A Second Course in Elementary Differential Equations PDF

A Second Course in Elementary Ordinary
Differential Equations
Marcel B. Finan
Arkansas Tech University
c
All
Rights Reserved
Contents
26 Calculus of Matrix-Valued Functions of a Real Variable
27 nth Order Linear Differential Equations:Existence and Uniqueness

18
28 The General Solution of nth Order Linear Homogeneous Equations
25
29 Fundamental Sets and Linear Independence
34
30 Higher Order Homogeneous Linear Equations with Constant

Coefficients
40
31 Non Homogeneous nth Order Linear Differential Equations 47
32 Existence and Uniqueness of Solution to Initial Value First
Order Linear Systems
56
33 Homogeneous First Order Linear Systems
63
34 First Order Linear Systems: Fundamental Sets and Linear

Independence
74
35 Homogeneous Systems with Constant Coefficients
83
36 Homogeneous Systems with Constant Coefficients: Complex

Eigenvalues
93
37 Homogeneous Systems with Constant Coefficients: Repeated
Eigenvalues
97
38 Nonhomogeneous First Order Linear Systems
108
39 Solving First Order Linear Systems with Diagonalizable Constant Coefficients Matrix
118
40 Solving First Order Linear Systems Using Exponential Matrix
126
41 The Laplace Transform: Basic Definitions and Results
131
42 Further Studies of Laplace Transform
142
43 The Laplace Transform and the Method of Partial Fractions155

44 Laplace Transforms of Periodic Functions
162
45 Solving Systems of Differential Equations Using Laplace Transform

171
46 Convolution Integrals
178
47 The Dirac Delta Function and Impulse Response
186
48 Numerical Solutions to ODEs: Eulers Method and its Variants

194
26
Calculus of Matrix-Valued Functions of a

Real Variable
In establishing the existence result for second and higher order linear differential equations one transforms the equation into a linear system and tries
to solve such a system. This procedure requires the use of concepts such
as the derivative of a matrix whose entries are functions of t, the integral
of a matrix, and the exponential matrix function. Thus, techniques from
matrix theory play an important role in dealing with systems of differential
equations. The present section introduces the necessary background in the
calculus of matrix functions.
Matrix-Valued Functions of a Real Variable
A matrix A of dimension m n is a rectangular array of the form
a11 a12 ... a1n

a21 a22 ... a2n
A=
...
... ... ...
am1 am2 ... amn
where the aij s are the entries of the matrix, m is the number of rows, n
is the number of columns. The zero matrix 0 is the matrix whose entries
are all 0. The n n identity matrix In is a square matrix whose main
diagonal consists of 10 s and the off diagonal entries are all 0. A matrix A can
be represented with the following compact notation A = (aij ). The entry aij
is located in the ith row and jth column.
Example 26.1
Consider the matrix
5 0
1
A(t) = 10 2 0
5 2 7
Find a22 , a32 , and a23 .

Solution.
The entry a22 is in the second row and second column so that a22 = 2.
Similarly, a32 = 2 and a23 = 0
An m n array whose entries are functions of a real variable defined on

a common interval is called a matrix function. Thus, the matrix
a11 (t) a12 (t) a13 (t)

A(t) = a21 (t) a22 (t) a23 (t)
a31 (t) a32 (t) a33 (t)
is a 3 3 matrix function whereas the matrix
x1 (t)
x(t) = x2 (t)
x3 (t)
is a 3 1 matrix function also known as a vector-valued function.
We will denote an m n matrix function by A(t) = (aij (t)) where aij (t) is
the entry in the ith row and jth coloumn.
Arithmetic of Matrix Functions
All the familiar rules of matrix arithmetic hold for matrix functions as well.
(i) Equality: Two m n matrices A(t) = (aij (t)) and B(t) = (bij (t)) are
said to be equal if and only if aij (t) = bij (t) for all 1 i m and 1 j n.
That is, two matrices are equal if and only if all corresponding elements are
equal. Notice that the matrices must be of the same dimension.
Example 26.2
Solve the following matrix equation for a, b, c, and d

ab
b+c
8 1
=
3d + c 2a 4d
7 6
Solution.
Equating corresponding entries we get the system
a b
=
b + c
=
c
+
3d
=
2a
4d =
8
1
7
6
Adding the first two equations to obtain a + c = 9. Adding 4 times the third
equation to 3 times the last equation to obtain 6a + 4c = 46 or 3a + 2c = 23.
5
Solving the two equations in a and c one finds a = 5 and c = 4. Hence,

b = 3 and d = 1
(ii) Addition: If A(t) = (aij (t)) and B(t) = (bij (t) are m n matrices
then the sum is a new m n matrix obtained by adding corresponding elements
(A + B)(t) = A(t) + B(t) = (aij (t) + bij (t))
Matrices of different dimensions cannot be added.
(iii) Subtraction: Let A(t) = (aij (t)) and B(t) = (bij (t)) be two m n
matrices. Then the difference (A B)(t) is the new matrix obtained by
subtracting corresponding elements,that is
(A B)(t) = A(t) B(t) = (aij (t) bij (t))
(iv) Scalar Multiplication: If is a real number and A(t) = (aij (t)) is an
m n matrix then (A)(t) is the m n matrix obtained by multiplying the
entries of A by the number ; that is,
(A)(t) = A(t) = (aij (t))
(v) Matrix Multiplication: If A(t) is an mn matrix and B(t) is an np
matrix then the matrix AB(t) is the m p matrix
AB(t) = (cij (t))
where
cij (t) =
n
X
aik (t)bkj (t)
k=1
That is the cij entry is obtained by multiplying componentwise the ith row
of A(t) by the jth column of B(t). It is important to realize that the order
of the multiplicands is significant, in other words AB(t) is not necessarily
equal to BA(t). In mathematical terminology matrix multiplication is not
commutative.
Example 26.3

A=
1 2
3 2

,B =
2 1
3 4
Show that AB 6= BA. Hence, matrix multiplication is not commutative.

6
Solution.
Using the definition of matrix multiplication we find

1 2
4 7
AB =
, BA =
0 5
9 2
Hence, AB 6= BA
(vi) Inverse: An n n matrix A(t) is said to be invertible if and only if
there is an n n matrix B(t) such that AB(t) = BA(t) = I where I is the
matrix whose main diagonal consists of the number 1 and 0 elsewhere. We
denote the inverse of A(t) by A1 (t).
Example 26.4
Find the inverse of the matrix

a b
c d
A=
given that ad bc 6= 0. The quantity ad bc is called the determinant of

A and is denoted by detA
Solution.
Suppose that
1
A
Then
x y
z t

=
a b
c d
x y
z t

=
1 0
0 1
This implies that

ax + cy bx + dy
az + ct bz + dt
1 0
0 1
Hence,
ax + cy
bx + dy
az + ct
bz + dt
=
=
=
=
1
0
0
1
Applying the method of elimination to the first two equations we find

7
d
adbc
x=
and y =
b
adbc
Applying the method of elimination to the last two equations we find

z=
c
adbc
and t =
a
adbc
Hence,
1
1
=
ad bc
d b
c a
Norm of a Vector Function

The norm of a vector function will be needed in the coming sections. In one
dimension a norm is known as the absolute value. In multidimenesion, we
define the norm of a vector function x with components x1 , x2 , , xn by
||x|| = |x1 | + |x2 | + + |xn |.
From this definition one notices the following properties:
(i) If ||x|| = 0 then |x1 | + |x2 | + + |xn | = 0 and this implies that
|x1 | = |x2 | = = |xn | = 0. Hence, x = 0.
(ii) If is a scalar then ||x|| = |x1 | + |x2 | + + |xn | = ||(|x1 | + |x2 | +
+ |xn |) = ||||x||.
(iii) If x is vector function with components x1 , x2 , , xn and y with components y1 , y2 , , yn then
||x + y|| =
|x1 + y1 | + |x2 + y2 | + + |xn + yn |
(|x1 | + |x2 | + + |xn |) + (|y1 | + |y2 | + + |yn |)
=
||x|| + ||y||
Limits of Matrix Functions
If A(t) = (aij (t)) is an m n matrix such that limtt0 aij (t) = Lij exists for
all 1 i m and 1 j n then we define
lim A(t) = (Lij )
tt0
Example 26.5
Suppose that

A(t) =
t2 5t t3
2t
3
Find limt1 A(t).

8
Solution.

lim A(t) =
t1
limt1 (t2 5t) limt1 t3

limt1 2t
limt1 3

=
4 1
2 3
If one or more of the component function limits does not exist, then the limit
of the matrix does not exist. For example, if

t t1
A(t) =
0 et
then limt0 A does not exist since limt0 1t does not exist.
We say that A(t) is continuous at t = t0 if
lim A(t) = A(t0 )
tt0
Example 26.6
Show that the matrix

A(t) =
t t1
0 et
is continuous at t = 1.
Solution.
Since

lim A(t) =
t1
2 1/2
0 e2

= A(1)
then A(t) is continuous at t = 1

Most properties of limits for functions of a single variable are also valid for
limits of matrix functions.
Matrix Differentiation
Let A(t) be an m n matrix such that each entry is a differentiable function
of t. We define the derivative of A(t) to be
A(t + h) A(t)
h0
h
A0 (t) = lim
provided that the limit exists.
Example 26.7
Let
a11 (t) a12 (t)

a21 (t) a22 (t)
A(t) =
where the entries a11 , a12 , a21 , and a22 are differentiable. Find A0 (t).
Solution.
We have
A0 (t) =
limh0
"
limh0
limh0
a11 (t+h)a11 (t)

h
a21 (t+h)a21 (t)
h

=
A(t+h)A(t)
h
limh0
limh0
a011 (t) a012 (t)

a021 (t) a022 (t)
a12 (t+h)a12 (t)

h
a22 (t+h)a22 (t)
h
It follows from the previous example that the derivative of a matrix function
is the matrix of derivatives of its component functions. From this fact one
can check easily the following two properties of differentiation:
(i) If A(t) and B(t) are two m n matrices with both of them differentiable
then the matrix (A + B)(t) is also differentiable and
(A + B)0 (t) = A0 (t) + B0 (t)
(ii) If A(t) is an mn differentiable matrix and B(t) is an np differentiable
matrix then the product matrix AB(t) is also differentiable and
(AB)0 (t) = A0 (t)B(t) + A(t)B0 (t)
Example 26.8
Write the system
y10 = a11 (t)y1 (t) + a12 (t)y2 (t) + a13 (t)y3 (t) + g1 (t)
in matrix form.
10
Solution.
Let
y1 (t)
a11 (t) a12 (t) a13
g1 (t)
y(t) = y2 (t) , A(t) = a21 (t) a22 (t) a23 , g(t) = g2 (t)
y3 (t)
a31
a32 a33
g3 (t)
Then the given system can be written in the matrix form
y0 (t) = A(t)y(t) + g(t)
Matrix Integration:
Since the derivative of a matrix function is a matrix of derivatives then
it should not be surprising that antiderivatives of a matrix function can
be evaluated by performing the corresponding antidifferentiation operations
upon each component of the matrix function. That is, if A(t) is the m n
matrix
a11 (t) a12 (t) a1n (t)

a21 (t) a22 (t) a2n (t)
A(t) =
..
..
.
.
am1 (t) am2 (t) amn (t)
then
R
R
R
a
(t)dt
a
(t)dt
a
(t)dt
11
12
1n
Z
R a21 (t)dt R a22 (t)dt R a2n (t)dt
A(t)dt =
..
..
.
.
R
R
R
am1 (t)dt
am2 (t)dt
amn (t)dt
Example 26.9
Determine the matrix function A(t) if

2t
1
0
A (t) =
cos t 3t2
Solution.
We have

A(t) =
t2 + c11
t + c12
sin t + c21 t3 + c22

=
11
t2
t
sin t t3

+
c11 c12
c21 c22
Finally, we conclude this section by showing that

Z t
Z t
||x(s)||ds
x(s)ds||
||
t0
t0
To see this,
||
Rt
t0
x(s)ds|| =
R t

x
(s)ds
1
t

R t0

t0 x2 (s)ds

..

.

R t

x
(s)ds
n
t
Rt
Rt 0
Rt
| t0 x1 (s)ds| + | t0 x2 (s)ds| + | t0 xn (s)ds|
Rt
Rt
Rt
|x
(s)|ds
+
|x
(s)|ds
+
+
|x (s)|ds
1
2
tR0
t0
R t t0 n
t
(|x1 | + |x2 | + + |xn |)ds = t0 ||x(s)||ds
t0
Problem 26.1
Consider the following matrices

t1
t2
t 1
t+1
A(t) =
, B(t) =
, c(t) =
2
2t + 1
0 t+2
1
(a) Find 2A(t) - 3tB(t)
(b) Find A(t)B(t) - B(t)A(t)
(c) Find A(t)c(t)
(d) Find det(B(t)A(t))
Problem 26.2
Determine all values t such that A(t) is invertible and, for those t-values,
find A1 (t).

t+1
t
A(t) =
t
t+1
Problem 26.3
Determine all values t such that A(t) is invertible and, for those t-values,
find A1 (t).

sin t cos t
A(t) =
sin t cos t
12
Problem 26.4
Find

lim
t0
Problem 26.5
Find
sin t
t
3t

lim
t0
t cos t
sec t
3
t+1
2t
t2 1
tet tan t
t2 2 esin t
Problem 26.6
Find A0 (t) and A00 (t) if

A(t) =
sin t 3t
2
t +2 5
Problem 26.7
Express the system
y10 = t2 y1 + 3y2 + sec t
y20 = (sin t)y1 + ty2 5
in the matrix form
y0 (t) = A(t)y(t) + g(t)
Problem 26.8
Determine A(t) where
0
A (t) =
2t
1
cos t 3t2

, A(0) =
2 5
1 2
Problem 26.9
Determine A(t) where

1 t
1 1
1 2
00
0
A (t) =
, A(0) =
, A (0) =
0 0
2 1
2 3
Problem 26.10 R
t
Calculate A(t) = 0 B(s)ds where

es
6s
B(s) =
cos 2s sin 2s
13
Problem 26.11
Construct a 2 2 a nonconstant matrix function A(t) such that A2 (t) is a
constant matrix.
Problem 26.12
(a) Construct a 2 2 differentiable matrix function A(t) such that
d
d 2
A (t) 6= 2A A(t)
dt
dt
That is, the power rule is not true for matrix functions.
(b) What is the correct formula relating A2 (t) to A(t) and A(t)?
Problem 26.13
Transform the following third-order equation
y 000 3ty 0 + (sin 2t)y = 7et
into a first order system of the form
x0 (t) = Ax(t) + b(t)
Problem 26.14
By introducing new variables x1 and x2 , write y 00 2y + 1 = t as a system of
two first order linear equations of the form x0 + Ax = b
Problem 26.15
Write the differential equation y 00 + 4y 0 + 4y = 0 as a first order system.
Problem 26.16
Write the differential equation y 00 + ky 0 + (t 1)y = 0 as a first order system.
Problem 26.17
Change the following second-order equations to a first-order system.
y 00 5y 0 + ty = 3t2 , y(0) = 0, y 0 (0) = 1
Problem 26.18
Consider the following system of first-order linear equations.
3
x0 = 2 1 x
1
Find the second-order linear differential equation that x satisfies.
14
The Determinant of a Matrix

The determinant of a matrix function is the same as the determinant with
constant entries. So we will introduce the definition of determinant of a matrix with constant entries.
A permutation of the set S = {1, 2, . . . , n} is an arrangement of the elements of S in some order without omissions or repetitions. We write
= ((1)(2) (n)). In terms of functions, a permutation is a one-toone function from S onto S.
Let Sn denote the set of all permutations on S. How many permutations are
there in Sn ? We have n positions to be filled by n numbers. For the first
position, there are n possibilities. For the second there are n 1 possibilities,
etc. Thus, according to the multiplication rule of counting there are
n(n 1)(n 2) . . . 2.1 = n!
permutations.
Is there a way to list all the permutations of Sn ? The answer is yes and one
can find the permutations by using a permutation tree which we describe
in the following example
Problem 26.19
List all the permutations of S = {1, 2, 3, 4}.
An inversion is said to occur whenever a larger integer precedes a smaller
one. If the number of inversions is even (resp. odd) then the permutation is
said to be even (resp. odd). We define the sign of a permutation to be
a function sgn with domain Sn and range {1, 1} such that sgn() = 1
if is odd and sgn() = +1 if is even. For example, the permutation in
S6 defined by (1) = 3, (2) = 6, (3) = 4, (5) = 2, (6) = 1 is an even
permuatation since the inversions are (6,1),(6,3),(6,4),(6,5),(6,2),(3,2),(4,2),
and (5,2).
Let A be an n n matrix. An elementary product from A is a product of
n entries from A, no two of which come from the same row or same column.
Problem 26.20
List all elementary products from the matrices
(a)

a11 a12
,
a21 a22
15
(b)
a11 a12 a13

a21 a22 a23
a31 a32 a33
Let A be an n n matrix. Consider an elementary product of entries of
A. For the first factor, there are n possibilities for an entry from the first
row. Once selected, there are n 1 possibilities for an entry from the second
row for the second factor. Continuing, we find that there are n! elementary
products. They are the products of the form a1(1) a2(2) . . . an(n) , where is
a permutation of {1, 2, . . . , n}, i.e. a member of Sn .
Let A be an n n matrix. Then we define the determinant of A to be
the number
X
det(A) =
sgn()a1(1) a2(2) . . . an(n)
where the sum is over all permutations of {1, 2, . . . , n}.
Problem 26.21
Find det(A) if
(a)

A=
a11 a12
a21 a22

,
(b)
a11 a12 a13

A = a21 a22 a23
a31 a32 a33
The following theorem is of practical use. It provides a technique for evaluating determinants by greatly reducing the labor involved.
Theorem 26.1
Let A be an n n matrix.
(a) Let B be the matrix obtained from A by multiplying a row or a column
by a scalar c. Then det(B) = cdetA.
(b) Let B be the matrix obtained from A by interchanging two rows or two
16
columns of A. Then det(B) = det(A).

(c) If A has two identical rows or columns then its determinant is zero.
(d) Let B be the matrix obtained from A by adding c times a row (or a
column) to another row (column). Then det(B) = det(A).
(e) The determinant of the product of two n n matrices is the product of
their determinant.
(g) If B is the matrix whose columns are the rows of A then det(B) = det(A).
The proof of this theorem can be found in any textbook in elementary linear
algebra.
17
27
nth Order Linear Differential Equations:Existence

and Uniqueness
In the following three sections we carry the basic theory of second order linear
differential equations to nth order linear differential equation
y (n) + pn1 (t)y (n1) + + p1 (t)y 0 + p0 (t)y = g(t)
where the functions p0 , p1 , , pn1 and g(t) are continuous functions for
a < t < b.
If g(t) is not identically zero, then this equation is said to be nonhomogeneous; if g(t) is identically zero, then this equation is called homogeneous.
Existence and Uniqueness of Solutions
We begin by discussing the existence of a unique solution to the initial value
problem
(n1)
y(t0 ) = y0 , y 0 (t0 ) = y00 , , y (n1) (t0 ) = y0
, a < t0 < b
The following theorem is a generalization to Theorems 3.2 and 15.1

Theorem 27.1
The nonhomogeneous nth order linear differential equation
(1)
with initial conditions

(n1)
y(t0 ) = y0 , y 0 (t0 ) = y00 , , y (n1) (t0 ) = y0
, a < t0 < b
(2)
has a unique solution in a < t < b.

Proof.
Existence: The existence of a local solution is obtained here by transforming the problem into a first order system. This is done by introducing the
variables
x1 = y, x2 = y 0 , , xn = y (n1) .
18
In this case, we have

x01
x02
..
.
=
=
x2
x3
..
.
=
x0n1 =
xn
0
xn = pn1 (t)xn + p1 (t)x2 p0 (t)x1 + g(t)
Thus, we can write
0
x1
x2

x3
+
..
.
xn
the problem as a system:

1 0 0
0 1 0
..
..
..
.
.
.
0
0 0
p1 p2 p3
0
0
..
.
0
p0
0
0
..
.
pn1
x1
x2
x3
..
.
0
0
..
.

=

0
xn
g(t)
or in compact form
x0 (t) = A(t)x(t) + b(t), x(t0 ) = y0
where
A(t) =
x(t) =
0
0
..
.
1
0
..
.
0
1
..
.
0
0
..
.
(3)
0
0
..
.
pn1
0
0
0
0
p0 p1 p2 p3
x1
0
y0
0
x2
y00
.
x3
, b(t) = .. , y0 =
..
.
..
0
.
(n1)
y0
xn
g(t)
Note that if y(t) is a solution of (1) then the vector-valued function
y
y0
x(t) =
..
.
(n1)
y
19
is a solution to (3). Conversely, if the vector
x1
x2
x(t) = x3
..
.
xn
(n1)
(n)
= xn . Hence, x1
is a solution of (3) then x01 = x2 , x001 = x3 , , x1
0
xn = pn1 (t)xn pn2 (t)xn1 p0 (t)x1 + g(t) or
(n)
(n1)
x1 + pn1 (t)x1
(n2)
+ pn2 (t)x1
+ + p0 (t)x1 = g(t)
which means that x1 is a solution to (1).

Next, we start by reformulating (3) as an equivalent integral equation. Integration of both sides of (3) yields
Z t
Z t
0
x (s)ds =
[A(s)x(s) + b(s)]ds
(4)
t0
t0
Applying the Fundamental Theorem of Calculus to the left side of (4) yields
Z t
x(t) = x(t0 ) +
[A(s)x(s) + b(s)]ds
(5)
t0
Thus, a solution of (5) is also a solution to (3) and vice versa.

To prove the existence and uniqueness, we shall use again the method of
successive approximation as described in Theorem 8.1.
Letting
y0
y00
y0 =
..
.
(n1)
y0
we can introduce Picards iterations defined recursively as follows:
y0 (t)
y1 (t) =
y0
y0 + t0 [A(s)y0 (s) + b(s)]ds
y2 (t) =
..
.
y0 +
Rt
yN (t) = y0 +
Rt
t0
Rt
t0
[A(s)y1 (s) + b(s)]ds

..
.
[A(s)yN 1 (s) + b(s)]ds

20
Let
yN (t) =
y1,N
y2,N
..
.
yn,N
For i = 1, 2, , n, we are going to show that the sequence {yi,N (t)} converges
uniformly to a function yi (t) such that y(t) (with components y1 , y2 , , yn )is
a solution to (5) and hence a solution to (3).
Let [c, d] be a closed interval containing t0 and contained in (a, b). Then there
exist constants k0 , k1 , , kn1 such that
max |p0 (t)| k0 , max |p1 (t)| k1 , , max |pn1 (t)| kn1 .
ctd
ctd
ctd
This implies that

||A(t)x(t)|| =
|x2 | + |x3 | + + |xn1 + |p0 ||x1 | + |p1 ||x2 | + + |pn1 ||xn |
k0 |x1 | + (1 + k1 )|x2 | + + (1 + kn2 )|xn1 | + kn1 |xn | K||x||
for all c t d, where we define
||y|| = |y1 | + |y2 | + + |yn |
and where
K = k0 + (1 + k1 ) + + (1 + kn2 ) + kn1 .
It follows that for 1 i n
Rt
|yi,N yi,N 1 | ||yN yN 1 || t0 ||A(s) (yN 1 yN 2 )||ds
Rt
K t0 ||yN 1 yN 2 ||ds
But
||y1 y0 ||
Rt
t0
||A(s) y0 + b(s)||ds
M (t t0 )
where
M = K||y0 || + max ||b(t)||
ctd
An easy induction yields that

|yi,N +1 yi,N | ||yN +1 yN || M K N
21
(b a)N +1
(t t0 )N +1
M KN
.
N!
N!
Since
(b a)N +1
M K(ba)
=
(e
1)
(N
+
1)!
K
N =0
P
then by Weierstrass M-test we conclude that the series
N =0 [yi,N yi,N 1 ]
converges uniformly for all c t d. But
M KN
yi,N (t) =
N
1
X
[yi,k+1 (t) yi,k (t)] + yi,0
k=0
Thus, the sequence {yi,N } converges uniformly to a function yi (t) for all
c t d.
The function yi (t) is a continuous function (a uniform limit of a sequence
of continuous function is continuous). Also we can interchange the order of
taking limits and integration for such sequences. Therefore
y(t) =
limN yN (t)
Rt
= y0 + limN t0 (A(s)yN 1 + b(s))ds
Rt
= y0 + t0 limN (A(s)yN 1 + b(s))ds
Rt
=
y0 + t0 (A(s)y + b(s))ds
This shows that y(t) is a solution to the integral equation (5) and therefore
a solution to (3).
Uniqueness:
The uniqueness follows from Gronwall Inequality (See Problem 8.11). Suppose that y(t) and z(t) are two solutions to the initial value problem, it
follows that for all a < t < b we have
Z t
||y(t) z(t)||
K||y(s) z(s)||ds
t0
Letting u(t) = ||y(t) z(t)|| we have

Z t
u(t)
Ku(s)ds
t0
so that by Gronwalls inequality u(t) 0 and therefore y(t) = z(t) for all
a < t < b. This completes a proof of the theorem
22
Example 27.1
Find the largest interval where
(t2 16)y (4) + 2y 00 + t2 y = sec t, y(3) = 1, y 0 (3) = 3, y 00 (3) = 1
is guaranteed to have a unique solution.
Solution.
We first put it into standard form
y (4) +
2
t2
sect
00
y
+
y
=
t2 16
t2 16
t2 16
The coefficient functions are continuous for all t 6= 4 and t 6= (2n + 1) 2 .

Since t0 = 3 then the largest interval where the given initial value problem
is guaranteed to have a unique solution is the ineterval 2 < t < 4
For Problems 28.1 - 28.3, use Theorem 27.1 to find the largest interval
a < t < b in which a unique solution is guaranteed to exist.
Problem 27.1
y 000
t2
1
y 00 + ln (t + 1)y 0 + (cos t)y = 0, y(0) = 1, y 0 (0) = 3, y 00 (0) = 0
9
Problem 27.2
y 000 +
1 0
y + (tan t)y = 0, y(0) = 0, y 0 (0) = 1, y 00 (0) = 2
t+1
Problem 27.3
y 00
t2
1
y 00 + ln (t2 + 1)y 0 + (cos t)y = 0, y(0) = 1, y 0 (0) = 3, y 00 (0) = 0
+9
Problem 27.4
Determine the value(s) of r so that y(t) = ert is a solution to the differential
equation
y 000 2y 00 y 0 + 2y = 0
23
Problem 27.5
Transform the following third-order equation
y 000 3ty 0 + (sin 2t)y = 7et
into a first order system of the form
x0 (t) = Ax(t) + b(t)
24
28
The General Solution of nth Order Linear

Homogeneous Equations
In this section we consider the question of solving the homogeneous equation

y (n) + pn1 (t)y (n1) + + p1 (t)y 0 + p0 (t)y = 0
(6)
where p0 (t), p1 (t), , pn1 (t) are continuous functions in the interval a <
t < b.
The next theorem shows that any linear combination of solutions to the
homogeneous equation is also a solution.
In what follows and for the simplicity of arguments we introduce the function
L defined by
L[y] = y (n) + pn1 (t)y (n1) + + p1 (t)y 0 + p0 (t)y
Theorem 28.1 (Linearity)
If y1 and y2 are n times differentiable and 1 and 2 are scalars then L
satisfies the property
L[1 y1 + 2 y2 ] = 1 L[y1 ] + 2 L[y2 ]
Proof.
Indeed, we have
L[1 y1 + 2 y2 ] =
(1 y1 + 2 y2 )(n) + pn1 (t)(1 y1 + 2 y2 )(n1) +
+
p0 (t)(1 y1 + 2 y2 )
(n1)
(n)
+ + 1 p1 (t)y10 + 1 p0 (t)y1 )
= (1 y1 + 1 pn1 (t)y1
(n)
(n1)
+ (2 y2 + 2 pn1 (t)y2
+ + 2 p1 (t)y20 + 2 p0 (t)y2 )
(n)
(n1)
=
1 (y1 + pn1 (t)y1
+ + p1 (t)y10 + p0 (t)y1 )
(n)
(n1)
+
2 (y2 + pn1 (t)y2
+ + p1 (t)y20 + p0 (t)y2 )
=
1 L[y1 ] + 2 L[y2 ]
The above property applies for any number of functions.
An important consequence of this theorem is the following result.
Corollary 28.1 (Principle of Superposition)
If y1 , y2 , , yr satisfy the homogeneous equation (6) and if 1 , 2 , , r
are any numbers, then
y(t) = 1 y1 + 2 y2 + + r yr
also satisfies the homogeneous equation (6).
25
Proof.
Since y1 , y2 , , yr are solutions to (6) then L[y1 ] = L[y2 ] = = L[yr ] = 0.
Now, using the linearity property of L we have
L[1 y1 + 2 y2 + + r yr ] = 1 L[y1 ] + 2 L[y2 ] + + r L[yr ]
=
0 + 0 + + 0 = 0
The principle of superposition states that if y1 , y2 , , yr are solutions to
(6) then any linear combination is also a solution. The next question that we
consider is the question of existence of n solutions y1 , y2 , , yn of equation
(6) such that every solution to (6) can be written as a linear combination
of these functions. We call such a set a functions a fundamental set of
solutions. Note that the number of solutions comprising a fundamental set
is equal to the order of the differential equation. Also, note that the general
solution to (6) is then given by
y(t) = c1 y1 (t) + c2 y2 (t) + + cn yn (t).
Next, we consider a criterion for testing n solutions for a fundamental set.
For that we first introduce the following definition:
For n functions y1 , y2 , , yn , we define the Wronskian of these functions
to be the determinant

y1 (t)

y
(t)
y
(t)
2
n

0
0
y10 (t)
y2 (t)
yn (t)

00

y200 (t)
yn00 (t)
W (t) = y1 (t)

..
..
..

.
.
.

(n1)
(n1)
(n1)
y1
(t) y2
(t) yn
(t)
Theorem 28.2 (Criterion for identifying fundamental sets)
Let y1 (t), y2 (t), , yn (t) be n solutions to the homogeneous equation (6).
If there is a a < t0 < b such that W (t0 ) 6= 0 then {y1 , y2 , , yn } is a
fundamental set of solutions.
Proof.
We need to show that if y(t) is a solution to (6) then we can write y(t) as a
linear combination of y1 , y2 (t), , yn (t). That is
y(t) = c1 y1 (t) + c2 y2 (t) + + cn yn .
26
So the problem reduces to finding the constants c1 , c2 , , cn . These are

found by solving the following linear system of n equations in the unknowns
c 1 , c2 , , cn :
c1 y1 (t0 ) + c2 y2 (t0 ) + + cn yn (t0 )
c1 y10 (t0 ) + c2 y20 (t0 ) + + cn yn0 (t0 )
...................................................
(n1)
(n1)
(n1)
c1 y1
(t) + c2 y2
(t) + + cn yn
(t)
=
y(t0 )
=
y 0 (t0 )
=
...
(n1)
= y
(t0 )
Solving this system using Cramers rule we find

ci =
Wi (t0 )
,
W (t0 )
1in
where

y1 (t0 )
y2 (t0 )

y10 (t0 )
y20 (t0 )

00

y200 (t0 )
Wi (t0 ) = y1 (t0 )

..
..

