Phase Plane For Linear Control System and Design

Ch 9.
1: The Phase Plane: Linear Systems

There are many differential equations, especially nonlinear
ones, that are not susceptible to analytical solution in any
reasonably convenient manner.
Numerical methods provide one means of dealing with these
equations.
Another approach, presented in this chapter, is geometrical in
character and leads to a qualitative understanding of the
solutions rather than to detailed quantitative information.
Solutions of Second Order Linear Systems

Consider a second order linear homogeneous system with
constant coefficients of the form x' = Ax, where A is a 2 x 2
constant matrix and x is a 2 x 1 vector.
Recall from Chapter 7 that if we assume x = ert, then
re rt = Ae rt
r = A
(A rI ) = 0
Therefore x = ert is a solution of x' = Ax provided that r is an

eigenvalue and is an eigenvector of the coefficient matrix A.
The eigenvalues are the roots of the polynomial equation
det(A-rI) = 0, and the eigenvectors are determined up to an
arbitrary constant from the equation (A-rI) = 0.
Equilibrium Solution, Phase Portrait

Solutions x for which Ax = 0 correspond to equilibrium
solutions, and are called critical points.
We assume A is nonsingular, or detA 0, and hence x = 0 is
the only critical point for the system x' = Ax.
A solution of x' = Ax is a vector function x = (t) that satisfies
the differential equation, and can be viewed as a parametric
representation for a curve in the x1 x2-plane.
This curve can be regarded as a trajectory traversed by a
moving particle whose velocity dx/dt is specified by the
differential equation.
The x1x2-plane is called the phase plane, and a representative
set of trajectories is a phase portrait.
Characterizing Equation by Trajectory Pattern

In analyzing the system x' = Ax, we must consider several
cases, depending on the nature of the eigenvalues of A.
These cases also occurred in Sections 7.5 7.8, where we
were primarily interested in finding a convenient formula for
the general solution.
Now our main goal is to characterize the differential equation
according to the geometric pattern formed by its trajectories.
In each case we discuss the behavior of the trajectories in
general an illustrate it with an example.
It is important to become familiar with the types of behavior
that the trajectories have for each case, as they are the basic
ingredients of the qualitative theory of differential equations.
Case 1: Real Unequal Eigenvalues

of the Same Sign (1 of 3)
When the eigenvalues r1 and r2 are both positive or both
negative, the general solution for x' = Ax is
x = c1 (1) e r1t + c2 ( 2) e r2t
Suppose first that r1 < r2 < 0, and that the eigenvectors (1)
and (2) are as shown below.
It follows that x 0 as t for all solutions x, regardless
of the values of c1 and c2.
Case 1: Nodal Sink
(2 of 3)
If the solution starts at an initial point on the line through

(1), then c2 = 0 and the solution remains on this line for all t.
Similarly if the initial point is on the line through (2).
The solution can be rewritten as
x = c1 (1) e r1t + c2 ( 2) e r2t = e r2t c1 (1) e (r1 r2 )t + c2 ( 2 )
Since r1 - r2 < 0, for c2 0 the term c1(1)e(r1 - r2)t is negligible

compared to c2 (2), for large t.
Thus all solutions are tangent to (2) at the
critical point x = 0 except for solutions
that start exactly on the line through (1).
This type of critical point is called a node
or nodal sink.
Case 1: Nodal Source
(3 of 3)
The phase portrait along with several graphs of x1 versus t

are given below. The behavior of x2 versus t is similar.
If 0 < r2 < r1, then the trajectories will have the same pattern
as in figure (a) below, but the direction will be away from
the critical point at the origin. In this case the critical point
is again called a node or a nodal source.
Case 2: Real Eigenvalues

of Opposite Sign (1 of 3)
Suppose now that r1 > 0 and r2 < 0, with general solution
x = c1 (1) e r1t + c2 ( 2 ) e r2t ,
and corresponding eigenvectors (1) and (2) as shown below.
If the solution starts at an initial point on the line through (1),
then c2 = 0 and the solution remains on this line for all t.
Also, since r1 > 0, it follows that ||x|| as t .
Similarly if the initial point is on the line through (2), then
||x|| 0 as t since r2 < 0.
Solutions starting at other initial points
have trajectories as shown.
Case 2: Saddle Point
(2 of 3)
For our general solution

x = c1 (1) e r1t + c2 ( 2 ) e r2t , r1 > 0, r2 < 0,
the positive exponential term is dominant for large t, so all
solutions approach infinity asymptotic to the line determined
by the eigenvector (1) corresponding to r1 > 0.
The only solutions that approach the critical point at the origin
are those that start on the line determined by (2).
This type of critical point is called a saddle point.
Case 2: Graphs of x1 versus t
(3 of 3)
The phase portrait along with several graphs of x1 versus t

are given below.
For certain initial conditions, the positive exponential term is
absent from the solution, so x1 0 as t .
For all other initial conditions the positive exponential term
eventually dominates and causes x1 to become unbounded.
The behavior of x2 versus t is similar.
Case 3: Equal Eigenvalues
(1 of 5)
Suppose now that r1 = r2 = r. We consider the case in which

the repeated eigenvalue r is negative. If r is positive, then the
trajectories are similar but direction of motion is reversed.
There are two subcases, depending on whether r has two
linearly independent eigenvectors or only one.
If the two eigenvectors (1) and (2) are linearly independent,
then the general solution is
x = c1 (1) e rt + c2 ( 2) e rt
The ratio x2/x1 is independent of t, but depends on the
components of (1) and (2) and on c1 and c2.
A phase portrait is given on the next slide.
Case 3: Star Node
(2 of 5)
The general solution is

x = c1 (1) e rt + c2 ( 2) e rt , r < 0
Thus every trajectory lies on a line through the origin, as seen
in the phase portrait below. Several graphs of x1 versus t are
given below as well, with the case of x2 versus t similar.
The critical point at the origin is called a proper node, or a
star point.
Case 3: Equal Eigenvalues
(3 of 5)
If the repeated eigenvalue r has only one linearly independent

eigenvector , then from Section 7.8 the general solution is
x = c1e rt + c2 (te rt + e rt )
For large t, the dominant term is c2tert. Thus every trajectory
approaches origin tangent to line through the eigenvector .
Similarly, for large negative t the dominant term is again
c2tert, and hence every trajectory is asymptotic to a line
parallel to the eigenvector .
The orientation of the trajectories
depends on the relative positions
of and , as we will see.
Case 3: Improper Node
(4 of 5)
We can rewrite the general solution as

x = (c1 + c2 + c2t )e rt = ye rt , y = c1 + c2 + c2t
Note that y determines the direction of x, whereas the scalar
quantity ert affects only the magnitude of x.
For fixed values of c1 and c2, the expression for y is a vector
equation of line through the point c1 + c2 and parallel to .
Using this fact, solution trajectories can be sketched for given
coefficients c1 and c2. See phase portrait below.
When a double eigenvalue has only
one linearly independent eigenvalue,
the critical point is called an improper
or degenerate node.
Case 3: Phase Portraits
(5 of 5)
The phase portrait is given in figure (a) along with several

graphs of x1 versus t are given below in figure (b).
When the relative orientation of and are reversed, the
phase portrait given in figure (c) is obtained.
Case 4: Complex Eigenvalues
(1 of 5)
Suppose the eigenvalues are i, where and are real,

with 0 and > 0.
It is possible to write down the general solution in terms of
eigenvalues and eigenvectors, as shown in Section 7.6.
However, we proceed in a different way here.
Systems having eigenvalues i are typified by

x =

x1 = x1 + x2
x
x2 = x1 + x2
We introduce the polar coordinates r, given by

r 2 = x12 + x22 , tan = x2 x1
Case 4: Polar Equations
(2 of 5)
Differentiating the polar equations

r 2 = x12 + x22 , tan = x2 x1
with respect to t, we have
dx
d x1 (dx2 / dt ) x2 (dx1 / dt )
dx
dr
2r = 2 x1 1 + 2 x2 2 , sec 2
=
x12
dt
dt
dt
dt
or
rr = x1 x1 + x2 x2 , sec 2 = ( x1 x2 x2 x1 ) x12
Substituting
x1 = x1 + x2 , x2 = x1 + x2
into these derivative equations, we obtain
r = r , =
Case 4: Spiral Point
(3 of 5)
Solving the differential equations

r = r , =
we have
r = ce t , = t + 0 , 0 = (0)
These equations are parametric equations in polar coordinates

of the solution trajectories to our system x' = Ax.
Since > 0, it follows that decreases as t increases, so the
direction of motion on a trajectory is clockwise.
If < 0, then r 0 as t , while r if > 0.
Thus the trajectories are spirals, which approach or recede
from the origin depending on the sign of , and the critical
point is called a spiral point in this case.
(4 of 5)
The phase portrait along with several graphs of x1 versus t are

given below.
Frequently the terms spiral sink and spiral source are used to
refer to spiral points whose trajectories approach, or depart
from, the critical point.
x = Ax
x1 x1
=

x2 x2
r 2 = x12 + x22 , tan = x2 x1
r = ce t , = t + 0
Case 4: General System
(5 of 5)
It can be shown that for any system with complex eigenvalues

i, where 0, the trajectories are always spirals.
They are directed inward or outward, respectively, depending
on whether is negative or positive.
The spirals may be elongated and skewed with respect to the
coordinate axes, and the direction may be either clockwise or
counterclockwise. See text for more details.
Case 5: Pure Imaginary Eigenvalues
(1 of 2)
Suppose the eigenvalues are i, where = 0 and real.

Systems having eigenvalues i are typified by
x1 = x2
0
x
x =
x2 = x1
0
As in Case 4, using polar coordinates r, leads to

r = c, = t + 0
The trajectories are circles with center at the origin, which are
traversed clockwise if > 0 and counterclockwise if < 0.
A complete circuit about the origin occurs in a time interval of
length 2 /, so all solutions are periodic with period 2 /.
The critical point is called a center.
(2 of 2)
In general, when the eigenvalues are pure imaginary, it is

possible to show that the trajectories are ellipses centered at
the origin.
The phase portrait along with several graphs of x1 versus t are
given below.
Behavior of Individual Trajectories

As t , each trajectory does one of the following:
approaches infinity;
approaches the critical point x = 0;
repeatedly traverses a closed curve, corresponding to a periodic
solution, that surrounds the critical point.
The trajectories never intersect, and exactly one trajectory

passes through each point (x0, y0) in the phase plane.
The only solution passing through the origin is x = 0. Other
solutions may approach (0, 0), but never reach it.
Behavior of Trajectory Sets

As t , one of the following cases holds:
All trajectories approach the critical point x = 0. This is the case when
the eigenvalues are real and negative or complex with negative real
part. The origin is either a nodal or spiral sink.
All trajectories remain bounded but do not approach the origin, and
occurs when eigenvalues are pure imaginary. The origin is a center.
Some trajectories, and possibly all trajectories except x = 0, tend to
infinity. This occurs when at least one of the eigenvalues is positive or
has a positive real part. The origin is a nodal source, a spiral source, or
a saddle point.
Summary Table
The following table summarizes the information we have
derived about our 2 x 2 system x' = Ax, as well as the stability
of the equilibrium solution x = 0.
Eigenvalues
r1 > r2 > 0
r1 < r2 < 0
r2 < 0 < r1
r1 = r2 > 0
r1 = r2 < 0
r1 , r2 = i
>0
<0
r1 = i , r2 = i
Type of Critical Point

Node
Node
Saddle Point
Proper or Improper Node
Proper or Improper Node
Spiral Point
Center
Stability
Unstable
Asymptotically Stable
Unstable
Unstable
Unstable
Stable
Ch 9.2: Autonomous Systems and Stability

In this section we draw together and expand on geometrical
ideas introduced in Section 2.5 for certain first order equations
and Section 9.1 for second order linear homogeneous systems
with constant coefficients.
These ideas concern the qualitative study of differential
equations and the concept of stability.
Initial Value Problem

We are concerned with systems of two simultaneous
differential equations of the form
dx / dt = F ( x, y ), dy / dt = G ( x, y )
We assume that the functions F and G are continuous and have

continuous partial derivatives in some domain D of xy-plane.
If (x0, y0) is a point in D, then by Theorem 7.1.1 there exists a
unique solution x = (t), y = (t), defined in some interval I
containing t0, satisfying the initial conditions
x(t0 ) = x0 , y (t0 ) = y0
Vector Form
We can write the initial value problem
dx / dt = F ( x, y ), dy / dt = G ( x, y ), x(t0 ) = x0 , y (t0 ) = y0
in vector form:
x F ( x, y ) 0 x0
, x =
x = =
y G ( x, y )
y0
or
dx / dt = f (x), x(t0 ) = x 0
where x = xi + yj, f(x)=F(x,y)i + G(x,y)j, x0 = x0i + y0j, and

1
0
i = , j =
0
1
In vector form, the solution x = (t) = (t)i + (t)j is a curve

traced out by a point in the xy-plane (phase plane).
Autonomous Systems
For our initial value problem
note that the functions F and G depend on x and y, but not t.

