Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Beginning with your liN algehra course you have encountered problems such as lhc
following:
and/or equations in the system-the more difficult !t often is to s�! ve the system. For
instance, suppose we needed to find the partial l'ract1on dccompos11Ion of
Two systems of equations are said to be equh•alent if they have the same solutions.
It is not difficult to see that applying an elementary operation to a system produces an
equivalent system.
I Named in honor of Karl Friedrich Gauss ( 1777-1855>. who is one of the greatest mathematicians of all time
and is often referred to as the "prince of mathematics," and Wilhelm Jordan l I S,l2-1899), a German engineer.
Chupfer I Mah,iccs and Det.erm.inants
13
z =-
3,
whid1 tcrb us the solution.
1.1 Systems of Linear Equations 5
You might notice that we only really need to keep track of the coefficients as we
transform our system. To keep track of them, we will indicate a system such as
x-y+z=O
2x - 3y + 4z = -2
J
-n
-2x-y+z=7
by the following array of numbers:
This array is called the augmented matrix for the system. The entries appearing to the
left of the dashed vertical line are the coefficients of the variahles as they appear in the
system. This part of the augmented matrix is called the coefficient matrix of the system.
The numbers to the right of the dashed vertical line are the constants on the right-hand
side of the system as they appear in the system. ln general, the augmented matrix for
the system
i:J
is
a2n
am2
The portion of the augmented matrix to the left of the dashed line with entries aii is the
coefficient matrix of the system.
Corresponding to the elementary operations for systems of equations are elementary
row operations that we perform on the augmented matrix for a linear system. These are
as follows.
1. Interchange two rows. 2
2. Multiply a row by a nonzero number.
3. Replace a row by itself plus a multiple of another row.
2 A line of numbers going across the matrix from left to right i� called a ro\l; a line of numbers going down
the matrix is called a column.
6 Chapter I Matrices and Determinants
As our first formal example of this section, we are going to redo the work we did in
solving the system
x-y+z=O
2x - 3y + 4z = -2
-2x-y+z = 7
with augmented matrices.
S0lutio11 Our work will consist of four steps. In the first step, we shall use the first row and row
operations to make all other entries in the first column zero. In the second step, we shall
use the second row to make al I other entries in the second column zero. In the third step,
we shall use the third row to make all other entries in the third column zero. In the fourth
step. we shall make the nonzero entries in the coefficient matrix 1 at which point we will
be able to read off our solution. To aid you in following the steps, an expression such as
R1 - 2R 1 next to the second row indicates that we are replacing the secon<l row by itself
-n
plus -2 times the first row; an expression such as R 1 /3 next to the first row indicates
we are dividing this row by 3. Arrows are used to indicate the progression of our steps.
[
-]
2 -3 4 R2 - 2R,
--n
2 -I I R3 + 2R,
-
I
-
-1 I R 1 - R2
u
-I 2
[ � -3 3 I R3 - 3R2
-[ i -I
0 -1
0 -3
2
I
I
-q 13
3R 1 - R3
3R2 + 2R3
0
-3
0 -3
0
0 20
-
R 1 /3
-Ri /3
137] -R3/3
[ I O O •
- 0 I O i -20/3
0 0 I : -13/3
•
-7/3]
. The solution is then x = -7/3, y = -20/3, z = -13/3.
1.1 Systems of Linear Equations 7
Looking back at Example I, you will see that the coefficient matrix in our final
augmented matrix is in reduced row-echelon form. Once we have the coefficient matrix
in reduced row-echelon form, the solutions to the system are easily determined.
Let us do some more examples.
X1 +Xz - X3 + 2X4
=1
.Xt +x2 +x4 = 2
x1 + 2x2 - 4x3 = l
2x, +x2 + 2x3 + 5x4 = I.
ll
Solution We try to proceed as we did in Ex.ample I. Notice, however, that we will have to modify
our approach here. The symbol R2 tt R3 after the first step is used to indicate that we
are interchanging the second and third rows.
[i
-1 2
0 I 2 R2 - R,
2 -4 0 J R3 - R1
=� ' _; l
2 5 1 R4 -2R,
-1 2
:
I
-3
4
-1 2
-3 -2 I
-1
4
.1.1 Systc, ms of Linear Equations 9
where z is any real number. In particular, we have infinitely many solutions in this
example. (Any choice of z gives us a solution. If z = 0, we have x = -3, y = 3, z =0
as a solution; if z = 1, we have x = 5, y = -2, z = 1 as a solution; if z = ./I7, we
have x = -3 + 8,,/17, y = 3 - SJ17, z = Jf7 as a solution; and so on.) In a case
such as this, we refer to z as the free var'iab1e and x and y as the dependent variables
in our solutions. When specifying our solutions to systems like this, we will follow the
convention of using variables that correspond to l, eading ones as dependent variables and
those that do not as free variables. It is not necessary to specify our solutions this way,
however. For instance, in this exampile we could solve for z in terms of x, obtaining
X 3
z=- +-
8 S
and
giving us the solutions with x as the free variable and y and z as the dependent variables.•
Solution W.e again begin by reducing the augmented matrix to the point where its coefficient
matrix. is in reduced row-echelon form:
[:
-8 -1 3 ' 0 ]
-IO -1 2 3 \ 0 4R2 -5Ri
-6 -I 2 , 0 4R3 - 3R1
10 Chapter I Matrices and Determinants
� [ O� -�O -1
-�
[ 4 -8 0 4 0
-+ 0 0 J 3 -3 I
0 0 0 4 -4 I
-2 0 I O '�0]
0 I 3 -3
0 0 -I
l -2 0 0 l I
Q
-+ [ 0 0 I 0 0 0 ].
0 0 0 I -1 0
We now have arrived at the equivalent system
X( - 2X2 + X5 = Q
X3=Q
X4 -x5 = Q,
which ha1, solutions
l
o rows than columns
reduced row-echelon fonn. Why when the coefficient
is this the case?) If matrix is in
\ eol,mos, as we hd ; ther e are fe.wer nonze.ro
" Examples 3 aod 4, the rows than
) If we h,,c as many oonrem system w;u ha,e ;ofio;,
rows aseol,mos, as e1y many soMions.
one solution. Recall that it was occurr ed io Exarople
mentioned at the begi l , we ha,e exactly
system of linear equations either nning of this section
has exactly one solu that e.very
no solutions. Now we can see tion, infinitely many
why this is true. solutions, or
w1•rr..-. ·-
1.1 Systems of Linear Equations 11
x, =0, X11 =0
is a solution to the homogeneous system in Equations (] ). This is called the trivial
solution of the homogeneous system. Because homogeneous systems always have a
trivial solution, they are never inconsistent systems. Homogeneous systems will occur
frequently in our future work and we will often be interested in whether such a system
has solutions other than the trivial one, which we naturally call nontrivial solutions.
The system in Example 4 has nontrivial solutions. For instance, we would obtain one
(among the infinitely many such nontrivial solutions) by letting x2 = I and x 5 = 2, in
which case we have the nontrivial solution x, = 0, x2 = I, X3 = 0, X4 = 2, x5 = 2.
i Actually, we can tell ahead of time that the system in Example 4 has nontrivial solutions.
Because this system has fewer equations than variables, the reduced row-echelon form of
I
the coefficient matrix will have fewer nonzero rows than columns and hence must have
infinitely many solutions (only one of which is the trivial solution) and consequently must
have infinitely many nontrivial solutions. This reasoning applies to any homogeneous
system with fewer equations than variables. and hence we have the following theorem.
THEOREM 1.1 A homogeneous system of m linear equations in II variables with m < 11 has infinitely
many nontrivial solutions.
Of course, if a homogeneous system has at least as many equations as variables such
as the systems
2x+y+z=0
x+y+z=O
X - 2y - Z = 0
x - y- z = 0 anJ
3x -y = 0
2x+y+z=0
4x -3y- z = 0
we would have to do some work toward solving these systems before we would be able
to see whether they have nontrivial solutions. We shall do this for the second system a
bit later.
Gaussian elimination, which is another systematic approach for solving linear
systems, is similar to the approach we have been using but does not require that all the
other entries of the column containing a leading I be zero. That is, it uses row operations
to transform the augmented matrix so that the coefficient matrix has the following fonn:
1. Any zero rows appear at the bottom.
12 Chapter I Matrices and Determinants
u -n
Such a form is called a row-echelon form for the coefficient matrix. In essence, we do
not eliminate (make zero) entries above the leading Is in Gaussian elimination. Here is
how this approach can be applied to the system in Example l.
I
-1 I
u
-3 4 R2 -2R 1
-I R3 + 2R 1
-1 : 0]
� -I 2 -2
-3 3 • 7 R3 - 3R2
-�]
I I
I -I I -I
� [ 0 -I 2 I
-R2 � [ 0 1 -2 I
0 0 -3 13 -R3/3 0 0 I I
-13/n
We now have the coefficient matrix in a row-echelon form and use this result to find the
,olutions. The third row tells us
13
z=-3·
The value� of the remaining variables are found by a process called back substitution.
From the �ccond row, we have the equation
y-2z=2
from which we can find y:
26
v+-=2
. 3
20
y=--
3'
Finally. the first row represents the equation
x-y+z==O
from which we can find x:
20 13
x+--- -O
3 3 -
X == --
3
1.1 Systems of Linear Equations 13
On the plus side, Gaussian elimination requires fewer row operations. But on the minus
side, the work is sometimes messy when doing the back substitutions. Often, we find
ourselves having to deal with fractions even if our original system involves only integers.
The back substitutions are also cumbersome to do when dealing with systems that have
infinitely many solutions. Try the Gaussian elimination procedure in Example 3 or 4 if
you would like to see how it goes.
As a rule we will tend to use Gauss-Jordan elimination when we have to find the
solutions to a linear system in this text. Sometimes, however, we will not have to
completely solve a system and will use Gaussian elimination since it will involve Jess
work. The next example illustrates an instance of this. In fact, in this example we will
not even have to bother completing Gaussian elimination by making the leading entries
one.
I ]
has solutions.
[: :] -[ i
Solutio11 We begin doing row operations as follows.
-1 2 -I 2
-1 3 -5 b-2:
]
I
I
R2 - 2R1 I
I
2 -3 I R3 - R1 3 -5 , c-a R3 - R2
-[:
-1 2 a
3 -5 b - 2a
0 0 a-h+c
Now we can see that this system has solutions if and only if a, b, and c satisfy the
•
equation
a -b +c = 0.
Another place where we will sometimes use an abbreviated version of Gaussian
elimination is when we are trying to see if a homogeneous system has nontrivial solutions.
t] i]
S0l11tio11 Perform row operations:
1 I I
� -2 -1 ' 2R2 - Ri -5 -3
l
[ 2R3 - 3R1 -5 -3 R3 - R2
3 -1 ()
4 -3 -1 R4 - 2R1 -5 -3 R4 - R2
:
[�
J 1 0
-5 -3 I
0
-+
I
'
Q
I
() 0 I
0 0 0
It is now apparent that this system has nontrivial solutions. In fact, you should be able
to sec this after the first set of row operations. •
It is not difficult to write computer programs for solving systems of linear equations
using the Gauss-Jordan or Gaussian elimination methods. T hus it is not surprising that
there are computer software packages for solving systems of linear systems. 3 Maple is
one among several available mathematical software packages that can be used to find
the solutions of linear systems of equations..
In the preface we mentioned that we will use Maple as our accompanying software
package within this text. The use of Maple is at the discretion of your instructor. Some
may use it, others may prefer to use a different software package, and yet others may
choose to not use any such package (and give an excellent and complete course). For
those intructors who wish to useMaple-or for students who are independently interested
in gaining some knowledge of its capabilities-we will include occasional remarks about
how to use it when we deem it appropriate. On many other occasions we will not include
any remarks and will simply provide some exercises asking you to use indicated Maple
commands. In these cases, you are expected to look up the command in the Help menu
under Topic Search to see how to use it. This is one place where we will include a few
remarks to get you started. For those who wish to use the software packagesMathematica
or MATLAB, the accompanying Technology Resource Manual contains corresponding
command!) for these software packages.
Here we explain how to use Maple to find solutions to linear systems. One way to
do this is to use the /in.wive command. To use this command in a Maple worksheet, you
will first have to load Maple's linear algebra package by typing
with(linalg);
at �he comman d �rompt > and then hitting the enter key. After doing this, you will get
,
a hst ofM �ple s hn�ar algebra commands. To solve the system in Example 1, first enter
the coefficient matnx of the system by typing
A: = matrix ( [ ( l, -1, l] , [ 2 , -3 , 4) , [ -2 , -1, 1] ] ) ;
·3 Often the�e packages employ methods that are more efficient than Gauss ·
, -Jordan or Gaussian · · ·
. . . e1 1m111auon,
,. not concern ourselves wnh
but we 11,111 these issues m this text.
1.1 Systems of Linear Equations 15
at the command prompt and then hitting the enter key. (The symbol := is used in Maple
for indicating that we are defining A to be the coefficient matrix we type on the right.)
The constants on the right-hand side of the system are typed and entered as
b:=vector([0,-2,7]);
at the command prompt. Finally, type and enter
linsolve (A,b);
at the commmand prompt and Maple will give us the solution as
4
-7 -20 -13
[ ]
3'_3_'_3_
Doing the same set of steps for the system in Example 2 results in no output, indicating
there is no solution. Doing them in Example 3 yields the output
EXERCISES 1.1
4 Software packages such as Maple often will have several ways of doing things. This is the case for solving
systems of linear equations. One variant is to enter b as a matrix with one column by typing and entering
b: %matrix ( [ [OJ , (-2) , [7) J ) ;
When we then type and enter
linsolve(A,b);
our solution is given in column form. Another way is to use Maple's solve command for solving equations
and systems of equations. (With this approach it is not necessary to load Maple's linear algebra package.) To
do it this way for the system in Example l, we type and enter
solve({x-y+z�0,2*x-3*y+4•z=-2,-2*x-y+z�7),{x,y,z});
16 Chaptt>r I Matrices and Determinants
J.l X - 2y =2 14. 2x + 3y = 5 3x - y - 2z =0
X + l!y =-4 2x + y = 2 2x-2y - 4z = 0
2x + y = I x -2y = I X + 3y + 6z = 0
15. 2x1 -x2 -x, +x4+xs =0 25. We have seen that homogeneous linear systems with
fewer equations than variables always have infinitely
x1 - X2 + x, + 2x4 - 3x5 = 0
many solutions. What possibiIi ties can arise for t�on
3.r 1 - 2x2 - x, - X4 + 2xs = 0 homogeneous linear systems with fewer equat10n�
16. x1 -3x2+x3-X4-X5=I than variables? Explain your answer.
2x, + X2 - X3 + 2x4 + X5 = 2 26. Give an example of a system of linear equations with
-X1 + 3x2 -X3 - Lt4 -X5 = 3 more equations than variables that illustrates each of
the following possibilities: Has exactly one solution.
2.tt + X2 - X3 - X4 - X5 = 6 has infinitely many solutions, and has no solution.
Determine conditions on a. h, and c so that the systems 27. Describe graphically the possible solutions to a sys
of equations in Exercise!, 17 and 18 have solutions. tem of two linear equations in x, y, and z.
17. 2x - y + 3z = a 18. x + 2y - z = a 28. Describe graphically the possible solutions to a sys
X - 3)' + 2;: = h + )' - 2z = b tem of three linear equations in x, y, and z.
X
X + 2y + Z = C 2x + y-3z = c Use Maple or another appropriate software package to
solve the systems of equations in Exercises 29-32. If
Determine condition� on a. h. c, and d so that the .�ys you are using Mathematica or MATLAB, see the Tech
tem� of equation� in Exercises 19 and 20 have solutions.
nology Resource Manual for appropriate commands.
19. Xt + X2 + X3 - X4 = Cl (To become more comfortable with the software pack
.t1 - X2 - X3 + .1"4 = /J age you are using. you may wish to practice using it
Xt + Xz + X3 + X4 = C
to solve some of the smaller systems in Exercises 1-16
hefore doing these.)
X1 - Xz + X3 + X4 =d
29. 7x 1 - 3x2 + 5x3 - 8x4 + 2 x5 = 13
20. X1 -X2 + X3 + X4 = ii
12x, +4x 2 - 16x3 - 9x4 + 7x5 = 21
xi + xi - 2x3 + 3x4 = h
-22x1 - 8x2 + 25x3 -16x4 - 8x5 =4 7
3x1 -2x2 + 3x3 -2.r4 = c
-52x1 - 40x2 + l18x3 - 37x4 - 29x5 = 62
2x2 - 3x3 + 2x4 = d
30. 46x1 + 82x2 - 26x3 + 44x4 = 122
Determine if the homogcneou� �ystems of linear equa 69x1 + IOlx2 + 43x3 + 30x4 = 261
tion� in Ewrci!-.e, 21-24 have nontrivial solutions. You -437x1 -735x2 + 335x3 + 437x4 = -406
do not have to solve the systems.
299x1 + 379x2 - 63!x3 - 2 50Jx4 = -4146
21. 9x-2y+lh=O
1863x1 + 2804x2 + 62x3 - 1983x4 = 4857
13x + I! I y - 27z = 0
1748x1 + 229 Ix2 - 46 lx3 - 9863x4 = 4166
\
1.2 Matrices and Matrix Operations 17
In the previous section we introduced augmented matrices for systems of linear equations
as a convenient way of representing these systems. This is one of many uses of matrices.
Tn this section we will look at matrices from a general point of view.
We should be explicit about exactly what a matrix is, so let us begin with a definition.
A matrix is a rectangular array of objects called the entries of the matrix. (For us, the
objects will be numbers, but they do not have to be. For example, we couId have matrices
whose entries are automobiles or members of a marching band.) We write matrices down
by enclosing their entries within brackets (some use parentheses instead) and, if we wish
to give a matrix a name, we will do so by using capital letters such as A, B, or C. Here
are some examples of matrices:
B = [ -7 4 4 0 3 ],
ln2
-I
-391/6298 l
18 Chapter I Matrices and Determinants
Augmented matrices of systems of linear equations have these fonns if we delete the
dashed line. In fact. the dashed line is included merely as a convenience to help distin
guish the left- and right-hand sides of the equations. If a matrix has m rows (which go
across) and II columns (which go up and down), we say the size (or dimensions) of the
matrix is (or are) 111 x 11 (read "m by n"). Thus, for the matrices just given, A is a 2 x 3
matrix. B is a I x 5 matrix, C is a 4 x I matrix, and Dis a 4 x 4 matrix. A matrix such
as B that has one row is called a row matrix or row vector; a matrix such as C that has
one column is called a column matrix or column vector. Matrices that have the same
number of rows as columns (that is, n x n matrices) are called square matrices. The
matrix Dis an example of a square matrix.
As you would expect, we consider two matrices A and B to be equal, written A = B,
if they have the same size and entries. For example,
while
[
-
l
I� J=[ -I 8/4
2-1 3.4 J
[ -: 1� J I[ : I�
J
and
[�]
1[1 2 ].
A=[aij ],
.
If we wish to single out the ij-entry of a matrix A, we wt·11 wnte
-n
ent ij (A).
For instance. if B is the matrix
-\ 2
5
•=[
4
3 -4
then
ent23(8) = -9.
1.2 Matrices and Matrix Operations 19
If A = [aii] is an n x n matrix, the entries a 11, a22, ... , a,, are called the diagonal
11
MmxnOR).
and
i n this book. More generally, the set of n x l column matrices Mn x 1 (IR) will be denoted
IR" and we will refer to the elements of Rn as vectors in ][{n or n-dimensional vectors.
We next turn our attention to the "arithmetic" of matrices beginning with the op
erations of addition and a multiplication by numbers called scalar multiplication.6 If
A and B are matrices of the same size, we add A and B by adding their corresponding
5 In set notation, the vertical bar, I, denotes "such that" (some use a colon,:, instead of a vertical bar) and the
symbol E denotes "element of' (or "member of'). One way of reading
such that a11, a12, a21, a12 are elements of the set of real numbers."
6 These two operations are extensions of the ones you already know for vectors in JR2 or R3 to matrices in
general.
20 Chapter I Matrices and Determinants
A+ B = [aij + bij]-
For instance, if
then
1+8 2 1
+9
A+B= 3+10 4tl l = 139 151 .
[ ] [ ]
5+12 6+13 17 19
Note that we have only defined sums of matrices of the same size. The sum of matrices
of different sizes is undefined. For example, the sum
[-23 5]
I +[ 3 0 -2 ]
\ cA = c[aij] = [caij] .
For example,
THEOREM 1.2 If A, 8, and C are matrices of the same size and if c and dare scalars, then:
l. A+B = B + A (commutative law of addition).
2. A + (B + C) = (A + B) + C (associative law of addition).
3. c(dA)= (cd)A.
4. c(A+B)=cA+cB.
5. (c + d)A = cA + dA.
1.2 Matrices and Matrix Operations 21
Proof We prove these equalities by showing that the matrices on each side have the same
entries. Let us prove parts (I) and (4) here. The proofs of the remaining parts will be
left as exercises (Exercise 24). For notational purposes, we set
A= [aij] and B= [bij ].
Part (1) follows since
entij (A + 8) = aii + bij = bij + aij = entij(B + A).
To obtain part (4),
ent;j (c(A + B)) = c(aij + bij ) = caii + cbi j = entij (cA + cB). •
One special type of matrix is the set of zero matrices. The m x n zero matrix,
denoted Omxn , is them x n matrix that has all of its entries zero. For example,
and 04x3 = [ � � � ] ·
0 0 0
0 0 0
Notice that if A is an m x n matrix, then:
1. A+ Omxn = A.
2, O·A = Omxn ·
We often will indicate a zero matrix by simply writing 0. (To avoid confusion with the
number zero, we put this in boldface print in this book.) For instance, we might write
the first property as A+ 0= A. The second property could be written as O · A = 0.
The negative of a matrix A = [aij ], denoted -A, is the matrix whose entries are
the negatives of those of A:
-A= [-aiJ],
Notice that
-A= (-l)A and A+ (-A)= 0.
Subtraction of matrices A and B of the same size can be defined in terms of adding the
I I
negative of B:
A - B =A+ (-B).
Of course, notice that A - B could also be found by subtracting the entries of B from
the corresponding entries of A.
Up to this point, alI of the operations we have introduced on matrices should seem
relatively natural. Our final operation wi!J be matrix multiplication, which upon first
glance may not seem to be the natural way to multiply matrices. However, the manner of
multiplying matrices you are about to see is the one that we will need as we use matrix
multiplication in our future work.
-·
A= [aij] is an m x n matrix
Here is how we do matrix multiplication: Suppose that
· nd B is defined to be the � x I
and B == ( b,j] is an II x I matrix. The product of A a
matrix
[AB== [pij] l
where
In other words, for each I :::: i :::: m an d 1 :::: j :::: l the ij-entry of AB is found by
multiplying each entry of row i of A times its corresponding entry of column j ofBand
then summing these products.
Here is an example illustrating our matrix multiplication procedure.
a
EXAMPLE l Find the product AB for
A� ;: aad B�
[ ] [�
= 1·5+2·7 1-6+2·8
]-
19 22
[ 3 · 5 +4 ·7 3 · 6 + 4 . 8 -[ 43 50
.J •
Once you practice t�is sum of row entries times column entries a few times, Y ou
should find yourself gettmg the hang of it.7 Let us do anothe1. examp 1e of matnx
· mu I t1-
plication.
·
-:1-
EXAMPLE 2 Find the product CD for
-
C� [ :
-: ] and D
� [ -;
J[-� -�]
Solution
CD-[-:-!
--[
(-1)·1+2(-3)-3·1 (-1)(-2)+2· 4-3-J
0 · I - 1 ( -3) + 1 · I 0( -2) - 1 · 4 + I · 1 ] =
-[ 1� -�]
4-1 +2 (-3)- 1· l 4(- 2)+2-4-1-1
Notice that for the product AB of two matrices A and B to be defined, it is necessary
-3
• -1
that the number of columns of A be the same as the number of rows of B. If this is not
the case, the product is not defined. For instance, the product DC for the matrices in
Example2 is not defined. In particular. CD is not the same as DC. This is an illustration
of the fact that matrix multiplication is not commutative; that is, AB is not in general the
same as BA for matrices A and B. Sometimes these products are not the same because
one is defined while the other is not, as the matrices C and D illustrate. But even if hoth
products are defined, it is often the case that they are not the same. If you compute the
product BA for the matrices in Example 1, you will find (try it)
3 34
- [ 2
31 46 J '
BA-
which is not the same as AB. 8 In the case when AB = BA for two matrices A and B,
we say A and B commute.
While matrix multiplication is not commutative, some prope1ties that you are used
to having for multiplication of numbers do carry over to matrices when the products are
defined.
THEOREM 1.3 Provided that the indicated sums and products are defined, the following properties hold
where A, B, and C are matrices and dis a scalar.
1. A(BC) = (AB)C (associative law of multiplication)
2. A(B + C)=AB+ AC (left-hand distrihutive law)
3. (A+ B)C = AC+ BC (right-hand distributive law)
4. d(AB) = (dA)B = A(dB)
We wiJI prove the first two parts here and leave proofs of the remaining parts as exercises
(Exercise25). For notational purposes. suppose
Proof
8 This is not the first time you have encountered an e,camplc of a noncommutative operation. Composition
of functions is noncommutative. The cross product of two three-dimensional vectors is another example of a
noncommutative operation.
2.J Chaptl'r 1 Matrices and Determinants
C.
some notation for the sizes of A, B, and
To prove part (I). we also have 10 introduce x matr ix. Both
x l matrix, and C is an l h
Suppose A is an III x II matrix, 8 is an 11
?) To see that these products are the same,
1
A(BC) and (A B)C are 111 x h matr ices. (Why
n (/ ) II(/
is
we work out the ij-entry of each. For A(BC), this
t. t. (t
Carrying out the same steps for (AB)C,
Since the summations over k and q are interchangeable, we see that the ij-entries of
A(BC) and (AB)C are the same and hence A (BC)= (AB)C.
To prove part (2). we again introduce. notation for the sizes of our matrices. Suppose
A is an m x II matrix and 8 and C are n x I matrices. Both A (B + C) and AB + AC
are III x I matrices. We have
= L(a;kbki +a;kCkj)
k=I
and
If
L(a,kb j +a,kcki ).
II
= k
•
k=I
L_
rn
1.2 Matrices and Matrix Operations 25
l
OJ 1X1
A-[
a11 a,2 a,11
a21 a22 a211
B=[�l- bm
Observe that our system can then be conveniently written as the matrix equation
AX= B.
For instance, the system
2x -y +4z = I
x-7y+z=3
-x+2y+z=2
26 Chapter I latriccs and Determinants
would be written
as a matrix�quation. Notice that a homogeneous linear system takes oo the form AX=
0 when written as a matrix equation.
EXERCISES 1.2
In E,crci�c� 1-18, either perfom1 the indicated opera Write the matrix equations as systems of equations in
tion� or ,talc that the expression is undefined where A, Exercises 21 and 22.
21.[� -; 1:
-
-I
5 J•
D::: [
3
0
-I
I
J' H�
23. Suppose that A and B are n x II matrices.
a) Show that (A+ 8) 2 = A1 +AB+ BA+ 82 .
b) Explain why (A+ 8) 2 is not equal to
A2 + 2AB+ 8 2 in general.
24. Prove the following parts of Theorem 1.2.
I. A + B 2. D-C 3. 28 a) Part (2)
.a. -iF 5. A-48 6. 3D +2C b) Part (3)
7. CD 8. DC 9. EF c) Part (5)
JO.FE 11. AE 12. EA 25. Prove the following parts of Theorem 1.3.
13. (£ + f)A 14. B(C + D) 15. 3AC a) Part (3)
b) Part (4)
16. F(-28} 11. c2 18. A3
26. Suppose A is an m x n matrix and B is an II x I
Write the �y,tems of equations in Exercises 19 anl.i 20 matrix. Further, suppose that A has a row of zeros.
in the matrix form AX = B. Does A _B have a row of zeros? Why or why not?
Does th is also hold if B has a row of zeros? Why or
2x - y + 4z = I
why not?
x+y-z=4
27· Sup�ose A is an m x n matrix and B is an n x I
y + 3z = 5 matnx. Further, suppose that B has a column of ze
X + J = 2, ros. Does AB have a column of zeros? Why or why
not? Does this also hold if A has a column of zeros?
xi - 3x2 + x3 - 5x4 = 2
Why or why not?
XJ +X2 - X3 + X4 ::: I
28. Give an example of two matrices A and B for which
x, - x2 -XJ + 6x4 ::: 6 AB= 0 with A -:j:. 0 and B -:j:. 0.
1.2 Matrices and Matrix Operations 27
29. a) Suppose that A is the row vector The matrix command introduced in the previous section
is one way of entering matrices on a Maple worksheet.
A= ( a1 a2 an
Maple uses the evalm command along with+, - , *, &*,
and B is an n x'l matrix. View Bas the column and I\ to find sums, differences, scalar products, matrix
of row vectors products, and matrix powers, respectively. For instance,
to find A - B + 4C 4- AB - C 3 where A, B, and C
are matrices already entered on a Maple worksheet, we
would type and enter
ev�lm(A-B+4*C+A&*B-CA3);
r�r:; n
at the command prompt. A scalar product cA also may
where B1 , B2 , .•• , B,, are the rows of B. Show be found with the scalarmul command by typing
that scalarmul (A, c) ;
AB = a1B1 + a2B2 + · · · + a,,B,,. at the comm,and prompt. Products of two or more ma
:\ U�: th
< �•: �
I trices can be found by using the multiply command. For
of�: :
(
:
instance, typing and entering
A multiply(B,A,C);
will give us the product BAC. Use these Maple com
30. a) Suppose that B is the column vector mands or appropriate commands in another suitable soft
ware package (keep in mind that corresponding Math
ematica and MATLAB commands 1can be found in the
b,, 4 -2 16 27 -11
9 43 9 -8 -1
and A is an m x n matrix. View A as the row of
column vectors A= 34 20 -3 0 21
A= [ A1 A2 A" ] -S 4 4 7 41
0 12 -2 -2 3
where A 1• A2, .•. , An are the columns of A.
Show that 0
0 0
AB= h1A1 + b2A2 + · · · + h11 A,, . 2 2 2 0 0
b) Use the result of part (a) to find AB for B= 0 3 3 3 0 ' and
0 0 4 4 4
0 0 0 s s
-: ]
c-[
(
-2 0 3
31. The trace of a square matrix A, denoted tr(A), is 0 2
the sum of the diagonal entries of A. Find tr(A) for 3 -1 -1
=[� -
- -[ 0 2 2 -3
A t� : ].
2 10 3 in Exercises 33--40.
32. Prove the following where A and R are square ma 33. A-2B 34. SA+ 6C
trices of the same size and c is a scalar. 35. ABC 36. CB +c
a) tr(A + B) = tr(A) + tr(B)
=
37. (A+ B)2 38. 4A +CB
b) tr(cA) = c tr(A)
c) tr(AB) tr(BA) 39. 4C A - SC B - 2C 40. B2 - 4AB + 2A2
28 Chapter l l\latriccs and Determinants
I I
II
AB= BA= I
B = [-35 -12]
a� an inverse si nee
and
BA=
[ -5
(How we obtain B will be seen later.)
3 _: l [ : : ]- [ � � ] .
Not all square matrices have inverses. Certainly
, square zero matrices do not have
inverses. (The products OB and BO are O, not
I.) But even a nonzero square matrix
may fail to have an inverse. As a simple example,
the matr ix
B=[a b],
C d
we have
AB= [ I O ][ a h]= [ a
0 0 c d o b
o]1[ I o O ].
Square matrices that have inverses
arc called invertible or
that do not have inverses are called nonsingular matrices; those
noninvertible or singu lar
When a matrix has an inverse, it matrices.
has only one inverse.
THEOREM 1.4 If A is an invertible matrix, then the
inverse of A is unique.
1.3 Inverses of Matrices 29
Proof Suppose that A did have two inverses Band C. Consider the product BAC. Ifwe group
B and A together,
BAC=(BA)C=IC=C
sinee BA = I. If we group A and C together,
BAC = B(AC) =Bl=B
since AC= 1. Thus,
•
The uniqueness ofthe inverse of an invertible matrix A allows us to speak of the
inverse ofA rather than an inverse of A. It also allows us to introduce a symbol for the
inverse of A. Henceforth we shall denote the inverse ofA by
in much the same manner as we use the exponent -1 for denoting inverses offunctions.9
Let us now turn our attention to a method for finding inverses of square matrices.
Consider again the matrix
x,2 + 2x22 = 0
3x12 + 5x22 = 1.
9 Do note that A -t does not stand for I/A any more than sin-1 x stands for I/ sin .r: indeed writing I/ A for
a matrix A amounts to writing nonsense.
30 Chapter l Matrices and Determinants
We then will have a unique matrix A-1 so that AA-1 = I if and only if each of these
systems of equations has a unique solution (which occurs if and only if the reduced
row-echelon form of A is/). Let us solve these systems to see if this is the case. To save
writing, notice that since both of these systems have the same coefficient matrices, any
set of row operations that leads to the solution of one system leads to the solution of the
other system too. Thus we can simultaneously solve these two systems by forming the
augmented matrix
. I 2 : 1 0
[Alli= [ ]
3 5 O 1 I
and then using row operations to reduce its left-hand portion A to reduced row-echelon
form (which will be /):
1 2 : I O 1 0 : -5
� [ I
] - [ I
0 -1 , -3 I O 1 , 3
(We have not indicated the row operations here. Can you determine the ones we used?)
The right-hand portion of our final augmented matrix tells us xi 1 = -5, x21 = 3,
x12 = 2, X22 = -1. and hence
-5
A-I = [ 2 ].
3 -1
We must be honest, however. There is a gap in our development here. The procedure
we have just illustrated produces a matrix B so that AB = I. (We describe this by saying
Bi� a right-hand inverse.) But the inverse of A must also have the property that BA = I.
(When BA = I, we say B is a left-hand inverse.) You can check that the right-hand
inverse we have just found for the given 2 x 2 matrix A is also a left-hand inverse and
hence is A- 1• Shortly we will fill in this left-hand inverse gap. Once we do so, we then
will have that the inverse of a square matrix A (if any) can be found by the following
procedure:
l. Form the augmented matrix [A I/] where I is the identity matrix with the same
size as A.
2. Use row operations to reduce the left-hand portion into reduced row-echelon
fonn.
3. If the augmented matrix after step 2 has the form [/I BJ, then B = A-l; if it
does not have this form (or equivalently, if the reduced row-echelon form of A
contains a zero row), A does not have an inverse.
Examples I and 2 illustrate our procedure for finding inverses.
EXA'.\-1PLE I If po�sible, find the inverse of the following matrix.
1.3 Inverses of Matrices 31
n
Solution We first form the augmented matrix and then apply row operations:
[;
I 3 0 OJ [2 1 3 0
u
l 0 IO -.1- 00 -2 -1 I
5 I 0 0 1 0 3 -5 -2 0
4u n 4
-i]
3 : 0 0 14 : 5 0
3 -5 -2 0 3 -5 -2 0
0 -2 I -I 0 -2 -1
[ � =; _: ]
Solution We again form the augmented matrix and apply row operations:
[ I -2 2 : J OO I -2 2 : I OO
2 -3 I !0 I O ] -.1- [0 I -3 I -2 I O ] .
I -I -1 , 00 l O I -3 , -I O I
At this point it is apparent that the left-hand portion cannot be reduced to I and hence
the matrix in this example does not have an inverse. •
When the inverse of a square matrix A is known, we can easily find the solutions to
a system of linear equations
AX=B.
:]
Solutio 11 From Example I, we have that the inverse of the coefficient matrix
A-P 1
4 5
of this system is
[-1/3 7/6 -1/6]
1/6 -5/6 1/3
[]
1/2 -1/2 0
The solution is then given by
X [-1/3 7/6 -1/6 6 -33/2
X= y =A-18= 1/6 -5/6 1/3 ][ -12 ] = [ 12 ] ;
•
z 1/2 -1/2 0 3 9
that is, .x = -33/2, y = 12, z = 9.
The following theorem gives us a characterization of when a system of n linear
equations in II unknowns has a unique solution. ,.
THEO REI\I 1.5 A system AX = B of n linear equations inn unknowns has a unique solution if and only
if A is invertible.
Proof If A is invertible. the solutions to the system are given by X = A-1 B and hence are
unique. Conversely, suppose AX = B has a unique solution. Considering the result of
Gauss-Jordan elimination on the system, it follows that the reduced row-echelon fonn
of A is I. Hence A is invertible. e
We now develop some mathematics that will justify why B = A- 1 when we are
able to reduce[A I/] to[/ I BJ.
Matrices obtained from an identity matrix I by applying an elementary row operation
to I are called elementary matrices. We classify elementary matrices into the following
three types.
1.3 Inverses of Matrices 33
\.
£2 = [ � � ] (multiply row I by 2),
[ 1 0 ]
£3. = (add 3 times row I to row 2).
3 1
An interesting fact is that multiplication hy elementary matrices on the left of another
matrix performs the corresponding row operation on the other matrix. Notice how this
works when we multiply £ 1 , E2 , and £3 times a 2 x 2 matrix:
[� �][: :)-[; :]
2
[ � � l [: : l - [ � :]
[: �][: :]-[
a b
]
3a + C 3b +d
These illustrate the following theorem whose proof we leave as an exercise
" (Exercise 13).
as a useful theoretical tool from time to time. Our first instance of this involves seeing
why our procedure for finding inverses does in fact produce the inverse. Look at our
procedure in the following way. We begin with the augmented matrix LAI/J and use
elementary row operations to reduce it to [JIB). Suppose this takes k elementary row
operations and £1, £2, ••• , Ek are the elementary matrices that perform the successive
row operations. Since performing these elementary operations on [Al/] is the same as
performing them on A and I individually, it follows that
Ek ··· £2 £1 [AI/I =!Ek··· E2E1AIEk · · · E2E1 /J = IIIB].
From the right-hand portion of this augmented matrix, we see
Thus Bis not only the right-hand inverse of A as we saw from conception of our method
for finding inverses, but is the necessary left-hand inverse too.
Let us proceed to further develop the theory of invertible matrices. We begin with
the following theorem.
THEOREM 1.7 If A and 8 are invertible matrices of the same size, then AB is invertible and
(AB)-1 = S- 1A-1.
It suf ices to show that B -1A-1 is the inverse of AB. This we do by showing the
f
Proof
necessary products are I:
and
•
Notice that (AB)- 1 is not A- 1 B- 1• The result of Theorem 1.7 generalizes to
products of invertible matrices with more factors as follows: If A1, A 2, ..., A" are
invertible matrices of the same size, then
2 1 A1 1
1
(A1 A2 · · · A,,)- = A;;- 1 • • · A
since
(Ai A2 · · · A,, )-1 = (A,A3 • • • A ) -1A -1
- • 11 I
= (A3 .. ·A )-1A-1A-1
. n 2 l = ... -
-A-1
" ... A2-I A -I
J •
Proof In each part, let B denote the described matrix. We can then prove each part by showing
that EB = I and BE = I. This can b e done by either directly calculating these products
or by using Theorem 1.6. We leave these details as exercises (Exercise 14). •
Up to this point we have been careful to show that B is both a right-hand inverse
(that is, AB = I) and a left-hand inverse (that is, BA = 1) when verifying that a square
matrix Bis the inverse of a square matrix A. There are places where a right-hand inverse
need not be a left-hand inverse or vice versa. The next theorem tells us that this is not
the case for square matrices, however.
THEOREM 1.9 Suppose that A and B are n x n matrices such that either AB = I or BA = l. Then A
is an invertible matrix and A_, = B.
Proof Let us prove this in the case when AB = I; the case BA = I will be left as an exercise
(Exercise 17). Suppose A is not invertible. Since the reduced row-echelon form of A is
not l, there are then elementary matrices E 1 , £2 , ••• , E,,, so that EI E2 · · · Em A contains
a zero row. Consequently
E1E2 · · · E AB = E1E2 · · · E
111 111 (I)
contains a zero row. But a matrix with a zero row is not invertible (Exercise l 6) and hence
£1 E2 • - - E111 is not invertible. This gives us a contradiction since each Ei is inve1tible
(Theorem 1.8) and products of invertible matrices are invertible (Theorem 1.7). Now
that we know A is invertible, we can choose E1 , E2, ... , E111 so that
•
E1 E2 ···E111 A=l.
This along with Equation (I) gives us that B = £1 £2 • • • E111 = A- 1 .
Because of Theorem 1.9, from now on we will only have to verify one of AB = l
or BA = I to see if a square matrix B is the inverse of a square matrix A.
Our final result gives a characterization of invertible matrices in terms of elementary
matrices.
THEOREM 1.10 A square matrix A is invertible if and only if A is a product of elementary matrices.
Proof If A is a product of elementary matrices, then A is invertible by Theorems 1.7 and 1.8.
Conversely, if A is invertible there are elementary matrices £1 , E2, ... Em so that
Thus,
A = Em-t ••· E- 1
2 £-
1 -1 -1 - 1
I £I E 2··· Em A = £m "' £2 £ l ·
36 Chapter I Matrices and Determinants
EXERCISES l.3
2
For each of the following matrices. either find the in 11. For A = [ � J find an elementary matrix E
,cr,e of the matrix or detennin e that the matrix is not 4 ,
r · · () so that:
imcrtiblc.
r- ' ·� [ _: -:
=: •r
I. [ ; : a) EA=
_ l [ .6, 28 ]·
- I
b) £
J, [:
!] I
� =[� � l
5. [ � =� � ]
-! _: ] 6. [ �
-: ] c) EA= [ 3 42 ]
I
12. Express the matrix in Exercise 5 as a product of el
ementary matrices.
,t :! _: -i]
2 2 2 -] 3 Prove the following parts of Theorem 1.6.
,"'.fa.
a) Part ( 1)
b) Part (2)
c) Part (3)
114. Prove the following parts of Theorem 1.8.
-I 0 _, a) Part (I)
-I I b) Part (2) J
-2
-I -1 c) Part (3)
0
0 -1 8 4
1
X1 - 4x2 + .t3 + 5x4 = I b) Give an example to show that the result of part
3.t1+ X2 + X3 + 6X4 = 2 (a) need not hold if A is not invertible.
21. Suppose that A and B are n x n matrices such that
AB is invertible. Show that A and Bare invertible.
�-"r
1.4 Special Matrices and Additional Properties of Matrices 37
l
Use the inverse command in Maple or the appropriate 18 -24 25
command in another suitable software package to find -12 - 2JT 14- rr -17 -27
the inverses of the matrices in Exercises 22-25 (if pos-
15+ 3J3 -18- 4/3 21 +4v'3 27
,ible).
G.
-10 12 -14 -18
27
14 -25 39 29 6 Use Maple or another appropriate software package
58 -41 88 24 18 to find p-1 AP where
22. 15 -6 31 -23 12 -46 192 36 -23 -84
-3 -22 -25 73 -24 -122 437 73 -45 -194
3 6 12 -24 9 A= 45 -191 -37 22 84
8 -21 14 26 -3 -120 438 74 -48 -193
�- r
7 -28 -6 66 -18 -200 686 110 -67 -306
23. 23 -12 45 -13 15
-1 2 0 -I
-9 14 -10 26 3
l
0 5 -1 3
2 4 8 -16 6
P= -I 1 -3 2 4
13.2 -11 13 27
I 0 4 2 4
-4.4 -10.2 ·5 -9
8 -2 8
15 -17.6 21.2 32.4
-10 12 -14 -18
A-[f-f ! �]
whose off diagonal entries are zero. The matrix
I Orn Maple a square root such as J3 is indicated by typing sqrt ( 3) ; Jr is indicated by typing pi. Products
require a multiplication star so that -12 - 211' is typed as -12-2*pi; likewise, 15 + 3J3 is typed as
15+3•sqrt { 3).
38 Chapter 1 Matrices and Determinants
Two other special types of square matrices are triangular matrices, which come in
two forms: upper triangular matrices in which all entries below the diagonal are zero
and lower triangular matrices in which all entries above the diagonal are zero. T he
matrix
[�; n
is an example of an upper triangular matrix, and the matrix
2. If A and B are both upper triangular, then so is AB; if A and B are both lower
triangular, then so is AB.
3. Ais invertible if and only if each of the diagonal entries of Ais nonzero.
I ent;j(AT) = aii·
is
h[ � ! ! l
THEOREM 1.13 If Aand Bare matrices so that the indicated sum or product is defined and c is a scalar,
then:
1. (AT ?= A.
2. (A+ Bl= AT + BT . M rs Y-1 rs (� f)
3. (cA) T = cAr .
b� '1)(5 'Jf) ) )< �
4. (ABl = BT AT .
5. (AT) -1 = (A-l )T . A' �' �q f'r."'?
� {l.. :) ..,. l( S '!2
Proof Part (4) is the most difficult to prove. We will prove it here and leave proofs of the
remaining parts as exercises (Exercise 23). Suppose A is an m x n matrix and Bis an
n x l matrix. The matrix (A Bl is an l x m matrix whose ij-entry is
As the results i n Equations (I) and (2) are the same, we have (AB) T = BT A r . •
40 Chapter 1 Matrices and Determinants
is not symmetric. Notice that a symmetric matrix must necessarily be a square matrix.
We leave the proofs of the properties of symmetric matrices listed in the following
theorem as exercises (Exercise 24).
-----
of a matrix and say "a matrix is row equivalent to its reduced row-echelon form
." When
we do not require property 4, the matrix is in row-echelon form. Row-echelon
'---
form is
1.4 Special Matrices and Additional Properties of Matrices 41
not uniq�so we would have to say '·a matrix is row equivalent to any of its row-echelon
forms."
The notion of row equivalence gives us a relationship between matrices of the same
size that possesses the properties listed in Theorem 1.15.
THEOREM 1.15
I. Every matrix A is row equivalent to itself.
2. If a matrix A is row equivalent to a matrix B, then Bis row equivalent to A.
3. If a matrix A is row equivalent to a matrix B and B is row equivalent to a
matrix C, then A is row equivalent to C.
We will leave the proof of Theorem 1.15 as another exercise (Exercise 25). ln
Theorem 1.15, the first property is called the reflexive property, the second is called
the symmetric property, and the third is called the transitive property of row equiva
lence. A relation that has all three of these properties is called an equivalence relation.
Equivalence relations are important types of relations occurring frequently throughout
mathematics. A couple of other important equivalence relations you have encountered
before are congruence and similarity of triangles. Not all relations are equivalence rela
tions. The inequality < on the set of real numbers IR is not an equivalence relation since
it is neither reflexive nor symmetric (although it is transitive).
We conclude this section by pointing out that just as we perform elementary row
operations, it is also possible to perform elementary column operations on a matrix.
As you might expect, these are the following:
1 . Interchange two columns.
2. Multiply a column by a nonzero number.
3. Replace a column by itself plus a multiple of another column.
When we apply a finite number of elementary column operations to a matrix A obtaining
a matrix B, we say A is column equivalent to R.
Many (we authors included) are so used to performing row operations that they
find it awkward to perform column operations. For the most part, we will avoid using
column operations in this book. But we will see one place in the next chapter where
column operations arise. If you too feel uncomfortable doing them, notice that column
operations may be performed on a matrix A by first performing the corresponding row
operations on A r and then transposing again.
EXERCISES 1.4
:J
Let A be the matrix Find:
A-[
I. A2 2. A-1
-1 0
0 -2 3. A 5 4. (A-t)4
0 0
Chapter J
u -� -n
42 Matrices and Determinants
u -! � ]
25. Prove the following parts of Theorem 1.15.
Let A and 8 be the matrices
a) Part(I )
a B b) Part (2)
A d
� , � c) Part(3)
Find: 26. Show that two matrices A and B are row equivalent
if and only if there is an invertible matrix C so that
5. AB
CA =B.
[-2 I ]
Let A and fl be rhe matrices
27. Show that two matrices A and B are row equivalent if
and only if they have the same reduced row-echelon
-
A= [ : � � ] and fl= 3 5 . form.
-
-4 I 28. Use the result of Exercise 27 to show that the matri
ces
[ i -1 ]
If possible, fin :
d -1
7. Ar . 8.B r. 2
n
9. AT +48. 10. 2A - 5B r . 2 and
II.(AB) r . 12.B T A T. -5 4
13. AT 8 1• 14. A T A.
[
2 -3
Let and B be the matrices -5
!]
A
- -2 -11
A= -� -� and B = : � ; .
[ [ ] are row equivalent.
3 4 5 _1 2
29.Show that any two invertible matrices of the same
Determine which of the following are symmetric matri size are row equivalent.
ce, in facrcises 15-20.
30. Just as we speak of the reduced row-echelon form
15. A 16. B of a matrix, we may speak of the reduced column
17. A+ 8 18. A-I echelon form of a matrix. Write a statement that
19.BBT 20.B T B describes the form of a reduced column-echelon ma
21.Prove the following parts of Theorem 1.11. trix.
a) Part (I) 31. Find the reduced column-echelon form of the matrix
[ _; ]
b) Part (2)
c) Part(3)
A-u f n
22. Prove the following parts of Theorem 1.12.
-� -2
-� -3 .
a) Part (I)
32.Find A for
3
b) Part (2)
c) Part(3)
23.Prove the following parts of Theorem 1.13.
a) Part (I) h) Part (2)
c) Part (3) d) Pan (5) 33.A square matrix A is called a nilpotent matrix if
24. Prove the following parts of Theorem 1.14. there is a positive integer m so that Am = O. Prove
a) Part(I) b) Part(2) that if A is a triangular n x n matrix whose diagonal
entries are all zero, then A is a nilpotent matrix by
c) Part (3) d) Part (4)
showing An == O.
1.5 Determinants 43
34. The Maple command for finding the transpose of a commands in another appropriate software
matrix is transpose. Use Maple or another appro package to find the reduced row-echelon form
priate software package to find AA T - 3A 7 for of matrix A in Exercise 34.
2 -3 4 b) Apply the gausselim command of Maple to
matrix A. Describe the form of the answer
-2 1 0 7 Maple is giving us.
A= 6 2 -8 4 9 c) Is A an invertible matrix?
-2 0 2 -6 -2
36. How could Maple commands or commands of an
IO 3 -13 14 12 other appropriate software package be used to find
35. a) Either of the Maple commands rrefor gaussJo rd the reduced column-echelon form of a matrix? Use
can be used to find the reduced row-echelon them to find the reduced column-echelon form of
form of a matrix. Use them or corresponding matrix A in Exercise 34.
1.5 DETERMINANTS
You already may have had some exposure to determinants. For instance, you might have
-
encountered them for finding cross products of three-dimensional vectors. Or perhaps
you have learned a method for finding solutions to some systems of linear equations
involving determinants called Cramer's rule. (We shalt discuss this rule in the next
section.) The Jacobian of a transformation is yet another example of a determinant you
might have encountered. Even if you have had some prior experience with determinants,
however, itis likely that yourknowledge of them is not thorough. The purpose of this and
the remaining sections of this chapter is to a give a thorough treatment of determinants.
]
There are a number of equivalent ways of defining determinants. We are going to
begin with a process called a cofactor expansion approach. Suppose that A is a square
matrix:
hr
a11 a12 a,,
a21 022 G2n
a12
I !There are two different ways in which the concept of a minor is commonly defined. Quite a few books
as well as Maple take our approach. Many other books, however, prefer to define the minor of a;j as the
determinant of the matrix obtained from A by deleting row i and column j. In other words. minors in these
other books are the determinants of our minors.
-l4 Chapter J Matrices and Determinants
M1 I
=[
some minors of A are:
a22 a23
], M12 = [
a21 an
a31 a33
],
a,3
a 23
].
a32 a33
We are going to define the determinant of an n x 11 ma�rix A (als? called an 11 � 11
_ _
determinant), denoted det(A), with what is called an mduct1ve defi111twn as follows. If
A is a I x I matrix,
I A = [a11 ],
I
we define the determinant of A to be its entry,
det(A) = a,,.
det(A) = a11 det(M11) - a,2 det(M12) + a13 det(Mn) - · · · + (-1) l+na1n det(M1 n)
+·
L..,(-1) 1 1 a 1j det(Mtj ).
=�
j=I
In effect, we have reduced det(A) to the determinants of the smaller matrices M j, which 1
(by repeating this reduction procedure if necessary) we already know how to find.
To illustrate, if A is a 2 x 2 matrix,
(1)
12we do not do thi� for a I x I matrix A == [ar 1] to avoid confusion with the absolute value of a11.
1.5 Determinants 45
For instance,
2 -3
=2·6-(-3)·5=27.
15 61
[
where MiJ is the minor of aiJ. Some cofactors of a 3 x 3 matrix,
a,1
:::
A= a21
]'
are:
a31 a 32 a33
I = I ::: ::: I
I a22
C 11 = (-1) 1+1
a32 a33
= -I
I
a21 a23
a31 a33
I = -I
a 11 a13
a21 a23
The signs of the cofactors can be easily remembered by noting that they form the
checkerboard pattern:
46 Chapter 1 Matrices and Determinants
+ +
+ +
+ +
det(A) = I::auC1j,
j=l
which we will refer to as the cofactor expansion about the.first row or simply the expan
sion about the. first mw. What is remarkable is that the same procedure may be followed
for any row or column.
or any I � j � n,
3 - (-2)
3 +I -12
-I 6 3
= 4(21) + 2 4) + 15 = 107,
4 -2 6
(
or the second column,
2 3 -2
- 1- - 1 ( 21
! 1+61: � -- ) �
-I 6
3 2
-
3 3 -
4 -2 � 3
= -3(-13) + 6(10) + 2(4) = 107.
1.5 Determinants 47
So that you first gain an overview of the theory of detenninants, we are going to
postpone many of the proofs of our results about determinants to the end of this chapter
in Section 1.7. The proofof Theorem 1.16 is the first of these we postpone.
You may raise the question: Why is it important to be able to expand a determinant
about any row or column? One reason is that sometimes we can make the work easier
by choosing a particular row or column. Consider the following example.
Solution Since the third column contains three zeros, let us begin by expanding about it obtaining
7 -3 4
-2 0 3
0 4 6
The remaining 3 x 3 determinant is now quickly found by expanding ahout its first
column. Doing so, we get our answer:
•
The fact that Theorem 1.16 allows us to expand about any row or column gives us
some cases in which determinants can be quickly found. One of these is described in
the following corollary.
COROLLARY 1.17 If an n x n matrix A has a zero row or zero column, then det(A) = 0.
Corollary 1.18 describes another case in which we can quickly see the value of a
determinant.
COROLLARY 1.18 The determinant of a triangular matrix is the product of its diagonal ent1ies.
Proof We will do the upper triangular case here. The lower triangular case will be left as an
exercise (Exercise 15). Suppose A is an upper triangular matrix:
48 Chapter I Matrices and Determinants
0 0 0 a,rn
Again the result is immediate if n = 1, so assume n 2'.: 2. Expanding det(A) about the
first column, we have
0 a33
det(A) = a11
0 0
If we continue to expand each remaining determinant about the first column, w e obtain
•
det(A) = a1 1a22 · · · a""
as desired.13
In Section 1.7 we shall use the fact that the determinant of a square matrix can be
found hy performing a cofactor expansion about either its first row or first column to
obtain the result stated in Theorem 1.19.
As the square matrices grow in size, the calculations of determinants using cofactor
expansions become lengthy. We next develop a more efficient method for calculating
determinants of large square matrices involving row operations. To use this approach,
we will have to know the effects of elementary row operations on a determinant. These
effects are listed in the following theorem.
Because det(A) = det(Ar), we can replace the elementary row operation by the
corresponding elementary column operation in each part of Theorem 1.20 and. obtain
13rf you are familiar with mathema1ical induction. you will notice that this proof could be more effectively
"'riltcn by using induc1ion on n.
1.5 Determinants 49
the same result. The proof of Theorem 1.20 is another one that we postpone until
Section 1.7.
We use row operations to calculate the determinant of a square matrix A in a manner
similar to the way we use them to solve a system of linear equations with Gaussian
elimination: Use row operations to reduce to row-echelon form with the exception of
making the leading entries one. This reduced matrix is an upper triangular matrix whose
determinant is easily found by Corollary 1.18.
Of course, when we apply the row operations, we must be careful to compensate
for their effects. The following example illustrates how we may do this.
-H
EXAMPLE2 Find the determinant of the matrix
-1 2
2 2
'A -[
-I
-]
Solution We begin with a first set of row operations toward the goal of getting A into an upper
triangular form. To help you follow our work, we have indicated the row operations that
will be performed.
-1 2 3
2 2 1 R2 - 2R 1
det(A) =
-1 -2 R3 - Ri
-1 4 R4 - Ri
By part (3) of Theorem 1.20, performing these elementary row operations does not affect
the determinant and hence
1 -I 2 3
0 3 -2 -5
det(A) =
0 2 -3 -5 3R3 - 2R2
0 0 -1
where again we have indicated the row operation we will perform in the next step. This
row operation is a combination of two elementary row operations: (1) multiplying the
third row by 3 and (2) adding -2 times the second row. The second of these has no
effect, but the first changes the determinant hy a factor of 3 by part (2) of Theorem I .20.
Notice how we multiply by 1/3 to compensate, obtaining
1 -1 2 3
1 0 3-5 -2
det(A) =
3 0 0 -5 -5 R3 ++ R4
0 0 -]
50 Chapter l Matrices and Determinants
The next indicated row operation we will perform changes the sign by part(l) of Theorem
1.20. Observe how we compensate:
l -1 2 3
1 0 3 -2 -5
det(A) = --
3 ·o 0 -I I
0 0 -5 -5 R4 - 5R3
Performing our last indicated elementary row operation, which has no effect on the
determinant, we have a desired upper triangular form from which we easily compute the
determinant:
2
(t -]
-2 -5
3
1 0 3 1
det(A) = -- = - .]. 3(-1)(-10) = -10.
•
3 0 0 -I l 3
0 0 �IO
EXERCISES 1.5
[ � -: -� ]
Find dct(AJ for 0 3 2 5 0
0 -2 0 0 0
A= 10. 2 4 0 I
-3 2 I 3 7 3 -2 0
4 5 0
by expanding ab<iut the indicated row or column in Ex
ercise� 1-6.
1. Row I 2. Row 2 3. Row 3 Use row operations to find the determinants in Exercises
11-14.
4. Column I 5. Colum� 2 6. Column 3
-2 1
Find the determinants in Exercises 7-1 O by expanding 11. 2 3
about appropriate rows or columns. -1 4 5
-3 0 4 2 -I 5 6 2 -1 3 I
7. 2 -1 3 0 3 4 0 -l 2 -1 4
8. 12.
4 0 5 2 I -I 3
0 5
0 -3 0 3 2 -I 5
4 3 2 I
I -2 1 -1
-2 5 -I -2
9. 2 -4 3 2
0 1 0 0 13.
5 -11 2 -6
0 2 0 -2
-1 3
1.6 Further Properties of Determinants 51
Proof Suppose that A is invertible. Then A is row equivalent to I. From Theorem 1.20, we
can see that the determinants of two row equivalent matrices are nonzero multiples of
one fillQJ:he.r.. finis'
det(A) is a nonzero multiple of det(/) = I and hence det(A) -:/= 0.
Conversely, suppose det(A) -:/= 0. Were A not invertible, the reduced row-echelon
form of A, let us call this rpatrix B, would have a zero row. But then det(B) = 0 by
Corollary 1.17. Since det(A) is a multiple of det(B), we obtain det(A) = 0, which is a
contradiction. •
Theorem 1.21 is useful for determining if a square matrix A is invertible when
det(A) can be quickly calculated. For instance, since
-2 3 0
-2 L{ -� -2
-:)-
4 10 2 =- =-2-:/= 0,
n
� (,;> I
� 2 -5
c.., 7
-5 7 0 1
- ltl ,H
the matrix
[
- 3
2
4 10
-5 7
is invertible.
Chapter 1 Matrices and Determinants
of a p roduct
Our next major objective will be to obtain a result about the determinant
about elementary
of matrices. To reach this objective, we first prove a couple of lemmas
matrices.
Proof These are all consequences of Theorem 1.20. For example, part (I) follows because
dct (£) = - det( /) = -I by part (I) of Theorem 1.20. We leave the proofs of the
remaining two parts as exercises (Exercise 13). •
Proof The first part is an immediate consequence of Theorem 1.6, which tells us that left
multiplication by an elementary matrix performs an elementary row operation, Theorem
1.20. which tells us the effect of an elementary row operation on a determinant, and
Lemma 1.22. The second part follows by repeated use of the first part:
det(E1E2 · · · E111 A) = <let(£ 1 ) det(E2 · · · E111 A)
det(AB) = det(A)det(B).
Proof We wi I reak our �r�of int ? two cases: one in which A is invertible and the other in
. � �
wb1 ch it is not. If A is rnvert1ble, then A is a product of elementary matrices by Theorem
1.10 and the result now follows from Lemma 1.23.
1.6 Further Properties of Determinants 53
det(AB) = 0
since det(E 1 £2 . • · E,,, ) -/= 0, and we again have the desired result that det(A B)
det(A) det(B). •
Proof Since
is called the cofactor matrix of A. The transpose of this cofactor matrix is called the
adjoint of A and is denoted adj(A); that is,
A=[��]
54 Chapter l Matrices and Determinants
is
and
The proof of Theorem 1.26 is another one we postpone until the next section. Notice
this theorem is telling us that adj(A) is almost the inverse of A. Indeed, we have
Corollary 1.27. �
A- 1 - I
- --ad· A - -
(
det(A) J ) - -2
I [ 4 -2 ] -- [ -2
-3
As a rule, however, Corollary 1.27 is not an effic
because of all the determinants that must be calcu
I 3/2 -1/�
l
ient way of finding inverses of matrices
lated. The approach used in Section
1.3 is usually the better way to go. This adjo
int method for finding inverses is used
primarily as a theoretical tool.
To develop our final property of this sect
ion, consider a linear system with two
equations and two unknowns:
a,,x + a12Y = b 1
a21x + a22y = b2 .
Let us start to solve this system. We coul
d use our matrix method, but we will
for such a small system. Instead, let us not bother
eliminate y by multiplying the first equ
a22 and subt racting a12 times the second equation. ation by
This gives us
(a11a22 - a21a12h = a22b 1 - a12h2.
- �-- ----
1.6 Further Properties of Determinants 55
x=---- - -
a11a22 - a12a21
a1 1
A= [
a2 1
and A 1 and A2 be the matrices obtained from A by replacing the first and second columns,
respectively, of A by the column
so that
then
det(A 1 )
x=
det(A) '
We have just discovered what is known as Cramer's rule, named after Gabriel
Cramer (1704-1752). Variations of this rule were apparently known prior to Cramer,
but his name became attacheu to it when it appeared in his 1750 work Introduction a
!'analyse des lignes courbes algebriques. The rule extends to any system of n linear
equations in n unknowns provided the determinant of the coefficient matrix is nonzero
(or equivalently, provided the coefficient matrix is invertible).
THEOREM 1.28 (Cramer's Rule) Suppose that AX = Bis a system of n linear equations in II unknowns
such that det(A) =I= 0. Let A1 be the matrix obtained from A by replacing the first column
of A by B, A2 be the matrix obtained from A by replacing the second column of A by
B, ... , A,, be the matrix obtained from A by replacing the nth column of A by B. Then
2
----
det(A ) det(A,, )
X Xn =
2 - det(A) ' det(A) ·
The proof of Cramer's rule for a general positive integer n will be given in tbe next
section. Let us look at an example using it.
56 Chapter 1 Matrices and Determinants
Solution We firsl calculate the necessary determinants (the details of which we leave out).
-l ·,
det(A)= 2 -I =3
-2
2 -1
det(A 1 ) = 3 -I =5
-2 I
2 -1
det(A2) = 2 3 =I
2
det(A3) = 2 -1 3 =0
-2
Our solulion is then:
z = - = 0.
3
•
As is the case with lhe adjoint method for finding inverses, Cramer's rule is usually
nol a very ef icient way to solve a linear system. The Gauss-Jordan or Gaussian elimi
f
nation methods are normally much better. Note too that Cramer's rule may not be used
should the system not have the same number of equations as unknowns or, if it does have
as many equations as unknowns, should the coefficient matrix not be invertible. One
place where it is often convenient to use Cramer's rule, however, is if the coefficients
involve functions such as in the following example. \
-e21 cost
det(A1) = I 2e2' sin t
I = 2e2' sin t + te2' cost
det(A2) = I
e21 sin t
2e21 cost t
I= et 21
sin t - 2e21 cos t
EXERCISES l.6
Use Theorem 1.21 to determine whether the following 13. Prove the following parts of Lemma 1.22.
J ! -! J
matrices are invertible.
- a) Part (2)
1. [ _: � .
2 [
-! ]
b) Part (3)
- -
3. � � 14. a) An invertible matrix A with integer entries is
-� ] 4. � :
[ [ said to be unimodular if A-1 also has integer
3 0 I 6 0 0 entries. Show that if A is a square matrix with
integer entries such that det(A) = ±l, then A is
Use the adjoint method to find the inverse of the follow
a unimodular matrix.
ing matrices.
s. [ -� � J [ -2
6. I
b) Prove the converse of the result in part (a); i.e.,
prove that if A is a unimodular matrix, then
det(A) = ±1.
Use Cramer's rule to solve the following systems.
7. 3x - 4y = l
[
15. a) Find the determinants of the following matrices.
J J
8. 7X + y = 4
2x+3y=2 2x -5y = 8 -2 I 2
9. - y +z = I
3x 10. 5x - 4y + z = 2
A=
[ 13 4
and B = _
2 3
2x + y - 3z = 3 2x - 3y- 2z = 4
b) Find det(AB), det(A-1), and det(B r A-1)
X - 2y + Z = 7 3x + y + 3z = 2 without finding AB, A-1, or B T A-1•
11. Use Cramer's rule to solve the following system for
X and y. c) Show that det(A + B) is not the same as
det(A) + det(B).
xe' sJn 2t + ye' cos2t = t
2xe1 cos 2t - 2ye' sin2t = t2 16. Show that if A and B are square matrices of the same
12. Use Cramer's rule to solve the following system for size, then det(AB) = det(BA).
x, y, and z.
17. a) Either of the commands adj or adjoint can be
e' x + e21y + e-1 z = I used in Maple to find the adjoint of a square
matrix. Use either one of these Maple
e1 x + 2e21 y - e-1 z = t
commands or corresponding commands in
e1 x + 4e2'y + e- z = t2
1
another appropriate software package to find
58 Chapter I Matrices and Determjnants
A=[ 2
1.2 2 -3.1 c) By part (b), what is the value of det(A)?
1.2 -2 2.6
.
-2.1 3.7
2.3 4
I
-3 6.5
-4
]
1. 7 PROOFS OF THEOREMS ON DETERMINANTS
In this section we will prove those results about detenninants whose proofs were omitted
in the previous two sections. Many of these proofs will use the technique of mathematical
induction, a technique of proof with which we will assume you are familiar.
Recall that we defined the determinant of an n x n matrix A = [aij] as the cofactor
expansion about the first row:
II
'
II
det(A) = �
L..a1 jCJi = �
L..(-1) I+lalJ det(Mtj),
j=I i=I
As we prove some of our results, we will sometimes have minors of minors; that is,
we will have matrices obtained by deleting two rows and two columns from A. For
notational purposes, let us use
M(ij, kl)
to denote the matrix obtained from A by deleting rows i and k (i =fa k) and columns j
and I (j -:/= l).
Our first theorem about determinants (Theorem 1.16) was that we could expand
ahout any row or column. As a first step toward obtaining this result, we show that we
can expand about any row.
det(A) = Laijcij •
j=I
Proof The v �rific �tion is easy for n = 2 and is left as an exercise (Exercise t ). Assume the
result 1s valid �or all� x k matrices and suppose that A = [%] is a (k + 1) x (k + J)
_
matrix. There 1s nothing to show if i = I, so assume i > 1. By definition,
k+l
Using the induction hypothesis, we may expand each det(M1i) about its row and obtain
k+l j-1
l=j+I
(1)
= LL(-l) i+Hla1jail det(M(lj, ii))
j=I l=I
j=l l=j+l
(On the second summation, l - l occurs in the exponent of -1 instead of l since the
Ith column of A becomes the (l - l)st column of M1i when l > j.) Now consider the
cofactor expansion about the ith row, which we write as
L.)-1)''+Iail det(Mu).
k+I
'\"""'
1=1
j=l+I
(2)
= LL(-l) i +Hl+taualj det(M(il, l j))
k+I 1-1
k+I k+I
l=I j=l
While they may look different, the results in Equations (1) and (2) are the same. To see
why, consider a term with a j and an l. If j > l, this term in Equation (I) is
(-1/+i+1aljau det(M(lj, ii)),
which is exactly the same tem1 as we have in Equation (2) for j > l. We leave it as an
exercise (Exercise 2) to show that the terms with j < l in Equations (I) and (2) are the
gm� •
clet(A) = La;1C;1.
i=l
ci A
Proof We again use induction on II. We �ill let you verify this for n = 2 (Exe: rJ\ �:�;x�
this result holds for all k x k matnces and let A= [ aij] be a (k + 1) ( )
By definition
k+l
j
det(A) = :�:)-l) I+ a1jdet(M1j)
j=l
k+l
= £111 det(M11) + 2)-1)1+ja1j det(M1j).
j=2
Using the induction hypothesis, we expand each det(M1j) about its first column for
I�(- I
j � 2 and obtain
I
i=I i=2
and then expanding each det(M;1) for i � 2 about its first row, we obtain
l
k+l k+I
a11 det(M1 1) + �(-l);+ ia;1 ?;(-1) 1+i-la1jdet(M(il. lj))
•
i=2 j=2
Since the results of Equations (3) and (4) are the same, our proof is complete.
Before completing the proof of Theorem 1.16, we use Lemma 1.30 to prove Theorem
1.19. For convenience, let us restate Theorem 1.19 as Theorem 1.31.
THEOREM 1.31 If A is an II x II matri x,
det(A r) = det(A).
Proof Herc we use induction too, only we may start with n = 1 where the result is trivial for a
I x I matrix A = [a , i]. Assume the result holds for any k x k matrix and let A = [aij J
1.7 Proofs of Theorems on Determinants 61
� aijCij = �
det(A) = L., ·+·
L..,(-1)' laij det(M ij )
i=I i=I
where A = [aij] is an n x n matrix with 11 ::: 2. To obtain this, we first expand det(A7)
det(AT ) = L., T
�(-1)1·+; a;j det(M;)· .
i=I
Now applying the result of Theorem J .31to det(A7) and each det(M{), we obtain the
desired result.
Another result we have not proved that we now prove is Theorem 1.20, which we
restate as Theorem 1.32.
Proof
1. We proceed by induction on n leaving the first case with n = 2 as an exercise
(Exercise 4.) Assume part (I) holds for all k x k matrices and Jet A = [aij] be a
(k + l) x (k + 1) matrix . Suppose that B is obtained from A by interchanging
rows i and I. We are going to expand det(B) about a row other than row i or
row l. Pick such a row. Let us call this the mth row. We have entmj(B) = amj ·
If we interchange the rows of the minor Mmj of the entry amj of A that come
from rows i and f of A, we obtain the minor of entmj(B) in B. By the
62 Chapter I Matrices and Determinants
1·
j=I
of A bi �he mi_�or
2. Suppose that B is obtained from A by multiplying row i
. ence 1
of entij (B) of B is the same as the minor Mij of the. entry aij o
we expand about the ith row,
"
det(B) = I)-l ) i +j ca;j det(M;j) = cdet(A).
j=l
LEMMA 1.33 If A is an II x II matrix where n � 2 with two rows that have the same entries, then
det(A) = 0.
Proof Suppose that row i and row j of A have the same entries. Let B be the matrix obtained
from A by interchanging rows i and j. On the one hand, by part (I) of Theorem 1.32.
we have
det(B) = - det(A).
On the other hand . /3 = A and hence
•
det(B ) = det(A).
Thus dct(A) = -det(A), which implies det (A) = 0.
Proof of Suppose that B is obtained from A hy replacing row i of A by itself plus c times row
Theorem 1.32, I of A. Then ent,j(B) = ai + ca1j and the minor of ent ij(B) of B is the same as the
j
Part (3) minor M;1 of the entry aij of A. If we expand det ( B) about row i,
n
The second sum is the same as the determinant of the matrix obtained from A by replacing
the i th row of A by row m and hence is zero by Lemma 1.33. Thus
n
�he proof of our result about the product of a square matrix and its adjoint (Theorem
)
1.26 1s another place where Lemma 1.33 is used. We restate this result as Theorem 1.34.
1.7 Proofs of Theorems on Determinants 63
Proof We will prove that A adj(A) = det(A)/ and leave the proof for adj(A)A as an exercise
(Exercise 5). Suppose A is an n x n matrix. Notice that
II
If i = ),
II
Ifif:j,
L a;kC k
II
entij (A adj(A)) = i
k=I
is the determinant of the matrix obtained from A by replacing the jth row of A by the
ith row of A. Since this determinant contains two rows with the same entries, we have
II
•
k=l
The final result about determinants we have yet to prove is Cramer's rule, restated
as Theorem .1 35.
THEOREM 1.35 Suppose that AX = B is a system of n linear equations in n unknowns such that
det(A) =/: 0. Let A I be the matrix obtained from A by replacing the first column of A by
B, A2 be the matrix obtained from A by replacing the second column of A by B, ... , A"
be the matrix obtained from A by replacing the nth column of A by B. Then
det(A1) ' det(A2 ) det(A11)
= x2 =
det(A)
Xt
det(A) ' det(A)
Proof Since
1 1
A- = --ad'(A),
det(A) J
we have
l .
X = A- I B = --adJ(A)B.
det(A)
M Chapter I Matrices and Determinants
The summation is exactly the determinant of A; expanded about the ith column and
hence
det(A;)
X; =- - -.
det(A)
•
EXERCISES 1.7
a) Show that if 11 = 2,
det(V) = x2 - x1.
b) Use row operations to show that if n = 3,
det(V) = (x2 - X1)(x3 - x1)(x3 - x2).
c) Use row operations to show that if n = 4,
det(V) = (X2 - X1 )(x3 - xi)(;4 - X1 )(X3 - X2)
(X4 - X2)(X4 - X3),
d) In general,
det(V) = 11
J=2
{fi
t=I
(Xj - x;)}.
14 Named for
Alexandre Theophile Vandennonde (1735-1 796) who studied
!he theory of equations and de-
1em1inant�.
Vector Spaces
Your first encounter with vectors in two or three dimensions likely was for modeling
physical situations. For example, winds blowing with speeds of 5 and IO miles per hour
45 ° east of north may be illustrated by the velocity vectors u and v in Figure 2. I drawn
as directed line segments of lengths 5 and IO units pointing 45 ° east of north. A force
pushing a block up an inclined plane might be illustrated by drawing a force vector F
as in Figure 2.2. In your calculus courses you should have encountered many uses of
vectors in two and three dimensions in the study of equations of lines and planes, tangent
and normal vectors to curves, and gradients, just to name a few.
I Figure 2.1
Figure 2.2
65
66 Chapter 2 Vector Spaces
Were someone to ask you to briefly tell them about vectors you might well r� spond
_
by saying simply that a vector v is a direc� ed line se�� ent in �wo or thr�e d 1men� 10ns. If
_ _
we place a vector so that its initial point 1s at the ongm (chmce � f th� m1t1� I pornt does
not matter since we are only trying to indicate magnitude and directlon with � vector)
and its terminal point is (a, b) in two dimensions or (a, b, c) in three dimens10ns, we
can denote the vector by its terminal point as
v = (a, b) or v = (a,b,c)
in two or three dimensions, respectively. (Other standard notations you might have
used instead are v = (a, b} or v = ai + bj in two dimensions and v = (a, b, c} or
v = ai + bj + ck in three dimensions. In Chapter I we mentioned that in this text we
will write our vectors in two, three, and even 11 dimensions as column matrices or column
vectors.) Vectors are added by the rules
(a,.b1) + (a2,b2) = (a, +a 2,b1 +b2)
or
(which, of course, are special cases of matrix addition and scalar multiplication).
