CHAPTER

1 Survey of the
Elementary Principles

The motion of material bodies formed the subject of some of the earliest research
pursued by the pioneers of physics. From their efforts there has evolved a vast
field known as analytical mechanics or dynamics, or simply, mechanics. In the
present century the term “classical mechanics” has come into wide use to denote
this branch of physics in contradistinction to the newer physical theories, espe-
cially quantum mechanics. We shall follow this usage, interpreting the name to
include the type of mechanics arising out of the special theory of relativity. It is
the purpose of this book to develop the structure of classical mechanics and to
outline some of its applications of present-day interest in pure physics. Basic to
any presentation of mechanics are a number of fundamental physical concepts,
such as space, time, simultaneity, mass, and force. For the most part, however,
these concepts will not be analyzed critically here; rather, they will be assumed as
undefined terms whose meanings are familiar to the reader.

1.1 MECHANICS OF A PARTICLE

Let r be the radius vector of a particle from some given origin and v its vector
velocity:

dr
v= . (1.1)
dt

The linear momentum p of the particle is defined as the product of the particle
mass and its velocity:

p = mv. (1.2)

In consequence of interactions with external objects and fields, the particle may
experience forces of various types, e.g., gravitational or electrodynamic; the vec-
tor sum of these forces exerted on the particle is the total force F. The mechanics
of the particle is contained in Newton’s second law of motion, which states that
there exist frames of reference in which the motion of the particle is described by
the differential equation

dp
F= ≡ ṗ, (1.3)
dt
1

“M01 Goldstein ISBN C01” — 2011/2/15 — page 1 — #1

2 Chapter 1 Survey of the Elementary Principles

or
d
F= (mv). (1.4)
dt
In most instances, the mass of the particle is constant and Eq. (1.4) reduces to

dv
F=m = ma, (1.5)
dt
where a is the vector acceleration of the particle defined by

d 2r
a= . (1.6)
dt 2
The equation of motion is thus a differential equation of second order, assuming
F does not depend on higher-order derivatives.
A reference frame in which Eq. (1.3) is valid is called an inertial or Galilean
system. Even within classical mechanics the notion of an inertial system is some-
thing of an idealization. In practice, however, it is usually feasible to set up a co-
ordinate system that comes as close to the desired properties as may be required.
For many purposes, a reference frame fixed in Earth (the “laboratory system”)
is a sufficient approximation to an inertial system, while for some astronomical
purposes it may be necessary to construct an inertial system (or inertial frame) by
reference to distant galaxies.
Many of the important conclusions of mechanics can be expressed in the form
of conservation theorems, which indicate under what conditions various mechan-
ical quantities are constant in time. Equation (1.3) directly furnishes the first of
these, the

Conservation Theorem for the Linear Momentum of a Particle: If the total force,
F, is zero, then ṗ = 0 and the linear momentum, p, is conserved.

The angular momentum of the particle about point O, denoted by L, is defined

as

L = r 3 p, (1.7)

where r is the radius vector from O to the particle. Notice that the order of the
factors is important. We now define the moment of force or torque about O as

N = r 3 F. (1.8)

The equation analogous to (1.3) for N is obtained by forming the cross product of
r with Eq. (1.4):

d
r3F = N = r3 (mv). (1.9)
dt

“M01 Goldstein ISBN C01” — 2011/2/15 — page 2 — #2

 1.1 Mechanics of a Particle 3

Equation (1.9) can be written in a different form by using the vector identity:

d d
(r 3 mv) = v 3 mv + r 3 (mv), (1.10)
dt dt

where the first term on the right obviously vanishes. In consequence of this iden-
tity, Eq. (1.9) takes the form

d dL
N= (r 3 mv) = ≡ L̇. (1.11)
dt dt

Note that both N and L depend on the point O, about which the moments are
taken.
As was the case for Eq. (1.3), the torque equation, (1.11), also yields an imme-
diate conservation theorem, this time the

Conservation Theorem for the Angular Momentum of a Particle: If the total

torque, N, is zero then L̇ = 0, and the angular momentum L is conserved.

Next consider the work done by the external force F upon the particle in going
from point 1 to point 2. By definition, this work is
! 2
W12 = F ? ds. (1.12)
1

For constant mass (as will be assumed from now on unless otherwise specified),
the integral in Eq. (1.12) reduces to
! ! !
dv m d 2
F ? ds = m ? v dt = (v ) dt,
dt 2 dt

and therefore
m 2
W12 = (v − v12 ). (1.13)
2 2

The scalar quantity mv 2 /2 is called the kinetic energy of the particle and is
denoted by T , so that the work done is equal to the change in the kinetic energy:

W12 = T2 − T1 . (1.14)

If the force field is such that the work W12 is the same for any physically
possible path between points 1 and 2, then the force (and the system) is said to be
conservative. An alternative description of a conservative system is obtained by
imagining the particle being taken from point 1 to point 2 by one possible path
and then being returned to point 1 by another path. The independence of W12 on
the particular path implies that the work done around such a closed circuit is zero,

“M01 Goldstein ISBN C01” — 2011/2/15 — page 3 — #3

4 Chapter 1 Survey of the Elementary Principles

i.e.:
"
F ? ds = 0. (1.15)

Physically it is clear that a system cannot be conservative if friction or other dis-

sipation forces are present, because F ? ds due to friction is always positive and
the integral cannot vanish.
By a well-known theorem of vector analysis, a necessary and sufficient condi-
tion that the work, W12 , be independent of the physical path taken by the particle
is that F be the gradient of some scalar function of position:

F = −∇V (r), (1.16)

where V is called the potential, or potential energy. The existence of V can be

inferred intuitively by a simple argument. If W12 is independent of the path of
integration between the end points 1 and 2, it should be possible to express W12
as the change in a quantity that depends only upon the positions of the end points.
This quantity may be designated by −V , so that for a differential path length we
have the relation

F ? ds = −d V

or
∂V
Fs = − ,
∂s
which is equivalent to Eq. (1.16). Note that in Eq. (1.16) we can add to V any
quantity constant in space, without affecting the results. Hence the zero level of V
is arbitrary.
For a conservative system, the work done by the forces is

W12 = V1 − V2 . (1.17)

Combining Eq. (1.17) with Eq. (1.14), we have the result

T1 + V1 = T2 + V2 , (1.18)

which states in symbols the

Energy Conservation Theorem for a Particle: If the forces acting on a particle

are conservative, then the total energy of the particle, T + V , is conserved.

The force applied to a particle may in some circumstances be given by the

gradient of a scalar function that depends explicitly on both the position of the
particle and the time. However, the work done on the particle when it travels a
distance ds,
∂V
F ? ds = − ds,
∂s

“M01 Goldstein ISBN C01” — 2011/2/15 — page 4 — #4

 1.2 Mechanics of a System of Particles 5

is then no longer the total change in −V during the displacement, since V also
changes explicitly with time as the particle moves. Hence, the work done as the
particle goes from point 1 to point 2 is no longer the difference in the function V
between those points. While a total energy T + V may still be defined, it is not
conserved during the course of the particle’s motion.

