Nonlinear Systems

Mohammed Rida MOKHTARI

École Supérieure en Sciences Appliquées de Tlemcen-ESSAT

August 30, 2022

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 1 / 59
 Chapter 2: Second Order Systems

Chapter 2: Second Order Systems

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 2 / 59
 Chapter 2: Second Order Systems

Concept of Phase Plane

A second-order autonomous system is represented by two scalar differential
equations
ẋ1 = f1 (x1 , x2 )
(1)
ẋ2 = f2 (x1 , x2 )
Let x(t) = (x1 (t), x2 (t)) be the solution of (1) that starts at a certain initial state
x0 = (x10 , x20 ).
f(.) is called a vector field

The set of points {(t, x1 (t), x2 (t)); t ∈ R} with (x1 , x2 ) a solution of (1)
(and x1 (t0 ) = x10 and x2 (t0 ) = x20 for some t0 ) is called the trajectory
or solution curve (through (x10 , x20 )).

The set of points {(x1 (t), x2 (t)); t ∈ R} with (x1 , x2 ) a solution of (1)
(and x1 (t0 ) = x10 and x2 (t0 ) = x20 for some t0 ) is called the f Orbit or
Phase Curve (through (x10 , x20 )).
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 3 / 59
 Chapter 2: Second Order Systems

Concept of Phase Plane

An orbit that forms a closed curve is called a closed orbit.
The family of all trajectories of a dynamical system is called the phase
portrait.

phase-plane

1.0

0.5

0.0

0.5-

1.0-

1.0- 0.5- 0.0 0.5 1.0

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 4 / 59
 Chapter 2: Second Order Systems

Vector fields and Orbits

Example: Uncoupled system

 f is a vector field in R2 .
ẋ = 2x = fx (x, y)
ẏ = −3y = fy (x, y) Solution : (x0 e2t , y0 e−3t )

Orbits / Possible trajectories

. = (−3 ," 2)

Direction
and
speed of
solution
for any
(x, y )
Vector field
Rate of change, velocity Phase portrait
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 5 / 59
 Chapter 2: Second Order Systems

Vector fields and Orbits

Example2:
(1,1)


ẋ = y = fx (x, y) (x, y) → f (x, −x − y 2 )
ẏ = −x − y 2 = fy (x, y) (1, 1) → (1, −2) (−2 ,1)

Rescaled) vector field)

Closed (periodic) orbit

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 6 / 59
 Chapter 2: Second Order Systems

Phase plan analysis

Problem

When x(t) ∈ R2 , study state trajectories around an equilibrium state

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 7 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Consider the following linear system

 
a b
ẋ = Ax, A= (2)
c d

(a, b, c, d) ∈ R. Change of coordinates: z(t) = T −1 x(t), T ∈ R2×2 invertible.

ż(t) = T −1 ẋ(t) = T −1 Ax(t) = T −1 AT z(t) = Jz(t)

The system ż = Jz is equivalent to the system ẋ = Ax.

Remark

A and J = T −1 AT are similar =⇒ they have the same eigenvalues

One can always choose T such that J is in real Jordan form

1 the new coordinates are called normal
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 8 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

There are three possible Jordan forms for A:

Different real eigenvalues.
Equal real eigenvalues.
Complex conjugate.
     
λ1 0 λ k α β
0 λ2 0 λ −β λ

where k=0 or 1.
In addition, we need to consider the case where at least one of the
eigenvalues is zero.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 9 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

 
λ1 0
Case 1: λ1 ∈ R, and independent eigenvectors
0 λ2
In this case
Av1 = λ1 v1 , Av2 = λ2 v2 =⇒ T = [v1 v2 ]
where v1 and v2 are the real eigenvectors of A associated with λ1 and λ2 ,
respectively.
The change of coordinate z = T −1 x, transforms the system into two decoupled
first-order differential equations, i.e.,

ż1 = λ1 z1 , ż2 = λ1 z2

with solution

z20 λ /λ
z1 (t) = z10 eλ1 t , z2 (t) = z20 eλ2 t =⇒ z2 (t) = z 2 1
(z10 )λ2 /λ1 1

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 10 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Case 1a: λ1 < 0 and λ2 < 0

The origin is called stable node

1.0

0.5
y(t)

