AC 2007-172: USING ELLIPTIC INTEGRALS AND FUNCTIONS TO STUDY

LARGE-AMPLITUDE OSCILLATIONS OF A PENDULUM

Josue Njock-Libii, Indiana University-Purdue University-Fort Wayne

Josué Njock Libii is Associate Professor of Mechanical Engineering at Indiana University-Purdue
University Fort Wayne, Fort Wayne, Indiana, USA. He earned a B.S.E in Civil Engineering, an
M.S.E. in Applied Mechanics, and a Ph.D. in Applied Mechanics (Fluid Mechanics) from the
University of Michigan, Ann Arbor, Michigan. His areas of interests are in mechanics,
particularly fluid mechanics, applied mathematics, and their applications in engineering, science,
and education.

Page 12.1543.1

© American Society for Engineering Education, 2007

 Using elliptic integrals and functions to study
large-amplitude oscillations of a pendulum

Abstract

The solution to the oscillations of a pendulum that includes large amplitudes is presented
for the purpose of comparing it to that for small amplitudes. Such a comparison allows
for the determination of the limits of applicability of the linearized equation. It is shown
that, in both cases, the angle of swing is a periodic function of time but that the nature of
the functions involved varies with the amplitude of motion. For small angular
displacements, the period of oscillation is a constant and the ensuing angle of swing can
be represented accurately by means of circular functions. However, for large amplitudes,
the period is represented by Jacobi’s complete elliptic integral of the first kind and varies
with the initial amplitude, while the corresponding angle of swing is represented by
elliptic functions of Jacobi. It is shown that the period of the linearized motion is always
smaller than, or equal to, that from the nonlinear motion. The errors induced by the
linearization process are determined analytically and represented graphically. It is
demonstrated that those in the magnitude and phase of swing vary with time and the
initial amplitude of the pendulum. Consequently, as a general rule, it is inaccurate to use
the error in the angle as an estimate of the accuracy of how well the linearized solution
approximates the actual motion.

1. Introduction

The motion of a pendulum is studied in the first college physics course; and its governing
differential equation is amongst the first ones that are solved in an introductory course on
ordinary differential equations. This equation is encountered again and again in courses
such as dynamics, controls, vibrations, and acoustics. In all these cases, however, it is
linearized by assuming that the amplitude of oscillation is small. As a consequence,
students do not see what happens to the oscillation of a pendulum when the amplitudes
are large and the restoring force becomes nonlinear. More importantly, they do not know
the limits of applicability of the linearized solution they have studied.

In this article, we present the solution to the oscillations of a pendulum that includes large
amplitudes and compare the general solution to that which is valid only for small
Page 12.1543.2
amplitudes. This allows one to determine the errors induced by and the limits of
applicability of the linearized equation.

2. The Basic equation

Consider a rigid body that is suspended from a point O about which it oscillates in the
vertical plane. Let the angular displacement about the vertical axis be denoted by ,
measured in radians. After applying either Newton’s second law of motion, or the
conservation of mechanical energy, it is found that undamped oscillations about point O
can be obtained by solving the equation1

2
n sin( ) 0, (1)

In general, the conditions at the starting time, t = ts, are given by 2

t t s , (t s ) s , (t s ) s. (1a)

In these equations, the dots represent differentiation with respect to time t and the
quantity n , which has units of rad/s, is related to the natural frequency of the system.

As an example, for a compound pendulum swinging in the vertical plane about a

horizontal axis that goes through point O,

mtotal gd
n , (1b)
J0

where, mtotal is the total mass of the pendulum; g is the acceleration of gravity; d is the
distance between point O and the center of mass of the pendulum; and J0 is the (polar)
mass moment of inertia of the body about point O. It can be seen that n is a physical
parameter that does not depend on time.1

3. The solution for small angles: circular functions

For small amplitudes, it is conventional to linearize Eq.(1) by expanding the sin into a
power series as shown below
3 5 7
( 1) n 2 n 1
sin( ) (2)
3! 5! 7! (2n 1)! Page 12.1543.3
and replacing the sin with , the first term in that series. Doing so gives

2
n 0 (3)
This is the equation that is used in all the courses mentioned above. Its solution is

(t ) A sin( n t ) B cos( n t) (4)

In this case, n is the circular frequency of the motion expressed in radians per second.

