Chap04 PDF

Chapter 4
Lagrangian mechanics
Motivated by discussions of the variational principle in the previous chapter,

together with the insights of special relativity and the principle of equivalence
in finding the motions of free particles and particles in uniform gravitational
fields, we seek now a variational principle for the motion of nonrelativistic
particles subject to arbitrary forces. This will lead us to introduce Hamil-
ton’s principle and the Lagrangian to describe physical systems in me-
chanics, both for single particles and systems of particles. These concepts
together are so elegant that we are encouraged to place them at the very
heart of classical mechanics. We are further encouraged to do so in the fol-
lowing chapter, the capstone chapter to Part I of the book, where we show
how they naturally emerge as we take the classical limit of the vastly more
comprehensive theory of quantum mechanics.
4.1 The Lagrangian in Cartesian coordinates

At the end of Chapter 3 we reached the intriguing conclusion that the correct
equations of motion for a nonrelativistic particle of mass m in a uniform
gravitational field can be found by making stationary the functional
Z Z
1 2
I → dt m v − U = dt (T − U ), (4.1)
2
where
1
T ≡ m v2 (4.2)
2
139
4.1. THE LAGRANGIAN IN CARTESIAN COORDINATES
is the particle’s kinetic energy and
U = mgy (4.3)
is its gravitational potential energy. It was the difference between the kinetic
and gravitational potential energy that was needed in the integrand.
Now suppose that a particle is subject to an arbitrary conservative force
for which a potential energy U can be defined. Does the form
Z Z
1 2
I → dt m v − U = dt (T − U ) (4.4)
2
still work? Do we still get the correct F = ma equations of motion?
Let us do a quick check using Cartesian coordinates. Note that if U =
U (x, y, z) and T = T (ẋ, ẏ, ż), then the integrand in the variational problem,
which we now denote by the letter L, is
1
L(x, y, z, ẋ, ẏ, ż) ≡ T (ẋ, ẏ, ż) − U (x, y, z) = mv 2 − U (x, y, z), (4.5)
2
where v 2 = ẋ2 + ẏ 2 + ż 2 . Writing out the three associated Euler equations,
we get the differential equations of motion
∂L d ∂L ∂L d ∂L ∂L d ∂L
− = 0, − = 0, − = 0, (4.6)
∂x dt ∂ ẋ ∂y dt ∂ ẏ ∂z dt ∂ ż
where
∂L ∂U d ∂L d
=− = Fx and = mẋ = mẍ, so Fx = mẍ, (4.7)
∂x ∂x dt ∂ ẋ dt
with similar results in the y and z directions. That is, we have derived the
three components of F = ma,
Fx = mẍ, Fy = mÿ, Fz = mz̈ ⇒ −∇U = ma . (4.8)
The quantity
L=T −U (4.9)
is called the Lagrangian of the particle. As we have seen, using the La-
grangian as the integrand in the variational problem gives us the correct
equations of motion, at least in Cartesian coordinates, for any conservative
force!
140
CHAPTER 4. LAGRANGIAN MECHANICS
4.2 Hamilton’s principle

We now have an interesting proposal at hand: reformulate the equations of
motion of nonrelativistic mechanics, F = dp/dt, in terms of a variational
principle making stationary a certain functional. This has two benefits:
(1) It is an interesting and intuitive idea to think of dynamics as arising
from making a certain physical quantity stationary; we will appreciate some
of these aspects in due time, especially when we get to the chapter on the
connections between classical and quantum mechanics;
(2) This reformulation provides powerful computational tools that can allow
one to solve complex mechanics problems with greater ease. The formalism
also lends itself more transparently to implementations in computer algo-
rithms.
The Lagrange technique makes brilliant use of what are called generalized
coordinates, particularly when the particle or particles are subject to one
or more constraints. Suppose that a particular particle is free to move in
all three dimensions, so three coordinates are needed to specify its position.
The coordinates might be Cartesian (x, y, z), cylindrical (r, ϕ, z), spherical
(r, θ, ϕ), as illustrated in Figure 4.1, or they might be any other complete set
of three (not necessarily orthogonal) coordinates.1
A different particle may be less free: it might be constrained to move
on a tabletop, or along a wire, or within the confines of a closed box, for
example. Sometimes the presence of a constraint means that fewer than
three coordinates are required to specify the position of the particle. So
if the particle is restricted to slide on the surface of a table, for example,
only two coordinates are needed. Or if the particle is a bead sliding along a
frictionless wire, only one coordinate is needed, say the distance of the bead
from a given point on the wire. On the other hand, if the particle is confined
to move within a closed three-dimensional box, the constraint does not reduce
the number of coordinates required: we still need three coordinates to specify
1
Note that in spherical coordinates the radius r is the distance from the origin, while in
cylindrical coordinates r is the distance from the vertical (z) axis. Because these r’s refer to
different distances, some people use ρ instead of r in cylindrical coordinates to distinguish
it from the r in spherical coordinates. However, retaining the symbol r in cylindrical
coordinates has the great advantage that on any z = constant plane the coordinates (r, ϕ)
automatically become a good choice for conventional planar polar coordinates.
141
4.2. HAMILTON’S PRINCIPLE
Cartesian Cylindrical Spherical
FIGURE 4.1 : Cartesian, cylindrical, and spherical coordinates
the position of the particle inside the box.

A constraint that reduces the number of coordinates needed to specify
the position of a particle is called a holonomic constraint. The requirement
that a particle move anywhere on a tabletop is a holonomic constraint, for
example, because the minimum set of required coordinates is lowered from
three to two, from (say) (x, y, z) to (x, y). The requirement that a bead
move on a wire in the shape of a helix is a holonomic constraint, because the
minimum set of required coordinates is lowered from three to one, from (say)
cylindrical coordinates (r, ϕ, z) to just z. The requirement that a particle
remain within a closed box is nonholonomic, because a requirement that
x1 ≤ x ≤ x2 , y1 ≤ y ≤ y2 , z1 ≤ z ≤ z2 does not reduce the number of
coordinates required to locate the particle.
For an unconstrained particle, three coordinates are needed; or if there is
a holonomic constraint the number of coordinates is reduced to two or one.
We call a minimal set of required coordinates generalized coordinates
and denote them by qk , where k = 1, 2, 3 for a single particle (or k = 1, 2,
or just k = 1, for a constrained particle). For each generalized coordinate
there is a generalized velocity q̇k = dqk /dt. Note that a generalized velocity
does not necessarily have the dimensions of length/time, just as a generalized
coordinate does not necessarily have the dimensions of length. For example,
the polar angle θ in spherical coordinates is dimensionless, and its generalized
velocity θ̇ has dimensions of inverse time.
142
Having chosen a set of generalized coordinates qk for a particle, the inte-

grand L in the variational problem, where L is called the Lagrangian, can
be written2
L = L(t, q1 , q2 , .., q̇1 , q̇2 , ...) = L(t, qk , q̇i ) (4.10)
in terms of the generalized coordinates, generalized velocities, and the time.

We can now present a more formal statement of the Lagrangian approach
to finding the differential equations of motion of a system.
Given a mechanical system described through N dynamical generalized
coordinates labeled qk (t), with k = 1, 2, . . . , N , we define its action S[qk (t)]
as the functional of the time integral over the Lagrangian L(t, q1 , q2 , ..., q̇1 , q̇2 , ...),
from a starting time ta to an ending time tb ,
Z tb Z tb
S[qk (t)] = dt L(t, q1 , q2 , ..., q̇1 , q̇2 , ...) ≡ dt L(t, qk , q̇k ) . (4.11)
ta ta
It is understood that the particle begins at some definite position (q1 , q2 , ...)a
at time ta and ends at some definite position (q1 , q2 , ...)b at time tb . We then
propose that, for trajectories qk (t) where S is stationary — i.e., , when
Z tb
δS = δ L(t, qk , q̇k ) dt = 0 (4.12)
ta
the qk (t)’s satisfy the equations of motion for the system with the prescribed
boundary conditions at ta and tb . This proposal was first enunciated by the
Irish mathematician and physicist William Rowan Hamilton (1805 – 1865),
and is called Hamilton’s principle.3 From Hamilton’s principle and our
discussion of the previous chapter, we get the N Lagrange equations
d ∂L ∂L
− = 0. (k = 1, 2, . . . , N ) (4.13)
dt ∂ q̇k ∂qk
2
Here, we are assuming that the Lagrangian does not involve dependence on higher
derivatives of qk , such as q̈k . It can be shown that such terms lead to differential equations
of the third or higher orders in time (See Problems section). Our goal is to reproduce
traditional Newtonian and relativistic mechanics involving differential equations that are
no higher than second order.
3
It is also sometimes called the Principle of Least Action or the Principle of Stationary
Action. This can be confusing, however, because there is an older principle called the
“principle of least action” that is quite different.
143
These then have to be the equations of motion of the system if Hamilton’s

principle is correct. Note that we need to use generalized coordinates since the
variational principle assumes that the perturbed variables in the functional
are independent.
Consider a general physical system involving only conservative forces and
a number of particles — constrained or otherwise. We propose that we
can describe the dynamics of this system fully through Hamilton’s principle,
using the Lagrangian L = T − U , the difference between the total kinetic
energy and the total potential energy of the system — written in generalized
coordinates. For a single particle under the influence of a conservative force,
and described with Cartesian coordinates, we have already shown that this
is indeed possible. The question is then whether we can extend this new
technology to more general situations with several particles, constraints, and
described with arbitrary coordinate systems. We will show this step by step,
looking at explicit examples and generalizing from there. There are three
main issues we would need to tackle in this process:
1. Does changing the coordinate system in which we express the kinetic
and potential energies generate any obstacles to the formalism? The
answer to this is “no”, since the functional we extremize – which is
the action – is a scalar quantity: its value does not change under
coordinate transformations qk → qk0
Z Z
S= dt L(t, qk , q̇k ) = dt L(t, qk0 , q̇k0 ) . (4.14)
The coordinate change simply relabels the stationary path of the func-
tional; that is, the path at the extremum transforms as qksol (t) → qk0 sol (t)
where qk0 sol (t) is the stationary path of S expressed in the new coordi-
nates. Hence, we can safely perform coordinate transformations as long
as we always write the Lagrangian as kinetic minus potential energy in
our preferred coordinate system.
2. Constraints provide for relations between the variables describing a

mechanical system, and hence reduce their number to a minimal set
of generalized coordinates. This was the premise of the variational
principle: the generalized coordinates must be independent. Hence,
no constraint would interfere with the variational principle as long as
we express L = T − U in terms of the generalized coordinates. But
144
constraints on the coordinates are due to forces in the system that

restrict the dynamics. For example, the normal force pushes upward
to make sure a block stays on the floor; likewise, the tension force in
a rope constrains the motion of a bob pendulum. Can we be certain
that these forces should not be included in the potential energy U that
appears in L? To ensure that this is the case, we need to ascertain
that such constraint forces do no work, and hence do not have any
net energetic contribution to U . This is not always easy to see. We
will demonstrate the mechanism at work through examples, and then
identify the general strategy.
3. Finally, should we expect any obstacles to the formalism when we have
more than one particle? Do we simply add the kinetic and potential
energies of all the particles? With two or more particles, shouldn’t
we worry about Newton’s third law? We will see soon that the La-
grangian formalism incorporates Newton’s third law and indeed can
handle many-particle systems very well.
The punchline of all this is that, for arbitrarily complicated systems with
many constraints and involving many particles interacting with conserva-
tive forces, the Lagrangian formalism works, and is very powerful. Newton’s
second law follows from Hamilton’s principle, and the third law arises, as we
will see, from symmetries of the Lagrangian. How about the first law? That
is indeed an important potential pitfall: one should always write the kinetic
energy and potential energy in L = T − U as seen from the perspective of an
inertial observer. This is because our contact with mechanics is through the
reproduction of Newton’s second law, F = ma — which is valid only in an
inertial frame. With this in mind, we now have a proposal that reformulates
Newtonian mechanics through a powerful new formalism.
It is important to emphasize that the Lagrangian formalism does not
introduce new physics. It is a mathematical reformulation of good old me-
chanics, non-relativistic and relativistic. What it does is give us powerful
new technical tools to tackle problems with greater ease and less work, a
deep insight into the laws of physics and how Nature ticks, and how the
classical world is linked to the quantum realm.
145
EXAMPLE 4-1: A simple pendulum
A small plumb bob of mass m is free to swing back and forth in a vertical x-z plane at
the end of a string of length R. The position of the bob can be specified uniquely by its angle
θ measured up from its equilibrium position at the bottom, so we choose θ as the generalized
coordinate. The bob’s kinetic energy is
1 1 1 1
T = m v 2 = m (ẋ2 + ż 2 ) = m (ṙ2 + r2 θ̇2 ) = m (R2 θ̇2 ) . (4.15)
2 2 2 2
Here, we switched to polar coordinates, and implemented the constraint equations ṙ = 0 and
r = R. Its potential energy is U = mgh = mgR(1 − cos θ), measuring the bob’s height h up
from its lowest point. The Lagrangian of the bob is therefore
1
L=T −U = mR2 θ̇2 − mgR(1 − cos θ). (4.16)
2
The constraint reduces the dynamics from two planar coordinates to only one, the single degree
of freedom that is the angle θ. The single Euler equation in this case is
∂L d ∂L d
− = −mgR sin θ − mR2 θ̇ = 0, (4.17)
∂θ dt ∂ θ̇ dt
equivalent to the well-known “pendulum equation”
θ̈ + (g/R) sin θ = 0. (4.18)
Note that equation (4.17)) (or (4.18)) is equivalent to τ = I θ̈, where the torque τ =
−mgR sin θ is taken about the point of suspension (negative because it is opposite to the
direction of increasing θ), and the moment of inertia of the bob is I = m R2 .
We had two twists in this problem. First, we switched from Cartesian to polar coordinates.
But we know that this is not a problem for the Lagrangian formalism since the action is a
scalar quantity. Second, we implemented a constraint r = R, implying ṙ = 0. This constraint
is responsible for holding the bob at fixed distance from the pivot and hence is due to the
tension force in the rope. By implementing the constraint, we reduced the problem from
two to only one degree of freedom. Furthermore, the tension in the rope does no work: it is
always perpendicular to the motion of the bob, and hence the work contribution T · dr = 0,
where T is the tension force and dr is the displacement of the bob. Thus, our potential energy
U — related to work done by a force — was simply the potential energy due to gravity alone,
U = mgh. In general, whenever a contact force is always perpendicular to the displacement
of the particle it is acting on, it can safely be ignored in constructing the Lagrangian.
146
EXAMPLE 4-2: A bead sliding on a vertical helix

