Chapter 1

Lagranges equations
Starting with dAlemberts principle, we now arrive at one of the most elegant and useful
formulations of classical mechanics, generally referred to as Lagranges equations. This treatment is taken from Goldsteins graduate mechanics text, as his treatment seems somewhat
more clear to me than Sommerfelds.

1.1

Lagranges equations from dAlemberts principle

We begin with dAlemberts principle written in its most fundamental and general form,
X
(Fi + Fi )xi = 0
(1.1)
i

where the subscript i ranges over all three components of all particles involved in the system
of interest. The first step is to rewrite the particle positions, represented by the xi in groups
of three for each particle, in terms of independent generalized coordinates qj . If there are
constraints in the system, then there are fewer q variables than x variables. For example, a
wheel rotating on a fixed axle has only one q, the angle of rotation, whereas there are three
times as many x variables as there are atoms in the wheel.
For holonomic constraints we can write
xi = xi (qj , t)

(1.2)

where we allow for the possibility that the relationship between the q and x variables to
depend on time.
We can rewrite dAlemberts principle by noting that
xi =

X xi
j

qj

qj

(1.3)

CHAPTER 1. LAGRANGES EQUATIONS

where the time dependence is not exercised since virtual changes are assumed to take place
at a fixed time. Thus,
X
X
X xi
qj =
Qj qj
(1.4)
Fi xi =
Fi
qj
j
i
i,j
where the
Qj =

Fi

xi
qj

(1.5)

are called the generalized forces. Notice that just as the qi need not have the units of length,
the Qi need not have the units of force. However, the product must have the units of energy.
For instance, if q is a dimensionless angle, then the corresponding Q would be a torque,
which has energy units.
Turning to the inertial forces in dAlemberts principle, we note that
X
i

Fi xi =

mi

X d2 xi xi
d 2 xi
x
=

qj
mi 2
i
dt2
dt
q
j
i,j

(1.6)

where we have used equation (1.3) in the last step. Using the product rule backwards, we
see that


X d2 xi xi X  d  xi 
d xi
=
mi
mi 2
x i
x i
(1.7)
dt qj
dt
qj
dt qj
i
i
To make further progress, we take the total time derivative of equation (1.2), resulting in
x i =

X xi
j

qj

qj +

xi
t

(1.8)

where x i = dxi /dt and qj = dqj /dt. The x i are the actual particle velocity components.
We call the qj the generalized velocities. Taking the partial derivative of equation (1.8) with
respect to a particular qj , we immediately conclude that
xi
x i
=
,
qj
qj

(1.9)

where the derivative of the second term in equation (1.8) is zero because the velocities are
not functions of the positions at an instant in time. (Ultimately, the positions can be derived
from the velocities by integration, but this relationship depends on knowing the complete
time history of the velocities, not just the values at a particular time.)
Finally, we note that
d
dt

xi
qj


=

x i
.
qj

(1.10)

To show this, change the dummy summation index in equation (1.8) from j to k to avoid
confusion, and take the partial derivative of this equation with respect to qj :
x i X 2 xi
2 xi
=
qk +
.
qj
q
q
j qk
j t
k

(1.11)

CHAPTER 1. LAGRANGES EQUATIONS

This is possible again because qk is not an explicit function of the qj . Then compare this
with

 X 2
d xi
xi
2 xi
=
qk +
.
(1.12)
dt qj
qk qj
tqj
k
Aside from the order of partial derivatives, the right sides of equations (1.11) and (1.12) are
identical, which proves equation (1.10).
Substituting equations (1.9) and (1.10) in equation (1.7) results in


X d2 xi xi X  d  x i 
x i
mi 2
=
x i
x i
mi
dt
q
dt

qj
j
j
i
i
X mi x 2i
d X mi x 2i

=
dt qj i
2
qj i
2


d T
T
=

dt qj
qj

(1.13)

where we recognize the sums as the total kinetic energy T of the system.
Combining equations (1.4), (1.6), and (1.13) yields



X
T
d T
+
qj = 0.
Qj
dt qj
qj
j

(1.14)

Since the qi are independent of each other, the coefficients of the qi are individually zero,
resulting in Lagranges equations:


d T
T

= Qj .
(1.15)
dt qj
qj
Often forces are conservative and possible to represent as the gradient of a potential energy
Fi = V /xi . Starting with the definition of generalized forces in equation (1.5), we find
that
X V xi
X xi
V
=
=
.
(1.16)
Qj =
Fi
q
x
q
q
j
i
j
j
i
i
If in addition, V is not an explicit function of time or of the generalized velocities, equation
(1.15) may be written


d L
L

= 0,
(1.17)
dt qj
qj
where L = T V is called the Lagrangian. The lack of dependence on time and the
generalized velocities allows the V to be incorporated in the first as well as the second terms
of this equation. If some of the forces are conservative and others are not, then the more
general form


d L
L

= Qj
(1.18)
dt qj
qj
may be used.

