Professional Documents
Culture Documents
by
Zsuzsanna Gonye
Department of Mathematics
Polytechnic University
Brooklyn, New York
2003
Contents
1 Linear Equations
1.1 Systems of Linear Equations
1.2 Gauss-Jordan Elimination .
1.3 Examples . . . . . . . . . .
1.4 The Zp field . . . . . . . . .
1.5 Systems of Linear Equations
.
.
.
.
.
1
1
3
6
13
14
.
.
.
.
.
17
17
25
26
28
30
.
.
.
.
.
.
.
.
.
33
33
34
36
37
39
40
42
44
45
4 Vector Spaces
4.1 Introduction to the Euclidean n-space . . . . . . . . . . . . . . . . . .
4.2 Linear Transformation from Rn to Rm . . . . . . . . . . . . . . . . .
4.3 Real Vector Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
47
49
52
. . .
. . .
. . .
. . .
in Zp
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
2 Matrix Algebra
2.1 Matrix Arithmetic . . . . . . . . . . . . . . .
2.2 Elementary Matrices . . . . . . . . . . . . . .
2.3 Matrix Inverse . . . . . . . . . . . . . . . . . .
2.4 Method for Finding the Inverse . . . . . . . .
2.5 Diagonal, Triangular and Symmetric Matrices
.
.
.
.
.
.
.
.
.
.
3 Determinant
3.1 The Determinant Function . . . . . . . . . . . .
3.2 Calculating the Determinant for 2 2 Matrices
3.3 Geometric Meaning of the Determinant . . . . .
3.4 Properties of the Determinant Function . . . . .
3.5 Evaluating the Determinant by Row Reduction
3.6 Determinant, Invertibility and Systems of Linear
3.7 Cofactor Expansion, Adjoint Matrix . . . . . . .
3.8 Calculating the Determinant for 3 3 Matrices
3.9 Block-Triangular Matrices . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
Equations
. . . . . .
. . . . . .
. . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
ii
CONTENTS
4.4
4.5
4.6
4.7
4.8
Subspaces . . . . . . . . . . . .
Spanning . . . . . . . . . . . . .
Linear Independence . . . . . .
Basis and Dimension . . . . . .
Column Space, Row Space, Null
.
.
.
.
.
53
56
57
59
61
.
.
.
.
67
67
69
73
76
.
.
.
.
79
79
79
82
83
A Supplementary Material
A.1 Cayley-Hamilton Theorem . . . . . . . . . . . . . . . . . . . . . . . .
A.2 Exponential of Matrices . . . . . . . . . . . . . . . . . . . . . . . . .
85
85
85
B The
B.1
B.2
B.3
89
89
90
91
93
93
96
. . . .
. . . .
. . . .
. . . .
Space,
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
Rank and Nullity
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Bibliography
97
Index
97
Chapter 1
Linear Equations
1.1 Systems of Linear Equations
A linear equation in the n variables x1 , x2 , . . . , xn is one that can be expressed in the
form
a1 x 1 + a2 x 2 + + an x n = b
where a1 , a2 , . . . , an and b are constants. The constants a1 , a2 , . . . , an are called the
coefficients, b is the constant term, and the variables x1 , x2 , . . . xn are also called
unknowns. A solution of a linear equation a1 x1 + a2 x2 + + an xn = b is a sequence
of n numbers, so that the equation is satisfied when we substitute x1 = t1 , x2 =
t2 , . . . , xn = tn . We will also call it a solution vector and write it as (t1 , t2 , . . . , tn ).
The collection of all solutions of the equation is called the solution set.
A finite set of linear equations is called a system of linear equations or a linear
system. A sequence of numbers (t1 , t2 , . . . , tn ) is called a solution of the system if
x1 = t1 , x2 = t2 , . . . , xn = tn is a solution of every equation in the system. If a
system of equation has at least one solution (maybe infinitely many), then we say the
system is consistent. If a system of equations has no solutions, then it is said to be
inconsistent.
A linear equation in two unknowns, for example 2x1 5x2 = 3, can be graphed
in the Descartes coordinate system, and its graph is a straight line.
In the study of two equations
a11 x1 + a12 x2 = b1
a21 x1 + a22 x2 = b2
in two unknowns x1 , x2 over R there are (assuming each equation determines a line)
three possibilities:
The two lines are distinct and not parallel. In this case they intersect in a point
and there is a unique solution.
Linear Equations
The two lines are parallel. In this case there is no solution.
The two lines coincide. In this case any solution of one equation is also a solution
of the other, so there are infinitely many solutions.
Z. G
onye
Linear Equations
As we did this example, you may see that we have to keep track only the coefficients. Lets look at a general linear system.
a11 x1 + a12 x2 + + a1n xn = b1
a21 x1 + a22 x2 + + a2n xn = b2
..
.
am1 x1 + am2 x2 + + amn xn = bm .
The coefficients of each variable aligned in columns
a
21 a22 a2n
. . . . . . . . . . . . . . . . . . .
am1 am2 amn
is called the coefficient matrix , and if we also include the constant terms of the
equations, then the matrix
is called the augmented matrix of the system. The size of a matrix tells you how
many rows and columns it has. The coefficient matrix above is an m n (read m by
n) matrix. The augmented matrix above is m (n + 1). The following operations
are allowed on the augmented matrix in order to get a solution.
Definition 1.2.1. Elementary row operations:
1. Multiply a row through by a nonzero number.
2. Interchange two rows.
3. Add a multiple of one row to another row.
Z. G
onye
Example 1.2.3. Lets redo Example 1.2.2 using the augmented matrix. The augmented matrix of
x + 2y 3z = 9
y 2z = 4
3x + 5y 6z = 20
is
1 2 3 9
0 1 2 4 .
3 5 6 20
We add 3 times the first row to the third row (R3 R3 3R1 ), then the matrix
changes to
1 2 3 9
0 1 2 4 ,
0 1 3
7
then we add the second row to the third one
1 2 3
0 1 2
0 0 1
(R3 R3 + R2 ):
9
4 .
3
From here you can again use the equations given by this matrix. Now it is a triangular
system, so use back substitution to find the solution.
The goal is to bring the augmented matrix to an easy-to-solve form, like
1 2 3 9
0 1 2 4 .
0 0 1
3
This matrix is in a row echelon form. To be of this form a matrix must have the
following three properties:
1. If there are rows that consist entirely of zeroes, then they are all at the bottom
of the matrix.
2. If a row does not consist entirely of zeroes, then the first nonzero number in the
row must be a 1. We will call it leading 1.
Z. G
onye
Linear Equations
3. In any two successive rows that do not consist entirely of zeroes, the leading
1 in the lower row occurs farther to the right than the leading 1 in the higher
row.
Example
1 ? ?
0 1 ?
0 0 1
0 0 0
?
1 ? ? ? ?
0 0 0 1 ?
?
,
0 0 0 0 1 ,
?
1
0 0 0 0 0
of matrices
0 1 ?
0 0 1
0 0 0
0 0 0
? ?
1 ? ?
? ?
,
0 1 ?
0 0
0 0 0
0 0
form:
? ?
? ?
1 ?
1 0 0
0 1 0
0 0 1
0 0 0
1.2.5. Here are some examples of matrices that are in reduced row echelon
0
0
,
0
1
1
0
0
0
?
0
0
0
?
0
0
0
0
1
0
0
0
0
,
1
0
0
0
0
0
1
0
0
0
0
1
0
0
?
?
0
0
?
?
,
0
0
1 0 ? 0 ?
0 1 ? 0 ?
0 0 0 1 ?
1.3 Examples
Note: This section is also available in txt format at
http://www.math.poly.edu/courses/ma2012/classnotes.phtml
Z. G
onye
1.3 Examples
that you can save (Save as ... File... choose a name, like class2.txt) and use with
MATLAB. Open MATLAB, set your current directory to your working directory.
Among the files you should see class2.txt. By clicking twice on this file you can open
it. MATLAB will open the file using its own text editor. If you want to try something
out, just cut and paste the command into the Command Window of the MATLAB.
Hit the enter key to evaluate the command. MATLAB commands are written as:
>> command
Example 1.3.1. Solve the following system of linear equations over the real numbers:
x + y + 4z = 15
2x + 4y 3z = 1
3x + 6y 6z = 3.
First write down the augmented matrix:
>> M=[1 1 4 15;2 4 -3 1;3 6 -6 -3]
1 1 4 15
M = 2 4 3 1 .
3 6 6 3
Then use the Gaussian algorithm to bring the augmented matrix to row-echelon or
reduced row-echelon form. Add 2 times the first row to the second:
>> M(2,:)=M(2,:)-2*M(1,:)
1 1 4
15
M = 0 2 11 29 .
3 6 6 3
Notice that the third row can be divided by three, so divide the third row by 3:
>> M(3,:)=M(3,:)/3
1 1 4
15
M = 0 2 11 29 .
1 2 2 1
Subtract the first row from the third:
>> M(3,:)=M(3,:)-M(1,:)
1 1 4
15
M = 0 2 11 29 .
0 1 6 16
Z. G
onye
Linear Equations
1 1 4
15
M = 0 1 6 16 .
0 2 11 29
Add 2 times the second row to the third
>> M(3,:)=M(3,:)-2*M(2,:)
1 1
M= 0 1
0 0
one:
4
15
6 16 .
1
3
Now you have the row-echelon form. From here you can get the solution by back
substitution: From the last row you have that z = 3. Second row says that y 6z =
16, so y = 2. First row says that x + y + 4z = 15, so x = 1.
Instead of the back substitution you can work more with the matrix to get the
reduced row-echelon form. Subtract the second row from the first one:
>> M(1,:)=M(1,:)-M(2,:)
1 0 10 31
M = 0 1 6 16 .
0 0 1
3
Add 10 times the third row the the first
>> M(1,:)=M(1,:)-10*M(3,:)
1 0
M = 0 1
0 0
row:
0
1
6 16 .
1
3
1 0 0 1
M = 0 1 0 2 .
0 0 1 3
Now you have the reduced row-echelon form. From this you can easily see the solutions. The first row says x = 1, the second row says y = 2, and the third row says
z = 3. So the solution of the system is: x = 1, y = 2, z = 3. You can also write down
the solution in the vector form: (1, 2, 3). Notice that the first number is the value of
Z. G
onye
1.3 Examples
x, the second is the value of y, and the third is the value of z. Keep the order of the
variables!