0.0

0.5

-1.0
-1.0 -0.5 0.0 0.5 1.0
x(t)
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 11 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Example: Consider the linear system
ẋ = −6x − 2y
ẏ = −2x − 9y
eigenvalues : λ1 = −10;
 λ2 = −5 =⇒ stable
 node

−
→
eigen-vectors : v1 =
0.5 −
→
, v2 =
0.5
,
1 1

D = (-0.93, 2.06)
B = (2.28, 1.6)

C = (1.09, -2.28)
A = (-2.6, -2.3 . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 12 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Case 1b: λ1 > 0 and λ2 > 0

The origin is called Unstable Node

1.0

0.5
y(t)

0.0

-0.5

-1.0
-1.0 -0.5 0.0 0.5 1.0
x(t)
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 13 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Example: Consider the linear system
ẋ = x − 2y
ẏ = x + 4y

eigenvalues : λ1 = 2;λ2 = 3=⇒ unstable

 node 
−
→ −2 −
→ −1
eigen-vectors : v1 = , v2 = ,
1 1
y

C = (16.03, 12.17)
B= (16.64 ,-6.54)

A = (-21.46, -1.38)
D = (7.4, -14.64)

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 14 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Case 1d: λ1 =λ2

The origin is called stable/unstable degenerate node

Stable degenerate node: λ1=λ2 < 0 Unstable degenerate node: λ1=λ2 > 0

1.0 1.0

0.5 0.5

y(t)
y(t)

0.0 0.0

-0.5 -0.5

-1.0 -1.0
-1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0
x(t) x(t)
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 15 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Case 1b: λ1 < 0 < λ2

The origin is called saddle

1.0 1.0

0.5 0.5
y(t)

y(t)
0.0 0.0

-0.5 -0.5

-1.0 -1.0
-1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0
x(t) x(t)
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 16 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Example: Consider the linear system
ẋ = 3x + 4y
ẏ = x

eigenvalues : λ1 = 4;λ2 =−1 =⇒ Saddle

 
−
→ 4 −
→ −1
eigen-vectors : v1 = , v2 = ,
1 1
y

D = (-2.32, 2.39)
A = (-2.95, 2.66)

(1 ,4)
(-1 ,1) x

B = (2.17, -1.68)

C = (1.81, -2.14)
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 17 / 59
 Chapter 2: Second Order Systems

ż1 = λ1 z1 → ż1 (t) = z1 (0)eλ1 t

ż2 = λ2 z2 → ż2 (t) = z2 (0)eλ2 t

Case 1e: degenerate saddle

λ1 < λ2 = 0→ all states on the z2 axis are equilibrium states.
0 = λ1 < λ2 → all states on the z1 axis are equilibrium states.
λ1 < λ 2 = 0 0 = λ1 < λ 2
1.0 1.0

0.5 0.5
y(t)

y(t)
0.0 0.0

-0.5 -0.5

-1.0 -1.0
-1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0
x(t) . . . . . . . x(t) .
. . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 18 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Example: Consider the linear system
ẋ = 3x − y
ẏ = −3x + y

eigenvalues : λ1 = 4;λ2 = 0=⇒ degenerate

 Source

−
→ −1 −
→ 0.3333
eigen-vectors : v1 = , v2 =
1 1
y

D = (-0.24, 1.44)

B = (1.35, 1.84)

A = (-1.18, -1.55) C = (0.34, -2.01)

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 19 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Example: Consider the linear system
ẋ = x − 2y
ẏ = −2x − 4y

eigenvalues : λ1 = −5;
 λ2 =0 =⇒ Degenerate
 Sink
−2
eigen-vectors : −
→ , −→
0.5
v1 = v2 =
1 1
y
D = (-0.66, 3.7)

B = (3.25, 2.1)

A = (-2.73, -1.43)

C = (-0.04, -2.87)
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 20 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

 
λ 1
Case 2: λ ∈ R, One can show that the state trajectories are given by
0 λ

ż1 = z10 eλt + z20 teλt

(3)
ż2 = z20 eλt

Assume z20 ̸= 0. If λ ̸= 0, from (3-b-) one gets

 
λt z2 (t) 1 z2 (t)
e = =⇒ t = ln
z20 λ z20

and using (3-a-) one obtains

 
z2 (t) 1 z2 (t)
z1 (t) = z10 + ln z2 (t)
z20 λ z20

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 21 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Case 2: λ ̸= 0

The origin is called stable/unstable improper node

Only the z1 axis is invariant.