1.2 MECHANICS OF A SYSTEM OF PARTICLES

In generalizing the ideas of the previous section to systems of many particles,

we must distinguish between the external forces acting on the particles due to
sources outside the system, and internal forces on, say, some particle i due to all
other particles in the system. Thus, the equation of motion (Newton’s second law)
for the ith particle is written as
#
F ji + Fi(e) = ṗi , (1.19)
j

(e)
where Fi stands for an external force, and F ji is the internal force on the ith
particle due to the jth particle (Fii , naturally, is zero). We shall assume that the
(e)
Fi j (like the Fi ) obey Newton’s third law of motion in its original form: that the
forces two particles exert on each other are equal and opposite. This assumption
(which does not hold for all types of forces) is sometimes referred to as the weak
law of action and reaction.
Summed over all particles, Eq. (1.19) takes the form

d2 # # (e) #
2
m i ri = Fi + F ji . (1.20)
dt
i i, j i
i̸= j

The first sum on the right is simply the total external force F(e) , while the second
term vanishes, since the law of action and reaction states that each pair Fi j + F ji
is zero. To reduce the left-hand side, we define a vector R as the average of the
radii vectors of the particles, weighted in proportion to their mass:
$ $
m i ri m i ri
R= $ = . (1.21)
mi M

The vector R defines a point known as the center of mass, or more loosely as the
center of gravity, of the system (cf. Fig. 1.1). With this definition, (1.20) reduces
to

d 2 R # (e)
M = Fi ≡ F(e) , (1.22)
dt 2
i

“M01 Goldstein ISBN C01” — 2011/2/15 — page 5 — #5

6 Chapter 1 Survey of the Elementary Principles

mi

Center of mass

ri R mj

rj

FIGURE 1.1 The center of mass of a system of particles.

which states that the center of mass moves as if the total external force were
acting on the entire mass of the system concentrated at the center of mass.
Purely internal forces, if the obey Newton’s third law, therefore have no effect
on the motion of the center of mass. An oft-quoted example is the motion of
an exploding shell—the center of mass of the fragments traveling as if the
shell were still in a single piece (neglecting air resistance). The same princi-
ple is involved in jet and rocket propulsion. In order that the motion of the
center of mass be unaffected, the ejection of the exhaust gases at high veloc-
ity must be counterbalanced by the forward motion of the vehicle at a slower
velocity.
By Eq. (1.21) the total linear momentum of the system,

# dri dR
P= mi =M , (1.23)
dt dt

is the total mass of the system times the velocity of the center of mass. Conse-
quently, the equation of motion for the center of mass, (1.23), can be restated as
the

Conservation Theorem for the Linear Momentum of a System of Particles: If the

total external force is zero, the total linear momentum is conserved.

We obtain the total angular momentum of the system by forming the cross
product ri 3 pi and summing over i. If this operation is performed in Eq. (1.19),
there results, with the aid of the identity, Eq. (1.10),

# # d # (e)
#
(ri 3 ṗi ) = (ri 3 pi ) = L̇ = ri 3 Fi + ri 3 F ji . (1.24)
dt
i i i i, j
i̸= j

“M01 Goldstein ISBN C01” — 2011/2/15 — page 6 — #6

 1.2 Mechanics of a System of Particles 7

FIGURE 1.2 The vector ri j between the ith and jth particles.

The last term on the right in (1.24) can be considered a sum of the pairs of the
form

ri 3 F ji + r j 3 Fi j = (ri − r j ) 3 F ji , (1.25)

using the equality of action and reaction. But ri − r j is identical with the vector
ri j from j to i (cf. Fig. 1.2), so that the right-hand side of Eq. (1.25) can be written
as

ri j 3 F ji .

If the internal forces between two particles, in addition to being equal and
opposite, also lie along the line joining the particles—a condition known as
the strong law of action and reaction—then all of these cross products vanish.
The sum over pairs is zero under this assumption and Eq. (1.24) may be written in
the form
dL
= N(e) . (1.26)
dt
The time derivative of the total angular momentum is thus equal to the moment
of the external force about the given point. Corresponding to Eq. (1.26) is the

Conservation Theorem for Total Angular Momentum: L is constant in time if the

applied (external) torque is zero.

(It is perhaps worthwhile to emphasize that this is a vector theorem; i.e., L z

(e) (e) (e)
will be conserved if Nz is zero, even if N x and N y are not zero.)
Note that the conservation of linear momentum in the absence of applied forces
assumes that the weak law of action and reaction is valid for the internal forces.
The conservation of the total angular momentum of the system in the absence of
applied torques requires the validity of the strong law of action and reaction—that

“M01 Goldstein ISBN C01” — 2011/2/15 — page 7 — #7

8 Chapter 1 Survey of the Elementary Principles

the internal forces in addition be central. Many of the familiar physical forces,
such as that of gravity, satisfy the strong form of the law. But it is possible to
find forces for which action and reaction are equal even though the forces are not
central (see below). In a system involving moving charges, the forces between
the charges predicted by the Biot-Savart law may indeed violate both forms of
the action and reaction law.* Equations (1.23) and (1.26), and their corresponding
conservation theorems, are not applicable in such cases, at least in the form given
here. Usually it is then possible to find some generalization of P or L that is
conserved. Thus, in an isolated system of moving charges it is the sum of the
mechanical angular momentum and the electromagnetic “angular momentum” of
the field that is conserved.
Equation (1.23) states that the total linear momentum of the system is the same
as if the entire mass were concentrated at the center of mass and moving with it.
The analogous theorem for angular momentum is more complicated. With the
origin O as reference point, the total angular momentum of the system is

#
L= ri 3 pi .
i

Let R be the radius vector from O to the center of mass, and let ri′ be the radius
vector from the center of mass to the ith particle. Then we have (cf. Fig. 1.3)

ri = ri′ + R (1.27)

FIGURE 1.3 The vectors involved in the shift of reference point for the angular
momentum.

*If two charges are moving uniformly with parallel velocity vectors that are not perpendicular to the
line joining the charges, then the net mutual forces are equal and opposite but do not lie along the
vector between the charges. Consider, further, two charges moving (instantaneously) so as to “cross
the T,” i.e., one charge moving directly at the other, which in turn is moving at right angles to the first.
Then the second charge exerts a nonvanishing magnetic force on the first, without experiencing any
magnetic reaction force at that instant.

“M01 Goldstein ISBN C01” — 2011/2/15 — page 8 — #8

 1.2 Mechanics of a System of Particles 9

and

vi = vi′ + v

where
dR
v=
dt
is the velocity of the center of mass relative to O, and

dri′
vi′ =
dt
is the velocity of the ith particle relative to the center of mass of the system. Using
Eq. (1.27), the total angular momentum takes on the form
% &
# # # d #
L= R 3 mi v + ri′ 3 m i vi′ + m i ri′ 3v + R3 m i ri′ .
dt
i i i i
$
The last two terms in this expression vanish, for both contain the factor m i ri′ ,
which, it will be recognized, defines the radius vector of the center of mass in the
very coordinate system whose origin is the center of mass and is therefore a null
vector. Rewriting the remaining terms, the total angular momentum about O is
#
L = R 3 Mv + ri′ 3 pi′ . (1.28)
i

In words, Eq. (1.28) says that the total angular momentum about a point O is
the angular momentum of motion concentrated at the center of mass, plus the
angular momentum of motion about the center of mass. The form of Eq. (1.28)
emphasizes that in general L depends on the origin O, through the vector R. Only
if the center of mass is at rest with respect to O will the angular momentum be
independent of the point of reference. In this case, the first term in (1.28) vanishes,
and L always reduces to the angular momentum taken about the center of mass.
Finally, let us consider the energy equation. As in the case of a single particle,
we calculate the work done by all forces in moving the system from an initial
configuration 1, to a final configuration 2:

#! 2 #! 2
(e)
#! 2
W12 = Fi ? dsi = Fi ? dsi + F ji ? dsi . (1.29)
i 1 i 1 i, j
1
i̸= j

Again, the equations of motion can be used to reduce the integrals to

#! 2 #! 2 #! 2 ' (
1 2
Fi ? dsi = m i v̇i ? vi dt = d m i vi .
1 1 1 2
i i i

“M01 Goldstein ISBN C01” — 2011/2/15 — page 9 — #9

10 Chapter 1 Survey of the Elementary Principles

Hence, the work done can still be written as the difference of the final and initial
kinetic energies:

W12 = T2 − T1 ,

where T , the total kinetic energy of the system, is

1#
T = m i vi2 . (1.30)
2
i

Making use of the transformations to center-of-mass coordinates, given in

Eq. (1.27), we may also write T as

1#
T = m i (v + vi′ ) ? (v + vi′ )
2
i
% &
1# 1# d #
= m i v2 + m i vi′2 + v ? m i ri′ ,
2 2 dt
i i i

and by the reasoning already employed in calculating the angular momentum, the
last term vanishes, leaving