We could continue discussing things such as the geometric impact of vector addi
tion (the parallelogram rule), the geometric impact of scalar multiplication (stretching,
shrinking, and reflecting), dot products, cross products, and so on, but that is not our
purpose here. Our purpose is to study sets of vectors forming a type of structure called a
vector space from an algebraic point of view rather than a geometric one. To us a vector
space will be a set on which we have defined an addition and a scalar multiplication
satisfying certain properties. Two old friends, vectors in two dimensions and vectors in
three dimensions, are two examples of vector spaces, but they are not the only ones as
you are about to see.
D �FI� ITION A nonempty set Vis called a vector space if there are operations
_
of add,t,on and scalar multiplication on V such that the following eight propenies
_
are satisfied:
1. tt + v = v +11 foral\ u and v in V. re"""" c.,,{cl
2. u + (v +w) = (u + v) +w for all u, v, and win V.- d i,� '-'J,,
2.1 Vector Spaces 67
The eight properties of a vector space are also called the laws, axioms, or postulates
of a vector space. The elements of the set V when V is a vector space are called the
vectors of V and, as we have done already with matrices, real numbers are �ailed scalars
in connection with the scalar multiplication on V .1 Actually, not all vector spaces are
formed using real numbers for the scalars. Later we shall work with some vector spaces
where the complex numbers are used as scalars. But for now, all scalars will be real
numbers.
Some terminology is associated with the vector space properties. Property 1 is
called the c ommutative law of addition, and property 2 is called the associative law
of addition. The clement O of V in property 3 is called an additive identity or a zero
vector,2 and the element -v of V in property 4 is called an additive inverse or a negative
of the vector v. Because of commutativity of addition, we could have equally well put
our zero and negative vectors on the left in the equations in properties 3 and 4, writing
them as
0+v =v and - v + v = 0.
Properties 5 and 6 are distributive properties: Property 5 is a left-hand distributive
property saying that scalar multiplication distributes over vector addition, and property
6 is a right-hand distributive property saying that scalar multiplication distributes over
scalar addition. Property 7 is an associative property for scalar multiplication.
Let us now look at some examples of vector spaces.
EXAMPLE 1 From our matrix addition and scalar multiplication properties in Chapter I, we imme
diately see the set of n x J column vectors or n-dimensional vectors JR." satisfies the
eight properties of a vector space under our addition and scalar multiplication of column
vectors,
I In print, vectors are often set in boldface type and, in handwritten work, marked with an arrow over the top
to distinguish them from scalars. Such confusion will not arise in this text, however, since we will reserve the
lowercase letters u, v, and w for vectors. Scalars usually will be denoted by letters such as a, b, c, and d.
2 In cases where zero vectors could be confused with the scalar zero, we will put zero vectors in boldface
print as O as we did for zero matrices in Chapter I.
68 Chapter 2 Vector Spaces
Hence JR" (of which vectors in l wo and three dimensions are special cases when we write
these vectors in column form) is a vector space for each positive integer n. •
Row vectors of a fixed length n would also form a vector space under our addition and
scalar multiplication of row vectors. More generally, so would matrices of a fixed size
under matrix addition and scalar multiplication. Let us make this our second example.
EXAl\lPLE 2 The set of m x II matrices Mmx11 (JR) satisfies the eight properties of a vector space under
matrix addition and scalar multiplication and hence is a vector space. •
Making our definition of a vector space as general as we have done will prove
valuable to us in the future. For instance, variou�ets of real-valued functions3 form
vector spaces as is illustrated in the next example. Because of this, sets of real-valued
functions wi II have many properties similar to those enjoyed by our matrix vector spaces,
which we shall exploit later in our study of differential equations.
EXAMPLE 3 Let F(a, b) denote the set of all real-valued functions defined on the open interval (a, b). 4
We can define an addition on F(a. b) as follows: If f and g are two functions in F(a, b),
we let f + g be the function defined on (a, b) by
(f + g)(x) = f (x) + g(x).
We can also define a scalar multiplication on F(a, b): If c is a real number and f is a
function in F(a, h), we let cf be the function defined on (a, b) by
(cf)(x) = cf (x).
Show that F(a, b) is a vector space under this addition and this scalar multiplication.
Solution We verify that the eight properties of a vector space are satisfied.
:u��:�-7��:! � t: ;/e�n:
n i n 1 nctio n whose r:�ge i s contai ned in the set of
real numbers. For i nstance, the
by /�x) '."' x IS real-valued; so is the function f from JR2 to IR defi
ned by
/(,;, ·) = xi + y2 . For a fixed pos1t1vc i n teger n the detenni nant is a real-valued function
toR.) from Mnxn(lR)
4 For in stance, the fu nction s given by J(x) = x2 · f(x ) -
integer function 'A-ould be elements of F(-oo
fi lx l, f(x) = sin x, ((x) = e', and the greatest
) th ese ve funct10ns along with the functions defined
f(x) = l/x. f(x) = lnx • and f(x) - cot x wou by
• oot d'be elements of F(O, n).
2.1 Vector Spaces 69
O(x) = 0.
The zero function is then an element of F(a, b), and for any fin F(a, b) and
x in (a, b),
(-f)(x) = - f (x)?
I
The function -J is in F (a, b), and for any x in (a, b) we have
(f + (- f))(x) ·= f (x) + (-f)(x) = f (x) + (- f (x)) = 0 = O(x).
Hence - f serves as a negative of f.
5. This and the remaining properties are verified in much the same manner as the
first two, so we will go a little faster now. For any real number c and any
functions f and gin F(a, b),
(c(j + g))(x) = c(f + g)(x) = c(f(x) + i(x)) = cf(x) +cg(x)
= (cf)(x) + (cg)(x)
for any x in (a, b), and hence c(f + g) =cf + cg.
6. For any real numbers c and d and any function fin F(a, b),
((c + d)j)(x) = (c + d)f(x) = cf(x) + df(x) = (cf)(x) + (df)(x)
for any x in (a. b), and hence (c + d)f =cf+ df.
Lest we b e accused of doing everything for you, we will let you verify the last two
properties as an exercise (Exercise 1 ). e
There is nothing special about using an open interval in Example 3. The set of real
valued functions defined on a closed interval [a,b], which we will denote by F[a, b],
also forms a vector space under the addition and scalar multiplication of functions we
used in Example 3. More generally, the set of real-valued functions defined on any set
70 Chapter 2 Vector Spaces
s, which we will denote by f(S),is a vector space under the types of addition and scalar
multiplication of functions we used in Example 3.
..
The next example illustrates that not e very se.t on which is defined an add1t1on and
a scalar multiplication is a vector space.
EXAMPLE 4 On the set of pairs of real numbers (x, y), define an addition by
(xi. Y1) + (x2, Y2) = (x1 + xz+1,Yi + Y2)
and a scalar multiplication by
c(x,y) = (cx,cy).
anJ
while
c(x,, y,) + c(x2, y2) = (ex,, cy,} + (cx2, cy2} = (cxi + cx2 + L cy, + cy2),
we see that property 5 does not hold for every real number c. (In fact, it holds
only when c = J .) Thus, this is not a vector space. Once we see one property
that does not hdld we are done with determining whether we have a vector
space in this example. Were we to continue going through the properties, we
would also find that property 6 does not hold. (Try it.) Properties 7 and 8 will
hold. (Verify this.) •
Do not let Example 4 mislead you into thinking that unusual operations for addition
or scalar multiplication will not produce a vector space. Consider the next example.
j
EXAMPLE 5 Let JR+ denote the set of positive real numbers. Define addition on JR+ by
xEBy = xy
and scalar multiplication by
C Ox= x'
where x and y are in R+ and c is a real number. (We use the symhols EB and 0 to avoid
confusion with usual addition and multiplication.) Determine if JR+ is a vector space
under this addition and scalar multiplication.
Solution
1. Since
x EB y = xy = yx = y EB x
this addition is commutative.
2. Since
7. c0(d0x)=(cd)0x,and
8. 10 X = X
•
J ,. ·-
Proof
1. Suppose that O and O' are zero vectors of V. On the one hand, since O is a zero
vector, we have that
O' +O =0'.
On the other hand,
O' +O =0
since O' is a zero vector. Thus we see O' = 0.
2. Suppose that -u and -v' are negatives of v. Notice that
-v' + (u +(-u)) =-u' + 0 = -v' = (-v' +v) + (-v) = 0 + (-u) = -v,
and hence we see -v = -v'. •
Because of Theorem 2.1, we may now say the zero vector instead of a zero vector
and the negative of a vector instead of a negative of a vector.
Theorem 2.2 contains some more properties that we shall use often.
Proof We will prove parts (1) and (3) and leave the proof of part (2) as an exercise
(Exercise I 0).
I. One way to prove this is to first notice that since
0 · V = (0 + O)v = 0 · V + 0 · V
5 Here is a place where we have put the zero vector in boldface print to distinguish it from the scalar zero.
2.1 Vector Spaces 73
we have
0 · V = 0 • V + 0 · V.
Adding -(0 · v) to each side of the preceding equation, we obtain
Q· V + (-(o.lv)) = 0 · V + 0 • V + (-(0 · v))
from which the desired equation
0 = Q. V
now follows.
2. Noting on the one hand that
(1 + (-l))v = 1 · v + (-l)v = v + (-l)v
and on the other hand that
(l+(-l))v=O·v=O
by part ( 1 ), we have
V + (-})v = 0.
Adding -v to each side of the preceding equation, we obtain
-V + V + (-l)V = -V + 0
and hence
(-l)v = -v. •
Finally, we point out that we can define subtraction on a vector space V by setting
the difference of two vectors u and v in V to be
U - V = U+(-V).
You might notice that we could have equally well subtracted the vector whenever we
added its negative in the proofs of parts (1) and (3) in Theorem 2.2.
EXERCISES 2.1
1. Complete Example 3 by showing that properties 7 properties of a vector space fail to hold.
and 8 of a vector space hold.
a) (x1, Yi)+ (xi, Y2) = (x 1, Y2),
2. Complete Example 5 by showing that properties 6, c(x, y) = (ex, cy)
7, and 8 of a vector space hold.
b) (xi, y,)+ (x2, Y2) = (x, + X2, YI+ Y2),
3:Yn each of the following, determine whether the indi
cated addition and scalar multiplication on ordered c(x,y)=(c+x,c+y)
pairs of real numbers yields a vector space. For c) (x1, Y1)+(x2, Y2) = (x, + Y2,x2 + Y1),
those that are not vector spaces, detenniue which. c(x, y) = (ex,cy)
74 Chapter 2 Vector Spaces
4. In each of the following. determine whether the indi addition and scalar multiplication
cated addition and scalar multiplication of ordered {a.}+ {b,,J ={a,,+ b11}. c{a.) = {ca.)?
triples of real numbers yields a vector space. For
those that are not vector spaces, determine which If not, why not?
properties of a vector space fail to hold.
a) (x1, YI, Zt) + (X2,Y 2, Z2) =
(xi + X2, )'1 + Y2, ZI + z2),
numbers I::
8. Let S denote the set of all convergent series of real
1 a11 • Is S a vector space under the
addition and scalar multiplication
c(x,y,z) = (cx,y,cz)
00 00 00
CX) 00
6. Docs the set of complex numbers under the addition Show that V is a vector space. Such a vector space
and scalar multiplication is called a zero vector space.
(a +bi)+ (c + di) = (a+ c) + (b + d)i, 10. Prove part (2) of Theorem 2.2.
c(a + bi) = ca + cbi 11. Prove that i f c is a real number and vis a vector in a
where a, b, c, and,/ are real numbers form a vector vector space V such that cv = 0, then either c = 0
space? If not, why not? or v = 0.
7. Let C denote the set of all convergent sequences of 12. Show that subtraction is not an associative operation
real numbers {a.}. Is C a vector space under the on a vector space.
[J
EXAMPLE 1 Let W be the set of all column vectors of the fonn
2.2 Subspaces and Spanning Sets 75
The set W is a subset of JR3 . In fact, W is a subspace of JR3 . To see this, first notice that
the addition of JR3 on elements of W gives us an addition on W: For two elements
1
[ XJ ] [ X2 ] [ X1 + X2 ]
� + = YI; Y2 ,
which is also an element of W. The fact that the sum of two elements of Wis again
an element of W is usually described by saying that W is closed under addition. Next
notice that the scalar multiplication of JR3 gives us a multiplication on W: If c is a scalar
and
is an eleme� of W, then
UJ
properties I and 2 of a vector space hold for W. Property 3 holds since
[ �]
are in W, property 4 holds. As was the case with commutativity and a ssociativity of
addition, the scalar multiplication properties 5-8 will carry over from JR3 to the subset
W. Hence W is a vector space. •
-----__
76 Chapter 2 Vector Spaces
Looking back at Example 1, notice that properties 1, 2, and 5-7 are immediately
inherited by any subset of a vector space, so we really do not need to check for them.
In fact. the next theorem tells us that the closure properties are really the crucial ones in
determining whether a nonempty subset of a vector space is a subspace.
THEOREM 2.3 Let W be a nonempty subset of a vector space V. Then W is a subspace of V if and only
if for all 11 and win Wand for all scalars c, 11 + w is in Wand cu is in W.
[ �' J + [ x, ]- [ x, ; ,
x
],
ha� 2 �ot I for its sec ond entry, the set of such vectors is nol closed under addition and
hence ts 00_1 a su�space. It is also easily seen that this set of vectors is not cl osed under
_
scalar mult1phcat1on. •
Solution Adding two such vectors, we obtain a vector of the same form:
Hence we have closure under addition. We also have closure under scalar multiplication
[:]
since
C
= [ :: ].
•
x - 2y ex -2cy
Thus these vectors do form a subspace.
Solutions to systems of homogeneous linear equations form subspaces. Indeed this
will be such an important fact for us that we record it as a theorem.
THEOREM 2.4 If A is an m x n matrix, then the solutions to the system of homogeneous lincarequations �
AX= 0 is a subspace of�".
Proof First notice that the set of solutions contains the trivial solution X = 0 and hence is a
nonempty subset of IR". If X 1 and X2 are two solutions of AX= 0, then AX, = 0 and
AX2 = 0 so that
A(X1 + X2) = AX1 + AX2 = 0 + 0 = 0
and hence the set of solutions is closed under addition. If X is a solution and c is a scalar,
then
A(cX) = cAX = cO = 0
and hence the set of solutions is closed under scalar multiplication. Thus the set of
solutions to AX = 0 is a subspace of IR". •
In Section 2.1, we noted that sets of real-valued functions on intervals form vector
spaces. There are numerous examples of suhspaces of such function spaces that will
come up in our future work listed in Examples 4-10.
EXAMPLE 4 Let C(a, b) denote the set of continuous real-valued functions on the open interval (a, b),
which is a nonempty subset of F (a, b). From calculus, we know that sums of continuous
functions and constant multiples of continuous functions are continuous. Hence C (a, b)
is closed under addition and scalar multiplication of functions and is a subspace of
F�,�- e
78 Chapter 2 Vector Spaces
EXA.MPLE7 For each nonnegative integer n, we will use C" (a, b) to denote the set of all functions
that have a continuous 11th derivative on (a, b). Notice that C0 (a, b) = C(a, b), each
c +1 (a, h) is a subspace of C" (a, h ), and C"(a, b) is a subspace of D" (a, b) for each
II� I. •
EXAMPLES We will let C (a, b) denote the set of functions that have a continuous nth derivative
00
for every nonnegative integer n. The set C00 (a, h) is a subspace of C" (a, b) for every
nonnegative integer 11. •
EXAMPLE9 We will let P denote the set of all polynomials; that is, P consists of all expressions
p(x) of the form
EXAMPLE HI Forcach nonnegative integerk, we will let Pk denote theset ofall polynomials of degree
less than or equal to k along with the polynomial 0. In particular, Po is the set of all
constant functions p(x) = a, P1 is the set of all linear functions p(x) = mx + b, and
P2 is the set of all functions of the form p(x) = ax2 +bx+ c. Each Pk is a subspace
of P. Also. Po is a subspace of P1 , PJ is a subspace of P2 , and so on. •
We could equally well use other types of intervals in Examples 4-8. When doing
so, we will adjust the notation accordingly. For example, C[a, b J will denote the set of
continuous functions on the closed interval [a, b], which is a subspace of F[a, b].
We next tum our attention to spanning sets. If Vis a vector space and v1, v2, ••• , v 1
1
of VI , V2, .•. , Vn • As
CJVJ + C2V2 + · · · + c,,v,. + d1v1 + d2v2 + · · · +d,,vn
= (CJ + d1)V1 + (c2 + d2)v2 + · · · + (c,. + d11)Vn
is a linear combination of v,, v2, .•• , v11 , we have closure under addition. Also, for any
scalar c, the fact that
shows we have closure under scalar multiplication and completes our proof. •
The subspace of a vector space V consisting of all linear combinations of vec
tors v 1, v2, ... , v11 of V will henceforth be called the subspace of V spanned by
I I
Vt, v2, ... , v11 and will be<lenoted
have a solution,
1 -1 2
_
-1
-2 0 -1 -1 -3
�[i :l
---+
[ -;
-] I -3 I
-
10 -1 6
:
2 3 I
1 -1
-1 -1
0 6 I 6 '
0 7 I 7
we see the system does have a solution and hence our answer to this problem is yes. •
Solution By comparing coe fficients of x2, x, and the constant tenns, we see that there are scalars
Ct, C2, C3 SO that
2 2
c1 (x +x) + c2(x2 - 1) + c3(x +I)= 2x +x + I
if and only if the syste m of equations
-c2 +c3 = I
has solutions. Starting to reduce the augmented matrix for this system,
I I 0:2 I 10: 2]
[ I 0 I ! I ] ---+ [ 0 -1 I I -1 ,
0 -1 , I O -] 1 , 1
we can see that we have arrived at a system with no solution and hence our answer to
this problem is no. e
We say thatthe vectors vi, v2, ... , Vn of a vector space V span V if Span{ v 1, li2, ... ,
Vn} = V. To put it another way, Vt, V2, . • • , Vn span V if every vector in V is a linear
combination of vi. v2, ... , Vn. In our final two examples of this section we determine if
some given vectors span the given vector space.
EXA \1PLE 13 Do
2.2 Subspaces and Spanning Sets 81
[: J
of R , we must determine whether there are scalars c1, c2 so that
J J=[: J,
2
Ct [ -� + C2 [ _:
EXAMPLE 14 Do x 2 + x - 3, x - 5, 3 span P2 ?
Here we must determine whether for an arbitrary element ax2 +bx+ c there are scalars
CJ, C2, C3 SO that
Solution
CJ (x
2
+ X - 3) + C2 (X - 5) + C3 • 3 = ax 2 + bx + C.
Comparing coefficients, we are led to the system
C2 = b
CJ= a
which obviously has a solution. Thus the answer to this problem is yes. •
EXERCISES 2.2
1. Determine which of the following sets of vectors are d) All vectors [ ; J where x + y = 0
subspaces of JR2 •
2. Determine which of the following sets of vectors are
subspaces of JR3 .
a) AIJ vectors of the form [ � J
z
R2 Chapter 2 Vector Spaces
[ � -� ]. [ _: � J
d) The II x II matrices of determinant zero
spa n M2x2(R).
e) The II x II invcniblc matrices
S. If A is an m x II matrix and B is a nonzero element
19. Determine if x 2 - 1, x 2 + l, x2 + x span h
of IR"' , do the solutions to the system AX = B form
a �ubspace of IR"? Why or why not? 20. Determine if x3 + x2 , x 2 + x, x + 1 span h
6. Complex numbers a+ bi where a and bare integers
arc called Gaussian integers. Do the Gaussian inte
gcn, form a subspace of the vector space of complex Use the system of linear equation solving capabilities of
numbers? Why or why not? Maple or another appropriate software package in Exer
7. Do the sequences that converge to zero form a sub cises 21-24.
�pace of the vector space of convergent sequences?
How abou1 the sequences that converge 10 a rational 2
number?
-2
8. Do the series that converge to a positive number form 21. Determine if is in
a l.ubspace of the vector space of convergent series? ,. 0
How about the series that converge absolutely? -1
4
9. h [ � }n Spa
n{[ -: ] , [ � ] ?
} 4 2 -17 -31
-
17 J 5 -2 -1 ./3 7 -l
44 0
[ ] [
2 6 -3 ' 2 7C ] ,
1 l -7 9 21 0 3 - 1
-22 [ ] [ ]
15 14 -22 8 ' 1 2
]I -3
span M2x3(!R).
1.9 3
25. Suppose that Vt, v2, ••• , Vk are vectors in JR". How
can we tell from a row-echelon fonn of the matrix
22. Determine if x4 + x2 + I is in
Span{x4 - x 3 + 3x -4, x 4 - x 3 - x 2 + x -4,
if VJ, v2, ... , Vt span JR"?
x 3 + x 2 - 3x + 3,
26. Use the answer to Exercise 25 and one of the gausse
x4 + x 3 - 2x2 + 4x - 8, lim, gaussjord, or rrefcommands of Maple or corre
5x 4 - 1x3 - 2x2 - x + 9, 2x 4 - 7x3 + I}. sponding commands in another appropriate software
package to determine if the vectors
2.1. Determine if 0 -2
x5 - x4 + x 3 + x, -1 2
x4 - x3 + 2x -4, 2 1 5
x5 - sJ + 6x 2
- 8x + 2, 4 -3 -3
i
xs + x4 -x3 + 2x 2 + 3.x - 1, 5 ./2
-2x 3 -4x 2 +3x -9, -3
x - 3x 3 + 7i x 2 - 2x + 1
4
8
span P5 • 7
-9 --4
24. Determine if
11
[ -� -� ; l [ �
-4
-1 -1
J span JR
2
5
2 •
DEFINITION Suppose that Vt, v2, •.. , v,, are vectors in a vector space V. We
say that VJ, v2 , ••• , v. are linearly dependent if there are scalars ct, c2, .•. , c,,
not all zero so that
I C1V1+c2v2+.. ·+CnVn=0.1
If Vt, v2 , ... , Vn are not linearly dependent, we say v1, v2, ... , Vn are linearly
independent.
84 Chapter 2 Vector Spaces
We can always get a linear combination of vectors v1, v2, ... , v" equal to the zero
vector by using zero for each scalar:
0 •Vt+ 0 • V2 + • • • + 0 • V11 = 0.
Carrying over the terminology we used for solutions of homogenous systems of linear
equations, let us call this the trivial linear c ombination of Vt, v2, ... , v, We then
1•
could equally well say Vt, v2, ••• , v11 are linearly dependent i f there is a nontrivial linear
combination of them equal to the zero vector; saying Vt, v2, ... , v11 are linearly inde
pendent would mean that the trivial linear combination is the only linear combination of
Vi, v2, . . . , v11 equal to the zero vector.
[�JI Jr: J
EXAMPLE l Are the vectors
Solutio11 Consider a linear combination of these vectors equal to the zero vector of R3:
3 I 5 : 0 0 -8 8 : O
we c�n s�e that our system has nontrivial solutions. Thus there are
nontrivial linear
•
comb1nat10ns of our three vectors equal to the zero vector and
hence they are rmear 1y
dependent.
Comparing coefficients of x 2 , .x and the constant tenns on each side of the preceding
equation, we obtain the system:
CJ+C2 = Q
-C2 + C3 = 0
CJ + c2 + 2c3 = 0.
We do not need to set up an augmented matrix here. Notice that subtracting the first
equation from the third will give us c3 = 0. The second equation then tells us c2 = 0
from which we can now see c 1 = 0 from the first equation. Since our system has only the
trivial solution, it follows that the three given polynomials are linearly independent. •
The next theorem gives us another characterization of linear dependence.
THEOREM 2.6 Suppose VJ, v2, ••• , v. are vectors in a vector space V. Then VJ, v2, ... , v11 are linearly
dependent if and only if one of v 1, v2, ... , v11 is a linear combination of the others.
Proof Suppose v 1, v2, ••• , v. are linearly dependent. Then there are scalars c1, c2, ... , Cn not
all zero so that
and hence obtain that v; is a linear combination of VJ, •.. , v;-J, v;+J, .•• , v •.
To prove the converse, suppose one of VJ, v2, ... , v11 let us call it v;, is a linear
,
dependence. But if not, we would then have to look at v2 and see if v 2 is a linear
combination of v 1, v3, ... , v,,. If so, we again would have linear dependence. If not,
we would move on to v3 and so on. Notice that checking to see if there is a nontrivial
linear combination of v 1, v2, .•• , v11 equal to the zero vector is much more efficient.
86 Chapter 2 Vector Spaces
One exception is in the case of two vectors v1 and u2, for having one vector a linear
combination of the other is the same as saying one vector is a scalar multiple of the
other, which is often easily seen by inspection. For example,
are linearly dependent since the second vector is 3 times the first (or the first is 1/3 times
the second). The polynomials x2 + x and x2 - 1 are linearly independent since neither
is a scalar multiple of the other.
We next introduce the concept of a basis.
DEFINITION We say that the vectors v1, v2, ... , Vn of a vector space V are a
basis for V if both of the following two conditions are satisfied:
1. v,, v2, ... , v11 are linearly independent.
2. v1, v2, ... , v" span V.
J J,
are easily seen to be linearly independent since if
c1 e1 + c 2 e2 = [ �� =[�
rl , n
Thus e1, e2 form a basis for ]R2 •
=
,. = [ ] ,, = [ [
both are linearly independent and span JR3 · Hence e1, e2, form
e3 a basis for JR3. •
1 0 0
0 1 0
0 0
e1 = e2 = en=
0 0 0
0 0
Here again the vectors e 1, e2, ••• , e,, are easily seen to both be linearly independent and
span JR" and consequently form a basis for JR".6 •
EXAMPLE 6 Generalizing even further, them x n matrices E;1 that have l in the ij-position and Os
elsewhere for i = l,2, ... ,m and j = 1, 2, ... ,n are linearly independent and span
Mmx11 (1R). Hence they are a basis for M111 x 11 (R). For instance,
£ 11 = [ � � l £12 = [ � � l 1
E2 = [ � � l
form a basis for M2x2 (IR).
write the polynomials in Pn as linear combinations of x", x"-1, ••• , x, I when we write
them as a11x11 + a 11_1 x"-1 + · · · + a1x + ao. Hence x",x"-1,••• , x, I form a basis
for P11 . e
The bases given in each of Examples 3-7 are natural bases to use and are called the
standard bases for each of these respective vector spaces. Standard bases are not the
only bases,however, for these vector spaces. Consider Examples 8 and 9.
6 The vectors e1 and e2 of JR2 in Example 3 are oflen denoted by i and j, respectively; the vectors e1, ez, and
e3 of IR.3 in Example 4 are often denoted by i, j, and k, respectively.
u
88 Chapter 2 Vector Spaces
, [ '. ] + c, [ : ] + c, [ -: ] l
ly independent. Suppose
Solution Let us first show that these three vectors are linear
[ � � -: \ � ]-[ � : -: I � J'
Setting up and reducing the augmented matrix for the resulting homogeneous system,
I I 1:0 00 2 , 0
we see that we have the trivial solution c, = 0, c2 = 0, c3 = 0. �ence thes� vectors
arc
the
linearly independent. Similar work applied to the system of equations resulting from
[ '. ] c, [ : J [ -: J [ � J
vector equation
+
C,
+ c, =
shows us that our three vectors span IR3 . Thus these vectors do form a basis for lR3 . •
S0lutio11 Let us first check to see if these three polynomials are linearly independent. If
C1(x 2 + x - 3) + c2(x - 5) + C3 · 3 = 0,
we have the homogeneous system:
CJ=0
CJ+ C2 = 0
Since this system has only the trivial solution, the three polynomials are linearly inde
pendent. Likewise. the system of equations resulting from
c, (x2 + x - 3) + Cz(x - 5) + 3c3 = ax 2 +bx+ c
has a solution and hence our three polynomials span P2 • (Indeed, if you have a good
memory, you will note that we already did the spanning part in the last example of the
previous section.) Thus we have shown x2 + x - 3, x - 5, 3 form a basis for JR3 • •
Keep in mind that we must have both linear independence and spanning to have a
basis. If either one (or both) fails to hold. we <lo not have a basis.
\ �
-Ct+ C2 = C,
which does not have a solution for all a, b, and c since adding the second and third
equations gives us
0 = b +c.
Since x 2 + x - I, x 2 - x+ I do not span P2 , they do not form a basis for P2. •
EXAMPLE 11 Do
then
Ct - C2 + 2c3 = 0
- C Ct + c2 + 3c3 = 0.
f'r)
0
Since we know that a homogeneous linear system with more variables than equations
always has a nontrivial solution (Theorem I.I), the three given vectors are linearly
dependent and hence I.lo not form a basis for IR.2 . •
There are other ways of characterizing bases besides saying they are linearly inde
pendent spanning sets. Theorem 2.7 describes one other way.
THEOREM 2.7 Suppose that v 1, v2, ..., Vn are vectors in a vector space V. Then v1, v2, ... , Un form a
basis for V if and only if each vector in Vis uniquely expressible as a linear combination
ofv1, v2, ... , V11 ,
Proof First suppose v 1 , v2, .•. , v11 form a basis for V. Let v be a vector in V. Sinee u1, vi, ... ,
Vn span V, there are scalars Ct, c2, ... , c,, so that
90 Chapter 2 Vector Spaces
Since the trivial linear combination of v1, v2, ... , v,, is the zero vector, the uniqueness
property gives us
c, = 0, c2 = 0, Cn = 0.
Hence v1, v2, ... , v,. are linearly independent, which completes our proof. •
l :. ]
the scalars c1, c2, ... , en are called the coordinates of v relative to the basis a of V.
The column vector
[v]c,. =
c,,
we have
we then get
In other words, the coordinates and coordinate vectors relative to the standard basis of JR2
are just the usual coordinates and column vectors. More generally, the same thing occurs
when we use the standard basis for JR". If we use the standard basis for Mmxn(lR), the
coordinates of a matrix relative to it are the entries of the matrix. If we use the standard
basis for P,,, the coordinates of a polynomial in Pn relative to it are the coefficients of
the polynomial. If we do not use standard bases, we have to work harder to determine
the coordinates relative to the basis. Consider Examples 12 and 13.
we see
C3 = 3, c2 = 0, CJ= 4.
•
EXAMPLE 13 Find the coordinate vector of v = x + l relative to the basis y for P2 in Example 9
consisting of x2 + x - 3, x -5, 3.
CJ= 0, C3 = 2.
Hence
•
We can reverse the procedure of translating vectors into coordinate vectors as the
following example illustrates.
EXAMPLE 14 Find v in P2 if
EXI:RCISES 2.3
: n
4
1 2
. [ -� l[� ] · [ -� l [ -� ] basis for JR3 .
H:l
-
15. Show that
l
,. [ -n-[ 4 [
. [� : ], [ �
�J
2
[l
,. -: Ultl
-:
[_� �].[
16. Show that [
-1
1 O
0 2
form a basis for M2x2(lR).
J[ 0
' , -]
I -I
2 0 ]'
l
·· [ _][ _H[ =i l [ � �
[
� �
: ]'[ � � ],[ �
� ] form a basis for M2x3(JR).
-� �
7. [ �: 0
l[� � l[ ] 17 . Show that x2 + x + 1, x 2 x + l. x 2 - I form a
basis for P2.
-
-l
8. [ � 18. Show thatx 3 +x, x 2 -x, x + l, x3 + I forrn a basis
l [ � -� l
l[ -� ]
for P3. •
r
12. Show that [ � J, [ ! ] form a basis for ne .
21. Show that x + l, x + 2�; +j}I not f9rm a basis
forP1• \ "'
�,11
>,- \o-\
:z- c Q,'J.1
[-n, [ i] l[ _:]
.)c.J'"\ � l,� [)
-;:
22. Show th ,t
8) (vk fo, V -[_iJ Use the system oflinear equation solving capabi Iitie, of
.
Maple or another appropriate software package 111 Exer
r-:
b) v ;r(v). 31. a) Show that the vectors
-5
24. If /J is the basis in Exercise 14, find:
8
15 22
13
- 14
13 14
81
II 12
a) (V)p ;f V - J 10 IOI
-9
3 11
-]
53 18
16 77 15
b) V ff(v)p - [-:}
n
3 7
25. If f3 is the basis in Exercise I 7, find:
3 3
a) (v]fi ifv = 2x2 +3.x, 16
15
2
b) V ff[v(p -[
99
88
-49
-68
26. If y is the basis ofExercise 18, find:
a) (v] y ifv=x3 +x2 +x+I, form a basis for IR.6 •
-Pl
b) Find the coordinates of
,;r1v1, 29
4
b)
-9
27. Show that any set of vectors that contains the zero
vector is a linearly dependent set of vectors. 13
28. Show thal if v1, v2, ••• , v,. is a linearly independent 71
set of vectors, then any subset ofthese vectors is also - 51
linearly independent.
relative to the basis in part (a).
29. Suppose v, and V2 arc nonzero vectors in JR3 and L
i, a line in R3 parallel to v1• What are necessary and 32. a) Show that
�uflicient condition� in terms ofthe line L for v 1 and
xs + 3x + I, xs + x4 + x 2 + x + I,
1 2 to be linearly independent?
x6
-x6 + x4 + x3 + 3x2 + x + I,
-
30. Let 1!1 and 112 be veclors in iR3 that are linearly inde
1
3x3 + 2x2 - 2x - l,
linear combinations ofv1 and v2 determine?
x6 + 3x5 + 2x4
2x - x + 2x3 + x 2 - 3x + I
-
b) Find the coordinates of 7x 6 + 6x 5 - 5x4 - 4x3 34. Let v1, v2, ... , Vn be vectors in Rn. Show that
-3x 2 + 2x - I relative to the basis in part (a). v1, v2, ••. , v,, form a basis for !Rn if and only if the
33. W� saw in the solution to Example 11 that the three matrix [v 1 v2 • • • v11 ] is nonsingular.
1. ,:ctors given in this example are linearly dependent 35. Use the result of Exercise 34 and a suitable test for
,n ne . Show that this can be generalized to the fol invertibility of a matrix in Maple or another appro
'owing: If v1 , v2, ••• , Vm are vectors in IR" and if priate software package to show that the vectors in
,;, > n, then v 1, v2, ... , Vm are linearly dependent. Exercise 31(a) fonn a basis for R6 .
LEMMA 2.8 Ifv,, v2, ... , v11 are a basis for a vector space V, then every set of vectors W1, w2, ... , w111
in V where m > n. is line.arly dependent.
wi =at1V1 +a2,v2+···+a,,,v,,
w2 =a12v1 + a22v2 + · · · + a,,zv,,
(2)
Since 111 > n, we know by Theore m l. I that this system has nontrivial solutions. Thus
we have nontrivial linear combinations equal to the zero vector in Equation (I) and hence
w1, w2, ... , Wm are linearly dependent. e
Now we are ready to prove the r esult referr ed to at the beginning of this section,
which we state as follows.
THEOREM 2.9 If vi, u2, ••• , v,, and w1• w2, ..• , w111 both form bases for a ve ctor space V, then n = m.
Proof Applying Lemma 2.8 with v1, vi, ... , v,, as the basis, we must have m :S n or else
wi, w2, ... , w,,, would be linearly dependent. Interchanging the roles of vi, v1, ... , v,,
and wi, w2, ... , w.,, we obtain n :Sm. Hence n = m. e
The number of vectors in a basis (which Theorem 2.9 tells us is always the same) is
what we call th e dimension of the vector space.
DEFINITION !fa vector space V has a basis ofn vectors, we say the dimension
of V is 11.
I dim(V). ,
Thus, for example, since the standard basis e1, e2, .•. , e,, for JR." has 11 vectors.
J dim{JR.") = n. J
Since the standard basis for Mm x11 (JR) consists of the mn matrices E .. with I in the
ij-position and Os elsewhere,
'J
I dim(MmxnOR)) = mn. ,
Since the standard basis x", ... , x, J for P,, has n + I ele ments,
I dim(P,,) = n + 1. J
Not every vector space V has· a b as1s ·
. . · · cons1st · of a ·fimte number of vectors but
· · mg
� �
e c n still rntroduce dimens!ons for such vector spaces. One such case occurs wh;n V
ts the zero vector space consisting of only the zero vector.
The zero vector space does
2.4 Dimension; Nullspace, Row Space, and Column Space 97
not have a basis. fn fact, the set consisting of the zero vector is the only spanning set for
the zero vector space. But it is not a basis since, as you were asked to show in Exercise
27 in the previous section, no set containing the zero vector is linearly independent. For
obvious reasons, if V is the zero vector space, we take the dimension of V to be O and
write dim(V) = 0. The zero vector space along with vector spaces that have bases with
a finite number of vectors are caJled finite dimensional vector spaces. Vector spaces
V that are not finite dimensional are called infinite dimensional vector spaces and we
write dim(V) = oo. It can be proven that infinite dimensional vector spaces have bases
with infinitely many vectors, but we will not attempt to prove this here.7 The set of
all polynomials P is an example of an infinite dimensional vector space. Indeed, the
polynomials 1, x, x2 , x3 , ... form a basis for P. Many of the other function spaces we
looked at in Sections 2.1 and 2.2, such as F(a, b), C(a, b), D(a, b), and cx (a, b), are
infinite dimensional, but we will not attempt to give bases for these vector spaces.
We next develop some facts that will be useful to us from time to time. We begin
with the following lemma.
LEMMA 2.10 Let VJ, v2, ... , v11 and w 1, w2, ... , w,,, be vectors in a vector space V. Then Span{v,, v2,
..., v,,} = Span{WJ, w2, .•• , w,,, } if and only if each v; is a linear combination of
w 1• w2, .•. , w111 and each w; is a linear combination of v 1, v2, ••• , v11 •
Proof Suppose that Span{v1, v2, ... , v11 } = Span {w,, w2, .... w111}. Since each v; 1s m
Span{VJ, v 2, ••• , v,,} (use I for the scalar on v; andOfor the scalaron all otherui to write
v; as a linear combination of VJ, v2, •.. , v,,), v; is in S pan{ w,, W2, .••, Wm}. Hence v;
is a linear combination of WJ, w2, ••• , w,,,. Likewise each w; is a linear combination of
VJ, V2, ... , V,,.
To prove the converse, first note that if eac h v; is a linear combination of WJ, w2,
..., Wm, then each v; is in Span{WJ, w2, ... , wm}. Since subspaces are closed under
addition and scalar multiplication, they are also closed under linear combinations. Hence
any linear combination of VJ, v2, ... , V lies in Span{wJ, w2, ••• , Wm l giving us that
11
Span{VJ, v2, .•. , v,,} is contained in Span(wJ, w2, ... , Wml- Likewise we will have
Span{w 1, w2 , • • • , w111 } is contained in Span{ v,, v2, ..., v,,} so that these two suhspaces
are equal as desired. e
The next lemma tells us that we can extend linearly independent sets of vectors to
bases and reduce spanning sets to bases by eliminating vectors if necessary.
2. If VJ, v2, • . . , vk span V, then there exists a subset of VJ, v2, . . • , Vk that forms
a basis of V.
7 Typical proofs involve using a result called Zorn's Lemma. which you may well encounter if you continue
your study of mathematics.
98 Chapter 2 Vector Spaces
Proof
I. Notice that k � n by Lemma 2.8. If k < n, vi, V2, ••. , Vk can.not span V;
otherwise dim(V) =knot 11. Thus there is a vector Vk+J not m
Span{v,,v2, .•• ,vd. We must have v1, v2, . .• , vk, Vk+t are linearly
independent. To see this, suppose
CJ V1 + C2V2 + · · · + CkVk + Ck+! Vk+t = 0
where c,, c2, .•• , c., Ck+t are scalars. Were ck+t =fa 0, we coul� solve this .
equation for vk+l ohtaining Vk+t is in Span{VJ, v2, · ., � k}, wht�h we know ts
: _
not the case. Now that ck+ J = 0 in this linear combmauon, the l 111ear 111-
dependence of v1, v2, ••• , Uk tells us we also have c1 = 0, c2 = U, ... , Ck = 0.
Hence we have the linear independence of VJ, V2, ••• , Vk+t. If k + 1 < n,
repeat this procedure again. After n - k steps we arrive at a set of II linearly
independent vectorsv1, v2, ... , v,,. These II vectors must span V; otherwise
we could repeat our procedure again obtaining n + I linearly independent
vectorsv1, v2, ••• , v,,,v,,+ 1, which is impossible by Lemma 2.8. Thus we have
arrived at the desired basis.
2. Ifv1, v2, ••• , Vt are linearly independent, they then form the desired basis. If
not, one of them must be a linear combination of the others. Relabeling if
necessary to make the notation simpler, we may assume Vt is a linear com
bination ofvJ, v2, ..• , Vk-l· Since each VJ, v2, ..• , vk is a linear combination
of VJ, v2, ••• , Vk-J and vice versa, it follows by Lemma 2.10 that
v1. v2 •... , vk-J span V. If v1, v2, ... , Vk-J are also linear] y independent, we
have our desired basis; if not, repeat the procedure we just did again. Since
such steps cannot go on forever, we must obtain the desired type of basis after
a finite number of steps. •
At the beginning of this section we noted that the examples and exercises of the
previous section suggested that once one hasis has n elements, all other bases also have
11 elements. We then went on to prove this (Theorem 2.9). Another thing you might notice
from the examples and exercises in the last section is that every time we had n vectors
in a vector space of dimension n, having one of the properties of linear independence or
spanning seemed to force the other property to hold too. The next theorem tells us this
is indeed the case.
J. lf the vectors VJ, v2, ... , v,, are linearly independent, then v1, v2, •.. ,v,, form
a basis for V.
2. If the vectors VJ, v2, •.. , v,, span V, then v1,v2, ••• ,v 1 form a basis for V.
1
Proof
1. By part (I) of Lemma 2.11, we can extend v 1,v2, ••• , v,, to a basis of V. But
since every basis of V has n elements, v1, v2, ••• , v11 must already form a basi s
for V.
2.4 Dimension; Nullspace, Row Space, and Column Space 99
2. By part (2) of Lemma 2.11, we know some subset of the set of vectors
v,, v2, ... , v,, forms a basis of V. Again since every basis has n elements, this
subset forming the basis must be the entire set of vectors v,, v2, ... , v,,. •
Solution Since dim (P2 ) = 3 and we are given three vectors in P2 , Theorem 2.12 tells us we can get
by with showing either linear independence or spanning. Let us do linear independence.
If
C3 = 0
-c1 t Cz + C3 = 0.
It is easily seen that this system has only the trivial solution. Thus x 2 - 1, x 2 + I, x + I
are linearly independent and hence form a basis for P2. •
Of course, notice that Theorem 2.12 allows us to get by witb checking for linear
independence or spanning only when we already know the dimension of a finite dimen
sional vector space V and are given the same number of vectors as dim(V). If we d o
not know dim(V) for a finite dimensional vector space V, we must check both linear
independence and spanning. Notice too that if we do know the dimension of a vector
space V is n, no set of vectors with fewer or more than n elements could form a basis
since all bases must have exactly n elements. For example, neither the set of two vectors
(I, 1, )), (1, -1, 3) nor the set of four vectors (I, I, 1), (1, -1, 3), (0, I, I), (2, I, -1)
could form a basis for the three-dimensional space IR3 .
We conclude this section with a discussion of techniques for finding bases for three
important subspaces associated with a matrix. Recall that the solutions to the homoge
neous system AX = 0 where A is an m x n matrix form a subspace of IR" (Theorem
2.4). This vector space of solutions is called the nullspace or kernel of the matrix A
and we shall denote it by
The manner in which we wrote our solutions to homogeneous systems in Chapter 1 leads
us naturally to a basis for N S(A). Consider the following example.
�l
EXAMPLE2 Find a basis for N S(A) if
A-[
2 -1 3
0 4
4 -3 -2
1 OO Chapter 2 Vector Spaces
[:
2 -] 3 0 I O 1 2 -1 3 0 : 0
0 4 1 \ O ] - [ 0 -I I 1 : 0]
-
4 -3 2 , C.) 0 2 -2 -2 -2 : 0
X5 X5
0]
EXAMPLE 3 Find a basis for the row space of the matrix
2 -1 3
0 4
4 -3 -2
of Example 2.
I -� -! -]
S0Lutio11 From the solution to Example 2, we see the reduced row-echelon form of A is
B-[�
The nonzero rows of 8 span RS(B) = RS(A). They also are linearly independent. (If
CJ [ 1 0 5 2 ] + c2 [ 0
we see from the first two entries that CJ
-I -1 -1 ]
= 0 and c2 = 0.) Hence
�.
=[ 0 0 0 0 0 ], 1
•
[ I 0 5 2 ),[ 0 1 -1 -I -I )
form a b asis for RS(A).
There is a relationship between the dimensions of RS(A) and N S(A) and the num
ber of columns n of an m x n matrix A. Notice that dim(RS(A)) is the number of
nonzero rows of the reduced row-echelon form of A, which is the same as the number of
nonfree variables in the solutions of the homogeneous system AX = 0. Also notice that
dim(NS(A)) is the number offree variables in the solutions of the homogeneous system
AX = 0. Since the number of nonfree variables plus the number of free variables is
the total number of variables in the system AX = 0, which is n, we have the following
theorem.
102 Chapter 2 Vector Spaces
THEOREM 2. 14 If A is an m x II matrix,
dim(RS(A)) + dim(N S(A)) = n.
I CS (A ). I
For the matrix A in Examples 2 and 3, the column space is
In the same manner as we use row operations to find bases for the row space of a matrix
A, we could find a basis for the column space of A by using column operations to get
the column reduced echelon form of A. The nonzero columns of the column reduced
echelon form will form a basis for C S(A). But if you feel more comfortable using row
operations as we authors do. notice that a basis for CS(A) can be found by reducing A T
to row-echelon form and then transposing the basis vectors of RS(A 7) back to column
vectors. This is the approach we take in the next example.
l'=Ul = [ _!J
from which we see
[ I 0 3 [ 0 I -2 r
form a basis for C S(A).
•
2.4 Dimension; Nullspace, Row Space, and Column Space 103
You might note that we actually do not have to go all the way to reduced row-echelon
form to see a basis. For instance, it is easy to see that the first two rows in the second
matrix of the solution to Example 4 would have to give us a basis so that we could equally
well use
I I
always the case; that is, for any matrix A,
dim(RS(A)) = dim(CS(A)).
This common dimension is called the rank of the matrix A and is denoted rank(A). For
instance, if A is the matrix in Examples 3 and 4, then rank(A) = 2. Observe that the
rank of a matrix is the same as the number of nonzero rows (columns) in its reduced row
(column)-echelon form.
Sometimes we have to find a basis for a subspace of JR" spanned by several vectors
of IR.". This is the same as finding a basis for the column space of a matrix as the final
example of this section illustrates.
-l
0
Reducing A T ,
-1
' 0
104 Chapter 2 Vector Spaces
It is apparent that
[ -i] .[; l
form a basis for C S(A) and hence for the subspace spanned by the three given vectors. e
EXERCISES 2.4
1. In each of parts (a)-{d), determine whether the given 4. In each of parts (a)-(d), determine whether the given
vectors form a basis for IR2. matrices form a basis for M2x2(1R).
a) [
� J
b) [ J. [ -� ]
c) [ -:! t [ _
J 1�]
d) [ =� ] ' [ ; ] ' [ -: ]
2. In each of part� (a)-{d). determine whether the given
vectors form a basis for Dt3 .
. 6. [ - �
5 [ : � J _; ]
-
3. In each o� parts (a}-(d), determine whether the given
polynom1als fonn a basis for p2 •
a) x + x -1. 2..t -3,x +x +2
b) 5 - 4x 2, 3 -2x
2 2 2
.
7 [
-
: :: :]
8.
[ :
-l -:]
c) x2 +x-J,x2 +x + 2,xi +x+l4
d) x2 +x,x+l,x2+ 1,I
·-[: : ; =:] 10. [ : -� _: : l
2.4 Dimension; Nullspace, Row Space, and Column Space 105
-� H: H =n
Jn E,erciscs 13-15, find a basis forthe subspace spanned 24. Use the system of linear equation solving capabili
by the given vectors. (These do not have unique an ties of Maple or another appropriate sotiware pack
swers.} age to show that the matrices
r
r H-J[ _J[ =n
l �!
- 3
13.
[ �� -� ] , [ ! _! l
14. -2
i
JUH-�n
32 -9 -1 I
L
r -2
15. 0 0 -4 0
3
form a basis for M3x2(IR).
-n
-4
25. Let A be the matrix
[ -:
16. Without formally showing that there is a nontrivial 3 -1 4
l111ear combination ofx4 -2x, x4 +x3 - l, x 3 +x+3,
7 -5
-1 2 6
x - x 2 - x 4, lOx - 91, ;rrx 4 + J3x 3 - 7x 2 equal A=
to the zero polynomial, we can conclude these poly I I
nomials are linearly dependent. Why is this? 0 12 -5 15
17. Show that det(A) f:- 0 for an 11 x 11 matrix A if and
only if rank(A) = n. a) In Maple, the command nullspace (or
18. Suppose that A is an m x n matrix. Show that if equivalently, kernel) can be used to find a basis
111 > n, then the rows of A are linearly dependent. for the nullspace of a matrix. (The basis vectors
will be given as row vectors instead of column
19. Suppose that A is an m x n matrix. Show that if
vectors.) Use this command or a corresponding
m < n, then the columns of A are linearly depen
command in an appropriate software package to
dent.
find a basis for the nu lispace of the matrix A.
20. Consider the linearly independent polynomials
b) Bases for the row space of a matrix can be found
P1(x) = x 2 +x,p2(x) = x + I. Find a poly with Maple by using the gausselim or gaussjord
nomial p 3(x) so that p 1 (x), p2(x) , p3(x) form a (or equivalently, rre/) commands. Such bases
basis for P2 in the manner of the proofof part (I) of may also be found using the rowspace and
Lemma 2.11 . rowspan commands. Find bases for row space
21. a) Show that the polynomials p 1 (x) =x 2
- I, of the matrix A using each of these four Maple
P2 (x) = x +1, p3 (x) = x - 1, p4(x) =x + I
2 commands or corresponding commands in
span P2 . another appropriate software package. If you
106 Chapter 2 Vector Spaces
are using Maple. compare your results obtained Compare the result you obtain here with your rcsuIts
with these four different commands. in Exercise 2S(b).
c) Bases for the column space of a matrix can be 27. Suppose that v1, v2, ••. , Vk are linearly indep�nde�t
found with Maple by using the gausselim or vectors in IR". Let vk+ 1 be another vector m IR .
,:aussjord commands on the transpose of th� How could the gausselim or gaussjord commands
matrix. Such bases may also be found by using of Maple or corresponding commands in another a�-
.
the co/space and co/span commands. Find propriate software package be used to determme 11
bases for the column space of the matrix A Vk+l is in Span{v1, v2, •.. , vd?
using each of these four methods in Maple or 28. Use the result of Exercise 27 to extend the set con
UH-Il
corresponding commands in another sisting of tbe vectors
appropriate software package. If you are using
Maple, compare your results obtained with
these four different approaches.
26. The basis command of Maple may be used to find a
hasis for the subspace spanned by a finite number of
row vectors in M1 ,n(IR). Use this command or the
corresponding command in an appropriate software to a basis of JR in the manner of the proof of part
4
package to find a basis for the subspace of M1,s(IR) (1) ofLemma2.ll.
spanned by the vectors
-2 3 -I 4 -2 ] , [ 3 -1 2 6 8 ] ,
7 -5 I 4],[0 12 -5 15 8].
2.5 WRONSKIANS
In our work with linear differential equations in Chapter 4 we will have instances where
we will have to determine whether a set of functions forms a linearly independent set.
For instance, problems such as the following will arise: Are the functions given by
Er , cos x, sin x linearly independent? Consider a linear combination of these functions
that equals the zero function:
cie + c2 cosx + c3 sinx = 0.
.r
The only such linear combination is the trivial one. One way to see this is to choose
three values of x such as x = 0, rr/2, rr. Substituting these values of x into our linear
combination, we have the system:
CJ+ Cz =0
e:n:12 c1 + c3 = 0
e" Ct - C2 = 0.
Adding the first and third equations, we obtain (I + e")c, = O from which we get
�1= 0. It then follows that c2 = 0 and c3 = 0. Hence e-", cos x, and sin x are linearly
mdepe�dent. �ut there is another way to arrive at a system of equations involving
denvat1ves that 1s often more convenient to use for showing linear independence.
2.5 Wronskians 107
1
cif1<n-l\x) + czf?- \x) + · · · + c11f?- 1)(x) = 0.
If there is some x in (a, b) for which this system has only the trivial solution, then
/1 , h, ... , !,, will be linearly independent. Having such an x is the same as having an
x in (a, b) for which the the matrix
Since a square matrix is nonsingular if and only if its determinant is nonzero (Theorem
1.21), we have the following theorem.
8 In fact, what wc arc about to do docs not require the interval lo be open: it will work on other types of
intervals (such as closed ones) as well.
9 The Wronskian is named in honor of the Polish-French mathematician Josef Maria Hoene-Wronski ( 177&---
1853).
108 Chapter 2 Vector Spaces
Let us use the Wronskian to show eX , cosx, sinx are linearly independent (with
(-oo, oo) for the interval).
S0lutio11 We have
cosx sin x
w(e·',cosx,sinx)= ex
-sinx cosx
e - cos x -sin x
x
2 2
= e (sin x + cos x) - cos x(-e sin x - e cosx)
x x x
Since 2e' is not zero for some x (in fact, it is not zero for every x), we have that
e-<, cos x, sin x are linearly in dependent by Theorem 2.15. •
Be careful not to read too much into Theorem 2.15. It only tells us that if the
Wronskian is nonzero for some x, then the function s are linearly independent. It does
not tell us that the converse is true; that is, it does not tell us that linearly independent
functions have their Wronskian being nonzero for some x (or equivalently, that if the
Wronskian is zero for all x, then the functions are linearly dependent). In fact, the
converse of Theorem 2.15 does not hold in general. Here is an example illustrating this.
EXAI\IPLE 2 Show that the Wronskian of the functions f and g where f(x) = x2 and g(x) = xlxl is
zero for every x and that f and g are linearly independent on (-oo, oo).
Solution To calculate the Wronskian off and g, we need their derivatives. This is easy for f. To
I
get the derivative of g notice that we can also express g(x) as
x2 if X � Q
g(x) =
x2 if X < Q
-
The graph of g appears in Figure 2.3.
I
We have
2x if X > 0
g'(x) =
-2x if X < Q
Atx = 0,
Figure 2.3
If x 2: 0,
g'(x) =
I 2x if X
-2x if X < Q
� 0
I x2 x2
w(f (x), g(x)) =
2x 2.x
I =0.
If X < 0,
I x2 x2 I
w(f(x), g(x)) = = 0.
2x -2.x
Hence we have that the Wronskian off and g is zero for every x.
To see that J and g are linearly independent, suppose that
J
where c 1 and c2 are scalars. Substituting x = I and x = -1 into this equation. we arrive
at the system
CJ +c2 = 0
CJ - C2 = 0,
which has only the trivial solution. Hence f and g are linearly independent as desired. •
-
I HI Chapter 2 Vector Spaces
EXERCISES 2.5
In each of Excrci�cs 1-9. show that the given functions 14. Suppose that a< c < d < band that /1, h, ... , f�
arc linearly independent on (-oo, oo). are functions in F(a, b).
I. e3', e-2.x a) If /1 , Ji, ... , fn are linearly independent on
2. cos 5x. �in 5x (a, b), are /1, h, ..., fn necessarily linearly
independent on (c, d)? Why or why not?
3. e ix co�x. eix sin x
b) lf !1 , h, ... , f,. are linearly independent on
4. e-•, xe-• (c, d), are /1 , h, ... , f,. necessarily linearly
5. x 2 - I. x2 + I.x+ I independent on (a, b)? Why or why not?
6. e-' , e h, e 1' IS. a) Find the Wronskian of I, x, x 2 , ••• , x•- 1•
7. e�x. xe4 '. x 2e 4' b) Show that
8. e·', e·' cos x. e' sin x w(g(x)f1 (x), g(x)h(x), ... , g(x)f,,(x))
9. x3 . lxl-'
= [g(x)]"w(/1 (x). h(x), ... , f,. (x)).
10. Show that I/x, x arc linearly independent on (0, oo).
c) Show that e'-',xe'·', x2 e"x, ... , x"- 1 e'·' are
11. Show that x + I, x - I,x arc linearly dependent on
linearly independent on (-oo, oo).
(-00, 00).
· 1 x, cos2x, cos 2x are linearly depen- 16. Use the wronskian and det commands in Maple or
I'-· Sh
. ow th at sm
appropriate commands in another software package
do.:nt on (-oo. oo).
to find the Wronskian of the functions e3x , xe3x,
13. Show that the two functions in Example 2 are lin e3.:r cos 2x, e3x sm
· 2 x,xcosx, x smx. Are these
early dependent on [0. oo). functions linearly independent?
First Order Ordinary
Differential Equations
In Chapters 1 and 2 you were introduced to linear algebra. A field of mathematics that has
a deep relationship with linear algebra is the field of differential equations, as you will see
in Chapters 4, 6, and 7. A differential equation is an equation that involves a function, its
variables, and its derivatives. Such equations arise in modeling many physical situations.
Here is one you probably have encountered in your calculus courses:
It is well documented that most radioactive materials decay at a rate
proportional to the amount present. What is an equation describing
this activity?
Suppose we use Q(t) to denote the amount of radioactive material p resent at time t. The
fact that "the materials decay at a rate proportional to the amount present" means that
the rate Q'(t) is a constant (called the constant of proportionality or the constant of
variation) multiple of the amount Q(t). Moreover, since Q' (t) is negative because Q(t)
is decreasing, the constant of proportionality must be negative. Thus we can model this
activity by the equation
Q'(t) = -k Q(t)
where k > O is a constant, which is a differential equation since it involves the derivative
Q'(t).
As we proceed through this and later chapters, we wi II see several other examples
of phenomena modeled by differential equations. However, we will merely scratcb
the surface on applications of differential equations in this text. There are widespread
applications of differential equations throughout the sciences, engineering, and finance
and economics that must necessarily be left for later courses. In thjs text we will study
some of the mathematical aspects of differential equations and learn how to model some
111
112 Chapter 3 First Order Ordinary Differential Equations
.
. t'10ns usi·no"' differential equations. We begin this
appI 1ca chapter with a general discussion
.
sections
.
deal with techm
·
ues �or fi ndmg
·
of differential equations in the first section. Later _ �
tions of them.
and approximating solutions of differential equations and some applica
y = x2 + y 2, (2)
'
and
2xy dx + (x 2 + 3y2 ) dy = 0
f
in dif erential fo1111.
A solution of a differential equation is a function that satisfies the equation. For
example, Y = x 2 /2 is a solution toy' = x. We leave it for you to check that any function
of the form Q (t) = C e-kr where C is a constant is a solution of the differen equation
, tial
Q (t) � -_kQ(t) �odeling radioactive decay given in the introduction to
this chapter.
As a thml 1llustrat10n, you can easily verify that y = e-x
and y = e3x are solutions to
y" - 2y' - 3y = 0.
A type of differential e4uation that you already know how to solve
is one of the
form
y' = f(x).
3.1 Introduction to Differential Equations 113
y = f f(x) dx +c.'
As an illustration, if
y'=x,
all the solutions to this differential equation have the form
f
x2
y= xdx+C= +c.
2
A general form of a function giving us all the solutions to a given differential equation
is called the general solution of the differential equation. Thus, for instance, the general
solution of a differential equation of the fon11 y' = f(x) is y = f(x) dx + C. InJ
particular, the general solution of y' = x is y = x2 /2 + C. Using techniques presented
in the next section, we will be able to obtain that the general solution of Q' (t) = -k Q(t)
is Q(t) = ce-k1 where C is a constant. Using techniques presented in Chapter 4, we
shalJ see that the general solution of y" - 2y' - 3y = 0 is y = c 1e x- + c2 e1r where c1
and c2 are constants.
A first order differential equation along with a condition specifying a value Yo of
y at a particular value xo of x is called a first order initial value problem. We will
indicate such an initial value problem by writing the differential equation along with the
condition in a manner illustrated by the following three examples:
y'=x, y(1)=2;
2xy dx + (x 2 + 3y 2 ) dy = 0, y(l) = l.
While there are infinitely many solutions given by the general solution y == x2 /2+ C
to the differential equation y' = x, there is only one solution also satisfying the initial
condition y(l) = 2. Indeed, substituting x = I into the general solution and equating
with y( 1) = 2 gives us
l
y(l) = + C = 2.
2
Solving for C we get C = 3/2 and hence y = :x2 /2 + 3 /2 is the one and only one
solution to the initial value problem y' = x, y(l) = 2.
As another illustration, consider the initial value problem
Q'(t) = -kQ(t), Q(O) = l.
I In this book we use the symbol f f(x) dx to indicate one antiderivative of f(x). For example, J x dx =
x2 /2.)
114 Chapter 3 First Order Ordinary Differential Equations
ns to this d�;e�ential
We have noted that functions of the form Q(t) = ce-k are solutio
1
,
= 1. H nce Q(t) = e. 1s one
equation. The initial condition gives us Q(O) = C �
another solut1on, w(t).
solution satisfying this initial value problem. Suppose there ts
Then
- (t))
d [Q(t)J- Q'(l)w(t) Q(t)w'(t) = -ke-k'w(t) - e :- (-kw
1
= O,
dt w(r) - [w(t)]2 [w(t)]
since w'(t) = -kw(t). Therefore, w e know that Q(t)/w(t) is a constant. Since
Q(O)/w(O) = 1. the constant must be I and Q(t) = w(t). Once again, we see that
imposing an initial condition results in one and only one solution.
The fact that both of the initial value problems y' = .x, y(l) = 2 and Q'(t) =
-kQ(t), Q(O) = I had unique solutions illustrates the following theorem whose proof
will be sketched in Section 3.8.
THEOREM 3.1 Let a. b > 0 and suppose f and Bf/By are continuous on the rectangle Ix - xol < a
and IY - Yol < b. Then there exists an h > 0 so that the initial value problem
2 In Maple diff
(y (x) ,x) is used for d\'/dx or y'. Notice that you must explicit
ly indicate that y is a
function of x.
3.1 Introduction to Differential Equations 115
y(x)
y(x)
(a) (b)
y(x)
y(x)
12
7
6
5
4
3
-12
(c) (d)
Figure 3.1
EXAMPLE 1 y' == y - y 2
If ·Y > 1, we see that y" > O, so th e s. oIutt_. ons are concave up .m this
. . .
· · regi on. In the reg10n
0 < ·v < 1 we see that y' > O, hence Y is · · ·
mcreasrng 111 this region In this region we
see that y" > 0 if y < 1/2 an·d Y < o for y > 1/2. Hence in the regi
. II
on O < y < I ,
solutions are concave up for Y < 11,.,� an . cl concave down for ·y > 1/2· In the region
Y < 0• we see that y' < O and Y < O so th at solut10ns
II ·
. . . . are dec reasing and concave down
m this regwn. In Figure 3 · 2 we have used Maple to generate a phase portrait makrng
3.1 Introduction to Differential Equations 117
sure to use initial conditions that give the equilibrium solutions and solutions in each of
the regions by using the initial conditions y(O) = k/2, k = -1, 0, 1, 2, 3. This graph
was obtained by typing and entering
phaseportrait(diff (y(x) , x) =Y (x) -(y (x) ) /\2,y (x) ,X=-2 ..2,
[seq( [y(O}=k/2] ,k=-l .. 3)] ,y=-2..2,arrows=none);
at the command prompt. Notice how we restricted y so that -2 � y � 2. Initially, we
tried to get Maple to draw these phase portraits without including this restriction on y but
got an error message. After some experimentation, we found we could fix the problem
by introducing this restriction. •
y(x)
Figure 3.2
EXAMPLE 2 y' = xy
Solution We see that y = 0 is the only equilibrium solution. In the region y > 0, we see that y is
increasing when x > 0 and y is decreasing when x < 0. Since
y" = xy' + y = x y + y = y(x 2 + 1)
2
we see that y" > 0 if y > 0. Hence solutions are concave up when y > 0. In the region
y < O, we see that y is decreasing when x > 0 and y is increasing when x < 0 and that
y" < O; hence solutions are concave down on this region. These features are illustrated
by the pbase portraits in Figure 3.3 obtained by typing and entering
phaseportrait(diff(y(x) ,x)=x*y(x) ,y(x) ,x=-3 .. 3,
[seq( [y(O}=k/2] ,k=-2 ..2)] ,y=-3..3,arrows=none);
at the command prompt. Maple did give us these phase portraits without including the
restriction -3 s y :S 3. However, the five curves were closely bunched about the origin
since a scale containing large values of lyl was used. We then decided to "zoom in" by
Equations
118 Chapter 3 First Order Ordinary Differential
y(x)
Figure 3.3
EXAMPLE 3 y' = x2 + y2
Sul11tio11 If y(x) = C is an equilibrium solution, then y 1 (x) = 0 = x 2 + C 2 for all x, which is
impossible. Therefore, there are no equilibrium solutions. Since y' > 0 except at (0, 0)
y is always increasing. The second derivative is
y = 2x + 2yy' = 2x
11
+ 2y(x + y2) = 2(x + y(x + y2)).
2 2
We see that y" > 0 if x > 0 and y > 0, and hence solutions are concave up in the
first quadrant. Similarly, y" < 0 in the third quadrant, and hence solutions are concave
down in this quadrant. In the second and fourth quadrants, the concavity is determined
by whether or not x + y(x2 + y2 ) is positive or negative. Figure 3.4 illustrates these
properties. These phase portraits were obtained by typing and entering
phaseportrait(diff(y(x) ,x),X)=xA2 +(y(x) )A2,y(x),x=
O.8 ..0.8, [seq(y (0) =k/2], k=-3 .. 3) J ,y=-2 .. 2, arrows=none);
at the command prompt. As in Example 1, we found ourselves having to include a
restriction on y in order to avoid an error message. •
EXAMPLE4 y1 = x sin y
y(x)
Figure 3.4
y(x)
Figure 3.5
EXERCISES 3.1
In Exercises 1-4, determine the order of the differential In Exercises 5 8,- determine whether the given function
equation. is a solution of the differential equation. If this function
is a solution, determine whether it satisfies the indicated
I. x2 - 3y + 4(y") 3 = 0 initial condition.
In Exercises 9-12, determine the slopes of the tangent 19. Show that both Jl - e-2x and -.JI -=-r2x satisfy
lines to the solutions of the differential equation at the the initial value problem y' = -y + Y , y(O) = 0.
_ Explain why this is not a contradiction to Theorem
given points. Also use Maple (or another appropn�te
software package) to graph a direction field for the dif 3.1.
ferential equation. 20. Show that for every real number c,
9. y' = 2xy; ( I. 0). (2, 4), (-2, -2) </J(x) = (x + cJx)2
JO. y' = y cos x; (0, 2), ( 3. 2), (-3, -5)
satisfies the initial value problem
J 1. y' = xy2 + 2x; (1.-1), (4, 2), (-4, 2)
12. y' = u• y -2 sinx; (0, -3), (rr /2, 0), (-11"/4, -1) y' = y/x + y 12, y(O) = 0.
1
24. (x2 - y2 ) dx + xy dy = 0
13. y' = 2y + 4 14. y' = xy + 4x
25. y' = xy2 +2x, y(I) = -1
15. y' = y2 -
4y 16. y' = ycosx
26. 2 y sin(xy) dx + (2x sin(xy) + 3y2 ) dy = 0,
17. y' = - 9y
y3
18. y' = xy -xy 3 y(O) = I
y' = M(x)N(y)
or
dy
= M(x)N(y)
dx
we obtain an equation in x and y that implicitly gives us the solutions to the differential
equation. Following are three examples illustrating this process.
EXAMPLE 1 Solve
y =ayI
where a is a constant.
dy
Solution Replacing y' by , our equation becomes
dx
dy
-=ay.
dx
Dividing by y and multiplying by dx, we have
dy
- =adx.
y
Integrating each side,
f d: =fa dx,
gives us
lnJyl=ax+C.
We will usually leave our answers as an equation in x and y, but let us solve for y
in this example. Applying the exponent function to each side, we have
eln IYI = eax+C.
This gives us
Hence
EXAMPLE 2 Solve
y' = xy - 4x.
122 Chapter 3 First Order Ordinary Differential Equations
•
After integrating each side, we find
xi
In IY - 41 = - t C.
2
------------------
3.2 Separable Differential Equations 123
Figure 3.6
EXAMPLE 4 Solve
x2
lnlyl =sinx- +c.
2
Since our initial condition requires y to be positive, we drop the absolute value on y and
have
x2
lny = sinx -
2 +c.
Substituting in the initial condition x = 0 and y = I gives us
In I = sin O -0 + C,
from which we see C = 0. The solution is given by th e equation
x2
lny = sinx --
2
•
124 Chapter 3 First Order Ordinary Differential Equations
EXERCISES 3.2
In Exercises 1-4, determine if the differential equation 16. ye' dy-secy dx =0, y(O) =n
is separable. 17. a) Solve the equation in the answer to Exercise 13
1. y' = X COS)' t 2xy2 for y.
2. 3. xy2 clx - 6siny dy = 0 b) Use the result of part (a) to determine the values
d)•
3. - =x2 +2xy 4. sin xy' =xy+4x of x for which the solution to the initial value
dx problem in Exercise 13 is valid.
Solve the differential equations in Exercises 5-12. 18. a) Solve the equation in the answer to Exercise 14
5. y' == 2x 2 y - 4y 6. e·• y dx+4y 3 dy == 0 for y .
d b) Use the result of part (a) to determine the values
7. (x2+ I) y ==xy2 +x
dx of x for which the solution to the initial value
8. sec xy' -xy/(y + 4) = 0 problem inExercise 14 is valid.
9. e-< (y2 - 4y) dx + 4 d y = 0
19. Use Maple (or another appropriate software pack
dy dy dy
10. - = x y +3xy
2 3 3
11. t - + - =te
Y age) to graph the solution in Exercise 13.
dx 2
2x y dI clt
20. Use Maple (or another appropriate software pack
dr . age) to graph the solution in Exercise 14.
12. (r
2
+ I) cos O == r sm 0
d()
21. a) Show that the equation y' = (y + x + 1) 2 is not
Solve the initial value problems in Exercises 13-16.
separable.
13. y' + x2 / y = 0, y(O) = I
dy
b) Show that the substitution v = y + x+ l
14. 3 - = 2xy -y, y(2) = I converts the differential equation in part (a) into
dx
dy 4xy a separable one.
15. dx = y2+4 ' y(I) = I c) Solve the differential equation in part (a).
In the last section we learned how to solve separable differential equations. In this section
we will consider a type of equation whose solutions arise from implicit differentiation.
If we use the Chain Rule to differentiate the equation
F(.x, y) =C
implicitly with respect to x,
That is, if Equation (3) is exact, then Equation (4) is true. Amazingly enough, the
converse of this also holds so that we have the following theorem.
'.iHEOREM 3.2 Suppose the functions M, N, My , and Nx are continuous on a rectangular region Q.
I I
Then
M(x,y)dx+N(x,y)dy=O
My (x, y) = Nx(x, y)
on Q.
We are going to sketch the proof of the converse part of Theorem 3.2. Before doing
so, we point out that the rectangular region is needed in Theorem 3.2 to guarantee the
existence of antiderivatives.3 We will assume that antiderivatives exist whenever we
need them. A rigorous proof of the c onverse part of Theorem 3.2 that includes the
construction of the necessary antiderivatives is left for other courses.
To prove the converse part, we must construct a function F so that
Fx(x,y) = M(x, y), Fy (x, y) = N (x, y).
3 Jn fact, Theorem 3.2 holds under the weaker assumption that the region is simply connected. Roughly
speaking, a simply connected region is one that is a single piece and contains no holes.
126 Chapter 3 First Order Ordinary Differential Equations
The function h(y) is the constant of integration, which may involve y since Y is treated as
a constant in this integral. This gives us (up to a yet to be determined h (y)) a function F
so that F, (x, y) = M(x,y). Now we need to detennine h(y) so that Fy (x, y) = N (x, y)
for the function F(x, y) constructed in Equation (5). This means we must have
of=
ay
!_
ay
f M(x, y) d.t +h'(y) = N(x, y).
A theorem from calculus allows us to interchange the order of the partial derivative and
integral in this equation. Doing so, we get
a (N(x, y) -
ax
f My (X, y) dx) = N,(x, y) - My (X, y),
, ex cosy-2xy
y = ex sin y + x2 •
� t is_easily �een that this equation is not separable. Let us see if it is exact. We rewrite it
Solution
m d1fferent1al form obtaining
F(x, y) = ex cosy-x2y.
Therefore, solutions are given by
ex cos y - x 2 y = C
where C is an arbitrary constant. •
The proof of the converse part of Theorem 3.2 may also be done by integrating with
respect to y first. We leave the proof of this approach as an exercise (Exercise 18). The
next example illustrates how this metbod looks in practice.
Since
g(x) = - cosx.
We now have
•
3y 2 ex - x 2 y - cos x = 11.
The graph of the solution appears in Figure 3.7.
2
1.8
0.6
0.4
0.2
Figure 3.7
Notice that
a
-M=3x+2y and -N=2x+y
a
ay ax
so that this equation is not exact. Suppose we try to solve this equation as we did in
Example 1. We have
Next,
a
-F 3 2
= -x +2xy+h (y).
ay 2
However, there is no function h(y) that would make this equal to N = x 2 + xy, for were
this the case,
which implies
1
hI (y) = - .x2 -xy.
2
But this is impossible since the right-hand side depends on x and is not a function of y
alone. We will see how to solve an equation such as this in Section 3.5.
EXERCISES 3.3
In Ext:r..:ises 1-10, determine if the differential equation 13. 2y sin(xy) dx + (2.x sin(xy) + 3y2 ) dy = 0,
is exact. lf it is exact, find its solution. y(O) = I
(1 - X+,)
1. (3x 2 - 4y2) dx - (Sxy - 12y3) dy = 0
dx = 0,y(O) = I
y
2. (3xy + 4y2) dx + (5x 2y + 2x 2 ) dy = 0 14. dy + 2
+Y X +y2
3. (2.xy + y x) dx + (x2 + x ) dy = 0
e e 15. Use Maple (or another appropriate software pack
4. (2.xe'Y + x2 xy - 2) dx + x 3 xy dy = 0
ye
e
age) to graph the solution in Exercise 11.
S. (2x cosy - x2) dx + x 2 sin y dy = 0 16. Use Maple (or another appropriate software pack
age) to graph the solution in Exercise 12.
6. (y cos x+3e' cosy) dx+(sin x-3ex sin y) dy = 0
1 - 2xy , _ x 2 - 2xy + l
17. Show a separable differential equation is exact.
7• y, = 8" y -
.x2 .x2 _ y 2 18. Show the converse of Theorem 3.2 can be proved by
(} -r cos/I
integrating with respect to y first.
ds _ e - 2t coss
1
dr
9. _ 10.-=---- 19. Determine conditions on a, b, c, and d so that the
dt e" - t2 sins d(} r + sin/I
differential equation
In Exercises 11-14, find the solution of the initial value
problem. , ax + by
y =
11. 2xy dx + (x2 + 3y2) dy = 0, y(l) = 1 cx +dy
is exact and, for a differential equation satisfying
12. y' = 2xe - 3x y y(I) = 0
Y 2
these conditions, solve the differential equation.
x3 -x2eY '
130 Chapter 3 First Order Ordinary Differential Equations
I I
An ordinary differential equation that can be written in the form
1
q1(x)y +qo(x)y=g(x)
is called a linear first order differential equation. If g (x) = 0, the differential equation
is called a homogeneous linear differential equation. We shall assume that the functions
q 1, q0, and g are continuous on some open interval (a, b) throughout this section. We
will also assume thatq 1 (x) ¥= 0 for all x in this interval (a, b). Doing so allows us to
I I
divide this differential equation by q 1 (x). This puts the differential equation in the fonn
where p(x) = q0(x)/q1 (x) and q(x) = g(x)/q 1 (x), which is the form we will work with
as we solve first order linear differential equations in this section. These assumptions
on q,, qo, and g also will ensure that the hypothesis of Theorem 3.1 is satisfied and that
the initial value problem
by u gives us
Iu = ef p(x) dx. j
With this u our equation takes on the form
d
-
dx
(uy) = uq(x).
Integrating this equation with respect to x, we obtain
uy = J uq(x) dx+C.
Solving for y,
We call the function u an integrating factor because it allows us to solve the differential
equation by integrating.