A bead of mass m is slipped onto a frictionless wire wound in the shape of a helix of
radius R, whose symmetry axis is oriented vertically in a uniform gravitational field, as shown in
Figure 4.2. Using cylindrical coordinates r, θ, z, the base of the helix is located at z = 0, θ = 0,
and the angle θ is related to the height z at any point by θ = αz, where α is a constant with
dimensions of inverse length. The gravitational potential energy of the bead is U = mgz,
and its kinetic energy is T = (1/2)mv 2 = (1/2)m[ṙ2 + r2 θ̇2 + ż 2 ]. However, the constraint
that the bead slide along the helix tells us that the bead’s radius is constant at r = R, and
(choosing z as the single generalized coordinate), θ̇ = αż. Therefore the kinetic energy of the
bead is simply
1 1 1
T = m v 2 = m (ẋ2 + ẏ 2 + ż 2 ) = m (ṙ2 + r2 θ̇2 + ż 2 )
2 2 2
= (1/2)m[0 + α2 R2 + 1]ż 2 , (4.19)
where we switched to cylindrical coordinates and implemented the two constraints θ = αz and
r = R. So the Lagrangian of the bead is
1
L = T − U = m[1 + α2 R2 ]ż 2 − mgz (4.20)
2
in terms of the single generalized coordinate z and its generalized velocity ż. Two constraints
reduced the dynamics from three to only one degree of freedom. Note that in Newtonian
mechanics we often need to take into account the normal force of the wire on the bead as one of
the forces in F = ma; however, the normal force appears nowhere in the Lagrangian, because
it does no work on the bead in this case — it is always perpendicular to the displacement
of the bead4 . In general, whenever a normal force is perpendicular to the displacement of a
particle it is acting on, we can safely ignore it in setting up the Lagrangian.
EXAMPLE 4-3: Block on an inclined plane

A block of mass m slides down a frictionless plane tilted at angle α to the horizontal, as
shown in Figure 4.3. The gravitational potential energy is mgh = mgX sin α, where X is the
distance of the block up along the plane from its lowest point. Using X as the generalized
coordinate, the velocity is Ẋ, and the Lagrangian of the block is
4
In this simple case with a single normal force, the simplification is not very obvious.
However, in more complicated scenarios we shall see later, the advantages of dropping the
normal force from the Lagrangian — versus the Newtonian approach – will become more
apparent.
147
FIGURE 4.2 : A bead sliding on a vertically-oriented helical wire
FIGURE 4.3 : Block sliding down an inclined plane
148
1 1
L=T −U = mv 2 − U = mẊ 2 − mgX sin α, (4.21)
2 2
which depends explicitly upon a single coordinate (X) and a single velocity (Ẋ). Our single
degree of freedom is then X. The Euler equation is
∂L d ∂L d
− = −mg sin α − mẊ = 0, or − mg sin α = mẌ, (4.22)
∂X dt ∂ Ẋ dt
which is indeed the correct F = m a equation for the block along the tilted plane.
In this problem, we judiciously chose our only degree of freedom as X, the distance along
the inclined plane. We can think of this as a coordinate transformation from x,y as shown in
the Figure, to X,Y . We then have a constraint Ẏ = 0, since the block cannot tunnel into
the plane. We are then left with a single degree of freedom, X. This constraint is associated
with the normal force. Furthermore, the normal force in this problem is always perpendicular
to the displacement of the block and thus does no work. Once again, the Lagrangian formal-
ism demonstrates its elegance and power by dropping from the computation a force – and a
corresponding equation – and reducing the effective number of degrees of freedom.
4.3 Generalized momenta and cyclic coordi-

nates
In Cartesian coordinates the kinetic energy of a particle is T = (1/2)m(ẋ2 +
ẏ 2 + ż 2 ), whose derivatives with respect to the velocity components are
∂L/∂ ẋ = mẋ, etc., which are the components of momentum. So with gener-
alized coordinates qk , it is natural to define the generalized momenta pk
as
∂L
pk ≡ . (4.23)
∂ q̇k
In terms of pk , the Lagrange equations become simply
dpk ∂L
= . (4.24)
dt ∂qk
Now sometimes a particular coordinate ql is absent from the Lagrangian. Its
generalized velocity q̇l is present, but not ql itself. A missing coordinate is
said to be a cyclic coordinate or an ignorable coordinate.5 For any
5
Neither “cyclic” nor “ignorable” is a particularly appropriate or descriptive name for a
coordinate absent from the Lagrangian, but they are nevertheless the conventional terms.
In this book we will most often call any missing coordinate “cyclic”.
149
4.3. GENERALIZED MOMENTA AND CYCLIC COORDINATES
FIGURE 4.4 : Particle moving on a tabletop
such coordinate the Lagrange equation (4.24) tells us that the time deriva-
tive of the corresponding generalized momentum is zero, so that particular
generalized momentum is conserved.
One of the first things to notice about a Lagrangian is whether there are
any cyclic coordinates, because any such coordinate leads to a conservation
law that is also a first integral of motion. This means that the equation of
motion for that coordinate is already half solved, in that it is only a first-
order differential equation rather than the second-order differential equation
one typically gets for a noncyclic coordinate.
EXAMPLE 4-4: Particle on a tabletop, with a central force
For a particle moving in two dimensions, such as on a tabletop, it is often useful to use polar
coordinates (r, ϕ) about some origin, as shown in Figure 4.4. The kinetic energy of the particle
is
1 1 1 2
T = m v 2 = m (ẋ2 + ẏ 2 ) = m ṙ2 + (rϕ̇ ). (4.25)
2 2 2
Alternatively, we could have started in three dimensions in cylindrical coordinates with the
addition of a ż 2 , then use the constraint ż 2 = 0 that gets rid of the normal force being applied
by the table onto the particle vertically. This force does no work and can be ignored using a
constraint.
150
We will assume here that any force acting on the particle is a central force, depending upon r
alone, so the potential energy U of the particle also depends upon r alone. The Lagrangian is
therefore
1
L= m(ṙ2 + r2 ϕ̇2 ) − U (r). (4.26)
2
Our two degrees of freedom are r and ϕ. We note right away that in this case the coordinate
ϕ is cyclic, so there must be a conserved quantity
∂L
pϕ ≡ = mr2 ϕ̇, (4.27)
∂ ϕ̇
which we recognize as the angular momentum of the particle. In Lagrange’s approach,
pϕ is conserved because ϕ is a cyclic coordinate; in Newtonian mechanics, pϕ is conserved
because there is no torque on the particle, since we assumed that any force is a central force.
The various partial derivatives of L are
∂L ∂L
= mṙ = mr2 ϕ̇ (4.28)
∂ ṙ ∂ ϕ̇
∂L ∂U (r, ϕ) ∂L
= mrϕ̇2 − = 0, (4.29)
∂r ∂r ∂ϕ
so the Lagrange equations

∂L d ∂L ∂L d ∂L
− =0 and − =0 (4.30)
∂r dt ∂ ṙ ∂ϕ dt ∂ ϕ̇
become
∂U (r)
mrϕ̇2 − − mr̈ = 0 and − mrϕ̈ − 2mṙϕ̇ = 0 (4.31)
∂r
or (equivalently)
Fr = m(r̈ − rϕ̇2 ) ≡ mar and Fϕ = m(rϕ̈ + 2ṙϕ̇) = 0, (4.32)
where the radial force is Fr = −∂U/∂r and the radial and tangential accelerations are
ar = r̈ − rϕ̇2 and aϕ = rϕ̈ + 2ṙϕ̇, (4.33)
and where the tangential acceleration aϕ is zero in this case.6

In an example of Chapter 1 we found (using F = ma) the equations of motion of a particle
of mass m on the end of a spring of zero natural length and force constant k, where one end
6
Note how easy it is to get the expressions for radial and tangential accelerations in
polar coordinates using this method. They are often found in classical mechanics by
differentiating the position vector r = rr̂ twice with respect to time, which involves rather
tricky derivatives of the unit vectors r̂ and θ̂.
151
of the spring was fixed and the particle was free to move in two dimensions, as on a tabletop.
There we used Cartesian coordinates (x, y). Now we are equipped to write the equations of
motion in polar coordinates instead. Equations (4.30) with Fr = −kr and Fϕ = 0 give
−kr = m(r̈ − rϕ̇2 ) and pϕ = mr2 ϕ̇ = constant. (4.34)
That is, since the Lagrangian is independent of ϕ, we get the immediate first integral of motion
pϕ = constant. Eliminating ϕ̇ between the two equations, we find the purely radial equation
(pϕ )2
r̈ − + ω02 r = 0 (4.35)
m2 r 3
p
where ω0 = k/m is the natural frequency the spring-mass system would have if the mass were
oscillating in one dimension (which in fact it would do if the angular momentum pϕ happened
to be zero.) Note that even though the motion is generally two-dimensional, equation (4.35))
contains only r(t); we can therefore find a first integral of this equation because it has the
form of a one-dimensional F = ma equation with F = mr̈ = (p2ϕ /mr3 − mω02 r. An effective
one-dimensional potential energy can be found by setting F (r) = −dUeff (r)/dr; that is,
r r
(pϕ )2 (pϕ )2

1
Z Z
2
Ueff (r) = − F (r) dr = − 3
− mω0 r dr = 2
+ kr2 (4.36)
mr 2mr 2
plus a constant of integration, which we might as well set to zero. Therefore, a first integral
of motion is
1 2 (pϕ )2 1 1
mṙ + + kr2 = mṙ2 + Ueff = constant. (4.37)
2 2mr2 2 2
A sketch of Ueff is shown in Figure 4.5. Note that Ueff has a minimum, which is the loca-
tion of an equilibrium point (the value of r for which dUeff (r)/dr = 0 is of course also the
radius for which r̈ = 0.) If r remains at the minimum of Ueff , the mass is actually circling
the origin. The motion about this point is stable because the potential energy is a minimum
there. For small displacements from equilibrium the particle oscillates back and forth about
this equilibrium radius as it orbits the origin. In Section 4.7 we will calculate the frequency
of these oscillations. This example demonstrates the use of coordinate transformations in
the Lagrangian formalism. We already knew the formalism goes through under a coordinate
change. In this case, we see how useful it can be to use this freedom at the Lagrangian level.
EXAMPLE 4-5: The spherical pendulum
152
FIGURE 4.5 : The effective radial potential energy for a mass m moving with an effective
potential energy Ueff = (pϕ )2 /2mr2 + (1/2)kr2 for various values of pϕ , m, and k.
A ball of mass m swings on the end of an unstretchable string of length R in the presence
of a uniform gravitational field g. This is often called the “spherical pendulum”, because the
ball moves as though it were sliding on the frictionless surface of a spherical bowl. We aim to
find its equations of motion.
The ball has two degrees of freedom:
(i) It can move horizontally around a vertical axis passing through the point of support, cor-
responding to changes in its azimuthal angle ϕ. (On Earth’s surface this would correspond to
a change in longitude.)
(ii) It can also move in the polar direction, as described by the angle θ. (On Earth’s surface
this would correspond to a change in latitude.)
These angles are illustrated in Figure 4.6. In spherical coordinates, the velocity square appears
as
v 2 = ṙ2 + r2 θ̇2 + r2 sin2 θϕ̇2 ⇒ R2 θ̇2 + R2 sin2 θϕ̇2 (4.38)
using the constraint r = R, which gets rid of the tension force and gets us to two degrees of
freedom. We know once again that the tension force is always perpendicular to the displace-
ment in this problem, and hence can be thrown away through the use of a constraint. The
velocities in the θ and ϕ directions are vθ = Rθ̇ and vϕ = R sin θϕ̇, which are perpendicular
to one another. Hence, the kinetic energy becomes
1
T = m R2 (θ̇2 + sin2 θϕ̇2 ) (4.39)
2
The altitude h of the ball, measured from its lowest possible point, is h = R(1 − cos θ),
so the potential energy can be written
U = mgh = mgR(1 − cos θ). (4.40)
153
FIGURE 4.6 : Coordinates of a ball hanging on an unstretchable string
The Lagrangian is therefore

