Mathematical Methods of Classical Mechanics

Complements and exercises

Guido Carlet

Version 0.3

Contents
1 Newtonian mechanics 1
1 Galilean spacetime and inertial frames . . . . . . . . . . . . . . . . . . . . . . 1
2 Galilean group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
3 Newton’s principle of determinacy and Newton equation . . . . . . . . . . . . 2
4 One dimensional systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
5 System of particles in space . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Lagrangian formalism on Rn 3
6 Lagrangian systems and Euler-Lagrange equations . . . . . . . . . . . . . . . 3
7 Euler-Lagrange equations for a system of particles in a potential . . . . . . . 5
8 Lagrangians depending on the velocities; the e/m field . . . . . . . . . . . . . 6
9 Calculus of variations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
10 Holonomic constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
11 Noether theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
12 Similarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
13 Central fields and the Kepler problem . . . . . . . . . . . . . . . . . . . . . . 9

3 Hamiltonian formalism on Rn 9
14 Hamilton equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
15 Legendre transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
16 Equivalence of Euler-Lagrange and Hamiltonian equations . . . . . . . . . . . 11
17 Variational formulation of Hamilton equations . . . . . . . . . . . . . . . . . . 12
18 Poisson brackets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
19 Cyclic coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
20 Symplectic and Hamiltonian matrices . . . . . . . . . . . . . . . . . . . . . . 13
21 Hamiltonian vector fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
22 Generating functions for canonical transformations . . . . . . . . . . . . . . . 15
23 Canonical transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
24 Hamilton-Jacobi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
25 Separation of variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

i
4 Lagrangian mechanics on manifolds 16
26 Smooth manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
27 The tangent bundle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
28 Euler-Lagrange equations on a manifold . . . . . . . . . . . . . . . . . . . . . 16
29 The Lagrangian vector field . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

5 Hamiltonian mechanics on manifolds 17

30 The cotangent bundle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
31 The Poisson bracket on the cotangent bundle . . . . . . . . . . . . . . . . . . 17
32 Poisson manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
33 Hamiltonian vector fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
34 Canonical transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
35 The fibre derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
36 The symplectic structure on the cotangent bundle . . . . . . . . . . . . . . . 20
37 Symplectic manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
38 Poisson and symplectic manifolds . . . . . . . . . . . . . . . . . . . . . . . . . 20
39 Liouville theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
40 Darboux theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
41 Poisson structure on the dual of a Lie algebra . . . . . . . . . . . . . . . . . . 21
42 Completely integrable systems . . . . . . . . . . . . . . . . . . . . . . . . . . 21

ii
Introduction
These notes are work in progress and have to be considered as a complement to the notes
taken during classes. When a section is incomplete, a summary of the content and indications
for further reading appear in italics.

This course aims bridge the gap between the usual treatment of analytical mechanics1 -
the Lagrangian and Hamiltonian formulations of Newtonian mechanics - and the modern
mathematical approach based on symplectic and Poisson geometry. Familiarity with elemen-
tary Newtonian mechanics is assumed, but several important topics will be recalled during
exercise classes.

1 Newtonian mechanics
The elementary treatment of Newtonian mechanics is reviewed for example in the lecture
notes [Ton13] which are available online. For an outline of the basic facts of Newtonian
mechanics you can also refer to Chapters 1 and 2 of the classic book [Arn89].

1 Galilean spacetime and inertial frames

An experimenter in Galilean spacetime2 sets out some devices to measure lengths, angles
and time intervals. He ideally fixes an origin in spacetime and a set of three orthogonal axis
that allow him to assign to each event coordinates in space r = (x, y, z) and in time t. He
calls this a reference frame.
In doing so he discovers that the space he inhabits is Euclidean, namely he can sum
displacements in space as if they were vectors in R3 and the distances and angles he measures
behave in accordance to the rules encoded in the Euclidean inner product r·r0 = xx0 +yy 0 +zz 0 .
He also finds that he agrees with all other experimenters about the lengths, angles and time
intervals he measures.
Investigating the laws of nature he finds that there are certain frames of reference, which
he calls inertial frames, in which such laws assume a particularly simple form. In an
inertial frame a mass which is left alone and is not subject to external influences continues
indefinitely in its uniform motion, that is to say constant in direction and velocity. This in
particular implies that any reference frame which moves uniformly with respect to an inertial
frame is also inertial. He observes that the laws of nature are the same in all inertial frames.
He calls this the Galilean relativity principle.
The following description of the relativity principle was given by Galileo in his “Dialogue
concerning the Two Chief World Systems” (1632):
Shut yourself up with some friend in the main cabin below decks on some large
ship, and have with you there some flies, butterflies, and other small flying
1 The main historical sources are the original writings of the scientists that contributed to the development

of classical mechanics:
• Galileo Galilei (1564-1642), Dialogo sui due massimi sistemi del mondo, 1632.
• Isaac Newton (1642-1727), Philosophiae Naturalis Principia Mathematica, 1687.
• Giuseppe Luigi Lagrange (1736-1813), Méchanique Analytique, 1788.
• William Rowan Hamilton (1805-1865), On a general method in dynamics, 1834.
• Carl Jacobi (1804-1851), Vorlesungen über Dynamik, 1866.
• Henri Poincaré, Les Méthodes Nouvelles de la Mécanique Céleste, 1892-99.

2 The notion of absolute space and universal time was introduced by Newton.

1
 animals. Have a large bowl of water with some fish in it; hang up a bottle that
empties drop by drop into a wide vessel beneath it. With the ship standing still,
observe carefully how the little animals fly with equal speed to all sides of the
cabin. The fish swim indifferently in all directions; the drops fall into the vessel
beneath; and, in throwing something to your friend, you need throw it no more
strongly in one direction than another, the distances being equal; jumping with
your feet together, you pass equal spaces in every direction. When you have
observed all these things carefully (though doubtless when the ship is standing
still everything must happen in this way), have the ship proceed with any speed
you like, so long as the motion is uniform and not fluctuating this way and that.
You will discover not the least change in all the effects named, nor could you tell
from any of them whether the ship was moving or standing still. (Galileo Galilei)

2 Galilean group
The coordinates of on event measured in different inertial frames are related by an element
of the Galilean group Gal(3), that is generated by:

• translations in spacetime Ta,s : (r, t) 7→ (r + a, t + s), for a ∈ R3 and s ∈ R,

• rotations in space R : (r, t) 7→ (Rr, t), for R ∈ O(3),
• and uniform motions Bv : (r, t) 7→ (r + tv, t), for v ∈ R3 .
Exercise 2.1. Denote G(R, v, a, s) = Ta,s ◦ Bv ◦ R an arbitrary element of the group Gal(3).
Compute the element corresponding to the composition G(R, v, a, s) ◦ G(R0 , v 0 , a0 , s0 ). Show
that under the identification R4 ,→ R5 given by (r, t) 7→ (r, t, 1) the element G(R, v, a, s) acts
as matrix multiplication by  
R v a
 0 1 s . (2.1)
0 0 1
Exercise 2.2. The Galilean group Gal(3) is a Lie group of dimension 10. Find the generators
of its Lie algebra and their commutation relation.