Using our formulas for u and y, we can write the solutions to a linear differential
equation as
In general, we w ill not use this formula. Instead, we will determjne the integrating factor
u, multiply the differential equation by u, and then integrate. We illustrate the procedure
with some examples.
132 Chapter 3 First Order Ordinary Differential Equations
Thus
e-2xy = f xe-2x dx + C.
e-2xy =
1 xe-2x
-2 -t-
1 2x
+c.
yI - -y =0
1
X
for X > 0.
u = ef -� dx = e-lnx = _!_
Multiplying through by this integrating
factor gives us x
1 , I
:;-Y - x2y =0.
3.4 Linear Differential Equations 133
I 11
Therefore
I
j,
!!__ (� y) = 0
I dx X
so that
I
-y = C or y = Cx.
X
•
You might notice that the differential equation in Example 2 can also be solved as
a separable equation. 1,
EXAMPLE 3 Solve the initial value problem
y=xy-x, y(l) = 2.
: I
I
II II
Solution We see that this equation is both separable and linear, but we will solve it as a linear
equation. Writing the equation as
I
yI -xy = -x,
and
• y = I+ Cex 12 .
2
An integrating factor is
ef � dx = x 2
--�Clill
I� ---.a--
----------------
J 34 Chapter 3 First Order Ordinary Differential Equations
Using the initial condition, we find C = -1 and consequently the solution to the initial
•
value problem is
y=x-x -2 .
y' +xy = I.
We find that an integrating factor is
J.
of the form F(x) = J(t) dt fore /2:
2
We now have
so that
ex2I2y = J{ er /2 dt + C.
"'
2
o
Using the initial condition y(O) = 1 gives us
o
y(O) = lo e,2 12 dt + C = 0
+C = C = 1.
3.4 Linear Differential Equations 135
Therefore we have
or
•
We can use Maple to graph the solution to Example 5 by typing and entering
plot (exp (-xA2/2) * int (exp (t/\2/2) , t=O .. x} +exp (-xA2/2} , x=O .. 5) ;
This graph appears in Figure 3.8.
y
Figure 3.8
EXERCISES 3.4
dv
In Exercises 1-12, determine the integrating factor and 11. __::_ + e1 y = e1
dt
solve the differential equation.
dr
1. y' + y / x2 = 0 2. y' = yIx -2, x > 0 12. = r tan O + sm (), 0 < e < n/2
d ()
3. y' - 2xy = x 4. y' = 4y + 2x
In Exercises 13-18, solve the initial value problems.
5. y' = 1 + _Y_ x > 0 13. y' + 4y = 2, y(l) = 2
1 +2x'
14. 2xy' + y = I, y(4) = 0
6. y' = xz - 2xy
I +x 2 IS. y' + _Y_ =2, y(O) = 2
x+l
7. xy' + y = x2 , x > O 8. xy' + y = x2 , x < 0 16 . y'+ 2xy = 1, y(O) = -1
9. (I + xy) dx - x2 dy = 0, x < 0 17. y' = cos2x -y/x, y(n/2 ) = 0
10. (I+ xy) dx - x2 dy = 0, x > 0 18. x(x + l)y' = 2 + y, y(l) = 0
L36 Chapter 3 First Order Ordinary Differential Equations
y' + p(x)y = q(x), y(xo) = Yo 22. Use the result of Exercise 21 to do Exercise 13.
In the previous t hree sections we considered three types of first order differential equa
t ions:separable equations, ex act equations, and linear equations. If a first order equation
is not of one of these types, there are methods that sometimes can be used to convert the
equation to one of these three types, which then gives us a way of solving the differential
equation. In t his section we will give a sampling of some of these techniques.
I I
The first approach we consider involves attempting to convert a differential e1,u at ion
that is not exact into an exact one. The idea is t o try to find a function I of x and y so
that when we multiply our differential equation in(!) by J(x, y) obtaining
Numerous techniques have been devised in attempts to find integrating factors that satisfy
I I
the above condition. One ploy is to try to find an integrating factor of the form
l(x, y) = xmyn
where m and n are constants. The following example illustrates this approach.
With
the same as
for Equation (4) to be exact. At this point we make the observation that My (x, y) and
N,(x, y) will be the same if the coefficients of the xm+2y" terms are the same and the
coefficients of the x my •+1 terms are the same in these two expressions. This gives us
the system of linear equations
n+l=m+3
n + 2 = 2(m + 1),
which you can easily solve finding
m =2 and n =4.
Substituting these values ofm and n into Equation (4), we have the exact equation
(x 4 y5 + x
2 6
y ) dx+ (x5 y4 +2x 3y5 ) dy = 0.
138 Chapter 3 First Order Ordinary Differential Equations
I
Let us now solve this exact equation. We have
4 5
F(x,y) = (x y + x2y6 ) dx + h(y)
3 6
= :2'._ + X y
5 5
+h(y)
5 3
and
= x 5y4 + 2x3 y + h (y) = X Y + 2x Y
5 5 4
I 3 5
Fy(X, y)
so that
h'(y) =0
and
h(y) = 0.
•
Do not expect the technique of Example I to always work. Indeed, if you were to
try to apply it to the equation
(x
2
+ y2 + I) dx + (xy + y) dy = 0,
you would find that the resulting system of equations in n and m obtained by comparing
coefficients a s we did in Example 1 has no solution. (Try it.) There is, however, a way
of finding an integrating factor for this equation. It involves looking for one that is a
function of x only; that is, we look for an integrating factor of the form
I /(x, y) = f(x).1
Indeed, Equation (3) tells us the differential equation is exact if and only if
a a
a (f(x)r(x,y)) = ax (f(x)s(x, y))
y
or
f'(x) = ry(x,y)-sx(x,y)
f(x )
s(x,y )
3.5 More Techniques for Solving First Order Differential Equations 139
du
-- ry (x,y)-s;i;(..t,y)
u (5)
dx - s(x,y)
If the expression
ry (x, y) - sx(x,y)
(6)
s(x, y)
is a function of x only, Equation (5) is a separable equation that can be solved for
u = f (x). In other words, we have just arrived at 1he following procedure for finding
an integrating factor that is a function of x:
1. Calculate the e,c.pression in (6).
2. If the expression in (6) is a function of x only, solve the separable differential
equation in (5) to obtain an integrating factor u = f(x).
3. Use this integrating factor to solve the differential equation.
Of course, our procedure does not apply if the expression in (6) involves y. A similar
approach can be used to attempt to obtain integrating factors that are functions of y only.
See Exercise 8 for the details.
Let us use the procedure we have just obtained to solve a differential equation.
Solution With
r(x, y) = x 2 + y2 + l and s(x, y) = xy + y,
we have
ry (X, y) - Sx(X, y) 2y - y
=
=
s(x, y) xy + y x + 1
is a function of x only. Now we solve the separable equation
du I
-=--•U
dx X +l
for the integrating factor u. This gives
du I
-=--dx
u x+I
In u = ln(x + 1)
or
140 Chapter 3 First Order Ordinary Differential Equations
(We leave off the constant of integration since we only have to find one solution for u
to obtain an integrating factor. ) Multiplying the differential equation in the statement of
this example by u = x + 1 we have the exact equation
(x 3 + xy + x+ x 1 + y + 1) dx +(x 2y +2xy + y) dy = 0.
1 2
Solving this exact equation, we find that our solutions are given by the equation
y2
x4 x2y2 x2 x3 2 x+ = C. •
+ + + x y +
4+-- 2 2 3 2
I I
A first order equation of the form
M(x,y)dx+N(x,y)dy=O
for all values of a for which these fractional expressions are defined. The differential
equation
(x1 + y2 ) dx + xy dy =O
M(x,y) dx + N(x,y) dy = O
has homogeneous coefficients. To solve it, we first write it in the fonn
dy M(x,y)
-=---- (8)
dx N(x, y) ·
I! Ii
and hence our differential equation has the form
dy M(l, y/x)
-= N(l, y/x)
dx
In other words, our differential equation can be expressed as one where we have dy/dx
as a function of y/x:
dy
= f (!). (10)
dx X
Let us set
(11)
1,
Since y = ux,
d
y
- = u+x -.
du
(12) ,t
'I
dx dx
I
Substituting the results of Equations ( 11) and ( 12) into Equation (10), we have
du
I,
u +x - = f(u). II
dx
This is a separable equation in u and x which we can solve for u and then use the fact
that y = ux to find our solution for y. The following example illustrates the procedure
we have just obtained.
I (x 2 - y 2 ) dx + xy dy = 0.
�-�----���------------------------.JII-
I
142 Chapter 3 First Order Ordinary Differential Equations
Making the substitutions from Equations (11) and (12), we arrive at the differential
equation
du
u+x -=u--.
dx u
where K is a constant.
X
•
The differential equation in Example 3 also may be solved by using the techniques
of Examples 1 or 2. Try doing it these two other ways. Among the three methods for
solving it, which do you find the easiest to use?
For our final technique, we consider a type of equation that can be converted to a
linear dif erential equation. An equation of the form
I I
f
I I
differential equation by making the substitution
V = yl-11· (13)
y' + xy = xy1.
Solution Making the substitution in Equation (13), which is
V = y-1
here, our differential equation becomes
v' -xv= -x
and hence
y
I
= v -1 = I + Cexz/2 . •
EXERCISES 3.5
In each of Exercises 1--6, find integrating factors for the Use the result of Exercise 8 to find an integrating fac
differ- ,;Li,tl equation and then solve it. tor and solve the differential equations in Exercises 9
and 10.
1. (x2+ y2 ) dx - xy dy = 0
9. (xy +x) dx+ (x 2 + y2 - I) dy = 0
2. (2y2 - xy) dx + (xy - 2x2 ) dy = 0 10. y dx + (2x - ye>' ) dy = 0
dy
3. (x - 3x 2y)- = xy2 + y In each of Exercises 11-14, show that the differential
dx
equation has homogeneous coefficients and then solve
4. (x2 + y2 )y' = 2xy it.
5. (2x 2 + 3y 2 - 2) dx - 2xy dy =0 lJ. (x 2 - y 2 ) dx +xy dy = 0
6. (3x +2eY ) dx +xeY dy = 0 12. (x 2 +y2 ) dx +xy dy = 0
7. Show that multiplying the differential equation 13. (y +xe>"fx) dx - x dy = 0
14. (xcos ! -ysin ! ) dx+xsin l'. dy=O
y' + p(x)y = q(x) by ef p(x) dx X X X
15. Show that if the differential equation
converts the differential equation into an exact one.
8. Show that if M(x,y) dx + N(x, y) dy = 0
Sx (X, y) - ry (x, y) has homogeneous coefficients, then
r(x, y)
l(x.y) = ------
is a function of y only, then a solution to the separa xM(x, y) + yN(.x, y)
ble differential equation
is an integrating factor for the differential equation.
du sx ( x, y) - ry (x, y) 16. Use the result of Exercise 15 to solve the differential
-= u
dy r(x,y) e4uation in Exercise 11.
is an integrating factor for the differential equation Solve the differential equations in Exercises 17 and 18.
r(x, y) dx +s(x, y) dy = 0. 17. xy' + y = y2 18. y' = 2y - y3
1-14 Chapter 3 First Order Ordinary Differential Equations
bacteria.
A similar analysis may be used to get a formula modeling the amount of a radioactive
substance in tenns of its half-life.5 To illustrate, suppose that we have a samplecontaining
radioactive carbon-14, which has a half-life of about 5730 years. By considenng the
amounts of radioactive carbon-14 after 5730 years, 2. 5730 years, 3 . 5730 years, and so
on, you should be able to convince yourself that the quantity Q of radioactive carbon-14
in the sample after t years is given approximately by
I r/5730
= (O) ( ) = Q(O)e-(ln2)r/5730
Q Q 2
where Q (0) is the initial quantity of radioactive carbon-14 present in the sample.
As a final elementary illustration of an exponential model, consider compounded
interest. If A(O) dollars are invested at 5% interest compounded yearly, the formula for
the value of the investment after t years is
A = A (0)( 1 + 0.05) = A(O)e'
1 In 1.os.
(Why?) If interest is compounded daily (ignoring leap years), the formula becomes
0 05 3651 A e365rln (l+0.05/365>.
A= A(O) (1 + · ) = (O)
365
(Why?) In general, if A(O) dollars is invested at r%
interest compounded k times per
year, the fonnula for the value of the investment after
t years is
A= A(O) (1 + _r_)
lOOk
kr
= A(O) irlnO+r/(IOOkJ)
A = A(O)err/100_ (1)
The population, radioactive decay, and compound interest examples we have just
considered are all modeled by functions of the form
I
(2)
I
where C > 0 and k are constants. In these models, we say we have exponential growth
if k > 0 and exponential decay if k < 0. The derivative of the function in Equation (2)
is y' = kCekr or '1,
I
I y' =ky.1
(3)
Of course, now you know that if you solve the differential equation in (3) you find its
solutions are given by Equation (2). In effect, we have found ourselves on a two-way
street. At the beginning of this section we applied common sense to obtain the expo
f
nential models of the form in Equation (2) from which we get dif erential equations
of the form in Equation (3). We could equally well start with the differential equa
tions in the form of Equation (3) and solve them to obtain the exponential models in
Equation (2).
You may well feel that the approach we used at the beginning of this section to
I arrive at exponentiaJ models is more natural than the differential equations approach and
wonder why we bother with the differential equation approach. One reason for taking
I the differential equation point of view is that there are many instances where we have
I to adjust the rate of change of y in order to model the problem at hand. Consider the
following example.
EXAMPLE 1 Suppose that an investment program pays 6% interest a year compounded continuously.
If an investor initially invests $1000 and then contributes $500 a month at a continuous I
I
rate in this account, how much will have accumulate.d in her account after 30 years?
Also, how much of this will be interest? II
I
I Solution Were it not for the contributions, Equation (I) gives us that this investor would have
A = A( O)e't/100 = lOOOe 0.06r
in her account after t years and the annual rate of growth of her investment would be
governed by the differential equation
0 1
A' = (0.06) 1000e ·06 or A' = 0.06A .
i
__,,�------------------------:1111111------------.1111�
.::::.l!-l!lil...______
146 Chapter 3 First Order Ordinary Differential Equations
In this problem, w e adjust the growth rate by noting that her continous contributions of
$500 per month increase her investment's growth rate by 12 · $500 = $6000 per year.
Hence w e can model the amount A of her investment after t years by the differential
equation
A' = 0.06A + 6000,
so that
The graph of the solution to the initial value problem in this example appears in Figure
3.9. The amount in her account after 30 years is
A(30) � $511,014.39
(rounded to the nearest cent). After 30 years she will have contributed $1000 + 30 ·
$6000 = $181,000, and hence she w ill have received $330,I 04.39 in interest (rounded
to the nearest cent). e
500,000
400,000
200,000
100,000
O 2 4 6 8 IO l 2 14 16 18 20 22 24 26 28 30
Figure 3.9
EXAMPLE 2 Ten pounds of salt is dissolved in a 200-gallon tank containing 100 gallons of water.
A saltwater solution containing I lb salt/gal is then poured into the tank at a rate of
2 gal/min. The tank is continuously well-stirred and drained at a rate of I gal/min. How
much salt is in the tank after a half hour? How much salt is in the tank when the tank
overfills?
--... .
2 gal/min
Figure 3.10
Note that the volume of salt water in the tank is increasing at a rate of l gal/min.
(Why?) Therefore, the volume of salt water in the tank after t min is (100 t t) gal.
(Why?) The amount of salt in the tank is also changing. To solve this problem, we first
determine the differential equation describing the rate of change of the salt.
The rate of pounds of salt entering the tank is given by
gal
I�· 2 = 2�.
gal min min
The amount of salt leaving the tank is determined in a similar manner. Denote the number
of pounds of salt in the tank at time t by S(t). Then
S(t) lb
100 + t gal
is the concentration of salt in the tank and is also the concentration of salt that will leave
the tank. Multiplying this concentration by I gal/min gives us
� � . 1 gal = � _!_I:_
100 + t gal min 100 + t min
for the rate of pounds of salt leaving the tank every minute. Since the rate of change of
salt in the tank is given by the rate of salt entering the tank minus the rate of salt leaving
the tank,
S'(t)=2-�.
IOO+t
This is a Linear differential equation. Its general solution is
C
S(t) = (100 + t) + - .
I 00 t t
148 Chapter 3 First Order Ordinary Differential Equations
140
120
100
80
60
40
20
0 20 40 60 80 100
Figure 3.11
EXAMPLE 3 Newton's law of cooling states that a body of uniform composition when placed in an
environment with a constant temperature will approach the temperature of the environ
ment at a rate that is proportional to the difference in temperature of the body and the
environment. A turkey has been in a refrigerator for several days and has a uniform
temperature of 40 ° F. An oven is preheated to 325° F. The turkey is placed in the oven for
20 minutes and then taken out and its temperature is found to be 60° F. Approximately
how long does the turkey have to stay in the oven to have a temperature of 185° F?
Solutio11 Let e be the temperature of the turkey at time t and assume the turkey has unifo rm
composition. By Newton's law of cooling
e' = k(325 - e).
We can solve this equation either as a separable or linear differential equation. Doin g
either, we find the solutions are
e = ce-k, + 325.
Using 0(0) = 40, we find
C = -285.
3.6 Modeling with Differential Equations 149
The graph of this function appears i n Figure 3.12. To finish the problem, we need to find
tso that
() = -285e- 10(57/53)
20 I+ 325 = (85.
•
::::: 195.397 minutes,
ln(57/53)
or about 3.26 hours.
260
240
220
200
180
160
140
120
100
80
60
40 c..._��..l-��--'--��--'���L---��_J_--
o JOO 200 300 400 500
Figure 3.12
The final two examples of this section involve the motion of an object.
EXAMPLE 4 Newton's Jaw of momentum says that the force acting on a body is equal to its mass times
acceleration. An object falling in the earth's atmosphere is subject to the force of gravity
and the friction of air. Assuming that the force due to friction of air is proportional to
the velocity of the object, determine the speed at which an object dropped from rest will
hit the ground.
Solution Let v be the velocity of the object. Then dv/dt is the acceleration of the object. If we
let M be the mass of the object and g represent the acceleration due to gravity, then
150 Chapter 3 First Order Ordinary Differential Equations
V =
Mg 1
-(1 - e-ii1).
k
•
EXAMPLE 5 Newton's law of gravity states that if the radius of a planet is Rand if xis the distance
of an object of mass M in space from the p lanet, then the weight of the object is given
by
MgR2
W=---
(R + x)2
where g is the acceleration due to the gravity of the planet. Suppose an object of mass
Mis projected upward from the earth's surface with initial velocity v0. Determine the
velocity and the maximum distance from the earth of the object.
Solutio11 The differential equation (ignoring air friction) describing the velocity of the object is
dv MgR2
M- =-w=
dt (R+x)2
w�ere the minus sign indicates that the weight is opposing the upward motion of the
obJect. From the Chain Rule we know that
dv dv dx dv
-=--=u-
dt dx dt dx ·
Using this our differential equation takes on the fonn
V-=
dv gR2
dx (R +x)2•
This is a separable equation whose solutions are given by
v2 gR2
2 = R+x +C.
Since x = 0 when t = O we have
3.6 Modeling with Differential Equations 151
EXERCISES 3.6
I. Afi(·r his company goes public, an entrepreneur re 5. Cesium 137 Jhas a half-life of approximately 30
tires al the age of 32 with a retirement ponfolio worth years. If cesium 137 is entering a cesium-free pond
$2 mi! lion. If his portfolio grows at a rate of 10% per at a rate of I lb/year, approximately how many
year compounded continuously and he withdraws pounds of cesium 137 will be in the pond after 30
$250.000 per year at a continuous rate to live on, years?
how old will he be when he runs out of money? 6. A drum contains 30 g of radioactive cobalt 60 sus
2. A company offers a retirement plan for its employ pended in liquid. Because the drum has a small leak,
ees. If the retirement plan bas a yearly interest rate 0.1 g of radioactive cobalt 60 escapes from the drum
of 10% compounded continuously and an employee each year. If the half-life of radioactive. cobalt 60 is
continuously invests in this plan at a rate of$ I 00 per about S.3 years, about how many grams of radioac
month, how much money will the employee have in tive cobalt 60 are left in the drum after 10 years?
this plan after 30 years? 7. A l 00-gal tank contains 40 gal of an alcohol-water
solution 2 gal of which is alcohol. A solution con
3. An amoeba population in a jar of water starts at
taining 0.S gal alcohol/gal runs into the tank at a rate
IOOO amoeba and doubles in an hour. After this
of 3 gal/min and the well-stirred mixture leaves the
time, amoeba are removed continuously from the tank at a rate of 2 gal/min.
jar at a rate of 100 amoeba per hour. Assuming con
tinuous growth, what will the amoeba population in a) How many gallons of alcohol are in the tank
the jar be after 3 hours? after 10 minutes?
4. Supp ose that a culture initially contains I million b) How many gallons of alcohol are in the tank
when it overflows?
bacteria that increases to 1.5 million in a half hour.
Further suppose after this time that 0.5 million bac 8. A 500-gal tank contains 200 gal fresh water. Salt wa
teria are added to this culture every hour at a contin ter containing 0.5 lb/gal enters the tank at 4 gal/min.
uous rate. Assuming continuous growth, how many The well-stirred solution leaves the tank at the same
bacteria will be in the culture after 5 hours? rate.
152 Chapter 3 First Order Ordinary Differential Equations
(As�ume that normal body temperature is 98.6° F.) (Hint: Recall that the slopes of perpendicular
lines are the negative reciprocals of each other.)
14. A dead body is discovered by an investigator. At
th :, tune of discovery, the body's temperature is b) Solve the differential equation in part (a) to
_
85 E The a1nemperature is 75° F. The body is then determine the orthogonal trajectories of the
put into a refngerated system with a temperature of family of circles x 2 + y2 = a2.
40 F for 2 hours. At the end of these 2 hours the
20. Find the orthogonal trajectories of the family of hy
temperature of the body is 650 F. Assuming the con-
perbolas xy = k where k is a constant.
3.7 Reduction of Order 153
21. A5 ;10pulations increase in size, the growth rate of 23. In thermodynamics, the equation relating pressure
the population tends to slow as the environment in P, volume V. number of moles N, and temperature
h•;>il'> the growth rate. The logistic growth curve is Tis given by PV = N RT where R is Avogadro's
Oil(' mPans of trying to model this. This curve is constant.
gi\�11 by a differential equation of the form a) Show that PdV + VdP = N RdT.
dP kP(C - P) b) Using the adiabatic condition dU = -PdV
dt C and the energy equation d U = N cv dT where
U is the total energy and Cv is the molar heat
when· k is a constant, P is the population, and C is
the maximum population the environment can sus capacity. show the differential equation in part
tain, ca1 led the carrying capacity of the environment. (a) can be expressed as
PdV + VdP = -(R/cv)PdV.
Nl,tice that if P is small relative to C, the differen
ti.i! equation for logistic growth is approximately the c) Solve the differential equation in part (b) to
s:m•c as the one dP/dt = kP modeling exponen obtain the equation for the ideal gas process
tial growth since (C - P)/C is close to one. As P relating P and V.
approaches C, (C- P)/C is close to zero so that 24. Financial analysts use the following differential
the �rowth rate in logistic growth is close to zero. equation to model the interest rate process of a mar
Find the logistic growth curve of a population P if ket economy
thro, �rrying capacity of the environment is I trillion
and the initial population is IO bilJion. d(f(r)) = (8(t) - a(r)) dt + a(t) dz
22. A� i.\O epidemic spreads through a population, the where dz models the randomness of the economy.
rate at which people are infected is often propor a) The Ho-Lee process uses /(r) = r, a(r) = 0,
t1 ,n.J to the product of the number of people infected and 8(t) = R(I) + a(t). Solve the Ho-Lee
anct the number of people who are not infected. equation if a(t) = 0 and R(t) = 0.05t.
a) Jf C denotes the total population, P denotes the b) Another process has f (r) = In r and
nmnber of infected people, and Po Jenotes the a(r) = lnr. Solve this equation assuming
initial number of infected people, write down O(t) = R(t) + a(t), a(t) = 0, and
an initial value problem modeling such an R(t) = 0.05t.
epidemic. (Solutions for a f= 0 are called stochastic solutions
b) Solve the initial value problem in part (a). and are beyond the scope of this text.)
3. 7 REDUCTION OF ORDER
In this section we will consider some higher order differential equations that can be
solved by reducing them to first order differential equations. One case where we can do
this is when w e have a second order differential equation that does not contain y. Such
an equation can be reduced to a first order equation by substituting
I IV =y'
as the first two examples of this section illustrate.
EXAMPLE 1 Solve
This is a linear equation. Solving this equation (you will find that you have to use
integration by parts), we obtain
l I
v = C 1 e -2x +-x- -.
2 4
Since y' = v, integrating v gives us
I
y = --C1 e-2x
2
+ -x
4
I
1 2 - -x + C2.
4
•
We have seen that the general solution to a first order differential equation contains
a constant. Notice that the general solution to the second order differential equation in
Example 1 has two constants. This pattern continues; that is, the general solution to an
nth order differential equation will haven constants.
v = -2 + Cex 12.
2
-2+ C = 1,
C =3,
and hence
v = -2 + 3ex 12.
2
2 2
Let us use Jo' e' /2 dt as an antiderivative of e-' /2. Doing so, we have
•
The approach of Examples 1 and 2 can be extended to third order differential equa
tions that do not have y' and y. In fact, it can be extended to nth order differential
equations not having y<•-2J, y<•-3>, ... , y. (How would we do this?)
The technique we used in Examples I and 2 worked because the equation did not
contain y. There is another technique we can use to reduce the order of second order
equations that have y but not x. We do this by again letting
Instead of using v' for y", however, we use the Chain Rule to write y" as
dv dv dy dv
Y =-=--=v-.
11
dx dy dx dy
The next two examples illustrate how these two substitutions are used.
Solution Substituting
dV
y' = V and yII = v -,
dy
our equation becomes
dv
JV = V2 ,
dy
which is a separable equation. Upon re writing this equation as
dv dy
-=-
V y
and integrating, we find
In lvl = In lyl + C.
Now we need to solve this first order differential equation. To do so, let us first solve
for v:
e•n lvl = eln IYl+C
c
lvl = e lYI
v = ±ec y .
156 Chapter 3 First Order Ordinary Differential Equations
•
we find its solutions have the form
dv.
S0lutio11 0 the substitutions v
Makino =y I
and y
I/
= v - mth1s" equat10n,
• we ob tam
.
�
dv
V - + y =0.
dy
This is a separable differential equation whose solutions are given by
v2 y2
2 =-2+c.
Since y' = v, we have
y
,2
+ y 2 = c:
where Ct = 2C. (The choice of C1 was made in hindsight to make the form of our
answer nicer.) Solving this equation for y' = dy/dx gives us the separable differential
equation
dy = ±(Cf -y2)1;2
dx
or
dy
dx.
(Cf - y2) 1/2 = ±
Integrating we get
sin-1 ( ;J = ±x + C,
so that
y = C1 sin(±x + C).
Using the fact that sin(±8) = ± sin(t9),
•
we express the solutions in the form
Y = C 1 sin(x + C2).
...,,._��-
---------•�•--•T,_..-,._
3.8 The Theory of' First Order Differential Equations 157
EXERUSES 3.7
In Exe',·, •ses 1-8, solve the differential equations. Determine the height y of the object at a time t if it
1. y'' - 9y' = x 2. y" + y' = 1 is dropped from a height yo.
3. y'' - · (y ')3 = 0 4. y" +xy' /(I +x) = 0 17. The initial value problem in Example 5 of Section
5. yy 1 - 4(y')2 = 0 6. )' 11 - eYy' = 0 3.6 could also be written as
7. )' I +- 4 )' = 0 8. )'11 - 25)' = 0 d2x MgR2
=- v(O) = v0 , x(O) = x0
In E.r.i:.rcises 9-15, solve the initial value problems. M dt2 ( + x)2;
R
9. y" y ' = l; y(O) = 0, y'(O) = .I where x0 is the initial distance of the object from the
10. y" = x(y') 2; y(O) = 0, y'(O) = I planet. Solve this initial value problem. You may
leave your answer i n terms of an integral.
11. x2 y" + 2xy' = I; y(l) = 0, y'(I) = I
12. y" - y = O ; y(O) = 1, y'(O) = l 18. The motion of a simple pendulum of length I is de
13. y" - 3y 2 = O; y(O) = 2, y'(O) = 4 scribed by the initial value problem
14. y 3y" = 9; y(O) = 1, y'(O) = 4 d2e
15. x./" ,= y"; y(l) = 1, y'(I) = 2, y"(l) = 3 I 2 + g sin (:I = 0; 8'(0) = 0, 0(0) = 80
dt
16. Ii y is the height of the object in Example 4 ofSec- where (J is the angle the pendulum makes with the
6on 3.6, the differential equation could also be vertical, 80 is the initial angle the pendulum m akes
writ!en as with the vertical, and g is the acceleration of grav
d2 y dy ity. Solve this initial value problem. You may leave
-=Mg-k-.
2 your answer in terms of an integral.
dt dt
THEOREM 3.3 Let a, b > 0 and suppose f and afjay are continuous on the rectangle R given by
j.x - xol < a and jy - Yol < b. Then there exists an h > 0 so that the initial value
problem
i
o
x
G(xo) = F(t) dt = 0
XO
and hence G(.x) is a solution to the initial value problem y' = F (x), y (xo) = 0. Suppose
y(x) is a solution to the initial value problem
If we let
F(x) = f(x,y(x)),
We will call Equation (I) an integral equation. A solution to it is a solution to the initial
value problem and vice versa. In particular, if the integral equation has a unique solution,
then so does the initial value problem.
We now prove the uniqueness of the solution to the initial value problem by proving
the uniqueness of the solution to the integral equation. Suppose that u(x) and v(x) are
solutions to the integral equation
"'
y(x) = Yo +1 XO
f(t,y(t)) dt.
That is,
u(x) = Yo+ r
fxo
f(t, u(t)) dt
and
11:
Thus
Using the Mean Value Theorem, for each t we can express f(t, u(t)) - f(t, v(t)) as
f (t, u(t)) - f (t, v(t)) = fy (t, y*(t))(u(t) - v(t))
for some y*(t) between v(t) and u(t). Substituting this, the inequality in Equation (2)
becomes
lu(x) - v(x)I :5 r
lxo
Mlu(t) - v(t)ldt :s M !..-
xo
lu(t) - v(t)I dt
on this rectangle.
We now let
We have that
w(x)::: 0,
r
and
L'
Setting
or
W'(x) - MW(x) � 0.
This looks like a first order linear differential equation only with an inequiility instead
of an equality. Let us multiply by the integrating factor
ef -M dx = e-M
x
to obtain
e-Mx(W'(x) - MW(x)) = (W(x)e -Mx)' ::S 0.
Replacing x by t, integrating each side of this inequality from x0 to x, and u�ing the fact
that
[ XO
W(xo) = w(t) dt = 0
.ro
gives us
W(x) = t
fxo
w(t) dt = {' lu(t) - v(t)I dt 2: 0
fro
W(x) = 0 ·
and
w(x) = W'(x) =O
for Ix - xol <a.Since
u(x) = v(x)
for Jx - Xol < a, completing our proof of the uniqueness part
We now turn oura uent1on
· to the existence part. Notice that if
then
Consequently, the initial value problem determines the first degree Taylor polynomial
dx dx
= fx (X,y) t /y (X, y)y'.
This allows us to find y"(xo) as
which is equivalent to the initial value problem. The idea is to create a sequence of
functions that converge to the solution of the integral equation. (This is like creating the
sequence of Taylor polynomials that converge to the Taylor series.)
To start the sequence of functions, we define the 0th Picard iterate to be the constant
function
I Po(x) = Yo· I
By this definition po agrees with the solution to the integral equation at x0:
ro(xo) = y(xo) =yo.
XO
r·
Thus, for example,
J(t,yo)dt
x
and
Pk(xo) = Yo
for each k = I, 2, 3 ....
To give you a feel for this iterative process, let us apply it to some initial value
problems.
EXAMPLE I Determine the Picard iterates for the following initial value problem.
y' = y, y(O) = I
f(x,y)=y
and that the equivalent integral equation to this problem is
Po(x) = y(O)= 1.
The first Picard iterate is
x
Pt (x) =Yo+ 1� f(t, Po(t))dt = l + �(" f(t, 1)dt = 1 + (" 1dt = 1 + x.
�
The second Picard iterate is
x
P2(x) =Yo+ 1 f(t, P1
xo
(t)) dt =I+ f" f(t, l + t) dt = t + f" (1 + t)dt
lo lo
1
= 1+ X +-x2
2
3.8 The Theory of First Order Differential Equations 163
Ii
The third Picard iterate is
=l+
1'( l ) I I
l+t+-t2 dt=l+x+-x 2 +-x3 .
2 2 3!
o
From the pattern developing here, we see that the kth Picard iterate is
lz 13 lk = "ln
Pk(X) = I+ X+ -X + -x + · · · + -X L., -x .
k
!
2 3! k! n=O n!
Notice that Pk (x) is the kth Maclaurin p olynomial fore' and that ex is the solution to
this initial value problem. The Picard iterates converge to the solution to the initial value
problem in this example. •
111,
EXAMPLE 2 Determine the Picard iterates for the following initial value problem.
Solution We have
{I '
J (x, y) = 1 + y2
L'
and the equivalent integral equation is
,,
The 0th Picard iterate is
I
Po(x) = y(O) = 0.
I
I 2
= x + 3x 3 + 15x5
1
+ 63x7 . [· 1
: fa
164 Chapter 3 First Order Ordinary Differential Equations
= r (1
}0
(1
+ +�t + l:_t
3
3
15
5
+ _2__, ) ) dt
63
1
2
I 3 2 5 17 7 38 9 134 11 4 _l_x 15
= x+ 3 x +15 x · + 315 x + 2835 x + 51975 x + 1 285 x +59535
13
2
The pattern developing here is not as apparent as it was in Example 1. However, it can
be verified that the terms of degree less than or equal to k of /Jk (x ) fonn the kth degree
Maclamin polynomial for tan x and that tan xis the solution to the initial value prohlem.
Here again the Picard iterates converge to the solution of the initial value prohlem. e
EXAMPLE 3 Determine the Picard iterates for the following initial value problem.
'
y = x +y , y(-1) = 0
2 2
S0l11tio11 We have
f(x,y)=x +y2
2
Po(x) =y(-1) = 0.
The first Picard iterate is
x x
Pi(x) =0+ f f(t,0)dt =l t2 dt = � + ! x3 •
-I -I 3 3
The second Picard iterate is
x I I
r2(x)=O+! f (r.-+-t3 ) dr
-I 3 3
_ x 2 (I I )2 17 J 1 J 1
- f (t + -+- t3 ) dt=- +-x+-J x 3+ -x4+ -x .
-I 3 3 42 9 3 18 63
Even if we continue, no pattern for Pk (x) is discernible as
i n Examples I and 2. However,
as we shall argue next, Picard •·terates sueh as these
. .. do converge to the solution to the
initial val�e problem. Exercises 11-13 illustrate how
to modify an initial value problem
su�� as t hi s so that the Picard iterates generate the
_ _. Taylor polynomials about x0 of the
ongmal m1tial value problem.
•
The questions that have to be add ressed . . . that
regarding the Picard iterates to obtam
they converge to the soluti. on of the initial value
problem in Theorem 3.1 are:
3.8 The Theory of First Order Differential Equations 165
I
I 1. Are the Picard iterates defined for each k?
2. Do the Picard iterates converge?
3. If the Picard iterates converge, do they converge to the solution of the initial
value problem?
In answer to question I, for the Picard iterates to be defined the integrals defining
them must exist. This is why Theorem 3.1 requires that f and aflay be continuous. In
Examples 1-3 f is not only continuous, but infinitely differentiable. However, continuity
off and fv is all we need to guarantee the existence of the integral. The number h in
Theorem 3.1 is needed to guarantee that (x, pk(x)) always lies in the rectangle R where
f and fy are continuous. Exercise 15 addresses this issue. Assuming that the Picard
I
iterates are defined for each k, we move on to question 2.
Note that I,
pi(x) = Po(x) + (p1(x) - Po(x)),
converges. In fact, it will be shown that this series converges absolutely for !xi � h for
an h > 0.
I
Consider the partial sum
t
I
ip0(x)I +
11=0
IPn +t (x) - P11(x)I = IYol + lYo + 1: f (t, Po(t)) dt - Yol + · · ·
11
,_
166 Chapter 3 First Order Ordinary Differential Equations
= I.Yol + 11: f(t, po(t)) dt' + 't. lL\J(t, Pn(t)) -f(t, p,,_,(r)) dtl
Using a Mean Value Theorem approach similar to the one we used in the proof of the
uniqueness part, we can obtain (the details of this are left as Exercise 16)
L IPn+I (x)-pn(x)ISIYol+Klx-xol+ L M
k k
IPo(x)I+ \'
f.....J IPn+I (x) -P11(x)ISI.Yol + Klx - xol+� M
k
K
11=0 xo (n .
M2 K 11-I
n=I
lx-xol"
k
SIYol+Klxl+L
n=l
n.I
for Ix -xol Sh. We leave it as Exercise 18 for you to show that the series
2K
IYol + Klx -xol+ L M n!n-1 Ix -xol"
oo
n=l
converges. Since
'°"'
IPo(x)l+i.....JIP11+1(x)-p11(x)ISIYol+Klx
-x0 1+L
oo
M 2 K 11- 1
jx-xol",
11=0 n.1
the sequence of partial sums
n=l
x
Y =Yo+ i f(t, p(t))dt. (5)
xo
Finally, because of the continuity of f, we can move the limit inside f and get
x
p(x) = Yo+ i f(t, lim Pk(t))dt
k-+00
XO
x
= Yo +l f (t, p(t))dt
XO
EXERCISES 3.8
In Exercises 1-2, give the equivalent integral equation In Exercises 7-8, find the first two Picard iterates for the
to the initial value problem. initial value problem.
1. y' =x + 2y, y (O) = 0 7. y' = x 2 - y 2 , y(l) = 0
2. y' = eY + sinx, y(l) =-1
8. y' = X+ y 2, y(- J) = l
In Exercises 3-6, solve the initial value problem, find
the first four Picard iterates, and compare these iterates In Exercises 9-10, generate the Picard iterates until you
to the first four Maclaurin polynomials of the solution can no longer compute the integrals in a closed form.
to the initial value problem. These exercises illustrate that Picard iterates must some
3. y' = 2y, y(O) = J 4. y' = -y, y(O)=2 times be expressed in integral form.
5. y' = 2y 2 , y(O) = I 6. y' =xy, y(O) = -1 9. y' =cosy, y(O) =0 10. y' = eY, y(O) = 0
7 If you take an advanced calculus course, you will learn that interchanging an integral and limit as we are
doing here requires a stronger type of convergence called uniform convergence. It is possible to show that the
convergence of the sequence of functions pk(x) is uniform. but the details of this are left for more advanced
courses.
168 Chapter 3 First Order Ordinary Differential Equations
11. Show that if u(x) is a solution to the initial value show that if h is the minimum of b/ Kand a, then
problem 11 1 = f(x + .t0, 11), 11(0) = ? 'o, then the Picard iterates are defined for Ix -xol ::: h.
.
v(x) = u(x _ xo) is a solution to the 1111 1ml value 16. a) Show that if Ix - xol .:'.S h, then
problem .v';;; f(x, y), y(xo) = Yo- lpi(x)-po(x)I:;: Klx-xol � Kh.
b) Using the Mean Value Theorem and part (a),
In Exercises I 2-13. use the approach of Exercise 11 to show that
generate the first three Picard iterates for u and then use
_
\'(X) = u( x-x0 ) to obtain the first three Picard iterates
L IP +l (x) -p,,{x)I
k
for y. Also verify that the first three terms of the third lpo(x)I + n
mial about x0 of the solution to the initial value problem :s IYol + Klx -xol
at xo.
12. y' = x 2 + y2 , y(-1) = 0 + tM L'
n=I
xo
IP11U) - Pn--1(t)I dt.
13. y' = x -y , y(I) = I
2
14. Consider the �cquence of functions Pn (X) = 17. Using induction on the Picard iterates show that
nxe-ni show that
1
MK"-1
Jim t
P (I)dt ,f.
n-+00 lo n
t lim Pn (t)dt.
lo ,,-oo
lp.(1)-Pn-1(t)I:::
(n - l )!
ltl"-I.
M2 x,-' x x I"
18. Show that IYol+Klx-xol + I:11oo=1 -, -I - o
,!-
Thi� illustrates that limits and integrals cannot be converges for all x.
intcn.:hanged in general.
19. We chose Po(x) = y0 for the Picard iteration. Show
15. Thi� exercise outlines the proof of the fact that the that for any choice o f p0 (x) that is continuous for
Picard iterates arc defined. Show that if (x, p.(x)) Ix -xol < a with IPo(x)I < b for Ix-.to! < a the
lies in the rectangle Ix -Xol < a and IY -Yol < h, Picard iterates converge to the unique solution of the
then lp�(x)I = 1/(x, P11 -1(x))I::: K. Use this to initial value problem.
If x is close to x0, then the tangent line is close to the solution y = </>(x) for x near xo.
Therefore, if x1 is close to xo, then
is given by
at the x-values
1.1, 1.2, 1.4, 1.5.
Solution Substituting x 1 = t. l , x2 = J .2, x3 = 1.4, and x4 = 1.5 into Equation (1), we have
YI =0+ f(I, 0)(1.1 -1) = 0.1
Y2 = 0.1 + f(l . I, 0.1)(1.2 -1.1) = 0.22
•
y3 = 0.22 + f (1.2, 0.22)(1.4 - 1.2) = 0.504
y4 = 0.504 + f(1.4, 0.504)(1.5 - 1.4) = 0.6944.
Since the differential equation in Example 1 is linear, we have the exact solution
y = t/>(x) = 2ex -l -x- I
to the initial value problem. The values of 4> at x1 = I. I, x2 l.2, x3 = 1.4, and
x4 = 1.5 are
Figure3.13
Recall that Taylor's theorem tells us that if a function pis n + I times continuously
differentiable on an interval containing a, then
p"(a) p (a)
"'
p(x) = p(a) + p'(a)(x -a)+ --(x - a)2 + --(x - a)3
2 3!
p(n)(a) p<n+l>(c)
+ · .. + --(x - a)"+ (x - a)" +1
n! (n + I)!
for some number c between x and a. The polynomial
' p"(a) 2 p (a) p<n> (a)
111
T,,(x) = p(a)+p (a)(x-a)+- -(x-a) + (x-a) +-··+ x-a)"
�(
3
2 �
(2)
is called the nth degree Taylor polynomial for pat a. This polynomial, T,, (x), and its
first n derivatives at a agree with p(x) and its first II derivatives at a. That is,
is called the Lagrange form of the remainder and is the error in approximating p(x) by
Tn(x). Determining techniques for minimizing this error is an important topic in the
study of numerical solutions to initial value problems.
As in Euler's method, suppose we wish to obtain approximations at the x-values
I xo,x1,X2,--·, xN
J 72 Chapter 3 First Order Ordinary Differential Equations
to the solution v = <jJ(x). In the Taylor method approach we begin by substituting y for
p in Equation (2), which gives us
y"(a) y111(a) 3
T,,(x) = y(a) + y'(a)(x -a)+ --(x -a) + (x -a)
2
2 31 (3)
<">
+ .. · + -y (a)-
"
(x - a) .
n!
We will use T,,(x1) as our approximation to ¢(xi). Substituting xi for a and x2 for x in
Equation (3), we obtain
<nl
y (x1)
+ .. · + ---(x2 -xi)"
n!
+· · · + �y<k\xk)(xk+i -xk)"
n.
Solution For the second order Taylor approximation we need the second
derivative of the solution.
3.9 Numerical Solutions of Ordinary Differential Equations 173
This is
d d
y = - y = -(y + x) = y + I.
II I I
dX dx
Since
we have that
, J 11
T2 (x1) = y(I)+y (1)(x, -I)+ y (l)(x, -I)2
2
I
=y(l)+y'(l)(l.1-J)+ y"(I)(J.l -1)2
2
I
=0+0.1 + 2(2)(0.1)2 =0.11.
Wenowdevelopthe secondorderTaylorpolynomialat(x1, T2(x1)) = (1.1, 0.11). Since
we have that
I
� 0.11 + J.21(0.1)+ -(2.21)(0.1) 2 � 0.24205.
2
(We have used the approximation symbol since each of the values we substituted for
y'(l.1), and y" (l.l) are approximations.) Continuing, we have
y(l.l),
so that
I 11
= y(l.2)+y (1.2)(1.4 - l.2) + _v (I.2)(1.4 - 1.l)
1 2
2
I
� 0.24205 + 1.44205(0.2)+ (2.44205)(0.2)2 � 0.579301
2
and
174 Chapter 3 First Order Ordinary Differential Equations
so that
•
2
� 0.7921 276.
Comparing with Exampl e I, notice that the second order Taylor polynomial gives a
b etter approximation of </J(x) at e ach of the x -values than Euler's method.
In Example 3 we use the fourth order Taylor method.
Ifd d 2
I
Y =d y = (x +xy)=2x+y+xy',
x dx
x dx
d d
yf4> = -y"'= -(2 + 2y' +xy")= 3y" + xy
111
dx dx
Using the formula for the nth order Taylor polynomial with n = 4 gives us
- y,, + (x,2, +x,,y,,)(xn+l - x,,) + (2 x,, + y,. + x y' (x,,, y,,))(Xn+l - x ,,)2
Y11+1 _ 1
2 ,,
Yi � 1.00 5345833
Y2 � 1.022887920
•
Y3 � 1.055189331
Y4 � 1.053149.
3.9 Numerical Solutions of Ordinary Differential Equations 175
In practice, one usually chooses x1, x2, •.• Xn so that they are equally spaced. That
is,
x1 =xo +h
X2 = XI + h = XQ + 2h
XN+l = XN + h = Xo + Nh,
for some fixed number h. For example. h = 0.1 in Example 3. In Exercise 9 we ask
you to obtain formulas for the Euler and Taylor methods in this special case.
is a weighted average of values off (x, y). We present the fourth order Runge-Kutta
method only for equally spaced x-values.
I xo, xo + h, xo + 2h ...xo + N h
We set
bn = f (Xn + i, Yn + ian)
Cn = f (xn + i, Yn + ibn)
dn = f (Xn + h, Yn +hen ),
176 Chapter 3 First Order Ordinary Differential Equations
18.
The term
I
-(a,,+ 2b,, + 2c. + d,,)
6
may be interpreted as a weighted average slope.
In the exercises we ask you to compare the numerical solutions of the Euler method,
fourth order Taylor method, and fourth order Runge-Kutta method for some initial value
problems.
EXERCISES 3.9
In Exercbes 1-4, calculate the Euler method approxi In Exercises 14-19, use a software package such as
mations to the solution of the initial value problem at Maple to help with the calculations.
the given x-values. Compare your results to the exact 14. Consider the initial value problem y' = x tan y,
solution at these x-values. y(O) = I. Use Euler's method with the following
I. y' = 2y - x; y(O) = I, x = 0.1, 0.2, 0.4, 0.5 values for (h, N): (0. I, I 0), (0.05. 20), (0.025, 40),
2. y' = xy + 2: y(O) = 0. x = 0.1, 0.2, 0.4, 0.5 (0.0 I 25, 80). Solve this initial value problem and
compare these approximating values with those of
3. y' = x 2 y2 ; y(I) = I, x = 1.2, 1.4, 1.6. 1.8 the actual solution. What conclusions can you make
4. y' = )' + y2 ; y(I) = -!, x = 1.2, 1.4, 1.6, 1.8 regarding the accuracy o f the Euler metlicd for this
initial value problem? What happens if you increase
In Exercises 5-8, use the Taylor method of the stated N?
order II to obtain numerical solutions to the indicated
15. Consider the initial value problem y' = x y,
2
excrci�c of thi� �ection. Compare the results from this
y(O) = 1. Use Euler's method with the following
method with the method of this exercise.
values for(h, N}: (0.1, I 0), (0.05, 20), (0.025, 40},
5. Excrci�c I. 11 =2 6. Exercise 2, n = 2 (0.0125, 80). Solve this initial value problem and
7. Exercise 3, 11 = 3 8. Exercise 4, 11 = 3 compare these approximating values with those of
9. Derive formulas in the Euler and Taylor metho<ls for the actual solution. What conclusions can you
Xn+I = Xn +Ir== Xo + nh, II = 0, I. 2, .. ,, N. make regarding the accuracy of the Euler method for
this initial value problem? What happens if you in
In Exercis es 10-13, use the fourth order Runge-Kutta crease N?
method with h == 0.1 and N = 5 to obtain numerical
_
16. Compare the fourth order Taylor and fourth order
i.Olutiom,. In Exercise 10, compare with the results of Runge-Kutta methods for y' = x 2 y + 4x, y(O) == 0
Exercii,es 2 and 6. and in Exercise I I, compare with the with h = 0.1 and N = 40. Solve this initial value
re,ultl. of Exercises I and 5. problem and compare these approximating values
JO. y' = xy + 2, y(O) = O with those of the actual solution. Which approxi
II. y' = 2y-x,y(O) = I mations are more accurate?
I 2. y' = x2 - y2 , y(O) = O 17. Compare the fourth order Taylor and fourth order
13. y' = x2 + y2 , y(O) = O R�nge-Kutta methods for y' = xy - 6, y(O) == 0
with h = 0.1 and N = 40. Solve this initial value
II
3.9 Numerical Solutions of Ordinary Differential Equations 177
probkm and compare these approximating values 19. Do Exercise 18 for the initial value problem y' =
with those of the actual solution. Which approxi cosy + sin x, y(O) = 0. whicb also does not have a 1,
mation'- are more accurate? closed form solution.
!I
18. A closed form solution cannot be obtained for the In Exercises 20-23, use the dsolve command in Maple
,,
initU value problem y' = x 2 + y 2 , y(O) = O. with the numeric option to determine a numerical solu
Use the fourth order Taylor and fourth order Runge tion at x = 0.5 with Maple's default method. Then use
Kun .. methods on the interval O < x < I first with
h = 0.2 and then with h = 0.1 to obtain numeri
Maple's odeplot command to graph Maple's numerical
solution over the interval [O. I]. (Or use another appro I!
(I III
cal �,,tutions to this initial value problem. Then use priate software package.) Finally, use Maple to graph
Mapie (or an appr opriate software package) to plot the phase portrait of the initial value problem over this II
the points obtained with these numerical solutions interval and compare this graph with the graph of the 1,
and to graph the phase portrait of this initial value numerical solution.
proi•l�m. Which method appears to give a moreac 20. y' = x2 + y2,y(O) = 0
cuwL a:1swer?
21. y' = cosy+ sin x, y(O) = 0
22. y' = sin y + e x,
- y( I) = 0
II
23. y' = x4 + y2 , y(l) = -I
1
I ::1
Ii
I;
11
I t
�a����--��---------------�·---------1·
................................ .
-..
Linear Differential Equations
In the previous chapter we focused most of our attention on techniques for solving first
order differential equations. Among the types of equations considered were the first
order linear differential equations
qi (x)y' + qo(x)y = g(x).
In this chapter we are going to study higher order linear differential equations. As we
do so, you will see how many of the vector space concepts from Chapter 2 come into
play in this study. Higher order linear differential equations arise in many applications.
Once you have learned techniques for solving these differential equations, we will tum
our attention to some of the applications involving them, such as mass-spring systems
and electrical circuits in the last section of this chapter. To get staned, we first discuss
the theory of these equations.
Throughout this chapter, we will assume that the functions q q -1, . . . , qo and g are
11, 11
continuous on an interval (a, b). We shall further assume that q (x) is not zero for any x
11
in (a, b). (Recall that we included such assumptions for the first order linear differential
equations q1 (x)y' + q0(x)y = g(x) in Chapter 3.) Note that if y = f (x) is a solution
to the differential equation (I) on (a, b), then f cenainly must have an nth derivative on
179
180 Chapter 4 Linear Differential Equations
(a, b ) and consequently is a function in the vector space of functions with nth derivatives
on (a, b), D"(a, b). Actually, we can say a bit more. Since
q (x)f(x) ,
g(x) _ qn-1(x)f -l)(x) _ ... _ o
(n
t<nl(x) =
q,i(x) q,,(x) q n (X)
showing that J< > is continuous on (a, b). Hence the solutions of an nth order linear
11
differential equation on (a, b) will be elements of the vector space of runctions with
continuous nth derivatives on (a, b), C(a, b).
If g(x) = 0 for all x in (a,b) in Equation (1), then we. call the 11th order linear
differential equation homogeneous; otherwise we call it nonhomogeneous. If g is
nonzero, we refer to
Ill + 2 II
Y Y X Y - 2xy/ + y = Q (5)
x y" - 2x(y')2 + y = x
2
(6)
II /
Y + x cosy -2xy + y = sin x (1)
(4) 2
form
Theorem 3.1 of the last chapter told us that first order initial value problems have
unique so lutions (provided the function f in the differential equation y' = f(x, y) is
sufficiently well-behaved). Similar theorems ho ld for higher order differential equations.
The next theorem, whose proof wi ll be omitted, is the version of this for nth order linear
differential equations.
THEOREM 4.1 Suppose q,,, q,, _ 1, •.. , q 1, q0, and g are continuous on an interval (a, b) containing xo
and suppose %(X) f= 0 for all x in (a, b). Then the initial va lue problem
q" (x)y{") + qn -1 (x)y(n-l) +···+qi (x) y' + qo(x)y = g(x);
y(xo) = ko , y' ( xo) = k1 , .,., y(n -l)(Xo) = kn I
-
where ko, k1 , ••• , k11 1 are constants has one and only one so lution on the interval (a, b).
_
Using Theorem 4.1, we can prove the following crucial theorem about homogeneous
linear differential equations.
THEOREM 4.2 The solutions to an nth order homogeneous linear differential equation
n
q,, (x )/ > + qn -1 (x)y n -l + · · · + qo(x)y = 0
on an interval (a, b) whereq,,, q,,_ 1, ... , qo are continuous on (a, b) and q,, (x) is nonzero
/
Proof We have already noted that the solutions are functions in C"(a, b). To show that they
form a vector space, we shall show that they are a subspace of the vector space C"(a, b).
Suppose that /1 (x) and h( x) are solutions of the homogeneous linear differential equa-
tion. Since
q,,(x)(/1 (x) + Ji(x)) ( n) +qn 1 (x)(/1 (x) + !z(x) )<n-I) + · · · + qo(x )(/1 (x ) + fi(x))
-
= q (x) J/"' cx) + q,,_, (x) fi"- 1' c x) + · · · + qo(x)fi (x)
11
-
+ q (x )f?'\x) + q,. _1 ( x)J;" '\x) + · · · +qo (x)fi(x)
11
= o+o = o,
/1(x) + fi(x) is a so lution and the solutions are then c losed under addition. For any
scalar c,
q,.(x)(c/1 (x))<11 > + qn -1 (x)(cf, (x))<n-l) + · · · + qo(x)(c/1 ( x))
1
= c(q,,(x)fi"l (x) + qn -1 (x)f?'- \x) + · · · + qo(x)f1 (x)) = c · 0 = 0
and hence c/1 (x) is a solution giving us the solutions are closed under scalar multipli
cation. Thus the so lutions Jo form a subspace of C"(a, b).
182 Chapter 4 Linear Differential Equations
To complete this proof, we show that the solutions have a basis of n fu� ctions.
. .
Choose a value xo in (a, b). By Theorem 4. 1, there is a solut10n Y1 (x) satisfying the
initial value problem
q.(x)y<n ) +q n -1( x)y n -l + ... + qo(x)y = O;
n I)
y1 (xo) = I, y((xo) = 0, y;'(xo) = 0, Yi - (xo) = 0,
a solution y2 (x) satisfying the initial value problem
q.(x)/"> +q n -1 (x)y n -l + · · · + qo(x)y = 0;
(n-1>( )
y2(xo) = 0, y�(xo) = I, y{(xo) = 0, Y2 xo = O,
0 0
To see thaty1(x), Y2(x), ... , Yn(x) span the vector space of solutions, suppose that f(x)
is a solution of the homogeneous linear differential equation. Let us denote the values
of f(X), J'(x), ·, •, f(n-l)(x) at xo by CJ, C2, ..., C11:
f(xo) = c1, f'(xo) = c2,
Now let
Then
Xo) = C1Y1
f(n-1)
(xo) + c2y/-
(n-1)
( xo) + ... + Cn y�11 -l\xo) = c•.
( ( I)
The last theorem indicates that we should look for a set of solutions to the homoge
neous equation that forms a basis for the vector space of all solutions to the homogeneous
equation. A set of n linearly independent functions y 1, ••• , y,. each of which is a solu
tion of an nth order homogeneous linear differential equation is called a fundamental
set of solutions for the homogeneous equation. Recalling part ( 1) of Theorem 2.12,
which tells us that n linearly independent vectors in an 11-dimensional vector space fonn
a basis for the vector space, it follows that a fundamental set of solutions is a basis for the
solutions to the homogeneous equation. If y 1, ••• , Yn is a fundamental set of solutions
to an nth order homogeneous linear differential equation, then the general solution to
the homogeneous equation is
I YH = C1Y1 + · · · + CnYn· I
The next theorem shows how we use the general solution to the corresponding
homogeneous equation to obtain the general solution to a nonhomogeneous equation.
THEOREM 4.3 Suppose that y 1, ••• , Yn forms a fundamental set of solutions to the homogeneous linear
differential equation
1
qn (x)y < ) +q11-1(x)/"- > + · · · + qi(x)y' + qo(x)y = 0
n
Proof We first show that any function given in the form ciy, + · ·. · + CnY11 + YP is a solution to
the nonhornogeneous equation. Substituting this into the left-hand side of the differential
equation, we have
q.(x)(c1y1 + · · · + Cn )'11 +yp)(n) +q,,_, (,,)(C1)'1 + · · · + CnY11 + yp) (n-l)
+ · · • + qo(x)(C1Y1 + '· · + c,.y,, + YP)
= q,,(x)(C1Y1 + · · · +cn y,. )<n) + q,. _1(x)(C1Y1 +· · · + C11Yn ) Cn-l)
+ · · · + qo(x)(c1Y1 + · · · + c.y,. ) + q,, (x)yt > +qn-1Yt-l) + · · · + qo(x)yp
= 0 + g(x) = g(x).
an<l hence c, y 1 + ... + c,,y + yp is indeed a solution to the nonhomogeneous solution.
11
To see that the general solution is c1 y1 + · · · + c,,y,, + yp, we must show every
solution of the nonhomogeneous equation has this fonn for some constants c,, ... , Cn-
184 Chapter 4 Linear Differential Equations
equation. Since
Suppose y is a solution of the non homogeneous
q,, (x)(y - YP )(nl + q,._1 (x)(y - YP / -1) + · · ·
n
+ qo(x)(y - YP)
) (11)
= q.(x)/•l + qn -l (x)y -1) + · · · +qo(x)y - (q,, (x Y p + q,,
<n
+ · · · + qo(x)yp)
= g(x) - g(x) = 0,
y - yp is a solution to the homogeneous equation. Thus there are scahr� c1,... , c. so
that
•
Y = C1 YI + · · · + Cn Y11 + YP
as needed.
We call the function yp of Theorem 4.3 a particular solution to the nonhomoge
neous dif erential equation. This theorem tells us that we may find the general solution
f
y" - y =x.
(a) Determine the largest interval for which a unique solution t o the initial value
problem
Solution
y" - y = 0.
Substituting in YI = e" and y2 = e-x shows they are solutions of this homo
geneous equation. Since the Wronskian of YI and y2 is
(c) Since
in order for this YH to solve the homogeneous initial value problem c1 and c2
must satisfy
y(O) = cI + ci = I
1
)1 (Q) = CJ - C2 = 0.
Solving this system gives us CI = 1/2, c2 = 1/2. Therefore, the unique
solution to the homogeneous initial value problem is given by
1 X I X = -(e
I X + e-X)
·YH = -e + -e-2 2
2
(d) Substituting yp = -x into the nonhomogeneous equation
y" -y =X,
we see it satisfies this equation. The general solution to the nonhomogeneous
equation is
(e) Since
x x
y = c1e + c2e-x - x and y' = CI e - c2e-x - 1,
CI and c2 must satisfy
y(O) = CJ + C2 = 1
y'(O) = CJ - C2 - I = 0.
Solving this s ystem gives us c 1 = 1, c2 = 0. Therefore, the unique solution to
•
the nonhomogeneous initial value problem is
y=e.. -x.
186 Chapter 4 Linear Diferential Equations
f
EXAMPLE 2 Show I, cos 2x, sin 2x form a fundamental set of solutions for
y
111
+ 4y' = 0.
Also , give the general solution to this homogeneous equation.
S0/utio11 Substituting each of I, cos 2x, sin 2x into this homogeneous equation, we find each is a
solution of the differential equation. Their Wronskian is
1 cos2x sin2x
w(I, cos 2x, sin 2x) = 0 -2sin2x 2cos2x
0 -4cos2x -4sin 2x
= 8.
Since this is not zero for any x, we have that I, cos 2x, sin 2x is a fundamental set of
solutions and
Y
m
+ 4y' = g(x).
For this g(x), solve the initial value problem for the initial conditions y(l) = 0, y'(l) =
-I, y"(l) = 0.
Y
111
+ 4y' = 24x + 16x3 •
Therefore,
Y = CJ + c2 cos2x + c3 sin 2 x + x4 .
Also
We conclude this section with some important results regarding the Wronskian of
solutions to nth order homogeneous linear differential equations. In Section 2.5 we saw
that the W ronskian of a set of functions could be zero and the set of functions is still
linearly independent. The following theorem shows this cannot be true for solutions to
an nth order homogeneous linear differential equation.
Proof We prove this for n =2 and leave the proof for general n as an exercise (Exercise 19).
I YI'(xo)
If
I
Y2(xo)
w(yt(Xo),y2(xo))= ' =0,
Yr ( Xo) Y2( xo)
COROLLARY 4.5 Let Yi, ... , y,, be a fundamental set of solutions to the differential equation
q,. (x)/"> + q,,_ 1 (x)y < -l) +.··+qi (x)y ' + qo(x)y = 0
n
on an interval (a, h). Then w(y1, ••• , y,,) i s never zero on (a, b).
Proof If w(y 1(xo),... , y,,(xo)) = O for some x0 in (a, h), then y 1, ••• , y11 are linearly depe
dent by Theorem 4.4. ;
EXERCISES 4.1
In Exercises 1-4, determine if the differential equation 12. {x,x 2, l/x),x 3 y"' + x 2y" - 2xy' + 2y = O;
is linear or not. If it is linear, give the order of the dif y(-1) = 0, y'(-1) = 0, y"(-1) = 0
ferential equation.
l. exy"' + (2 sinx)y" - 4x 2y' - 5y = 3 In Exercises 13-16, show yp is a solutiun of the nonho
2. eY y" -4 y' + 5x y = 0 mogeneous differential equation, give the general solu
tion to the nonhomogeneous equation by using the �act
3. 5x sinx = 3 yl4l -4x y"' + �y" - 2y' + x 3 y that Exercises 9 -12 contain the respective correspondmg
4. y" = 2y' sinx - 3x y homogeneous equation, and then solve the initial value
In Exercises 5-8,give the largest interval for which The problems for the given initial conditions.
orem 4.1 guarantees a unique solution to the initial value 13. YP = -2,y" - y' - 2y = 4; y(O) == 1, y'(O) == 0
problem.
14. YP = -x/4 + (1/2)x lnx, x2y" + xy' - Y == x ;
5. y" -2xy' +4y = 5x; y(O) = I, y'(O) = 0 y(I) =0,y'(I) = -1
6. x(x 2 -t)y"-2xy'+4y = O; y(2) = 0,y'(2) = -1
15. YP = e-x, y'" - 7y' + 6y = 12e--'; y(O) == I,
7. e' l nxy'" -xy" + 2y' - 5xy = 2; y(l) = 0, y'(O) == 0, y"(O) = 0
y'(I) = O ,y"(l) = -I 4.
8. 4y'41 - 3y"' + 2y" + y' - 6y = 7x 2 - 3x + 4; 16. YP = x 4/15,x 3 y'" + x 2 y " - 2xy' + 2y == 2 x'
y(-1) = 0, y'(-1) = 2 , y"(-1) =0, y(-1) =0, y'(-1) = 0, y"(-1) == 0
"'
y (-1) = -7[
19. Pre":· l'heorem 4.4 for an arbitrary positive inte three solutions if the Wronskian at x = I is
ger 11.
I
where a0, a 1, ••. , a are constant real numbers is called a constant coefficient 11th order
11
linear differential equation. In this section we shall see how to find the general solution
I I
of the homogeneous equation
After we master this homogeneous case, we will turn our attention to finding particular
solutions in the nonhomogeneous case in the next two sections.
To solve constant coefficient linear homogeneous differential equations. we will
seek solutions of the form y = e4. Since
gives us
,11-I /.x , ).,: + aoe/.x -
+ · · · + 01.,,,e - 0.
a,,,.,,n e/J +a11 _,.,,, e (l)
�hat have a characteristic polynomial with n distinct real roots�The foJlowing examples
illustrate how we determine the general solution to this type of differential equation.
A2 +4:>.. + 3 = 0.
Factoring, we have
(:>.. + 3) (:>.. + 1) = 0,
which has roots :>.. - -3 ' -1· Th
us we have solutions
1- :=: _:=: I
4.2 Homogeneous Constant Coefficient Linear Differential Equations 191
= 2e-4x F 0.
3
Hence e -3 x, e-x are linearly independent and form a fundamental set of solutions for
•
this second order equation. The general solution is then
)...3 - 4)... = 0.
This factors into
)...()... + 2)()... - 2) = 0.
The roots are
0, -2, 2.
Hence these solutions are linearly independent and form a fundamental set of solutions
•
for this third order differential equation. The general solution is then
Possible rational roots of this characteristic polynomial are ± I and ±2. 1 We begin to
substitute the values J, -I, 2, -2 into our characteristic polynomial and find I is a root.
This tells us ).. - I is a factor of our characteristic p o lynomial. The othC'r factor can be
found by dividing)..- I into our characteristic polynomial (using either long or synthetic
division). Doing so, we find the quotient (which is the other factor: the remaind er is
necessarily zero) is )..2 - 4).. + 2 and hence our characteristic equatio n becomes
( ).. - I)()..2 - 41 + 2) = 0.
Using the quadratic formula, we find the roots of the quadratic factor )..2 - 4).. + 2 are
2+Jz,2-h
so that all the ro ots of the characteristic polynomial are
1.2+Jz,2-Jz.
You can also find these roots using the solve command in Maple or the appropriate
command in a suitable software package. Typing and entering
solve (lambda/\3-5 * lambda I\ 2+6 * lambda-2=0, lambda)
Maple also gives us the roots l, 2 + Ji, 2 - J?.. We now have that
ex , e<2+../2)x, /2-../2)x
are solutions to this differential equation. Computing the Wronskian of these three
functio ns, we find it to be nonzero. (Try it.) Hence these three functions are linearly
independent and fonn a fundamental set of s olutions for this third order equation. The
•
general solution is then
In Examples 1 3- you will no tice that the functions c orresponding t o the distinct real
oots of the characteristic polynomial have nonzero W ronskian and hence are linearly
�
inde �endent. We leave it as an exercise (Exercise 35) to prove this is always the case.
Thus we have the followmg theo rem.
THEOREM 4.6 If r,, ... , 'k are distinct real roots for the characteristic po lynomial of
a <n) +
ny Gn-lY
(n-1)
+ ... +a1y, + aoy = 0,
then e'•X '· · ·' e'tX are rmear IY .independent s olutions
of this differential equation.
A well-known result .
· about poIynomials .
over the integers called the Rational Roots Theorem states that if
+
P(x) = an x" an-IXn-t +···+a1x
+ao (aniO)
is a polynomial where each a· is an . Pand
integer and .if p/q •s a rational
. root of P(x) in reduced form where
q are integers• then pd'1v1·des' ao and t
have p dividing 2 and q d',v,.ding
. d'
.q. i 'd
v, e s n· Thus a rational root p/q of A3 _ 5A2 + 6). - 2 == 0 mus
0
I givmg us± t and ±2 as the only
possible rational roots.
4.2 Homogeneous Constant Coefficient Linear Differential Ec1uations 193
If the number of distinct real roots is equal to the order of the linear constant co
efficient differential equation (that is, k = 11 in Theorem 4.6), then the solutions corre
sponding to these roots must form a fundamental set of solutions which we record as the
following corollary.
COR(KL\.RY 4.7 If r1, ... , rn are distinct real roots for the characteristic polynomial of
an y<n) + an-IY(n-t) + ... + a,y' + aoy = 0.
then e''x, ... , e'"x form a fundamental set of solutions for this differential equation.
As a final example of this subsection, we <lo an initial value problem.
Factoring,
.l,.(). + 2)(A + l ) = 0.
By Corollary 4.7, it follows that the general solution to the differential equation is
y = Ct + c2e-2:t + c3e -x.
Using the initial conditions, we get the system of equations
y(O) = Ct + C2 + C3 = I
y'(O) = -2c2 - c3 = 0
y"(0) = 4c2 + C3 = -J
whose solution is c1 = I/2, c2 = -1 /2, c3 = 1. Therefore, the solution to the initial
value problem is
I l -2:t
Y = 2 - 2e +e
-x
•
4.2.2 Characteristic Equations with Real Repeated Roots
If the characteristic polynomial has repeated roots, our method for obtaining solutions to
a constant coefficient homogeneous linear differential equation does not produce enough
solutions to form a fundamental set of solutions. To illustrate, consider
y" - 2y' + y = 0.
Its characteristic equation is
A2 - 2). + I = (). - 1) 2 = 0
J 94 Chapter 4 Linear Differential Equations
Thus if u" = 0, y2 = ue gives us a second solution. Since the functions with zero
x
second derivative are linear functions, any u of the form u = mx + b will work. We
cannot use one where m = 0, however, since the resulting solution be' will not produce
a solution linearly independent of e . Let us use the one with m = I and b = 0 giving
x
these two solutions are linearly independent and the general solution of
y -2 y' + y = 0
11
is
Y = ciex +c2xe'.
The second solution Y2 = xex that was produced for the differential equation
2y' + Y = 0 is x times the solution � that we obtained from the root of mul·
11
Y -
tiplicity 2,3 A = I. This works in general. That is, if r is a root of the characteristic
polynomial with multiplicity 2, then
resu1ts in an equat10n that can be solved ,or& II using one of our reduction of order techniques from Section 3 ·7·
• •
are solutions. Further, this extends to roots of higher multiplicity. For example, if r is a
root of the characteristic polynomial with multiplicity 3, then
THfOREM 4.8 If)..= r is a root of multiplicity m of the characteristic polynomial of the homogeneous
linear differential equation
any<n> + Gn-lY(n-l) + ... +ail+ y = 0,
then
•
The general solution is
y = c1e-2..r + c2:xe-2..r .
•
The general solution is
Solution The roots of the characteristic equation can be found by using the rational root approach
as in Example 3 or by using a software package like Maple. They are 2, -3, and 4 where
•
4 is a root of multiplicity 2. The general solution is
Solution The roots of the characteristic equation are easily found to be O and 8 •:here 8 is a root
of multiplicity 2. The general solution is
Y= c, + c2e8x + c3xe8x .
We have
y'(x) = 8c2e 8x + c3( 8xe 8.r + e 8.r)
y"(x) = 64c2e8x + C3(8e 8x + 64xe 8·' + 8e 8x ).
Using the initial conditions gives us
y(O) = c, + C2 =0
y' (0) = 8c2 + C3 = 0
y"(O) = 64c2 + l6c3 = 2.
The solution to this system of equations is c, = J/32, c2 = -1 /32, c3 :::: l/4. Therefore,
the solution to the initial value problem is
I I I
v;;:: - - -e8x + -xe&x
• 32 32 4
•
�e now can determine the general solution to a constant coefficient Iinear differential
equation whose characteristic equation has only real roots.
+
II 4
y y =0,
the characteristic equation is
'A.2 +4 =0,
4.2 Homogeneous Constant Coefficient Linear Differential Equations 197
A= ±J=4 = ±2i.
We are now going to show you how to use such imaginary roots to produce real
valued functions that are solutions to the differential equation. To get this, we are first
going to show you how we raise e to an imaginary exponent. You will then see that this
gives us complex-valued solutions e'x when r is an imaginary number, which will lead
to two real-valued functions that are solutions.
Let us review some basic aspects of the complex numbers. Recall that if a. bare
real numbers, then z = a + bi is a complex number where i 2 = -1. We call a the
real part of z and write Re(z) = a; b is called the imaginary part of z and we write
z, z
I m(z) = b. The conjugate of z, denoted is = a - bi. For example, if z = 2 + 3i,
then Re(z) = 2, lm(z) = 3, and z = 2 + 3i = 2 - 3i.
We also note that since i 2 = - I, we bave
.
I. 3 =-!, .5 .
I =!, ...
00 tk ,2 t3 t4
/='\"_=I +t+-+-+-+···
L, k!
k=O
2 3! 4!
We use this same series to define e' when t is the imaginary number
t = ifJ: 4
'°' --
00 2 3 4
. (ifJ)" (i8) (i8) (i8)
e.r1 = = I + i8 + -- + -- + -- + · · ·
L, k!
k=O
2 3! 4!
Rearranging thjs series, we obtain
. j2()2 ;383 i484
e'9 = I +ie+-+-+-+· ..
- .. ·) .
2 3! 4!
2
= ( I - (} +
2
8 4
4!
- .. ·) + i 83
- +e
s
3! 5!
(e
Notice that Re(e;o) is the Maclaurin series for cos fl and that Im (ei8 ) is the Maclaurin
series for sin(). This gives us an important identity discovered by Euler called Euler's
formula:
e i9 = cos8 + i sin8. , p
To illustrate, we have that
e2ix = cos 2x + i sin 2x
4 We will work formally and ignore questions about the convergence of the series we work with here. All of
this can be carefully justified. These details are left for a course in complex analysis.
198 Chapter 4 Linear Differential Equations
cc
and
e-2ix = cos(-2x) + j sin(-2x) = cos 2x - i sin 2x.
For example,
COROLLARY 4.10 If a+ bi i s a root of a characteristic equation for a constant coeffici ent linear homoge
neous differential equ ation, then eaz cos bx an d eax sin bx are two linearly in d epende nt
soluti ons to the different ial equation.
Proof The fact that they are solutions follows from Theorem 4.9. C a lculating the Wronskian
0z
of e"x cos bx and e sin bx shows they are li nea rly in depen dent . (We leave this as
Exercise 38.) •
To illustrate how to use Corollary 4.10, co nsider a gain the differential equ ation
y" +4y = 0.
We know that the roots of its chara cteristic equation are ±2i. Using the root 2i, Corollary
4.10 tells us that y1 = cos 2x and y2 = sin 2x form a fundamental set of so lutions to
this second or de r di fferential equation. Its gen eral solution is then given by
y = c1 c os 2x + c2 sin 2x.
If we use the c onjuga te root -2i, we have YI = c os 2x and Y2 = - sin 2x. However,
their linear combinations would give the same general solution as Y1 = co s 2x an d
y2 = si n 2x.
It is a lways the case that if r is a complex root of a polynomial with real coefficients,
then its conjugate, r, i s also a root of the polynomial . Thus complex roots of a charac
teristic p olynomia l come in conjugate pairs. If a + bi is one o f these roots p roducing
the solutions eax cos bx and e"x sin bx, we will a lways have the solutions eax cos bx and
-e0" sin bx pr oduced by the conjugate root a - bi. Howeve r, these solutio ns produced
by the conjugate a - bi are just linear combinations of the solutions p roduce d by a+ bi
and hence we do n ot n eed them when fo rming the genera l solution .
We now look at some more examples involvi ng complex roots.
EXAMPLE 9 Determi ne the genera l solution to the foll owing differential equation.
2y" + 8y' + 26y = 0
Solution The charact erist ic equation is 2A2 +8). + 26 = 0 and its roots using the quadratic formula
are foun d to be
•
Thus the general solu ti on is
y = c1 e-ix cos 3x + c2 e-2x sin 3x = e-2z (c1 cos 3x + c2 sin 3x).
