Differential

Equations
y

t
5 10 15 20 25 30

t
5 10 15 20 25 30

t
5 10 15 20 25 30

y
t
5 10 15 20 25 30

⟞ MATH 280 • FALL 2022 ⟝

Applied Differential Equations
Fall 2022

Chad Sprouse
chad.sprouse@csun.edu
 Here are some of our conventions:

1. When a word or phrase is defined, it is written in boldface.

2. Most examples are ended with a □, so that you can tell where they end.
✎ ☞
3. Important procedures to follow are usually surrounded by a round box .
✍ ✌

Important facts/definitions are surrounded by a shadow box .

Important formulas are surrounded by a double box .

If it has a box around it, it is something that you should remember, and usually
something you should memorize.

4. Sections, subsections, or exercises that will not be on the final exam are marked
with ∗.

Other texts: any edition of Boyce and DiPrima, Elementary Differential Equations (with
or without boundary value problems) can be used as a supplement. This is the standard
book which covers much of the same material. It does not perfectly correspond to how I
would do things (hence these notes), but has a good number of examples and all answers
to the exercises in the back, not just odd ([1]).
As for other textbooks, [14] is also pretty common, and is the book I took this class
from. [17] is a bit streamlined, with slightly more emphasis on theoretical aspects also
covered here. Some of the pictures later on in these notes use the Java applet pplane.jar
from this book, which is a great applet for drawing vector fields. It can be downloaded
from https://www.cs.unm.edu/~joel/dfield/ if one also has Java installed. A good
text for a more theoretical course, but still at about the same level would be [20].
The origin of these notes was to have something covering Linear Algebra as well as
ODE, a requirement for Math 280 at the time, which is preserved in Chapter 2. The
material there is somewhat dated compared to the rest of the notes, but has been left in
just-in-case. A standard textbook covering both of these topics would be [5]. The dif-
ference is that this and similar texts attempt a complete introduction to Linear Algebra,
somewhat at the expense of the Differential Equations. Here, we focus only on the Linear
Algebra directly relevant to the theory of ODE’s that we will need.

Thanks: Long ago, class notes were given to me by Suvimol Nilprapa, which eventually
became this. Special thanks to Corine Sutherland, Andrew Knightly, and Sungjin Kim
who have been particularly helpful with comments and proofreading, along with Kyle
Booth, Joe Castagna, Gyana Cureg, Jose Gonzales Sanchez, Paul Graves, Abdullah Hendy,
and Kayla Saelee who have found several typos in previous semesters. More have surely
been generated since then.
Contents

0 Preliminaries 1

1 First Order ODE’s 5

1.1 Differential Equations: Basic Notions and Definitions . . . . . . . . . . . . . 5
1.1.1 Elementary Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.2 Linearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.1.3 Initial Value Problems . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2 Separation of Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.2.1 Solutions by definite integrals . . . . . . . . . . . . . . . . . . . . . . 27
1.2.2 Autonomous First Order ODE’s ∗ . . . . . . . . . . . . . . . . . . . . 29
1.3 Linear First Order ODE’s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.4 Graphical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
1.4.1 Slope Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
1.4.2 Autonomous ODE’s and sign charts . . . . . . . . . . . . . . . . . . . 51
1.5 Existence and Uniqueness ∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
1.5.1 The Noncrossing Principle . . . . . . . . . . . . . . . . . . . . . . . . 69
1.6 Substitutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
1.6.1 Homogeneous First Order ODE’s . . . . . . . . . . . . . . . . . . . . 75
1.6.2 Bernoulli equations ∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
1.7 Applications: Linear First Order ODE’s . . . . . . . . . . . . . . . . . . . . . 87
1.7.1 Exponential Growth and Population Models . . . . . . . . . . . . . . 87
1.7.2 Newton’s Law of Cooling: . . . . . . . . . . . . . . . . . . . . . . . . 89
1.7.3 Mixing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
1.7.4 Circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
1.7.5 Skydiving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

2 Higher Order Linear ODE’s 113

2.1 General Theory: Homogeneous Linear n-th order ODE’s . . . . . . . . . . . 113
2.1.1 The Wronskian and Fundamental Solution Sets . . . . . . . . . . . . 116
2.1.2 Existence and Uniqueness: The importance of being nonsingular . . 125
2.1.3 Linear Independence and Fundamental Solution Sets ∗ . . . . . . . 128
2.2 Reduction of Order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
2.3 Constant Coefficient ODE’s: Fundamental Solution Sets . . . . . . . . . . . 145
2.3.1 Second Order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
2.3.2 Higher Order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

iii
 2.4 Nonhomogeneous Linear ODE’s . . . . . . . . . . . . . . . . . . . . . . . . . 160
2.5 Undetermined Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
2.6 Variation of Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
2.6.1 Definite Integrals and the forced solution yF (x) . . . . . . . . . . . . 187
2.7 Oscillations: Amplitude, Phase, and Transients . . . . . . . . . . . . . . . . 198
2.7.1 Sinusoids ∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
2.7.2 Steady-State vs. Transients . . . . . . . . . . . . . . . . . . . . . . . . 202
2.8 Applications of Linear Second Order ODE’s . . . . . . . . . . . . . . . . . . 208
2.8.1 Springs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
2.8.2 RLC Circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
2.9 Oscillations II: Frequency Response and Resonance ∗ . . . . . . . . . . . . 226
2.9.1 Frequency Response ∗ . . . . . . . . . . . . . . . . . . . . . . . . . . 226
2.9.2 Resonance ∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230

3 Laplace transforms 241

3.1 Definitions and Basic Transforms . . . . . . . . . . . . . . . . . . . . . . . . 241
3.1.1 Derivative and Integral Properties of L . . . . . . . . . . . . . . . . . 246
3.1.2 Table of Common Transforms . . . . . . . . . . . . . . . . . . . . . . 253
3.2 Solving ODE’s by Laplace Transform . . . . . . . . . . . . . . . . . . . . . . 257
3.3 Shift Properties and Step Functions . . . . . . . . . . . . . . . . . . . . . . . 270
3.4 Impulse Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
3.5 Convolution and Impulse Response ∗ . . . . . . . . . . . . . . . . . . . . . . 296
3.6 Systems and Circuits ∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
3.6.1 Systems ∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
3.6.2 Circuits ∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306

4 First Order Systems 313

4.1 Constant Coefficient Systems: Substitution/Elimination . . . . . . . . . . . 313
4.2 Applications of First Order Systems ∗ . . . . . . . . . . . . . . . . . . . . . 322
4.2.1 Mixing ∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
4.2.2 Newton’s Law of Cooling ∗ . . . . . . . . . . . . . . . . . . . . . . . 325
4.2.3 Circuits ∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
4.2.4 Springs ∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330
4.3 First Order Systems: General Theory . . . . . . . . . . . . . . . . . . . . . . 336
4.3.1 Basic Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
4.3.2 Linear Systems: Fundamental Solution Sets and Fundamental Ma-
trices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
4.3.3 n-th Order ODE’s to First Order Systems . . . . . . . . . . . . . . . . 344
4.4 Eigenvalues and Eigenvectors . . . . . . . . . . . . . . . . . . . . . . . . . . 351
4.5 2-Dimensional Systems by Eigenvalues/Eigenvectors . . . . . . . . . . . . . 354
4.5.1 Real distinct eigenvalues . . . . . . . . . . . . . . . . . . . . . . . . . 354
4.5.2 complex eigenvalues . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
4.5.3 repeated eigenvalues, λ1 = λ2 . . . . . . . . . . . . . . . . . . . . . . 358
4.6 Linear Systems Gallery ∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
4.6.1 Real Distinct Eigenvalues ∗ . . . . . . . . . . . . . . . . . . . . . . . 368
4.6.2 Complex Eigenvalues: λ = α ± βi ∗ . . . . . . . . . . . . . . . . . . . 372
 4.6.3 Real Repeated Eigenvalues ∗ . . . . . . . . . . . . . . . . . . . . . . 376
4.7 Nonhomogeneous First Order Linear Systems ∗ . . . . . . . . . . . . . . . . 380

A Appendices 387
A.1 Complex Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387
A.1.1 Trigonometric Identities . . . . . . . . . . . . . . . . . . . . . . . . . 390
A.1.2 Roots of Unity: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
A.2 A Survey of Partial Fractions Methods . . . . . . . . . . . . . . . . . . . . . 396
A.2.1 Solving for the coefficients . . . . . . . . . . . . . . . . . . . . . . . . 398
Chapter 0

Preliminaries

You will need to remember the following techniques of integration from calculus:
• u-substitution.
• Integration by Parts.
• Partial Fractions.
In the calculus book used at CSUN (Briggs 3rd), these are sections 5.5, 8.2, and 8.5.
But any calculus book has sections on these, and you should review them if you feel you
might be rusty. They will be coming up soon! Speaking of integrals, the number one
scourge in Math 280 is the idea that ∫ whatever
1
dx = ln ∣whatever∣ + C. For instance stuff
like this:

dx = ln(x ) + C
1 2 1
dx = ln( x) + C
x2 x

This sort of thing is almost never true, but only works in the case ∫ x+a
1
dx = ln ∣x + a∣ +
C. If you have been inclined to do this in the past, then the first step to success in Math
280 is to

Stop Doing It.

These integrals are actually
1 −2 −1 1 − 12
dx = x dx = −x +C dx = x dx = 2 x + C.
x2 x
There are no logs in the answers at all. Of course it is possible to get logs for some inte-
grals like this, but only after doing the appropriate u-substitution first.

1
 2

You will need to know the following integrals. So remember them, especially the two
at the bottom.

x x
• e dx = e + C

dx = ln ∣x∣ + C
n+1
(n ≠ −1)
x −1
•
n
x dx = +C • x
n+ 1

• sin(x) dx = − cos(x) + C • cos(x)dx = sin(x) + C

sec (x) dx = tan(x) + C csc (x) dx = − cot(x) + C

2 2
• •

• sec(x) tan(x) dx = sec(x) + C • csc(x)cot(x) dx = − csc(x) + C

• tan(x) dx = − ln ∣ cos(x)∣ + C • cot(x)dx = ln ∣ sin(x)∣ + C

= ln ∣sec(x)∣ + C
csc(x)dx = − ln ∣csc(x)+cot(x)∣+C
sec(x) dx = ln ∣sec(x) + tan(x)∣ + C
•
•

√
1 1
• dx = arctan(x) + C • dx = arcsin(x) + C
2
x +1 1 − x2

Here is one more algebraic issue:

x+C x C x+C x
e = e +e e = e +C

x+c x c x c
Of course the correct answer is e = e e = Ae , where A = e . You must respect the
exponential rules, which will occur frequently in the first chapter.

C. Sprouse —Math 280 Notes Page 2 DRAFT August 26, 2022

 3

Homework 0.1:
A) Practice with the following integrals.
1. 9.
1 2 3x
dx x e dx
2x + 1
2. 10.
√
x 2 x3
dx x e dx
x2 + 1
3. 11. x
√
x
3 e
dx dx
4
x +1 1 − e 2x
4. 12.
x cos(ln(x))
dx x dx
x4 + 1
5. 13.
x
dx sec(3x) dx
x+2
6. 14.
1 sin(x)
cos2(x)
2
dx dx
x −1
7. 15.
3x − 1
(x − 2)(x 2 + 1) x sec (x) dx
dx 2

8. 16.
2x + 1
x 2 (x + 1)
dx ln(x) dx

C. Sprouse —Math 280 Notes Page 3 DRAFT August 26, 2022

 4

C. Sprouse —Math 280 Notes Page 4 DRAFT August 26, 2022

Chapter 1

First Order ODE’s

1.1 Differential Equations: Basic Notions and Definitions

1.1.1 Elementary Terminology
Definition 1.1.1. An Ordinary Differential Equation is an equation of the form
(n) ′
F(y , ..., y , y, x) = 0,

or
n
d y dy
F ,..., , y, x = 0.
dx n dx

This just says that a differential equation is an equation that involves a variable x,
a function y(x) and some derivatives of y. That is, F is just notation for any specific
relation between y(n) (x), ..., y ′(x), y(x), and x. When writing such an equation we will
often suppress the dependence of y on x. In other words, we will usually just write y
instead of y(x). Also note that for us y (n) always mean the n-th derivative of y, as opposed
n
to y , which is y raised to the n-th power.

Example 1.1.2.
′
y − 2xy = 0.

solution: This is a differential equation with F(y′ , y, x) = y ′ − 2xy.

Note that the definition above does not imply that the right-hand side of a differential
equation is necessarily zero.

Example 1.1.3.
′′ 3
y + y = 1.

solution: This is a differential equation with F(y ′′ , y ′, y, x) = y ′′ + y 3 − 1.

Example 1.1.4.
dy 2
x = sin(y) − 3x .
dx

5
1.1. Differential Equations: Basic Notions and Definitions 6

dy dy 2
solution: This is a differential equation with F dx
, y, x = x dx − sin(y) + 3x .

The variables can be represented by different letters. For instance x often represents
position, while t represents time. In this case one can have a differential equation for the
function x(t), such as

Example 1.1.5.
2
d x dx −t
2
+3 + 2x = e .
dt dt
2 2
d x dx d y dy −t
solution: This is a differential equation with F , ,x,t
dx2 dt
= d x2
+ 3 dx + 2x − e □.

We use the abbreviation ODE for the words “ordinary differential equation”, which is
very standard. We will typically leave off the word ‘ordinary’, as we’ve done above. The
word ordinary is used to distinguish these equations from partial differential equations
(PDE’s) which involve partial derivatives.

Example 1.1.6.
2
∂u ∂ u
= 3 2 + u(1 − u)
∂t ∂x
is a partial differential equation.

Example 1.1.7.
2 2
∂ f 1 ∂f 1 ∂ f
+ + 2 = r sin(θ)
∂r 2 r ∂r r ∂θ 2
is a partial differential equation.

Example 1.1.8.
2 2 2
∂ u ∂ u ∂ u 3 2
2
+ 2 + 2 = x − yz
∂x ∂y ∂z
is a partial differential equation.

The most fundamental attribute that determines how a differential equation behaves
is simply the highest order derivative that appears, which is called the order of the equa-
tion. For example all three partial differential equations above are second-order. For
ordinary differential equations, this is even more straightforward:

Definition 1.1.9. The order of a differential equation is the highest order of derivative which
occurs in the equation. That is,
(n) ′
F(y ,..., y , y,x) = 0
and
n
d y dy
F n , ..., ,y,x = 0
dx dx
are both n-th order ODE’s.

Example 1.1.10.

C. Sprouse —Math 280 Notes Page 6 DRAFT August 26, 2022

1.1. Differential Equations: Basic Notions and Definitions 7

dy 2
• x dx + y = 3 is a 1-st order ODE.

• (y ) + 2y = sin(x) is a 2-nd order ODE.

′′ 3 ′

′′ ′′′ 4
• y y = xy is a 3-rd order ODE.
3
d x
• dt 2
+ x dx
dt
− 2x = 5t is a 3-rd order ODE.
(5) ′′′ 2 xy
• y + xy + y = e is a 5-th order ODE.
For y to be a solution to the ODE
(n) ′
F(y , ..., y , y, x) = 0,
it is of course necessary that y is a function such that the equation
(n)
(x),...,y (x), y(x), x) = 0
′
F(y
is true for all x. We need to be more specific, though! In order for y to be a solution, n
derivatives of y must exist. But sometimes when solving an ODE one gets ‘solutions’ that
are functions not actually defined for all x. In this case we must restrict the domain on
which we consider y to be a solution. It would not make sense for y to solve a differential
equation if the graph of y were to be disconnected. More specifically, the domain D of
the function y ∶ D → R that can be called a solution must also be connected, which means
that D must be an interval. Therefore by definition we only allow y to be a solution to
a differential equation on a single interval (a, b). Even if y(x) is defined for x outside of
(a, b), we don’t consider y to be a solution there.

Definition 1.1.11. A solution to the differential equation F(y (n) , ..., y′ , y,x) = 0 is a func-
tion y defined on an interval (a,b), such that y is n-times differentiable on (a, b) and satisfies
(n)
F(y (x), ..., y (x), y(x), x) = 0 for all x ∈ (a, b).
′

Example 1.1.12. y = e 3x is a solution to y ′′ − 2y′ − 3y = 0 on (−∞, ∞).

solution: Plugging y = e 3x into y ′′ − 2y ′ − 3y gives

(e ) − 2(e ) − 3e
3x ′′ 3x ′ 3x 3x 3x 3x
= 9e − 6e − 3e = 0, ✓

so y satisfies the differential equation. Also, y(x) is defined and twice differentiable for
all x, so y is a solution on (−∞, ∞).

is a solution to y = y on (−∞, 3).

1 ′ 2
Example 1.1.13. y = 3−x
solution:
′ 2
= −(3 − x) (−1) =
′ 1 −2 1
y = .
3− x 3− x
But similarly
2
2 1
y = . ✓
3− x
′ 2
So we can say that y is a solution to y = y on some open interval where it exists and is
differentiable. So y is a solution on (−∞, 3) □.

C. Sprouse —Math 280 Notes Page 7 DRAFT August 26, 2022

1.1. Differential Equations: Basic Notions and Definitions 8

We could also say above that y is a solution to y = y on (3, ∞). But what we cannot
′ 2

say is that y is a solution to y ′ = y 2 on (−∞, 3) ∪ (3,∞). That set is not a single open
interval!
Next, we note that differential equations generally have more than just one solution.
2 2 2
For example, from first semester calculus we know that y = x , y = x + 5, and y = x − 1
′
all satisfy y = 2x. In fact, without any other constraints such as initial conditions, a
differential equation by itself will have infinitely many solutions. The set of all such
solutions is called the general solution:
(n)
,..., y , y,x) = 0 on (a, b) is the set
′
Definition 1.1.14. The general solution to an ODE F(y
of all solutions to F(y (n) , ..., y ′ , y, x) = 0 on (a,b).

Example 1.1.15. Find the general solution to

′′
y + 2x = 0.

solution: This is easy to do, by just writing y′′ = −2x and integrating. That is, we have
d dy
dx dx
= −2x. So
dy 2
= −2x dx = −x + C,
dx
2 1 3
y= −x + C dx = − x + Cx + D.
3
′′ 3
So the general solution to y + 2x = 0 is y = − 13 x + Cx + D, where C and D are arbitrary
constants.

Most differential equations are not as simple as this one (if they were, there would be
no reason for these notes!), and cannot be solved by just integrating. For instance a more
interesting example is
′ −3x
Example 1.1.16. The general solution to y + 3y = 0 is y = Ce .

Although this can be solved by integrating, it is not as straightforward as the previous

′ dy
equation. If one writes y = −3y, then we have dx = −3y. But since the right-hand side
involves y instead of x, one cannot simply integrate the right-hand side by using a dx in-
tegral. The equation must be rewritten in a different way before things can be integrated.
The technique for this is covered in the next section and is known as separation of vari-
−3x
ables. Although we don’t wish to derive that y = Ce is the solution at this point, we can
easily verify that it is the solution by checking that it does in fact satisfy the differential
−3x ′
equation. That is, we substitute Ce into y + 3y and check that we get 0:
−3x ′ −3x −3x −3x
Ce + 3Ce = −3Ce + 3Ce = 0. ✓
Again, it is much easier to check that a function is the solution to an ODE rather
than to derive the solution. However, since we didn’t derive the above solution, we may
−3x 2
wonder if there are any other solutions that are not of the form Ce . Can x − 3x + 1
be a solution? Can tan(x) be a solution? It is easy to check that they cannot, by plugging

C. Sprouse —Math 280 Notes Page 8 DRAFT August 26, 2022

1.1. Differential Equations: Basic Notions and Definitions 9

them into the equation... but one cannot check all other functions in this way. To show
that every solution is of the form Ae−3x , we will have to actually solve the equation, which
literally means to find all solutions. We’ll go back and solve this ODE in the beginning of
the next section.

Example 1.1.17. Find as many solutions as possible to

′′
y + y = 0.

solution: Since this equation looks pretty simple, we might try to just guess a solution.
′′
We can rewrite the equation as y = −y and think about which common functions could
satisfy this. After a while, one might remember how sin(x) and cos(x) work. The
derivative of sin(x) is cos(x) and the derivative of cos(x) is − sin(x). And this means if
you take 2 derivatives of sin(x) you get − sin(x)! This means that we have found a
solution. Just to make sure, we insert sin(x) into the original equation.

(sin(x)) + sin(x) = (cos(x)) + sin(x) = − sin(x) + sin(x) = 0.

′′ ′
✓

Thus sin(x) is indeed a solution. But then one should immediately recognize that this
game can also be played with cos(x). That is, since (cos(x))′ = − sin(x) and
(sin(x)) = cos(x), we have (cos(x)) = (−sin(x)) = −cos(x). Or,
′ ′′ ′

(cos(x)) + cos(x) = (− sin(x)) + cos(x) = −cos(x) + cos(x) = 0.

′′ ′
✓

Then one could realize that it is actually easy to construct many more solutions. For
instance, try y = 2cos(x) − sin(x):

(2cos(x) − sin(x)) + (2 cos(x) − sin(x))

′′

= (−2 sin(x) − cos(x)) + (2 sin(x) − sin(x))

′

= −2 cos(x) + sin(x) + 2 cos(x) − sin(x) = 0. ✓

Obviously there is nothing special about the coefficients 2 and −1, and this same trick
works with c1 cos(x) + c2 sin(x) for any c1 and c2. So we have that c1 cos(x) + c2 sin(x) is a
solution to y′′ + y = 0 for any choice of constants c1 , c2.

But still there is the question: is this all of the possible solutions? And now the answer
′′
is yes. All solutions to y + y = 0 are of the form y = c1 sin(x) + c2 cos(x), where c1 and c2
are constants. However we are not in a position to prove this yet. It is actually more than
a matter of simply ‘solving’ the equation, but instead follows from the general theory of
′′
Chapter 3. This theory applies here, because the equation y + y = 0 is linear.

1.1.2 Linearity
Definition 1.1.18. A linear differential equation is an equation of the form
(n) (n−1)
an(x)y (x) + an−1 (x)y (x) + ... + a1(x)y (x) + a0 (x)y = f (x).
′

Example 1.1.19.

C. Sprouse —Math 280 Notes Page 9 DRAFT August 26, 2022

1.1. Differential Equations: Basic Notions and Definitions 10

′′ −x
• y + 3y = e is a linear 2-nd order ODE.
3
d y 2 dy
• dx3
−x dx
+ y = 0 is a linear 3-rd order ODE.
′
• cos(x)y + sin(x)y = x is a linear 1-st order ODE.
2
3
• t ddt 2x + 2 dx
dt
= t is a linear 2-nd order ODE.

• (x + 1)y + xy − 2y = sin(x) is a linear 3-rd order ODE.

2 ′′′ ′′

Example 1.1.20.

• y′′ + y2 = x3 is not a linear ODE.

• y′ + 2xy = sin(y) is not a linear ODE.

2
d y dy
• dx2
+ y dx = 1 is not a linear ODE.
dy y x
• dx
+ x
= y
is not a linear ODE.

To get a linear ODE in terms of a function y, one is only allowed to multiply y and its
derivatives by ai (x), which are functions of the of the independent variable x. One cannot
do things like square y, take functions such as sin(y) or multiply y or its derivatives by
each other.
There is a big difference in the way that linear and nonlinear ODE’s behave. For
instance we will soon see in this chapter that first order linear ODE’s are always explicitly
solvable, whereas there is no general method to solve a nonlinear ODE, unless it is a more
special type. In Chapter 3 we will see there is a general theory which describes how the
general solution must look for higher order linear ODE’s as well. Such an explicit theory
is far from possible for general nonlinear ODE’s.
′ 3
As a preview of some issues in Chapter 3, consider the example cos(x)y +sin(x)y = x
above. Recall from algebra that when dealing with polynomials, it is usually nicer to have
2
a 1 as the leading coefficient. For instance, to factor 3x − 21x + 30 = 0, the first thing one
would do would be to divide by 3 and get x − 7x + 10 = 0, or (x − 2)(x − 5) = 0. Similarly
2

for linear ODE’s it is often convenient to have a 1 as the coefficient of the leading term
before attempting a solution. We will call this standard form.
Definition 1.1.21. An n-th order linear equation is said to be in standard form if it is of the
form
(n) (n−1)
y + pn−1 (x)y + ... + p1 (x)y + p 0 (x)y = f (x).
′

So, to put
′′
cos(x)y + sin(x)y = x
in standard form, one would divide by cos(x) and get:
′′
y + tan(x)y = x sec(x).

C. Sprouse —Math 280 Notes Page 10 DRAFT August 26, 2022

1.1. Differential Equations: Basic Notions and Definitions 11

Similarly, we can put any equation in standard form by dividing by the coefficient
an (x). However, notice that now tan(x) and sec(x) are discontinuous functions, which
are not defined at x = π2 + nπ. These values of x are actually the zeroes of cos(x). The
equation in standard form becomes undefined at points where we would divide by zero.
This happens in general: we can convert any linear ODE to standard form by simply di-
viding by the leading coefficient an (x). However, the equation in standard form is always
undefined at points x where an(x) = 0. Such points are called singular. In Chapter 3 we
will see that we can only get the general solution to a linear ODE on (a, b) if the interval
does not contain any singular points.
We will also need to define standard form for nonlinear equations. For this we can
restrict to first order ODE’s:
Definition 1.1.22. A first order ODE is in standard form if it is of the form
dy
= F(x, y).
dx
Unlike for linear ODE’s, nonlinear ODE’s cannot always easily be written in standard
form. For instance consider (y ) + y = x . If we try to solve for y we get y = ± x 4 − y 3 ,
′ 2 3 4 ′ ′

which is actually two different standard form equations! Since most of our theory starts
with an equation in standard form, we will assume from now on that we only allow ODE’s
which are equivalent to a single standard form equation.
But even in this case, there is the issue of the equation not always being defined in
standard form. For example
dy 2
y +y = x
dx
has standard form
dy x
= − y.
dx y
The second equation is not defined at y = 0, whereas the original form of the equation is
defined for all (x, y).
Points where the equations in standard form are undefined, in both the linear and
nonlinear case, are important to be careful of when solving initial value problems. This
brings us to the last definition that we need to discuss.

1.1.3 Initial Value Problems

Definition 1.1.23. An n-th order Initial Value Problem (IVP) is an equation of the form
(n) (n−1) ⎫
⎪
⎪
′
⎪
, ..., y , y, x) = 0,
⎪
F(y ,y
y(x0) = b1, ⎪
⎪
⎪
⎪
y (x0) = b2,
′
⎬
⎪
⎪
(1.1)
⎪
⎪
⋮ ⎪
⎪
(n−1) ⎪
⎪
y (x0 ) = bn . ⎭
(n−1)
The equations y1 (x0 ) = b1, ..., y (x0 ) = bn are called the initial values or initial
conditions. Notice that the number of initial values is always exactly the same as the
order of the ODE, which is what we will need to determine a specific solution.

C. Sprouse —Math 280 Notes Page 11 DRAFT August 26, 2022

1.1. Differential Equations: Basic Notions and Definitions 12

Recall throwing a ball in the air, as seen in calculus or physics. One gets a unique
solution for the path of the ball if one knows the initial velocity and position at which the
ball was released. The unique path is defined by the solution to a very simple ODE, given
= −g, y(0) = y0 , y (0) = v0 .
d2 y ′
by the constant downward acceleration from gravity: dt2

y ft. 2
15 y(t) = − 12 gt + v0 t + y0

10

t sec.
0.5 1 1.5 2

This is actually true in general: for an n-th order ODE, you need n initial values to be
specified in order to be able to guarantee a unique solution. Fewer than that would allow
more than one solution, while more might not allow any solution at all. This is easy to see
in concrete examples.

Example 1.1.24. Solve the initial value problem

y (π) = −1.
′′ ′
y + y = 0, y(π) = 2,
′′
solution: We have previously discussed that the solutions to y + y = 0 are of the form
y = c1 cos(x) + c2 sin(x). So we are looking for a function that satisfies y(π) = 2 and
y (π) = −1, where y = c1 cos(x) + c2 sin(x) and hence y = c1 sin(x) − c2 cos(x). That is,
′ ′

y(π) = 2 ⇒ c1 cos(π) + c2 sin(π) = 2 ⇒ c2 = −2

y (π) = −1 ⇒
′
−c1 sin(π) + c2 cos(π) = −1 ⇒ c1 = 1

Thus we have that y = −2 cos(x) + sin(x) is the solution to the IVP.

3
2 ⋅
1

π π 3π 2π
−1 2 2

−2
−3

Note that this worked nicely because we had two initial values which give two equa-
tions, for the two unknown constants c1 , c2 . The fact that there were two constants is not
a coincidence, but a general property of second order linear ODE’s. But again we need to
be specific about what we mean by ‘solution’. For situations like the above example, it is
very straightforward:

C. Sprouse —Math 280 Notes Page 12 DRAFT August 26, 2022

1.1. Differential Equations: Basic Notions and Definitions 13

(n) (n−1)
, ..., y , y, x) = 0 on (a, b),
′
Definition 1.1.25. Suppose y is a solution to the ODE F(y , y
and x0 ∈ (a, b). Then y is a solution to the IVP (1.1) on (a,b) if y also satisfies the initial values
at x0 .
Example 1.1.26. The general solution to x y − 2y = 0 on (0, ∞) is y = c1 x
2 ′′ −1 2
+ c2 x . Show
that there is a unique solution to the IVP on (0, ∞):

y (1) = 0.
2 ′′ ′
x y − 2y = 0, y(1) = 3,

solution: Note that 0 < 1 < ∞, so 1 ∈ (0,∞). If y solves the ODE x y − 2y = 0 on (a, b)
2 ′′
−1 2 ′ −2
then from the general solution, y = c1 x + c2 x , and then y = −c1x + 2c2 x. So
y(1) = 3 ⇒ c1(1) + c2 (1) = 3 ⇒ c 1 + c2 = 3
y (1) = 0 ⇒ −c1(1) + 2c2 (1) = 0 ⇒
′
−c1 + 2c2 = 0.
Adding the two equations gives 3c2 = 3, so c2 = 1. Then the second equation gives
c1 = 2c2 = 2. So the unique solution to the IVP on (0, ∞) is
−1 2
y = 2x +x .

8
7
6
5
4
3 ⋅
2
1

1 2 3 □.

The reason we need to be so careful about the precise meaning of ‘solution’ here is
that the following situation also comes up fairly often. It needs a separate definition.
(n) (n−1)
, ..., y , y, x) = 0 on (a, b),
′
Definition 1.1.27. Suppose y is a solution to the ODE F(y , y
and x0 = a. Then y is a solution to the IVP (1.1) on [a,b) if y satisfies the initial values at
x0 = a, and y is also continuous on [a, b).
Example 1.1.28. y = −1 + x is a solution on [0, ∞) to the IVP
dy
2x − y = 1, y(0) = −1.
dx
solution: We can see that y solves the ODE on (0, ∞) by plugging it in:

− y = 2x 0 + x 2 − (−1 + x) = x + 1 − x = 1.
dy 1 −1
2x ✓
dx 2
Note that y is not differentiable at x = 0, so does not satisfy the differential equation there.
However, y is continuous on [0, ∞) and also satisfies

y(0) = −1 + 0 = −1,

so y is a solution to the IVP on [0, ∞).

C. Sprouse —Math 280 Notes Page 13 DRAFT August 26, 2022

1.1. Differential Equations: Basic Notions and Definitions 14

2
y = −1 + x

1 2 3 4

−1

−2 □.

Actually, the general solution to 2x dx − y = 1 on (0, ∞) is y = −1 + C x, where C is any

dy

constant. All of these solutions are continuous on [0,∞) and satisfy y(0) = −1. So the
above solution to the IVP is not unique. There are infinitely many solutions to the IVP on
[0, ∞):

3
y = −1 + C x
2

1 2 3 4
−1

−2

−3 □.

So the above gives an example of how the solution to an IVP might not be unique. It is
also possible for there to be no solution to an IVP. In this case we say that a solution does
not exist.
Example 1.1.29. There are no solutions to the IVP
′
2xy − y = 1,

y(0) = 2.

solution: From the above, the general solution to the ODE on (0, ∞) is y = −1 + C x. If
we try to solve y(0) = 2, we get
−1 + C 0 = 2.
Since −1 ≠ 2 independent of what C is here, there can be no solutions to this IVP. □.

We will show in Chapter 3 that this kind of thing does not happen away from a singular
point. That is, an nth order linear ODE with continuous coefficients and an (x) nonzero
on (a, b) will have a unique solution to any IVP on (a,b), as long as a < x0 < b.

C. Sprouse —Math 280 Notes Page 14 DRAFT August 26, 2022

 1.1. Differential Equations: Basic Notions and Definitions 15

Homework 1.1:
A) For each ODE, find the order and classify as linear or nonlinear.
1. 7.
(x + 1)y + xy = x
dy 2 ′′′ 4
3
x +y = x
dx
8.
2. dy
3 ′′
x y − 2xy = y
5 cos(x) + sin(x)y = 1
dx
3. 9.
(5)
2
3 d y 3x − y
y − y + 2y = x =
dx2 x + 3y
4.
5 (3) 2x 10.
x y +e y = x ′′ 1 x
y + xy =
x +1
5.
d 2y dy 11.
t +3 = cos(t) dx 3
dt 2 dt + tx = x
dt
6. 12.
3 3
dy x d x 2 −t
+y =e +t x = e
dx dt3
B) Show that each function is a solution to the IVP. On what interval is the solution defined?
13.
′ 2x
y −y = e , y(0) = 3.
x 2x
y = 2e + e .
solution: We need to check two things.

(2e + e ) − (2e + e ) = 2e + 2e
x 2x ′ x 2x x 2x x 2x
− 2e − e
2x
= e
0 0
y(0) = 2e + e = 3.
x 2x
The function y = 2e + e is differentiable and continuous everywhere, so is a solution
on(−∞, ∞) □..
14.
′ 2
y = 2(1 − x)y , y(1) = −1.
1
y= .
x 2 − 2x
solution: y = ((x − 2x) ) = −(x − 2x) (2x − 2) =
′ 2 −1 ′ 2 −2 2−2x 2
.and2(1 − x)y =
(x2 −2x)2
(2 − 2x) (x2 −2x)
1
2 =
2−2x
(x 2−2x)2
.So y ′ = 2(1 − x)y 2.Also y(1) 1
= x2 −2x 1
= 1−2 = −1, so y satisfies the
1 1
initialvalues.y(x) = x2 −2x = x(x−2) is discontinuous at x =
0 and x = 2. Thus y can only be a
solution to an IVP on (−∞, 0), (0, 2), or (2, ∞). Since the initial value y(1) = −1 is at x = 1,
and 1 ∈ (0, 2), y is defined on the interval (0, 2) □.
15.
′ 2
y = 2(1 − x)y , y(3) = 1.
3
y= .
x 2 − 2x

C. Sprouse —Math 280 Notes Page 15 DRAFT August 26, 2022

1.1. Differential Equations: Basic Notions and Definitions 16

′ 2
solution: Checking that y = 2(1 − x)y is the same as above. And this time,
3
y(3) = 9−6 = 33 = 1.y(x) = x2 −2x
3 3
= x(x−2) is discontinuous at x = 0 and x = 2. Thus y can only
be a solution to an IVP on (−∞, 0), (0, 2), or (2,∞). Since the initial value y(3) = −1 is at
x = 3, and3 ∈ (2, ∞), y is defined on the interval (2, ∞) □.
16. π
′
y = tan(x)y, y − = 2.
3
y = sec(x).
solution: y = (sec(x)) = sec(x) tan(x) and
′ ′
1 1
tan(x)y = tan(x) sec(x) = sec(x) tan(x).y − π3 = cos(π/3)
= .5
= 2.sec(x) is discontinuous at
x = ... − 3π , − π2 , π2 , 3π , ....The maximum interval containing − π3 on which sec(x) is continuous
is then(− 2 , 2 ). So y = sec(x) is a solution to the IVP on (− π2 , π2 ).
2 2
π π

17. 24.
′ 3
′
y + y = x, y(0) = 0. y + 2y 2 = 0, y(3) = 1.

y = x − 1 +e
−x
. 1
y=
(x − 2)2
.
18.
y (1) = 0. 25.
2 ′′ ′
x y − 2y = 0, y(1) = 3,
y (0) = 0.
′′ ′ ′
y = 2x
−1
+x .
2 y − 2xy − 2y = 0, y(0) = 1,
2
19. x
y=e .
(x − 1) y = 12y,
2 ′′
y(0) = 2.
26.
y (0) = 6.
′
y (−1) = −2.
′′ ′ ′
xy + y = 0, y(−1) = 0,
2
y=
(1 − x)3 y = ln(x ).
. 2

20. 27.
′ 2 ′
y =1+y , y(π) = 0, yy + x = 0, y(3) = 4.
√
y = tan(x). y = 25 − x2 .
21. 28.
(4 + 2x)y = y,
′ 2 ′
y = y + 4, y(0) = 0. y(0) = 2.
y = 2 tan (2x) . y = 4 + 2x.
22. 29.
(1 − x )y = 2xy,
′ 2 2 ′
y − y = 1, y(0) = 1. y(0) = 1.
π 1
y = tan x + . y= .
4 1 − x2
23. 30.
(1 − x )y = 2xy,
′ 2 2 ′
y =y , y(2) = 1. y(2) = 1.
1 3
y= . y= .
3−x x2 − 1

C. Sprouse —Math 280 Notes Page 16 DRAFT August 26, 2022

1.2. Separation of Variables 17

1.2 Separation of Variables

Now we’ll discuss how to solve an equation such as
dy
+ 3y = 0,
dx
which was introduced in the previous section. First, recall what is usually called u-
substitution in calculus class. For example, to evaluate the integral

x cos(x ) dx,
2

2
we set u = x . Then
1
du = 2x dx ⇒ dx = du.
2x
So

x cos(x ) dx = x cos(x )
2 2 1
du
2x
1
= cos(u)du
2
sin(u) + C = sin(x ) + C.
1 1 2
=
2 2
The crucial fact here is that du is not the same as dx. Instead we have du = 2x dx,
which comes from the chain rule. The general expression for du is

du
du = dx.
dx

This is the point of the dudx

notation, as created by Leibniz. Making the derivative appear
to act like a fraction gives an easy to remember formula. Now, here is how we can use this
to solve differential equations.
Example 1.2.1. Solve
dy
+ 3y = 0.
dx
solution: First we rewrite the equation as
dy
= −3y.
dx
To integrate, we need to get the y on the other side. So we divide both sides by y:
1 dy
y dx = −3.
dy
After integrating both sides dx, we can then use dx
dx = dy for the left-hand side.

1 dy 1
y dx dx = −3 dx ⇒ y dy = −3 dx.

C. Sprouse —Math 280 Notes Page 17 DRAFT August 26, 2022

1.2. Separation of Variables 18

So
ln ∣y∣ = −3x + C.
We want to solve for y. First we have

∣y∣ = e
−3x+C
.
Using the properties of exponentials,

∣y∣ = e
−3x+C −3x C −3x
=e e = Ae ,

where A = e . Finally, ∣y∣ is either y or −y depending on whether y is positive or

C

negative. So
−3x −3x
±y = Ae ⇒ y = ±Ae .
But ±A is simply a constant again, so we can say
−3x
y = Ce □.

The ODE’s that can be solved in this way are called ‘separable’:
Definition 1.2.2. A first order ODE is called separable if if can be written in the form
dy
g(y) = h(x),
dx
for some h(x),g (y).
They can always be solved as above. By simply integrating both sides:

✬ ✩
To solve a separable ODE:
1. First separate the variables, by writing the equation as

dy
g(y) = h(x).
dx

2. Integrate both sides by dx, using

dy
g(y) dx = h(x) dx ⇒ g(y)dy = h(x) dx.
dx

3. After integration, we have

G(y) = H(x) + C.

This is called an implicit solution. If initial values are given, then

use them to solve for C.

• It may or may not be possible to solve for y from the implicit so-
lution. In the case that it is, we also have the explicit solution

(H(x) + C).
−1
y=G
✫ ✪

C. Sprouse —Math 280 Notes Page 18 DRAFT August 26, 2022

1.2. Separation of Variables 19

Example 1.2.3. Solve the initial value problem

dy −2
= 2xy ,
dx
y(−1) = 2.

solution: We rewrite the equation as

2 dy
y = 2x.
dx
Integrating both sides dx,
2 dy
y dx = 2x dx.
dx
So using the u-substition formula (with y in the place of u) gives

2
y dy = 2x dx.

Or,
1 3 2
y = x + C.
3
So
3 2
y = 3x + C.
Plugging in the initial values,
3 2
2 = 3(−1) + C ⇒ C = 5.

So we have the implicit solution

3 2
y = x + 5.
Then solving for y gives the explicit solution
1
2
y = 3x + 5 3

4 2
1
3
y = 3x + 5
3

⋅ 2

−2 −1 1 2 □.

Both integrals above produce arbitrary constants +C, but we only include it on the right-
hand side because we can combine them into a single constant C. Also you may have
noticed that when we multiplied by 3, we wrote C instead of 3C. This is because 3C is
also an unknown constant, and for simplicity we can rename any arbitrary constant to C:

C. Sprouse —Math 280 Notes Page 19 DRAFT August 26, 2022

1.2. Separation of Variables 20

Any single arbitrary constant can be written as C. So when doing

2
algebraic operations, C , 5C, ±C, etc. can be rewritten as C.

This example takes a little more work to separate the variables. That is, we must get
dy
the equation in the form g(y) dx = h(x) before integrating.
2 dy 2
Example 1.2.4. Find the general solution to x y dx + y = 1.

solution: First, put the y 2 on the other side:

2 dy 2
x y =1−y .
dx
2 2
Then we can divide both sides by 1 − y , and then by x , to get the equation in the right
form:
2 dy 2 dy 1 − y 2 y dy 1
x y = 1−y ⇒ y = 2
⇒ 2
= 2.
dx dx x 1 −y dx x
Integrating,

y dy 1 y −2
dx = dx ⇒ dy = x dx.
1 − y2 dx x2 1 − y2
2
The integral on the left can be solved by u = 1 − y , du = −2y dy. So we get

− ln ∣1 − y ∣ = − x + C.
1 2 1
2
Now let’s try to solve for y. Again, we rename the constant C each time that we need to:

ln ∣1 − y ∣ = x + C
1 2 1
2

ln ∣1 − y ∣ = x + C
2 2

∣1 − y ∣ = e x
2
2 +C 2/x C 2/x
=e e = Ce .
So
2
2 2/x
1 − y = ±Ce x = Ce
or 2
2 2/x
y = 1 − Ce x = 1 + Ce .
So the general solution is
y = ± 1 + Ce 2/x .

√
A choice of the ± sign can be determined if an initial condition is given. For in-
stance√ if y(1) = 3 the solution is y = 1 + 8e e . But if y(1) = −2, the solution is
−2 2/x

y = − 1 + 3e −2 e 2/x .

C. Sprouse —Math 280 Notes Page 20 DRAFT August 26, 2022

1.2. Separation of Variables 21

3⋅

−1

−2 ⋅

−3
1 2 3

Here is a similar example, with initial values.

Example 1.2.5. Solve the initial value problem
′ 2
y − y = 9,
y(0) = 3.
2
solution: Again, putting the y on the other side,
′ 2
y = y + 9,

we see that dividing both sides by y 2 + 9 gives the separated equation

1 dy
2
= 1.
y + 9 dx
Hence
1 dy 1
dx = 1 dx ⇒ dy = 1 dx
y2 + 9 dx y2 + 9
1 y
⇒ arctan = x + C.
3 3
The initial condition y(0) = 3 gives
1 3 1 1π
arctan = 0+C ⇒ C= arctan(1) = .
3 3 3 34
Solving for y,
y π π
arctan = 3x + ⇒ y = 3tan 3x + □.
3 4 4
15
3 tan 3x + π4 12
9
6
3 ⋅

π π
− 3π
12
− 2π
12
− 12 −3 12
−6
−9 □.

C. Sprouse —Math 280 Notes Page 21 DRAFT August 26, 2022

1.2. Separation of Variables 22

It is often not possible to invert H in H(y) = G(x) + C to solve for y. In this case we
must be satisfied with what is called an implicit solution.

Definition 1.2.6. An explicit solution to dx = f (x, y) is a solution where y is given as a

dy

function of x:
y = y(x).

= f (x, y) is relationship between x and y that any solution

dy
An implicit solution to dx
y(x) must satisfy:
Φ(x, y) = C.

Example 1.2.7. Solve the initial value problem

(y + 1)
4 dy
= 3 sin(x), y(π) = 2.
dx
solution: The equation is already separated, so we just integrate both sides.

(y + 1)
4 dy 4
dx = 3 sin(x) dx ⇒ y + 1 dy = 3sin(x)dx
dx
1 5
⇒ y + y = −3 cos(x) + C
5
5
⇒ y + 5y = −15 cos(x) + C.

Here we cannot solve for y, because unlike quadratics, there is no ‘quintic formula’ for
5th degree polynomials. We can however solve the for the initial values.
5
y(π) = 2 ⇒ 2 + 10 = −15 cos(π) + C ⇒ C = 27.

So the implicit solution is

5
y + 5y = −15 cos(x) + 27 □ .

2 ⋅

-π π 2π 3π

Example 1.2.8. Solve the initial value problem

(y + 1)y + 3x y = 0,
′ 2

y(2) = 1.

C. Sprouse —Math 280 Notes Page 22 DRAFT August 26, 2022

1.2. Separation of Variables 23

solution: Rewriting the equation gives

(y + 1)
dy 2
= −3x y,
dx

(1 + y )
−1 dy 2
= −3x .
dx
So
(1 + y ) (1 + y )dy = −3x dx.
−1 dy 2 −1 2
dx = −3x dx ⇒
dx
Or,
y + ln ∣y∣ = −x + C.
3

Since 1 > 0, we have ln ∣y∣ = ln(y) near the initial condition y = 1. Then plugging in
y(2) = 1,
3
1 + ln(1) = −2 + C ⇒ C = 9.
So the implicit solution is
3
y + ln(y) = 9 − x .

8
7
6
5
4
3
2
1 ⋅

−1 1 2

We left the solution in implicit form above, because it is not possible to solve for y in
terms of regular calculus functions. However if we take the exponential of both sides, we
get

3 3
y+ln(y) 9−x y ln(y) 9−x
e =e ⇒ e e =e
3
y 9−x
⇒ ye = e .
y
Again, this is also not solvable by algebra. But the inverse of ye comes up so often in
applications that it is given a name. It is called the Lambert-W function.

Definition 1.2.9. The Lambert-W function, W(x), is defined by

ye = x ⇔ y = W (x).
y

See Example 1.2.17 below for a graph of W (x).

C. Sprouse —Math 280 Notes Page 23 DRAFT August 26, 2022

1.2. Separation of Variables 24

This is a case where we can get an explicit solution by simply expanding our list of
known functions. The new function here is W (x). The explicit solution to (y + 1)y ′ +
3x y = 0 with y(2) = 1 can now be written as y = W (e ).
3
2 9−x

In the previous example we had to be a little careful about the absolute values. We
used y(2) = 1 to get ln ∣y∣ = ln(y). But if the initial values were instead y(2) = −1, then
at y = −1 we have ∣y∣ = 1 = −y. So these initial values would give ln ∣y∣ = ln(−y)! Here is
another example, where this sort of thing happens.
Example 1.2.10. Find an explicit solution to the initial value problem
dy 2
= 2−y −y ,
dx
y(0) = 3.
solution: Separating,
1 dy 1 dy
=1 ⇒ dx = 1 dx
2 − y − y 2 dx 2 − y − y 2 dx
1
(2 + y)(1 − y)
⇒ dy = x + C.

Recall that we can use partial fractions to evaluate the integral:

1 A B
(2 + y)(1 − y)
= + .
2 + y 1 − y

Rewriting as 1 = A(2 + y) + B(1 − y) and plugging y = −2,y = 1 gives A = 13 ,B = 13 . So

1 1
ln ∣2 + y∣ − ln∣1 − y∣ = x + C.
3 3 1 1
+ dy = x + C ⇒
2+ y 1−y 3 3
Here, the absolute values are not ignorable! Recall that absolute value works as

∣z∣ =
z if z≥0
.
−z if z<0
That is, the absolute value is accomplished by switching the sign of anything negative. In
situations like this, we can use the initial value to determine which signs need to be
switched. y(0) = 3, so
ln ∣2 + 3∣ = ln(5) = ln(2 + 3) ⇒ ln ∣2 + y∣ = ln(2 + y).
ln∣1 − 3∣ = ln(2) = ln(−(1 − 3)) ⇒ ln ∣1 − y∣ = ln(−1 + y).
The solution then becomes
1 1 2+y
ln(2 + y) − ln(−1 + y) = x + C ⇒ ln = 3x + C.
3 3 −1 + y
Or,
2+ y 3x
= Ce .
−1 + y
At this point, we can plug in y(0) = 3 to get C = 52 . Then with a little more algebra, we get
the explicit solution for y:
3x
5e + 4
y = 3x .
5e − 2

C. Sprouse —Math 280 Notes Page 24 DRAFT August 26, 2022

1.2. Separation of Variables 25

4
y = (5e
3x 3x
+ 4)/(5e − 2)

3 ⋅

−1 1 2
□.

Remark 1.2.11. In the next section (and often in the future), we will typically leave off the
absolute values. So we will write things such as ∫ 1x dx = ln(x) + C. But this means we are
implicitly assuming that x > 0. It is often okay to leave off absolute values, as long as one is
aware of what to do when it is not okay. The above example is a case where we cannot leave the
absolute values off, because the initial condition y(0) = 3 means ln ∣1 − y∣ cannot be ln(1 − y).
And if you tried to evaluate ln(1 − y) you would quickly get that ln(−3) does not exist!
Now we’ll discuss some other cases where we need to be careful to ensure a solution
y(x) satisfies the initial values. For instance we saw in Example 1.2.4 above that this
happens when taking square roots.
Example 1.2.12.
2
x ′ 3
e y = xy ,
y(0) = −1
solution: Separating variables,
x2 3 −3 dy x2
e y = −xy ⇒ y = −xe .
dx
So
−3 dy −x
2
−3 1 −x2
y dx = − xe dx ⇒ y dy = e + C.
dx 2
Or,
1 −2 1 −x2 −2 −x 2
− y = e +C y = −e + C.
2 2
So
1
y=± .
2
C − e −x
Now the initial condition y(0) = −1 means we must take the negative square root. And
then
−√
1
= −1 ⇒ C − 1 = 1 ⇒ C = 2.
C − e0

C. Sprouse —Math 280 Notes Page 25 DRAFT August 26, 2022

1.2. Separation of Variables 26

So the solution to the IVP is

1
y =− .
−x2
2− e

−0.5
y = −(2 − e −x )− 2
2 1
−0.6

−0.7

−0.8
−0.9

−1 ⋅
0 1 2 3 □.

One also might need to be careful with trigonometric functions. These are not invert-
ible unless one restricts the domain, which can cause issues when solving initial value
problems.

Example 1.2.13. Find the solution.

= cos (y),
dy 2
dx
y(0) = 3π.

solution:
sec (y) sec (y)
2 dy 2 dy
=1 ⇒ dy = 1 dx.
dx dx

sec (y) dy = x + C
2
⇒ ⇒ tan(y) = x + C.

The important subtlety in this example is that now we cannot say y = arctan(x + C)! The
tangent function is periodic, and not invertible (it doesn’t satisfy the horizontal line test).
The inverse is y = arctan(x + C) only if − π2 < y < π2 (the range of arctan). But here we
have y(0) = 3π which is clearly not in this interval.
To deal with this, we can use that tangent has period π:

tan(y) = x + C ⇒ tan y + kπ = x + C.

In this way, there will be k so that

π π
y + kπ ∈ − , .
2 2

Since y(0) = 3π, we have 3π + kπ ∈ − π2 , π2 , and we can choose k = −3, to get

0 ∈ − π2 , π2 . Inverting,

y − 3π = arctan(x + C) ⇒ y = arctan (x + C) + 3π.

And y(0) = 3π gives arctan(C) = 3π ⇒ C = 0, so

y = arctan(x) + 3π.

C. Sprouse —Math 280 Notes Page 26 DRAFT August 26, 2022

1.2. Separation of Variables 27

7π
2

3π ⋅

5π
2

2π

3π
2

π
2

−3 −2 −1 1 2 3 □.

1.2.1 Solutions by definite integrals

It is very common with separation of variables to get integrals that cannot be solved. That
√
is, the antiderivative cannot be written in terms of regular calculus functions. Some well-
known examples of integrals like this are ∫ e dx, ∫ cos(x ) dx, ∫ x dx, ∫ x 3 + 1 dx,
2
x 2 sin(x)

∫ ln(x)
1
dx. In this case, we replace the indefinite integral (which represents a general an-
tiderivative) by a definite integral, which is a specific function given by the area under
a curve. This allows us to then solve an initial value problem, even when we can’t write
down a more explicit formula for the antiderivative.
Example 1.2.14. Solve the initial value problem
dy 2 −x
2
=y e ,
dx
y(1) = 2.
Express the answer with a definite integral.
2
1 dy −x
solution: We write y 2 dx
=e and integrate both sides:

1 dy −x
2
dx = e dx.
y 2 dx
2
−2 −x
y dy = e dx.
2
−x
We cannot integrate e . So we need to replace the indefinite integral on the right by a
definite integral.
x 2
−1 −s
−y = e ds + C.
1

C. Sprouse —Math 280 Notes Page 27 DRAFT August 26, 2022

1.2. Separation of Variables 28

We have to change the variable of integration to s, because x must now be an endpoint, so

is already taken. The lower endpoint also cannot be overlooked. It must be the initial
value x0 , or we will not be able to solve for C. With the integral as written we can solve
for C we plug in y(1) = 2:

)= −(2 ) = 0 + C.
1 2
−1 −s −1
−(2 e ds + C ⇒
1

So C = − 12 . Then we can solve for y as

1 2
y= = .
− ∫1 e −s ds 1 − 2 ∫1 e −s
1 x 2 x 2
2
ds

3
y(x)

2 ⋅

0.5 1 1.5 2

Example 1.2.15. Solve the initial value problem

dy √
x = y sin(x),
dx
y(π) = 9.
Express the answer with a definite integral.

solution: Separating,
1 dy sin(x)
= x .
y dx
So
1 dy sin(x) 1 x sin(s)
dx = x dx ⇒ dy = s ds + C
y dx y π
√ x sin(s)
⇒ 2 y= s ds + C
π

The initial values y(π) = 1 give

π sin(s)
2 9= s ds + C ⇒ C = 6.
π

Solving for y gives

x 2
sin(s)
y= s ds + 3 .
π

C. Sprouse —Math 280 Notes Page 28 DRAFT August 26, 2022

1.2. Separation of Variables 29

10
y(x)
9 ⋅

2π 3π 4π 5π 6π 7π 8π 9π 10π

1.2.2 Autonomous First Order ODE’s ∗

Recall that a first order ODE in standard form is dx = f (x, y). Such an ODE is called
dy

autonomous if the right-hand side does not depend on x. That is,

= f (y).
dy
dx
Such ODE’s are very common in applications, so we will give them special focus. (See
also Section 1.4.) It is easy to see that these ODE’s are always separable. Dividing by f (y),

1 dy 1 dy
f (y) dx f (y) dx
=1 ⇒ dx = 1dx

Or,
1
f (y)
dy = x + C.

We can get an explicit solution if:

1) We can find the antiderivative G(y) = ∫ 1

f (y)
dy.

2) We can invert G to get y = G −1 (x + C).

We’ve actually already done this several times. Many of our examples of separable
ODE’s above were autonomous! We’ve seen that for general f (y), both 1) and 2) above
are not guaranteed. But if we can at least solve the integral in 1), then 2) can easily be
done graphically which an explicit formula is not possible. This is because we know from
−1
precalculus that to graph G we just flip the graph of G along the diagonal line y = x.

Example 1.2.16. Find an implicit solution and use this to sketch y(x):
3
dy y
= ,
dx 2 + y 3

y(0) = 1.

C. Sprouse —Math 280 Notes Page 29 DRAFT August 26, 2022

1.2. Separation of Variables 30

solution: The separated form is

3
(2y
2 + y dy −3 dy
dy = 1 ⇒ + 1) = 1.
y 3 dx dx

Integrating gives
−2
−y + y = x + C,
comm and y(0) = 1 then gives C = 0. So the implicit solution is
−2
x = −y + y.

One might think this is easy to invert and solve for y(x), but this isn’t the case. One
3 2
would have to solve the cubic y − xy − 1 = 0 to do this, which is possible but not so easy
as it involves the cubic formula. But x = y − y−2 is easy to graph, because it is asymptotic
to x = y and we know it goes through x = 0 at y = 1. Then we can get a graph of the
solution y(x) by flipping along y = x, as we see below on the right.
3 x −2
x = y−y 3 y Solution y(x)

2 2

y =x y =x

1 1 ⋅

y x
⋅
−1 1 2 3 −1 1 2 3

−1 −1 □.

Example 1.2.17.
−y
dy e
= ,
dx y + 1
y(e) = 1.

solution: Separating,

(y + 1)e
y + 1 dy y dy
=1 ⇒ = 1.
e −y dx dx
So
(y + 1)e (y + 1)e dy = x + C.
y dy dy y
dx = 1 dx ⇒
dx dx
The left-hand side can be integrated by parts:

(y + 1)e = (y + 1)e − e dy = (y + 1)e − e + C = ye + C.

y y y y y y

So an implicit solution is
y
ye = x + C.

C. Sprouse —Math 280 Notes Page 30 DRAFT August 26, 2022

1.2. Separation of Variables 31

1
Plugging in the initial condition, 1e = e + C, so C = 0, and
y
x = ye .

Again, we can flip along the line y = x to get the solution on the right:
5 x x = ye
y 5 y Solution y = W (x)

4 4
y=x y=x

3 3
⋅ (1,e)

2 2

1 1 ⋅
(e, 1)
y x
1 2 3 4 5 1 2 3 4 5
As we have mentioned above, this function has a name. The Lambert W-function is
y y
defined as the inverse of ye . That is, if G(y) = ye , then the Lambert W-function is
defined as W(x) = G (x). So the above solution can also be written as
−1

y = W (x) □ .

C. Sprouse —Math 280 Notes Page 31 DRAFT August 26, 2022

 1.2. Separation of Variables 32

Homework 1.2:
A) Solve for an explicit solution y(x).

√
1.
(x + 1)
4 dy
= x y.
dx
dy
solution: We need to write the equation in the form g(y) dx = h(x). So divide by y and
x4 + 1:
1 dy x
= .
y dx x 4 + 1
Now integrate both sides dx. (Do not skip this step.)

1 dy x
dx = dx.
y dx x4 + 1
dy
Then use dx
dx = dy to simplify the left hand side.

1 x
dy = 4
dx.
y x +1
Now you can integrate both sides. The integral on the right is solved by
2 1
u-substitution:u = x , x4x+1 dx = (x 2 )x2 +1 2x du = 12 u 21+1 du.

√ 1 2 1 2
2
2 y = arctan x + C ⇒ y= arctan x + C □.
2 4

3 dy 2 dy
2. x −y = 1 6. y = 2x y 2 − 1
dx dx
2 dy 6 dy 1 − y2
3. y = 1+ y 7. =
dx dx x
4. (x + 1)
2 dy 2 dy 2
+y = 0 8. x + y ln(x) = 0
dx dx
x dy √ dy x 2 −2y
5. e + y=0 9. = xe
dx dx
B) Find the explicit solutions y(x) to the initial value problems.
10. 14.
= cos (y),
dy dy 2 π
x = 3y, y(2) = 32. x y(1) = .
dx dx 4
11. 15.
′ dy 1
y − 5y = 10, y(0) = 1. = 2x 1 − y 2 , y(0) = .
dx 2
12. 16.
dy 2 dy 2
= 4x y + 1, y(−1) = 3. x − y = 1, y(1) = 0.
dx dx
13. 17.
(x + 1)
2 dy 2 dy 2
+ xy = 0, y(0) = 2. 2 +1=y , y(0) = 0.
dx dx

C. Sprouse —Math 280 Notes Page 32 DRAFT August 26, 2022

 1.2. Separation of Variables 33

C) Find the explicit solution y(x) by separation of variables. Simplify so that there are no logarithms in your
final answer.
18.
(x > 0)
dy
x = y(5 − x)
dx
1 dy 5−x
solution: Separate variables: y dx
= x
, or

1 dy 5 1 dy 5
y dx = x − 1 ⇒ y dx dx = x − 1dx
1 5
⇒ y dy = x − 1 dx
⇒ ln∣y∣ = 5 ln ∣x∣ − x + C = 5 ln(x) − x + C
⇒ ∣y∣ = e
5 ln(x)−x+C

ln(x ) −x
⇒ ∣y∣ = e
5
5 ln(x) −x C
e e = Ce e
5 −x 5 −x
⇒ y = ±Cx e = Cx e □.

19. 28.
= 3y. (x > 0) = (x + 2)(y − 3). (x > 0)
dy dy
x x
dx dx
20. 29.
(x > 0)
dy
(x + 1) (x > −1)
x + y = 0. dy
dx = xy.
dx
21.
30.
(x > 0)
dy
(x − 2) = (x + 1)y. (x > 2)
x + 2y = 1. dy
dx
dx
22.
(x > 0)
dy 31.
+ 6 = 3y.
(− <x < )
x dy π π
dx cos(x) = 3y sin(x).
23. dx 2 2
(1 + .2x) + 3 = y. (x > −5)
dy 32.
dx
+ 2x sin(y) = 0. (− < y < )
24. dy π π
cos(y)
(x + 1)
2 dy dx 2 2
= xy.
dx
33.
25. dy
2 dy 3 = y(y − 3), y(0) = 1.
y +1 = y . dx
dx
26. 34.
dy
= y(4x − 1). (x > .25)
dy 2 = y(2 − y), y(0) = 3.
x dx
dx
27. 35.
+ xy = 3y. (x > 0)
dy dy 2
x = y − 1, y(0) = 0.
dx dx

C. Sprouse —Math 280 Notes Page 33 DRAFT August 26, 2022

 1.2. Separation of Variables 34

D) Find the solutions to the IVP’s in implicit form (Do not solve for y!).
36.
dy 3
y = x sec(y)
dx
y(2) = π
dy 3
solution: Separating variables: y cos(y) dx =x .

dy 3 3
y cos(y) dx = x dx ⇒ y cos(y)dy = x dx
dx
1 4
→ y sin(y) + cos(y) = x + C.
4
We cannot solve explicitly for y, but we can still use y(2) = π to find C.

π sin(π) + cos(π) = (2 ) + C
1 4
⇒ −1 = 4 + C.
4
So
1 2
y sin(y) + cos(y) = x − 5 □.
4

37. 42.
(2y − 3)
dy 2x − 1 dy 2
= , y(2) = 3. + 2xy = 0, y(−1) = 1.
dx y 2 + 1 dx
38. 43.
y (−2) = π.
dy ′
y = sec(y), yy − xy = x, y(2) = 0.
dx
39. 44.
dy 2x−y
2 dy y y = 4e , y(0) = 0.
x y =e , y(−1) = 0. dx
dx
40. 45.
dy dy 2
ln(y) = 3 sin(x), y(0) = 1. y = y − 3y + 2, y(0) = 3.
dx dx
41. 46.
(2 − 7y) (y + y + 1)
dy 2 2 dy 2
+ x y = 0, y(3) = 1. = y(y + 1), y(0) = 1.
dx dx

C. Sprouse —Math 280 Notes Page 34 DRAFT August 26, 2022

 1.2. Separation of Variables 35

E) Write the explicit solution to the IVP using a definite integral.

47.
1
dy cos x
=
dx 3 y
y(3) = 4
dy 1
solution: Separate variables: 3 y dx = cos x
.

√ dy 1
3 y dx = cos x dx.
dx

√ 1
3 y dy = cos x dx.

The right-hand side integral cannot be evaluated directly, so we replace it by a definite

integral.
√ x
1
3 y dy = cos dt + C.
3 t
3 x 1
2y 2 = cos dt + C.
3 t
Using the initial conditions,

2(4 2 ) =
3 3 1
cos t dt + C ⇒ 16 = 0 + C.
3

Solving for y,
2
1 x 1 3
y= cos dt + 8 □.
2 3 t

48. 52.
2 dy x dy 2
x =e , y(1) = 5. ln(x) =y , y(e) = .25.
dx dx

dy √
49. 53.
= y sin(x ),
dy 2
x = sin(x), y(π) = 1. y(π) = 9.
dx dx

dy √
50. 54.
= cos(x ),
2 dy 2
y y(0) = 2. 2y = 1 + x3 , y(1) = 3.
dx dx
51. √ dy
55.
dy 2
x −y
1 + x3 = y, y(0) = e. =e , y(1) = 0.
dx dx

C. Sprouse —Math 280 Notes Page 35 DRAFT August 26, 2022

1.3. Linear First Order ODE’s 36

1.3 Linear First Order ODE’s

We introduced linear ODE’s in the first section. From the definition given there, we have
that a first order ODE is linear if it is of the form

a1 (x) + a0 (x)y = f (x).

dy
dx
To solve these, we will usually need to divide by a1(x) to make the coefficient of the
leading term be 1. For linear ODE’s, we have previously defined this as standard form.

Definition 1.3.1. A first order ODE is called linear, if it can be written in the standard form

dy
+ p(x)y = q(x).
dx
Example 1.3.2. These are all linear first order ODEs:
dy 3
1. + x y = 1.
dx
′
2. y = 2y + x.
′ 3
3. xy + 5y = x .
3
solution: The first has p(x) = x , q(x) = 1.
′
The second can be written y − 2y = x, so p(x) = −2, q(x) = x.
′ 2 2
The third can be written y + 5x y = x , so p(x) = 5x , q(x) = x .

Example 1.3.3. These are not linear first order ODEs:

dy 2
1. + xy = 1.
dx
′ x
2. y = 2y + y .

′ 2
3. yy − y = x .

Before discussing the general method of solution, we can give a brief demonstration.
Consider for instance the linear ODE
dy
+ 3y = x.
dx
3x
Now multiply both sides by e :

3x dy 3x 3x
e + 3e y = xe .
dx
Then, the trick is to notice that the product rule gives (e y) = 3e y + e y , which is
3x ′ 3x 3x ′

equal to the left-hand side of the equation above. This means that the equation can be
rewritten as

C. Sprouse —Math 280 Notes Page 36 DRAFT August 26, 2022

1.3. Linear First Order ODE’s 37

d 3x 3x
e y = xe .
dx
But when written in this form, it is now easy to solve the equation by integrating:

3x 3x
e y = xe dx
1 3x 1 3x
= xe − e dx
3 3
1 3x 1 3x
= xe − e + C.
3 9

And then we get y just by dividing by e 3x :

1 1 −3x
y= x − + Ce .
3 9
3x 3x
So the obvious question is, where did the e come from? Here, e is called an inte-
grating factor. In general, an integrating factor is any function that you can multiply an
ODE by, such that it can then be solved by integrating both sides. A significant feature of
linear first order ODE’s is that there is a simple formula which always gives an integrating
factor.

For the linear first order ODE

dy
+ p(x)y = q(x),
dx
the integrating factor is

∫ p(x) dx
µ(x) = e .

To see why this is, the chain rule gives

′ ′
dµ ∫ p(x) dx ∫ p(x) dx ∫ p(x) dx
= e = p(x) dx e = p(x)e = p(x)µ(x).
dx

dµ
⇒ = p(x)µ.
dx

So again our goal is to solve the equation

dy
+ p(x)y = q(x).
dx
After multiplying by µ(x),

C. Sprouse —Math 280 Notes Page 37 DRAFT August 26, 2022

1.3. Linear First Order ODE’s 38

dy dy dµ
µ + pµy = µq ⇒ µ + y = µq.
dx dx dx
But now the product rule gives

(µy) =
d dµ dy dy dµ
y +µ =µ + y,
dx dx dx dx dx
which is the left-hand side of 1.3. So 1.3 becomes

(µy) = µq.
d
dx

This can be integrated to get y:

(µy) dx =
d
µq dx ⇒ µy = µq dx + C.
dx
Dividing by µ gives an explicit formula for y:
1 C
y(x) = µ(x)q(x)dx + .
µ(x) µ(x)
Warning 1.3.4. Do not just memorize the formula above. The final answer is not of much
interest, unless the work is shown. What is important here is to follow the correct procedure,
✬ ✩
which we summarize below.

dy
To solve + p(x)y = q(x) ∶
dx

1. Find µ(x) = e ∫ p(x) dx and simplify.

dy
2. Multiply by µ to get µ(x) dx + µ(x)p(x)y = µ(x)q(x), then rewrite
as
′
µ(x)y = µ(x)q(x).

3. Integration gives

µ(x)y = µ(x)q(x) dx + C.

Complete the integral and divide by µ(x) to get y.

✫ ✪
Warning 1.3.5. Do not forget the +C! It is a crucial part of the solution. And when dividing
by µ, make sure to divide C by µ(x) as well:
1 C
y=µ µq dx + .
µ(x)
Forgetting these things is a very common mistake, which completely changes the nature of the
solution!

C. Sprouse —Math 280 Notes Page 38 DRAFT August 26, 2022

1.3. Linear First Order ODE’s 39

Example 1.3.6. Solve the initial value problem

dy x
+ 2y = 6e ,
dx
y(0) = 5.
∫ 2d x 2x
solution: 1) The integrating factor is µ(x) = e =e .
2) So,

2x dy 2x x 2x 3x
e + 2e y = 6e e = 6e ,
dx
which we rewrite as
2x ′ 3x
e y = 6e .
3) Then integrating,
2x 3x 3x
e y= 6e dx = 2e + C.
2x
And dividing by µ = e ,
x −2x
y = 2e + Ce .
The initial value y(0) = 5 gives 2 + C = 5, so C = 3 and
x −2x
y = 2e + 3e □.

30 x −2x
2e + 3e
25
20
15
10
5
0
0 0.5 1 1.5 2 2.5

Example 1.3.7. Find the general solution to y ′ + 3x y = x.

solution: The integrating factor is more interesting this time.
∫ 3x dx ln(x )dx
3
3ln(x)dx 3
µ(x) = e =e =e =x .
So
3 dy 2 4
x + 3x y = x ,
dx
which we rewrite as
d 3 4
x y =x .
dx
And upon integrating,
3 1 5
x y = x + C.
5
Or,
1 2 −3
y= x + Cx □ .
5

C. Sprouse —Math 280 Notes Page 39 DRAFT August 26, 2022

1.3. Linear First Order ODE’s 40

3
1 2
5
x + xC3
2

−1
1 2 3 4

Example 1.3.8. Find the general solution to

′
xy − y = 2.
∫ −1 dx −x
First of all, the integrating factor is not e =e . Because then the solution would
be given by
(e
−x ′ −x −x −x
y) = 2e ⇒ e y = −2e + C,
x
⇒ y − −2 + Ce .
But this is clearly not a solution to the ODE, as plugging it in gives
xy − y = x(Ce ) + 2 − Ce = C(x − 1)e
′ x x −x
+ 2 ≠ 2.
∫ p(x) dx
The reason is that the formula µ = e requires the ODE to be written in standard
dy
form: dx + p(x)y = q(x).
Warning 1.3.9. Before solving the linear ODE a1 (x) dx +a0 (x)y = f (x) by integrating factors,
dy

you must always divide by the leading term a1(x), so that the ODE is in standard form:
dy
dx
+
p(x)y = q(x).
solution: First divide by x to put the equation in standard form:
′ 1 2
y − xy = x.
then
)
1 −1
−∫ dx − ln(x) ln(x −1
µ(x) = e x =e =e =x .
So then

(x
−1 ′ −2 −1 −2 −1
y) = 2x ⇒ x y=2 x dx = −2x +C
⇒ y = −2 + Cx □ .

1
−2 + Cx
0

−1

−2

−3
0 1 2 3 4

C. Sprouse —Math 280 Notes Page 40 DRAFT August 26, 2022

1.3. Linear First Order ODE’s 41

The integrating factor µ(x) often needs to be simplified to be of any use. It is crucial to
follow the exponential and logarithm rules correctly when doing the algebra! Now we’ll
do some examples where µ(x) is more interesting.

Example 1.3.10. Solve the initial value problem

(x + 2)
dy −x
+ xy = e
dx
y(0) = 1.

solution: The equation in standard form is

−x
dy x e
+ y= .
dx x+2 x +2

x x u−2
dx = u dx = u du
x +2
2
= 1 − u du = u − 2ln(u) = x + 2 − 2 ln(x + 2) + C

For the integrating factor µ(x), the +C will always cancel if we include it. So we can
ignore it by setting it to 0. Then

(x+2)−2 ln(x+2) x+2 ln((x+2)−2 )

= (x + 2)
x+2 −2ln(x+2) −2 x+2
µ(x) = e =e e =e e e .

So now
−x 2
(x + 2) (x + 2) e
−2 x+2 ′ e −2 x+2 e
(x + 2)−3
e = = .
x+2

e2 e2
(x + 2) (x + 2) + C.
−2 x+2 −2
y= =−
(x + 2)−3
e
2
2
e −(x+2) 2 −(x+2) 1 −x 2 −(x+2)
y=− e + C(x + 2) e = e + C(x + 2) e .
2 2
−2 2
The initial condition y(0) = 1 ⇒ − 12 + 4Ce = 1 gives C = 38 e . So

y = − e + (x + 2) e = e (3x + 12x + 8) □ .
1 −x 3 2 −x 1 −x 2
2 8 8

2
1 e−x (3x 2 + 12x + 8)
8

0
0 1 2 3 4

C. Sprouse —Math 280 Notes Page 41 DRAFT August 26, 2022

1.3. Linear First Order ODE’s 42

Example 1.3.11. Find the general solution for −1 < x < 1.

(1 − x )
2 dy
− xy = 1.
dx
solution:
dy x 1
− y= .
dx 1 − x2 1 − x2
x
−∫ dx
µ(x) = e 1−x 2 .
2
From the substitution u = 1 − x we have
x x −1
− dx = − u 2x dx
1 − x2

u du = 2 ln ∣u∣ + C = 2 ln(1 − x ) + C.
1 1 1 1 2
=
2
Again, we don’t need the +C in the integrating factor, so
√
ln(1−x2 )
ln((1−x 2) 2 )
1
1
µ(x) = e 2 =e = 1 − x2 .
√
√ 1 − x2
( 1 − x )y = =√
′ 1
2
2
.
1− x 1 − x2
√
( 1 − x 2 )y = √
1
dx = arcsin(x) + C.
1 − x2
So
y= √ +√
arcsin(x) C
□.
2
1−x 1 − x2

3
√
arcsin(x)+C
2 1−x2
1
0
−1
−2
−3
−1 −0.5 0 0.5 1

Finally, it is very easy to produce examples where the integral ∫ µ(x)q(x)dx cannot be
explicitly solved. In this case, we must write the solution in terms of definite integrals. We
already saw this in the previous section for separation of variables. So to see how this is
done, refer to subsection 1.2.1, as well as the examples below.
Example 1.3.12. Solve the initial value problem in terms of a definite integral.
dy
+ .02xy = sin(x).
dx
y(0) = 2

C. Sprouse —Math 280 Notes Page 42 DRAFT August 26, 2022

1.3. Linear First Order ODE’s 43

solution:
∫ .02x dx .02 2 2
µ(x) = e =e 2 x =e
.01x
.
So

(e
2 2 2 2
.01x ′ .01x .01x .01x
y) = e sin(x) ⇒ e y = e sin(x) dx
x 2
−.01s
= e sin(s) ds + C.
0
2 x 2 2
−.01x .01s −.01x
y=e e sin(s) ds + Ce .
0

0 2
0 .01s 0
y(0) = 2 ⇒ e e sin(s)ds + Ce = 2.
0
⇒ 0 + C = 2 ⇒ C = 2.

So
2 x 2
−.01x .01s
y(x) = e e sin(s)ds + 2 □ .
0

4
y(x)
3

5 10 15 20 25 30
−1

Example 1.3.13. Solve the initial value problem in terms of a definite integral.

(x + 1)
3 dy 2
+ 1.5x y = 20.
dx
y(2) = 1.

solution: Putting the equation in standard form,

dy 1.5x 2 20
+ 3
y= .
dx 1 +x 1 + x3

The integrating factor is

1.5 ∫ x
2
1.5 3 3
ln(x +1) 2
1
√
dx ln(x +1)
µ(x) = e x 3+1 =e 3 =e = x 3 + 1.

√
√
So
x3 + 1
=√
′ 20
x 3 + 1y = 3
.
x +1 x3 + 1
√
√ √
20 x 20
⇒ x 3 + 1y = dx = ds + C.
3
x +1 2 s3 + 1

C. Sprouse —Math 280 Notes Page 43 DRAFT August 26, 2022

1.3. Linear First Order ODE’s 44

Dividing by µ(x),
y= √ √ ds + √
1 x 20 C
.
x3 + 1 2 s3 + 1 x3 + 1

√ √ ds + √
2
1 20 C
y(2) = 1 ⇒ =1
23 + 1 2 s3 + 1 23 + 1
C
⇒ 0 + = 1.
3
So C = 3 and the solution to the IVP is

y=√ √ ds + √
1 x 20 3
□.
x3 + 1 2 s3 + 1 x3 + 1

1.5
y(x)

0.5

2 4 6 8

2 √
Remark 1.3.14. It may be surprising, but like e x and sin(x) x
, the function 1/ x 3 + 1 cannot
√ calculus functions. That is, it does not have an elementary
be integrated in terms of standard
antiderivative. Integrals with p(x), where p(x) is a polynomial of degree 3 or 4 are typically
of this type. Their antiderivatives are called elliptic integrals.

C. Sprouse —Math 280 Notes Page 44 DRAFT August 26, 2022

 1.3. Linear First Order ODE’s 45

Homework 1.3:
A) Use the method of integrating factors to find the general solution.
1.
xy + (2 − 2x )y = 1
′ 2

solution: First rewrite as y′ + (2x−1 − 2x)y = 1x . Then the integrating factor is

∫ 2x ln(x ) −x
−1 2 2 2 2
−2x dx 2(ln(x))−x 2 −x
µ(x) = e =e =e e =x e .So

(x e
2 2
2 −x ′ −x
y) = xe

2 −x
2 1 −x 2
x e y =− e +C
2
2
1 Cex
y=− + □.
2x2 x2

2. 11.
+ (2 + x)y = 1
′ −3x dy
y + 5y = e x
dx
3.
′ 12.
y −y = x ′
xy + 2y = cos(x)
4.
′ 3x 13.
(x + 2)y + 3y = 3
y − 3y = e ′

5.
dy 2 14.
(x − 1)
+ y =3 dy
dx x + xy = 1
dx
6.
dy 1 15.
xy + (3 + 2x )y = x
− y = ln(x) ′ 2 1
dx x
7. 16.
dy
(x + 1)
x + 5y = 16x
3 2 dy
dx + xy = x
dx
8. 17.
dy
(x − 1)
x − 3y = x
2 2 dy
dx + 2y = 1
dx
9. 18.
dy x
3
dy
x + 3y = e cos(x) + sin(x)y = 1
dx dx
10. 19.
dy 2 2 dy
+ 3− x y = x + sec(x)y = cos(x)
dx dx

C. Sprouse —Math 280 Notes Page 45 DRAFT August 26, 2022

 1.3. Linear First Order ODE’s 46

B) Use the method of integrating factors to solve the IVP’s.

20.
′
xy − 3y = x, y(2) = 1.
solution: Rewriting, y + (−3/x)y = 1, so the integrating factor is
′

)
µ(x) = e ∫ −3/x dx = e −3 ln(x) = eln(x
−3
= x −3.

(x
−3 ′ −3
y) = x ,

−3 1 −2
x y = − x + C,
2
1 3
y = − x + Cx ,
2
y(2) = 0 so − 12 (2) + C(8) = 1 ⇒ C = 14 . So y = − 12 x + 14 x3 □ .

21. 30.
(x + 1)y + 3y = 3,
′ 3x ′
y − y = 2e , y(0) = 3. y(1) = 2.
22. 31.
(x − 1)y − y = x,
′ ′
y + 2y = 4x, y(0) = 1. y(2) = 1.
23. 32.
y + x y = 2 sin(x ), (2x + 1)y + y = 3x,
′ 1 2 ′
y( 2π) = 0. y(0) = 1.
24. 33.
1 1 xy + (2 − x)y = e ,
′ x
′
y + y = x, y(4) = 0. y(2) = 0.
2x
34.
xy + (2x − 1)y = 6x ,
25. ′ 2
′
y + 2xy = 2x, y(0) = 3. y(1) = 0.

