
∗
M.
The nondegeneracy assumption on φ is equivalent to the requirement that the lagrangian
submanifold (dφ)(B) of T
∗
B be transverse to E
⊥
. By Lemma 3.6(2), this implies that the
section (dφ)(B)[
Σ
φ
= (dφ)(B) ∩ E
⊥
is nowhere tangent to the distribution (
⊥
and therefore
immerses into T
∗
M. To complete the proof, we note that the equality
λ
∗
φ
α
M
= dφ[
Σ
φ
implies that L
φ
is exact lagrangian.
2
7
Some readers may ﬁnd it instructive to follow the ensuing discussion by writing everything in local
coordinates.
47
From the preceding deﬁnitions, it follows that the nullity k of the ﬁberhessian at p ∈ Σ
φ
equals dim(T
p
Σ
φ
∩E
p
), which in turn equals the dimension of the kernel of (π◦λ
φ
)
∗
on T
p
Σ
φ
.
This means in particular that the dimension of the ﬁbers of B must be at least k; if the ﬁber
dimension of B equals k, then the phase function φ is said to be reduced at p. This occurs
when its ﬁberhessian vanishes at p.
Example 4.16 Suppose that M = R and consider the function φ(x, θ) = xθ + θ
3
/3 on
B = RR. Here, d
θ
φ = θ
2
+x, and ∇(d
θ
φ) = 2θdθ +dx. Evidently the phase function φ is
nondegenerate, and its ﬁber critical set consists of the parabola x = −θ
2
, which fails to be
xprojectable precisely when x = 0. This is just the value of x for which φ(x, ) acquires a
degenerate critical point.
´
A phase function which generates a neighborhood of a point in a lagrangian submanifold
is not unique per se, but it is unique up to “stable equivalence,” a concept which we now
deﬁne. To begin, we introduce the following terminology.
Deﬁnition 4.17 A triple (B, p
B
, φ) is called a Morse family over a manifold M if p
B
:
B → M is a smooth (possibly nonsurjective) submersion, and φ is a nondegenerate phase
function on B such that λ
φ
is an embedding. A Morse family is said to be reduced at p ∈ B
if φ is a reduced phase function at p.
We will say that the lagrangian submanifold im(λ
φ
) = L
φ
is generated by the Morse family
(B, p
B
, φ). If ι : L → T
∗
M is a lagrangian immersion and p ∈ L, then we denote by
M(L, ι, p) the class of Morse families (B, p
B
, φ) which generate ι(U) for some neighborhood
U ⊂ L of p. For such Morse families, we denote by g
φ
: U →Σ
φ
the diﬀeomorphism deﬁned
by g
φ
= λ
−1
φ
◦ ι.
If (B, p
B
, φ) ∈ M(L, ι, p), then the following operations produce further elements of
M(L, ι, p).
1. Addition: For any c ∈ R, (B, p
B
, φ + c) ∈ M(L, ι, p).
2. Composition: If p
B
: B
t
→ M is a second submersion and g : B
t
→ B is a ﬁber
preserving diﬀeomorphism, then (B
t
, p
B
, φ ◦ g) ∈ M(L, ι, p).
3. Suspension: The suspension of (B, p
B
, φ) by a nondegenerate quadratic form Q on
R
n
is deﬁned as the Morse family comprised of the submersion ˜ p
B
: B R
n
→ M
given by composing p
B
with the projection along R
n
, together with the phase function
˜
φ = φ + Q. Evidently the ﬁbercritical set of
˜
φ equals the product Σ
φ
¦0¦, and
λ
˜
φ
(b, 0) = λ
φ
(b) for all (b, 0) ∈ Σ
˜
φ
. Thus (B R
n
, ˜ p
B
,
˜
φ) ∈ M(L, ι, p).
4. Restriction: If B
t
is any open subset of B containing λ
−1
φ
(p), then the restrictions of
p
B
and φ to B
t
deﬁne a Morse family on M which belongs to M(L, ι, p).
These operations generate an equivalence relation among Morse families called stable equiv
alence. The central result of this section is the following.
48
Theorem 4.18 Let ι : L →T
∗
M be a lagrangian immersion, and let p ∈ L. Then:
1. The class M(L, ι, p) contains a reduced Morse family over M.
2. Any two members of M(L, ι, p) are stably equivalent.
The next two subsections are devoted to the proof of this theorem.
Symplectic normal forms
The purpose of this section is to develop several results needed to prove Theorem 4.18, all of
which more or less rely on the socalled deformation method. This method was introduced
by Moser in [45] (it probably has a longer history) and applied to a variety of problems of
symplectic geometry in [62]. We will prove the following theorem.
Theorem 4.19 Let (P, ω) be a symplectic manifold, ι : L →P a lagrangian immersion, and
E a lagrangian subbundle of ι
∗
TP which is complementary to the image of ι
∗
: TL →ι
∗
TP.
Then there is a symplectic immersion ψ of a neighborhood U of the zero section Z ⊂ T
∗
L
into P such that ψ = ι ◦π on Z and which maps the vertical subbundle V
Z
L ⊂ T
Z
(T
∗
L) onto
E.
As a result of the existence of compatible complex structures on ι
∗
P, lagrangian subbundles
of ι
∗
P complementary to the image of ι
∗
: TL → ι
∗
TP always exist. Thus, Theorem 4.19
implies the following nonlinear “normal form” result, which states that, near any of its
lagrangian submanifolds L, a symplectic manifold looks like a neighborhood of the zero
section in T
∗
L.
Corollary 4.20 If L is a lagrangian submanifold of P, then a neighborhood of L in P is
symplectomorphic to a neighborhood of the zero section in T
∗
L, by a map which is the identity
on L.
Theorem 4.19 may be combined with Proposition 3.24 to yield:
Corollary 4.21 The map ψ gives a 11 correspondence between a neighborhood of L in the
space of all lagrangian submanifolds of (P, ω) and a neighborhood of the zero section in the
space Z
1
(L) of closed 1forms on L.
The proof of Theorem 4.19 will rely on the following
Relative Darboux theorem [62] . Let N be a submanifold of a manifold P, and let
ω
0
, ω
1
be two symplectic forms on P which coincide on T
N
P. Then there are neighborhoods
U and V of N and a diﬀeomorphism f : U →V such that
1. f
∗
ω
1
= ω
0
2. f[
N
= id and Tf[
T
N
P
= id.
49
Proof. Deﬁne a 1parameter family of closed forms on P by
ω
t
= ω
0
+ t(ω
1
−ω
0
).
Since all ω
t
agree on the submanifold N, there is a neighborhood of N in P (which for our
purposes we may assume to be P itself) on which all ω
t
are nondegenerate.
To ﬁnd f satisfying f
∗
ω
1
= ω
0
, we will construct a timedependent vector ﬁeld X
t
for
which the isotopy f
t
that it generates satisﬁes f
∗
t
ω
t
= ω
0
for all t ∈ [0, 1]. By the usual
properties of the Lie derivative, it is necessary and suﬃcient that such a vector ﬁeld solve
the equation
0 =
d
dt
(f
∗
t
ω
t
) = f
∗
t
dω
t
dt
+ f
∗
t
(L
Xt
ω
t
) = f
∗
t
(ω
1
−ω
0
+ d(X
t
ω
t
)).
In order to ﬁx N, we also want X
t
[
N
= 0. These conditions will be satisﬁed if we set
X
t
= −˜ ω
−1
t
(ϕ) for a 1form ϕ on a neighborhood of N in P, vanishing on N, such that
dϕ = ω
1
− ω
0
. If the submanifold N consisted of a single point, the form ϕ would be easy
to ﬁnd, since the closed 2form ω
1
−ω
0
is locally exact by the classical Poincar´e lemma. For
more general submanifolds, we use a more general version of this lemma:
Relative Poincar´e lemma . Let N be a submanifold of P, and let β be a closed kform on
P which vanishes on TN. Then there is a form ϕ on a neighborhood of N such that dϕ = β
and which vanishes on T
N
P. Furthermore, if β vanishes on T
N
P, then ϕ can be chosen so
that its ﬁrst derivatives vanish along N.
Proof. Since the statement is local around N, we will identify P with a tubular neighborhood
of N in P and let h
t
: P →P be a smooth isotopy such that
h
1
= id h
0
= ﬁberwise projection of P onto N.
Since β vanishes on N, we have
β = h
∗
1
β −h
∗
0
β =
1
0
d
dt
(h
∗
t
β) dt =
1
0
h
∗
t
(L
Yt
β) dt,
where Y
t
is the (timedependent) vector ﬁeld which generates the isotopy h
t
for t > 0. By
Cartan’s formula and the fact that β is closed, the last expression is equal to
1
0
h
∗
t
d(Y
t
β) dt = d
1
0
h
∗
t
(Y
t
β) dt
.
Our assertion follows if we take ϕ =
1
0
h
∗
t
(Y
t
β) dt.
2
If (P, ω) is a symplectic manifold and g
t
: P →P a continuous family of diﬀeomorphisms,
then homotopy invariance for deRham cohomology implies that the forms ω
t
= g
∗
t
ω lie in
the same cohomology class for all t. By the same method of proof as in the relative Darboux
theorem, we can deduce the following converse result:
50
Corollary 4.22 (Moser [45]) If ¦ω
t
¦ is a family of symplectic structures on a compact
manifold P and ω
t
1
− ω
t
2
is exact for all t
1
, t
2
, then there is a diﬀeomorphism f : P → P
with f
∗
1
ω
1
= ω
0
.
By restricting to a neighborhood of a point in N, Givental [6] proves an equivalence theorem
like the relative Darboux theorem under the weaker hypothesis that ω
0
and ω
1
coincide on
TN, i.e. just for vectors tangent to N. In contrast to Corollary 4.22, McDuﬀ [40] describes
a family ω
t
of symplectic forms on a compact sixdimensional manifold P such that ω
1
−ω
0
is exact, but there is no diﬀeomorphism f : P → P at all satisfying f
∗
ω
1
= ω
0
. Of course,
for intermediate t, the form ω
t
−ω
0
is not exact.
Proof of Theorem 4.19. Beginning with the lagrangian immersion ι : L → P, we let
f : Z →P denote the composition ι ◦ π, where Z is the zero section of T
∗
L and π: T
∗
L →L
is the natural projection. Using Theorem 3.20, we then choose a ωcompatible complex
structure J on f
∗
TP which satisﬁes J(f
∗
TZ) = E, where E is the lagrangian subbundle given
in the statement of the theorem. Similarly, we choose a ω
L
compatible complex structure J
t
on T
Z
(T
∗
L) which rotates TZ into V
Z
L. Now consider the symplectic bundle map
T
Z
(T
∗
L)
˜
f
→f
∗
TP
given by
˜
f(v ⊕ J
t
w) = f
∗
v ⊕ J(f
∗
w) for v, w ∈ TZ. In a neighborhood U ⊂ T
∗
L of Z, we
can extend f to obtain an immersion F : U → P which satisﬁes F
∗
=
˜
f on T
Z
(T
∗
L). To
ﬁnish the proof of the normal form theorem, we apply the relative Darboux theorem to the
forms ω
L
and F
∗
(ω) on U.
2
Next, we turn to two generalizations of the classical Morse lemma (see Appendix B)
which require the following version of Taylor’s theorem.
Lemma 4.23 Let N ⊂ B be a submanifold deﬁned by the vanishing of functions g
1
, , g
k
whose diﬀerentials are linearly independent along N. If f is any function such that f and
df vanish at all points of N, then there exist functions c
ij
such that
f =
¸
i,j
c
ij
g
i
g
j
on a neighborhood of N.
Parametrized Morse lemma . Let f : M R
k
→ R satisfy the condition that for
each x ∈ M, (x, 0) is a nondegenerate critical point for f[
x¦R
k. Then for each x
0
∈ M,
there exists a neighborhood U of (x
0
, 0), a nondegenerate quadratic form Q on R
k
, and a
diﬀeomorphism u of U ﬁxing M¦0¦ and preserving ﬁbers of the projection to M such that
f(u(x, θ)) = f(x, 0) + Q(θ).
51
Proof. By a preliminary change of coordinates linear on ﬁbers, we may assume that
f(x, θ) = f(x, 0) +Q(θ) + a(x, θ)
where the error term a(x, θ) is O([θ[
3
) for each x. To ﬁnd the diﬀeomorphism u we apply
the deformation method. Deﬁne f
t
= f −(1 −t)a. We seek a vector ﬁeld X
t
tangent to the
ﬁbers x = constant which generates an isotopy u
t
ﬁxing M ¦0¦ and satisfying f
t
◦ u
t
= f
0
for all t ∈ [0, 1]. This means that X
t
should be of the form
X
t
=
¸
i
h
i
t
∂
∂θ
i
,
for certain smooth functions h
i
t
which vanish on M ¦0¦, and it should satisfy
0 =
d
dt
(f
t
◦ u
t
) = u
∗
t
(X
t
f
t
+ a)
for all t ∈ [0, 1]. Evidently, the latter condition will be met if X
t
is chosen so that
X
t
f
t
+ a = 0.
To determine X
t
, we invoke Lemma 4.23 in order to ﬁnd smooth functions c
ij
t
which vanish
on M ¦0¦ and satisfy
a =
¸
c
ij
t
∂f
t
∂θ
i
∂f
t
∂θ
j
.
The required condition
¸
i
h
i
t
∂f
t
∂θ
i
= −
¸
i,j
c
ij
t
∂f
t
∂θ
i
∂f
t
∂θ
j
will be satisﬁed if we set
h
i
t
= −
¸
j
c
ij
t
∂f
t
∂θ
j
,
which vanishes (to second order) on M ¦0¦.
2
If φ is a phase function on M R
n+k
whose ﬁberhessian Hφ has rank k at a point b ∈ Σ
φ
,
then by the lowersemicontinuity of rank Hφ, there exists an integrable subbundle F
→E =
ker(p
M∗
) such that the ﬁberhessian
∗
◦ ∇d
θ
φ ◦ : F →F
∗
is nondegenerate at each zero of
the section
∗
◦d
θ
φ in a neighborhood of b. A change of coordinates near b and an application
of the parametrized Morse lemma then proves:
Thom splitting theorem . Let φ be a function on B = M R
n+k
whose ﬁberhessian
Hφ has rank k at a point b ∈ Σ
φ
. Then there exists a ﬁberpreserving diﬀeomorphism g of
B, a function η : M R
n
→R, and a nondegenerate quadratic form Q on R
k
such that
(φ ◦ g)(x, θ, θ
t
) = η(x, θ) + Q(θ
t
)
in a neighborhood of b in B. The function η is totally degenerate in the sense that its
ﬁberhessian is identically zero at b.
2
52
Existence and stable equivalence of Morse families
Returning to the proof of Theorem 4.18, we ﬁrst note the following reinterpretation of the
Thom splitting theorem in the terminology of Morse families.
Theorem 4.24 Any Morse family in M(L, ι, p) is stably equivalent to a reduced Morse
family in M(L, ι, p).
To show that M(L, ι, p) is nonempty, we ﬁrst prove
Theorem 4.25 Suppose that ι : L → T
∗
M is an exact lagrangian immersion such that
ι
∗
T(T
∗
M) admits a lagrangian subbundle F transverse to both ι
∗
V M and ι
∗
TL. Then L is
generated by a Morse family over M.
Proof. By Theorem 4.19, there exists a symplectic immersion ψ from a neighborhood U of
the zero section Z ⊂ T
∗
L into T
∗
M which satisﬁes ψ = ι ◦ π on Z and which maps V
Z
L
onto the subbundle F. Since ψ is symplectic, the 1form ψ
∗
α
M
− α
L
is closed on U. Its
restriction to Z is exact because α
L
is zero on Z. By the relative Poincar´e lemma, there
exists a function φ on U satisfying dφ = ψ
∗
α
M
−α
L
.
To complete the proof, we ﬁrst note that the restriction of dφ to a ﬁber of the compo
sition of ψ with the projection π
t
: T
∗
M → M equals −α
L
. Combined with the deﬁnition
of α
L
, a computation (using the transversality hypothesis) now shows that there exists a
neighborhood B of Z within U such that (B, π
t
(ψ(B)), π
t
◦ ψ, φ) is a Morse family such that
Σ
φ
= Z and λ
φ
= ι ◦ π for the projection π: T
∗
L →L.
2
Proof of Theorem 4.18. Using Lemma 3.9 we can construct a subbundle F which satisﬁes
the hypotheses of the preceding theorem over a neighborhood of any p in L. Consequently,
there exists a Morse family generating a neighborhood of p, and so the class M(L, ι, p) is
nonempty. Combined with Theorem 4.24, this proves part (1) of Theorem 4.18.
To prove part (2), it suﬃces by Theorem 4.24 to show that any two reduced Morse
families in M(L, ι, p) are stably equivalent. Since this property is local near p, we may
assume that M = R
m
and and that φ,
˜
φ are both deﬁned on R
m
R
k
. Consider the
mappings F,
˜
F : R
m
R
k
→R
m
R
m
deﬁned by
F(x, y) =
x,
∂φ
∂x
(x, y)
˜
F(x, y) =
x,
∂
˜
φ
∂x
(x, y)
.
By the assumption that φ,
˜
φ are nondegenerate phase functions, it follows that F and
˜
F are
embeddings near q = λ
−1
φ
(p) and ˜ q = λ
−1
˜
φ
(p) respectively. Moreover, F coincides with the
map λ
φ
on the ﬁbercritical set Σ
φ
near q, and similarly for
˜
F. Since the Morse families
are reduced, it follows that the images of DF
q
and D
˜
F
˜ q
coincide in T
p
(R
m
R
m
). By an
application of the implicit function theorem, we can therefore ﬁnd a ﬁberpreserving map g
53
of R
m
R
m
whose restriction to some neighborhood B of q is a diﬀeomorphism which sends
Σ
φ
to Σ
˜
φ
and in particular g(q) = ˜ q.
Recall from the proof of Theorem 4.15 that the identity
dφ = λ
∗
φ
α
M
holds on the ﬁber critical set Σ
φ
. Since λ
φ
= λ
˜
φ
◦ g, this identity implies that dφ = d(
˜
φ ◦ g)
on Σ
φ
, and so the phase functions φ and
˜
φ ◦ g diﬀer by an additive constant (which we may
assume to be zero) at points of Σ
φ
. In order for g to deﬁne an equivalence between φ and
˜
φ, this property would have to be valid on all of B. While there is no reason to expect this
of g itself, we will construct a similar diﬀeomorphism with this property by appealing once
again to the deformation method:
Let φ
0
= φ, φ
1
=
˜
φ ◦ g. Then Σ
φ
0
= Σ
φ
1
= Σ and λ
φ
0
= λ
φ
1
. Moreover φ
1
−φ
0
vanishes
to second order along Σ, and so there exist functions c
ij
deﬁned near b such that
φ
1
−φ
0
=
¸
i,j
c
ij
∂φ
0
∂θ
i
∂φ
0
∂θ
j
in a neighborhood of Σ
φ
.
As before, we seek a vector ﬁeld X
t
generating an isotopy f
t
that satisﬁes f
∗
t
φ
t
= φ
0
for
all t ∈ [0, 1]; to insure that each f
t
ﬁxes Σ and preserves ﬁbers, we must also require that X
t
be of the form
X
t
=
¸
i,j
h
ij
∂φ
0
∂θ
i
∂
∂θ
j
for certain functions h
ij
which vanish on Σ.
To arrive at an equation for these coeﬃcients, we note that the equation
0 =
d
dt
(f
∗
t
φ
t
) = f
∗
t
(X
t
φ
t
+ φ
1
−φ
0
)
will be satisﬁed provided that
0 = X
t
φ
t
+ φ
1
−φ
0
,
i.e.
0 =
¸
i,j
c
ij
∂φ
0
∂θ
i
∂φ
0
∂θ
j
+
¸
i,j
h
ij
∂φ
0
∂θ
i
∂
∂θ
j
φ
0
+ t
¸
u,v
c
uv
∂φ
0
∂θ
u
∂φ
0
∂θ
v
.
This equation holds if
0 = C + H(I + S),
where C = (c
ij
), H = (h
ij
), and S is a matrix function which vanishes at b for all t, since φ
0
is reduced. Hence we can solve for H in a neighborhood of b. This completes the proof.
2
54
Maslov objects
If ι : L → T
∗
M is any lagrangian immersion, then the symplectic vector bundle ι
∗
T(T
∗
M)
over L has two lagrangian subbundles, L
1
= ι
∗
TL and L
2
= ι
∗
V M. The Maslov class of
(L, ι) is deﬁned as the degree1 cohomology class
µ
L,ι
= µ(L
1
, L
2
) ∈ H
1
(L; R)
described in Example 3.21. From this deﬁnition, it follows that, unlike the Liouville classes,
the Maslov classes of two immersions (L, ι) and (L, ι
t
) are equal whenever ι and ι
t
are
homotopic through lagrangian immersions.
Associated to any Morse family (B, p
B
, φ) over a manifold M is an index function ind
φ
:
L
φ
→Z deﬁned by
ind
φ
(p) = index(Hφ
λ
−1
φ
(p)
),
where the index of a quadratic form is, as usual, the dimension of the largest subspace on
which it is negativedeﬁnite. Since the ﬁberhessian is nondegenerate where L
φ
is projectable
(see the discussion following Theorem 4.15), the index function ind
φ
is constant on any
connected projectable subset of L
φ
. From Theorem 4.18, it follows furthermore that any two
index functions ind
φ
, ind
φ
diﬀer by an integer on each connected component of L
φ
∩ L
φ
.
Example 4.26 Consider φ(x, θ) = θ
3
/3 + θ(x
2
− 1). We have ∂φ/∂θ = θ
2
+ x
2
− 1; thus
the ﬁber critical set Σ
φ
is the circle θ
2
+ x
2
= 1, and its image under λ
φ
is the ﬁgureeight.
The caustic set of the projection L → R consists of the points (−1, 0), (1, 0). To compute
the index function corresponding to φ, we note that
∂
2
φ
∂θ
2
= 2θ;
consequently ind
φ
(x, p) = 0 for (x, p) lying on the upper right part of the curve and
ind(x, p) = 1 for (x, p) on the upper left. Since this loop is generated by a single phase
function, there is a corresponding global index function, and the index of the loop is neces
sarily zero.
A diﬀerent situation occurs in the case of a circle. The circle has two caustics, and the
jump experienced by any index function as one passes through a caustic is given by the
example just computed: passing through the right caustic in the +p direction decreases the
index by 1, while passing through the left caustic in the same direction increases the index
by 1. Traversing the circle in a counterclockwise direction, we see that the total index must
change by −2. Consequently, the circle does not admit a global generating function.
Of course, the Liouville class of the circle is also nonzero. However, it is easy to deform
the circle to a closed curve with two transverse selfintersections which has zero Liouville
class, but still admits no global phase function, since its Maslov class is nonzero.
´
55
As a degree1 cohomology class on L, the Maslov class determines via the exponential
map R → U(1) an isomorphism class of ﬂat hermitian line bundles over L. A canonical
representative of this class can be constructed as follows. First, we consider the union
M(L, ι) =
¸
p∈L
M(L, ι, p)
with the discrete topology. On the subset of LM(L, ι)Z consisting of all (p, (B, p
B
, φ), n)
such that (B, p
B
, φ) ∈ M(L, ι, p), we now introduce an equivalence relation ∼ by setting
(p, (B, p
B
, φ), n) ∼ (˜ p, (
˜
B, p
˜
B
,
˜
φ), ˜ n) provided that p = ˜ p and
n + ind
φ
(p) = ˜ n + ind
˜
φ
(p).
The quotient space with respect to this relation is a principal Zbundle M
L,ι
over L which
we will call the Maslov principal bundle.
Associated to the Maslov principal bundle via the representation n →e
iπn/2
of Z in U(1) is
a complex line bundle ´
L,ι
over L called the Maslov line bundle. Having discrete structure
group, this line bundle carries a natural ﬂat connection with holonomy in ¦e
iπn/2
¦ · Z
4
.
Our main use of the Maslov line bundle will be to modify the halfdensities on L in order to
incorporate the Maslov correction into our quantization scheme.
4.4 WKB quantization
In this section we will combine the tools assembled in the preceding sections into a technique
for quantizing halfdensities on lagrangian submanifolds of arbitrary cotangent bundles.
The phase bundle associated to an immersed lagrangian submanifold ι : L →T
∗
M and
> 0 is deﬁned as the tensor product
Φ
L,ι,
def
= ´
L,ι
⊗ι
∗
c
M,
,
where we recall that c
M,
is the prequantum line bundle over T
∗
M (see Section 4.1). Observe
that the product of the natural ﬂat connections on ´
L,ι
and ι
∗
c
M,
deﬁnes a ﬂat connection
on the phase bundle whose holonomy is represented by the modZ
reduction of the real
cohomology class
λ
L,ι
+ πµ
L,ι
/2 ∈
ˇ
H
1
(L; R),
which we call the phase class of (L, ι). The phase bundle Φ
L,ι,
can be described explicitly
as the collection of all quintuples (p, t, (B, p
B
, φ), n, z) where (p, t, n, z) ∈ L T
Z C
and (B, p
B
, φ) ∈ M(L, ι, p), modulo the equivalence relation ∼ given by
(p, t, (B, p
B
, φ), n, z) ∼ (˜ p,
˜
t, (
˜
B, p
˜
B
,
˜
φ), ˜ n, ˜ z)
whenever p = ˜ p and
z e
−it/
e
iπ(n+ind
φ
(p))/2
= ˜ z e
−i
˜
t/
e
iπ(˜ n+ind
˜
φ
(p))/2
.
56
A Morse family (B, p
B
, φ) which generates an open subset L
φ
of L deﬁnes a nonvanishing
parallel section of Φ
L,ι,
over L
φ
by
s
φ,
(p) = [p, 0, (B, p
B
, φ), 0, e
−iφ(y)/
],
where λ
φ
(y) = p. A check of these deﬁnitions shows that whenever ι(p) = λ
φ
(y) = λ
˜
φ
(˜ y),
s
φ,
(p) e
iφ(y)/
e
−iπ ind
φ
(p)/2
= s
˜
φ,
(p) e
i
˜
φ(˜ y)/
e
−iπ ind
˜
φ
(p)/2
.
For each ∈ R
+
, we denote by Γ
par
(Φ
L,ι,
) the space of parallel sections of Φ
L,ι,
. If, for
a particular , the phase class of (L, ι) is integral, then Γ
par
(Φ
L,ι,
) is a complex vector
space isomorphic to C. Otherwise, Γ
par
(Φ
L,ι,
) consists of a single point (the zero section of
Φ
L,ι,
). The product
Γ
L,ι
=
¸
>0
Γ
par
(Φ
L,ι,
)
then has the structure of a Cmodule. An element s ∈ Γ
L,ι
is then a (generally discontinuous)
function which assigns to each > 0 an element s
in Γ
par
(Φ
L,ι,
), so that the map p →s
(p)
deﬁnes a parallel section of Φ
L,ι,
. The symbol space of (L, ι) is deﬁned as the complex
vector space
o
L,ι
def
= [Ω[
1/2
L ⊗
C
Γ
L,ι
.
The amplitude bundle /
φ
associated to a Morse family (B, p
B
, φ) over a smooth man
ifold M is deﬁned as the complex line bundle
/
φ
= [Λ[
1/2
B ⊗[Λ[
1/2
E
over B, where E again denotes the subbundle ker(p
B∗
) of TB. An amplitude is a section a of
/
φ
. We will say that a is properly supported provided that the restriction of p
B
: B →M
to Supp(a) is a proper map. The purpose of the space of amplitudes is to deﬁne a relation
between halfdensities on M and symbols on the subset L
φ
of L generated by φ. To describe
this relation, we begin by noting that from the exact sequence of vector bundles
0 →E →TB →p
∗
B
TM →0
over B, it follows that [Λ[
1/2
B is naturally isomorphic to [Λ[
1/2
p
∗
B
TM ⊗ [Λ[
1/2
E. This in
turn gives rise to the natural isomorphism
/
φ
· [Λ[
1/2
p
∗
B
TM ⊗[Λ[E.
The image of an amplitude a on B under this isomorphism can be written as p
∗
B
[dx[
1/2
⊗σ,
where σ is a family of 1densities on the ﬁbers of p
B
, i.e., σ
x
is a density on each nonempty
p
−1
B
(x). By ﬁberintegration we pass to a halfdensity on M:
I
(φ, a)(x) = (2π)
−n/2
e
−inπ/4
p
−1
B
(x)
e
iφ/
σ
x
[dx[
1/2
,
57
where n = dim(p
−1
B
(x)), setting I
(φ, a)(x) = 0 if p
−1
B
(x) = ∅. When a is properly supported,
we may diﬀerentiate under the integral to conclude that I
(φ, a) is a smooth halfdensity on
M.
To pass geometrically from a to a symbol on L
φ
, we ﬁrst recall from Section 4.3 that
the nondegeneracy of φ gives rise to the following exact sequence of vector bundles over the
ﬁbercritical set Σ
φ
0 →TΣ
φ
→T
Σ
φ
B
∇d
θ
φ
→ E
∗
[
Σ
→0.
Since [Λ[
−1/2
E is naturally isomorphic to [Λ[
1/2
E
∗
, this sequence induces an isomorphism
of the restriction of /
φ
to Σ
φ
with [Λ[
1/2
Σ
φ
. If the restriction of a to Σ
φ
corresponds to a
halfdensity a on Σ
φ
under this isomorphism, the associated symbol on L
φ
is deﬁned as
s
a
def
= g
∗
φ
a ⊗s
φ
.
(Here we recall from the discussion following Deﬁnition 4.17 that g
φ
is a diﬀeomorphism
from a neighborhood of p in L onto Σ
φ
deﬁned by the composition λ
−1
φ
◦ ι. Also, for each
, the section s
φ,
is the canonical element of Γ
par
(Φ
L
φ
,ι,
) deﬁned above). When (L, ι) is
projectable, the symbol s
a
and the halfdensity I
(φ, a) are linked by the following theorem.
Theorem 4.27 Suppose that two Morse families (B, p
B
, φ), (
˜
B, p
˜
B
,
˜
φ) generate the same
projectable lagrangian embedding (L, ι), and let a, ˜ a be amplitudes on B,
˜
B, respectively.
Then s
a
= s
˜a
on L if and only if
[I
(φ, a) −I
(
˜
φ, ˜ a)[ = O()
locally uniformly on V .
Proof. By Theorem 4.18, the Morse families (B, p
B
, φ), (
˜
B, p
˜
B
,
˜
φ) are stably equivalent, and
so there exists a diﬀeomorphism g : Σ
φ
→ Σ
˜
φ
deﬁned as the composition g = g
˜
φ
◦ g
−1
φ
. A
check of the deﬁnitions shows that the symbols s
a
and s
˜a
are equal precisely when
g
∗
a e
iφ/
e
−iπ ind
φ
/2
= ˜ a e
i
˜
φ/
e
−iπ ind
˜
φ/2
,
where a is the halfdensity on Σ
φ
induced by the amplitude a, and similarly for ˜ a. Since
˜
φ = φ ◦ g, up to a constant, Lemma B.3 implies that this occurs precisely when
[ det
ax
Hφ[
1/2
e
−iφ/
e
iπ ind
φ
/2
= [ det
˜ ax
H
˜
φ[
1/2
e
−i
˜
φ/
e
iπ ind
˜
φ
/2
. (∗)
Since (L, ι) is projectable, we can apply the principle of stationary phase to each ﬁber of
the projection p
B
: B →M to obtain
I
(φ, a)(x) =
e
iφ/
e
−iπ ind
φ
/2
[ det
ax
Hφ[
1/2
+ O()
and similarly for I
(
˜
φ, ˜ a)(x). The theorem follows by comparing these expressions with (∗)
above.
58
2
Morse families provide a general means for locally quantizing symbols on an immersed
lagrangian submanifold (L, ι) ⊂ T
∗
M. Suppose that (B, p
B
, φ) is a Morse family such
that the phase function φ generates an open subset L
φ
⊂ L, and consider a symbol s on
L supported in L
φ
. Then there exists a unique halfdensity a supported in L
φ
such that
s = s
φ
⊗a, and (g
φ
)
∗
a may be canonically identiﬁed with a section of the amplitude bundle
of B over the ﬁbercritical set Σ
φ
, which we can extend to an amplitude a on B, compactly
supported in ﬁbers. We then set
I
(L, ι, s) = I
(φ, a)(x).
From Theorem 4.27 we draw two conclusions about this tentative deﬁnition when L
φ
is
projectable. First, we note that if (
˜
B, p
˜
B
,
˜
φ) is a second Morse family which generates L
φ
,
and ˜ a is an amplitude on
˜
B obtained as above, then s
a
= s
a
= s, and so
[I
(φ, a) −I
(
˜
φ, ˜ a)[ = O().
Thus, I
(L, ι, s) is welldeﬁned, up to O() terms. Furthermore, we note that Theorem 4.27
asserts that this choice of I
(L, ι, s) coincides with pullback.
To quantize an arbitrary symbol s on L, we ﬁrst ﬁx a locally ﬁnite cover ¦L
j
¦ of L, such
that each L
j
is generated by a Morse family (B
j
, p
B
j
, φ
j
), and choose a partition of unity
¦h
j
¦ subordinate to ¦L
j
¦. We then set
I
(L, ι, s) =
¸
j
I
(L, ι, h
j
s).
Using the remarks above, it is easy to check that up to O() terms, this deﬁnition depends
only on the semiclassical state (L, ι, s).
Of course, although the technique above enables us to quantize all symbols on L in a
consistent way, the existence of nonzero symbols requires that the phase bundle Φ
L,ι
admit
nontrivial parallel sections, i.e. that the phase class of (L, ι) be integral. For this reason,
we introduce the following terminology.
Deﬁnition 4.28 An immersed lagrangian submanifold ι : L →T
∗
M is called quantizable
if, for some ∈ R
+
, its phase bundle Φ
L,ι
admits a global parallel section, or, equivalently,
if its phase class is integral. The set of for which this condition holds will be called
admissible for L.
This deﬁnition is known as the Maslov quantization condition. Note that it is a straight
forward generalization of the condition derived in the preceding section.
Example 4.29 Let N be a closed submanifold of a smooth manifold M, and let U be a
tubular neighborhood of N, i.e., U is the image of the normal bundle ν
N
⊂ T
N
M under an
embedding ψ : ν
N
→ M satisfying ψ = π on the zero section of ν
N
, where π : ν
N
→ N is
the natural projection. Consider the Morse family (r
∗
N
⊥
, p
N
, φ), where r = π ◦ ψ
−1
is a
59
retraction of U onto N, p
N
denotes the natural submersion r
∗
N
⊥
→M, and φ: r
∗
N
⊥
→R
is deﬁned by
φ(p) = 'p, ψ
−1
(p
N
(p))`.
Since ψ is an embedding, a computation shows that the ﬁber critical set of φ is given by
Σ
φ
= p
−1
N
(N) = N
⊥
, and the map λ
φ
: N
⊥
→T
∗
M equals the inclusion. Thus, the conormal
bundle of N is a lagrangian submanifold of T
∗
M which admits a global generating function,
and therefore both the Liouville and Maslov classes of N
⊥
are zero. In particular, this
implies that the conormal bundle of any submanifold of M satisﬁes the Maslov quantization
condition.
´
Quantum states as distributions
Unfortunately, the interpretation of I
(L, ι, s) at regular values of π
L
is not valid at caustics.
Indeed, this remark is suggested by the fact that I
(L, ι, s) is smooth, whereas we saw in
Section 2.2 that classical solutions to the transport equation are singular at caustic points.
The basic technical diﬃculty is that the principle of stationary phase (the basic underpinning
of Theorem 4.27) no longer applies to the integral
p
−1
B
(x)
e
iφ/
σ
x
when x is a caustic point, since the phase function φ has a degenerate critical point in the
ﬁber p
−1
B
(x).
Example 4.30 The ﬁber critical set of the phase function φ(x, θ) = θ
3
/3 + xθ consists of
the parabola x = −θ
2
, at whose points the ﬁberhessian assumes the form ∂
2
φ/∂θ
2
= 2x.
The origin is therefore a degenerate critical point for φ, and stationary phase cannot be used
to estimate the integral
R
e
iθ
3
/3
a(0, θ) dθ.
´
A more appropriate way to interpret the expression I
(L, ι, s) in the presence of caustic points
is as a family of distributional halfdensities on M deﬁned as follows. For each Morse family
(B, p
B
, φ) ∈ M(L, ι) and compactly supported amplitude a on B, we deﬁne a distributional
halfdensity on M by
'I
(φ, a), u` = (2π)
−b/2
e
−iπb/4
B
e
iφ/
a ⊗p
∗
B
u,
where b = dim(B) and u ∈ [Ω[
1/2
0
M. The family I
(L, ι, s) then consists of all distributional
halfdensities I
on M obtained by choosing a locally ﬁnite open cover ¦L
j
¦ of L and a
60
partition of unity ¦h
j
¦ subordinate to ¦L
j
¦. Then set
I
=
k
¸
j=1
I
(φ
j
, a
j
).
The family I
(L, ι, s) consists of those distributional halfdensities obtained in this way using
amplitudes a
j
such that s
a
j
= s over each L
φ
j
.
Although the class I
(L, ι, s) may appear very large, a link among its members can be
described as follows. We ﬁrst prove
Theorem 4.31 Suppose that two Morse families (B, p
B
, φ), (
˜
B, p
˜
B
,
˜
φ) generate the same
lagrangian embedding (L, ι), and let a, ˜ a be amplitudes on B,
˜
B, respectively. If ψ: M → R
is a smooth function whose diﬀerential intersects ι(L) at exactly one point ι(p) transversely,
then s
a
(p) = s
˜a
(p) if and only if
'I
(φ, a), e
−iψ/
u` −'I
(
˜
φ, ˜ a), e
−iψ/
u`
= O().
Proof. Set R = φ −ψ ◦ p
B
. Since ψ ◦ p
B
is constant on the ﬁbers of p
B
, the function R is a
phase function having the same ﬁber critical set as φ.
As in the proof of Theorem 4.27, equality of the symbols s
a
and s
˜a
occurs precisely when
the diﬀeomorphism g : Σ
φ
→Σ
˜
φ
satisﬁes
g
∗
a
y
e
iφ(y)/
e
−iπ ind
φ
(p)/2
= ˜ a
˜ y
e
i
˜
φ(˜ y)/
e
−iπ ind
˜
φ(p)/2
,
where a is the halfdensity on Σ
φ
induced by the amplitude a, and similarly for ˜ a. Since
˜
R = R ◦ g, up to a constant, Lemma B.3 shows that this condition is equivalent to
[ det
a
R
tt
(y)[
1/2
e
−iφ(y)/
e
iπ ind
φ
(p)/2
= [ det
˜ a
˜
R
tt
(˜ y)[
1/2
e
−i
˜
φ(˜ y)/
e
iπ ind
˜
φ
(p)/2
.
Since g is ﬁberpreserving, we have R(y) −
˜
R(˜ y) = φ(y) −
˜
φ(˜ y). Moreover, it is easy
to check that stable equivalence of the Morse families (B, p
B
, φ), (
˜
B, p
˜
B
,
˜
φ) implies that
ind
φ
(p) −ind
˜
φ
(˜ y) = ind R
tt
(y) −ind
˜
R
tt
(˜ y), and so the preceding equation gives
e
iR(y)/
e
−iπ ind R
(y)/2
[ det
a
R
tt
(y)[
1/2
=
e
i
˜
R(˜ y)/
e
−iπ ind
˜
R
(˜ y)/2
[ det
˜ a
˜
R
tt
(˜ y)[
1/2
. (∗∗)
The critical point y of R is nondegenerate precisely when the intersection of ι(L) with
im(dψ) is transverse. In this case, an application of the principle of stationary phase gives
'I
(φ, a), e
−iψ/
u` =
e
iR(y)/
e
−iπ ind R
(y)/2
[ det
a
R
tt
(y)[
1/2
+ O()
and similarly for 'I
(
˜
φ, ˜ a), e
−iψ/
u`. Comparing this expression with (∗∗) above completes
the proof.
2
61
Now consider a semiclassical state (L, ι, s). As before, we let ¦L
j
¦ be a locally ﬁnite
cover of L such that each L
j
is generated by a Morse family (B
j
, p
B
j
, φ
j
) over M, and choose
a partition of unity ¦h
j
¦ subordinate to ¦L
j
¦. We then deﬁne
'I
(L, ι, s), e
−iψ/
u`
def
=
¸
j
'I
(φ
j
, a
j
), e
−iψ/
u`,
where a
j
is the amplitude on B
j
obtained in the usual way from the symbol h
j
s on L
j
. When
the image of dψ is transverse to L, then, up to O(), the principal part of 'I
(L, ι, s), e
−iψ/
`
depends only on the principal part of s. For a thorough exposition of this topic, we refer to
[21, 28, 31].
Example 4.32 Suppose that (B, p
B
, φ) is a Morse family over a manifold M, and suppose
that S : V → R is a smooth function. By setting
˜
S = S ◦ p
B
, we obtain a new Morse
family (B, p
B
, φ −
˜
S) for which Σ
φ
= Σ
φ−
˜
S
and λ
φ
= f
dS
◦ λ
φ−
˜
S
, where f
dS
is the ﬁberwise
translation map deﬁned by dS.
Note in particular that if dS(p
B
(x)) = λ
φ
(x) for some x ∈ Σ
φ
, then x is a critical point
of φ −
˜
S. This critical point is nondegenerate provided that λ
φ−
˜
S
is transverse to the zero
section of T
∗
M near x, or, equivalently, if λ
φ
is transverse to the image of dS near x.
´
By allowing I
(φ, a) to be a distribution, we can sometimes deﬁne it even when a does
not have compact support.
Equivalent semiclassical states
A further important modiﬁcation of our quantization picture is based on the conceptual
distinction between an “immersion” and an “immersed submanifold”. In terms of our dis
cussion, this means the following. If ι : L → T
∗
M and ι
t
: L
t
→ T
∗
M are lagrangian
immersions, we will say that (L, ι) and (L
t
, ι
t
) are equivalent provided that there exists a
diﬀeomorphism f : L → L
t
such that ι = ι
t
◦ f. In this way, any lagrangian immersion
ι : L →T
∗
M deﬁnes an equivalence class of lagrangian immersions in T
∗
M which, for nota
tional simplicity we will denote by L. We refer to the equivalence class L as an immersed
lagrangian submanifold of T
∗
M.
A check of the deﬁnitions shows that the main objects in our quantization scheme behave
nicely with respect to this notion of equivalence. On the classical level, a diﬀeomorphism
f : L →L
t
as above induces an isomorphism of symbol spaces
o
L,ι
f∗
→o
L
,ι
.
Thus, a symbol on the immersed lagrangian submanifold L is welldeﬁned as an equivalence
class of symbols on the members of L. Moreover, if E is a regular value of some hamiltonian
function H on T
∗
M, then the HamiltonJacobi equation
H ◦ ι = E
62
deﬁnes a condition on the class L which we denote by H(L) = E. In this case, the vector
ﬁelds X
H,ι
and X
H,ι
induced on (L, ι), (L
t
, ι
t
) by the hamiltonian vector ﬁeld of H (see
Example 3.13) satisfy X
H,ι
= f
∗
X
H,ι
, and therefore the homogeneous transport equation
L
X
H,ι
s = 0
is a welldeﬁned condition on (L, s) which we denote by L
X
H
s = 0.
From these remarks, we are led to view the equivalence class (L, s) as a semiclassical
state in T
∗
M. The state is stationary with respect to a classical hamiltonian H on T
∗
M
provided that (L, s) satisﬁes the associated HamiltonJacobi and homogeneous transport
equations, as described above. On the quantum level, it is easy to check that for any
members (L, ι, s), (L
t
, ι
t
, s
t
) of the equivalence class (L, s), we have
I
(L, ι, s) = I
(L
t
, ι
t
, s
t
),
and thus we may deﬁne the quantization of (L, s) as the (unique) distributional halfdensity
I
(L, s) on M obtained by quantizing any member of (L, s). If (L, s) is a stationary semi
classical state, then I
(L, s) is a semiclassical approximate solution to the timeindependent
Schr¨odinger equation on M.
Our list of classical and quantum correspondences now assumes the form:
Object Classical version Quantum version
basic space T
∗
M H
M
state (L, s) as above distributional halfdensity
on M
timeevolution Hamilton’s equations Schr¨odinger equation
generator of evolution function H on T
∗
M operator
ˆ
H on H
M
stationary state state (L, s) satisfying eigenvector of
ˆ
H
H(L) = E and L
X
H
s = 0
With these correspondences between classical and quantum mechanics in mind, we are fur
ther led to deﬁne a semiclassical state in an arbitrary symplectic manifold (P, ω) to be an
immersed lagrangian submanifold equipped with a halfdensity and possibly some “phase
object” corresponding to a parallel section of the phase bundle in the cotangent bundle case.
Note, however, that it is unclear what the quantum states are which are approximated by
these geometric objects, since there is no underlying conﬁguration space on which the states
can live. Extending some notion of quantum state to arbitrary symplectic manifolds is one
of the central goals of geometric quantization.
63
5 The Symplectic Category
There are many symplectic manifolds which are not cotangent bundles. For instance, an
important process known as symplectic reduction generates many examples of such manifolds,
starting with cotangent bundles. We begin this section with a discussion of reduction and
then turn to the classical and quantum viewpoints in the context of general symplectic
manifolds, concluding with Dirac’s formulation of the quantization problem.
5.1 Symplectic reduction
The technique of symplectic reduction geometrizes the process in mechanics in which ﬁrst
integrals of the hamiltonian are used to eliminate variables in Hamilton’s equations.
Linear symplectic reduction
Degenerate skewsymmetric bilinear forms yield symplectic vector spaces in the following
way.
Lemma 5.1 A skewsymmetric bilinear form ω on a vector space Y induces a symplectic
structure on Y/Y
⊥
.
Proof. First note that ω gives rise to a skewsymmetric bilinear form ω
t
on Y/Y
⊥
by the
equation
ω
t
([x], [y]) = ω(x, y).
To prove the nondegeneracy of ω
t
, we use the following commutative diagram, where π is
the projection.
Y
∗
π
∗
←−−− (Y/Y
⊥
)
∗
˜ ω
¸
¸
˜ ω
Y
π
−−−→ Y/Y
⊥
If ˜ ω
t
([x]) = 0, then ˜ ω(x) = 0, and so x ∈ Y
⊥
and hence [x] = 0.
2
The symplectic quotient space Y/Y
⊥
described in this lemma is known as the (linear) re
duced space associated to Y . A special case of linear reduction arises when Y is a coisotropic
subspace of a symplectic vector space and ω equals the restriction of the symplectic form to
Y . Lagrangian subspaces behave remarkably well with respect to this reduction.
Lemma 5.2 Let V be a symplectic vector space and L, C ⊂ V a lagrangian and a coisotropic
subspace respectively. Then
L
C
= L ∩ C + C
⊥
is a lagrangian subspace of V contained in C, and
L
C
= (L ∩ C)/(L ∩ C
⊥
)
is a lagrangian subspace of C/C
⊥
.
64
Proof. Note that
(L
C
)
⊥
= (L + C
⊥
) ∩ C.
Since by assumption C
⊥
⊂ C, the selforthogonality of L
C
follows from the simple fact that
if E, F, G are any three subspaces of V , then (E+F) ∩G = E∩G+F if and only if F ⊂ G.
Next, we observe that since L
C
= L
C
/C
⊥
, the second assertion follows from the equality
L
⊥
C
= (L
C
)
⊥
/C
⊥
.
2
Example 5.3 If (V, ω), (V
t
, ω
t
) are symplectic vector spaces and L is a lagrangian subspace
of V , then then C = V
t
⊕L is a coisotropic subspace of the direct sum V
t
⊕V (see Exam
ple 3.4), and C
⊥
= 0 ⊕L. Consequently, the linear reduced space C/C
⊥
equals (V
t
, ω
t
).
Now suppose that V and V
t
are of equal dimension and T : V →V
t
is a linear symplectic
map. If Γ
T
⊂ V
t
⊕V is the lagrangian subspace deﬁned by the graph of T, then the reduced
lagrangian subspace (Γ
T
)
C
of C/C
⊥
equals T(L) ⊂ V
t
.
´
Observe that if C is a coisotropic subspace of (V, ω), then C
⊥
⊕V/C also carries a natural
symplectic structure induced by ω. This gives rise to the following decomposition of V .
Lemma 5.4 If (V, ω) is a symplectic vector space with a coisotropic subspace C, then there
exists a linear symplectomorphism
V →C
⊥
⊕ V/C ⊕C/C
⊥
.
Proof. Let J be a ωcompatible complex structure on V , and set A = JC
⊥
and B =
C∩JC. Then A, B are orthogonal to C, C
⊥
with respect to the innerproduct g
J
on V . The
projections V →V/C and C →C/C
⊥
restricted to A, B give rise to an isomorphism
V = C
⊥
+ A + B →C
⊥
⊕ V/C ⊕C/C
⊥
.
2
Recall that if L, L
t
are lagrangian subspaces of a symplectic vector space V and W ⊂ V
is a lagrangian subspace transverse to both L and L
t
, then there exists a natural linear
symplectomorphism from V to L⊕L
∗
which sends W onto the subspace 0 ⊕L
∗
and L
t
onto
the graph of some selfadjoint linear map T : L → L
∗
. Denoting by Q
T
the quadratic form
Q
T
on L induced by T, we deﬁne
ind(L, L
t
; W) = index Q
T
sgn(L, L
t
; W) = signature Q
T
These quantities will be useful in our study of the Maslov bundle under reduction, and we
collect some useful facts about them in the following examples.
65
Example 5.5 To begin, we leave to the reader the job of checking the following elementary
identities for the case when the lagrangian subspaces L, L
t
are themselves transverse:
sgn(L, L
t
; W) = −sgn(L
t
, L; W) = −sgn(L, W; L
t
).
´
Example 5.6 Suppose that V is a ﬁnitedimensional vector space with a subspace E ⊂ V
and its algebraic orthogonal E
⊥
⊂ V
∗
. Then C = V ⊕E
⊥
is a coisotropic subspace of V ⊕V
∗
with its usual symplectic structure, and C
⊥
= E ⊕0. The reduced symplectic vector space
is then
C/C
⊥
= (V/E) ⊕E
⊥
.
The composition of a selfadjoint linear map A : V
∗
→ V with the projection V → V/E
restricts to a selfadjoint linear map A
E
: E
⊥
→ V/E (note that V/E identiﬁes canonically
with (E
⊥
)
∗
). Evidently, the lagrangian subspace W of V ⊕V
∗
given by the graph of A passes
under reduction by C to the graph W
C
of A
E
. Now if we denote by L, L
t
the lagrangian
subspaces V ⊕ 0 and 0 ⊕ V
∗
, respectively, then L
C
= (V/E) ⊕ 0 and L
t
C
= 0 ⊕ E
⊥
. If Q
∗
denotes the quadratic form on V
∗
deﬁned by A, we therefore have
sgn(Q
∗
) = sgn(L
t
, W; L) and sgn(Q
∗
[
E
⊥) = sgn(L
t
C
, W
C
; L
C
).
From [21, p.130] we recall that if Q
∗
is nondegenerate and Q is the quadratic form it induces
on V , then
sgn(Q) = sgn(Q[
E
) + sgn(Q
∗
[
E
⊥).
Combined with the preceding equations, this formula gives
sgn(L
t
, W; L) = sgn(Q[
E
) + sgn(L
t
C
, W
C
; L
C
).
´
Presymplectic structures and reduction
By deﬁnition, a symplectic form is closed and nondegenerate. In some sense, the next best
structure a manifold M may possess along these lines is a closed twoform ω of constant
rank, i.e., with the dimension of the orthogonal (T
x
M)
⊥
the same for all x ∈ M. In this
case, ω is called a presymplectic structure on M with characteristic subbundle (TM)
⊥
.
Theorem 5.7 The characteristic subbundle of a presymplectic manifold (M, ω) is integrable.
Proof. Leaving for the reader the veriﬁcation that (TM)
⊥
is actually a subbundle of TM,
we recall that Lie brackets and inner products are related by the formula
[X, Y ] ω = L
X
(Y ω) −Y L
X
ω.
If the vector ﬁeld Y belongs to (TM)
⊥
, then the ﬁrst term on the right hand side of this
equation vanishes identically. If X belongs to (TM)
⊥
, then Cartan’s formula combined with
the assumption that ω is closed implies that ω is invariant under the ﬂow of X. This means
that L
X
ω = 0, so the second term vanishes as well, and [X, Y ] lies in (TM)
⊥
.
66
2
The foliation ´
⊥
deﬁned by the characteristic subbundle of M is known as the characteris
tic foliation of M. If the quotient space M/´
⊥
is a smooth manifold, then we say that M
is reducible. A pointwise application of Lemma 5.2, together with the fact that L
X
ω = 0
for characteristic X, shows that the presymplectic structure ω induces a smooth, nonde
generate 2form ω
/
on M/´
⊥
. Since dω = 0 by hypothesis, and since the quotient map
M → M/´
⊥
is a submersion, the form ω
/
is necessarily closed and therefore symplectic.
The symplectic manifold (M/´
⊥
, ω
/
) is called the reduced manifold of M.
For the most part, we will be interested in presymplectic manifolds which arise as
coisotropic submanifolds of some symplectic manifold (P, ω). Recall that a submanifold
C ⊂ P is called coisotropic if, for each p ∈ C, the tangent space T
p
C contains its symplectic
orthogonal
(T
p
C)
⊥
= ¦v ∈ T
p
P : ω(v, w) = 0 for all w ∈ T
p
C¦.
In this case, we can view C as an abstract manifold and note that if ω
t
is the pullback
of the symplectic form ω by the natural inclusion C → P, then ker(ω
t
) = (TC)
⊥
. Since
dim(TC)
⊥
= dim(P) −dim(C) is constant, the form ω
t
deﬁnes a presymplectic structure on
P.
Example 5.8 As noted in Example 3.13, the hamiltonian ﬂow associated to a function
H : R
2n
→ R generates the characteristic foliation of any regular energy surface H
−1
(E).
The corresponding reduced symplectic manifold (when it exists) will be the symplectic model
for the space of quantum states of energy E.
If P, Q are symplectic manifolds, and L is a lagrangian submanifold of Q, then P L is a
coisotropic submanifold of P Q whose characteristic foliation consists of leaves of the form
¦p¦ L for p ∈ P. The product P L is therefore reducible, and the reduction projection
coincides with the usual cartesian projection P L →P.
´
Our goal for the remainder of this section is to describe two operations on immersed
lagrangian submanifolds deﬁned by a reducible coisotropic submanifold; in eﬀect, these op
erations will be nonlinear analogs of the mappings L → L
C
and L → L
C
described in the
linear setting in Lemma 5.2. To begin, we recall that the ﬁber product of two maps
ι : N →V and : M →V is deﬁned as the subset
N
V
M = (ι )
−1
∆
V
of N M, where ι : N M → V V is the product map and ∆
V
⊂ V V is the
diagonal. This gives rise to the following commutative diagram
N
V
M
r
M
−−−→ M
r
N
N
ι
−−−→ V
where r
M
, r
N
denote the restrictions of the cartesian projections NM →M, N to the ﬁber
product N
V
M.
67
Deﬁnition 5.9 If N, M, V are smooth manifolds, then two smooth maps ι : N → V and
: M → V are said to intersect cleanly provided that their ﬁber product N
V
M is a
submanifold of N M and
ι
∗
TN ∩
∗
TM = ( ◦ r
M
)
∗
T(N
V
M)
as subbundles of ( ◦ r
M
)
∗
TV .
For brevity, we will say that a map ι : N → V intersects a submanifold W ⊂ V cleanly if ι
and the inclusion map W →V intersect cleanly.
Example 5.10 1. If the product ι of two smooth maps ι : N → V and : M → V
is transverse to the diagonal ∆
V
, then ι and intersect cleanly. A particular case of this
situation occurs when is a submersion and ι is any smooth map.
2. Suppose that ι : N →V and : M →V are any smooth maps whose ﬁber product N
V
M
is a smooth submanifold of NM. Then for any tangent vector (v, w) of N
V
M, the vector
(ι )
∗
(v, w) is tangent to the diagonal in V V , and so ι
∗
v =
∗
w. Since r
M∗
(v, w) = w,
it follows that r
M∗
(v, w) = 0 only if w = 0, in which case ι
∗
v =
∗
w = 0. Thus, if ι is an
immersion, we can conclude that r
M
: N
V
M →M is an immersion as well.
In particular, if ι and are cleanly intersecting immersions, then the maps r
N
and r
M
are themselves immersions.
3. A basic example of two maps which do not intersect cleanly is provided by two embedded
curves in the plane which intersect tangentially at a single point.
´
The usefulness of clean intersection in symplectic geometry lies in its compatibility with
symplectic orthogonalization, as in the following lemma.
Lemma 5.11 Let P be a symplectic manifold. If a lagrangian immersion ι : L → P inter
sects a coisotropic submanifold C ⊂ P cleanly, then the map (p
C
◦ r
C
): L
P
C →C/(
⊥
has
constant rank.
Proof. Let F, G denote the lagrangian and coisotropic subbundles r
∗
L
TL and r
∗
C
TC of the
symplectic vector bundle r
∗
C
TP over the ﬁber product L
P
C. The kernel of (p
C
◦r
C
)
∗
then
equals
G
⊥
∩ r
C∗
T(L
P
C).
The clean intersection hypothesis implies that r
C∗
T(L
P
C) = F ∩ G, and so the basic
properties of symplectic orthogonalization applied ﬁberwise to these bundles give
ker(p
C
◦ r
C
)
∗
= G
⊥
∩ G∩ F
= G
⊥
∩ F
= (G+ F)
⊥
.
68
Since F ∩ G has constant rank, the internal sum F + G and its orthogonal (F + G)
⊥
also
have constant rank, indicating that the dimension of ker(p
C
◦r
C
)
∗
is independent of the base
point in L
P
C. In fact, we have dim(ker(p
C
◦ r
C
)
∗
) = dim(L) + dim(L
P
C) −dim(C).
2
If the quotient of L
P
C by the ﬁbers of the map p
C
◦ r
C
is a smooth, Hausdorﬀ manifold,
then we obtain an induced immersion ι
C
from the reduced space L
C
of L
P
C into C/(
⊥
.
In this case, we deﬁne L
C
as the ﬁber product of ι
C
and the quotient map p
C
: C →C/(
⊥
.
By
C
: L
C
→ C/(
⊥
we denote the restriction of the cartesian projection L
C
C → C to
L
C
.
Recall from Chapter 4 that an immersed submanifold N of a manifold V is an equivalence
class N of immersions, where ι : N →V is considered equivalent to ι
t
: N
t
→V provided that
there exists a diﬀeomorphism f : N →N
t
which intertwines ι and ι
t
, i.e. ι = ι
t
◦f. Evidently,
if a member ι : N →V of the equivalence class N intersects a smooth map : M →V cleanly,
then the same is true of any other member of N. In this case, we will say that the immersed
submanifold N and the map : M →V intersect cleanly.
With these remarks in mind, we will say that a coisotropic submanifold C and an im
mersed lagrangian submanifold L in a symplectic manifold P form a reducible pair (C, L)
provided that C is reducible, and L, C intersect cleanly. We leave it to the reader to check
that any reducible pair (C, L) in P induces immersed submanifolds L
C
and L
C
of C/(
⊥
and
P, respectively.
Theorem 5.12 If (C, L) is a reducible pair in a symplectic manifold P, then L
C
and L
C
are immersed lagrangian submanifolds of C/(
⊥
and P, respectively.
Proof. Let ι : L → P be a ﬁxed element of L. Since L
C
is the quotient of L
P
C by the
ﬁbers of p
C
◦ r
C
, the induced map ι
C
: L
C
→ C/(
⊥
is a smooth immersion. A pointwise
application of Lemma 5.2 shows that this immersion is lagrangian.
Since ι
C
: L
C
→ C/(
⊥
is an immersion, Example 5.10(2) implies that the map
C
:
L
C
→C is a smooth immersion, and a pointwise application of Lemma 5.2 implies that this
immersion is lagrangian.
Finally, since the immersed submanifolds L
C
and L
C
are welldeﬁned by the reducible
pair (C, L), it follows that these are immersed lagrangian submanifolds.
2
Example 5.13 If (M, g) is a riemannian manifold, all of whose geodesics are closed and
have the same length, then for each E > 0, the constant energy hypersurface C = k
−1
M
(E)
(see Section 3.2) is reducible, and the reduced manifold CM = C/(
⊥
is the space of oriented
geodesics on M. The tangent space to CM at a point p identiﬁes naturally with the space
of Jacobi ﬁelds normal to the geodesic represented by p. Moreover, it is easy to check that
for any pair J
1
, J
2
of normal Jacobi ﬁelds along p, the symplectic form on CM is given in
terms of the metric by
ω(J
1
, J
2
) = g(J
1
, J
t
2
) −g(J
t
1
, J
2
).
69
For further details, see [12].
The conormal bundle N
⊥
to a smooth submanifold N ⊂ M is a lagrangian submanifold
of T
∗
M which intersects C cleanly. More precisely, N
⊥
∩ C is a sphere bundle over N
transverse to the characteristic foliation of C. From Theorem 5.12 it therefore follows that
the space of geodesics in M normal to N at some point comprises an immersed lagrangian
submanifold of CM.
To give a concrete case of this example, we consider the sphere S
n
equipped with its stan
dard metric. Each oriented geodesic in S
n
identiﬁes naturally with an oriented 2dimensional
subspace of R
n+1
, and therefore CS
n
is represented by the grassmannian Gr
+
(2, n + 1).
´
Example 5.14 Note that the standard symplectic structure on R
2n+2
is equivalent to the
imaginary part of the standard hermitian metric on C
n+1
. The unit sphere S
2n+1
is a
coisotropic submanifold of C
n+1
, the reduction of which is complex projective space CP
n
.
Since the symplectic form on C
n+1
is invariant under unitary transformations, the resulting
symplectic form on CP
n
is invariant under transformations induced by unitary transforma
tions.
The maximal real subspace R
n+1
⊂ C
n+1
is a lagrangian submanifold of C
n+1
whose inter
section with S
2n+1
is clean and coincides with the real unit sphere S
n
. From Theorem 5.12,
it follows that the real projective space RP
n
is a lagrangian submanifold of CP
n
.
´
Conormal submanifolds
A particular example of symplectic reduction which we will use in the next chapter begins
with the following setup. Suppose that M is a smooth manifold, and let N ⊂ M be a smooth
submanifold equipped with a tangent distribution T, i.e., a subbundle of TN ⊂ T
N
M. By
C
N
we denote the annihilator of T, that is
C
N
= ¦(x, p) ∈ T
∗
M : x ∈ N, T
x
⊂ ker(p)¦.
By deﬁnition, C
N
is a subbundle of T
∗
N
M, but we wish to consider various properties of C
N
when it is viewed as a smooth submanifold of the symplectic manifold (T
∗
M, ω
M
). We begin
with the following lemma.
Lemma 5.15 In the notation above, the annihilator C
N
is a coisotropic submanifold of T
∗
M
if and only if the distribution T is integrable.
Proof. If T is integrable, then we can consider its leaves as submanifolds of M via the
inclusion N →M. The conormal bundles of the leaves deﬁne a foliation of C
N
by lagrangian
submanifolds; by Lemma 3.6, this implies that C
N
is coisotropic.
Now let Z denote the intersection of C
N
with the zero section of T
∗
M. Then the natural
projection T
∗
M →M maps Z diﬀeomorphically onto N, and moreover, T
Z
C
N
= TN ⊕T
⊥
,
70
where T
⊥
is the subbundle of V
Z
M which annihilates T. From the natural lagrangian
splitting of T(T
∗
M) along the zero section of T
∗
M, it follows that
(T
Z
C
N
)
⊥
= T ⊕(TN)
⊥
,
and thus the intersections of Z with the isotropic leaves of C
N
are integral manifolds of the
distribution T. This completes the proof.
2
Lemma 5.16 If the leaf space N/T is a smooth, Hausdorﬀ manifold, then the reduced space
C
N
/(
⊥
is canonically symplectomorphic to T
∗
(N/T).
Proof. Since, by deﬁnition, each (x, p) ∈ C
N
contains the subspace T
x
of T
x
N in its kernel,
it follows that (x, p) deﬁnes an element of T
∗
(N/T). This gives a welldeﬁned surjective
submersion f : C
N
→T
∗
(N/T), and it is easy to check that the level sets of this map are the
ﬁbers of the projection C
N
→C
N
/(
⊥
. Thus we get a diﬀeomorphism C
N
/(
⊥
→T
∗
(N/T).
To prove that the preceding diﬀeomorphism is symplectic, it suﬃces to note that by the
commutative diagram
C
N
f
−−−→ T
∗
(N/T)
N −−−→ N/T
the Liouville forms of T
∗
M and T
∗
(N/T) are related by α
M
[
C
N
= f
∗
α
N/T
.
2
Example 5.17 Suppose that T is a foliation of a manifold N. By Lemma 5.15, it follows
that the annihilator C
N
⊂ T
∗
N of T is a coisotropic submanifold which, by deﬁnition,
contains the conormal bundle F
⊥
to any leaf F of T. Since F
⊥
is a lagrangian submanifold
of T
∗
N, Lemma 3.6 implies that F
⊥
is foliated by isotropic leaves of C
N
. Moreover, the
proof of Lemma 5.16 shows that each such leaf projects diﬀeomorphically onto F, and so the
bundle F
⊥
is equipped in this way with a ﬂat Ehresmann connection (see Ehresmann [22])
known as the Bott connection associated to the foliation T.
´
Example 5.18 As a special case of Example 5.17, suppose that (B, p
B
, φ) is a Morse family
over a manifold M, and let C
B
⊂ T
∗
B be the conormal bundle to the ﬁbers of the submersion
p
B
: B →V . If L ⊂ T
∗
B is the lagrangian submanifold given by the image of the diﬀerential
dφ, then the nondegeneracy condition on φ implies that L intersects C
B
transversally along a
submanifold I which maps diﬀeomorphically onto the ﬁber critical set Σ
φ
under the natural
projection π: T
∗
B →B. The lagrangian immersion λ
φ
: Σ
φ
→T
∗
M equals the composition
Σ
φ
π
−1
→ L ∩ C
B
p
C
B
→ C
V
/(
⊥
.
In other words, Theorem 4.15 can be reformulated as an application of Theorem 5.12 when
the lagrangian submanifold im(dφ) is transverse to E, i.e., when φ is nondegenerate.
71
´
We remark that since C
N
is a bundle over N, the intersection TC
N
∩ V M is a (constant
rank) subbundle of TC
N
. A check of the deﬁnitions shows that this subbundle is mapped
onto V (N/T) under the quotient map C
N
→T
∗
(N/T) described above.
