CHAPTER

Matrix Algebra

In Chapter 1 we discussed the basic steps involved in any finite element analysis. These
steps include discretizing the problem into elements and nodes, assuming a function
that represents behavior of an element, developing a set of equations for an element, assembling the elemental formulations to present the entire problem, and applying the
boundary conditions and loading. These steps lead to a set of linear (nonlinear for some
problems) algebraic equations that must be solved simultaneously. A good understanding
of matrLx algebra is essential in formulation and solution of finite element models. As
is the case with any topic, matrix algebra has its own terminology and follows a set of
rules. We provide an overview of matrix terminology and matrix algebra in this chapter.
The main topics discussed in Chapter 2 include
2.1
2.2
2.3

Basic Definitions
MatrLx Addition or Subtraction
MatrL-1: Multiplication

2.4
2.5
2.6

Partitioning of a Ma trLx
Transpose of a Matrix
Determinant of a Matrix

2.7
2.8
2.9

Solutions of Simultaneous Linear Equations

Inverse of a Matrix
Eigenvalues and Eigenvectors

2.10 Using MATLAB to Manipulate Matrices

2.11 Using Excel to Manipulate Matrices

2.1

BASIC DEFINITIONS

A matrix is an alTay of numbers or mathematical terms. The numbers or the mathematical terms that make up the matrix are called the elements of matrix. The size
of a matrix is defined by its number of rows and columns. A matrix may consist

Section 2.1

Basic Definitions

67

of m rows and n columns. For example,

~[ ~

[i\1

5
26
8

-5

{L}

~4]

[T]

Df(x , y, z)
dX
af(x , y, z)
ay
Df(x ,y,z)

'r

COS

-sin e
cos e
0
0

0
0
cos e
sin e

jXdX

jYdY

jx2

az

cos e

[I] =

-,L]

dx

j i dy

Matrix [N] is a 3 by 3 (or 3 X 3) matrix whose elements are numbers, [1] is a 4 X 4 that
has sine and cosine terms as its elements, {L} is a 3 X 1 matrix with its elements representing partial derivatives, and [/] is a 2 X 2 matrix with integrals for its elements. The
[N] , [T], and [I] are square matrices. A square matrix has the same number of rows
and columns. The element of a matrix is denoted by its location. For example, the element in the first row and the third column of a matrix [8] is denoted by bu , and an element occulTing in matl-lX [A] in row 2 and column 3 is denoted by the term a 23. In this
book, we denote the matrix by a bold-facc Ictter in brackets [] and {} , for example: [K] ,
[T], {F} , and the elements of matrices are represented by regular lowercase letters. The
{} is used to distinguish a column matrix.

Column Matrix and Row Matrix

A column matrix is defined as a matrix that has one column but could have many rows.
On the other hand, a row matrix is a matrix that has one row but could have many
columns. Examples of column and row matrices follow.

af(x , y, z)

{A}

~ ~2}' ~ {;:}..nd ~
{

{X}

{L}

ax
at(x, y, z)
by

are examples of column matrices.

at(x, y, z)
ilz

whereas [C]

= [5

-3] and [Y]

= [YI Y2 Y3] are examples of row matrices.

Diagonal, Unit, and Band (Banded) Matrix

A diagonal matrix is one that has elements only along its principal diagonal ;
the elements are zero everywhere else (aij = 0 when i * j). An example of a

68

Chapter 2

Matrix Algebra

4 X 4 diagonal matrix follows.

0
0

a2

o
o

a3

The diagonal along which at> a2, a3, and a4lies is called the principal diagonal. An identity
or unilmatrix is a diagonal matrix whose elements consist of a value 1. An example of
an identity matrix follows.

10000
01000
00100

[I]

=
00010
00001

A banded matrix is a matrix that has a band of nonzero elements parallel to its principal diagonal. As shown in the example that follows, all other elements outside the band
are zero.