1
L=T −U = mR2 (θ̇2 + sin2 θϕ̇2 ) − mgR(1 − cos θ) . (4.41)
2
Our two degrees of freedom are then θ and ϕ. The derivatives are
∂L ∂L
= mR2 sin θ cos θϕ̇2 − mgR sin θ = mR2 θ̇ (4.42)
∂θ ∂ θ̇
∂L ∂L
=0 = mR2 sin2 θ ϕ̇. (4.43)
∂ϕ ∂ ϕ̇
The Lagrange equations become

∂L d ∂L ∂L d ∂L
− =0 and − = 0, (4.44)
∂θ dt ∂ θ̇ ∂ϕ dt ∂ ϕ̇
but before writing them out, note that ϕ is cyclic, so the corresponding generalized momentum
is
∂L
pϕ = = mR2 sin2 θ ϕ̇ = constant, (4.45)
∂ ϕ̇
which is an immediate first integral of motion. (We identify pϕ as the angular momentum
about the vertical axis.)
Now since
d
mR2 θ̇ = mR2 θ̈, (4.46)
dt
154
FIGURE 4.7 : A sketch of the effective potential energy Ueff for a spherical pendulum. A ball
at the minimum of Ueff is circling the vertical axis passing through the point of suspension, at
constant θ. The fact that there is a potential energy minimum at some angle θ0 means that if
disturbed from this value the ball will oscillate back and forth about θ0 as it orbits the vertical
axis.
the θ equation can be written
mR2 θ̈ = mR2 sin θ cos θϕ̇2 − mgR sin θ, (4.47)
so
g
θ̈ − sin θ cos θ ϕ̇2 + sin θ = 0. (4.48)
R
We can eliminate the ϕ̇2 term using ϕ̇ = pϕ /(mR2 sin2 θ), to give
p 2 cos θ g
ϕ
θ̈ − 2 3 + sin θ = 0, (4.49)
mR sin θ R
a second-order differential equation for the polar angle θ as a function of time.

To make further progress, do we have to tackle this differential equation head-on? Not if
we can find a first integral instead! In fact, we have already identified one first integral, the
conservation of angular momentum
mR2 sin2 θ ϕ̇ = pϕ = constant (4.50)
about the vertical axis. Another first integral is energy conservation
1
E =T +U = mR2 (θ̇2 + sin2 θϕ̇2 ) + mgR(1 − cos θ), (4.51)
2
155
4.4. SYSTEMS OF PARTICLES
valid because no work is being done on the ball aside from the work done by gravity, which is
already accounted for in the potential energy. By combining the two conservation laws we can
eliminate ϕ̇:
1 (pϕ )2
E= mR2 θ̇2 + + mgR(1 − cos θ) (4.52)
2 2mR2 sin2 θ
or
1
E= mR2 θ̇2 + Ueff (4.53)
2
where the “effective potential energy” is
(pϕ )2
Ueff = + mgR(1 − cos θ). (4.54)
2mR2 sin2 θ
This effective potential energy Ueff is sketched in Figure 4.7; it includes the terms in E
that depend only on position. The second term is the actual gravitational potential energy,
while the first term is really a piece of the kinetic energy that has become a function of position
only, thanks to angular momentum conservation.
Equations (4.50)) and (4.51 form a first-order differential equation that can be reduced
to an integral by solving it for θ̇ and then separating the variables t and θ and integrating both
sides. The result is, if we choose t = 0 when θ = θ0 ,
r
mR2 θ dθ
Z
t(θ) = q . (4.55)
2 θ0 (E − mgR) − (p )2 /(2mR2 sin2 θ) + mgR cos θ
ϕ
Once again, the constraint reduce the number of degrees of freedom, from three to two in this
case. And the associated tension constraint force does no work since it is perpendicular to the
trajectory; and hence we need not include its contribution to the Lagragian.
4.4 Systems of particles

So far we have been thinking about the motion of single particles only, de-
scribed by at most three generalized coordinates and three generalized ve-
locities. But often we want to find the motion of systems of particles, in
which two or more particles may interact with one another, like two blocks
on opposite ends of a spring, or several stars orbiting around one another.
Can we still use the variational approach, by writing down a Lagrangian that
contains the total kinetic energy and the total potential energy of the entire
system?
156
FIGURE 4.8 : Two interacting beads on a one-dimensional frictionless rail. The interaction
between the particles depends only on the distance between them.
EXAMPLE 4-6: Two interacting particles
Consider a system of two particles, with masses m1 and m2 , confined to move along
a horizontal frictionless rail. Figure 4.8 shows a picture of the setup, where we label the
coordinates of the particles x1 and x2 . We can then write an action for the system

1 1
Z
S = dt m1 ẋ21 + m2 ẋ22 − U (x2 − x1 ) (4.56)
2 2
where, in additional to the usual kinetic energy terms, there is some unknown interaction
between the particles described by a potential U (x2 − x1 ). Note that we use the total kinetic
energy, and we assume that the potential – hence the associated force law – depends only
upon the distance between the particles. We then have two equations of motion with two
generalized coordinates, x1 and x2 , so that

d ∂L ∂L ∂U
− = 0 ⇒ m1 ẍ1 = − (4.57)
dt ∂ ẋ1 ∂x1 ∂x1
and

d ∂L ∂L ∂U ∂U
− = 0 ⇒ m2 ẍ2 = − =+ (4.58)
dt ∂ ẋ2 ∂x2 ∂x2 ∂x1
where in the last step we used the fact that U = U (x2 − x1 ). We see that we have a second
law of Newton for each of the two particles: kinetic energy is additive and each of its terms
will generally give the ma part of Newton’s second law for the corresponding particle. Hence,
157
in multi-particle systems, we need to consider the total kinetic energy T minus the total
potential energy. Terms that mix the variables of different particles, such as U (x2 − x1 ), will
give the correct forces on the particles as well. In this case, we see that the action-reaction
pair, ∂U/∂x1 = −∂U/∂x2 , comes out for free, and arises from the fact that the force law
depends only on the distance between the particles! That is, Newton’s third law is naturally
incorporated in the formalism and originates from the fact that forces between two particles
depend only upon the distance between the interacting entities, and not (say) their absolute
positions.
Suppose for example that the particles are connected by a Hooke’s-law spring of force
constant k. If we choose the coordinates x1 and x2 appropriately, the spring stretch will be
x2 − x1 , so the potential energy is U = (1/2)k(x2 − x1 )2 . The Lagrange equations then give
m1 ẍ1 = +k(x2 − x1 ) and m2 ẍ2 = −k(x2 − x1 ). (4.59)
The forces on the two particles are obviously equal but opposite: In such a case the total
momentum of the systems must be conserved, which is easily verified simply by adding the
two equations, to show that
d
(m1 ẋ1 + m2 ẋ2 ) = 0. (4.60)
dt
There are actually two conserved quantities in this problem, the momentum and the energy,
each of which leads to a first integral of motion.
These results suggest that there must be a more transparent set of generalized coordinates
to use here, in which one of the new coordinates is cyclic, so that its generalized momentum
will be conserved automatically. These new coordinates are the center of mass and relative
coordinates
m1 x1 + m2 x2
X≡ and x ≡ x2 − x1 , (4.61)
M
where M = m1 + m2 is the total mass of the system: Note that X and x are simply linear
combinations of x1 and x2 . Then in terms of X and x, it is straightforward to show that the
Lagrangian of the system becomes
1 1
L= M Ẋ 2 + µẋ2 − U (x) (4.62)
2 2
where µ ≡ m1 m2 /M is called the reduced mass of the system (note that µ is in fact smaller
than either m1 or m2 .) Using this Lagrangian, it is obvious that the center of mass coordinate
X is cyclic, so the corresponding momentum
∂L
P = = M Ẋ ≡ m1 ẋ1 + m2 ẋ2 (4.63)
∂ Ẋ
is conserved (as we saw before).
This problem is an example of reducing a two-body problem to an equivalent one-body
problem through a coordinate transformation. The motion of the center of mass of the system
is trivial: the center of mass just drifts along at constant velocity. The interesting motion of
158
FIGURE 4.9 : A contraption of pulleys. We want to find the accelerations of all three weights.
We assume that the pulleys have negligible mass so they have negligible kinetic and potential
energies.
the particles is their relative motion x, which behaves as though it were a single particle of
mass µ and position x(t) subject to the potential energy U (x) with Lagrangian
1 2
L= µẋ − U (x) . (4.64)
2
Hence, coordinate transforming at the Lagrangian level can be very powerful technique.
EXAMPLE 4-7: Pulleys everywhere

Another classic set of mechanics problems involves pulleys, lots of pulleys. Consider the
setup shown in Figure 4.9. Two weights, with masses m1 and m2 , hang on the outside of
a three-pulley system, while a weight of mass M hangs on the middle pulley. We assume the
pulleys and the connecting rope have negligible mass, so their kinetic and potential energies
are also negligible. We will suppose for now that all three pulleys have the same radius R, but
this will turn out to be of no importance. We want to find the accelerations of m1 , m2 , and
M . We construct a Cartesian coordinate system as shown in the figure, which is at rest in an
inertial frame of the ground. First of all, note that there are three massive objects moving in
two dimensions, so we might think that we have six variables to track, x1 and y1 for weight
m1 , x2 and y2 for weight m2 , and X and Y for weight M . We can then write the total kinetic
energy
1 1 1
T = m1 (ẋ21 + ẏ12 ) + m2 (ẋ22 + ẏ22 ) + M (Ẋ 2 + Ẏ 2 ). (4.65)
2 2 2
159
But that is obviously overkill. We know the dynamics is entirely vertical, so we can focus on
y1 , y2 , and Y only and set ẋ1 = ẋ2 = Ẋ = 0. But that is still too much. There are only two
degrees of freedom in this problem! Just pick any two of y1 , y2 , or Y , and we can draw the
figure uniquely, as long as we know the length of the rope. Another way of saying this is to
write
Length of rope = (H − y1 ) + 2 (H − Y ) + (H − y2 ) + 3πR, (4.66)
where H is the height of the ceiling, as shown in the figure. We can therefore write in general
y1 + 2 Y + y2 = constant , (4.67)
which can be used to eliminate one of our three variables. We choose to get rid of Y , using
y1 + y2
Y =− + constant, (4.68)
2
which implies
ẏ1 + ẏ2
Ẏ = − . (4.69)
2
We can now write our kinetic energy in terms of two variables only, y1 and y2 ,
2
1 2 1 2 1 ẏ1 + ẏ2
T = m1 ẏ1 + m2 ẏ2 + M . (4.70)
2 2 2 2
We next need the potential energy, which is entirely gravitational. We can write
U = m1 g y1 + m2 g y2 + M g Y

y1 + y2
= m1 g y1 + m2 g y2 − M g + constant (4.71)
2
where the zero of the potential was chosen at the ground, and we can also drop the addi-
tive constant term since it does not affect the equations of motion. In summary we have a
variational problem with a Lagrangian
2
1 1 1 ẏ1 + ẏ2
L = T − U = m1 ẏ12 + m2 ẏ22 + M
2 2 2 2

y1 + y2
− m1 g y1 − m2 g y2 + M g . (4.72)
2
There are two dependent variables y1 and y2 , so we have two equations of motion,

d ∂L ∂L M Mg
− = 0 ⇒ m1 ÿ1 + (ÿ1 + ÿ2 ) = −m1 g + (4.73)
dt ∂ ẏ1 ∂y1 4 2
and

d ∂L ∂L M Mg
− = 0 ⇒ m2 ÿ2 + (ÿ1 + ÿ2 ) = −m2 g + . (4.74)
dt ∂ ẏ2 ∂y2 4 2
160
We can now solve for ÿ1 and ÿ2 ,