Solution The char acter istic equation is }..3 +}.. 2 - J.. + I 5 = 0. Using t he�ional Root Theorem
or a software package like Maple, the ro ots are fou nd to be -3, I + 2i, I - 2i. (If you
200 Chapter 4 Linear Differential Equations
use Maple, you will see it uses I for i.) The general solution is, therefore,
y = c1 e-3,. + ex (c2 cos 2x + c2 sin 2x). •
Solution The roots of the characteristic equation are -1 + i ./3, -1 - i ./3. The general solution
is
c, =0
-c, + ./3c2 = 1
from which we obtain
y= ./3 -x srn
. 1
3e x1
as the solution to the initial value prob
lem.
If the characteristic polynom ial has
•
a repeated real root >.. = r of multiplic ity m,
we know we can produce addition
al solutions by multiplying e'x by x2, ••• , xm-l
(Theorem 4.8). The same procedur x,
e will work for complex roots: If>.. = a+ bi is
a complex root of the characte
ristic polynomial of multiplicity m, it foliows from our
differentiation rule w'(x) = u'(x
) + iv'(x) for a complex-valued function w(x) =
u(x) + iv(x) that multiplying
y<
4)
+ 2y" + y = 0.
Solution The chara cteristic equa
tion is
,
4
+ 2, 2 + l = 0.
w•
4.2 Homogeneous Constant Coefficient Linear Differential E quations 20J
(,2 + 1)2 = 0
and hence i and -i are both roots of multi plicity 2. Using i, we obtain the solutions
cosx, sinx, xcosx, x sinx.
Calculating the Wronskian of these four functions, we find them to be linearly indepen
•
dent (try i t) and consequently the general solution i s
Ct cosx + C2 sin X + C3X cosx + C4X sinx.
We now can determine the general solution for any constant coefficient homogeneous
linear differential equati on provided we can determine all the roots of its characteristic
polynomial. The final example of this section makes us use all of the techniques we have
developed here.
Solution The roots of the characteristic equation are 0, 3, -4, 8, l + 2i, I - 2i, - 2 - 3i, -2 + 3i.
The root 3 has multipli city 4 while the complex roots -2- 3i, -2+ 3i have multip licity
2. This leads to the general soluti on
•
y = Ct + c2e3x + c3xe3.r + c4x2e3x + csx3e3x
+ c6e-4x + c7 e8.r + ex (cs cos 2x + c9 sin 2x)
+ e-2.x (c10 cos 3x + c 1 t sin 3x + c12x cos 3x + c13x sin 3x).
EXERCISES 4.2
In Exercises 15-26, find the general solution of the dif 33. (I + x)y" + xy' - y = 0, Yi = e-"
ferential equation in part (a) and the solution to the initial 34. x2y" - 2xy' + 2y = 0, y, = x2
value problem in part (b) for the differential equation in
pan (a). 35. Prove Theorem 4.6. (Suggestion: Use the result of
Exercise 7(d) of Section 1.7.)
IS. a) y" - y = 0
b) y(l) = 0,y'(I) =-I 36. Show ..:!_ (/a+bi)r) = (a + bi )e(a+bi)i.
dx
16. a) y" + y = 0
7
3 . Prove Theorem 4.9.
b) y(n) = -1,y'(n) = I
17.a) y"+4y '+8y =0 38. Show that eax cos bx and e0x sin bx are linwly in
b) y(O) = 0, y'(O) = -1 dependent.
18.a) y"+4 y '+4y=0 39. For the initial value problems in part (b) of Exer
b) y(O)=0, y'(O) = -1 cises 15-26, determine the limits as x-+ oo. Also,
use Maple (or an appropriate software package) to
19. a) 4y" + 4 y ' + y = 0 plot the solutions to these initial value problems to
b) y(O) = 0, y'(O) = -1 illustrate these limits.
20. a) 4y"+Sy '+y = 0
b) y(O) = 0,y'(O) = -1 In many cases it is important to determ1,1e what a solu
21.a) y111 + Sy" = 0 tion to a differential equation tends to (that is,determine
b) y(-1) = l,y'(-1) = 0, y"(-1)= 2 its limit as x � oo). Use the results from Exercise 39
to help with Exercises 40-42.
22. a) y111 Sy"= 0
-
b) y(-1) = I, y'(-1) = 0, y"(-1)= 2 40. Suppose the roots r 1 , ••• , r11 of the characteristic
equation are real and distinct. What is (are) the con
23.a) y111 y" - 6y' = 0
dition(s) on r 1, ••• , r so that the limit is O?
-
Y dx x2dz2 - x2dz
....
4.3 The Method of Undetermined Coefficients 203
II
b) Substitute y' and y" from part (a) into the Euler 2 2
46. x y" +4xy' +2y = 0 47. x y" - xy' + y = 0
type equation to get the constant coefficient
48. Suppose that y1 is a solution to the differential equa
linear differential equation
tion
d2 y dy
+ (A - 1) + By = O q2(x)y"+q,(x)y' +qo(x)y = 0.
dz2 dz
Show that substituting y = u y1 into this differential
In Ex.en..;�..::; 44--47, use Exercise 43 to transform the equation results in the differential equation
given Eu;�r type differential equation into a constant
1
coefficienr ,inear differential equation. Then determine
the genetal sr.lution to the given Euler type differential
y,q,(x)u" + (2y;qz(x) + y,q1(x))u' = 0.
Ii I
Observe that this latter second order equation does 11
equatior,. not contain u and hence can be solved by the first 1111
44. x2y"+2xy'-2y = 0 45. x2 y"-3xy'-5y = 0 reduction of order technique in Section 3.7.
ii
4.3 THE METHOD OF UNDETERMINED COEFFICIENTS
In the last section we learned how to solve the constant coefficient homogeneous equation
a n y"+ an -lY <n -l) + · · · + a1y' + aoy = 0.
I
In this and the next section we are going to consider nonhomogeneous equations. Recall
from Theorem 4.3 that we can obtain the general solution to a non homogeneous equation ii'
by adding together the general solution YH to its corresponding homogeneous equation I
and a particular solution yp of the nonhomogeneous equation. Now that the constant
coefficient homogeneous case has been studied, we focus our attention on finding a II Ii
particular solution to the nonhomogeneous constant coefficient equation. In this section 1
11
we consider a method for finding particular solutions called the method of undetermined I
coefficients. The basic idea behind this method is trial and error using common sense.
We illustrate it with some examples.
Ii
EXAMPLE 1 Determine the general solution to the following nonhomogeneous equation.
y" + 2y' - 3y = 4e2r
II
I, I
Solution We first solve the corresponding homogeneous equation
I '1
I
y" + 2y' - 3y = 0.
I!.
Its characteristic equation is
1:I ,. ,ii
). 2 + 2).. - 3 = (.l.. - l )(..l.. + 3) = 0. 111 II
Consequently, it has general solution '
...x + cze
YH = c1e
-3x
Ii
Now, let us try to find a particular solution to the nonhomogeneous equation. We
must find a function y p so that when w e substitute JP into the left-hand side of the I
differential equation, we get the right-hand side, 4e2r . Would it not make sense (from I I
what we know about the derivatives of e 2x ) to look for a particular solution o f the same
-- Ls, i,,!:.,•
204 Chapter 4 Linear Differential Equations
we obtain
5Ae2r = 4e2r
upon substituting yp == Ae2x into the differential equation. Hence w e obtain a solution
if
4
5A=4 or A=-;
5
that is,
4 2x
YP = -e
is a particular solution. The general solution to the nonhomogeneous cyuation in this
example is then
•
EXAMPLE 2 Detennine the general solution to the following nonhomogeneous equation.
y" +2y' - 3y == x2
Solution From Example I we know that the general solution to the corresponding homogeneous
equation is
YH =c1e-3x +c2e".
Based on Example I, you might be tempted to try YP = Ax2 as a particular solution.
However, if we substitute this into the differential equation, we obtain
2A + 4Ax - 3Ax 2 = x 2 •
Notice that we have three equations
for A:
2A =O,
4A =0,
-3A = I.
4.3 The Method of Undetermined Coefficients 205
Obviously, there is no real number A that satisfies these three equations. What caused
this not to work is that we introduced an x-term and a constant term on the left-hand side
of the differential equation when differentiating yp. If we add an x-term and a constant
term to YP we can get this approach to work. That is, let us instead look for a particular
solution of the form
2A +28- 3C = 0
4A-3B = 0
-3A = 1,
y=c1e-3x +c2e·r--x
4 14
l 2 --x--.
3 9 27
•
Notice how we included terms involving the derivatives of x 2 in yp in Example 2.
To put it another way, we sought a particular solution yp that was a linear combination
of the type of function on the right-hand side of the equation and its derivatives. In
Example I, the derivatives of e2x are expressions of this same form so that we did not
have to include additional terms.
The name method of undetermined coefficients for the approach of Examples I
and 2 derives from the fact that the scalars we seek in these linear combinations are
coefficients that we must determine. Let us do two more examples.
Solution Again from Example 1 we know that the general solution to the corresponding homoge
neous equation is
-3x + _x
YH = qe c2e •
206 Chapter 4 Linear Differential Equations
Because the derivative of sin x is cosx, we try the linear combination YP == A cosx +
B sin x. Substituting this into the differential equation gives us
y; + 2y� - 3yp = (-A cosx - B sinx) + 2(-A sin x + B cosx)
-3(Acosx+Bsinx)
= (-4A + 28) cosx + (-2A - 48) sin x = 3 sinx.
In order for A cos x + 8 sin x to be a particular solution, we see that A and B must satisfy
the system of equations
-4A + 28 = 0
-2A -48 = 3.
JO 5
•
EXAMPLE 4 Find the general solution to the equation.
A= 2_' B= --
17 17.
Therefore, the general solution is
y = cie
-3.r + c2eX + -e
17
8 ( .
2 _x cosx - -tr
17
srnx. •
Notice that in Exa mples 1-4 none of the tenns in the linear combination we tried for
a particular solution appeared in the general solution to the corresponding homogeneous
equation. The next example illustrates that this procedure will not work if this does not
hold and shows how we modify the procedure to ohtain a particular solution.
•
The next exam pie i s another example where we have to modify our original approach.
y" + y = xcosx.
Solution The roots of the characteristic equation for the corresponding homogeneous equ ation
y" + y = 0
are ±i. The general solution to this homogeneous equation is
YH = c1 cosx + c2 sinx.
The derivatives of x cosx involve terms with itself, x sinx, cosx. and sinx. We do not
need to include terms with cosx and sin x since these functions are solutions to the
208 Chapter 4 Linear Differential Equations
THEOREM 4.11 Suppose that>.. = r is a root of multiplicity m of the characteristic polynomial p(>..) of
the homogeneous linear differential equation
The proof of this theorem will be discussed in Section 5.2. While this theorem is
stated for a root r of the characteristic polynomial p()..), its conclusions hold even if r is
not a root of p()..) by considering the multiplicity of r to be m = 0. Thus this theorem
guarantees us that we can always find a particular solution to a nonhomogeneous linear
differential equation
polynomials as terms in g(x) since polynomials are linear combinations of terms of the
form xk e rx with r = 0.) We use Theorem 4.11 in the following examples.
Solution The roots of the characteristic equation for the corresponding homogeneous equation
are O, ±2i. Therefore, the general solution to the corresponding homogeneous equation
is
YH =C1 +c2cos2x+c3sin2x.
210 Chapter 4 Linear Differential Equations
13
Y"(0) = -4c2 - - = -1
16
Solving this system, the solution to the initial value problem is
3 .
Y = _!2 + 2_ cos,,�x+-sm-x ? +-x I 2 3 . 3 2 3 �- 15 2x
--x sm2x-- x e2x+-xe""" --e I
16 64 64 4 8 16 64
8
EXERCISES 4. 3
InEx.erc1ses 1-16, de termine the general s olu ti on of the 19. y" - y = 5ex ; y(l ) = 0, y'(I) = -1
given diffe1ential equation.
20. y" + y = 2cos x - 3si n x; y(,r) = I, y1(,r) = I
1. y" - y' -· 6y = 3e2x 2. y" +2y' +2y = 4
21. y -4y" +5y' = 7-2cos x; y(O) = I, y'(O) = 0,
3. y" + 25 y = X + 25 4, 2y"+3y'+y = COS X y"(O) = 2
111
q2(x)y
11
+ q,(x)y ' +qo(x)y = g(x).
Suppose that we have determined that y1, Y2 form a fundamental set of solutions for the
corresponding homogeneous equation giving us the general solution
YH = C1Y1 + C2Y2
to the homogeneous differential equation. We vary the parameters c1, c2 in this general
homogeneous solution to functions u 1 (x) and u2(x) and try to find a particular solution
of the form
of equations
u'1 y, + u2y2 = 0
g(X)
U1Y1 + U2Y2 = -
q2(x)
I I I I
[
Yt Y2
Y; Y2
u'
] [ u
2
1] [ =
0
g(x)/q2(x)
]
·
Let us use Cramer's rule to solve this system for u'1 and u2. Notice that the determinant
of the coefficient matrix of this system is
I :; :� I·
which is the Wronskian w(y,, yz). Thus
0 Y2
g(x)/q2(x) y2
u', =
w(y,, Y2)
and
Yt 0
y; g(x)/q2(x)
w( yi, Y2)
IY
I O I
+ y; g(x)/qz(X)
Y1 f .!..__-----�dx.
w(.v,, Y1)
For notational purposes let
I w = w(y1,Y2),
Y2
I O I
w, = g(x)/q2(x) Y2 '
214 Chapter 4 Linear Differential Equations
and
Y1
W2 =I
y\
YP = U1Y1 + U2Y2 I
and
J
= W
dx,
-;--
UJ
/
Y1 = cos2x , Y2 = sin 2x
for a fundamental set of solutions. We have
cos2x
w = I -2sin2x
sin2x I =2
2cos2x '
wi =
I O sin2x I = - tan2x sin 2x,
tan2x 2cos2x
and
=
I cos2x O I = cos2x tan 2x = sin 2x.
Wz -2sin2x tan 2x
Using the formulas for u I and u 2,
� j- tan2x sin2x
I
uJ = f dx = 1 sin2 2x I
2 =-- j - - = - (cos2 x-sec2x)dx
2 cos2x
w
2
1 . l
= sm2x - ln I sec2x + tan2x I
4 4
and
- 7
u2 = J w2 dx - sin 2x d l
I " .x = - cos2x.
w 4
4.4 The Method of Variation of Parameters 215
THEOREM 4.12 Suppose that q,,, q,,_ 1, ••• , qo, g are continuous on an interval (a, b) and q,,(x) #- 0 for
all x in (a, b). Further suppose that Y1, Y2, • • . , y,. form a fundamental set of solutions
of the homogeneous differential equation
qn (x)y<n> +q11 -1(x)y <n -l) + · · · + qo(x)y = 0.
Let w denote the Wronskian of y1, Y2, ... , y,. , B denote then x I column vector
and w; denote the determinant obtained from w by replacing the ith column of w by 8
for i = 1, 2, ... , n. Then
f w·
U; = � dx
sec2x)dx Let us use Theorem 4.12 to solve a third order nonhomogeneous equation.
Solution Solving the characteristic equation, we findits roots are -3, l, 4. We let
zp
and
u3 = J W3
-:; dx = J 4e2x
84e 2x dx
1
= 21x,
which leads to the particular solution
•
By Corollary 4.5,
the Wronskian w in Theorem 4.12 is neve zero on the interval
(a, b). Hence, once we know a r
fundamental set of solutions, the met hod of �an. ation
of parameters always gives us a
particular solution (but it may be in terms of rntegral s
that do not have closed forms)
on the interval (a, b). Consequently, unlike the thod
of und etermined coefficients, whi
. ch works only for certain typ es of g(x), the m et�:d of
van. at1o n o f parameters works for the
method of undetermined coeffici
any continuous g(x). When app11ca · ble, however,
ents often is easier to use than variation of parameters.
Exercise I of the following
exercise set illustrates this.
-- -·
4.5 Some Applications of Higher Order Differential Equations 217
EXERC15ES 4.4
Many second order differential equations arise from using Newton's law of motion:
"The net force applied to an object equals the object's mass times its <1cceleration." If u
represents the position of the object moving along a line, then
u(t) is the object's position along the line at time t,
v(t) = u' (t) is the object's velocity along the line at time 1,
and
a(t) = u"(t) is the object's acceleration along the line at time t.
Therefore, if m is the mass of the object and Fis the force, then from Newton's law of
motion we obtain
F = ma =mu",
and we have an equation involving a second derivative. If we can write F in terms of 11
and u, then we will have a differential equation involving u, v = u', and a= u".
The first example we consider is a famous historical problem i1!volving springs,
which arises in many engineering applications. We consider an object ol mass m attached
to a spring. The spring can lie on a horizontal (line) track o r be suspended vertically
from a ceiling. (We will assume the motion of the object is always along lhe same line.)
If you stretch or compress the spring and then let go, the spring wili apply a force to
the object and the object will move back and forth or up and down. The differential
equation describing the horizontal motion will be equivalent to the differential equation
describing the vertical motion so we develop only the equation for vertical motion and
use this same equation for horizontal motion.
The British scientist Robert Hooke (1635-1703) discovered that if a spring is
stretched a distance L, then the force Fs that the spring exerts is proportional to L
(provided L is not too large). This is known as Hooke's law and, in equation form, is
written
F.i = kl
where� is the constant of proportionality. Since k is a property of the spring, it is called
the sprmg constant.
If a� object of mass m is hung from a vertical spring with spring constant k, then the
.
sprmg will be stretched a length L by the object,
as illustrated in Figure 4.1. The force due
Figure 4.1
4.5 Some Applications of Higher Order Differential Equations 219
to gravity is given by mg where gis the acceleration due to gravity and acts downward.
The spring force is F., = -kl by Hooke's law. (The minus sign is there because the
spring force acts upward.) Since the system is in equilibrium, the acceleration of the
object is 0. S umming the forces actin g o n the object and using Newton's law gives us
ma = 0 = F = mg - kl.
Tbis equation tells us that we can detennine the spring co nstant by hanging an object
of weight w = mg from the spring and dividing by the elongation L of the spri n g. That
is,
mg w
k= = _
l L
When a mass-sprin g system is positioned so that the length of the spring is this distance
L the spring is vertically stretched by the mass, we say the system is in equilibrium.
Now, suppose that the mass attached to the spring is moved u units from equilibrium,
where we take u to be positive when the spring is stretched and u to be negative when
the spring is compressed. (See Figure 4.2.) Then
mu"(t) = F(t)
where F is the sum of all t he forces actin g on the object. We now determine these forces.
Equilibrium
Figure 4.2
The force due to gravity is the weight of the mass w = mg. The force exerted by the
spring is -k(u + L). (The minus sign is there because the spring tries to pull the o bject
up when u > O an d the spring tries to push the object do wn when u < 0.) There is also
force due to friction. The force due to friction has been studied exten sively an d there
is experimental evidence showing that it is proportional to the velocity of the obj�ct.
Since the fric tion al force opposes the mo tion, it has the opposite sign of the velocity.
Therefore, we use -J v = -fu' for the force due to friction, where f is a positive
constant called the friction constant. If another force (such as a person pushi ng on the
mass, for example) is applied to the o bject, we call such a force an external force. We
will let h(t) represent the external force. We now have that
220 Chapter 4 Linear Differential Equations
which we know has a unique solution. Therefore, if we are given or can lind m, J, k, h,
u 0, and u 1, we can detennine the solution giving the motion of the object.
The characteristic equation for this differential equation is
m)..2+J>..+k=0
and its roots are
-J-----
A=- ±Jf2-4km
-
2m
Since J2 - 4km can be positive, zero, or negative, we can have real, imaginary, or
complex roots depending on the size of the friction constant J. In particular, we see that
if there is no friction (that is, if J = 0), then the roots are imaginat"). If there is friction
(that is, if J > 0), then Re(>..) < 0. If there is no friction the mass-�pring system is
called undamped, and if there is friction the system is called damped.
We now look at examples that include each of these possibilities. First we consider
an undamped system.
EXAMPLE 1 An obje�t weighing 4 lb stretches a spring 6 in. Suppose the mass-spring system is
on a honzontal tra �k and that the mass is kept off the track by a cushion of air fro� a
cornpre�sor. (In this case, the friction is virtually zero and will be ignored.) Determine
the �Ott on of the mass if no external force is applied and the object is pulled 2 in. from
.
equ1ltbnum and then released.
w 4
k= Z = o.s = 8·
4.5 Some Applications of Higher Order Differential Equations 221
f =0.
Since there is no external force, h(t) = 0. Because the object is moved to an initial
position 2 in. beyond equilibrium, we have
I
u(O) = u0 = -.
6
The fact that it is released from rest means
u'(O) = u 1 = 0.
The initial value problem describing the motion of the object attached to the spring is
then
I I
-u II +8u =0· u(O) = . u'(O) =0.
s 6
The roots of the characteristic equation to this differential equation are ±Si. There
fore,
Since
8 4
and the a mplitude is
I
6
This means the object oscillates back and forth along the track covering a distance of
2/3 ft every rr /4 sec. The value 8 is called the frequency of the motion. •
The next example illustrates the difference in a damped system from the undamped
system in Example I.
222 Chapter 4 Linear Differential Equations
� � I
0.15 � � � � �
0.1
0.05
I I I I '
.5 2 2.D 4 45 ti
3 3.5
0 0. 5.5 (i
-0.05 ,-
-0.l ,-
-0.15
\ \ \ \ \ \ \ \
Figure4.3
EXAMPLE 2 An object of mass 2 kg is attached to a vertical spring and stretches che spring 25 cm.
The spring is hung in oil that offers a resistance to the motion of 8 kg/sec. Dete1mine
the motion of the mass if no external force is applied and the object i� given an initial
velocity of 25 cm/sec from its equilibrium position.
m = 2 and f = 8.
k == mg _ 2(10) _ SO.
I - 0.25 -
The initial value problem describing the motion of the mass is then
2u" + 8u' + 80u == 0; u(O) == O, u'(O) = 0.25.
Th e
The roots of the characteristic equation for this differential equation are -2 ± 6i.
general solution is given by
u(t) = e-21 (c, cos 6t + c2 sin 6t).
Using the initial condi tions, we obtain
u(O) = c 1 == O
and
2 3 4 5 6
Figure 4.4
EXAMPLE 3 An object of mass 25 g is attached to a vertical spring and stretches the spring 25 cm.
A 250-g mass is then attached to the spring and the spring is hung in an oil that offers
a resistance to the motion of I kg/sec. Determine the motion of the mass if no external
force is applied and the object is given an initial velocity of 5 cm/sec after being pushed
up 10 cm from equilibrium.
The characteristic equation of the differential equation has 2 as a repeated root. The
general solution is given by
0.2
0.1
-0.2
Figure 4.5
u(0)=
Since (from Example J) the· gener
6 u'(O)=O.
a1 soluti. on to the homo geneous equation is
u(t) = c1 cos St+ c2 sin St,
the method of undetermined coe
ffictentste
· lls us to look for a particular solution of Ihe
4.5 Some Applications of Higher Order Differential Equations 225
form
A =-2. B =0.
The general solution of the nonhomogeneous differential equation is then
u = c 1 cos 8t + c2 sin 8t - 21 cos 8t.
Finally, using the initial conditions, we find the solution for the position is
I I
11 = - cos 8r + -4 sin 8t - 2t cos 81.
6
As t -+ oo the term -2t cos 8r grows without bound. Notice the effect of this in
our graph of the solution in Figure 4.6. This solution indicates that the spring may be
damaged as time becomes large. It is well known in physics and engineering that this will
happen if a mass-spring system is oscillated by a forcing term with the same frequency
as the system. This phenomenon is known as resonance. •
Figure4.6
EXAMPLE 5 Do Example 2 if an exte.rnal oscillating force is applied to the object given by h(t) =
2 cos 4t newtons. 5
0.04
Figure 4.7
J
4.5 Some Applications of Higher Order Differential Equations 227
Resistance, R Capacitance. C
Ii) lod"'Wore, L
Impressed
voltage, E(t)
Figure 4.8
Another area where second order linear constant coefficient differential equations
arise is in the flow of electric current in a circuit, as in Figure 4.8. The resistance. R,
is measured in ohms; the capacitance, C, is measured in farads; and the inductance. L,
is measured in henrys. We assume that all of these are constant values. The applied
voltage, E(t), is measured in volts and can change over time. We let I be the current in
the circuit. If Q is the charge (measured in coulombs) on the capacitor, then
dQ
= I.
dt
The flow of the current in the circuit is governed by Kirchoff's second law, which
states: "In a closed circuit, the applied voltage is equal to the sum of the voltage drops
across the circuit." Voltage drops are determined as follows:
The voltage drop across the resistor is IR.
The voltage drop across the capacitor is Q / C.
di
The voltage drop across the inductor is L-.
dt
Combining these with Kirchoff's law, we obtain
Q
Ll'+Rl+ =E(t).
C
Differentiating this equation and using the fact that
dQ
- =I,
dt
gives us
I
LI"+ RI'+ -I=E'(t),
C
which is a second order differential equation in I. If we also are given the initial charge
and initial current or the initial current and the initial current's derivative, we have an
initial value problem. Exercises 17-18 ask you to solve such initial value problems.
228 Chapter 4 Linea r Differential Equations
EXERCISES 4.5
In Exercises 1-8, use 32 ft/sec 2 or IO m/sec2 for the ac In Exercises 9-15, an external force is applied to the
celeration of gravity, g, as appropriate. Also, use Maple mass-spring system in the indicated exercise of this sec
or another suitable software package to graph your so tion. Determine the motion of the object when this ad
lution and observe the behavior of the motion. ditional external force is applied. Also, use Maple or
another suitable software package to 2raph your solu
I. A 4-kg mass is attached to a vertically hanging tion and observe the behavior of the n otion.
spring. The object stretches the spring JO cm. As
sume air resistance on the object is negligible and 9. The external force 3 newtons is Jpplied to the mass
the mass is pulled down an additional 50 cm and spring system in Exercise 1.
released from rest. Determine the motion of this
10. The external force -3 cos I Ot m:v.,nns is applied to
mass. What are the frequency and amplitude of this
the mass-spring system i n Exerci1,,.; 2.
motion?
2. Do Exercise I if the object is further given an initial 11. The external force 2 cos 2t lb is ap�'lied to the mass
velocity of 2 cm/sec. spring system in Exercise 5.
3. A 2-lb object stretches a spring 6 in. The mass 12. The external force 6 sin t lb is applied to the mass
spring system is placed in equilibrium on a horizon spring system in Exercise 5.
tal track that has a cushion of air eliminating friction. 13. The external force 6e-21 newtor1� is applied to the
Determine the motion of this object if the object is mass-spring system in Exerci�c 7.
given an initial velocity of 2 ft/sec. What are the
frequency and amplitude of this motion? 14. The external force Je-1 1 10 ncwtor,-. 1s applied to the
mass-spring system in Exercise 7.
4. Do Exercise 3 if the spring is first stretched 2 in.
from equilibrium and then given an initial velocity 15. Consider the mass-spring initial ,·alue problem
of 2 ft/sec from its resting position.
mu"+fu'+ku = 0: u(O) = tlo, u'(O) = u1.
5. An 8-lb object stretches a spring 4 ft. The mass
spring system is placed on a horizontal track that a) Determine the value of the friction constant f
has a friction constant of I lb-sec/ft. The object is where the roots of the characteristic polynomial
pulled 6 in. from equilibrium and released. Deter for the differential equation change from
mine the motion of the object. complex to real. This value off is called the
6. A 5-kg ma�s stretches a spring 25 cm. The mass critical damping value.
spring system is hung vertically in a tall tank filled · b) Determine the conditions form, f, and k so
with oil offering a resistance to the motion of 40 that the system is (i) undamped. (ii) damped
kg/sec. The object is pulled 25 cm from rest and and oscillating, (iii) critically damped , and
given an initial velocity of 50 cm/sec. Determine (iv) overdampcd.
the motion of the object.
=
c) Let m = 1 = k, u 0 = 1 u 1• Choose values
7. A 2-kg mass stretches a spring I m. This mass is for f so that the system is (i) undamped,
hung vertically on the spring and then a shock ab (ii) damped and oscillating, (iii) critically_ ..
sorber is attached t �at exerts a resistance of I 4 kg/sec damped, and (iv) overdamped. Solve the miu al
to mot .1011. Determrne the motion of the mass if it is value problems for each of these systems and
pulled down 3 m and then released. graph the four solutions on one coordinate
8. A �-lb object stretches a spring 4 in. The mass system using Maple or another suitable
s �nng system .1s hung vertically and air offers a re software package.
_
s1stanc � to the motion of the object of 12 lb-sec/ft. 16. Determine the general solution t o
.
:�� object 1 � pushed .up 3 in. from rest and given an
1mt1al ve!oc1ty of 6 m./sec. Determine the motion mu"+ ku. = sin wt.
of the object.
Determine the value of w that induces resonance.
4.5 Some Applications of Higher Order Differential Equations 229
17. Deterriiine the current in an RLC circuit that has 19. Determine the conditions on L, R, C, and w so that
L = 0.5, R = 2, C = 1, and E(t) = 0 with initial the voltage h(t) = sin wt produces resonance in an
chars.c 1 and initial current l. RLC circuit.
18. Detrrir.ine the current in an RLC circuit that has
...__, L = 0.5, R = 0.2, C = l, and E(t) = 3 sin(2t) 20. Do Exercise 16 of Section 3.7 using the methods of
with ,;r, ;.1itial charge or current. this chapter.
I
• J
Linear Transformations
and Eigenvalues
and Eigenvectors
231
232 Chapter 5 Linear Transformations and Eigenvalues and Eigenvectors
both the domain and codomain of f are IR; the range of this function f is the set of
positive real numbers. (In particular, notice that the range of a functior. can be a proper
subset of its codomain.) If g : JR2 -+ JR by
[ X J 2 2
g =x -y,
y
the domain of g is JR2 and its codomain is JR; here the range of the funcuon g is JR, the
same as its codomain (why?).
With those preliminaries out of the way, we now state what we mean by a linear
transformation.
A function T from one vector space to another satisfying property I of our definition
is said to preserve addition; if the function satisfies property 2, the function is said to
presen,e scalar multiplication. Thus a linear transformation is a function from one
vector space to another that preserves both addition and scalar multiplication. In the
special case of a linear transformation T : v -+ v from a vector space v to itself, T is
sometimes called a linear operator.
Let us l �k at some examples of functions from one vector space t o another that are
and are not hnear transformations.
T
[ X]
:
x+y-z
� [ x+2y+
,J
5.1 Linear Transformations 233
We have
+ [ :: ])
Solution
T [ :: ] = T [ :: ::: ]
( z1 z2 z1 + z2
= x1 + x i+ Y1 + Y2 -z1 - z2
[ XI + X2 + 2y1 + 2y2 + Z1 + 22 ]
X1 + YI - Zt x2 + Y2 - z2
X1 + 2y1 + ZI ] + X2 + 2�2 + 22 ]
[ [
=
X + y-z
� c [ x+ Zy +Z ] � cT [ X] :
+
[ :: : : ] � [ , + x� �::: y, + I ] '
2
[ : ]) �
x
([ :: ]
T T
T
[ :: ]
+
T [ : ] � [ x, +'!, + I ] + [ ,2 /,� + I ]
�[ x, +x,t:l y,+2]
234 Chapter 5 Linear Transformations and Eigenvalues and Eigenvectors
r
([ ;;
for any two vectors
H: ]) is not the same as T [ ;: J + [: J
r
in JR2 • Since T does not preserve addition, Tis not a linear transformation. It is also the
case that this function does not preserve scalar multiplication. (Verify this.) t
T(cX) == A(cX)
== c (AX) = cT(X).
Hence T is a linear tran
sfo nnation.
•
vwr•-• •-•
5.1 Linear 1ransformations 235
We shall call the type of linear transformation T(X) = AX of Theorem 5.1 a matrix
transformation. The linear tranformation T : JR3 -+ JR2 by
then
In Section 5.3 we shall see that every linear transformation from JR" to !Rm is in fact a
matrix transformation. Even more generally, by using coordinate vectors we will see
that every linear transformation from one finite dimensional vector space to another is,
in a sense, given by such a matrix multiplication.
Differentiation and integration give rise to many important linear transformations.
In the case of differentiation, recall from Chapter 2 that D(a, b) denotes the set of all
differentiable functions on an open interval (a, b), which is a subspace of the set of all
functions F(a, b) defined on (a, b). W riting the derivative of a function f as DJ (that
is, DJ = f'), we may regard Das a function from D(a, b) to F(a, b),
D: D(a, b) -+ F(a, b).
This function D is called the differential operator on D(a, b). Since we know from
calculus that
D(f + g) = (f + g)' = f' + g'=DJ+ Dg and D(cf) =(cf)'= cf'=cDf
for any two differentiable functions f and g and any constant (scalar) c, it follows that
Dis a linear transformation from D(a, b) to F(a, b). In the next section we will look
back at the last chapter and see that much of the work we did there can be viewed in
tenns of linear transformations involving this differential operator D.
To give an illustration of how integration arises in connection with linear transfor
mations, consider the vector space of continuous functions on a closed interval [a, b] for
finite numbers a and b, which we have been denoting as C[a, b ]. Let Int(/) denote the
definite integral of a function fin C[a, b] over [a, h]:
and
Int(cf) = 1 b b
r .f(x)dx = clnt(fl
cf(x)dx = c la
for any f and g in C[a, bJ and any constant c and hence Int : C[a. b] - JR is a linear
transformation. In Chapter 7, we shall employ this type of linear traf' formation (only
on an interval of the form [O, oo)) in our study of Laplace transforms
The following theorem gives some elementary properties of linear transfonnations.
1. T(O) = 0.
2. T(-v) = -T(v) for any v in V.
3. T(u - v) = T(u) - T(v} for any u and v in V.
4. T(c1v1 +c2v2 + · · · +ckvk) = c1T(u 1) + c2T(v2) + ... + nT<vk) for any
scalars c,, c2, ... , ck and any vectors u1, v2, ••. , Uk in V.
Proof We prove the first two parts here and leave the proof of the third and 1t 1 urth parts as an
exercise (Exercise 14). We obtain part (I) by fi rst observing
T (v + (-v)) = T(O) = O
by part (I). Thus we have T(v)
+ T(-v) = 0, which implies T(-v) = -T(v). I
Part (4) of Theorem 5 ·2 can · . ·
. . b e d escnb
bmation of vectors .1s the Iinear com ed m words by saying' "T of a linear com·
.
b.rnat1on of T of the vectors " A consequence 0f it
IS
· th at once w� know what a linear
transformation T : V - � does to a basis of V,
we can determine what T does
to any other vector of v of V. The following example
illustrates how we do this.
f"
II Ii
I It 11Ii I
I I
Find: iI I
:
I I
I
iii 11
II
II !
II
II 1,
111 1, I
Solution Before solving this example, we point out that the vectors
ii
,!I
',
[l]f l[�]
i
I
do form a basis of �3. (Verify this if you are in doubt.)
1,
(a) We first write Ii II I
I:
i
[i]
II
i
,,
as a linear combination of our basis vectors:
1, '
'i
,,
I
This gives us the system
Ct+ C3 = I 11
Ct+ C2 = 3
C2 + C3 = 0, i I
'
Ct= 2. C3 = -1.
I II
I
i.J LI ,.
238 Chapter 5 Linear Transformations and Eigenvalues and Eigenvectors
Thus
J
(b) We proceed in the same manner as in part (a), only using
[ : ] fo plare of [
, [ l ] +,, [ : ] +" [ i ] = [ : ] .
==
CJ+ C3 = X
= - - .::...
C1 + C2 y
= + =-
C2 + C3 Z
x y z x y z x v -z
2 2 - 2' 2 2 2,
T [ : ] = c, T [ l ] T [ : ] , T [ � ]
C1 C2 + C3
+
2 2+2
+ ,, +
= G+ i- D [ '. ] +� + t + D [ � ]
+G-i+D[�]
=[ ;/2:::] •
From the answer to part (b) of
E ation
in this example is the same I'mear xample �· we can see that the linear transform
transfonnation as in Example I. Example 4 illustrates
5.1 Linear Transformations 239
a common way of specifying a linear transformation: Simply indicate how the linear
transformation acts on a basis for the domain. Once its action is known on a basis, the
action of the linear transformation on any other vector may be determined by writing
the other vector as a linear combination of the basis vectors and applying part (4) of
Theorem 5.2.
An important set associated with a linear transformation T : V - W is the kernel
of T, denoted ker(T), which is the set of all vectors v in V so that T(v) = 0: in set
notation,
AX =0.
Equivalently, N S(A) is the subspace of!R" consisting of the solutions to the homogeneous
system AX = 0. Notice that the kernel of the matrix transformation
T(X) = AX
is then exactly the same as the nullspace of A:
kcr(T) = N S(A).
of Example 1.
]-[
l 1 -1 : I -I ' 0 0 -3 I O ],
[ 2 0
I 2 I , 0 0 2 I 0 � I
][i l
a11 a, 2 a,11
a 2, a 22 a211
AX-[
we see that the set of vectors AX as X runs over the elements of JR", wnich is the range T
of the function T, consists of the linear combinations of the colu1w · of A, which is
I
C S(A). We are going to denote the range of a linear transformation T tiy
range(T
). 1
We could then restate what we have just discovered by saying if T is a i"latrix transfor
mation
T(X) = AX,
then
range(T) = CS(A).
of Example 1.
as we d id i n Secti on 2.4,
c:T(v) = T(cv)
er scala r multipli
e(T) and hence range(T) is close d und
gives us that cT(v) lies in rang •
cati on.
e(T),
relationship between the dimensions ofker(T), rang
The next theorem gives us a
sional.
and V when V is finite dimen
242 Chapter 5 Linear Transformations and Eigenvalues and Eigenvectors
THEOREM 5.4 If T : V --+ Wis a linear transformation where V is a finite dimensional vector space,
then
dim(ker(T)) + dim(range(T)) = dim(V).
Proof We prove this here in the case whenO < dim(ker(T)) < dim(V). Othe;casesthatmust
be considered are dim(V) = 0, 0 = dim(ker(T)) < dim(V), and O < u1m(ker(T)) =
dim(V), which we leave as exercises (Exercise 30).
Let us set dim(V) = n and dim(ker(T)) = k. Choose a basis vi, vi, . .. , vk for
k.er(T). By Lemma 2.11 we can extend this set of vectors to a basis v1, v2, .. ., Vk, Vk+ l;
..., v,. of V. Our proof hinges on the truth of the following claim.
Claim T(ut+1), ... , T(vn ) form a basis for range(T).
We first show these vectors span range(T). Consider a vector T(v) in range(T). Ex
pressing v as
we have
T(v) = T(c1V1 + · · · + CkVk + Ck+JVk+I + · · · +cnVn )
= Ct T(vi) + · .. + ckT(vk) + Ck+l T(vk+1) + · · · + C Tl Vn)
11
Then
dim(RS(A)) = dim(CS(A)).
5.1 Linear Transformations 243
EXERCISES 5.1
Detennine whether the given function is a linear trans 14. Prove the following parts of Theorem 5.2.
formation in Exercises 1-12. a) Part ( )
3
y
x-
1. T: nt2-+ R2 by T [ X = [ 35x + 2y ] 3y ]·
b) Part (4)
J !: J
l
x-y
IS. T [ : = [ �: :
n Jiif���]
3, U' � R' by T [ � ]
= [ : �:,;; 16. T [; ] = [-� :: � :, ]
5x + 2y - Sz
z
x z
x - 2y+ 5
4. T: iR.3--+ IR2 by T[ ] =[ 2 ].
: x + 2z
!7. T [ [
5. T: P2--+ P1 by T(ax 2 +bx+ c) = 2ax+ b. =
2 /2+ bx.
6. T: Pi -+ P2 by T(ax +b ) = ax
a(x + 1)2 +
XJ
7. T: P2 --+ P2 by T(ax +bx+ c) =
2 XJ - X2+.3X3 - X4
b(x + 1) + C. 18. T [ :: ] = [ 2x1 + 3x2 - X3 - 2x4 ]
8. T: P2-+ P1 by T(ax2+bx+c) = ax +bx+c+I.
2
3x1 + 7x2 - 5x3 - 3x4
9. T : F(-oo, oo) --+ F(-oo, oo) by T(f(x)) :;:
X4
so
19. Suppose T: JR. -+ JR. is a linear transformation
3 3
f(x) - 2.
10. T: F(-00,00) --+ F(-00,00) by T(f(x)) =
that
f(x - 2).
11. T: MmxnOR)-+ Mnxm (JR) by T(A) = A .
r
20. Suppose T : JR3 4 JR.4 is a linear transformation so In Exercises 23-26, find bases for th� kernel and the
that range of the linear transformation in the indicated exer
cise.
23. Exercise 15 24. Exerci,c 16
25. Exercise 17 26. Exerci,e 18
27. Describe the vectors in the kernel of the differential
operator D: D(a, b) -4 F(a, b).
28. Describe the vectors in the kernel of the linear trans
formation Int : C[a, b] -4 JR where Int(!) ==
J: f(x)dx.
29. In the verification of the claim of proof of Theorem
5.4, show that ifck+I Vk+l +···+en Vn lies in ker(T),
then l'k+l = 0, ... , Cn = 0.
30. Prove Theorem 5.4 in the case wheli:
a) dim(V) = 0.
_n
b) 0 = dim(ker(T)) < dim(V)
•) HodT[ c) 0 < dim(ker(T)) = dim(V).
31. Suppose that f : lR -4 IR preserve� ,calar multipli
cation. Show that f (x) = mx for some constant m.
(Hint: View f(x) as f(x · 1).)
b) FiOO T [;]
32. Suppose that V is a vector space and k is a scalar.
Show that T: V -4 V by T(v) = kv is a lin
21. Suppose T : P1 --+ P1 is a linear transformation so ear transfom1ation. Such a linear transformation is
that called a dilation if k > I, a contraction if O < k < I,
T(x+ I) =2x+ I, T(x - I) = 2x - I. and a reflection if k = -1. Why would they be given
these names? (Suggestion: Consider the results of
a) Find T(x). such transformations in JR.2 or JR.3.)
b) Find T(ax + b).
33. Let T : IR2 --+ IR.2 be defined by letting
22. Suppose T : IR3 --+ Pi is a linear tranformation so
that
Y - r sin(} ·
5.2 The Algebra of Linear Transformations 245
J J.
1. For each vector v in V there is a vector u in V
Obser v that so that T(u) = v.2
T c s(O + a)
T[ X =[ � 2. dim(ker (T)) > 0.
y rsm(8+a)
36. Recall that a function f : X --+ Y is one-to-one if
34. Let T : f�2 --+ IR.2 be defined by letting whenever J(x1 ) = f(x2), x1 = x2• Show that a
linear transformation T : V --+ W is one-to-one if
and only if ker(T) consists of only the zero vector.
37. Suppose V is a vector space, Xis a set, and there is
be the vector obtained by rotating a function f : V --+ X that is one-to-one and onto.
Thus, every element in X is uniquely expressible as
f ( v) for some vector v in V. This allows us to define
an addition on Xas
clockw :;� through the angle a. Show that T is the + J(u) = f(v + u)
J [ J.
f (v)
J
matri;, ,ransformation
where v and u are elements o f V and a scalar mul
1' \'" x
co� a sin a
=[ x tiplication on X by
- sm a cos a y
_ y
cf (v) = f(cv)
35. Prove the following fact, which is sometimes called where c is a real number. Show that X is a vector
the Frt•dholm alternative: 1 If V is an n-dimensional space under this addition and this scalar multiplica-
vector $pace and T : V --+ V is a linear transforma tion.
tion, thr.n exactly one of the following holds:
2
p ysicist Ivar Fredholm. 1866-19 7.
I Named for the mathematical h
g function T is onto.
2 This is the same as sayin the
- ---.- --------...
and Eigenvectors
246 Chapters Linear Transformations and Eigenvalues
and
s[
x
y
]=[ 2x-y
X +2y
l·
[ l [ l [ l [ l [ Yl
then
(T + S) X =T X +S X X +Y 2x -
y y y = x-y + x + 2�
and
l
� [ 2, : Y
3
T
l d [ : � : l � [ !: � !; l
(5 ) [ :
Our addition and scalar multiplication of linear transformations give us linear trans
formations back again.
Proof We verify this for T +Sand leave the proof for cT as an exercise (Exercise 15). Suppose
that u and v are vectors in V and a is a scalar. We have
(T + S)(u + v) = T(u + v) + S(u + v) = T(u) + T(v) + S(u) + S(v)
= T(u) + S(u) + T(v) + S(v) = (T + S)(u) + (T + S)(v)
and
COROLLARY 5.7 If T1' T2, ... , Tn : V -+ Ware linear transformations and c1, c2, ... , Cn are scalars, then
and
y .
x]=[2x- ]
s[ y x+2y
the composite ST is
2(x + y) - (x - y) ] = [ x + 3y
=[ ].
x+y+2(x-y) 3x-y
THEOREM 5.8 If T : V � W and S : W � U are linear transformat ions, then the composite
ST : V � U is a linear transformation.
Proof For any two vectors u and v of V and any scalar c, we have
ST(u + v) = S(T(u + v)) = S(T(u) + T(v)) = S(T(u)) + S(T(v))
= ST(u) + ST(v)
and
ST(cv) = S(T(cv)) = S(cT(v)) = cS(T(v)) = cST(V).
•
id ed size s are the same), a scalar multiplica
With matrices we have an addition (prov
the number of columns in the left-hand factor equals
tion, and a multiplication (provided
factor). Now, with linear transformat i ons, we also
the number of rows in the right-hand
ins and codomains of both line ar transformations are
have an add ition (provided the doma
and a multiplication (provided the domain of the left
the same), a scalar multiplication,
of the right-hand factor). Properties we have for these
hand factor equals the codomain 1.2
to linear transformations (compare with Theorems
operations on matrice s carry over
and 1.3).
R. S.
ns are defined. the following properties hold where
THEOREM 5.9 Provided the indicated ope ratio
and c and d are scalars .
and T are linear transfor mations
I. S+T=T+S
2. R + (S + T) = (R + S)
+T
3. c(dT) = (cd)T
248 Chapter 5 Linear Transformations and Eigenvalues and Eigenvectors
4. c(S + T) = cS + cT
5. (c + d)T = cT + dT
6. R(ST) = (RS)T
7. R(S + T) =RS+ RT
8. (R + S)T =RT+ ST
9. c(ST) = (cS)T = S(cT)
Proof We will prove the associativity property in part (6) and leave the proofs of the remaining
parts as exercises (Exercise 16). In fact, this associativity property holds or all functions,
not just linear transformations. To prove it (which is the approach used to prove all of
these properties), we take a vector v in the domain and verify that the 1unctions on each
side of the equation applied to v give us the same result: Applying thl! definition of the
composite of functions, we have
R(ST)(v) = R(ST(v)) = R(S(T(v)))
and
•
(RS)T(v) = RS(T(v)) = R(S(T(v))).
Therefore, the functions R(ST) and (RS)T are the same.
One other similarity between matrices and linear transfonnations we mention at this
time involves exponents. Recall that if A is a square matrix and II is a positive integer,
we have defined the nth power of A, An , as the product of n factors of A. Notice that
we can do likewise for a linear transformation that has the same domain and codomain.
That is, if T : V -. Vis a linear transformation and n is a positive integer, then T" is
the product (composite) of n factors of T.
We are now going to see how we can use the ideas we have developed involving
li �ear tr �nsfonnations to obtain a particularly elegant approach to the study of linear
differential equations. To simplify our presentation, let us assume that our functions have
derivatives of all orders on an open interval (a, b); that is, we will use C "°(a, b) as our
vector spa�e of functions. We will make use of two basic types of Iinear transformations.
O e type rnvolves the di�erential operator D : coo(a, b) ---+ coo(a, b) and its ?owers
� _ .
D ' �htch �mount lo taking the nth derivative. The other type involves multiplying
functions f m C (a, b) by a fixed function gin C (a' b): If g is a function in c (a, b),
00 00 00
define
toy:
L(y) = (T,q (x)D" + T,q _,(x) Dn-i + · · · + Tq,(x)D + Tq cx))(y)
0 (2)
To simpl i fy our notation, we leave off the Ts and write the linear transformation in
Equation (1) as
-1
L= q,,(x)D" + q.-i(x)D" + · · · + 41(x)D + l/o(x) (3)
indicat e s mul tiplicatio n
wh e re it is to be unde rstood that a factor q; (x) appearing in a tenn
also th e parenthe s e s about y,
by q; (x). In Equation (2), we l eave off not on ly th e Ts but
writing this e xpression as
i
Ly= (q,,(x)D" + l/11-1 (x)D "- +···+qi (x)D 4o(x))
+ y. (4)
Ly =0
hat finding th e solutio ns to such a l ine
ar
in this differential operator notatio n. Notic e t r tra sfor
nd i ng the vectors in t he kern e l of the linea
n
differ ential equation is the same as fi
ct, t e techniques you learned for solvin g homo geneous
mation in Equation (3). In effe
h
finding a basis
linear differ e ntial equa
tions in the pr e vious chapt e r amount to methods for
for this kern e l.
ene o us l inear differ ential equation has constant
In the special case where the homog
coefficients,
an/"> +an -i/ ) + · · · + a1y' aoy=
n-1 + 0,
eristic polynomial
whic h is th e sam e as the charact
n i
+ · · · + a1A + ao
p(A) =a11 +an-i>.. -
)..11
250 Chapter 4 Linear Transformations and Eigenvalues and Eigenvectors
with D substituted for .l.. Because products of polynomials in D comn1ute (see Exercise
18), we can factor polynomials in Din the same manner we factor polynomials int In
particular, if). = ris a root of the characteristic polynomial p ()...)so that
p ()...) = q()...)(. l. - r)
where q()...)is a polynomial in A, then
p(D)= q(D)(D- r). (5)
Equation (5) gives a very nice way of seeing why if ). = r is a root (either real or
imaginary )of the characteristic polynomial p(.l.), then eri is a solution to the differential
equation p(D)y = 0:
p(D)e'x = q(D)(D- r)e'x = q(D)(De'x - re'x)= q(D)(re'' - re'x)
= q(D)O = 0.
The fact that (D - r)e'x = 0 can be extended to obtain a jusu .,cation for our
method for constructing additional solutions when the characteristic polynomial has
repeated roots . Notice that
(D- r)2xe'x = (D- r)(D- r)xe'" = (D- r)(Dxe'x - rxe'x)
= (D- r)(rxe'x + e'" - rxe'x)= (D - r)e'-� = 0.
Likewise,
(D- r)3 x2 e'x = (D- r)2(D- r)x2erx = (D - r) 2(Dx2erx - rx2 e'x)
= (D- r)2 {r x 2 e'x + 2xe'" - r x2 e'x) = (D- r) 2 2xe'x
= 2(D- r)2 xe'r = 0.
Continuing, we obtain
THEOREM4.8 If). = ' is a root of multiplicity m of the characteristic polynom p(A)of the homo
. . _ ial
geneous lmear d1ffe rent1al equation
a y(rr)
n + an-lY(n-1) + - .. + a1y + y = 0, 1
3 A polynom · · ·
ial p(D) in D 1ssa1
· ·d toanmh1late
Thus, m. . . afu nction/inCoo(a b) ifp(D)f(x) = OforallxlR(O, b)·
this tennm ology' Eg�aJOn
r (6) . ' · sp(D)
saysthat(D- r)' annihilatesx -le,x Thesetofall polynol lllal
t hatanm'h'1lates a funct.mn f m C""(
i
. · ·1ator
a, b) .is called th e annihilator off. It can be shown thatthe annih1
ofxi-I erx con sists
. .
of all polynom1als in D that have (D
- r); asa factor.
5.2 The Algebra of Linear Transformations 251
then
Proof Since (D - r); is a factor of p(D) for each I s i ::: m, Equation (6) gives us that
p(D)xi-lerx = 0 for each l � i � m. •
Our differential operator point of view also can be used to prove Theorem 4. I I that
we used for finding particular solutions to nonhomogeneous linear differential equations
with the method of undetem1ined coefficients. Recall this theorem was the following.
THEO)[{f.M 4.11 Suppose that 'A. = r is a root of multiplicity m of the characteristic polynomial p()..) of
the homogeneous linear differential equation
an /n> + · · · + a1y +aoy = 0.
1
A complete proof of Theorem 4.11 is long and arduous and will not be included
bere. But to give you an idea of how it goes, let us prove part (I) when k = l. We must
show that
a,./ > + · · · + a1 y' + ao.Y = Axe'x
11
EXERCISES 5.2
l l
Find:
Find bases for the kernels of the following linear trans
l
1. cs+T) [ � 2. (S - T) [ ; formations from C00(-oo, oo) to C ""(-oo, oo).
J.
4. (S - 4T) [ �
13. D4 + D2 14. D 4 + 2D2 + I
· that
5. ST [ ; 6.rs[;J. 15. Complete the proof of Theorem 5.6 by sho wing
and
.
c is a
if T : V � W is a linear transformation
Let S, T : P1 � Pi be the linear transformations scalar, then cT is a linear transformation.
16. Prove the following parts of Theorem 5.9.
S(ax + b) = ax - 2a + b,
a) Part (I) b) Part (2)
T(ax + b) =ax+ 2a + b. c) Part (3) d) Part (4)
-
5.3 Matrices for Linear Transformations 253
]
e) Pa,! (5) f) Part (7) 24. Parts (b) and (c) of Exercise 23 illustrate the follow
g) P:irt(B) h) Part (9) ing fact: If T : JR• -+ JR"' is the linear transforma
tion given by
17. Show th,1t if g is a functi on in C00 (a, h), then Tg(x) is
•• •
a linear tnnsformation from C00 (a, h) to C°" (a, b). XJ G11X1 + ll12 X2 + ·' • + G1n Xn
18. Suppv:'f' X, a21X1 + ll22X2 + • • · + G2n Xn
[ ] [
p(D) = a,, o
n n
+ • • • + a1 D + ao
=
+ Gn_,D -l
T ;: Clm(Xl +llm2X2;+ +amn Xn
and
S : JR"' -+ R1 is the linear transformation given by
q(J)\ = bm Dm + hm -1Dm -i + · · · + h,D + ho
wheP' a., ... , a0 and b111 ho are constants.
, • • • ,
]
21. Use t!.� results of Exercise 20 to prove Theorem 4.3. Clm2
f
22. Show ,hat the set of linear transormations from a and
vector space V to a vector space W is a vector space. [ h,, h12 ... b,.
23. Let T and S be the linear transformations in Exer ...
h21 C122 bm
cises l-6. B= �
a) Find matrices A and B so that T and S are
expressed as the matrix transformations b11 h12 him
T(X) = AX and S(X) = BX. then ST(X) = BAX. Prove this fact. (ln Theorem
b) Find the matrix C so that the composite ST in 5.10 of the next �ection. we will general i ze this fact.)
Exercise 5 is expressed in the form 25. Suppose T : V -+ W is a linear transforma
ST(X) = CX. Then verify that C = BA. tion and suppose that T has an inverse function
c) Find the matrix D so that the composite TS in r-1 : W -.. V. Show that r-1 is also a linear
Exercise 6 is expressed in the form transfonnation. (You will have to use the fact that
TS(X) =DX.Then verify that D = AB. r- 1 is defined by r- 1 (w) = v where T(v) = w.)
l
a11
a1
[T]� = [ ;
You may wish to remember how to obtain [T)e by noting that its columns are the
coordinate vectors of T(vi), ... , T(v,,) with respect to /3:
[T]� = [[T(vi)] ,B ... [T(v11 )]p].
A special case that comes up often is when T maps V to itself (that is, T : V --+ V)
and the same basis a is used for hoth the domain and the codomain. In this case, [TJ:
is simply called the matrix of T with respect to a.
So that you get a feel for these matrices, let us do an example.
l
EXAMPLE I Let T : IR3 --+ IR3 be the linear transformation
T [ ]-[ 3x�E�3z
of!R3 , find:
[ J [ -: l[:]
5.3 Matrices for Linear Transformations 255
Solution
(a) We have
and
Thus,
in terms of the vectors in f3. This means we have to solve each of the three
vector equations
256 Chapter 5 Linear Transformations and Eigenvalues and Eigenvectors
and
[ ;] ] 2 [ : ] + c3 [
-
= c,U +,
n
We can simultaneously solve these three systems as follows.
I
I 7 6
-2
[;
-1 -1
5 5
_:]
I I
7 6
4u
I
-2 0 -8 -8
-1 -I I -9 -7 -7
2 6 4 8 ]
4[:
0
-2 0 -8 -8 -4
0 -2 I -10 -6 -10
-
2
4 [:
0 0 -4 -2
]
-2 0 -8 -8 -4
0 -2 -10 -6 -10
i
I
4[ -�]
0 0 I
-2 -1
I 0 4 4
0 I 5 3
The r!ght-hand side columns are the respective solutions to our three vector
equations and hence
-
-2 -1
1
]
[T]: = [ 4 4 2
5 3 5
:l
(c) Here we have to write
and
1
1 5 Q
-
I
[�
3 2 3 ]
I 5 0 -0
ented matrix
to reduced row-echelon form. we obtain the augm
I O O : 0 0 -1]
[ 0 I O , I -I 2 .
I
0 0 I : 4 I 0
(If in doubt, perform row operations and verify this.) This gives us
-
2
I
•
0]
three vector
You can see from Example I that finding a matrix of a transformation with respect
to standard bases as in part (a) is easy, while parts (b) and (c) illustrate that finding
matrices with respect to other bases is more involved. Shortly, we will see another way
of doing problems such as parts (b) and (c) of Example I that many find more convenient
-�]
to use. Toward this ohjective. we ne xt consider t he question: What happens to matrices
with respect to bases when we form composites of linear transformations? You might
conjecture that if composition is a type of multiplication of functions, perhaps these
matrices should be multiplied. If you did so. you are exactly corre ct.
258 Chapter 5 Linear Transformations and Eigenvalues and Eigenvectors
Proof Suppose that a consists of the vectors v1, ..• , Vn , f3 consists of the vectors w1, .•• , Wm ,
and y consists of the vectors u 1, ••• , Uk. Setting
T(vi) = a11W1 + ··· +a 1W m m
and
we have
[ a1 b11
[T]� = : and [SJp = [ :
am , bk l
Notice that
ST(v1) = S(T(v,)) = S(a11w, +·,,+am tWm )
= a11(b11u1 +· · · + bk1Uk) +· · · + a1111 (b1 U 1 + · · · + bkmU k)
m
The matrix
]
[ Pu P12 ... Pin
PI P22 ... P2tr
p= ; ...
Pnl P112 ... P nn
is called the change of basis matrix from a to /J. Note that the columns of P are the
coordinate vectors of w1, ••• , w,. with respect to a:
I (v) = v.
It is easily verified that I is a linear transformation (Exercise 11 ). Notice that since
J(wi) = W1 = P11V1 + · ·· + Pn1 V11
Theorem 5. IO,
[/]�[/]� = [/]�
and hence we have the following theorem.
THEOREM 5.11 If P is the change of basis matrix from a basis ex to a basis f3 of a vector s pace, then the
change of basis matrix from f, to a is p-l.
EXAMPLE 2 Leto: be the standard basis for JR3 and the basis f3 b e the has is of JR� consisting of
(the bases in Example !). Find the change of basis matrices from a to f3 and from P
to a.
Solution Getting the change of basis matrix from a to f, is easy since it is easy to \,rite the vectors
in f3 in terms of those in a:
n
Let us use Theorem 5.11 to get the change of basis matrix from f, to a, w hich is p-l .
[;
�u n
I 0 0] [ I I 1 1 0
-1 0 1 0 - 0 -2 0 -1 1
n-[�
I I 0 0 I O -1 -1 -2 0
0 2 1 I 0 0 -2 0
-2 0 -1 1 -2 0 -1 1
0 -2 -3 -1 0 -2 I -3 -1
The change of basis matrix from f3 to o: is
then
�l •
I 0
p -l = [- 1/2 -1/2
3/2 1/2 -)
We are now ready.to o�tain the alternative approach mentioned after Example is
bas
!·
S ppose T .. V - W is a hnear transformation. If a is a basis for V and f, is a
u
5.3 Matrices for Linear Transformations 261
for W, we get the matrix [T]�. Suppose we use different bases a' for V and fJ' for W.
Then we get the matrix ITJ!:. Is there a way of converting IT J� to [TJ!:? The answer
is yes, and it is shown in the following theorem.
THEOREM 5.12 Let T : V --+ W be a linear transformation, a and a' be bases for V, and f3 and /J' be
bases for W. If P is the change of basis matrix from a to a' and Q is the change of b asis
matrix from fJ to fJ', then
Proof Since P = [[]�. and Q = [/]:,,Theorem 5.11 and then Theorem 5.10 give us
rn:·rr1cu1�. = u1�·rn1�.
Q- 1 rTJ�P =
/J'
= ([TlJc,,
fJ'
= [Tla •· •
An important special case of Theorem 5.12 is Corollary 5.13.
CORO[ LARY 5.13 If T : V --+ V is a linear transformation, a and fJ are bases of V, and P is the change
of basis matrix from a to /J, then
EXAMPLE 3 Use the results of part (a) Example I and Example 2 to do p art (b) of Ex ample I.
-1
Solution By Corollary 5.13, we have:
3/2 1/2 0
-1 62 ] = [-42 •
:] [ _;
6 -I
0
]
-2 4 2
[
1/2 -1/2
5 5 3 5
3/2 1/2 -1 5 5
I, and 3.
of examples similar in nature to Examples 2,
Let us look at another set
ar transformation
EXAMPLE 4 Let T : p2 --+ P2 be the line
T(ax +bx+ c) = 2ax +
2 2
(2a + 2c)x - 2a + 2b.
2 2.
the standard basis consisting of x , x, I for P
(a) Find [Tl� where a is
matrix from a to the basis fJ for Pi consisting of
(b) Find the change of basis
x 2 +x,x
2
- I,x-1.
262 Chapter 5 Linear Transformations and Eigenvalues and Eigenvectors
S0l11tio11
(a) As
n
T(I) = 2x = Ox 2 + 2x + 0,
we see
1n: �
[ _� �
(b) Writing the vectors in f3 in terms of the vectors of a,
x2 + x = x 2 + x 0,
+ - I = x2 + Ox - 1.
x2
x - I = Ox 2 + x - I,
P�[i �l
we see
I
n
0
-1 -1 T
(c) This is p-1 .
[ 1 1 1 0
n
0:100] [1 0
I 0 1\010-+ O -1 -1 1
0 -1 -1 , 00 I o -1 -1 0 0
-[:
-u
0 0
-1 I -1 I
0 -2 -1
]
0 0 I
:
-2 0 -1
0 -2 -1
Hence our answer is
[ 1/2
1/2 1/2 ]
p-l = 1/2 -1/2 -1/2
-1/2 1/2 -1/2
5.3 Matrices for Linear Transformations 263
J[
(d) By Corollary 5.13,
;J[i _: J
[T]: = p-J[TJ:P
1/2 2 0
=
-: J [ �
1/2 1/2
[
1/2 -1/2 -1/2 2 0 0
_:]
1
l.x + 2,
= [:
-1 0
-I -1
=U �l •
0
2
0 -2
THEOREM 5.14 Suppose T : V -+ Wis a linear transformation. Let a be a hasis for V and /3 be a basis
for W. If vis a vector in V,
I .0 [T(v}]11 = [TJ![v]a,
-1
n-
0 0 ists of w1, ••• , Wm.
••• , Vn and the basis /3 cons
Proof Suppose the basis a consists of VJ,
We have
[v]. = [
[T]� =
[ a11
:
OmJ
where
264 Chapter S Linear Transformations and Eigenvalues and Eigenvectors
Hence
G.1JC1 + · · · + a,,, c n
[T(v)]p = [ J
:.: . J [ : J-
:
G.mJ Ct + · · · + G.mnC11
[TJ/[,
.J •. •
Solution
]
I I
� ] � � -2 ()
:]
I [
I 3 0 -I -1
0 2 I 3] [2 0 0
-2 0 I � 0 -2 0
0 -2 I O O -2
Ct= 2, C3 = --
2
we see
[ V ]Ji = [ -1/� ] .
•
-1/2
-1 -I ] [ 2 ] [-3 ]
(b) By Theorem 5.14 and Example 3,
-2
[T(v)]p = [TJ�[v)p = 4 2 -1/2 = 5 .
[ ; 3 5 -1/2 6
(c) From the coordinates with respect to f3 we found in part (b),
Notice the result of part (c) agrees with the resul t we obtain by applying the formula
for T in Exampl e I to v.
Solution
(a) Here we need c1, c2, c3 so that
3x2+4x-5=c1(x2 +x)+c2(x2 -l)+c3(x - 1)
= (c1+ C2)x2 + (Ct +C3)x - c2 - C3.
nvectors
266 Chapter S Linear Transformations and Eigenvalues and Eige
l � [; J
Solving the resulting system,
[;
0
u _n
!
1 0 I
0 -1 1
-1 -1 I -5 0 -1 -1
I 0
--+ -1 l
0 -2
C3 = 3, C 2 =2, Ct= I
!l
we obtain
[vJ, -[
(b) We have
Again, notice if we apply the formula for T (from Example 4 this time), we will
obtain the same result as in part (c).
Theorem 5.14 is what we were referring to when we earlier made the stateme?t
that all linear transformations on finite dimensional vector spaces are, in a sense, mat�
multplication. In the special case when we have a linear transformation T : IR" --+ JR
we can see that T is in fact a matrix transformation as a corollary to Theorem 5.14.
COROLLARY 5.15 If T : !Rn --+ !Rm is a linear transformation and A is the matrix of T with respect to the
standard bases for ]Rn and !Rm, then
T(X) = AX.
Proof If adenotes the standard basis for IR" and f3 denotes the standard basis for IR'"''
Theorem 5.14 tells us
[: HiH-il
1 . Let T : ne � IR 2
be the linear transformation r
[ Xt ] = XJ +x2 ] . ��: �
T [ }. s
X2 Xt - X2 'll"'°
a) Fird [T]� where a is the standard basis for JR2 .
b) Let f3 be the basis consisting of
•=[-!]
4. T : JR3 - R.3 by
0 hm 1 [T(v)]ft.
g) Use the result of part (f) to find T(v).
;J
2. T : JR2 - iR2 by 5. T : P2 � Pi by
[�l[�l
sis for P2; f3 the basis consisting of x 2+ I, x + I,
2x 2 + I; V = x 2 + 2x - 2.
O;
7. D : V � V where V is rhe set of solutions to the
differential equation y"+ y = a the basis of V
� = [ _: ] consisting of sin x, cos x; fj the basis consisting of
sin x + cosx. sinx - cosx; 11 = 3 cosx - 2 sinx.
3 8. S: M2x2(IR) - M2x2(R.) by S(A) = A7 ; a the
3. T : JR3 � JR by
standard basis for M 2 x2 (lR); f3 the basis consisting
17x1 - 8x2 -12x3 of
16Xt - 1X2 - )2X3 ] ;
16x1 - 8x2 - l lx3 01·
[ I O ] [ 0 I ] [ -1 0 ] [ 0 -I ]
o· 01·1 o·
a the standard basis for JR3; f3 the basis consisting v=[�
of !]·
268 Chapter 5 Linear Transformations and Eigenvalues and Eigenvectors
9. Suppose that v1, v2, v3 form a basis ex for a ve�tor IX = 0 or OX= 0, respectively. A,sume Wis not
space V and T : V --+ V is a linear transfonnatJon one of these two subspaces. Let V1 , Vz, ... , vk be a
basis for W. Extend this to a basis v,, v2, ..• , vk,
such that
vk+1, ••• , v,, of IR". Let T : !Rn -+ JR" be the linear
T(v1 )=v, - v2, T(v2)=v2 - V3, transformation so that
T(V3) = V3 - VJ. T(v 1 )=0, T(vz)=O.
a) Find [T]�. T(vk) = 0, T(vk+1) = VH\,
b) Fine.I [T(v)lu if v = v, - 2v2 + 3v3.
c) Use the result of part (b) to find T(v) in terms
T(v11 ) = Vn·
Of VJ, V2, V3. Now use T to obtain a matrix A so that W is the set
10. Suppose that v1 , v2, v3, v4 form a basis a for a vec of solutions to the homogeneous sy�tem AX= 0.)
tor space V and w J, w2 form a basis f3 for a vector 16. In Exercise 25 of Section 5.2 we saw that if a lin
space W. Suppose that T : V --+ . W is a linear ear transformation T : V --+ W has an inverse,
transformation such that then r- 1 : W --+ V is also a linear transfonua
T(v2) = -W1 + 3w2, tion. Suppose that a is a basi� for V and f3 is a
basis for W. Show that if T has an inverse, then
T(v3) = w1 +2w2, T(v4) = 3wz. 1
[T-l]p = ([T]�)- .
a) Find [T]�.
l
In Exercises 17-20, use Maple or another appropriate
h) Find (T(v)]fi if v = 4v, + 3v2 + 2v3+ V4. software package as an aid in doing the parts of Ex
c) Use the result of part (b) to find T ( v) in terms ercise I for the given linear transformation, bases, and
of w 1 and w2. vector v.
11. Suppose that Vis a vector space. Show that 17. T : JR4 --+ l!4 by
I : V -+ V by I (v) = v is a linear transformation.
XI - X z + X 3 - X 4
12. Suppose that T : V -+ W and S : V --+ W are
linear transformations, a is a basis for V, and f3 is a 2x1 - X2 + 2.t3 + X4
;
basis for W. Show that: 3x l + Xz - X3 + X4
a) IT+ SJ�= [T]�+ [SJ�. -X1 + 3x2 - 5X3 + X4
b) If c is a scalar, [cT]� = c[T]�.
a the standard basis for JR4, fJ the basis consisting of
13. Suppose that T : V-+ Vis a linear transfonnation,
the vectors v,, v2, v3 fonn a basis a for V, and
5
et the :,tandard basis for JR , {J the basis consisting of 19. T : P., _.. p3 by
T(a3.t 3 + a2x2 +t11x + aol = (a3 + u2}x
3
J 0 2 -] 2
0 I 0 2 + (a2 - a, }.t 2 + (a3 + a2 + a, }x - £11 + a2;
0 0 2 CJ the �tanda rd ba�i� for P.1: f1 the 3basi� consist
I 0 2 ing of x3 - x2 + I . x2 + x + I. x - I. x - L
2 -1 u = 3x3 - 5x 2 + 2x- I.
0
I
20. D : V f -+ V where V i� the set of �olution� to
-3 1°'
'
the dif erential equation y + 2l•" + y = ()· a
V= -2 the ba�i, con�isting of si� x, cos�. x si� x. x c�s x
I for V; f1 the basi� consisting of sinx + xsinx.
3 sinx - xsinx. co�x + xcosx. co�x - xcosx:
v = 3sinx - 2cosx+xsinx - 3x cosx.
OF MATRICES
5.4 EIGENVALUES AND EIGENVECTORS
ero column vector v in IR" ,o that
If A is an n x n matrix. an eigenvector of A is a nonz
the vector
r g
eige nvec tor and eige nval ue are partial translations from the cor espondin
The terms wor d for va lue. The
Eigemrert, Wert being the German
German words Eigenvektorand rent,
s translation s. some of which are prope r. inhe
German adjective eigen has
variou
tors and eige val ues ca lled
Sometime� you will find eigenvec
n
special, and characteristic. espective ly. As men tion ed in the in
characteristic values. r
characteristic vectors and s in a umb er of p lace s in
, eigenva lues and eigenvector arise n
troduction to this chapter will encounter in the next chapter.
ation�. one of which you
mathematics and its app lic d eigenvalues of square
next section is to study eigenvectors an
Our purpose in this and the nv cto s exte nd to linea r tran�fonnations
eigenvalues a nd eige
e r
matrices. The concepts of Section 5.6.
itself. as we �hall sec in
from a vector space to . If:,, is an eigenvalue
how we can find eigenvectors and eig envalues
Let u s investigate d with >. ,� ill be the nonzero �olutions to
eigenvalue s a s�ociate
of an n x n matrix A, the
rs
270 Chapter s Linear Transformations and Eigenvalues and Eigenvecto
the equation
AX= ).X,
or
(U-A)X=O
where. I is then x n identity matrix. This last matrix equation is that of a homogeneous
system of n linear equations in n unknowns with coefficient matrix Al - A. Since
w e know that such a homogeneous system has nontrivial solutiom if and only if its
coefficient matrix is not invertible, which in tum is equivalent to its coefficient matrix
having determinant zero, we have just discovered the follow ing fact.
The equation
det(U - A)= 0
is called the characteristic equation of the matrix A. Upon expand in;� the determinant
det(H - A) we will have a polynomial of degree n in ).. called the: characteristic
polynomial of A. Theorem 5.16 then tells us that we can find all the eigenvalues
of a square matrix A by finding all of the solutions of its characteristic equation (or,
equivalently, all of the roots of its characteristic polynomial).
A=
[ -2
•
= (). - l)(J... - 2) - 6 =).2 - 3). - 4 = ().. - 4)(,\ +I)= 0.
fts solutions, which are the eigenvalues of A, are).
= 4 and ,\ = -1.
Now that we have a method for finding eigenvalu der
es of a matrix A, let us consi
h?w w e may find the eigenvectors of A.
Suppose ). is an eigenvalue of A. Since the
eigenvectors associated with ).. are the nontri
vial solutions to the homogeneous system
(H - A)X = 0• they aI ong w·ith the tnvial · ·
solution form the nullspace of)./ - A '
5.4 Eigenvalues and Eigenvectors of Matrices 271
N S(H - A). We call this nullspace the eigenspace of). and denote it by £).. Since we
learned how to find bases of nullspaces of matrices in Section 2.4, we know how to find
a basis for an eigenspace E,,_ . All the nonzero linear combinations of these basb vector�
give us all the eigenvectors associated with>...
-3 J
A
= -2 2
[
in Example I.
Solution In Example I, we found that A has two eigenvalues,>.. = 4 and). = -1. We individually
l
consider each of these eigenvalues. For). = 4, the cigcnspace is the null space of
4/ - A= [ 32 23 .
[:: i �J-[� 0. a
we see our solutions are
[ : ]- [ -: J � [ -: l y
and hence the column vector
forms a basis for £4• The eigenvectors associated with the eigenvalue>.. = 4 are then the
vectors of the form
c
[ -I l
where c is a nonzero scalar. For>.. = -1, we need to find a basis for the nullspace of
-I-A =
[-2 3 ]·
2 -3
l
Here we find (try it) that we can use
[ 3/2
272 Chapter 5 Linear Transformations and Eigenvalues and Eigenvectors
as a basis for £_ 1 (although some prefer to multiply this vector by 2 to eliminate the
fraction and use
[!l
as the basis vector). Using the former basis vector, the eigenvector:- associated with
).. = - I are of the fon11
EXAMPLE 3 Find the eigenvalues and bases for the eigenspaces of the following matrix.
A= [ 2 -\ 3]
0·-1 0
0 0 -1
The eigenvalues of A are then).. = 2 and).. = -1. Next we find bases for the eigenspaces.
: 0] [ 0
For l = 2, we find a basis for the nullspace of 2 / - A.
]
. '
-3
-u
1 -3
p3 0 0
0 i O
3 O
- O 0
0 0
9
3
�]
I I
�
I 0
0 I
0 0
[ � J � U J-x [ D
The homogeneous system has solution
s
5.4 Eigenvalues and Eigenvectors of Matrices 273
[ �]
Next we carry out the same procedure for ).. = -1. The augmented matrix for the
as a basis vector for £2.
[ �]
-
-� -�
0 �
0 0 0
=:]
eigenspaces of the following matrix .
. EXAMPLE4 Find the eigenvalues and bases for the
A= [ � -�
.. -
2 A -2 5
det(H - A) =
-2 -1 )..-8
= (>.. - !)(().. - 2)().. - 8) + 5) - 2(2().. - 8) + JO)+ 6(-2 + 2().. - 2))
u O]
from which we see the eigenvalues are). = 3 and A= 5. Consider). = 3. From
6:o] [2 2 6
�u
2 I
l 5 i O -+ 0 -1 - I ! 0 �
-1 -5 , 0 0 I 1 : 0
�]
0 4
-1 -1
0 0 '
we see the solutions of (3/ - A)X = 0 are
u
forms a basis for £3 .