[8]

b l2
bn

b ll
b21
0
0
0
0
0

b32

0
0
0
0

0
b23
b 33
b 43

0
0
0

0
0

0
0
0

0
0
0
0

b 34
b44
b 54

b 45
b 55

b 56

b 65

b66

0
0
0
0
0
b67

b 76

~7

Upper and Lower Triangular Matrix

An upper triangular matrix is one that has zero elements below the principal diagonal
(uii = 0 when i > j) , and the lower triangular matrix is one that has zero elements above
the principal diagonal (Iii = 0 when i < j). Examples of upper triangular and lower
triangular matrices are shown below.
llll

[U]

=[

o
o

lll2

lil3

lln

Un

0
0

U33

1124

122

1134

132

133

1144

142

143

""J

I~J

Section 2.3

2.2

Matrix Multiplication

69

MATRIX ADDITION OR SUBTRACTION

Two matrices can be added together or subtracted from each other provided that
they are of the same size-each matrix must have the same number of rows and
columns. We can add matrix [A ]mxn of dimension m by 11 to matrix [B]mxn of the same
dimension by adding the like elements. Matrix subtraction follows a similar rule , as
shown.

[A] [B]

au

al2

a 21

an

al n
a2n

bu
b21

bl 2
bn

bin
b2n

b ml

bm2

bmn

ami

amn

am2

(au b u )
(a21 b21)

(a 12 b 12 )
(an b22 )

(al n bin)
(azn b2n)

(ami b ml )

(a m2 b m2)

(a mn b mn )

The rule for matrix addition or subtraction can be generalized in the following
manner. Let us denote the elements of matrix [A] by ai; and the elements of matrix [B]
by bi;, where the number of rows i varies from 1 to m and the number ofcolumnsjvaries
from 1 to Il . If we were to add matrix [A] to matrix [B] and denote the resulting matrix
by [C] , it follows that

[A] + [B] = [C]

and
Cij

2.3

= ai; + bijfori = 1, 2, ... , mandj = 1, 2, ... , 11

(2.1)

MATRIX MULTIPLICATION

In this section we discuss the rules for multiplying a matrix by a scalar quantity and by
another matrix.

Multiplying a Matrix by a Scalar Quantity

When a matrix [A] of size m X 11 is multiplied by a scalar quantity such as 13, the
operation results in a matrix of the same size m X 11 , whose elements are the product
of elements in the original matrix and the scalar quantity. For example, when we multiply matrix [A] of size m X n by a scalar quantity [3 , this operation results in another

70

Chapter 2

Matrix Algebra

matrix of size m X n, whose elements are computed by multiplying each element of

matrix [A] by 13 , as shown be low.

13 [A ]

011

012

In

13 0 11

13 0 12

130ln

021

022

2n

13 0 21

13 0 22

1302n

= 13

=
0 111 1

0 1112

(2.2)
130111 1

llln

13 0 llln

130m2

Multiplying a Matrix by Another Matrix

Whereas any size matrix can be multiplied by a scalar quantity, matrix multiplication can
be performed only when the number of columns in the premultiplier matrix is equal to
the number of rows in the pastmultiplier matrix. For example, matrix [A] of size m X n
can be premultiplied by matrix [D] of size n X p because the number of columns n in
matrix [A] is equal to number of rows n in matrix [D]. Moreover, the multiplication
results in another matrix, say [C], of size m X p. Matrix multiplication is carried out
according to the following rule .
[A ]III X,,[D]n x p

= [C]III Xp

L-J

must match

[AHD]

011

012

Oln

b ll

b l2

b lp

CI1

CI2

Clp

0 21

022

2n

b 21

b 22

b 2p

C21

C22

C2p

CIIII

Cm2

cmp

=
0ml

0m2

0nlll

b nl

b n2

bnp

where the elements in the first column of the [C] matrix are computed from
cZ I

= ol1 b l1 + 01 2b 21 + ... + Olnbnl

= 0 21b ll + 0 22 b 21 + ... + 0 2n b nl

Cml

= Omlbll + Om2b 21 + ... + Omnbnl

ClI

and the e le me nts in the second column of the [C] matrix are

C22

= 0l1 b l2 + 012b 22 + ... + 0lnb n2

= 021 b l2 + 0 22b 22 + ... + 0 2"b,,2

Cm2

= 01lllb l 2 + 01ll2b 22 + ... + 01ll"b,,2

C12

and similarly, the elements in the other columns are computed, leading to the last column
of the [C] matrix
Clp
C2p

= Ol1 b lp + 012b 2p + ... + Olnbnp

= 021 b lp + 0 22b 2p + ... + 02nb np

Section 2.3

Matrix Multiplication

71

The multiplication procedure that leads to the values of the elements in the [q matrix
may be represented in a compact summation form by
n

cmp = '2:,a mk bkP

(2.3)

k- I

When multiplying matrices, keep in mind the following rules. Matrix multiplication is not commutative except for very special cases.

