Chapter 4. Systems of ODEs.

Phase Plane 1
§4.1 Systems of ODEs as Models
Example 1. (Mixing Problem Involving Two Tanks)
Initially, tank T1: 100 gal of pure water,
tank T2: 100 gal of water dissolved 150 lb of fertilizer.
By circulating liquid at a rate of 2 gal/min and stirring the amounts of
fertilizer y1(t) in T1 and y2(t) in T2 change with time t.
How long should we let the liquid circulate so that T1 will contain at least
half as much fertilizer as there will be left in T2?

Fig. 78. Fertilizer contents in Tank T1 (lower curve) and T2

Sol. Step 1. Setting up the model
2 2
y1′ = Inflow/min − Outflow/min = y2 − y1 (Tank T1)
100 100
2 2
y2′ = Inflow/min − Outflow/min = y1 − y2 (Tank T2)
100 100
y1(0) = 0, y2(0) = 150.

system of first-order ODEs
 ′
 y1 = −0.02y1 + 0.02y2
y2′ = 0.02y1 − 0.02y2

 y (0) = 0, y (0) = 150
1 2
matrix form
y′ = Ay, y = [0 150]T
where
[ ] [ ]
y1 −0.02 0.02
y= , A= .
y2 0.02 −0.02
Computational Science & Engineering (CSE) T. Jeong
 Chapter 4. Systems of ODEs. Phase Plane 2

Step 2. General solution

Try
y = xeλt.
Then y′ = λxeλt = Axeλt and
Ax = λx.
−0.02 − λ 0.02
det(A − λI) =
0.02 −0.02 − λ

= λ(λ + 0.04) = 0, λ = 0, −0.04.

[ ] [ ]
1 1
λ1 = 0 : x(1) = , λ2 = −0.04 : x(2) = .
1 −1
We obtain a solution (called a general solution)
y = c1x(1)eλ1t + c2x(2)eλ2t
[ ] [ ]
1 1
= c1 + c2 e−0.04t.
1 −1
Step 3. Use [of initial
] conditions
[ ] [ ]
1 1 0
y(0) = c1 + c2 = ⇒ c1 = 75, c2 = −75.
1 −1 150
[ ] [ ]
1 1
∴ y = 75 − 75 e−0.04t.
1 −1
Thus, y1 = 75 − 75e−0.04t and y2 = 75 + 75e−0.04t.

Step 4. Answer
1
y1 = y2, y1 + y2 = 150 ⇒ y1(t) = 50.
2
−0.04t −0.04t 1
y1 = 75 − 75e = 50 ⇒ e = .
3
ln 3
∴ t= = 27.5.
0.04
See Fig. 78.
Computational Science & Engineering (CSE) T. Jeong
 Chapter 4. Systems of ODEs. Phase Plane 3

Example 2. (Electrical Network)

Assume that all currents and charges to be zero at t = 0, that is,
I1(0) = I2(0) = 0, Q(0) = 0.
Find the currents I1(t) and I2(t) in the network in Fig. 78.

Fig. 79. Electrical network in Example 2

Sol. Step 1. Setting up the model
In the left loop, [voltage drops: LI1′ = I1′ , R1(I1 − I2) = 4(I1 − I2)]
I1′ + 4(I1 − I2) = 12 ⇒ I1′ = −4I1 + 4I2 + 12.
In the right loop, [voltage drops: R2I2 = 6I2, R1(I2 −I
∫ 1) = 4(I∫2 −I1),
(1/C) I2dt = 4 I2dt]
∫
6I2 + 4(I2 − I1) + 4 I2dt = 0

⇒ I2′ = 0.4I1′ − 0.4I2 = −1.6I1 + 1.2I2 + 4.8.

matrix form
J′ = AJ + g
[ ] [ ] [ ]
I1 −4.0 4.0 12.0
where J = , A= , g= .
I2 −1.6 1.2 4.8
Step 2. Solving
Consider
J′ = AJ.
Try J = xeλt. Then
J′ = λxeλt = Axeλt =⇒ Ax = λx.