1 1#
T = Mv 2 + m i vi′2 (1.31)
2 2
i

The kinetic energy, like the angular momentum, thus also consists of two parts:
the kinetic energy obtained if all the mass were concentrated at the center of mass,
plus the kinetic energy of motion about the center of mass.
Consider now the right-hand side of Eq. (1.29). In the special case that the
external forces are derivable in terms of the gradient of a potential, the first term
can be written as
# ! 2 (e) #! 2 # ))2
Fi ? dsi = − ∇i Vi ? dsi = − Vi )) ,
i 1 i 1 i 1

where the subscript i on the del operator indicates that the derivatives are with
respect to the components of ri . If the internal forces are also conservative, then
the mutual forces between the ith and jth particles, Fi j and F ji , can be obtained
from a potential function Vi j . To satisfy the strong law of action and reaction, Vi j
can be a function only of the distance between the particles:

Vi j = Vi j (| ri − r j |). (1.32)

The two forces are then automatically equal and opposite,

F ji = −∇i Vi j = +∇ j Vi j = −Fi j , (1.33)

“M01 Goldstein ISBN C01” — 2011/2/15 — page 10 — #10

 1.2 Mechanics of a System of Particles 11

and lie along the line joining the two particles,

∇Vi j (| ri − r j |) = (ri − r j ) f, (1.34)

where f is some scalar function. If Vi j were also a function of the difference of

some other pair of vectors associated with the particles, such as their velocities
or (to step into the domain of modern physics) their intrinsic “spin” angular mo-
menta, then the forces would still be equal and opposite, but would not necessarily
lie along the direction between the particles.
When the forces are all conservative, the second term in Eq. (1.29) can be
rewritten as a sum over pairs of particles, the terms for each pair being of the
form
! 2
− (∇i Vi j ? dsi + ∇ j Vi j ? ds j ).
1

If the difference vector ri − r j is denoted by ri j , and if ∇i j stands for the gradient

with respect to ri j , then

∇i Vi j = ∇i j Vi j = −∇ j Vi j ,

and

dsi − ds j = dri − dr j = dri j ,

so that the term for the ij pair has the form

!
− ∇i j Vi j ? dri j .

The total work arising from internal forces then reduces to

! )
1# 2 1 # ))2
− ∇i j Vi j ? dri j = − Vi j ) . (1.35)
2 1 2 1
i, j i, j
i̸= j i̸= j

The factor 12 appears in Eq. (1.35) because in summing over both i and j each
member of a given pair is included twice, first in the i summation and then in the
j summation.
From these considerations, it is clear that if the external and internal forces are
both derivable from potentials it is possible to define a total potential energy, V ,
of the system,
# 1#
V = Vi + Vi j , (1.36)
2
i i, j
i̸= j

such that the total energy T + V is conserved, the analog of the conservation
theorem (1.18) for a single particle.

“M01 Goldstein ISBN C01” — 2011/2/15 — page 11 — #11

12 Chapter 1 Survey of the Elementary Principles

The second term on the right in Eq. (1.36) will be called the internal potential
energy of the system. In general, it need not be zero and, more important, it may
vary as the system changes with time. Only for the particular class of systems
known as rigid bodies will the internal potential always be constant. Formally,
a rigid body can be defined as a system of particles in which the distances ri j
are fixed and cannot vary with time. In such case, the vectors dri j can only be
perpendicular to the corresponding ri j , and therefore to the Fi j . Therefore, in a
rigid body the internal forces do no work, and the internal potential must remain
constant. Since the total potential is in any case uncertain to within an additive
constant, an unvarying internal potential can be completely disregarded in dis-
cussing the motion of the system.

1.3 CONSTRAINTS

From the previous sections one might obtain the impression that all problems in
mechanics have been reduced to solving the set of differential equations (1.19):
(e)
#
m i r̈i = Fi + F ji .
j

One merely substitutes the various forces acting upon the particles of the system,
turns the mathematical crank, and grinds out the answers! Even from a purely
physical standpoint, however, this view is oversimplified. For example, it may be
necessary to take into account the constraints that limit the motion of the system.
We have already met one type of system involving constraints, namely rigid bod-
ies, where the constraints on the motions of the particles keep the distances ri j
unchanged. Other examples of constrained systems can easily be furnished. The
beads of an abacus are constrained to one-dimensional motion by the supporting
wires. Gas molecules within a container are constrained by the walls of the ves-
sel to move only inside the container. A particle placed on the surface of a solid
sphere is subject to the constraint that it can move only on the surface or in the
region exterior to the sphere.
Constraints may be classified in various ways, and we shall use the following
system. If the conditions of constraint can be expressed as equations connecting
the coordinates of the particles (and possibly the time) having the form

f (r1 , r2 , r3 , . . . , t) = 0, (1.37)

then the constraints are said to be holonomic. Perhaps the simplest example of
holonomic constraints is the rigid body, where the constraints are expressed by
equations of the form

(ri − r j )2 − ci2j = 0.

A particle constrained to move along any curve or on a given surface is another

obvious example of a holonomic constraint, with the equations defining the curve
or surface acting as the equations of a constraint.

“M01 Goldstein ISBN C01” — 2011/2/15 — page 12 — #12

 1.3 Constraints 13

Constraints not expressible in this fashion are called nonholonomic. The walls
of a gas container constitute a nonholonomic constraint. The constraint involved
in the example of a particle placed on the surface of a sphere is also nonholo-
nomic, for it can be expressed as an inequality

r 2 − a2 ≥ 0

(where a is the radius of the sphere), which is not in the form of (1.37). Thus, in
a gravitational field a particle placed on the top of the sphere will slide down the
surface part of the way but will eventually fall off.
Constraints are further classified according to whether the equations of con-
straint contain the time as an explicit variable (rheonomous) or are not explicitly
dependent on time (scleronomous). A bead sliding on a rigid curved wire fixed
in space is obviously subject to a scleronomous constraint; if the wire is moving
in some prescribed fashion, the constraint is rheonomous. Note that if the wire
moves, say, as a reaction to the bead’s motion, then the time dependence of the
constraint enters in the equation of the constraint only through the coordinates
of the curved wire (which are now part of the system coordinates). The overall
constraint is then scleronomous.
Constraints introduce two types of difficulties in the solution of mechanical
problems. First, the coordinates ri are no longer all independent, since they are
connected by the equations of constraint; hence the equations of motion (1.19)
are not all independent. Second, the forces of constraint, e.g., the force that the
wire exerts on the bead (or the wall on the gas particle), is not furnished a pri-
ori. They are among the unknowns of the problem and must be obtained from the
solution we seek. Indeed, imposing constraints on the system is simply another
method of stating that there are forces present in the problem that cannot be spec-
ified directly but are known rather in terms of their effect on the motion of the
system.
In the case of holonomic constraints, the first difficulty is solved by the intro-
duction of generalized coordinates. So far we have been thinking implicitly in
terms of Cartesian coordinates. A system of N particles, free from constraints,
has 3N independent coordinates or degrees of freedom. If there exist holonomic
constraints, expressed in k equations in the form (1.37), then we may use these
equations to eliminate k of the 3N coordinates, and we are left with 3N − k inde-
pendent coordinates, and the system is said to have 3N − k degrees of freedom.
This elimination of the dependent coordinates can be expressed in another way,
by the introduction of new, 3N − k, independent variables q1 , q2 , . . . , q3N −k in
terms of which the old coordinates r1 , r2 , . . . , r N are expressed by equations of
the form

r1 = r1 (q1 , q2 , . . . , q3N −k , t)
..
. (1.38)
r N = r N (q1 , q2 , . . . , q3N −k , t)

“M01 Goldstein ISBN C01” — 2011/2/15 — page 13 — #13

14 Chapter 1 Survey of the Elementary Principles

FIGURE 1.4 Double pendulum.

