VB Bhatia 1

10.
Nonlinear Oscillations and

Chaos
10.1 Introduction
This chapter is meant to be a quick introduction to nonlinear oscillations
and chaos. In the last few decades this subject has assumed great
importance, mainly due to the discovery of chaos and its application to
phenomena not only in various branches of science but also in social
sciences. It has therefore been argued that this subject should form an
essential part of a course in Classical Mechanics. Moreover, personal
computers (PCs) have now become available at many places and a degree
course in physics includes some elements of computer programming. A
PC can be an extremely useful tool in understanding some of the concepts
introduced in this chapter because many of the examples and problems
have been worked out on a PC. Many diagrams have been taken from
the computer screen. Some computer programs are given in the Appendix.
The aim of this brief introduction is to convey the flavour of this
subject and introduce the key concepts, some through numerical work
on the computer, or what may be called 'computer experiments. No
attempt is made to give a cömplete, self contained treatment with proofs
and other supporting material because this is not possible in the short
span of a chapter in a book meant for a course on classical mechanics.
For details of this subject reference can be made to some of the sources
referred to here.
Section 7.9 discussed a special case of nonlinear oscillations. We now
discuss briefly general nonlinear
oscillations.To introduce the various
concepts which will be used in connection with the nonlinear oscillations,
we make use of the now all too familiar linear harmonic
oscillator with
angularfrequency o- Its equation is
i+x=0 (10.1)
when there is no dissipation. We know that its general solution is
A sin (@0t + ), where the amplitude A and the phase d are constants.
Nonlinear Oscillations and Chaos 269
Rut this 1S not wnat we dre

concerned with here; we are more
with what is called the phase concernea
space trajectory or simply the
trajectory, the trajectory in the (x, x) plane. It is customary to denote x phase
by y, so that Equation (1) is equivalent to the set
y -w
(10.2)
of coupled first order differential equations. The phase trajectories can

be easily shown to be ellipses (Fig. 10.1a) (see also Section 5.2).
y/o
*
al (b)
Fig. 10.1
PROBLEM 10.1. Show that the phase trajectories given by Equation (10.2) are
ellipses. Show also that if we plot y/@ against x, the trajectories are circles
(Fig. 10.1b)
Let us now introduce dissipation proportional to the speed. The equation
of the harmonic oscillator can now be written as
i+ 2bi + ox =0 (10.3)
or equivalently,
y= - w x - 2by
(10.4)
X =y
Recall that the substitution of the trial solution e in Equation (10.3)

gives a quadratic in A whose roots are 1,2 = - b t (b - w). The
Solution, therefore, depends upon the nature of the quantity b - aw)
When b < oj. the roots, 12 are complex, and the solution of Equation
(10.3) is in the form of damped harmonic oscillations, as we saw in
Chapter 7. In case b> aof. or b' = o6, the solution is aperiodic damped
motion. The phase trajectories in all these cases can be obtained
270 Classical Mechanics
conveniently by numerical methods. The trajectories are shown in

Fig. 10.2.
(a) (b)
(c)
Fig. 10.2
For the sake of completeness, we also consider the
case when b =0
and o6 is a negative quantity, say -o. In this case the
motion is obviously
unstable (one solution goes as e). The
phase trajectories are hyperbolic
curves as shown in Fig. 10.3.
Fig. 10.3
Computer programs for the numerical integration of differential equations are
of the book in the Appendix. given at the end
10.2
Singular Points of Trajectories
Consider a system of equations
J X1, X2, .., X t) j = 1, 2,.. n (10.5)

Here f are such-that their partial derivatives with respect to the n +l
variables x, and t exist and are-continuous. The functions f are then said
to satisfy Lipschitz
conditions. Let the initial conditions at t = to, namely,
xto)=Xoj J=1, 2, . .
n, be prescribed The Cauchy theorem then ensures
that the solution of this system of equations exists and is determined
uniquely by the initial conditions
In physical problems of interest to us here the functionsf will not depend
on t. In that eventEquations (10.5) are said to be autonomous. For example
the set
x = P(x, y). j = Qx, y) (10.6)
offshoot is that be eliminated and we can write

is autonomous. The t can
Equations (10.6) as
dxldy =
Qx, yVP(x. y); Plx. y) *0 (10.7)
flx). In case P(x, y) =
0,
integrate to get the integral curve as
=
and y
we can reverse the role ofx and y and get the integral curve as x = g(y).
and
pertinent at this point to notice that though Equations (10.6) the
It is
curve while
(10.7) are equivalent, the latter give a geometrical in time and
former tell us, in addition, how this curve is described
of the system at
direction thereby giving us information about the state
and Q(x, y) vanish
each instant of time#A point at which both P(x, y)
of the differential equations.
simultaneously is said to be a singularpoint
other point is
It is also called an equilibrium point of the system. Any
an ordinary point of the equations.
Since the functions P(x, y) and Q(x, y) satisfy Lipschitz conditions,
It implies that
the Cauchy theorem is applicable to the system (10.6).
and only one integral
through each apoint of the (x, y) plane passes one
-
curve. In fact, i and y may be thought of as defining the components