Finally, we consider).= 5. From
2 6: OJ [ 2 3 ' 0
n�u n
1
3 5 : 0 -+ 2 3 ]
5
() -2 -1 -3 0
j ()
-1 -3 I
[:
3 0 2
2 2 1
� 0 0 0 0
we see the solutions to (5/ _ A)X = 0 are
as a basis for E5 •
[ =:] •
-
5.4 Eigenvalues and Eigenvectors of Matrices 275
From Our study of linear algebra (matrices, dete nninants, vector spaces. and linear trans-
formations) up to this point has involved the use of the re al numbers as our unde rlying
number system. But imaginary numbers can arise as roots of th e characteristic poly
nomial of a square matrix just as they did for the characteristic polynomia l of a homo
geneous linear differential equation with constant coefficients in the previous chapter.
Consequently, imaginary numbers will come up in our work with eigenvalues and eigen
vectors. All of th e linear algebra material we have covered works equally well if complex
numbers are used in pl ac e of real numbe rs. Th e final example of this section illustrates
how we adapt our method for finding eigenvectors to imaginary roots of th e characte ristic
polynomial.
Using the quadratic formula, we find the roots of the characteristic equation are
>.. = 2±� = ±i
I .
2
For')...= l + i, w e find the complex solutions to ((I + i)l - A)X = 0.
[ _; : ! � ] R, � [ � 0 ! n -;R, � [ � -� : �]
rn, +
The complex solutions are
[ 1 ]
numbers.
forms a basis for E t +i over the complex .
Finall y, we do the corres pondin g work for the eigenvalue >.. = I - , .
i 0 -i 0
1
j Rt l
•
1
- [
[ 0 0 : 0 ] 0
[ -1 -i : 0 ] iR2 - R1 - -
276 Chapters Linear Transformations and Eigenvalues and Eigenvectors
A= [
THEOREM 5.17 If A is a square matrix with real entries, 11. = r is an eigenvalue of A, and v1, v2,. · ·, Vk
form a basis for the eigenspace E,, then r is also an eigenvalue of A and VJ, v2, ···•vi
form a basis for the eigenspace E;.
Proof e th at
�i .nce the characteristic polynomial of A has real coefficients, we immediately hav
' is also a root of the characteristic polynomial of A and hence r is an eigenvalue of A.
To see why v1, v2, · · · , -
vk "
1orm a bas1s
· for E;: , we first note that smce
Av= rv
if and only if
AV = rV or A ij = A ij = r ij'
e 19)
the vectors of E, and E, are conjugates of one another. This gives us (see Exercis
5.4 Eigenvalues and Eigenvectors of Matrices 277
that VJ, v2, ... , vk span £,. To see why the,y are linearly independent, suppose
CJ iij" + C2 V2 + • • · + Ck Vk = 0.
Then
or
Because of Theorem 5.17, there is now no need to separately find a basis for the
eigenspace of a conjugate root as we did in Example 5-all we have to do is conjugate
the basis vectors. For instance, once we find the basis vector
EXERC1SES 5.4
.
4 [ : =! ] 5. [ � � ] 6. [ � !]
1 .
5 [ ! -I
-2] 16. -
[ -4
4 4
5 ]
I O O 3 I -1
[ ] [ ]
17. 0 0 I 18. 0 0 -2
7. - 8. � : -� ] 0 1 0 0 I 2
[ : � � ] [ -
. [ � � n_! ] ,. [ =: ; =n
-2 0 I O O 2 19. Suppose that U and W arc subspaces of a vector
space V over the complex numbers and suppose
that the vectors of V and W are conjugates of one
another. Show that if u 1, u 2, ... , u, span V, then
ut, u2, ... , uk span W.
20. If D = diag(d 1 • d2, ... , dn ) is a diagonal matrix,
what are the eigenvalues of D? Also, find bases for
11.
[
�
I
-�
O -2
12.
[
: �
0 -1 0
!] the eigenspaces of D involving the standard ba�is
vectors for iR" .
278
rs
Chapter 5 Linear Transformations and Eigenvalues and Eigenvecto
21. Show that the eigenvalues of a triangular diatrix are will i nclude each eigenvalue, its multiplicity as a root
its diagonal entries.
of the characteristic polynomial, and a basi s for the as
sociated ei genspace. Use Maple or ,wother appropriate
22. Prove that a square matrix A is not invertible if and
software package 'to find the eigenvalues and bases for
-n -n
only if zero is an eigenvalue of A. the eigenspaces of the following matrices.
-! ] _!
23. Show that if A is an invertible matrix and i f v is an
27. [ � -� 28. [ � -� l
eigenvector of A with associated eigenvalue }.. = r,
then v is an eigenvector of A- 1 wi th associated
eigenvalue I /r. -2 3 2 -2 3
24. Use the result of Exercise 23 to find the eigenvalues 1
and bases for the eigenspaces of the inverse of the ; ::
29. [ ·: :: 30. [
matrix in Exerc ise 11.
l ·� :; :; =; l ·1
25. Show that i f v is an eigenvector of a square matrix
A with associ ated eigenvalue }.. = r, then v i s an
eigenvector of A' with associated eigenvalue rk for
any positive integer k.
26. Show that i f A is a square matrix, then A and Ar 5 10 15 -27
have the same ei genvalues. Jl.
20 1 -4 -10 -5
In Maple, the command eigenvalues or eigenvals can be 36 3 -3 -18 -16
J
32.
used to find the eigenvalues of a square matrix. The
commands eigenvectors or eigenvects can be used to 12 -2 -7 -2
find the ei genvectors of a square matrix.4 When us 21 -4 -11 -5
ing either of the latter two commands, Maple's output 21 -4 -10 -6
IB=P-1 AP.,
One place we have already encountered similar matri ces is when changing bases f�r
matrices of linear transformations. If T : v -+ v is a l i near transformation where V .18
a nonz�ro finite dimensio al vector space, we saw in Section 5.3 that if A is th ma�i
� � �
of T with respect to a basis a of V {so that A = [T]a) and if B is the matn_ x of Twit
respect to another basis f3 of v {so that B = [T]'\
�hen B = p-1 A p where Pis the
change of basis matrix from the basi s a to the ba:is an
/3. In fact, all similar matrices �
be obtained in this manner. To see why, let T : IR" mau on
-+ JR" be the matrix transfor
·
4 If any entries arc entered as decimal numbers, Maple will give
decimal approximations in its answer s· ol.h-
erwise
· MapIe w1·11 give
_ · ·
its answers in a symbo lic exact form, although these exact forms can be d1'fficult to
interpret if the roots of the characteristic polyn
omial are not nice.
5.5 Similar Matrices, Diagonalization, and Jordan Canonkal Form 279
0
and if the basis vectors in f3 are w1, w2, ••• , Wn, then
T(w;)=Aw;= d;w;
for each i. That is, each basis vector in {3 is an eigenvector of A. Let us record what we
have just observed as
THEOREM 5.18 Ann x n matrix A is diagonalizable if and only there is a basb for IR" consisting of
eigenvectors of A.
n
diagonalizable or. equivalently, if !R has a
How can we tell if a square matrix A is
see momentarily, the following lemma is one of
basis of eigenvectors of A? As we shall
the keys.
LEMMA 5.19 Suppose that r1, r2, ••. , rk are distinct eigenvalues of a square matrix A. If the vectors
vu, v12, ••., v 11, fonn a basis for E,,, the vector� v21, v22, .••, v211 form a basi� for
£,l' ... , the vectors Vk!, vk2, .••, Vk1, fonn a basis for E,., then the vectors
V11, v1 2, •••, v11,, V21, v22, .••, V2J2 , •• ·, Vkl, Vk2· · , , , Vki,
Thus c1 1 = 0, ... , c 1 1, = 0 since v11, ••• , v11 1 are linearly independent. Repeating this
procedure with the matrices
82 = (r1/ - A)(r3/ -A)··· (rkl - A), ..., Bk= (r, I - A)··· (rk-tl -A),
we obtain the remaining scalars c ji are all zero, completing the proof. •
Let us now see the impact of Lemma 5.19 on the diagonalizability ot an n x n matrix
A. Since the basis vectors from the distinct eigenspaces of A are linearly independent,
it follows that if the total number of vectors in these eigenspace base, is n, then these
vectors form a basis for JR" and hence A is diagonalizable. To put it ...1other way, if, in
the notation of Lemma 5.19,
(2)
THEOREM 5.20 Suppose A is an n x n matrix with distinct eigenvalues r1, ,2, •••, rk. Then A is diago
nalizable if and only if
-3
[ -2I
A=
2]
5.5 Similar Matrices, Diagonalization, and Jordan Canonical Form 281
Solutioll From the solutions to Examples I and 2 of the previous section, "'e have that the eigen
values of A are .l. = 4 and >.. = -1 and
dim(£4 ) + dim(L 1) = 1 + I = 2.
Consequently, A is diagonalizable. Together, the basis vectors
- 3/2
, otl [
[ :] ]
l
we gave as bases for the individual eigenspaces of A form a basis
__:z..
-)
[
-I l[ 3/2 0
for !R2• The matrix of the linear transformation T(X) = AX with respect to this basis is
which then is a diagonal matrix similar to A.5 A matrix P so that p- 1 AP is this diagonal
matrix is the change of basis matrix from the standard basis to our basis of eigenvectors,
l·
which is
-I 3/� •
[
p ==
I
EXAMPLE2 Determine whether the given matrix A is diagonalizable and, if it is, give a diagonal
matrix similar to A as well as a matrix P so that P I AP is thi� Jiagonal matrix.
(
A==[�=: �]
0 0 -I
vO u
Solution Let us go faster now. Look back at Example 3 in the previous section. From its solution,
we can see that A is diagonalizable since
dim(£2 ) + dim(E-1) == I + 2 = 3
5 The diagonal matrix is not unique. For instance, were we to interchange the order of the basis vecton here,
the diagonal matrix would become
I! is possible to show that the diagonal matrix is unique up to a pennutation (that is, a rearrangement) of the
diagonal entries.
282 Chapter 5 Linear Transformations and Eigenvalues and Eigenvectors
[ � -� � l
0 0 -I J
•
1/3 -1 J
1 0 .
0 1
EXAMPLE 3 Determine whether the given matrix A is diagonalizable and, if it is, give a diagonal
matrix similar to A as well as a matrix P so that p-l AP is this diagonal matrix.
From the solution to Example 4 in the previous section, we see that this matrix is not
diagonalizable since
Solution
Look back at the solution to Example 5 in the previous section. If we work over JR,
t�is matrix is not diagonalizable since it has no real eigenvalues (and hence no rea l
S0illtio11
_
eige�v�ctors either). But if we work over the complex numbers, we have success: The
matnx ts diagonalizable and it is similar to the diagonal matrix
•
P
=
[ l -i
5.5 Similar Matrices, Diagonalization, and Jordan Canonical Form 283
>.. 0 0 0
0 >.. 0 0
0 0 >.. 0 0
0 0 0 >..
0 0 0 0 >..
The Jordan canonical form of a square matrix is comprised of such Jordan blocks.
that
THEOREM 5.21 Suppose that A is an n x n matrix and suppose
m2 • • . - rt) mt
det(H - A)= (.l.. - r,r (>.. - r2) • (>.
1
[� ]
rix of the form
where each B; is an m; x m; mat
0 0
l;z
B; =
0 0 1;,
( 183S-1922).
mathematician Camille Jordan
6 Named for the the French rmuting the vectors in a basis f3 for R" giving
us
canonical f�s cotespond to �
7 These different Jordan the mat nx tran sfor mauo n T(X ) = AX .
form as the matnx [T).8 for
one Jordan canonical
284 Chapter 5 Linear Transformations and Eigenvalues and Eigenvectors
EXAMPLE 5 List the possible Jordan canonical forms of a 4 x 4 matrix A whose characteristic poly
nomial is
Solution For the first eigenvalue,).. = 4, there is only possibility for B1 since it must be a 1 x I
matrix:
B i = [4].
For the second eigenvalue, ).. = 2, the matrix B2 is a 3 x 3 matrix and there are several
possible ways we can form a 3 x 3 matrix consisting of basic Jordan nlocks associated
with ).. = 2. One way is for B2 to consist of an entire 3 x 3 basic Jordan block:
B,�u � !l
Another possibility is for B2 to consist of a 2 x 2 basic Jordan block ,md a 1 x I basic
Jordan block:
82 = [ � � : ].
0 0 2
The last possibility is for B2 to consist of three I x I basic Jordan blocks:
[ 2 0 0]
B2 = 0 2 0 .
0 0 2
iH� ][� �]
Putting this all together, there are then three possible Jordan canoniaJ forms for A:
[� •
0 0 0 0 0 0
2 I 2 2 0
0 2 0 2 0 2
0 0 0 0 0 0
It is �nt�resting to observe that while there are infinitely many 4 x 4 matrices with
char�cten�tic polynomial ().. - 4)().. - 2)3 all of these are similar to one of the three
matrices given at the end of our solution to Example 5!
The proof of Theorem 5.21 is beyond the scope of this book and is omitted. Further,
there are general methods for finding the Jordan canonical form of a gi vcn square m atrlx,
but th�se too are beyond the scope of this book. Let us notice howe ver, th at it is
' ·
sometimes possi'ble to see the Jordan canonical form of an n x n matrix A by knowing
the bases for the eigenspaces. We already have noted that sions
if the sum of the dimen
5.5 Similar Matrices, Diagonalization, and Jordan Canonical Form 285
of the eigenspaces is 11, the matrix is diagonalizable. The resulting diagonal matrix is the
Jordan canonical form of A, so we know the Jordan canonical form in this case. Thus
=n
in Examples 1, 2, and 4, we could say that w e found the Jordan canonical form!> of the
given matrices in these examples.
Another case in which we, can determine the Jordan canonical form of A is when
[ -� -�
the multiplicity of each eigenvalue is at most 2. To illustrate how. consider the matrix
A=
[ :].[� �]·
0� �0 5 0 �0 5
Because A is not diagonalizable, the first form is out and the second matrix is the Jordan
canonical form of A.
=n
While we ourselves will not study methods for finding Jordan canonical fonns,
software packages such as Maple employ these methods to find them. Maple also will
find a change of basis matrix P so that p- I A P is the Jordan canonical fonn of an II x 11
matrix A. To illustrate, let A be the matrix in the previous paragraph:
A=[-� -�
l form and
ed on a Maple worksheet, its Jordan canonica
Once the matrix has been enter
typing and entering
the matrix P can be found by
jordan(A.'P'):
canonical form of A as
U 1 :l
which gives us the Jordan
in the previous
basic Jordan blocks in a different order than
(Notice that Maple has the
to displ�y P. we type and enter
paragrnph.) To get Maple
print(P):
286 Chapters Linear Transformations and Eigenvalues and Eigenvectors
-3]
[ 1/2 -I -1/2
-1/2 3/2
Exercises 39 and 40 ask you to use Maple or another appropriate software package to
find both a Jordan canonical form of a square matrix A as well as a matrix P so that
p-J AP is this Jordan canonical form of A.
EXERCISES 5.5
1-18. Let A be the matrix given in the corresponding 34. Suppose that A is an invertible matrix and the ma
exercise of Section 5.4 (see page 277). Determine trix B is similar to A. Show that H is an invertible
if A is diagonalizable and, if it is, give a diagonal matrix and that B -1 is similar to A- 1•
matrix similar to A as well as a matrix P so that 35. Suppose that A is an n x 11 J1agonalizable ma
p-t AP is this diagonal matrix. trix with distinct eigenvalues r 1, r2, ... , rt, Fur
List the possible Jordan canonical forms for the matrices ther suppose that v 1, v2, .•.• v11 are eigenvectors of
with the given characteristic polynomials in Exercises A forming a basis for JR" arran£ed in such a way
19-24. that v 1, ••• , vm , are eigenvectors associated with
20. >..2 - 3>.. - l 0 r,, v111 ,+ 1, ••• , vm1 are eigenvectors associated with
r2, ..• , vm, ,+1, ..• , Vn are eigenvtctors associated
21.>..3
->.. 2 22.>..5-2>..4+>..3 with rk, Show that for each i, Vm,_ 1+ 1, ••• , v,,,, span
23.(A - 5)(>.. - 1) (>.. + 2)
2 3 E, and hence form a basis for E,,. (Hint: Let v
1
29. Exercise 11 30. Exercise 8 (not necessarily distinct) eigenvalues r1, r2, · · ·, '•
that fonn a basis for JR". Let v be a vector in JR"·
31. Show that similar matrices have the same character Express v as
istic polynomial.
32. Suppose that A, B, and Care square matrices of the U = C1V1 + C2t12 + • · • + CnVn,
same size. Show that for any positive integer k,
a) Show that A is similar to A.
b) Show tbat if Bis similar to A. then A is similar
Ak V = C1f k1 V1 + C2rJ:.2 V1 + · · · + Cn n"vn·
f
to B.
c) Show that if A is similar to B and B is similar 38. Suppose that A is square matrix that has �nly 0��
to C, then A is similar to C. eigenvalue>.. = r. Show that A is diagonaltzable 1
33. Show that if the square matrix B is similar to the and only if A = r I.
squ � �atrix A, then Bk is similar to Ak for any In Exercises 39 and40' use Maple or another appropriate
pos11tve integer k. software package to find a Jordan canonical form 0f A
5.6 Eigenvectors and Eigenvalues or Linear Transformations 287
I -I 2
-4 3 -1 3
-I
and a matrix P so that p-l AP is this Jorda� canonical 0
1 -4 -II
21
I 0
l
form of A.
43 4 -3 -22 -19
-4 0 3 -1 -2
r -1
40. A=
-I 2
0 -I 4 0 -2
22 -4 -12
-7
-I 2
39. A� 10 0 0 2 -2
-11 -5
-4 -I 0
22 -4
I T(v) = AV I
e of T. For eltamp lc, x + 3x + 2 is an
2
where >.. is a scalar: >.. is called an eigenvalu ation T : P2 - P2
eigenvector with associated eigenvalue>.. = -3
of the linear transform
by
T(ax 2 + bx +c) = (Sa -4c)x 2 + (12a + b - 12c)x + Sa - 7c
since
+ 2) = -3x2 - 9x - 6 = -3(x + 3x + 2).
2
T(x 2 + 3x
the
of eigenvectors a��ociated with >.. along with
If>.. is an eigenva lue of T, the set is the iden tity
linear transformation Al - T where I
zero vector is the kernel of the nspace of>..
= u on V. We will ca l l this subspace the eige
linear transformation /(u)
and denote it by V,, so that
Vi. = ker(>../ - T).
are matrix. then
transformation T(X) = AX where A is a squ
Notice that if Tis a mat rix the eigenvectors and eigenval ues
l ues of T are the same as
the eigenvectors and eigenva
of A. <;formation?
tors and eigenvalues if Tis not a matrix tran
How do we find the eigenvec relative to a basis of V.
es of T
The answer is to use matric
of T with associated
:r,
288 Chapter 5 Linear Transformations and Eigenvalues and Eigenvectors
Proof Suppose vis an eigenvector of T with associated eigenvalue A. Sine,; T(v) = .l..v and
[T(v)]a = A[v]a
. by Theorem 5.14, we have
[Av]a = A[vJc,.
Hence
A[V]a = A[v]a
giving us that [v] a is-an eigenvector of A with associated eigenvalue A.
Conversely, suppose the column vector
[ :: ]
a,,
is an eigenvector of A with associated eigenvalue A. Set
in which case
[v]a =
[ CIJ] at .
We then have
Cln
A.a,, J
5.6 Eigenvectors and Eigenvalues of Linear Transformations 289
Hence
[ V ]a = CJ [ ::: ] + • • · + Ck [ ::: ]
Gnl G11A
f(v) = ,l,.v and
= CJ[UJ la+···+ cdukla
from which i t follows that
then
[(lJ ] [CIJk]
or
= 0.
a21 a2k
CJ + • · • + Ck :
:
a.1 tlnk
•
Hence
CJ :::::0, c2 = 0. Ck= 0.
Following is an illustration of how we use Theorem 5.22 to find the eigenvalues and
eigenvectors of a linear transformation.
T : P2
EXAMPLE 1 Find the eigenvalues and bases for the eigenspaces of the linear transformation �
P2 by
T(ax 2 +bx+ c) = (Sa - 4c)x
2
+ (12a +b - 12c)x + 8a - 1c.
290 Chapter S Linear Transformations and Eigenvalues and Eigenvectors
Solution Letting a be the standard basis consisting of x2, x, I for P2, the matri>.. of T with respect
to a is
�]-[�
-4 0 4 : -1 : 0
[ -12 0 12 i 0 0]
I
-8 0 8 I 0 0 �
0 0 : 0
The soluti ons to (I - A)X =Oare
[JUJ
as a basis for E 1. These two vectors in IR.3 are the coordinate vectors of the polynomials
x, X2 + 1, which form a basis for V1 where v = p2
Finally, consider).. = -3.
-8 0 4 : 0 O -1 2
[ -12 -4 12 l O ] - [ Q -4 6
-8 0 4 O O 0I 0 I 0 ]
�
The solutions to (-3/ -A)X = o are
5.6 Eigenvectors and Eigenvalues of Linear Transfor mations 291
as a basis for £_3 . Thi s vector in !R3 is the coordina te vector of the poly nomial
x 2 3x
+ +
•
2 2 !,
which forms a basis for V_3 where again V = P2•
The theory of diagonalizability for square matrices extends to linear tra nsfom1ations
T : V --+ V. We say that T is diagonalizable if there is a basis f3 of V where each vector
in f3 is an eigenvector of T or, equiva lently, if V has a basis f3 s o that [ T J� is a diagonal
matrix. We can see from our solution to Example I that the linear transformation of this
example is diagon alizable. The notion of J ordan canonical form also exte nds to such
linear transformations T : V _. V by using the J ordan canonical form of a matrix of T
with respect to any basis of V.
EXERO:jES 5.6
Do the following for each linear transfonnation in Ex 6. D2 : V --+ V where V is the vector space of solu
ercises l·-6. tions to the differential equation y" + y = 0.
(a) Hnd the eigenvalues and bases for the 7. The determinant of a linear transformation
cigenspaces of the linear transformation. T : V --+ V, denoted det(T), is defined to be
(b) Detennine whether the linear transformation dct(T) = det([Tl:)
is diagonalizable.
(c) Find the Jordan canonical form of the linear where a is a basis for V. Show that this defini
transformation. tion of det(T) is independent of the choice of ba
sis of V; that is, if f3 is another basis for V, then
1. T : R3 --+ JR3 by det([TJ:)) = det([T]�). (This is described by say
ing that the definition of the determinant of a linear
7x -5z
transformation from a nonz.ero finite dimensional
15x+2y-15z ] vector space to itself is we/I-defined).
I Ox - 8z
8. Find the determinant of the linear transformation in
Exercise 2.
3 42
2. T : !it3 -+ JR3 by T [ : ] = [ \: � ;z ] · 9. Find the determinant of the linear transformation in
Exercise 3.
z 4x + y -5z
3. T: P1 -+ P1 by T(ax + b) = 2bx +a+ b. IO. Suppose that T : V --+ V is a diagonalizable linear
transformation. Show that if T has only one eigen
4. T : P2 -+ P2 by T(ax 2+bx+ c) == value>..= r, then T(v) = rv for all v in V.
ax 2 + (2b + 2c - 3a)x + 3c.
S. D : V -+ v where v is the vector space of solutions 11. You may have observed in every example and exer
to the differential equation y" - Y == 0. cise involving eigenvalues and eigenvectors in this
292 Chapter 5 Linear Transformations and Eigenvalues and Eigenvectors
chapter that ifA = r is an eigenvalue ofan x n matrix b) Show that (A - rl is a factor o: the charac
A, then dim(£,) is less than or equal to the multi teristic polynomial of [TJ�. This then gives us
plicity of the root r in the characteristic polynomial the desired result. Why is thi��
of A. This is always true. Prove it as follows:
12. Let T : P3 ---+ P3 be the linear lr.i !;formation given
a) Let v 1, v2, .•• , Vk be a basis for E,. Extend this
by
to a basis /J consisting of v 1, v2, ••. , vk,
T(ax 3 + bx 2 +ex+ d) = (3a + 6b- 7c + 7d)x3
n
VHJ, ... , v. for IR . Show that the matrix of
the linear transfonnation T(X) = AX with + (c -a - 3b)x2 + (26a + 27b - 31c + 29d)x
respect to /J has the form
+ 26a + 26b - 27c + 24d.
p [ rh
[T]p = Use Maple or another appropriate ,oftware package
O(n-k)xk
to aid in finding the eigenvalue� and bases for the
where B 12 is a k x (n - k) matrix and 822 is an eigenspaces of T and determining whether T is di
(11 - k) x (11 - k) matrix. agonalizable.
Systems of Differential
Equations
We begin this chapte r with an example that. upon first glance. may appear to be a problem
you already have encountered in your calculus classes.
Find the position of the object at time t if the initial position of the
object is the point (I, 1).
you.
ver, you will notice that tbi� problem is new to
Upon closer examination, howe
is
Realizing the velocity vector v
v=[�J. dy
dt
a� the �y�tcm of
the statement of the problem is the same
the vector equation given in
equations:
dx
- = 2x - 5v
dt
dy
� = X -2y.
dt
293
••
294 Chapter 6 Systems of Differential Equations
giving us an initial value problem. A solution to this initial value problem would, of
course, consist of functions of the form x = x(t), y = y(t) satisfying the system of
differential equations and the initial condjtions. Software packages such as Maple can
be used to graph phase portraits to an initial value problem such as the one we have here.
The phase portrait in Figure 6.1 for this initial value problem, which was obtained by
typing and entering
phaseportrait ( [D (x) ( t)= 2* x ( t) -5* y(t} ,
D(y) (t) =X (t) -2* y(t) I, [x (t), y (t)],
t=0 .. 7, [[x(O)=l,y(O)=l]], arrows = none:;
illustrates the path of the object.
-1.5
Figure 6.1
1:1e main purpose of this chapter is to develop techniques for solving systems of dif·
feren�ial equations and to consider some applications of them. The system of differential
equatmns
dx
dt = 2x -5y
dy
df = X -2y
ns.
we �ust consid�red is an example of a first order linear system of differential equatio
er
While there will �e a fe � places in this chapter where we will consider some hig�
nt'.a l
order systems 0� hnear differential equations and some nonlinear systems of dif fere
_ al
equat'.ons, we will focus most of our attention on first order systems of linea r diff erenu
der
equations. As you are about to see, there is a great deal of similarity betw een first or
6.1 The Theory or Systems of Linear Differential Equations 295
systems of linear differential equations and the linear differential equations we studied
in Chapter 4. Indeed, we begin this chapter with a section on the theory of these systems
that parallels the first section of Chapter 4.
as we
endent variable in our functions rather than t
(Note that we are using x as the indep x = 0 ... . , 8n (x ) = 0. the
chapter.) If g1( )
had in the example in the introduction of this omo gene ous. Thro ugho ut
e it is called nonh
system is called homogeneous; otherwis on an interv al (a , b). If we add
is continous
this chapter we assume each aii and each 8i
the initial conditions
y.(xo) = bn
tions
b , we call the system of linear differential equa
for fixed real numbers b 1 , 2 , • • • , n
an initial value problem. b xI
r differential equations in Equation ( l) is an II
A solution to the system of linea
column vector
Y
=
[:]
h ation of the system.
of x so that Y1. y2, ..• , Yn satisfy eac equ
where each y; is a function sfie s the initial conditions.
the initial value problem if it sati
Further, y is a solution to
That is,
y,(.xo)
b2
b
Y2(.xo)
Y(xo) = [ .. ] = [ .1 ] ·
. .
Yn(.xo) b.
296 Chapter 6 Systems of Differential Equations
j[
We can use matrices to write our system of linear differential equations as
: : : :�:� ] [ : ] + [ : : ; ] .
y; a11(x) adx)
y1 _ a21(x) a22(x)
[ . . .
.. .. ..
y� ani(x) a.2(x) ann(X) Yn
+
g,. (x)
Letting A(x) be then x n matrix [a;j(x)], Y be then x I vector [y; ], G�x) be then x I
vector [g; (x) ], and Y' be then x l vector [y;J, 1 our system then becom, s
I Y' = A(x)r. j
As mentioned in the introduction to this chapter, the theory of Hrsr order systems
of linear differential equations parallels the theory of linear differcm·�i equations we
learned in Chapter 4. Corresponding to Theorem 4.1, we have the folk>•,, 'lg uniqueness
and existence theorem.
+
THEOREM 6.1 If a;,j(x) and g;(x) are continuous on the interval (a, b) containing xv /or 1 ::S i :'.Sn
and I S j S n, then the initial value problem
Y' = A(x) Y
[] :b2n1
Recall from Theorem 4.2 and its proof that the solutions to an 11th order homogeneous
linear differential equation form ann-dimensional subspace of the vector �pace C"(a, b).
[
The solutions to a homogeneous system of n linear differential equations Y' = AY fonn
a subset of the set of all n x I column vector functions of the form
f1(x)
h(x) j
f,, (x)
l The denva
. t1veofa matrix M(x) = (f;j(X)] is the matrix
M' x) = Jim (M( x + tu) - M(x)) = [ I'
( (fij(X + liX) - /;j(x)) ]- [ ''·( )).
ax-o t,,,.x ��o t,,,x - ;,, x
Many properties of derivatives of funct .
ions extend to matn·ces. (See Exerc1se 23. )
6.1 The Theory of Systems of Linear Differential Equations 297
the�c
where each f; is in C 1 (a, b). We have an addition and �calar multiplication on
e just as we do with column
column functions (add and multiply them by scalars entrywis
the
vectors) making the set of such column vector functions into a vector space. Viewing
vector functions.
solutions to Y' = A Y as a subset of the vector space of these column
m 4.2 (see Exercise 18) to obtain
we can proceed along the lines of the proof of Theore
Theorem 6.2.
THEnREM 6.2 The solutions to a homogeneous system of n first order linear differential equations
Y' = A (x) Y form a vector space of dimension 11.
ly independent solutions
As we did in Chapter 4, we will call a set of 11 linear
first order linear differential equations a
Y1, Y2, ••. , Yn to a homogeneous system of 11
form a funda menta l set of solutions.
fundamental set of solutions. If Y1 , Y2 , ••• , Y,.
ons of the homo geneo us system. and the
they form a basis for the vector space of soluti entia l equations
of first order linear differ
general solution to the homogeneous system 11
is given by
l
If we let M be the matrix
[M = [ ft Y2 ·.. Y11 ]
ar differential
the homogeneous sy�tem of first order line
then the general solution to
equations can be written as
and that Yp is a solution to the nonhomogeneous system of first order I inear differential
equations
Then every solution to this nonhomogeneous system of first order linear differential
I I
equations has the form
The proof of Theorem 6.3 is similar to the proof of Theorem 4.3 and is left as
Exercise 19. As in Chapter 4, we call Y p a particular solution to the nonhomogeneous
system.
For our last part of the theory of systems of first order linear difkn ,1tial equations,
we discuss how the Wronskian is modified to the system setting. Given column vector
functions
Y11 (x) [ y!2(x) ]
Yi (x) =
2 (x) Y22(x)
Yi (x) = [ Y \ ], ,
:
Yn l (x) Yn2(x)
we define the Wronskian of Y 1 (x), Y2 (x), ..., Y.(x), denoted
w(Y 1 (x), Y2(x), ... , Y,,(x)),
to be the determinant
Ynn (x)
Let xo be a value in (a, b) for which w(Y1 (xo), Y2 (xo), .... Yn(xo)) i= 0. The system
Y11 (xo)c1 + ydxo)c2 + · .. + Yin(xo) c,, =0
Y21 (xo)c1 + )'22(xo)c2 + · · · + y:z,, (xo)Cn =0
THEOREM 6.5 If Y1, Y2 , ... , Y are s olutions to a homogeneous first order linear system Y' = A(x)Y
on an interval (a, b) such that w(Y1 (xo), Y2(xo), ... , Y,,(xo)) = 0 for some xo in (a. b),
11
Y' = AY
ly to solve such
constant. One case whe re it i s especial easy
where each entry of A is a r the sy
diagonal matrix. To illust ra te. conside
stem
a system i s when A is a
y: [ ] [
Yi
=
3
O -1
0
] [ I ]•
Y
Y2
on
Considering each equati
and
y; = -y2
or 4 we determin e t hat t he general
t her Chapter
of the system, using the method s of ei 3
quations are
solutions to these two e
300 Chapter 6 Systems of Differential Equations
and
y= [ : l ,
l 0
= [ ;;;: = c, [ :· ] +" [ , • ] = [ ,:, ,�.
is a solution to this system of linear differential equations. If we let
J [ :: l
Yi =
[ e 3. r
0
J
and
Yi and Y2 form a fundamental set of solutions to this system of linear differential equa
tions. The corresponding matrix of fundamental solutions is
Y2 ]=
[ e3.r
O
In general, if A is the diagonal matrix
di 0 0 0
0 d2 0 0
A=
0 0 0 dn
the matrix
ed x
1
0 0 0
0 ed2.r 0 0
M=
0 0 0 ed•.r
i � a matrix of fundamental solutions to Y' = AY. We let you verify this fact in Exer
cise 17.
le. We
. In the next section we will consider the case in which A is diagonalizab for
'"'.111 ma�e . use of �ur ability to solve diagonal systems along with the tech ue
niq
diagonahzmg matrices we learned in Chapter 5 to solve Y'= A y in this case .
6.1 The Theory of Systems of Linear Differential E<1uations 30 I
EXERCJ:;;::,s 6.1
I [2 J
is a solution of
4 -J
y = Y.
II. E><,n,;se 7; Y(0) =
-I ]
�
[ ]
J J.
is a sol:ltion of
-
Y' = [ : � y+[ �
In Exercises 13-16, determine the general solution to the
nonhomogeneous equation Y' = AY + G(x) by indi
In Exercic.t.s 3 and 4, show that the given columns vidually solving each equation of the system. Compare
of functiO,·'> are linearly independent on the interval your solution in each of the se exercise� with your solu
J
tion to each of Exercises 5-8, respectively, and ob�rvc
J[
(-oo. oo;.
how these solutions illu�trate Theorem 6.3.
[ e2' cos 3x
J J
e2x sin 3x
3.
= [ � -�
-e2.x sin 3x ' e 2" cos 3x
13. Y' Y+[ ;
e-:•.x 0 0
4.
[
0
0
,
] [
3 co� 5x
-3 srn 5x ]
, sin 5x
[ cos 5x ]
to
14. Y' = [ �
�
J Y
+[
s
�:
r
J
In Exercises 5-8, determine the general solution
o] [
a mat rix
Y' = AY for the give n matrix A. Also give -1
of fundamental solutions.
5 = I
.A [
0 -2
O
J 6. A= [ � � J 15. Y' =
[
0
-
�
0
3i
e ]
- -2 0 0 0
7. b J6. Y' = � 2 0
y+
I - 2x 2
0
� � [
; ] 0 () 5 sin 3x
[ 3
0 0
17. Show that a matrix of fundamental solutions to
-2 0
8.h [-� di O O 0
0 2
0 d2 0 0
0 0
initial y' = y
In Exercises 9-12, determine the s olution to the
value problem y' = AY, y (0) = Yo for t he sy st�n:i .of
and he giv en rniu al
the indicated exercise of this section t
0 0 () d.
condition Y(O).
302 Chapter 6 Systems of Differential Equations
-x
22. Let V be the set of all column functio ns For
-e:
A(x) = [ 2e-x4x J. B(x) =
[ ea
2x e3x ] '
C=[ 42 -1 1
where .Yt, ... , Yn are in F(a, b).
J.
a) Prove that V is a vector space. find the following.
J
b) Let W be the set of all column functions 24. (A+ B)' 25. (3A)'
I� Section 6.1 we saw how to solve the homogeneous system of linear differential equa
tions
Y' = AY
for a diagonal matrix A. In Chapter 5 we saw
that some matrices are simi lar to a diagonal
r�al'.ix. The?rem 6.7 sho ws how sol utions t
to linear systems with constant coefficien
similar matnces are rel ated. We will use s o l ve
this theorem later in this secti on to
Y' "".' AY when A is similar to a diagonal
matrix. In the next section we wil l see ho w to
use it lo solve such systems when A is not
similar to a diagonal matrix .
6.2 Homogeneous Systems with Constant Coefficients: The Diagonalizable Case 303
THEOREM 6.7 Suppose that A and B are similar n x n matrice� with B = p- 1 AP where P is an
invertible n x n matrix. If Z is a solution of Y' = BY. then P 2 is a 1,olution of
Y ' = AY. Moreover, if 2 1, Z2,... , Z,, forms a fundamental set o f solutions to Y' = 8 Y,
then P Z1,P Z2, ..., P 211 fom1s a fundamental set of solution-. to Y' = A Y.
of Theorem 6. 7 is to say if
Another way to state the last part
[us ::;; [ Z1 Z2 .. · Zn ]l
Y'=AY,
D = p- 1 AP
di 0 0 0
0 d2 0 0
D=
l 1
0 0 0 d,,
where each of d1 , di, ... , d11 is a real number. (We will consider the c.ise where all or
some of d 1 , d2 , ... , d11 are imaginary numbers later in this section. l We know from the
last section that
c1ed,x
c2 d2x
�
c,,ed.x
l1
is the general solution to the system
Y' = DY.
It follows from Theorem 6.7 that the general solution to Y' = AY is then
c 1ed,x
p Cz d2x
�
c,,ed.x
A= [ I -3].
-2 2
6.2 Homogeneous Systems with Constant Coefficients: The Dia�onaliwhlc Case 305
Solution The matrix A is the matrix of Examples I and 2 of Section 5. 4 and Example I of Section
5.5. In these examples we found that A has eigenvalues A = 4 aml A = - I (Example I
of Section 5.4), EJ, has
[] �
as a basis vector (Example 2 of Section 5.4). and A is similar to the diagonal matrix
D=P- 1 AP=[
4
O]
0 -1
where
that
(Example I of Section 5.5). We know
u =� =; ] n
an initial value problem.
In the next example we solve
m
EXAMPLE 2 Solve the initial value proble
[
y' = Y. Y(OJ =
A= [� -2
=� =;]
that the eigenvalues of
I
Systems of Differential Equations
= : - J
306 Chapter 6
[ � H- :l ·· d [
are bases for the eigenspaces £_ 1, Eo , and E2, respectively. Lettin
g
p -[ � _: =: ]
Hl
gives us
1-
D = P-1AP =
�
e2x
-1 + e2x ]
•
- e2x
l
Up to this point, all of our matrices A h ave had real eigenvalues. If A has i maginar
y
eigenvalues, we proceed in a manner similar to the m ethod we used in Section 4.2 when
· · · nt
the c h aractenstic equation of a homogeneous lmear differenti al equati· on with consta
· at ·r
coefficients had imaginary roots. Correspond ing ·
· to Theorem 4.9, which told us th .�
w(x) = u(x) + iv(x) is a complex-valued solution to a homogeneous linear differenll
equation, then u(x) and v(x) are real-valued solutions, we have the following theorem.
THEOREM 6.8 If U(x) + iV(x) is a solution to Y' = A(x)Y, then U(x) and V(x) are solutions 10
Y' = A(x)Y.
The proof of Theorem 6.8 is straightforward and is left as an exercise (Exercise 29).
6.2 Homogeneous Systems with Constant Coefficients: The Diagonalizable Case 307
along with Theorem 4.9 to obtain the solutions e0 cos bx and e0x sin bx to a constant
-'
A=
[
Solution Here A is the matrix in Example 5 of Section 5.4 and Example 4 of Section 5.5. The
eigenvalues of A are ).. = 1 + i and ).. = I - i. The vector
[]
forms a basis for E 1 +;, and the vector
nal
Section 5.4). The matrix A is similar to the diago
forms a basis for Ei-i (Example 5 of
[ 1+i
matrix
1
D=P- AP= O
= [ -: ]
where
i] [
solutions to Y' = AY .) By the first part of Theorem 6.7,
- ex cosx+ iex sinx -ex sinx +iex cosx
PZ = [ ] =[ ]
1 0 ex cosx + :ex sinx
=[
is a (complex.-valued) solution to Y' = AY. By Theorem 6.8,
are real-valued solutions to Y' = AY. Calculating their Wronskian (11} it), we find these
solutions are linearly independent. Hence they form a fundamenta · �:�l of solutions to
Y' = AY. The general solution to Y' = AY is then
C1 [
-ex sin x
e COS X ] + C2 [
e cos x
ex sin X
] =[
-c1 ex sin x + cze: l:OS x
C1 ex COS X + C]l:"' S'l'U:
] •
Notice that we did not multiply the matrix P (whose columns c,,nsist of the eigen·
vectors) times the general solution to the diagonal system in Example 3 as we did in
the case of only real eigenvalues in Examples I and 2. Instead. we multiplied P times
one of the. complex-valued solutions to the diagonal system from which we obtained
two real-valued solutions. This process may be continued if we ha,·c ·nore eigenvalues
some of which are imaginary. To do so, multiply P times each solution to the diagonal
system. As we do this for imaginary eigenvalues, however, we can omit the conjugate
eigenvalues since each imaginary eigenvalue produces two real-valued solutions. The
final example of this section illustrates this procedure.
: �].
[ 0 -1 -I
0 0 I
0 -1 0 : 0
UJ
The solutions to this system have the form
UJ
from which we see
forms a basis for E 1• When >.. = I + i, the augmented matrix for the homo
geneous system (H - A)X = 0 is
[ -:
-1 : 0]
' 0 .
�0 -1 : 0
us to
Solving this system (try it) leads
Thus if we set
is a basis vector for E 1-i.
� 1�;
[
P=
0 0]
D = P- AP = [ � 1
1 +j 0
0 I -i
310 Chapter 6 Systems of Differential Equations
[, l
to Y' = DY. This gives us the complex-valued solution
] [1 1-i l+i][ 0
cosx: ie' sin, = � i i ex cosx iex sinx
l
P
� :
EXERCISES 6.2
V::
[
28. An object is moving in 3-space with velocity vector
2x+y
3x + 4y J
5x - 6y + 3z
16. fae,dso 8; Y(O) - [ : ]
_ Find the position of the object at time t if the initial
position of the object is the point (5, 3, 4).
We know from Chapter 5 that not all matrices are diagonalizable. r onsequently, we
need to develop additional methods in order to be able to solve all ho ,v1.1eneous constant
coefficient linear systems. In preparation for this, we first consider, :1·-tems of the fonn
Y' = AY where A is an upper triangular matrix. These are easily solve··i -:,y backtracking
through the system individuaJly solving each equation in the syste:i . The following
example illustrates this.
-1]
EXAMPLE 1 Determine the general solution to the following system of differentid ,' 1uations
[ 3 1
Y' = 0 2 Y = AY.
�
0 0
Solution W riting our system with individual equations, our system has the fon,1:
y; = 3yi + Y2 - Y3
Y� = 2y2 + Y3
Y3 = 2y3.
Using the techniques of either Section 3.5 or Chapter 4, the last equation has general
solution
=[ =[
Therefore, the general solution is given by
�2::
] [ :J
-e2:t. -x e2x
]
3 2:t.
:::
x
e e
e:
c1 :
2 3
•
YH xe2
C3e2x 0 0 eix
l
Note that a matrix of fundamental solutions is
M= [
�
, �
: •
-:;
Now that we know how to solve linear systems Y' = AY where A is an upper
triangular matr ix, let us consider solving Y' = AY when A is not uppe r triangular. If we
could find an upper triangular matrix B simi lar to A and an invertible matrix P r,o that
B = p- 1 AP, we could find the general solution to Y' = BY and then u!>c Theorem 6.7
to obtain the general solution of Y' = A Y. How could we find such a matrix B'! One
way is to use the Jordan canonical form of A (which is an upper t riangular matrix) for
B. This is the approach we take in the following exampl e.
whe re we �aw
is the matrix of Examp le 3 of Section 5.5
Solution The matrix A in this example to find a Jordan
At the end of Section 5.5 we used Mapl e
that A is not diagonalizable.
canonical form of A to be
=� -�:].
matrix P so that p-· AP = B to be
1
and found a change of basis
1/2
p= [ 1/2
-1/2 3/2
18. Let
l
we find its general sol ution to be
sx eige
c ,e
:;� =� -�;2 ] [
Z 3 xe3x
= c2e x + c3
[ C3 e
.r 3
It follows from Theorem 6.7 that the general solution to Y' = AY is then
c
�
c2e 3x :3xe 3x ]
- J /2 3/2 C3 e3'
l
! c1 esx - 2cie3x -"2c3 xe3x - 3c3 e3x
! c, esx - c2e3.x - C3 Xe3x - !c3 e3x •
-! c1 esx + cie 3x + c3xe3x + fc3e3x
Using the Jordan canonical form is just one way to solve these systems of linear
differentfa l equations. We wil l see another way in Section 9.3.
EXERCISES 6.3
[
18. Let A be a matrix. Determi ne the conditions on the
is a solution to Y ' = AY. then limx-"- y;(x) = 0
J
eigenvalues of A so t hat if for i = I. .. . , n.
(x)
y,
Y2t)
Y(x) =
Yn(X)
is a particular solution to
Y' = A(x)Y + G(x).
equation
Rearranging leads to the
(M' - A(x)M)V + MV' = G(x). (l)
Notice that if
then
A(x)Y" ]
M' == [ r: Y� ... Y� ] == [ A(x)Y1
A(x)Y2
= A(x)[ Y, Y2 Y. ] == A(x)M
316 Chapter 6 Systems of Differential Equations
so that
M' - A(x)M = 0.
MV' = G(x).
Since det(M) is the Wronskian of a fundamental set of solutions Y1, l'�, ... , Y11 to Y' =
A(x)Y on an interval (a, b), det(M)-/=- 0 foreachx in (a, b) by Corollary 6.6 and hence
Mis invertible for each x in (a, b). We can then solve for V' by multiplying by M-1 on
the left, obtaining
We can now find V by integrating each of the entries in the column ,.:ctor
M-1G(x),
which we will indicate by writing
V= f M- 1 G(x)dx.
Yp = MV = M f M- 1 G(x)dx.
Let us summarize what we have just done with the following theorem.
THEOREM 6.9 Suppose that the entries of A(x) and G(x) are continuous on an interval (a, b). Further
suppose that Yi, Y2, ... , Y,. form a fundamental set of solutions of the homogeneous
system of linear differential equations
Y' = A(x)Y
on (a, b). If Mis the matrix of fundamental solutions
EX ·, \1PLE 1 Determine the general solution to the following system of linear differential equation�
Solution We leave it to you to show that the general solution to the corresponding homogeneou�
equation is
2c1e2.r + c2e3.r .
]
c 1 e 2.r + c2e1'
M= [
2e2.< e3.r
.
2.re e3.r ]
We have
e3.r
2e2.r e3.<
�]
2:··
[ e2.r e3 .r 0 �
]- [ eJ.r -1
-+ [
2:.r 0
e3
.r
�- -� l
which leads to
ral is then
The product inside the integ
e-2.r -e-2.r (2 - .x)e -2.x 1
M-IG(x) = [ -e-3.r 2e-3x ][�]=[ us
(-2 + 2.x)e-
3.r
We thus have
2 ][ (-�+f)e-2.r ] = [-*+51·
r, = M j w'G(x)dx = [ :: �: (:!9 - �x)e-3" _ .!!
36
- :!.
6
3
318 Chapter 6 Systems of Differential Equations
EXERCISES 6.4
In Exercises 1-10, determine the general solution to In Exercises 11-14, determine the soluLon to the initial
Y' = AY + G(x)for the given G(x) where Y' = AY value problem Y' = AY+G(x), Y (0) = Yo for thegiven
is the homogeneous system in the indicated exercise of column vector Y(O) and the system Y' = A.(x)Y+G(x)
J
Section 6.2 (see page31 l)or6.3 (see page 314). in the indicated exercise of thi� section.
1. Exercise 1 ofSection 6.2; G ( x. )= [ �
11. Exercise I; Yo = [ � ]
J ]
1-x
0
3
4. Exercise IO ofSection6.2; G(x) = [ ; 13. EwciseS; Yo - [
-1
1
x�t
]
5. Emcise 5 ofS,ctioo6.2;G(x)-[ ] 0
14. famise 6; Yo - [
-1
6. E,e,ds, 12 ofSccUoo 6.2;G(x) - [ ,�� ] 15. An object is moving with velocity vector
2x+y-t
=[ ].
7. Exercise I ofSection6.3;G(.x) = [ � J 2X+3y+ t
V
8. Exercise 2 ofSection6.3; G(x) = [ e: J Find the position of the object at time t if the initi al
position of the object is the point ( 1,1).
16. An object is moving with velocity vector
9. fae,cise 5 ofSecUoo6.3;G(x)- [ ,�� ] 2x+y+2
V = [ 3x +4y - t ] ,
1 5x - 6y + 3z
10. Emdse6ofSectioo6.3;G(x)-[ x�;� ] Find the position of the object at time I if the initial
position of the object is the point (5, 3, 4).
6.5 Converting Differential Equations to First Order Systems 319
EXAMPLE 1 Convert the following linear differential equation to a system of linear equations.
y" + 3y' + 2y = 0
Solution W e let v, = y and v2 = y'. Using the differential equation. this gives us
v; = y' = v2
v; = y " = -3y' - 2y = -2u1 - 3112.
[ 1 l � [ _: l [ : l ·
which can be rewritten as the sys tem
-3
the
The characteris tic equation of y" + 3y'
+ 2y = 0 is >.. + 3>- + 2 = 0, which is also
2
[ -� _; l tution, Vt = y and
equations obtained from the substi
of the system of differential
V2 = y'.
of linear equation:..
=
ar differential equation to a system
EXAMPLE 2 Convert the following line
y"' + 4y" - y' - 4y 0
le I, we �ee that
v3 = y". Proceeding as in Examp
=
and
Solution We let Vt= y, v2 = y',
v; y' = u2
v; = y" = 1/3
v3 = y"' = -4y" + y' + 4y = 4v 1 + u2 - 4u3.
as the sy stem
which can be rewrit ten
[ �l ] = [ � �
Ii,3 4 -4
� ] [ �: ] .
1/3
2
is >..3 + 4). - >.. - 4::: 0, \\hich
•.•..
=0
· · equation ofv"' + 4y" - y' - 4y
charactens11c
The
,,.,
320 Chapter 6 Systems of Differential Equations
.
of the system of differential equations formed from the substitution� t11 = y, v2 = y',
and v3 = y".
In Exercises I and 2 we ask you to solve the systems obtained in l�.>..-irnples I and 2,
respectively. Notice that the first entries of the general solutions lo the�� systems will be
v 1 and, since v 1 = y, they will give us the general solutions to the original differential
���-
Examples I and 2 lead to the following theorem. We leave the pro(){ 0f this theorem
for Exercise I 0.
l
0 1 0 0
[
0 y
0 0
Q(x) = y'
and V =
0 0
-� -�
ytn-1)
_ q._,(x)
q.(x) q.(x) q,(x)
Furthermore, if qo, q1, ... , q,, are constant functions, then the characteristic equation
of the nth order constant coefficient homogeneous linear differential equation and the
characteristic equation of the matrix Q(x) are the same.. The general solution to the nth
order homogeneous linear differential equation is the first entry of the general solution
of the system V' == Q(x) V.
Not only can we solve homogeneous linear differential equations by converting
the_m �o systems of first order linear homogeneous differential equations, but we can �s
e
[
� = �
]
[
: ]
[ +[
si� X
: ] ]•
was the case in the homogeneous
We will let you solve this system in Exercise 3. As
system will be the general <,Olution
setting, the first entry of the general solution to this •
to the original differential equation.
rential equations into first order linear
Our technique for converting single linear diffe
r systems of linear differential equations
systems can also be used t o convert higher orde
trates. Indeed, the fact that we ma) do
into first order ones as the next example illus
of studying fir�l order linear systems when
these conversions gives us the sufficiency
al equations in general.
considering systems of linear differenti
rential equation�:
m into a first order system of linear diffe
EX.\�. IPLE 4 Coovert the following syste
y;' = )' 1 + )'2
2
Y2 = )'1 + y; + Y2 + Y ·
Solution Letting
V1 = YI
vi= y; = v;
V3 = Y2
system:
we obtain the first order
v; = V2
v2 = V1 + V3
v3 = V�
V4 = V1 + V2 + V3 + V4.
:[ !]=[! i ! �][::].
ystem is:
In matrix form, our s
J
the solution
I I �
EXERCISES 6.5
In Exercises J 3,
- detem1ine the general solution to the 7. Section 4.2, Exercise 13
system obtained in the indicated example of this section. 8. Section 4.3, Exercise I
Use your general solution to the system to find the gen 9. Section 4.3, Exercise 11
eral solution to the original differential equation in the
example. 10. Prove Theorem 6.10
t. Example I 2. Example 2 3. Example 3 11. Reformulate Theorem 6.10 for the nonhomogeneous
case and prove this result.
In Exercises 4-9, do the indicated exercise of the section
12. Find the general solution of the sy,tem in Exam
from Chapter 4 (see pages 20 I and 211) by converting
ple 4.
to a first order system of linear differential equations.
13. Find the general solution of the system:
4. Section 4.2, Exercise 1
5. Section 4.2, Exercise 5 y;' = 2y; + Y1
6. Section 4.2, Exercise 9 y� = 6y; + 3y2,
There are many applications where systems of linear differential equations arise besides
the one we saw in the introduction of this chapter involving a given velocity vector. In this
section we consider a few of these. Our first example is a mixing problem. Recall that
we considered mixing problems involving a single tank in Section 3.6. Mixing problems
with more than one tank lead to systems of first order linear differential equations.
EXAMPLE I Consider two tanks each with volume l 00 gallons connected together by two pipes. The
first tank initially contains a well-mixed solution of 5 lb salt in 50 gal water. The second
tank initially contains 100 gal salt-free water. A pipe from tank l to tank 2 allows the
solution in tank I to enter tank 2 at a rate of 5 gal/min. A second pipe from tank 2 to
tank I allows the solution from tank 2 to enter tank 1 at a rate of 5 gal/min. (See Figure
6.2.) Assume that the salt mixture in each tank is well-stirred. How much salt is in each
tank after 5 min? Graph the amount of salt in each tank for the first 25 min.
r,
--.
-
5 gal/min /
-
--.
T2
'-
gal/min '- �
Figure 6.2
Letting
[ :: ]
Q =
we have
-I I 10 I 2
/ 0
[ ] = AQ.
Q' = 1/10 -12 / 0 Q
. ng
Eo and E-3/20, respectively U�i
-[ 1 -I] •
are bases for the e igenspaces
P- I
2
l
on is
we see the general soluti
Q =[
p ,,:� •• ] = [ : -: ] [ ,,:� •• ]
ive us
= [ ;;, �':::-·�.
When t = 5 min,
5 10
3 + 3 e-• � 3.24 lb
J
q 1 (5) =
IO IO 3
q2 (5) = - e-• � 1.76 lb.
3 3
Graphs of the two salt amounts obtained by letting Maple graph the unctions q1 and
q2 appear in Figure 6.3. Notice that q 1 and qz approach 5/3 and 10/'.:<, respectively, as
t--+ 00. •
0 2 4 6 8 I O 12 14 16 18 20 22 24
Figure 6.3
For our next application involving a system of linear differential equations, we con
sider a problem involving a circuit. In Section 4.5 we presented a single loop closed
circuit and found a differential equation modeling the circuit using Kirchoff's law. Sup
pose we have a double loop closed circuit as in Figure 6.4. In this setting Kirchoff's law
leads to the following system of differential equations for the current'- i 1 and i2 where
the derivatives are with respect tot:
Lii + R(i1 - i2) = E(t)
R(Zz-11
•I •I) 1.
+ l2 = 0.
C
E C
Figure 6.4
6.6 Applications Involving Systems of Linear Differential Equations 325
L
O ] [ �'.
[ _R R 12
R ][ �
] = [ -R0 -1/C 12
1
] + [ E(t) ] .
0
on on the left by
Solving this system for;; and i2 by multiplying each side of this equati
1/L
1/L
[]i;
i2
-R/L i
1
= [ -R/L R/L-1/RC ] [ i2
R/L
+
E(t)/L
] [ E(t)/L ]
= 2. L = 2,
the configuration of Figure 6.4 if R
EXd,1PLE 2 Determine i 1 and i2 for a circuit in
e is no initial current.
C = l/9, and E(t) = 2sint and ther
[i(] sin
R
+[ r
i2
[-I
== -1 -7/2 i2 sin t ] ·
Letting
i1
y:::: [ .
12
] and G = [ sin t
.
smr
]
we have
-1 1 sin t
= ] Y + [ smt
. ] = AY + G.
[ -1 -7/2
Y'
s of A.
values and bases for the eigenspace
again leav e it to you to find the eigen -3/2 and the vectors
We are -3 and
the eigenvalues of A
Doing so you will find
and
[ _:] [ -� ]
ly. Solving the hom ogeneous
o
e1· enspaces E-3 and E-3/2• respective
1or
are b ases � the o
326 Chapter 6 Systems of Differential Equations
system, we take P to be
p = [ -2
I -2]
1
where
M
2 -31
=[ -e
2c2 - 130 0
Y(O) = §}_ ] = [
].
[ Ct
-
-2c1 +c2 + i 0
Solving for c1 and c2 we have c1 = -1 /1O and c2 = -4/13 and hence
. 1 8 ) 67 81 .
11(t) = --e- 31 + -e-1' - -cost+ - sm t
10 13 130 130
. 1 31 4 3 t 7 9
12 (t ) = -e- - -e-,. + - cost+ - sin t.
5 13 65 65
Graphs of the currents appear in Figure 6.5. So that we can see which curve is whi�h .
we individually graphed i1 and i 2 in Figures 6.S(a) and (b), respectively, before graphin
:
them together i n Figure 6.S(c).
6.6 Applications Involving Systems of Linear Differential Equations 327
0 20
2 -0.04
-0.2
-0.08
-0.4 -0.12
-0.6 -0.16
-0.8
(b)
(a)
-0.2
-0.4
-0.6
-0.8
(c)
Figure 6.5
vector
EXAMPLE 3 A force given by th e
[ 36x ]
F= 12x+l6J
zero vector.
328 Chapter 6 Systems of Differential Equations
Solution Since
F =ma or a= -
[
m
where a is the acceleration vector and m is the mass of the object, and f'ince
d2x
dt2
a= ]
d2y
dt2
we have the second order system:
2x
d
-=9
dt 2 x
d2y
dt2 =3x+4y
This gives us the first order system
v; = v2
v; = 9v1
v; = V4
v� = 3v1 + 4v3
where v, = x, v2 = dx/dt, v3 = y, and v4 = dy/dt. The eigenvalues nfthe matrix
A=[l; � �]
for the system of differential equations are
A= 2, -2, 3, -3
and bases for the eigenspaces £2, E_2, £3 , £_3, each of which are one-di mensional
l[ l[ l l
spaces, are
order system is
[}]-[)
5 5
-C3 - -C4 = 10
9 . 9
5 5
-C3 + -C4 = 0
9 9
I I I I
-CJ - -c2 + -C3 - -C4 =8
2 2 3 . 3
CJ + C2 + C3 + C4 = 0
Y = V3 = e + e- 1 + 3e + 3e-
21 2 31 31•
8.0030
8.0025
8.0020
8.0015
8.0010
8.0005
8.0000=----'------'---� - --�-...
10 10.001 10.002 10.003 10.004
(a) (b)
Figure 6.6
where k is the spring constant, f is the frictional constant, and h(c: is the external
force. The reasoning we used can be extended to mass-spring systems such as the one
illustrated in Figure 6.7 involving two masses, m I and m2, and two �prings. Here it can
be argued that the motion of the two masses is governed by the system o f second order
linear differential equations
m,x;' = -k,x1 - fi x;+ k2 (x2 - xi)+ h 1 (t)
m2x{ = -k2(x2 - x1) - fix!z + h2(t)
where k1 and k2 are the associated spring constants, J, and Ji are the associated frictional
constants, and h 1 (t) and h2 (t) are the associated external forces. Converting this system
to a system of first order linear differential equations by letting v 1 = x i , v1 == x;,
Figure 6.7
6.6 Applications Involving Systems of Linear DitTercntial Equation,; 331
V3 = U4
/2 I
= -V1 k2
k2 - -U3 -V4 + -h2(t).
1112
U1 -
4
m2 m 2 m2
In matrix form, this system i s
[-�
0
_il. .!l. � )
'=
m1 "11 m1
] V + [ �' (f ] ·
V 0 0 ()
'. ;;;-;h2(t)
s
"12
0 _.!l,
m2 -/:;
quickl y
some mass-spring systems. You w ill
In Exercises 11-18, we will give you Our advice
long and difficult to solve by hand.
discover t hat these problems become r systems in
Maple to solve t he resulting fir�t orde
is to use a software package such as
these problems.
EXERCISES 6.6
min. The
Each into room I allows air 10 enter at r1.1 gal/
l. Tank 1 and tank 2 are arranged as i n Figure 6.2. l air ven t from room I to room 2 allow , air to enter
l quid ferti r0<im 2
has volume 500 gal and is filled wi th a i
onia room 2 at r1.2 gal/min. The a i r vent from
s l O lb amm r •1 gal/min.
izer. The solu tion in tank I contain to room I allow s air to ente r room I at
ammonia. leave rcxim 2
2
and the solution in tank 2 contains l lb The air vent out of room 2 allow, air to
gal/min. De monoxide.
Assume the flow rate in t he pipes is 25 at r2.2 gal/mi n. The
air cont ain!, carb on
ank at any
termine t he amount of ammonia in each t ns that
a) Set up a system of dif erential equatio each
f
tank l to initially
2 gal/min, solution flows in tbe pipe from in the volume 1000 cubic met ers. Ass ume
ion flows on mon oxid e and the air
tank 2 at a rate of 6 g al/min, and solut that room I has no carb
pipe from tank 2 to tank I at a rate of 4 gal/min. De in room 2 is 10% carbon monox
ide.
tank under these ine the system of differ
termine the amount of salt in each 5. Using Kirchoff"s law, determ
currenb for the cir
condi t ions. ential equations describing the
arat ed by a system if L = I. R =
4· Two perfectly insulated rooms are sep
air vents. cuit in Figure 6.8. Sol ve thi s
d by 2, C = 1/16. i i (O) = I. i2(0 ) = 0.
hallway. The two rooms are connecte The air vent
ve nt.
Each room also has an outside air
332 Chapter 6 Systems of Differential Equations
�
R E C
ht
L
Figure 6.11
co
i (f (x. y, z) dx + g(x. y. z) dy + h(.x, y, z)
dz)
a, 1x + a,2Y J
F=[
+ g(x(t), y(t), z(t))y'(t)
a21x + a22Y + h(x(t), y(t), :;(t))z'(t)) dt.
g to
is A� in Exercise 21. the line integ ral corre�pondin
a force vector
--a.!2/ m)-(a11/ m)().. -au/ m)-a12a2i/ m ·
/(x,y,z)
2
2
)..2(),2
F = [ g(x. y. Z) J
the c. o ·
20. Show th;• t the characteristic polynomial of h(x. y, ;:)
of linear
efficient matrix of the first order system
mas s-spring fo rce along th.:
differential equations associated with a in 3-space is the work done by the
tion al or external wo k done by the force in Ex.:r-
system as in Figure 6.7 with no fric curve C. Find the
forces has four im aginary r oots. cise 10 from t = I tot = 2.
r
) dx +
21. The line integral of a tlifferential form f (x, Y d
y -plan e, denote the large�t incrca,c in _a
g(x, y) d y over a curve C in the x 23. Recall from calculus that s occur!> in the tl1·
fc U (x, y) d.x + g(.x, y) dy),
is function of two or th ree variable
ient of the func tion.2 Thu, if
rection of the grad
Y, z), it
• ,
variables w = f (x.
- •
,
z)j :)k.
334 Chapter 6 Systems of Differential Equations
x(t), y = y(t) solve the init i al value problem 24. The largest decrease in a function of two or three
vari ables occurs i n the direction of the negative of
x' = fx (x, y) the gradient of the function. Suppose the tempera
y' = fy (x, y) ture Tat a point (x, y, z) in space is
x(O) = Xo, y(O) = YO· T = f(x, y, z) = x2 +4y + 3yz.
2
S uppose the elevation at a point is If a bug is initially at the point (0, 0, 1 ), find the curve
2
J(x,y) =x +4xy + y 2 along whi ch the bug must fly for the t emperature to
decrease most rapidly.
and a hiker is at the point with x = 10 and y = 15.
Find the cu rve the hi ker follows in ascend ing most
rapi dly.
dx
- = F(x,y)
dt
dy
- = G(x,y)
dt
or
x'=F(x,y)
(1)
y' = G(x,y)
and
y' = 2y -xy.
We will refer to these as simply 2 x 2 systems throughout this section. Two places
where these systems arise are in modeling the interaction of two competing species and
in modeling the motion of a pendulum, as w i ll be seen later in this section.
The theory of Taylor polynomials and series for functions of one variable that you
studied in calculus can be extended to functions of several variables. A full-scale d evel
opment of the topic is left for an advanced calculus course. In this section we are goin g
to make use of the fact that (under appropriate con ditions) a function f in two variables
x and Y has a first degree Taylor polynomial representation of the fonn
f(x • y) = f(xo, Yo) + fx(xo, Yo)(x - xo) + /y (xo, Yo)(y Yo) + R J (x - Xo, Y - yo)
-
The function R f is called the rema i nder and has the p roperty that
As you are about to see, the idea will be to approximate the functions F and G in th..:
2 x 2 system shown in Equation ( I) by their firs t degree Taylor polynomial s . Thi., \\ill
result in a linear syste m that we know we can s olve. Sol, ing i t, we then obtain a ,olution
that gives us qualitative information about the original 2 x 2 syste m .
An equilibrium solution to the 2 x 2 sy stem is a con stant solution x = Xo, y = Yo
We will denote this solution as (xo, y0). At an equilibrium solution (xo. Jo). we ha,..:
F(xo, yo) = 0 and G(xo, yo) = 0 since the derivatives of the constant function.,
x(t) = xo, y(t) = Yo
are 0. The first degree Taylor polynomial repre sentation of the 2 x 2 system at (.ro, .vo)
is
x' = F(x, y) = Fx(xo, yo)(x - xo) + Fy (Xo, YoHY -)'o) + R, (x - Xo, Y - Yo)
y' = G(x, y) = Gx(xo, yo)(X -xo) + Gy (xo. YoHY - Yo)+ Rc;(x -xo, Y - Yo),
To simplify notation, we let 11 = x -x0 and v = y - Yo, We then have 11' = x' and
v' = y'. Our system becomes
.
= f,(xo, y0)u + Fy (xo, Yo)v+ R, (11. v)
1
11. v}.
v' = G,(xo, yo)u + G.,.(xo, yo)v + Rc;(
11
In this notation
. RF(u, v )
R F (X - xo, y - yo)
= l,m �
= (l
Jim (w.•>-tO, ) vu· -r v-
(x - .x0)2 + (y - yo)
J 2 O
(x,y)-+(xo,Yo)
and
. Rc;(II. v)
Rc;(X - Xo, y -yo) = < 1,1m 1 � = (I.
li m !( 2 + ' u.t• - < o, o v 11· T + v·
(.:c,y)-+(xo.yol (X - Xo) (y - Jo)·
(why?) so we ignore
.
v
u' = Fx (xo, y0)u + F,(xo, yo)
v.
v' = G.:c(xo, yo)u +G ,(xo, Yo)
_
have ___ ______ _
In matrix form we:_:
[
� _=__
]
_
u' = Fx(xo, Yo)[ F,(xo.
G (xo,
YO
}' o
)
)
][
11
v
1
]•
v' Gx (Xo, Yo) y
The matri x
F;c(xo, yo) Fy(xo, Yo)
A= [ G , (xo, yo) G ,.(xo, Yo) J
.
336 Chapter 6 Systems of Differential Equations
is called the linear part of the 2 x 2 system at the equilibrium solutio11 (.xo, Yo). We can
solve. this linear system using the techniques of this chapter.
To demonstrate consider:
x' = - 4x2 + xy,
y' = 2y -xy.
We have
and
We see that
F(x, y) = 0 if x =0 or y = 4x
and that
G(x, y) = 0 if y = 0 or x = 2.
We can now see that (0, 0) and (2, 8) are the equilibrium solutions. Fi iure 6.12 illus
trates the graphs of F(x, y) = 0 and G(x, y) = 0. The equilibrium �··!,1tions are the
intersection points of these two curves and the origin.
Figure 6.12
Fx(x, y) = -8x + y,
Fy (X, y) = X,
Gx(X, y) = -y,
6.7 2 x 2 Systems of Nonlinear Differential Equations .B7
G_,.(x, y) = 2 - x.
and
G(x. y) = G(2, 8) + G,:(2,8)(x -2) + G_,.(2, 8)(y - 8)
+ R<;(x -2, J- 8)
= -8(x -2) + Rc(x-2, y - 8).
We leave it a s Exercise 7 to show th at RF(x-2. y -8)
= -4(x-2) 2 + (x -2)(y -8)
Rdx-2, y-8) and Rc(x-2, y-8)
and Ra(x-2, y-8) =-(x-2)(y- 8) and that
sa tisfy the limit conditions.
.t 8, we tran�form
Using th e substitutions u == x- 2 a nd y = -
x' =-8(x -2) + 2(y - 8) + RF(x -2. y- 8)
y' = -S(x -2) + RG(x -2. )'- 8)
tem:
(u, v), we o btain th e linea r sy�
Ifwe ignore RF(u, v) and RG
u' = -811 + 2v
v' == -811.
m is:
Th e solutio n to this syste 4
11 == ((c2- 4ci) - 4c2t)e- ',
41
v == -8(c 1 + c2t)e- .
Notice th at
lim u(t) = lim x(t) - 2 =0
,_,.,._
t-+-00
and
lim v(t) = ,lim y(t) - 8 = 0 .
.... oo
t-H>J
ial nditionsx(0) =
"f w e ave an init ial value problem with init. co
As a conse.quence, o n (x( r). y(t)) to th e
iently close to (2. 8). the soluu
t h
- o th a t ar e suffic
b y(O) = b2 att _ 8) as t -t oo.
v lue pro blem will also approach (2.
i�itial a
338 Chapter 6 Systems of Differential Equations
and
G(x, y) = Gx(O, O)x + Gy (O, O)y + RG(x, y)
= Ox+2y-xy
and
__,,,
11/ -
......._......._,, "- \12 I///
_,,,, \ 1 J///.,/'- ////
1/ / /- \' ///-
1 // -
_......._,,, \IQ ///--
11 //- y..,..,. ____
__...__....._, \ 1
___,,,\
__,,,\g J / // .,/'-
__,,,\1 1 / // //)
______,,,
11 /// \\ --
__,,,\6 1 I / / --�-------
,. !II I .-�-�
\ \ 1 I I --�-�....-.-
�
----"-'\
___
�-�2�-c:=�==-�1��===�/�� ::=:1
--///
..,..,./
-== ��=:f== ==:::;:::1:=====::±=---x
�3�4·
Figure 6.13
6.7.1 Predator-Prey
island Isle
interaction of moose and wolves on the
Wildlife biologists have studied the hes e wo spe cies . The
e the po pulat ion cycle of t t
Royale in Lake Superior to determin p y the moo se. This
d to the isl and. The wolves re on
mo ose and wolves are confine u s syste ms of diff eren tial equ ations
r-pre.y problem. Var io
proble m is known as a predato n bi log i t A. J. Lo t ka ( 188 0-
interactions. The Austria o s
are used to model these types of 0-1 940 ) are cred i ed wi th the
atician Vito Volterra ( 186
t
1949) and the Italian mathem he predator-prey population
equations models describing t
first system of di ffe rential
relationship. of the pre dat or and
ate n s h mo del, let w(t) represent the population
To illu st r o e uc
t time t. As su me the
mo ose have a birthrate
the p p la tion of the prey a
m(t) repre sen t u he moo�e
the n umber of interaction s betwee n t
o
th e wolves di e
off at a r ate proportion n i ncreases
to int e ractions with
themselves) and that their p opul atio
e off due and wolves since the
(wol ves d i
interactions of the moose
di g on t he number of s describin g
at a rate depe n n
t eraction. The e qu ation
and
= w(t)(-y w(t) + 8m(t)).
w' (t) ::::: -y w(t}2 + 8m(t)w(t)
ution s of this pred ator-p rey m odel are
The equilibrium sol
(0, 0)
340 Chapter 6 Systems of Differential Equations
and
(y
a
8/J' �
a)
The equilibrium solution (0, 0) implies that the predators and prey bom Jie off. The
equilibrium solution (ya/8{3, a//3) implies that the predators and prey C'• • , .�tat constant
populations with the population of the moose being ya/8/3 and the !'·\··ilation of the
wolves being a/{J.
The linear part of the predator-prey problem at the equilibrium (0, 0) is
The eigenvalues of A are a and 0. Since a > 0, the equilibrium solution (0, 0) is
unstable. Hence this model indicates that if the moose and wolf popula, ,1>n� are initially
small they will increase. The linear part of the predator-prey problem at the equilibrium
_Y;]
(ya/8/3, a/{J) is
a8
fJ
The eigenvalues of this system are more complicated to analyze for stab ty. In Figure
6.14 we present a direction field for a = 0.1, {J = 0.0025, 8 = 0.025. a,1d y = 0.05.
In Exercises 8-11, we will ask you to determine the eigenvalues for d,ffcrent sets of
a, {3, y, and 8.
w
Figure 6.14
6.7 2 x 2 Systems of Nonlinear Differential Equations 3-il
Figurc,'i.15 x' =y
y = - s mx.
g ,
1
I
a
. .
. (11'. 0) and the linear part 1s
a fixed frequency. the equilibriu solut ton
Fork = 1 we hav e
m
·=[
g�,
342 Chapter 6 Systems of Differential Equations
EXERCISES 6.7
In Exercises 1--6, detennine (a) the equilibrium solu In Exercises 8 -11, do parts (a-e) of':,ercises 1--6 for
tions of the given 2 x 2 system, (b) the linear system the predator-prey problem
at each equilibrium solution, (c) the general solutions
of these linear systems, and (d) the stability of these m'(t) = am(t) - f:Jm(t)w(t) = m(r)(a - f3w(t))
2
linear systems. Use Maple or another appropriate soft w'(t) =-yw(t) + Sm(t)u·tn
ware package to (e) graph the curves F(x, y) = 0 and =w(t)(-y w(t) + Sm(t)i
G(x, y) = 0 and the direction field for the nonlinear
system. Observe the stable or unstable behavior of the with the given values.
solutions of the nonlinear systems near the equilibrium 8. a= 1/2, f3 = 1/4, y = 1/4,8 = 114
solutions. 9. a= 1/2,/3 = 1/4, y = 1/4, 8 = 3/4
1. x' = x + xy, y' = 2y - xy 10. a= 1/2, /3 = 1/4, y = 1/2, 8 = 3/4
2. x' = x - x2 - xy, y' = 2y - y2 - 3xy 11. a= 3/4,/3 = 1/4, y = 1/4.8 = 1/4
12. Derive the differential equation mllJ" = -mg sin IJ
3. x' = x - 2xy + xy2 , y' = y + xy for the motion of the pendulum in f'1gure 6.15.
4. x' = 2y - xy, y' = x2 - y2 13. For the system
5. x' = 2 + x - 2e-2Y, y' = x - sin y x' =y
14. For the Lmdamped pendulum, determine the solu and y by solving the differential equation
tion of th,; linear system at the equilibrium solution
(0, 11') foe G(O) = 0 and 8'(0) = I if the length of dy dy/dr
-=--=-
y'
the pendulum is 6 in. dx dx/tfr x'
15. The mcl;<,n of a pendulum damped by friction is 17. Do Exercise 16witha =3/4,/J = 1/4,y = !, and
given b:, the differential equation mlO" + ce' + 8 = I /4. Also, use Maple or another appropriate
mg sin () = 0 where c > 0 is the friction constant. software package to graph the direction field of the
Convert tbis into a 2 x 2 system and determine the system and observe the stable or unstable behavior
equilibn..rn solutions. Find the linear systems at the near the equilibrium �olution,.
equilibrit•111 �olutions. Also, use Maple or another 18. One model for describing the competition between
appropri2te software package to graph the direction two species in which neither species is predator or
field of ·1e �ystem and observe the stable or un prey is:
stable behavior near the equilibrium solutions for
m = 0.0125 kg, c = 0.02, and l = 10 cm. x' = (a - /Jx - yy).t
16. The on,,. 11:11 Lotka-Volterra equations for a predator y' = (8 - iu - vy)y.
prey prnrkm are Determine the equilibrium \Olutions of this system.
