Oscillations of a simple pendulum with extremely large amplitudes

Abstract

Large oscillations of a simple rigid pendulum with amplitudes close to are treated on the basis of a physically
justified approach in which the cycle of oscillation is divided into several stages. The major part of the almost closed
circular path of the pendulum is approximated by the limiting motion, while the motion in the vicinity of the inverted
position is described on the basis of the linearized equation. The accepted approach provides additional insight into the
dynamics of nonlinear physical systems. The final simple analytical expression gives values for the period of large
oscillations that coincide with high precision with the values given by the exact formula.
(Some figures may appear in colour only in the onlinejournal)
1. Introduction
The old problem of large oscillations of a simple planar pendulum continues to attract the attention of the academic
community. Dozens of papers on the subject appeared during the last decade in European Journal of Physics, American

Journal of Physics, and otherjournals—see, for example, and references therein. In most ofthe papers various

approximation schemes have been developed to express the large‐angle pendulum period by simple formulae in terms
of elementary functions. Each of the authors usually claims that the formula proposed by him is more simple and
accurate when compared with other approximate formulae. A detailed comparison of several approximate expressions

that have appeared in recent publications can be found in . The common feature of all suggested approximation

schemes can be reduced to a search for some empirical expression for the period which gives for large

amplitudes an acceptable numerical agreement with the values obtained from the exact formula given by the

complete elliptic integral of the first kind :

, (1)

0143‐0807/12/061555 09$33.00 © 2012 Publishing Ltd Printed in the UK& the USA 1555 where

is the natural period and is the frequency of oscillations with infinitely small

amplitude, is the effective length of the pendulum and is the acceleration due to gravity.
The approximate expressions for the period that can be found in the literature (see and references therein) give
indefinitely increasing errors as the amplitude of the pendulum tends to . Moreover, all these exercises with
various approximation schemes give little physical insight in the nonlinear dynamics of the pendulum behaviour at large
amplitudes.
In the present paper we suggest a radically different approach to the problem of extremely large amplitudes. Our
approach is based on the physically clear presentation of large oscillations as consisting of several stages during which
the motion can be described analytically with high precision in terms of elementary functions. The principal idea of our
approach is very simple: the motion of the pendulum in the close vicinity of the inverted position can be described by a
linear differential equation (if we choose as a variable the angle which the pendulum makes with the upper
vertical line), while the remaining part of the pendulum’s path (constituting nearly a full circle) is almost
indistinguishable from the limiting motion (motion along the separatrix), for which a simple solution in elementary

functions is available. The precision of the final (very simple) formula for the period, equation , increases as the

amplitude approaches . We have already used this idea earlier in while comparing large amplitude
oscillations of the pendulum with full revolutions. The aim of the present paper is to draw attention to this fruitful
approach that gives additional physical insight into the nonlinear dynamics of the pendulum—a very popular physical
model often encountered in various undergraduate courses.
2. The phase portrait of the pendulum
Next we recall several peculiarities in the behaviour of the simple pendulum which are essential for understanding the

problem of large‐amplitude oscillations. The solution to the differential equation of a conservative simple

pendulum

(2)

can be expressed in elementary functions in the limiting case of oscillations with infinitely small amplitude: when

, equation becomes linear and describes simple harmonic motion with the frequency .

Oscillations with large amplitudes, as well as revolutions in a full circle, require special functions (elliptic functions) for
their description. However, the general character of variation with time of the mechamcal state of a nonlinear system

such as the pendulum can be graphically demonstrated by trajectories in the phase plane , i.e. the graphs which

plot the angular velocity versus the angular displacement . The family of these trajectories, corresponding to
different values of energy, constitutes the phase portrait of the system. The phase portrait tells us a great deal about the
possible motions of a nonlinear system.
We can construct a phase portrait for a conservative system (e.g., for the pendulum) without explicitly solving the

differential equation of motion of the system. The equations for phase trajectories follow directly from the law of

energy conservation. In the absence of friction, the total energy of the pendulum, i.e. the sum of its kinetic energy,

and potential energy, , remains constant during the

motion:

. (3)

This equation gives the relation between and , and therefore it is the equation of the phase trajectory which
corresponds to a definite value of total energy. It is convenient to express
Oscillations of a simple pendulum with extremely large amplitudes
Figure 1. Potential well of the simple pendulum and the phase portrait in the absence of friction. Closed phase
trajectories that enclose the origin of the phase plane correspond to oscillations with different amplitudes. Trajectories
passing over and below the separatrix correspond to counterclockwise and clockwise revolutions respectively.

equation in a slightly different form. We define the quantity that has the physical sense

of the kinetic energy of the pendulum rotating with the angular velocity , or, which is the same, of the potential

energy of the inverted pendulum. Then we rewrite equation :

. (4)

Several phase trajectories are shown in figure under the graph of .

If the total energy of the pendulum is less than the height of the potential barrier in figure ,

the pendulum swings between the extreme deflections and . If the amplitude is small , the

oscillations are nearly sinusoidal, and the phase trajectory is nearly an ellipse. The greater the total energy , the
greater the divergence of the phase trajectory from an ellipse and the greater the difference of the motion from simple
harmonic. At large amplitudes the pendulum spends more time near the extreme (turning) points where its direction of
motion reverses. The period of motion grows with the amplitude. If the total energy of the pendulum is greater than

the height of the potential barrier , the pendulum occurs in the inverted position with non‐zero

angular velocity. This means that it makes full revolutions.

3. The limiting motion
The phase trajectory corresponding to a total energy which is equal to the maximal possible potential energy,

namely , is of special interest. It separates the central region of the phase plane which is occupied by

the closed phase trajectories of oscillations from the outer region, occupied by the phase trajectories of rotations. This
boundary is called the separatrix. The separatrix divides the phase plane of a conservative pendulum into regions
Figure 2. Time dependent graphs of and for the limiting motion along the separatrix from to

in the absence of friction. The time origin corresponds to the moment at which the pendulum crosses the
lower equilibrium position

which correspond to different types of motion. The equation of the separatrix follows from equation by setting

, or from equation by setting :

. (5)

When the pendulum with energy approaches the inverted position at or , its velocity

tends to zero, becoming zero at . This state is represented in the phase plane by the saddle points and

where the upper and lower branches ofthe separatrix (equation ( )) meet on the ‐axis. Both these points

represent the same mechanical state of the system, namely the state in which the pendulum is at rest in the unstable
inverted position. The slightest initial displacement of the pendulum from this point to one side or the other results in its
swinging with an amplitude which almost equals , and the slightest initial push causes rotational motion of the
pendulum in a full circle. Executing such swinging or rotation, the pendulum spends an extended time in the vicinity of
the inverted position.

For the case of motion with energy (motion along the separatrix) there exists an analytical solution

(in elementary functions) for the angle of deflection and for the angular velocity . Indeed, integration of

equation for the positive sign at the initial condition yields:

. (6)

This solution describes a counterclockwise motion beginning at from At the pendulum

passes through the bottom of its circular path, and continues its motion until , asymptotically approaching

. Differentiating given by equation with respect to time , we find the following time dependence
of the angular velocity for the limiting motion of the pendulum:

. (7)

The graphs of and for the limiting motion are shown in figure . The graph of has the form of an

isolated impulse. In equation the time origin is the instant at which the pendulum passes through the

equilibrium position with angular velocity This moment corresponds to the peak value of the impulse. For

time on either side of the peak equation gives an angular velocity of only . Therefore the

period of small natural oscillations gives an estimate for the duration of the impulse on the velocity

graph, that is, for the time needed for the pendulum to execute almost all of its circular path, from the vicinity of the
inverted position through the lowest point and back.
Oscillations of a simple pendulum with extremely large amplitudes

Figure 3. Graphs of and for oscillations with the amplitude and 179 in the absence of

friction, obtained in the simulation experiment.