of a velocity vector and then Equations (10.6) can be interpreted as
giving a flow in the phase plane (x, y). The integral curves are then
completely determined by being tangential to the velocity vector at every
point. If the two integral curves were to intersect at a point, this would
imply two directions of the velocity vector at that point. This is ruled out
'For a discussion of the existence theorems, see, for example, Ince, op. cit, Chapter I1.
by the uniqueness of the solution. Thus, no twointegral curves can have

point in common.
A Singular point is a stationary point of the flow, since at this point
Pa, y) Qr, y) =0. t
= asymptotically stableif alltrajectories starting
is
it 'attracts'
sutficiently near it tend to it asymptotically as t>0o. Since
all trajectories, it is called an attractor. Jf there are trajectories which
SYmptotically tend away from the singular point as i 0o, then the
In the case of an harmonic
Sungular point is asymptotically unstable.
oScillator [Equation (10.1)] the singular point is the point (0, 0) which
is neither stable nor unstable since the trajectories are closed curves
around it which neither approach the singular point nor do they move
away from it. The singular point in this case is called a centre, It may
be noticed that the linear harmonic oscillator is obtainéd by putting
b =0 in Equation (10.3). For real systems b is never exactly zero. The
singularity typified by a centre is therefore a result of a

convenient
mathematical idealization. In real life b is nonzero and the trajectories
are spirals converging to or diverging from foci, as in Figs (10.2a) andd
(10.4). Ifb is + ve the trajectory approaches the singular point (as in

if b is
Fig. 10.2a) and the singular point is a stable focus. However,
which
ve, the trajectory spirals away from the singular point (Fig. 10.4),
is then an unstable focus.
b=-1
Fig. 10.4
For a detailed discussion of the singular points of autonomous systems

see, for example, Minorsky (1962).
of the system whose equation of
PROBLEM 10.2. Show that the phase trajectories
motion is
- dr =0 (10.8)
are hyperbolic curves with y = t o as asymptotes.
PROBLEM 10.3. Show that the phase trajectories of a damped harmonic oscillator
(b< ) are logarithmic spirals.

The trajectories of the kind shown in Fig. 10.2(b) and (c) tend towards
.arigin with a limiting direction. The origin is therefore a singular
the
int which is called a node. And, as all the trajectories in these two
point
examples approach the singular point, it is, therefore, stable. If the

eaiectories were all to tend away from the node as in Fig. 10.5, then it
trajed
would be an unstable singular point.
Fig. 10.5
The trajectories shown in Fig. 10.3 are hyperbolic curves. The singular
point in this case is said to be a saddle point. Since all trajectories
move
a saddle point is an unstable

away from the singular point as too,
of
singular point. We shall obtain later (Fig. 10.7) the phase trajectories
a pendulum. There we shall see saddle points again. The physical meaning
of the curves around a saddle point, as in Fig. 10.3, will be explained
there with reference to the pendulum.
PROBLEM 10.4. Show that for the set of equations
2x -4y (10.9)
i =
2x 2y. y
- =
the origin is a saddle point.
PROBLEM 10.5. Solve the equations
i 20 -), j =
6y -
7x (10.10)
and find the nature of the singular point.

PROBLEM 10.6. Solve the equations
(10.11)
Show that the origin is a node for trajectories in

and sketch the phase trajectories.
the regionx<0 and a saddle point for trajectories in the region x > 0. Such a singular
point is called a saddle node.
10.3 Nonlinear Oscillations

concepts associated with oscillations, we
Having explained the various nonlinear oscillations. Before we discuss
are now in a position to consider
case of a mathematical
consider the simple
let us
the general case,
be written as
pendulum whose equation may
* + @ sinx =
0, w =
gll (10.12)
the familiar equation of a harmonic
If we approximate sin x by x we get
second term in the sine series,
However, if we retain the
Oscillator.
nonlinear equation
namely r'/3!, we get the
+ afx (0/6)x*
-
= 0 (10.13)
approximate solution of this equation is
A simple method of getting an
that the solution is still sinusoidal

but the frequency now
to assume
therefore put
depends upon the amplitude. Let
us
x(0) = A sin ør, @ =

