Math 18500 Week 4: Second Order Linear Equations

Damped Oscillators. We have seen how to solve linear equations with constant coefficients, but the
solution was rather unmotivated - we have not explained how such an equation would come up in practice.
So, it is worth introducing some physical examples which you can have in mind when thinking about second
order linear equations.
For the first example, consider a mass hanging on a spring. The mass has an equilibrium position, in which
it is at rest - let y denote the displacement from this equilibrium position. Then combining Newton’s law
for the force acting on an object of mass m,
F = ma = my ′′
with Hooke’s law for the force exerted by a spring,
F = −ky
we see that y(t) satisfies the equation
d2 y
+ ky = 0.
m
dt2
This homogeneous and we know how to solve it - the corresponding characteristic equation
mλ2 + k = 0
has purely imaginary roots r
k
λ = ±i = ±iω0 ,
m
so a fundamental pair of solutions is
y1 (t) = Re eiω0 t = cos(ω0 t) and y2 (t) = Im eiω0 t = sin(ω0 t),
   

and the general solution is

y(t) = a cos(ω0 t) + b sin(ω0 t).
The constant ω0 is called the natural frequency of the mass-spring system.
Sometimes it is nice to write the solution y(t) in phase-shifted form
y(t) = A cos(ω0 t − ϕ).
In this context, A is called the amplitude of the oscillation, and ϕ is called the phase shift.
To bring the solution into phase-shifted form, we can first the substitution
a = A cos ϕ , b = A sin ϕ,
or equivalently we draw a triangle:

A
b
ϕ
a
We then apply the angle subtraction formula, cos(x − y) = cos x cos y + sin x sin y:
y(t) = A cos(ω0 t) cos(ϕ) + A sin(ω0 t) sin(ϕ) = A cos(ω0 t − ϕ).

Unfortunately, a sinusoidal solution is not physically realistic - the energy of an actual spring would be
dissipated gradually in the form of heat, and the spring would eventually come back to rest at its equilibrium
position. This effect can be modeled1 by introducing a damping force which resists the velocity of the spring,
rather than its displacement from equilibrium. Adding a damping force leads to an equation of the form
d2 y dy
m +l + ky = 0,
dt2 dt
1In reality, the damping of a mass-spring system is nonlinear - the model we are describing is only an approximation, and
this approximation is only accurate when the system is very close to its equilibrium state.
1
 2

where m, l, and k are positive constants.

In the damped case, the characteristic equation
mλ2 + lλ + k = 0
has two roots √
−l ± l2 − 4mk
q
λ= = −b ± b2 − ω02 ,
2m
where r
l k
d= and ω0 =
2m m
Note that the roots may be either real or complex, depending on whether d2 − ω02 is positive or negative.
In the real case (d2 > ω02 ), the system is said to be overdamped, and the corresponding differrential equation
has two real exponential solutions,
y1 = er1 t and y2 = er2 t .
Since d and ω02 are positive, we have
q √
± d2 − ω02 < d2 = d,
which implies that q
−d ± d2 − ω02 < −d + d = 0.
Therefore, the roots r1 and r2 are both negative, and the corresponding exponential solutions are decreasing.
In the complex case (d2 < ω02 ), the system is said to be underdamped, and the exponential solutions are
y = e(r+iω)t , y = e(r−iω)t
where q
r = −d and ω = ω02 − d2 .
The real and imaginary parts of the complex solutions are a fundamental pair of solutions,
y1 = ert cos (ωt) and y2 = ert sin (ωt) .
To graph these functions, one draws an “exponential envelope” y = ±ert , together with a sinusoid oscillating
between the upper and lower boundaries of the envelope:
1

0.5

1 2 3 4 5

−0.5

−1

Note that both solutions still decay to 0 exponentially (since r = −d < 0), but now they also oscillate as
they decay. Also note that damping reduces the frequency of the oscillation.
There is also a third case - if d2 = ω02 , then the characteristic equation has a repeated root and the oscillator
is said to be critically damped. In this case, a fundamental pair of solutions is
y1 = ert , y2 = tert ,
 3

where
r = −b < 0.
Note that both of these solutions decay to zero as well.
With this model in mind, we can also understand the physical meaning of an inhomogeneous equation
my ′′ + ly ′ + ky = f (t).
Such an equation describes a mass on a spring which is subject to an additional external force f (t).
Any physical system which can be accurately modeled by an equation of this form is called a damped oscillator
(in the case l > 0) or a harmonic oscillator (in the case l = 0).
We would like to apply the physical intuition provided by either of these models to our understanding of
second order linear equations. It is common to visualize a damped oscillator using the following “black box”
diagram:
f (t) y(t)

In this picture, f (t) represents the input to our physical system (e.g. external forces in the mass-spring case),
and y(t) represents the response to that input. To find the response, we must solve an equation of the form
my ′′ (t) + ly ′ (t) + ky(t) = f (t),
subject to whatever initial conditions happen to be physically reasonable.
 4

Exponential Response. We now turn our attention to inhomogeneous second order equations
my ′′ + ly ′ + ky = f (t)
where m, l, and k are constants, and f (t) is a function of t. Recall that equations like this can be written
in operator form as
Oy = f,
where
O = mD2 + lD + k
is a linear differential operator.
Before we get started solving inhomogeneous equations, recall the three crucial properties of linear equations:
(1) Suppose that y1 (t) and y2 (t) are two solutions of a linear homogeneous equation:
Oy1 = 0
Oy2 = 0.
If c1 and c2 are arbitrary constants, and
y(t) = c1 y1 (t) + c2 y2 (t),
then y(t) is a solution of the same homogeneous equation:
O[c1 y1 + c2 y2 ] = c1 Oy1 + c2 Oy2 = 0 + 0 = 0
In other words, solutions of homogeneous equations can be superimposed.