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 4

Eigenvalues and eigenvectors of A:

[ ] [ ]
2 1
λ1 = −2 : x(1) = , λ2 = −0.8 : x(2) = .
1 0.8
A general solution of the homogeneous system is
[ ] [ ]
2 1
Jh = c1 e−2t + c2 e−0.8t.
1 0.8
[ ]
a
Set Jp = , then 0 = J′p = AJp + g.
b
{ [ ]
−4a + 4b + 12 = 0 3
⇒ a = 3, b = 0, Jp = .
−1.6a + 1.2b + 4.8 = 0 0
A general solution of the system is
[ ] [ ] [ ]
2 1 3
J = Jh + J p = c 1 e−2t + c2 e−0.8t + .
1 0.8 0
The initial conditions I1(0) = 0, I2(0) = 0 give c1 = −4, c2 = 5.
Hence the solution is
[ ] [ ] [ ]
2 1 3
J = −4 e−2t + 5 e−0.8t + ,
1 0.8 0
that is, {
I1 = −8e−2t + 5e−0.8t + 3
.
I2 = −4e−2t + 4e−0.8t

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 5

[Conversion of an nth-order ODE to a system]

Consider an nth-order ODE
y (n) = F (t, y, y ′, · · · , y (n−1)).
Setting
y1 = y, y2 = y ′, y3 = y ′′, · · · , yn−1 = y (n−2), yn = y (n−1),
the ODE is converted to a system


 y ′
= y2

 1





 y2′ = y3





...








′
yn−1 = yn






 y ′ = F (t, y , y , · · · , y ).
n 1 2 n

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 6

Example 3. Mass on a spring

By conversion of the ODE to a system, solve the ODE
my ′′ + cy ′ + ky = 0, m = 1, c = 2, k = 0.75.

Sol. The ODE is

y ′′ + 2y ′ + 0.75y = 0.
Set y1 = y, y2 = y ′, then
y1′ = y2
y2′ = −0.75y1 − 2y2.
matrix form [ ][ ]
0 1 y1
y′ = Ay =
−0.75 −2 y2

−λ 1
det(A − λI) =
−0.75 −2 − λ

= λ2 + 2λ + 0.75 = 0
∴ λ = −0.5, − 1.5
[ ] [ ]
2 1
λ1 = −0.5 : x(1) = , λ2 = −1.5 : x(2) = .
−1 −1.5
Hence [ ] [ ]
2 1
y = c1 e−0.5t + c2 e−1.5t
−1 −1.5
and the solution is
y = y1 = 2c1e−0.5t + c2e−1.5t.
Note that
y2 = y1′ = y ′ = −c1e−0.5t − 1.5c2e−1.5t.

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 7

§4.2 Basic Theory of Systems of ODEs. Wronskian

First-Order System


 y1′ = f1(t, y1, · · · , yn)



 y ′
= f2(t, y1, · · · , yn)

 2
... (1)

 yn′ = fn(t, y1, · · · , yn)






y1(t0) = K1, · · · , yn(t0) = Kn (2)
matrix form
y′ = f (t, y), y(t0) = K
where

y = [y1 · · · yn]T , f = [f1 · · · fn]T , K = [K1 · · · Kn]T .

A solution of (1) on (a, b) is a set of n differentiable functions
y1 = h1(t), · · · , yn = hn(t), a<t<b
that satisfy (1) throughout (a, b).
 
h1
 .. 
For the solution vector h =  . , we can write
hn

y = h(t).

Theorem 1. Existence and Uniqueness Theorem

Let f1, · · · , fn in (1) be continuous functions having continuous partial
∂f1
derivatives ∂y , · · · , ∂y
∂f1
n
, · · · , ∂fn
∂yn
in some domain R of ty1 · · · yn-space
1
containing (t0, K1, · · · , Kn).
Then (1) has a solution on some interval t0 − α < t < t0 + α satisfying
the initial condition (2), and this solution is unique.

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 8

Linear System

 ′
 y1 = a11(t)y1 + · · · + a1n(t)yn + g1(t)
··· (3)

 y ′ = a (t)y + · · · + a (t)y + g (t)
n n1 1 nn n n

vector form,
     
a11 · · · a1n y1 g1
′  .. . . .  
..  , y =  ..  , g =  ... 
.  
y = Ay + g, A =  . . 
an1 · · · ann yn gn
If g = 0, then this system is homogeneous.
y′ = Ay
If g ̸= 0, then this system is nonhomogeneous.