4 m2 g
ÿ1 = −g +
m1 + m2 + 4 m1 m2 /M
4 m1 g
ÿ2 = −g + . (4.75)
m1 + m2 + 4 m1 m2 /M
Note that these accelerations have magnitudes less than g, as we might expect intuitively. We
can also find Ÿ from (4.69)
ÿ1 + ÿ2 2 (m1 + m2 )g
Ÿ = − ⇒ Ÿ = g − . (4.76)
2 m1 + m2 + 4 m1 m2 /M
The constraint given by (4.66), which eliminated one of our three original variables, imple-
ments the physical condition that the rope has constant length. This is related to the tension
force in the rope. The astute reader may rightfully wonder whether we have overlooked some-
thing in this treatment: we never encountered the tension force of the rope on each of the
masses! Consider the two tension forces T1 and T2 at the end of this (or any) massless rope. If
we wanted to account for such forces in a Lagrangian, we would need the associated energy, or
work they contribute to the system. Since the rope has zero mass, we know that |T1 | = |T2 |.
The two tension forces however point in opposite directions. When one end of the rope moves
by ∆x1 > 0 parallel to T1 , T1 does work W1 = |T1 |∆x1 . At the same time, the other end must
move the same distance ∆x2 = ∆x1 . However, at this other end, the tension force points
opposite to the displacement, and the work is W2 = −|T2 |∆x2 = −|T1 |∆x1 . The total work
is W = W1 + W2 = 0. Hence, the tension forces along a massless rope will always contribute
zero work, and hence cannot be associated with energy in the Lagrangian. Similarly, this is
the case for any force that appears in a problem in an action-reaction pair, as we shall see next.
EXAMPLE 4-8: A block on a movable inclined plane
Let us return to the classic problem of a block sliding down a frictionless inclined plane, as
in Example 2, except we will make things a bit more interesting: now the inclined plane itself
is allowed to move! Figure 4.10(a) shows the setup. A block of mass m rests on an inclined
plane of mass M : Both the block and the inclined plane are free to move without friction.
The plane’s angle is denoted by α. The problem is to find the acceleration of the block.
The observation deck is the ground, which is taken as an inertial reference frame. We
set up a convenient set of Cartesian coordinates, as shown in the figure. The origin is shifted
to the top of the incline at zero time to make the geometry easier to analyze. We start by
identifying the degrees of freedom. At first, we can think of the block and inclined plane as
161
(a) (b)
FIGURE 4.10 : A block slides along an inclined plane. Both block and inclined plane are free
to move along frictionless surfaces.
moving in the two dimensions of the problem. The block’s coordinates could be denoted by
x and y, and the inclined plane’s coordinates by X and Y . But we quickly realize that this
would be overkill: if we specify X, and how far down the top of the incline the block is located
(denoted by D in the figure), we can draw the figure uniquely. This is because the inclined
plane cannot move vertically, either jumping off the ground or burrowing into it (that’s one
condition), so Y is unnecessary, and the horizontal position x of the block is determined by X
and D (that’s a second condition.) We then start with four coordinates, add two conditions
or restrictions, and we are left with two degrees of freedom. The choice of the two remaining
degrees of freedom is arbitrary, as long as the choice uniquely fixes the geometry. We will pick
X and D; another choice might be X and x, for example.
Next, we need to write the Lagrangian. The starting point for this is the total kinetic
energy of the system,
1 1
T = m ẋ2 + ẏ 2 + M Ẋ 2 + Ẏ 2 (4.77)
2 2
in the inertial frame of the ground. Note the importance of writing the kinetic energy in an
inertial frame, even if it means using more coordinates than the generalized coordinates that
will be used in the Lagrangian.
Now we need to rewrite the kinetic energy in terms of the two degrees of freedom X and
D alone. This requires a little bit of geometry. Looking back at the figure, we can write
Y =0, x = X + D cos α , y = −D sin α . (4.78)
This implies
Ẏ = 0 , ẋ = Ẋ + Ḋ cos α , ẏ = −Ḋ sin α . (4.79)
162
We can now substitute these into (4.77) and get

1 1 1
T = M Ẋ 2 + m Ẋ 2 + m Ḋ2 + m Ẋ Ḋ cos α . (4.80)
2 2 2
Note that this result, in terms of the generalized coordinates and velocities, would have been
very difficult to guess, especially the Ẋ Ḋ term. Again, it is very important to start by writing
the kinetic energy first in an inertial frame, and often important as well to use Cartesian
coordinates in this initial expression, to be confident that it has been done correctly. We now
need the potential energy of the system, which is entirely gravitational. The inclined plane’s
potential energy does not change. Since it is a constant, we need not add it to the Lagrangian:
the Lagrange equations of motion involve partial derivatives of L and, hence, a constant term
in L is irrelevant to the dynamics. The block’s potential energy on the other hand does change.
We can choose the zero of the potential at the origin of our coordinate system and write
U = m g y = −m g D sin α . (4.81)
The Lagrangian is now

1 1 1
L=T −U = M Ẋ 2 + m Ẋ 2 + m Ḋ2 + m Ẋ Ḋ cos α + m g D sin α (4.82)
2 2 2
We observe immediately that X is cyclic, so its corresponding momentum is conserved; also
the total energy is conserved. Therefore we can obtain the complete set of two first integrals
of motion.
Nevertheless, to illustrate a different approach, we will tackle the full second-order differ-
ential equations of motion obtained directly from the Lagrange equations. Since we have two
degrees of freedom X and D, we’ll have two second-order equations. The equation for X is

d ∂L ∂L
− = 0 ⇒ (m + M )Ẍ + m D̈ cos α = 0, (4.83)
dt ∂ Ẋ ∂X
and the equation for D is

d ∂L ∂L
− = 0 ⇒ m D̈ + m Ẍ cos α = m g sin α. (4.84)
dt ∂ Ḋ ∂D
This is a system of two linear equations in two unknowns Ẍ and D̈. The solution is
g cos2 α sin α g sin α cos α
Ẍ = D̈ = − . (4.85)
(1 + M/m) sin2 α + M/m sin2 α + M/m

Since we want the acceleration of the block in our inertial reference frame, we need to find
ẍ ≡ ax and ÿ ≡ ay . Differentiating (4.79) with respect to time, we get
ax = ẍ = Ẍ + D̈ cos α , ay = ÿ = −D̈ sin α. (4.86)
Substituting our solution from (4.85) into these, we have
(M/m)g cos2 α sin α g sin2 α cos α

ax = − ay = . (4.87)
(1 + M/m) sin2 α + M/m sin2 α + M/m

163
4.5. THE HAMILTONIAN
It is always useful to look at various limiting cases, to see if a result makes sense. For example,
what if α = 0, i.e., what if the block moves on a horizontal plane? Both accelerations then
vanish, as expected: if started at rest, both block and incline just stay put. Now what if the
inclined plane is much heavier than the block, with M m? We then have
m m
ax ' − g cos2 α sin α ay ' g sin2 α cos α, (4.88)
M M
so that ay /ax ' − tan α, which is what we would expect if the inclined plane were not moving
appreciably.
The computational step at the beginning where we zeroed onto the degrees of freedom
of the problem – going from X, Y , x, and y to X and D – is associated with the normal
forces. The most impressive aspect of this example is the absence of any normal forces from
our computations! With the traditional approach of problem solving, we would need to include
several normal forces in the computation, shown in Figure 4.10(b): the normal force exerted by
the inclined plane on the block, the normal force exerted by the ground on the inclined plane,
and the normal force exerted by the block on the inclined plane as a reaction force. The role
of these normal forces is to hold the inclined plane on the ground and to hold the block on
the inclined plane. If we think of the contributions of the normal forces to the Lagrangian, we
would want to include some potential energy terms for them. But potential energy is related
to work done by forces. The normal force is often perpendicular to the direction of motion,
and hence does no work, N · ∆r = 0. Hence, there is no potential energy term to include in
the Lagrangian to account for such normal forces. In our example, this is not entirely correct...
While it is true for the normal force exerted by the ground onto the inclined plane, it is not
true for the normal forces acting between the block and inclined plane. This is because the
inclined plane is moving as well and the trajectory of the block is not parallel to the incline!
However, there is another reason why this normal force is safely left out of the Lagrangian
method. These normal forces occur as an action-reaction pair. And the displacement of the
interface between the block and incline is the same for both forces, and hence the contributions
to the total work or energy of the system from these two normal forces cancel. As we saw
from the previous example, such forces do not appear in the Lagrangian.
In a later chapter we will also learn of a way to impose the inclusion of normal and ten-
sion forces in a Lagrangian even when we need not do so — for the purpose of finding the
magnitude of a normal force if it is desired. For now, we are very happy to drop normal and
tension forces from consideration. This can be a big simplification for problem solving: fewer
variables, fewer forces to consider, less work to do (no pun intended).
4.5 The Hamiltonian

We will now prove an enormously useful mathematical consequence of the
Lagrange equations, providing one more potential way to achieve a first in-
tegral of motion, which we can add to the arsenal of approaches already
164
summarized in the preceding section. First, take the total derivative of the
Lagrangian L(t, qk , q̇k ) with respect to time t. There are many ways in which
L can change: it can change because of explicit changes in t, and also because
of implicit changes in t due to the time dependence of one or more of the
coordinates qk (t) or velocities q̇k (t). Therefore, from multivariable calculus,
dL(qk , q̇k , t) ∂L ∂L ∂L
= + q̇k + q̈k , (4.89)
dt ∂t ∂qk ∂ q̇k
using the Einstein summation convention from Chapter 2. That is, since the
index k is repeated in each of the last two terms, a sum over k is implied in
each term; we have also used the fact that dq̇k /dt ≡ q̈k . Now, take the time
derivative of the quantity q̇k (∂L/∂ q̇k ); again, sums over k are implied:

d ∂L ∂L d ∂L ∂L ∂L
q̇k = q̈k + q̇k = q̈k + q̇k (4.90)
dt ∂ q̇k ∂ q̇k dt ∂ q̇k ∂ q̇k ∂qk
using the product rule. We have also used the Lagrange equations to simplify
the second term on the right. Note that this expression contains the same
two summed terms that we found in equation (4.89)). Therefore, subtracting
equation (4.90) from equation (4.89 gives

∂L d ∂L
− L − q̇k = 0. (4.91)
∂t dt ∂ q̇k
This result is particularly interesting if L is not an explicit function of time,

i.e., , if ∂L/∂t = 0. In fact, define the Hamiltonian H of a particle to be
H ≡ q̇k pk − L (4.92)
where we have already defined the generalized momenta to be pi = ∂L/∂ q̇k ,