3 Newton’s principle of determinacy and Newton equation

Newton’s principle of determinacy, the form of Newton equation, inertial mass, constraints
by the Galilean principle of relativity on the form of the forces, configuration space, potential
systems, simple examples of mechanical systems. See class note or any book on classical
mechanics, e.g. [Arn89] §1-2.

4 One dimensional systems

Newton equation for systems with one degree of freedom, conservation of energy, phase flow,
integration by separation of variables, examples. See class notes, exercise classes, and [Arn89]
§4. For a more detailed mathematical analysis we refer to a course on ODEs and in particular
to §12 in [Arn92].
Exercise 4.1. Show that the Newton equation mẍ = F (x) in one dimension is conservative
(i.e. F is always given by a potential V ) and that it can be integrated by separation of
variables.
Exercise 4.2. Show that the Newton equation mẍ = F (ẋ) in one dimension can be integrated
by separation of variables.

2
5 System of particles in space
Forces of interaction, closed systems, conservation of total momentum, total angular momen-
tum, and energy, center of mass. See class notes and [Arn89] §10.

2 Lagrangian formalism on Rn
We discuss the Lagrangian formalism for systems with configuration space Rn . For introduc-
tory lecture notes on analytical mechanics see [Ton04]. Classic textbooks include [LL69], [GPS01]
and [Arn89].

6 Lagrangian systems and Euler-Lagrange equations

Let us consider a system with configuration space Rn . The generalised position or configura-
tion of the system is given by a vector q in the configuration space Rn . The components of the
vector q are called generalised coordinates. The evolution in time of the configuration of the
system is described by a map q = q(t) : R → Rn which we call a time-dependent path. At a
time t the generalised velocity is q̇(t) = dq n
dt (t). The map (q, q̇) = (q(t), q̇(t)) : R → R × R
n

is called a lift of the path q(t).

A Lagrangian is a function

L = L(q, q̇, t) : Rn × Rn × R → R (6.1)

of the generalised position q ∈ Rn , generalised velocity q̇ ∈ Rn and time t.

The Euler-Lagrange equations
 
d ∂L ∂L
− =0 i = 1, . . . , n (6.2)
dt ∂ q̇i ∂qi
are a system of n ODEs for the variables q = (q1 , . . . , qn ) with independent variable t.
A motion of the system is a time-dependent path q = q(t) that solves the Euler-Lagrange
equations, namely such that substituting its lift (q(t), q̇(t)) in (6.2) gives an identity for all
t ∈ R. An orbit or trajectory of the system is the image of a motion in configuration space.
Given a Lagrangian L we define the functions
∂L
p = p(q, q̇, t) = , E = E(q, q̇, t) = p · q̇ − L (6.3)
∂ q̇
respectively called generalised momentum and energy.
Remark 6.1. Notice that the coordinates q̇ are the components of the tangent vector to a
path expressed in coordinates q. Under a coordinate transformation they indeed transform
as the components of a tangent vector
∂Qj
Q̇j = q̇i (6.4)
∂qi
for Q = Q(q) a coordinate transformation in Rn . For a time-dependent transformation of
coordinates Q = Q(q, t) we have
∂Qj ∂Qj
Q̇j = q̇i + . (6.5)
∂qi ∂t
In both cases however we get that
∂ Q̇j ∂Qj
= , (6.6)
∂ q̇i ∂qi

3
from which it is easy to see that the momenta transform as components of a cotangent vector
∂qj
Pi = pj . (6.7)
∂Qi

Observe that in (6.4)-(6.5) what we are really saying is that the function Q̇(q, q̇, t) is
defined in such way that when evaluated on (q(t), q̇(t)) is coincides with the total t-derivative
of Q(q(t), t). Therefore, more precisely, the definition
∂Qj ∂Qj
Q̇j (q, q̇, t) = q̇i (q, t) + (q, t) (6.8)
∂qi ∂t
implies
∂qj
Pi (Q, Q̇, t) = pj (q, q̇, t) q=q(Q,t) ∂Q (Q, t). (6.9)
i
q̇=q̇(Q,Q̇,t)

Exercise 6.2. Compute the transformation law of the energy E under a coordinates transfor-
mation Q = Q(q, t) in configuration space.
Proposition 6.3. The Euler-Lagrange equations are invariant under coordinates transfor-
mations Q = Q(q, t) in the configuration space Rn .
Proof. Show that:
 
d   ∂L d ∂L ∂qj
Pi (Q, Q̇, t) − (Q, Q̇, t) = (pj (q, q̇, t)) − (q, q̇, t) (Q, t).
dt ∂Qi dt ∂qj q=q(Q,t) ∂Qi
q̇=q̇(Q,Q̇,t)
(6.10)

 2 
We say that a Lagrangian L is regular if det ∂∂q̇∂Lq̇ 6= 0 at all points (q, q̇, t). For a
regular Lagrangian the Euler-Lagrange equations satisfy Newton’s principle of determinacy:
Proposition 6.4. For a regular Lagrangian L, the Euler-Lagrange equations are n second
order ODEs that uniquely determine the motion q(t) in terms of the initial configuration
q(0) = q(0) and velocity q̇(0) = q̇(0) of the system.
Proof. As en exercise, prove that

∂2L ∂2L
 
∂L
q̈j = A−1
ji − q̇k − (6.11)
∂qi ∂ q̇i ∂qk ∂ q̇i ∂t
∂L
where A is the matrix ∂ q̇∂ q̇ .

In the following two exercises we investigate what happens when the Lagrangian is not
regular.
Exercise 6.5. Consider a time-independent Lagrangian L = L(q) that does not depend on
the velocities q̇. Assuming L is generic, i.e. has isolated critical points, find the motions.
Exercise 6.6. Consider a time-independent Lagrangian L that depends affinely on q̇:

L = L0 (q) + L1 (q) · q̇, (6.12)

where L1 : Rn → Rn . Study the solvability of the Euler-Lagrange equations, for generic

L1 (q).
Different Lagrangians may give the same Euler-Lagrange equations, as we see in the
following two exercises.