1.0 1.0

0.5 0.5
y(t)

y(t)
0.0 0.0

-0.5 -0.5

-1.0 -1.0
-1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0
x(t) Figure: (a) λ < 0, (b) λ > 0 . x(t)
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 22 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Case 3: Complex conjugate eigenvalues λ1,2 = α ± jβ ∈ C
Let v1 = u + jv, v2 = u − jv be the eigenvectors associated to the eigenvalues
λ1 = α + jβ, λ2 = α − jβ. One has
A(u + jv) = (α + jβ)(u + jv) A(u − jv) = (α − jβ)(u − jv)
Summing and subtracting: Au = αu − βv Av = βu + αv

⇒ T = [v1 v2 ]
v1 and v2 are the real eigenvectors of A associated with λ1 and λ2 , respectively.
Defining the change of coordinates
q  
2 2 −1 z2
r = z1 + z2 θ = tan
z1
we can write the dynamic equations in polar coordinates as
ṙ = αr, θ̇ = β
with solution
r(t) = r0 eαt , θ(t) = θ0 + βt . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 23 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Case 3a: α < 0

The origin is called Stable Focus

1.0

0.5
y(t)

0.0

-0.5

-1.0
-1.0 -0.5 0.0 0.5 1.0
x(t)
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 24 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Example: Consider the linear system
ẋ = −2.2x − 2.9y
ẏ = 2.9x + 2y

 + 2j; λ2 =−0.1 − 2j =⇒ Stable Focus

eigenvalues : λ1 = −0.1 
−
→
eigen-vectors : v1 =
2.9 −
→
, v2 =
2.9
−2.1 − 2j −2.1 + 2j
y

B = (0.86, 2.07)

D = (-0.25, 1.25)
x

A = (-1.5, -2.5)
C = (3.82, -3.63)

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 25 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Case 3a: α > 0

The origin is called Unstable Focus

1.0

0.5
y(t)

0.0

-0.5

-1.0
-1.0 -0.5 0.0 0.5 1.0
x(t)
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 26 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Case 3a: α
The origin is called Center

1.0 1.0

0.5 0.5

y(t)
y(t)

0.0 0.0

-0.5 -0.5

-1.0 -1.0
-1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 E0.0 0.5 1.0
x(t) x(t)
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 27 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Example: Consider the linear system
ẋ = −3x + 10y
ẏ = −x + 3y

+ j; λ2 =0 − j =⇒ 
eigenvalues : λ1 = 0  Center 
eigen-vectors : −
→ , − →
10 10
v1 = v2 =
3+j 3−j
y

B = (0.86, 2.07)
D = (0.99, 0.79)

C = (2.63, -3.22)

A = (-0.83, -2.95)

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 28 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

Phase Diagram All of these behaviors can be classified according to the trace Tr
and the determinant Det of the matrix A. Recall that for a matrix
 
a b
A=
c d

Find the eigenvalues of A :

a−λ b
det(A − λI) = 0 ⇒ = λ2 − Tr λ + Det = 0
c d−λ

Thus the eigenvalues are

T r(A) ≡ a + d = λ1 + λ2 ⇒ 12 Tr (A) = m (mean)

Det(A) ≡ ad − p
bc = λ1 λ2 = p (product)
λ1 , λ 2 = m± m −p 2

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 29 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

The values of (m, p) determine the equilibrium type.

If p < 0, then the eigenvalues are real with opposite signs (saddle node).
if m2 < p, then the eigenvalues are complex with a real part (spiral: unstable
if m > 0 and stable if m < 0).
If m = 0 and p > 0, then the eigenvalues are purely imaginary (a center).
p > 0 and m2 > p then the eigenvalues are real with the same sign (a node:
stable if m > 0 and unstable if m < 0).

For linear system

The global qualitative behavior is determined by the type of equilibium point.

