Jim Lambers

MAT 605
Fall Semester 2015-16
Lecture 1-2 Notes

These notes correspond to Section 1.1 in the text.

Examples of Dynamical Systems

This course is devoted to the study of systems of ordinary differential equations (ODEs), in terms
of analytical and numerical solution techniques, and also acquiring insight into the qualitative
behavior of solutions. We begin our study of a general system of n ODEs of the form

y0 = F(t, y),

where y(t) : R → Rn and F : Rn+1 → Rn , with some examples that illustrate very different types of
qualitative behavior. The behavior that we can observe has led to the following common synonym.

Definition 1 (Dynamical System) A dynamical system is a system of ordinary

differential equations.

Example 1 (Circular Flow) We begin with the simple dynamical system

x0 = y (1)
0
y = −x. (2)

By differentiating the first equation, we obtain x00 = −x, which has the general solution

x(t) = A cos t + B sin t

where A and B are constants. It follows from x0 = y that y(t) = B cos t − A sin t. If we express the
point (A, B) in polar coordinates A = r cos θ and B = r sin θ, then, using trigonometric identities,
we conclude that this dynamical system has the solution

x(t) = r cos(θ − t), y(t) = r sin(θ − t).

That is, the graph of any solution is a circle of radius r centered at the origin. Because the graph
is a closed curve, we say that this circle is the orbit of the solution. 2

Example 2 (A Row of Stagnation Points) Next, we consider the dynamical system

x0 = sinh y (3)
0
y = − sin x (4)

Unlike the preceding example, an explicit formula for the solution cannot be found analytically.
However, we can still understand the behavior of the solution. The following concept is helpful.

Definition 2 (Streamline) A streamline, or solution curve or integral curve, of

a dynamical system y0 = F(t, y) is a curve r(t) defined for t in an interval I, such that
r0 (t) = F(t, r(t)) for t ∈ I. That is, at each point on the curve, its tangent vector is given
by F(t, r(t)).

1
Streamlines of this dynamical system can be computed using numerical methods, which will be
discussed in an upcoming lecture.
Upon plotting the streamlines of this system, it can be seen that at certain points, there are no
visible streamlines; that is, the flow isn’t “going anywhere”. That is, the velocity of the streamline
is zero. This leads to the following concept.

Definition 3 (Fixed Point) A fixed point, also known as a stagnation point, sta-
tionary point, or equilibrium point of a dynamical system y0 = F(y) is a point at
which F(y) = 0.

It can readily be determined that the dynamical system in this example has fixed points at (x, y) =
(kπ, 0) for each integer k. Furthermore, as |y| increases, the streamlines approach horizontal lines,
with flow from left to right for y > 0 and flow from right to left for y < 0. This can be seen by
examining the velocity vectors of the streamlines, (x0 , y 0 ) = (sinh y, sin x), as |y| increases. 2

Example 3 (Karmen Vortex Sheet) Consider the system

cos x sinh y − sinh c
x0 = (5)
M (x, y)
− sin x cosh y
y0 = , (6)
M (x, y)
where c = 0.8828 and
[cosh(y + c) + cos x][cosh(y − c) − cos x]
M (x, y) = . (7)
2 cosh c
It can be seen that y 0 = 0 when x = kπ, for any integer k, but for such x, the numerator in x0 is
nonzero except when y = ±c. However, when this is the case, M (x, y) = 0, so the derivatives are
actually undefined. That is, this system has no fixed points.
That said, these points at which M (x, y) = 0 are still noteworthy. For x = kπ for even k, and
y = c, or when x = kπ for odd k and y = −c, the velocity of the streamline passing through these
points is infinite, as can be determined using l’Hospital’s Rule. Such a point is referred to as a
vortex. 2

Example 4 (Heat Flow in a Cube) Next, we examine a three-dimensional dynamical system

√
x0 = b sin(2πx) cos(2πy) sinh(2 2πz)
√ √
y 0 = b cos(2πx) sin(2πy) sinh(2 2πz), b = 2π/ sinh(2 2π) (8)
0
√ √
z = −1 − 2b cos(2πx) cos(2πy) cosh(2 2πz)

that models heat flow in the unit cube [0, 1]3 . Let S(x, y, z) denote the temperature throughout
the cube. Then, because heat energy flows from hot to cold, the heat flux vector field is −∇S,
the direction of steepest descent of S. That is, −∇S indicates the direction of streamlines, and
therefore we have
x0 = −Sx , y 0 = −Sy , z 0 = −Sz .
Comparing these equations to the original dynamical system (8) yields
√ √
S(x, y, z) = z + [cos(2πx) cos(2πy) sinh(2 2πz)]/ sinh(2 2π).

Note that this dynamical system has no fixed points in the interior of the cube. 2

2
Example 5 (The Two-Body Problem)

m1 r001 = Gm1 m2 (r2 − r1 )/r12

3
(9)
m2 r002 = Gm1 m2 (r1 − 3
r2 )/r12 (10)

Unlike the previous examples, this system consists of second-order equations rather than first-order.
However, any higher-order system can be rewritten as a first-order system.

Definition 4 (Phase Space) Let F : D → Rn , where D ⊆ Rn . The domain D of F is

called the phase space of the dynamical system y0 = F(y).

