Differential Equations Massoud Malek

Equilibrium Points
♣ Limit-Cycle. A limit-cycle on a plane or a two-dimensional manifold is a closed trajec-
tory in phase space having the property that at least one other trajectory spirals into it
either as time approaches infinity or as time approaches minus-infinity. Such behavior
is exhibited in some nonlinear systems. In the case where all the neighboring trajecto-
ries approach the limit-cycle as time t → +∞, it is called a stable or attractive limit-cycle.
If instead all neighboring trajectories approach it as time t → −∞, it is an unstable or
non-attractive limit-cycle. In all other cases it is neither ”stable” nor ”unstable”.

Figure illustrating a stable limit cycle for the Van der Pol oscillator
d2 x dx
2
− µ(1 − x2 ) + x = 0.
dt dt
As seen in the figure, trajectories for various initial states of this system converge to the
limit cycle.
♣ Qualitative Analysis of Systems with Complex Eigenvalues. Consider the linear homo-
geneous system (
x0 (t) = ax + by,
,
y 0 (t) = cx + dy.

where the eigenvalues are λ1 = α + iβ and λ2 = α − iβ. Then the general solution is given
by    
x(t) x
= eαt (cos βt + i sin βt) 0 ,
y(t) y0
   
x a b
where X0 = y 0 is a complex eigenvector of the matrix A = c d
.
0

The fact that sin and cos are periodic function implies that the solutions spiral around
the origin with a period of 2π/β . This quantity is called the natural period of the system.
Every solution takes this amount of time to wind once around the origin. The natural
frequency is the reciprocal of the natural period, or β/2π . Moreover
• If α < 0, then the solutions tend to the origin (when t → ∞) while spiraling. In this
case, the equilibrium point is called a spiral sink.
• If α > 0, then the solutions explode or get away from the origin (when t → ∞) while
spiraling. In this case, the equilibrium point is called a spiral source.
• If α = 0, then the solutions are periodic. This means that the trajectories are closed
curves or cycles. In this case, the equilibrium point is called a center.

California State University, East Bay

Massoud Malek Equilibrium Points Page 2

P hase P ortrait

α<0 α>0 α=0

The question of “which way” the solutions spiral (clockwise or counterclockwise) can be
answered by looking at the direction field. Even one vector of the vector field is enough
to tell which way a system with complex eigenvalues spirals. For example, if the direction
field at (1, 0) points down into the fourth quadrant, then the solutions must spiral in the
clockwise direction. If the direction field at (1, 0) points up into the first quadrant, then
the solutions must spiral in the counterclockwise direction.
♣ Qualitative Analysis of Systems with Zero as an Eigenvalue. Consider a linear system
where at least one of the eigenvalues is zero. In fact, it is easy to see that this happen
if and only if we have more than one equilibrium point (which is (0, 0)). In this case, we
will have a line of equilibrium points (the direction vector for this line is the eigenvector
associated to the eigenvalue zero).
Consider the linear homogeneous system
(
x0 (t) = 2x − y,
.
y 0 (t) = −2x + y.
 
The eigenvalues are λ1 = 0 and λ2 = 3. The associated eigenvectors are v1 = 1 and
  2
1
v2 = respectively. Therefore the general solution is:
−1
     
x(t) 1 1
= C1 e0t + C2 e3t .
y(t) 2 −1

Note that all the solutions are line parallel to the line x + y = 0. When t → ∞, the
trajectory goes to infinity. But when t → −∞, the trajectory converge to the equilibrium
point on the line of equilibrium points 2x − y = 0. The picture below explains more what
is happening.

♣ When Linearization Fails. Unfortunately, in some cases the information given by the
linearized system is not enough to completely determine the behavior of the nonlinear
system near the equilibrium point. We shall explain this problem with the help of two
examples.

California State University, East Bay

Massoud Malek Equilibrium Points Page 3

Consider the systems:

( (
x0 (t) = y − (x2 + y 2 )x, x0 (t) = y + (x2 + y 2 )x,
(a) (b)
y 0 (t) = −x − (x2 + y 2 )y. y 0 (t) = −x + (x2 + y 2 )y.