(2) Suppose that y1 (t) and y2 (t) are solutions of linear inhomogeneous equations with different right
hand sides:
Oy1 = f1
Oy2 = f2
If c1 and c2 are arbitrary constants, and
y(t) = c1 y1 (t) + c2 y2 (t)
then y(t) is a solution of the inhomogeneous equation
O [c1 y1 + c2 y2 ] = c1 f1 + c2 f2
where the right hand side is a linear combination of the right hand sides of the original equations.
(3) Suppose that yp (t) is a particular solution of an inhomogeneous equation,
Oyp = f
If yh (t) is any solution of the corresponding homogeneous equation,
Oyh = 0.
and if
y(t) = yp (t) + yh (t),
then y(t) is a solution of the original inhomogeneous equation:
Oy = Oyp + Oyh = f + 0 = f
From property (3) you can see that the only difficult part of solving an inhomogeneous equation is finding
one particular solution. Once a particular solution yp has been found, more general solutions can always be
constructed by adding solutions of the homogeneous equation.
As a simple example to get us started, suppose we want to solve an inhomogeneous equation like
y ′′ + 4y ′ + 3y = et .
In general, we can often find particular solutions of equations of the form
my ′′ + ly ′ + ky = ert ,
using an ansatz of the form
y = Cert ,
 5

where C is an unspecified constant (or undetermined coefficient).

In the case of the equation
y ′′ + 4y ′ + 3y = et ,
we can try the ansatz y = Cet . This gives us
Cet + 4Cet + 3Cet = et .
Dividing by et we find that
8C = 1,
so C = 81 . This gives us a particular solution,
1 t
yp =
e.
8
Once we have a particular solution, more general solutions can be obtained by adding solutions of the
corresponding homogeneous equation
y ′′ + 4y ′ + 3y = 0.
The general solution of this equation works out to be
yh = c1 e−t + c2 e−3t ,
and we can conclude that
1 t
e + c1 e−t + c2 e−3t
y = yp + yh =
8
is a solution of the original inhomogeneous equation. 2
This idea allows us to solve arbitrary inhomogeneous initial value problems (once a particular solution is
known!). For example, to solve the initial value problem
y ′′ + 4y ′ + 3y = et , y(0) = 0 , y ′ (0) = 0,
we can just write out the solution we found above and its derivative,
1
y = et + c1 e−t + c2 e−3t ,
8
1 t
y = e − c1 e−t − 3c2 e−3t .
′
8
Setting t = 0 in both equations gives us a system of equations for c1 and c2 ,
1
0 = y(0) = + c1 + c2 ,
8
1
0 = y ′ (0) = − c1 − 3c2 .
8
Adding the equations, we find that
1 1
0 = − 2c2 =⇒ c2 = ,
4 8
and substituting in the first equation, we find that
1 1 1
0 = + c1 + =⇒ c1 = − .
8 8 4
So, the solution of the initial value problem is
1 1 1
y = et − e−t + e−3t .
8 4 8
To reiterate, the only difficult part of solving inhomogeneous equations is finding particular solutions.
2In fact, this is the general solution of the inhomogeneous equation! If y was any other solution, then we would have

Oy = Oyp = f
O(y − yp ) = Oy1 − Oy2 = f − f = 0.
Therefore, yh = y − yp is a solution of the homogeneous equation, and
y = yp + yh .
 6

As a next simplest case, suppose we are given an inhomogeneous equation

my ′′ + ly ′ + ky = c1 er1 t + c2 er2 t ,
where the right hand side is a linear combination of exponentials. Equations of this form can be solved by
applying crucial property (2) of linear equations.
For example, consider the equation
y ′′ + 4y ′ + 3y = 5et − 7e3t .
To find a particular solution of this equation, we first seek solutions y1 (t) and y2 (t) of the equations
y1′′ + 4y1′ + 3y1 = et ,
y2′′ + 4y2′ + 3y2 = e3t .
If we can find theses solutions, then the linear combination y(t) = 5y1 (t) − 7y2 (t) will be a solution of the
equation
y ′′ + 4y ′ + 3y = 5et − 7e3t .
Of course, we have already seen that y1 = 81 et is a solution of the first equation. To find a solution of the
second equation, we can try an ansatz of the form y2 = Be3t , where B is an undetermined constant. This
leads to the equation
9Be3t + 12Be3t + 3Be3t = e3t ,
1
and solving this equation gives the value B = 24 , so
1 3t
y2 = e .
24
Therefore,
5 t 7
yp (t) = 5y1 (t) − 7y2 (t) =e − e3t
8 24
is the desired particular solution. Then the general solution is
5 7
y(t) = et − e3t + c1 e−t + c2 e−3t .
8 24