1

Concourse 18.03 Lecture #8

Today we continue with some of the details of 2nd order (and higher) linear ODEs. We will focus primarily on
the constant coefficient case in which we can easily construct exponential solutions from the roots of the
characteristic equation and then combine these solutions via superposition. Linear algebra idea regarding
spanning sets and linear independence will be introduced as needed, and well define the Wronskian as a tool
for checking independence of solutions. Well apply these methods to the study of mass-spring-dashpot
systems.
Higher order linear ordinary differential equations with constant coefficients
In general, an nth order linear ordinary differential equation is a differential equation of the form
1
1 1 1 0
( ) ( ) ( ) ( ) ( )
n n
n n n
d x d x dx
dt dt dt
p t p t p t x t q t

+ + + + = , where
1 1 0
( ), , ( ), ( ), ( )
n
p t p t p t q t

are functions of the

independent variable t. We solve this by (1) finding an expression for all homogeneous solutions ( )
h
x t ,
(2) using some productive method to find one particular solution ( )
p
x t to the inhomogeneous equation, and
then (3) adding these to get the general solution ( ) ( ) ( )
h p
x t x t x t = + . If we are solving an initial value problem,
we would then use the initial conditions to determine any unknown constants in the expression for ( ) x t .
One case of special interest is the case where all of the coefficient functions ( )
i i
p t a = are constant. In this case
the differential equation simplifies to
1
1 1 1 0
( ) ( )
n n
n n n
d x d x dx
dt dt dt
a a a x t q t

+ + + + = .
If we seek exponential solutions of the form
rt
e for the homogeneous equation
1
1 1 1 0
( ) 0
n n
n n n
d x d x dx
dt dt dt
a a a x t

+ + + + = , we calculate
rt
dx
dt
re = ,
2
2
2 rt
d x
dt
r e = , ,
n
n
n rt
d x
dt
r e = , and substitution
gives
1 2 1 2
1 2 1 0 1 2 1 0
( ) 0
n rt n rt rt rt rt n n rt
n n
r e a r e a r e a re a e r a r a r a r a e


+ + + + + = + + + + + = . This yields a
solution only when the characteristic polynomial
1 2
1 2 1 0
( ) 0
n n
n
p r r a r a r a r a

= + + + + + = .
So for any root
i
r of the characteristic polynomial,
i
r t
e will be a homogeneous solution. The Fundamental
Theorem of Algebra guarantees (in principle) that we can factor ( ) p r into a product of linear factors and
irreducible quadratic factors. As long as there are no repeated roots, and since we can use the quadratic formula
to produce a complex conjugate pair of roots for each irreducible quadratic factor, we will be able to produce n
distinct roots and a corresponding set of exponential solutions
{ }
1 2
, , ,
n
r t r t r t
e e e . In the case of repeated roots,
this will yield fewer solutions of this form.
By linearity, any function of the form
1 2
1 2
( )
n
r t r t r t
h n
x t c e c e c e = + + + will solve the homogeneous equation.
Question: Does this yield all solutions?
A second order example should explain why the answer is YES. Suppose we wish to solve the ODE
3 2 0 x x x + = . Any exponential solution
rt
e would give
2
( ) 3 2 ( 2)( 1) 0 p r r r r r = + = = . Its characteristic
roots are
1
2 r = and
2
1 r = , and these yield solutions
2t
e and
t
e . Why are ALL homogeneous solutions of the
form
2
1 2
( )
t t
x t c e c e = + ?
If we write the differential equation in terms of linear differential operators, we might write this as
2 1 ( ) 0
d d
dt dt
x t
( (
=

, i.e. as a composition of two 1st order linear differential operators. If we let
1 ( ) ( )
d
dt
x t y t
(
=

, this gives two 1st order equations: ( )
dx
dt
x y t = and 2 0
dy
dt
y = . The latter equation is
easily solved to give all solutions
2
1
( )
t
y t c e = where
1
c is a constant. We then substitute this into the former
equation to get
2
1
t
dx
dt
x c e = . This is an inhomogeneous equation with integrating factor
t
e