[A][B]

'* [BHA]

(2.4)

= ([A][B])[q

(2.5)

Matrix multiplication is associative; that is

[A]( [B][C])

The distributive law holds true for matrix multiplication; that is

([A] + [B])[q = [A][q + [B][q

(2.6)

= [A][B] + [A][q

(2.7)

or

[A]([B] + [CD

For a square matrix , the matrix may be raised to an integer power n in the following
manner:
ntimcs

[AY

= [AHA] .. . [A]

(2.8)

This may be a good place to point out that if [I] is an identity matrix and [A] is a square
matrix of matching size, then it can be readily shown that the product of
[JHA] = [A][J] = [A]. See Example 2.1 for the proof.

EXAMPLE 2.1
Given matrices

[A]

[~ ~2

n G~ ~:l
[B]

perform the following operations:

ll.

[A] + [B]

b. [A] - [B]

=?
=?

c.3[A]=?

d.
c.
f.
g.

[AHB] = ?
[A]{C} = ?
[A]2 = ?
Show that [J][A] = [AHJ]

[A]

and {e}

n}

72

Chapter 2

Matrix Algebra

'Ve will use the operation rules discussed in the preceding sections to answer these
questions.
a. [A] + [B]

[AJ

=?

+ ~ G !]+n ~2]
[(0 +4)
(0 +(-2))] = [4
=
5
3
-2

[BJ

6
2
3

(8 + 7)
(9 + 1)

b. [A] - [B]

[AJ - [BJ

~G

-2]

11

15
10

5
1

15

0]7 - [47 26 ~2]

5
3
-2

(5-6)
(3-2)
(-2-3)

(9-1)

3[AJ

(7 + 3)
(9 + (-4))

=?

[(0-4)
= (8-7)
c. 3[A]

(5 + 6)
(3 + 2)
(-2 + 3)

-4

-4

(0-(
-2))] = [-41
(7-3)
(9-(-4))

-1
1
-5

1~]

=?

~ {~ ~2 ~] =

d. [A][B]

[(3)(8)
(3)(9)

g~g~

(3)(-2)

(3)(7)] =
(3)(9)

[2~
27

=?

[A][B] =

[~ ~ ~][; ~ ~2] =
9

-2

-4

(0)(4) + (5)(7) + (0)(1) (0)(6) + (5)(2) + (0)(3) (0)(-2) + (5)(3) + (0)(-4) ]

[ (8)(4) + (3)(7) + (7)(1) (8)(6) + (3)(2) + (7)(3) (8)(-2) + (3)(3) + (7)(-4)
(9)(4) + (-2)(7) + (9)(1) (9)(6) + (-2)(2) + (9)(3) (9)(-2) + (-2)(3) + (9)( -4)

[~~ ~~ ~~5]
31

c. [A]{C}

[A]{C}

=?

77

-60

= { ~~~~=~~ g~g~ : ~~~g~ } = {~~}

~ G -2~ 9~]{~1}
5
(9)(-1)+(-2)(2)+(9)(5)
32

Section 2.4

f. [A]2

[AJ'

73

=?

~ [A][A] ~ [~

g. Show that

[A][I]

m~

5
3
-2

0] = [40

5
3
-2

87

65

15
35
21

35]

5
3

nand

84
67

= [AHI] = [A]

[IHA]

[/HA]

2.4

Partitioning of a Matrix

~ [~

0
1
0

~ [~

5
3
-2

m~

n ~[:
n~[:

5
3
-2
0
1
0

nu

-2
5
3
-2

PARTITIONING OF A MATRIX

Finite element formulation of complex problems typically involves relatively large sized
matrices. For these situations, when performing numerical analysis dealing with matrix
operations, it may be advantageous to partition the matrix and deal with a subset of
elements. The partitioned matrices require less computer memory to perform the operations. Traditionally, dashed horizontal and vertical lines are used to show how a matrix is partitioned. For example , we may partition matrix [A] into four smaller matrices
in the following manner:

[A]

al2

a\3

al4

a l5

a l6

a21

a22

a23

! a24

a 25

a26

a31

a32

a33

I a34

a 35

a36

a41

a42

a43

a45

a46

a51

a52

a53

! a44
! a54

a5S

a56

A I2

--------------- ---- ---- ---~----------- ------ ---- ----

and in terms of submatrices

all

[A]

= [ All
A21

A22

where
[All]

[A2tl

=
=

[an

a12

a21

a22

[a"

a32

a41

a42

a51

a52

a 13
a23

a,,]
a43
a53

[A I2]

[Azzl

a l6
a26

a45

a55

a56

al4
a24

al5

[a"

a35

a44
a54

=[

a25

36

a46

It is important to note that matrix [A] could have been partitioned in a number of
other ways, and the way a matrix is partitioned would define the size of sub matrices.

74

Chapter 2

Matrix Algebra

Addition and Subtraction Operations

Using Partitioned Matrices
Now let us turn our attention to matrix operations dealing with addition , subtraction ,
or multiplication of two matrices that are partitioned. Consider matrix [B] having the
same size (5 X 6) as matrix [A]. If we partition matrix [B] in exactly the same way we
partitioned [A] previously, then we can add the submatrices in the following manner:

[B]

b l2

b l1
b21

b13
b 23

I b l4 b ls b l6

Ib b b
b;-;-----b;;-----b;-;T-b;~-----b~;-----b;~
b
b
b Ib
b
b
bn

24

2S

26

41

42

43

44

4S

46

bSI

bS2

bS3

I bS4

bss

bS6

b 13 ]
b23

[B I2 ]

= [b 14

b24

b 2S

b26

bbn ]
43

[b~
[Bn] = b44

bb

bSJ

bS4

b3S
b 4S
bss

where

[Bid

[B2d

= [b ll

b l2

b21

b 22

[b"b

b32
b 42
bS2

41

bSI

b ls

b 16]

36
46 ]

b S6

Then, using submatrices we can write

[A] + [8]

= [All + Bl1
A21

+ 8 21

Matrix Multiplication Using Partitioned Matrices

As mentioned earlier, matrix multiplication can be performed only when the number of
columns in the prel11l1/tip/ier matrix is equal to the number of rows in the postrl1l1lliplier
matrix. RefelTing to [A] and [8] matrices of the precedi.ng section, because the number of
columns in matrix [A] does not match the number of rows of matrix [8], then mntrjx [B]
cannot be premultiplied by matrix [AJ. To demonstrate matrix multiplication using submatrices, consider matrix [q of size 6 X 3, which is partitioned in the manner shown below.

[e]

Cll

CI2

CI3

C21

C22: C23

C41

C42

=~!------~~~-L~~-~
I
C51

c52

C61

C62

[e] = [ell

C43

I
i

el2 ]

e21 e22

C53
C63

Section 2.4

Partitioning of a Matrix

75

where

Next , consider premultiplying matrix [C] by matrix [A]. Let us refer to the results
of this multiplication by matrix [D] of size 5 X 3. In addition to paying attention to the
size requirement for matrix multiplication , to carry out the multiplication using partitioned matrices, the premultiplying and postmultiplying matrices must be partitioned in
such a way that the resulting submatrices conform to the multiplication rule. That is, if
we partition matrix [A] be tween the third and the fourth columns, then matrix [C] must
be partitioned betwee n the third and the fourth rows. However, the column partitioning of matrix [C] may be done arbitrarily, because rega rdless of how th e columns are partitioned, the resulting submatrices will still conform to the multiplication rule. In other
words, instead of partitioning matrix [C] betwee n columns two and three , we could
have partitioned the matrix between columns one and two and still carried out the multiplication using the resulting submatrices.

[A][C]

= [D] = [Att

A2l

Al2][C:ll
A 22
C2l

C l2 ]
C22

= [AttC:tt + A l2 C:2l
A21 Ctt +

A 22 C21

D1 2]
D22

where

EXAMPLE 2.2

Given the following partitioned matrices, calculate the product of [A ][E]

the submatrices shown.

