Chapter 10

Qualitative Theory of Differential

Equations

by

Dwijendra Narain Pandey

Department of Mathematics
Indian Institute of Technology Roorkee
Roorkee 247667, Uttarakhand.
E-mail: [Link]@[Link]
Module 4: Stability of equilibrium solutions.

1 Stability of equilibrium solutions

Consider the differential equation

ẋ = Ax + g(x) (1)

where the nonlinear part  

g1 (x)
g(x) = 
 .. 
. 
gn (x)
is weak in comparison to x, that is

g1 (x) gn (x)
,...,
max{kx1 k, . . . , kxn k} max{kx1 k, . . . , kxn k}

are continuous functions of x1 , . . . , xn and vanishes for x1 = . . . = xn = 0. If

g(0) = 0 then x(t) ≡ 0 is an equilibrium solution of (1). Such a system is called
a weakly nonlinear system.

Stability of every solution for weakly nonlinear system may not be possible
but stability of equilibrium solutions may be discussed in some cases with the
help of stability of zero solution of linearized problem. The following theorem
is useful in this regard.

Theorem 1. [Braun 1993] Suppose that the vector-valued function

g(x)/kxk = g(x)/max{kx1 k, . . . , kxn k}

is a continuous function of x1 , . . . , xn which vanishes for x = 0.

Then,

(a) The equilibrium solution x(t) ≡ 0 of (1) is asymptotically stable if the

equilibrium solution x(t) ≡ 0 of the ”linearized” equation ẋ = Ax is
asymptotically stable. Equivalently, the solution x(t) ≡ 0 of (1) is asymp-
totically stable if all the eigenvalues of A have negative real part.

(b) The equilibrium solution x(t) ≡ 0 of (1) is unstable if at least one eigen-
value of A have positive real part.

1
 (c) The stability of the equilibrium solution x(t) ≡ 0 of (1) can not be deter-
mined from the stability of the equilibrium solution x(t) ≡ 0 of ẋ = Ax if
all the eigenvalues of A have real part ≤ 0 but at least one eigenvalue of
A has zero real part.
Proof. We may use the variation of parameter formula to write any solution
x(t) of (1) in the form
Z t
At
x(t) = e x(0) + eA(t−s) g(x(s))ds. (2)
0

Part (a)
We want to show that kx(t)k → 0 as t → ∞. Now we know that if all the
eigenvalues of A have negative real part, then we can find positive constants K
and α such that

keAt x(0)k 6 Ke−αt kx(0)k

and

keA(t−s) g(x(s))k 6 Ke−α(t−s) kg(x(s))k.

Moreover we can find a positive constant δ such that

α
kg(x)k 6 kxk if kxk 6 δ.
2K
This follows from our assumption that g(x)/kxk is continuous and vanishes at
x = 0. Consequently, equation (1) implies that
Z t
kx(t)k 6 keAt x(0)k + keA(t−s) g(x(s))kds
0
α t −α(t−s)
Z
6 Ke−αt kx(0)k + e kx(s)kds
2 0

as long as kx(s)k 6 δ, 0 6 s 6 t. By multiplying both sides by eαt , we get

α t αs
Z
αt
e kx(t)k 6 Kkx(0)k + e kx(s)kds.
2 0

This inequality can be simplified by setting z(t) = eαt kx(t)k, for then

α t
Z
z(t) 6 Kkx(0)k + z(s)ds.
2 0
We see that

kx(t)k = e−αt kz(t)k 6 e−αt [Kkx(0)k + U (t)]

−αt
6 Kkx(0)ke 2

2
as long as kx(s)k 6 δ, 0 6 s 6 t. or if kx(0)k 6 δ/K, then the above inequality
gives that kx(t)k 6 δ for all future time t. Consequently, this inequality is true
for all t > 0 if kx(0)k 6 δ/K. Finally we observe that kx(t)k 6 Kx(0) and
kx(t)k → 0 as t → ∞. Therefore the equilibrium solution x(t) ≡ 0 of (1) is
asymptotically stable.
(b) The proof of (b) is more involving and lengthy so we skip the proof.
(c) We will present two differential equations of the form (1) where the
nonlinear term g(x) determines the stability of the equilibrium solution x(t) ≡ 0.
Consider the first system of differential equations
dx1 dx2
= x2 − x1 (x21 + x22 ), = −x1 − x2 (x21 + x22 ). (3)
dt dt
The linearized equation is
    
d x1 0 1 x1
= ,
dt x2 −1 0 x2

and the eigenvalues of the matrix

 
0 1
−1 0

are ±i. To analyze the behavior of the nonlinear system (3) we multiply the
first equation by x1 , the second equation by x2 and add; this gives
dx1 dx2
x1 + x2 = −x21 (x21 + x22 ) − x22 (x21 + x22 )
dt dt
= −(x21 + x22 )2 .

But
dx1 dx2 1 d 2
x1 + x2 = (x + x22 ).
dt dt 2 dt 1
Hence,
d 2
(x + x22 ) = −2(x21 + x22 )2 .
dt 1
This implies that
c
x21 (t) + x22 (t) = ,
1 + 2ct
where

c = x21 (0) + x22 (0).

Thus x21 (t) + x22 (t) → 0 as t → ∞ for any solution x1 (t), x2 (t) of (3).

3
 The value of x21 + x22 at any time t is always less than its value at t = 0.
We conclude therefore x1 (t) ≡ 0, x2 (t) ≡ 0 is asymptotically stable.
On the other hand, consider the first system of differential equations
dx1 dx2
= x2 + x1 (x21 + x22 ), = −x1 − x2 (x21 + x22 ). (4)
dt dt
Here too, the linearized system is
 
0 1
ẋ = x.
−1 0

But in this case , (d/dt)(x21 + x22 ) = 2(x21 + x22 )2 .