@(A) (10.14)
With this the left hand side of Equation (10.13) takes the form
(05A-/8)] Asin øt + (0AS/24) sin 3wt (10.15)

o- -
This is obviously not equal to zero, which it should be if Equation
out is to make
(10.14) is to be the solution of Equation (10.13). One way
the
the coefficient of A sin wt vanish and assume A to be small so that
second term in Equation(10.15) is small, thus making the two terms in
this equation to be nearly zero. This gives
(10.16)
The smallness of A implies
oa1 (10.17)
While this is a good enough approximation, a better approximation may

be obtained by including higher harmonics in the assumed solution of x(t).
We now consider how one can get an exact solution for a nonlinear
oscillator. We write the equation of the nonlinear oscillator in the general
form
+ fx) = 0 (10.18)
where f(x) is any integrable function. Equation (10.18) may be regarded
as having the form
+k(x) x =0 (10.19)
which, if k(x)> 0, 1s the equation of an oscillator with a variable spring
constant. The system equivalent to Equation (10.18) is
=
y, y -ft)
=
(10.20)
so that the integral curves are given by
dyld = -fx/y (10.21)
This equation shows that the tangent to the integral curves at the roots
of the equation fx) 0 (y # 0) are parallel to the x-axis, and that the
=
integral curves are perpendicular to the x-axis at the points where they
cross it except at the critical points. This is clear from Fig.
10.6, where
have drawn the integral curves in a special case f(x) =x(x a).
+
we
---e
Fig. 10.6
curves are
points
located at
of the integral
Moreover, the singular
points
Multiplying throughout by i
simultaneously.
0 get
0 and flx)
=
where =
with respect
to time, we
y (10.18)
Equation
and integrating
(10.22)
(/2)y+ fo) dr=E
integral since (1/2)y is
the energy
identified with
easily
The integral is be thought of as the potential
and fx) dr may
the kinetic energy
energy V(x) defined to within arbitrary constant. So, at the singular

an
points of the integral curves, where f(x) 0, we have d V(r)/dx 0, which

=
=
the point of equilibrium

Simply reproduces the well known result that at
the potential energy of the system must go through an extremum. Equation
know is constant for
(10.22) implies that E is the total energy which we
a conservative system. So, if a single-valued first integral exists for a
be conservative. From
system such as (10.18), the system is said
to
Equation (10.22) we have

y =t [2(E - V)]2 (10.23)
which arè the phase trajectories. To explain how one really obtains the
trajectories graphically, we show the process in Fig. 10.6 for the special
case when f(x) = x (x + a).
PROBLEM 10.7. Apply this method to Equation (10.12) for a mathematical pendulum
and show that the phase trajectories are given by
y=t [2 (cosx - (1 -E*)] (10.24)

where E*= El@. Notice that E* has to be between 0 and 2 for the pendulum to
oscillate. When E* >2 the pendulum has sufficient energy to rotate about the point
of suspension. The trajectories are shown in Fig. 10.7.
ww.*** e 3.2
*2
e#12,
*******
wwwwe w ww
Fig. 10.7
The closed curves in Fig. 10.7 indicate oscillations, while the open
curves far from the x-axis show the rotation of the pendulum. E* = 2 is
the critical case; the corresponding curve passes through the saddle point.
In this case, the pendulum simply stands inverted (in an unstable position)
If we put
cos A =
(1 E*) (0<E* <2) then
-
Abecomes the amplitude

of the oscillation of the pendulum, then Equation (10.24) can be integrated
to give
t=t(1/@o) d (10.25)
[(2(cos- cos A]
ave assumed that the measurement of time
where we hav mences from
the moment when x = 0. If we use the substitution
sin /2) = sin (A/2) sin z (10.26)

Equation (10.25) becomes
t(1/@ sin"
dz z (10.27)
where
k sin (A/2), and a =sin [sin (E/2)/sin (A/2)1. (10.28)

Equation (10.27) expresses t in terms of the elliptic integral of the first
kind. This is the exact solution of the mathematical pendulum.
To find the time period of the oscillations, we have
t/2
T (4/0o) dz (10.29)
-Rsin z
in terms of the complete elliptic integral K(k). Since
K(k) = (T/2) [1 + (1/2)'* +(1.3/(2.4) K* +...] (10.30
we have for small amplitudes [sin (A/2) = (A/2)],
T (27t/wo) (1 + A/16) (10.31)
implying
= @1-A/16) (10.32)
contained in Equation (10.17).
which is identical with the result
solutions of
There are several other methods for getting approximate
are called the perturbation
methods.
nonlinear oscillator equations'. These
or a@, or both, in terms of a
small
In all these methods we expand x,
how good the
to some order depending on
parameter and retain terms up
we illustrate one such method,
desired approximation is. As an example,
method by applying it to the problem of
namely, the Linstedt-Poincaré
that this method gives the same result
mathematical pendulum. We show
as that given by
the one approximated from the
Tor amplitudes
small
exact solution (10.29).
Oscillations.
Mood, Nonlinear
See, for example, Nayfeh and
of the pendulum becomes