CHAPTER 1. LAGRANGES EQUATIONS

The quantities
pj =

L
qj

(1.19)

are called the generalized momenta. Note that when the Lagrangian is not a function of a
particular generalized coordinate and the associated non-conservative force Qj is zero, then
the associated generalized momentum is conserved, since equation (1.18) reduces to
dpj
= 0.
dt

(1.20)

To summarize, these equations are valid for systems obeying the following conditions:
1. The constraints on the system are holonomic, so that the qj are independent for both
finite and infinitesimal displacements. The constraints may be time dependent.
2. The potential energy V is a function only of the qj . If there are forces for which no
such potential exists, then they can be included on the right side of equation (1.18) in
the Qj .

1.2

Examples of use

We now look at several examples to see how Lagranges equations are used.

1.2.1

Simple pendulum

We start with the simple pendulum, a problem that is easily solvable by elementary methods.
Imagine the pendulum arm to have negligible mass and length D. A mass M is attached to
the one end of the arm and the other is attached to a support that allows the pendulum to
pivot freely in the x-z plane. The angle between the pendulum arm and the vertical is .
Taking zero elevation as the pivot point, the potential energy of the pendulum mass is
V = M gz = M gD cos .

(1.21)

Motion in the x-z plane is constrained to be in the form of a circular arc of radius D, which
means that the kinetic energy of the pendulum is
1
T = M D2 2 ,
2

(1.22)

which means that the Lagrangian function is

= T V = 1 M D2 2 + M gD cos .
L(, )
2

(1.23)

CHAPTER 1. LAGRANGES EQUATIONS

D
x1

z1

x2

z2

Figure 1.1: Sketch for the double pendulum.

The generalized coordinate (only one, as there is only one effective degree of freedom) is .
Recall that and are considered to be independent variables in Lagrangian dynamics, so
L
= M D2

and

d
dt

= M D2

(1.24)

d2
.
dt2

(1.25)

The second term in Lagranges equation is also easily calculated:

L
= M gD sin .

(1.26)

Combining equations (1.25) and (1.26), we arrive at the governing equation for the pendulum:
 
2
d L
L
2d

= M D 2 + M gD sin = 0
(1.27)
dt

dt
which simplifies to the usual form
d2
g
+ sin = 0.
2
dt
D

(1.28)

For small we have sin and we have a harmonic oscillator equation for the pendulum.

1.2.2

Double pendulum

A particular form of the double pendulum is illustrated in figure 1.1. The masses are free
to swing in the x-z plane, with the second pendulum swinging from the bob on the first
pendulum. (Other forms include a version with massive rods instead of weights attached to
rods of negligible mass.)

CHAPTER 1. LAGRANGES EQUATIONS

The Cartesian coordinates of the two masses are related to the angles and as follows
(x1 , z1 ) = (D sin , D sin )

(1.29)

(x2 , z2 ) = [D(sin + sin ), D(cos + cos )

(1.30)

and
where the origin of the coordinate system is located where the pendulum attaches to the
ceiling. The kinetic energies of the two pendulums are
1
T1 = m(x 21 + z12 )
2
1
= mD2 2
2

(1.31)

1
T2 = m(x 22 + z22 )
2
h
i
1
= mD2 2 + 2 + 2 cos( )
2

(1.32)

and the potential energy of the two pendulum bobs together is

V = mg(z1 + z2 ) = mgD (2 cos + cos ) .

(1.33)

The angles and and their time derivatives are the generalized coordinates and velocities
and the Lagrangian is
h
i
1
(1.34)
L = mD2 2 2 + 2 + 2 cos( ) + mgD (2 cos + cos ) .
2
From

d
dt

L
=0

d
dt

L
=0

(1.35)

we get the governing equations

2
and

d2
d2
2g
+
cos(

)
+ sin( )2 +
sin = 0
2
2
dt
dt
D

(1.36)

d2
d2
g
+
cos(

)
sin( ) 2 + sin = 0.
(1.37)
2
2
dt
dt
D
Thus, two degrees of freedom result in two coupled governing equations. These happen to
be complicated nonlinear equations and the physical system demonstrates chaotic behavior
for large amplitude oscillations. Behavior at small amplitudes is more benign and will be
investigated in the next chapter.

CHAPTER 1. LAGRANGES EQUATIONS

Figure 1.2: Schematic of the motion of a puck on an air table constrained by a string to
which a force F is applied.

1.2.3

Air table problem

We now analyze a problem with a non-conservative, externally applied force. Imagine a puck
of mass m sliding frictionlessly on a horizontal air table with a string attached that passes
through a hole in the air table. A force F that may be varied at will is applied to the string,
constraining the puck to move in some orbit around the hole.
There is no potential energy in this problem, so the Lagrangian is simply the kinetic energy.
Referring to figure 1.2,
1
1
(1.38)
L = T = mr 2 + mr2 2 .
2
2
The generalized force in the +r direction is just Qr = F , since a positive F is toward the
origin. There are two degrees of freedom for this problem, r and . The r component of
Lagranges equation is therefore
 
d L
L
d2 r

= m 2 mr 2 = F
(1.39)
dt r
r
dt
and the component is
d
dt

L
d  2 
=
mr = 0.

dt

(1.40)

Notice that L/ = 0, with the consequence that the generalized momentum p = mr2 is
constant in time, i.e., it is conserved. We identify this quantity as the angular momentum
about an axis of rotation normal to the air table and located at the hole in the table. Solving
for and substituting in equation (1.39) results in
p2
d2 r
m 2
= F.
dt
mr3

(1.41)

In equilibrium with constant r, this reduces to an equation for F ,

F = mr 2 ,
which we recognize as the centripetal force for circular motion.