Such a system is said to be autonomous.
The system x' = Ax, where A is a constant matrix, is an
example of an autonomous system. However, if one or more
of the elements of the coefficient matrix A is a function of t,
then the system is nonautonomous.
The geometrical qualitative analysis of Section 9.1 can be
extended to two-dimensional autonomous systems in general,
but is not as useful for nonautonomous systems.
Phase Portraits for Autonomous Systems

Our autonomous system
dx / dt = F ( x, y ), dy / dt = G ( x, y )
has a direction field that is independent of time.

It follows that only one trajectory passes through each point
(x0, y0) in the phase plane.
Thus all solutions to an initial value problem of the form
lie on the same trajectory, regardless of the time t0 at which

they pass through (x0, y0).
Hence a single phase portrait displays important qualitative
information about all solutions of the system.
Stability and Instability

For the following definitions, we consider the autonomous
system x' = f(x) and denote the magnitude of x by ||x||.
The points, if any, where f(x) = 0 are called critical points. At
these points x' = 0 also, and hence critical points correspond to
constant, or equilibrium, solutions of the system of equations.
A critical point x0 is said to be stable if, for all > 0 there is a
> 0 such that every solution x = (t) satisfying ||(0) - x0|| <
exists for all positive t and satisfies ||(t) - x0|| < for all t 0.
Otherwise, x0 is unstable.
Asymptotic Stability
A critical point x0 is said to be asymptotically stable if it is
stable and if there exists a 0 > 0 such that if a solution x = (t)
satisfies ||(0) - x0|| < 0, then (t) x0 as t .
Thus trajectories that start sufficiently close to x0 not only
stay close to x0 but must eventually approach x0 as t .
This is the case for the trajectory in figure (a) below but not for
the one in figure (b) below.
Thus asymptotic stability is a stronger property than stability.
However, note that
lim (t ) = x 0
t
does not imply stability.
The Oscillating Pendulum

The concepts of asymptotic stability, stability, and instability
can be easily visualized in terms of an oscillating pendulum.
Suppose a mass m is attached to one end of a weightless rigid
rod of length L, with the other end of the rod supported at the
origin O. See figure below.
The position of the pendulum is described by the angle , with
the counterclockwise direction taken to be positive.
The gravitational force mg acts downward,
with damping force c|d /dt |, c > 0, always
opposite to the direction of motion.
Pendulum Equation
The principle of angular momentum states that the time rate of
change of angular momentum about any point is equal to the
moment of the resultant force about that point.
The angular momentum about the origin is mL2(d /dt ), and
hence the governing equation is
2
d
d
mL2 2 = cL
mgL sin
dt
dt
Here, L and Lsin are the moment arms

of the resistive and gravitational forces.
This equation is valid for all four sign
possibilities of and d /dt.
Autonomous Pendulum System

Rewriting our equation in standard form, we obtain
+ + 2 sin = 0, = c mL, 2 = g L
To convert this equation into a system of two first order
equations, we let x = and y = d /dt. Then
x = y, y = 2 sin x y
To find the critical points of this autonomous system, solve
y = 0, 2 sin x y = 0 x = n , y = 0
These points correspond to two physical equilibrium positions,
one with the mass directly below point of support ( = 0), and
the other with the mass directly above point of support ( = ).
Intuitively, the first is stable and the second is unstable.
Stability of Critical Points: Damped Case

If mass is slightly displaced from lower equilibrium position, it
will oscillate with decreasing amplitude, and slowly approach
equilibrium position as damping force dissipates initial energy.
This type of motion illustrates asymptotic stability.
If mass is slightly displaced from upper equilibrium position, it
will rapidly fall, and then approach lower equilibrium position.
This type of motion illustrates instability
See figures (a) and (b) below.
Stability of Critical Point: Undamped Case

Now consider the ideal situation in which the damping
coefficient c (or ) is zero.
In this case, if the mass is displaced slightly from the lower
equilibrium position, then it will oscillate indefinitely with
constant amplitude about the equilibrium position.
Since there is no dissipation in the system, the mass will
remain near the equilibrium position but will not approach it
asymptotically. This motion is stable but not asymptotically
stable. See figure (c) below.
Determination of Trajectories
Consider the autonomous system
dx / dt = F ( x, y ), dy / dt = G ( x, y )
It follows that
dy / dx = G ( x, y ) / F ( x, y ),
which is a first order equation in the variables x and y.

If we can solve this equation using methods from Chapter 2,
then the implicit expression for the solution, H(x,y) = c, gives
an equation for the trajectories of
dy / dx = G ( x, y ) / F ( x, y )
Thus the trajectories lie on the level curves of H(x,y).

Note this approach is applicable only in special cases.
Example 1
Consider the system
dx / dt = y, dy / dt = x
It follows that
dy / dx = x / y
ydy = x dx
The solution of this separable equation is

H ( x, y ) = y 2 x 2 = c
Thus the trajectories are hyperbolas, as shown below.

The direction of motion can by inferred
from the signs of dx/dt and dy/dt in the
four quadrants.
Example 2: Separable Equation
(1 of 2)
Consider the system

dx / dt = 4 2 y, dy / dt = 12 3 x 2
It follows that
dy 12 3 x 2
=
dx
4 2y
(4 2 y )dy = (12 3x 2 )dx
The solution of this separable equation is

H ( x, y ) = 4 y y 2 12 x + x 3 = c
Note that the equilibrium points are found by solving

4 2 y = 0, 12 3 x 2 = 0 x = 2, y = 2
and hence (-2, 2) and (2, 2) are the equilibrium points.
Example 2: Phase Portrait
(2 of 2)
We have
H ( x, y ) = 4 y y 2 12 x + x 3 = c
A graph of some level curves of H are given below.

Note that (-2, 2) is a center and (2, 2) is a saddle point.
Also, one trajectory leaves the saddle point (at t = -), loops
around the center, and returns to the saddle point (at t = ).
Ch 9.3: Almost Linear Systems

In Section 9.1 we gave an informal description of the stability
properties of the equilibrium solution x = 0 of the 2 x 2 system
x' = Ax. The results are summarized in Table 9.1.1.
We required detA 0, and hence x = 0 is the only critical point
of the system x' = Ax.
Now that we have precisely defined the concepts of asymptotic
stability, stability, and instability, we can restate these results
in the Theorem 9.3.1, on the next slide.
Theorem 9.3.1
The critical point x = 0 of the 2 x 2 linear system x' = Ax is
(1) asymptotically stable if the eigenvalues r1 and r2 are real
and negative or have negative real part;
(2) stable, but not asymptotically stable, if r1 and r2 are pure
imaginary;
(3) unstable if r1 and r2 are real and either is positive, or if they
have positive real part.
Perturbations
Thus by Theorem 9.3.1, or Table 9.1.1, the eigenvalues r1, r2
of A determine the type of critical point at x = 0 and its
stability characteristics.
Now r1, r2 depend on the coefficients in the system x' = Ax,
which in turn may depend on physical measurements.
Since these measurements are typically subject to small
uncertainties, it is of interest to investigate whether the small
changes (perturbations) in the coefficients can affect the
stability or instability of a critical point and/or significantly
alter the pattern of the trajectories.
Perturbations in Pure Imaginary Eigenvalues

Recall that the eigenvalues r1 and r2 of A are the roots of the
polynomial equation det(A-rI) = 0.
It can be shown that small perturbations in some or all of the
coefficients are reflected in small changes in the eigenvalues.
The most sensitive situation occurs when r1 = i and r2 = - i,
that is, when the critical point is a center.
Small changes to the coefficients results in r1, r2 taking on new
values: r1 = ' + i' and r2 = ' - i', where ' 0 and ' .
Perturbations and Centers

Thus r1, r2 become r1 = ' + i' and r2 = ' - i'.
If ' 0, which almost always happens, then trajectories of
perturbed system are spirals rather than closed curves.
The system is asymptotically stable if ' < 0; unstable if ' > 0.
Thus small perturbations in coefficients may change a stable
system into an unstable one, and in any case may be expected
to alter radically the trajectories in the phase plane.
Perturbations in Equal Eigenvalues

Another slightly less sensitive case occurs when r1 = r2, that is,
when the critical point is a node.
Small perturbations in the coefficients will normally cause the
two equal roots to separate (bifurcate).
If the separated roots are real, then the critical point remains a
node, but if the separated roots are complex conjugates, then
the critical point becomes a spiral point.
Here, the stability or instability of the system is not affected by
small changes in the coefficients, but the trajectories may be
altered considerably.
Perturbations in Other Cases

In all other cases the stability or instability of the system is not
changed, nor is the type of critical point altered, by sufficiently
small perturbations in the coefficients of the system.
For example, if r1 and r2 are real, negative, and unequal, then a
small change in the coefficients will neither change the sign of
r1 and r2 nor allow them to coalesce. Thus the critical point
remains an asymptotically stable node.
Nonlinear Systems
Consider a nonlinear two-dimensional autonomous system
x' = f(x)
Our main object is to investigate the behavior of trajectories of
this system near a critical point x0.
We do this by approximating the nonlinear system near a
critical point x0 by an appropriate linear system, whose
trajectories are easy to describe.
The crucial question is whether the trajectories of the linear
system are good approximations to those of nonlinear system.
For convenience, assume critical point is at the origin, x0 = 0.
This involves no loss of generality, since in general the
substitution u = x x0 shifts the critical point to the origin.
Nonlinear and Nearby Linear Systems

First, we consider what it means for a nonlinear system to be
close to a linear system.
Accordingly, suppose that x' = f(x) = Ax + g(x).
Assume that x = 0 is an isolated critical point of this system.
This means that there is some circle about the origin within
which there are no other critical points.
In addition, assume that detA 0, and hence x = 0 is also an
isolated critical point of the system x' = Ax.
For the nonlinear system x' = f(x) to be close to the linear
system x' = Ax, we must assume that g(x) is small.
Almost Linear Systems: Vector Form

Recall: x' = f(x) = Ax + g(x), where g(x) is small.
More precisely, assume the components of g have continuous
first partial derivatives and satisfy the limit condition
lim
x 0
g ( x)
x
=0
Thus ||g(x)|| is small compared to ||x|| near the origin.

Such a system is called an almost linear system in the
neighborhood of the critical point x = 0.
Almost Linear Systems: Scalar Form

Recall: x' = f(x) = Ax + g(x), where
lim
x 0
g ( x)
x
=0
In scalar form, if we let

g ( x, y )
x
x = , g (x) = 1
y
g 2 ( x, y )
then
x = x2 + y 2 = r,
g (x) = g12 ( x, y ) + g 22 ( x, y ) ,
and the corresponding limit condition is

g1 ( x, y )
g 2 ( x, y )
lim
= 0 and lim
=0
0
r 0
r
r
r
Example 1: Nonlinear System
(1 of 2)
Consider the system
x 2 xy
x 1 0 x
=
+
2
y 0 0.5 y 0.75 xy 0.25 y
Note that (0, 0) is a critical point and detA 0.
It can be shown that the other critical points are (0, 2), (1, 0)
and (0.5, 0.5). Thus (0, 0) is an isolated critical point.
The limit condition
lim
x 0
g ( x)
x
=0
is more readily verified if we use polar coordinates x = rcos

and y = rsin .
Example 1: Almost Linear System
(2 of 2)
Using polar coordinates x = rcos and y = rsin, we have

g1 ( x, y ) x 2 xy r 2 cos 2 r 2 cos sin
=
=
r
r
r
= r cos 2 + cos sin ,
g 2 ( x, y ) 0.75 xy 0.25 y 2 0.75r 2 cos sin 0.25r 2 sin 2

=
=
r
r
r
= r 0.75 cos sin + 0.25 sin 2
Thus
lim
r 0
g1 ( x, y )
g ( x, y )
= 0 and lim 2
=0
0
r
r
r
and hence our system is almost linear near the origin.
Example 2: Pendulum System

The motion of a pendulum system is given by
or
1 x
0
x 0
2
sin x x
y y
Note detA 0, and critical points are (n , 0) for n = 1, 2,.

