You are on page 1of 18

Ordinary Differential Equations

Jagmohan Tyagi

Discipline of Mathematics
Indian Institute of Technology Gandhinagar

Ordinary Differential Equations ( ) 1 / 18


Nonlinear Systems
Consider the nonlinear system of the form
x0 = A(t)x + f (t, x), t ≥ 0, (1.1)
where x and f are n-vectors, A(t) is a continuous n × n matrix for t ≥ 0,
and f (t, x) is a continuous function of t and x for t ≥ 0 and kxk < ∞.
Theorem 1
Suppose there exist positive consatnts M and α such that

kΦ(t)Φ−1 (s)k ≤ M e−α(t−s) , for t ≥ s ≥ t0 (1.2)

and let f satisfy the inequality

kf (t, x)k ≤ Kkxk, (1.3)

where K < α/M . Then, every solution x(t) of (1.1) for which
Ordinary Differential Equations ( ) 2 / 18
Theorem 2 (Continuation)
kx(t1 )k < c/M is defined for all t1 ≥ t0 and satisfies

kx(t)k ≤ M e−β(t−t1 ) kx(t1 )k for all t ≥ t1 , (1.4)

where β = α − KM > 0.

Theorem 3
Suppose all the characteristic roots of A have negative real parts and f
satisfies the property

kf (t, x)k = 0(kxk). (1.5)

Then, the zero solution of

x0 = Ax + f (t, x) (1.6)

is asymptotically stable.
Ordinary Differential Equations ( ) 3 / 18
Proof. Let x(t) be a solution of the system (1.6) with x(0) = x0 existing
on the interval 0 ≤ t ≤ ∞. Then, by the variation of constant formula,
we have
Z t
x(t) = Φ(t)x0 + Φ(t − s)f (s, x(s)) ds, (1.7)
0

where Φ(t) is a fundamental matrix of x0 = Ax such that Φ(0) = I.


Since all the characteristic roots of A have negative real parts, there
exist positive numbers R and α such that

kΦ(t)k ≤ Re−αt , t ≥ 0. (1.8)

Relations (1.7) and (1.8) yield


Z t
kx(t)k ≤ Rkx0 ke−αt + R e−α(t−s) kf (s, x(s))k ds. (1.9)
0

From (1.5), it follows that, given any m > 0, there exists a positive
Ordinary Differential Equations ( ) 4 / 18
number d such that, for all t ≥ 0, kxk ≤ d,

kf (t, x)k ≤ mkxk. (1.10)

Thus, as long as kx(t)k ≤ d, it follows, from the inequality (1.9), that


Z t
kx(t)keαt ≤ Rkx0 k + mR eαs kx(s)k ds for 0 ≤ t < ∞. (1.11)
0

By Gronwall-Reid-Bellman inequality, this implies

kx(t)k ≤ Rkx0 kexp((mR − α)t) for 0 ≤ t < ∞. (1.12)

Since x0 and m are arbitrary, we choose m such that mR < α. Then,


kx0 k < d/(2R) implies kx(t)k < d/2 for t ∈ [0, ∞). This means that the
zero solution of (1.6) is stable. Further, the condition mR < α gives the
asymptotic stability of the zero solution of (1.6).
Remark: If all the charactistic roots of λk , (k = 1, 2, . . . , n) of A have
the property max (Reλk ) = −µ, where µ > 0, then every solution x(t)
Ordinary Differential Equations ( ) 5 / 18
of (1.6) which leads to zero as t → ∞ satisfies

log kx(t)k
limt→∞ ≤ −µ. (1.13)
t
To verify this remark, for any given ε = mR > 0, choose α = µ − ε.
Since kx(t)k → 0 as t → ∞, we can make kx(t0 )k sufficiently small for
sufficiently large t0 > 0. Therefore, by Theorem 3, for t ≥ t0 we have

e(α−mR)(t−t0 ) kx(t)k = 0(1) as t → ∞. (1.14)

Since α = µ − ε, where ε = mR, by taking logarithm on both sides, we


get

log kx(t)k
limt→∞ ≤ −µ + 2ε. (1.15)
t
This implies inequality (1.13) as ε > 0 is arbitrary.