For nonlinear system Only local qualitative behavior in the vicinty of equi-
librium point is determined by the type of equilibium point.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 30 / 59
 Chapter 2: Second Order Systems

Behavior of Linear second Order Systems

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 31 / 59
 Chapter 2: Second Order Systems

Qualitative Behavior Near Equilibria

Given the nonlinear system
ẋ1 = f1 (x1 , x2 )
(4)
ẋ2 = f2 (x1 , x2 )

let us assume xe = (x1e , x2e ) is an equilibrium point of (4) i,e.,

f1 (x1e , x2e ) = f2 (x1e , x2e ) = 0
f1 , f2 are continuously differentiable about (x1e , x2e )
Since we are interested in trajectories near (x1e , x2e ), define
x1 = x1e + x̃1 , x2 = x2e + x̃2
x̃1 , x˜2 are small perturbations form equilibrium point.
Expanding (4) into its Taylor series
δf1 (x) δf1 (x)
ẋ1 = ẋ1e + x̃˙ 1 = f1 (x1e , x2e ) + x̃1 + x̃2 + H.O.T
| {z } δx1
xe
δx2
xe
0
δf1 (x) δf1 (x)
ẋ1 = ẋ1e + x̃˙ 1 = f1 (x1e , x2e ) + x̃1 + x̃2 + H.O.T
| {z } δx1
xe
δx2
xe
0 . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 32 / 59
 Chapter 2: Second Order Systems

For sufficiently small neighborhood of equilibrium points, H.O.T. are

negligible

x̃˙ 1 = a11 x̃1 + a12 x̃2 δfi

, ai,j = , i = 1, 2.
x̃˙ 2 = a21 x̃1 + a22 x̃2 δx xe

The equilibrium point of the linear system is

 
  δf1
δx1
δf1
δx2
a11 a12
x̃ = Ax̃, A = = = δf
δx
a21 a22 δf2 δf2 xe
δx1 δx2

Matrix δfδx is called Jacobian Matrix .

The trajectories of the nonlinear system in a small neighborhood of an
equilibrium point are close to the trajectories of its linearization about that
point:
if the origin of the linearized state equation is a
stable (unstable) node, or a stable (unstable) focus or a saddle point,
then in a small neighborhood of the equilibrium point, the trajectory of the
nonlinear system will behave like a
stable (unstable) node, or a stable (unstable) focus
. . or
. . a. saddle
. . . . . point.
. . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 33 / 59
 Chapter 2: Second Order Systems

Qualitative Behavior Near Equilibria

Example ẋ1 = x2 − x21 + 2
ẋ2 = 2(x21 − x22 )

Equilibrium points: f (xe ) = 0 : (−1, −1), (2, 2), (1, −1), (−2, 2)

Linearization :
 
∂f (x) −2x1 1
=
∂x 4x1 −4x2

Linearization around (−1, −1) Linearization around x̄ = (2, 2)

   
∂f (x) 2 1 ∂f (x) 3 0
= =
∂x x̄=(−1,−1) −4 4 ∂x 0 −4
x̄=(2,2)
√ √
Eigenvalues : = {3 ± j 3} Eigenvalues : = {−6 ± 2 3}

⇒ Unstable focus type of equilibrium. ⇒ stable node type of equilibrium.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 34 / 59
 Chapter 2: Second Order Systems

Qualitative Behavior Near Equilibria

Linearization around x̄ = (−1, −1)
  3
∂f (x) −2 1
=
∂x x̄=(−1,0) 4 4
2

√
; Eigenvalues : = λ1 = 1 + 13 > 0
√ 1

λ2 = 1 − 13 < 0 (18)
⇒ Saddle type of equilibrium. 0

Linearization around x̄ = (−2, 2) -1

 
∂f (x) 4 1 -2
=
∂x x̄=(−2,2) −8 −8
√ -3
Eigenvalues : = λ1 = −2 + 2 7 > 0
√ -3 -2 -1 0 1 2 3

λ2 = −2 − 2 7 < 0 (19)
⇒ Saddle type of equilibrium. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 35 / 59
 Chapter 2: Second Order Systems