Thus (0, 0) is an isolated critical point, with g1(x,y) = 0 and
2 (sin x x)
2 r 3 cos 3
g 2 ( x, y )
lim
= lim
= lim
= 0,
r 0
r
0
0
r
r
r
where, from the Taylor series representation of sinx,

sin x x = a1 x 3 + a2 x 5 + L
Thus this system is almost linear.
General Nonlinear System
(1 of 3)
Consider the general nonlinear system x' = f(x), or

x = F ( x, y ), y = G ( x, y )
We will show that this system is almost linear near a critical

point (x0, y0) whenever the functions F and G have continuous
partial derivatives up to order 2.
To see this, we use Taylor expansions about the point (x0, y0)
to write F(x,y) and G(x,y) in the form
F ( x, y ) = F ( x0 , y0 ) + Fx ( x0 , y0 )( x x0 ) + Fy ( x0 , y0 )( y y0 ) + 1 ( x, y ),
G ( x, y ) = G ( x0 , y0 ) + Gx ( x0 , y0 )( x x0 ) + G y ( x0 , y0 )( y y0 ) + 2 ( x, y )
where
lim
( x , y ) ( x0 , y 0 )
1 ( x, y )
( x x0 ) + ( y y0 )
2
lim
( x , y ) ( x0 , y 0 )
2 ( x, y )
( x x0 ) + ( y y0 )
2
=0
Rewriting General Nonlinear System
(2 of 3)
We have F(x,y) and G(x,y) in the form

F ( x, y ) = F ( x0 , y0 ) + Fx ( x0 , y0 )( x x0 ) + Fy ( x0 , y0 )( y y0 ) + 1 ( x, y ),
G ( x, y ) = G ( x0 , y0 ) + Gx ( x0 , y0 )( x x0 ) + G y ( x0 , y0 )( y y0 ) + 2 ( x, y )
Since (x0, y0) is a critical point, F(x0, y0) = G(x0, y0) = 0. Also,
note that dx /dt = d(x - x0)/dt and dy /dt = d(y - y0)/dt.
Thus the original system of equations
x = F ( x, y ), y = G ( x, y )
reduces to
d x x0 Fx ( x0 , y0 ) Fy ( x0 , y0 ) x x0 1 ( x, y )
dt y y0 Gx ( x0 , y0 ) G y ( x0 , y0 ) y y0 2 ( x, y )
General Nonlinear System: Almost Linear (3 of 3)

Thus our system of equations can be written as
d x x0 Fx ( x0 , y0 ) Fy ( x0 , y0 ) x x0 1 ( x, y )
dt y y0 Gx ( x0 , y0 ) G y ( x0 , y0 ) y y0 2 ( x, y )
In vector notation,
x x0
1 ( x, y )
du df 0
, (x ) =
x u + (x ), where u(x ) =
=
dt dx
2 ( x, y )
y y0
( )
Thus if F and G are twice differentiable, then the nonlinear

system of equations x' = f(x) is almost linear, and the linear
system that approximates the nonlinear system is given by
u1 x x0
d u1 Fx ( x0 , y0 ) Fy ( x0 , y0 ) u1
, where =
dt u2 Gx ( x0 , y0 ) G y ( x0 , y0 ) u2
u 2 y y0
(1 of 2)
The motion of a pendulum system is given by

Thus
F ( x, y ) = y, G ( x, y ) = 2 sin x y
The critical points are (n , 0) for n = 0, 1, 2,.

Since F and G are twice differentiable, the system of equations
is almost linear near each critical point. We have
Fx = 0, Fy = 1, Gx = 2 cos x, G y =
To find the approximating linear system at (x0, y0), we use
u x x0
d u1 Fx ( x0 , y0 ) Fy ( x0 , y0 ) u1
, where 1 =
dt u2 Gx ( x0 , y0 ) G y ( x0 , y0 ) u2
u 2 y y0
Example 3:
Approximating Linear Systems
(2 of 2)
To find the approximating linear system at (x0, y0), we use

u1 x x0
d u1 Fx ( x0 , y0 ) Fy ( x0 , y0 ) u1
, where =
dt u 2 Gx ( x0 , y0 ) G y ( x0 , y0 ) u 2
u 2 y y0
with
Fx = 0, Fy = 1, Gx = 2 cos x, G y =
At the origin, the approximating linear system is

1 x
d x 0

=
2
dt y y
At ( , 0), the approximating linear system is

1 x
d x 0
= 2
dt y y
Theorem 9.3.2
Consider the almost linear system x' = Ax + g(x). Let r1 and r2
be the eigenvalues of A. Then the type and stability of the
critical point (0,0) of the linear system x' = Ax and the almost
linear system x' = Ax + g(x) are as given in the table below.
Eigenvalues
r1 > r2 > 0
r1 < r2 < 0
r2 < 0 < r1
r1 = r2 > 0
r1 = r2 < 0
r1 , r2 = i
>0
<0
r1 = i , r2 = i
Linear System
Type
Node
Almost
Type
Node
Linear System
Stability
Unstable
Asymptotically
Stable
Unstable
Node
Asymptotically
Stable
Unstable
Proper or
Improper Node
Proper or
Improper Node
Unstable
Node or
Spiral Point
Node or
Spiral Point
Unstable
Spiral Point
Spiral Point
Unstable
Asymptotically
Stable
Stable
Spiral Point
Spiral Point
Unstable
Asymptotically
Stable
Indeterminate
Node
Saddle Point
Center
Stability
Unstable
Asymptotically
Stable
Saddle Point
Center or
Spiral Point
Asymptotically
Stable
Theorem 9.3.2 Discussion
(1 of 2)
Since nonlinear term g(x) is small compared to the linear term

Ax when x is small, we hope that the trajectories of the linear
system x' = Ax are good approximations to those of nonlinear
system, at least near the origin.
Theorem 9.3.2 states that this is true in many but not all cases.
For small x, the nonlinear terms are small and do not affect the
stability and type of critical point as determined by linear term,
except in two sensitive cases: r1 and r2 pure imaginary, and r1
and r2 real and equal.
As we have seen earlier, small changes in eigenvalues can alter
the type and stability of the critical point for a linear system,
but only in these two cases. Thus a small nonlinear term might
have a similar effect for these two cases as well.
Theorem 9.3.2 Discussion
(2 of 2)
Even if the critical point is of the same type as that of the

linear system, the trajectories of the almost linear system may
be much different in appearance than for the linear system,
except very near the critical point.
However, it can be shown that the slopes at which the
trajectories enter or leave the critical point are given
correctly by the linear equation.
Damped Pendulum System
(1 of 2)
Recall that the motion of a pendulum system is given by

Near the origin these nonlinear equations are approximated by
1 x
x 0

=
2
y y
whose eigenvalues are

2 4 2
r1 , r2 =
2
The nature of the solutions to the linear and nonlinear systems

depends on the sign of 2 42, as examined on the next slide.
2 4 2
r1 , r2 =
2
Damped Pendulum System
(2 of 2)
Thus we have the following cases:

If 2 4 2 > 0, then the eigenvalues are real, unequal, and negative.
The critical point (0,0) is an asymptotically stable node of the linear
system, and of the almost linear system.
If 2 4 2 = 0, then the eigenvalues are real, equal, and negative. The
critical point (0,0) is an asymptotically stable (proper or improper) node
of the linear system. It may be either an asymptotically stable node or a
spiral point of the almost linear system.
If 2 4 2 < 0, then the eigenvalues are complex with negative real
part. The critical point (0,0) is an asymptotically stable spiral point of
the linear system, and of the almost linear system.
Hence the origin is a spiral point of the nonlinear system if the

damping is small, and a node if damping is large enough. In
either case, the origin is asymptotically stable.
Pendulum System: Small Damping
(1 of 5)
Consider the small damping case, 2 4 2 < 0.

The direction of motion of the spirals near (0,0) can be
obtained directly from the equations
dx / dt = y, dy / dt = 2 sin x y
For the point at which a spiral intersects the positive y-axis, at
x = 0 and y > 0, it follows that dx/dt > 0. Thus the point (x,y)
on the trajectory is moving to the right, so the direction of
motion spirals clockwise.
(2 of 5)
The direction of motion of the spirals near (2n , 0) is the

same as near the origin.
As before, this can be obtained directly from the equations
dx / dt = y, dy / dt = 2 sin x y
We can expect this on physical grounds, since all these critical
points correspond to the downward equilibrium position of the
pendulum.
(3 of 5)
Next, consider the critical point (, 0). Here, the nonlinear

equations are approximated by the linear system
1 x
d x 0
= 2
dt y y
whose eigenvalues are

2 + 4 2
r1 , r2 =
2
One eigenvalue (r1) is positive and the other (r2) is negative.

Therefore, regardless of the amount of damping, the critical
point ( , 0) is an unstable saddle point both of the linear
system and of the almost linear system.
(4 of 5)
To examine the trajectories near the saddle point (, 0) in more

detail, consider the general solution of the linear system:
1
1
u
u x
= C1 e r1 t + C2 e r2 t , where =
v
v y
r1
r2
Since r1 > 0 and r2 < 0, it follows that the solution that

approaches zero as t corresponds to C1 = 0.
For this solution v/u = r2, thus slope of entering trajectories is
negative; one lies in second quadrant (C2 < 0) and the other
lies in the fourth quadrant (C2 > 0).
For C2 = 0, we obtain a pair of trajectories exiting from the
saddle point. They have slope r1 > 0; one lies in first quadrant
(C1 > 0) and the other lies in the third quadrant (C1 < 0).
(5 of 5)
The analysis at (, 0) can be repeated to show that the critical

points (n, 0), n odd, are all saddle points oriented in the same
way as the one at (, 0).
These critical points all correspond to the upward equilibrium
position of the pendulum, so we expect them to be unstable.
Diagrams showing the trajectories in the neighborhood of two
saddles points are given below.
(1 of 6)
The motion of a certain pendulum system is given by

x = y, y = 9 sin x y / 5
where x = and y = d /dt.

Note that 2 = 9, and thus the damping coefficient, = 1/5, is
relatively small. It follows that 2 4 2 < 0 here.
The phase portrait for this system is given below.
Example 4: Critical Points
(2 of 6)
The critical points are (n , 0) for n = 0, 1, 2,.

Even values of n, including (0,0), correspond to the downward
position of the pendulum, while odd values correspond to the
upward position.
Near each of the asymptotically stable critical points, the
trajectories are clockwise spirals that represent a decaying
oscillation about the equilibrium solution.
Example 4: Whirling
(3 of 6)
The wavy horizontal portions of the trajectories for larger

values of |y| represent whirling motions of the pendulum.
A whirling motion cannot continue indefinitely, since
eventually the angular velocity y is so much reduced by
damping that the pendulum can no longer go over the top, and
instead begins to oscillate about its downward position.
Example 4: Separatrix
(4 of 6)
The trajectories that enter the saddle points separate the phase
plane into regions. Such a trajectory is called a separatrix.
Each region contains exactly one of the spiral points.
The initial conditions on x = and y = d /dt determine the
position of an initial point (x, y) in the phase plane.
The subsequent motion of the
pendulum is represented by the
trajectory passing through the
initial point as it spirals toward
the asymptotically stable
critical point in that region.
Example 4: Basins of Attraction
(5 of 6)
The set of all initial points from which the trajectories

approach an asymptotically stable critical point is called the
basin of attraction, or the region of asymptotic stability, for
that critical point.
Each asymptotically stable critical point has its own basin of
attraction, which is bounded by the separatrices through the
neighboring unstable saddle points.
The basin of attraction is shown
in blue on the graph.
Example 4: Asymptotic Stability
(6 of 6)
An important difference between nonlinear autonomous

systems and the linear systems x' = Ax discussed in Section
9.1 is illustrated by the pendulum equations.
Recall that x' = Ax has only the single critical point x = 0 if
detA 0. Thus if the origin is asymptotically stable, then not
only do the trajectories that start close to the origin approach
the origin, but every trajectory approaches the origin. In this
case the critical point x = 0 is globally asymptotically stable.
This property is not true in general for nonlinear systems, and
thus it is important to determine, or estimate, the basins of
attraction for each asymptotically stable critical point.
Ch 9.4: Competing Species

In this section we explore the application of phase plane
analysis to some problems in population dynamics.
These problems involve two interacting populations and are
extensions of those discussed in Section 2.5, which dealt with
a single population.
Although the relationships discussed here are overly simple
compared to the complex relationships in nature, it is still
possible to acquire some insight into ecological principles
from a study of these model problems.
Logistic Equations
Suppose that in some closed environment there are two similar
species competing for a limited food supply.
For example, two species of fish in a pond that do not prey on
each other but do compete for the available food.
Let x and y be the populations of the two species at time t.
As in Section 2.5, assume that the population of each species,
in the absence of the other, is modeled by the logistic equation.
Thus
dx / dt = x( 1 1 x), dy / dt = y ( 2 2 y ),
where 1 and 2 are the growth rates of the two populations,
and 1/1 and 2 /2 are their saturation levels.
Competing Species Equations

However, when both species are present, each will impinge on
the available food supply for the other. In effect, they reduce
each others growth rates and saturation populations.
The simplest expression for reducing growth rate of species x
due to the presence of species y is to replace the growth factor
1 - 1x by 1 - 1x - 1y, where 1 is a measure of the degree
to which species y interferes with species x.
Similarly, we model the reduced growth rate of species y due
to presence of species x by replacing the growth factor 2 - 2y
by 2 - 2y - 2x.
Thus we have the system of equations
dx / dt = x( 1 1 x 1 y ), dy / dt = y ( 2 2 y 2 x)
Example 1: Population Equations
(1 of 8)
Consider the system of equations

dx / dt = x(1 x y ), dy / dt = y (0.75 y 0.5 x)
To find the critical points, we solve
x(1 x y ) = 0, y (0.75 y 0.5 x) = 0,
obtaining (0,0), (0,0.75), (1,0), and (0.5,0.5). These critical
points correspond to equilibrium solutions.
The first three points involve the extinction of one or both
species. Only the fourth critical point corresponds to the long
term survival of both species.
Other solutions are represented as trajectories in the xy-plane
that describe the evolution of the populations over time.
Example 1: Direction Field
(2 of 8)
A direction field for our system of equations is given below.