26. 35.
(x − 1)
′ 3 2 dy 2
xy + 5y = x , y(1) = 0. + 2xy = x , y(2) = 1.
dx
27. 36.
′ 3
(x + 9)
xy − 2y = x , y(2) = 4. 2 dy
− xy = x, y(0) = 5.
dx
28.
′ 3 π2 37.
′ π
xy − 2y = x sin(2x), y(π) = . y + sin(x)y = sin(x), y = −1
2 2
29. 38.
′ cos(x) ′ π
xy + 3y = y(π) = 0. y + tan(x)y = cos(x), y = 0.
x , 4

C. Sprouse —Math 280 Notes Page 46 DRAFT August 26, 2022

 1.3. Linear First Order ODE’s 47

C) Find the solution y(x) to the initial value problem in terms of a definite integral.
′
39. xy − 3y = sin(x), y(2) = 24
solution: The equation is rewritten as y − (3/x)y = sin(x)/x. The integrating factor
′

)
−3
−3 ln(x) ln(x −3
isµ(x) = e =e =x . So

(x
−3 ′ −4
y) = x sin(x).
x
−3 −4
x y= t sin(t) dt + C
2
x
3 −4 3
y=x t sin(t) dt + Cx
2
y(2) = 24 gives 24 = 8(0) + C(8), or C = 3. So
x
3 −4 3
y=x t sin(t)dt + 3x □ .
2

40. 43.
′ ′
y − 2xy = 1, y(1) = 2e xy + 2y = sec(x), y(3) = −1.

41. 44.
′ 2 ′ 1 x
2
y + cos(x)y = x , y(π) = 3. xy + y = e , y(4) = 1.
2
42. 45.
(x + 8)y + 2x y = 1,
′ 2x 4 ′ 3
xy − y = e , y(3) = 6. y(−1) = 1.

C. Sprouse —Math 280 Notes Page 47 DRAFT August 26, 2022

1.4. Graphical Methods 48

1.4 Graphical Methods

1.4.1 Slope Fields

For first order ODE’s, we can in many situations determine what the solutions look like
without ever actually solving the ODE. We just need to look at what the equation means,
in terms of the slopes of the solutions. The first order ODE in standard form is

dy
= F(x,y).
dx

Suppose the function y = y(x) is a solution. Then at any fixed point x∗, we have

y (x∗ ) = F(x∗ ,y(x∗ )).

dy ′
= F(x,y(x)) ⇒
dx

That is, the slope of the solution y(x) at the point (x∗ , y(x∗ )) is F(x∗ , y(x∗ )), as indicated
in the figure below.

y(x)

tangent
line
1

F (x∗ ,y(x∗ ))

⋅
(x∗ ,y(x∗ ))

The triangle shows that the slope is

rise F(x∗, y(x∗))

run = = F(x∗ , y(x∗ )),
1

as required. To get a better visualization of this, we can replace this triangle by the tangent
vector, that points in the same direction as the tangent line:

v⃗ = ⟨1,F(x∗ , y(x∗ ))⟩.

C. Sprouse —Math 280 Notes Page 48 DRAFT August 26, 2022

1.4. Graphical Methods 49

y(x)

F (x∗ , y(x∗ ))

⋅
(x∗ ,y(x∗ ))

If now instead of just looking at the fixed point (x∗ , y∗) = (x∗ ,y(x∗ )), we think of what
is happening at each point (x, y), we get a vector field

⃗
v(x,y) = ⟨1, F(x, y)⟩ .

This is called the direction field or slope field of the ODE. We can sketch v(x, y) by
plotting the vector ⟨1, F(x,y)⟩ at any point (x, y). This is usually quite laborious to do
by hand, but can also be done by computer. For instance by typing ‘vector plot’ into
wolfram-alpha, you’ll get a window that lets you type in a vector field for plotting. Here
we use the java applet pplane.jar by John Polking, which is devoted specifically to this
type of application.

11

10

1 2 3 4

C. Sprouse —Math 280 Notes Page 49 DRAFT August 26, 2022

1.4. Graphical Methods 50

The arrows in the vector field above represent the tangent vectors of solutions, scaled
down for visualization. So now if we can graph the vector field given by an ODE with the
following correspondence

= ⟨1, F(x, y)⟩ ,

dy
= F(x, y) ⇒ ⃗
v(x,y)
dx

we can easily sketch solutions to the ODE by drawing curves which follow along the
vector field, making the vector field tangent to the curves. The above curve is a solution
dy
to dx = y − 2x. So the vector field is

= ⟨1, y − 2x⟩ .
dy
= y − 2x ⇒ ⃗
v(x,y)
dx

11
dy
10 dx = y − 2x

1 2 3 4

dy 2
Example 1.4.1. Use the direction field of dx
= xy to sketch the solutions of the following
initial value problems.

a) b) c)

= (x − 1)y , = (x − 1)y , = (x − 1)y ,

dy 2 dy 2 dy 2
dx dx dx
y(0) = 1. y(0) = 0. y(0) = −1.

solution: The vector field is below, on the left. One should try to start at each given
initial point, and follow the arrows to get a curve which has those arrows tangent to it.
The solutions are also below. Notice how the behavior of the solution can change quite a
bit, depending on the initial condition.

C. Sprouse —Math 280 Notes Page 50 DRAFT August 26, 2022

1.4. Graphical Methods 51

3 3

2 2

1 1 ⋅

1 2 3 1 2 3

−1 −1

−2 −2

−3 −3
3 3

2 2

1 1

⋅
1 2 3 1 2 3

−1 −1 ⋅

−2 −2

−3 −3

1.4.2 Autonomous ODE’s and sign charts

Again, graphing vector fields like the above can be tedious or difficult to do accurately by
hand. However, for most applications the vector field is too much information. To get an
initial rough sketch of solutions, we really only need to see where they are increasing or
decreasing. This is particularly easy for the class of ‘autonomous’ ODE’s.

Definition 1.4.2. A first-order ODE is called autonomous if it is of the form

= f (y).
dy
dx
That is, the right hand side does not depend on the independent variable x.

Autonomous ODE’s occur frequently in applications, so it is quite convenient to be

able to easily sketch their solutions. This can be done very quickly, and is much easier
than actually solving the ODE.
We can do this simply by making a sign chart, similar to what one uses to graph
dy
functions in Calculus I. Since d x = F(y), the sign of F(y) determines whether the solution

C. Sprouse —Math 280 Notes Page 51 DRAFT August 26, 2022

1.4. Graphical Methods 52

dy
y = y(x) is increasing or decreasing. That is, if F(y) > 0 then dx > 0, which says that any
solution y = y(x) is increasing when it passes through that y-value. Similarly, a solution
y(x) is decreasing for any y values with F(y) < 0.
Definition 1.4.3. For an autonomous ODE
dy
= F(y),
dx
A critical point is any value yc with F(yc ) = 0.
Critical points are of interest because they determine the constant solutions of an ODE.
dy
That is, if yc is a critical point of dx = F(y), then the function y(x) = yc is a solution. We
can check this by simply plugging in:

(y ) = 0 , F(yc ) = 0,
d
dx c
dy
⇒ = F(y) for y(x) = yc .
dx
dy 2
Example 1.4.4. Find the constant solutions to the ODE dx
= y − 7y + 10.

solution:
y − 7y + 10 = 0 (y − 2)(y − 5) = 0.
2
F(y) = 0 ⇒
So y(x) = 2 and y(x) = 5 are constant solutions. We can see these in the graph below.

6
y(x) = 5
5

3
y(x) = 2
2

1 2 3 4 5

We have included the slope field in the graph. Notice that the arrows are always
pointing downward between y = 2 and y = 5, while for y < 2 or y > 5 the arrows are
always pointing upwards. This is because the sign chart of dx = (y − 2)(y − 5) is the
dy

following:

C. Sprouse —Math 280 Notes Page 52 DRAFT August 26, 2022

1.4. Graphical Methods 53

F(y) = (y − 2)(y − 5)

+ − +

2 5

So solutions are decreasing for 2 < y < 5 and increasing for y < 2 and y > 5. We
can now include these solutions, and see that they are tangent to the slope field, as they
should be.
6
y(x) = 5
5

3
y(x) = 2
2

1 2 3 4 5

The same process works to sketch the solutions to any first order autonomous ODE:
✬ ✩
dy
To sketch the solutions to the autonomous ODE dx = F(y):
1. First solve F(y) = 0 for the critical points, yc .

2. Make a sign chart, which indicates sign of F(y) between the crit-
ical points yc . The sign may or may not change at each critical
point.

3. Sketch the constant solutions y = yc . Outside of the constant solu-

tions, use the sign chart to determine where the solutions must be
increasing or decreasing. Use this to fill in the rest of the solution.
✫ ✪
dy 2 3
Example 1.4.5. Sketch the solutions to dx
= 3y − y .
solution:
3y − y = y (3 − y),
2 3 2

So the critical points are y = 0,3. The sign chart is

C. Sprouse —Math 280 Notes Page 53 DRAFT August 26, 2022

1.4. Graphical Methods 54

F(y) = y (3 − y)
2

+ + −

0 3

Notice that the sign did not change at y = 0, because the factor of F(y) with that zero is to
2
an even power (=y ). The sketch of the general solution is then:

5
Gen. Sol.: y = y (1 − y)
′ 2
4
3 y =3

2
1
y =0
−1 1
−1
−2 □.

dy
Example 1.4.6. Sketch the solutions to dx
= cos(y).

solution: Don’t get distracted by the cosine here. The solutions have nothing to do with
the function y = cos(x). That is, they don’t look like the graph of cos(x) at all. Instead, it
is cos(y) that determines the critical points and sign chart:

3π π π 3π
cos(y) = 0 ⇒ y = ...,− ,− , , , ...
2 2 2 2

And the sign chart is

F(y) = cos(y)

+ − + − +

− 3π − π2 π
2
3π
2 2

So the general solution looks like:

C. Sprouse —Math 280 Notes Page 54 DRAFT August 26, 2022

1.4. Graphical Methods 55

′
Gen. Sol.: y = cos(y)
y = 3π
2

y = π2

−10 −5 5 10
y = − π2

y = − 3π
2

□.

Now we will change the question somewhat. Instead of looking for the general so-
lution, we will use the same technique to sketch the solution of an initial value problem
(IVP). We can still sketch the general solution for guidance, but the crucial distinction
is that there can only be one solution to the IVP. So the answer must be a textitsingle
function, not the many functions represented by the general solution.

Example 1.4.7. Sketch the solution to the IVP

dy 2
= 2−y −y ,
dx

y(0) = 3.

solution: F(y) factors as 2 − y − y 2 = (2 + y)(1 − y), so the critical points are y = −2, 1.
The sign chart is.

F(y) = (2 + y)(1 − y)

− + −

-2 1

Again, we can sketch the general solution to get a good idea of where the solution to the
IVP is, but the general solution is not the answer, as it is not being asked for. The answer
is a single function, which is the solution that satisfies y(0) = 3. So we have a single curve
that goes through the point (0, 3). This is the picture on the right below.

C. Sprouse —Math 280 Notes Page 55 DRAFT August 26, 2022

1.4. Graphical Methods 56

4 4
General Solution Solution to IVP
3 ⋅ 3 ⋅

2 2

1 1

−1 1 2 −1 1 2
−1 −1

−2 −2

−3 −3
□.
Example 1.4.8. Sketch the solution to the IVP
dy 2
= y − 6y + 9,
dx
y(1) = 0.
solution: F(y) factors as y − 6y + 9 = (y − 3) , so the sign chart quite simple:
2 2

F(y) = (y − 3)
2

+ +

So we have the solution to the IVP as the answer on the right:

5 5
General Solution Solution to IVP

4 4

3 3

2 2

1 1

⋅ ⋅
−1 1 2 3 4 −1 1 2 3 4
−1 −1

Example 1.4.9. Sketch the solution to the IVP

= (4 − y) ln(y),
dy
dx
y(3) = 2.

C. Sprouse —Math 280 Notes Page 56 DRAFT August 26, 2022

1.4. Graphical Methods 57

solution: Note that the function (4 − y) ln(y) is only defined for y > 0, because ln(y) is
undefined for y ≤ 0. The critical points are

(4 − y) ln(y) = 0 ⇒ 4 −y = 0 or ln(y) = 0

⇒ y = 4, 1.
Recall that ln(y) is negative for 0 < y < 1 and positive for y > 1. So the sign chart is

F(y) = (4 − y) ln(y)

− + −

0 1 4

This gives the solution on the right:

6 6
General Solution Solution to IVP
5 5

4 4

3 3

2 ⋅ 2 ⋅

1 1

1 2 3 4 5 6 1 2 3 4 5 6

C. Sprouse —Math 280 Notes Page 57 DRAFT August 26, 2022

 1.4. Graphical Methods 58

Homework 1.4:
A) Sketch the solutions to the initial value problems. Use the direction fields from the pictures below.
dy
1. = x − xy
dx
a) y(0) = 2 b) y(0) = 1 c) y(0) = −1

= 12 ye−x
dy
2.
dx

a) y(0) = 1 b) y(0) = 0 c) y(0) = −1

dy
3. = y −x
dx
a) y(0) = 2 b) y(0) = 1 c) y(0) = 0

dy 2 2
4. =y −x
dx
a) y(0) = 1 b) y(0) = 0 c) y(0) = −1
′ ′ −x
Figure 1.1: y = x − xy Figure 1.2: y = 12 ye

′ ′ 2 2
Figure 1.3: y = y − x Figure 1.4: y = y − x

C. Sprouse —Math 280 Notes Page 58 DRAFT August 26, 2022

 1.4. Graphical Methods 59

B) Sketch the general solutions.

1.
dy (y + 1) (y − 2)
2
=
dx y4 + 3

solution: Note that since y4 + 3 > 0, it does not change the sign. And the critical points only
depend on the numerator: (y + 1) (y − 2) = 0 ⇒ yc = −1, 2. The sign chart and
2

corresponding solution are below.

− − +

-1 2

y=2

0
y = −1

□.

2. 10.
dy dy 3 2
= y(y − 2) = y − 6y + 9y
dx dx
3. 11.
dy 2 dy 3
= 3y − y = y − 9y
dx dx
4. 12.
dy 2
dy 4
= y −y −2 = 16 − y
dx dx
5. 13.
dy 2
dy y−3
= y + 7y + 10 = 2
dx dx y + 1
6. 14.
= (1 − y)e
dy 2 dy
= 15 − 8y − y y
dx dx
7. 15.
dy 2 dy y
= 1 − 2y + y = 9− 3
dx dx
8. 16.
dy 2 4 dy
=y −y = arctan(y)
dx dx
9. 17.
dy 3 2 dy
= y − 4y = sin(y)
dx dx

C. Sprouse —Math 280 Notes Page 59 DRAFT August 26, 2022

 1.4. Graphical Methods 60

C) Sketch the solutions to the initial value problems. (Your answer should be a single function. You can use
the general solution to find the answer, but the general solution should not be a part of your answer!)
18.
′ (y + 1)(y − 2)
y = ,
y2 + 2
y(3) = 0.
solution: We know what the general solution looks like from the sign chart on the previous
page. To sketch the solution to the initial value problem,we simply choose the solution with
the correct initial values.
4
3
2 y =2

1
0 ⋅
1 2 3 4 5 6
−1 y = −1

−2
−3 □.

19. 25.
′ 2 ′ 3
y = y − 3y y = y − 25y
y(0) = 2 y(0) = 3
20. 26.
′ 2 ′ 2 3
y = y − 3y y = y − 2y
y(0) = 4 y(0) = 3
21. 27.
′ 2 ′ 3 2
y = y − 3y y = 10y − 2y
y(0) = −2 y(0) = 2
22. 28.
′ 2 ′ 2
y =6+y−y y = y − 10y + 25
y(2) = 1 y(3) = 0
23. 29.
(y > 0)
′ 2 ′
y =6+y−y y = ln(y)
y(2) = 5 y(0) = 3
24. 30.
′ 2 ′
y = y − 7y + 10 y = cos(y)
y(−1) = 2 y(0) = π

C. Sprouse —Math 280 Notes Page 60 DRAFT August 26, 2022

1.5. Existence and Uniqueness ∗ 61

1.5 Existence and Uniqueness ∗

In the previous section we saw how to graph the solutions to autonomous ODE’s which
may be difficult or impossible to solve explicitly by hand. More generally, we would like
to know if an ODE is solvable, even if it is not possible to write down a precise formula for
the solution. This is particularly important for initial value problems, where we usually
want to get a unique solution. Fortunately it is easy to characterize when this happens.
The following is the fundamental existence-uniqueness theorem for ODE’s, also called
the Picard-Lindelhöf Theorem. (Or rather, a simple consequence of it.)
Theorem 1.5.1 (Fundamental Existence-Uniqueness Theorem). Consider the first order
IVP in standard form,
dy
= F(x, y),
y(x0 ) = y0
dx ⋆

If F(x, y) and ∂F
(x, y) are both continuous in a neighborhood of (x0 , y0), then there exists an
interval (a,b), with a < x0 < b, such that there exists a unique solution to ⋆ on (a, b).
∂y

Example 1.5.2. Show that there exists a unique solution to

dy
= ln(y − x),
dx
y(1) = 2
on some interval (a, b).
solution: Since 2 > 1 we have that y − x > 0 on a neighborhood of (2, 1). So ln(y − x)
exists and is continuous in this neighborhood. Furthermore,
∂F 1
=
∂y y − x
also exists and is continuous as long as y > x. Hence the theorem applies and a solution
y(x) exists on some interval (a, b).

Notice that we did not solve the equation above. We have no method of solving this
equation. We have only used the theorem to show that a solution exists. Once we know
the unique solution exists, we can use a computer to graph it.

y
2.1
y(x)

2 ⋅

1.9

1.8
x
0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4

C. Sprouse —Math 280 Notes Page 61 DRAFT August 26, 2022

1.5. Existence and Uniqueness ∗ 62

Example 1.5.3. Show that there exists a unique solution to

(1 − x )
2 dy
+ cot(y) = 2x,
dx
y(0) = 1
on some interval (a, b).
dy
solution: We need to write this in standard form first. Solving for dx
gives

dy 2x − cot(y)
= .
dx 1 − x2
cos(y)
Note that cot(y) = sin(y)
is discontinuous at y = nπ. So the function

2x − cot(y)
F(x, y) =
1 − x2
is discontinuous at y = nπ and x = ±1. Since the initial values are y = 1 and x = 0, we
avoid points at which F(x, y) is discontinuous. And similarly

∂F csc (y)
2
1
sin2 (y)(1 − x 2 )
= =
∂y 1 − x2

is also continuous around (0, 1). Hence the theorem applies, and a unique solution exists
on some (a, b).

Like the previous example, the above equation is nonlinear, nonseparable, and we
have no technique to solve it analytically. Here is a graph of the solution by computer.

2 y
y(x)
1.8
1.6
1.4
1.2
1 ⋅
0.8
0.6
0.4 x
−0.9 −0.6 −0.3 0.3 0.6 0.9

It is an important part of the theorem that the solution only exists on some interval
(a,b) containing x0 . Even for relatively simple equations, the solution y(x) may not exist
for all x.
Example 1.5.4.
dy 2
=y ,
dx
y(2) = 1.

C. Sprouse —Math 280 Notes Page 62 DRAFT August 26, 2022

1.5. Existence and Uniqueness ∗ 63

solution: This is separable, so

−2 dy −2 dy
y =1 → y dx = 1 dx.
dx dx
Or,
−2 −1
y dy = x + C ⇒ −y = x + C.
−1
Setting y(2) = 1 gives −1 = 2 + C, so
1
− y = x − 3,

and
1
. y=
3−x
We can see here that y → ∞ as x → 3, so the solution y only exists on (−∞, 3) □.

5
1
4 3−x

1 ⋅

−1 1 2 3 4
The above is a common way that solutions can fail to exist for on (−∞, ∞). But it is
not the only way. Sometimes the issue is that the derivative of y approaches ∞ at some
point x0 , rather than y itself.
Example 1.5.5.
2 dy
3y + 2x = 0,
dx
y(1) = 2.
solution: Again this is separable.

2 dy 2
3y dx = −2x dx ⇒ 3y dy = − 2x dx.
dx
Or,
3 2
y = −x + C.
Setting y(1) = 2 gives 8 = −1 + C, so C = 9 and
1
2 3
y = 9−x .
Now note that the derivative of y is

dy 1 −2x
= .
dx 3 (9 − x 2 ) 23
dy
So the derivative does not exist at x = ±3. Therefore this solution can only exist on
(−3, 3) □.
dx

C. Sprouse —Math 280 Notes Page 63 DRAFT August 26, 2022

1.5. Existence and Uniqueness ∗ 64

3 1
9 − x2 3

2 ⋅

−3 −2 −1 1 2 3

The above examples both have in common the fact that they are nonlinear. It turns
out that for linear ODE’s, the solutions do exist for all x, provided that the coefficients are
well behaved.
Definition 1.5.6. A first order linear ODE of the form

a1 (x) + a0 (x)y = f (x)

dy
dx
is said to be nonsingular on (a,b) if a0 (x), a1(x), f (x) are continuous on (a, b), and a1 (x) is
never zero on (a,b).

Theorem 1.5.7. Consider the linear initial value problem

a1 (x) dx + a0(x)y = f (x),

dy

y(x0) = y0,
(1.2)

Suppose the ODE is nonsingular on (a, b). Then for any x0 ∈ (a, b) there exists a
unique solution to 1.2 on (a, b). on (a,b).

The above theorem is true for higher order linear ODE’s as well, and will be funda-
mental in Chapter 3 (Theorem 2.1.20). In the first order case, the proof is straightforward.
We write the ODE in standard form as

dy
+ p(x)y = q(x),
dx
where p = a0 (x)/a1 (x) and q = f (x)/a1(x). Since a1 (x) is never zero on (a,b) we have that
both p(x) and q(x) are continuous on (a,b). Then we have seen already in Section 3 that
the above equation can be solved in the form
x
1 C
y(x) = µ(t)q(t)dt + ,
µ(x) x0 µ(x)

where µ = e∫ p(x) dx . Since p and q exist and are continuous on (a,b), the solution here also
exists on (a, b). (Dividing by µ(x) is not a problem, as exponentials are never zero!)
Example 1.5.8.
(x + 1)
2 dy
− 2xy = 1,
dx
y(0) = 1.

C. Sprouse —Math 280 Notes Page 64 DRAFT August 26, 2022

1.5. Existence and Uniqueness ∗ 65

solution: The functions x + 1, −2x, and 1 are all continuous on (−∞,∞). The leading
2

coefficient x + 1 is never zero, so the ODE is nonsingular on (−∞, ∞). If we solve by

2

√
integrating factors, we get that
y = x + x 2 + 1.

This exists and is differentiable on all of (−∞, ∞), in agreement with the theorem above
□.

√ 8
x + x2 + 1 7
6
5
4
3
2
1 ⋅
−3 −2 −1 1 2 3

Example 1.5.9.
(x − 2)
dy
+ y = 1,
dx
y(3) = 0.

solution: The leading coefficient x − 2 is zero at x = 2, so we can only say the ODE is
nonsingular on (2,∞). (It is also nonsingular on (−∞, 2), but we need the initial value
x0 = 3 to be contained in the interval we choose.) Solving by integrating factors gives

x−3
y= .
x−2
Note that y exists on all of (2, ∞) □.

1 x−3
x−2

⋅
1 2 3 4 5
−1

−2

It is now easy to demonstrate examples where existence and uniqueness can fail.

Example 1.5.10. There is no solution to the initial value problem

(x − 1)
dy
= 3y,
dx

y(1) = 3.

C. Sprouse —Math 280 Notes Page 65 DRAFT August 26, 2022

1.5. Existence and Uniqueness ∗ 66

solution: Note that x0 = 1 does not satisfy the conditions of the existence-uniqueness
3y
theorem, as F(x,y) = x−1 is not continuous there. We can also solve the IVP above by
separating variables:

1 dy 2
y dx = x − 1 .
After integrating, we get

2 ln ∣x−1∣+C ln((x−1) )+C

ln ∣y∣ = 2 ln∣x − 1∣ + C = ∣y∣ = e
2
2
⇒ =e = C(x − 1) .
2 2
So y = ±C(x − 1) = C(x − 1) . Setting y(1) = 3 gives

3 = C(0).

But 3 = 0 is impossible, so there can be no solution to this IVP.

Example 1.5.11. There are infinitely many solutions to

(x − 1)
dy
= 2y,
dx
y(1) = 0.
2
solution: In the previous example we got solutions of the form y = C(x − 1) . With the
initial values y(1) = 0 this gives
2
C(1 − 1) = 0 ⇒ 0C = 0.

Note that this does not mean that C = 0. It means that C can be anything! So y = (x − 1)2 ,
2 2
y = −2(x − 1) , y = 3(x − 1) are all solutions to the IVP.

3
2
1

1 2
−1
−2
−3

In fact, there are even more solutions than we indicated above. When solving, we used
1
x−1
and ln(x−1), which are not defined at x = 1. So our solutions were actually only valid
for x > 1, or for x < 1. And, we can combine the solutions in the following way:

C1 (x − 1) x ≥ 1
2

C2 (x − 1)2 x < 1
y=

Surprisingly enough, this function is not only continuous, but also differentiable at
x = 1. And therefore it is also a solution to the IVP! So for instance, we have more solutions
such as this one, with C1 = 5, C2 = −2.

C. Sprouse —Math 280 Notes Page 66 DRAFT August 26, 2022

1.5. Existence and Uniqueness ∗ 67

5
4
3
2
1

−1 1 2
−2
−3

∂F
Finally we show that the continuity of ∂y
is also necessary in the theorem. In particu-
lar, it is necessary for uniqueness.
Example 1.5.12.
∂y 2
= y3,
∂x
y(1) = 0.
2 1
solution: Here, F(x, y) = y 3 is continuous everywhere, but ∂F
∂y
= 23 x − 3 is not.
3 3
2
− 13
F =y 3 ∂F
∂y
= 23 y
2 2

1 1
y y
−2 −1 1 2 −2 −1 1 2
−1 −1

−2 −2

−3 −3
Solving by separation of variables,
− 23 dy
y = 1.
dx
Integrating gives
1
3y 3 = x + C.
The initial condition gives C = −1, and
3
(x − 1)
1
y= .
3
So it looks like this would be the only solution. But there are actually more. You can
check yourself that y(x) ≡ 0 is a solution to the ODE, and also satisfies the initial values!
−2
We missed this solution because when solving by separation of variables, we took y 3
which is not defined for y = 0. In fact, we missed many more solutions. From above we
have that y = 13 (x − C) solves the ODE for any value of C. From this we can get that for
instance
(x − 2) x > 2
y=
0 x<0
also solves the IVP. This function is in fact differentiable for all x including x = 2.
So here are three solutions that we have found:

C. Sprouse —Math 280 Notes Page 67 DRAFT August 26, 2022

1.5. Existence and Uniqueness ∗ 68

10

−7 −6 −5 −4 −3 −2 −1 1 2 3 4 5 6 7 8
−2

−4

−6

−8

−10

To see what all solutions might be, take any two numbers x1 < 1 and x2 > 1. Then any
function of the following form is also a solution:

⎧
⎪ (x − x1 )
⎪
⎪
1 3
x < x1
⎪
y=⎨
3
⎪
⎪
0 x1 ≤ x ≤ x2 □ .
⎪
⎪
⎩ 3 (x − x2 )
1 3
x > x2

10
Sol’s to
′ 2 8
y =y3,
y(1) = 0. 6

−7 −6 −5 −4 −3 −2 −1 1 2 3 4 5 6 7 8
−2

−4

−6

−8

−10

We will not be able to show any examples of nonexistence in the case where F(x, y)
is continuous, but ∂F
∂y
is not. In this case there will always be at least one solution to any
initial value problem, due to the following:

Theorem 1.5.13 (Peano Existence Theorem). Suppose F(x, y) is continuous. Then for the

C. Sprouse —Math 280 Notes Page 68 DRAFT August 26, 2022

1.5. Existence and Uniqueness ∗ 69

initial value problem

(IV P),
dy
= F(x, y)
y(x0 ) = y0
dx

there is always at least one solution y ∶ (a, b) → R, for some a < x0 < b.
We cannot call this a fundamental theorem however, because in applications of ODE’s
the uniqueness of solutions is nearly always just as important as existence. So for us this
theorem is really more of an interesting curiosity. Another interesting fact is that when
the solutions to an IVP are not unique, there are never just two or three. There will always
be infinitely many, as we have seen so far in Examples 1.5.11, 1.5.12, and also 1.3.8. An
explanation of this is contained in [9], which gives an advanced topological approach to
this subject.

1.5.1 The Noncrossing Principle

In Section 1.4, we spent some time sketching the general solutions to ODE’s. Often when
making such a sketch it is useful to use that two different solutions cannot cross each
other. This is the case as long as one has an ODE for which solutions to an initial value
problem are always unique. We state this here in its most basic form.
∂F
Fact 1.5.14 (Noncrossing Principle). Suppose F(x, y) and ∂y
are continuous for a < x < b
and all y. If y(x) and z(x) are both solutions to the ODE
dy
= F(x, y)
dx
on (a, b), then the graphs of y and z can never cross each other. That is, there can be no point
x0 ∈ (a,b) such that y(x0 ) = z(x0 ) unless y(x) ≡ z(x) on (a, b).
∂F
Proof. This follows from Theorem ?? from the next section. F(x, y) and ∂y
are continuous
on all of R . So if x0 ∈ (a, b) is any point where y(x0 ) = z(x0 ), then y and z are both
2

solutions to the same initial value problem. Therefore y(x) ≡ z(x) on (a, b).

Example 1.5.15. Suppose y solves the initial value problem

dy 2 2
= x − y + 1, y(0) = 1
dx
on an interval (a, b). Show that y(x) ≥ x for all x ∈ (a, b).

solution: Note that this ODE is not linear or separable, and can’t be solved with any
techniques that we have seen. However, we can check that z(x) = x is also a solution on
any (a, b) by plugging in:

dz d
= x = 1,
dx dx
2 2 2 2
x − z + 1 = x − x + 1 = 1. ✓

C. Sprouse —Math 280 Notes Page 69 DRAFT August 26, 2022

1.5. Existence and Uniqueness ∗ 70

y(x)

⋅
z(x)

( )
a x0 b
Figure 1.5: The Noncrossing Principle: This cannot happen!

2 2 ∂F
We have that F(x, y) = x − y + 1 and = −2y are both continuous, and therefore the
noncrossing principle applies. So if y(x) is any other solution on (a, b), y can never cross
∂y

y1 . But y(0) = 1 > 0 = z(0). Since y(0) > z(0), this means we must have y(x) > x on (a, b)
by continuity. (More precisely, this follows from the Intermediate Value Theorem of
calculus.) □.

2 y
y(x)
z(x)

1 ⋅

x
−2 −1 1 2

−1

−2

Figure 1.6: Example 1.5.15: y(x) ≥ x.

C. Sprouse —Math 280 Notes Page 70 DRAFT August 26, 2022

 1.5. Existence and Uniqueness ∗ 71

Homework 1.5:
A) For each initial value problem, prove that there exists a unique solution on some interval (a, b).

3. (y − x )
dy 2 2 2 dy
1. = x + ln(y), y(2) = 1. = x, y(0) = 1.
dx dx
4. (x − 1)
dy 3 dy
2. x + y = 1, y(1) = 2. + 1 − y = cos(x), y(2) = 0.
dx dx
B) For each initial value problem, find all (x0 ,y0) such that the initial value problem with y(x0 ) = y0 does
not necessarily have aunique solution on some interval (a, b).
dy √
5. (x − 1) y(x0 ) = y0 + y = x, y(x0 ) = y0 .
2 dy 2
+ y = 1, 9.
dx dx
= x + ln(1 + cos(y)), y(x0 ) = y0 10. (3 − x) = x 3 + y 3 , y(x0 ) = y0 .
dy dy 2 2
6.
dx dx
+ (y + 1) = ln(x), y(x0) = y0 = 25 − y 3 , y(x0 ) = y0 .
1
dy −2 dy 2
7. 11.
√
dx dx
+ xy = 1 − x2 , y(x0 ) = y0 . + ∣y − 2∣ = 3, y(x0 ) = y0
dy dy
8. y 12. sin(x)
dx dx
C) Solve the initial value problem. Then find the maximum interval (a, b) on which a solution exists.
dy 2 dy 3
13. = y , y(3) = −2. 15. = 2xy , y(0) = 5.
dx dx
dy 2 dy x
14. − y = 4, y(π) = 2. 16. + = 0, y(3) = 4.
dx dx y
dy
D) 17. x + 2y = 6.
dx
a) Find the general solution for x > 0.
b) Find a value of y0 such that the IVP with y(0) = y0 has a unique solution.
c) What about the other values of y0 ?
(x − 5)
dy
18. − 3y = 3.
dx
a) Find the general solution for x > 5.
b) Find a choice of initial values with no solutions.
c) Find a choice of initial values with infinitely many solutions.

(1 − x)
dy
19. + y = 2.
dx
a) Find the general solution for x > 1.
b) Find a choice of initial values with no solutions.
c) Find a choice of initial values with infinitely many solutions.
dy
20. sin(x) − cos(x)y = 1.
dx
a) Find the general solution. (This will only really be the general solution for intervals on which
sin(x) ≠ 0.)
b) Find all initial values with infinitely many solutions. (Be careful, y0 is not always the same!)
a) Find at least three solutions to the initial value problem dx = (y − 2) 3 , y(1) = 2.
dy 1
21.
b) Explain how this can happen, in terms of the existence/uniqueness theorem.
dy
22. a) Explain why the ODE x dx = 3y does not satisfy the hypothesis of the Noncrossing Principle.
b) Find at least two solutions that cross. Draw a picture. (You can do this by trying to solve the
ODE and lookingfor the initial value that would be interesting.)

C. Sprouse —Math 280 Notes Page 71 DRAFT August 26, 2022

 1.5. Existence and Uniqueness ∗ 72

E) For each initial value problem, use the noncrossing principle to prove the inequality over the interval on
which y is a solution.
2
dy 2y − 6y
23. = 2 , y(2) = 5. y(x) > 3.
dx x + y2
2
dy 2y − 6y
24. = , y(5) = 1. 0 < y(x) < 3.
dx x2 + y 2
dy 2 2
25. = 2xy − x − y , y(2) = 0. y(x) > 1 − x.
dx
2
dy x 2
26. = y + 2x − 1, y(1) = 0. y(x) < x .
dx
dy 2 1
27. + xy + y = 1 , y(1) = 2. y(x) > x .
dx
= (x + y) sin(y),
dy
28. y(1) = 5. π < y(x) < 2π.
dx

C. Sprouse —Math 280 Notes Page 72 DRAFT August 26, 2022

1.6. Substitutions 73

1.6 Substitutions
Here we discuss a few standard types of equations which are solvable by a change of
variables. The techniques here might seem more specialized than what we have done
previously. But it is fairly common to have to reduce a given ODE to a more tractable one,
using a substitution that may or may not be obvious.
As a warmup, consider an equation of the form

= f (x + y),
dy
dx
such as dx = (x + y)3 or dx = cos(x +y). This equation is definitely not linear or separable,
dy dy

so the techniques we have learned so far cannot apply directly to this. But since the x + y
sticks out, we can try the substitution

u = x + y,

and see where it goes. Taking derivatives,

du dy dy du
= +1 ⇒ = − 1.
dx dx dx dx
= f (x + y) gives
dy
Then substituting this back into dx

− 1 = f (u) = f (u) + 1.
du du
⇒
dx dx
This equation is now separable, as it can be rewritten
1 du
f (u) + 1 dx
= 1.
✬ ✩

= f (x + y) ∶
dy
To solve dx

1) Make the substitution u = x + y. Taking derivatives of this gives

du dy
dx
= 1 + dx .

2) Substitute these into the given ODE to get

= f (u) + 1.
du
dx

3) Solve the ODE for u by separation of variables. Then remember

to replace u by x + y in the final answer.
✫ ✪
Example 1.6.1. Find an implicit solution to
dy
= sin(x + y),
dx
y(−1) = 1.

C. Sprouse —Math 280 Notes Page 73 DRAFT August 26, 2022

1.6. Substitutions 74

du dy
solution: Using u = x + y and dx
= 1 + d x , we have
dy du
= sin(u) ⇒ − 1 = sin(u).
dx dx
So by separation of variables,
du 1 du
= 1 + sin(u) ⇒ =1
dx 1 + sin(u) dx
1 du
⇒ dx = 1dx
1 + sin(u) dx
1
⇒ du = x + C.
1 + sin(u)

The integral on the left can be solved after some algebraic manipulation:

1 1 1 − sin(u)
du = du
1 + sin(u) 1 + sin(u) 1 − sin(u)
1 − sin(u)
=
1 − sin2 (u)
du

1 − sin(u)
cos2 (u)
= du

sec (u) − sec(u) tan(u) du = tan(u) − sec(u) + C.

2
=

So
tan(u) + sec(u) = x + C ⇒ tan(x + y) − sec(x + y) = x + C.
Plugging in y(−1) = 1 gives tan(0) − sec(0) = 0 + C. So C = 0 and the implicit solution is
tan(x + y) − sec(x + y) = x.

3
tan(x + y) − sec(x + y) = x
2

−1 1 2 3 4 5
−1 □.

One may notice, given the example above, that there are not many ODE’s like this that
can explicitly be solved by hand... the integrals usually end up being just too hard! It is
the trick itself that is interesting though, because we can try it in other, similar situations.
For instance, if we had
= f (xy),
dy
dx
we could try setting u = xy and use
du dy
u = xy ⇒ =y +x .
dx dx

C. Sprouse —Math 280 Notes Page 74 DRAFT August 26, 2022

1.6. Substitutions 75

The trick is that for simple substitutions like this, we will always get a separable equation,
which can be solved by integrating. One specific case of this occurs frequently enough
dy y
that it has a special name: an equation of the form dx = f x is called ‘homogeneous’.
We will now treat these in more detail.

1.6.1 Homogeneous First Order ODE’s

Definition 1.6.2. A homogeneous first order ODE is an equation of the form
dy y
=f x .
dx
We have emphasized that we are talking about first order ODE’s here, because in sub-
sequent chapters the adjective homogeneous will refer to something completely different.
(‘Homogeneous’, like elliptic, hyperbolic, and harmonic, is a particularly overused word
in math, so can mean many things.)
So to solve these, as indicated above we set

u = y/x.
du
Then we need to find dx
.

y dy du
u= x ⇒ y = xu ⇒ =u+x .
dx dx
dy y
So then the original ODE dx
=f x
becomes

= f (u).
du
u +x
dx
And✬
this is always separable! ✩
dy y
To solve dx
=f x
∶
y
1) Make the substitution u = x . Taking derivatives of y = xu gives
dy
dx
= u + x du
dx
.

2) Substitute these into the given ODE to get

= f (u) − u.
du
x
dx

3) Solve the ODE for u by separation of variables. Then replace u by

y
in the final answer.
✫ ✪
x

Example 1.6.3. Solve the initial value problem

dy y 2
=2− x ,
dx
y(1) = 0.

C. Sprouse —Math 280 Notes Page 75 DRAFT August 26, 2022

1.6. Substitutions 76

solution: Letting u = y/x one solves for y as y = xu and takes derivatives to get
dy
dx
= u + x dd ux . Then we have
du 2 du 2
u +x =2−u ⇒ x = 2− u −u .
dx dx
So
1 du 1 1 du 1
= ⇒ dx = x dx
2 − u − u dx x
2 2
2 − u − u dx
1
du = ln∣x∣ + C.
2 − u − u2
Recall that the left-hand side can be integrated by partial fractions (we omit the details):
1 1
(2 + u)(1 − u)
2
du = du
2−u−u
1 1
3 3
= + du
2+ u 1−u
1 1
= ln∣2 + u∣ − ln∣1 − u∣ + C.
3 3
So
ln ∣2 + u∣ + ln ∣1 − u∣ = ln ∣x∣ + C,
1 1
3 3
or
ln 2 + x − ln 1 − x = ln ∣x∣ + C.
1 y 1 y
3 3
At x = 1, y = 0 the inside of each logarithm above is positive, so we can drop the absolute
values. Furthermore at these values we have C = 13 ln(2), so
1 y 1 y 1
ln 2 + x − ln 1 − x = ln(x) + ln(2).
3 3 3
This can be simplified, and we can in fact get an explicit solution here. By the rules of
logs,
y
2+ x
y = 3ln(x) + ln(2) = ln(2x ).
3
ln
1− x
So taking e to both sides,
y
2+ x 3 2x + y 3
y = 2x ⇒ x − y = 2x .
1− x
Then finally a little more algebra gives an explicit solution:
4
2x − 2x
y= □.
2x 3 + 1

−0.5 0.5 1 1.5 2

−1 □.

C. Sprouse —Math 280 Notes Page 76 DRAFT August 26, 2022

1.6. Substitutions 77

Example 1.6.4. Find an explicit solution to

dy y x
= + ,
dx x y
y(1) = 2.
solution: The right hand side is
y y x y y −1
f x = x +y =x + x .
y
So substituting u = x
gives
dy −1
=u +u .
dx
dy
Since y = xu gives dx
= u + x du
dx
again,

du −1 du −1
u +x =u+u ⇒ x =u .
dx dx
Separating,
du 1 du 1
u = ⇒ u dx = x dx
dx x dx
⇒ u du = ln ∣x∣ + C

u = ln ∣x∣ + C.
1 2
⇒
2

So
= ln ∣x∣ + C.
1 y 2
2 x
With the initial values y(1) = 2, we get that C = 2, ln ∣x∣ = ln(x), and that we can use the
positive square root when finding the explicit solution:

y = x 2 ln(x) + 4

5 √
x 2 ln(x) + 4
4

0.5 1 1.5 2 □.

Many of the more interesting homogeneous ODE’s are often in disguise, unless you
know what to look for. For instance
3 2 2 3
dy x y − x y
= 4
dx x y + xy 4

C. Sprouse —Math 280 Notes Page 77 DRAFT August 26, 2022

1.6. Substitutions 78

5
is homogeneous. This can be seen by factoring an x from both the top and bottom of the
fraction.
x (x y − x y )
3 2 2 3 5 −2 2 −3 3 y 2 y 3
x y −x y x
− x
= 5 −1 =
x (x y + x−4y 4 )
.
x4 y + xy 4 y
+
y 4
x x
Fortunately these can all be dealt with as above. They are of the form

dy p(x,y)
= ,
dx q(x,y)

where p(x, y) and q(x, y) are both homogeneous polynomials of degree n. This means
p q
that all terms in p(x, y) and q(x, y) are constant multiples of x y , where p + q = n. And
these can always be written as a homogeneous equation by factoring out x n from both the
top and bottom of the fraction.

Example 1.6.5.
dy y2
= 2 ,
dx x + xy

y(1) = 1.

solution: The top and bottom of the fraction are both homogeneous polynomials of
degree 2. So
2 2 −2 2 y 2
y x x y x
= 2 = .
x + xy x 2 + x −1 y
2
2+
y
x
y
So setting u = x
gives
dy u2 du u2
= ⇒ u +x = ,
dx 1 + u dx 1 + u
2 2 2
du u u u+u u
x = −u = − =− .
dx 1 + u 1 + u 1 + u 1 + u
So
du u 1 + u du 1
x =− ⇒ u dx = − x
dx 1 +u
1 + u du 1
⇒ u dx dx = − x dx
1 1
⇒ u + 1du = − x dx
⇒ ln∣u∣ + u = − ln∣x∣ + C.

The initial values y(1) = e give C = 2, and that ln ∣x∣ = ln(x), ln∣u∣ = ln(u). So
y y
ln(u) + u = − ln(x) + 1 ⇒ ln x + x = − ln(x) + 1.

This gives the implicit solution.

C. Sprouse —Math 280 Notes Page 78 DRAFT August 26, 2022

1.6. Substitutions 79

If we allow the Lambert-W function, then we can also get an explicit solution. Taking e to
both sides of the above
u − ln(x)+1 e
ln(u) + u = − ln(x) + 1 ⇒ ue = e = x.

So
e y e e
u =W x ⇒ x =W x ⇒ y = xW x .

2 y y
ln x + x = − ln(x) + 1

1 2 3 4 5 □.

Example 1.6.6.
dy 2x + y
= x−y ,
dx
y(1) = 0.

solution: The right-hand side is

y y
x + 2y x 1 + 2 x 1+ 2x
=x = y .
2x − y 2− x
y
2− x

y
So setting u = x
we have

dy 2 + u du 2 + u
= ⇒ u+x = .
dx 1 − u dx 1 − u
So
2
du 2 + u 2 + u − u(1 − u) u + 1
x = −u = = .
dx 1 − u 1− u 1−u
Separating variables,
1 − u du 1
= ,
u 2 + 1 dx x

du = ln ∣x∣ + C dx
1 − u du 1 1 u
dx = x dx ⇒ −
u 2 + 1 dx u2 + 1 u2 + 1
arctan(u) − ln(u + 1) = ln ∣x∣ + C.
1 2
⇒
2
2
y 1 y
⇒ arctan x − ln 2 + 1 = ln∣x∣ + C.
2 x

C. Sprouse —Math 280 Notes Page 79 DRAFT August 26, 2022

1.6. Substitutions 80

Setting y(1) = 0 gives C = 0 and

2
y 1 y
arctan x − ln 2 + 1 = ln(x).
2 x

By combining logs, this can be written somewhat nicely as

arctan x = ln( x2 + y 2 ).
y

arctan x = ln( x 2 + y 2 )
1 y

1.05 1.1 1.15 1.2 1.25 □.

1.6.2 Bernoulli equations ∗

Our next substitution is less obvious than the previous ones. The following equations are
not linear, but they are close enough that we can transform them into linear first order
ODE’s.

Definition 1.6.7. A Bernoulli equation is a first order ODE of the form

dy n
+ p(x)y = q(x)y .
dx
We make the substitution

1−n
u =y .

Then the chain rule gives

n
= (1 − n)y
du −n dy dy y du
⇒ = .
dx dx dx 1 − n dx
So
n
dy n y du n
+ p(x)y = q(x)y ⇒ + p(x)y = q(x)y .
dx 1 − n dx
Now multiplying both sides of the above by (1 − n)/y gives
n

+ (1 − n)p(x)y = (1 − n)q(x).
du 1−n
dx

C. Sprouse —Math 280 Notes Page 80 DRAFT August 26, 2022

1.6. Substitutions 81

1−n
But then we have y = u! So now we get an equation in u, that we can solve for u(x):

+ (1 − n)p(x)u = (1 − n)q(x).
du
dx

✬ ✩
And, this is a linear ODE that we can solve for u. So, again:

dy n
To solve dx
+ p(x)y = q(x)y ∶
1−n
1) Make the substitution u = y . Taking derivatives gives
(1 − n)y dx ,
n
du −n dy dy y du
dx
= or dx
= 1−n dx
.

2) Substitute these into the given ODE to get

+ (1 − n)p(x)u = (1 − n)q(x)
du
dx

3) Solve the linear ODE for u by integrating factors. Then use y =

1
u 1−n to get the final answer.
✫ ✪

Example 1.6.8. Find the general solution to

dy 3
+y =y .
dx

−2 −3 dy dy 3 du
solution: We have n = 3, so u = y and du
dx
= −2y dx
. This gives dx
= − 12 y dx
and

1 3 du 3 du −2
− y +y = y ⇒ − 2y = −2.
2 dx dx

So
du
− 2u = −2
dx

−2x 2x −2
Solving by integrating factors gives µ = e and u = Ce + 1. Then u = y gives
y = ±1/ u, so

y=√
±1
.
Ce 2x + 1

−2
(We also have the solution y = 0, which was destroyed when we multiplied by y .)

C. Sprouse —Math 280 Notes Page 81 DRAFT August 26, 2022

1.6. Substitutions 82

2
1
y = ±(Ce2x + 1)− 2

0.5 1 1.5 2

−1

−2 □.

The general solution above should look familiar, as we did many examples like this in
Section 1.2. That is, the above ODE is actually autonomous, and we could have graphed
the solution with a sign chart, and solved for it by separation of variables. Now let’s do
some Bernoulli equations that are not linear or separable, so cannot be done in an easier
way.

Example 1.6.9.
dy 1 1
+ xy = 2 ,
dx y

y(1) = 2.

1−(−2) 3 du 2 dy
solution: Let u = y = y . Then dx
= 3y dx
. So

1 du 1 1 du 3 3
+ xy = 2 ⇒ + y =3
3y 2 dx
y dx x
du 3
⇒ + u = 3.
dx x
3∫ 1
dx 3ln(x) 3
This can be solved by the integrating factor µ(x) = e x =e =x .

(x u) = 3
3 ′ 3 −2 −3
⇒ x u = 3x + C ⇒ u = 3x + Cx .
1
Since y = u 3, we have
1
−2 −3
y = 3x + Cx 3
,

and y(1) = 2 gives

1
−2 −3
y = 3x + Cx 3
.

C. Sprouse —Math 280 Notes Page 82 DRAFT August 26, 2022

1.6. Substitutions 83

5
1

4 y = 3x −2 + 5x−3 3

0.5 1 1.5 2 2.5 3 □.

Example 1.6.10.
dy 2
− tan(x)y = cos(x)y ,
dx
1
y(0) = .
2
1−2 −1 du −2 dy dy 2 du
solution: We have u = y =y . So dx
= −y dx
, and dx
= −y dx
. So

2 du 2 du −1
−y − tan(x)y = cos(x)y ⇒ + tan(x)y = cos(x)
dx dx
du
⇒ + tan(x)u = − cos(x).
dx
∫ tan(x) dx − ln(cos(x)) ln(sec(x))
The integrating factor is µ(x) = e =e =e = sec(x). So

(sec(x)u) = −sec(x)cos(x) = −1
′
⇒ sec(x)u = −x + C.

So u = (−x + C) cos(x), and

1
(−x + C)cos(x)
y= .

1
y(0) = 2
gives C = 2, and
1
(2 − x) cos(x)
y= .

5
y = (2−x)1cos(x)
4

−1.5 −1 −0.5 0.5 1 1.5 □.

C. Sprouse —Math 280 Notes Page 83 DRAFT August 26, 2022

 1.6. Substitutions 84

Homework 1.6:
A) Find an implicit solution to the homogeneous first order ODE. Write the equation in the formy ′ = F(y/x)
y
if necessary, and use the substitution u = x
to solve.
1.
(x + y )
2 2 dy
= xy, y( 2) = 1
dx
solution: First we have to write in the form y ′ = F(y/x). First write as a fraction

dy xy
= .
dx x 2 + y 2

2 2
The highest power of x in the fraction is x . So divide x from the top and bottom of the
fraction to get
y
dy x dy u
= ⇒ = .
dx y 2 dx 1 + u 2
1+ x
dy
Now we use the substitution y = xu ⇒ dx
= u + x du
dx
. So

du u du u
x +u = ⇒ x = − u.
dx 1 + u2 dx 1 + u 2
Adding fractions and then separating variables,
3 2
du u 1 + u du 1
x =− ⇒ = −x.
dx 1 + u2 u 3 dx
So
1 + u2 du 1 −3 1 1
dx = − x dx ⇒ u + u du = − x dx.
u 3 dx
This is easy to integrate:
1 −2
− u + ln(u) = − ln(x) + C.
2
Then substituting back in for y,
2
1 x y
− + ln x = − ln(x) + C.
2 y2

The initial condition y( 2) = 1 gives C = −1.If we want to make it look nice, we can use
ln(y/x) = ln(y) − ln(x) and simplify to get
2
2y (1 + ln(y)) = x □ .
1 x 2 2
− + ln(y) = −1 ⇒
2 y2

2. 3.
dy x
=3−2 y , y(1) = 3.
dx
4.
dy y x 3 dy y
= + , y(1) = 2. = , y(2) = 1
dx x y 3 dx x + y

C. Sprouse —Math 280 Notes Page 84 DRAFT August 26, 2022

 1.6. Substitutions 85

5. 7.
(2x + y )
dy 2 2 2 2 dy
xy = 2y − x , y(1) = 2 = 2xy, y(2e) = e
dx dx
6. 8.
3
dy 2 2 dy x y y
xy = x +y , y(e) = 0 = + + , y(1) = 1
dx dx y x x3

B) Find an explicit solution y(x). Write the equation in the formy ′ = F(y/x) if necessary, and use the substi-
y
tution u = x
(y = xu).

9. 14.
dy y y
= + sec x , y(1) = 0. 2 dy 2 2
dx x x = x + xy + y , y(1) = 3.
dx
10.
dy y x2 15.
= 2x − 3 2 , y(2) = 6.
dx y 2 dy 2 2
x = x − xy + y , y(1) = 2.
dx
√
11.
dy
x = y + xy, y(1) = 100. 16.
dx
12. 2 dy 2
x = y − 2xy, y(1) = 1.
dy y y dx
= x ln x , y(1) = 1.
dx
17.
13.
dy 2 2 dy x − y
xy = y −x , y(1) = 2. = , y(−1) = 3.
dx dx x + y

C. Sprouse —Math 280 Notes Page 85 DRAFT August 26, 2022

 1.6. Substitutions 86

C) Find an explicit solution y(x), by using the given substitution then solving a linear ODE.
18.
dy −2 3
x −y = y , y(1) = 2. u=y .
dx
du 2 dy dy 1 du
solution: The chain rule gives dx
= 3y dx
. So dx
= 3y2 d x
.So

1 du −2 du 3
x −y = y ⇒ x − 3y = 3,
3y 2 dx dx

du du 3 3
x − 3u = 3 ⇒ − xu = x.
dx dx
−3
This can be solved by integrating factors. µ(x) = x , so

(x
−3 ′ −4 −3 −3
u) = 3x ⇒ x u = −x +C

, and y = (−1 + Cx ) 3 . Solving y(1) = 2 gives C = 9 □.

1
−3 −3
So u = −1 + Cx

19. 28.
dy 3 1 −2
+ y = y , y(0) = . u=y .
dx 2 2 dy 2 −1
x + xy + y = 0, y(1) = 2. u =y .
20. dx
dy 2 −1
+ 2xy = xy , y(0) = 1. u = y .
dx 29.
21. dy 3 −2
dy x −2 3 − tan(x)y = y , y(0) = 1. u = y
+ y = e y , y(0) = 1. u =y . dx
dx
22. 30.
dy 2x 2 dy √ √
+ y = y , y(0) = 1. u=y . + 2y = 4 y, y(0) = 1. u= y
dx dx
23.
dy 31.
√ √
2 4 −3
x + y = x y , y(−1) = 1. u = y . dy
dx x − 2y = 2 y, y(1) = 4. u= y
24. dx
dy 2 −1
x − 3y = y , y(1) = 3. u=y . 32.
√ √
dx
dy
25. x − y = x y, y(4) = 1. u= y
+ (x − 1)y = y , y(2) = 1.
dy 2 −1 dx
x u=y .
dx 33.
26. dy 8 3
dy y x 2 +y = , y(0) = 1. u = y2.
− = , y(1) = 2. u=y . dx y
dx x y
27. 34.
2 dy −1 2 dy 3
− 12
x + xy + y = 0, y(2) = 1. u=y . x + y = y 2 , y(4) = 1. u=y .
dx dx

C. Sprouse —Math 280 Notes Page 86 DRAFT August 26, 2022

1.7. Applications: Linear First Order ODE’s 87

1.7 Applications: Linear First Order ODE’s

In this section we present the traditional word problems associated with linear first order
ODE’s.

1.7.1 Exponential Growth and Population Models

This is the first example of a nontrivial differential equation that one usually encounters.
Suppose there is a population of bacteria in a petri dish, such that every day each individ-
ual has a 15 % chance of dividing and thus reproducing. If there are 500 individuals, then
after one day you would expect .15 × 500 = 75 more bacteria. More generally, if y(t) is
the population and t is time in days, then y(t + 1) = y(t) + .15y(t). The important feature
here is that the increase in population is proportional to the population already there. As
a differential equation, this can be represented by

dy
= ky.
dt
This equation is both linear and separable. In first year calculus, this is sometimes
used for a simple demonstration of separation of variables, which goes like this:

1 dy 1 dy
y dt = k ⇒ y dt dt = k
⇒ ln ∣y∣ = kt + c.

Since y is a population, it can’t be negative (you can’t have −5 bacteria), and thus ln∣y∣ =
ln(y). So
kt+c kt c kt
ln(y) = kt + c ⇒ y =e = e e = Ce ,
where C = e . If we are given inital values y(t0) = y0, then
c

kt 0 −kt0 −kt 0 kt
y0 = Ce ⇒ C = y0 e ⇒ y = y0 e e .
Applying the exponential rules again, we have a nice expression for y:
k(t−t0 )
y = y0 e .
In specific applications, constant k might not be given directly. In the bacteria exam-
ple above, k would not be exactly .15, as t is measured in days, whereas measuring the
rate of change by a derivative requires t to be infinitesimally small. But given enough
information, we can always determine k with a little more work. For the bacteria, you
k(t−t0)
can check for yourself that plugging y(t0 + 1) = 1.15y(t0 ) into y = y0 e will give
k = ln(1.15) ≈ .14.

Example 1.7.1. The White Garden Snail (Theba Pisana) is a snail species native to southern
Europe. It readily survives a container ship journey, reproduces quickly, and can maintain
2
high population densities (as much as 500 snails/m , or 2 million snails/acre). Thus it is a
major worldwide pest, particularly on coasts with Mediterranean climates. In coastal Southern

C. Sprouse —Math 280 Notes Page 87 DRAFT August 26, 2022

1.7. Applications: Linear First Order ODE’s 88

Figure 1.7: White garden snail on dead stalk of invasive black mustard. Filiorum Reserve,
Palos Verdes Peninsula.

California it is seen in large numbers feeding on the stalks of its fellow Mediterranean invaders
from the plant world: black mustard, fennel, and wild radish.
The snail is hermaphroditic, so that each adult can lay an average of 80 eggs at a time, 5
times a year. Suppose that you take a vacation on an uninhabited island, and inadvertently
bring 3 of these snails along with your luggage. Suppose that 14% of the eggs are able to hatch
into adults, and the adults have a lifespan of one year in this environment (many eggs won’t
survive to hatch, both snails and eggs are eaten by predators, etc.) How long until there are 1
million snails?

solution: Since no initial time has been given, we can assume t0 = 0 so

kt kt
y(t) = y0 e = 3e . From the above, we assume the number of eggs that hatch into adults
from one snail is
80 ⋅ 5 ⋅ .14 = 56 snails/year.

Then after one year the original 3 snails die off and we have 3(56) = 168 snails. This
allows us to solve for k:

k(1) k 168
y(1) = 3e = 168 ⇒ e = = 56.
3
4.025t
So k = ln(56) ≈ 4.025, and y(t) = 3e .
Then we can solve y(t) = 1,000, 000 for t:

6
4.025t 6 1 10
3e = 10 ⇒ t= ln ≈ 3.16 years □ .
4.025 3

C. Sprouse —Math 280 Notes Page 88 DRAFT August 26, 2022

1.7. Applications: Linear First Order ODE’s 89

100
6
10
80
4

snails
60
snails
10
40 2
10
20
3 0
0 0 1 2 3 10
0 0.5 1 4
years years

By the way, the 2 million snails/acre figure above is not irrelevant to the problem
at hand. Exponential growth is never sustainable in the long term, as eventually the
numbers become unrealistic. At some point the rate of growth must taper off for other
physical reasons. Here, those reasons would be mostly lack of food and space. If we say
that a population of 2 million snails is sustainable on a 1-acre island, we typically expect
the growth rate to be exponential until the population reaches about half that. So our
assumption of exponential growth until there are 1, 000,000 snails is reasonable.

1.7.2 Newton’s Law of Cooling:

Suppose that T (t) is the temperature of some object at time t. This object has initial
temperature T (t0 ) = T0 when it is placed in an environment of temperature Te . Then
Newton’s law of cooling says that the temperature of the object changes proportionately
to the difference between the object’s current temperature and its environment:

= k(T − Te ),
dT
dt
The constant k depends on the physical properties of the object, such as the exposed
surface area, and the volume and thermal conductivity of its inner material. For instance
a large potato might take longer to heat than a cup of water. At any rate, the above
equation is both separable and linear. So one can solve it by either separation of variables
or integrating factors. To solve with integrating factors,

= k(T − Te )
dT dT
⇒ − kT = −kTe .
dt dt

, and the method of integrating factors gives (e T ) = −kTe e

∫ −k dt −kt −kt ′ −kt
So µ(t) = e =e .
So
−kt −kt −kt
e T= −kTe e dt = Te e + C.

−kt
Dividing by e ,
T (t) = Te + Ce .
kt

With the initial condition T (0) = T0, we get C = T0 − Te and get

T (t) = (T0 − Te )e + Te .
kt

C. Sprouse —Math 280 Notes Page 89 DRAFT August 26, 2022

1.7. Applications: Linear First Order ODE’s 90

Now, here is what the graphs of T (t) looks like. Note that this is not a surprise. Anyone
knows that if they put a cup of coffee with T0 = 190◦ in a room of Te = 75◦ , then over
several hours the temperature T (t) will decrease and eventually become the temperature
of it’s environment, 75◦ .

T0 < Te T0 > Te

T0

Te

T (t) T (t)
Te

T0

t t

Figure 1.8: Temperature T (t) of object approaching temperature of environment.

kt
Remark 1.7.2. For the graphs to look like the above, we certainly need to have lim e = 0.
k→∞
This in particular shows that k must always be negative in the Newton’s law of cooling model.
Example 1.7.3. Suppose there is a six-pack of soda in the trunk of your car. It is summer at
◦
CSUN, so the cans have a temperature of 120 F. To cool them down, you put them in a freezer
◦ ◦
of temperature 5 F. After 10 minutes you check a can, and find it to have a temperature of 90
◦
F. You want to make sure you take the soda out before it freezes, at 32 F. What is the longest
amount of time you can leave the can in the freezer?

solution: We have T0 = 120 and Te = 5. So solving

T (0) = 120
dT dT
= k(T − 5), ⇒ − kT = −5k,
dt dt
−kt
we have µ = e and

(e T ) = −5ke
−kt ′ −kt −kt −kt −kt
⇒ e T = −5k e dt = 5e + C.

−kt
The initial conditions give C = 120 − 5 = 115, and dividing by e ,
−k t
T (t) =
5e + 115 kt
= 5 + 115e .
e −kt
To find k, use T (10) = 90:

k(10) 1 85
5 + 115e = 90 ⇒ k= ln ≈ −.03.
10 115

C. Sprouse —Math 280 Notes Page 90 DRAFT August 26, 2022

1.7. Applications: Linear First Order ODE’s 91

Then the total time until the cans achieve 32 is given by setting T (t) = 32, so
◦

−.03t 1 27
5 + 115e = 32 ⇒ t=− ln ≈ 48 minutes.
.03 115

120
−.03t
T = 115e +5
100

80

60
F
◦

40
32
20
5
0 20 40 48 60 80 100
t □.

Warning 1.7.4. We have derived the answer T (t) = (T0 − Te)e +Te as a solution to Newton’s
kt

law of cooling. It is fine to memorize this, and in fact you probably should. But don’t try to
use it to get out of solving the equation. You are still required to solve the ODE as part of any
solution. Moreover, the solution here only applies when Te is constant. As the next example
shows, this doesn’t always have to be the case!
◦
Example 1.7.5. You put a cold potato of 50 F into an oven. The oven is at room temperature,
and you set it to preheat to 450◦ in 25 minutes. The temperature of the oven is then 75 + 15t ◦ .
Solve for the temperature T (t) of the potato over this time period, if it has k = −.012. How hot
did it get?

solution: We solve

= k(T − Te ) = −.012(T − (75 + 15t)).

dT dT
⇒
dt dt
So
dT .012t
+ .012T = .9 + .18t ⇒ µ(t) = e
dt
= (.9 + .18t)e
.012t ′ .012t
⇒ e T .

and

(.9 + .18t)e dt = (75 + 15t)e

.012t .012t .012t .012t
e T = − 15 e dt

(75 + 15t)e + C = (−1175 + 15t)e

.012t .−12t .012t
= − 1250e + C,

where we have integrated by parts. Dividing by the integrating factor,

T (t) = −1175 + 15t + Ce

−.012t −.012t
= −1175 + 15t + 1225e .

At the end of 25 minutes the temperature of the potato is then T (25) = 107.5 .
◦

C. Sprouse —Math 280 Notes Page 91 DRAFT August 26, 2022

1.7. Applications: Linear First Order ODE’s 92

450 T (t)
Te 400

300
F
◦

200

100
75
50

0 5 10 15 20 25
t □.

Remark 1.7.6. You might be surprised to see how low the graph of T (t) is compared to the
oven. It seems to have hardly gone up. This is because an object of small k like a potato simply
can’t keep up with the oven’s speed of heating. That is why, even if you are impatient, you might
as well wait until the oven is preheated before you put the food in. The extra heat that the potato
gains during preheating isn’t enough to make much difference in overall cooking time.

1.7.3 Mixing
Suppose that you have a tank of water, of volume V0. Wa-
ter containing c lb/gal of salt enters the tank at a rate of R
R gal/min
gal/min, and also drains at a rate of R gal/min, so the total
volume of water in the tank remains constant at V (t) = V0. c lb/gal
Our goal is to measure how much salt is in the tank as a
function of time. Sometimes we might actually be interested
in the concentration of salt instead, but it is usually easier to
solve for the total amount of salt. So let x(t) be this amount.
Then x(t) changes by
V0 gallons
dx
= salt salt
in − out .
x(0) = x0 lb
dt
This can be turned into an actual equation, if we use the
units to guide us:
R gal/min
dx gal lb gal x lb
= R c − R .
dt min gal min V0 gal
Again, note the importance of units! In both the ‘in’ and ‘out’ terms above, we can see
that we can simplify the fractions to always get units of lb/min. This is important, as dx
dt
must also have units of pounds per minute. After we’ve gotten the units straight, we can
drop them temporarily, so

C. Sprouse —Math 280 Notes Page 92 DRAFT August 26, 2022

1.7. Applications: Linear First Order ODE’s 93

dx R dx R
= cR − x ⇒ + x = cR.
dt V0 dt V0

∫ R
dt R
t
We can solve this using the integrating factor µ(t) = e V0 = e V0 .

R ′ R
t t
e V0 x = cRe V0

R R R
t t t
e V0 x = cR e V0 dt = cV0 e V0 + C.

R
t
Setting t = 0 gives C = x0 − cV0 , and dividing by e V0 gives the solution

x(t) = cV0 + (x0 − cV0)e

− VR t
0 .

Note that this starts at x0 and approaches cV0 as t → ∞. This is because for large time
the contents of the tank become the same as whatever is being fed in. So when the salt
concentration in the tank is the same as the input concentration, the total amount of salt
is this times the volume, or cV0 .

Example 1.7.7. A 1000 gallon tank contains 50 pounds of salt. Ocean water containing .3
lb/gal of salt is poured into the tank at a rate of 20 gal/min, while the water from the tank
drains at the same rate of 20 gal/min. Find the amount of salt in pounds, x(t).

solution:

dx salt − salt
= in out
dt
gal lb gal x lb
= 20 .3 − 20 .
min gal min 1000 gal

So the ODE to solve is

dx dx
= 6 − .02x ⇒ + .02x = 6.
dt dt
∫ .02d t .02t
The integrating factor is µ(t) = e =e , and

(e
.02t ′ .02t .02t .02t
x) = 1.5e ⇒ e x=6 e dt
.02t .02t
⇒ e x = 300e + C.

Setting x(0) = 50 gives C = −250 and we have

−.02t
x(t) = 300 − 250e

C. Sprouse —Math 280 Notes Page 93 DRAFT August 26, 2022

1.7. Applications: Linear First Order ODE’s 94

400
350 x(t) = 300 − 250e−.02t
300
250
lbs salt

200
150
100
50
0
30min 1hr 90min 2hr
t □.

Example 1.7.8. For most of the time from 100BC to 1500AD, the Colorado River did not
directly enter the Sea of Cortez as it does now, but flowed into a very large lake known as
2
(Ancient) Lake Cahuilla. This covered a 2100 mi region from present-day Indio in the north
to Mexicali in the south. When the lake reached a height of 40 feet above sea level, the water
was then able to drain to the current Colorado River Delta and into the sea. The present day
remnant of this lake is the Salton Sea, a much smaller (but still quite large) lake which sits at
the much lower elevation of 220 feet below sea level.
Since the lake was very large, we need to use large units. Water from the Colorado River
3
flowed in at a rate of about 4.5 mi /year. At its full height, water evaporated from the lake at
3 3
a rate of 2.4 mi per year. The rest of the excess water (2.1 mi /year) would drain to the sea,
3
so the lake had a constant volume of 56 mi . The salinity of the Colorado River is naturally
6 3 3
about 330 ppm. In our giant units this would be 1.5 ⋅ 10 tons/mi , or 1.5 megatons/mi . The
salinity of the lake itself could fluctuate, rather than always staying at a constant equilibrium.
So assume the salt in the lake is at an arbitrary initial value x0 , and solve for the amount of salt
x(t) as it approaches equilibrium.

solution: The crucial fact in the above is that the lake has a constant volume of water
3
(56 mi ). This means we can set up the ODE for the amount of salt as before:
3 3
dx mi Mton mi x Mton
= 4.5 year 1.5 3
− 2.1 year .
dt mi 56 mi3

So
dx dx
= 6 − .0375x ⇒ + .0375x = 6.
dt dt
∫ .0375 dt .0375t
An integrating factor is µ(t) = e =e , so

.0375t ′ .0375t .0375t .0375t

e x = 6e ⇒ e x = 160e + C.

.0375t
Setting x(0) = x0 gives C = x0 − 160. Then, dividing by e ,

x(t) = 160 + (x0 − 160)e

−.0375t
Mton.

C. Sprouse —Math 280 Notes Page 94 DRAFT August 26, 2022

1.7. Applications: Linear First Order ODE’s 95

Figure 1.9: The shoreline of ancient Lake Cahuilla is clearly visible from Highway 86, near
the town of Desert Shores. Rocks that were under water for long periods are darkened
by mineral deposits (travertine). This can be seen as a horizontal line in the Santa Rosa
mountains at an elevation of 40 feet.

To get the concentration, we divide by the volume of water, 56 mi3 :

c(t) = 2.86 + (c0 − 2.86)e

−.0375t 3
Mton/mi .

The more conventional measure of concentration is ppm (=mg/l). Omitting the

arithmetic, changing units gives

c(t) = 622 + (c0 − 622)e

−.0375t
ppm.

5 −.0375t
c(t) = 2.86 + Ce
1000
4
3

800
Mton/mi

ppm
2.86
600
2
400
1 200

0 3 6 9 12 15 18 21 24 27 30
years
□.

Remark 1.7.9. The equilibrium of 622 ppm salinity is acceptable for drinking water, which
begins to taste bad around 1000 ppm. So Lake Cahuilla was quite healthy back then. But
unfortunately the salinity of its replacement, the Salton Sea, is now 44000 ppm, exceeding that
of the ocean (35000 ppm). There is also a modern ‘Lake Cahuilla’ in the same region near the
town of La Quinta. Its salinity as of 2017 is 59000 ppm!

C. Sprouse —Math 280 Notes Page 95 DRAFT August 26, 2022

1.7. Applications: Linear First Order ODE’s 96

Remark 1.7.10. The solution above shows that our model is necessarily very rough. Because
e−.0375t decays so slowly, on the order of decades, any natural fluctuations in the salinity enter-
ing the lake would overwhelm the relatively slow progress towards equilibrium. So, contrary to
the graph above, we could never really expect a constant level of salinity to be reached.

Now we will make things more interesting. The last two examples involved equations
of the form dx
dt
+ αx = β, where α and β are constants. This is exactly the same as the basic
form of Newton’s law of cooling, where we had dT dt
−kT = −kTe . These constant coefficient
equations are both linear and separable. However, if we consider a tank for which the
volume of water changes over time, the resulting ODE will have non-constant coefficients.
Since these equations are not separable, this gives a more convincing application of linear
first order ODE’s.
Suppose that the water enters the tank at a rate Rin that is
differs from the exiting rate Rout . Then the volume of water
Rin gal/min
in the tank will change at a rate of
cin lb/gal
dV
= Rin − Rout .
dt
So
V (t) = V0 + (Rin − Rout )t.
Then our ODE for the amount of salt x(t) becomes V (0) = V0 gal
x(0) = x0 lb
dx salt − salt
= in out
dt
gal lb gal x lb
V0 + (Rin − Rout )t gal
= Rin cin − Rout . Rout gal/min
min gal min

Example 1.7.11. You have a 100 gallon feeding tank attached to a large saltwater aquarium.
The tank currently contains 40 gallons of water and no krill. You fill the tank with water than
contains .3 lb/gal of krill at a rate of 20 gal/min, while draining the tank into the aquarium at
a rate of 15 gal/min. How many pounds of krill are in the tank when the tank is full?

solution: The ODE is

dx salt − salt
= out
dt in
gal lb gal x lb
= 20 .3 − 15 .
min gal min 40 + 5t gal

So
dx 15
+ x = 6,
dt 40 + 5t

C. Sprouse —Math 280 Notes Page 96 DRAFT August 26, 2022

1.7. Applications: Linear First Order ODE’s 97

and
= (40 + 5t) .
∫ 15
dt 3 ln(40+5t) 3
µ(t) = e 40+5t =e
So

((40 + 5t) x) = 6(40 + 5t) ⇒ (40 + 5t) x

3 ′ 3 3 3
= 6(40 + 5t) dt

(40 + 5t) + C
6 1 4
=
4 5
4
= .3(40 + 5t) + C.
4
x(0) = 0 gives C = −.3(40) , and

x(t) = .3(40 + 5t) − .3(40) (40 + 5t)

4 −3
.

The tank is full when 40 + 5t = 100 ⇒ t = 12min. Then we have

x(12) = .3(100) − .3(40 )/(100 ) ≈ 29.2 lb □ .

4 3

30

20

12
10

0
0 2 4 6 8 10 12
t □.

The above answer makes physical sense, as for large time the concentration of krill
should approach the concentration entering the tank, in this case .3 lb/gal. So in 100
gallons of water, the total amount of krill should be about .3(100) = 30 lb. This is true in
general, and so we can infer that the amount of salt x(t) will always look like

x(t) = cin V0 + (Rin − Rout )t + C V0 + (Rin − Rout )t

−p

−p
= cin Vol Water + C Vol Water .

where cin is the concentration of salt in the water going in, and p = R R−R out
. When Rin >
in out
Rout −p < 0 but Vol Water → +∞, so the right hand term → 0. And when Rin < Rout ,
−p > 0 but Vol Water → 0, meaning that the right hand term still goes to 0! This means
that x(t) always approaches the line cin ( V0 +(Rin −Rout )t ) = cin (Vol Water) as t increases.
We can also divide x(t) by the volume of water to get the concentration of salt in the
tank:

C. Sprouse —Math 280 Notes Page 97 DRAFT August 26, 2022

1.7. Applications: Linear First Order ODE’s 98

Rin > Rout Rin < Rout

x0
x0
cin V0

cin V0

x0
x0

t t

Figure 1.10: The possibilities for the solution x(t) = cin (Vol Water) + C(Vol Water)
−p
.

c(t) = cin + C V0 + (Rin − Rout )t

−q −q
= cin + C Vol Water ,

Rin
where q = p + 1 = Rin −Rout
. Again, the term on the right always → 0 as t increases.
Example 1.7.12. Lake Havasu is a man made reservoir on the Colorado River, with outflow
controlled by Parker Dam. It has a maximum capacity of about 650 kilo-acre-feet (kaf). Suppose
that the volume of the lake is currently 500 kaf and the salinity is 1200 tons/kaf, which is rather
high (880ppm). The EPA sets a standard of 1020 tons/kaf (749ppm) for this section of the river.
Suppose water with a much better salinity of 800 tons/kaf (590ppm) is now coming in
at a rate of 20 kaf/day. You set the outflow to 16 kaf/day for one month (30 days). What is the
volume and salinity of the water after this? Did you meet the standard?

solution: We have
dx kaf tons kaf x tons
= 20 800 − 16 .
dt day kaf day 500 + 4t kaf
So the ODE to solve is
dx 16
+ x = 16000.
dt 500 + 4t
The integrating factor is

= (500 + 4t) .
∫ + 500+4t
16
dt 4ln(500+4t) 4
µ(t) = e =e
So
((500 + 4t) x) = 16000(500 + 4t)
4 ′ 4

(500 + 4t) x (500 + 4t) dt

4 4
⇒ = 16000
5
= 800(500 + 4t) + C.

C. Sprouse —Math 280 Notes Page 98 DRAFT August 26, 2022

1.7. Applications: Linear First Order ODE’s 99

Figure 1.11: Parker Dam, with water exiting to the Colorado River. Lake Havasu is on the
other side.

Since the concentration of salt at t = 0 is 1200, and the volume of water is 500 we have
x(0) = (500)(1200) = 600, 000. This gives C = 200, 000 × (500)4 = 1.25 ⋅ 1016 . Dividing by
(500 + 4t) ,
4

x(t) = 800(500 + 4t) + 1.25 ⋅ 10 ⋅ (500 + 4t) □ .

16 −4

1200 16
900
1.25⋅10
c(t) = 800 + (500+4t)5
1100
800
tons/kaf

ppm
1000
700
900

600
800
0 5 10 15 20 25 30 35 40
days

1.7.4 Circuits
Circuits with passive elements (resistors, capacitors, inductors) provide an abundant
source of linear ODE’s with constant coefficients. This is due to the voltage/current rela-
tionship on each element. We won’t get into the physics much, but recall that current is
the flow of electrons through some pathway, and voltage is the ‘force’ (more accurately,
the potential energy from an electric field) that moves the electrons down the path. The
resistance R of the path measures how difficult it is to move the electrons, which by
Ohm’s law has the linear relationship I = V /R. So, we have V = IR across a resistor. For
inductors and capacitors, the voltage-current relationship involves a derivative. Recall

C. Sprouse —Math 280 Notes Page 99 DRAFT August 26, 2022

1.7. Applications: Linear First Order ODE’s 100

that the most basic inductor is a wire, wound in a tight coil. The basic capacitor is two
small conducting plates separated by a thin gap. The resulting current/voltage relation-
ships are:

dVC
VR = RI VL = L dI
dt I =C dt
+ − + − + −

I I I

An RL circuit is one that has a voltage source, resistor, and inductor connected in
series.

Vin + L
− I

Figure 1.12: RL Circuit

For an RL circuit, deriving the ODE is particularly easy by using Kirchoff’s voltage
law: the sum of the voltages around any closed loop must = 0. It is a standard convention
to consider the applied voltage to be negative with respect to the other elements. So going
around clockwise, Kirchoff’s law gives:

−Vin + VR + VL = 0.

Using VL = L dI
dt
from above, we immediately get a first order ODE for the current:

dI
L + RI = Vin
dt

Let’s solve this in the case where Vin is constant. We divide by L and get the integrating
R
factor µ(t) = e L t .
R
t ′ Vin RL t R
t Vin RL t
eL I = e ⇒ eL I = e + C.
L R
So the general solution is
Vin −R t
I= + Ce L .
R

C. Sprouse —Math 280 Notes Page 100 DRAFT August 26, 2022
1.7. Applications: Linear First Order ODE’s 101

V
Notice that as t → ∞, the current goes to Rin , which would be the current through
the resistor if there was no inductor present at all. This means there can be no voltage
left to go across the inductor. We can also see this by observing that VL = L dIdt
gives
− Rt
VL = −C RL e L → 0 as t → ∞. Physically, inductors welcome constant current, but avoid
nonzero constant voltage. (Constant voltage across an inductor is possible, but requires a
steadily increasing current I = at + b, which is not sustainable for very long.)

current I (t) voltage VL (t)

I0

Vin V0
R

I V 0

I0
V0
0
t t

Figure 1.13: Current and voltage across inductor, for Vin constant.