In this case, the phase space consists of U × R6 , where U ≡ {(r1 , r2 )|r1 6= r2 }. The additional 6
variables are used for the conversion of the original second-order system into a first-order system,
which adds 6 equations, one for each dependent variable in the second-order system. 2

Example 6 (Pendulum/Ball in a Hoop) The motion of a pendulum, or really, a weight on a

string, can be modeled using the second-order equation

θ00 = −k sin θ,

where θ is the angle that the pendulum makes with the downward axis. As in the previous example,
this equation can easily be converted into a system of two first-order equations,

θ0 = v,
v 0 = −k sin θ.

The phase space of this first-order system is all of R2 .

This equation is too simplified to accurately model the motion of the pendulum, because, for
example, from the position θ = 3π/4, the weight would first drop straight down until the slack in
the string is taken up. Therefore, we impose constrained motion and instead use this equation to
model the motion of a ball in a hoop.
It can be seen from the first-order system that the points (θ, v) = (nπ, 0), where n is an integer,
are fixed points. However, these fixed points are associated with very different behavior of the
solution, depending on whether n is odd or even. If n is even, the ball is at the bottom, so if the
ball is released, it remains stationary. Therefore, we say that these points are stable fixed points. If
n is odd, then the ball is at the top of the hoop, and if it is released, it will quickly move far away
from the top, so the fixed point is unstable.
A phase portrait, or a graph of streamlines in phase space, illustrates the very different behavior
of solutions to this system. The phase portrait for this system is shown in Figure 1. First, it can
be seen that the stable fixed points are the centers of closed orbits. Second, once |v| is sufficiently
large, the streamlines are no longer orbital; instead they are wavy curves that oscillate around a
constant value of v. Physically, this means that once the ball has enough initial velocity, it can
cycle through the hoop indefinitely. Finally, there are two streamlines. that pass through the
unstable fixed points, and clearly define a separation between the orbital and wavy streamlines.
These streamlines are called separatrices for the system.
The phase portrait in Figure 1 was generated in Matlab, using the following function:

function phaseportrait(xp,yp,a,b,c,d,h,xname,yname)
x=a:h:b;

3
 Figure 1: Phase portrait for dynamical system θ0 = v, v 0 = −k sin θ with k = 1

y=c:h:d;
[x2,y2]=meshgrid(x,y);
u2=xp(x2,y2);
v2=yp(x2,y2);
z2=u2./sqrt(u2.^2+v2.^2);
w2=v2./sqrt(u2.^2+v2.^2);
quiver(x2,y2,z2,w2);
xlabel(xname)
ylabel(yname)
axis tight
Then, the function was used as follows:
>> xp=inline(’y’,’x’,’y’);
>> yp=inline(’-sin(x)’,’x’,’y’);
>> phaseportrait(xp,yp,-10,10,-4,4,0.2,’\theta’,’v’);
2

Example 7 (Perturbed Pendulum/Ball in a Hoop) In this example, we modify the previous

motion of the ball in a hoop by allowing the hoop to rotate through an angle a with frequency b.
Assuming the hoop itself has radius 1, we obtain the following modified ODE
a2 b2
θ00 = −g sin θ + sin2 (bt) sin(2θ)
2
which can be written as a first-order dynamical system

θ0 = v (11)
a2 b2
v 0 = −g sin(θ) + sin2 (bt) sin(2θ). (12)
2
This system differs from all of the preceding examples in that v 0 depends directly on the independent
variable t. It follows that this system cannot be written in the form y0 = F(y) for some function
F; instead, we must use the more general form y0 = F(t, y). This leads to the following definition.

4
 Definition 5 (Autonomous System) A dynamical system y0 = F(t, y) is said to be
autonomous, or time-independent, if F is independent of t; that is, an autonomous
system can be written in the form y0 = F(y). When this is not the case, the system is
said to be non-autonomous, or time-dependent.

It is far easier to study the behavior of autonomous systems, as we can look for fixed points,
etc. but fortunately, any non-autonomous system can be rewritten as an autonomous system by
introducing an extra variable:

τ0 = 1
θ0 = v (13)
a2 b2
v 0 = −g sin(θ) + sin2 (bτ ) sin(2θ).
2
If α̃(t) = (t, θ(t), v(t)) is a solution curve of the autonomous system (13), then a solution curve
of the original first-order system (11), (12) can easily be obtained by projecting α̃ onto the θ-v
plane. It follows that a solution curve of a non-autonomous system can cross itself, whereas a
solution curve for an autonomous system cannot. 2

For some dynamical systems, even if is not possible to obtain an explicit formula for solution
curves, it may be possible to describe them implicitly. Consider a system of the form

x0 = f (x, y), y 0 = g(x, y).

Multiplying the first equation by y 0 and the second equation by x0 leads to the equation

f (x, y) dy = g(x, y) dx.

This is a first-order ODE that can be solved if, for example, it is separable or exact, or if an
integrating factor can be found that makes it separable or exact.
If an implicit solution of the form F (x, y) = 0 can be found, then the function F is said to
be a first integral of the dynamical system. The solution curves are the level curves of F ; that is,
F (x, y) = k describes a solution curve for any constant k in the range of F . It is possible to find a
first integral for the system in Example 2, so at least an implicit analytical solution can be found,
if not an explicit one.

Exercises
Section 1.1: Exercise 4