4
Exercise 6.7. Show that the Euler-Lagrange equations do not change if we add to L the total
derivative in time of an arbitrary function f (q, t), i.e.
df ∂f ∂f
L 7→ L + =L+ · q̇ + . (6.13)
dt ∂q ∂t
Compute how the generalised momentum and energy transform under such change of
Lagrangian.
Exercise 6.8. Show that the Euler-Lagrange equations are invariant under the transformation
L 7→ aL + b for a, b ∈ R, a = 6 0. Observe that the momentum and the energy are not
invariant.
We say that a function f = f (q, q̇, t) is conserved (or is a constant of motion or is an
integral of the motion or is a conserved quantity) if its derivative in time along any motion
of the system q(t) (i.e. its total derivative) is zero:
df ∂f ∂f ∂f
= · q̇ + · q̈ + = 0. (6.14)
dt ∂q ∂ q̇ ∂t
Proposition 6.9. If the Lagrangian is time-independent, L = L(q, q̇), then the energy E is
conserved.
Proof. One can check that  
dE dp ∂L
= − · q̇ = 0. (6.15)
dt dt ∂q

The component pi = ∂∂L q̇i of p is called momentum conjugate to qi . We say that a

∂L
coordinate qi is cyclic, if L does not depend on it, ∂qi
= 0. If follows immediately from the
Euler-Lagrange equations that:
Proposition 6.10. The conjugate momentum to a cyclic coordinate is conserved.

7 Euler-Lagrange equations for a system of particles in a potential

Let us consider a system of particles in a potential V . The configuration space is Rn , where
n = 3N for N particles in space.3 The potential (energy) is a function V = V (r, t) of the
configuration r and of time t. The motion r(t) satisfies Newton equations
∂V
mi r̈i = − , i = 1, . . . , n. (7.1)
∂ri
The kinetic energy is the function T = T (ṙ) defined by
1X
T = mi ṙi2 . (7.2)
2 i

We define the Lagrangian L = L(r, ṙ, t) as the difference of the kinetic and potential energies.
Proposition 7.1. Newton equations for a system of particles in a potential V (r, t) are
equivalent to Euler-Lagrange equations for the Lagrangian L = T − V .
Proof. Just observe that in this case
∂L ∂L ∂V
= mi ṙi , =− . (7.3)
∂ ṙi ∂ri ∂ri

3 We might as well consider particles on a line or on a plane, e.g. for a single particle on the plane n = 2,

so here we just require n to be a positive integer.

5
8 Lagrangians depending on the velocities; the e/m field
The Lagrangian formulation of Newton equations can hold also for systems of particles in
potentials that depend on the velocity. The main example is that of the Lorentz force due to
an e/m field.
Let us consider the case of a single particle of mass m and charge e moving in space in
an e/m field specified by the vector fields E(r, t) and B(r, t). The configuration space R3
coincides with the Euclidean space. The Lorentz force is F = e(E + ṙ × B).
Exercise 8.1. Show that the Lorentz force can be expressed in terms of a velocity-dependent
potential V (r, ṙ, t) via
∂V d ∂V
F =− + , V = e(φ − ṙ · A), (8.1)
∂r dt ∂ ṙ
where φ and A are the scalar and vector potentials of the e/m field.
Let L = T − V . It follows from (8.1) that
d ∂L ∂L d ∂T d ∂V ∂V
− = − + = mr̈ − F, (8.2)
dt ∂ ṙ ∂r dt ∂ ṙ dt ∂ ṙ ∂r
hence Newton equations are equivalent to Euler-Lagrange equations.
Clearly this generalises to any system of particles in a potential V = V (r, ṙ, t) depending
on the velocities such that Newton equations take the form
∂V d ∂V
mi r̈i = − + , i = 1, . . . , n, (8.3)
∂ri dt ∂ ṙi
as in the case of a system of particles in an e/m field.
Proposition 8.2. Newton equations of the form (8.3) are equivalent to Euler-Lagrange
equations for the Lagrangian L = T − V with T given by (7.2).
Exercise 8.3. Show that under a gauge transformation
∂f ∂f
A 7→ A + , φ 7→ φ − (8.4)
∂r ∂t
where f (r, t) is an arbitrary function, the e/m fields E, B are unchanged. Show that the
Lagrangian changes as follows
df
L 7→ L + e . (8.5)
dt
Conclude that the Euler-Lagrange equations are unaffected by gauge transformations.
Exercise 8.4. Find the motion of a particle in a constant e/m field.

9 Calculus of variations
See class notes.
Exercise 9.1 (Surface of revolution). Let S be the area of the surface of revolution parametrised
by
x = r cos φ, y = r sin φ, z = z(r) (9.1)
with r ∈ [r1 , r2 ]. Show that
Z r2 p
S = 2π r 1 + (z 0 (r))2 dr (9.2)
r1

and find the minimal surface with z(ri ) = zi for i = 1, 2.

6
Exercise 9.2 (Fundamental lemma of the calculus of variations). Let F([t1 , t2 ]) be the space
of smooth functions on the interval [t1 , t2 ], t1 < t2 . Let f ∈ F([t1 , t2 ]) such that
Z t2
f (t)δq(t)dt = 0 (9.3)
t1

for every function δq ∈ F([t1 , t2 ]) such that δq(t1 ) = δq(t2 ) = 0. Show that f ≡ 0. Hint:
show the existence of bump functions on R.

10 Holonomic constraints
See class notes.

11 Noether theorem
Consider a time-independent Lagrangian L on the configuration space Rn . A symmetry for
L is a coordinates transformation Q(q) on Rn such that L(Q(q), Q̇(q, q̇)) = L(q, q̇). Recall
that
∂Qi
Q̇i (q, q̇) = (q)q̇j . (11.1)
∂qj
A one-parameter s ∈ R family of coordinates transformations

Q = Q(q, s) (11.2)

is called a one-parameter group of symmetries for L if

1. Qi (q, 0) = qi for all q,

2. Qi (Q(q, s1 ), s2 ) = Qi (q, s1 + s2 ) for all q, and for all s1 , s2 ∈ R,

and Q(q, s) is a symmetry for L for all s ∈ R. The infinitesimal generator X of the
one-parameter group of symmetries is the vector field defined by
∂Qi
Xi (q) := (q, 0). (11.3)
∂s
Theorem 11.1 (Noether). If the Lagrangian L admits a one-parameter group of symmetries
with infinitesimal generator X, then the Euler-Lagrange equations for L have the integral of
motion
∂L
I(q, q̇) = (q, q̇) · X(q). (11.4)
∂ q̇
Proof. By definition we have that