Both systems have the same unique equilibrium point (0, 0) and the same linearized
system:  0    
x (t) 0 1 x(t)
(c) = .
y 0 (t) −1 0 y(t)
The eigenvalues of the linearized system (c) are λ1 = i and λ2 = −i. Since the real part of
the eigenvalues is zero, the equilibrium point (0, 0) of (c) is a center. All nonzero solutions
of the linearized system lie on periodic solutions that wind around (0, 0) in the clockwise
direction. But there are no periodic solution for the nonlinear systems (a). To see this,
we regard the vector field as a sum of two vector fields. The linear field (y, −x) and the
nonlinear vector field (−(x2 + y 2 )x, −(x2 + y 2 )y). The linear field (y, −x) is always tangent to
circles centered at the origin. But the vector field (−(x2 + y 2 )x, −(x2 + y 2 )y) always point
toward (0, 0), since this vector field is basically a multiple of (−x, −y) by a positive number
x2 + y 2 . The net effect of adding these two vector fields is a vector field that always
point “into” a circle centered at the origin. This forces solutions to spiral slowly into the
equilibrium point (0, 0) rather than to circulate on periodic orbit.
Note that the system (b) is obtained from (a) by changing the signs of its higher order
terms, so solutions of (b) spiral away from the origin.
The solutions of the nonlinear systems (a) and (b), and the solutions of the linearized
system (c) are still approximately the same, at least for a short amount of time. The
problem is that because the fixed point of the linearized system is a center,any small
perturbation caused by the inclusion of the nonlinear terms can turn the center into a
spiral sink or a spiral source.
Fortunately there are only two situations in which the long term behavior of solutions
near an equilibrium point of the nonlinear system and its linearization can differ. One
is when the linearized system is a center, the other is when the linearized system has
zero as eigenvalue. In any other case, the local picture of the nonlinear system near an
equilibrium point looks like its linearization.
♣ The Distance Formula. Suppose e = (0, 0) is an equilibrium point of the following system
of differential equations (
x0 (t) = f (x(t), y(t))
y 0 (t) = g(x(t), y(t)).

Let p(t) = (x(t), y(t)) 6= (0, 0) = e be a solution point. The distance between p(t) and e is
given by p
d(p(t), e) = [x(t)]2 + [y(t)]2
and its instantaneous rate of change is obtain as follows:
" #
0 1 2x(t)x0 (t) + 2y(t)y 0 (t) x(t) [f (x(t), y(t))] + y(t) [g(x(t), y(t))]
[d(p(t), e)] = p = p
2 2
[x(t)] + [y(t)] 2 [x(t)]2 + [y(t)]2
p
let D(t) = x(t) [f (x(t), y(t))] + y(t) [g(x(t), y(t))], then since [x(t)]2 + [y(t)]2 > 0, the distance
between p(t) and e is increasing (resp. decreasing) if D(t) > 0 (resp. D(t) < 0).
If the equilibrium point (a, b) is different from (0, 0), then by choosing u = x−a and v = y −b,
we obtain a system, where (0, 0) is an equilibrium point.

California State University, East Bay

Massoud Malek Equilibrium Points Page 4

Examples. Consider the following nonlinear differential equations systems:

( 0 ( 0
x = − x3 − y x = y − xy 2
(a) and (b) .
y0 = x − y3 y 0 = − x + x2 y

The origin e = (0, 0) is the equilibrium point of both (a) and (b). The linearized systems
around the origin are:
( (
x0 = − y x0 = y
(a) and (b) .
y0 = x y0 = − x

The characteristic equation of both systems is λ2 + 1 = 0 and the eigenvalues are λ1 = i

and λ2 = −i. Thus the equilibrium point e of the linearized systems is a center. Notice
that
(a) D(t) = x(−y) + yx = 0 and (b) D(t) = xy + y(−x) = 0;
hence the origin is a center and the trajectories are circles. This is a case when the
linearizations fails to describe the solutions of the original systems around the origin.
Using the nonlinear systems, we find that
(a) D(t) = (−x4 − xy) + (xy − y 4 ) = −(x4 + y 4 ) < 0. This says that the distance between points
on the trajectory and the origin is decreasing with time; that is, the trajectory spiral
toward the origin.
(b) D(t) = (xy − x2 y 2 ) + (−yx + x2 y 2 ) = 0. Thus all trajectories are circles around the origin.
1
But there are also other equilibrium points; in fact any point on the hyperbola y =
x
is an equilibrium point; so for a 6= 0, (a, 1/a) is an equilibrium point. By choosing
u = x − a and v = y − 1/a, the system becomes:
 