Example 5.19 An important special case of Examples 5.6 and 5.18 arises when (B, V, p
V
, φ)
is a Morse family which generates a lagrangian submanifold in L ⊂ T
∗
M and x ∈ B is a
(nondegenerate) critical point of φ, so that p = λ
φ
(x) lies in the zero section Z
M
of T
∗
M.
Denoting by x
t
the point in the zero section of T
∗
B lying over x, we recall that T
x
(T
∗
B)
admits the natural lagrangian splitting T
x
B ⊕ T
∗
x
B. Let let E ⊂ T
x
B be the kernel of the
linearized projection p
B∗
: TB →TM.
Now suppose that Q is the nondegenerate quadratic form on T
x
B deﬁned by the hessian
of φ. By deﬁnition, the restriction of Q to E equals the form induced by the ﬁberhessian
Hφ
x
of φ at x, and so
ind(Q[
E
) = ind Hφ
x
.
On the other hand, it is easy to check that the tangent space to the conormal submanifold
C
V
deﬁned in Example 5.18 at x equals the coisotropic subspace C = T
x
B ⊕ E
⊥
, so that
the last formula in Example 5.6 implies that the Morse index ι
M
(x, φ) of the critical point
x satisﬁes
ι
M
(x, φ) = ind Hφ
x
+ ind(V
p
M, T
p
L; T
p
Z
M
).
´
Reduction and the Maslov bundle
Using linear reduction we can give a useful alternative description of the Maslov bundle of
an immersed lagrangian submanifold L in T
∗
M. Given two lagrangian subspaces L, L
t
of a
symplectic vector space V , we denote by L
L,L
the subset of the lagrangian grassmannian
L(V ) comprised of those lagrangian subspaces which are transverse to both L and L
t
. For
W, W
t
∈ L
L,L
, the cross index of the quadruple (L, L
t
; W, W
t
) is deﬁned as the integer
σ(L, L
t
; W, W
t
) = ind(L, L
t
; W) −ind(L, L
t
; W
t
).
Immediately from this deﬁnition and Example 5.5, we obtain the following “cocycle” property
of the cross index.
Lemma 5.20 If L, L
t
are lagrangian subspaces of V and W, W
t
, W
tt
∈ L
L,L
, then
σ(L, L
t
; W, W
t
) + σ(L, L
t
; W
t
, W
tt
) + σ(L, L
t
; W
tt
, W) = 0.
2
72
We denote by T
L,L
(V ) the space of functions f : L
L,L
→Z such that
f(W) −f(W
t
) = σ(L, L
t
; W, W
t
)
for all W, W
t
∈ L
L,L
. Since any such function is determined up to an additive constant by
its value at a single point of L
L,L
, it follows that Z acts simply and transitively by addition
on T
L,L
(V ).
Now we may assign a principal Z bundle to a triple (E, λ, λ
t
), where E is a symplectic
vector bundle and λ, λ
t
are lagrangian subbundles. Associated to E is its lagrangian grass
mannian bundle L(E) whose ﬁber over x ∈ M is simply L(E
x
). The lagrangian subbundles
λ, λ
t
then correspond to smooth sections of L(E), and we denote by M
λ,λ
(E) the principal
Z bundle whose ﬁber over x ∈ M equals T
λx,λ
x
(E
x
).
A special case of this setup occurs when L is an immersed lagrangian submanifold of
T
∗
M so that E = ι
∗
T(T
∗
M) is a symplectic vector bundle over L. Natural lagrangian
subbundles of E are then given by λ = ι
∗
TL and λ
t
= ι
∗
V M.
Theorem 5.21 In the notation above, the Maslov bundle M
L,ι
is canonically isomorphic to
M
λ,λ
(E).
Proof. Let (B, p
B
, φ) be a Morse family over a manifold M, ﬁx a point x in the ﬁber
critical set Σ
φ
, and set p = λ
φ
(x). From Example 4.32, we recall that if S : M → R is a
smooth function such that the image of dS intersects λ
φ
(Σ
φ
) transversely at p, then x is a
nondegenerate critical point of φ −
˜
S, where
˜
S = S ◦ p
B
. From Example 5.19 it follows that
the Morse index ι
M
(x, φ −
˜
S) at x is given by
ι
M
(x, φ −
˜
S) = ind Hφ
x
+ ind(V
p
M, T
p
L
φ−S
; T
p
Z
M
),
where p
t
= p − dS(p
V
(x)) and L
φ−
˜
S
is the lagrangian submanifold generated by φ −
˜
S.
Fiberwise translation by dS does not change the index on the right, and moreover it maps
T
p
L
φ−S
to T
p
L
φ
, the subspace V
p
M to the vertical subspace V
p
M at p, and T
p
Z
M
to the
tangent space Y of the image of dS at p. Thus, the preceding equation and Example 5.5
imply
ι
M
(x, S) = ind Hφ
x
+ ind(T
p
L
φ
, V
p
M; Y ).
Now if Y is any lagrangian subspace of T
p
(T
∗
M) transverse to V
p
M, then there exists a
function S: V →R such that Y = imdS(p
V
(x)). A function f
φ
: L
TpL
φ
,VpM
→Z is obtained
by setting
f
φ
(Y ) = ι
M
(x, φ −S
Y
).
From the discussion above, it follows that this function satisﬁes
f
φ
(Y ) −f
φ
(Y
t
) = σ(T
p
L
φ
, V
p
M; Y, Y
t
).
Using these remarks, we deﬁne a map LM(L, ι) Z →M
λ,λ
(E) by sending the triple
(p, (B, p
B
, φ), n) to the function (p, f
φ
). From the preceding paragraph and the deﬁnition of
M
λ,λ
(E), it follows that this map passes to an isomorphism of principal Z bundles M
L,ι
→
M
λ,λ
(E).
73
2
We refer to [3] and [31] for a proof that the Maslov bundle described above is equivalent
to the Maslov bundle constructed using index functions in Chapter 4. For the remainder
of this section, our goal is to establish the following property of the bundle M
λ,λ
(E) under
symplectic reduction. The proofs are rather technical and can be passed over in a ﬁrst
reading.
Theorem 5.22 Suppose that E is a symplectic vector bundle over M with lagrangian sub
bundles λ, λ
t
and a coisotropic subbundle η. If λ ∩ η and λ
t
∩ η have constant rank, then
M
λ,λ
(E) is canonically isomorphic to M
λη,λ
η
(η/η
⊥
).
The starting point of the proof is the following special case of the theorem.
Lemma 5.23 Suppose that E is a symplectic vector bundle with lagrangian subbundles λ, λ
t
and a coisotropic subbundle η. If λ, λ
t
⊂ η, then M
λ,λ
(E) is canonically isomorphic to
M
λη,λ
η
(η/η
⊥
).
Proof. First consider a symplectic vector space V with lagrangian subspaces L, L
t
and a
coisotropic subspace C. If L, L
t
⊂ C, then there exists a natural linear symplectomorphism
of reduced spaces
(L
C
+ L
t
C
)/(L
C
+ L
t
C
)
⊥
→(L + L
t
)/(L + L
t
)
⊥
which maps L
C
/(L
C
+ L
t
C
)
⊥
onto L/(L + L
t
)
⊥
and L
t
C
/(L
C
+ L
t
C
)
⊥
onto L
t
/(L + L
t
)
⊥
.
Moreover, if W ∈ L
L,L
, then L
C
is transverse to L
C
, L
t
C
and W
C
/(L
C
+ L
t
C
)
⊥
maps onto
W/(L + L
t
)
⊥
under the symplectomorphism above. Thus,
ind(L, L
t
; W) = ind(L
C
, L
t
C
; W
C
).
Using this remark, we can deﬁne the isomorphism M
λ,λ
(E) → M
λη,λ
η
(η/η
⊥
) ﬁberwise
by sending (x, f) ∈ M
λ,λ
(E) to (x, f
η
), where
f
η
(W
ηx
) = f(W)
for each W ∈ L
λx,λ
x
.
2
Lemma 5.24 Suppose that E is a symplectic vector bundle over Y . If λ, λ
t
are lagrangian
subbundles such that λ ∩ λ
t
has constant rank, then M
λ,λ
(E) is canonically isomorphic to
Y Z.
Proof. If λ, λ
t
are ﬁberwise transverse, then there exists a vector bundle symplectomorphism
E → λ ⊕λ
∗
which maps λ onto λ ⊕0 and λ
t
onto 0 ⊕λ
∗
. Any choice of a positivedeﬁnite
quadratic form on λ is induced by a section T of Hom(λ, λ
∗
) comprised of selfadjoint maps.
74
The lagrangian subbundle λ
tt
of E which maps to the subbundle of λ ⊕ λ
∗
deﬁned by the
graph of T is then transverse to λ, λ
t
and clearly satisﬁes
ind(λ
x
, λ
t
x
; λ
tt
x
) = 0
for all x ∈ Y . A natural section s(x) = (x, f) of M
λ,λ
(E) is therefore given by
f(λ
tt
x
) = 0.
In general, our hypotheses imply that η = λ + λ
t
is a coisotropic subbundle of E which
contains the lagrangian subbundles λ and λ
t
. By Lemma 5.23, the bundle M
λ,λ
(E) is
canonically isomorphic to M
λη,λ
η
(η/η
⊥
). Since λ
η
, λ
t
η
are transverse lagrangian subbundles
of η/η
⊥
, the latter bundle has a canonical trivialization, and the assertion follows.
2
Proof of Theorem 5.22. An application of the cocycle identity of Lemma 5.20 shows
that if λ, λ
t
, λ
tt
are lagrangian subbundles of a symplectic vector bundle E, then M
λ,λ
(E) is
canonically isomorphic to the principalbundle product (see Appendix D for the deﬁnition)
M
λ,λ
(E) M
λ
,λ
(E). Applying this remark, we obtain the following canonical isomor
phisms:
M
λ,
˜
λ
(E) · M
λ,λ
η (E) M
λ
η
,
˜
λ
(E)
· M
λ,
˜
λ
(E)
M
λ
η
,
˜
λ
η
(E) M
˜
λ,
˜
λ
η
(E)
.
From the assumption that λ∩η has constant rank, it follows that λ∩λ
η
has constant rank, and
thus M
λ,λ
η (E) is canonically trivial by Lemma 5.24. Similarly, M
˜
λ,
˜
λ
η
(E) is canonically trivial.
Combined with the canonical isomorphisms above, this shows that M
λ,
˜
λ
(E) is canonically
isomorphic to M
λ
η
,
˜
λ
η
(E). Since the lagrangian subbundles λ
η
,
˜
λ
η
of E are contained in the
coisotropic subbundle η, Lemma 5.23 implies that M
λ
η
,
˜
λ
η
(E) is in turn canonically isomorphic
to M
λη,
˜
λη
(η/η
⊥
), completing the proof.
2
Example 5.25 Suppose that E, E
t
are trivial symplectic vector bundles over M with la
grangian subbundles λ ⊂ E and λ
t
⊂ E
t
, and let T be a section of the bundle Hom
ω
(E, E
t
)
of symplectic vector bundle maps. The graph of the section T deﬁnes a lagrangian subbundle
λ
Γ(T)
of E
t
⊕E, while the restriction of T to the subbundle λ deﬁnes a lagrangian subbundle
T(λ) of E
t
.
It is easy to check that the lagrangian subbundles λ
Γ(T)
and λ
t
⊕λ intersect the coisotropic
subbundle E
t
⊕ λ along constant rank subbundles, whereas their reductions by E
t
⊕ λ are
given by T(λ) and λ
t
, respectively. Thus, Theorem 5.22 implies that M
L
Γ(T)
,L
⊕L
(E
t
⊕ E)
and M
T(L),L
(E
t
) are canonically isomorphic.
´
75
Example 5.26 Any symplectomorphism F : T
∗
M →T
∗
N induces a lagrangian embedding
ι
F
: T
∗
M →T
∗
(M N), deﬁned as the composition of the graph Γ
F
: T
∗
M →T
∗
N T
∗
M
of F with the Schwartz transform S
M,N
: T
∗
N T
∗
M →T
∗
(M N). A second lagrangian
embedding z
F
: M →T
∗
N is deﬁned by the composition of F with the zero section s
0
: M →
T
∗
M. Observe that these deﬁnitions imply that the Liouville class λ
M,z
F
of the embedding
(M, z
F
) identiﬁes with λ
T
∗
M,ι
F
under the natural isomorphism H
1
(M; R) · H
1
(T
∗
M; R)
induced by the projection T
∗
M →M.
To see that a similar relation holds for the Maslov classes of (M, z
F
) and (T
∗
M, ι
F
),
consider the symplectic vector bundles E = s
∗
0
T(T
∗
M) and E
t
= z
∗
F
T(T
∗
N) over M, along
with the section T of Hom
ω
(E, E
t
) deﬁned by the restriction of F
∗
to T
Z
(T
∗
M). Moreover, we
have the lagrangian subbundles λ = s
∗
0
V M of E and λ
t
= z
∗
F
V N of E
t
, and the lagrangian
subbundles λ
Γ(T)
⊂ E
t
⊕ E and T(λ) ⊂ E
t
, as described in Example 5.25. Theorem 5.21
shows that s
∗
0
M
T
∗
M,ι
F
is canonically isomorphic to M
λ
Γ
T
,λ
⊕λ
(E
t
⊕ E) and that M
M,z
F
is
canonically isomorphic to M
T(λ),λ
(E
t
). Thus, it follows from Example 5.25 that the Maslov
classes are also related by pullback: s
∗
0
µ
T
∗
M,ι
F
= µ
M,z
F
.
´
Example 5.27 We can apply the conclusion of the preceding example to ﬁberpreserving
symplectomorphisms f : T
∗
M →T
∗
N. From Section 3.2, we recall that any such f is equal
to a ﬁberwise translation of T
∗
M by a closed 1form β on M composed with the cotangent
lift of some diﬀeomorphism N → M. Evidently, the phase class of the induced lagrangian
embedding (M, z
f
) equals [β] ∈
ˇ
H
1
(M; R), and thus ∈ R
+
is admissible for the lagrangian
embeddings (M, z
f
), (T
∗
M, ι
f
) if and only if [β] is integral.
´
5.2 The symplectic category
To systematize the geometric aspects of quantization in arbitrary symplectic manifolds, we
now introduce the symplectic category S. As objects of S we take the class of all smooth,
ﬁnitedimensional symplectic manifolds. Given two objects (P, ω) (Q, ω
t
) of S, we deﬁne
their product as the symplectic manifold (P Q, π
∗
1
ω + π
∗
2
ω
t
), where π
1
, π
2
denote the
cartesian projections. The symplectic dual of an object (P, ω) is the object (P, −ω).
From Lemma 3.14, we recall that a smooth diﬀeomorphism from a symplectic manifold
P to a symplectic manfold Q is a symplectomorphism if and only if its graph is a lagrangian
submanifold of Q P. More generally, an immersed lagrangian submanifold of Q P is
called a canonical relation from P to Q. The morphism set Hom(P, Q) is then deﬁned
to consist of all canonical relations in Q P. Since immersed lagrangian submanifolds of
the product QP coincide with those of its dual, we can therefore deﬁne the adjoint of a
canonical relation L ⊂ Hom(P, Q) as the element L
∗
∈ Hom(Q, P) represented by the same
equivalence class L of immersions into the product P Q.
Composition of morphisms is unfortunately not deﬁned for all L
1
∈ Hom(P, Q) and
L
2
∈ Hom(Q, R), and so S is therefore not a true category. Nevertheless, we can describe
76
a suﬃcient condition for the composability of two canonical relations as follows. By Exam
ple 5.8, the product R ∆
Q
P is a reducible coisotropic submanifold of R QQP,
where ∆
Q
denotes the diagonal in QQ. The product L
2
L
1
of the canonical relations L
1
and L
2
is a welldeﬁned immersed lagrangian submanifold of R Q Q P, and we will
call L
2
L
1
clean if (R∆
Q
P, L
2
L
1
) is a reducible pair. Applying Theorem 5.12, we
then obtain
Proposition 5.28 If L
2
L
1
is clean, then L
2
◦ L
1
is an immersed lagrangian submanifold
of R P, i.e. L
2
◦ L
1
∈ Hom(P, R).
2
Associativity of compositions holds in the symplectic category in the sense that for canonical
relations L
1
, L
2
, L
3
, the equation
L
1
◦ (L
2
◦ L
3
) = (L
1
◦ L
2
) ◦ L
3
is valid provided that both sides are deﬁned.
Among the members of S, there is a minimal object Z, the zerodimensional symplectic
manifold consisting of a single point ∗ equipped with the null symplectic structure. Mor
phisms from Z to any other object P ∈ S identify naturally with immersed lagrangian
submanifolds of P. Thus, the “elements” of P are its immersed lagrangian submanifolds,
and in particular, we may identify the set Hom(P, Q) of morphisms with the “elements” of
QP for any P, Q ∈ S.
8
A canonical relation L ∈ Hom(P, Q) is said to be a monomorphism if the projection
of L onto P is surjective and the projection of L onto Q is injective, in the usual sense.
Equivalently, L is a monomorphism if L
∗
◦ L = id
P
. Dually, one deﬁnes an epimorphism
in the symplectic category as a canonical relation L for which L ◦ L
∗
= id
P
. From these
deﬁnitions we see that a canonical relation is an isomorphism if and only if it is the graph
of a symplectomorphism.
Canonical lifts of relations
The full subcategory of S consisting of cotangent bundles possesses some special properties
due to the Schwartz transform S
M,N
: T
∗
N T
∗
M → T
∗
(M N), which enables us to
identify canonical relations in Hom(T
∗
M, T
∗
N) with immersed lagrangian submanifolds in
T
∗
(M N).
Example 5.29 A smooth relation between two manifolds M, N is a smooth submanifold S
of the product M N. Under the Schwartz transform, the conormal bundle of S identiﬁes
with a canonical relation L
S
∈ Hom(T
∗
N, T
∗
M) called the cotangent lift of S. In view of
Example 3.27, this deﬁnition generalizes tangent lifts of diﬀeomorphisms. Note in particular
8
Another point of view is to deﬁne Hom(P, Q), not as a set, but rather as the object Q P in the
symplectic category; then the composition Hom(P, Q) Hom(Q, R) → Hom(P, R) is the canonical relation
which is the product of three diagonals.
77
that although a smooth map f : M → N does not in general give rise to a welldeﬁned
transformation T
∗
N →T
∗
M, its graph in MN nevertheless generates a canonical relation
L
f
∈ Hom(T
∗
N, T
∗
M).
Example 5.30 The diagonal embedding ∆ : M → M M induces a canonical relation
L
∆
∈ Hom(T
∗
(M M), T
∗
M) which in local coordinates assumes the form
L
∆
= ¦((x, α), (x, x, β, γ)) : x ∈ M and α, β, γ ∈ T
∗
x
M satisfy α = β + γ¦.
In other words, L
∆
is the graph of addition in the cotangent bundle. If L
1
, L
2
∈ Hom(∗, T
∗
M)
and L
2
L
1
is identiﬁed with an element of Hom(∗, T
∗
(M M)) by the usual symplecto
morphism T
∗
M T
∗
M → T
∗
(M M), then the sum L
1
+ L
2
∈ Hom(∗, T
∗
M) is deﬁned
as
L
1
+ L
2
= L
∆
◦ (L
2
L
1
).
Note that if L
1
, L
2
coincide with the images of closed 1forms ϕ
1
, ϕ
2
on M, then L
1
+ L
2
equals the image of ϕ
1
+ ϕ
2
.
The dual of the isomorphism T
∗
M T
∗
M →T
∗
(M M) identiﬁes L
∆
with an element
L
t
∆
∈ Hom(T
∗
M, Hom(T
∗
M, T
∗
M)). This canonical relation satisﬁes L
1
+L
2
= L
t
∆
(L
1
)(L
2
),
and if L is the image of a closed 1form ϕ on M, then
L
t
∆
(L) = f
ϕ
,
where f
ϕ
is the ﬁberwise translation mapping introduced in Section 3.2.
´
Example 5.31 (The Legendre transform) As a particular example of this situation, let
V be a smooth manifold, and consider the ﬁber product TV
V
T
∗
V along with its natural
inclusion ι : TV
V
T
∗
V →TV T
∗
V and “evaluation” function ev : TV
V
T
∗
V →R given
by
ev((x, v), (x, p)) = 'p, v`.
If L ⊂ T
∗
(TV
V
T
∗
V ) is the lagrangian submanifold given by the image of the diﬀerential
d(ev), then the pushforward L
leg
= ι
∗
L deﬁnes an isomorphism in Hom(T
∗
(TM), T
∗
(T
∗
M))
given in local coordinates by
((x, v), (ξ, η)) →((x, η), (ξ, −v)).
As noted in [57], this canonical relation can be viewed as a geometric representative of the
Legendre transform in the following sense: If L: TM → R is a hyperregular Lagrangian
function, in the sense that its ﬁberderivative deﬁnes a diﬀeomorphism TM → T
∗
M, then
the composition of L
leg
with the image of dL equals the image of dH, where H: T
∗
M → R
is the classical Legendre transform of L.
´
78
Morphisms associated with coisotropic submanifolds
If C ⊂ P is a reducible coisotropic submanifold, then the reduction relation R
C
∈
Hom(P, C/(
⊥
) is deﬁned as the composition of the quotient map C →C/(
⊥
and the inclu
sion C/(
⊥
→P. Somewhat more concretely, the relation R
C
is the subset
¦([x], x) : x ∈ C¦
of C/(
⊥
P, where [x] is the leaf of the characteristic foliation passing through x ∈ C.
Evidently R
C
is an epimorphism, and so R
C
◦ R
∗
C
= id
P
. On the other hand, R
C
is not a
monomorphism unless C = P, and we deﬁne the projection relation K
C
∈ Hom(P, P) as
the composition R
∗
C
◦ R
C
, i.e.
K
C
= ¦(x, y) : x, y ∈ C, [x] = [y]¦.
By associativity, we have K
C
= K
C
◦ K
C
, and K
∗
C
= K
C
. Thus, K
C
is like an orthogonal
projection operator.
These relations give a simple interpretation of Theorem 5.12 in the symplectic category.
If L is an immersed lagrangian submanifold of P, i.e. L ∈ Hom(Z, P), then L is composable
with both R
C
and K
C
, provided that it intersects C cleanly. In this case, we have
L
C
= R
C
◦ L L
C
= K
C
◦ L.
In particular, K
C
ﬁxes any lagrangian submanifold of C.
Example 5.32 Suppose that T is a foliation on a manifold B whose leaf space B
T
is smooth.
In the notation of Example 5.17, we then ﬁnd that if T
∗
(B/T) is canonically identiﬁed with
the reduced space E/(
⊥
, then the reduction relation R
E
is the Schwartz transform of the
conormal bundle to the graph of the projection B →B/T.
´
5.3 Symplectic manifolds and mechanics
In general, an arbitrary symplectic manifold has no associated “conﬁguration space,” and
therefore the classical and quantum mechanical viewpoints must be adapted to this new
context based on the available structure.
The classical picture
The central objects in the classical picture of mechanics on an arbitrary symplectic manifold
(P, ω) are the semiclassical states, represented as before by lagrangian submanifolds of P
equipped with halfdensities, and the vector space of observables C
∞
(P).
With respect to pointwise multiplication, C
∞
(P) forms a commutative associative alge
bra. Additionally, the symplectic form on P induces a Lie algebra structure on C
∞
(P) given
by the Poisson bracket
¦f, g¦ = X
g
f,
79
where X
g
= ˜ ω
−1
(dg) denotes the hamiltonian vector ﬁeld associated to g. These structures
satisfy the compatibility condition
¦fh, g¦ = f ¦h, g¦ +¦f, g¦ h,
and are referred to collectively by calling C
∞
(P) the Poisson algebra of P.
The classical system evolves along the trajectories of the vector ﬁeld X
H
associated to
the choice of a hamiltonian H : P → R. If f is any observable, then Hamilton’s equations
assume the form
˙
f = ¦H, f¦.
Note that in local Darboux coordinates, the Poisson bracket is given by
¦f, g¦ =
¸
j
∂f
∂q
j
∂g
∂p
j
−
∂f
∂p
j
∂g
∂q
j
.
Setting f = q
j
or f = p
j
in the Poisson bracket form of Hamilton’s equations above yields
their familiar form, as in Section 3.3.
Two functions f, g ∈ C
∞
(P) are said to be in involution if ¦f, g¦ = 0, in which case the
hamiltonian ﬂows of f and g commute. An observable in involution with the hamiltonian
H is called a ﬁrst integral or constant of the motion of the system. A collection f
i
of functions in involution on P is said to be complete if the vanishing of ¦f
i
, g¦ for all i
implies that g is a function of the form g(x) = h(f
1
(x), , f
n
(x)).
The quantum mechanical picture
Quantum mechanical observables should be the vector space /(H
P
) of selfadjoint linear op
erators on some complex Hilbert space H
P
associated to P. The structure of a commutative,
nonassociative algebra is deﬁned on / by Jordan multiplication:
A ◦ B =
1
2
(AB + BA),
representing the quantum analog of pointwise multiplication in the Poisson algebra of P.
Similarly, the dependent commutator
[A, B]
=
i
(AB −BA)
deﬁnes a Lie algebra structure on / analogous to the Poisson bracket on C
∞
(P).
Quantum mechanical states are vectors in H
P
. The timeevolution of the quantum system
is determined by a choice of energy operator
ˆ
H, which acts on states via the Schr¨odinger
equation:
˙
ψ =
i
ˆ
Hψ.
A collection ¦A
j
¦ of quantum observables is said to be complete if any operator B which
commutes with each A
j
is a multiple of the identity. This condition is equivalent to the
irreducibility of ¦A
j
¦, in other words, no nontrivial subspace of H
P
is invariant with respect
to each j.
80
Quantization
Although neither an underlying conﬁguration manifold nor its intrinsic Hilbert space is
generally available in the context of arbitrary symplectic manifolds, the basic goal of quan
tization remains the same as in the case of cotangent bundles: starting with a symplectic
manifold P, we wish to identify a ∗algebra / of operators which give the quantum analog
of the “system” at hand. Early approaches to this problem were based on the principle
that, regardless of how the algebra / is identiﬁed, the correspondence between classical and
quantum observables should be described by a linear map from the Poisson algebra C
∞
(P)
to / which satisﬁes the following criteria known as the Dirac axioms
Deﬁnition 5.33 A linear map ρ : C
∞
(P) →S(H
P
) is called a quantization provided that
it satisﬁes
1. ρ(1) = identity.
2. ρ(¦f, g¦) = [ρ(f), ρ(g)]
3. For some complete set of functions f
1
, , f
n
in involution, the operators
ρ(f
1
), , ρ(f
n
) form a complete commuting set.
As was eventually proven by Groenwald and Van Hove (see [1] for a proof), a quantization
of all classical observables in this sense does not exist in general.
To make the basic quantization problem more tractable, we ﬁrst enlarge the class of
classical objects to be quantized, but then relax the criteria by which quantum and classical
objects are to correspond. Based on an idea of Weyl and von Neumann, we ﬁrst replace
the classical observables by the groups of which they are the inﬁnitesimal generators. The
basic classical objects are then symplectomorphisms, which should be represented by unitary
operators on quantum Hilbert spaces.
A general formulation of the quantization problem is then to deﬁne a “functor” from
the symplectic category to the category of (hermitian) linear spaces. This means that to
each symplectic manifold P we should try to assign a Hilbert space H
P
in such a way that
H
P
is dual to H
P
, and H
PQ
is canonically isomorphic to a (completed) tensor product
H
P
´
⊗H
Q
. However this is accomplished, each canonical relation L ∈ Hom(P, Q) must then
be assigned to a linear operator T
L
∈ Hom(H
P
, H
Q
) · H
Q
⊗ H
∗
P
in a way which commutes
with compositions, i.e.,
T
˜
L◦L
= T
˜
L
◦ T
L
for each
˜
L ∈ Hom(Q, R). In these abstract terms, our classical and quantum correspondences
can be expressed as follows.
81
Object Classical version Quantum version
basic spaces symplectic manifold (P, ω) hermitian vector space H
P
QP H
Q
´
⊗H
P
QP Hom(H
P
, H
Q
)
point ∗ C
state lagrangian submanifold L element of H
P
space of observables Poisson algebra C
∞
(P) symmetric operators on H
P
At this stage, two relevant observations are apparent from our earlier study of WKB
quantization. First, in addition to just a lagrangian submanifold L ⊂ P, it may require
more data (such as a symbol) to determine an element of H
P
in a consistent way. Second,
the Hilbert space H
P
may carry some sort of ﬁltration (e.g. by powers of or by degree
of smoothness), and the quantization may be “correct” only to within a certain degree of
accuracy as measured by the ﬁltration. By “correctness” we mean that composition of
canonical relations should correspond to composition of operators. As we have already seen,
it is too much to require that this condition be satisﬁed exactly. The best we can hope for
is a functorial relation, rather than a mapping, from the classical to the quantum category.
82
6 Fourier Integral Operators
For our ﬁrst examples of a quantization theory which gives a functorial relation between
the classical (symplectic) category and the quantum category of hermitian vector spaces, we
return to WKB quantization in cotangent bundles. In this context, semiclassical states are
represented by pairs (L, s) consisting of an immersed lagrangian submanifold L in T
∗
(MN)
equipped with a symbol s. Our discussion begins by deﬁning a suitable notion of composition
for such states. The WKB quantization I
(L, s) of the state (L, s) is then regarded as the
Schwartz kernel of the corresponding operator in Hom(H
M
, H
N
). A particular concrete case
of this classicalquantum correspondence is given by the symbol calculus of Fourier integral
operators.
6.1 Compositions of semiclassical states
In Section 4.4, we deﬁned a semiclassical state in a cotangent bundle T
∗
M as a pair (L, s)
consisting of a quantizable lagrangian submanifold L ⊂ T
∗
M and a symbol s on L. In this
section, we study certain natural transformations of semiclassical states.
Reduction of semiclassical states
We will say that a reducible pair (C, L) in a symplectic manifold P is properly reducible
if the quotient of I = L
P
C by its characteristic foliation is a smooth, Hausdorﬀ manifold
and the map I →L
C
is proper.
Lemma 6.1 If (C, L) is a properly reducible pair in a symplectic manifold P, then there
exists a natural linear map
[Ω[
1/2
L ⊗[Ω[
1/2
C →[Ω[
1/2
L
C
.
Proof. First note that if V is a symplectic vector space, together with a lagrangian subspace
L and a coisotropic subspace C, then the exact sequence
0 →L ∩ C
⊥
→L ∩ C →L
C
→0
gives rise to the isomorphism
[Λ[
1/2
L
C
⊗[Λ[
1/2
(L ∩ C
⊥
) · [Λ[
1/2
(L ∩ C).
The linear maps v →(v, −v) and (x, y) →x + y deﬁne a second exact sequence
0 →L ∩ C →L ⊕C →L + C →0,
from which we get
[Λ[
1/2
L
C
⊗[Λ[
1/2
(L ∩ C
⊥
) ⊗[Λ[
1/2
(L + C) · [Λ[
1/2
L ⊗[Λ[
1/2
C.
83
Finally, the exact sequence
0 →(L + C)
⊥
→V
˜ ω
→(L + C)
∗
→0
combined with the halfdensity on V induced by the symplectic form deﬁnes a natural
isomorphism
[Λ[
1/2
(L + C) · [Λ[
1/2
(L + C)
⊥
.
Since L ∩ C
⊥
= (L + C)
⊥
, we arrive at
[Λ[
1/2
L
C
⊗[Λ[(L ∩ C
⊥
) · [Λ[
1/2
L ⊗[Λ[
1/2
C.
Now consider a properly reducible pair (C, L) in a symplectic manifold P. By the pre
ceding computations, there is a linear map
[Ω[
1/2
L ⊗[Ω[
1/2
C →[Ω[
1/2
L
C
⊗[Ω[(T
I
).
Since the quotient map I →L
C
is proper, integration over its ﬁbers is welldeﬁned and gives
the desired linear map
[Ω[
1/2
L ⊗[Ω[
1/2
C →[Ω[
1/2
L
C
.
2
Let M be a smooth manifold and consider a submanifold N ⊂ M equipped with a
foliation T such that the leaf space N
T
is a smooth, Hausdorﬀ manifold. From Section 5.1
we recall that the integrability of T implies that
C
N
= ¦(x, p) ∈ T
∗
M : x ∈ N, T
x
⊂ ker(p)¦
is a coisotropic submanifold of T
∗
M whose reduced space is the cotangent bundle T
∗
N
T
of
the leaf space N
T
. If L is an immersed lagrangian submanifold of T
∗
M such that (C
N
, L)
form a reducible pair, then we denote by I the ﬁber product L
T
∗
M
C
N
of L and C
N
and
consider the following commutative diagram
L −−−→ T
∗
M
r
L
¸
ι
¸
I
r
C
N
−−−→ C
N
π
p
C
L
C
−−−→ T
∗
N
Our goal is describe how a symbol s on L naturally induces a symbol s
C
on L
C
.
Lemma 6.2 In the notation of the diagram above, there is a natural isomorphism
r
∗
L
Φ
L,
→π
∗
Φ
L
C
,
.
84
Proof. From the proof of Lemma 5.16 it follows easily that the pullback to I of the Liouville
forms on T
∗
M and T
∗
N coincide, and thus, the pullback to I of the prequantum line bundles
over T
∗
M and T
∗
N are canonically isomorphic.
Similarly, let E be the symplectic vector bundle over I given by the pullback of T(T
∗
M).
Lagrangian subbundles λ, λ
t
of E are then induced by the immersed lagrangian submanifold
L and the vertical subbundle V M of T(T
∗
M). By deﬁnition, the pullback of the Maslov
bundle of L to I is canonically isomorphic to the bundle M
λ,λ
(E).
Now the tangent bundle of C
N
induces a coisotropic subbundle C of E, and one must
check that the pullback of the Maslov bundle of L
C
to I is canonically isomorphic to
M
λ
C
,λ
C
(C/C
⊥
). From Theorem 5.22 we therefore obtain a canonical isomorphism of r
∗
L
M
L
with π
∗
M
L
C
. By tensoring this isomorphism with the isomorphism of prequantum bundles
described in the preceding paragraph, we arrive at the desired isomorphism of phase bundles.
2
A parallel section s of Φ
L,
pulls back to a parallel section of r
∗
L
Φ
L
C
,
, which, under the
isomorphism of Lemma 6.2, identiﬁes with a parallel section of π
∗
Φ
L
C
,
. Since parallel
sections of π
∗
Φ
L
C
,
identify naturally with parallel sections of Φ
L
C
,
, we obtain a map
(parallel sections of Φ
L,
) →(parallel sections of Φ
L
C
,
).
By tensoring with the map of density spaces given above, we obtain, for each halfdensity σ
on C, a natural map of symbol spaces
S
L
→S
L
C
.
We will denote the image of a symbol s on L under this map by s
C,σ
.
Composition of semiclassical states
If L
1
, L
2
are immersed lagrangian submanifolds of T
∗
M, T
∗
N, then their product L
2
L
1
gives
a welldeﬁned immersed lagrangian submanifold of T
∗
(M N) via the Schwartz transform
S
M,N
: T
∗
NT
∗
M →T
∗
(MN). By the fundamental properties of the Schwartz transform
described in Proposition 3.32, it follows easily that the phase bundle Φ
L
2
L
1
,
is canonically
isomorphic to the external tensor product Φ
L
2
,
Φ
−1
L
1
,
. From the discussion in Appendix A
we also have a linear isomorphism of density bundles [Λ[
1/2
(L
2
L
1
) →[Λ[
1/2
L
2
[Λ[
1/2
L
1
,
and thus there is a natural linear map of symbol spaces
o
L
2
⊗o
L
1
→o
L
2
L
1
which we denote by (s
1
, s
2
) →s
2
s
∗
1
.
Deﬁnition 6.3 The product of semiclassical states (L
1
, s
1
), (L
2
, s
2
) in T
∗
M, T
∗
N is the
semiclassical state (L
2
L
1
, s
2
s
∗
1
) in T
∗
(M N).
2
85
Turning to compositions, we ﬁrst note that if M, V are smooth manifolds, then the
Schwartz transform S
M,V
: T
∗
V T
∗
M → T
∗
(M V ) identiﬁes each canonical relation in
Hom(T
∗
M, T
∗
V ) with an immersed lagrangian submanifold in T
∗
(M V ). Our goal is now
to describe how semiclassical states (L, s) in T
∗
(M N) and (L
t
, s
t
) in T
∗
(N V ) deﬁne
a semiclassical state in T
∗
(M V ) when L, L
t
are composable as canonical relations in
Hom(T
∗
M, T
∗
N) and Hom(T
∗
N, T
∗
V ), respectively.
We begin with the following commutative diagram
T
∗
V T
∗
N T
∗
N T
∗
M −−−→ T
∗
(M N N V )
T
∗
V T
∗
M −−−→ T
∗
(M V )
where the upper horizontal arrow denotes the composition S
MN
◦(S
N,V
S
M,N
) of Schwartz
tranforms, and the lower horizontal arrow equals the Schwartz transform S
M,V
. The left
vertical arrow represents the reduction relation deﬁned by the coisotropic submanifold C =
T
∗
V ∆
T
∗
N
T
∗
M, while the right vertical arrow is the reduction relation deﬁned by
the image of C under the Schwartz transform. L
t
◦ L ∈ Hom(T
∗
M, T
∗
V ) is deﬁned as the
reduction of L
2
L
1
by the coisotropic submanifold T
∗
V ∆
T
∗
N
T
∗
M of T
∗
V T
∗
N
T
∗
N T
∗
M. Under the Schwartz transform,
T
∗
V T
∗
N T
∗
N T
∗
M →T
∗
(M N N V ),
the submanifold T
∗
V ∆
T
∗
N
T
∗
M maps to the conormal submanifold C
V,N,M
of T
∗
(V
NNM) deﬁned by V ∆
N
M ⊂ V NNM and its product foliation by subsets of
the form v∆
N
m for (v, m) ∈ V M. Thus, the Schwartz transform of L
2
L
1
reduces by
C
V,N,M
to yield the image of L
2
◦L
1
under the Schwartz transform T
∗
V T
∗
M →T
∗
(V M).
From Lemma 6.1, it follows that for any properly reducible pair L
1
, L
2
, we obtain a
natural linear map of symbol spaces
o
L
2
⊗o
L
1
→o
L
2
◦L
1
given by the composition of the product map above and the reduction map o
L
2
L
1
→o
L
2
◦L
1
deﬁned by the natural halfdensity on C
V,N,M
induced by the symplectic forms on T
∗
V, T
∗
N,
and T
∗
M. We denote the image of s
2
s
1
under this map by s
2
◦ s
1
.
We will say that semiclassical states (L
1
, s
1
) in T
∗
(M N) and (L
2
, s
2
) in T
∗
(N V )
are composable if the conormal submanifold C
V,N,M
and immersed lagrangian submanifold
L
2
L
1
form a properly reducible pair.
Deﬁnition 6.4 If a semiclassical state (L
1
, s
1
) in T
∗
(M N) is composable with a semi
classical state (L
2
, s
2
) in T
∗
(N V ), then their composition is deﬁned as the semiclassical
state (L
2
◦ L
1
, s
2
◦ s
1
) in T
∗
(M V ).
2
86
Example 6.5 Consider a semiclassical state (L
F
, ˜s) in T
∗
(MN), where L
F
is the Schwartz
transform of the graph of a symplectomorphism F : T
∗
M → T
∗
N. If L is an immersed la
grangian submanifold of T
∗
M, then Deﬁnition 6.4 provides that for each symbol s ∈ o
L
,
the semiclassical state (L, s) in T
∗
M is transformed by F into the semiclassical state
(L
F
◦ L, ˜s ◦ s) in T
∗
N.
To make this correspondence more explicit, let us choose a particular lagrangian immer
sion ι : L →T
∗
M representing L (see the discussion in Section 4.4). We then note that the
natural halfdensity [ω
n
M
[
1/2
on L
F
≈ T
∗
M enables us to identify the symbol space o
L
F
with
the product
Γ
L
F
⊗
C
C
∞
(L
F
, C).
Given a parallel section ˜ s and smooth complexvalued function h on L
F
, we ﬁnd by a com
putation that the isomorphism o
L
→o
L
F
◦L
determined by the symbol ˜s = ˜ s ⊗h is given by
s ⊗a →s
t
⊗(a ι
∗
h), where a is any halfdensity on L and s
t
is the unique parallel section of
Φ
L
F
◦L,
such that ι
∗
˜ s →s
t
⊗s
−1
under the canonical isomorphism ι
∗
Φ
L
F
,
· Φ
L
F
◦L,
⊗Φ
−1
L,
.
´
Example 6.6 As a special case of Example 6.5, note that if the symplectomorphism F :
T
∗
M → T
∗
N is the cotangent lift of a diﬀeomorphism M → N, then for each > 0, the
phase bundle of of L
F
admits a canonical parallel section, and so the symbol space o
L
F
identiﬁes naturally with R
+
C
∞
(L
F
, C). If h ∈ C
∞
(L
F
, C), then with this identiﬁcation,
the composition of (L
F
, h) with a semiclassical state (L, s) equals the semiclassical state
(L
F
◦ L, (h ◦ ι) s).
Similarly, if β is a closed 1form on M and the symplectomorphism F : T
∗
M → T
∗
M
equals ﬁberwise translation by β, then ∈ R
+
is admissible for L
F
if and only if [β] is
integral. For such , each parallel section of the phase bundle Φ
L
F
,
identiﬁes with an
oscillatory function of the form ce
iS/
for c ∈ R and some S : M → T
satisfying S
∗
dσ =
β. A computation then shows that the composition of a semiclassical state of the form
(L
F
, e
iS/
⊗h) with (L, ι, s) yields the semiclassical state (L
F
◦ L, e
i(S◦π
L
)/
(h ◦ ι) s).
´
6.2 WKB quantization and compositions
To deﬁne the composition of semiclassical states (L, s) ∈ Hom(T
∗
M, T
∗
N) and (
˜
L, ˜s) ∈
Hom(T
∗
N, T
∗
V ) as a semiclassical state (
˜
L ◦ L, ˜s ◦ s) ∈ Hom(T
∗
M, T
∗
V ), we used the
Schwartz transform to identify the immersed lagrangian submanifold
˜
L ◦ L ∈ T
∗
V T
∗
M
with an immersed lagrangian submanifold in T
∗
(M V ), since it is the cotangent bundle
structure of the latter space which gives meaning to “symbols” (in the sense of Chapter 4)
on
˜
L ◦ L. On the quantum level, an analogous correspondence is furnished by the Schwartz
kernel theorem.
Let M be a smooth manifold and let [Ω[
1/2
0
M be equipped with the topology of C
∞
convergence on compact sets. A distributional halfdensity on M is then a continuous,
Clinear functional on [Ω[
1/2
0
M. We denote the space of distributional halfdensities on M
87
by [Ω[
1/2
−∞
M, and we equip this space with the weak
∗
topology. If N is another smooth
manifold, a kernel is any element of [Ω[
1/2
−∞
(M N). A kernel K deﬁnes a linear map
/: [Ω[
1/2
0
M →[Ω[
1/2
−∞
N by duality via the equation
'/(u), v`
def
= 'K, u ⊗v` (∗)
Schwartz kernel theorem . Every K ∈ [Ω[
1/2
−∞
(MN) deﬁnes a linear map /: [Ω[
1/2
0
M →
[Ω[
1/2
−∞
N by (∗) above, which is continuous in the sense that /(φ
j
) →0 in [Ω[
1/2
−∞
N if φ
j
→0
in [Ω[
1/2
0
M. Conversely, to every such linear map, there is precisely one distribution K such
that (∗) is valid.
2
Of course, each w ∈ [Ω[
1/2
N deﬁnes an element ˜ w ∈ [Ω[
1/2
−∞
N by the equation
' ˜ w, v` =
N
v ⊗w.
In all of the examples we will consider, the image of the map / will lie within the subspace
of [Ω[
1/2
−∞
N represented by elements of [Ω[
1/2
N in this way, and so we can consider / as
a continuous linear map [Ω[
1/2
0
M → [Ω[
1/2
0
N, thereby giving a (densely deﬁned) operator
H
M
→H
N
on the intrinsic Hilbert spaces of M, N. In practical terms, the value of the half
density Ku at each y ∈ N can be computed in these cases by, roughly speaking, integrating
the product K ⊗ u over the submanifold ¦y¦ M of N M. Thus, in the same way that
the Schwartz transform provides a natural correspondence
Hom(T
∗
M, T
∗
N) ↔T
∗
(M N),
the Schwartz kernel theorem gives the identiﬁcation
[Ω[
1/2
−∞
(M N) ↔Hom(H
M
, H
N
)
diﬀerential operators
With respect to linear coordinates ¦x
j
¦ on R
n
, we deﬁne dependent operators
D
j
def
= −i
∂
∂x
j
.
An diﬀerential operator of asymptotic order
9
k ∈ Z is then an asymptotic series of the
form
P
=
∞
¸
m=0
P
m
m+k
,
9
Note that this order generally diﬀers from the order of a diﬀerential operator.
88
where each P
m
is a polynomial in the operators D
j
. By formally substituting ξ
j
for D
j
in
each term of P
, we obtain an dependent function σ
P
on T
∗
R
n
, known as the symbol of
P
, which is related to P
by the asymptotic Fourier inversion formula: for any compactly
supported oscillatory test function e
−iψ/
u (we permit ψ = 0), we have
(P
e
−iψ/
u)(x) = (2π)
−n
e
i(/x−y,ξ)−ψ)/
σ
P
(x, ξ) u(y) dy dξ.
The principal symbol p
of the operator P
is deﬁned as the symbol of the ﬁrst nonvanishing
term in its series expansion, i.e.,
p
(x, ξ) = σ
Pm
0
(x, ξ)
m
0
+k
if P
m
= 0 for all m < m
0
and P
m
0
= 0. By replacing σ
P
by p
in the formula above, we
obtain
(P
e
−iψ/
u)(x) = (2π)
−n
e
i(/x−y,ξ)−ψ)/
p
(x, y, ξ) u(y) dy dξ + O(
m
0
+k+1
)
where p
is of course independent of y. (Later we will drop this assumption to obtain a more
symmetric calculus). For ﬁxed x, the function (y, ξ) → 'x − y, ξ` − ψ has a nondegenerate
critical point when y = x and ξ = dψ(x). An application of the principle of stationary phase
therefore gives
(P
e
−iψ/
u)(x) = e
−iψ(x))/
u(x) p
(x, x, dψ(x)) h
m
0
+k
+ O(
m
0
+k+1
) (∗∗)
To interpret this expression in the context of WKB quantization, we ﬁrst note that the phase
function φ(x, y, ξ) = 'x − y, ξ` on B = R
n
R
n
(R
n
)
∗
, together with V = R
n
R
n
and
the cartesian projection p
V
: B → V deﬁne a Morse family (B, V, p
V
, φ) which generates
the conormal bundle to the diagonal ∆ ⊂ R
n
R
n
. The principal symbol p
, written as a
function of the variables (x, y, ξ) deﬁnes an amplitude a = p
[dx dy[
1/2
[dξ[ on B, whose
restriction to the ﬁbercritical set Σ
φ
= ¦(x, x, ξ) : 'x, ξ` = 0¦ induces a welldeﬁned symbol
s
P
= p
(x, x, ξ) on L
∆
.
Now, e
−iψ/
u = I
(L, s), where L is the projectable lagrangian submanifold of T
∗
R
n
deﬁned by im(dψ), and s is obtained from the pullback of u to L. By Example 6.6, we have
(L, s) = (L
∆
◦ L, s
P
◦ s), that is, composition with (L
∆
, s
P
) multiplies the symbol s by the
values of the principal symbol p
on L · Σ
φ
. Combined with the preceding equation, this
gives
(P
e
−iψ/
u)(x) = I
(L
∆
◦ L, s
P
◦ s)
m
0
+k
+ O(
m
0
+k+1
).
The Schwartz kernel for the operator P
is given by the distribution family I
(L
∆
, s
P
).
An diﬀerential operator on a manifold M is an operator P
on [Ω[
1/2
0
M which coincides
in local coordinates with a series of the form P
as above. While the symbol of P
depends
on the choice of local coordinates, its principal symbol p
is a welldeﬁned function on the
cotangent bundle T
∗
M. Using a global generating function for the conormal bundle of
∆ ⊂ M M (see Example 4.29) and arguing as above, we obtain the following geometric
version of (∗∗) above:
89
Theorem 6.7 If (L, s) is an exact, projectable semiclassical state in T
∗
M and P
is an
diﬀerential operator of order k on M, then:
P
(I
(L, s)) = I
(L
∆
◦ L, s
P
◦ s) + O(
m
0
+k+1
),
where s
P
is the symbol on L
∆
induced by the principal symbol p
of P
.
Roughly speaking, this theorem asserts that polynomial functions on the identity relation L
∆
quantize as the Schwartz kernel of diﬀerential operators on M. As usual, this correspondence
is only asymptotic; many diﬀerential operators share the same principal symbol, and their
actions on a given function coincide only up to terms of higher order in .
As a particular case of Theorem 6.7, the vanishing of the principal symbol p
on L implies
that
P
(I
(L, s)) = O(
m
0
+k+1
),
i.e., that I
(L, s) is a ﬁrstorder approximate solution to the equation P
u = 0. This remark
suggests the quantum analog of certain coisotropic submanifolds of T
∗
M. The zero set of the
principal symbol p
is called the characteristic variety of the operator P
. If 0 is a regular
value of p
, then C
P
= p
−1
(0) is a coisotropic submanifold of T
∗
M, and semiclassical states
contained in C
P
represent solutions of the asymptotic diﬀerential equation P
u = 0. In this
sense, C
P
, or, more properly, the reduced manifold of C
P
, corresponds to the kernel of P
in H
M
, and the projection relation K
C
P
quantizes as the orthogonal projection onto this
subspace.
A familiar illustration of these concepts is provided by the WKB approximation.
Example 6.8 Recall that the Schr¨odinger operator associated to a given potential V on a
riemannian manifold M is given by
ˆ
H = −
2
2m
∆ + m
V
.
For E > 0, the timeindependent Schr¨odinger equation is then
P
ψ = 0,
where P
=
ˆ
H −E is the zerothorder asymptotic diﬀerential operator on M with principal
symbol
p
(x, θ) = −
[θ[
2
2m
+ (V (x) −E).
The characteristic variety of p
is simply the level set H
−1
(E) of the classical hamiltonian
of the system, and as in previous sections we see that ﬁrstorder approximate solutions to
the timeindependent Schr¨odinger equation arise from semiclassical states represented by
quantizable lagrangian submanifolds in C
P
.
´
90
The diﬃculty of quantizing more general symbols on the identity relation L
∆
lies in the
convergence of the integral
(P
u)(x) = (2π)
n
e
i/x−y,θ)/
a(x, y, ξ) u(y) dξ dy.
For diﬀerential operators, integration over the phase variables θ was welldeﬁned due to the
fact that the symbol of a diﬀerential operator has a polynomial growth rate with respect to
these variables. Weakening this condition while still guaranteeing that the integral converges
leads to the deﬁnition of pseudodiﬀerential operators.
In R
n
, an pseudodiﬀerential operator of order µ is given by
(Au)(x) = (2π)
n
e
i/x−y,θ)/
a
(x, θ) u(y) dθ dy,
where the symbol a(x, θ) is an asymptotic series in of the form
a
(x, θ) ∼
¸
a
j
(x, θ)
µ+j
.
Each coeﬃcient a
j
is a smooth function on R
n
R
n
` ¦0¦ and is positively homogeneous of
degree µ −j, i.e.,
a
j
(x, cθ) = c
µ−j
a
j
(x, θ)
for c > 0. As in the special case of diﬀerential operators, the invariance of pseudodiﬀerential
operators under coordinate changes enables one to extend this theory to any smooth manifold
M. In this case, the principal symbol of a pseudodiﬀerential operator is a welldeﬁned
function on T
∗
M, and there is a corresponding version of Theorem 6.7 (see for example
[51]).
Fourier integral operators
In its simplest form, the theory of Fourier integral operators provides a means for quantizing
halfdensities on more general lagrangian submanifolds L of T
∗
M by replacing the function
(x, y, ξ) → 'x − y, ξ` in the deﬁnition of pseudodiﬀerential operators by a phase function
which generates L:
(Au)(x) = (2π)
n
e
iφ(x,y,ξ)/
a(x, y, ξ) u(y) dξ dy.
As before, the class of quantizable halfdensities (and canonical relations) is constrained
by the necessary conditions for this integral to converge. In this section, we will only be
concerned with a few specialized cases; a detailed description of this theory can be found in
[21, 28, 31, 43, 56].
Perhaps the simplest generalization of the picture of (pseudo)diﬀerential operators given
above is provided by quantizing semiclassical states whose underlying lagrangian subman
ifolds are conormal bundles. Let M be a smooth manifold with a smooth submanifold N.
As constructed in Example 4.29, the conormal bundle L
N
⊂ T
∗
N is generated by a single
91
Morse family (B, V, p
V
, φ) with the properties that V is a tubular neighborhood of N in M
and B is a vector bundle over V with ﬁber dimension equal to the codimension of N in M.
As in the case of (pseudo)diﬀerential operators, an amplitude a on B having an asymptotic
expansion in terms which are positively homogeneous with respect to the natural R
+
action
on the ﬁbers of p
V
: B → V gives rise to a welldeﬁned distributional halfdensity family
I
(L, s) on M.
It is easy to see that the ﬁbercritical set Σ
φ
is invariant under the R
+
action on B, and
that the identiﬁcation L
N
· Σ
φ
deﬁned by λ
φ
is equivariant with respect to the natural
R
+
action on L
N
. Consequently, positively homogeneous symbols on L
N
induce positively
homogeneous amplitudes on B of the same order, and we can proceed to deﬁne I
(L, s) as
before by requiring that s be homogeneous.
When M = X Y is a product manifold, the distributions I
(L, s) represent Schwartz
kernels for continuous linear operators [Ω[
1/2
0
X →[Ω[
1/2
∞
Y . Under certain additional restric
tions on N which are always fulﬁlled, for example, when N is the graph of a diﬀeomorphism
X →Y , these operators map [Ω[
1/2
0
X to [Ω[
1/2
0
Y and extend continuously to [Ω[
1/2
−∞
X. More
over, they satisfy the composition law
I
(
˜
L, ˜s) ◦ I
(L, s) = I
(
˜
L ◦ L, ˜s ◦ s).
Example 6.9 If X, Y are smooth manifolds and f : X → Y is a smooth diﬀeomorphism,
then the procedure described above can be applied to the graph Γ
f
of f, viewed as a
smooth submanifold of X Y . The family I
(L
f
, s) of distributions on X Y deﬁned
above corresponds via the Schwartz kernel theorem to a family of continuous linear maps
[Ω[
1/2
0
X →[Ω[
1/2
−∞
Y
A diﬀeomorphism f : X →Y induces a unitary operator on intrinsic Hilbert spaces given
by pullback:
(f
∗
)
−1
: H
X
→H
Y
.
At the same time, f gives rise to a symplectomorphism of cotangent bundles:
(T
∗
f)
−1
: T
∗
X →T
∗
Y.
Of course, the more interesting Fourier integral operators from [Ω[
1/2
X to [Ω[
1/2
Y come
from quantizing canonical relations from T
∗
X to T
∗
Y which do not arise from diﬀeomor
phisms from X to Y . For instance, quantizing the hamiltonian ﬂow ¦ϕ
t
¦ of a hamiltonian
H on T
∗
X gives the solution operators exp(−it
ˆ
H/) of the Schr¨odinger equation.
´
92
7 Geometric Quantization
7.1 Prequantization
In the preceding chapter, we observed that the characteristic variety C = C
P
of a diﬀerential
operator P
on a riemannian manifold M represents the classical analog of the kernel of P
in
H
M
, in the sense that semiclassical states contained in C quantize to ﬁrstorder approximate
solutions of the equation P
= 0. When C is reducible, the lagrangian submanifolds of M
contained in C correspond to lagrangian submanifolds of the reduced manifold C/(
⊥
. This
suggests that if C is reducible, then a quantum Hilbert space H
C
somehow associated to the
reduced manifold C/(
⊥
should map isometrically onto the kernel of P
via an appropriate
quantization of the adjoint reduction relation R
∗
C
∈ Hom(C/(
⊥
, T
∗
M). Before turning to a
systematic means for deﬁning H
C
, let us consider what conditions the reduced manifold C/(
⊥
must satisfy in order that quantizations of the reduction and projection relations be well
deﬁned for arbitrary quantizable semiclassical states. Speciﬁcally, we ask what assumptions
on C guarantee that the classical projection operation K
C
preserves the class of quantizable
lagrangian submanifolds in T
∗
M.
For the time being, we will ignore the Maslov correction and assume that C/(
⊥
is simply
connected. A simple argument then shows that K
C
preserves the class of prequantizable
lagrangian submanifolds provided that the Liouville class of each (
⊥
leaf is integral.
To interpret this condition in terms of C/(
⊥
, ﬁrst note that if C/(
⊥
is simplyconnected,
any class [a] ∈ H
2
(C/(
⊥
; Z) is represented by a continuous map
(D, ∂D)
f
→(C/(
⊥
, [I])
for a ﬁxed leaf I of (
⊥
. By the homotopy lifting property, f lifts to a continuous map
(D, ∂D)
˜
f
→(C, I).
Applying Stokes’ theorem, we then have
[a]
ω
C
=
∂D
˜
f
∗
α
M
∈ Z
by the assumption that the Liouville class of I is integral. Thus, we conclude that K
C
preserves prequantizable lagrangian submanifolds provided that the reduced symplectic form
ω
C
is itself integral. In this case, the reduced manifold C/(
⊥
is said to be prequantizable.
Example 7.1 From Example 5.13 we recall that the standard metric on the unit nsphere
S
n
induces a kinetic energy function k
n
whose constant energy surfaces C
E
are reducible
coisotropic submanifolds of T
∗
S
n
for E > 0. Each leaf of the characteristic foliation of C
E
is a circle S which projects diﬀeomorphically to a great circle in S
n
. If γ parametrizes a
such a geodesic, then its lift to a parametrization of S satisﬁes ˙ γ = (2E)
−1/2
X
kn
. Since
α
M
(X
kn
) = 2E, the Liouville class of S is determined by the number
S
α
M
= 2π (2E)
1/2
.
Thus C
E
is prequantizable if and only if E = (n)
2
/2 (compare Example 4.2).
93
´
By Theorem D.2, this condition is equivalent to the existence of a principal T
bundle
Q over C/(
⊥
. In fact, the bundle Q can be constructed explicitly from the prequantum
T
bundle Q
M
over T
∗
M by noting that for any leaf I of (
⊥
, the modZ
reduction of λ
I
represents the holonomy of Q
M
[
I
; if λ
I
is integral, there exists a parallel section of Q
M
over I. If C/(
⊥
is prequantizable, parallel sections over the leaves of (
⊥
deﬁne a foliation
O of Q
M
[
L
whose leaf space is a principal T
bundle Q over C/(
⊥
. The connection on Q
induced by the connection on Q
M
has curvature equal to the reduced symplectic form ω
C
.
A linear mapping ρ : C
∞
(P) → S(H
P
) satisfying the ﬁrst two Dirac axioms will be
called a prequantization of the classical system represented by P. The basic example of
prequantization is that of cotangent bundles due to Segal [52], Koopman, and Van Hove.
In this case, the prequantum Hilbert space is taken to be the completion of the space of
smooth, complexvalued functions on the cotangent bundle P = T
∗
M itself, with respect to
the innerproduct
'f, g` =
P
f g ω
n
M
.
The mapping C
∞
(P) →S(H
P
) is then given explicitly by the formula
ρ(f) = −i X
f
+ m
L(f)
,
where as usual m
L(f)
denotes multiplication by the lagrangian L(f) = f −α
M
(X
f
). Clearly
the map ρ is linear and satisﬁes the ﬁrst Dirac axiom; veriﬁcation of the second is also
straightforward and will be carried out in somewhat more generality below.
The Segal prequantization is an important ﬁrst step towards attempts at (pre)quantizing
more general symplectic manifolds. Roughly speaking, the idea is the following: Although
the symplectic form of an arbitrary symplectic manifold (P, ω) is not exact, we can choose
a covering of P by open sets U
j
on which the restriction of the symplectic form satisﬁes
ω = −dα
j
for appropriately chosen 1forms α
j
on U
j
. In direct analogy with the Segal
prequantization, we can then associate to a function f ∈ C
∞
(P) the operator
ρ(f)
j
= −iX
f
+ m
L
j
(f)
,
on C
∞
(U
j
). In order to associate a “global” operator to f, we hope to piece these local
operators together. One way of doing this would be to impose the condition that each ρ(f)
j
and ρ(f)
k
coincide as operators on the function space C
∞
(U
j
∩U
k
). This essentially requires
that the 1forms α
j
agree on overlaps, and we arrive at nothing new.
A true generalization of the Segal construction due to Kostant and Souriau is achieved
by ﬁrst reinterpreting ρ(f) as an operator on sections of a line bundle over P = T
∗
M. More
precisely, let E be the complex line bundle over P associated to the trivial principal T
bundle Q = P T
via the representation x → e
−ix/
of T
in U(1). The space of sections
of E identiﬁes canonically with C
∞
(P) by means of the constant section s = 1, and we have
ρ(f) = −i∇
X
f
+ m
f
,
94
where ∇ denotes the connection on E induced by the connection 1form ϕ = −α
M
+dσ on Q.
Returning to the general case, we ﬁnd that the operators ρ(f)
j
enjoy a similar interpretation
provided that −α
j
+ dσ deﬁnes a local representative of a connection 1form on a (possi
bly nontrivial) principal T
bundle Q over P. The curvature of such a bundle necessarily
coincides with the symplectic form on P; according to the discussion in Appendix C, this
condition can be satisﬁed if and only if ω is integral.
Deﬁnition 7.2 A prequantization of a symplectic manifold (P, ω) is a principal T
bundle
Q over P equipped with a connection 1form ϕ having curvature ω.
The upshot of the KostantSouriau construction is that prequantizable symplectic manifolds
have prequantizable Poisson algebras. While a direct proof of this fact follows the outline of
the preceding paragraph, we will study prequantizable manifolds in somewhat more detail
below. The prequantum Hilbert space in this case will be the completion of the vector
space of smooth sections of a hermitian line bundle associated to Q. Although several more
modiﬁcations of this choice must be made in order to arrive at a reasonable substitute for
the intrinsic Hilbert space of the base of a cotangent bundle, this is an important ﬁrst step
in our general quantization program.
For the remainder of this section, we will focus on geometric properties of prequantum
circle bundles and prove that their existence coincides with the prequantizability of C
∞
(P).
Automorphisms of (Q, ϕ)
Let P be a prequantizable symplectic manifold with prequantumT
bundle Q and connection
ϕ. To a function f ∈ C
∞
(P), we associate an operator on C
∞
(Q):
ξ
f
= X
f
−fX,
where X
f
denotes the horizontal lift of the hamiltonian vector ﬁeld of f, and X is the
fundamental vector ﬁeld on Q deﬁned by the equations
X ϕ = 1 X dϕ = 0.
A direct computation shows that the connection form ϕ is invariant under the ﬂow of ξ
(k)
f
.
Conversely, if L
ξ
ϕ = 0 for some vector ﬁeld ξ on Q, then we can decompose ξ into its
horizontal and vertical parts:
ξ = ξ −gX
for some realvalued function g on Q satisfying dg = ξ dϕ. From the deﬁnition of X it
follows that X g = 0 and [ξ, X] = 0, and consequently, [ξ, X] = 0. Thus, ξ is the horizontal
lift of X
g
, and so ξ = ξ
g
. Moreover, the requirement that the curvature of ϕ equal the
symplectic form ω implies
[ξ
f
, ξ
g
] = [X
f
, X
g
] + ω(X
f
, X
g
)X −2¦f, g¦X = ξ
f,g¦
.
95
The association f → ξ
f
therefore deﬁnes a Lie algebra isomorphism between the Poisson
algebra C
∞
(P) and the space χ(Q, ϕ) of ϕpreserving vector ﬁelds on Q with the standard
Lie bracket. This produces the exact sequence of Lie algebras
0 →R →χ(Q, ϕ) →χ(P, ω) →H
1
(P; R) →0,
where H
1
(P; R) is assigned the trivial bracket, and the image of χ(Q, ϕ) in χ(P, ω) consists
precisely of the hamiltonian vector ﬁelds on P. This sequence can be integrated to give
an exact sequence of automorphism groups as follows. Let (Q, ϕ) be a prequantization of
(P, ω), and let Aut(Q, ϕ), Aut(P, ω) denote those groups of diﬀeomorphisms which preserve
ϕ, ω respectively. By the deﬁnition of X it follows that every F ∈ Aut(Q, ϕ) preserves X and
is therefore T
equivariant. In particular, this means that F is the lift of a diﬀeomorphism f
of P; from the fact that π
∗
ω = dϕ it follows furthermore that f
∗
ω = ω, that is, f ∈ Aut(P, ω).
The association F →f deﬁnes a group homomorphism
Aut(Q, ϕ) →Aut(P, ω).
To determine its kernel, we simply note that the identity map on P is covered by precisely
those automorphisms of Q given by the action of elements of T
on its connected components.
This implies that the kernel is isomorphic to H
0
(P, T
); if P (and hence Q) is connected,
this is just the circle T
and we have the exact sequence
0 →T
→Aut(Q, ϕ) →Aut(P, ω),
i.e. Aut(Q, ϕ) is a central extension of Aut(P, ω) by T
. We interpret this observation to
mean that an automorphism of Q is determined “up to phase” by an automorphism of P.
If we equip the space C
∞
(Q) of complexvalued functions on Q with the innerproduct
'u, v` =
Q
uv µ,
where µ denotes the volume form ϕ∧(dϕ)
n
on Q, then each F ∈ Aut(Q, ϕ) preserves µ, and
therefore deﬁnes a unitary operator U
F
on C
∞
(Q) by composition:
U
F
(u) = u ◦ F.
Clearly the correspondence F →U
F
deﬁnes a unitary representation of Aut(Q, ϕ) on L
2
(Q);
from the exact sequence above, we therefore obtain a projective unitary representation of the
image of Aut(Q, ϕ) in Aut(P, ω).