L(Q(q, s), Q̇(q, q̇, s)) = L(q, q̇) (11.5)

for all s ∈ R. Deriving w.r.t. s and setting s = 0 we obtain

∂L ∂L ∂ Q̇i
(q, q̇)Xi (q) + (q, q̇) (q, q̇, 0) = 0 (11.6)
∂qi ∂ q̇i ∂s
where we have used the fact that
∂Qi
(q, 0) = δij (11.7)
∂qj

7
and therefore that Q̇i (q, q̇, 0) = q̇i . We also have that the last term in (11.6) is equal to

∂ Q̇i ∂ 2 Qi ∂Xi
(q, q̇, 0) = (q, 0)q̇j = (q)q̇j . (11.8)
∂s ∂s∂qj ∂qj
Finally let us compute the total derivative w.r.t. time of I(q, q̇) along a solution q(t) of the
Lagrange equations associated to L. We obtain
 
dI d ∂L ∂L ∂Xi
= Xi + q̇j (11.9)
dt dt ∂ q̇i ∂ q̇i ∂qj
which vanishes because of (11.6) and (11.8).
Exercise 11.2 (Cyclic coordinates). Show that the conservation of the conjugate momentum
to the cyclic coordinate qj follows from the fact that the Lagrangian admits the one-parameter
group of symmetries qi 7→ qi + sδij .
Exercise 11.3 (Momentum conservation). Consider a system of particles in space in a potential
V invariant under translations in the direction of d ∈ R3 . Show the component of the total
momentum P along d, i.e. P · d, is conserved.
Exercise 11.4 (Homogeneity of space). Consider a system of particles in space in a potential
V depending only on the differences between the positions of pairs of particles. Show that
the translations in the direction of an arbitrary d ∈ R3 define a one-parameter group of
symmetries for the Lagrangian. Conclude that the total momentum P of the system is
conserved. On the contrary show that a potential that is invariant under any translation
depends only on the differences between the positions of pairs of particles.
Exercise 11.5 (Angular momentum conservation). Consider a system of particles in space in
a potential V invariant under rotations w.r.t. an axis d ∈ R3 . Show that such rotations form
a one-parameter group of symmetries for the Lagrangian with infinitesimal generator
X = (d × r1 , . . . , d × rN ). (11.10)
Conclude that the component along d of the total angular momentum is conserved.
Exercise 11.6 (Isotropy of space). Consider a system of particles in space in a potential V
depending only on the distances between pairs of particles. Show that rotations about an
arbitrary axis are symmetries for the Lagrangian. Conclude that the total angular momentum
is conserved.
Notice that we can interpret the energy conservation for a time-independent Lagrangian
as the conservation law associated to the invariance of the Lagrangian under time translations,
which is usually referred to as “homogeneity of time”.

12 Similarity
Consider a system of particles in configuration space Rn in a homogeneous potential V (r) of
degree k, i.e. such that
V (λr) = λk V (r) (12.1)
for all λ ∈ R. The Lagrangian
X1
L= mi ṙi2 − V (r) (12.2)
i
2
k
is then homogeneous under rescaling of space r 7→ λr and time t 7→ µt variables if µ = λ1− 2
and rescales as
L 7→ λk L. (12.3)

8
Proposition 12.1. If r(t) is a motion for a system of particles in a homogeneous potential
k
of degree k, then λr(λ 2 −1 t) also is.

Proof. Recall that the Euler-Lagrange equations are invariant under rescaling of the La-
grangian.
Exercise 12.2. Show that, for an homogeneous potential, if γ is an orbit of the system, then
also λγ is.
Exercise 12.3. For the case of the harmonic oscillator, k = 2, deduce that the period of any
motion is the same.
Exercise 12.4. For the case of the Kepler problem, k = −1, deduce the third Kepler law.

13 Central fields and the Kepler problem

See exercise classes. For a treatment in the framework of Newtonian mechanics see [Arn89]
§7-8.

3 Hamiltonian formalism on Rn
In this section we introduce the Hamiltonian formalism for systems with configuration space
Rn . Hamilton equations define a dynamical system on the phase space Rn × Rn , i.e. the
space with coordinates (q, p), where q ∈ Rn is the configuration of the system and p its
momentum, in terms of a function H(q, p, t).

14 Hamilton equations
Consider a system with phase space Rn × Rn with coordinates (q, p). An Hamiltonian is a
function (in general time-dependent) on the phase space

H = H(q, p, t) : Rn × Rn × R → R. (14.1)

The Hamilton equations

∂H ∂H
q̇i = , ṗi = − , i = 1, . . . , n (14.2)
∂pi ∂qi
are a system of 2n first order ODEs in t for the 2n variables qi and pi for i = 1, . . . , n. A
motion in phase space is a time-dependent path in phase space (q(t), p(t)) : R → R2n that
satisfies Hamilton equations.
It is an immediate consequence of Hamilton equations that:
Proposition 14.1. The total time derivative of the Hamiltonian H along a motion is equal
to its partial derivative w.r.t. t, i.e.
dH ∂H
= . (14.3)
dt ∂t
Therefore if the Hamiltonian H is time-independent, then it is an integral of motion.

Exercise 14.2. Consider a system of particles in a potential V (q, t). Show that Newton
equations (or Euler-Lagrange equations for the Lagrangian L(q, q̇, t) = T (q̇) − V (q, t)) are
equivalent to Hamilton equations for the Hamiltonian H(q, p, t) = T (q̇(p)) + V (q, t), where
the velocities are expressed as functions of the momenta by inverting p = ∂L
∂ q̇ .

9
15 Legendre transform
See class notes.
Exercise 15.1. Consider a smooth function L = L(q̇) defined on an interval I ⊂ R and convex
i.e. with L00 (q̇) > 0. Then its derivative p(q̇) := L0 (q̇) is a diffeomorphism (the Legendre
coordinates transformation) of I on its image J. Denote by q̇(p) its inverse.
1. Find a primitive H(p) (the Legendre transform) of q̇(p) in terms of L (use integrations
by parts).

2. Compute the second derivative of H(p) and show it is also convex.

3. Show that by applying the same procedure to H(p) one re-obtains the function L(q̇)
(the Legendre transform is an involution).
4. Show that the Young inequality is satisfied

pq̇ 6 L(q̇) + H(p) (15.1)

for all p ∈ J and q̇ ∈ I.

Hint: Consider the function F (q̇, p) = q̇p − L(q̇) and show that it has a maximum for
fixed p at q̇ = q̇(p).

Exercise 15.2. Show that the Legendre transform g(p) of f (x) is equal to the maximum
vertical distance between the graphs of the functions y = f (x) and the straight line y = px,
namely
h(p) = sup(px − f (x)). (15.2)
x

Exercise 15.3. Compute the Legendre transform g(p) of the following functions f (x), specify-
ing the domains of definition of the functions f and g. Check that the Legendre transform is
an involution.