Theorem 2. Existence and Uniqueness in the linear case

Let ajk ’s and gj ’s in (3) be continuous functions of t on (a, b) containing
t = t0. Then (3) has a solution y(t) on this interval satisfying (2), and
this solution is unique.
Theorem 3. Superposition Principle
If y(1) and y(2) are solutions of the homogeneous linear system
y′ = Ay on some interval I,
so is y = c1y(1) + c2y(2)(a linear combination of y(1), y(2)).
Proof. Let y(1) and y(2) be solutions of y′ = Ay on I.
Insert y = c1y(1) + c2y(2) into the system, we obtain
′ ′
y′ = c1y(1) + c2y(2)
= c1Ay(1) + c2Ay(2)
= A(c1y(1) + c2y(2))
= Ay.
Thus y = c1y(1) + c2y(2) is a solution of y′ = Ay on I.
Computational Science & Engineering (CSE) T. Jeong
 Chapter 4. Systems of ODEs. Phase Plane 9

[Basis, General solution, Wronskian]

Let y(1), y(2), · · · , y(n) be solutions of
y′ = Ay (4)
on some interval I.
If y(1), y(2), · · · , y(n) are linearly independent on I, they form a basis
or fundamental system of solutions on I. Their linear combination of
y(1), y(2), · · · , y(n)
y = c1y(1) + c2y(2) + · · · + cny(n) (5)
is called a general solution of y′ = Ay of the system (4) on I.

Notation:
1.
 (1) (2) (n)

y1 ··· y1 y1
 (1) (2) (n) 
 y2 ··· y2 y2 
Y = [y(1), y(2), · · · , y(n)] =  . ... ... ... .
 .. 
yn(1) yn(2) · · · yn(n)

2. Wronskian of y(1), y(2), · · · , y(n) = det(Y ),

that is,
(1) (2) (n)
y1 y1 · · · y 1
(1) (2) (n)
W (y , y , · · · , y ) =
(1) (2) (n) y 2 y 2 · · · y 2 = |Y |.
... ... ... ...
yn(1) yn(2) · · · yn(n)
Note 1. We can rewrite y = c1y(1) + c2y(2) + · · · + cny(n) as
 
c1
 
(n)  c2 
y = [y , y , · · · , y ]  .. 
(1) (2)
 . 
cn
= Y c.

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 10

2. y(1), y(2), · · · , y(n) form a basis on I

⇐⇒ W = det(Y ) ̸= 0 at any t ∈ I.

3. (Sec. 2.6)
Let y and z be solutions of a second order homogeneous linear ODE.
Then their Wronskian is
y z
W (y, z) = .
y′ z′

(1) ′ (2) ′
Set y = y1 , y ′ = y1 = y2 and z = y1 , z ′ = y1
(1) (1) (2) (2)
= y2 .
Then
[ ] [ ]
(1)
y1 y
y(1) = = ,
(1)
y2 y′
[ ] [ ]
(2)
y1 z
y(2) = =
(2)
y2 z′

and
(1) (2)
y z y1 y 1 (1)
W (y, z) = ′ ′ = (1) (2) = W (y , y(2)).
y z y2 y 2

4. If y(1), y(2), · · · , y(n) form a basis on I, then Y is called a Funda-

mental matrix.

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 11

§4.3 Constant-coefficient systems. Phase plane method.

Consider a homogeneous linear system
y′ = Ay (1)
where A = [ajk ], ajk are constants.
Try
y = xeλt.
Substitution into (1) gives y′ = λxeλt = Axeλt, and
Ax = λx.
Let x(1), · · · , x(n) be eigenvectors corresponding to eigenvalues λ1, · · · , λn,
respectively. Then the corresponding solutions are
y(1) = x(1)eλ1t, y(2) = x(2)eλ2t, · · · , y(n) = x(n)eλnt,
and their Wronskian is
(1) (2) (n)
x1 eλ1t x1 eλ2t · · · x1 eλnt
(1) (2) (n)
W (y(1), · · · , y(n)) = x2 eλ1t x2 eλ2t · · · x2 eλnt
... ... ... ...
x(1)
n e
λ1 t
x(2)
n e
λ2 t
· · · x(n)
n e
λn t