[A]

7
2 I 0
3
5
5
3
8
-3 i -5 0
8
---------------------------j----------------------------and
1
4
0 i 7
15
9
-1
0 10
12
3
5
2 -5
9 I 2
18 -10

= [C]

2 10 I 0
8
7 I 5
! -4
-5
2
I
[E] = ----------------,-------4
8 I 13
3 12 I 0
1
5 I 7
I

using

76

Chapter 2

Matrix Algebra

where

[~5 ~

[Ad

[An]

= [1272

1; ~

1]
-10

18

and

where

where

[ClI ]

= [AlI][BlIl + [A 12][B2tl = [53 87

+
[Cd

[~5

= [AlI]{Bd + [A 12]{B22 } = [53 87

~3 ~5
J[

o3

5JU
8 1

= [A 21 HB lI ] +

[A22HB2tl

5
9

8
-5

7
2

9
-1
-10

3
1

10
-5

+ 12
2

15
3
18

164J
80

mn ~ {:;}

[0

12
10]
r[7 r
~

1~]
~ [7~7.;)
5

~3JUJ

+ -5 0

[C2tl

r]

8]
12
5

319]

= [116
111 207
-29

185

Section 2.5

Transpose of a Matrix

15
3
18

77

nUJ
9 ]{ 13}
-1
0
-10
7

= {174}
179
-105

The final result is given by

70

164 II 62

73
80 i 43
---------- -----------~--- --------

[A][B] = [C] = [Att

116
111
-29

A21

319 I 174
207 I 179
I
185 i -105

As explained earlier, the column partitioning of matrix [B] may be done arbitrarily because the resulting submatrices still conform to the multiplication rule. It is left as an exercise for you (see problem 3 at the end of this chapter) to show that we could have
partitioned matrix [B] between columns one and two and used the resulting submatrices to compute[A ][B] = [C].

2.5

TRANSPOSE OF A MATRIX

As you will see in the following chapters, the finite element formulation lends itself to
situations wherein it is desirable to rearrange the rows of a matrix into the columns of
another matrix. To demonstrate this idea, let us go back and consider step 4 in Example 1.1. In step 4 we assembled the elemental stiffness matrices to obtain the global
stiffness matrix. You will recall that we constructed the stiffness matrix for each element with its position in the global stiffness matrix by inspection. Let's recall the stiffness matrix for element (1), which is shown here again for the sake of continuity and
convenience.

[K](1)

[K](1C)

=[

kl
-kl

-k l ] and its position in the global matrix

kl

kl
-k l
0
0
0

-kl
kl
0
0
0

0
0
0
0
0

0
0
0
0
0

0
0
0
0
0

Instead of putting together [K](IC) by inspection as we did, we could obtain [K](1C) using
the following procedure:

(2.9)

78

Chapter 2

Matrix Algebra

where

[Ad=[~

0
0

0
0

~J

and

[A1t =

0
0
0
0

0
0
0

[Ad\ called th e transpose of [Ad , is obtained by taking the first and the second rows of
[A d and making them into the first and the second columns of the transpose matrix. It
is easily verified that by carrying out th e multiplication given by Eq. (2.9), we will arrive
at th e same result that was obtained by inspection.

[K](IG)

u u u

o o OJ =
o o 0

= 0

o
o

o o
o o
o o

0
0
0

Similarly, we could have performed th e following operation to obtain [K](2G):

[K](2G)

= [A 2Y[K](2) [A 2]

where

[A 2]

[~

0
1

0
0

~J

and
0
1
0
0
0

[A 2Y =

0
0

1
0
0

and

[K](2G) =

0
1
0
0
0

0
0
1 [ k2
-k2
0
0

J[ 00

-k 2
k2

1
0

0
1

0
0

~J =

0
0
0
0
0

k2
-k2
0
0

0
-k2
k2
0
0

0
0
0
0
0

0
0
0
0
0

Section 2.5

Transpose of a Matrix

79

As you have seen from the previous examples, we can use a positioning matrix,
such as [A] and its transpose, ancl methodically create the global stiffness matrix for
finite element models.
In general, to obtain the transpose of a matrix [B] of size m X n, the first row of
the given matrix becomes the first column of the [BY, the second row of [B] becomes
the second column of [B]\ and so on, leading to the mth row of [B] becoming the mth
column of the [B]\ resulting in a matrix with the size of n X m. Clearly, if you take the
transpose of [B]\ you will end up with [B]. That is,