containing the constraints in them implicitly. These are transformation equations

from the set of (rl ) variables to the (ql ) set, or alternatively Eqs. (1.38) can
be considered as parametric representations of the (rl ) variables. It is always
assumed that we can also transform back from the (ql ) to the (rl ) set, i.e., that
Eqs. (1.38) combined with the k equations of constraint can be inverted to obtain
any qi as a function of the (rl ) variable and time.
Usually the generalized coordinates, ql , unlike the Cartesian coordinates, will
not divide into convenient groups of three that can be associated together to
form vectors. Thus, in the case of a particle constrained to move on the surface
of a sphere, the two angles expressing position on the sphere, say latitude and
longitude, are obvious possible generalized coordinates. Or, in the example of
a double pendulum moving in a plane (two particles connected by an inexten-
sible light rod and suspended by a similar rod fastened to one of the particles),
satisfactory generalized coordinates are the two angles θ1 , θ2 . (Cf. Fig. 1.4.) Gen-
eralized coordinates, in the sense of coordinates other than Cartesian, are often
useful in systems without constraints. Thus, in the problem of a particle moving
in an external central force field (V = V (r )), there is no constraint involved,
but it is clearly more convenient to use spherical polar coordinates than Cartesian
coordinates. Do not, however, think of generalized coordinates in terms of con-
ventional orthogonal position coordinates. All sorts of quantities may be invoked
to serve as generalized coordinates. Thus, the amplitudes in a Fourier expansion
of r j may be used as generalized coordinates, or we may find it convenient to
employ quantities with the dimensions of energy or angular momentum.
If the constraint is nonholonomic, the equations expressing the constraint can-
not be used to eliminate the dependent coordinates. An oft-quoted example of
a nonholonomic constraint is that of an object rolling on a rough surface with-
out slipping. The coordinates used to describe the system will generally involve
angular coordinates to specify the orientation of the body, plus a set of coordi-
nates describing the location of the point of contact on the surface. The constraint
of “rolling” connects these two sets of coordinates; they are not independent.
A change in the position of the point of contact inevitably means a change in

“M01 Goldstein ISBN C01” — 2011/2/15 — page 14 — #14

 1.3 Constraints 15

FIGURE 1.5 Vertical disk rolling on a horizontal plane.

its orientation. Yet we cannot reduce the number of coordinates, for the “rolling”
condition is not expressible as a equation between the coordinates, in the manner
of (1.37). Rather, it is a condition on the velocities (i.e., the point of contact is
stationary), a differential condition that can be given in an integrated form only
after the problem is solved.
A simple case will illustrate the point. Consider a disk rolling on the horizontal
x y plane constrained to move so that the plane of the disk is always vertical.
The coordinates used to describe the motion might be the x, y coordinates of the
center of the disk, an angle of rotation φ about the axis of the disk, and an angle
θ between the axis of the disk and say, the x axis (cf. Fig 1.5). As a result of the
constraint the velocity of the center of the disk, v, has a magnitude proportional
to φ̇,

v = a φ̇,

where a is the radius of the disk, and its direction is perpendicular to the axis of
the disk:

ẋ = v sin θ,
ẏ = −v cos θ.

Combining these conditions, we have two differential equations of constraint:

d x − a sin θ dφ = 0,
(1.39)
dy + a cos θ dφ = 0.

Neither of Eqs. (1.39) can be integrated without in fact solving the problem; i.e.,
we cannot find an integrating factor f (x, y, θ, φ) that will turn either of the equa-
tions into exact differentials (cf. Derivation 4).* Hence, the constraints cannot be
reduced to the form of Eq. (1.37) and are therefore nonholonomic. Physically we
*In principle, an integrating factor can always be found for a first-order differential equation of con-
straint in systems involving only two coordinates and such constraints are therefore holonomic. A
familiar example is the two-dimensional motion of a circle rolling on an inclined plane.

“M01 Goldstein ISBN C01” — 2011/2/15 — page 15 — #15

16 Chapter 1 Survey of the Elementary Principles

can see that there can be no direct functional relation between φ and the other
coordinates x, y, and θ by noting that at any point on its path the disk can be
made to roll around in a circle tangent to the path and of arbitrary radius. At the
end of the process, x, y, and θ have been returned to their original values, but φ
has changed by an amount depending on the radius of the circle.
Nonintegrable differential constraints of the form of Eqs. (1.39) are of course
not the only type of nonholonomic constraints. The constraint conditions may
involve higher-order derivatives, or may appear in the form of inequalities, as we
have seen.
Partly because the dependent coordinates can be eliminated, problems involv-
ing holonomic constraints are always amenable to a formal solution. But there is
no general way to attack nonholonomic examples. True, if the constraint is nonin-
tegrable, the differential equations of constraint can be introduced into the prob-
lem along with the differential equations of motion, and the dependent equations
eliminated, in effect, by the method of Lagrange multipliers.
We shall return to this method at a later point. However, the more vicious cases
of nonholonomic constraint must be tackled individually, and consequently in the
development of the more formal aspects of classical mechanics, it is almost invari-
ably assumed that any constraint, if present, is holonomic. This restriction does
not greatly limit the applicability of the theory, despite the fact that many of the
constraints encountered in everyday life are nonholonomic. The reason is that the
entire concept of constraints imposed in the system through the medium of wires
or surfaces or walls is particularly appropriate only in macroscopic or large-scale
problems. But today physicists are more interested in atomic and nuclear prob-
lems. On this scale all objects, both in and out of the system, consist alike of
molecules, atoms, or smaller particles, exerting definite forces, and the notion of
constraint becomes artificial and rarely appears. Constraints are then used only
as mathematical idealizations to the actual physical case or as classical approxi-
mations to a quantum-mechanical property, e.g., rigid body rotations for “spin.”
Such constraints are always holonomic and fit smoothly into the framework of the
theory.
To surmount the second difficulty, namely, that the forces of constraint are
unknown a priori, we should like to so formulate the mechanics that the forces
of constraint disappear. We need then deal only with the known applied forces.
A hint as to the procedure to be followed is provided by the fact that in a particular
system with constraints, i.e., a rigid body, the work done by internal forces (which
are here the forces of constraint) vanishes. We shall follow up this clue in the
ensuing sections and generalize the ideas contained in it.

1.4 D’ALEMBERT’S PRINCIPLE AND LAGRANGE’S EQUATIONS

A virtual (infinitesimal) displacement of a system refers to a change in the con-

figuration of the system as the result of any arbitrary infinitesimal change of the
coordinates δri , consistent with the forces and constraints imposed on the system
at the given instant t. The displacement is called virtual to distinguish it from an
actual displacement of the system occurring in a time interval dt, during which

“M01 Goldstein ISBN C01” — 2011/2/15 — page 16 — #16

 1.4 D’Alembert’s Principle and Lagrange’s Equations 17

the forces and constraints may be changing. Suppose the system is in equilibrium;
i.e., the total force on each particle vanishes, Fi = 0. Then clearly the dot product
Fi ? δri , which is the virtual work of the force Fi in the displacement δri , also
vanishes. The sum of these vanishing products over all particles must likewise be
zero:
#
Fi ? δri = 0. (1.40)
i

As yet nothing has been said that has any new physical content. Decompose Fi
into the applied force, Fi(a) , and the force of constraint, fi ,

(a)
Fi = Fi + fi , (1.41)

so that Eq. (1.40) becomes

# (a)
#
Fi ? δri + fi ? δri = 0. (1.42)
i i

We now restrict ourselves to systems for which the net virtual work of the
forces of constraint is zero. We have seen that this condition holds true for rigid
bodies and it is valid for a large number of other constraints. Thus, if a particle is
constrained to move on a surface, the force of constraint is perpendicular to the
surface, while the virtual displacement must be tangent to it, and hence the virtual
work vanishes. This is no longer true if sliding friction forces are present, and we
must exclude such systems from our formulation. The restriction is not unduly
hampering, since the friction is essentially a macroscopic phenomenon. On the
other hand, the forces of rolling friction do not violate this condition, since the
forces act on a point that is momentarily at rest and can do no work in an infinites-
imal displacement consistent with the rolling constraint. Note that if a particle is
constrained to a surface that is itself moving in time, the force of constraint is
instantaneously perpendicular to the surface and the work during a virtual dis-
placement is still zero even though the work during an actual displacement in the
time dt does not necessarily vanish.
We therefore have as the condition for equilibrium of a system that the virtual
work of the applied forces vanishes:
#
Fi(a) ? δri = 0. (1.43)
i

Equation (1.43) is often called the principle of virtual work. Note that the coef-
ficients of δri can no longer be set equal to zero; i.e., in general Fi(a) ̸= 0, since
the δri are not completely independent but are connected by the constraints. In
order to equate the coefficients to zero, we must transform the principle into a
form involving the virtual displacements of the qi , which are independent. Equa-
tion (1.43) satisfies our needs in that it does not contain the fi , but it deals only
with statics; we want a condition involving the general motion of the system.