After the initial conditions given in Eq (1a) are used in Eq (4), the constants A and B are
found to be given, respectively, by
s
A s sin( n s t) n
cos( t)
n s
(5)
s
B s cos( t
n s ) n
sin( t
n s )
In order to obtain a solution with a simple mathematical form, it is conventional to let
2
be the maximum amplitude of oscillation and set t s 0, s 0, s n . Incorporating
these assumptions into Eq. (5) leads to A and B 0 ; and Eq (4) becomes

(t ) sin( n t) (6)

Here the period of oscillation, n, is related to the circular frequency, n, by

2
n (6a)
n

It can readily be observed from Eq.(6) that the instantaneous position of the pendulum
during oscillation is a circular function of time and is directly proportional to the
amplitude of motion, . From Eq.(6a), it is seen that the period of oscillation of the
pendulum is a constant that is independent of the amplitude of motion. It follows that all
amplitudes that are within the limits of applicability of the governing equation yield the
same period of oscillation. Consequently, the period and frequency of oscillation are not
affected by the initial conditions. We will compare these results to those obtained when
the pendulum assumes large amplitudes of oscillation.

4. The solution for any angle: elliptic functions and integrals

Page 12.1543.4
When swinging angles may be large, Eq.(1) is transformed into Jacobi’s elliptic integral
of the first kind by two successive integrations and a change of variables 2. The exact
solution to Eq.(1) is 2

(t ) 2 Arc sin sin

sn( t ) , (7)
2
where sn represents Jacobi’s elliptic function with the elliptic modulus suppressed 3-7.

The elliptic functions of Jacobi are defined as inverses of Jacobi’s elliptic integral of the
first kind. Thus, if one writes

d
u ,
0
1 k 2 sin 2 ( )
then, for example, sn(u, k ) sin( ) , cn(u, k ) cos( ) and dn(u, k ) 1 k 2 sin 2 ( ) .

For the derivation of Eq. (7), it is conventional to transform the original differential
equation into an integral as

1 d
t . (7a)
2 0
sin 2 sin 2
2 2
Then, by setting

sin
2
u , (7b)
sin
2
and using this change of variables in Eq. (7a), one gets

du
t (7c)
2
0
1 u 1 k 2u2
where

Page 12.1543.5
 sin
2
, (7d)
sin
2
and
k2 sin 2 . (7e)
2

From Eq.(7c), the period of oscillation is given by Eq.(8) as

4
K(k 2 ) , (8)
where K denotes Jacobi’s complete elliptic integral of the first kind 3-7, which is
defined as
2
du
K(k ) . (8a)
0
1 u2 1 k 2u2
Expanding K into a power series 4, one gets

2 2 2 2
1 .
13 ..
135 . . .7
135
0 1 k2 k4 k6 k 8 ... , (9)
2 2.4 2.4.6 2.4.6.8
where
2
0 .
Rearranging Eq.(9), the ratio of the two periods is found to be
2 2 2 2
1 2 13. 4 135
.. 6 135
. . .7
1 k k k k 8 ... , (9a)
0 2 2.4 2.4.6 2.4.6.8

It can readily be observed from Eq.(7) that the instantaneous position of the pendulum
during oscillation is a nonlinear function of the amplitude of motion, , and an elliptic
function of time. From Eq.(8), it is seen that the period of oscillation of the pendulum
depends upon the amplitude of motion. Although the period varies with of the amplitude
in a nonlinear way, one can see from Eq.(9) that it increases monotonically with the
amplitude. Consequently, the period and frequency of oscillation are affected by the
initial conditions. We will compare results obtained assuming small amplitudes of
oscillation to those obtained assuming large amplitudes. Page 12.1543.6
5. Comparing the solutions

5.a. Comparing the two periods

It can be seen from Eq.(9) that, , the period obtained from the nonlinear equation,
increases with the amplitude; and from Eq.(9a) that is always larger than, or equal to 0,
the period obtained from the linearized equation. This relationship is illustrated
graphically by the plot of vs. k that is shown in Fig. 1.
0
t
t0

3.5

2.5

1.5

k
0.2 0.4 0.6 0.8 1

Fig.1. Plot of the ratio of periods: vs. k, Eq. (9a)

0
We define the error in the computation of the period as the difference between the exact
and the approximate periods divided by the exact period, as shown in Eq (10).