This implies that
c
x21 (t) + x22 (t) = , c = x21 (0) + x22 (0).
1 − 2ct
Notice that every solution x1 (t), x2 (t) of (4) with x21 (0) + x22 (0) 6= 0 approches
infinity in finite time. Therefore we conclude that the equilibrium solution
x1 (t) ≡ 0, x2 (t) ≡ 0 is unstable.

Autonomous Differential Equations

Theorem (1) is also useful in determining the stability of equilibrium solutions
of arbitrary autonomous differential equations. Let x0 be an equilibrium value
of the differential equation
ẋ = f (x) (5)
and set z(t) = x(t) − x0 . Then

ż = ẋ = f (x0 + z) (6)

Clearly z(t) ≡ 0 is an equilibrium solution of (6) and the stability of x(t) ≡ x0

is equivalent to the stability of z(t) ≡ 0.
Lemma 1. Let f (x) have two continuous partial derivatives with respect to each
of its variables x1 , x2 , . . . , xn . Then, f (x0 + z) can be written in the form

f (x0 + z) = f (x0 ) + Az + g(z) (7)

where g(z)/ max{kz1 k, . . . , kzn k} is a continuous function of z which vanishes

for z = 0.
Theorem 1 and Lemma 1 provide us the following useful algorithm for de-
termining the stability of an equilibrium solution x(t) ≡ x0 of ẋ = f (x):

1. Set z = x − x0 .

4
 2. Write f (x0 + z) in the form Az + g(z) where g(z) is a vector-valued poly-
nomial in z1 , . . . , zn beginning with terms of order two or more and the
matrix A is given as the following Jacobean matrix
 ∂f (x ) 
1 (x0 )
1
∂x1
0
· · · ∂f∂x n

A=
 .. .. 
.
 . . 
∂fn (x0 ) ∂fn (x0 )
∂x1 · · · ∂xn

3. Compute the eigenvalues of A. If all the eigenvalues of A have negative

real part, then x(t) ≡ x0 is asymptotically stable. If one eigenvalue of A
has positive real part, then x(t) ≡ x0 is unstable.
Example 1. Find all equilibrium solutions of the system of differential equa-
tions
dx dy
= 1 − xy, = x − y3 (8)
dt dt
and determine whether they are stable or unstable.

Solution The equilibrium solutions of the system (8) can be obtained by solving
the equations x′ (t) = 0, y ′ (t) = 0, which gives 1 − xy = 0 and x − y 3 = 0 i.e.
x = 1, y = 1 or x = −1, y = −1. Hence x ≡ 1, y ≡ 1 or x ≡ −1, y ≡ −1 are
the only two equilibrium solutions of (8).

(i) For stability of first equilibrium solution E1 = (1, 1) we proceed as follows.

Set u = x − 1, v = y − 1. Then,
du dx
= = −u − v − uv
dt dt
dv dy
= = u − 3v − 3v 2 − v 3 .
dt dt
Re-writing the system
      
d u −1 −1 u uv
= − .
dt v 1 −3 v 3v 2 + v 3
 
−1 −1
The matrix has single eigenvalue λ = −2. Hence, the equi-
1 −3
librium solution x(t) ≡ 1, y(t) ≡ 1 of (8) is asymptotically stable.
(ii) For stability of second equilibrium solution E2 = (−1, −1). Set u = x + 1,
v = y + 1. Then,
du dx
= = −u − v − uv
dt dt
dv dy
= = u − 3v + 3v 2 − v 3 .
dt dt

5
 Rewriting the system we have
      
d u 1 1 u −uv
= − .
dt v 1 −3 v 3v 2 − v 3

√
 
1 1
The matrix has eigenvalues λ1 = −1 − 5, which is negative
1 −3
√
and λ2 = −1 + 5, which is positive. Hence, the equilibrium solution
x(t) ≡ −1, y(t) ≡ −1 of (8) is unstable.

Alternate Method:
We can
 do the above said calculation
 directly by calculation the Jacobean matrix
∂f1 (x0 ) ∂f1 (x0 )
· · ·
 ∂x. 1 ∂xn
.. 
A=  .
. .
.

∂fn (x0 ) ∂fn (x0 )
∂x1 · · · ∂xn

For the given example n = 2, and nonlinear functions are given by

f1 (x, y) = 1 − xy, and f2 (x, y) = x − y 3 .

This shows that the given nonlinear system can be rewritten as x′ = Ax + g(x)
where g(x) is a vector-valued polynomial in x1 , . . . , xn beginning with terms of
order two or more and
 
−y −x
A(x, y) = .
1 −3y 2

So to check the stability of nonlinear (weakly) system at the critical points

(x0 , y0 ) we evaluate the eigenvalue of the jacobean matrix A(x0 , y0 ).

1. For stability of first equilibrium solution E1 = (1, 1): Consider

 
−1 −1
A(x0 , y0 ) = A(1, 1) = .
1 −3

Eigenvalues of A(1, 1) are λ1 = −2 = λ2 . Hence, as discussed above the

equilibrium solution x(t) ≡ 1, y(t) ≡ 1 of (8) is asymptotically stable.
2. For stability of second equilibrium solution E2 = (−1, −1): Consider
 
1 1
A(x1 , y1 ) = A(−1, −1) = .
1 −3

We can√ easily check that the jacobean matrix

√ A has eigenvalues λ1 =
−1 − 5, which is negative and λ2 = −1 + 5, which is positive. Hence,
the equilibrium solution x(t) ≡ −1, y(t) ≡ −1 of (8) is unstable.