Example 1.7.13. DC motors are a common application of inductance, where the current through
a coil creates an electromagnetic field sufficient to turn an axle, creating motion. A Moog
C23− L50 motor has a resistance of 20 and inductance of .032 H seen at the input terminals,
and is designed to run at a rated current of 1.1 amps. Suppose the motor is initially turned off,
and you apply a 24 Volt source to the inputs. How long until the rated current is achieved?

solution: The circuit equation is

dI dI
.032 + 20I = 24 ⇒ + 625I = 750,
dt dt
with I (0) = 0. The integrating factor is µ(t) = e
−635t
, so

(e
625t ′ 625t 625t 625t
I) = 750e ⇒ e I = 750 e dt
625t
= 1.2e + C.
So using I (0) = 0 and dividing by µ(t),

I (t) = 1.2 − 1.2e

−625t
.
The time it takes to get to 1.1 amps is then
−625t 1 .1
1.2 − 1.2e = 1.1 ⇒ t=− ln ≈ .004 sec. □ .
625 1.2

C. Sprouse —Math 280 Notes Page 101 DRAFT August 26, 2022
1.7. Applications: Linear First Order ODE’s 102

Figure 1.14: A 12V DC Motor (much cheaper than the one used in the example!).

current I(t)

1.2 A
1.1 A

0
0 2 4 6 8
t (msec)
□.

Next we consider the RC circuit, with one resistor and one capacitor:

Vin + C
− I

Figure 1.15: RC Circuit

It is slightly less straightforward to derive the ODE for this, but not too much harder.

C. Sprouse —Math 280 Notes Page 102 DRAFT August 26, 2022
1.7. Applications: Linear First Order ODE’s 103

Summing voltages gives

−Vi n + VR + VC = 0 ⇒ −Vi n + RI + VC = 0

.
dVC
But now we have I = C dt
, so we cannot replace VC in the above. But then now we
can replace I :
dVC dVC
−Vin + R C + VC = 0 ⇒ RC + VC = Vin .
dt dt
Setting V (t) = VC for simplicity of notation gives the RC circuit equation

dV
RC + V = Vin
dt

Once again this is a constant coefficient linear first order ODE. And again we solve for
the case Vin constant:

dV 1 Vin 1
t
+ V = ⇒ µ(t) = e RC .
dt RC RC
1 ′ Vin RC
1 1 1
t t t t
e RC V = e ⇒ e RC V = Vin e RC + C,
RC
1
− RC t
⇒ V = Vin + Ce .
As t → ∞ we have that VC = V (t) → Vi n . Since −Vin +VR +VC = 0, this can only happen
if the voltage across the resistor goes to zero. But that means the current across the resistor
gives T (t) = − RC
1
C − RC t
goes to zero. We can see this is the case, since I = C ddt
VC
e → 0 as t →
∞. So a capacitor avoids nonzero constant current. On the other hand, VC → Vi n shows
the capacitor easily holds nonzero voltage. This is in perfect contrast to the inductor,
which we found to accept constant current and reject nonzero constant voltage.
Example 1.7.14. Consider an RC circuit with R = 4000 ,C = .005 F, and Vin = −5 V.
1. Solve for the voltage V (t) = VC (t) across the capacitor if V (0) = 5 V.
2. Suppose an LED with large series resistance (so that it doesn’t change the output VC too
much) is attached across the capacitor. It requires VC > 4 volts to operate. Find T such
that VC (t) ≥ 4 for 0 ≤ t ≤ T . In this way, an LED designed to operate at VC = 5 volts
will not be effected if there is a brief reversal of voltage for that amount of time.
solution: We have
dV dV dV
RC + V = Vin ⇒ 20 + V = −5 ⇒ + .05V = −.25.
dt dt dt
∫ .05d t .05t
The integrating factor is µ(t) = e =e .

(e V ) = −.25e
.05t ′ .05t .05t .05t −.25 .05t
⇒ e V = −.25 e dt = e + C.
.05

C. Sprouse —Math 280 Notes Page 103 DRAFT August 26, 2022
1.7. Applications: Linear First Order ODE’s 104

current I (t) voltage VC (t)

I0 V0

Vin

I 0 VC

I0
V0

0
t t

Figure 1.16: Current and voltage across capacitor, for Vin constant.

So dividing by e.05t , V = −5 + Ce −.05t . And setting x(0) = 5 gives C = 10. So

V (t) = −5 + 10e
−.05t
.

To determine how long the LED can remain operational, set V (t) = 4.

−.05t 1 9
−5 + 10e = 4. ⇒ t=− ln ≈ 2.1.
.05 10

So the LED can operate for 2.1 more seconds.

5
4
3
2
1
voltage VC (t) 0
−1
−2
−3
−4
−5
0 2.1sec 5 10 15 20
t (sec)
□.

−5
Example 1.7.15. Consider an RC circuit with R = 2000 , C = 5 ⋅ 10 F, and Vin = sin(ωt).
If V (t) is the general solution for the voltage VC across the capacitor, then the steady-state
voltage VS (t) is a periodic solution such that V (t) approaches VS (t) for large time t. Find the
steady-state voltage across the capacitor.

C. Sprouse —Math 280 Notes Page 104 DRAFT August 26, 2022
1.7. Applications: Linear First Order ODE’s 105

solution: The equation is

dV dV dV
RC + V = Vin ⇒ .1 + V = sin(ωt) ⇒ + 10V = 10sin(ωt).
dt dt dt
∫ 10dt 10t
So the integrating factor is µ(t) = e =e and

10t ′ 10t 10t 10t

e V = 10e sin(ωt) ⇒ e x= 10e sin(ωt)dt.

The above integral is usually solved by integrating by parts twice. It’s a little long so we
omit the computation. It gives

10t −10ω 10t 10 10t

e V= e cos(ωt) + e sin(ωt) + C.
100 + ω2 100 + ω 2
So the general solution is

V (t) =
−ω 10 −10t
cos(ωt) + sin(ωt) + Ce .
100 + ω 2 100 + ω2
−10t −10t
We can see that as t increases, the Ce term becomes small. That is, lim e = 0. So
t→∞
the steady-state solution is

VS (t) =
−ω 10
cos(ωt) + sin(ωt).
100 + ω2 100 + ω 2

.1
.08 ω=2
.06 ω = 10
.04 ω = 50
.02
VC

0.0
-.02
-.04
-.06
-.08
-.1 π π 3π π
4 2 4
t

Figure 1.17: Output voltage VC (t) for Vin = sin(ωt). □.

Remark 1.7.16. In the above picture, one should notice that the solution amplitudes get smaller
as ω increases. We will show in a future section that VS (t) can be written as √ 10 2 sin(ωt +
100+ω
φ) for some constant φ that depends on ω. So, the amplitude of VS (t) is √ 10 . This
100+ω 2
amplitude goes to zero for large ω. So the output voltage VC will be low for high frequency
signals, and relatively high for low frequency signals. A circuit with this type of output response
is called a low pass filter. The RC circuit above is the simplest basic example of this.

C. Sprouse —Math 280 Notes Page 105 DRAFT August 26, 2022
1.7. Applications: Linear First Order ODE’s 106

1.7.5 Skydiving
Suppose an object is dropped from a certain height above the surface of the earth. We
2
know that gravity provides a constant downward acceleration of g = 32 ft/sec = 9.8
2
m/sec . So Newton’s second law gives that the force of gravity on the object is Fgravity =
ma = mg. To derive a first order ODE from this, we consider the velocity of the object,
rather than the position. We have
dv dv
ma = m ⇒ m = mg.
dt dt
We know that in reality when an object falls through the air, its speed does not increase
indefinitely as time increases. This is because there is a force of wind resistance, which
opposes motion in proportion to the velocity: Fresistance = −kv. So the total force on the
object is

dv
m = Ftotal = Fgravity − Fresistance
dt
Or,

dv
m = mg − kv.
dt

Again, this is a constant coefficient linear ODE.

dv dv k
m + kv = mg ⇒ + v =g
dt dt m
k
So the integrating factor is e m t and
′
k
t k
t k
t mg k t
em v = ge m ⇒ em v = e m + C.
k
So the general solution for the velocity is
mg −k t
v= + Ce m .
k
mg
The quantity k is called the terminal velocity, which is the constant equilibrium speed
that the falling object will eventually obtain.
The position above is given by
t t mg mg
−ks −k t
p(t) = p0 − v(s) ds = p 0 − + Ce m ds = p 0 − t + Ce m ,
0 0 k k
the negative sign due to the velocity being downward.
Example 1.7.17. A weather balloon carries a 3.3-lb instrument, with a parachute designed to
achieve a terminal velocity of = 16.133 ft/sec (=11 mph). Find k for this instrument. Then
solve for v(t) if the parachute deploys near freefall, v(0) = 105 mph (= 154 ft/sec).

C. Sprouse —Math 280 Notes Page 106 DRAFT August 26, 2022
1.7. Applications: Linear First Order ODE’s 107

velocity v(t) position p(t)

mg
v0 > k p0

mg
k
mg
p0 − k t

p0

v0 = 0
t t

Figure 1.18: Velocity and position for falling object with parachute.

solution: Recall that pounds measure force (not mass), and the weight of the object is
mg = 3.3 lb. So

mg 3.3 3.3
= 16.133 ⇒ = 16.133 ⇒ k= ≈ .205.
k k 16.133

So the ODE is
dv
m = 3.3 − .205v.
dt

But again,
3.3
mg = 3.3 ⇒ m= = .103125.
32

So dividing the above ODE by .103125,

dv dv
= 32 − 1.99v ⇒ + 1.99v = 32.
dt dt

µ(t) = e1.99t and

(e
1.99t ′ 1.99t 1.99t 1.99t
v) = 32e ⇒ e v = 16.1e v + C.

Setting v(0) = 154 and dividing by the integrating factor gives

−1.99t
v(t) = 16.1 + 137.9e □.

C. Sprouse —Math 280 Notes Page 107 DRAFT August 26, 2022
1.7. Applications: Linear First Order ODE’s 108

velocity v(t)

154 105
ft/sec.

mph
16.1 11

0 0.5 1 1.5 2
t sec. □.

C. Sprouse —Math 280 Notes Page 108 DRAFT August 26, 2022
 1.7. Applications: Linear First Order ODE’s 109

Homework 1.7:

= k(T − Te ) , RL circuit ∶ L + RI = Vin

dT dI
NLC ∶
dt dt
dv dV
skydiving ∶ m = mg − kv , RC circuit ∶ RC + V = Vin
dt dt

A) Solve the word problems.

1. Suppose that an alien spore infects the earth. Every two years a single spore divides into 3 spores, by
consuming mass from the earth.
dy
a) Solve the ODE dt
= ky for y(t), and find k.
b) The earth has mass 6 ⋅ 1024 -kg. If each spore weighs 5 milligrams and there was initially one
spore, how many years until the earth is completely consumed?
2. A population y(t) has a time-dependent rate of reproduction R(t), so that

dy
= R(t)y.
dt
a) Solve for y(t), if y(0) = y0 . Your answer should be in terms of a definite integral.
b) Use your solution from a) to show that after T years, the population is

y(T ) = y0 e
AT
,

where A is the average value of R(t) over the interval [0, T ].

3. You have a worm farm, and on average 5 worms make a new worm every week, which means that
the population increasesat a rate of .2, times the number of worms already there, per week. Simulta-
neously, you also remove 50 worms/week to feed to birds.
a) Write down an ODE for the number of worms, and find the general solution.
b) Find the number of worms y(t), supposing that you start with 300 worms. How many worms
are there at the end of six weeks?
◦
4. Coffee is brewed at a temperature of about 200 , and a standard cup of coffee has k = −.025 in
minutes.
a) Solve for T (t), if a fresh brewed cup of coffee is placed in a room of temperature 75 .
◦

b) Graph T (t).
◦
c) If you want to drink the coffee at a temperature of 160 , how long does it take to cool to this
temperature?
◦ ◦
5. You have an oven preheated to 425 , and you want to heat a potato to a temperature of 210 .
a) Solve for T (t) for a potato that is initially at room temperature, 75 .
◦

b) Suppose that a room temperature potato takes 40-minutes to cook to the correct temperature in
this oven.Find k.
◦
c) Now suppose instead you have frozen potato at an initial temperature of −10 , which is placed
in the same425 oven. Solve for T (t) using the value of k from above.
◦

d) Graph T (t).
◦
e) How long does it take to get to 210 for the frozen potato?
◦
6. You take a can of soda from the trunk of your car, which is at a temperature of 180 , and put it into
◦
a freezer of temperature 10 .
a) Solve for T (t).

C. Sprouse —Math 280 Notes Page 109 DRAFT August 26, 2022
1.7. Applications: Linear First Order ODE’s 110

◦
b) Suppose that after 15 minutes the can has temperature 120 . Solve for k.
c) How long does it take until the can is at a temperature of 40◦ ?
◦
7. Suppose that you take an outside thermometer into your house, where the temperature is 70 .
a) Solve for the temperature of the thermometer, with T (0) = T0 .
◦ ◦
b) After 5 minutes the thermometer reads 95 , and after 10 minutes the thermometer reads 85 .What
is the temperature outside?
◦
8. You take some waffles from the freezer of temperature 5 . Your waffles are known to have k = −.05
in minutes.
a) You put the waffles in an oven, and then turn on the oven so that its temperature is 80 + 15t.
Solve for T (t).
◦
b) How long does it take for the oven to reach a temperature of 350 ?
c) What is the temperature of the waffles at this time?
9. A tank contains 100 gallons of water and 50 lbs of salt. Water containing .2 lb/gal of salt enters the
tank at a rate of 3 gal/min. Water drains from the tank at a rate of 3 gal/min.
a) Find the amount of salt in the tank, x(t).
b) Graph it.
c) When is the concentration of salt in the tank equal to .25 lb/gal?
10. A tank contains 500 liters of water and no chlorine. Water containing 10 mg/liter of chlorine enters
the tank at a rate of 5 L/min and water drains from the tank at a rate of 5 L/min.
a) Find the amount of chlorine in the tank, x(t).
b) Graph it.
c) When is the concentration of chlorine in the tank equal to 2 mg/liter?
11. A 600 gallon tank contains 400 gallons of water and 80 pounds of salt. Water containing .1 pounds of
salt per gallon enters the tank at a rate of 3 gal/min and drains from the tank at a rate of 1 gal/min.
a) Find the amount of salt in the tank, x(t).
b) What is the concentration of salt when the tank overflows?
12. A jug contains 10 liters of water and no iodine. Water containing 5 mg/liter of iodine is poured in at
a rate of 4 liters/min. Water drains out at a rate of 3 liters/min.
a) Solve for the amount of iodine as a function of time.
b) What is the concentration of iodine when the jug contains 15 liters of water?
13. A tank contains 200 gallons of water and 100 pounds of salt. Water containing .1 pounds of salt per
gallon enters the tank at a rate of 2 gal/min and drains from the tank at a rate of 4 gal/min.
a) Find the amount of salt in the tank, x(t).
b) Find the concentration of salt in the tank, c(t).
c) When is the concentration of salt in the tank equal to .2 lb/gal?
14. A pond contains 500 kL of water.The natural amount of phosphorus in a pond is .02 g/kL. During a
rainstorm, agricultural runoff containing .5 g/kL of phosphorus enters the pond at a rate of 6 kL/day,
while water from the pond drains out at a rate of 2 kL/day.
a) What is the initial amount of phosphorus in the pond?
b) Set up an initial value problem for the amount of phosphorus in the pond, and solve it.
c) How much phosphorus is in the pond after 10 days?
d) What is the concentration of phosphorus in the pond after 10 days?
15. A tank contains 10 gallons of water and no salt. Water with .3 lb/gallon of salt is poured in at a rate
of 4 gal/min. Water drains out at a rate of 6 gal/min.
a) Set up the differential equation for the amount of salt in the tank, x(t). Then, without solving for
x, find when the concentration of salt is maximumby setting dx/dt = 0. (Hint: The concentration
c = x/V occurs in the right-hand side of the equation, where V is the volume of water. For what
value of c is the right-hand side zero?)

C. Sprouse —Math 280 Notes Page 110 DRAFT August 26, 2022
1.7. Applications: Linear First Order ODE’s 111

b) Solve for x(t).

c) Find the time that the maximum occurs at, by setting x′ (t) = 0.
d) Find x(t) and V (t) for the value of t in part c). Then verify that your answer in part a) was
correct.
16. An RL circuit has L = 10 H, R = 500 , Vin = 0. Suppose the initial current through the circuit is
i(0) = .02 A.
a) Find the current through the inductor.
b) Graph it.
c) Find the voltage through the inductor.
d) Graph it.
17. An RL circuit has L = 5 H, R = 200 , Vi n = 10 V. Suppose the initial current through the circuit is
i(0) = .02 A.
a) Find the current through the inductor.
b) Graph it.
c) Find the voltage through the inductor.
d) Graph it.
18. A charged capacitor holding an initial voltage of V (0) = 100 V is hooked to the circuit below, and
theswitch is closed at t = 0.
250

+
20 V .04F
−

a) Solve for the voltage VC (t).

b) How long does it take for the capacitor to discharge to a voltage of VC = 21 volts?
c) Sketch the voltage VC (t).
19. In the circuit below, the voltage across the capacitor is zero when the switch is closed.Suppose the
switch is opened at time t = 0 to charge the capacitor.
2000

I
+
24 V .01F
−

a) Solve for the capacitor voltage VC (t) for t ≥ 0 by using the RC-circuit equation.
b) Sketch VC (t) for −∞ < t < ∞.
c) Find I (t) for t > 0, and use V = IR to find I(t) for t < 0.
d) Verify that I (t) is continuous at t = 0. Sketch I (t) for −∞ < t < ∞.
20. An RL circuit has L = 2 H, R = 50 , Vin = 250t mV. Suppose the initial current through the circuit
is i(0) = 1 mA.
a) Find the current through the inductor.

C. Sprouse —Math 280 Notes Page 111 DRAFT August 26, 2022
1.7. Applications: Linear First Order ODE’s 112

b) Find the voltage through the inductor.

c) Graph it.
21. An RC circuit has R = 100 , C = .005 F, and Vin = 10cos(t). Find VC (t) if VC (0) = 10.You can use
aeat cos(bt)+be at sinbt
that ∫ e cos(bt) dt =
at
a2 +b 2
.
22. An RL circuit has L = 2 H, R = 160 , Vin = 3000 sin(60t) mV.
at at
a) Find the current through the inductor. You can use that ∫ e sin(bt) dt =
at ae sin(bt)−be cosbt
a2 +b 2
.

✗ ✔
b) Find the voltage through the inductor.

2 2
For the following questions, use either g = 32 ft/sec or g = 10 m/sec .
Be careful about the difference between mass and weight.
✖ ✕
23. A parachute with k = .2 N-sec/m carries an object of mass .5-kg.
a) Without solving for v(t), find the terminal velocity by setting dv dt
= 0.
b) Suppose the object thrown upwards, so that v(0) = −10 m/sec. Solve for v(t).
c) Graph v(t).
24. A parachute has k = 2 ft-lb/sec.
a) Suppose that an object dropped from the air is to reach a maximum velocity of 64-ft/sec. What
should the weight of the object be?
b) Suppose the object is ejected downwards with an initial velocity of 16 ft/sec. Solve for v(t).
c) Graph v(t).
25. Suppose that a parachute is designed so that for a skydiver of weight 160-lb, the equilibrium velocity
will be 10 ft/sec.
a) Solve for k.
b) The free fall velocity of a human is about 180 ft/sec. Solve for v(t) assuming the above parachute
opens with v(0) = 180.
c) Graph it.
d) How long does it take for the velocity to reach 11 ft/sec?
26. Suppose that a parachute is designed so that for a skydiver of mass 100-kg, the equilibrium velocity
will be 5 m/s.
a) Solve for k.
b) The free fall velocity of a human is about 60 m/sec. Solve for v(t) assuming the above parachute
opens with v(0) = 60.
c) Graph it.
d) How long does it take for the velocity to reach 6 m/sec?
27. A dropsonde is a weather sensor attached to a parachute, which is dropped from an airplane over
the ocean.These are used off the coast of California to collect data for future weather prediction.
a) The data sheet for the Vaisala RD94 dropsonde gives mass of .35 kg, and a terminal velocity of
11 m/sec. Find kfor the RD94.
b) Solve for the velocity v(t), if the RD94 is dropped from a plane with v(0) = 0.
c) How far would this RD94 have fallen after 10 seconds? (Hint: position is the antiderivative of
velocity.)

C. Sprouse —Math 280 Notes Page 112 DRAFT August 26, 2022
Chapter 2

Higher Order Linear ODE’s

2.1 General Theory: Homogeneous Linear n-th order ODE’s

Definition 2.1.1. An n-th order ODE is called Linear if it of the form

(n) (n−1)
an (x)y + an−1 (x)y + ... + a1 (x)y + a0 (x)y = f (x).
′
(⋆)

A very simple example of ⋆ occured in Section 1.1:

d2 y
= −g.
dt 2

Recall that this represents the height of an object, such as a ball thrown into the air, with
constant downward acceleration due to gravity. In calculus this equation is introduced
with antiderivatives, to show how integration constants can be used to solve initial value
problems. That is, suppose the initial position y0 and initial velocity v0 of the ball are
given as
y(0) = y0 ,

y (0) = v0 .
′

Then one can solve for the exact position of the ball as a function of time by integrating
twice:
1 2
y(t) = − gt + v0 t + y0 .
2

Thus we are guaranteed a unique solution as long as we are given two initial values y(0)
and y (0). This is extremely important in real life: it means that when we throw a ball
′

with a given initial position and velocity, we can predict exactly where it will go. So, as
done in Calculus, it is possible to determine the maximum height the ball attains, the
time that it hits the ground, etc.

113
2.1. General Theory: Homogeneous Linear n-th order ODE’s 114

y ft. 2
150 y(t) = − 12 gt + v0t + y0

100
v0
50
y0 ⋅
t sec.
1 2 3 4 5 6

The above is in fact true for a general n-th order linear ODE ⋆ as well, provided the
ODE is nonsingular:
Definition 2.1.2. The n-th order linear ODE
(n) (n−1)
an (x)y + an−1(x)y + ... + a1 (x)y + a0(x)y = f (x)
′
(2.1)

is called nonsingular on (a, b) if a0 (x),...,a n (x), f (x) exist and are continuous on all of (a, b)
and an(x) is never zero on (a, b).
The condition that an (x) cannot be zero on (a, b) is crucial. Note that we can also write
⋆ in standard form as
(n) an−1 (x) (n−1) a1 (x) ′ a0(x) f (x)
an (x) an (x) an(x) an(x)
y + y + ... + y + y= .

So the requirement that an (x) is never zero is required for the equation in standard form
to also be nonsingular on (a, b), because any zeroes of an (x) would make the above co-
efficients discontinuous. The importance of being nonsingular is that we always get a
unique solution given an appropriate set of initial values, exactly like what we had for
the thrown ball above.

Theorem 2.1.3 (Existence/Uniqueness Theorem for Linear IVP’s). Consider the

n-th order linear initial value problem

an (x)y (n) + an−1 (x)y (n−1) + ... + a1 (x)y ′ + a0 (x)y = f (x), ⎫

⎪
⎪
⎪
y(x0 ) = b1 , ⎪
⎪
⎪
⎪
y (x0 ) = b2 , ⎪
⎬
⎪
′
⎪
(IVP)
⎪
⎪
⋮ ⎪
⎪
(n−1) ⎪
⎪
y (x0 ) = bn . ⎭

Suppose the ODE is nonsingular on (a, b). That is, an (x), ..., a0 (x), f (x) are contin-
uous and an(x) is never zero on (a, b). Then for any x0 ∈ (a, b) there exists a unique
solution to IVP on (a,b).

So for nonsingular initial value problems, things are extremely straightforward: an n-

(n−1)
th order ODE has unique solution if and only if n initial values y(x0 ),y (x0 ), ..., y (x0 )
′

C. Sprouse —Math 280 Notes Page 114 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 115

are specified. But to actually find these solutions, we need to look more closely at what
the general solution looks like. For the linear ODE ⋆, the general solution always has a
very specific structure.
For a simple example, again we go back to Section 1.1 (pg. 9) where we looked at
′′
y + y = 0.
′′
Example 2.1.4. Find the unique solution to the Initial Value problem y + y = 0, y(0) =
b1 , y (0) = b2 .
′

solution: Recall from Example 1.1.17, it is easy to check cos(x) and sin(x) are solutions
′′
to y + y = 0 by plugging in. It is then a small step to recognize that
y = c1 cos(x) + c2 sin(x) also satisfies the ODE for any pair of constants c1 , c2 . Again, this
can be checked by plugging into y ′′ + y and verifying that one gets 0 out. With a solution
y = c1 cos(x) + c2 sin(x) we have that y (x) = −c1 sin(x) + c2 cos(x) and we can plug in for
′

the initial values:

✟✟✯1 ✟✟✯0
y(0) = b1 ⇒ cos(0)
c1✟ + c2✟
sin(0) = b1 ⇒ c1 = b1 ,

✯0 ✯1
y (0) = b2 ✟✟ ✟✟
′
⇒ −c1✟
sin(0) + c2✟
cos(0) = b2 ⇒ c 2 = b2 .
Therefore we have a solution to the IVP:

y = b1 cos(x) + b2 sin(x).

Note that 2.1.3 then gives that it is the only solution to the IVP, as solutions to
nonsingular initial value problems are unique.

Going a little further, we also get the following:

′′ ′′
Example 2.1.5. Find the general solution to y +y = 0. That is, find all solutions to y +y = 0.

solution: Let y be any solution to y′′ + y = 0. Then define c1 , c2 by c1 = y(0) and

c2 = y (0). Then replacing b1 ,b2 in 2.1.4 by c1 , c2 gives that the function
′

c1 cos(x) + c2 sin(x) is a solution to y′′ + y = 0 with the same initial values as y. Again,
′′
Theorem 2.1.3 gives that this can be the only solution to y + y = 0 with these initial
′′
values. Therefore any solution to y + y = 0 must be of the form

y = c1 cos(x) + c2 sin(x).

So the set of functions {cos(x), sin(x)} gives all solutions to y + y = 0, and also can be
′′

used to get the unique solution to any initial value problem. {cos(x), sin(x)} is called a
′′
fundamental solution set to the ODE y + y = 0.
The distinguishing feature of linear ODE’s is that they all work just like this. In the
rest of the section we will explore this in the general case, which will necessarily be some-
what theoretical compared to what we have done so far.

C. Sprouse —Math 280 Notes Page 115 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 116

2.1.1 The Wronskian and Fundamental Solution Sets

Recall that for an n × n matrix A, the determinant of A, det(A) is a number. For 2 × 2
matrices we have

a b a b
det = = ad − bc.
c d c d
We will mostly be concerned with the 2 × 2 matrices, but in larger cases the determi-
nant can be computed by expansion. For instance, expanding along the top row of a 3 × 3
matrix gives

a b c
e f d f d e
d e f = a −b +c
h i g i g h
g h i
= a(ei − f h) − b(di − f g ) + c(dh − eg)
= aei − af h − bdi + bf g + cdh − ceg.

Definition 2.1.6. If y1(x),..., yn (x) are a collection of functions which are differentiable n − 1-
times, then the Wronskian of y1 (x), ..., yn (x) is the function

y1 (x) ⋯ yn (x)
y1 (x) yn (x)
′ ′
W (x) =
⋯
⋮ ⋮
(n−1) (n−1)
y1 (x) ⋯ yn (x)

Example 2.1.7. Find the Wronskian of {x, x , x }.

5 −1

solution:
5 −1
x x x
W (x) = 1 5x
4
−x
−2
.
3 −3
0 20x 2x
Expanding along the left column gives
5 −1
x x x
W (x) = 1 5x 4 −x −2
3 −3
0 20x 2x
4 −2 5 −1 5 −1
5x −x x x x x
= x 3 −3 − 1 3 −3 + 0 4 −2
20x 2x 20x 2x 5x −x
= x(5x (2x − (−x )(20x ) − 1(x (2x )−x (20x ))
4 −1 −2 3 5 −3 −1 3

= x(10x + 20x) − (2x − 20x ) = 48x □ .

2 2 2
=

The significance of the Wronskian is that it completely determines when an initial

value problem can be solved in terms of the functions {y1 , ..., yn }:

C. Sprouse —Math 280 Notes Page 116 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 117

Fact 2.1.8. Let {y1 (x), ..., yn (x)} be any set of n − 1-times differentiable functions, W (x) be
their Wronskian, and x0 ∈ R. Then W (x0 ) ≠ 0 if and only if for any set of initial values b1 , ...b n
there exists a unique set of coefficients c1 , ..., cn such that the function

y = c1 y1 (x) + ... + cnyn (x)

has initial values

y(x0 ) = b1
y (x0 )
′
= b2
⋮
(n−1)
y (x0 ) = bn

Proof. Let y = c1 y1 (x) + ... + cnyn (x). Then taking derivatives gives

y(x) = c1 y1 (x) + ... + cn yn (x)

y (x) = c1 y1 (x) + ... + cn yn (x)
′ ′ ′

⋮
(n−1) (n−1) (n−1)
y (x0 ) = c1 y1 (x) + ... + cn yn (x)

Plugging in x0 and using the above initial conditions, we have

c1 y1 (x) + ... + cn yn (x) = b1

c1y1 (x0) + ... + cn yn (x) = b2
′ ′

⋮
(n−1) (n−1)
c 1 y1 (x0) + ... + cn yn (x0 ) = bn .

In matrix form, this is

⎡
⎢ y1 (x0 ) yn (x0 ) ⎤ ⎥ ⎡ c1 ⎤ ⎡ b1 ⎤
⎢
⎢
...
⎥
⎥ ⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥
⎢
⎢ (x ) (x ) ⎥
⎥ ⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
′ ′
⎢ ⎥ ⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥
y ... y
⎢ ⎥
c b
⎢ ⎥ ⎢ ⎥
1 0 0
⎢ ⎥
n 2 2
⎢ ⎥ ⎢ ⎥
=
⎢ ⎥
.
⎢ ⎥
⎥⎢⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎥ ⎢ ⎢ ⎥
⋮ ⋮ ⋮ ⋮
⎢
⎢ ⎥
(x0 )⎥
(n−1) (n−1)
⎣y1 (x0 ) ... yn ⎦ ⎣cn ⎦ ⎣bn ⎦
From linear algebra, we have that an n×n matrix A is invertible if and only if det(A) ≠
0. The determinant of the above matrix is given by W (x0 ), the Wronskian at x0 . Therefore,
we get a unique solution for c1, ..., c n by the formula

⎡ c1 ⎤ ⎡
⎢ y1 (x0 ) yn (x0 ) ⎤ ⎥
−1
⎡b1 ⎤
⎢
⎢ ⎥
⎥ ⎢
⎢
...
⎥
⎥ ⎢
⎢ ⎥
⎥
⎢ ⎥
⎢ c2 ⎥ ⎢
⎢ (x ) (x ) ⎥
⎥ ⎢
⎢ ⎥
⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
′ ′
⎢
⎢ ⎥
⎥ ⎢ ⎥ ⎢
⎢ ⎥
⎥
y ... y
⎢ ⎥
b
⎢ ⎥ ⎢ ⎥
1 0 0
⎢ ⎥
n 2
⎢ ⎥ ⎢ ⎥
=
⎢ ⎥
,
⎢
⎢ ⎥
⎥ ⎢ ⎥ ⎢
⎢ ⎥
⎥
⋮ ⋮ ⋮ ⋮
⎢ ⎢ ⎥
⎣cn ⎥⎦ ⎢ (x0 )⎥ ⎢ ⎥
(n−1) (n−1)
⎣y1 (x0 ) ... yn ⎦ ⎣bn ⎦

if and only of W (x0) ≠ 0.

From now on in this section, we will only be concerned with homogeneous linear
ODE’s:

C. Sprouse —Math 280 Notes Page 117 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 118

Definition 2.1.9. The linear ODE ⋆ is called homogeneous if f (x) ≡ 0. That is,

(n) (n−1)
an (x)y + an−1 (x)y + ... + a1 (x)y + a0 (x)y = 0,
′
(⋆⋆)

Recall that the general solution to ⋆⋆ on (a, b) is the set of all solutions to ⋆⋆ on (a,b).
The crucial property of homogeneous linear ODE’s is that their general solution is always
determined by a ‘Fundamental Solution Set’. That is the general solution is always of the
form
y = c1 y1 (x) + ... + cn yn (x),
for some set of functions {y1 , ..., yn }. For instance in Example 2.1.5 above, we saw that the
′′
general solution to y + y = 0 was y = c1 cos(x) + c2 sin(x).
There are many different (but equivalent) ways to define what a fundamental solution
set is. The simplest, and easiest to check, is in terms of the Wronskian:

Definition 2.1.10. Suppose {y1 ,..., yn } is a set of solutions on (a, b) to the n-th order linear
homogeneous ODE ⋆⋆, which is nonsingular on (a,b). Then a fundamental solution set on
(a,b) is a set of solutions {y1, ..., yn} to ⋆⋆ such that the Wronskian of {y1 , ..., yn } is never zero
on (a, b).

Example 2.1.11. {sin(x), cos(x)} is a fundamental solution set to y + y = 0 on (−∞, ∞).

′′

solution: We’ve already seen that y1 = cos(x) and y2 = sin(x) are solutions to y ′′ + y = 0.
Their Wronskian is

W (x) = = (cos(x)) + (sin(x)) = 1.

cos(x) sin(x) 2 2
−sin(x) cos(x)

Since W (x) = 1 is never 0 for −∞ < x < ∞, we have that {cos(x), sin(x)} is a fundamental
solution set to y + y − 0 on (−∞,∞).
′′

Example 2.1.12. {e ,e } is a fundamental solution set to y − 2y − 3y = 0 on (−∞, ∞).

−x 3x ′′ ′

−x 3x
solution: First we check that e and e are in fact solutions:

(e ) − 2(e ) − 3(e )=e

−x ′′ −x ′ −x −x −x −x
+ 2e − 3e =0 ✓

(e ) − 2(e ) − 3(e ) = 9e
3x ′′ 3x ′ 3x 3x 3x 3x
− 6e − 3e =0 ✓
The Wronskian is then

e −x e3x
W (x) =
−x 3x 3x −x 2x
−x 3x = 3e e + e e = 4e .
−e 3e

The exponential function is never equal to zero. So W (x) = 4e is never 0 on (−∞, ∞)

2x

and we have that {e ,e } is a fundamental solution set to y −2y −3y = 0 on (−∞, ∞).
−x 3x ′′ ′

Example 2.1.13. {x, x , x } is a fundamental solution set to x y −2x y − 5xy + 5y = 0 on

5 −1 3 ′′′ 2 ′′ ′

(0, ∞).

C. Sprouse —Math 280 Notes Page 118 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 119

solution: Note that a3 (x) = x is zero at x = 0. So this ODE is singular on any interval
3

that contains the point x = 0, but nonsingular on (0, ∞). Next we need to check that
{x, x ,x } are in fact solutions to this ODE:
5 −1

x (x) − 2x (x) − 5x(x) + 5(x) = −5x + 5x = 0

3 ′′′ 2 ′′ ′
✓
x (x ) − 2x (x ) − 5x(x ) + 5y = 60x − 40x − 25x + 20x = 0
3 5 ′′′ 2 5 ′′ 5 ′ 5 5 5 5
✓

x (x ) − 2x (x ) − 5x(x ) + 5(x ) = −6x

3 −1 ′′′ 2 −1 ′′ −1 ′ −1 −1 −1 −1 −1
− 4x + 5x + 5x =0 ✓
Since we have a set of solutions to an ODE that is nonsingular on (0, ∞), we just need to
verify that their Wronskian is never zero on (0, ∞). But we have already shown above
that W (x) = 48x , which is only zero at x = 0, and therefore nonzero on (0, ∞).
2

Therefore {x, x ,x } is a Fundamental Solution Set to the ODE on (0, ∞). □.

5 −1

Theorem 2.1.14. Suppose {y1, ..., yn} is a set of solutions on (a, b) to the n-th order
linear homogeneous ODE ⋆⋆, which is nonsingular on (a, b). Then the following are
equivalent:

A) {y1 , ..., yn } is a fundamental solution set. That is, their Wronskian W (x) is
never zero on (a,b).

B) For any solution y to ⋆⋆, there is a unique set of constants c1 , ..., cn such that
y = c1 y1 (x) + ... + cn yn (x).

Proof. If {y1 , ..., yn } is any set of solutions to ⋆⋆ on (a, b) then it is easy to check that
z = c1 y1 + ... + cn yn is also a solution, by plugging in. For any x0 in (a, b), Fact 2.1.8 shows
that we can solve for z with the same initial values as any other solution y to ⋆⋆ if and
only if W (x0 ) ≠ 0, where in this case we have

⎡ ⎤ ⎡
⎢ y1(x0 ) yn (x0 ) ⎤ ⎥ ⎡
⎢ y(x0 ) ⎤ ⎥
⎢ ⎥
−1
⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥
...
⎢ ⎥ ⎢ ⎥
c
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
1
⎢
⎢ ⎥
⎥ ⎢
⎢
⋮ ⋮ ⎥
⎥ ⎢
⎢
⋮ ⎥
⎥
⎢ ⎥
⋮ =
⎢ ⎥ ⎢ ⎥
.
⎢ ⎥
⎣cn ⎦ ⎢
(n−1)
(x )
(n−1)
(x ) ⎥ ⎢ (n−1)
(x ) ⎥
⎣ 1
y 0 ... y n 0 ⎦ ⎣ y 0 ⎦

Finally, since z has the same initial values as y, the Fundamental Theorem of Linear
IVP’s gives that z = y, so
y = c1 y1 (x) + ... + cnyn (x).
So if {y1 , ..., yn } is a fundamental solution set, we can pick any x0 in (a,b), and since
W (x0 ) ≠ 0 we have by the above that condition B) holds. Conversely, if {y1 , ..., yn } is any
set of functions such that condition B) holds, then for any x0 in (a,b) the above implies
that W (x0 ) ≠ 0. So the Wronskian of {y1 , ..., yn } must be nonzero on all of (a,b) and hence
{y1, ..., yn} is a fundamental solution set.

So if {y1 , ..., yn } is a fundamental solution set on (a, b), then any solution y to ⋆⋆ on
(a, b) is of the form y = c1y(x) + ... + cn yn (x). This immediately gives:

C. Sprouse —Math 280 Notes Page 119 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 120

Corollary 2.1.15. Suppose {y1 , ..., yn } is a fundamental solution set to ⋆⋆, which is
nonsingular on (a, b). Then the general solution to ⋆⋆ on (a, b) is of the form

y = c1 y1 (x) + ...cn yn (x).

So to completely solve any homogeneous ODE ⋆⋆, we first find a set of solutions
{y1 , ..., yn }. Then if we can verify that {y1 , ..., yn } is a fundamental solution set by using the
Wronskian, we will know that all solutions to ⋆⋆ are of the form y = c1 y1 (x)+...+cn yn (x).
We already know a few fundamental solution sets from Examples 2.1.11, 2.1.12, and
2.1.13. From these we get:

Example 2.1.16. The general solution to y + y = 0 on (−∞, ∞) is

′′

y = c1 cos(x) + c2 sin(x).

Example 2.1.17. The general solution to y − 2y − 3y = 0 on (−∞,∞) is

′′ ′

−x 3x
y = c1 e + c2 e .

Example 2.1.18. The general solution to x y − 2x y − 5xy + 5y = 0 on (0, ∞) is

3 ′′′ 2 ′′ ′

5 −1
y = c1 x + c2 x + c3 x .

Finally, we still need to show that all n-th order homogeneous ODE’s ⋆⋆ do in fact
work this way. That is, we need to show that every ODE ⋆⋆ which is nonsingular on (a, b)
actually has a fundamental solution set. The easiest way to construct such a fundamental
solution set is to simply find a set of solutions {y1, ..., yn } to ⋆⋆ such that their wronskian
is nonzero at a single point x0 in (a,b). Then the proof of Theorem 2.1.14 shows that part
B) of Theorem 2.1.14 is satisfied, and hence W (x) is nonzero for all x in (a,b). That is, we
have:

Corollary 2.1.19. Let {y1 , ..., yn } be any set of solutions to ⋆⋆, which is nonsingular
on (a, b). Suppose also that their Wronskian is nonzero for some x0 in (a, b), W (x0 ) ≠
0. Then {y1, ..., yn } is a Fundamental Solution Set to ⋆⋆ on (a, b).

From this it is easy to use the Fundamental Theorem of Linear IVP’s to produce a
fundamental solution set on any interval where ⋆⋆ is nonsingular. The fact that any n-th
order linear homogeneous ODE always has a fundamental solution set shows that their
solutions are always completely characterized by such a set.

C. Sprouse —Math 280 Notes Page 120 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 121

Theorem 2.1.20 (Fundamental Theorem of Linear ODE’s). Consider the n-th order
Linear Homogeneous ODE
(n) (n−1)
an (x)y + an−1(x)y + ... + a1 (x)y + a0(x)y = 0, ,
′
(⋆⋆)

which is nonsingular on (a, b): ai (x) are continuous and an (x) is never zero on (a, b).
Then there is always a fundamental solution set {y1 ,..., yn } to ⋆⋆ on (a, b). Hence (by
Corollary 2.1.15), there is always a set of functions {y1 , ..., yn } such that the general
solution to ⋆⋆ is of the form

y = c1 y1(x) + ... + cn yn (x).

Proof. Let x0 be any point in (a,b), where ⋆⋆ is nonsingular on (a,b). Then from the
Fundamental Theorem of Linear IVP’s (2.1.3), we have solutions {ŷ1 , ..., ŷn } to ⋆⋆ such
that

ŷ1(x0 ) = 1 ŷ2 (x0 ) = 0 ŷn (x0 ) = 0

ŷ1(x0 ) ŷ2 (x0 ) ŷn (x0 )
′ ′ ′
= 0 = 1 = 0
⋮ , ⋮ , ⋯ , ⋮
(n−1) (n−1) (n−1)
ŷ1 (x0 ) = 0 ŷ2 (x0) = 0 ŷn (x0 ) = 1
Their Wronskian at x0 is then

⎛⎡⎢ ŷ1 (x0 ) ŷn (x0 ) ⎤ ⎥ ⎞ 0 1 ⋯ 0

1 0 ⋯ 0
⎢
⎢ ⎥
⎥
...
W (x0 ) = det ⎜
⎜
⎜⎢
⎢
⎢
⎥
⎥
⎥
⎟
⎟
⎟=
⎢ ⎥
⋮ ⋮ = 1.
⎝⎢⎢ ⎥
(x0 )⎥
⎦⎠ 0 0 ⋯ 1
(n−1) (n−1)
(x0 ) ... ŷn
⋱
⎣ŷ1

Hence by Corollary 2.1.19, {ŷ1, ..., ŷn } is a Fundamental Solution Set to ⋆⋆ on (a,b).

The above Fundamental Solution Set is often very useful, so we will give it a special
name.
Definition 2.1.21. Suppose ⋆⋆ is an n-th order linear homogeneous ODE which
is nonsingular on (a,b). Then for any x0 ∈ (a, b), the Canonical Fundamental
Solution Set to ⋆⋆ at x0 is the fundamental solution set {ŷ1 , ..., ŷn } such that

ŷ1 (x0) = 1 ŷ2 (x0) = 0 ŷn (x0 ) = 0

ŷ1′ (x0) = 0 ŷ2′ (x0) = 1 ŷn′ (x0 ) = 0
, , ⋯ ,
⋮ ⋮ ⋮
(n−1) (n−1) (n−1)
ŷ1 (x0 ) = 0 ŷ2 (x0 ) = 0 ŷn (x0 ) = 1

Example 2.1.22. The canonical fundamental solution set to y +y = 0 at x0 = 0 is {cos(x), sin(x)}

′′

Proof. We already know that cos(x) and sin(x)

C. Sprouse —Math 280 Notes Page 121 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 122

Note that for a given ODE ⋆⋆, the fundamental solution set on (a, b) is never unique,
and there are many possible choices. The ‘simplest’ fundamental solution set is usually
not the same as the canonical fundamental solution set. Here is just one example.
′′ ′
Example 2.1.23. Find the canonical fundamental solution set to y − 2y − 3y = 0 at x0 = 0.

Proof. We know that {y1 , y2 } = {e , e } is a fundamental solution set to y −2y − 3y = 0.

−x 3x ′′ ′
′′ ′ −x 3x
Therefore any solution y to y − 2y − 3y = 0 is of the form y = c1 e + c2 e . So we can set

−x 3x
y1 = c11 e + c12e ,

−x 3x
y2 = c21 e + c21e .

Plugging in the initial conditions ŷ1 (0) = 1, ŷ1 (0) = 0 gives

′

3
c11 + c12 = 1 c11 = 4
⇒ 1
−c11 + 3c12 = 0 c12 = 4

Plugging in the initial conditions ŷ2 (0) = 0, ŷ2 (0) = 1 gives

′

c21 + c22 = 0 c11 = − 14

⇒ .
−c21 + 3c22 = 1 c22 = 14

Therefore the canonical fundamental solution set at x0 = 0 is

{ŷ1, ŷ2 } =
3 −x 1 3x 1 −x 1 3x
e + e ,− e + e .
4 4 4 4

Here are both fundamental solution sets, graphed for comparison.

3 3
y1 y2

2 2

1 1
t t
−1 1 −1 1
−1
−1

Figure 2.1: The fundamental solution set {e ,e }

−x 3x

C. Sprouse —Math 280 Notes Page 122 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 123

3 3
ŷ1 ŷ2
2 2

1 1
t t
−1 1 −1 1
−1 −1

Figure 2.2: The canonical fundamental solution set {ŷ1 , ŷ2} at x0 = 0.

Finally, an amazing thing about the Wronskian is that there is a way to compute it for
any linear homogeneous linear ODE, even without knowing a fundamental solution set.
It will turn out that in general finding explicit formulas for a fundamental solution set
can be extremely difficult. (For this reason, most of the rest of the notes will only concern
constant coefficient ODE’s, which are far more tractable.) But for the Wronskian, we have
the following relatively easy formula:

Theorem 2.1.24 (Abel’s Theorem). Consider the homogeneous n-th order

linear ODE ⋆⋆ which is nonsingular on (a, b). Then if {y1 ,..., y n } is any set
of solutions to ⋆⋆ on (a,b), their Wronskian is of the form
an−1 (x)
W (x) = Ce
−∫ an (x)
dx
.

More explicitly, we have

x a n−1 (s)
W (x) = W (x0 )e
− ∫x0 an (s)
ds

for any x0 ∈ (a, b).

The proof of this is that the Wronskian itself solves a simple first order ODE, which
we leave to the exercises. This formula is true for any set of solutions, whether or not
{y1, ..., yn} is a fundamental solution set. In fact, if {y1 , ..., yn } is not a fundamental solution
set, then there exists some x0 in (a, b) with W (x0 ) = 0. Then the above formula gives
x a n−1 (s)
W (x) = 0e
− ∫x 0 an (s)
ds
= 0.

So if {y1 , ..., yn} is a set of solutions that is not a fundamental solution set, W (x) ≡ −0 on
(a, b) (which we already know from Fact 2.1.19).
Example 2.1.25. Let {y1 ,y2 } be the canonical fundamental solution set on (−3, 3) for the ODE

(9 − x )y − xy + 5y = 0.
2 ′′ ′

That is, y1 satisfies the initial conditions y1(0) = 1,y1(0) = 0, and y2 satisfies the initial con-
′

ditions y2 (0) = 0, y2(0) = 1. Find the Wronskian of y1 , y2 .

′

C. Sprouse —Math 280 Notes Page 123 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 124

solution: By Abel’s formula we have

− 12 ln(9−x ) dx
=C√
∫ x
dx 2 1
W(x) = Ce 9−x 2 = Ce .
9 − x2

We can use the initial conditions on y1 ,y2 to get

y1 (0) y2(0)
W (0) =
1 0
y1 (0) y2(0)
′ ′ = = 1.
0 1

√
So C 9 − 02 = 1 ⇒ C = 3. So W (x) = √ 3 □.
9−x 2

Again, in the above example we have no explicit formula for y1 ,y2 . In fact, for this
ODE such a formula is not very nice or easy to come by. However, we do know that the
fundamental solution set exists, by the Fundamental Theorem of Linear ODE’s. In fact, a
fundamental solution set on (−3, 3) is graphed below:

3 3
y1 y2
2 2

1 1
t t
−3 −2 −1 1 2 3 −3 −2 −1 1 2 3
−1 −1
−2 −2

−3 −3

Figure 2.3: The fundamental solution set {y1 , y2 } for (9 − x )y − xy + 5y = 0.

2 ′′ ′

Here is one more example to show how exotic the fundamental solution set can be.

Example 2.1.26. For the ODE y + (9 + 8 cos(t))y = 0, the canonical fundamental solution
′′

set {ŷ1, ŷ2 } at x0 = 0 has Wronskian W (x) = 1.

The ODE can be thought of as describing the motion of a ball at the bottom of a
spring, where the top of the spring is attached to a device that oscillates vertically, as in
the picture below. It shouldn’t be too hard to believe that these solutions do not have
formulas that can be written in terms of regular calculus functions. These solutions are
called Mathieu functions.

C. Sprouse —Math 280 Notes Page 124 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 125

y1
2

1
t
2 4 6 8 10 12 14 16
−1

−2

y2
2

1
t
2 4 6 8 10 12 14 16
−1

−2

Figure 2.4: The fundamental solution set {y1 , y2 } for y + (9 + 8 cos(t))y = 0.

′′

R
ωt
h(t) = Rcos(ωt)

(t) = 0 + y(t)

2.1.2 Existence and Uniqueness: The importance of being nonsingular

In Fact 2.1.8 and Theorem 2.1.14 we saw how the Wronskian is connected to solutions to
initial value problems. Namely, we have

C. Sprouse —Math 280 Notes Page 125 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 126

Fact 2.1.27. Let {y1 (x), ..., yn (x)} be any set of solutions to an n-th order linear ho-
mogeneous ODE, with Wronskian W (x). Then W (x0 ) ≠ 0 if and only if for any
choice of constants b1 ,..., bn there is a unique choice of c1 , ..., cn such that

y = c1 y1 (x) + ... + cn yn (x)

is the unique solution that solves the initial value problem with

y(x0) = b1
y (x0)
′
= b2
.
⋮
(n−1)
y (x0 ) = bn

Recall also that the exact formula for the coefficients is given in the proof of Fact 2.1.8:

⎡ c1 ⎤ ⎡
⎢ y1 (x0) yn (x0 ) ⎤ ⎥
−1
⎡ b1 ⎤
⎢
⎢ ⎥
⎥ ⎢
⎢
...
⎥
⎥ ⎢
⎢ ⎥
⎥
⎢
⎢ ⎥ ⎢ y1 (x0) yn (x0 ) ⎥ ⎥ ⎢ b2 ⎥
⎢c2 ⎥⎥=⎢
⎢ ⎥ ⎢
⎢ ⎥
⎥
′ ′
⎢
⎢ ⎥
⎥ ⎢ ⎥ ⎢
⎢ ⎥
⎥
...
⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥
⎢ ⎥
,
⎢ ⎥
⎢ ⎥ ⎢ ⎥ ⎢
⎢ ⎥
⎥
⋮ ⋮ ⋮ ⋮
⎢ ⎢ (n−1) ⎥ ⎢b ⎥
⎣c n ⎥
⎦ ⎢ (x0 )⎥
(n−1)
⎣ y1 (x0 ) ... yn ⎦ ⎣ n⎦
The above inverse exists if and only if W (x0 ) ≠ 0. Again, this comes from the linear
−1
algebra fact that for an n × n matrix A, A exists if and only if det(A) ≠ 0. The vast
majority of the ODE’s we will consider here are 2nd order, and for this we only need the
−1
2 × 2 case, for which A has a simple and very useful formula:

Fact 2.1.28.
−1
a b 1 d −b 1 d −b
= =
c d det(A) −c a ad − bc −c a

y(0) = 3, y (0) = 5.
′′ ′ ′
Example 2.1.29. Solve the initial value problem y − 2y − 3y = 0,

solution: From Example 2.1.12, we have that the general solution to the ODE is
−x 3x ′ −x 3x
y = c1 e + c2 e . So y = −c1 e + 3c2 e , and
0 0
y(0) = 3 c1e + c2 e = 3 c1 + c2 = 3
y (0) = 5
′ ⇒ 0 0 ⇒
−c1 e + 3c2 e = 5 −c1 + 3c2 = 5
In matrix notation
−1
1 1 c1 3 c1 1 1 3
= ⇒ =
−1 3 c2 5 c2 −1 3 5

Then Fact 2.1.28 gives

c1 1 3 −1 3 1
= = .
c2 4 1 1 5 2

C. Sprouse —Math 280 Notes Page 126 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 127

So the unique solution to the IVP is

−x 3x
y =e + 2e .

To get a feel for why the emphasis on nonsingularity and W (x0 ) ≠ 0 is really necessary,
we need to look at examples where W (x0) = 0 at some point x0. We can do this by taking
an ODE which is nonsingular on (0, ∞), but then choosing x0 = 0.
Example 2.1.30. There are infinitely many solutions to the initial value problem
2 ′′ ′
x y − 3xy + 3y = 0,

y(0) = 0,
y (0) = 1.
′

solution: The general solution to the ODE on (0, ∞) is y = c1 x + c2 x . One can check
3
3
this by showing that x and x are both solutions by plugging in, and then showing that
W (x) = 2x , which is nonzero on (0,∞). But setting y = c1 x + c2 x and solving for the
3 3

initial conditions gives

y(0) = 0 ⇒ c1 (0) + c2 (0) = 0 ⇒ 0 = 0.

y (0) = 0 c1 (1) + c2 (0) = 1

′
⇒ ⇒ c1 = 1.
3
Since there are no restrictions on c2 , we have that y = x + Cx solves the IVP for any
C ∈ R. Therefore there are infinitely many solutions to the initial value problem.

1.5 y

0.5
x
−1.5 −1 −0.5 0.5 1 1.5
−0.5

−1

−1.5

Figure 2.5: Infinitely many solutions with y(0) = 0, y (0) = 1.

′

This shows that if W (x0 ) = 0, the solution to an IVP with initial values at x0 may not
be unique. A slight change in initial values gives an example where the solution does not
exist:

C. Sprouse —Math 280 Notes Page 127 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 128

Example 2.1.31. There are no solutions to the IVP

2 ′′ ′
x y − 3xy + 3y = 0,

y(0) = 1,

y (0) = 0.
′

solution: If {y1 , y2 } solve x y − 3xy + 3y = 0 on (0, ∞), then we know from the
2 ′′ ′

previous example that {x, x3 } is a fundamental solution set on (0, ∞) and so any solution
2 ′′ ′
y to x y − 3xy + 3y = 0 must be of the form
3
y = c 1 x + c2 x

on the interval (0, ∞). But as a solution to a 2nd order ODE we have that y is
differentiable, and thus continuous at x = 0. So

y(0) = lim+ c1 x + c2 x = c1 (0) + c2 (0) = 0.

3
x→0

This means it is not possible to solve the initial condition y(0) = 1.

The solution above is unfortunately technical, in that we had to take limits rather than
3 2 ′′ ′
just plugging x = 0 into y = c1 x+c2 x . The reason for this is that x y −3xy +3y = 0 is only
nonsingular on (0, ∞) and not (−∞, ∞) so the general solution to the ODE on (−∞, ∞)
3
is not of the form y = c1 x + c2 x ! For instance the piecewise-defined function

3
x x>0
y(x) =
0 x≤0

solves x y −3xy + 3y = 0 on (−∞, ∞) and is not of the form y = c1 x + c2 x . This function

2 ′′ ′ 3

also shows that x3 is not the unique solution to the ODE on (−∞, ∞) which satisfies the
initial conditions y(0) = 0, y (0) = 0. There are actually infinitely many such solutions, of
′

the form
3
Cx x > 0
y(x) = 3 ,
Dx x ≤ 0
for any two numbers C, D ∈ R!

2.1.3 Linear Independence and Fundamental Solution Sets ∗

In subsection 2.1.1, we defined what a ‘fundamental solution set’ is in terms of the Wron-
skian. There is another common definition of fundamental solution set to ⋆⋆, which is
simply a set of n solutions that is ‘linearly independent’, which we define below. These
definitions turn out to be equivalent, but the equivalence is not at all obvious. In particu-
lar there is a major difference between a linearly independent set of functions, and a lin-
early independent set of functions which are also solutions to the ODE ⋆⋆. In this section
we will show that if {y1 , ..., yn } is a linearly independent set of functions that is also a set
of solutions to ⋆⋆, then the general solution to ⋆⋆ must be of the form y = c1y1 + ...+ cn yn ,

C. Sprouse —Math 280 Notes Page 128 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 129

and {y1 ,..., yn } is in fact a fundamental solution set in the sense of 2.1.1.

Definition 2.1.32. {y1 (x), ..., yn (x)} is called linearly independent on (a, b) if and
only if the only solution to

c1y1 (x) + ... + cn yn(x) ≡ 0 on (a, b)

is c1 , ..., cn = 0.

Example 2.1.33. Show {cos(x), sin(x)} is linearly independent on (−∞, ∞).

solution: Write
c1 cos(x) + c2 sin(x) ≡ 0.
Since the left hand side is = 0 for all x, we can plug in different values of x and use that
the equation holds.

x=0∶ c1 cos(0) + c2 sin(0) = 0 ⇒ c1 (1) + c2 (0) = 0 ⇒ c1 = 0,

⇒ c1 (0) + c2(1) = 0
π π π
x= ∶ c1 cos + c2 sin =0 ⇒ c2 = 0.
2 2 2
So from c1 cos(x) + c2 sin(x) ≡ 0 we derived that we must have c1 = c2 = 0. So
{cos(x), sin(x)} is linearly independent □.

Note that if one doesn’t mention an interval (a, b) on which {y1 ,..., yn } is independent,
then it is standard to assume the interval is (−∞, ∞).

Definition 2.1.34. If {y1, ..., yn} is not linearly independent, then {y1 ,..., yn } is said to be lin-
early dependent.. That is, {y1, ..., yn } is linearly dependent if there exist a set of constants
c1 , ..., cn not all zero such that

c1 y1(x) + ... + cn yn (x) ≡ 0.

Example 2.1.35.
{cos (x), sin (x),1}
2 2

is linearly dependent.

solution: Because of the well-known identity sin (x) + cos (x) = 1, we can write
2 2

(1) sin (x) + (1) cos (x) + (−1)1 = 0,

2 2

which gives the nonzero solutions c1 = 1,c2 = 1,c3 = −1.

Unfortunately it is not as easy to check that a set of functions is linearly dependent as

it is to check independence. In particular, we cannot plug in points to check dependence.
That is, to show
c1 y1(x) + ... + cn yn (x) ≡ 0,

C. Sprouse —Math 280 Notes Page 129 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 130

we must show that it is true for all x. So finding c1 , ..., cn that solve the equations for a
finite number of x values would not prove it is true for all other possible x values. If one
already knows a relationship between the functions y1, ..., yn, then {y1, ..., yn } can readily
seen to be dependent. Trigonometric identities such as the above example give some
common instances of this. As another example, recall the half-angle formula often used
in calculus:
cos (x) = (1 + cos(2x)) .
2 1
2
This can be rearranged to form 2 cos (x)−cos(2x)−1 = 0, which shows that {cos (x), cos(2x),1}
2 2

is linearly dependent.
For certain sets of functions like polynomials we can test for independence by rear-
ranging terms rather than plugging in points. In this case we can check for dependence
as well as independence by same method.
Example 2.1.36. Show that the set of polynomials

{x − 1, x + 2x, 3x + 1}
2 2

is linearly independent.

solution: Write
c1 (x − 1) + c2(x + 2x) + c3 (3x + 1) ≡ 0
2 2

. Rearranging,
(c1 + c2 )x + (2c2 + 3c3 )x + (−c1 + c3 )1 = 0.
2

Setting each term equal to zero gives the linear system

c1 + c2 = 0
2c2 + 3c3 = 0.
−c1 + c3 = 0
In general this sort of equation be solved by Gaussian elimination, but substitution is
quite easy here. The first and third equations give c2 = c1 and c3 = −c1 . So plugging these
into the second equation gives −c1 = 0, and hence c2 , c3 are also zero. Since the only
solution is c1 = c2 = c3 = 0, we have that {x −1, x +2x, 3x + 1} is linearly independent □.
2 2

Example 2.1.37. Show that the polynomials are linearly dependent.

{x, x + 1, 2x − 1}.

solution: Write
c1 x + c2 (x + 1) + c3 (2x − 1) ≡ 0
. Rearranging,
(c1 + c2 + 2c3 )x + (c2 − c3 ) = 0.
This means c2 − c3 = 0, so c2 = c3 . But we also have c1 + c2 + 2c3, which then gives
c1 + 3c3 = 0, or c1 = −3c3 . So if c3 is any number s, this gives the solution

c1 = −3s , c 2 = s , c 3 = s.

For instance if s = 1 this is a solution where not all ci ’s are zero. So {x, x + 1, 2x − 1} is
linearly dependent □.

C. Sprouse —Math 280 Notes Page 130 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 131

In the examples above, we used c1 y1(x) + ... + cn yn (x) ≡ 0 to derive a new system of
simultaneous equations that can be used to solve directly for the ci ’s. This requires n
equations, as there are n unknowns, c1 ,...,c n . But we can also get a system for c1 , ..., cn by
differentiating the equation c1 y1 (x) + ... + cn yn (x) ≡ 0:

c1 y1(x) + ... + cn yn (x) ≡ 0

c1 y1(x) + ... + cn yn (x) ≡ 0
′ ′
.
⋮
(n−1) (n−1)
c1y1 (x) + ... + cn yn (x) ≡ 0
Now for any fixed number x0 this becomes
c1 y1 (x0 ) + ... + cn yn (x0 ) = 0
c1 y1′ (x0 ) + ... + cn yn′ (x0 ) = 0
.
⋮
(n−1) (n−1)
c1 y1 (x0 ) + ... + cn yn (x0) = 0
Note this is very similar to what we’ve already seen leads to the Wronskian in Facts 2.1.8
and 2.1.27. In matrix form it becomes
⎡
⎢ y1 (x0 ) yn(x0 ) ⎤ ⎥ ⎡ c1 ⎤ ⎡0⎤
⎢
⎢
⋯
⎥
⎥ ⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥
⎢
⎢ (x ) (x ) ⎥
⎥ ⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
′ ′
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⋯
⎢ ⎥ ⎢ ⎥
y y 0
⎢ ⎥
c
⎢ ⎥ ⎢ ⎥
1 0 n 0
⎢ ⎥
2
⎢ ⎥ ⎢ ⎥
=
⎢ ⎥
.
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⋮ ⋮
⎥⎢⎢ ⎥ ⎥ ⎢ ⎢ ⎥
⋮ ⋮
⎢
⎢ ⎥
(x0 )⎥
(n−1) (n−1)
⎣y1 (x0 ) ⋯ yn ⎦ ⎣cn⎦ ⎣0⎦
Recall again from linear algebra that system has the unique solution c1 = ... = cn = 0
if and only if the above matrix is invertible, which is equivalent to it having a nonzero
determinant. So if
y1 (x0 ) ⋯ yn (x0 )
y1′ (x0 ) ⋯ yn′ (x0 )
≠ 0,
⋮ ⋮
(n−1) (n−1)
y1 (x0 ) ⋯ yn (x0 )
then all ci′ s are zero, and {y1 , ..., yn } is linearly independent. That is, we have

Fact 2.1.38. Let {y1, ..., yn} be any set of functions on (a, b). If their Wronskian W (x)
satisfies W (x0) ≠ 0 for some x0 in (a,b), then {y1 , ..., yn } is linearly independent on
(a,b).

Example 2.1.39. cos(x) and sin(x) are linearly independent on (−∞,∞).

solution: As we have already seen,

W (x) = = (cos(x)) + (sin(x)) = 1

cos(x) sin(x) 2 2
− sin(x) cos(x)
So W (x0 ) = 1 ≠ 0 for any x0 in (−∞,∞), and thus cos(x) and sin(x) are linearly
independent □.

C. Sprouse —Math 280 Notes Page 131 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 132

Example 2.1.40. Show that the polynomials are linearly independent.

{x − x , x , x − 1}
3 2 2

solution:
3 2 2
x −x x x−1
W (x) 1 = (x − 1) 6x − 2
2 3 2 2
2 3x − 2x 2x x −x x
= 3x − 2x 2x −1
2 6x − 2 2
6x − 2 2 0
= (x − 1)(2(3x − 2x) − 2x(6x − 2)) − 1(2(x − x ) − x (6x − 2))
2 3 2 2

= −2x (x − 3).
2

Since W (x) = −2x (x − 3) is not zero for all x (for instance W (1) = 4), we have that
2

{x − x , x , x − 1} is linearly independent. Note that there are some values here at which
3 2 2

the Wronskian is zero (x = 0, 3). But we only need W (x0 ) ≠ 0 for a single x0 to get linearly
independence □.

It is not as easy to show linear dependence. In particular, the converse of Fact 2.1.38
is not true. That is, W (x) ≡ 0 does not imply that {y1 , ...yn} are linearly dependent. It
is certainly true that any linearly dependent set of functions on (a, b) gives W (x) ≡ 0.
(Try W (x) for {x + 2, 2x + 4}, for instance). But there are also some linearly independent
functions that can also give W (x) ≡ 0. Here is the standard example:

Example 2.1.41. x and x∣x∣ are linearly independent on (−∞, ∞) but their Wronskian sat-
2

isfies W (x) ≡ 0 on (−∞, ∞).

solution: Note that even though ∣x∣ by itself is not differentiable at 0, the function
f (x) = x∣x∣ is actually differentiable at 0, with derivative f (0) = 0 (exercise). To compute
′

the Wronskian, one needs to do it in 3 cases. First,

W (0) =
0 0
= 0.
0 0

Then for x > 0,

W (x) =
2 2
x x 3 3
= 2x − 2x = 0.
2x 2x
And for x < 0,
2 2
W (x) =
x −x 3 3
= −2x + 2x = 0.
2x −2x
Therefore W (x) ≡ 0 on R □.

However, these two functions cannot both be solutions to the same nonsingular 2-nd
order ODE ⋆⋆ on (−∞, ∞). We will show below that any set of functions {y1 , ..., yn } that
is also a set of solutions to ⋆⋆ which is nonsingular on (a,b) must in fact be a fundamen-
tal solution set, which means that such a linearly independent set of functions must have
nonzero Wronskian. The proof of this relies on the fact below, which is typically why
linear independence is important in practice.

C. Sprouse —Math 280 Notes Page 132 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 133

Fact 2.1.42. A set of functions {y1 , ..., yn } is linearly independent if and only if for
any function y with y = c1 y1 (x) + ... + cn yn (x), the set of constants c1 ,...,c n must be
unique.

Uniqueness here means the following: If y can be written both as

y = c1 y1 (x) + ... + cnyn (x),
and
y = d1 y1 (x) + ... + dnyn (x),
then
c1 = d1 , ... , cn = dn .
That is, the coefficients must be the same in order to sum to the same function y.
Proof. Let {y1, ..., yn} be linearly independent. Then subtracting the two equations above
gives
(c1 − d1 )y1 (x) + ... + (cn − dn )yn (x) = y − y ≡ 0.
Then by definition of linear independence we have c1 − d1 = 0, ... cn −dn = 0, which means
that c1 = d1 , ... ,cn = dn , as we want to show.
Conversely, suppose {y1 ,...,y n } is any set of functions and the set of constants such
that y = c1 y1 (x) + ... + cn yn (x) is unique for any y. Then setting y(x) ≡ 0 gives
c1 y1 (x) + ... + cn yn (x) ≡ 0
But c1 = 0,...,cn = 0 is clearly a solution to this. Therefore by uniqueness it must be the
only solution, and hence {y1 , ..., yn } must be linearly independent.
We will now show that if {y1 , ..., yn } is any set of solutions to an n-th order homoge-
neous ODE ⋆⋆ which is nonsingular on (a,b), then if {y1 , ..., yn } is linearly independent
on (a,b), it will automatically be a fundamental solution set. This means that x and x∣x∣
2

can never both be solutions to the same 2nd order nonlinear ODE which is nonsingular
on (−∞, ∞). This is because they are linearly independent, but their Wronskian is zero
they can never form a fundamental solution set.

Theorem 2.1.43. Suppose {y1, ..., yn} is a set of solutions on (a, b) to the n-th order
linear homogeneous ODE ⋆⋆, which is nonsingular on (a, b). Then the following are
equivalent:

1) {y1 , ..., yn } is linearly independent on (a, b).

2) The general solution to ⋆⋆ on (a, b) is of the form y = c1 y1(x) + ... + cn yn (x).

3) {y1 , ..., yn } is a fundamental solution set to ⋆⋆ on (a, b).

C. Sprouse —Math 280 Notes Page 133 DRAFT August 26, 2022
2.1. General Theory: Homogeneous Linear n-th order ODE’s 134

Note that by Fact 2.1.42 above, Fact 2.1.14 shows that 1) implies both 2) and 3). What
needs to be shown is that for a set of solutions {y1 ,..., yn } to ⋆⋆ 2) implies 3), and 3) implies
2). Then Fact 2.1.14 again gives that both 2) and 3) together imply 1). We will omit the
1
proof, as it is somewhat lengthy without using some standard results in Linear Algebra.
Combining the results of Theorem 2.1.14, Fact 2.1.19, and Theorem 2.1.43, we now
have that the following conditions are all equivalent to being a fundamental solution set.
We have used ii) as our definition in subsection 2.1.1. However iii) is another very com-
mon definition, and iv) is also sometimes used.

Theorem 2.1.44. Suppose {y1 , ..., yn } is a set of solutions on (a, b) to the n-th order
linear homogeneous ODE ⋆⋆, which is nonsingular on (a, b). Then the following are
equivalent. So if any one of these conditions holds, then all of the conditions hold.
Therefore any one of these conditions could be used as the definition of fundamental
solution set:

i) W (x0 ) ≠ 0 for some x0 in (a, b).

ii) W (x) is nonzero for all x in (a, b).

iii) {y1 , ..., yn } is linearly independent on (a, b).

iv) The general solution to ⋆⋆ on (a, b) is of the form y = c1y1 (x) + ... + cn yn(x).

v) Any solution to ⋆⋆ can be written as y = c1 y1 (x) + ... + cn (x) for a unique set
of constants c1 ,..., cn .

1
The Linear Algebra proof the following: The set of solutions to ⋆⋆ is a vector space, which is of dimension
n, as the Fundamental Theorem of Linear ODE’s shows there is a basis of n vectors. In this case, any set of
exactly n linearly independent solutions must be a basis, and hence a fundamental solution set by Fact
2.1.14. Similarly, if {y1 ,..., yn } is a set of solutions such that the general solution to ⋆⋆ is of the form y =
c1 y1 + ...+ cn yn , then this is a set of exactly n solutions that spans the space of all solutions, and is hence also
a basis and again a fundamental solution set by Fact 2.1.14.

C. Sprouse —Math 280 Notes Page 134 DRAFT August 26, 2022
 2.1. General Theory: Homogeneous Linear n-th order ODE’s 135

Homework 2.1:
A) a) Show that each function is a solution to the ODE.
b) Show that given functions form a fundamental solution set on some interval (a, b).
c) Identify the largest such interval (a, b) which contains x0 .
d) Write the general solution to the ODE on that interval.
{x ,x },
2 ′′ ′ 2 3
1) x y − 4xy + 6y = 0, x0 = 2.
2 ′′ ′
2) x y + xy − y = 0, x, 1x , x0 = −1.
3) (1 − x )y + 2xy − 2y = 0, {x, x + 1},
2 ′′ ′ 2
x0 = 0.
4) (x + 2)y + (x + 1)y + y = 0, {x + 1, e },
′′ ′ −x
x0 = 0.
5) (x − 1)y − (x + 1)y + 2y = 0,
′′ ′
{x + 1, e
2 x−1
}, x0 = 0.
{cos(x ), sin(x )},
′′ ′ 3 2 2
6) xy − y + 4x y = 0, x0 = 1.
{x,cos(x), sin(x)},
′′′ ′′ ′
7) xy − y + xy − y = 0, x0 = π.
B) Use the given general solution to solve the initial value problems. Is there a unique solution, infinitely
many solutions, or no solution?
8) x 2 y ′′ − 4xy ′ + 6y = 0, {x 2 ,x3 }.

a) y(2) = 4, y ′ (2) = 0. b) y(0) = 0, y′ (0) = 1. c) y(0) = 0, y ′ (0) = 0.

9) x 2 y ′′ + 3xy ′ − 3y = 0, x, x3 .

a) y(0) = 1, y (0) = 0. b) y(0) = 0, y (0) = 1. c) y(0) = 0, y (0) = 0.

′ ′ ′

2 ′′ ′
10) x y + xy − y = 0, x, 1x .

a) y(0) = 1, y (0) = 0. b) y(0) = 0, y (0) = 1. c) y(0) = 0, y (0) = 0.

′ ′ ′

11) (1 − x )y + 2xy − 2y = 0, {x, x + 1}.

2 ′′ ′ 2

a) y(1) = 1, y (1) = 2. b) y(1) = 0, y (1) = 0. c) y(0) = 2, y (0) = −1.

′ ′ ′

12) (1 − x )y + 2xy − 2y = 0, {x, x + 1}.

2 ′′ ′ 2

a) y(0) = 1, y (0) = 2. b) y(−1) = 2, y (−1) = 1. c) y(−1) = 3, y (−1) = −3.

′ ′ ′

13) (x + 2)y + (x + 1)y + y = 0, {x + 1, e }.

′′ ′ −x

a) y(0) = 3, y (0) = −1. b) y(−2) = 1, y (−2) = −1. c) y(−2) = 1, y (−2) = 2.

′ ′ ′

14) (x − 1)y − (x + 1)y + 2y = 0, {x + 1, e }.

′′ ′ 2 x−1

a) y(1) = 2, y (1) = 0. b) y(1) = 0, y (1) = 0. c) y(0) = 0, y (0) = 2.

′ ′ ′

{cos(x ), sin(x )}.

′′ ′ 3 2 2
15) xy − y + 4x y = 0,

a) y(0) = 1, y (0) = 0. b) y(0) = 0, y (0) = 0. c) y(0) = 0, y (0) = 1.

′ ′ ′

{x, x ,x }.
3 ′′′ 2 ′′ ′ 5 −1
16) x y − 2x y − 5xy + 5y = 0,

C. Sprouse —Math 280 Notes Page 135 DRAFT August 26, 2022
 2.1. General Theory: Homogeneous Linear n-th order ODE’s 136

a) y(0) = 1, y ′ (0) = 0, y ′′ (0) = 0. b) y(0) = 0, y ′ (0) = 1, y′′(0) = 0. c) y(0) = 0, y′ (0) = 0, y ′′ (0) = 1.

{x,cos(x), sin(x)}.
′′′ ′′ ′
17) xy − y + xy − y = 0,

a) y(0) = 0, y ′ (0) = 0, y ′′ (0) = 1. b) y(0) = 0, y ′ (0) = 1,y ′′(0) = 0. c) y(0) = 1, y′ (0) = 0, y ′′ (0) = −1.

C) Is the set of functions linearly independent or linearly dependent? In the linearly dependent case, find
nonzero constants c1 , ..., cn such that c1 y1 + ... + cn yn = 0.
18.
{x − 3x, x + 3,x + 1}
2 2

solution:
2 2
x − 3x x + 3 x+1
W (x) = 2x − 3 2x 1
2 2 0
(x + 1)((4x − 6) − 4x) − ((2x − 6x) − (2x + 6)) = −6x − 6 + 6x + 6 = 0.
2 2
=

So we should look for c1, c2 , c3 with c1 (x − 3x) + c2 (x + 3) + c3 (x + 1) = 0..

2 2

c1 (x − 3x) + c2 (x + 3) + c3 (x + 1) = (c1 + c2 )x + (−3c1 + c3 )x + (3c2 + c3 ) = 0.

2 2 2

So
c1 + c2 = 0
−3c1 + c3 = 0.
3c2 + c3 = 0
If we use the first equation to get c1 = −c2 then the second and third equations both become
′
3c2 + c3 = 0. This has nonzero solution c2 = −1, c3 = 3. So a nonzero set of ci s is
c1 = 1,c2 = −1, c3 = 3 □.

19. 29.
{e ,e ,e } {x ,x + 2, 3x + 4}.
−x x 2x 2

20. 30.
{x, cos(x), sin(x)} {x ,x − 2, 4 − 2x}.
2

21. 31.
{sin (x), cos (x), 2} {x − 1,2x + 1, 3x}.
2 2

22. 32.
{cos(2x), sin(2x),1} {2x − 1,x − x, x + 2}
2 2

23. 33.
{sin(2x), sin(x) cos(x)} {x + 2,x + 1,x − 1}
2 2 2

24. 34.
{cos(2x),sin (x), cos (x)} {2x + 1,x − x,2x + 1}.
2 2 2 2

25. 35.
{sec (x), tan (x), 1} {x + x, x + 1,x + 1}.
2 2 2 2

26. 36.
{x , e ,e } {x − x, x − 1,x − 1}.
2 −x x 2 2

27. 37.
{e {x , x + 1, x − 1}.
−x x 3 2 2
,e , cosh(2x)}
28. 38.
{cosh(x), sinh(x), e } {(x + 1) , x + 1, x}.
−x 2 2

C. Sprouse —Math 280 Notes Page 136 DRAFT August 26, 2022
2.2. Reduction of Order 137

2.2 Reduction of Order

Suppose that you are given an n-th order linear homogeneous ODE. From the previous
section, we know that to get the general solution we need a fundamental solution set
which consists of n solutions, {y1 , ..., yn }. Suppose you know only one solution, y1 (x).
Then if we make the simple substitution y = uy1 , plugging into the ODE results in an
n − 1-st order linear ODE that can be solved to find u. In general this might just be a
curiosity. But when n = 2 this gives us a first order linear ODE, which we know that
we can solve by integrating factors, from Chapter 1. This means we can get the general
solution from only one solution.
So we will focus exclusively on the second order case now. Since we will have to plug
in the product y = uy1 , it is important to know the correct product rule. You are certainly
well aware from calculus that for any differentiable f ,g we have the product rule,

(f g ) = f g + f g .
′ ′ ′

But from now on, we will need to be just as familiar with the product rule for second
derivatives:

(f g ) = (f g + f g ) = f g + f g + f g + f g = f g + 2f g + f g .
′′ ′ ′ ′ ′′ ′ ′ ′ ′ ′′ ′′ ′ ′ ′′

That is, the product rules that we’ll now need are:

(f g ) = f g + f g .
′ ′ ′

(f g) = f g + 2f g + f g .
′′ ′′ ′ ′ ′′

Memorize this!! You will need it often from this point forward.

Example 2.2.1. Find (x sin(5x)) .

3 ′′

solution:

(x sin(5x)) = (x ) sin(5x) + 2(x ) (sin(5x)) + x (sin(5x))

3 ′′ 3 ′′ 3 ′ ′ 3 ′′

2 3
= 6x sin(5x) + 30x cos(5x) − 25x sin(5x) □ .

Now we’ll show how this applies to reduction of order.

3x ′′ ′
Example 2.2.2. Suppose you know that y1 = e is a solution to y − y − 6y = 0. Find the
general solution.

solution: We can see that

(e ) − (e ) − 6e
3x ′′ 3x ′ 3x 3x 3x 3x
= 9e − 3e − 6e = 0 ✓,
3x 3x
so y1 = e is in fact a solution to the ODE. Now let y = ue and plug into the ODE:

C. Sprouse —Math 280 Notes Page 137 DRAFT August 26, 2022
2.2. Reduction of Order 138

(ue ) − (ue ) − 6(ue ) = 0

3x ′′ 3x ′ 3x

− (u e + 3ue ) − 6ue
′′ 3x ′ 3x 3x ′ 3x 3x 3x
⇒ u e + 6u e + 9ue = 0.

✘✘✿0
✘
e u + (6e − e )u + (✘ − 3e − 6e )u = 0.
3x ′′ 3x 3x ′ 3x ✘✘ 3x ✘ 3x
⇒ 9e✘✘
3x ′′ 3x ′
⇒ e u + 5e u = 0.
But the above can be written as just
′′ ′
u + 5u = 0.
′
Let w = u . Then this is a first order equation for w that is easily solved by integrating
factors:
w + 5w = 0 ⇒ (e w) = 0 ⇒ w = C1 e .
′ 5x ′ −5x

Using this we solve for u ∶

′ −5x −5x 1 −5x

u = C1 e ⇒ u= C1 e dx = − C1 e + C2 .
5

Simplifying constants,
−5x
u = C1 e + C2 .
3x
This is not the final answer though! We need to now use y = ue to get the general
solution:
y = (C1 e + C2 )e = C1 e
−5x 3x −2x 3x
+ C2 e □ .