(1.42)

CHAPTER 1. LAGRANGES EQUATIONS

spring

m
x

wire

Figure 1.3: Sketch of a mass moving along a wire with a spring force.
This is an example of a general phenomenon with Lagrangian dynamics: if the Lagrangian
doesnt depend on a particular generalized coordinate, in this case , then there exists a
conserved quantity which equals the partial derivative of the Lagrangian with respect to the
associated generalized velocity variable.
A Lagrangian that does not depend on a generalized coordinate q is symmetric to changes
in this coordinate, q q + q. In the above example, such changes correspond to a rotation
about the above-defined axis. This is the classical mechanics analog to the corresponding idea
in quantum mechanics that symmetry under a transformation is related to the conservation
of a dynamic variable.

1.3

Problems

1. Hoop rolling down a ramp: A hoop of radius R and mass m rolls down a ramp of
inclination without slipping. Derive the Lagrangian for the hoop and use it to
determine the acceleration of the hoop down the ramp. Make a sketch of the setup
and the parameters that you use.
2. A mass m with a hole in it slides frictionlessly on a straight wire as shown in figure
1.3. The mass is connected to a wall a distance d from the wire by a spring with spring
constant k and the unstretched length of the spring is also d.
(a) Assuming that gravity does not act, derive the equation for the movement of the
mass along the wire using the method of Lagrange.
(b) Approximate this equation for x2  d2 .
3. A disk of radius a and mass M is rolling around the inside of a cylinder of radius b as
shown in figure 1.4.

CHAPTER 1. LAGRANGES EQUATIONS

Figure 1.4: Disk of radius a rolling around inside a cylinder of radius b.

side view

top view
y
m
l

m
z

Figure 1.5: Sketch for spherical pendulum.

(a) Assuming the disk rolls without slipping, determine the relationship between
and .
(b) Assuming that is defined to be zero when = 0, write down the Lagrangian for

this system in terms of and .

(c) Derive the governing equation.
4. Spherical pendulum. A mass m which hangs from a string of length l attached to the
ceiling can move in both horizontal directions as illustrated in figure 1.5. Its position
is specified by the azimuthal angle and the elevation angle .
and .

(a) Find the Lagrangian for the motion of the mass in terms of , , ,
(b) Derive the two Lagrange equations for and .
(c) Note that the Lagrangian does not depend on . Derive the corresponding conserved generalized momentum. Show that this quantity is the vertical component
of the angular momentum of the mass.
(d) Derive an equation for the angular momentum of the mass in terms of the equilibrium elevation angle 0 that occurs when d2 /dt2 = 0. Simplify this expression
in the small and large angular momentum limits when

CHAPTER 1. LAGRANGES EQUATIONS

10

(X,Y)
m

center of mass

Figure 1.6: Sketch for pucks connected by a spring on an air table.

A

Figure 1.7: Illustration for forced mass-spring system.

i. |0 |  1 and
ii. |0 | = |/2 0 |  1.
5. Two pucks of equal mass m are connected on a horizontal air table with a spring of
spring constant k and equilibrium length of zero as shown in figure 1.6. Other than
the spring constraint, the pucks are assumed to move freely on the air table without
friction. The orientation of the pucks is specified by the aximuthal angle .
(a) Derive the Lagrangian for this system. Hint: Use the theorem that separates the
kinetic energy into a part associated with the motion of the center of mass and a
part related to the motions relative to the center of mass. There are 4 degrees of
freedom associated with X, Y , d, and .
(b) Derive the governing equations for each of these variables. Which of the above
degrees of freedom are associated with conserved generalized momenta and what
are these quantities?
(c) Rewrite the d equation eliminating any generalized velocities in favor of the conserved momenta associated with these velocities. Find the equilibrium value of d
(i.e., with d2 d/dt2 = 0.)
6. Two masses m and an oscillating support point are connected by two springs with
spring constant k and equilibrium length l as shown in figure 1.7. The forcing point
moves back and forth according to a = a0 cos(0 t).
(a) Find the Lagrangian for this system.

CHAPTER 1. LAGRANGES EQUATIONS

11

(b) Determine the coupled equations of motion for the two masses in terms of their
positions x and y.
(c) Find the equilibrium values of x and y with a set to zero, x0 and y0 , and rewrite
the equations in terms of the perturbation lengths x0 = x x0 and y 0 = y y0 .