MATLAB can give you the reduced row-echelon form in one step, lets see how.
Enter the augmented matrix (same as above):
>> M=[1 1 4 15;2 4 -3 1;3 6 -6 -3]
1 1 4 15
M = 2 4 3 1 .
3 6 6 3
Ask MATLAB to find the reduced row-echelon form:
>> rref(M)
1 0 0 1
ans = 0 1 0 2 .
0 0 1 3
Thats it. The solution of the system is (1, 2, 3).
Example 1.3.2. Solve the system of linear equations over the real numbers:
2y + 3z v = 1
2x + 6y 4z + 10v = 8
3x + 5y 12z + 17v = 7
Enter the augmented matrix:
>> M=[0 2 3 -1 1;2 6 -4 10 8;3
M= 2
3
5 -12 17 7]
2 3 1 1
6 4 10 8 .
5 12 17 7
0 2 3 1
M = 1 3 2 5
3 5 12 17
1
4 .
7
1 3 2 5
M = 0 2 3 1
3 5 12 17
Z. G
onye
4
1 .
7
10
Linear Equations
1 3 2 5
4
3 1 1 .
M = 0 2
0 4 6 2 5
Divide the second row by two, to get a 1 for you
>> M(2,:)=M(2,:)/2
0
1.0000
1.5000
M=
0
4.0000 6.0000
5.0000
4.0000
0.5000 0.5000 .
2.0000 5.0000
There are fractions in this matrix, and MATLAB shows you the numbers up to the
4th decimal place. If you would rather see the matrix in rational form, then type:
>> format rat
>> M
1 3 2
5
4
M = 0 1 3/2 1/2 1/2 .
0 4 6
2
5
However, we note here, that MATLAB uses rationales to APPROXIMATE the exact
value. Sometimes it is a problem, please see Example 1 in the on-line MATLAB
manual for further explanation on this.
Add 4 times the second row to the third:
>> M(3,:)=M(3,:)+4*M(2,:)
1 3 2
5
4
M = 0 1 3/2 1/2 1/2 .
0 0 0
0
3
Look at the last row. That says: 0 x + 0 y + 0 z + 0 v = 3. There are no values
for (x, y, z, v) for which this can be true. So the system HAS NO SOLUTION. The
system is inconsistent.
Again we could have used MATLAB build-in function to get the reduced rowechelon form in one step:
Enter the augmented matrix:
>> M=[0 2 3 -1 1;2 6 -4 10 8;3 5 -12 17 7]
0 2 3 1 1
M = 2 6 4 10 8 .
3 5 12 17 7
Z. G
onye
11
1.3 Examples
Get the reduced row-echelon form:
>> rref(M)
1 0 13/2 13/2 0
ans = 0 1 3/2 1/2 0 .
0 0
0
0
1
Look at the last row and conclude that the system is inconsistent.
Example 1.3.3. Find all real solutions (p, q, r, s) of the system:
p + 2r = 0
2p 2q + 4r 3s = 1
q + 3s = 5
2p + 8q + 4r + 15s = 13.
Enter the augmented matrix:
>> M=[1 0 2 0 0;2 -2 4 -3 -1;0 1 0
1 0
2 2
M =
0 1
2 8
Bring to reduced row-echelon form:
>> rref(M)
1
0
ans =
0
0
0
1
0
0
3 5;2 8 4 15 13]
2 0
0
4 3 1
.
0 3
5
4 15 13
2
0
0
0
0 0
0 4
.
1 3
0 0
12
Linear Equations
So the solutions are: p = 2r, q = 4, r, s = 3. Writing this down with the vector
notation, the solutions are: (2r, 4, r, 3). The system has infinitely many solutions.
Example 1.3.4. For which value(s) of the constant k does the system
x + (k 4)y = k + 3
kx + (2k 3)y = 2
have
(a) no solution
(b) exactly 1 solution
(c) infinitely many solutions
over the field of real numbers?
First we have to teach MATLAB that k is a parameter (a symbolic variable):
>> syms k
>> M=[1 k-4 k+3;-k 2*k-3 2]
1,
k 4,
k+3
M=
.
k, 2 k 3,
2
Add k times the first row to the second row:
>> M(2,:)=M(2,:)+k*M(1,:)
1
k4
k+3
M=
.
0 2 k 3 + k (k 4) 2 + k (k + 3)
Factor the nonzero terms in the last row:
>> M(2,2)=factor(M(2,2));
>> M(2,3)=factor(M(2,3))
1
k4
k+3
M=
.
0 (k + 1) (k 3) (k + 2) (k + 1)
(a) The system has no solution if the last row becomes: 0,0,nonzero. So if (k + 1)
(k 3) = 0 but (k + 2) (k + 1) is not zero. This happens if k = 3.
(b) The system has exactly one solution, if the second row also has a leading one,
that is (k +1)(k 3) is not zero. Therefore the system has exactly one solution
if k is neither equal to 1 nor equal to 3. Notice that (k + 2) (k + 1) can be
anything, zero or not zero, you do still get exactly one solution.
Z. G
onye
13
(c) The system has infinitely many solutions, if the second row has no leading and
the system has solution(s). That is when both (k+1)(k3) and (k+2)(k+1)
are equal to 0. This happens when k = 1.
Remark 1.3.1. I do NOT suggest using the command rref(M) for this problem. Try
it, and see what happens! How would you answer the questions from that form? Why
dont you get the same answer for part (c)? You will not be able to answer part (c)
correctly, because MATLAB (and most of the calculators) will divide the second row
by (k + 1). You know that you cannot divide by (k + 1) if that is zero, MATLAB just
assumes that is not zero. Which is not correct.
(mod p),
if p divides b a.
Example 1.4.1.
1 1 (mod 2)
22 1 (mod 3)
12 2 (mod 5)
12 3 (mod 5)
Definition 1.4.2. Zp denotes the residue classes modulo p, that is
Zp = {0, 1, 2, . . . , p 1}.
Example 1.4.2.
Z2 = {0, 1}
Z3 = {0, 1, 2}
Z5 = {0, 1, 2, 3, 4}
You can add, subtract and multiply numbers in Zp , but no division!
Example 1.4.3.
Z. G
onye
14
Linear Equations
1 + 1 0 (mod 2)
2 + 2 1 (mod 3)
3 + 4 2 (mod 5)
3 4 4 (mod 5)
2 2 1 (mod 3)
4 4 1 (mod 5)
3 4 2 (mod 5)
6 3 4 (mod 7)
5 3 1 (mod 7)
1
2
1
is:
1 2 0
0 2 1 .
2 0 2
1 1
0 1
1 2
2 0
1 1 ,
0 2
1 1 2 0
0 1 1 1 ,
0 1 1 2
Z. G
onye
15
1 1 2 0
0 1 1 1 .
0 0 0 1
The last row says: 0 x + 0 y + 0 z = 1, which is not possible, therefore the system
is inconsistent and it has no solution in Z3 .
Example 1.5.2. Solve the following system in Z2 :
x1 + x2
x2 + x3 + x4
x1 + x4
x1 + x2 + x3
=1
=0
=0
= 0.
1
0
1
1
Using row operations in Z2 , add the
1
0
0
0
add the second row to the third row:
1
0
0
0
add the third row to the fourth:
1
0
0
0
1
1
0
1
0
1
0
1
0
1
1
0
1
0
.
0
0
1 0 0 1
1 1 1 0
,
1 0 1 1
0 1 0 1
1
1
0
0
0
1
1
1
0
1
0
0
1
0
,
1
1
1
1
0
0
0
1
1
0
0
1
0
0
1
0
,
1
0
This is now in row-echelon form. Notice that there are only 3 leading 1, so one of the
four unknowns will be free. We can use back-substitution to solve:
Z. G
onye
16
Linear Equations
The fourth column has no leading one, that means x4 is a free variable. That
means x4 is either 0 or 1, since there are the only elements in Z2 .
The first row says: x1 + x2 = 1. By solving for x1 : x1 = 1 + x2 (remember we
are calculating in Z2 !)
The second row says: x2 + x3 + x4 = 0. By solving for x2 : x2 = x3 + x4 .
The third row says: x3 = 1.
Z. G
onye
Chapter 2
Matrix Algebra
2.1 Matrix Arithmetic
Definition 2.1.1. A matrix is a rectangular array of numbers. We will always denote
matrices by capital letters. The numbers in the array are called entries in the matrix.
If we want to refer to the entry of matrix A standing in the ith row and jth column,
then we write (A)ij or aij . If a matrix has m rows and n columns, then the matrix
is said be of size m n. The entries of a matrix can be real numbers, then we say
the matrix is over R. If the entries are complex numbers, then the matrix is said to
be over C. If we want to refer to the entry of matrix A that is in the ith row and
jth column, we write (A)ij or aij . A matrix which has only one row is also called a
row matrix . A matrix which has only one column is also called a column matrix or
vector . A matrix with n rows and n columns is called a square matrix of order n, and
the entries a11 , a22 , . . . , ann are said to be on the main diagonal of the matrix.
Example 2.1.1.
2 4 1 0
A = 3 3 1 2 ,
0 2
1 0
1
0
C=
2 ,
7
B = 2 0 1 2 3 ,
D=
2 3
,
1 7
E= 5 .
18
Matrix Algebra
0
0
0
0 0 0 ,
0 ,
0 0 0 0 ,
0 0 0
0 0 0 0
0 0 0 0 ,
0 0 0 0
...
Definition 2.1.3. A square matrix whose entries along its main diagonal are all 1s
and whose all other entries are 0s is called an identity matrix . The n n identity
matrix is denoted by In .
Example 2.1.3. The following are examples for identity matrices:
1 0
1 0 0
1 0
0 1
I1 = 1 ,
I2 =
,
I3 = 0 1 0 ,
I4 =
0 1
0 0
0 0 1
0 0
0
0
1
0
0
0
,
0
1
...
Definition 2.1.4. Two matrices are equal if they have the same size and their corresponding entries are equal.
Definition 2.1.5. If two matrices, A and B, are of the same size, then the sum A+B
is the matrix obtained by adding the entries of B to the corresponding entries of A.
Definition 2.1.6. If two matrices, A and B, are of the same size, then the difference
A B is the matrix obtained by subtracting the entries of B from the corresponding
entries of A.
Example 2.1.4. Consider the matrices
2 1 0 1
A=
and
3 2 2 0
B=
0
2 3 1
.
1 5 3 7
Both matrices have the same size, 2 4, so the sum and difference are defined.