Only the first quadrant is depicted, as this corresponds to
positive population size.
The heavy dots in the figure correspond to the critical points.
Based on the direction field, (0.5,0.5) attracts other solutions
and therefore appears to be asymptotically stable.
The other three critical points
appear to be unstable.
To confirm these observations,
we can examine the linear
approximations to each point.
Example 1: Linearization
(3 of 8)
Our system of equations,

dx / dt = x(1 x y ), dy / dt = y (0.75 y 0.5 x)
is almost linear, since F and G are twice differentiable.
To obtain the linear system near a critical point (x0, y0), we use
the results of Section 9.3, given below.
u x x0
d u Fx ( x0 , y0 ) Fy ( x0 , y0 ) u
, where =
dt v Gx ( x0 , y0 ) G y ( x0 , y0 ) v
v y y0
Thus
d u 1 2 x0 y0
=
dt v 0.5 y0
x0
u

0.75 2 y0 0.5 x0 v
Example 1: Critical Point at (0,0)
(4 of 8)
For the critical point (0,0), the approximating linear system is

0 x
d x 1
=

dt y 0 0.75 y
The eigenvalues and eigenvectors are

r1 = 1,
(1)
1
= ;
0
r2 = 0.75,
( 2)
0
=
1
and hence the general solution for this linear system is

x
1 t
0 0.75 t
= c1 e + c2 e
y
0
1
Thus the origin is an unstable node of both the linear and

nonlinear systems. The trajectories near origin are all tangent
to y-axis, except for one trajectory that lies along the x-axis.
(5 of 8)

1 u
d u 1
=

dt v 0 0.25 v

r1 = 1,
(1)
1
= ;
0
r2 = 0.25,
( 2)
4
=
5

u
1 t
4 0.25 t
= c1 e + c2 e
v
0
5
Thus (1,0) is an unstable saddle point of both the linear and

nonlinear systems. One pair of trajectories approach (1,0)
along the x-axis, while all other trajectories depart from (1,0).
Example 1: Critical Point at (0,0.75)
(6 of 8)
For the critical point (0,0.75), the linear system is

0 u
d u 0.25
=

dt v 0.375 0.75 v

r1 = 0.25,
(1)
8
= ;
3
r2 = 0.75,
( 2)
0
=
1

u
8 0.25 t
0 0.75 t
= c1 e
+ c2 e
v
3
1
Thus (0,0.75) is an unstable saddle point of both the linear and

nonlinear systems. One pair of trajectories approach (0,0.75)
along y-axis, while all other trajectories depart from (0,0.75).
Example 1: Critical Point at (0.5,0.5)
(7 of 8)
For the critical point (0.5,0.5), the linear system is

d u 0.5 0.5 u
=

dt v 0.25 0.5 v

2
2
2+ 2
2 2
(1)
( 2)
0.146, = ; r2 =
0.854, =
r1 =
4
4
1
1

2 0.146 t
2 0.854 t
u
= c1 e
+ c2 e
v
1
1
Thus (0.5,0.5) is an asymptotically stable node of both the

linear and nonlinear systems.
(8 of 8)
A phase portrait is given below, along with the direction field,

for our nonlinear system
dx / dt = x(1 x y ), dy / dt = y (0.75 y 0.5 x)
Note that the quadratic terms in these equations are all
negative, and hence x' < 0 and y' < 0 for large x and y.
Thus the trajectories are directed inward towards (0.5,0.5).
(1 of 9)

dx / dt = x(1 x y ), dy / dt = y (0.5 0.25 y 0.75 x)
The critical points are (0,0), (1,0), (0,2), and (0.5,0.5). These
critical points correspond to equilibrium solutions.
The first three points involve the extinction of one or both
species. Only the fourth critical point corresponds to the long
term survival of both species.
Other solutions are represented as trajectories in the xy-plane
that describe the evolution of the populations over time.
Example 2: Direction Field
(2 of 9)
A direction field for our system of equations is given below,

where the heavy dots correspond to the critical points.
Based on the direction field, (0.5,0.5) appears to be a saddle
point, and hence unstable, while (1,0) and (0,2) appear to be
asymptotically stable.
Thus only one species will eventually
survive, and this species is determined
by the initial conditions.
(3 of 9)

0 x
d x 1
=

dt y 0 0.5 y

r1 = 1,
(1)
1
= ;
0
r2 = 0.5,
( 2)
0
=
1

x
1 t
0 0.5 t
= c1 e + c2 e
y
0
1
Thus the origin is an unstable node of both the linear and

nonlinear systems. All trajectories leave the origin tangent to
y-axis, except for one trajectory that lies along the x-axis.
(4 of 9)

1 u
d u 1
=

dt v 0 0.25 v
The general solution for this linear system is

u
1 t
4 0.25 t
= c1 e + c2 e
v
0
3
Thus (1,0) is an asymptotically stable node of both the linear

and nonlinear systems.
One pair of trajectories approach (1,0) along the x-axis
All other trajectories approach (1,0) tangent to the line with
slope 3/4, determined by the eigenvector (2).
(5 of 9)
For the critical point (0,2), the linear system is

0 u
d u 1
=

dt v 1.5 0.5 v
The general solution for this linear system is

u
1
0
= c1 e t + c2 e 0.5 t
v
3
1
Thus (0,2) is an asymptotically stable node of both the linear

and nonlinear systems.
One trajectory approaches (0,2) along the line with slope 3,
determined by the eigenvector (1).
All other trajectories approach (0,2) along the y-axis.
(6 of 9)
For the critical point (0.5,0.5), the linear system is

0.5 u
d u 0 .5
=

dt v 0.375 0.125 v

1
1

;

=
3 57 / 8 1.3187
1
1
5 57
( 2)

0.7844 , =
r2 =
16
3 + 57 / 8 0.5687
5 + 57
r1 =
0.1594 ,
16
(1)

1
u
0.1594 t
1 0.7844 t
= c1
e
e
+ c2
v
1.3187
0.5687
(7 of 9)
Thus for the critical point (0.5,0.5), we have

1
1 0.7844 t
0.1594 t
u
e
e
= c1
+ c2
0.5687
1.3187
v
It follows that (0.5,0.5) is an unstable saddle point of both the

linear and nonlinear systems.
One pair of trajectories approaches (0.5,0.5) as t , while
the others depart from (0.5,0.5).
The entering trajectories approach (0.5,0.5) tangent to the line
with slope of 0.5687, determined by the eigenvector (2).
(8 of 9)
A phase portrait is given below, along with the direction field.

Of particular interest is the pair of trajectories that enter the
saddle point. These trajectories form a separatrix that divides
the first quadrant into two basins of attraction.
(9 of 9)
Trajectories starting above the separatrix approach the node at

(0,2), while those below approach the node at (1,0).
If initial state lies on separatrix, then the solution will approach
the saddle point, but the slightest perturbation will send the
trajectory to one of the nodes instead.
Thus in practice, one species will
survive the competition and the
other species will not.
Coexistence Analysis
(1 of 7)
Examples 1 and 2 show that in some cases the competition

between two species leads to an equilibrium state of
coexistence, while in other cases the competition results in
eventual extinction of one of the species.
We can predict which situation will occur by examining the
governing equations
dx / dt = x( 1 1 x 1 y ), dy / dt = y ( 2 2 y 2 x)
Note that this system is almost linear, since F and G are
quadratic polynomials.
There are four cases to be considered, depending on the
relative orientation of the lines
1 1 x 1 y = 0, 2 2 y 2 x = 0
Coexistence Analysis: Nullclines
(2 of 7)
The graphs below show the relative orientation of the lines

1 1 x 1 y = 0, 2 2 y 2 x = 0
The lines are called the x and y nullclines, respectively,
because x' = 0 on the first and y' = 0 on the second.
The heavy dots represent equilibrium solutions.
Thus sustained coexistence is not possible in cases (a) and (b).
We show that sustained coexistence happens only in case (d).
Coexistence Analysis: Linear System
(3 of 7)
Let (X,Y) be a critical point, with corresponding linear system

d u 1 2 1 X 1Y
=
2Y
dt v
1 X
u

2 2 2Y 2 X v
Since (X,Y) is a critical point, we solve

1 1 X 1Y = 0, 2 2Y 2 X = 0
to obtain nonzero values of X and Y, as given below:
1 2 21
2 1 1 2
X=
, Y=
1 2 1 2
1 2 1 2
Further, with (X,Y) a critical point, the linear system reduces to
d u 1 X
=
dt v 2Y
1 X u

2Y v
Coexistence Analysis: Eigenvalues
(4 of 7)
The eigenvalues of the linear system are

( 1 X + 2Y ) ( 1 X + 2Y ) 2 4( 1 2 1 2 ) XY
r1, 2 =
2
If 12 12 < 0, then the eigenvalues are real and of

opposite sign, and hence (X,Y) is an unstable saddle point, and
thus sustained coexistence is not possible.
If 12 12 > 0, then the eigenvalues are real, negative and
unequal, or complex. It can be shown that the eigenvalues are
not complex, and hence (X,Y) is an asymptotically stable node,
and thus sustained coexistence is possible.
In Example 1, 12 12 = (1)(1) - (1)(0. 5) = 0.5 > 0.
In Example 2, 12 12 = (1)(0.25) - (1)(0.75) = -0.5 < 0.
Coexistence Analysis: Case (c)
(5 of 7)
In case (c), we have

1 2
2 1
>
or 1 2 > 2 1 ,
>
or 21 > 1 2
1 2
2 1
These inequalities, together with

X = 1 2 2 1 > 0, Y = 2 1 1 2 > 0,
1 2 1 2
1 2 1 2
yield 12 12 < 0.
Therefore sustained coexistence is not
possible in case (c).
Coexistence Analysis: Case (d)
(6 of 7)
In case (d), we have

1 2
2 1
<
or 1 2 < 2 1 ,
<
or 21 < 1 2
1 2
2 1
These inequalities, together with

1 2 21
2 1 1 2
X=
> 0, Y =
> 0,
1 2 1 2
1 2 1 2
yield 12 12 > 0.
We can also show that the other critical
points are unstable, and thus the two
populations approach the equilibrium
state of coexistences in case (d).
Discussion of Coexistence Conditions
(7 of 7)
For our competitive species equations,

dx / dt = x( 1 1 x 1 y ), dy / dt = y ( 2 2 y 2 x)
sustained coexistence is possible when 12 12 > 0.

The s are a measure of the inhibitory effect that the growth
of each population has on itself.
The s are a measure of the inhibitory effect that the growth
of each population has on the other species.
Thus when 12 12 > 0, the competition is weak, and the
species can coexist.
When 12 12 < 0, the competition is strong, and the
species cannot coexist one must die out.
Ch 9.5: Predator-Prey Systems

In Section 9.4 we discussed a model of two species that
interact by competing for a common food supply or other
natural resource.
In this section we investigate the situation in which one
species (the predator) preys on the other species (the prey),
while the prey lives on some other source of food.
For example, foxes and rabbits in a closed forest.
Again we emphasize that a model involving only two species
cannot fully describe the complex relationships among species
that occur in nature.
Nevertheless, the study of simple models is the first step
toward an understanding of more complicated phenomena.
Assumptions
Let x and y be the populations of the prey and predator,
respectively, at time t.
We make the following assumptions:
In the absence of the predator, the prey grows at a rate proportional to
the current population; thus dx/dt = a x, a > 0, when y = 0.
In the absence of the prey, the predator dies out at a rate proportional to
the current population; thus dy/dt = -cy, c > 0, when x = 0.
The number of encounters between predator and prey is proportional to
the product of their populations. Each such encounter tends to promote
the growth of the predator and to inhibit the growth of the prey. Thus
the growth rate of the predator is increased by a term of the form xy,
while the growth rate of the prey is decreased by a term xy, where
and are positive constants.
Predator-Prey Equations
Thus we have the system of equations
dx / dt = ax xy = x(a y ),
dy / dt = cy + xy = y ( c + x )
The constants a, c, , are all positive, where a, c are the
growth rate of prey and death rate of predator, respectively,
and , are measures of the effect of the interaction between
the two species.
The predator-prey equations are known as the Lotka-Volterra
equations. Although they are rather simple equations, they do
characterize a wide class of problems.
Our goal here is to determine the qualitative behavior of the
solutions for arbitrary positive initial values x and y.
(1 of 8)

dx / dt = x 0.5 xy, dy / dt = 0.75 y + 0.25 xy,
x, y > 0
The critical points are (0,0) and (3,2).