Expanding sin x in powers of x, the equation
0
+ wfx -x/6 +..) =
and @ as follows:
We define a quantity T= ot and expand x
X Exi(T) + Erz(7) +...
= 0 + E02 + ..
of the oscillation.
where is a small parameter related to the amplitude
e
because there is n0 x term in
E term is absent from the expansion of x
been omitted in @ because
the original equation. Also, the e@ term has
the frequency should not depend on the sign of E (or the amplitude).
in the original equation and
Substituting the expansions of x and @
for x1, X3, etc. Let
comparing coefficients of E, E1, etc., we get equations
derivative with respect to
us deal with the first two. Denoting the
just
Twith an accent, these equations are:
@x+ x) = 0
o(xs+x3)= -
20,02xf+ (1/6) ax
and B are
The equationfor x has a solution x = a cos (T+ B) where a
constants. This is the leading term in the
solution. We substitute this in
the equation for x3. We get
o6(x + x3) =
[20%aha + (1/8) wf*] cos (T + B)
+(1/24) or'oß cos (3r + 38)
The first term on the right hand side will produce undesirable secular
terms in the solution, so the coefficient of cos (T+ B) must be made zero.
This gives the first order correction @2 = - (1/16) @a to wo. This was
the result obtained in Equation (10.32). We now write down the first
approximation to x identifying ea with the amplitude A.
x =Acos |@o
A look at Figs 10.6 and 10.7 reveals another interesting aspect of the
topology of phase trajectories. In these figures the phase plane has been
mapped into regions in which trajectories have different oscillatory
properties. These regions are separated by asymptotic trajectories called
separatrices, marked by the letter S. In both the figures the trajectories
inside the separatrices are closed curves representing periodic oscillations.
tn Fig. 10.6, for example, we see
that when the has energy E
there are two turninE points, the motion of the system
system is periodic and its
nhase space trajectory 1s
closed. The
separatrix
equal to the maximum value of V. The corresponds to energy Ea
equ.
trajectories outside the
correspond to the energy E being larger than the maximum in V separatrix
and imply
non-periodic motion.
In many physical problems the
function fof Equation (10.18)
also on a parameter, say A. In the case of a plane pendulum on a depends
rotating
plane, for example, A depends on the frequency of rotation of the plane.
In the (2, x) plane then we draw the curve f(, ) = 0. Each point on this
curve is a point of equilibrium because f(x, 2) = 0 implies a Vldx = 0,
which is the condition for equilibrium. The curve fir, ) may be quite
complicated in some cases as, for example, sketched in Fig. 10.8. Notice
that at critical points ike A and B, either two values of x coalesce
into
one, or one value gives way to two values. Values of 1 at these points
are said to be critical or bifurcation values. Notice that at a critical point
Fig. 10.8
ofx, A/dx vanishes and consequently tangent to the curve fx, A) = 0

becomes vertical.
We have already noted that each point on the curve f(x, A) 0 is a
=
point of equilibrium. The equilibrium would be stable if the potential

energy has a minimum there, i.e., d-VIdx* > 0. This means that
of x, A/dx = - a*Vid should be negative at the point of stable
equilibrium. So, if f (x, A) changes sign from positive below the curve
fx, A) = 0 to negative above the curve, then each point on the curve is
the point of stable equilibrium. Butiff(x, A) undergoes a change of sign
in the opposite direction, then points on the curve will be points of
unstable equilibrium. This rule is due to Poincaré .
JULES HENRI POINCARË (1854-1912), the great French mathematician and physicist, is
remembered as the last of the universalists. Besides studying the theory of orbits and linear
(Contd)
a situation
consider a simple pendulum of
example of such
As an
which itself is rotating about a
to oscillate in a plane
length I confined velocity S2.
Its equation of motion
vertical line with a constant angular
as
may be written
+f(0, 2) =0 (10.33)
S(0, 2) =
Q°(a-cos 6) sin 6 (10.34)
with the parameter 2 =g/21.

PROBLEM 10.8. Establish Equation (10.33) for a rotating pendulum.
The system equivalent to Equation (10.33) is
7,= - j(6, A) (10.35)

(0, 0), (0, tT) and (0, cos A) if a< 1.
=
located at
Singular points are
The curve fl6, A) = 0 is shown in Fig. 10.9. We notice that tl are the
the of stable
bifurcation values of A. Applying the rule of Poincar points
and unstable equilibrium can be easily identified.
/2
-1.5 1 -0.5 0.51 1.5 A
-T/2
Fig. 10.9
(Footnote Contd)
differential equations, his most celebrated work is the development of the theory of automorphic
functions and the introduction of Fuchsian functions. In physics he made lasting contributions
in almost all areas of classical physics. His solution of the three-body problem fetched him the
coveted prize from the king of Sweden, though later it was found that the solution was not only
incomplete but was also incorrect. As a note of historical interest, his work on electromagnetism
is specially interesting. In 1905 he showed that
Maxwell equations were not consistent with the
classical Newtonian mechanics and
suggested different theory. Independently, and at almost
a
the same time, Einstein arrived at the same
conclusion as a part of his development of the
theory of special relativity.
PROBLEM 10.9. Show that the points shown as filled