.
.
(n1)
(n1)
y1
(t0 ) y2
(t0 )

(n1)
(n1)
y
(t0 ) yn
(t0 )
y(t0 )
y 0 (t0 )
y 00 (t0 )
..
.
yn (t0 )
yn0 (t0 )
yn00 (t0 )
That is, Wi (t0 ) is the determinant of W with the ith column being replaced
by the right-hand column of the above system. Note that c1 , c2 , , cn exist
since W (t0 ) 6= 0
As a first application to this result, we establish the existence of fundamental
sets
Theorem 28.3
The linear homogeneous differential equation
y (n) + pn1 (t)y (n1) + + p1 (t)y 0 + p0 (t)y = 0
where pn1 (t), , p1 (t), p0 (t) are continuous functions in a < t < b has a
fundamental set {y1 , y2 , , yn }.
Proof.
Pick a t0 in the interval a < t < b and consider the n initial value problems
y (n) +pn1 (t)y (n1) + +p1 (t)y 0 +p0 (t)y = 0, y(t0 ) = 1, y 0 (t0 ) = 0, y 00 (t0 ) = 0, , y (n1) (t0 ) = 0
27
..
.
Then by Theorem 27.1, we can find unique solutions {y1 , y2 , , yn }. This

set is a fundamental set by the previous theorem since

1 0 0 0

0 1 0 0

W (t) = .. .. ..
. = 1 6= 0
. . . ..

0 0 0 1
Theorem 28.2 says that if one can find a < t0 < b such that W (t0 ) 6= 0
then the set {y1 , y2 , , yn } is a fundamental set of solutions. The following
theorem shows that the condition W (t0 ) 6= 0 implies that W (t) 6= 0 for all
t in the interval (a, b). That is, the theorem tells us that we can choose our
test point t0 on the basis of convenience-any test point t0 will do.
Theorem 28.4 (Abels)
Let y1 (t), y2 (t), , yn be n solutions to equation (6). Then
(1) W (t) satisfies the differential equation W 0 (t) + pn1 (t)W (t) = 0;
(2) If t0 is any point in (a, b) then
W (t) = W (t0 )e
Rt
t0
pn1 (s)ds
Thus, if W (t0 ) 6= 0 then W (t) 6= 0 for all a < t < b.

Proof.
(1) By introducing the variables x1 = y, x2 = y 0 , x3 = y 00 , , xn = y (n1) we
can write the differential equation as a first order system in the form
x0 (t) = A(t)x(t)
where
A(t) =
0
0
..
.
1
0
..
.
0
1
..
.
0
0
..
.
0
0
..
.
, x(t) =
0
0
0
0
p0 p1 p2 p3 pn1
28
x1
x2
x3
..
.
xn
We will show in Section 33 that for any linear system of the form
x0 (t) = A(t)x(t)
we have
W 0 (t) = (a11 + a22 + . . . + ann )W (t)
In our case
a11 + a22 + . . . + ann = pn1 (t)
so that
W 0 (t) + pn1 (t)W (t) = 0
(2) The previous differential equation can be solved by the method of integrating factor to obtain
Rt
W (t) = W (t0 )e
t0
pn1 (s)ds
Example 28.1
Use the Abels formula to find the Wronskian of the DE: ty 000 + 2y 00 t3 y 0 +
2
et y = 0
Solution.
The original equation can be written as
2
et
2
y=0
y + y 00 t2 y 0 +
t
t
000
By Abels formula the Wronskian is

W (t) = Ce
2
dt
t
C
t2
Example 28.2
Consider the linear system
x0 (t) = A(t)x(t)
where

A=
a11 a12
a21 a22
Show that for any two solutions Y1 and Y2 we have

W 0 (t) = (a11 + a22 )W (t).
29
Solution.
Suppose that

Y1 =
u1
u2

, Y2 =
v1
v2
are solutions to the given system. Then we have

u
v
1
1
d

W 0 (t) =
dt u

v
2
2

0

u1 v10 u1 v1

+
=
u2 v2 u02 v20
But
and

0
u1 v10 a11 u1 + a12 u2 a11 v1 + a12 v2

u2 v 2 =
u2
v2

= a11 W (t)

u1 v 1
u1
v1

0
u2 v20 = a21 u1 + a22 u2 a21 v1 + a22 v2

= a22 W (t)

It follows that
W 0 (t) = (a11 + a22 )W (t)
We end this section by showing that the converse of Theorem 28.2 is also
true.
Theorem 28.5
If {y1 , y2 , , yn } is a fundamental set of solutions to
y (n) + pn1 (t)y (n1) + + p1 (t)y 0 + p0 (t)y = 0
where pn1 (t), , p1 (t), p0 (t) are continuous functions in a < t < b then
W (t) 6= 0 for all a < t < b.
Proof.
Let t0 be any point in (a, b). By Theorem 27.1, there is a unique solution y(t)
to the initial value problem
y (n) +pn1 (t)y (n1) + +p1 (t)y 0 +p0 (t)y = 0, y(t0 ) = 1, y 0 (t0 ) = 0, , y (n1) (t0 ) = 0
30
Since {y1 , y2 , , yn } is a fundamental set then there exist unique constants

c1 , c2 , , cn such that
c1 y1 (t) + c2 y2 (t)
c1 y10 (t) + c2 y20 (t)
..
.
(n1)
c1 y1
(n1)
(t) + c2 y2
cn yn (t)
cn yn0 (t)
(n1)
(t) cn yn
=
=
y(t)
y 0 (t)
(t) = y (n1) (t)
for all a < t < b. In particular for t = t0 we obtain the system
c1 y1 (t) + c2 y2 (t)
c1 y10 (t) + c2 y20 (t)
..
.
(n1)
c1 y1
(n1)
(t) + c2 y2
cn yn (t)
cn yn0 (t)
(n1)
(t) cn yn
= 1
= 0
(t) = 0
This system has a unique solution (c1 , c2 , , cn ) where

ci =
Wi
W (t0 )
and Wi is the determinant W with the
1
0
..
.
0
ith column replaced by the column
Note that for c1 , c2 , , cn to exist we must have W (t0 ) 6= 0. By Abels theorem we conclude that W (t) 6= 0 for all a < t < b
In Problems 28.1 - 28.3, show that the given solutions form a fundamental set for the differential equation by computing the Wronskian.
Problem 28.1
y 000 y 0 = 0, y1 (t) = 1, y2 (t) = et , y3 (t) = et
31
Problem 28.2
y (4) + y 00 = 0, y1 (t) = 1, y2 (t) = t, y3 (t) = cos t, y4 (t) = sin t
Problem 28.3
t2 y 000 + ty 00 y 0 = 0, y1 (t) = 1, y2 (t) = ln t, y3 (t) = t2
Use the fact that the solutions given in Problems 28.1 - 28.3 for a fundamental
set of solutions to solve the following initial value problems.
Problem 28.4
y 000 y 0 = 0, y(0) = 3, y 0 (0) = 3, y 00 (0) = 1
Problem 28.5
) = 2 + , y 0 ( 2 ) = 3, y 00 ( 2 ) = 3, y 000 ( 2 ) = 1.
y (4) + y 00 = 0, y( |pi
2
Problem 28.6
t2 y 000 + ty 00 y 0 = 0, , y(1) = 1, y 0 (1) = 2, y 00 (1) = 6

Problem 28.7
In each question below, show that the Wronskian determinant W (t) behaves
as predicted by Abels Theorem. That is, for the given value of t0 , show that
W (t) = W (t0 )e
Rt
t0
pn1 (s)ds
(a) W (t) found in Problem 28.1 and t0 = 1.

(b) W (t) found in Problem 28.2 and t0 = 1.
(c) W (t) found in Problem 28.3 and t0 = 2.
32
Problem 28.8
Determine W (t) for the differential equation y 000 +(sin t)y 00 +(cos t)y 0 +2y = 0
such that W (1) = 0.
Problem 28.9
Determine W (t) for the differential equation t3 y 000 2y = 0 such that W (1) =
3.
Problem 28.10
Consider the initial value problem
y 000 y 0 = 0, y(0) = , y 0 (0) = , y 00 (0) = 4.
The general solution of the differential equation is y(t) = c1 + c2 et + c3 et .
(a) For what values of and will limt y(t) = 0?
(b) For what values and will the solution y(t) be bounded for t 0, i.e.,
|y(t)| M for all t 0 and for some M > 0? Will any values of and
produce a solution y(t) that is bounded for all real number t?
Problem 28.11
Consider the differential equation y 000 + p2 (t)y 00 + p1 (t)y 0 = 0 on the interval
1 < t < 1. Suppose it is known that the coefficient functions p2 (t) and p1 (t)
are both continuous on 1 < t < 1. Is it possible that y(t) = c1 + c2 t2 + c3 t4
is the general solution for some functions p1 (t) and p2 (t) continuous on 1 <
t < 1?
(a) Answer this question by considering only the Wronskian of the functions
1, t2 , t4 on the given interval.
(b) Explicitly determine functions p1 (t) and p2 (t) such that y(t) = c1 + c2 t2 +
c3 t4 is the general solution of the differential equation. Use this information,
in turn, to provide an alternative answer to the question.
Problem 28.12
(a) Find the general solution to y 000 = 0.
(b) Using the general solution in part (a), construct a fundamental set
{y1 (t), y2 (t), y3 (t)} satisfying the following conditions
y1 (1) = 1, y10 (1) = 0, y100 (1) = 0.
y2 (1) = 0, y10 (1) = 1, y100 (1) = 0.
y1 (1) = 0, y10 (1) = 0, y100 (1) = 1.
33
29
Fundamental Sets and Linear Independence
In Section 28 we established the existence of fundamental sets. There remain two questions that we would like to answer. The first one is about the
number of fundamental sets. That is how many fundamental sets are there.
It turns out that there are more than one. In this case, our second question
is about how these sets are related. In this section we turn our attention to
these questions.
We start this section by the following observation. Suppose that the Wronskian of n solutions {y1 , y2 , , yn } to the nth order linear homogeneous
differential equation is zero. In terms of linear algebra, this means that one
of the columns of W can be written as a linear combination of the remaining
columns. For the sake of argument, suppose that the last column is a linear
combination of the remaining columns:
yn1
y2
y1
yn
0
yn1
y20
y10
yn0
+ + cn1
+ c2
= c1
..
..
..
..
.
.
.
.
(n1)
(n1)
(n1)
y2
y1
yn
(n1)
yn1
This implies that

yn (t) = c1 y1 (t) + c2 y2 (t) + + cn1 yn1 (t).
Such a relationship among functions merit a name:
We say that the functions f1 , f2 , , fm are linearly dependent on an
interval I if at least one of them can be expressed as a linear combination
of the others on I; equivalently, they are linearly dependent if there exist
constants c1 , c2 , , cm not all zero such that
c1 f1 (t) + c2 f2 (t) + + cm fm (t) = 0
(7)
for all t in I. A set of functions that is not linearly dependent is said to be

linearly independent. This means that a sum of the form (7) implies that
c1 = c2 = = cm = 0.
Example 29.1
Show that the functions f1 (t) = et , f2 (t) = e2t , and f3 (t) = 3et 2e2t are
linearly dependent on (, ).
34
Solution.
Since f3 (t) = 3f1 (t) 2f2 (t) then the given functions are linearly dependent
on (, )
The concept of linear independence can be used to test a fundamental set of
solutions to the equation
y (n) + pn1 (t)y (n1) + + p1 (t)y 0 + p0 (t)y = 0
(8)
Theorem 29.1
The solution set {y1 , y2 , , yn } is a fundamental set of solutions to
y (n) + pn1 (t)y (n1) + + p1 (t)y 0 + p0 (t)y = 0
where pn1 (t), , p1 (t), p0 (t) are continuous functions in a < t < b if and
only if the functions y1 , y2 , , yn are linearly independent.
Proof.
Suppose first that {y1 , y2 , , yn } is a fundamental set of solutions. Then by
Theorem 28.5 there is a < t0 < b such that W (t0 ) 6= 0. Suppose that
c1 y1 (t) + c2 y2 (t) + + cn yn (t) = 0
for all a < t < b. By repeated differentiation of the previous equation we find
c1 y10 (t) + c2 y20 (t) + + cn yn0 (t)
c1 y100 (t) + c2 y200 (t) + + cn yn00 (t)
..
.
(n1)
c1 y1
(n1)
(t) + c2 y2
(n1)
(t) + + cn yn
= 0
= 0
(t) = 0
Thus, one finds c1 , c2 , , cn by solving the system

c1 y1 (t0 ) + c2 y2 (t0 ) + + cn yn (t0 )
c1 y10 (t0 ) + c2 y20 (t0 ) + + cn yn0 (t0 )
c1 y100 (t0 ) + c2 y200 (t0 ) + + cn yn00 (t0 )
..
.
(n1)
c1 y1
(n1)
(t0 ) + c2 y2
(n1)
(t0 ) + + cn yn
Solving this system using Cramers rule one finds

c 1 = c 2 = , cn =
35
0
=0
W (t0 )
= 0
= 0
= 0
(t0 ) = 0
Thus, y1 (t), y2 (t), , yn (t) are linearly independent.

Conversely, suppose that {y1 , y2 , , yn } is a linearly independent set. Suppose that {y1 , y2 , , yn } is not a fundamental set of solutions. Then by
Theorem 28.2, W (t) = 0 for all a < t < b. Choose any a < t0 < b. Then
W (t0 ) = 0. But this says that the matrix
y1 (t0 )
y2 (t0 )
yn (t0 )
y10 (t0 )
yn0 (t0 )
y20 (t0 )
..
..
..
..
.
.
.
.
(n1)
y1
(n1)
(t0 ) y2
(n1)
(t0 ) yn
(t0 )
is not invertible which means that there exist c1 , c2 , , cn not all zero such
that
c1 y1 (t0 ) + c2 y2 (t0 ) + + cn yn (t0 )
= 0
0
0
0
c1 y1 (t0 ) + c2 y2 (t0 ) + + cn yn (t0 )
= 0
= 0
c1 y100 (t0 ) + c2 y200 (t0 ) + + cn yn00 (t0 )
..
.
(n1)
c1 y1
(n1)
(t0 ) + c2 y2
(n1)
(t0 ) + + cn yn
(t0 ) = 0
Now, let y(t) = c1 y1 (t)+c2 y2 (t)+ +cn yn (t) for all a < t < b. Then y(t) is a
solution to the differential equation and y(t0 ) = y 0 (t0 ) = = y (n1) (t0 ) = 0.
But the zero function also is a solution to the initial value problem. By
the existence and uniqueness theorem (i.e, Theorem 27.1) we must have
c1 y1 (t) + c2 y2 (t) + + cn yn (t) = 0 for all a < t < b with c1 , c2 , , cn not
all equal to 0. But this means that y1 , y2 , , yn are linearly depedent which
is contradicts our assumption that y1 , y2 , , yn are linearly independent
Remark 29.1
The fact that {y1 , y2 , , yn } are solutions is very critical. That is, if y1 , y2 , , yn
are merely differentiable functions then it is possible for them to be linearly
independent and yet have a vanishing Wronskian. See Section 17.
Next, we will show how to generate new fundamental sets from a given one
and therefore establishing the fact that a linear homogeneous differential
equation has many fundamental sets of solutions. We also show how different
fundamental sets are related to each other. For this, let us start with a fundamental set {y1 , y2 , , yn } of solutions to (8). If {y 1 , y 2 , , y n } are n solutions then they can be written as linear combinations of the {y1 , y2 , , yn }.
36
That is,
a11 y1 + a21 y2 + + an1 yn = y 1
a12 y1 + a22 y2 + + an2 yn = y 2
..
.
a1n y1 + a2n y2 + + ann yn = y n
or in matrix form as
y1 y2 yn
y1 y2 yn
a11 a12 a13

a21 a22 a23
..
..
..
.
.
.
an1 an2 an3
a1n
a2n
..
.
ann
Theorem 29.2
{y 1 , y 2 , , y n } is a fundamental set if and only if det(A) 6= 0 where A is the
coefficient matrix of the above matrix equation.
Proof.
By differentiating (n-1) times the system
a11 y1 + a21 y2 + + an1 yn = y 1
a12 y1 + a22 y2 + + an2 yn = y 2
..
.
a1n y1 + a2n y2 + + ann yn = y n
one can easily check that
y1
y 01
..
.
(n1)
y1
y2
y 02
..
.
(n1)
y2
yn
y 0n
..
.
(n1)
yn
y1
y10
..
.
(n1)
y1
y2
y20
..
.
(n1)
y2
yn
yn0
..
.
(n1)
yn
a11
a21
.
..
a12
a22
..
.
an1
an2
ann
By taking the determinant of both sides and using the fact that the determinant of a product is the product of determinants then we can write
W (t) = det(A)W (t)
Since W (t) 6= 0 then W (t) 6= 0 (i.e., {y 1 , y 2 , , y n } is a fundamental set) if
and only if det(A) 6= 0
37
a1n
a2n
..
.
Example 29.2
The set {y1 (t), y2 (t), y3 (t)} = {1, et , et } is fundamental set of solutions to
the differential equation
y 000 y 0 = 0
(a) Show that {y 1 (t), y 2 (t), y 3 (t)} = {cosh t, 1 sinh t, 2 + sinh t} is a solution
set.
(b) Determine the coefficient matrix A described in the previous theorem.
(c) Determine whether this set is a fundamental set by calculating the determinant of the matrix A.
Solution.
000
0
(a) Since y 01 = sinh t, y 001 = cosh t, and y 000
1 (t) = sinh t then y 1 y 1 = 0 so that
y 1 is a solution. A similar argument holds for y 2 and y 3 .
(a) Since y 1 (t) = 0 1 + 12 et + 12 et , y 2 (t) = 1 1 21 et + 21 et , y 3 (t) =
2 1 + 12 et 21 et then
0
1
2
A = 1/2 1/2 1/2
1/2 1/2 1/2
(c) One can easily find that det(A) =
fundemantal set of solutions
3
2
6= 0 so that {y 1 (t), y 2 (t), y 3 (t)} is a
Problem 29.1
Determine if the following functions are linearly independent
y1 (t) = e2t , y2 (t) = sin (3t), y3 (t) = cos t
Problem 29.2
Determine whether the three functions : f (t) = 2, g(t) = sin2 t, h(t) = cos2 t,
are linearly dependent or independent on < t <
Problem 29.3
Determine whether the functions, y1 (t) = 1; y2 (t) = 1 + t; y3 (t) = 1 + t + t2 ;
are linearly dependent or independent. Show your work.
Problem 29.4
Consider the set of functions {y1 (t), y2 (t), y3 (t)} = {t2 + 2t, t + 1, t + }. For
what value(s) is the given set linearly depedent on the interval < t <
?
38
Problem 29.5
Determine whether the set {y1 (t), y2 (t), y3 (t)} = {t|t| + 1, t2 1, t} is linearly
independent or linearly dependent on the given interval
(a) 0 t < .
(b) < t 0.
(c) < t < .
In Problems 29.6 - 29.7, for each differential equation, the corresponding set
of functions {y1 (t), y2 (t), y3 (t)} is a fundamental set of solutions.
(a) Determine whether the given set {y 1 (t), y 2 (t), y 3 (t)} is a solution set to
the differential equation.
(b) If {y 1 (t), y 2 (t), y 3 (t)} is a solution set then find the coefficient matrix A
such that

y1
a11 a12 a13
y1
y 2 = a21 a22 a23 y2
y3
a31 a32 a33
y3
(c) If {y 1 (t), y 2 (t), y 3 (t)} is a solution set, determine whether it is a fundamental set by calculating the determinant of A.
Problem 29.6
y 000 + y 00 = 0
{y1 (t), y2 (t), y3 (t)} = {1, t, et }
{y 1 (t), y 2 (t), y 3 (t)} = {1 2t, t + 2, e(t+2) }
Problem 29.7
t2 y 000 + ty 00 y 0 = 0, t > 0
{y1 (t), y2 (t), y3 (t)} = {t, ln t, t2 }
{y 1 (t), y 2 (t), y 3 (t)} = {2t2 1, 3, ln (t3 )}
39
30
Higher Order Homogeneous Linear Equations with Constant Coefficients
In this section we investigate how to solve the nth order linear homogeneous
equation with constant coefficients
y (n) + an1 y (n1) + + a1 y 0 + a0 = 0
(9)
The general solution is given by

y(t) = c1 y1 + c2 y2 + + cn yn
where {y1 , y2 , , yn } is a fundamental set of solutions.
What was done for second-order, linear homogeneous equations with constant
coefficients holds, with the obvious modifications, for higher order analogs.
As for the second order case, we seek solutions of the form y(t) = ert , where
r is a constant (real or complex-valued) to be found. Inserting into (9) we
find
(rn + an1 rn1 + a1 r + a0 )ert = 0
We call P (r) = rn + an1 rn1 + a1 r + a0 the characteristic polynomial
and the equation
rn + an1 rn1 + a1 r + a0 = 0
(10)
the characteristic equation. Thus, for y(t) = ert to be a solution to (9) r

must satisfy (11).
Example 30.1
Solve: y 000 4y 00 + y 0 + 6y = 0
Solution.
The characteristic equation is
r3 4r2 + r + 6 = 0
We can factor to find the roots of the equation. A calculator can efficiently
do this, or you can use the rational root theorem to get
(r + 1)(r 2)(r 3) = 0
40
Thus, the roots are

r = 1, r = 2, r = 3
The Wronskian
t
e
e2t e3t

t
2t
3t
e
t 2e2t 3e3t
e
4e 9e

= 12e4t 6= 0

Hence, {et , e2t , e3t } is a fundamental set of solutions and the general solution
is
y = c1 et + c2 e2t + c3 e3t
In the previous example, the characteristic solution had three distinct roots
and the corresponding set of solutions formed a fundamental set. This is
always true according to the following theorem.
Theorem 30.1
Assume that the characteristic equation
rn + an1 rn1 + a1 r + a0 = 0
has n distinct roots r1 , r2 , , rn (real valued or complex valued). Then the
set of solutions {er1 t , er2 t , , ern t } is a fundamental set of solution to the
equation
y (n) + an1 y (n1) + + a1 y 0 + a0 = 0
Proof.
For a fixed number t0 we consider the Wronskian

er1 t0
er2 t0
ern t0

r1 er1 t0
r2 er2 t0 rn ern t0

r2 er1 t0
r22 er2 t0 rn2 ern t0
W (t0 ) = 1

..
..

.
n1. r1 t0 n1 r2 t0
n1 rn t0
r1 e
r2 e
rn e
Now, in linear algebra one proves that if a row or a column of a matrix

is multiplied by a constant then the determinant of the new matrix is the
determinant of the old matrix multiplied by that constant. It follows that

1

1
1

r1

r
r
2
n

2
2
2

r2 rn
W (t) = er1 t0 er2 t0 ern t0 r1
..
..
.
.
n1 n1
n1
r1
r2
rn
41
The resulting determinant above is the well-known Vandermonde determinant. Its value is the product of all factors of the form rj ri where j > i.
Since rj 6= ri for i 6= j then this determinant is not zero and consequently
W (t0 ) 6= 0. This establishes that {er1 t , er2 t , , ern t } is a fundamental set of
solutions
Next, we consider characteristic equations whose roots are not all distinct.
For example, if is a real root that appears k times (in this case we say that
is a root of multiplicity k), that is, P (r) = (r )k q(r), where q() 6= 0,
then the k linearly independent solutions are given by
et , tet , t2 et , , tk1 et
The remaining n k solutions needed to complete the fundamental set of
solutions are determined by examining the roots of q(r) = 0.
If, on the other hand, i are conjugate complex roots each of multiplicity
k, that is
P (r) = (r r1 )k (r r1 )k p(r)
where r1 = + i and p(r1 ) 6= 0, p(r1 ) 6= 0 then the 2k linearly independent
solutions are given by
et cos t, tet cos t, , tk1 et cos t
and
et sin t, tet sin t, , tk1 et sin t
Example 30.2
Find the solution to
y (5) + 4y 000 = 0, y(0) = 2, y 0 (0) = 3, y 00 (0) = 1, y 000 (0) = 1, y(4)(0) = 1
Solution.
We have the characteristic equation
r5 + 4r3 = r3 (r2 + 4) = 0
Which has a root of multiplicity 3 at r = 0 and complex roots r = 2i and
r = 2i. We use what we have learned about repeated roots and complex
roots to get the general solution. Since the multiplicity of the repeated root
is 3, we have
y1 (t) = 1, y2 (t) = t, y3 (t) = t2 .
The complex roots give the other two solutions
42
y4 (t) = cos (2t) and y5 (t) = sin (2t)

The general solution is
y(t) = c1 + c2 t + c3 t2 + c4 cos (2t) + c5 sin (2t)
Now Find the first four derivatives
y 0 (t)
y 00 (t)
y 000 (t)
y(4)(t)
= c2 + 2c3 t 2c4 sin (2t) + 2c5 cos (2t)

=
2c3 4c4 cos (2t) 4c5 sin (2t)
=
8c4 sin (2t) 8c5 cos (2t)
=
16c4 cos (2t) + 16c5 sin (2t)
Next plug in the initial conditions to get

2
3
1
1
1
= c1 + c4
= c2 + 2c5
= 2c3 4c4
=
8c5
=
16c4
Solving these equations we find

c1 = 31/16, c2 = 23/4, c3 = 5/8, c4 = 1/16, c5 = 1/8
The solution is
y(t) =
5
1
1
31 23
+ t + t2 +
cos (2t) + sin (2t)
15
4
8
16
8
Solving the Equation y (n) ay = 0.

The characteristic equation corresponding to the differential equation y (n)
ay = 0 is rn a = 0. The fundamental theorem of algebra asserts the existence
of exactly n roots (real or complex-valued). To find these roots, we write a
in polar form a = |a|ei where = 0 if a > 0 and = is a < 0 (since
ei = cos + i sin = 1). Also, since ei2k = 1 for any integer k then we
can write
a = |a|e(+2k)i
Thus, the charactersitic equation is
rn = |a|e(+2k)i .
43
Taking the nth root of both sides we find

1
r = |a| n e
(+2k)i
n
The n distinct roots are generated by taking k = 0, 1, 2, , n 1. We illustrate this in the next example.
Example 30.3
Find the general solution of y (6) + y = 0.
Solution.
In this case the characteristic equation is r6 + 1 = 0 or r6 = 1 = ei(2k+1) .
(2k+1)
Thus, r = ei 6 where k is an integer. Replacing k by 0,1,2,3,4,5 we find
r0
r1
r2
r3
r4
r5
3
=
+ 2i
2
=
i
3
= 2 + 2i
= 23 2i
= i
3
=
2i
2
Thus, the general solution is

y(t) = c1 e
3
t
2
3
3
3
t
t
t
t
t
t
2
2
cos +c2 e sin +c3 e
cos +c4 e 2 t sin +c5 cos t+c6 sin t
2
2
2
2
Problem 30.1
Solve y 000 + y 00 y 0 y = 0
Problem 30.2
Find the general solution of 16y (4) 8y 00 + y = 0.
Problem 30.3
Solve the following constant coefficient differential equation :
y 000 y = 0.
Problem 30.4
Solve y (4) 16y = 0
44
Problem 30.5
Solve the initial-value problem
y 000 + 3y 00 + 3y 0 + y = 0, y(0) = 0, y 0 (0) = 1, y 00 (0) = 0.
Problem 30.6
Given that r = 1 is a solution of r3 + 3r2 4 = 0, find the general solution
to
y 000 + 3y 00 4y = 0
Problem 30.7
Given that y1 (t) = e2t is a solution to the homogeneous equation, find the
general solution to the differential equation
y 000 2y 00 + y 0 2y = 0
Problem 30.8
Suppose that y(t) = c1 cos t + c2 sin t + c3 cos (2t) + c4 sin (2t) is the general
solution to the equation
y (4) + a3 y 000 + a2 y 00 + a1 y 0 + a0 y = 0
Find the constants a0 , a1 , a2 , and a3 .
Problem 30.9
Suppose that y(t) = c1 + c2 t + c3 cos 3t + c4 sin 3t is the general solution to
the homogeneous equation
y (4) + a3 y 000 + a2 y 00 + a1 y 0 + a0 y = 0
Determine the values of a0 , a1 , a2 , and a3 .
Problem 30.10
Suppose that y(t) = c1 et sin t+c2 et cos t+c3 et sin t+c4 et cos t is the general
solution to the homogeneous equation
y (4) + a3 y 000 + a2 y 00 + a1 y 0 + a0 y = 0
Determine the values of a0 , a1 , a2 , and a3 .
45
Problem 30.11
Consider the homogeneous equation with constant coefficients
y (n) + an1 y (n1) + + a1 y 0 + a0 = 0
Suppose that y1 (t) = t, y2 (t) = et , y3 (t) = cos t are several functions belonging to a fundamental set of solutions to this equation. What is the smallest
value for n for which the given functions can belong to such a fundamental
set? What is the fundamemtal set?
Problem 30.12
y (n) + an1 y (n1) + + a1 y 0 + a0 = 0
Suppose that y1 (t) = t2 sin t, y2 (t) = et sin t are several functions belonging
to a fundamental set of solutions to this equation. What is the smallest value
for n for which the given functions can belong to such a fundamental set?
What is the fundamemtal set?
Problem 30.13
y (n) + an1 y (n1) + + a1 y 0 + a0 = 0
Suppose that y1 (t) = t2 , y2 (t) = e2t are several functions belonging to a
fundamental set of solutions to this equation. What is the smallest value for
n for which the given functions can belong to such a fundamental set? What
is the fundamemtal set?
46
31
Non Homogeneous nth Order Linear Differential Equations
We consider again the nth order linear nonhomogeneous equation

(11)
where the functions p0 , p1 , , pn1 and g(t) are continuous functions for
a < t < b.
The solution structure established for second order linear nonhomogeneous
equations applies as well in the nth order case.
Theorem 31.1
Let {y1 (t), y2 (t), , yn (t)} be a fundamental set of solutions to the homogeneous equation
y (n) + pn1 (t)y (n1) + + p1 (t)y 0 + p0 (t)y = 0
and yp (t) be a particular solution of the nonhomogeneous equation
y (n) + pn1 (t)y (n1) + + p1 (t)y 0 + p0 (t)y = g(t).
The general solution of the nonhomogeneous equation is given by
y(t) = yp (t) + c1 y1 (t) + c2 y2 (t) + + cn yn (t)
Proof.
Let y(t) be any solution to equation (11). Since yp (t) is also a solution then
(y yp )(n) + pn1 (t)(y yp )(n1) + + p1 (t)(y yp )0 + p0 (t)(y yp ) =
y (n) +pn1 (t)y (n1) + +p1 (t)y 0 +p0 (t)y(yp(n) +pn1 (t)yp(n1) + +p1 (t)yp0 +p0 (t)yp ) =
g(t) g(t) = 0
Therefore yyp is a solution to the homogeneous equation. But {y1 , y2 , , yn }

is a fundamental set of solutions to the homogeneous equation so that there
exist unique constants c1 , c2 , , cn such that y(t)yp (t) = c1 y1 (t)+c2 y2 (t)+
+ cn yn (t). Hence,
47
Since the sum c1 y1 (t) + c2 y2 (t) + + cn yn (t) represents the general solution
to the homogeneous equation then we will denote it by yh so that the general
solution of (11) takes the form
y(t) = yh (t) + yp (t)
It follows from the above theorem that finding the general solution to nonhomogeneous equations consists of three steps:
1. Find the general solution yh of the associated homogeneous equation.
2. Find a single solution yp of the original equation.
3. Add together the solutions found in steps 1 and 2.
The superposition of solutions is valid only for homogeneous equations and
not true in general for nonhomogeneous equations. (Recall the case n = 2
in Section 22). However, we can have a property of superposition of nonhomogeneous if one is adding two solutions of two different nonhomogeneous
equations. More precisely, we have
Theorem 31.2
Let y1 (t) be a solution of y (n) + pn1 (t)y (n1) + + p1 (t)y 0 + p0 (t)y = g1 (t)
and y2 (t) a solution of y (n) + pn1 (t)y (n1) + + p1 (t)y 0 + p0 (t)y = g2 (t).
Then for any constants c1 and c2 the function Y (t) = c1 y1 (t) + c2 y2 (t) is a
solution of the equation
y (n) + pn1 (t)y (n1) + + p1 (t)y 0 + p0 (t)y = c1 g1 (t) + c2 g2 (t)
Proof.
We have
(n)
(n1)
L[Y ] = c1 (y1 + pn1 (t)y1