4. Oscillations with amplitudes approaching

If the pendulum is released with zero initial velocity near the inverted position (say, at an initial angle of about ), it
slowly starts moving toward the down position with a small imtial acceleration because the torque of gravity, being
proportional to the sine of deviation from the inverted position, is small. After the pendulum gains some speed, it
rapidly makes almost a full circular path through the lower equilibrium position. When the pendulum occurs on the
opposite side of the inverted position, its motion gradually slows down as it climbs up along the slope of the potential
barrier to its summit. In the absence of friction the pendulum stops when its angular distance to the vertical becomes
equal to the initial deviation. From this turning point all the motion repeats in the opposite direction, and after a period
the pendulum occurs at the initial point with zero velocity.
A computer simulation of motion of the rigid planar pendulum developed by the author can be found on the web .
The simulation program (applet) runs directly in any web browser with Java runtime environment (JRE) installed. To
observe the oscillations discussed in this section, we should switch off the yiscous friction (using the corresponding

check‐box on the ‘parameters’ panel), and choose appropriate initial conditions (imtial angle about , initial

velocity zero). The program allows the user to plot the time dependences of and , and to draw the phase

trajectory simultaneously with the visualization of oscillations.

Graphs of and for oscillations with amplitudes 179.90o and 179.99o in the absence of friction are shown

in figure
Comparing these graphs, we can see that for the most part of the angular excursion from to these graphs for

amplitudes and are nearly identical. We guess that for these stages of motion the deflection angle

and angular velocity are characterized by almost the same time dependence as for the limiting motion

along the separatrix, shown in figure . This time dependence is described (in elementary functions) by the simple

expression . Hence the duration of this stage of oscillation for all these cases

Figure 4. The phase trajectory of oscillatory motion with a large amplitude and its portion (increased) that

corresponds to the motion of the pendulum in the vicimty of the inverted position (b).

of large amplitudes approaching is about (the period of small oscillations) and can be calculated with high

precision with the help of the same expression . The duration of the remaining stage, during which the pendulum

lingers near the inverted position, depends critically on the amplitude . This is clearly seen from comparison of the

upper and lower panels of figure . This duration increases indefinitely as . In order to calculate the duration

of this stage for certain large amplitudes, we can make use of the linearized differential equation, applicable for small
deviations from the inverted position. We will do this on page 7.

The closed phase trajectory of oscillatory motion with a large amplitude is shown in figure . Most parts ofthe phase

trajectory almost coincide with the separatrix. The representing point goes around the whole closed curve during one
period of oscillation. Next we consider one quarter ofthis curve which starts in the phase plane at the initial point of

maximal deflection and initial velocity , and ends at the point (marked as in

figure (a)). To calculate this time , we choose on this curve an arbitrary point not far from the

inverted position (see figure (b)), which divides the curve into two parts. The first part between and

lies in the vicinity of the inverted position, so that the duration of motion along this part can be calculated

with the help of a linearized differential equation of motion (see below). The second part between and is

almost indistinguishable from the separatrix, so that the duration of motion along this part can be immediately

expressed with the help of equation :

. (8)

Here we introduced the notation for the angle that the pendulum makes with the upward vertical line at

. When is close to , the angle is small, so that in equation we can assume

. Therefore .

When considering the motion of the pendulum in the vicinity of the inverted position, we find it convenient to define
the pendulum position (instead of the angle ) by the angle of deflection from the position of unstable equilibrium.

This angle equals , so that . Substituting angular acceleration and in

equation , we find the differential equation for the pendulum in terms of . Since near the inverted position

Oscillations of a simple pendulum with extremely large amplitudes

, we can replace in this equation by . Thus we get the following linear differential equation

approximately valid for the pendulum’s motion between and :

. (9)

The general solution to this linear equation can be represented as a superposition of two exponential functions of time
:
 (10)

The initial conditions for the motion from to are and Applying these

conditions, we find the constants and in equation :

. (11)

To find the duration of motion from to (from to ), we substitute in equation

(12)

We have omitted the second term in the right‐hand side of equation . This is admissible if the arbitrary angle