circles in Fig. 10.9 are the
af stable equilibrium and those shown as open circles are the points of unstable
P o i
e q u i l i b r i u m .
The phase trajectories of the pendulum on a rotating plane are given
by
/2
n=+E-(zcos 20-22cos e (10.36)
where E' is the constant of integration and are shown in Fig. 10. 10.
Comparison with the phase trajectories of the nonrotating pendulum
depicted in Fig. 10.7 shows the effect of rotation. Now there are two
centres O and O2 placed symmetrically with respect to the origin. The
periodic motion abut these centres is no longer symmetric. The origin is
a saddle point through which passes the separatrix S. The trajectories
outside S are stilll closed, representing periodic oscillation, the velocity
in each oscillation the origin. There is also a second
decreasing near
are
separatrix S2 passing through t r. Outside S2 the phase trajectories the
the system about the axis through
open and they represent rotation of
point of support.
-
***
S2
*o****
A.5,a=2
Fig. 10.10
conductor is fixed
placed near a
PROBLEM 10.10. A movable current carrying
current carrying conductor. The motion of the movable conductor is constrained byy
a spring of stiffness constant k. The equation of motion of the conductor may be
written as
+ fx, A) = 0
where
flx, ) = -k [x - (A/(a - x)
currents in the two conductors and the

on the
Here is a parameter which depends integral of the system
and hence
Obtain a first
dimensions of the system. A <a14, (i) l =a'14
physical trajectories when (i)
trajectories. Draw
the equation of phase s0.
Examine also the c a s e s a
and > a14.
10.31 Volterra's Problem

nonlinear system is provided
example of a conservative
An interesting connection with the
equations derived by Volterra in
by the coupled one of which thrives on
(or decay) of populations of two species
growth of the type (10.22)
The does not possess an integral
system
the other. single
as conservative
since it possesses a
but is yet termed formally
valued first integral".
be stated in the following
The problem which Volterra posed may
there exists a species x,
which lives on, say,
In a closed region
way. rate of growth of x
sufficient vegetation, then the
vegetation. If there is
is proportional to its number:
dxldt = kix (10.37)
subsists
species, say y, which
on
In the same region there exists another
at rate proportional to its number:
x. So, if there is nox, it will decay a
dyldt =- kzy (10.38)

and (10.38) is quite clear. If the
The implication of Equations (10.37)
Similarly, if y is left
species x is left alone, it will grow exponentially.
Volterra argued that if the
alone it will decay exponentially with time.
interaction of the two species is taken into account, then x will not grow
as fast as given by Equation (10.37)

because some of its members will
members consumed
be consumed by the species y. Since the number of
is proportional to y, it is appropriate to replace k by (ki KY), where -
a constant. By the same token, since the food for y

is available, their
is
number will not dwindle as fast as that indicated by Equation (10.38)
and k must be replaced by (k2 x), where ^ is another constant. The
equations of growth and decay of the two species may now be written
as
dxldt = (k - Yy)x (10.39)

dyldt = - k2 - T)y (10.40)
It may be noted that k and k summarize the birth and mortality conditions
of the two populations and, in the main, determine the sizes of the
'For a discussion of nonlinear conservative systems, see Minorsky op. cit.

populations. 7 and 72 measure the interaction between the species, n

measuring of the ability of the first population to protect itself and
measuring the ability of the second one to attack the first one. As a first
it may suffice to use Equations (10.37) and
approximation, therefore,
(10.38) to get k^x and ky.
Multiplying Equation (10.39) by kzlx and Equation (10.40) by ki/y,
adding the two and using Equations (10.37) and (10.38), we get
h dxddt+ Y dyldt -k, d In xldt - k d In yldt = 0 (10.41)
which can be readily integrated to give a first integral

hx+y-k In x -kjln y = C (10.42)
which is single valued, which makes the system conservative according
to the criterion stated above. It is seen that the origin, which is a singular
therefore a
point of Equations (10.39) and (10.40) is a saddle point and
15
point of unstable equilibrium. The other singular point, (&/h. K/7)
a centre and is the point of stable equilibrium.
PROBLEM 10.11. Shift the origin in Equations (10.39) and (10.40) to (k/h» + ),
k / +) where and n are small quantities. Show then that (k2/¥» ks/y) is a centre.
The integral curves of Equations (10.39) and (10.40) are shownin

Fig. 10.11. Obviously only the first quadrant ofthe (x, y) plane is relevant.
ka/M2
Fig. 10.11
The nature of singular point at the origin as a saddle point and that at
centre is confirmed. It is clear that if the original
ka h, k/y) as a
populations of the species given by the coordinates of the centre,

are
remain unchanged
namely, =k/h and yj =ki/7. then these populations
with time. The integral closed curves, indicating that the
curves are
Near the centre the curves are
populations execute stable oscillations.
almost circles, but as we move away from the centre, the integral curves
are compressed on the side of the origin. On the side away from the
284
curious phenomenon,
well spread out. This is a very
origin the curves are large fluctuations
have very
we can
that if we start with small populations show that the density
The observations, however,
and in the long
in the two populations. constant on the average
of each species remains roughly formulation of Equations
leading to the
run. Actually. the assumptions that the
rather simplistic; there is no certainty
(10.39) and (10.40) are
accordance with these
behaves in
growth and decay of populations has
of view the main interest in the problem
equations. From our point of equations
first integral by a system
been the existence of a single-valued
the point of view of the
Such as The
this. of populations from
growth
phenomena of chaos is discussed later.
10.4 Limit Cycles

oscillator in the presence of
Let us now consider the equation of an
amplitude
instantaneous and
dissipation (or excitation) dependent
on
of motion of this system is

proportional to velocity. The equation
x+bx)i +fx) = 0 (10.43)
may transform it to the form