“M01 Goldstein ISBN C01” — 2011/2/15 — page 17 — #17

18 Chapter 1 Survey of the Elementary Principles

To obtain such a principle, we use a device first thought of by James Bernoulli

and developed by D’Alembert. The equation of motion,
Fi = ṗi ,
can be written as
Fi − ṗi = 0,
which states that the particles in the system will be in equilibrium under a force
equal to the actual force plus a “reversed effective force” −ṗi . Instead of (1.40),
we can immediately write
#
(Fi − ṗi ) ? δri = 0, (1.44)
i

and, making the same resolution into applied forces and forces of constraint, there
results
# (a) #
(Fi − ṗi ) ? δri + fi ? δri = 0.
i i

We again restrict ourselves to systems for which the virtual work of the forces of
constraint vanishes and therefore obtain
# (a)
(Fi − ṗi ) ? δri = 0, (1.45)
i

which is often called D’Alembert’s principle. We have achieved our aim, in that
the forces of constraint no longer appear, and the superscript (a) can now be
dropped without ambiguity. It is still not in a useful form to furnish equations
of motion for the system. We must now transform the principle into an expression
involving virtual displacements of the generalized coordinates, which are then
independent of each other (for holonomic constraints), so that the coefficients of
the δqi can be set separately equal to zero.
The translation from ri to q j language starts from the transformation equations
(1.38),
ri = ri (q1 , q2 , . . . , qn , t) (1.45′ )
(assuming n independent coordinates), and is carried out by means of the usual
“chain rules” of the calculus of partial differentiation. Thus, vi is expressed in
terms of the q̇k by the formula
dri # ∂ri ∂ri
vi ≡ = q̇k + . (1.46)
dt ∂qk ∂t
k

Similarly, the arbitrary virtual displacement δri can be connected with the virtual
displacements δqi by
# ∂ri
δri = δq j . (1.47)
∂q j
j

“M01 Goldstein ISBN C01” — 2011/2/15 — page 18 — #18

 1.4 D’Alembert’s Principle and Lagrange’s Equations 19

Note that no variation of time, δt, is involved here, since a virtual displacement
by definition considers only displacements of the coordinates. (Only then is the
virtual displacement perpendicular to the force of constraint if the constraint itself
is changing in time.)
In terms of the generalized coordinates, the virtual work of the Fi becomes
# # ∂ri
Fi ? δri = Fi ? δq j
∂q j
i i, j
#
= Q j δq j , (1.48)
j

where the Q j are called the components of the generalized force, defined as
# ∂ri
Qj = Fi ? . (1.49)
∂q j
i

Note that just as the q’s need not have the dimensions of length, so the Q’s do
not necessarily have the dimensions of force, but Q j δq j must always have the
dimensions of work. For example, Q j might be a torque N j and dq j a differential
angle dθ j , which makes N j dθ j a differential of work.
We turn now to the other other term involved in Eq. (1.45), which may be
written as
# #
ṗi ? δri = m i r̈i ? δri .
i i

Expressing δri by (1.47), this becomes

# ∂ri
m i r̈i ? δq j .
∂q j
i, j

Consider now the relation

# ∂ri #* d ' ∂ri
( '
d ∂ri
(+
m i r̈i ? = m i ṙi ? − m i ṙi ? . (1.50)
∂q j dt ∂q j dt ∂q j
i i

In the last term of Eq. (1.50) we can interchange the differentiation with respect
to t and q j , for, in analogy to (1.46),
' ( # ∂ 2 ri
d ∂ri ∂ ṙi ∂ 2 ri
= = q̇k + ,
dt ∂q j ∂q j ∂q j ∂qk ∂q j ∂t
k
∂vi
= ,
∂q j
by Eq. (1.46). Further, we also see from Eq. (1.46) that
∂vi ∂ri
= . (1.51)
∂ q̇ j ∂q j

“M01 Goldstein ISBN C01” — 2011/2/15 — page 19 — #19

20 Chapter 1 Survey of the Elementary Principles

Substitution of these changes in (1.50) leads to the result that

# ∂ri #* d ' ∂vi
(
∂vi
+
m i r̈i ? = m i vi ? − m i vi ? ,
∂q j dt ∂ q̇ j ∂q j
i i

and the second term on the left-hand side of Eq. (1.45) can be expanded into
, - % &. % & /
# d ∂ #1 ∂ #1
2 2
m i vi − m i vi − Q j δq j .
dt ∂ q̇ j 2 ∂q j 2
j i i

$
Identifying i 21 m i vi2 with the system kinetic energy T , D’Alembert’s principle
(cf. Eq. (1.45)) becomes
# 0* d ' ∂ T ( ∂ T + 1
− − Q j δq j = 0. (1.52)
dt ∂ q̇ j ∂q j
j

Note that in a system of Cartesian coordinates the partial derivative of T with

respect to q j vanishes. Thus, speaking in the language of differential geometry,
this term arises from the curvature of the coordinates q j . In polar coordinates,
e.g., it is in the partial derivative of T with respect to the angle coordinate that the
centripetal acceleration term appears.
Thus far, no restriction has been made on the nature of the constraints other
than that they be workless in a virtual displacement. The variables q j can be any
set of coordinates used to describe the motion of the system. If, however, the con-
straints are holonomic, then it is possible to find sets of independent coordinates
q j that contain the constraint conditions implicitly in the transformation equations
(1.38). Any virtual displacement δq j is then independent of δqk , and therefore the
only way for (1.52) to hold is for the individual coefficients to vanish:
' (
d ∂T ∂T
− = Q j. (1.53)
dt ∂ q̇ j ∂q j

There are n such equations in all.

When the forces are derivable from a scalar potential function V ,

Fi = −∇i V.

Then the generalized forces can be written as

# ∂ri # ∂ri
Qj = Fi ? =− ∇i V ? ,
∂q j ∂q j
i i

which is exactly the same expression for the partial derivative of a function
−V (r1 , r2 , . . . , r N , t) with respect to q j :

∂V
Qj ≡ − . (1.54)
∂q j

“M01 Goldstein ISBN C01” — 2011/2/15 — page 20 — #20

 1.5 Velocity-Dependent Potentials and the Dissipation Function 21

See Eq. (1.47). Equations (1.53) can then be rewritten as

' (
d ∂T ∂(T − V )
− = 0. (1.55)
dt ∂ q̇ j ∂q j

The equations of motion in the form (1.55) are not necessarily restricted to conser-
vative systems; only if V is not an explicit function of time is the system conserva-
tive (cf. p. 4). As here defined, the potential V does not depend on the generalized
velocities. Hence, we can include a term in V in the partial derivative with respect
to q̇ j :
' (
d ∂(T − V ) ∂(T − V )
− = 0.
dt ∂ q̇ j ∂q j

Or, defining a new function, the Lagrangian L, as

L = T − V, (1.56)

the Eqs. (1.53) become

' (
d ∂L ∂L
− = 0, (1.57)
dt ∂ q̇ j ∂q j

expressions referred to as “Lagrange’s equations.”

Note that for a particular set of equations of motion there is no unique choice
of Lagrangian such that Eqs. (1.57) lead to the equations of motion in the given
generalized coordinates. Thus, in Derivations 8 and 10 it is shown that if L(q, q̇, t)
is an approximate Lagrangian and F(q, t) is any differentiable function of the
generalized coordinates and time, then
dF
L ′ (q, q̇, t) = L(q, q̇, t) + (1.57′ )
dt
is a Lagrangian also resulting in the same equations of motion. It is also often
possible to find alternative Lagrangians beside those constructed by this prescrip-
tion (see Exercise 20). While Eq. (1.56) is always a suitable way to construct a
Lagrangian for a conservative system, it does not provide the only Lagrangian
suitable for the given system.