+ + p1 (t)y10 + p0 (t)y1 )
(n)
(n1)
+ c2 (y2 + pn1 (t)y2
+ + p1 (t)y20 + p0 (t)y2 )
=
c1 g1 (t) + c2 g2 (t)
Problem 31.1
Consider the nonhomogeneous differential equation
t3 y 000 + at2 y 00 + bty 0 + cy = g(t), t > 0
Determine a, b, c, and g(t) if the general solution is given by y(t) = c1 t +
c2 t2 + c3 t4 + 2 ln t
48
Problem 31.2
y 000 + ay 00 + by 0 + cy = g(t), t > 0
Determine a, b, c, and g(t) if the general solution is given by y(t) = c1 + c2 t +
c3 e2t + 4 sin 2t
Next, we discuss methods for determining yp (t). The techinque we discuss
first is known as the method of undetermined coefficients.
This method requires that we make an initial assumption about the form of
the particular solution yp (t), but with the coefficients left unspecified, thus
the name of the method. We then substitute the assumed expression into
equation (11) and attempt to determine the coefficients as to satisfy that
equation.
The main advantage of this method is that it is straightforward to execute
once the assumption is made as to the form of yp (t). Its major limitation is
that it is useful only for equations with constant coefficients and the nonhomogeneous term g(t) is restricted to a very small class of functions, namely
functions of the form et Pn (t) cos t or et Pn (t) sin t where Pn (t) is a polynomial of degree n.
In the following table we list examples of g(t) along with the corresponding form of the particular solution.
Form of g(t)
Pn (t) = an tn + an1 tn1 + + a0
Pn (t)et
Pn (t)et cos t or Pn (t)et sin t
Form of yp (t)
tr [An tn + An1 tn1 + + A1 t + A0 ]
tr [An tn + An1 tn1 + + A1 t + A0 ]et
tr et [(An tn + An1 tn1 + + A1 t + A0 ) cos t
+(Bn tn + Bn1 tn1 + + B1 t + B0 ) sin t]
The number r is chosen to be the smallest nonnegative integer such that

no term in the assumed form is a solution of the homogeneous equation
y (n) +pn1 (t)y (n1) + +p1 (t)y 0 +p0 (t)y = 0. The value of r will be 0 r n.
Equivalently, for the three cases, r is the number of times 0 is a root of the
characterstic equation, is a root of the characteristic equation, and + i
is a root of the characteristic equation, respectively.
Example 31.1
Solve
y 000 + y 00 = cos (2t), y(0) = 1, y 0 (0) = 2, y 00 (0) = 3
49
Solution.
We first solve the homogeneous differential equation
y 000 + y 00 = 0
r3 + r2 = 0
Factoring gives
r2 (r + 1) = 0
Solving we find r = 0 (repeated twice) and r = 1. The homogeneous
solution is
yh (t) = c1 + c2 t + c3 et
The trial function generated by g(t) = cos (2t) is
yp (t) = A cos (2t) + B sin (2t)
Then
yp0 = 2A cos (2t) + 2B sin (2t)

yp00 = 4A sin (2t) 4B cos (2t)
yp000 = 8A cos (2t) + 8B sin (2t)
Plugging back into the original differential equation gives

[8A cos (2t) + 8B sin (2t)] + [4A sin (2t) 4B cos (2t)] = cos (2t)
Combining like terms gives
(8A 4B) cos (2t) + (8B 4A) sin (2t) = cos (2t)
Equating coefficients gives
8A 4B = 1
4A + 8B = 0
Solving we find A = 0.1 and B = 0.05. The general solution is thus
y(t) = c1 + c2 t + c3 et 0.1 cos (2t) 0.05 sin (2t)
50
Now take derivatives to get

y 0 = c2 c3 et + 0.2 sin (2t) 0.1 cos (2t)
y 00 =
c3 et + 0.4 cos (2t) + 0.2 sin (2t)
Plug in the initial values to get
c1 + c3 = 1.1
c2 c3 = 2.1
c3
= 2.6
Solving we find c1 = 3.6, c2 = 4.7, c3 = 2.6. The final solution is
y(t) = 3.6 + 4.7t + 2.6et 0.1 cos (2t) 0.05 sin (2t)
Problem 31.3
Solve
y (4) + 4y 00 = 16 + 15et
Problem 31.4
Solve: y (4) 8y 00 + 16y = 64e2t
Problem 31.5
Given that y1 (t) = e2t is a solution to the homogeneous equation, find the
general solution to the differential equation,
y 000 2y 00 + y 0 2y = 12 sin 2t
Finally, we discuss a second method for finding a particular solution to a
nonhomogeneous differential equation known as the method of variation
of parameters. This method has no prior conditions to be satisfied by
either pn1 (t), , p1 (t), p0 (t), or g(t). Therefore, it may sound more general
than the method of undetermined coefficients.
The basic assumption underlying the method is that we know a fundamental
set of solutions {y1 , y2 , , yn }. The homogeneous solution is then
yh (t) = c1 y1 + c2 y2 + + cn yn
Then the constants c1 , c2 , , cn are being replaced by functions u1 , u2 , , un
so that the particular solution assumes the form
yp (t) = u1 y1 + u2 y2 + + un yn
51
(12)
We find u1 , u2 , , un by solving a system of n equations with the n unknowns

u1 , u2 , , un . We obtain the system by first imposing the n 1 constraints
y1 u01 + y2 u02 + + yn u0n = 0
y10 u01 + y20 u02 + + yn0 u0n = 0
..
.
(n2) 0
u1
y1
(n2) 0
u2
+ y2
(13)
+ + yn(n2) u0n = 0
This choice of constraints is made to make the successive derivatives of yp (t)

have the following simple forms
yp0
yp00
..
.
(n1)
yp
y10 u1 + y20 u2 + + yn0 un

y100 u1 + y200 u2 + + yn00 un
=
=
(n1)
= y1
(n1)
u1 + y 2
(n1)
u2 + + y n
un
Substituting (12) into (11), using (13) and the fact that each of the functions
y1 , y2 , , yn is a solution of the homogeneous equation we find
(n1) 0
u1
y1
(n1) 0
u2
+ y2
+ + yn(n1) u0n = g(t)
(14)
Take together, equations (13) and (14) form a set of n linear equations for
the n unknowns u01 , u02 , , u0n . In matrix form that system takes the form

y1
y2
yn
0
u01
0
0
0
y10
y2
yn u2 0
.. = ..
..
..
.
.
.
.
(n1)
y1
(n1)
y2
(n1)
yn
u0n
Solving this system we find

Wi
g
W
where 1 i n, W is the Wronskian of {y1 , y2 , , yn } and Wi is the
determinant obtained after replacing the ith column of W with the column
vector

0
0

..
.
1
u0i =
52
It follows that
Z
Z
Z
W1 (t)
W2 (t)
Wn (t)
yp (t) = y1
g(t)dt + y2
g(t)dt + + yn
g(t)dt
W (t)
W (t)
W (t)
Example 31.2
Solve
y 000 + y 0 = sec t
Solution.
We first find the homogeneous solution. The characteristic equation is
r3 + r = 0 or r(r2 + 1) = 0
so that the roots are r = 0, r = i, r = i.
We conclude
yh (t) = c1 + c2 cos t + c3 sin t
We have
yp (t) = u1 + u2 cos t + u3 sin t
and the Wronskian is

1 cos t
sin t

W (t) = 0 sin t cos t
0 cos t sin t
So

W1 (t) =

W2 (t) =

W3 (t) =

=1

0 cos t
sin t
0 sin t cos t = 1
1 cos t sin t

1 0 sin t
0 0 cos t = cos t
0 1 sin t

1 cos t 0
0 sin t 0 = sin t
0 cos t 1
Hence,
u1 (t) =
u2 (t) =
u3 (t) =
R
W1 (t)
g(t)dt
=
sec tdt = ln | sec t + tan t|
W (t)
R W2 (t)
R
g(t)dt = dt = t
W (t)
R W3 (t)
R sin t
g(t)dt
=
cos t dt = ln | cos t|
W (t)
53
Hence, the general solution is

y(t) = c1 + c2 cos t + c3 sin t + ln | sec t + tan t| t cos t + ln | cos t| sin t
Problem 31.6
Find the general solution of the equation
y 000 6y 00 + 12y 0 8y =
2te2t
Problem 31.7
(a) Verify that {t, t2 , t4 } is a fundamental set of solutions of the differential
equation
t3 y 000 4t2 y 00 + 8ty 0 8y = 0
(b) Find the general solution of
t3 y 000 4t2 y 00 + 8ty 0 8y = 2 t, t > 0

Problem 31.8
(a) Verify that {t, t2 , t3 } is a fundamental set of solutions of the differential
equation
t3 y 000 3t2 y 00 + 6ty 0 6y = 0
(b) Find the general solution of by using the method of variation of parameters
t3 y 000 3t2 y 00 + 6ty 0 6y = t, t > 0
Problem 31.9
Solve using the method of undetermined coefficients: y 000 y 0 = 4 + 2 cos t
Problem 31.10
Solve using the method of undetermined coefficients: y 000 y 0 = 4et
Problem 31.11
Solve using the method of undetermined coefficients: y 000 y 00 = 4e2t
Problem 31.12
Solve using the method of undetermined coefficients: y 000 3y 00 +3y 0 y = 12et .
Problem 31.13
Solve using the method of undetermined coefficients: y 000 + y = et + cos t.
54
In Problems 31.14 and 31.15, answer the following two questions.

(a) Find the homogeneous general solution.
(b) Formulate an appropriate for for the particular solution suggested by the
method of undetermined coefficients. You need not evaluate the undetermined coefficients.
Problem 31.14
y 000 3y 00 + 3y 0 y = et + 4et cos 3t + 4
Problem 31.15
y (4) + 8y 00 + 16y = t cos 2t
y 000 + ay 00 + by 0 + cy = g(t)
In Problems 31.16 - 31.17, the general solution of the differential equation is
given, where c1 , c2 , and c3 represent arbitrary constants. Use this information
to determine the constants a, b, c and the function g(t).
Problem 31.16
y(t) = c1 + c2 t + c3 e2t + 4 sin 2t.
Problem 31.17
y(t) = c1 + c2 t + c3 t2 2t3
Problem 31.18
t3 y 000 + at2 y 00 + bty 0 + cy = g(t), t > 0
Suppose that y(t) = c1 t + c2 t2 + c3 t4 + 2 ln t is the general solution to the
above equation. Determine the constants a, b, c and the function g(t)
55
32
Existence and Uniqueness of Solution to

Initial Value First Order Linear Systems
In this section we study the following initial-value problem

y10 = p11 (t)y1 + p12 (t)y2 + + p1n (t)yn + g1 (t)
y20 = p21 (t)y1 + p22 (t)y2 + + p2n (t)yn + g2 (t)
..
.
yn0 = pn1 (t)y1 + pn2 (t)y2 + + pnn (t)yn + gn (t)
y1 (t0 ) = y10 , y1 (t0 ) = y20 , , yn (t0 ) = yn0 , a < t0 < b
where all the pij (t) and gi (t) functions are continuous in a < t < b. The
above system can be recast in matrix form as
y0 (t) = P(t)y(t) + g(t), y(t0 ) = y0
where
y(t) =
y1 (t)
y2 (t)
..
.
, g(t) =
yn (t)
and
g1 (t)
g2 (t)
..
.
(15)
, y0 =
gn (t)
P(t) =
y10
y20
..
.
yn0
p11 (t) p12 (t) p1n (t)

p21 (t) p22 (t) p2n (t)
..
.
pn1 (t) pn2 (t) pnn (t)
We refer to differential equation in (15) as a first order linear system. If

g(t) is the zero vector in a < t < b then we call
y0 (t) = P(t)y(t)
a first order homogeneous linear system. Otherwise, we call the system
a first order nonhomogengeous linear system.
Next we discuss the conditions required for (15) to have a unique solution.
In order to establish the next theorem we state an important result from
analysis.
56
Theorem 32.1 (Weierstrass M-Test)

Assume {yN (t)}
N =1 is a sequence of functions defined in an open interval
a < t < b. Suppose that {MN }
N =1 is a sequence of positive constants such
that
|yN (t)| MN
P
P
for all a < t < b. If N =1 MN is convergent then
N =1 yN converges uniformly for all a < t < b.
Theorem 32.2
If the components of the matrices P(t) and g(t) are continuous in an interval
a < t < b then the initial value problem (15) has a unique solution on the
entire interval a < t < b.
Proof.
We start by reformulating the matrix differential equation in (15) as an integral equation. Integration of both sides of (15) yields
Z t
Z t
0
y (s)ds =
[P(s)y(s) + g(s)]ds
(16)
t0
t0
Applying the Fundamental Theorem of Calculus to the left side of (16) yields
Z t
y(t) = y0 +
[P(s)y(s) + g(s)]ds
(17)
t0
Thus, a solution of (17) is also a solution to (15) and vice versa.

Existence: To prove the existence we shall use again the method of successive approximations as described in Theorem 8.1.
y0 (t)
y1 (t) =
y0
y0 + t0 [P(s)y0 (s) + g(s)]ds
y2 (t) =
..
.
y0 +
Rt
yN (t) = y0 +
Write
Rt
t0
Rt
t0
[P(s)y1 (s) + g(s)]ds

..
.
[P(s)yN 1 (s) + g(s)]ds
yN (t) =
y1,N
y2,N
..
.
yn,N
57
For i = 1, 2, , n, we are going to show that the sequence {yi,N (t)} converges
uniformly to a function yi (t) such that y(t) (with components y1 , y2 , , yn )is
a solution to (17) and hence a solution to (15).
Let [c, d] be a closed interval containing t0 and contained in (a, b). Then by
continuity there exist positive constants kij , 1 i, j n, such that
max |pij (t)| kij
ctd
This implies that

P
P
P
||P(t)y(t)|| =
| nj=1 p1j yj | + | nj=1 p2j yj | + + | nj=1 pnj yj |
P
P
P
K 0 nj=1 |yj | + K 0 nj=1 |yj | + + K 0 nj=1 |yj | = K||y||
for all c t d, where we define
||y|| = |y1 | + |y2 | + + |yn |
and where
0
K =
n X
n
X
kij , K = nK 0 .
i=1 j=1
It follows that for 1 i n

Rt
|yi,N yi,N 1 | ||yN yN 1 || = || t0 [P(s)(yN 1 yN 2 )ds||
Rt
||P(s)(yN 1 yN 2 )||ds
t0 R
t
K t0 ||yN 1 yN 2 ||ds
But
||y1 y0 ||
Rt
t0
||P(s)y0 + g(s)||ds
M (t t0 )
where
M = K||y0 || + max |g1 (t)| + max |g2 (t)| + + max |gn (t)|.
ctd
ctd
ctd
An easy induction yields that for 1 i n

|yi,N +1 yi,N | ||yN +1 yN || M K N
58
(t t0 )N +1
(b a)N +1
M KN
(N + 1)!
(N + 1)!
Since
(b a)N +1
M K(ba)
=
(e
1)
(N
+
1)!
K
N =0
P
then by the Weierstrass M-test we conclude that the series
N =0 [yi,N yi,N 1 ]
converges uniformly for all c t d. But
M KN
yi,N (t) =
N
1
X
[yi,k+1 (t) yi,k (t)] + yi,0 .
k=0
Thus, the sequence {yi,N } converges uniformly to a function yi (t) for all
c t d.
The function yi (t) is a continuous function (a uniform limit of a sequence of
continuous functions is continuous). Also, we can interchange the order of
taking limits and integration for such sequences. Therefore
y(t) =
limN yN (t)
Rt
= y0 + limN t0 (P(s)yN 1 + g(s))ds
Rt
= y0 + t0 limN (P(s)yN 1 + g(s))ds
Rt
=
y0 + t0 (P(s)y + g(s))ds
This shows that y(t) is a solution to the integral equation (17) and therefore
a solution to (15).
Uniqueness:
The uniqueness follows from Gronwall Inequality (See Problem 8.11). Suppose that y(t) and z(t) are two solutions to the initial value problem, it
follows that for all a < t < b we have
Z t
||y(t) z(t)||
K||y(s) z(s)||ds
t0
Letting u(t) = ||y(t) z(t)|| we have

Z t
u(t)
Ku(s)ds
t0
so that by Gronwalls inequality u(t) 0 and therefore y(t) = z(t) for all
a < t < b. This completes a proof of the theorem
59
Example 32.1
y10 = t1 y1 + (tan t)y2 , y1 (3) = 0
y20 = (ln |t|)y1 + et y2 , y2 (3) = 1
Determine the largest t-interval such that a unique solution is guaranteed to
exist.
Solution.
The function p11 (t) = 1t is continuous for all t 6= 0. The function p12 (t) = tan t
is continuous for all t 6= (2n + 1) 2 where n is an integer. The function
p21 (t) = ln |t| is continuous for all t 6= 0. The function p22 (t) = et is continuous
for all real numbers. All these functions can be continuous on the common
domain t 6= 0 and t 6= (2n + 1) 2 . Since t0 = 3 then the largest t-interval for
which a unique solution is guaranteed to exist is 2 < t < 3
2
Problem 32.1
y10 = y1 + (tan t)y2 + (t + 1)2 , y1 (0) = 0
y20 =
(t2 2)y1 + 4y2 ,
y2 (0) = 0
exist.
Problem 32.2
(t + 2)y10 = 3ty1 + 5y2 , y1 (1) = 0
(t 2)y20 = 2y1 + 4ty2 , y2 (1) = 2
exist.
Problem 32.3
Verify that the functions y1 (t) = c1 et cos t + c2 et sin t and y2 (t) = c1 et sin t +
c2 et cos t are solutions to the linear system
y10 = y1 + y2
y20 = y1 + y2
60
Problem 32.4
Consider the first order linear system
y10 = y1 + y2
y20 = y1 + y2
(a) Rewrite the system in matrix form y0 (t) = Ay(t) and identify the matrix
A.
(b) Rewrite the solution to this system in the form y(t) = c1 y1 (t) + c2 y2 (t).
Problem 32.5
y0 (t) = Ay(t), y(0) = y0
where

3 2
1
A=
, y0 =
4 1
8

1
1
(a) Verify that y(t) = c1 e5t
+ c2 et
is a solution to the first
1
2
order linear system.
(b) Determine c1 and c2 such that y(t) solves the given initial value problem.
Problem 32.6
Rewrite the differential equation (cos t)y 00 3ty 0 + ty = t2 + 1 in the matrix

form y(t) = P(t)y(t) + g(t).
Problem 32.7
Rewrite the differential equation 2y 00 + ty + e3t = y 000 + (cos t)y 0 in the matrix
form y(t) = P(t)y(t) + g(t).
Problem 32.8
The initial value problem

0 1
0
1
0
y (t) =
y+
, y(1) =
3 2
2 cos (2t)
4
was obtained from an initial value problem for a higher order differential
equation. What is the corresponding scalar initial value problem?
61
Problem 32.9
The initial value problem
y2
y3
, y(1) =
y0 (t) =
y4
2
y2 + y3 sin y1 + y3
0
0
1
2
was obtained from an initial value problem for a higher order differential
equation. What is the corresponding scalar initial value problem?
Problem 32.10
Consider the system of differential equations
y 00 = tz 0 + y 0 + z
z 00 = y 0 + z 0 + 2ty
Write the above system in the form
y0 = P(t)y + g(t)
where
y(t)
y 0 (t)
y(t) =
z(t)
z 0 (t)
Identify P(t) and g(t).

Problem 32.11
Consider the system of differential equations
y 00 = 7y 0 + 4y 8z + 6z 0 + t2
z 00 = 5z 0 + 2z 6y 0 + 3y sin t
Write the above system in the form
y0 = P(t)y + g(t)
where
y(t)
y 0 (t)
y(t) =
z(t)
z 0 (t)
Identify P(t) and g(t).

62
33
Homogeneous First Order Linear Systems
In this section we consider the following system of n homogeneous linear differential equations known as the first order homogeneous linear system.
y10 = p11 (t)y1 + p12 (t)y2 + + p1n (t)yn
y20 = p21 (t)y1 + p22 (t)y2 + + p2n (t)yn
..
.
yn0 = pn1 (t)y1 + pn2 (t)y2 + + pnn (t)yn
where the coefficient functions are all continuous in a < t < b. The above
system can be recast in matrix form as
y0 (t) = P(t)y(t)
(18)
where
y(t) =
y1 (t)
y2 (t)
..
.
, P(t) =
yn (t)
p11 (t) p12 (t) p1n (t)

p21 (t) p22 (t) p2n (t)
..
.
pn1 (t) pn2 (t) pnn (t)
Example 33.1
(a) Rewrite the given system of linear homogeneous differential equations as
a homogeneous linear system of the form y0 (t) = P(t)y.
y2 + y3
y10 =
0
y2 = 6y1 3y2 + y3
y30 = 8y1 2y2 + 4y3
(b) Verify that the vector function
et
y(t) = et
2et
is a solution of y0 (t) = P(t)y.
Solution.
(a)

y1
y2
0

0
1 1
y
1
= 6 3 1
y2
8 2 4
63
(b) We have
et
y0 = et
2et
and
t t
0
1 1
e
e
t
e
P(t)y = 6 3 1
= et = y0
t
8 2 4
2e
2et
Our first result shows that any linear combinations of solutions to (18) is
again a solution.
Theorem 33.1
If y1 , y2 , , yr are solutions to (18) then for any constants c1 , c2 , , cr , the
function y = c1 y1 + c2 y2 + + cr yr is also a solution.
Proof.
Differentiating we find
y0 (t) =
(c1 y1 + c2 y2 + + cr yr )0
=
c1 y10 + c2 y20 + + cr yr0
=
c1 P(t)y1 + c2 P(t)y2 + + cr P(t)yr
= P(t)(c1 y1 + c2 y2 + + cr yr ) = P(t)y
In Problems 33.1 - 33.3 answer the following two questions.
(a) Rewrite the given system of linear homogeneous differential equations as
a homogeneous linear system of the form y0 (t) = P(t)y.
(b) Verify that the given function y(t) is a solution of y0 (t) = P(t)y.
Problem 33.1
y10 = 3y1 2y2
y20 = 4y1 + 3y2
and

y(t) =
et + et
2et et
Problem 33.2
y10 =
y20 =
2
y
t2 1
64
y2
2t y2
and

y(t) =
t2 + 3t
2t + 3
Problem 33.3
y10 = 2y1 + y2 + y3
y20 = y1 + y2 + 2y3
y30 = y1 + 2y2 + y3
and
2et + e4t
y(t) = et + e4t
et + e4t
Next, we pose the following question: Are there solutions {y1 , y2 , , yn }

such that every solution to (18) can be written as a linear combination of
y1 , y2 , , yn . We call such a set of functions a fundamental set of solutions. With such a set, the general solution is
y = c1 y1 + c2 y2 + + cn yn
Our next question is to find a criterion for testing n solutions to (18) for a
fundamental set. For this purpose, writing the components of the vectors
y1 , y2 , , yn
y1,1 (t)
y1,2 (t)
y1,n (t)
y2,1 (t)
y2,2 (t)
y2,n (t)
y1 (t) =
,
y
(t)
=
,
,
y
(t)
=
,
..
..
..
2
n
.
.
.
yn,1
yn,2
yn,n
we define the matrix (t) whose columns are the vectors y1 , y2 , , yn . That
is,
y1,1 y1,2 y1,n

y2,1 y2,2 y2,n
(t) = ..
yn,1 yn,2 yn,n
65
We call (t) a solution matrix of y0 = P(t)y. In this case, (t) is a

solution to the matrix equation 0 (t) = P(t)(t). Indeed,
0 (t) =
[y10 (t) y20 (t) yn0 (t)]
= [P(t)y1 (t) P(t)y2 (t) P(t)yn (t)]
=
P(t)[y1 (t) y2 (t) yn (t)]
=
P(t)(t)
We define the Wronskian of y1 , y2 , , yn to be the determinant of ; that
is
W (t) = det((t)).
The following theorem provides a condition for the solution vectors y1 , y2 , , yn
to form a fundamental set of solutions.
Theorem 33.2
Let {y1 , y2 , , yn } be a set of n solutions to (18). If there is a < t0 < b
such that W (t0 ) 6= 0 then the set {y1 , y2 , , yn } forms a fundamental set
of solutions.
Solution.
Let u(t) be any solution to (18). Can we find constants c1 , c2 , , cn such
that
u(t) = c1 y1 + c2 y2 + + cn yn ?
A simple matrix algebra we see that
c1 y1 + c2 y2 + + cn yn = (t) =
=
where
c1
c2
..
.
c=
y1,1 y1,2 y1,n

y2,1 y2,2 y2,n
..
.
yn,1 yn,2 yn,n
(t)c
cn
Thus,
u(t) = (t)c, a < t < b.
66
c1
c2
..
.
cn
In particular,
u(t0 ) = (t0 )c.
Since W (t0 ) = det((t0 )) 6= 0 then the matrix (t0 ) is invertible and as a
result of this we find
c = 1 (t0 )u(t0 )
When the columns of t) form a fundamental set of solutions of y0 (t) =
P(t)y(t) then we call (t) a fundamental matrix.
Example 33.2
(a) Verify the given functions are solutions of the homogeneous linear system.
(b) Compute the Wronskian of the solution set. On the basis of this calculation can you assert that the set of solutions forms a fundamental set?
(c) If the given solutions are shown in part(b) to form a fundamental set,
state the general solution of the linear homogeneous system. Express the
general solution as the product y(t) = (t)c, where (t) is a square matrix
whose columns are the solutions forming the fundamental set and c is a column vector of arbitrary constants.
(d) If the solutions are shown in part (b) to form a fundamental set, impose
the given initial condition and find the unique solution of the initial value
problem.
t
21 10 2
3
5et
e
11 2 y, y(0) = 10 , y1 (t) = 11et , y2 (t) = 0
y0 = 22
110 50 11
16
0
11et
et
y3 (t) = et
5et
Solution.
(a) We have
5et
y10 = 11et
0
and
21 10 2
5et
5et
22
11 2 11et = 11et = y10
110 50 11
0
0
67
Similarly,
et
y20 = 0
11et
and
and
t t
21 10 2
e
e
22
11 2
0
0 = y20
=
t
110 50 11
11e
11et
et
y30 = et
5et
t
et
21 10 2
e
22
11 2 et = et = y30
5et
5et
110 50 11
(b) The Wronskian is given by

5et
et
et

W (t) = 11et 0 et
0
11et 5et

= 11et

Since W (t) 6= 0 then the set {y1 , y2 , y3 } forms a fundamental set of solutions.
(c) The general solution is
c1
5et
et
et
y(t) = c2 y1 + c2 y2 + c3 y3 = 11et 0 et c2
c3
0
11et 5et
(d) We have
5
1 1
c1
3
11 0 1 c2 = 10
0 11 5
c3
16
Solving this system using Cramers rule we find c1 = 1, c2 = 1, c3 = 1.
Therefore the solution to the initial value problem is
t t
5et
e
e
4et et
y(t) = 11et 0 et = 11et + et
0
11et
5et
11et 5et
68
In Problems 33.4 - 33.7

(a) Verify the given functions are solutions of the homogeneous linear system.
(b) Compute the Wronskian of the solution set. On the basis of this calculation can you assert that the set of solutions forms a fundamental set?
(c) If the given solutions are shown in part(b) to form a fundamental set,
state the general solution of the linear homogeneous system. Express the
general solution as the product y(t) = (t)c, where (t) is a square matrix
whose columns are the solutions forming the fundamental set and c is a column vector of arbitrary constants.
(d) If the solutions are shown in part (b) to form a fundamental set, impose
the given initial condition and find the unique solution of the initial value
problem.
Problem 33.4

3t

9 4
0
2e3t 4et
4e + 2et
0
y =
y, y(0) =
, y1 (t) =
, y2 (t) =
1
3e3t 10et
6e3t + 5et
15 7
Problem 33.5

3 5
5
5e2t cos 3t
0
y =
y, y(0) =
, y1 (t) =
,
2 1
2
e2t (cos 3t 3 sin 3t)

y2 (t) =
5e2t sin 3t
2t
e (3 cos 3t + sin 3t)
Problem 33.6

1 1
2
1
e3t
0
y =
y, y(1) =
, y1 (t) =
, y2 (t) =
2 2
4
1
2e3t
Problem 33.7
2t
2 0 0
3
e
0
1 4 y, y(0) = 4 , y1 (t) = 0 , y2 (t) = 2et cos 2t
y0 = 0
0 1 1
2
0
et sin 2t
0
y3 (t) = 2et sin 2t
et cos 2t
69
The final result of this section is Abels theorem which states that the Wronskian of a set of solutions either vanishes nowhere or it vanishes everywhere
on the interval a < t < b.
Theorem 33.3 (Abels)
Let {y1 (t), y2 , , yn (t)(t)} be a set of solutions to (18) and let W(t) be the
Wronskian of these solutions. Then W(t) satisfies the differential equation
W 0 (t) = tr(P(t))W (t)
where
tr(P(t)) = p11 (t) + p22 (t) + + pnn (t).
Moreover, if a < t0 < b then
Rt
W (t) = W (t0 )e
t0
tr(P(s))ds
Proof.
Since {y1 , y2 , , yn } is a set of n solutions to (18) then we have
0
yi,j
n
X
pik yk,j , 1 i, j n
(19)
k=1
Using the definition of determinant we can write

X
W (t) =
sgn()y1,(1) y2,(2) yn,(n)
where the sum is taken over all one-to-one functions from the set {1, 2, , n}
to itself. Taking the derivative of both sides and using the product rule we
find
P
0
W 0 (t) =
(
2,(2) yn,(n) )
sgn()y1,(1) yP
P
0
0
=
y
+ sgn()y1,(1) y2,(2)
yn,(n)
sgn()y1,(1) y2,(2)
P n,(n)
0
+ + sgn()y1,(1) y2,(2) yn,(n)

0
0
y0
1,1 y1,2 y1,n
y2,1 y2,2 y2,n

= ..
.

yn,1 yn,2 yn,n

y1,1 y1,2 y1,n

0
0
y0
2,1 y2,2 y2,n
+ ..
.

yn,1 yn,2 yn,n
70

y1,1 y1,2 y1,n

y2,1 y2,2 y2,n

+ + ..

.

0
0
0

yn,1 yn,2
yn,n
But

0
0
y0
1,1 y1,2 y1,n
y2,1 y2,2 y2,n

..
.

yn,1 yn,2 yn,n
Pn
Pn
Pn

k=1 p1k yk,1
k=1 p1k yk,2
k=1 p1k yk,n

y
y
y2,n
2,1
2,2

=
..

.

yn,1
yn,2
yn,n
We evaluate the determinant of the right-side using elementary row operations (See Theorem 26.1). We multiply the second row by p12 , the third by
p13 , and so on, add these n 1 rows and then subtract the result from the
first row. The resulting determinant is

0
0

y0
1,1 y1,2 y1,n p11 y1,1 p11 y1,2 p11 y1,n
y2,1 y2,2 y2,n y2,1
y2,2 y2,n

=
= p11 W (t)
..
..