Everything worked because the u term cancels out above. The magic of reduction of
order is that this happens for any linear homogeneous ODE. So suppose y1 (x) is a solution
to
′′ ′
y + p(x)y + q(x)y = 0.
Then plugging in y = uy1 gives

(uy1 ) + p(uy1) + q(uy1 ) = 0,

′′ ′

u y1 + 2u y1 + uy1 + p(u y1 + uy1 ) + quy1 = 0,

′′ ′ ′ ′′ ′ ′
⇒
✿0
✘
✘
y1 u + (2y1 + py1)u + (✘ + qy1 )u = 0.
′ ✘
py✘1 ✘
′′ ′ ′ ′′
⇒ y1✘ +✘
′′ ′
Here we use that y1 solves y + py + qy = 0, so that the u-term always goes to zero! The
equation then becomes
′
′′ y ′
u + 2 y1 + p u = 0.
1

′
If we let w = u then the first order equation for w is both linear and separable. So you
can solve it any way you like. This gives us the following procedure, reduction of order:

C. Sprouse —Math 280 Notes Page 138 DRAFT August 26, 2022
2.2. Reduction of Order 139

✬ ✩

To find the general solution to y ′′ + p(x)y′ + q(x)y = 0, given a single

solution y1 (x):
1. Set y = uy1 (x). Plug into y′′ + py ′ + qy = 0 to get

y1 u + (2y1 + py1 )u = 0.
′′ ′ ′

′
2. Set w = u and solve the first order equation for w. (It is both linear
and separable.)
′
dw y1
+ 2 y + p w = 0.
dt 1

3. Integrate u = w to get u = ∫ w dx + C2 , so the general solution is

′

y = u(x)y1 (x) = w dx + C2 y1(x).

✫ ✪

Note that not only do we get the general solution, but also a fundamental solution set.
Our general solution is of the form

y = C1 y2 (x) + C2 y1(x).

It stands to reason that {y2 (x), y1 (x)} (or {y1 (x), y2(x)}) will be a fundamental solution
set, which can be verified with the Wronskian. For instance in the above example, we got
′′ ′
that the general solution to y − y − 6y = 0 was
−2x 3x
y = C 1e + C2 e .

The Wronskian of {e , e } is
−2x 3x

e −2x e 3x
W (x) = e − (−2)e e
−2x 3x 3x −2x x
−2x 3x = 3e = 5e ≠ 0.
−2e 3e

So {e , e } is a fundamental solution set to y − y − 6y = 0 on (−∞, ∞).

−2x 3x ′′ ′

2 ′′ ′ 9
Example 2.2.3. Find a fundamental solution set to x y − 6xy − 18y = 0. Use that y1 = x is
a solution.
9
solution: Again, we can quickly verify that x solves the ODE, as

x (x ) − 6x(x ) − 18x = x (72x ) − 6x(9x ) − 18x = (72 − 54 − 18)x = 0

2 9 ′′ 9 ′ 9 2 7 8 9 6
✓.

So we can plug y = ux 9 into the ODE:

x (ux ) − 6x(ux ) − 18(ux ) = 0

2 9 ′′ 9 ′ 9

x (u x + 18u x + 72ux ) − 6x(u x + 9ux ) − 18ux = 0

2 ′′ 9 ′ 8 7 ′ 9 8 5
⇒

C. Sprouse —Math 280 Notes Page 139 DRAFT August 26, 2022
2.2. Reduction of Order 140

✘✘✿ 09
x u + (18 − 6)x u + (✘ 54✘− 18)x u = 0.
9 ′′ 9 ′
⇒ −✘
72✘
So
7 ′′ 6 ′ ′′ 12 ′
x u + 12x u = 0 ⇒ u + x u = 0.
′
Letting w = u we solve
′ 12
w + x w = 0.
∫ ln(x )
12 12
12
The integrating factor is µ = e x dx = e = x , so

(x w) = 0
12 ′ 12 −12
⇒ x w = C1 ⇒ w = C1x .

Then
′ −12 −12 C1 −11 −11
u = C1 x ⇒u = C1 x dx = − x + C2 = C 1 x + C2 .
11
So the general solution is

y = (C1x + C2 )x = C1x
−11 9 −2 9
+ C2 x .

So we try {x ,x }. The Wronskian is

−2 9

−2 9
W (x) =
x x −2 8 −3 9 6
= 9x x + 2x x = 11x .
−2x−3 9x 8

This is nonzero on (0,∞) so {x , x } is linearly independent and hence a fundamental

−2 6

solution set on (0, ∞) □.

Example 2.2.4. Find a fundamental solution set to

2 ′′ ′
x y − xy + y = 0

on (0, ∞). Use that y1 (x) = x is a solution.

2 ′′ ′
solution: As with the previous example, the leading term of x y − xy + y = 0 is
a2 (x) = x , which is zero at x = 0. So we can only get a fundamental solution set on either
2

(−∞, 0) or (0, ∞). We plug y = uy1 = ux into the ODE:

x (ux) − x(ux) + (ux) = 0.

2 ′′ ′

x (u x + 2u (1) + u(0)) − x(xu + u(1)) + ux = 0.

2 ′′ ′ ′
⇒

x u + (2x − x )y + (✘ ✿
✘ 0
−x✘+✘
3 ′′ 2 2 ′ 3 ′′ 2 ′
⇒ x)u = 0 ⇒ x u + x u = 0.
′
So letting w = u gives
3 ′ 2 ′ 1
x w +x w = 0 ⇒ w + x w = 0.
1
dx ln(x)
The integrating factor is µ = e x =e = x, and

(xw) = 0
′ −1
⇒ xw = C1 ⇒ w = C1 x .

Then
−1
u= w dx = C1 x dx = C1 ln(x) + C2 ,

C. Sprouse —Math 280 Notes Page 140 DRAFT August 26, 2022
2.2. Reduction of Order 141

and the general solution is

y = uy1 = (C1 ln(x) + C2 )x = C1x ln(x) + C2x.

So {x, x ln(x)} is a good candidate for a fundamental solution set. The Wronskian is

W (x) =
x x ln(x)
= x ln(x) + x − x ln(x) = x.
1 ln(x) + 1

This is nonzero on (0, ∞) so {x, x ln(x)} is linearly independent and hence an

fundamental solution set on (0, ∞).

y1 (x)
y
y2 (x)

x
1 2 3 4 5

Example 2.2.5. Find a fundamental solution set to

xy + (x − 1)y − y = 0
′′ ′

on (0, ∞). Use that y1 = e

−x
is a solution.
−x
solution: Plugging in y = ue ,

) + (x − 1)(ue ) − (ue ) = 0,
−x ′′ −x ′ −x
x(ue

) + (x − 1)(u e ) − ue
′′ −x ′ −x −x ′ −x −x −x
⇒ x(u e − 2u e + ue − ue = 0.
✿ 0 −x
✘✘
u + (−2x + x − 1)e u + (x − (x ✘−
−x ′′ −x ′ ✘1)
⇒ xe ✘ ✘− 1)ue = 0.
✘
So
u − (x + 1)e u − (1 + x )u = 0.
−x ′′ −x ′ ′′ 1 ′
xe u =0 ⇒

The integrating factor for w − (1 + 1x )w = 0 is

′

− ∫ 1+ 1x dx )
−1
−x−ln(x) −x ln(x −1 −x
µ(x) = e =e =e e =x e .

Then
(x
−1 −x ′ −1 −x x
e w) = 0 ⇒ x e w = C1 ⇒ w = C1 xe .
So
C1xe dx = C1 (xe − e ) + C2 ,
x x x
u=

y = uy1 = (C1 (xe − e ) + C2 )e = C1 (x − 1) + C2 e

x x −x −x
.
So we have {y1 ,y2 } = {e , x − 1}. The Wronskian is W(x) = xe
−x −x
which is nonzero on
(0, ∞), so this is a fundamental solution set on (0, ∞).

C. Sprouse —Math 280 Notes Page 141 DRAFT August 26, 2022
2.2. Reduction of Order 142

y1 (x)
y
y2 (x)

x
−1 1 2

Except for ODE’s of a few special types, the ability to express the fundamental solution
set as a nice set of explicit functions {y1(x),y2 (x)} like the above is actually relatively rare.
But if there is at least one explicit solution y1(x), we can use reduction of order to express
y2 (x) in terms of definite integrals.

Example 2.2.6. Find the general solution to xy − (x + 2)y + 3y = 0 in terms of a definite

′′ ′
3
integral. Use that y1 = x is a solution.
3
solution: Plugging in y = ux ,

x(ux ) − (x + 2)(ux ) + 3(ux ) = 0,

3 ′′ 3 ′ 3

x(u x + 2u (3x ) + u(6x)) − (x + 2)(u x + u(3x )) + 3ux = 0,

′′ 3 ′ 2 ′ 3 2 3
⇒
x u + (6x − x − 2x )u + (6x − 3x − 6x + 3x )u = 0,
4 ′′ 3 4 3 ′ 2 3 2 3
⇒
x u + (−x + 4x )u = 0,
4 ′′ 4 3 ′
⇒

u + (−1 + x )u = 0.
′′ 4 ′
⇒

The integrating factor of w ′ + (−1 + 4x )w = 0 is

∫ −1+ 4x dx −x+4 ln(x) −x ln(x 4 ) 4 −x

µ(x) = e =e =e e =x e .

So x
(x e w) = 0
4 −x ′ 4 −x e
⇒ x e w = C1 w = C1 .
x4
Then
ex
u= wdx = C1
dx.
x4
The integral here cannot be evaluated with standard calculus functions, so we replace it
by a definite integral:
x es
u = C1 4
ds + C2 .
1 s
The lower endpoint could be any fixed value in (0, ∞), just not zero because the integral
becomes improper there. The general solution is then
x s
3 e 3
y = uy1 = C1 x ds + C2 x □ .
1 s4

C. Sprouse —Math 280 Notes Page 142 DRAFT August 26, 2022
 2.2. Reduction of Order 143

Homework 2.2:
A) For each question:
a) verify that y1 (x) is a solution.
b) Use reduction of order to find the general solution.
c) Find a fundamental solution set.
d) Find the Wronkskian, and list it’s zeroes and discontinuities. Verify that theWronskian is nonzero
and continuous on the given interval.
1.
(0, ∞).
2 ′′ ′ 5
x y − 11xy + 35y = 0, y1 = x .
solution: a)
x (x ) − 11x(x ) + 35x = x (20x ) − 11x(5x ) + 35x = 20x − 55x + 35x = 0.b) Plug
2 5 ′′ 5 ′ 5 2 3 4 5 5 5 5

y = ux into the ODE and use (f g ) = f g + 2f g + f g .

5 ′′ ′′ ′ ′ ′′

x (u x + 10u x + 20ux ) − 11x(u x + 5ux ) + 35x = 0.

2 ′′ 5 ′ 4 3 ′ 5 4 5

Collect terms. If done correctly, the u-term must always cancel.

(x )u + (10x − 11x )u + (20x − 55x + 35x )u = 0

7 ′′ 6 6 ′ 5 5 5 7 ′′ 6 ′
⇒ x u − x u = 0.
′
So if we let w = u then
7 ′ 6 ′ 1
x w −x w=0 ⇒ w − x w = 0.
The integrating factor is

−∫ )
1 −1
x dx − ln(x) ln(x −1
µ(x) = e =e =e =x .

So
(x
−1 ′ −1
w) = 0 ⇒x w =C ⇒ w = Cx,
and
C 2 2
u= w dx = C x dx = x + D = Cx + D.
2

y = y1 u = x (Cx + D)
5 2 7 5
⇒ y = Cx + Dx .

c) Set C = 0, D = 1 and C = 1, D = 0 above to get the solution set {x ,x }.

5 7
5 7
W (x) = 6 = 7x − 5x = 2x .We can see that W (x) is nonzero on (0, ∞)
x x 11 11 11
d) 4
5x 7x
□.

(−∞, ∞). y1 = x. (0, ∞).

′′ ′ 3x 2 ′′ ′
2. y − y − 6y = 0, y1 = e . 9. x y − xy + y = 0,
3. y − 10y + 25y = 0, y1 = e . (−∞, ∞). y1 = x . (0, ∞).
′′ ′ 5x 2 ′′ ′ 2
10. x y − 2xy + 2y = 0,
. (−∞, ∞). y1 = x . (0, ∞).
′′ ′ −x 2 ′′ ′ 3
4. y + 6y + 5y = 0, y1 = e 11. x y − 5xy + 9y = 0,

. (−∞,∞). y1 = x . (0, ∞).

2 ′′ ′ 5
′′ ′
5. y + 4y + 4y = 0, y1 = e
−2x 12. x y + xy − 25y = 0,
y1 = x. (0, ∞).
6. y + y = 0, y1 = sin(x). (−∞, ∞).
2 ′′
′′ 13. x y − 9xy + 9y = 0 = 0,
(0, ∞).
2 ′′ ′
14. 2x y + 3xy − y = 0, y1 = x,
7. y − x y − 2 y = 0, y1 = x . (0, ∞).
′′ 1 ′ 3 3
(0, ∞).
2 ′′ ′
x 15. 2x y − xy + y = 0, y1 = x,
(0,∞). y1 = sin(ln(x)). (0, ∞).
2 ′′ ′ 2 2 ′′ ′
8. x y − 6xy + 10y = 0, y1 = x . 16. x y + xy + y = 0,

C. Sprouse —Math 280 Notes Page 143 DRAFT August 26, 2022
 2.2. Reduction of Order 144

B∗) For each question:

a) verify that y1 (x) is a solution.
b) Use reduction of order to find the general solution.
c) Find a fundamental solution set.
d) Find the Wronskian, and list it’s zeroes and discontinuities. Verify thatthe Wronskian is nonzero
and continuous on the given interval.
17.
(x + 1)y + xy − y = 0, . (−1, ∞).
′′ ′ −x
y1 = e
solution: a) (x + 1)(e ) + x(e ) − e = xe
−x ′′ −x ′ −x −x −x −x −x −x
+e − xe −e = 0.b) Plug y = ue into
the ODE and use (f g ) = f g + 2f g + f g .
′′ ′′ ′ ′ ′′

u (x + 1)e − 2u (x + 1)e
′′ −x ′ −x −x −x ′ −x −x
+ u(x + 1)e − uxe + u xe −e = 0.

Collect terms.
(x + 1)e u + (x − 2(x + 1))e
−x ′′ −x ′
u = 0.

w = 0. (w = u )
′′ x+2 ′ ′ x +2 ′
u − u =0 ⇒ w−
x+1 x +1

= (x + 1) e
∫ − x+2 dx − ∫ 1+ x+1
1
dx −1 −x
The integrating factor is µ(x) = e x+1 =e .

((x + 1) (x + 1) e
−1 −x ′ −1 −x x
e w) = 0 ⇒ w=C ⇒ w = C(x + 1)e .

(x + 1)e dx = Cxe + D.
x x
u=
−x −x
y=e u ⇒ y = Cx + De .
c) Set C = 0,D = 1 and C = 1,D = 0 above to get the solution set {e
−x
, x}.d)
−x
W (x) = = e (1 + x).
e x −x
−x
−e 1

We can see that W (x) is nonzero on (−1, ∞) □.

18. x y − (x + 2x)y + (2 + x)y = 0, (0, ∞).
2 ′′ 2 ′
y1 = x
19. x y + (x − 4x)y + (6 − 2x)y = 0, y1 = x . (0, ∞).
2 ′′ 2 ′ 2

20. y − 4xy + (4x − 2)y = 0, y1 = e . (0, ∞).

2
′′ ′ 2 x

21. xy − y − (x + 1)y = 0, (0,∞).

′′ ′ −x
y1 = e .
22. xy + (4x − 1)y + (4x − 2)y = 0, . (0, ∞).
′′ ′ −2x
y=e
23. xy + (2x + 2)y + (x + 2)y = 0, (0, ∞).
′′ ′ −x
y=e .
24. (x − 1)y − xy + y = 0, y1 = e . (1, ∞).
′′ ′ x

′′ ′ π π
25. y + 3 tan(x)y − 2y = 0, − , . y1 = sin(x).
2 2
26. y + 2 tan(x)y + (sec (x) + tan (x))y = 0, y1 = cos(x).
′′ ′ 2 2 π π
− , .
2 2
27. y − tan(x)y + sec (x)y = 0, y1 = sec(x).
′′ ′ 2 π π
− , .
2 2

C. Sprouse —Math 280 Notes Page 144 DRAFT August 26, 2022
2.3. Constant Coefficient ODE’s: Fundamental Solution Sets 145

2.3 Constant Coefficient ODE’s: Fundamental Solution Sets

Definition 2.3.1. We say that a linear ODE is constant coefficient if it is of the form

(n) (n−1)
(x) + an−1 y (x) + ... + a1 y (x) + a0 y(x) = g(x)
′
y

where a0, ..., an−1 are constants.

If the variable is t for time, the constant-coefficient ODE

n
d y dy
n + ... + a1 + a0 y = g(t)
dt dt
is also called Linear Time Invariant (LTI).
But for now, we will use x as the variable. The importance of constant coefficients,
aside from being the most basic form for applications, is that we can completely describe
the general solution to the homogeneous case,

(n) (n−1)
(x) + an−1 y (x) + ... + a1 y (x) + a0 y(x) = 0.
′
y

That is, we can always find the fundamental solution set in an explicit way. This is in
fact extremely rare for general linear ODE’s, and almost exclusive to constant-coefficient
ODE’s and closely related equations, as a general class.
We already know the first order case from Example 1.2.1 of Chapter 1, where the first
dy
nontrivial ODE that we solved was d x + 3y = 0, or more generally of the form

′
y + ay = 0.

By now we know very well that the solution here is

−ax
y = Ce .

So next we will cover the second order case in detail. These are the most common
when it comes to applications, but also demonstrate all of the interesting features of what
happens in higher order.

2.3.1 Second Order

We now want to find the fundamental solution sets to
′′ ′
y + by + cy = 0,

where b and c are constants.

We can first try to look for a simple solution, keeping in mind the first order case. In
sx
particular, we’ll just substitute e into the equation and see what happens. Although this
might seem like a naive thing to try at first, the nice thing about it is that it’s easy to take
derivatives of e : (e ) = se , and (e ) = s e .
sx sx ′ sx sx ′′ 2 sx

C. Sprouse —Math 280 Notes Page 145 DRAFT August 26, 2022
2.3. Constant Coefficient ODE’s: Fundamental Solution Sets 146

′′ ′
Substituting this into y + by + cy = 0,

(e ) + (be ) + ce
sx ′′ sx ′ sx 2 sx sx sx
= s e + sbe + ce
= (s + bs + c)e
2 sx

= 0

So can we solve (s + bs + c)e

2 sx sx
≡ 0? Since e is never 0, we can divide by it to get
2 −sx
s + bs + c = 0e = 0.
2
Now notice that s + bs + c = 0 is a polynomial equation that doesn’t depend on x
anymore! We know from algebra that quadratic polynomials have two roots (although
the two roots might be the same, or they might be complex). And in fact, the quadratic
formula gives the solutions as
√
−b ± b2 − 4c
s 1 , s2 = .
2

This means that for s1 , s2 as above, e s1 x and es2 x solve y′′ + by′ + cy = 0.
′′ ′
So we have discovered that the solutions to y + by + cy = 0 are determined by the
2
roots of the polynomial s + bs + c. Therefore we give this polynomial a special name.
′′ ′
Definition 2.3.2. For the ODE y + by + cy = 0, its characteristic equation is given by
2
s + bs + c = 0.
′′ ′
For the general solution to y + by + cy = 0, the most obvious case is when the char-
acteristic polynomial has two real roots s1 , s2 that are distinct (s1 ≠ s2). This means that
s x s x
e 1 and e 2 are distinct real-valued functions. From Section ??, we know the general so-
lution if these functions are linearly independent. To check this, we can just check that
the Wronskian is nonzero. The Wronskian is

s x s x
W (x) =
s1 x s x
e1 e2 e e2
(e ) (e 2 )
s1 x ′ s x ′ = s 1x s x
s1 e s2e 2
s 1 x s2 x s1 x s 2 x
= s2 e e − s1 e e
(s1 +s2 )x
= (s2 − s1 )e .

And, remembering that we have assumed here that s1 ≠ s2 , we know that the above quan-
s x
tity is definitely nonzero. Therefore from Section generaltheorylinear we have that e 1
s2 x ′′ ′
and e are linearly independent, and all solutions to y + by + cy = 0 are of the form
s 1x s2 x
y = c1 e + c2 e .

Example 2.3.3. Find the general solution to

′′ ′
y − y − 6y = 0.

C. Sprouse —Math 280 Notes Page 146 DRAFT August 26, 2022
2.3. Constant Coefficient ODE’s: Fundamental Solution Sets 147

solution: This was Example 2.2.2 in the previous section. Now we have a much easier
way to solve it. The characteristic equation factors as

(s + 2)(s − 3) = 0
2
s −s−6 =0 ⇒ ⇒ s = −2, 3.

So the general solution is

−2x 3x
y = c1 e + c2 e □.

We next look at the case of repeated roots. The quadratic with double root s1 looks
like (s − s1 ) = 0. Multiplying out, we have that this is is s − 2s1 s + s1 . Therefore the ODE
2 2 2

corresponding to this situation is

′′ ′ 2
y − 2s1 y + s1 y = 0.
s x
Now, we already know that one solution to this problem is y = e 1 . We need to find
another solution. And conveniently we have a method for doing precisely that: the re-
duction of order method from the previous section.
s x s x
We have y1 = e 1 . We assume that y2 = ue 1 where u is a function of x. Then we have

y2 = (ue ) =ue
′ s1 x ′ ′ s1 x s1 x
+ s1 ue

y2 = ((ue ) ) = (u e )
′′ s1 x ′ ′ ′ s1 x s1 x ′ ′′ s1 x ′ s1 x ′ s1 x 2 s1 x
+ s1ue = u e + u s1 e + s1 u e + s1 ue
′′ s x ′ s x 2 s1 x
= u e 1 + 2u s1 e 1 + s1 ue

Now insert this into the equation y ′′ − 2s1y ′ + s12 y = 0 to get

− 2s1(u e ) + s1 e
′′ s1x ′ s 1x 2 s1 x ′ s 1x s1 x 2 s1 x
u e + 2u s1 e + s1ue + s1ue =0

This simplifies (a lot) to

′′ s1 x
u e = 0.
Now remember when solving ODE’s we always divide by whatever is beside the highest
order term. in this case one gets
′′ −s x
u = 0e 1 = 0
′′
Therefore we are left with the very simple equation u = 0 which has solution u(x) =
Cx + D. Now recall to get y2 we multiply this by y1 to get y2 (x) = Cxe 1 + De 1 .
s x s x
s1 x s1 x
We can also check that e and xe are linearly independent:
s s x
s x s x
e 1x xe 1 e1 xe 1 2s1 x 2s x 2s x 2s x
(e ) (xe )
s1 x ′ s1 x ′ = s1 x s1 x s x s x =e + s1 xe 1 − s1 xe 1 = e 1 ≠ 0.
s1 e e + s1 xe 1 xe 1

So {e , xe 1 } is the fundamental solution set and we have the general solution

s1 x s x

s1 x s1 x
y = c 1e + c2 xe .
′′ ′
Example 2.3.4. Find the general solution to y − 16y + 64y = 0.

C. Sprouse —Math 280 Notes Page 147 DRAFT August 26, 2022
2.3. Constant Coefficient ODE’s: Fundamental Solution Sets 148

solution: The characteristic equation factors as

(s − 8) = 0
2 2
s − 16s + 64 = 0 ⇒ ⇒ s = 8, 8.

So the general solution is

8x 8x
y = c1 e + c2 xe □.

2
Finally suppose the roots of the characteristic equation s + bs + c = 0 are complex.
Recall that the quadratic formula gives
√
b b2 − 4c
s 1 , s2 = − ± .
2 2
2 2
The roots are complex when 4c > b , or when c > 2b . This is a convenient way to
remember when a quadratic might have complex roots. This happens when c is positive
and big (meaning bigger than (b/2) ). It will be very important for you to be able to detect
2

when complex roots could happen with your ODE’s from now on, so this is worth writing
down one more time:

2
s + bs + c = 0 has complex roots
if and only if
b 2
c> .
2

The formula for the roots s1 ,s2 above shows they must then be of the form α ± βi,
(α+iβ)x (α−iβ )x
where α and β are real. Therefore we have a fundamental solution set {e ,e },
and the general solution is
(α+iβ )x (α−iβ)x
y = c1 e + c2 e .
However, this is now a complex-valued solution, which we do not want. When all quan-
tities in an ODE are real, we are generally looking for the real-valued solution. This is
especially true for the type of applications that we are interested in. You can’t have a
temperature or position of 5 − 3i!
This is important enough to be worth boxing.
′′ ′
For the ODE y + by + cy = 0 with real coefficients, an answer for the
general solution of the form
(α+βi)x (α−βi )x
y = c1 e + c2 e

is not correct, because it is the complex general solution.

For an ODE with real coefficients, it is always assumed that
any answer must be a real solution.

C. Sprouse —Math 280 Notes Page 148 DRAFT August 26, 2022
2.3. Constant Coefficient ODE’s: Fundamental Solution Sets 149

Now, it is always possible to derive the real general solution from the complex general
solution. But that is work that we must do! So we now use Euler’s formula to see what the
α+βi α−βi
complex solutions e and e really look like:
(α+βi )x
(cos(βx) + i sin(βx))
αx βix αx
e =e e =e

and
(α−βi )x
= e (cos(−βx) + i sin(−βx)) = e (cos(βx) − i sin(βx).
αx −βix αx αx
e =e e

This is interesting, because it shows that if you add e (cos(βx)+i sin(βx)) to e (cos(βx)−
αx αx

i sin(βx) then the imaginary terms cancel each other out!

Thus, if we let c1 = 12 and c2 = 12 we get

(α+βi)x (α−βi)x
e (cos(βx) + i sin(βx)) + e (cos(βx) − i sin(βx))
1 αx 1 αx
c1 e + c2 e =
2 2
αx
= e cosβx
′′ ′
which is a real solution to y + by + cy = 0. Now we need to find another real solution,
which is not simply a constant multiple of the one we just found. The trick is to remember
that for the complex general solution, the constants c1 ,c2 can also be complex! If c1 , c2
are purely imaginary, this will interchange the real and imaginary parts of the complex
general solution. So if we choose c1 = −i
2
and c2 = 2i then we can show the imaginary part
of the complex solution is another real solution:

(α+βi )x (α−βi )x
e (cos(βx) + i sin(βx)) + e (cos(βx) − i sin(βx))
−i αx i αx
c1 e + c2 e =
2 2
−i αx 1 αx i αx
= e cos(βx) + e sin βx + e cos(βx)
2 2 2
αx
= e sin(βx)
αx αx
We’ll leave it as an exercise to show that e sin(βx) and e cos(βx) are linearly inde-
pendent by taking the Wronskian.
So the real general solution now is
αx αx
y = c1 e cos(βx) + c2 e sin(βx).
′′ ′
Example 2.3.5. Find the general solution to y − 10y + 34y = 0.

solution: We see that c = 34 is positive, and pretty big. So we check

2 2
b −10
34 = c > = = 25 ✓.
2 2

This means that there are complex roots, which are

√
−b ± b2 − 4c 10 ± 100 − 136
s 1, s 2 = =
2 2
10 ± 6i
= = 5 ± 3i.
2

C. Sprouse —Math 280 Notes Page 149 DRAFT August 26, 2022
2.3. Constant Coefficient ODE’s: Fundamental Solution Sets 150

So the general solution is

5x 5x
y = c1 e cos(3x) + c2 e sin(3x).

(5+3i)x (5−3i)x
Warning 2.3.6. Once again, we emphasize that y = c1 e + c2 e is not correct for an
answer in the above example. It should always be understood that the final answer to a real
ODE must be real.
So, to summarize:
′′ ′
To find the general solution to y + by + cy = 0, let s1, s2 be the roots of
2
s + bs + c = 0. Then,

s1 x s x
s1 , s2 real, s1 ≠ s2 ⇒ y = c1 e + c2 e 2 .
s1 x s1 x
s1 , s2 real, s1 = s2 ⇒ y = c1 e + c2 xe .
αx αx
s1 , s2 complex, s1 , s2 = α ± iβ ⇒ y = c1e sin(βx) + c2 e cos(βx).

Warning 2.3.7. From now on, most of the ODE’s that you solve will be constant coefficient.
That’s because the above table makes them easy to explicitly solve. Just don’t assume that non-
constant coefficient equations work the same way! For instance here are some non-constant
coefficient ODE’s, where you won’t be able to naively apply the formulas in the box above to get
the general solution:
2 ′′ ′ 3
2x y − 5xy − 3y = 0 ⇒ y = c1 x + c2 x .

′′ ′ e −x ex
xy + 2y − xy = 0 y = c1 x + c2 x .
⇒
There are similar examples in the exercises from the last section. Or for even more, refer to [?]
or [?], where you can see that in general the story becomes quite a bit more complicated!
For a few more concrete examples, we can do an initial value problem in each case:
Example 2.3.8.
′′ ′
y + 4y + 3y = 0
y(0) = 1 .
y (0) = 0
′

The characteristic equation is s + 4s + 3 = 0 which factors as (s + 1)(s + 3) = 0. So the

2

roots are s = −1, −3. These are distinct real numbers so we have
−x −3x
y(x) = c1 e + c2 e .

To solve for c1 and c2 ,

y (x) = −c1 e
′ −x −3x
− 3c2 e .

C. Sprouse —Math 280 Notes Page 150 DRAFT August 26, 2022
2.3. Constant Coefficient ODE’s: Fundamental Solution Sets 151

So the initial conditions give

c1 + c2 = 1,
−c1 + 3c2 = 0.
3
Which are easily solved to get c1 = 2
and c2 = − 12 . So,

3 −x 1 −3x
y(x) = e − e .
2 2
Example 2.3.9.
′′ ′
y + 4y + 4y = 0
y(0) = 1 .
y (0) = 0
′

The characteristic equation is s + 4s +4 = 0, which factors as (s +2)(s + 2) = 0. So this

2

has a double root at s = −2. So we have

−2x −2x
y(x) = c1 e + c2 xe .

To solve for c1 and c2 ,

y (x) = −2c1 e
′ −2x −2x −2x
+ c2 e − 2c2xe .

So the initial conditions give

c1 = 1,
−2c1 + c2 = 0 ⇒ c2 = 2.
So the solution to the initial value problem is
−2x −2x
y(x) = e + 2xe .

Example 2.3.10.
′′ ′
y + 4y + 5y = 0
y(0) = 1 .
y (0) = 0
′

2
The characteristic equation s + 4s + 5 is not so easy to factor this time. So instead we
use the quadratic formula

−4 ± 42 − 4(1)(5) −4 ± −4 −4 ± 2i
s= = = = −2 ± i.
2(1) 2 2
So,
−2x −2x
y(x) = c1e cos(x) + c2 e sin(x).
And
y (x) = −2c1 e
′ −2x −2x −2x −2x
cos(x) − c1 e sin(x) − 2c2 e sin(x) + c2e cos(x),
which gives
c1 = 1

C. Sprouse —Math 280 Notes Page 151 DRAFT August 26, 2022
2.3. Constant Coefficient ODE’s: Fundamental Solution Sets 152

−2c1 + c2 = 0 ⇒ c2 = 2

So
−2x −2x
y(x) = e cos(x) + 2e sin(x).

For each example above, the real parts of both roots were negative. The resulting the
solution is then called damped. The cases above are is called overdamped, critically
damped, and underdamped, respectively. The graphs below only give a hint of where
these names come from, but we will return to this in Section 2.8, when we look at appli-
cations!
y
overdamped
critically damped
underdamped

x
1 2 3 4 5

Warning 2.3.11. You need to be careful when solving constant-coefficient ODE’s from now
on, because a slight change in the characteristic equation could give a dramatically different
solution. For instance, you should check that you get the following general solutions here:
′′ −x x
y −y = 0 ⇒ y = c 1 e + c2 e ,
′′
y +y = 0 ⇒ y = c1 cos(x) + c2 sin(x),
′′ ′ −x
y +y = 0 ⇒ y = c1 e + c2 .

It is a very common mistake in a class like this to confuse these three, which will lead to a
completely wrong answer!

2.3.2 Higher Order

Now let’s consider the higher order case

(n) (n−1) ′
y + an−1 y + ... + a1 y + a0 y = 0.

We can again derive the characteristic equation by looking for solutions of the form y =
sx sx
e , for constant s. Again, it is easy to take the derivatives of e and plug in to see what
happens.

C. Sprouse —Math 280 Notes Page 152 DRAFT August 26, 2022
2.3. Constant Coefficient ODE’s: Fundamental Solution Sets 153

sx (n) sx (n−1)
(e ) + an−1(e ) + ... + a1 (e ) + a0 (e ) = s e + an−1 s
sx ′ sx n sx n−1 sx sx sx
e + ... + a1se + a0 e
= (s + an−1s + ... + a1 s + a0 )e
n n−1 sx

So we need to solve (s + an−1s + ... + a1 s + a0 )e ≡ 0.

n n−1 sx

Like before, e is nonzero so we can divide by it to get (s + an−1s + ... + a1 s + a0 ) =

sx n n−1
−sx sx
0e = 0. So e is a solution to the ODE as long as s is a root of the characteristic equation.
That is, the characteristic equation is

n n−1
s + an−1 s + ... + a1s + a0 .

We know from algebra that n-th degree polynomials have n solutions (although some of
them might be repeated, and some might be complex). So denoting the n solutions by
sx (n) (n−1) ′
s1 , ..., sn , we have that e i is a solution to y + an−1 y + ... + a1 y + a0 y = 0 for all
i = 1, ..., n.
To find the general solution, we again first consider the easiest case, when all of the
solutions s1 , ..., sn are real and distinct (si ≠ sj for all i ≠ j). This means that e s1 , ..., esn are all
distinct real-valued functions. So if we can show that they are linearly independent, they
form a fundamental solution set. Again this just comes from the Wronskian, though it is
a larger determinant now:
s x s x s x s x
e1 ⋯ en e1 ⋯ en 1 ⋯ 1
(e ) (e )
s1 x ′ sn x ′ s x s x
W (x) =
⋯ s1 e 1 ⋯ s ne n s x s x s 1 ⋯ s n
= = e 1 ⋯e n .
⋮ ⋮ ⋮ ⋮ ⋮ ⋮
s1 x (n−1) sn x (n−1)
(e ) ⋯ (e )
n−1 s1x n−1 s x n−1 n−1
s1 e ⋯ sn e n s1 ⋯ sn

s x s x
The function e 1 ⋯e n is never 0, just because exponentials are never zero. so what must
be checked is that the determinant

1 ⋯ 1
s1 ⋯ sn
⋮ ⋮
n−1 n−1
s1 ⋯ sn

is nonzero. This looks like a pretty complicated thing to compute. Fortunately, it is a

famous determinant called the Vandermonde determinant, and its value is well-known
from linear algebra to be

1 ⋯ 1
= ∏(si − sj ).
s1 ⋯ sn
⋮ ⋮
n−1 n−1 i≠j
s1 ⋯ sn
We can leave the proof of this formula as a (hard) exercise. The point of the formula
is that it shows that the above determinant definitely not zero, because we have assumed

C. Sprouse —Math 280 Notes Page 153 DRAFT August 26, 2022
2.3. Constant Coefficient ODE’s: Fundamental Solution Sets 154

that si ≠ sj for i ≠ j. This is exactly the right condition to give that the terms in the product
are all nonzero: si − sj ≠ 0 for i ≠ j! Therefore

1 ⋯ 1
= ∏(si − sj ) ≠ 0.
s1 ⋯ sn
⋮ ⋮
n−1 n−1 i≠j
s1 ⋯ sn

So {e s1x , ..., esn x } is a fundamental solution set, and we have the general solution

s1 x sn x
y = c1 e + ... + cn e .

Example 2.3.12. The characteristic equation of

(5) (4) ′′′ ′′

y − 3y − 5y + 15y + 4y − 12 = 0

factors as

s − 3s − 5s + 15s + 4s − 12 = (s + 2)(s + 1)(s − 1)(s − 2)(s − 3).

5 4 3 2

The general solution is

−2x −x x 2x 3x
y = c1 e + c2e + c 3 e + c4 e + c5 e .

Note that the above polynomial would be quite difficult to factor by hand. There is
no formula for 5-th degree polynomials corresponding to the quadratic formula. So the
only way to factor the above would be to find some roots, by perhaps guessing, and then
reduce the order by polynomial division, and try to find more roots, until one reduces to
a quadratic which is finally solvable by the quadratic formula. (There is a cubic formula,
but even this is far too complicated to do by hand).

Remark 2.3.13. For practical problems, one is often not interested in exactly what the roots
are, but only whether the real parts are positive or negative. In this case, an algebraic method
called the Routh Stability Criterion can determine exactly how many roots have positive real
parts. This is an important technique for engineering applications, but saying more about
it here would take us too far afield. Other than this, the characteristic equation can only be
factored by hand if it is of a special form. Otherwise one has to rely on a computer to get the
roots of whatever the specific polynomial is.

Here is a specific example where polynomial division helps.

Example 2.3.14. Find the general solution to y ′′′ − 1000y = 0.

3
solution: The characteristic equation is s − 1000 = 0. This has the obvious root of
s1 = (1000) = 10. So we can do the long division
1
3

C. Sprouse —Math 280 Notes Page 154 DRAFT August 26, 2022
2.3. Constant Coefficient ODE’s: Fundamental Solution Sets 155

2
s + 10s + 100
3
s − 10 s − 1000
3 2
− s + 10s
2
10s
− 10s 2 + 100s
100s − 1000
− 100s + 1000
0
2
Then s + 10s + 100 has roots s2 , s3 = −5 ± 5 3 by the quadratic formula. So the general
solution is
10x −5x −5x
y = c1 e + c2 e cos(5 3 x) + c3 e sin(5 3 x) □ .

n
The n complex solutions to s −1 = 0 are called roots of unity. These can be used to get
the general solution to y (n) + ay = 0 for any real coefficient a (details are in the appendix).
This is important for certain applications, but we won’t go into it any further here.
We can see here though that everything is the same as in second order. For a char-
acteristic equation with real coefficients, any complex roots must still come in conjugate
pairs. So if some of the roots in {s1 , ..., sn } are complex, the roots can be grouped as:

s = α1 ± iβ1 , ..., αk ± iβ k , s2k+1 ,..., sn ,

where α1 ± iβ1 , ..., αk ± iβk are the complex roots s1, ..., s 2k and s2k+1, ..., s n are real. Then,
the general solution is
α1 x α1 x αk x αk x s2k+1 x
y = c1 e sin(β1 x) + c2 e cos(β1 x) + ... + c2k−1e sin(βx) + c2k e cos(βk x) + c2k+1 e

This might seem a little complicated when written out in general form like this. But in
practice it is the obvious thing to do. For example,
(5) (4) ′′′ ′′ ′
Example 2.3.15. The characteristic equation of y +7y +21y +33y +28y +10y = 0 has
−2x
roots s = −2+i, −2−i, −1+i, −1−i, −1. The general solution is of the form y = c1 e sin(x)+
−2x −x −x −x
c2 e cos(x) + c3 e sin(x) + c4 e cos(x) + c5 e .

Finally, we must consider the case of repeated roots. Although we could write a formal
description of this case as well, things have already gotten complicated enough. Again
what happens is exactly like second order, with the following additions:

• A root can be repeated arbitrarily many times. Each repetition gives another solu-
tion multiplied by x. So for instance a real root si that occurs p times will give the p
s x s x p−1 s x
solutions e 1 ,xe 1 , ..., x e 1 .

• Complex roots can also be repeated, in which case the solutions corresponding to
the repeated roots must also be multiplied by x.
′′′ ′′ ′
Example 2.3.16. Find the general solution to y − 6y + 12y − 8y = 0.

C. Sprouse —Math 280 Notes Page 155 DRAFT August 26, 2022
2.3. Constant Coefficient ODE’s: Fundamental Solution Sets 156

solution: The characteristic equation here is actually a special one. With a little
experience one can recognize it as

s − 6s + 12s − 8 = (s − 2) .
3 2 3

So the general solution is

2x 2x 2 2x
y = c1 e + c2 xe + c3 x e .

′′′′′ ′′′′
′′′
Example 2.3.17. Find the general solution to y − 6y + 9y = 0.

solution: The characteristic equation factors as

s − 6s + 9s = s (s − 6s + 9) = s (s − 3) .
5 4 3 3 2 3 2

So the general solution is

2 3x 3x
y = c1 + c2 x + c 3 x + c 4 e + c5 xe □.

(4) ′′′ ′′ ′
Example 2.3.18. Find the general solution to y + 4y + 14y + 20y + 25y = 0.

solution: Here the factorization is far less obvious, but the characteristic equation is

s + 4s + 14s + 20s + 25 = (s + 2s + 5) .
4 3 2 2 2

So by the quadratic formula, the roots are s = −1 + 2i, −1 + 2i, −1 − 2i, −1 − 2i. The general
solution is
−x −x −x −x
y = c1 e sin(2x) + c2 e cos(2x) + xc3 e sin(2x) + xc4 e cos(2x) □ .

All of these things can be mixed together:

Example 2.3.19. Find the general solution to y (6) + y (4) − y ′′ − y = 0.

solution: The characteristic equation factors as

s + s − s − 1 = (s − 1)(s + 2s + 1) = (s + 1)(s − 1)(s + 1) .

6 4 2 2 4 2 2 2

This has roots s = −1,1, i, −i, i, −i. So the general solution is

−x x
y = c1 e + c2 e + c3 cos(x) + c4 sin(x) + c5 x cos(x) + c6 x sin(x) □ .

C. Sprouse —Math 280 Notes Page 156 DRAFT August 26, 2022
 2.3. Constant Coefficient ODE’s: Fundamental Solution Sets 157

Homework 2.3:
A) Find the homogeneous solution to the second order constant coefficient ODE’s. Your solution must be real.
Note:
• These are pretty easy, but you should practice them. A great many points in Math 280 are taken from
′′ ′′ ′′ ′
thosewho fail to distinguish between y −y = 0, y + y = 0, and y + y = 0. So practice these and don’t
do that!
(3−i)x (3+i)x
• An answer of the form y = c1 e + c2 e is always incorrect in this class. Complex numbers are
used for intermediate steps, but your final answers can never have anything imaginary.
• For as + bs + c = 0, look for complex roots when c > (b/2) . If c is negative the roots are always real.
2 2

1.
′′ ′
y + 4y + 5y = 0.
solution: The characteristic polynomial is s + 4s + 5 = 0. Since 5 > (4/2) = 2 = 4, the
2 2 2

roots are complex. The quadratic formula gives

−4 ± 16 − 20 −4 ± 2i
s= = = −2 ± i.
2 2
−2x −2x
So y = c1e cos(x) + c2 e sin(x) □.

2. 11. 20.
′′ ′ ′′ ′ ′′ ′
y − 3y + 2y = 0. y + 2y + y = 0. y − 6y + 9y = 0.
3. 12.
′′ ′ ′′ ′ 21.
y + 2y − 3y = 0. y + 2y = 0. ′′ ′
y − 6y + 34y = 0.
4. 13.
′′ ′
′′
y + y = 0. y + 2y + 5y = 0. 22.
′′ ′
14. y + 20y + 99y = 0.
5. ′′
′′ ′ y + 9y = 0.
y − y = 0. 23.
15.
6. ′′ ′ ′′ ′
′′ y + 9y + 8y = 0. y + 20y + 100y = 0.
y − y = 0.
16.
7. ′′ ′ 24.
′′
y − 2y + y = 0.
′ y + 9y = 0.
8. 17. ′′
y + 20y + 101y = 0.
′
′′
′′
y − 2y + 2y = 0.
′ y − 9y = 0.
18. 25.
9. ′′ ′ ′′
′′ ′
y − y − 2y = 0. y − y − 6y = 0. y + 100y = 0.
10. 19.
′′ ′ ′′ ′
y + y − 2y = 0. y − 2y + 17y = 0.

C. Sprouse —Math 280 Notes Page 157 DRAFT August 26, 2022
 2.3. Constant Coefficient ODE’s: Fundamental Solution Sets 158

B) Find the homogeneous solution to the higher order constant coefficient ODE’s. Your solution must be
real.
26.
′′′′ ′′
y + 8y + 16y = 0.

solution: s + 8s + 16 factors as (s + 4) . So the roots are s = ±2i, ±2i. So

4 2 2 2

y = c1 cos(2x) + c2 sin(2x) + c3 x cos(2x) + c4 x sin(2x) □ .

27. 35.
′′′ ′′ ′ ′′′′ ′′′
y − 2y − y + 2y = 0. y + 2y = 0.
(Hint: s = 1 is a root. 36.
′′′′ ′′′ ′′ ′
Polynomial division.) y + 3y + 3y + y = 0.
28.
′′′ ′′ ′ (s = −1 is a root.)
y + y − y − y = 0.
37.
(s = 1 is a root. Polynomial division.) ′′′ ′
y + 4y = 0.
29.
′′′ ′′ ′ 38.
y − y + y − y = 0. ′′′′ ′′
y − y = 0.
(s = 1 is a root.)
39.
30. ′′′′ ′′
′′′ ′′ ′ y − 2y + y = 0.
y − 4y + y + 6y = 0.
40.
(s = −1 is a root.) ′′′′ ′′
y + 2y + y = 0.
31.
′′′ ′′
y − 3y + 4 = 0. 41.
′′′′
y − 16y = 0.
(s = 2 is a root.)
32. 42.
′′′ ′′ ′ ′′′′
y − 3y + 3y − y = 0. y + 64y = 0.

(s = 1 is a root.) (Hint: s + 64 = (s − 4s + 8)(s 2 + 4s + 8).)

4 2

33. 43.
′′′ ′′′′ ′′
y + y = 0. y − 18y + 81y = 0.
(s = −1 is a root.) 44.
(5) (4) (3)
34. y + 6y + 9y = 0.
′′′
y − 8y = 0.
45.
(5) (3)
(s = 2 is a root.) y +y = 0.

C. Sprouse —Math 280 Notes Page 158 DRAFT August 26, 2022
 2.3. Constant Coefficient ODE’s: Fundamental Solution Sets 159

C) Find a homogeneous ODE with the given general solution.

46.
−2x −2x
y = c1 cos(x) + c2 sin(x) + c3 e cos(x) + c4 e sin(x).
solution: The characteristic polynomial is

(s − i)(s + i)(s − (−2 + i))(s − (−2 − i)) = (s + 1)(s + 4s + 5).

2 2

Note for the complex arithmetic, we use that

(a + bi)(a − bi) = a + (ab − ab)i − b i = a + b .

2 2 2 2 2

and

(s − (a + bi))(s − (a − bi)) = s − ((a + bi) + (a − bi))s + (a + bi)(a − bi) = s − 2as + (a + b ).

2 2 2 2

Then multiplying out,

(s + 1)(s + 4s + 5) = s + 4s + 6s + 4s + 1,
2 2 4 3 2

′′′′ ′′′ ′′ ′
so the ODE is y + 4y + 6y + 4y + y = 0 □.

47. 59.
y = c1 cos(8x) + c2 sin(8x). −x x
y = c1 e + c2e + c3 cos(x) + c4 sin(x).
48. 60.
x 3x
y = c1 e + c 2 e . x −x 2x −2x
y = c1 e + c2 e + c3 e + c4 e .
49.
x
y = c1 e cos(3x) + c2e sin(3x).
x 61.
50. y = c1 cos(x) + c2 sin(x) + c3 cos(2x) + c4 sin(2x).
−3x −3x
y = c1 e cos(4x) + c2e sin(4x).
51. 62.
3x x x
y = c1 + c2 e . y = c1 e + c2 xe + c3 cos(x) + c4 sin(x).
52. 63.
−2x −2x
y = c1 e + c2 xe . y = c1 e
−x
+ c2 xe
−x x
+ c3 e + c4xe .
x

53.
−x x 3x 64.
y = c1 e + c 2 e + c3 e .
x −x −x
54. y = c1 + c2 e + c3 e cos(x) + c4 e sin(x).
−2x x x
y = c1 e + c2 e cos( 3x) + c3 e sin( 3x).
55. 65.
−x −x
y = c1 e + c2 xe + c3 .
y = c1 cos(2x) + c2 sin(2x)
56.
y = c1 e
2x
+ c2xe
2x
+ c3 x e .
2 2x +c3 x cos(2x) + c4 x sin(2x).

57. 66.
−x
y = c1 e + c2 + c3 x.
−x −x
58. y = c1 e cos(x) + c2 e sin(x)
2 −x −x
y = c 1 + c2 x + c 3 x . +c3 xe cos(x) + c4 xe sin(x).

C. Sprouse —Math 280 Notes Page 159 DRAFT August 26, 2022
2.4. Nonhomogeneous Linear ODE’s 160

2.4 Nonhomogeneous Linear ODE’s

For the first three sections of this chapter we have focused almost exclusively on homo-
geneous linear ODE’s. Fortunately it is only a short step to go from there to the general
situation. Recall that the general (nonhomogeneous) linear ODE is
(n) (n−1)
an (x)y + an−1(x)y + ... + a1 (x)y + a0(x)y = f (x),
′
(⋆)
The function f (x) is called a forcing function, and typically represents something like
an applied force, or input signal. Again, our interest is to find the general solution, which
represents all solutions to ⋆. It turns out this is easy to do if we just know one solution to
⋆. We can use this one solution to reduce finding any other solutions to the homogeneous
case. All of the other solutions could be particular solutions as well, but we just need to
choose one of them. So now any single solution to ⋆ will be called a particular solution,
yp (x).

Fact 2.4.1. Let yp (x) be a single solution to

(n) (n−1)
an (x)y + an−1 (x)y + ... + a1(x)y + a0 (x)y = f (x),
′
(⋆)

Then the general solution to ⋆ is of the form

y = c1y1 (x) + ... + cn yn(x) + yp (x),

where {y1, ..., yn } is a fundamental solution set to the homogeneous equation

(n) (n−1)
an (x)y + an−1 (x)y + ... + a1 (x)y + a0(x)y = 0.
′
(⋆⋆)

The proof of this is as follows. Suppose that y is any solution to ⋆. Then, since yp also
solves ⋆, we have
(n) (n−1)
an (x)yp + an−1(x)yp + ... + a1 (x)yp + a0 yp = f (x).
′

This implies that

(n) (n−1)
an(x)(y − yp ) + an−1 (x)(y − yp ) + ... + a1(x)(y − yp ) + a0yp
′

(n) (n−1)
= an(x)y + an−1 (x)y + ... + a1 (x)y + a0 y
′

(n) (n−1)
− an (x)y + an−1 (x)y + ... + a1 (x)y + a0 y
′

= f (x) − f (x) = 0.
Definition 2.4.2. For any nonhomogeneous ODE ⋆, the homogeneous solution yh (x) is the
solution to the homogeneous equation ⋆⋆:
yh (x) = c1 y1 (x) + ... + cn yn (x).

C. Sprouse —Math 280 Notes Page 160 DRAFT August 26, 2022
2.4. Nonhomogeneous Linear ODE’s 161

The above is actually a very general fact from linear algebra. If L is any linear trans-
formation, then the set all of solutions to L(v) = w can be written as a single solution vp ,
plus all of the solutions to L(v) = 0. We don’t need to get into this level of abstractness,
but this is exactly the same result that we ended Chapter 2 with, where we were solving
Ax⃗ = b⃗ . In this case L was given by the matrix A, v was the unknown vector x,
⃗ and w was
the fixed vector b⃗ .
The important thing here is that we already know that the solutions to the homoge-
neous ODE are always of the form

yh = c1 y1 (x) + ... + cnyn (x),

where {y1 , ..., yn } is a fundamental solution set.

We can denote the general solution to an ODE by yg (x). That is, yg (x) represents
all solutions to the ODE. Again, a single solution to the nonhomogeneous ODE ⋆ will be
denoted yp (x) and be called a particular solution. Finally, we will let yh (x) denote the
general solution to the homogeneous equation ⋆⋆. Then, our above fact can be written:

Fact 2.4.3. The general solution to

(n) (n−1)
an (x)y + an−1 (x)y + ... + a1(x)y + a0 y = f (x)
′

is always of the form

yg (x) = yh (x) + yp (x),
where yh (x) is the homogeneous solution and yp (x) is a particular solution.

Example 2.4.4. {cos( x), sin( x)} is a fundamental solution set to 4xy + 2y + y = 0. Show
′′ ′

that yp = x − 2 is a particular solution to

′′ ′
4xy + 2y + y = x,
′′ ′
and find the general solution to 4xy + 2y + y = x.
′′ ′
solution: The first statement gives that the homogeneous solution to 4xy + 2y + y = x is

yh = c1 cos( x) + c2 sin( x).

To check that x − 2 is a particular solution, we can just plug it in:

4x(x − 2) + 2(x − 2) + (x − 2) = 4x(0) + 2(1) + x − 2 = 0

′′ ′
✓.

So yp (x) = x − 2, and the general solution is

yg (x) = c1 cos( x) + c2 sin( x) + x − 2 □ .

C. Sprouse —Math 280 Notes Page 161 DRAFT August 26, 2022
2.4. Nonhomogeneous Linear ODE’s 162

Example 2.4.5. Solve the initial value problem

′′ ′
4xy + 2y + y = x

y(π ) = π ,
2 2

y (π ) = 0.
′ 2

solution: We know from before that the general solution is

y = c1 cos( x) + c2 sin( x) + x − 2.

(After this section we will usually just write just y instead of yg , since the general solution
simply all of the solutions y.) So

y(π ) = π
2 2 2 2
⇒ c1 cos(π) + c2 sin(π) + π − 2 = π .

⇒ c1 (−1) + c2 (0) = 2 c1 = −2.

Also
y (x) = −
′ c1 c2
sin( x) + cos( x) + 1.
2 x 2 x
So
y (π ) = 0
′ 2 c1 c2
⇒ − sin(π) + cos(π) + 1 = 0.
2π 2π
(0) + (−1) = −1
c1 c2
− ⇒ c2 = 2π.
2π 2π
This gives the unique solution

y = −2 cos( x) + 2π sin( x) + x − 2 □ .

y
y(x)
20

(π , π )
2 2

10 ⋅

x
5 10 15 20 25

Example 2.4.6. Consider the nonhomogeneous ODE

2 ′′ ′
x y − xy = 2.
2
Show that yh = c1 + c2 x is the homogeneous solution and yp = − ln(x) is a particular solution
on (0, ∞). Find the general solution on (0, ∞).

C. Sprouse —Math 280 Notes Page 162 DRAFT August 26, 2022
2.4. Nonhomogeneous Linear ODE’s 163

solution: To find the homogeneous solution, we first show that {1, x } is a fundamental
2

solution set to the homogeneous equation

2 ′′ ′
x y − xy = 0.

We plug each function in, to check that they are solutions:

x (1) − x(1) = x (0) − x(0) = 0

2 ′′ ′ 2
✓.

x (x ) − x(x ) = x (2) − x(2x) = 2x − 2x = 0

2 2 ′′ 2 ′ 2 2 2
✓.
The Wronksian is
2
W (x) =
2
1 x 1 x
(1) (x )
′ 2 ′ = = 2x − 0 = 2x.
0 2x

So {1,x2 } is a linearly independent solution set, and 2.1.20 gives that the general solution
to xy − y = 0 on (0, ∞) is
′′ ′

yh = c1 (1) + c2 x = c1 + c2 x .
2 2

Then we can also check that −ln(x) is a particular solution to the original ODE:

x (− ln(x)) − x(−ln(x)) = x (−(x ) − x(−x ) = −x (−x ) + x(x ) = 1 +1 = 2

2 ′′ ′ 2 −1 ′ −1 2 −2 −1
✓.

So we can have that yp = − ln(x) is a particular solution, and

yg (x) = yh (x) + yp (x) = c1 + c2 x − ln(x) □ .

2

2 ′′ ′
Now, note that any solution to is also a particular solution to x y − xy = 2. So given
the general solution above, we could also say that

yp = 1 − ln(x)

or
2
yp = x − ln(x)
are also particular solutions. This would give general solutions of the form
2
yg = c1 + c2 x + 1 − ln(x)

or
2 2
yg = c1 + c2 x + x − ln(x).
These are all different forms of the same general solution! We should put this in a box,
since it will come up later.

The particular solution to a nonhomogeneous ODE ⋆ is never unique.

There are always infinitely many possible choices of a particular solu-
tion. A particular solution represents a specific choice of any one of
these, which is a single solution yp (x).

C. Sprouse —Math 280 Notes Page 163 DRAFT August 26, 2022
2.4. Nonhomogeneous Linear ODE’s 164

So any solution from the general solution could be a particular solution. But it must
be a single choice. For instance we could pick
2
yp = 3 − 5x − ln(x),
as long as the coefficients are chosen. That is, the coefficients of the particular solution
yp (x) can never be arbitrary constants! The arbitrary constants are always part of the
homogeneous solution, yh (x). This is worth remembering in the next section, where we
solve for the coefficients in yp (x). Another thing we will have to keep in mind is that
when solving the initial value problems, we always of course use the general solution to
solve for the initial values, rather than the homogeneous solution, because yg (x) is the
solution to the ODE that we are supposed to be solving!
1 2 1 2
x ′′ 2 x
Example 2.4.7. Show that yp = e 2 − 3x is a particular solution to y − y = x e 2 + 3x. Then
solve the initial value problem Solve the initial value problem
1 2
′′ 2 x
y −y = x e2 + 3x,
y(0) = 0,
y (0) = 0.
′

1 2
solution: First, plug e 2 x − 3x into the ODE to make sure that it is a particular solution:

1 2 ′′ 1 2 1 2 ′ 1 2
x x x x
e2 − 3x − e2 − 3x = xe 2 − 3 −e2 + 3x
1 2 1 2 1 2 1 2
x 2 x x 2 x
= e2 +x e2 −e2 + 3x = x e 2 + 3x ✓.
So then we need to find the homogeneous solution, which we can now do easily because
′′
the homogeneous equation is constant-coefficient. y − y = 0 has characteristic equation
2
s − 1 = 0, so
⇒ (s + 1)(s − 1) = 0
2
s −1 = 0 ⇒ s = ±1.
yh (x) = c1 e
−x x
+ c2 e .
The general solution is then

y(x) = yh (x) + yp (x) = c1 e

1 2
−x x x
+ c2 e + e 2 − 3x.
The derivative is
y (x) = −c1e
′ −x x 1 x2
+ c3 e + xe 2 − 3.
So plugging in the initial values,
0
y(0) = 0 ⇒ c1 + c2 + e − 3(0) = 0 ⇒ c1 + c2 = −1,
y (0) = 0
′ 0
⇒ −c1 + c2 + 0e − 3 = 0 ⇒ −c1 + c2 = 3.
This easily gives c2 = 1 and c1 = −2. So the final answer is
1 2
−x x x
y(x) = −2e +e +e2 − 3x □ .

C. Sprouse —Math 280 Notes Page 164 DRAFT August 26, 2022
2.4. Nonhomogeneous Linear ODE’s 165

y
10 y(x)

x
1 2

Warning 2.4.8. The point of this example is that there is a little bit of work to do before solving
for the coefficients c1 , c2 . A common mistake in this situation is to forget what the original
question was, and plug into the homogeneous solution yh (x) to find c1 , c2 , rather than the
general solution yg (x).
In this example, that would give c1 = 0, c2 = 0, which when plugging into yh (x) gives
yh (x) = 0 as the answer. That would definitely not be the correct solution! This is a particular
issue in the next section, where one might have to do a lot of other work before solving for these
constants near the very end.

C. Sprouse —Math 280 Notes Page 165 DRAFT August 26, 2022
2.5. Undetermined Coefficients 166

2.5 Undetermined Coefficients

Now we look for ways to find the particular solution yp (x) of a nonhomogeneous equation

(n) (n−1)
an (x)y + an−1 (x)y + ... + a1 (x)y + a0(x)y = f (x), .
′
(⋆)

The most common equations of this form occur when ai (x) are all constant, and f (x)
is particularly nice. In this case, we can guess the particular solution to be of the same
form as f (x), with some unknown constants. If we can plug this into the equation ⋆ and
actually solve for these constants, then we have yp . It is traditional to call the general
form of yp (x) a guess, but when done correctly it isn’t really a guess at all. You will
always know the right form of yp if you follow the procedure below. But to understand
this procedure, you can’t memorize a single formula. It requires doing several examples
and a lot of practice.
′′ ′ 5x
Example 2.5.1. Find a particular solution to y + 3y + 2y = e .
5x
solution: We will guess that the particular solution is of the form Ae where A is an
unknown constant that we must solve for. That is, we will plug Ae 5x into y′′ + 3y′ + 2y
5x
and see if we can choose A such that the result is e .

(Ae ) + 3(Ae ) + 2(Ae ) =

5x ′′ 5x ′ 5x 5x 5x 5x
25Ae + 15Ae + 2Ae
5x
= 42Ae

We want
5x 5x
42Ae =e ,
so we have
1
42A = 1 ⇒ A = .
42
So
yp (x) =
1 5x
e □.
42

Again, there are two main requirements for undetermined coefficients to work:

1. The coefficients ai (x) in ⋆ must be constant.

2. The forcing function f (x) in ⋆ must be nice.

The word nice has a special meaning here. Abstractly it means that f (x) is a solution to
any other constant coefficient homogeneous ODE. That is, the kind of functions we saw
as the solutions in Section 2.3. But a more concrete way to express it is that f (x) must be
of the following type:

Definition 2.5.2. A function f (x) is called nice if it is one of either e

αx
, cos(βx), sin(βx), a
polynomial, or any sum or product of these.

Example 2.5.3. Undetermined coefficients will work these forcing functions.

C. Sprouse —Math 280 Notes Page 166 DRAFT August 26, 2022
2.5. Undetermined Coefficients 167

• f (x) = e • f (x) = 3x − 1 • f (x) = 1 + xe

−x x

• f (x) = 3 cos(2x) − sin(2x) • f (x) = x − e • f (x) = e

2x −x
+ cos(3x) sin(2x)

• f (x) = x • f (x) = x sin(x) • f (x) = 2xe cos(x)

2 2 x

Again, f (x) can only contain terms of the form e , cos(βx), sin(βx), or polynomials.
αx

These terms can be added to each other, multiplied by constants, and multiplied by each
other. There are a few functions like this that we can do undetermined coefficients on but
must be simplified first:

• f (x) = sin (x) f (x) =

2 1
⇒ 2
− 12 cos(2x)

• f (x) = sin(3x) cos(3x) ⇒ f (x) = 1

2
sin(6x)

• f (x) = cosh(2x) f (x) = 12 e

−2x 2x
⇒ + 12 e
Now lets list some non-examples.
Example 2.5.4. Undetermined coefficients will not work with these functions:

• f (x) = tan(x) • f (x) = x • f (x) = arcsin(x)

• f (x) = sec(x) • f (x) = 1

x • f (x) = e
−x 2

• f (x) = ln(x) • f (x) = ex

• f (x) = cos(x )
3
1+x

Note that even though tan(x) can be written as sin(x)/ cos(x), and cos(x) and sin(x)
are the kinds of terms allowed in f (x), one cannot divide the functions and still get some-
thing eligible for undetermined coefficients. similarly one cannot compose such functions.
2
For instance e−x is the composition of e x with the polynomial −x 2 , but we can’t do unde-
termined coefficients with f (x) = e
2
−x
.
′′ ′ 2
Example 2.5.5. Find the general solution to y − 2y + y = x .

solution: The characteristic equation factors as s − 2s + 1 = (s − 1) , so the

2 2
x x 2
homogeneous solution is yh = c1 e + c2 xe . To get yp , we cannot use yp = Ax , because
when taking derivatives gives (x ) = 2x and (x ) = 2, which are not of the form Ax . To
2 ′ 2 ′′ 2

cancel these terms out, we need to choose yp to be the most general quadratic polynomial:
yp = Ax2 + Bx + C.

(Ax + Bx + C) − 2(Ax + Bx + C) + (Ax + Bx + C)

2 ′′ 2 ′ 2

2A − 2(2Ax + B) + (Ax + Bx + C)
2
=
Ax + (−4A + B)x + (2A − 2B + C) = x .
2 2
=

So
Ax + (−4A + B)x + (2A − 2B + C) = 1x + 0x + 0.
2 2

C. Sprouse —Math 280 Notes Page 167 DRAFT August 26, 2022
2.5. Undetermined Coefficients 168

Setting coefficients equal gives:

A = 1
−4A + B = 0
2A − 2B + C = 0.

The solutions are A = 1,B = 4, and C = 6, so

yp (x) = x + 4x + 6,
2

yg (x) = c1 e + c2 xe + x + 4x + 6 □ .
x x 2

′′ ′
Example 2.5.6. Find the general solution to y + 2y = 5 cos(x).
2
solution: The characteristic equation factors as s + 2s = s(s + 2). So s = 0,−2 gives
−2x
yh = c1 + c2 e . To get yp , we have a similar situation as the last example: we cannot just
use yp = Acos(x), because the derivative of cos(x) is − sin(x), which must be canceled
out. So we need to choose yp = A cos(x) + B sin(x). Inserting this into the ODE gives:

(A cos(x) + B sin(x)) + 2(Acos(x) + B sin(x))

′′ ′
= −A cos(x) − B sin(x) − 2A sin(x) + 2B cos(x)
= (−A + 2B) cos(x) + (−2A − B) sin(x) = 5cos(x).

So
(−A + 2B) cos(x) + (−2A − B)sin(x) = 5 cos(x) + 0 sin(x).
Setting coefficients equal gives
−A + 2B = 5
.
−2A − B = 0
The solutions are A = −1 and B = 2. So,

yp (x) = − cos(x) + 2sin(x),

yg (x) = c1 + c2 e
−2x
− cos(x) + 2 sin(x) □ .

As with the previous two examples, if f (x) = xe , then the particular solution must
αx
αx αx
also include an e term, since the derivative of xe produces these terms. This gives
yp = Axe αx + Be αx .

Example 2.5.7. y′′ − y = 18xe2x

solution: The characteristic equation factors as s 2 − 1 = (s + 1)(s − 1), so the

−x x 2x 2x
homogeneous solution is yh = c1 e + c2e . We guess yp = Axe + Be for the particular
solution. Plugging into the equation gives

(Axe + Be ) − (Axe + Be ) = (A(0)e + 2A(1)(2e ) + Ax(4e ) + 4Be

2x 2x ′′ 2x 2x 2x 2x 2x 2x 2x 2x
− Axe − Be
(3A)xe + (4A + 3B)e
2x 2x 2x
= = xe .

So
(3A)xe + (4A + 3B)e
2x 2x 2x 2x
= 1xe + 0e .

C. Sprouse —Math 280 Notes Page 168 DRAFT August 26, 2022
 2.5. Undetermined Coefficients 169

3A = 18 ⇒ A = 6
4A + 3B = 0 ⇒ B = − 43 A = −8.
So
2x 2x
yp = 6xe − 8e ,
−x x 2x 2x
yg = c 1 e + c2e + 6xe − 8e □.

Here, we’ve again used the product rule for second derivatives. You should too, or
undetermined coefficients will be much harder than it should be!

(f g) = f g + 2f g + f g .
′′ ′′ ′ ′ ′′

The above examples indicate how things work in general. Below is a formal table
which shows the possibilities for f (x) and the corresponding yp :
f (x) Guess for yp

αx αx
ae Ae

a cos(βx) + b sin(βx) A cos(βx) + B sin(βx)

n n
an x + ...+ a1 x + a0 An x + ...+ A1 x + A0

αx αx αx αx
ae cos(βx) + be sin(βx) Ae cos(βx) + Be sin(βx)

n αx αx αx n αx αx αx
an x e + ... + a1 xe + a0 e An x e + ... + A1 xe + A0e

n n
an x cos(βx) + ... + a1 x cos(βx) + a0 cos(βx) Anx cos(βx) + ... + A1 x cos(βx) + A0 cos(βx)
n n
+bn x sin(βx) + ... + b1 x sin(βx) + b0 sin(βx) +Bn x sin(βx) + ... + B1 x sin(βx) + B0 sin(βx)

n αx αx αx n αx αx αx
an x e cos(βx) + ... + a1 xe cos(βx) + a0 e cos(βx) An x e cos(βx) + ... + A1 xe cos(βx) + A0 e cos(βx)
αx n αx αx n αx αx αx
+bn e x sin(βx) + ... + b1xe sin(βx) + b0 e sin(βx) +Bn x e sin(βx) + ... + B1 xe sin(βx) + B0 e sin(βx)

Note that everything in the table is obvious, if one understands that the guess is al-
ways the most general thing of the same form as the forcing function f (x). But, one has
to be careful about what “general thing of the same form” means. That is, the guess cor-
responding to f (x) = x is Ax + Bx + C, not just Ax . Since x is a quadratic polynomial,
2 2 2 2
2
the guess has to be a general quadratic polynomial. Note that the guesses for 2x − 1
2 2 2 −x 2 −x
and x + 2x + 3 are also Ax + Bx + C. Similarly, the guess of x e is not Ax e , but
2 −x −x −x
Ax e + Bxe + Ce . So again:

C. Sprouse —Math 280 Notes Page 169 DRAFT August 26, 2022
2.5. Undetermined Coefficients 170

Fact 2.5.8. The particular solution yp will be of the same general form as f (x), provided that
no terms of this form match any terms of the homogeneous solution yh (x).
Note that there is an important condition here. Things change when any part of the
guess for yp matches yh , which we have not discussed yet. In this case, the above proce-
dure must be modified. Here is a very simple example of what goes wrong, and how it is
fixed.
′′ x
Example 2.5.9. Find a particular solution to y − y = e .

solution: The characteristic equation factors as s − 1 = (s + 1)(s − 1), so the

2
−x x x
homogeneous solution is yh = c1 e + c2e . And now we can see that choosing yp = Ae
doesn’t work:
(Ae ) − (Ae ) = e
x ′′ x x x x x x
⇒ Ae − Ae = e ⇒ 0=e .
This means there are no solutions for A, and we chose the wrong thing. This will always
happen when we plug in something from the homogeneous solution. To fix it, we
x
multiply by x: yp = Axe . More precisely, we multiply any term that matches the
homogeneous solution by x. Now let’s see that this works:

(Axe ) − (Axe )
x ′′ x x x x
= A(0) + 2A(1)e + Axe − Axe
x
= e .
So
(0)xe + (2A)e = 1e .
x x x

Note that the introduced term has 0 for the coefficient, since there is no term like that on
the right-hand side. This always happens in the matching cases, and can be used to check
that you are on the right track. We are left with a new term that we can use, though.
1
2A = 1 ⇒ A= ,
2
so
1 x
yp = xe □ .
2

Note the difference between yp in the above two examples.

Do not get these confused!

• y ′′ − y = xe2x ⇒ yp = Axe 2x + Be 2x (no matching yh ).

• y ′′ − y = ex ⇒ yp = Axe x (matching yh ).

In particular, don’t add a second term that shouldn’t be there in the second case! Also,
this is why we have always found the homogeneous solution before we choose a guess for
yp . If choose yp without knowing the homogeneous solution yh , and don’t notice that a
matching occurs, then your guess for yp will be wrong, and you will either get a wrong
answer or be wasting a lot of time fixing it. To take this one step further:

C. Sprouse —Math 280 Notes Page 170 DRAFT August 26, 2022
2.5. Undetermined Coefficients 171

′′ x
Example 2.5.10. Find a particular solution to y − y = xe .
−x x x x
Again, yh = c1 e + c2e . For the particular solution, we would have yp = Axe + Be ,
but the Bex term matches the homogeneous solution and we multiply it by x. But we
x x
cannot use yp = Axe + Bxe either, because now both terms are the same, and we can’t
afford to lose any terms. So we must also multiply the Axe x term by x:
2 x x
yp = Ax e + Bxe .

Let’s plug this in to see if it works:

(Ax e + Bxe ) − (Ax e + Bxe )

2 x x ′′ 2 x x

= (2Ae + 4Axe + Ax e + 2Be + Bxe − Ax e − Bxe

x x 2 x x x 2 x x

= (A − A)x e + (4A)xe + (2A + 2B)e = xe .

2 x x x x

So
(0)x e + (4A)xe + (2A + 2B)e = 18xe + 0e .
2 x x x x x

(Note that again the coefficient of the introduced term is zero.)

4A = 1 ⇒ A = 14
2A + 2B = 0 ⇒ B = −A = − 14 .
2 x
So yp = 14 x e − 14 xe □ .
x

In general, when a part your guess matches part of the homogeneous solution, you
must multiply that part in your guess by x. You should multiply the terms in your guess
by x until no terms in the guess match the homogeneous solution, and no terms in the
guess match any other terms in the guess. However, you should not multiply unrelated
terms by x!
′′ ′ 2x 3x
Example 2.5.11. Find the form of the particular solution. y − 2y = 3e −e

solution: yh = c1 + c2 e 2x , so yp = Axe 2x + Be3x . We do not multiply the e3x term by x

because it does not match the homogeneous solution □.
′′ ′ −x
Example 2.5.12. Find the form of the particular solution. y − 2y = 2x + 3e .
2x 2 −x
solution: yh = c1 + c2 e , so yp = Ax + Bx + Ce . We had to multiply the first two terms
−x
by x to avoid them matching yh or each other. But e does not match, so is not multiplied
by x.
′′ ′ x
Example 2.5.13. Find the form of the particular solution. y − 2y + y = e □.

solution: The characteristic equation factors as s − 2s + 1 = (s − 1) so the homogeneous

2 2
x x 2 x x
solution is yh = c1 e + c2 xe . This means yp = Ax e . We have to multiply Ae by x twice,
x x
so that it doesn’t match either c1 e or c2 xe in the homogeneous solution □.

Again...

C. Sprouse —Math 280 Notes Page 171 DRAFT August 26, 2022
2.5. Undetermined Coefficients 172

Do not get these confused!

• y ′′ − y = x2 e 2x ⇒ yp = Ax 2 e2x + Bxe 2x + Ce 2x (no matching yh ).

• y ′′ − y = xex ⇒ yp = Ax2 e x + Bxex (matching yh ).

′′ ′ x 2 x
• y − 2y + y = e ⇒ yp = Ax e (matching yh twice).

Now let’s do some with sin and cos.

Example 2.5.14. Find the general solution to y ′′ + 4y = 10cos(3x) − 5 sin(3x).

solution: The homogeneous solution is

yh = c1 cos(2x) + c2 sin(2x).

This does not match Acos(3x) + Bsin(3x). So the correct guess is

yp (x) = A cos(3x) + B sin(3x). And, omitting the computations this time, we have

yp (x) = −2 cos(3x) + sin(3x),

yg (x) = c1 cos(2x) + c2 sin(2x) − 2 cos(3x) + sin(3x) □ .

Example 2.5.15. Find the general solution to y ′′ + 4y = cos(2x).

solution: Again, the homogeneous solution is yh = c1 cos(2x) + c2 sin(2x). But now this
does match A cos(2x) + B sin(2x). So the correct guess is yp (x) = Ax cos(2x) + Bx sin(2x).

(Ax cos(2x) + Bx sin(2x)) + 4(Ax cos(2x) + Bx sin(2x))

′′

= −4A sin(2x) − 4Ax cos(2x) + 4Bcos(2x) − 4Bx sin(2x) + 4Ax cos(2x) + 4Bx sin(2x)
= −4A sin(2x) + 4B cos(2x)
= cos(2x).

So,
4A = 1 ⇒ A = 14
.
−4B = 0 ⇒ B = 0
And,
yp (x) =
1
x sin(2x),
4

yg (x) = c1 cos(2x) + c2 sin(2x) + x sin(2x) □ .

1
4

Example 2.5.16. Find the general solution to y + y = 4 cos (x).

′′ ′ 2

C. Sprouse —Math 280 Notes Page 172 DRAFT August 26, 2022
2.5. Undetermined Coefficients 173

−x
solution: This time we have yh = c1 + c2 e . Now for yp , we cannot say
yp = A cos2 (x) + B sin2 (x). This is not one of the forms of yp we have discussed above!
There is a big difference between cos(x) and cos (x)!! To get yp in the correct form we can
2

use the half-angle formula: cos (x) = 12 (1 + cos(2x)). This lets us rewrite the question as
2

′′ ′
y + y = 2 + 2 cos(2x).

Now, since A + B cos(2x) + C sin(2x) has the constant term matching the homogeneous
solution, we must use yp = Ax + B cos(2x) + C sin(2x). Plugging in,

(Ax + Bcos(2x) + C sin(2x)) + (Ax + B cos(2x) + C sin(2x))

′′ ′

= −4B cos(2x) − 4C sin(2x) + A − 2Bsin(2x) + 2C cos(2x)

= A + (−4B + 2C) cos(2x) + (2B − 4C) cos(2x) = 2 + 2cos(2x),

Or
A + (−4B + 2C)cos(2x) + (2B − 4C) cos(2x) = 2 + 2 cos(2x) + 0 sin(2x).
So A = 2, and the equations −4B + 2C = 2, 2B − 4C = 0 give B = − 25 , C = 15 . So

2 1
yp = 2x − cos(2x) + sin(2x),
5 5
−x 2 1
yg = c 1 + c 2 e + 2x − cos(2x) + sin(2x) □ .
5 5

The above was our first example where f (x) = 2 + 2 cos(2x) had more than one term to
take care of. Here is another.
′′ ′ 2x
Example 2.5.17. Find the general solution to y − y − 2y = 4x + 3e .

solution: The characteristic equation factors as s − s − 2 = (s + 1)(s − 2) so the

2
−x 2x 2x
homogeneous solution is yh = c1 e + c2 e . Since Ax + B + Ce has a term that matches
2x
the homogeneous solution, we must use yp = Ax + B + Cxe . Again, we do not multiply
the Ax + B by x, because it wasn’t involved in the matching. Plugging in,

(Ax + B + Cxe ) − (Ax + B + Cxe ) − 2(Ax + B + Cxe )

2x ′′ 2x ′ 2x

2x 2x 2x 2x 2x
= 4Ce + 4Cxe − A − Ce − 2Cxe − 2Ax − 2B − 2Cxe
= (−2A)x + (−A − 2B) + (4C − 2C − 2C)xe + (4C − C)e = 4x + 3e .
2x 2x 2x

(−2A)x + (−A − 2B) + (0)xe + (3C)e

2x 2x 2x
⇒ = 4x + 0 + 3e .
−2A = 4 ⇒ A = −2
−A − 2B = 0 ⇒ B = − 12 A = 1
3C = 3 ⇒ C = 1.
So
yp (x) = −2x + 1 + xe
2x

yg (x) = c1 e
−x 2x 2x
+ c2 e − 2x + 1 + xe □.

C. Sprouse —Math 280 Notes Page 173 DRAFT August 26, 2022
2.5. Undetermined Coefficients 174

One might wonder in this example (and the one before), if one could find yp for each
term individually, and then add the terms. This is true:

Fact 2.5.18 (‘Superposition’). A particular solution yp (x) for a linear differential equation

(n) (n−1)
+ ... + a1 y + a0 y = f 1 (x) + f2 (x)
′
an y + an−1 y

is given by
yp (x) = yp,1 (x) + yp,2 (x),
where yp,1 (x) and yp,2 (x) are the particular solutions to

(n) (n)
+ ... + a1 y + a0 y = f 1 (x) + ... + a1 y + a0 y = f 2 (x).
′ ′
any and an y

In other words, to get a particular solution to

′′ ′ 2x
y − y − 2y = 4x + 3e

above, we could solve the two equations

′′ ′ ′′ ′ 2x
y − y − 2y = 4x , y − y − 2y = 3e

separately and then add the results. This is a fundamental property of linear equations.
But it isn’t really that helpful in practice, since it is easier to find yp all at once, as we did
above. But it does explain in the matching situation why some things get multiplied by
x while other things don’t. If the form of f1 (x) has terms that match yh (x) while f 2 (x)
doesn’t, there is certainly no reason to multiply the particular solution for f 2 (x) by x,
since the matching doesn’t happen in the second equation!
Finally, we have completely ignored initial value problems in this section up to now.
The important thing here is that when solving for the initial conditions, one must use the
general solution yg (x) for y, not the homogeneous solution yh . That is why you spent all
the time solving for yp !
Example 2.5.19. Solve the initial value problem
−x 2x 2x
c1 e + c2 e − 2x + 1 + xe ,

y(0) = 5,
y (0) = −2.
′

solution: We previously solved for the general solution yg (x). Setting y = yg (x) gives
−x 2x 2x
y = c1 e + c2 e − 2x + 1 + xe ,
′ −x 2x 2x 2x
y = −c1e + 2c2 e − 2 +e + 2xe .
So
y(0) = 2 ⇒ c1 + c2 + 2(0) + 1 + 0(1) = 5 ⇒ c1 + c2 = 4.

