You are on page 1of 10

APPENDIX A

MATRIX INTRODUCTION

This appendix presents a matrix introduction to support the material of


the textbook. Topics covered are matrix basics, matrix arithmetic, matrix
determinants, and matrix inversion.

A.1 BASICS

A matrix is a two-dimensional array of real or complex numbers. Here a


few examples:
 
−1 + j 2 3
    
2 0 −1  0 −j 
 −3 1
  [0 2 4]
1 5 3  2−j 1  2 −4
1 + j3 0
 
−1
 0
 −2 
7

The number of rows and columns specify the dimension of the matrix.
That is, a matrix with m rows and n columns is said to have dimension
m × n or is referred to as an m × n matrix. The preceding matrices have
dimension 2 × 3, 4 × 2, 2 × 2, 1 × 3, and 4 × 1, respectively. Note that a
row vector can be regarded as a matrix with a single row, and a column
vector can be regarded a matrix with a single column. Matrices with the
same number of rows and columns are called square.

Linear State-Space Control Systems. Robert L. Williams II and Douglas A. Lawrence


Copyright  2007 John Wiley & Sons, Inc. ISBN: 978-0-471-73555-7
407
408 MATRIX INTRODUCTION

An m × n matrix will be represented generically as


 
a11 a12 · · · a1n
 a21 a22 · · · a2n 
A=  ... .. .. 
. 
..
. .
am1 am2 · · · amn
or, using the shorthand notation, A = [aij ], where aij denotes the element
in the ith row and j th column. Hence the first subscript indexes the row
in which the element lies, and the second subscript indexes the column
in which the element lies.
The n × n identity matrix, denoted by In (or simply I when the dimen-
sion is clear from the context), is the square matrix with ones along the
main diagonal and zeros elsewhere. For example,
 
1 0 0
I3 =  0 1 0 
0 0 1

The m × n matrix with a zero in every element is called a zero matrix


and will be denoted either by 0m×n or just 0, again when the dimension
is clear from the context.
The transpose of an m × n matrix A = [aij ] is the n × m matrix given
by AT = [aj i ]. That is, the transpose is obtained by interchanging rows
and columns. The conjugate transpose of an m × n matrix A = [aij ] is
the n × m matrix given by A∗ = [a j i ]. Hence the conjugate transpose
is obtained by combining matrix transposition with element-wise con-
jugation. The conjugate transpose is also referred to as the Hermitian
transpose.

Example A.1 For  


−1 + j 2 3
 −j 
 0 
A= 
 2−j 1
1 + j3 0
we obtain
 
−1 + j 2 0 2 − j 1 + j 3
A =
T
and
3 −j 1 0
 
∗ −1 − j 2 0 2 + j 1 − j 3
A =
3 j 1 0
MATRIX ARITHMETIC 409

If A is a real matrix, meaning that all elements are real numbers, then
A∗ = AT . 

A.2 MATRIX ARITHMETIC

Matrix addition and subtraction operations are performed on an element-


by-element basis and therefore are defined only for matrices of the
same dimension. For m × n matrices A = [aij ] and B = [bij ], their sum
C = A + B is specified by

[cij ] = [aij + bij ]

and their difference D = A − B is specified by

[dij ] = [aij − bij ]

Multiplication of a matrix A of any dimension by a real or complex scalar


α is specified by
αA = [αaij ]

that is, every element of A is multiplied by α to produce αA. Note that


matrix subtraction can be represented as A − B = A + (−1)B.

Example A.2 For


   
−1 + j 2 3 0 1+j
 −j  1 
 0   2 
A=  and B= 
 2−j 1 j −1 
1 + j3 0 5 −1 − j 4

we obtain
 
−1 + j 2 4+j
 2−j 
 1 
A+B =  and
 2 0 
6 + j 3 −1 − j 4
 
−1 + j 2 2 − j
 −1 −2 − j 
 
A−B = 
 2 − j2 2 
−4 + j 3 1 + j 4
410 MATRIX INTRODUCTION

Also,
   
−2 + j 4 6 0 1−j
 −j 2   −j −j 2 
 0   
2A =   and − jB =  
 4 − j2 2   1 j 
2 + j6 0 −j 5 −4 + j


Based on the definitions, it is clear that matrix addition and scalar mul-
tiplication have general properties very similar to their counterparts for
real and complex numbers: For any m × n matrices A, B, and C and any
pair of scalars α and β,