1.5 VELOCITY-DEPENDENT POTENTIALS

AND THE DISSIPATION FUNCTION

Lagrange’s equations can be put in the form (1.57) even if there is no potential
function, V , in the usual sense, providing the generalized forces are obtained from
a function U (q j , q̇ j ) by the prescription
' (
∂U d ∂U
Qj = − + . (1.58)
∂q j dt ∂ q̇ j

“M01 Goldstein ISBN C01” — 2011/2/15 — page 21 — #21

22 Chapter 1 Survey of the Elementary Principles

In such case, Eqs. (1.57) still follow from Eqs. (1.53) with the Lagrangian given
by

L = T − U. (1.59)

Here U may be called a “generalized potential,” or “velocity-dependent poten-

tial.” The possibility of using such a “potential” is not academic; it applies to one
very important type of force field, namely, the electromagnetic forces on moving
charges. Considering its importance, a digression on this subject is well worth-
while.
Consider an electric charge, q, of mass m moving at a velocity, v, in an other-
wise charge-free region containing both an electric field, E, and a magnetic field,
B, which may depend upon time and position. The charge experiences a force,
called the Lorentz force, given by

F = q [E + (v × B)] . (1.60)

Both E(x, y, z, t) and B(x, y, z, t) are continuous functions of time and position
derivable from a scalar potential φ(x, y, z, t) and a vector potential A(x, y, z, t)
by

∂A
E = −∇φ − (1.61a)
∂t
and

B = ∇ × A. (1.61b)

The force on the charge can be derived from the following velocity-dependent
potential energy

U = qφ − qA ? v, (1.62)

so the Lagrangian, L = T − U , is

L = 12 mv 2 − qφ + qA ? v. (1.63)

Considering just the x-component of Lagrange’s equations gives

' ( ' (
∂ Ax ∂ Ay ∂ Az ∂φ d Ax
m ẍ = q vx + vy + vz −q + . (1.64)
∂x ∂x ∂x ∂x dt

The total time derivative of A x is related to the partial time derivative through

d Ax ∂ Ax
= + v ? ∇ Ax
dt ∂t
∂ Ax ∂ Ax ∂ Ax ∂ Ax
= + vx + vy + vz . (1.65)
∂t ∂x ∂y ∂z

“M01 Goldstein ISBN C01” — 2011/2/15 — page 22 — #22

 1.5 Velocity-Dependent Potentials and the Dissipation Function 23

Equation (1.61b) gives

' ( ' (
∂ Ay ∂ Ax ∂ Az ∂Ax
(v × B)x = v y − + vz − .
∂x ∂y ∂x ∂z
Combining these expressions gives the equation of motion in the x-direction

m ẍ = q [Ex + (v × B)x ] . (1.66)

On a component-by-component comparison, Eqs. (1.66) and (1.60) are identical,

showing that the Lorentz force equation is derivable from Eqs. (1.61) and (1.62).
Note that if not all the forces acting on the system are derivable from a poten-
tial, then Lagrange’s equations can always be written in the form
' (
d ∂L ∂L
− = Q j,
dt ∂ q̇ j ∂q j

where L contains the potential of the conservative forces as before, and Q j rep-
resents the forces not arising from a potential. Such a situation often occurs when
frictional forces are present. It frequently happens that the frictional force is pro-
portional to the velocity of the particle, so that its x-component has the form

F f x = −k x vx .

Frictional forces of this type may be derived in terms of a function F, known as

Rayleigh’s dissipation function, and defined as
1 #2 2 3
F= k x vi x + k y vi2y + k z vi2z , (1.67)
2
i

where the summation is over the particles of the system. From this definition it is
clear that
∂F
F f xi = − ,
∂vxi

or, symbolically,

F f = −∇v F. (1.68)

We can also give a physical interpretation to the dissipation function. The work
done by the system against friction is
2 3
dW f = −F f ? dr = −F f ? v dt = k x vx2 + k y v 2y + k z vz2 dt.

Hence, 2F is the rate of energy dissipation due to friction. The component of the
generalized force resulting from the force of friction is then given by
# ∂ri # ∂ri
Qj = F fi ? =− ∇v F ?
∂q j ∂q j
i

“M01 Goldstein ISBN C01” — 2011/2/15 — page 23 — #23

24 Chapter 1 Survey of the Elementary Principles

# ∂ ṙi
=− ∇v F ? , by (1.51),
∂ q̇ j
∂F
=− . (1.69)
∂ q̇ j
An example is Stokes’ law, whereby a sphere of radius a moving at a speed v,
in a medium of viscosity η experiences the frictional drag force F f = −6π ηav.
The Lagrange equations with dissipation become
' (
d ∂L ∂L ∂F
− + = 0, (1.70)
dt ∂ q̇ j ∂q j ∂ q̇ j
so that two scalar functions, L and F, must be specified to obtain the equations
of motion.

1.6 SIMPLE APPLICATIONS OF THE LAGRANGIAN FORMULATION

The previous sections show that for systems where we can define a Lagrangian,
i.e., holonomic systems with applied forces derivable from an ordinary or gen-
eralized potential and workless constraints, we have a very convenient way of
setting up the equations of motion. We were led to the Lagrangian formulation
by the desire to eliminate the forces of constraint from the equations of motion,
and in achieving this goal we have obtained many other benefits. In setting up the
original form of the equations of motion, Eqs. (1.19), it is necessary to work with
many vector forces and accelerations. With the Lagrangian method we only deal
with two scalar functions, T and V , which greatly simplifies the problem.
A straightforward routine procedure can now be established for all problems
of mechanics to which the Lagrangian formulation is applicable. We have only to
write T and V in generalized coordinates, form L from them, and substitute in
(1.57) to obtain the equations of motion. The needed transformation of T and V
from Cartesian coordinates to generalized coordinates is obtained by applying the
transformation equations (1.38) and (1.45′ ). Thus, T is given in general by
⎛ ⎞2
#1 #1 # ∂ri ∂ri⎠
T = m i vi2 = mi ⎝ q̇ j + .
2 2 ∂q j ∂t
i i j

It is clear that on carrying out the expansion, the expression for T in generalized
coordinates will have the form
# 1#
T = M0 + M j q̇ j + M jk q̇ j q̇k , (1.71)
2
j j,k

where M0 , M j , M jk are definite functions of the r’s and t and hence of the q’s
and t. In fact, a comparison shows that
# 1 ' ∂ri (2
M0 = mi ,
2 ∂t
i

“M01 Goldstein ISBN C01” — 2011/2/15 — page 24 — #24

 1.6 Simple Applications of the Lagrangian Formulation 25
# ∂ri ∂ri
Mj = mi ? , (1.72)
∂t ∂q j
i

and
# ∂ri ∂ri
M jk = mi ? .
∂q j ∂qk
i

Thus, the kinetic energy of a system can always be written as the sum of three
homogeneous functions of the generalized velocities,

T = T0 + T1 + T2 , (1.73)

where T0 is independent of the generalized velocities, T1 is linear in the velocities,

and T2 is quadratic in the velocities. If the transformation equations do not contain
the time explicitly, as may occur when the constraints are independent of time
(scleronomous), then only the last term in Eq. (1.71) is nonvanishing, and T is
always a homogeneous quadratic form in the generalized velocities.
Let us now consider simple examples of this procedure:

1. Single particle in space

(a) Cartesian coordinates
(b) Plane polar coordinates
2. Atwood’s machine
3. Time-dependent constraint—bead sliding on rotating wire

1. (a) Motion of one particle: using Cartesian coordinates. The generalized

forces needed in Eq. (1.53) are obviously Fx , Fy , and Fz . Then
8 9
T = 12 m ẋ 2 + ẏ 2 + ż 2 ,
∂T ∂T ∂T
= = = 0,
∂x ∂y ∂z
∂T ∂T ∂T
= m ẋ, = m ẏ, = m ż,
∂ ẋ ∂ ẏ ∂ ż
and the equations of motion are
d d d
(m ẋ) = Fx , (m ẏ) = Fy , (m ż) = Fz . (1.74)
dt dt dt
We are thus led back to the original Newton’s equations of motion.
(b) Motion of one particle: using plane polar coordinates. Here we must
express T in terms of ṙ and θ̇ . The transformation equations, Eqs. (1.38), are

x = r cos θ
y = r sin θ.