1. f (x) = x2 ,
m 2
2. f (x) = 2x ,

xα 1 1
3. f (x) = α , α > 1, denote β the number given by α + β = 1,

4. f (x) is a convex polygonal line,

5. f (x) = ex ,
6. f (x) = x2 on the interval [2, 3].
Exercise 15.4. Let f (x, y) a smooth function with fxx > 0. Denote by g(p, y) its Legendre
transform w.r.t. the variable x, considering the variable y as a parameter. Prove that

df = pdx + vdy =⇒ dg = d(px − f ) = xdp − vdy. (15.3)

Give precise definitions of the symbols p, v and x that appear in the previous equations.
Exercise 15.5. Compute the Legendre transform g(p), p ∈ Rn of the quadratic form f (x) =
n
P
i,j ij i xj for x ∈ R , with (fij ) invertible.
f x
Exercise 15.6. Show that the results of Exercise 15.1 extend to the case of a smooth function
L(q̇) defined on Rn with positive definite Hessian matrix.

10
16 Equivalence of Euler-Lagrange and Hamiltonian equations
Given a Lagrangian L = L(q, q̇, t) : R2n+1 → R we call Legendre coordinate transform
the map
∂L
(q, q̇) 7→ (q, p) where p = . (16.1)
∂ q̇
Since the Jacobian of such map is
!
1 0
J = ∂2L ∂2L , (16.2)
∂q∂ q̇ ∂ q̇∂ q̇

the Legendre coordinate transform for a regular Lagrangian is a local diffeomorphism.

We say that a Lagrangian L is hyperregular if the Legendre coordinate transform is a
diffeomorphism. The Legendre transform of L is
H(q, p, t) = p · q̇(q, p, t) − L(q, q̇(q, p, t), t) (16.3)
where q̇(q, p, t) is the inverse to the Legendre coordinate transform.
Theorem 16.1. If a Lagrangian L is hyperregular, the Euler-Lagrange equations are equiv-
alent to the Hamilton equations with Hamiltonian H given by the Legendre transform of
L.
Remark 16.2. By equivalent we mean that if q(t) is a solution of the Euler-Lagrange equations
and p(t) is defined by
∂L
p(t) = (q(t), q̇(t), t), (16.4)
∂ q̇
then they satisfy Hamilton equations. Viceversa, given a solution (q(t), p(t)) of the Hamilton
equations, then q(t) solves Euler-Lagrange equations.
Proof. Let q(t) be a solution to Euler-Lagrange equations. First prove that from the definition
of Legendre transform it follows that
∂H ∂H ∂L
(q, p, t) = q̇i (q, p, t), (q, p, t) = − (q, q̇(q, p, t), t). (16.5)
∂pi ∂qi ∂qi
Then, prove that substitution of the solution q(t) and p(t) = p(q(t), q̇(t), t) gives the Hamilton
equations. In a similar way, one can prove the converse. See lecture notes for details.
Exercise 16.3. Show that an hyperregular Lagrangian is regular.
Exercise 16.4. Show that the Lagrangian of a system of particles in a potential
1X
L= mi q̇i2 − V (q, t) (16.6)
2 i
is hyperregular.
Exercise 16.5. Show that, as a consequence of Legendre transform we have that
∂H ∂L
(q, p, t) = − (q, q̇(q, p, t), t). (16.7)
∂t ∂t
Exercise 16.6. Consider a Lagrangian of the form L = T − V where
1
T = aij (q, t)q̇i q̇j , V = V (q, t). (16.8)
2
Show that
H=T + V. (16.9)
q̇=q̇(q,p,t)

Exercise 16.7. Write the Hamiltonian of a particle in e/m field.

11
17 Variational formulation of Hamilton equations
Let us denote by x = (q, p) the coordinates on the phase space R2n . Let us consider the
functional on the space of time-dependent paths γ = x(t) : [t0 , t1 ] → R2n in phase space
Z t1
S[γ] = f (x(t), ẋ(t), t)dt (17.1)
t0

where f = f (x, ẋ, t) is a function on R4n × R. Define a variation γs of γ to be a smooth

curve in the space of paths in phase space, namely a smooth function

γs = x(t, s) : [t0 , t1 ] × [−, ] → R2n (17.2)

for some  > 0.

Exercise 17.1. 1. Complete the following formula:
Z t1  
dS[γs ] ∂f d ∂x t
= − (...) · dt + [..]t10 . (17.3)
ds s=0 t0 ∂x dt ∂s s=0

2. Choosing f (x, ẋ, t) = p · q̇ − H(q, p, t), compute the boundary term.

3. Discuss what boundary conditions on γ and on the variation γs are required to obtain
Hamilton equations.

18 Poisson brackets
Given two smooth functions f, g ∈ F(R2n ) of the variables q and p, define their Poisson
bracket as the function in F(R2n ) given by
n  
X ∂f ∂g ∂g ∂f
{f, g} = − . (18.1)
i=1
∂qi ∂pi ∂qi ∂pi

We can rewrite the Hamilton equations in the symmetric form

q̇i = {qi , H}, ṗi = {pi , H} (18.2)

and more generally we have that, for any function f = f (q, p, t), its total derivative along a
motion is
df ∂f
= {f, H} + , (18.3)
dt ∂t
as one can easily verify using Hamilton equations. Therefore we have that:

Proposition 18.1. The function f is an integral of motion iff {f, H} + ∂f ∂t = 0 and, in

particular, if f does not depend on t, it is an integral of motion iff {f, H} = 0.
Two functions f , g such that {f, g} = 0 are said to be in involution. The Poisson
bracket endows F(R2n ) with the structure of Lie algebra, i.e. it is bilinear, skewsymmetric
and satisfies the Jacobi identity

{f, {g, h}} + {g, {h, f }} + {h, {f, g}} = 0 (18.4)

for any f, g, h ∈ F(R2n ). Moreover it satisfies the Leibniz rule in both its entries

{f g, h} = f {g, h} + g{f, h}. (18.5)

12
Exercise 18.2. Prove that the Poisson bracket satisfies the Jacobi identity and the Leibniz
rule.
As a consequence of the Jacobi identity, we have the Poisson theorem:
Proposition 18.3. If f and g are integrals of motion, then also {f, g} is.
In other words the integrals of motion for the dynamical system defined by an Hamiltonian
H form a Lie subalgebra of (F(R2n ), {, }).
Exercise 18.4. Prove the previous proposition, including the case when both f and g are
time dependent.
Exercise 18.5. Let pi and Li with i = 1, 2, 3 be the components of the momentum and of the
angular momentum of a particle in space R3 . Show that

{pi , Lj } = ijk pk , {Li , Lj } = ijk Lk ,

where ijk is the Levi-Civita symbol, i.e., antisymmetric in any exchange of two indices, with
123 = 1. You might want to use the identity

ijk ilm = δjl δkm − δjm δkl . (18.6)

Show that if any two components of the angular momentum are conserved, then the third is.