Ordinary Differential Equations ( ) 6 / 18


Two Dimensional Systems
Consider the coupled systems
u0 = P (u, v), v 0 = Q(u, v), (1.16)
where u = u(t) and v = v(t) are scalar functions and P, Q, together with
their first and second order partial derivativesm are continuous in some
domain D of the (u, v)-plane containing the origin. Equations (1.16)
constitute an autonomous system since P and Q do not depend upon t.
If x = (u, v), then (1.16) is of the form x0 = f (x) = (P (u, v), Q(u, v))
and our hypotheses on P and Q guarantee the existence and uniqueness
of its solutions.
Consider the Lienard’s equation
u00 + g(u)u0 + h(u) = 0. (1.17)
If we let u0 = v, then this equation takes the form
u0 = v, v 0 = −h(u) − g(u)v. (1.18)
Ordinary Differential Equations ( ) 7 / 18
An important characteristic of autonomous systems is that if (u(t), v(t))
is a solution of (1.16) existing on (a, b) then (u(t − a), v(t − a)) is a
solution of (1.16) on (a + α, b + α) for any real number α. However, this
property is not true for nonautonomous system.
Example: For the nonautonomous system

u0 = u, v 0 = tu, (1.19)

u(t) = et , v(t) = tet − et is a solution; but v 0 (t + α) = (t + α)et+α 6=


tu(t + α) unless α = 0.
From the property of (1.16), it is clear that a solution (u(t), v(t)) of (1.16)
describes parametrically a curve in D with t as a parameter. This curve
is called an orbit or a trajectory or a path of (1.16). It follows from the
uniqueness of solution of (1.16) that through a given point there passes
one and only trajectory. Our discussion here clearly indicates that a
trajectory is a curve in D that is represented parametrically by more
than one solution.

Ordinary Differential Equations ( ) 8 / 18


Example: (u(t), v(t)) and (u(t + α), v(t + α)), α 6= 0 are distinct solution
of (1.16) but represent the same curve parametrically. In particular, the
functions u(t) = sin(t + α), v(t) = cos(t + α), −∞ < t < ∞, α ∈ (0, 2π),
describe an infinite number of distinct solutions of the system u0 = v,
v 0 = −u but represent the same trajectory, that is, circle u2 + v 2 = 1.
Definition 1
Any point (u0 , v0 ) ∈ D at which both P and Q vanish simultaneously is
called a critical point of (1.16). Such a point is also referred to as a
point of equilibrium or stationary point or rest point.
A critical point (u0 , v0 ) is said to be isolated if there exits no other
critical point in some neighbourhood of (u0 , v0 ). By a critical point, we
shall hereafter mean isolated critical point.

Example: Consider the simple undamped pendulum equation


g
u00 + sin u = 0. (1.20)
L

Ordinary Differential Equations ( ) 9 / 18


The autonomous system corresponding to this equation is
g
u0 = v v0 = − sin u. (1.21)
L
The critical points of this system are (nπ, 0), where n = 0, ±1, ±2, . . . ,
and D is the whole (u, v)-plane.
If R(u, v) = (P (u, v), Q(u, v)) with (u, v) ∈ D is a field of vectors, then
(1.16) describes the motion of a particle (u, v) whose velocity (u0 , v 0 ) at
every in D is given by R(u, v). The fixed path along which this particle
moves independent of its starting point is called the trajectory, and a
critical point is a point of equilibrium. Viewed in this way, we shall call
D the phase space of (1.16).