([BYl

= [B]

(2.10)

As you will see in succeeding chapters, in order to save space, we write the solution
matrices , which are column matrices, as row matrices using the transpose of the
solution-another use for transpose of a matrix. For example, we represent the
displacement solution

V,
V2
{V}

V3

by [VY

= [V,

V 2 V3

Vn]

When pe rforming matrix operations dealing with transpose of matrices, the following identities are true:

([A] + [B] + ... + [N]l = [AY + [BY + ... + [NY

([AHB] ... [N])T = [NY ... [BY[AY

(2.11)
(2.12)

In Eq. (2.12), note the change in the order of multiplication.

This is a good place to define a symmetric matrix. A symmetric matrix is a square
matrix whose elements are symmetrical with respect to its principal diagonal. An example of a symmetric matrix follows.

[A] = [

~
-5

1~3
20

-5
20
8

Note that for a symmetric matrix, element amn is equal to anm . That is, amn
values of n andm. Therefore, for a symmetric matrix, [A] = [A

EXAMPLE 2.3
Given the following matrices:

[A]

~ [~

5
3
-2

~]9

and [B]

[~1

6
2

~2]

-4

= anm for all

80

Chapter 2

Matrix Algebra

perform the following operations:

a. [Af = ? and [Bf = ?
b. verify that ([A] + [BDT = [Af + [Bf
c. verify that ([A][BDT = [Bf[Af
a. [AV = ? and [Bf = ?
As explained earlier, the first , second, third, ... , and the mth rows of a matrix become the first , second, third, .. . , and the mth columns of the transpose matrix
respectively.

Similarly,

7
2

=[ :

[BY

-2
b. Verify that ([A]

[A[ + [B]

+ [B]l = [AV +

=[I~
=[~

10

[A[T + [B[T

-2]

11
5
1
8
3
7

10
5

[BY.
and ([A]

~2]
+[ :
9
-2

7
2
3

= [4
11

15
5

-2

10

~ ]=[1~

15
5
10

[BDT

-4

-2

In
In

Comparing results, it is clear that the given identity is true.

c. Verify that ([AHB]f = [BV[AY.
In Example 2.2 we computed the product of [A][B] :

[AIIB]

[~

5
,.,

.)

-2

and using the results we get

([A][BDT

[35
= 10

m~

6
2
3

-2]
= [35
60
3

10
75

-4

77

31

IS ]

-35
-60

3177 ]

15

60
75
-35

7
2
3

q[~

-60

Alternatively,

[BJ'[A]' = [ :
-2

-4

8
3
7

~2]
[~~15
9
=

60
75
-35

31 ]
77

-60

Again , by comparing results we can see that the given identity is true.

Section 2.6

2.6

Determinant of a Matrix

81

DETERMINANT OF A MATRIX

Up to this point we have defined essential matrix terminology and discussed basic matrix operations. In this section we define what is meant by a determinant of a matrix. As
you will see in the succeeding sections, determinant of a matrix is used in solving a set
of simultaneous equations, obtaining the inverse of a matrix, and forming the characteristic equations for a dynamic problem (eigenvalue problem).
Let us consider the solution to the following set of simultaneous equations:
a11xI
a 21x I

+ al2x 2 = b l
+ a 22 x 2 = b2

(2.13a)
(2.13b)

We can represent Eqs. (2.13a) and (2.13b) in a matrix form by

(2.14)
or in a compact form by

[A]{X}

= {B}

To solve for the unknowns XI and X2, we may first solve for X2 in terms of XI> using
Eq. (2.13b), and then substitute that relationship into Eq. (2.13a). These steps are shown

next.