2
3 3 0
A+B =
,
2 7 5 7
and
AB =
2 1 3 2
.
4 3 1 7
Definition 2.1.7. If A is a matrix and c is a scalar, then the product cA is the matrix
obtained by multiplying each entry of the matrix A by c.
Z. G
onye
19
5A =
2 1 0 1
.
3 2 2 0
10 5 0 5
.
15 10 10 0
Theorem 2.1.1 (Properties of Matrix Arithmetic I). Assuming that the sizes
of the matrices are such that the indicated operations can be performed, the following
rules of matrix arithmetic are valid. Here A, B and C denote matrices, 0 is a zero
matrix, and a and b are scalars.
1. A + B = B + A
Commutative law for addition
2. A + (B + C) = (A + B) + C
Associative law for addition
3. a(B + C) = aB + aC = (B + C)a
4. a(B C) = aB aC = (B C)a
5. (a + b)C = aC + bC = C(a + b)
6. (a b)C = aC bC = C(a b)
7. a(bC) = (ab)C
8. A + 0 = 0 + A = A
9. A A = 0
10. 0 A = A
Definition 2.1.8. If A is an m r matrix and B is an r n matrix, then the matrix
product AB is the m n matrix whose entries are obtained as follows:
For the entry (AB)ij choose the ith row of matrix A and the jth column of matrix
B, and multiply the corresponding entries of this row and column together and then
add up the resulting products. (See Figure 2.1 for the sizes.)
With sum notation we can write
(AB)ij =
r
X
k=1
Z. G
onye
(A)ik (B)kj .
20
Matrix Algebra
A
m
B
r
AB
n
Inside
Outside
b11
b21
b31
b41
a11
a21
a31
a41
a51
a12
a22
a32
a42
a52
a13
a23
a33
a43
a53
a14
a24
a34
a44
a54
c11
c21
c31
c41
c51
b12
b22
b32
b42
b13
b23
b33
b43
c12
c22
c32
c42
c52
c13
c23
c33
c43
c53
1
2
B=
0
3
Then
AB =
1
, and
0 24
0 1
2 0
.
3 2
5 2 43
3 3 4
,
7 2
1 23
but BA is not defined since the number of columns in B and the number of rows in
A are not the same.
Z. G
onye
21
AB =
B = 5 3 ,
and
10 6
5 3
and
BA = 7 .
Definition 2.1.9. For a positive integer p and a square matrix A we define the pth
power of A by
Ap = AAA
| {z A},
p
22
Matrix Algebra
21. A0 = I
Definition 2.1.10. If A is any m n matrix, then the transpose of A, denoted by
AT , is defined to be the n m matrix, that is obtained by interchanging the rows
and columns of A.
Example 2.1.8. If
A=
2 1 0 1
,
3 2 2 0 24
then
2
3
1 2
.
AT =
0
2
1 0 42
Theorem 2.1.3 (Properties of Matrix Arithmetic III). Assuming that the sizes
of the matrices are such that the indicated operations can be performed, the following
rules of matrix arithmetic are valid. Here A and B denote matrices, and a is a scalar.
22. (A + B)T = AT + B T
23. (A B)T = AT B T
24.* (AB)T = B T AT
25. (AT )T = A
26. (aB)T = aB T
Definition 2.1.11. If A is a square matrix, then the trace of A, denoted by tr(A),
is defined to be the sum of the entries along the main diagonal of A.
Example 2.1.9. The trace of
A=
2 1 0 1
3 2 2 0
23
Theorem 2.1.4 (Properties of Matrix Arithmetic IV). Assuming that the sizes
of the matrices are such that the indicated operations can be performed, the following
rules of matrix arithmetic are valid. Here A and B denote matrices, and c is a scalar.
27. tr(A + B) = tr(A) + tr(B)
28.* tr(A B) = tr(A) tr(B)
29.* tr(cB) = c tr(B)
30.* tr(AB) = tr(BA)
Proof. Let
A=
. . . . . . . . . . . . . . . . . .
an1 an2 ann
and
B=
. . . . . . . . . . . . . . . . . ,
bn1 bn2 bnn
then tr(A) = a11 + a22 + + ann and tr(B) = b11 + b11 + + bnn .
Since
A+B =
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ,
an1 + bn1 an2 + bn2 ann + bnn
tr(A + B) = a11 + b11 + a22 + b22 + + ann + bnn
= a11 + a22 + + ann + b11 + b22 + + bnn
= tr(A) + tr(B).
This proves that tr(A + B) = tr(A) + tr(B). Similarly you can prove that tr(A B) =
tr(A) tr(B).
Since
cA =
. . . . . . . . . . . . . . . . . . . . . ,
can1 can2 cann
tr(cA) = ca11 + ca22 + + cann
= c(a11 + a22 + + ann )
= c tr(A).
Z. G
onye
24
Matrix Algebra
Finally, we would like to emphasize two very important things about matrix calculations.
Remark 2.1.2. For matrices the cancellation law usually does not hold!
Consider the matrices
0 1
1 1
A=
,
B=
,
0 2
3 4
Z. G
onye
25
2 5
,
3 4
and
D=
3 7
.
0 0
3 4
,
6 8
and
AD =
0 0
.
0 0
1 0 0
E1 = 0 1 0 ,
0 0 4
Z. G
onye
0 1 0
E2 = 1 0 0 ,
0 0 1
1 0 2
E3 = 0 1 0
0 0 1
26
Matrix Algebra
a11
a12
a13
a21 a22 a23
a22
a23 ,
E1 A = a21
E2 A = a11 a12 a13 ,
4a31 4a32 4a33
a31 a32 a33
and
Theorem 2.2.1. The matrix EA that results by multiplying a matrix A on the left
by an elementary matrix E is the same as the matrix that results by applying the
corresponding elementary row operation to A.
2 0
0
1/2 0
0
0 1/3 0 = 0 3
0 .
0 0 4
0 0 1/4
Z. G
onye
27
1/2 3/2
2 3
Example 2.3.3. The matrix B =
is an inverse of A =
,
0
1
0 1
because
2 3
1/2 3/2
1 0
AB =
=
,
0 1
0
1
0 1
and
BA =
1/2 3/2
2 3
1 0
=
.
0
1
0 1
0 1
0 0 0
3 1 4
2 5 7
is singular, it has no inverse, because the first row of the matrix product AB is always
a zero row, so AB cannot give the identity matrix.
Example 2.3.5. Every elementary matrix is invertible, and the inverse is also an
elementary matrix. The inverse of Scale(I, i, c) is Scale(I, i, 1c ). The inverse of
Swap(I, i, j) is Swap(I, i, j). The inverse of Shear(I, i, j, c) is Shear(I, i, j, c).
1 0
0 1
0 0
0
1
0
1
0
0
1
1
0
0
= 0
0
4
1
1 0
0
0 0
= 1
0 1
0
1
0 2
1
1 0
= 0
0 1
0
0
0
1
0 ,
0 1/4
1 0
0 0 ,
0 1
0 2
1 0 .
0 1
Example 2.3.6. Let A and B be two invertible matrices of the same size. Then
(B 1 A1 )(AB) = B 1 (A1 A)B = B 1 IB = B 1 B = I,
and
(AB)(B 1 A1 ) = A(BB 1 )A1 = AIA1 = AA1 = I.
This shows that the product AB is invertible, and
(AB)1 = B 1 A1 .
Z. G
onye
28
Matrix Algebra
29
1 2 1
A = 0 4 5
2 3 3
over R.
Construct the matrix (A|I) and use row operations to bring the left side to the
form of I3 .
1 2 1 1 0 0
0 4 5 0 1 0
2 3 3 0 0 1
Add two times the first row to the third row:
1 2 1 1 0 0
0 4
0 1 0 .
5
0 1 1 2 0 1
Multiply the third row by 1, and switch
1 2 1
0 1 1
0 4 5
1 0 0
2 0 1 .
0 1 0
1 2 1 1 0 0
0 1 1
2 0 1 .
0 0 1 8 1 4
Subtract two times the second row from the first:
1 0 3 3 0 2
0 1 1
2 0 1 .
0 0 1 8 1 4
Add three times the third row
second:
1
0
0
Z. G
onye
the the first, and subtract the third row from the
0 0 27 3 14
1 0 10 1 5 .
0 1 8 1
4
30
Matrix Algebra
27 3 14
A1 = 10 1 5 .
8 1
4
Example 2.4.2. Show that the matrix
1 0 2
B = 0 2 1
2 0 1
is singular in Z3 .
Construct the matrix (B|I) and use row
form of I3 .
1 0 2 1
0 2 1 0
2 0 1 0
0 0
1 0 ,
0 1
subtracting 2 times the first row from the third row (modulo 3):
1 0 2 1 0 0
0 2 1 0 1 0 .
0 0 0 1 0 1
The reduced row-echelon form of the matrix is not I3 , so B is singular in Z3 .
2
3 0
0
,
0 7
0
diagonal matrices:
0 0
2 0 0
0 6 0 .
4 0 ,
0 0
0 0 5
Definition 2.5.2. A square matrix in which the entries below the main diagonal are
all zero is called upper triangular .
Z. G
onye
31
2 0 3
2 7
3
3 4
0 4 0 ,
0 4 5 .
,
0 7
0 0 0
0 0 3
Definition 2.5.3. A square matrix in which the entries above the main diagonal are
all zero is called lower triangular .
Example 2.5.3. Some examples for lower triangular matrices:
2 0 0
2
0 0
3 0
0 4 0 ,
5 4 0 .
,
9 7
6 5 0
3 0 0
Definition 2.5.4. A matrix that is either upper triangular or lower triangular is
called triangular .
Example 2.5.4. Some examples for
2
3 0
0
,
9 7
0
triangular matrices:
0 0
2 7 0
0 4 7 .
4 0 ,
0 0
0 0 6
2 7
7 4
0 3
1 21 9
0
21 11 0 .
3 ,
9 0 17
6
32
Matrix Algebra
Example 2.5.7.
0
7 0
7 0 3 ,
0 3 0
0 21 9
21 0
0
9
0
0
are skew-symmetric matrices over R (and also over C). Notice that the entries along
the main diagonal must be zero.
Example 2.5.8.
1 1 0
1 0 1 ,
0 1 0
1 1 1
1 0 0
1 0 1
are skew-symmetric matrices in Z2 . Notice that the entries along the main diagonal
can be either 0 or 1, because 1 1 (mod 2).