As in the case of Equation (10.22), we
T+V) = - b(x)i? (10.44)
where T= (1/2)** and V=| fx) dx Suppose the solution of Equation
(10.43) is quasiperiodic, i.e,

r= A cos (@t + Ù) (10.45)
where w and o are slowly varying functions ofx. Suppose further that
the dissipative term can be separated from the exciting term in b(r). Then
we may integrate over one cycle and write
2 b(x) dx =
AEdis -
AEexe (10.46)
where AEdis is the energy dissipated over one cycle and AEexe is the energy
pumped into the system over a complete cycle. When the dissipation is
linear, say given by 26x, then over one cycle
AE dis = 28t dt =24 or (10.47)
This quantity is sketched in
Fig, 10.12a. Also sketched in this is
AEexe on the assumption that it is independent of the amplitude.figure
At the
AEdis
AE exc
/AE dis
AEext
AA A2
(a)
Limit
Cycle
(b)
Fig. 10.12
point A = A*, the energy dissipated is compensated fully by the energy

fed to the system, so that the net gain or loss of energy by the system
is zero. So, the system behaves like an ideal oscillator whose trajectory
in the phase space is a closed curve (Fig. 10.12b). At a point like Aj < A*,
there is net gain of energy by the system and therefore the phase space
trajectory is a spiral issuing origin. At A2, the trajectory is again

from the
a spiral, but in this case the spiral

tends to wind in. The isolated closed
curve among spirals in this case
is an example of a limit cycle. In case
in 10.13, there are two limit
AEdis and AEexc vary with A as depicted Fig.
cycles.
It was shown by Poincaré that a system of equations of the form
i= P(x, y; j =Qr, y)
of which Equations (10.18) and (10.43) are special cases, possess
closed trajectories in the phase
occasionally solutions which give isolated
If all other trajectories approach the
plane. These he called limit cycles.
limit cycle. For example, the limit
limit cycle, it is said to be a stable
one. A stable limit cycle is another example
cycle of Fig. 10.12 is a stable
of an attractor: it attracts all the trajectories in its neighbourhood. On
the other hand, if all the trajectories move away from it (for example,
the curve C2 in Fig. 10.13), the limit cycle is called an unstable limit
A Edis
Axc /AE dis

AE xc
(a)i
Limit
cycles
(b)
Fig. 10.13
approaches the limit

cycle. If one set of trajectories (inside or outside)
from it, the limit cycle is said to
cycle, and the other set moves away
because it does not satisfy
be semi-stable, although it is actually unstable
above. The existence of
the requirement of a stable limit cycle outlined
of stable
stable limit cycles in dissipative systems indicates the presence
oscillations.
PROBLEM 10.12. Solve the system of equations
i= - y+x(x* + y - 1; y =x+ y(r* + y - 1)
sketch the phase space trajectories and show that there exists a limit cycle which is
unstable. [Hint: you can solve the system numerically, or alternatively you can
express the system in terms of polar coordinates, in which form the equations can
be solved analytically.]
PROBLEM 10.13. The system of equations
=y y= -f*-k{r* - 1y
where k is a positive parameter, is equivalent to the Van der Pol equation:
+k(x 1)i+fx = 0
Solve the system numerically and show that there exists a stable limit cycle shown
in Fig. 10.14a. Examine also what happens when k is negative [Fig. 10.146].
For solving the Van der Pol

equation approximately, reference may
be made to the books of
Hagedorn (1981) and Minorsky op. cit.
Limit v.k0.2,f-1
cycle
(a)
k-0.3, f1
Limt
Cycle
(b)
Fig. 10.14
for the damped nonlinear

The existence of limit cycles is very important
electron tube oscillator. First, the oscillations
Oscillatory system, such as an of the initial conditions. This
indicated by limit cycles are independent
can be understood by the following
argument. Consider a trajectory C
outside it (Fig. 10.15). We
inside the limit cycle C and a trajectory C2
ot the phase space there passes only one
recall that through each point
the initial conditions, and therefore C and C
trajectory determined by
different initial conditions. As C and C, (and
must have been caused by
C and ultimately merge to
for that matter any other trajectory) approach
that C cannot have anything to do with the
give way to C, it is clear
not matter when and how the circuit of the
initial conditions. So, it does
OsCillations are established if the circuit
tube oscillator is closed, the
the oscillations indicated by
parameters have the right values. Secondly,
the limit cycles are self-exciting, that is they take place in the absence
Limit
cycle
C2
Fig. 10.15
tube oscillator, for

of any periodic forces acting on the system. In the
oscillations to take place; closing
example, no impulse is necessary for the
the circuit is all that is required. Systems in which very
little impulse is
needed to start stable oscillations are called soft self-exciting systems. In
these systems the focus is an unstable point. Contrast these with systems
of the kind shown in Fig. 10.13. Here the focus is a stable point and C2
is an unstable limit cycle. In this case we need an impulse which is large
enough to place the system in a state between the limit cycles Ci and C2
for the stable oscillations to occur. Such systems in which the initial
conditions have to be far from equilibrium position are called hard self
exciting systems.
Since limit cycles are so important for the nonlinear oscillating systems,
lot of work has been done to discover conditions under which systems
governed by differential equations such as
i= Px, y). y Qx, y)