(1) (2) (n)

x1 x1 · · · x1
(1) (2) (n)
= eλ1t+λ2t+···+λnt x2 x2 · · · x2 .
... ... ... ...
x(1)
n x(2)
n · · · x(n)
n

If A has the n linearly independent eigenvectors x(1), x(2) · · · , x(n)

(rank(x(1) x(2) · · · x(n)) = n), then W ̸= 0.
Ex. 1. A has n-distinct eigenvalues.
2. AT = A(symmetric)
3. AT = −A(skew-symmetric)
4. etc
Thus y(1), y(2), · · · , y(n) are linearly independent on I.
Computational Science & Engineering (CSE) T. Jeong
 Chapter 4. Systems of ODEs. Phase Plane 12

Theorem 1. (general solution)

If the constant matrix A in (1) has a linearly independent set of n eigen-
vectors, then the corresponding solutions
y(1) = x(1)eλ1t, · · · , y(n) = x(n)eλnt
form a basis and the corresponding general solution is
y = c1x(1)eλ1t + · · · + cnx(n)eλnt.

[How to graph solutions in the phase plane]

We consider
{ ′
y1 = a11y1 + a12y2
′ , y′ = Ay. (2)
y2 = a21y1 + a22y2

Solution of the system [ ]

y1(t)
y(t) =
y2(t)
⇐⇒ two curves over the t-axis.

We can graph a solution (y1(t), y2(t)) as a single curve in the y1y2-plane.

This is a parametric representation(parametric equation) with parameter
t. Such a curve is called a trajectory (or orbit, path) of (2), and the
y1y2-plane is called the phase plane.

If we fill the phase plane with trajectories of (2), we obtain the so-called
phase portrait.

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 13

Example 1.
Find and graph solutions of the system
[ ]
−3 1
y′ = Ay = y.
1 −3

Sol. Substitute y = xeλt, and |A − λI| = λ2 + 6λ + 8 = 0.

[ ] [ ]
1 1
λ1 = −2 : x(1) = , λ2 = −4 : x(2) = .
1 −1
[ ] [ ] [ ]
y1 1 1
∴ y= = c1 e−2t + c2 e−4t
y2 1 −1
or
y1(t) = c1e−2t + c2e−4t,
y2(t) = c1e−2t − c2e−4t.
If c1 = 0, then y2 = −y1, and if c2 = 0 then y2 = y1.
If t is sufficiently large, then y1, y2 ≈ c1e−2t, and y1, y2 → 0 as t → ∞.

Fig. 82. Trajectories of the system(Improper node)

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 14

[Critical points of the system]

{
y1′ = a11y1 + a12y2
, y′ = Ay (3)
y2′ = a21y1 + a22y2

Assume that |A| ̸= 0. From (3)

dy2 y2′ dt y2′ a21y1 + a22y2
= = = .
dy1 y1′ dt y1′ a11y1 + a12y2
This gives a unique tangent direction of trajectory passing through P :
dy2
(y1, y2), except for P0 : (0, 0). The point P0, at which dy becomes
1
undetermined, is called a critical point of (3).

[Five types of critical points]

Example 1. (Improper Node)
An improper node is a critical point P0 at which all the trajectories,
except for two of them, have the same limiting direction of tangent. The
two exceptional trajectories also have a limiting direction of the tangent at
P0 which is different.
See Example 1.

Fig. 82. Trajectories of the system(Improper node)

Note two (negative) real eigenvalues

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 15

Example 2. (Proper Node)

A proper node is a critical point P0 at which every trajectory has a definite
limiting direction and for any given direction d at P0 there is a trajectory
having d as its limiting direction.
Ex. [ ]
1 0
y′ = Ay = y.
0 1
Substitute y = xeλt, and |A − λI|[ = ](1 − λ)2 =[ 0. ]
1 0
λ = λ1 = λ2 = 1 : x(1) = , x(2) = .
0 1
(Any x ̸= 0 is an eigenvector.)
[ ] [ ]
1 0
∴ y = c1 et + c2 et
0 1
{
y1 = c1et
⇐⇒ ⇐⇒ c1y2 = c2y1.
y2 = c2et

Fig. 83. Trajectories of the system(Proper node)

Note (positive) real eigenvalues.