= -ax + f]xy
Find the linear systems at the equilibrium �olutiOn\,
x' solve these �ystems. and determine the stability of
y' = yy-8xy these solutions.
where x i;, the predator and y is the prey. Deter 19. Do Exercise 18 with a = I, fJ = 1/4, y = 1/2.
mine thti l'quilibrium solutions of this system. Find 8 =I,µ= 1/2. and v = 1/4. Also. U\C Maple or
the linc:,r �ystems a t the equilibrium solutions, solve another appropriate software package to graph the
these sy.-:tcms, and determine the stahility of these direction field of the sy�tem and ob!>enc the �table
solution, Finally, determine the relation between x or unstable behavior near the equilibrium solution�.
7.1
The Laplace Transform
tion. In th i s chapter
you are goi ng to sec anothe r
containing a linear differential equa ial val ue proble ms than
ient to use for solving these init
method that is often more conven re the function g
especially effective in the case whe
the Chapter 4 techniques. It is apter 4. Many engineering
we judiciously avoided in Ch
has disc ontinu ities, a case that nutics, so having a method
problems where g has disconti
problems involv e initial value st i ntroduced by the
s qui te valuab le . The
method involves a function fir
a case lled the Laplace
for such uis de Laplace (1749- 1827) ca
i
When we apply the Laplace transform[, to a function f, .C(f) is a function of s, as the S0lutio11
following examples illustrate.
Solution We have
"°
e-st(l)dt = lim
b
_!e-s1\
b
•
b-+oo s2
0 s2
ifs > 0. Again, ifs S 0, the. integral diverges.
O
s-a
ifs > a. Ifs Sa, the integral diverges. •
0 St::: 1,
f(t) =
I,
I t, t>1
7.1 Definition and Properf,es of the Laplace Transform
J,� ,-" .
Solution o
L(f)
_- I --I e_ ,
= -�e-s _ 2-2 e-s + 2_2 + � e-s -
5 s s s 2 51
s
ifs > O; othe rwis e , it diverge s.
.
Figure 7.1
al
functions f de fin e d on the inter
v
st of all
l cons ider the domain of C to consi
We shal
imp roper integral
[O, oo) for which the
-
L e sr f(t)dt
oe
that the
e op n interval of the form (a. oo). Notice
in som il lustrat e. If we
a ll val ues of s especially Examp le 3)
e
c onverges for ples J -4 (an d tor
on f as Exam is our notati on for the vec
value of a de pends main of[, and reca ll that F[O. oo) is,
w, wi!l lot you show that
V
do
use V to den ote the tl,fioed on tho iotoc,al [O, oo), that L is a liooa,
space of all function
s
e 43 ). w, als<> shall kt you show
oo) (see Exercis
subspace of F[O, e 44). l inear combination
V (s ee Exercis onnation to find l of a
transfonnatio n on ar tra nsf
that[, is a in functions. For instanc e . supp
We can us e the fact w the action of£ on each of the
l e ose
we kno
of functions when
we want to find 2 ).
[,(5 - 3t + 6e '
348 Chapter 7 The Laplace Transform
THEORl!.1'• l 7.2
Since we know
l l I
L'.(l ) = -, [,(t) = -, and l(e2, ) = --
s2 s- 2
from Examples I, 2, and 3, we have
5 3 6
£(5 - 3t + 6e2') = 5£(1) - 3£(t) + 6£(e21 ) = � - + _ for s > 2.
52 5 2
Besides the properties of a linear transformation possessed by the Laplac�· transfonn,
here are some other important properties of£:
Proof We will show property I and leave the remaining two as exercises (see Exerc',es 45 and
46). Note that since
1 °"
Solution We could calculate
00
L(te"') = e-st · teat dt = fo te <a -.<)t dt
directly. Doing so, we will find we have to use integration by parts. (Try it.) Let us
instead use property 3 ofTheorem 7.1 and the result of Example 3:
[,(te"') - - d £(ea') - - d [ I ] -
ds ds s-a (s
1
- a)2
. •
Another way to do Example 5 is to use property I of Theorem 7.1 with J (t) = 1 in
conjunction with Example 2.
We do not consider the Laplace transform of a function J to exist if the integral
Jo"° e-stf (t) dt does not converge for all sin some open interval of the fonn (a, oo).t
Exercise 48 has an example of a function J for which .C(f) does not exist. The nex
theorem gives us a wide class of functions that do have a Laplace transform.
7.1 Definition and Properties of the Laplace Transform
where a i s a con,tant
THEORl<:�l 7.2 If f is continuous on [O, oo) with lf(t)I ::: Me ' for all tin [O. oo)
a
in the ne xt section wh n we
functions have Laplace transforms
01
on [O, oo ) with
THEORl \l 7.3 If f and g are continu ous
l(f (t)) = .C(g(t)).
then
f (t) = g(t)
by .3. the
·
r e em 7
. . . .
.
somethmg sum · ·1ar h by Theor
· t d to W is a one-to-on e function as fol ow:-.: II
. .
f 1s
lace
rang of [, on . . ,
· rse . on the r e-s f (t)dtfor alls m ,111
transf·orm has an inve
e l
. egral Jo
c
· ervaI [O,
e
.
conti nuous on t he 1nt
o
oo) and if
open interval (a,
F(s) == .C(f(t)
).
then
1 ).
C (F(s)) = f(t
we have
ExampI e s J-4 respectively.
To illu strat e, from
.c- 1 G)=1,
c 1 C'i)=t.
£-I - ) ::::e,
( s-
I
a,
a
••••
7.1 Definition and Properties of the Laplare Transform 351
AB
=-+-
F(s)=
I
s2 + 2s - 3
s + 3 s-I ·
Sol ving for A and B, we find A= -1/4, B = l /4. Therefore. we have
l /
F(s) = 2 =- 1 4 + �
s+2s-3 s+3 ,f-1'
and hence
= -�.C- 1
4
(-l-)
s+3
+ � _c- 1 (-'-) = _.!_e-3'
4 s-1 4
+ �e'.
4
•
T here are more extensive tables of Laplace transforms than the one in Table 7.1.
However, as is now the case with many tables, they a re falling into disuse since mathe
matical software packages such as Map le can b e used to obtain Laplace transfom1, and
inverse Laplace transforms of many functions. See Exercises 49-60 for some problem �
asking you to use a software package to do this. Table 7. I ha� been designed to he large
enough for our needs as we apply the Laplace transform and its inver�c to �olvc the initial
value problem s of this chapter.
EXERCISES 7.1
Use the definiti< ,n of the Laplace transform to determine In Exercises l 7-22 use Table 7.1 and pr opertic� of the
the Laplace trnns form of the functions in Exercises 1-6. Laplace transform to find the Laplace tran�form of the
I. 4 given functions.
2 Jr 2t2
5. -41 3. 17. cosh 3t 18. sinh 21 19. 21 \inh t
•e 6. 2e 31
20. 3t cosh 2t 21. e ' cosh bt 22. e"1 ,inh ht
0
Use Table 7.1 and the properties of Theorem 7.1 if nec 23. Show that if f is continuou, on (0. oc) and if
essary to find the Laplace transform of the functions in lim,-:xi f (t)/e'11 is a finite numher l for some con
Exercises 7-16. stant a. then f is exponentially bounded on (0. oc).
7. cos 3t 8. sin 4t 9. t sin t
Jn Exercises 24-34 use 1he result of Exercise 23 and
IO. tcos 3t 12. 3te-u rH6pital's rule to show that the given function� are t:X·
13, e21 sin 3t 15. e"' cos ht ponentially bounded on (0, oc).
14. re- cost
1
e' +e- 33. t" e"' cos bt 3;$. r" tt'' sin ht
1
cosht= �
Find the inverse Laplace tr,msform of the function, in
and the hyperb olic sine, sinh, is defined by
Exercises 35--t2 with the aid of Table 7. I.
8 4
e' - e-'
4
35. - 36 · - 37. --
5 - 6
sinhr=-· s 52
2
350 Chapter 7 The Laplace Transform
and
[,_1 ( -
s 2
I
I - -e
s 2
-s) -_ ! t,
1,
0 St .:'.5 1,
t > I.
This linear combination property is frequently used along with tab les of L:lplace trans
forms of commonly occurring functions to determine the inverse Laplace transforms.
Table 7.1 contains the Laplace transforms (and their corresponding inverses) of some
functions frequently arising when solving initial value problems.
I l
s
n!
2 t" ---:;:T , n = 0, I. 2, 3, ...
s"
s
4 cosbt
s2 +b2
b
5 sinbt
s2 + b2
Table 7.1
We use Table 7. I along with our linear combination property for ,e- 1 in the following
examples to detennine some inverse Laplace transforms.
F(s) = __I __
s2 + 2s - 3
7.2 So
352 Chapter 7 The Laplace Transform
2s 2
any value of s. 60.----- -
-
3s -2 s 2 + 2s + I 0
00
.C(y') = [''° e-•1 y '(t)dt
b
= li m e •- 1 \ +s f e-s1 y (t)dt
lo y(t)
(1)
b--+oo
O lo
= -y(O) + sl(y).
Doing so we get
1 n · + ai/ + aoy) == {,(g( t ))
.C(a,,/' + On-lY( -1) + · ·
or
.
lution to the follow
ing initial value problem
EXAMPLE 1 Deter mine the s o
y(O) = 2
y' -2y = 0,
differential equation , we have
Lapl ace tra nsfor m to each side of the
Solution Applying the
{,(y' - 2y) = .c(O)
or
[,(y') - 2.C(y) = o.
we obtain
Using Equation (1), 2.C(y) = 0
-y(O) + s.C(y) -
(s - ).C(y) =
or
2 2.
We thus have 2
.C(y) = - 2·
s-
354 Chapter 7 The Laplace Transform
•
Applying the inverse Laplace transform gives us
2r
y = 2e .
Solution Applying the Laplace transform to each side of the differential equation gives us
.C(y") + .C(y) = £(0).
Using Equations ( I) and (2), this equation becomes
-y'(O) - sy(O) + s 2 .C(y) + L(y) = -s + (s2 + 1)£(y) = 0.
Solving for .C(y) gives us
s
.C(y) = --.
s2 + 1
•
Finally, using the inverse Laplace transform, we find our solution is
y = cost.
S0l11tio11 After applying the Laplace transform and simplifying (try it), we are led to
-1 + (2s 2 + 3s - 2).C(y) = -1
and hence
.C(y) =
1 +_ _
I __
s(2s2 + 3s - 2 ) 2s
2 + 3s - 2
We now use the method of partial fractions to determine the inverse Laplace transform.
Let us work on the second fraction first. Its partial fraction decomposition has the form
I I A B
+
_ 2_ = (2s - 1) (s + 2) = 2s - 1
-2s__ +_3_s _
2 s + 2'
where we find A= 2/5 and B = -1/5. Thus
+
7.2 Solving Constant Coefficient Linear Initial Value Problems w1'th La pIace Transforms 355
�XERCISES 7.2
In Exercises 1-16, solve the initial value problems using 12. y' - 4y = 5, y(O) = I
) = 0, y'(O) = 0
13. y" + 4y' + 3y = 2 - e '; y(O
3
Laplace transforms.
-1
1. y' - 2y = 0, y(O) = 1 14. y'' - y = Se'; y(O) = 0, y'(O) =
-
4y" +5y' = 7 2co sr; y(O) = l, y'(O) = O.
2• y' + 3y = 0, y(O) = 2 1 5. y
111
-
3 + y" (O) = 2
• Y 8y' = O; y(O) = -1. y'(O) = 0 - 3e-1: y(O) = O.
16. y"' - Sy"+ 3y' + 9)' = 2
II
4 y11 I
' - Y - 6y = O; y(O) = 2, y'(O) = 1 y'(O) = -1,y"(O) = I
5' y11 +6Y + 9y = O; y(O) = I, y'(O) = 2
solve the initial value prob
6· Y - 4y" + 4y = O; y(O) = 0, y'(O) = l Use Laplace transforms to 17-23.
Exerci ses
lem s that arise in
I/
18. A 500-gal tank contains 200 gal fresh water. Salt 22. Determine the current at time t in an rnc circuit that
water containing l /2 lb sa lt per gallon enters the has L = 0.5. R = 2, C = I, and E(t) = 0with
tank at 4 gal/min. The wcl l-stirreu solution leaves initial charge I and initial current 1.
the tank at the same rate. How much salt is in the 23. Determine the current at time tin an RLC circuit that
tank. after IO min? has L = 0.5, R = 2, C = 2/3, and E(I) = 3 sin(2t)
19. I\ cup of liquiu is put in a microwave oven and heated with no initial charge or current.
to 180° F. The cup is then taken out of the microwave 24. Verify Equation (3). 25. Verify Equation (4).
and put into a room with a constant temperature of
70° F. After 2 min the liquid has cooled to 160° F. 26. Verify that when the Laplace transform :s applied to
What wi 11 the temperature of the liquid be at IO min? the differential equation in the initial 1..ilue problem
20. A 2-lb object stretches a spring 6 in. The mass a,,y(rr) + an-lin-1) + ... + a1 y' + .ioy = g(t);
spring system is placed on a horizontal track that
y<rr-ll(Q) =kn-I, y'(0) = k1' :v(O) =ko
has a cushion of air eliminating friction. The object
is given an initial velocity of 2 ft/sec. An external we obtain the equation
force of 3 lb is applied to this system. Determine
the position of the object at time t. -a" y<n-ll(O) - (an s+ an t)y <"-21(0) - · · ·
-
21. A 5-kg mass stretches a spring 25 cm. The mass - (a,,s n -l + a,, _1s"-2 + • · · + ai)y(O)
�pring system is hung vertical ly in a tall tank filled + p(s)l(y) = .C(g(t))
with oi I offering a resistance to the motion of
20 kg/sec. The object is pulleu 25cm from rest and where p()..) is the characteristic polynomial of the
given an initial velocity of 50cm/sec. If an external corresponding homogeneous differential equation.
force of 3e-11 10 newtons is applieu to this system,
Jctcrmi ne the position of the object at time t.
One of the strengths of the Laplace transform lies in the fact that we can use it to solve
constant coefficient linear differential equations
a ,, ),(11) + an -tY (n-1) + · · · + a 1 yI + a0 y = g(t)
with initial conditions
y'(O)=k1, y(O) = ko
fo_r a large class of g. As promised at the end of the previous section, in thjs section you
will see that we can use the Laplace transform to solve such initial value problems for
�ome _ dis�ontinuous functions g. In many applications g is discontinuous. For example,
111 a circu1 t problem g might model a switch that can be turned on or off i n a discontinuous
manner.
O �e cas� of practical significance where discontinuous functions g arise invol ves a
type of f unction called a unit step function or Heaviside function. These are functions
of the form
0, t < a,
(
U a (t) =
1, t�a
7.3 Step Functions, Impulse Functions, and the Delta Function 357
where a is a positive constant. We can think of u u (I) as a �witch that is off (\\ith
u0 (t) = 0) before time a and is on (with ua U) = I) at and after time a. Notice that thc-,c
unit step functions are discontinuous at t = a. The graph of 114 appearing in Figure 7 .2
was obtained using Maple. Observe how Maple puts in a vertical line i.egment at the
discontinuity t = 4. This vertical line segment is not part of the graph of u.,. Graphing
devices frequently do this since they graph by "connecting the dot!>"; that is. they plot
points and sketch the graph connnecting these plotted points whether the point!> shou Id be
connected or not. As we use Maple to graph functions in this text. it b to be undcn,tood
that any vertical line segments at discontinuities should be erased to obtain a correct
sketch of the graph.
0.8
0.6
0.4
0.2
0 2 4 6 8 10 12 14
Figure 7.2
nn of lla, we have
Calcu lating the Laplace transfo
t
.C(ua (t)) = Jo
e -sr . Odt + 100
a
e -sr dt =1
a
oo
e-sr
I
dt = lim --e-s' l
b-+oo s
b
a
l -a'
= -e
s
0.8 �
0.6
0.4
0.2 �
0 2 3 4
Figure 7.3
g(t) =
I 0,
f (t - a),
t < a,
I� a.
An example of a graph of such a g appears in Figure 7.4 where the portion of th� graph
of g fort 2'.: 4 is the graph of /(t) with t ::=:: 0 shifted to the right 4 units. Notice that our
unit step function allows us to write g as
g(t) = u"(t)f(t -a).
1
0.8
0.6
0.4
0.2
0
2
-0.2
-0.4
-0.6
Figure 7.4
7.3 Step Functions' ImpuIse Functions, and the Delta Function ]59
L"°
The Laplace transform of g is
L:io
.C(g(t))
For example,
os(t - n)) = -e-"'. _!.__
..C(uir(t)cost)= l(-uir (t)c s2 + 1 ·
se Laplace transform I of
e we have to find the inver
As anothe r example, suppos
4e--2s
F(s) = - .
s-2
We write F(s) as
4
-2s .-
F(s ) =e s-2
f(t) == C
function� and
valu e prob lems that involve the unit step
me initial
We now solve so
Laplace transforms.
functions. Theorem
e S ince we are de aling with discontinuous en tly. >'e no longer
y her . er a one-10--0ne function.
Consequ
otation and terminolog sfonn
l ace tran''"'
I We ar c abusing n and the Laplace 1ransform is no long we s till will speak of the inve rse Lap- �ct.
appl ies ert heless. " miru!. I
7.3 no longer . Nev K«p '
ction in the usual sensef (<)J. will .,,re c' <F!')) • !<'>· < U l. Som• lile '" d<�i'
have a n inverse fun C( •
,ml. if J '" r"""' '" so '"" F(,) a be °'"'""""""' g s o "'' ,- ' ( f(,))
�, il i" is, ,h,re ""
ru<d "" b< ""iq uedfunction.
.c-1 is a m11ltiral
this by suying that
,..
360 Chapter 7 The Laplace Transform
y "+2y'-3y=
y(O)
I
= 2,
t
0,,
I :'.St< 2,
0 :'.St< I or t :::: 2
y'(O) = 0. EXA\IPI E 2
Solution We express the right -hand side in tenns of unit step functions as
g(t) = tu1 (t) - tu2(t). Solution
In order to apply the Laplace formula for the unit step functions in Equati1Jn (1), we
rewrite this as
g(t) = (t - l)u1 (t) - (t - 2)u2(t) + ui (I) - 2u2(t).
Applying the Laplace transform, we have
.C.(g(I)) = .C.((t- I )u1 (r) - (t - 2)u2(t) + u1 (t)- 2u2(t))
s -
= e-s.C.(t) - e-2s.C.(t) + e-s- - 2 · e-2.
-=-
r e- e 2.s e-s ,e-2s
-- + - - --.
s s s 2 s 2 s s
When we apply the Laplace transform to the differential equation, we obtain
-s e-2s e-s 2e- s
(s 2 + 2s - 3).C(y) = 2s + 4 + e--- + - - --
2
2 s
2 s s s '
which gives us
2s + 4 e-s -2s
.C.(y) = + 2 2 ___e_ _ __
s +2s-3 s (s +2s-3) s2 (s2 +2s-3)
2
-
e .,
-2�-:-e___
-2s
+
s(s2 +2s-3) s(s2 +2s-3).
Next (leaving out the details) we find
r,-t 2s + 4 _3 r 1 _,
( ) 3
s2 + 2s -3 - 2 + 2 e
e
.c-1 ( _
+ 12 e -31
1 I I
-3 + 4e
1
s(s2 + 2s -3) ) -
I
and
' 1 I 2 I 1
.C.- (
s 2(s2 + 2 s -3) ) = -3t - 9 + ;/ - 36e-3r.
Hence using Equation (I), we have that
3 l -3, 5 1
_ e' + 2e
y (t) - - u1(t) + 18u1(t)e-JC1-t J + -I u1(t)e 1 -1 _t- I
2 9 - -u1(t}
g
2 3
+ 9u2(t) - : u2(t)e-3C1- 2J - �u2(t)e1-2 + t; 2 u2( •
6 t).
7.3 Step Functions, Impulse Functions, and the Delta Function 361
In Section 4.5 we saw that i f a mass-s pring system wi th no friction was put in motion
by a sinu soidal ex ternal force w ith the natural frequency, then resonance occurred. We
now consider a model where we turn the force off.
EXAI\IPLE 2 Suppose a mas s-spring system is on a horizo ntal track a nd that the mass i s kept off the
track by a cu shion of a i r from a compressor. The mass is an object that weighs 4 lh and
s tretch es the spring 6 i n. Detennine the mo tio n o r the object if the object is released
2 in. from equi librium and an external for ce sin81 is applied for the first rr sec.
Solution If we let y be the position of the mass, the differential equation with ini t ial conditions
de scr ibing the motion of the object attached to the spring is
J sin81. 0 � I � rr.
-y" +Sy= ( >
8 0, t j'(
equation as
Let us r ewrite the di fferential
8s in8t, 0 � t � rr. (2)
y" + 64y= ( t > rr.
0,
as
and side of Eq uation (2)
We can express the right-h
8(t-rr).
g(t) :8 sin81-8u r (1)sin
r
nsform to Equa
tion (2) gives us
Applying the Laplace tra 64e-n'
s 64
(s 2 + 64)l(y) = 6 + r+64 - �,
s
so that
64 64e-n
+ - 2
64)2 ·
s
l(y) = � (s2 + 64)2 (s +
We th en find
y(t) = � cos 81
I
+ 16 sin81 - z' I
co s81
1 s8(t - rr).
1 ()
t ·
sin 8 (I _ rr) + -u ir (t)(t-rr)co
- -urr 2 · ·
16 ampl i tudes of the
osc1llatto ns do
. . �1. gure 7 5 Note that the
that or
utto n m (Compare this res ult with
We graph the sol ume, bu ; � ste ad stabilize. •
to grow over
i
not continue
tion 4.5.) . ulses (for
Exa mple 4 in Sec tha t arises in
modeli. ng imp
. .
. to conside . ra u
f cu·o
to obtatn a function
n n
1.4
1.2
1
0.8
0.6
0.4
0.2
o�---1--\--1--'-l--,J--Jl-'--1--1-'-1f-+-+'-t--t--'t-__.
-0.2
-0.4
-0.6
-0.8
-1
-l.2
-l.4
Figure 7.5
Oa,.1:(t) =
[ k,
O,
a < t <a+ I/k,
0 � t � a or t �a+ 1/ k
= kua(t) - kua+ IJ.l:(t).
The graph of the unit impulse function 82•10 appears in Figure 7 .6.
EXAl\lPI
10
8 Soh
0 2 3 4
Figure 7.6
Jim .C(llo,t(t)) = 1.
k-+00
.C(8(t)) = 1 or
1
.c-
(1) = 8(t).
orm gives us
So lu tion Applying the Laplace transf
e-2s ·
.C(y ) = --i--
s +1
ce transfonn we find
From the inverse
Lapla
2) =
0
,
0 � t � 2, •
y == uz(t) sin(t - ! sin(t - 2). t > 2.
le 3 is
derivati ve of
the solution in Examp
Notice that the ,
0 0 � t < 2,
y' == cos(t - 2), t > 2.
!
).
Dirac (1902 1984
-
after the Nobe l physicist Paul
2 This is named
364 Chapter 7 The Laplace Transform
The graphs of y and y' appear in Figures 7.7 and 7.8, respectively. Observe that the
solution to Example 3 is continuous, but that its derivative is discontinuous at t = 2.
The initial value problem in Example 3 describes the motion of a frictionless mass
spring system at rest that is hit with an impulse force at 2 sec. The solution in Example
3 indicates that a discontinuous forcing function causes a jump in the velocity of the
mass. Exercise 23 also asks you to consider the jump in the derivative caused by 8(t)
for a specific initial value problem.
I
0.8
0.6
0.4
0.2
Ot--��.J-���L-.�\-----_L���.L,--+
-0.2
-0.4
-0.6
-0.8
-I
Figure 7.7
1
0.8
0.6
0.4
0.2
Or-��� �-\-L��----1..----Jc__�L---
-0.2
-0.4
-0.6
-0.8
-1
Figure 7.8
EXERCISES 7.3
4. v(t) ={ 310
0,
0 5 I <rr,
t,
t ::: rr
tank at the same rate. Suppose for the first IO min
the salt concentration of the so lution entering the
tank is l/2 lb/gal and 1hen the salt concentration of
s. h(t) ={ ,:�s 2t,
01
n:;:t<2T(,
0 5I<T( or t ::: 21'(
the entering solution is reduced to 1/4 lb/gal. How
much salt is in the tank after 30
min?
at I 000
0, 051<21'(, 25. An amoeba population in ajar of water :..tan,
6. �(I) == { After this time.
3 �in 31, t � 2rr amoeba and doubles in 30 minutes.
nuou sly from the jar at a
(b) amoeba are removed conti
In Exercise s 7·-10, (a) graph the given function and rate of I 000 amoe b a per hour for 2 hr. Afte r the end
determine the ! .aplace transform of the funct ion. ved from
of these 2 hr. no more amoeba are remo
8. u 1 (t)e3 e- r grow th. ho"' many
the jar. Assuming continuous
3
7. tui(t) - 'Ltu.l(t)
amoeba will be in the jar after 4 hr?
9. u"(t) co� t 10. tu" (t) sin t
g 6 in. The ma�,
= 26. A 2-lb object stretches a sprin
In Exercise � I l--14, find a function f so that f(t) spring syst em is pla c ed on a hori zontal track that
.c-1( F(s)) for the giv en function F(s). has a c ushion of air eliminating frict ion. The object
e-3
s
city of 2 ft/se c. An external
is given an initial velo
s
e-2
11. - 12. - to this sys tem for 5 ,cc and
s2 force of 3 lb is appl ied
n of the ohje<.:t.
s
e-
s e - - 2s then shut off. Dete rmine the positio
13.-- 14.- +4 ng 25 cm. The ma,s
s2 - 4s s2 27. A 5-kg mass stretches a spri
us spring system is hun g verti cally in a tall tanf.. filhxl
In Exercises 1 'i-22, solve the initial value problems a resis tanc to the motion of
with oil offering
e
I+
ing Laplace ,r,msforms. ed 25 cm fro m re,t and
l•�. y 3y =
0.
{
t >2
(,
y(O) = 0
0 5 t.::: 2,
20 kg/sec . The object is pull
given an initial velocity of cm/
force of 3e-
50 sec. If an external
11 10 newtons is applied to thi'> sy�tcm for
rmine the motion of the
!6. y' -4y = 5tu 1(t), y(O) = I 3 sec and then stopped, dete
0 mass.
17. y" + 4y' + 3y = e3r-3ui(t); y(O) = O, y'(O) =
t has L == 0.5. R = 2. C = 2/3
0 _::: t< 3, 28. An RLC circuit tha
, current is connected to a
r
18. y'1 _ y = { Se with no initial cha or rge
0, t�3 vo ltage £ ( t) = 3 :.in (21).
generator tbat supplies the
19
y(O) = 0, y'(O) = -1
I I
4 {
7 - 3 COS 2f 05 t <21'(, 1 If the generator is turned
off, what is the vo ltage
on for T( sec and then turned
in the system?
that has L = 0.5. R = 2, C =
• y - y +Sy= t ::: 2rr 2/3
0,
29. An RLC circuit or c urrent b connect ed to a
y(O) = I, y'(O) == O with no initial charge
={ 051 <2, a switch . If he swi tch is
6-volt batte ry through
t
0. off. what i� the
20. y" - 5 y' - 36)•
t::: 2 d on fo r 2 s e c and then turned
2- 3e-', turne
Y( O) = 0, y' (0) = - I voltage in the system?
osition of
function, detern1ine the p
Zl. Y11 + 2y' - 8y = 8(t); y(O) = 0, y'(O) =
0
30. Using the delta 26 if the external force of 3 lb
Z2. y" - 4y' - 3y = 8(t - 2); y(O) == 0, y'(O ==
) O the object in Exercise
0 sec.
e p roblem is an impulse at t ==
ZJ. Co mpare the solution to the initial valu tion to = 2/3
that has L = 0.5, R = 2. C
y" + y = O; y( O) = O, y'(O)+= 1 to
the solu 31. An RLC circuit cted to a
e or curren t is conne
== 8(t); y(O) == O, with no initia l charg itch. u�
the initial v al ue problem ning a
-
y" y
respect to. the a wire contai !ilA
the delta function to find the charge in the system if 35. u ,.. t
( )cosr+28(t-2)
the switch is turned on and off at 1T sec.
In Exercises 36-39, use Maple or another ,1ppropriate
In Maple the unit step function u ,, t
( ) is typed as software package to find a function f so 1hat f (t) =
Heaviside(t - a) and the Dirac 8 function typed as .c-1 (F(s)) for the given functionF(s).
Dirac(t). In faercises 32-35, use Maple or another e-1.s e-•
36. s - 37. s2
appropriate software package to determine the Laplace 2 -9
transform of the function. 4
38. e2-,rs
+ -4 39. e- 3• (� + 2)
32. u1(t) s s2 + I
As we have seen, one way to do this is to determine the partial fraction decompo
sition of
I I
Hs( ) = 2 - 2s - 3 = - (s
_ _ _ 3
_ )_s( +
_ _l)'
s
We now are going to develop another way of determining e,-1 (H(s)) in terms of
.c- 1 (F(s)) and .c- 1 (G(s)).
We do this by letting f (t) = .c-1 F
( s
( )) and g(t) = .c-1 (G(s))and note that
( ) = .C(g(t - r)u t (t))
e-srGs
{ 00
-
=J e st g(t-r)u,(t)dt
o (1)
= f 00 e-st ( -
g t r) dt.
7.4 Convolution Integrals 367
We have that
F(s)G(s) = G(s)£(f(r:))
oo e-sr f
=G
(s)
l (r:) dr
: e-srG(s)dr.
fo f(r )
00
=
J. ro ,-" L f(,)g(t
J.ro 1 f(r),-" -
ro
of integration to obtain
d, - r) Jr dr
g(t ,)dr -
= [, (L f (r)g(t - r) dr).
that
We have, therefore, shown
1
C (F(s)G(s)) = L f(t)g(t - r)dr.
Figure 7.9
368 Chapter 7 The Laplace Transform
*
We will also write (f g)(t) as f (t) g(t). *
We now apply the convolution integral to some examples. We consider the P· oblem
introduced at the beginning of this section.
s2 - 2s - 3·
s+I s-3
Since
and
.C(e3') = _1_.
s-3
we have that
e,-1 ( I )-
-
s2 - 2s -3
e-1 * e3' =
0t i () 1'
(Compare this to the problem at the beginning of the section.)
41
e-4T dr = -e3' e- - I =---
e-r eJ<1-rJ dr: = e3t
4 4 4
•
EXAMPLE 2 Solve the f o llowing initial value problem.
- � *smt
y-e Jo
sm(t-r)dr= 2 (e- +sin1-cost).
. = ('e�- 1 I •
Convolution integrals are also useful for solving some integral equations. which are
equations similar to differential equations except that they involve integrals ins tead of
derivatives. See Exercises 13-17 for details about th is.
EXERCISES 7.4
0 � t < 1,
t �I
Solution We use the unit step function to first write the system as:
y; = 2y1 - Y2
Y2 = 5y1 - 4y2 + U1(t).
Now applying the Laplace transform to each equation in the system gives us
.C(y;) = .C(2y1 - y2)
.C(y2) = .C(Sy1 - 4 y2 + UJ (t))
or
£(y;) = 2.C(v1) - .C(Y2)
.C(y2) = S.C(y1) - 4.C(y2) + � e-s
s
Using the property EXAMPLE2
.C(y') = -y(O) + s.C(y)
in Equation (I) of Section 7.2, we have
-I +s.C(y1) = 2.C(y1)-.C(Y2)
or Solution
(s - 2).C(y1) + .C(y2) = I
1
-S.C(y1) + (s + 4).C(y2) = - e-s.
s
Solving this system for .C(yi) and .C(y2), we find
£(y ) = s+4 _ l -s
i e
(s +3)(s -1) s(st3)(s - l)
r )= 5 s-2
.t...(y2 + e-s
(s + 3)(s - I) s(s + 3)(s - 1)
7.S Systems of Linear Differential Equations 371
4
Y2 = e - e
4 3 4 -
12 u1(t)e
The graph of the solutions is in Figure 7.10. (Which one i s the graph of y, and which
one. is the graph of Y2? Hint: Use the initial conditions.) •
EXAMPLE 2 Solve
x;'::::-lOx1+4x2
8(t)
xf:::: 4x1 - 4x2+
x 2(0) = 0, xi(O) = 0.
property
Lap lace tran sfonn and using the
the
sy(O) + s C(y)
Solution Apply ing 2
C(/ ):::: -y'(O) -
us
Section 7 .2 gives
in Equation (2) of (x1) + 4C(x2)
52.c(xi) = -10C
.
52 .C(xz) =
4.C(x1) - 4C(x2)+l
system as
We rewrite this - 4C(x2) = 0
(s + !O)C(x1)
2
4.c(xi) _ (s + 4)C
2 (x2) == -1.
372 Chapter 7 The Laplace Transform
C (x i ) =
(s2 + 2)(s2 + 12)
s2 + 10
.L: ( x2) = (s2 + 2)(s2 + 12)
Applying the inverse Laplace transform gives us
.Ji sm
x1 = - . ./it v'3 . r,:;
- -sm3v2t
5 15
2J2 � .
x2 = -- sin ht+ -sm3v2t.
r,:;
5 30
The graph of the solutions appears in Figure 7.11. The lighter curve is the t;raph
of xz. •
Figure 7.11
In the exercises we will ask you to solve systems that arose from applications
considered in Chapter 6. Some of these will now involve the unit step function or the
delta function.
EXERCISES 7.5
In Exercises 1-8, solve the initial value problems using 2. y; = 3y1 + 5y2
the Laplace transform.
Y2 = Y1 - Y2 +t
1. y; = Yt - Y2 Y1(0) = l, y2(0) = 1
0,
{
0 � t < 2,
y2 = 2y1 +4y2 + , 3. y; = Yt + Y2
I
4 t::: 2
Y2 = 4y1 - 2y2 + e'ui(t)
Yt (0) = 0, Y2(0) = I
Yi(O) = 1, Y2(0) = 0
7.5 Systems of Linear Differential Equations 373
Ui&W
Power Series Solutions
to Linear Differential
Equations
The techniques w e have seen for sol ving higher order linear differential equations in
Chapters 4 and 7 center around the constant coefficient ca�e. One commonly used
vc
approach for solving nonconstantcoefficientlinear differential equations invol s finding
tia l equati on s . Initial l y you might react 10 thi� by
power series sol utions to the differen
unnatu ra l and w ou l d not be usefu l solution� to know
thinking that pow er series are
ctive. Whi le you may feel
to a differential equation. But this is not a proper perspe
funct ions repres ented as power series. in many
some apprehensio n about dealing with
way t o r epr es ent f unction s. One in stance of thi�
instances power series are the best
occu rs w hen appr o ximat ing values of functio ns.
y ou have seen in your calculus classes °
sin 1 = sin(.7!' /180 ), for example , is to use a
A standard method for approximating
al .7l' I 180, which is the same as a truncation of
Macl aurin polynomial of sin x evaluated
d at .7l' I 180.
the Macl a urin se1ies of sin x evaluate
view , if y ou have studied Section 3.9 you might ask:
From the approximationpoint of
techniques and just appl y numerical methods ?"" One
"Why not l eave out power s eries el for approx
which numerical methods do not work w l
reason is that there a re points at series tech niq ues are important
l equations. Al so power
imating so l utions to differentia numerical method s are ba� ed on po wer
l utions since many
in the study of numerical so the on l y rea son for study
numerica l considerations are not
series techniques. Further, g nom ena in ma th ematical
For instance, many inter�stin phe
ing power series s olutions. e thods .
d through t�e power series -� . ..
physics have been disco vere l w ith the comp l ex1t1es and pecuhant1es o f power
dea
There are entire texts that We give onl y an overview here and Jea,e
for differ entia l equations. ·
· methods
s eries · the ti rst section o f th'1s chapter
ourses. We begm
e s for future c · so I · ons to d'1!1e -� ·
many of thc technical issu w e con
.
sid er pow er sen e s utt rent1al
· series before
wit· h a review o 1· power
equations.
375
376 Chapter 8 Power Series Solutions to Linear Differential Equations
f" o (x ) o (x )
"'
(Of course, for this nth degree Taylor polynomial to exist, f must have an 11th derivative
at x0. Throughout this chapter we will assume any indicated derivatives exist.) In the
particular case when x0 = 0, the Taylor polynomial is called the Maclaurin polynomial
of degree n. Taylor polynomials are useful for approximating values of functions and
(n+I> c
()
Rn (xo) = f( + I)! (x -xot.
n
THEOREM 8
f"( o) "'
f (xo) + f' (xo)(x - xo) + x (x - x0) 2+ ! (xo) (x - x0) 3
2 3!
)
+···+ f(n)(xo (x-xat+···
n!
)
� f(n)(xo
=� 1 (x -xo)".
n=O n.
8.1 Introduction to Power Series Solutions 377
In the spec ial case when xo = 0, we obtain the Maclaurin series of J(x):
•l(Q) tt
J"(O)x2 + J"'(O) X3 + "• + --X n + • •
f(O) + J'(O)x + 2 3! n! •
00 f
'"l(Q)
=""""-x" •
� n.I
n=O
If
fun R.(x,xo) =0
n-HXl
3! 5!
x2 x4 x6 -00 < X 00,
2 4! - 6!
COS X = 1 - + +•·•, <
and - 1)4
(x - 1)2 (x - 1) - (x
3
+ .. ·, 0 < X � 2.
-1)- + 4
lnx = (x � 3
r series about a fixed value x0
of power series. A powe
Taylor series are examples
is a series of the form
00
[a.(x - xo)".
n=O
rning
s from your calculus courses. First, conce
us rec l some facts abo ut these serie
Let
have Theorem 8.1.
a l
their convergence, we
real values of x satisfying Ix - xo\ < R and diverges for all real values of x satisfying COROLLA RY 8.3
Ix -xo\ > R.
The value of R in Theorem 8.1 is called the radius of convergence of the power
ser ies. The set of all values for which the power series converges is called the interval
of convergence of the power series. Since the real values of x satisfying Ix - xo I < R
are the same as those in the open interval (xo - R, xo + R), the interval of convergence
always consists of this open interval with or without its endpoints depending on whether
o r not we have convergence at an endpoint. In the case when R = 0, we do have
co nvergence at the (only) endpoint xo and the interval of convergence is the dosed
interval [x0, x 0] consisting of the single value xo. In the case when O < R < oo,
the interval of convergence has one of the forms (xo - R, xo + R), [xo - R, xo + R),
(xo - R, x0 + R], or [x0 - R, xo + R] depending on whether or not the power series
converges at the endpoints xo - R and x0 + R. In the case when R = oo, thC' interval
of convergence is (-oo, oo). The value of R is often found by applying the ratio test or
the root test to the absolute values of the terms of the power series.
One property that power series possess that will enter into our work here is that,
between the endpoints of the interval of convergence, these infinite sums can be differ
entiated term by term just like finite sums.
Y = La,,(x -xo)"
n=O
00
[a,,(x - xor
) LOO
has radius of convergence O < R S oo. Then for all x in the open interval \x -x0\ < R ,
d (
=
d
(an(X - xot)
y' = dx dx
n=O n=O
EXAMPLE I
L
00 00
Further, the power series 1::1 na (x - xoy-• also has radius of convergence
n R.
Solution
We are going to use the fact that we can differentiate power series term by term to
find solutions to initial value problems for linear differential equations where the initial
conditions are specified at a value xo as a power series about x0; that is, we will seek a
solution of the form
00
Y = Lan(x -xo)".
n=O
An �ther way to :iew this is to realize that we are going to specify the solution in terms
of its Tay�or series about xo since, as the following corollary to Theorem 8.2 tells us,
power senes and Taylor series are one and the same.
8.1 Introduction to Power Series Solutions 379
1ROLLA RY 8.3 lf
00
J<"\xo
a. = -- ) -
n!
is" a (x - xO)" ,
"°
and hence the Taylor s eries of f(x) expanded about xo L...n=O n
if J (x) is equal to a power series
A functio� f is said to be analytic at a point x0
(that is, if f(x) can be expressed as
on some open interval jx - xol < R about xo
00
y = La,, (x - xof
n=O
the coefficients a. of the tenns to
em at xo involv es finding
to an initial value probl using the initial conditions and the
coefficients are found by
the power series. The se rate.
following examples illust
differential equation as the
linear initial val ue problem.
nine a pow er seri es solution to the following
EXAMPLE 1 Deten
y' =2Y, y(O):::: I
the fo':11 y =
desir ed solution will have _
L� a. x ".
= O so that our utmg == Ln=<> a. x•
Solution In this problem, xo tion in the form y' -2 y =
0 and_subsut y
renti al equa
writingour diffe rem 8.2, we obtarn
using our form ula for y' from Theo
into it
I:
00
a.x• = 0.
00
2
y' - 2y == �)n +
l)an+ i x" -
n::O
n&
This leads to
L)" + l)a.+ 2a.)x" == 0.
00
1 -
,1:0
if
This equation holds if and only
n == 0, 1, 2, ....
Zan == 0 for
(n + l)an+l _
380 Chapter 8 Power Series Solutions to Linear Differential Equations
EXAl'\!PLE 2
Solving for a11 +1 gives us
2an
a,,+1 =--.
n+l
• 1/utio11
From this equation we can generate all the coefficients in the power series for y. This
equation is called a recurrence relation. Using n =0 in Corollary 8.3, we have that
y<O)(O)
a0 = -- =y(O) = I.
O!
Now, from the recurrence relation we have:
2ao
a1 = -=2
1
2a1
a2 = - =2
2
2a2 4-
a3 = - =
3 3
Clearly, we can generate as many coefficients of tbe power series as we want using the
recurrence relation. We can even use the recurrence relation to obtain a formula for a,,.
Notice that
2
a1 = -ao
1!
2a1 22
a2 = = ao
2 2!
2a2 23
a3 = = ao
3 3!
2a3 24
a4 = !ao
4
=
4
from which we see
L..,
1
n!
'°' -
(2x
'
t
n=O n=O
which is the Maclaurin series for e2x . This is the same solution that would have been
obtained using the techniques of Chapter 3 or 4. •
8.1 lntroduction to Power Series Solutions 381
EXMiPLE 2 De termine a power series solution to the following linear initial value problem.
2
y '=(x-l) y , y(l)=-1
., ,[lltio11 In this problem, xo = I and we subst itute in y = E:;':0 an (x -I)" into y' -(x - l)2 y = O
to obtain
00 00
+2
11(X-1r = 0.
y' - (x - 1)2y =I>+ l)an+ 1<x - 1)" - I:a
n=O
n=O
11=0
giving us
00 2
l)(n + 3)a11+3 - a,,)(x
- !)"+ = 0.
a1 + 2a2(x - 1) +
n=O
Therefore,
and (n + 3)an+J - a,, = 0
a1 = 0, a2 =0,
relation is
and our recurrence 011
an+3 = -·
· n +3
:
Consequently we have
a 1 == 0
a2 = 0
ao
(13 = 3
== - = 0
OJ
(14
4
a3 -ao
Cl6--= 3 ,6
- 6
382 Chapter 8 Power Series Solutions to Linear Differential Equations
as
a8 = - =0
8
It follows that only coefficients of the form a3,, for n = 1, 2, 3, ... are nonzero and that
a311 = --ao.
311 n!
Since a0 = y( I) =-1, we have that
'°'---
00 00
I [(x - 1) 3 /3]"
Y = - �-(x-1)3" =- � n!
�3"n!
11=0 11=0
This is the Maclaurin series for -e <x -1) 3, which is the same solution as obtaint>d by
3
/
EXAMPLE 3 Determine a power series solution to the following linear initial value problem.
Solution Here x0 = 0 and we substitute y = I::o a11 x 11 into y" + y = 0. In order to do so, we
needy" which. applying Theorem 8.2 twice, is
y" = 11 11 +2x".
11=2 n=O
11=0
Consequently:
ao - - ao
a2 = _
2 - 2!
a3=-� =-�
2·3 3!
a2 a0
a4=-- = -
3 · 4 4!
G3 OJ
as=--= -
. 4· 5 5!
00 00
+ La2n+tX ,
2n+l
= I: a2.x
2n
n=O n= O
we have
00
00 a1 x 2n+l
y = L(-l)"�x + n=
L(-lt (2n + I)! 2n
.
""'O (2n)! O
Thus
n I x 2n
L.,(-1) -
y = \'
•"'°
2n)! (
same solution obtained
t xo = 0, which is the
series for cosx abou •
This is the Maclaurin
Chapter 4.
using the methods of
�CISES 8.1
= 0; y(O) = I, y'(O) == 0
· soIuuo 5. y" - y
In Exerc1ses I-8 , ti mJ thc power series · n to the
· · y'(O) = 2
initial value problem. 6. y" + 9y =0: y(O) = 0.
1. y' + 2. y' = y, y(O) = -
1 = I, y'(O) = - I
) = 0. y(O) = 2 7. y" + y' =0; y(O)
3. y1 -
- (?-X - 2)y, y(l) = I = 2. y'(Ol = I
8. y" - xy' = O;
y(O)
'
4. y' + (X + 1)-y = 0. \'(-1):::: 0
384 Chapter 8 Power Series Solutions to Linear Differential Equations
In Exercises 9-12, find the power series solution to the and hence
initial value problem and determine the interval of con y"(O) 2y'(O)
vergence of the power series solution. a2 =-- = - - =2.
2 2
9. )'1 =X}', y(O) = I Differentiating again,
JO. y' + (x - 2); y = 0, y(2) = I flf
y = 2y"
11. y" + 4y = 0; y (0) = 1. y'(0) = 0
so that
12. y" + y' = O; y(O) = 0. y'(O) = I
y111 (0) 2y"(O) 4
a3 - - = _3__ = 3·
=
00
The Maclaurin series is then
y = I:011 (.x - xo)" 4 3
I +2x+2x + -x +···.
2
n=O 3
of an initial value problem in the form Continuing this procedure, we obtain more terms of the
oo y (n)( ) series. Use this alternative method to obtain the fiN four
L
n=O
(
XO
--X-Xo
n!
)
" nonzero terms of the Taylor series solutions in Exer-'ises
13-16.
gives rise to an alternative method for finding the Taylor 13. y' = 2xy, y(O) = I
series solution: Use the initial conditions to get the first 14. y' +x 2 y = 0, y(O) = I
terms of the series and the differential equation and its 15. y" = 4(x - 2)y; y(O) = I, y'(O) = 0
derivatives to get the remaining terms. To illustrate, we
16. y" + (x + l)y = O; y(-1) = 0, y'(-1) = I
will generate the first three (em1s of the Maclaurin series
of the solution to the initial value problem In Exercises 17-20, use the dsolve command in Maple
with the series optio11 or the corresponding command in
y' =2y . y(O) =I another appropriate software package to find the first six
in Example I. We have a0 = y(O) = I from the initial terms of the Taylor series solutions of the initial value
f
condition. From the dif erential equation, problems.
17. y" = 2xy, y(O) = I
a 1 =y'(O) = 2y(O) = 2.
18. y" + x2 y =0, y(O) = I
Differentiating the differential equation, we have 19. y" = xy' - y; y(O) = I, y'(O) = 0
y" = 2y ' 20. y" - y' + (x + l)y = O; y(-1) = 0, y'(-1) = I
4. 1 we know there is a unique solution to this initial value problem. We first consider
the case where q, r, and g are polynomials. (Polynomials are analytic functions whose
Taylor series have only a finite number of terms-see Exercise 25.)
EXAMPLE 1 Determine a power series solution to the following initial value problem.
y"+xy=O; y(O) = l, y'(O) = 0
00 00
giving us
2a2 = 0 and (n + 3)(n + 2)an+3 + a,. = 0
or
an
a1 = 0 and an+3 = - (n +J)(n + 2)
·
Consequently:
ao
a3 = - 3!
a1 2a1
a =-
4 12 = -4!
a2
a5 = -- = 0
20
a3 4ao
·
ll6 ____
- 6 5 =-6!
a4 == 2 · 5a1
a7- --- -
- 7·6 7!
a8 = 0
4 · 7ao
a9 =---
a6
=--,
9-8 9
386 Chapter 8 Power Series Solutions to Linear Differential Equations
a7 2 · 5 · 8a1
a 10 = - 0 · 9 = - 1 !
1 0
a11 = --as - =0
11 · 10
2a 1 2 · 5 .. · (3n - 1)
00
+ a1x - -x 4 + L 1
a1x311 + .
n
(-1)
4! n=2 (3n + l)!
y" -2xy' + y =I:<,,+ 2)cn + 1)a11+2x" -2x [en+ oan+ 1x" +I: a,,x"
n=O n=O n=O
00 00 00
This gives us
oo (2n + l)Gn+l
a2=-- and
+ 2)·
�..:.:...:_::__
On+J=--
2 (n + 3)(n
Thus:
a1
=
3!
a3
3 3
a4 = --02 = --a0
4·3 4!
5 5
as = - a3 - = -ai
5· 4 5!
7 3.7
a6 = . 04 = -6!a0
6 5
9 5.9
a7 = --05 = -a1
7· 6 7!
I· 5 .. · (4n - 3)
a211+1 = a1 for n = l, 2, 3....
(Zn+ l)!
refore. we
= y(O) = I anda, = y'(O) = .1 The
The initial conditions tell us that ao
have that
I · 5 .. ·(4n - 3) 2n+t
aox2 � 3 · 7 · .. (4n - 5) a x211 +a, x + �
o f__, alx
Y = ao - -- - 2n + 1)!
t
f__,
(2n) ! n=I (
2 n=2
·1 · · (· 4n -3
_ )_ x 2n+t ·
_1_
-
2
-
x2
n=2
3 · 7· · · (4n - 5) x2n + x + f-- 5_ __ _ _
(2n)! ,, 1
�
(2n + 1)! •
s with one depe nd
tha t the pow er seri es solu tion can be expressed as two sum nd
No tice
er dep e nd ing on o 1 in Examp
les I and 2. (Of course, the seco
ing on 0 and the oth for a hom oge neo us
e a 1 = 0.) This is always true
a
ppe d out in Exa mple I sinc
sum dro
equation
y" + q(x)y' + r(x)y
=0
inary
atx . When q and rare ana
lytic atx0, we say xo is an ord
when q and ra re ana lyti c de lead to the followi ng
Indeed the commen ts just ma
0
caw
388 Chapter 8 Power Series Solutions to Linear Differential Equations
where y1 (x) and y2(x) are power series in x - xo and ao = y (xo) and a 1 = y' (xo).
But we can say more. Concerning the radii of convergence of YI (x) and y2(x) we
have Theorem 8.5.
THEOREM 8.5 Under the hypotheses and notation of Theorem 8.4, the radius of convergence of each
power series y 1 (x) and y2(x) is at least R where R is the smaller of the radii of conver
gence of the Taylor series of q and r about xo.
Finally, we can even use these power series to obtain general solutions. Omittii;g
specific values for ko and k 1 in Theorem 8.4 means that a0 and a 1 become arbitrary
constants. This leads us to Theorem 8.6.
As an illustration of Theorem 8.6, we have that the general solution to the differential
equation
y" - 2xy' +y =0
(i _ n) ( �
in Example 2 is given by
Up to this point in this section we have only looked at homogeneous equations. But
the approach of Examples l and 2 also works for nonhomogeneous equations. as the
following example illustrates.
11=0
00 00 �
y11 +2y' -xy = L(n + 2)(n + l)a,, +2x" + 2 L)1 + l)an+1x" - x L a11x"
11 =0 11=0 n=O
= I +4x.
Equating coefficients, we have
la2 +2a, = l, (603 + 4a2 - ao) = 4, and
2 1
a•=
., --a4+2-a2
5 0
l 1
a6 = - as + 0a3
3 3
Listing the first seven terms of the power series, we obtain:
I 2 l 3 I 4 11 5 7 6
y = 2X + 3X - 6X + 120X - 360X +... •
The last example is not atypical. In many cases it is not possible to determine the
general term of the power series solution. This is especially true if either q or r are not
polynomials. In this case we have to express q(x) or r(x) as power series to find power
series solutions as the final example of this section illustrates.
EXAMPLE 4 Determine the first four terms of the power series solution aboutxo = 0 to
y" + (cosx)y = 0
in terms of a0 and a 1.
oo
.m=O a11 x
oo
and L
"00 or y to get
°"(n+2)(n+l)a 1+2x 1 +
L., 1 1
( x2 x4
1--+---+··· °"ax11 =0
x6 )
(1)
2 4! 6! L., Tl�
II '
II�
In order to get the first four terms of the power series solution, we only need to look at
terms involving I, x, x2 , and x3 in Equation ( 1). Consequently, terms involving powers
of four or more of x in the Maclaurin series of cos x will not come into play. This enables
us to view Equation (I) as
00 00 2 00
L(n + 2)(11+ l)an+2x" + La,,x" - � I::a11 x 11 + ... = 0.
11=0 11=0 11=0
I I
differential equations of the form
n=O
00 0C
-L (n + 2)(11 + l)an+2X
n+i
00
+ v(v + l)a1x + x
n=2
= 2a2 + v(v + l)ao + ([v(v + 1) - 2]a1 + 6a3)x In Exr �t,tSes 1 3-16, determine the first three terms of the a) Determine the power series solution to this
powr:- ,l"ries soluti on to the differenti al equation. equation for Ix I < 1.
13. y" - y COS X = 0 b) Show that if a is a natural number, then there is
00
polynomial, use reduction of order to find the second a 1 x + a o and a number x0, determine ho, b1, ... , hn
v(v + 1) linearly independent solution. so that
a2 = - ao,
2 17. (l - x 2 )y" - 2xy' + 3/4y = 0 p(x) = ho +h1 (x -xo) +b2(x -xo)2 + ...
18. (l - x 2 )y " - 2xy' -1/4y = 0 + hn (X -xo)" + O(x -xo)"+t
a3 = 19. (t -x 2 )y" - 2xy' +2y = 0
+ O(x -xo)"+2 + · · ·
2-v(v+l)
a1,
6 20. (1 - x 2 )y" - 2xy' + 6y = 0
= bo +h1 (x -x0) + b2(x -xo)2 + ...
and
In Ex,·rc ises 21-22, determine the polynomial solution + hn (X -Xo)"
to thr ',egendre equation.
+ 2)(n + 3) - v(v + 1) for all x and hence conclude that every polynomial
an+4 = 21. (l - x 2 )y11 2xy' + 20y = 0 equal to its Taylor series a�ut xo for al I x and also
is
(n
Gn+2·
that its Taylor series has a finite number of terms.
-
The recurrence relation gives us two l i nearly i ndependent power series solutions to the 23. The equati on 26. a) Show that in Theorem8.4 ao = ko, a1 = ki,
Legendre equation as described i n Theorem 8.4. An interest i ng result occurs in the case ai = y"(0)/2 = -(q(O)k 1 + r(O)ko)/2.
y" -2x y' + AY = 0 b) Continuing the process in part (a), show that we
when
· known as the Hermite equa-
�-here J.. is a constant 1s 3 can generate a power series solution for
v = k+2 t:vn. This equation i s important i n physics and nu- Theorem 8.4 that satisfies y(O) = ko and
merical analysis. y'(O) = k1.
for some nonnegative integer k. Noti ce that, in this case, ak+4 = O, wh i ch then forces To complete the proof of Theorem 8.4 we need to
coefficients depending on ak+4, wh ich are the coeffic ients a) Show that the power series solution is � show this power series converges.
.
polynomial (called a Hermite polynomial) if>..
is an even natural number. In Exercises 27-3 0, use the dsolve command in Maple
b) Determine the Hermi te polynomials for with the type=series option or the corresponding
to be 0. Consequently one of the l i nearly independent power seri e s solutions in Theorem command in another appropriate soft ware pa.ckage to
J.. = o, 2,4 , 6, and 8.
8.4 is a polynomial Pk+2(x) of degree k + 2. (Why?) When the condition Pk+2( I) = I is obtain the first six terms of the power senes �lu-
imposed, Pk+2(x) is called the Legendre polynomial of degree k + 2. In this setting, it 24. The equati on .
non. Compare this answer to your results for Exemses
is sometimes easier to obtai n the two solutions by first finding the Legendre polynomial (1 - x 2)y" -xy' + cx zy = 0 13-16.
solution and then using the method of reduction of order introduced in Chapter 4 to find . 4 27. y" - ycosx = 0 28. y" + e7 )' = 0
the other solution. Exercises 17-20 ask you to apply thi s approach. where a is a constant 1s known as the Chebysbev
29. e7y" + xy = 0 30. y"-2y'+ysinx = 0
equation.
EXERCISES 8.2
EULER TYPE EQUATIONS
. solut1·ons about xo when xo is an ordinary
8.3
. power senes
In Exercises 1 8,- determi ne the power series solution to 6. y" + xy' - y = x; y(O) = 0, In the last section we cons1· dered
y'(O) = 0 series solutions when xo is not an ord'mary
the initial value problems. point. Our �ext objec tive is t on
7. y" -xy = I +x; y(O) = 1 , y'(O �:;�:�i�:;:;tial equation called an Euler type equati
1. y" - xy = 0; y(O) = I , y'(O) = 0 )=0 poi nt. In this study a type. o: homo -
8. y" + x 2 y = -2x; y(O) = 0, . •
h ns1ona
1me I equation) will arise. In the, second order
2. y" + x2 y = O; y(O) = 0, y'(O) = -1 y'(O) = -I (sometimes calied an equid
ln Exercises 9 1-2,determine the
3. y" -xy' + 2y = O;y(O) = 0, y'(O) = 2 power series solution ..
to the differential equation. and analysis.
1) was a French mathemat1.c1an who studied algebra
4. y" - 2y' - xy = O; y(O) = 2, y'(O) = 0 9 . y" -xy = 0 3 Charles Hermite (1822-190 of the most m fluential Russian mathem
. .
aticians of all ume and is
10. y " + x2 y = 0 (182 1-1 894) as one . .
5. y" - 2y' + 2xy = 4- 2x; y(O) = 0. y'(O) = 0 4 Pafnuty Cheb yshev
. om1�a I approximations. number theory.and probab1hty
.
11. y" - xy' + 2y = 0 12. y" - 2y' + x2 y = 0 well known for his wor k •·n p0lyn
392 Chapter 8 Power Series Solutions to Linear Differential Equations 8.3 Euler Type Equations 393
= 2a2 + v(v + l)ao + ([v(v + 1) - 2]a1 + 6a3)x In Exr �t,tSes 1 3-16, determine the first three terms of the a) Determine the power series solution to this
powr:- ,l"ries soluti on to the differenti al equation. equation for Ix I < 1.
13. y" - y COS X = 0 b) Show that if a is a natural number, then there is
00
polynomial, use reduction of order to find the second a 1 x + a o and a number x0, determine ho, b1, ... , hn
v(v + 1) linearly independent solution. so that
a2 = - ao,
2 17. (l - x 2 )y" - 2xy' + 3/4y = 0 p(x) = ho +h1 (x -xo) +b2(x -xo)2 + ...
18. (l - x 2 )y " - 2xy' -1/4y = 0 + hn (X -xo)" + O(x -xo)"+t
a3 = 19. (t -x 2 )y" - 2xy' +2y = 0
+ O(x -xo)"+2 + · · ·
2-v(v+l)
a1,
6 20. (1 - x 2 )y" - 2xy' + 6y = 0
= bo +h1 (x -x0) + b2(x -xo)2 + ...
and
In Ex,·rc ises 21-22, determine the polynomial solution + hn (X -Xo)"
to thr ',egendre equation.
+ 2)(n + 3) - v(v + 1) for all x and hence conclude that every polynomial
an+4 = 21. (l - x 2 )y11 2xy' + 20y = 0 equal to its Taylor series a�ut xo for al I x and also
is
(n
Gn+2·
that its Taylor series has a finite number of terms.
-
The recurrence relation gives us two l i nearly i ndependent power series solutions to the 23. The equati on 26. a) Show that in Theorem8.4 ao = ko, a1 = ki,
Legendre equation as described i n Theorem 8.4. An interest i ng result occurs in the case ai = y"(0)/2 = -(q(O)k 1 + r(O)ko)/2.
y" -2x y' + AY = 0 b) Continuing the process in part (a), show that we
when
· known as the Hermite equa-
�-here J.. is a constant 1s 3 can generate a power series solution for
v = k+2 t:vn. This equation i s important i n physics and nu- Theorem 8.4 that satisfies y(O) = ko and
merical analysis. y'(O) = k1.
for some nonnegative integer k. Noti ce that, in this case, ak+4 = O, wh i ch then forces To complete the proof of Theorem 8.4 we need to
coefficients depending on ak+4, wh ich are the coeffic ients a) Show that the power series solution is � show this power series converges.
.
polynomial (called a Hermite polynomial) if>..
is an even natural number. In Exercises 27-3 0, use the dsolve command in Maple
b) Determine the Hermi te polynomials for with the type=series option or the corresponding
to be 0. Consequently one of the l i nearly independent power seri e s solutions in Theorem command in another appropriate soft ware pa.ckage to
J.. = o, 2,4 , 6, and 8.
8.4 is a polynomial Pk+2(x) of degree k + 2. (Why?) When the condition Pk+2( I) = I is obtain the first six terms of the power senes �lu-
imposed, Pk+2(x) is called the Legendre polynomial of degree k + 2. In this setting, it 24. The equati on .
non. Compare this answer to your results for Exemses
is sometimes easier to obtai n the two solutions by first finding the Legendre polynomial (1 - x 2)y" -xy' + cx zy = 0 13-16.
solution and then using the method of reduction of order introduced in Chapter 4 to find . 4 27. y" - ycosx = 0 28. y" + e7 )' = 0
the other solution. Exercises 17-20 ask you to apply thi s approach. where a is a constant 1s known as the Chebysbev
29. e7y" + xy = 0 30. y"-2y'+ysinx = 0
equation.
EXERCISES 8.2
EULER TYPE EQUATIONS
. solut1·ons about xo when xo is an ordinary
8.3
. power senes
In Exercises 1 8,- determi ne the power series solution to 6. y" + xy' - y = x; y(O) = 0, In the last section we cons1· dered
y'(O) = 0 series solutions when xo is not an ord'mary
the initial value problems. point. Our �ext objec tive is t on
7. y" -xy = I +x; y(O) = 1 , y'(O �:;�:�i�:;:;tial equation called an Euler type equati
1. y" - xy = 0; y(O) = I , y'(O) = 0 )=0 poi nt. In this study a type. o: homo -
8. y" + x 2 y = -2x; y(O) = 0, . •
h ns1ona
1me I equation) will arise. In the, second order
2. y" + x2 y = O; y(O) = 0, y'(O) = -1 y'(O) = -I (sometimes calied an equid
ln Exercises 9 1-2,determine the
3. y" -xy' + 2y = O;y(O) = 0, y'(O) = 2 power series solution ..
to the differential equation. and analysis.
1) was a French mathemat1.c1an who studied algebra
4. y" - 2y' - xy = O; y(O) = 2, y'(O) = 0 9 . y" -xy = 0 3 Charles Hermite (1822-190 of the most m fluential Russian mathem
. .
aticians of all ume and is
10. y " + x2 y = 0 (182 1-1 894) as one . .
5. y" - 2y' + 2xy = 4- 2x; y(O) = 0. y'(O) = 0 4 Pafnuty Cheb yshev
. om1�a I approximations. number theory.and probab1hty
.
11. y" - xy' + 2y = 0 12. y" - 2y' + x2 y = 0 well known for his wor k •·n p0lyn
394 Chapte r 8 Power Series Solutions to Linear Differential Equations 8.3 Euler Type Equations 395
geneous case, these Euler type equations are differen tial equati o ns that can be written in This is a polynomial equation in r similar t o the characteristic equatio ns we had for
the form constant coefficient linear differential equations in Chapter 4. We will call this e quation
I,= 1
the Euler indicial equation. Solving this equatio n for r gives us
I (x -xo)2y" + a(x -xo)y' + fJy =0 I I -a± J(�-a)'-4P·
wher e a and f3 are constan ts. W hen we write one of these Euler type equatio n s in the
form
\ As with c onstant coefficient lin ear differential equations, we have to consider distinct
y"+ q(x)y' + r(x)y = 0, real, repeated real, and imaginary root cases. Here these respective cases arise as follows:
extend either of these two methods to cases when x0 -I 0. In Case 2 we have one real root, r, and hence only one soluti o n,
C onsider the Euler type equatio n x',
I x2 y" +axy' + {Jy =0.1 when .:c > 0. ln Exercise 16 we will ask you to obtain that
We will determine solution s for x > 0. It is als o possible to determine s olutions for �
x < 0. See E xercise 21 for details about this. Since differentiating reduces the p ower
. f x'. Consequen tly, the general s olution for
on the exponent by one, it makes sense to try a s olution of the form 1s a second soIut·10n 11·nearly in d ependent o
x > O is
= xr .
Iy 1
Substituting this into th e differential equati o n gives us
in Case 2.
x2 r(r - l)x r-2 +axrxr -l + fJxr = O ex r ots, r =a± ib. As in Chapters 4. and 6, we will
In C ase 3 we have two comp1 o nt . I be
ons a n d w1I
or
- a+ ib to produce two linearly independe soluti
only need the root r--:
- a _ ib If x >
root, r _ Owe hav
able t o ignore the conJugate
e
geneous case, these Euler type equations are differen tial equati o ns that can be written in This is a polynomial equation in r similar t o the characteristic equatio ns we had for
the form constant coefficient linear differential equations in Chapter 4. We will call this e quation
I,= 1
the Euler indicial equation. Solving this equatio n for r gives us
I (x -xo)2y" + a(x -xo)y' + fJy =0 I I -a± J(�-a)'-4P·
wher e a and f3 are constan ts. W hen we write one of these Euler type equatio n s in the
form
\ As with c onstant coefficient lin ear differential equations, we have to consider distinct
y"+ q(x)y' + r(x)y = 0, real, repeated real, and imaginary root cases. Here these respective cases arise as follows:
extend either of these two methods to cases when x0 -I 0. In Case 2 we have one real root, r, and hence only one soluti o n,
C onsider the Euler type equatio n x',
I x2 y" +axy' + {Jy =0.1 when .:c > 0. ln Exercise 16 we will ask you to obtain that
We will determine solution s for x > 0. It is als o possible to determine s olutions for �
x < 0. See E xercise 21 for details about this. Since differentiating reduces the p ower
. f x'. Consequen tly, the general s olution for
on the exponent by one, it makes sense to try a s olution of the form 1s a second soIut·10n 11·nearly in d ependent o
x > O is
= xr .
Iy 1
Substituting this into th e differential equati o n gives us
in Case 2.
x2 r(r - l)x r-2 +axrxr -l + fJxr = O ex r ots, r =a± ib. As in Chapters 4. and 6, we will
In C ase 3 we have two comp1 o nt . I be
ons a n d w1I
or
- a+ ib to produce two linearly independe soluti
only need the root r--:
- a _ ib If x >
root, r _ Owe hav
able t o ignore the conJugate
e
equation to obtain its general solution. 22. a) Let L be the linear transformation
which hasr = -1 as a repeated root. The general solution is
13. (x - !)2 y" + (5x - 5)y' + 3y = 0 L = x 2 D 2 + ax D + /J so that kernel of L
c1
y=-+
c2 lnx 14. (x + 2)2 y" - 2(x + 2)y'2y = 0 consists of the solutions to the Euler type
--.
X X equationx2y" + axy' + tJy = 0. Further. let
Exerci,c� 15-17 deal with the three cases for the roots F(r) = r(r - I)+ ar + /J. v. hich is the
Using the initial conditions, we obtain of the Euler indicial equation. polynomial in the Euler indicial equation ofth1�
15. Show that in Case 1 the solutions are linearly inde Euler type equation. Show that
c, = 0, C2 = l.
pendent for x > 0.
The solution t o this initial value problem is then L(x' In x) = F(r)x' In x + F'(r):<'.
16. Use the reduction of order method in Section 4.2
lnx
y=-. • to find a second linearly independent solution for
x > 0 in Case 2.
17. Show that the two solutions we gave in Case 3 are
b) Suppose that the indicial equation ha, a
repeated root ,1 so that F (r) = (r - r,)2 . U�e
the result of part (a) to show that
L (x'' In x) = 0 and hence conclude that x:• In x
solutions of the Euler type equation and are linearly
EXAMPLE3 Determine the general solution of x2 y" + 2.xy' + y = O. is another solution to the Euler type equauon.
independent.
18. Determi ne the behavior of the solutions to th_e Euler
Solution The equation with r for this example is type equations as x --+ 0 if the real parts ofri and
r2 are positive.
r (r - 1 ) + 2, + 1 = 0.
POINT
NEAR A REGULAR SINGULAR
Its solutions are SERIES SOLUTIONS
./3
8.4 . . reason we studied the Euler type
introduction of Section 8.3 one
l As mentioned in the
r = -- ±-i equations
2 2 .
+fir = 0
The general solution is (x -xo )2 Y" +a (x -xo)J"
equation to obtain its general solution. 22. a) Let L be the linear transformation
which hasr = -1 as a repeated root. The general solution is
13. (x - !)2 y" + (5x - 5)y' + 3y = 0 L = x 2 D 2 + ax D + /J so that kernel of L
c1
y=-+
c2 lnx 14. (x + 2)2 y" - 2(x + 2)y'2y = 0 consists of the solutions to the Euler type
--.
X X equationx2y" + axy' + tJy = 0. Further. let
Exerci,c� 15-17 deal with the three cases for the roots F(r) = r(r - I)+ ar + /J. v. hich is the
Using the initial conditions, we obtain of the Euler indicial equation. polynomial in the Euler indicial equation ofth1�
15. Show that in Case 1 the solutions are linearly inde Euler type equation. Show that
c, = 0, C2 = l.
pendent for x > 0.
The solution t o this initial value problem is then L(x' In x) = F(r)x' In x + F'(r):<'.
16. Use the reduction of order method in Section 4.2
lnx
y=-. • to find a second linearly independent solution for
x > 0 in Case 2.
17. Show that the two solutions we gave in Case 3 are
b) Suppose that the indicial equation ha, a
repeated root ,1 so that F (r) = (r - r,)2 . U�e
the result of part (a) to show that
L (x'' In x) = 0 and hence conclude that x:• In x
solutions of the Euler type equation and are linearly
EXAMPLE3 Determine the general solution of x2 y" + 2.xy' + y = O. is another solution to the Euler type equauon.
independent.
18. Determi ne the behavior of the solutions to th_e Euler
Solution The equation with r for this example is type equations as x --+ 0 if the real parts ofri and
r2 are positive.
r (r - 1 ) + 2, + 1 = 0.
POINT
NEAR A REGULAR SINGULAR
Its solutions are SERIES SOLUTIONS
./3
8.4 . . reason we studied the Euler type
introduction of Section 8.3 one
l As mentioned in the
r = -- ±-i equations
2 2 .
+fir = 0
The general solution is (x -xo )2 Y" +a (x -xo)J"
where a and f3 are constants was to prepare us for the study of power series solutions to EXAMPLE 1 ff possible, determine solutions to the differential equation
differential equations of the type 5 2 +x
+ - y =0
II I
I I
11
11=0 11=0
y" + p(x)y' + q(x)y = 0
Solution Note that x = o is a regular singular point for this differential equation. Indeed, Equa
where
tion (I) is
(x - xo)p(x) and (x -xo)2 q(x) 5 5
xp(x) = 2 = 2 + 0 . x + 0. x 2 + ... ,
are analytic at xo. When this occurs, x o is called a regular singular point of the
and Equation (2) is
differential equation y" + p(x)y'+q(x)y = 0. Notice that the Euler type equation
2+x 1
x 2q(x)=---=-I- x+O · X2 + O ·X 3 + ....
y"+ --y
a '+ /3 2 y= 0 2 2
x- xo ( x - x0} multipl y our di�e.rential. equatio� by x before
first
2
It will turn out to be conveni ent ifwe
. becomes
has a regular singular point at xo since ( x - x 0)p(x} = a and (x - x 0)2 q(x) = /3 are we attempt to determme series solutions. Doing so, our d1tferent1al equation
analytic at xo. Power series solutions about x0 toy"+p(x)y' + q(x)y = 0 when this 5 2+X
differential equation has a regular singular point at x0 are related to the solutions of x 2y " + -xy - --y = 0. 1
(3)
2 2
Euler type equations. As we did for Euler type equations in the previous section, we
For our desired solution Y we have
will restrict our attention to the case when xo = 0 and our solutions will be for x > 0.
The method we develop here can be easily modified to handle nonzero values for xo and n+r -l
y, = "'""'
Lan(n+ )r x
other values of x.
Suppose that xo = 0 is a regular singular point of y" + p(x)y' + q(x)y = 0. Then
n=O
xp(x) and x2 q(x) are equal to their Maclaurin series in some open interval !xi < R. To and
put it another way, we can express xp(x) and x2 q(x) as 00
n+r-2
y" = [ a n(n+ r )(n+r- ]) x
L p x"
00
xp(x) =Po+ pix+ JJ2X 2 + p3x3 + · · · = (1)
n=O
. . . to Equat'1on (3) oives us
,,
11=0 Substttutmg m O
and oo 5
00
n+r-L
x 1
"'""'-a n+r+l = 0.
L n
00
" + r
) x -�� a,, nx
(n
+r
( n+)(
r n+ r
- l )x +'+[ - a .t
2
a11 2
n
L q x"
00
n=O n=O
x2 q(x ) = qo+qix+q2x2+q3x3+· · · = ,, (2) 11=0
n=O
xp(x) = a, x2 q(x) = /3, and the solutions xr are x' times the Maclaurin series of the
11=0 le
constant function f(x) =I.Could it be that solutions toy" + p(x)y' + q(x)y =Oar e first sum and reindexing. we can rewn
.·
Ierm .
with - 0 m
- . the
of the form x r times the Maclaurin series of some function? That is, are there solutio ns Separating o ff the
11
where a and f3 are constants was to prepare us for the study of power series solutions to EXAMPLE 1 ff possible, determine solutions to the differential equation
differential equations of the type 5 2 +x
+ - y =0
II I
I I
11
11=0 11=0
y" + p(x)y' + q(x)y = 0
Solution Note that x = o is a regular singular point for this differential equation. Indeed, Equa
where
tion (I) is
(x - xo)p(x) and (x -xo)2 q(x) 5 5
xp(x) = 2 = 2 + 0 . x + 0. x 2 + ... ,
are analytic at xo. When this occurs, x o is called a regular singular point of the
and Equation (2) is
differential equation y" + p(x)y'+q(x)y = 0. Notice that the Euler type equation
2+x 1
x 2q(x)=---=-I- x+O · X2 + O ·X 3 + ....
y"+ --y
a '+ /3 2 y= 0 2 2
x- xo ( x - x0} multipl y our di�e.rential. equatio� by x before
first
2
It will turn out to be conveni ent ifwe
. becomes
has a regular singular point at xo since ( x - x 0)p(x} = a and (x - x 0)2 q(x) = /3 are we attempt to determme series solutions. Doing so, our d1tferent1al equation
analytic at xo. Power series solutions about x0 toy"+p(x)y' + q(x)y = 0 when this 5 2+X
differential equation has a regular singular point at x0 are related to the solutions of x 2y " + -xy - --y = 0. 1
(3)
2 2
Euler type equations. As we did for Euler type equations in the previous section, we
For our desired solution Y we have
will restrict our attention to the case when xo = 0 and our solutions will be for x > 0.
The method we develop here can be easily modified to handle nonzero values for xo and n+r -l
y, = "'""'
Lan(n+ )r x
other values of x.