19 Cyclic coordinates
See class notes.
Exercise 19.1. Considering the system with the following Lagrangian in polar coordinates
m 2  k
L= ṙ + r2 θ̇2 + n . (19.1)
2 r
Find the associated Hamiltonian H = H(r, pθ , pr ). Observing that the variable θ is cyclic,
write the reduced system for the variables r and pr . Write θ(t) in terms of r(t).
Exercise 19.2. Consider two particles with mass mi , i = 1, 2 on a plane interacting via a
potential that depends only on their distance. Show that the Lagrangian can be written as
1 1
L= M Ṙ2 + µṙ2 − V (19.2)
2 2
where M is the total mass, µ the reduced mass, r = r1 − r2 and R = m m2
M r1 + M r2 . Rewrite
1

the Lagrangian in polar coordinates and find the Hamiltonian. Observe that the presence of
three cyclic coordinates allows to reduce the system from eight to two degrees of freedom.

20 Symplectic and Hamiltonian matrices

A matrix A ∈ M2n (R) is called symplectic if AT πA = π where
 
0 I
π= . (20.1)
−I 0

Notice that π 2 = −I, π −1 = π T = −π. The symplectic matrices in M2n (R) form a matrix
Lie group denoted Sp(n, R).
The Lie algebra sp(n, R) of Sp(n, R) is given by the Hamiltonian matrices, namely
matrices B ∈ M2n (R) that satisfy

B T π + πB = 0. (20.2)

13
Exercise 20.1. Prove that symplectic matrices form a subgroup Sp(n, R) of Gl(2n, R).
Exercise 20.2. Compute the inverse of a symplectic matrix A.
Exercise 20.3. Show that a matrix is symplectic iff its transpose is.
Exercise 20.4. Prove if A ∈ Sp(n, R) then det A = ±1. One actually always has det A = +1,
but this identity is more difficult to prove. (Look up “Pfaffians” for a hint.)
Exercise 20.5. Derive (20.2) by considering a symplectic matrix close to the identity, namely
A = I + B + o(2 ).
Exercise 20.6. Show that B ∈ M2n (R) is Hamiltonian iff B = πS for a symmetric matrix S.
 
a b
Exercise 20.7. Show that B = for a, b, c, d ∈ Mn (R) is Hamiltonian iff b, c are
c d
symmetric and a = −dT .
Exercise 20.8. Let B ∈ M2n (R). Prove that etB ∈ Sp(n, R) for t ∈ R iff B ∈ sp(n, R).
Exercise 20.9. Consider the symplectic product ω(x, y) := xT πy of two vectors x, y ∈ R2n .
Show that it is skew-symmetric and bilinear. Show that for n = 1 it gives the (oriented)
area of the parallelogram defined by the two vectors. Can we interpret in a similar way the
symplectic product of two vectors in R2n ?
Exercise 20.10. Show that a 2 × 2 matrix is symplectic iff it has determinant equal to one.

21 Hamiltonian vector fields

Notice that in the phase space variables x = (q, p) ∈ R2n the Poisson brackets are written

∂f ∂g
{f, g} = πij (21.1)
∂xi ∂xj

and Hamilton equations have the form

∂H
ẋi = {xi , H} = πij . (21.2)
∂xj

A vector field X(x, t), i.e. a map R2n × R → R2n , is Hamiltonian if

∂H
X=π (21.3)
∂x
for a function H(x, t) on R2n × R.
∂X
Proposition 21.1. A vector field X is Hamiltonian iff its Jacobian matrix J(x, t) = ∂x
belongs to sp(n, R), i.e. it is an Hamiltonian matrix for all (x, t).
Proof. See class notes.
Exercise 21.2. Consider the system of differential equations

ṗ = −pα+1 q δ , q̇ = pα q β . (21.4)

For which values of the real constants α, β and δ does it define a Hamiltonian vector field ?
Find the corresponding Hamiltonian.

14
Exercise 21.3. Let X be a vector field on R2n seen as a derivation of F(R2n )
X ∂f
X(f ) = Xi . (21.5)
i
∂xi

Recall that the space of vector fields on R2n is a Lie algebra w.r.t. the commutator defined
by
[X, Y ](f ) = X(Y (f )) − Y (X(f )) (21.6)
for all f ∈ F(R2n ). Let XH be the Hamiltonian vector field associated the Hamiltonian H,
i.e.
XH (f ) = {f, H}. (21.7)
Using the Jacobi identity of the Poisson bracket, prove that

X{H,K} = −[XH , XK ] (21.8)

and conclude that the space of Hamiltonian vector fields is a Lie subalgebra.

22 Generating functions for canonical transformations

See class notes.
Exercise 22.1. Find the coordinate transformation associated to the generating function
Q
F =− (22.1)
q
and prove explicitly that it preserves the canonical form of Hamilton equations.
Exercise 22.2. Prove that the coordinate transformation
 2 
p q p
Q = log , P =− +1
q 2 q
preserves the canonical form of Hamilton equations by showing that it is obtained via a
generating function. Prove again the same fact by explicit change of variables in Hamilton
equations.
Exercise 22.3. Consider the Hamiltonian of the harmonic oscillator
p2 q2
H= +k , k = mω 2
2m 2
Consider the coordinate transformation
√ α √
p = α 2mP cos Q, q= 2mP sin Q.
mω
For which values of the constant α are they canonical? Find the Hamiltonian and the
Hamilton equations in the new coordinates Q, P . Hint: compute p/q, find a generating
function F (q, Q).
Exercise 22.4. Consider the time-dependent coordinate transformation
1
Qi = a(t)qi , Pi = pi (22.2)
a(t)
where a(t) is a non-vanishing function. Find a generating function for this coordinate
transformation, find the new Hamiltonian in terms of the old one. Show explicitly that it
preserves the canonical form of Hamilton’s equations.

15
23 Canonical transformations
See class notes.

24 Hamilton-Jacobi
See class notes.

25 Separation of variables
See class notes.

4 Lagrangian mechanics on manifolds

We have seen that the Euler-Lagrange equations are covariant under changes of coordinates
in configuration space. This allows to define Lagrangian mechanics in the more general
setting where the configuration space is a manifold. For simplicity we will only consider
time-independent Lagrangians.