Given below is a direction field for this system of equations,
with the critical points indicated as heavy dots.
The trajectories appear to be closed curves surrounding the
critical point (3,2).
(2 of 8)

0 x
d x 1
=

dt y 0 0.75 y

r1 = 1,
(1)
1
= ;
0
r2 = 0.75,
( 2)
0
=
1

x
1 t
0 0.75 t
= c1 e + c2 e
y
0
1

nonlinear systems. One trajectory approaches (0,0) along the
y-axis, while all other trajectories depart from (0,0).
(3 of 8)

d u 0 1.5 u
=
,
0 v
dt v 0.5
u x 3
=
v y 2

r1 =
3i (1) 1
;
, =
2
i / 3
r2 =
3i ( 2) 1
, =
2
i / 3
Thus (3,2) is stable center point of the linear system, but is

indeterminate for the nonlinear systems, by Theorem 9.3.2.
To find the trajectories for the linear system, we have
dv / dt dv
0.5u
u
=
=
=
du / dt du
1.5v
3v
(4 of 8)
Near the critical point (3,2), we thus have

dv / du = u / 3v
The solution to this separable equation is
u 2 + 3v 2 = k , k > 0
Thus the trajectories of the linear system are ellipses centered
at the critical point (3,2), and are elongated horizontally.
Returning to the nonlinear system
dx / dt = x 0.5 xy, dy / dt = 0.75 y + 0.25 xy, x, y > 0
we have
dy 0.75 y + 0.25 xy
=
dx
x 0.5 xy
(5 of 8)
Thus
0.75 + 0.25 x
dy y ( 0.75 + 0.25 x )
1 0.5 y
=
dy =
dx
dx
x(1 0.5 y )
y
x
0.75 ln x + ln y 0.5 y 0.25 x = c
It can be shown that the graph of this equation, for a fixed c, is
a closed curve about the point (3,2).
Thus the critical point (3,2) is also a center of our nonlinear
system, and hence the predator and prey populations exhibit a
cyclic variation about the equilibrium solution (3,2).
This behavior is seen in the phase portrait on the next slide.
(6 of 8)
Given below is a phase portrait for our nonlinear system.

For some initial conditions, the trajectories represent small
variations in x and y about (3,2), and are almost elliptical in
shape, as the linear analysis suggests.
For other initial points, the oscillations in x and y are more
pronounced, and the shape of the trajectories are significantly
different from an ellipse.
Note that the trajectories are
traversed counterclockwise.
(7 of 8)
A phase portrait along with population graphs x(t) and y(t), for
a typical set of initial conditions, are given below.
Note from both of these figures that the oscillation of the
predator population lags behind that of the prey.
(8 of 8)
Starting with a state in which both populations are relatively

small, the prey first increase because of little predation.
Then the predators, with abundant food, increase in population.
This causes heavy predation, and the prey tend to decrease.
Finally, with a diminished food supply, the predator population
also decreases, and the system returns to original state.
General Predator-Prey Equations
(1 of 7)
The general system of equations

dx / dt = ax xy = x(a y ),
dy / dt = cy + xy = y ( c + x )
can be analyzed in the same way as in Example 1.
The critical points are the solutions of the equations
x(a y ) = 0, y ( c + x ) = 0,
yielding the points (0,0) and (c/, a/).
We next examine the solutions of the corresponding linear
system near each critical point.
General System: Critical Point at (0,0)
(2 of 7)

0 x
d x a

=
dt y 0 c y

r1 = a,
(1)
1
= ;
0
r2 = c,
( 2)
0
=
1

0 c t
1 at
x
= c1 e + c2 e
1
0
y

nonlinear systems. One trajectory approaches (0,0) along the
y-axis, while all other trajectories depart from (0,0).
Critical Point at (c/, a/): Linear System
(3 of 7)
For the critical point (c/, a/), the linear system is

d u 0
=
dt v a /
c / u
,
0 v
u x c /
=
v y a /
The eigenvalues are

r1 = i ac , r2 = i ac
Thus (c/, a/) is stable center point of the linear system, but is
indeterminate for the nonlinear systems, by Theorem 9.3.2.
To find the trajectories for the linear system, we have
dv / dt dv
( a / )u
2a u
=
=
= 2
du / dt du
( c / )v
cv
Critical Point at (c/, a/): Ellipses
(4 of 7)
Near the critical point (c/, a/), we thus have

dv
2a u
= 2
( 2 a u )du + ( 2 c v )dv = 0
du
cv
2 a u 2 + 2c v 2 = k , k > 0
Thus the trajectories of the linear system are ellipses centered
at the critical point (c/, a/).
Returning to the nonlinear system
dx / dt = x(a y ), dy / dt = y ( c + x )
we have
dy y ( c + x )
=
dx
x(a y )
Critical Point at (c/, a/):

Nonlinear System (5 of 7)
Thus
c + x
dy y ( c + x )
a y
=
dy =
dx
dx
x(a y )
y
x
a ln y y + c ln x x = C
It can be shown that the graph of this equation, for a fixed C, is
a closed curve about the point (c/, a/).
Thus the critical point (c/, a/) is also a center of our
nonlinear system, and hence the predator and prey populations
exhibit a cyclic variation about (c/, a/).
Critical Point at (c/, a/):

Small Deviations and Linear System (6 of 7)
The cyclic variation of the predator and prey populations can
be analyzed in more detail when the deviations from the point
(c/, a/) are small and the linear system can be used.
The solution of the linear system
d u 0
=
dt v a /
c / u

0 v
can be written in the form

u (t ) =
K cos act + , v(t ) =
c
K sin act +
a
where the constants K and are determined by the initial

conditions.
Critical Point at (c/, a/): Small Deviations and

Elliptical Approximation (7 of 7)
The equations below are good approximations to the nearly
elliptical trajectories near (c/, a/).
u (t ) =
K cos act + , v(t ) =
c
K sin act +
a
We can use them to draw several conclusions about the cyclic

variation of the predator and prey on such trajectories:
The predator and prey population sizes vary sinusoidally with period
2 /(ac). This oscillation period is independent of initial conditions.
The predator and prey populations are out of phase by one-quarter of a
cycle. The prey leads and the predator lags.
The amplitudes of oscillations are Kc/ for the prey, and (c/a)Ka/
for the predator, and hence depend on initial conditions and parameters.
The average predator and prey populations over one complete cycle are
c/ and /a, respectively same as equilibrium populations.
Modified Predator-Prey Equations

Cyclic variations of predator and prey, as predicted by our
equations, have been observed in nature; see text.
One criticism of the Lotka-Volterra equations is that in the
absence of the predator, the prey will grow without bound.
This can be corrected using a logistic model for x when y = 0.
As a result of this modification, the critical point at (c/, a/)
moves to (c/, a/ - c/ ) and becomes an asymptotically
stable point. It is either a node or a spiral point, depending on
the parameters in the differential equations.
In either case, other trajectories are no longer closed curves but
approach the critical points as t .
Ch 9.6: Liapunovs Second Method

In Section 9.3 we showed how the stability of a critical point
of an almost linear system can usually be determined from a
study of the corresponding linear system.
However, no conclusion can be drawn when the critical point
is a center of the corresponding linear system.
Also, for an asymptotically stable critical point, we may want
to investigate the basin of attraction, for which the localized
almost linear theory provides no information.
In this section we discuss Liapunovs second method, or direct
method, in which no knowledge of the solution is required.
Instead, conclusions about the stability of a critical point are
obtained by constructing a suitable auxiliary function.
Physical Principles
Liapunovs second method is a generalization of two physical
principles for conservative systems.
The first principle is that a rest position is stable if the
potential energy is a local minimum, otherwise it is unstable.
The second principle states that the total energy is a constant
during any motion.
To illustrate these concepts, we again consider the undamped
pendulum, which is a conservative system.
Undamped Pendulum Equation
(1 of 5)
The governing equation for the undamped pendulum is

d 2 g
+ sin = 0
2
dt
To convert this equation into a system of two first order

equations, we let x = and y = d /dt, obtaining
dy
g
dx
= y,
= sin x
dt
L
dt
The potential energy U is the work done in

lifting pendulum above its lowest position:
U ( x, y ) = mgL(1 cos x)
Undamped Pendulum System:

Potential Energy (2 of 5)
The critical points of our system
dx
dy
g
= y,
= sin x
dt
dt
L
are x= n , y = 0, for n = 0, 1, 2,.

Physically, we expect the points (2n , 0) to be stable, since
the pendulum bob is vertical with the weight down, and the
points ((2n+1) , 0) to be unstable, since the pendulum bob
is vertical with the weight up.
Comparing this with the potential energy U,
U ( x, y ) = mgL(1 cos x),
we see that U is a minimum equal to zero at (2n , 0) and U

is a maximum equal to 2mgL at ((2n+1) , 0).

Total Energy (3 of 5)
The total energy V is the sum of potential and kinetic energy:
V ( x, y ) = mgL(1 cos x) + (1 / 2)mL2 y 2
On a solution trajectory x = (t), y = (t), V is a function of t.

The derivative of V((t), (t)) with respect to t is called the
rate of change of V following the trajectory.
For x = (t), y = (t), and using the chain rule, we obtain
dy
dx
d
d
dV ( , )
= Vx ( , )
+ V y ( , )
= mgL sin x + mL2 y
dt
dt
dt
dt
dt
Since x and y satisfy the differential equations

dx / dt = y, dy / dt = g sin x / L,
it follows that dV(, )/dt = 0, and hence V is constant.

Small Energy Trajectories (4 of 5)
Observe that we computed the rate of change dV(, )/dt of
the total energy V without solving the system of equations.
It is this fact that enables us to use Liapunovs second method
for systems whose solution we do not know.
Note that V = 0 at the stable critical points (2n , 0), where
we recall
2 2
V ( x, y ) = mgL(1 cos x) + (1 / 2)mL y
If the initial state (x1, y1) of the pendulum is sufficiently near a

stable critical point, then the energy V(x1, y1) is small, and the
corresponding trajectory will stay close to the critical point.
It can be shown that if V(x1, y1) is sufficiently small, then the
trajectory is closed and contains the critical point.

Small Energy Elliptical Trajectories
(5 of 5)
Suppose (x1, y1) is near (0,0), and that V(x1, y1) is very small.
The energy equation of the corresponding trajectory is
V ( x1 , y1 ) = mgL(1 cos x) + (1 / 2)mL2 y 2
From the Taylor series expansion of cos x about x = 0, we have

1 cos x = 1 (1 x 2 / 2! x 4 / 4! K) x 2 / 2
Thus the equation of the trajectory is approximately

x2
y2
+
=1
2
2V ( x1 , y1 ) / mgL 2V ( x1 , y1 ) / mL
This is an ellipse enclosing the origin. The smaller V(x1, y1) is,
the smaller the axes of the ellipse are.
Physically, this trajectory corresponds to a periodic solution,
whose motion is a small oscillation about equilibrium point.
Damped Pendulum System:

Total Energy (1 of 2)
If damping is present, we may expect that the amplitude of
motion decays in time and that the stable critical point (center)
becomes an asymptotically stable critical point (spiral point).
Recall from Section 9.3 that the system of equations is
dx / dt = y, dy / dt = ( g / L) sin x (c / Lm) y
The total energy is still given by
V ( x, y ) = mgL(1 cos x) + (1 / 2)mL2 y 2
Recalling
dV ( , )
dx
2 dy
= mgL sin x + mL y ,
dt
dt
dt
it follows that dV/dt = -cLy2 0.
Damped Pendulum System:

Nonincreasing Total Energy
(2 of 2)
Thus dV/dt = -cLy2 0, and hence the energy is nonincreasing

along any trajectory, and except for the line y = 0, the motion
is such that the energy decreases.
Therefore each trajectory must approach a point of minimum
energy, or a stable equilibrium point.
If dV/dt < 0 instead of dV/dt 0, we can expect this to hold for
all trajectories that start sufficiently close to the origin.
General Autonomous System:

Total Energy
To pursue these ideas further, consider the autonomous system
dx / dt = F ( x, y ), dy / dt = G ( x, y ),
and suppose (0,0) is an asymptotically stable critical point.
Then there exists a domain D containing (0,0) such that every
trajectory that starts in D must approach (0,0) as t .
Suppose there is an energy function V such that V(x, y) 0
for (x, y) in D with V = 0 only at (0,0).
Since each trajectory in D approaches (0,0) as t , then
following any particular trajectory, V approaches 0 as t .
The result we want is the converse: If V decreases to zero on
every trajectory as t , then the trajectories approach (0,0)
as t , and hence (0,0) is asymptotically stable.
Definitions: Definiteness
Let V be defined on a domain D containing the origin. Then
we make the following definitions.
V is positive definite on D if V(0,0) = 0 and V(x, y) > 0 for all
other points (x, y) in D.
V is negative definite on D if V(0,0) = 0 and V(x, y) < 0 for all
other points (x, y) in D.
V is positive semi-definite on D if V(0,0) = 0 and V(x, y) 0
for all other points (x, y) in D.
V is negative semi-definite on D if V(0,0) = 0 and V(x, y) 0
for all other points (x, y) in D.
Example 1
Consider the function
V ( x, y ) = sin (x 2 + y 2 )
Then V is positive definite on the domain

D = {( x , y ) : x 2 + y 2 < / 2 }
since V(0,0) = 0 and V(x, y) > 0 for all other points (x, y) in D.
Example 2
Consider the function
V ( x, y ) = ( x + y )
Then V is only positive semi-definite on the domain

D = {( x , y ) : x 2 + y 2 < / 2 }
since V(x, y) = 0 on the line y = -x.
Derivative of V With Respect to System

We also want to consider the function
V& = Vx ( x, y ) F ( x, y ) + V y ( x, y )G ( x, y ),
where F and G are the functions given in the system
dx / dt = F ( x, y ), dy / dt = G ( x, y ),
The function V& can be identified as the rate of change of V
along the trajectory that passes through (x, y), and is often
referred to as the derivative of V with respect to the system.
That is, if x = (t), y = (t) is a solution of our system, then
dV ( , )
d
d
= Vx ( , )
+ V y ( , )
dt
dt
dt
= Vx ( x, y ) F ( x, y ) + V y ( x, y )G ( x, y )
= V&
Theorem 9.6.1
Suppose that the origin is an isolated critical point of the
autonomous system
dx / dt = F ( x, y ), dy / dt = G ( x, y ),
If there is a function V that is continuous and has continuous
first partial derivatives, is positive definite, and for which
V& = Vx ( x, y ) F ( x, y ) + V y ( x, y )G ( x, y )
is negative definite on a domain D in the xy-plane containing
(0,0), then the origin is an asymptotically stable critical point.
If V& negative semidefinite, then (0,0) is a stable critical point.
See the text for an outline of the proof for this theorem.
Theorem 9.6.2
autonomous system
dx / dt = F ( x, y ), dy / dt = G ( x, y )
Let V be a function that is continuous and has continuous first
partial derivatives.
Suppose V(0,0) = 0 and that in every neighborhood of (0,0)
there is at least one point for which V is positive (negative).
If there is a domain D containing the origin such that
is positive definite (negative definite) on D, then the origin is
an unstable critical point.
See the text for an outline of the proof for this theorem.
Liapunov Function
The function V in Theorems 9.6.1 and 9.6.2 is called a
Liapunov function.
The difficulty in using these theorems is that they tell us
nothing about how to construct a Liapunov function, assuming
that one exists.
In the case where the autonomous system represents a physical
problem, it is natural to consider first the actual total energy of
the system as a possible Liapunov function.
However, Theorems 9.6.1 and 9.6.2 are applicable in cases
where the concept of physical energy is not pertinent.
In these cases, a trial-and-error approach may be necessary.
Example 3: Undamped Pendulum
(1 of 3)
For the undamped pendulum system

du / dt = v, dv / dt = ( g / L )sin u ,
use Theorem 9.6.1 show that (0,0) is a stable critical point, and
use Theorem 9.6.2 to show (, 0) is an unstable critical point.
Let V be the total energy function
V ( x, y ) = mgL(1 cos x) + (1 / 2)mL2 y 2
and let
D = { ( x, y ) : / 2 < x < / 2, < y < }
Thus V is positive definite on D, since V > 0 on D, except at

the origin, where V(0,0) = 0.
(2 of 3)
Thus V is positive definite on D,

D = { ( x, y ) : / 2 < x < / 2, < y < }
Further, as we have seen,
V& = (mgL sin x ) y + mL2 y x = 0

for all x and y. Thus V& is negative semidefinite on D.
Thus by Theorem 9.6.1, it follows that the origin is a stable

critical point for the undamped pendulum.
To examine the critical point (, 0) using Theorem 9.6.2, we
cannot use the same Liapunov function
V ( x, y ) = mgL(1 cos x) + (1 / 2)mL2 y 2 ,
since V& is not positive or negative definite.
Example 3: Critical Point at (, 0)
(3 of 3)
Consider the change of variable x = + u, and y = v. Then

our system of differential equations becomes
du / dt = v, dv / dt = ( g / L )sin u ,
with critical point (0, 0) in the uv-plane. Let V be defined by

V (u , v) = v sin u
and let D be the domain
D = { ( x, y ) : / 4 < u < / 4, < v < }
Then V(u, v) > 0 in the first and third quadrants, and
V& = (v cos u ) v + (sin u )[( g / L )sin u ] = v 2 cos u + ( g / L )sin 2 u
is positive definite on D.
Thus (0, 0) in the uv-plane, or (, 0) in xy-plane, is unstable.
Theorem 9.6.3
autonomous system
dx / dt = F ( x, y ), dy / dt = G ( x, y )
Let V be a function that is continuous and has continuous first
partial derivatives.
If there exists a bounded domain DK containing the origin on
which V(x, y) < K, with V is positive definite and
negative definite, then every solution of the system above that
starts at a point in DK approaches the origin as t .
Thus DK gives a region of asymptotic stability, but may not be
the entire basin of attraction.
Liapunov Function Discussion

Theorems 9.6.1 and 9.6.2 give sufficient conditions for
stability and instability, respectively.
However, these conditions are not necessary, nor does our
failure to determine a suitable Liapunov function mean that
there is not one.
Unfortunately, there are no general methods for the
construction of Liapunov functions.
However, there has been extensive work on the construction of
Liapunov functions for special classes of equations.
An algebraic result that is often useful in constructing positive
or negative definite functions is stated in the next theorem.
Theorem 9.6.4
Let V be the function defined by
V = ax 2 + bxy + cy 2
Then V is positive definite if and only if
a > 0 and 4ac b 2 > 0,
and is negative definite if and only if
a < 0 and 4ac b 2 > 0.
Example 4
Consider the system
dx / dt = x xy 2 , dy / dt = y x 2 y
We try to construct a Liapunov function of the form

V = ax 2 + bxy + cy 2
Then
V& = (2ax + by )( x xy 2 ) + (bx + 2cy )( y x 2 y )
= 2a ( x 2 + x 2 y 2 ) + b(2 xy + xy 3 + x 3 y ) + 2c( y 2 + x 2 y 2 )
If we choose b = 0, and a, c to be any positive numbers, then

V& = 2a ( x 2 + x 2 y 2 ) + 2c( y 2 + x 2 y 2 )
is negative definite, and V positive definite by Theorem 9.6.4.

Hence (0,0) is asymptotically stable, by Theorem 9.6.1.
Example 5: Competing Species System
(1 of 7)
Consider the system

dx / dt = x(1 x y ), dy / dt = y (0.75 y 0.5 x)
In Example 1 of Section 9.4 we found that this system models
a certain pair of competing species, and that the point (0.5,0.5)
is asymptotically stable. We confirm this conclusion by
finding a suitable Liapunov function.
We transform (0.5,0.5) to the origin by letting x = 0.5 + u, and
y = 0.5 + v. Our system then becomes
du / dt = 0.5u 0.5v u 2 uv,
dv / dt = 0.25u 0.5v 0.5uv v 2
Example 5: Liapunov Function
(2 of 7)
We have
du / dt = 0.5u 0.5v u 2 uv,
dv / dt = 0.25u 0.5v 0.5uv v 2
Consider a possible Liapunov function of the form

V (u , v) = u 2 + v 2
Then V is positive definite, so we only need to determine
whether there is region of the origin in the uv-plane where
V& = 2u ( 0.5u 0.5v u 2 uv ) + 2v( 0.25u 0.5v 0.5uv v 2 )
[(
) (
= u 2 + 1.5uv + v 2 + 2u 3 + 2u 2 v + uv 2 + 2v 3
is negative definite.
)]
Example 5:
Derivative With Respect to System
To show that
[(
) (
V& = u 2 + 1.5uv + v 2 + 2u 3 + 2u 2 v + uv 2 + 2v 3
(3 of 7)
)]
is negative definite, it suffices to show that
) (
H (u, v) = u 2 + 1.5uv + v 2 + 2u 3 + 2u 2 v + uv 2 + 2v 3
is positive definite, at least for u and v sufficiently small.

Observe that the quadratic terms of H can be written as
u 2 + 1.5uv + v 2 = 0.25 u 2 + v 2 + 0.75(u + v ) ,

2
and hence are positive definite.

The cubic terms may be of either sign. We show that in some
neighborhood of the origin, the following inequality holds:
2u 3 + 2u 2 v + uv 2 + 2v 3 < 0.25 u 2 + v 2 + 0.75(u + v )
Example 5: Negative Definite
(4 of 7)
To estimate the left side of the desired inequality
2u 3 + 2u 2 v + uv 2 + 2v 3 < 0.25 u 2 + v 2 + 0.75(u + v )
we introduce polar coordinates u = rcos and v = rsin.

Then
2u 3 + 2u 2 v + uv 2 + 2v 3
= r 3 2 cos 3 + 2 cos 2 sin + cos sin 2 + 2 sin 3
r 3 2 cos 3 + 2 cos 2 sin + cos sin 2 + 2 sin 3
7r 3 ,
since |cos |, |sin | 1. It is then sufficient to require

7 r 3 < 0.25(u 2 + v 2 ) = 0.25r 2 r < 1 / 28 0.0357
Example 5:
Asymptotically Stable Critical Point
(5 of 7)
Thus for the domain D defined by
{(u, v) : u
+ v 2 < 1 / 28 },
the hypotheses of Theorem 9.6.1 are satisfied, so the origin is

an asymptotically stable critical point of the system
du / dt = 0.5u 0.5v u 2 uv,
dv / dt = 0.25u 0.5v 0.5uv v 2
It follows that the point (0.5,0.5) is an asymptotically stable
critical point of the original system of equations
dx / dt = x(1 x y ),
dy / dt = y (0.75 y 0.5 x)
Example 5:
Region of Asymptotic Stability
(6 of 7)
Recall that the Liapunov function for this example is

V ( x, y ) = ( x 0.5) 2 + ( y 0.5) 2
If we refer to Theorem 9.6.3, then the preceding argument also
shows that the disk
D1/ 28 = ( x, y ) : ( x 0.5) 2 + ( y 0.5) 2 < 1 / 28
is a region of asymptotic stability for the system of equations

dx / dt = x(1 x y ), dy / dt = y (0.75 y 0.5 x)
Example 5:
Estimating Basin of Attraction
The disk
(7 of 7)
D1/ 28 = ( x, y ) : ( x 0.5) 2 + ( y 0.5) 2 < 1 / 28
is a severe underestimate of the full basin of attraction.