and again a sum over k is implied. Then equation (4.91) can be written
∂L dH
=− , (4.93)
∂t dt
an extremely useful result! It shows that if a Lagrangian L is not an explicit
function of time, then the Hamiltonian H is conserved.
What is the meaning of H? Suppose that our particle is free to move
in three dimensions in a potential U (x, y, z) without constraints, and that
165
P
we are using Cartesian coordinates. Then px = mẋ, etc., so i q̇k pk =
m(ẋ2 + ẏ 2 + ż 2 ). Therefore,
1
H = m(ẋ2 + ẏ 2 + ż 2 ) − m(ẋ2 + ẏ 2 + ż 2 ) + U (x, y, z)
2
1
= m(ẋ2 + ẏ 2 + ż 2 ) + U (x, y, z) = T + U = E, (4.94)
2
which is simply the energy of the particle!
Is H always equal to E = T + U ? The answer is no, although very often
it is. The precise conditions for which H 6= E are derived in Appendix A.
EXAMPLE 4-9: Bead on a rotating parabolic wire
Suppose we bend a wire into the shape of a vertically oriented parabola defined by z = αr2 ,
as illustrated in Figure 4.11, where z is the vertical coordinate and r is the distance of a point
on the wire from the vertical axis of symmetry. Using a synchronous motor, we can force the
wire to spin at constant angular velocity ω about its symmetry axis. Then we slip a bead of
mass m onto the wire and find its equation of motion, assuming that it slides without friction
along the wire.
We first have to choose generalized coordinate(s) for the bead. The bead moves in three
dimensions, but because of the constraint we need only a single generalized coordinate to
specify the bead’s position. For example, if we know the distance r of the bead from the
vertical axis of symmetry, we also know its altitude z, because it is constrained to move along
the parabolic wire. And the bead also has no freedom to choose its azimuthal angle, because
the synchronous motor turns the wire around at a constant rate, so given its angle ϕ0 at time
t = 0, its angle at other times is constrained to be ϕ = ϕ0 + ωt. So it is convenient to
choose the cylindrical coordinate r as the generalized coordinate, although we could equally
well choose the vertical coordinate z.
In cylindrical coordinates the square of the bead’s velocity is the sum of squares of the
velocities in the r, ϕ, and z directions,
v 2 = ṙ2 + r2 ϕ̇2 + ż 2 = ṙ2 + r2 ω 2 + (2αrṙ)2 (4.95)
since ϕ̇ = ω = constant, and z = αr2 for the parabolic wire. The gravitational potential
energy is U = mgz = mgαr2 , so the Lagrangian is
1
L=T −U = m[(1 + 4α2 r2 )ṙ2 + r2 ω 2 ] − mgαr2 . (4.96)
2
We implemented two constraints, ϕ̇ = ω and z = αr2 , and thus reduced the problem from
three variables to one degree of freedom r. The contact force that keeps the bead on the wire
166
FIGURE 4.11 : A bead slides without friction on a vertically-oriented parabolic wire that is
forced to spin about its axis of symmetry.
is the normal force associated with these two constraints. And we thus have chosen not to
include its contribution to the Lagrangian assuming that this normal force has no contribution
to the energy of the system and can be packaged into the two constraints safely. . . That this
is correct to do is far from obvious in this setup since this normal force does work! This work
is associated with the energy input by the motor to keep the wire turning at constant rate ω.
Let us proceed anyways with this arrangement, and revisit the issue at the end of the section
— hopefully justifying the absence of contributions from the normal force to our Lagrangian.
The partial derivatives ∂L/∂ ṙ and ∂L/∂r are easy to find, leading to the Lagrange equation
∂L d ∂L d
− = m[4α2 rṙ2 + rω 2 − 2gαr] − m (1 + 4α2 r2 )ṙ = 0 (4.97)
∂r dt ∂r dt
which results finally in a second-order differential equation of motion
(1 + 4α2 r2 )r̈ + 4α2 rṙ2 + (2gα − ω 2 )r = 0. (4.98)
Are we stuck with having to solve this second-order differential equation? Are there no
first integrals of motion? The coordinate r is not cyclic, so pr is not conserved. However, note
that L is not an explicit function of time, so the Hamiltonian H is conserved! Conservation
of H — which we will refer to as energy by definition — will give us a first integral of motion,
so we are rescued; we do not have to solve the second-order differential equation (4.98) after
all.
The generalized momentum is pr = ∂L/∂ ṙ = m(1 + rα2 r2 )ṙ, so the Hamiltonian is
1
H = ṙpr − L = m(1 + 4α2 r2 )ṙ2 − m[(1 + 4α2 r2 )ṙ2 + r2 ω 2 ] + mgαr2
2
1
= m[(1 + 4α r )ṙ − r ω ] + mgαr2 = constant,
2 2 2 2 2
(4.99)
2
167
FIGURE 4.12 : The effective potential Ueff for the Hamiltonian of a bead on a rotating
parabolic wire with z = αr2 , depending
√ upon whether the angular velocity ω is less than,
greater than, or equal to ωcrit = 2g α.
which differs from the energy

1
E =T +U = m[(1 + 4α2 r2 )ṙ2 + r2 ω 2 ] + mgαr2 (4.100)
2
by H − E = −mr2 ω 2 .
Equation (4.99) is a first-order differential equation, which can be reduced to quadrature.
Without going that far, we can understand a good deal about the motion just by using equa-
tion (4.99), and noting that it has a similar mathematical form to that of energy-conservation
equations. That is, rewrite the equation as
1
H= m(1 + 4α2 r2 )ṙ2 + Ueff , (4.101)
2
where
1 1
Ueff = m(−r2 ω 2 ) + mgαr2 = mr2 (2gα2 − ω 2 ). (4.102)
2 2
This effective potential is quadratic in r with the interesting feature that everything depends
upon how the angular velocity ω of the wire is related to a critical angular velocity ωcrit ≡
(2g)1/2 α, as illustrated in Figure 4.12. If ω < ωcrit , then Ueff rises with r, so the bead is
stable at r = 0, the potential minimum. But if ω > ωcrit , then Ueff falls off with increasing
r, so r = 0 is an unstable equilibrium point in that case; if the bead wanders even slightly
from r = 0 at the bottom of the parabola it will be thrown out indefinitely far. The stability
is neutral if ω = ωcrit , meaning that if placed at rest at any point along the wire the bead
will stay at that point indefinitely, but if placed at any point and pushed outward it will keep
moving outward, or if pushed inward it will keep moving inward.
168
This example shows that although the Hamiltonian function H is often equal to E ≡ T +U ,
this is not always so. Appendix A at the end of the chapter explains when and why they can
differ. In any case, the Hamiltonian can be very useful, because it provides a first integral of
motion if L is not explicitly time-dependent. It is also the starting point for an alternative
approach to classical dynamics, as we will see in a later chapter, and it turns out to be an
important bridge between classical and quantum mechanics7 .
Let us come back to the issue of dropping the normal force’s contribution to our La-
grangian. Since the bead is sliding along the wire while rotating with it, and since the normal
force is some vector perpendicular to the wire at any instant in time, we can see that this
normal force is not necessarily perpendicular to the bead’s displacement. Hence, it can have
non-zero contribution to the energy of the system. This is why E is not conserved in the
problem; the conserved quantity is H which is not equal to E in this case. Nevertheless, how
can we justify dropping the normal force that we know can do work and hence may perhaps
have a piece of U in L = T − U ?
The answer is a delicate one: in constructing a Lagrangian, we need to include all objects
that are moving around. The full system is bead plus rotating parabolic wire. The parabolic
wire is not a free dynamical object since its motion is prescribed externally through the action
of the motor. You can think of such a non-dynamical object as one with zero mass; i.e., , it
has no contribution to the kinetic energy of the system. In reality, it has a constant non-zero
contribution, but a constant term in a Lagrangian does not effect the equations of motion.
Hence, for simplicity, we can drop such a term altogether. However, forces can still act on it
and do work. In this case, there is a normal force — equal but opposite in direction to the one
acting on the bead — acting on the wire. The point of action of this normal on the wire is
displaced by the same amount as the bead. Hence, the contribution to the work of the system
of this normal force is equal in magnitude but opposite in sign to the contribution coming from
the normal force acting on the bead. This is simply a reflection of Newton’s third law: for
every action, there is an equal but opposite reaction. The net contribution to work from these
two normal forces adds up to zero. Hence, when we write U for the system, we only need to
consider the contribution from gravity! We have L = T − U as before, without a trace of the
normal forces.
Once again, the Lagrangian formalism avoids dealing with contact forces and accounts for
them through constraints – simplifying the problem significantly. We leave it as an exercise
for the reader to solve this same problem using traditional force body diagram methods so as
to appreciate the power of the Lagrangian formalism.
7
In a different approach we will see in Chapter 6, conserved quantities are defined
through associated symmetries in Nature. Within such a convention, the Hamiltonian
would be defined as energy, and the combination T + U loses its privileged name and
place.
169
4.6. THE MORAL OF CONSTRAINTS
4.6 The moral of constraints

Let us summarize the steps we have used so far in setting up and preparing
to solve Lagrange’s equations.
1. We identify the degrees of freedom of each particle or object consis-
tent with any constraints, and choose an appropriate set of generalized
coordinates qk for each.
2. Write the square of the velocity for each particle in terms of any conve-
nient coordinates, usually Cartesian coordinates, in some inertial ref-
erence frame. Then reexpress the kinetic energy in terms of the gen-
eralized coordinates qk , the generalized velocities q̇k , and the time if
needed; i.e., , v 2 = v 2 (qk , q̇k , t). Then write the total kinetic energy T
in terms of these v’s.
3. Write the total potential energy in the form U = U (qk , t). Do not
include any contributions from constraint forces.
4. Write the Lagrangian L(t, q1 , q2 , ..., q̇1 , q̇2 , ...) = T − U
5. Identify any cyclic coordinates in L; that is, identify any coordinate
ql missing in the Lagrangian, even though its corresponding general-
ized velocity q̇l is present. In this case the corresponding generalized
momentum pl ≡ ∂L/∂ q̇l is conserved. This gives a highly-valued first
integral of motion, i.e., , a differential equation that is first order rather
than second order.
6. If there are more generalized coordinates in the problem than first
integrals identified in the preceding steps, then one or more of the
Lagrange equations of motion
∂L d ∂L
− = 0. (4.103)
∂qk dt ∂ q̇k
must be used as well, to obtain a complete set of differential equations.

That is, if there are N generalized coordinates, we will generally need
N mutually independent differential equations whose solutions will give
the coordinates as functions of time. Some of these may be first-order
equations, each corresponding to a conserved quantity, while others
may be second-order equations.
170
All these steps but one were rigorously justified: that the Lagrangian
L = T − U accounts for all conservative forces, that coordinate transforma-
tions at the Lagrangian level are justified and in fact very useful, and that
constraints assure that the generalized coordinates are independent and a
reduced set of degrees of freedom describes the dynamics fully. But what
about dropping constraint forces from the Lagrangian? We saw example
after example that this works out fine. But now we see the emerging pat-
tern in general. Constraint forces implement restrictions on the dynamics
between two objects in contact. If both objects in question are part of the
dynamical system (i.e., , they both have kinetic energy contributions to the
Lagrangian), we know that these constraint forces must come in equal and
opposite pairs. Since the contact point is the same, this always implies that
such forces will not do work and hence need not appear in the Lagrangian.
On the other hand, if only one of the two objects is part of the dynamical
system, the other one must then have prescribed time evolution by definition:
i.e., , the ground just sits there as a function of time, the pivot of the pen-
dulum is fixed in position, a parabolic wire – on which a bead is sliding –
is rotating at a given constant angular speed driven by a motor. In some of
these cases, the constraint forces do no work because they are perpendicular
to the displacement. But it is easy to see this statement in more generality:
Extend the Lagrangian to include the non-dynamical system – the ground,
the parabolic wire connected to a motor – by adding their constant kinetic
energies to the Lagrangian. Then the constraint force becomes part of an
internal action-reaction pair, which we know does not contribute to the La-
grangian! And the cost of adding the kinetic energy of the external agent to
the Lagrangian is irrelevant: it is a constant shift to the Lagrangian since the
relevant dynamics is, by definition, prescribed. This, we now see one of the
central advantages of the Lagrangian formalism: drop all constraint forces
from the outset!
As one gets used to the steps outlined above, many stages of this algorith-
mic process become second nature and can be done mentally. With practice,
you may be able to stare at a complex mechanical system, write down the
Lagrangian immediately on a single line, and in a few more lines, write the
equations of motion! To get there however, one needs to first practice the
steps summarized here ad nauseum, problem after problem.
171
4.7. SMALL OSCILLATIONS ABOUT EQUILIBRIUM
4.7 Small oscillations about equilibrium

When we look around at many common mechanical systems, we find that
energy is often approximately conserved and the system is in a more or less
stable equilibrium state. For example, a nearby chair is resting on the floor,
happily doing nothing as expected from a chair. It is in its minimal energy
configuration. If we bump it, it wobbles a bit for some time, and then quickly
finds itself again at rest in some new equilibrium state. When we bumped
the chair, we added energy to the system, and the chair eventually dissipated
this energy through friction (sound, heat, etc...) and found another minimal-
energy motionless state.
In general, most mechanical systems can be accorded an energy of the
form
Constant × q̇k2 + Ueff (q) = E (4.104)
where the qk ’s are the generalized coordinates and Ueff is an effective poten-
tial. We saw this in example after example in this chapter. For simplicity,
imagine we have only one such coordinate we’ll call qk . If the effective po-
tential energy Ueff has an extremum at some particular point (qk )0 , then
that point is an equilibrium point of the motion, so if placed at rest at (qk )0
the particle will stay there. If (qk )0 happens to be a minimum of Ueff , as
illustrated in Figure 4.13, (qk )0 is a stable equilbrium point, so that if the
particle is displaced slightly from (qk )0 it will oscillate back and forth, never
wandering far from that point.
It is sometimes interesting to find the frequency of small oscillations about
an equilibrium point. We can do this by fitting the bottom of the effective
energy curve to a parabolic bowl, because that is the shape of the potential
energy of a simple harmonic oscillator. That is, by the Taylor expansion
dU 1 d2 U
U (x) = U (x0 ) + |x0 (x − x0 ) + 2
|x0 (x − x0 )2 + ... (4.105)
dx 2! dx
So if x0 is the equilibrium point, by definition the second term vanishes,
and the third term has the form (1/2)keff (x − x0 )2 , like that for a simple
harmonic oscillator with center at x0 , where the effective force constant is
given by keff = U 00 (x0 ). The frequency of small oscillations is therefore
p p
ω = keff /m = U 00 (x0 )/m. (4.106)
172
FIGURE 4.13 : An effective potential energy Ueff with a focus near a minimum. Such a
point is a stable equilibrium point. The dotted parabola shows the leading approximation to
the potential near its minimum. As the energy drains out, the system settles into its minimum
with the final moments being well approximated with harmonic oscillatory dynamics.
Note that this explains the pervasiveness of the harmonic oscillator in Nature:
since system will try to find their lowest energy configurations by dissipating
energy, they will often find themselves near the minima of their effective
potential. As we just argued, in the vicinity of such minima, systems will
generically oscillate harmonically. An example will demonstrate how this
works.
EXAMPLE 4-10: Particle on a tabletop with a central spring force