Computational Science & Engineering (CSE) T. Jeong
 Chapter 4. Systems of ODEs. Phase Plane 16

Example 3. (Saddle Point)

A saddle point is a critical point P0 at which there are two incoming trajec-
tories, two outgoing trajectories, all the other trajectories in a neighborhood
of P0 bypass P0.
Ex. [ ]
1 0
y′ = Ay = y.
0 −1
Substitute y = xeλt, and |A − λI| = λ2 − 1 = 0.
[ ] [ ]
1 0
λ1 = 1 : x(1) = , λ2 = −1 : x(1) = .
0 1
[ ] [ ]
1 0
∴ y = c1 et + c2 e−t
0 1
{
y1 = c1et
⇐⇒
y2 = c2e−t

⇐⇒ y1y2 = c1c2 = const.

Fig. 84. Trajectories of the system(Saddle point)

Note a negative real eigenvalue and a positive real eigenvalue.

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 17

Example 4. (Center)
A center is a critical point that is enclosed by infinitely many closed trajec-
tories.
Ex. [ ]
0 1
y′ = Ay = y.
−4 0
Substitute y = xeλt, and [ |A −
] λI| = λ 2
+ 4 = 0. [ ]
1 1
λ1 = 2i : x(1) = , λ2 = −2i : x(2) = .
2i −2i
[ ] [ ]
1 1
∴ y = c1 e2it + c2 e−2it
2i −2i
{
y1 = A cos 2t + B sin 2t
⇐⇒
y2 = 2B cos 2t − 2A sin 2t
or
{ ′ {
y1 = y 2 4y1y1′ = 4y1y2
=⇒
y2′ = −4y1 y2y2′ = −4y1y2

=⇒ 4y1y1′ + y2y2′ = 0
2 1 2
=⇒ 2y1 + y2 = const.
2

Fig. 85. Trajectories of the system(Center)

Note pure imaginary eigenvalues.

Computational Science & Engineering (CSE) T. Jeong
 Chapter 4. Systems of ODEs. Phase Plane 18

Example 5. (Spiral Point)

A spiral point is a critical point P0 about which the trajectories spiral,
approaching P0 as t → ∞.
[ ]
−1 1
Ex. y′ = Ay = y.
−1 −1
Substitute y = xeλt, and |A [− λI| ] = λ 2
+ 2λ + 2 = 0. [ ]
1 1
λ1 = −1 + i : x(1) = , λ2 = −1 − i : x(2) = .
i −i
[ ] [ ]
1 1
∴ y = c1 e(−1+i)t + c2 e(−1−i)t
i −i
{
y1 = e−t[A cos t + B sin t]
⇐⇒
y2 = e−t[B cos t − A sin t]
{ ′
y1 = −y1 + y2 × y1
Or ′ =⇒ y1y1′ + y2y2′ = −(y12 + y22).
y2 = −y1 − y2 × y2
Let r 2 = y12 + y22.
Then 2rr ′ = 2y1y1′ + 2y2y2′ = −2(y12 + y22) = −2r 2 and
r ′ = −r.
By integration, we get
ln r = −t + c̃, r = ce−t(or y12 + y22 = c2e−2t).

Fig. 86. Trajectories of the system(Spiral point)

Note complex eigenvalues (a ± bi, a, b ̸= 0).

Computational Science & Engineering (CSE) T. Jeong
 Chapter 4. Systems of ODEs. Phase Plane 19

Example 6. (Degenerate Node)

Find and graph a general solution of
[ ]
4 1
y′ = Ay = y.
−1 2

Sol. |A − λI| = λ2 [− 6λ ]+ 9 = 0, λ = 3(double), x(1) = [1, −1]T .