. Multiplication by
2
this gives
1
( )
t t t t
dx d
dt dt
e xe xe c e

= = , so
1 2
t t
xe c e c

= + . Finally, multiplying both sides by

t
e gives
2
1 2
( )
t t
x t c e c e = + .
It should be clear that this approach can be generalized to the nth order case as long as the characteristic
polynomial can be factored into distinct linear factors. (We write the differential equation as a composition of n
1st order linear operators and iterate the above process.) This even works in the case of complex roots as long as
they are not repeated. The more difficult case is when there are repeated roots of the characteristic polynomial,
but, as well soon see, this case also yields a relatively simple solution.
In Linear Algebra terms, we say that
{ }
1 2
, , ,
n
r t r t r t
e e e span all solutions in the above case. It is a valid question
to ask whether all of these solutions are necessary, i.e. if we could span all solutions with a subset of these
exponential solutions. In Linear Algebra terms, we would ask: Are these solutions are linearly independent?
In other words, is it possible to express any of these solutions as a linear combination of the other solutions?
Definition: A set of functions { }
1 2
, , ,
n
f f f is called linearly independent if the equation
1 1 2 2
( ) ( ) ( ) 0
n n
c f t c f t c f t + + + = (for all t) implies that
1 2
0
n
c c c = = = = .
When seeking solutions to an nth order linear differential equation, we actually want more than this. In
order to guarantee unique solutions to any well-posed initial value problem with initial conditions given for the
function and its derivatives up to order ( 1) n , we would also want that:
{ }
1 1 0 2 2 0 0 0
1 1 0 2 2 0 0 0
1 2
( 1) ( 1) ( 1) ( 1)
1 1 0 2 2 0 0 0
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
unique , , ,
( ) ( ) ( ) ( )
n n
n n
n
n n n n
n n
c f t c f t c f t x t
c f t c f t c f t x t
c c c
c f t c f t c f t x t

+ + + =

' ' '
+ + + =

`

+ + + =
)

In terms of matrices, we can express this as:
1 0 2 0 0
0 1 1
0 2 2
1 0 2 0 0
( 1)
( 1) ( 1) ( 1)
0
1 0 2 0 0
( ) ( ) ( )
( )
( )
( ) ( ) ( )
unique
( )
( ) ( ) ( )
n
n
n
n n n
n n
n
f t f t f t
x t c c
x t c c
f t f t f t
c c x t
f t f t f t


(
( ( (
(
( ( (
' ' '
(
= ( ( (
(
( ( (
(
( ( (

Two fundamental results in linear algebra say that this will only be the case when the above matrix is invertible,
and this will only be the case when its determinant is never equal to 0.
Definition:
1 2 1 2
1 2 1 2
( 1) ( 1) ( 1) ( 1) ( 1) ( 1)
1 2 1 2
det
n n
n n
n n n n n n
n n
f f f f f f
f f f f f f
f f f f f f

(
(
' ' ' ' ' '
(
=
(
(





is called the Wronskian determinant.
Corollary: If the Wronskian determinant is never 0, the given ODE will yield unique solutions in the form
1 1 2 2
( ) ( ) ( ) ( )
h n n
x t c f t c f t c f t = + + + for any given initial conditions given for the function and its derivatives up
to order ( 1) n .
Though not routinely used to ensure a linearly independent set of solutions, (there are arguments with less
tedious calculations that can be made), the Wronskian is one tool for ensuring that a set of solutions to a
homogeneous linear ODE is valid for uniquely expressing all homogeneous solutions.
3
Example: Solve the initial value problem 5 4 3sin2 x x x t + + = with initial conditions (0) 3 x = , (0) 2 x = .
Solution: We first solve the homogeneous equation 5 4 0 x x x + + = . Its characteristic polynomial is
2
( ) 5 4 ( 4)( 1) p r r r r r = + + = + + and this yields two distinct roots 4 r = and 1 r = . The corresponding
exponential solutions are
4t
e

and
t
e

. We can check that these are, in fact, linearly independent by calculating

the Wronskian determinant:
4
5 5 5
4
4 3 0
4
t t
t t t
t t
e e
e e e
e e



= + = =

. From our previous arguments, we know
that all homogeneous solutions are of the form
4
1 2
( )
t t
h
x t c e c e

= + .
Next, we seek a particular solution. There are at least two good ways to do this. We could do complex
replacement and simultaneously solve 5 4 3cos2 x x x t + + = and 5 4 3sin2 y y y t + + = by solving the
inhomogeneous equation
2
5 4 3
it
z z z e + + = and then taking the imaginary part. Its perhaps easier to solve
using undetermined coefficients.
If we let cos2 sin2 x a t b t = + , we get
cos2 sin2
2 cos2 2 sin2 5 4 (10 )cos2 ( 10 )sin2
4 cos2 4 sin2
x a t b t
x b t a t x x x b t a t
x a t b t
= +

= + + = +
`

=
)