.
.

yn,1 yn,2 yn,n yn,1
yn,2 yn,n
Proceeding similarly with the other determinants we obtain
W 0 (t) = p11 W (t) + p22 W (t) + + pnn W (t)
=
(p11 + p22 + + pnn )W (t)
=
tr(P(t))W (t)
This is a first-order scalar equation for W(t), whose solution can be found
by the method of integrating factor
Rt
W (t) = W (t0 )e
t0
tr(P(s))ds
It follows that either W (t) = 0 for all a < t < b or W (t) 6= 0 for all a < t < b
Example 33.3
(a) Compute the Wronskian of the solution set and verify the set is a fundamental set of solutions.
(b) Compute the trace of the coefficient matrix.
(c) Verify
Abels theorem by showing that, for the given point t0 , W (t) =
Rt
tr(P(s))ds
W (t0 )e t0
.

4t
9
5
5e2t
e
0
y =
y, y1 (t) =
, y2 (t) =
, t0 = 0, < t <
7 3
7e2t
e4t
71
Solution.
(a) The Wronskian is

5e2t
e4t
W (t) =
7e2t e4t

= 2e6t

Since W (t) 6= 0 then the set {y1 , y2 } forms a fundamental set of solutions.
(b) tr(P(t)) = 9 3 = 6 R
Rt
t
tr(P(s))ds
(c) W (t) = 2e6t and W (t0 )e t0
= 2e 0 6ds = 2e6t
In Problems 33.8 - 33.9, the given functions are solutions of the homogeneous linear system.
(a) Compute the Wronskian of the solution set and verify the set is a fundamental set of solutions.
(b) Compute the trace of the coefficient matrix.
(c) Verify
Abels theorem by showing that, for the given point t0 , W (t) =
Rt
tr(P(s))ds
W (t0 )e t0
.
Problem 33.8

t
6
5
5et
e
0
y =
y, y1 (t) =
, y2 (t) =
, t0 = 1, < t <
t
7 6
7e
et
Problem 33.9

t
1
t
1
e
0
, t0 = 1, t 6= 0, 0 < t <
y =
y, y1 (t) =
, y2 (t) =
1
1
0
0 t
t
Problem 33.10
The functions

y1 (t) =
5
1

,
y2 (t) =
2e3t
e3t
are known to be solutions of the homogeneous linear system y0 = Py, where

P is a real 2 2 constant matrix.
(a) Verify the two solutions form a fundamental set of solutions.
(b) What is tr(P)?
(c) Show that (t) satisfies the homogeneous differential equation 0 = P,
where

5 2e3t
(t) = [y1 (t) y2 (t)] =
1 e3t
72
(d) Use the observation of part (c) to determine the matrix P.[Hint: Compute
the matrix product 0 (t)1 (t). It follows from part (a) that 1 (t) exists.]
Are the results of parts (b) and (d) consistent?
Problem 33.11
The homogeneous linear system
0
y =
3 1
2

y
has a fundamental set of solutions whose Wronskian is constant, W (t) =

4, < t < . What is the value of ?
73
34
First Order Linear Systems: Fundamental

Sets and Linear Independence
The results presented in this section are analogous to the ones established
for nth order linear homogeneous differential equations (See Section 5.3).
We start by showing that fundamental sets always exist.
Theorem 34.1
The first-order linear homogeneous equation
y0 = P(t)y,
a<t<b
where the entries of P are all continuous in a < t < b, has a fundamental set
of solutions.
Proof.
Pick a number t0 such that a < t0 < b. Consider the following n initial value
problems
y0 = P(t)y, y(t0 ) = e1
y0 = P(t)y, y(t0 ) = e2
..
.
y0 = P(t)y, y(t0 ) = en
where
e1 =
1
0
..
.
, e2 =
0
1
..
.
, , en =
0
0
..
.
By the existence and uniqueness theorem we find the solutions {y1 , y2 , , yn }.

Since W (t) = det([e1 , e2 , , en ]) = det(I) = 1 where I is the n n identity
matrix we see that the solution set {y1 , y2 , , yn } forms a fundamental set
of solutions
Next, we establish the converse to Theorem 33.2
Theorem 34.2
If {y1 , y2 , , yn } is a fundamental set of solutions to the first order linear
homogeneous system
y0 = P(t)y, a < t < b
then W (t) 6= 0 for all a < t < b.
74
Proof.
It suffices to show that W (t0 ) 6= 0 for some number a < t0 < b because by
Abels theorem this implies that W (t) neq0 for all a < t < b. The general
solution y(t) = c1 y1 + c2 y2 + + cn yn to y0 = P(t)y can be written as the
matrix equation
y(t) = c1 y1 + c2 y2 + + cn yn = (t)c, a < t < b
where (t) = [y1 y2 yn ] is the fundamental matrix and
c1
c2
c = ..
.
cn
In particular,
y(t0 ) = (t0 )c.
This matrix equation has a unique solution for c. This is possible only when
1 (t0 ) exists which is equivalent to saying that W (t0 ) = det((t0 )) 6= 0.
This completes a proof of the theorem
We next extend the definition of linear dependence and independence to
vector functions and show that a fundamental set of solutions is a linearly
independent set of vector functions on the t-interval of existence.
We say that a set of n 1 vector functions {f1 (t), f2 (t), , fr (t)}, where
a < t < b, is linearly dependent if one of the vector function can be written as a linear combination of the remaining functions. Equivalently, this
occurs if one can find constants k1 , k2 , , kr not all zero such that
k1 f1 (t) + k2 f2 (t) + + kr fr (t) = 0, a < t < b.
If the set {f1 (t), f2 (t), , fr (t)} is not linearly dependent then it is said to
be linearly independent in a < t < b. Equivalently, {f1 (t), f2 (t), , fr (t)}
is linearly independent if and only if
k1 f1 (t) + k2 f2 (t) + + kr fr (t) = 0
implies k1 = k2 = = 0.
75
Example 34.1
Determine whehter the given functions are linearly dependent or linearly
independent on the interval < t < .

1
0
f1 (t) = t , f2 (t) = 1
0
t2
Solution.
Suppose that k1 f1 (t) + k2 f2 (t) = 0 for all t. This implies that for all t we have
k1
= 0
k1 t + k2 = 0
k2 t2
= 0
Thus, k1 = k2 = 0 so that the set functions f1 (t) and f2 (t) are linearly independent
In Problems 34.1 - 34.4, determine whehter the given functions are linearly
dependent or linearly independent on the interval < t < .
Problem 34.1

f1 (t) =
t
1

, f2 (t) =
t2
1
Problem 34.2

f1 (t) =
et
1

, f2 (t) =
et
1

, f3 (t) =
et et
2
Problem 34.3

1
0
0
f1 (t) = t , f2 (t) = 1 , f3 (t) = 0

0
t2
0
Problem 34.4

1
0
1
2
2
f1 (t) = sin t , f2 (t) = 2(1 cos t) , f3 (t) = 0

0
2
1
76
Theorem 34.3
The solution set {y1 , y2 , , yn } is a fundamental set of solutions to
y0 = P(t)y
where the n n matrix P(t) is continuous in a < t < b, if and only if the
functions y1 , y2 , , yn are linearly independent.
Proof.
Suppose first that y1 , y2 , , yn is a fundamental set of solutions. Then by
Theorem 34.2 there is a < t0 < b such that W (t0 ) 6= 0. Suppose that
c1 y1 (t) + c2 y2 (t) + + cn yn (t) = 0
for all a < t < b. This can be written as the matrix equation
(t)c = 0, a < t < b
where
c1
c2
..
.
c=
cn
In particular,
(t0 )c = 0.
Since W (t0 ) = det((t0 )) 6= 0 then 1 (t0 ) exists so that c = 1 (t0 )(t0 )c =
1 (t0 ) 0 = 0. Hence, c1 = c2 = = cn = 0. Therefore, y1 , y2 , , yn are
linearly independent.
Conversely, suppose that {y1 , y2 , , yn } is a linearly independent set. Suppose that {y1 , y2 , , yn } is not a fundamental set of solutions. Then by Theorem 33.2, W (t) = det((t)) = 0 for all a < t < b. Choose any a < t0 < b.
Then W (t0 ) = 0. But this says that the matrix (t0 ) is not invertible. In
terms of matrix theory, this means that (t0 ) c = 0 for some vector
c1
c2
c = .. 6= 0
.
cn
77
Now, let y(t) = c1 y1 (t) + c2 y2 (t) + + cn yn (t) for all a < t < b. Then y(t)
is a solution to the differential equation and y(t0 ) = (t0 )c = 0. But the
zero function also is a solution to the initial value problem. By the existence
and uniqueness theorem we must have c1 y1 (t) + c2 y2 (t) + + cn yn (t) = 0
for all a < t < b with c1 , c2 , , cn not all equal to 0. But this means that
y1 , y2 , , yn are linearly depedent which contradicts our assumption that
y1 , y2 , , yn are linearly independent
Remark 34.1
The fact that y1 , y2 , , yn are solutions to y0 = Py is critical in the above
theorem. For example the vectors

2
1
t
t
t
f1 (t) = 0 , f2 (t) = 2 , f3 (t) =
0
0
0
are linearly independent with det(f1 , f2 , f3 ) 0.
Example 34.2
Consider the functions

f1 (t) =
et
0

, f2 (t) =
t2
t
(a) Let (t) = [f1 (t) f2 (t)]. Determine det((t)).

(b) Is it possible that the given functions form a fundamental set of solutions
for a linear system y0 = P(t)y where P(t) is continuous on a t-interval containing the point t = 0? Explain.
(c) Determine a matrix P(t) such that the given vector functions form a
fundamental set of solutions for y0 = P(t)y. On what t-interval(s) is the
coefficient matrix P(t) continuous?(Hint: The matrix (t) must satisfy
0 (t) = P(t)(t) and det((t)) 6= 0.)
Solution.
(a) We have
t 2
e t
= tet
F (t) =
0 t
(b) Since F (0) = 0 then the given functions do not form a fundamental set
for a linear system y0 = P(t)y on any t-interval containing 0.
78
(c) For (t) to be a fundamental matrix it must satisfy the differential equation 0 (t) = P(t)(t) and the condition det((t)) 6= 0. But det((t)) = tet
and this is not zero on any interval not containing zero. Thus, our coefficient
matrix P(t) must be continuous on either < t < 0 or 0 < t < . Now,
from the equation 0 (t) = P(t)(t) we can find P(t) = 0 (t)1 (t). That
is,
t

e 2t
t t2
1
0
1
P(t) = (t) (t) = tet
0 1
0 et
1
tet
tet (2t t2 )et

0
et

t 2t t2
0
1
2t
1
Problem 34.5
Consider the functions

f1 (t) =
t2
0

, f2 (t) =
(a) Let (t) = [f1 (t) f2 (t)]. Determine det((t)).

(b) Is it possible that the given functions form a fundamental set of solutions
for a linear system y0 = P(t)y where P(t) is continuous on a t-interval containing the point t = 0? Explain.
(c) Determine a matrix P(t) such that the given vector functions form a
fundamental set of solutions for y0 = P(t)y. On what t-interval(s) is the
coefficient matrix P(t) continuous?(Hint: The matrix (t) must satisfy
0 (t) = P(t)(t) and det((t)) 6= 0.)
Finally, we will show how to generate new fundamental sets from a given
one and therefore establishing the fact that a first order linear homogeneous
system has many fundamental sets of solutions. We also show how different
fundamental sets are related to each other. For this, let us start with a fundamental set {y1 , y2 , , yn } of solutions to y0 = P(t)y. If {y1 , y2 , , yn } are
n solutions then they can be written as linear combinations of the {y1 , y2 , , yn }.
79
That is,
a11 y1 + a21 y2 + + an1 yn = y1
a12 y1 + a22 y2 + + an2 yn = y2
..
.
a1n y1 + a2n y2 + + ann yn = yn
or in matrix form as
y1 y2 yn
y1 y2 yn
a11 a12 a13

a21 a22 a23
..
..
..
.
.
.
an1 an2 an3
a1n
a2n
..
.
ann
That is
(t) = (t)A
Theorem 34.4
{y1 , y2 , , yn } is a fundamental set if and only if det(A) 6= 0 where A is
the coefficient matrix of the above matrix equation.
Proof.
Since (t) = (t)A and W (t) = det((t)) 6= 0 then W (t) = det((t)) 6= 0
if and only if det(A) 6= 0. That is, {y1 , y2 , , yn } is a fundamental set of
solutions if and only if det(A) 6= 0
Example 34.3
Let

t

0 1
e
et
sinh t cosh t
0
, (t) =
y =
y, (t) =
et et
1 0
cosh t sinh t
(a) Verify that the matrix (t) is a fundamental matrix of the given linear
system.
(b) Determine a constant matrix A such that the given matrix (t) can be
represented as (t) = (t)A.
(c) Use your knowledge of the matrix A and assertion (b) of Theorem 34.4 to
determine whether (t) is also a fundamental matrix, or simply a solution
matrix.
80
Solution.
(a) Since

(t) =
and

P(t)(t) =
0 1
1 0

et et
et et
et et
et et

=
et et
et et
Thus, is a solution matrix. To show that (t) is a fundamental matrix

we need to verify that det((t)) 6= 0. Since det((t)) = 2 6= 0 then (t)
is a fundamental matrix.
(b) First write

1 et et et + et
sinh t cosh t
(t) =
=
cosh t sinh t
2 et + et et et
Thus, the question is to find a, b, c, and d such that

t
t

1 et et et + et
a b
ae + cet bet + det
e
et
=
=
aet cet bet det
et et
c d
2 et + et et et
Comparing entries we find a = 1/2, b = 1/2, c = 1/2, and d = 1/2.
(c) Since det(A) = 21 then (t) is a fundamental matrix
Problem
Let
1
0
y = 0
0
34.6
t
1 1
e + et 4e2t et + 4e2t
e
et
4e2t
1 1 y, (t) = 0 2et e2t , (t) = 2et e2t

e2t
2t
2t
2t
0
3e
3e
0 2
0
0
3e
(a) Verify that the matrix (t) is a fundamental matrix of the given linear
system.
(b) Determine a constant matrix A such that the given matrix (t) can be
represented as (t) = (t)A.
(c) Use your knowledge of the matrix A and assertion (b) of Theorem 34.4 to
determine whether (t) is also a fundamental matrix, or simply a solution
matrix.
81
Problem 34.7
Let
0
y =
1 1
0 2

y, (t) =
et
e2t
0 3e2t
where the matrix (t) is a fundamental matrix of the given homogeneous

linear system.

Find a constant matrix A such that (t) = (t)A with
1 0
(0) =
.
0 1
82
35
Homogeneous Systems with Constant Coefficients
In this section, we consider solving linear homogeneous systems of the form

y0 = Py where P is a matrix with real-valued constants. Recall that the
general solution to this system is given by
y(t) = c1 y1 (t) + c2 y2 (t) + + cn yn (t)
where {y1 (t), y2 (t), , yn (t)} is a fundamental set of solutions. So the problem of finding the general solution reduces to the one of finding a fundamental
set of solutions.
Lets go back and look at how we solved a second order linear homogeneous
equation with constant coefficients
y 00 + ay 0 + by = 0
(20)
To find the fundamental set of solutions we considered trial functions of the

form y = ert and find out that r is a solution to the characteristic equation
r2 + ar + b = 0. But (20) is a first order homogeneous linear system

0
0
1
y1
y1
=
(21)
b a
y2
y2
where y1 = y and y2 = y 0 .
Now, if r is a solution to the characteristic equation r2 + ar + b = 0 then one
can easily check that the vector function
rt

e
1
rt
y=
=e
rert
r
is a solution to (21).
Motivated by the above discussion we will consider trial functions for the
system
y0 = Py
(22)
of the form y = ert x where x is a nonzero vector. Substituting this into (22)
we find rert x = Pert x. This can be written as a linear system of the form
(P rI)x = 0
83
(23)
where I is the n n identity matrix.

Since system (23) has a nonzero solution x then the matrix P rI cannot
be invertible (otherwise x=0). This means that
p(r) = det(P rI) = 0.
(24)
We call (24) the characteristic equation associated to the linear system

(22). Its solutions are called eigenvalues. A vector x corresponding to an
eigenvalue r is called an eigenvector. The pair (r, x) is called an eigenpair.
It follows that each eigenpair (r, x) yields a solution of the form y(t) = ert x.
If there are n different eigenpairs then these will yield n different solutions.
We will show below that these n different solutions form a fundamental set
of solutions and therefore yield the general solution to (22). Thus, we need
to address the following questions:
(1) Given an n n matrix P, do there always exist eigenpairs? Is it possible
to find n different eigenpairs and thereby form n different solutions of (22)?
(2) How do we find these eigenpairs?
As pointed out earlier, the eigenvalues are solutions to equation (24). But

a11 r
a12
a1n

a21
a22 r
a2n

p(r) =
..
.. = 0

.
.

an1
an2
ann r
The determinant is the sum of elementary products each having n factors
no two come from the same row or column. Thus, one of the term has
the form (a11 r)(a22 r) (ann r). From this we see that p(r) is a
polynomial of degree n. We call p(r) the characteristic polynomial. By
the Fundamental Theorem of Algebra, the equation p(r) = 0 has n roots,
and therefore n eigenvalues. These eigenvalues may be zero or nonzero, real
or complex, and some of them may be repeated.
Now, for each eigenvalue r, we find a corresponding eigenvector by solving
the linear system of n equations in n unknowns: (P rI) = 0.
Example 35.1
Consider the homogeneous first order system

4 2
0
y =
y
1 1
84

1
2
(a) Show that x1 =
and x1 =
are eigenvectors of P. Deter1
1
mine the corresponding eigenvalues.
(b) For each eigenpair found in (a), form a solution yk of the system y0 = Py.
(c) Calculate the Wronskian and decide if the two solutions form a fundamental set.
Solution.
(a) Since

Px1 =
4 2
1 1

1
1

=
2
2

= 2x1
Thus, x1 is an eigenvector corresponding to the eigenvalue 2. Similarly,

4 2
2
6
Px2 =
=
= 3x2
1 1
1
3
Thus, x2 is an eigenvector corresponding
tothe eigenvalue 3.

1
2
2t
3t
(b) The two solutions are y1 (t) = e
and y2 (t) = e
.
1
1
(c) The Wronskian is

2t
e
2e3t
= e5t .
W (t) =
e2t
e3t
Since W (t) 6= 0 then the set {y1 , y2 } forms a fundamental set of solutions
In Problems 35.1 - 35.3, a 2 2 matrix P and vectors x1 and x2 are given.
(a) Decide which, if any, of the given vectors is an eigenvector of P, and
determine the corresponding eigenvalue.
(b) For the eigenpair found in part (a), form a solution yk (t), where k = 1
or k = 2, of the first order system y0 = Py.
(c) If two solution are found in part (b), do they form a fundamental set of
solutions for y0 = Py.
Problem 35.1

P=
7 3
16 7

, x1 =
85
3
8

, x2 =
1
2
Problem 35.2

P=
5 2
18 7
2 1
4 2
1
3
, x1 =
1
2
, x2 =
Problem 35.3

P=

, x1 =
1
2

, x2 =
1
2
Example 35.2

Find the eigenvalues of the matrix P =
8 0
5 2

.
Solution.

8r
0
The characteristic polynomial is p(r) =
3
2r
Thus, the eigenvalues are r = 8 and r = 2

= (8 r)(2 r).

Example 35.3

Suppose that r = 2 is an eigenvalue of the matrix P =
4 3
4 4
1
2
. Find the
eigenvector corresponding to this eigenvalue.

Solution.
We have (P 2I)x = 0 or

6 3
4 2

x1
x2

=
0
0
Solving this system we find 2x1 = x2 . Thus, an eigenvector is
In Problems 35.4 - 35.6, an eigenvalue is given of the matrix P. Determine a

corresponding eigenvector.
Problem 35.4

P=
5
3
4 3
86

, r = 1
Problem 35.5
1 7 3
P = 1 1 1 , r = 4
4 4 0
Problem 35.6
1 3 1
P = 2 1 2 , r = 5
4 3 2
In Problems 35.7 - 35.10, Find the eigenvalues of the matrix P.
Problem 35.7

P=
5 1
0 4
Problem 35.8

P=
3 3
6 6
Problem 35.9
5 0 0
P= 0 1 3
0 2 2
Problem 35.10
1 7 3
P = 1 1 1
4 4 0
We next list some properties of eigenvalues and eigenvectors.
Theorem 35.1
(a) If (r, x) is an eigenpai then for any 6= 0, (r, x) is also an eigenpair.
This shows that eigenvectors are not unique.
(b) A matrix P can have a zero eigenvalue.
(c) A real matrix may have one or more complex eigenvalues and eigenvectors.
87
Proof.
(a) Suppose that x is an eigenvector corresponding to an eigenvalue r of a
matrix P. Then for any nonzero constant we have P(x) = Px = r(x)
with x 6= 0. Hence, (r, x is an eigenpair.

0 0
(b) The characteristic equation of the matrix P =
is r(r 1) = 0
0 1
so that r = 0 is an eigenvalue.

1 1
(c) The characteristic equation of the matrix P =
is r2 2r+2 = 0.
1 1
Its roots are r = 1 + i and r = 1 i. For the r = 1 + i we have the system

i 1
x1
1 i
x2

1
A solution to this system is the vector x1 =
. Similarly, for r = 1 i
i
we have

i 1
x1
x2
1 i

1
A solution to this system is the vector x2 =
i
Theorem 35.2
Eigenvectors x1 , x2 , , xk corresponding to distinct eigenvalues r1 , r2 , , rk
are linearly independent.
Proof.
Let us prove this by induction on k. The result is clear for k = 1 because
eigenvectors are nonzero and a subset consisting of one nonzero vector is
linearly independent. Now assume that the result holds for k 1 eigenvectors. Let x1 , x2 , , xk be eigenvectors corresponding to distinct eigenvalues
r1 , r2 , , rk . Assume that there is a linear combination
c1 x1 + c2 x2 + + ck xk = 0.
88
Then we have
c1 x1 + c2 x2 + + ck xk = 0
P(c1 x1 + c2 x2 + + ck xk ) = 0
c1 Px1 + c2 Px2 + + ck Pxk = 0
c1 r1 x1 + c2 r2 x2 + + ck rk xk = 0
(c1 r1 x1 + c2 r2 x2 + + ck rk xk ) (c1 rk x1 + c2 rk x2 + + ck rk xk ) = 0
c1 (r1 rk )x1 + c2 (r2 rk )x2 + + ck1 (rk1 rk )xk1 = 0
=
=
=
=
=
But by the induction hypothesis, the vectors x1 , x2 , , xk1 are linearly independent so that c1 (r1 rk ) = c2 (r2 rk ) = = ck1 (rk1 rk ) = 0. Since
the eigenvalues are all distinct then we must have c1 = c2 = = ck1 = 0.
In this case we are left with ck xk = 0. Since xk 6= 0 then ck = 0. This shows
that {x1 , x2 , , xk } is linearly independent
The next theorem states that n linearly independent eigenvectors yield a
fundamental set of solutions to the equation y0 = Py.
Theorem 35.3
Consider the homogeneous system y0 = Py, < t < . Suppose that P
has eigenpairs (r1 , x1 ), (r2 , x2 ), , (rn , xn ) where x1 , x2 , , xn are linearly
independent. Then the set of solutions
{er1 t x1 , er2 t x2 , , ern t xn }
forms a fundamental set of solutions.
Proof.
We will show that the vectors er1 t x1 , er2 t x2 , , ern t xn are linearly independent. Suppose that
c1 er1 t x1 + c2 er2 t x2 + + cn ern t xn = 0
for all < t < . In particular, we can replace t by 0 and obtain
c1 x1 + c2 x2 + + cn xn = 0.
Since the vectors x1 , x2 , , xn are linearly independent then we must have
c1 = c2 = = cn = 0. This shows that er1 t x1 , er2 t x2 , , ern t xn are linearly
independent. Since each vector is also a solution then by Theorem 34.3 the
set {er1 t x1 , er2 t x2 , , ern t xn } forms a fundamental set of solutions
Combining Theorem 35.2 and Theorem 35.3 we obtain
89
Theorem 35.4
Consider the homogeneous system y0 = Py, < t < . Suppose that P
has n eigenpairs (r1 , x1 ), (r2 , x2 ), , (rn , xn ) with distinct eigenvalues. Then
the set of solutions
forms a fundamental set of solutions.
Proof.
Since the eigenvalues are distinct then by Theorem 35.2 the eigenvectors
x1 , x2 , , xn are linearly independent. But then by Theorem 35.3 the set of
solutions
forms a fundamental set of solutions
Example 35.4
Solve the following initial value problem

2 1
3
0
y =
y, y(0) =
1 2
1
Solution.

2 r
1

1
2 r

= (r + 1)(r + 3) = 0

Solving this quadratic equation we find r1 = 1 and r2 = 3. Now,

1 1
x1
x1 + x2
0
=
(P + I)x =
x1 x2
1 1
x2
0
Solving this system we find x1 = x2 . Letting x1 = 1 then x2 = 1. Thus, an
eigenvector is

1
x1 =
1
Similarly,

(P + 3I)x =
1 1
1 1

x1
x2
90

x1 + x2
x1 + x2

=
0
0
Solving this system we find x1 = x2 . Letting x1 = 1 then x2 = 1. Thus, an

eigenvector is

1
x2 =
1
By Theorem 35.4, a fundamental set of solutions is given by {et x1 , e3t x2 }.
The general solution is then
y(t) = c1 et x1 + c2 e3t x2 .
Using the initial condtion we find c1 + c2 = 3 and c1 c2 = 1. Solving this
system we find c1 = 2 and c2 = 1. Hence, the unique solution is given by
2et x1 + e3t x2
y(t) =

=
2et + e3t
2et e3t
In Problems 35.11 - 35.13, the matrix P has distinct eigenvalues. Using

Theorem 35.4 determine a fundamental set of solutions of the system y0 =
Py.
Problem 35.11

P=
0.09 0.02
0.04 0.07
Problem 35.12
1 2 0
P = 4 7 0
0 0 1
Problem 35.13
3
1 0
P = 8 6 2
9 9 4
91
Problem 35.14
Solve the following initial value problem.

5
3
2
0
y =
y, y(1) =
4 3
0
Problem 35.15
Solve the following initial value
4 2
0
y = 0 1
0 0
problem.
0
1
3 y, y(0) = 0
2
3
Problem 35.16
Find so that the vector x is an eigenvector of P. What is the corresponding
eigenvalue?

2
1
P=
, u=
1 5
1
Problem 35.17
Find and so that the vector x is an eigenvector of P corresponding the
eigenvalue r = 1.

1
P=
, u=
2
1
92
36
Homogeneous Systems with Constant Coefficients: Complex Eigenvalues
We continue the study of finding the general solution of y0 = Py where P is

an n n matrix with real entries. In this section, we consider the case when
P possesses complex eigenvalues. We start with the following result.
Theorem 36.1
If (r, x) is an eigenpair of P then (r, x) is an eigenpair of P. Thus, complex
eigenvalues always occur in conjugate pairs.
Proof.
Write r = + i. Then we have Px = ( + i)x. Take the conjugate of
both sides to obtain Px = ( i)x. But P is a real matrix so that P = P.
Thus, Px = ( i)x. This shows that i is an eigenvalue of P with
corresponding eigenvector x
In most applications, real-valued solutions are more meaningful then complex
valued solutions. Our next task is to describe how to convert the complex
solutions to y0 = Py into real-valued solutions.
Theorem 36.2
Let P be a real valued n n matrix. If P has complex conjugate eigenvalues
r1 = + i and r2 = i, where 6= 0, and corresponding (complex
conjugate) eigenvectors x1 = a + ib and x2 = a ib then y1 = et (a cos t
b sin t) and y2 = et (a sin t+b cos t) are two solutions of y0 = Py.(These
functions are the real and imaginary parts of the two solutions, e(+i)t x1 and
e(i)t x2 ).
Proof.
By Eulers formula we have
e(+i)t x1 =
et (cos t + i sin t)(a + ib)
= et (a cos t b sin t) + et i(a sin t + b cos t)
=
y1 + iy2
and
e(i)t x2 =
et (cos t i sin t)(a ib)
t
= e (a cos t b sin t) et i(a sin t + b cos t)
=
y1 iy2
93
We next show that y1 and y2 are solutions to y0 = Py. Indeed,

[e(+i)t x1 ]0 = y1 0 + iy2 0
and
Pe(+i)t x1 = Py1 + iPy2
Since Pe(+i)t x1 = [e(+i)t x1 ]0 then we must have Py1 = y10 and Py2 = y20
Example 36.1
Solve
0
y =
21 1
1 12
Solution.
1
r
1
2
1
1
2 r

y

= (r + 1 )2 + 1 = 0

2
Solving this quadratic equation we find r1 = 21 i and r2 = 21 + i. Now,

1
i 1
x1
ix1 + x2
0
(P + ( + i)I)x =
=
1 i
x2
x1 ix2
0
2
Solving this system we find x1 = ix2 . Letting x2 = i then x1 = 1. Thus, an
eigenvector is

1
x1 =
i
An eigenvector corresponding to the eigenvalue 12 + i is then

1
x2 =
i
The general solution is then

1
0
1
0
2t
2t
y(t) = c1 e
cos t
sin t + c2 e
sin t +
cos t
0
1
0
1
t

e 2 (c1 cos t + c2 sin t)
=
t
e 2 (c1 sin t + c2 cos t)
94
Problem 36.1
Find the eigenvalues and the eigenvectors of the matrix

0 9
P=
1 0
Problem 36.2
Find the eigenvalues and the eigenvectors of the matrix

3 1
P=
2 1
Problem 36.3
Find the eigenvalues and the eigenvectors
1 4
P= 3 2
1 1
of the matrix
1
3
3
In Problems 36.4 - 36.6, one or more eigenvalues and corresponding eigenvectors are given for a real matrix P. Determine a fundamental set of solutions
for y0 = Py, where the fundamental set consists entirely of real solutions.
Problem 36.4
P is a 2 2 matrix with an eigenvalue r = i and corresponding eigenvector

2 + i
x=
5
Problem 36.5
P is a 22 matrix with an eigenvalue r = 1+i and corresponding eigenvector

1 + i
x=
i
Problem 36.6
P is a 44 matrix with eigenvalues r = 1+5i with corresponding eigenvector

i
1
x=
0
0
95
and eigenvalue r = 1 + 2i with corresponding eigenvector

0
0
x=
i
1
Problem 36.7
Solve the initial value problem

0 9
6
0
y, y(0) =
y =
1 0
2
Problem 36.8

3 1
8
0
y =
y, y(0) =
2 1
6
Problem 36.9
1 4 1
1
3 y, y(0) = 9
y0 = 3 2
1 1
3
4
96
37
Homogeneous Systems with Constant Coefficients: Repeated Eigenvalues
In this section we consider the case when the characteristic equation possesses
repeated roots. A major difficulty with repeated eigenvalues is that in some
situations there is not enough linearly independent eigenvectors to form a
fundamental set of solutions. We illustrate this in the next example.
Example 37.1
Solve the system
0
y =
Solution.

1r
2

0
1r
1 2
0 1

y

= (r 1)2 = 0

and has a repeated root r = 1. We find an eigenvector as follows.