(which divides the phase trajectory into two parts) is chosen to be large compared to . From equation we get

for :

. (13)

The desired period of oscillations is four times greater than the duration of motion from to the

lower equilibrium position . Adding from equation and from , we finally obtain the following

expression for the period of oscillations with large amplitude approaching :

. (14)

(Here ) We note that both and depend on the value of the angle which we have chosen to

divide the trajectory into one part that corresponds to the motion in the vicimty of the inverted position, and the other

that almost merges with the separatrix. Nevertheless, this dependence on disappears when we add and : the

final expression for the period is independent of the arbitrarily chosen value of (provided ).
The approximation given by expression is more accurate the closer the amplitude is to . The table below

illustrates the precision of this simple expression for oscillations of extremely large amplitudes. The values of in the

middle column are calculated on the basis of an exact formula ; the right column corresponds to the approximate

expression .

2.877664 2.876395

3.202109 3.201597

3. 3.900995

5.366867 5.366866

6.832737 6.832737

8.298608 8.298608

We note that according to this table one cycle of the pendulum oscillation at large amplitudes covers several periods of
small oscillations. As an assignment for a student activity, it would be expedient to suggest them to verify the values
cited in the table by direct measurements of the period in a simulation experiment using the software available on the

web
The above described oscillations with extremely large amplitudes occur if the energy supplied to the pendulum at initial

excitation is slightly less than the height of the potential barrier (see figure ). Ifthe pendulum is excited

from the lower equilibrium position by an initial push, the initial yelocity should be a little less than . If

the pendulum makes revolutions in a full circle. If is only slightly greater than , it is also

expedient to divide the motion of the pendulum into two stages. The stage of motion at crossing the inverted position

and in a small vicinity of it can be described with good precision by the linearized equation . The remaining almost

closed part of the circular path can be approximated, like in the above analysis of oscillations, by the known analytical

solution for the limiting motion, equation . In this way a simple analytical expression similar to equation for

the period of such non‐uniform revolutions can be obtained (see [ ]).

5. Another derivation of the expression for the period of large oscillations

In the above derivation of expression we have chosen arbitrarily some small angle for dividing the motion into

stages described by different analytical time dependences. Another way is to choose for this conventional boundary of

the two stages, instead of the angular position , some arbitrary small angular velocity , which the

pendulum gains while moving from the turning point at which its angular velocity is zero. To find the duration

of this stage occurring in the vicinity of the inverted position, we can make use of the above obtained solution to

the linearized equation , according to which

. (15)

Substituting in equation and taking into account that , we find

. (16)

The further motion towards the equilibrium position is almost indistinguishable from the limiting motion. Hence the

time dependence of the angular velocity for this stage can be assumed the same as for the limiting

motion, see equation . Therefore for calculating the duration of this stage we can substitute in

and take into account that . This yields

. (17)

Adding from equation and from , we finally obtain the same simple expression for the period of

oscillations with a very large amplitude :

. (18)

Again, the arbitrarily chosen angular velocity which we have used to divide the motion into different
stages vanish from the final expression .

6. Concluding remarks

We have considered the old problem of large oscillations of a simple rigid pendulum with amplitudes close to on
the basis of an approach in which the cycle of oscillation is divided
Oscillations of a simple pendulum with extremely large amplitudes into several stages. The major part of the almost
closed circular path of the pendulum in such an oscillation is approximated with good precision by the limiting motion,
for which there exists an analytic solution in elementary functions. The remaining small parts of the path, occurring in
the vicinity of the inverted position, are described on the basis of the linearized equation, which is valid for the new
variable . The point that divides the path of the pendulum into stages described by different approximations
is chosen to some extent arbitrarily, but this arbitrariness does not influence the final approximate expression for the

period of oscillations. The final analytical formula is very simple and gives for the period of large oscillations

rather accurate values that coincide with high precision with the values given by the exact expression in terms of

the complete elliptic integral of the first kind. More importantly, the approach to the problem described in this paper
provides additional physical insight into the dynamics of nonlinear systems.