Example 7.3 Translations of R
2n
are generated by the hamiltonian vector ﬁelds associated
to linear functionals on R
2n
. A basis for this space is given by the vector ﬁelds
X
q
i
= −
∂
∂p
i
X
p
i
=
∂
∂q
i
,
which assume the form
ξ
q
i
= −
∂
∂p
i
−q
i
X ξ
p
i
=
∂
∂q
i
96
when lifted to Q = R
2n
T
via the Segal prescription ξ
f
= X
f
− L(f)X, where again
L(f) = f − α
M
(X
f
). Our earlier results show that [ξ
q
i
, ξ
p
j
] = ξ
q
i
,p
j
¦
= δ
ij
X, and so the
vector ﬁelds ξ
q
i
, ξ
p
i
generate a Lie subalgebra h
n
⊂ χ(Q, ϕ) isomorphic to R
2n
R with
bracket given by
[(v, a), (w, b)] = (0, ω(v, w)).
By exponentiating, we ﬁnd that h
n
corresponds to a subgroup H
n
⊂ Aut(Q, ϕ) which
comprises a central extension of the translation group:
0 →T
→H
n
→R
2n
→0.
The group H
n
is known as the Heisenberg group of R
2n
with its usual symplectic structure.
In explicit terms, H
n
is diﬀeomorphic to R
2n
T
, with group multiplication given by
(Q, P, σ) (Q
t
, P
t
, σ
t
) = (Q + Q
t
, P + P
t
, σ
t
) = (Q + Q
t
, P + P
t
, σ + σ
t
+
¸
j
P
j
Q
t
j
).
´
KostantSouriau prequantization
To prequantize the Poisson algebra C
∞
(P), we ﬁrst recall that complex line bundles E
k
associated to Q arise via representations of T
in U(1) of the form x → e
ikx/
. Smooth
sections of E
k
identify with functions on Q satisfying the equivariance condition
f(p a) = e
−ika/
f(p)
for a ∈ T
; in other words, the space of sections of E
k
is isomorphic to the −ik/eigenspace
c
k
of the fundamental vector ﬁeld X. Under this correspondence, covariant diﬀerentiation
by a vector ﬁeld η on P is given simply by the Lie derivative with respect to the horizontal
lift η of η to Q.
To each f ∈ C
∞
(P) and integer k, we assign an operator on C
∞
(Q) by
ξ
(k)
f
= −
i
k
ξ
f
.
Since [ξ
f
, X] = 0, the operator ξ
(k)
f
restricts to an operator on each eigenspace c
k
of X having
the form
ρ
k
(f) = −
i
k
X
f
+ m
f
.
Evidently the map ρ
k
satisﬁes the ﬁrst Dirac axiom; moreover
ρ
k
(¦f, g¦) = k [ρ
k
(f), ρ
k
(g)]
.
To verify that ρ
k
(f) is selfadjoint, is suﬃces to prove that ρ
k
(f) acts as a symmetric operator
on the real subspace of c
k
. For this purpose, we note that if the hamiltonian vector ﬁeld of
97
f is integrable (for example, if f is compactly supported), then the fact that ξ
f
preserves ϕ
implies that
'ξ
f
u, v` = −'u, ξ
f
v`
for realvalued functions u, v on Q. Combined with the deﬁnition of ρ
k
, conjugatesymmetry
of the innerproduct implies that
'ρ
k
(f) u, v` = 'u, ρ
k
(f) v`.
We now deﬁne the prequantum line bundle associated to Q as E = E
−1
, equipped
with a hermitian metric ' , ` compatible with its induced connection, and let H
P
denote
the L
2
completion of the space of compactly supported sections of E with respect to the
innerproduct
's
1
, s
2
` =
P
's
1
, s
2
` ω
n
.
For each f ∈ C
∞
(P), the operator ρ
−1
(f) extends to an essentially selfadjoint operator on
H
P
; from the general remarks above, it follows that ρ
−1
deﬁnes a prequantization of the
Poisson algebra C
∞
(P).
7.2 Polarizations and the metaplectic correction
The KostantSouriau prequantization of symplectic manifolds fails to satisfy the third Dirac
axiom and is therefore not a quantization. In fact, the position and momentum operators
ρ(q) = −i
∂
∂p
+ m
q
ρ(p) = −i
∂
∂q
on T
∗
R both commute with ∂/∂p and therefore do not form a complete set. According to
our earliest concepts of quantization (see the Introduction), the operator corresponding to q
should act by multiplication alone, whereas ˆ q involves the spurious ∂/∂p term. If we restrict
to the pindependent subspace C
∞
q
(R
2
) · C
∞
(R), however, this diﬃculty is overcome: on
C
∞
q
(R
2
),
ρ(q) = m
q
ρ(p) = −i
∂
∂q
.
This association agrees with our earlier heuristic quantization of the classical position and
momentum observables and suggests in general that the space of sections of a prequantum
line bundle E is too “large” for the third Dirac axiom to be satisﬁed. Indeed, in quantum
mechanics, wave functions depend on only half of the phase space variables, whereas the
space of sections of E has the “size” of the space of functions on P. For an interpretation of
this argument in terms of the Heisenberg uncertainty principle, see [53].
From the standpoint of WKB quantization, the appropriate quantum state space asso
ciated with a classical conﬁguration space M is (the L
2
completion of) [Ω[
1/2
0
M, which we
temporarily identify with C
∞
(M) using a metric on M. Since the Liouville form α
M
vanishes
on each ﬁber of the projection T
∗
M
π
→M, we can interpret C
∞
(M) as the space of sections
of the prequantum line bundle E over T
∗
M which are parallel along each ﬁber of π. In other
98
words, the correct adjustment to the size of the prequantum Hilbert space is determined by
the (canonical) foliation of T
∗
M by lagrangian submanifolds.
This adjustment is generalized in the framework of geometric quantization by the intro
duction of the following concept.
Deﬁnition 7.4 A polarization of a symplectic manifold (P, ω) is an involutive lagrangian
subbundle T of T
C
P.
Here, T
C
P denotes the complexiﬁcation of the tangent bundle of P, equipped with the
complexvalued symplectic form ω
C
given by the complexlinear extension of ω to T
C
P. In
tegrability of T means that locally on P there exist complexvalued functions whose hamil
tonian vector ﬁelds (with respect to ω
C
) span T. The quantum state space associated to P
is then given by the space of sections s of a prequantum line bundle E over P which are
covariantly constant along T. That is, for each complex vector ﬁeld X on P lying in T, we
have ∇
X
s = 0.
Real polarizations
The standard polarization of a classical phase space T
∗
M is given by the complexiﬁcation
T = V M ⊕ iV M of the vertical subbundle of T(T
∗
M). In this case, the polarization T
satisﬁes T = T, meaning that T
p
is a totally real subspace of T
C,p
(T
∗
M) for each p ∈
T
∗
M. In general, any polarization T of a symplectic manifold P which satisﬁes T = T
is the complexiﬁcation of an integrable lagrangian subbundle of TP and is called a real
polarization of P.
Example 7.5 The most basic examples of real polarizations are given by either the planes
q=constant or p=constant in R
2n
. These correspond to the δ
q
and e
iπ/p,)/
bases of L
2
(R
n
)
respectively, in the sense that
f =
R
n
f(q) δ(q) dq = (2π)
−n/2
R
n
e
iπ/p,q)/
ˆ
f dp.
Covariant diﬀerentiation along vector ﬁelds X tangent to the q=constant polarization T
q
is given by the ordinary Lie derivative with respect to X, and so the space of T
q
parallel
sections of the prequantum line bundle E identiﬁes with the space of smooth functions on
qspace. For the p=constant polarization T
p
, we consider covariant derivatives of the form
∇ ∂
∂q
=
∂
∂q
−p
∂
∂σ
.
A T
p
parallel section ψ of E must therefore satisfy
∂ψ
∂q
−2πipψ
and is therefore of the form ψ(q, p) = v(p)e
iπ/p,q)/
. Note that the T
p
representation of the
freeparticle H(q, p) = p
2
/2 is ρ(H) = m
p
2
/2
, and so the ppolarization appears to be better
adapted to this case.
99
´
A real polarization T of a symplectic 2nmanifold P deﬁnes a subspace T(P) ⊂ C
∞
(P)
of functions constant along the leaves of T. According to the HamiltonJacobi theorem, the
hamiltonian vector ﬁeld of any member of T(P) is contained in T. Thus ¦f, g¦ = 0 for
any f, g ∈ T(P), i.e., T(P) is an abelian Poisson subalgebra of C
∞
(P). By modifying the
normal form results of Chapter 4, one can furthermore prove
Theorem 7.6 Every real polarization is locally isomorphic to the q=constant polarization
on R
2n
.
By pulling back the position functions q
i
to P, we have
Corollary 7.7 For each p ∈ P, there exist f
i
∈ T(P), i = 1, , n, such that the hamilto
nian vector ﬁelds X
i
span T in a neighborhood of p.
An aﬃne structure on a manifold M is a curvature and torsionfree connection on TM.
Since any polarizationpreserving symplectic transformation of T
∗
R
n
· R
2n
is aﬃne on ﬁbers
(see Theorem 3.29), we have the following result.
Corollary 7.8 Each leaf of any real polarization carries a natural aﬃne structure.
The main remark we wish to make for the moment is that our earlier prequantization ρ
−1
of C
∞
(P) now represents the subalgebra T(P) ⊂ C
∞
(P) as multiplication operators on H
T
.
Since any f ∈ T(P) has the property that its hamiltonian vector ﬁeld X
f
lies completely
within the polarization T, it follows that ∇
X
f
s = 0 for all s ∈ H
T
. By the deﬁnition of ρ
−1
,
this implies that ρ
−1
(f) = m
f
, the multiplication operator on H
T
. More generally, those
functions whose hamiltonian vector ﬁelds have ﬂows which leave the polarization invariant
(not necessarily leaf by leaf) are those which are “aﬃne” along the leaves.
Example 7.9 Following [28], we call a polarization T on P ﬁbrating if each leaf of T is
simplyconnected and the leaf space P
T
= P/T is a smooth manifold. In this case, the
quotient map p : P → P
T
satisﬁes T = ker p
∗
, and so p
∗
T
∗
P
T
identiﬁes with the normal
bundle T
⊥
⊂ T
∗
P. Since T is a lagrangian distribution, the map ˜ ω: TP →T
∗
P sends T to
T
⊥
as well, thus deﬁning a natural identiﬁcation
T · p
∗
T
∗
P
T
.
A function H: P →R constant on the leaves of T induces a function H
T
on P
T
satisfying
dH = p
∗
dH
T
.
From this remark, it follows that a section s of T over a leaf L is parallel with respect to
the connection described above if and only if (˜ ω ◦ s) deﬁnes a ﬁxed element of T
∗
L¦
P
T
.
Similarly, if E is a prequantum line bundle over P, we may use parallel sections of E over
the leaves of T to construct a foliation of the total space of E. The quotient of E by this
100
foliation deﬁnes a hermitian line bundle E
T
over the leaf space P
T
whose sections identify
with Tparallel sections of E.
The simplest example of a ﬁbrating polarization is given by the vertical polarization T
M
of a cotangent bundle T
∗
M. The basic geometry of this situation remains the same if the
symplectic structure on T
∗
M is perturbed by a form on M. More precisely, given a closed
2form η on M, we can “twist” the standard symplectic structure on T
∗
M as follows:
ω = ω
M
+ π
∗
η.
Since π
∗
η vanishes on the ﬁbers of the projection, ω is a symplectic form on T
∗
M and ker π
∗
is again a polarization. This general type of symplectic manifold is known as a twisted
cotangent bundle. A check of deﬁnitions shows that the standard and twisted symplectic
structures on T
∗
M are equivalent precisely when their diﬀerence is the pullback of an exact
form on M. Moreover, the symplectic manifold (T
∗
M, ω
M
+ π
∗
η) is prequantizable if and
only if the cohomology class of η in H
2
(M; R) is integral. Finally, an application of
Corollary 7.8 proves that if each leaf of a ﬁbrating polarization T on a symplectic manifold
(P, ω), is complete, then (P, ω) is symplectomorphic to a twisted symplectic structure on
T
∗
P
T
.
´
Complex polarizations
Associated to any polarization T are the distributions
D
C
= T ∩ T E
C
= T +T
which arise as the complexiﬁcations of distributions D, E in TP. Although their dimensions
vary in general from point to point, D and E are pointwise ωorthogonal, i.e.
D
⊥
x
= E
x
for all x ∈ P. From the deﬁnition of T, the distribution D is involutive. The polarization
T is called strongly admissible provided that E is involutive, the leaf spaces P
D
and P
E
are smooth manifolds, and the natural projection
P
D
π
→P
E
is a submersion. In this case, each ﬁber of π carries a K¨ahler structure, and geometric quan
tization attempts to construct a quantum state space for P from sections of an appropriate
line bundle over P which are parallel along D and holomorphic along the ﬁbers of π. For a
discussion of quantization in this general setting, we refer to [53].
A polarization satisfying T ∩T = ¦0¦ is called totally complex and identiﬁes with the
graph of a complex structure J on P, i.e.,
T = ¦(v, iJv) : v ∈ TP¦
101
under the identiﬁcation T
C
P · TP ⊕ iTP. We emphasize that J is a genuine complex
structure on P due to the integrability condition on the polarization T. Moreover, J is
compatible in the usual sense with the symplectic form ω on P, so the hermitian metric
' , ` = g
J
+ iω is a K¨ahler structure on P. In this section, we will only cite two examples
involving totally complex T, in which case P
D
= P is K¨ahler, and P
E
= ¦pt¦.
Example 7.10 Consider the complex plane C with its usual complex and symplectic struc
tures. With respect to Darboux coordinates (q, p), the CauchyRiemann operator is deﬁned
as
∂
∂z
=
1
2
∂
∂q
+ i
∂
∂p
.
A function f : C →C is holomorphic if ∂f/∂z = 0.
If we identify sections of the (trivial) prequantum line bundle E with smooth, complex
valued functions on C, then Tparallel sections correspond to those functions annihilated by
the covariant derivative
∇ ∂
∂z
= 2
∂
∂z
+ m
z
.
To determine the general form of these sections, we note that ∇ ∂
∂z
ψ = 0 if and only if for
some branch of the logarithm,
∂
∂z
log ψ = −
z
2
,
or
log ψ = −
zz
2
+ h
for some holomorphic function h. Hence
ψ(z) = ϕe
−[z[
2
/2
with ϕ holomorphic is the general form for Tparallel sections of E. The space H
C
thus
identiﬁes with the space of holomorphic ϕ: C →C satisfying
ϕ =
C
[ϕ(z)[
2
e
−[z[
2
dz < ∞,
or, in other words, the space of holomorphic functions which are squareintegrable with
respect to the measure e
−[z[
2
dz, known as the Fock or BargmanSegal space. For a func
tion to be quantizable in this picture, its hamiltonian ﬂow must preserve both the metric
and symplectic structure of C and therefore consist solely of euclidean motions. Among
such functions are the usual position and momentum observables, as well as the harmonic
oscillator.
´
Example 7.11 If (P, ω) is any prequantizable symplectic manifold with a totally complex
polarization T, then the prequantum line bundle E associated to ω can be given the structure
of a holomorphic line bundle by taking the (0, 1) component of the connection on E with
102
respect to the complex structure J arising from T. The space H
T
of Tparallel sections of E
then equals the space of holomorphic sections of E and is therefore completely determined
by the complex geometry of P.
Using some machinery from algebraic geometry (see [15]), it can be shown that for com
pact P, the space H
P
is ﬁnitedimensional and that its dimension is given asymptotically by
the symplectic volume of P. More precisely, we note that for k ∈ Z
+
, the line bundle E
⊗k
deﬁnes a holomorphic prequantum line bundle associated with (P, kω). For suﬃciently large
k, the dimension of the quantum state space H
P
k
of holomorphic sections of E
⊗k
is given by
the HirzebruchRiemannRoch formula:
dimH
P
k
=
P
e
kω
Td(P),
where Td denotes the Todd polynomial in the total Chern class of P. The ﬁrst conclusion
to be drawn from this fact is that for k large, dimH
P
k
is a symplectic invariant of P inde
pendent of its complex structure. Roughly speaking, k plays the role of
−1
, and so “large
k” means that we are approaching the classical limit. A second point to note is that dimH
P
k
is a polynomial in k (or
−1
) whose leadingorder term is
P
(kω)
n
n!
= k
n
vol P.
Consequently the number of quantum states is determined asymptotically by the volume of
P.
´
Metalinear structures and halfforms
For the remainder of this chapter, we will focus on the quantization of symplectic manifolds
P equipped with a prequantum line bundle E and a “suﬃciently nice” real polarization T.
Although the space H
T
of Tparallel sections of E appears to have the right “size” in the
simplest examples, there are still several problems to be resolved before we have a suitable
quantum state space. The ﬁrst arises as soon as we attempt to deﬁne a preHilbert space
structure on H
T
. On P, the square of an Tparallel section s is constant along the leaves of
T, and thus the integral
s
2
=
P
's, s` ω
n
diverges in general. On the other hand, there is no canonical measure on the leaf space P
T
with which to integrate the induced function 's, s`.
Regardless of how this ﬁrst diﬃculty is resolved, we will again try to quantize the Poisson
algebra C
∞
(P) by representing elements of Aut(P, ω) as (projective) unitary operators on
H
T
. The most obvious quantization of a symplectomorphism f : P → P (or more general
canonical relation), however, is an operator from H
T
to H
f(T)
. If f does not preserve the
polarization T, then these spaces are distinct. Thus, we will need some means for canonically
identifying the quantum state spaces H
T
associated to diﬀerent polarizations of P.
103
Finally, the polarization T may have multiplyconnected leaves, over most of which the
prequantum line bundle may admit only the trivial parallel section. A preliminary solution
to this problem is to admit “distributional” states, such as are given by Tparallel sections
of E whose restrictions to each leaf are smooth. Leaves on which E has trivial holonomy are
then said to comprise the BohrSommerfeld subvariety of the pair (P, T).
Example 7.12 Consider the punctured phase plane
˙
R
2
= R
2
` ¦0¦ polarized by level sets
of the harmonic oscillator H(q, p) = (q
2
+ p
2
)/2. The holonomy of the usual prequantum
line bundle on the level set H
−1
(E) is given by the modZ
reduction of the area it encloses;
thus the BohrSommerfeld variety consists of those circles of energy E = n. As in the
case of WKB prequantization, these energy levels do not correspond to actual physical
measurements, and so we will again need to incorporate a sort of Maslov correction.
´
The solution to manhy of these diﬃculties relies on the use of a metaplectic structure on
P, a concept which we now introduce. Let P be a principal G bundle over a manifold M.
A meta Gbundle associated to P and a central extension
1 →K →
˜
G
ρ
→G →1
of G is a principal
˜
G bundle
˜
P over M together with a map Φ :
˜
P → P satisfying the
equivariance condition
Φ(p a) = Φ(p) ρ(a)
for all a ∈
˜
G. Two meta Gbundles (
˜
P
1
, Φ
1
), (
˜
P
2
, Φ
2
) over P are considered equivalent if
there exists a
˜
Gequivariant diﬀeomorphism ψ:
˜
P
1
→
˜
P
2
such that Φ
1
= Φ
2
◦ ψ.
Example 7.13 A riemannian structure together with an orientation of an nmanifold M
deﬁnes a bundle of oriented orthonormal frames in TM, i.e. an SO(n)structure on M. A
spin structure on M is then a meta SO(n)bundle corresponding to the extension
1 →K →Spin(n) →SO(n) →1
given by the double cover Spin(n) of SO(n). An orientable manifold admits a spin structure
if and only if its second StiefelWhitney class vanishes.
´
Example 7.14 Since π
1
(Sp(n)) · Z, there exists a unique connected doublecovering group
of Sp(n), known as the metaplectic group Mp(n). A symplectic 2nplane bundle F admits
a reduction of its structure group from GL(2n) to Sp(n); a metaplectic structure on F
then corresponds to a lifting of the symplectic frame bundle to a principal Mp(n) bundle. In
general, a symplectic vector bundle admits a metaplectic structure if and only if its second
StiefelWhitney class vanishes.
A symplectic manifold P whose tangent bundle is equipped with a metaplectic structure
is called a metaplectic manifold. If the StiefelWhitney class w
2
(P) is zero, metaplectic
104
structures are classiﬁed by the set H
1
(P; Z
2
) (see [28]). We emphasize that this classiﬁcation
depends on the bundle of metaplectic frames and its covering map to the bundle of symplectic
frames. To see this explicitly in a special case, note that up to topological equivalence, the
punctured plane
˙
R
2
admits only trivial Sp(1) and Mp(1) principal bundles, since both Sp(1)
and Mp(1) are connected. On the other hand, a metaplectic structure
Φ:
˙
R
2
Mp(1) →
˙
R
2
Sp(1)
is deﬁned for any choice of continuous map s:
˙
R
2
→Sp(1) by
Φ(p, A) = (p, s(p) ρ(A))
for A ∈ Mp(1). Two such maps s
0
, s
1
deﬁne equivalent metaplectic structures if and only if
(s
1
)
−1
s
0
admits a continuous lift ˜ s:
˙
R
2
→Mp(1), in which case an equivalence is given by
the map ψ:
˙
R
2
Mp(1) →
˙
R
2
Mp(1) deﬁned as
ψ(p, A) = (p, ˜ s(p) A)
By the usual homotopy theory for continuous groups, this occurs precisely when [s
0
] = [s
1
]
as elements of π
1
(Sp(1))/ρ
#
π
1
(Mp(1)) · Z
2
.
´
Example 7.15 Lying within the metaplectic group is a doublecover Ml(n) of the subgroup
Gl(n) preserving the usual lagrangian splitting of R
2n
. The double covering of GL(n) then
induced by the identiﬁcation GL(n) · Gl(n) (see Example 3.1) is called the metalinear
group ML(n). It is trivial as a topological covering, but as a group it is not the direct
product GL(n) Z
2
. More explicitly, the group ML(n) is isomorphic to the direct product
GL
+
(n) Z
4
with the covering map ML(n)
ρ
→GL(n) given by
ρ(A, a) = A e
iπa
.
A metalinear lifting of the frames of an nplane bundle E is called a metalinear structure
on E. If E is orientable, its structure group can be reduced to GL
+
(n), which can be
interpreted as the identity component of ML(n) in order to give a metalinear structure on
E. More generally, a vector bundle admits a metalinear structure if and only if the square
of its ﬁrst StiefelWhitney class is zero (see [28]).
´
The importance of the metalinear group for our purposes is that it admits 1dimensional
representations which are not the lifts of representations of GL(n). Bundles associated to
metalinear structures via these representations will be the key to the metaplectic correction
of the prequantization procedure. First note that by means of the quotient homomorphism
ML(n) →ML(n)/GL
+
(n) · Z
4
,
105
we can associate a principal Z
4
bundle to any metalinear structure (MB(E), Φ) on a vector
bundle F. From the representation
a →e
iπa/2
of Z
4
in U(1), we then obtain a complex line bundle Λ
1/2
F called the bundle of halfforms
associated to the triple (F, MB(F), Φ). The reason for this terminology is that the bundle
Λ
1/2
F can be constructed directly from the metalinear frame bundle via the representation
ML(n) →C
∗
given by the squareroot of (det ◦ρ)
(A, a) →(det A)
1/2
e
iπa/2
.
A section of Λ
1/2
F then identiﬁes with a complexvalued function λ on MB(F) satisfying
λ(e) = (det A)
1/2
e
iπa/2
λ(e (A, a)),
and therefore represents, loosely speaking, the squareroot of a volume form on F. By
inverting and conjugating the preceding representation of ML(n), one similarly deﬁnes the
bundles Λ
1/2
F and Λ
−1/2
F of conjugate and negative halfforms on F. Evidently, the product
of two halfforms on F yields an nform; similarly, a halfform λ and a conjugate halfform
µ can be multiplied to give a 1density λµ on the bundle F.
An important link between metalinear and metaplectic structures on symplectic mani
folds can be described as follows.
Theorem 7.16 Let (P, ω) be a symplectic 2nmanifold and L ⊂ TP a lagrangian subbundle.
Then TP admits a metaplectic structure if and only if L admits a metalinear structure.
Proof. If J is any ωcompatible almost complex structure on P, then TP = L⊕JL · L⊕L.
By the Whitney product theorem, we then have
w
2
(P) = w
1
(L)
2
,
and our assertion follows from the remarks in the preceding examples.
More explicitly, we may use L and J to reduce the structure group of TP to the subgroup
Gl(n) ⊂ Sp(n) corresponding to GL(n). The resulting frame bundle is isomorphic to the
frame bundle of L, and thus, a metalinear structure on L can be enlarged to a metaplectic
structure on TP, while a metaplectic structure on TP reduces to a metalinear structure on
L.
2
Now let P be a metaplectic manifold equipped with a prequantum line bundle E and a
real polarization T. By the preceding theorem, T inherits a metalinear structure from the
metaplectic structure on TP, and thus Λ
−1/2
T is deﬁned; it is equipped with a natural ﬂat
connection inherited from the one on T.
Deﬁnition 7.17 The quantum state space H
T
associated to P is the space of sections of
E ⊗Λ
−1/2
T which are covariantly constant and along each leaf of T.
106
In the following examples, we will sketch how in certain cases this deﬁnition enables us to
overcome the diﬃculties mentioned at the beginning of this section.
Example 7.18 We ﬁrst return to Example 7.12 to illustrate how the introduction of half
forms also enables us to incorporate the BohrSommerfeld correction in the case of the
1dimensional harmonic oscillator.
To quantize the punctured plane
˙
R
2
with its polarization by circles centered at the origin,
we will use the trivial metaplectic structure on
˙
R
2
(see Example 7.14). In this case, both the
trivial Mp(1) and Sp(1) bundles over
˙
R
2
can be identiﬁed with the set of triples (r, θ, a e
iφ
),
where a, r > 0 and 0 ≤ θ, φ ≤ 2π, and the covering map
˙
R
2
Mp(1) →
˙
R
2
Sp(1) is given
by
(r, θ, a e
iφ
) →(r, θ, a e
2iφ
).
The pullback of
˙
R
2
Sp(1) of T
˙
R
2
to the circle S
r
of radius r can be reduced to the trivial
Gl(1) principal bundle over S
r
0
by means of the lagrangian splitting TS
r
0
⊕JTS
r
0
. In terms
of the identiﬁcations above, the subbundle
˙
R
2
Gl(1) equal the subset
¦(r, θ, a e
iθ
) : r = r
0
, a > 0, 0 ≤ θ ≤ 2π¦
of
˙
R
2
Sp(1) while its corresponding metalinear bundle equals the subset
¦(r, θ, a e
i(θ/2+cπ)
) : r = r
0
, a > 0, 0 ≤ θ ≤ 2π, c ∈ ¦0, 1¦¦
of
˙
R
2
Mp(1). Note that this is a nontrivial Ml(1) bundle over S
r
0
.
From the expression for the covering map given above, it follows easily that the parallel
transport of a negative halfform associated to TS
r
0
around S
r
0
amounts to multiplication
by −1. Consequently, the bundle E ⊗ Λ
−1/2
T admits a parallel section over S
r
0
precisely
when
πr
2
0
=
Sr
0
p dq = π(2n + 1)
for some integer n. That is, the BohrSommerfeld variety consists of circles of energy E =
(n + 1/2)π, in accordance with the corrected BohrSommerfeld quantization conditions
described in Chapter 4.
´
Example 7.19 Over each leaf L of the polarization T, the Bott connection on TL described
in Example 5.17 deﬁnes a foliation of the preimage of L in B(T). As proven in [28], the
quotient of B(T) by this leafwise foliation deﬁnes a metalinear structure on the leaf space
P
T
in such a way that halfforms on P
T
are in 11 correspondence with Tparallel negative
halfforms on T. In this way, the quantum state space H
T
identiﬁes with the space of
compactly supported sections of
E
T
⊗Λ
1/2
P
T
.
The latter space is equipped with a natural innerproduct
's
1
⊗λ
1
, s
2
⊗λ
2
` =
P
F
's
1
, s
2
` λ
1
λ
2
.
Note that integration is welldeﬁned, since λ
1
λ
2
is a density on P
T
.
107
´
BlattnerKostantSternberg kernels
If T
1
and T
2
are ﬁbrating polarizations of P with the property that T
1,x
is transverse to
T
2,x
at each x ∈ P, then the quantum state spaces H
T
1
and H
T
2
can be related as follows.
By assumption, the symplectic form deﬁnes an isomorphism of T
1
with T
∗
2
; similarly, the
metaplectic structure on P deﬁnes a natural isomorphism between the space of halfforms
on T
2
and the space of conjugate halfforms on T
1
(see [28]). Using this identiﬁcation, we
can pair a halfform λ
1
on T
1
with a halfform λ
2
on T
2
to obtain a function (λ
1
, λ
2
) on P.
If we identify elements σ
1
, σ
2
of H
T
1
and H
T
2
with T
i
parallel sections s
i
⊗λ
i
of E ⊗Λ
1/2
T
i
on P, then a sesquilinear pairing H
T
1
H
T
2
→C is deﬁned by
''σ
1
, σ
2
`` =
1
(2π)
n/2
P
's
1
, s
2
` (λ
1
, λ
2
) ω
n
,
where dim(P) = 2n. This pairing is nondegenerate and thus determines a linear operator
B: H
T
2
→H
T
1
satisfying
''σ
1
, σ
2
`` = 'σ
1
, Bσ
2
`
for all σ
1
∈ H
1
and σ
2
∈ H
2
. This construction, which can be extended to more general pairs
of polarizations, is due to Blattner, Kostant, and Sternberg deﬁnes a sesquilinear pairing on
the associated quantum state spaces via a BlattnerKostantSternberg kernel.
Example 7.20 The basic example of a BlattnerKostantSternberg kernel arises from the
classical Fourier transform. From Example 7.5, we recall that for the q =constant polariza
tion T
q
of R
2n
, the space H
Tq
consists of pindependent functions σ
q
(q, p) = u(q) on R
2n
,
which we may identify with the space of smooth functions on qspace. Similarly, if T
p
is the
p =constant polarization, then a smooth function v(p) on pspace identiﬁes with an element
σ
p
(q, p) = v(p)e
i/q,p)/
of H
Tp
. The sesquilinear pairing described above is therefore given by
''σ
q
, σ
p
`` =
1
(2π)
n/2
R
2n
e
i/q,p)/
u(q) v(p) [dq dp[.
Note that this corresponds to the usual association of a distribution ˆ v(q) on qspace to a
function v(p) on pspace via the inverse asymptotic Fourier transform.
´
Given a symplectic manifold P with the structures above, geometric quantization at
tempts to represent the Poisson algebra C
∞
(P) on H
T
as follows. First, if the hamiltonian
ﬂow ϕ
t
of f ∈ C
∞
(P) preserves T, then it lifts to a 1parameter family ˜ ϕ
t
of operators on
smooth sections of E ⊗Λ
1/2
T. The quantum operator associated to T is then
ρ(f) σ = i
d
dt
( ˜ ϕ
t
σ)[
t=0
.
In particular, ρ(f) acts as pointwise multiplication for any f ∈ T(P).
108
For those f ∈ C
∞
(P) whose hamiltonian ﬂow does not preserve T, we assume that
for t ∈ (0, ε), the diﬀerential Dϕ
t
maps T into a polarization T
t
for which there exists a
BlattnerKostantSternberg kernel and a corresponding operator U
t
: H
Tt
→ H
T
. The ﬂow
ϕ
t
deﬁnes an operator ˜ ϕ
t
: H
T
→H
Tt
, and the quantum operator ρ(f) is deﬁned as
ρ(f) σ = i
d
dt
((U
t
◦ ˜ ϕ
t
) σ)[
t=0
.
We refer to [28] for further details.
7.3 Quantization of semiclassical states
In this brief section, we suppose that P is a metaplectic manifold with a prequantum line
bundle E and a ﬁbrating polarization T. Our goal is to comment brieﬂy on how an ele
ment of H
T
can be constructed from a “semiclassical state” in P consisting of a lagrangian
submanifold L ⊂ P with some extra structure, such as a halfdensity or halfform with
coeﬃcients in a prequantum line bundle over P.
Example 7.21 If L ⊂ P intersects each leaf of T transversely in at most one point, then
the section s of E⊗Λ
−1/2
T over L corresponding to a section of E⊗[Λ[
1/2
L can be extended
to a section ˜ s over P which is covariantly constant on each leaf and which vanishes on leaves
which are disjoint from L. In this way, we can regard ˜ s as the quantization of the pair (L, s).
We can generalize this viewpoint by considering a lagrangian submanifold L which has
possibly multiple transverse intersections with the leaves of T (that is, each intersection of
L with a leaf of T is transverse). If s is again a section of E ⊗ Λ
−1/2
T over L, then we
can construct and element of H
T
by “superposition”. More precisely, we think of L as the
union of lagrangian submanifolds L
j
, such that each L
j
intersects any leaf of T at most
once. Applying the procedure above to each (L
j
, s[
L
j
) and summing the results produces an
element of H
T
as long as this sum converges.
Finally, if L is a union of leaves of T and s is a section of E ⊗ Λ
−1/2
T over L which is
covariantly constant along each leaf of T, then we can pair it with arbitrary elements of H
T
by integration over L, thus obtaining a linear functional on H
T
, which, via the inner product
on H
T
, can be considered as a generalized section.
´
Example 7.22 To indicate brieﬂy how “distributional” elements of H
T
arising as in the
preceding example are paired, we consider the following situation. Let (L, s) be a semi
classical state in R
2n
such that L equals the zero section of R
2n
· T
∗
R
n
and s is any section
of E⊗Λ
−1/2
T
q
over L. Similarly, let
˜
L be the ﬁber of T
∗
R
n
over 0 ∈ R
n
along with a constant
section ˜ s of E ⊗ Λ
−1/2
T
q
over
˜
L. According to the preceding example, the quantization of
(L, s) is given by an ordinary smooth function on R
n
, whereas the quantization of (
˜
L, ˜ s) is
“concentrated” at the origin of R
n
.
The meaning of this last statement is made more precise by the fact that if T : R
2n
→R
2n
is a linear symplectomorphism, then the associated quantum operator H
Tq
→ H
T(Tq)
is
109
required to be unitary. By choosing T so that both L and
˜
L are transverse to T(T
q
), we can
quantize (L, s) and (
˜
L, ˜ s) with respect to this new polarization to obtain smooth functions
on the leaf space R
2n
T(Tq)
whose pairing is deﬁned by integration. In terms of H
Tq
, this means
that, up to a normalization, the quantum state obtained from (
˜
L, ˜ s) is a Dirac δfunction
concentrated at the origin of R
n
.
´
110
8 Algebraic Quantization
An approach to quantization going back to Dirac and recently revived in [9][10][11] is based
on the idea that the multiplicative structure of the ∗algebra of quantum observables is more
central to quantization than the representation of observables as operators on Hilbert space,
and that most of quantum mechanics can be done without regards for the precise nature of
observables. Quantum observables comprise a noncommutative algebra / which is assumed
to belong to a family /
of algebras, all with a common underlying vector space, but with
an dependent multiplication ∗
. As → 0, the algebra /
approaches a commutative
algebra /
0
, which may be interpreted as the algebra of functions on a classical phase space.
Additionally, it is assumed that the ∗
commutator approaches the Poisson bracket in the
limit of large quantum numbers:
¦f, g¦ = lim
→0
(f ∗
g −g ∗
f)/i.
The abstract goal of algebraic quantization is to construct the family /
of noncommutative
algebras from a Poisson algebra. In this chapter, we sketch two approaches towards this
goal, known as deformation quantization and the method of symplectic groupoids.
8.1 Poisson algebras and Poisson manifolds
The main objects in algebraic quantization theory are Poisson algebras and Poisson mani
folds.
Deﬁnition 8.1 A Poisson algebra is a real vector space / equipped with a commutative,
associative algebra structure
(f, g) →fg
and a Lie algebra structure
(f, g) →¦f, g¦
which satisfy the compatibility condition
¦fg, h¦ = f ¦g, h¦ +¦f, h¦ g.
A Poisson manifold is a manifold P whose function space C
∞
(P) is a Poisson algebra
with respect to the usual pointwise multiplication of functions and a prescribed Lie algebra
structure.
If P, Q are Poisson manifolds, a smooth map ψ: P →Q is called a Poisson map provided
that it preserves Poisson brackets, i.e. ¦f, g¦ = ¦f ◦ψ, g ◦ψ¦ for all f, g ∈ C
∞
(Q). Similarly,
ψ is called an antiPoisson map if ¦f, g¦ = −¦f ◦ ψ, g ◦ ψ¦ for all f, g ∈ C
∞
(Q).
On a Poisson manifold, the Leibniz identity implies that the Poisson bracket is given by
a skewsymmetric contravariant tensor ﬁeld π via the formula
¦f, g¦ = π(df, dg).
The main examples of Poisson manifolds we will be concerned with are the following.
111
Example 8.2 1. Any smooth manifold M is a Poisson manifold when C
∞
(M) is given the
trivial bracket ¦f, g¦ = 0.
2. Any symplectic manifold is a Poisson manifold with respect to its standard Poisson
structure
¦f, g¦ = X
g
f.
3. Let g be a ﬁnitedimensional Lie algebra with dual g
∗
. The diﬀerential of a smooth
function F : g
∗
g
∗
→ R is a map DF : g
∗
g
∗
→ g g whose composition with the Lie
bracket on g deﬁnes a smooth map [DF] : g
∗
g
∗
→ g. The LiePoisson operator on
C
∞
(g
∗
g
∗
) is the diﬀerential operator T deﬁned by
TF =
1
2
'[DF]
(x,y)
, x + y`.
The dual space g
∗
is then a Poisson manifold when equipped with the LiePoisson bracket
¦f, g¦ = ∆
∗
T(f g),
where ∆: g
∗
→g
∗
g
∗
is the diagonal and f, g ∈ C
∞
(g
∗
). More concretely,
¦f, g¦(µ) = µ([Df, Dg]),
where µ ∈ g
∗
, and Df, Dg : g
∗
→g
∗∗
· g are the diﬀerentials of f and g.
4. If V is a ﬁnitedimensional vector space and π is a skewsymmetric bilinear form on V
∗
,
then
¦f, g¦
def
= π(df, dg)
deﬁnes a Poisson algebra structure on C
∞
(V ), making V a Poisson manifold.
´
8.2 Deformation quantization
The aim of deformation quantization is to describe the family ∗
of quantum products on a
Poisson algebra / as an asymptotic series (in ) of products on /. In accordance with the
introductory remarks above, the zeroth and ﬁrst order terms of this series are determined
by the Poisson algebra structure of /; the role of the higherorder terms is roughly speaking
to give a more precise dependent path from classical to quantum mechanics.
Deﬁnition 8.3 Let / be a complex vector space equipped with a commutative associative
algebra structure, and let ∗
be a family of associative multiplications on / given by a formal
power series
f ∗
g =
∞
¸
j=0
B
j
(f, g)
j
where each B
j
: // →/ is a bilinear map. Then ∗
is called a ∗deformation of / if
112
1. B
0
(f, g) equals the product in /.
2. B
j
(g, f) = (−1)
j
B(f, g)
3. B
j
(1, f) = 0 for j ≥ 1
4. B
j
is a diﬀerential operator in each argument.
5. (f ∗
g) ∗
h = f ∗
(g ∗
h)
For condition (5) to make sense, we must extend the product ∗
in the obvious way from /
to the space /[[]] of formal power series. We emphasize that the power series above is not
in general assumed to converge for = 0. Instead it should be viewed as an “asymptotic
expansion” for the product, and manipulations of this series will be purely formal.
In any case, conditions (15) imply that
¦f, g¦
def
=
1
2i
B
1
(f, g)
deﬁnes a Poisson algebra structure on /. This observation suggests the question of whether
every Poisson structure on a function algebra C
∞
(P) can be realized as the ﬁrst order term
in a ∗deformation of /.
Example 8.4 If V is a ﬁnitedimensional real vector space and π : V
∗
V
∗
→ R is a
skewsymmetric bilinear form on V
∗
, then the Hessian of π at 0 ∈ V
∗
V
∗
is a linear map
A = Hπ : V
∗
V
∗
→ V V . We deﬁne the Poisson operator associated to π as the
secondorder diﬀerential operator on C
∞
(V V )
T
π
= 'A(∂/∂y
∗
, ∂/∂z
∗
), (∂/∂y, ∂/∂z)`,
where (y, z) are linear coordinates on V V arising from a single set of linear coordinates on
V , and (y
∗
, z
∗
) are dual to (y, z). The MoyalWeyl operator is then the pseudodiﬀerential
operator given by exponentiation:
´
π,
= e
−iTπ/2
.
Pointwise multiplication of functions in f, g ∈ C
∞
(R
2n
) can be deﬁned as the pullback
f g = ∆
∗
f g,
where ∆ : R
2n
→ R
2n
R
2n
is the diagonal embedding, and (f g)(y, z) = f(y) g(z) for
f, g ∈ C
∞
(R
2n
). Similarly, a straightforward computation shows that the Poisson operator
is related to the Poisson bracket induced by π via the equation
¦f, g¦ = ∆
∗
T(f g).
The diagonal pullback
f ∗
g = ∆
∗
´
(f g)
113
of the Moyal operator deﬁnes a ∗deformation of the Poisson algebra C
∞
(V ) called the
MoyalWeyl product. In terms of the linear coordinates (y, z) on V V , the operator B
j
in the expansion of the MoyalWeyl product is
B
j
(f, g) =
1
j!
i
2
¸
r,s
π
r,s
∂
∂y
r
∂
∂z
s
j
f(y)g(z)
y=z=x
.
If π is nondegenerate, then an application of the principle of stationary phase shows that
an integral expression for f ∗
g is given by
(f ∗
g)(x) =
e
iQ(y−x,z−x)
f(y) g(z) dy dz,
where Q is the skewsymmetric bilinear form on V induced by π.
´
A Poisson manifold P is called regular if its Poisson tensor π has constant rank, in
which case a theorem of Lie [39] asserts that P is locally isomorphic to a vector space with a
constant Poisson structure. In view of Example 8.4, any regular Poisson manifold is locally
deformation quantizable. To construct a ∗product on all of C
∞
(P), one may therefore
attempt to “patch together” the local deformations to arrive at a global ∗product.
This technique has succeeded. A theorem of DeWilde and Lecomte [19] asserts that the
Poisson algebra of any ﬁnitedimensional symplectic manifold admits a ∗deformation. Using
similar techniques, M´elotte [44] extended their result to arbitrary regular Poisson manifolds.
A simpliﬁed proof of these results has recently been given by Fedosov [23] (see [68] for a
survey of deformation quantization, emphasizing Fedosov’s construction).
8.3 Symplectic groupoids
The method of symplectic groupoids also attempts to directly construct a noncommutative
algebra /
of quantum observables without explicitly identifying a quantum state space.
Unlike deformation quantization, however, this approach involves a geometric procedure
which attempts to construct /
for a particular value of and in particular incorporates
geometric objects with certain quantum properties.
Groupoids
In this section we collect some basic deﬁnitions and examples of groupoids and their coun
terparts in symplectic geometry.
Deﬁnition 8.5 A groupoid is a set Γ endowed with a product map (x, y) → xy deﬁned
on a subset Γ
2
⊂ Γ Γ called the set of composable pairs, and an inverse map ι : Γ → Γ
satisfying the conditions
1. ι
2
= id
114
2. If (x, y), (y, z) ∈ Γ
2
, then (xy, z), (x, yz) ∈ Γ
2
, and (xy)z = x(yz).
3. (ι(x), x) ∈ Γ
2
for all x ∈ Γ, and if (x, y) ∈ Γ
2
, then ι(x) (xy) = y.
4. (x, ι(x)) ∈ Γ
2
for all x ∈ Γ, and if (z, x) ∈ Γ
2
, then (zx)ι(x) = z.
Note that by (3), the map ι is bijective and thus inverses in Γ are unique. An element x
of Γ can be thought of as an arrow with source α(x) = ι(x)x and target β(x) = xι(x); a
pair (x, y) then belongs to Γ
2
if and only if the source of y equals the target of x. The set
Γ
0
of all sources (and targets) is called the base of Γ, and Γ is said to be a groupoid over
Γ
0
. Elements of Γ
0
are units of Γ in the sense that xα(x) = x and β(x)x = x for all x ∈ Γ.
Finally, the multiplication relation of Γ is the subset
m = ¦(xy, x, y) : (x, y) ∈ Γ
2
¦
of Γ Γ Γ. In abstract terms, a groupoid is a small category in which all morphisms have
inverses.
Example 8.6 1. Any group is a groupoid over its identity element, and conversely, any
groupoid whose base is a singleton comprises a group.
2. A disjoint union of groupoids is a groupoid over the union of their bases. If Γ is a groupoid
with base Γ
0
and Γ
t
0
⊂ Γ
0
, then Γ
t
= ¦x ∈ Γ : α(x), β(x) ∈ Γ
t
0
¦ is a groupoid over Γ
t
0
.
3. Combining (1) and (2), we see that any vector bundle E deﬁnes a groupoid Γ(E) over its
zero section.
4. The pair groupoid associated to a set X consists of Γ = X X, endowed with the
multiplication (x, y) (y, z) = (x, z). Thus Γ
0
is the diagonal, and α, β are the projections
α(x, y) = (y, y) and β(x, y) = (x, x). In this groupoid, there is exactly one arrow from any
object to another.
´
We will be interested in groupoids with some geometric structures. Maintaining the notation
above, we make the following deﬁnition.
Deﬁnition 8.7 A groupoid Γ is called a Lie groupoid if
1. Γ
0
is a submanifold of Γ.
2. The mappings α, β : Γ →Γ
0
are submersions.
3. Multiplication Γ
2
→Γ and inversion Γ
ι
→Γ are smooth.
Condition (2) implies that the map α β is transverse to the diagonal ∆ in Γ
0
Γ
0
, and so
both Γ
2
= (α β)
−1
(∆) ⊂ Γ Γ and the multiplication relation m ⊂ Γ Γ Γ are smooth
submanifolds.
A submanifold L of Γ is called unitary if the restriction of α and β to L are diﬀeomor
phisms L →Γ
0
. The unitary submanifolds form a group under the natural multiplication of
subsets.
115
Deﬁnition 8.8 A Lie groupoid Γ is called a symplectic groupoid if Γ is a symplectic
manifold and the multiplication relation m is a lagrangian submanifold of Γ Γ Γ.
Two immediate consequences of this deﬁnition and the calculus of canonical relations are that
Γ
0
is lagrangian, and ι : Γ →Γ is antisymplectic, i.e., its graph is a lagrangian submanifold
of Γ Γ. Thus, a symplectic groupoid Γ is characterized by the three canonical relations
(recall that Z is a point):
Γ
0
∈ Hom(Z, Γ) L
ι
∈ Hom(Γ, Γ) m ∈ Hom(Γ Γ, Γ)
linked by the equation
Γ
0
= m◦ L
ι
.
As a consequence of the axioms and the assumption that m is lagrangian, there is a unique
Poisson structure on the base Γ
0
of a symplectic groupoid Γ such that α: Γ →Γ
0
is a Poisson
map and β : Γ → Γ
0
is antiPoisson. This Poisson structure gives meaning to the following
concept.
Deﬁnition 8.9 A symplectic groupoid Γ is said to integrate a Poisson manifold P if there
exists a Poisson isomorphism from the base Γ
0
of Γ onto P.
If there exists a groupoid Γ which integrates P, we will say that P is integrable and refer
to Γ as a symplectic groupoid over P.
Example 8.10 1. A simple example of a Lie groupoid is given by any Lie group. From
the requirement that the base of a symplectic groupoid be a lagrangian submanfold, only
discrete Lie groups can be symplectic groupoids.
2. If M is any manifold with the zero Poisson structure, then T
∗
M, equipped with the group
oid structure of a vector bundle and its standard symplectic structure, deﬁnes a symplectic
groupoid over M.
3. If P is any symplectic manifold, then the pair groupoid structure on P P deﬁnes a
symplectic groupoid over P. Its unitary lagrangian submanifolds are precisely the graphs of
symplectomorphisms of P.
4. If g is any Lie algebra, then the dual space g
∗
with its LiePoisson structure is integrable.
For any Lie group G whose Lie algebra is g, we deﬁne a groupoid structure on T
∗
G by taking
α and β to be the right and left translations of covectors to the ﬁber at the identity e ∈ G..
A simple computation shows that the base of T
∗
G is its ﬁber at the identity, while L
ι
and m
identify with the conormal bundles to the inversion and multiplication relations of G under
the identiﬁcations T
∗
GT
∗
G · T
∗
(GG) T
∗
GT
∗
GT
∗
G · T
∗
(GGG). Equipped
with these operations and its usual symplectic structure, T
∗
G is a symplectic groupoid over
g
∗
.
´
116
Quantization via symplectic groupoids
From our discussion of geometric quantization, we know that certain symplectic manifolds
quantize to give vector spaces V , and lagrangian submanifolds correspond to elements in V .
If we wish V to be an associative ∗algebra with unit element, like the algebras of quan
tum mechanics, then the underlying symplectic manifold must possess a groupoid structure
compatible with its symplectic structure.
The ﬁrst step in the quantization of a Poisson manifold P by the method of symplectic
groupoids is to construct a symplectic groupoid (Γ, Γ
0
, ι, m) over P. This is known as the
integration problem for Poisson manifolds. Using the techniques of geometric quantization
(prequantizations, polarizations), we then attempt to associate a vector space /
Γ
to Γ in
such a way that the canonical relations Γ
0
, L
ι
, and m quantize as elements of /
Γ
, /
Γ
⊗/
∗
Γ
,
and /
Γ
⊗/
∗
Γ
⊗/
∗
Γ
which deﬁne the structure of an associative ∗algebra on /
Γ
. If successful,
it is in this sense that the symplectic groupoid Γ represents a classical model for the quantum
algebra /
Γ
.
To conclude this chapter, we will give several examples illustrating the spirit of the
symplectic groupoid method. Throughout, we will only deal with groupoids Γ equipped
with reasonable (e.g. ﬁbrating) polarizations, so that the construction of the vector space
/
Γ
follows unambiguously from the procedure deﬁned in the preceding chapter.
Example 8.11 Consider a trivial Poisson manifold M and its associated symplectic group
oid T
∗
M. As described in the preceding example, the identity relation Γ
0
coincides with
the zero section of T
∗
M; under the identiﬁcations T
∗
M T
∗
M · T
∗
(M M) and T
∗
M
T
∗
MT
∗
M · T
∗
(MMM), the relations L
ι
and m identify with the conormal bundles
of the diagonals ∆
2
⊂ M M and ∆
3
⊂ M M M respectively.
As in Chapter 7, we may identify the quantum Hilbert spaces associated to T
∗
M, T
∗
(M
M), and T
∗
(MMM) with (completions of) the function spaces C
∞
(M), C
∞
(MM) and
C
∞
(MMM). The relation Γ
0
then quantizes as the function 1 on M. Since L
ι
and m are
the conormal bundles of the diagonals in MM and MMM, respectively, our heuristic
discussion in Section 7.3 shows that, after an appropriate normalization, these relations are
quantized by the δfunctions δ(x, y) and δ(x, z)δ(y, z) supported on the diagonals ∆
2
and
∆
3
respectively. Thus, the quantization of the groupoid Γ yields the usual identity element,
complex conjugation, and pointwise multiplication in the associative ∗algebra C
∞
(M, C).
´
Example 8.12 Arguing as in the preceding example, we ﬁnd that the quantization of the
canonical relations Γ
0
, L
ι
, and m associated to the groupoid T
∗
G of Example yields the
distributions δ(e) on G, δ(g
1
, g
−1
1
) on GG, and δ(g
1
g
2
, g
1
, g
2
) on GGG. If the Haar
measure on G is used to identify the quantum Hilbert space H
G
with C
∞
(G, C), then the
relations Γ
0
, L
ι
and m quantize as evaluation (at e ∈ G), antiinvolution f(g) →f(g
−1
), and
convolution.
There is actually a ﬂaw in the preceding two examples, since geometric quantizaton
produces halfdensities rather than functions, and furthermore, the natural domain of the
117
convolution operation on a group consists not of functions but of densities. It appears then
that the construction of V itself, and not just the multiplication, should depend on the
groupoid structure on Γ.
Although the two multiplications associated with the two groupoid structures on T
∗
G
described above live on diﬀerent spaces, it is possible to relate them more closely by dualizing
one of them, say convolution. In this way, we obtain the coproduct ∆ on C
∞
(G) deﬁned as
a map from C
∞
(G) to C
∞
(G) ⊗C
∞
(G) = C
∞
(GG) by the formula (∆f)(x, y) = f(xy).
The coproduct satisﬁes a coassociative law and is related to pointwise multiplication by the
simple identity ∆(fg) = ∆(f)∆(g), making C
∞
(G) into a Hopf algebra. The compatibility
between the two structures on C
∞
(G) reﬂects a compatibility between the two groupoid
structures on T
∗
G.
´
Example 8.13 If we take P = R
2n
with its standard symplectic structure, there is a po
larization of P P which does not depend on any polarization of P but only on the aﬃne
structure. In fact, the polarization comes from the isomorphism of Γ = R
2n
R
2n
with
T
∗
R
2n
given by
(x, y) →((x + y)/2, ˜ ω
n
(y −x)).
Using this polarization to quantize Γ, we get as V the space of smooth functions on the
diagonal, which we identify with R
2n
itself, and the multiplication on V turns out to be the
Moyal product (see Example 8.4).
´
Example 8.14 If V is any ﬁnitedimensional vector space, then a skewadjoint linear map
π : V
∗
→ V deﬁnes a skewsymmetric bilinear form on V
∗
and thus a translationinvariant
Poisson structure on V . If T is a torus equal to the quotient of V by some lattice, then
T inherits a translationinvariant Poisson structure from V . A symplectic groupoid which
integrates T is given by the cotangent bundle T
∗
T, with Γ
0
equals the zero section of T
∗
T,
and L
ι
equals the conormal bundle of the diagonal in T T. To describe the multiplication
relation, we identify T
∗
T with T V
∗
and let m ⊂ T
∗
(T T T) consist of all triples
(q, q
t
, q
tt
, p, p
t
, p
tt
) such that q
tt
= q
t
+
1
2
T(p
t
+ p
tt
) and
q = q
t
+
1
2
T(p
tt
) p = p
t
+ p
tt
.
When the map T is zero, the Poisson structure on T is trivial, and the groupoid product
quantized to the usual pointwise multiplication of functions on T. Otherwise one gets a
noncommutative multiplication on C
∞
(T) which is precisely that of (the functions on) a
noncommutative torus, one of the basic examples of noncommutative geometry. (The
relation between Poisson tori and noncommutative tori was studied from the point of view
of deformation quantization in [50]).
´
118
A Densities
An nform ν on an ndimensional manifold M can be viewed as a scalar function on the
space of bases in the tangent bundle which satisﬁes
ν(eA) = ν(e) det(A),
where e = (e
1
, , e
n
) is any frame in a tangent space of M and A = (a
ij
) is any invertible
n n matrix. Because the “change of variables” formula for integration involves absolute
values of Jacobians, integration of nforms on M requires a choice of orientation. The use of
densities instead of forms circumvents this need.
A density on a real vector space V of dimension n is a complexvalued function η, deﬁned
on the set B(V ) of bases in V , which satisﬁes η(eA) = η(e) [ det(A)[. The collection of such
functions is denoted [Λ[V . This concept can be generalized as follows.
Deﬁnition A.1 For α ∈ C, an αdensity on V is a map λ : B(V ) →C such that
λ(eA) = λ(e) [ det(A)[
α
.
We denote the vector space of αdensities on V by [Λ[
α
V. Since GL(V ) acts transitively on
B(V ), an αdensity is determined by its value on a single basis. As a result, [Λ[
α
V is a
1dimensional complex vector space.
Operations on densities
1. If σ ∈ [Λ[
α
V and β ∈ C, then σ
β
is a welldeﬁned αβdensity on V .
2. A linear map T : V →V
∗
induces a realvalued 1density T on V given by
T e
def
= [ det('Te
i
, e
j
`)[
1/2
.
Equivalently, any real bilinear form ω on V induces a 1density ˜ ω on V .
3. Multiplication of densities is deﬁned by multiplication of their values and gives rise to a
bilinear map:
[Λ[
α
V [Λ[
β
V →[Λ[
α+β
V.
4. If W is a subspace of V , then a basis of V , unique up to transformation by a matrix of
determinant ±1, is determined by a choice of bases for W and V/W. Consequently, there is
a natural product
[Λ[
α
W [Λ[
α
(V/W) →[Λ[
α
V
In particular, if V = V
1
⊕V
2
, then there is a natural product
[Λ[
α
V
1
[Λ[
α
V
2
→[Λ[
α
V.
119
5. Operations 3 and 4 induce natural isomorphisms
[Λ[
α
V ⊗[Λ[
β
V →[Λ[
α+β
V
[Λ[
α
W ⊗[Λ[
α
(V/W) →[Λ[
α
V.
[Λ[
α
V
1
⊗[Λ[
α
V
2
→[Λ[
α
(V
1
⊕V
2
)
6. If T : V →V
t
is an isomorphism, there is a welldeﬁned isomorphism
T
∗
: [Λ[
α
V →[Λ[
α
V
t
.
7. A natural map B(V )
∗
→ B(V
∗
) is deﬁned by associating to each basis e of V its dual
basis e
∗
. Since (eA)
∗
= e
∗
(A
∗
)
−1
for any A ∈ GL(V ), the map ∗ gives rise to a natural
isomorphism
[Λ[
α
V →[Λ[
−α
V
∗
.
Example A.2 Suppose that A, C are vector spaces and
0 →A →A ⊕C
F
→C
∗
→0
is an exact sequence such that F[
C
= T. By the operations described above, we obtain an
isomorphism
[Λ[
α
A · [Λ[
α
A ⊗[Λ[
2α
C
given explicitly by
σ →σ ⊗T
2α
= [ det
θ
T[
α
σ ⊗θ
2α
,
where θ is any realvalued 1density on C, and the positive real number [ det
θ
T[ is deﬁned
by the equation
[ det
θ
T[
1/2
θ = T.
´
It is easy to check that the association V →[Λ[
α
V deﬁnes a diﬀerentiable functor, so we
can associate the αdensity bundle [Λ[
α
E to any smooth vector bundle E over a manifold
M. The remarks above imply that if
0 →E →F →G →0
is an exact sequence of vector bundles over M, then there is a natural densitybundle iso
morphism
[Λ[
α
E ⊗[Λ[
α
G · [Λ[
α
F.
We denote by [Ω[
α
E the vector space of smooth sections of [Λ[
α
E and by [Ω[
α
c
E the space of
smooth, compactlysupported sections. The density spaces associated to the tangent bundle
of M are denoted [Ω[
α
M. An nform on an nmanifold M induces an αdensity [ν[
α
in the
manner of (4) above.
120
A natural mapping [Ω[
1
c
M →C is deﬁned by integration
σ →
M
σ.
Similarly, a preHilbert space structure on the space [Ω[
1/2
c
M of smooth, compactlysupported
halfdensities on M is deﬁned by
'σ, τ` =
M
σ τ.
The completion H
M
of this space is called the intrinsic Hilbert space of M.
B The method of stationary phase
We begin with a few useful formulas involving the asymptotic Fourier transform. Let o
denote the usual Schwartz space of rapidly decreasing, complex valued functions on R
n
. For
u ∈ o, the asymptotic Fourier transform and its inverse are deﬁned respectively as
(T
u)(ξ) = (2π)
−
n
2
R
n
e
−i/x,ξ)/
u(x) dx
(T
−1
v)(x) = (2π)
−
n
2
(R
n
)
∗
e
i/x,ξ)/
v(ξ) dξ
To see that these transforms are actually inverse to one another, note that the change of
variables η = ξ/ gives
(2π)
−
n
2
(R
n
)
∗
e
i/x,ξ)/
(T
u)(ξ) dξ = (2π)
−n
e
i/x−y,η)
u(y) dy dη,
which equals u(x) by the usual Fourier inversion formula. This observation also veriﬁes the
asymptotic inversion formula:
u(x) = (2π)
−n
e
i/x−y,ξ)/
u(y) dy dξ.
A simple application of this formula shows that the asymptotic diﬀerential operator D
j
=
−i∂
j
satisﬁes the familiar equations
T
(D
j
u) = ξ
j
T
u T
(x
j
u) = −D
j
T
u.
A similar check of deﬁnitions proves the asymptotic Parseval formula:
R
n
uv dx =
(R
n
)
∗
T
uT
v dξ.
We study next the asymptotic behavior of integrals of the form
I
=
R
n
e
iR(x)/
a(x) [dx[ a ∈ C
∞
0
(R
n
), R ∈ C
∞
(R
n
)
as →0. As a ﬁrst step, we will prove that if the critical point set of R is not contained in
the support of a, then I
is rapidly decreasing in :
121
Lemma B.1 If dR = 0 on Supp(a), then I
= O(
∞
) as →0.
Proof. Suppose for the moment that R
x
1
= ∂R/∂x
1
= 0 on Supp(a). Then
e
iR/
a = −i
a
R
x
1
∂
∂x
1
e
iR/
,
and so integration by parts with respect to the x
1
variable gives
[I
[ =
R
n
e
iR/
∂
∂x
1
a
R
x
1
[dx[
,
implying that I
is O(). Our assertion follows by noting that (a/Q
x
1
)
x
1
∈ C
∞
0
(R
n
) and
repeating the same argument.
For the general case, we can use a partition of unity to break up Supp(a) into ﬁnitely
many domains as above and then applying the same argument (with x
1
possibly replaced by
another coordinate) to each piece.
2
The upshot of this lemma is that the main (asymptotic) contribution to the integral I
must come from the critical points of R.
Lemma B.2 If the quadratic form Q is nondegenerate, then for each nonnegative integer
K,
R
n
e
iQ/
a [dx[ = (2π)
n/2
e
iπ sgn(Q)/4
[ det
[dx[
T[
1/2
K
¸
k=0
1
k!
(D
k
a)(0)
k
+ O
K+1+n/2
,
where T : R
n
→ (R
n
)
∗
is the selfadjoint map associated to Q and D is the secondorder
diﬀerential operator given by
D =
i
2
¸
j,k
T
−1
jk
∂
∂x
j
∂
∂x
k
.
Proof. From [32, Vol.1, Thm.7.6.1], we recall that for ξ ∈ (R
n
)
∗
,
R
n
e
−i/x,ξ)
e
iQ(x)/
dx = (2π)
n/2
e
iπsgn(Q)/4
e
−iQ
∗
(ξ)
[ det
[dx[
T[
1/2
,
where Q
∗
(ξ) = 'T
−1
ξ, ξ`/2 and the determinant det
[dx[
T is deﬁned as in Appendix A. Con
sequently, the asymptotic Fourier transform of the function x →e
iQ(x)/
equals
ξ →
e
iπsgn(Q)/4
e
−Q
∗
(ξ)/
[ det
[dx[
T[
1/2
.
Combining this expression with the asymptotic Parseval formula, we obtain
I
=
R
n
e
iQ/
a [dx[ =
e
iπsgn(Q)/4
[ det
[dx[
T[
1/2
(R
n
)
∗
e
−iQ
∗
/
T
a (ξ) dξ.
122
A simple computation shows that T
a (ξ) = T
a (−ξ); the change of variables ξ → −ξ
doesn’t aﬀect the integral, and consequently we have
I
=
e
iπsgn(Q)/4
[ det
[dx[
T[
−1/2
(R
n
)
∗
e
−iQ
∗
(ξ)/
T
a (ξ) dξ.
To deal with the integral on the right, we use the the Taylor series expansion (with remainder)
of e
−iQ
∗
/
to write
(R
n
)
∗
e
−iQ
∗
/
T
a(ξ) dξ =
∞
¸
k=0
1
k!
−i
k
(R
n
)
∗
(Q
∗
(ξ))
k
T
a(ξ) dξ
= (2π)
n/2
K
¸
k=0
1
k!
−i
k
T
−1
((Q
∗
)
k
T
a)(0) + O
K+1+
n
2
= (2π)
n/2
K
¸
k=0
1
k!
(D
k
a)(0)
k
+ O
K+1+
n
2
,
the last expression on the right following from the asymptotic Fourier inversion formula.
2
We now wish to apply this lemma to evaluate integrals of the form
I
=
M
e
iR/
σ,
where M is a smooth nmanifold equipped with a compactly supported density σ, and
R: M →R is a smooth function. To this end, we require the following two lemmas. Recall
that the hessian of R : M → R at a critical point p ∈ M is a welldeﬁned selfadjoint
linear map R
tt
(p) : T
p
M →T
∗
p
M. The critical point p is called nondegenerate if R
tt
(p) is an
isomorphism. In this case, the function R has the following normal form near p.
Morse Lemma . If p is a nondegenerate critical point of a function R : M → R, then
there exists a nondegenerate quadratic form Q on R
n
and an embedding g : U → M, where
U is a neighborhood of 0 in R
n
, such that g(0) = p and
(R ◦ g)(x) = R(p) + Q(x).
2
(This theorem is a special case of the Parametrized Morse Lemma, proven in Section 4.3).
If g : M
t
→M is any diﬀeomorphism, then the “change of variables” formula states that
M
g
∗
e
iR/
σ
=
M
e
iR/
σ.
The same role in the stationary phase formula will be played by the following lemma.
123
Lemma B.3 In the notation above, suppose that p ∈ M is a critical point of R and set
p
t
= g
−1
(p). If σ
t
is any density on M
t
such that σ
t
p
= (g
∗
σ)
p
, then
[ det
σ
(R ◦ g)
tt
(p
t
)[ = [ det
σ
R
tt
(p)[.
2
Finally, we mention that deﬁnition of [ det
σ
(R)
tt
(p)[ implies furthermore that
f(p) [ det
fσ
R
tt
(p)[
1/2
= [ det
σ
R
tt
(p)[
1/2
for any function f on M. Combining these observations, we obtain
Principle of Stationary Phase . Let M be a smooth nmanifold and σ ∈ [Ω[
c
M. If
R: M →R has only nondegenerate critical points p
j
, j = 1, , k in Supp(σ), then
M
e
iR/
σ = (2π)
n/2
k
¸
j=1
e
iR(p
j
)/
e
iπsgn(R
(p
j
))/4
[ det
σ
(R
tt
(p
j
))[
1/2
+ O(
1+n/2
).
2
C
ˇ
Cech cohomology
This appendix will give some of the basic deﬁnitions of
ˇ
Cech cohomology for manifolds and
describe its relation to deRham cohomology. A more general treatment is available in [14].
An open cover U = ¦U
α
¦
α∈I
of a manifold M is said to be good if every intersection of
ﬁnitely many members of U is either contractible or empty. For ease of notation, we will
denote by U
α
0
..α
k
the intersection
¸
k
i=0
U
α
i
.