Qualitative Behavior Near Equilibria

Nonlinear phase portrait Linear phase portrait
3 3
2 2
1 1
y(t)

y(t)
0 0
-1 -1 Example: The
-2 -2 Liénard
-3 -3 equation (red)
-2 -1 0 1 2 -2 -1 0 1 2 and its
x(t) x(t) linearization
Displacement Velocity (black).
x(t) y(t) Parameter
1 µ = 0.95
6
t
2 4 6 8 10 12 14 4
-1
-2 2
-3 t
-4 2 4 6 8 10 12 14
-2
-5 . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 36 / 59
 Chapter 2: Second Order Systems

Qualitative Behavior Near Equilibria

Nonlinear phase portrait Linear phase portrait
3 3
2 2
1 1
y(t)

y(t)
0 0
-1 -1
Example: The
-2 -2
Liénard
-3 -3 equation (red)
-2 -1 0 1 2 -2 -1 0 1 2 and its
x(t) x(t) linearisation
Displacement Velocity (black).
x(t) y(t)
Parameter
3
2 µ = −0.35
2
1 1
t t
5 10 15 20 25 5 10 15 20 25
-1
-1
-2 . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 37 / 59
 Chapter 2: Second Order Systems

Example: ambiguous borderline case


 ẋ1 = −x2 + |µx1 (x{z
2 2
 1 + x2 )
}
z }| {
nonlinear terms (5)


ẋ2 = x1 + µx2 (x21 + x22 )
2
Fixed point : (x1e , x2e ) = (0, 0).
1
Linearization :
0
 
∂ ẋ1 ∂ ẋ1  
∂x1 ∂x2 0 −1 -1
J=  =
∂ ẋ2 ∂ ẋ2 1 0 -2
∂x1 ∂x2 0,0 -2 -1 0 1 2

It is important to note that the linearized system does not depend on the
control parameter µ.
Classification of the fixed point of the linearized system.
Trace of the system matrix is Tr = 0.
Determinant of the system matrix is p = 1.
The linear fixed point is a centre. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 38 / 59
 Chapter 2: Second Order Systems

Example : The Lotka-Volterra competitive cohabitation model2 from ecology

competitive cohabitation of rabbits and sheep. The model has the following form:

ẋ = x(3 − x) − 2xy
(6)
ẏ = y(2 − y) − xy
where x and y are the sizes of rabbit and sheep populations, respectively.
{0, 2} {1, 1}
1.0 1.0
3.0
0.5 0.5

Sheep v(t)
Sheep v(t)
2.5 0.0 0.0

-0.5 -0.5
2.0
Sheep y(t)

-1.0 -1.0
0.0 0.2 0.4 0.6 0.8 1.0 -1.0 -0.5 0.0 0.5 1.0
1.5 Rabbits Rabbits
{3, 0} u(t)
{0, 0}u(t)
1.0 1.0
1.0 0.8 0.8

Sheep v(t)
Sheep v(t)

0.6 0.6
0.5 0.4
0.4
0.2 0.2
0.0
0.0 0.0
0 1 2 3 4 0.0 0.2 0.4 0.6 0.8 1.0 -1.0 -0.5 0.0 0.5 1.0
. Rabbits u(t)
Rabbits u(t). . . . . . . . . . . . . . . . . . .
Rabbits x(t) . . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 39 / 59
 Chapter 2: Second Order Systems

Phase Portraits of Nonlinear Systems Near Hyperbolic Equilibria

Hyperbolic equilibrium point: linearization has no eigenvalues on the
imaginary axis.

Hartman-Grobman Theorem:
If xe is a hyperbolic equilibrium of a planar dynamical system ẋ = f (x),
x ∈ R2 then there is neighborhood U around xe and a homeomorphism1
h : U → R2
that maps the nonlinear trajectories in U to the linear trajectories in R2 .
homeomorphism: a continuous map with a continuous inverse (i.e. a change of
coordinates)
x2 x2
0.2 0.2

0.1 0.1

0.0
x1 0.0
x1

0.1- 0.1-

0.2- 0.2-

0.2- 0.1- 0.0 0.1 0.2 0.2- 0.1- 0.0 . . 0.1. . .0.2 . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 40 / 59
 Chapter 2: Second Order Systems

Example: Consider the non-linear autonomous system

ẋ1 = −x1
ẋ2 = x2 + x21

Equilibrium point : ẋ = 0 ⇒ x̄ = (0, 0)T .