Example 2.5.9. Show that: If A is an invertible skew-symmetric matrix, then A1
is also skew-symmetric.
Definition 2.5.7. A square matrix A is called nilpotent if Ak = 0 for some positive
integer k.
Example 2.5.10. The matrices
0 1 0
A = 0 0 1
0 0 0
and
0
7
B=
0
0
0
0
0
3
0
0
0 0
0 0
0 0
4 0
0 1
0
0
0
0
!
1 0 0
1 0 0
3
1
2
0 1
0 0 1 .
0 1 0 ,
2
,
,
3
1
1 0
2
2
0 1 0
0 0 1
Z. G
onye
Chapter 3
Determinant
3.1 The Determinant Function
We follow an intuitive approach to introduce the definition of determinant. We already have a function defined on certain matrices: the trace. The trace assigns a
number to a square matrix by summing the entries along the mail diagonal of the
matrix. So the trace is a function; its domain is the set of all square matrices, its
range is the set of numbers. We also showed that the trace is a linear function, that
is
tr(A + B) = tr(A) + tr(B)
tr(cA) = c tr(A)
where A and B are two n n matrices, and c is a constant.
The determinant is also a function which assigns a number to every square matrix.
So its domain will be again the set of square matrices, and its rage is the set of
numbers. The notation of the determinant of a matrix A is
det(A).
The determinant function has some nice properties, and we should emphasize two of
them at this point:
det(AB) = det(A) det(B)
det(I) = 1
where A and B are two n n matrices. We will see the other properties later in this
chapter.
34
Determinant
1 0
det(I2 ) = det
= 1.
0 1
Example 3.2.2.
3 2
det
= 3 4 2 (2) = 16.
2 4
Theorem 3.2.1. For any two 2 2 matrices A and B
det(AB) = det(A) det(B).
Proof. Let
A=
a11 a12
a21 a22
and
B=
b11 b12
.
b21 b22
Then
det(A) det(B) = a11 a22 a12 a21 b11 b22 b12 b21
= a11 a22 b11 b22 a11 a22 b12 b21 a12 a21 b11 b22 + a12 a21 b12 b21 .
(3.2.1)
Z. G
onye
35
1
.
det(A)
Using the definition of the determinant we can easily show some properties of the
determinant function for 2 2 matrices.
Corollary 3.2.3.
1 0
det
= 1.
0 1
Corollary 3.2.4. If a row or column of a matrix is 0, then the determinant of the
matrix is 0.
Corollary 3.2.5. If a matrix is triangular, then its determinant is the product of the
diagonal entries:
a 0
det
= ad,
c d
a b
det
= ad.
0 d
Corollary 3.2.6. If we swap two rows or two columns, then the determinant changes
its sign:
c d
a b
det
= det
,
a b
c d
b d
a b
det
= det
.
a c
c d
Z. G
onye
36
Determinant
and
~y =
x1
x2
!
.
We can draw these two vectors in the Descartes coordinate system, and we can see
that any two such vectors determine a parallelogram. (You may see Section 4.1 for
more about vectors.)
Example 3.3.1. The area of the parallelogram with vertices (0, 0), (1, 0), (1, 0.8)
and (2, 0.8) is 0.8, see Figure 3.1. This is the same as the determinant of the matrix
formed by the two column vectors which determine the parallelogram:
1 1
det
= 0.8.
0 0.8
Z. G
onye
37
(1,0.8)
(0,0)
(2,0.8)
(1,0)
38
Determinant
(x1+x2,y1+y2)
(y1,y2)
y2
x2
(x1,x2)
y1
x1
(a)
y2
y2
x2
x2
y1
y1
x1
(b)
y2
x2
A
A
x1
(c)
y2
x2
x1
y1
y1
(d)
x1
(e)
Z. G
onye
39
3. If B is the matrix that results when two rows or two columns of A are interchanged, then det(B) = det(A).
4. If B is the matrix that results when a multiple of one row is added to another
row or when a multiple of one column of A is added to another column, then
det(B) = det(A).
Some further properties of the determinant that we showed for 2 2 matrices and
remain true for larger matrices:
Theorem 3.4.2. Let A and B be n n matrices.
1. If A has a row or a column of zeroes, then det(A) = 0.
2. If A is a triangular matrix, then its determinant is the product of the diagonal
entries.
3. det(A) = det(AT ).
4. det(AB) = det(A) det(B).
5. A square matrix A is invertible if and only if det(A) 6= 0.
6. If A is invertible, then det(A1 ) =
1
.
det(A)
1
x
2
x
x3
Z. G
onye
1 1 1
1 1 1
.
x2 1 1
x3 x3 1
40
Determinant
1 1 1 1
1
1
1
1
x 1 1 1
0 1 x
1
1
det
2
x2 x2 1 1 = det 0
0
1x
1
x3 x3 x3 1
0
0
0
1 x3
= (1 x)(1 x2 )(1 x3 ).
Example 3.5.2. Since we already calculated the determinant of
1 1 1 1
x 1 1 1
A=
x2 x2 1 1 ,
x3 x3 x3 1
we can easily tell that this matrix is singular (has no inverse) over the real numbers
if the determinant is equal to zero, that is if x = 1.
Notice, that det(A) = 0 has two more complex roots, so over the complex numbers
4
2
A is singular not only for x = 1 but also for x = ei 3 , or x = ei 3 .
41
2k k + 1
det
6= 0,
k+6 k+3
that is when 2k(k + 3) (k + 1)(k + 6) = k 2 k 6 = (k 3)(k + 2) 6= 0, i.e. if
k 6= 2, 3. However if k = 2 or 3, then we have to use the Gaussian algorithm.
Z. G
onye
42
Determinant
If k = 2, then the system becomes
4x y = 2
4x + y = 3
1 1/4 1/2
0 0
5
,
1 2/3 1/3
0
0 0
.
In this case y is a free variable, and the system has infinitely many solutions over R
and also over C.
n
X
k=1
The recursive definition of the determinant using cofactor expansion along the jth
column of A:
det A = a1j C1j + a2j C2j + a3j C3j + anj Cnj .
With sum notation:
det(A) =
n
X
k=1
Z. G
onye
43
Here Cij denotes the cofactor of the entry aij : the determinant of the minor you get
from A by cancelling the ith row and jth column, with a plus or minus sign according
to the checkerboard:
+ +
+ +
+ +
+ +
.................
Definition 3.7.2. The cofactor matrix of A:
. . . . . . . . . . . . . . . . . . . . . . . . .
Cn1 Cn2 Cn3 Cnn
is the matrix you get by replacing each entry in A by its cofactor.
Definition 3.7.3. The adjoint matrix of A is the transpose of the cofactor matrix:
.
C
C
C
C
adj(A) =
13
23
33
n3
. . . . . . . . . . . . . . . . . . . . . . . . .
C1n C2n C3n Cnn
Theorem 3.7.1.
det(A)
0
0
0
0
det(A) 0
0
A adj(A) =
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . = det(A) In .
0
0
0 det(A)
Theorem 3.7.2. If A is an invertible matrix, then
A1 =
1
adj(A).
det(A)
1 1 0
A = 3 4 5 .
3 2 1
Z. G
onye
44
Determinant
To get the determinant of A we can use cofactor expansion. The first row would be
the best choice, since it has a zero in it and the other entries are 1, which makes the
calculations easier:
det(A) = a11 C11 + a12 C12 + a13 C13
4 5
3 5
= 1 det
+ 1 (1) det
+0
3 1
2 1
= 6 + 12
= 6.
The cofactor matrix of A is:
6 12 6
1 1
1 .
5 5 1
6 1 5
adj(A) = 12 1 5 .
6 1
1
Calculating A adj(A):
1 1 0
6 1 5
6 0 0
A adj(A) = 3 4 5 12 1 5 = 0 6 0 = det(A) I3 .
3 2 1
6 1
1
0 0 6
Therefore we can use the adjoint matrix to find the inverse of A:
6 1 5
1 1/6 5/6
1
1
1/6 5/6 .
adj(A) = 12 1 5 = 2
A1 =
det(A)
6
6 1
1
1 1/6
1/6
@
a11 @
a12 @
a13 a11 a12
@
@
@
@
@
@
a21 a@22 a@23 a@21 a22
@
@
@
@
@
@
a31 a32 a@
a@
a@
33
31
32
@
@
@
Z. G
onye
45
1 1
3 4
3 2
det
0 0
0 0
0 0
0 4 5 6
5 9 4 6
1
1
0
2
1
1 4 3 1
= det 3 4 5 det 1 det
5 2
0 1 0 2
3 2 1
0 0 2 1
0 0 5 2
= 6 (1) (1)
= 6.
Z. G
onye
46
Determinant
Z. G
onye
Chapter 4
Vector Spaces
4.1 Introduction to the Euclidean n-space
Definition 4.1.1. A vector is an ordered sequence of numbers:
u1
u2
~u = ..
.
un
where u1 , u2 , . . . un are the coordinates of the vector ~u.
We may write a vector in a row format to save space:
~u = (u1 , u2 , u3 , . . . , un ).
Example 4.1.1.
(3, 5) is a vector in R2 (in the plane)
(2, 3, 3) is a vector in R3 (in the space)
(3, 5, 7, 0, 1) R5 .
Let us consider a simple example. Nonzero vectors in R2 (the plane) can be
represented geometrically by directed line segments. For example, the vector
!
3
5
can be represented by the directed line segment starting at the origin and pointing
to (3, 5), see Figure 4.1. In general, the vector
!
x1
~x =
x2
48
Vector Spaces
(6,7)
(3,5)
x
x
(3,2)
(0,0)
(7,2)
x
(4,3)
Z. G
onye
49
50
Vector Spaces
x1
x1 + 3x2
T x2 = 2x1 + x2 + 2x3 .
x3
4x1 + 5x2 3x3
This is also a linear transformation, its standard matrix is:
1 3 0
A = 2 1 2 .
4 5 3
Then the transformation is just a multiplication by the matrix A, that is T (~x) = A~x:
x1
1 3 0
x1
2 1 2
x2 .
T x2 =
x3
4 5 3
x3
Why do we call the matrix the standard matrix of T ? If you calculate the image
of the standard coordinate vectors of Rn under the map T you will find the answer.
1
1
2 ,
T 0 =
0
4
Z. G
onye
51
T 1 = 1 ,
0
5
the second column of the matrix.