=
(10.48)
may possess limit cycles. For this Poincaré devised a scheme of associating
indices with the critical points and the closed phase
plane trajectories.
We have seen above that x and y may be considered as components of
a vector field in the phase plane. Suppose we draw a closed curve C
round a critical point in the phase plane and let a
point S move on this
closed curve in the positive
we note down the
(counterclockwise) direction. At each point
angle between the tangent to the vector at that point
and the x-axis. Then as S
completes one full circuit of the closed curve,
the tangent will resume its initial
direction having
angle 2, where J is an integer. Jis called the index ofturned through an
the critical point.
Since tan o = ýli = Qx,
yVP(x, y) (Fig. 10.16), the index can be calculated
from a line integral round C We
have
Nonlinear Oscillations and
Chaos 289
Fig. 10.16
J=dtan (O/P)]=E Jc P dQP2+02Q - dP

(10.49)
Ifthe singular point is a node, focus or centre, then.J= + 1; if the singular
point is a saddle point, then J=-1. If there is no critical point within
C, then J = 0. The integral (10.49) can also be computed if there are
more than one critical points inside C. It is then called the index of the
curve C. It is given by
I= N + F+ C- S
where Nis the number of nodes, F is the number of foci, C is the number
of centres and S is the number of saddle points enclosed within the
closed curve. Poincaré then proved that a necessary condition for an
isolated closed curve to be a limit cycle is that its index value is +1. This
that closed curve must contain within it at least a
necessarily
implies a
one critical points, then
focus, a node, or a centre. If there are more than
one less than the
it means that the number of saddle points must be
number of all other critical points
condition, its main use
The condition stated above is only a necessary
exist. There is also another criterion,
is to tell us where limit cycles cannot
when limit cycles will not exist. For
due to Bendixson, which tells us
the negative criterion of Bendixson. To
this reason it is usually called
understand it consider the quantity
(Plax + dQldy)
does not vanish
on a domain of thespace. If this quantity
phase
values on D then no
take positive and negative
identically or does not the Green's theorem,
can exist on
. To get this quantity we use
limit cycle
according to which
(OPlk +00ldy) drdy (10.50)

(Pdy-Q dr) | | =
C
where C is the boundary of I and the integral on the right hand side
extends over the whoie region 9 Now, if C were a phase trajectory of
vanish identically. This
the system (10.48) then the left hand side would
vanish, which it cannot
would require that the right hand side must also
do if it always has the same sign on D. Thus, if the quantity
(OPldx + dQldy)
9, no limit cycle
does not vanish identically or does not change sing on
can exist on D.
EXAMPLE 10a. Consider the equation
Ixl )* +x =0 (10.51)
*+(1 -
It be written as
equation for negative k.
can
which is similar to the van der Pol
the set
i = y; y= -x- (1 -Ix)y (10.52)
We havee
Ixl
OPldx + OQidy =- 1 +
this strip
l <x<lofthe xy-plane. In
only negative values in the strip
-
which takes
the negative criterion.
no limit cycles can exist according to
and show that in the strip
PROBLEM 10.14. SolveEquation (10.52) numerically
10.17).
-
1 <x <l of the xy-plane there exist no limit cycles (Fig.
Limit cycle
Fig. 10.17
if limit cycles are possible for the system whose

PROBLEM 10.15. Investigate
equation is given by
i - [O.1 (10/3)*]+ x-x ' = 0
Verify your answer by integrating the equation numerically.

10.5 Chaos
f
One the difficulties
introduced by the nonlinearities
in a system is
that its long term behaviour may becomes sensitive to the
Amall change in one or more of the initial conditions initial conditions.
or the course that the system would follow, and the could completely
eventual outcome
could no longer be predicted. It is this
unpredictability
associated with
the sensitivity to the initial conditions that has been termed chaos.
To be sure the unpredictability referred to above is not due to the
uncertainty which results from the
application of the laws of quantum
mechanics. Here the system is classical; its motion is governed by the
laws of classical mechanics and is, therefore, completely deterministic;
and yet its long term behaviour is chaotic. The phenomenon of chaos in
classical systems is, therefore, called deterministic chaos. Following gives
a brief description of this phenomenon.
As an example of a simple system whose motion becomes chaotic,
consider a forced damped nonlinear oscillator whose equation may be
written as
ar2
2b+w sin x =fcos wt (10.53)
where fis the amplitude of the periodic forcing term of frequency . This
seemingly simple equation cannot be solved exactly. Numerically, the
behaviour of x as a function of t is shown in Fig. 10.18. Here we observe
= 1 , Wo=1, b=.22,f= 2.7
Fig. 10.18
characteristic
ofa chaotic
is
the disorderly
excursion of x with t which
motion.
10.51 Logistic Map indeed the phenomenon of