26 Smooth manifolds
We will work with real manifolds which are smooth i.e. of class C ∞ . We denote by F(Q) the
algebra of smooth functions on the manifold Q. Some useful references to review the basic
material in differential geometry are [Lee03] and [War83].

27 The tangent bundle

Given a manifold Q, recall the definition of the tangent space Tq Q at q ∈ Q, of the tangent
bundle T Q, and of the tangent coordinates (q, q̇) on T Q associated to coordinates q on Q.

28 Euler-Lagrange equations on a manifold

Let the configuration space be given by an n-dimensional manifold Q. A Lagrangian L is a
function on the tangent bundle T Q

L : T Q → R. (28.1)

A time dependent path γ is a map

γ:R→Q (28.2)
that associates to each time t the configuration of the system γ(t) in Q. At each time t the
tangent vector γ̇(t) belongs to Tγ(t) Q, hence it defines a time dependent path γ̃ in T Q called
the lift of γ:
γ̃ : R → T Q. (28.3)
In a coordinate chart Ũ on T Q with coordinates (q, q̇) the Lagrangian function is repre-
sented by a function L(q, q̇), a time dependent path γ by a function q(t), and the lift of γ by
a function (q(t), q̇(t)).
We say that a path γ in Q satisfies the Euler-Lagrange equations associated with the
Lagrangian L if its coordinate representation q(t) in any chart satisfies the Euler-Lagrange
equations in the usual sense.
Exercise 28.1. Show that this definition does not depend on the chart chosen.

16
29 The Lagrangian vector field
A Lagrangian L : T Q → R is regular if in any coordinate chart (q, q̇) the Hessian matrix
∂2L
∂ q̇∂ q̇ is non-degenerate.
For a regular Lagrangian L on T Q in local coordinates (q, q̇) the formula
!
 
Zq q̇
=   −1   (29.1)
Zq̇ ∂2L ∂L ∂2L
∂ q̇∂ q̇ ∂q − ∂ q̇∂q q̇

defines a vector field ZL on T Q, called Lagrangian vector field associated with the
Lagrangian L.
Proposition 29.1. A time dependent path γ satisfies the Euler-Lagrange equations for L
iff its lift γ̃ is an integral curve of ZL .
Exercise 29.2. Verify that the previous statements are valid and in particular that they are
independent of the choice of coordinates.

5 Hamiltonian mechanics on manifolds

30 The cotangent bundle
Recall the definition of cotangent bundle T ∗ Q associated to a smooth manifold Q of dimension
n, with its natural projection π : T ∗ Q → Q, which is in particular a smooth surjective map
(and a submersion) between smooth manifolds. The cotangent bundle is a vector bundle
over Q of rank n. To any choice of local coordinates q on Q, namely to any choice of chart
φ : U → Rn on Q, we can associate cotangent coordinates (q, p) on T ∗ Q, namely a chart
φ̃ : Ũ := π −1 (U ) → R2n on T ∗ Q. This construction provides a distinguished atlas on T ∗ Q.
For a reference, see for example Ch. 4 in [Lee03].

31 The Poisson bracket on the cotangent bundle

Let Q be a smooth manifold and T ∗ Q its cotangent bundle.
We define the Poisson bracket of two functions f, g ∈ F(T ∗ Q) as
n  
X ∂f ∂g ∂g ∂f
{f, g} = − , (31.1)
i=1
∂q i ∂pi ∂q i ∂pi

where (q, p) are cotangent coordinates associated with a choice of coordinates q on U ⊂ Q.

The Poisson bracket gives a skew-symmetric bilinear map
{, } : F(T ∗ Q) × F(T ∗ Q) → F(T ∗ Q) (31.2)
which satisfies the Jacobi identity and the Leibniz rule in each entry.
Exercise 31.1. Prove that the definition of Poisson bracket of two functions f and g does not
depend on the choice of local coordinates q, hence {f, g} is a well-defined smooth function in
F(T ∗ Q).
Denoting x = (q, p) the cotangent coordinates, we can write
2n
X ∂f ∂g
{f, g} = π ij , (31.3)
i,j=1
∂xi ∂xj

where π is the skew-symmetric matrix (20.1).

Hamiltonian vector fields can be defined exactly as we did over Rn . However their
definition makes sense in larger setting, that we consider in the following.

17
32 Poisson manifolds
A Poisson manifold is a manifold P endowed with a Poisson bracket, namely a skewsym-
metric bilinear map
{, } : F(P ) × F(P ) → F(P ) (32.1)
that satisfies the Jacobi identity

{{f, g}, h} + {{g, h}, f } + {{h, f }, g} = 0 (32.2)

and the Leibniz rule with respect to the product of functions

{f g, h} = f {g, h} + g{f, h} (32.3)

for all f, g, h ∈ F(P ).

We have seen in §31 that the cotangent bundle T ∗ Q is a Poisson manifold.
The Leibniz rules ensures that the Poisson bracket is a derivation of F(P ) w.r.t. to
both entries, which in turn means that the Poisson bracket defines a bivector, namely a
contravariant skew-symmetric tensor π, called Poisson bivector. In coordinates x we have
that
n
X ∂f ∂g
{f, g}(x) = π ij (x) i j , (32.4)
i=1
∂x ∂x

for a skew-symmetric matrix π ij (x) which depends smoothly on x. The Jacobi identity is
equivalent to the following nonlinear PDE for π(x)

∂π jk ∂π ki ∂π ij
π il l
+ π jl l
+ π kl = 0. (32.5)
∂x ∂x ∂xl
Notice that we do not require any assumption about nondegeneracy of the Poisson bracket.

33 Hamiltonian vector fields

Let P be a Poisson manifold.
The Hamiltonian vector field XH on P is associated to the Hamiltonian H ∈ F(P )
via
XH (f ) = {f, H} (33.1)
for all f ∈ F(P ). Clearly this formula defines a derivation of F(P ), which is the same as a
vector field on P . In coordinates x we have that the components of the vector field XH are
given by
∂H
XH (xi ) = {xi , H} = π ij (x) j . (33.2)
∂x
Notice that π ij in general depends on x. When P = T ∗ Q we get the usual form of the
Hamiltonian vector field, where the Poisson bivector is actually constant and equal to the
standard skew-symmetric matrix π.
The map that associates to a function H ∈ F(P ) the vector field XH is a anti-
homomorphism of Lie algebras, namely

X{f,g} = −[Xf , Xg ] (33.3)

for any f, g ∈ F(P ).

Exercise 33.1. Prove that (33.3) follows from the Jacobi identity of {, }.