0
0
T 0 = 2 ,
1
3
the third column of the matrix. Actually this gives another method to find the
standard matrix of a linear transformation. We just have to calculate the image of
the standard basis vectors of Rn and put them in the column of the matrix.
Definition 4.2.3. Let T be a linear transformation with standard matrix A. This
transformation T is invertible if its matrix A is invertible (i.e. when det(A) 6= 0). The
inverse of the transformation T is denoted by T 1 , and its standard matrix is A1 .
Example 4.2.6. Let T : R3 R3 defined by
x1
1 1 0
x1
x2 .
T x2 = 3 4 5
x3
3 2 1
x3
1 1 0
In Example 3.7.1 we calculated that det 3 4 5 = 6 6= 0 , so the transformation
3 2 1
is invertible, and
w1
1 1/6 5/6
w1
1
w2 =
2
1/6 5/6
w2 .
T
w3
1 1/6
1/6
w3
Example 4.2.7. The transformation T : Rn Rm , T (~x) = ~0 is called the zero
transformation. Its standard matrix is the m n zero matrix.
Example 4.2.8. The transformation T : Rn Rn , T (~x) = ~x is called the identity
transformation. Its standard matrix is the n n identity matrix.
Example 4.2.9. The projection from R3 into the xy-plane is also a linear transformation:
x
x
3
3
P : R R , P y = y .
z
0
Z. G
onye
52
Vector Spaces
Example 4.2.10. The rotation on the xy-plane by angle is also a linear transformation, and its standard matrix is:
cos sin
.
sin cos
For example,
2
x
x
3/2
1/2
R
=
y
y
1/2
3/2
R:R R ,
T :R R ,
3
1
x
x
T
=
y
y
1
3
53
4.4 Subspaces
8. The scalar multiplication is distributive: (k + l)~u = k~u + l~u.
9. The scalar multiplication is associative: k(l~u) = (kl)~u.
10. Multiplication by the unit: 1~u = ~u.
Example 4.3.1. Here are some examples for real vector spaces:
n o
V = ~0
V =R
V = R2
V = R3
V = Rn
V ={ all the m n real matrices }
V ={ all the 2 2 real matrices }
V ={ all continuous functions on R }
V ={ all polynomials of degree n }
4.4 Subspaces
Definition 4.4.1. A subset W of a vector space V is called a subspace of V if W is
itself a vector space under addition and scalar multiplication. We have to check the
following three things:
1. The zero vector of V is in W .
2. For any two elements ~u and ~v from W : ~u + ~v W .
3. If ~u W and k is any scalar, then k~u W .
Example 4.4.1. Let
V = P = {all polynomials} and
W = Pn = {polynomials of degree n}.
Z. G
onye
54
Vector Spaces
1. The zero vector of P is the zero polynomial, which has degree n, so it is
also in Pn .
2. The sum of two polynomials of degree n is also a polynomial of degree n,
so Pn is closed for addition.
3. If an xn + an1 xn1 + + a0 Pn , and k is a scalar, then k(an xn + an1 xn1 +
+ a0 ) = kan xn + kan1 xn1 + + ka0 is also in Pn .
Therefore Pn is a subspace of P.
Example 4.4.2. Let
V = {all n n real matrices} and
W = {all n n upper triangular matrices}.
1. The zero vector in V is the n n zero matrix. The zero matrix is upper
triangular, so it is in W .
2. The sum of two upper triangular matrices is also upper triangular, so W is
closed for addition.
3. k times an upper triangular matrix is also upper triangular, so W is closed for
scalar multiplication.
Therefore W is a subspace of V .
Example 4.4.3. Any plane through the origin is a subspace of R3 .
Example 4.4.4. Let V = R3 . Show that
W = {(a, b, c) : a2 + b2 + c2 1}
is not a subspace of V . The zero vector is in W . But you can find two vectors in
W whose sum is no longer in W . For example: (1, 0, 0) + (0, 1, 0) = (1, 1, 0) 6 W .
Or, you can find a vector in W and a scalar, so that the product is not in W . For
example: 2 (1, 0, 0) = (2, 0, 0) 6 W .
Example 4.4.5. Let V be the vector space of functions f : R R. Show that
W = {f (t) : f (t) = f (t)}
is a subspace of V . Notice that W is the set of all even functions.
Z. G
onye
55
4.4 Subspaces
1. The zero vector in V is the zero function: f (t) = 0. This is an even function,
so it is also in W .
2. Let f and g be two functions from W , i.e. f and g are even functions. Is the
sum h(t) = f (t) + g(t) in W , i.e. is h even?
h(t) = f (t) + g(t) = f (t) + g(t) = h(t)
so h = f + g is in W . W is closed for addition.
3. Let f be a function from W (i.e. f is an even function) and k be a scalar. Is
h(t) = k f (t) also in W , i.e. is h even?
h(t) = kf (t) = kf (t) = h(t)
so h = kf (t) is in W . W is closed for scalar multiplication.
Therefore W is a subspace of V .
56
Vector Spaces
Therefore W is a subspace of V .
x1
Problem 4.4.8. Let W = {~x : |x1 | = |x2 |}, where ~x =
. The set W is a subset
x2
of R2 , but is W a subspace of R2 ?
1
Solution. We can show that W is not closed for addition. Take for example ~u =
1
3
4
and ~v =
from W . Then ~u + ~v =
does not belong to W , because
3
2
|4| =
6 |2|.
4.5 Spanning
Definition 4.5.1. Take a set of vectors {~v1 , ~v2 , . . . , ~vr } from a vector space V . Define
W as all the linear combinations of the vectors ~v1 , ~v2 , . . . , ~vr . That is all vectors in
the form
k1~v1 + k2~v2 + + kr~vr
where k1 , k2 , . . . , kr are arbitrary scalars. Then W is called the space spanned by
~v1 , ~v2 , . . . , ~vr . Notation: W = span{~v1 , ~v2 , . . . , ~vr }.
Example 4.5.1. span{1, t, t2 , t3 , . . . , tn } = Pn
Example 4.5.2. span{(1, 0, 0), (0, 1, 0), (0, 0, 1)} = R3 .
Example 4.5.3. Is the vector u = (8, 4, 3) in the span of {(1, 2, 1), (2, 0, 1)}? It is
the same question as: Can we write the vector u as the linear combination of (1, 2, 1)
and (2, 0, 1)? Are there constants a and b so that
a(1, 2, 1) + b(2, 0, 1) = (8, 4, 3)?
This is a system of linear equations:
a + 2b = 8
2a = 4
a b = 3
whose augmented matrix is
8
1 2
2 0
4 .
1 1 3
You can solve this system using the Gaussian algorithm, the system has no solution,
which means that u is not in the span{(1, 2, 1), (2, 0, 1)}.
Z. G
onye
57
Example 4.5.5. The vectors ~v1 = (1, 0, 2), ~v2 = (0, 1, 0), and ~v3 = (2, 1, 4) do
not span R3 .
~v3 = (3, 2, 1)
1
5 3
A = 2 6 2 .
3 1 1
Since det(A) = 0, the matrix A is not invertible, so the homogeneous matrix equation
A~k = ~0 has infinitely many solution. For example: k1 = 1, k2 = 1, k3 = 2 is also a
solution:
1~v1 + 1~v2 2~v3 = ~0.
Example 4.6.2. The set of vectors
~v1 = (1, 2, 0),
~v3 = (3, 2, 1)
58
Vector Spaces
only if the scalars k1 , k2 , and k3 are all equal to 0. We can easily see this if we write
down the coefficient matrix for the unknowns k1 , k2 , k3
1
5 0
A = 2 6 2 .
3 1 1
Since det(A) 6= 0, the matrix equation A~k = ~0 only has the trivial solution, that is
k1 = 0, k2 = 0, k3 = 0.
So if the coefficient matrix is a square matrix, then the determinant is a good
device. If the coefficient matrix in not a square matrix, then we have to use the
Gaussian algorithm to determine how many solutions the vector equation has.
Example 4.6.3. We would like to determine the values of the parameter p for which
the set of vectors
~v1 = (1, 0, 2, 4),
~v3 = (c, 3, 2, 4)
1 0 c 0
0 1 3 0
A=
2 0 2. 0
4 0 4 0
Using the Gaussian algorithm we can get a row echelon form:
1 0
c
0
0 1
3
0
.
0 0 c + 1 0 .
0 0
0
0
The vectors are linearly independent, if this system has only the (0, 0, 0) solution,
that is when c 6= 1.
Example 4.6.4. The standard basis vectors ~e1 = (1, 0, 0), ~e2 = (0, 1, 0), ~e3 = (0, 0, 1)
of R3 are linearly independent.
Z. G
onye
59
Example 4.6.5. The vectors ~u = (2, 0, 1, 3, 5) and ~v = (4, 0, 2, 6, 10) are linearly dependent, because 2~u + ~v = ~0. Notice, that two vectors are linearly dependent
if one is a constant multiple of the other.
Example 4.6.6. The set {1, t, t2 , t3 } is linearly independent in P3 .
Example 4.6.7. If r > n then the set any set of r vectors from Rn is linearly
dependent.
Example 4.6.8. The set of vectors S = {~v1 , ~v2 , . . . , ~vk , ~0} is linearly dependent.
~v3 = (3, 3, 4)
60
Vector Spaces
1 0 0
1
+ c2
0
0 0 0
1
+ c5
1
1 0
1
+ c3
0 0
0
1 1
1
+ c6
1 0
1
1 1
1
+ c4
0 0
1
1 1
0 0
=
1 1
0 0
1 1
+
0 0
0
0
=0
=0
=0
=0
=0
= 0.
We can easily solve this system and see that the only solution is the trivial (zero)
solution. In more difficult cases, form the augmented matrix
1 1 1 1 1 1 0
0 1 1 1 1 1 0
0 0 1 1 1 1 0
0 0 0 1 1 1 0
0 0 0 0 1 1 0
0 0 0 0 0 1 0
Z. G
onye
4.8 Column Space, Row Space, Null Space, Rank and Nullity
61
and use the Gauss-Jordan elimination to solve the system. Since the coefficient matrix
1 1 1 1 1 1
0 1 1 1 1 1
0 0 1 1 1 1
A=
0
0
0
1
1
1
0 0 0 0 1 1
0 0 0 0 0 1
is a square matrix we can also use the determinant. The determinant of A is equal to
1, therefore the coefficient matrix A is invertible, so the homogeneous system A~c = ~0
only has the trivial solution.