In Example 4 we considered a particle moving on a frictionless tabletop, subject to a
central Hooke’s-law spring force. There is an equilibrium radius for given energy and angular
momentum for which the particle orbits in a circle of some radius r0 . Now we can find the
oscillation frequency ω for the mass about the equilibrium radius if it is perturbed slightly from
this circular orbit.
The effective potential in Example 4 was Ueff = (pϕ )2 /2mr2 +(1/2)kr2 ; the first derivative
0
of this potential is Ueff (r) = −(pϕ )2 /mr3 + kr. The equilibrium value of r is located where
U (r) = 0; namely, where r = r0 = ((pϕ )2 /mk)1/4 . The second derivative of Ueff (r) is
0
U 00 (r) = 3(pϕ )2 /mr4 + k, (4.107)

so
3(pϕ )2
U 00 (r0 ) = + k = 3k + k = 4k. (4.108)
m((pϕ )2 /mk)
173
4.8. RELATIVISTIC GENERALIZATION
FIGURE 4.14 : The shape of the two-dimensional orbit of a particle subject to a central
spring force, for small oscillations about the equilibrium radius.
The frequency of small oscillations about the equilibrium radius r0 is therefore

p p
ω = U 00 (r0 )/m = 4k/m = 2ω0 . (4.109)
That is, for the mass orbiting the origin and subject to a central Hooke’s law spring force, the
frequency of small oscillations about a circular orbit is just twice what it would be for the mass
if it were oscillating back and forth in one dimension.
We can also find the shape of the orbits if the radial oscillations are small. The angular
frequency of rotation is ωrot = vφ /ro = (pϕ /mr0 )/r0 = pϕ /(mr02 ), where vφ is the tangential
component of velocity. But the equilibrium radius is r0 = ((pϕ )2 /mk)1/4 , so the angular
p
frequency of rotation is ωrot = pϕ /(mr02 ) = pϕ /[m (pϕ )2 /k] = k/m = ω0 , which is the
p
same as the frequency of oscillation of the system as if it were moving in one dimension!
Therefore the frequency of radial oscillations (4.109) is just twice the rotational frequency,
so the orbits for small oscillations are closed: that is, the path retraces itself in every orbit,
as shown in Figure 4.14. The small-oscillation path appears to be elliptical, and in fact it is
exactly elliptical, even for large displacements from equilibrium, as we already saw in Chapter
1 using Cartesian coordinates.
4.8 Relativistic generalization

By now, we have established the power of the Lagrangian formalism in dealing
with mechanics problems involving one or more particles, interacting with
174
conservative forces, with or without a large class of constraints. But all

this was within Newtonian mechanics — valid at speeds much less than the
speed of light. Can we use the Lagrangian formalism for situations requiring
relativistic treatment? The answer is rather simple to find. Our original setup
for deriving the Lagrangian formalism started by integrating proper time to
construct the action functional (see equation (3.67))). We then took the limit
of low speeds in (3.73 to identify a piece of the future Lagrangian — the
kinetic energy. Through the example of a particle in a uniform gravitational
field, we identified the second piece in L = T − U — the potential energy.
We then proceeded to show that this works for any conservative potential,
with one or more particles, with or without constraints. Going back to the
beginning, we then must have
r
2 v2
L = T − U = mc 1− 2 −U (4.110)
c
to tackle a fully relativistic problem. We can even still write L = T − U
using relativistic kinetic energy
r
v2
T = m c2 1 − 2 − m c2 (4.111)
c
since the additional piece m c2 is a constant and hence does not effect the
equations of motion. For a single particle in Cartesian coordinates, using our
results from equations (3.72)), the equations of motion from (4.110 now look
like
d ∂U d ∂U d ∂U
(γ ẋ) = − , (γ ẏ) = − , (γ ż) = − . (4.112)
dt ∂x dt ∂y dt ∂z
This is the expected relativistic form of Newton’s second law from (2.93))
or (2.100) if we identify F = −∇U . Hence, our entire Lagrangian formalism
extends through the relativistic regime as long as we replace kinetic energy
in L = T − U by (4.111 instead of T = (1/2)m v 2 .
There is, however, a possible new pitfall. As always, the Lagrangian
T − U is to be written from the perspective of an inertial observer. In rel-
ativistic settings, the correct transformations linking inertial observers are
the Lorentz transformations. This implies that our Lagrangian should now
be invariant under Lorentz transformations, not Galilean. The kinetic en-
ergy term (4.111)) is indeed Lorentz invariant since it arises from the integral
175
4.9. SUMMARY
over proper time (see equation (3.67)). We thus have to make sure that the
potential energy term U in the Lagrangian is also Lorentz invariant, indepen-
dently. Not any old force law is allowed! We will tackle the transformation
properties of the action in Chapter 6. And we will come back to this issue
in Chapter 8 when we encounter a full Lorentz invariant force law — the
electromagnetic force. In the meantime, it is worthwhile emphasizing that
traditional mechanics force laws, such as Newtonian gravity, are not Lorentz
invariant and hence should only be considered in approximate Newtonian
settings with L = (1/2)m v 2 − U — requiring only Galilean invariance from
the action.
4.9 Summary
In this chapter we have presented a variational approach to classical mechan-
ics, which is at the very heart of the subject. The variational approach is in
fact the central theme of this book.
We began the chapter by describing a conservative mechanical system by
N generalized coordinates qk (t), with k = 1, 2, . . . , N , and then defining the
Lagrangian
L(t, q1 , q2 , ..., q̇1 , q̇2 , ...) = T − U (4.113)
as the difference between the kinetic and potential energies of the system,
expressed in terms of the generalized coordinates qk , generalized velocities q̇k ,
and time t. We then define the action S[qk (t)] of the system as the functional
consisting of the time integral over the Lagrangian L(t, q1 , q2 , ..., q̇1 , q̇2 , ...),
from a starting time ta to an ending time tb ,
Z tb Z tb
S[qk (t)] = dt L(t, q1 , q2 , ..., q̇1 , q̇2 , ...) ≡ dt L(t, qk , q̇k ) , (4.114)
ta ta
where it is understood that the particle or system of particles begins at some

definite position (q1 , q2 , ...)a at time ta and ends at some definite position
(q1 , q2 , ...)b at time tb .
Hamilton’s principle then proposes that, for trajectories qk (t) where
the action S is stationary — i.e., when
Z tb
δS = δ L(t, qk , q̇k ) dt = 0 (4.115)
ta
176
the coordinates qk (t)’s satisfy the equations of motion for the system with
the prescribed boundary conditions at ta and tb . That is, the variational
principle δS = 0 gives the Lagrange equations
d ∂L ∂L
− = 0. (k = 1, 2, . . . , N ) (4.116)
dt ∂ q̇k ∂qk
which are the differential equations of motion of the system.
We also defined the Hamiltonian of the system to be
H ≡ q̇k pk − L (4.117)
where a sum over k is implied, and the generalized momenta pk are defined
to be
∂L
pk = . (4.118)
∂ q̇k
Then
∂L dH
=− , (4.119)
∂t dt
so if L is not an explicit function of time, it follows that the Hamiltonian is
conserved.
One of the strengths of the Lagrangian formalism is that it prescribes a
quite straightforward algorithmic process to solve a problem. Here are the
typical steps:
1. Identify the degrees of freedom of each particle or object consistent

with any constraints, and choose an appropriate set of generalized co-
ordinates qk for each.
2. Write the square of the velocity for each particle in terms of any conve-
nient coordinates, usually Cartesian coordinates, in some inertial ref-
erence frame. Then reexpress the kinetic energy in terms of the gen-
eralized coordinates qk , the generalized velocities q̇k , and the time (if
needed); i.e., , v 2 = v 2 (qk , q̇k , t). Then write the total kinetic energy T
in terms of these v’s.
3. Write the total potential energy in the form U = U (qk , t).
4. Write the Lagrangian L(t, q1 , q2 , ..., q̇1 , q̇2 , ...) = T − U
177
4.9. SUMMARY
5. Identify any cyclic coordinates in L; that is, identify any coordinate

ql missing in the Lagrangian, even though its corresponding general-
ized velocity q̇l is present. In this case the corresponding generalized
momentum pl ≡ ∂L/∂ q̇l is conserved. This gives a highly-valued first
integral of motion, i.e., , a differential equation that is first order rather
than second order.
6. If L is not an explicit function of time, then the Hamiltonian H ≡

q̇k pk − L is conserved, providing another first integral of motion. Here
the generalized momentum pi = ∂L/∂ q̇i .
7. If there are more generalized coordinates in the problem than first

integrals identified in the preceding steps, then one or more of the
Lagrange equations of motion
∂L d ∂L
− = 0. (4.120)
∂qk dt ∂ q̇k
must be used as well, to obtain a complete set of differential equations.

That is, if there are N generalized coordinates, we will generally need
N mutually independent differential equations whose solutions will give
the coordinates as functions of time. Some of these may be first-order
equations, each corresponding to a conserved quantity, while others
may be second-order equations.
If the problem is simple, exact analytic solutions of the equations may be

possible. If not, we can always solve the equations numerically on a computer.
There is also a very common intermediate situation, when one or more of the
equations is too difficult to solve exactly, but approximate analytic techniques
can be used to find the answer as accurately as required. Finding a sufficiently
accurate approximate analytic technique to solve a particular problem can
be very entertaining, requiring as much creativity as setting up the physical
problem in the first place. One of the pleasures of doing physics is to find
a way, no holds barred, to solve a problem to the extent we need it solved,
using back-of-the-envelope calculations, dimensional reasoning, approximate
analytical techniques, numerical calculations, or whatever it takes to get the
job done.
178
Problems
PROBLEM 4-1: Example 2 featured a bead sliding on a vertically-oriented helix of radius
R. The angle θ about the symmetry axis was related to its vertical coordinate z on the
wire by θ = αz. There is a uniform gravitational field g vertically downward. (a) Rewrite
the Lagrangian and find the Lagrange equation, using z as the generalized coordinate. (b)
Are there any conserved dynamical quantities? (c) Write the simplest differential equation of
motion of the bead, and go as far as you can to solve analytically for the z as a function of
time.
PROBLEM 4-2: In Example 3 we found the equation of motion of a block on an inclined

plane, using the generalized coordinate s, the distance of the block from the bottom of the
incline. Solve the equation for s(t), in terms of an arbitrary initial position s(0) and velocity
ṡ(0).
PROBLEM 4-3: A bead of mass m is placed on a vertically-oriented circular hoop of radius

R and negligible mass that is free to rotate about a vertical axis through its center. There
is a uniform downward-directed gravitational field g. (a) Using polar and azimuthal angles θ
and ϕ as generalized coordinates, find the kinetic and potential energies of the bead. (b) Find
the equations of motion using Lagrange’s equations. (c) Write down two first integrals of the
motion. What is the physical significance of each?
PROBLEM 4-4: A particle of mass m slides inside a smooth hemispherical bowl of radius R.
Beginning with spherical coordinates r, θ and ϕ to describe the dynamics, select generalized
coordinates, write the Lagrangian, and find the differential equations of motion of the particle.
PROBLEM 4-5: A particle moves in a cylindrically symmetric potential V (r, z). Use cylin-
drical coordinates r, θ, and z to parameterize the space.
(a) Write the Lagrangian for an unconstrained particle of mass m (using cylindrical coor-
dinates) in the presence of this potential.
(b) Write the Lagrange equations of motion for r, θ and z.
(c) Identify and cyclic coordinates, and write a first integral corresponding to each.
PROBLEM 4-6: A spring pendulum has the pendulum bob of mass m attached to a spring
of force-constant k and unstretched length R0 . The pendulum is constrained to swing in a
vertical plane (i.e., two degrees of freedoms are at work).
(a) Write the Lagrangian for the blob. Do not forget the potential for the spring.
(b) Write the equations of motion for θ and r. Stare at the resulting differential equations
and write me a brief thank you note for not asking you to solve them.
PROBLEM 4-7: A particle of mass m slides inside a smooth paraboloid of revolution whose
axis of symmetry z is vertical. The surface is defined by the equation z = αr2 , where z and
r are cylindrical coordinates, and α is a constant. There is a uniform gravitational field g. (a)
Select two generalized coordinates for m. (b) Find T, U, and L. (c) Identify any ignorable
179
4.9. SUMMARY
coordinates, and any conserved quantities. (d) Show that there are two first integrals of
motion, and find the corresponding equations.
PROBLEM 4-8: Repeat the preceding problem for a particle sliding inside a smooth cone
defined by z = αr.
PROBLEM 4-9: A spring pendulum features a pendulum bob of mass m attached to one
end of a spring of force-constant k and unstretched length R. The other end of the spring is
attached to a fixed point on the ceiling. The pendulum is allowed to swing in a plane. Use
r, the distance of the bob from the fixed point, and θ, the angle of the spring relative to the
vertical, as generalized coordinates. (a) Find the kinetic and potential energies of the bob in
terms of the generalized coordinates and velocities. (b) Find the Lagrangian. Are there any
ignorable coordinates? (c) Are there any conserved quantities? (d) Find a complete set of
equations of motion, including as many first integrals as possible.
PROBLEM 4-10: A pendulum is constructed from a bob of mass m on one end of a light
string of length D. The other end of string is attached to the top of a circular cylinder of
radius R (R < 2D/π). The string makes an angle θ with the vertical, as shown below, as
the pendulum swings in the plane. There is a uniform gravity g directed downward. (a) Find
the Lagrangian and write out Lagrange’s equation using θ as the generalized coordinate. (b)
Identify any first integrals of motion. (c) Find the frequency of small oscillations about the
stable equilibrium point.
PROBLEM 4-11: A particle moves with a cylindrically symmetric potential energy U =