If Γ is an abelian group, a Γvalued
ˇ
Cech cochain with respect to the cover U is then a
rule which assigns an element c
α
0
,..,α
k
of Γ to every list (α
0
, .., α
k
) for which the intersection
U
α
0
..α
k
is nonempty. The group of all such cochains is denoted C
k
U
(M, Γ), and a coboundary
operator
δ
k
: C
k
U
(M, Γ) →C
k+1
U
(M, Γ)
is deﬁned by
δ
k
(c)(α
0
, .., α
k+1
) =
k+1
¸
j=0
(−1)
j
c(α
0
, .., ´ α
j
, .., α
k+1
).
where the symbol ´ indicates which member of the list is to be deleted. The groups of
degreek
ˇ
Cech cocycles and coboundaries are deﬁned respectively by
ˇ
Z
k
U
(M; Γ) = ker(δ
k
)
ˇ
B
k
U
(M; Γ) = im(δ
k−1
),
and the kth
ˇ
Cech cohomology group of M with coeﬃcients in Γ and relative to the covering
U is the quotient
ˇ
H
k
U
(M; Γ) =
ˇ
Z
k
U
(M; Γ)/
ˇ
B
k
U
(M; Γ).
124
Now consider a closed kform ω ∈ Z
k
DR
(M). Since each U
α
is contractible, there exists
on each U
α
a (k −1)form ϕ
α
satisfying dϕ
α
= ω. For any indices α, β, we have
d(ϕ
α
−ϕ
β
) = 0
on U
αβ
. Since U
αβ
is itself contractible, there exist (k − 2)forms ψ
αβ
deﬁned on U
αβ
such
that
dψ
αβ
= ϕ
α
−ϕ
β
and
d(ψ
αβ
+ ψ
βγ
−ψ
αγ
) = 0
on the set U
αβγ
for any indices α, β, γ. Continuing in this way, we see that ω determines a
ˇ
Cech kcocycle with coeﬃcients in R. This association deﬁnes for each k ∈ Z
+
a homomor
phism
Z
k
DR
(M)
w
→
ˇ
H
k
U
(M; R).
Theorem C.1 ([61]) The map w induces an isomorphism between the deRham cohomology
of M and the
ˇ
Cech cohomology of M with real coeﬃcients.
One consequence of this theorem is that
ˇ
H
k
U
(M; Γ) does not depend on the choice of U.
A group homomorphism Γ →Γ
t
induces a homomorphism
ˇ
H
k
U
(M; Γ) →
ˇ
H
k
U
(M; Γ
t
)
in the obvious way. Of particular interest in these notes is the subgroup Z
= 2π Z of R.
D Principal T
bundles
In this appendix, we record some standard facts about principal bundles over paracompact
manifolds, referring to [17] for more details. Throughout this section, denote by Z
the group
2π Z and set T
= R/Z
h
.
A principal T
bundle over a manifold P is a locally trivial T
bundle Q
π
→P together
with a nonsingular, ﬁberpreserving action T
Q →Q. Two principal T
bundles Q
π
→P
and Q
t
π
→ P are said to be isomorphic if there exists a smooth map f : Q → Q
t
which is
equivariant with respect to the T
actions, i.e. f(a p) = a f(p) for all a ∈ T
and p ∈ Q,
and satisﬁes π = π
t
◦ f.
Local triviality of a T
ﬁber bundle Q
π
→ P implies that for any good cover U of P,
there exist homeomorphisms h
j
: U
j
T
→π
−1
(U
j
) such that h
j
(x, t +s) = s h
j
(x, t) and
π(h
j
(x, t)) = x for all (x, t) ∈ U
j
T
and s ∈ T
. These maps give rise to the transition
functions g
jk
: U
jk
→T
of Q, deﬁned by the requirement that
h
j
(x, t) = h
k
(x, t + g
jk
(x))
for all x ∈ U
jk
. This equation implies that for each x ∈ U
ijk
,
h
i
(x, t) = h
i
(x, t + g
ij
(x) + g
jk
(x) + g
ki
(x)),
125
and so the transition functions satisfy the cocycle condition g
ij
+g
jk
+g
ki
= 0 (mod Z
). If
˜ g
jk
: U
jk
→R is any lift of g
jk
, then the numbers
c
ijk
= ˜ g
ij
+ ˜ g
jk
+ ˜ g
ki
are therefore elements of Z
which deﬁne a
ˇ
Cech cocycle [c
ijk
]. The corresponding class
[Q] ∈
ˇ
H
2
(P; Z
) is known as the Chern class of Q. The fundamental theorem describing
this space is the following (see [17, Cor.2.1.4] for a proof).
Theorem D.1 Two principal T
bundles over P are isomorphic if and only if their Chern
classes are equal. Moreover, the assignment Q →[Q] induces a bijective map from the space
of isomorphism classes of principal T
bundles over P to
ˇ
H
2
(P; Z
).
Corresponding to the abelian group structure on
ˇ
H
2
(P; Z
) are the following operations on
principal T
bundles. If Q → P is a principal T
bundle having transition functions ¦g
jk
¦
with respect to some good cover of P, then the inverse of Q is deﬁned as the principal
T
bundle −Q over P obtained from the transition functions ¦−g
jk
¦. Similarly, if Q, Q
t
are principal T
bundles over P with transition functions ¦g
jk
¦ and ¦g
t
jk
¦ respectively, then
the product of Q and Q
t
is deﬁned as the principal T
bundle Q
P
Q
t
over P having
transition functions ¦g
jk
+g
t
jk
¦. From the deﬁnitions above, it follows easily that the Chern
classes of inverses and products of principal T
bundles are given by [−Q] = −[Q] and
[Q
P
Q
t
] = [Q] +[Q
t
]. On the level of the bundles themselves, we can describe the product
Q
P
Q
t
as the quotient of the usual ﬁberproduct Q
P
Q
t
(which in this case is a T
T

bundle over the base), modulo the antidiagonal action of T
, i.e. t (p, p
t
) = (t p, −t p
t
).
T
bundles with connection
The inﬁnitesimal generator of the T
action on a principal bundle Q is a vector ﬁeld X on
Q deﬁned by the equation
X(p) =
d
dt
t=0
t p.
A connection on Q is a T
invariant 1form ϕ on Q such that ϕ(X) = 1. In terms of a
good cover U of P, the form ϕ satisﬁes h
∗
j
ϕ = dσ + π
∗
ϕ
j
, where dσ denotes the usual form
on T
and the ϕ
j
are 1forms on the U
j
satisfying
ϕ
j
−ϕ
k
= d˜ g
jk
,
where ˜ g
jk
are again Rvalued lifts of the transition functions of Q. The curvature of the
connection ϕ is the unique closed 2form ω on P such that
dϕ = π
∗
ω.
From the compatibility condition for the ϕ
j
, it follows that the Chern class of Q is the
ˇ
Cech
representative of the deRham cohomology class [ω].
Theorem D.2 A closed 2form ω on a manifold P is the curvature form of a connection
on a principal T
bundle Q over P if and only if 'ω, a` ∈ Z
for any a ∈ H
2
(P; Z).
126
Proof. Most published proofs of this result (e.g. [37]) use
ˇ
Cech cohomology and the deRham
isomorphism. We prefer to give the following direct proof by P.Iglesias; see [33] for further
details. Let A(P, p
0
) denote the space of smooth paths γ : [0, 1] →P such that γ(0) = p
0
, and
let e: A(P, p
0
) →P be the endpoint map e(γ) = γ(1). Since the interval [0, 1] is contractible,
there exists a natural contraction of A(P, p
0
) onto the constant map [0, 1] →p
0
. If Y
t
is the
vector ﬁeld which generates this contraction, we deﬁne the 1form Kω on A(P, p
0
) by
Kω =
1
0
(Y
t
e
∗
ω) dt.
Then dσ + Kω is a connection form on the product A(P, p
0
) T
having curvature e
∗
ω.
To complete the proof, we will deﬁne the T
bundle (Q, ϕ) over P with curvature ω as an
appropriate quotient of A(P, p
0
)T
. For this purpose, we call two elements γ, γ
t
∈ A(P, p
0
)
homologous if their diﬀerence is the boundary of a singular 2chain σ in P. The quotient
of A(P, p
0
) by this equivalence relation is the covering space
´
P of P corresponding to the
commutator subgroup of π
1
(P), and thus, H
1
(
´
P, Z
) = 0. An equivalence relation on the
product A(P, p
0
) T
is then deﬁned by the condition that (γ, t) ∼ (γ
t
, t
t
) if γ, γ
t
are
homologous and
t −t
t
=
σ
ω,
where ∂σ = γ −γ
t
. The quotient of A(P, p
0
) T
by this equivalence relation is a principal
T
bundle
´
Q over
´
P with connection ´ ϕ having curvature π
∗
ω, where π:
´
P →P is the natural
projection.
If H
1
(P; Z
) = 0, then
´
P = P, and the proof is complete. Otherwise, let s be any map
from H
1
(P; Z
) into the space of loops based at p
0
which assigns a representative to each
homology class. A T
valued group cocycle on H
1
(P; Z
) is then deﬁned by
φ(h, h
t
) =
σ
ω,
where ∂σ = s(h + h
t
) −(s(h) + s(h
t
)) . Since this cocycle is symmetric, it deﬁnes a central
extension Γ of H
1
(P; Z
) which acts naturally on
´
Q by
(h, τ)[γ, z] = [s(h) γ, z + τ].
Since T
is divisible, the extension Γ is isomorphic to the product T
H
1
(P; Z
), and any
choice of isomorphism deﬁnes an action of H
1
(P; Z
) on
´
Q which preserves the connection
´ ϕ. The quotient of
´
Q is the desired principal T
bundle Q over P.
2
A connection ϕ on a principal T
bundle Q
π
→ P induces a connection on the inverse −Q
of Q whose local representatives are of the form dσ − ϕ
j
, where ϕ is locally represented by
dσ+ϕ
j
. Similarly, connections ϕ and ϕ
t
on the T
bundles Q, Q
t
over P induce a connection
ϕ + ϕ
t
on the product Q
P
Q
t
deﬁned locally by dσ + ϕ
j
+ ϕ
t
j
.
127
Two principal T
bundles with connection (Q, ϕ), (Q
t
, ϕ
t
) over P are said to be isomorphic
provided that there exists an isomorphism f : Q → Q
t
of the underlying principal bundles
which satisﬁes f
∗
ϕ
t
= ϕ. A check of the deﬁnition shows that (Q, ϕ), (Q
t
, ϕ
t
) are isomorphic
if and only if (Q
P
−Q
t
, ϕ − ϕ
t
) is isomorphic to the trivial bundle (P T
, dS). Since
curvature is obviously invariant under isomorphisms of principal T
bundles with connection,
the classiﬁcation of such bundles reduces to the classiﬁcation of ﬂat connections on trivial
T
bundles over P.
A connection ϕ on a principal T
bundle Q
π
→ P is said to be ﬂat if dϕ = 0. In this
case, the local representatives h
∗
j
ϕ = dσ + ϕ
j
have the property that ϕ
j
= d˜ s
j
for functions
˜ s
j
: U
j
→ R. Denoting by s
j
the composition of ˜ s
j
with the projection R → T
, we ﬁnd
that the functions s
jk
= s
j
−s
k
deﬁne transition functions for a trivial T
principal bundle
over P. On the other hand, the compatibility condition for the ϕ
j
implies that s
jk
− g
jk
is
constant for each j, k, and so there exists an isomorphism f : Q → P T
. It is easy to
check that if ϕ
0
is the trivial connection on P T
, then
ϕ −f
∗
ϕ
0
= π
∗
β
for some closed 1form β on P. The cohomology class [β] induced by this form in
ˇ
H
1
(P; T
)
is called the holonomy of the (ﬂat) connection ϕ.
Theorem D.3 Two ﬂat connections ϕ,ϕ
t
on the trivial principal T
bundle Q = P T
over
P are isomorphic if and only if they have equal holonomy. Moreover, the map (Q, ϕ) →[β
ϕ
]
induces a bijection from the space of isomorphism classes of ﬂat connections on the trivial
T
bundle with
ˇ
H
1
(P; T
).
A section s of a principal T
bundle Q over P with connection ϕ is called parallel provided
that s
∗
ϕ = 0. The map s
Q
: Q →Q associated to a parallel section deﬁnes an isomorphism
of (Q, ϕ) with the trivial T
bundle equipped with the trivial connection ϕ = dσ. Thus, we
have:
Corollary D.4 A principal T
bundle Q over P with a ﬂat connection admits a parallel
section if and only if (Q, ϕ) has zero holonomy.
2
Associated line bundles
A representation ρ: T
→U(1) enables us to associate to any principal T
bundle Q →P a
complex line bundle E →P deﬁned explicitly as the quotient of QC by the T
action
t (p, z) = (t p, ρ
−1
(t)z).
The space of functions g : Q →C satisfying the condition
g(t p) = ρ
−1
(t)g(p)
128
is identiﬁed with the space of sections of E by the assignment g → s
G
, where the section
s
g
: P →E is deﬁned by the requirement that
s
g
(x) = [(p, g(p))],
for any element p of π
−1
(x).
129
References
[1] Abraham, R., and Marsden, J.E., Foundations of Mechanics, 2nd edition, Ben
jamin/Cummings, Reading, 1978.
[2] Arnol’d, V.I., Mathematical Methods of Classical Mechanics, Graduate Texts in
Math. 60, SpringerVerlag, New York, 1978.
[3] Arnol’d, V.I., On a characteristic class entering into conditions of quantization,
Funct. Anal. Appl. 1 (1967) 113.
[4] Arnol’d, V.I., Catastrophe theory (3rd edition), SpringerVerlag, New York, 1992.
[5] Arnol’d, V.I., Symplectic geometry and topology, preprint 1993.
[6] Arnol’d, V.I., and Givental’, A.B., Symplectic geometry, Dynamical systems IV,
V.I. Arnol’d and S.P. Novikov, eds., Encyclopedia of Mathematical Sciences, Vol. 4,
SpringerVerlag, New York, 1990, pp. 1136.
[7] Atiyah, M.F., The Geometry and Physics of Knots, Cambridge Univ. Press, Cam
bridge, 1990.
[8] Audin,M., The Topology of Torus Actions on Symplectic Manifolds, Progress in
Math. vol. 93, Birkh¨auser, Basel, 1991.
[9] Bayen, F., Flato, M., Fronsdal, C., Lichnerowicz, A., and Sternheimer, D., Defor
mation theory and quantization, I and II, Ann. Phys. 111, (1977), 61151.
[10] Berezin, F.A., Some remarks about the associated envelope of a Lie algebra, Funct.
Anal. Appl. 1, (1967), 91102.
[11] Berezin, F.A., Quantization, Math USSR Izv. 8 (1974), 11091165.
[12] Besse, A. Manifolds All of Whose Geodesics are Closed, Ergebnisse der Math. 93
SpringerVerlag, Berlin, 1978.
[13] Bohm, D., Quantum Theory, PrenticeHall, New York, 1951.
[14] Bott, R. and Tu, L, Diﬀerential Forms in Algebraic Topology, SpringerVerlag, New
York, 1982.
[15] Boutet de Monvel, L., and Guillemin, V., The Spectral Theory of Toeplitz Operators,
Annals. of Math. Studies 99, Princeton University Press, Princeton, 1981.
[16] Br¨ uning, J. and Guillemin, V, Mathematics Past and Present: Fourier Integral Op
erators, SpringerVerlag, Berlin, 1994.
[17] Brylinski, J.L., Loop Spaces, Characteristic Classes, and Geometric Quantization,
Progress in Mathematics vol. 107, Birkh¨auser, Boston, 1992.
130
[18] Connes, A., Noncommutative Geometry, Academic Press, San Diego, 1995.
[19] De Wilde, M., and Lecomte, P., Formal deformations of the Poisson Lie algebra
of a symplectic manifold and starproducts. Existence, equivalence, derivations, in
M. Hazewinkel and M. Gerstenhaber, eds., Deformation Theory of Algebras and
Structures and Applications, Kluwer Acad. Pub., Dordrecht (1988), 897960.
[20] Dirac, P. A. M., The Principles of Quantum Mechanics, Clarenden Press, Oxford,
1930.
[21] Duistermaat, J. J., Fourier Integral Operators, Courant Institute, NYU, New York,
1973. (New edition to be published by Birkh¨auser Boston in 1995.)
[22] Ehresmann, C., Les connexions inﬁnit´esimales dans un espace ﬁbr´e diﬀ´erentiable,
Colloque de Topologie (Espaces Fibr´es) tenu a Bruxelles du 5 au 8 juin, 1950,
Georges Thone, Li`ege and Masson et Cie., Paris, 1951, pp. 2955.
[23] Fedosov, B., A simple geometrical construction of deformation quantization, J. Diﬀ.
Geom. 40 (1994) 213238.
[24] Folland, G.B., Harmonic Analysis in Phase Space, Annals of mathematics studies
122, Princeton University Press, 1989.
[25] Fomenko, A.T., Symplectic Geometry, Gordon and Breach, New York, 1988.
[26] Golubitsky, M., Guillemin, V., Stable Mappings and Their Singularities, Graduate
Texts in Math., vol. 14, SpringerVerlag, (1973).
[27] Gompf, R., New constructions of symplectic manifolds, Ann. Math., to appear.
[28] Guillemin, V., and Sternberg, S., Geometric Asymptotics, Math. Surveys 14, Amer.
Math. Soc., Providence, 1977.
[29] Guillemin, V., and Sternberg, S., Symplectic Techniques in Physics, Cambridge Uni
versity Press, Cambridge, 1984.
[30] Helﬀer, B., Semiclassical analysis for the Schrodinger operator and applications,
Lecture Notes in Math. 1336 (1988).
[31] H¨ormander, L., Fourier integral operators I., Acta Math. 127 (1971) 79183.
[32] H¨ormander, L., The Analysis of Linear Partial Diﬀerential Operators, Vols. 14,
SpringerVerlag, New York, 198390.
[33] Iglesias, P., La trilogie du moment, Ann. Inst. Fourier (to appear).
[34] Karasev, M.V. and Maslov, V.P., Nonlinear Poisson Brackets : Geometry and Quan
tization Translations of mathematical monographs vol. 119, American Mathematical
Society, Providence, R.I., 1993.
131
[35] Kirillov, Geometric quantization, Dynamical Systems IV, V.I. Arnol’d and S.P.
Novikov, eds., Encyclopedia of Mathematical Sciences, Vol. 4, SpringerVerlag, New
York, 1990, pp. 137172.
[36] Klingenberg, W., Riemannian Geometry, deGruyter Studies in Math., Berlin, 1982.
[37] Kostant, B., Quantization and unitary representations, Lecture Notes in Math. 170
(1970), 87208.
[38] Libermann, P., and Marle, C.M., Symplectic Geometry and Analytical Mechanics,
Reidel, Dordrecht, 1987.
[39] Lie, S., Theorie der Transformationsgruppen, (Zweiter Abschnitt, unter Mitwirkung
von Prof. Dr. Friedrich Engel), Leipzig, Teubner, 1890.
[40] McDuﬀ, D., Examples of symplectic structures, Invent. Math. 89 (1987), 1336.
[41] Mackey, G., Mathematical Foundations of Quantum Mechanics, W.A.Benjamin,
Reading, 1963.
[42] Marsden, J., Lectures on Mechanics, London Math. Soc. Lecture Note Series 174,
Cambridge Univ. Press, Cambridge, 1992.
[43] Maslov, V.P., Theory of Perturbations and Asymptotic Methods, Moskov. Gos. Univ.,
Moscow, 1965.
[44] M´elotte, D., Invariant deformations of the Poisson Lie algebra of a symplectic man
ifold and star products, in M. Hazewinkel and M. Gerstenhaber, eds., Deformation
Theory of Algebras and Structures and Applications, Kluwer Acad. Pub., Dordrecht
(1988), 961972.
[45] Moser, J., On the volume elements on a manifold, Trans. Amer. Math. Soc. 120
(1965), 280296.
[46] Nazaikinskii, V., Sternin, B., and Shatalov, V., Methods of Noncommutative Analy
sis, Theory and Applications, Berlin, New York, W. de Gruyter (to appear).
[47] Olver, F.W.J., Asymptotics and Special Functions, Academic Press, New York, 1974.
[48] Omohundro, S., Geometric Perturbation Theory in Physics, World Scientiﬁc, Singa
pore, 1986.
[49] Omori, I., Inﬁnitedimensional Lie Transformation Groups, Lecture Notes in Math
ematics, vol.427, SpringerVerlag, New York, 1975.
[50] Rieﬀel, M., Deformation quantization of Heisenberg manifolds, Comm. Math. Phys.
122 (1989), 531562.
132
[51] Robert, D., Autour de l’Approximation Semiclassique, Progress in Mathematics,
vol. 68, Birkh¨auser, Boston, 1987.
[52] Segal, I., Quantization of nonlinear systems, J. Math. Phys. 1 (1960), 468488.
[53] Sniatycki, J., Geometric Quantization and Quantum Mechanics, SpringerVerlag,
New York, 1980.
[54] Souriau, J.M., Structure des Syst`emes Dynamiques, Dunod, Paris, 1970.
[55] Souriau, J.M., Groupes diﬀ´erentiels, Lecture Notes in Math. 836 (1980), 91128.
[56] Treves, F., Introduction to Pseudodiﬀerential and Fourier Integral Operators.. Vols
I,II. Plenum Press, New York, 1980.
[57] Tulczyjew, W.M., The Legendre transform, Ann. Inst. H. Poincar´e, 27 (1977) 101
114.
[58] Unterberger, A., Quantiﬁcation et analyse pseudodiﬀ´erentielle, Ann. Sci.
´
Ec. Norm.
Sup. 21 (1988), 133158.
[59] Vaisman, I., Symplectic Geometry and Secondary Characteristic Classes, Birkh¨ auser,
Boston, 1987.
[60] Wallach, N. R., Symplectic Geometry and Fourier Analysis, Math. Sci. Press, Brook
line, 1977.
[61] Weil, A., Sur les th´eor`emes de de Rham, Comm. Math. Helv. 26 (1952) 119145.
[62] Weinstein, A., Symplectic manifolds and their lagrangian submanifolds, Advances
in Math. 6 (1971), 329346.
[63] Weinstein, A., Lectures on Symplectic Manifolds, Regional conference series in math
ematics 29, Amer. Math. Soc., Providence, 1977.
[64] Weinstein, A., Symplectic geometry, Bulletin A.M.S. (new series) 5 (1981), 1–13.
[65] Weinstein, A., Coisotropic calculus and Poisson groupoids, J. Math. Soc. Japan 40
(1988), 705727.
[66] Weinstein, A., Noncommutative geometry and geometric quantization, in Symplectic
Geometry and Mathematical Physics: actes du colloque in l’honneur de JeanMarie
Souriau, P.Donato et al. eds., Birkh¨auser, Basel, 1991.
[67] Weinstein, A., Symplectic groupoids, geometric quantization, and irrational rotation
algebras, in Symplectic Geometry, Groupoids, and Integrable Systems, P.Dazord and
A.Weinstein, eds., MSRI Publications, Springer, New York, 1991.
[68] Weinstein, A., Deformation quantization, S´eminaire Bourbaki, no. 789, Juin 1994.
133
[69] Weinstein, A., and Xu, P., Extensions of symplectic groupoids and quantization, J.
Reine Angew. Math. 417 (1991), 159189
[70] Weyl, H., The Theory of Groups and Quantum Mechanics, Dover Publications, New
York, 1950.
[71] Woodhouse, N.M.J., Geometric Quantization, 2nd Ed., Oxford U. Press, Oxford,
1992.
[72] Yano, K., and Ishihara, S., Tangent and Cotangent Bundles; Diﬀerential Geometry
Dekker, New York, 1973.
134
.
.
.
. . . . . . . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . . . . . . . . . . . 5 The 5. . . . . . . . . . . . . . . . . 8. . . . . . . . . . . . .1 Compositions of semiclassical states . . . . . . . . 109 8 Algebraic Quantization 8. . . . . . . . . . . . 93 7. . .2 The WKB approximation . . . . 6. . . . . . . . . . . . . . . . . . . . . . .3 Symplectic groupoids . . . . .2 WKB quantization and compositions . . . . . . . . . .3 Phase functions and lagrangian submanifolds 4. . . . . . . . . . . . . . . .2 Deformation quantization . 3. . . . . . . . . . 6 Fourier Integral Operators 6. . . . . . . . 7 Geometric Quantization 93 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Poisson algebras and Poisson manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. . . . . . . . . . .1 Some HamiltonJacobi preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . .2 The Maslov correction . . . . . . . . . . .1 Prequantization . . . . . . . . . . . . . . . . . . . . . . . 2. . . . . . . .4 WKB quantization . . . . . 3 Symplectic Manifolds 3. . . . . . . . . . .3 Quantization of semiclassical states . 8. . . . A Densities B The method of stationary phase ˇ C Cech cohomology D Principal T bundles 111 111 112 114 119 121 124 125 2 . . . . . . . . . . . . . .Contents 1 Introduction: The Harmonic Oscillator 2 The WKB Method 2. . . . . . . . . . .2 Cotangent bundles . Symplectic manifolds and mechanics . . . . . The symplectic category . . . . . . . . . . . Symplectic Category Symplectic reduction . . . . . . .1 Prequantization . . . . . . . .2 Polarizations and the metaplectic correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Quantization in Cotangent Bundles 4. . . . .3 5 8 8 11 17 17 28 32 36 36 41 46 56 64 64 76 79 83 83 87 . . . . . . . . . . . . . . . . 98 7. . . . . . . . . . . . . . . . . . . . . .2 5. . . . . . . . . . . . . . .1 5. . . . . . . . . . . . . . . . . . . . . . 4. . .1 Symplectic structures . . . . . . .3 Mechanics on manifolds . . . . .
.
but on all the mathematics at once. These notes are still not in ﬁnal form. we show how symplectic geometry arises from the study of semiclassical solutions to the Schr¨dinger equation. operators. a topic pretty much ignored in the present notes. A previous paper by one of us [64] referred to “the symplectic creed”. and [56]. and which can be bypassed on a ﬁrst reading. The pamphlet [63] is in some sense a precursor to these notes. These notes are meant to function as a guide to the literature. For classical mechanics and symplectic geometry. Symplectisation belongs to the small set of highest level operations. Although symplectic geometry is like any ﬁeld of mathematics in having its deﬁnitions. with special attention to the role of symmetry groups. we refer to [43] for a physicsoriented presentation and to the notes [21] and the treatises [32]. symplectic and contact geometries have encroached on all areas of mathematics. algebraisation. [59]. we refer to other sources for many details which are omitted here. For “geometric quantization”. it is also a special way of looking at a very broad part of mathematics and its applications. but most of them provide too much detail for the reader who just wants to ﬁnd out what the subject is about. one may consult [35]. The only prerequisite for the course (and for these notes) was a knowledge of the basic notions from the theory of diﬀerentiable manifolds (diﬀerential forms. and in o turn provides a geometric foundation for the further analysis of this and other formulations of quantum mechanics. it is almost a religion. complexiﬁcation. [6]. and so on. superisation. is [41]. is [28]. with an emphasis on the role which these ideas play in formalizing the transition between the mathematics of classical dynamics (hamiltonian ﬂows on symplectic manifolds) and that of quantum mechanics (unitary ﬂows on Hilbert spaces). theorems. a much more complete reference on the subject. [2]. Although some such highest level operations are presently known (for example. [54]. etc. [46]..1 In these notes. An earlier work. two basic references on quantum mechanics itself are [13] and [20]. Finally. but they have already beneﬁtted from the comments 1 We like the following quotation from [4] very much: In recent years.Preface These notes are based on a course entitled “Symplectic geometry and geometric quantization” taught by Alan Weinstein at the University of California. symplectisation) there is as yet no axiomatic theory describing them. acting not on details (functions. [60] or [71]. functors). [38]. [8]. [25]. The book [29] treats further topics in symplectic geometry and mechanics. etc. written at about the same time. in the fall semester of 1992 and again at the Centre Emile Borel (Institut Henri Poincar´) in the spring semester e of 1994. As each skylark must display its comb.). one of the ﬁrst to treat the connections between classical and quantum mechanics from a geometric viewpoint. For many “symplecticians”. 3 . In mathematics there exist operations on diﬀerent levels: functions acting on numbers. [53]. Bourbakisation. vector ﬁelds. functors acting on operators. On the other hand. Berkeley. For more extensive treatment of the PDE aspects of the subject. transversality. The aim of the course was to give students an introduction to the ideas of microlocal analysis and the related symplectic geometry. so every branch of mathematics must ﬁnally display symplectisation. There already exist many books on the subjects treated here. operators acting on functions. we suggest [1].
W. and Dmitry Roytenberg. S.B. 4 . We welcome further comments. A. Jim Morehead. was supported by NSF graduate and postdoctoral fellowships in mathematics. especially Maurice Garay.and suggestions of many readers. During the preparation of these notes. We would like to thank the Centre Emile Borel and the Isaac Newton Institute for their hospitality. was partially supported by NSF Grants DMS9001089 and 9301089.
˙ so that Newton’s equation becomes the pair of equations: q= ˙ p m p = −kq. solution curves in the case of the harmonic oscillator are ellipses centered at the origin.1 Introduction: The Harmonic Oscillator In these notes. The classical picture The harmonic oscillator in 1 dimension is described by Newton’s diﬀerential equation: m¨ = −kx. we will take a “spiral” approach toward the quantization problem. ˙ ˙˙ ˙˙ dt ∂q ∂p If we had chosen x rather than mx as the second coordinate of our phase plane. p). which can describe a wide variety of classical mechanical systems with appropriate choices of the function H. we can convert this secondorder ordinary diﬀerential equation into a system of two ﬁrstorder equations. we would not have ˙ ˙ arrived at this universal form of the equations. x By a standard procedure. Speciﬁcally. and then returning to the same kind of problem at progressively higher levels of generality. 2 5 . We note two qualitative features of the hamiltonian description of a system: 1. are called Hamilton’s equations. we will start with the harmonic oscillator as described classically in the phase plane R2 and work toward the problem of quantizing arbitrary symplectic manifolds. Introducing the “phase plane” R2 with position and momentum coordinates (q. and points in the phase plane move clockwise around each ellipse. ˙ If we now introduce the hamiltonian function H : R2 → R representing the sum of kinetic and potential energies. The latter problem has taken on a new interest in view of recent work by Witten and others in the area of topological quantum ﬁeld theory (see for example [7]).2 Hamilton’s equations deﬁne a ﬂow on the phase plane representing the timeevolution of the classical system at hand. p2 kq 2 H(q. we set q=x p = mx. p) = + 2m 2 then we ﬁnd ∂H ∂H p=− ˙ q= ˙ ∂p ∂q These simple equations. The derivative of H along a solution curve is dH ∂H ∂H = q+ ˙ p = −pq + q p = 0. beginning with a very concrete example and its proposed solution.
t) = ψ(x. C) representing the time evolution of the quantum system. a wave function satisfying Schr¨dinger’s equation then corresponds to an integral o curve of the associated ﬂow.. i ψ ∗ ψ = 1. the motion of the harmonic oscillator is described by a complexvalued wave function ψ(x.e. 6 . and its ﬂow preserves area in the phase plane. − ∂H ) is ˙ ˙ ∂p ∂q · XH = ∂2H ∂2H − = 0. instead. The quantum mechanical picture In quantum mechanics. we may rewrite the Schr¨dinger equation o ˆ def H = − 2 then its squarenorm ρ(x. the value of H is constant along integral curves of the hamiltonian vector ﬁeld.e. ∂t 2m ∂x2 2 Here. t) satisfying the 1dimensional Schr¨dinger equation: o 2 ∂ψ ∂2ψ k 2 =− + x ψ. Since H represents the total energy of the system. We shall see in Chapter 3 that the use of diﬀerential forms leads to a coordinatefree description and generalization of the hamiltonian viewpoint in the context of symplectic geometry. From the latter point of view. The divergence of the hamiltonian vector ﬁeld XH = (q. t) itself may be viewed alternatively as a tdependent function of x. C). The description of classical hamiltonian mechanics just given is tied to a particular coordinate system. ∂q ∂p ∂p ∂q Thus the vector ﬁeld XH is divergencefree. t)2 is interpreted as a probability density for observing the oscillator at the position x at time t. ∂t A solution ψ of this equation does not represent a classical trajectory. this property of the ﬂow is interpreted as the law of conservation of energy.i. Interpreting the right hand side of this equation as the result of applying to the wave function ψ the operator i ∂ψ ˆ = Hψ. 2. the Schr¨dinger equation is a general o form for the quantum mechanical description of a large class of systems. i. The wave function ψ(x. 2m ∂x2 2 is the operator of multiplication by x2 . Schr¨dinger’s o ∞ equation deﬁnes a vector ﬁeld on C (R. if ψ is normalized. R where mx2 as k ∂2 + m x2 . Like Hamilton’s equations in classical mechanics. or as a path in the function space C ∞ (R. p) = ( ∂H . Planck’s constant has the dimensions of action (energy × time).
To the classical position and momentum observables q. ∂x The classical hamiltonian H(q. What is the meaning of “macroscopic” in mathematical terms? It turns out that good approximate solutions of Schr¨dinger’s equation can be generated from classical information o when is small enough. . p) = p2 /2m+kq 2 /2 then corresponds naturally to the operator ˆ H. from which a unit of action appropriate to the system can be derived. the classical motions described by solutions of Hamilton’s equations lead to approximate solutions of Schr¨dinger’s equation. not only in verifying that the theories are consistent with the fact that we “see” classical behavior in systems which are “really” governed by quantum mechanics. we would like to establish that. so ˆ ˆ 2 2 that we are forced to choose between (1 + q )ˆ and p2 (1 + q 2 ). But how can a constant with physical dimensions be small? Although there remain some unsettled issues connected with the question. p we associate the diﬀerential operators q → q = mx ˆ ∂ . . “How can become small?” the answer is essentially the following.Quantization and the classical limit The central aim of these notes is to give a geometric interpretation of relationships between the fundamental equations of classical and quantum mechanics.” That is. we run in to the problem that the operators q and p do not commute with one another. For any particular mechanical system. As soon as we wish to “quantize” a more complicated energy function. we will often regard mathematically as a formal parameter or a variable rather than as a ﬁxed number. velocities. Based on the present discussion of the harmonic oscillator. and the classical limit is applicable when divided by this unit is much less than 1. The diﬀerence between these choices turns out to become small when → 0. such as (1 + q 2 )p2 . for systems which are in some sense macroscopic. one tenuous connection can be drawn as follows. there are usually characteristic distances. masses. Establishing this relation o between classical and quantum mechanics is important. In these notes. but also as a tool for developing approximate solutions to the quantum equations of motion. But how can a constant approach zero? Besides the problem of “quantization of equations. among a number of other ˆ p ˆ ˆ possibilities. . p → p = −i ˆ 7 .” we will also treat that of “quantization of solutions.
e. o As a ﬁrst step toward solving the Schr¨dinger equation. we look for stationary states. t) = ϕ(x) e−iωt .1 Some HamiltonJacobi preliminaries In this section we will carry out the ﬁrst step in the WKB method to obtain an approximate solution to the “stationary state” eigenvalue problem arising from the Schr¨dinger equation. and Brillouin. Consider a 1dimensional system with hamiltonian H(q. Substituting this expression for ψ in the Schr¨dinger equation. including Liouville. Hamilton’s equations now become q= ˙ For ﬁxed p m p = −V (q). we will still refer to the method as WKB. o we obtain ˆ ω ϕ(x) e−iωt = (Hϕ)(x) e−iωt . Kramers. Green. (Other names.2 The WKB Method A basic technique for obtaining approximate solutions to the Schr¨dinger equation from o classical motions is called the WKB method. so the following discussion is absolutely central to these notes. solutions of the form ψ(x.) A good part of what is now called microlocal analysis can be understood as the extension of the basic WKB idea to more precise approximations and more general situations. and Jeﬀreys are sometimes attached to this method. 8 . 2. o The geometric interpretation of this technique will lead to a correspondence between classical and quantum mechanics which goes beyond the one described in Chapter 1. Also see [5]. ∂t ∂2 + mV 2m ∂x2 2 where ˆ H=− is the Schr¨dinger operator. after Wentzel. o i. Eliminating the factor e−iωt above. References [13] and [47] contain a discussion of some of its history. these solutions keep the same form (up to multiplication by a complex scalar of norm 1). socalled because as time evolves. nevertheless. 2m where V (x) is a potential (for example the potential kx2 /2 for the harmonic oscillator). For convenience. ˙ ∈ R+ . we arrive at the timeindependent Schr¨dinger equao tion: ˆ (H − E) ϕ = 0. where the method is traced all the way back to 1817. Schr¨dinger’s equation assumes the form o i ∂ψ ˆ = Hψ. p) = p2 + V (q).
we ﬁnd that ˆ (H − E) ϕ = 0 ⇔ ( ξ)2 = 2m(E − V ). our ﬁrstorder approximation attempt will ignore the last term in brackets. For V < E. then ξ should vary with x as well. and there are only real exponential solutions. the ˆ constant ξ is imaginary. it represents a particle which is equally likely to be anywhere in space. the eigenvalue E represents the energy of the system. This equation means that ϕ is to be an eigenfunction of the linear diﬀerential ˆ operator H. to kill the other two terms. S (x) = ± 2m(E − V (x)). we consider the classical phase4 plane R2 T ∗ R with coordinates (q. if V varies with x. To understand the phase function S geometrically. and so we are dealing with a free particle. The basic idea at this stage of the WKB method is that. p). 2m See [55] for a grouptheoretic interpretation of such states. and one has an abundance of exact solutions of the Schr¨dinger equation which are oscillatory and bounded. and in this case we have ˆ (H − E) ϕ = (S (x))2 i + (V − E) − S (x) eiS(x)/ . This proposed form of the solution is the simplest version of the WKB ansatz. The diﬀerential dS = S dx can be viewed as a mapping dS : R → T ∗ R. in which case the force −V (x) is zero. This observation establishes a fundamental link between classical and quantum mechanics: When the image of dS lies in a level manifold of the classical hamiltonian. the (real) value of ξ is thus determined up to a choice of sign.3 When E < V . o 3 4 (S (x))2 + V (x) = E. which are unbounded and admit no physical interpretation. we require that S satisfy the eikonal or HamiltonJacobi equation: H(x. where as usual we set p = S . Trying a solution of the form ϕ(x) = eixξ for some constant ξ. Suppose for the moment that the potential V is constant. the function S may be taken as the phase function of a ﬁrstorder approximate solution of Schr¨dinger’s equation. which has a “deﬁnite value” in this state. 2m 2m Since we will consider to be “small”.where E = ω. for some realvalued function S known as a phase function. o Such a wave function is not squareintegrable and as such is said to be “unnormalizable”. but which has a deﬁnite momentum (since it is an eigenfunction of the momentum operator p).e. a more general solution candidate is then ϕ(x) = eiS(x)/ . Then S satisﬁes the HamiltonJacobi equation if and only if the image of dS lies in the level manifold H −1 (E). These two uses of the term “phase” seem to be unrelated! 9 . S (x)) = i.
the level sets for the 1dimensional harmonic oscillator are lagrangian submanifolds in the phase plane. Such submanifolds are called lagrangian. The pullback to L of the form αn = j pj dqj on R2n is exact. the ansatz ϕ again constitutes a ﬁrstorder approximate solution to the timeindependent Schr¨dinger equation. L is an ndimensional submanifold of H −1 (E). In other words. ∂x1 ∂xn = S(x) 2m 2 + V (x) = E. The restriction of the canonical projection π : T ∗ Rn → Rn to L induces a diﬀeomorphism L Rn . 2. For example. It is also clear that the curve fails to project diﬀeomorphically onto R. which we will clarify later. suggesting that in general the state of a system should be represented by the submanifold L (projectable or not) rather than by the phase function S. if we consider a WKB ansatz of the form ϕ = eiS/ . the pullback to L of the form p dq is closed. The image L = im(dS) of the diﬀerential of an admissible phase function S is characterized by three geometric properties: 1. .. the Schr¨dinger o operator corresponding to the classical hamiltonian H(q. ∂S ∂S . so its pullback to L is not exact. o n We will call a phase function S : R → R admissible if it satisﬁes the HamiltonJacobi equation. While many of the basic constructions of microlocal analysis are motivated by operations on these projectable submanifolds of T ∗ Rn R2n . but the integral of p dq around the curve equals the enclosed nonzero area. ˆ Since ϕ is of order zero in . applications of the theory require us to extend the constructions to more general ndimensional submanifolds of R2n satisfying only a weakened version of condition (2) above. L is projectable.. p) = is ˆ H=− p2 i + V (q) 2m 2 ∆ + mV . As before. 2m where ∆ denotes the ordinary Laplacian. 3. xn . In Rn . This idea. 10 . . then ˆ (H − E) ϕ = S 2m 2 + (V − E) − i ∆S eiS/ 2m will be O( ) provided that S satisﬁes the HamiltonJacobi equation: H x1 . is the starting point of the geometrical approach to microlocal analysis. A regular level curve of the hamiltonian is an ellipse L.. while (H − E) ϕ = O( ). . From the classical standpoint. in which “exact” is replaced by “closed”. Since L is 1dimensional. . the behavior of an oscillator is nevertheless completely described by its trajectory. .The preceding discussion generalizes easily to higher dimensions.
where its potential energy is higher. ∂pj ∂qj ∂qj ∂pj j A simple computation shows that XH and the form αn are related by the equation XH i. the quantity ϕ(x)2 represents the probability of the particle being at the position x. HamiltonJacobi theorem . dαn = −dH. It is also clear on physical grounds that our ansatz for ϕ is too restrictive because it satisﬁes ϕ(x) = 1 for all x.e.e. 2. The restriction of dαn to the tangent space Tp R2n of R2n at any point p deﬁnes a nondegenerate. ˆ Up to terms of order . These remarks imply that XH is tangent to L.. dαn (XH . in other words. then ϕ(x) = eiS(x)/ satisﬁes ˆ (H − E) ϕ = O( ). There is no way to improve the order of approximation simply by making a better choice of S.2 The WKB approximation Returning to our WKB ansatz for a stationarystate solution of the Schr¨dinger equation. i. A function H : R2n → R is locally constant on a lagrangian submanifold L ⊂ R2n if and only if the hamiltonian vector ﬁeld XH is tangent to L. in fact. skewsymmetric bilinear form. v) = −dH(v) for every tangent vector v. as we will see in the next chapter. and there is no reason for this to be constant. this theorem implies that L is invariant under the ﬂow of XH . If the lagrangian submanifold L is locally closed. and thus. ϕ is an eigenfunction of H with eigenvalue E. in other words. 11 . We may therefore hope to ﬁnd a better approximate solution by multiplying ϕ by an “amplitude function” a ϕ(x) = eiS(x)/ a(x). Recall that to a function H : R2n → R. and we have the following result. or. In quantum mechanics.For now. subspaces of Tp R2n on which dαn vanishes can be at most ndimensional. the 2form dαn vanishes on the subspace of Tp R2n generated by Tp L and XH (p) for each p ∈ L. Hamilton’s equations associate the vector ﬁeld XH = q ˙ ∂ ∂ +p ˙ = ∂q ∂p ∂H ∂ ∂H ∂ − . o n we recall that if S : R → R is an admissible phase function. we want to note an important relationship between lagrangian submanifolds of R2n and hamiltonian ﬂows. then T L lies in the kernel of dH at all points of L. If L is a lagrangian submanifold of a level set of H. it is at least intuitively plausible that a particle is more likely to be found where it moves more slowly.
we obtain a2 S + 2aa S = (a2 S ) = 0 c ⇒ a= √ S for some constant c. [4(E − V )] 4 If E > V (x) for all x ∈ R. the function a assumes complex values. for example. Consider. ∂xj ∂xj j If S is an admissible phase function. Notice that a = ϕ is largest where V is largest. we can extend the procedure above by adding to the original amplitude a = a0 certain appropriately chosen functions of higher order in . ˆ If a is chosen to kill the coeﬃcient of on the right. then ϕ will be an eigenfunction of H modulo terms of order O( 2 ). the unbounded potential V (x) = √ x2 in the case of the harmonic oscillator. 12 . the homogeneous transport equation amounts to aS + 2a S = 0. we have S = 2m(E − V ). This condition on a is known as the homogeneous transport equation: ∂a ∂S a∆S + 2 = 0. then the secondorder solution ϕ = eiS/ a is called the semiclassical approximation. we now obtain: ˆ (H − E) ϕ = − 1 i 2m a∆S + 2 j ∂a ∂S ∂xj ∂xj + 2 ∆a eiS/ .1 In 1 dimension. the function a is still welldeﬁned √ up to a multiplicative constant.If S is again an admissible phase function. Solving this equation directly. Consider the next approximation ϕ = eiS/ (a0 + a1 ). For x < E. observe that these occur precisely at the projected image of those points of L where the √ projection itself becomes singular. then a is a smooth solution to the homogeneous transport equation. however. Since S is assumed to satisfy the HamiltonJacobi equation. and thus c a= 1 . At x = E. Example 2. To generate better approximate solutions to the eigenfunction problem. a has (asymptotic) singularities. Outside the interval x ≤ E. Since the expression above for a does not depend explicitly on the phase function S. and a is an amplitude which satisﬁes the homogeneous transport equation. we might naively attempt to use the same formula when im(dS) is replaced by a nonprojectable lagrangian submanifold of H −1 (E). as our physical reasoning predicted.
the diﬀerence K f − k=0 ak k is O( K+1 ) locally uniformly in x. where is viewed as a parameter ranging in R+ . we obtain: ˆ (H − E) ϕ = − 1 i 2m 2 a1 ∆S + 2 j ∂a1 ∂S − i∆a0 ∂xj ∂xj + 3 ∆a1 eiS(x)/ .Borel (see [28. The order of a is the index of its principal part. conversely. Such a function is said to be represented by a formal asymptotic expansion of the form ∞ ak k . ∂xj ∂xj n In general. It is obvious that any dependent function which extends smoothly to = 0 is represented by an asymptotic series. p. A WKB ˆ “solution” to the eigenfunction problem Hϕ = Eϕ is then an equivalence class of functions of the form ϕ = eiS/ a. Evidently.Assuming that eiS/ a0 is a semiclassical approximate solution. O( k ) for all k. if.e. where S is an admissible phase function and a is an dependent function represented by a formal asymptotic series ∞ a∼ k=p ak k 13 . a0 satisﬁes the homogeneous transport equation. where each coeﬃcient ak is a smooth k=0 complexvalued function on Rn . the function ak satisﬁes the inhomogeneous transport equation: ak ∆S + 2 j ∂ak ∂S = i∆ak−1 . a solution to the eigenfunction problem modulo terms of order O( by a WKB ansatz of the form ϕ = eiS/ (a0 + a1 + · · · + an n ) is given ). for each K ∈ Z+ . i. ∂xj ∂xj This situation can be described in the terminology of asymptotic series as follows. and for each k > 0. The principal part of an asymptotic series ∞ ak k is deﬁned as its ﬁrst term which is k=0 not identically zero as a function of x. where S satisﬁes the HamiltonJacobi equation. When f admits such an expansion. then each asymptotic series determines a unique equivalence class. its coeﬃcients ak are uniquely determined. If we consider as equivalent any two dependent functions whose diﬀerence is O( ∞ ). and a theorem of E.28]) tells us that. any asymptotic series can be “summed” to yield such a function. ϕ will be a solution of the timeindependent Schr¨dinger equation modulo terms o of order O( 3 ) provided that a1 satisﬁes the inhomogeneous transport equation a1 ∆S + 2 j ∂a1 ∂S = i∆a0 . By an dependent function f on Rn we will mean a function f : Rn × R+ → C.
suppose that a is a function on Rn which satisﬁes the homogeneous transport equation: ∂a ∂S = 0.e. we can rewrite it as: ∂ ∂xj a2 ∂S ∂xj = 0. Notice ﬁrst that the restriction to L of the hamiltonian vector ﬁeld associated to H(q. a function S satisfying the HamiltonJacobi equation H(x.1. we saw that a ﬁrstorder WKB approximate solution ϕ = eiS/ to the timeindependent Schr¨dinger equation depended on the choice of an admissible phase function. we can lift all of this activity to the lagrangian submanifold L = im(dS).. H 14 . and so the homogeneous transport equation says that a2 XH is divergencefree for the canonical density dx = dx1 ∧ · · · ∧ dxn  on Rn . o Geometry of the transport equation In Section 2. We now wish to interpret and generalize in a similar way the semiclassical approximation with its amplitude included. p) = p2 /2 + V (q) is i XH L = j (x) ∂S ∂ ∂V ∂ − ∂xj ∂qj ∂qj ∂pj . ∂S ) = E. The projection XH of XH L onto Rn (the (x) reminds us of the coordinate x on Rn ) there(x) fore coincides with S. The generalized or ∂x geometric version of such a solution was a lagrangian submanifold of the level set H −1 (E). j which means that the divergence of the vector ﬁeld a2 S is zero. To begin. a∆S + 2 ∂xj ∂xj j After multiplying both sides of this equation by a. But it is better to reformulate this condition as: LX (x) (a2 dx) = 0. ∂xj ∂xj and for k > p. ∂xj ∂xj This means that the dependent function ϕ (or any function equivalent to it) satisﬁes the Schr¨dinger equation up to terms of order O( ∞ ). Rather than considering the transport equation as a condition on the vector ﬁeld a2 S (on Rn ) per se. the ak satisfy the recursive transport equations: ak ∆S + 2 j ∂ak ∂S = i∆ak−1 .with the property that its principal part ap satisﬁes the homogeneous transport equation ap ∆S + 2 j ∂ap ∂S = 0. o i.
(x) Notice that while a solution to the homogeneous transport equation in the case of the 1dimensional harmonic oscillator was necessarily singular (see Example 2.1). Any such level curve L.2 Recall that in the case of the 1dimensional harmonic oscillator. a (geometric) semiclassical state should be deﬁned as a lagrangian submanifold L of R2n equipped with a halfdensity a.) In other words. which we will denote o ˆ . namely that the quantum states themselves should be represented. p) = (p2 + kq 2 )/2. invariant by XH .e. elements of the intrinsic Hilbert space HRn (see Appendix A). suggests that a solution of the homogeneous transport equation should be represented geometrically by a halfdensity on L. we transfer the factor of a2 from the vector ﬁeld XH = S to the density dx. There is a unique (up to a constant) invariant volume element for the hamiltonian ﬂow of H on each level curve H. not by functions. Example 2. stationary classical states are simply those lagrangian submanifolds of R2 which coincide with the regular level sets of the classical hamiltonian H(q. the semiclassical state described in the preceding example is a perfectly smooth object everywhere on the lagrangian submanifold L.that is. and since the Lie derivative is invariant under diﬀeomorphism. we can express the result of our analysis as follows: If S is an admissible phase function and a is a halfdensity on L = im(dS) which is invariant under the ﬂow of the classical hamiltonian. Such a state is stationary when L lies in a level set of the classical hamiltonian and a is invariant under its ﬂow. then eiS/ (dS)∗ a is a secondorder approximate solution to the timeindependent Schr¨dinger equation. From this new point of view. by the equation momentarily as Hfun ˆ ˆ H(adx1/2 ) = (Hfun a)dx1/2 . constitutes a semiclassical stationary state for the harmonic oscillator. (See Appendix A for a discussion of densities of fractional order. o 15 . but by halfdensities on conﬁguration space Rn . This observation. This discussion leads us to another change of viewpoint. i. this equation is satisﬁed if and only if the pullback of a2 dx to L via the projection π is invariant under the ﬂow of XH . Since XH is tangent to L by the HamiltonJacobi theorem. Another substantial advantage of the geometric interpretation of the semiclassical approximation is that the concept of an invariant halfdensity depends only on the hamiltonian vector ﬁeld XH and not on the function S. which is deﬁned on the space of smooth o halfdensities in terms of the old Schr¨dinger operator on functions. so it makes sense on any lagrangian submanifold of R2n lying in a level set of H. together with a square root of this volume element. Stationary states ˆ are then eigenvectors of the Schr¨dinger operator H. The singularities arise only when we try to transfer the halfdensity from L down to conﬁguration space. together with the fact that it is the square of a which appears in the density π ∗ (a2 dx).
we have noted the following correspondences between classical and quantum mechanics: Object basic space state Classical version R2n Quantum version H Rn lagrangian submanifold of R2n with halfdensity on Rn halfdensity Hamilton’s equations Schr¨dinger equation o ˆ operator H on smooth halfdensities ˆ eigenvector of H timeevolution generator of evolution function H on R2n stationary state lagrangian submanifold in level set of H with invariant halfdensity Proceeding further. o however. In particular. Speciﬁcally. which is introduced in the following chapter. we will extend the geometric picture presented above to systems with more general phase spaces. we could attempt to interpret a solution of the recursive system of inhomogeneous transport equations on Rn as an asymptotic halfdensity on L in order to arrive at a geometric picture of a complete WKB solution to the Schr¨dinger equation. Instead. 16 . we will try to construct a space of quantum states corresponding to a general classical phase space. This will require the concept of symplectic manifold.In summary. Second. involves some additional diﬃculties. we will focus on two aspects of the semiclassical approximation. we will start with an invariant halfdensity on a (possibly nonprojectable) lagrangian submanifold of R2n and attempt to use this data to construct an explicit semiclassical approximate solution to Schr¨dinger’s o equation on Rn . First. notably the lack of a geometric interpretation of the inhomogeneous transport equations. which lie beyond the scope of these notes. This. Then we will try to construct asymptotic quantum states corresponding to halfdensities on lagrangian submanifolds. we will “quantize” semiclassical states in these symplectic manifolds.
or in other words if V ⊥ = {0}. A linear endomorphism of a symplectic vector space (V. we ﬁrst study such forms pointwise. collecting pertinent facts about nondegenerate.e. ˜ If ω is an isomorphism. and the group of all such transformations is denoted by Sp(V ). skewsymmetric bilinear forms. 17 . Finally. ˜ The ωorthogonal to a subspace W ⊂ V is deﬁned as W ⊥ = {x ∈ V : W ⊂ ker ω (x)}. we saw that Hamilton’s equations associate a vector ﬁeld XH on R2n satisfying XH dαn = −dH. our proof of the HamiltonJacobi theorem relied on the nondegeneracy of the 2form dαn . 3. More general symplectic manifolds will reappear as the focus of more sophisticated quantization programs in later chapters. to cotangent bundles. In this section. 29. then ω is called a linear symplectic structure on V .1 Symplectic structures In Section 2. We refer to [6. ω) which preserves the form ω is called a linear symplectic transformation.3 Symplectic Manifolds In this chapter. y). 63] for thorough discussions of the topics in this chapter. to a function 2n H : R → R.1. a lagrangian submanifold of R2n was deﬁned as an ndimensional submanifold L ⊂ R2n on which the exterior derivative of the form αn = pi dqi vanishes. the correct generalization of the hamiltonian picture to arbitrary conﬁguration spaces relies similarly on the use of 2forms with certain additional properties. A bilinear form ω : V × V → R gives rise to a linear map ω: V → V ∗ ˜ deﬁned by contraction: ω (x)(y) = ω(x. i. we will introduce the notion of a symplectic structure on a manifold. then the form ω is said to be nonde˜ generate. We then turn to the deﬁnition of symplectic manifolds. While some discussion will be devoted to certain general properties of symplectic manifolds. motivated for the most part by the situation in R2n . These points already indicate the central role played by the form dαn in the study of hamiltonian systems in R2n . Linear symplectic structures Suppose that V is a real. if in addition ω is skewsymmetric. our main goal at this point is to develop the tools needed to extend the hamiltonian viewpoint to phase spaces associated to general ﬁnitedimensional conﬁguration spaces. mdimensional vector space.
e. J·) deﬁnes a symmetric. J is compatible with ω (we also call it ωcompatible) if J : V → V is a linear symplectomorphism and gJ (·. 18 . (K −1 )∗ λ) deﬁnes a linear symplectomorphism between E ⊕ E ∗ and F ⊕ F ∗ equipped with these linear symplectic structures. Note in particular that these conditions are satisﬁed if A ∈ GL(E). ·) = ω(·. With respect to a basis {xi } of E and a dual basis {λi } of E ∗ . ω) admits a compatible complex structure. then the association (x.Example 3. λ )) = λ (x) − λ(x ). then T is symplectic provided that At C. positivedeﬁnite bilinear form on V . the form ω is represented by the matrix 0 I ω= . D = (At )−1 . and so GL(E) is isomorphic to a subgroup Gl(E) of Sp(V ). and B = C = 0. then a linear symplectic structure on V = E ⊕ E ∗ is given by ω((x.2 Every symplectic vector space (V. B t D are symmetric. y) and ω(x. λ). Jy) = ω(x. so that ω is represented by an invertible skewadjoint operator K : V → V . λ) → (Kx. y ∈ V . y) = Kx. Theorem 3.e. a linear endomorphism J such that J 2 = −I. ω(x. Let . be a symmetric. A complex structure is said to be compatible with a symplectic structure on V if ω(Jx. More generally. if K : E → F is an isomorphism. i. −I 0 It follows that if a linear operator on V is given by the real 2n × 2n matrix T = A B C D . Since the determinant of a skewsymmetric m × m matrix is zero if m is odd. Jx) > 0 for all x. the existence of a linear symplectic structure on a vector space V implies that V is necessarily evendimensional and therefore admits a complex structure. positivedeﬁnite inner product on V .1 If E is any real ndimensional vector space with dual E ∗ . Proof. In other words. i. y . (x . and At D − C t B = I.
2 Corollary 3. ·). J).. g. this implies the corollary. · = gJ (·. O(V ). and so J 2 = −JJ t = − id. since R and . and ·. Proof. J. w ∈ V . ﬁrst note that ω(Jx. J = R−1 K is orthogonal. the complex structure J constructed in the preceding proof is continuous. Jx = JRx. 2 If J is ωcompatible. and the composition J → P → J of these maps equals the identity on J . ·) + iω(·. J). Aw). As is easily checked. Jy) = KJx. a linear transformation T ∈ GL(V ) which preserves any two of the structures ω. etA w) = ω(Av. a hermitian structure on V is deﬁned by ·. consider a 1parameter family of maps etA associated to some linear map A : V → V . O(V ) equals U (V ). J is a complex structure on V . Jy = JKx. · . t=0 and so A ∈ sp(V) if and only if the linear map ω ◦ A : V → V ∗ is selfadjoint. The association J → gJ described above deﬁnes a continuous map from J into the space P of symmetric. ω(x. w) + ω(v. Jy = Kx. By the uniqueness of the polar decomposition. positivedeﬁnite bilinear forms on V . y). GL(V. Also. ˜ dim(V ) = 2k implies dim(sp(V)) = dim(Sp(V)) = k(2k + 1). Jx) = Kx. GL(V. ω) is contractible.√ The operator K admits a polar decomposition K = RJ. From the nondegeneracy of the symplectic form. we have d dt ω(etA v. For any v. From the skewsymmetry of K it follows that J t = −J.e. ω) is called the symplectic orthogonal to W . it follows that the map which assigns to a form . and U (V ) of ω. i. x > 0. Distinguished subspaces The ωorthogonal to a subspace W of a symplectic vector space (V. gJ . it follows that W ⊥⊥ = W and dim W ⊥ = dim V − dim W 19 . and RJ = JR. To determine the Lie algebra sp(V) of Sp(V ). J on V preserves the third and therefore preserves the hermitian structure. where R = KK t is positivedeﬁnite symmetric. To see that J is ωcompatible. this means that the intersection of any two of Sp(V ). Jx = Rx. Since P is contractible. y = ω(x. Consequently.3 The collection J of ωcompatible complex structures on a symplectic vector space (V. are positivedeﬁnite. In terms of the automorphism groups Sp(V ).
Various subspaces of a symplectic vector space are related as follows. we denote by V ⊕V the vector space V ⊕V equipped with the symplectic structure ω ⊕ −ω. then W ⊥ is 1dimensional. there exists a nonzero vector w ∈ W ⊥ \W . i. W ⊥ = W . Since 2n = dim(W ) + dim(W ⊥ ). and is called lagrangian. ˜ Example 3. Also. Beginning with any 1dimensional subspace of V . then the graph of T is a lagrangian subspace of V ⊕ V . For instance. the restriction of ω to certain subspaces Z ⊂ V may again be nondegenerate. Any selforthogonal subspace is simultaneously isotropic and coisotropic. the skewsymmetry of ω implies that W ⊂ W ⊥ . From this observation it follows that for every isotropic subspace W of a (ﬁnitedimensional) symplectic vector space V which is not lagrangian. Lemma 3. Note that the symplectic orthogonal W ⊥ might not be an algebraic complement to W . Finally. ω) is a symplectic vector space. At the other extreme. B of V . The subspace W of V spanned by W ∪{w} is then isotropic and dim(W ) = dim(W )+1. we note that if codim W = 1. there exists an isotropic subspace W of V which properly contains W . and W ⊥ ⊂ W ⊥⊥ = W . The kernel of a nonzero covector α ∈ V ∗ is a codimension1 coisotropic subspace ker α of V whose symplectic orthogonal (ker α)⊥ is the distinguished 1dimensional subspace of ker α spanned by ω −1 (α). (A + B)⊥ = A⊥ ∩ B ⊥ and (A ∩ B)⊥ = A⊥ + B ⊥ for any pair of subspaces A. spaces W satisfying the condition W ⊥ ⊂ W are called coisotropic or involutive. If (V. then: 20 . both E and E ∗ are lagrangian subspaces. any subspace contained in its orthogonal will be called isotropic. Example 3. a subspace W ⊂ V is isotropic if the restriction of the symplectic form to W is identically zero. then the dimension relation above implies that 1 dim W = 2 dim V . and A ⊂ V is an arbitrary subspace.5 Suppose that (V.6 If L is a lagrangian subspace of a symplectic vector space V . In general. According to these deﬁnitions. if dim W = 1. Such subspaces are called symplectic. hence isotropic.4 In E ⊕ E ∗ with its usual symplectic structure. It also follows from the deﬁnition of this structure that the graph of a linear map B : E → E ∗ is a lagrangian subspace of E ⊕ E ∗ if and only if B is selfadjoint. we can apply this remark inductively to conclude that every ﬁnitedimensional symplectic vector space contains a lagrangian subspace.e. If T : V → V is a linear symplectic map. this is equivalent to saying that Z ∩Z ⊥ = {0} or Z + Z ⊥ = V . B ⊥ ⊂ A⊥ whenever A ⊂ B. if W is selforthogonal. ω) is a 2ndimensional symplectic vector space and W ⊂ V is any isotropic subspace with dim(W ) < n. More generally. In particular.for any subspace W ⊂ V . Dually.
Theorem 3. which in turn ∗ gives rise to a linear symplectomorphism between V and L ⊕ L equipped with its canonical symplectic structure (see Example 3. we have L ⊂ A.5 implies that there is a lagrangian subspace L with A⊥ ⊂ L. and so W ⊥ ∩ I ⊂ I ⊥ ∩ L. Tε y) = (1 + ε2 ) ω(x. if A is coisotropic. ωn ). the map ω deﬁnes an isomorphism L ˜ L∗ . If J is a ωcompatible complex structure on V and L ⊂ V a lagrangian subspace. we can choose orthonormal bases 21 . a subspace C ⊂ V is coisotropic if and only if it contains a lagrangian subspace.e.6 implies that if L ⊂ A. L ⊂ A if and only if A⊥ ⊂ L. W ⊥ ∩ I ⊂ I ∩ L = 0. A pair L.1. If L. then L. Similarly. then A⊥ ⊂ A. Statement (1) follows from the properties of the operation ⊥ and the equation L = L⊥ . By Example 3. For any two lagrangian subspaces L. there is a lagrangian subspace L transverse to both L and L. If W ⊂ L is any complementary subspace to I ⊥ ∩ L. 2. Suppose that V is a symplectic vector space with an isotropic subspace I and a lagrangian subspace L such that I ∩ L = 0. it follows that I + W is a symplectic subspace of V .2 also implies the following useful result. Passing to orthogonals.1). L are any lagrangian subspaces of V . then I + L ⊂ W + I ⊥ . 2. Conversely.5. proving statement (2). It is easy to check that Tε is a conformal linear symplectic map. and since every ndimensional vector space is isomorphic to Rn .8 Every 2ndimensional symplectic vector space is linearly symplectomorphic to (R2n . Lemma 3. L of V . Thus. JL is a lagrangian splitting. Thus. L + A = V if and only if (L + A)⊥ = L ∩ A⊥ = {0}. Thus. Proof. every symplectic vector space contains a lagrangian subspace. Using the innerproduct gJ on V induced by J. Proof. and so A is a coisotropic subspace.7 Note that statement (1) of Lemma 3. i. then A⊥ is isotropic. In this case. the preceding remarks prove the following linear “normal form” result: Theorem 3. and Example 3. Since I ⊥ ∩ W = 0 by our choice of W . L is transverse to A if and only if L ∩ A⊥ = {0}. then Lε = Tε (L) is a lagrangian subspace transverse to L for small ε > 0. 1. L of transverse lagrangian subspaces of V is said to deﬁne a lagrangian splitting of V . 2 Example 3. y).9 Suppose that V is a symplectic vector space with a ωcompatible complex structure J and let Tε : V → V be given by Tε (x) = x + εJx. an isomorphism of V satisfying ω(Tε x. Lε is a lagrangian subspace for all ε > 0.
wj ) + ε gJ (vi . Proof. wj + εJwj )} = {ω(vi . the lagrangian subspace Lε is transverse to L. wj )}n i.1). To prove (2). Then {wi + εJwi } form a basis of Lε . L by (1). which induces a ﬁbration L(V ) → S 1 with 1connected ﬁber SU (n)/SO(n). 2 In fact. {wi } of L and L.j=k+1 and B is some (n−k)×(n−k) matrix. giving an isomorphism of fundamental groups π1 (L(V )) π1 (S 1 ) Z. the function t → det(Ai + tBi ) is a nonzero polynomial and therefore has ﬁnitely many zeros. Our choice of bases implies that ε · id 0 M= 0 A+ε·B n where A = {ω(vi .{vi }. and L . Thus. Consequently. respectively. and let Ai . · · · . Bi be the matrices obtained with respect to L as in the proof above. the statement of preceding lemma can be improved as follows. the lagrangian subspace Tt (L) is transverse to all Li for almost every t ∈ R. Setting I = span{vi }i=k+1 and W = span{wi }n i=k+1 . Lemma 3. L2 ∈ L(V ). and assertion (1) follows. L) = L (T ) = T (L). observe that for small ε > 0. k. For each i. J The choice of J also deﬁnes a complex determinant U (V ) → S 1 . wj )} is nonsingular.1. an orthogonal transformation L1 → L2 induces a symplectic transformation L1 ⊕ L∗ → L2 ⊕ L∗ in the manner of Example 3. A natural action of the group Sp(V ) on L(V ). so that for i = 1. we can apply Example 3. For arbitrary L1 . 2 The stabilizer of L ∈ L(V ) under the U (V )action is evidently the orthogonal subgroup of Gl(L) deﬁned with respect to the innerproduct and splitting L ⊕ JL of V induced by J (see Example 3.7 to conclude that A is nonsingular. a (noncanonical) identiﬁcation of the lagrangian grassmannian with the homogeneous space U (n)/O(n) is obtained from the map L U (V ) → L(V ). which in turn 1 2 gives rise to a unitary transformation L1 ⊕ JL1 → L2 ⊕ JL2 mapping L1 onto L2 . det2 22 . the vectors vi = wi span L ∩ L . Lε are transverse precisely when the matrix M = {ω(vi . denoted : Sp(V ) × L(V ) → L(V ) is deﬁned by (T. The lagrangian grassmannian The collection of all unoriented lagrangian subspaces of a 2ndimensional symplectic vector space V is called the lagrangian grassmannian L(V ) of V .10 The unitary group associated to an ωcompatible complex structure J on V acts transitively on L(V ). Let {Li } be a countable family of lagrangian subspaces.
f2 : M → L(V ) are continuous maps. we ﬁrst recall some features of the n diﬀerential form −dαn = ωn = j=1 dqj ∧ dpj which appeared in our earlier discussion. the multiplication map S 1 × S 1 → S 1 induces the diagonal map H 1 (S 1 ) → H 1 (S 1 ) ⊕ H 1 (S 1 ) H 1 (S 1 × S 1 ) on cohomology). then a check of the preceding deﬁnitions shows that mL ((T. connectedness of the unitary group together with Lemma 3.3). then ∗ ∗ the deﬁnition of the universal Maslov class shows that (f1 − f2 )µV equals the pullback of 1 1 the canonical generator of H (S .This isomorphism does not depend on the choices of J and L made above. R) by the map (mL ◦ f1 )(mL ◦ f2 )−1 . Independence of J follows from the fact that J is connected (Corollary 3. fi ). The image of the canonical generator of H 1 (S 1 . Z) → H 1 (L(V ).11 If (V. µV . If f1 . ∂qj ∂qk 23 =0 ωn ∂ ∂ . Z) under this map is called the universal Maslov class. From the ﬁrst paragraph. L )) = mL (L ) · det2 (T ) J for any T ∈ U (V ) and L ∈ L(V ). Passing to homology and dualizing. If T : M → Sp(V ) is any map. ∂pj ∂pk =0 . Example 3. Symplectic manifolds To motivate the deﬁnition of a symplectic manifold. it follows that this product equals (mL ◦f1 )(mL ◦f2 )−1 . On the other hand. it follows that T is homotopic to a map T : M → U (V ). ω) is any symplectic vector space with ωcompatible complex structure J and lagrangian subspace L. n First. (Recall that : Sp(V )×L(V ) → L(V ) denotes the natural action of Sp(V ) on L(V )). we set (T · fi ) = (T. The result of the following example will be useful when we extend our discussion of the Maslov class from vector spaces to vector bundles. (Here we use the fact that when S 1 is identiﬁed with the unit complex numbers. ∂qj ∂pk = δjk ωn ∂ ∂ . from which we conclude that ∗ ∗ (f1 − f2 )µV = ((T · f1 )∗ − (T · f2 )∗ )µV . Z). In fact: ωn ∂ ∂ . we obtain a natural homomorphism H 1 (S 1 .10 gives independence of L. Since Sp(V ) deformation retracts onto U (V ). and so ((T · f1 )∗ − (T · f2 )∗ )µV is obtained via pullback by (mL ◦ (T · f1 ))(mL ◦ (T · f2 ))−1 . Now consider any topological space M . we note that j=1 dqj ∧ dpj deﬁnes a linear symplectic structure on the tangent space of R2n at each point.