Ordinary Differential Equations ( ) 10 / 18


Two Dimensional Linear Autonomous Systems
Consider the two dimensional real linear system
x0 = Ax, (1.22)
where
   
x1 a11 a12
x= A= , (1.23)
x2 a21 a22
a11 , a12 , a21 , a22 are constants, and det A 6= 0. Clearly, (x1 , x2 ) = (0, 0)
is the only critical point of (1.22). Using the Jordon canonical form, we
shall examine the qualitative behaviour of the trajectories of (1.22). To
do this, let us consider the transformation x = T y, where T is a real
constant nonsingular matrix. Now, (1.22) reduces to
y 0 = Dy, (1.24)
where D = T −1 AT . ALthough the nonsingular transformation matrix T
distorts the shape of the phase portrait of (1.22) yet the basic character
Ordinary Differential Equations ( ) 11 / 18
of the phase portrait remains unchanged. Also, the critical point (0, 0)
of (1.22) is the critical point of the canonical system (1.24). Moreover,
the matrices A and D are similar, and hence they have the same charac-
teristic polynomials. Therefore, without any loss of generality, we migth
as well study the qualitative behaviour of the solutions of (1.22) around
its critical point (0, 0) in the canonical form (1.24). Let λ1 and λ2 be the
characteristic roots of A. We now examine the various cases.
Case 1: (real and distinct roots).
Here it follows that
 
λ1 0
D= , (1.25)
0 λ2
and hence (1.24) takes the form
y10 = λ1 y1 , y20 = λ2 y2 . (1.26)
Therefore, the solution of system (1.26) through (y1 (0), y2 (0)) = (c1 , c2 ) 6=
(0, 0) is y1 (t) = c1 eλ1 t , y2 (t) = c2 eλ2 t . We shall discuss each possibility
seperately.
Ordinary Differential Equations ( ) 12 / 18
Both λ1 and λ2 are nonzero and have the same sign. From (1.26), it
follows that
dy2 λ 2 y2
= , (1.27)
dy1 λ 1 y1
λ /λ
and hence y2 = cy1 2 1 , where c is an arbitrary constant. If λ2 < λ1 < 0,
then (y1 (t), y2 (t)) → (0, 0) as t → ∞. Thus, every orbit tends to the
origin as t → ∞. Therefore, the critical point (0, 0) is asymptotically
stable. The phase portrait is as shown in Fig. 3.5.1; the arrows indicate
the direction of increasing t. If λ1 > λ2 > 0, then the critical point (0, 0)
is unstable and every orbit goes away from the origin as t increases. The
phase portrait for this possibility would be as in Fig. 3.5.2. The origin
(0,0) in Fig. 3.5.1. and Fig. 3.5.2. corresponding to the possibilities is
called an improper node.

Ordinary Differential Equations ( ) 13 / 18


Ordinary Differential Equations ( ) 14 / 18
Ordinary Differential Equations ( ) 15 / 18
Ordinary Differential Equations ( ) 16 / 18
Both λ1 and λ2 are nonzero and have a different sign. Let λ1 < 0 and
λ2 > 0. Then, y1 (t) → 0 as t → ∞, and y2 (t) → ±∞ as t → ∞ according
as c2 > 0 or c2 < 0. If |λ1 | = |λ2 |, then it is easy to see that the orbits
would be rectangular hyperbolas. For arbitrary λ1 < 0 and λ2 < 0, the
phase portrait will have the form shown in Fig. 3.5.3. The critical point
(0,0) is called a saddle point.
One root is zero and the other is nonzero. If λ1 = 0 and λ2 < 0, then
y1 (t) = c1 and y2 (t) → 0 as t → ∞. Also, if λ1 = 0 and λ2 > 0, then
y1 (t) = c1 , and y2 (t) → ±∞ as t → ∞ according as c2 > 0 or c2 < 0.
The origin (0,0) is called a Proper node. The phase portrait for these
possibilities would be as in Fig. 3.5.4 and Fig 3.5.5.

Ordinary Differential Equations ( ) 17 / 18


Ordinary Differential Equations ( ) 18 / 18

You might also like