0 2
x1
2x2
0
=
=
0 0
x2
0
0
It follows that x2 = 0 and x1 is arbitrary. Letting x1 = 1 then an eigenvector
is

1
x1 =
0
This is the only eigenvector. It yields the solution
t
e
y1 =
0
But we need two linearly independent solutions to form the general solution
of the given system and we only have one. How do we find a second solution
y2 (t) such that {y1 , y2 } is a fundamental set of solutions?
Let y(t) be a solution. Write

y1 (t)
y(t) =
y2 (t)
97
Then we have
y10 (t) = y1 + 2y2

y2
y20 (t) =
Solving the second equation we find y2 (t) = c2 et . Substituting this into the
first differential equation we find y10 (t) = y1 + 2c2 et . Solving this equation
using the method of integrating factor we find y1 (t) = c1 et + c2 tet . Therefore
the general solution to y0 = Py is

1
0
1
c1 et + c2 tet
t
t
t
= c1 y1 (t)+c2 y2 (t).
+ te
+c2 e
= c1 e
y(t) =
0
1
0
c2 et
Thus, a second solution to y0 = Py is

0
1
t
t
y2 (t) = e
+ te
1
0
Finally, letting (t) = [y1 y2 ] we find

1 0
W (0) = det((0)) =
0 1

=1

so that {y1 , y2 } is a fundamental set of solutions

Example 37.2

13 11
1
0
y =
y, y(0) =
11 9
2
Solution.

13 r
11

11 9 r

= (r 2)2 = 0

and has a repeated root r = 2. We find an eigenvector as follows.

11
11
x1
11x1 + 11x2
0
=
=
11 11
x2
11x1 11x2
0
98
It follows that x2 = x1 . Letting x1 = 1 then an eigenvector is

1
x1 =
1
Therefore, one solution of y0 = Py is

y1 (t) =
e2t
e2t

.
The second solution has the form

y2 (t) = te2t x1 + e2t x2
where x2 is to be determined. Substituting y2 into the equation y0 = Py we
find
(1 + 2t)e2t x1 + 2e2t x2 = P(te2t x1 + e2t x2 ).
We can rewrite this equation as
te2t (Px1 2x1 ) + e2t (Px2 2x2 x1 ) = 0
But the set {e2t , te2t } is linearly independent so that
Px1 2x1 = 0
Px2 2x2 = x1
From the second equation we find

11
11
x1
11x1 + 11x2
1
=
=
11 11
x2
11x1 11x2
1
This shows that 11x1 + 11x2 = 1. Thus,

1 1 11x2
1 1
1
x2 =
=
x2
1
11x2
11
11 0
Letting x2 = 0 we find
1
x2 =
11
1
0
e2t
+
11
1
0
Hence,
2t
y2 (t) = te
1
1
99

=
2t
te2t + e11
te2t
Computing the Wronskian of the two solutions we find

1 1
11
= 1 6= 0

W (0) =
1 0
11
Therefore, the two solutions form a fundamental set of solutions and the
general solution is given by

2t
2t
c1
e
te2t + e11
y(t) =
c2
e2t
te2t
Imposing the initial condition,

1
1
y(0) =
=
1
2
1
11

c1
c2
Solving this system we find c1 = 2 and c2 = 33. Hence, the unique solution
to the initial value problem is
2t

e + 33te2t
y(t) =
2e2t 33te2t
Multiplicity of an Eigenvalue
As you have seen from the discussion above, when an eigenvalue is repeated
then one worries as to whether there exist enough linearly independent eigenvectors. These considerations lead to the following definitions.
Let P be an n n matrix and
det(P rI) = (r r1 )n1 (r r2 )n2 (r rk )nk .
The numbers n1 , n2 , , nk are called the algebraic multiplicities of the
eigenvalues r1 , r2 , , rk . For example, if det(P rI) = (r 2)3 (r 4)2 (r + 1)
then we say that 2 is an eigenvalue of P of multiplicity 3, 4 is of multiplicity
2, and 1 is of multiplicity 1.
We define the geometric multiplicity of an eigenvalue to be the number
of linearly independent eigenvectors corresponding to the eigenvalue.
Example 37.3
Find the algebraic and geometric multiplicities of the matrix
2 1 1 1
0 2 0 1
P=
0 0 2 1
0 0 0 3
100
Solution.
The characteristic equation is given by

2r

1
1
1

2r
0
1
0

2r
0
1

2r
1
= (2r) 0
0
0
2r
1

0
0
3r
0
0
0
3r

= (2r)3 (3r) = 0

Thus, r = 2 is an eigenvalue of algebraic multiplicity 3 and r = 3 is an

eigenvalue of algebraic multiplicity 1.
Next, we find eigenvector(s) associated to r = 2. We have

0 1 1 1
x1
0
0 0 0 1 x2 0

0 0 0 1 x3 = 0
0 0 0 1
x4
0
x1
x2
= x1
x3
x4
0
+ x2
0
0
0
1
1
0
Hence, the linearly independent eigenvectors are

1
0
0
1
x1 =
0 , x2 = 1
0
0
It follows that r = 2 has geometric multiplicity 2.

Similarly, we find an eigenvector associated to r = 3.
0
1 1
1 1
x1
0 1 0 1 x2 0
0
0 1 1 x3 0
0
0
0
0 0
x4
3
1
x3 =
1
1
101
It follows that r = 1 has geometric multiplicity 1

All the above examples discussed thus far suggest the following theorem.
For the proof, we remind the reader of the following definition: A set of
vectors {v1 , v2 , , vn } forms a basis of Rn if every vector in Rn is a linear combination of v1 , v2 , , vn and the vectors v1 , v2 , , vn are linearly
independent.
Theorem 37.1
Let A be an nn matrix with eigenvalue r1 . Then the geometric multiplicity
of r1 is less than or equal to the algebraic multiplicity of r1 .
Proof.
Let r1 be an eigenvalue of A with algebraic multiplicity a and geometric multiplicity g. Then we have g linearly independent eigenvectors B =
{v1 , v2 , , vg } with eigenvalues r1 . We next extend B to a basis
{v1 , v2 , , vg , wg+1 , , wn }
of Rn as follows. Let W be the set of all linear combinations of the vectors
of B. If g = n then W = Rn and we are done. If g < n then W is a
proper subset of Rn . Then we can find wg+1 that belongs to Rn but not
in W. Then wg+1 6 W and the set S1 = {v1 , v2 , , vg , wg+1 } is linearly
independent. If g + 1 = n then the set of all linear combinations of elements
of S1 is equal to Rn and we are done. If not, we can continue this extension
process. In n g steps we will get a set of n linearly independent vectors
{v1 , v2 , , vg , wg+1 , , wn } in Rn which will be a basis of Rn .
Now, let
P = [v1 v2 vg wg+1 wn ] = [P1 P2 ]
where P1 is the first g columns and P2 is the last n g columns. Since the
columns of P form a basis of Rn then
c1 v1 + c2 v2 + + cg vg + cg+1 wg+1 + + cn wn = 0
implies that c1 = c2 = = cn = 0 then P is an invertible matrix.
Next, write

B11 B12
1
P AP =
B21 B22
102
where B11 is g g matrix. Now, comparing

AP = [r1 v1 r1 v2 r1 vg Awg+1 Awn ] = [r1 P1 AP2 ]
with

P
B11 B12
B21 B22

= [P1 B11 + P2 B21 P1 B12 + P2 B22 ]
we get B11 = r1 Ig and B21 = 0. Thus,

r1 Ig B12
1
P AP =
0 B22
It follows that
det(A rIn ) = det(P1 (A rIn )P) = det(P1 AP rIn )
=
=

(r1 r)Ig
B12
det(
)
0
B22 r1 I(ng)(ng)
(r1 r)g det(B22 rI(ng)(ng) )
In particular, r1 appears as a root of the characteristic polynomial for at

least g times. Since the algebraic multiplicity a is the total number of times
r1 appears as a root, we conclude that a g
If ki is the geometric multiplicity of an eigenvalue ri of an n n matrix
P and ni is its algebraic multiplicity such that ki < ni then we say that the
eigenvalue ri is defective (its missing some of its eigenvalues) and we call
the matrix P a defective matrix. A matrix that is not defective is said to
have a full set of eigenvectors.
There are important family of square matrices that always have a full set of
eigenvectors, namely, real symmetric matrices and Hermitian matrices that
we discuss next.
The transpose of a matrix P, denoted by PT , is another matrix in which the
rows and columns have been reversed. That is, (PT )ij = (P)ji . For example,
the matrix
a11 a12 a13

P = a21 a22 a23
a31 a32 a33
103
would have the transpose
a11 a21 a31

PT = a12 a22 a32
a13 a23 a33
Theorem 37.2
(a) If P and Q are n n matrices then (P + Q)T = PT + QT .
(b) If P is an n m matrix and Q is an m p matrix then (PQ)T = QT PT .
Proof.
(a) We have [(P + Q)T ]ij = (P + Q)ji = (P)ji + (Q)ji = (PT )ij + (QT )ij .
(b) We have
Pm
((PQ)T )ij = (PQ)
Pjim= Tk=1 (P)Tjk (Q)ki
=
k=1 (Q )ik (P )kj
=
(QT PT )ij
An n n matrix P with real entries and with the property P = PT is called
a real symmetric matrix. For example, the matrix
1 2 3
P = 2 4 5
3 5 6
is a real symmetric matrix.
Real symmetric matrices are a special case of a larger class of matrices, known
T
as Hermitian matrices. An n n matrix P is called Hermitian if P = P ,
where P is the complex conjugate of P(The conjugate of a complex matrix
is the conjugate of all its entries.) For example,

3
2+i
P=
2i
1
T
is a Hermitian matrix. Note that PT = P when P is real matrix. Also note

that a real symmetric matrix is a Hermitian matrix.
Theorem 37.3
If P is a real symmetric matrix or Hermitian matrix then its eigenvalues are
all real.
104
Proof.
We prove the theorem for real Hermitian matrices. Suppose r is an eigenvalue
of P with corresponding eigenvector x. We will show that r is real. That
is, r = r. Since Px = rx then we can multiply both sides of this equation
from the left by xT to obtain xT Px = rxT x. On the other hand, we have
T
T
P x = r x. Thus, xT P x = r xT x. Since P = P then rxT x = r xT x. Since
xT x = ||x||2 6= 0 , where ||x||2 is the two norm of x,(x is an eigenvector) we
see that r = r, that is, r is real
The following theorem asserts that every Hermitian or real symmetric matrix
has a full set of eigenvectors. Therefore, when we study the homogeneous
linear first order system y0 = Py, where P is an n n a real symmetric
matrix we know that all solutions forming a fundamental set are of the form
ert x, where (r, x) is an eigenpair.
Theorem 37.4
If P is a Hermitian matrix (or a symmetric matrix) then for each eigenvalue,
the algebraic multiplicity equals the geometric multiplicity.
Proof.
We will prove the result for real symmetric matrices. In Section 39, we will
show that a real symmetric matrix has a set of n linearly independent eigenvectors. So if a1 , a2 , , ak are the algebraic multiplicities with corresponding geometric multiplicities g1 , g2 , , gk then we have a1 + a2 + + ak =
g1 + g2 + + gk = n. By Theorem 37.1, this happens only when ai = gi
In Problems 37.1 - 37.4, we consider the initial value problem y0 = Py, y(0) =
y0 .
(a) Compute the eigenvalues and the eigenvectors of P.
(b) Construct a fundamental set of solutions for the given differential equation. Use this fundamental set to construct a fundamental matrix (t).
(c) Impose the initial condition to obtain the unique solution to the initial
value problem.
Problem 37.1

P=
3 2
0 3

, y0 =
105
4
1
Problem 37.2

P=
3 0
1 3

, y0 =
2
3
0
2
4
4
Problem 37.3

P=
3 36
1
9

, y0 =
Problem 37.4

P=
6 1
1 4

, y0 =
Problem 37.5
Consider the homogeneous linear system
2 1 0
y0 = 0 2 1 y
0 0 2
(a) Write the three component differential equations of y0 = Py and solve
these equations sequentially, first finding y3 (t), then y2 (t), and then y1 (t).
(b) Rewrite the component solutions obtained in part (a) as a single matrix
equation of the form y = (t)c. Show that (t) is a fundamental matrix.
In Problems 37.6 - 37.8, Find the eigenvalues and eigenvectors of P. Give the
geometric and algebraic multiplicity of each eigenvalue. Does P have a full
set of eigenvectors?
Problem 37.6
5 0 0
P= 1 5 0
1 0 5
Problem 37.7
5 0 0
P= 0 5 0
0 0 5
106
Problem 37.8
2
0
P=
0
0
0
2
0
0
0
0
2
1
0
0
0
2
Problem 37.9
Let P be a 2 2 real matrix with an eigenvalue r1 = a + ib where b 6= 0. Can
P have a repeated eigenvalue? Can P be defective?
Problem 37.10
Dtermine the numbers x and y so that
symmetric.
0 1
P= y 2
6 2
the following matrix is real and
x
2
7
Problem 37.11
Dtermine the numbers x and y so that the following matrix is Hermitian.
2
x + 3i
7
5
2 + yi
P = 9 3i
7
2 + 5i
3
Problem 37.12
(a) Give an example of a 2 2 matrix P that is not invertible but have a full
set of eigenvectors.
(b) Give an example of a 2 2 matrix P that is invertible but does not have
a full set of eigenvectors.
107
38
Nonhomogeneous First Order Linear Systems
In this section, we seek the general solution to the nonhomogeneous first

order linear system
y0 = P(t)y + g(t)
(25)
where the components of the n n matrix P(t) and the n 1 vector g(t)
are continuous on a < t < b.
The solution structure is similar to one for nth order linear nonhomogeneous
equations and is the result of the following theorem.
Theorem 38.1
Let {y1 (t), y2 (t), , yn (t)} be a fundamental set of solutions to the homogeneous equation y0 = P(t)y and yp (t) be a particular solution of the
nonhomogeneous equation y0 = P(t)y + g(t). Then the general solution of
the nonhomogeneous equation is given by
Proof.
Let y(t) be any solution to equation (25). Since yp (t) is also a solution then
(y yp )0 =
y0 yp0
= P(t)y + g(t) [P(t)yp + g(t)]
=
g(t) g(t) = 0
Therefore yyp is a solution to the homogeneous equation. But {y1 , y2 , , yn }
is a fundamental set of solutions to the homogeneous equation so that there
exist unique constants c1 , c2 , , cn such that y(t)yp (t) = c1 y1 (t)+c2 y2 (t)+
+ cn yn (t). Hence,
Since the sum c1 y1 (t) + c2 y2 (t) + + cn yn (t) represents the general solution
to the homogeneous equation then we will denote it by yh so that the general
solution of (25) takes the form
y(t) = yh (t) + yp (t)
Superposition theorem for nth order linear nonhomogeneous equations holds
as well for linear systems.
108
Theorem 38.2
Let y1 (t) be a solution of y0 = P(t)y + g1 (t) and y2 (t) a solution of y0 =
P(t)y+g2 (t). Then for any constants c1 and c2 the function yp (t) = c1 y1 (t)+
c2 y2 (t) is a particular solution of the equation
y0 = P(t)y + c1 g1 (t) + c2 g2 (t),
Proof.
We have
a < t < b.
(c1 y1 (t) + c2 y2 (t))0

yp0 =
=
c1 y10 (t) + c2 y20 (t)
= c1 (P(t)y1 + g1 (t)) + c2 (P(t)y2 + g2 (t))
= P(t)(c1 y1 + c2 y2 ) + c1 g1 (t) + c2 g2 (t)
=
P(t)yp + c1 g1 (t) + c2 g2 (t)
Example 38.1
Consider the system
0
y =
1 2
2 1

y+
e2t
2t
(a) Find yh (t).

(b) Find yp (t).
(c) Find the general solution to the given system.
Solution.
(a) The characteristic equation is

1r
2

2
1r

= (r 1)2 4 = 0

Thus, the eigenvalues are r1 = 1 and r2 = 3. An eigenvector corresponding

to r1 = 1 is found as follows

2 2
x1
2x1 + 2x2
0
(P + I)x1 =
=
=
2 2
x2
2x1 + 2x2
0
Solving this system we find x2 = x1 . Letting x1 = 1 we find x2 = 1 and
an eigenvector is

1
x1 =
1
109
Similarly, for r2 = 3 we have

2 2
x1
2x1 + 2x2
0
(P 3I)x2 =
=
=
2 2
x2
2x1 2x2
0
Solving this system we find x2 = x1 . Letting x1 = 1 we find x2 = 1 and an
eigenvector is

1
x2 =
1
Hence,

1
1
t
3t
yh (t) = c1 e
+ c2 e
1
1
(b) To find yp (t) we note first that
2t

e
1
0
2t
=e
g(t) =
+t
= g1 (t) + g2 (t).
2t
0
2
By Superposition Theorem above, we will find a particular solution to y0 =
P(t)y + g1 (t) as well as to y0 = P(t)y + g2 (t). For the first equation, we use
the method of undetermined coefficients.
That is, we seek a solution of the

a
1
form up (t) = e2t a where a =
is a constant vector to be determined.
a2
Substituting up into the equation y0 = P(t)y + g1 (t) to obtain

1
2t
2t
2t
2e a = P(t)(e )a + e
0
This equation reduces to

a1
1
=
a2
0

1
. Hence,
Solving this system we find a = 31
2

1 2t 1
up (t) = e
.
2
3
1 2
2 1

Now, for the system y0 = P(t)y + g2 (t) we consider the guess function
vp (t) = tb + c where b and c are vectors whose components are to be
determined. Substituting this guess into the differential equation we find

1 2
0
b=
(tb + c) + t
2 1
2
110
and this reduces to

1 2
0
1 2
b+
+t
cb =0
2 1
2
2 1
Since the set {1, t} is linearly independent then the last equation implies the
two systems

1 2
b1
0
=
2 1
b2
2
and
1 2
2 1

c1
c2

=
b1
b2
Solving these systems we find

2
2 4
2
b=
, c=
5
3 1
9
Hence,
2
vp (t) = t
3
2
1
2
+
9
4
5
By the Superposition Theorem, we have

yp (t) = up (t) + vp (t).
(c) The general solution is given by

2t
2 4
1 2t 1
2
1
1
t
3t
+
+
y(t) = c1 e
+ c2 e
e
2
5
1
1
3
3 1
9
The Variation of Parameters Method
Next, we consider a method for finding a particular solution to (25) and the
unique solution to the initial-value problem
y0 = P(t) + g(t), y(t0 ) = y0
(26)
To solve the above initial-value problem, we start by looking at a fundamental

set {y1 , y2 , , yn , } of the homogeneous equation. Then we construct the
fundamental matrix (t) = [y1 y2 yn ]. Recall that (t) (See Section
33) satisfies the differential equation
0 = P(t), a < t < b.
111
Now, since
y(t) = c1 y1 + c2 y2 + + cn yn = (t)c
we vary the parameter and look for a solution to the initial-value problem (26) of the form y = (t)u(t), where u is an unknown vector to be
determined. Substituting this vector function into (26) to obtain
0 (t)u(t) + (t)u0 (t) = P(t)(t)u(t) + g(t).
Using the fact that 0 = P(t) the last equation reduces to
(t)u0 (t) = g(t)
Since (t) is a fundamental matrix then det((t)) 6= 0 and this implies that
the matrix (t) is invertible. Hence, we can write
u0 (t) = 1 (t)g(t).
Integrating both sides we find
Z
u(t) = u(t0 ) +
1 (s)g(s)ds
t0
where u(t0 ) is an arbitrary constant vector. It follows that the general solution to y0 = P(t)y + g(t) is given by
Z t
y(t) = (t)u(t) = (t)u(t0 ) + (t)
1 (s)g(s)ds = yh (t) + yp (t)
t0
Rt
where yh (t) = (t)u(t0 ) and yp (t) = (t) t0 1 (s)g(s)ds. Finally, since
y(t0 ) = (t0 )u(t0 ) then u(t0 ) = 1 (t0 )y(t0 ) and the unique solution to the
initial value problem is given by
Z t
1
y(t) = (t) (t0 )y(t0 ) + (t)
1 (s)g(s)ds.
t0
We refer to the last expression as the variation of parameters formula for

the solution of the initial-value problem.
112
Remark 38.1
y 0 = p(t)y + g(t), y(t0 ) = y0 , a < t < b.
Rt
Rt
p(s)ds
p(s)ds
Clearly, {e
} is a fundamental set. Letting (t) = e t0
in the
variation of parameters formula we see that the unique solution is given by
Z t R
Rt
Rt
t p(s)ds
p(s)ds
p(s)ds
e t0
g(s)ds
y(t) = y0 e t0
+ e t0
t0
t0
which is nothing than the method of integrating factor.

Example 38.2
0
y =
1 1
1 1

y+
e2t
0

, y(0) =
0
0
Solution.
0
We first find a fundamental matrix of the linear system y =

characteristic equation is

1r
1

1
1r
1 1
1 1

y. The

= r(r 2) = 0

and has eigenvalues r1 = 0 and r2 = 2. We find an eigenvector corresponding

to r1 = 0 as follows.

1 1
x1
x1 + x2
0
=
=
x1 + x2
0
1 1
x2
It follows that x1 = x2 . Letting x1 = 1 then x2 = 1 and an eigenvector is

1
x1 =
1
An eigenvector corresponding to r2 = 2

1 1
x1
x1 + x2
0
=
=
1 1
x2
x1 x2
0
113
Solving we find x1 = x2 . Letting x1 = 1 we find x2 = 1 and an eigenvector is

1
x2 =
1
Thus, a fundamental matrix is

=
1 e2t
1 e2t
Therefore,

= 0.5
1
e2t
1
e2t
But the variation of parameters formula is

Z
1 (s)g(s)ds.
y(t) = (t) (0)y(0) + (t)

0
Thus,

y(t) =

=
1 e2t
1 e2t

1 e2t
1 e2t
0.25
Rt
0

0.5

(0.25)
e2s
1

ds
1e2t 1
2t
e2t 1 + 2te2t
(e2t 1) + 2te2t
In Problems 38.1 - 38.3, we consider the initial value problem y0 = Py +

g(t), y(t0 ) = y0 .
(a) Find the eigenpairs of the matrix P and form the general homogeneous
solution of the differential equation.
(b) Construct a particular solution by assuming a solution of the form suggested and solving for the undetermined constant vectors a,b, and c.
(c) Form the general solution of the nonhomogeneous differential equation.
(d) Find the unique solution to the initial value problem.
Problem 38.1
0
y =
2 1
1 2

y+
Try yp (t) = a.
114
1
1

, y0 =
3
1
Problem 38.2

y =
0 1
1 0

y+
t
1

, y0 =
2
1
Try yp (t) = ta + b.
Problem 38.3
0
y =
3 2
4
3
y+
sin t
0

, y0 =
0
0
Try yp (t) = (sin t)a + (cos t)b.

Problem 38.4

0 2
0
y =
y + g(t), y
= y0 .
2 0
2
Suppose we know that

y(t) =
1 + sin 2t
et + cos 2t
is the unique solution. Determine g(t) and y0 .

Problem 38.5

1 t
2
0
y =
y + g(t), y(1) =
.
t2 1
1
Suppose we know that

y(t) =
t+
t2 +
is the unique solution. Determine g(t) and the constants and .

Problem 38.6
Let P(t) be a 2 2 matrix with continuous entries. Consider
the
differ1
ential equation y0 = P(t)y + g(t). Suppose that y1 (t) =
is the
et
115
2
0
et
1
solution to y = P(t)y +
and y2 (t) =
is the solution to
t
e
0
y = P(t)y +
. Determine P(t). Hint: Form the matrix equation
1
[y10 y20 ] = P[y1 y2 ] + [g1 g2 ].
Problem 38.7
Consider the linear system y0 = Py + b where P is a constant matrix and b
is a constant vector. An equilibrium solution, y(t), is a constant solution
of the differential equation.
(a) Show that y0 = Py + b has a unique equilibrium solution when P is
invertible.
(b) If the matrix P is not invertible, must the differential equation y0 =
Py + b possess an equilibrium solution? If an equilibrium solution does exist
in this case, is it unique?
Problem 38.8
Determine all the equilibrium solutions (if any).

2 1
2
0
y =
y+
1 1
1
Problem 38.9
Determine all the equilibrium solutions (if any).

1 1 0
2
0
y = 0 1 2 y + 3
0 0 1
2
Consider the homogeneous linear system y0 = Py. Recall that any associated fundamental matrix satisfies the matrix differential equation 0 = P.
In Problems 38.10 - 38.12, construct a fundamental matrix that solves the
matrix initial value problem 0 = P, (t0 ) = 0 .
Problem 38.10
0
1 1
1 1

, (1) =
116
1 0
0 1
Problem 38.11

1 1
1 1
1 0
2 1
1 0
0 1
, (0) =
Problem 38.12

1 4
1 1

,

4
In Problems 38.13 - 38.14, use the method of variation of parameters to solve

the given initial value problem.
Problem 38.13

y =
9 4
15 7

y+
et
0
2
5
0
0
, y(0) =
Problem 38.14
0
y =
1 1
0 1

y+
1
1
117

, y(0) =
39
Solving First Order Linear Systems with

Diagonalizable Constant Coefficients Matrix
In this section we discuss a method for solving the initial value problem
y0 = Py + g(t), y(t0 ) = y0 , a < t < b
where P is a nondefective constant matrix and the entries of g(t) are continuous in a < t < b.. This type of matrices is always diagonalizable, a concept
that we will introduce and discuss below.
Similar Matrices
An n n matrix A is said to be similar to an n n matrix B if there is an
invertible nn matrix T such that T1 AT = B. Note that if we let R = T1
then B = RAR1 so whether the inverse comes first or last does not matter.
Also, note that from this definition we can write A = (T1 )1 BT1 so that
the matrix B is similar to A. Thats why, in the literature one will just say
that A and B are similar matrices.
The first important result of this concept in the following theorem.
Theorem 39.1
If A and B are similar then they have the same characteristic equation and
therefore the same eigenvalues.
Proof.
Since A and B are similar then B = T1 AT for some invertible matrix T.
From this one notices the following
det(A rI) =
det(T1 (A rI)T) =
= det(T1 (AT rI) = det(B rI)
This shows that A and B have the same characteristic equation and therefore the same eigenvalues. We point out the following equality that we used
in the above discussion: det(T1 T) = det(T1 )det(T) = det(I) = 1
The second important result is the following.
118
Theorem 39.2
Suppose that W(t) is a solution to the system y0 = By and B and A are
similar matrices with B = T1 AT. Then y(t) = TW(t) is a solution to
y0 = Ay.
Proof.
Since W(t) is a solution to y0 = By then W0 = BW. But B = T1 AT so
we can write W0 = T1 ATW. Thus, TW0 = ATW. That is, (TW(t))0 =
A(TW). But this says that y(t) = TW(t) is a solution to y0 = Ay
Diagonalizable Matrices
An n n matrix A is said to be diagonalizable if there is an invertible
matrix T such that
d1 0 0 0
0 d2 0 0
0 0 d3 0
1
T AT = D =
..
..
.
.
0 0 0 dn
In other words, the matrix A is similar to a diagonal matrix.
Our first question regarding diagonalization is the question of whether every
square matrix is diagonalizable.
Example 39.1
Show that the matrix

A=
0 1
0 0
is not diagonalizable.
Solution.
If
A is diagonalizable then we expect to find an invertible matrix T =
a b
such that T1 AT is a diagonal matrix. But
c d

1
1
d b
0 1
a b
cd
d2
1
T AT =
0 0
c d ad bc c2 cd
ad bc c a
119
Now, if c = 0 then d 6= 0 and therefore the above product does not result
in a diagonal matrix. Similar argument applies if d = 0. Hence, the given
matrix is not diagonalizable
Note that the above matrix does not have a full set of eigenvectors. Indeed,
the characteristic equation of the matrix A is

r 1

0 r = 0
Expanding the determinant and simplifying we obtain
r2 = 0.
The only eigenvalue of A is r = 0. Now, an eigenvector is found as follows.

0 1
x1
x2
0
=
=
0 0
x2
0
0
Thus, we find that x2 = 0 and x1 is arbitrary. Hence, an eigenvector is

1
x=
0
Since the geometric multiplicity of r = 0 is less than its algebraic multiplicity
then A is defective.
So, is having a full set of eigenvectors results in the matrix to be diagonalizable? The answer to this question is provided by the following theorem.
Theorem 39.3
An n n matrix A is diagonalizable if and only if it has a set of n linearly
independent eigenvectors.
Proof.
(=) : Suppose first that A is diagonalizable. Then there are an ivnertible
matrix T and a diagonal matrix D such that T1 AT = D. By Theorem 39.1,
the diagonal entries of D are the eigenvalues of A. Now, let c1 , c2 , , cn be
the n columns of T so that T = [c1 c2 cn ]. Since T1 AT = D then
AT = TD. That is,
r1 0 0 0
0 r2 0 0
A[c1 c2 cn ] = [c1 c2 cn ] 0 0 r3 0
..
..
.
.
0 0 0 rn
120
where r1 , r2 , , rn are the eigenvalues of A. The above matrix equation is

the same as
[Ac1 Ac2 Acn ] = [r1 c1 r2 c2 rn cn ]
and this equality yields
Ac1 = r1 c1 , Ac2 = r2 c2 , , Acn = rn cn .
This shows that the eigenvectors of A are just the columns of T. Now, if
c1 c1 + c2 Ac2 + + cn cn = [c1 c2 cn ]c = 0
then the invertibility of T forces c = 0. This shows that the columns of T,
and therefore the eigenvectors of A, are linearly independent.
(=) : Now, suppose that A has n linearly independent eigenvectors c1 , c2 , , cn
with corresponding eigenvalues r1 , r2 , , rn . Let T = [c1 c2 cn ] and D
be the diagonal matrix with diagonal entries r1 , r2 , , rn . Then AT = TA.
Also, since the eigenvectors are linearly independent then T is invertible and
therefore T1 AT = D. This shows that A is diagonalizable
Remark 39.1
We have seen in Section 37 that real symmetric matrices and Hermitian
matrices have full set of eigenvectors. According to the previous theorem
these matrices are always diagonalizable.
Solution Method of First Order Linear Systems by Uncoupling
We finally describe a method based on matrix diagonalization for solving the
initial value problem
y0 = Py + g(t), y(t0 ) = y0 , a < t < b.
(27)
where the components of g(t) are continuous in a < t < b and the matrix P
is a diagonalizable constant matrix, that is, there is an invertible matrix T
such that
r1 0 0 0
0 r2 0 0
0 0 r3 0
1
T PT = D =
..
..
.
.
0 0 0 rn
121
where r1 , r2 , , rn are the eigenvalues of P. Thus, P = TDT1 . Substituting

into the differential equation we find
y0 = TDT1 y + g(t)
or
(T1 y)0 = D(T1 y) + T1 g(t).
Letting z(t) = T1 y then the previous equation reduces to
z0 (t) = Dz(t) + T1 g(t).
Letting
z(t) =
z1
z2
..
.
, T1 g(t) =
, z(t0 ) = T1 y0 =
hn
zn
We can write
h1
h2
..
.
z10
z20
..
.
zn0
r1 z1
r2 z2
..
.
rn zn
h1
h2
..
.
z10
z20
..
.
zn0
hn
Thus, for 1 i n we have the scalar initial value problem

zi0 = ri zi + hi , zi (t0 ) = zi0 .
Solving this equation using the method of integrating factor we find
Z t
ri (tt0 ) 0
zi (t) = e
zi +
eri (ts) hi (s)ds, 1 i n.
t0
Having found the vector z(t) we then find the solution to the original initial
value problem by forming the matrix product y(t) = Tz(t).
Example 39.2
Solve the following system by making the change of variables y = Tz.