0
ErrorP (10)

Page 12.1543.7
Similarly, we define the error in the computation of the initial swing angle as the
difference between the initial angle of swing and the sine of divided by the sine of ,
as shown in Eq.(11).

sin
Error (11)
sin

The ratio between the errors found in Eqs. (11) and (10) is shown in Eq.(12) and plotted
in Fig. (2).

Errorp
Ratio (12)
Error
It can be seen from that figure that the error in the angle is always larger than that in the
period. Thus, since the error that is made in using the angle itself instead of its sine is
much easier to compute, it can be obtained and used as an upper bound on the error to be
expected in the period.

Ratio
0.4

0.3

0.2

0.1

k
0.2 0.4 0.6 0.8 1

Fig. 2. A ratio of the error in the period to that in the angle .

5.b. Comparing the swing angles

To illustrate the differences between the swing angles obtained from the nonlinear and
linear equations, six starting angular amplitudes have been chosen; and, for each, a
Page 12.1543.8
solution was obtained using the linearized equation and another with the nonlinear
equation. The initial angles used are 10o, 30.32o, 63o, 88.420o, 121.3o, and 147o and
they have been identified in Fig.1b with dots. They correspond, respectively, to
k = {0.08716, 0.26148, 0.52296, 0.69728, 0.8716, 0.95876}. Plots of the corresponding
variations of the angular positions of the pendulum with time are shown in Fig. 3, where
the solid lines represent the linear solution and the dashed lines the nonlinear (exact)
solution.

From Fig. 3, it can be seen that, as the initial amplitude , that is given to the pendulum to
initiate its motion, increases (from 10o, to 30o, 63o, 88o, 121o, and 147o), so does, , the
0
ratio between the exact period and the approximate period 0, Eq.(9). The widening
difference between the periods prevents the two curves from being in lock-step; this, in
turn, increases the discrepancy between the corresponding angular positions of the
pendulum.

For each initial angle used in Fig. 3, differences between the two solutions were
computed using Eq.(13) and plotted in Fig. 4 in which the initial angle is a parameter.

Differ = 2 Arc sin sin sn( t ) - sin( n t ) (13)

2
It can be seen from Fig. 4 that these differences vary considerably with both time and
amplitude and can change algebraic signs during the motion.

Page 12.1543.9
 Angle Angle
0.15 0.4
0.1 0.2
0.05
u u
-0.05 1 2 3 4 5 6 -0.2 1 2 3 4 5 6
-0.1
-0.15 -0.4

Angle Angle
1 1.5
1
0.5 0.5
u u
1 2 3 4 5 6 7 -0.5 2 4 6 8
-0.5 -1
-1 -1.5
Angle Angle
2
2
1 1
u u
2 4 6 8 10 -1 2 4 6 8 10 12
-1
-2 -2

Fig.3. Swing angle vs. time, for 10o, 30o, 60o, 90o, 120o, and 150o.
Exact values are for 10 , 30.32 , 63o, 88.420o, 121.3o, and147o.
o o

Solid lines (approximate solution); dashed lines (exact solution)

Page 12.1543.10
 Differ Differ
0.15 0.6
0.1 0.4
0.05 u 0.2 u
-0.05 5 10152025 -0.2 5 10152025
-0.1
-0.15 -0.4
-0.6

Differ Differ
2
1 1
0.5
u u
-0.5 5 10152025 -1 5 1015202530
-1
-1.5 -2
Differ Differ
2 3
2
1 1
u u
-1 5 1015202530 -1 51015202530
-2 -2
-3
Differ Differ
4 4
2 2
u u
-2 5101520253035 -2 10 20 30 40
-4 -4

Fig. 4. Differences in Eq.(13) are plotted over four consecutive cycles, k is a parameter.
(k = 0.12, 0.24, 0.36, 0.48, 0.60, 0.72, 0.84, and 0.96, respectively).
Page 12.1543.11
6. Conclusions

We presented the solution to the oscillations of a pendulum that encompasses large

amplitudes of swing and compared it to that which is valid only for small amplitudes.
This allowed for the determination of the limits of applicability of the linearized equation.