C. Sprouse —Math 280 Notes Page 174 DRAFT August 26, 2022
2.5. Undetermined Coefficients 175

y (0) = −5
′
⇒ −c1 + 2c2 − 2 + 1 − 3(0)(1) = −2 ⇒ −c1 + 2c2 = −1.
Solving c1 + c2 = 4 and −c1 + 2c2 = −1 gives c1 = 3,c2 = 1. So the unique solution is
−x 2x 2x
y = 3e +e − 2x + 1 + xe .

10 y −x 2x 2x
y = 3e +e − 2x + 1 + xe .
9

5
x
0.2 0.4 0.6 0.8 1 □.

Again, the important thing is that we used the full general solution when solving for the
initial values, which includes yp . This may seem obvious here, but can sometimes be
forgotten at the end of lengthy problems.

When solving an IVP after undetermined coefficients, always remember

to include yp when solving for the initial values. That is, always use the
general solution:
y = yg (x) = yh (x) + yp (x).

Example 2.5.20. Solve the initial value problem

′′ ′
y + 2y + 5y = 10,

y(0) = 0,
y (0) = 0.
′

2
solution: The characteristic s + 2s + 5 has complex roots, s = −1 ± 2i. So
−x −x
yh = c 1 e cos(2x) + c2 e sin(2x).

The particular solution is yp = A, and plugging in gives

(A) + 2(A) + 5A = 10
′′ ′
⇒ A = 2.

So yp = 2, and
−x −x
y = c1 e cos(2x) + c2e sin(2x) + 2.
Then to solve for the initial values,
0 0
y(0) = c1 e cos(0) + c2 e sin(0) + 2 = 0 ⇒ c1 = −2.

C. Sprouse —Math 280 Notes Page 175 DRAFT August 26, 2022
2.5. Undetermined Coefficients 176

Then
′ −x −x −x
y = −c1 e cos(2x) − 2c1 e sin(2x) − c2 e sin(2x) + 2c2 cos(2x).
So
y (0) = 0
′ 1
⇒ −c1 + 2c2 = 0 ⇒ c2 = c1 = −1,
2
and
−x −x
y = −2e cos(2x) − e sin(2x) + 2.

3
y(x)

1 2 3 4 5 □.

Note that if we used yh in the above example to solve for c1 and c2 instead of the
general solution with yp , we would get c1 = c2 = 0, which is definitely not the answer! So
don’t do this!

Example 2.5.21. Solve the initial value problem

′′ ′ −2x
y + 3y + 2y = e ,

y(0) = 1,

y (0) = 0.
′

solution: s + 3s + 2 = (s + 1)(s + 2) so yh = c1 e + c2 e . Since this matches Ae

2 −x −2x −2x
, the
−2x
guess for yp must be yp = xe . Plugging in and solving for A gives A = −1. So

−2x
yp = −xe .

Then to solve for the initial conditions,

−x −2x −2x
y = c1 e + c2 e − xe ⇒ y(0) = c1 + c2 .

y (0) = −c1 − 2c2 − 1 = 0.

′ −x −2x −2x −2x ′
y = −c1 e − 2c2 e −e + 2xe ⇒
So the initial conditions give

c1 + c2 = 1 c1 = 3
⇒ .
−c1 − 2c2 = 1 c2 = −2

So
−x −2x −2x
y = 3e − 2e − xe .

C. Sprouse —Math 280 Notes Page 176 DRAFT August 26, 2022
2.5. Undetermined Coefficients 177

1.5 y(x)

0.5

1 2 3 4 5 □.

Notice that in each of the graphs above, the solutions do satisfy the correct initial
values.

C. Sprouse —Math 280 Notes Page 177 DRAFT August 26, 2022
 2.5. Undetermined Coefficients 178

Homework 2.5:
A) Find the homogeneous solution, and then write the general form of yp (x). Do not solve for the coefficients.
1.
′′ ′ 2x 3x
y − 3y = x − cos(x) + 3xe +e .
solution: First find the homogeneous solution. s 2 − 3s = 0 ⇒ s = 0, 3 ⇒ yh = c1 + c2 e 3x .

Ax + B (match!)
2
x ⇒ ⇒ Ax + Bx.

−cos(x) ⇒ C cos(x) + D sin(x).

2x 2x 2x
3xe ⇒ Exe + Fe .
(match!)
3x 3x 3x
e ⇒ Ge ⇒ Gxe .
So
2 2x 2x 3x
yh = Ax + Bx + C cos(x) + D sin(x) + Exe + Fe + Gxe .
3x
Note that we DON’T multiply the nonmatching groups by x. Also, note that Gxe is
NOTGxe3x + He 3x. Do not get these two different things confused! □.
′′ ′ 2 ′′
2. y + 2y = x . 14. y + y = cos(x) + sin(2x).
′′ ′ 2x ′′
3. y − 3y = 3x + 2 − e . 15. y + 4y = x sin(2x).
′′ 2 2x x ′′ ′
4. y − y = x e −e . 16. y + y = 2x sin(x) + cos(x).
′′ x x
5. y − y = 3xe − 2e . ′′
17. y + 4y = x sin(x) + cos(2x).
′′ x 2x
6. y − 4y = 3xe + e . ′′ ′ x
′′ 2 3x 3x
18. y − 2y + y = e cos(x).
7. y − 9y = x e − 2e . ′′ ′ −x
′′ ′ 2x 19. y + 2y + 2y = e cos(x).
8. y − 3y + 2y = 2 + 3e . ′′ 2x
′′ ′ 2x 3x 20. y + y = xe sin(x).
9. y − 3y = 3e + 2e .
′′
′′ ′ 3x 2x 21. y − 9y = cosh(3x).
10. y − 3y + 2y = e + xe .
′′ ′
′′
11. y + 2y + y = x e
′ 2 −x
. 22. y + 3y + 2y = sinh(2x).
23. y + 3y = sin (x).
′′ 2x ′′ ′ 2
12. y + y = e − cos(x).
24. y + 4y = cos (x).
′′ ′′ 2
13. y + 4y = cos(x) + 2 sin(x).

B) Find the general solution by undetermined coefficients.

25.
′′ ′ 3x 3x
y − 2y + 2y = 20xe +e .
2 x x
solution: Always find yh first. s − 2s + 2 = 0 ⇒ s = 1 ± i ⇒ yh = c1 e cos(x) + c2 e sin(x).yp
3x 3x
is then of the form yp = Axe + Be , which does not match yh . So we plug into theleft-hand
side. Using (f g ) = f g + 2f g + f g , we have
′′ ′′ ′ ′ ′′

(Axe + Be )
3x 3x ′′ 3x 3x 3x 3x
= 0e + 6Ae + 9Axe + 9Be ,
(Axe + Be )
3x 3x ′ 3x 3x 3x
= Ae + 3Axe + 3Be .

So,

= (6Ae + 9Be ) − 2(Ae + 3Be ) + 2(Axe + Be )

′′ ′ 3x 3x 3x 3x 3x 3x 3x 3x
y − 2y + 2y + 9Axe + 3Axe
= (5A)xe + (4A + 5B)e
3x 3x

C. Sprouse —Math 280 Notes Page 178 DRAFT August 26, 2022
 2.5. Undetermined Coefficients 179

So
5A = 20,
4A + 5B = 1.
3x 3x x x 3x 3x
A = 4, B = −3 so yp = 4xe − 3e and y = c1 e cos(x) + c2 e sin(x) + 4xe − 3e □.
′′ ′ 2 ′′ ′
26. y + 3y + 2y = 4x + 2x − 1. 32. y + 2y + 5y = 3 cos(2x) + 5 sin(2x).
′′ 2x 2
27. y − y = 3e −x . ′′ ′
33. y + y = 2 cos(x) − 10 sin(2x).
′′ ′ 3x
28. y − 2y + y = 4xe . ′′ 2x
′′ ′ −x −x 34. y − y = 10e sin(x).
29. y + 4y + 5y = 4xe − 2e .
′′ x x
′′
30. y + y = 2x e .
2 −x 35. y + y = 6e cos(2x) + 2e sin(2x).
′′ ′ ′′
31. y + 2y + y = −20 cos(3x) + 10 sin(3x). 36. y − 4y = 16x cos(2x).

C) Solve the initial value problems.

37.
y(0) = 0, y (0) = 1.
′′ ′ 2x ′
y − 5y + 4y = 4x + 3 + 2e ,
solution: The homogeneous solution is
s − 5s + 4 = 0 ⇒ (s − 1)(s − 4) = 0 ⇒ s = 1, 4 ⇒ yh = c1 e + c2 e . Thus there is no matching
2 x 4x
2x
with the terms of yp , and we have yp = Ax + B + Ce . Solving for the coefficients gives
2x
yp = x + 2 − e . Therefore
x 4x 2x
y = c 1 e + c 2e + x+ 2 −e .
′ x 4x 2x
y = c1e + 4c2 e + 1 − 2e .
This gives
c1 + c2 + 2 − 1 = 0, c1 + c2 = −1, c1 = −2
⇒ ⇒ .
c1 + 4c2 + 1 − 2 = 1 c1 + 4c2 = 2 c2 = 1
So
x 4x 2x
y = −2e + e +x +2−e □.

y(0) = 0, y (0) = 1.
′′ ′ 2x ′
38. y − 8y + 15y = 3e ,
y(0) = 1, y (0) = 0.
′′ ′ −x ′
39. y − 6y + 9y = 32e ,
y(0) = 0, y (0) = 0.
′′ ′ ′
40. y + 3y + 2y = 2x − 1,
y(0) = 3,y (0) = 2.
′′ ′ 2 ′
41. y − 2y + y = x ,
y(0) = 3, y (0) = 0.
′′ ′ −x ′
42. y + 6y + 9y = 8e ,
y(0) = −1,y (0) = 4
′′ ′ x ′
43. y − y − 2y = 4 − 2e ,
y(0) = 0, y (0) = 0.
′′ ′ 2x ′
44. y − y = 4xe ,
y(0) = 0,y (0) = 0.
′′ ′ x ′
45. y − 4y + 5y = e ,
y(0) = 0, y (0) = 1.
′′ ′ ′
46. y + 3y + 2y = 2 cos(x) + 4 sin(x),
y(0) = 1, y (0) = 0.
′′ ′ ′
47. y + y − 2y = 20 cos(2x),
y(0) = 0, y (0) = 2.
′′ ′ ′
48. y + 2y + 2y = cos(x) − 2 sin(x),
D) Find the general solution by undetermined coefficients.
49.
′′ 2x
y + 4y = 8e + 4 sin(2x).

C. Sprouse —Math 280 Notes Page 179 DRAFT August 26, 2022
 2.5. Undetermined Coefficients 180

solution: The homogeneous solution is

2 2x 2x
s + 4 = 0 ⇒ s = ±2i ⇒ yh = c1 cos(2x) + c2 sin(2x).For the particular solution, 8e ⇒ Ae
which does not match yh . But B cos(2x) + C sin(2x) does match yh so we must multiply by x.
So
2x
yp = Ae + Bx cos(2x) + Cx sin(2x).
Using (f g ) = f g + 2f g + f g , we have
′′ ′′ ′ ′ ′′

(Bx cos(2x)) = 0 cos(2x) − 4B sin(2x) − 4Bx cos(2x),

′′

(Cx sin(2x)) = 0 sin(2x) + 4C cos(2x) − 4Cx sin(2x).

′′

So plugging yp in for y,
′′ 2x 2x
y + 4y = 4Ae − 4Bsin(2x) − 4Bx cos(x) + 4C cos(2x) − 4Cx sin(2x) + 4Ae + 4B cos(2x) + 4C sin(2x)
2x
= 8Ae + 4C cos(2x) − 4B sin(2x)
2x
= 8e + 0 cos(2x) + 4 sin(2x).
2x
So we have 8A = 8,4C = 0, −4B = 4, giving A = 1, C = 0, B = −1. So yp = e − x cos(2x) and
y = c1 cos(2x) + c2 sin(2x) + e 2x − x cos(2x) □ .
′′ ′ ′′ 3x
50. y + 2y = 4x − 5 cos(x). 56. y − 9y = 36xe .
′′ ′ 2 2x
51. y − 2y = 12x − 24x + 4e . ′′ x
57. y − y = 4e + 3xe
2x 2x
+e .
′′ ′ −2x
52. y + 3y + 2y = e − 10sin(x). ′′ ′ x
′′ ′ x 2x 58. y − y = 2xe .
53. y − 7y + 10y = 4e + 3e .
′′ ′
′′
54. y + y = 6cos(x) + 4 sin(x). 59. y + 2y + y = 8 cosh(x).
60. y + 4y = 8cos (x).
′′ ′ 5x ′′ 2
55. y − 10y + 25y = 2e .

E) Solve the initial value problems.

61. y + y = 3x , y(0) = 3, y (0) = 2.
′′ ′ 2 ′

62. y − y = 2x − 5, y(0) = 1,y (0) = 0.

′′ ′ ′

y(0) = 0, y (0) = 0.
′′ ′ x ′
63. y + 2y = 4x + 3e ,
y(0) = 0, y (0) = 1.
′′ ′ 2x ′
64. y − y = 2 + 2e ,
y(0) = 1,y (0) = 0.
′′ ′ 3x ′
65. y − 4y + 3y = 2e ,
66. y − y = 2e + 3e , y(0) = 0,y (0) = 0.
′′ x 2x ′

67. y − 3y + 2y = 4x − e , y(0) = 1, y (0) = 2.

′′ ′ x ′

y(0) = 1,y (0) = 2.

′′ x ′
68. y − y = 4xe ,
y(0) = 0,y (0) = 0.
′′ x 2x ′
69. y − y = 2e − 9xe ,
y(0) = 0, y (0) = 1.
′′ ′ ′
70. y + 2y = 2 − 5cos(x),
71. y + 2y − 3y = 8e + 10cos(x), y (0) = 1, y (0) = 0.
′′ ′ x ′ ′

72. y + y = 4cos(x) − 2 sin(x), y(0) = 1,y (0) = 0.

′′ ′

73. y + 4y = 4 sin(2x) + 8 cos(2x), y(0) = 0,y (0) = 1.

′′ ′

74. y + 9y = 6 sin(3x), y(0) = −1,y (0) = 2.

′′ ′

y(0) = 0, y (0) = 1.
′′ ′ 2x ′
75. y − 4y + 4y = 8 − 2e ,
y(0) = 3, y (0) = 2.
′′ ′ x −x ′
76. y + 2y + y = 4e − 2e ,
77. y + 2y = 16sin (x), y(0) = 0,y (0) = 0.
′′ ′ 2 ′

y(0) = 1, y (0) = 0.
′′ ′ ′
78. y − 2y = 16cosh(2x),

C. Sprouse —Math 280 Notes Page 180 DRAFT August 26, 2022
2.6. Variation of Parameters 181

2.6 Variation of Parameters

The method of Undetermined Coefficients in the previous section allows us to solve a non-
homogeneous linear ODE when it is constant-coefficient and the forcing function f (x) is
nice. What about other cases? It turns out there is a method to solve all nonhomogeneous
linear ODE’s, provided that one can solve the homogeneous equation. That is, consider
the general linear 2-nd order ODE, written in the form

y (x) + p(x)y (x) + q(x)y = f (x).

′′ ′
(4.6.1)

If we know the homogeneous solution yh(x), then there is actually a formula that gives a
particular solution yp (x) and the general solution, for any f (x)!
To derive this, assume we know the homogeneous solution. That is, we know linearly
independent functions y1 (x) and y2(x) such that
′′ ′ ′′ ′
y1 + py1 + qy1 = 0 , y2 + py2 + qy2 = 0.

And then
yh (x) = c1y1 (x) + c2y2 (x).
The method of variation of parameters is to assume that a solution to (4.6.1) is of the
form
y(x) = u1 (x)y1 (x) + u2 (x)y2 (x).
Then we try to solve for the functions u1(x) and u2(x) that will make this true. That is,
′′ ′
we will try to plug u1 y1 + u2 y2 into the equation y + py + qy = f and see if we can solve
for the functions u1 , u2 . Furthermore we’ll make one more assumption in advance that
simplifies things:
u1 (x)y1 (x) + u2 (x)y2 (x) = 0.
′ ′

Again, we will show in the end that we can solve for such a u1 ,u 2 . So now the first
derivative becomes

(u1 y1 + u2 y2 )
′ ′ ′ ′ ′
= u1 y1 + u1 y1 + u2 y2 + u2 y2
′ ′
= u1 y1 + u2 y2 .

And plugging that into the second derivative gives

(u1y1 + u2 y2 ) = (u1 y1 + u2y2 )

′′ ′ ′ ′

′ ′ ′′ ′ ′ ′′
= u1 y1 + u1 y1 + u2 y2 + u2 y2 .

With this, inserting y = u1 y1 + u2 y2 into (4.6.1) becomes

(u1 y1 + u1 y1 + u2 y2 + u2y2 ) + p(u1 y1 + u2 y2 ) + q(u1y1 + u2 y2 ) = f .

′ ′ ′′ ′ ′ ′′ ′

This is still a pretty long equation, but we have not yet used that y1 and y2 are solutions
to the homogeneous equation. That is, rearranging the above equation gives

✘✘ ✿0
✘ ✿0
✘
✘
u1 y1 + u2 y2 + u1 (✘ + qy1 ) + u2 (✘ + qy 2) = f
′ ✘
py✘1✘ py✘2✘
′ ′ ′ ′ ′′ ′ ′′
y1✘ +✘ y2✘ +✘

C. Sprouse —Math 280 Notes Page 181 DRAFT August 26, 2022
2.6. Variation of Parameters 182

′′ ′
So using that y1 and y2 both solve y + py + qy = 0 gives the much simpler equation
′ ′ ′ ′
u1y1 + u2 y2 = f .
′ ′
So if we can solve both this, and our previous assumption u1y1 + u1 y2 = 0, then we will be
done. This is actually a system of two equations:

y1u1′ + y2 u2′ = 0
′ ′ ′ ′ .
y1u1 + y2 u2 = f
′ ′
We put the u1 , u2 terms on the right, because those are actually the unknowns that we
′ ′
want to solve for, while y1 , y1 , y2 , y2 are the coefficients! In matrix form, this is
′
y1 y2 u1 0
′ ′ ′ = .
y1 y2 u2 f

This could be solved by taking the inverse of the matrix on the left. And we know that
the determinant of that matrix is the Wronskian, and hence nonzero. So this matrix is
invertible! Moreover, we have a formula for the inverse from Section 2.6. This leads to:

u1′ 1 y2′ −y2 0

′ = ′ ′ ′ .
u2 y1 y2 − y2 y1 −y1 y1 f

And multiplying out,

′
u1 1 −y2f
′ = ′ ′
.
u2 y1 y2 − y2 y1 y1 f
′ ′
Using that y1 y2 − y2 y1 is the Wronskian, we get

u1 (x) = (−y2 (x)f (x))

′ 1
W (x)
u2 (x) = (y (x)f (x))
′ 1 .
W (x) 1

Integrating,

−y2 (x)f (x) y1 (x)f (x)

u1 (x) = u2 (x) =
W (x) W (x)
dx , dx

And again, the general solution to the ODE is:

yg (x) = u1(x)y1(x) + u2 (x)y2 (x).

To show how this works, we can revisit the first example in the previous section:

C. Sprouse —Math 280 Notes Page 182 DRAFT August 26, 2022
2.6. Variation of Parameters 183

′′ ′ 5x
Example 2.6.1. Find a particular solution to y + 3y + 2y = e .

solution: The characteristic equation factors as s + 3s + 2 = (s + 1)(s + 2), so the

2

homogeneous solution is yh = c1 e −2x + c2 e −x . The Wronskian is then

e −x e −2x
W (x) =
−3x
−x −2x = −e .
−e −2e
So,
−2x 5x
e e 6x 1 6x
u1 = − dx = e dx = e + c1.
−e−3x 6
e −x e 5x 7x 1 7x
u2 = dx = − e dx = − e + c2.
−e −3x 7

y = u 1 y1 + u 2 y2
1 6x −x 1 7x −2x
= e + c1 e + − e + c2 e
6 7
−x −2x 1 5x 1 5x
= c 1 e + c2 e + e − e
6 7
−x −2x 1 5x
= c 1 e + c2 e + e .
42
This is actually the general solution for y, which one always gets by variation of
−x −2x 1 5x
parameters. Since c1 e + c2 e is the homogeneous solution, we have that 42 e is a
particular solution. And this is the same thing we got with undetermined coefficients in
the last section! □

Actually, we don’t always get the same particular solution as undetermined coeffi-
cients in this way. This is because a particular solution is not unique. There are many
possibilities for yp (x). More precisely, for any fixed choice of c1 ,c 2 in the above, we get
a different particular solution. We’ll do one more example from the previous section to
elaborate on this.
′′ x
Example 2.6.2. Find a particular solution to y − y = e .

solution: The characteristic equation factors as s − 1 = (s + 1)(s − 1) so yh = c1 e

2 −x x
+ c2 e .
−x x
W (x) =
e e 0 0
= e + e = 2.
−e −x ex

So,
x x
e e 1 2x 1 2x
u1 = − dx = − e dx = − e + c1 .
2 2 4
−x x
e e 1 1
u2 = dx = dx = x + c2 .
2 2 2

y = u1 y1 + u2 y2
1 2x −x 1 −2x
= − e + c1 e + x + c2 e
4 2
−x x 1 x 1 x
= c 1 e + c2 e − e + xe .
4 2

C. Sprouse —Math 280 Notes Page 183 DRAFT August 26, 2022
2.6. Variation of Parameters 184

So it looks like yp (x) above is yp = − 14 e + 12 xe . Actually, this is what we get if we choose

x x

c1 = c2 = 0. But we could choose c1 = 0,c2 = − 14 , instead! This choice of ci ’s cancels the 14 e x

term in the above. This gives yp = 12 xe , the answer from undetermined coefficients! □.
x

x
′′ ′ e
Example 2.6.3. Find a particular solution to y − 2y + y = x2
.

solution: Note that since f (x) = e x /x2 is not nice, this cannot be done with
undetermined coefficients! The characteristic equation factors as s − 2s + 1 = (s − 1) so
2 2

yh (x) = c1e + c2xe .

x x

W (x) =
x x
e xe 2x 2x 2x 2x
x x x = e + xe − xe = e ,
e e + xe

−xe ( xe 2 )
x
x
1
u1 = 2x
dx = − x dx = −ln(x) + c1 ,
e
ex ( xe 2 )
x
1 −1
u2 = dx = dx = −x + c2 .
e 2x x2
So the general solution is

(− ln(x) + c1)e + (− x + c2 )xe

x 1 x
y =
x x x x
= c1 e + c2 xe − ln(x)e − e .

Choosing c1 = c2 = 0 gives the particular solution yp = − ln(x)ex − e x . But choosing

x
c1 = 1, c2 = 0 gives the nicer particular solution yp = − ln(x)e .

′′
Example 2.6.4. Find a particular solution to y + 4y = sec(2x) csc(2x).

solution: The forcing function f (x) = sec(2x) csc(2x) is definitely not nice, so we can’t
use undetermined coefficients. The characteristic equation is s 2 + 4 = 0 ⇒ s = ±2i. So
yh = c1 cos(2x) + c2 sin(2x).

= 2(cos (2x) + sin (2x)) = 2.

cos(2x) sin(2x) 2 2
W(x) =
−2 sin(2x) 2 cos(2x)

−sin(2x) sec(2x) csc(2x) 1 1

u1 = dx = − sec(2x)dx = − ln∣ sec(2x) + tan(2x)∣ + c1 ,
2 2 4

csc(2x)dx = − ln ∣cot(2x) + csc(2x)∣ + c2 .

cos(2x) sec(2x) csc(2x) 1 1
u2 = dx =
2 2 4
So a particular solution is

yp = − ln ∣ sec(2x) + tan(2x)∣ cos(2x) − ln∣ cot(2x) + csc(2x)∣ sin(2x) □ .

1 1
4 4

2 ′′ ′ 5
Example 2.6.5. Find the general solution to x y − 6xy + 10y = x . Use the homogeneous
2 5
solution yh = c1 x + c2 x .

C. Sprouse —Math 280 Notes Page 184 DRAFT August 26, 2022
2.6. Variation of Parameters 185

5
solution: Here, the right-hand side (= x ) is ‘nice’, according to the definition in the
previous section, however the ODE is not constant-coefficient, so we cannot use
undetermined coefficients. To use variation of parameters, we need yh (x) and the forcing
function f (x). This is not a constant coefficient ODE, so yh (x) must be given, as we only
have a way to solve for it in the constant coefficient case. That’s why yh (x) is given above.
With this we can get the Wronskian,
2 5
W (x) =
x x 6 6 6
4 = 5x − 2x = 3x .
2x 5x

Now note that when using our variation of parameters formula, the forcing function f (x)
is not x ! To get f (x), we must have the ODE written in the form (4.6.1). In particular,
5
2 ′′ ′ 5
there can’t be anything in front of the leading term. So the ODE x y − 6xy + 10y = x
2
must be divided by x to get
′′ 6 ′ 10 3
y − xy + x y = x .

That is, f (x) = x .

3

5 3
x x 1 2 1 3
u1 = − dx = − x dx = − x + c1,
3x6 3 9
2 3
x x 1 1 1
u2 = dx = x dx = 3 ln(x) + c2 .
3x 6 3
So the general solution is
1 3 2 1 5
yp = − x + c1 x + ln(x) + c2 x
9 3
2 5 1 5 1 5
= c1 x + c2 x − x + x ln(x).
9 3

Choosing c1 = c2 = 0 gives yp (x) = − 19 x + 13 x ln(x), while choosing c1 = 0, c2 =

5 5 1
9
gives
the nicer particular solution yp (x) = 1 5
3
x ln(x) □.

So we have seen that the f (x) in variation of parameters must be when the equation
is written in the form (4.6.1). For this reason, we refer to (4.6.1) as the standard form of
a second order linear ODE, and when applying variation of parameters, we only refer to
the right-hand side f (x) as a forcing function when the equation is in standard form.
2 ′′ ′
Example 2.6.6. Find the general solution to x y − 2xy + 2y = 4ln(x). Use the homogeneous
2
solution yh = c1 x + c2 x .
Here we have both a non constant-coefficient ODE, and the function 4 ln(x) is not nice.
So again variation of parameters is required.
solution: The Wronskian is
2
W (x) =
x x 2 2 2
= 2x − x = x .
1 2x

Again, 4ln(x) is not the forcing function. To get f (x) must write the equation in standard
form
′′ 2 ′ 2 4 ln(x)
y − xy + 2y = .
x x2

C. Sprouse —Math 280 Notes Page 185 DRAFT August 26, 2022
2.6. Variation of Parameters 186

So f (x) =
4 ln(x)
x2
, giving

2 4 ln(x)
x x2 −2 −1 −1
u1 = − dx = −4 x ln(x)dx = 4x ln(x) + 4x + c1 ,
x2

4 ln(x)
x
x2 −3 −2 −2
u2 = dx = 4 x ln(x)dx = −2x ln(x) − x + c2 .
x2

We omitted the evaluation of the above integrals, but recall when k ≠ −1 that ∫ xk ln(x)dx
is evaluated by parts, differentiating the ln(x). So the general solution is

−1 −1 −2 −2 2
y = 4x ln(x) + 4x + c1 x + −2x ln(x) − x + c2 x
2
= c1 x + c2 x + 2 ln(x) + 3 □ .

′′ ′ 3 3
Example 2.6.7. Find a particular solution to xy − y + 4x y = x . Use
yh = c1 cos(x ) + c2 sin(x ).
2 2

solution: Again, since the ODE is not constant-coefficient, we must be given the
homogeneous solution. The Wronskian is

cos(x ) sin(x )
2 2
W (x) = = 2(cos (x ) + sin (x )) = 2x.
2 2 2 2
−2x sin(x 2 ) 2x cos(x 2 )

And again we must rewrite the equation first to get f (x). Dividing by the leading
coefficient,
′′ 1 ′ 2 2
y − x y + 4x y = x .

So f (x) = x and
2

sin(x )x
2 2
x sin(x )dx = cos(x ) + c1 ,
1 2 1 2
u1 = − dx = −
2x 2 4

cos(x )x
2 2
x cos(x )dx = sin(x ) + c2 .
1 2 1 2
u2 = dx =
2x 2 4
So the general solution is

cos(x ) + c1 cos(x ) + sin(x ) + c2 . sin(x )

1 2 2 1 2 2
y =
4 4

= c1 cos(x ) + c2 sin(x ) + cos (x ) + sin (x )

2 2 1 2 2 2 2
4
= c1 cos(x ) + c2 sin(x ) + .
2 2 1
4

So we simply have yp (x) = 14 ! Note that in retrospect, this is obviously a particular

solution. Plug it into the left hand side of the ODE above to see why □.

C. Sprouse —Math 280 Notes Page 186 DRAFT August 26, 2022
2.6. Variation of Parameters 187

2.6.1 Definite Integrals and the forced solution yF (x)

So far we have seen how variation of parameters can be used to find the general solution
to nonhomogeneous ODE’s by evaluating some (indefinite) integrals. But the truth is
that there are very few ODE’s that can be solved as we have done so far that can’t also
be solved with undetermined coefficients. Check the exercises in any introductory ODE
book to verify this fact. The variety of exercises you will find in the section on variation
of parameters will be extremely limited, because ODE’s that give solvable (elementary)
antiderivatives for u1 (x) and u2 (x) are exceptionally rare. This does not mean variation
of parameters is useless, however! What it means is that this is not how variation of
parameters is actually used.
Since the issue is that the indefinite integrals for u1 and u2 can hardly ever both be
computed explicitly, the solution is to replace them both by definite integrals:

−y2 (s)f (s) y1 (s)f (s)

u1 (x) = u2 (x) =
x x

W (s) W (s)
ds + c1 , ds + c2 .
x0 x0

But we need to understand the details of what happens when we do this. Our solution
to the nonhomogeneous ODE (4.6.1) looks like
−y2(s)f (s) y1 (s)f (s)
ds + c1 y1 (x) + ds + c2 y2(x).
x x

W (s) W (s)
y=
x0 x0

This is the general solution. Or if we choose specific values for c1 , c2 , we get a particular
solution. With no other information to use, the obvious choice is c1 = c2 = 0, which gives
the particular solution
−y2 (s)f (s) y1(s)f (s)
yp (x) = ds y1 (x) + ds y2 (x).
x x

x0 W (s) x0 W (s)
But since we are most often interested in solving initial value problems, we will need
to find the initial values of yp (x) above. That is, we need yp (x0) and yp (x0 ). Any definite
′

integral of the form ∫a g(s) ds is always zero. So we easily have that

a

−y2 (s)f (s) y1 (s)f (s)

yp (x0 ) = ds y1 (x0 ) + ds y2 (x0) = 0.
x0 x0

x0 W (s) x0 W (s)
yp (x0 )?
′
Now what about To find this, we have to differentiate yp , using the product rule
and the Fundamental Theorem of Calculus:
−y2 (x)f (x) x −y (s)f (s)
yp (x) = y1 (x) + ds y1 (x)
′ 2 ′
W (x) x0 W (s)
y1 (x)f (x) x y (s)f (s)
y2 (x) + ds y2 (x)
1
W (x) W (s)
+
x0
x −y (s)f (s) x y (s)f (s)
ds y1(x) + ds y2 (x).
2 ′ 1 ′
(s) (s)
=
x0 W x0 W

C. Sprouse —Math 280 Notes Page 187 DRAFT August 26, 2022
2.6. Variation of Parameters 188

Magically, half of the terms generated by the product rule cancel each other out! So we
have a much simpler form of yp′ than anticipated, which allows us to plug in x0 again:

−y2 (s)f (s) y1 (s)f (s)

yp (x0 ) = ds y1 (x0 ) + ds y2 (x0 ) = 0.
x0 x0
′ ′ ′
x0 W (s) x0 W (s)

Definition 2.6.8. The forced solution yF (x) to a nonhomogeneous linear ODE

y + p(x)y + q(x)y = f (x)

′′ ′

is the unique solution with initial values y(x0 ) = 0, y (x0 ) = 0.

′

We have shown:

Theorem 2.6.9 (Variation of Parameters). The forced solution yF (x) to

y + p(x)y + q(x)y = f (x)

′′ ′

is
−y2 (s)f (s) y1 (s)f (s)
yF (x) = ds y1 (x) + ds y2(x),
x x

x0 W (s) x0 W (s)
or

y1 (s)y2 (x) − y2(s)y1(x)

yF (x) = f (s) ds.
x

x0 W (s)

Example 2.6.10. Find the solution to the IVP as a definite integral:

′′ ′
y + 2y + y = x

y(4) = 0,

y (4) = 0.
′

solution: The characteristic equation factors as s + 2s + 1 = (s + 1) , so the

2 2
−x −x
homogeneous solution is yh = c1 e + c2xe . The Wronskian is
−x −x
W (x) =
e xe −2x −2x −2x −2x
−x −x −x = e − xe + xe =e .
−e e − xe

So

e (xe ) − se e
−s −x −s −x
yF (x) =
x
s ds
4 e −2s

e (xe ) − (se )e
x
s −x s −x
= s ds □ .
4

C. Sprouse —Math 280 Notes Page 188 DRAFT August 26, 2022
2.6. Variation of Parameters 189

Example 2.6.11. Find the solution to the IVP as a definite integral. Use the homogeneous
solution yh = c1 x + c2 x ln(x).
2 ′′ ′
x y − xy + y = sin(x)
y(1) = 0,
y (1) = 0.
′

solution: The Wronskian is

W (x) =
x x ln(x)
= x ln(x) + x − x ln(x) = x.
1 ln(x) + 1

Recall that the forcing function is not sin(x), but rather we must rewrite the equation as

′′ 1 ′ 1 sin(x)
y − xy + 2y =
x x2

to get f (x) =
sin(x)
x2
. So

yF (x)
x s(x ln(x)) − s ln(s)x sin(s)
= s ds
1 s2
(x ln(x) − x ln(s))
x
sin(s)
= ds □ .
1 s2

Now, the point of knowing the forced solution yF (x) is that it allows us to solve any
initial value problem, with initial conditions y(x0) = b1 , y (x0 ) = b2 . Since yF (x) is a
′

particular solution to the ODE, we have that

y = yh (x) + yF (x)

is the general solution. Then it is an easy matter to plug the initial values into this, because
we know that yF (x0) = yF (x0 ) = 0.
′

Example 2.6.12. Solve the IVP

′′
y + 4y = 1x ,
y(π) = −1,
y (π) = 4.
′

2
solution: The characteristic equation is s + 4 = 0 so s = ±2i and
yh (x) = c1 cos(2x) + c2 sin(2x). This gives W (x) = 2, so

yF (x) =
x cos(2s) sin(2x) − sin(2s) cos(2s) 1
π 2 s ds.

Then to solve the IVP,

y = c1 cos(2x) + c2 sin(2x) + yF (x),

y = −2c1 sin(2x) + 2c2 cos(2x) + yF (x),

′ ′

C. Sprouse —Math 280 Notes Page 189 DRAFT August 26, 2022
2.6. Variation of Parameters 190

So
y(π) = −1 ⇒ (1)c1 + 0c2 + yF (π) = −1 ⇒ c1 = −1.

y (π) = 4 (0)c1 + 2c2 + yF (π) = 4

′ ′
⇒ ⇒ c2 = 2.

(Again, y(π) = y (π) = 0!) So then

′

x cos(2s) sin(2x) − sin(2s) cos(2s)

y = −cos(2x) + 2sin(2x) + ds □ .
π 2s

Example 2.6.13. Solve the IVP

x 2 y ′′ + 2xy ′ − 2y = e x ,
y(1) = 2,
y (1) = 5.
′

−2
Use the homogeneous solution yh = c1 x + c2 x.

solution: The Wronskian is

−2
W (x) =
x x −2 −2 −2
= x + 2x = 3x .
−2x−3 1

The equation in standard form is

x
′′ 2 ′ 2 e
y + xy − 2y = 2,
x x

so f (x) = ex
x2
and
s −2x − sx−2 es
yF (x) =
x
ds.
1 3s −2 s2
To solve the IVP, the general solution is

+ c2 x + yF (x),
−2
y = c1 x

and
+ c2 + yF (x)
′ −3 ′
yh = −2c1 x
So
y(1) = 2 ⇒ c1 + c2 + yF (1) = c1 + c2 + 0 = 2,

y (1) = 5 −2c1 + c2 + yF (1) − 2c1 + c2 + 0 = 5.

′ ′
⇒
Solving gives c1 = −1,c2 = 3, so

1 1 x x s s
y =− + 3x + − 2 e ds □ .
x 2 3 1 s 2
x

C. Sprouse —Math 280 Notes Page 190 DRAFT August 26, 2022
2.6. Variation of Parameters 191

So, if yF (x) is our particular solution, then we only need to use the homogeneous solu-
tion to get the initial values. (This is definitely not the case with other particular solutions,
such as the one that comes from undetermined coefficients.) A more formal way of saying
this is:

Fact 2.6.14. Consider the initial value problems

y + p(x)y + q(x)y = 0,⎫

⎪
⎪ y + p(x)y + q(x)y = f (x),⎫
⎪
⎪
′′ ′ ′′ ′
⎪
⎪ ⎪
⎪
y(x0 ) = b1 , ⎬
⎪ (I ) y(x0 ) = 0, ⎬
⎪ (II ).
⎪ ⎪
,
y (x0 ) = b2 . ⎪
⎪ y (x0 ) = 0. ⎪
⎪
⎭ ⎭
′ ′

If yI (x) solves (I ) and yF (x) solves (II), then the solution to the IVP

y + p(x)y + q(x)y = f (x),

′′ ′

y(x0 ) = b1 ,
y (x0 ) = b2
′

is
y = yI (x) + yF (x).

So in Example 2.6.12, we have

yI (x) = − cos(2x) + 2 sin(2x),

and in Example 2.6.13 we have

yI (x) = −
1
+ 3x.
x2
Note that yI (x) is not the ‘homogeneous solution’, but rather a specific function, cho-
sen from the homogeneous solution to satisfy the initial value problem.

C. Sprouse —Math 280 Notes Page 191 DRAFT August 26, 2022
 2.6. Variation of Parameters 192

Homework 2.6:
A) a) Find a particular solution by undetermined coefficients.
b) Find a particular solution by variation of parameters.
c) Are the answers the same? If not, what is the difference?
1.
′′ ′ 3x
y − 2y − 3y = 16xe
solution: a) The homogeneous solution is yh = c1 e −x + c2 e 3x , and the particular solution is
2 3x 3x
yp = Ax e + Bxe . (There is matching with yh).Plugging in yp and solving for the
2 3x 3x
coefficients gives yp = 2x e − xe .
−x 3x
b) yh = c1 e + c2 e , so
−x 3x
W (x) =
e e 2x
−x 3x = 4e .
−e 3e

−e (16xe )
3x 3x
4x 4x 1 4x
u1 = =− 4xe dx = −xe + e + c1 .
4e 2x 4
e −x (16xe3x ) 2
u2 = 2x
= 4x dx = 2x + c2 .
4e
So

y = (−xe + e + c1 )e + (2x + c2 )e = c1 e + c2 e + 2x e − xe + e
4x 1 4x −x 2 3x −x 3x 2 3x 3x 1 3x
4 4
1 3x 2 3x 3x
+ e . yp = 2x e − xe
4
c) Here, the answers from i) and ii) for yp are not the same! This can happen because the
particular solution is not unique.The difference will always be part of the homogeneous
3x
solution. In this case, subtracting the answer of i) from ii) gives a difference of 14 e . Notice
this is part of the homogeneous solution and will give 0 when plugged in to the left hand
side of the ODE□.

2. 7.
′′ ′ 5x ′′ 2x
y − 3y + 2y = e y − 4y = xe

3. 8.
′′ 3x
′′
y − 5y + 6y = e
′ 3x y − y = 32xe

4. 9.
′′
′′ ′ x y + y = cos(x)
y + y − 2y = e
10.
5. ′′
y + 9y = sin(3x)
′′ ′ 3x
y + 2y + y = e
11.
6. ′′ ′
y + y = 2x
′′ ′ 5x
y − 10y + 25y = e
12.
′′ ′
y − 3y + 2y = 4x

C. Sprouse —Math 280 Notes Page 192 DRAFT August 26, 2022
 2.6. Variation of Parameters 193

B) Use variation of parameters to find the general solution y and the particular solution yp .
13.
3x
′′ ′ e
y − 3y + 2y =
1 + ex
solution: s − 3s + 2 = (s − 1)(s − 2) = 0 gives s = 1, 2, so the homogeneous solution is
2
x 2x
yh = c 1 e + c 2 e .
2x
W (x) = x
x
e e x 2x x 2x 3x
2x = 2e e − e e =e .
e 2e
3x
2x e
e 2x
1+e x e
u1 = − dx = − dx.
e 3x 1 + ex
3x
x e
e 1+e x ex
(1 + e x )
u2 = dx = dx.
e 3x
x x
Both integrals can be done with the substitution u = 1 + e , du = e dx.
x
e u −1
u1 = − u du = − u du
1
= − 1 − u du

= −u + ln(u) + c1 = −1 − e + ln(1 + e ) + c1 .
x x

u du = ln(u) + c2 = ln(1 + e ) + c2
1 x
u2 =

So,
y = −1 − e + ln(1 + e ) + c1 . e + ln(e + 1) + c2 . e
x x x x 2x

+ e ln(1 + e ) + e ln(1 + e ).
x 2x x 2x x x 2x x
y = c1 e + c2 e −e −e
So
+ e ln(1 + e ) + e ln(1 + e ).
x 2x x x 2x x
yp = −e − e
x 2x
Note that the particular solution is not unique! In this case we can see that −e − e is part
of the homogeneous solution, another particular solution would be
yp = e ln(1 + e ) + e ln(1 + e ). □.
x x 2x x

′′ 5x
14. y + 9y = sec(3x) ′′ ′ e
24. y − 10y + 25y = x
15. y + 25y = sec (5x)
′′ 2

16. y + y = sec (x)

′′ ′ 3x
′′ 3 25. y − 6y + 9y = xe .
′′ −x
26. y + 2y + y = √
17. y + y = sec(x) csc(x) ′′ ′ e
′′
18. y + y = tan(x) 1 − x2
27. y + 3y + 2y = cos(e ).
′′ ′′ ′
19. y + y = sec(x) tan(x) x
′′ ′ x
20. y − 2y + 2y = e csc(x). ′′ ′ 3
′′ ′ x 28. y − 3y + 2y =
21. y − 2y + 5y = e cos(2x). 1 + e −x
−x
′′ ′ e ′′ ′ 1
22. y + 2y + y = 3 . 29. y + 3y + 2y = 2x
x e +1
2x
′′ ′ e ′′ ′ 1
23. y − 4y + 4y = 2 30. y + 6y + 8y = 2x
x +1 e +1

C. Sprouse —Math 280 Notes Page 193 DRAFT August 26, 2022
 2.6. Variation of Parameters 194

C) For each ODE with non-constant coefficients, use the given homogeneous solution to find a particular
solution by variation of parameters.
31.
2 ′′ ′ 4
x y − 4xy + 6y = x ln(x),
2 3
yh = c 1 x + c 2 x .
solution: The Wronskian is
x2 x3
W (x) =
4
2 =x .
2x 3x
To write the formulas for u 1 and u2 , remember that the variation of parameters formula was
derived fory + p(x)y + q(x)y = f (x). So the above equation must be divided by x to get
′′ ′ 2

f (x):
′′ −1 ′ −2 2
y − 4x y + 6x y = x ln(x).
So
x3 (x 2 ln(x)) 1 2 1 2
u1 = − dx = − x ln(x) dx = − x ln(x) − x + c1
x4 2 4
x (x ln(x))
2 2
u2 = dx = ln(x) dx = x ln(x) − x + c2
x4
k
Recall that both of the integrals above are done by parts: to integrate x ln(x) (for k ≠ −1)
use u = ln(x), u ′ = 1x ,v ′ = x k , v = k+1
1 k+1
x , so that

k 1 k+1 1 k xk+1 x k+1

x ln(x) dx = x dx =
(k + 1)2
x ln(x) − ln(x) − + C.
k +1 k +1 k+1

So using u1 and u2 from above,

− x ln(x) + x + c1 x + (x ln(x) − x + c2 )x
1 2 1 2 2 3
y =
2 4
2 3 1 4 3 4
= c1 x + c2 x + x ln(x) − x
2 4
Or,
1 4 3 4
yp = x ln(x) − x □ .
2 4

′′ 4 ′ 6 3 2 3
32. y − x y + 2 y = x , yh = c 1 x + c 2 x
x
′′ −2 −1 2
33. y − 2x y = x, yh = c1 x + c2 x
′′ 1 ′ 1 1
34. y + x y + 2 y = 2 , yh = c1 cos(ln(x)) + c2 sin(ln(x))
x x
′′ 3 ′ 4 2 2
35. y − x y + 2 y = 1 yh = c1 x + c2 x ln(x)
x
2 ′′ 3 −1
36. x y − 2y = x , yh = c1 x + c2 x
2 ′′ ′ 3 5
37. x y − 7xy + 15y = x, yh = c1 x + c2 x
′′ ′ 2 2
38. xy − y = x , yh = c1 + c2x .
2 ′′ ′ 2 2
39. x y − 2xy + 2y = x , yh = c1 x + c2 x
′′ ′ 2 3
40. xy − 2y = x , yh = c1 + c2 x

C. Sprouse —Math 280 Notes Page 194 DRAFT August 26, 2022
 2.6. Variation of Parameters 195

2 ′′ ′
41. x y − xy + y = x, yh = c1 x + c2 x ln(x).
2 ′′ ′ 2 2 2
42. x y − 3xy + 4y = x ln(x), yh = c1 x + c2 x ln(x)
2 ′′ ′ 3 2 5
43. x y − 6xy + 10y = 4x ln(x), yh = c 1 x + c 2 x
2 ′′ ′ −1 −2
44. x y + 4xy + 2y = 4 ln(x), yh = c1x + c2 x
′′ ′ 3
45. xy − 2y = ln(x), yh = c1 + c2 x
2 ′′ 2 −1 2
46. x y − 2y = 18x ln(x), yh = c1 x + c2 x
′′ ′ 2 x 2
47. xy − y = x e , yh = c1 + c2x
48. x y − 2xy + (2 + x )y = x ,
2 ′′ ′ 2 3
yh = c1 x cos(x) + c2 x sin(x)
′′ ′
49. cos(x)y − sin(x)y − sec(x)y = cos(x), yh = c1 sec(x) + c2 tan(x)
50. xy + (x − 1)y − y = x , + c2 (x − 1)
′′ ′ 2 −x
yh = c1 e
51. xy − (2x − 1)y + (x + 1)y = e ,
′′ ′ x x 2 x
yh = c1 e + c2 x e
52. x y − 2xy + (2 − x )y = x e ,
2 ′′ ′ 2 3 x −x x
yh = c1 xe + c2 xe
D) Use variation of parameters to write the forced solution in terms of a definite integral.
53.
(yh = c1 x + c2 x )
2 ′′ ′ 2
x y − 2xy + 2y = x cos(x), x0 = 3.

solution: In general, to write the variation of parameters formula in terms of definite

integrals,
x −y (s)f (s)
u1 (x) =
2
W (s)
ds
a

y1 (s)f (s)
u2 (x) =
x

W (s)
ds
a

So
−y2(s)f (s) y1 (s)f (s)
yp (x) = ds y1 (x) + ds y2 (x).
x x

a W(s) a W (s)
Or, combining into a single integral,

y1(s)y2 (x) − y2 (s)y1(x)

yp (x) = f (s) ds.
x

a W (s)

yh (x) and hence y1 (x) and y2 (x) is given, so we need to find f (x) and W (x).To find f (x),
divide by whatever is in front of the highest derivative:

f (x) = x
′′ −1 ′ −2 −1 −1
y − 2x y + 2x y =x cos(x) ⇒ cos(x),

W (x) =
x x2 2
=x .
1 2x
So
2
yF (x) =
x x
−s −1 s −1 2
s cos(s) ds x + s cos(s) ds x .
3 s2 3 s2
x x
−1 −2 2
= −s cos(s) ds x + s cos(s) ds x .
3 3

C. Sprouse —Math 280 Notes Page 195 DRAFT August 26, 2022
 2.6. Variation of Parameters 196

Or, we can plug directly into the single integral:

2 2
yF (x)
x sx − s x −1
= s cos(s) ds
3 s2
x
−2 2 −1
= s x −s x cos(s) ds □ .
3

′′
54. y − y = x, x0 = 4.
′′ 1
55. y + y = x , x0 = π.
′′ ′ 1
56. y − 6y + 8y = , x0 = e.
ln(x)
57. y + 2y + y = cos(x ),
′′ ′ 2
x0 = π.
2 ′′ ′ −x 2 3
58. x y − 4xy + 6y = e , x0 = ln(2). yh = c 1 x + c 2 x .
2 ′′ ′
59. x y − xy + y = sin(x), x0 = 3. y h = c1 x + c2 x ln(x).
2 ′′ 2x −1 2
60. x y − 2y = e , x0 = −1 yh = c 1 x + c2 x .
′′ ′ 2
61. xy − y = arctan(x), x0 = 3. y h = c 1 + c2 x .
yh = c1 cos(x ) + c2 sin(x ).
′′ ′ 3 2 2
62. xy − y + 4x y = 1, x0 = 1.
63. (x − 1)y − (x + 1)y + 2y = e , yh = c1 (x + 1) + c2 e .
′′ ′ x 2 x
x0 = 2.
E) Find the solution to the initial value problem in terms of a definite integral.
64.
y(3) = 0, y (3) = 6. (yh = c1 x + c2 x ).
2 ′′ ′ ′ 2
x y − 2xy + 2y = x cos(x),

solution: From question 53. above we have yF = ∫3 s

x −2 2 −1
x −s x cos(s) ds.Using
y = c1 x + c2 x + yF (x), we have
2

✯0
y(3) = c1 (3) + c2 (9) + ✟ (3)
✟
yF✟ ⇒ 3c1 + 9c2 = 0

✯0
y (3) = c1 + 2c2(3) + ✟ (3)
′ ′ ✟
yF✟ ⇒ c1 + 6c2 = 6.
This gives c1 = −6, c2 = 2, so the solution to the IVP is
x
2 −2 2 −1
y = −6x + 2x + s x −s x cos(s)ds.
3

y(1) = 0, y (1) = 0.
′′ ′
65. y + 4y = ln(x),
y(4) = 0,y (4) = 0.
′′ ′ 1 ′
66. y − 2y + 2y = ,
x
y(2) = 0, y (2) = 0.
2 ′′ ′ ′ −3 −1
67. x y + 5xy + 3y = cos(x), y h = c1 x + c2 x .
y(π) = 0,y (π) = 0,
2 ′′ ′ ′
68. x y + xy + y = tan(x), yh = c1 cos(ln(x)) + c2 sin(ln(x))
y(0) = 2, y (0) = −1.
′′ ′ ′
69. y − 5y + 6y = sec(x),
y(0) = 2, y (0) = 1.
′′ ′ 1 ′
70. y + 2y + y = 2 ,
x +1
′′ π ′ π
71. y + y = ln(x), y = 3, y = 1.
2 2

C. Sprouse —Math 280 Notes Page 196 DRAFT August 26, 2022
2.6. Variation of Parameters 197

y(π) = −2, y (π) = 3.

′′ ′
72. y + 9y = x,
y(1) = 2, y (1) = 3.
2 ′′ ′ ′
73. x y + xy + y = tan(x), yh = c1 cos(ln(x)) + c2 sin(ln(x))
y(2) = 6, y (2) = −1.
2 ′′ ′ x ′ −1
74. x y + xy − y = e , yh = c 1 x + c2 x
y(1) = 0, y (1) = 2.
2 ′′ ′ ′ −3 −1
75. x y + 5xy + 3y = sin(x), yh = c1 x + c2 x .
y(1) = 3, y (1) = 2.
2 ′′ ′ ′ −1 −1
76. x y + 3xy + y = cos(x), yh = c1 x + c2 x ln(x)
77. (x + 1)y + xy − y = x, y(0) = 2, y (0) = 1.
′′ ′ ′ −x
y h = c1 e + c2 x.

C. Sprouse —Math 280 Notes Page 197 DRAFT August 26, 2022
2.7. Oscillations: Amplitude, Phase, and Transients 198

2.7 Oscillations: Amplitude, Phase, and Transients

2.7.1 Sinusoids ∗
Several solutions to the second order ODE’s in this section have terms of the sinusoidal
form y = C1 cos(ωt) + C2 sin(ωt), or the same form multiplied by an exponential. These
solutions are of particular interest, because they occur often in applications. We will
discuss the most basic of these here. As in section 3.1, the dependent variable in the
applications is time, so we use t as the variable here.

Example 2.7.1.
′′ ′
y + 2y + y = 25cos(2t)

−t −t
solution: The general solution is y = c1 e + c2 te − 3 cos(2t) + 4 sin(2t).

Example 2.7.2.
y ′′ + 6y ′ + 25y = 0
y(0) = 3
y ′ (0) = −1

−3t −3t
solution: The solution is y = 3e cos(4t) + 2e sin(4t).

It turns out that one can use trigonometric identities to rewrite these solutions as

y = A cos(ωt − φ)

or
at
y = Ae cos(ωt − φ)
where A and φ are new constants, called the amplitude and phase angle of the sinusoid
y = A cos(ωt − φ).
One way to derive this is with complex numbers. Recall Euler’s formula
iθ
e = cos(θ) + i sin(θ).

This clearly gives

cos(θ) = Re{e }.
iθ

But what about sin(θ)? Again from Euler’s formula we see that sin(θ) = Im{e }. But if
iθ
iθ
we multiply Euler’s formula by −i, we also get −ie = sin(θ) − i cos(θ). Therefore we
also have
sin(θ) = Re{−ie }.
iθ

Adding things together gives

} + C2 Re{−ie }.
iωt iωt
C1 cos(ωt) + C2 sin(ωt) = C1 Re{e
= Re{(C1 − iC 2 )e }.
iωt

C. Sprouse —Math 280 Notes Page 198 DRAFT August 26, 2022
2.7. Oscillations: Amplitude, Phase, and Transients 199

Now from the Appendix, we know that the complex number C1 − iC2 = C1 + (−C2 )i
can be written in polar form as Ae , where A = C12 + C22 and tan(θ) = −C2/C1 . Or, if
iθ

θ = −φ, then tan(φ) = tan(−θ) = −tan(θ) = C2 /C1 .

And then we have

}
−iφ iωt
C1 cos(ωt) + C2 sin(ωt) = Re{Ae e
(−iφ+iωt)
= ARe{e }
}
i(ωt−φ)
= ARe{e
= A cos(ωt − φ).

The formula for φ is tan(φ) = C2 /C1 . Recall also from the Appendix that this does not
always mean that φ = arctan(C2 /C1 ). If C1 is negative, then the correct formula is φ =
arctan(C2 /C1) + π. So with that in mind, we have derived:

The function y(t) = C1 cos(ωt) + C2 sin(ωt) can be rewritten as

y(t) = A cos(ωt − φ),

where
A= C12 + C22 ,
arctan(C2 /C1 ) for C1 > 0
arctan(C2 /C1 ) + π
φ=
for C1 < 0

Also, one has of course has that φ = π/2 when C1 = 0, c2 > 0, and φ = 3π/2 when
c1 = 0, c2 < 0. The formula above is important enough that we will derive it again, this
time without complex numbers.

2nd proof. Start with y = C1 cos(ωt) + C2 sin(ωt). Then (C1 ,C2 ) is a point in the plane
with polar coordinates (R, φ). Therefore C1 = R cos(φ) and C2 = R sin(φ), where R =
C12 + C22 = A. Therefore we have

C1 cos(ωt) + C2 sin(ωt) = R cos(φ) cos(ωt) + R sin(φ)sin(ωt)

= R cos(φ) cos(ωt) + sin(φ)sin(ωt)
= A cos(ωt)cos(φ) + sin(ωt) sin(φ) .

Then we use a standard trig identity. The angle addition formula for cosine is cos(a +b) =
cos(a) cos(b) − sin(a) sin(b). If a = ωt and b = −φ, then

cos(ωt − φ) = cos(ωt) cos(−φ) − sin(ωt)sin(−φ)

= cos(ωt) cos(φ) + sin(ωt) sin(φ).

Inserting this into the above gives

C1 cos(ωt) + C2 sin(ωt) = Acos(ωt − φ)

as before.

C. Sprouse —Math 280 Notes Page 199 DRAFT August 26, 2022
2.7. Oscillations: Amplitude, Phase, and Transients 200

We can graph functions like y = A cos(ωt − φ) using the rules of shifting and stretch-
ing functions from precalculus. But be careful! The shift to the right when graphing
cos(ωt − φ) by shifting cos(ωt) is not the phase angle φ, but rather φ/ω. This comes
from the fact that if we are shifting f (t) = cos(ωt) by a, we get f (t − a) = cos(ω(t − a)) =
cos(ωt − ωa). So ωa = φ, which means the horizontal shift is a = φ/ω.

For f (x) = A cos(ωt − φ) with phase angle φ,

the horizontal shift is
φ
a = ω.

Warning: The term phase shift is used differently, depending on the source, to refer
to either the phase angle φ or the horizontal shift φ/ω. This causes a lot of confusion, and
is reason we use the term ‘horizontal shift’ instead of ‘phase shift’.
To graph a function of the form Acos(ωt − φ) or Ae αt cos(ωt − φ), we use that the
φ
amplitude A stretches cos(ωt), and the horizontal shift ω shifts the zeroes of cos(ωt) by
φ π π 3π (2k+1)π
ω
. The zeroes of cos(ωt) are ..., − 2ω , 2ω , 2ω , ... = 2ω
. So,

αt
The zeros of Ae cos(ωt − φ) are located at

φ π φ π φ 3π
tk = ..., ω − , ω+ , + , ...
2ω 2ω ω 2ω
(2k + 1)π φ
= + ω.
2ω

t
φ π φ π φ π φ 5π
ω
− 2ω ω
+ 2ω ω
+ 3ω ω
+ 2ω

−A

C. Sprouse —Math 280 Notes Page 200 DRAFT August 26, 2022
2.7. Oscillations: Amplitude, Phase, and Transients 201

Example 2.7.3. Graph the solution to the IVP

′′
y + 4y = 0,
y(0) = −1,
y (0) = 2.
′

solution: The solution is y = − cos(2t) + sin(2t) which we rewrite as y = 2cos(2t − 3π 4

).
3π
The graph is obtained by shifting the graph of cos(2x) to the right by 8 . Some of the
zeroes are
3π 3π 3π π 3π π
− + =− , − + = ,
4 8 8 4 8 8
π 3π 5π 3π 3π 9π
+ = , + = .
4 8 8 4 8 8
So here is the graph. You should check that y actually satisfies the initial conditions □.
y
3π
y = 2 cos 2t −
2 4

t
π ≈ 0.4
− 3π
8
≈ −1.2 8
5π ≈ 2.0
8
9π ≈ 3.5
8

−1

− 2

The reason we use the zeroes to fix the graph rather than the extrema (which for cosine
αt
have a slightly simpler form), is that when multiplying by another function such as e ,
the location of the zeroes does not change. So we can use this to graph solutions like
αt
Ae cos(ωt − φ) as well.

Example 2.7.4. Graph the solution to the IVP

′′ ′
y + 2y + 17y = 0,
y(0) = 1,
y (0) = −7.
′

solution: Here the solution to the IVP is y = e −t cos(4t) − 2e −t sin(4t). This gives
A= (2)2 + (−1)2 = 5 and φ = arctan 2
−1
≈ −1.1. So we can rewrite the solution as
−t −t
y = 5e cos(4t + 1.1). 5e is then an envelope function in the sense that the solution y
−t
bounces around between ± 5e . We will look at envelope functions in more detail soon,
−t
when we discuss resonance. Also in this case, multiplying by e changes the locations of
the local maxima and minima of cos(4t + 1.1), so the only specific points that we can

C. Sprouse —Math 280 Notes Page 201 DRAFT August 26, 2022
2.7. Oscillations: Amplitude, Phase, and Transients 202

easily place on the graph are the zeroes. Also note that φ = −1.1 is negative this time, so
the zeroes will move to right by −1.1
4
, which means to the left by 1.1
4
. Some zeroes are

π 1.1 3π 1.1
− ≈ .117 , − ≈ .903,
8 4 8 4
5π 1.1 7π 1.1
− ≈ 1.69 , − ≈ 2.47.
8 4 8 4
Again, one should look at the graph to see that the initial conditions are satisfied □.

y −t
y = 5e cos(4t + 1.1)
5 −t
± 5e

t
.117 .903 1.69 2.47 3.25

−1

− 5

2.7.2 Steady-State vs. Transients

For a non-homogeneous ODE y +by +cy = f (t), we typically think of the left-hand side as
′′ ′

some sort system, and the right-hand side as an applied force. Then the resulting output
is the solution y(t). If b, c > 0 then the system is passive in the sense that if f (t) ≡ 0, then
y(t) will approach 0 as t → ∞, no matter what the initial values of y are. So for the initial
value problem
y + by + cy = f (t),
′′ ′

y(0) = b1 ,
y (0) = b2
′

it stands to reason that for t large (after a long amount of time), the solution y will only
depend on the applied force, and not the initial conditions (which happened a long time
ago). In particular, if f (t) is periodic, then y(t) should settle down to some solution that
is also periodic, which is called the ‘steady state’, and this solution is independent of the
whatever initial conditions y has.

Definition 2.7.5. Suppose the solution to an IVP is of the form

y(t) = yT (t) + yS (t)

where yS is periodic, and lim yT (t) = 0. Then yT is called the transient solution to the IVP,
t→∞
and yS is called the steady-state solution.

C. Sprouse —Math 280 Notes Page 202 DRAFT August 26, 2022
2.7. Oscillations: Amplitude, Phase, and Transients 203

The transient solution depends on the given ODE and initial values, while the steady-
state solution only depends on the ODE.

Example 2.7.6. Find the transient and steady-state solution to

y ′′ + 4y ′ + 5y = 8 sin(t)
y(0) = 1
y (0) = 0
′

−2t −2t
solution: The solution to the IVP is y = 3e cos(t) + 2e sin(t) + cos(t) − sin(t). We
can see that limt→∞ 3e cos(t) + 2e sin(t) = 0, and cos(t) − sin(t) = 2 cos(t + π4 ) is
−2t −2t

−2t −2t
periodic. Therefore yT = 3e cos(t) + 2e sin(t) and yS = cos(t) − sin(t) □.

Note that the transient solution yT in the above example only contained terms from the
homogeneous solution of the given ODE. The steady-state solution yS contained no such
terms. This is not a coincidence. It will always be the case for a constant-coefficient ODE
with nice forcing function. More specifically, if y has a steady-state and transient solu-
tion, then yT will contain only terms of the homogeneous solution, and yS will be the
particular solution that comes from undetermined coefficients. Recall that when doing un-
determined coefficients, the particular solution yp can never contain terms that are in the
homogeneous solution. So, in general

Consider the constant-coefficient ODE

(n)
+ ... + a1 y + a0 y = f (t).
′
an y

If all roots of the characteristic equation have negative real parts, and f (t) is a sum of
sine and cosine terms, then for any initial value problem, the steady-state solution yS (t) is
precisely the same as the particular solution derived by undetermined coefficients.

It is for this reason that sometimes the terminology is slightly abused. If one has that
all roots of the characteristic equation have negative real part, then the homogeneous so-
lution yh = c1 y1 (t) +... + cn yn (t) satisfies lim yh (t) = 0 for any choice of constants c1 , ..., cn .
t→∞
So once the constants c1 , ..., cn are determined to solve an initial value problem, that part
of the solution is still referred to as transient, even if f (t) and the particular solution yp
from undetermined coefficients are not periodic.
Variation of parameters does not give the steady-state solution for yp (t), however. In
particular the forced solution yF (t) that one gets from variation of parameters satisfies
yF (t0 ) = yF (t0) = 0. This means yF definitely does depend on initial conditions, whereas
′

yS must be independent of initial conditions.

′′ ′
Example 2.7.7. Consider the ODE y + 4y + 3y = 6 + 10 sin(t).
−t −3t
solution: The homogeneous solution is yh = c1 e + c2 e and by undetermined
coefficients the particular solution then takes the form yp = A + Bcos(t) + C sin(t).
Plugging in gives A = 2,B = −2, C = 1. So this gives yp = 2 − 2 cos(t) + sin(t). Since this

C. Sprouse —Math 280 Notes Page 203 DRAFT August 26, 2022
2.7. Oscillations: Amplitude, Phase, and Transients 204

particular solution is periodic, we know it must be the steady-state solution, as

−t −3t
c 1 e + c2 e always decays to zero. So

yS (t) = 2 − 2 cos(t) + sin(t).

On the other hand we can compute the forced solution yF (t) by either using variation of
′′ ′
parameters with definite integrals, or simply solving y + 4y + 3y = 6 + 10 sin(t) with
initial values y(0) = y (0) = 0. The solution is
′

yF (t) = − e + e + 2 − 2cos(t) + sin(t).

1 −t 1 −3t
2 2
So we can see that yF (t) contains transients, and is not the same as yS (t) □.

We can use this to graph the solutions to IVP’s. The steady-state solution is a periodic
function, which is often reasonable to graph. In the examples here, we can graph yS given
that we have just learned how to graph A cos(ωt − φ).
Example 2.7.8. Graph the solution to the IVP
′′ ′
y + 2y + y = 2 sin(t),
y(0) = 2,
y (0) = 3.
′

solution: By undetermined coefficients, the solution to the IVP is

y = 3e −t + 6te −t − cos(t). So the steady-state solution is yS = − cos(t) and the transient
−t −t
solution is yT = 3e + 6te . We can check below that y satisfies the initial conditions,
and y → −cos(t) as t → ∞.

y −t −t
y = 3e + 6te − cos(t)
2 yS = − cos(t)
1
t
π 3π 5π 7π
−1 2 2 2 2
−2

The above technique is how we can graph the solutions to many IVP’s.

Suppose any initial value problem has solution y(t) = yT (t)+yS (t), where yS (t) is a steady-
state solution. Then to graph y(t),

1. First graph the steady-state solution yS (t).

2. Graph the part of y(t) around t = t0 , by making sure that y satisfies the initial condi-
tions there.

3. Graph the rest of y(t) by using that y(t) → yS (t) as t increases.

C. Sprouse —Math 280 Notes Page 204 DRAFT August 26, 2022
2.7. Oscillations: Amplitude, Phase, and Transients 205

Note that the transient solution yT (t) is never used in the above. The specific form
of yT (t) is whatever is needed to make sure that y satisfies the initial conditions. But we
can ensure that y satisfies the initial conditions on the graph without ever knowing the
explicit form of yT .
Example 2.7.9. Graph the solution to the IVP
′′ ′
y + 3y + 2y = 10 cos(t),
y(0) = 0,
y (0) = 0.
′

solution: By undetermined coefficients, the solution to the IVP is

√ + 3 sin(t). So the steady-state solution is yS = cos(t) + 3sin(t).
−t −2t
y = −5e + 4e + cos(t)
This has amplitude A = 12 + 33 = 10 and phase angle arctan(3) ≈ 1.25 and the
−t −2t
transient solution is yT = −5e + 4e .

y −t −2t
y = −15e + 9e + 10 cos(t − 1.25)
5 yS = 10cos(t − 1.25)
10

1 t
-1 π 3π 5π 7π
2 2 2 2

10
-5

In the above examples we have only seen how to find the steady-state solution when the
forcing function and steady-state are simple sinusoids. More general examples could be
obtained by having a forcing function that is a sum of sinusoids of varying frequencies,
such as f (t) = 3 sin(t) − sin(2t). In fact, a very large class of periodic forcing functions
f (t) can be represented by infinite sums of sins and cosines (a.k.a. Fourier series). For
example, here is an infinite sum of sin’s that turns out to be a sawtooth wave:
(−1)
∞ n+1
f (t) = π
2
n sin(nπt).
n=1

4
y
2
t
5 10 15 20 25 30
−2

−4
With all such forcing functions, we can find a formula for the steady-state solution by
undetermined coefficients.

C. Sprouse —Math 280 Notes Page 205 DRAFT August 26, 2022
2.7. Oscillations: Amplitude, Phase, and Transients 206

Example 2.7.10. Find the steady-state solution to y + 10y + 25y = 100f (t), where f (t) =
′′ ′

(−1)n+1
∞
2∑ n
sin(nt).
n=1

solution:
We plug y = A cos(nt) + B sin(nt) into y′′ + 10y ′ + 25y = 200 (−1)
n

n
sin(nt). Rearranging this
gives

(−1)
n
((25 − n )A + 10B) cos(nt) + (−10A + (25 − n )B) sin(nt) = 200
2 2
n sin(nt),
so
(25 − n )A + 10B
2
= 0
(−1)
−10A + (25 − n )B
n
2
= 200 n
.

200(25−n )
n 2 2
−2000(−1)
This can be solved to get A = n((25−n2 )2 +100)
, B= n((25−n2 )2 +100)
. So

∞
−2000(−1)n 200(25 − n2 )2
yS (t) =
n((25 − n2 )2 + 100) n((25 − n2 )2 + 100)
cos(nt) + sin(nt) □ .
n=1

4
y
2
t
5 10 15 20 25 30
−2

−4

Note that the formula above actually doesn’t give much insight into why the graph of
yS (t) looks like it does. Here, we simply numerically solved for it. There is a way of
understanding what the solution looks like, which comes from the Chapter on Laplace
transforms. There we will see that the forced solution is a convolution of the forcing
function with the impulse response, which should look just like the function above.
Finally, we have discussed steady-state solutions in the context of constant-coefficient
ODE’s with periodic forcing functions. However the concept of steady-state solution ap-
plies to any ODE, linear or nonlinear, homogeneous or not.

Example 2.7.11.
y + 5(1 − y )y + y = 0,
′′ 2 ′

y(0) = 4
y (0) = 0
′

solution: This is a Van der Pol equation, which is the standard example of a type of
nonlinear oscillation important in electrical and biological applications. We cannot
derive an analytic formula for the solution, but the graph is indicated below. The solution

C. Sprouse —Math 280 Notes Page 206 DRAFT August 26, 2022
2.7. Oscillations: Amplitude, Phase, and Transients 207

y(t) to the IVP converges to a periodic solution yS (t). So there is a periodic solution
yS (t), with lim y(t) = yS (t). Or, we can define yT (t) = y(t) − yS (t) and we can see that
t→∞

y(t) = yT (t) + yS (t),

where yS is periodic and lim yT (t) = 0. Hence yS is a steady-state solution as we have

t→∞
defined above □.
4 y yS (t)
y(t)
2

t
10 20 30 40 50 60

−2

Notice that the solution y to the IVP moves very slowly towards the steady-state solu-
tion yS at first. But as soon as y(t) gets close to yS (t), it locks on and follows the steady-
state very closely. This can be seen as a general feature of Lienhard systems, of which the
Van der Pol equation is the standard example. The corresponding steady-state solution is
called a relaxation oscillation.

C. Sprouse —Math 280 Notes Page 207 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 208

2.8 Applications of Linear Second Order ODE’s

Here we will discuss the standard simple applications of second order ODE’s: Springs
and Circuits. Although this class is more about solving the equations than deriving them
from physics, we’ll try to present the derivations in a way that is simple enough to be
worth reading.

2.8.1 Springs
Consider an object attached to a spring, which is attached at the other end to something
like a wall or ceiling. We want to consider how the mass moves subject to the forces on it.
Since we’ll usually be neglecting gravity, we will draw the spring horizontally.

Equilibrium
x=0

m Unstretched

F
m Stretched

The equilibrium position of the mass is the point where the mass sits when no exter-
nal forces are acting on the mass/spring. The spring is unstretched at equilibrium, so is
resting at its natural length. At any time t, we define x(t) to be the distance of the mass
from equilibrium. So x = 0 when the mass is at equilibrium. Since x is position, we know
2 2
the velocity of the mass is dx/dt and its acceleration is d x/dt . Hence Newton’s second
law becomes
2
d x
F = ma ⇒ F = m 2 .
dt
Now we need to determine the forces acting on the object. They add together to give the
total force F.

1. Foremost there will be the stiffness of the spring. This is the tendency for a stretched
spring to pull back towards its natural length. Hence the mass will always be pulled
back towards equilibrium. This is usually encountered in calculus in the form of
Hooke’s Law, which says that the force is proportional to distance from equilibrium,
x. That is, we have a force of −kx on the object, where k is the spring constant.

C. Sprouse —Math 280 Notes Page 208 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 209

2. Next there will be internal friction in the spring. This is different from stiffness.
Recall that friction is a force that resists motion. Assuming that friction is directly
proportional to the motion, we have a force of −Rdx/dt, where R is a constant.
Friction in the context of ODE’s is also called damping.

3. Finally, there may be external forces acting on the spring, such as something pulling
on the mass. A more interesting model might have F = F0 sin(γt) for the effect of
winds blowing the mass over time. In this way the spring could represent an ab-
stract model for a different physical structure, such as a bridge. For another exam-
ple, the ball could represent the position of the axis/wheel of a moving car, and the
spring is just a simplified model of any of the components that provide a restoring
force.
Using Newton’s second law and adding the three forces above, we have

d 2x
− kx(t) + f (t).
dx
F =m = −R
dt2 dt
Or in a more familiar form,

+ kx = f (t).
2
m ddt2x + R dx
dt

Example 2.8.1. Suppose a ball of mass m is attached to a spring with stiffness k = 100, and
resistance 40 times the velocity.
1. Solve for the mass of the ball such that the motion is critically damped.
2. If the motion is critically damped, solve for the position of the ball, if the ball is initially
stretched 1 meter past equilibrium, and moving at a rate of 2 m/sec away from equilib-
rium.
solution:
′′ ′ b 2
1. Recall that x + bx + cx = 0 is critically damped when c = 2
. So

′′ ′ ′′ 40 ′ 100
mx + 40x + 100x = 0 ⇒ x + m x + m x=0

means the spring is critically damped when

2
100 20
m = m ⇒ 100m = 400 ⇒ m = 4.

2. So the motion is critically damped when

′′ ′ ′′ ′
4x + 40x + 100x = 0 ⇒ x + 10x + 25x = 0.

The characteristic equation factors as (s + 5) = 0 so the general solution is

2

x = c1 e + c2 te . The initial values are x(0) = 1, x (0) = 2. Solving for the ci ’s

−5t −5t ′
−5t −5t
gives x = e + 7te □.

C. Sprouse —Math 280 Notes Page 209 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 210

1.5 x −5t −5t

x(t) = e + 7te

0.5

t
0.5 1 1.5 2 □.

Example 2.8.2. Suppose a ball of mass 2-kg is attached to a spring of with spring constant
k = 100 and resistance of 4-times the velocity. The spring is attached to a ceiling, and gravity
pulls on the mass with a force of F = mg ≈ 20N . Solve for the motion of the ball, if it passes
equilibrium at an initial velocity of 3 m/sec. Is the motion underdamped, critically damped, or
overdamped? Sketch the solution.

2 kg

20 N

solution: The IVP to solve is

′′ ′
2x + 4x + 100x = 20,
x(0) = 0,
x (0)
′
= 3.
′′ ′
Dividing by the leading coefficient, x + 2x + 50x = 10. The roots of the characteristic
−t −t
equation are −1 ± 7i so xh = c1 e cos(7t) + c2e sin(7t) and xp = A ⇒ xp = .2 by
undetermined coefficients. So
−t −t
x(t) = c1 e cos(7t) + c2 e sin(7t) + .2,

x (t) = −c1 e
′ −t −t −t −t
cos(7t) − 7c1 e sin(7t) − c2 e sin(7t) + 7c2 e cos(7t).
Plugging in the initial conditions gives c1 + 0c2 + .2 = 0 ⇒ c1 = −.2,
−c1 + 7c2 = 3 ⇒ c2 = .4. So
−t −t
x(t) = −.2e cos(7t) + .4e sin(7t) + .2 .

C. Sprouse —Math 280 Notes Page 210 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 211

When sketching a solution like this, you must make sure the initial conditions are
satisfied. We can see in the picture below that x(0) = 0, x′ (0) = 3. We also have that
lim x(t) = .2. Since the roots of the characteristic equation are complex, the motion is
t→∞
underdamped. So x(t) oscillates as x → .2□.

x −t −t
0.6 x(t) = −.2e cos(7t) + .4e sin(7t) + .2

0.4

0.2

t
1 2 3 4 5 □.

Notice that here the position of the mass settles on +.2 rather than −.2, even though
the ball is pulled down. This is because when hanging from the ceiling, the spring is
actually upside-down. So pulling the ball downward increases the length of the spring.
Again, x(t) is not the position of the ball, but how far the string is stretched past the
equilibrium length.
Recall that in the language of the previous section, the solution x(t) above goes to a
steady state as t → ∞. More precisely, from Definition 2.7.5 we have that if x(t) = xT (t) +
xS (t) where xS (t) is periodic and limt→∞ xT (t) = 0, then xT (t) is called the transient
solution, and xS (t) is called the steady-state solution. So in the above example we have

xT (t) = −.2e xS (t) = .2

−t −t
cos(7t) + .4e sin(7t) ,

To better describe the steady-state solution, we should do an example where the forc-
ing function f (t) is sinusoidal. For a literal spring and ball as above, it is not clear what
force could be pulling the ball in this way, but there are many practical applications for
which this is a basic abstract model. For instance the spring could represent an auto-
mobile suspension, or one of many structural components in a bridge which supplies a
restorative force.
Suppose for instance a bumpy road has ≈ 10 meters between consecutive humps. Then
for an approximation of the form sin(ωr), the frequency is

2π 2π
ω= = ≈ .63.
T 10

So the height varies like y ∼ sin(.63r), where r is the distance along the road in feet.
(Again, this is not to be taken as an actual road, but instead a useful abstraction.)

C. Sprouse —Math 280 Notes Page 211 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 212

Mass

70 mph
10 m

If the car is moving at 70 mph = 31.3 m/sec, then R ≈ 31.3t, giving roughly y ∼
sin(20t).
Example 2.8.3. Suppose a vehicle has mass 1000-kg, and four wheels. We model the suspen-
sion at each wheel by a mass of 250-kg attached to a spring with stiffness 24000 N/m, and
damping force of 5000 N-sec/m times the velocity. Suppose x(t) is the displacement from equi-
librium of the mass above a single wheel. Find x(t) if the force acting on the suspension above
the wheel is f (t) = 500 sin(20t) N, and x(0) = 0, x ′ (0) = 0. What are the transient and
steady-state solutions?

solution: The above gives

′′ ′
250x + 5000x + 24000x = 500 sin(20t).

So we need to solve the initial value problem

′′ ′
x + 20x + 96x = 2 sin(20t).

The characteristic equation factors as s 2 + 20s + 96 = (s + 12)(s + 8), so

xh (t) = c1 e
−12t −8t
+ c2 e . The particular solution is xp = A cos(20t) + Bsin(20t), and
plugging in gives

(−304A + 400B) cos(20t) + (−400A − 304B) sin(20t) = 2sin(20t),

from which we get A ≈ −.00317, B ≈ −.00241. So

−12t −8t
x(t) = c1 e + c2 e − .00317cos(20t) − .00241 sin(20t).

The initial conditions give

x(0) = 0 ⇒ c1 + c2 = .00317,
x (0) = 0 ⇒
′
−12c1 − 8c2 = .0482,

and then c1 = −.0184,c2 = .0216. So

−12t −8t
x(t) = −.0184e + .0216e − .00317 cos(20t) − .00241 sin(20t).

Or,

xT (t) = −.0184e xS (t) = −.00317 cos(20t) − .00241 sin(20t).

−12t −8t
+ .0216e ,

C. Sprouse —Math 280 Notes Page 212 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 213

1 x (cm) x(t)

0.5

t
0.5 1 1.5

−0.5

So the maximum displacement from equilibrium is about .7 cm, which is probably ac-
ceptable. Here is the decomposition into the transient and steady-state solutions:

1 x (cm) xT (t)
xS (t)
0.5

t
0.5 1 1.5

−0.5

Remark 2.8.4. The above is a bit simplified, but not too far from the basic model of a four-
wheel automobile suspension, called the quarter-car model. The difference is that the actual
model typically uses two springs in sequence, much like Example 4.2.6 from Chapter 5. The
second spring is the tire, which of course also has a restorative force! There are also models
which consider two or four seperate wheels. Again, this will result in a system, as in Chapter 5.