Eigenvalues: λ1 = −1, λ2 = 1 =⇒ saddle type of equilibrium.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 41 / 59
 Chapter 2: Second Order Systems

Phase Portraits of Nonlinear Systems Near Hyperbolic Equilibria

Example
ẋ1 = −x2 + µx1 (x21 + x22 )
(7)
ẋ2 = x1 + µx2 (x21 + x22 )

There is only one equilibium point at (0,0), and the linearized system at this
point is
2

1
 
0 −1
ẋ = =⇒ λ1,2 = {±J} 0
1 0
-1

-2
-2 -1 0 1 2

=⇒ the equilibrium point is center. Since this equilibrium point is

non-hyperbolic =⇒ No conclusion about the behavior of the nonlinear
system near (0,0)
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 42 / 59
 Chapter 2: Second Order Systems

Frame Title
System (7) is analysed in polar coordinates. Usually a direct coordinate transform
  p
x1 = r cos θ r = x21 + x22
, (9)
x2 = r sin θ θ = tan (x2 /x1 )

where r = r(t) and θ = θ(t), is used. We are searching a system in the form:

ṙ = f1 (r, θ)
(10)
θ̇ = f2 (r, θ)

where functions f1 (r, θ) and f2 (r, θ) are to be determined. We are interested in

temporal dynamics of (9)

 x1 f1 (x1 , x2 ) + x2 f2 (x1 , x2 )
 2rṙ = 2x1 ẋ1 + 2x2 ẋ2 ⇒ ṙ =
r
 x ẋ − x ẋ x f (x , x ) − x2 f1 (x1 , x2 )
 sec2 θθ̇ = (1 + tan2 θ)θ̇ = 1 2 2 2 1 ⇒ θ̇ = 1 2 1 22
x1 x1 + x22
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 43 / 59
 Chapter 2: Second Order Systems

Phase Portraits of Nonlinear Systems Near Hyperbolic Equilibria

After developing (11). The system (7) has been represented in polar coordinates.
Resulting decoupled equations

ẋ1 = −x2 + µx1 (x21 + x22 ) ṙ = µr3
2 2 =⇒
ẋ2 = x1 + µx2 (x1 + x2 ) θ̇ = 1

2 2 2

1 1 1

x2(t)
x2(t)

x2(t)

0 0 0

-1 -1 -1

-2 -2 -2
-2 -1 0 1 2 -2 -1 0 1 2 -2 -1 0 1 2
x1(t) x1(t) x1(t)

Figure: µ < 0, µ = 0, µ>0

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 44 / 59
 Chapter 2: Second Order Systems

Non-existence of Periodic Orbits

Bendixson criterion gives a sufficient condition for detecting the absence of
periodic orbits for second-order systems (Limit cycles or neutrally stable cycles).

Bendixson criterion:
For a time-invariant planar system

ẋ1 = f1 (x1 , x2 ), ẋ2 = f2 (x1 , x2 )

 
f1
If div(f ) = ∇.f (x) = [∂/∂x1 ∂/∂x2 ] = ∂f1 /∂x1 + ∂f2 /∂x2 is
f2
not identically zero and does not change sign in a simply connected region
D, then there are no periodic orbits lying entirely in D.

Example 1: ẋ = Ax, x ∈ R2 can have periodic orbits only if div f= trace(A)=0.,

 
0 −β
A=
β 0
Unless trace(A) = 0 =⇒ non periodic orbits. .
.
.
.
.
. . . . .
. . . .
. . . .
. . . .
. . . .
. . . . .
.
.
.
.
.
.
.
.
.

R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 45 / 59

 Chapter 2: Second Order Systems

Example 2:
ẋ1 = x2
ẋ2 = −δx2 + x1 − x31 + x21 x2 , δ>0
∂f1 ∂f2
∇.f (x) =
+ = x21 − δ
∂x1 ∂x2
√
∇.f (x) = 0, then x1 = ± δ
Therefore, no periodic orbit can lie entirely in the region
√ √ √ √
x1 ∈] − ∞, − δ[, ] − δ, δ[, ] δ, +∞[

x2
x2
not possible
possible

x1 x1

√
√ √ x1=− δ √
x1=− δ x1 = δ x1 = δ

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 46 / 59
 Chapter 2: Second Order Systems

Periodic Orbits in the Plane

Let ϕ(t, x0 ) denotes the solution of ẋ = f (x) with initial condition x(0) = x0 .