1
one solution : y(1) = e3t.
−1
Set another solution y(2) = xteλt + ueλt, u = [u1 u2]T , and
′
y(2) = xeλt + λxteλt + λueλt = Axteλt + Aueλt.
Thus
Ax = λx, (A − λI)u = x.
Since x is an eigenvector of A corresponding to the eigenvalue λ, λ = 3
and take
x = [1 − 1]T .
From (A − λI)u = x, we have
[ ] [ ] [ ]
u1 + u2 = 1 u1 1 0
⇒ u= = u1 + .
−u1 − u2 = −1 1 − u1 −1 1
Take u1 = 0, then
([ 1 ] [ ])
0
y(2) = t+ e3t.
−1 1
Thus
y = c1y(1) + c2y(2)
[ ] ([ 1 ] [ ])
1 0
= c1 e3t + c2 t+ e3t
−1 −1 1
or
y1 = (c1 + c2t)e3t, y2 = (−c1 + c2(1 − t))e3t.
Note The critical point at the origin is called a degenerate nod.
Computational Science & Engineering (CSE) T. Jeong
 Chapter 4. Systems of ODEs. Phase Plane 20

Fig. 87. Degenerate node in Example 6

Note Consider
y′ = Ay.
1. Let λ be a double eigenvalue of A with only one linearly independent
eigenvector x. The system has a solution
y(1) = xeλt.
One gets another (linearly independent) solution from
y(2) = xteλt + ueλt.
2. Let λ be a triple eigenvalue of A with only one linearly independent
eigenvector x. The system has a solution
y(1) = xeλt.
One gets another two (linearly independent) solutions from
y(2) = xteλt + ueλt,
1
y(3) = xt2eλt + uteλt + veλt.
2

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 21

§4.4 Criteria for Critical Points. Stability

We consider
{ ′
y1 = a11y1 + a12y2
, y′ = Ay (det A ̸= 0). (1)
y2′ = a21y1 + a22y2
Let y(t) = [y1(t), y2(t)]T be a solution of the system (1).
We can graph solutions (y1(t), y2(t)) as curves in the y1y2-plane (called
the phase plane). Such a curve is called a trajectory of (1), and the
totality is the phase portrait of (1).
dy2 y2′ a21y1 + a22y2
= = .
dy1 y1′ a11y1 + a12y2
This gives a unique tangent direction of trajectory passing through P , except
for P0 : (0, 0). The point P0, at which dy 2
dy1
becomes undetermined, is called
a critical point of (1).
Let λ = λ1, λ2 be solutions of the characteristic equation
a11 − λ a12
det(A − λI) = = λ2 − pλ + q = 0,
a21 a22 − λ
where p = a11 + a22 = Trace(A), q = a11a22 − a12a21 = detA.
Let ∆ = p2 − 4q. Note that p = λ1 + λ2, q = λ1λ2.

Table 4.1 Eigenvalue Criteria for Critical Points

Note
(a) q > 0, △ ≥ 0 → λ1, λ2 > 0 or λ1, λ2 < 0 : Node.
(b) q < 0(→ △ > 0) → if λ1 < λ2, λ1 < 0 < λ2: Saddle point.
√
(c) p = 0, q > 0(→ △ < 0) → λ1, λ2 = ± qi : Center.
√
(d) p ̸= 0, △ < 0 → λ1, λ2 = (p ± i −∆)/2 : Spiral point.

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 22

Definition
1. A critical point P0 of (1) is called stable if all trajectories of (1) that
at some instant are close to P0 remain close to P0 at all future times.
2. P0 is called unstable if P0 is not stable.
3. P0 is called stable and attractive (or asymptotically stable)if P0
is stable and every trajectory that has a point in Dδ approaches as t → ∞.

Table 4.2 Stability Criteria for Critical Points

Note
(a) p < 0, q > 0 → λ1, λ2 < 0 or Re λ1, Re λ2 < 0
: stable and attractive.
(b) p ≤ 0, q > 0 → λ1, λ2 < 0 or Re λ1, Re λ2 ≤ 0 : stable.
(c) p > 0 or q < 0 → ∃λ s.t. λ > 0 or Re λ > 0 : unstable.

Fig. 92. Stability chart of the system with p, q, ∆

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 23

Example 1 Application of the Criteria

Determine the type and stability of the critical point.
[ ]
−3 1
y′ = y.
1 −3

Sol. p = −6, q = 8, △ = 4 (λ = −2, −4) → node, stable and

attractive.