Suppose that xo = 0 is a regular singular point of y" + p(x)y' + q(x)y = 0. Then
n=O
xp(x) and x2 q(x) are equal to their Maclaurin series in some open interval !xi < R. To and
put it another way, we can express xp(x) and x2 q(x) as 00
n+r-2
y" = [ a n(n+ r )(n+r- ]) x
L p x"
00
xp(x) =Po+ pix+ JJ2X 2 + p3x3 + · · · = (1)
n=O
. . . to Equat'1on (3) oives us
,,
11=0 Substttutmg m O
and oo 5
00
n+r-L
x 1
"'""'-a n+r+l = 0.
L n
00
" + r
) x -�� a,, nx
(n
+r
( n+)(
r n+ r
- l )x +'+[ - a .t
2
a11 2
n
L q x"
00
n=O n=O
x2 q(x ) = qo+qix+q2x2+q3x3+· · · = ,, (2) 11=0
n=O
xp(x) = a, x2 q(x) = /3, and the solutions xr are x' times the Maclaurin series of the
11=0 le
constant function f(x) =I.Could it be that solutions toy" + p(x)y' + q(x)y =Oar e first sum and reindexing. we can rewn
.·
Ierm .
with - 0 m
- . the
of the form x r times the Maclaurin series of some function? That is, are there solutio ns Separating o ff the
11
for X > 0. . . .
Let us now apply the procedure of Example I to a general differential equation
11+ - 2n2 -n -3
a ,----- n=<I n=<I
400 Chapter 8 Power Series Solutions to Linear Differential Equations 8.4 Series Solutions Near a Regular Singular Point 401
for X > 0. . . .
Let us now apply the procedure of Example I to a general differential equation
11+ - 2n2 -n -3
a ,----- n=<I n=<I
402 Chapter 8 Power Series Solutions to Linear Differential Equations 8.4 Series Solutions Near a Regular Singular Point 403
2 3 F (r) = 0
00
THEOREM 8.7 Suppose that x0 = 0 is a regular singular point of the d i fferential equation
F(r) = r(r - 1) + por +qo
y" + p(x)y' + q(x)y = 0
and ,1 and ,2 are the roots of its indicial equation with r1 > r2. If r1 - r2 is not a positive
integer, then this differential equation has a solution of the form
to save writing. Indeed, with this notation, observe that Equation (7) now becomes:
aoF(r)x' + (a1 F(l + r) + ao(p1r + q1))x'+l
2
Yl = x'
1
L Gn (ri)x
n
+ (a3 F(3 + r) + a2(P1 (2 + r)+qi)+a1(p2(1 + r) + q2) and a second solution of the form
+ ao(p3r + q3))x'+3 + · · · = 0.
We then have:
y2 = x '2 L a (r2)x .
n==O
n
n
F(r) = 0
a1 F(l + r) + ao(Pir +qi)= 0 it illustrates Theorem 8.7 with r1 = 1/2
Looking back at Example I, notice that
a2 F(2 + r) + a1(pi (1+r)+qi)+ ao(p2r + q2) = 0 andr2 = -2.
rem 8.7 converge to a solut 10n for O < x < R
It can be argued that y1 and y22 in Theo
+ a1(p2(1 + r) + q2) + ao(p3r +q3) = 0 in Equa tions (I) and (2) converge to xp(x) and
a3 F{3+ r ) + a2(p1(2 + r )+ qi) when the series for xp(x) and x q (x)
al. Furth er, if r, are r2 are real numbers and we choose
x 2q (x) , respec11·vely, on this interv
and y2 are linearly .indepen dent on th.1s mterva
· 1
nonzero vaIues of ao(r1) and a0(r2) y1
402 Chapter 8 Power Series Solutions to Linear Differential Equations 8.4 Series Solutions Near a Regular Singular Point 403
2 3 F (r) = 0
00
THEOREM 8.7 Suppose that x0 = 0 is a regular singular point of the d i fferential equation
F(r) = r(r - 1) + por +qo
y" + p(x)y' + q(x)y = 0
and ,1 and ,2 are the roots of its indicial equation with r1 > r2. If r1 - r2 is not a positive
integer, then this differential equation has a solution of the form
to save writing. Indeed, with this notation, observe that Equation (7) now becomes:
aoF(r)x' + (a1 F(l + r) + ao(p1r + q1))x'+l
2
Yl = x'
1
L Gn (ri)x
n
+ (a3 F(3 + r) + a2(P1 (2 + r)+qi)+a1(p2(1 + r) + q2) and a second solution of the form
+ ao(p3r + q3))x'+3 + · · · = 0.
We then have:
y2 = x '2 L a (r2)x .
n==O
n
n
F(r) = 0
a1 F(l + r) + ao(Pir +qi)= 0 it illustrates Theorem 8.7 with r1 = 1/2
Looking back at Example I, notice that
a2 F(2 + r) + a1(pi (1+r)+qi)+ ao(p2r + q2) = 0 andr2 = -2.
rem 8.7 converge to a solut 10n for O < x < R
It can be argued that y1 and y22 in Theo
+ a1(p2(1 + r) + q2) + ao(p3r +q3) = 0 in Equa tions (I) and (2) converge to xp(x) and
a3 F{3+ r ) + a2(p1(2 + r )+ qi) when the series for xp(x) and x q (x)
al. Furth er, if r, are r2 are real numbers and we choose
x 2q (x) , respec11·vely, on this interv
and y2 are linearly .indepen dent on th.1s mterva
· 1
nonzero vaIues of ao(r1) and a0(r2) y1
404 Chapter 8 Power Series Solutions to Linear Differential Equations 8.4 Series Solutions Near a Regular Singular Point 405
0 < x < R and consequently the general solution toy"+ p(x)y' + q(x)y = 0 is and there is a constant a so that it has a second solution of the form
C1Y1+ C2Y2 on this interval. We will only use this theorem in the case when the indicial
equation has real roots. The complex case as well as a complete proof of Theorem
00
YI = x' 1
x y" + 2xy' + (x - x 3 )y = 0.
2
11=0
and give the forms of the power series solutions for this case.
and a second solution of the form
When written in the form y" + p(x)y' + q(x)y = 0, this differential equation i\
L b.x".
00 Solution
Yi = Y1 ln x + x'• I -x 2
n=O y
11
+ -2 + --y = 0.
X X
lf r 1 - r2 is a positive integer, our process does produce a power series solution for We have
the larger root, but breaks down for the smaller root, r2• Here again, can we produce xp(x) = 2 and x 2q(x) = x - x3
a second solution? Frobenius found a positive answer in this case too, as described in
Theorem 8.9. so that Po = 2 and q0 = 0. The indicial equation is then
00
0 < x < R and consequently the general solution toy"+ p(x)y' + q(x)y = 0 is and there is a constant a so that it has a second solution of the form
C1Y1+ C2Y2 on this interval. We will only use this theorem in the case when the indicial
equation has real roots. The complex case as well as a complete proof of Theorem
00
YI = x' 1
x y" + 2xy' + (x - x 3 )y = 0.
2
11=0
and give the forms of the power series solutions for this case.
and a second solution of the form
When written in the form y" + p(x)y' + q(x)y = 0, this differential equation i\
L b.x".
00 Solution
Yi = Y1 ln x + x'• I -x 2
n=O y
11
+ -2 + --y = 0.
X X
lf r 1 - r2 is a positive integer, our process does produce a power series solution for We have
the larger root, but breaks down for the smaller root, r2• Here again, can we produce xp(x) = 2 and x 2q(x) = x - x3
a second solution? Frobenius found a positive answer in this case too, as described in
Theorem 8.9. so that Po = 2 and q0 = 0. The indicial equation is then
00
We conclude this section with a type of differential equation that has a regular sin
gular point at x0 = 0 that arises in many physical applications called a Bessel equation.6
These differential equations have the form
r(r - 1) + r - v2 = r 2 -v2 = 0.
The roots of the indicial equation are
I '1 = v. r2 = -v. j
Observe that we are then in the case of Theorem 8.7 if 2 v is not a nonnegative integer,
in the case of Theorem 8.8 if v = 0, and in the case of Theorem 8.9 if 2 v is a fX>Sitive
integer.
Our final set of examples invol ve power series solutions to Bessel equations in some
special cases.
EXAMPLE 3 Determine the series solutions to the following Bessel equation: the Bessel equation of
order 0,
Solution Here v = 0 and we are in the case of Theorem 8.8 with r = 0. Let us find the solution
of the form
00 00
6 Friedrich Wilhelm Bessel (1784-1846) was a German astronomer and malhematician who served as the
director of the Konigsberg observatory.
8.4 Series Solutions Near a Regular Singular Point
(n + 2)2
so that
a2
04=--=-- a0
42 24(2!)2
04 ao
06=--=- --
62 26(3!)2
6 a a
as=--=- -0
g2 28(4!)2
It appears that
(-J)"ao
a2 = ---2
" 22"(n!) .
We will let you verify that this is lrue in general in Exercise 9(a).
The first solution in Theorem 8.8 is then
_ � (-l)"ao 211
.Yt - L., 2 2n (n !)2 X
11=0
� (-1) " n
1,o (x) = L., 22n(n!)2 x 2
11=0
called the Bessel function of the first kind of order O. We leave it for you to find lhe
second solution, y2, in Theorem 8.8 in Exercise 9(b). •
EXAMPLE 4 Determine the series solutions to the following Bessel equations: the Bessel equation of
order I /2,
x 2y " +xy' + (x 2 - i) Y = 0.
Solution Here v = J /2, the roots of the indicial equation are r, = I /2 and r2 = - l /2 . We arc
then in the case of Theorem 8.9. We again find the first solution
- Substituting, we have
408 Chapter 8 Power Series Solutions to Linear Differential Equations 8.4 Series Solutions Near a Regular Singular Point 409
EXAMPLE 5 Determine the series solutions to the following Bessel equation: the Bessel equation of
order 1,
Substituting it in,
4
.�
00 00 �
1 1
[ n(n + l)a11 x•+ + [(11 + l)a,,x"+1 + (x - I) [a,, x "+
2
Thus, � �
a1 = 0 = a3 =as = a1 =··· = 3a1x 2 + L([(,r + 3) 2 - 1 Ja,,+ 2 + Gn)Xn+J = 0
11=0
and
from which we obtain
ao ao
a2 = -- = -- CIJ = 0
6 3!
a2 ao
a4=-- =- and
20 5!
a,,
a4 ao a,,+2 =
) -- =--
Cl(:= (n + 3) 2 - 1 ·
42 7!
We have
a1 =0=a3 = as = 01 = ···
The second set of equations suggest that
(-1 )"a
o
a2" ---
- and
(2n + !) !. a ao
We will let you prove that this in fact is the case in Exercise 9(c). We then have
a2 = -8o = 2 2 (2!)(1!)
Y1 = xl/2
f ( -l)"ao x2,'
(2n + I)!
= aox-112 f (-1)" 2n 1
x + = a o x-111 sinx.
4
a2
G = - 24
= 4
ao
2 (3!)(2!)
11=0 11=0 (2n + l)!
( 'l
1/2
l1;2(x) = ;x ) sinx
. that the second set of equations generalizes to
We w11l let you ven'fy .m Exercise 9(e)
is called the Bessel function of the first kind of order l/2. We will let you determine (-I tao
the second solution, .Y2, in Exercise 9(d). • ai,, = 2n n I !11!
2 ( + ) ·
408 Chapter 8 Power Series Solutions to Linear Differential Equations 8.4 Series Solutions Near a Regular Singular Point 409
EXAMPLE 5 Determine the series solutions to the following Bessel equation: the Bessel equation of
order 1,
Substituting it in,
4
.�
00 00 �
1 1
[ n(n + l)a11 x•+ + [(11 + l)a,,x"+1 + (x - I) [a,, x "+
2
Thus, � �
a1 = 0 = a3 =as = a1 =··· = 3a1x 2 + L([(,r + 3) 2 - 1 Ja,,+ 2 + Gn)Xn+J = 0
11=0
and
from which we obtain
ao ao
a2 = -- = -- CIJ = 0
6 3!
a2 ao
a4=-- =- and
20 5!
a,,
a4 ao a,,+2 =
) -- =--
Cl(:= (n + 3) 2 - 1 ·
42 7!
We have
a1 =0=a3 = as = 01 = ···
The second set of equations suggest that
(-1 )"a
o
a2" ---
- and
(2n + !) !. a ao
We will let you prove that this in fact is the case in Exercise 9(c). We then have
a2 = -8o = 2 2 (2!)(1!)
Y1 = xl/2
f ( -l)"ao x2,'
(2n + I)!
= aox-112 f (-1)" 2n 1
x + = a o x-111 sinx.
4
a2
G = - 24
= 4
ao
2 (3!)(2!)
11=0 11=0 (2n + l)!
( 'l
1/2
l1;2(x) = ;x ) sinx
. that the second set of equations generalizes to
We w11l let you ven'fy .m Exercise 9(e)
is called the Bessel function of the first kind of order l/2. We will let you determine (-I tao
the second solution, .Y2, in Exercise 9(d). • ai,, = 2n n I !11!
2 ( + ) ·
8.4 Series Solutions Near a Regular S ingular Point 411
410 Chapter 8 Power Series Solutions to Linear Differential Equations
b) Use part (a) to determine the two linearly a) Show x0 = 0 and x0 = I are regular singular
Thus
independent solutions to the Bessel equation of points of this differential equation.
� (-l)"ao � (-lt order 1 /2. (Hint: User= -1/2.) b) Determine the indicial equation for the regular
YI = x L., 22"(n + l)!n! x = ao L., 22•(n + J)!n! x
2n 2n+I
.
c) Using the result of part(b), solve the singular point x0 = 0.
11=0 11=0
non homogeneous equation c) Determine the recurrence relations at the
When ao = I /2, we obtain the solution x 2y" + xy' + (x 2 -1/4)y = x3f2 . regular singular point xo = 0.
14. Use Maple or another appropriate software package
f1 (x) -
i
- -
00
I: <-o" X 211+1 to obtain a second solution to the Bessel equation in
d) Use the result of part (c) and Maple or another
appropriate software package to determine two
2 n=O
2 211 (n + l)!n! Example 5. linearly independent solutions at the regular
In Exercises 15-18 determine one solution to the differ singular point x0 = 0 for o: = 3/2.
called the Bessel function of the first kind oforder 1. Finding the second solution, Y2,
is best done by using a software package such as Maple. We will ask you to do this in ential equation at x0 = 0 an<l use Maple to determine a 20. Consider the differential equation
Exercise 14. second linearly independent solution at xo = 0.
x2y " + COS.X)' = 0.
15 x 2 y" - x(2 + x)y' + (2 + x2 )y = 0
a) Show x0 = 0 is a regular singular point of thi�
16. x(x + l)y" + y' +xy = 0
EXERCISES 8.4 differential equation.
17. x2 y " + xy ' + (x2 - 4)y = 0
b) Determine the indicial equation of thb
18. x 2 y " + xy ' + (x 2 - 9)y = 0 differential equation.
In Exercises 1-2, determine the regular singular points d) Determine the second linearly independent
for the differential equation. solution for the Bessel equation of order 1/2 in 19 fhe differential equation
c) Determine the first four terms of one of the
I. x 2 y" + sin x y' + y = 0 Example 4. x(l - x)y " + (o: -(1 + f3 + y)x)y ' - f3y y = 0 solutions of this differential equation
e) Show that guaranteed by Theorems 8.7-8.9.
2. (I - x2)y" - 2xy' + 20y = 0 occurs often in mathematical physics.
In Exercises 3-4, determine the indicial equation for the (-l)"ao
differential equation at x0 = 0. 2 211 (n + l)!n!
a2n -
-
---
8.4 Series Solutions Near a Regular S ingular Point 411
410 Chapter 8 Power Series Solutions to Linear Differential Equations
b) Use part (a) to determine the two linearly a) Show x0 = 0 and x0 = I are regular singular
Thus
independent solutions to the Bessel equation of points of this differential equation.
� (-l)"ao � (-lt order 1 /2. (Hint: User= -1/2.) b) Determine the indicial equation for the regular
YI = x L., 22"(n + l)!n! x = ao L., 22•(n + J)!n! x
2n 2n+I
.
c) Using the result of part(b), solve the singular point x0 = 0.
11=0 11=0
non homogeneous equation c) Determine the recurrence relations at the
When ao = I /2, we obtain the solution x 2y" + xy' + (x 2 -1/4)y = x3f2 . regular singular point xo = 0.
14. Use Maple or another appropriate software package
f1 (x) -
i
- -
00
I: <-o" X 211+1 to obtain a second solution to the Bessel equation in
d) Use the result of part (c) and Maple or another
appropriate software package to determine two
2 n=O
2 211 (n + l)!n! Example 5. linearly independent solutions at the regular
In Exercises 15-18 determine one solution to the differ singular point x0 = 0 for o: = 3/2.
called the Bessel function of the first kind oforder 1. Finding the second solution, Y2,
is best done by using a software package such as Maple. We will ask you to do this in ential equation at x0 = 0 an<l use Maple to determine a 20. Consider the differential equation
Exercise 14. second linearly independent solution at xo = 0.
x2y " + COS.X)' = 0.
15 x 2 y" - x(2 + x)y' + (2 + x2 )y = 0
a) Show x0 = 0 is a regular singular point of thi�
16. x(x + l)y" + y' +xy = 0
EXERCISES 8.4 differential equation.
17. x2 y " + xy ' + (x2 - 4)y = 0
b) Determine the indicial equation of thb
18. x 2 y " + xy ' + (x 2 - 9)y = 0 differential equation.
In Exercises 1-2, determine the regular singular points d) Determine the second linearly independent
for the differential equation. solution for the Bessel equation of order 1/2 in 19 fhe differential equation
c) Determine the first four terms of one of the
I. x 2 y" + sin x y' + y = 0 Example 4. x(l - x)y " + (o: -(1 + f3 + y)x)y ' - f3y y = 0 solutions of this differential equation
e) Show that guaranteed by Theorems 8.7-8.9.
2. (I - x2)y" - 2xy' + 20y = 0 occurs often in mathematical physics.
In Exercises 3-4, determine the indicial equation for the (-l)"ao
differential equation at x0 = 0. 2 211 (n + l)!n!
a2n -
-
---
Inner Product Spaces
ACES
9 .1 INNER PRODUCT SP dot product
tors. the
writing vectors in R" as column vec
Following our convention of
vect ors
J
( or scalar product) of two
u= [
01
a2 J = bi
anti v [
b 2
in JR2, denoted u · v, is
414 Chap,tcr 9 Inner Product Spaces 9.1 Inner Product Spaces 415
l
If u and v are vectors in JR3 , Notice that we can equally well describe the length of a vector v in JR2 or JR3 as
II · V = UT V. Also notice that the vectors u and v are orthogonal (or perpendicular) if a nd only if
U • V = 0.
llvll =}bf+ h�. The dot product on JR2 or JR 3 is a function fr om the set of pairs of vector<. in 1R2 or
Of course, if we follow the usual convention of i ndicating such a vector v graphically by JR3 to the set of real n umbers JR. Some easily verified properties of it are
drawing a directed line segment from the origin to the point (b 1, b 2), as in Figure 9.1,
= V • II
then II v II is the length of this directed line segment. If v is a vector in JR3 ,
U • V
l
(u + v) · w = 11 • w + v · w
(cu)· v = c(u · v)
v=
.
v v = 0 if and only if v = 0
[ �
where c is a scalar. Any vector space possessing such a fu nction satisfying these prop
then its length (magnitude, or no1m) is e11ies is called an inner product space.
l
If u and v are vectors in JR3 , Notice that we can equally well describe the length of a vector v in JR2 or JR3 as
II · V = UT V. Also notice that the vectors u and v are orthogonal (or perpendicular) if a nd only if
U • V = 0.
llvll =}bf+ h�. The dot product on JR2 or JR 3 is a function fr om the set of pairs of vector<. in 1R2 or
Of course, if we follow the usual convention of i ndicating such a vector v graphically by JR3 to the set of real n umbers JR. Some easily verified properties of it are
drawing a directed line segment from the origin to the point (b 1, b 2), as in Figure 9.1,
= V • II
then II v II is the length of this directed line segment. If v is a vector in JR3 ,
U • V
l
(u + v) · w = 11 • w + v · w
(cu)· v = c(u · v)
v=
.
v v = 0 if and only if v = 0
[ �
where c is a scalar. Any vector space possessing such a fu nction satisfying these prop
then its length (magnitude, or no1m) is e11ies is called an inner product space.
b
(!, g) = l J(x)g(x)d.t.
3. (cu, u) = �(u, v) for all u and v i n Vand all scalars c.
4. (v, v) = 0 1f and only if v = O. We lea�e it to y ou t� verify that the four required properties for this to be an inner product
are.satisfied (Exerc ise 7). Hence C[a, h] with this inner product involving integ ration is
an mner product space. •
T�e function in this definition assigning t o the pair of vector s u and v the V3Iue Other examples of some inner pro ducts you may encounter in futu re co urses can be
( u, v) is called an inner product on V. We read (u' v) as "the inner . product of u and found in Exer cises 8, 9, JI, 12, and 13.
.
v." The dot pr o ducts on JR.2 and JR3 give us two examples of mner product spaces (where The next two theorems contain so me basic p roperties of inner products.
( u, v ) - u. v). Thts. . naturally extends to higher dimensions· • whi ch we make as our next
example. If u, v, and w are vectors in an inner pro duc t space V and c is a scalar, then:
THEOREM 9.1
EXAMPLE 1 For vectors 1. (0, v) = (v, 0) = 0.
2. (v,u+w} = (v,u)+(v, w }.
3. (v, cu) = c(v, 11).
4. (v - u, w) =(v, w) - (u.w}.
5. (v, u - w) = (v,u) - (v, w).
r ]•
T
(u +v,w) (u +vl w (uT +v ) w -
= = - urw + v T w = (u, ,, '
w)+ (v, w),
(cu,v) = (cu)T v = c(uTv) = c(u, v), of thi s theorem:
(w, v) = (v, u)+(u, w). •
and (v, u + w) = (u + w. v) = (u, v}+
product, or Euclidean inner product o n �,, 'and (u, v) is usually denot hat we earlier noted that the length
Recall space�:
define lengths of vectors in inner product
t
ed as u . v as is
done in JR2 and JR3 .;v:v. We use th is same approach to
space, we defi ne the length (magn itude, or norm) of a \ector
Dot products are not the only type of in • ner products. Here is a nother one that arises If v is an inner p ro duct
frequently in mathematic s and ·,t·s ap·p1.1cat1ons. v in V, deno ted llvJI, to be
EXAMPLE2 Rec all that the set of c ontinuous functi o ns o n. a cl osed mte
. r val [a'b], denoted Cla, b]' Uvll = J{v:0.
is a vector space. For two func tions f a n d g m C[a,b], define
9.1 Inner Product Spaces 417
416 Chapter 9 Inner Product Spaces
b
(!, g) = l J(x)g(x)d.t.
3. (cu, u) = �(u, v) for all u and v i n Vand all scalars c.
4. (v, v) = 0 1f and only if v = O. We lea�e it to y ou t� verify that the four required properties for this to be an inner product
are.satisfied (Exerc ise 7). Hence C[a, h] with this inner product involving integ ration is
an mner product space. •
T�e function in this definition assigning t o the pair of vector s u and v the V3Iue Other examples of some inner pro ducts you may encounter in futu re co urses can be
( u, v) is called an inner product on V. We read (u' v) as "the inner . product of u and found in Exer cises 8, 9, JI, 12, and 13.
.
v." The dot pr o ducts on JR.2 and JR3 give us two examples of mner product spaces (where The next two theorems contain so me basic p roperties of inner products.
( u, v ) - u. v). Thts. . naturally extends to higher dimensions· • whi ch we make as our next
example. If u, v, and w are vectors in an inner pro duc t space V and c is a scalar, then:
THEOREM 9.1
EXAMPLE 1 For vectors 1. (0, v) = (v, 0) = 0.
2. (v,u+w} = (v,u)+(v, w }.
3. (v, cu) = c(v, 11).
4. (v - u, w) =(v, w) - (u.w}.
5. (v, u - w) = (v,u) - (v, w).
r ]•
T
(u +v,w) (u +vl w (uT +v ) w -
= = - urw + v T w = (u, ,, '
w)+ (v, w),
(cu,v) = (cu)T v = c(uTv) = c(u, v), of thi s theorem:
(w, v) = (v, u)+(u, w). •
and (v, u + w) = (u + w. v) = (u, v}+
product, or Euclidean inner product o n �,, 'and (u, v) is usually denot hat we earlier noted that the length
Recall space�:
define lengths of vectors in inner product
t
ed as u . v as is
done in JR2 and JR3 .;v:v. We use th is same approach to
space, we defi ne the length (magn itude, or norm) of a \ector
Dot products are not the only type of in • ner products. Here is a nother one that arises If v is an inner p ro duct
frequently in mathematic s and ·,t·s ap·p1.1cat1ons. v in V, deno ted llvJI, to be
EXAMPLE2 Rec all that the set of c ontinuous functi o ns o n. a cl osed mte
. r val [a'b], denoted Cla, b]' Uvll = J{v:0.
is a vector space. For two func tions f a n d g m C[a,b], define
418 Chapter 9 Inner Product Spaces 9.1 Inner Product Spaces 419
In JR2 or JR 3 where v . u = II v II II u II cos e, we have that The next notion about inner product spaces we introduce in this section is that of
the angle between two vectors v and u in an inner product space V. If either vor II is
\v · u\ = \\\v\\ llull cos()\= \\v\l llu\\\ cos()\ S \\vii llu\l
the zero vector, we take the angle () between v and II to be zero just as we do in iR2 or
since I cos 81 s I. Remarkably, this inequalily, called the Cauchy-Schwartz inequality, 2 IR3 . Suppose that v and u are both nonzero. Recall that if v and u arc in R2 or :R.\ the
holds in any inner product space. angle between them is e = cos-1 ( v · u / II v \\ 11 u II). Let us use the �ame type of equation
in V; that is, we define the angle e between v and II in the inner product space V to be
THEOREM 9.3 Cauchy-Schwartz Inequality If v and u are vectors in an inner product space V, then
{v, u)
= cos 1 ---.
£l _
17
l{v, u)I S llv\1 1\ul\. llvll llull
There is, however, one potential flaw in this definition of angle between vectors. We
Proof If either v = 0 or u = 0, then {v, u) = 0 by part (1) ofTheorem 9.1 and either l\vll or need
lluII will be zero by part ( 4) of the definition of an inner product. Hence we in fact have
equality of l{v, u)I and l \v\\ l\u\l when either v = 0 or u = 0. Suppose then that both I IIv\1(v,11u)1111 I .s 1
v I= 0 and u i= 0. Dividing by \\vii \lull , we have that the inequality we wish to ohtain
is equivalent to the inequality for this inverse cosine to be defined. Do we have this for any nonzero vectors v and 11?
By the Cauchy-Schwartz inequality we do, and hence our definition does make <,ensc.
As in JR2 or JR3, we will say that two nonzero vectors v and II in the inner product
space V are orthogonal (or perpendicular) if the angle between v and II is n/2. We
Since also consider the zero vector of V to be orthogonal to each vector in V. Since R =
cos-1 ( (v, u)/ \I vii llu \\) is rr/2 if and only if (v, u) = 0 in the case when both 1 and 11 1
I-
\IVII I - \lull I
I- u - are nonzero and (u, u) = 0 when either v or II is the zero vector, we have Theorem 9.4.
V
- I and -I
(see Exercise 16), it suffices to assume thal both v and u have length I and prove that THEOREM 9.4 Two vectors v and u in an inner product space V are orthogonal if and only if {v. 11) = 0.
l{v, u)\ S 1.
We have that EXERCISES 9.1
-1 -6
0 S {v - u, v - u) = (v, v - u) - {u, v - u) = (v, v) - {v, u) - {u, v) + (u, u) In Exercises 1-6, find (v, u), l\vl\ , l\uI\, and the angle() 3 -I
= \lv\1 2 - 2{v, u) + llu11 = 2- 2{v, u).
2 between v and u for the given vectors and the indicated
; (v. u) = t' · 11
l l
inner product. 4. V = -4 ,II= 2
Solving the inequality OS 2 - 2{v, u) for (v, u), we see 2 I
-
{v, u) S 1. ]. v = [ _ � u=[ � (v, ) = v · u -3
u
Carrying out similar steps with (v+u, v +u), the details of which we leave as an exercise 5. v(x) = x2 - .t, u(x) = 3x + 5; (v, u) =
(Exercise I 7), we obtain
J�
v(x)u(x)dx
{v, u) :::. -1.
6. v(x) = sin x, u(x) = cosx; (v. u) =
Thus we have J::..,, v(x)u(x)dx
•
\(v, u)\ S I 7. Verify that the function
as needed. b
(/,g) = 1 f(x)g(x)dx
2 Named for the French mathematician Augustin Louis Cauchy (1789-1857) and the Gennan mathematician in Example 2 satisfies the four properties of an inner
Herman Amandus Schwanz (1843-192 I). product.
418 Chapter 9 Inner Product Spaces 9.1 Inner Product Spaces 419
In JR2 or JR 3 where v . u = II v II II u II cos e, we have that The next notion about inner product spaces we introduce in this section is that of
the angle between two vectors v and u in an inner product space V. If either vor II is
\v · u\ = \\\v\\ llull cos()\= \\v\l llu\\\ cos()\ S \\vii llu\l
the zero vector, we take the angle () between v and II to be zero just as we do in iR2 or
since I cos 81 s I. Remarkably, this inequalily, called the Cauchy-Schwartz inequality, 2 IR3 . Suppose that v and u are both nonzero. Recall that if v and u arc in R2 or :R.\ the
holds in any inner product space. angle between them is e = cos-1 ( v · u / II v \\ 11 u II). Let us use the �ame type of equation
in V; that is, we define the angle e between v and II in the inner product space V to be
THEOREM 9.3 Cauchy-Schwartz Inequality If v and u are vectors in an inner product space V, then
{v, u)
= cos 1 ---.
£l _
17
l{v, u)I S llv\1 1\ul\. llvll llull
There is, however, one potential flaw in this definition of angle between vectors. We
Proof If either v = 0 or u = 0, then {v, u) = 0 by part (1) ofTheorem 9.1 and either l\vll or need
lluII will be zero by part ( 4) of the definition of an inner product. Hence we in fact have
equality of l{v, u)I and l \v\\ l\u\l when either v = 0 or u = 0. Suppose then that both I IIv\1(v,11u)1111 I .s 1
v I= 0 and u i= 0. Dividing by \\vii \lull , we have that the inequality we wish to ohtain
is equivalent to the inequality for this inverse cosine to be defined. Do we have this for any nonzero vectors v and 11?
By the Cauchy-Schwartz inequality we do, and hence our definition does make <,ensc.
As in JR2 or JR3, we will say that two nonzero vectors v and II in the inner product
space V are orthogonal (or perpendicular) if the angle between v and II is n/2. We
Since also consider the zero vector of V to be orthogonal to each vector in V. Since R =
cos-1 ( (v, u)/ \I vii llu \\) is rr/2 if and only if (v, u) = 0 in the case when both 1 and 11 1
I-
\IVII I - \lull I
I- u - are nonzero and (u, u) = 0 when either v or II is the zero vector, we have Theorem 9.4.
V
- I and -I
(see Exercise 16), it suffices to assume thal both v and u have length I and prove that THEOREM 9.4 Two vectors v and u in an inner product space V are orthogonal if and only if {v. 11) = 0.
l{v, u)\ S 1.
We have that EXERCISES 9.1
-1 -6
0 S {v - u, v - u) = (v, v - u) - {u, v - u) = (v, v) - {v, u) - {u, v) + (u, u) In Exercises 1-6, find (v, u), l\vl\ , l\uI\, and the angle() 3 -I
= \lv\1 2 - 2{v, u) + llu11 = 2- 2{v, u).
2 between v and u for the given vectors and the indicated
; (v. u) = t' · 11
l l
inner product. 4. V = -4 ,II= 2
Solving the inequality OS 2 - 2{v, u) for (v, u), we see 2 I
-
{v, u) S 1. ]. v = [ _ � u=[ � (v, ) = v · u -3
u
Carrying out similar steps with (v+u, v +u), the details of which we leave as an exercise 5. v(x) = x2 - .t, u(x) = 3x + 5; (v, u) =
(Exercise I 7), we obtain
J�
v(x)u(x)dx
{v, u) :::. -1.
6. v(x) = sin x, u(x) = cosx; (v. u) =
Thus we have J::..,, v(x)u(x)dx
•
\(v, u)\ S I 7. Verify that the function
as needed. b
(/,g) = 1 f(x)g(x)dx
2 Named for the French mathematician Augustin Louis Cauchy (1789-1857) and the Gennan mathematician in Example 2 satisfies the four properties of an inner
Herman Amandus Schwanz (1843-192 I). product.
420 Chapter 9 Inner Product Spaces
9.2 Orthonormal Bases -'21
8. a) Let Verify that this defines an inner product on the 16. Show that ifv is a nonzero vector in an inner product b) Show that if the angle between II and l' is zero.
vector space P.-1. space V, then v/llvll is a vector of length I. then u = (111111/llv\l )v. (Hint: Suppose
b) For the inner product space in part (a) with 17. Complete the proof ofTheorem 9.3 by showino that w = u - (111111/llvll )v 'I- 0. Consider (ui. w).)
X1 = -1, X2 = 0, X3 = 1, find (f(x), g(x)), if v and u are vectors of length I, then (v, u) { -1. c) Show that if the angle between u and vis n.
11/(x)II, llg(x) II, and the angle between f(x) then LI= -(llull/ llvll)v.
and g(x) if f(x) = x and g(x) = l -x 2 •
18. Show that sin x and sin 2x define two orthooonal"'
vectors in the inner product space C[O. JT] with in- d) Show that if 11 is parallel to v, then II i� a l>Calar
12. a) Suppose that w(x) is a continuous nonnegative ner product (f , g) = f0" f (x)g(x) dx. multiple of v.
be vectors in IR" and let w 1• w2, ... , w,, be
positive real numbers. Define (v, u) to be function on [a, b] for which J: w(x) dx > 0. 25. The vector projection of II onto v, proj,,(11), is
For f and gin C[a, b], define (f, g) to be 19. Show that 2../Jx and 3v'5x 2 - ../5 define two or
+ · · · + a"w"b". thogonal vectors in the inner product space C[-1, l J (u, v)
(v, u) = a1 w1b1 +a2w2h2
((a11 }, (h,,}) = Lan b ,,. function T: V-> V given by T(w) = w+(u. w)v
Verify that T is a linear transformation. What are is called a shear parallel to the vector v.
l l[. llull, and the angle between v
Find (v, u), l v n=I the elements in ker(T)?
and u for the inner product on JR2 determined a) Show that T is a linear transfonnation.
Show that this defines an inner product on [2.
J- ! J.
by A if In Exercises 24-26, V denotes an inner product space b) Show that T(w) - w is parallel to l' for all
e) Verify that {a.} = { I /2"} and and u and v denote vectors in V with v -/= 0. vectors win V.
1 (h,.} = { (-3/4)"} are vectors in [2 and then find
v= and u = ({a11 J. {b 11 }), ll(a,,}11, ll(h11 JII, and the angle 24. We say that LI and v arc parallel if the angle between c) Show tha1 the only eigenvectors of T arc
[ [ _
between {a11} and {b,,J for these vectors. u and v is O or rr. nonzero vectors orthogonal to II and the
eigenvalue associated with these eigem.:ctors
10. Find an II x n matrix A so that the weighted inner 14. Prove the following parts of Theorem 9.1. a) Show that ifu is a scalar multiple of v, then 11
is l.
product in Exercise 8(a) is expressed as the inner and v are parallel.
a) Part (3)
product on JR" detennincd by A.
b) Part (4)
11. a) Let x,, x2, ... , Xn be distinct real numbers. For 9.2 ORTHONORMAL BASES
c) Part (5)
J (x) and g(x) in Pn-h define (f (x), g(x)) to
A set of vectorsv,. v2, . ·.·, v. in Vis called an
be 15. Prove the following parts of Theorem 9.2. Suppose that v is an inner product space.
gonal to each vi or. e_q�1valentl_y. (v;. Vj) "'.' 0
a) Part (1) orthogonal set of vectors if each v; is ortho
(f(x), g(x)) = f(x1)g(x1 ) + J (x2)g(x2) + · · · length I, or II 11; II = I (this 1s described b} �aymg
-
for each i I= j. Further, if each v; has
+ f(x.)g(x.). b) Part (2) v , v , .•. , v,, form an orthonormal set of ,cctori>.
v; is a unit vector), then we say that 1 2
420 Chapter 9 Inner Product Spaces
9.2 Orthonormal Bases -'21
8. a) Let Verify that this defines an inner product on the 16. Show that ifv is a nonzero vector in an inner product b) Show that if the angle between II and l' is zero.
vector space P.-1. space V, then v/llvll is a vector of length I. then u = (111111/llv\l )v. (Hint: Suppose
b) For the inner product space in part (a) with 17. Complete the proof ofTheorem 9.3 by showino that w = u - (111111/llvll )v 'I- 0. Consider (ui. w).)
X1 = -1, X2 = 0, X3 = 1, find (f(x), g(x)), if v and u are vectors of length I, then (v, u) { -1. c) Show that if the angle between u and vis n.
11/(x)II, llg(x) II, and the angle between f(x) then LI= -(llull/ llvll)v.
and g(x) if f(x) = x and g(x) = l -x 2 •
18. Show that sin x and sin 2x define two orthooonal"'
vectors in the inner product space C[O. JT] with in- d) Show that if 11 is parallel to v, then II i� a l>Calar
12. a) Suppose that w(x) is a continuous nonnegative ner product (f , g) = f0" f (x)g(x) dx. multiple of v.
be vectors in IR" and let w 1• w2, ... , w,, be
positive real numbers. Define (v, u) to be function on [a, b] for which J: w(x) dx > 0. 25. The vector projection of II onto v, proj,,(11), is
For f and gin C[a, b], define (f, g) to be 19. Show that 2../Jx and 3v'5x 2 - ../5 define two or
+ · · · + a"w"b". thogonal vectors in the inner product space C[-1, l J (u, v)
(v, u) = a1 w1b1 +a2w2h2
((a11 }, (h,,}) = Lan b ,,. function T: V-> V given by T(w) = w+(u. w)v
Verify that T is a linear transformation. What are is called a shear parallel to the vector v.
l l[. llull, and the angle between v
Find (v, u), l v n=I the elements in ker(T)?
and u for the inner product on JR2 determined a) Show that T is a linear transfonnation.
Show that this defines an inner product on [2.
J- ! J.
by A if In Exercises 24-26, V denotes an inner product space b) Show that T(w) - w is parallel to l' for all
e) Verify that {a.} = { I /2"} and and u and v denote vectors in V with v -/= 0. vectors win V.
1 (h,.} = { (-3/4)"} are vectors in [2 and then find
v= and u = ({a11 J. {b 11 }), ll(a,,}11, ll(h11 JII, and the angle 24. We say that LI and v arc parallel if the angle between c) Show tha1 the only eigenvectors of T arc
[ [ _
between {a11} and {b,,J for these vectors. u and v is O or rr. nonzero vectors orthogonal to II and the
eigenvalue associated with these eigem.:ctors
10. Find an II x n matrix A so that the weighted inner 14. Prove the following parts of Theorem 9.1. a) Show that ifu is a scalar multiple of v, then 11
is l.
product in Exercise 8(a) is expressed as the inner and v are parallel.
a) Part (3)
product on JR" detennincd by A.
b) Part (4)
11. a) Let x,, x2, ... , Xn be distinct real numbers. For 9.2 ORTHONORMAL BASES
c) Part (5)
J (x) and g(x) in Pn-h define (f (x), g(x)) to
A set of vectorsv,. v2, . ·.·, v. in Vis called an
be 15. Prove the following parts of Theorem 9.2. Suppose that v is an inner product space.
gonal to each vi or. e_q�1valentl_y. (v;. Vj) "'.' 0
a) Part (1) orthogonal set of vectors if each v; is ortho
(f(x), g(x)) = f(x1)g(x1 ) + J (x2)g(x2) + · · · length I, or II 11; II = I (this 1s described b} �aymg
-
for each i I= j. Further, if each v; has
+ f(x.)g(x.). b) Part (2) v , v , .•. , v,, form an orthonormal set of ,cctori>.
v; is a unit vector), then we say that 1 2
422 Chapter 9 Inner Product Spaces 9.2 Orthonormal Bases 423
An orthonormal basis is a basis that forms an orthonormal set. Next we observe that
I I
VJ • V2 = -- - - = 0,
v'12 v'f2
DEFINITION The vectors Vt, v2, ... , Vn in an inner product space V form an I I
orthonormal basis for V if: VJ • V3 = -- - = 0
./6 ./6
1. v1, v2, •••• Vn form a basis for V; and
2. ( v;, vi) = 0 for each i i= j;
3. II v; II = I for each i. I 1 2
Vz • V3 = -- + -- - - = 0.
v"l8 Jig Jig
and consequently VJ, vi, V3 form an orthogonal set of vectors.
The standard basis Finally,
�
= V 2 -r 2 = I
l!vill .
for 1!11 is easily seen to be an orthonormal basis for IR" when the dot product is used as and
the inner product. It is not the only one, however, as the following example illustrates.
I I I •
J
llv31i=-+-+-= I.
1[ /./2 ] 1[ /v'3 ]
EXAMPLE 1 Verify that 3 3 3
2/./6
/
-1/v'3
The next theorem shows us that it is easy to express a vector in an inner product
J/ , 1 ./6] ,
= [ -1/y'6 V3 = -J/v'J space as a linear combination of the vectors in an orthonormal basis.
-:-
Vt = V2
THEOREM 9.5 Suppose that v,, v2, .•., vn form an orthonormal basis for an inner product space V. If
form an orthonormal basis for !R3 with the dot product as the inner product.
v is a vector in V, then
Solution Let us first show VJ, v2, v3 are linearly independent. If
d
422 Chapter 9 Inner Product Spaces 9.2 Orthonormal Bases 423
An orthonormal basis is a basis that forms an orthonormal set. Next we observe that
I I
VJ • V2 = -- - - = 0,
v'12 v'f2
DEFINITION The vectors Vt, v2, ... , Vn in an inner product space V form an I I
orthonormal basis for V if: VJ • V3 = -- - = 0
./6 ./6
1. v1, v2, •••• Vn form a basis for V; and
2. ( v;, vi) = 0 for each i i= j;
3. II v; II = I for each i. I 1 2
Vz • V3 = -- + -- - - = 0.
v"l8 Jig Jig
and consequently VJ, vi, V3 form an orthogonal set of vectors.
The standard basis Finally,
�
= V 2 -r 2 = I
l!vill .
for 1!11 is easily seen to be an orthonormal basis for IR" when the dot product is used as and
the inner product. It is not the only one, however, as the following example illustrates.
I I I •
J
llv31i=-+-+-= I.
1[ /./2 ] 1[ /v'3 ]
EXAMPLE 1 Verify that 3 3 3
2/./6
/
-1/v'3
The next theorem shows us that it is easy to express a vector in an inner product
J/ , 1 ./6] ,
= [ -1/y'6 V3 = -J/v'J space as a linear combination of the vectors in an orthonormal basis.
-:-
Vt = V2
THEOREM 9.5 Suppose that v,, v2, .•., vn form an orthonormal basis for an inner product space V. If
form an orthonormal basis for !R3 with the dot product as the inner product.
v is a vector in V, then
Solution Let us first show VJ, v2, v3 are linearly independent. If
d
424 Chapter 9 Inner Product Spaces 9.2 Orthonormal Bases 425
In the second step. we first use v 1 and 112 to construct a vector w1 orthogonal to Vt
EXAMPLE 2 Express the vector
by setting
[ 1/./2]
as a linear combination of the vectors
1/./6 We then normalize w2 to obtain a unit vector v2 orthogonal to vi. In summary. our
v, = 1/-:- , v2 = [ -1/./6 ] , second step is:
2/./6
in Example 1. In the third step, we set
Solution We know from Example I that v1 , v2, v3 form an orthonormal basis for ll • Since 3
k 1 = {v1, u3), k2 = {v2, u3),
3 Since
{V, V1 } = '
.ji
5
{v, V2} = '
,/6 w3 is orthogonal to v1• Likewise, w3 is orthogonal to v2. (Verify this.) NormaliLing u1l
and gives us v3 • In summary, our third step is:
4 Step 3. Let k1 = (v1, u3 ), k2 = (v2, u3 ), and W3 = 113 - k1 vi - k2v2. Then
{V, V3)
· = - r,,•
v3 V3 = W3/1iw3II-
Theorem 9.5 then tells us that
•
Can you see the pattern developing here? Just for good measure, let u� Matt: the
3 5 4 fourth step:
v = -v, + -v2 - -v3.
..fl. ./6 v'3
Step 4. Let k, = (v1, u4), k2 = (v2, u4). k3 = (v3, u4), and
Orthonormal bases often are convenient bases to use for inner product spaces. The W4 = U4 - k1 v, - k2 v2 - k3V3. Then V4 = W4/llw411.
ease of expressing other vectors as linear combinations of the vectors in an orthonormal
5 n) produces a set ofm orth�normal
basis as illustrated in Example 2 is one reason for this. In fact, you may well have Them th step of the Gram-Schmidt process (m
encountered cases where vectors are expressed as linear combinations of orthonormal . orthonormal sets of vectors are linearl y independent (see Exercise 18).
vectors. Smce . · depe�dent
· n steps will lea<l us to a 1-rnearI y 111
basis vectors in previous courses. For instance, in calculus, physics, or engineering applyrng the Gram-Schmidt process with onnal bam for
v which will then be an orthon
courses, you may have seen the acceleration vector of a particle moving along a curve set of n orthonorrnaI vectors vi, v2, ... , n,
expressed in terms of the unit tangent and unit normal vectors. then-dimensional vector space V.
Given a basi& of vectors u1 , u2, ... , u,, for an inner product space V, there is a m-Schmidt process.
Here is an example illustrating the Gra
procedure called the Gram-Schmidt process3 for converting a basis u 1, 112, .•• , u,, into
3 Named for the Danish mathematician and actuary Jtirgen Pederson Gram ( 1850-1916) and the German t.
-
<luc
. re the inner product is the dot pro
mathematician Erhardt Schmidt (l 876-1959). into an orth onormaI basis whe
424 Chapter 9 Inner Product Spaces 9.2 Orthonormal Bases 425
In the second step. we first use v 1 and 112 to construct a vector w1 orthogonal to Vt
EXAMPLE 2 Express the vector
by setting
[ 1/./2]
as a linear combination of the vectors
1/./6 We then normalize w2 to obtain a unit vector v2 orthogonal to vi. In summary. our
v, = 1/-:- , v2 = [ -1/./6 ] , second step is:
2/./6
in Example 1. In the third step, we set
Solution We know from Example I that v1 , v2, v3 form an orthonormal basis for ll • Since 3
k 1 = {v1, u3), k2 = {v2, u3),
3 Since
{V, V1 } = '
.ji
5
{v, V2} = '
,/6 w3 is orthogonal to v1• Likewise, w3 is orthogonal to v2. (Verify this.) NormaliLing u1l
and gives us v3 • In summary, our third step is:
4 Step 3. Let k1 = (v1, u3 ), k2 = (v2, u3 ), and W3 = 113 - k1 vi - k2v2. Then
{V, V3)
· = - r,,•
v3 V3 = W3/1iw3II-
Theorem 9.5 then tells us that
•
Can you see the pattern developing here? Just for good measure, let u� Matt: the
3 5 4 fourth step:
v = -v, + -v2 - -v3.
..fl. ./6 v'3
Step 4. Let k, = (v1, u4), k2 = (v2, u4). k3 = (v3, u4), and
Orthonormal bases often are convenient bases to use for inner product spaces. The W4 = U4 - k1 v, - k2 v2 - k3V3. Then V4 = W4/llw411.
ease of expressing other vectors as linear combinations of the vectors in an orthonormal
5 n) produces a set ofm orth�normal
basis as illustrated in Example 2 is one reason for this. In fact, you may well have Them th step of the Gram-Schmidt process (m
encountered cases where vectors are expressed as linear combinations of orthonormal . orthonormal sets of vectors are linearl y independent (see Exercise 18).
vectors. Smce . · depe�dent
· n steps will lea<l us to a 1-rnearI y 111
basis vectors in previous courses. For instance, in calculus, physics, or engineering applyrng the Gram-Schmidt process with onnal bam for
v which will then be an orthon
courses, you may have seen the acceleration vector of a particle moving along a curve set of n orthonorrnaI vectors vi, v2, ... , n,
expressed in terms of the unit tangent and unit normal vectors. then-dimensional vector space V.
Given a basi& of vectors u1 , u2, ... , u,, for an inner product space V, there is a m-Schmidt process.
Here is an example illustrating the Gra
procedure called the Gram-Schmidt process3 for converting a basis u 1, 112, .•• , u,, into
3 Named for the Danish mathematician and actuary Jtirgen Pederson Gram ( 1850-1916) and the German t.
-
<luc
. re the inner product is the dot pro
mathematician Erhardt Schmidt (l 876-1959). into an orth onormaI basis whe
426 Chapter 9 Inner Product Spaces 9.2 Orthonormal Bases 427
1
II
1 /
,
I:a,;a,j
../2
= � = - - [ lI ] = [ 1/../2 ] This is the same as the dot product of column i of A and column j of A. To put it another
k=I
way, if we let
VJ
lluill ./2
0 0
Step 2:
1 then
1 / l/v'6
Vz = � = - - -1 2 ] = -1/v f,
../614 [ / [ ' ]
1 2/v'
6 Consequently, if the columns of A fonn an orthononnal basis for IRn , then
llw2II
A T A= I.
Step 3:
k1 = .ji' kz = •U3 = - Conversely, if A r A = I, then the columns of A form an orthonormal set. Also, since
v'6
1 5
A is an invertible matrix (with A- 1 = A r ), the columns of A form a basis for JR•. Thus
=VJ· U3 V2
-2/3
THEOREM 9.6 A n x ,, matrix A is an orthogonal matrix if and only if the columns of A fonn an
= = JI� 9
/
[ 2/3 ] [ 1 .J3 ]
-2/3 = -II/.J3 orthonormal basis for IR".
The Gram-Schmidt process gives us one way offinding orthonormal bases. Another
V3 11:: 11
-1/../3
)l.- [
-2/3
The orthonormal basis we have found is then procedure that is sometimes used is based on the following theorem.
�=
1
/../2
J/.. ]
0 /2
.
V2 = [ t/J6]
-J/vf6
2 v'
, V3 = [ -1/v'J
1/.,/3 ] . • THEOREM 9.7 Suppose that
/ 6
-I/.J3
Notice the orthononnal basis we found in Example 3 is the orthononnal basis given
in Example . If you were wondering how we came up with the orthonormal basis in
[
l can now see how we got it.
]
Example l, you
If A is an n x n matrix, is a unit vector with a I -:ft -1. Let w denote the column
a,. a12 a,,
a21 a22 a2n
A= . '
an! a.,,2 ... a••
426 Chapter 9 Inner Product Spaces 9.2 Orthonormal Bases 427
1
II
1 /
,
I:a,;a,j
../2
= � = - - [ lI ] = [ 1/../2 ] This is the same as the dot product of column i of A and column j of A. To put it another
k=I
way, if we let
VJ
lluill ./2
0 0
Step 2:
1 then
1 / l/v'6
Vz = � = - - -1 2 ] = -1/v f,
../614 [ / [ ' ]
1 2/v'
6 Consequently, if the columns of A fonn an orthononnal basis for IRn , then
llw2II
A T A= I.
Step 3:
k1 = .ji' kz = •U3 = - Conversely, if A r A = I, then the columns of A form an orthonormal set. Also, since
v'6
1 5
A is an invertible matrix (with A- 1 = A r ), the columns of A form a basis for JR•. Thus
=VJ· U3 V2
-2/3
THEOREM 9.6 A n x ,, matrix A is an orthogonal matrix if and only if the columns of A fonn an
= = JI� 9
/
[ 2/3 ] [ 1 .J3 ]
-2/3 = -II/.J3 orthonormal basis for IR".
The Gram-Schmidt process gives us one way offinding orthonormal bases. Another
V3 11:: 11
-1/../3
)l.- [
-2/3
The orthonormal basis we have found is then procedure that is sometimes used is based on the following theorem.
�=
1
/../2
J/.. ]
0 /2
.
V2 = [ t/J6]
-J/vf6
2 v'
, V3 = [ -1/v'J
1/.,/3 ] . • THEOREM 9.7 Suppose that
/ 6
-I/.J3
Notice the orthononnal basis we found in Example 3 is the orthononnal basis given
in Example . If you were wondering how we came up with the orthonormal basis in
[
l can now see how we got it.
]
Example l, you
If A is an n x n matrix, is a unit vector with a I -:ft -1. Let w denote the column
a,. a12 a,,
a21 a22 a2n
A= . '
an! a.,,2 ... a••
428 Chapter 9 Inner Product Spaces 9.2 Orthonormal Bases 429
A= [
OJ
1
-ww
l+a
+-
W
1
7
T
l -I
0 1/
0./2
]
where I is the (n - I) x (n - I) identity matrix is an orthogonal matrix. 0 - t/./2
-J
Proof It suffices to show that A T A = 111 Keep in mind that since v is a unit vector, v 7 v = 1. The orthonormal basis we then obtain consists of
Because of this, / 1/
•
./2
10 ] , -I0 ] , 0./2 ] •
[ [ [
T 2
w w= -a,.
We have
J
I/./2 0 -1/./2
1
-I+ l+t11
--ww T
W
T
l [: EXERCISES 9.2
(Why?) It then follows that
1. a) Verify that the vectors form an orthonormal ba�i� for IR4 "'ith the dot
A7 A=
[ I
w+ t+a ,ww w
T WW T
a I W T -W T +-1-W
+/-
l+a,
2
T
WW T
T
+ (l+ai)2 WW WW
I T T l [
1/2
1/2 l ,
product as the inner product.
b) Express
a I w 7 -w 7 + - 1-( - a 2 )w 7
OJW-
l+ai WW
I+a 1 I
l l
Vt =
2 I ( I - a 2)ww T
./2/2
=[aw ww + J - T+ii;'ww + (I+a,)2
1
w + t1a, (1-a})w
l
T T
1 -
/
1
= I,,
./2 2
= [ ./z/6
-2/3 as a linear combination of v1. v2, VJ, 1•4 .
=[ •
01x(11-J) V3
I
form an orthonormal basis for R with the dot
0(,1-l)xl
The following example illustrates how. starting with a nonzero vector in R", we can In Exercises 3-6, use the Gram-Schmidt proce\� to con
use this vector in conjunction with Theorem 9.7 to find an orthonormal basis for R". product as the inner product.
3
I1/../3 [ J
find an orthonormal basis for JR3 .
2. a) Verify that the vectors
·f l[J[:]
We begin by normalizing u to get a unit vector v:
0
Solution
1/ 1/./3 -2/ 3
./2 [ ,;-/3
' Vz = 1/3 '
[ ]. J/./3
7. Use the approach of Example .t to find an onhonor
[ 1/../31 [ j
1/3
= VJ =
1/3
=
428 Chapter 9 Inner Product Spaces 9.2 Orthonormal Bases 429
A= [
OJ
1
-ww
l+a
+-
W
1
7
T
l -I
0 1/
0./2
]
where I is the (n - I) x (n - I) identity matrix is an orthogonal matrix. 0 - t/./2
-J
Proof It suffices to show that A T A = 111 Keep in mind that since v is a unit vector, v 7 v = 1. The orthonormal basis we then obtain consists of
Because of this, / 1/
•
./2
10 ] , -I0 ] , 0./2 ] •
[ [ [
T 2
w w= -a,.
We have
J
I/./2 0 -1/./2
1
-I+ l+t11
--ww T
W
T
l [: EXERCISES 9.2
(Why?) It then follows that
1. a) Verify that the vectors form an orthonormal ba�i� for IR4 "'ith the dot
A7 A=
[ I
w+ t+a ,ww w
T WW T
a I W T -W T +-1-W
+/-
l+a,
2
T
WW T
T
+ (l+ai)2 WW WW
I T T l [
1/2
1/2 l ,
product as the inner product.
b) Express
a I w 7 -w 7 + - 1-( - a 2 )w 7
OJW-
l+ai WW
I+a 1 I
l l
Vt =
2 I ( I - a 2)ww T
./2/2
=[aw ww + J - T+ii;'ww + (I+a,)2
1
w + t1a, (1-a})w
l
T T
1 -
/
1
= I,,
./2 2
= [ ./z/6
-2/3 as a linear combination of v1. v2, VJ, 1•4 .
=[ •
01x(11-J) V3
I
form an orthonormal basis for R with the dot
0(,1-l)xl
The following example illustrates how. starting with a nonzero vector in R", we can In Exercises 3-6, use the Gram-Schmidt proce\� to con
use this vector in conjunction with Theorem 9.7 to find an orthonormal basis for R". product as the inner product.
3
I1/../3 [ J
find an orthonormal basis for JR3 .
2. a) Verify that the vectors
·f l[J[:]
We begin by normalizing u to get a unit vector v:
0
Solution
1/ 1/./3 -2/ 3
./2 [ ,;-/3
' Vz = 1/3 '
[ ]. J/./3
7. Use the approach of Example .t to find an onhonor
[ 1/../31 [ j
1/3
= VJ =
1/3
=
430 Chapter 9 Inner Product Spaces 9.3 Schur's Theorem and Symmetric Matrices 431
8. Use the approach of Example 4 to find an orthonor 15. Show that if A and 8 are n x n orthogonal matrices, we say that B is orthogonally similar to A. Also. recall that P is the change of ba\is
mal basis for IR4 starting with the vector then AB is an orthogon al matr ix. matr ix of the matrix transformation T (X) = AX from the standard basi� a for ;R" to the
16. Suppose that A is an n x n orthogonal matrix. Prove basis fJ for JR" consisting of the columns of P when B = p-1 AP. If P is an orthogonal
that for any vector v in R11 with inner product the <lot matrix, then the basis f3 is an orthonormal bas is for JR" by Theorem 9.6.
product, IIAvll = llvll, Ou r first major result of this section is a result known as Schur 's Theorem.5 Th i�
17. a) Apply the Gram-Schmidt p rocess to find an theorem gives us a case in which a square matrix is orthogonally simi lar to a triangular
orthonormal basis for the subspace of!R3 matrix.
spanned by the vectors
Let P11 have the inne r product THEOREM 9.8 Schur's Theorem Suppose that A is an 11 x n matrix. If all of the eigenvalues of A
are real numbers, then A is orthogonally s i milar to an upper triangular matri x.
(p, q) = fI p(x)q(x) dx.
-1 Proof We use inducti on on n. The result is immed iate if k = 1. Assume the result holds for all
The polynomials obtained hy applyi ng the Gram where the inner product is the dot product. k x k matrices that have only real eigenvalues and suppose A is a (k + I) x (k + I) matri x
Schmidt process to the standard basis of P11 consisting b) Use the result of part (a) in conjunction with the w ith only real e igenvalues. Let). = , 1 be one of the eigenvalues of A and let 11 1 be an
of l, x, x , ••• , x are called normalized Legendre p oly
2 n
cross product of vectors you learned in calculus eigenvector of A associated with the eigenvalue r 1• By part (I) of Lemma 2.11. we can
nomials.4 to find an orthonormal basi s for JR3• extend the l inearly i ndepen dent set consisting of the vectoru I to a basis u 1, u2, ... , ui+ 1
9. Find the normalized Legen dre polynomials when 18. Show that if v 1, v2, .••, Um are orthonormal vectors of JRk+ 1• Apply ing the Gram-Schmidt process to this basis, we obtain an orthonormal
n = 2. in an inner product space V, then u 1, v2, ••• , v111 arc bas is v 1, v2, ••• , uk+l w ith v 1 = u1 / II u, II also being an ei genvector associated v. ith the
10. Find the normalized Legendre polynomials when linearly independent. (Hint: Use an approach along e igenvalue ,1 . 6 Letting P1 den ote the orthogonal matrix
n = 3.
11. Apply the G ram-Schmidt process to convert the
the lin es of the one in the proof of Theorem 9.5.)
19. The GramSchmidt command in Maple converts a set
P1 =[ V1 V2 Vt+I ],
standard bas i s of IR2 to an orthononnal bas is if the in of vectors (the vectors do not have to be a basis or the matr ix of the l inear tmnsfonnation T (X) = AX with respect to the ba�i,; consistin g
ner product on R2 is that of Exercise 8(b) i n Section even lin early independent) in JR" in to an orthogonal of v1, v2, ••• , Uk+! has the form
9.1. set of vectors but does not normalize them. Apply
12. Do Exercise 11 if the inner product is that of Exercise the GramSchmidl command or the correspond ing
P1-1 AP, = P1
T
AP1 =
[ r1
-
4 These nonnalizcd Legend re polynomials can also be obtained by nonnalizing the Legendre p olynomials of 6 If the first entry of l 1s not -1,
8. Use the approach of Example 4 to find an orthonor 15. Show that if A and 8 are n x n orthogonal matrices, we say that B is orthogonally similar to A. Also. recall that P is the change of ba\is
mal basis for IR4 starting with the vector then AB is an orthogon al matr ix. matr ix of the matrix transformation T (X) = AX from the standard basi� a for ;R" to the
16. Suppose that A is an n x n orthogonal matrix. Prove basis fJ for JR" consisting of the columns of P when B = p-1 AP. If P is an orthogonal
that for any vector v in R11 with inner product the <lot matrix, then the basis f3 is an orthonormal bas is for JR" by Theorem 9.6.
product, IIAvll = llvll, Ou r first major result of this section is a result known as Schur 's Theorem.5 Th i�
17. a) Apply the Gram-Schmidt p rocess to find an theorem gives us a case in which a square matrix is orthogonally simi lar to a triangular
orthonormal basis for the subspace of!R3 matrix.
spanned by the vectors
Let P11 have the inne r product THEOREM 9.8 Schur's Theorem Suppose that A is an 11 x n matrix. If all of the eigenvalues of A
are real numbers, then A is orthogonally s i milar to an upper triangular matri x.
(p, q) = fI p(x)q(x) dx.
-1 Proof We use inducti on on n. The result is immed iate if k = 1. Assume the result holds for all
The polynomials obtained hy applyi ng the Gram where the inner product is the dot product. k x k matrices that have only real eigenvalues and suppose A is a (k + I) x (k + I) matri x
Schmidt process to the standard basis of P11 consisting b) Use the result of part (a) in conjunction with the w ith only real e igenvalues. Let). = , 1 be one of the eigenvalues of A and let 11 1 be an
of l, x, x , ••• , x are called normalized Legendre p oly
2 n
cross product of vectors you learned in calculus eigenvector of A associated with the eigenvalue r 1• By part (I) of Lemma 2.11. we can
nomials.4 to find an orthonormal basi s for JR3• extend the l inearly i ndepen dent set consisting of the vectoru I to a basis u 1, u2, ... , ui+ 1
9. Find the normalized Legen dre polynomials when 18. Show that if v 1, v2, .••, Um are orthonormal vectors of JRk+ 1• Apply ing the Gram-Schmidt process to this basis, we obtain an orthonormal
n = 2. in an inner product space V, then u 1, v2, ••• , v111 arc bas is v 1, v2, ••• , uk+l w ith v 1 = u1 / II u, II also being an ei genvector associated v. ith the
10. Find the normalized Legendre polynomials when linearly independent. (Hint: Use an approach along e igenvalue ,1 . 6 Letting P1 den ote the orthogonal matrix
n = 3.
11. Apply the G ram-Schmidt process to convert the
the lin es of the one in the proof of Theorem 9.5.)
19. The GramSchmidt command in Maple converts a set
P1 =[ V1 V2 Vt+I ],
standard bas i s of IR2 to an orthononnal bas is if the in of vectors (the vectors do not have to be a basis or the matr ix of the l inear tmnsfonnation T (X) = AX with respect to the ba�i,; consistin g
ner product on R2 is that of Exercise 8(b) i n Section even lin early independent) in JR" in to an orthogonal of v1, v2, ••• , Uk+! has the form
9.1. set of vectors but does not normalize them. Apply
12. Do Exercise 11 if the inner product is that of Exercise the GramSchmidl command or the correspond ing
P1-1 AP, = P1
T
AP1 =
[ r1
-
4 These nonnalizcd Legend re polynomials can also be obtained by nonnalizing the Legendre p olynomials of 6 If the first entry of l 1s not -1,
We leave it as an exercise (Exercise 15) to verify that P = Pi Pi is an orthogonal matrix Unitary matrices are the analogue of orthogonal matrices in Mnx11 (C). An II x II matrix
and that P 7 AP is an upper triangular matrix. 0 B is unitarily similar to an 11 x II matrix A if there is an II x 11 unitary matrix P �o
The theory of inner product spaces that we have developed for vector spaces over that B = P* AP. ln the case when A has complex eigenvalues, Schur·s Theorem hold,
the real numbers can be extended to vector spaces over the complex numbers, but we provided we replace orthogonally similar by unitarily similar.
will not do so fully here. We do mention, however, one special example of a complex
l
inner product space: the vector space of n x I column vectors over the complex numbers
C, denoted en , with the dot product of two vectors THEOREM 9.9 Schur's Theorem (continued) If A is an II x II matrix, then A is unitarily similar to
an upper triangular matrix.
To prove Theorem 9.9, we adjust the proof of Theorem 9.8 to the complex setting.
v= :: and u = :: We will let you go through these details as an exercise (Exercise 16).
[ [ ] The inductive step in the proof of Schur's Theorem gives us an algorithm for trian
�II �II gularizi ng matrices. The following example illustrates how we do this in practice.
in C" defined to be
[ � : =�].
EXAMPLE 1 Find an 01thogonal matrix P and a matrix B = p r AP so that JJ is a triangular matrix
for
This dot product is then a function from pairs of vectors from C" to C. I n the case when
v and u are vectors in JR", their dot product in C" is the same as it is in JR". Many,
but not all, of the properties we have for dot products of vectors in IR" carry over to the A=
dot product on C". Perhaps the most notable exception is with scalars in the right-hand
factor. Notice that because of the conjugate of u, we have
4 5 -[
v · (cu) = c(v · u)
rather than c(v · u) as in the real case. We will leave out many details and just outline how we arrive at our solution. In fact.
With the introduction of the dot product on C11 we can extend Schur's Theorem to we authors did many of the calculations that follow using Maple. When doing 50. we
Solution
the case where A has imaginary eigenvalues. If C is a matrix whose entries are complex sometimes applied Maple's simplify command to some of the matrices since Maple did
,
[�]
c• = c:7 .
l
For instance. if
C=[ 2
4+ 3i
m 9.8. We could use the
,./2; ,
1orms a b as ·1 s t·or E3· Let us· use this for u i in the proof of Theore
. . .for iR'1 contam
I - 2i
mg
l l·
then approach m · the proof of Theorem 9.8. which involves findmg a basis .
· ·
].
i
An 11 x n matrix P with complex entries i s called a unitary matrix if take Pi to be
in this example is ui. Consequently, we may
p•p = /. II:i _� II:
Pi =
-
[
7 Named for Charles Hcrrnile ( I 822-190 I). 1/-./2 0 -1/.fi
432 Chapter 9 Inner Product Spaces 9. Schur's Theorem and Symmetric Matrices 433
3
We leave it as an exercise (Exercise 15) to verify that P = Pi Pi is an orthogonal matrix Unitary matrices are the analogue of orthogonal matrices in Mnx11 (C). An II x II matrix
and that P 7 AP is an upper triangular matrix. 0 B is unitarily similar to an 11 x II matrix A if there is an II x 11 unitary matrix P �o
The theory of inner product spaces that we have developed for vector spaces over that B = P* AP. ln the case when A has complex eigenvalues, Schur·s Theorem hold,
the real numbers can be extended to vector spaces over the complex numbers, but we provided we replace orthogonally similar by unitarily similar.
will not do so fully here. We do mention, however, one special example of a complex
l
inner product space: the vector space of n x I column vectors over the complex numbers
C, denoted en , with the dot product of two vectors THEOREM 9.9 Schur's Theorem (continued) If A is an II x II matrix, then A is unitarily similar to
an upper triangular matrix.
To prove Theorem 9.9, we adjust the proof of Theorem 9.8 to the complex setting.
v= :: and u = :: We will let you go through these details as an exercise (Exercise 16).
[ [ ] The inductive step in the proof of Schur's Theorem gives us an algorithm for trian
�II �II gularizi ng matrices. The following example illustrates how we do this in practice.
in C" defined to be
[ � : =�].
EXAMPLE 1 Find an 01thogonal matrix P and a matrix B = p r AP so that JJ is a triangular matrix
for
This dot product is then a function from pairs of vectors from C" to C. I n the case when
v and u are vectors in JR", their dot product in C" is the same as it is in JR". Many,
but not all, of the properties we have for dot products of vectors in IR" carry over to the A=
dot product on C". Perhaps the most notable exception is with scalars in the right-hand
factor. Notice that because of the conjugate of u, we have
4 5 -[
v · (cu) = c(v · u)
rather than c(v · u) as in the real case. We will leave out many details and just outline how we arrive at our solution. In fact.
With the introduction of the dot product on C11 we can extend Schur's Theorem to we authors did many of the calculations that follow using Maple. When doing 50. we
Solution
the case where A has imaginary eigenvalues. If C is a matrix whose entries are complex sometimes applied Maple's simplify command to some of the matrices since Maple did
,
[�]
c• = c:7 .
l
For instance. if
C=[ 2
4+ 3i
m 9.8. We could use the
,./2; ,
1orms a b as ·1 s t·or E3· Let us· use this for u i in the proof of Theore
. . .for iR'1 contam
I - 2i
mg
l l·
then approach m · the proof of Theorem 9.8. which involves findmg a basis .
· ·
].
i
An 11 x n matrix P with complex entries i s called a unitary matrix if take Pi to be
in this example is ui. Consequently, we may
p•p = /. II:i _� II:
Pi =
-
[
7 Named for Charles Hcrrnile ( I 822-190 I). 1/-./2 0 -1/.fi
434 Chapter 9 Inner Product Spaces
9.3 Schur's Theorem and Symmetric Matrices 435
Cal cul ating the product Pt AP, we find EXAMPLE 2 Find the general solution to Y' = A Y where A is the matrix in Example I.
-: J
-9./2/2
Solution So l ving the t riangular system Y' = BY for the matrix B in the sol ution to Example I,
A,= Pi AP= 4 or equivalently, solving the system
[ �
./2/2 2v'3 23J6
Y/1 = 3y1 - --y2- --y3
Now we have to triangularize (find Q) for the submatrix 3 6
3v'2
- Y2 = 3y2 + --y3
I
-
81 - 2
[ ./2/2
4 2 J
./2 Y3 = 3y3,
of A 1• The only eigenvalue of 8 1 is)..= 3 and E3 has we find its general so lution to be
[�J 31 2v'3 3x
c 1e - - -ci.xe - c3
3 (
,/6 2 23,,/6
2 x +
6
-x e
)
3x
-
l
as a basis vector. Applying the same set of steps we just appl ied t o A to 8 1, we find that Z= 3
c2e31 + --c3xe31
./6/3 ../3/3 2
Q = ./3 ./2
[ /3 -./6/3
l
and It fol lows from Theorem 6.7 that the general solution to Y' = A Y is as fo llows.
[
3 3 /2 Y = PZ
T
Q B1Q= 0 3
./2 2
c1e3x - -vf3 31
-c2xe - c3
( .J6
23,/6
.x2 + - -x e31
0 ]
Setting ,,/6/6 -./3/3 3 2 6
0 =
[ �/2 -./6/3 -./3/3 3./2
c2eJx + - -c3. xe3x
)
v'3/3 �
./2/2 -..;o/6 r,,
-v3/3 2
./6/3 1
./3/3 -v'fi/3
and (The entries are l ong and complicated if we calculate this product, so let us l ea,e our
answer in this product form.) •
./6/6 -,/3/3 Of course if the matrix A of a system Y' = A Y has imaginary eigenvalues. !he type
-./6/3 -v'3/3 J ' f work we did in Examples I and 2 would yield complex solutions which we would use
-./6/6 �o obtain real sol utions in a manner simi l ar to the one we used in Cha�ter 6.
v/3/3
In Chapter 5 we discussed the diagonalizabi lity of squ�e matnces and saw that
er
we have not all square matrices are diagonalizable. Further, we saw m Chapt�r 6 tha.t \lhet�
a constant coefficie nt homogen eous system of hnear d 1fferen�1al
or not the matrix of
•
-2,/3/3 -23,,/6/6 soluuon
· ct·1 agonal izable has a significant impact on how we find the general
equati· ons 1s . .
3 Usl · ng Schur's Theorem · we wi show that all symmetnc matnces
3v12/2 ] to sueh a syst em. ll
.
matnces !hat '.\e
0 3 ·th real entries are diagonalizable. This then gives us a large class of
we first prove Theorem
:mediatel y know are diagonalizable. In order to obtain this,
Our abi l ity to triangularize square matrices gives us an al ternative method to the 9.10.
· eigenvalues of A are real
methods of Chapter 6 for sol ving homogeneous systems of linear differential equations
THEOREM 9.10 If A 1s an n x n symmetn·c matrix with real entries, then all the
with constant coefficients. numbers.
434 Chapter 9 Inner Product Spaces
9.3 Schur's Theorem and Symmetric Matrices 435
Cal cul ating the product Pt AP, we find EXAMPLE 2 Find the general solution to Y' = A Y where A is the matrix in Example I.
-: J
-9./2/2
Solution So l ving the t riangular system Y' = BY for the matrix B in the sol ution to Example I,
A,= Pi AP= 4 or equivalently, solving the system
[ �
./2/2 2v'3 23J6
Y/1 = 3y1 - --y2- --y3
Now we have to triangularize (find Q) for the submatrix 3 6
3v'2
- Y2 = 3y2 + --y3
I
-
81 - 2
[ ./2/2
4 2 J
./2 Y3 = 3y3,
of A 1• The only eigenvalue of 8 1 is)..= 3 and E3 has we find its general so lution to be
[�J 31 2v'3 3x
c 1e - - -ci.xe - c3
3 (
,/6 2 23,,/6
2 x +
6
-x e
)
3x
-
l
as a basis vector. Applying the same set of steps we just appl ied t o A to 8 1, we find that Z= 3
c2e31 + --c3xe31
./6/3 ../3/3 2
Q = ./3 ./2
[ /3 -./6/3
l
and It fol lows from Theorem 6.7 that the general solution to Y' = A Y is as fo llows.
[
3 3 /2 Y = PZ
T
Q B1Q= 0 3
./2 2
c1e3x - -vf3 31
-c2xe - c3
( .J6
23,/6
.x2 + - -x e31
0 ]
Setting ,,/6/6 -./3/3 3 2 6
0 =
[ �/2 -./6/3 -./3/3 3./2
c2eJx + - -c3. xe3x
)
v'3/3 �
./2/2 -..;o/6 r,,
-v3/3 2
./6/3 1
./3/3 -v'fi/3
and (The entries are l ong and complicated if we calculate this product, so let us l ea,e our
answer in this product form.) •
./6/6 -,/3/3 Of course if the matrix A of a system Y' = A Y has imaginary eigenvalues. !he type
-./6/3 -v'3/3 J ' f work we did in Examples I and 2 would yield complex solutions which we would use
-./6/6 �o obtain real sol utions in a manner simi l ar to the one we used in Cha�ter 6.
v/3/3
In Chapter 5 we discussed the diagonalizabi lity of squ�e matnces and saw that
er
we have not all square matrices are diagonalizable. Further, we saw m Chapt�r 6 tha.t \lhet�
a constant coefficie nt homogen eous system of hnear d 1fferen�1al
or not the matrix of
•
-2,/3/3 -23,,/6/6 soluuon
· ct·1 agonal izable has a significant impact on how we find the general
equati· ons 1s . .
3 Usl · ng Schur's Theorem · we wi show that all symmetnc matnces
3v12/2 ] to sueh a syst em. ll
.
matnces !hat '.\e
0 3 ·th real entries are diagonalizable. This then gives us a large class of
we first prove Theorem
:mediatel y know are diagonalizable. In order to obtain this,
Our abi l ity to triangularize square matrices gives us an al ternative method to the 9.10.
· eigenvalues of A are real
methods of Chapter 6 for sol ving homogeneous systems of linear differential equations
THEOREM 9.10 If A 1s an n x n symmetn·c matrix with real entries, then all the
with constant coefficients. numbers.