Definition: Invariant sets

A set M ⊆ Rn is positively invariant if, for each x0 ∈ M, ϕ(t, x0 ) ∈ M

for all t ≥ 0.

Theorem n( x )

If V : Rn → R is of class C 1 and M = {x : V (x) ≤

c}, then M is invariant if
M f (x)
f (x).∇V (x) ≤ 0 ∀x : V (x) = c

i.e. if x is on the boundary of M , then the vector

f (x) points into M .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 47 / 59
 Chapter 2: Second Order Systems

Non-existence of Periodic Orbits

Example: Consider a closed orbit ẋ = f (x).

f (x) is tangential to the trajectory x. Along this closed M
trajectory : f T (x).⃗n = 0.
The interior of any closed trajectory is a positively
invariant set.

n( x )

Example:
If f (x)T .⃗n ≤ 0 then M is positively invariant.
⃗n: outward normal on a boundary of M. M f (x)
Along boundary of M :for all x ∈ ∂M ⇒ [f (x)]T .⃗n ≤ 0.

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 48 / 59
 Chapter 2: Second Order Systems

Periodic Orbits in the Plane

y
Example: Predator-prey model

prey : ẋ1 = (a − bx2 )x1 ( dc , ba )

predator : ẋ1 = (cx1 − d)x2

a,b,c,d positive parameters.

   
0 d/c saddle
Equilibrium points : x̄ = , x̄ = . %
0 a/b x

Clearly [f (x)]T .⃗n = 0 along the boundary of M = {x1 ≥ 0, x2 ≥ 0}, which means
the first quadrant M is positively invariant.  
a 0
Linearization around x̄ = (0, 0) : A = ∂f = .
∂x
x̄ 0 −d
λ1 = a > 0 and λ2 = −d < 0 =⇒ saddle type of equilibrium.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 49 / 59
 Chapter 2: Second Order Systems

Periodic Orbits in the Plane

Example 2:

ẋ1 = x1 + x2 − x1 (x21 + x22 )

ẋ2 = −2x1 + x2 − x2 (x21 + x22 )

Show that Br := {x ∈ R2 /x21 + x2 ≤ r2 } is positively invariant for sufficiently

large r (to be determined). We want to calculate [f (x)]T n(x)
   
∂V /∂x1 2x1
V (x) = x1 + x2 = r ⇒ n(x) = ∇V (x) =
2 2 2
=
∂V /∂x2 2x2  
x1
n( x ) =
x2 x2
T ∂V ∂V
[f (x)] .n(x) = = f1 ∂x 1
+ f2 ∂x 2
= −2(x1 + x2 ) + 2x1 + 2x22
2 2 2 2
− 2x1 x2
f (x)
−2x2 x2 ≤ x21 + x22 (completion of squares). x1

Therefore [f (x)]T .n(x) ≤ −2r2 (r2 − 3/2)) ≤ 0 if

r2 ≥ 3/2. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 50 / 59
 Chapter 2: Second Order Systems

Existence Theorem of Limit Cycle

Poincaré-Bendixson Theorem:

Let M be a compact (closed and bounded) set in R2 , which is positively

invariant for ẋ = f (x), x ∈ R2 . If M does not contain an equilibrium point,
then it contains a periodic orbit.

The "no equilibrium condition" in PB Theorem can be relaxed as: " M can
have one equilibrium point which is either an unstable focus or an unstable
node, then there is a periodic orbit.
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 51 / 59
 Chapter 2: Second Order Systems

x2
Example: harmonic oscillator
    
ẋ1 0 1 x1
=
ẋ2 −1 0 x2 x1

For any R > r > 0, the ring {x : r2 ≤ x2 + x2 ≤ R2 } is

compact, invariant and contain no equilibria.
[f (x)]T .n(x) = 0 everywhere and PB Theorem state there exists a periodic orbit
(or more) in M.
Example:2
ẋ1 = x2 + x1 x22
ẋ1 = −x1 + x21 x2
Linearization around the equilibrium at x̄ = (0 0)T yields
    
ẋ1 0 1 x1
=
ẋ2 −1 0 x2
which exhibits a continuum of periodic solutions. However, for this nonlinear
system, we have
∇.f (x) = x21 + x22 > 0, ∀x ̸= 0
Hence, Bendixson theorem leads to the conclusion that this dynamical system has
no nontrivial periodic solutions. .
.
.
.
.
. . . . .
. . . .
. . . .
. . . .
. . . .
. . . . .
.
.
.
.
.
.
.
.
.