U (r, z) where r, θ, z are cylindrical coordinates. (a) Write the Lagrangian for an unconstrained
particle of mass m in this case. (b) Are there any cyclic coordinates? If so, what symmetries
do they correspond to, and what are the resulting constants of the motion? (c) Write the
Lagrange equation for each cyclic coordinate. (d) Find the Hamiltonian H. Is it conserved?
(e) Find the total energy E. Is E = H? is E conserved? (f) Write the simplest (i.e., ,
lowest-order) complete set of differential equations of motion of the particle.
PROBLEM 4-12: A plane pendulum is made with a plumb bob of mass m hanging on a
Hooke’s-law spring of negligible mass, force constant k, and unstretched length `0 . The spring
can stretch but is not allowed to bend. There is a uniform downward gravitational field g.
(a) Select generalized coordinates for the bob, and find the Lagrangian in terms of them. (b)
Write out the Lagrange equations of motion (c) Are there any conserved quantities? If so,
write down the corresponding conservation law(s). (d) If the bob is swinging back and forth,
find the frequency of small oscillations in the general case where the spring can change its
length while the bob is swinging back and forth.
PROBLEM 4-13: Motion in a slowly-changing uniform electric field

A particle of mass m and charge q moves within a parallel-plate capacitor whose charge
Q is decays exponentially with time, Q = Q0 e−t/τ , where τ is the time constant of the decay.
Find the equations of motion of the particle.
180
PROBLEM 4-14: A particle of mass m travels between two points x = 0 and x = x1 on

Earth’s surface, leaving at time t = 0 and arriving at t = t1 . The gravitational field g is
uniform. (a) Suppose m moves along the ground (keeping altitude z = 0) at steady speed.
Find the total action S to go by this path. (b) Suppose instead that m moves along the
least-action parabolic path. Show that the action in this case is
mx21 mg 2 t31
S= − (4.121)
2t1 24
and verify that it is less than the action for the straight-line path of part (a).
PROBLEM 4-15: Suppose the particle of the preceding problem moves instead at constant
speed along a isoceles triangular path between the beginning point and the end point, with the
high point at height z1 above the ground, at x = x1 /2 and t = t1 /2. (a) Find the action for
this path. (b) Find the altitude z1 corresponding to the least-action path among this class of
constant-speed triangular paths. (c) Verify that the total action for this path is greater than
that of the parabolic path of the preceding problem.
PROBLEM 4-16: A plane pendulum consists of a light rod of length R supporting a plumb
bob of mass m in a uniform gravitational field g. The point of support of the top end of the
rod is forced to oscillate back and forth in the horizontal direction with x = A cos ωt. Using
the angle θ of the bob from the vertical as the generalized coordinate, (a) find the Lagrangian
of the plumb bob. (b) Are there any conserved dynamical quantities? (c) Find the simplest
differential equation of motion of the bob.
PROBLEM 4-17: Solve the preceding problem if instead of being forced to oscillate in the
horizontal direction, the upper end of the rod is forced to oscillate in the vertical direction with
y = A cos ωt.
PROBLEM 4-18: A particle of mass m on a frictionless table top is attached to one end
of a light string. The other end of the string is threaded through a small hole in the table
top, and held by a person under the table. If given a sideways velocity v0 , the particle circles
the hole with radius r0 . At time t = 0 the mass reaches an angle defined to be θ = 0 on
the table top, and the person under the table pulls on the string so that the length of the
string above the table becomes r(t) = r0 − αt for a period of time thereafter, where α is a
constant. Using θ as the generalized coordinate of the particle, find its Lagrangian, identify
any conserved quantities, finds its simplest differential equation of motion, and get as far as
you can using analytic means alone toward finding the solution θ(t) (or t(θ) ).
PROBLEM 4-19: A rod is bent in the middle by angle α. The bottom portion is kept vertical
and the top portion is therefore oriented at angle α to the vertical. A bead of mass m is slipped
onto the top portion and the bottom portion is forced by a motor to rotate at constant angular
speed ω about the vertical axis. (a) Define a generalized coordinate for the bead and write
down the Lagrangian. (b) Identify any conserved quantity or quantitiies and explain why it (or
they) are conserved. (c) Find the generalized momentum of the bead and the Hamiltonian.
(d) Are there any equilibrium points of the bead? If so, are they stable or unstable?
181
4.9. SUMMARY
PROBLEM 4-20: A wire is bent into the shape of a cycloid, defined by the parametric
equations x = A(ϕ + sin ϕ) and y = A(1 − cos ϕ), where ϕ is the parameter (−π < ϕ < π),
and A is a constant. The wire is in a vertical plane, and is spun at constant angular velocity
ω about a vertical axis through its center. A bead of mass m is slipped onto the wire. (a)
Find the Lagrangian of the bead, using the parameter ϕ as the generalized coordinate. (b)
Identify any first integral of motion of the bead. (c) Are there any equilibrium points of the
bead? Are they stable or unstable? For any stable equilibrium point, find the frequency of
small oscillations about that point.
PROBLEM 4-21: Center of mass and relative coordinates. Show that for two particles moving
in one dimension, with coordinates x1 and x2 , with a potential that depends only upon their
separation x2 − x1 , then the Lagrangian
1 1
L= m1 ẋ21 + m2 ẋ22 − U (x2 − x1 ) (4.122)
2 2
can be rewritten in the form
1 1
L= M Ẋ 2 + µẋ02 − U (x0 ), (4.123)
2 2
where M = m1 + m2 is the total mass and µ = m1 m2 /M is the “reduced mass” of the
system, and X = (m1 x1 + m2 x2 )/M is the center of mass coordinate and x0 = x2 − x1 is
the relative coordinate.
PROBLEM 4-22: Two blocks of equal mass m, connected by a Hooke’s-law spring of un-
stretched length `, are free to move in one dimension. Find the equations of motion of the
system, using the relative and center of mass coordinates introduced in the preceding problem.
PROBLEM 4-23: A small block of mass m and a weight of mass M are connected by a string
of length D. The string has been threaded through a small hole in a tabletop, so the block can
slide without friction on the tabletop, while the weight hangs vertically beneath the tabletop.
We can let the hole be the origin of coordinates, and use polar coordinates r, θ for the block,
where r is the block’s distance from the hole, and z for the distance of the weight below the
tabletop. (a) Using generalized coordinates r and θ, write down the Lagrangian of the system
of block plus weight. (b) Write down a complete set of first integrals of the motion, explaining
the physical meaning of each. (c) Show that the first integrals can be combined to give an
equation of the form
1
E= (M + m)ṙ2 + Ueff (r) (4.124)
2
and write out an expression for Ueff (r). (d) Find the radius of a circular orbit of the block in
terms of constants of the motion. (e) Now suppose the block executes small oscillations about
a circular orbit. What is the frequency of these oscillations? Is the resulting orbit of the block
open or closed? That is, does the perturbed orbit of the block continually return to its former
position or not?
182
PROBLEM 4-24: The Moon has no atmosphere, so it would be possible in principle to shoot
projectiles off its surface with the escape velocity or even higher. The projectiles might be mined
Moon material, shot into space to use as raw material for building structures there. One way to
raise material to the escape velocity would be to construct a huge boom that would continually
swing around in a horizontal plane with constant angular velocity ω. Buckets of material would
be dropped onto the boom at some small distance r0 from the rotation axis, with no initial
radial velocity. The buckets would then be thrown outward by the boom’s rotation and would
come off the end of the boom with the escape velocity if the boom is long enough and if ω
is large enough. (a) Find the Lagrangian for a bucket of material of mass m. (b) Find its
equation of motion. (c) Solve it for r(t), subject to the given initial conditions (i.e., , be sure
that r = r0 and ṙ = 0 at t = 0.) (d) If the boom has radius R, find the radial and tangential
components of the bucket’s velocity, and its total speed, as it emerges from the end of the
boom. (e) Find the power input P = dE/dt into a bucket of mass m as a function of time.
Is the power input larger at the beginning or end of the bucket’s journey along the boom? (f)
Find the torque exerted by the boom on the bucket, as a function of the position r of the
bucket on the boom. There would be an equal but opposite torque back on the boom, caused
by the bucket, which might break the boom. At what part of the bucket’s journey would this
torque most likely break the boom? (g) If R = 100 meters and r0 = 1 meter, what must be
the rotational period of the boom so that buckets will reach the Moon’s escape speed as they
fly off the boom?
PROBLEM 4-25: Consider a vertical circular hoop of radius R rotating about a vertical axis
with constant angular velocity Ω. A bead of mass m is threaded on the hoop. Denote the
angle along the hoop of the bead as measured from the vertical as θ.
(a) Write the Lagrangian and equations of motion.
(b) For small angles θ 1 (in radians), we have the approximate expressions cos θ ∼ 1
and sin θ ∼ θ. In this regime, simplify the equation of motion for θ. Say a sentence about the
solution θ(t) that you expect.
(c) In the result of part (a), rewrite things in terms of the new coordinate α ≡ θ − π. Note
the mildly exciting identities cos(α + π) = − cos α and sin(α + π) = − sin α. Now consider
the regime α ∼ 0 and simplify the equation of motion for α.
PROBLEM 4-26: Consider the Lagrangian L0 = m ẋ ẏ − k x y for a particle free to move in

two dimensions, where x and y are Cartesian coordinates, and m and k are constants.
(a) Find the equations of motion for the system.
(b) Confirm that the answer to part (a) is the same if we were to use instead the Lagrangian
for the harmonic oscillator L = (1/2)m(ẋ2 + ẏ 2 ) − (1/2)k(x2 + y 2 ).
(c) Show that L and L0 do not differ by a total derivative!
PROBLEM 4-27: In certain situations, it is possible to incorporate frictional effects in a simple

way into a Lagrangian problem. As an example, consider the Lagrangian

γt 1 2 1 2
L=e mq̇ − kq . (4.125)
2 2
183
4.9. SUMMARY
(a) Find the equation of motion for the system.

(b) Do a coordinate change s = eγt/2 q. Rewrite the dynamics in terms of s.
(c) How would you describe the system?
PROBLEM 4-28: Consider a particle moving in three dimensions with Lagrangian L =

(1/2)m(ẋ2 + ẏ 2 + ż 2 ) + aẋ = b, where a and b are constants. (a) Find the equations of motion
and show that the particle moves in a straight line at constant speed, so that it must be a free
particle. (b) The result of (a) shows that there must be another reference frame (x0 , y 0 , z 0 ) such
that the Lagrangian is just the usual free-particle Lagrangian L0 = (1/2)m(ẋ02 + ẏ02 + ż02 ).
However, L0 may also be allowed an additive constant, which cannot show up in Lagrange’s
equations. Find the Galilean transformation between (x, y, z) and (x0 , y 0 , z 0 ) and find the
velocity of the new primed frame in terms of a and b.
PROBLEM 4-29: Consider a Lagrangian L0 = L + df /dt, where the Lagrangian is L =