To obtain a better estimate of the actual basin of attraction
from Theorem 9.6.3, we would need to estimate the terms in
2u 3 + 2u 2 v + uv 2 + 2v 3 < 0.25 u 2 + v 2 + 0.75(u + v )
more accurately, use a better (and possibly more complicated)

Liapunov function, or both.
Ch 9.7: Periodic Solutions and Limit Cycles

In this section we discuss further the possible existence of
periodic solutions of second order autonomous systems
x' = f(x)
Such solutions satisfy the relation x(t + T) = x(t) for all t and
for some nonnegative constant T called the period.
Periodic solutions are often important in physical problems
because they represent phenomena that occur repeatedly.
In many situations a periodic solution represents a final state
toward which all neighboring solutions tend as the transients
due to the initial conditions die out.
Nonconstant Periodic Solutions

Thus a periodic solution satisfies x(t + T) = x(t) for all t and for
some nonnegative constant T.
Note that a constant solution x = x0 is periodic for any T.
In this section, the periodic solutions that are discussed refer to
nonconstant periodic solutions.
In this case the period T is positive and is usually chosen as the
smallest positive number for which x(t + T) = x(t) is valid.
Linear Autonomous Systems

Recall: The solutions of a linear autonomous system x' = Ax
are periodic if and only if the eigenvalues are pure imaginary.
Thus if the eigenvalues of A are pure imaginary, then every
solution of x' = Ax is periodic, while if the eigenvalues of A
are not pure imaginary, then there are no periodic solutions.
The predator-prey equations discussed in Section 9.5, although
nonlinear, behave similarly: All solutions in the first quadrant
are periodic. See graph below.
Example 1: Nonlinear System
(1 of 8)
Consider the nonlinear autonomous system

x x + y x(x 2 + y 2 )
=
2
2
y x + y y (x + y )
It can be shown that (0, 0) is the only critical point and that
this system is almost linear near the origin.
The corresponding linear system
x 1 1 x

=
y 1 1 y
has eigenvalues 1 i, and hence the
origin is an unstable spiral point.
Example 1: Unstable Spiral Point
(2 of 8)
Thus the origin is an unstable spiral point, and hence any

solution that starts near the origin in the phase plane will spiral
away from the origin.
Since there are no other critical points, we might think that all
solutions of our nonlinear system correspond to trajectories
that spiral out to infinity.
However, we will show that this is incorrect, because far away
from the origin the trajectories are directed inward.
Example 1: Polar Coordinates

Our nonlinear system can be written as
(3 of 8)
dx / dt = x + y x x 2 + y 2 , dy / dt = x + y y x 2 + y 2
Then
x
dy
dx
+y
= x 2 + xy x 2 x 2 + y 2 xy + y 2 y 2 x 2 + y 2
dt
dt
) (
= x +y x +y
2
2 2
Using polar coordinates x = rcos and y = rsin, note that

dr
dx
dy
r = x + y , and r
=x +y
dt
dt
dt
2
Thus
dr
r
= r 2 (1 r 2 )
dt
Example 1:
Critical Points for Equation of Radius
(4 of 8)
The critical points (for r 0) of

r (dr / dt ) = r 2 (1 r 2 )
are r = 0 (the origin) and r = 1, which corresponds to the unit
circle in the phase plane.
Note that dr/dt > 0 if r < 1 and dr/dt < 0 if r > 1. Thus inside
the unit circle, the trajectories are directed outward, while
outside the unit circle they are directed inward.
The circle r = 1 appears to be a limiting trajectory for system.
We next determine an equation for .
Example 1: Equation for Angle
(5 of 8)
Recall our nonlinear system:
Then
) (
dx
dy
y x
= xy + y 2 xy x 2 + y 2 + x 2 xy + xy x 2 + y 2 = x 2 + y 2
dt
dt
Using polar coordinates x = rcos and y = rsin, note that

dx
d
d dy
d
d
r =x +y ,
= r sin
= y
,
= r cos
=x
dt
dt
dt
dt
dt
dt
2
It follows that
2 d
2 d
y
dt
dt
= r2 r2
d
= r2
dt
d
= 1
dt
Example 1:
A Solution to Polar Equations
Our original nonlinear system
(6 of 8)
is therefore equivalent to the system
r (dr / dt ) = r 2 (1 r 2 ), d / dt = 1
One solution to this system is
r = 1, = t + t0
where t0 is an arbitrary constant.
As t increases, a point on this solution
trajectory moves clockwise around the
unit circle.
Example 1:
General Solution to Polar Equations
(6 of 8)
Other solutions of
dr
r
= r 2 (1 r 2 )
dt
can be found by separation of variables: For r 0 and r 1,

dr
= dt ,
2
r (1 r )
and after using a partial fraction expansion and some algebra,

r=
1
1 + c0 e
2t
, = t + t 0
where c0 and t0 are arbitrary constants.

Note that c0 = 0 yields r = 1, = -t + t0, as before.
Example 1:
Initial Value Problem in Polar Form
The solution satisfying the initial value problem
dr
d
2
2
= r (1 r ),
= 1; r (0) = , (0) =
r
dt
dt
is given by
r=
1
1 + [(1 / ) 1] e
2
2t
, = (t )
We have the following two cases:

If < 1, then r 1 from the inside as t .
If > 1, then r 1 from the outside as t .
See phase portrait on right.
(8 of 8)
Limit Cycle
In the previous example, the circle r = 1 not only corresponds
to periodic solutions of the system
but it also attracts other nonclosed trajectories that spiral

toward it as t .
In general, a closed trajectory in the phase plane such that
other nonclosed trajectories spiral toward it, either from the
inside or the outside, as t , is called a limit cycle.
Stability of Closed Trajectories

If all trajectories that start near a closed trajectory spiral
toward the closed trajectory as t , both from the inside and
the outside, then the limit cycle is asymptotically stable.
In this case, since the closed trajectory is itself a periodic orbit
rather than an equilibrium point, this type of stability is often
called orbital stability.
If the trajectories on one side spiral toward a closed trajectory ,
while those on the other side spiral away as t , then the
closed trajectory is semistable.
If the trajectories on both sides of a closed trajectory spiral
away as t , then the closed trajectory is unstable.
Closed trajectories for which other trajectories neither
approach nor depart from are called stable.
Theorem 9.7.1
dx / dt = F ( x, y ), dy / dt = G ( x, y ),
Let F and G have continuous first partial derivatives in a
domain D in the xy-plane.
A closed trajectory of the system must necessarily enclose at
least one critical (equilibrium) point.
If encloses only one critical point, the critical point cannot be a
saddle point.
Note: It follows that in any region not containing a critical

point, there cannot be a closed trajectory within that region.
Theorem 9.7.2
dx / dt = F ( x, y ), dy / dt = G ( x, y ),
simply connected domain D in the xy-plane.
If Fx + Gy has the same sign throughout D, then there is no
closed trajectory of the system lying entirely within D.
Note: A simply connected domain in the xy-plane is a domain

with no holes.
Also, If Fx + Gy changes sign in D, then no conclusion can be
drawn.
Example 2: Applying Theorem 9.7.2
(1 of 2)
Consider again the nonlinear autonomous system
Then
Fx + G y = 2 4(x 2 + y 2 ) = 2(1 2r 2 )
Thus Fx + Gy > 0 on 0 r < (1/2), so there is no closed
trajectory in this simply connected circular disk.
From Example 1, there is no closed trajectory in r < 1.
Thus the information given in
Theorem 9.7.2 may not be the
best possible result.
Example 2:
Annular Region and Theorem 9.7.2
(2 of 2)
Note that
Fx + G y = 2(1 2r 2 ) < 0 on r > 1 / 2
However, Theorem 9.7.2 does not apply since the annular
region r > (1/2) is not simply connected.
Thus we cannot use Theorem 9.7.2 to conclude that there is no
closed trajectory lying entirely within r > (1/2).
In fact, from Example 1, we know that r = 1 is a closed
trajectory for the system that lies entirely within r > (1/2).
Theorem 9.7.3 (Poincar-Bendixson)

dx / dt = F ( x, y ), dy / dt = G ( x, y ),
domain D in the xy-plane.
Let D1 be a bounded subdomain in D, and let R be the region
that consists of D1 plus its boundary (all points of R are in D).
Suppose that R contains no critical point of the system.
If there exists a constant t0 such that x = (t), y = (t) is a
solution of the system that exists and stays in R for all t > t0,
then x = (t), y = (t) either is a periodic solution (closed
trajectory) or spirals toward a closed trajectory as t .
In either case, the system has a periodic solution in R.
Example 3: Applying Theorem 9.7.3

Consider again the nonlinear autonomous system
Since the origin is a critical point, it must be excluded from R.

Consider the region R defined by 0.5 r 2.
Recall from Example 1 that dr/dt = r(1 - r) for 0.5 r 2.
For r = 0.5, dr/dt > 0 and hence r increases, while for r = 2,
dr/dt < 0 and hence r decreases.
Thus a trajectory that crosses the boundary of R is entering R.
Consequently, any solution that starts in R
at t = t0 cannot leave but must stay in R for
all t > t0, and is either a periodic solution or
approaches one as t .
Example 4: Van der Pol Equation
(1 of 13)
The van der Pol equation describes the current u in a triode

oscillator:
u (1 u 2 )u + u = 0, > 0
If = 0, then the equation reduces to u'' + u = 0, whose
solutions are sine or cosine waves of period 2.
If > 0, then -(1 u2) is the resistance coefficient.
For large , the resistance term is positive and acts to reduce
the amplitude of the response.
For small , the resistance term is negative and causes the
response to grow.
This suggests that perhaps there is a solution of intermediate
size that other solutions approach as t increases.
Example 4: Unstable Critical Point
(2 of 13)
Let x = u and y = u'. Then the van der Pol equation

u (1 u 2 )u + u = 0, > 0
becomes
x = y, y = x + (1 x 2 )y
The only critical point is the origin. This system is almost
linear, with linear approximation
x 0 1 x
=
,
y 1 y
whose eigenvalues are [ (2 4)]/2.

Thus the origin is an unstable spiral point for 0 < < 2, and an
unstable node for 2. In all cases, a solution that starts near
the origin grows as t increases.
Example 4: Theorems 9.7.1 and 9.7.2
(3 of 13)
With regard to periodic solutions, Theorems 9.7.1 and 9.7.2

provide only partial information.
From Theorem 9.7.1 we conclude that if there are closed
trajectories, then they must enclose the origin.
To apply Theorem 9.7.2, we first calculate
Fx ( x, y ) + G y ( x, y ) = (1 x 2 )
It follows that closed trajectories, if there are any, are not
contained in the strip |x| < 1, where Fx + Gy > 0.
To apply Theorem 9.7.3, we introduce polar coordinates to
obtain the following equation for r:
r = (1 r 2 cos 2 )r sin 2
Example 4: Theorems 9.7.1 and 9.7.2
(4 of 13)
We have the following equation for r:

r = (1 r 2 cos 2 )r sin 2
Consider the annular region R given by r1 r r2, where r1 is
small and r2 is large.
When r = r1, the linear term in the equation for r' dominates,
and r' > 0 except on the x-axis, where sin = 0, hence r' = 0.
Thus the trajectories are entering R at every point on the circle
r = r1, except possibly those on the x-axis, where trajectories
are tangent to the circle.
Example 4: Theorem 9.7.3
(5 of 13)
We have the following equation for r:

r = (1 r 2 cos 2 )r sin 2
and R given by r1 r r2, where r1 is small and r2 is large.
When r = r2, the cubic term in the equation for r' dominates,
and r' < 0 except on the x-axis, where r' = 0, and for points
near the y-axis where r2cos2 < 1, and hence r' > 0.
Thus no matter how large a circle is chosen, there will be
points on it (namely, the points on or near the y-axis) where
trajectories are leaving R.
Therefore Theorem 9.7.3 is not applicable unless we consider
more complicated regions.
It is possible to show that this system does have a unique limit
cycle, but we will not pursue that here.
Example 4: Numerical Solutions
(6 of 13)
We next plot numerically computed solutions.

Experimental observations show that the van der Pol equation
has an asymptotically stable periodic solution whose period
and amplitude depend on the parameter .
Graphs of trajectories in the phase plane and of u versus t can
provide some understanding of periodic behavior.
Example 4: Phase Portrait ( = 0.2)
(7 of 13)
The graph below shows two trajectories when = 0.2.

The trajectory starting near the origin spirals outward in the
clockwise direction. This is consistent with the behavior of the
linear approximation near the origin.
The other trajectory passes through (-3, 2) and spirals inward,
again in the clockwise direction.
Both trajectories approach a
closed curve that corresponds
to a stable periodic solution.
Example 4: Limit Cycle ( = 0.2)
(8 of 13)
Given below are the graphs for the two trajectories previously
mentioned, along with corresponding graphs of u versus t.
The solution solution that is initially smaller gradually
increases in amplitude, while larger solution gradually decays.
Both solutions approach a stable periodic motion that
corresponds to the limit cycle.
Example 4: Phase Difference ( = 0.2)
(9 of 13)
Given below are the graphs for the two trajectories previously
mentioned, along with corresponding graphs of u versus t.
The graph of u versus t shows that there is a phase difference
between the two solutions as they approach the limit cycle.
The plots of u versus t are nearly sinusoidal in shape,
consistent with the nearly circular limit cycle in this case.
Example 4: Solution Graphs ( = 1)
(10 of 13)
The graphs below show similar plots for the case = 1.

Trajectories again move clockwise in the phase plane, but the
limit cycle is considerably different from a circle.
The graphs of u versus t tend more rapidly to the limiting
oscillation than before, and again show a phase difference.
The oscillations are somewhat less symmetric in this case,
rising somewhat more steeply than the fall.
Example 4: Phase Portrait ( = 5)
(11 of 13)
The graph below shows a phase portrait for the case = 5.