R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 52 / 59

 Chapter 2: Second Order Systems

Existence Theorem of Limit Cycle

Example:3

ẋ1 = x1 + x2 − x1 (x21 + x22 )

ẋ2 = −2x1 + x2 − x2 (x21 + x22 )

px1 + x2 > 3/2. i.e. Br = x ∈ R /x1 + x2 ≤ r is

[f (x)]T .n(x) ≤ 0 iff r2 = 2 2 2 2 2 2

positively invariant r ≥ 3/2 but contains the equilibrium xe = 0.

2
 
∂f 1 1 √
= , λ1,2 = 1 ± j 2, unstable focus 1
∂x xe =0
−2 1
0
Therefore, Br must contain a periodic orbit.
-1

-2
. . . 1- 2-
. . . .. . . .
0
. . . . .
1 . .
2. .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 53 / 59
 Chapter 2: Second Order Systems

Limit Cycle: Stable Limit Cycle

All trajectories in the vicinity of the limit cycle converges to it as t → ∞

Example:

ẋ1 = x2 − x1 (x21 + x22 − 1) 1

ẋ1 = −x1 − x2 (x21 + x22 − 1)


ṙ = −r(r2 − 1)
=⇒ (12)
θ̇ = −1 0

if r > 1 → ṙ > 0 converging -1

if r > 1 → r < 0 converging

if r = 1 → ṙ = 0 remaining
-2

-2 -1 0 1 2

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 55 / 59
 Chapter 2: Second Order Systems

Limit Cycle: Unstable Limit Cycle

All trajectories in the vicinity of the limit cycle diverges from it as t → ∞

Example 2

ẋ1 = x2 + x1 (x21 + x22 − 1) 1

ẋ1 = −x1 + x2 (x21 + x22 − 1)


ṙ = r(r2 − 1)
=⇒ 0

θ̇ = −1

if r < 1 → ṙ < 0
-1
diverging
if r > 1 → r > 0 diverging
if r = 1 → ṙ = 0 remaining -2

-2 -1 0 1 2

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 56 / 59
 Chapter 2: Second Order Systems

Limit Cycle: half-stable Limit Cycle

Some of the trajectories in the vicinity of the limit cycle converges to it,
while others diverge from it as t → ∞

Example: 2

ẋ1 = x2 − x1 (x21 + x22 − 1)2 1

ẋ1 = −x1 − x2 (x21 + x22 − 1)2


ṙ = −r(r2 − 1)2
=⇒
θ̇ = −1 0

if r < 1 → ṙ < 0
-1
diverging
if r > 1 → r < 0 converging
if r = 1 → ṙ = 0 remaining -2

-2 -1 0 1 2

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 57 / 59
 Chapter 2: Second Order Systems

Bifurcation: Holf Bifurcation

Example: Supercritical Hopf bifurcation
 
ẋ1 = −x2 + x1 (µ − x21 − x22 ) ṙ = r(µ − r2 )
=⇒ (15)
ẋ2 = +x1 − x2 (µ − x21 − x22 ) θ̇ = 1
Equilibrium points : r(µ − r̄2 ) = 0.
Note that a positive equilibrium for the r subsystem means a limit cycle in the
(x1, x2) plane.
µ < 0 : stable equilibrium at r = 0. x2

µ > 0 unstable equilibrium point ar r = 0

√ x1
and stable limit cycle at r = µ.
The origin loses stability at µ = 0 and a
stable limit cycle emerges. µ

. . . . . . . . . . . . . . . . . . . .
µ µ µ . . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 58 / 59
 Chapter 2: Second Order Systems

Bifurcation: Holf Bifurcation

In Supercritical Hopf bifurcation by increase of µ near zero, the stable equi-

librium point becomes unstable but a stable limit cycle appears. Hence, this
is a Safe bifurcation.

'µ

. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
R. MOKHTARI (SNL) Nonlinear Systems August 30, 2022 59 / 59