24 . Here. LXH H = XH · H = ωn (XH . we recall that the hamiltonian vector ﬁeld associated via Hamilton’s equations to H : R2n → R satisﬁes XH ωn = dH. Since ωn is closed. This equation implies that the ﬂow of XH preserves the form ωn and therefore generalizes our earlier remark that the hamiltonian vector ﬁeld associated to the 1dimensional harmonic oscillator is divergencefree. This again reﬂects the fact that the ﬂow of XH preserves energy. XH = ωn (dH).and so ωn ˜ ∂ ∂qj = dpj ωn ˜ ∂ ∂pj = −dqj . nondegenerate 2form ω on P . from which it is clear that ωn is invertible. the term d(XH ω) = d2 H is automatically zero. XH ) = 0. ˜ Next. we impose the condition that ω be closed. We now see what is needed to do hamiltonian mechanics on manifolds. By assuming that each of these bilinear forms is nondegenerate. to guarantee the vanishing of the second term. ˜ we see that the conservation of energy follows from the skewsymmetry of the form ω. XH ) = 0. Computing the Lie derivative of H with respect to XH LXH H = XH · H = ω (XH )(XH ) = ω(XH .12 A symplectic structure on a manifold P is a closed. ˜ −1 so we see that the symplectic form ωn is all that we need to obtain XH from H. we also have by Cartan’s formula (see [1]) LXH ωn = d(XH ωn ) + XH dωn = d2 H = 0. we guarantee that the equation XH = ω −1 (dH) ˜ deﬁnes a hamiltonian vector ﬁeld uniquely for any H. This description of the hamiltonian vector ﬁeld leads immediately to the following two invariance results. invariance of ω under the hamiltonian ﬂow is satisﬁed if LXH ω = d(XH ω) + XH dω = 0. or in other words. implying that XH is tangent to the level sets of H. Thus we make the following deﬁnition: Deﬁnition 3. First note that by the skewsymmetry of ωn . Finally. A 2form ω on a manifold P is a smooth family of bilinear forms on the tangent spaces of P .
Example 3. ω1 ) to (P2 . ι) is a lagrangian immersion whose image is contained in C. the characteristic subspace Cι(p) ⊂ Tι(p) C is contained in ι∗ Tp L. Generalizing our earlier discussion of distinguished subspaces of a symplectic vector space. we call a submanifold C ⊂ P (co)isotropic provided that each tangent space Tp C of C is a (co)isotropic subspace of Tp P . their product P1 ×P2 admits a symplectic structure given by the sum ω1 ⊕ω2 . Lemma 3. which are (co)isotropic submanifolds of dimension 1 dim(P ). we will see that the cotangent bundle of any smooth manifold carries a natural symplectic structure. nonsingular vector ﬁeld XH. If (L. ι) a lagrangian immersion. this form is the sum of the pullbacks of ω1 and ω2 by the projections of P1 × P2 to P1 and P2 . Darboux’s theorem (Section 4.The condition that ω be nondegenerate means that ω deﬁnes an isomorphism of vector ˜ bundles T P → T ∗ P . that the top exterior power of ω is a volume form on P . In this case. the subspaces Cp = (Tp C)⊥ comprise a subbundle (T C)⊥ of T C known as the characteristic distribution of C. ω1 ) and (P2 . The collection Aut(P. In the next section. then C is a coisotropic submanifold. More explicitly. or equivalently. ω) of symplectomorphisms of P becomes an inﬁnitedimensional Lie group when endowed with the C ∞ topology (see [49]).13 If C ⊂ P is a hypersurface. More generally.3) will tell us that this is the local model for the general case. ω2 ). this assertion generalizes the HamiltonJacobi theorem (see the end of Section 2. Of particular interest in our discussion will be lagrangian submanifolds of P . ω2 ) is a smooth diﬀeomorphism f : P1 → P2 compatible with the symplectic structures: f ∗ ω2 = ω1 . and symplectomorphisms is provided by the following lemma. that ω deﬁnes a linear symplectic structure on each tangent space of P . A useful connection among duals. ω) consists of the same underlying manifold endowed with the symplectic structure −ω. New symplectic manifolds can be manufactured from known examples by dualizing and by taking products. In view of the remarks above. the corresponding Lie 25 . and thus XH induces a smooth. A simple check of deﬁnitions shows that if H : P → R is a smooth function having C as a regular level set. then the hamiltonian vector ﬁeld XH is tangent to the characteristic foliation of C. then Lemma 3. A symplectomorphism from (P1 . products. An immediate example of a symplectic manifold is furnished by R2n with its standard structure ωn = n dqj ∧ dpj (a diﬀerential form with constant coeﬃcients and not just j=1 a single bilinear form). It is integrable because ω is closed. Evidently P and its dual P share the same (co)isotropic submanifolds. or ﬁnally. the product of (co)isotropic submanifolds of P1 and P2 is a (co)isotropic submanifold of P1 × P2 . Given two symplectic manifolds (P1 . As is easily veriﬁed.1) to arbitrary symplectic manifolds and lagrangian immersions. When C is coisotropic.6 implies that for each p ∈ L. The symplectic dual of a manifold (P.ι on L.14 A diﬀeomorphism f : P1 → P2 between symplectic manifolds is a symplectomorphism if and only if its graph is a lagrangian submanifold of the product P2 × P 1 . 2 we will call the pair (L. if L is a 2 smooth manifold of dimension 1 dim(P ) and ι : L → P is an immersion such that ι∗ ω = 0.
deﬁned in analogy with Example 3. the proof of Theorem 3. More generally.algebra is the space χ(P. ω) of smooth vector ﬁelds X on P satisfying LX ω = 0. Example 3.16 If F → B is any vector bundle.) Note. v). any vector bundle E → B with this structure is called a symplectic vector bundle. 26 . 2 Symplectic vector bundles Since a symplectic form on a 2nmanifold P deﬁnes a smooth family of linear symplectic forms on the ﬁbers of T P . i. Example 3.17.18 Despite Theorem 3. (See [27] and the numerous earlier references cited therein. With the aid of an arbitrary riemannian metric. ω) consisting of vector ﬁelds of the form X(v) = Av for some A ∈ sp(V). ω) are called locally hamiltonian vector ﬁelds or symplectic vector ﬁelds. Example 3. there exist examples of symplectic manifolds which are not complex (the almost complex structure coming from the theorem cannot be made integrable). the frame bundle of P can be reduced to a principal Sp(n) bundle over P . and the Lie algebra sp(V) identiﬁes with the subalgebra of χ(V. however. Consequently. ω).15 A linear symplectic form ω on a vector space V induces a symplectic structure (also denoted ω) on V via the canonical identiﬁcation of T V with V ×V . functions satisfying Q(tv) = t2 Q(v) for all real t. F are said to be symplectomorphic if there exists a vector bundle isomorphism E → F which preserves their symplectic structures. Two symplectic vector bundles E.e. ω) and the space of closed 1forms on P . The symplectic group Sp(V ) then embeds naturally in Aut(V. then the sum F ⊕ F ∗ carries a natural symplectic vector bundle structure. Since LX ω = d(X ω). that the K¨hler form of any K¨hler manifold a a is a symplectic form. the association X → X ω deﬁnes an isomorphism between χ(P.2 can be generalized by a ﬁberwise construction as follows.17 Every symplectic vector bundle admits a compatible complex vector bundle structure. sp(V) is canonically identiﬁed with the space of such functions via the correspondence 1 A ↔ QA (v) = ω(Av. Note that these are precisely the hamiltonian vector ﬁelds of the homogeneous quadratic polynomials on V . The elements of χ(P. those X which map to exact 1forms are simply the hamiltonian vector ﬁelds on P . Theorem 3.1. and of complex manifolds which are not symplectic.
Z) is the universal Maslov class. For nontrivial E.A lagrangian subbundle of a symplectic vector bundle E is a subbundle L ⊂ E such that Lx is a lagrangian subspace of Ex for all x ∈ B. A simple check of the deﬁnition then shows that J = T −1 J0 T is a compatible complex structure on E which satisﬁes JL = L . Proof. we denote by fL . L are lagrangian subbundles such that Lx is transverse to Lx for each x ∈ B. then E is symplectomorphic to L ⊕ L∗ . γ ∗ L ). In general. the pullback bundle γ ∗ E is trivial. γ ∗ µ(L. and T L ⊂ TL P is a lagrangian subbundle. the automorphism group of a symplectic vector bundle E does not act transitively on the lagrangian subbundles of E. Also note that if C ⊂ P is any submanifold such that T C contains a lagrangian subbundle of TC P .19 If L is a lagrangian submanifold of a symplectic manifold P . Example 3. L ) is welldeﬁned by the requirement that for every smooth loop γ in M . then the restricted tangent bundle TL P is a symplectic vector bundle over L. it follows that for any loop γ : S 1 → M . L ) ∈ H 1 (M . we note that since the symplectic group Sp(V ) is connected.20 Let E → B be a symplectic vector bundle and suppose that L.6). Nevertheless. L ) = µ(γ ∗ L. Theorem 3. fL : M → L(V ) the maps induced by the lagrangian subbundles f (L). 2 Example 3. a pair of transverse lagrangian subbundles can be related as follows. If E admits a lagrangian subbundle. then any pair L. Let J0 be any compatible complex structure on E. Since L and J0 L are both transverse to L. Thus µ(L. L of lagrangian subbundles of E deﬁne a cohomology class µ(L. 27 . Z) as follows. and the frame bundle of E admits a further reduction to a principal GL(n) bundle over B (compare Example 3. where µV ∈ H 1 (L(V ). Then there exists a compatible complex structure J on E satisfying JL = L .21 If E is a symplectic vector bundle over M . f (L ) of M × V . From Example 3. Assuming ﬁrst that E admits a symplectic trivialization f : E → M × V for some symplectic vector space V . then C is coisotropic (see Lemma 3.11 it follows that this class is independent of the choice of trivialization f . we can ﬁnd a symplectomorphism T : L ⊕ L → L ⊕ J0 L which preserves the subbundle L and maps L to J0 L. Then ∗ ∗ µ(L.1). L ) = (fL − fL ) µV .
p1 . pn ) on T ∗ M . · · · . Finally. where π : T ∗ Rn → Rn is the natural projection. (See [3] for an interpretation of the Maslov class of a loop γ in L as an intersection index of the loop G ◦ γ with a singular subvariety in the lagrangian grassmannian). If ι : L → R2n is a lagrangian immersion.2 Cotangent bundles The cotangent bundle T ∗ M of any smooth manifold M is equipped with a natural 1form. −dαM = dqj ∧ dpj in these coordinates. ι) is said to be projectable if πL is a diﬀeomorphism. To do this.Example 3. and αM = αn . we consider the symplectic manifold R2n T ∗ Rn with its standard symplectic structure. from which it follows that the form ωM = −dαM is a symplectic structure on T ∗ M . Lagrangian immersions and the Liouville class Given a lagrangian immersion ι : L → T ∗ M . where π : T ∗ M → M is the canonical projection. Then the tangent bundle T (R2n ) is a symplectic vector bundle over R2n with a natural “vertical” lagrangian subbundle V Rn deﬁned as the kernel of π∗ . L2 ) ∈ H 1 (L. the image L1 of ι∗ : T L → ι∗ T (T ∗ (Rn )) and L2 = ι∗ V Rn .ι = µ(L1 . A check of these deﬁnitions shows that the Maslov class of (L. known as the Liouville form. · · · . Critical points and critical values of πL are called respectively singular points and caustic points of L. we set πL = π ◦ ι.21. Z) by the Gauss map G : L → L(R2n ) deﬁned by G(p) = ι∗ Tp L ⊂ R2n . 3. 28 . A nice property of the Liouville 1form is that it can be used to parametrize the set of projectable lagrangian submanifolds. qn . Z) is called the Maslov class of (L. deﬁned by the formula αM ((x. ι) equals the pullback of the universal Maslov class µn ∈ H 1 (L(R2n ). the equations qj (x. then the symplectic vector bundle ι∗ T (T ∗ (Rn )) has two lagrangian subbundles. Note that if M = Rn . b))(v) = b(π∗ v). b) = xj (x) imply that n pj (x.22 As a particular case of Example 3. In local coordinates (x1 . xn ) on M and corresponding coordinates (q1 . b) = b ∂ ∂xj αM = j=1 n j=1 pj dqj . · · · . The class µL. Thus. (L. we use the notation ιϕ to denote a 1form ϕ on M when we want to think of it as a map from M to T ∗ M . ι). where π : T ∗ M → M is the natural projection. then ωM is just the symplectic structure ωn on T ∗ Rn R2n discussed previously.
29 . Taking exterior derivatives on both sides of the equation in Lemma 3. L is the image of an exact 1form on M if and only if the restriction of the Liouville form to L is itself exact. it satisﬁes π ◦ ιϕ = idM . Tx N ⊂ ker(p)}.Lemma 3. then Lemma 3. αM is often described as the “tautological” 1form on T ∗ M .26 If L. we get dϕ = dι∗ αM = ι∗ dαM = −ι∗ ωM . then the union of the conormal bundles to the leaves of F is a smooth submanifold of T ∗ M foliated by lagrangian submanifolds and is thus coisotropic (see Example 3. ιϕ ) deﬁnes a natural bijective correspondence between the the vector space of closed 1forms on M and the set of projectable lagrangian submanifolds of T ∗ M . This motivates the following deﬁnition. we will call S : M → R a phase function for a projectable lagrangian embedding (L. known as the conormal bundle to N . If F is a smooth foliation of M .25 suggests that the primitive of ι∗ αM is a sort of generalized phase function for (L. Because ιϕ is a section of T ∗ M . ϕ Proof. A general class of exact lagrangian submanifolds can be identiﬁed as follows. Deﬁnition 3. Generalizing our WKB terminology. π∗ (ιϕ∗ v) = ιϕ (p). This proves Proposition 3. ι) ⊂ T ∗ M provided that ι(L) = dS(M ). Then ι∗ αM = ϕ. while the Liouville form of T ∗ M vanishes on N ⊥ for any N .19). then S : M → R is a phase function for L if and only if d(S ◦ πL ◦ ι) = ι∗ αM . ι) which lives on the manifold L itself. ι) ⊂ T ∗ M is a projectable lagrangian embedding.25 If (L. ϕ ϕ ϕ From this equation we see that the image of ϕ is a lagrangian submanifold of T ∗ M precisely when the form ϕ is closed.24 The relation ϕ ↔ (M. it follows that for each v ∈ Tp M . The preceding remarks imply a simple link between phase functions and the Liouville form: Lemma 3. M are nmanifolds and ι : L → T ∗ M is an immersion such that ι∗ αM is exact. From this deﬁnition it follows easily that dim T ∗ M = 2 dim N ⊥ . v . ϕ 2 For this reason. Example 3. Associated to a smooth submanifold N ⊂ M is the submanifold N ⊥ = {(x.23 Let ϕ ∈ Ω1 (M ). ι∗ αM (p)(v) = αM (ιϕ (p))(ιϕ∗ v) = ιϕ (p). Thus. then ι is called an exact lagrangian immersion. We will return to this important viewpoint in the next chapter. By the deﬁnition of αM . although it is the image of an exact lagrangian immersion of R. If ι : L → T ∗ M is an exact lagrangian immersion.23. p) ∈ T ∗ M : x ∈ N.27 A simple application of Stokes’ theorem shows that an embedded circle in the phase plane cannot be exact.
29 If a symplectomorphism f : T ∗ M → T ∗ M preserves each ﬁber of the projection π : T ∗ M → M . Since αM vanishes on the zero section ZM ⊂ T ∗ M . ι). we can conclude from Example 3. then f = fβ for a closed 1form β on M . Example 3. Since f is symplectic. since dι∗ αM = ι∗ ωM = 0. p) = (x. the pullback of αM vanishes. p0 ) equals the identity. Theorem 3. the ﬁberderivative of f at the arbitrary point (x0 . it follows that ZM is lagrangian as well. R) induced by this form will play an important role in the quantization procedures of the next chapter and is known as the Liouville class of (L. p0 ) ∈ T ∗ M and let ψ be a closed 1form on M such that ψ(x0 ) = (x0 . and the subbundles T ZM and V M deﬁne a canonical lagrangian splitting of T (T ∗ M ) over ZM . Fix a point (x0 . the vertical bundle V M = ker π∗ is a lagrangian subbundle of T (T ∗ M ). 30 . Deﬁning β(x) = f (x. where ⊥ πN : Nβ → N is here the restriction of the natural projection π : T ∗ M → M . Thus. since the derivative Dh preserves the lagrangian splitting of T (T ∗ M ) along ZM and equals the identity on T ZM . Then ⊥ Nβ = {(x. Fiberpreserving symplectomorphisms On each ﬁber of the projection π : T ∗ M → M . Moreover.28 To generalize the picture described in Example 3. we have f = fβ .1 that Dh is the identity at all points of ZM . Proof. A 1form β on M deﬁnes a diﬀeomorphism fβ of T ∗ M by ﬁberwise aﬃne translation fβ (x.ι ∈ H 1 (L. we consider a smooth manifold M . It is easy to see that this map satisﬁes ∗ fβ αM = αM + π ∗ β.27. and thus the map −1 h = fµ ◦ f ◦ fψ is a symplectomorphism of T ∗ M which preserves ﬁbers and ﬁxes the zero section ZM ⊂ T ∗ M . Consequently. p + β(x)). the form ι∗ αM is always closed. p0 ).Although many lagrangian immersions ι : L → T ∗ M are not exact. so the ﬁbers are lagrangian submanifolds. p) ∈ T ∗ M : x ∈ N pTx N = β(x)} ∗ ⊥ is a lagrangian submanifold of T ∗ M whose Liouville class equals [πN β] ∈ H 1 (Nβ . R). together with a submanifold N ⊂ M and a closed 1form β on N . so fβ is a symplectomorphism of T ∗ M if and only if β is closed. 0). the form µ = f ◦ ψ is also closed. so f is a translation on each ﬁber. The deRham cohomology class λL.
Corollary 3. the HamiltonJacobi theorem implies furthermore that the ﬂow ft satisﬁes the hypotheses of Theorem 3.29.29. and so ft∗ π ∗ β = π ∗ β for all t. Proof. Proof. the time1 map f = f1 of the ﬂow of Xβ equals fβ . 2 Using Theorem 3. 2 31 . By Theorem 3. we can furthermore classify all ﬁberpreserving symplectomorphisms from T ∗ M to T ∗ N . By composing F with a ﬁbertranslation in T ∗ M we may assume that F maps the zero section of T ∗ M to that of T ∗ N .2 If β is a closed 1form on M . To this end. the assertion will follow provided that we can show that f ∗ αM = αM + π ∗ β. note that the deﬁnition of the Lie derivative shows that f satisﬁes 1 f αM = αM + 0 ∗ d ∗ (f αM ) dt = αM + dt t 1 ft∗ (LXβ αM ) dt. since V M ⊂ ker π ∗ β. Inserting these computations into the expression for f ∗ αM above. From the preceding theorem.31 Any ﬁberpreserving symplectomorphism F : T ∗ M → T ∗ N can be realized as the composition of a ﬁbertranslation in T ∗ M with the cotangent lift of a diﬀeomorphism N → M.30 For any closed 1form β on M . we obtain f ∗ αM = αM + π ∗ β. we conclude that F = f ∗ . Another application of Cartan’s formula. 0 By Cartan’s formula for the Lie derivative. then the ﬂow ft of the vector ﬁeld Xβ = −˜ M (π ∗ β) is ω −1 symplectic.29. we have LXβ αM = d(Xβ αM ) − Xβ ωM = π ∗ β. Corollary 3. The restriction of F −1 to the zero sections then induces a diﬀeomorphism f : N → M such that the composition F ◦ (f −1 )∗ is a ﬁberpreserving symplectomorphism of T ∗ N which ﬁxes the zero section. the latter equality following from the fact that Xβ ⊂ V M ⊂ ker αM and dαM = −ωM . combined with the assumption that β is closed shows that LXβ π ∗ β = 0.
then the map SM. 3. we associate to any symplectomorphism F : T ∗ M → T ∗ N the lagrangian embedding ιF : T ∗ M → T ∗ (M × N ) deﬁned as the composition of SM. Finally. 5 32 . if F is the cotangent lift of g.3 Mechanics on manifolds With the techniques of symplectic geometry at our disposal. In particular.N equals the submanifold Γ⊥∗ β ⊂ T ∗ (M × N ) deﬁned in Example 3.32 If M.N with the graph ΓF : T ∗ M → T ∗ M × T ∗ N . then the Schwartz transform SM. −ξ.N )∗ αM ×N = αM ⊕ −αN .N induces a diﬀeomorphism of zero sections ZM × ZN and an isomorphism of vertical bundles VM ⊕VN V (M × N ). N are smooth manifolds. ξ).2). (y. setting the stage for the quantization problem in the next chapter.31.The Schwartz transform If M. N are smooth manifolds. Example 3. ZM ×N Using the Schwartz transform. ιF ) equals the conormal bundle of Γ. then the image of the composition of the lagrangian embedding (T ∗ M. but fundamental property of this mapping can be described as follows. largely parallelling the earlier material on the 1dimensional harmonic oscillator given in the introduction. We then turn to the semiclassical approximation and its geometric counterpart in this new context. The name comes from the relation of this construction to the Schwartz kernels of operators (see Section 6. Proposition 3. A computation shows that if Γ ⊂ M × N is the graph of g and p : Γ → M is the natural projection. η) is a symplectomorphism which we will call the Schwartz transform. In particular.28.33 By Corollary 3. SM.N satisﬁes (SM. Thus in the special case of cotangent bundles. ιF ) with the Schwartz transform SM. η)) → (x. Our description begins with a comparison of the classical and quantum viewpoints. we are ready to extend our discussion of mechanics to more general conﬁguration spaces. we note that multiplying the cotangent vectors in T ∗ M by −1 deﬁnes a symplectomorphism T ∗ M → T ∗ M which can be combined with the Schwartz transform SM.N : T ∗ M × T ∗ N → T ∗ (M × N ) deﬁned in local coordinates by ((x.N to arrive at the usual symplectomorphism T ∗ M × T ∗ N T ∗ (M × N ). y.5 An elementary. dualizing and taking products leads to nothing new. p then the image of (T ∗ M. a ﬁberpreserving symplectomorphism F : T ∗ M → T ∗ N equals the composition of ﬁberwise translation by a closed 1form β on M with the cotangent lift of a diﬀeomorphism g : N → M .
j where g ij is the inverse matrix to gij . we ﬁrst deﬁne the operator on the function space C ∞ (M ) by ˆ H=− 2 2m ∆ + mV . Theorem 3. Theorem 3. 33 . As before. realvalued potential V : M → R induces the hamiltonian function H(q. Integral curves of the hamiltonian ﬂow of H then project to classical trajectories of a particle on M subject to the potential V . p) = 1 2 g ij (q) pi pj . ij ˙ For details. i. A smooth. ˆ where ∆ denotes the LaplaceBeltrami operator. then integral curves of the cogeodesic ﬂow project via π to geodesics on M . A riemannian metric g = (gij ) on M induces an inner product on the ﬁbers of the cotangent bundle T ∗ M . p) = kM (q. The quantum mechanical picture For the time being. p) + V (q) on T ∗ M .v ∂g uv pu pv ∂qi to derive the geodesic equation qk + ¨ i. That is. p) is given by kM (q. A proof of this theorem can be given in local coordinates by using Hamilton’s equations qi = ˙ ∂H = ∂pi g ij pj j pi = − ˙ ∂H 1 =− ∂qi 2 u.j Γk qi q˙j = 0. H induces a (densely deﬁned) ˆ on the intrinsic Hilbert space HM of M by the equation operator H ˆ ˆ H(adx1/2 ) = (Ha)dx1/2 . Regular level sets of kM are sphere bundles over M .34 states that a free particle on a manifold must move along a geodesic. The hamiltonian ﬂow associated to kM is called the cogeodesic ﬂow due to its relation with the riemannian structure of M described in the following theorem.The classical picture The hamiltonian description of classical motions in a conﬁguration space M begins with the classical phase space T ∗ M . In physical terms.34 If M is a riemannian manifold. we will assume that the Schr¨dinger operator on a riemannian manifold o M with potential function V is deﬁned in analogy with the ﬂat case of Rn with its standard metric. and a “kinetic energy” function which in local coordinates (q. see [36].
a “generalized” phase function φ : L → R satisfying dφ = ι∗ αM . then (L. Now if H : T ∗ M → R is any smooth function. the embedding ι and hamiltonian vector ﬁeld XH of H induce a nonsingular vector ﬁeld XH. If these conditions are satisﬁed. The semiclassical approximation The basic WKB technique for constructing semiclassical solutions to the Schr¨dinger equao tion on M proceeds as in Section 2. and Lemma 3.25 implies that for −1 any primitive φ : L → R of ι∗ αM . this means that πL = π ◦ ι is a diﬀeomorphism. If a is a halfdensity on L.ι on L (see Example 3. and a halfdensity a on L.ι a = 0. ι). ι. This interpretation of the WKB approximation leads us to consider a semiclassical state as a quadruple (L. and the halfdensity a satisﬁes the homogeneous transport equation. exact lagrangian embedding ι : L → T ∗ M . then the requirement −1 that (πL )∗ a satisfy the homogeneous transport equation on M becomes LXH. as above. By deﬁnition.2. and the timeindependent Schr¨dinger equation on M assumes the familiar form o ˆ (H − E)ϕ = 0. where π : T ∗ M → M is the natural projection.where dx is the natural density associated to the metric on M . a halfdensity of the form eiS/h a is a ˆ secondorder approximate solution of the eigenvector problem (H − E) ϕ = 0 provided that the phase function S : M → R satisﬁes the HamiltonJacobi equation H ◦ ιdS = E.e. a) comprised of a projectable. The advantage of this viewpoint is that both the classical state space T ∗ M and the quantum state space HM are objects intrinsically associated to the underlying diﬀerential manifold M . We can formulate this construction abstractly by considering ﬁrst a projectable. The dynamics on both objects are determined by the choice of metric on M . φ. Speciﬁcally. Our correspondence table now assumes the form 34 . ι) satisﬁes the HamiltonJacobi equation provided that E is a regular value of H and H ◦ ι = E. In this case. exact lagrangian embedding ι : L → T ∗ M . the composition S = φ ◦ πL is a phase function for (L.13). which assumes the coordinatefree form a∆S + 2L S a = 0. then the halfdensity eiφ/ a on L can be quantized (i. −1 pulledback) to yield a secondorder approximate solution (πL )∗ eiφ/ a to the Schr¨dinger o equation on M .
Object basic space state timeevolution Classical version T ∗M (L. a) as a further generalization of the concept of semiclassical state in T ∗ M . ι. in which we drop the conditions of projectability and exactness. φ. it is tempting to regard (L. Our goal in the next chapter will be to determine when and how such “geometric” semiclassical states can be used to construct “analytical” semiclassical approximate solutions to the timeindependent Schr¨dinger equation. a) consisting of an arbitrary lagrangian immersion ι : L → T ∗ M and a halfdensity a on L. a) as above Hamilton’s equations Quantum version HM on M Schr¨dinger equation o ˆ operator H on HM generator of evolution function H on T ∗ M stationary state ˆ state (L. ι. ι. o 35 .ι a = 0 Since the HamiltonJacobi and transport equations above make sense for any triple (L. a) such that eigenvector of H H ◦ ι = E and LXH. φ. ι.
ι.” 4. we will often speak loosely of “quantizing lagrangian submanifolds” or “quantizable lagrangian submanifolds. Given a halfdensity a on L which is invariant under the ﬂow induced by H. aj ) in the sense above: −1 Ij = (πLj )∗ eiφj / aj . then the Liouville class will be 2π times an integral class for all only if it is zero. we are forced to confront once more the nature of . and so we can choose a good cover {Lj } of L (see Appendix C) e and functions φj : Lj → R such that dφj = ι∗ αM Lj . At this point. it is locally exact by the Poincar´ lemma. a) can be quantized is answered largely in terms of the geometry of the lagrangian immersion (L. (It is. ιLj . To quantize (L. The choice of φ is unique only up to an additive constant. an approximate solution to Schr¨dinger’s equation on M . which leads to an ambiguity in the overall phase of I (L. i. this is precisely the condition that the Liouville class λL. the question of whether or not a given semiclassical state (L. in a more or less systematic way.1 Prequantization The simplest quantization procedure consists of pullingback halfdensities from projectable lagrangian embeddings in T ∗ M . we set aj = aLj and deﬁne a halfdensity on πL (Lj ) by quantizing (Lj . ι.) As we shall see. To generalize this procedure. As seen in Chapter 3. If it is a formal variable. According to the discussion in Appendix C. we must piece together the Ij to form a welldeﬁned global halfdensity I (L. however. we must have def φj − φk ∈ Z = 2π · Z on each Lj ∩ Lk .ι be integral. suppose now that ι : L → T ∗ M is a projectable. o This process will be referred to as quantization. ι.e. a) = (πL )∗ eiφ/ a on M . Since the 1form ι∗ αM on L is closed. For this reason. The starting point will be a lagrangian immersion ι : L → T ∗ M whose image in contained in a regular level set of the classical hamiltonian H associated to some metric and potential on the conﬁguration space M .4 Quantization in Cotangent Bundles This chapter deals with the problem of constructing semiclassical approximate solutions to the timeindependent Schr¨dinger equation from the data contained in a “geometric” semio classical state. our goal is to use this data in order to construct. ι. φj . the notion of integrality is meaningless. This is possible for arbitrary a provided that the functions φj can be chosen so that the oscillatory coeﬃcients eiφj / agree where their domains overlap. ι. (This ambiguity was overcome in Chapter 3 by including the choice of φ in the deﬁnition of a semiclassical state). a triple (L. a). a) on M . only one of many operations which go by this name. but not necessarily exact lagrangian embedding. a) for which ι : L → T ∗ M is a projectable exact lagrangian embedding is quantized by choosing a primitive φ of ι∗ αM and forming the halfdensity def −1 I (L. that is. ι. a). only for an exact lagrangian 36 . Given a halfdensity a on L. ι). If denotes a number ranging over an interval of the real numbers. of course.
then by linearity we should have I (L1 ∪ L2 . where 0 is the largest such number. ι. a1 ) + I (L2 . it is desirable that quantization be linear with respect to halfdensities: I (L. the set of all admissible forms a sequence converging to zero. The situation changes as soon as we consider the closed 1form τ = a dθ1 + b dθ2 on the torus S 1 × S 1 . a1 ). and it is reasonable to adopt as a general rule. ι2 . Deﬁnition 4. Instead. ι1 ∪ ι2 . a2 )) = I (L1 ∪ L2 . 37 . Certainly this condition holds for the procedure we have been using in the projectable case. ι1 ∪ ι2 . ι1 . a2 )) = I (L1 . ι1 ∪ ι2 . meaning that a/b is a rational number if b = 0. the class of projectable quantizable lagrangian submanifolds of T ∗ M is rather limited whenever dim(H 1 (M . but with diﬀering sets of admissible .ι is integral for some ∈ R+ . Since we are speciﬁcally trying to go beyond the exact case. In general. ι) ⊂ T ∗ M . this interpretation is not acceptable either.submanifold. ι) ⊂ T ∗ M is quantizable if its Liouville class λL. If L is exact. for p ∈ R. (a1 . a2 ). (A period of a closed 1form β on M is any number obtained by integrating β around some closed loop in M ).2 Let M = S 1 and consider the closed 1form β = p dθ on S 1 . (0. (a1 . ι. ι1 . ι). and so the admissible values of are the numbers {p/k : k ∈ Z+ }. however.1 A projectable lagrangian submanifold (L. then a/ and b/ are both integers. ι2 . 0)) + I (L1 ∪ L2 . ι. Geometrically this means that all horizontal circles in the cylinder T ∗ S 1 are quantizable in the sense deﬁned above. and k runs over the positive integers. ι. the condition that closed 1form β on a manifold M be quantizable is equivalent to the requirement that the ratio of any two nonzero periods of β be rational. we will focus on the set of regular points of πL in order to pass from halfdensities on L to halfdensities deﬁned near noncaustic points of L. Example 4. R)) > 1. we will make the following compromise. all > 0 are admissible. The cohomology class represented by this form is integral provided that p S1 dθ ∈ Z . The values of for which this condition holds will be called admissible for (L. One consequence is I (L. since πL cannot be used to pushforward halfdensities from L to M . If a semiclassical state is represented by the union of a disjoint pair (L1 . Thus. sa1 + a2 ) = s · I (L. If L is quantizable but not exact. 0) = 0. The simple quantization technique described above does not generalize immediately to nonprojectable immersed lagrangian submanifolds (L. consisting of the numbers 0 /k. a2 ) of lagrangian submanifolds carrying halfdensities. a2 ). (L2 . a1 ) + I (L. For the time being. If τ is integral for some ∈ R+ . Regardless of how this is to be carried out.
a) is given for q > 0 by I (L. ∂q ∂x the transformation rule for halfdensities implies −1 (πL+ )∗ a = 2−1/2 q −1/4 B(q 1/2 ) dq1/2 −1 (πL− )∗ a = 2−1/2 q −1/4 B(−q 1/2 ) dq1/2 . a halfdensity a on L is invariant under the ﬂow of XH. a) should look something like −1 (πLj )∗ eiφj / a j on U . p) = (p2 − q). To quantize halfdensities on L consistently. ι. then there is a contractible −1 neighborhood U ⊂ M of p for which πL (U ) consists of ﬁnitely many disjoint open subsets Lj ⊂ L such that each (Lj .3 For L = R. From the expression above. 2 and it is easy to check that the induced vector ﬁeld XH. If p ∈ πL (L) is noncaustic and πL is proper. ι. ι. L− the right and left projectable components of L. the lagrangian embedding ι : L → T ∗ R given by ι(x) = (x2 . a) = e2iq 3/2 /3 + e−2iq 38 3/2 /3 2−1/2 q 1/4 B dq1/2 . we obtain I (L. If a = B(x) dx1/2 is any halfdensity on L. Thus. ιLj ) is a projectable lagrangian submanifold of T ∗ U . A phase function φ : L → R for (L. As before. then we may use any function φ : L → R satisfying dφ = ι∗ αM to ﬁx phases. the quantization of (L. Choosing a generalized phase function φj : Lj → R for each Lj . ι. and so the meaning of the preceding sum is ambiguous. respectively (these correspond to the upper and lower components of the parabola ι(L)).ι = (1/2) ∂/∂x.ι if and only if a = B dx1/2 for some B ∈ R. the prequantization of (L. we note that by the preceding remarks. Thus. ι) is given by φ(x) = 2x3 /3. we must therefore decide how to specify the relative phases of the oscillatory coeﬃcients eiφj / . x) has a singular point at x = 0. then since −1 (πL )∗ ∂ ∂ = ±2−1 q −1/2 . as in the following example.ι on L equals XH. Example 4. The parabola ι(L) lies in the regular level set H −1 (0) of the hamiltonian for a constant force ﬁeld 1 H(q. the requirement that φj be a generalized phase function for Lj determines each φj only up to an additive constant. ι) ⊂ T ∗ M and halfdensity a on L. and we denote by L+ . a)(q) = e2iq 3/2 /3 B(q 1/2 ) + e−2iq 3/2 /3 B(−q 1/2 ) 2−1/2 q −1/4 dq1/2 .Now consider an arbitrary immersed lagrangian submanifold (L. If L is exact.
Deﬁnition 4. In particular. 2 According to Deﬁnition 4. energy levels of the 1dimensional harmonic oscillator corresponding to level sets for which a particular value of is admissible are given by E = n . For this purpose. If is admissible for some prequantizable lagrangian immersion (L. Consequently.ι is integral for some ∈ R+ . φ. the actual quantum energy levels are E = (n + 1/2) . 2 ∂x2 2 2 Unfortunately. 39 . we can treat certain nonexact cases in a similar way. ι. ι) ⊂ T ∗ M is said to be prequantizable if its Liouville class λL. we have no way as yet to check that letting φ be continuous at 0 as a function on L is the right way to assure that we have a good approximation in the immediate vicinity of q = 0. φ. ι).4 An immersed lagrangian submanfold (L. ι. we can now quantize (L. a) by summing the pullbacks of eiφ/ a to M . a) on U is deﬁned as −1 (πVj )∗ eiφ/ a. then there exists a good cover {Vj } of the manifold L and functions φj : Vj → R such that dφj = ι∗ αM Vj and φj − φk ∈ Z on each Vj ∩ Vk . j Example 4. generalizing Deﬁnition 4.1. The additional 1/2 can be explained geometrically in terms of the nonprojectability of the classical energy level curves. as we shall soon see. the value of I (L. If a is a halfdensity on L. H −1 (E) Thus. so we have more work ahead of us. The values of for which this condition holds will again be called admissible for (L. p) = (q 2 + p2 ).4.as a semiclassical approximate solution to the Schr¨dinger equation o − ∂2ψ x − ψ = Eψ. the φj describe a single function φ : L → T = R/Z which satisﬁes dφ = ι∗ αM and which deﬁnes a global oscillatory function eiφ/ on L. this solution blows up at q = 0 (and is not deﬁned for q < 0). In fact.5 Consider the 1dimensional harmonic oscillator with hamiltonian 1 H(q. ι). In the previous notation. As one can read in any textbook on quantum mechanics. a number that ∈ R+ is admissible for the level set H −1 (E) provided α1 = 2πE ∈ Z . we make the following provisional deﬁnition. we will see later that this is the wrong choice! Since exactness is only used to insure that the function eiφ/ is welldeﬁned on L.
dϕ is a linear symplectic form on ker(ϕ). If we associate to QM.8 A contact form on a (2n + 1)dimensional manifold Q is a 1form ϕ such that ϕ ∧ (dϕ)n vanishes nowhere on Q. φ) consisting of a lagrangian immersion ι : L → T ∗ M . the odddimensional counterpart of symplectic geometry. 40 . the prequantum line bundle EM. ωM ) consists of the trivial principal bundle QM. This says that the distribution is integrable (in the sense of Frobenius) iﬀ dϕ is zero on ker(ϕ). and the “largest” integral submanifolds of ker(ϕ) are ndimensional (i. Digressing brieﬂy from quantization. The holonomy of this connection is represented by the modZ reduction of the Liouville class λL. If ι : L → T ∗ M is any lagrangian immersion. Consequently. If ξ. ι. To interpret the condition on ϕ we note ﬁrst that the kernel of any nowhere vanishing 1form ϕ deﬁnes a 2ndimensional distribution in Q. This remark proves the following geometric characterization of prequantizability. by means of the representation x → e−ix/ of T in U (1). and a parallel lift φ of L to the T bundle ι∗ QM. the basic geometric object representing a classical state is therefore a quadruple (L. Theorem 4. η]) = −ϕ([ξ. The condition ϕ ∧ (dϕ)n = 0. corresponding to the (inverse of the) oscillatory function eiφ/ on L which appeared in the preceding section. From the prequantum standpoint.6 For ﬁxed ∈ R+ . η are (local) vector ﬁelds lying in this distribution. ι) ⊂ T ∗ M is prequantizable if and only if there exists a nonzero parallel section over L of the line bundle ι∗ EM. η]). ϕ) is an ndimensional integral submanifold for ϕ. then the curvature of the induced connection on ι∗ QM. σ denotes the multiplevalued linear variable in T and π : QM. Deﬁnition 4. η) = ξ · ϕ(η) − η · ϕ(ξ) − ϕ([ξ.7 An immersed lagrangian submanifold (L. we have dϕ(ξ. we assemble here a few facts about contact manifolds. It is customary to make the following deﬁnition. = T ∗ M × T together with the connection 1form ϕ = −π ∗ αM + dσ.ι . for some > 0. → T ∗ M is the bundle projection. Prequantum T bundles and parallel lifts of lagrangian submanfolds constitute our ﬁrst examples of the fundamental objects of contact geometry. then φ induces a parallel section of ι∗ EM. means that the kernel of dϕ is 1dimensional and everywhere transverse to ker(ϕ). on the other hand. . A manifold endowed with a contact form is called a strict contact manifold. coincides with ι∗ ωM and therefore vanishes.9 A legendrian submanifold of a 2n+1dimensional strict contact manifold (Q. a. Here. Deﬁnition 4. ker(ϕ) is “maximally nonintegrable”). a halfdensity a on L.Prequantum bundles and contact manifolds Prequantizability can be described geometrically in terms of principal T bundles with connection over T ∗ M .e. Deﬁnition 4. the prequantum T bundle associated to a cotangent bundle (T ∗ M.
a possible generalization of the quantization procedure described above might therefore begin by reinterpreting geometric semiclassical states as triples (R. or ﬁber. so that dτ = ι∗ (−q dp) for some phase function τ on L. and then quantizing to obtain a function on pspace.10 A contact structure on a manifold M is a codimension one subbundle E ⊂ T M which is locally deﬁned by contact forms. If a is a halfdensity on L. Deﬁnition 4. ι) is said to be pprojectable. we obtain −1 (πp )∗ eiτ / dx1/2 = e−iq0 p/ q0 dp1/2 . E is the kernel of a globally deﬁned contact form. Following an idea of Maslov. . ι) ⊂ T ∗ R is pprojectable. 41 . This factor will be incorporated into our quantization procedure using a procedure due to Maslov. the image of a parallel lift of a lagrangian immersion ι : L → T ∗ M to QM. Since the wave function corresponding to a constant halfdensity a on L should correspond to a probability distribution describing the position of a particle at q0 with completely indeterminate momentum. and a manifold endowed with such a structure is called simply a contact manifold. then f · ϕ is again a contact form which is said to be equivalent to ϕ. A simple example of an embedded lagrangian submanifold of T ∗ R which does not project diﬀeomorphically onto the qaxis is a vertical line. over a cotangent bundle T ∗ M . we will illustrate this idea in the case of lagrangian submanifolds of the phase plane. If πp is a diﬀeomorphism. in which case there exists an “alternate” generalized phase function τ : L → R satisfying dτ = ι∗ (−q dp). The result is exactly the asymptotic Fourier transform (see Appendix B) of the delta function which we guessed above! In its simplest form. Such a contact structure is called coorientable. then (L. we analyze this situation by pretending that p is position and q momentum. When the quotient T M/E is trivial. Given a lagrangian immersion ι : L → T ∗ R R2 . it should be a delta function supported at q0 . we deﬁne a function B on pspace by the equation −1 B dp1/2 = (πp )∗ eiτ / a. is an immersed legendrian submanifold of QM. a) consisting of a legendrian immersion (R.2 The Maslov correction It turns out that the naive quantization procedure of the preceding section is incorrect. 4. obtained by thinking of R2 as the cotangent bundle of pspace. Although we will not pursue the idea in these notes. since they have the same legendrian submanifolds. Maslov’s technique is to suppose that (L. we denote by πp the composition of ι with the projection of R2 onto the paxis. This leads to the following deﬁnition. Our basic example of a strict contact manifold is furnished by a prequantum T bundle QM. ) in a contact manifold together with a halfdensity a on R. Using the phase function τ (x) = −q0 x on L. To begin.If ϕ is a contact form and f a nowherevanishing function on Q. . x) for x ∈ R. of the form ι(x) = (q0 . since it ignores a certain structure which arises from the relation of L to the ﬁbers of the π projection T ∗ M → M .
Quantization by pullback therefore gives I (L. we begin with the example of linear lagrangian subspaces. φ.6. ι) ⊂ R2 with phase functions φ and τ corresponding to p dq and −q dp. Next. the halfdensity obtained from (L. To establish a similar correspondence in somewhat greater generality. On the other hand. For simplicity.and pspace by pullback: −1 S = φ ◦ πL −1 T = τ ◦ πp . Example 4. i.The Maslov quantization of (L. ι. τ. Vol. τ.and pdirections. 6 See.1] . a) = F −1 (B) dq1/2 on qspace. 42 . If a is a constant halfdensity on L.11 For real k = 0. so that S and T satisfy the Legendre transform relation (see [2]) S(q) = −p(q) T (p(q)) + T (p(q)). Thm. a) = eikq 2 /2 A dq1/2 . we must compare the results in the case of lagrangian submanifolds L ⊂ R2 which are biprojectable. [32. Thus. ι. τ. let S(q) and T (p) be the functions deﬁned on q. we have (πp )∗ τ (p) = −p2 /2k. respectively. For simplicity. a). we will assume that the additive constants in φ and τ are chosen so that φ = τ + ι∗ (qp). kx). consider the lagrangian embedding ι : R → T ∗ R given by ι(x) = (x. ι. where F denotes the asymptotic Fourier transform. Generalized phase functions on (L.1. ι. a) = e−iπ·sgn(k)/4 I (L. then the transformation rule for halfdensities implies −1 (πL )∗ a = A dq1/2 −1 (πp )∗ a = k−1/2 A dp1/2 def for a real constant A determined by a. a) by Maslov quantization diﬀers from the simple pullback by a constant phase shift. φ. a) is then given by the halfdensity J (L.e.7. ι. for example. ι. ι) for the forms p dq and −q dp are given by φ(x) = kx2 /2 and τ (x) = −kx2 /2. projectable in both the q. consider an arbitrary biprojectable lagrangian embedding (L. and a computation6 shows that −1 F −1 ((πp )∗ eiτ / )(q) = k1/2 e−iπ·sgn(k)/4 eikq 2 /2 . respectively. and so Maslov’s technique yields J (L. To relate this procedure to our earlier quantization by pullback.
1 and the Maslov quantization of a pprojectable lagrangian embedding (L. a) = F −1 e−ip 3 /3 B(p) dq1/2 . Hence. when (q.e. a) = eiS/ A dq1/2 . For biprojectable (L. it follows easily that T (p(q)) = −(S (q))−1 .12 A phase function associated to −q dp for the lagrangian embedding ι(x) = (x2 . For each q > 0. a) = e−iπ·sgn(k)/4 I (L.where p(q) = S (q). The Maslov quantization of a halfdensity a = B(x) dx1/2 on L is thus J (L. ι. To this end. ι. ι. The essential diﬀerence between the naive prequantization of Section 4. as illustrated by the following example. . ι. ι. a) = (2π )−1/2 R ei(pq+T (p))/ B(p) dp dq1/2 with that obtained by pullback: I (L. i. a) = e−iπ/4 e2iq 3/2 /3 B(−q 1/2 ) + eiπ/4 e−2iq 43 3/2 /3 B(q 1/2 ) 2−1/2 q −1/4 dq1/2 + O( ). φ. A halfdensity a on L determines functions A(q) and B(p) such that −1 (πL )∗ a = A dq1/2 −1 (πp )∗ a = B dp1/2 and A(q) = S (q)1/2 B(p(q)). we set k(q) = T (p(q)) and apply the principle of stationary phase (see Appendix B). ι). we must now compare the Maslov halfdensity J (L. ι. ei(pq+T (p))/ B(p) dp = (2π )1/2 e−iπ·sgn(k)/4 eiS(q)/ k(q)−1/2 B(p(q)) + O( R 3/2 ). Example 4. ι. τ. and an application of the principle of stationary phase therefore yields two terms corresponding to the upper and lower halves of L. p) ∈ ι(L).e. critical points of the function R(p) = pq − p3 /3 occur precisely when q = p2 . we have J (L. The critical point of the exponent pq + T (p) occurs where q = −T (p). Thus J (L. a) + O( ). we therefore conclude that Maslov’s technique coincides with quantization by pullback up to a constant phase factor and terms of order . From this relation. where p = S (q) = p(q). x) of L = R into R2 is given by τ (x) = −x3 /3. For each q. τ. Speciﬁcally. i. ι) which is not qprojectable lies in the relative phase constants of the summands of I (L. φ. τ. −1 since (πp )∗ a = B(p) dp1/2 . a). τ.
τ. ι) by summing these changes while traversing L in a prescribed direction. ι) is a closed. suppose that (L.or pprojectable and no intersection Lj ∩ Lk contains a critical point of πL . a). and we can assign an index to (L. The reader is invited to check that if L is a circle with a ﬁxed orientation ν ∈ H1 (L. The result is twice an integer known as the Maslov index mL. p) is a singular point of L. if T were concave. In fact. both of which are positive according our rule when the circle is traversed counterclockwise. the full expression for J (L. The basic idea is as follows. a) is. sgn(T ) changes by ±2 in the vicinity of a critical point of T . This smoothness at caustics is a clear advantage of Maslov quantization. The relative phase factor in the preceding example can be attributed to the fact that the function T (p) = −p3 /3 has an inﬂection point at p = 0. Next.ι of (L. ι). immersed curves in the phase plane. but not if both are.ι . like I (L. downward motion to the right of the ﬁber is positive. a factor of eiπ/2 arises in passing from the upper to the lower half of the parabola. In other words. then T has only nondegenerate critical points. Given an immersed lagrangian submanifold (L. a ﬁgureeight has Maslov index zero. However.Compare this with the result of Example 4. while the term of order 0 of J (L. τ. a). Under this assumption. ι) ⊂ R2 . The Maslov index of the circle therefore equals 2. we ﬁx a partition of unity {hj } subordinate 44 . Z). at least if a has compact support. (Compare [3]). then mL. On the other hand. φ.ι is the Maslov class of (L. ι.ι equals µL.3. the same is true of any closed embedded curve traversed counterclockwise.12 that the integer 1 should be assigned to a critical point of πL which is traversed in the −p direction to the right of the ﬁber. The preceding observation leads us to assign an index to closed. the situation would be reversed. Our goal is now to modify the prequantization procedure for arbitrary lagrangian submanifolds of the phase plane by incorporating the Maslov index. ι. τ. Since the index only involves the sign of T . ι. ι. immersed curve in R2 which is nondegenerate in the sense that if T is a pdependent phase function for a subset of L. A pdependent phase function T for L ⊂ R2 will have inﬂection points at precisely those p for which (T (p). a) essentially diﬀerent from the prequantization of (L. ν . it follows from Example 4. since T is convex. a) as an integral is perfectly smooth there. With these remarks in mind. A circle in the phase plane has a right and a left singular point. Example 4.22. ι) deﬁned in Example 3. More precisely. ι. while the sign changes if either the direction of motion or the side of the ﬁber is reversed. The extra phase factors of e iπ/4 make J (L. Moreover. singular at the caustic point q = 0. we ﬁrst choose a good cover {Lj } of L such that the image of each Lj under ι is either q. the sign of T at nearby points depends only on L and not on the choice of T . where µL.13 The computation of the Maslov index can be interpreted geometrically if we ﬁrst observe that the nondegeneracy condition requires L to remain on the same side of a ﬁber π −1 (q) near a singular point.
a) to be welldeﬁned. or. a) as the sum I (L. E = (n + 1/2) . If Lj is quantized by pullback. 45 . Evidently.14 Returning to the harmonic oscillator of Example 4. Allowable energy levels in this case therefore correspond to the BohrSommerfeld condition. a) = j Ij . (More precisely.to {Lj }. that the Maslov index mL. ι.12. For I (L. otherwise sj equals sgn(T ) for a suitable pdependent phase function). we quantize each (Lj . ι) as follows. ι. L This is the simplest version of the Maslov quantization condition. Here. Example 4.ι of (L. so that our requirement becomes ajk ∈ Z . we must choose the functions ˜ ˜ φj so that Ij = Ik on each intersection Lj ∩ Lk regardless of the particular halfdensity a. in other words. if Maslov’s technique is applied to Lj . it follows from an application of −1 the principle of stationary phase as in Example 4. and to make this deﬁnition independent of the choice of cover {Lj } and partition of unity. On Lj ∩ πL (U ).5. then this condition implies that the sum of any of the ajk lies in Z . this statement is obvious. This condition can be precisely formulated in terms of the Maslov index and Liouville class of (L. In other words. which actually gives the precise energy levels for the quantum harmonic oscillator. In order to specify the relative phases of the Ij . φ is a realvalued function on Lj satisfying dφj = ι∗ α1 . ι) satisﬁes π mL. we can deﬁne ajk as the (constant) value of (φj − φk ) − π (sj − sk )/4 on Lj ∩ Lk . On an open interval U of noncaustic points. we would then like to deﬁne the quantization of (L. sj is zero if Lj is qprojectable and is quantized by pullback. these halfdensities are of the form ˜ Ij = e−iπsj /4 eiφj / a. each halfdensity Ij is the sum of the pull−1 back of halfdensities on each component of Lj ∩ πL (U ). this condition can be fulﬁlled on any arc of a curve in the phase plane. a · hj ) to obtain a halfdensity Ij on R either by pullback or by Maslov’s technique. we require ei(φj −φk )/ e−iπ(sj −sk )/4 = 1 at each point of Lj ∩ Lk . a) coincide up to order with the usual quantization by pullback for any halfdensity a supported in a projectable subset of L. ι. ι. As before. Since φj − φk is constant on Lj ∩ Lk . we see that the level set H −1 (E) satisﬁes the Maslov condition provided that for some integer n. ι. we will require that I (L. while sj are integers depending −1 only on the component of Lj ∩ πL (U ) in question. If L is a circle.ι + 2 ι∗ α1 ∈ Z . To quantize a halfdensity a on L.
For ∗ each p ∈ U . v) ∈ U ×Rm and attempt to deﬁne dϕp ∈ Tp N by dϕp . As we saw in Chapter 3. the idea is the following. the expression for dϕp produces a welldeﬁned element ˜ ˜ ∗ of Tp N .A general quantization scheme Motivated by the simple results above. a projectable lagrangian submanifold of T ∗ N can be locally parametrized by the diﬀerential of a function f on U ⊂ N . we can begin with a function ϕ : U ×Rm → R together ∗ with a point p = (p.3 Phase functions and lagrangian submanifolds In this section. ι) ⊂ T ∗ M deﬁned by such phase functions will ﬁrst enable us to quantize a given halfdensity locally on L by means of a slightly more general version of the (inverse) asymptotic Fourier transform. which relies in this context on the concept of generalized phase functions. 4. it is necessary and suﬃcient that L satisfy a general version of the Maslov quantization condition. the restriction of the projection U × Rm → U to Σϕ is noninjective. The result will be a collection of halfdensities on M . which we formulate in the next section using the Maslov class. we generalize the concept of phase functions to nonprojectable lagrangian submanifolds of cotangent bundles. γ(0) = (ϕ◦˜ ) (0) ˜ γ ˜ ˜ ˙ m for any lift γ of γ to the product U × R such that γ (0) = p. Roughly speaking.θ)/ a(q. and the assignment p → dϕp deﬁnes a lagrangian ˜ ˜ ∗ immersion of Σϕ into T N . In order for these halfdensities to piece together appropriately. In general. As we shall see. 46 . In general. the meaning of dfp as an element of Tp N is that for any smooth curve γ in N satisfying γ(0) = p.q +T (p))/ a(p) dp dq1/2 Rn for the solution of Schr¨dinger’s equation by the more general form o (2π )−m/2 Rm eiφ(q. we have dfp . this generalization amounts to replacing Maslov’s ansatz (2π )−n/2 ei( p. Local parametrizations of an immersed lagrangian submanifold (L. If. γ(0) = (f ◦γ) (0). however. To parametrize more general lagrangian ˙ ∗ submanifolds of T N in a similar way. this fails. the ˜ ˜ ﬁberderivative ∂ϕ/∂θ vanishes at p. and thus the image of Σϕ is a nonprojectable lagrangian submanifold. our aim in the next sections will be to develop a systematic method for quantizing lagrangian submanifolds of cotangent bundles. since ˜ ˜ ˜ the value of the directional derivative (ϕ ◦ γ ) (0) depends on the lift γ . viewed as a mapping df : U → T ∗ N . The basis of this method will again be Maslov’s technique. θ)dθdq1/2 . From the point of view of the WKB method. the assumption that the map ∂ϕ/∂θ : U × Rm → Rm is transverse to 0 implies that the ﬁber critical set Σϕ = p ∈ U × Rm : ˜ ∂ϕ =0 ∂θ def is a smooth submanifold of U × Rm .
ι∗ d φ ∗ The ﬁberhessian Hφ of φ at p ∈ Σφ is deﬁned as the composition dθ φ ◦ ι : Ep → Ep . The nondegeneracy assumption on φ is equivalent to the requirement that the lagrangian submanifold (dφ)(B) of T ∗ B be transverse to E ⊥ . unlike the previous Fourier transform picture.6(2). (We will denote the zero section of a vector bundle F by ZF ). Dualizing the inclusion E = ker(pM ∗ ) → T B gives rise to an exact sequence of vector bundles over B 0 ← E ∗ ← T ∗B ← E ⊥ ← 0 where E ⊥ ⊂ T ∗ B denotes the annihilator of E.where θ is an auxiliary variable in Rm which may have nothing to do with the variable p dual to q. the characteristic distribution C ⊥ of E ⊥ is tangent to the ﬁbers of the mapping E ⊥ → T ∗ M . 2 Some readers may ﬁnd it instructive to follow the ensuing discussion by writing everything in local coordinates. this implies that the section (dφ)(B)Σφ = (dφ)(B) ∩ E ⊥ is nowhere tangent to the distribution C ⊥ and therefore immerses into T ∗ M . The ﬁberderivative of a function φ : B → R is the composition dθ φ = ι∗ ◦ dφ. which induces for nondegenerate φ an exact sequence of vector bundles over Σφ θ 0 → T Σφ → TΣφ B → E ∗ Σφ → 0.19 that E ⊥ is a coisotropic submanifold of T ∗ B. By Lemma 3. To begin. 7 47 . On the ﬁbercritical set Σφ . which requires linear structures on p. giving rise to a natural M p projection E ⊥ → T ∗ M . At points of Σφ . Proof. the diﬀerential dφ deﬁnes a section of E ⊥ whose composition with p we denote by λφ : Σφ → T ∗ M . denoted dθ φ.and qspace. then the map λφ : Σφ → T ∗ M is an exact lagrangian immersion. we ﬁx some notation and terminology. To complete the proof. B be smooth manifolds. and its ﬁber critical set is deﬁned as Σφ = (dθ φ)−1 ZE ∗ . and ι let pM : B → M be a smooth submersion. we note that the equality λ∗ αM = dφΣφ φ implies that Lφ is exact lagrangian.7 Let M. The function φ is said to be nondegenerate if its ﬁber derivative is transverse to ZE ∗ . Moreover. An advantage of this generalization will be to allow a calculus which is more clearly invariant under changes of coordinates. Since E ⊥ is the union of the conormal bundles of the ﬁbers of pM . it follows from Example 3. the section dθ φ has a welldeﬁned intrinsic derivative (see [26]). Theorem 4. in which case Σφ is a smooth submanifold of B.15 If φ is nondegenerate. The bundle E ⊥ may be identiﬁed with the pullback p∗ T ∗ M .
1. p). A Morse family is said to be reduced at p ∈ B if φ is a reduced phase function at p. Restriction: If B is any open subset of B containing λ−1 (p). ι. then the phase function φ is said to be reduced at p. ·) acquires a degenerate critical point. Example 4. Evidently the ﬁbercritical set of φ equals the product Σφ × {0}. Addition: For any c ∈ R.From the preceding deﬁnitions. This is just the value of x for which φ(x. pB . which in turn equals the dimension of the kernel of (π ◦ λφ )∗ on Tp Σφ . pB . and n λφ (b. A phase function which generates a neighborhood of a point in a lagrangian submanifold is not unique per se. p). φ). φ) ∈ M(L. φ) by a nondegenerate quadratic form Q on Rn is deﬁned as the Morse family comprised of the submersion pB : B × Rn → M ˜ n given by composing pB with the projection along R . 3. Composition: If pB : B → M is a second submersion and g : B → B is a ﬁberpreserving diﬀeomorphism. 48 . The central result of this section is the following. ι. pB . If ι : L → T ∗ M is a lagrangian immersion and p ∈ L. it follows that the nullity k of the ﬁberhessian at p ∈ Σφ equals dim(Tp Σφ ∩ Ep ).16 Suppose that M = R and consider the function φ(x. which fails to be xprojectable precisely when x = 0.” a concept which we now deﬁne.17 A triple (B. and its ﬁber critical set consists of the parabola x = −θ2 . This occurs when its ﬁberhessian vanishes at p. φ If (B. ι. ι. ι. θ) = xθ + θ3 /3 on B = R × R. and φ is a nondegenerate phase function on B such that λφ is an embedding. 0) ∈ Σφ . For such Morse families. but it is unique up to “stable equivalence. p). then we denote by M(L. Deﬁnition 4. ˜ ˜ ˜ ˜ 4. p). if the ﬁber dimension of B equals k. p). then (B . then the following operations produce further elements of M(L. ι. pB . pB . Evidently the phase function φ is nondegenerate. we introduce the following terminology. φ) is called a Morse family over a manifold M if pB : B → M is a smooth (possibly nonsurjective) submersion. φ ◦ g) ∈ M(L. we denote by gφ : U → Σφ the diﬀeomorphism deﬁned by gφ = λ−1 ◦ ι. p). To begin. These operations generate an equivalence relation among Morse families called stable equivalence. together with the phase function ˜ ˜ φ = φ + Q. then the restrictions of φ pB and φ to B deﬁne a Morse family on M which belongs to M(L. dθ φ = θ2 + x. and (dθ φ) = 2θdθ + dx. φ) which generate ι(U ) for some neighborhood U ⊂ L of p. φ) ∈ M(L. pB . This means in particular that the dimension of the ﬁbers of B must be at least k. Here. Thus (B × R . pB . Suspension: The suspension of (B. pB . 2. p) the class of Morse families (B. 0) = λφ (b) for all (b. We will say that the lagrangian submanifold im(λφ ) = Lφ is generated by the Morse family (B. φ + c) ∈ M(L. (B. ι.
ω1 be two symplectic forms on P which coincide on TN P . Then there is a symplectic immersion ψ of a neighborhood U of the zero section Z ⊂ T ∗ L into P such that ψ = ι ◦ π on Z and which maps the vertical subbundle VZ L ⊂ TZ (T ∗ L) onto E. Theorem 4. by a map which is the identity on L.20 If L is a lagrangian submanifold of P .Theorem 4. p) are stably equivalent. a symplectic manifold looks like a neighborhood of the zero section in T ∗ L. 2. ι. lagrangian subbundles of ι∗ P complementary to the image of ι∗ : T L → ι∗ T P always exist. Symplectic normal forms The purpose of this section is to develop several results needed to prove Theorem 4.24 to yield: Corollary 4. Theorem 4. Theorem 4. all of which more or less rely on the socalled deformation method. ω) be a symplectic manifold.19 will rely on the following Relative Darboux theorem [62] . Thus. As a result of the existence of compatible complex structures on ι∗ P . Let N be a submanifold of a manifold P . The class M(L. p) contains a reduced Morse family over M .19 Let (P. 49 . and E a lagrangian subbundle of ι∗ T P which is complementary to the image of ι∗ : T L → ι∗ T P . ι.21 The map ψ gives a 11 correspondence between a neighborhood of L in the space of all lagrangian submanifolds of (P. f N = id and T f TN P = id. We will prove the following theorem. This method was introduced by Moser in [45] (it probably has a longer history) and applied to a variety of problems of symplectic geometry in [62]. then a neighborhood of L in P is symplectomorphic to a neighborhood of the zero section in T ∗ L. ι : L → P a lagrangian immersion. The next two subsections are devoted to the proof of this theorem. ω) and a neighborhood of the zero section in the space Z 1 (L) of closed 1forms on L. and let p ∈ L.19 implies the following nonlinear “normal form” result. Any two members of M(L. The proof of Theorem 4.18. Corollary 4. f ∗ ω1 = ω0 2. Then: 1. which states that. near any of its lagrangian submanifolds L.19 may be combined with Proposition 3.18 Let ι : L → T ∗ M be a lagrangian immersion. Then there are neighborhoods U and V of N and a diﬀeomorphism f : U → V such that 1. and let ω0 .