Example 2 Free Motions of a Mass on a Spring

What kind of critical point does
my ′′ + cy ′ + ky = 0
in Sec. 2.4 have?
Sol. Set y1 = y, y2 = y ′. Then y1′ = y2, y2′ = −k/my1 − c/my2,
hence [ ]
0 1
y′ = y.
−k/m −c/m
p = −c/m, q = k/m, △ = p2 − 4q = (c2 − 4mk)/m2.

No damping : c = 0, p = 0, q > 0 center, stable.

Under-damping : c2 < 4mk, p < 0, q > 0, △ < 0
spiral point, stable and attractive.
Critical damping : c2 = 4mk, p < 0, q > 0, △ = 0
node, stable and attractive.
Over-damping : c2 > 4mk, p < 0, q > 0, △ > 0
node, stable and attractive.

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 24

§4.5 Qualitative methods for nonlinear systems

(up to p.155)
• Qualitative methods : methods of obtaining qualitative information
on solutions without actually solving a system.
• Autonomous nonlinear system(t does not occur explicitly)
{ ′
y1 = f1(y1, y2)
y′ = f (y), (1)
y2′ = f2(y1, y2).

Critical points (y1, y2) of (1) :

f1(y1, y2) = 0, f2(y1, y2) = 0.
(1) may have several critical points. When P0 : (a, b) is the critical point,
by a translation,
ỹ1 = y1 − a, ỹ2 = y2 − b,
we can assume P0 to be (0, 0). (Assume that P0 is isolated.)

[Linearization of nonlinear systems]

How can we determine the kind and stability property of a critical point
P0 : (0, 0) of the nonlinear system (1)?
• Linearization of the nonlinear system(1) at P0
{ ′
y1 = a11y1 + a12y2 + h1(y1, y2)
y′ = Ay + h(y),
y2′ = a21y1 + a22y2 + h2(y1, y2).
{ ′
y1 = a11y1 + a12y2
=⇒ y′ = Ay,
y2′ = a21y1 + a22y2.
Use tables of Sec 4.4.
Note [Taylor series of f (x, y) at (a, b)]
f (x, y) = f (a, b) + [fx(a, b)(x − a) + fy (a, b)(y − b)]
1[
+ fxx(a, b)(x − a)2 + 2fxy (a, b)(x − a)(y − b)
2! ]
+fyy (a, b)(y − b) + · · · .
2

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 25

Example 1. Free Undamped Pendulum. Linearization

Figure shows a pendulum consisting of a body of mass m and a rod of length
L. Determine the locations and types of the critical points.

Sol. Setting up the mathematical model

restoring force: mg sin θ.
force of acceleration: mass × acceleration = mLθ ′′.
mLθ ′′ = −mg sin θ
=⇒ θ ′′ + k sin θ = 0, k = g/L.

Note When θ is very small, sin θ ≈ θ. √ √

approximate solution θ = A cos kt + B sin kt.

Set θ = y1, θ ′ = y2, then

y1′ = y2,
y2′ = −k sin y1.
Critical points
y2 = 0, −k sin y1 = 0
⇒ critical points : (nπ, 0), n = 0, ±1, ±2, · · · .
Computational Science & Engineering (CSE) T. Jeong
 Chapter 4. Systems of ODEs. Phase Plane 26

Critical points (2nπ, 0)

Linearized system at (0, 0)
Note that sin y1 = y1 − 3! y1 + · · · ≈ y1 .
1 3

{ ′ [ ]
y1 = y2 0 1
′ y′ = Ay = y.
y2 = −ky1, −k 0
p = 0, q = k > 0, ∆ = −4k < 0
or
1 k
ky1y1′ + y2y2′ = 0 ⇒ y12 + y22 = const.
2 2
Hence the critical points (2nπ, 0) are all centers and stable.

Critical points ((2n + 1)π, 0)

Linearized system at (π, 0)
Set θ − π = y1 and y1′ = (θ − π)′ = θ ′ = y2.
sin θ = sin(y1 + π) = − sin y1 = −y1 + 3! y1 + · · · ≈ −y1.
1 3

{ ′ [ ]
y1 = y2 0 1
′ y′ = Ay = y
y2 = ky1, k 0
p = 0, q = −k < 0, ∆ = 4k > 0
or ky1y1′ − y2y2′ = 0 ⇒ k2 y12 − 12 y22 = const.
Hence the critical points ((2n + 1)π, 0) are all saddle points and unstable.