It was shown that, in both cases, the angle of swing is a periodic function of time but the
mathematical nature of the functions involved changes with the amplitude of motion. For
small initial angular displacements, the period of oscillation is a constant that is
independent of the initial displacement of the pendulum; and the ensuing angle of swing
is represented accurately by circular functions. For large amplitudes, however, the period
of oscillations is not a constant, for it varies with the initial amplitude given to the
pendulum; it is represented mathematically by Jacobi’s complete elliptic integral of the
first kind; and the corresponding angle of swing is expressed by means of elliptic
functions of Jacobi 2-8. As can be seen from Eq.9, the period of the linearized motion is
always smaller than, or equal to, that of the nonlinear motion.

The approximation sin that is used to linearize the differential equation introduces
three kinds of errors in the solution: one is in the magnitude of the period of oscillation;
the second one is in the magnitude of the swing angle; and the third one is in the phase of
motion. These errors were determined exactly and represented graphically.

When one uses the approximate equation, approximation errors that are introduced in the
angle of swing affect the swing period of oscillation. The error induced in the period is
fixed for a given motion of the pendulum (Eq. 10). So is the error in the angle itself (Eq.
11). These two errors were compared in Eq.(12). From the plot of their ratio that is
shown in Fig. (2), it can be seen that the error in the period is always smaller than that
made in approximating the sin( ) by the angle . Therefore, the error in the angle can be
used as shortcut to the determination of an upper bound on the error to be expected in the
period. Numerical experimentation showed that the period of oscillation of the pendulum
can be estimated reasonably well using the linearized equation up to angles of 30o.

Errors in the amplitude of swing and in the phase vary with time and the initial
amplitude. If the time elapsed is large, even what might ordinarily be considered to be
small angles of swing can lead to large errors in the predicted position of the pendulum.
Indeed, numerical experimentation showed that the linearized equation represents the
position of the pendulum reasonably well only up to about 10o 8. Consequently, care
must be taken in determining the amplitude and the phase of swing of a pendulum from Page 12.1543.12
the linearized equation. As a general rule, therefore, it is inaccurate to use the magnitude
of the difference between sin( ) and the angle as an indication of the accuracy of the
solution obtained from the linearized equation.

7. References

1. Rao, S.S., Mechanical Vibrations, 4th edition, Addison-Wesley, Reading, Massachusetts, 2004, 129-152.

2. Walker, Peter L., Elliptic Functions: A Constructive Approach, John Wiley and Sons, Chichester,
England, 1996.

3. http://mathworld.wolfram.com/JacobiEllipticFunctions.html

4. Page, Leigh, Introduction to Theoretical Physics, third edition –Sixth Printing, D. Van Nostrand
Company, Inc. Princeton, New Jersey, 1965, 104-107.

5. Temme, Nico M., Special functions an introduction to the classical functions of mathematical physics,
John Wiley and Sons, New York, 1996, pp. 315-332.

6. Whittaker, E. T. and Watson, G. N. A Course in Modern Analysis, 4th ed. Cambridge, England:
Cambridge University Press, 1990.

7. Morse, P. M. and Feshbach, H. Methods of Theoretical Physics, Part I. New York: McGraw-Hill, 1953,
433.

8. Thomson, J.M.T. and Stewart H.B., Nonlinear Dynamics and Chaos: Geometric Methods for Engineers
and Scientists, John Wiley and Sons, Chichester, England, 1986, 15-25

9. A.B. Basset, A Treatise on Hydrodynamics, Vol. 2, Dover Publications, New York, New York, 1961,
260-267.

Page 12.1543.13