associated with chaos,
of the concepts introduced following the study
Many sense,
was
particular
chaos itself in this
of nonlinear difference equation
Fi(x) (10.54)
yl Ax, (1 -xn) =
Such difference
the value of x afterj iterations.
constant and x;
where A is a (10.54) applies
known as logistic equations or maps. Equation socio-economic
equations are and even
chemical, biological
to many physical, connection with the growth of
We derive this equation in
phenomena.
biological populations. the death rates
difference between the birth and
the
Suppose r(t) is P may
of population, then the rate of growth of population
per head
a
be written as
dP(O)/dt = r()PO)
or simply as
dPldt = rP (10.55)
the solution of which is

P Ce" (10.56)
We see that if r is positive, the population grows
where C is a constant.
to infinity. On the other hand, if r is negative, the population dies after

are known to persist for long
times
some time. Actually, populations
infinite sizes. This means that r
without either vanishing or reaching
function of P and must
must be modulated by P; it must be a decreasing
become negative when P becomes large. Keeping this in view, we try
relation between r and P:

a simple
r= roll- (PIK)]| (10.57)
where ro is a constant and K, another constant, is called the carrying
capacity, for the simple reason that when P= K, the growth rate becomes
zero, indicating that the resources are just sufficient to support this much
population.
PROBLEM 10.16. Show that the assumption contained in Equation (10.57) 1is
equivalent to either birth rate falling linearly with P, or the death rate rising linearly
with P, or both.

VB Bhatia 1

Uploaded by

Copyright:

Available Formats

VB Bhatia 1

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

VB Bhatia 1

Uploaded by

Copyright:

Available Formats

10.

Nonlinear Oscillations and

Rut this 1S not wnat we dre

of coupled first order differential equations. The phase trajectories can

Recall that the substitution of the trial solution e in Equation (10.3)

Solution, therefore, depends upon the nature of the quantity b - aw)

conveniently by numerical methods. The trajectories are shown in

J X1, X2, .., X t) j = 1, 2,.. n (10.5)

x = P(x, y). j = Qx, y) (10.6)

offshoot is that be eliminated and we can write

curve. In fact, i and y may be thought of as defining the components

by the uniqueness of the solution. Thus, no twointegral curves can have

singularity typified by a centre is therefore a result of a

(10.4). Ifb is + ve the trajectory approaches the singular point (as in

For a detailed discussion of the singular points of autonomous systems

(b< ) are logarithmic spirals.

examples approach the singular point, it is, therefore, stable. If the

a saddle point is an unstable

the origin is a saddle point.

PROBLEM 10.5. Solve the equations

and find the nature of the singular point.

Show that the origin is a node for trajectories in

10.3 Nonlinear Oscillations

that the solution is still sinusoidal

x(0) = A sin ør, @ =

(05A-/8)] Asin øt + (0AS/24) sin 3wt (10.15)

The smallness of A implies

While this is a good enough approximation, a better approximation may

energy V(x) defined to within arbitrary constant. So, at the singular

points of the integral curves, where f(x) 0, we have d V(r)/dx 0, which

the point of equilibrium

Equation (10.22) we have

y=t [2 (cosx - (1 -E*)] (10.24)

Abecomes the amplitude

sin /2) = sin (A/2) sin z (10.26)

k sin (A/2), and a =sin [sin (E/2)/sin (A/2)1. (10.28)

T (27t/wo) (1 + A/16) (10.31)

of the pendulum becomes

X Exi(T) + Erz(7) +...

ofx, A/dx vanishes and consequently tangent to the curve fx, A) = 0

point of equilibrium. The equilibrium would be stable if the potential

with the parameter 2 =g/21.

The system equivalent to Equation (10.33) is

7,= - j(6, A) (10.35)

-1.5 1 -0.5 0.51 1.5 A

PROBLEM 10.9. Show that the points shown as filled

The phase trajectories of the pendulum on a rotating plane are given

currents in the two conductors and the

10.31 Volterra's Problem

dyldt =- kzy (10.38)

as fast as given by Equation (10.37)

a constant. By the same token, since the food for y

dxldt = (k - Yy)x (10.39)

'For a discussion of nonlinear conservative systems, see Minorsky op. cit.

populations. 7 and 72 measure the interaction between the species, n

h dxddt+ Y dyldt -k, d In xldt - k d In yldt = 0 (10.41)

which can be readily integrated to give a first integral

The integral curves of Equations (10.39) and (10.40) are shownin

populations of the species given by the coordinates of the centre,

10.4 Limit Cycles