“M01 Goldstein ISBN C01” — 2011/2/15 — page 25 — #25

26 Chapter 1 Survey of the Elementary Principles

By analogy to (1.46), the velocities are given by

ẋ = ṙ cos θ − r θ̇ sin θ,
ẏ = ṙ sin θ + r θ̇ cos θ.

The kinetic energy T = 12 m(ẋ 2 + ẏ 2 ) then reduces formally to

: 8 92 ;
T = 12 m ṙ 2 + r θ̇ . (1.75)

An alternative derivation of Eq. (1.75) is obtained by recognizing that the plane

polar components of the velocity are ṙ along r, and r θ̇ along the direction per-
pendicular to r , denoted by the unit vector û. Hence, the square of the velocity
expressed in polar coordinates is simply ṙ 2 +(r θ̇)2 . With the aid of the expression

dr = r̂ dr + r û dθ + k̂ dz

for the differential position vector, dr, in cylindrical coordinates, restricted to the
plane z = 0 where r̂ and û are unit vectors in the r and u-directions, respec-
tively, the components of the generalized force can be obtained from the defini-
tion, Eq. (1.49),
∂r
Qr = F ? = F ? r̂ = Fr ,
∂r
∂r
Qθ = F ? = F ? r û = r Fθ ,
∂θ
since the derivative of r with respect to θ is, by the definition of a derivative, a
vector in the direction of θ̂ (cf. Fig. 1.6). There are two generalized coordinates,
and therefore two Lagrange equations. The derivatives occurring in the r equation
are
' (
∂T ∂T d ∂T
= mr θ̇ 2 , = m ṙ , = m r̈ ,
∂r ∂ ṙ dt ∂ ṙ

FIGURE 1.6 Derivative of r with respect to θ.

“M01 Goldstein ISBN C01” — 2011/2/15 — page 26 — #26

 1.6 Simple Applications of the Lagrangian Formulation 27

and the equation itself is

m r̈ − mr θ̇ 2 = Fr ,

the second term being the centripetal acceleration term. For the θ equation, we
have the derivatives

∂T ∂T d 2 2 3
= 0, = mr 2 θ̇, mr θ̇ = mr 2 θ̈ + 2mr ṙ θ̇ ,
∂θ ∂ θ̇ dt

so that the equation becomes

d 2 2 3
mr θ̇ = mr 2 θ̈ + 2mr ṙ θ̇ = r Fθ .
dt

Note that the left side of the equation is just the time derivative of the angular
momentum, and the right side is exactly the applied torque, so that we have simply
rederived the torque equation (1.26), where L = mr 2 θ̇ and N (e) = r Fθ .
2. Atwood’s machine—(See Fig. 1.7) an example of a conservative system
with holonomic, scleronomous constraint (the pulley is assumed frictionless and
massless). Clearly there is only one independent coordinate x, the position of
the other weight being determined by the constraint that the length of the rope
between them is l. The potential energy is

V = −M1 gx − M2 g(l − x),

while the kinetic energy is

T = 1
2 (M1 + M2 ) ẋ 2 .

FIGURE 1.7 Atwood’s machine.

“M01 Goldstein ISBN C01” — 2011/2/15 — page 27 — #27

28 Chapter 1 Survey of the Elementary Principles

Combining the two, the Lagrangian has the form

L =T −V = 1
2 (M1 + M2 ) ẋ 2 + M1 gx + M2 g(l − x).

There is only one equation of motion, involving the derivatives

∂L
= (M1 − M2 ) g,
∂x
∂L
= (M1 + M2 ) ẋ,
∂ ẋ
so that we have

(M1 + M2 ) ẍ = (M1 − M2 ) g,

or
M1 − M2
ẍ = g,
M1 + M2
which is the familiar result obtained by more elementary means. This trivial prob-
lem emphasizes that the forces of constraint—here the tension in the rope—
appear nowhere in the Lagrangian formulation. By the same token, neither can
the tension in the rope be found directly by the Lagrangian method.
3. A bead (or ring) sliding on a uniformly rotating wire in a force-free space.
The wire is straight, and is rotated uniformly about some fixed axis perpendicular
to the wire. This example has been chosen as a simple illustration of a constraint
being time dependent, with the rotation axis along z and the wire in the x y plane.
The transformation equations explicitly contain the time.

x = r cos ωt, (ω = angular velocity of rotation)

y = r sin ωt. (r = distance along wire from rotation axis)

While we could then find T (here the same as L) by the same procedure used to
obtain (1.71), it is simpler to take over (1.75) directly, expressing the constraint
by the relation θ̇ = ω:
8 9
T = 12 m ṙ 2 + r 2 ω2 .

Note that T is not a homogeneous quadratic function of the generalized velocities,

since there is now a term not involving ṙ . The equation of motion is then

m r̈ − mr ω2 = 0

or

r̈ = r ω2 ,

which is the familiar simple harmonic oscillator equation with a change of sign.
The solution r = eωt for a bead initially at rest on the wire shows that the

“M01 Goldstein ISBN C01” — 2011/2/15 — page 28 — #28

 Derivations 29

bead moves exponentially outwards. Again, the method cannot furnish the force
of constraint that keeps the bead on the wire. Equation (1.26) with the angular
momentum, L = mr 2 ω = mωr02 e2ωt , provides the force F = N /r , which pro-
duces the constraint force, F = 2mr0 ω2 eωt , acting perpendicular to the wire and
the axis of rotation.

DERIVATIONS

1. Show that for a single particle with constant mass the equation of motion implies the
following differential equation for the kinetic energy:
dT
= F ? v,
dt
while if the mass varies with time the corresponding equation is
d(mT )
= F ? p.
dt
2. Prove that the magnitude R of the position vector for the center of mass from an
arbitrary origin is given by the equation
# 1#
M 2 R2 = M m i ri2 − m i m j ri2j .
2
i i̸= j

3. Suppose a system of two particles is known to obey the equations of motion,

Eqs. (1.22) and (1.26). From the equations of the motion of the individual parti-
cles show that the internal forces between particles satisfy both the weak and the
strong laws of action and reaction. The argument may be generalized to a system with
arbitrary number of particles, thus proving the converse of the arguments leading to
Eqs. (1.22) and (1.26).
4. The equations of constraint for the rolling disk, Eqs. (1.39), are special cases of
general linear differential equations of constraint of the form
n
#
gi (x1 , . . . , xn )d xi = 0.
i=1
A constraint condition of this type is holonomic only if an integrating function
f (x1 , . . . , xn ) can be found that turns it into an exact differential. Clearly the func-
tion must be such that
8 9
∂ ( f gi ) ∂ fgj
=
∂x j ∂ xi
for all i ̸= j. Show that no such integrating factor can be found for either of
Eqs. (1.39).
5. Two wheels of radius a are mounted on the ends of a common axle of length b such
that the wheels rotate independently. The whole combination rolls without slipping on
a plane. Show that there are two nonholonomic equations of constraint,
cos θ d x + sin θ dy = 0
8 9
sin θ d x − cos θ dy = 12 a dφ + dφ ′ ,

“M01 Goldstein ISBN C01” — 2011/2/15 — page 29 — #29

30 Chapter 1 Survey of the Elementary Principles

(where θ, φ, and φ ′ have meanings similar to those in the problem of a single vertical
disk, and (x, y) are the coordinates of a point on the axle midway between the two
wheels) and one holonomic equation of constraint,
a
θ =C− (φ − φ ′ ),
b
where C is a constant.
6. A particle moves in the x y plane under the constraint that its velocity vector is always
directed towards a point on the x axis whose abscissa is some given function of time
f (t). Show that for f (t) differentiable, but otherwise arbitrary, the constraint is non-
holonomic.
7. Show that Lagrange’s equations in the form of Eqs. (1.53) can also be written as