1 1
t
0
y =
y+
2 2
t + 3
122
Solution.

1r
1

1
2r

= r(r 3) = 0

Thus, the eigenvalues are r1 = 0 and r2 = 3. An eigenvector corresponding

to r1 = 0 is found as follows

1 1
x1
x1 + x2
0
(P + 0I)x1 =
=
=
2 2
x2
2x1 + 2x2
0
Solving this system we find x1 = x2 . Letting x2 = 1 we find x1 = 1 and
an eigenvector is

1
x1 =
1
Similarly, for r2 = 3 we have

2 1
x1
2x1 + x2
0
(P 3I)x2 =
=
=
2 1
x2
2x1 x2
0
Solving this system we find 2x1 = x2 . Letting x1 = 1 we find x2 = 2 and an
eigenvector is

1
x2 =
2
Therefore
1 1
1 2
T=
Thus,
T
2 1
1 1
1
=
3
Letting y = Tz we obtain
0
z =
0 0
0 3
z+
t1
1
That is,
z10 = t 1
z20 = 3z2 + 1
123

1 2
t t + c1
2
31 + c2 e3t

z(t) =
Thus, the general solution is

1 2

1 2
1 e3t
c1
t t + c1
t t 13
1 1
2
2
=
y(t) = Tz(t) =
1 2e3t
c2
31 + c2 e3t
12 t2 + t 23
1 2
In Problems 39.1 - 39.4, the given matrix is diagonalizable. Find matrices T
and D such that T1 PT = D.
Problem 39.1

3
4
2 3
P=
Problem 39.2
2 3
2 3
1 2
2 1
P=
Problem 39.3
P=
Problem 39.4

P=
2 2
0 3
In Problems 39.5 - 39.6, you are given the characteristic polynomial for the
matrix P. Determine the geometric and algebraic multiplicities of each eigenvalue. If the matrix P is diagonalizable, find matrices T and D such that
T1 PT = D.
Problem 39.5
7 2 2
P = 8 1 4 , p(r) = (r 3)2 (r + 1).
8 4 1
124
Problem 39.6
5 1 1
P = 14 3 6 , p(r) = (r 2)2 (r 3).
5 2 5
Problem 39.7
At leat two (and possibly more) of the following four matrices are diagonalizable. You should be able to recognize two by inspection. Choose them and
give a reason for your choice.

5 6
3 6
3 0
1 3
(a)
, (b)
, (c)
, (d)
3 4
6 9
3 4
1 4
Problem 39.8

2t

4 6
e 2et
0
y =
y+
e2t + et
3
5
Problem 39.9

3 2
4t + 4
0
y =
y+
1 4
2t + 1
Problem 39.10
Solve the following system by making the change of variables x = Tz.

6
7
00
x =
x
15 16
Problem 39.11
Solve the following system by making the change of variables x = Tz.

4 2
00
x =
x
2 1
125
40
Solving First Order Linear Systems Using

Exponential Matrix
The matrix exponential plays an important role in solving systems of linear

differential equations. In this section, we will define such a concept and study
some of its important properties.
Recall from calculus the power series expansion of et given by
X
t2 t3
tn
t
t
e = 1 + + + + =
1! 2! 3!
n!
n=0
and this series converges for all real numberes t.
To develop something similar with number t replaced by a matrix A one
proceeds as follows: The absolute value used for measuring the distance
between numbers is now replaced by a matrix norm given by
v
uX
n
u n X
t
||A|| =
(aij )2
i=1 j=1
Next, we construct the sequence of partial sums

I+ A
1!
I+ A
+
1!
S1 =
S2 =
..
.
Sn = I +
A
1!
A
2!
+ +
A
n!
With little effort which we dont pursue here, it can be shown that the
sequence of partial sums converges and its limit is denoted by eA . That is,
lim Sn = eA
or
X An
A A2 A3
e =I+ +
+
+ =
1!
2!
3!
n!
n=0
A
and this series converges for any square matrix A.

Example 40.1
Suppose that

A(t) =
Find eA .
126
0 t
t 0
Solution.
One can easily see that for any nonnegative odd integer n we have

0 tn
n
A =
tn 0
and for nonnegative even integer n

A =
tn 0
0 tn
Thus,
" P 2n
t
eA = Pn=0 t(2n)!
2n+1
n=0 (2n+1)!
P
Pn=0
t2n+1
(2n+1)!
t2n
n=0 (2n)!

=
cosh t sinh t
sinh t cosh t
The following theorem describes some of the important properties of the

exponential matrix.
Theorem 40.1
(i) If AB = BA then eA+B = eA eB .
(ii) For any square matrix A, eA is invertible with (eA )1 = eA .
(iii) For any invertible n n matrix P and any n n matrix A
eP
1 AP
= P1 eA P
(Thus, if A is similar to B; then eA is similar to eB ).

(iv) If A has eigenvalues r1 , r2 , , rn (not necessarily distinct), then eA
has eigenvalues er1 , er2 , , ern .
(v) det(eA ) = etr(A) .
(vi) AeA = eA A.
When dealing with systems of differential equations, one has often to deal
with expressions like ePt , where is a matrix and t is a real number or real
variable. With the above formula of the exponential matrix function we get
ePt = I +
X Pn tn
Pt P2 t2
+
+ =
.
1!
2!
n!
n=0
127
Lets find the derivative of ePt . To prove this, we calculate this derivative via
the usual limit formula:
eP(t+h) ePt
d Pt
e = lim
.
h0
dt
h
But
eP(t+h) = ePt ePh .
since Pt and Ph commute.(These matrices are scalar multiples of the same
matrix P, and P commutes with itself.) Going back to the derivative, we get
d Pt
e
dt
eP(t+h) ePt
h
limh0
ePt limh0
ePh I
h

= ePt Ph + 2!1 (Ph)2 + 3!1 (Ph)3 +
ePt P = PePt
Now, consider the initial value problem

y0 = Py + g(t), y(t0 ) = y0 , a < t < b
where P is a constant square matrix and the entries of g(t) are continuous
in a < t < b. Then one has
y0 Py
0
eP(tt0 ) y
eP(tt0 ) y
y(t)
=
g(t)
P(tt0 )
=
e
g(t)
R t P(st )
0
=
y0 + t0 e
g(s)ds
R t (ts)P
(tt0 )P
= e
y0 + t0 e
g(s)ds
Remark 40.1
The above procedure does not apply if the matrix P is not constant! That
is, dtd eP(t) 6= P0 (t)eP(t) . This is due to the fact that matrix multiplication is
not commutative in general.
Example 40.2

6
9
Pt
Find e if P =
.
4 6
128
Solution.
Since P2 = 0 then
P(t)

= I + tP =
1 + 6t
9
4 1 6t
Problem 40.1

0 2
P(t)
Find e
if P =
.
2 0
Problem 40.2
Consider the linear differential system

y = Py, P =
1
4
1 3
(a) Calculate ePt . Hint: Every square matrix satisfies its characteristic equation.
(b) Use the result from part (a) to find two independent solutions of the
differential system. Form the general solution.
Problem 40.3
Show that if
then
d1 0 0
D = 0 d2 0
0 0 d3
ed1 0
0
eD = 0 ed2 0
0
0 ed3
Problem 40.4

3 0
0
y =
y, y(0) = y0
0 1
Problem 40.5
Show that if r is an eigenvalue of P then er is an eigenvalue of eP .
129
Problem 40.6
Show that det(eA ) = etr(A) . Hint: Recall that the determinant of a matrix
is equal to the product of its eigenvalues and the trace is the sume of the
eigenvalues. This follows from the expansion of the characteristic equation
into a polynomial.
Problem 40.7
Prove: For any invertible n n matrix P and any n n matrix A
eP
1 AP
= P1 eA P
(Thus, if A is similar to B; then eA is similar to eB ).

Problem 40.8
Prove: If AB = BA then eA+B = eA eB .
Problem 40.9
Prove: For any square matrix A, eA is invertible with (eA )1 = eA .
Problem 40.10
Consider the two matrices

1 0
0 1
A=
, B=
0 1
1 0
Show that AB 6= BA and eA+B 6= eA eA .
130
41
The Laplace Transform: Basic Definitions

and Results
Laplace transform is yet another operational tool for solving constant coefficients linear differential equations. The process of solution consists of three
main steps:
The given hard problem is transformed into a simple equation.
This simple equation is solved by purely algebraic manipulations.
The solution of the simple equation is transformed back to obtain the solution of the given problem.
In this way the Laplace transformation reduces the problem of solving a differential equation to an algebraic problem. The third step is made easier by
tables, whose role is similar to that of integral tables in integration.
The above procedure can be summarized by Figure 41.1
Figure 41.1
In this section we introduce the concept of Laplace transform and discuss
some of its properties.
The Laplace transform is defined in the following way. Let f (t) be defined
for t 0. Then the Laplace transform of f, which is denoted by L[f (t)]
or by F (s), is defined by the following equation
Z T
Z
st
L[f (t)] = F (s) = lim
f (t)e dt =
f (t)est dt
T
The integral which defined a Laplace transform is an improper integral. An

improper integral may converge or diverge, depending on the integrand.
When the improper integral in convergent then we say that the function f (t)
possesses a Laplace transform. So what types of functions possess Laplace
transforms, that is, what type of functions guarantees a convergent improper
integral.
Example 41.1
Find the Laplace transform, if it exists, of each of the following functions
(a) f (t) = eat
(b) f (t) = 1 (c) f (t) = t (d) f (t) = et

131
Solution.
(a) Using the definition of Laplace transform we see that
Z T
Z
(sa)t
at
e(sa)t dt.
e
dt = lim
L[e ] =
T
But
(sa)t

dt =
T
1e(sa)T
sa
if s = a
if s 6= a.
For the improper integral to converge we need s > a. In this case,

L[eat ] = F (s) =
1
, s > a.
sa
(b) In a similar way to what was done in part (a), we find

Z
Z T
1
st
L[1] =
e dt = lim
est dt = , s > 0.
T 0
s
0
(c) We have
Z
L[t] =
st
te
0

test est
1
dt =
2
= 2 , s > 0.
s
s 0
s
(d) Again using the definition of Laplace transform we find

Z
2
t2
L[e ] =
et st dt.
0
R 2
2
If
1 and this implies that 0 et st dt
R s 0 then t st 0 so that e
. Since the integral on the right is divergent then by the comparison the0
orem of improper integrals (see TheoremR 41.1 below) theR integral on the left
is also divergent. Now, if s > 0 then 0 et(ts) dt s dt. By the same

reasoning the integral on the left is divergent. This shows that the function
2
f (t) = et does not possess a Laplace transform
t2 st
The above example raises the question of what class or classes of functions
possess a Laplace transform.
closely at Example 41.1(a), we notice
R Looking
(sa)t
that for s > a the integral 0 e
dt is convergent and a critical component for this convergence is the type of the function f (t). To be more specific,
if f (t) is a continuous function such that
|f (t)| M eat ,
132
tC
(28)
where M 0 and a and C are constants, then this condition yields

Z
Z C
Z
st
st
e(sa)t dt.
f (t)e dt + M
f (t)e dt
0
Since f (t) is continuous in 0 t C then by letting A = max{|f (t)| : 0

t C} we have

Z C
Z C
1 esC
st
st
f (t)e dt A
e dt = A
< .
s
s
0
0
R
On the other hand, Now, by Example 41.1(a), the integral C e(sa)t dt is
convergent for s > a. By the comparison theorem of improper integrals (see
Theorem 41.1 below) the integral on the left is also convergent. That is, f (t)
possesses a Laplace transform.
We call a function that satisfies condition (28) a function with an exponential order at infinity. Graphically, this means that the graph of f (t) is
contained in the region bounded by the graphs of y = M eat and y = M eat
for t C. Note also that this type of functions controls the negative exponential in the transform integral so that to keep the integral from blowing
up. If C = 0 then we say that the function is exponentially bounded.
Example 41.2
Show that any bounded function f (t) for t 0 is exponentially bounded.
Solution.
Since f (t) is bounded for t 0 then there is a positive constant M such that
|f (t)| M for all t 0. But this is the same as (28) with a = 0 and C = 0.
Thus, f (t) has is exponentially bounded
Another question that comes to mind is whether it is possible to relax the
condition of continuity on the function f (t). Lets look at the following situation.
Example 41.3
Show that the square wave function whose graph is given in Figure 41.2
possesses a Laplace transform.
133
Figure 41.2
Note that the function is periodic of period 2.
Solution.
R
R
Since f (t)est est then 0 f (t)est dt 0 est dt. But the integral on
the right is convergent for s > 0 so that the integral on the left is convergent
as well. That is, L[f (t)] exists for s > 0
The function of the above example belongs to a class of functions that we
define next. A function is called piecewise continuous on an interval if
the interval can be broken into a finite number of subintervals on which the
function is continuous on each open subinterval (i.e. the subinterval without
its endpoints) and has a finite limit at the endpoints (jump discontinuities
and no vertical asymptotes) of each subinterval. Below is a sketch of a
piecewise continuous function.
Figure 41.3
Note that a piecewise continuous function is a function that has a finite
number of breaks in it and doesnt blow up to infinity anywhere. A function
defined for t 0 is said to be piecewise continuous on the infinite interval if it is piecewise continuous on 0 t T for all T > 0.
Example 41.4
Show that the following functions are piecewise continuous and of exponential
order at infinity for t 0
134
(a) f (t) = tn
(b) f (t) = tn sin at
Solution.
P
tn
tn
n
t
n
(a) Since et =
n=0 n! n! then t n!e . Hence, t is piecewise continuous
and exponentially bounded.
(b) Since |tn sin at| n!et then tn sin at is piecewise continuous and exponentially bounded
Next, we would like to establish the existence of the Laplace transform for
all functions that are piecewise continuous and have exponential order at
infinity. For that purpose we need the following comparison theorem from
calculus.
Theorem 41.1
Suppose that f (t) and g(t) are both
R integrable functions for allRt t0 such
that |f (t)| |g(t) for t t0 . If t0 g(t)dt is convergent, then t0 f (t)dt is
R
R
also convergent. If, on the other hand, t0 f (t)dt is divergent then t0 f (t)dt
is also divergent.
Theorem 41.2 (Existence)
Suppose that f (t) is piecewise continuous on t 0 and has an exponential
order at infinity with |f (t)| M eat for t C. Then the Laplace transform
Z
f (t)est dt
F (s) =
0
exists as long as s > a. Note that the two conditions above are sufficient, but
not necessary, for F (s) to exist.
Proof.
The integral in the definition of F (s) can be splitted into two integrals as
follows
Z
Z C
Z
st
st
f (t)e dt =
f (t)e dt +
f (t)est dt.
0
Since f (t) is piecewise continuous in 0 t C then it is bounded there. By

letting A = max{|f (t)| : 0 t C} we have

Z C
Z C
1 esC
st
st
< .
f (t)e dt A
e dt = A
s
s
0
0
135
R
Now, by Example 41.1(a), the integral C f (t)est dt is convergent for s > a.
By Theorem 41.1 the integral on the left is also convergent. That is, f (t)
possesses a Laplace transform
In what follows, we will denote the class of all piecewise continuous functions with exponential order at infinity by PE. The next theorem shows that
any linear combination of functions in PE is also in PE. The same is true for
the product of two functions in PE.
Theorem 41.3
Suppose that f (t) and g(t) are two elements of PE with
|f (t)| M1 ea1 t ,
t C1
and
|g(t)| M2 ea1 t ,
t C2 .
(i) For any constants and the function f (t) + g(t) is also a member of
PE. Moreover
L[f (t) + g(t)] = L[f (t)] + L[g(t)].
(ii) The function h(t) = f (t)g(t) is an element of PE.
Proof.
(i) It is easy to see that f (t) + g(t) is a piecewise continuous function.
Now, let C = C1 + C2 , a = max{a1 , a2 }, and M = ||M1 + ||M2 . Then for
t C we have
|f (t) + g(t)| |||f (t)| + |||g(t)| ||M1 ea1 t + ||M2 ea2 t M eat .
This shows that f (t) + g(t) is of exponential order at infinity. On the
other hand,
RT
L[f (t) + g(t)] =
limT 0 [f (t) + g(t)]dt
= limT
=
RT
0
f (t)dt + limT
RT
0
g(t)dt
L[f (t)] + L[g(t)]
(ii) It is clear that h(t) = f (t)g(t) is a piecewise continuous function. Now,

letting C = C1 + C2 , M = M1 M2 , and a = a1 + a2 then we see that for t C
we have
|h(t)| = |f (t)||g(t)| M1 M2 e(a1 +a2 )t = M eat .
136
Hence, h(t) is of exponential order at infinity. By Theorem 41.2 , L[h(t)]

exists for s > a
We next discuss the problem of how to determine the function f (t) if F (s)
is given. That is, how do we invert the transform. The following result on
uniqueness provides a possible answer. This result establishes a one-to-one
correspondence between the set PE and its Laplace transforms. Alternatively, the following theorem asserts that the Laplace transform of a member
in PE is unique.
Theorem 41.4
Let f (t) and g(t) be two elements in PE with Laplace transforms F (s) and
G(s) such that F (s) = G(s) for some s > a. Then f (t) = g(t) for all t 0
where both functions are continuous.
The standard techniques used to prove this theorem( i.e., complex analysis,
residue computations, and/or Fouriers integral inversion theorem) are generally beyond the scope of an introductory differential equations course. The
interested reader can find a proof in the book Operational Mathematics
by Ruel Vance Churchill or in D.V. Widder The Laplace Transform.
With the above theorem, we can now officially define the inverse Laplace
transform as follows: For a piecewise continuous function f of exponential
order at infinity whose Laplace transform is F, we call f the inverse Laplace
transform of F and write f = L1 [F (s)]. Symbolically
f (t) = L1 [F (s)] F (s) = L[f (t)].
Example 41.5

1
, s > 1.
Find L1 s1
Solution.
1
, s > a. In particular, for
From Example 41.1(a), we have that L[eat ] = sa

1
1
t
1
a = 1 we find that L[e ] = s1 , s > 1. Hence, L
= et , t 0 .
s1
The above theorem states that if f (t) is continuous and has a Laplace transform F (s), then there is no other function that has the same Laplace transform. To find L1 [F (s)], we can inspect tables of Laplace transforms of
known functions to find a particular f (t) that yields the given F (s).
When the function f (t) is not continuous, the uniqueness of the inverse
137
Laplace transform is not assured.

uniqueness issue.
The following example addresses the
Example 41.6
Consider the two functions f (t) = h(t)h(3 t) and g(t) = h(t) h(t 3).
(a) Are the two functions identical?
(b) Show that L[f (t)] = L[g(t).
Solution.
(a) We have

1, 0 t 3
0,
t>3
1, 0 t < 3
0,
t3
f (t) =
and
g(t) =
So the two functions are equal for all t 6= 3 and so they are not identical.
(b) We have
Z 3
1 e3s
, s > 0.
est dt =
L[f (t)] = L[g(t)] =
s
0
Thus, both functions f (t) and g(t) have the same Laplace transform even
though they are not identical. However, they are equal on the interval(s)
where they are both continuous
The inverse Laplace transform possesses a linear property as indicated in
the following result.
Theorem 41.5
Given two Laplace transforms F (s) and G(s) then
L1 [aF (s) + bG(s)] = aL1 [F (s)] + bL1 [G(s)]
for any constants a and b.
Proof.
Suppose that L[f (t)] = F (s) and L[g(t)] = G(s). Since L[af (t) + bg(t)] =
aL[f (t)]+bL[g(t)] = aF (s)+bG(s) then L1 [aF (s)+bG(s)] = af (t)+bg(t) =
aL1 [F (s)] + bL1 [G(s)]
138
Problem 41.1
R
Determine whether the integral 0
verges, give its value.
1
dt
1+t2
converges. If the integral con-
Problem 41.2
R
Determine whether the integral 0
verges, give its value.
t
dt
1+t2
converges. If the integral con-
Problem 41.3
R
Determine whether the integral 0 et cos (et )dt converges. If the integral
converges, give its value.
Problem 41.4
Using the definition, find L[e3t ], if it exists. If the Laplace transform exists
then find the domain of F (s).
Problem 41.5
Using the definition, find L[t 5], if it exists. If the Laplace transform exists
Problem 41.6
2
Using the definition, find L[e(t1) ], if it exists. If the Laplace transform
exists then find the domain of F (s).
Problem 41.7
Using the definition, find L[(t 2)2 ], if it exists. If the Laplace transform
exists then find the domain of F (s).
Problem 41.8
Using the definition, find L[f (t)], if it exists. If the Laplace transform exists

0,
0t<1
f (t) =
t 1,
t1
Problem 41.9
Using the definition, find L[f (t)], if it exists. If the Laplace transform exists
0t<1
0,
t 1, 1 t < 2
f (t) =
0,
t 2.
139
Problem 41.10
Let n be a positive integer. Using integration by parts establish the reduction
formula
Z
Z
tn est n
n st
+
t e dt =
tn1 est dt, s > 0.
s
s
Problem 41.11
For s > 0 and n a positive integer evaluate the limits
limt0 tn est
(b) limt tn est
Problem 41.12
(a) Use the previous two problems to derive the reduction formula for the
Laplace transform of f (t) = tn ,
L[tn ] =
n n1
L[t ], s > 0.
s
(b) Calculate L[tk ], for k = 1, 2, 3, 4, 5.

(c) Formulate a conjecture as to the Laplace transform of f (t), tn with n a
positive integer.
From a table of integrals,
R u
sin u
e sin udu = eu sin u
2 + 2
R u
sin u
e cos udu = eu cos u+
2 + 2
Problem 41.13
Use the above integrals to find the Laplace transform of f (t) = cos t, if it
exists. If the Laplace transform exists, give the domain of F (s).
Problem 41.14
Use the above integrals to find the Laplace transform of f (t) = sin t, if it
Problem 41.15
Use the above integrals to find the Laplace transform of f (t) = cos (t 2),
if it exists. If the Laplace transform exists, give the domain of F (s).
Problem 41.16
Use the above integrals to find the Laplace transform of f (t) = e3t sin t, if it
140
Problem 41.17
Use the linearity property of Laplace transform to find L[5e7t + t + 2e2t ].
Find the domain of F (s).
Problem 41.18
Consider the function f (t) = tan t.
(a) Is f (t) continuous on 0 t < , discontinuous but piecewise continuous on 0 t < , or neither?
(b) Are there fixed numbers a and M such that |f (t)| M eat for 0 t < ?
Problem 41.19
Consider the function f (t) = t2 et .
Problem 41.20
Consider the function f (t) =
et
e2t +1
Problem 41.21
Consider the floor function f (t) = btc, where for any integer n we have
btc = n for all n t < n + 1.
Problem 41.22

3
Find L1 s2
.
Problem 41.23
Find L1 s22 +
1
s+1
Problem 41.24

2
2
Find L1 s+2
+ s2
.
141
42
Further Studies of Laplace Transform
Properties of the Laplace transform enable us to find Laplace transforms

without having to compute them directly from the definition. In this section, we establish properties of Laplace transform that will be useful for
solving ODEs.
Laplace Transform of the Heaviside Step Function
The Heaviside step function is a piecewise continuous function defined by

1, t 0
h(t) =
0, t < 0
Figure 42.1 displays the graph of h(t).
Figure 42.1
Taking the Laplace transform of h(t) we find
st
Z
Z
e
1
st
st
h(t)e dt =
e dt =
= , s > 0.
L[h(t)] =
s 0
s
0
0
A Heaviside function at 0 is the shifted function h(t ) ( units to the
right). For this function, the Laplace transform is
st
Z
Z
e
es
st
st
L[h(t )] =
h(t )e dt =
e dt =
=
, s > 0.
s
s
0
Laplace Tranform of eat

The Laplace transform for the function f (t) = eat is
(sa)t
Z
e
1
at
(sa)t
L[e ] =
e
dt =
=
, s > a.
sa 0
sa
0
142
Laplace Tranforms of sin at and cos at

Using integration by parts twice we find
R st
L[sin at]
=
e sin atdt
0
=
st
e ssin at
aest cos at
s2
sa2
=

s2 +a2
s2
a2
s2
R
0
est sin atdt
a2
L[sin at]
s2
a
s2
L[sin at] =
L[sin at]
a
,
s2 +a2
s>0
A similar argument shows that

L[cos at] =
s2
s
, s > 0.
+ a2
Laplace Transforms of cosh at and sinh at

Using the linear property of L we can write
1
2
L[cosh at] =
=
1
2
(L[eat ] + L[eat ])
1
sa
1
s+a
s
,
s2 a2
, s > |a|
s > |a|
A similar argument shows that

L[sin at] =
s2
a
, s > |a|.
a2
Laplace Transform of a Polynomial

Let n be a positive integer. Using integration by parts we can write
n st
Z
Z
t e
n n1 st
n st
t e dt =
+
t e dt.
s
s 0
0
0
By repeated use of LHopitals rule we find limt tn est = limt
for s > 0. Thus,
n
L[tn ] = L[tn1 ], s > 0.
s
143
n!
sn est
=0
Using induction on n = 0, 1, 2, one can easily eastablish that

L[tn ] =
n!
sn+1
, s > 0.
Using the above result together with the linearity property of L one can find
the Laplace transform of any polynomial.
The next two results are referred to as the first and second shift theorems.
As with the linearity property, the shift theorems increase the number of
functions for which we can easily find Laplace transforms.
Theorem 42.1 (First Shifting Theorem)
If f (t) is a piecewise continuous function for t 0 and has exponential order
at infinity with |f (t)| M eat , t C, then for any real number we have
L[et f (t)] = F (s ), s > a +
where L[f (t)] = F (s).
Proof.
From the definition of the Laplace transform we have
Z
Z
at
st at
L[e f (t)] =
e e f (t)dt =
e(sa)t f (t)dt.
0
Using the change of variable = s a the previous equation reduces to

Z
Z
at
st at
L[e f (t)] =
e e f (t)dt =
et f (t)dt = F () = F (sa), s > a+
0
Theorem 42.2 (Second Shifting Theorem)

If f (t) is a piecewise continuous function for t 0 and has exponential order
at infinity with |f (t)| M eat , t C, then for any real number 0 we
have
L[f (t )h(t )] = es F (s), s > a
where L[f (t)] = F (s) and h(t) is the Heaviside step function.
Proof.
From the definition of the Laplace transform we have
Z
Z
st
L[f (t )h(t )] =
f (t )h(s )e dt =
144
f (t )est dt.
Using the change of variable = t the previous equation reduces to

R
L[f (t )h(t )] =
f ()es(+) d
0
= es
R
0
f ()es d = es F (s), s > a
Example 42.1
Find
(a) L[e2t t2 ] (b) L[e3t cos 2t] (c) L1 [e2t s2 ]
Solution.
(a) By Theorem 42.1, we have L[e2t t2 ] = F (s 2) where L[t2 ] = s2!3 =
2
F (s), s > 0. Thus, L[e2t t2 ] = (s2)
3 , s > 2.
3t
(b) As in part (a), we have L[e cos 2t] = F (s3) where L[cos 2t] = F (s3).
But L[cos 2t] = s2s+4 , s > 0. Thus,
L[e3t cos 2t] =
(c) Since L[t] =
1
s2
s3
, s>3
(s 3)2 + 4
then by Theorem 42.2, we have

e2t
= L[(t 2)h(t 2)].
s2
Therefore,
1

e2t
0,
0t<2
= (t 2)h(t 2) =
2
t
2,
t2
s
The following result relates the Laplace transform of derivatives and integrals
to the Laplace transform of the function itself.
Theorem 42.3
Suppose that f (t) is continuous for t 0 and f 0 (t) is piecewise continuous
of exponential order at infinity with |f 0 (t)| M eat , t C Then
(a) f (t) is of exponential order at infinity.
(b) L[f 0 (t)] = sL[f (t)] f (0) = sF (s) f (0), s > max{a, 0} + 1.
(c) L[f 00 (t)] = s2 L[f (t)] sf (0) f 0 (0) = s2 F (s) sf (0) f (0), s >
max{a,
hR0} + 1. i
t
(d) L 0 f (u)du = L[fs(t)] = F (s)
, s > max{a, 0} + 1.
s
145
Proof.
Rt
(a) By the Fundamental Theorem of Calculus we have f (t) = f (0) 0 f 0 (u)du.
Also, since f 0 is piecewise continuous then |f 0 (t)| T for some T > 0 and
all 0 t C. Thus,

Rt
RC
Rt 0

|f (t)| = f (0) 0 f (u)du = |f (0) 0 f 0 (u)du C f 0 (u)du|
|f (0)| + T C + M
Note that if a > 0 then

Z
eau du
1
eat
eau du = (eat eaC )
a
a
and so
|f (t)| [|f (0)| + T C +
If a = 0 then
Rt
M at
]e .
a
eau du = t C
and therefore
|f (t)| |f (0)| + T C + M (t C) (|f (0)| + T C + M )et .
Now, if a < 0 then
Z
1
1
eau du = (eat eaC )
a
|a|
so that
|f (t)| (|f (0)| + T C +
M t
)e
|a|
It follows that
|f (t)| N ebt , t 0
where b = max{a, 0} + 1.
(b) From the definition of Laplace transform we can write
Z A
0
L[f (t)] = lim
f 0 (t)est dt.
A
146
Since f 0 (t) may have jump discontinuities at t1 , t2 , , tN in the interval

0 t A, we can write
Z A
Z t2
Z t1
Z A
0
st
0
st
0
st
f 0 (t)est dt.
f (t)e dt + +
f (t)e dt +
f (t)e dt =
tN
t1
Integrating each term on the RHS by parts and using the continuity of f (t)
to obtain
R t1 0
Rt
f (t)est dt =
f (t1 )est1 f (0) + s 0 1 f (t)est dt
0
R t2
t1
f 0 (t)est dt
..
.
R tN
tN 1
RA
tN
f (t2 )est2 f (t1 )est1 + s
R t2
t1
f 0 (t)est dt = f (tN )estN f (tN 1 )estN 1 + s
f 0 (t)est dt
f (A)esA f (tN )estN + s
f (t)est dt
R tN
RA
tN
tN 1
f (t)est dt
Also, by the continuity of f (t) we can write

Z A
Z t1
Z t2
Z
st
st
st
f (t)e dt =
f (t)e dt +
f (t)e dt + +
0
t1
f (t)est dt
f (t)est dt.
tN
Hence,
Z
A
0
st
f (t)e
sA
dt = f (A)e
f (0) + s
f (t)est dt.
Since f (t) has exponential order at infinity then limA f (A)esA = 0.