The determinant of the coefficient matrixA easily gives answer for our other question: Does the set S span M23 ? Since the determinant of A is nonzero, so the
coefficient matrix A is invertible, therefore the matrix equation A~c = ~b has a unique
solution for any vector ~b. This shows that the set S spans M23 .
4.8 Column Space, Row Space, Null Space, Rank and Nullity
Column space
How can we find a basis for the subspace spanned by vectors which are not linearly
independent?
Example 4.8.1. Lets show first that the vectors
~v1 = (1, 2, 1, 0),
~v4 = (1, 1, 1, 2)
are not linearly independent. That is we have to show that the vector equation
k1~v1 + k2~v2 + k3~v3 + k4~v4 = ~0
has a nonzero solution. The augmented matrix is:
1 2 0 1 0
2 1 3 1 0
1 2 0 1 0 .
0
1 1 2 0
Z. G
onye
62
Vector Spaces
You can either calculate the determinant, or you can bring this augmented matrix to
row-echelon form:
1 2 0 1 0
0 1 1 1 0
0 0 0 1 0 ,
0 0 0 0 0
from this k3 is a free variable, so the vector equation
k1~v1 + k2~v2 + k3~v3 + k4~v4 = ~0
has nonzero solution, and therefore the set of vectors {~v1 , ~v2 , ~v3 , ~v4 } is linearly dependent.
How can we find a basis (a set of linearly independent vectors) for the subspace
they span? We will use the Gaussian algorithm to find a basis. Form a matrix
from these vectors, so that the columns of the matrix are the given vectors. Using
elementary row operations find the row-echelon form of the matrix. Then the column
vectors from the original matrix which correspond to the columns with the leading
1 of the row-echelon matrix form a basis spanned by the original (column) vectors
(i.e. form a basis for the column space of the original matrix).
Definition 4.8.1. Let A be an m n matrix. The vectors formed by each of the
columns of A are called the column vectors of A. The space spanned by the column
vectors is the column space of A.
So form the matrix from the given vectors:
1 2 0 1
2 1 3 1
A=
1 2 0 1 .
0
1 1 2
We already found its row-echelon form:
1
0
R=
0
0
2
1
0
0
0 1
1 2
.
0 1
0 0
A basis for the space spanned by the vectors ~v1 , ~v2 , ~v3 , and ~v4 are the columns from the
original matrix A which correspond to the columns with leading 1 in the row-echelon
matrix R, that is the first, second and fourth columns:
1
2
1
2 1 1
, , .
1 2 1
0
1
2
Z. G
onye
4.8 Column Space, Row Space, Null Space, Rank and Nullity
63
Row space
Definition 4.8.2. The vectors formed by the rows of a matrix A are called row
vectors. The space spanned by these row vectors is called the row space of the matrix
A.
Example 4.8.2. Find a basis for
1
2
B=
2
1
3 4 2 5
4
6 9 1 8
2
,
6 9 1 9
7
3 4 2 5 4
Again, use elementary row operations to bring the matrix into row-echelon form: So
form the matrix from the given vectors:
1 3 4 2 5
4
0 0 1 3 2 6
.
R=
0 0 0 0
1
5
0 0 0 0
0
0
The nonzero row vectors of R a basis for the row-reduced matrix, and also for the
original matrix B. The basis vectors for the row space of B are:
~r1 = (1, 3, 4, 2, 5, 4) ~r2 = (0, 0, 1, 3, 2, 6) ~r3 = (0, 0, 0, 0, 1, 5).
Null space
Definition 4.8.3. The solution space of the homogeneous system A~x = ~0 is called
the null space of A. The dimension of the null space is called the nullity of A and is
denoted by nullity(A).
Example 4.8.3. Find the null space of the system
x+y =0
3x y + 4z = 0.
The augmented matrix is
Z. G
onye
1 1 0 0
3 1 4 0
1 1 0 0
0 1 1 0
64
Vector Spaces
=0
=0
=0
= 0.
1
1 2 0 1 0
2
2 1 0
1 0
1 1 2 3 1 0
0
0
1
1
1 0
its row-echelon form is
1
0
0
0
1 2 0 1 0
0 1 0 1 0
,
0 0 1 0 0
0 0 0 0 0
4.8 Column Space, Row Space, Null Space, Rank and Nullity
65
1 1
A = 2 3
3 2
Find a basis for its row space. Find a basis for its solution space. Find its solution
space, a basis for the solution space. What is rank(A), and nullity(A)? (Remark: the
system A~x = ~0 is so called over-determined, because there are more equations then
unknowns.)
First we have to solve the system A~x = ~0 so we set up the augmented matrix:
1 1 0
2 3 0 .
3 2 0
Its row-echelon form is
1 1 0
0 1 0 .
0 0 0.
3
2
those column of A which corresponds to the columns of the row-echelon matrix with
the leading 1.
The only solution of the system is the zero solution: (0, 0), so the null space
consists of the zero vector only:
nullspace = {(0, 0)}.
There is no basis for the null space. Therefore, rank(A) = 2, nullity(A) = 2 2 = 0.
Z. G
onye
66
Vector Spaces
Z. G
onye
Chapter 5
Eigenvalues and Eigenvectors
5.1 Eigenvalues and Eigenvectors
Let T : Rn Rn be a linear transformation. Then T can be represented by a matrix
(the standard matrix), and we can write
T (~v ) = A~v .
Example 5.1.1. Consider the transformation T : R2 R2 given by its standard
matrix
3 1
A=
,
5 3
and lets calculate the image of some vectors under the transformation T .
3 1
1
2
1
T
=
=
,
1
5 3
1
2
3
3 1
3
6
T
=
=
.
3
5 3
3
6
We may notice that the image of (x, x) is 2(x, x). Lets calculate some more images:
1
3 1
1
2
T
=
=
,
5
5 3
5
10
2
3 1
2
4
=
=
.
T
10
5 3
10
20
We may notice that the image of a vector (x, 5x) is 2(x, 5x). A couple of more
images:
2
3 1
2
3
T
=
=
,
3
5 3
3
1
1
3 1
1
4
T
=
=
.
8
1
5 3
1
There are no such nice patterns for these vectors.
68
69
5.2 Examples
5.2 Examples
Example 5.2.1. Let
0 0 2
A = 1 2 1
1 0 3
The characteristic polynomial of A is
0
2
1 = 3 + 52 8 + 4.
det(A I) = det 1 2
1
0
3
To get the eigenvalues, find the zeroes of the characteristic polynomial:
3 + 52 8 + 4 = ( 2)2 ( 1),
the eigenvalues of A are: = 2, which has algebraic multiplicity of 2, (that is = 2 is a
double root of the characteristic equation) and = 1, which has algebraic multiplicity
of 1 (that is = 1 is a simple root of the characteristic equation).
Lets find the eigenspace and a basis for the eigenspace for each of the eigenvalues
of A. To find the eigenspace corresponding to , we have to find the solutions space
of the equation (A I)~v = ~0. So for = 2, the augmented matrix is:
2 0 2 0
1 0 1 0
1 0 1 0
whose row-echelon form is
Z. G
onye
1 0 1 0
0 0 0 0 .
0 0 0 0
70
Since there are two free variables the solution space of (A 2I)~(v) = ~0, and therefore
the eigenspace of A corresponding to = 2 has dimension two. The geometric multiplicity of the eigenvalue = 2 is 2 (the dimension of the corresponding eigenspace).
The solutions of (A 2I)~v = ~0 are (v3 , v2 , v3 ) where v2 and v3 are free variables.
These are the eigenvectors of A corresponding to = 2. The eigenspace corresponding
to = 2 is
v3
v 2 : v 2 , v3 C .
v3
A basis for the eigenspace is
1
0
1 , 0 .
0
1
To find the eigenspace corresponding to = 1 we have to repeat the same procedure. We have to find the solutions space of the equation (A 1I)~v = ~0, the
augmented matrix is:
1 0 2 0
1 1 1 0 ,
1 0 2 0
whose row-echelon form is
1 0 2 0 0
0 1 1 0 0 .
0 0 0 0 0
Since there is only one free variable the solution space of (A 1I)~v = ~0, and therefore
the eigenspace of A corresponding to = 1 has dimension one. The geometric multiplicity of the eigenvalue = 1 is 1. The solutions of (A I)~v = ~0 are (2v3 , v3 , v3 ),
where v3 is a free variable. These are the eigenvectors of A corresponding to = 1.
The eigenspace corresponding to = 1 is
2v3
v3 : v3 C .
v3
A basis for the eigenspace is
2
1 .
1
Z. G
onye
71
5.2 Examples
Example 5.2.2. Let
5 0 4
B = 0 3 1 .
0 0 2
It is a triangular matrix. The eigenvalues of B are = 5, 3, and 2. Each has
algebraic multiplicity of one. For each eigenvalue we can find the eigenspace, and a
basis for the eigenspace. The eigenspace corresponding to = 5 is
v1
0 : v1 C .
0
A basis for the eigenspace is
1
0 .
0
The eigenspace corresponding to = 3 is
v 2 : v 2 C .
0
A basis for the eigenspace is
0
1 .
0
The eigenspace corresponding to = 2 is
7
1
5 v3 : v3 C .
v3
A basis for the eigenspace is
4
7
1
5
1
or a more convenient one is: (20, 7, 35).
Z. G
onye
72
5 1
.
1 3
1 1 0
.
0 0 0
Since there is only one free variable the solution space, and therefore the eigenspace
corresponding to = 4 has dimension one. The geometric multiplicity of the eigenvalue = 4 is 1. The solutions, so the eigenvectors are (v2 , v2 ), where v2 is a free
variable. The eigenspace corresponding to = 4 is
v2
: v2 C .
v2
A basis for the eigenspace is
1
.
1
Example 5.2.4. Let
2 3 1
N = 0 4 0 ,
0 0
1
a triangular matrix. Then the matrix N I
2
3
1
4
0
N = 0
0
0
1
Z. G
onye
5.3 Diagonalization
73
5.3 Diagonalization
Definition 5.3.1. A square matrix is called diagonalizable if there exists an invertible
matrix P so that P 1 AP is diagonal.
74
1 0 . . . 0
0 ... 0
2
D=
. . . . . . . . . . . . . . .
0 0 . . . n
where ~v1 corresponds to 1 , ~v2 corresponds to 2 , and so on.