∂ Ṫ ∂T
−2 = Q j.
∂ q̇ j ∂q j

These are sometimes known as the Nielsen form of the Lagrange equations.
8. If L is a Lagrangian for a system of n degrees of freedom satisfying Lagrange’s equa-
tions, show by direct substitution that

d F(q1 , . . . , qn , t)
L′ = L +
dt
also satisfies Lagrange’s equations where F is any arbitrary, but differentiable, func-
tion of its arguments.
9. The electromagnetic field is invariant under a gauge transformation of the scalar and
vector potential given by

A → A + !ψ(r, t),
1 ∂ψ
φ→φ− ,
c ∂t
where ψ is arbitrary (but differentiable). What effect does this gauge transformation
have on the Lagrangian of a particle moving in the electromagnetic field? Is the motion
affected?
10. Let q1 , . . . , qn be a set of independent generalized coordinates for a system of n
degrees of freedom, with a Lagrangian L(q, q̇, t). Suppose we transform to another
set of independent coordinates s1 , . . . , sn by means of transformation equations

qi = qi (s1 , . . . , sn , t), i = 1, . . . , n.

(Such a transformation is called a point transformation.) Show that if the Lagrangian

function is expressed as a function of s j , ṡ j , and t through the equations of transfor-
mation, then L satisfies Lagrange’s equations with respect to the s coordinates:
' (
d ∂L ∂L
− = 0.
dt ∂ ṡ j ∂s j

In other words, the form of Lagrange’s equations is invariant under a point transfor-
mation.

“M01 Goldstein ISBN C01” — 2011/2/15 — page 30 — #30

 Exercises 31

EXERCISES

11. Check whether the force F⃗ = yz î + zx ĵ + x y k̂ is conservative or not.

12. Compute the orbital period and orbital angular velocity of a satellite revolving around
the Earth at an altitude of 720 km. [Given: radius of Earth R = 6000 km and
g = 9.83 m/s2 .]
13. Rockets are propelled by the momentum reaction of the exhaust gases expelled from
the tail. Since these gases arise from the reaction of the fuels carried in the rocket, the
mass of the rocket is not constant, but decreases as the fuel is expended. Show that the
equation of motion for a rocket projected vertically upward in a uniform gravitational
field, neglecting atmospheric friction, is
dv dm
m = −v ′ − mg,
dt dt
where m is the mass of the rocket and v ′ is the velocity of the escaping gases relative to
the rocket. Integrate this equation to obtain v as a function of m, assuming a constant
time rate of loss of mass. Show, for a rocket starting initially from rest, with v ′ equal
to 2.1 km/s and a mass loss per second equal to 1/60th of the initial mass, that in
order to reach the escape velocity the ratio of the weight of the fuel to the weight of
the empty rocket must be almost 300!
14. Two points of mass m are joined by a rigid weightless rod of length l, the center of
which is constrained to move on a circle of radius a. Express the kinetic energy in
generalized coordinates.
15. A point particle moves in space under the influence of a force derivable from a gener-
alized potential of the form

U (r, v) = V (r ) + s ? L,

where r is the radius vector from a fixed point, L is the angular momentum about that
point, and s is a fixed vector in space.
(a) Find the components of the force on the particle in both Cartesian and spherical
polar coordinates, on the basis of Eq. (1.58).
(b) Show that the components in the two coordinate systems are related to each other
as in Eq. (1.49).
(c) Obtain the equations of motion in spherical polar coordinates.
16. A particle moves in a plane under the influence of a force, acting toward a center of
force, whose magnitude is
% &
1 ṙ 2 − 2r̈r
F = 2 1− ,
r c2

where r is the distance of the particle to the center of force. Find the generalized
potential that will result in such a force, and from that the Lagrangian for the motion
in a plane. (The expression for F represents the force between two charges in Weber’s
electrodynamics.)
17. A nucleus, originally at rest, decays radioactively by emitting an electron of momen-
tum 1.73 MeV/c, and at right angles to the direction of the electron a neutrino with

“M01 Goldstein ISBN C01” — 2011/2/15 — page 31 — #31

32 Chapter 1 Survey of the Elementary Principles

momentum 1.00 MeV/c. (The MeV, million electron volt, is a unit of energy used in
modern physics, equal to 1.60 × 10−13 J. Correspondingly, MeV/c is a unit of lin-
ear momentum equal to 5.34 × 10−22 kg ? m/s.) In what direction does the nucleus
recoil? What is its momentum in MeV/c? If the mass of the residual nucleus is
3.90 × 10−25 kg what is its kinetic energy, in electron volts?
18. A Lagrangian for a particular physical system can be written as
m8 2 9 K 8 2 9
L′ = a ẋ + 2b ẋ ẏ + c ẏ 2 − ax + 2bx y + cy 2 ,
2 2

where a, b, and c are arbitrary constants but subject to the condition that b2 − ac ̸= 0.
What are the equations of motion? Examine particularly the two cases a = 0 = c
and b = 0, c = −a. What is the physical system described by the above Lagrangian?
Show that the usual Lagrangian for this system as defined by Eq. (1.56) is related to L ′
by a point transformation (cf. Derivation 10). What is the significance of the condition
on the value of b2 − ac?
19. Obtain the Lagrange equations of motion for a spherical pendulum, i.e., a mass point
suspended by a rigid weightless rod.
20. A particle of mass m moves in one dimension such that it has the Lagrangian

m 2 ẋ 4
L= + m ẋ 2 V (x) − V 2 (x),
12
where V is some differentiable function of x. Find the equation of motion for x(t) and
describe the physical nature of the system on the basis of this equation.
21. Two mass points of mass m 1 and m 2 are connected by a string passing through a
hole in a smooth table so that m 1 rests on the table surface and m 2 hangs suspended.
Assuming m 2 moves only in a vertical line, what are the generalized coordinates for
the system? Write the Lagrange equations for the system and, if possible, discuss
the physical significance any of them might have. Reduce the problem to a single
second-order differential equation and obtain a first integral of the equation. What is
its physical significance? (Consider the motion only until m 1 reaches the hole.)
22. Obtain the Lagrangian and equations of motion for the double pendulum illustrated in
Fig. 1.4, where the lengths of the pendula are l1 and l2 with corresponding masses m 1
and m 2 .
23. Two masses 2 kg and 3 kg, respectively, are tied to the two ends of a massless,
inextensible string passing over a smooth pulley. When the system is released,
calculate the acceleration of the masses and the tension in the string.
24. A spring of rest length L a (no tension) is connected to a support at one end and has
a mass M attached at the other. Neglect the mass of the spring, the dimension of the
mass M, and assume that the motion is confined to a vertical plane. Also, assume that
the spring only stretches without bending but it can swing in the plane.
(a) Using the angular displacement of the mass from the vertical and the length
that the string has stretched from its rest length (hanging with the mass m), find
Lagrange’s equations.
(b) Solve these equations for small stretching and angular displacements.
(c) Solve the equations in part (a) to the next order in both stretching and angular
displacement. This part is amenable to hand calculations. Using some reasonable

“M01 Goldstein ISBN C01” — 2011/2/15 — page 32 — #32

 Exercises 33

assumptions about the spring constant, the mass, and the rest length, discuss the
motion. Is a resonance likely under the assumptions stated in the problem?
(d) (For analytic computer programs.) Consider the spring to have a total mass
m ≪ M. Neglecting the bending of the spring, set up Lagrange’s equations
correctly to first order in m and the angular and linear displacements.
(e) (For numerical computer analysis.) Make sets of reasonable assumptions of the
constants in part (a) and make a single plot of the two coordinates as functions of
time.

“M01 Goldstein ISBN C01” — 2011/2/15 — page 33 — #33