Hence,
L[f 0 (t)] = sL[f (t)] f (0).
(c) Using part (b) we find
L[f 00 (t)] =
sL[f 0 (t)] f 0 (0)
=
s(sF (s) f (0)) f 0 (0)
2
= s F (s) sf (0) f 0 (0), s > max{a, 0} + 1
R

t
d
(d) Since dt 0 f (u)du = f (t) then by part (b) we have
Z
F (s) = L[f (t)] = sL

f (u)du
147
and therefore
Z t

F (s)
L[f (t)]
=
, s > max{a, 0} + 1
L
f (u)du =
s
s
0
The argument establishing part (b) of the previous theorem can be extended
to higher order derivatives.
Theorem 42.4
Let f (t), f 0 (t), , f (n1) (t) be continuous and f (n) (t) be piecewise continuous
of exponential order at infinity with |f (n) (t)| M eat , t C. Then
L[f (n) (t)] = sn L[f (t)]sn1 f (0)sn2 f 0 (0) f (n1) (0), s > max{a, 0}+1.
We next illustrate the use of the previous theorem in solving initial value
problems.
Example 42.2
y 00 4y 0 + 9y = t, y(0) = 0, y 0 (0) = 1.
Solution.
We apply Theorem 42.4 that gives the Laplace transform of a derivative. By
the linearity property of the Laplace transform we can write
L[y 00 ] 4L[y 0 ] + 9L[y] = L[t].
Now since
L[y 00 ] = s2 L[y] sy(0) y 0 (0) = s2 Y (s) 1
L[y 0 ] =
sY (s) y(0) = sY (s)
1
L[t] =
s2
where L[y] = Y (s), we obtain
s2 Y (s) 1 4sY (s) + 9Y (s) =
Rearranging gives
(s2 4s + 9)Y (s) =
148
s2 + 1
.
s2
1
.
s2
Thus,
Y (s) =
and
1
y(t) = L
s2 + 1
s2 (s2 4s + 9)
s2 + 1
s2 (s2 4s + 9)
In the next section we will discuss a method for finding the inverse Laplace
transform of the above expression.
Example 42.3
Consider the mass-spring oscillator without friction: y 00 + y = 0. Suppose
we add a force which corresponds to a push (to the left) of the mass as it
oscillates. We will suppose the push is described by the function
f (t) = h(t 2) + u(t (2 + a))
for some a > 2 which we are allowed to vary. (A small a will correspond
to a short duration push and a large a to a long duration push.) We are
interested in solving the initial value problem
y 00 + y = f (t), y(0) = 1, y 0 (0) = 0.
Solution.
To begin, determine the Laplace transform of both sides of the DE:
L[y 00 + y] = L[f (t)]
or
1
1
s2 Y sy(0) y 0 (0) + Y (s) = e2s + e(2+a)s .
s
s
Thus,
e(2+a)s
e2s
s
Y (s) =
+ 2
.
2
2
s(s + 1) s(s + 1) s + 1
Now since
1
s(s2 +1)
1
s
(2+a)s
Y (s) = e
s
s2 +1
we see that

1
s
s
s
2s 1
2
e
2
+ 2
s s +1
s s +1
s +1
149
and therefore

s
y(t) =
h(t (2 + a)) L1 1s s
(t (2 + a))
2 +1
h(t 2) L1 1s s2s+1 (t 2) + cos t

= h(t (2 + a))[1 cos (t (2 + a))] u(t 2)[1 cos (t 2)]
+
cos t
We conclude this section with the following table of Laplace transform pairs.
f(t)
F(s)

h(t) =
1, t 0
0, t < 0
1
,
s
s>0
tn , n = 1, 2,
n!
,
sn+1
s>0
et
s
,
s
s>
sin (t)
,
s2 + 2
s>0
cos (t)
s
,
s2 + 2
s>0
sinh (t)
,
s2 2
s > ||
cosh (t)
s
,
s2 2
s > ||
et f (t), with |f (t)| M eat
F (s ), s > + a
et h(t)
1
,
s
et tn , n = 1, 2,
n!
,
(s)n+1
et sin (t)
,
(s)2 + 2
s>
et cos (t)
s
,
(s)2 + 2
s>
f (t )h(t ), 0
with |f (t)| M eat
es F (s), s > a
150
s>
s>
f(t)
h(t ), 0
F(s) (continued)
es
, s>0
s
tf (t)
-F 0 (s)
t
2
s
,
(s2 + 2 )2
s>0
1
,
(s2 + 2 )2
s>0
sin t
1
[sin t
2 3
t cos t]
f 0 (t), with f (t) continuous

and |f 0 (t)| M eat
sF (s) f (0)
s > max{a, 0} + 1
f 00 (t), with f 0 (t) continuous

and |f 00 (t)| M eat
s2 F (s) sf (0) f 0 (0)

s > max{a, 0} + 1
f (n) (t), with f (n1) (t) continuous

and |f (n) (t)| M eat
sn F (s) sn1 f (0)

-sf (n2) (0) f (n1) (0)
s > max{a, 0} + 1
Rt
F (s)
,
s
f (u)du, with |f (t)| M eat

Table L
Problem 42.1
Use Table L to find L[2et + 5].
Problem 42.2
Use Table L to find L[e3t3 h(t 1)].
Problem 42.3
Use Table L to find L[sin2 t].
Problem 42.4
Use Table L to find L[sin 3t cos 3t].
Problem 42.5
Use Table L to find L[e2t cos 3t].
151
s > max{a, 0} + 1
Problem 42.6
Use Table L to find L[e4t (t2 + 3t + 5)].
Problem 42.7
+
Use Table L to find L1 [ s210
+25
4
].
s3
Problem 42.8
5
Use Table L to find L1 [ (s3)
4 ].
Problem 42.9
2s
Use Table L to find L1 [ es9 ].
Problem 42.10
3s
Use Table L to find L1 [ e s2(2s+7)
].
+16
Problem 42.11
Graph the function f (t) = h(t 1) + h(t 3) for t 0, where h(t) is the
Heaviside step function, and use Table L to find L[f (t)].
Problem 42.12
Graph the function f (t) = t[h(t 1) h(t 3)] for t 0, where h(t) is the
Problem 42.13
Graph the function f (t) = 3[h(t 1) h(t 4)] for t 0, where h(t) is the
Problem 42.14
Graph the function f (t) = |2 t|[h(t 1) h(t 3)] for t 0, where h(t) is
the Heaviside step function, and use Table L to find L[f (t)].
Problem 42.15
Graph the function f (t) = h(2 t) for t 0, where h(t) is the Heaviside step
function, and use Table L to find L[f (t)].
Problem 42.16
Graph the function f (t) = h(t 1) + h(4 t) for t 0, where h(t) is the
152
Problem 42.17
The graph of f (t) is given below. Represent f (t) as a combination of Heaviside step functions, and use Table L to calculate the Laplace transform of
f (t).
Problem 42.18
The graph of f (t) is given below. Represent f (t) as a combination of Heaviside step functions, and use Table L to calculate the Laplace transform of
f (t).
Problem 42.19
h
i
12
Using the partial fraction decomposition find L1 (s3)(s+1)
.
Problem 42.20
h 5s i
.
Using the partial fraction decomposition find L1 24e
s2 9
Problem 42.21
Use Laplace transform technique to solve the initial value problem
y 0 + 4y = g(t), y(0) = 2
where
0, 0 t < 1
12, 1 t < 3
g(t) =
0,
t3
153
Problem 42.22
Use Laplace transform technique to solve the initial value problem
y 00 4y = e3t , y(0) = 0, y 0 (0) = 0.
Problem 42.23
Rt
Obtain the Laplace transform of the function 2 f ()d in terms of L[f (t)] =
R2
F (s) given that 0 f ()d = 3.
154
43
The Laplace Transform and the Method

of Partial Fractions
In the last example of the previous section we encountered the equation

s2 + 1
1
.
y(t) = L
s2 (s2 4s + 9)
We would like to find an explicit expression for y(t). This can be done using
the method of partial fractions
which
is the topic of this section. According
1 N (s)
to this method, finding L
, where N (s) and D(s) are polynomials,
D(s)
require decomposing the rational function into a sum of simpler expressions
whose inverse Laplace transform can be recognized from a table of Laplace
transform pairs.
The method of integration by partial fractions is a technique for integrating
rational functions, i.e. functions of the form
R(s) =
N (s)
D(s)
where N (s) and D(s) are polynomials.

The idea consists of writing the rational function as a sum of simpler fractions called partial fractions. This can be done in the following way:
Step 1. Use long division to find two polynomials r(s) and q(s) such that
r(s)
N (s)
= q(s) +
.
D(s)
D(s)
Note that if the degree of N (s) is smaller than that of D(s) then q(s) = 0
and r(s) = N (s).
Step 2. Write D(s) as a product of factors of the form (as + b)n or (as2 +
bs+c)n where as2 +bs+c is irreducible, i.e. as2 +bs+c = 0 has no real zeros.
r(s)
Step 3. Decompose D(s)
into a sum of partial fractions in the following
way:
(1) For each factor of the form (s )k write
A1
A2
Ak
+
+ +
,
2
s (s )
(s )k
155
where the numbers A1 , A2 , , Ak are to be determined.

(2) For each factor of the form (as2 + bs + c)k write
B2 s + C2
B1 s + C1
Bk s + Ck
+
+
+
,
as2 + bs + c (as2 + bs + c)2
(as2 + bs + c)k
where the numbers B1 , B2 , , Bk and C1 , C2 , , Ck are to be determined.
Step 4. Multiply both sides by D(s) and simplify. This leads to an expression of the form
r(s) = a polynomial whose coefficients are combinations of Ai , Bi , and Ci .
Finally, we find the constants, Ai , Bi , and Ci by equating the coefficients of
like powers of s on both sides of the last equation.
Example 43.1
Decompose into partial fractions R(s) =
s3 +s2 +2
.
s2 1
Solution.
3
2 +2
Step 1. s s+s
= s + 1 + ss+3
2 1
2 1 .
2
Step 2. s 1 = (s 1)(s + 1).
s+3
A
B
Step 3. (s+1)(s1)
= s+1
+ s1
.
Step 4. Multiply both sides of the last equation by (s 1)(s + 1) to obtain
s + 3 = A(s 1) + B(s + 1).
Expand the right hand side, collect terms with the same power of s, and
identify coefficients of the polynomials obtained on both sides:
s + 3 = (A + B)s + (B A).
Hence, A + B = 1 and B A = 3. Adding these two equations gives B = 2.
Thus, A = 1 and so
s3 + s2 + 2
1
2
=s+1
+
.
2
s 1
s+1 s1
Now, after decomposing the rational function into a sum of partial fractions
all we need to do is to find the Laplace transform of expressions of the form
A
or (as2Bs+C
.
(s)n
+bs+c)n
156
Exampleh 43.2 i
1
.
Find L1 s(s3)
Solution.
We write
1
A
B
= +
.
s(s 3)
s
s3
Multiply both sides by s(s 3) and simplify to obtain

1 = A(s 3) + Bs
or
1 = (A + B)s 3A.
Now equating the coefficients of like powers of s to obtain 3A = 1 and
A + B = 0. Solving for A and B we find A = 31 and B = 13 . Thus,
h
i

1
1
L1 s(s3)
= 31 L1 1s + 13 L1 s3
=
13 h(t) + 31 e3t , t 0
where h(t) is the Heaviside unit step function
Example 43.3
Find L1 s3s+6
2 +3s .
Solution.
We factor the denominator and split the integrand into partial fractions:
3s + 6
A
B
= +
.
s(s + 3)
s
s+3
Multiplying both sides by s(s + 3) to obtain
3s + 6 = A(s + 3) + Bs
= (A + B)s + 3A
Equating the coefficients of like powers of x to obtain 3A = 6 and A + B = 3.
Thus, A = 2 and B = 1. Finally,

3s + 6
1
1
1 1
1
= 2L
L
+L
= 2h(t) + e3t , t 0.
2
s + 3s
s
s+3
157
Exampleh 43.4 i
s2 +1
.
Find L1 s(s+1)
2
Solution.
We factor the denominator and split the rational function into partial fractions:
s2 + 1
B
C
A
+
= +
.
2
s(s + 1)
s
s + 1 (s + 1)2
Multiplying both sides by s(s + 1)2 and simplifying to obtain
s2 + 1 =
A(s + 1)2 + Bs(s + 1) + Cs
= (A + B)s2 + (2A + B + C)s + A.
Equating coefficients of like powers of s we find A = 1, 2A + B + C = 0
and A + B = 1. Thus, B = 0 and C = 2. Now finding the inverse Laplace
transform to obtain

2
1
s +1
1 1
1
1
=L
2L
= h(t) 2tet , t 0.
L
2
2
s(s + 1)
s
(s + 1)
Example 43.5
Use Laplace transform to solve the initial value problem
y 00 + 3y 0 + 2y = et , y(0) = y 0 (0) = 0.
Solution.
By the linearity property of the Laplace transform we can write
L[y 00 ] + 3L[y 0 ] + 2L[y] = L[et ].
Now since
L[y 00 ] = s2 L[y] sy(0) y 0 (0) = s2 Y (s)

L[y 0 ] =
sY (s) y(0) = sY (s)
1
t
L[e ] =
s+1
where L[y] = Y (s), we obtain

s2 Y (s) + 3sY (s) + 2Y (s) =
158
1
.
s+1
Rearranging gives
(s2 + 3s + 2)Y (s) =
1
.
s+1
Thus,
Y (s) =
and
1
(s +

y(t) = L
1
.
+ 3s + 2)
1)(s2

1
.
(s + 1)(s2 + 3s + 2)
Using the method of partial fractions we can write

(s +
1
1
1
1
=
+
.
+ 3s + 2)
s + 2 s + 1 (s + 1)2
1)(s2
Thus,
1
y(t) = L

1
1
1
1
1
L
+L
= e2t et +tet , t 0
s+2
s+1
(s + 1)2
In Problems 43.1 - 43.4, give the form of the partial fraction expansion for
F (s). You need not evaluate the constants in the expansion. However, if the
denominator has an irreducible quadratic expression then use the completing
the square process to write it as the sum/difference of two squares.
Problem 43.1
F (s) =
s3 + 3s + 1
.
(s 1)3 (s 2)2
Problem 43.2
s2 + 5s 3
F (s) = 2
.
(s + 16)(s 2)
Problem 43.3
F (s) =
s3 1
.
(s2 + 1)2 (s + 4)2
159
Problem 43.4
F (s) =
s4 + 5s2 + 2s 9
.
(s2 + 8s + 17)(s 2)2
Problemh43.5 i
1
.
Find L1 (s+1)
3
Problem43.6
Find L1 s22s3
.
3s+2
Problemh43.7 i
2
.
Find L1 4ss3+s+1
+s
Problemh43.8
i
2 +6s+8
Find L1 s4s+8s
2 +16 .
Problem 43.9
y 0 + 2y = 26 sin 3t, y(0) = 3.
Problem 43.10
y 0 + 2y = 4t, y(0) = 3.
Problem 43.11
y 00 + 3y 0 + 2y = 6et , y(0) = 1, y 0 (0) = 2.
Problem 43.12
y 00 + 4y = cos 2t, y(0) = 1, y 0 (0) = 1.
Problem 43.13
y 00 2y 0 + y = e2t , y(0) = 0, y 0 (0) = 0.
160
Problem 43.14
y 00 + 9y = g(t), y(0) = 1, y 0 (0) = 0
where

g(t) =
6, 0 t <
0, t <
Problem 43.15
Determine the constants , , y0 , and y00 so that Y (s) =
transform of the solution to the initial value problem
2s1
s2 +s+2
is the Laplace
y 00 + y 0 + y = 0, y(0) = y0 , y 0 (0) = y00 .

Problem 43.16
Determine the constants , , y0 , and y00 so that Y (s) =
transform of the solution to the initial value problem
s
(s+1)2
y 00 + y 0 + y = 0, y(0) = y0 , y 0 (0) = y00 .
161
is the Laplace
44
Laplace Transforms of Periodic Functions
In many applications, the nonhomogeneous term in a linear differential equation is a periodic function. In this section, we derive a formula for the Laplace
transform of such periodic functions.
Recall that a function f (t) is said to be T periodic if we have f (t+T ) = f (t)
whenever t and t + T are in the domain of f (t). For example, the sine and
cosine functions are 2periodic whereas the tangent and cotangent functions are periodic.
If f (t) is T periodic for t 0 then we define the function

f (t), 0 t T
fT (t) =
0,
t>T
The Laplace transform of this function is then
Z
Z
st
L[fT (t)] =
fT (t)e dt =
0
f (t)est dt.
The Laplace transform of a T periodic function is given next.

Theorem 44.1
If f (t) is a T periodic, piecewise continuous fucntion for t 0 then
L[f (t)] =
L[fT (t)]
, s > 0.
1 esT
Proof.
Since f (t) is piecewise continuous then it is bounded on the interval 0 t
T. By periodicity, f (t) is bounded for t 0. Hence, it has an exponential
order at infinity. By Theorem 41.2, L[f (t)] exists for s > 0. Thus,
Z
L[f (t)] =
st
f (t)e
0
dt =
Z
X
n=0
fT (t nT )h(t nT )dt,
where the last sum is the result of decomposing the improper integral into a
sum of integrals over the constituent periods.
By the Second Shifting Theorem (i.e. Theorem 42.2) we have
L[fT (t nT )h(t nT )] = enT s L[fT (t)], s > 0
162
Hence,
L[f (t)] =
enT s L[fT (t)] = L[fT (t)]
n=0
!
enT s
n=0
Since s > 0, it follows that 0 < enT s < 1 so that the series
convergent geoemetric series with limit 1e1sT . Therefore,
L[f (t)] =
n=0
enT s is a
L[fT (t)]
, s>0
1 esT
Example 44.1
Determine the Laplace transform of the function
1, 0 t T2
f (t + T ) = f (t), t 0.
f (t) =
0, T2 < t < T
Solution.
The graph of f (t) is shown in Figure 44.1.
Figure 44.1
By Theorem 44.1,
RT
est dt
, s > 0.
1 esT
Evaluating this last integral, we find
2
L[f (t)] =
sT
L[f (t)] =
1e 2
s
1 esT
1
sT
s(1 + e 2 )
, s>0
Example 44.2
Find the Laplace transform of the sawtooth curve shown in Figure 44.2
163
Figure 44.2
Solution.
The given function is periodic of period b. For the first period the function
is defined by
a
fb (t) = t[h(t) h(t b)].
b
So we have
L[fb (t)] = L[ ab t(h(t) h(t b))]
d
= ab ds
L[h(t) h(t b)]
But
L[h(t) h(t b)] = L[h(t)] L[h(t b)]

bs
1
=
es , s>0
s
Hence,
a
L[fb (t)] =
b
bsebs + ebs
1
s2
s2

.
Finally,

a 1 ebs bsebs
L[fb (t)]
=
L[f (t)] =
1 ebs
b
s2 (1 ebs )
Exampleh 44.3
i
es
Find L1 s12 s(1e
s ) .
Solution.
Note first that
1
es
1 es ses
=
.
s2 s(1 es )
s2 (1 es )
164
According
to theiprevious example with a = 1 and b = 1 we find that
h
es
1 1
is the sawtooth function shown in Figure 44.2
L
s(1e
s )
s2
Linear Time Invariant Systems and the Transfer Function
The Laplace transform is a powerful technique for analyzing linear timeinvariant systems such as electrical circuits, harmonic oscillators, optical devices, and mechanical systems, to name just a few. A mathematical model
described by a linear differential equation with constant coefficients of the
form
an y (n) + an1 y (n1) + + a1 y 0 + a0 y = bm u(m) + bm1 u(m1) + + b1 u0 + b0 u
is called a linear time invariant system. The function y(t) denotes the
system output and the function u(t) denotes the system input. The system is
called time-invariant because the parameters of the system are not changing
over time and an input now will give the same result as the same input later.
Applying the Laplace transform on the linear differential equation with null
initial conditions we obtain
an sn Y (s)+an1 sn1 Y (s)+ +a0 Y (s) = bm sm U (s)+bm1 sm1 U (s)+ +b0 U (s).
The function
bm sm + bm1 sm1 + + b1 s + b0
Y (s)
=
(s) =
U (s)
an sn + an1 sn1 + + a1 s + a0
is called the system transfer function. That is, the transfer function of
a linear time-invariant system is the ratio of the Laplace transform of its
output to the Laplace transform of its input.
Example 44.4
Consider the mathematical model described by the initial value problem
my 00 + y 0 + ky = f (t), y(0) = 0, y 0 (0) = 0.
The coefficients m, , and k describe the properties of some physical system,
and f (t) is the input to the system. The solution y is the output at time t.
Find the system transfer function.
165
Solution.
By taking the Laplace transform and using the initial conditions we obtain
(ms2 + s + k)Y (s) = F (s).
Thus,
(s) =
1
Y (s)
=
F (s)
ms2 + s + k
(29)
Parameter Identification
One of the most useful applications of system transfer functions is for system
or parameter identification.
Example 44.5
Consider a spring-mass system governed by
my 00 + y 0 + ky = f (t), y(0) = 0, y 0 (0) = 0.
(30)
Suppose we apply a unit step force f (t) = h(t) to the mass, initially at
equilibrium, and you observe the system respond as
1
1
1
y(t) = et cos t et sin t + .
2
2
2
What are the physical parameters m, , and k?
Solution.
Start with the model (30)) with f (t) = h(t) and take the Laplace transform
1
of both sides, then solve to find Y (s) = s(ms2 +s+k)
. Since f (t) = h(t) then
1
F (s) = s . Hence
Y (s)
1
(s) =
=
.
2
F (s)
ms + s + k
On the other hand, for the input f (t) = h(t) the corresponding observed
output is
1
1
1
y(t) = et cos t et sin t + .
2
2
2
Hence,
Y (s) = L[ 12 et cos t 12 et sin t + 21 ]
s+1
1
1
1
= 12 (s+1)
2 +1 2 (s+1)2 +1 + 2s
1
=
s(s2 +2s+2)
166
Thus,
1
Y (s)
= 2
.
F (s)
s + 2s + 2
By comparison we conclude that m = 1, = 2, and k = 2
(s) =
Problem 44.1
Find the Laplace transform of the periodic function whose graph is shown.
Problem 44.2
Problem 44.3
167
Problem 44.4
Problem 44.5
State the period of the function f (t) and find its Laplace transform where
sin t, 0 t <
f (t + 2) = f (t), t 0.
f (t) =
0, t < 2
Problem 44.6
State the period of the function f (t) = 1 et , 0 t < 2, f (t + 2) = f (t),
and find its Laplace transform.
Problem 44.7
Using Example 44.3 find
1

es
s2 s
+
.
s3
s(1 es )
Problem 44.8
An object having mass m is initially at rest on a frictionless horizontal surface.
At time t = 0, a periodic force is applied horizontally to the object, causing
it to move in the positive x-direction. The force, in newtons, is given by
f0 , 0 t T2
f (t + T ) = f (t), t 0.
f (t) =
0, T2 < t < T
The initial value problem for the horizontal position, x(t), of the object is
mx00 (t) = f (t), x(0) = x0 (0) = 0.
168
(a) Use Laplace transforms to determine the velocity, v(t) = x0 (t), and the
position, x(t), of the object.
(b) Let m = 1 kg, f0 = 1 N, and T = 1 sec. What is the velocity, v, and
position, x, of the object at t = 1.25 sec?
Problem 44.9
ay 00 + by 0 + cy = f (t), y(0) = y 0 (0) = 0, t > 0
Suppose that the transfer function of this system is given by (s) =
(a) What are the constants a, b, and c?
(b) If f (t) = et , determine F (s), Y (s), and y(t).
1
.
2s2 +5s+2
Problem 44.10
ay 00 + by 0 + cy = f (t), y(0) = y 0 (0) = 0, t > 0
Suppose that an input f (t) = t, when applied to the above system produces
the output y(t) = 2(et 1) + t(et + 1), t 0.
(a) What is the system transfer function?
(b) What will be the output if the Heaviside unit step function f (t) = h(t)
is applied to the system?
Problem 44.11
y 00 + y 0 + y = f (t), y(0) = y 0 (0) = 0,
where
f (t) =
1,
0t1
f (t + 2) = f (t)
1, 1 < t < 2
(a) Determine the system transfer function (s).

(b) Determine Y (s).
Problem 44.12
y 000 4y = et + t, y(0) = y 0 (0) = y 00 (0) = 0.
(a) Determine the system transfer function (s).
(b) Determine Y (s).
169
Problem 44.13
y 00 + by 0 + cy = h(t), y(0) = y0 , y 0 (0) = y00 , t > 0.
Suppose that L[y(t)] = Y (s) =
and y00 .
s2 +2s+1
.
s3 +3s2 +2s
170
Determine the constants b, c, y0 ,
45
Solving Systems of Differential Equations

Using Laplace Transform
In this section we extend the definition of Laplace transform to matrix-valued

functions and apply this extension to solving systems of differential equations.
Let y1 (t), y2 (t), , yn (t) be members of PE. Consider the vector-valued
function
y1 (t)
y2 (t)
y(t) = ..
.
yn (t)
The Laplace transform of y(t) is
L[y(t)] =
R
0
y(t)est dd
R
y (t)est dt
0 1
R
y2 (t)est dt
= 0
..
.
R
st
y
(t)e
dt
n
0
L[y1 (t)]
L[y2 (t)]
..
.
L[yn (t)]
In a similar way, we define the Laplace transform of an m n matrix to

be the m n matrix consisting of the Laplace transforms of the component
functions. If the Laplace transform of each component exists then we say
y(t) is Laplace transformable.
Example 45.1
Find the Laplace transform of the vector-valued function
2
t
y(t) = 1
et
171
Solution.
The Laplace transform is
6
s3
L[y(t)] =
1
s
, s > 1
1
s1
The linearity property of the Laplace transform can be used to establish the
following result.
Theorem 45.1
If A is a constant n n matrix and B is an n p matrix-valued function
then
L[AB(t)] = AL[B(t)].
Proof.
P
Let A = (aij ) and B(t) = (bij (t)). Then AB(t) = ( nk=1 aik bkp ). Hence,
n
n
X
X
aik L(bkp )] = AL[B(t)]
aik bkp )] = [
L[AB(t)] = [L(
k=1
k=1
Theorem 42.3 can be extended to vector-valued functions.

Theorem 45.2
(a) Suppose that y(t) is continuous for t 0 and let the components of the
derivative vector y0 be members of PE. Then
L[y0 (t)] = sL[y(t)] y(0).
(b) Let y0 (t) be continuous for t 0, and let the entries of y00 (t) be members
of PE. Then
L[y00 (t)] = s2 L[y(t)] sy(0) y0 (0).
(c) Let the entries of y(t) be members of PE. Then
Z t

L[y(t)]
L
y(s)ds =
.
s
0
172
Proof.
(a) We have
L[y0 (t)] =
L[y10 (t)]
L[y20 (t)]
..
.
L[yn0 (t)]
sL[y1 (t)] y1 (0)

sL[y2 (t)] y2 (0)
..
.
sL[yn (t)] yn (0)
sL[y(t)] y(0)
(b) We have
L[y00 (t)] =
sL[y0 (t)] y0 (0)
= s(sL[y(t)] y(0)) y0 (0)
= s2 L[y(t)] sy(0) y0 (0)
(c) We have
Z
L[y(t)] = sL

y(s)ds
so that
Z
L

y(s)ds
L[y(t)]
s
The above two theorems can be used for solving the following initial value
problem
y0 (t) = Ay + g(t),
y(0) = y0 , t > 0
(31)
where A is a constant matrix and the components of g(t) are members of

PE.
Using the above theorems we can write
sY(s) y0 = AY(s) + G(s)
or
(sI A)Y(s) = y0 + G(s)
173
where L[g(t)] = G(s). If s is not an eigenvalue of A then the matrix sI A

is invertible and in this case we have
Y(s) = (sI A)1 [y0 + G(s)].
(32)
To compute y(t) = L1 [Y(s)] we compute the inverse Laplace transform

of each component of Y(s). We illustrate the above discussion in the next
example.
Example 45.2

2t

1 2
e
1
0
y =
y+
, y(0) =
2 1
2t
2
Solution.
We have
1
(sI A)
1
=
(s + 1)(s 3)
1
s2
s22
and

G(s) =
s1
2
2
s1
Thus,
1
(sI
A) [y0 + G(s)]

1
1 + s2
s1
2
1
= (s+1)(s3)
2
s1
2 s22
Y(s) =
"
=
s4 6s3 +9s2 4s+8

s2 (s+1)(s2)(s3)
2s4 +8s3 8s2 +6s4
s2 (s+1)(s2)(s3)
Using the method of partial fractions we can write

Y1 (s) =
4 1
3 s2
Y2 (s) = 23 s12 +
81
9s
10 1
9 s
7 1
3 s+1
7 1
3 s+1
1 1
3 s2
2 1
3 s2
1 1
9 s3
1 1
9 s3
Therefore
y1 (t) =
L1 [Y1 (s)] = 43 t 89 + 37 et 13 e2t 19 e3t
y2 (t) = L1 [Y2 (s)] = 23 t +
10
9
174
73 et 23 e2t 91 e3t , t 0
Hence, for t 0

4 8
9
t
3
y(t) = t
+ 10 + e
23
9
7
3
7
3
2t
+e
13
23
3t
+e
19
91
System Transfer Matrix and the Laplace Transform of etA

The vector equation (31) is a linear time invariant system whose Laplace
input is given by y0 + G(s) and the Laplace output Y(s). According to
(32) the system tranform matrix is given by (sI A)1 . We will show that
this matrix is the Laplace transform of the exponential matrix function etA .
Indeed, etA is the solution to the initial value problem
0 (t) = A(t), (0) = I,
where I is the nn identity matrix and A is a constant nn matrix. Taking
Laplace of both sides yields
sL[(t)] I = AL[(t)].
Solving for L[(t)] we find
L[(t)] = (sI A)1 = L[etA ].
Problem 45.1
Find L[y(t)] where
et cos 2t
d
0
y(t) =
dt
t + et
Problem 45.2
Find L[y(t)] where
t
Z
y(t) =
0
1
u du
eu
Problem 45.3
Find L1 [Y(s)] where
Y(s) =
1
s
2
2
s +2s+2
1
s2 +s
175
Problem 45.4
Find L1 [Y(s)] where
1 1 2
L[t3 ]
Y(s) = 2 0 3 L[e2t ]
1 2 1
L[sin t]
Problem 45.5
Use the Laplace transform to solve the initial value problem

5 4
0
0
0
y =
y+
, y(0) =
5 4
1
0
Problem 45.6

5 4
3
0
y =
y, y(0) =
3 2
2
Problem 45.7

1 4
0
3
0
y =
y+
, y(0) =
t
1 1
3e
0
Problem 45.8

3 2
1
0
00
0
y =
y, y(0) =
, y (0) =
4
3
0
1
Problem 45.9

1 1
2
0
0
00
0
y =
y+
, y(0) =
, y (0) =
1 1
1
1
0
Problem 45.10
Use the Laplace transform to solve the
1 0 0
y0 = 0 1 1 y +
0 0 2
initial value problem

et
0
1
, y(0) = 0
2t
0
176
Problem 45.11
0
The Laplace transform
was

applied to the initial value problem y = Ay,
y(0)
=
y1 (t)
y1,0
y0 , where y(t) =
, A is a 2 2 constant matrix, and y0 =
.
y2 (t)
y2,0
The following transform domain solution was obtained

1
s 2 1
y1,0
.
L[y(t)] = Y(s) = 2
4
s7
y2,0
s 9s + 18
(a) what are the eigenvalues of A?
(b) Find A.
177
46
Convolution Integrals
We start this section with the following problem.