We will look at the three examples we did in Section 5.2, and see whether the
matrices A, B, and C are diagonalizable.
Example 5.3.1.
0 0 2
A = 1 2 1
1 0 3
is
diagonalizable,
1 0 2
P = 0 1 1
1 0 1
and
2 0 0
D = 0 2 0 .
0 0 1
Note: Since we could have found anther basis for the eigenspaces, this matrix P
is not unique.
Example 5.3.2. The matrix
5 0 4
B = 0 3 1
0 0 2
is also diagonalizable, because we found 3 basis vectors for the eigenspaces combined.
Therefore
1 0 20
P = 0 1 7
0 0 35
Z. G
onye
75
5.3 Diagonalization
and
5 0 0
D = 0 3 0 .
0 0 2
5 1
.
1 3
is not diagonalizable, because it only has one basis vector for its eigenspace(s).
Example 5.3.4. If all eigenvalues are different, then the matrix is diagonalizable,
because for each eigenvalue there will be one basis vector for the corresponding
eigenspace. For example:
0 1 0
M = 0 0 1
4 17 8
4
0
0
0 .
D = 0 2 + 3
0
0
2 3
Example 5.3.5. The triangular matrix
2 3 1
N = 0 4 0 ,
0 0
1
has eigenvalues = 2, 4 and 1. The eigenvalues of N are all different, so N is
diagonalizable, and D can be
2 0 0
0 4 0 .
0 0 1
To find the corresponding matrix P , to each eigenvalue you will have to find a corresponding eigenvector.
Example 5.3.6. The matrix
Z. G
onye
0 2
3 0
76
has complex eigenvalues, = 6i. In the diagonal form we would see these complex
entries. Since the diagonal form is not a matrix over R, we say this matrix is not
diagonalizable over R.
2 0 0
D = 0 2 0
0 0 1
is
D100
2100 0 0
= 0 2100 0 .
0
0 1
77
0
A = 1
1
power of
0 2
2 1 .
0 3
1 0 2
P = 0 1 1 ,
1 0 1
and then
2 0 0
D = 0 2 0 .
0 0 1
So
A15 = P D15 P 1
1
15
1 0 2
1 0 2
2 0 0
= 0 1 1 0 2 0 0 1 1
1 0 1
1 0 1
0 0 1
15
2
0 0
1 0 2
1 0 2
= 0 1 1 0 215 0 1 1 1
1 0 1
0
0 1
1 0 1
15
16
22
0 22
15
= 2 1 215 215 1 .
215 1 0 216 1
For further applications you may see Section A.2 in the appendix.
Z. G
onye
78
Z. G
onye
Chapter 6
General Linear Transformations
6.1 Linear Transformations from Rn to Rm
We studied linear transformations in Section 4.2 and used the following definition to
determine whether a transformation from Rn to Rm is linear.
Definition 6.1.1. A transformation T : Rn Rm is linear if the following two
properties hold for all vectors ~u and ~v in Rn and every scalar c:
(a) T (~u + ~v ) = T (~u) + T (~v ), and
(b) T (c~u) = cT (~u)
Example 6.1.1. Using this theorem we can show for example that
1. T : R2 R2 defined by T (x, y) = (2x, x y) is linear transformation.
2. T : R2 R2 defined by T (x, y) = (x2 , y) is not linear.
80
Example 6.2.1. Let V be the vector space of all real-valued functions that are
differentiable, and let W be the vector space of all real-valued functions. Define
D : V W by
D(f ) = f 0
where f 0 is the derivative of f . We will show that this transformation is a linear
transformation.
(a) Checking addition. Take two functions f and g from V , then
D(f + g) = (f + g)0 = f 0 + g 0 = D(f ) + D(g).
(b) Checking scalar multiplication. Take a function f form V and let c be a scalar.
Then
D(cf ) = (cf )0 = cf 0 = cD(f ).
Both of the properties of Definition6.2.2 are satisfied, so D is a linear transformation.
Actually, this transformation is called the differential operator .
Example 6.2.2. Let V be the vector space of all real-valued functions that are
integrable over the interval [a, b]. Let W = R. Define I : V W by
Z b
I(f ) =
f (x) dx.
a
(b) Checking scalar multiplication. Take a function f form V and let c be a scalar.
Then
Z b
Z b
I(cf ) =
cf (x) dx = c
f (x) dx = cI(f ).
a
81
(a) Checking addition. Take two polynomials from P2 (t), say p(t) = at2 + bt + c,
and q(t) = t2 + t + . Then
T p(t) + q(t) = T (at2 + bt + c + t2 + t + )
= T (a + )t2 + (b + )t + c +
(6.2.1)
= t (a + )t2 + (b + )t + c + + 1
= (a + )t3 + (b + )t2 + (c + ) + 1.
But
T p(t) + T q(t) = T (at2 + bt + c) + T (t2 + t + )
= t(at2 + bt + c) + 1 + t(t2 + t + ) + 1
= (a + )t3 + (b + )t2 + (c + ) + 2.
(6.2.2)
Since 6.2.1 and 6.2.2 are not equal, the first property of the definition 6.2.2
fails. Therefore this is not a linear transformation.
(As a practice, show that
the other property, T cp(t) = cT p(t) , also fails.)
Example 6.2.4. Let P4 (x) be the vector space of all real polynomials of degree 4 or
less. Let T : P4 (x) P4 (x) be given by
T =
d
d2
+3 .
2
dx
dx
First, lets just see how to calculate the image of a polynomial, say the image of
3x4 5x2 7x + 10. That is, we have to find T (3x4 5x2 7x + 10).
T (3x4 5x2 7x + 10) = (3x4 5x2 7x + 10)00 + 3(3x4 5x2 7x + 10)0
= 36x2 10 + 3(12x3 10x 7)
= 36x3 + 36x2 30x 31.
Is this transformation linear? We have to check the two properties given in the
definition 6.2.2.
(a) Checking addition. Let p(x) and q(x) be two polynomials from P4 (x).
d2
d
(p(x) + q(x)) + 3 (p(x) + q(x))
2
dx
dx
d2
d2
d
d
= 2 p(x) + 2 q(x) + 3
p(x) + q(x)
dx
dx
dx
dx
2
2
d
d
d
d
= 2 p(x) + 3 p(x) + 2 q(x) + 3 q(x)
dx
dx
dx
dx
= T (p(x)) + T (q(x)).
T (p(x) + q(x)) =
Z. G
onye
82
(b) Checking scalar multiplication. Let p(x) be a polynomial from P4 (x), and let k
be any constant.
d2
d
(kp(x))
(kp(x))
+
3
dx2
dx
d
d2
= k 2 p(x) + 3k p(x)
dx
dx
= kT (p(x)).
T (kp(x)) =
83
Using this, decode T (x4 ), T (x3 ), T (x2 ), T (x), T (1), the vectors we get are the columns
of the standard matrix:
0 0 0 0 0
12 0 0 0 0
.
12
9
0
0
0
A=
0 6 6 0 0
0 0 2 3 0
1 0 0 0 0
0 1 0 0 0
rref(A) =
0 0 1 0 0 ,
0 0 0 1 0
0 0 0 0 0
so its null space is {(0, 0, 0, 0, c) : c R}, and a basis for this null space is (0, 0, 0, 0, 1).
Using the decoding again, the vector (0, 0, 0, 0, c) corresponds to 0 x4 + 0 x3 + 0
x2 + 0 x + c 1 = c, which is the constant polynomial c. So
Ker(T ) = {all constant polynomials}.
The nullity(T ) is the dimension of the kernel of T , therefore nullity(T ) = 1.
Definition 6.4.2. The range of a transformation T : V W is the set of all vectors
w
~ W for which there is a vector ~v V such that T (~v ) = w.
~
Z. G
onye
84
For linear transformation on Euclidean spaces, the range is equal to the column
space of the standard matrix of the transformation. To find the range of a general
linear transformation T , we can use its matrix representation.
So in our example, lets find the column space of its matrix representation. The
first 4 columns of A form a basis for the column space, so
0
0
0
0
0
0
0
12
column space = a 12 + b 9 + c 0 + d 0 : a, b, c, d R .
0
6
6
0
0
0
2
3
A basis for the column space is:
0
0
0
0
12 0 0 0
12 , 9 , 0 , 0 .
0 6 6 0
0
2
3
0
Remark 6.4.1. We could choose a simpler basis:
0
0
0
0
1 0 0 0
1 , 3 , 0 , 0 .
0 2 3 0
0
0
1
1
To get the range of T you have to decode the column space of A.
Range(T ) = a(12t3 + 12t2 ) + b(9t2 + 6t) + c(6t + 2) + d 3 : a, b, c, d R .
A basis for the range is:
12t3 + 12t2 , 9t2 + 6t, 6t + 2, 3 .
Z. G
onye
Appendix A
Supplementary Material
A.1 Cayley-Hamilton Theorem
Theorem A.1.1. Every matrix is a root of its characteristic polynomial.
Example A.1.1. Let q(t) = (t 4)4 t2 + 8t. Evaluate q(C), where
5 1
C=
.
1 3
The characteristic equation of the matrix is 2 8 + 16. Cayley-Hamilton Theorem
says that the matrix C is a root of this polynomial, which means if we plug C in to
the equation, then we get zero: C 2 8C + 16I. So we try to rearrange the given
polynomial q(t):
2
q(t) = (t 4)4 t2 + 8t = (t 4)2 (t2 8t + 16) + 16.
Using the Cayley-Hamilton theorem:
2
q(C) = (C 4I)2 (C 2 8C + 16I) + 16I = 16I =
16 0
.
0 16
An
A2 A3
+
+ +
+
2!
3!
n!
86
Supplementary Material
(At)2 (At)3
(At)n
+
+ +
+
2!
3!
n!
(a) If A is a diagonal matrix, then it is very easy to calculate eA . For example, let
!
3 0
A=
.
0 2
eAt = I + At +
Then
eA =
=
=
!
1 0
+
0 1
1+3+
!
3 0
+
0 2
(3)2
2!
0
(2)2
2!
0
e3 0
0 e2
(3)2
2!
!
+
0
1+2+
(2)2
2!
(3)3
3!
0
!
0
(2)3
3!
!
+
7 0 0
Problem A.2.1. What is eB , where B = 0 0 0 ?
0 0 3
Solution.
e 0 0
eB = 0 1 0 .
0 0 e3
eA = I + A +
= P eD P 1
Z. G
onye
87
5 3 0
A = 0 3 1 .