To ﬁnd f satisfying f ∗ ω1 = ω0 . the form ϕ would be easy to ﬁnd. Deﬁne a 1parameter family of closed forms on P by ωt = ω0 + t(ω1 − ω0 ). 1]. ∗ then homotopy invariance for deRham cohomology implies that the forms ωt = gt ω lie in the same cohomology class for all t. such that ω −1 dϕ = ω1 − ω0 . ω) is a symplectic manifold and gt : P → P a continuous family of diﬀeomorphisms. β) dt . we also want Xt N = 0. we will identify P with a tubular neighborhood of N in P and let ht : P → P be a smooth isotopy such that h1 = id Since β vanishes on N . it is necessary and suﬃcient that such a vector ﬁeld solve the equation 0= d ∗ (f ωt ) = ft∗ dt t dωt dt + ft∗ (LXt ωt ) = ft∗ (ω1 − ω0 + d(Xt ωt )). By the usual properties of the Lie derivative. since the closed 2form ω1 − ω0 is locally exact by the classical Poincar´ lemma. Then there is a form ϕ on a neighborhood of N such that dϕ = β and which vanishes on TN P . Furthermore. By the same method of proof as in the relative Darboux theorem. For e more general submanifolds. Since all ωt agree on the submanifold N . if β vanishes on TN P . we will construct a timedependent vector ﬁeld Xt for which the isotopy ft that it generates satisﬁes ft∗ ωt = ω0 for all t ∈ [0. we have 1 h0 = ﬁberwise projection of P onto N. Our assertion follows if we take ϕ = h∗ (Yt t 2 If (P. the last expression is equal to 1 1 h∗ d(Yt t 0 β) dt = d 0 1 0 h∗ (Yt t β) dt. then ϕ can be chosen so that its ﬁrst derivatives vanish along N . there is a neighborhood of N in P (which for our purposes we may assume to be P itself) on which all ωt are nondegenerate. In order to ﬁx N . β = h∗ β − h∗ β = 1 0 0 d ∗ (h β) dt = dt t 1 h∗ (LYt β) dt. Since the statement is local around N . t 0 where Yt is the (timedependent) vector ﬁeld which generates the isotopy ht for t > 0. and let β be a closed kform on e P which vanishes on T N . By Cartan’s formula and the fact that β is closed. we use a more general version of this lemma: Relative Poincar´ lemma . Let N be a submanifold of P . If the submanifold N consisted of a single point. Proof. we can deduce the following converse result: 50 . vanishing on N . These conditions will be satisﬁed if we set Xt = −˜ t (ϕ) for a 1form ϕ on a neighborhood of N in P .Proof.
By restricting to a neighborhood of a point in N . Lemma 4. where Z is the zero section of T ∗ L and π : T ∗ L → L is the natural projection. we let f : Z → P denote the composition ι ◦ π. Givental [6] proves an equivalence theorem like the relative Darboux theorem under the weaker hypothesis that ω0 and ω1 coincide on T N . 51 . Parametrized Morse lemma . Proof of Theorem 4. Then for each x0 ∈ M .22. we choose a ωL compatible complex structure J on TZ (T ∗ L) which rotates T Z into VZ L.20.Corollary 4. gk whose diﬀerentials are linearly independent along N . i. the form ωt − ω0 is not exact. Let f : M × Rk → R satisfy the condition that for each x ∈ M . just for vectors tangent to N . If f is any function such that f and df vanish at all points of N . · · · . 2 Next. McDuﬀ [40] describes a family ωt of symplectic forms on a compact sixdimensional manifold P such that ω1 − ω0 is exact. t2 .19. Using Theorem 3. 0) is a nondegenerate critical point for f {x}×Rk . but there is no diﬀeomorphism f : P → P at all satisfying f ∗ ω1 = ω0 . To ﬁnish the proof of the normal form theorem. then there is a diﬀeomorphism f : P → P ∗ with f1 ω1 = ω0 . and a diﬀeomorphism u of U ﬁxing M × {0} and preserving ﬁbers of the projection to M such that f (u(x. we then choose a ωcompatible complex structure J on f ∗ T P which satisﬁes J(f∗ T Z) = E. we apply the relative Darboux theorem to the forms ωL and F ∗ (ω) on U .23 Let N ⊂ B be a submanifold deﬁned by the vanishing of functions g1 . we turn to two generalizations of the classical Morse lemma (see Appendix B) which require the following version of Taylor’s theorem.22 (Moser [45]) If {ωt } is a family of symplectic structures on a compact manifold P and ωt1 − ωt2 is exact for all t1 . for intermediate t. θ)) = f (x. Beginning with the lagrangian immersion ι : L → P . there exists a neighborhood U of (x0 . where E is the lagrangian subbundle given in the statement of the theorem. 0) + Q(θ). (x.e. In a neighborhood U ⊂ T ∗ L of Z. Of course. a nondegenerate quadratic form Q on Rk . In contrast to Corollary 4. w ∈ T Z. 0).j ˜ f cij gi gj on a neighborhood of N . Similarly. Now consider the symplectic bundle map TZ (T ∗ L) → f ∗ T P ˜ given by f (v ⊕ J w) = f∗ v ⊕ J(f∗ w) for v. we ˜ can extend f to obtain an immersion F : U → P which satisﬁes F∗ = f on TZ (T ∗ L). then there exist functions cij such that f= i.
Deﬁne ft = f − (1 − t)a. a function η : M × Rn → R. Evidently. By a preliminary change of coordinates linear on ﬁbers. We seek a vector ﬁeld Xt tangent to the ﬁbers x = constant which generates an isotopy ut ﬁxing M × {0} and satisfying ft ◦ ut = f0 for all t ∈ [0. This means that Xt should be of the form Xt = i hi t ∂ . Let φ be a function on B = M × Rn+k whose ﬁberhessian Hφ has rank k at a point b ∈ Σφ . and it should satisfy d (ft ◦ ut ) = u∗ (Xt · ft + a) t dt for all t ∈ [0.23 in order to ﬁnd smooth functions cij which vanish t on M × {0} and satisfy ∂ft ∂ft a= cij . then by the lowersemicontinuity of rank Hφ. ∂θi for certain smooth functions hi t which vanish on M × {0}.Proof. we may assume that f (x. θ. θ) where the error term a(x. The function η is totally degenerate in the sense that its ﬁberhessian is identically zero at b. θ ) = η(x. t ∂θi ∂θj The required condition ∂ft ∂ft ∂ft =− cij hi t t ∂θi ∂θi ∂θj i. To ﬁnd the diﬀeomorphism u we apply the deformation method. θ) = f (x. 1]. ∂θj 2 which vanishes (to second order) on M × {0}. 1]. 2 52 . To determine Xt .j i will be satisﬁed if we set hi = − t j cij t ∂ft . If φ is a phase function on M × Rn+k whose ﬁberhessian Hφ has rank k at a point b ∈ Σφ . A change of coordinates near b and an application of the parametrized Morse lemma then proves: Thom splitting theorem . and a nondegenerate quadratic form Q on Rk such that (φ ◦ g)(x. θ) + Q(θ ) in a neighborhood of b in B. there exists an integrable subbundle F → E = ker(pM ∗ ) such that the ﬁberhessian ∗ ◦ dθ φ ◦ : F → F ∗ is nondegenerate at each zero of the section ∗ ◦dθ φ in a neighborhood of b. we invoke Lemma 4. θ) is O(θ3 ) for each x. Then there exists a ﬁberpreserving diﬀeomorphism g of B. 0) + Q(θ) + a(x. the latter condition will be met if Xt is chosen so that 0= Xt · ft + a = 0.
p) is nonempty. Consequently. p) is stably equivalent to a reduced Morse family in M(L. To show that M(L. and so the class M(L.24 Any Morse family in M(L. Since this property is local near p.18. we ﬁrst prove Theorem 4. By the relative Poincar´ lemma. φ) is a Morse family such that Σφ = Z and λφ = ι ◦ π for the projection π : T ∗ L → L. F coincides with the ˜ ˜ ˜ map λφ on the ﬁbercritical set Σφ near q. By an application of the implicit function theorem. p) are stably equivalent.18. ι. ι. we may ˜ assume that M = Rm and and that φ. φ are both deﬁned on Rm × Rk . y) = x.9 we can construct a subbundle F which satisﬁes the hypotheses of the preceding theorem over a neighborhood of any p in L.24.19. ∂x ˜ ˜ By the assumption that φ. Since ψ is symplectic. p). Then L is generated by a Morse family over M . p) is nonempty. 2 Proof of Theorem 4. π ◦ ψ. F : Rm × Rk → Rm × Rm deﬁned by F (x. φ are nondegenerate phase functions. Since the Morse families ˜˜ are reduced. there exists a Morse family generating a neighborhood of p. ˜ ∂φ (x.24 to show that any two reduced Morse families in M(L. and similarly for F . we can therefore ﬁnd a ﬁberpreserving map g 53 . it follows that the images of DFq and DFq coincide in Tp (Rm × Rm ). Combined with the deﬁnition of αL . Using Lemma 3. it suﬃces by Theorem 4. π (ψ(B)). ∂φ (x. Theorem 4. there e ∗ exists a function φ on U satisfying dφ = ψ αM − αL . To prove part (2). Moreover. there exists a symplectic immersion ψ from a neighborhood U of the zero section Z ⊂ T ∗ L into T ∗ M which satisﬁes ψ = ι ◦ π on Z and which maps VZ L onto the subbundle F . we ﬁrst note the following reinterpretation of the Thom splitting theorem in the terminology of Morse families. Its restriction to Z is exact because αL is zero on Z. To complete the proof. it follows that F and F are −1 −1 embeddings near q = λφ (p) and q = λφ (p) respectively. y) = x. we ﬁrst note that the restriction of dφ to a ﬁber of the composition of ψ with the projection π : T ∗ M → M equals −αL . the 1form ψ ∗ αM − αL is closed on U .18. By Theorem 4.Existence and stable equivalence of Morse families Returning to the proof of Theorem 4. ι. this proves part (1) of Theorem 4.25 Suppose that ι : L → T ∗ M is an exact lagrangian immersion such that ι∗ T (T ∗ M ) admits a lagrangian subbundle F transverse to both ι∗ V M and ι∗ T L. Combined with Theorem 4. a computation (using the transversality hypothesis) now shows that there exists a neighborhood B of Z within U such that (B. Consider the ˜ mappings F. ι. y) ∂x ˜ F (x. Proof. ι. y) .
and so the phase functions φ and φ ◦ g diﬀer by an additive constant (which we may assume to be zero) at points of Σφ . 1]. To arrive at an equation for these coeﬃcients. we seek a vector ﬁeld Xt generating an isotopy ft that satisﬁes ft∗ φt = φ0 for all t ∈ [0. Hence we can solve for H in a neighborhood of b. H = (hij ). φ1 = φ ◦ g. and so there exist functions cij deﬁned near b such that φ1 − φ0 = i. ˜ Recall from the proof of Theorem 4. where C = (cij ). Then Σφ0 = Σφ1 = Σ and λφ0 = λφ1 .15 that the identity dφ = λ∗ αM φ ˜ holds on the ﬁber critical set Σφ . As before. While there is no reason to expect this of g itself. we will construct a similar diﬀeomorphism with this property by appealing once again to the deformation method: ˜ Let φ0 = φ.j cij ∂φ0 ∂φ0 ∂θi ∂θj in a neighborhood of Σφ . and S is a matrix function which vanishes at b for all t. we must also require that Xt be of the form ∂φ0 ∂ Xt = hij ∂θi ∂θj i. since φ0 is reduced. Moreover φ1 − φ0 vanishes to second order along Σ. This equation holds if 0 = C + H(I + S). 0= i. this identity implies that dφ = d(φ ◦ g) ˜ ˜ on Σφ . 2 54 .e. to insure that each ft ﬁxes Σ and preserves ﬁbers. In order for g to deﬁne an equivalence between φ and ˜ φ.of Rm × Rm whose restriction to some neighborhood B of q is a diﬀeomorphism which sends ˜ Σφ to Σφ and in particular g(q) = q .j for certain functions hij which vanish on Σ. Since λφ = λφ ◦ g.j d ∗ (ft φt ) = ft∗ (Xt · φt + φ1 − φ0 ) dt cij ∂φ0 ∂φ0 + ∂θi ∂θj hij i. This completes the proof. we note that the equation 0= will be satisﬁed provided that 0 = Xt · φt + φ1 − φ0 .v cuv ∂φ0 ∂φ0 ∂θu ∂θv . this property would have to be valid on all of B. i.j ∂φ0 ∂ ∂θi ∂θj φ0 + t u.
Maslov objects If ι : L → T ∗ M is any lagrangian immersion, then the symplectic vector bundle ι∗ T (T ∗ M ) over L has two lagrangian subbundles, L1 = ι∗ T L and L2 = ι∗ V M . The Maslov class of (L, ι) is deﬁned as the degree1 cohomology class µL,ι = µ(L1 , L2 ) ∈ H 1 (L; R) described in Example 3.21. From this deﬁnition, it follows that, unlike the Liouville classes, the Maslov classes of two immersions (L, ι) and (L, ι ) are equal whenever ι and ι are homotopic through lagrangian immersions. Associated to any Morse family (B, pB , φ) over a manifold M is an index function indφ : Lφ → Z deﬁned by indφ (p) = index(Hφλ−1 (p) ),
φ
where the index of a quadratic form is, as usual, the dimension of the largest subspace on which it is negativedeﬁnite. Since the ﬁberhessian is nondegenerate where Lφ is projectable (see the discussion following Theorem 4.15), the index function indφ is constant on any connected projectable subset of Lφ . From Theorem 4.18, it follows furthermore that any two index functions indφ , indφ diﬀer by an integer on each connected component of Lφ ∩ Lφ . Example 4.26 Consider φ(x, θ) = θ3 /3 + θ(x2 − 1). We have ∂φ/∂θ = θ2 + x2 − 1; thus the ﬁber critical set Σφ is the circle θ2 + x2 = 1, and its image under λφ is the ﬁgureeight. The caustic set of the projection L → R consists of the points (−1, 0), (1, 0). To compute the index function corresponding to φ, we note that ∂2φ = 2θ; ∂θ2 consequently indφ (x, p) = 0 for (x, p) lying on the upper right part of the curve and ind(x, p) = 1 for (x, p) on the upper left. Since this loop is generated by a single phase function, there is a corresponding global index function, and the index of the loop is necessarily zero. A diﬀerent situation occurs in the case of a circle. The circle has two caustics, and the jump experienced by any index function as one passes through a caustic is given by the example just computed: passing through the right caustic in the +p direction decreases the index by 1, while passing through the left caustic in the same direction increases the index by 1. Traversing the circle in a counterclockwise direction, we see that the total index must change by −2. Consequently, the circle does not admit a global generating function. Of course, the Liouville class of the circle is also nonzero. However, it is easy to deform the circle to a closed curve with two transverse selfintersections which has zero Liouville class, but still admits no global phase function, since its Maslov class is nonzero.
55
As a degree1 cohomology class on L, the Maslov class determines via the exponential map R → U (1) an isomorphism class of ﬂat hermitian line bundles over L. A canonical representative of this class can be constructed as follows. First, we consider the union M(L, ι) =
p∈L
M(L, ι, p)
with the discrete topology. On the subset of L×M(L, ι)×Z consisting of all (p, (B, pB , φ), n) such that (B, pB , φ) ∈ M(L, ι, p), we now introduce an equivalence relation ∼ by setting (p, (B, pB , φ), n) ∼ (˜, (B, pB , φ), n) provided that p = p and p ˜ ˜ ˜ ˜ ˜ n + indφ (p) = n + indφ (p). ˜ ˜ The quotient space with respect to this relation is a principal Zbundle ML,ι over L which we will call the Maslov principal bundle. Associated to the Maslov principal bundle via the representation n → eiπn/2 of Z in U (1) is a complex line bundle ML,ι over L called the Maslov line bundle. Having discrete structure group, this line bundle carries a natural ﬂat connection with holonomy in {eiπn/2 } Z4 . Our main use of the Maslov line bundle will be to modify the halfdensities on L in order to incorporate the Maslov correction into our quantization scheme.
4.4
WKB quantization
In this section we will combine the tools assembled in the preceding sections into a technique for quantizing halfdensities on lagrangian submanifolds of arbitrary cotangent bundles. The phase bundle associated to an immersed lagrangian submanifold ι : L → T ∗ M and > 0 is deﬁned as the tensor product ΦL,ι, = ML,ι ⊗ ι∗ EM, , where we recall that EM, is the prequantum line bundle over T ∗ M (see Section 4.1). Observe that the product of the natural ﬂat connections on ML,ι and ι∗ EM, deﬁnes a ﬂat connection on the phase bundle whose holonomy is represented by the modZ reduction of the real cohomology class ˇ λL,ι + π µL,ι /2 ∈ H 1 (L; R), which we call the phase class of (L, ι). The phase bundle ΦL,ι, can be described explicitly as the collection of all quintuples (p, t, (B, pB , φ), n, z) where (p, t, n, z) ∈ L × T × Z × C and (B, pB , φ) ∈ M(L, ι, p), modulo the equivalence relation ∼ given by (p, t, (B, pB , φ), n, z) ∼ (˜, t, (B, pB , φ), n, z ) p ˜ ˜ ˜ ˜ ˜ ˜ whenever p = p and ˜
n ˜ z · e−it/ eiπ(n+indφ (p))/2 = z · e−it/ eiπ(˜ +indφ (p))/2 . ˜ ˜ def
56
A Morse family (B, pB , φ) which generates an open subset Lφ of L deﬁnes a nonvanishing parallel section of ΦL,ι, over Lφ by sφ, (p) = [p, 0, (B, pB , φ), 0, e−iφ(y)/ ], where λφ (y) = p. A check of these deﬁnitions shows that whenever ι(p) = λφ (y) = λφ (˜), ˜ y
y ˜ sφ, (p) eiφ(y)/ e−iπ indφ (p)/2 = sφ, (p) eiφ(˜)/ e−iπ indφ (p)/2 . ˜ ˜
For each ∈ R+ , we denote by Γpar (ΦL,ι, ) the space of parallel sections of ΦL,ι, . If, for a particular , the phase class of (L, ι) is integral, then Γpar (ΦL,ι, ) is a complex vector space isomorphic to C. Otherwise, Γpar (ΦL,ι, ) consists of a single point (the zero section of ΦL,ι, ). The product ΓL,ι = Γpar (ΦL,ι, )
>0
then has the structure of a Cmodule. An element s ∈ ΓL,ι is then a (generally discontinuous) function which assigns to each > 0 an element s in Γpar (ΦL,ι, ), so that the map p → s (p) deﬁnes a parallel section of ΦL,ι, . The symbol space of (L, ι) is deﬁned as the complex vector space def SL,ι = Ω1/2 L ⊗C ΓL,ι . The amplitude bundle Aφ associated to a Morse family (B, pB , φ) over a smooth manifold M is deﬁned as the complex line bundle Aφ = Λ1/2 B ⊗ Λ1/2 E over B, where E again denotes the subbundle ker(pB∗ ) of T B. An amplitude is a section a of Aφ . We will say that a is properly supported provided that the restriction of pB : B → M to Supp(a) is a proper map. The purpose of the space of amplitudes is to deﬁne a relation between halfdensities on M and symbols on the subset Lφ of L generated by φ. To describe this relation, we begin by noting that from the exact sequence of vector bundles 0 → E → T B → p∗ T M → 0 B over B, it follows that Λ1/2 B is naturally isomorphic to Λ1/2 p∗ T M ⊗ Λ1/2 E. This in B turn gives rise to the natural isomorphism Aφ Λ1/2 p∗ T M ⊗ ΛE. B
The image of an amplitude a on B under this isomorphism can be written as p∗ dx1/2 ⊗ σ, B where σ is a family of 1densities on the ﬁbers of pB , i.e., σx is a density on each nonempty p−1 (x). By ﬁberintegration we pass to a halfdensity on M : B I (φ, a)(x) = (2π )−n/2 e−inπ/4
p−1 (x) B
eiφ/ σx
dx1/2 ,
57
and −1 so there exists a diﬀeomorphism g : Σφ → Σφ deﬁned as the composition g = gφ ◦ gφ .3 that the nondegeneracy of φ gives rise to the following exact sequence of vector bundles over the ﬁbercritical set Σφ θ 0 → T Σφ → TΣφ B → E ∗ Σ → 0. pB . a be amplitudes on B. we ﬁrst recall from Section 4. When (L. Then sa = s˜ on L if and only if a ˜ ˜ I (φ. ) deﬁned above). a) − I (φ. def (Here we recall from the discussion following Deﬁnition 4. a) is a smooth halfdensity on M. Lemma B. Since ˜ ˜ φ = φ ◦ g. a) are linked by the following theorem. is the canonical element of Γpar (ΦLφ . (B. the associated symbol on Lφ is deﬁned as ∗ sa = gφ a ⊗ sφ . ˜ ˜ ˜ Proof. a)(x) = eiφ/ e−iπ indφ /2 + O( )  detax Hφ1/2 ˜ ˜ and similarly for I (φ. setting I (φ. for each φ . pB . ι) is projectable. ˜ ˜ ˜ (∗) Since (L. the symbol sa and the halfdensity I (φ. pB . 58 . ˜ ˜ ˜ Theorem 4. pB . a)(x) = 0 if p−1 (x) = ∅.ι. (B. this sequence induces an isomorphism of the restriction of Aφ to Σφ with Λ1/2 Σφ . The theorem follows by comparing these expressions with (∗) above. φ) are stably equivalent. ι). φ) generate the same ˜ ˜ projectable lagrangian embedding (L. To pass geometrically from a to a symbol on Lφ .where n = dim(p−1 (x)). A ˜ ˜ check of the deﬁnitions shows that the symbols sa and s˜ are equal precisely when a g ∗ a · eiφ/ e−iπ indφ /2 = a · eiφ/ e−iπ ind φ/2 . a) = O( ) locally uniformly on V . the section sφ. ˜ where a is the halfdensity on Σφ induced by the amplitude a. B. φ). respectively. a)(x). we can apply the principle of stationary phase to each ﬁber of the projection pB : B → M to obtain I (φ. d φ Since Λ−1/2 E is naturally isomorphic to Λ1/2 E ∗ . up to a constant. When a is properly supported. the Morse families (B. If the restriction of a to Σφ corresponds to a halfdensity a on Σφ under this isomorphism. B B we may diﬀerentiate under the integral to conclude that I (φ. and let a. φ). ι) is projectable.27 Suppose that two Morse families (B. By Theorem 4.3 implies that this occurs precisely when ˜ ˜ ˜  detax Hφ1/2 e−iφ/ eiπ indφ /2 =  detax Hφ1/2 e−iφ/ eiπ indφ /2 .17 that gφ is a diﬀeomorphism from a neighborhood of p in L onto Σφ deﬁned by the composition λ−1 ◦ ι. Also.18. and similarly for a.
s) = j I (L. the existence of nonzero symbols requires that the phase bundle ΦL. or.27 asserts that this choice of I (L. it is easy to check that up to O( ) terms. s) is welldeﬁned. Furthermore. such that each Lj is generated by a Morse family (Bj . From Theorem 4. Deﬁnition 4. φ). its phase bundle ΦL. ι. a) = O( ). ι. Of course. ι. Thus. U is the image of the normal bundle νN ⊂ TN M under an embedding ψ : νN → M satisfying ψ = π on the zero section of νN . hj · s). that the phase class of (L. For this reason.ι admit nontrivial parallel sections.. pN . To quantize an arbitrary symbol s on L.27 we draw two conclusions about this tentative deﬁnition when Lφ is ˜ ˜ ˜ projectable. a)(x). although the technique above enables us to quantize all symbols on L in a consistent way. φ) is a second Morse family which generates Lφ . we introduce the following terminology. and consider a symbol s on L supported in Lφ . which we can extend to an amplitude a on B. we ﬁrst ﬁx a locally ﬁnite cover {Lj } of L. I (L. ι. The set of for which this condition holds will be called admissible for L. We then set I (L.e. s). ι. then sa = sa = s.e. for some ∈ R+ . a) − I (φ. Consider the Morse family (r∗ N ⊥ . if its phase class is integral. ˜ ˜ and a is an amplitude on B obtained as above. ι) ⊂ T ∗ M .28 An immersed lagrangian submanifold ι : L → T ∗ M is called quantizable if. up to O( ) terms. compactly supported in ﬁbers. ι. and so ˜ ˜ I (φ. i. pB . we note that if (B. s) coincides with pullback. where r = π ◦ ψ −1 is a 59 .29 Let N be a closed submanifold of a smooth manifold M . pBj . Suppose that (B. we note that Theorem 4. and choose a partition of unity {hj } subordinate to {Lj }. and (gφ )∗ a may be canonically identiﬁed with a section of the amplitude bundle of B over the ﬁbercritical set Σφ .ι admits a global parallel section. where π : νN → N is the natural projection. Then there exists a unique halfdensity a supported in Lφ such that s = sφ ⊗ a. φj ). i. Note that it is a straightforward generalization of the condition derived in the preceding section. s) = I (φ. equivalently. Example 4. this deﬁnition depends only on the semiclassical state (L. pB .2 Morse families provide a general means for locally quantizing symbols on an immersed lagrangian submanifold (L. Using the remarks above. First. We then set I (L. φ) is a Morse family such that the phase function φ generates an open subset Lφ ⊂ L. This deﬁnition is known as the Maslov quantization condition. and let U be a tubular neighborhood of N . ι) be integral.
ι. Quantum states as distributions Unfortunately. and stationary phase cannot be used to estimate the integral eiθ R 3 /3 a(0. s) in the presence of caustic points is as a family of distributional halfdensities on M deﬁned as follows. Indeed. the interpretation of I (L. and therefore both the Liouville and Maslov classes of N ⊥ are zero. ι. this implies that the conormal bundle of any submanifold of M satisﬁes the Maslov quantization condition. pN denotes the natural submersion r∗ N ⊥ → M . and φ : r∗ N ⊥ → R is deﬁned by φ(p) = p. The origin is therefore a degenerate critical point for φ. pB . u = (2π )−b/2 e−iπb/4 B 1/2 eiφ/ a ⊗ p∗ u. this remark is suggested by the fact that I (L. s) is smooth. Thus. θ) dθ. s) at regular values of πL is not valid at caustics. The family I (L. since the phase function φ has a degenerate critical point in the ﬁber p−1 (x). B where b = dim(B) and u ∈ Ω0 M . a computation shows that the ﬁber critical set of φ is given by Σφ = p−1 (N ) = N ⊥ . Since ψ is an embedding. a). whereas we saw in Section 2.2 that classical solutions to the transport equation are singular at caustic points. ι. φ) ∈ M(L.27) no longer applies to the integral eiφ/ σx p−1 (x) B when x is a caustic point. B Example 4. at whose points the ﬁberhessian assumes the form ∂ 2 φ/∂θ2 = 2x. ψ −1 (pN (p)) . In particular.retraction of U onto N . ι) and compactly supported amplitude a on B. ι. the conormal N bundle of N is a lagrangian submanifold of T ∗ M which admits a global generating function. A more appropriate way to interpret the expression I (L. and the map λφ : N ⊥ → T ∗ M equals the inclusion. s) then consists of all distributional halfdensities I on M obtained by choosing a locally ﬁnite open cover {Lj } of L and a 60 . θ) = θ3 /3 + xθ consists of the parabola x = −θ2 . we deﬁne a distributional halfdensity on M by I (φ.30 The ﬁber critical set of the phase function φ(x. For each Morse family (B. The basic technical diﬃculty is that the principle of stationary phase (the basic underpinning of Theorem 4.
a). a link among its members can be described as follows. ˜ y  deta R (y)1/2  deta R (˜)1/2 ˜ ˜ ˜ (∗∗) The critical point y of R is nondegenerate precisely when the intersection of ι(L) with im(dψ) is transverse. φ) implies that ˜ y indφ (p) − indφ (˜) = ind R (y) − ind R (˜). s) may appear very large. ˜˜ ˜ ˜ where a is the halfdensity on Σφ induced by the amplitude a. equality of the symbols sa and s˜ occurs precisely when a the diﬀeomorphism g : Σφ → Σφ satisﬁes ˜ y g ∗ ay · eiφ(y)/ e−iπ indφ (p)/2 = ay · eiφ(˜)/ e−iπ ind φ(p)/2 . a be amplitudes on B. and so the preceding equation gives ˜ y y y eiR(y)/ e−iπ ind R (y)/2 eiR(˜)/ e−iπ ind R (˜)/2 = . φ). respectively. aj ). pB . pB . In this case. e−iψ/ u = eiR(y)/ e−iπ ind R (y)/2 + O( )  deta R (y)1/2 ˜ ˜ and similarly for I (φ. Since ˜ ˜ = R ◦ g. Lemma B. a). We ﬁrst prove ˜ ˜ ˜ Theorem 4. B. it is easy ˜ ˜ ˜ to check that stable equivalence of the Morse families (B. and let a. The family I (L.3 shows that this condition is equivalent to R ˜y ˜ ˜ y  deta R (y)1/2 e−iφ(y)/ eiπ indφ (p)/2 =  deta R (˜)1/2 e−iφ(˜)/ eiπ indφ (p)/2 . Since ψ ◦ pB is constant on the ﬁbers of pB . (B. If ψ : M → R is a smooth function whose diﬀerential intersects ι(L) at exactly one point ι(p) transversely. e−iψ/ u − I (φ. pB . Moreover. e−iψ/ u = O( ). ˜ ˜y ˜ y Since g is ﬁberpreserving. φ) generate the same ˜ ˜ lagrangian embedding (L. As in the proof of Theorem 4. we have R(y) − R(˜) = φ(y) − φ(˜). e−iψ/ u . an application of the principle of stationary phase gives I (φ. up to a constant. ι.partition of unity {hj } subordinate to {Lj }. ι. 2 61 . Set R = φ − ψ ◦ pB . (B. s) consists of those distributional halfdensities obtained in this way using amplitudes aj such that saj = s over each Lφj . a). then sa (p) = s˜ (p) if and only if a ˜ ˜ I (φ. and similarly for a. φ). the function R is a phase function having the same ﬁber critical set as φ. Although the class I (L. Proof. a). Then set k I = j=1 I (φj .31 Suppose that two Morse families (B. pB . ι).27. Comparing this expression with (∗∗) above completes the proof.
the principal part of I (L. or. 31]. As before. we can sometimes deﬁne it even when a does not have compact support.ι → SL . s). φ) is a Morse family over a manifold M . Example 4. and suppose ˜ that S : V → R is a smooth function. e−iψ/ depends only on the principal part of s. equivalently. then x is a critical point ˜ of φ − S. Moreover. This critical point is nondegenerate provided that λφ−S is transverse to the zero ˜ ∗ section of T M near x. e−iψ/ u = def j I (φj . On the classical level. We then deﬁne I (L. up to O( ). a) to be a distribution. Note in particular that if dS(pB (x)) = λφ (x) for some x ∈ Σφ . this means the following.Now consider a semiclassical state (L. By setting S = S ◦ pB .ι . Equivalent semiclassical states A further important modiﬁcation of our quantization picture is based on the conceptual distinction between an “immersion” and an “immersed submanifold”. ι. then the HamiltonJacobi equation H ◦ι=E 62 f∗ . aj ).32 Suppose that (B. ι. When the image of dψ is transverse to L. s). ι. e−iψ/ u . then. where fdS is the ﬁberwise ˜ ˜ translation map deﬁned by dS. if E is a regular value of some hamiltonian function H on T ∗ M . In terms of our discussion. φ − S) for which Σφ = Σφ−S and λφ = fdS ◦ λφ−S . where aj is the amplitude on Bj obtained in the usual way from the symbol hj ·s on Lj . φj ) over M . pB . and choose a partition of unity {hj } subordinate to {Lj }. any lagrangian immersion ι : L → T ∗ M deﬁnes an equivalence class of lagrangian immersions in T ∗ M which. ι ) are equivalent provided that there exists a diﬀeomorphism f : L → L such that ι = ι ◦ f . ι) and (L . pB . If ι : L → T ∗ M and ι : L → T ∗ M are lagrangian immersions. 28. A check of the deﬁnitions shows that the main objects in our quantization scheme behave nicely with respect to this notion of equivalence. In this way. we refer to [21. we obtain a new Morse ˜ family (B. pBj . for notational simplicity we will denote by L. We refer to the equivalence class L as an immersed lagrangian submanifold of T ∗ M . we let {Lj } be a locally ﬁnite cover of L such that each Lj is generated by a Morse family (Bj . s). we will say that (L. Thus. a symbol on the immersed lagrangian submanifold L is welldeﬁned as an equivalence class of symbols on the members of L. a diﬀeomorphism f : L → L as above induces an isomorphism of symbol spaces SL. For a thorough exposition of this topic. if λφ is transverse to the image of dS near x. By allowing I (φ.
s). ι.ι = f∗ XH. since there is no underlying conﬁguration space on which the states can live.ι induced on (L.13) satisfy XH. Extending some notion of quantum state to arbitrary symplectic manifolds is one of the central goals of geometric quantization. ω) to be an immersed lagrangian submanifold equipped with a halfdensity and possibly some “phase object” corresponding to a parallel section of the phase bundle in the cotangent bundle case. that it is unclear what the quantum states are which are approximated by these geometric objects. (L . we are led to view the equivalence class (L. s) on M obtained by quantizing any member of (L. s). s) satisfying eigenvector of H H(L) = E and LXH s = 0 With these correspondences between classical and quantum mechanics in mind. as described above. On the quantum level. s) satisﬁes the associated HamiltonJacobi and homogeneous transport equations. s) = I (L . In this case.ι . s) as the (unique) distributional halfdensity I (L. Note. The state is stationary with respect to a classical hamiltonian H on T ∗ M provided that (L. s) is a semiclassical approximate solution to the timeindependent Schr¨dinger equation on M . s) is a stationary semiclassical state.ι s = 0 is a welldeﬁned condition on (L. and therefore the homogeneous transport equation LXH.deﬁnes a condition on the class L which we denote by H(L) = E. From these remarks. s) which we denote by LXH s = 0. ι . s) as above Quantum version HM distributional halfdensity on M Schr¨dinger equation o ˆ operator H on HM timeevolution Hamilton’s equations generator of evolution function H on T ∗ M stationary state ˆ state (L. the vector ﬁelds XH. o Our list of classical and quantum correspondences now assumes the form: Object basic space state Classical version T ∗M (L. 63 . ι. ι ) by the hamiltonian vector ﬁeld of H (see Example 3. If (L. it is easy to check that for any members (L. ι . (L . s). we are further led to deﬁne a semiclassical state in an arbitrary symplectic manifold (P. s ) of the equivalence class (L. s) as a semiclassical state in T ∗ M .ι and XH. s ). and thus we may deﬁne the quantization of (L. then I (L. we have I (L. however. ι).
A special case of linear reduction arises when Y is a coisotropic subspace of a symplectic vector space and ω equals the restriction of the symplectic form to Y . Linear symplectic reduction Degenerate skewsymmetric bilinear forms yield symplectic vector spaces in the following way.2 Let V be a symplectic vector space and L. an important process known as symplectic reduction generates many examples of such manifolds. For instance. [y]) = ω(x.5 The Symplectic Category There are many symplectic manifolds which are not cotangent bundles. π∗ −− Y ∗ ← − (Y /Y ⊥ )∗ ω ˜ ˜ ω Y −− −→ π Y /Y ⊥ 2 If ω ([x]) = 0. To prove the nondegeneracy of ω .1 A skewsymmetric bilinear form ω on a vector space Y induces a symplectic structure on Y /Y ⊥ . We begin this section with a discussion of reduction and then turn to the classical and quantum viewpoints in the context of general symplectic manifolds. and so x ∈ Y ⊥ and hence [x] = 0. Lemma 5. First note that ω gives rise to a skewsymmetric bilinear form ω on Y /Y ⊥ by the equation ω ([x]. Proof. and LC = (L ∩ C)/(L ∩ C ⊥ ) is a lagrangian subspace of C/C ⊥ . y). we use the following commutative diagram. 64 . Lagrangian subspaces behave remarkably well with respect to this reduction. then ω (x) = 0. starting with cotangent bundles. Then LC = L ∩ C + C ⊥ is a lagrangian subspace of V contained in C. Lemma 5. 5.1 Symplectic reduction The technique of symplectic reduction geometrizes the process in mechanics in which ﬁrst integrals of the hamiltonian are used to eliminate variables in Hamilton’s equations. concluding with Dirac’s formulation of the quantization problem. C ⊂ V a lagrangian and a coisotropic subspace respectively. ˜ ˜ The symplectic quotient space Y /Y ⊥ described in this lemma is known as the (linear) reduced space associated to Y . where π is the projection.
65 . C 2 Example 5.3 If (V. the linear reduced space C/C ⊥ equals (V .4 If (V. Next.4). L . then there exists a linear symplectomorphism V → C ⊥ ⊕ V /C ⊕ C/C ⊥ . then (E + F ) ∩ G = E ∩ G + F if and only if F ⊂ G.Proof. Consequently. and we collect some useful facts about them in the following examples. (V . This gives rise to the following decomposition of V . ω) is a symplectic vector space with a coisotropic subspace C. Proof. then then C = V ⊕ L is a coisotropic subspace of the direct sum V ⊕ V (see Example 3. W ) = signature QT These quantities will be useful in our study of the Maslov bundle under reduction. B are orthogonal to C. then the reduced lagrangian subspace (ΓT )C of C/C ⊥ equals T (L) ⊂ V . Now suppose that V and V are of equal dimension and T : V → V is a linear symplectic map. If ΓT ⊂ V ⊕ V is the lagrangian subspace deﬁned by the graph of T . B give rise to an isomorphism V = C ⊥ + A + B → C ⊥ ⊕ V /C ⊕ C/C ⊥ . and C ⊥ = 0 ⊕ L. ω ) are symplectic vector spaces and L is a lagrangian subspace of V . F. the second assertion follows from the equality L⊥ = (LC )⊥ /C ⊥ . Note that (LC )⊥ = (L + C ⊥ ) ∩ C. ω). C ⊥ with respect to the innerproduct gJ on V . The projections V → V /C and C → C/C ⊥ restricted to A. then C ⊥ ⊕V /C also carries a natural symplectic structure induced by ω. Denoting by QT the quadratic form QT on L induced by T . Then A. 2 Recall that if L. W ) = index QT sgn(L. Since by assumption C ⊥ ⊂ C. G are any three subspaces of V . the selforthogonality of LC follows from the simple fact that if E. we deﬁne ind(L. ω ). Lemma 5. Let J be a ωcompatible complex structure on V . then there exists a natural linear symplectomorphism from V to L ⊕ L∗ which sends W onto the subspace 0 ⊕ L∗ and L onto the graph of some selfadjoint linear map T : L → L∗ . L . Observe that if C is a coisotropic subspace of (V. L are lagrangian subspaces of a symplectic vector space V and W ⊂ V is a lagrangian subspace transverse to both L and L . ω). and set A = JC ⊥ and B = C ∩ JC. we observe that since LC = LC /C ⊥ .
Proof. then the ﬁrst term on the right hand side of this equation vanishes identically. Evidently. Leaving for the reader the veriﬁcation that (T M )⊥ is actually a subbundle of T M . L. Theorem 5.e. then Cartan’s formula combined with the assumption that ω is closed implies that ω is invariant under the ﬂow of X. WC . the lagrangian subspace W of V ⊕V ∗ given by the graph of A passes under reduction by C to the graph WC of AE . then sgn(Q) = sgn(QE ) + sgn(Q∗ E ⊥ ).6 Suppose that V is a ﬁnitedimensional vector space with a subspace E ⊂ V and its algebraic orthogonal E ⊥ ⊂ V ∗ . W . the next best structure a manifold M may possess along these lines is a closed twoform ω of constant rank. If the vector ﬁeld Y belongs to (T M )⊥ . respectively. with the dimension of the orthogonal (Tx M )⊥ the same for all x ∈ M . In this case. If Q∗ denotes the quadratic form on V ∗ deﬁned by A. Then C = V ⊕E ⊥ is a coisotropic subspace of V ⊕V ∗ with its usual symplectic structure. and C ⊥ = E ⊕ 0. WC . so the second term vanishes as well. this formula gives sgn(L . L . we therefore have sgn(Q∗ ) = sgn(L .5 To begin. LC ). we recall that Lie brackets and inner products are related by the formula [X. Y ] lies in (T M )⊥ . The reduced symplectic vector space is then C/C ⊥ = (V /E) ⊕ E ⊥ . Combined with the preceding equations. p. The composition of a selfadjoint linear map A : V ∗ → V with the projection V → V /E restricts to a selfadjoint linear map AE : E ⊥ → V /E (note that V /E identiﬁes canonically with (E ⊥ )∗ ). and [X. Y ] ω = LX (Y ω) − Y LX ω. L are themselves transverse: sgn(L. Example 5. W ) = − sgn(L . W . ω is called a presymplectic structure on M with characteristic subbundle (T M )⊥ . LC ). L ). Now if we denote by L. If X belongs to (T M )⊥ .7 The characteristic subbundle of a presymplectic manifold (M.Example 5. L) = sgn(QE ) + sgn(LC . Presymplectic structures and reduction By deﬁnition. then LC = (V /E) ⊕ 0 and LC = 0 ⊕ E ⊥ . L the lagrangian subspaces V ⊕ 0 and 0 ⊕ V ∗ . a symplectic form is closed and nondegenerate. From [21. This means that LX ω = 0. i. we leave to the reader the job of checking the following elementary identities for the case when the lagrangian subspaces L. In some sense. ω) is integrable.. W . L) and sgn(Q∗ E ⊥ ) = sgn(LC . W ) = − sgn(L. 66 .130] we recall that if Q∗ is nondegenerate and Q is the quadratic form it induces on V .
where ι × : N × M → V × V is the product map and ∆V ⊂ V × V is the diagonal. these operations will be nonlinear analogs of the mappings L → LC and L → LC described in the linear setting in Lemma 5. Since dω = 0 by hypothesis. N to the ﬁber product N ×V M . then P × L is a coisotropic submanifold of P × Q whose characteristic foliation consists of leaves of the form {p} × L for p ∈ P . shows that the presymplectic structure ω induces a smooth. we will be interested in presymplectic manifolds which arise as coisotropic submanifolds of some symplectic manifold (P. Q are symplectic manifolds. rN denote the restrictions of the cartesian projections N × M → M. together with the fact that LX ω = 0 for characteristic X. The product P × L is therefore reducible. A pointwise application of Lemma 5. the hamiltonian ﬂow associated to a function H : R2n → R generates the characteristic foliation of any regular energy surface H −1 (E). Since dim(T C)⊥ = dim(P ) − dim(C) is constant. in eﬀect. the form ωM is necessarily closed and therefore symplectic. then we say that M is reducible.2 The foliation M⊥ deﬁned by the characteristic subbundle of M is known as the characteristic foliation of M .13. and since the quotient map M → M/M⊥ is a submersion. For the most part. the tangent space Tp C contains its symplectic orthogonal (Tp C)⊥ = {v ∈ Tp P : ω(v. Recall that a submanifold C ⊂ P is called coisotropic if. Our goal for the remainder of this section is to describe two operations on immersed lagrangian submanifolds deﬁned by a reducible coisotropic submanifold. The symplectic manifold (M/M⊥ . the form ω deﬁnes a presymplectic structure on P. If P. The corresponding reduced symplectic manifold (when it exists) will be the symplectic model for the space of quantum states of energy E.2. 67 ι r . nondegenerate 2form ωM on M/M⊥ . Example 5. ωM ) is called the reduced manifold of M .2. and the reduction projection coincides with the usual cartesian projection P × L → P . In this case. we can view C as an abstract manifold and note that if ω is the pullback of the symplectic form ω by the natural inclusion C → P . for each p ∈ C. If the quotient space M/M⊥ is a smooth manifold.8 As noted in Example 3. w) = 0 for all w ∈ Tp C}. then ker(ω ) = (T C)⊥ . and L is a lagrangian submanifold of Q. To begin. we recall that the ﬁber product of two maps ι : N → V and : M → V is deﬁned as the subset N ×V M = (ι × )−1 ∆V of N × M . ω). This gives rise to the following commutative diagram N ×V M − − M −M→ rN N −− V −→ where rM .
10 1. V are smooth manifolds. if ι and are cleanly intersecting immersions. Since rM ∗ (v. For brevity. then the map (pC ◦ rC ) : L ×P C → C/C ⊥ has constant rank. A basic example of two maps which do not intersect cleanly is provided by two embedded curves in the plane which intersect tangentially at a single point. ∗ ∗ Proof. then two smooth maps ι : N → V and : M → V are said to intersect cleanly provided that their ﬁber product N ×V M is a submanifold of N × M and ι∗ T N ∩ ∗ T M = ( ◦ rM )∗ T (N ×V M ) as subbundles of ( ◦ rM )∗ T V . the vector (ι × )∗ (v. w) = w. Then for any tangent vector (v. and so ι∗ v = ∗ w. Suppose that ι : N → V and : M → V are any smooth maps whose ﬁber product N ×V M is a smooth submanifold of N ×M . we can conclude that rM : N ×V M → M is an immersion as well. 68 . M.9 If N. w) is tangent to the diagonal in V × V . The kernel of (pC ◦ rC )∗ then equals G⊥ ∩ rC∗ T (L ×P C).Deﬁnition 5. we will say that a map ι : N → V intersects a submanifold W ⊂ V cleanly if ι and the inclusion map W → V intersect cleanly. Let F. and so the basic properties of symplectic orthogonalization applied ﬁberwise to these bundles give ker(pC ◦ rC )∗ = G⊥ ∩ G ∩ F = G⊥ ∩ F = (G + F )⊥ . w) of N ×V M . it follows that rM ∗ (v. then ι and intersect cleanly. A particular case of this situation occurs when is a submersion and ι is any smooth map.11 Let P be a symplectic manifold. Lemma 5. in which case ι∗ v = ∗ w = 0. The usefulness of clean intersection in symplectic geometry lies in its compatibility with symplectic orthogonalization. Thus. G denote the lagrangian and coisotropic subbundles rL T L and rC T C of the ∗ symplectic vector bundle rC T P over the ﬁber product L ×P C. Example 5. The clean intersection hypothesis implies that rC∗ T (L ×P C) = F ∩ G. then the maps rN and rM are themselves immersions. In particular. w) = 0 only if w = 0. 3. as in the following lemma. if ι is an immersion. 2. If a lagrangian immersion ι : L → P intersects a coisotropic submanifold C ⊂ P cleanly. If the product ι × of two smooth maps ι : N → V and : M → V is transverse to the diagonal ∆V .
With these remarks in mind. it follows that these are immersed lagrangian submanifolds. all of whose geodesics are closed and −1 have the same length. 2 Example 5. L). C intersect cleanly. Proof.13 If (M. 69 . J2 ).e. then LC and LC are immersed lagrangian submanifolds of C/C ⊥ and P . Let ι : L → P be a ﬁxed element of L. where ι : N → V is considered equivalent to ι : N → V provided that there exists a diﬀeomorphism f : N → N which intertwines ι and ι .2 implies that this immersion is lagrangian. the symplectic form on CM is given in terms of the metric by ω(J1 . g) is a riemannian manifold. the internal sum F + G and its orthogonal (F + G)⊥ also have constant rank. L) in P induces immersed submanifolds LC and LC of C/C ⊥ and P . In this case. then the same is true of any other member of N . In this case. respectively. By C : LC → C/C ⊥ we denote the restriction of the cartesian projection LC × C → C to LC . we have dim(ker(pC ◦ rC )∗ ) = dim(L) + dim(L ×P C) − dim(C).10(2) implies that the map C : LC → C is a smooth immersion. indicating that the dimension of ker(pC ◦ rC )∗ is independent of the base point in L ×P C. if a member ι : N → V of the equivalence class N intersects a smooth map : M → V cleanly. Recall from Chapter 4 that an immersed submanifold N of a manifold V is an equivalence class N of immersions. since the immersed submanifolds LC and LC are welldeﬁned by the reducible pair (C. Since LC is the quotient of L ×P C by the ﬁbers of pC ◦ rC . ι = ι ◦ f . A pointwise application of Lemma 5. and L. Since ιC : LC → C/C ⊥ is an immersion. J2 ) = g(J1 .12 If (C. i. we will say that a coisotropic submanifold C and an immersed lagrangian submanifold L in a symplectic manifold P form a reducible pair (C. we deﬁne LC as the ﬁber product of ιC and the quotient map pC : C → C/C ⊥ . Moreover. J2 of normal Jacobi ﬁelds along p. then we obtain an induced immersion ιC from the reduced space LC of L ×P C into C/C ⊥ . We leave it to the reader to check that any reducible pair (C. the constant energy hypersurface C = kM (E) (see Section 3. In fact. it is easy to check that for any pair J1 . the induced map ιC : LC → C/C ⊥ is a smooth immersion. Theorem 5. Hausdorﬀ manifold.2 shows that this immersion is lagrangian. J2 ) − g(J1 .2) is reducible. Evidently. L) is a reducible pair in a symplectic manifold P .Since F ∩ G has constant rank. L) provided that C is reducible. Example 5. and a pointwise application of Lemma 5. Finally. 2 If the quotient of L ×P C by the ﬁbers of the map pC ◦ rC is a smooth. The tangent space to CM at a point p identiﬁes naturally with the space of Jacobi ﬁelds normal to the geodesic represented by p. and the reduced manifold CM = C/C ⊥ is the space of oriented geodesics on M . then for each E > 0. we will say that the immersed submanifold N and the map : M → V intersect cleanly. respectively.
the annihilator CN is a coisotropic submanifold of T ∗ M if and only if the distribution F is integrable. we consider the sphere S n equipped with its standard metric. and therefore CS n is represented by the grassmannian Gr+ (2. From Theorem 5. then we can consider its leaves as submanifolds of M via the inclusion N → M . From Theorem 5. ωM ). and moreover. N ⊥ ∩ C is a sphere bundle over N transverse to the characteristic foliation of C. TZ CN = T N ⊕ F ⊥ . it follows that the real projective space RP n is a lagrangian submanifold of CP n . To give a concrete case of this example. the reduction of which is complex projective space CP n . i. but we wish to consider various properties of CN when it is viewed as a smooth submanifold of the symplectic manifold (T ∗ M.For further details. Now let Z denote the intersection of CN with the zero section of T ∗ M . that is CN = {(x.. Since the symplectic form on Cn+1 is invariant under unitary transformations. this implies that CN is coisotropic. The unit sphere S 2n+1 is a coisotropic submanifold of Cn+1 . Fx ⊂ ker(p)}. We begin with the following lemma.6.e. Then the natural projection T ∗ M → M maps Z diﬀeomorphically onto N . The conormal bundles of the leaves deﬁne a foliation of CN by lagrangian submanifolds. ∗ By deﬁnition.14 Note that the standard symplectic structure on R2n+2 is equivalent to the imaginary part of the standard hermitian metric on Cn+1 .12. p) ∈ T ∗ M : x ∈ N. 70 .12 it therefore follows that the space of geodesics in M normal to N at some point comprises an immersed lagrangian submanifold of CM . The conormal bundle N ⊥ to a smooth submanifold N ⊂ M is a lagrangian submanifold of T ∗ M which intersects C cleanly. Proof. a subbundle of T N ⊂ TN M . Example 5. The maximal real subspace Rn+1 ⊂ Cn+1 is a lagrangian submanifold of Cn+1 whose intersection with S 2n+1 is clean and coincides with the real unit sphere S n .15 In the notation above. Each oriented geodesic in S n identiﬁes naturally with an oriented 2dimensional subspace of Rn+1 . and let N ⊂ M be a smooth submanifold equipped with a tangent distribution F. CN is a subbundle of TN M . By CN we denote the annihilator of F. by Lemma 3. see [12]. the resulting symplectic form on CP n is invariant under transformations induced by unitary transformations. Lemma 5. n + 1). Conormal submanifolds A particular example of symplectic reduction which we will use in the next chapter begins with the following setup. More precisely. Suppose that M is a smooth manifold. If F is integrable.
Moreover. the proof of Lemma 5. π −1 pC In other words. Since F ⊥ is a lagrangian submanifold of T ∗ N .16 shows that each such leaf projects diﬀeomorphically onto F . and it is easy to check that the level sets of this map are the ﬁbers of the projection CN → CN /C ⊥ .15 can be reformulated as an application of Theorem 5. From the natural lagrangian splitting of T (T ∗ M ) along the zero section of T ∗ M . by deﬁnition.17 Suppose that F is a foliation of a manifold N . i. Proof. Since. and thus the intersections of Z with the isotropic leaves of CN are integral manifolds of the distribution F.e. and so the bundle F ⊥ is equipped in this way with a ﬂat Ehresmann connection (see Ehresmann [22]) known as the Bott connection associated to the foliation F. and let CB ⊂ T ∗ B be the conormal bundle to the ﬁbers of the submersion pB : B → V . To prove that the preceding diﬀeomorphism is symplectic. it follows that the annihilator CN ⊂ T ∗ N of F is a coisotropic submanifold which.15. ∗ ∗ 2 Example 5.18 As a special case of Example 5. This gives a welldeﬁned surjective submersion f : CN → T ∗ (N/F). 71 . p) deﬁnes an element of T ∗ (N/F). then the reduced space CN /C ⊥ is canonically symplectomorphic to T ∗ (N/F). each (x. If L ⊂ T ∗ B is the lagrangian submanifold given by the image of the diﬀerential dφ. it follows that (TZ CN )⊥ = F ⊕ (T N )⊥ . p) ∈ CN contains the subspace Fx of Tx N in its kernel. Hausdorﬀ manifold. Lemma 3. Theorem 4. then the nondegeneracy condition on φ implies that L intersects CB transversally along a submanifold I which maps diﬀeomorphically onto the ﬁber critical set Σφ under the natural projection π : T ∗ B → B. By Lemma 5.. contains the conormal bundle F ⊥ to any leaf F of F. The lagrangian immersion λφ : Σφ → T ∗ M equals the composition B Σφ → L ∩ CB → CV /C ⊥ . This completes the proof.17. suppose that (B. 2 Lemma 5. it suﬃces to note that by the commutative diagram f CN − − T ∗ (N/F) −→ N −− − → N/F the Liouville forms of T M and T (N/F) are related by αM CN = f ∗ αN/F . Thus we get a diﬀeomorphism CN /C ⊥ → T ∗ (N/F). when φ is nondegenerate.12 when the lagrangian submanifold im(dφ) is transverse to E. φ) is a Morse family over a manifold M .16 If the leaf space N/F is a smooth. Example 5. by deﬁnition.where F ⊥ is the subbundle of VZ M which annihilates F. it follows that (x. pB .6 implies that F ⊥ is foliated by isotropic leaves of CN .
the restriction of Q to E equals the form induced by the ﬁberhessian Hφx of φ at x. W ∈ LL. Reduction and the Maslov bundle Using linear reduction we can give a useful alternative description of the Maslov bundle of an immersed lagrangian submanifold L in T ∗ M . W ) is deﬁned as the integer σ(L. L . it is easy to check that the tangent space to the conormal submanifold CV deﬁned in Example 5. W. so that p = λφ (x) lies in the zero section ZM of T ∗ M .20 If L.6 implies that the Morse index ιM (x. L are lagrangian subspaces of V and W.5. the cross index of the quadruple (L.L . Denoting by x the point in the zero section of T ∗ B lying over x. 2 72 . W . L of a symplectic vector space V . Tp ZM ). L . Let let E ⊂ Tx B be the kernel of the linearized projection pB∗ : T B → T M . W . so that the last formula in Example 5. L .6 and 5. φ) is a Morse family which generates a lagrangian submanifold in L ⊂ T ∗ M and x ∈ B is a (nondegenerate) critical point of φ.We remark that since CN is a bundle over N . W ) + σ(L. V. φ) = ind Hφx + ind(Vp M.L . Immediately from this deﬁnition and Example 5. By deﬁnition. On the other hand. L . W ) + σ(L. we obtain the following “cocycle” property of the cross index. then σ(L. For W. Now suppose that Q is the nondegenerate quadratic form on Tx B deﬁned by the hessian of φ. and so ind(QE ) = ind Hφx . φ) of the critical point x satisﬁes ιM (x.19 An important special case of Examples 5. L . pV .18 at x equals the coisotropic subspace C = Tx B ⊕ E ⊥ .L the subset of the lagrangian grassmannian L(V ) comprised of those lagrangian subspaces which are transverse to both L and L . W. W ) = ind(L. Lemma 5. W. Given two lagrangian subspaces L. L . W ) − ind(L. W . we recall that Tx (T ∗ B) ∗ admits the natural lagrangian splitting Tx B ⊕ Tx B. W ). W ) = 0. we denote by LL. Example 5.18 arises when (B. W ∈ LL. the intersection T CN ∩ V M is a (constant rank) subbundle of T CN . L . Tp L. A check of the deﬁnitions shows that this subbundle is mapped onto V (N/F) under the quotient map CN → T ∗ (N/F) described above.
then there exists a function S : V → R such that Y = im dS(pV (x)).λ (E).21 In the notation above. 73 . W ) for all W. ˜ where p = p − dS(pV (x)) and Lφ−S is the lagrangian submanifold generated by φ − S.L (V ). Y ).λx (Ex ). A special case of this setup occurs when L is an immersed lagrangian submanifold of ∗ T M so that E = ι∗ T (T ∗ M ) is a symplectic vector bundle over L. we deﬁne a map L × M(L. the subspace Vp M to the vertical subspace Vp M at p. W. and Tp ZM to the tangent space Y of the image of dS at p. φ − S) at x is given by ˜ ιM (x. Y ). pB . Using these remarks. Thus.L (V ) the space of functions f : LL.L . W ∈ LL. Now if Y is any lagrangian subspace of Tp (T ∗ M ) transverse to Vp M . and set p = λφ (x). λ ).5 imply ιM (x. Vp M .Vp M → Z is obtained by setting fφ (Y ) = ιM (x. Proof.ι is canonically isomorphic to Mλ. From the preceding paragraph and the deﬁnition of Mλ. Tp Lφ−S . φ) be a Morse family over a manifold M . Tp ZM ). ι) × Z → Mλ. λ then correspond to smooth sections of L(E). Theorem 5. Let (B.λ (E). S) = ind Hφx + ind(Tp Lφ . we recall that if S : M → R is a smooth function such that the image of dS intersects λφ (Σφ ) transversely at p.λ (E). From the discussion above. where E is a symplectic vector bundle and λ. the preceding equation and Example 5.ι → Mλ. φ − S) = ind Hφx + ind(Vp M. n) to the function (p. ˜ Fiberwise translation by dS does not change the index on the right. it follows that this function satisﬁes fφ (Y ) − fφ (Y ) = σ(Tp Lφ . it follows that this map passes to an isomorphism of principal Z bundles ML.L → Z such that f (W ) − f (W ) = σ(L. L . Natural lagrangian subbundles of E are then given by λ = ι∗ T L and λ = ι∗ V M . From Example 4. Now we may assign a principal Z bundle to a triple (E. then x is a ˜ ˜ nondegenerate critical point of φ − S. it follows that Z acts simply and transitively by addition on FL. Since any such function is determined up to an additive constant by its value at a single point of LL. Vp M . ﬁx a point x in the ﬁber critical set Σφ . λ. The lagrangian subbundles λ. φ). pB . λ are lagrangian subbundles. φ − SY ).λ (E) by sending the triple (p. (B. From Example 5. and we denote by Mλ. Associated to E is its lagrangian grassmannian bundle L(E) whose ﬁber over x ∈ M is simply L(Ex ). where S = S ◦ pB .32. Y. fφ ). the Maslov bundle ML.19 it follows that ˜ the Morse index ιM (x. and moreover it maps Tp Lφ−S to Tp Lφ .λ (E) the principal Z bundle whose ﬁber over x ∈ M equals Fλx .We denote by FL.L . A function fφ : LTp Lφ .
λ∗ ) comprised of selfadjoint maps. 2 Lemma 5. Proof. then there exists a natural linear symplectomorphism of reduced spaces (LC + LC )/(LC + LC )⊥ → (L + L )/(L + L )⊥ which maps LC /(LC + LC )⊥ onto L/(L + L )⊥ and LC /(LC + LC )⊥ onto L /(L + L )⊥ . Thus. then LC is transverse to LC . we can deﬁne the isomorphism Mλ. then Mλ. Any choice of a positivedeﬁnite quadratic form on λ is induced by a section T of Hom(λ. If L.24 Suppose that E is a symplectic vector bundle over Y . then Mλ.λ (E) → Mλη . λ and a coisotropic subbundle η. WC ).L . if W ∈ LL. The starting point of the proof is the following special case of the theorem. First consider a symplectic vector space V with lagrangian subspaces L.λη (η/η ⊥ ). then Mλ. λ are ﬁberwise transverse.λη (η/η ⊥ ). L and a coisotropic subspace C.λ (E) is canonically isomorphic to Mλη . where fη (Wηx ) = f (W ) for each W ∈ Lλx . λ ⊂ η.23 Suppose that E is a symplectic vector bundle with lagrangian subbundles λ. Theorem 5. If λ. f ) ∈ Mλ. LC . If λ ∩ η and λ ∩ η have constant rank. λ are lagrangian subbundles such that λ ∩ λ has constant rank. If λ.2 We refer to [3] and [31] for a proof that the Maslov bundle described above is equivalent to the Maslov bundle constructed using index functions in Chapter 4.λx . λ and a coisotropic subbundle η. W ) = ind(LC . L ⊂ C.λη (η/η ⊥ ) ﬁberwise by sending (x.λ (E) under symplectic reduction. Proof. For the remainder of this section. Moreover. LC and WC /(LC + LC )⊥ maps onto W/(L + L )⊥ under the symplectomorphism above.22 Suppose that E is a symplectic vector bundle over M with lagrangian subbundles λ. If λ.λ (E) is canonically isomorphic to Y × Z. fη ). L . The proofs are rather technical and can be passed over in a ﬁrst reading. ind(L.λ (E) to (x. Lemma 5. 74 . our goal is to establish the following property of the bundle Mλ. then there exists a vector bundle symplectomorphism E → λ ⊕ λ∗ which maps λ onto λ ⊕ 0 and λ onto 0 ⊕ λ∗ .λ (E) is canonically isomorphic to Mλη . Using this remark.
Similarly. Thus.λη (E) . ˜ ˜ ˜˜ From the assumption that λ∩η has constant rank.λ (E) × Mλη .λη (η/η ⊥ ). λη are transverse lagrangian subbundles of η/η ⊥ . ˜˜ Combined with the canonical isomorphisms above.λ (E). Since the lagrangian subbundles λ . E are trivial symplectic vector bundles over M with lagrangian subbundles λ ⊂ E and λ ⊂ E .λη (E).λ (E) ˜ 75 .λ (E) is canonically ˜ η ˜η isomorphic to Mλη . completing the proof. A natural section s(x) = (x. λ are lagrangian subbundles of a symplectic vector bundle E. and thus Mλ.L (E ) are canonically isomorphic. and the assertion follows. λ and clearly satisﬁes ind(λx .L ⊕L (E ⊕ E) and MT (L). In general. it follows that λ∩λη has constant rank. then Mλ.λ (E) is canonically isomorphic to the principalbundle product (see Appendix D for the deﬁnition) Mλ. Mλ.24. respectively. the bundle Mλ.22. λ .22 implies that MLΓ(T ) . ˜ 2 Example 5.λ (E) is canonically isomorphic to Mλη . our hypotheses imply that η = λ + λ is a coisotropic subbundle of E which contains the lagrangian subbundles λ and λ . E ) of symplectic vector bundle maps. λx . By Lemma 5.20 shows that if λ. and let T be a section of the bundle Homω (E.The lagrangian subbundle λ of E which maps to the subbundle of λ ⊕ λ∗ deﬁned by the graph of T is then transverse to λ.λη (E) × Mλη . we obtain the following canonical isomorphisms: Mλ. It is easy to check that the lagrangian subbundles λΓ(T ) and λ ⊕λ intersect the coisotropic subbundle E ⊕ λ along constant rank subbundles.λη (η/η ⊥ ).23.λ (E) is therefore given by f (λx ) = 0. Theorem 5.λ (E) × Mλ . Mλ. λx ) = 0 for all x ∈ Y . the latter bundle has a canonical trivialization.λη (E) is in turn canonically isomorphic ˜ to Mλη .λη (E) × Mλ. The graph of the section T deﬁnes a lagrangian subbundle λΓ(T ) of E ⊕ E. Lemma 5. this shows that Mλ. f ) of Mλ. Applying this remark.λ (E) ˜ Mλ. Since λη .λη (E) is canonically trivial.25 Suppose that E. λ of E are contained in the ˜ coisotropic subbundle η. while the restriction of T to the subbundle λ deﬁnes a lagrangian subbundle T (λ) of E . whereas their reductions by E ⊕ λ are given by T (λ) and λ .23 implies that Mλη .λη (E) is canonically trivial by Lemma 5. An application of the cocycle identity of Lemma 5. 2 Proof of Theorem 5.