L(t, qk , q̇k ), and the function f = f (qk , t). (a) Show that L0 = L0 (qk , q̇k , t), so that it
depends upon the proper variables. Show that this would not generally be true if f were
allowed to depend upon the q̇k0 s. (b) Show that L0 obeys Lagrange’s equations if L does, by
substituting L0 into Lagrange’s equations. Therefore the equations of motion are the same
using L0 as using L, so the Lagrangian of a particle is not unique. (This problem requires care
in taking total and partial derivatives!)
PROBLEM 4-30: Show that the function L0 given in the preceding problem must obey La-
grange’s equations if L does, directly from the principle of stationary action. Lagrange’s
equations do not have to be written down for this proof!
PROBLEM 4-31: Consider the Lagrangian L0 = mẋẏ − kxy for a particle free to move
in two dimensions, where x and y are Cartesian coordinates, and m and k are constants.
(a) Show that his Lagrangian gives the equations of motion appropriate for a two-dimension
simple harmonic oscillator. Therefore as far as the motion of the particle is concerned, L0 is
equivalent to L = (1/2)m(ẋ2 + ẋ2 ) − (1/2)k(x2 + y 2 ). (b) Show that L0 and L do not differ
by the total time derivative of any function f (x, y). Therefore L0 is not a member of the
class of Lagrangians mentioned in the preceding problems, so there are even more Lagrangians
describing a particle than suggested before.
PROBLEM 4-32: Consider a Lagrangian that depends on second derivatives of the coordi-
nates
L = L(qk , q̇k , q̈k , t) . (4.126)
Through the variational principle, show that the resulting differential equations of motion are
third order in the time derivative.
PROBLEM 4-33: A pendulum consists of a plumb bob of mass m on the end of a string that
swings back and forth in a plane. The upper end of the string passes through a small hole in
the ceiling, and the angle θ of the bob relative to the vertical changes with time as it swings
back and forth. The string is pulled upward at constant rate through the hole, so the length
184
R of the pendulum decreases at a constant rate, with dR/dt = −α. (a) Find the Lagrangian
of the bob, using θ as the generalized coordinate. (b) Find the Hamiltonian H. Is H equal to
the energy E? Why or why not? (c) Is either H or E conserved? Why or why not?
PROBLEM 4-34: A spherical pendulum consists of a particle of mass m on the end of a

string of length R. The position of the particle can be described by a polar angle θ and an
azimuthal angle ϕ. The length of the string decreases at the rate dR/dt = −f (t), where
f (t) is a positive function of time. (a) Find the Lagrangian of the particle, using θ and ϕ as
generalized coordinates. (b) Find the Hamiltonian H. Is H equal to the energy? Why of why
not? (c) Is either E or H conserved? Why or why not?
PROBLEM 4-35: Equation 4.100 for the Hamiltonian of a bead on a parabolic wire turning
with constant angular velocity ω is
1
H= m[(1 + 4α2 r2 )ṙ2 − r2 ω 2 ] + mgαr2 , (4.127)
2
where H is a constant. Reduce the problem to quadrature: That is, find an equation for the
time t is terms of an integral over r.
PROBLEM 4-36: A bead of mass m is placed on a vertically-oriented circular hoop of radius

R that is forced to rotate with constant angular velocity ω about a vertical axis through its
center. (a) Using the polar angle θ measured up from the bottom as the single generalized
coordinate, find the kinetic and potential energies of the bead. (Remember that the bead has
motion due to the forced rotation of the hoop as well as motion due to changing θ. ) (b) Find
the bead’s equation of motion using Lagrange’s equation. (c) Is its energy conserved? Why
or why not? (c) Find its Hamiltonian. Is H conserved? Why or why not? (d) Is E = H?
Why or why not? (e) Find the equilibrium angle θ0 for the bead as a function of the hoop’s
angular velocity ω. Sketch a graph of θ0 versus ω. Notice that there is a “phase transition” at
a certain critical velocity ωcrit . (b) Find the frequency of small oscillations of the bead about
the equilibrium angle θ0 , as a function of ω.
PROBLEM 4-37: A bead of mass m is placed on a vertically-oriented elliptical hoop that is

forced to rotate with constant angular velocity ω about a vertical axis through its center. The
ellipse is defined by (x/a)2 + (y/b)2 = 1 where a and b are the semimajor and semiminor axes
of the ellipse, and suppose that the vertical axis is the semiminor axis. (a) Choose a generalized
coordinate for the bead and find the Lagrangian (b) Is the bead’s energy conserved? Why or
why not? (c) Is the bead’s angular momentum conserved about the vertical axis? Why or why
not? (d) Find the bead’s Hamiltonian. Is H conserved? Why or why not? (d) Is E = H?
Why or why not? (e) Given ω, is there an equilibrium position of the bead, and is it stable or
unstable?
PROBLEM 4-38: In Example 9 we analyzed the case of a bead on a rotating parabolic wire.
The energy of the bead was not conserved, but the Hamiltonian was:
1
H= m(1 + 4α2 r2 )ṙ2 + “Ueff
00
= constant,
2
185
4.9. SUMMARY
where
00 1 2
“Ueff = mr (2gα2 − ω 2 ).
2
√
There is an equilibrium point at r = 0 which is unstable if ω > ω0 ≡ 2g α, neutrally stable if
ω = ω0 , and stable if ω < ω0 . Find the frequency of small oscillations about r = 0 if ω < ω0 .
Hint: Note that for very small oscillations near r = 0 the quantity (1 + 4α2 r2 ) ∼ = 1 to an
excellent approximation, so the Hamiltonian is
1 1
H∼
= mṙ2 + kr2
2 2
where the constant k = m(2gα2 − ω 2 ).
186
Appendix A
When is H 6= E?
In Example 9 the Hamiltonian H was not equal to E. Why were they differ-
ent, and why was H conserved while E was not?
The definition H = q̇k ∂L/∂ q̇k − L (using the Einstein summation conven-
tion, implying a sum over k in the first term on the right since k is repeated
in that term) contains the Lagrangian L = T − U , where only the kinetic
energy T depends upon the generalized velocities q̇k ; therefore
∂T
H = q̇k + U − T. (A.1)
∂ q̇k
Now let r(qk , t) be the position vector of the particle from some arbitrary
origin fixed in an inertial frame, in terms of the time and any or all of the
generalized coordinates qk . Then the velocity of the particle is
dr(qk , t) ∂r ∂r
v= = + q̇l , (A.2)
dt ∂t ∂ql
because r can change with time either from an explicit time dependence
or because one or more of the generalized coordinates changes with time.
Therefore its kinetic energy is T = (1/2)mv 2 = (1/2)mv · v, which is

1 ∂r ∂r ∂r ∂r ∂r ∂r
T = m · +2 · q̇l + q̇l · q̇m (A.3)
2 ∂t ∂t ∂t ∂ql ∂ql ∂qm
where we have used a different dummy summation index m in the final factor
to distinguish it from the sum over l is the preceding factor. That is, by the
Einstein summation convention the final term above is actually the product of
187
APPENDIX A. WHEN IS H 6= E?
two sums, one over l and one over m. Now we can take the partial derivative
of T with respect to a particular one of the generalized velocities q̇k ,

∂T 1 ∂r ∂r ∂r ∂r
= m 2 · +2 · q̇l (A.4)
∂ q̇k 2 ∂t ∂qk ∂qk ∂ql
where there is a factor of two in the second term because q̇k occurs in both
summations in the last term of the expression for T . Therefore the sum

∂T ∂r ∂r ∂r ∂r
q̇k = m · q̇k + q̇k · q̇l
∂ q̇k ∂t ∂qk ∂qk ∂ql

∂r ∂r ∂r
= 2T − m · + q̇k
∂t ∂t ∂qk
∂r dr
= 2T − m · , (A.5)
∂t dt
using equations (A.3)) and (A.4. The Hamiltonian H can therefore be written
dr ∂r ∂r
H =E−m · =E−p· (A.6)
dt ∂t ∂t
where p is the momentum of the particle in the chosen inertial frame. If the
transformation r = r(qk , t) between the Cartesian coordinates r = (x, y, z)
and the generalized coordinates qk happens not to be an explicit function of
time, i.e., , if ∂r/∂t = 0, then the Hamiltonian is just T + U . This case
occurs when there are no constraints or when any constraints are fixed in
space. But if the constraints are moving, then the transformation r = r(qk , t)
does generally depend upon time, and so in the likely case that there is
a component of ∂r/∂t in the direction of the particle’s momentum p, the
Hamiltonian is not equal to T + U .
For the problem of the bead on a rotating parabolic wire, where the
constraint is obviously moving, the position vector of the bead can be taken
to be r = (x, y, z) = (r cos ωt, r sin ωt, αr2 ). In that case we found that
H = T +U −mω 2 r2 , and it is easy to show that mω 2 r2 = p·∂r/∂t, as required
by equation (A.6)). It is clear that E is not conserved in this case because
the rotating wire is continually doing work on the bead. The wire exerts a
force F θ in the tangential direction, which causes work to be done at the
rate dW/dt = F θ v θ = F θ rω. From the elementary relationship N z = dLz /dt,
with torque N z = F θ r and angular momentum Lz = (r × pz = mr2 ω, it
follows that
dW d(mr2 ω) dr2
=ω = mω 2 , (A.7)
dt dt dt
APPENDIX A. WHEN IS H 6= E?
so that the work done by the wire is W = mω 2 r2 plus a constant of in-

tegration, which depends upon the initial location of the bead. Thus the
energy’ E = T + U of the bead increases by the work done upon it by the
wire, so that E minus the work done must be conserved, and that difference
E − mω 2 r2 = H.
Note that:
1. H is conserved if the Lagrangian L is not an explicit function of time.
2. H = E if the coordinate transformation r = r(qk , t) is not an explicit

function of time. Therefore it is possible to have E = H, with both E
and H conserved, or neither conserved, and it is also possible to have
E 6= H, with both conserved, neither conserved, or only one of the two
conserved. Examples are given in the problems.
189

Chap04 PDF

Uploaded by

Copyright:

Available Formats

Chap04 PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chap04 PDF

Uploaded by

Copyright:

Available Formats

Chapter 4

Motivated by discussions of the variational principle in the previous chapter,

4.1 The Lagrangian in Cartesian coordinates

is the particle’s kinetic energy and

Fx = mẍ, Fy = mÿ, Fz = mz̈ ⇒ −∇U = ma . (4.8)

4.2 Hamilton’s principle

Cartesian Cylindrical Spherical

FIGURE 4.1 : Cartesian, cylindrical, and spherical coordinates

the position of the particle inside the box.

Having chosen a set of generalized coordinates qk for a particle, the inte-

L = L(t, q1 , q2 , .., q̇1 , q̇2 , ...) = L(t, qk , q̇i ) (4.10)

in terms of the generalized coordinates, generalized velocities, and the time.

These then have to be the equations of motion of the system if Hamilton’s

2. Constraints provide for relations between the variables describing a

constraints on the coordinates are due to forces in the system that

EXAMPLE 4-1: A simple pendulum

θ̈ + (g/R) sin θ = 0. (4.18)

EXAMPLE 4-2: A bead sliding on a vertical helix

EXAMPLE 4-3: Block on an inclined plane

FIGURE 4.2 : A bead sliding on a vertically-oriented helical wire

FIGURE 4.3 : Block sliding down an inclined plane

4.3 Generalized momenta and cyclic coordi-

FIGURE 4.4 : Particle moving on a tabletop

EXAMPLE 4-4: Particle on a tabletop, with a central force

so the Lagrange equations

Fr = m(r̈ − rϕ̇2 ) ≡ mar and Fϕ = m(rϕ̈ + 2ṙϕ̇) = 0, (4.32)

ar = r̈ − rϕ̇2 and aϕ = rϕ̈ + 2ṙϕ̇, (4.33)

and where the tangential acceleration aϕ is zero in this case.6

−kr = m(r̈ − rϕ̇2 ) and pϕ = mr2 ϕ̇ = constant. (4.34)

EXAMPLE 4-5: The spherical pendulum

FIGURE 4.6 : Coordinates of a ball hanging on an unstretchable string

The Lagrangian is therefore

The Lagrange equations become

the θ equation can be written

mR2 θ̈ = mR2 sin θ cos θϕ̇2 − mgR sin θ, (4.47)

a second-order differential equation for the polar angle θ as a function of time.

mR2 sin2 θ ϕ̇ = pϕ = constant (4.50)

about the vertical axis. Another first integral is energy conservation

4.4 Systems of particles

EXAMPLE 4-6: Two interacting particles

m1 ẍ1 = +k(x2 − x1 ) and m2 ẍ2 = −k(x2 − x1 ). (4.59)

EXAMPLE 4-7: Pulleys everywhere

Length of rope = (H − y1 ) + 2 (H − Y ) + (H − y2 ) + 3πR, (4.66)

We can now solve for ÿ1 and ÿ2 ,

EXAMPLE 4-8: A block on a movable inclined plane

Y =0, x = X + D cos α , y = −D sin α . (4.78)

Ẏ = 0 , ẋ = Ẋ + Ḋ cos α , ẏ = −Ḋ sin α . (4.79)

We can now substitute these into (4.77) and get

The Lagrangian is now

ax = ẍ = Ẍ + D̈ cos α , ay = ÿ = −D̈ sin α. (4.86)

Substituting our solution from (4.85) into these, we have

(M/m)g cos2 α sin α g sin2 α cos α

4.5 The Hamiltonian

This result is particularly interesting if L is not an explicit function of time,

where we have already defined the generalized momenta to be pi = ∂L/∂ q̇k ,

EXAMPLE 4-9: Bead on a rotating parabolic wire

v 2 = ṙ2 + r2 ϕ̇2 + ż 2 = ṙ2 + r2 ω 2 + (2αrṙ)2 (4.95)

(1 + 4α2 r2 )r̈ + 4α2 rṙ2 + (2gα − ω 2 )r = 0. (4.98)