We must therefore have 10 0 b = and 10 3 a = , so
3
10
a = and 0 b = . So
3
10
( ) cos2
p
x t t = .
The general solution is therefore
4
1 2
3
10
( ) cos2
t t
x t c e c e t

= + , and we have
4
1 2
3
5
( ) 4 cos2
t t
x t c e c e t

= + .
If we substitute the initial conditions (0) 3 x = , (0) 2 x = , we have:
53 3 33 33
1 2 1 2 30 1 1 10 10 10
76
2 2 1 2 1 2 15
(0) 3 1 1
4 1 (0) 4 2 4 2 2
x c c c c c c
c c x c c c c
= + = + = ( ( ( ( (
= =
` `
( ( ( ( (
= = =
) )

.
We certainly dont have to use matrices to solve these two equations, but its worth noting that the nonvanishing
of the Wronskian determinant is precisely why there is a unique solution for these constants. The unique
solution to this initial value problem is therefore
4
53 76
30 15
3
10
( ) cos2
t t
x t e e t

= + .
Note: In this example, the exponential terms are transients (they decay quickly) and the steady state solution
is the particular solution that we calculated.
Mass-Spring-Dashpot systems
Of particular interest to us (for a variety of reasons) are mass-spring-dashpot systems in which a spring is
governed by Hookes Law but also subject to friction that is proportional to the velocity. [A picture was drawn
in class illustrating a spring with an attached mass and the friction supplied by a piston (dashpot).] The simplest
case is where this system is confined with the spring attached to one fixed wall, the dashpot on the other side of
the mass attached to another fixed wall, and the mass moving relative to its equilibrium position. In this case,
we would express the force acting on the mass as F kx cv = where v x = and F ma mx = = . This gives the
system 0 mx cx kx + + = or 0
c k
m m
x x x + + = .
We could also imagine a system that is driven by moving either the fixed end of the spring or by moving the
fixed end of the dashpot. If we incorporate this additional acceleration, the resulting system would be governed
by an inhomogeneous ODE of the form ( )
c k
m m
x x x q t + + = .
Note: We get similar equations in the case of an electric circuit with inductance (L), resistance (R), and
capacitance (C), i.e. and LRC circuit.
4
Spring only case
The simplest case is a pure spring with no friction and no external driving force. In this case, the differential
equation governing the motion would be simply 0
k
m
x x + = . In anticipation of what will follow, its useful to
let
2
k
m
e = or
k
m
e = . This gives the differential equation
2
0 x x e + = . Its characteristic polynomial is
2 2
( ) 0 p r r r i e e = + = = . So all solutions to this homogeneous equation can be expressed as the span of
{ }
,
i t i t
e e
e e
, i.e. in the form
1 2
( )
i t i t
x t c e c e
e e
= + where
1 2
, c c are complex constants. We would, of course, prefer
to express solutions as real-valued functions. Using Eulers Formula, we could rewrite the solutions as
1 2 1 2 1 2
( ) (cos sin ) (cos sin ) ( )cos ( )sin x t c t i t c t i t c c t i c c t e e e e e e = + + = + + and then hope that any given
initial condition will produce real coefficients (they will). Another way to look at this is to note that since
cos sin
i t
e t i t
e
e e = + and cos sin
i t
e t i t
e
e e

= and we can also solve for

2
cos
i t i t
e e
t
e e
e

+
= and
2
sin
i t i t
e e
i
t
e e
e

= , it must be the case that

{ } { } Span , Span cos ,sin
i t i t
e e t t
e e
e e

= . That is, all solutions must be

of the form ( ) cos sin x t a t b t e e = + . We also have the option of expressing this as ( ) cos( ) x t A t e | = where
2 2
A a b = + and tan
b
a
| = .
Note: If we felt the urge to inquire whether the set
{ }
,
i t i t
e e
e e
or the set { } cos ,sin t t e e were linearly
independent solutions, the corresponding Wronskians would give either 2 0
i t i t
i t i t
e e
i
i e i e
e e
e e
e
e e

= =

or
2 2
cos sin
(cos sin ) 0
sin cos
t t
t t
t t
e e
e e e e
e e e e
= + = =

. They both provide a linearly independent spanning set for

the solutions, i.e. a basis for the solutions (in linear algebra terms).
In the next lecture well look at the cases where there is friction and classify these as underdamped,
overdamped, or critically damped. The underdamped case is characterized by decaying oscillatory solutions.
The overdamped case is characterized by exponential decay. The critically damped case will be most interesting
in terms of the concepts linear independence and spanning sets we just introduced. After that well move on to
driven systems.
Notes by Robert Winters