From Lemma 3. we deﬁne ∗ ∗ their product as the symplectic manifold (P × Q. From Section 3. zf ). zF ) and (T ∗ M. ﬁnitedimensional symplectic manifolds. as described in Example 5. we ∗ have the lagrangian subbundles λ = s∗ V M of E and λ = zF V N of E .ιF under the natural isomorphism H 1 (M . 5.2 The symplectic category To systematize the geometric aspects of quantization in arbitrary symplectic manifolds. Moreover. As objects of S we take the class of all smooth. R) H 1 (T ∗ M . The morphism set Hom(P. deﬁned as the composition of the graph ΓF : T ∗ M → T ∗ N × T ∗ M of F with the Schwartz transform SM. we can therefore deﬁne the adjoint of a canonical relation L ⊂ Hom(P.14. To see that a similar relation holds for the Maslov classes of (M. Thus. Evidently. ιf ) if and only if [β] is integral. the phase class of the induced lagrangian ˇ embedding (M. Q) is then deﬁned to consist of all canonical relations in Q × P . Since immersed lagrangian submanifolds of the product Q × P coincide with those of its dual.21 shows that s∗ MT ∗ M. π2 denote the cartesian projections. we now introduce the symplectic category S.25 that the Maslov classes are also related by pullback: s∗ µT ∗ M. ∗ consider the symplectic vector bundles E = s∗ T (T ∗ M ) and E = zF T (T ∗ N ) over M . Q) as the element L∗ ∈ Hom(Q. E ) deﬁned by the restriction of F∗ to TZ (T ∗ M ).ιF = µM. Given two objects (P. Theorem 5. Observe that these deﬁnitions imply that the Liouville class λM. and thus ∈ R+ is admissible for the lagrangian embeddings (M.27 We can apply the conclusion of the preceding example to ﬁberpreserving symplectomorphisms f : T ∗ M → T ∗ N . and the lagrangian 0 subbundles λΓ(T ) ⊂ E ⊕ E and T (λ) ⊂ E . The symplectic dual of an object (P. Composition of morphisms is unfortunately not deﬁned for all L1 ∈ Hom(P. ιF ).ιF is canonically isomorphic to MλΓT . along 0 with the section T of Homω (E. ω) is the object (P. R). P ) represented by the same equivalence class L of immersions into the product P × Q. we recall that any such f is equal to a ﬁberwise translation of T ∗ M by a closed 1form β on M composed with the cotangent lift of some diﬀeomorphism N → M . and so S is therefore not a true category.26 Any symplectomorphism F : T ∗ M → T ∗ N induces a lagrangian embedding ιF : T ∗ M → T ∗ (M × N ).N : T ∗ N × T ∗ M → T ∗ (M × N ). we recall that a smooth diﬀeomorphism from a symplectic manifold P to a symplectic manfold Q is a symplectomorphism if and only if its graph is a lagrangian submanifold of Q × P . Nevertheless.zF .25. ω) (Q.λ (E ). where π1 . zf ) equals [β] ∈ H 1 (M . −ω). ω ) of S.zF is 0 canonically isomorphic to MT (λ). π1 ω + π2 ω ). More generally. (T ∗ M. R).2. A second lagrangian embedding zF : M → T ∗ N is deﬁned by the composition of F with the zero section s0 : M → T ∗ M . we can describe 76 .Example 5. zF ) identiﬁes with λT ∗ M. an immersed lagrangian submanifold of Q × P is called a canonical relation from P to Q. R) induced by the projection T ∗ M → M . it follows from Example 5. Q) and L2 ∈ Hom(Q.zF of the embedding (M.λ ⊕λ (E ⊕ E) and that MM. 0 Example 5.
8. this deﬁnition generalizes tangent lifts of diﬀeomorphisms. Canonical lifts of relations The full subcategory of S consisting of cotangent bundles possesses some special properties due to the Schwartz transform SM. Under the Schwartz transform. In view of Example 3. which enables us to identify canonical relations in Hom(T ∗ M.a suﬃcient condition for the composability of two canonical relations as follows. L3 . Applying Theorem 5. R) → Hom(P. N is a smooth submanifold S of the product M × N . 2 Associativity of compositions holds in the symplectic category in the sense that for canonical relations L1 . then the composition Hom(P. T ∗M ) called the cotangent lift of S. in the usual sense. R). where ∆Q denotes the diagonal in Q × Q. Q) × Hom(Q.12. the conormal bundle of S identiﬁes with a canonical relation LS ∈ Hom(T ∗N. the product R × ∆Q × P is a reducible coisotropic submanifold of R × Q × Q × P .29 A smooth relation between two manifolds M. L2 ◦ L1 ∈ Hom(P. From these deﬁnitions we see that a canonical relation is an isomorphism if and only if it is the graph of a symplectomorphism. Equivalently. one deﬁnes an epimorphism in the symplectic category as a canonical relation L for which L ◦ L∗ = idP . Thus. we may identify the set Hom(P. then L2 ◦ L1 is an immersed lagrangian submanifold of R × P . the “elements” of P are its immersed lagrangian submanifolds. Q) of morphisms with the “elements” of Q × P for any P. Among the members of S.27. not as a set. Example 5. 8 77 . there is a minimal object Z. Dually. By Example 5.N : T ∗ N × T ∗ M → T ∗ (M × N ). L is a monomorphism if L∗ ◦ L = idP . L2 × L1 ) is a reducible pair. but rather as the object Q × P in the symplectic category. Note in particular Another point of view is to deﬁne Hom(P. and we will call L2 × L1 clean if (R × ∆Q × P . i. Q) is said to be a monomorphism if the projection of L onto P is surjective and the projection of L onto Q is injective. the zerodimensional symplectic manifold consisting of a single point ∗ equipped with the null symplectic structure. Q ∈ S. L2 .e. the equation L1 ◦ (L2 ◦ L3 ) = (L1 ◦ L2 ) ◦ L3 is valid provided that both sides are deﬁned. T ∗ N ) with immersed lagrangian submanifolds in T ∗ (M × N ). we then obtain Proposition 5. Morphisms from Z to any other object P ∈ S identify naturally with immersed lagrangian submanifolds of P . Q). The product L2 × L1 of the canonical relations L1 and L2 is a welldeﬁned immersed lagrangian submanifold of R × Q × Q × P .8 A canonical relation L ∈ Hom(P.28 If L2 × L1 is clean. R) is the canonical relation which is the product of three diagonals. and in particular.
that although a smooth map f : M → N does not in general give rise to a welldeﬁned transformation T ∗N → T ∗M , its graph in M × N nevertheless generates a canonical relation Lf ∈ Hom(T ∗N, T ∗M ). Example 5.30 The diagonal embedding ∆ : M → M × M induces a canonical relation L∆ ∈ Hom(T ∗ (M × M ), T ∗ M ) which in local coordinates assumes the form
∗ L∆ = {((x, α), (x, x, β, γ)) : x ∈ M and α, β, γ ∈ Tx M satisfy α = β + γ}.
In other words, L∆ is the graph of addition in the cotangent bundle. If L1 , L2 ∈ Hom(∗, T ∗ M ) and L2 × L1 is identiﬁed with an element of Hom(∗, T ∗ (M × M )) by the usual symplectomorphism T ∗ M × T ∗ M → T ∗ (M × M ), then the sum L1 + L2 ∈ Hom(∗, T ∗ M ) is deﬁned as L1 + L2 = L∆ ◦ (L2 × L1 ). Note that if L1 , L2 coincide with the images of closed 1forms ϕ1 , ϕ2 on M , then L1 + L2 equals the image of ϕ1 + ϕ2 . The dual of the isomorphism T ∗ M × T ∗ M → T ∗ (M × M ) identiﬁes L∆ with an element L∆ ∈ Hom(T ∗ M, Hom(T ∗ M, T ∗ M )). This canonical relation satisﬁes L1 +L2 = L∆ (L1 )(L2 ), and if L is the image of a closed 1form ϕ on M , then L∆ (L) = fϕ , where fϕ is the ﬁberwise translation mapping introduced in Section 3.2.
Example 5.31 (The Legendre transform) As a particular example of this situation, let V be a smooth manifold, and consider the ﬁber product TV ×V T ∗V along with its natural inclusion ι : TV ×V T ∗V → TV × T ∗V and “evaluation” function ev : TV ×V T ∗V → R given by ev((x, v), (x, p)) = p, v . If L ⊂ T ∗(TV ×V T ∗V ) is the lagrangian submanifold given by the image of the diﬀerential d(ev), then the pushforward Lleg = ι∗ L deﬁnes an isomorphism in Hom(T ∗(TM ), T ∗(T ∗M )) given in local coordinates by ((x, v), (ξ, η)) → ((x, η), (ξ, −v)). As noted in [57], this canonical relation can be viewed as a geometric representative of the Legendre transform in the following sense: If L : TM → R is a hyperregular Lagrangian function, in the sense that its ﬁberderivative deﬁnes a diﬀeomorphism TM → T ∗M , then the composition of Lleg with the image of dL equals the image of dH, where H : T ∗M → R is the classical Legendre transform of L.
78
Morphisms associated with coisotropic submanifolds If C ⊂ P is a reducible coisotropic submanifold, then the reduction relation RC ∈ Hom(P, C/C ⊥ ) is deﬁned as the composition of the quotient map C → C/C ⊥ and the inclusion C/C ⊥ → P . Somewhat more concretely, the relation RC is the subset {([x], x) : x ∈ C} of C/C ⊥ × P , where [x] is the leaf of the characteristic foliation passing through x ∈ C. ∗ Evidently RC is an epimorphism, and so RC ◦ RC = idP . On the other hand, RC is not a monomorphism unless C = P , and we deﬁne the projection relation KC ∈ Hom(P, P ) as ∗ the composition RC ◦ RC , i.e. KC = {(x, y) : x, y ∈ C, [x] = [y]}.
∗ By associativity, we have KC = KC ◦ KC , and KC = KC . Thus, KC is like an orthogonal projection operator. These relations give a simple interpretation of Theorem 5.12 in the symplectic category. If L is an immersed lagrangian submanifold of P , i.e. L ∈ Hom(Z, P ), then L is composable with both RC and KC , provided that it intersects C cleanly. In this case, we have
LC = R C ◦ L
LC = KC ◦ L.
In particular, KC ﬁxes any lagrangian submanifold of C. Example 5.32 Suppose that F is a foliation on a manifold B whose leaf space BF is smooth. In the notation of Example 5.17, we then ﬁnd that if T ∗ (B/F) is canonically identiﬁed with the reduced space E/C ⊥ , then the reduction relation RE is the Schwartz transform of the conormal bundle to the graph of the projection B → B/F.
5.3
Symplectic manifolds and mechanics
In general, an arbitrary symplectic manifold has no associated “conﬁguration space,” and therefore the classical and quantum mechanical viewpoints must be adapted to this new context based on the available structure. The classical picture The central objects in the classical picture of mechanics on an arbitrary symplectic manifold (P, ω) are the semiclassical states, represented as before by lagrangian submanifolds of P equipped with halfdensities, and the vector space of observables C ∞ (P ). With respect to pointwise multiplication, C ∞ (P ) forms a commutative associative algebra. Additionally, the symplectic form on P induces a Lie algebra structure on C ∞ (P ) given by the Poisson bracket {f, g} = Xg · f, 79
where Xg = ω −1 (dg) denotes the hamiltonian vector ﬁeld associated to g. These structures ˜ satisfy the compatibility condition {f h, g} = f {h, g} + {f, g} h, and are referred to collectively by calling C ∞ (P ) the Poisson algebra of P . The classical system evolves along the trajectories of the vector ﬁeld XH associated to the choice of a hamiltonian H : P → R. If f is any observable, then Hamilton’s equations assume the form f˙ = {H, f }. Note that in local Darboux coordinates, the Poisson bracket is given by {f, g} =
j
∂f ∂g ∂f ∂g · − · ∂qj ∂pj ∂pj ∂qj
.
Setting f = qj or f = pj in the Poisson bracket form of Hamilton’s equations above yields their familiar form, as in Section 3.3. Two functions f, g ∈ C ∞ (P ) are said to be in involution if {f, g} = 0, in which case the hamiltonian ﬂows of f and g commute. An observable in involution with the hamiltonian H is called a ﬁrst integral or constant of the motion of the system. A collection fi of functions in involution on P is said to be complete if the vanishing of {fi , g} for all i implies that g is a function of the form g(x) = h(f1 (x), · · · , fn (x)). The quantum mechanical picture Quantum mechanical observables should be the vector space A(HP ) of selfadjoint linear operators on some complex Hilbert space HP associated to P . The structure of a commutative, nonassociative algebra is deﬁned on A by Jordan multiplication: 1 A ◦ B = (AB + BA), 2 representing the quantum analog of pointwise multiplication in the Poisson algebra of P . Similarly, the dependent commutator [A, B] = i (AB − BA)
deﬁnes a Lie algebra structure on A analogous to the Poisson bracket on C ∞ (P ). Quantum mechanical states are vectors in HP . The timeevolution of the quantum system ˆ is determined by a choice of energy operator H, which acts on states via the Schr¨dinger o equation: i ˆ ˙ ψ = Hψ. A collection {Aj } of quantum observables is said to be complete if any operator B which commutes with each Aj is a multiple of the identity. This condition is equivalent to the irreducibility of {Aj }, in other words, no nontrivial subspace of HP is invariant with respect to each j. 80
To make the basic quantization problem more tractable. 81 . In these abstract terms. R). 2. As was eventually proven by Groenwald and Van Hove (see [1] for a proof). HQ ) HQ ⊗ H∗ in a way which commutes P with compositions.e. However this is accomplished. i. · · · . A general formulation of the quantization problem is then to deﬁne a “functor” from the symplectic category to the category of (hermitian) linear spaces.33 A linear map ρ : C ∞ (P ) → S(HP ) is called a quantization provided that it satisﬁes 1. ρ(fn ) form a complete commuting set. TL◦L = TL ◦ TL ˜ ˜ ˜ for each L ∈ Hom(Q. we wish to identify a ∗algebra A of operators which give the quantum analog of the “system” at hand. the basic goal of quantization remains the same as in the case of cotangent bundles: starting with a symplectic manifold P . each canonical relation L ∈ Hom(P. which should be represented by unitary operators on quantum Hilbert spaces. Based on an idea of Weyl and von Neumann. For some complete set of functions f1 . ρ({f. ρ(1) = identity. and HP ×Q is canonically isomorphic to a (completed) tensor product HP ⊗HQ .. the operators ρ(f1 ). regardless of how the algebra A is identiﬁed. our classical and quantum correspondences can be expressed as follows.Quantization Although neither an underlying conﬁguration manifold nor its intrinsic Hilbert space is generally available in the context of arbitrary symplectic manifolds. we ﬁrst enlarge the class of classical objects to be quantized. This means that to each symplectic manifold P we should try to assign a Hilbert space HP in such a way that HP is dual to HP . The basic classical objects are then symplectomorphisms. fn in involution. Early approaches to this problem were based on the principle that. but then relax the criteria by which quantum and classical objects are to correspond. the correspondence between classical and quantum observables should be described by a linear map from the Poisson algebra C ∞ (P ) to A which satisﬁes the following criteria known as the Dirac axioms Deﬁnition 5. we ﬁrst replace the classical observables by the groups of which they are the inﬁnitesimal generators. · · · . a quantization of all classical observables in this sense does not exist in general. Q) must then be assigned to a linear operator TL ∈ Hom(HP . g}) = [ρ(f ). ρ(g)] 3.
The best we can hope for is a functorial relation. Second. ω) Q×P Q×P point ∗ lagrangian submanifold L Quantum version hermitian vector space HP HQ ⊗HP Hom(HP . it may require more data (such as a symbol) to determine an element of HP in a consistent way. from the classical to the quantum category. by powers of or by degree of smoothness). rather than a mapping. two relevant observations are apparent from our earlier study of WKB quantization. 82 . it is too much to require that this condition be satisﬁed exactly.g. By “correctness” we mean that composition of canonical relations should correspond to composition of operators. in addition to just a lagrangian submanifold L ⊂ P . First. the Hilbert space HP may carry some sort of ﬁltration (e.Object basic spaces Classical version symplectic manifold (P. HQ ) C element of HP symmetric operators on HP state space of observables Poisson algebra C ∞ (P ) At this stage. and the quantization may be “correct” only to within a certain degree of accuracy as measured by the ﬁltration. As we have already seen.
In this section. then there exists a natural linear map Ω1/2 L ⊗ Ω1/2 C → Ω1/2 LC . Our discussion begins by deﬁning a suitable notion of composition for such states. First note that if V is a symplectic vector space. A particular concrete case of this classicalquantum correspondence is given by the symbol calculus of Fourier integral operators. s) of the state (L. In this context. s) consisting of an immersed lagrangian submanifold L in T ∗ (M ×N ) equipped with a symbol s. s) is then regarded as the Schwartz kernel of the corresponding operator in Hom(HM . −v) and (x. then the exact sequence 0 → L ∩ C ⊥ → L ∩ C → LC → 0 gives rise to the isomorphism Λ1/2 LC ⊗ Λ1/2 (L ∩ C ⊥ ) Λ1/2 (L ∩ C). L) is a properly reducible pair in a symplectic manifold P . s) consisting of a quantizable lagrangian submanifold L ⊂ T ∗ M and a symbol s on L.6 Fourier Integral Operators For our ﬁrst examples of a quantization theory which gives a functorial relation between the classical (symplectic) category and the quantum category of hermitian vector spaces. Reduction of semiclassical states We will say that a reducible pair (C. we deﬁned a semiclassical state in a cotangent bundle T ∗ M as a pair (L. L) in a symplectic manifold P is properly reducible if the quotient of I = L ×P C by its characteristic foliation is a smooth. we return to WKB quantization in cotangent bundles. The linear maps v → (v. 83 . together with a lagrangian subspace L and a coisotropic subspace C.1 If (C. Lemma 6. semiclassical states are represented by pairs (L. Proof. The WKB quantization I (L.4. y) → x + y deﬁne a second exact sequence 0 → L ∩ C → L ⊕ C → L + C → 0. Hausdorﬀ manifold and the map I → LC is proper.1 Compositions of semiclassical states In Section 4. from which we get Λ1/2 LC ⊗ Λ1/2 (L ∩ C ⊥ ) ⊗ Λ1/2 (L + C) Λ1/2 L ⊗ Λ1/2 C. 6. HN ). we study certain natural transformations of semiclassical states.
there is a natural isomorphism ∗ rL ΦL. L) form a reducible pair. If L is an immersed lagrangian submanifold of T ∗ M such that (CN . we arrive at Λ1/2 LC ⊗ Λ(L ∩ C ⊥ ) Λ1/2 L ⊗ Λ1/2 C. Since L ∩ C ⊥ = (L + C)⊥ . L) in a symplectic manifold P . Hausdorﬀ manifold. By the preceding computations. From Section 5.Finally. the exact sequence 0 → (L + C)⊥ → V → (L + C)∗ → 0 combined with the halfdensity on V induced by the symplectic form deﬁnes a natural isomorphism Λ1/2 (L + C) Λ1/2 (L + C)⊥ . Since the quotient map I → LC is proper.1 we recall that the integrability of F implies that CN = {(x. Lemma 6. integration over its ﬁbers is welldeﬁned and gives the desired linear map Ω1/2 L ⊗ Ω1/2 C → Ω1/2 LC . there is a linear map Ω1/2 L ⊗ Ω1/2 C → Ω1/2 LC ⊗ Ω(FI ). Fx ⊂ ker(p)} is a coisotropic submanifold of T ∗ M whose reduced space is the cotangent bundle T ∗ NF of the leaf space NF . then we denote by I the ﬁber product L ×T ∗ M CN of L and CN and consider the following commutative diagram L − − T ∗M −→ rL ι I −− −N → CN pC π LC − − T ∗ N −→ Our goal is describe how a symbol s on L naturally induces a symbol sC on LC . ω ˜ Now consider a properly reducible pair (C. 2 Let M be a smooth manifold and consider a submanifold N ⊂ M equipped with a foliation F such that the leaf space NF is a smooth. rC 84 . p) ∈ T ∗ M : x ∈ N. . → π ∗ ΦLC .2 In the notation of the diagram above.
Composition of semiclassical states If L1 . Since parallel sections of π ∗ ΦLC . ) → (parallel sections of ΦLC . λ of E are then induced by the immersed lagrangian submanifold L and the vertical subbundle V M of T (T ∗ M ). s2 ) in T ∗ M.σ .3 The product of semiclassical states (L1 . pulls back to a parallel section of rL ΦLC . (L2 .N : T ∗ N ×T ∗ M → T ∗ (M ×N ). and thus. it follows easily that the phase bundle ΦL2 ×L1 . is canonically isomorphic to the external tensor product ΦL2 . . the pullback to I of the prequantum line bundles over T ∗ M and T ∗ N are canonically isomorphic. for each halfdensity σ on C. under the isomorphism of Lemma 6. By tensoring with the map of density spaces given above. identiﬁes with a parallel section of π ∗ ΦLC . T ∗ N is the semiclassical state (L2 × L1 . we arrive at the desired isomorphism of phase bundles.λ (E). and one must check that the pullback of the Maslov bundle of LC to I is canonically isomorphic to ∗ MλC . 2 ∗ A parallel section s of ΦL.Proof.16 it follows easily that the pullback to I of the Liouville forms on T ∗ M and T ∗ N coincide. we obtain. Now the tangent bundle of CN induces a coisotropic subbundle C of E. . s1 ). a natural map of symbol spaces SL → SLC . the pullback of the Maslov bundle of L to I is canonically isomorphic to the bundle Mλ. s2 ) → s2 s∗ . L2 are immersed lagrangian submanifolds of T ∗ M.λC (C/C ⊥ ). From Theorem 5.32. ). s2 s∗ ) in T ∗ (M × N ). From the proof of Lemma 5. Similarly. T ∗ N . . Lagrangian subbundles λ. and thus there is a natural linear map of symbol spaces SL2 ⊗ SL1 → SL2 ×L1 which we denote by (s1 . By the fundamental properties of the Schwartz transform described in Proposition 3. identify naturally with parallel sections of ΦLC . By tensoring this isomorphism with the isomorphism of prequantum bundles described in the preceding paragraph. we obtain a map (parallel sections of ΦL.2.22 we therefore obtain a canonical isomorphism of rL ML with π ∗ MLC . By deﬁnition. let E be the symplectic vector bundle over I given by the pullback of T (T ∗ M ). which. 1 Deﬁnition 6. From the discussion in Appendix A L1 we also have a linear isomorphism of density bundles Λ1/2 (L2 × L1 ) → Λ1/2 L2 Λ1/2 L1 . 1 2 85 . then their product L2 ×L1 gives a welldeﬁned immersed lagrangian submanifold of T ∗ (M × N ) via the Schwartz transform SM. We will denote the image of a symbol s on L under this map by sC. . Φ−1.
the Schwartz transform of L2 ×L1 reduces by CV. s2 ) in T ∗ (N × V ) are composable if the conormal submanifold CV. T ∗ V × T ∗ N × T ∗ N × T ∗ M → T ∗ (M × N × N × V ). and the lower horizontal arrow equals the Schwartz transform SM. We will say that semiclassical states (L1 .N. we ﬁrst note that if M.V . Under the Schwartz transform. respectively. V are smooth manifolds. 2 86 . s ) in T ∗ (N × V ) deﬁne a semiclassical state in T ∗ (M × V ) when L. T ∗ V ) is deﬁned as the reduction of L2 × L1 by the coisotropic submanifold T ∗ V × ∆T ∗ N × T ∗ M of T ∗ V × T ∗ N × T ∗ N × T ∗ M . s1 ) in T ∗ (M × N ) is composable with a semiclassical state (L2 . L are composable as canonical relations in Hom(T ∗ M. T ∗ N .1. Thus.V ×SM. and T ∗ M . The left vertical arrow represents the reduction relation deﬁned by the coisotropic submanifold C = T ∗ V × ∆T ∗ N × T ∗ M . while the right vertical arrow is the reduction relation deﬁned by the image of C under the Schwartz transform. T ∗ N ) and Hom(T ∗ N. s1 ) in T ∗ (M × N ) and (L2 .V : T ∗ V × T ∗ M → T ∗ (M × V ) identiﬁes each canonical relation in Hom(T ∗ M. then the Schwartz transform SM. L ◦ L ∈ Hom(T ∗ M. From Lemma 6. m) ∈ V ×M .M and immersed lagrangian submanifold L2 × L1 form a properly reducible pair. s2 ◦ s1 ) in T ∗ (M × V ).M to yield the image of L2 ◦L1 under the Schwartz transform T ∗ V ×T ∗ M → T ∗ (V ×M ).M of T ∗ (V × N ×N ×M ) deﬁned by V ×∆N ×M ⊂ V ×N ×N ×M and its product foliation by subsets of the form v ×∆N ×m for (v. We begin with the following commutative diagram T ∗ V × T ∗ N × T ∗ N × T ∗ M − − T ∗ (M × N × N × V ) −→ T ∗V × T ∗M −− −→ T ∗ (M × V ) where the upper horizontal arrow denotes the composition SM ×N ◦(SN.M induced by the symplectic forms on T ∗ V.Turning to compositions.N. Deﬁnition 6.N ) of Schwartz tranforms. s) in T ∗ (M × N ) and (L . then their composition is deﬁned as the semiclassical state (L2 ◦ L1 . L2 . the submanifold T ∗ V × ∆T ∗ N × T ∗ M maps to the conormal submanifold CV. we obtain a natural linear map of symbol spaces SL2 ⊗ SL1 → SL2 ◦L1 given by the composition of the product map above and the reduction map SL2 ×L1 → SL2 ◦L1 deﬁned by the natural halfdensity on CV.N. Our goal is now to describe how semiclassical states (L. We denote the image of s2 s1 under this map by s2 ◦ s1 . T ∗ V ) with an immersed lagrangian submanifold in T ∗ (M × V ). T ∗ V ). s2 ) in T ∗ (N × V ). it follows that for any properly reducible pair L1 .N.4 If a semiclassical state (L1 .
We denote the space of distributional halfdensities on M 87 . 6. an analogous correspondence is furnished by the Schwartz kernel theorem. T ∗ V ) as a semiclassical state (L ◦ L. 1/2 Clinear functional on Ω0 M . h) with a semiclassical state (L. ⊗ Φ−1 . ˜) ∈ ˜ Hom(T ∗ N.5. A distributional halfdensity on M is then a continuous. since it is the cotangent bundle structure of the latter space which gives meaning to “symbols” (in the sense of Chapter 4) ˜ on L ◦ L.4). 1/2 Let M be a smooth manifold and let Ω0 M be equipped with the topology of C ∞ convergence on compact sets. For such . note that if the symplectomorphism F : T ∗ M → T ∗ N is the cotangent lift of a diﬀeomorphism M → N . (h ◦ ι) · s). ˜) in T ∗ (M ×N ). the semiclassical state (L. if β is a closed 1form on M and the symplectomorphism F : T ∗ M → T ∗ M equals ﬁberwise translation by β. If h ∈ C ∞ (LF . s) in T ∗ M is transformed by F into the semiclassical state (LF ◦ L. we ﬁnd by a com˜ putation that the isomorphism SL → SLF ◦L determined by the symbol ˜ = s ⊗ h is given by s ˜ ∗ s ⊗ a → s ⊗ (a · ι h). ˜ ◦ s) ∈ Hom(T ∗ M. If L is an immersed lagrangian submanifold of T ∗ M .5 Consider a semiclassical state (LF .4 provides that for each symbol s ∈ SL . then Deﬁnition 6. and so the symbol space SLF identiﬁes naturally with R+ × C ∞ (LF . the phase bundle of of LF admits a canonical parallel section.2 WKB quantization and compositions ˜ s To deﬁne the composition of semiclassical states (L. A computation then shows that the composition of a semiclassical state of the form (LF . C). ˜ ΦLF ◦L.Example 6. where a is any halfdensity on L and s is the unique parallel section of ΦLF ◦L. On the quantum level. T ∗ N ) and (L. identiﬁes with an oscillatory function of the form ceiS/ for c ∈ R and some S : M → T satisfying S ∗ dσ = β. T ∗ V ). Given a parallel section s and smooth complexvalued function h on LF . where LF is the Schwartz s transform of the graph of a symplectomorphism F : T ∗ M → T ∗ N . then for each > 0. C). We then note that the n natural halfdensity ωM 1/2 on LF ≈ T ∗ M enables us to identify the symbol space SLF with the product ΓLF ⊗C C ∞ (LF . then ∈ R+ is admissible for LF if and only if [β] is integral. s To make this correspondence more explicit. Example 6. eiS/ ⊗ h) with (L. we used the s ˜ Schwartz transform to identify the immersed lagrangian submanifold L ◦ L ∈ T ∗ V × T ∗ M ∗ with an immersed lagrangian submanifold in T (M × V ).6 As a special case of Example 6. C). let us choose a particular lagrangian immersion ι : L → T ∗ M representing L (see the discussion in Section 4. Similarly. then with this identiﬁcation. s) equals the semiclassical state (LF ◦ L. s) ∈ Hom(T ∗ M. ˜ ◦ s) in T ∗ N . ei(S◦πL )/ (h ◦ ι) · s). the composition of (LF . each parallel section of the phase bundle ΦLF . s) yields the semiclassical state (LF ◦ L. L. such that ι∗ s → s ⊗ s−1 under the canonical isomorphism ι∗ ΦLF . ι.
T ∗ N ) ↔ T ∗ (M × N ). Every K ∈ Ω−∞ (M ×N ) deﬁnes a linear map K : Ω0 M → 1/2 1/2 Ω−∞ N by (∗) above. N . each w ∈ Ω1/2 N deﬁnes an element w ∈ Ω−∞ N by the equation ˜ w. In all of the examples we will consider. 88 . we deﬁne dependent operators Dj = −i def 1/2 ∂ . In practical terms. v = K. there is precisely one distribution K such that (∗) is valid. and so we can consider K as 1/2 1/2 a continuous linear map Ω0 M → Ω0 N . the Schwartz kernel theorem gives the identiﬁcation Ω−∞ (M × N ) ↔ Hom(HM . the image of the map K will lie within the subspace 1/2 of Ω−∞ N represented by elements of Ω1/2 N in this way. which is continuous in the sense that K(φj ) → 0 in Ω−∞ N if φj → 0 1/2 in Ω0 M . Thus. A kernel K deﬁnes a linear map 1/2 1/2 K : Ω0 M → Ω−∞ N by duality via the equation K(u). HN ) diﬀerential operators With respect to linear coordinates {xj } on Rn . integrating the product K ⊗ u over the submanifold {y} × M of N × M . to every such linear map. and we equip this space with the weak∗ topology. a kernel is any element of Ω−∞ (M × N ). thereby giving a (densely deﬁned) operator HM → HN on the intrinsic Hilbert spaces of M. u ⊗ v 1/2 1/2 def 1/2 (∗) Schwartz kernel theorem . the value of the halfdensity Ku at each y ∈ N can be computed in these cases by. 2 Of course. in the same way that the Schwartz transform provides a natural correspondence Hom(T ∗ M. Conversely. Note that this order generally diﬀers from the order of a diﬀerential operator. If N is another smooth 1/2 manifold.by Ω−∞ M . roughly speaking. ∂xj An diﬀerential operator of asymptotic order9 k ∈ Z is then an asymptotic series of the form ∞ P = m=0 9 Pm m+k . v = ˜ N 1/2 v ⊗ w.
The Schwartz kernel for the operator P is given by the distribution family I (L∆ . y. By formally substituting ξj for Dj in each term of P . together with V = Rn × Rn and the cartesian projection pV : B → V deﬁne a Morse family (B. (Later we will drop this assumption to obtain a more symmetric calculus). dψ(x)) hm0 +k + O( m0 +k+1 ) (∗∗) To interpret this expression in the context of WKB quantization..e. y. An application of the principle of stationary phase therefore gives (P e−iψ/ u)(x) = e−iψ(x))/ u(x) · p (x. the function (y. we obtain an dependent function σP on T ∗ Rn . its principal symbol p is a welldeﬁned function on the cotangent bundle T ∗ M . Now. ξ) u(y) dy dξ + O( m0 +k+1 ) where p is of course independent of y. s). we obtain (P e−iψ/ u)(x) = (2π )−n ei( x−y.29) and arguing as above. While the symbol of P depends on the choice of local coordinates. Combined with the preceding equation. y. By replacing σP by p in the formula above. The principal symbol p of the operator P is deﬁned as the symbol of the ﬁrst nonvanishing term in its series expansion. sP ◦ s). The principal symbol p . composition with (L∆ . x. x. we obtain the following geometric version of (∗∗) above: 89 . By Example 6. this gives (P e−iψ/ u)(x) = I (L∆ ◦ L.ξ −ψ)/ σP (x. sP ) multiplies the symbol s by the values of the principal symbol p on L Σφ . ξ − ψ has a nondegenerate critical point when y = x and ξ = dψ(x). where L is the projectable lagrangian submanifold of T ∗ Rn deﬁned by im(dψ). ξ = 0} induces a welldeﬁned symbol sP = p (x. pV . ξ) → x − y. sP ). ξ) = x − y. e−iψ/ u = I (L. ξ) on L∆ . For ﬁxed x. ξ on B = Rn × Rn × (Rn )∗ . known as the symbol of P . Using a global generating function for the conormal bundle of ∆ ⊂ M × M (see Example 4. sP ◦ s) · m0 +k + O( m0 +k+1 ).ξ −ψ)/ p (x. ξ) deﬁnes an amplitude a = p · dx dy1/2 dξ on B. ξ) = σPm0 (x. we have (L. 1/2 An diﬀerential operator on a manifold M is an operator P on Ω0 M which coincides in local coordinates with a series of the form P as above. V. p (x. ξ) m0 +k if Pm = 0 for all m < m0 and Pm0 = 0. i. we ﬁrst note that the phase function φ(x. which is related to P by the asymptotic Fourier inversion formula: for any compactly supported oscillatory test function e−iψ/ u (we permit ψ = 0). written as a function of the variables (x. s) = (L∆ ◦ L.6.where each Pm is a polynomial in the operators Dj . ξ) u(y) dy dξ. ξ) : x. that is. we have (P e−iψ/ u)(x) = (2π )−n ei( x−y. and s is obtained from the pullback of u to L. whose restriction to the ﬁbercritical set Σφ = {(x. x. φ) which generates the conormal bundle to the diagonal ∆ ⊂ Rn × Rn .
many diﬀerential operators share the same principal symbol.e. s)) = I (L∆ ◦ L. the timeindependent Schr¨dinger equation is then o P ψ = 0. sP ◦ s) + O( m0 +k+1 ). Roughly speaking. or. s) is an exact. As a particular case of Theorem 6. ˆ where P = H − E is the zerothorder asymptotic diﬀerential operator on M with principal symbol θ2 p (x. Example 6. For E > 0. this theorem asserts that polynomial functions on the identity relation L∆ quantize as the Schwartz kernel of diﬀerential operators on M . i. A familiar illustration of these concepts is provided by the WKB approximation. The zero set of the principal symbol p is called the characteristic variety of the operator P .Theorem 6.8 Recall that the Schr¨dinger operator associated to a given potential V on a o riemannian manifold M is given by ˆ H=− 2 2m ∆ + mV . this correspondence is only asymptotic. and their actions on a given function coincide only up to terms of higher order in . then CP = p−1 (0) is a coisotropic submanifold of T ∗ M . s) is a ﬁrstorder approximate solution to the equation P u = 0. more properly. CP . and semiclassical states contained in CP represent solutions of the asymptotic diﬀerential equation P u = 0. As usual. In this sense. This remark suggests the quantum analog of certain coisotropic submanifolds of T ∗ M .7. θ) = − + (V (x) − E). s)) = O( m0 +k+1 ). the reduced manifold of CP . that I (L.7 If (L. where sP is the symbol on L∆ induced by the principal symbol p of P . then: P (I (L. corresponds to the kernel of P in HM . projectable semiclassical state in T ∗ M and P is an diﬀerential operator of order k on M . and as in previous sections we see that ﬁrstorder approximate solutions to the timeindependent Schr¨dinger equation arise from semiclassical states represented by o quantizable lagrangian submanifolds in CP .. the vanishing of the principal symbol p on L implies that P (I (L. If 0 is a regular value of p . and the projection relation KCP quantizes as the orthogonal projection onto this subspace. 90 . 2m The characteristic variety of p is simply the level set H −1 (E) of the classical hamiltonian of the system.
θ) . aj (x. Let M be a smooth manifold with a smooth submanifold N . we will only be concerned with a few specialized cases. an pseudodiﬀerential operator of order µ is given by (Au)(x) = (2π )n ei x−y. the conormal bundle LN ⊂ T ∗ N is generated by a single 91 . For diﬀerential operators. of the form µ+j where the symbol a(x. 28. θ) for c > 0. cθ) = cµ−j aj (x. As before. ξ) → x − y. i. As constructed in Example 4.7 (see for example [51]). the theory of Fourier integral operators provides a means for quantizing halfdensities on more general lagrangian submanifolds L of T ∗ M by replacing the function (x. y. 31. ξ) u(y) dξ dy. y.θ / a (x.θ / a(x. Fourier integral operators In its simplest form.The diﬃculty of quantizing more general symbols on the identity relation L∆ lies in the convergence of the integral (P u)(x) = (2π )n ei x−y. a detailed description of this theory can be found in [21.y.ξ)/ a(x.e. integration over the phase variables θ was welldeﬁned due to the fact that the symbol of a diﬀerential operator has a polynomial growth rate with respect to these variables. Perhaps the simplest generalization of the picture of (pseudo)diﬀerential operators given above is provided by quantizing semiclassical states whose underlying lagrangian submanifolds are conormal bundles. ξ) u(y) dξ dy. 43. y. As in the special case of diﬀerential operators. Weakening this condition while still guaranteeing that the integral converges leads to the deﬁnition of pseudodiﬀerential operators. θ) u(y) dθ dy. Each coeﬃcient aj is a smooth function on Rn × Rn \ {0} and is positively homogeneous of degree µ − j. the class of quantizable halfdensities (and canonical relations) is constrained by the necessary conditions for this integral to converge. θ) ∼ aj (x. ξ in the deﬁnition of pseudodiﬀerential operators by a phase function which generates L: (Au)(x) = (2π )n eiφ(x.. 56]. In this section.29. θ) is an asymptotic series in a (x. and there is a corresponding version of Theorem 6. In Rn . the invariance of pseudodiﬀerential operators under coordinate changes enables one to extend this theory to any smooth manifold M . the principal symbol of a pseudodiﬀerential operator is a welldeﬁned function on T ∗ M . In this case.
Under certain additional restrictions on N which are always fulﬁlled. pV . V. for example. an amplitude a on B having an asymptotic expansion in terms which are positively homogeneous with respect to the natural R+ action on the ﬁbers of pV : B → V gives rise to a welldeﬁned distributional halfdensity family I (L. Consequently. they satisfy the composition law ˜ s ˜ I (L. s) = I (L ◦ L. positively homogeneous symbols on LN induce positively homogeneous amplitudes on B of the same order.9 If X.Morse family (B. φ) with the properties that V is a tubular neighborhood of N in M and B is a vector bundle over V with ﬁber dimension equal to the codimension of N in M . s) on M . then the procedure described above can be applied to the graph Γf of f . As in the case of (pseudo)diﬀerential operators. quantizing the hamiltonian ﬂow {ϕt } of a hamiltonian ˆ H on T ∗ X gives the solution operators exp(−itH/ ) of the Schr¨dinger equation. Moreover. and we can proceed to deﬁne I (L. s) of distributions on X × Y deﬁned above corresponds via the Schwartz kernel theorem to a family of continuous linear maps 1/2 1/2 Ω0 X → Ω−∞ Y A diﬀeomorphism f : X → Y induces a unitary operator on intrinsic Hilbert spaces given by pullback: (f ∗ )−1 : HX → HY . these operators map Ω0 X to Ω0 Y and extend continuously to Ω−∞ X. o 92 . when N is the graph of a diﬀeomorphism 1/2 1/2 1/2 X → Y . f gives rise to a symplectomorphism of cotangent bundles: (T ∗ f )−1 : T ∗ X → T ∗ Y. the more interesting Fourier integral operators from Ω1/2 X to Ω1/2 Y come from quantizing canonical relations from T ∗ X to T ∗ Y which do not arise from diﬀeomorphisms from X to Y . s) represent Schwartz 1/2 1/2 kernels for continuous linear operators Ω0 X → Ω∞ Y . and that the identiﬁcation LN Σφ deﬁned by λφ is equivariant with respect to the natural R+ action on LN . The family I (Lf . When M = X × Y is a product manifold. Y are smooth manifolds and f : X → Y is a smooth diﬀeomorphism. s Example 6. ˜ ◦ s). the distributions I (L. s) as before by requiring that s be homogeneous. Of course. viewed as a smooth submanifold of X × Y . For instance. At the same time. It is easy to see that the ﬁbercritical set Σφ is invariant under the R+ action on B. ˜) ◦ I (L.
ﬁrst note that if C/C ⊥ is simplyconnected. By the homotopy lifting property. To interpret this condition in terms of C/C ⊥ . we ask what assumptions on C guarantee that the classical projection operation KC preserves the class of quantizable lagrangian submanifolds in T ∗ M . This suggests that if C is reducible. we then have ωC = [a] ∂D ˜ f f ˜ f ∗ αM ∈ Z by the assumption that the Liouville class of I is integral. let us consider what conditions the reduced manifold C/C ⊥ must satisfy in order that quantizations of the reduction and projection relations be welldeﬁned for arbitrary quantizable semiclassical states. the Liouville class of S is determined by the number αM = 2π · (2E)1/2 . in the sense that semiclassical states contained in C quantize to ﬁrstorder approximate solutions of the equation P = 0.1 Geometric Quantization Prequantization In the preceding chapter.7 7.2). Speciﬁcally.1 From Example 5. we observed that the characteristic variety C = CP of a diﬀerential operator P on a riemannian manifold M represents the classical analog of the kernel of P in HM . In this case. Applying Stokes’ theorem. f lifts to a continuous map (D. Thus. ∂D) → (C/C ⊥ . Each leaf of the characteristic foliation of CE is a circle S which projects diﬀeomorphically to a great circle in S n . Before turning to a systematic means for deﬁning HC . we conclude that KC preserves prequantizable lagrangian submanifolds provided that the reduced symplectic form ωC is itself integral. If γ parametrizes a such a geodesic. [I]) for a ﬁxed leaf I of C ⊥ .13 we recall that the standard metric on the unit nsphere S n induces a kinetic energy function kn whose constant energy surfaces CE are reducible coisotropic submanifolds of T ∗ S n for E > 0. T ∗ M ). Z) is represented by a continuous map (D. Since ˙ αM (Xkn ) = 2E. ∂D) → (C. I). S Thus CE is prequantizable if and only if E = (n )2 /2 (compare Example 4. When C is reducible. any class [a] ∈ H2 (C/C ⊥ . the reduced manifold C/C ⊥ is said to be prequantizable. then its lift to a parametrization of S satisﬁes γ = (2E)−1/2 Xkn . For the time being. A simple argument then shows that KC preserves the class of prequantizable lagrangian submanifolds provided that the Liouville class of each C ⊥ leaf is integral. Example 7. we will ignore the Maslov correction and assume that C/C ⊥ is simplyconnected. 93 . the lagrangian submanifolds of M contained in C correspond to lagrangian submanifolds of the reduced manifold C/C ⊥ . then a quantum Hilbert space HC somehow associated to the reduced manifold C/C ⊥ should map isometrically onto the kernel of P via an appropriate ∗ quantization of the adjoint reduction relation RC ∈ Hom(C/C ⊥ .
the bundle Q can be constructed explicitly from the prequantum T bundle QM over T ∗ M by noting that for any leaf I of C ⊥ . . the modZ reduction of λI represents the holonomy of QM I . parallel sections over the leaves of C ⊥ deﬁne a foliation Q of QM L whose leaf space is a principal T bundle Q over C/C ⊥ . Clearly the map ρ is linear and satisﬁes the ﬁrst Dirac axiom. ω) is not exact. A true generalization of the Segal construction due to Kostant and Souriau is achieved by ﬁrst reinterpreting ρ(f ) as an operator on sections of a line bundle over P = T ∗ M . This essentially requires that the 1forms αj agree on overlaps. The mapping C (P ) → S(HP ) is then given explicitly by the formula ρ(f ) = −i Xf + mL(f ) . g = P ∞ n f g ωM . A linear mapping ρ : C ∞ (P ) → S(HP ) satisfying the ﬁrst two Dirac axioms will be called a prequantization of the classical system represented by P . Roughly speaking. Koopman. we can then associate to a function f ∈ C ∞ (P ) the operator ρ(f )j = −i Xf + mLj (f ) . let E be the complex line bundle over P associated to the trivial principal T bundle Q = P × T via the representation x → e−ix/ of T in U (1). the idea is the following: Although the symplectic form of an arbitrary symplectic manifold (P. If C/C ⊥ is prequantizable. on C ∞ (Uj ). The space of sections of E identiﬁes canonically with C ∞ (P ) by means of the constant section s = 1. the prequantum Hilbert space is taken to be the completion of the space of smooth. there exists a parallel section of QM over I. we hope to piece these local operators together. and we arrive at nothing new. In this case. complexvalued functions on the cotangent bundle P = T ∗ M itself. with respect to the innerproduct f. and Van Hove.By Theorem D.2. In fact. if λI is integral. One way of doing this would be to impose the condition that each ρ(f )j and ρ(f )k coincide as operators on the function space C ∞ (Uj ∩ Uk ). this condition is equivalent to the existence of a principal T bundle Q over C/C ⊥ . In direct analogy with the Segal prequantization. In order to associate a “global” operator to f . The Segal prequantization is an important ﬁrst step towards attempts at (pre)quantizing more general symplectic manifolds. we can choose a covering of P by open sets Uj on which the restriction of the symplectic form satisﬁes ω = −dαj for appropriately chosen 1forms αj on Uj . and we have ρ(f ) = −i 94 Xf + mf . More precisely. The connection on Q induced by the connection on QM has curvature equal to the reduced symplectic form ωC . The basic example of prequantization is that of cotangent bundles due to Segal [52]. veriﬁcation of the second is also straightforward and will be carried out in somewhat more generality below. where as usual mL(f ) denotes multiplication by the lagrangian L(f ) = f − αM (Xf ).
the requirement that the curvature of ϕ equal the symplectic form ω implies [ξf . then we can decompose ξ into its horizontal and vertical parts: ξ = ξ − gX for some realvalued function g on Q satisfying dg = ξ dϕ. Conversely. Automorphisms of (Q. Deﬁnition 7. The curvature of such a bundle necessarily coincides with the symplectic form on P . ξ is the horizontal lift of Xg . (k) A direct computation shows that the connection form ϕ is invariant under the ﬂow of ξf . 95 . where X f denotes the horizontal lift of the hamiltonian vector ﬁeld of f . this condition can be satisﬁed if and only if ω is integral. ϕ) Let P be a prequantizable symplectic manifold with prequantum T bundle Q and connection ϕ. X] = 0. To a function f ∈ C ∞ (P ). and so ξ = ξg . ξg ] = [Xf . Xg ] + ω(Xf . Although several more modiﬁcations of this choice must be made in order to arrive at a reasonable substitute for the intrinsic Hilbert space of the base of a cotangent bundle. and X is the fundamental vector ﬁeld on Q deﬁned by the equations X ϕ=1 X dϕ = 0.g} . X] = 0. we will study prequantizable manifolds in somewhat more detail below. we associate an operator on C ∞ (Q): ξf = X f − f X. and consequently. While a direct proof of this fact follows the outline of the preceding paragraph. Xg )X − 2{f. this is an important ﬁrst step in our general quantization program. The prequantum Hilbert space in this case will be the completion of the vector space of smooth sections of a hermitian line bundle associated to Q. if Lξ ϕ = 0 for some vector ﬁeld ξ on Q. we ﬁnd that the operators ρ(f )j enjoy a similar interpretation provided that −αj + dσ deﬁnes a local representative of a connection 1form on a (possibly nontrivial) principal T bundle Q over P .2 A prequantization of a symplectic manifold (P. Thus. From the deﬁnition of X it follows that X · g = 0 and [ξ. Moreover. For the remainder of this section. according to the discussion in Appendix C. we will focus on geometric properties of prequantum circle bundles and prove that their existence coincides with the prequantizability of C ∞ (P ).where denotes the connection on E induced by the connection 1form ϕ = −αM +dσ on Q. g}X = ξ{f. [ξ. ω) is a principal T bundle Q over P equipped with a connection 1form ϕ having curvature ω. The upshot of the KostantSouriau construction is that prequantizable symplectic manifolds have prequantizable Poisson algebras. Returning to the general case.
we therefore obtain a projective unitary representation of the image of Aut(Q. This sequence can be integrated to give an exact sequence of automorphism groups as follows. and let Aut(Q. ϕ) → Aut(P. ω) denote those groups of diﬀeomorphisms which preserve ϕ. ϕ) in Aut(P. ∂qi ∂ ∂qi ∂ − qi X ∂pi 96 ξpi = . If we equip the space C ∞ (Q) of complexvalued functions on Q with the innerproduct u. A basis for this space is given by the vector ﬁelds Xqi = − which assume the form ξqi = − ∂ ∂pi Xpi = ∂ . then each F ∈ Aut(Q. ϕ) preserves µ. ω).3 Translations of R2n are generated by the hamiltonian vector ﬁelds associated to linear functionals on R2n . We interpret this observation to mean that an automorphism of Q is determined “up to phase” by an automorphism of P . Clearly the correspondence F → UF deﬁnes a unitary representation of Aut(Q. R) is assigned the trivial bracket. if P (and hence Q) is connected. The association F → f deﬁnes a group homomorphism Aut(Q. ϕ) on L2 (Q). ϕ) is a central extension of Aut(P. ϕ) → Aut(P. ω) consists precisely of the hamiltonian vector ﬁelds on P . This implies that the kernel is isomorphic to H 0 (P. f ∈ Aut(P. Let (Q. This produces the exact sequence of Lie algebras 0 → R → χ(Q. By the deﬁnition of X it follows that every F ∈ Aut(Q. where µ denotes the volume form ϕ ∧ (dϕ)n on Q. ϕ) in χ(P. ϕ) of ϕpreserving vector ﬁelds on Q with the standard Lie bracket.e. ω). that is. ϕ) be a prequantization of (P. ϕ). ω). v = Q u v µ. R) → 0. T ). To determine its kernel. ϕ) → χ(P. where H 1 (P . from the fact that π ∗ ω = dϕ it follows furthermore that f ∗ ω = ω. this means that F is the lift of a diﬀeomorphism f of P . and therefore deﬁnes a unitary operator UF on C ∞ (Q) by composition: UF (u) = u ◦ F. ω). ω) → H 1 (P .The association f → ξf therefore deﬁnes a Lie algebra isomorphism between the Poisson algebra C ∞ (P ) and the space χ(Q. Aut(P. ϕ) preserves X and is therefore T equivariant. Aut(Q. ω) by T . i. Example 7. from the exact sequence above. and the image of χ(Q. ω respectively. we simply note that the identity map on P is covered by precisely those automorphisms of Q given by the action of elements of T on its connected components. In particular. this is just the circle T and we have the exact sequence 0 → T → Aut(Q. ω).
KostantSouriau prequantization To prequantize the Poisson algebra C ∞ (P ). ξpi generate a Lie subalgebra hn ⊂ χ(Q. ξpj ] = ξ{qi . ϕ) isomorphic to R2n × R with bracket given by [(v. a). covariant diﬀerentiation by a vector ﬁeld η on P is given simply by the Lie derivative with respect to the horizontal lift η of η to Q. σ ) = (Q + Q . we ﬁnd that hn corresponds to a subgroup Hn ⊂ Aut(Q. (w. in other words. g}) = k [ρk (f ). b)] = (0. the operator ξf restricts to an operator on each eigenspace Ek of X having the form i ρk (f ) = − X f + mf . with group multiplication given by (Q. k Evidently the map ρk satisﬁes the ﬁrst Dirac axiom. To each f ∈ C ∞ (P ) and integer k. we note that if the hamiltonian vector ﬁeld of 97 . σ) · (Q . σ ) = (Q + Q . Smooth sections of Ek identify with functions on Q satisfying the equivariance condition f (p · a) = e−ika/ f (p) for a ∈ T . For this purpose. ϕ) which comprises a central extension of the translation group: 0 → T → Hn → R2n → 0. we ﬁrst recall that complex line bundles Ek associated to Q arise via representations of T in U (1) of the form x → eikx/ . P + P . w)). P. where again L(f ) = f − αM (Xf ).pj } = δij X. σ + σ + j Pj · Qj ). is suﬃces to prove that ρk (f ) acts as a symmetric operator on the real subspace of Ek . k Since [ξf . Our earlier results show that [ξqi . Hn is diﬀeomorphic to R2n × T . we assign an operator on C ∞ (Q) by ξf = − (k) (k) i ξf . P . To verify that ρk (f ) is selfadjoint. moreover ρk ({f. and so the vector ﬁelds ξqi . P + P .when lifted to Q = R2n × T via the Segal prescription ξf = X f − L(f )X. In explicit terms. ρk (g)] . By exponentiating. The group Hn is known as the Heisenberg group of R2n with its usual symplectic structure. Under this correspondence. the space of sections of Ek is isomorphic to the −ik/ eigenspace Ek of the fundamental vector ﬁeld X. ω(v. X] = 0.
whereas the space of sections of E has the “size” of the space of functions on P . s2 ω n . In other 98 . the position and momentum operators ρ(q) = −i ∂ + mq ∂p ρ(p) = −i ∂ ∂q on T ∗ R both commute with ∂/∂p and therefore do not form a complete set. ρ(q) = mq ρ(p) = −i ∂q This association agrees with our earlier heuristic quantization of the classical position and momentum observables and suggests in general that the space of sections of a prequantum line bundle E is too “large” for the third Dirac axiom to be satisﬁed. then the fact that ξf preserves ϕ implies that ξf · u. we can interpret C ∞ (M ) as the space of sections of the prequantum line bundle E over T ∗ M which are parallel along each ﬁber of π. We now deﬁne the prequantum line bundle associated to Q as E = E−1 . whereas q involves the spurious ∂/∂p term. s2 = P ∞ s1 . If we restrict ˆ ∞ to the pindependent subspace Cq (R2 ) C ∞ (R). ∂ . if f is compactly supported). the operator corresponding to q should act by multiplication alone. in quantum mechanics. wave functions depend on only half of the phase space variables. 7. v = − u. this diﬃculty is overcome: on ∞ Cq (R2 ). it follows that ρ−1 deﬁnes a prequantization of the Poisson algebra C ∞ (P ). In fact. compatible with its induced connection. the appropriate quantum state space asso1/2 ciated with a classical conﬁguration space M is (the L2 completion of) Ω0 M . Since the Liouville form αM vanishes π on each ﬁber of the projection T ∗ M → M . Indeed. equipped with a hermitian metric . v on Q. which we temporarily identify with C ∞ (M ) using a metric on M . From the standpoint of WKB quantization. and let HP denote the L2 completion of the space of compactly supported sections of E with respect to the innerproduct s1 . however.f is integrable (for example. conjugatesymmetry of the innerproduct implies that ρk (f ) · u. from the general remarks above. For an interpretation of this argument in terms of the Heisenberg uncertainty principle. Combined with the deﬁnition of ρk .2 Polarizations and the metaplectic correction The KostantSouriau prequantization of symplectic manifolds fails to satisfy the third Dirac axiom and is therefore not a quantization. According to our earliest concepts of quantization (see the Introduction). v = u. ρk (f ) · v . ξf · v for realvalued functions u. see [53]. the operator ρ−1 (f ) extends to an essentially selfadjoint operator on HP . For each f ∈ C (P ).
words. Note that the Fp representation of the freeparticle H(q. p) = p2 /2 is ρ(H) = mp2 /2 .5 The most basic examples of real polarizations are given by either the planes q=constant or p=constant in R2n . equipped with the complexvalued symplectic form ωC given by the complexlinear extension of ω to TC P .q / f dp. ω) is an involutive lagrangian subbundle F of TC P . Example 7. we have X s = 0. the polarization F satisﬁes F = F. A Fp parallel section ψ of E must therefore satisfy ∂ψ − 2πipψ ∂q and is therefore of the form ψ(q. in the sense that f= Rn f (q) δ(q) dq = (2π )−n/2 Rn ˆ eiπ p. for each complex vector ﬁeld X on P lying in F. Integrability of F means that locally on P there exist complexvalued functions whose hamiltonian vector ﬁelds (with respect to ωC ) span F. 99 . That is. For the p=constant polarization Fp . TC P denotes the complexiﬁcation of the tangent bundle of P .· / bases of L2 (Rn ) respectively. any polarization F of a symplectic manifold P which satisﬁes F = F is the complexiﬁcation of an integrable lagrangian subbundle of T P and is called a real polarization of P . Real polarizations The standard polarization of a classical phase space T ∗ M is given by the complexiﬁcation F = V M ⊕ iV M of the vertical subbundle of T (T ∗ M ). These correspond to the δq and eiπ p.p (T ∗ M ) for each p ∈ T ∗ M . Deﬁnition 7. This adjustment is generalized in the framework of geometric quantization by the introduction of the following concept.q / . we consider covariant derivatives of the form ∂ ∂q = ∂ ∂ −p ∂q ∂σ . p) = v(p)eiπ p. the correct adjustment to the size of the prequantum Hilbert space is determined by the (canonical) foliation of T ∗ M by lagrangian submanifolds. Here. The quantum state space associated to P is then given by the space of sections s of a prequantum line bundle E over P which are covariantly constant along F. and so the space of Fq parallel sections of the prequantum line bundle E identiﬁes with the space of smooth functions on qspace. meaning that Fp is a totally real subspace of TC. In general. and so the ppolarization appears to be better adapted to this case.4 A polarization of a symplectic manifold (P. In this case. Covariant diﬀerentiation along vector ﬁelds X tangent to the q=constant polarization Fq is given by the ordinary Lie derivative with respect to X.
According to the HamiltonJacobi theorem. Thus {f. one can furthermore prove Theorem 7. g ∈ F(P ).A real polarization F of a symplectic 2nmanifold P deﬁnes a subspace F(P ) ⊂ C ∞ (P ) of functions constant along the leaves of F. An aﬃne structure on a manifold M is a curvature. ω Similarly.8 Each leaf of any real polarization carries a natural aﬃne structure. The main remark we wish to make for the moment is that our earlier prequantization ρ−1 of C ∞ (P ) now represents the subalgebra F(P ) ⊂ C ∞ (P ) as multiplication operators on HF . A function H : P → R constant on the leaves of F induces a function HF on PF satisfying dH = p∗ dHF . if E is a prequantum line bundle over P . By the deﬁnition of ρ−1 . the multiplication operator on HF . The quotient of E by this 100 . From this remark.e. the map ω : T P → T ∗ P sends F to ˜ ⊥ F as well. More generally. we have Corollary 7.29). Corollary 7.9 Following [28]. Example 7. g} = 0 for any f.7 For each p ∈ P . n. those functions whose hamiltonian vector ﬁelds have ﬂows which leave the polarization invariant (not necessarily leaf by leaf) are those which are “aﬃne” along the leaves. we have the following result.and torsionfree connection on T M . i = 1.6 Every real polarization is locally isomorphic to the q=constant polarization on R2n . it follows that a section s of F over a leaf L is parallel with respect to ∗ the connection described above if and only if (˜ ◦ s) deﬁnes a ﬁxed element of T{L} PF . In this case. By modifying the normal form results of Chapter 4. · · · . Since F is a lagrangian distribution. and so p∗ T ∗ PF identiﬁes with the normal bundle F ⊥ ⊂ T ∗ P . By pulling back the position functions qi to P . there exist fi ∈ F(P ). such that the hamiltonian vector ﬁelds Xi span F in a neighborhood of p. Since any polarizationpreserving symplectic transformation of T ∗ Rn R2n is aﬃne on ﬁbers (see Theorem 3. it follows that Xf s = 0 for all s ∈ HF .. we call a polarization F on P ﬁbrating if each leaf of F is simplyconnected and the leaf space PF = P/F is a smooth manifold. we may use parallel sections of E over the leaves of F to construct a foliation of the total space of E. the hamiltonian vector ﬁeld of any member of F(P ) is contained in F. the quotient map p : P → PF satisﬁes F = ker p∗ . Since any f ∈ F(P ) has the property that its hamiltonian vector ﬁeld Xf lies completely within the polarization F. this implies that ρ−1 (f ) = mf . F(P ) is an abelian Poisson subalgebra of C ∞ (P ). i. thus deﬁning a natural identiﬁcation F p ∗ T ∗ PF .
i. F = {(v. the distribution D is involutive.e. is complete.e. the leaf spaces PD and PE are smooth manifolds. From the deﬁnition of F. The basic geometry of this situation remains the same if the symplectic structure on T ∗ M is perturbed by a form on M . Although their dimensions vary in general from point to point. A check of deﬁnitions shows that the standard and twisted symplectic structures on T ∗ M are equivalent precisely when their diﬀerence is the pullback of an exact form on M . we can “twist” the standard symplectic structure on T ∗ M as follows: ω = ωM + π ∗ η. each ﬁber of π carries a K¨hler structure. the symplectic manifold (T ∗ M. A polarization satisfying F ∩ F = {0} is called totally complex and identiﬁes with the graph of a complex structure J on P . Moreover. and the natural projection PD → PE is a submersion. ω). Complex polarizations Associated to any polarization F are the distributions DC = F ∩ F EC = F + F which arise as the complexiﬁcations of distributions D.. For a discussion of quantization in this general setting. ω) is symplectomorphic to a twisted symplectic structure on T ∗ PF . R) is integral. The polarization F is called strongly admissible provided that E is involutive. In this case. given a closed 2form η on M .foliation deﬁnes a hermitian line bundle EF over the leaf space PF whose sections identify with Fparallel sections of E. then (P. More precisely. ω is a symplectic form on T ∗ M and ker π∗ is again a polarization. E in T P . ωM + π ∗ η) is prequantizable if and only if the cohomology class of η in H 2 (M . Finally. and geometric quana tization attempts to construct a quantum state space for P from sections of an appropriate line bundle over P which are parallel along D and holomorphic along the ﬁbers of π. This general type of symplectic manifold is known as a twisted cotangent bundle. an application of Corollary 7. we refer to [53]. D and E are pointwise ωorthogonal.8 proves that if each leaf of a ﬁbrating polarization F on a symplectic manifold (P. iJv) : v ∈ T P } 101 π . i. Since π ∗ η vanishes on the ﬁbers of the projection. ⊥ Dx = Ex for all x ∈ P . The simplest example of a ﬁbrating polarization is given by the vertical polarization FM of a cotangent bundle T ∗ M .
under the identiﬁcation TC P T P ⊕ iT P . We emphasize that J is a genuine complex structure on P due to the integrability condition on the polarization F. Moreover, J is compatible in the usual sense with the symplectic form ω on P , so the hermitian metric , = gJ + iω is a K¨hler structure on P . In this section, we will only cite two examples a involving totally complex F, in which case PD = P is K¨hler, and PE = {pt}. a Example 7.10 Consider the complex plane C with its usual complex and symplectic structures. With respect to Darboux coordinates (q, p), the CauchyRiemann operator is deﬁned as ∂ 1 ∂ ∂ = +i . ∂z 2 ∂q ∂p A function f : C → C is holomorphic if ∂f /∂z = 0. If we identify sections of the (trivial) prequantum line bundle E with smooth, complexvalued functions on C, then Fparallel sections correspond to those functions annihilated by the covariant derivative ∂ + mz . ∂ = 2 ∂z ∂z To determine the general form of these sections, we note that ∂ ψ = 0 if and only if for ∂z some branch of the logarithm, z ∂ log ψ = − , ∂z 2 or zz log ψ = − + h 2 for some holomorphic function h. Hence ψ(z) = ϕe−z
2 /2
with ϕ holomorphic is the general form for Fparallel sections of E. The space HC thus identiﬁes with the space of holomorphic ϕ : C → C satisfying ϕ =
C
ϕ(z)2 e−z dz < ∞,
2
or, in other words, the space of holomorphic functions which are squareintegrable with 2 respect to the measure e−z dz, known as the Fock or BargmanSegal space. For a function to be quantizable in this picture, its hamiltonian ﬂow must preserve both the metric and symplectic structure of C and therefore consist solely of euclidean motions. Among such functions are the usual position and momentum observables, as well as the harmonic oscillator.
Example 7.11 If (P, ω) is any prequantizable symplectic manifold with a totally complex polarization F, then the prequantum line bundle E associated to ω can be given the structure of a holomorphic line bundle by taking the (0, 1) component of the connection on E with 102
respect to the complex structure J arising from F. The space HF of Fparallel sections of E then equals the space of holomorphic sections of E and is therefore completely determined by the complex geometry of P . Using some machinery from algebraic geometry (see [15]), it can be shown that for compact P , the space HP is ﬁnitedimensional and that its dimension is given asymptotically by the symplectic volume of P . More precisely, we note that for k ∈ Z+ , the line bundle E ⊗k deﬁnes a holomorphic prequantum line bundle associated with (P, kω). For suﬃciently large k, the dimension of the quantum state space HPk of holomorphic sections of E ⊗k is given by the HirzebruchRiemannRoch formula: dim HPk =
P
ekω T d(P ),
where T d denotes the Todd polynomial in the total Chern class of P . The ﬁrst conclusion to be drawn from this fact is that for k large, dim HPk is a symplectic invariant of P independent of its complex structure. Roughly speaking, k plays the role of −1 , and so “large k” means that we are approaching the classical limit. A second point to note is that dim HPk is a polynomial in k (or −1 ) whose leadingorder term is (kω)n = k n vol P. n!
P
Consequently the number of quantum states is determined asymptotically by the volume of P.
Metalinear structures and halfforms For the remainder of this chapter, we will focus on the quantization of symplectic manifolds P equipped with a prequantum line bundle E and a “suﬃciently nice” real polarization F. Although the space HF of Fparallel sections of E appears to have the right “size” in the simplest examples, there are still several problems to be resolved before we have a suitable quantum state space. The ﬁrst arises as soon as we attempt to deﬁne a preHilbert space structure on HF . On P , the square of an Fparallel section s is constant along the leaves of F, and thus the integral s
2
=
P
s, s ω n
diverges in general. On the other hand, there is no canonical measure on the leaf space PF with which to integrate the induced function s, s . Regardless of how this ﬁrst diﬃculty is resolved, we will again try to quantize the Poisson algebra C ∞ (P ) by representing elements of Aut(P, ω) as (projective) unitary operators on HF . The most obvious quantization of a symplectomorphism f : P → P (or more general canonical relation), however, is an operator from HF to Hf (F ) . If f does not preserve the polarization F, then these spaces are distinct. Thus, we will need some means for canonically identifying the quantum state spaces HF associated to diﬀerent polarizations of P . 103
Finally, the polarization F may have multiplyconnected leaves, over most of which the prequantum line bundle may admit only the trivial parallel section. A preliminary solution to this problem is to admit “distributional” states, such as are given by Fparallel sections of E whose restrictions to each leaf are smooth. Leaves on which E has trivial holonomy are then said to comprise the BohrSommerfeld subvariety of the pair (P, F). ˙ Example 7.12 Consider the punctured phase plane R2 = R2 \ {0} polarized by level sets of the harmonic oscillator H(q, p) = (q 2 + p2 )/2. The holonomy of the usual prequantum line bundle on the level set H −1 (E) is given by the modZ reduction of the area it encloses; thus the BohrSommerfeld variety consists of those circles of energy E = n . As in the case of WKB prequantization, these energy levels do not correspond to actual physical measurements, and so we will again need to incorporate a sort of Maslov correction.
