18 Systems of first order ODE.

Introduction
Up till now we were studying one equation at a time. In this last part of the course we will be studying
systems of first order ODE. It is quite a straightforward procedure to rewrite any explicit ODE of the
n-th order as a system of n first order equations.
Example 1. Consider the IVP for the pendulum:
g
θ̈ + sin θ = 0, θ(0) = θ0 , θ′ (0) = v0 .
l
Introduce new variables
y1 (t) = θ(t), y2 = θ′ (t).
We have

ẏ1 = y2 ,
g
ẏ2 = − sin y1 ,
l
with the initial conditions
y1 (0) = θ0 , y2 (0) = v0 .
Therefore, instead of one second order differential equation we end up with a system of two first order
equations. It turns out that this particular form is arguably the most convenient way to look at the
ordinary differential equations from both theoretical and computational points of view.
Example 2. Rewrite the equation
y ′′′ + 5ty ′′ + 3y = 0
as a system of three first order equations. Again, the same approach: Consider new variables:

y1 = y, y2 = y ′ , y3 = y ′′ .

Therefore,

ẏ1 = y2 ,
ẏ2 = y3 ,
ẏ3 = −5ty3 − 3y1 ,

and we obtain a system of three first order ordinary differential (linear in this case) equations. If
together with the original equation we had to solve an IVP, i.e., we also had the initial conditions

y(t0 ) = a0 , y ′ (t0 ) = a1 , y ′′ (t0 ) = a2 ,

(the number of initial conditions coincides with the order of the equations), then for the system our
initial conditions turns into
y1 (t0 ) = a0 , y2 (t0 ) = a1 , y3 (t0 ) = a2 ,
and the number of initial conditions is equal to the number of the variables and the number of equations
in the system of first order ODE.
MATH266: Intro to ODE by Artem Novozhilov, e-mail: artem.novozhilov@ndsu.edu. Spring 2024

1
 Is it possible to go in the opposite direction — from systems to one equation? In most of the cases
the answer is negative, but for linear systems there is a simple approach.

Example 3. Consider the system of two first order ODE:

ẏ1 = ay1 + by2 ,

ẏ2 = cy1 + dy2 ,

where a, b, c, d ∈ R are some parameters. Differentiate, for example, the first equation:

ÿ1 = aẏ1 + bẏ2 = aẏ1 + b(cy1 + dy2 ).

From the first equation we also can find

1
y2 = (ẏ1 − ay1 ),
b
therefore,
( d )
ÿ1 = ay1 + b cy1 + (ẏ1 − ay1 ) ,
b
which is a linear second order ODE with constant coefficients. To obtain a final answer for the system,
we need to present a pair of functions as a solution. After finding y1 from the last equation, you can
use the expression for y2 to write it down. In the final answer there should be two arbitrary constants,
which can be present in both expressions for y1 and y2 .

Most of the time we will be studying linear systems of ODE with constant coefficients. To formulate
the problem, I will use the matrix notations (more on this in the next lecture). First, we have a square
matrix A = [aij ]n×n , which is given explicitly as
 
a11 . . . a1n
 ..  .
A =  ... ..
. . 
an1 . . . ann

To denote matrices I usually use capital letters and bold font.

We also have vector-function  
f1 (t)
 
f (t) =  ...  ,
fn (t)
and the unknown vector-function  
y1 (t)
 
y(t) =  ...  ,
yn (t)
both of which are column vectors with n components. In these lecture notes I use small letters and
bold font to point that this is a vector, whereas in the class I usually use the notations f⃗(t) and ⃗y (t).
Finally, recall that the derivative of a vector function means the vector of derivatives of its components.

2
Definition 4. The system
ẏ = Ay + f (t) (1)
is called the linear system of n first order ODE with constant coefficients. This system is called ho-
mogeneous if f (t) ≡ 0, and nonhomogeneous otherwise. System (1) together with the initial condition

y(t0 ) = y 0 ∈ Rn (2)

is called the initial value problem (IVP).

Note how the matrix notations allow to write in a concise way. In coordinates system (1) takes
the form

ẏ1 = a11 y1 + . . . + a1n yn + f1 (t),

..
.
ẏn = an1 y1 + . . . + ann yn + fn (t).

In most of the cases we will be dealing with systems of two or three first order equations. However,
the general theory remains valid for any n.
Example 5. Consider the system

ẏ1 = −y1 ,
ẏ2 = 2y2 .

Actually, the only thing that makes this pair of equations a system is the fact that I announced that
this is a system. Otherwise, the equations are not connected to each other, or decoupled. This is
always true for any linear system, if the matrix is diagonal, in this particular case:
[ ]
−1 0
A= .
0 2
We can solve these equations separately, obtaining

y1 (t) = C1 e−t ,
y2 (t) = C2 e2t .

Of course, we can plot these solutions in the usual manner, in coordinates ( (t, y1 ) and
) (t, y2 ).
Another nice geometric way to look at this solution is to note that the pair y1 (t), y2 (t) defines a
parametric curve on the plane (y1 , y2 ). Moreover, we have a direction on this curve, which is given
by the increase of the parameter t, which is often associated with time. Let us figure out what kind
of curves on the plane (y1 , y2 ) are given by our solution. For this we eliminate t from the solutions,
getting
y12 C2 C
2 = e−2t = =⇒ y2 = 2 ,
C1 y2 y1
which is straightforward to sketch (left as an exercise). The plane (y1 , y2 ) is called the phase plane,
and the parametric curves defined by the solutions of ODE are called orbits. One of our main goals
is to learn how to figure out the structure of orbits on the phase plane for any linear homogeneous
system of two first order ODE.

3
 In the last example we were able to solve our system because the equations were decoupled. It
turns out that this is the general approach to the linear homogeneous system: In order to solve a
system one needs to find a change of variables such that in the new coordinates the system becomes
decoupled (or as simple as possible, if we cannot decouple the equations). Much more on this will be
covered in the following lectures.