18
34 Canonical transformations
Let P be a Poisson manifold.
A canonical transformation is a diffeomorphism Φ : P → P such that

{Φ∗ f, Φ∗ g} = Φ∗ {f, g} (34.1)

for any f, g ∈ F(P ).

Exercise 34.1. A diffeomorphism φ : Q → Q on a smooth manifold Q induces a diffeomorphism
Φ : T ∗ Q → T ∗ Q on the Poisson manifold T ∗ Q by pulling back of forms with φ−1 . Write
the explicit form of Φ in cotangent coordinates (q, p) and prove that it is a canonical
transformation.

35 The fibre derivative

Recall that a Lagrangian on the configuration space given by a manifold Q is a smooth
function on the tangent bundle T Q

L : T Q → R. (35.1)

Given a Lagrangian L ∈ F(T Q), the fibre derivative F L associates to a point vq ∈ T Q

an element in the cotangent bundle F L(vq ) ∈ Tq∗ Q defined by

d
< F L(vq ), wq >= L(vq + swq )|s=0 (35.2)
ds
for all wq ∈ Tq Q.
The energy is the function E : T Q → R defined by

E(vq ) =< F L(vq ), vq > −L(vq ) (35.3)

for vq ∈ T Q.
Proposition 35.1. The energy E is conserved along any solution of the Euler-Lagrange
equations.
A Lagrangian is hyperregular if the map LF : T Q → T ∗ Q is a diffeomorphism.
Let L be a hyperregular Lagrangian. The Hamiltonian function H : T ∗ Q → R
corresponding to L is defined by
H := E ◦ F L−1 . (35.4)
We have the following diagram:

Ro / T ∗Q /R
FL H
TQ (35.5)
L
E
π
π
~  |
R Q

Exercise 35.2. Check that these definitions, when spelled out in coordinates correspond to
the definitions given in the previous sections.
Recall that the hyperregular Lagrangian L defines a vector field ZL on T Q, corresponding
to the Euler-Lagrange equations, and the Hamiltonian H defines an Hamiltonian vector
field XH on T ∗ Q. As we have seen in the Rn case, the two vector fields coincide, or more
precisely:

19
Proposition 35.3. The push-forward by F L of the Lagrangian vector field ZL coincides
with the Hamiltonian vector field XH , namely

XH = F L∗ (ZL ). (35.6)

Remark 35.4. By an analogous construction one can start with an hyperregular Hamiltonian
H ∈ F(T ∗ Q) and define an equivalent Lagrangian vector field on T Q.

36 The symplectic structure on the cotangent bundle

See class notes.
Let us first introduce the notion of tautological4 one-form on the cotangent bundle T ∗ Q
of a manifold Q.
Let Q be a smooth manifold and T ∗ Q its cotangent bundle. Let us define a 1-form η (i.e.
a smooth section of the cotangent bundle T ∗ (T ∗ Q) → T ∗ Q) by
n
X
η= pi dq i (36.1)
i=1

where (q, p) are cotangent coordinates. This definition in local coordinates is actually invariant
under change of coordinate chart on Q and defines a smooth one-form on T ∗ Q called the
tautological one-form.
Exercise 36.1. Prove that the previous definition is independent of the choice of coordinates
on Q and that it indeed defines a one-form η on T ∗ Q.
A coordinate-free definition can also be given as in the following exercise.
Exercise 36.2. Show that η at the point m ∈ T ∗ Q with q = π(m) is given by ηm = m ◦ dπm
where dπ : T (T ∗ Q) → T Q is the tangent map associated to π : T ∗ Q → Q and m is seen as a
linear map m : Tq Q → R.
The name "tautological" is justified by the following equivalent characterisation:
Proposition 36.3. η is the unique one-form on T ∗ Q such that α∗ η = α for any one form
on Q.
Exercise 36.4. Prove the previous proposition. For a proof, see e.g. [LM87] p. 61.
It follows that ω = dη is a closed 2-form on T ∗ Q called the canonical symplectic form
on T ∗ Q. In coordinates x = (q, p) we have
2n
1 X
ω = dpi ∧ dq i = − πij dxi ∧ dxj , (36.2)
2 i,j=1

where π is the standard skew symmetry matrix defined in (20.1).

37 Symplectic manifolds
See class notes. General definition of symplectic manifold and of symplectomorphism. The
cotangent bundle is a symplectic manifold.

38 Poisson and symplectic manifolds

See class notes. Symplectic manifolds are the same as non-degenerate Poisson manifolds.
Symplectomorphisms coincide with canonical transformations.
4 In the literature this is sometimes referred to also as Liouville form, see for example [LM87].

20
39 Liouville theorem
See class notes. The symplectic form defines a canonical volume form on any symplectic man-
ifold. Canonical transformations preserve the volume. Poincaré-Cartan integral invariants.
The Hamiltonian flow preserves the volume in phase space. Poincaré recurrence theorem.

40 Darboux theorem
See class notes. Local form of the Poisson brackets of constant rank. Darboux theorem. Local
symplectic foliation of a Poisson manifold with Poisson bracket of constant rank.

41 Poisson structure on the dual of a Lie algebra

See class notes. Definition of the Lie-Poisson bracket (also called Konstant-Kirillov Poisson
bracket) on the dual of a Lie algebra. Symplectic leaves correspond to the coadjoint orbit of
the Lie group.

42 Completely integrable systems

See class notes. Definition of completely integrable Hamiltonian system. Liouville-Arnold
theorem.

References
[Arn89] Vladimir I Arnold, Mathematical methods of classical mechanics, second ed., Grad-
uate Texts in Mathematics, vol. 60, Springer-Verlag, New York, 1989.
[Arn92] Vladimir I. Arnold, Ordinary Differential Equations, Universitext, Springer-Verlag,
1992.

[GPS01] Herbert Goldstein, Charles Poole, and John Safko, Classical Mechanics, third ed.,
Addison-Wesley, 2001.
[Lee03] John M Lee, Introduction to smooth manifolds, Graduate Texts in Mathematics,
vol. 218, Springer-Verlag, New York, New York, NY, 2003.
[LL69] L. D. Landau and E. M. Lifshitz, Mechanics, second ed., Course of Theoretical
Physics, vol. 1, 1969.
[LM87] Paulette Libermann and Charles-Michel Marle, Symplectic Geometry and Analytical
Mechanics, Springer Netherlands, Dordrecht, 1987.
[Ton04] David Tong, Classical Dynamics, 2004.

[Ton13] , Dynamics and Relativity, 2013.

[War83] Frank W Warner, Foundations of differentiable manifolds and Lie groups, Graduate
Texts in Mathematics, vol. 94, Springer-Verlag, New York-Berlin, 1983.

21