1. Commutativity
A+B =B +A
2. Associativity
a. (A + B) + C = A + (B + C)
b. (αβ)A = α(βA)
3. Distributivity
a. α(A + B) = αA + αB
b. (α + β)A = αA + βA
4. Identity elements
a. The zero matrix 0 satisfies A + 0 = A
b. 0A = 0m×n and 1A = A

The matrix product C = AB of an m × n matrix A and a p × q matrix


B can be defined only when n = p; i.e., the number of columns of the
left factor A matches the number of rows of the right factor B. When this
holds, the matrices A and B are said to be conformable, and the matrix
product is the m × q matrix given by


n
C = [cij ] cij = aik bkj
k=1

It is often helpful to visualize the computation of cij as the inner or dot


product of the ith row of A and the j th column of B, which are vectors
of equal length because of conformability.
MATRIX ARITHMETIC 411

Example A.3 Given


 
−1 + j 2 3
  
 0 −j 
 −3 1
A=  and B=
 2−j 1 2 −4
1 + j3 0

we obtain
 
(−1 + j 2)(−3) + (3)(2) (−1 + j 2)(1) + (3)(−4)
 0(−3) + (−j )(2) 0(1) + (−j )(−4) 
 
AB =  
 (2 − j )(−3) + (1)(2) (2 − j )(1) + (1)(−4) 
(1 + j 3)(−3) + (0)(2) (1 + j 3)(1) + (0)(−4)
 
9 − j 6 −13 + j 2
 −j 2 j4 
 
= 
 −4 + j 3 −2 − j 
−3 − j 9 1 + j3

In this example, it is not possible to define the product BA because the


number of columns of B does not match the number of rows of A. 
It is important to note that matrix multiplication is not commutative.
Even when both products AB and BA can be formed and have the same
dimension (which happens when A and B are square matrices of the same
size), in general, AB  = BA. Consequently, the order of the factors is very
important. Matrix multiplication does satisfy other properties reminiscent
of the scalar case. For matrices A, B, and C with dimensions such that
every matrix product below is defined:

1. Associativity
(AB)C = A(BC)
2. Distributivity
a. A(B + C) = AB + AC
b. (B + C)A = BA + CA

Finally, matrix multiplication and matrix transposition are related as


follows:
(AB)T = B T AT
412 MATRIX INTRODUCTION

and this can be extended to any number of factors. In words, the transpose
of a product is the product of the transposes with the factors arranged in
reverse order. An analogous result holds for conjugate transposition.

A.3 DETERMINANTS

Determinants are only defined for square matrices. The determinant of


an n × n matrix A is denoted by |A|. For dimensions n = 1, 2, 3, the
determinant is computed according to

n = 1 : |A| = a11
n = 2 : |A| = a11 a22 − a12 a21
n = 3 : |A| = a11 a22 a33 + a12 a23 a31 + a13 a32 a21 − a31 a22 a13
− a32 a23 a11 − a33 a12 a21

For n > 3, the determinant is computed via the Laplace expansion spec-
ified as follows: First, the cofactor Cij of the matrix element aij is
defined as
Cij = (−1)i+j Mij

in which Mij is the minor of the matrix element aij , which, in turn,
is defined as the determinant of the (n − 1) × (n − 1) submatrix of A
obtained by deleting the ith row and j th column. In terms of these defi-
nitions, the determinant is given by the formula(s):

n
|A| = aij Cij for fixed 1 ≤ i ≤ n
j =1


n
|A| = aij Cij for fixed 1 ≤ j ≤ n
i=1

A few remarks are in order. First, there are multiple ways to com-
pute the determinant. In the first summation, the row index i is fixed,
and the sum ranges over the column index j . This is referred to as
expanding the determinant along the ith row, and there is freedom to
expand the determinant along any row in the matrix. Alternatively, the
second summation has the column index j fixed, and the sum ranges
over the row index i. This is referred to as expanding the determinant
along the jth column, and there is freedom to expand the determinant
along any column in the matrix. Second, the Laplace expansion specifies
DETERMINANTS 413

a recursive computation in that the determinant of an n × n matrix A


involves determinants of (n − 1) × (n − 1) matrices, namely, the minors
of A, which, in turn, involve determinants of (n − 2) × (n − 2) matri-
ces, and so on. The recursion terminates when one of the cases above
is reached (n = 1, 2, 3) for which a closed-form expression is available.
Third, observe that whenever aij = 0, computation of the cofactor Cij is
not necessary. It therefore makes sense to expand the determinant along
the row or column containing the greatest number of zero elements to
simplify the calculation as much as possible.