Example 46.1
A spring-mass system with a forcing function f (t) is modeled by the following
initial-value problem
mx00 + kx = f (t), x(0) = x0 , x0 (0) = x00 .
Find solution to this initial value problem using the Laplace transform method.
Solution.
Apply Laplace transform to both sides of the equation to obtain
ms2 X(s) msx0 mx00 + kX(s) = F (s).
Solving the above algebraic equation for X(s) we find
F (s)
ms2 +k
X(s) =
1 F (s)
m s2 + k
m
msx0
ms2 +k
sx0
k
s2 + m
+
+
mx00
ms2 +k
x00
k
2
s +m
Apply the inverse Laplace transform to obtain

L1 [X(s)]
n
o
x(t) =
=
=
1 1
L
m
1 1
L
m
F (s)
k
s2 + m
n
F (s)
1
k
s2 + m
+ x0 L
+ x0 cos
s
k
s2 + m
q
k
m
x00 L1
t+
x00
pm
k
1
k
s2 + m
sin
q
k
m
n
o
Finding L1 F (s) s2 +1 k ,i.e., the inverse Laplace transform of a product,
m
requires the use of the concept of convolution, a topic we discuss in this
section
Convolution integrals are useful when finding the inverse Laplace transform
of products H(s) = F (s)G(s). They are defined as follows: The convolution
of two scalar piecewise continuous functions f (t) and g(t) defined for t 0
is the integral
Z
t
(f g)(t) =
f (t s)g(s)ds.
0
178
Example 46.2
Find f g where f (t) = et and g(t) = sin t.
Solution.
Using integration by parts twice we arrive at
R t (ts)
(f g)(t) =
e
sin sds
0
(ts)
t
1
(sin s cos s) 0
= 2 e
t
= e 2 + 12 (sin t cos t)
Graphical Interpretation of Convolution Operation
For the convolution
Z t
f (t s)g(s)ds
(f g)(t) =
0
we perform the following:

Step 1. Given the graphs of f (s) and g(s).(Figure 46.1(a) and (b))
Step 2. Time reverse f (s). (See Figure 46.1(c))
Step 3. Shift f (s) right by an amount t to get f (t s). (See Figure 46.1(d))
Step 4. Determine the product f (t s)g(s). (See Figure 46.1(e))
Step 5. Determine the area under the graph of f (t s)g(s) between 0 and t.
(See Figure 46.1(e))
Figure 46.1
Next, we state several properties of convolution product, which resemble
those of ordinary product.
179
Theorem 46.1
Let f (t), g(t), and k(t) be three piecewise continuous scalar functions defined
for t 0 and c1 and c2 are arbitrary constants. Then
(i) f g = g f (Commutative Law)
(ii) (f g) k = f (g k) (Associative Law)
(iii) f (c1 g + c2 k) = c1 f g + c2 f k (Distributive Law)
Proof.
(i) Using the change of variables = t s we find
Rt
(f g)(t) =
f (t s)g(s)ds
0
=
=
Rt
0
R0
t
f ( )g(t )d
g(t )f ( )d = (g f )(t)
(ii) By definition, we have

Rt
[(f g) k)](t) =
=
R t hR tu
0
(f g)(t u)k(u)du
i
f (t u w)g(w)k(u)dw du
For the integral in the bracket, make change of variable w = s u. We have

Z t Z t
[(f g) k)](t) =
f (t s)g(s u)k(u)ds du.
0
This multiple integral is carried over the region

{(s, u) : 0 u s t}
as depicted by shaded region in the following graph.
Figure 46.2
180
Changing the order of integration, we have

R t R s
[(f g) k)](t) = 0 0 f (t s)g(s u)k(u)du ds
Rt
f (t s)(g k)(s)ds
=
0
=
[f (g k)](t)
(iii) We have
Rt
(f (c1 g + c2 k))(t) =
= c1
Rt
0
f (t s)(c1 g(s) + c2 k(s))ds
f (t s)g(s)ds + c2
Rt
0
f (t s)k(s)ds
c1 (f g)(t) + c2 (f k)(t)
Example 46.3
Express the solution to the initial value problem y 0 + y = g(t), y(0) = y0
in terms of a convolution integral.
Solution.
Solving this initial value problem by the method of integrating factor we find
Z t
t
e(ts) g(s)ds = et y0 + et g(t)
y(t) = e y0 +
0
Example 46.4
If f (t) is an m n matrix function and g(t) is an n p matrix function then
we define
Z t
(f g)(t) =
f (t s)g(s)ds, t 0.
0
Express the solution to the initial value problem y0 = Ay + g(t), y(0) = y0

in terms of a convolution integral.
Solution.
The unique solution is given by
Z t
tA
y(t) = e y0 +
eA(ts) g(s)ds = etA y0 + etA g(t)
0
The following theorem, known as the Convolution Theorem, provides a way

for finding the Laplace transform of a convolution integral and also finding
the inverse Laplace transform of a product.
181
Theorem 46.2
If f (t) and g(t) are piecewise continuous for t 0, and of exponential order
at infinity then
L[(f g)(t)] = L[f (t)]L[g(t)] = F (s)G(s).
Thus, (f g)(t) = L1 [F (s)G(s)].
Proof.
First we show that f g has a Laplace transform. From the hypotheses we
have that |f (t)| M1 ea1 t for t C1 and |g(t)| M2 ea2 t for t C2 . Let
M = M1 M2 and C = C1 + C2 . Then for t C we have
R
R
t

t
|(f g)(t)| = 0 f (t s)g(s)ds 0 |f (t s)||g(s)|ds
M1 M2

Rt
0
ea1 (ts) ea2 s ds
M tea1 t ,
a1 = a2
ea2 t ea1 t
6 a2
M a2 a1 , a1 =
This shows that f g is of exponential order at infinity. Since f and g are

piecewise continuous then the first fundamental theorem of calculus implies
that f g is also piecewise continuous. Hence, f g has a Laplace transform.
Next, we have
R

R
t
L[(f g)(t)] = 0 est 0 f (t )g( )d dt
=
R Rt
t=0
=0
est f (t )g( )d dt
Note that the region of integration is an infinite triangular region and the
integration is done vertically in that region. Integration horizontally we find
Z Z
est f (t )g( )dtd.
L[(f g)(t)] =
=0
t=
We next introduce the change of variables = t . The region of integration

becomes 0, t 0. In this case, we have
R R s(+ )
L[(f g)(t)] =
e
f ()g( )d d
=0 =0
=
=
es g( )d
=0
R
es f ()d
=0
G(s)F (s) = F (s)G(s)

182
Example 46.5
Use the convolution theorem to find the inverse Laplace transform of
H(s) =
Solution.
Note that

H(s) =
(s2
1
2
s + a2
1
.
+ a2 )2

1
2
s + a2

.
1
1
So, in this case we have, F (s) = G(s) = s2 +a
2 so that f (t) = g(t) = a sin (at).
Thus,
Z
1
1 t
sin (at as) sin (as)ds = 3 (sin (at) at cos (at))
(f g)(t) = 2
a 0
2a
Convolution integrals are useful in solving initial value problems with forcing
functions.
Example 46.6
4y 00 + y = g(t), y(0) = 3, y 0 (0) = 7
Solution.
Take the Laplace transform of all the terms and plug in the initial conditions
to obtain
4(s2 Y (s) 3s + 7) + Y (s) = G(s)
or
(4s2 + 1)Y (s) 12s + 28 = G(s).
Solving for Y (s) we find
12s28
4(s2 + 14 )
Y (s) =
=
3s
2
s2 +(( 12 )
+
2
G(s)
4(s2 + 14 )
( 12 )
( 21 )
1
+
G(s)
2
2
4
s2 +( 12 )
s2 +( 12 )
Hence,

Z
s
t
t
1 t
y(t) = 3 cos
7 sin
+
sin
g(t s)ds.
2
2
2 0
2
So, once we decide on a g(t) all we need to do is to evaluate the integral and
well have the solution
183
Problem 46.1
Consider the functions f (t) = g(t) = h(t), t 0 where h(t) is the Heaviside
unit step function. Compute f g in two different ways.
(a) By directly evaluating the integral.
(b) By computing L1 [F (s)G(s)] where F (s) = L[f (t)] and G(s) = L[g(t)].
Problem 46.2
Consider the functions f (t) = et and g(t) = e2t , t 0. Compute f g in
two different ways.
Problem 46.3
Consider the functions f (t) = sin t and g(t) = cos t, t 0. Compute f g in
two different ways.
Problem 46.4
Use
transform to
comput
the convolution P y, where |bf P (t) =
Laplace
t
h(t)
h(t) e
and y(t) =
.
0
t
et
Problem 46.5
Compute and graph f g where f (t) = h(t) and g(t) = t[h(t) h(t 2)].
Problem 46.6
Compute and graph f g where f (t) = h(t) h(t 1) and g(t) = h(t 1)
2h(t 2)].
Problem 46.7
Compute t t t.
Problem 46.8
Compute h(t) et e2t .
Problem 46.9
Compute t et et .
184
Problem 46.10
n f unctions
}|
{
z
Suppose it is known that h(t) h(t) h(t) = Ct8 . Determine the constants C and the poisitive integer n.
Problem 46.11
Use Laplace transform to solve for y(t) :
Z t
sin (t )y()d = t2 .
0
Problem 46.12
Z t
y(t)
e(t) y()d = t.
0
Problem 46.13
t y(t) = t2 (1 et ).
Problem 46.14

y = h(t) y, y(0) =
1
2

.
Problem 46.15
Solve the following initial value problem.
Z t
0
y y =
(t )e d, y(0) = 1.
0
185
47
The Dirac Delta Function and Impulse Response
In applications, we are often encountered with linear systems, originally at

rest, excited by a sudden large force (such as a large applied voltage to an
electrical network) over a very short time frame. In this case, the output
corresponding to this sudden force is referred to as the impulse response.
Mathematically, an impulse can be modeled by an initial value problem with
a special type of function known as the Dirac delta function as the external
force, i.e., the nonhomogeneous term. To solve such IVP requires finding the
Laplace transform of the delta function which is the main topic of this section.
An Example of Impulse Response
Consider a spring-mass system with a time-dependent force f (t) applied to
the mass. The situation is modeled by the second-order differential equation
my 00 + y 0 + ky = f (t)
(33)
where t is time and y(t) is the displacement of the mass from equilibrium.
Now suppose that for t 0 the mass is at rest in its equilibrium position, so
y(0) = y 0 (0) = 0. Hence, the situation is modeled by the initial value problem
my 00 + y 0 + ky = f (t), y(0) = 0, y 0 (0) = 0.
(34)
Solving this equation by the method of variation of parameters one finds the
unique solution
Z t
y(t) =
(t s)f (s)ds
(35)
0
where
q
k
e(/2m)t sin t m
q
(t) =
2
k
m m
4m
2
2
4m2

.
Next, we consider the problem of strucking the mass by an instantaneous

hammer blow at t = 0. This situation actually occurs frequently in practice-a
system sustains a forceful, almost-instantaneous input. Our goal is to model
the situation mathematically and determine how the system will respond.
186
In the above situation we might describe f(t) as a large constant force applied
on a very small time interval. Such a model leads to the forcing function
1
, 0t

f (t) =
0, otherwise
where is a small positive real number. When is close to zero the applied
force is very large during the time interval 0 t and zero afterwards. A
possible graph of f (t) is given in Figure 47.1
Figure 47.1
In this case its easy to see that for any choice of we have
Z
f dt = 1
and
lim f (t) = 0, t 6= 0,
0+
lim f (0) = .
0+
(36)
Our ultimate interest is the behavior of the solution to equation (33) with
forcing function f (t) in the limit 0+ . That is, what happens to the
system output as we make the applied force progressively sharper and
stronger?.
Let y (t) be the solution to equation (33) with f (t) = f (t). Then the unique
solution is given by
Z
t
(t s)f (s)ds.
y (t) =
0
187
For t the last equation becomes

Z
1
y (t) =
(t s)ds.
0
Since (t) is continuous for all t 0 we can apply the mean value theorem
for integrals and write
y (t) = (t )
for some 0 . Letting 0+ and using the continuity of we find
y(t) = lim+ y (t) = (t).
0
We call y(t) the impulse response of the linear system.

The Dirac Delta Function
The problem with the integral
Z
(t s)f (s)ds
0
is that lim0+ f (0) is undefined. So it makes sense to ask the question of

whether we can find a function (t) such that
Rt
lim0+ y (t) = lim0+ 0 (t s)f (s)ds
Rt
=
(t s)(s)ds
0
=
(t)
where the role of (t) would be to evaluate the integrand at s = 0. Note that
because of Fig 47.1 and (36), we cannot interchange the opeartions of limit
and integration in the above limit process. Such a function exist in the
theory of distributions and can be defined as follows:
If f (t) is continuous in a t b then we define the function (t) by the
integral equation
Z b
Z b
f (t)f (t t0 )dt.
f (t)(t t0 )dt = lim+
a
0
The object (t) on the left is called the Dirac Delta function, or just the
delta function for short.
188
Finding the Impulse Function Using Laplace Transform

For > 0 we can solve the initial value problem (34) using Laplace transforms. To do this we need to compute the Laplace transform of f (t), given
by the integral
Z
Z
1 st
1 es
st
f (t)e dt =
e dt =
.
L[f (t)] =
0
s
0
Note that by using LHopitals rule we can write
lim+ L[f (t)] = lim+
0
0
1 es
= 1, s > 0.
s
Now, to find y (t), we apply the Laplace transform to both sides of equation
(33) and using the initial conditions we obtain
ms2 Y (s) + sY (s) + kY (s) =
1 es
.
s

Y (s) =
1
1 es
.
ms2 + s + k s
Letting 0+ we find
Y (s) =
ms2
1
+ s + k
which is the transfer function of the system. Now inverse transform Y (s) to
find the solution to the initial value problem. That is,

1
1
y(t) = L
= (t).
ms2 + s + k
Now, impulse inputs are usually modeled in terms of delta functions. Thus,
knowing the Laplace transform of such functions is important when solving
differential equations. The next theorem finds the Laplace transform of the
delta function.
189
Theorem 47.1
With (t) defined as above, if a t0 < b
Z
f (t)(t t0 )dt = f (t0 ).

a
Proof.
We have
Rb
a
Rb
f (t)(t t0 ) = lim0+ a f (t)f (t t0 )dt
R t +
=
lim0+ 1 t00 f (t)dt
= lim0+ 1 f (t0 + ) = f (t0 )
where 0 < < 1 and the mean-value theorem for integrals has been used
Remark 47.1
Rb
Since p (tt0 ) = 1 for t0 t t0 + and 0 otherwise we see that a f (t)(t
Rb
a)dt = f (a) and a f (t)(t t0 )dt = 0 for t0 b.
It follows immediately from the above theorem that
Z
L[(t t0 )] =
est (t t0 )dt = est0 , t0 0.
0
In particular, if t0 = 0 we find
L[(t)] = 1.
The following example illustrates the formal use of the delta function.
Example 47.1
A spring-mass system with mass 2, damping 4, and spring constant 10 is
subject to a hammer blow at time t = 0. The blow imparts a total impulse of
1 to the system, which was initially at rest. Find the response of the system.
Solution.
The situation is modeled by the initial value problem
2y 00 + 4y 0 + 10y = (t), y(0) = 0, y 0 (0) = 0.
190
Taking Laplace transform of both sides we find

2s2 Y (s) + 4sY (s) + 10Y (s) = 1.
Y (s) =
2s2
1
.
+ 4s + 10
The impulsive response is

1
y(t) = L
1
1
2 (s + 1)2 + 22
1
= e2t sin 2t
4
Example 47.2
A 16 lb weight is attached to a spring with a spring constant equal to 2
lb/ft. Neglect damping. The weight is released from rest at 3 ft below the
equilibrium position. At t = 2 sec, it is struck with a hammer, providing an
impulse of 4 lb-sec. Determine the displacement function y(t) of the weight.
Solution.
This situation is modeled by the initial value problem
16 00
y + 2y = 4(t 2), y(0) = 3, y 0 (0) = 0.
32
Apply Laplace transform to both sides to obtain
s2 Y (s) 3s + 4Y (s) = 8e2s .
Y (s) =
e2s
3s
+
.
s2 + 4 s2 + 4
Now take the inverse Laplace transform to get

y(t) = L1 [Y (s)] = 3 cos 2t + 8h(t 2)f (t 2)
where
1
f (t) = L
1
2
s +4
191

=
1
sin 2t.
2
Hence,
y(t) = 3 cos 2t + 4h(t 2) sin 2(t 2) = 3 cos 2t + 4h(t 2) sin 2t
or more explicitly

y(t) =
3 cos 2t,
t < 2
3 cos 2t + 4 sin 2t, t 2
Problem 47.1
Evaluate
R3
(a) 0 (1 + et )(t 2)dt.
R1
(b) 2 (1 + et )(t 2)dt.

R2
cos 2t
(c) 1
(t)dt.
tet
(t + 2)
R 2 2t
(d) 1 (e + t) (t 1) dt.
(t 3)
Problem 47.2
Let f (t) be a function defined and continuous on 0 t < . Determine
Z t
f (t s)(s)ds.
(f )(t) =
0
Problem 47.3
R1
Determine a value of the constant t0 such that 0 sin2 [(t t0 )](t 12 )dt = 34 .
Problem
47.4
R5 n
If 1 t (t 2)dt = 8, what is the exponent n?
Problem 47.5
RtRs
Sketch the graph of the function g(t) which is defined by g(t) = 0 0 (u
1)duds, 0 t < .
Problem 47.6
Rt
The graph of the function g(t) = 0 et (t t0 )dt, 0 t < is shown.
Determine the constants and t0 .
192
Problem 47.7
(a) Use the method of integarting factor to solve the initial value problem
y 0 y = h(t), y(0) = 0.
(b) Use the Laplace transform to solve the initial value problem 0 =
(t), (0) = 0.
(c) Evaluate the convolution h(t) and compare the resulting function with
the solution obtained in part(a).
Problem 47.8
y 0 + y = 2 + (t 1), y(0) = 0, 0 t 6.
Graph the solution on the indicated interval.
Problem 47.9
y 00 = (t 1) (t 3), y(0) = 0, y 0 (0) = 0, 0 t 6.
Problem 47.10
y 00 2y 0 = (t 1), y(0) = 1, y 0 (0) = 0, 0 t 2.
Problem 47.11
y 00 + 2y 0 + y = (t 2), y(0) = 0, y 0 (0) = 1, 0 t 6.
193
48
Numerical Solutions to ODEs: Eulers Method

and its Variants
Whenever a mathematical problem is encountered in science or engineering,

which cannot readily or rapidly be solved by a traditional mathematical
method, then a numerical method is usually sought and carried out. In
this section, we study Eulers method and its variants for approximating the
solution to the initial value problem
y 0 = f (t, y), y(t0 ) = y0 , a < t < b.
(37)
Eulers Method
Eulers method is a very simple numerical method for solving the first-order
initial value problem (37). Eulers method can be intuitively derived from
several different avenues and, for this reason, is often considered to be the
cornerstone method on which the framework to formulate and discuss subsequent methods is built.
The first avenue that we use is a geometrical one, by using the definition of
the derivative. The derivative of y(t0 ) is
y(t0 + h) y(t0 )
.
h0
h
y 0 (t0 ) = lim
Since y 0 (t0 ) = f (t0 , y(t0 )) then
y(t0 + h) y(t0 )
.
h0
h
f (t0 , y(t0 )) = lim
One way of deriving Eulers method comes about by considering the quotient
in this limit to be a good approximation when h is small, that is
f (t0 , y(t0 ))
y(t0 + h) y(t0 )
.
h
(38)
Eulers method is derived from Equation (38) by asserting equality in this

approximation and replacing y(t0 + h) with the approximation y (t0 + h).
That is,
y (t0 + h) y(t0 )
f (t0 , y(t0 )) =
.
h
Rearranging terms yields the relation
y (t0 + h) = y(t0 ) + hf (t0 , y(t0 ))
194
(39)
This is Eulers method approximation to the solution value y(t0 + h). (See
Figure 48.1). By letting t1 = t0 + h and y1 = y (t1 ) the last equation can be
written in the form
y1 = y0 + (t1 t0 )f (t0 , y0 ).
Figure 48.1
Next, we would like to proceed in a similar manner, but we dont have the
value of the solution at t1 and so we wont know the slope of the tangent
line to the solution at this point. However, since we assumed that y1 is a
good approximation to y(t1 ) then we can use that to estimate the slope of
the tangent line at t1 and construct a line through the point (t1 , y1 ) that has
slope f (t1 , y1 ). This gives
y = y1 + (t t1 )f (t1 , y1 ).
Now, to get an approximation to the solution at t = t2 we will hope that this
new line will be fairly close to the actual solution at t2 and use the value of
the line at t2 as an approximation to the actual solution. This gives.
y2 = y1 + (t2 t1 )f (t1 , y1 ).
We can continue in this fashion. Use the previously computed approximation
to get the next approximation. So,
y3 = y2 + (t3 t2 )f (t2 , y2 )
y4 = y3 + (t4 t3 )f (t3 , y3 )
..
.
195
In general, if we have tn and the approximation to the solution at this point,

yn , and we want to find the approximation at tn+1 all we need to do is use
the following
yn+1 = yn + (tn+1 tn )f (tn , yn ).
(40)
Equation (40) is known as Eulers method.

When implementing Eulers method, we will assume that the step sizes between the points t0 , t1 , t2 , to a uniform size of h. In other words, we will
often assume that
tn+1 tn = h.
In this case, Equation (40) can be written as
yn+1 = yn + hf (tn , yn ).
Remark 48.1
One anticipates Eulers method to be more and more accurate by making
h smaller and smaller. This may lead to a significant amount of computation. Therefore, numerical methods are usually programmed and run on a
computer or programmable calculators.
We illustrate Eulers method in the next example.
Example 48.1
Suppose that y(0) = 1 and
method.
Solution.
The step size is h =
0.50
5
k
0
1
2
3
4
5
dy
dt
= y. Estimate y(0.5) in 5 steps using Eulers
= 0.1. The following chart lists the steps needed:

tk
yk
f (tk , yk )h
0
1
0.1
0.1
1.1
0.11
0.2
1.21
0.121
0.3 1.331
0.1331
0.4 1.4641
0.14641
0.5 1.61051
Thus, y(0.5) 1.61051. Note that the exact value is y(0.5) = e0.5 = 1.6487213
196
Figure 48.2
The Improved Eulers Method: Heuns Method
One can derive Eulers method in a different way. Let t0 , t1 , t2 , be as above
but with uniform step size h. We begin by integrating Equation (37) between
tn and tn+1 obtaining
Z tn+1
Z tn+1
0
y (s)ds =
f (s, y(s))ds.
tn
tn
By the fundamental theorem of calculus we can write

Z tn+1
y 0 (s)ds = y(tn+1 ) y(tn ).
tn
Thus,
Z
tn+1
y(tn+1 ) = y(tn ) +
f (s, y(s))ds.
(41)
tn
Computationally, this last equation can not be used since we do not know
f (s, y(s)) for tn s tn+1 . If we estimate the integral by a left Riemann
sum we find
Z tn+1
f (s, y(s))ds hf (tn , y(tn )).
tn
Hence,
y(tn+1 ) y(tn ) + hf (tn , y(tn ))
197
or
yn+1 = yn + hf (tn , yn ).
Therefore, we can view Eulers method as a left Riemann sum approximation
of the integral equation (41).
The Euler method has a serious flaw in its approach to determining the slope
to use in taking each step of the iteration. The iteration step from tn to tn+1
uses only the slope at the endpoint tn of the interval. Unless the function y(t)
is essentially a straight line in the ty-plane, the use of the slope evaluated at
only one endpoint of the step results in using a slope that does not necessarily
represent very well the average slope over the interval. To take the average
slope in consideration, we estimating the definite integral in (41) by using
the trapezoid rule obtaining
Z tn+1
h
f (s, y(s))ds [f (tn , y(tn )) + f (tn+1 , y(tn+1 ))].
2
tn
This leads to the estimate
h
yn+1 = yn + [f (tn , yn ) + f (tn+1 , yn+1 )].
2
The difficulty here is that since we do not know the function y(t) yet, we do
not know the value of yn+1 at the end of the step, so that a scheme based on
taking the average slope over the interval is not directly possible. Note that
yn+1 is defined implicitely so solving for yn+1 requires solving a nonlinear
equation.
Heuns method is based on the evaluation of an estimate of f (tn+1 , yn+1 ) by
replacing the value of yn+1 with an estimate derived from the original Euler
method, with the resulting iteration scheme:
h
yn+1 = yn + h(f (tn , yn ), yn+1 = yn + [f (tn , yn ) + f (tn+1 , yn+1 )]
2
or
h
yn+1 = yn + [f (tn , yn ) + f (tn+1 , yn + hf (tn , yn ))].
2
This equation is known as Heuns method or the improved Eulers
method.
Example 48.2
improved Eulers method.
dy
dt
= y. Estimate y(0.5) in 5 steps using the
198
Solution.
y0
y1
y2
y3
y4
y5
0.50
5
= 0.1 The following chart lists the steps needed:
=
1
h
=
y0 + 2 [f (t0 , y0 ) + f (t1 , y0 + hf (t0 , y0 ))] = 1.105
=
y1 + h2 [f (t1 , y1 ) + f (t2 , y1 + hf (t1 , y1 ))] = 1.221025
=
y2 + h2 [f (t2 , y2 ) + f (t3 , y2 + hf (t2 , y2 ))] = 1.349232625
=
y3 + h2 [f (t3 , y3 ) + f (t4 , y3 + hf (t3 , y3 ))] = 1.490902050625
= y4 + h2 [f (t4 , y4 ) + f (t5 , y4 + hf (t4 , y4 ))] = 1.647446765940625
Thus, y(0.5) 1.647446765940625. Note that the exact value is y(0.5) =

e0.5 = 1.6487213
The Modified Eulers Method
Another numerical integration scheme is the midpoint rule. In this case, we
have

Z tn+1
h
h
f (s, y(s))ds hf tn + , y tn +
.
2
2
tn
This implies that

y(tn+1 ) y(tn ) + hf

h
h
tn + , y tn +
2
2
Using Eulers method we can write

h
h
y tn +
y(tn ) + f (tn , y(tn )).
2
2
Hence,

yn+1 = yn + hf

h
h
tn + , yn + f (tn , y(tn )) .
2
2
This last equation is known as the modified Eulers method.

Example 48.3
modified Eulers method.
dy
dt
= y. Estimate y(0.5) in 5 steps using the
199
Solution.
y0
y1
y2
y3
y4
y5
0.50
5
= 0.1. The following chart lists the steps needed:
=
1
h
=
y0 + hf (t0 + 2 , y0 + h2 f (t0 , y0 )) = 1.105
=
y1 + hf (t1 + h2 , y1 + h2 f (t1 , y1 )) = 1.221025
=
y2 + hf (t2 + h2 , y2 + h2 f (t2 , y2 )) = 1.349232625
=
y3 + hf (t3 + h2 , y3 + h2 f (t3 , y3 )) = 1.490902050625
= y4 + hf (t4 + h2 , y4 + h2 f (t4 , y4 )) = 1.647446765940625
Thus, y(0.5) 1.647446765940625. Note that the exact value is y(0.5) =

e0.5 = 1.6487213
In Problems 48.1 - 48.3, answer the following questions:
(a) Solve the initial value problem analytyically, using an appropriate solution technique.
(b) For the given initial value problem express yn+1 in terms of yn using
Heuns method.
(c) For the given initial value problem express yn+1 in terms of yn using the
Modified Eulers method.
(d) Use a step size y = 0.1. Compute the first three approximations y1 , y2 , y3
using the method in part (b).
(e) Use a step size y = 0.1. Compute the first three approximations y1 , y2 , y3
using the method in part (c).
(f) For comparison, calculate and list the exact solution values y(t1 ), y(t2 ), y(t3 ).
Problem 48.1
y 0 = 2t 1, y(1) = 0.
Problem 48.2
y 0 = ty, y(0) = 1.
Problem 48.3
y 2 y 0 + t = 0, y(0) = 1.
200
Problem 48.4
y 0 = 1 + y 2 , y(0) = 1.
(a) Find the exact solution of the given initial value problem.
(b) Use step size h = 0.05. Compute 20 steps of Eulers method, Heuns
method, and modified Eulers method. Compare the numerical values obtained at t = 1 by calculating the error |y(1) y20 |.
Problem 48.5
y 0 + 2y = 4, y(0) = 3.
(a) Find the exact solution of the given initial value problem.
(b) Use step size h = 0.05. Compute 20 steps of Eulers method, Heuns
method, and modified Eulers method. Compare the numerical values obtained at t = 1 by calculating the error |y(1) y20 |.
In Problems 48.6 - 48.8, the given iteration is the result of Applying Eulers
method, Heuns method, or the modified Eulers method to an initial value
problem of the form
y 0 = f (t, y), y(t0 ) = y0 , t0 t t0 + T.
Identify the numerical method and determine t0 , T, and f (t, y).
Problem 48.6
yn+1 = yn + h(yn + t2n yn3 ), y0 = 1
tn = 2 + nh, h = 0.02, n = 0, 1, 2, , 49.
Problem 48.7

h
h
2
2
sin yn + tn sin yn , y0 = 1
= yn + h tn +
2
2
tn = nh, h = 0.01, n = 0, 1, 2, , 199.

yn+1
Problem 48.8
yn+1

h
h
= yn + h sin tn + + yn + sin (tn + yn ) , y0 = 1
2
2
tn = 1 + nh, h = 0.05, n = 0, 1, 2, , 199.
201

A Second Course in Elementary Differential Equations PDF

Uploaded by

Copyright:

Available Formats

A Second Course in Elementary Differential Equations PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Second Course in Elementary Differential Equations PDF

Uploaded by

Copyright:

Available Formats

A Second Course in Elementary Ordinary

27 nth Order Linear Differential Equations:Existence and Uniqueness

30 Higher Order Homogeneous Linear Equations with Constant

34 First Order Linear Systems: Fundamental Sets and Linear

36 Homogeneous Systems with Constant Coefficients: Complex

41 The Laplace Transform: Basic Definitions and Results

42 Further Studies of Laplace Transform

43 The Laplace Transform and the Method of Partial Fractions155

45 Solving Systems of Differential Equations Using Laplace Transform

47 The Dirac Delta Function and Impulse Response

48 Numerical Solutions to ODEs: Eulers Method and its Variants

Calculus of Matrix-Valued Functions of a

a11 a12 ... a1n

Find a22 , a32 , and a23 .

An m n array whose entries are functions of a real variable defined on

a11 (t) a12 (t) a13 (t)

Solving the two equations in a and c one finds a = 5 and c = 4. Hence,

aik (t)bkj (t)

Show that AB 6= BA. Hence, matrix multiplication is not commutative.

given that ad bc 6= 0. The quantity ad bc is called the determinant of

This implies that

Applying the method of elimination to the first two equations we find

Applying the method of elimination to the last two equations we find

Norm of a Vector Function

Find limt1 A(t).

limt1 (t2 5t) limt1 t3

then A(t) is continuous at t = 1

a11 (t) a12 (t)

a11 (t+h)a11 (t)

a011 (t) a012 (t)

a12 (t+h)a12 (t)

a11 (t) a12 (t) a1n (t)

Finally, we conclude this section by showing that

The Determinant of a Matrix

a11 a12 a13

a11 a12 a13

columns of A. Then det(B) = det(A).

nth Order Linear Differential Equations:Existence

y(t0 ) = y0 , y 0 (t0 ) = y00 , , y (n1) (t0 ) = y0

The following theorem is a generalization to Theorems 3.2 and 15.1

with initial conditions

y(t0 ) = y0 , y 0 (t0 ) = y00 , , y (n1) (t0 ) = y0

has a unique solution in a < t < b.

In this case, we have

the problem as a system:

Note that if y(t) is a solution of (1) then the vector-valued function

is a solution to (3). Conversely, if the vector

which means that x1 is a solution to (1).

Thus, a solution of (5) is also a solution to (3) and vice versa.

[A(s)y1 (s) + b(s)]ds

[A(s)yN 1 (s) + b(s)]ds

This implies that

An easy induction yields that

[yi,k+1 (t) yi,k (t)] + yi,0

Letting u(t) = ||y(t) z(t)|| we have

The coefficient functions are continuous for all t 6= 4 and t 6= (2n + 1) 2 .

The General Solution of nth Order Linear

In this section we consider the question of solving the homogeneous equation

So the problem reduces to finding the constants c1 , c2 , , cn . These are