The solution to manhy of these diﬃculties relies on the use of a metaplectic structure on P , a concept which we now introduce. Let P be a principal G bundle over a manifold M . A meta Gbundle associated to P and a central extension ˜ ρ 1→K→G→G→1 ˜ ˜ ˜ of G is a principal G bundle P over M together with a map Φ : P → P satisfying the equivariance condition Φ(p · a) = Φ(p) · ρ(a) ˜ ˜ ˜ for all a ∈ G. Two meta Gbundles (P1 , Φ1 ), (P2 , Φ2 ) over P are considered equivalent if ˜ ˜ ˜ there exists a Gequivariant diﬀeomorphism ψ : P1 → P2 such that Φ1 = Φ2 ◦ ψ. Example 7.13 A riemannian structure together with an orientation of an nmanifold M deﬁnes a bundle of oriented orthonormal frames in T M , i.e. an SO(n)structure on M . A spin structure on M is then a meta SO(n)bundle corresponding to the extension 1 → K → Spin(n) → SO(n) → 1 given by the double cover Spin(n) of SO(n). An orientable manifold admits a spin structure if and only if its second StiefelWhitney class vanishes.
Example 7.14 Since π1 (Sp(n)) Z, there exists a unique connected doublecovering group of Sp(n), known as the metaplectic group M p(n). A symplectic 2nplane bundle F admits a reduction of its structure group from GL(2n) to Sp(n); a metaplectic structure on F then corresponds to a lifting of the symplectic frame bundle to a principal M p(n) bundle. In general, a symplectic vector bundle admits a metaplectic structure if and only if its second StiefelWhitney class vanishes. A symplectic manifold P whose tangent bundle is equipped with a metaplectic structure is called a metaplectic manifold. If the StiefelWhitney class w2 (P ) is zero, metaplectic 104
s(p) · ρ(A)) for A ∈ M p(1). s1 deﬁne equivalent metaplectic structures if and only if (s1 )−1 · s0 admits a continuous lift s : R2 → M p(1). Bundles associated to metalinear structures via these representations will be the key to the metaplectic correction of the prequantization procedure.1) is called the metalinear group M L(n). . the group M L(n) is isomorphic to the direct product ρ GL+ (n) × Z4 with the covering map M L(n) → GL(n) given by ρ(A. a metaplectic structure ˙ ˙ Φ : R2 × M p(1) → R2 × Sp(1) ˙ is deﬁned for any choice of continuous map s : R2 → Sp(1) by Φ(p. Z2 ) (see [28]). but as a group it is not the direct product GL(n) × Z2 . The importance of the metalinear group for our purposes is that it admits 1dimensional representations which are not the lifts of representations of GL(n). note that up to topological equivalence. It is trivial as a topological covering. a vector bundle admits a metalinear structure if and only if the square of its ﬁrst StiefelWhitney class is zero (see [28]). the ˙ punctured plane R2 admits only trivial Sp(1) and M p(1) principal bundles. First note that by means of the quotient homomorphism M L(n) → M L(n)/GL+ (n) 105 Z4 . Example 7. since both Sp(1) and M p(1) are connected. A metalinear lifting of the frames of an nplane bundle E is called a metalinear structure on E. On the other hand. which can be interpreted as the identity component of M L(n) in order to give a metalinear structure on E. More explicitly. this occurs precisely when [s0 ] = [s1 ] as elements of π1 (Sp(1))/ρ# π1 (M p(1)) Z2 . To see this explicitly in a special case. Two such maps s0 . in which case an equivalence is given by ˜ ˙ ˙ ˙ the map ψ : R2 × M p(1) → R2 × M p(1) deﬁned as ψ(p. s(p) · A) ˜ By the usual homotopy theory for continuous groups. We emphasize that this classiﬁcation depends on the bundle of metaplectic frames and its covering map to the bundle of symplectic frames. More generally. The double covering of GL(n) then induced by the identiﬁcation GL(n) Gl(n) (see Example 3. A) = (p. a) = A · eiπa . its structure group can be reduced to GL+ (n). If E is orientable.structures are classiﬁed by the set H 1 (P . A) = (p.15 Lying within the metaplectic group is a doublecover M l(n) of the subgroup Gl(n) preserving the usual lagrangian splitting of R2n .
Deﬁnition 7. By the preceding theorem. and thus Λ−1/2 F is deﬁned. similarly. we may use L and J to reduce the structure group of T P to the subgroup Gl(n) ⊂ Sp(n) corresponding to GL(n).16 Let (P. Evidently. a metalinear structure on L can be enlarged to a metaplectic structure on T P . An important link between metalinear and metaplectic structures on symplectic manifolds can be described as follows. . one similarly deﬁnes the bundles Λ1/2 F and Λ−1/2 F of conjugate and negative halfforms on F . If J is any ωcompatible almost complex structure on P . a halfform λ and a conjugate halfform µ can be multiplied to give a 1density λµ on the bundle F . and therefore represents. M B(F ). the product of two halfforms on F yields an nform. 2 Now let P be a metaplectic manifold equipped with a prequantum line bundle E and a real polarization F. a)). we then obtain a complex line bundle Λ1/2 F called the bundle of halfforms associated to the triple (F. while a metaplectic structure on T P reduces to a metalinear structure on L. The resulting frame bundle is isomorphic to the frame bundle of L. Then T P admits a metaplectic structure if and only if L admits a metalinear structure.we can associate a principal Z4 bundle to any metalinear structure (M B(E). the squareroot of a volume form on F . Φ). Proof. a) → (det A)1/2 eiπa/2 . and our assertion follows from the remarks in the preceding examples. 106 L⊕L. loosely speaking. then T P = L⊕JL By the Whitney product theorem. ω) be a symplectic 2nmanifold and L ⊂ T P a lagrangian subbundle. The reason for this terminology is that the bundle Λ1/2 F can be constructed directly from the metalinear frame bundle via the representation M L(n) → C∗ given by the squareroot of (det ◦ρ) (A. and thus. Φ) on a vector bundle F . F inherits a metalinear structure from the metaplectic structure on T P . By inverting and conjugating the preceding representation of M L(n). From the representation a → eiπa/2 of Z4 in U (1). A section of Λ1/2 F then identiﬁes with a complexvalued function λ on M B(F ) satisfying λ(e) = (det A)1/2 eiπa/2 · λ(e · (A.17 The quantum state space HF associated to P is the space of sections of E ⊗ Λ−1/2 F which are covariantly constant and along each leaf of F. Theorem 7. we then have w2 (P ) = w1 (L)2 . More explicitly. it is equipped with a natural ﬂat connection inherited from the one on F.
In this way.17 deﬁnes a foliation of the preimage of L in B(F). ˙ ˙ where a. the bundle E ⊗ Λ−1/2 F admits a parallel section over Sr0 precisely when 2 p dq = π (2n + 1) πr0 = Sr0 for some integer n. the BohrSommerfeld variety consists of circles of energy E = (n + 1/2)π . a · eiφ ) → (r. In this case. Note that this is a nontrivial M l(1) bundle over Sr0 . Note that integration is welldeﬁned. Consequently. a > 0. θ.19 Over each leaf L of the polarization F. From the expression for the covering map given above. the quantum state space HF identiﬁes with the space of compactly supported sections of EF ⊗ Λ1/2 PF . r > 0 and 0 ≤ θ.18 We ﬁrst return to Example 7. a · eiθ ) : r = r0 . and the covering map R2 × M p(1) → R2 × Sp(1) is given by (r. c ∈ {0. the subbundle R2 × Gl(1) equal the subset {(r. since λ1 λ2 is a density on PF . θ. Example 7. In terms ˙ of the identiﬁcations above. the quotient of B(F) by this leafwise foliation deﬁnes a metalinear structure on the leaf space PF in such a way that halfforms on PF are in 11 correspondence with Fparallel negative halfforms on F. 0 ≤ θ ≤ 2π} ˙ of R2 × Sp(1) while its corresponding metalinear bundle equals the subset {(r. ˙ To quantize the punctured plane R2 with its polarization by circles centered at the origin. θ. ˙ we will use the trivial metaplectic structure on R2 (see Example 7. both the ˙ trivial M p(1) and Sp(1) bundles over R2 can be identiﬁed with the set of triples (r. The latter space is equipped with a natural innerproduct s1 ⊗ λ1 . it follows easily that the parallel transport of a negative halfform associated to T Sr0 around Sr0 amounts to multiplication by −1. θ. Example 7.12 to illustrate how the introduction of halfforms also enables us to incorporate the BohrSommerfeld correction in the case of the 1dimensional harmonic oscillator. φ ≤ 2π. 0 ≤ θ ≤ 2π. in accordance with the corrected BohrSommerfeld quantization conditions described in Chapter 4. we will sketch how in certain cases this deﬁnition enables us to overcome the diﬃculties mentioned at the beginning of this section. a > 0.In the following examples. s2 λ1 λ2 . 1}} ˙ of R2 × M p(1). a · e2iφ ). θ. a · eiφ ). As proven in [28]. a · ei(θ/2+cπ) ) : r = r0 . the Bott connection on T L described in Example 5. That is.14). 107 . ˙ ˙ The pullback of R2 × Sp(1) of T R2 to the circle Sr of radius r can be reduced to the trivial Gl(1) principal bundle over Sr0 by means of the lagrangian splitting T Sr0 ⊕ JT Sr0 . s2 ⊗ λ2 = PF s1 .
x at each x ∈ P . Example 7. p) = u(q) on R2n . ρ(f ) acts as pointwise multiplication for any f ∈ F(P ).p / u(q) v(p) dq dp. This construction. 108 . s2 (λ1 . ∗ By assumption. the space HFq consists of pindependent functions σq (q. p) = v(p)ei q.BlattnerKostantSternberg kernels If F1 and F2 are ﬁbrating polarizations of P with the property that F1.20 The basic example of a BlattnerKostantSternberg kernel arises from the classical Fourier transform. if the hamiltonian ﬂow ϕt of f ∈ C ∞ (P ) preserves F. Bσ2 for all σ1 ∈ H1 and σ2 ∈ H2 . The quantum operator associated to F is then ρ(f ) σ = i d (ϕt σ)t=0 . the metaplectic structure on P deﬁnes a natural isomorphism between the space of halfforms on F2 and the space of conjugate halfforms on F1 (see [28]). σ2 = 1 (2π )n/2 s1 . λ2 ) ω n . This pairing is nondegenerate and thus determines a linear operator B : HF2 → HF1 satisfying σ1 . R2n Note that this corresponds to the usual association of a distribution v (q) on qspace to a ˆ function v(p) on pspace via the inverse asymptotic Fourier transform.x is transverse to F2. Using this identiﬁcation. is due to Blattner. Given a symplectic manifold P with the structures above. then the quantum state spaces HF1 and HF2 can be related as follows. and Sternberg deﬁnes a sesquilinear pairing on the associated quantum state spaces via a BlattnerKostantSternberg kernel. λ2 ) on P . Kostant. If we identify elements σ1 . σ2 = σ1 . which we may identify with the space of smooth functions on qspace. we can pair a halfform λ1 on F1 with a halfform λ2 on F2 to obtain a function (λ1 . we recall that for the q =constant polarization Fq of R2n . then it lifts to a 1parameter family ϕt of operators on ˜ 1/2 smooth sections of E ⊗ Λ F. Similarly. The sesquilinear pairing described above is therefore given by σq . then a smooth function v(p) on pspace identiﬁes with an element σp (q.p / of HFp . P where dim(P ) = 2n. if Fp is the p =constant polarization.5. First. then a sesquilinear pairing HF1 × HF2 → C is deﬁned by σ1 . From Example 7. σp = 1 (2π )n/2 ei q. ˜ dt In particular. σ2 of HF1 and HF2 with Fi parallel sections si ⊗ λi of E ⊗ Λ1/2 Fi on P . which can be extended to more general pairs of polarizations. the symplectic form deﬁnes an isomorphism of F1 with F2 . similarly. geometric quantization attempts to represent the Poisson algebra C ∞ (P ) on HF as follows.
then we can construct and element of HF by “superposition”. ˜ We can generalize this viewpoint by considering a lagrangian submanifold L which has possibly multiple transverse intersections with the leaves of F (that is. s) is n “concentrated” at the origin of R . and the quantum operator ρ(f ) is deﬁned as ˜ ρ(f ) σ = i We refer to [28] for further details. The ﬂow ϕt deﬁnes an operator ϕt : HF → HFt . we suppose that P is a metaplectic manifold with a prequantum line bundle E and a ﬁbrating polarization F. whereas the quantization of (L. which. In this way. The meaning of this last statement is made more precise by the fact that if T : R2n → R2n is a linear symplectomorphism. via the inner product on HF . we consider the following situation.22 To indicate brieﬂy how “distributional” elements of HF arising as in the preceding example are paired. s) be a semiclassical state in R2n such that L equals the zero section of R2n T ∗ Rn and s is any section ˜ of E ⊗Λ−1/2 Fq over L. ε). such that each Lj intersects any leaf of F at most once. sLj ) and summing the results produces an element of HF as long as this sum converges. Example 7.21 If L ⊂ P intersects each leaf of F transversely in at most one point. Applying the procedure above to each (Lj . can be considered as a generalized section. s) is given by an ordinary smooth function on Rn . Similarly. s). each intersection of L with a leaf of F is transverse). If s is again a section of E ⊗ Λ−1/2 F over L. Let (L. the quantization of ˜ ˜ ˜ (L. According to the preceding example. More precisely. then we can pair it with arbitrary elements of HF by integration over L. such as a halfdensity or halfform with coeﬃcients in a prequantum line bundle over P . Our goal is to comment brieﬂy on how an element of HF can be constructed from a “semiclassical state” in P consisting of a lagrangian submanifold L ⊂ P with some extra structure. the diﬀerential Dϕt maps F into a polarization Ft for which there exists a BlattnerKostantSternberg kernel and a corresponding operator Ut : HFt → HF . Finally. we think of L as the union of lagrangian submanifolds Lj .For those f ∈ C ∞ (P ) whose hamiltonian ﬂow does not preserve F. we can regard s as the quantization of the pair (L. let L be the ﬁber of T ∗ Rn over 0 ∈ Rn along with a constant ˜ section s of E ⊗ Λ−1/2 Fq over L. then the associated quantum operator HFq → HT (Fq ) is 109 . d ((Ut ◦ ϕt ) σ)t=0 . we assume that for t ∈ (0. if L is a union of leaves of F and s is a section of E ⊗ Λ−1/2 F over L which is covariantly constant along each leaf of F. ˜ dt 7. then the section s of E ⊗ Λ−1/2 F over L corresponding to a section of E ⊗ Λ1/2 L can be extended to a section s over P which is covariantly constant on each leaf and which vanishes on leaves ˜ which are disjoint from L. thus obtaining a linear functional on HF .3 Quantization of semiclassical states In this brief section. Example 7.
up to a normalization. s) is a Dirac δfunction n concentrated at the origin of R . s) with respect to this new polarization to obtain smooth functions 2n on the leaf space RT (Fq ) whose pairing is deﬁned by integration. the quantum state obtained from (L.˜ required to be unitary. In terms of HFq . By choosing T so that both L and L are transverse to T (Fq ). we can ˜ ˜ quantize (L. s) and (L. this means ˜ ˜ that. 110 .
Additionally. h} = f {g. g} = {f ◦ ψ. Similarly. On a Poisson manifold. but with an dependent multiplication ∗ .1 Poisson algebras and Poisson manifolds The main objects in algebraic quantization theory are Poisson algebras and Poisson manifolds. g} which satisfy the compatibility condition {f g. Deﬁnition 8. all with a common underlying vector space. g) → f g and a Lie algebra structure (f. If P. it is assumed that the ∗ commutator approaches the Poisson bracket in the limit of large quantum numbers: {f.8 Algebraic Quantization An approach to quantization going back to Dirac and recently revived in [9][10][11] is based on the idea that the multiplicative structure of the ∗algebra of quantum observables is more central to quantization than the representation of observables as operators on Hilbert space. known as deformation quantization and the method of symplectic groupoids. g} = −{f ◦ ψ. the Leibniz identity implies that the Poisson bracket is given by a skewsymmetric contravariant tensor ﬁeld π via the formula {f. i. dg). Q are Poisson manifolds.1 A Poisson algebra is a real vector space A equipped with a commutative. we sketch two approaches towards this goal. h} + {f. 8. g ∈ C ∞ (Q). {f. g ◦ ψ} for all f. Quantum observables comprise a noncommutative algebra A which is assumed to belong to a family A of algebras. and that most of quantum mechanics can be done without regards for the precise nature of observables. h} g. g ∈ C ∞ (Q). 111 . →0 The abstract goal of algebraic quantization is to construct the family A of noncommutative algebras from a Poisson algebra. g ◦ ψ} for all f. associative algebra structure (f. As → 0. g) → {f. g} = lim (f ∗ g − g ∗ f )/i . A Poisson manifold is a manifold P whose function space C ∞ (P ) is a Poisson algebra with respect to the usual pointwise multiplication of functions and a prescribed Lie algebra structure. ψ is called an antiPoisson map if {f. a smooth map ψ : P → Q is called a Poisson map provided that it preserves Poisson brackets. g} = π(df. the algebra A approaches a commutative algebra A0 . In this chapter. The main examples of Poisson manifolds we will be concerned with are the following. which may be interpreted as the algebra of functions on a classical phase space.e.
2 The dual space g∗ is then a Poisson manifold when equipped with the LiePoisson bracket {f. x + y . the zeroth and ﬁrst order terms of this series are determined by the Poisson algebra structure of A. The diﬀerential of a smooth function F : g∗ × g∗ → R is a map DF : g∗ × g∗ → g × g whose composition with the Lie bracket on g deﬁnes a smooth map [DF ] : g∗ × g∗ → g. where µ ∈ g∗ . and let ∗ be a family of associative multiplications on A given by a formal power series ∞ f∗ g= j=0 Bj (f. Any smooth manifold M is a Poisson manifold when C ∞ (M ) is given the trivial bracket {f. then def {f. making V a Poisson manifold. g ∈ C ∞ (g∗ ). Dg : g∗ → g∗∗ g are the diﬀerentials of f and g. g} = π(df. g) j where each Bj : A × A → A is a bilinear map. More concretely. where ∆ : g∗ → g∗ × g∗ is the diagonal and f. g}(µ) = µ([Df. Dg]). The LiePoisson operator on C ∞ (g∗ × g∗ ) is the diﬀerential operator D deﬁned by DF = 1 [DF ](x. and Df.y) . Let g be a ﬁnitedimensional Lie algebra with dual g∗ . 2. g} = Xg · f.2 Deformation quantization The aim of deformation quantization is to describe the family ∗ of quantum products on a Poisson algebra A as an asymptotic series (in ) of products on A. dg) deﬁnes a Poisson algebra structure on C ∞ (V ).Example 8. In accordance with the introductory remarks above. 4. 3. 8. Then ∗ is called a ∗deformation of A if 112 . {f. g} = ∆∗ D(f g). Any symplectic manifold is a Poisson manifold with respect to its standard Poisson structure {f. g} = 0.3 Let A be a complex vector space equipped with a commutative associative algebra structure.2 1. Deﬁnition 8. If V is a ﬁnitedimensional vector space and π is a skewsymmetric bilinear form on V ∗ . the role of the higherorder terms is roughly speaking to give a more precise dependent path from classical to quantum mechanics.
We deﬁne the Poisson operator associated to π as the secondorder diﬀerential operator on C ∞ (V × V ) Dπ = A(∂/∂y ∗ .1. f ) = 0 for j ≥ 1 4. then the Hessian of π at 0 ∈ V ∗ × V ∗ is a linear map A = Hπ : V ∗ × V ∗ → V × V .4 If V is a ﬁnitedimensional real vector space and π : V ∗ × V ∗ → R is a skewsymmetric bilinear form on V ∗ . z). z) are linear coordinates on V × V arising from a single set of linear coordinates on V . and (f g)(y. z) = f (y) g(z) for f. g) equals the product in A. z ∗ ) are dual to (y. The MoyalWeyl operator is then the pseudodiﬀerential operator given by exponentiation: Mπ. and manipulations of this series will be purely formal. Bj (1. g ∈ C ∞ (R2n ). f ) = (−1)j B(f. This observation suggests the question of whether every Poisson structure on a function algebra C ∞ (P ) can be realized as the ﬁrst order term in a ∗deformation of A. g) 3. we must extend the product ∗ in the obvious way from A to the space A[[ ]] of formal power series. g ∈ C ∞ (R2n ) can be deﬁned as the pullback f · g = ∆∗ f g. ∂/∂z) . B0 (f. g) 2i deﬁnes a Poisson algebra structure on A. Instead it should be viewed as an “asymptotic expansion” for the product. In any case. = e−i Dπ /2 . where (y. ∂/∂z ∗ ). and (y ∗ . (f ∗ g) ∗ h = f ∗ (g ∗ h) For condition (5) to make sense. Similarly. Pointwise multiplication of functions in f. Bj (g. (∂/∂y. g} = def 1 B1 (f. 5. . We emphasize that the power series above is not in general assumed to converge for = 0. Example 8. where ∆ : R2n → R2n × R2n is the diagonal embedding. g} = ∆∗ D(f The diagonal pullback f ∗ g = ∆∗ M (f 113 g) g). a straightforward computation shows that the Poisson operator is related to the Poisson bracket induced by π via the equation {f. conditions (15) imply that {f. 2. Bj is a diﬀerential operator in each argument.
A Poisson manifold P is called regular if its Poisson tensor π has constant rank. In view of Example 8. and an inverse map ι : Γ → Γ satisfying the conditions 1. z) on V × V . then an application of the principle of stationary phase shows that an integral expression for f ∗ g is given by (f ∗ g)(x) = ei Q(y−x. Unlike deformation quantization. e A simpliﬁed proof of these results has recently been given by Fedosov [23] (see [68] for a survey of deformation quantization. in which case a theorem of Lie [39] asserts that P is locally isomorphic to a vector space with a constant Poisson structure. y) → xy deﬁned on a subset Γ2 ⊂ Γ × Γ called the set of composable pairs.3 Symplectic groupoids The method of symplectic groupoids also attempts to directly construct a noncommutative algebra A of quantum observables without explicitly identifying a quantum state space.5 A groupoid is a set Γ endowed with a product map (x. This technique has succeeded.4. g) = j! i 2 ∂ ∂ πr. M´lotte [44] extended their result to arbitrary regular Poisson manifolds. however. Deﬁnition 8. A theorem of DeWilde and Lecomte [19] asserts that the Poisson algebra of any ﬁnitedimensional symplectic manifold admits a ∗deformation. In terms of the linear coordinates (y. r. any regular Poisson manifold is locally deformation quantizable. the operator Bj in the expansion of the MoyalWeyl product is 1 Bj (f. emphasizing Fedosov’s construction). To construct a ∗product on all of C ∞ (P ). Groupoids In this section we collect some basic deﬁnitions and examples of groupoids and their counterparts in symplectic geometry.of the Moyal operator deﬁnes a ∗deformation of the Poisson algebra C ∞ (V ) called the MoyalWeyl product.z−x) f (y) g(z) dy dz. one may therefore attempt to “patch together” the local deformations to arrive at a global ∗product. where Q is the skewsymmetric bilinear form on V induced by π. Using similar techniques. 8.s If π is nondegenerate.s ∂yr ∂zs j f (y)g(z) y=z=x . this approach involves a geometric procedure which attempts to construct A for a particular value of and in particular incorporates geometric objects with certain quantum properties. ι2 = id 114 .
z). Thus Γ0 is the diagonal. In this groupoid. 4. We will be interested in groupoids with some geometric structures. and (xy)z = x(yz). x) ∈ Γ2 . (x. If Γ is a groupoid with base Γ0 and Γ0 ⊂ Γ0 . y) then belongs to Γ2 if and only if the source of y equals the target of x. then (xy. If (x. 2. Finally. a groupoid is a small category in which all morphisms have inverses. y) : (x. Combining (1) and (2). then (zx)ι(x) = z. x) ∈ Γ2 for all x ∈ Γ. Elements of Γ0 are units of Γ in the sense that xα(x) = x and β(x)x = x for all x ∈ Γ. The pair groupoid associated to a set X consists of Γ = X × X. In abstract terms. The mappings α. x). and if (x. 2. Condition (2) implies that the map α × β is transverse to the diagonal ∆ in Γ0 × Γ0 . and if (z. Multiplication Γ2 → Γ and inversion Γ → Γ are smooth. y) · (y. A disjoint union of groupoids is a groupoid over the union of their bases. there is exactly one arrow from any object to another. y) = (x. and α. y) and β(x. then ι(x) (xy) = y. (y. The unitary submanifolds form a group under the natural multiplication of subsets. 3. and Γ is said to be a groupoid over Γ0 . 115 ι . β(x) ∈ Γ0 } is a groupoid over Γ0 . endowed with the multiplication (x.6 1. z). x. Note that by (3). y). we make the following deﬁnition. the map ι is bijective and thus inverses in Γ are unique. Γ0 is a submanifold of Γ. we see that any vector bundle E deﬁnes a groupoid Γ(E) over its zero section.7 A groupoid Γ is called a Lie groupoid if 1. y) ∈ Γ2 . and so both Γ2 = (α × β)−1 (∆) ⊂ Γ × Γ and the multiplication relation m ⊂ Γ × Γ × Γ are smooth submanifolds. (ι(x). Maintaining the notation above. β : Γ → Γ0 are submersions. any groupoid whose base is a singleton comprises a group. 4. z) ∈ Γ2 . the multiplication relation of Γ is the subset m = {(xy. ι(x)) ∈ Γ2 for all x ∈ Γ. 3. The set Γ0 of all sources (and targets) is called the base of Γ. and conversely. y) ∈ Γ2 } of Γ × Γ × Γ. Example 8. y) = (y. Deﬁnition 8. yz) ∈ Γ2 . Any group is a groupoid over its identity element. z) = (x. (x.2. 3. β are the projections α(x. a pair (x. An element x of Γ can be thought of as an arrow with source α(x) = ι(x)x and target β(x) = xι(x). then Γ = {x ∈ Γ : α(x). A submanifold L of Γ is called unitary if the restriction of α and β to L are diﬀeomorphisms L → Γ0 .
3.Deﬁnition 8. Γ) linked by the equation Γ0 = m ◦ Lι . A simple computation shows that the base of T ∗ G is its ﬁber at the identity.8 A Lie groupoid Γ is called a symplectic groupoid if Γ is a symplectic manifold and the multiplication relation m is a lagrangian submanifold of Γ × Γ × Γ. its graph is a lagrangian submanifold of Γ × Γ. Γ) 116 . and ι : Γ → Γ is antisymplectic. Two immediate consequences of this deﬁnition and the calculus of canonical relations are that Γ0 is lagrangian. while Lι and m identify with the conormal bundles to the inversion and multiplication relations of G under the identiﬁcations T ∗ G × T ∗ G T ∗ (G × G) T ∗ G × T ∗ G × T ∗ G T ∗ (G × G × G). we deﬁne a groupoid structure on T ∗ G by taking α and β to be the right and left translations of covectors to the ﬁber at the identity e ∈ G. we will say that P is integrable and refer to Γ as a symplectic groupoid over P . only discrete Lie groups can be symplectic groupoids. then the pair groupoid structure on P × P deﬁnes a symplectic groupoid over P . 4. then the dual space g∗ with its LiePoisson structure is integrable. then T ∗ M . Example 8. A simple example of a Lie groupoid is given by any Lie group. This Poisson structure gives meaning to the following concept. Its unitary lagrangian submanifolds are precisely the graphs of symplectomorphisms of P . Γ) m ∈ Hom(Γ × Γ.10 1. From the requirement that the base of a symplectic groupoid be a lagrangian submanfold. T ∗ G is a symplectic groupoid over g∗ . If P is any symplectic manifold..9 A symplectic groupoid Γ is said to integrate a Poisson manifold P if there exists a Poisson isomorphism from the base Γ0 of Γ onto P . As a consequence of the axioms and the assumption that m is lagrangian. 2. For any Lie group G whose Lie algebra is g. Thus. If there exists a groupoid Γ which integrates P . If M is any manifold with the zero Poisson structure. i. equipped with the groupoid structure of a vector bundle and its standard symplectic structure.. there is a unique Poisson structure on the base Γ0 of a symplectic groupoid Γ such that α : Γ → Γ0 is a Poisson map and β : Γ → Γ0 is antiPoisson. a symplectic groupoid Γ is characterized by the three canonical relations (recall that Z is a point): Γ0 ∈ Hom(Z. Deﬁnition 8. deﬁnes a symplectic groupoid over M . Equipped with these operations and its usual symplectic structure. Lι ∈ Hom(Γ.e. If g is any Lie algebra.
g1 ) on G × G. Lι . so that the construction of the vector space AΓ follows unambiguously from the procedure deﬁned in the preceding chapter. To conclude this chapter. we ﬁnd that the quantization of the canonical relations Γ0 . There is actually a ﬂaw in the preceding two examples. and m associated to the groupoid T ∗ G of Example yields the −1 distributions δ(e) on G. z) supported on the diagonals ∆2 and ∆3 respectively.g. C). respectively. polarizations). Using the techniques of geometric quantization (prequantizations. and convolution. C). Example 8. we know that certain symplectic manifolds quantize to give vector spaces V . we will only deal with groupoids Γ equipped with reasonable (e. complex conjugation. As in Chapter 7. Example 8.Quantization via symplectic groupoids From our discussion of geometric quantization. T ∗ (M × M ). the relations Lι and m identify with the conormal bundles of the diagonals ∆2 ⊂ M × M and ∆3 ⊂ M × M × M respectively. ﬁbrating) polarizations. Lι and m quantize as evaluation (at e ∈ G). antiinvolution f (g) → f (g −1 ). and δ(g1 g2 . under the identiﬁcations T ∗ M × T ∗ M T ∗ (M × M ) and T ∗ M × T ∗ M × T ∗ M T ∗ (M × M × M ). Γ0 . the natural domain of the 117 . then the underlying symplectic manifold must possess a groupoid structure compatible with its symplectic structure. g1 . and pointwise multiplication in the associative ∗algebra C ∞ (M. If we wish V to be an associative ∗algebra with unit element. we will give several examples illustrating the spirit of the symplectic groupoid method. then the relations Γ0 . and T ∗ (M ×M ×M ) with (completions of) the function spaces C ∞ (M ). Lι . we then attempt to associate a vector space AΓ to Γ in such a way that the canonical relations Γ0 . since geometric quantizaton produces halfdensities rather than functions. δ(g1 . This is known as the integration problem for Poisson manifolds. our heuristic discussion in Section 7. the quantization of the groupoid Γ yields the usual identity element.12 Arguing as in the preceding example. these relations are quantized by the δfunctions δ(x. we may identify the quantum Hilbert spaces associated to T ∗ M. AΓ ⊗ A∗ . The relation Γ0 then quantizes as the function 1 on M . Thus.3 shows that. and lagrangian submanifolds correspond to elements in V . If successful. As described in the preceding example. C ∞ (M ×M ) and C ∞ (M ×M ×M ). and furthermore. the identity relation Γ0 coincides with the zero section of T ∗ M . and m quantize as elements of AΓ . Throughout. Γ and AΓ ⊗ A∗ ⊗ A∗ which deﬁne the structure of an associative ∗algebra on AΓ . z)δ(y. ι. after an appropriate normalization. If the Haar measure on G is used to identify the quantum Hilbert space HG with C ∞ (G. y) and δ(x. m) over P . like the algebras of quantum mechanics. Γ Γ it is in this sense that the symplectic groupoid Γ represents a classical model for the quantum algebra AΓ .11 Consider a trivial Poisson manifold M and its associated symplectic groupoid T ∗ M . g2 ) on G × G × G. The ﬁrst step in the quantization of a Poisson manifold P by the method of symplectic groupoids is to construct a symplectic groupoid (Γ. Since Lι and m are the conormal bundles of the diagonals in M × M and M × M × M .
y) → ((x + y)/2. When the map T is zero. The compatibility between the two structures on C ∞ (G) reﬂects a compatibility between the two groupoid structures on T ∗ G. say convolution. then T inherits a translationinvariant Poisson structure from V . which we identify with R2n itself. p ) such that q = q + 2 T (p + p ) and 1 q = q + T (p ) 2 p=p +p . the Poisson structure on T is trivial.14 If V is any ﬁnitedimensional vector space. Although the two multiplications associated with the two groupoid structures on T ∗ G described above live on diﬀerent spaces. Example 8.convolution operation on a group consists not of functions but of densities. q . making C ∞ (G) into a Hopf algebra. Example 8. it is possible to relate them more closely by dualizing one of them.13 If we take P = R2n with its standard symplectic structure. p. we get as V the space of smooth functions on the diagonal. there is a polarization of P × P which does not depend on any polarization of P but only on the aﬃne structure. and the multiplication on V turns out to be the Moyal product (see Example 8. (The relation between Poisson tori and noncommutative tori was studied from the point of view of deformation quantization in [50]). A symplectic groupoid which integrates T is given by the cotangent bundle T ∗ T. y) = f (xy). ωn (y − x)). and the groupoid product quantized to the usual pointwise multiplication of functions on T. p . with Γ0 equals the zero section of T ∗ T. 118 . one of the basic examples of noncommutative geometry. ˜ Using this polarization to quantize Γ. we identify T ∗ T with T × V ∗ and let m ⊂ T ∗ (T × T × T ) consist of all triples 1 (q. If T is a torus equal to the quotient of V by some lattice. should depend on the groupoid structure on Γ. To describe the multiplication relation. then a skewadjoint linear map π : V ∗ → V deﬁnes a skewsymmetric bilinear form on V ∗ and thus a translationinvariant Poisson structure on V .4). Otherwise one gets a noncommutative multiplication on C ∞ (T) which is precisely that of (the functions on) a noncommutative torus. q . the polarization comes from the isomorphism of Γ = R2n × R2n with T ∗ R2n given by (x. and Lι equals the conormal bundle of the diagonal in T × T. It appears then that the construction of V itself. The coproduct satisﬁes a coassociative law and is related to pointwise multiplication by the simple identity ∆(f g) = ∆(f )∆(g). and not just the multiplication. In this way. we obtain the coproduct ∆ on C ∞ (G) deﬁned as a map from C ∞ (G) to C ∞ (G) ⊗ C ∞ (G) = C ∞ (G × G) by the formula (∆f )(x. In fact.
2. Λα V is a 1dimensional complex vector space. The collection of such functions is denoted ΛV . then there is a natural product Λα V1 × Λα V2 → Λα V. then σ β is a welldeﬁned αβdensity on V . A density on a real vector space V of dimension n is a complexvalued function η. · · · . The use of densities instead of forms circumvents this need. unique up to transformation by a matrix of determinant ±1. As a result. A linear map T : V → V ∗ induces a realvalued 1density T on V given by T e =  det( T ei . Since GL(V ) acts transitively on B(V ). where e = (e1 . an αdensity is determined by its value on a single basis. any real bilinear form ω on V induces a 1density ω on V . if V = V1 ⊕ V2 . an αdensity on V is a map λ : B(V ) → C such that λ(eA) = λ(e) ·  det(A)α . ej )1/2 . 4.1 For α ∈ C. If σ ∈ Λα V and β ∈ C. en ) is any frame in a tangent space of M and A = (aij ) is any invertible n × n matrix. Consequently. which satisﬁes η(eA) = η(e) ·  det(A). Multiplication of densities is deﬁned by multiplication of their values and gives rise to a bilinear map: Λα V × Λβ V → Λα+β V. is determined by a choice of bases for W and V /W . If W is a subspace of V . def 119 .A Densities An nform ν on an ndimensional manifold M can be viewed as a scalar function on the space of bases in the tangent bundle which satisﬁes ν(eA) = ν(e) · det(A). Deﬁnition A. Operations on densities 1. ˜ 3. This concept can be generalized as follows. We denote the vector space of αdensities on V by Λα V. there is a natural product Λα W × Λα (V /W ) → Λα V In particular. Equivalently. deﬁned on the set B(V ) of bases in V . integration of nforms on M requires a choice of orientation. Because the “change of variables” formula for integration involves absolute values of Jacobians. then a basis of V .
the map ∗ gives rise to a natural isomorphism Λα V → Λ−α V ∗ . then there is a natural densitybundle isomorphism Λα E ⊗ Λα G Λα F. We denote by Ωα E the vector space of smooth sections of Λα E and by Ωα E the space of c smooth. The density spaces associated to the tangent bundle of M are denoted Ωα M . 120 .2 Suppose that A. The remarks above imply that if 0→E→F →G→0 is an exact sequence of vector bundles over M . where θ is any realvalued 1density on C. and the positive real number  detθ T  is deﬁned by the equation  detθ T 1/2 θ = T . C are vector spaces and 0 → A → A ⊕ C → C∗ → 0 is an exact sequence such that F C = T . compactlysupported sections. It is easy to check that the association V → Λα V deﬁnes a diﬀerentiable functor. so we can associate the αdensity bundle Λα E to any smooth vector bundle E over a manifold M . Example A. By the operations described above. we obtain an isomorphism Λα A Λα A ⊗ Λ2α C given explicitly by σ→σ⊗ T 2α F ∗ =  detθ T α σ ⊗ θ2α . Λα V1 ⊗ Λα V2 → Λα (V1 ⊕ V2 ) 6. Since (eA)∗ = e∗ (A∗ )−1 for any A ∈ GL(V ).5. If T : V → V is an isomorphism. Operations 3 and 4 induce natural isomorphisms Λα V ⊗ Λβ V → Λα+β V Λα W ⊗ Λα (V /W ) → Λα V. 7. A natural map B(V ) → B(V ∗ ) is deﬁned by associating to each basis e of V its dual basis e∗ . An nform on an nmanifold M induces an αdensity να in the manner of (4) above. there is a welldeﬁned isomorphism T∗ : Λα V → Λα V .
A natural mapping Ω1 M → C is deﬁned by integration c σ→ M σ. which equals u(x) by the usual Fourier inversion formula. 1/2 Similarly.ξ / u(y) dy dξ. a preHilbert space structure on the space Ωc M of smooth. τ = M σ τ. As a ﬁrst step. For u ∈ S. complex valued functions on Rn . note that the change of variables η = ξ/ gives (2π )− 2 n ei x. This observation also veriﬁes the asymptotic inversion formula: u(x) = (2π )−n ei x−y. A similar check of deﬁnitions proves the asymptotic Parseval formula: u v dx = Rn (Rn )∗ F u F v dξ.ξ (Rn )∗ / (F u)(ξ) dξ = (2π)−n ei x−y. Let S denote the usual Schwartz space of rapidly decreasing.ξ Rn / u(x) dx / (F −1 v)(x) = (2π )− 2 n ei x.ξ (Rn )∗ v(ξ) dξ To see that these transforms are actually inverse to one another. A simple application of this formula shows that the asymptotic diﬀerential operator Dj = −i ∂j satisﬁes the familiar equations F (Dj u) = ξj F u F (xj u) = −Dj F u. The completion HM of this space is called the intrinsic Hilbert space of M . compactlysupported halfdensities on M is deﬁned by σ. We study next the asymptotic behavior of integrals of the form I = Rn eiR(x)/ a(x) dx ∞ a ∈ C0 (Rn ). we will prove that if the critical point set of R is not contained in the support of a.η u(y) dy dη. B The method of stationary phase We begin with a few useful formulas involving the asymptotic Fourier transform. the asymptotic Fourier transform and its inverse are deﬁned respectively as (F u)(ξ) = (2π )− 2 n e−i x. R ∈ C ∞ (Rn ) as → 0. then I is rapidly decreasing in : 121 .
2 If the quadratic form Q is nondegenerate.  detdx T 1/2 Combining this expression with the asymptotic Parseval formula. then I = O( ∞ ) as → 0. Rx1 ∂x1 and so integration by parts with respect to the x1 variable gives I  = Rn eiR/ ∂ ∂x1 a R x1 dx . we can use a partition of unity to break up Supp(a) into ﬁnitely many domains as above and then applying the same argument (with x1 possibly replaced by another coordinate) to each piece. From [32. Vol. Then eiR/ a = −i a ∂ iR/ e . ∞ implying that I is O( ).6. ξ /2 and the determinant detdx T is deﬁned as in Appendix A. Lemma B. we obtain I = Rn ∗ e iQ/ eiπ·sgn(Q)/4 a dx =  detdx T 1/2 122 e−iQ (Rn )∗ ∗/ F a (ξ) dξ. )  detdx T 1/2 where Q∗ (ξ) = T −1 ξ. Thm. 2 The upshot of this lemma is that the main (asymptotic) contribution to the integral I must come from the critical points of R. K eiπ sgn(Q)/4 1 eiQ/ a dx = (2π )n/2 (Dk a)(0) k + O K+1+n/2 . 1/2  detdx T  k! Rn k=0 where T : Rn → (Rn )∗ is the selfadjoint map associated to Q and D is the secondorder diﬀerential operator given by ∂ i −1 ∂ Tjk .k ∂xj ∂xk Proof. e Rn −i x. Consequently.Lemma B.1. the asymptotic Fourier transform of the function x → eiQ(x)/ equals eiπsgn(Q)/4 e−Q (ξ)/ ξ→ . D= 2 j. we recall that for ξ ∈ (Rn )∗ . Proof.1 If dR = 0 on Supp(a). then for each nonnegative integer K. Our assertion follows by noting that (a/Qx1 )x1 ∈ C0 (Rn ) and repeating the same argument. .1].ξ e iQ(x)/ dx = (2π iπsgn(Q)/4 −i Q∗ (ξ) e n/2 e . For the general case.7. Suppose for the moment that Rx1 = ∂R/∂x1 = 0 on Supp(a).
2 We now wish to apply this lemma to evaluate integrals of the form I = M eiR/ σ. Morse Lemma . then the “change of variables” formula states that g ∗ eiR/ σ = M M eiR/ σ. the last expression on the right following from the asymptotic Fourier inversion formula. The same role in the stationary phase formula will be played by the following lemma. the change of variables ξ → −ξ doesn’t aﬀect the integral. then there exists a nondegenerate quadratic form Q on Rn and an embedding g : U → M . and consequently we have I = eiπ·sgn(Q)/4  detdx T −1/2 e−iQ (Rn )∗ ∗ (ξ)/ F a (ξ) dξ. If p is a nondegenerate critical point of a function R : M → R. where M is a smooth nmanifold equipped with a compactly supported density σ. the function R has the following normal form near p. In this case. such that g(0) = p and (R ◦ g)(x) = R(p) + Q(x).A simple computation shows that F a (ξ) = F a (−ξ).3). The critical point p is called nondegenerate if R (p) is an isomorphism. where U is a neighborhood of 0 in Rn . To deal with the integral on the right. we require the following two lemmas. proven in Section 4. 2 (This theorem is a special case of the Parametrized Morse Lemma. To this end. and R : M → R is a smooth function. we use the the Taylor series expansion (with remainder) ∗ of e−iQ / to write ∞ e−iQ (Rn )∗ ∗/ F a(ξ) dξ = k=0 1 k! n/2 −i K k (Q∗ (ξ))k F a(ξ) dξ (Rn )∗ = (2π ) k=0 K 1 k! −i k F −1 ((Q∗ )k F a)(0) + O k K+1+ n 2 K+1+ n 2 = (2π )n/2 k=0 1 k (D a)(0) k! +O . Recall that the hessian of R : M → R at a critical point p ∈ M is a welldeﬁned selfadjoint ∗ linear map R (p) : Tp M → Tp M . 123 . If g : M → M is any diﬀeomorphism.
. Γ). A more general treatment is available in [14]. we mention that deﬁnition of  detσ (R) (p) implies furthermore that f (p) ·  detf σ R (p)1/2 =  detσ R (p)1/2 for any function f on M .. Γ) → CU (M. .. The groups of ˇ degreek Cech cocycles and coboundaries are deﬁned respectively by ˇk ZU (M . Combining these observations. Let M be a smooth nmanifold and σ ∈ Ωc M . For ease of notation. 124 . 2 Finally.αk of Γ to every list (α0 . 2 C ˇ Cech cohomology ˇ This appendix will give some of the basic deﬁnitions of Cech cohomology for manifolds and describe its relation to deRham cohomology. then k e M iR/ σ = (2π ) n/2 j=1 eiR(pj )/ eiπ·sgn(R (pj ))/4 + O(  detσ (R (pj ))1/2 1+n/2 ). Γ)/BU (M . Γ) = ZU (M . Γ) = ker(δ k ) ˇk BU (M . · · · . An open cover U = {Uα }α∈I of a manifold M is said to be good if every intersection of ﬁnitely many members of U is either contractible or empty. αk+1 ). ˇ and the kth Cech cohomology group of M with coeﬃcients in Γ and relative to the covering U is the quotient ˇk ˇk ˇk HU (M . .. . . αj . j = 1. αk+1 ) = j=0 k (−1)j c(α0 . k in Supp(σ).. The group of all such cochains is denoted CU (M. If R : M → R has only nondegenerate critical points pj .αk is nonempty. Γ) = im(δ k−1 ). and a coboundary operator k+1 k δ k : CU (M. we will denote by Uα0 . αk ) for which the intersection k Uα0 . a Γvalued Cech cochain with respect to the cover U is then a rule which assigns an element cα0 .. Γ). then  detσ (R ◦ g) (p ) =  detσ R (p).. i=0 ˇ If Γ is an abelian group...3 In the notation above. where the symbol indicates which member of the list is to be deleted. Γ) is deﬁned by k+1 δ (c)(α0 ..αk the intersection k Uαi . If σ is any density on M such that σp = (g ∗ σ)p .Lemma B... we obtain Principle of Stationary Phase .. suppose that p ∈ M is a critical point of R and set p = g −1 (p).
t)) = x for all (x. π A principal T bundle over a manifold P is a locally trivial T bundle Q → P together π with a nonsingular. ﬁberpreserving action T × Q → Q. Since each Uα is contractible. Continuing in this way. we record some standard facts about principal bundles over paracompact manifolds. t) and π(hj (x. Of particular interest in these notes is the subgroup Z = 2π Z of R. f (a · p) = a · f (p) for all a ∈ T and p ∈ Q. we have d(ϕα − ϕβ ) = 0 on Uαβ . A group homomorphism Γ → Γ induces a homomorphism ˇk ˇk HU (M . referring to [17] for more details. Theorem C. D Principal T bundles In this appendix. Γ) → HU (M . 125 . there exist (k − 2)forms ψαβ deﬁned on Uαβ such that dψαβ = ϕα − ϕβ and d(ψαβ + ψβγ − ψαγ ) = 0 on the set Uαβγ for any indices α. π Local triviality of a T ﬁber bundle Q → P implies that for any good cover U of P . R). β. This association deﬁnes for each k ∈ Z+ a homomorphism w ˇk k ZDR (M ) → HU (M . t + s) = s · hj (x.k Now consider a closed kform ω ∈ ZDR (M ). This equation implies that for each x ∈ Uijk . γ. there exist homeomorphisms hj : Uj × T → π −1 (Uj ) such that hj (x. These maps give rise to the transition functions gjk : Ujk → T of Q. deﬁned by the requirement that hj (x. Γ ) in the obvious way. Γ) does not depend on the choice of U. Since Uαβ is itself contractible. t) ∈ Uj × T and s ∈ T .1 ([61]) The map w induces an isomorphism between the deRham cohomology ˇ of M and the Cech cohomology of M with real coeﬃcients. t + gij (x) + gjk (x) + gki (x)). hi (x. t) = hi (x. we see that ω determines a ˇ Cech kcocycle with coeﬃcients in R. t) = hk (x. β. ˇk One consequence of this theorem is that HU (M . Throughout this section. For any indices α.e. t + gjk (x)) for all x ∈ Ujk . Two principal T bundles Q → P π and Q → P are said to be isomorphic if there exists a smooth map f : Q → Q which is equivariant with respect to the T actions. i. denote by Z the group 2π · Z and set T = R/Zh . and satisﬁes π = π ◦ f . there exists on each Uα a (k − 1)form ϕα satisfying dϕα = ω.
and so the transition functions satisfy the cocycle condition gij + gjk + gki = 0 (mod Z ). t · (p. The fundamental theorem describing this space is the following (see [17.1 Two principal T bundles over P are isomorphic if and only if their Chern classes are equal. Z ) is known as the Chern class of Q.2. Z). Z ) are the following operations on principal T bundles. Z ). it follows that the Chern class of Q is the Cech representative of the deRham cohomology class [ω]. If Q → P is a principal T bundle having transition functions {gjk } with respect to some good cover of P . i. T bundles with connection The inﬁnitesimal generator of the T action on a principal bundle Q is a vector ﬁeld X on Q deﬁned by the equation d t · p.1. we can describe the product Q ×P Q as the quotient of the usual ﬁberproduct Q ×P Q (which in this case is a T × T bundle over the base). it follows easily that the Chern classes of inverses and products of principal T bundles are given by [−Q] = −[Q] and [Q ×P Q ] = [Q] + [Q ]. the form ϕ satisﬁes h∗ ϕ = dσ + π ∗ ϕj . The curvature of the ˜ connection ϕ is the unique closed 2form ω on P such that dϕ = π ∗ ω. p ) = (t · p. Cor. In terms of a good cover U of P .2 A closed 2form ω on a manifold P is the curvature form of a connection on a principal T bundle Q over P if and only if ω. X(p) = dt t=0 A connection on Q is a T invariant 1form ϕ on Q such that ϕ(X) = 1. The corresponding class 2 ˇ [Q] ∈ H (P . where dσ denotes the usual form j on T and the ϕj are 1forms on the Uj satisfying ϕj − ϕk = d˜jk . Similarly. ˇ Corresponding to the abelian group structure on H 2 (P . a ∈ Z for any a ∈ H2 (P . ˇ From the compatibility condition for the ϕj . then the numbers ˜ cijk = gij + gjk + gki ˜ ˜ ˜ ˇ are therefore elements of Z which deﬁne a Cech cocycle [cijk ].e. then the inverse of Q is deﬁned as the principal T bundle −Q over P obtained from the transition functions {−gjk }. the assignment Q → [Q] induces a bijective map from the space ˇ of isomorphism classes of principal T bundles over P to H 2 (P . Theorem D. On the level of the bundles themselves. From the deﬁnitions above. modulo the antidiagonal action of T . 126 .4] for a proof). Theorem D. Moreover. −t · p ). then the product of Q and Q is deﬁned as the principal T bundle Q ×P Q over P having transition functions {gjk + gjk }. If gjk : Ujk → R is any lift of gjk . if Q. Q are principal T bundles over P with transition functions {gjk } and {gjk } respectively. g where gjk are again Rvalued lifts of the transition functions of Q.
If H1 (P . H1 (P . Z ) = 0. ϕ) over P with curvature ω as an appropriate quotient of A(P. 1] → P such that γ(0) = p0 . and let e : A(P. 127 π . where ϕ is locally represented by dσ + ϕj . An equivalence relation on the product A(P. p0 ) × T is then deﬁned by the condition that (γ.ˇ Proof. p0 ) by this equivalence relation is the covering space P of P corresponding to the commutator subgroup of π1 (P ). For this purpose. Z ). t ) if γ.Iglesias. 1] is contractible. To complete the proof. h ) = σ ω. The quotient of A(P. the extension Γ is isomorphic to the product T × H1 (P . Most published proofs of this result (e. γ are homologous and t−t = σ ω. Z ) = 0. Z ) which acts naturally on Q by (h. The quotient of Q is the desired principal T bundle Q over P . Z ) on Q which preserves the connection ϕ. let s be any map from H1 (P . p0 ) → P be the endpoint map e(γ) = γ(1). p0 ) × T by this equivalence relation is a principal T bundle Q over P with connection ϕ having curvature π ∗ ω. p0 ) denote the space of smooth paths γ : [0. Q over P induce a connection ϕ + ϕ on the product Q ×P Q deﬁned locally by dσ + ϕj + ϕj . 1] → p0 . Otherwise. If Yt is the vector ﬁeld which generates this contraction. [37]) use Cech cohomology and the deRham isomorphism. and the proof is complete. and any choice of isomorphism deﬁnes an action of H1 (P . where ∂σ = γ − γ . Since this cocycle is symmetric. where π : P → P is the natural projection. 2 A connection ϕ on a principal T bundle Q → P induces a connection on the inverse −Q of Q whose local representatives are of the form dσ − ϕj . We prefer to give the following direct proof by P. it deﬁnes a central extension Γ of H1 (P . τ )[γ. then P = P . where ∂σ = s(h + h ) − (s(h) + s(h )) . p0 ) homologous if their diﬀerence is the boundary of a singular 2chain σ in P . Since T is divisible. p0 ) × T having curvature e∗ ω. z] = [s(h) · γ. and thus. see [33] for further details. connections ϕ and ϕ on the T bundles Q. p0 ) by 1 Kω = 0 (Yt e∗ ω) dt. z + τ ]. Z ) is then deﬁned by φ(h. p0 ) × T . The quotient of A(P. we call two elements γ. we will deﬁne the T bundle (Q. Let A(P. we deﬁne the 1form Kω on A(P. there exists a natural contraction of A(P. Then dσ + Kω is a connection form on the product A(P. p0 ) onto the constant map [0. γ ∈ A(P. Since the interval [0. Similarly. t) ∼ (γ . A T valued group cocycle on H1 (P .g. Z ) into the space of loops based at p0 which assigns a representative to each homology class.
T ) is called the holonomy of the (ﬂat) connection ϕ. The space of functions g : Q → C satisfying the condition g(t · p) = ρ−1 (t)g(p) 128 . ϕ ) are isomorphic if and only if (Q ×P −Q . Moreover. ϕ) → [βϕ ] induces a bijection from the space of isomorphism classes of ﬂat connections on the trivial ˇ T bundle with H 1 (P . Theorem D. ϕ) with the trivial T bundle equipped with the trivial connection ϕ = dσ. It is easy to check that if ϕ0 is the trivial connection on P × T . T ). the compatibility condition for the ϕj implies that sjk − gjk is constant for each j. ϕ ) over P are said to be isomorphic provided that there exists an isomorphism f : Q → Q of the underlying principal bundles which satisﬁes f ∗ ϕ = ϕ. Denoting by sj the composition of sj with the projection R → T . z) = (t · p. then ϕ − f ∗ ϕ0 = π ∗ β ˇ for some closed 1form β on P . dS). (Q .Two principal T bundles with connection (Q. π A connection ϕ on a principal T bundle Q → P is said to be ﬂat if dϕ = 0. On the other hand.4 A principal T bundle Q over P with a ﬂat connection admits a parallel section if and only if (Q. the classiﬁcation of such bundles reduces to the classiﬁcation of ﬂat connections on trivial T bundles over P . ϕ) has zero holonomy. k. (Q .ϕ on the trivial principal T bundle Q = P ×T over P are isomorphic if and only if they have equal holonomy. A check of the deﬁnition shows that (Q. ϕ − ϕ ) is isomorphic to the trivial bundle (P × T . The cohomology class [β] induced by this form in H 1 (P . the map (Q. we have: Corollary D. we ﬁnd ˜ ˜ that the functions sjk = sj − sk deﬁne transition functions for a trivial T principal bundle over P . the local representatives h∗ ϕ = dσ + ϕj have the property that ϕj = d˜j for functions j sj : Uj → R. A section s of a principal T bundle Q over P with connection ϕ is called parallel provided that s∗ ϕ = 0. In this s case.3 Two ﬂat connections ϕ. 2 Associated line bundles A representation ρ : T → U (1) enables us to associate to any principal T bundle Q → P a complex line bundle E → P deﬁned explicitly as the quotient of Q × C by the T action t · (p. ϕ). Thus. and so there exists an isomorphism f : Q → P × T . ϕ). The map sQ : Q → Q associated to a parallel section deﬁnes an isomorphism of (Q. ρ−1 (t)z). Since curvature is obviously invariant under isomorphisms of principal T bundles with connection.
for any element p of π −1 (x). where the section sg : P → E is deﬁned by the requirement that sg (x) = [(p. 129 . g(p))].is identiﬁed with the space of sections of E by the assignment g → sG .
A. 4. D. 1.A. [13] Bohm.M.B. Basel. Birkh¨user. and Guillemin. eds. Boston.. Flato. Annals. L. SpringerVerlag. 111. M. [17] Brylinski. Some remarks about the associated envelope of a Lie algebra.. Graduate Texts in Math. Dynamical systems IV. 93. A. and Guillemin. Vol. Ann. pp.I. V. Loop Spaces..I. New York. and Marsden. [10] Berezin.. [8] Audin.. Princeton University Press. V. SpringerVerlag.F. 61151.. On a characteristic class entering into conditions of quantization. (1967). New York. Foundations of Mechanics. Deformation theory and quantization. [16] Br¨ning. The Geometry and Physics of Knots. 1978. A. 1990. and Geometric Quantization. V. of Math. I and II. 1978. Mathematics Past and Present: Fourier Integral Opu erators. Diﬀerential Forms in Algebraic Topology. 11091165. Benjamin/Cummings. The Spectral Theory of Toeplitz Operators.. and Givental’. V. Berlin. [3] Arnol’d. R.I. 1992.References [1] Abraham. 107. 1 (1967) 113. Berlin.L.I.. Quantum Theory.. Lichnerowicz. Press. [6] Arnol’d. 1981.. (1977). [12] Besse. Symplectic geometry and topology.P. Progress in Math. SpringerVerlag. Characteristic Classes.I. Anal. 2nd edition. 8 (1974).. D. [5] Arnol’d. [14] Bott.. [11] Berezin.E. J. Birkh¨user.. The Topology of Torus Actions on Symplectic Manifolds. a [9] Bayen.. J. 93 SpringerVerlag. F. Progress in Mathematics vol. and Tu..I. Encyclopedia of Mathematical Sciences.. Appl. Funct. Cambridge. R. vol. 1994. Studies 99. and Sternheimer. 1951. [15] Boutet de Monvel.. SpringerVerlag. 1991.. V. 1982. Princeton. Anal. Reading. Phys.. 60. Quantization. New York. Novikov.. Mathematical Methods of Classical Mechanics. [7] Atiyah. 1990. SpringerVerlag. Symplectic geometry. Fronsdal. PrenticeHall. F.. J. Appl. [4] Arnol’d. 1136. Catastrophe theory (3rd edition).. New York. preprint 1993. L. V. C. Manifolds All of Whose Geodesics are Closed. M. Funct. a 130 . V. [2] Arnol’d. Arnol’d and S. Ergebnisse der Math. Math USSR Izv. V. 1978. Cambridge Univ. 1992. F. 91102. New York.A.
vol.. Kluwer Acad. [25] Fomenko. [28] Guillemin. o SpringerVerlag. 1950. A. Harmonic Analysis in Phase Space. Annals of mathematics studies 122. American Mathematical Society. C. Surveys 14. Soc. 14. M. New constructions of symplectic manifolds. pp. M. [24] Folland. Inst. Deformation Theory of Algebras and Structures and Applications. [33] Iglesias. Math. [27] Gompf. Lecture Notes in Math.. The Analysis of Linear Partial Diﬀerential Operators. [31] H¨rmander.T. Providence. Existence.. 1930. and Sternberg.. e Georges Thone.[18] Connes. San Diego.. New York.B. and Lecomte. 1973. Noncommutative Geometry. Semiclassical analysis for the Schrodinger operator and applications..V. Hazewinkel and M. [29] Guillemin.P. Courant Institute. S. Cambridge. The Principles of Quantum Mechanics. Princeton University Press. 1989. e e e Colloque de Topologie (Espaces Fibr´s) tenu a Bruxelles du 5 au 8 juin. 119. Geom. Math. NYU. Fourier Integral Operators. SpringerVerlag. Clarenden Press. 2955.I. V. 131 . A. Pub. New York. Ann.. Dordrecht (1988).. [26] Golubitsky. 1988.. Amer. [34] Karasev. and Maslov.. eds.. Geometric Asymptotics. V. V. P. Li`ge and Masson et Cie. Oxford. [21] Duistermaat. J. (New edition to be published by Birkh¨user Boston in 1995. M. B. 40 (1994) 213238.. S. Acta Math.. Formal deformations of the Poisson Lie algebra of a symplectic manifold and starproducts.. L... Guillemin. R.. Providence. 1993. G. New York. 1995. 14.. to appear. Graduate Texts in Math. Paris. Gordon and Breach. P. [20] Dirac. Vols... o [32] H¨rmander.. 1336 (1988).. (1973). Diﬀ. equivalence..) a [22] Ehresmann. La trilogie du moment. L. Symplectic Techniques in Physics. Fourier (to appear). Academic Press. derivations. A simple geometrical construction of deformation quantization. M. [19] De Wilde. 1977. e [23] Fedosov. P. 897960.. Les connexions inﬁnit´simales dans un espace ﬁbr´ diﬀ´rentiable. J. Nonlinear Poisson Brackets : Geometry and Quantization Translations of mathematical monographs vol. Ann. in M. J. V. 198390. and Sternberg.. 127 (1971) 79183. [30] Helﬀer. R. Cambridge University Press. Symplectic Geometry. Stable Mappings and Their Singularities. Math. Fourier integral operators I.. Gerstenhaber.. 1951. 1984. B.. A..
New York. in M. unter Mitwirkung von Prof. 1986. 1336.. [49] Omori. Soc. Lecture Note Series 174. 1975. Moscow. [43] Maslov. Mathematical Foundations of Quantum Mechanics. and Marle. 137172. Math.I. vol. W. [37] Kostant. J. Examples of symplectic structures. Deformation Theory of Algebras and Structures and Applications. Singapore. P. S.W. Dordrecht. Inﬁnitedimensional Lie Transformation Groups. Encyclopedia of Mathematical Sciences. D. Geometric quantization. Lecture Notes in Math. V.. S. Vol. eds.. [38] Libermann. [40] McDuﬀ. Lectures on Mechanics. W. 120 (1965).. Reading. 1987.. V. New York. Reidel. 122 (1989). 89 (1987). [39] Lie. [46] Nazaikinskii.. pp. 961972. Berlin.. Quantization and unitary representations.. C. W.. M. Deformation quantization of Heisenberg manifolds... 1992. Teubner.P. Trans. 4. 1890. 132 . SpringerVerlag. G.. New York. [47] Olver. 531562.427. Theory and Applications. Theorie der Transformationsgruppen. On the volume elements on a manifold.. Hazewinkel and M.. Asymptotics and Special Functions. 1965. Univ.. Comm. 1963. Methods of Noncommutative Analysis.. [36] Klingenberg. and Shatalov. Cambridge Univ. Leipzig. Kluwer Acad. London Math. Dynamical Systems IV. Geometric Perturbation Theory in Physics. Invent. Novikov.M..P. Amer.. World Scientiﬁc. Math. D. Riemannian Geometry. Soc. [44] M´lotte. J. 1990. deGruyter Studies in Math. B. Dr.Benjamin. (Zweiter Abschnitt. Press. Math. Gerstenhaber.. 170 (1970). Academic Press. Phys. Symplectic Geometry and Analytical Mechanics. Berlin. Invariant deformations of the Poisson Lie algebra of a symplectic mane ifold and star products.A. B.. de Gruyter (to appear). 1982. [50] Rieﬀel. Friedrich Engel). Lecture Notes in Mathematics. 1974. Sternin. SpringerVerlag. I. Dordrecht (1988). New York. eds. F. 280296. [41] Mackey. 87208.[35] Kirillov. V. Gos..J. V. [48] Omohundro. Theory of Perturbations and Asymptotic Methods. [45] Moser. [42] Marsden. Cambridge.. Pub.. Moskov. Arnol’d and S.
.. Paris. W. 705727. Amer. Phys. Lectures on Symplectic Manifolds. Ann.. Brookline. Boston. Bulletin A. e [56] Treves. eds. F. Comm. J.M.. 1987. 836 (1980). Birkh¨user. geometric quantization. 329346. Math. Birkh¨user.Weinstein. The Legendre transform.M. 1987. [63] Weinstein.. P. Vols I. Japan 40 (1988). Groupes diﬀ´rentiels. Inst.. in Symplectic Geometry.M. a [67] Weinstein.[51] Robert. [53] Sniatycki. Geometric Quantization and Quantum Mechanics.. 1 (1960).. in Symplectic Geometry and Mathematical Physics: actes du colloque in l’honneur de JeanMarie Souriau. Symplectic Geometry and Fourier Analysis. ´ [58] Unterberger. Symplectic manifolds and their lagrangian submanifolds. Providence.S. New York.Dazord and A. Symplectic groupoids. A. 6 (1971). 91128. and Integrable Systems.. Structure des Syst`mes Dynamiques. New York. SpringerVerlag. [59] Vaisman. Lecture Notes in Math. Sci. Advances in Math. eds.. Autour de l’Approximation Semiclassique. Ec. e Sup. Sci. 1977.. 1–13.. H. A. J. Coisotropic calculus and Poisson groupoids. J. a Boston. P. Math. New York. A. N. no. Symplectic geometry.. Dunod. Symplectic Geometry and Secondary Characteristic Classes. Quantization of nonlinear systems. I. S´minaire Bourbaki.. and irrational rotation algebras.. Sur les th´or`mes de de Rham. [64] Weinstein. 68. Norm. Press. 1991. Noncommutative geometry and geometric quantization. (new series) 5 (1981). A.. Helv.II.. [54] Souriau. Math. e 133 . 789.M. 1980. Poincar´. Springer. [61] Weil. R. A. Math.. 26 (1952) 119145. Progress in Mathematics. 468488...Donato et al. [68] Weinstein. [66] Weinstein. Soc. Plenum Press. Math. [65] Weinstein. J. 1970. A. J. 1991. Introduction to Pseudodiﬀerential and Fourier Integral Operators. MSRI Publications. 1977. Juin 1994. Soc. A. Quantiﬁcation et analyse pseudodiﬀ´rentielle. D. a [52] Segal. 133158. e e [62] Weinstein. Birkh¨user. Regional conference series in mathematics 29. A. vol. [57] Tulczyjew. A. 21 (1988). Deformation quantization.. Ann. [60] Wallach. e [55] Souriau. I. Basel. 27 (1977) 101e 114. 1980. Groupoids..
Dover Publications. and Xu.J. 134 . P. Tangent and Cotangent Bundles.. The Theory of Groups and Quantum Mechanics. Oxford. Press.. [71] Woodhouse.. A. Diﬀerential Geometry Dekker. N. Geometric Quantization. New York. 1992. Math. Oxford U. S. and Ishihara. Reine Angew. 417 (1991).. 159189 [70] Weyl. 1950.[69] Weinstein. Extensions of symplectic groupoids and quantization. 1973. [72] Yano.M. New York.. 2nd Ed.. J.. K. H.