Example A.4 For the 4 × 4 matrix


 
1 0 0 −3
 0 5 2 1
A=
−1 0 0 −1 
3 1 0 0
there are several choices for a row or column with two zero elements
but only one choice, namely, the third column, for which there are three
zero elements. Expanding the determinant about the third column (j = 3)
gives
4
|A| = ai3 Ci3
i=1

= (0)C13 + (2)C23 + (0)C33 + (0)C43


= (2)C23

so the only cofactor required is C23 = (−1)2+3 M23 . The associated minor
is the determinant of the 3 × 3 submatrix of A obtained by deleting row
2 and column 3:


1 0 −3

M23 =

−1 0 −1


3 1 0

= (1)(0)(0) + (0)(−1)(3) + (−3)(1)(−1) − (3)(0)(−3)


− (1)(−1)(1) − (0)(0)(−1)
=4

so C23 = (−1)2+3 (4) = −4. Finally,


|A| = 2C23
= −8 
414 MATRIX INTRODUCTION

Several useful properties of determinants are collected below:

1. |A| = |AT |
2. |AB| = |A||B| = |BA|
3. |αA| = α n |A|
4. If any row or column of A is multiplied by a scalar α to form B,
then |B| = α|A|.
5. If any two rows or columns of A are interchanged to form B, then
|B| = −|A|.
6. If a scalar multiple of any row (respectively, column) of A is
added to another row (respectively, column) of A to form B, then
|B| = |A|.

A.4 MATRIX INVERSION

An n × n matrix A is called invertible or nonsingular if there is another


n × n matrix B that satisfies the relationship

AB = BA = In

In such cases, B is called the inverse of A and is instead written as A−1 .


If A has an inverse, the inverse is unique. If A has no inverse, then A is
called singular. The following basic fact provides a test for invertibility
(nonsingularity).

Proposition A.1 A is invertible (nonsingular) if and only if |A|  = 0. 

The inverse of an invertible matrix is specified by the formula


adj(A)
A−1 =
|A|
in which adj(A) is the adjugate or adjoint of A and is given by

adj(A) = [Cij ]T

where Cij is the cofactor of the (i, j )th element of A. That is, the adjugate
of A is the transpose of the matrix of cofactors. The fraction appearing in
the preceding formula for the inverse should be interpreted as multiplica-
tion of the matrix adj(A) by the scalar 1/|A|.

Example A.5 The 4 × 4 matrix A from Example A.4 is invertible


(nonsingular) because |A| = −8  = 0. Construction of the 4 × 4 adjugate
MATRIX INVERSION 415

matrix requires the calculation of a total of 16 3 × 3 determinants. From


a previous calculation, C23 = −4, and as another sample calculation


5 2 1

M11 =
0 0 −1

1 0 0
= (5)(0)(0) + (2)(−1)(1) + (1)(0)(0) − (1)(0)(1)
− (0)(−1)(5) − (0)(2)(0)
= −2

from which C23 = (−1)1+1 M11 = −2. After 14 more such calculations,
 
−2 6 −16 2 T
 0 0 −4 0 
adj(A) = 
6 −18 44 2 
0 −8 20 0
 
−2 0 6 0
 6 0 −18 −8 
=
−16 −4 44 20 
2 0 2 0

which leads to
 
−2 0 6 0
1  6 0 −18 −8 
A−1 =  −16 −4
−8 44 20 
2 0 2 0
 
1 −3
 4 0 0 
 4 
 −3 
 9
1 
 0 
 4 4 
= 
 1 −11 −5 
 2 
 2 2 2 
 
 −1 −1 
0 0
4 4

The correctness of an inverse calculation can always be verified by check-


ing that AA−1 = A−1 A = I . This is left as an exercise for the reader.

416 MATRIX INTRODUCTION

The transpose of a matrix is invertible if and only if the matrix itself


is invertible. This follows from Proposition A.1 and the first determinant
property listed in Section A.3. When A is invertible,

(AT )−1 = (A−1 )T

In general, an asserted expression for the inverse of a matrix can be veri-


fied by checking that the product of the matrix and the asserted inverse (in
either order) yields the identity matrix. Here, this check goes as follows:

AT (AT )−1 = AT (A−1 )T


= (A−1 A)T
= (I )T
=I

This result shows that matrix transposition and matrix inversion can be
interchanged, and this permits the unambiguous notation A−T .
Also, the product of two n × n matrices is invertible if and only if
each factor is invertible. This follows from Proposition A.1 and the second
determinant property listed in Section A.3. When the product is invertible,
the inverse is easily verified to be

(AB)−1 = B −1 A−1

That is, the inverse of a product is the product of the inverses with the
factors arranged in reverse order.

You might also like