2.8.2 RLC Circuits

R L

Vin + C
− I

We consider a closed electrical circuit with the following elements in series: A voltage
source Vin (t), a resistor with resistance R, an inductor with inductance L, and a capacitor
with capacitance C. We will see that its behavior is closely related to the just considered

C. Sprouse —Math 280 Notes Page 213 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 214

spring-mass system. This is called a series RLC circuit. It looks like the RL and RC circuits
considered in the first chapter, and we have the same voltage laws across the components:

1. VR = IR

2. VL = L dI
dt
dVC
3. I = C dt

Vin is the voltage of the source, VR the voltage across the resistor, VL the voltage across
the inductor, and VC the voltage across the capacitor. Then Kirchoff’s voltage law says
that the sum of the voltages around the circuit is 0, which gives

−Vin + VR + VL + VC ⇒ VR + VL + VC = Vin .

(Note that this is because of the way the + and − signs are placed on the circuit indicating
in which direction the voltages are measured. If the + and − signs on Vin in the diagram
were reversed, one would have VL + VR + VC = −Vin .
Now the current through each element is the same. Therefore we can use the voltage-
current relationships to try to get an equation only in terms of current and the source
voltage.

dI
IR + L + VC = Vin .
dt

There are several ways to get a second-order ODE out of this, depending on what
we want to solve for. We need to use the relationship between VC and I to get an ODE:
dV
I = C d tC . For instance if we differentiate the above equation, we get
2
dI d I dVC dVin
R +L 2 + = .
dt dt dt dt
We can use the fact that dVC /dt = (1/C)I from above and rearrange to get

2
d I dI 1 dVin
L 2
+R + I = .
dt dt C dt

So this gives a second-order ODE for the current.

If we want to solve for the voltage across the capacitor instead, let V = VC so I = C dV
dt
.
Then
2
dI dV d V
IR + L + VC = Vin ⇒ C + LC 2 + V = Vin .
dt dt dt
Rearranging and dividing by C gives

C. Sprouse —Math 280 Notes Page 214 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 215

d 2V dV 1 1
L 2 +R + V = Vin .
dt dt C C

Finally, sometimes one might be interested in the charge on the capacitor. The current
dV
is the rate of change of this charge, and we already know the relationship I = C dtC . In
terms of the charge, this is dQ
dt
C
= C dV
dt
C
. This is in fact the derivative of the equation
QC = CVC , which is actually the definition of capacitance. So again, starting with the
equation given by summing voltages around the loop,
2
dI dQ d Q 1
IR + L + VC = Vin ⇒ R + L 2 + Q = Vin .
dt dt dt C
Or,

2
d Q dQ 1
L 2 +R + Q = Vin .
dt dt C

So we have three slightly different ODE’s, depending on what we want to solve for:

1. Current:
2
d I dI 1 dVin
L 2
+R + I =
dt dt C dt

2. Capacitor Voltage, V = VC :

d 2V dV 1 1
L 2
+R + V = Vin
dt dt C C

3. Capacitor Charge, Q = QC :
2
d Q dQ 1
L 2
+R + Q = Vin
dt dt C

Note that the coefficients on the left-hand side are always the same. These indicate
how the components of the circuit affect the output. Since electrical circuits are some-
what less tangible than than a mechanical spring, we can make an analogy with the
spring equation to understand the role of inductance, resistance, and capacitance in a
circuit. (This goes back to Maxwell.)

Electrical Mechanical
R Resistance R Damping
L Inductance m Mass
1
C
Elastance k Stiffness
1
C Capacitance k
Flexibility

C. Sprouse —Math 280 Notes Page 215 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 216

1
The inverse of Stiffness, k
is also called Compliance.

Example 2.8.5. A problem with inductance is that if a circuit is broken so that current instan-
taneously goes to zero, then VL = L dd It means that VL would become infinite. This isn’t possible,
so often results in electrical arcing across the circuit break, as the current attempts to maintain
itself. To prevent this, one can put a capacitor across the switch as in the circuit below. This is
called a snubber, and such capacitors are common across switches in high power circuits.

50

240 V + 100mH
− I
250 µ F

This circuit represents a DC motor, which operates at a constant 4.8 Amps when the switch
is closed. Suppose the switch is suddenly opened, so that I (0) = 4.8, I (0) = 0. Solve for the
′

current I (t). What happens with the voltage across the inductor?

solution: Using 100mH= .1H and 250µF= .00025F, the current equation is

2
I = (240),
d I dI 1 d
.1 + 50 +
dt dt .00025 dt
or
d 2I dI
+ 500 + 40000I = 0.
dt dt
The characteristic equation factors as

s + 500s + 40000 = (s + 400)(s + 100),

2

so
−400t −100t
I(t) = c1 e + c2e .
The initial conditions are

I (0) = c1 + c2 = 4.8,
I (0) =
′
−400c1 − 100c2 = 0,

which gives c1 = −1.6, c2 = 6.4. Note that the voltage across the inductor is
′
dI −400t −100t
VL = L = .1 − 1.6e + 6.4e
dt
−400t −100t
= 64e − 64e .

C. Sprouse —Math 280 Notes Page 216 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 217

4.8 Amp I (t)

3.6 Amp VL (t)

2.4 Amp

1.2 Amp t (msec)

5 10 15 20 25
-10 V

-20 V

-30 V

If we set VL (t) = 0 we can get that the maximum (negative) voltage across the induc-
′

tor happens at t = ln(2)/150 ≈ .0046, this gives −30.1 volts, which we can see in the
above graph. That could easily be handled by the inductor, so this is an effective snubber
capacitor.

Example 2.8.6. An RLC Circuit has L = .2 H, R = 4 , and C = .001 F. Suppose that the
current and voltage through the capacitor are initially zero, and a voltage of Vin = 5cos(100t)
is applied. Find the voltage VC (t) across the capacitor, including the transient and steady-state
solution.

4 .2H

5 cos(100t) + .001 F
− I

solution: The RLC equation for the voltage V = VC (t) gives

2
d V dV
.2 2
+4 + 1000VC = 5000 cos(100t),
dt dt
or
2
d V dV
+ 20 + 5000VC = 25000cos(100t).
dt 2 dt
The characteristic equation s 2 + 20s + 5000 has complex roots

−20 ± 400 − 20000 −20 ± 140i

s1, s2 = = = −10 ± 70i.
2 2
The particular solution by undetermined coefficients is

yp = Acos(100t) + B sin(100t),

which plugging into the ODE gives

(−5000A + 2000B)cos(100t) + (−2000A − 5000B) sin(100t) = 25000 cos(100t).

C. Sprouse —Math 280 Notes Page 217 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 218

So after solving for A and B we get

yp = −4.31 cos(100t) + 1.72 sin(100t).

So
−10t −10t
y = c1 e cos(70t) + c2 e sin(70t) − 4.31 cos(100t) + 1.72 sin(100t).
The initial conditions are VC (0) = 0 and I (0) = 0. But since I = C
dVC
, we have that
VC′ (0) = 0. So
dt

V (0) = 0 ⇒ c1 = 4.31,
V (0) = 0 ⇒
′ ,
−10c1 + 70c2 = −172 ⇒ c2 = −1.84.

and

V (t) = 4.31e
−10t −10t
cos(70t) − 1.84e sin(70t) − 4.31cos(100t) + 1.72 sin(100t).

V (t)

t (msec)

100 200 300 400 500

−5

The transient and steady-state solutions are

VT (t) = 4.31e VS (t) = −4.31 cos(100t)+1.72 sin(100t).

−10t −10t
cos(70t)+1.72e sin(70t) ,

VS (t)
VT (t)
5

t (msec)

100 200 300 400 500

−5

Example 2.8.7. For the RLC circuit in the previous example, find the current I (t).

C. Sprouse —Math 280 Notes Page 218 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 219

dVC
solution: There are two ways to do this. One is to use I = C . Since we already solved
for VC (t) above, we have
dt

′
d −10t −10t
I = .001 4.31e cos(70t) − 1.84e sin(70t) − 4.31 cos(100t) + 1.72sin(100t)
dt
−10t −10t
= .172 cos(100t) + .431 sin(100t) − .172e cos(70t) − .283e sin(70t).

The other way would be to go back and solve the RLC equation for current:
2
d I dI d
.2 2
+ 4 + 1000I = 5cos(100t) .
dt dt dt

But we need to find initial conditions for this! The condition I (0) = 0 is obvious, as it was
already given in the previous example. To get I (0) we need to use VL = L ddtI . This actually
′

takes summing voltages across the circuit. Since I (0) = 0, VR (0) = 4(0) = 0. And
VC (0) = 0 again as given in the previous example. This leaves

−Vin (0) + ✘ (0)

VR✘✘✿+0V (0) + V ✘
✘
L ✘C (0)
✘✿=00
✘ ⇒ VL (0) = Vin (0) = 5.

(0), or I (0) = 25. So the IVP for I would be

′
So 5 = .2 dI
dt

I (0) = 0
2
d I
dt2
+ 20 dI
dt
+ 5000I = −2500 sin(100t),
I (0)
′
= 25.
The solution is again
−10t −10t
I = .172 cos(100t) + .431sin(100t) − .172e cos(70t) − .283e sin(70t),

which is graphed below.

I (t)

0.5

t (msec)

100 200 300 400 500

−0.5

To get more detail about the steady-state solution (and to graph it by hand), we can
also find the amplitude and phase angle, as in the previous section.

C. Sprouse —Math 280 Notes Page 219 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 220

100 4H

−68 cos(5t) + I .0025F

−

Example 2.8.8. An RLC circuit has L = 4H, R = 100 , and C = .0025F. Suppose that at
time t = 0 an input voltage Vin = −68 cos(5t) is applied. Assume I (0) = I ′(0) = 0.

1. Find the amplitude and phase angle of the steady-state current.

2. Find the current which satisfies the initial values. What are the transients?

3. Find the voltage across the inductor (without solving another ODE). What is the transient
and steady-state voltage?

solution:
dVin
1. 1/C = 400 and dt
= 340sin(5t) so the equation we need to solve is

′′ ′
4I + 100I + 400I = 340 sin(5t).

Dividing by the leading coefficient, I ′′ + 25I ′ + 100I = 85sin(5t). The homogeneous

−5t −20t
solution is Ih = c1 e + C2 e and the particular solution by undetermined
coefficients is Ih = A cos(5t) + B sin(5t) = −.5 cos(5t) + .3 sin(5t). So

A= (−.5)2 + (.3)2 = .34 ≈ .583.

.3
φ = arctan + π ≈ −.54 + π ≈ 2.60.
−.5

So
IS = −.5 cos(5t) + .3 sin(5t) = .583cos(5t − 2.60).

−5t −20t
2. The general solution for the current is I(t) = c1 e + c2 e − .5 cos(5t) + .3 sin(5t),
with I (t) = −5c1e
′ −5t −20t
− 20c2e + 2.5 sin(5t) + 1.5 cos(5t). So the initial conditions
give c1 + c2 − .5 = 0, −5c1 − 20c2 + 1.5 = 0. Solving gives c1 ≈ .57, c2 ≈ −.07.
3. Finally, to get VL we do not solve another differential equation, but rather use
VL = L dI
dt
.

d −5t −20t −5t −20t

VL = 4 −.57e + .07e = .583 cos(5t − 2.60) = −11.4e +5.6e −2.9 sin(5t−2.60).
dt
−5t −20t
The transients are VL,T = 11.4e − 5.6e .

C. Sprouse —Math 280 Notes Page 220 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 221

I I (t)
1 IS (t)

t
0.5 1 1.5 2 2.5 3 3.5

−1
The voltage across the inductor is out of phase with the current:
VL VL (t)
VL,S (t)
10

t
0.5 1 1.5 2 2.5 3 3.5

−10

C. Sprouse —Math 280 Notes Page 221 DRAFT August 26, 2022
 2.8. Applications of Linear Second Order ODE’s 222

Homework 2.8:

2
+ kx = f (t)
d x dx
Spring ∶ m 2
+R
dt dt
2
d I dI 1 dVi n dI
RLC current ∶ L 2
+R + I= , VL = L
dt dt C dt dt
2
RLC voltage VC (t) ∶ L
d VC dVC 1 1 dVC
2
+R + VC = Vin , I=C
dt dt C C dt

A) Solve the word problems.

1. A 3 kg mass is attached to a spring with spring constant k = 9 N/m. Suppose the spring has resistance
of 12 times the velocity.
a) Solve for the position of the spring if the initial position of the mass is at equilibrium and the
initial velocity is 6-m/s.
b) Sketch the solution.
2. A 2 kg mass is attached to a spring with spring constant k = 4 N/m. Suppose the spring has resistance
of 6 times the velocity.
a) Solve for the position of the mass if the initial position of the mass is 1 meter below equilibrium
and the initial velocity is3-m/s towards equilibrium.
b) From your solution above, find the time at which the mass reaches maximum displacement from
equilibrium, and the value of the displacement.
c) Use the above to sketch the position as a function of time.
3. A mass of 1 kg is attached to a spring with resistance of 6 times the velocity has spring constant k.
a) Find the motion if k = 8. Use x(0) = 2, x (0) = 0.
′

b) Sketch it.
c) Find the motion if k = 9. Use x(0) = 3, x′ (0) = 0.
d) Sketch it.
e) Find the motion if k = 10. Use x(0) = 3, x (0) = 0.
′

f) Sketch it.
4. A mass of 4 kg is attached to a spring with resistance of 40 times the velocity.
a) Find the value of the spring constant k so that the spring is critically damped.
b) Find the motion if k = 84. Use x(0) = 0, x (0) = 1.
′

c) Sketch it.
d) Find the motion if k = 116. Use x(0) = 0, x (0) = 1.
′

e) Sketch it.
5. For a 1 kilogram mass on a spring with spring constant k = 4 N/m and resistance of 4 times the
velocity, suppose there is an applied force of f (t) = 8 cos(2t) N.
a) If the mass begins at 2 meters below equilibrium, with no initial velocity, solve for the position
of the mass, x(t).
b) Identify the transient and steady state solutions.
c) Sketch the steady-state solution.
d) Sketch x(t), the solution to the initial value problem from part a).
6. For a 2 kilogram mass on a spring with spring constant k = 10 N/m, and resistance of 8 times the
velocity, suppose there is an applied force of f (t) = 32 sin(t) N.

C. Sprouse —Math 280 Notes Page 222 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 223

a) Solve for the position of the spring x(t) if x is initially at 5 meters beyond equilibrium, moving
away at a speed of 3 m/sec.
b) Find the transient and steady-state solutions.
c) Find the amplitude and phase of the steady state solution.
d) Sketch the transient and steady-state solutions.
e) Sketch the solution to the initial value problem.
7. The capacitor is charged at VC (0) = 8 V and VC (0) = 0 when the switch is closed.
′

4 0.2 H

+
6V I .04 F
−

a) Find the voltage across the capacitor, VC (t).

b) Find the transient and steady-state solutions.
8. The RLC circuit below is first running at a constant 240 volts, such that I (0) = 4, and I ′ (0) =
0.Suppose that at t = 0, the voltage ramps down, so that Vin = 240 − 120t.
60 4H

+
Vin I .005F
−

a) Solve for the current in the circuit.

b) What is the current when the voltage becomes zero?
c) How much of that current came from the transient solution?
9. The RLC circuit below is turned on at t = 0 such that Vin = 10t, for 0 < t < 6.
30 2H

+
Vin I .01 F
−

a) Find the voltage VC (t) across the capacitor, if VC (0) = 0, VC (0) = 0.

′

b) Use I = C dV
dt
to find the current.
10. A series RLC circuit has L = 5 H and C = .002 F.
a) Find the value of resistance R that would cause the circuit to be critically damped.
b) Find the voltage VC (t) if V (0) = 0, V (0) = 1, given your value of R above.
′

c) Sketch it.
d) Find the voltage in the inductor, given the current above.
e) Sketch it.
11. When the switch is closed, the current is a constant 12 Amps. Suppose the switch is opened at t = 0,
causing the current to instead flow through the capacitor.

C. Sprouse —Math 280 Notes Page 223 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 224

20

+
240 V I .05 H
− CF

a) Find the value of C such that the circuit is critically damped.

Solve for I, using I (0) = 12, and I (0) = 0. (Use C from part a))
′
b)
c) Find the maximum (negative) voltage across the inductor.
12. a) Find the steady-state current if Vin = 100sin(2t) Volts. (You do not need to find the homogeneous
solution to the ODE.)
2 .5 H

+
Vin I .2 F
−

b) Write in the form IS = A cos(ωt + φ).

13. a) Find the steady state voltage across the capacitor if Vin = sin(30t) Volts. (You do not need to find
the homogeneous solution to the ODE.)
2 .1 H

Vi n + .01F
I
−

b) Write in the form VS = A cos(ωt + φ).

14. The series RLC circuit below has an input voltage ofVin (t) = 16 cos(5t) Volts.
8 2H

+
Vin I .1 F
−

a) Solve for the steady-state voltage across the capacitor, VC (t).

b) Find the steady-state current. (This can be done by using I = C dV dt
instead of solving another
ODE. If you want, you can check that these give the same answer.)
15. The series RLC circuit below has an input voltage Vin (t) = 50 cos(10t) Volts.
20 2H

+
Vin I .01 F
−

C. Sprouse —Math 280 Notes Page 224 DRAFT August 26, 2022
2.8. Applications of Linear Second Order ODE’s 225

a) Find the steady-state current.

b) Find the steady-state voltage across the inductor.
16. A series RLC circuit has L = 1 H, R = 6 , and C = .04 F. Suppose there is an input voltage ofVin (t) =
6 cos(5t) Volts.
a) Find the steady-state voltage across the capacitor.
b) Find the transient voltage, if V (0) = 0, V (0) = 85 Volts/sec.
′

17. A series RLC circuit has L = 5 H, R = 10 , and C = .1 F. Suppose there is an input voltage ofVin (t) =
25sin(t) Volts.
a) Find the steady-state current.
b) Find the transient current, if I (0) = 3 Amps, I (0) = 0.
′

18. The switch is closed at time t = 0, with Vin = f (t) an arbitrary input voltage. Find the forced current
in the circuit, as a definite integral.
10 5H

+
Vin I .02 F
−

19. The switch is closed at time t = 0, with Vin = f (t) an arbitrary input voltage.
15 .5 H

+
Vin I .01 F
−

a) Find the forced voltage across the capacitor VF (t), as a definite integral.
b) Using your forced solution VF (t) from above, find the voltage across the capacitor VC (t) if
VC (0) = 12, VC (0) = 0.
′

C. Sprouse —Math 280 Notes Page 225 DRAFT August 26, 2022
2.9. Oscillations II: Frequency Response and Resonance ∗ 240

C. Sprouse —Math 280 Notes Page 240 DRAFT August 26, 2022
Chapter 3

Laplace transforms

3.1 Definitions and Basic Transforms

An integral transform is a specified way of transforming one function into another, by
taking a definite integral. For instance if K (s, t) is a given (fixed) function, then for any
f (t) one can evaluate
f (t)K (s, t)dt.
b
F(s) =
a
Since the variable t is integrated away, the function F takes the new variable s. The (one-
sided) Laplace transform is a specific type of integral transform, which is particularly
suited to solving initial value problems.

Definition 3.1.1. Given a function f ∶ [0, ∞) → R, the

Laplace transform L{f } is the function F given by

L{f }(s) = F(s) = f (t)e

∞
−st
dt.
0

So, for the Laplace transform we have a = 0, b = ∞, and K (s, t) = e . We can compute
−st

some examples to see what F(s) looks like in simple situations. If f (t) = 1 we have
∞ 1 −st ∞
−st
F(s) = e dt = − s e
0 0
1 −st 0
= − s lim e − e
t→∞
0
(s > 0).
1 −∞✯
✟ 1
e ✟ −1 = s
= −s ✟

Remark 3.1.2. Here we will always consider L to be restricted to piecewise-continuous func-

tions, f (t). However even for many of these functions, there can be values of s such that the
improper integral above diverges. For functions like f (t) = e , we will see that this means F(s)
2t

241
3.1. Definitions and Basic Transforms 242

must be left undefined for some values of s. Also, though t is always real, s is often taken to be
complex. In this case, it turns out that when the domain of F is not D = C, it will always be a
“half-plane” ,
D = {s ∈ C∣ Re(s) > a},
or
D = {s ∈ C∣ Re(s) ≥ a}.
Although this domain will be mostly irrelevant to the applications here, we specify it below for
completeness.
1
So in the above we have F(s) = s
for Re(s) > 0, or

(Re(s) > 0)
1
L{1} = s .

For f (t) = t we can integrate by parts to get F(s):

∞
−st
L{t} = te dt
0
−t −st ∞ 1 ∞
−st
= s e +s e dt
0 0
0
t ✁✕ 1 −st ∞
0
= lim − ✁st + 0e − 2 e
t→∞ ✁e s 0
0
(Re(s) > 0)
1 −∞ ✯ 0
✟ 1
= − 2
e ✟ −e = 2.
✟
s s

And we have

(Re(s) > 0)
1
L{t} = .
s2

Similarly, integrating by parts twice gives

∞
L{t } = (Re(s) > 0).
2 2 −st 2
t e dt = ... =
0 s3
And in general, one has:

L{t } = (Re(s) > 0)

n n!
.
sn+1

C. Sprouse —Math 280 Notes Page 242 DRAFT August 26, 2022
3.1. Definitions and Basic Transforms 243

The Laplace transform of an exponential function is particularly simple:

L{e } =
∞
at at −st
e e dt
0
(a−s)t 1 (a−s)t
∞ ∞
= e dt = a − s e
0 0
0
(Re(s) > a)
1 ✯ 0
✟
e ✟ −e
−∞
= a −s ✟
1
= s −a.

L{e } = s − a . (Re(s) > a)

at 1

Now that we have some functions f that we can take L, we can point out that L is a
linear operation:

L{αf ± βg } = αL{f } ± βL{g}.

The proof is straightforward from the definition, so we leave it as an exercise. We will

use this frequently, for instance to compute things like this
Example 3.1.3.

L{3t − e } = 3L{t } − L{e } = 3

2 5t 2 5t 2 1 6 1
− = − .
s3 s − 5 s3 s − 5

Now that we have linearity and L of exponentials, we can also get the transforms of
sine and cosine. The trick is to recall Euler’s identity from the previous chapter,
iθ
e = cos(θ) + i sin(θ).

From this, one can directly verify that

1 it −it
cos(t) = e +e ,
2
1 it −it
sin(t) = e −e .
2i

C. Sprouse —Math 280 Notes Page 243 DRAFT August 26, 2022
3.1. Definitions and Basic Transforms 244

L{e } + L{e }
1 iωt 1 −iωt
L{cos(ωt)} =
2 2
1 1 1
= +
2 s − iω s + iω
1 1 1 1 s + iω + s − iω
2 (s − iω)(s + iω)
= + =
2 s − iω s + iω
1 2s s
= = 2 .
2 s +ω
2 2
s + ω2
Similarly,

L{e } − L{e }
1 iωt 1 −iωt
L{sin(ωt)} =
2i 2i
1 1 1 1 s + iω − (s − iω)
2 (s + iω)(s − iω)
= − =
2i s − iω s + iω
1 2iω ω
= = 2 .
2i s 2 + ω 2 s + ω2
So we have derived

s
L{cos(ωt)} = ,
s + ω2
2

(Re(s) > 0).

ω
L{sin(ωt)} =
s2 + ω2

We can also generalize the above by using

−iωt
1 (a+iω)t 1 (a−iω)t
iωt
at at e +e
e cos(ωt) = e = e + e ,
2 2 2

−iωt
1 (a+iω)t 1 (a−iω)t
iωt
at at e −e
e sin(ωt) = e = e − e .
2i 2i 2i
Then after taking L, we get

at s−a
(s − a)2 + ω2
L{e cos(ωt)} = ,

(Re(s) > a).

at ω
L{e sin(ωt)} =
(s − a)2 + ω2

We leave the details as an exercise. But by now we should be able to see a pattern.
The f (t) for which we have been finding L{f }of are functions that are quite familiar
from the last chapter. That is, f (t) has always been function that could also be a solution

C. Sprouse —Math 280 Notes Page 244 DRAFT August 26, 2022
3.1. Definitions and Basic Transforms 245

to a constant-coefficient homogenous ODE. For instance f (t) = sin(ωt) is a solution to

y ′′ + ω 2 y = 0. When learning the method of undetermined coefficients, these functions
f (t) were called nice. Recall from Section 2.5 that these are just the functions that can
be written as sums or products of exponentials, polynomials, sines, and cosines. We have
seen that for these f , the Laplace transform L{f } can be computed explicitly, directly
from the definition of L . We haven’t computed L{te } so far, but this also just integration
at

by parts:

(a−s)t
∞ ∞
L{te } =
at at −st
te e dt = te dt
0 0
t (a−s)t ∞ 1 ∞
(a−s)t
= a − se + a −s e dt
0 0
0
(a−s)t
∞
1 t ✚✚
❃ 0 1
=
(a − s)
lim − 0e −
e (s−a)t
a−s ✚ 2
e
t→∞ ✚
0
0
(Re(s) > a)
1 ✯ 0
✟ 1
e ✟−e =
−∞
(s − a) (s − a)2
= − 2 ✟ .

And similarly we get

L{t e } = (Re(s) > a)

n at n!
(s − a)n+1
.

As for other functions, we can also get L{f } from the definition whenever f has piece-
wise components consisting of the above functions. For instance

Example 3.1.4. Find L{f }, where

5 u(t) f (t)
4

1 3
f (t) = 2
t 0 ≤ t ≤ 6,
3 t > 6. 2

1
t
2 4 6 8 10

C. Sprouse —Math 280 Notes Page 245 DRAFT August 26, 2022
3.1. Definitions and Basic Transforms 246

solution: From the definition,

∞
L{f } = f (t)e
−st
dt
0
6 1 −st ∞
−st
= te dt + 3e dt
0 2 6
1 −st 6 1 6
3 −st ∞
−st
= − te + e dt − s e
2s 0 2s 0 6
3 −6s 1 −st 6 3 −6s
= −s e − 2e +s e
2s 0
1 −6s 1
= − 2e + 2.
2s 2s
The integral converges when Re(s) > 0 □.

In the future, we usually won’t need to compute L by the definition, as we will already
know L for the functions that we need. This is true even for piecewise-defined functions
like the above, as long as we know the shift properties of L. It takes a little bit of setup to
explain how shifting works, so we will save that until Section 3.3.
But we can still use this section to describe some other important properties of L.
Namely, the derivative and integral properties are easy enough to do right now. These
could have been used to derive several of the transforms above much more quickly, avoid-
ing integration by parts. In fact, from the derivative and shift properties of L alone, we
could derive all of the above Laplace transforms above from just knowing the first one:
L{1} = 1s .

3.1.1 Derivative and Integral Properties of L

Soon we will use the Laplace transform to solve ODE’s. This is possible because of how
L interacts with derivatives. There are actually two derivative properties. One with the
derivative on the right, and the other with the derivative on the left. This will be true
with all of the basic properties of L. For the derivative we have

Derivative Properties of L:

(I) L{tf (t)} = −

dF
,
ds

(II) = sF (s) − f (0).

df
L
dt

✎ ☞
Derivative Property I
✍ ✌
The first property is very straightforward and easy to derive. We can use it to simplify
computations of L.

C. Sprouse —Math 280 Notes Page 246 DRAFT August 26, 2022
3.1. Definitions and Basic Transforms 247

Theorem 3.1.5.
L{tf (t)} = −
dF
.
ds
Proof.
∞
f (t)e
−st
F(s) = dt,
0
so
∞ ∞
f (t)e f (t) (e )dt
dF d −st d −st
= dt =
ds ds 0 0 ds
∞
tf (t)e dt = −L{tf (t)}.
−st
= −
0

Then multiplying by −1 gives the property as written.

Now we can get a different derivation of many of the Laplace transforms we have
already seen. We start with L{1} = 1s . Then

d d 1 1
L{t} = − L{1} = − s = 2.
ds ds s

L{t } = −
d 2 d 1 2
L{t} = − = 3.
ds ds s 2 s
We can keep going, and then it is easier to see what we already have asserted from
before

L{t } =
n n!
.
s n+1

Similarly, if instead of L{1} = 1s , we start with L{e } = 1

at
s−a
, we get

L{te } = − L{e } = −
at d at d 1 1
(s − a)2
= .
ds ds s − a
And, continuing gives an easy derivation of

L{t e } =
n at n!
(s − a)n+1
.

We can also do several others that we haven’t done so far. The most obvious are

d d ω 2ωs
L{t sin(ωt)} = − L{cos(ωt)} = − = 2
(s + ω 2 )2
.
ds ds s 2 + ω 2
2 2
d d s s −ω
(s + ω 2 )2
L{t cos(ωt)} = − L{cos(ωt)} = − 2 2
= 2 .
ds ds s + ω
So

C. Sprouse —Math 280 Notes Page 247 DRAFT August 26, 2022
3.1. Definitions and Basic Transforms 248

s2 − ω2
L{t cos(ωt)} =
(s 2 + ω 2)2
,

2ωs
(s 2 + ω 2)2
L{t sin(ωt)} = .

We will avoid these for the most part, since they are a little too elaborate to deal with
at at
in practice. Similarly one can also get L{te cos(ωt)} and L{te sin(ωt)} in this way, but
again these are a bit much for our applications here.
✎ ☞
Derivative Property II
✍ ✌
The second derivative property is also easy to prove. It is just what falls out of inte-
grating by parts.

Theorem 3.1.6.
L{f } = sL{f } − f (0).
′

Proof.

∞ ∞
f (t)e f (t)e (t)(−se )dt
′ −st −st ∞ st
dt = 0
− f
0 0

= −f (0) + s f (t)e dt = −f (0) + sL{f }.

∞
−st
0

The appearance of the initial value f (0) is actually important. It means we can use L
to solve initial value problems. Also, once we know the property for first derivatives, we
can repeatedly apply it to get similar properties for higher derivatives.

L{f } = s(L{f }) − f (0) = s(sL{f } − f (0)) − f (0)

′′ ′ ′ ′

= s L{f } − sf (0) − f (0).

2 ′

L{f } = s(L{f }) − f (0) = s(s L{f } − sf (0) − f (0)) − f (0)

′′′ ′′ ′′ 2 ′ ′′

= s L{f } − s f (0) − sf (0) − f (0).

3 2 ′ ′′

Summarizing, we get

C. Sprouse —Math 280 Notes Page 248 DRAFT August 26, 2022
3.1. Definitions and Basic Transforms 249

L{f } = sL{f } − f (0)

′

L{f } = s L{f } − sf (0) − f (0)

′′ 2 ′

⋮
(n) (n−1) (n−1)
} = s L(f ) − s f (0) − ... − f (0)
n
L{f
n−k (k−1)
n
= s L(f ) − (0).
n
s f
k=1

This is the property of L that we will use the most often, as it allows us to solve
differential equations.
Example 3.1.7. Solve by Laplace transforms:
′′
y + 4y = 0,
y(0) = 5,
y (0) = 6.
′

solution: Taking L of the ODE,

L{y } + 4L{y} = 0
′′ ′′
L{y + 4y} = L{0} ⇒
5 6
s Y (s) − s✟ ✯− y ′ (0)
✟✟ + 4Y (s) = 0
2 ✟✟ ✯
⇒ y(0) ✟
(s + 4)Y (s) = 5s + 6.
2
⇒
Solving for Y,
Y (s) =
5s + 6 s 2
=5 2 +3 2 .
s2 + 4 s +4 s +4
From the transforms in the previous sections, we can see that a y(t) corresponding to this
would be
y(t) = 5 cos(2t) + 3 sin(2t) □ .

If we rearrange the derivative properties, we can also get integral properties:

Integral Properties of L:

(I) f (t) =
1 ∞
L F(ξ)dξ,
t s

(II) L f (τ)dτ = s F(s).

t 1
0

C. Sprouse —Math 280 Notes Page 249 DRAFT August 26, 2022
3.1. Definitions and Basic Transforms 250

✎ ☞
Integral Property I
✍ ✌
One may notice that the right hand side of I is a little unusual. The reason for this is
that to get an unambiguous definite integral with endpoint s0 , we need to know the value
F(s0 ). Recall that our formulas for F(s) have looked like 1s , s+a
1 s
, s2 +ω 2 , etc. So if s 0 ∈ R,
we can get many different things for F(s0 ). But there is one value common to all of them:
F(∞) = lim F(s) = 0. This is true in general:
s→∞

Fact 3.1.8. If F(s) = L{f } for some f , then

lim F(s) = 0.
s→∞

Proof.
∞
f (t)e
−st
lim F(s) = lim dt
s→∞ s→∞ 0
∞
f (t) lim e
−st
= dt
0 s→∞
∞
= f (t)(0) dt = 0.
0

Note that it is not always possible to put a limit inside a definite integral, as we have done
here. But for this and similar integrals in the present chapter, rest assured it can be done
without worry.

Theorem 3.1.9. ∞
f (t) =
1
L F(ξ)dξ
t s

Proof. Let g(t) = 1t f (t). Applying the corresponding derivative property I to g,

L{f } = −
dG dG
L{tg(t)} = − ⇒
ds ds
dG
⇒ = −F(s).
ds
This says that
F(ξ)dξ + F(s0 ).
s
G(s) = − F(s) ds = −
s0
Again, the only value we know for F(s0) with general F(s) is with s0 = +∞. Then we have
s
G(s) = − F(ξ)dξ + F(∞)
∞
s
= − F(ξ)dξ
∞
∞
= F(ξ) dξ.
s

C. Sprouse —Math 280 Notes Page 250 DRAFT August 26, 2022
3.1. Definitions and Basic Transforms 251

This property is not one of the ones we use very often. But it does allow us to find a
few more Laplace transforms. Here is an interesting one.
sin(t)
Example 3.1.10. Find L t
.

solution:

L{sin(t)}(ξ )dξ
sin(t) ∞
L =
t s
∞ 1
= dξ
s ξ2 + 1
∞ π
= arctan(ξ)∣s = − arctan(s) □ .
2

f (t) =
π π
1.5
sin(t)
2 F(s) = 2
− arctan(s)
t

1 1

0.5 0.5
t s

5 10 15 20 2 4 6

✎ ☞
Integral Property II
✍ ✌
The second integral property is simpler to prove. It is also much more useful, as it
allows us to solve integral equations.
Theorem 3.1.11.
f (τ) dτ = s F(s).
t 1
L
0

Proof. let g = ∫0 f (τ) dτ, so that g (t) = f (t). Then applying Derivative Property II to
t ′

g(t),

L{g } = sG(s) − g(0).

′

Since g(0) = ∫0 f (τ) dτ = 0, this gives L{g } = sG(s). Rewriting,

0 ′

G(s) = s L{g } f (t) dt = s F(s).

1 t 1
′
⇒ L
0

Example 3.1.12. Solve by Laplace transforms:

t
y+2 y(τ) dτ = 1.
0

C. Sprouse —Math 280 Notes Page 251 DRAFT August 26, 2022
3.1. Definitions and Basic Transforms 252

solution: Taking L,
t t 1
L y +2 y(τ) dτ = L{1} ⇒ L y + 2L y(τ)dτ = s
0 0

Y (s) + s Y (s) = s Y (s) = s .

2 1 s+2 1
⇒ ⇒ s

Solving for Y ,
s 1 1
Y(s) = = .
s+2 s s+2
We recall from before that L{e } = 1 1
at
s+a
. So a solution y(t) with L{y} = s+2
is

−2t
y(t) = e □.

−2t
To conclude definitively that y(t) = e in the above would require a fact that we
haven’t stated yet: If L{z(t)} = Y (s) for some z(t) then L{y(t)} = Y(s) implies y(t) = z(t).
That is, the operation L needs to be invertible. That is true, at least when y and z are
continuous. We’ll continue with this in the next section.

C. Sprouse —Math 280 Notes Page 252 DRAFT August 26, 2022
3.1. Definitions and Basic Transforms 253

3.1.2 Table of Common Transforms

As is customary, here is a table of the transforms we have derived.

f (t) L(f )

1
1 s
1
t s2
n!
n
t s n+1

1
at
e s −a
1
(s − a)2
at
te

n!
(s − a)n+1
n at
t e

s
cos(ωt) s + ω2
2

ω
sin(ωt) s2 + ω2
s −a
(s − a)2 + ω 2
at
e cos(ωt)

ω
(s − a)2 + ω 2
at
e sin(ωt)
2 2
s −ω
t cos(ωt) (s 2 + ω 2 )2
2sω
t sin(ωt) (s + ω 2 )2
2

C. Sprouse —Math 280 Notes Page 253 DRAFT August 26, 2022
 3.1. Definitions and Basic Transforms 254

Homework 3.1:
A) Use the definition of Laplace transform to find L{f }. (Do the integrals.) For what values of s is L{f }
defined?
1.
f (t) = te
2t

solution:
∞
2t 2t −st
L te = te e dt
0
(2−s)t
∞
= te dt
0
(2−s)t ∞
(2−s)t
∞
te 1
(2 − s) (2 − s)
= − e dt
0
0
(2−s)t ∞
e (2−s)t
∞
te
(2 − s)
=
(2 − s)2
− .
0 0

The infinite limits above go to 0 when 2 − s is negative, or s > 2. Then

= (0 − 0) − 0 −
2t 1 1
=
(2 − s)2 (s − 2)2
L te .

So
2t 1
= for s > 2 □ .
(s − 2)2
L te

2. 7.
f (t) = 3 f (t) = (2t + 1)e
t

3.
f (t) = e
3t 8.
f (t) = cos(t)
4.
f (t) = t + 2 ∫ e cos(βt) dt =
αt
αt e
α 2 +β2
α cos(βt) + β sin(βt) .
5. 9.
f (t) = t f (t) = sin(ωt)
2

6.
f (t) = te ∫ e sin(βt) dt =
αt
at αt e
α 2 +β2
−β cos(βt) + α sin(βt) .

C. Sprouse —Math 280 Notes Page 254 DRAFT August 26, 2022
 3.1. Definitions and Basic Transforms 255

B) First sketch the function f (t). Then evaluate L{f } using the definition of Laplace transforms. (Do some
integrals.)

⎧
10.
⎪
⎪ 2 0≤t ≤1
f (t) = ⎪
⎨
⎪
⎪
3−t 1≤t ≤3
⎪
⎩ 0 t >3
solution: The integral needs to be done piecewise:
∞
L(f ) = f (t)e
−st
dt
0

(3 − t)e
1 3 ∞
−st −st −st
= 2e dt + dt + 0e dt
0 1 3
2 −st 1 (3 − t) −st 3
1 3
−st
= −s e − s e −s e dt
0 1 1
2 −st 1 (3 − t) −st 3
1 −st
3
= −s e − s e + 2
e
0 1 s 1
2 −s 2 2 −s 1 −3s 1 −s
= −se + s + se + 2e − 2e .
s s
So
L(f ) = s + 2 e
2 1 −3s 1 −s
− se □.
s

11. 18.
f (t) =
0≤t ≤2
f (t) =
3 1 0 ≤t ≤ 2
0 t≥2 e
−t+2
t>2
12. 19.
f (t) =
0 0≤t ≤3
f (t) =
3t 0≤t≤2
2 t≥3
0 t >2

⎧
13.
⎪
20.
⎪
0 ≤t ≤ 2
f (t) = ⎪
−1
⎨
⎪ f (t) =
2 − 2t 0≤t ≤1
⎪
1 2 ≤t ≤ 4
⎪
⎩0 t>4 0 t>1
21.
⎧
14.
⎪
⎪
0 0 ≤t ≤3 f (t) =
t 0≤t≤2
f (t) = ⎪
⎨
⎪
2 t >2
⎪
2 3 ≤t ≤5
⎪
⎩0 t≥5 22.
⎧
⎪
⎪
0≤t ≤1
f (t) = ⎪
0
⎧
15.
⎪ ⎨
⎪
⎪ ⎪
2 0 ≤t ≤1 − 1≤t ≤2
f (t) = ⎪
t 1
⎨ ⎪
⎩
⎪
t>2
⎪
3 1 ≤t ≤3 0
⎪
⎩1 t≥3
23.
⎧
⎪
⎪
16. 0≤t ≤1
⎪
t
f (t) = ⎨
−t
f (t) = ⎪
0≤t ≤2
⎪
e 2 − t t≥1
0 t>2 ⎪
⎩ 0 t≥2
17. 24.
t
f (t) = 3
0 ≤t ≤ 3
f (t) =
e 2
t 0≤t<3
e t>3 9 t ≥3

C. Sprouse —Math 280 Notes Page 255 DRAFT August 26, 2022
 3.1. Definitions and Basic Transforms 256

25. 26.
f (t) =
− cos(t) 0 ≤ t ≤ π
1 t >π

∫ e cos(βt) dt = e αt
αt
2 α cos(βt) + β sin(βt)
f (t) =
25 − t 0≤t ≤5 α 2 +β2
.
0 t >5 27.
f (t) =
sin(πt) 0 ≤ t ≤ 2
0 t >2

∫ e sin(βt) dt =
αt
αt e
α 2 +β2
−β cos(βt) + α sin(βt) .

C) Find L{f } using known transforms. (Don’t do any integrals.)

28.
f (t) = (t + e )(2 + e )
2 t −t

solution: This needs to be multiplied out:

f (t) = (t + e )(3 + e ) = 3t + 3e + t e
2 t −t 2 t 2 −t
+ 1.

So using the standard Laplace transform formulas,

L{f } =
6 3 2 1
s − 1 (s + 1)3 s
3
+ + + □.
s

29. 36.
f (t) = (2t − 1) f (t) = cos (2t)
2 2

30.
f (t) = (1 + te )
−t 2 37.
f (t) = cos(3t)sin(3t)
31.
f (t) = t (t − e )
3 t
38.
32. f (t) = cosh(2t)
f (t) =
2+ t
et 39.
f (t) = 4 sinh (t)
2
33.
f (t) = 5
t
40.
f (t) = t sinh(t)
34.
f (t) = t2
−t

35. 41.
f (t) = 2 sin (t) f (t) = 8 cosh(t)sinh(2t)
2

C. Sprouse —Math 280 Notes Page 256 DRAFT August 26, 2022
3.2. Solving ODE’s by Laplace Transform 257

3.2 Solving ODE’s by Laplace Transform

In practice, the most common method used for solving linear ODE’s by hand is unde-
termined coefficients. Recall that this method applies specifically to constant coefficient
ODE’s with nice forcing functions f (t). In the previous section, we showed how to get the
Laplace transform F(s) for any nice function f (t). This means given a constant-coefficient
ODE for y(t) with forcing function f (t), with the derivative properties of L and a little al-
gebra, we can get Y (s). This allows us to solve any such constant coefficent ODE by
Laplace transforms. Here is the general second order case:

Example 3.2.1. Find Y (s) if y solves the second order IVP

y + by + cy = f (t),
′′ ′

y(0) = b1
y (0) = b2 ,
′

solution: Taking L of both sides,

L{y + by + cy} = L{f }

′′ ′

s Y − sb1 − b2 + b(sY − b1) + cY = F(s)

2
⇒
(s + bs + c)Y = F(s) + sb1 + bb1 + b2
2
⇒

Y (s) = (F(s) + sb1 + bb1 + b2) .

1
⇒ 2
s + bs + c

Then to solve for y(t), we need to find y(t) such that L{y} = Y (s). If one thinks of L
as an invertible transformation, then one would be looking for

y(t) = L {Y (s)}.
−1

For such an idea to work in general, we would need to know that L is in fact invertible,
and in particular there is a unique y(t) with L{y(t)} = Y (s). This is in fact true, if we
restrict the domain of L to be piecewise-continuous functions, and we make the technical
allowance to consider two piecewise continuous functions to be the same if they differ
only at their points of discontinuity. In fact, there is even an explicit formula for L−1 .
If Y(s) is defined on the half-plane, {s ∈ C ∣ Re(s) > a}, then for any α > a we have the
Mellin Inverse Formula:

(Y (s)) = Y (s)e ds.

1 α+iβ
−1 st
L lim
2πi β→∞ α−iβ

This means in the above for instance that the forced solution y(t) with b1 = b2 = 0 is
α+iβ
1 F(s) st
y(t) = lim e ds.
2πi β→∞ α−iβ s 2 + bs + c

C. Sprouse —Math 280 Notes Page 257 DRAFT August 26, 2022
3.2. Solving ODE’s by Laplace Transform 258

Although this is useful for proving certain theorems and general results, it involves
integrating a somewhat complicated function over a line in the the complex plane. So
it isn’t really suitable for deriving explicit solutions to specific ODE’s by hand, and we
won’t make any use of it here. Instead, we will have to look for what y(t) must be given
a specific Y(s), using the Laplace transforms that we already know. If f (t) is nice, the
function Y (s) will always be a rational function:

Y (s) =
p(s)
,
q(s)
where p and q are polynomials. Recall from Calculus that when degree(p) < degree(q),
we can decompose such functions by partial fractions, as a sum of simpler fractions. Then
it will be easy to see what the corresponding y(t) is, given a table of transforms that we
know.
Example 3.2.2.
′′ ′ t
y + 3y + 2y = 6e
y(0) = 2
y (0) = 1.
′

solution: Taking Laplace transforms as above gives

′′ ′ t
L{y + 3y + 2y} = L 6e

L{y } + 3L{y } + 2L{y} = 6L e

′′ ′ t
⇒
2 6
⇒ s Y − 2s − 1 + 3(sY − 2) + 2Y = ,
s −1
2 6
⇒ s Y + 3sY + 2Y = + 2s + 7,
s−1
6 + (s − 1)(2s + 7)
⇒ (s + 3s + 2)Y =
2
.
s−1
Simplifying the right hand side fraction first,
2
(s + 3s + 2)Y =
2 2s + 5s − 1
s−1
2
Y (s) =
2s + 5s − 1
(s 2 + 3s + 2)(s − 1)
.

Factoring completely, we have

2
Y (s) =
2s + 5s − 1
(s + 1)(s + 2)(s − 1)
.

This can be decomposed by the most basic method of partial fractions:

2
2s + 5s − 1 A B C
(s + 1)(s + 2)(s − 1) s + 1 s + 2 s − 1
= + + .

This can be solved by plugging s = −1, −2, and 1 into

2
2s + 5s − 1 = A(s + 2)(s − 1) + B(s + 1)(s − 1) + C(s + 1)(s + 2)

C. Sprouse —Math 280 Notes Page 258 DRAFT August 26, 2022
3.2. Solving ODE’s by Laplace Transform 259

to get A = 2, B = −1, and C = 1. So

Y (s) =
2 −1 1
+ + ,
s +1 s +2 s−1
{Y } is
−1
from which we can easily see that L
−t −2t t
y(t) = 2e −e +e □.

Notice that even a simple ODE such as this can result in a good deal of algebra, such as
adding and simplifying fractions. It is important to get the rational function Y (s) correct,
and also factored correctly, or the partial fractions step will go completely wrong. And
then additionally one must be very well practiced with partial fractions to get y(t) in the
different situations that can come up. For instance, here is an example that is superficially
similar, but leads to a different type of partial fractions.
Example 3.2.3.
′′ t
y − y = 6e ,
y(0) = 3,
y (0) = 2.
′

solution: Taking the Laplace transform of both sides again gives

′′ t
L{y − y} = L 6e
L{y } − L{y} = 6L e
′′ t
⇒
2 6
⇒ s Y − 3s − 2 − Y =
s −1
2 6
⇒ s Y −Y = + 3s + 2
s−1
6 + (s − 1)(3s + 2) 3s − s + 4
2
(s − 1)Y =
2
⇒ = .
s −1 s −1
So
2 2 2
Y (s) =
3s − s + 4 3s − s + 4 3s − s + 4
(s 2 − 1)(s − 1) (s + 1)(s − 1)(s − 1) (s + 1)(s − 1)2
= = .

Again, it is crucial to get the factorization on the right hand side correct here. From this
we can see the partial fractions looks like
2
3s − s + 4 A B C
=
(s + 1)(s − 1) s + 1 (s − 1)
+ + ,
2 2 s−1
and
2 2
3s − s + 4 = A(s − 1) + B(s + 1) + C(s + 1)(s − 1).
Plugging points (or rearranging) gives A = 2, B = 3, C = 1. So
2 3 1
Y(s) =
s + 1 (s − 1)2 s − 1
+ + ,

−t t t
y(t) = 2e + 3te + e .
This gives the same answer as undetermined coefficients, although the terms appear in a
slightly different order □.

C. Sprouse —Math 280 Notes Page 259 DRAFT August 26, 2022
3.2. Solving ODE’s by Laplace Transform 260

Example 3.2.4.
′
y + y = 20 sin(3t),

y(0) = 2.

solution: Taking Laplace transforms gives

3
sY − 2 + Y = 20
s2 + 9
60
⇒ sY + Y = +2
s2 + 9
2
(s + 1)Y = 2
2s + 78
⇒ .
s +9
So
2
Y (s) =
2s + 78
(s + 1)(s 2 + 9)
.

2
In this case, s + 9 does not factor (with real roots). So in this case the partial fractions
looks like
2
2s + 78 A Bs + C
=
(s + 1)(s + 9)
+ 2 ,
2 s + 1 s +9

2s + 78 = A(s + 9) + (Bs + C)(s + 1).

2 2

Solving by plugging points or rearranging gives A = 8, B = −6,C = 6. So

Y (s) = 8
1 s 1
−6 2 +6 2 ,
s+1 s +9 s +9

and
−t
y(t) = 8e − 6 cos(3t) + 2 sin(3t).

−1 1 1
(Recall that L = ω sin(ωt)) □.
s2 + ω 2

Again, you will need to be confident with partial fractions from this point forward. If
you only remember from Calculus how to plug points to get the most basic fractions, it
will not be enough. You must be familiar with both plugging points and rearranging to
be able to do these by hand. See the Appendix, Page 396 for a review of the techniques
that you need to know.

Example 3.2.5.
′′ ′
y + 4y + 5y = 10,

y(0) = 3,

y (0) = 0.
′

C. Sprouse —Math 280 Notes Page 260 DRAFT August 26, 2022
3.2. Solving ODE’s by Laplace Transform 261

solution:

L {10}
′′ ′
L{y + 4y + 5y} =
L{y } + 4L{y } + 5L{y} = 10L {1}
′′ ′
⇒
2
⇒ s Y − 3s + 4(sY − 3) + 5Y = 10L{1}
2 1
⇒ s Y + 4sY + 5Y = 10 s + 3s + 12
2
(s + 4s + 5)Y =
2 10 + 3s + 12s
⇒ s .

So
3s 2 + 12s + 10
Y (s) = .
s(s 2 + 4s + 5)
Recall that s + bs + c = 0 has complex roots when c > (b/2) . Since 5 > (4/2) = 4, we
2 2 2

can’t factor s + 4s + 5. Instead we must complete the square: s + 4s + 5 = (s + 2) + 1.

2 2 2

2
Y (s) =
3s + 12s + 10
⇒ .
s((s + 2)2 + 1)
This means the partial fractions is

3s 2 + 12s + 10 A B(s + 2) C
= s +
(s + 2) + 1 (s + 2)2 + 1
2 2
+ .
s((s + 2) + 1)
To actually do the partial fractions, we rewrite again.
2
3s + 12s + 10 A B(s + 2) C
= s + 2 + 2
s(s 2 + 4s + 5) s + 4s + 5 s + 4s + 5
2 2
⇒ 3s + 12s + 10 = A(s + 4s + 5) + B(s + 2)s + Cs.
This is easy to solve by rearrangement. Multiplying out gives

3s + 12s + 10 = (A + B)s + (4A + 2B + C)s + 5A,

2 2

so
A+B = 3 A = 2,
4A + 2B + C = 12 ⇒ B = 1, .
5A = 10 C = 2.
Then
Y (s) = s +
2 1(s + 2) 2
(s + 2)2 + 1 (s + 2)2 + 1
+ ,

−2t −2t
y(t) = 2 + e cos(t) + 2e sin(t) □ .

Example 3.2.6.
′′ −t
y + 4y = 25te ,
y(0) = 0,
y (0) = 1.
′

C. Sprouse —Math 280 Notes Page 261 DRAFT August 26, 2022
3.2. Solving ODE’s by Laplace Transform 262

solution: Taking L,

L{y } + 4L{y} 25L{te }

′′ −t
=
2 1
s Y − 1 + 4Y = 25
(s + 1)2
⇒

2 25
s Y + 4Y =
(s + 1)2
⇒ +1

25 + (s + 1)
2
(s + 4)Y =
2
(s + 1)2
⇒ .

Simplifying,
2 2
Y (s) =
s + 2s + 26
(s 2 + 4)(s + 1)2
.

The partial fractions is

2
s + 2s + 26 As + B C D
(s + 4)(s + 1) s + 4 (s + 1)
= 2 + +
2 2 2 s+1

s + 2s + 26 = (As + B)(s + 1) + C(s + 4) + D(s + 1)(s + 4).

2 2 2 2
⇒
To find the coefficients here is now quite lengthy. Plugging points:

s = −1 ⇒ 5C = 25,
s=0 ⇒ B + 4C + 4D = 26,
s=1 ⇒ 4A + 4B + 5C + 10D = 29,
s=2 ⇒ 18A + 9B + 8C + 24D = 34.

After solving, we get A = −2, B = −2, C = 5, D = 2. So

Y (s) = −2
s 2 5 2
(s + 1)2
− + + ,
s2 + 4 s2 + 4 s+1
−t −t
y(t) = −2 cos(2t) − sin(2t) + 5te + 2e □.

Now recall that L also has integral properties. So we can use the Laplace transform to
solve constant-coefficient integral equations in the same way as we did the ODE’s above.
(In fact, the integral equations tend to be slightly easier!)
Example 3.2.7. Find the solution y to
t
y +3 y(τ) dτ = 3t
0

solution: Taking L,
3 3 3 3
Y+ sY= 2 ⇒ 1+ s Y = 2.
s s
So
s+3 3 3
s = 2 ⇒ Y= .
s s(s + 3)

C. Sprouse —Math 280 Notes Page 262 DRAFT August 26, 2022
3.2. Solving ODE’s by Laplace Transform 263

After partial fractions,

1 1 −3t
Y=s − ⇒ y(t) = 1 − e □.
s +3

Notice that the above did not have an initial value, because there was no derivative.
The number of initial values that are needed is still the same as the number of derivatives
in the equation.

Example 3.2.8. Find the solution to the IVP

t
′
y +y =6 y(τ)dτ,
0

y(0) = 5.

solution: Taking L,
6
sY − 5 + Y = s Y .
If we multiply by s and rearrange,

(s + s − 6)Y = 5s
2 5s 5s
(s + 3)(s − 2)
⇒ Y= = .
s2 + s − 6
Partial fractions gives

3 2 −3t 2t
Y= + ⇒ y(t) = 3e + 2e □.
s + 3 s −2

Example 3.2.9. Find the solution to the IVP

t
′
y +y =6 y(τ)dτ,
0

y(0) = 5.

solution: Taking L,
6
sY − 5 + Y = s Y .
If we multiply by s and rearrange,

(s + s − 6)Y = 5s
2 5s 5s
(s + 3)(s − 2)
⇒ Y= = .
s2 + s − 6
Partial fractions gives

Y (s) =
3 2 −3t 2t
+ ⇒ y(t) = 3e + 2e □.
s +3 s− 2

C. Sprouse —Math 280 Notes Page 263 DRAFT August 26, 2022
3.2. Solving ODE’s by Laplace Transform 264

Example 3.2.10. Find the solution to the IVP

t
′ 2t
y + y(τ) dτ = 5e ,
0

y(0) = 1.

solution: We have
1 5
sY − 1 + s Y = ,
s−2
So
2
1 5 s +1 s+3
s+ s Y = +1 ⇒ Y= ,
s−2 s s−2

and thus
2
s(s + 3) s + 3s
(s + 1)(s − 2) (s + 1)(s − 2)
Y= 2
= 2 .

The partial fractions is

2
s + 3s As + B C
=
(s + 1)(s − 2)
+ .
2
s +12 s−2
So

s +3s = (As + B)(s − 2) + C(s + 1) s +3s + 0 = (A+ C)s + (−2A + B)s + (−2B +C).
2 2 2 2
⇒

Then
A + C = 1,
−2A + B = 3,
−2B + C = 0.
The first equation gives C = 2B, so then we have

A = −1,
A + 2B = 1
⇒ B = 1.
−2A + B = 3
C = 2B = 2.

So
2t
y(t) = − cos(t) + sin(t) + 2e □.

10
2t
y(t) = − cos(t) + sin(t) + 2e
8

2
t
−5 −4 −3 −2 −1 1

C. Sprouse —Math 280 Notes Page 264 DRAFT August 26, 2022
 3.2. Solving ODE’s by Laplace Transform 265

Homework 3.2:
A) Use partial fractions to find L−1 (F). (That is, find f (t) with L(f ) = F(s).)
1. 7.
4s − 6 8s + 4
s 2(s 2 + 4)
F(s) = F(s) =
s(s − 1)(s − 2)
2. 8.
4s + 2 2
s − 3s + 1
(s − 1)(s + 1)(s + 2)
F(s) = F(s) =
s 2(s − 1)2
3. 9.
2s − 1 2
F(s) = s − 8s + 1
F(s) =
(s + 1)(s 2 + 9)
s(s + 1)2 2

4. 10.
3s + 2 2s − 1
F(s) =
(s − 2)(s2 + 4) F(s) =
(s − 1)(s 2 − 2s + 2)
5. 11.
2s + 32 2s 2 + 10s + 15
F(s) =
s(s 2 + 16) F(s) =
s(s 2 + 4s + 5)
6. 12.
2
s + 2s + 2 2s + 5
s (s + 1) s 2 (s2 + 2s + 5)
F(s) = 3 F(s) =

B) Solve the initial value problems using Laplace transforms.

13.
y(0) = 3, y (0) = −1
′′ ′
y − y = 5 sin(2t),
solution: Taking the Laplace transform of both sides,

s Y − sy(0) − y (0) − Y =
2 ′ 10 2 10
⇒ s Y − 3s + 1 − Y = .
s2 + 4 s2 + 4

3s(s 2 + 4) − 1(s 2 + 4)
(s − 1)Y
2 10 10
= + 3s − 1 = +
s2 + 4 s2 + 4 s2 + 4
3 2
3s − s + 12s + 6
=
s2 + 4
So
3 2 3 2
3s − s + 12s + 6 3s − s + 12s + 6
Y= =
(s 2 − 1)(s2 + 4) (s + 1)(s − 1)(s 2 + 4)
.

Partial Fractions gives

3 2
3s − s + 12 + 6 A B Cs + D
=
(s + 1)(s − 1)(s + 4)
+ + 2 ,
2 s + 1 s − 1 s +4
where A = 1, B = 2, C = 0, D = −2. So

Y (s) =
1 2 2
+ −
s + 1 s − 1 s2 + 4
−t t
y(t) = e + 2e − sin(2t) □ .

C. Sprouse —Math 280 Notes Page 265 DRAFT August 26, 2022
 3.2. Solving ODE’s by Laplace Transform 266

14. 23.
y(0) = 2, y (0) = 1.
′ ′′ ′
y − y = 3, y(0) = 2. y − y = 1,
15. 24.
y(0) = 0, y (0) = 1.
′ −t
y + 2y = 2e , y(0) = 3. ′′
y − y = 2e ,
′ −t ′

16. 25.
′
y(0) = 0, y (0) = 1.
y + y = 5 sin(2t), y(0) = 1. ′′ t ′
y + y = 2e ,
17.
′ 26.
y(0) = 1,y (0) = 2.
y − 2y = 10 cos(4t), y(0) = 0. ′′ ′ 3t ′
y − 2y = 6e ,
18.
y(0) = 2, y (0) = 1.
′′ ′ ′ 27.
y − 7y + 10y = 0,
y(0) = 1,y (0) = −1.
′′ ′ 2t ′
19. y + y − 2y = 4e ,
y(0) = 2, y (0) = 3.
′′ ′ ′
y − y − 2y = 2,
28.
y(0) = 1,y (0) = 0.
20.
y(0) = 3,y (0) = 5.
′′ −t ′
′′
y + 25y = 0,
′ y + 4y = 10e ,

21. 29.
y(0) = 0,y (0) = 6. y(0) = 1,y (0) = 0.
′′ ′ ′′ 2t ′
y + 9y = 9, y − y = 3e ,

22. 30.

y(0) = 0, y (0) = 0. y(0) = 0, y (0) = 0.

′′ ′ −2t ′ ′′ ′ ′
y + 4y + 3y = 2e , y − 2y = 8cos(2t),

C) Solve the initial value problems using Laplace transforms.

31. 41.
y (0) = 1. y − 5y + 4y = 3e , y(0) = 4,y (0) = 0.
′ ′′ ′ t ′
y + 2y = 4t,
42.
y + 6y + 9y = 0, y(0) = 1, y (0) = −1.
32. ′′ ′ ′
′ t
y + y = 4te , y(0) = 2.
43.
33. y(0) = 1,y (0) = 2.
′′ ′ ′
′ −2t y − 2y + y = 2,
y + 3y = te , y(0) = 1.
44.
y(0) = 0, y (0) = 0.
′′ ′ 2t ′
34. y − 6y + 9y = e ,
′ 2t
y − 2y = e , y(0) = 2. 45.
y(0) = 0, y (0) = 3.
′′ ′ t ′
35. y + 2y + y = 4e ,
′
y − y = 2t + 1, y(0) = 0. 46.
y(0) = 2, y (0) = 1.
′′ ′ t ′
y − 2y + y = 4e ,
36.
y(0) = 2, y (0) = 3. 47.
′′ ′ ′
y + y = 1,
y(0) = 0, y (0) = 2.
′′ ′ −2t ′
37. y + 4y + 4y = 2e ,
y (0) = 3, y (0) = 0.
′′ ′
y − y = t, 48.
y (0) = 1,y (0) = 2.
′′ ′ ′
y + 2y + y = t,
38.
y(0) = 0, y (0) = 2.
′′ ′ −t ′ 49.
y + 3y + 2y = e ,
y(0) = 0, y (0) = 1.
′′ ′ ′
39. y + 2y + y = 2 sin(t),
y(0) = 1, y (0) = 0.
′′ ′ t ′
y −y = e , 50.
y(0) = 1, y (0) = 0.
′′ ′ ′
40. y + y = 2t,
51.
y(0) = 1, y (0) = 3. y(0) = 2,y (0) = 0.
′′ ′ 2t ′ ′′ ′
y − 3y + 2y = 3e , y + 4y = 8t,

C. Sprouse —Math 280 Notes Page 266 DRAFT August 26, 2022
 3.2. Solving ODE’s by Laplace Transform 267

52. 54.
y(0) = 0, y (0) = 0. y(0) = 0, y (0) = 0.
′′ ′ ′ ′′ t ′
y − y = 2t − 3, y + y = 2te ,
53.
y(0) = 0,y (0) = 0.
′′ ′ 2t ′
y − y = 4te ,

D) Solve the initial value problems using Laplace transforms.

55.
y(0) = 0, y (0) = 1.
′′ ′ −2t ′
y + 6y + 10y = 4e ,
solution: Taking Laplace transforms,

s Y − sy(0) − y (0) + 6(sY − y(0)) + 10Y =

2 ′ 4
.
s +2

(s + 6s + 10)Y =
2 4 s+6
+1= ,
s +2 s+2
s+6
Y=
(s + 2)(s 2 + 6s + 10)
.

Completing the square gives s + 6s + 10 = (s + 3) + 1. It is important to do this first, so we

2 2

can see that the formof partial fractions we need is

s+6 A B(s + 3) + C
(s + 2)(s2 + 6s + 10) s + 2 s2 + 6s + 10
= + .

So
s + 6 = A(s + 6s + 10) + (B(s + 3) + C)(s + 2)
2

We can either multiply out to get equations for A, B, C or plug in s = −2, s = −3, and s = 0 to
find them.This gives A = 2,B = −2, C = −1. So
−2t −3t −3t
y(t) = 2e − 2e cos(t) − e sin(t) □ .

56. 62.
y(0) = 0,y (0) = 4.
′ t ′′ ′ ′
y − y = e sin(t), y(0) = 1. y − 2y + 5y = 8e ,
t

57.
′
y + y = 2e
−t
cos(2t), y(0) = 2. 63.

y(0) = 0,y (0) = 3.

58. ′′ ′ −t ′
′ t y + 4y + 5y = 6e ,
y + y = 4e sin(2t),y(0) = 1.
59. 64.
y(0) = 1,y (0) = 0. y(0) = 0,y (0) = 0.
′′ ′ ′ ′′ ′ ′
y + 4y + 5y = 10, y + 2y + 2y = 2t,
60.
65.
y(0) = 0,y (0) = 5. y (0) = 0,y (0) = 1.
′′ ′ ′ ′′ ′ ′
y − 2y + 10y = 10, y + 2y + 5y = 25t,
61. 66.
y(0) = 0, y (0) = 0. y(0) = 2, y (0) = 0.
′′ ′ −2t ′ ′′ ′ t ′
y + 2y + 5y = 10e , y − 2y + 2y = e ,

C. Sprouse —Math 280 Notes Page 267 DRAFT August 26, 2022
 3.2. Solving ODE’s by Laplace Transform 268

E) Solve the initial value problem using Laplace transforms.

67.
t
′
y +4 y(τ)dτ = 4t − 2, y(0) = 3
0
solution: Taking Laplace transforms,

4 4 2 4 4 2
sY − y(0) + s Y = 2 − s ⇒ s + s Y = 2 − s + 3.
s s
Adding fractions,
2 2
s +4 4 − 2s + 3s
s Y= .
s2
So
2
4 − 2s + 3s
Y= .
s(s 2 + 4)
Partial fractions gives
3s2 − 2s + 4 A Bs + C
2
= s + 2
s(s + 4) s +4
where
3s − 2s + 4 = A(s + 4) + (Bs + C)s = (A + B)s + Cs + 4A.
2 2 2

So A + B = 3, C = −2,4A = 4, which gives A = 1,B = 2,C = −2.

1 2s −2
Y= s + 2 + 2 ⇒ y = 1 + 2 cos(2t) − sin(2t) □ .
s +4 s +4

68. 75.
t t
′
y+ y(τ)dτ = 2. y − y(τ) dτ = t, y(0) = 3.
0 0
69. 76.
t t
t ′ 2
y−2 y(τ)dτ = e . y +4 y(τ) dτ = 4t , y(0) = 1.
0 0
70. 77.
t t
′ −2t
y− y(τ) dτ = 5 sin(2t). y + 16 y(τ) dτ = 10e , y(0) = 2.
0 0
71. 78.
t t
2 ′ 2t
y+ y(τ)dτ = t . y −9 y(τ) dτ = 5e ,y(0) = 3.
0 0
72. 79.
t t
−t ′ −t
y+ y(τ) dτ = 2te . y − y(τ) dτ = 4te , y(0) = 3.
0 0
73. 80.
t t
t ′
y −3 y(τ) dτ = 4e cos(2t) y − 6y + 8 y(τ) dτ = 0, y(0) = 1.
0 0
74. 81.
t t
′ t ′
y + y(τ)dτ = 2e , y(0) = 0. y − 2y − 3 y(τ) dτ = 5, y(0) = 1.
0 0

C. Sprouse —Math 280 Notes Page 268 DRAFT August 26, 2022
3.2. Solving ODE’s by Laplace Transform 269

82. 86.
t t
′ 3t ′
y − 3y + 2 y(τ) dτ = 2e ,y(0) = 3. y − 2y + y(τ) dτ = 2t − 2, y(0) = 3.
0 0

83. 87.
t
t ′
′
y − 5y + 6
3t
y(τ) dτ = e , y(0) = 2. y − 2y + 5 y(τ) dτ = 3, y(0) = 1.
0
0
88.
84. ′
t
2t
′
t
−2t y − 2y + 2 y(τ) dτ = e , y(0) = 0.
y + 2y + y(τ) dτ = e , y(0) = 1. 0
0
89.
85.
t t
′ 4t ′
y − 8y + 16 y(τ) dτ = e , y(0) = 1. y + 2y + 5 y(τ) dτ = 5t + 7, y(0) = 0.
0 0

C. Sprouse —Math 280 Notes Page 269 DRAFT August 26, 2022
3.3. Shift Properties and Step Functions 270

3.3 Shift Properties and Step Functions

Just like with the derivative and integral properties, there are two shifting properties of
L, depending on whether it is F(s) or f (t) that is being shifted.

Shift Properties of L:

(I) L e f (t) = F(s − a).

at

(II) L{f (t − a)u(t − a)} = e

−as
F(s).

In the right-hand side of I, we have F(s −a), which is the horizontal rightward shift of
F(s) by a. In II we have a shift of f (t) by a. Except we will have to explain what u(t − a)
is below, and why it must be included.
✎ ☞
Shift Property I
✍ ✌
As with the previous properties, the first is more straightforward and easier to prove.

Theorem 3.3.1.
L e f (t) = F(s − a).
at

Proof.
∞
L{e f (t)} = e f (t)e
at at −st
dt
0

f (t)e f (t)e
∞ ∞
at−st −(s−a)t
= dt = dt.
0 0

Now let σ = s − a. Then

f (t)e f (t)e
∞ ∞
−(s−a)t −σ t
dt = dt = F(σ).
0 0

So
L{e f (t)} = F(σ ) = F(s − a).
at

This property makes it easier to evaluate some Laplace transforms, in particular the
ones we already know. For instance we can derive L{e } now from just knowing L{1}.
at

f (t) = 1
1 at 1
⇒ F(s) = s , L{e ⋅ 1} = s − a .

Similarly we can get L{te }:

at

f (t) = t L{te } = L{e ⋅ t} =

1 at at 1
F(s) =
(s − a)2
⇒ , .
s2

C. Sprouse —Math 280 Notes Page 270 DRAFT August 26, 2022
3.3. Shift Properties and Step Functions 271

And, replacing f (t) by t , cos(ωt), or sin(ωt) gives us several of the other transforms
n

that we have used.

f (t) = t L{t e } = L{e t } =

n n! n at at n n!
F(s) =
(s − a)n+1
⇒ n+1
, .
s

f (t) = cos(ωt)
s at s −a
F(s) = L{e cos(ωt)} =
(s − a)2 + ω2
⇒ 2 2
, .
s +ω

f (t) = sin(ωt)
ω at ω
F(s) = L{e sin(ωt)} =
(s − a)2 + ω 2
⇒ 2 2
, .
s +ω
✎ ☞
Shift Property II
✍ ✌
Before we get to shifting general f (t), we need to go over step functions. The unit
step function u(t) is defined by

u(t)

1 ⋅

1 t≥0
u(t) = .
0 t<0

-1 0 1
t

From the perspective of the Laplace transform, f (t) = u(t) and f (t) = 1 are the same.
This is because the Laplace transform is only considered for f ∶ [0, ∞) → R. Or, in terms
of computing L,
∞ ∞
−st −st
L{u(t)} = u(t)e dt = 1e dt = L{1}.
0 0

So this doesn’t look like very much at first. The advantage of u(t) is what we get when we
shift it:

C. Sprouse —Math 280 Notes Page 271 DRAFT August 26, 2022
3.3. Shift Properties and Step Functions 272

u(t − a)

1 ⋅
1 t ≥a
u(t − a) = .
0 t <a

t
a

Then we have
∞ ∞
−st −st
L{u(t − a)} = u(t − a)e dt = e dt
0 a
1 −st ∞ 1 −sa
= − se = se .
a

And so now we can get something out of this. Any piecewise-constant function f (t) can
actually be written as a sum of shifted step functions. For instance let’s look at the general
such function with four constant pieces:

⎧
⎪
⎪
A 0 0 ≤ t < t1,
⎪
⎪
f (t) = ⎪
⎨
A1 t1 ≤ t < t2 ,
⎪
⎪
⎪
⎪
A2 t2 ≤ t < t3 ,
⎪
⎩A 3 t ≥ t3 .

f (t) A2

A0

A3

A1

t
t1 t2 t3

Written in terms of step functions, this is

f (t) = A0 u(t) + (A1 − A0 )u(t − t1 ) + (A2 − A1 )u(t − t2) + (A3 − A2 )u(t − t3 ).

To see this, you should check what happens when 0 < t < t1 : All of the terms are zero
except for the first one, where you just get f (t) = A0. For t1 < t < t2, the second term
becomes nonzero and you get f (t) = A0 + (A1 − A0 ) = A1 . For t2 < t < t3 , the first three

C. Sprouse —Math 280 Notes Page 272 DRAFT August 26, 2022
3.3. Shift Properties and Step Functions 273

terms are nonzero now, and f (t) = A0 + (A1 − A0 ) + (A2 − A1 ) = A2. For t > t3, all terms
are nonzero and you get A3 .
After writing f with step functions, it is then easy to get L{f }.

L{f } = L{A0 u(t) + (A1 − A0 )u(t − t1 ) + (A2 − A1)u(t − t2 ) + (A3 − A2 )u(t − t3 )}

= A0 L{u(t)} + (A1 − A0 )L{u(t − t1 )} + (A2 − A1)L{u(t − t2)} + (A3 − A2 )L{u(t − t3 )}

= A0 s + (A1 − A0 ) s e 1 + (A2 − A1 ) s e 2 + (A3 − A2 ) s e 3 .

1 1 −t s 1 −t s 1 −t s

The above is of course very easy to generalize to N pieces. Here is an explicit example
with numbers.

Example 3.3.2. Find L{f } for

⎧
⎪
⎪
−1 0 ≤ t < 2,
⎪
f (t) = ⎨
⎪
⎪
3 2 ≤ t < 5,
⎪
⎩0 t >5

3 f (t)

t
1 2 3 4 5
−1

solution: Using the technique above, we have

f (t) = −1u(t) + (3 − (−1))u(t − 2) + (0 − 3)u(t − 5)

= −u(t) + 4u(t − 2) − 3u(t − 5)

So
L{f } = s + s e
−1 4 −2s 3 −5s
− se □.

Now we need to see what the step function has to do with shifting another function
f (t). Again, the Laplace transform only depends on f ∶ [0, ∞) → R. That is, L{f } does
not depend on the values for f (t) where t < 0.

C. Sprouse —Math 280 Notes Page 273 DRAFT August 26, 2022
3.3. Shift Properties and Step Functions 274

So when shifting such a function to the right, the shift should be as follows:
f (t) correct shift

Shift
−−−→
t t
-1 1 a

The shift should not look like this:

(t − a)
2 2
t

Shift
−−−→
t t
-1 1 a

So we cannot use f (t − a) to represent f (t) shifted by a. We need to make sure that

any part of the shifted f that doesn’t come from the original f is eliminated. We do this
by multiplying f (t − a) by 0 for 0 < t < a. But this is the same as multiplying by u(t − a)!

(t − a) u(t − a)
2 2
t

Shift
−−−→
t t
-1 1 a

When computing L, the shift of f (t) to the

right by a is

f (t) → f (t − a)u(t − a)

This explains why Shift Property II looks the way it does. Now we are ready to prove
it.

Theorem 3.3.3.

L {f (t − a)u(t − a)} = e L {f }.
−as

C. Sprouse —Math 280 Notes Page 274 DRAFT August 26, 2022
3.3. Shift Properties and Step Functions 275

Proof.
∞
L{f (t − a)u(t − a)} = f (t − a)u(t − a)e
−st
dt
0

f (t − a)e
∞
−st
= dt
a

where we have used that u(t − a) = 0 for t < a. Now we can make the change of variables
τ = t − a with dτ = dt and then
∞
L{f (t − a)u(t − a)} = f (τ)e
−s(τ+a)
dτ
0

f (τ)e
∞
−as −sτ
= e dτ.
0
L{f }.
−as
= e

This shift property is what is used to compute the Laplace transform of piecewise-
defined functions. For example, the function f (t) = (t − a) u(t − a) above can also be
2

written
(t − a) t ≥ a
f (t) =
2
.
0 t<a

So this is a piecewise-defined function. By the shift property, we have

L{f } = L{(t − a) u(t − a)} =

2 2 −as
e .
s3
For most other piecewise-defined functions however, things are not as simple. Here is
an example that shows why.

Example 3.3.4. Find L{f } where

40
u(t) f (t)
30

f (t) =
0 0≤t <3 20
.
t2 t≥3
10 ⋅
t
1 2 3 4 5 6

solution: It is easy to write f in terms of step-functions: f (t) = t u(t − 3). But this is not
2
2
in a form that can be used with the shift property. So for the t term we need to replace t
by (t − 3) + 3.

f (t) = ((t − 3) + 3) u(t − 3) = (t − 3) + 6(t − 3) + 9 u(t − 3).

2 2

C. Sprouse —Math 280 Notes Page 275 DRAFT August 26, 2022
3.3. Shift Properties and Step Functions 276

So

L{f } = L{(t − 3) u(t − 3)} + 6L{(t − 3)u(t − 3)} + 9L{u(t − 3)}

2

2 −3s 6 −3s 9 −3s

= e + 2e + se □.
s3 s

For polynomials or powers of t, sometimes it is helpful to use Taylor series for this
type of thing:

3
Example 3.3.5. Find L{t u(t − 2)}.

solution: We have

t = ((t − 2) + 2) = (t − 2) + 6(t − 2) + 12(t − 2) + 8.

3 3 3 2

To get what is written on the right, one has to do the algebra of cubing ((t − 2) + 2),
which we did not show since it is a little bit of work to write out. Another way to get the
3
same thing is to find the Taylor series of the polynomial t centered at a = 2 from
(n)
Calculus 2, where the coefficients are cn = f (a)/n!:
(n) (n)
n f (t) f (2) cn
3
0 t 8 8
⇒ t = 8 + 12(t − 2) + 6(t − 2) + (t − 2) .
2 3 2 3
1 3t 12 12
2 6t 12 6
3 6 6 1
So

L{8u(t − 2) + 12(t − 2)u(t − 2) + 6(t − 2) u(t − 2) + (t − 2) u(t − 2)}

3 2 3
L{t u(t − 3)} =
8 −2s 12 −2s 12 −2s 6 −2s
= se + 2e + 3e + 4e .
s s s

It is this operation that generally makes the shifting property somewhat tricky. But if
you remember the below, you can automate the process to make it less difficult:

To compute L{f } where f (t) = g(t)u(t − a),

you must write

f (t) = g( (t − a) + a)u(t − a),

and simplify g( (t − a) + a ) to a sum of func-

tions of the form gi (t − a)u(t − a) if possible.

Example 3.3.6. Find L{f } where

C. Sprouse —Math 280 Notes Page 276 DRAFT August 26, 2022
3.3. Shift Properties and Step Functions 277

1 u(t) f (t)
0.8

1 0.6 ⋅
f (t) =
0 0≤t< 5
−3t .
e t ≥ 15 0.4
0.2
t
0.2 0.4 0.6 0.8 1

solution: We have f (t) = e

−3t
u(t − .2). Write

−3t −3((t−.2)+.2) t−.2 −.6

e =e =e e .

Therefore f (t) = e−3(t−.2) e −.6 u(t − .2) = e −.6 e −3(t−.2) u(t − .2). So we can now apply the
shift theorem:

L{f (t)} = L{e

−.6 −3(t−.2)
e u(t − .2)}
−.6 −3(t−.2) −.6 1 −.2s
= e L{e u(t − .2)} = e e .
s +3

This can be written nicely as

−3t 1 −.2(s+3)
L{e u(t − .2)} = e □.
s+3

The above were examples with only one step function. We need to be able to get
more general piecewise-defined functions. In fact it is easy to get any piecewise-defined
function, as long as we use the same technique that we did for step functions above. For
instance in this picture,
f (t)
f2

f0

f1
t
t1 t2

we can see by using the previous technique that

f (t) = f 0 (t) + u(t − t1 )(f 1 (t) − f 0(t)) + u(t − t2 )(f2 (t) − f 1 (t)).

Example 3.3.7. Write in terms of step functions. Then use the shift properties to find L{f }.

C. Sprouse —Math 280 Notes Page 277 DRAFT August 26, 2022
3.3. Shift Properties and Step Functions 278

f (t)
7
6

⎧
5
⎪
⎪
2t + 1 0 ≤ t < 2
f (t) = ⎪
4
⎨
⎪
⎪
7−t 2 ≤ t < 5
⎪
3
⎩ 2 t >5 2
1
t
1 2 3 4 5 6 7

solution: From the above,

f (t) = (2t + 1) + ((7 − t) − (2t + 1))u(t − 2) + (2 − (7 − t))u(t − 5)

= 2t + 1 + (6 − 3t)u(t − 2) + (t − 5)u(t − 5)
= 2t + 1 − 3(t − 2)u(t − 2) + (t − 5)u(t − 5).

So
L{f } =
2 1 3 −2s 1 −5s
2
+ s − 2e + 2e □.
s s s

Notice that in the above example that the terms naturally took the form g(t − a)u(t −
a), without doing much extra work. This sometimes happens when the pieces connect
together nicely, such as when f (t) is piecewise linear and continuous.