436 Chapter 9 Inner Product Spaces 9.3 Schur's Theorem and Symmetric Matrices 437
Proof Suppose that)... = r is an eigenvalue of A. Let v be an eigenvector of A associated Of course, step l is nothing new. We have been doing this all along. But next:
with the eigenvalue r. On the one hand, notice that (Av)· v (viewing this dot product
Step 2. Apply the Gram-Schmidt process to the basis of each eigenspace to
in C") is
obtain an orthormal basis for the eigenspace.
(Av)· v = (rv) · v = r(v · v).
Since we know A is diagonalizable, we have from Chapter 5 that altogether the
On the other hand, vectors from the eigenspace bases in step I form a basis of eigenvectors of A for !Rn .
The same then holds for all of the vectors we obtain in step 2. Because of Theorem 9.12.
(Av). V = (Av)T v = VT A T V= V T Av= V T Av= V. (rv) = r(v. v).
the vectors obtained in step 2 will be orthonormal. Thus they give us an orthonormal
Since v . v =I= 0, these two equations give us that r =rand hence r is a real number. • basis for IR", which forms the desired matrix P making our third and final step:
Now we are ready to prove the aforementioned result about diagonalizability. Step 3. Use for P the matrix whose columns are the vectors from step 2.
THEOREM 9.11 If A is a symmetric matrix with real entries, then A is diagonalizable. The following example illustrates how this procedure looks in practice.
Proof Since all of the eigenvalues of A are real by Theorem 9.IO, Schur's Theorem tells us that
A=[�! �l
there is an orthogonal matrix P so that p TAP is an upper triangular matrix. Using the EXAMPLE 3 Find an orthogonal matrix P so that p T AP is a diagonal matrix for
property (A8) 1' = B TA T , we have
(P T AP/ = p T AT( p T) 7 = P T AP
since A is symmetric. Hence p T AP is symmetric. However, the only way p TAP can
be both upper triangular and symmetric is for it to be a diagonal matrix and our proof is
complete. •
We can even say a bit more for the matrix A in Theorem 9.11. From its proof we can Solution
see that not only is A diagonalizable, but it is orthogonally similar to a diagonal matrix.
Step 1. The characteristic polynomial is
One way to obtain an orthogonal matrix P so that pr AP is a diagonal matrix is to use
the approach of Example I. However, there is a more commonly used approach. Before )...-3 0 -I
stating it, we prove the following theorem, which will be used to justify our technique.
det(U - A) = 0 A-4 0 = (). - 4){(A - 3)2 - I) = (A - 4)2(A - 2) .
THEOREM 9.12 If A is a symmetric matrix with real entries and v1 and v2 are eigenvectors of A with -1 0 ). - 3
different associated eigenvalues, then v1 and v2 are orthogonal. Let us next find a basis for £4.
Proof Suppose that A v 1 = r1 vi and Av2 = r2 vi. We proceed in a manner similar to the p roof I O -1
ofTheorem 9.10. On the one hand, [ 0
-� �
[!l[�]
Now we state a commonly used set of steps for finding an orthogonal matrix P that
diagonalizes an n x n symmetric matrix A with real entries.
-
Step I. Find bases for the eigenspaces of A.
436 Chapter 9 Inner Product Spaces 9.3 Schur's Theorem and Symmetric Matrices 437
Proof Suppose that)... = r is an eigenvalue of A. Let v be an eigenvector of A associated Of course, step l is nothing new. We have been doing this all along. But next:
with the eigenvalue r. On the one hand, notice that (Av)· v (viewing this dot product
Step 2. Apply the Gram-Schmidt process to the basis of each eigenspace to
in C") is
obtain an orthormal basis for the eigenspace.
(Av)· v = (rv) · v = r(v · v).
Since we know A is diagonalizable, we have from Chapter 5 that altogether the
On the other hand, vectors from the eigenspace bases in step I form a basis of eigenvectors of A for !Rn .
The same then holds for all of the vectors we obtain in step 2. Because of Theorem 9.12.
(Av). V = (Av)T v = VT A T V= V T Av= V T Av= V. (rv) = r(v. v).
the vectors obtained in step 2 will be orthonormal. Thus they give us an orthonormal
Since v . v =I= 0, these two equations give us that r =rand hence r is a real number. • basis for IR", which forms the desired matrix P making our third and final step:
Now we are ready to prove the aforementioned result about diagonalizability. Step 3. Use for P the matrix whose columns are the vectors from step 2.
THEOREM 9.11 If A is a symmetric matrix with real entries, then A is diagonalizable. The following example illustrates how this procedure looks in practice.
Proof Since all of the eigenvalues of A are real by Theorem 9.IO, Schur's Theorem tells us that
A=[�! �l
there is an orthogonal matrix P so that p TAP is an upper triangular matrix. Using the EXAMPLE 3 Find an orthogonal matrix P so that p T AP is a diagonal matrix for
property (A8) 1' = B TA T , we have
(P T AP/ = p T AT( p T) 7 = P T AP
since A is symmetric. Hence p T AP is symmetric. However, the only way p TAP can
be both upper triangular and symmetric is for it to be a diagonal matrix and our proof is
complete. •
We can even say a bit more for the matrix A in Theorem 9.11. From its proof we can Solution
see that not only is A diagonalizable, but it is orthogonally similar to a diagonal matrix.
Step 1. The characteristic polynomial is
One way to obtain an orthogonal matrix P so that pr AP is a diagonal matrix is to use
the approach of Example I. However, there is a more commonly used approach. Before )...-3 0 -I
stating it, we prove the following theorem, which will be used to justify our technique.
det(U - A) = 0 A-4 0 = (). - 4){(A - 3)2 - I) = (A - 4)2(A - 2) .
THEOREM 9.12 If A is a symmetric matrix with real entries and v1 and v2 are eigenvectors of A with -1 0 ). - 3
different associated eigenvalues, then v1 and v2 are orthogonal. Let us next find a basis for £4.
Proof Suppose that A v 1 = r1 vi and Av2 = r2 vi. We proceed in a manner similar to the p roof I O -1
ofTheorem 9.10. On the one hand, [ 0
-� �
[!l[�]
Now we state a commonly used set of steps for finding an orthogonal matrix P that
diagonalizes an n x n symmetric matrix A with real entries.
-
Step I. Find bases for the eigenspaces of A.
438 Chapter 9 Inner Product Spaces 9.3 Schur's Theorem and Symmetric Matrices 439
n
form a basis for £4. To complete step l, we find a basis for £2. which forms an orthononnal basis for £2 .
0 -1 : 0] 0 I
o
1 i/./2
-1J .Ji
•
[-� [:
Step 3. P =
[ o o
-2 0 0 � 1 ]
1/..fi
I
0 1
[ ]-[-:]
-1 0 -1 : 0 0 0 I /./2
EXERCISES 9.3
�
[ v'2 ,Ii
in Exercises 1-4, find an orthogonal matrix Panda ma
=! ! ]
The vector trix B = p r AP so that Bis a triangular matrix forthe 11. A
given matrix A. = �
0
=� _: � ]
�
1. A= [ -� �] 2. A= [
40 3
=� l
[
12. A=
O O
forms a basis for £2.
3. A = [
-3 0
Step 2. Applying the Gram-Schmidt process to the basis consisting of 0 1
13. Find the general solution to the system of linear dif
4. A= [ ! �
8 12 -7
ferential equation� Y' = A Y for the matrix A in
Exercise 9.
14. Find the general solution to the system of linear dif
ferential equations Y' = AY for the matri., A in
S. Find the general solution to the system of linear dif
for £4 we obtain: Exercise I 0.
ferential equations Y' = AY for the matrix A in
Exercise 3. 15. Complete the proof of Theorem 9.8 by showing:
UJ = UJ, 6. Find the general solution to the system of linear dif a) P = P1 P2 is an orthogonal matrix:
ferential equations Y' = A Y for the matrix A in b) pT APis an upper triangular matrix.
�]' 1 / :
Exercise 4.
16. Prove Theorem 9.9.
[ [ ]
The vectors In Exercises 7-12, find an orthogonal matrix P so that 17. Find a unitary matrix U and a matrix B = u· AU
pT AP is a diagonal matrix for the given matrix A.
l l
so that B is a triangular matrix for
I O]
u-[ -! -HJ
7. A - [ 8. A - [ _ - -1 +i
� � 2: �
0 A= [ -I I+ i O
If.Ji
2+i 1 I
=: :� =: ]
form an orthonormal basis for £4 . Applying the Gram-Schmidt process to the basis -
consisting of 18. Prove that if an II x n matrix A is orthogonall) �imilar
to a diagonal matrix. then A is a symmetric matri'<..
10. A 19. Prove that for any matrix A \\ith real entric,. A T A
[ and AAT are diagonalizable matrice�.
-
n
form a basis for £4. To complete step l, we find a basis for £2. which forms an orthononnal basis for £2 .
0 -1 : 0] 0 I
o
1 i/./2
-1J .Ji
•
[-� [:
Step 3. P =
[ o o
-2 0 0 � 1 ]
1/..fi
I
0 1
[ ]-[-:]
-1 0 -1 : 0 0 0 I /./2
EXERCISES 9.3
�
[ v'2 ,Ii
in Exercises 1-4, find an orthogonal matrix Panda ma
=! ! ]
The vector trix B = p r AP so that Bis a triangular matrix forthe 11. A
given matrix A. = �
0
=� _: � ]
�
1. A= [ -� �] 2. A= [
40 3
=� l
[
12. A=
O O
forms a basis for £2.
3. A = [
-3 0
Step 2. Applying the Gram-Schmidt process to the basis consisting of 0 1
13. Find the general solution to the system of linear dif
4. A= [ ! �
8 12 -7
ferential equation� Y' = A Y for the matrix A in
Exercise 9.
14. Find the general solution to the system of linear dif
ferential equations Y' = AY for the matri., A in
S. Find the general solution to the system of linear dif
for £4 we obtain: Exercise I 0.
ferential equations Y' = AY for the matrix A in
Exercise 3. 15. Complete the proof of Theorem 9.8 by showing:
UJ = UJ, 6. Find the general solution to the system of linear dif a) P = P1 P2 is an orthogonal matrix:
ferential equations Y' = A Y for the matrix A in b) pT APis an upper triangular matrix.
�]' 1 / :
Exercise 4.
16. Prove Theorem 9.9.
[ [ ]
The vectors In Exercises 7-12, find an orthogonal matrix P so that 17. Find a unitary matrix U and a matrix B = u· AU
pT AP is a diagonal matrix for the given matrix A.
l l
so that B is a triangular matrix for
I O]
u-[ -! -HJ
7. A - [ 8. A - [ _ - -1 +i
� � 2: �
0 A= [ -I I+ i O
If.Ji
2+i 1 I
=: :� =: ]
form an orthonormal basis for £4 . Applying the Gram-Schmidt process to the basis -
consisting of 18. Prove that if an II x n matrix A is orthogonall) �imilar
to a diagonal matrix. then A is a symmetric matri'<..
10. A 19. Prove that for any matrix A \\ith real entric,. A T A
[ and AAT are diagonalizable matrice�.
-
CHAPTER I
=:
X7 = 92659ti f27)536 + 122549,i/271536, Xg = Q
-3 3
1. [ � _; ] 3. [ _: ] 5. [ �; � ] 7·
[ 15 -4 ]
2 3 0 8 2 -17
u -· -n [ n u J
11. Undefined 11·
13. [ : -� ] 15. [ :� -�: ]
7 2 9 -21
[ 73 -7]
24
-
+U
Answers to Odd Numbered
Exercises
CHAPTER I
=:
X7 = 92659ti f27)536 + 122549,i/271536, Xg = Q
-3 3
1. [ � _; ] 3. [ _: ] 5. [ �; � ] 7·
[ 15 -4 ]
2 3 0 8 2 -17
u -· -n [ n u J
11. Undefined 11·
13. [ : -� ] 15. [ :� -�: ]
7 2 9 -21
[ 73 -7]
24
-
+U
n
442 Answers to Odd Numbered Exercises
Answers to Odd Numbered Exercises 4"3
2 -4 16 27 -11 0 -I 0 0 2
I
-3 4
5 -8 -1 0
I
5 39 0 I -I 0 0 0 2 -2 4 8
31. -3 33. 34 14 -9 -6 21 35. Undefined 31. [ ; 35. (a) 0 0 0 I 0 (b) 0 0 0 -6 10
-5 4 -4 -1 33 -1 -I 0 0 0 0 I 0 0 0 0 -35/3
0 12 -2 -12 -7 0 0 0 0 0 0 0 0 0 0
423 294 307 455 1409 row-echelon form except for having leading entries I. (c)No
-52 234 -114 -230 516
964 2223 609 -121 -303
278 258 -28 -23 323 Exercises 1.5, pp. 5�51
408 1265 779 830 -94 39 [
37. · -155 -200 183 344 -83]
236 953 -38 113 1074 1. 41 3. 41 5. 41 7. 31 9. 0 11. 28 13. 15
223 31 -106 -27 545 15. 177652
r
49 602 140 -45 145
Exercises 1.6, pp. 57-58
Exercises 1.3, pp. 36-37 1/7 -3/7
1. Not invertible 3. Invertible
5.
[ 2/7 1/7 J 7. x=ll/17,y=4/17
3 5/6 -1/3
1 /7 -2/7 ��
1. [ 3. Not invertible 5. [ -1/3 1/3] 9. x = -2/5, Y = -26/5, z=-3 1 l. x =(t2 cos 2t + 21 sin 2t)/2e', y =(2t cos 2t - r2 sin 21)/:!e'
[
3/7 1/7 J
-1 -2/3 2/3 15. (a) det(A) = 14, det(B) = 7 (b) det(AB) =98, det(A -1 ) = I /14, det(Br A-1)= I /2
J J
11/12 3/4 1/12 -1/4 (c) det(A + B) = 28, det(A) + det(B) = 21
-1/4 -1/4 -1/4 1/4 +34 24 -143.355 -27.260 51.I02 72.513
7. [ 9. X = 1/2, y = 0, Z = 2 � 72.513�
1 -1/2 -1/2 0 • 4:00 -28.395 7.828 17.406
]
17 (a) � �
1.424 -90.780 - 6.640 32.664
(b)
-7/12 -1/4 1/12 1/4 0 0 72.5136 0
�J �J �J
� -13.92 26.301 l.764 -2.562 0 0 0 72.5136
ll. (a) [ �
(b)
[
� (c) [ � (c) 72.5136
·U n n n -n
21. Is in span 23. Do span 25. It cannot have a zero row.
-I
Exercises 1.4, pp. 41-43
�rn
0 0 7 Exercises 2.3, pp. 93-95
4 3. -32 5. -4 Linearly independent 7. Linearly independent
7. [ :
� 1. Linearly independent 3. Linearly dependent 5.
[ -I
0 [: 0 -3
[ 1/5]
0
I
-7 8 -1 2 5. (a) [ -:] (b)6x 2 -x
J
9. Linearly dependent 23. (a)
16 -12
n -3]
14 11. [ 13. - -4 -10
9. [
15. Is 17. Is not 19. Is 2/5
-8 �
-19 -4 13 29. v 1 and v2 are linearly independent if and only if v2 is not para llel 10 L.
n
442 Answers to Odd Numbered Exercises
Answers to Odd Numbered Exercises 4"3
2 -4 16 27 -11 0 -I 0 0 2
I
-3 4
5 -8 -1 0
I
5 39 0 I -I 0 0 0 2 -2 4 8
31. -3 33. 34 14 -9 -6 21 35. Undefined 31. [ ; 35. (a) 0 0 0 I 0 (b) 0 0 0 -6 10
-5 4 -4 -1 33 -1 -I 0 0 0 0 I 0 0 0 0 -35/3
0 12 -2 -12 -7 0 0 0 0 0 0 0 0 0 0
423 294 307 455 1409 row-echelon form except for having leading entries I. (c)No
-52 234 -114 -230 516
964 2223 609 -121 -303
278 258 -28 -23 323 Exercises 1.5, pp. 5�51
408 1265 779 830 -94 39 [
37. · -155 -200 183 344 -83]
236 953 -38 113 1074 1. 41 3. 41 5. 41 7. 31 9. 0 11. 28 13. 15
223 31 -106 -27 545 15. 177652
r
49 602 140 -45 145
Exercises 1.6, pp. 57-58
Exercises 1.3, pp. 36-37 1/7 -3/7
1. Not invertible 3. Invertible
5.
[ 2/7 1/7 J 7. x=ll/17,y=4/17
3 5/6 -1/3
1 /7 -2/7 ��
1. [ 3. Not invertible 5. [ -1/3 1/3] 9. x = -2/5, Y = -26/5, z=-3 1 l. x =(t2 cos 2t + 21 sin 2t)/2e', y =(2t cos 2t - r2 sin 21)/:!e'
[
3/7 1/7 J
-1 -2/3 2/3 15. (a) det(A) = 14, det(B) = 7 (b) det(AB) =98, det(A -1 ) = I /14, det(Br A-1)= I /2
J J
11/12 3/4 1/12 -1/4 (c) det(A + B) = 28, det(A) + det(B) = 21
-1/4 -1/4 -1/4 1/4 +34 24 -143.355 -27.260 51.I02 72.513
7. [ 9. X = 1/2, y = 0, Z = 2 � 72.513�
1 -1/2 -1/2 0 • 4:00 -28.395 7.828 17.406
]
17 (a) � �
1.424 -90.780 - 6.640 32.664
(b)
-7/12 -1/4 1/12 1/4 0 0 72.5136 0
�J �J �J
� -13.92 26.301 l.764 -2.562 0 0 0 72.5136
ll. (a) [ �
(b)
[
� (c) [ � (c) 72.5136
·U n n n -n
21. Is in span 23. Do span 25. It cannot have a zero row.
-I
Exercises 1.4, pp. 41-43
�rn
0 0 7 Exercises 2.3, pp. 93-95
4 3. -32 5. -4 Linearly independent 7. Linearly independent
7. [ :
� 1. Linearly independent 3. Linearly dependent 5.
[ -I
0 [: 0 -3
[ 1/5]
0
I
-7 8 -1 2 5. (a) [ -:] (b)6x 2 -x
J
9. Linearly dependent 23. (a)
16 -12
n -3]
14 11. [ 13. - -4 -10
9. [
15. Is 17. Is not 19. Is 2/5
-8 �
-19 -4 13 29. v 1 and v2 are linearly independent if and only if v2 is not para llel 10 L.
444 Answers to Odd Numbered Exercises Answers to Odd Numbered Exercises 445
-3697930773/391460975
4I 30489956/391460975
- l 445251861/78292195
- l 926239514/391460975
31. ( b)
3619315327 /391460975 27. �pply th� ga�sselim or gaussjord command to the matrix with rows vf, vf, ... , v[, v[+i · If thi� docs not rc�ult
3387950282/391460975 m a matnx with a zero row, Vk+I will not be in Span{ v1, v2, ••• , vk).
CHAPTER3
Exercises 2.4, pp.104-106
Exercises 3.1, pp.119-120
I. (a) Do not Do (c) Do not (d) Do not 3. (a) Do (b) Do not (c) Do not (d) Do not
1. Second order 3. Second order 5. Solves the differential equation. Docs not satisfy the initial condition.
(b)
(b) [ I O ], [ 0 I ] 7. Solves the differential equation. Does not satisfy the initial condition. 9. 0, 16. 8 11. 3, 24, -24
5.(a) No basis (c) [ �
l[ � J
(d) 2
13. (1) y = -2, (2) y' > 0 for y > -2, y' < 0 for y < -2; y" = 2y' = 4y +8, y" > 0 for y > -2 and y'' < 0 for
y < -2
[i ] (b) [ 1 -1 0 ] , [ 0 0 l]
n [:] ( d) 2 15. (I) y = 0, 4, (2) y' > 0 for y > 4 and y < 0, y' < 0 for O < y < 4; y" = 2yy' - 4y' = (2y - 4)(y2 -4y),
y" > 0 for y > 4 and O < y < 2, y" < 0 for 2 < y < 4 and y < 0.
n
7. ( ) (c) [
a
17. (1) y = 0, 3, -3, (2) y' > 0 for y > 3 and -3 < y < 0, y' < 0 for O < y < 3 and y < -3;
[-n [-� l
y" = (3y2 - 9)y' = (3y2 - 9)(y 3 - 9y)
l[
[-:;nr:;�]
(b) [ 1 1 0 3], [ 0 0 I -5 ] (d) 2 Exercises 3.2, p.124
1. Separable 3. Not separable 5. ln IYI = 2x3 /3 - 4x + C 7. 2 arctan y = ln(x 2 + IJ + C
9. (a) (c) [
:
-1
Exercises 3.4, pp. 135-136
(c) (d) 2
7. x; y = x2/3 + C/x
21. (b) ..t 2 - 1, x 2 + 1, x - I
9. 1/x; y = -1/(2..t)+ Cx 11. t/: y = I +ce-<'
15. y = (x2 + 2x +2)/(x + I) 17. y = (cos(2x) + l)/(4x) + sin(2x)/2
25. (a) [ -17/47 -135/47 -183/47 I O ], [ -102/47 -58/47 -64/47 O I ],
(b) gausselim: [ -2 3 -1 4 -2], [ 0 12 -5 15 8], [ 0 O -47/24 -8/3]
-61/8
1. 2x2 Jn lxl - y2 = Cx2 3. 2 x-112y 112 -2x1t2y3/2 = C 5. -ix-' - .t-3y2 + 2.x-J/ 3 = C
Exercises 3.5, p. 143
gaussjord: ( I O O 17/47 102/47]. ( 0 I O 135/47 58/47 ], [ O O 183/47 64/ 47] 11. y = -2t2 ln lxl +Cx 2
9. x2(y + 1)2 /2 + y4/4 + y /3 - y /2 - y = C
rowspace: [ 1 0 0 17/47 102/47 ]. ( 0 0 1 183/47 64/47]. [ O 1 O 135/47 58/47 ]
2
1 3. e-y/x + In lxl = C
[-i H l [ =d ]
rowspan: [ 0 -7 -1 -24 -10 ]. [ 0 0 47 183 64]. [ -2 3 -1 4 -2]
H;H�l
Exercises 3.6, pp. 151-153
1. � 48 years 3. � 7567 amoeba 5. � 21.64 poun<ls 7. (a) 13.48 gal (b) -H.12 gal
9. (a) � 46.2 hours (b) No solution 11. � 55.09 minutes 13. ::::: 1.12 hour�
15. � 4.91 meters per second 17. 20R/(R - 20) where R is the earth's radius 19. Y = mx
(c) a�<.uUm
g f g
a,,,jam [ �
444 Answers to Odd Numbered Exercises Answers to Odd Numbered Exercises 445
-3697930773/391460975
4I 30489956/391460975
- l 445251861/78292195
- l 926239514/391460975
31. ( b)
3619315327 /391460975 27. �pply th� ga�sselim or gaussjord command to the matrix with rows vf, vf, ... , v[, v[+i · If thi� docs not rc�ult
3387950282/391460975 m a matnx with a zero row, Vk+I will not be in Span{ v1, v2, ••• , vk).
CHAPTER3
Exercises 2.4, pp.104-106
Exercises 3.1, pp.119-120
I. (a) Do not Do (c) Do not (d) Do not 3. (a) Do (b) Do not (c) Do not (d) Do not
1. Second order 3. Second order 5. Solves the differential equation. Docs not satisfy the initial condition.
(b)
(b) [ I O ], [ 0 I ] 7. Solves the differential equation. Does not satisfy the initial condition. 9. 0, 16. 8 11. 3, 24, -24
5.(a) No basis (c) [ �
l[ � J
(d) 2
13. (1) y = -2, (2) y' > 0 for y > -2, y' < 0 for y < -2; y" = 2y' = 4y +8, y" > 0 for y > -2 and y'' < 0 for
y < -2
[i ] (b) [ 1 -1 0 ] , [ 0 0 l]
n [:] ( d) 2 15. (I) y = 0, 4, (2) y' > 0 for y > 4 and y < 0, y' < 0 for O < y < 4; y" = 2yy' - 4y' = (2y - 4)(y2 -4y),
y" > 0 for y > 4 and O < y < 2, y" < 0 for 2 < y < 4 and y < 0.
n
7. ( ) (c) [
a
17. (1) y = 0, 3, -3, (2) y' > 0 for y > 3 and -3 < y < 0, y' < 0 for O < y < 3 and y < -3;
[-n [-� l
y" = (3y2 - 9)y' = (3y2 - 9)(y 3 - 9y)
l[
[-:;nr:;�]
(b) [ 1 1 0 3], [ 0 0 I -5 ] (d) 2 Exercises 3.2, p.124
1. Separable 3. Not separable 5. ln IYI = 2x3 /3 - 4x + C 7. 2 arctan y = ln(x 2 + IJ + C
9. (a) (c) [
:
-1
Exercises 3.4, pp. 135-136
(c) (d) 2
7. x; y = x2/3 + C/x
21. (b) ..t 2 - 1, x 2 + 1, x - I
9. 1/x; y = -1/(2..t)+ Cx 11. t/: y = I +ce-<'
15. y = (x2 + 2x +2)/(x + I) 17. y = (cos(2x) + l)/(4x) + sin(2x)/2
25. (a) [ -17/47 -135/47 -183/47 I O ], [ -102/47 -58/47 -64/47 O I ],
(b) gausselim: [ -2 3 -1 4 -2], [ 0 12 -5 15 8], [ 0 O -47/24 -8/3]
-61/8
1. 2x2 Jn lxl - y2 = Cx2 3. 2 x-112y 112 -2x1t2y3/2 = C 5. -ix-' - .t-3y2 + 2.x-J/ 3 = C
Exercises 3.5, p. 143
gaussjord: ( I O O 17/47 102/47]. ( 0 I O 135/47 58/47 ], [ O O 183/47 64/ 47] 11. y = -2t2 ln lxl +Cx 2
9. x2(y + 1)2 /2 + y4/4 + y /3 - y /2 - y = C
rowspace: [ 1 0 0 17/47 102/47 ]. ( 0 0 1 183/47 64/47]. [ O 1 O 135/47 58/47 ]
2
1 3. e-y/x + In lxl = C
[-i H l [ =d ]
rowspan: [ 0 -7 -1 -24 -10 ]. [ 0 0 47 183 64]. [ -2 3 -1 4 -2]
H;H�l
Exercises 3.6, pp. 151-153
1. � 48 years 3. � 7567 amoeba 5. � 21.64 poun<ls 7. (a) 13.48 gal (b) -H.12 gal
9. (a) � 46.2 hours (b) No solution 11. � 55.09 minutes 13. ::::: 1.12 hour�
15. � 4.91 meters per second 17. 20R/(R - 20) where R is the earth's radius 19. Y = mx
(c) a�<.uUm
g f g
a,,,jam [ �
446 Answers to Odd Numbered Exercises
Answers to Odd Numbered Exercises .W7
15. y = (x3 + x)/2 17. dx/dt = ±J2gR2 /(R +x) + v5 - 2gR 7. y = C t e-x + ci+..r - x2 /2 + e" /2 + (1/10) sin 2x - (1/5) cos 2x
9. y = Ct sin 2x+c2 cos2x + (J/J6)x sin 2x 11. y =Cte-8.r +c2 +x3/12 - (15/32)x2 + (63/128).t
Exercises 3.8, pp. 167-168
13. y = Ct ei. cos3x +c2 e2, sin 3x - (2/3)xe2' cos3x
J;
1.y(x) = (s + 2y(s)) ds 3. Maclaurin and Picard are the same: I +2x + 2x 2 + 4x3 /3 + 2x 4/3
15. y = C t e-x+ c2e4x + C3 - x 2/8 + (3/16)x - (l/20)xe4x
5. Maclaurin: I + 2x+4x2+ 8x 3 + 16x4 , Picard: I+ 2x + 4x2 + 8x3 + 16x4 + (416/15)x5 + (l28/3)x6
+(3712/63)x7 + (4544/63)x8 + (44032/567)x9+(22528/315)x 10+ (10240/189)x11 + (2048/63)x 12 17. y = 1/32 - x/8- (31/96)e4x - (41/24)e-2'
+ (8192/567)x 13 + (16384/3969)x 14 + (32768/59535)x 15 19.y = (5/2)xe' - (15e + 2)e x /(4e) + ((5/4)e2 + e/2)e-·'
7. -ll/42 - x/9+x3 /3+x 4/18-x7 /63 9.f�'cos( sins) ds 21. y = -(1/4) cosx-(l /4) sin x - (127/ IOO)e2x cos x + (139/ IOO)elt sinx + (7/5)x+ (63/25)
2 3
13. I + (x - I ) /2 - (x - 1) /3 + (x - 1) /6 - (x - 1)5 /20 + 13(x - 1)6 /180- (x - 1)7 /63+ (x - 1)8/160
4 23. yp = Ax2 +Bx+ C+ (Dx2 + Ex) sinx + (Fx 2 + Gx) cosx
- (x - 1)9 /270 - (x - I)11/4400 25. )'p = (Ax2 + Bx) sin3x + (Cx 2 + Dx) cos3x +Exe-·'
17. g(x) = e"(x2 /2-x+ I), y =C t X +c2x2 + e',y =(2- e)x -x2 + ex 17. e-1/5((6/7) sin(7t/5) + cos(7t/5)) 19. R =0. w = k
Exercises 4.2, pp. 201-203 CHAPTERS
71. y. =
_ Cte-Bx + c2 =Cte <-2+vS)x+c2e-c2
3. Y +./fo 5. )'= CJ e-4.r + c2e-x+ c3e-" Exercises 5.1, pp. 243--245
.y - c1e-3.t + c2xe-3.r 9,)=C
, t e -2x+ c2xe-2x + c3x2 e-2, ll.y=ct cos2x + c 2sin2x
13. y = c, e-2' + ex (c2 cosx + C3 sin .x) 15. (a) y = c,ex + c2 e -x (b) y=-e"/ (2e) + e t -.r /2 3
5. Is 7. Is 9. ls not 11. ls 15. [ 2 ]
17. (a) y =e-2x(c1 sin 2x+ c2 cos2x) (b) y =-(e-2' sin 2x)/2 19. (a) (c, + c 2 x)e-t/2x (b) -xe-tf2x I. Is 3. Is not 2 3
5 5
21. (a) c, + c2x + c3 e- ·' (b) 33/25 + (2/5)x +2e- ' /(25e ) 5
15. y = (x3 + x)/2 17. dx/dt = ±J2gR2 /(R +x) + v5 - 2gR 7. y = C t e-x + ci+..r - x2 /2 + e" /2 + (1/10) sin 2x - (1/5) cos 2x
9. y = Ct sin 2x+c2 cos2x + (J/J6)x sin 2x 11. y =Cte-8.r +c2 +x3/12 - (15/32)x2 + (63/128).t
Exercises 3.8, pp. 167-168
13. y = Ct ei. cos3x +c2 e2, sin 3x - (2/3)xe2' cos3x
J;
1.y(x) = (s + 2y(s)) ds 3. Maclaurin and Picard are the same: I +2x + 2x 2 + 4x3 /3 + 2x 4/3
15. y = C t e-x+ c2e4x + C3 - x 2/8 + (3/16)x - (l/20)xe4x
5. Maclaurin: I + 2x+4x2+ 8x 3 + 16x4 , Picard: I+ 2x + 4x2 + 8x3 + 16x4 + (416/15)x5 + (l28/3)x6
+(3712/63)x7 + (4544/63)x8 + (44032/567)x9+(22528/315)x 10+ (10240/189)x11 + (2048/63)x 12 17. y = 1/32 - x/8- (31/96)e4x - (41/24)e-2'
+ (8192/567)x 13 + (16384/3969)x 14 + (32768/59535)x 15 19.y = (5/2)xe' - (15e + 2)e x /(4e) + ((5/4)e2 + e/2)e-·'
7. -ll/42 - x/9+x3 /3+x 4/18-x7 /63 9.f�'cos( sins) ds 21. y = -(1/4) cosx-(l /4) sin x - (127/ IOO)e2x cos x + (139/ IOO)elt sinx + (7/5)x+ (63/25)
2 3
13. I + (x - I ) /2 - (x - 1) /3 + (x - 1) /6 - (x - 1)5 /20 + 13(x - 1)6 /180- (x - 1)7 /63+ (x - 1)8/160
4 23. yp = Ax2 +Bx+ C+ (Dx2 + Ex) sinx + (Fx 2 + Gx) cosx
- (x - 1)9 /270 - (x - I)11/4400 25. )'p = (Ax2 + Bx) sin3x + (Cx 2 + Dx) cos3x +Exe-·'
17. g(x) = e"(x2 /2-x+ I), y =C t X +c2x2 + e',y =(2- e)x -x2 + ex 17. e-1/5((6/7) sin(7t/5) + cos(7t/5)) 19. R =0. w = k
Exercises 4.2, pp. 201-203 CHAPTERS
71. y. =
_ Cte-Bx + c2 =Cte <-2+vS)x+c2e-c2
3. Y +./fo 5. )'= CJ e-4.r + c2e-x+ c3e-" Exercises 5.1, pp. 243--245
.y - c1e-3.t + c2xe-3.r 9,)=C
, t e -2x+ c2xe-2x + c3x2 e-2, ll.y=ct cos2x + c 2sin2x
13. y = c, e-2' + ex (c2 cosx + C3 sin .x) 15. (a) y = c,ex + c2 e -x (b) y=-e"/ (2e) + e t -.r /2 3
5. Is 7. Is 9. ls not 11. ls 15. [ 2 ]
17. (a) y =e-2x(c1 sin 2x+ c2 cos2x) (b) y =-(e-2' sin 2x)/2 19. (a) (c, + c 2 x)e-t/2x (b) -xe-tf2x I. Is 3. Is not 2 3
5 5
21. (a) c, + c2x + c3 e- ·' (b) 33/25 + (2/5)x +2e- ' /(25e ) 5
I a33 a31
: ; � -: -: ; � -
Note that the answers in Exercises 23 and 25 are not unique. 13. [ a13 a,, ::: ] ; rearrange the row and column entries of [T]� in the same way to get [T]:.
=: � -: l [-: � � _: l [
- a23 a21 a22
J. [ �
23. No basis for kernel 25. Basis for kernel: [ � ] 27. Constant func tions 1
} l [i ]
Basis for range: [ � 0
: : :l
J 17.(a) [ i ( (
-I 1 b) 1 I -1 1 c) 1/2 0 -1/2 ]
[-1 j
�
Bas;, -1 3 -5 1 0 0 I I -1/2 0 1/2
forn,
� [
(d [ (e (f [ ] ( g) [ ]
) ; ) ) �: _;
Exercises 5.2, pp. 252-253
J J
[-i � : ; ] [ :;: -:;� :;:
-10 -6 8 -8 -2 -30 -25
2x -2y
3 +2y [ x+?y
2x + y
y
1. [ x J. [ 2x+6 5 7. 4ax+4a+ 4b 1/4
3x+Y ]
•
-
� 1/4
J J
19. (a) [ � : (b) (c -1/4
� ] ) 3/4 1/2 -1/4
Jn=[ � -� (c D = [
(b)C=[ � � 51 -35 ]
23.(a A= [: 1 1 -1 -1 -1/4 -1/2 3/4 -1/4
) -� ) 0 I -1 0
(g) -2x 3 - 7x - 7
2
(d) [ ( ) (f [
1/2 -1/2 e -1/2 )
3/4 5/4 1/2 3/2 7/2 23/4
J
Exercises 5.4, pp. 277-278
.
0 [ 1 /2 [ 3/2
1.-l,3;L 1 : [ ] ,£3: . £:
I ] 3 4; 4 I
J J
1
./3/2 -/3/2
5. 2./3, - 2./3; Ez.Jf [ • E_2.fj· [
-n
I 1
.
-1 · 0 0
-1/2
J ,_, ri H-i i
£ [ E:[ ] 9. 4;£4 : [ �
7. 2, 3, l; £2 : [ : ], 3 : : ], , �
l
,.-.,£, [ £, [:
n -n '-· [ln
11. 2
£ ,_ [ �: J 13···- '
,..
15.1+ 2i,l- 2i;E1+2;: [
I /2 + i /2
l - ,
.
J [ E
1-2;-
l /2 - i /2
I J
l
17. ,. ;. _;, E, [ £, [
I 0 £, ; [ I: ; ]
;
[ 1 /2 ->/2 £,_./l ' [ 1/2 � Ji'/2
9. (a) [ -1 1 27. 3+,1J, 3 - Ji'. O; Es, J5
0 -1
Answers to Odd Numbered Exercises 4'9
448 Answers to Odd Numbered Exercises
I a33 a31
: ; � -: -: ; � -
Note that the answers in Exercises 23 and 25 are not unique. 13. [ a13 a,, ::: ] ; rearrange the row and column entries of [T]� in the same way to get [T]:.
=: � -: l [-: � � _: l [
- a23 a21 a22
J. [ �
23. No basis for kernel 25. Basis for kernel: [ � ] 27. Constant func tions 1
} l [i ]
Basis for range: [ � 0
: : :l
J 17.(a) [ i ( (
-I 1 b) 1 I -1 1 c) 1/2 0 -1/2 ]
[-1 j
�
Bas;, -1 3 -5 1 0 0 I I -1/2 0 1/2
forn,
� [
(d [ (e (f [ ] ( g) [ ]
) ; ) ) �: _;
Exercises 5.2, pp. 252-253
J J
[-i � : ; ] [ :;: -:;� :;:
-10 -6 8 -8 -2 -30 -25
2x -2y
3 +2y [ x+?y
2x + y
y
1. [ x J. [ 2x+6 5 7. 4ax+4a+ 4b 1/4
3x+Y ]
•
-
� 1/4
J J
19. (a) [ � : (b) (c -1/4
� ] ) 3/4 1/2 -1/4
Jn=[ � -� (c D = [
(b)C=[ � � 51 -35 ]
23.(a A= [: 1 1 -1 -1 -1/4 -1/2 3/4 -1/4
) -� ) 0 I -1 0
(g) -2x 3 - 7x - 7
2
(d) [ ( ) (f [
1/2 -1/2 e -1/2 )
3/4 5/4 1/2 3/2 7/2 23/4
J
Exercises 5.4, pp. 277-278
.
0 [ 1 /2 [ 3/2
1.-l,3;L 1 : [ ] ,£3: . £:
I ] 3 4; 4 I
J J
1
./3/2 -/3/2
5. 2./3, - 2./3; Ez.Jf [ • E_2.fj· [
-n
I 1
.
-1 · 0 0
-1/2
J ,_, ri H-i i
£ [ E:[ ] 9. 4;£4 : [ �
7. 2, 3, l; £2 : [ : ], 3 : : ], , �
l
,.-.,£, [ £, [:
n -n '-· [ln
11. 2
£ ,_ [ �: J 13···- '
,..
15.1+ 2i,l- 2i;E1+2;: [
I /2 + i /2
l - ,
.
J [ E
1-2;-
l /2 - i /2
I J
l
17. ,. ;. _;, E, [ £, [
I 0 £, ; [ I: ; ]
;
[ 1 /2 ->/2 £,_./l ' [ 1/2 � Ji'/2
9. (a) [ -1 1 27. 3+,1J, 3 - Ji'. O; Es, J5
0 -1
l
_n
450 Answers to Odd Numbered Exercises Answers to Odd Numbered Exercises 45 I
- I .038008500l [
29. r1 = 4.732050808, r2 = 1.267949199, r3 == -0.2 x 10-8;
0.562305740 l [ -0.3297195484 l 2 0 0
25. [ 0 3 0 27. [ �
] [:
I
3
l l
£,, : [ 0.3799374765 , £,,: 0.7681239226 , £,,: -2.143177059
0 0 I �
�- 0
-1.038008498 0.5623057380 , -1.318878189 -1 0 0 0 0 -] -3 18 -I
1 0 3 l 0 0 3359/9 -3 51 7 3422/9
39. 0 0 3 0 -970/9
i
-3 1 8 -2 -988/9
i
u � j]
] 467/9
i
l I 476/9
I . I . I /2 + i l,f2 =._ i ' 0 0 0 0 -I 476/9 -3 18 . I 485/9
I J [ :�]
Exercises 5.5, pp. 286-Z87 Exercises 5.6, pp. 291-292
-
1. Diagonali�able, [ � � J.
,J
P =[ �
l/
�J 3. Not diagonalizable
I. (a) 2, -3; v, [� J[ (b) Is diagonalizable (c)
r]
v_,
O p = [ ./3/2 -v'3/
l
5. Diagonalizable, [ ?../3 � ]
0 -2v'3 · I -
(b) s diagonalizable (c) [ � � ]
3. (a) -1, 2; V _1 : -2x + l, V 2 :x "'t- I
-I / - �,
7. Diagonali,ablc, [ � : ; P- [ : : 9. Not diagorutli,ablc 1 O
5. (a) I, -1; V I :e", V _I: e-x '(b)Isdiagonalizable (c) [ 9. -2
0 - 1J
- -
11 . Not diagonalizable 13. Diagonalizable, [ � -� � ], P = [ I : � ] CHAPTER6
0 0 -2 I O 1
. .
15. Diagonahzable, [
I +2i
0
O
1- 2i
J [ ,p=
l/2+i/2
1
1 / 2- i/2
1
]
Exercises6.1, pp. 301-302
5. YI =c1e' 7. Yt = cie-x 9. YI== 2ex 11. y, == 2e-• 13. Y1 = C1ex - 2
Y2 == I )'2 = c2e- + (I/2).t - I/4
1 O Yz = c2e-2x Y2 = c2
2x
)'3 = 0
71 . Diagonalizable, [ 0 , = � - � ;� ] 19. [ � � ]
Y3 = C3e -'
4
"-U�HUin
Y2 = C2 - e-x (x + I)
y3 =c3e -' -(l/5) cos2x + ( 1/l0)sin2x
4
Exercises6.2, p. 311
5 0 0 0 0 0 5 0 0 0 0 ·o 5 0 0 0 0 0 2 •
1. YI = Cie3x 3. YI = c1e2Jj' + c2e- •11 5. YI == c2e - C3e
2x 3x
0 l O O O 0 0 I l 0 0 0 0 I O O O 0 )'2 =Ct e' - 2c2e + c3e
Y2 = 2cie + c2e-x )'2 = (2/J3)c,e ./3x - (2/v'3)c2e- ../l•
2x
0 0 1 0 0 0 0 0 I
3x
0 0 0 0 0 I O O 0
3x 2 2
)'3 = -2c2e + c1 e3'
2x
0 0 0 -2 0 0
23.
0 0 0 -2 0 0 0 0 0 -2 1 0 '
7• Y1 = -c1f!\2:< - c2e-2x + c3e 9. y 1 == ex(c1 (cos 2x + sin2x)+ c2(cos2x -sin 2x))
0 0 0 0 -2 0 0 0 0 0 -2 0 0 0 0 0 -2 0
4x
21-
0 0 0 0 0 -2 0 0 0 0 0 -2 0 0 0 0 0 -2 Y2 = CJ e-- + c3e4"'
5 0 0 0 0 0 5 0 0 0 0 0 y3 = c2e- x + c3e
2
5 0 0 0 0 0
4"
I
l
_n
450 Answers to Odd Numbered Exercises Answers to Odd Numbered Exercises 45 I
- I .038008500l [
29. r1 = 4.732050808, r2 = 1.267949199, r3 == -0.2 x 10-8;
0.562305740 l [ -0.3297195484 l 2 0 0
25. [ 0 3 0 27. [ �
] [:
I
3
l l
£,, : [ 0.3799374765 , £,,: 0.7681239226 , £,,: -2.143177059
0 0 I �
�- 0
-1.038008498 0.5623057380 , -1.318878189 -1 0 0 0 0 -] -3 18 -I
1 0 3 l 0 0 3359/9 -3 51 7 3422/9
39. 0 0 3 0 -970/9
i
-3 1 8 -2 -988/9
i
u � j]
] 467/9
i
l I 476/9
I . I . I /2 + i l,f2 =._ i ' 0 0 0 0 -I 476/9 -3 18 . I 485/9
I J [ :�]
Exercises 5.5, pp. 286-Z87 Exercises 5.6, pp. 291-292
-
1. Diagonali�able, [ � � J.
,J
P =[ �
l/
�J 3. Not diagonalizable
I. (a) 2, -3; v, [� J[ (b) Is diagonalizable (c)
r]
v_,
O p = [ ./3/2 -v'3/
l
5. Diagonalizable, [ ?../3 � ]
0 -2v'3 · I -
(b) s diagonalizable (c) [ � � ]
3. (a) -1, 2; V _1 : -2x + l, V 2 :x "'t- I
-I / - �,
7. Diagonali,ablc, [ � : ; P- [ : : 9. Not diagorutli,ablc 1 O
5. (a) I, -1; V I :e", V _I: e-x '(b)Isdiagonalizable (c) [ 9. -2
0 - 1J
- -
11 . Not diagonalizable 13. Diagonalizable, [ � -� � ], P = [ I : � ] CHAPTER6
0 0 -2 I O 1
. .
15. Diagonahzable, [
I +2i
0
O
1- 2i
J [ ,p=
l/2+i/2
1
1 / 2- i/2
1
]
Exercises6.1, pp. 301-302
5. YI =c1e' 7. Yt = cie-x 9. YI== 2ex 11. y, == 2e-• 13. Y1 = C1ex - 2
Y2 == I )'2 = c2e- + (I/2).t - I/4
1 O Yz = c2e-2x Y2 = c2
2x
)'3 = 0
71 . Diagonalizable, [ 0 , = � - � ;� ] 19. [ � � ]
Y3 = C3e -'
4
"-U�HUin
Y2 = C2 - e-x (x + I)
y3 =c3e -' -(l/5) cos2x + ( 1/l0)sin2x
4
Exercises6.2, p. 311
5 0 0 0 0 0 5 0 0 0 0 ·o 5 0 0 0 0 0 2 •
1. YI = Cie3x 3. YI = c1e2Jj' + c2e- •11 5. YI == c2e - C3e
2x 3x
0 l O O O 0 0 I l 0 0 0 0 I O O O 0 )'2 =Ct e' - 2c2e + c3e
Y2 = 2cie + c2e-x )'2 = (2/J3)c,e ./3x - (2/v'3)c2e- ../l•
2x
0 0 1 0 0 0 0 0 I
3x
0 0 0 0 0 I O O 0
3x 2 2
)'3 = -2c2e + c1 e3'
2x
0 0 0 -2 0 0
23.
0 0 0 -2 0 0 0 0 0 -2 1 0 '
7• Y1 = -c1f!\2:< - c2e-2x + c3e 9. y 1 == ex(c1 (cos 2x + sin2x)+ c2(cos2x -sin 2x))
0 0 0 0 -2 0 0 0 0 0 -2 0 0 0 0 0 -2 0
4x
21-
0 0 0 0 0 -2 0 0 0 0 0 -2 0 0 0 0 0 -2 Y2 = CJ e-- + c3e4"'
5 0 0 0 0 0 5 0 0 0 0 0 y3 = c2e- x + c3e
2
5 0 0 0 0 0
4"
I
452 Answers to Odd Numbered Exercises
Answers to Odd Numbered ExerciSC!I .JS3
21. Yt = Ct e<2 +MJ.r + c2e(2 -Jio).r 23. YI = -c1 e5' + c2e2.r + C3 e-x 11. Y1 = (8/3)e3' - 2/3
Y2 = (- I + .Jlo)c1e (2+MJ.r - ( I + ../fo)c2e(7-,/ili).r Y2 = 2cie5x + c2e2.r + 2c3e-x 1r
Y2 = (16/3)e +2e-.r +x - 19/3
Y3 = C1e
5x
13. y1 = e2x - (7/9)e3:r - 2/9 + x/3
25. y,= ex (ci cos 3x +c2 sin3x) 27. x=cost - 3 sin t
Y2 = -2e2x + (7/9)e3:c + 2/9 + (2/3)x + (9/4)er - (l/4)e-x - (x/2)e-x
Y2= ex (ci sin3x -c2 cos3x) y=cost - sin t
y 3 = -2e2x + (7/9)e3x + 2/9 + (2/3)x
Exercises 6.3, pp. 314 - 315 15. x = 1 + t - (2/3)e1 + (2/3)e4'
6 1 · Yi = (6c1 + C2 + 6c2x)e4x y=-1 - t + (2/3)e' +(4/3)e
41
1.P= [ 4 ]' 4
0 Y2 = (4c1 + 4c2x)e x
4 Exercises 6.5, p. 322
0 0 I YI =c3e .r
1. y 1 = cie-.r +c2e-2x 3. y1=c1e-'+c2e2.r+(l/lO)cosx -(3/IO)sinx
3. P = [ 0 I O ] ; Y2 = (c2 + c3 x)e4x
Y2 = -cie-x -2c2e- 2.r
y2 = -c1e-x+ 2c2 e2r - (I/JO) sin x - (3/10) cos x
I O O y3=( I /2)(2c, + 2c2x+c3x2 )e4:c
Y = YI, Y = )'2 Y = yi,y' = Y2
1
6/49 20/7 43/49 Yi = 6c1e- 4x + 140(c2 + c3x)e 3.r + 43c3e3.r
5. y 1 =c 1 e +c2e-.r+c3 e- 4.r
.r
7. Y1=e'(c1cosx+c2sinx)+c3 e -lr
5. P = [ -8/49 -8/7 8/49 ] ; Y2 = -8c,e-4x - 56(c2 + c3 x)e3.r + 8c 3e3x
J
y2 = c1ex - c2e-, - 4c3 e- 4 . r Y2 = ex (c1(cosx -sinx) +c2(cosx+sinx)) - ll')e-2.r
Y3 = -3c,e- + 28(c2+ C3X)e + 3c3e3
4
-3/49 4/7 3/49 x 3x
y3 = c1 �+c2e-x + 16c3e-4x
.t
Y2= c,e-r - 2c2e2x + C3e3.r + 2/9 + (2/3).x - e--</4 - (x/2)e-x 13. .x, = (3/50)cos2v'3t +(l/25)cosJ21
y3 = -2c2e2.r + c3e3x + 2/9 + (2/3)x X2 = (-1/50) COS 2.J3t + (3/25) COS Jzt
7. Yt =6(c, + c2x)e4' + cie4x - (9/ I 6)x - I/32 15. x, = e-1/2((2./3/5) sin v'3t /2 + (4v'23/115)sin../fjr
/2)
Y2= (4c, +4c2x)e4' - (5/8)x + 1/4 2 /2 - (2v'23/l15)sin ..ffj,/2)
X?- = e-,1 ((4./3/5) sin ,J3t
9. Y1 = 6c,e-4"" + 140(c2+ c3.x)e3-' +43c3e3.r . + (I/15) cos t-(2/5)cos./6t+(-./6/15)sin./6t+ (l/3)c
os2t
17. x - (3/5) smt
XI2 : (6/5)sin I+ (2/ J 5)cosI +
Y2=-8c1e--4x - 56(c2+C3X )e3..r + 8c3e3x - e-2.r/25 ( I /5) cos ./6t - ( ../6/30)sin ./6, - ( I/3) 2,
cos
y3=-3c1e-4-'" + 28(c2 + c3x)e3x +3c 3e3.r + (3/25)e- 2 .r , -� 1s +e20 -2e's + 2e� + (13/4)e'o - 17/2)
21. 250(e-20 + (13/4)e- o + 2e · - 2e-
452 Answers to Odd Numbered Exercises
Answers to Odd Numbered ExerciSC!I .JS3
21. Yt = Ct e<2 +MJ.r + c2e(2 -Jio).r 23. YI = -c1 e5' + c2e2.r + C3 e-x 11. Y1 = (8/3)e3' - 2/3
Y2 = (- I + .Jlo)c1e (2+MJ.r - ( I + ../fo)c2e(7-,/ili).r Y2 = 2cie5x + c2e2.r + 2c3e-x 1r
Y2 = (16/3)e +2e-.r +x - 19/3
Y3 = C1e
5x
13. y1 = e2x - (7/9)e3:r - 2/9 + x/3
25. y,= ex (ci cos 3x +c2 sin3x) 27. x=cost - 3 sin t
Y2 = -2e2x + (7/9)e3:c + 2/9 + (2/3)x + (9/4)er - (l/4)e-x - (x/2)e-x
Y2= ex (ci sin3x -c2 cos3x) y=cost - sin t
y 3 = -2e2x + (7/9)e3x + 2/9 + (2/3)x
Exercises 6.3, pp. 314 - 315 15. x = 1 + t - (2/3)e1 + (2/3)e4'
6 1 · Yi = (6c1 + C2 + 6c2x)e4x y=-1 - t + (2/3)e' +(4/3)e
41
1.P= [ 4 ]' 4
0 Y2 = (4c1 + 4c2x)e x
4 Exercises 6.5, p. 322
0 0 I YI =c3e .r
1. y 1 = cie-.r +c2e-2x 3. y1=c1e-'+c2e2.r+(l/lO)cosx -(3/IO)sinx
3. P = [ 0 I O ] ; Y2 = (c2 + c3 x)e4x
Y2 = -cie-x -2c2e- 2.r
y2 = -c1e-x+ 2c2 e2r - (I/JO) sin x - (3/10) cos x
I O O y3=( I /2)(2c, + 2c2x+c3x2 )e4:c
Y = YI, Y = )'2 Y = yi,y' = Y2
1
6/49 20/7 43/49 Yi = 6c1e- 4x + 140(c2 + c3x)e 3.r + 43c3e3.r
5. y 1 =c 1 e +c2e-.r+c3 e- 4.r
.r
7. Y1=e'(c1cosx+c2sinx)+c3 e -lr
5. P = [ -8/49 -8/7 8/49 ] ; Y2 = -8c,e-4x - 56(c2 + c3 x)e3.r + 8c 3e3x
J
y2 = c1ex - c2e-, - 4c3 e- 4 . r Y2 = ex (c1(cosx -sinx) +c2(cosx+sinx)) - ll')e-2.r
Y3 = -3c,e- + 28(c2+ C3X)e + 3c3e3
4
-3/49 4/7 3/49 x 3x
y3 = c1 �+c2e-x + 16c3e-4x
.t
Y2= c,e-r - 2c2e2x + C3e3.r + 2/9 + (2/3).x - e--</4 - (x/2)e-x 13. .x, = (3/50)cos2v'3t +(l/25)cosJ21
y3 = -2c2e2.r + c3e3x + 2/9 + (2/3)x X2 = (-1/50) COS 2.J3t + (3/25) COS Jzt
7. Yt =6(c, + c2x)e4' + cie4x - (9/ I 6)x - I/32 15. x, = e-1/2((2./3/5) sin v'3t /2 + (4v'23/115)sin../fjr
/2)
Y2= (4c, +4c2x)e4' - (5/8)x + 1/4 2 /2 - (2v'23/l15)sin ..ffj,/2)
X?- = e-,1 ((4./3/5) sin ,J3t
9. Y1 = 6c,e-4"" + 140(c2+ c3.x)e3-' +43c3e3.r . + (I/15) cos t-(2/5)cos./6t+(-./6/15)sin./6t+ (l/3)c
os2t
17. x - (3/5) smt
XI2 : (6/5)sin I+ (2/ J 5)cosI +
Y2=-8c1e--4x - 56(c2+C3X )e3..r + 8c3e3x - e-2.r/25 ( I /5) cos ./6t - ( ../6/30)sin ./6, - ( I/3) 2,
cos
y3=-3c1e-4-'" + 28(c2 + c3x)e3x +3c 3e3.r + (3/25)e- 2 .r , -� 1s +e20 -2e's + 2e� + (13/4)e'o - 17/2)
21. 250(e-20 + (13/4)e- o + 2e · - 2e-
454 Answers to Odd Numbered Exercises Answers to Odd Numbered Exercises 455
1. (a) (0, 0), (2, -1) (b) A(O, 0) = [ � � J. A(2, -1) = [ � � J Exercises 7.1, pp. 351-352
1. 4/s 3. 4/s3 5. 1/(s + 4) 7. s/(s2 + 9) 9. 2s/(s2 + 1)2 11. 2/(s - 3)3
e (c1 /2)e./'i.'+(c2/2)e-./L - -
(c) (0.0)[ c1 :, J.(2,-1)[ ] (d) (0,0 ) unstable, (2,- J) unstable 13. 3/((s - 2)2 + 9) 15. (s - a)/((s - a)2 + b2 ) 17. s/(s2 9) 19. 4s/(s2 1)2
c2 e ( ./2ci 14)e./'i., _ (J2c2/4)e-./'i., 21. (s - a)/((s - a) - b ) 37. 4e 39. (2/3)e - (2/3)e-41
J
2 2
35. 4 61 21
41. 2 cos3t
3. (a) (0, 0), (-1, I) (b) A(O, 0) =[ I O , A(-1 I)= [ O O Exercises 7.2, pp. 355-356
J
0 I ] ' I 0
cie ],(- ci 1.y=e2' 3.y=-1 S.y=e-3'+5te-31 7.y=(l/2)sin2t
(c) (0,0)[ ; I.I)[ (d) (O,O)unstable,(-1,l)unstable
c2e c1 + cit 9. y = -(3/ IO)e' + (2/15)e- - (5/6)e_,
41
11. y = (2/3) - (2/3)e-31
(1/2 + ./5/to)c,e ./5'+(1/2- J5/IO)c2e-./5, 13. y =(2/3) - e31 /24 + e 3' /4 - 1e-1 /8
5. (a) (0,0) (b) A(O. 0) = [ : � (c) (O,O) [
- J (J5 IlO)c1 e./5 , -(J5/ 10)c2 e-./5,
] 15. y = e 2'((139/100) sin t - (127 /JOO) cost) - (I /4) sin t - (1/4) cost+ 7t/5 +63/25
(d) (0. 0) unstable 17.:::::: 328,056 19. 7 0+I 10e5 1n9/ 1 1
I /2 O ], O -1/6] 21. u =60e-r/JO/3961 + e-21 ((3847/23766) sin 61 + (3721 / 15844) cos6t)
9. (a) (0, 0), (2/3, 2) (b) A(O, 0) = [ A(2/3, 2) =[
0 0 3/2 -1/2 23. (96/65) sin 2t - (12/65) cos 2t + 18e-3' /13 - 6e- 1 /5
, e-' (c,(cosJ'St/4+c2 sin,J3t/4))
c1 12 Exercises 7.3, pp. 364-366
14
(c) (0. 0)[ ; J,(2/3, 2) [ (3e-114 /2)(c 1 (cosJ'S,/4+ ...fjsin ./3, /4)) ]
1. (a) t - tu2 (t) (c) l/s 2 - e_,.. /s 2 - 2e-21 /s
+c2(-J3 cos-J3t/4 + sin ...fj1 /4) 3. (a) e e2(1-2> u2 (t) - e8e2Ct-4>u4(t) (c) e-2<•-2> /(s - 2) - e-4<•-2>/(s - 2)
2
S. (a) Urr (t) cos 2(t - 1r) - u2,.. (t) cos2(t - 2,r) (c) e_,.., s/(s 2 + 4) - e-2tus/(s 2 +4)
3/4 0 A 0 -3/4
11. (a) (0,0), (3. 3) (b) A(O, 0) = [ ], (3. 3) = [ 7. (b) e-2s(2/s+l/s2 )- 2e-41 (4/s + J/s2) 9. (b) -e-,..'s/(s 2 + I) II. u2 (1)
0 0 3/4 -3/4 J 31 3 2
3 18 13. u1 (t)(e <r-l) - 1)/4
4 - >
15. e- /9 + t/3 - 1/9 - u2(t)(-Se <r - /9 + t/3 - I /9)
e- ' (c, ((1/2) cos3J31 /8 +(...fj/6) sin 3-J3t/8)+
c 1 3,14 17. u1(t)(e 3<1-n/24+e-3<r- 1 l/t2 - e 1 -1/8)
(c) (0,0) [ : J.(3,3) [ c2(-(-J3/6)cos3J3t/8+(l/2) sin3J31/8)) ]
19. -(39/20)e' + 1e3' /12+ (I /30) cos 3t +(l/15) sin 3t + 7 /3 + u2rr(t)(69e,-,r /20 - 13e3<1 -2ir 1 /12-
( ,J3j3)e- 3'18 (c, sin 3../3t/8 - c2 cos 3../31/8)
2
(l/30) cos3t - (l/15) sin 3t - 7/3)
(d) (0. 0) unstable,(3, 3) stable
21. e 21 /6 - e-41 /6 25. � 503.660
13. y 2 - (2g/ I) cosx = C, c = -64 J2 3 1 3/
JO_
27. 60/396Ie- /lO + e-21 ((3721/ 15844) cos61 + (3847 /23766) sin 6t) - e- 110u3(1)/396 I (6Qe-f - >
r
15. (a) (0, 0). ((2k + l)rr, 0),(2k1r, 0), k = ±1,±2,... (b) A(O, O) = [
0 I 0 l
O
-(g/l) -c/(ml) J '
6oe -2 <r-
1
3l cos(6(t - 3)) - 19e-2<r- > sin(6(t - 3)))
-
3
3 2
29. 4 - 6e- +2e-3' u 2 (t)(4 + 2e- <r- > - 6e-<r- >)
2
31. 6u" (t)(e-<
1-"' -
"
e-3<r- >J
A ((2k+l)1r,O)=[ J,A(2b,0)=[ ] Exercises 7.4, p. 369
J
g/1 -c/(ml) -(g /l) -c/(ml)
3/4 O 1. e ' - I 3. e- ;2 - (1/2) cost+(l/2}sint S. 1/((s - l)s 2 ) 7. r
17. (a) (0. 0), (4, 3) (b) A(O. O) =[ - ]. A(4, 3) =[ 0 I
1
=� _� l
0 I -3/4 0 9. cos2t +(l/8)sin2t - (t/4) cos2t 11. t /24
4
13. (3/2)(1 +e21 ) 15. (l/3)(e' +ut'>
17. Se'/3+((4J3/3) sin ./Jr/2 - (2/3) cosJ'St/2)e <- 12>
c -3 4
(c) (O,O) [ , e :! ], (
4. 3) [
( I /2)(c 1 cos ./31/2+c2 sin ../3, /2)
1
A (4, 0) = [
-
]
0
-2
-I
J. A(4/3,4/3) = [ -1
/3
-
2/3
-2/3 -1/3 J
(c) (0, 0) [ C1e'
c2e1
]. 3. y1 = (l/5)(4e2' + e-31) + u2 (t)(-e /4 + e21-2 /5 + e8-3' /2 0)
Y2 = (l/5)(4e - 4e-31) + u2(t}(e 21 -2 -e8-3')/5
21
1
454 Answers to Odd Numbered Exercises Answers to Odd Numbered Exercises 455
1. (a) (0, 0), (2, -1) (b) A(O, 0) = [ � � J. A(2, -1) = [ � � J Exercises 7.1, pp. 351-352
1. 4/s 3. 4/s3 5. 1/(s + 4) 7. s/(s2 + 9) 9. 2s/(s2 + 1)2 11. 2/(s - 3)3
e (c1 /2)e./'i.'+(c2/2)e-./L - -
(c) (0.0)[ c1 :, J.(2,-1)[ ] (d) (0,0 ) unstable, (2,- J) unstable 13. 3/((s - 2)2 + 9) 15. (s - a)/((s - a)2 + b2 ) 17. s/(s2 9) 19. 4s/(s2 1)2
c2 e ( ./2ci 14)e./'i., _ (J2c2/4)e-./'i., 21. (s - a)/((s - a) - b ) 37. 4e 39. (2/3)e - (2/3)e-41
J
2 2
35. 4 61 21
41. 2 cos3t
3. (a) (0, 0), (-1, I) (b) A(O, 0) =[ I O , A(-1 I)= [ O O Exercises 7.2, pp. 355-356
J
0 I ] ' I 0
cie ],(- ci 1.y=e2' 3.y=-1 S.y=e-3'+5te-31 7.y=(l/2)sin2t
(c) (0,0)[ ; I.I)[ (d) (O,O)unstable,(-1,l)unstable
c2e c1 + cit 9. y = -(3/ IO)e' + (2/15)e- - (5/6)e_,
41
11. y = (2/3) - (2/3)e-31
(1/2 + ./5/to)c,e ./5'+(1/2- J5/IO)c2e-./5, 13. y =(2/3) - e31 /24 + e 3' /4 - 1e-1 /8
5. (a) (0,0) (b) A(O. 0) = [ : � (c) (O,O) [
- J (J5 IlO)c1 e./5 , -(J5/ 10)c2 e-./5,
] 15. y = e 2'((139/100) sin t - (127 /JOO) cost) - (I /4) sin t - (1/4) cost+ 7t/5 +63/25
(d) (0. 0) unstable 17.:::::: 328,056 19. 7 0+I 10e5 1n9/ 1 1
I /2 O ], O -1/6] 21. u =60e-r/JO/3961 + e-21 ((3847/23766) sin 61 + (3721 / 15844) cos6t)
9. (a) (0, 0), (2/3, 2) (b) A(O, 0) = [ A(2/3, 2) =[
0 0 3/2 -1/2 23. (96/65) sin 2t - (12/65) cos 2t + 18e-3' /13 - 6e- 1 /5
, e-' (c,(cosJ'St/4+c2 sin,J3t/4))
c1 12 Exercises 7.3, pp. 364-366
14
(c) (0. 0)[ ; J,(2/3, 2) [ (3e-114 /2)(c 1 (cosJ'S,/4+ ...fjsin ./3, /4)) ]
1. (a) t - tu2 (t) (c) l/s 2 - e_,.. /s 2 - 2e-21 /s
+c2(-J3 cos-J3t/4 + sin ...fj1 /4) 3. (a) e e2(1-2> u2 (t) - e8e2Ct-4>u4(t) (c) e-2<•-2> /(s - 2) - e-4<•-2>/(s - 2)
2
S. (a) Urr (t) cos 2(t - 1r) - u2,.. (t) cos2(t - 2,r) (c) e_,.., s/(s 2 + 4) - e-2tus/(s 2 +4)
3/4 0 A 0 -3/4
11. (a) (0,0), (3. 3) (b) A(O, 0) = [ ], (3. 3) = [ 7. (b) e-2s(2/s+l/s2 )- 2e-41 (4/s + J/s2) 9. (b) -e-,..'s/(s 2 + I) II. u2 (1)
0 0 3/4 -3/4 J 31 3 2
3 18 13. u1 (t)(e <r-l) - 1)/4
4 - >
15. e- /9 + t/3 - 1/9 - u2(t)(-Se <r - /9 + t/3 - I /9)
e- ' (c, ((1/2) cos3J31 /8 +(...fj/6) sin 3-J3t/8)+
c 1 3,14 17. u1(t)(e 3<1-n/24+e-3<r- 1 l/t2 - e 1 -1/8)
(c) (0,0) [ : J.(3,3) [ c2(-(-J3/6)cos3J3t/8+(l/2) sin3J31/8)) ]
19. -(39/20)e' + 1e3' /12+ (I /30) cos 3t +(l/15) sin 3t + 7 /3 + u2rr(t)(69e,-,r /20 - 13e3<1 -2ir 1 /12-
( ,J3j3)e- 3'18 (c, sin 3../3t/8 - c2 cos 3../31/8)
2
(l/30) cos3t - (l/15) sin 3t - 7/3)
(d) (0. 0) unstable,(3, 3) stable
21. e 21 /6 - e-41 /6 25. � 503.660
13. y 2 - (2g/ I) cosx = C, c = -64 J2 3 1 3/
JO_
27. 60/396Ie- /lO + e-21 ((3721/ 15844) cos61 + (3847 /23766) sin 6t) - e- 110u3(1)/396 I (6Qe-f - >
r
15. (a) (0, 0). ((2k + l)rr, 0),(2k1r, 0), k = ±1,±2,... (b) A(O, O) = [
0 I 0 l
O
-(g/l) -c/(ml) J '
6oe -2 <r-
1
3l cos(6(t - 3)) - 19e-2<r- > sin(6(t - 3)))
-
3
3 2
29. 4 - 6e- +2e-3' u 2 (t)(4 + 2e- <r- > - 6e-<r- >)
2
31. 6u" (t)(e-<
1-"' -
"
e-3<r- >J
A ((2k+l)1r,O)=[ J,A(2b,0)=[ ] Exercises 7.4, p. 369
J
g/1 -c/(ml) -(g /l) -c/(ml)
3/4 O 1. e ' - I 3. e- ;2 - (1/2) cost+(l/2}sint S. 1/((s - l)s 2 ) 7. r
17. (a) (0. 0), (4, 3) (b) A(O. O) =[ - ]. A(4, 3) =[ 0 I
1
=� _� l
0 I -3/4 0 9. cos2t +(l/8)sin2t - (t/4) cos2t 11. t /24
4
13. (3/2)(1 +e21 ) 15. (l/3)(e' +ut'>
17. Se'/3+((4J3/3) sin ./Jr/2 - (2/3) cosJ'St/2)e <- 12>
c -3 4
(c) (O,O) [ , e :! ], (
4. 3) [
( I /2)(c 1 cos ./31/2+c2 sin ../3, /2)
1
A (4, 0) = [
-
]
0
-2
-I
J. A(4/3,4/3) = [ -1
/3
-
2/3
-2/3 -1/3 J
(c) (0, 0) [ C1e'
c2e1
]. 3. y1 = (l/5)(4e2' + e-31) + u2 (t)(-e /4 + e21-2 /5 + e8-3' /2 0)
Y2 = (l/5)(4e - 4e-31) + u2(t}(e 21 -2 -e8-3')/5
21
1
Answers to Odd Numbered Exercises .a57
456 Answers to Odd Numbered Exercises
l· [
l • ( b) 3-5./2 V 1 _ 39+5J2 V2
2 6
J
• [ 2/../5 ] , [ -I/../5 ]
Exercises 8.1, pp. 383-384
1
1. 21=:oc-xt/n! 3. z::o<x - 1)2"/n! 2n
5. I::ax /(2n)
1 7. I::o<-x)"/n! /
2
9. I::o x " /(2" (211)!) JJ. ! I:�0 (-1Y(2x) 2 J
"+ /(2n+ I)! 13. I+ x2/2 +x4/2 +x6 /6 5. [ � ] , [ -3/� 4/
- � ]
7. [-:;� ]· [ (-;� ;)/6 ], [ (-3 ;�) 6]
/
15. I - 4x + 2x /3 + 8x /3
2 3 4 0 4/5 -3/5 1/../3 (-3 + ./3)/6 (-3 - ../3 )/6
JI: ] , [ I� ]
l[
Exercises 8.2, pp. 392-393 11. [
I
1. I +x3 /3!+4/6!x6+[:2(1)(4) · · · (311 + l)x3<n+ll /(3(n + 1))!
-:n-�n
3. 2(x - x3 /3! - .x 5 /5! - 3/7!x7 -I:� 2(1)(3) ...(211 + l)x "+ /(2n + 5)!
2 5 1/../3 1/../2 -
"· [ n[
9. ao(l+x /3!+4/6!x + [:, 2 0 )(4) · · · (311+ l)x3<n+I) /(3(11+ I))!)+aJ (x + 2x4/4!+(2)(5)/7!x7 +
3 6
1/../3 1/./3
I::2(2)(5) · · · (3n + 2)x3"+4 /(3n + 4)!)
11. ao(I -x2 ) +aJ(X -x3/3!-x5 /5! - I::1 (1)(3) .. ·(2n + l)x2n+S /(211+5)!)
13. ao +a1x+(ao/2)x2 +(aif6)x3 15. a0 + a1x - (a0/6)x3 +((a0 - a1 )/12)x4
17. ao + a,x - (3ao/8)x + (5a 1 /24)x - (21 /128)a0x4 + ·..
2 3
19. a1 x+a0(1 - (l /2)x In(() +x)/(1 - x)))
21. (ao/3)(35x4 - 30x 2 + 3)
23. (b)). = 0, Y = I;)...= 2, y = x;). = 4, y = I - 2x2 ;).. = 6, y = x - 2x3 /3;). = 8, y = I - 4x2 + (4/3)x4
25. p(xo) = bo, p'(xo) = b 1 , p"(xo) = 2b2, · · ·, 11!b0 = p 1"l(x0)
9. (b) lo(x) In x +x2/4 -(3/l28)x4 + (11/ !3824)x6 + .. · (d) x - 112(c 1 cos x + c2 sin x)
II. x-3f2(c 1 (cosx +x sin x) + c2(sin x - x cosx))
J
13. (b) x-•12 cos x, x- 112 sin x (c) x- 1 12(c 1 cos x+c2 sinx)+x - /l
15. x2+x 3+x4/ 3 + x5 /36 - (7/720)x6+ · · · 17. x2 - x4 / 12+x6/384 - x8/23040 + ...
+ )(n + y)
19. (h) r(r - 1) + ar = 0 (c) r = 0; a11 + 1 = (l1 fJ a
(n + 1)(11+a) "
r-_ 1 - a,a . _ (n + I - a+ fJ)(n + l - a + y) a
,,+ 1-
(n+2-a)(n+l) "
Answers to Odd Numbered Exercises .a57
456 Answers to Odd Numbered Exercises
l· [
l • ( b) 3-5./2 V 1 _ 39+5J2 V2
2 6
J
• [ 2/../5 ] , [ -I/../5 ]
Exercises 8.1, pp. 383-384
1
1. 21=:oc-xt/n! 3. z::o<x - 1)2"/n! 2n
5. I::ax /(2n)
1 7. I::o<-x)"/n! /
2
9. I::o x " /(2" (211)!) JJ. ! I:�0 (-1Y(2x) 2 J
"+ /(2n+ I)! 13. I+ x2/2 +x4/2 +x6 /6 5. [ � ] , [ -3/� 4/
- � ]
7. [-:;� ]· [ (-;� ;)/6 ], [ (-3 ;�) 6]
/
15. I - 4x + 2x /3 + 8x /3
2 3 4 0 4/5 -3/5 1/../3 (-3 + ./3)/6 (-3 - ../3 )/6
JI: ] , [ I� ]
l[
Exercises 8.2, pp. 392-393 11. [
I
1. I +x3 /3!+4/6!x6+[:2(1)(4) · · · (311 + l)x3<n+ll /(3(n + 1))!
-:n-�n
3. 2(x - x3 /3! - .x 5 /5! - 3/7!x7 -I:� 2(1)(3) ...(211 + l)x "+ /(2n + 5)!
2 5 1/../3 1/../2 -
"· [ n[
9. ao(l+x /3!+4/6!x + [:, 2 0 )(4) · · · (311+ l)x3<n+I) /(3(11+ I))!)+aJ (x + 2x4/4!+(2)(5)/7!x7 +
3 6
1/../3 1/./3
I::2(2)(5) · · · (3n + 2)x3"+4 /(3n + 4)!)
11. ao(I -x2 ) +aJ(X -x3/3!-x5 /5! - I::1 (1)(3) .. ·(2n + l)x2n+S /(211+5)!)
13. ao +a1x+(ao/2)x2 +(aif6)x3 15. a0 + a1x - (a0/6)x3 +((a0 - a1 )/12)x4
17. ao + a,x - (3ao/8)x + (5a 1 /24)x - (21 /128)a0x4 + ·..
2 3
19. a1 x+a0(1 - (l /2)x In(() +x)/(1 - x)))
21. (ao/3)(35x4 - 30x 2 + 3)
23. (b)). = 0, Y = I;)...= 2, y = x;). = 4, y = I - 2x2 ;).. = 6, y = x - 2x3 /3;). = 8, y = I - 4x2 + (4/3)x4
25. p(xo) = bo, p'(xo) = b 1 , p"(xo) = 2b2, · · ·, 11!b0 = p 1"l(x0)
9. (b) lo(x) In x +x2/4 -(3/l28)x4 + (11/ !3824)x6 + .. · (d) x - 112(c 1 cos x + c2 sin x)
II. x-3f2(c 1 (cosx +x sin x) + c2(sin x - x cosx))
J
13. (b) x-•12 cos x, x- 112 sin x (c) x- 1 12(c 1 cos x+c2 sinx)+x - /l
15. x2+x 3+x4/ 3 + x5 /36 - (7/720)x6+ · · · 17. x2 - x4 / 12+x6/384 - x8/23040 + ...
+ )(n + y)
19. (h) r(r - 1) + ar = 0 (c) r = 0; a11 + 1 = (l1 fJ a
(n + 1)(11+a) "
r-_ 1 - a,a . _ (n + I - a+ fJ)(n + l - a + y) a
,,+ 1-
(n+2-a)(n+l) "