1 1

0
 u = (v − a ) a u + av − uv


  .
 0 1
 v = − (u − a) u + av − uv

a

The linearized system is

u0 − a12
    
−1 u
= .
v0 1 a2 v

The eigenvalues are λ1 = 0 and λ2 = a2 − 1/a2 . This is the case when the linearized
system fails to describe the behavior of the phase plane of the original system. But
the distance formula tells us that every trajectory is a circle centered around the
origin with two stationary points at (a, 1/a) on the circle.
♣ Bifurcation Point. When a nonlinear system depends on a parameter, then as the
parameter changes, the equilibrium points can change. That is, as the parameter changes
a bifurcation can occur. Consider the one-parameter system:
(
x0 (t) = x2 − α
, where α is a parameter.
y 0 (t) = −(x2 + 1)y.

• If α < 0, then the there is no x-nullclines, hence the system has no equilibrium points.
• If α = 0, then the system has exactly one equilibrium point at (0, 0).
√ √
• If α > 0, then the system has two equilibrium points (− α, 0) and ( α, 0).

California State University, East Bay

Massoud Malek Equilibrium Points Page 5

The system changes from having no fixed points to having two fixed points when the
parameter α is increase through α = 0. We say that the system has a bifurcation when
α = 0 and that α is a bifurcation point.
 
2x 0
The Jacobian matrix is J = with
−2xy −x2 − 1
 √  √
√ √
   
0 0 −2 α 0 2 α 0
J(0, 0) = , J(− α, 0) = , and J( α, 0) = .
0 −1 0 −α − 1 0 −α − 1

• For α = 0, there will be a line equilibrium (since one of the eigenvalues is zero).
√ √
• For α > 0, the point (− α, 0) is a sink and ( α, 0) is a saddle point.
Saddle-Node Bifurcation
A saddle-node bifurcation or tangent bifurcation is a collision and disappearance of two equilibria
in dynamical systems. In autonomous systems, this occurs when the critical equilibrium
has one zero eigenvalue. This phenomenon is also called fold or limit point bifurcation.
In n-dimensional case with n ≥ 2, the Jacobian matrix at the saddle-node bifurcation has
• a simple zero eigenvalue λ1 = 0, as well as
• nr eigenvalues with λj < 0, and ns eigenvalues with λj > 0, with nr + ns + 1 = n.
Thus in the above example, for α = 0, there is a saddle-node bifurcation point at (0, 0).

Andronov-Hopf Bifurcation
A Hopf or Poincare-Andronov-Hopf bifurcation is a local bifurcation in which a fixed point
of a dynamical system loses stability as a pair of complex conjugate eigenvalues of the
linearization around the fixed point cross the imaginary axis of the complex plane.
Consider the one-parameter systems:
( (
x0 (t) = βx − y − (x2 + y 2 )x, x0 (t) = βx − y + (x2 + y 2 )x,
(a) (b)
y 0 (t) = x + βy − (x2 + y 2 )y. y 0 (t) = x + βy + (x2 + y 2 )y.

Both systems have the same unique equilibrium point at the origin and the same lin-
earized system:  0    
x (t) β −1 x(t)
(c) = .
y 0 (t) 1 β y(t)

The eigenvalues of the linearized system (c) are λ1 = β + i and λ2 = β − i.

• The equilibrium point of the system (a) is asymptotically stable for β < 0, weakly
stable at β = 0, and unstable for β > 0. Moreover, there √ is a unique and stable
circular limit cycle that exists for β > 0 and has radius β . This is a supercritical
Andronov-Hopf bifurcation.

California State University, East Bay

Massoud Malek Equilibrium Points Page 6

• The origin is asymptotically stable for β < 0 and unstable for β ≥ 0, and weakly stable
β = 0. Moreover, there is a unique and unstable limit cycle for β < 0. This is a
subcritical Andronov-Hopf bifurcation.

Supercritical Hopf bifurcation Subcritical Hopf bifurcation

California State University, East Bay