Example 3.3.8. Find L{f } where

20 f (t)

2
f (t) =
t 0≤t <3 10
.
0 t≥3

t
1 2 3 4 5 6

solution: This is just

f (t) = t + (0 − t )u(t − 3) = t − t u(t − 3).

2 2 2 2

2
Now one needs to do some work to get t u(t − 3) into the right form before finding L.
This was already done in Example above. So we can write

f (t) t − (t − 3) + 6(t − 3) + 9 u(t − 3)

2 2
=

t − (t − 3) u(t − 3) − 6(t − 3)u(t − 3) − 9u(t − 3),

2 2
=

and
L{f } =
2 2 −3s 6 −3s 9 −3s
3
− 3e − 2e −s e .
s s s

C. Sprouse —Math 280 Notes Page 278 DRAFT August 26, 2022
3.3. Shift Properties and Step Functions 279

Now let’s see how these fit into solving ODE’s.

Example 3.3.9. Find the forced solution yF (t) to y + y = f (t) ∶
′

20 f (t)

2
f (t) =
t 0≤t <3 10
.
0 t≥3

t
1 2 3 4 5 6

solution: The function f (t) is of course from the previous example. So taking L of both
sides gives
2 2 −3s 6 −3s 9 −3s
sY + Y = 3 − 3 e − 2e − se .
s s s
So
2
(s + 1)Y = 3 −
2 2 + 6s + 9s −3s
3
e .
s s
Solving for Y gives
9s2 + 6s + 2
Y (s) =
2 −3s
(s + 1)s (s + 1)s
3
− 3
e .

To get y(t) now, we need to do two partial fractions.

2 A B C D
=
(s + 1)s
+ + + .
3 s + 1 s 3 s2 s
This one can be done fairly easy by rearranging to get
2 −2 2 2 2
=
(s + 1)s
+ − + .
3 s + 1 s 3 s2 s
For the second partial fractions, we have

9s2 + 6s + 2 A B C D
=
(s + 1)s
+ 3+ 2+ s.
3 s+1 s s
−3s
Note that the e is not part of the partial fractions! This is very important, as partial
fractions only works for rational functions, which can’t have any exponentials. We get
2
9s + 6s + 2 −5 2 4 5
=
(s + 1)s
+ 3 + 2 +s .
3 s + 1 s s
So the full decomposition of Y is
−2 2 2 2 −5 9 3 5 −3s
Y(s) = + − + − + − + e .
s + 1 s3 s2 s s + 1 s3 s 2 s
Then using the shift theorem gives

+ (t − 3) + 4(t − 3) + 5 u(t − 3) □ .
−t 2 −(t−3) 2
y(t) = −2e + t − 2t + 2 − − 5e

C. Sprouse —Math 280 Notes Page 279 DRAFT August 26, 2022
3.3. Shift Properties and Step Functions 280

5 y(t)
4

1
t
1 2 3 4 5 6 7

Again, it is very important for these that the exponential is never part of the partial
fractions. It remains on the side during partial fractions, and is used afterwards, with the
shift theorem.

Example 3.3.10. Solve y + 6y + 5y = f (t):

′′ ′

50 f (t)
40
⎧
⎪
⎪
0 0≤t <2
f (t) = ⎪
30
⎨
⎪
⎪
40 2≤t <6.
⎪
⎩0
20
t≥6
10
t
2 4 6 8 10

solution: It is easy to see that f (t) = 40u(t − 2) − 40u(t − 6). So

2 40 −2s 40 −6s
s Y + 6sY + 5Y = s e − s e

Y (s) =
40 −2s 40 −5s
⇒ e − e
s2 + 6s + 5 s 2 + 6s + 5

So again we have two partial fractions to do, but fortunately this time they are both the
same:
40 40 A B C
= = s + + .
2
s(s + 6s + 5) s(s + 1)(s + 5) s + 1 s + 5

Again, there are never exponentials in the partial fractions! Heaviside coverup gives

40 40 40
(1)(5) (−1)(4) (−5)(−4)
A= =8 , B= = −10 , C= = 2.

So
Y (s) = s −
8 10 2 −2s 8 10 2 −6s
+ e − s− + e ,
s +1 s +5 s +1 s+ 5

y(t) = (8 − 10e )u(t − 2) − (8 − 10e )u(t − 6) □ .

−(t−2) −5(t−2) −(t−6) −5(t−6)
+ 2e + 2e

C. Sprouse —Math 280 Notes Page 280 DRAFT August 26, 2022
3.3. Shift Properties and Step Functions 281

u(t) y(t)
10

t
2 4 6 8 10

Example 3.3.11. Solve y + 2y + 50y = f (t) ∶

′′ ′

100 f (t)
80

60
f (t) =
100 0 ≤ t < 5
.
0 t ≥5 40
20
t
5 10

solution: As usual,

f (t) = 100 − 100u(t − 5)

100 100 −5s
⇒ F(s) = s − s e .
So we have
2 100 100 −5s
s Y + 2sY + 50Y = s − s e
Y (s) =
100 100 −5s
⇒ − e .
s(s 2 + 2s + 50) s(s 2 + 2s + 50)

Completing the square, s 2 + 2s + 50 = (s + 1)2 + 49, so partial fractions gives

100 A B(s + 1) C
= s + s 2 + 2s + 50 + s2 + 2s + 50
s(s 2 + 2s + 50)
2 −2(s + 1) −2
= s + s 2 + 2s + 50 + s2 + 2s + 50 .
So
2 s+1 1 2 s+1 1 −5s
Y(s) = s − 2
(s + 1) + 49 (s + 1) + 49 (s + 1) + 49 (s + 1)2 + 49
2
−2 2
− s −2 2
−2 e .

And then
−t 2 −7t −(t−5) 2 −(t−5)
y(t) = 2−2e cos(7t)− e sin(7t)− 2−2e cos(7(t−5))− e sin(7(t−5)) u(t−5) □.
7 7

C. Sprouse —Math 280 Notes Page 281 DRAFT August 26, 2022
3.3. Shift Properties and Step Functions 282

3 y(t)

t
5 10

−1

C. Sprouse —Math 280 Notes Page 282 DRAFT August 26, 2022
 3.3. Shift Properties and Step Functions 283

Homework 3.3:
A) Sketch the function, f (t).
1. f (t) = 1 + 2u(t − 3) 6. f (t) = sin(2(t − π))u(t − π)
2. f (t) = 3u(t − 2) − u(t − 5)
3. f (t) = (t − 3) u(t − 3)
∞
7. f (t) =
2
u(t − k)
4. f (t) = 2 + e
−(t−1)
u(t − 1) k=1

5. f (t) = (t − 2)e 8. f (t) = u(t) − u(t − 1) + u(t − 2) − u(t − 3) + ...

−(t−2)
u(t − 2)

B) a) Graph the function.

b) Express f (t) in terms of unit step functions.
c) Find the Laplace transform using the shift property, L{f (t − a)u(t − a)} = e F(s).
−as

9.
f (t) =
5 − 2t 0≤t≤2
.
t−1 t ≥2
solution: f 0 (t) = 5 − 2t is a line of slope −2 that goes through 5 at t = 0 and 1 at t = 2.
f 1 (t) = t − 1 is a line ofslope 1 that goes through 1 at t = 2. So we have the following graph.
5 f (t)

1
t
1 2 3 4 5
In terms of unit step functions, this is

f (t) = (5 − 2t) + ((t − 1) − (5 − 2t))u(t − 2) = 5 − 2t + (3t − 6)u(t − 2).

So then

L{f } = L{5 − 2t + 3(t − 2)u(t − 2)}

= 5L{1} − 2L{t} + 3L{(t − 2)u(t − 2)}
5 2 3 −2s
= s − 2 + 2e
s s

10. 12.
⎧
⎪
⎪
−2 0 ≤ t < 1
f (t) = f (t) = ⎪
0 ≤ t < t0
⎨
3
⎪
⎪
3 1≤t <3
⎪
0 t ≥ t0
⎩1 t ≥3
11. 13.
⎧
⎪ ⎧
⎪
⎪ ⎪
0 0 ≤t <3 0≤t<2
f (t) = ⎪ f (t) = ⎪
5
⎨
⎪ ⎨
⎪
⎪ ⎪
2 3 ≤t ≥8 3 2≤t<5
⎪
⎩0 t≥8 ⎪
⎩6 t ≥5

C. Sprouse —Math 280 Notes Page 283 DRAFT August 26, 2022
 3.3. Shift Properties and Step Functions 284

⎧
14. 19.
⎪
⎪ 0≤t <1
⎪
0
f (t) =
1 0≤t ≤3
⎪
f (t) = ⎪
⎨
3 1≤t <2 2(t−3) .
⎪
t≥3
⎪
e
⎪
⎪
5 2≤t <3
⎪
⎩6 t≥3 20.
f (t) =
t −2 0≤t ≤2
15. −(t−2) .
t≥2
f (t) =
5−t 0≤t≤3 e
.
2 t ≥3
⎧
21.
⎪
⎪
0≤t ≤1
f (t) = ⎪
16. 1
⎨
⎪
f (t) = ⎪
−1 0≤t≤2 t 1 ≤ t ≤ 3.
3t − 7 t ≥2
. ⎪
⎩3 t≥3
17. 22.
f (t) = f (t) =
2t 0≤t≤1 t +2 0≤t ≤3
. .
3−t t ≥1 2t − 1 t≥3
18. 23.
f (t) = f (t) =
0 0 ≤ t ≤ t0 t2 0≤t <1
(t − t0 )
3 . .
t ≥ t0 2t − 1 t≥1

C) a) Express f (t) in the form f (t − a)u(t − a)

b) Use this to find L{f }.
24. f (t) = (t − e )u(t − 3).
2 2t

solution: We need to rewrite t as t = (t − 3) + 3. So

f (t) = (((t − 3) + 3) + e )u(t − 3)

2 2((t−3)+3)

= ((t − 3) − 6(t − 3) + 9 + e e )u(t − 3).

2 6 t−3

So
6
2 −3s 6 −3s 9 −3s e −3s
F(s) = e − e + se + e
s3 s2 s −2
2 −3s 6 −3s 9 −3s 1 −3(s−2)
= e − e + se + e □.
s3 s 2 s − 2

25. f (t) = 2tu(t − 1) 28. f (t) = (t + 3)u(t − 1)

2

26. f (t) = (3t + 1)u(t − 2) 29. f (t) = e u(t − 5)

3t

27. f (t) = t 2u(t − 5) 30. f (t) = e −2t+1 u(t − 3)

D) a) Use the method of Example 3.3.5 to write f (t) in the form f (t − a)u(t − a).
b) Use this to find L{f }.

31. f (t) = (3t − 5t)u(t − 2) 34. f (t) = (2t + t)u(t − 2)

2 3

32. f (t) = t 3u(t − 5) 35. f (t) = t 4 u(t − 2)

33. (t − t )u(t − 3) 36. f (t) = (t − t )u(t − 1)
3 2 4 3

E) a) Write f (t) in terms of unit step functions.

b) Rewrite if necessary and use the shift theorem L{f (t − a)u(t − a)} to find L{f }.

C. Sprouse —Math 280 Notes Page 284 DRAFT August 26, 2022
 3.3. Shift Properties and Step Functions 285

37. 43.
f (t) =
5 0≤t <3
f (t) =
2t 0≤t<1
t t≥3 t
2
t ≥1
38.
f (t) =
3t 0≤t <2 44.
f (t) =
1 t≥2 0 0≤t<3
2t
39. e t ≥3
f (t) =
2
t 0≤t <3
45.
0 t≥3 −3t
f (t) =
e 0≤t <2
⎧
40.
⎪
0 t≥2
⎪
0 0 ≤t <2
f (t) = ⎪
⎨
⎪
⎪
t 2≤t ≤5
⎪
46.
⎩ 0 t≥5
f (t) =
0 0 ≤ t < t0
at
41. e t ≥ t0 .
f (t) =
3t 0≤t <1
t +2 t≥1 47.
⎧
⎪
⎪ at
0 0≤t<1
f (t) = ⎪
42.
⎨
⎪
f (t) = ⎪
2t + 1 0≤t <2 1≤t<2
⎪
e
5 −t t≥2 ⎩0 t ≥2

F) Find L{f }. To derive any of the trig identities needed, you can use
cos(a + b) = cos(a) cos(b) − sin(a) sin(b) , sin(a + b) = sin(a)cos(b) + cos(a) sin(b).

48. f (t) = sin(ω(t − t0 ))u(t − t0 ). 51. f (t) = 2 cos(t)u t − π3 .

49. f (t) =
π
52. f (t) =
sin(2t) 0 ≤ t < π cos(t) 0 ≤ t < 2
0 t≥π 0 t ≥ π2

50. f (t) = 53. f (t) =

sin(t) 0≤ t <π cos(3t) 0< t <π
0 t≥π −1 t>π

G) Solve the initial value problem by Laplace transforms.

y(0) = 0, y (0) = 2.
′′ ′
54. y − y = u(t − 3),
y(0) = 0, y (0) = 1.
′′ ′ ′
55. y + y = u(t − 2),
56. y + 4y = 4u(t − t0 ), y(0) = 3, y (0) = 2.
′′ ′

y(0) = 0, y (0) = 2.
′′ ′ 3(t−1) ′
57. y − 2y = 6e u(t − 1),
58. Find the current through an RL-circuit with R = 4 and L = 200mH (=.2H), and Vin = 12(t − 5)u(t −
5),if the current is initially I (0) = 2 Amps.
4
60 Vin
50
40
+ 30
Vin I 0.2 H
− 20
10
t
1 2 3 4 5 6 7 8 9 10 11
59. A 3-kg ball is on a spring of spring-constant 6 N/m, which has a damping force of 9 times the
velocity.Suppose the ball is released at 1-m from equilibrium with no initial velocity, and a force of
F(t) = 30u(t − 2) is applied to the ball. Solve for the position of the ball, x(t).

C. Sprouse —Math 280 Notes Page 285 DRAFT August 26, 2022
 3.3. Shift Properties and Step Functions 286

H) a) Find f (t) in terms of step functions.

b) Find L{f }.
c) Use Laplace transforms to find the forced solution to the ODE. (All initial values zero.)
60.
y + 5y = f (t), f (t) =
′ 10 0 < t < 7
0 t>7
61.
y + 3y = f (t), f (t) =
′ 9 0≤t<5
.
3 t ≥5
62.
y + 2y = f (t), f (t) =
′ e −2t 0≤t <3
.
0 t ≥3
63.
y + y = f (t), f (t) = e
′ −t
+ 2u(t − 5).
64.
y + 3y = f (t), f (t) = 6 + 2e
′ −5(t−1)
u(t − 1).
65.
y + y = f (t), f (t) =
′ 0 0≤t <π
.
20 sin(3t) t≥π
66.
y − 2y = f (t), f (t) =
′ 0 0≤t <2
.
3t − 6 t≥2
67.
y + 2y = f (t), f (t) = 4t + 2u(t − 1).
′

68.
y − y = f (t), f (t) =
′ 1 0≤t <1
.
3 − 2t t≥1
69.
y − y = f (t), f (t) =
′′ 2t 0≤t <1
.
2t − 2 t>1
70.
2t
y + 4y = f (t), f (t) =
′′ 8e 0≤t <3
.
0 t>3
71. A 2-kg ball is on a spring of spring-constant 30 N/m, which has a damping force of 16 times the
velocity.Find the displacement from equilibriumgiven the applied force

⎧
⎪
⎪
0 0≤t<2
⎪
f (t) = ⎨
⎪
⎪
60 2≤t<5.
⎪
⎩ 0 t ≥5

72. A ball of mass 2-kg is on a spring of stiffness k = 50-N/m with no resistance/damping.Find the
displacement from equilibrium given the applied force

f (t) =
100 − 50t 0≤t <2
.
0 t≥2

C. Sprouse —Math 280 Notes Page 286 DRAFT August 26, 2022
3.3. Shift Properties and Step Functions 287

73. Find the current through an RL-circuit with R = 10 and L = 500mH (= .5H).
10

Vin (t) = f (t) =

200t 0≤t <3
. +
0 t ≥3 Vin I 0.5 H
−

74. Find the capacitor voltage in an RC-circuit with R = 5000 and C = 10µF (= 10−5F).
5000

Vin(t) = f (t) =
10 0 ≤ t < 2
. +
0 t ≥2 Vin I 10
−5
F
−

75. Find the voltage V = VC (t) in the RLC circuit below. (V (0) = V (0) = 0.)
′

60 4H

Vin (t) = f (t) =

240 − 120t 0<t≤2
0 t >2
+
Vin I .005 F
−

76. Find the voltage V = VC (t) in the RLC-circuit below. (V (0) = V (0) = 0.) Compare with Question
′

3.8) #10.

30 2H

Vin (t) = f (t) =

10t 0≤t <6
60 t≥6
Vin + .01 F
I
−

C. Sprouse —Math 280 Notes Page 287 DRAFT August 26, 2022
3.4. Impulse Functions 288

3.4 Impulse Functions

In the language of the previous section, consider the function kh (t) = 1h (1−u(t −h)). That
is,
kh(t)
1
h
1
kh(t) = h
0≤ t <h
0 t>h

t
h
If h > 0 is very small, then kh (t) will look like a tall, thin spike near zero. If f (t) is any
continuous function, then multiplying by k(t) and integrating gives

f (t)k(t) dt = f (t) dt.

∞ 1 h

0 h 0
The right-hand side is the average value of f on the interval [0, h]. Since f is continuous,
the Mean Value Theorem for Integrals from Calculus gives that as h → 0 this average value
converges to simply f (0). That is,
∞
f (t)k(t)dt = lim f (t) dt = f (0).
h
1
lim
h→0 0 h→0 h 0
We want to think of lim kh (t) as a function, even though such a limit does not actually
h→0
exist. As the limit of tall thin spikes, we think of this function as ‘+∞’ at t = 0, and zero
everywhere else. This is the ‘impulse’ function at t = 0, which we call δ(t).

δ(t)

We would like δ to be defined by

δ(t)” = ”lim kh (t).

h→0

But since that limit doesn’t exist, we can only think of this as intuition. What does exist
however is that limh→0 ∫0 f (t)k(t) dt = f (0), from above. So thinking of passing the limit
∞

to the inside of the integral, we define δ(t) to be the ‘function’ such that
∞
f (t)δ(t) dt = f (0).
0

C. Sprouse —Math 280 Notes Page 288 DRAFT August 26, 2022
3.4. Impulse Functions 289

The discussion above also applies to kh (t−a) and δ(t−a), so that the impulse function
δ satisfies the important property

The impulse function δ(t) is defined by

∞
f (t)δ(t − a) dt = f (a)
0

for all a > 0 and continuous f .

δ(t − a)

t
a

Since Laplace transforms are defined in terms of integrals, this property is enough to
allow δ(t) to be considered like a function within this context. In particular the Laplace
transform of the impulse function is
∞
−st −s⋅0
L{δ(t)} = δ(t)e dt = e = 1.
0

∞
−st −s⋅a −as
L{δ(t − a)} = δ(t − a)e dt = e =e .
0
The second equation also follows from the first by the shift properties of the previous
section. The impulse function δ(t) is a very useful concept, because for nonhomogeneous
ODE’s such as y + by + cy = f (t), it is often useful to see what happens when the forcing
′′ ′

function f (t) acts like a sudden spike, in which we will say f (t) = δ(t). And now these
types of equations can easily be dealt with by Laplace transforms.

L{δ(t)} = 1,
−as
L{δ(t − a)} = e .

We have seen how to solve any constant-coefficient initial value problem by Laplace
transforms.

Definition 3.4.1. For the constant coefficient ODE

(n)
+ ... + a1 y + a0 y = f (t)
′
an y

C. Sprouse —Math 280 Notes Page 289 DRAFT August 26, 2022
3.4. Impulse Functions 290

the impulse response is the function h(t) which solves

(n) ′
an y + ... + a1 y + a0 y = δ(t),
(n−1)
y(0) = 0, y (0) = 0, ... , y (0) = 0.
′

Example 3.4.2. Find the impulse response to y ′′ + 6y ′ + 34y = f (t).

solution: We solve y + 6y + 34y = δ(t), y(0) = y (0) = 0 by Laplace transforms.
′′ ′ ′

′′ ′
L{y + 6y + 34y} = L{δ(t)}
2
⇒ s Y + 6sY + 34Y = 1
⇒ (s + 6s + 34)Y =
2
1
1
⇒ Y= 2 .
s + 6s + 34
Completing the square,
1 1 −3t
(s + 3)2 + 25
Y(s) = ⇒ y(t) = e sin(5t).
5
−3t
So the impulse response is h(t) = 15 e sin(5t) □ .
The impulse response is a fundamental characteristic of any linear ODE. We will show
in the next section that the solution to any linear ODE with forcing function f (t) is given
by the convolution of f (t) with the equation’s impulse response:

yF (t) = h(τ)f (t − τ)dτ.

t

0
This gives a new way of looking at Variation of Parameters, which is much more conve-
nient for applications. For example, this shows that the output of any speaker can be sim-
ulated by simply taking the input and convolving with the speakers’ impulse response.
The impulse response is obtained by playing a short pulse into the speaker and simply
recording the output. There are now many such impulse responses available both pub-
licly and commercially, which allow one to digitally simulate the output from a speaker,
microphone, amplifier, or even the acoustic properties of an enclosed room. This is com-
monly used to simulate vintage guitar amplifiers, where there are even hardware effect
pedals to do the convolution. Here is what the impulse response of a Marshall 1960A am-
plifier (of the famous Marshall Stack) looks like, from a public domain library at sound-
woofer.se:

1 h(t)
0.8
0.6

0.4
0.2
msec
1 2 3 4 5 6 7 8 9 10
−0.2

C. Sprouse —Math 280 Notes Page 290 DRAFT August 26, 2022
3.4. Impulse Functions 291

We will cover the details of convolution in the next section. For the rest of this section
we will get some more practice with using the impulse functions δ(t) and in particular
shifting.

Example 3.4.3. Solve

′′ ′
y + 4y + 3y = 2δ(t − 5)
y(0) = 0,

y (0) = 0.
′

solution: Taking Laplace transforms gives

2 −5s 2 −2s 2 −5s

(s + 1)(s + 3)
s Y + 4sY + 3Y = 2e ⇒ Y= e = e .
s 2 + 4s + 3
After partial fractions,
1 1 −5s
Y= − e .
s +1 s +3
−(t−5) −3(t−5)
So y(t) = e −e u(t − 5). The graph below shows how y arises from an
impulse applied at t = 5 □.

y(t)

t
1 2 3 4 5 6 7 8 9 10 □.

Example 3.4.4. Solve

y ′′ + 4y ′ + 20y = 4δ(t − 3)
y(0) = 1 .
y (0)
′
= 0

solution:
′′ ′
L{y + 4y + 20y} = L{4δ(t − 3)},
so
(s + 4s + 20)Y = s + 4 + 4e
2 −3s 2 −3s
s Y − s(1) + 4(sY − 1) + 20Y = 4e ⇒ ,
and
Y (s) =
s+1 4 −3s
+ e .
s 2 + 4s + 20 s2 + 4s + 20
We must deal with each term separately. Completing the square,
s + 4s + 20 = (s + 2) + 16. So partial fractions on the first term gives
2 2

s+4 (s + 2) + 2 s+2 2
s + 4s + 20 (s + 2) + 4 (s + 2)2 + 42
= = + .
s 2 + 4s + 20 2 2 2

C. Sprouse —Math 280 Notes Page 291 DRAFT August 26, 2022
3.4. Impulse Functions 292

So
Y (s) =
s+2 2 4 −3s
(s + 2) + 4 (s + 2) + 4
2 2
+ 2 2
+ 2 e
s + 4s + 20
gives
−2t 1 −2t −2(t−3)
y(t) = e cos(4t) + e sin(4t) + e sin(4(t − 3))u(t − 3).
2
This function is graphed below. Again we can see the result of the impulse applied at
t = 3.
1 y(t)

t
1 2 3 4 5 6 7 8

−1
□.

Example 3.4.5. Solve

y ′′ + 3y ′ + 2y = 2 − 2δ(t − 3) + 3δ(t − 7)
y(0) = −2 .
y (0)
′
= 1
solution: Taking Laplace transforms gives
2 2 −3s −7s
s Y + 2s − 1 + 3(sY + 2) + 2Y = s − 2e + 3e

(s + 3s + 2)Y
2 2 −3s −7s
⇒ = −2s + 5 + s − 2e + 3e
−2s 2 + 5s + 2 −s −2s
= s − e + 3e .

2 −s −2s
−2s + s + 2 −2e 3e
⇒ Y = 2
+ 2 + 2
s(s + 3s + 2) s + 3s + 2 s + 3s + 2
2
−2s + 5s + 2 −2 −3s 3 −7s
(s + 1)(s + 2) (s + 1)(s + 2)
= + e + e
s(s + 1)(s + 2)
We have three partial fractions that must be done individually. (We leave out the details.)
2
−2s + 5s + 2 1 −5 2
=s + +
s(s + 1)(s + 2) s + 1 s + 2
−2 −2 2
(s + 1)(s + 2) s + 1 s + 2
= +

3 3 3
(s + 1)(s + 2)
= −
s + 1 s + 2
So
Y (s) = s +
1 −5 2 −2 2 −3s 3 3 −7s
+ + + e + − e .
s + 1 s +2 s+1 s +2 s +1 s+ 2
−t −2t −(t−3) −2(t−3) −(t−7) −2(t−7)
y(t) = 1 − 5e + 2e + −2e + 2e u(t − 3) + 3e − 3e

C. Sprouse —Math 280 Notes Page 292 DRAFT August 26, 2022
3.4. Impulse Functions 293

2 y(t)

1
t
1 2 3 4 5 6 7 8
−1

−2
□.

Example 3.4.6. Find yF (t) for y + 4y + 4y = f (t), where f (t) = ∑ δ(t − k).
∞
′′ ′
k=0

f (t)

t
1 2 3 4 5

solution:
∞
′′ ′
L{y + 4y + 4y} = L δ(t − k)
k=0
∞ ∞
(s + 4s + 4)Y =
2 −k s
⇒ L{δ(t − k)} = e ,
k=0 k=0
so
1 −ks 1 −ks
(s + 2)2
Y(s) = e = e .
s 2 + 4s + 4
−1 1 −2t
No partial fractions is necessary here, as L (s+2)2
= te . So the shift theorem gives

∞
(t − k)e
−2(t−k)
y(t) = u(t − k) □ .
k=0

You should look at the graph of this and see if it matches your intuition.

.5
y(t)

t
1 2 3 4 5
□.

C. Sprouse —Math 280 Notes Page 293 DRAFT August 26, 2022
 3.4. Impulse Functions 294

Homework 3.4:
A) Find the impulse response, h(t).
1.
y + 6y + 34y = f (t).
′′ ′

solution: The impulse response h(t) here is by definition the solution to

y(0) = 0 , y (0) = 0.
′′ ′ ′
y + 6y + 34y = δ(t),

Taking L gives

s Y + 6Y + 34Y = 1 ⇒ (s + 6s + 34)Y = 1 ⇒ Y =
2 2 1
.
s 2 + 6s + 34

Since 34 > (6/2)2 = 9, the denominator cannot be factored. So completing the square gives
1
Y=
(s + 3)2 + 25
and hence
1 −3t
h(t) = y(t) = e sin(5t) □ .
5

2. y − 2y − 15y = f (t). 5. y + 9y = f (t)

′′ ′ ′′′ ′

3. y + 16y + 64y = f (t). 6. y + 2y = f (t)

′′ ′ ′′′ ′′

4. y + 2y + 50y = f (t) − y = f (t)

′′ ′ ′′′′
7. y

B) Use Laplace transforms to solve the initial value problems.

8.
y(0) = 3, y (0) = 0.
′′ ′
y + 25y = 10δ(t − 1),
solution: Taking Laplace transforms,

(s + 25)Y = 3s + 10e
2 −s 2 −s
s Y − 3s + 25Y = 10e ⇒

3s 10 −s
Y= + 2 e .
s 2 + 25 s + 25
The inverses of each of these is done separately! The right term involves a shift, while the
left term does not. We have

y = 3 cos(5t) + 2sin(5(t − 1))u(t − 1) □ .

y(0) = 3, y (0) = 0.
′′ ′ ′
9. y + 8y + 12y = 2δ(t),
y(0) = 0, y (0) = 0.
′′ ′ ′
10. y + 2y + y = 2 − δ(t),
y(0) = 0, y (0) = 0.
′′ ′
11. y + 16y = 32t + 6δ(t),
y(0) = 0, y (0) = 0.
′′ ′ ′
12. y − 5y + 6y = δ(t − 1),
y(0) = 3, y (0) = 1.
′′ ′
13. y − y = 2δ(t − 3),
y(0) = 2, y (0) = 3.
′′ ′ ′
14. y + 3y = 6δ(t − 5),
y(0) = 1, y (0) = 0.
′′ ′
15. y + 4y = 2δ(t − 3),

C. Sprouse —Math 280 Notes Page 294 DRAFT August 26, 2022
 3.4. Impulse Functions 295

y(0) = 1, y (0) = −3.

′′ ′ ′
16. y + 4y + 4y = δ(t − 1),
y(0) = 0, y (0) = 0.
′′ ′ −t ′
17. y + 4y + 3y = 4e + 6δ(t − 2),
y(0) = 0, y (0) = 0.
′′ ′ ′
18. y + y − 2y = 6 + 3δ(t − 5),
y(0) = 1, y (0) = 0.
′′ ′ ′
19. y + 6y + 9y = 2δ(t) − δ(t − 1),
′
20. y + y = 10cos(3t) + 2δ(t − 5), y(0) = 0.
y(0) = 0, y (0) = 1.
′′ ′ ′
21. y + 2y + 5y = 5δ(t) − 10δ(t − 3),
y(0) = 2, y (0) = −1.
′′ ′ ′
22. y + 4y + 5y = δ(t − 3),
∞
C) 23. a) Solve y ′′ + 10y ′ + 25y = y(0) = 0, y (0) = 0.
′
δ(t − k),
k=0
b) Sketch the solution, y(t).
∞
y(0) = 0, y (0) = 0.
′′ ′
24. a) Solve y + y = δ(t − kπ),
k=0
b) Use that sin(t − kπ) = (−1) sin(t) to show that
k

sin(t) t ∈ [k, k + 1], k even

t ∈ [k, k + 1], k odd
y(t) =
0

c) Sketch the solution, y(t).

D) 25. An RL circuit with R = 10 , L = 200 mH (=.2 H), has I (0) = 0 Amps. Find the current I(t) if there is
an input voltage of Vin (t) = 6 + δ(t − t0 ) Volts.
10
10 Vin
8
6
+
Vin I 0.2 H 4
−
2
t
t0

26. An RLC circuit with R = 30 , L = 5 H, and C = .04 F has VC (0) = 12 Volts, VC (0) = 0. Find V = VC (t)
′

if an impulse of Vi n (t) = 20δ(t − 3) Volts is applied.

30 5H

Vin + .04 F
I
−

2
27. Suppose a ball on a spring has mass m = 3 kg. The spring has spring constant k = 15 kg/m and
resistance to motion of6 times the velocity. Find the motion of the ball if the spring is initially
stretched 2 meters past equilibrium, with no initial velocity, and a force of 10δ(t − t0 ) is applied.

C. Sprouse —Math 280 Notes Page 295 DRAFT August 26, 2022
3.6. Systems and Circuits ∗ 312

10
2
A(γ )

1
10

0
10
A
−1
10

0 1 2
10 10 10
γ

C. Sprouse —Math 280 Notes Page 312 DRAFT August 26, 2022
Chapter 4

First Order Systems

4.1 Constant Coefficient Systems: Substitution/Elimination

In this chapter we will discuss systems of differential equations. Again, a system is a set
of several equations that hold simultaneously, with more than one unknown variable to
solve for. This is the same idea as linear systems of algebraic equations that were the main
subject of Chapter 2, but now the equations will involve functions and their derivatives.

Definition 4.1.1. A First Order System of ODE’s is a set of equations of the form

= f 1 (x1 ,..., xn , t),

dx1
dt
⋮
= f 1 (x1 ,..., xn , t).
dxn
dt

It is important that if there are n unknowns x1 (t), ..., xn (t), there must also be exactly n equa-
tions. The number of unknowns is called the dimension, so this is an n-dimensional (n-d)
first order system.

For now, we want to stick to equations we can solve explicitly using the techniques of
the last chapter. So we need to restrict to constant-coefficient systems.

Definition 4.1.2. A first order system of ODE’s is called constant-coefficient, or linear time
invariant (LTI), if it is of the form
dx1
dt
= a11x1 + ...a1n xn + f 1(t),
⋮
dxn
dt
= an1x1 + ...ann xn + f n(t).

These constant-coefficient systems can in fact always be solved by Gaussian elimina-

tion and back-substitution. exactly like the linear systems from Section 3.1. We indicate
how this can be done for 2-dimensional systems here, in which case we call this the elim-
ination method. The other way to solve 2-d systems is the substitution method. These
both correspond to the substitution and elimination methods for solving the 2-d linear
systems which are commonly introduced in high school. For example,

313
4.1. Constant Coefficient Systems: Substitution/Elimination 314

Example 4.1.3. Solve the linear system

2x − y = 3,
x + 3y = 5.
✬ ✩
Substitution:

This is the most basic method. We can use the first equation to get y = 2x − 3 and plug
into the second equation, or use the second equation to get x = 5 − 3y and plug into the
first equation. So using the first to plug into the second gives

y = 2x − 3 ⇒ x + 3(2x − 3) = 5 ⇒ 7x = 14.

So x = 2, and back-substituting, y = 2(2) − 3 = 1 □.

✫ ✪
✬ ✩
Elimination:

We need to add a multiple of one equation to the other to eliminate either x or y. Follow-
ing the order of Gaussian elimination, we eliminate x:

2x − y = 3
−2( x + 3y = 5)

−7y = −7

So y = 1. Then back-substituting into the original second equation, x = 5 − 3y = 5 − 3(1) =

2 □.
✫ ✪

Now we indicate how each of these methods can be done with the 2-d constant-
coefficient system:
dx
dt
= ax + by + f (t),
dy
dt
= cx + dy + g(t).
First we give an example of substitution. This the most elementary way to deal with
2-dimensional systems. The important thing is that you can use the first equation to solve
for y:
− ax − f (t) ,
1 dx
y=
b dt
but you cannot use the first equation to solve for x! That would involve solving a differ-
ential equation, since the first equation has both x and dx
dt
. Similarly, the second equation
cannot be used to solve for y. The second equation can be used to solve for x:
1 dy
x= c − dy − g(t) .
dt
Either way, you plug this substitution into the other equation. Using the first equation to
plug into the second gives a second order ODE for x. Using the second equation to plug
into the first gives a second order ODE for y.

C. Sprouse —Math 280 Notes Page 314 DRAFT August 26, 2022
4.1. Constant Coefficient Systems: Substitution/Elimination 315

Example 4.1.4. Find the general solution by substitution.

dx 3t
dt
= x − y + 7e ,
dy 3t
dt
= 2x − y + 8e .

solution: Let’s use the first equation to solve for y in terms of x. That is, we get:
dx 3t
y=− + x + 7e .
dt
Then we insert this into the second equation to get
d 3t 3t 3t
dt − dd xt + x + 7e = 2x − − dx
dt + x + 7e + 8e
2
3t 3t
− ddt 2x + dx
dt
+ 21e = x + dx
dt
+e

Or,
2
d x 3t
2
+ x = 20e .
dt
2
The characteristic equation is s + 1 = 0 ⇒ s = ±i so yh = c1 cos(t) + c2 sin(t). The forcing
3t 3t
function 20e means we have yp = Ae by undetermined coefficients, and plugging in
gives 10A = 20 ⇒ A = 2. So
3t
x(t) = c1 cos(t) + c2 sin(t) + 2e .

Now, here comes something important. In order to find y, should not start over from the
beginning and repeat the above process, using the second equation to solve a differential
equation for y! Not only will this take much more work than necessary, it will give a
wrong answer, because solving another differential equation gives two more arbitrary
constants c3 ,c4 , which in fact should not be arbitrary. They are actually determined by
whatever c1 and c2 are. We can see this when we find y the correct way, which is simply
to back-substitute into the equation for y that we already know.

dx 3t
y = − + x + 7e
dt
d 3t 3t 3t
= − c cos(t) + c2 sin(t) + 2e + c1 cos(t) + c2 sin(t) + 2e + 7e
dt 1
3t 3t
= − −c1 sin(t) + c2 cos(t) + 6e + c1 cos(t) + c2 sin(t) + 9e
= (c1 − c2 ) cos(t) + (c1 + c2 ) sin(t) + 3e .
3t

Therefore the general solution to the system is

3t
x(t) = c1 cos(t) + c2 sin(t) + 2e
(c1 − c2) cos(t) + (c1 + c2) sin(t) + 3e3t
□.
y(t) =

Next we will solve the same system by elimination.

Example 4.1.5. Find the general solution by elimination.
dx 3t
dt
= x − y + 7e ,
dy 3t
dt
= 2x − y + 8e .

C. Sprouse —Math 280 Notes Page 315 DRAFT August 26, 2022
4.1. Constant Coefficient Systems: Substitution/Elimination 316

solution: Using D to denote differentiation, we have

3t 3t
Dx = x − y + 7e Dx − x + y = 7e
⇒ ⇒
Dy = 2x − y + 8e 3t Dy − 2x + y = 8e 3t

(D − 1)x + y
3t
= 7e
−2x + (D + 1)y
⇒ 3t
= 8e .
To eliminate x we multiply the first equation by 2, the second by D − 1, and add.

2(D − 1)x + 2y = 14e 3t

+ −2(D − 1)x + (D − 1)(D + 1)y (D − 1)(8e )
3t
=
.

2y + (D − 1)(D + 1)y = 14e + (D − 1)(8e )

3t 3t

d
Now we simplify the left and right side of the above, using that D is just the derivative dt
.

2y + (D − 1)(D + 1)y = 2y + (D + D − D − 1)y = 2y + (D − 1)y = 2y + y − y = y + y,

2 2 ′′ ′′

14e + (D − 1)(8e ) = 14e + 24e − 8e

3t 3t 3t 3t 3t 3t
= 30e .
So we now have a second order ODE for y,
′′ 3t
y + y = 30e .
2
The characteristic equation is s + 1 = 0 ⇒ s = ±i so yh = c1 cos(t) + c2 sin(t). The forcing
3t 3t
function 30e means we have yp = Ae by undetermined coefficients, and plugging in
gives 10A = 30 ⇒ A = 3. So the general solution is y = c1 cos(t) + c2 sin(t) + 3e 3t .
Now again, to find x(t) we do not start from the beginning and solve another differential
equation. Instead, we use one of the original equations to back-substitute. Since we need
to solve for x in terms of y, we need the second equation. (The first equation is unusable
to solve for x, since it contains both x and dx
dt
).

dy 3t 1 dy 1 3t
= 2x − y + 8e ⇒ x= + y − 4e ,
dt 2 dt 2
So
1 d 3t 1 3t 3t
x = c cos(t) + c2 sin(t) + 3e + c cos(t) + c2 sin(t) + 3e − 4e
2 dt 1 2 1
1 1 9 3t 1 1 3 3t 3t
= − c1 sin(t) + c2 cos(t) + e + c1 cos(t) + c2 sin(t) + e − 4e
2 2 2 2 2 2
1 1 1 1 3t
= c + c cos(t) + − c1 + c2 sin(t) + 2e .
2 1 2 2 2 2

So the general solution of the system is then

1 3t
x(t) = c + 12 c2
2 1
cos(t) + − 12 c1 + 12 c2 sin(t) + 2e
3t .
y(t) = c1 cos(t) + c2 sin(t) + 3e

Note that this solution looks slightly different than when we solved the same system by
substitution above. The reason is that we solved for y first and then x here, but
previously we solved for x first and then y. If we had eliminated y above and solved for x
first, we would have gotten the general solution in the same form as substitution □.

C. Sprouse —Math 280 Notes Page 316 DRAFT August 26, 2022
4.1. Constant Coefficient Systems: Substitution/Elimination 317

Example 4.1.6. Find the general solution to the system.

dx
dt
= y + 3t − 2
dy
dt
= 9x − 18t + 6

solution: We’ll solve this by elimination.

dx
dt
−y = 3t − 2 Dx − y = 3t − 2
⇒ .
dy
− 9x = −18t + 6 −9x + Dy = −18t + 6
dt

So to eliminate x,

9Dx − 9y = 27t − 18
+ −9Dx + D 2 y = D(−18t + 6)
.
2
D y − 9y = 27t − 18 + D(−18t + 6)

Remember that D means derivative, so D(−18t + 6) = d

dt
(−18t + 6) = −18. So

′′
y − 9y = 27t − 36.

s2 − 9 = 0 ⇒ s = ±3 ⇒ yh = c1 e−3t + c2 e 3t , and yp = At + B gives y = −3t + 4 after plugging

−3t 3t
in, so y = c1 e + c2 e − 3t + 4.
Back substituting for x into the second equation,

(c1 e
1 −3t 3t ′ 1 −3t 1 3t
x=− + c2 e − 3t + 4 − 18t + 6 = c1 e − c2 e − 3 + 2t + 2.
9 3 3
So the general solution is
1 −3t 3t
x = c e − 13 c2e + 2t − 1,
3 1
−3t 3t □.
y = c1 e + c2 e − 3t + 4.

Example 4.1.7. Solve the initial value problem

dx
dt
= 2x − y, x(0) = 0,
, .
dy
= x + 2 sin(t) y(0) = 0.
dt

solution: We solve by substitution. To avoid fractions, we use the second equation to

solve for x.
dy
x= − 2 sin(t).
dt
Plugging into the first equation gives

dy ′ dy
− 2 sin(t) = 2 − 2 sin(t) − y,
dt dt

Or
′′ ′
y − 2y + y = 2 cos(t) − 4 sin(t).

C. Sprouse —Math 280 Notes Page 317 DRAFT August 26, 2022
4.1. Constant Coefficient Systems: Substitution/Elimination 318

Solving by undetermined coefficients,

t t
y = c1e + c2te − 2 cos(t) − sin(t).

Back substituting into the second equation gives

− 2 sin(t) = (c1 + c2 )e + c2 te + cos(t).

dy t t
x=
dt
To solve the IVP,
x(0) = c1 + c2 + 1 = 0,
y(0) = c1 − 2 = 0
So c1 = 2 and c2 = −1, and
t t
x(t) = e − te − cos(t),
t t □.
y(t) = 2e − te − 2 cos(t) − sin(t).

1
t
0.5 1 1.5 2
−1

−2

−3

−4
x(t)
−5 y(t)

C. Sprouse —Math 280 Notes Page 318 DRAFT August 26, 2022
 4.1. Constant Coefficient Systems: Substitution/Elimination 319

Homework 4.1:
A) Find the general solution by substitution or elimination.
1.
dx 5t
dt
= x + 3y + 11e ,
dy 5t
dt
= x − y − 8e .
solution: For this example we use the method of substitution. We can either use the first
equation to solve for y, or the second equation tosolve for x. To avoid fractions, we use the
second equation to solve for x:
dy 5t
x= + y + 8e .
dt
We then plug this into the first equation:
′
dy 5t dy 5t 5t
+ y + 8e = + y + 8e + 3y + 11e
dt dt

d2y dy 5t dy 5t 5t d2y 5t
+ + 40e = + y + 8e + 3y + 11e ⇒ − 4y = −21e .
dt 2 dt dt dt 2
This can be solved by undetermined coefficients. The homogenous solution is
−2t 2t 5t 5t
y = c1 e + c2 e and the particular solution isyp = Ae = −e . So
−2t 2t 5t
y(t) = c1 e + c 2e − e .
To solve for x, we do NOT solve another differential equation. Aside from being more work,
this is because the constants for x are now fixed in terms of y, and solving another
differential equation would introduce new constants which we don’t want. To solve for x we
simplyuse the equation for x that we derived above, and plug in y:
dy 5t −2t 2t 5t ′ −2t 2t 5t 5t
x = + y + 8e = c1 e + c2 e − e + c1 e + c2 e − e + 8e
dt
−2t 2t 5t −2t 2t 5t 5t
= −2c1 e + 2c2e − 5e + c1 e + c2 e − e + 8e
−2t 2t 5t
= −c1 e + 3c2 e + 2e □.

2.
dx
dt
= 2x + y − 3t,
dy
dt
= 4x − y + 21t.
d
solution: We replace the derivative by dt
= D, and rewrite as an algebraic system:

Dx = 2x + y − 3t (D − 2)x − y = −3t
−4x + (D + 1)y
⇒ .
Dy = 4x − y + 21t = 21t
We can eliminate either x or y. To eliminate x, multiply the top equation by 4 and the
bottom by D − 2:
4(D − 2)x − 4y = 4(−3t) 4(D − 2)x − 4y = −12t
−4(D − 2)x + (D − 2)(D + 1)y (D − 2)(21t) −4(D − 2)x + (D − 2)(D + 1)y
⇒ .
= = −42t + 21
d
In the above, remember that when applying D to a function, D means dt
.Adding the
equtions,

(D − 2)(D + 1)y − 4y = (−42t + 21) − 12t (D − D − 6)y = −54t + 21.

2
⇒

C. Sprouse —Math 280 Notes Page 319 DRAFT August 26, 2022
4.1. Constant Coefficient Systems: Substitution/Elimination 320

′′ ′ −2t 3t
This gives y −y −6y = −54t +21. Using undetermined coefficients, yh = c1 e +c2 e and yp = At+B =
9t − 5.Now we have to find x. Again, we cannot solve another differential equation, but instead must
′
find x in terms of y. Going backto the original equations, y = 4x − y + 21t can be solved for x:
1 ′ 1 21
x = y + y+ t
4 4 4
1 −2t 3t ′ 1 −2t 3t 21
= c1 e + c2e + 9t − 5 + c e + c2 e + 9t − 5 + t
4 4 1 4
−2 −2t 3 3t 9 1 −2t 1 3t 9 5 21
= c e + c2 e + + c2 e + c 2e + t − − t
4 1 4 4 4 4 4 4 4
1 −2t 3t
= − c1 e + c2e − 3t + 1 □ .
4

3. 12.
dx dx
dt
= −y + 1 dt
= 2x + y
dy dy
dt
= x−2 dt
= 4x + 2y
4. 13.
dx dx
dt
= 4y − sin(t) dt
= 3x − y
dy dy
dt
= x + 4 cos(t) dt
= x+y
5. 14.
dx −t dx
dt
= y − 3e dt
= x −y
dy dy
dt
= −2x + 3y dt
= 2x − y
6. 15.
dx dx
dt
= x − 2y dy
= 2x − y
dy 3t dy
dt
= −x + 2e = x + 2y
dt
7. 16.
−x + y + 2e −t
dx
dt
= y − 2t dx
=
dt
dy −t
dt
= 2x − y − 1 dy
= x−y−e
dt
8. 17.
dx 2t
dy
= −y + t dx
= 3x + y − 2e
dt
dy dy 2t
dt
= x − 2y dt
= 2x + 2y + 2e
9. 18.
dx dx
dt
= 2x − y dt
= x − 2y + t
dy dy
dt
= x + 2 cos(t) dt
= x − y + 2t
10. 19.
dx dx
dt
= −y + 5 dt
= x −y + t
dy dy
dt
= 5x + 2y dt
= 2x − 2y
11. 20.
dx dx
dt
= 3x + y dt
= x + 3y − 5 sin(t)
dy dy
dt
= 2x + 4y dt
= x − y + 5 cos(t)

C. Sprouse —Math 280 Notes Page 320 DRAFT August 26, 2022
 4.1. Constant Coefficient Systems: Substitution/Elimination 321

B) Solve the initial value problems by substitution or elimination.

21. 31.
dx 2t dx
dt
= y + 3e x(0) = 0, dt
= 2x + y x(0) = 2,
, . , .
dy
= x y(0) = 1 dy
= 2x + 3y y(0) = 1
dt dt
22. 32.
dx −t
dt
= 4y + 2e x(0) = 0, dx
= 2x − 3y x(0) = 1,
, . dt , .
dy −t y(0) = 0
= −x + 3e dy
= x − 2y y(0) = −1
dt dt
23. 33.
dx
dt
= 2y + 4 x(0) = 2, dx
= x−y x(0) = 1,
, . dt , .
dy
= 2x − 6 y(0) = 3 dy
= x + 3y y(0) = 2
dt dt
24. 34.
dx −t
= −y + 3e x(0) = −1, dx
= 2x − 5y
dt , . dt x(0) = 1,
dy
= 2x + 3y y(0) = 1 dy ,
y(0) = −1
.
dt
dt
= x − 2y
25. 35.
dx
= y+3 x(0) = −2, dx
= 2x + y
dt , . x(0) = 2,
dy
= 3x − 2y y(0) = 1 dt , .
dt
dy
= −x + 2y y(0) = 1
dt
26.
dx
= x+y 36.
x(0) = 1, dx
dt , . = x+y −5 x(0) = 4,
dy
= 2x + 4t y(0) = 0 dt , .
dt dy
= 3x − y + 1 y(0) = 3
dt
27.
dx
= 2x − y x(0) = 0, 37.
dx
dt , . = 2x − 2y x(0) = 1,
= x + 4e−t y(0) = 0
dy dt
2t , .
dt dy
= x −y + e y(0) = 0
dt
28.
dx
= −y + 2 sin(t) x(0) = 1, 38.
dt dx t
dy ,
y(0) = −1
. = x + y − 3e x(0) = 5,
= x − 2y dt , .
dt dy
= −x + 3y + 2e t y(0) = 1
dt
29.
39.
dx 2t
dt
= y + 4 cos(t) x(0) = −1, dx
= x − y − 3e x(0) = 0,
, . dt , .
dy
= 3x − 2y − 2 sin(t) y(0) = 2 dy
= 2x − y + e
2t y(0) = 0
dt dt
30. 40.
dx dx
dt
= 4x + 5y − 2 x(0) = 0, dt
= x−y +3 x(0) = 3,
, . , .
dy
= −x + 3 y(0) = 0 dy
= −x + y − 1 y(0) = 1
dt dt

C. Sprouse —Math 280 Notes Page 321 DRAFT August 26, 2022
4.3. First Order Systems: General Theory 336

4.3 First Order Systems: General Theory

4.3.1 Basic Definitions
Definition 4.3.1. A first order system of differential equations is a set of equations of the
form
d x1
dt
= F(x1 , ...xn , t),
⋮ .
d xn
dt
= F(x1 , ..., xn , t).
In vector form, this can be written as
d x⃗
⃗ x,
= F( ⃗ t).
dt
Here, x1, ..., xn are functions of t, which (as usual) we would like to explicitly solve for.
Example 4.3.2. The following are examples of first order systems.
d x1
dt
= 3x1 − 2x2 dx
dt
= x2 − y 2 dx
dt
= 2y + t3
d x2 , dy , dy 2
dt
= x1 + 5x2 + cos(t) dt
= 2xy dt
= x − ty
From the above examples, we have that the definition of first order systems is quite
general and includes both nonlinear and time-dependent systems. However, it is still
required that we have both n equations and n unknowns (That is, we only consider
‘square’systems.) This is because we want to have unique solutions to initial value prob-
lems. So, like the algebraic systems of Chapter 3, we need to have the same number of
equations as unknowns.
Theorem 4.3.3. Suppose that
d x⃗
⃗ x,
= F( ⃗ t)
dt
⃗ x,
is a general first order system as above. If F( ⃗ t) and ∂
(x⃗, t)
exist and are continuous around
(⃗
x0 , t⃗0 ), then there exists ε > 0 such that there is a unique solution x(t)
∂x i
⃗ to the IVP
dx⃗ ⃗ x,t)
= F(⃗
⃗ 0)
dt
x(t = x⃗0
on (−ε, ε).
This theorem is analogous to, and proved in exactly the same way as, the Fundamental
Existence/Uniqueness Theorem for First Order IVP’s in Chapter 2.
In order to say anything more about the solutions to first order systems, we need to
restrict to more specific cases. We have two basic kinds of systems, the treatment of each
being very different: autonomous systems and linear systems.
Definition 4.3.4. A first order system is said to be autonomous if the right hand side does not
depend on t. That is, an autonomous system first order system is of the form

F(x1 , ...xn ),
dx1
dt
=
⋮
dxn
dt
= F(x1 , ..., xn ).

C. Sprouse —Math 280 Notes Page 336 DRAFT August 26, 2022
4.3. First Order Systems: General Theory 337

Example 4.3.5. The systems

= x1 (3 − x1 − x2 )
dx dx 2 2 dx
= x −y = y
= x2 (5 + x1 − x2 )
dt , dt , dt
dy dy dy
dt dt
= 2xy dt
= − sin(x)

are autonomous.

The above systems are all autonomous, but not linear. For these, it is typically difficult
or impossible to write explicit solutions. They can still be understood very well though,
by looking at them qualitatively, and in particular graphically by sketching the solutions.
But since that topic would take us far afield at this point, we will now focus on linear
systems instead.

Definition 4.3.6. A first order system is said to be linear if it is of the form

= a11(t)x1 + ... + a1n(t)xn + f 1 (t),

d x1
dt
⋮
= an1 (t)x1 + ... + ann (t)xn + f n (t).
d xn
dt

Or, in matrix notation, a first order linear system is an equation of the form

d x⃗
= A(t)x⃗ + f⃗(t).
dt

Example 4.3.7.

d x1 dx t dx 2 3
dt
= 3x1 − 2x2 dt
= 3x + ty − e dt
= t x−t y
, ,
= t 3x + t2 y
dy dy
d x2
dt
= x1 + 5x2 + cos(t) dt
= 2x − ty + et dt

are linear first order systems. In matrix notation, we have

d x⃗ 3 −2 0 3 t −e
t
d x⃗ 2
t −t
3
= x⃗ + , x⃗ + t , = 3 ⃗
2 x.
dt 1 5 cos(t) 2 −t e dt t t

As we can see above, for the linear system

d x⃗
= A(t)x⃗ + f⃗(t)
dt
we can write
⎡
⎢ a11(t) ⋯ a1n (t)⎤ ⎥
⎢
⎢ ⎥
⎥
⎢
A(t) = ⎢
⃗ ⎥
⎥
⎢
⎢
⋮ ⋮ ⎥
⎥
.
⎢
⎣ a n1 (t) ⋯ ann (t)⎥
⎦
If all of the functions aij (t) are in fact constant, that is A is a constant matrix, we say that
the system is constant coefficient. Such systems are also often called time invariant, or
linear time invariant (LTI), particularly in engineering applications.

C. Sprouse —Math 280 Notes Page 337 DRAFT August 26, 2022
4.3. First Order Systems: General Theory 338

Example 4.3.8.
dx1 dx t dx
dt
= 3x1 − 2x2 dt
= x−y +e dt
= 3x + 2y
dx2 , dy 3 , dy
dt
= x1 + 5x2 + cos(t) dt
= 2x − t dt
= 4x − y
Are examples of constant coefficient (linear time invariant) systems. In matrix notation,

d x⃗ d x⃗ d x⃗
t
3 −2 0 1 −1 e 3 2
= x⃗ + , = x⃗ + 3 , = ⃗
x.
dt 1 5 cos(t) dt 2 0 −t dt 4 −1

Warning 4.3.9. In the vast majority of applications, the linear systems involved are constant-
coefficient. Constant coefficient systems are far more tractable than general linear systems.
Their solutions can always be solved for explicitly, as we discuss in the rest of this chapter. For
this reason people will often refer to ‘Linear Systems’, when what they really mean is constant-
coefficient, or LTI systems. So it is important to be mindful of context whenever linear systems
are being discussed.
Finally, as with linear ODE’s before, a linear system is said to be homogeneous if the
functions f i on the right hand side are = 0. That is,
Definition 4.3.10. A homogeneous linear system is of the form

= a11(t)x1 + ... + a1n(t)xn,

dx 1
dt
⋮ ,
dxn
dt
= an1 (t)x1 + ... + ann (t)xn.
Or in matrix notation,
d x⃗
= A(t)⃗
x.
dt
Example 4.3.11.
d x1 dx dx 2 3
dt
= x2 dt
= 3x + 2y dt
= t x−t y
, ,
= t 3 x + t 2y
d x2 dy dy
dt
= −x1 dt
= 4x − y dt
are homogeneous linear first order systems. In matrix notation these can be written as

d x⃗ 0 1 d x⃗ 3 2 d x⃗ 2
t −t
3
= ⃗
x, = ⃗
x, = 3 ⃗
2 x.
dt −1 0 dt 4 −1 dt t t
Again, as we have seen previously with linear ODE’s, there is a general theory of linear
systems that allows us to understand them fairly well. In particular, we have a simpler
and more satisfactory existence/uniqueness theorem:
Theorem 4.3.12. Consider the linear IVP
⎡ a11(t) ⋯ a1n (t)⎤ ⎡ f 1 (t)⎤
d x⃗ ⎢ ⎢
⎢
⎥
⎥
⎥
⎢
⎢
⎢
⎥
⎥
⎥
=⎢ ⎢ ⎥
⎥ ⃗ ⎢
⎢ ⎥
⎥
dt ⎢ ⎢ ⋮ ⋮ ⎥
⎥
x + ⎢
⎢
⋮ ⎥
⎥
,
⎢
⎣ n1
a (t) ⋯ a nn (t) ⎥
⎦ ⎢
⎣ n ⎦
f (t) ⎥

⃗ 0) = x⃗0.
x(t
Suppose aij (t) and f i (t) are continuous for all t and all 1 ≤ i, j ≤ n. Then for any x⃗0 there
⃗ to the IVP on (−∞, ∞).
exists a unique solution x(t)

C. Sprouse —Math 280 Notes Page 338 DRAFT August 26, 2022
4.3. First Order Systems: General Theory 339

So if the system is constant-coefficient, we only need the forcing functions f i (t) to

⃗
be continuous to get existence and uniqueness of solutions x(t) for all time. In the next
subsection we will go into the details of what these solutions must look like.

4.3.2 Linear Systems: Fundamental Solution Sets and Fundamental Matrices

Definition 4.3.13. For the homogeneous linear system

⎡ a11 (t) ⋯ a1n (t)⎤

d x⃗ ⎢
⎢ ⎥
⎥
=⎢
⎢
⎢
⎢
⎥
⎥
⎥
⎥ ⃗
dt ⎢ ⎥
⋮ ⋮ x, (⋆)
⎢
⎣ a n1 (t) ⋯ a nn (t) ⎥
⎦

suppose that ai j (t) are all continuous on (a, b). Then a set of n solutions {⃗
x1 (t), ...x⃗n (t)} is
called a fundamental solution set if

⎡
⎢ ∣ ∣ ⎤ ⎥
⎢
⎢ ⎥
⎥
⎢
det ⎢ ⃗ (t) ⃗ (t)⎥
⎥
⎢
⎢
x ⋯ x ⎥
⎥
⎢ ⎥
1
∣ ∣
n
⎣ ⎦

is nonzero on all of (a,b).

Alternatively, we could say that a set of solutions {⃗ x1 (t), ...x⃗n (t)} is a fundamental
solution set if for each t0 ∈ (a, b), we have that the n vectors {x⃗1 (t0 ), ...x⃗n (t0 )} are linearly
independent.

Example 4.3.14. Find a fundamental solution set to the system

dx 1 2
= ⃗
x.
dt −3 6

solution: We can solve the system

dx
dt
= 2x − y
dy
dt
= 3x + 6y

3t 5t
by substitution/elimination. We get x(t) = c1 e + c2 e , and then substituting into the
3t 5t
first equation gives y(t) = −c1 e − 3c2 e . Setting c1 = 1, c2 = 0 and then c1 = 0, c2 = 1 gives

3t 5t
x⃗1 (t) = x⃗2 (t) =
e e
3t , 5t .
−e −3e

This is then a fundamental solution set on (−∞, ∞) because

3t 5t
e e 3t 5t 3t 5t 8t
det 5t = −3e e + e e = −2e
−e3t −3e

is nonzero for all t ∈ R.

C. Sprouse —Math 280 Notes Page 339 DRAFT August 26, 2022
4.3. First Order Systems: General Theory 340

Note that the general solution in the above example can be seen to be

3t 5t
e e
⃗ = c1
x(t) 3t + c2 5t .
−e −3e

This will be true in general:

Theorem 4.3.15. 1. For the homogeneous system ⋆, if aij (t) are continuous on
(a,b) then ⋆ has a fundamental solution set {x⃗1 (t),..., x⃗n (t)} on (a,b).

2. If {x⃗1 (t), ..., x⃗n (t)} is any fundamental solution set to ⋆, then the general solu-
tion to ⋆ is
⃗ = c1 x⃗1(t) + ... + cn x⃗n (t).
x(t)

Definition 4.3.16. For a linear homogeneous system

d x⃗
= A(t)⃗
x,
dt

a fundamental matrix Φ(t) is an n × n matrix such that

Φ (t) = A(t)Φ(t),
′

and det(Φ(t)) is nonzero on all of (a,b).

Finding a fundamental matrix Φ(t) is the same thing as finding a fundamental solu-
tion set {x⃗1 (t), ..., x⃗n (t)}. This is because it is easy to see that Φ(t) is a fundamental matrix
if {x⃗1 (t),..., x⃗n (t)} form the columns of Φ. Because if

d x⃗1 d x⃗n
= A(t)⃗
x1 , ... , = A(t)x⃗n ,
dt dt
then

⎡∣ ∣⎤ ⎡
⎢ ∣ ∣ ⎤
⎥ ⎡ ∣ ∣ ⎤ ⎡ ∣ ∣⎤
d ⎢⎢
⎢
⎥
⎥
⎥ ⎢
⎢ ⎥
d x⃗n ⎥
⎢
⎢
⎢
⎥
⎥
⎥
⎢
⎢
⎢
⎥
⎥
⎢x⃗1 ⋯ x⃗n ⎥ ⎢
⎥=⎢ ⎥
⎥= ⎢ xn ⎥
⎥ = A(t) ⎢ x⃗1 ⋯ x⃗n ⎥
⎥
dt ⎢ ⎢ dt ⎥ ⎢A(t)x⃗1 ⋯ A(t)⃗ ⎢ ⎥.
x⃗1
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
d
⎢ ⎥ ⎢ ∣ ⎥ ⎢ ⎥ ⎢ ⎥
⋯
⎢
⎣∣ ∣⎥⎦ ⎢ ∣ ⎥ ⎢
⎣ ∣ ∣ ⎥
⎦ ⎢
⎣ ∣ ∣ ⎥
⎦
⎣ ⎦
dt

So
⎡
⎢ ∣ ∣⎤ ⎥
⎢
⎢ ⎥
⎥
Φ(t) = ⎢
⎢
⎢ ⃗ ⃗ ⎥
⎥
⎥
⎢ ⎥
x ⋯ x .
⎢
⎣∣ ∣⎦ ⎥
1 n

Example 4.3.17. Find a fundamental matrix for the system

dx 1 2
= ⃗
x.
dt −3 6

C. Sprouse —Math 280 Notes Page 340 DRAFT August 26, 2022
4.3. First Order Systems: General Theory 341

solution: We have seen that

3t 5t
{x⃗1(t), x⃗2(t)} =
e e
3t , 5t
−e −3e

is a fundamental solution set to the system, from the previous example. Therefore a
fundamental matrix is
3t 5t
e e
Φ(t) = 3t 5t .
−e −3e

d x⃗
Conversely, if Φ(t) is any fundamental matrix to dt
= A(t)⃗
x, the columns of Φ(t) will
always be a fundamental solution set.
2 cos(3t) 2 sin(3t)
Example 4.3.18. Show that Φ(t) = is a funda-
cos(3t) + 3 sin(3t) −3 cos(3t) + sin(3t)
d x⃗ 1 −2
mental matrix to = ⃗ Then use this to find the general solution to
x.
dt 5 −1
dx
dt
= x − 2y
dy .
dt
= 5x − y

solution: Taking derivatives,

d 2 cos(3t) 2 sin(3t) −6 sin(3t) 6 cos(3t)

= .
dt cos(3t) + 3sin(3t) −3 cos(3t) + sin(3t) −3sin(3t) + 9cos(3t) 9sin(3t) + 3 cos(3t)

On the other hand,

1 −2 2 cos(3t) 2 sin(3t)
5 −1 cos(3t) + 3 sin(3t) −3 cos(3t) + sin(3t)
2 cos(3t) − 2 cos(3t) − 6 sin(3t) 2 sin(3t) − 2 sin(3t) + 6 cos(3t)
=
10 cos(3t) − cos(3t) − 3 sin(3t) 10sin(3t) + 3 cos(3t) − sin(3t)
−6 sin(3t) 6cos(3t)
= .
−3 sin(3t) + 9cos(3t) 9 sin(3t) + 3cos(3t)

Since we got the same matrix both times, this shows

Φ (t) =
′ 1 −2
Φ(t).
5 −1

Also we have
2 cos(3t) 2 sin(3t)
det(Φ(t)) =
cos(3t) + 3 sin(3t) 3 cos(3t) − sin(3t)
6 cos (3t) − 2cos(3t) sin(3t) + 2cos(3t) sin(3t) + 6 sin (t) = 6,
2 2
=

which is nonzero on (−∞, ∞). So Φ(t) is a fundamental matrix on (−∞, ∞).

1 −2
The general solution to ddtx⃗ = x⃗ is then
5 −1

2 cos(3t) 2 sin(3t)
⃗ = c1
x(t) + c2 .
cos(3t) + 3 sin(3t) −3cos(3t) + sin(3t)

C. Sprouse —Math 280 Notes Page 341 DRAFT August 26, 2022
4.3. First Order Systems: General Theory 342

To get the general solution to the system

dx
dt
= x − 2y
dy
dt
= 5x − y

we can add the above vectors on the right-hand side to get

x(t) = 2c1 cos(3t) + 2c2 sin(3t),

y(t) = (c1 − 3c2 )cos(3t) + (3c1 + c2 ) sin(3t).

Finally, since fundamental solution sets to a system are not unique (there are many
possible choices), we also have that the fundamental matrix of a system is not unique.
There is a different fundamental matrix Φ(t) for any choice of fundamental solution set
{⃗
x1 (t), ..., x⃗n(t)}. However, there is one choice for Φ(t) that is generally nicest for appli-
cations. This is the fundamental matrix with Φ(t0 ) = I. It is unique because each column
satisfies an IVP for which the vector-valued solution x⃗i (t) is unique. This matrix is im-
portant enough to have a special name.

Definition 4.3.19. Given a system ddtx⃗ = A(t)⃗ ⃗ 0 ) = x⃗0 ,

x, with an initial value specified at x(t
the principal fundamental matrix, or transition matrix is the fundamental matrix Ψ(t)
such that Ψ(t0 ) = I.

It is easy to calculate Ψ(t), as long as you remember the order of multiplication. If B

is any invertible n×n matrix, and Φ(t) any fundamental matrix to ddtx⃗ = A(t)⃗
x, then Φ(t)B
is also a fundamental matrix, because

(ΦB) = Φ B = (AΦ)B = A(ΦB).

′ ′

But note that BΦ is not a fundamental matrix, as (BΦ) = B(Φ ) = B(AΦ) ≠ A(BΦ), since
′ ′

in general for matrices, AB ≠ BA. But with this in mind, since Φ(t)B is a fundamental
matrix for any invertible B, we can choose B = Φ (t0 ) so that
−1

(t0 ) = Φ(t0 )Φ (t0 ) = I.

−1 −1
Φ(t)Φ t=t0

So we have:

(t0 ).
−1
Ψ(t) = Φ(t)Φ

Example 4.3.20. Find the transition matrix Ψ(t) for

(t0 = 0).
dx 1 2
= ⃗
x.
dt −3 6

C. Sprouse —Math 280 Notes Page 342 DRAFT August 26, 2022
4.3. First Order Systems: General Theory 343

solution: From a previous example we have a fundamental matrix

3t 5t
e e
Φ(t) = .
−e 3t −3e 5t

So
3t 5t −1
(0) =
−1 e e 1 1
Ψ(t) = Φ(t)Φ 3t 5t
−e −3e −1 −3
3t 5t 3 1
e e 2 2
= 3t 5t
−e −3e − 12 − 12
3 3t 5t 1 3t 5t
e − 12 e e − 12 e
= 2
3t 5t
2
5t 5t .
− 32 e + 32 e − 12 e + 32 e

Recall that for Φ(t) to be a fundamental matrix on (a,b), it is also necessary for
det(Φ(t)) to be nonzero on (a,b). We did not need to verify this for Ψ(t) above, as Ψ(0) =
I implies that det(Ψ(0)) = det(I ) = 1. And furthermore, this implies that det(Ψ(t)) is
nonzero for all t ∈ R in this case, due to the following theorem.

Theorem 4.3.21 (Abel’s Theorem). Let Φ(t) be any matrix solution to

Φ (t) = A(t)Φ(t).
′

Suppose A(t) has continuous entries for a < t < b. Then for t0 ∈ (a,b) the determinant of Φ on
(a, b) is given by
det(Φ(t)) = det(Φ(t0 ))e
∫ tr(A) dt
,
where tr(A) = a11 (t) + ... + ann (t) is the sum of the diagonal entries of A.

This is the matrix form of what we called of Abel’s Theorem before, Theorem 2.1.24.
As we did in Chapter 3, we can leave the proof of this as an exercise. However we can
now see that it means the determinant of Ψ in the previous example is nonzero, as we
have
1 2
tr = 1+6 = 7
−3 6
so using t0 = 0 and Ψ(0) = I in the above formula gives
7t
det(Ψ) = e ,

which is nonzero on all of R.

Example 4.3.22. Find the transition matrix Ψ(t) for

d x⃗
(t0 = 0).
1 −2
= x⃗
dt 5 −1

C. Sprouse —Math 280 Notes Page 343 DRAFT August 26, 2022
4.3. First Order Systems: General Theory 344

solution: Again, from an example above we have a fundamental matrix

2cos(3t) 2 sin(3t)
Φ(t) = .
cos(3t) + 3 sin(3t) −3cos(3t) + sin(3t)
So
−1
(0) =
−1 2 cos(3t) 2 sin(3t) 2 0
Ψ(t) = Φ(t)Φ
cos(3t) + 3 sin(3t) −3 cos(3t) + sin(3t) 1 −3
1
2 cos(3t) 2 sin(3t) 0
= 2
1
cos(3t) + 3 sin(3t) −3 cos(3t) + sin(3t)
6
− 13
cos(3t) + 13 sin(3t) − 23 sin(3t)
= 5 .
3
sin(3t) cos(3t) − 13 sin(3t)

At At
Remark 4.3.23. If A is a constant matrix, then it turns out that Ψ(t) = e , where e is what
is called the matrix exponential, which by definition is
∞
(At) .
At 1 n
e =
n!
n=0

For a constant n × n matrix A, e satisfies (e ) = Ae , and e

At At ′ At A⋅0
= I . This characterization
At
gives concrete ways to compute Ψ(t) = e using linear algebra. However, this does not work
∫ A(s) ds
t
when A(t) is not constant. The matrix exponential e t0 is not necessarily a fundamental
⃗ More generally, (e
matrix for x⃗ = A(t)x. ) = B (t)e fails unless B(t)B (t) = B (t)B(t),
′ B(t) ′ ′ B(t) ′ ′

which is not guaranteed in general.

4.3.3 n-th Order ODE’s to First Order Systems

A general n-th order ODE in standard form looks like
n n−1
d y d y dy
=F , ..., , y, t .
dtn dt n−1 dt
For nonlinear ODE’s like this, it is extremely difficult to get explicit solutions. So to be
able to say something about them, it is often helpful to convert to a first order system,
where the n-th derivative is replaced by n equations with one derivative each. This turns
out to be quite easy to do. Just set
u1 = y
dy
u2 = dt
⋮ .
n−1
d y
un = dt n−1
Then we get the following system:
du1
dt
= u2
du2
= u3
dt .
⋮
dun
dt
= F(un , ..., u2 , u1 , t)

C. Sprouse —Math 280 Notes Page 344 DRAFT August 26, 2022
4.3. First Order Systems: General Theory 345

The last equation because

n n−1
dun d y d y dy
= n =F n−1
, ..., , y,t = F(un , ..., u2 , u1, t).
dt dt dt dt

= (y ) − (y ) + y cos(t).
′′′′ ′′′ 2 ′ 3 4
Example 4.3.24. Find a first order system for y
2 3
dy d y d y
solution: As above, set u1 = y, u2 = dt
, u3 = ,u
dt 2 4
= d t3
. Then

du 1
dt
= u2
du 2
dt
= u3
du 3
= u4
(u4 ) − (u2 ) + (u1 ) cos(t).
dt
du 4 2 3 4
dt
=

′′′ ′′ ′
Example 4.3.25. Find a first order system for y = 3y − yy .
′ ′′
solution: Set u = y, v = y ,w = y . Then the first order system is
′
u = v
′
v = w .
′
w = 3w − uv

For Linear ODE’s, we can write the new first order system as a matrix equation. Consider
the general n-th order linear ODE, with an (t) = 1:
(n) (n−1)
+ an−1(t)y + ... + a1 (t)y + a0(t)y = f (t).
′
y

With u1 = y, u2 = y ′, ..., un = y (n−1) the system is

du1
dt
= u2
du2
= u3
dt .
⋮
dun
dt
= −a0 (t)u1 − ... − an−1 (t)un + f (t)

The last line comes from using the original ODE to write
(n) (n−1)
y = −a0 (t)y − ... − an−1 (t)y + f (t).

In matrix form,

⎡ du 1 ⎤ ⎡
⎢ ⎤
⎥ ⎡
⎢ 0 ⎤⎥
⎢ dt ⎥ ⎢ ⎥ ⎢ ⎥
0 1 0 ⋯ 0
⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥ ⎢
⎢ 0 ⎥⎥
⎢
⎢ du 2 ⎥
⎥ ⎢
⎢
⋯ ⎥
⎥ ⎢
⎢ ⎥
⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
0 0 1 0
⎢
⎢ ⎥ ⎢
⎥=⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥
⎢
⎢ ⋮ ⎥ ⎥ ⎢
⎢ 0
⋮ ⎥
⎥
+ ⎢
⎢
⋮⋮ ⎥
⎥
.
dt
⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥
⎢
⎢ du n ⎥ ⎢ ⎥
⎥ ⎢ ⎥
⎣ dt ⎥ ⎦ ⎢ ⎢ ⎥
0 0 ⋯ 10
⎢
⎣ 0
−a (t) −a1 (t) −a 2 (t) ⋯ −an−1 (t)⎥
⎦ ⎢ ⎣f (t)⎥
⎦

C. Sprouse —Math 280 Notes Page 345 DRAFT August 26, 2022
4.3. First Order Systems: General Theory 346

′′′ ′′ ′ 3t
Example 4.3.26. Write y + ty − y + 2y = e as a first-order system.
′ ′′
solution: We have u = y, v = y ,w = y and
⎡
⎢ u⎤⎥
′ ⎡
⎢ 0⎤⎥ ⎡u⎤ ⎡ 0⎤
⎥⎢ ⎥ ⎢ ⎥
′
⎢ ⎥ ⎢ ⎢ ⎥ ⎢ ⎥
=
⎢ ⎥ ⎢ 1⎥ ⎢ ⎥ ⎢ 0⎥
u v 0 1
⎢
⎢ ⎥ ⎥
⎥ ⎢ ⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥
⎢ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
′
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
v = w ⇒ v = 0 0 v + .
w
′
= −tw + v − 2u + e
3t ⎢w ⎥
⎣ ⎦ ⎣ ⎢ −t 1 −2⎥ ⎢ w ⎥
⎦⎣ ⎦ ⎣ ⎦ ⎢ e
3t ⎥

In the case that the linear ODE is homogeneous (f (t) = 0), we can try to find the fun-
damental solution set and fundamental matrix, which we have recently defined. It turns
out that knowing a fundamental solution set to the first order system is equivalent to
knowing a fundamental solution set to the ODE!

Fact 4.3.27. If {y1(t), ..., yn (t)} is a fundamental solution set to the ODE
(n) (n−1)
+ an−1 (t)y + ... + a1(t)y + a0 (t)y = 0,
′
y

Then the fundamental solution set to the associated first order system is
⎧
⎪ ⎡ y1 (t) ⎤ ⎡ yn (t) ⎤ ⎫
⎪
⎪
⎪ ⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥ ⎪
⎪
⎪
⎪⎢⎢ (t) ⎥
⎥ ⎢
⎢ (t) ⎥
⎥ ⎪
⎪
⎪ ⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥ ⎪
′ ′
{u⃗1 , ..., u⃗n } = ⎨ ⎢ ⎥ ⎢ ⎥ ⎬
y y
⎪ ⎢ ⎥ ⎢ ⎥ ⎪
1
⎢ ⎥ ⎢ ⎥
n
⎪⎢ ⎥ ⎢ ⎥ ⎪
,⋯, .
⎪
⎪ ⎢ ⎥ ⎢ ⎥ ⎪
⎪
⋮ ⋮
⎪ ⎢
⎢ ⎥ ⎢ ⎥⎪
(t)⎥ ⎢ (t)⎥
(n−1) (n−1)
⎪
⎩⎣y1 ⎦ ⎣yn ⎦⎪⎭
The fundamental matrix is

⎡
⎢ y1 (t) yn(t) ⎤ ⎥
⎢ ⎥
⋯
⎢
⎢ (t) (t) ⎥
⎥
⎢ ⎥
Φ(t) = ⎢ ⎥
′ ′
⎢ ⎥
y ⋯ y
⎢ ⎥
1
⎢ ⎥
n
⎢ ⎥
.
⎢ ⎥
⋮ ⋮
⎢
⎢ ⎥
(t)⎥
(n−1) (n−1)
⎣ y1 (t) ⋯ yn ⎦

′′ ′
Example 4.3.28. Find a first order system for y − 6y + 9y = 0. Then find a fundamental
solution set and fundamental matrix for the system.
′
solution: The system is u = y, v = y and
′
u 0 1 u
= .
v −9 6 v

Since s − 6s + 9 = (s − 3) , the ODE y − 6y + 9y = 0 has F.S.S. {e ,te }. So the F.S.S. to

2 2 ′′ ′ 3t 3t

the first order system above is

e 3t e 3t
{u⃗1 , u⃗2 } = 3t , 3t 3t .
3e e + 3te
A fundamental matrix is
3t 3t
e te
Φ(t) = 3t 3t 3t .
3e e + 3te

C. Sprouse —Math 280 Notes Page 346 DRAFT August 26, 2022
4.3. First Order Systems: General Theory 347

C. Sprouse —Math 280 Notes Page 347 DRAFT August 26, 2022
 4.3. First Order Systems: General Theory 348

Homework 4.3:
A) Solve the system by substitution/elimination. Then use your general solution to:
a) Find the general solution in vector form.
b) Find the fundamental solution set in vector form.
c) Find a fundamental matrix.

1.
1 −1
x⃗ = ⃗
′
x.
5 −3
solution: The matrix gives the equations
dx
dt
= x−y
dy .
dt
= 5x − 3y

Plugging y = − dx
dt
+ x into the second equation gives
′′ ′ −t −t
x + 2x + 2x = 0 ⇒ x = c1 e cos(t) + c2e sin(t).

Then we can plug x into

y = (2c1 − c2)e cos(t) + (c1 + 2c2 )e

′ −t −t
y = −x + x ⇒ sin(t).

Note that y can be rearranged to

y(t) = c1 (2e sin(t)) + c2 (−e

−t −t −t −t
cos(t) + e cos(t) + 2e sin(t)).