0 0 2
Find the eigenvalues of A. For each of its eigenvalue, find a basis for the
corresponding eigenspace. Explain why the matrix A is diagonalizable, find its
diagonal form D and the matrix P so that A = P DP 1 . Calculate eA .
Solution. The eigenvalues
ofA are 5, 3, and 2. A basis for the eigenspace
1
corresponding to = 5 is 0.
0
3
1 3 1
5 0 0
P = 0 2 1 , and D = 0 3 0 .
0 0 1
0 0 2
Note: Depending the basis vectors you chose, you could have different matrix
for P .
5
1 3 1
e 0 0
2 3 1
1
1
eA = P eD P 1 = 0 2 1 0 e3 0 0 1
2
2
0 0
2
0 0 1
0 0 e
5
e 32 e5 + 32 e3 21 e5 + 32 e3 e2
e3
e3 e2
= 0
2
0
0
e
(c) Let
2 1 0
C= 0
0
1 .
0 4 4
Find the eigenvalue(s) of C. For each eigenvalue find a basis for the corresponding eigenspace. Explain why the matrix C is not diagonalizable.
Z. G
onye
88
Supplementary Material
In the case when the matrix C has only one eigenvalue , there is a trick to
calculate its exponent eC . Instead of C we write C I + I and we use the
property of exponents (ea+b = ea eb ).
eC = eCI+I = eCI eI = e(CI) eI .
In the last part I is diagonal so we can easily calculate eI as described in part
(a). The matrix C I is nilpotent (a consequence of the Cayley-Hamilton
Theorem), so the infinite Taylor series will truncate to a finite sum.
For the matrix C given above, show that (C I)2 6= 0 but (C I)3 = 0. So
the Taylor series of eCI truncates to the first three terms:
eCI = I + C I +
1
(C I)2 .
2!
Problem A.2.3. Calculate eC (by calculating eI and eCI and then using
that eC = eCI eI ).
Solution. The eigenvalue of C is = 2, and a basis for its eigenspace is
1
0 .
0
So C is not diagonalizable because there is only one basis vector for the eigenspace(s).
eC = e(C+2I) e2I
1
2
= I + C + 2I + (C + 2I) e2 I
2!
1 0 0
0 1 0
0 2 1
1
1 + 0 0
0 e2
= 0 1 0 + 0 2
2
0 0 1
0 4 2
0 0
0
1
1 2 2
1 .
= e2 0 3
0 4 1
Z. G
onye
Appendix B
The Laplace Expansion For Determinant
B.1 First Minors and Cofactors; Row and Column Expansions
To each element aij in the determinant |A| = |aij |n , there is associated a subdeterminant of order (n 1) which is obtained from |A| by deleting row i and column j. This
subdeterminant is known as a first minor of |A| and is denoted by Mij = D (A (i|j)).
The first cofactor (CA)ij is then defined as a signed first minor:
(CA)ij = (1)i+j Mij = (1)i+j D (A (i|j)) .
(B.1.1)
It is customary to omit the adjective first and to refer simply to minors and cofactors
and it is convenient to regard Mij and (CA)ij as quantities which belong to aij in
order to give meaning to the phrase an element and its cofactor.
The expansion of |A| by elements from row i and their cofactors is
|A| = D(A) =
n
X
aij (CA)ij =
j=1
n
X
1 i n.
(B.1.2)
j=1
n
X
i=1
aij (CA)ij =
n
X
1 j n.
(B.1.3)
i=1
Since (CA)ij belongs to but is independent of aij , another way to define (CA)ij is
(CA)ij =
|A|
.
aij
(B.1.4)
90
aij (CA)kj = 0,
1 i n,
1 k n,
k 6= i.
(B.2.1)
j=1
The elements come from row i of |A|, but the cofactors belong to the elements in row
k and are said to be alien to the elements. The identity is merely an expansion by
elements from row k of the determinant in which row k = row i and which therefore
zero.
The identity can be combined with the expansion formula for A with the aid of
the Kronecker delta function ik to form a single identity which may be called the
sum formula for elements and cofactors:
n
X
1 i n, 1 k n.
(B.2.2)
j=1
It follows that
0
..
.
0
n
X
,
(CA)ij Cj =
D(A)
j=1
0
.
..
1 i n,
(B.2.3)
where Cj is column j of the matrix A, and the element D(A) is in the row i of the
column vector and all the other elements are zero. If D(A) = 0, then
n
X
(CA)ij Cj = 0,
1 i n,
(B.2.4)
j=1
that is, the columns are linearly dependent. Conversely, if the columns are linearly
dependent, then D(A) = 0.
T. Pranayanuntana
91
1 i n,
aij xj = bi ,
(B.3.1)
j=1
Cj xj = B,
(B.3.2)
j=1
where
B=
b1
b2
b3
..
.
bn
So
Pn
C1 Cj1 B Cj+1 Cn = C1 Cj1
j x
j Cj+1 Cn
j=1 C
= C1 Cj1 x1 C1 Cj+1 Cn
+
+ C1 Cj1 xj1 Cj1 Cj+1
+ C1 Cj1 xj Cj Cj+1 Cn
+ C1 Cj1 xj+1 Cj+1 Cj+1
Cn
Cn
+
+ C1 Cj1 xn Cn Cj+1 Cn
= xj C1 Cj1 Cj Cj+1 Cn
= xj |A|.
(B.3.3)
T. Pranayanuntana
92
If |A| = |aij |n 6= 0, then the unique solution of the equations can also be expressed in
column vector notation.
1
xj =
C1 Cj1 B Cj+1 Cn
|A|
n
(B.3.4)
1 X
=
bi (CA)ij .
|A| i=1
T. Pranayanuntana
Appendix C
94
14. A(B C) = AB AC
15. (B C)A = BA CA
16. a(BC) = (aB)C = B(aC)
17. A0 = 0A = 0
18. AI = IA = A
19. Ar As = Ar+s
20. (Ar )s = Ars
21. A0 = I
22. (A + B)T = AT + B T
23. (A B)T = AT B T
24.* (AB)T = B T AT
25. (AT )T = A
26. (aB)T = aB T
27.* tr(A + B) = tr(A) + tr(B)
28.* tr(A B) = tr(A) tr(B)
29.* tr(aB) = a tr(B)
30.* tr(AB) = tr(BA)
31. (A1 )1 = A
32. (An )1 = (A1 )n
1
33. (aA)1 = A1 , for any nonzero scalar a.
a
34.* (AB)1 = B 1 A1 .
35. (AT )1 = (A1 )T .
36.* det(I) = 1.
37.* det(AB) = det(A) det(B).
Z. G
onye
95
1
.
det(A)
Z. G
onye
adj(A)
.
det(A)
96
Z. G
onye
Bibliography
[1] A. C. Aitken, Determinant and matrices, Oliver and Boyd, London, 1939.
[2] Howard Anton and Chris Rorres, Elementary linear algebra, Wiley & Sons, Inc.,
New York, 2000.
[3] Kenneth Hoffman and Ray Kunze, Linear algebra, Prentice-Hall, Inc., New Jersey, 1971.
[4] Roger A. Horn and Charles R. Johnson, Matrix analysis, Cambridge University
Press, New York, 1985.
[5] Bernard Kolman and David Hill, Elementary linear algebra, Prentice-Hall, Inc.,
New Jersey, 2000.
[6] Peter Lancaster and Miron Tismenetsky, The theory of matices, Academic Press
Inc., San Diego, CA, 1985.
[7] David C. Lay, Linear algebra and its applications, Addison Wesley, Boston, 2003.
[8] V. V. Prasolov, Problems and theorems in linear algebra, vol. 134, American
Mathmatical Society, United States, 1994.
[9] Joel W. Robbin, Matrix algebra, A K Peters, Ltd., MA, 1985.
[10] Max F. Stein, Introdution to matrices and determinants, Wadsworth Publishing
Company Inc., Belmont, Ca., 1967.
[11] Gilbert Strang, Linear algebra and its applications, Brooks/Cole, Thomson
Learning Inc., 1988.
[12] Robert Vein and Paul Dale, Determinants and their applications in mathematical
physics, Springer-Verlag New York Inc., New York, 1999.
Index
(A)ij , 17
Zp , 13
P, 53
Pn , 53
rref, 6
aij , 17
pth power, 21
adjoint matrix, 43
algebraic multiplicity, 69
augmented matrix, 4
back substitution, 3
basis, 59
diagonalizable, 73
difference, 18
differential operator, 80
dilatation, 52
eigenspace, 69
eigenvalue, 68
eigenvector, 68
elementary matrix, 25
Elementary row operations:, 4
entries, 17
equal, 18
free variable, 11
characteristic equation, 68
characteristic polynomial, 68
coefficient matrix, 4
coefficients, 1
cofactor, 43
cofactor expansion, 42
cofactor matrix, 43
column matrix, 17
column space, 62
congruent modulo p, 13
consistent, 1
constant term, 1
Gauss-Jordan elimination, 6
Gaussian algorithm, 6
Gaussian elimination, 6
General Linear Transformations, 79
geometric multiplicity, 70
determinant, 34
determinant function, 37
diagonal matrix, 30
leading 1, 5
linear equation, 1
linear system, 1
identity matrix, 18
identity transformation, 51
inconsistent, 1
inverse, 26
invertible, 26, 51
kernel, 83
99
INDEX
linear transformation, 49, 79
linearly dependent, 57
linearly independent, 57
lower triangular, 31
main diagonal, 17
map, 49, 79
matrix, 17
matrix product, 19
nilpotent, 32
non-invertible, 26
non-trivial solution, 41
null space, 63
nullity, 63, 83
operator, 49, 79
orthogonal, 32
product, 18
projection, 51
range, 83
rank, 64
real vector space, 52
reduced row echelon form, 6
residue classes modulo p, 13
rotation, 52
row echelon form, 5
row matrix, 17
row space, 63
Scale, 25
Shear, 25
singular, 26
size, 4
size m n, 17
skew-symmetric, 31
solution, 1
space spanned by, 56
square matrix of order n, 17
standard basis, 59
standard matrix, 50
subspace, 53
sum, 18
Swap, 25
symmetric, 31
system of linear equations, 1
trace, 22
transformation, 49, 79
transpose of A, 22
triangular, 31
trivial solution, 41
unknowns, 1
upper triangular, 30
vector, 17, 47
Vector arithmetic, 48
zero matrix, 17
zero transformation, 51