Trajectories again move clockwise in the phase plane.
Although solution starts far from the limit cycle, the limiting
oscillation is virtually reached in a fraction of the period.
Starting from one of its extreme values on
the x axis, the solution moves toward other
extreme slowly at first, but once a certain
point is reached, the rest of the transition
is completed swiftly. The process is
repeated in the opposite direction.
Example 4: Solution Graphs ( = 5)
(12 of 13)
Given below is a graph of u versus t for the case = 5, along

with the phase portrait discussed on the previous slide.
Note that the waveform of the limit cycle is quite different
from a sine wave.
Example 4: Discussion
(13 of 13)
Recall that the van der Pol equation is

u (1 u 2 )u + u = 0, > 0
The graphs on the previous slides show that, in the absence of
external excitation, the van der Pol oscillator has a certain
characteristic mode of vibration for each value of .
The graphs of u versus t show that the amplitude of oscillator
changes very little with , but period increases as increases.
At the same time, the waveform changes from one that is
nearly sinusoidal to one that is much less smooth.
The presence of a single periodic motion that attracts all
nearby solutions (asymptotically stable limit cycle), is one of
the characteristics associated with nonlinear equations.
Ch 9.8: Chaos and Strange Attractors:

The Lorenz Equations
In principle, the methods described in this chapter for second
order autonomous systems can be applied to higher order
systems as well. In practice, several difficulties arise.
One problem is that there is simply a greater number of
possible cases that can occur, and the number increases with
the order of the system (and dimension of the phase space).
Another problem is the difficulty of graphing trajectories
accurately in the phase space of higher than two dimensions.
Finally, different and very complex phenomena can occur in
higher order systems that are not present in second order case.
In this section we investigate some of these phenomena by
discussing one particular third order autonomous system.
Convection in a Layer of Fluid

An important problem in meteorology, and other applications
of fluid dynamics, concerns the motion of a layer of fluid, such
as the earths atmosphere, that is warmer at the bottom than at
the top, see the figure below.
If vertical temperature difference T is small, then there is a
linear variation of temperature with altitude but no significant
motion of the fluid layer.
If T is large enough, then a steady convection motion results.
As T increases, a more
complex and turbulent
motion eventually ensues.
Lorenz Equations
While studying these phenomena, the American meteorologist
Lorenz was led to a nonlinear autonomous third order system
dx / dt = ( x + y ),
dy / dt = rx y xz,
dz / dt = bz + xy
These equations are known as the Lorenz Equations.
Note that the second and third equations involve quadratic
nonlinearities.
However, other than being a third order system, superficially
these equations seem no more complicated that the competing
species or predator-prey equations.
Lorenz Equations: Variables and Parameters

The Lorenz equations are
dx / dt = ( x + y ),
dy / dt = rx y xz,
dz / dt = bz + xy
The variable x is related to the intensity of the fluid motion,

while the variables y and z are related to temperature variations
in the horizontal and vertical directions.
The parameters , r, b are all real and positive; and b depend
on the material and geometrical properties of the fluid layer.
For the earths atmosphere, = 10 and b = 8/3.
The parameter r is proportional to the temperature difference
T , and our purpose is to investigate how the nature of the
solutions of the Lorenz equations changes with r.
Critical Points
To find the critical points of the Lorenz equations, we solve
( x + y ) = 0, rx y xz = 0, bz + xy = 0
From the first equation, y = x. Then

x(r 1 z ) = 0, bz + x 2 = 0
If we choose x = 0, then y = z = 0 follows.

Alternatively, choose z = r 1. Then
x = b(r 1) y = b(r 1) , provided r 1
Thus P1 = (0,0,0) is a critical point for all r, and it is the only

critical point for r < 1.
However, when r > 1, we have two other critical points:
P2 =
b(r 1) , b(r 1) , r 1 , P3 = b(r 1) , b(r 1) , r 1
Critical Points: Birfurcation

Our critical points are as follows:
P1 = (0,0,0) , r > 0
( b(r 1) , b(r 1) , r 1), r 1

P = ( b(r 1) , b(r 1) , r 1), r 1
P2 =
3
Note that all three critical points coincide when r = 1.

As r increases through 1, the critical point P1 at the origin
bifurcates, and critical points P2 and P3 come into existence.
We next determine behavior of solns near each critical point.
Although much of the following analysis can done for and b
arbitrary, we will simplify our work by using = 10, b = 8/3.
Linear System for Critical Point P1

With these values for and b, the Lorenz equations are
dx / dt = 10( x + y ), dy / dt = rx y xz, dz / dt = 8 z / 3 + xy
The approximating linear system near the origin is

0 x
x 10 10

0 y
1
y = r
z 0
z
0
8
/
3
The eigenvalues are determined from the equation

10
10
r
0
1
0
0
= (8 / 3 + ) 2 + 11 10(r 1) = 0
8/3
8
11 81 + 40r
11 + 81 + 40r
, 3 =
1 = , 2 =
3
2
2
Solution Behavior Near Critical Point P1

Thus the eigenvalues are
8
3
1 = , 2 =
11 81 + 40r
11 + 81 + 40r
, 3 =
2
2
Note that all three eigenvalues are negative for r < 1.

It follows that the origin is asymptotically stable for r < 1, both
for the linear and the nonlinear systems.
However, 3 changes sign when r = 1, and is positive for r > 1.
The value r = 1 corresponds to the initiation of convective
flow in the physical problem described earlier.
The origin is unstable for r > 1; all solutions starting near the
origin tend to grow, except for those lying in plane determined
by the eigenvectors (1) and (2).
Linear System for Critical Point P2

The approximating linear system near the critical point
P2 =
b(r 1) , b(r 1) , r 1
is
10
u

1
v =
w 8(r 1) / 3

10
u

1
8(r 1) / 3 v
w
8(r 1) / 3
8/3

0
The eigenvalues are determined from the equation

33 + 412 + 8(r + 10) + 160(r 1) = 0
The solutions of this equation depend on r, as described on the
next slide.
Eigenvalues Corresponding to P2 and P3

The approximating linear system near the critical point
P2 =
b(r 1) , b(r 1) , r 1
has eigenvalues according to the following three cases:

For 1 < r < r1 1.3456, there are three negative real eigenvalues.
For r1 < r < r2 24.737, there are one negative real eigenvalue and two
complex eigenvalues with negative real part.
For r2 < r, there are one negative real eigenvalue and two complex
eigenvalues with positive real part.
The same results are obtained for the critical point
P3 = b(r 1) , b(r 1) , r 1
Solution Behavior Near Critical Points

With these eigenvalue results, we have several different cases
for the solution behavior near the critical points.
For r < 1, the only critical point is P1 and it is asymptotically stable.
All solutions approach this point (the origin) as t .
For 1 < r < r1, P2 and P3 are asymptotically stable and P1 is unstable.
All nearby solutions approach either P2 or P3 exponentially.
For r1 < r < r2, P2 and P3 are asymptotically stable and P1 is unstable.
All nearby solutions approach either P2 or P3; most of them spiral
inwardly to the critical point.
For r2 < r, all three critical points are unstable. Most solutions near P2
or P3 spiral away from the critical point.
However, this is not the end of the story, as we see next.
Solution Behavior for Large r

Consider solutions for which r is somewhat larger than r2.
Then P1 has one positive eigenvalue and each of P2 and P3 has
a pair of complex eigenvalues with positive real part.
A trajectory can approach any one of the critical points only on
certain highly restricted paths. The slightest deviation from
these paths causes trajectory to depart from the critical point.
Since none of these critical points is stable, one might expect
that most trajectories would approach infinity for large t.
However, it can be shown that all trajectories remain bounded
as t . In fact, all solutions ultimately approach a certain
limiting set of points that has zero volume. This is true not
only for r > r2, but for all r > 0.
Solution Graph
A plot of computed values of x versus t for a typical solution
with r > r2 is shown below.
This graph resembles a random vibration, even though the
Lorenz equations are deterministic and the solution is
completely determined by the initial conditions.
Nevertheless, the solution also exhibits a certain regularity in
that the frequency and amplitude of the oscillations are
essentially constant in time.
Graphs of Initially Nearby Solutions

The solutions of Lorenz equations are also extremely sensitive
to perturbations in the initial conditions.
The figure below shows the graphs of computed solutions
whose initial points are (5, 5, 5) and (5.01, 5, 5), with r = 28.
The dashed graph is the same one shown on the previous slide.
The two solutions remain close until t is near 10, after which
they seem unrelated.
Thus Lorenz concluded
that detailed long-range
weather predictions are
probably not possible.
Strange Attractors
The attracting set in this case, although of zero volume, has a
complicated structure and is called a strange attractor.
The term chaotic has come into general use to describe the
solutions shown on the previous slide.
To determine how and when the strange attractor is created, it
is illuminating to investigate solutions for smaller values of r.
For r = 21, solutions starting
at three different initial points
are shown here.
Solution Behavior for Smaller Values of r

As shown in figure (a), for the initial point (3, 8, 0) the
solution begins to converge to P3 almost at once.
In figure (b), for initial point (5, 5, 5) there is a short interval
of transient behavior, after which solution converges to P2.
In figure (c), for initial point (5, 5, 10) there is a longer interval
of transient behavior, after which the solution converges to P2.
Strange Attractor and r 24.06

As r increases, the duration of the chaotic transient behavior
also increases.
When r =r3 24.06, the chaotic transients appear to continue
indefinitely, and the strange attractor comes into existence.
Phase Portrait Projections

One can also show the trajectories of the Lorenz equations in
the three-dimensional phase space, or at least projections of
them in various planes.
The figures below show projections in the xy-plane (left graph)
and xz-plane (right graph).
Uniqueness of Solution in Phase Portraits

Observe that the graphs in these figures appear to cross over
themselves repeatedly.
Actual trajectories in three-dimensional space do not cross
each other, because of the uniqueness of solutions results.
The apparent crossings are due to the two-dimensional nature
of the graphs.
Implications for Numerical Computation

The sensitivity of solutions to perturbations of the initial data
also has implications for numerical computations.
Different step sizes, different numerical algorithms, or even
the execution of the same algorithm on different machines will
introduce small differences in the computed solution, which
eventually lead to large deviations.
For example, the exact sequence of positive and negative loops
in the calculated solution depends strongly on the precise
numerical algorithm and its implementation, as well as on the
initial conditions.
However, the general appearance of the solution and structure
of the attracting set are independent of all these factors.
Other Types of Solution Behavior

Solutions of the Lorenz equations for other parameter ranges
exhibit other interesting types of behavior.
For example, for certain values of r > r2, intermittent bursts of
chaotic behavior separate long intervals of apparently steady
periodic oscillation.
For other ranges of r, solutions show the period-doubling
described in Section 2.9 for the logistic differential equation.
Since about 1975, the Lorenz equations and other higher order
autonomous systems have been an active area of research.
Chaotic behavior of solutions appears to be more common
than expected at first, and many questions remain unanswered.

Phase Plane For Linear Control System and Design

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Phase Plane For Linear Control System and Design

Diunggah oleh

Hak Cipta:

Format Tersedia

Ch 9.

1: The Phase Plane: Linear Systems

Solutions of Second Order Linear Systems

Therefore x = ert is a solution of x' = Ax provided that r is an

Equilibrium Solution, Phase Portrait

Characterizing Equation by Trajectory Pattern

Case 1: Real Unequal Eigenvalues

Case 1: Nodal Sink

If the solution starts at an initial point on the line through

x = c1 (1) e r1t + c2 ( 2) e r2t = e r2t c1 (1) e (r1 r2 )t + c2 ( 2 )

Since r1 - r2 < 0, for c2 0 the term c1(1)e(r1 - r2)t is negligible

Case 1: Nodal Source

The phase portrait along with several graphs of x1 versus t

Case 2: Real Eigenvalues

Case 2: Saddle Point

For our general solution

Case 2: Graphs of x1 versus t

The phase portrait along with several graphs of x1 versus t

Case 3: Equal Eigenvalues

Suppose now that r1 = r2 = r. We consider the case in which

Case 3: Star Node

The general solution is

Case 3: Equal Eigenvalues

If the repeated eigenvalue r has only one linearly independent

Case 3: Improper Node

We can rewrite the general solution as

Case 3: Phase Portraits

The phase portrait is given in figure (a) along with several

Case 4: Complex Eigenvalues

Suppose the eigenvalues are i, where and are real,

We introduce the polar coordinates r, given by

Case 4: Polar Equations

Differentiating the polar equations

Case 4: Spiral Point

Solving the differential equations

These equations are parametric equations in polar coordinates

Case 4: Phase Portraits

The phase portrait along with several graphs of x1 versus t are

Case 4: General System

It can be shown that for any system with complex eigenvalues

Case 5: Pure Imaginary Eigenvalues

Suppose the eigenvalues are i, where = 0 and real.

As in Case 4, using polar coordinates r, leads to

Case 5: Phase Portraits

In general, when the eigenvalues are pure imaginary, it is

Behavior of Individual Trajectories

The trajectories never intersect, and exactly one trajectory

Behavior of Trajectory Sets

Type of Critical Point

Ch 9.2: Autonomous Systems and Stability

Initial Value Problem

We assume that the functions F and G are continuous and have

where x = xi + yj, f(x)=F(x,y)i + G(x,y)j, x0 = x0i + y0j, and

In vector form, the solution x = (t) = (t)i + (t)j is a curve

note that the functions F and G depend on x and y, but not t.

Phase Portraits for Autonomous Systems

has a direction field that is independent of time.

lie on the same trajectory, regardless of the time t0 at which

Stability and Instability

does not imply stability.

The Oscillating Pendulum

Here, L and Lsin are the moment arms

Autonomous Pendulum System