This gives the general solution

−t −t
e cos(t) e sin(t)
⃗ = c1
x(t) −t −t + c2 −t −t .
2e cos(t) + e sin(t) −e cos(t) + 2e sin(t)

Aside from rearranging, we could get the vectors in the general solutions by choosing
constants.
−t
x⃗1 (t) =
e cos(t)
c1 = 1, c2 = 0 ⇒ −t −t
2e cos(t) + e sin(t)
−t
x⃗2 (t) =
e sin(t)
c1 = 0, c2 = 1 ⇒ −t −t .
−e cos(t) + 2e sin(t)
A fundamental matrix is then
−t −t
e cos(t) e sin(t)
Φ(t) = −t −t −t −t □.
2e cos(t) + e sin(t) −e cos(t) + 2e sin(t)

2. 4.
0 1 1 1
x⃗ = ⃗ x⃗ = ⃗
′ ′
x. x.
−1 0 −2 4
3. 5.
0 1 2 3
x⃗ = ⃗ x⃗ = ⃗
′ ′
x. x.
9 0 1 0

C. Sprouse —Math 280 Notes Page 348 DRAFT August 26, 2022
 4.3. First Order Systems: General Theory 349

6. 7.
3 1 1 −1
x⃗ = ⃗ x⃗ = ⃗
′ ′
x. x.
−1 1 5 −1

B) For the given ODE:

a) Convert the higher order ODE to a first order system in matrix form.
b) Find a fundamental solution set to the ODE, and then use it to find a fundamental solution set to the
system in vector form.
c) Write the general solution to the system in vector form.
8.
′′ ′
y − 6y + 10y = 0.
′ ′
solution: To write as a first order system we let u = y and v = y . Then u = v and
′ ′′ ′ ′
v = y = 6y − 10y = −10y + 6y . The first order system is then
′
u 0 1 u
= .
v −10 6 v

′′ ′ 2
The characteristic equation for y − 6y + 10y = 0 is s − 6s + 10 = 0 which has complex roots
s = 3 ± i. So the fundamental solution set to the ODE is {e cos(t),e sin(t)}. Now using
3t 3t
′
that u = y, v = y

3t
y1 (t) = e cos(t) u⃗1(t) =
3t e cos(t)
⇒ 3t 3t .
3e cos(t),−e sin(t)

e 3t sin(t)
y2 (t) = e cos(t) u⃗2(t) =
3t
⇒ 3t 3t .
3e sin(t) + e cos(t)

So the fundamental solution set to the system is {u⃗1 (t), u⃗2 (t)}, and the general solution is

3t 3t
e cos(t) e sin(t)
⃗ = c1
u(t) + c2 □.
3e3t cos(t) − e 3t sin(t) 3e 3t sin(t) + e 3t cos(t)

9. 14.
′′ ′ ′′ ′
y − 11y + 24y = 0 y + 2y + 2y = 0
10. 15.
′′ ′′′ ′
y + 9y = 0 y − 25y = 0
11. 16.
′′ ′ ′′′ ′′
y − 6y + 9y = 0 y − 2y = 0
12. 17.
′′ ′ ′′′′
y + 3y + 2y = 0 y −y = 0
13. 18.
′′ ′ ′′′′ ′′
y + 2y + y = 0 y + 5y + 4y = 0

C) For the given system:

a) Verify that the given Ψ(t) is the transition matrix by checking that Ψ = AΨ and Ψ(t0 ) = I.
′

b) Use Ψ(t) to solve the initial value problem.

C. Sprouse —Math 280 Notes Page 349 DRAFT August 26, 2022
4.3. First Order Systems: General Theory 350

19.
1 7 −2 −t 3 + 2t 5 t 3 − t5 dx
= 7
x− 2
y, x(1) = −2,
x⃗ = x⃗, Ψ = (t0=1),
′ dt t t
3 5 3 5 .
t 4 1 −2t + 2t 2t − t dy
= 4
x+ 1
y. y(1) = 3.
dt t t
′
solution: First check that Ψ satisfies Ψ = AΨ:
3 5 3 5 ′ 2 4 2 4
Ψ (t) =
′ −t + 2t t −t −3t + 10t 3t − 5t
3 5 3 5 = 2 4 2 4 .
−2t + 2t 2t − t −6t + 10t 6t − 5t

3 5 3 5
1 7 −2 −t + 2t t −t
AΨ = 3 5 3 5
t 4 1 −2t + 2t 2t − t
1 7(−t 3 + 2t 5) − 2(−2t 3 + 2t 5 ) 7(t 3 − t 5 ) − 2(2t 3 − t 5) 2
−3t + 10t
4 2
3t − 5t
4

t 4(−t 3 + 2t 5) + 1(−2t 3 + 2t 5 ) 4(t 3 − t 5 ) + 1(2t 3 − t 5)

= = 2 4 2 4 . ✓
−6t + 10t 6t − 5t

Then we also have

−1 + 2 1−1 1 0
Φ(1) = = . ✓
−2 + 2 2−1 0 1
The solution to the initial value problem is then
3 5 3 5 3 5
x(t) −t + 2t t −t −2 5t − 7t
= 3 5 3 5 = 3 5 □.
y(t) −2t + 2t 2t − t 3 10t − 7t

20.
−t 2t −t 2t dx
−7 6 3e − 2e −2e + 2e = −7x + 6y, x(0) = 1,
x⃗ = x⃗, Ψ = (t0 =0),
′ dt
−t 2t −t 3t .
−9 8 3e − 3e −2e + 3e dy
= −9x + 8y. y(0) = 2.
dt
21.
2t 2t 2t dx
1 1 e − te −te = x + y, x(0) = −1,
x⃗ = x⃗, Ψ = (t0 =0),
′ dt
2t 2t 2t .
−1 3 te e + te dy
= −x + 3y. y(0) = 1.
dt
22.
2t 2t dx
2 −1 e cos(t) −e sin(t) = 2x − y, x(0) = 3,
x⃗ = x⃗, Ψ = 2t (t0=0),
′ dt .
1 2 e sin(t) e2t cos(t) dy
= x + 2y. y(0) = −1.
dt
23.
dx y
1 0 −1 cos(2 ln(t)) sin(2 ln(t)) = −t , x(1) = 3,
x⃗ = x⃗, Ψ = (t0 =1),
′ dt .
t 4 0 2 sin(2 ln(t)) −2 cos(2 ln(t)) dy
= 4x
. y(1) = 2.
dt t
24.
dx 2 1
1 2 −1 1 + 2 ln(t) − ln(t) = x− y, x(1) = 2,
x⃗ = x⃗, Ψ = (t0 =1),
′ dt t t .
t 4 −2 4 ln(t) 1 − 2 ln(t) dy
= 4
x− 2
y. y(1) = 1.
dt t t
25.
2 3 2 3 dx 1 2
1 1 −2 2t − t 2t − 2t = x− y, x(1) = −1,
x⃗ = x⃗, Ψ = (t0 =1),
′ dt t t
2 3 2 3 .
t 1 4 −t + t −t + 2t dy
= 1
x+ 4
y. y(1) = 2.
dt t t

C. Sprouse —Math 280 Notes Page 350 DRAFT August 26, 2022
4.4. Eigenvalues and Eigenvectors 351

4.4 Eigenvalues and Eigenvectors

We have seen some ways to solve first order systems: by substitution and elimination, and
also by Laplace transforms. Both of these methods end up being quite cumbersome to do
by hand. They also don’t give a lot of insight into why the solutions look the way they do.
In the next section we will solve both of these issues for constant-coefficient homogeneous
systems. That is, we will now focus on

d x⃗
⃗
= Ax,
dt
where A is an n ×n constant matrix. There is an excellent general theory for these. It gives
a complete understanding of which matrices give which types of solutions, and it also lets
us solve these systems very quickly compared to the previous methods. To get to this, we
have to go back to linear algebra. In particular we need to know what eigenvalues and
eigenvectors are, for a square matrix A.

Definition 4.4.1. If A is an n × n matrix, then we say that λ is an eigenvalue of A if there

exists some nonzero vector v with
Av⃗ = λv.
⃗
v⃗ is called an eigenvector associated to λ.

Given a fixed n × n matrix, we’d like a way to find all of its eigenvalues. Then for
each eigenvalue we need to find the eigenvectors, or at least one eigenvector. One thing
to notice about this is that we do not have a general method for solving equations of the
form Av⃗ = λv, ⃗ where v is some unknown. What we learned in Chapter 2 was how to
solving Ax⃗ = b⃗ where A and b⃗ are given, and x⃗ is the unknown. So we need to rewrite the
eigenvalue equation so we can use that:

Av⃗ = λv⃗ ⇔ Av⃗ − λv⃗ = 0 ⇔ Av⃗ − λI v⃗ = 0 ⇔ (A − λI )v⃗ = 0.

This is now something that we do know how to solve. The equation (A − λI )v⃗ = 0 is a
homogeneous system, which we can solve by Gaussian elimination. Usually this equation
will only have the trivial solution for v⃗ (⃗
v=0⃗). It is only for certain values of λ that one
will get nontrivial solutions, and these values of λ are by definition the eigenvalues of A.
How do we find such λ? Recall from Chapter 2 that if B is an n×n matrix, then an equation
of the form Bx⃗ = 0 has nontrivial solutions for x⃗ if and only if B is not invertible (that is, M
is “singular”). Furthermore recall that this happens if and only if det(B) = 0. Therefore,
we have nontrivial solutions to the system (A − λI )v⃗ = 0 if and only if det(A − λI ) = 0.
So, the eigenvalues of A are precisely the solutions to det(A − λI) = 0. Furthermore, it
turns out that the equation det(A − λI ) = 0 is something that we can easily deal with... it
is always a polynomial of degree n!

C. Sprouse —Math 280 Notes Page 351 DRAFT August 26, 2022
4.4. Eigenvalues and Eigenvectors 352

Fact 4.4.2.

• If A is an n×n matrix, the eigenvalues of A are precisely the solutions to det(A −λI ) = 0,
which is a polynomial of degree n.

• For each eigenvalue λ, the eigenvectors are the solutions v⃗ to the matrix equation (A −
λI )v⃗ = 0.

This gives:
✬ ✩
To find the eigenvalues and eigenvectors of an n × n matrix A:

1. Solve det(A − λI ) for eigenvalues λ. (There will be at most n dis-

tinct values).

2. For each λ from part 1, solve (A − λI )v.

⃗ Every nonzero solution,
v⃗ ≠ 0, is an eigenvector of λ.
✫ ✪
1 2
Example 4.4.3. Find the eigenvalues and eigenvectors of A = .
4 3
solution:
1−λ 2
det(A − λI) =
4 3−λ
= (1 − λ)(3 − λ) − 8 = 3 − −4λ + λ − 8
2

= λ − 4λ − 5 = (λ + 1)(λ − 5).
2

So the eigenvalues are −1, 5. Then for each eigenvalue we can find all eigenvectors.
λ = −1:
We need to solve (A − λI )v = 0, where λ = −1.

1 −λ 2
A − λI = ,
4 3−λ
so
1 − (−1)
A − (−1)I =
2 2 2
3 − (−1)
= .
4 4 4
x
Therefore if we let v = , we have
y
2 2 x 0
= .
4 4 y 0
Which is
2x + 2y = 0
.
4x + 4y = 0
Now you may notice that the second equation is just a multiple of the first. This is
actually good! This must happen if we are to get nontrivial solutions, and eigenvectors are
by definition nonzero. Therefore, if something like this doesn’t happen, you’ve likely

C. Sprouse —Math 280 Notes Page 352 DRAFT August 26, 2022
4.4. Eigenvalues and Eigenvectors 353

done something wrong (such as made a mistake with the eigenvalues...). Anyway, we are
left with the single equation 2x + 2y = 0, or x + y = 0. Setting the parameter y = s we get
x = −s. Therefore the set of all eigenvalues for λ = −1 is

−s
s ∈R .
s

λ = 5:

1−5 2 −4 2
A − 5I = = .
4 3−5 4 −2
So solving (A − 5I)v = 0 we have

−4 2 x 0
= ,
4 −2 y 0
or
−4x + 2y = 0
.
4x − 2y = 0
Again, the second equation is a multiple of the first, which is what we want! So now
letting x = s we get y = 2s. Therefore the set of all eigenvalues for λ = 5 is

s
s ∈ R □.
2s

It is always a good idea to check that one has the right answer after doing this. And it
is also very easy to do. We have
1 2
A= .
4 3
For both λ = −1 and λ = 5 in the above, we can just take the eigenvector corresponding to
−1 1
s = 1. That is, we can take and . (Any other value for s would also work). Then
1 2
we simply note that

= (−1)
1 2 −1 1 −1
= , ✓
4 3 1 −1 1

and
1 2 1 5 1
= =5 . ✓
4 3 2 10 2
This is exactly what should happen by Definition 4.4.1!

C. Sprouse —Math 280 Notes Page 353 DRAFT August 26, 2022
4.5. 2-Dimensional Systems by Eigenvalues/Eigenvectors 354

4.5 2-Dimensional Systems by Eigenvalues/Eigenvectors

Here we will consider in detail the equation x⃗ = Ax,
⃗ where x(t)
⃗ is a vector in R and A is
′ 2

a fixed 2 × 2 matrix. That is,

′
x1 a b x1
′ =
x2 c d x2
From the previous section we know that det(A − λI ) is a quadratic polynomial. So A
has 2 eigenvalues, λ1 , λ2 . This means that there are 3 cases that we need to deal with:
1. λ1 and λ2 are real and distinct (λ1 ≠ λ2 ).

2. λ1 and λ2 are complex.

3. λ1 and λ2 are real and equal (λ1 = λ2 ).

4.5.1 Real distinct eigenvalues

The first case is by far the easiest. First let A be an n × n matrix, and suppose λ is an
⃗ Then if x(t)
⃗ = e v, ⃗ we can see that
λt
eigenvalue of A, with eigenvector v.

x⃗ (t) = e v⃗ = (e ) v⃗ = λe v,
′
⃗
′ λt λt ′ λt

since v⃗ is a constant vector. On the other hand, we have

⃗ = e Av⃗ = e (λv) = λe v.
⃗ = A(e v) ⃗
λt λt λt λt
Ax(t)

These give the same results! In other words,

Fact 4.5.1. If v⃗ is an eigenvector of A, with eigenvalue λ, then x⃗(t) = e v⃗ is always a solution

λt

to x⃗ = Ax.
⃗
′

Now we need another fact from linear algebra, which we state without proof:
Theorem 4.5.2. If A is an n × n matrix, and λ1 , ..., λn are eigenvalues of A, with eigenvectors
v⃗1 , ..., v⃗n , then {v⃗1, ..., v⃗n} is always linearly independent.
So if a 2 × 2 matrix A has real eigenvectors λ1 ≠ λ2 , then we can find two linearly
independent eigenvectors v⃗1 , v⃗2 , and two solutions x⃗1(t) = e λ1 t v⃗1 , x⃗2 (t) = eλ2 t v⃗2 to x⃗′ =
⃗ Now we want to find the general solution. From Section 4.3, we need to show that
Ax.
{x1 (t), x2 (t)} is a fundamental solution set. This happens if

⎛⎡⎢
⎢
∣ ∣ ⎤ ⎥
⎥ ⎞
det ⎜ ⎢
⎢ (t) (t) ⎥
⎥ ⎟
⎜⎢ ⃗
⎢ 1 ⃗ ⎥
⎥ ⎟ ≠ 0.
⎝⎢ ⎥
x x
⎢ ∣ ∣ ⎥
⎦⎠
2
⎣
But from Chapter 2, we know that
⎡
⎢ ∣ ∣ ⎤ ⎥ ⎡ ∣ ∣⎤ ⎡ ∣ ∣⎤
⎢
⎢ ⎥ ⎢
⎢ ⎥
⎥ ⎢
⎢ ⎥
⎥
⎢ λ2 t ⎥ ⎥ ⎢
⎢ ⎥
⎥ (λ1 +λ2 )t ⎢
⎢ ⎥
⎥
det ⎢
⎢ ⃗ ⃗ ⎥
⎥ ⎢
⎢ ⃗ ⃗ ⎥
⎥ ⎢
⎢ ⃗ ⃗ ⎥
⎥
⎢ ⎥
λ1 t λ1 t λ2 t
⎢ ⎥ ⎢ ⎥
= det = det
⎢ ⎥
e v e v e e v v e v v ,
⎢ ⎥ ⎢ ∣ ∣ ⎥ ⎢ ∣ ∣ ⎥
1 2 1 2
∣ ∣
1 2
⎣ ⎦ ⎣ ⎦ ⎣ ⎦

C. Sprouse —Math 280 Notes Page 354 DRAFT August 26, 2022
4.5. 2-Dimensional Systems by Eigenvalues/Eigenvectors 355

and since v⃗1 , v⃗2 are linearly independent, the determinant on the right is nonzero. That is
{x⃗1(t), x⃗2(t)} is always a fundamental solution set. This means that we have the general
solution:

⃗ = c 1e v⃗1 + c2 e v⃗2
λ1 t λ2 t
x(t)

Example 4.5.3. Find the general solution to

′
x1 3 −2 x1
= .
x2 2 −2 x2

solution: First we find the eigenvalues and eigenvectors of A.

3 −2 λ 0 3− λ −2
A − λI = − = ,
2 −2 0 λ 2 −2 − λ

= 0 ⇒ (3 − λ)(−2 − λ) + 4 = 0.
3− λ −2
det
2 −2 − λ
So‘

−6 + 2λ − 3λ + λ + 4 = 0 ⇒ λ − λ − 2 = 0 ⇒ (λ − 2)(λ + 1) = 0 ⇒ λ = 2, −1.
2 2

Now we need to find the eigenvectors for each eigenvalue. For λ = 2:

3−2 −2 1 −2
A − 2I = = .
2 −2 − 2 2 −4

1 −2 x 0 x − 2y = 0
= ⇒ ⇒ x − 2y = 0 ⇒ y = s, x = 2s.
2 4 y 0 2x − 4y = 0
Letting s = 1 we get the eigenvector
2
v⃗1 = .
1
For λ = −1:
A − (−I ) =
3+1 −2 4 −2
=
2 −2 + 1 2 −1
4 −2 x 4x − 2y = 0
⇒ ⇒ 2x − y = 0 ⇒ x = s, y = 2s
2 −1 y 2x − y = 0
Letting s = 1 we get the eigenvector
1
v⃗2 = .
2
So the general solution is
2 −t 1
⃗ = c1 e
2t
x(t) + c2 e .
1 2
This can of course be rewritten as
2t −t
x1 = 2c1 e + c2 e
2t −t .
x2 = c1e + 2c2 e

C. Sprouse —Math 280 Notes Page 355 DRAFT August 26, 2022
4.5. 2-Dimensional Systems by Eigenvalues/Eigenvectors 356

4.5.2 complex eigenvalues

Even if the eigenvalues λ1 ,λ2 and their eigenvectors v⃗1 , v⃗2 for A are complex, we still
have that {v⃗1 , v⃗2 } is a fundamental solution set to x⃗ = Ax.
⃗ The problem is that these are
′

complex solutions, and what we need to find is the real solutions.

Recall that complex solutions to a quadratic polynomial (with real coefficients) come
in conjugate pairs. That is, we have λ1, λ2 = α ± iβ where α and β are real numbers. For
the same reason, the eigenvectors v⃗1 , v⃗2 must also come in conjugate pairs. So we have

v⃗1 , v⃗2 = a⃗ ± b⃗ i.

where a⃗ and b⃗ are vectors with real entries.

(α+iβ )t (α−iβ)t
⃗ = c1 e
x(t) (a⃗ + ⃗bi) + c2 e (a⃗ − b⃗i ).
By Euler’s identity this can be written as

⃗ = c1 (cos(αt) + i sin(βt))(⃗
x(t) a + ⃗bi) + c2 (cos(αt) − i sin(βt))(⃗
a − b⃗i).

By multiplying out, this equation can be rearranged as

⃗ = c1 e ((cos(βt)a⃗ − sin(βt)b⃗ ) + (cos(βt)b⃗ + sin(βt)⃗

a)i)
αt
x(t)
+c2 e (((cos(βt)⃗ a − sin(βt)b⃗ ) − (cos(βt)b⃗ + sin(βt)a)i).
⃗
αt

Although this looks somewhat complicated, all we have done is to separate x(t) ⃗ into its
real and imaginary parts. This makes it convenient to find real solutions, which is really
what we want to do. This works much the same way as things did in Section 2.3. That is,
choosing the right values of c1 and c2 will give us 2 linearly independent solutions.
We want to choose c1 and c2 to cancel out the imaginary parts of the equation. It isn’t
hard to see that choosing c1 = 12 and c2 = 12 gives

e ((cos(βt)a⃗ − sin(βt)b⃗ ) + (cos(βt)b⃗ + sin(βt)⃗

a)i)
1 αt
⃗
x(t) =
2
+ e (((cos(βt)⃗ a − sin(βt)b⃗ ) − (cos(βt)b⃗ + sin(βt)a)i)
1 αt
⃗
2
= e (cos(βt)⃗
a − sin(βt)b⃗ ).
αt

2
To find a second linear solution we use our knowledge that i = −1 and choose c1 = − 2i ,
c2 = 2i .Then,

= − e ((cos(βt)a⃗ − sin(βt)b⃗) + (cos(βt)b⃗ + sin(βt)a)i)

i αt
⃗
x(t) ⃗
2
+ e (((cos(βt)⃗ a − sin(βt)b⃗ ) − (cos(βt)b⃗ + sin(βt)a)i)
i αt
⃗
2
= e (sin(βt)⃗
a + cos(βt)b⃗ ).
αt

So we have two linearly independent solutions, and the general solution is then

C. Sprouse —Math 280 Notes Page 356 DRAFT August 26, 2022
4.5. 2-Dimensional Systems by Eigenvalues/Eigenvectors 357

⃗
x(t) = c1 e αt (cos(βt)a⃗ − sin(βt)b⃗ )
+ c2 e (sin(βt)a⃗ + cos(βt)b⃗ )
αt

Although this looks somewhat complicated, it is pretty easy to remember once you
learn the pattern and practice a few. It’s even easier to remember if you know the geomet-
ric interpretation, which we save until the next section.

Example 4.5.4.
′
x1 1 −2 x1
=
x2 5 −5 x2

det(A − λI ) = det
1 −2 λ 0 1− λ −2
− = det
5 −5 0 λ 5 −5 − λ
= (1 − λ)(−5 − λ) + 10 = λ + 4λ + 5.
2

This gives
−4 ± 16 − 20 −4 ± −4 −4 ± 2i
λ= = = = −2 ± i.
2 2 2
One nice thing about complex eigenvalues is that one only needs to compute the eigen-
vectors for one of the eigenvalues, λ = α + iβ . So for λ = −2 + i,

1 − (−2 + i) −2 x 0 3−i −2 x 0
−5 − (−2 + i)
= ⇒ = .
5 y 0 5 −3 − i y 0

So
(3 − i)x − 2y = 0,
5x + (−3 − i)y = 0.
We do not solve this directly, as we want to avoid any complex arithmetic. Instead we use
that if
px + qy = 0,
then two obvious solutions are

x =q , y = −p or x = −q , y = p.

(Plug them in to px + qy to see why!) Since we already know that the above system has a
nontrivial solution, the two equations must be multiples of each other, so we only have to
solve one of them. So we’ll choose the first one:

(3 − i)x − 2y = 0 ⇒ x =2 , y = 3 − i.
So

2 2 0
v= = + i.
3−i 3 −1

C. Sprouse —Math 280 Notes Page 357 DRAFT August 26, 2022
4.5. 2-Dimensional Systems by Eigenvalues/Eigenvectors 358

This gives
2 0
a⃗ = , b⃗ = .
3 −1
So the general solution to the system is

2 0 2 0
⃗ = c1 e
−2t −2t
x(t) cos(t) − sin(t) + c2 e sin(t) + cos(t) .
3 −1 3 −1

Note that if one wants to unravel the vector notation, this can be rewritten as

x1(t) =
−2t −2t
2c1e cos(t) + 2c2e sin(t),
x2(t) = 3c1 e cos(t) + c1e sin(t) + 3c2 e sin(t) − c2 e cos(t).
−2t −2t −2t −2t

Or, a nicer way of writing this is

x1 (t) =
−2t −2t
2c1e cos(t) + 2c2e sin(t),
x2 (t) = (3c1 − c2 )e −2t cos(t) + (c1 + 3c2 )e −2t sin(t).

4.5.3 repeated eigenvalues, λ1 = λ2

Finally we consider the case when the matrix A has repeated eigenvalues λ1 = λ2 . In
some sense this case is the most complicated. First of all, two things can happen: Either
there exist two linearly independent eigenvectors v1 and v2 , or there are no two such
eigenvectors. The first case is somewhat trivial. It turns out that even if λ1 and λ2 are the
same, one might still have two linearly independent eigenvectors. Then things are easy.
λ t
This is because even if λ1 = λ2 , as long as v1 and v2 are linearly independent, e 1 v1 and
λ t
e 2 v2 will be linearly independent for all t. Therefore the general solution in this case
⃗
is still x(t) = c1 eλ1 t v1 + c2 e λ2 t v2 . But the problem is that this only happens when A is a
multiple of I , such as
2 0
A= .
0 2
In this case, v⃗1 , v⃗2 can be literally any two linearly independent vectors. So we will not
consider this case any further.
The interesting case is when A does not have two linearly independent eigenvectors.
From now on, since there is just one eigenvalue we call the eigenvalue and eigenvector
⃗ Then one nonzero solution to the system is clearly e v. ⃗ To find the other
λt
simply λ and v.
⃗ but that doesn’t
λt
linearly independent solution, one might guess something like te v,
quite work out:
(te v)
⃗ = e v⃗ + λte v⃗
λt ′ λt λt

But,
⃗ = te Av = λte v.
⃗
λt λt λt
A(te v)
⃗
λt
The above two things are off by e v.
To correct this, we assume that the solution is of the form x⃗ = te v + e w.⃗ Then let’s
λt λt

again compare x⃗ and Ax⃗ to see if we can figure out what w⃗ is.
′

C. Sprouse —Math 280 Notes Page 358 DRAFT August 26, 2022
4.5. 2-Dimensional Systems by Eigenvalues/Eigenvectors 359

(te v + e w)
⃗ = e v⃗ + λte v⃗ + λe w,
⃗
λt λt ′ λt λt λt

and
A(te v⃗ + e w)
⃗ = te Av⃗ + e A w
⃗ = λte v⃗ + e Aw.
⃗
λt λt λt λt λt λt

For the left hand sides of the two above equations to be equal, we need:

e v⃗ + λte v⃗ + λe w
⃗ = λte v⃗ + e Aw.
⃗
λt λt λt λt λt

The λte v⃗ terms cancel to give

λt

e v⃗ + λe w
⃗ = e Aw.
⃗
λt λt λt

λt
And now the function e can also be factored out of this equation which gives

v⃗ + λw⃗ = Aw.
⃗

⃗ we write this as
Since we want to solve for w,

Aw⃗ − λw⃗ = v.
⃗

Or,
(A − λI )w
⃗ = v.
⃗
In summary, the general solution to the system x⃗ = Ax,
⃗ when A has repeated eigenvalues
′

λ1 = λ2 = λ is

⃗ = c1 e v⃗1 + c2 e (t v⃗ + w)
⃗
λt λt
x(t)
where
(A − λI )v⃗ = 0⃗
and
(A − λI )w⃗ = v⃗

w⃗ is called a generalized eigenvector of A.

′
x1 1 2 x1
Example 4.5.5. =
x2 −2 −3 x2

solution:

= (1 − λ)(−3 − λ) + 4 = λ + 2λ + 1.
1 2 λ 0 1−λ 2 2
det − = det
−2 −3 0 λ −2 −3 − λ

So the eigenvalues are λ1 = λ2 = −1. To find the eigenvectors,

1 2 −1 0 x 0 2 2 x 0
− = ⇒ = .
−2 −3 0 −1 y 0 −2 −2 y 0

C. Sprouse —Math 280 Notes Page 359 DRAFT August 26, 2022
4.5. 2-Dimensional Systems by Eigenvalues/Eigenvectors 360

This gives the equations

2x + 2y = 0
,
−2x − 2y = 0
which clearly give x = −y as the solution. So one eigenvector would be

1
v=
−1

and here it would not be possible to find a linearly independent set of eigenvectors,
because all other eigenvectors would be a multiple of that. So we need to find w such that
(A − λI )w = v:

1 2 −1 0 x 1 2 2 x 1
− = ⇒ = .
−2 −3 0 −1 y −1 −2 −2 y −1

This gives the equations

2y + 2x = 1
−2x − 2y = −1
Since one equation is a multiple of the other here, we only need to solve one of them. We
have that y = 12 − x. Choosing x = 1 then gives y = − 12 . (Note that one could choose
anything for x here, even 0.) So,
1
w⃗ = ,
− 12
and then the general solution is

x⃗(t) = c1 e
−t 1 −t 1 −t 1
+ c2 te +e □.
−1 −1 − 12

Something pretty remarkable actually happens here. In solving the equation

(A − λI)w
⃗ = v,
⃗

the matrix A − λI is singular. And for nonzero v⃗ it is very rare for such an equation to
have any solution for w.⃗ Of course, the fact that we could solve for w ⃗ in this case is no
coincidence. It is guaranteed when A has repeated eigenvalues and does not have two
linearly independent eigenvectors. In this case there is a theorem which says that A must
have the generalized eigenvector w.⃗ The full story for an n ×n matrix A is quite elaborate.
It is called “Jordan canonical form”.

C. Sprouse —Math 280 Notes Page 360 DRAFT August 26, 2022
4.5. 2-Dimensional Systems by Eigenvalues/Eigenvectors 361

Now we summarize everything:

For a 2 × 2 matrix A with eigenvalues λ1 ,λ2 , and eigenvectors v⃗1, v⃗2 , the
general solution to x⃗′ = Ax⃗ is:

• Distinct real eigenvalues λ1 ≠ λ2 :

v⃗1 + c2 e v⃗1 .
λ 1t λ2 t
x(t) = c1 e

• Complex eigenvalues λ1 , λ2 = α ± βi , v⃗1 , v2 = a⃗ ± b⃗ i:

⃗ = c1e (cos(βt)a⃗ − sin(βt)b⃗ )

αt
x(t)
+ c2e (sin(βt)⃗
a + cos(βt)b⃗ )
αt

• Repeated real eigenvalues λ1 = λ2 = λ:

⃗ = c1e v⃗1 + c2 e (t v⃗ + w),

⃗
λt λt
x(t)

where
(A − λI )v⃗ = 0
⃗,
(A − λI )w⃗ = v.
⃗

All three forms need to be memorized!

C. Sprouse —Math 280 Notes Page 361 DRAFT August 26, 2022
 4.5. 2-Dimensional Systems by Eigenvalues/Eigenvectors 362

Homework 4.5:
A) Solve the initial value problems by eigenvalues/eigenvectors (distinct real eigenvalues).
1.
6 3 2
X⃗ = ⃗ ⃗
′
X, X(0) =
−1 2 4
solution: Eigenvalues:

⃗ = 6− λ = (6 − λ)(2 − λ) + 3 = λ − 8λ + 15.
3 2
det(A − λI)
−1 2−λ

λ − 8λ + 15 = (λ − 3)(λ − 5) = 0
2
→ λ = 3, 5.
Eigenvectors:
(A − 3I )v⃗ = 0
3 3 x 0
⇒ = ⇒ x = −y.
−1 −1 y 0

(A − 5I)v⃗ = 0
1 3 x 0
⇒ = ⇒ x = −3y.
−1 −3 y 0
−1 −3
We choose the simplest possible eigenvectors:v⃗ = for λ = 3 and v⃗ = for λ = 5.So
1 1

3t −1 5t −3
⃗
X(t) = c1 e + c2 e .
1 1

Solving for the initial values, −c1 − 3c2 = 2, c1 + c2 = 4. This gives c1 = 7, c2 = −3.So

−1 5t −3 −7e 3t + 9e 5t
⃗
X(t) = 7e
3t
− 3e = 3t 5t □.
1 1 7e − 3e

2. 5.
1 2 −2
X⃗ = ⃗ ⃗
1 1 ⃗ 3 ′
X⃗ = ⃗ =
′ X, X(0) .
X, X(0) = . 0 3 3
−3 5 1
6.
3. 2 0 1
X⃗ = ⃗ ⃗
′
X, X(0) = .
3 1 3 3 1 2
X⃗ = ⃗ ⃗
′
X, X(0) = .
4 3 2
7.
0 −1 ⃗ 2
X⃗ = ⃗
′
4. X, X(0) = .
2 −3 1
1 −2 3
X⃗ = ⃗ ⃗
′
X, X(0) = . 8.
−2 4 −1
1 2 2
X⃗ = ⃗ ⃗
′
X, X(0) = .
4 −1 −1

C. Sprouse —Math 280 Notes Page 362 DRAFT August 26, 2022
 4.5. 2-Dimensional Systems by Eigenvalues/Eigenvectors 363

B) Solve the initial value problems by eigenvalues/eigenvectors (repeated eigenvalues).

9.
1 −1 3
X⃗ = ⃗ ⃗
′
X, X(0) = .
4 5 −1
solution: Eigenvalues:

⃗ )= = (1 − λ)(5 − λ) + 4 = λ − 6λ + 9.
1 −λ −1 2
det(A − λI
4 5− λ

λ − 6λ + 9 = (λ − 3)(λ − 3) = 0
2
→ λ = 3,3.
Eigenvectors:

(A − 3I )v⃗ = 0
−2 −1 x 0
⇒ = ⇒ y = −2x.
4 2 y 0

1
We choose v⃗ = . Then
−2

(A − 3I )w⃗ = v⃗
−2 −1 x 1
⇒ = ⇒ y = −2x − 1.
4 2 y −2

0
We can choose x = 0, so w⃗ = . Then
−1

= c1 e v⃗ + c2e (t v⃗ + w)
1 1 0
⃗
X(t)
λt λt
⃗ = c1 e
3t
+ c2 t + ,
−2 −2 −1
or
3t 3t
c1 e + c2 te
⃗
(−2c1 − c2 )e − 2c2 te
X(t) = 3t 3t .

The initial conditions give c1 = 3, −2c1 − c2 = −1, so c2 = 3, c2 = −5, and

3t 3t
3e − 5te
⃗
X(t) = 3t 3t □ .
−e + 10te

10. 14.
4 1 1 3 4 −1
X⃗ = ⃗ ⃗ X⃗ = ⃗ ⃗
′ ′
X, X(0) = . X, X(0) = .
−1 2 2 −1 7 1
11. 15.
3 −1 2 −2 8 3
X⃗ = ⃗ ⃗ X⃗ = ⃗ ⃗
′ ′
X, X(0) = . X, X(0) = .
4 −1 1 −2 6 2
12. 16.
4 −5 ⃗ 1 0 1 ⃗ 1
X⃗ = ⃗ X⃗ = ⃗
′ ′
X, X(0) = . X, X(0) = .
5 −6 −1 −9 6 2
13. 17.
2 −1 1 2 1 2
X⃗ = ⃗ ⃗ X⃗ = ⃗ ⃗
′ ′
X, X(0) = . X, X(0) = .
4 −2 −2 0 2 3

C. Sprouse —Math 280 Notes Page 363 DRAFT August 26, 2022
 4.5. 2-Dimensional Systems by Eigenvalues/Eigenvectors 364

C) Solve the initial value problems by eigenvalues/eigenvectors (complex eigenvalues).

18.
5 −5 2
X⃗ = ⃗ ⃗
′
X, X(0) = .
5 −3 1
solution: Eigenvalues:

⃗ )= = (5 − λ)(−3 − λ) + 25 = λ − 2λ + 10.
5 −λ −5 2
det(A − λI
5 −3 − λ

2 ± 4 − 40 2 ± 6i
λ= = = 1 ± 3i.
2 2
Eigenvectors:We only find an eigenvector for λ = 1 + 3i.

(A − (1 + 3i)I )v⃗ = 0
4 − 3i −5 x 0
⇒ = .
5 −4 − 3i y 0

The first row above gives (4 − 3i)x − 5y = 0, or 5y = (4 − 3i)x, or y = 15 (4 − 3i)x. To avoid

fractions we can choose x = 5. Then y = 4 − 3i.

5 5 0
v⃗ = = + i.
4 − 3i 4 −3

Recall that if v⃗ = a⃗ + b⃗ i is an eigenvector for λ = α + βi, then the general solution is

⃗ c1 e (cos(βt)a⃗ − sin(βt)b⃗ )
αt
X(t) =
+ c2 e (sin(βt)a⃗ + cos(βt)b⃗ ).
αt

So in our case,

5 0 5 0
⃗
X(t) = c1 e
t
cos(3t) − sin(3t) + c2 e
t
sin(3t) + cos(3t)
4 −3 4 −3
t t
5c1 e cos(3t) + 5c2 e sin(3t)
(4c1 − 3c2 )cos(3t) + (3c1 + 4c2 )e sin(3t)
= t .

The initial conditions then give 5c1 = 2 and 4c1 − 3c2 = 1. So c1 = 25 , c2 = 15 . Plugging these
back into the above gives
t t
⃗ 2e cos(3t) + e sin(3t)
X(t) = t t □.
e cos(3t) + 2e sin(3t)

19. 22.
3 −1 1 −2 5 −1
X⃗ = ⃗ ⃗ X⃗ = ⃗ ⃗
′ ′
X, X(0) = . X, X(0) = .
2 1 −2 −1 2 1
20. 23.
1 −2 ⃗ 1 0 5 ⃗ −3
X⃗ = ⃗ X⃗ = ⃗
′ ′
X, X(0) = . X, X(0) = .
5 −1 −1 −1 2 1
21. 24.
1 −1 2 −2 10 −3
X⃗ = ⃗ ⃗ X⃗ = ⃗ ⃗
′ ′
X, X(0) = . X, X(0) = .
5 −3 1 −5 0 1

C. Sprouse —Math 280 Notes Page 364 DRAFT August 26, 2022
 4.5. 2-Dimensional Systems by Eigenvalues/Eigenvectors 365

D) a) Solve by eigenvalues/eigenvectors and write the general solution in vector form.

b) Write the general solution in x(t), y(t) form.

25. 30.
dx
dx
= x +y dt
= x + 2y
dt dy
dy
= 3x − y dt
= −2x + y
dt
31.
26. dx
= x−y
dx dt
= 2x + 8y dy
dt
dy dt
= −2x + 2y
dt
= −x − 2y
32.
dx
27. dt
= x−y
dy
dx
= 5x − y dt
= x−y
dt
dy
dt
= 9x − y 33.
dx
dt
= 2x − y
28. dy
dx = 2x + 4y
dt
= −x − 2y dt
dy 34.
= 2x − 5y dx
dt
dt
= −y
dy
29. dt
= x + 2y
dx
dt
= x + 2y 35.
dy dx
dt
= 2x + y dt
= −5y
dy
dt = 5x + 6y

E) a) Write the initial value problem in vector form.

b) Solve by using eigenvalues/eigenvectors.
c) Write the solution in x(t), y(t) form.

36. 41.
dx
dx
= x + 2y x(0) = 2 dt
= 3x − y x(0) = 2
dt , ,
dy y(0) = 1
dy
= 9x − 3y y(0) = 1
dt
= 4x + 3y dt
42.
37. dx
= x−y
dx dt x(0) = −1
= x − 2y x(0) = 1 ,
dt ,
dy
= 5x + 5y y(0) = 1
dy
= 4x − 3y y(0) = −2 dt
dt
43.
dx
38. dt
= 8x − 9y x(0) = 2
,
dx
= x + 2y x(0) = 1
dy
= x + 2y y(0) = 1
dt dt
,
dy
= −2x − 3y y(0) = 2 44.
dt
dx
dt
= 3x − 5y x(0) = 3
39. dy ,
y(0) = 1
dx = 5x − 3y
dt
= x −y x(0) = 3 dt
, 45.
dy
= 2x − y y(0) = 2 dx
dt
dt
= −x x(0) = 2
,
40.
dy
= 3x − y y(0) = −3
dt
dx
dt
=x−y x(0) = 3 46.
,
dy
= 2x − 2y y(0) = 5 dx
= 6x − 2y x(0) = 1
dt dt ,
dy
= 5x y(0) = −2
dt

C. Sprouse —Math 280 Notes Page 365 DRAFT August 26, 2022
 4.5. 2-Dimensional Systems by Eigenvalues/Eigenvectors 366

F) For the given ODE:

a) Convert the higher order ODE to a first order system in matrix form. (See section 5.3)
b) Solve the system by eigenvalues/eigenvectors, and use this to find the general solution in vector form.
c) Find a fundamental matrix for the system.
47.
′′ ′
y − 6y + 10y = 0.
′ ′
solution: To write as a first order system we let u = y and v = y . Then u = v and
′ ′′ ′ ′
v = y = 6y − 10y = −10y + 6y . The first order system is then
′
u 0 1 u
= .
v −10 6 v

We have
−λ 1 2
det = λ − 6λ + 10.
−10 6 − λ
Note that this is the characteristic equation of the original second order ODE. (This will
always happen!)The roots are complex, λ = 3 ± i. So we need to find an eigenvector v⃗ = a⃗ + b⃗ i
for λ = 3 + i.The matrix equation is

−3 − i 1 x 0
= .
−10 3−i y 0

So, using the first of the two equations to get the eigenvector,

(−3 − i)x + y = 0, 1 1 0
v⃗ =
−10x + (3 − i)y
⇒ = + i.
= 0, 3 +i 3 1

And the general solution is

1 0 1 0
⃗ = c1 e
3t 3t
x(t) cos(t) − sin(t) + c2 e sin(t) + cos(t) .
3 1 3 1

If we set c1 = 1, c2 = 0 we get the first column of a fundamental matrix Φ, and c0 = 0, c1 = 1

gives the second column.So
3t 3t
e cos(t) e sin(t)
Φ(t) = 3t 3t 3t 3t .
3e cos(t) − e sin(t) 3e sin(t) + e cos(t)

Note that this is the same answer we got in the Exercises in 5.3, but in a completely different
way. (Your answer may or may not be exactly the same, dependin g on your choice of
eigenvectors!) □.
′′ ′ ′′ ′
48. y − 11y + 24y = 0 53. y + 2y + 2y = 0
′′
49. y + 9y = 0 ′′ ′
′′ ′
54. y + 2y + y = 0
50. y − 6y + 9y = 0 ′′ ′
′′ ′ 55. y + 3y = 0
51. y + 3y + 2y = 0
′′ ′ ′′ ′
52. y + 2y + y = 0 56. y + 6y + 25y = 0

C. Sprouse —Math 280 Notes Page 366 DRAFT August 26, 2022
Appendix A

Appendices

A.1 Complex Numbers

This is a very brief account of complex numbers. Hopefully you’ve been exposed to these
somewhere before, but even if not, this should include most of what you need to know.
A complex number is a number of the form

a + bi
2
Where a and b are real numbers, and i is some “imaginary” object such that i = −1.
Engineers also tend to use the letter j instead of i, to avoid conflict with i(t) being used
for current.

Example A.1.1. 2 − 3i is a complex number.

Complex numbers are added in the obvious way:

(a + bi) + (c + di) = (a + c) + (b + d)i.

Example A.1.2.
(2 − 3i) + (3 + i) = 5 − 2i.

Multiplication is slightly more interesting but still very natural:

(a + bi)(c + di) = ac + adi + bci + bdi = (ac − bd) + (ad + bc)i.

2

Example A.1.3.
(2 − 3i)(3 + i) = 6 + 2i − 9i + 3i = 3 − 7i.
2

Example A.1.4.

(3 − 5i)(3 + 5i) = 9 + 15i − 15i − 25i = 9 + 25 = 34.

2

387
A.1. Complex Numbers 388

The second example is something that comes up pretty often:

(a + bi)(a − bi) = a + b .
2 2

Another issue is dividing complex numbers. We need to write the answer in the form
α +βi. We can do this by the conjugate trick (called rationalizing the denominator in high
school). Then the formula we just found is useful here:

a + bi a + bi c − di (a + bi)(c − di) (ac + bd) + (bc − ad)i

= = = .
c + di c + di c − di c2 + d 2 c2 + d 2
This gives the somewhat complicated formula

a + bi ac + bd bc − ad
= 2 + 2 i.
c + di c +d 2
c + d2
This is why complex arithmetic is more difficult than you would think, and its usually
better to avoid it. There is one way to simplify multiplication and division though, which
is to use the polar form of complex numbers. This comes from Euler’s identity which is:

iθ
e = cos(θ) + i sin(θ).

iθ
Given a complex number a + bi let’s see if we can express it in the form Re , where
R > 0 and θ is an angle between 0 and 2π.

iθ
Re = a + bi ⇒ R cos(θ) + iR sin(θ) = a + bi
This gives Rcos(θ) = a and R sin(θ) = b. Using the definition of sin and cos, this means
that R and θ can be derived from the picture.

Im.

a + bi

R
b

θ Real
a

C. Sprouse —Math 280 Notes Page 388 DRAFT August 26, 2022
A.1. Complex Numbers 389

√
The Pythagorean Theorem easily gives R: R = a2 + b2 . We can also see that
b
tan(θ) = a .

If a > 0 this means that θ = arctan ba . But if a < 0 we cannot say this, because arctan only
returns angles between − π2 and π2 . To fix this, we use that if a < 0 −a > 0. But then the
angle of (a, b) is the angle of (−a, −b) plus π (draw it!), so
−b b
θ = arctan −a + π = arctan a + π.

iθ
Cartesian (a + bi) to polar (Re ):
√
R = a2 + b 2 ,

arctan(b/a) for a>0

θ=
arctan(b/a) + π for a<0

Note that going polar to Cartesian is much easier: a = R cos(θ) and b = R sin(theta).
iθ
So Re = R cos(θ) + R sin(θ)i.
Polar form often makes certain computations easier. In particular, it is much easier
to multiply complex numbers in polar form because of the laws of exponentials. For
instance:

i(θ1 +θ2 )
(R1 e )(R2 e ) = R1 R2 e
iθ 1 iθ 2

which is sometimes simpler to use than (a + bi)(c + di) = (ac − bd) + (ad + bc)i. If that
is not convincing enough, here is how raising a number to a power works in polar form:

(Re ) = R e
iθ n n inθ
.

So, for instance to compute (1 + 2i) , first write in polar form as

6 iφ
5e , where φ =
arctan(2) ≈ 1.107. Then

(1 + 2i) = ( 5e ) = 5 e
8 iφ 6 6 i6 arctan(2)

3
≈ 5 cos(6.642) + i sin(6.642)
≈ 125(.936 + .352i) = 117 + 44i.
Although the above computation is not entirely trivial, it is still much easier than
multiplying out (1 + 2i) = (1+ 2i)(1 + 2i)(1 + 2i)(1+ 2i)(1 + 2i)(1 + 2i)! In the next two
6

subsections we will look at other important uses of polar form. These are again closely
related to how polar form greatly simplifies the work involved in multiplication.

C. Sprouse —Math 280 Notes Page 389 DRAFT August 26, 2022
A.1. Complex Numbers 390

A.1.1 Trigonometric Identities

We first want to use polar form to derive the angle addition formulas. That is, we want to
find identities for cos(s+t) and sin(s+t). These can be derived in trigonometry class using
geometric methods, which are ‘elementary’, but far from easy to remember or reproduce.
The polar form of complex arithmetic makes finding these identities much easier. We
have:
is
e = cos(s) + i sin(s),
it
e = cos(t) + i sin(t).
The key trick is the exponential rule for multiplication:

is it is+it i(s+t)
e e =e =e

So multiplying this out,

i(s+t) is it
e = e e
= cos(s) + i sin(s) cos(t) + i sin(t)
2
= cos(s) cos(t) + i cos(s) sin(t) + i sin(s) cos(t) + i sin(s) sin(t)
= cos(s) cos(t) − sin(s) sin(t) + i cos(s) sin(t) + sin(s) cos(t) .

But Euler’s identity also gives

i(s+t)
e = cos(s + t) + i sin(s + t).

So setting the real and imaginary components equal, we get

Angle addition formulas:

cos(s + t) = cos(s)cos(t) − sin(s) sin(t),

sin(s + t) = cos(s) sin(t) + sin(s) cos(t).

The above derivation should be memorized. The point is that this derivation is easier
to remember than the formulas themselves! If one does this only a few times, then the
formulas become easy to remember. And they are used often enough that it is worth it to
know them, rather than always having to look them up. But even better, by going a little
further we can derive all of the other identities we will need.
Recall that cos(x) is an even function, and sin(x) is an odd function:

cos(−x) = cos(x) , sin(−x) = − sin(x).

We can apply this to the identity for cos(s + t):

C. Sprouse —Math 280 Notes Page 390 DRAFT August 26, 2022
A.1. Complex Numbers 391

cos(s + t) = cos(s) cos(t) − sin(s) sin(t),

cos(s − t) = cos(s) cos(t) + sin(s) sin(t).

Now if we add these formulas, we get

cos(s − t) + cos(s + t) = 2 cos(s)cos(t).

If we subtract the two formulas, we get

cos(s − t) − cos(s + t) = 2 sin(s)sin(t).

We can do the same thing with the identity for sin(s + t):

sin(s + t) = cos(s) sin(t) + sin(s)cos(t),

sin(s − t) = − cos(s) sin(t) + sin(s) sin(t).

Adding:
sin(s − t) + sin(s + t) = 2 sin(s)cos(t).

Subtracting:
− sin(s − t) + sin(s + t) = 2 cos(s)sin(t).

This then gives the very useful formulas:

Product to Sum formulas:

1 1
cos(s)cos(t) = cos(s − t) + cos(s + t),
2 2
1 1
sin(s)sin(t) = cos(s − t) − cos(s + t),
2 2
1 1
sin(s)cos(t) = sin(s − t) + sin(s + t),
2 2
1 1
cos(s)sin(t) = − sin(s − t) + sin(s + t).
2 2

These are the identities that you need to do integrals such as ∫ cos(2x)sin(3x) dx.
Again, this comes up often enough that it is nice to be able to derive them without having
to look them up. You have almost certainly already memorized these when s = t in Calcu-
lus, because these are used for the more common integrals ∫ cos (x)dx and ∫ sin (x) dx:
2 2

C. Sprouse —Math 280 Notes Page 391 DRAFT August 26, 2022
A.1. Complex Numbers 392

Half-Angle formulas:

cos (x) =
2 1
1 + cos(2x) ,
2

sin (x) = 1 − cos(2x) ,

2 1
2
1
sin(x)cos(x) = sin(2x).
2

There is one more set of formulas that we will need. Sometimes we need to go from
a sum of sines and cosines to a product, rather than from a product to a sum. Taking the
product-to-sum formulas and setting a = s − t, b = s + t, we have:
1 1
cos(s)cos(t) = cos(a) + cos(b),
2 2
1 1
sin(s)sin(t) = cos(a) − cos(b),
2 2
1 1
sin(s) cos(t) = sin(a) + sin(b),
2 2
1 1
cos(s)sin(t) = − sin(a) + sin(b).
2 2
But then we can also solve for s and t:

s−t = a
⇒ 2s = a + b, 2t = b − a.
s+t = b
a+b
This we get by just adding and subtracting the equations. So now s = 2
,t = − a−b
2
.
Plugging this into the above, and adjusting for +/- signs gives:

Sum to Product formulas:

a+b a−b
cos(a) + cos(b) = 2 cos cos ,
2 2
a+ b a−b
cos(a) − cos(b) = −2 sin sin ,
2 2
a+b a−b
sin(a) + sin(b) = 2 sin cos ,
2 2
a+b a−b
sin(a) − sin(b) = 2 cos sin .
2 2

These will be used to explain some very nice pictures when we discuss resonance in
Chapter 3.

C. Sprouse —Math 280 Notes Page 392 DRAFT August 26, 2022
A.1. Complex Numbers 393

A.1.2 Roots of Unity:

2 2
Consider the quadratic equation s − 1 = 0. This is easy to solve as s = 1 and therefore
s = ±1. The ± comes from the fact that we all know well from high school that when you
2
take square roots you should remember the signs. Therefore s − 1 = 0 has two solutions.
3
In Chapter 3 you may encounter a cubic equation such as s − 1 = 0. Since this is a cubic
it must have 3 solutions! You might say that the 3 solutions are all one... but that’s only
true for the cubic polynomial (s − 1) = 0, which multiplies out to s − 3s + 3s − 1 = 0...
3 3 2
3 2 3
not what we have here. If you try to solve s − 1 = 0 as we did for s − 1 = 0 you get s = 1.
3
1 is clearly a solution to this but −1 isn’t this time. (−1 is −1, not 1!) How do we get
more solutions then? Well it turns out that the other solutions are complex! That is why
they are hidden, and why we are discussing them in this section.
To find out what the complex solutions are, we write s as a complex number in polar
form:
iθ
s = Re .
Then we want to solve
iθ 3
Re =1
or using the regular exponentiation rules,
3 3iθ
R e = 1.

And then using Euler’s identity,

3 3iθ 3 3
R e = R cos(3θ) + iR sin(3θ).

We know that there are 3 solutions... we just have to find them. The first one that we
already know is given by setting R = 1 and θ = 0. Because cos(0) = 1 and sin(0) = 0
implies 1 cos(0) + i sin(0) = 1. To get the other solutions, we can try to find other values
of θ that give us cos(3θ) = 0 and sin(3θ) = 0. But this is easy if we just remember that
cos and sin are 2π-periodic! That is, cos(2π) and sin(2π) are also 1 and 0 respectively. So
we just need 3θ = 2π... or θ = 2π3
. Then we have
2π 3 2π 2π
e 3 = cos 3 + i sin 3
3 3
= cos(2π) + i sin(2π)
= 1 + i(0) = 1
i 2π 3
Therefore e 3 is a solution to s = 1, and if we want to write it in a + bi form we have
2π 2π 2π 1 3
e 3 = cos + i sin =− +i .
3 3 2 2

So that is one of the other solutions. To find the third one, we can use the same trick. We
also know that cos(4π) = 1 and sin(4π) = 0! So another solution would be

) + i sin( ) = − − i
4π 4π 4π 1 3
e 3 = cos( .
3 3 2 2

C. Sprouse —Math 280 Notes Page 393 DRAFT August 26, 2022
A.1. Complex Numbers 394

So now we have 3 solutions! To see why there are really only 3 solutions, you might want
to check what happens if you try the trick again with 6π, 8π, 10π, etc... .
So in summary we have found that the set of solutions to s −1 = 0 is {1, − 12 + 23 i, − 12 −
3
n
32i}. Now lets try to find a general formula for the solutions of s − 1 = 0 so we dont
have to do this kind of work over again every single time.

(Re ) = R e
iθ n n niθ n n
= R cos(kθ) + R i sin(nθ).
n
From the previous discussion we know that to find the solutions to s = 1 we can set R = 1
and nθ to multiples of 2π. That is, nθ = 0, nθ = 2π, nθ = 4π,..., nθ = 2(n − 1)π. We only
do it up to n − 1 because this gives us exactly k solutions. So the values for θ that we get
are 0, (2π)/n, (4π)/n, ..., ((n − 1)π)/n. That is
The solutions to sn − 1 = 0
are
i 2kπ
e n where k = 0, 1, ..., n − 1.
Or in other words,
2(n−1)π
i 2π i 4π i
0, e n ,e n , ..., e n .
If one needs the solutions in rectangular form (as we do in chapter 3), then it is easy to
convert from polar form by using Euler’s identity. For instance you can check that the
6
solutions to s − 1 = 0 are
1 3 1 3 1 3 1 3
1, + i, − + i, −1, − − i, + i.
2 2 2 2 2 2 2 2
Now it helps to take this a little further. Suppose that we want to find all solutions
to z n = c, where c is any real, positive constant. As before, there should be exactly n
solutions. But finding them is easy from what we have already done! Because if we now
1
( 2kπ i) n
let c n denote the real n-th root of c, and e n solves our previous equation s = 1, we
have n 1 2kπ
1 i2kπ
cne n =c e = c.
1
( 2kπ i)
So all solutions are of the form c n e n .
This is exactly the thing you are supposed to remember in high school for square roots!
2 n
That is, in exactly the same way that the solutions to z = 4 are ±2, the solutions to z = c
are everything that looks like
i 2πk 1
z=e n cn.
3
Example A.1.5. The set of solutions to z − 8 = 0 is

2, 2 + 3i, 2 − 3i.
n
We can do the same thing if we have z = −c... but there is a little more do do. Recall
2 2
that if we solve z = −c the solutions are ±i c. Here i and −i are the solutions to s = −1.
n
Similarly, if we want to find the solutions to z = −c we should know all of the solutions
n
to s = −1.
We’ll just say what they are:

C. Sprouse —Math 280 Notes Page 394 DRAFT August 26, 2022
A.1. Complex Numbers 395

n
The solutions to s + 1 = 0
are
(2k+1)π
i
e n where k = 0, 1, ..., n − 1.
Or in other words,
(2n−1)π
iπ i 3π i
e n ,e n ,..., e n .

You can derive this for yourself... either by repeating the whole process that we did for
n
s − 1 = 0, or more interestingly using in some way that we already know the answer for
n
s − 1 = 0.
It is easiest to remember how things work with pictures. The n-th roots of 1 (solutions
n
to s − 1 = 0) are n evenly spaced points on the unit circle starting at s = 1. The n-th roots
of −1 are also evenly spaced points on the unit circle, except these are rotated to start
at angle 2π/n from 1. If n is odd then this means −1 is always a root. But one should
4
keep in mind that this is not true when n is even. For instance the roots of s = −1 are
(±1 ± i)/ 2.

1 i 1 i 1 i

−1 1 −1 1 −1 1

−1 −1 −1
3 3 4
Roots of s = 1. Roots of s = −1. Roots of s = −1.

C. Sprouse —Math 280 Notes Page 395 DRAFT August 26, 2022
A.2. A Survey of Partial Fractions Methods 396

A.2 A Survey of Partial Fractions Methods

To be proficient at solving constant coefficient ODE’s by Laplace transforms, one must
become very adept at partial fractions. Recall that Partial fractions is a way to rewrite a
rational function p(s)/q(s) as a sum of simpler terms. Here p(s) and q(s) are polynomials,
and the degree of p(s) must be less than the degree of q(s) for the technique to work.
(Otherwise, one must do polynomial division first!)
The general pattern of partial fractions can be deduced by just looking at when the
denominator q(s) has degree 3. In this case there are exactly three types, that cover most
of what can happen in general.

• Type I: q(s) contains only linear factors.

2
1 3s + 2s + 1
(s + 1)(s − 2)(s − 3)
, .
s(2s − 1)(s + 2)
• Type II: q(s) contains linear factors to a power. (‘repeated’ linear factors.)

2
1 2s − 1 s
(s + 1)2 (s − 2) s (s + 2) (s − 1)3
, 2
,

• Type III: q(s) has an unfactorable quadratic.

1 s+1 s
(s + 1)(s − 2) (s − 1)(s + 2s + 5)
2
, ,
s(s2 + 4) 2

Type I: q(s) has distinct linear factors.

This is straightforward:
p(s) A B C
(s − a)(s − b)(s − c)
= s −a + + s − c.
s−b
Example A.2.1.
2
s +1 A B C
= s + +
s(s + 3)(s − 2) s +3 s −2

Type II: q(s) has a ‘repeated’ linear factor.

p(s) A B C
(s − a)2 (s − b) (s − a) (s − a) s − b
= 2
+ + .

Example A.2.2.
s A B C
s2 (s − 2)
= + s+ .
s 2 s −2
Example A.2.3.
2
s −1 A B C
(s + 2)(s + 3) s + 2 (s + 3)
= + + .
2 2 s+3

C. Sprouse —Math 280 Notes Page 396 DRAFT August 26, 2022
A.2. A Survey of Partial Fractions Methods 397

Linear factors to a higher power work similarly.

p(s) A B C
(s − a)3 (s − a) (s − a)
= 3
+ 2
+ s − a.

Example A.2.4.
2
2s − s + 1 A B C
=
(s − 5) (s − 5) (s − 5)
+ + .
3 3 2 s − 5

Type III: q(s) has an unfactorable quadratic.

Before presenting the general formula, the following type of examples should be familiar
from calculus.

Example A.2.5.
s As + B C
=
(s2 + 1)(s − 2)
+ .
s +1 s−2
2

Example A.2.6.
2s + 1 A Bs + C
2
= s + 2 .
s(s + 25) s + 25
2
Now an issue arises for the general quadratic factors s + bs + c. In calculus you may
have dealt with these using the obvious generalization of the above:

3s − 1 As + B C
(s2 + 2s + 5)(s − 3)
= +
s2 + 2s + 5 s−3

This works, and is suitable for the integration techniques covered in calculus. However,
when doing Laplace transforms a somewhat different form is much more convenient:

p(s) A s + 2b + B C
=
(s2 + bs + c)(s − d)
+ .
s 2 + bs + c s −d
b
Where did the 2
come from? It’s what you get when completing the square:

2 2
2 b b
s + bs + c = s + +c− .
2 2
αt
The reason for doing this should be clear, if you recall the Laplace transforms of e cos(βt)
αt
and e sin(βt), which is what one would be trying to recover here. So now the partial
fractions above will look like

Example A.2.7.
3s − 1 A(s + 1) + B C
(s + 2s + 5)(s − 3)
= 2 + .
2
s + 2s + 5 s−3

Similarly,

C. Sprouse —Math 280 Notes Page 397 DRAFT August 26, 2022
A.2. A Survey of Partial Fractions Methods 398

Example A.2.8.
2
s +1 A B(s − 3) + C
2
= s + 2 .
s(s − 6s + 10) s − 6s + 10
Example A.2.9.
2
s +1 A B(s + 5) + C
=
(s + 1)(s + 10s + 34)
+ 2 .
2 s + 1 s + 10s + 34
We now discuss the common methods of finding the coefficients A,B,C. You surely
saw the first method in Calculus, and perhaps also the second. But for Laplace transforms
you will need to be fully versatile with at least both of these methods, because often one
way is quite easy while the other is too cumbersome to deal with by hand.

A.2.1 Solving for the coefficients

Method 1: Plugging points.
This method is best for Type I, when each coefficient can be obtained from plugging a
single point.
Example A.2.10.
5s − 1 A B C
(s + 1)(s − 1)(s − 2)
= + + .
s + 1 s − 1 s − 2

solution: Multiplying through by the denominator,

5s − 1 = A(s − 1)(s − 2) + B(s + 1)(s − 2) + C(s + 1)(s − 1).

We plug in the points that can make the terms on the right zero.

s = −1 ⇒ −6 = A(−2)(−3) + B(0)(−3) + C(0)(−2) ⇒ A = −1.

s=1 ⇒ 4 = A(0)(−1) + B(2)(−1) + C(2)(0) ⇒ B = −2.

s=2 ⇒ 9 = A(1)(0) + B(3)(0) + C(3)(1) ⇒ C = 3.

In the Type II and III cases, there are less points that can zero out terms. We still have
to plug in as many points as there are coefficients (in this case n = 3), to have as many
equations as unknowns. But now getting all of the coefficients takes more work.
Example A.2.11.
s−1 A B C
(s − 2)(s − 3) s − 2 (s − 3)
= + + .
2 2 s−3
solution:
2
s − 1 = A(s − 3) + B(s − 2) + C(s − 2)(s − 3).
Plugging points,
2
s=2 ⇒ 1 = A(−1) + B(0) + C(0)(−1) ⇒ A = 1.
2
s=3 ⇒ 2 = A(0) + B(1) + C(1)(0) ⇒ B = 2.

C. Sprouse —Math 280 Notes Page 398 DRAFT August 26, 2022
A.2. A Survey of Partial Fractions Methods 399

To get C we need a third point. We can choose any point that we haven’t used. Usually it’s
best to choose s to be the smallest number available.
2
s=0 ⇒ −1 = A(−3) + B(−2) + C(−2)(−3) = 9A − 2B + 6C
C = (−1 − 9A + 2B) = (−1 − 9 + 4) = −1.
1 1
⇒
6 6
So
s−1 1 2 1
=
(s − 2)(s − 3) (s − 3)
+ − .
2 s − 2 2 s − 3

The most difficult case using this method is Type III. In this case there is only one
point that gives a coefficient directly.
Example A.2.12.
3s − 1 A Bs + C
(s − 2)(s + 1) s − 2 s + 1
2
= + 2 .

solution:
3s − 1 = A(s + 1) + (Bs + C)(s − 2).
2

There is now only one point to plug that is obviously good.

s=2 ⇒ 5 = A(5) + (B(2) + C)(0) ⇒ A = 1.

The next best point is s = 0, because it eliminates the B term.

s=0 ⇒ −1 = A(1) + (B(0) + C)(−2) = A − 2C

C = (A + 1) = (1 + 1) = 1.
1 1
⇒
2 2
The final point can be anything else, so again we just choose a small number.

s=1 ⇒ 2 = A(2) + (B(1) + C)(−1) = 2A − B − C

⇒ B = −2 + 2A − C = −2 + 2 − 1 = −1.

The second-best point is not necessarily always zero. In particular if the quadratic term
2
s + bs + c has nonzero b:
Example A.2.13.
9s − 8 A B(s − 3) + C
=
(s + 2)(s − 6s + 10)
+ 2 .
2 s + 2 s − 6s + 10
solution:
9s − 8 = A(s − 6s + 10) + (B(s − 3) + C)(s + 2).
2

The first point is still obvious.

s = −2 ⇒ −26 = A(26) + (B(−5) + C)(0) ⇒ A = −1.

The next best term is whatever eliminates B so then we can then get C.

s=3 ⇒ 19 = A(1) + (B(0) + C)(5) = A + 5C

C = (19 − A) = (19 + 1) = 4.
1 1
⇒
5 5

C. Sprouse —Math 280 Notes Page 399 DRAFT August 26, 2022
A.2. A Survey of Partial Fractions Methods 400

For the final point we can choose s = 0, since it hasn’t already been used.

s=0 ⇒ −8 = A(10) + (B(−3) + C)(2) = 10A − 6B + 2C

B = (8 + 10A + 2C) = (8 − 10 + 8) = 1.
1 1
⇒
6 6
So
9s − 8 −1 1(s − 3) + 4
=
(s + 2)(s2 − 6s + 10) s + 2 s2 − 6s + 10
+ .

Method 2: Rearranging.
This method complements the first, in that it is often much easier when Method 1 is hard.
And conversely for problems that are easy with Method 1, this method should be avoided
as it would be far more complicated. In particular, Type I partial fractions should hardly
ever be done using the method here. So we focus on the Type II and III expansions.
The method is to multiply everything out and rearrange the terms into a polynomial
of s. Then equating coefficients gives a system of equations we can solve.
Example A.2.14.
2
2s − 7s + 9 A B C
(s − 3)
= s + + .
s(s − 3)2 2 s − 3
solution:
2 2
2s − 7s + 9 = A(s − 3) + Bs + Cs(s − 3).
Multiplying out and rearranging,
2 2 2
2s − 7s + 9 = A(s − 6s + 9) + Bs + C(s − 3s)
= (A + C)s + (−6A + B − 3C)s + 9A.
2

Equating coefficients gives the system of equations

A+C = 2,
−6A + B − 3C = −7,
9A = 9.

Fortunately the third equation gives us A = 1 immediately, so then the first equation gives
C = 2 − A = 1, and then the second gives B = −7 + 6A + 3C = −7 + 6 + 3 = 2. So
2
2s − 7s + 9 1 2 1
(s − 3)
= s+ + .
s(s − 3) 2 2 s−3

We can see that this method works best when the denominator contains factors of s
or s , rather than only things like s − 2 and (s + 1) . This makes the rearranging a little
2 2

simpler, but more importantly it tends to give a system of equations that is more easily
solved.
Example A.2.15.
s−2 A B C
s2 (2s − 1)
= + s+ .
s 2 2s − 1

C. Sprouse —Math 280 Notes Page 400 DRAFT August 26, 2022
A.2. A Survey of Partial Fractions Methods 401

solution:
2
s − 2 = A(2s − 1) + Bs(2s − 1) + Cs
2 2
= 2As − A + 2Bs − Bs + Cs
= (2B + C)s + (2A − B)s − A.
2

Note that the left hand side is s − 2 = (0)s + (1)s + (−2)1. This gives the system of
2

equations
2B + C = 0,
2A − B = 1,
−A = −2.
So A = 2, and then from the second equation B = 2A − 1 = 3, and from the first equation,
C = −2B = −6.6 And so
s −2 2 3 6
=
s 2 (2s − 1) s2 s 2s − 1
+ − .

Example A.2.16.
2
2s + 7 A Bs + C
=
(s − 2)(s + 1) s − 2 s + 1
2
+ 2 .

solution: This one has a factor of s − 2 instead of s. So this isn’t quite the type you’d
definitely want to do Method 2 on. But plugging in points will be about the same amount
of work, so for this one you could use whichever method you prefer. Let’s try rearranging.

A(s + 1) + (Bs + C)(s − 2)

2 2
2s + 7 =
2 2
= As + A + Bs − 2Bs + Cs − 2C
(A + B)s + (−2B + C)s + (A − 2C).
2
=

So the system is
A+B = 2,
−2B + C = 0,.
A − 2C = 7.
The second equation gives C = 2B, so the third equation becomes A − 4B = 7. Subtracting
the first gives −5B = 5, or B = −1. So then the second equation gives C = −2, and the first
gives A = 2 − B = 2 + 1 = 3. So
2
2s + 7 3 −s − 2
=
(s − 2)(s2 + 1) s − 2 s2 + 1
+ .

Example A.2.17.
7s + 10 A B(s + 1) + C
= + 2 .
s(s2 + 2s + 5) s s + 2s + 5
solution:

A(s + 2s + 5) + (B(s + 1) + C)(s)

2
7s + 10 =
2 2
= As + 2As + 5A + Bs + Bs + Cs
(A + B)s + (2A + B + C)s + 5A.
2
=

C. Sprouse —Math 280 Notes Page 401 DRAFT August 26, 2022
A.2. A Survey of Partial Fractions Methods 402

The system is
A+B = 0,
2A + B + C = 7, .
5A = 10.
The third equation gives A = 2, so the first gives B = −2. Then by the second,
C = 7 − 2A − B = 7 − 4 + 2 = 5. So
7s + 10 2 −2(s + 1) + 5
= s+ 2 .
s(s 2 + 2s + 5) s + 2s + 5

Method 3: Heaviside Coverup

Heaviside Coverup is primarily used in engineering, where it is common to do Laplace
transforms by hand. Its use is often restricted to Type I expansions, where it is equivalent
to plugging in points, but faster. Consider the general Type I form:

p(s) A B C
(s − a)(s − b)(s − c)
= s −a + + s − c.
s−b

By Method 1, we can solve for A, B,C by plugging points into

p(s) = A(s − b)(s − c) + B(s − a)(s − c) + C(s − a)(s − b).

The solutions are easily seen to be

p(a) p(s)
(a − b)(a − c) (s − b)(s − c)
A= = .
s=a

p(b) p(s)
(b − a)(b − c) (s − a)(s − c)
B= = .
s=b

p(c) p(s)
(c − a)(c − b) (s − a)(s − b)
C= = .
s=c
The method of Heaviside coverup is simply to use the formulas on the right-hand side
for A,B,C , without going through the process of plugging in points, because they are of a
form that is easy to remember. For instance to get A, you just plug a into the original form
p(s)/((s −a)(s −b)(s −c)), but also ‘cover up’ the s −a term for which plugging in a would
give a division by 0. This is the formula for A above. You can check that the formulas for
B and C are obtained similarly. This takes a little bit of practice to get used to, but once
learned it becomes a fast shorthand for plugging in points. It reduces the time involved
in any Type I expansion, so can always be recommended for these.

Example A.2.18.
s+9 A B C
= s + +
s(s + 1)(s − 3) s + 1 s − 3

solution:
s+9 9
(s + 1)(s − 3) (1)(−3)
A= = = −3.
s=0

C. Sprouse —Math 280 Notes Page 402 DRAFT August 26, 2022
A.2. A Survey of Partial Fractions Methods 403

s+9 8
(−1)(−4)
B= = = 2.
s(s − 3) s=−1
s+9 12
(3)(4)
C= = = 1.
s(s + 1) s=3
So
s+9 −3 2 1
= s + + .
s(s + 1)(s − 3) s + 1 s − 2

Heaviside coverup also works for Type II expansions, but is more involved. For

p(s) A B C
(s − a) (s − b)
=
(s − a)(s − b)2
+ + ,
2 s − b

we can get A exactly as before. But for the other two terms we need to cover up the (s−b)
2

in the denominator, if we are to evaluate without dividing by zero. Correspondingly, we

can only get the coefficient B in this way.

p(s)
A=
(s − b)2
.
s=a

p(s)
B = s−a .
s=b
That is, we do not have a formula for C, until we work a little more. Suppose we take the
Taylor series of p(s)/(s − a) around s = b:

s − a = c0 + c1(s − b) + c2 (s − b) + c3 (s − b) ...
p(s) 2 3

Then the function we are trying to decompose is

+ c2 + c3 (s − b) + ...
p(s) c0 c1
(s − a)(s − b) (s − b)
= +
2 2 s−b

Applying the partial fraction expansion,

+ c2 + c3 (s − b) + ...
A B C c0 c1
s − a (s − b) =
s − b (s − b)
+ + +
2 2 s−b

If we multiply both sides of this by (s − b) and then set s = b, all terms will go to zero
2

except the constant terms B and c0 . This gives B = c0 . So canceling these terms from the
above equation,

s − a + s − b = s − b + c2 + c3 (s − b) + ...
A C c1

Now we can multiply this by s − b, and set s = b again. Then all terms are zero except the
constant terms C and c1 . This gives C = c0 . So B and C are the coefficients c0 , c1 in the
Taylor expansion of p(s)/(s − a). But from calculus we have a formula for these (the first
of which we already know!):

C. Sprouse —Math 280 Notes Page 403 DRAFT August 26, 2022
A.2. A Survey of Partial Fractions Methods 404

p(s)
B= s−a .
s=b

d p(s)
C= .
ds s − a s=b

Example A.2.19.
5s − 1 A B C
=
(s − 2)(s + 1) s − 2 (s + 1)
+ + .
2 2 s+1

solution:
5s − 1 9
A= = = 1.
(s + 1)2 s=2 (3)2
5s − 1 (−6)
(−3)
B= = = 2.
s−2 s=−1

d 5s − 1 −9 −9
(s − 2)2
C= = = = −1.
ds s − 2 s=−1 s=−1
9
So,
5s − 1 1 2 1
=
(s − 2)(s + 1)2 s − 2 (s + 1)2 s + 1
+ − .

A similar formula works for higher degree expansions as well. In general if the de-
nominator has a (s − a) term, we can write the expansion as
n

p(s) A0 A1 An−1 r(s)

(s − a) q(s) (s − a)
=
(s − a)
n n + k−1
+ ... + s − a + ,
q(s)

and then formula for Ak is

k
1 d p(s)
Ak = .
k! ds k q(s) s=a
Note that in practice, unless p(s) and q(s) are very simple it is prohibitively difficult
to take more than one or two derivatives of p(s)/q(s) by hand. But things might be
reasonable if for instance p(s) is constant.

Example A.2.20.
1 A0 A1 B0 B1 B2
=
s (s − 1) (s − 1) (s − 1)
+ s + + +
2 3 2
s 3 2 s−1

solution:
A0 = (s − 1)
−3 −4
s=0
= −1, A1 = −3(s − 1) s=0
= −3
−2 −3 −4
B0 = s s=1
= 1, B1 = −2s s=1
= −2, B2 = 6s s=1
=6
So,
1 1 3 1 2 6
=−
s2(s − 1)3 (s − 1) (s − 1)
−s+ − + .
s 2 3 2 s − 1

C. Sprouse —Math 280 Notes Page 404 DRAFT August 26, 2022
A.2. A Survey of Partial Fractions Methods 405

Finally, Type III expansions can theoretically also be done by Heaviside coverup in the
same way that Type I expansions can, by using complex arithmetic. That is, the quadratic
2
term can in fact be factored if we accept complex roots. In fact, s + bs + c has roots − 2b ±
√
4c−b2
by the quadratic formula. So for instance s + 9 can be factored as (s + 3i)(s − 3i),
2
2
i,
or s − 6s + 34 can be factored as (s −(3 +5i))(s −(3 −5i)). Although it is easy to write any
2

Type III expansion like a Type I expansion in this way, the complex arithmetic involved in
computing the coefficients quickly gets out of hand, in all but the very simplest of cases.
So we don’t encourage actually doing it.

Higher degree expansions

In the above we focused on expansion where the denominator had degree three. Higher
degree expansions are of course similar, and you should be familiar with the pattern from
calculus.

p(s) A B C D
= s−a +
(s − a)(s − b) (s − b) (s − b)
+ +
3 3 2 s − b
p(s) A B C D
=
(s − a) (s − b) (s − a) (s − b)
+ s−a + +
2 2 2 2 s − b
p(s) A B Cs + D
(s − a)2 (s 2 + bs + c) (s − a)
= 2
+ s−a + 2
s + bs + c
p(s) As + B Cs + D
=
(s + bs + c)(s + ds + e)
2 2 2
+ 2
s + as + b s + ds + e
p(s) As + B Cs + D
=
(s + bs + c) (s + bs + c)
2 2 2 2
+ 2 .
s + bs + c

Taylor Series
Sometimes we have a single term of the form
p(s)
(s − a)n
.

Above we used a Taylor series when deriving the Heaviside coverup formula for these.
When there is a single term in this form, we can easily get the partial fraction expansion
just by computing the Taylor series of p(s) at s = a. Since p(s) is a polynomial of degree
< n, this results in a finite series:

p(s) = c0 + c1 (s − a) + c2 (s − a) + ... + cn−1 (s − a)

2 n−1
.
Dividing this by (s − a) we then immediately get the partial fraction expansion!
n

p(s) c0 c1 cn−1
(s − a) (s − a)
=
(s − a)
n n + n−1
+ ... + s − a .

Note that since p(s) is a polynomial, computing the Taylor expansion of p(s) by hand is
in general quite easy to do.

C. Sprouse —Math 280 Notes Page 405 DRAFT August 26, 2022
A.2. A Survey of Partial Fractions Methods 406

Example A.2.21.
3
s
(s + 1)4
3
solution: We have p(s) = s . We compute the Taylor series at a = −1. p(−1) = −1,
(n)
p (−1) = 3, p (−1) = −6, p (−1) = 6, so using cn = f (−1)/n!,
′ ′′ ′′′

s = −1 + 3(s + 1) − 3(s + 1) + (s + 1) .
3 2 3

This gives
3
s −1 3 3 1
=
(s + 1) (s + 1) (s + 1) (s + 1)
+ − + .
4 4 3 2 s + 1

C. Sprouse —Math 280 Notes Page 406 DRAFT August 26, 2022
Index

Abel’s Theorem, 123, 343 generalized eigenvector, 359

amplitude, 198 guess for particular solution, 166
angle addition formulas, 390
autonomous, 51, 336 half-angle formulas, 392
homogeneous, 338
Bernoulli equation, 80 homogeneous (first order), 75
homogeneous (Linear ODE), 118
capacitor, 100 homogeneous polynomial, 78
characteristic equation, 146, 153 horizontal shift, 200
constant coefficient, 145, 337
constant-coefficient system, 313 implicit solution, 18
convolution, 297 Improper node, 377
critically damped, 226 impulse function, 289
impulse response, 290, 299, 310
damping, 209 initial condition, 11
dimension, 313 initial value, 11
direction field, 49 Initial values, 114
dropsonde, 112 integrating factor, 37
eigenvalue, 351 IVP, 11
eigenvector, 351 Lambert W-function, 31
elimination, 313 Laplace transform, 241
elliptic integrals, 44 linear, 9, 36, 337
envelope, 201, 235 Linear Time Invariant, 145
Euler’s identity, 388 linearly dependent, 129
explicit solution, 18 linearly independent, 129
filter, 304 low pass filter, 105
first order system, 336 LTI, 313, 337
forced oscillation, 227 matrix exponential, 344
forced solution, 299 Mellin Inverse Formula, 257
forcing function, 226
friction, 209 n-dimensional, 313
Fund Thm of LIVP’s, 114 nice, 166, 245
Fund. Thm. of LODE’s, 121 Nodal sink, 370
fundamental matrix, 340 nodal source, 368
Fundamental solution set, 115 Node, 368
Noncrossing Principle, 69
General Solution, 118
general solution, 8, 160, 161 ODE, 5

407
Index 408

order, 6 unit step function, 271

overdamped, 226 unstable node, 368
Unstable spiral, 374
particular solution, 160, 161
PDE, 6 Vandermonde determinant, 153
phase, 198
phase shift, 200 Wronskian, 116
Picard Lindelhöf Theorem, 61
polar form, 199, 388
principal fundamental matrix, 342
product rule, 137, 169
product to sum formulas, 391

quarter-car model, 213

reduction of order, 138

relaxation oscillation, 207
resistor, 99
resonance, 231
roots of unity, 155, 393
Routh Stability Criterion, 154

separable, 18
sign chart, 51
slope field, 49
snubber, 216
solution, 7
Source, 368
Spiral, 374
Spiral sink, 375
Spiral source, 374
spring constant, 208
Stable spiral, 375
standard form, 10, 11, 36, 185
steady-state, 104, 202, 211
stiffness, 208
sum to angle formula, 235
sum to product formulas, 392
Superposition, 174

terminal velocity, 106

time invariant, 313, 337
transfer function, 299
transient, 202, 211
transition matrix, 342
trigonometric identities, 390

underdamped, 226

C. Sprouse —Math 280 Notes Page 408 DRAFT August 26, 2022