Fig. 93(b). Solution curves y2 (y1 ) in the phase plane

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 27

§4.6 Nonhomogeneous Linear Systems of ODEs

y′ = Ay + g (1)
y′ = Ay (2)
A general solution of (1) is
y = y(h) + y(p),
where y(h) is a general solution of (2) and y(p) is a particular solution of (1).

How to find a particular solution y(p)

1. Method of Undetermined Coefficients

: A particular solution y(p) is assumed in a form similar to g.
(Ex. y(p) = xeλt, u + vt + wt2, · · · )

2. Method of Variation of Parameters

: When a general solution y(h) of y′ = Ay is known,
we can obtain a particular solution
y(p) = Y(t)u(t),
where Y (t) is the fundamental matrix(obtained from y(h)).

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 28

[Method of Undetermined Coefficients]

Example 1.
Find a general solution of
[ ] [ ]
−3 1 −6
y′ = Ay + g = y+ e−2t.
1 −3 2
′
Sol. The general solution [of homogeneous
] system
[ ]y = Ay is
1 1
y(h) = c1 e−2t + c2 e−4t.
1 −1
Put y(p) = ute−2t + ve−2t. (why)
′
y(p) = ue−2t − 2ute−2t − 2ve−2t
= Aute−2t + Ave−2t + g
[ ]
−6
=⇒ Au = −2u and u − 2v = Av + .
2
Clearly, u is an
[ eigenvector
] of A corresponding to the eigenvalue −2.
1
Thus u = a , a ̸= 0 and
1
[ ] [ ] [ ] [ ]
a v1 −3v1 + v2 −6
−2 = + .
a v2 v1 − 3v2 2
Collecting terms and reshuffling gives
{
v1 − v2 = −a − 6
and − a − 6 = a − 2,
−v1 + v2 = −a + 2
thus a = −2, v2 = v1 + 4 and v1 = k, v2 = k + 4.
Take k = 0(why), then a particular solution is
[ ] [ ]
1 0
y(p) = −2 te−2t + e−2t.
1 4
The general
[ solution
] is [ ] [ ] [ ]
1 1 1 0
y = c1 e−2t + c2 e−4t − 2 te−2t + e−2t.
1 −1 1 4

Computational Science & Engineering (CSE) T. Jeong

 Chapter 4. Systems of ODEs. Phase Plane 29

[Method of Variation of Parameters]

Example 2.
Solve [ ] [ ]
−3 1 −6
y′ = Ay + g = y+ e−2t.
1 −3 2
′
Sol. The general solution of the
[ homogeneous
] system
[ ]y = Ay is
1 1
y(h) = c1 e−2t + c2 e−4t.
1 −1
The solution y(h) may be written
[ −2t −4t
][ ]
e e c1
y(h) = −2t −4t = [y(1) y(2)] ≡ Y (t)c.
e −e c2
Replace c by u(t), and set y(p) = Y (t)u(t).
Substitution into the ODE gives
Y ′u + Y u′ = AY u + g (1)
′ ′
Since y(1) = Ay(1) and y(2) = Ay(2), we have Y ′ = AY .
Hence (1) reduces to
Y u′ = g
and [ ][ ] [ ]
e2t e4t −6e−2t −2
u′ = Y −1g = 1
= .
2 e2t −e4t 2e−2t −4e2t
Integration gives [ ]
∫t −2t
u(t) = Y −1gdt = .
0 −2e2t + 2
Thus a particular solution is [ ] [ ]
−2t − 2 2
y(p) = Y u = e−2t + e−4t,
−2t + 2 −2
and the general solution of the system is
y = y(h) + y(p)
[ ] [ ] [ ] [ ]
1 1 1 −2
= c1 e−2t + c∗2 e−4t −2 te−2t + e−2t
1 −1 1 2
| {z }
why!

Computational Science & Engineering (CSE) T. Jeong