Solving for

XI

(2.15a)
After we substitute for XI in either Eq. (2.13a) or (2.13b), we get
(2.15b)
Referring to the solutions given by Eqs. (2.15a) and (2.15b) , we see that the denominators in these equations represent the product of coefficients in the main diagonal
minus the product of the coefficient in the other diagonal of the [A] matrix.
The a11a22 - a 12a21 is the determinant of the 2 X 2 [A] matrix and is represented

in one of following ways:

Dct[A] or dct[A] orl al1

(2.16)

a21

Only the determinant of a square matrix is defined. Moreover, keep in mind that the determinant of the [A] matrix is a single number. That is, after we substitute for the values

82

Chapter 2

Matrix Algebra

of al1 , an, a12, and a2l into al1a22 - al2a21> we get a single number. In general , the
determinant of a matrix is a single value. However, as you will see later, for dynamic
problems the determinant of a matrix resulting from equations of motions is a polynomial expression.
Cramer's rule is a numerical technique that can be used to obtain solutions to a
relatively small set of equations similar to the previous example. Using Cramer's rule we
can represent the solutions to the set of simultaneous equations given by Eqs. (2.l3a)
and (2.l3b) with the following determinants:

and

I:~: ~~I

X2

(2.17)

= -71-a-11--a-12':..1
a21

a22

Let us now consider the determinant of a 3 X 3 matrix such as

which is computed in the following manner:

Cl1

CI2

CI3

C21

C22

C23

CllC22 C33

C31

Cn

C12 C23 C31

Cl1 C23 C32 -

C13 C21C32 -

C13 C22C31

(2.18)

CI2 C21C33

C33

There is a simple procedure called direct expansion, which you can use to obtain
the results given by Eq . (2.18). Direct expansion proceeds in the following manner.
First , we repeat and place the first and the second columns of the matrix [C] next to the
third column, as shown in Figure 2.1. Then we add the products of the diagonal elements lying on the solid arrows and subtract them from the products of the diagonal
elements lying on the dashed arrows. This procedure, shown in Figure 2.1 , results in the
determinant value given by Eq. (2.18).
The direct expansion procedure cannot be used to obtain higher order determinants. Instead, we resort to a method that first reduces the order of the determinantto what is called a minor-and then evaluates the lower order determinants. To
demonstrate this method, let's consider the right-hand side of Eq. (2.18) and factor out
CII> -CI2, and CI 3 from it. This operation is shown below.
Cl1 Cn C33

C12 C23C31

= Cl1 (C22C33

+ C13 C21C32

C23Cn ) -

C13 C22 C31 -

cd C21C33

CllC23 C32 -

C23C31)

C12 C21C33

+ Cl3( C21Cn

C22C31)

As you can see, the expressions in the parentheses represent the determinants of
reduced 2 X 2 matrices. Thus, we can express the determinant of the given 3 X 3 matrix

Section 2.6

Determinant of a Matrix

83

FIGURE 2.1 Direct expansion procedure

for computing the detenuinant of (a)
2 X 2 matrix, and (b) 3 X 3 matrix.

(a)

in terms of the determinants of the reduced 2

manner:

2 matrices (minors) in the following

A simple way to visualize this reduction of a third-order determinant into three

second-order minors is to draw a line through the first row, first column, second
column , and third column. Note that the elements shown in the box-the common
element contained in elimination rows and columns-are factors that get multiplied
by the lower order minors. The plus and the minus signs before these factors are assigned based on the following procedure. We add the row and the column number of
the factor, and if the sum is an even number, we assign a positive sign , and if the sum
is an odd number we assign a negative sign to the factor. For example, when deleting
the first row and the second column (1 + 2) , a negative sign should then appear before
the Cl2 factor.

31

C22

C23

C21

C 2

C23

C32

C33

C31

C 2

C33

It is important to note that alternatively to compute the determinant of [A] we

could have eliminated the second or the third row-instead of the first row-to reduce
the given determinant into three other second-order minors. This point is demonstrated
in Example 2.4.
Our previous discussion demonstrates that the order of a determinant may be
reduced into other lower order minors, and the lower order determinants may be used
to evaluate the value of the higher order determinant.
Here are two useful properties of determinants: (1) The determinant of a matrix
[A] is equal to the determinant of its transpose [Af This property may be readily
verified (see Example 2.4). (2) If you multiply the elements of a row or a column of
a matrix by a scalar quantity, then its determinant gets multiplied by that quantity as
well.

84

Chapter 2

Matrix Algebra

EXAMPLE 2.4

Given the following matrix:

5
3
-2
calculate
H. determinant of [A]
b. determinant of [A f

a. For this example, we use both the direct expansion and the minor methods to
compute the determinant of [A]. As explained earlier, using the direct expansion
method, we repeat and place the first and the second columns of the matrix next
to the third column as shown, and compute the products of the elements along
the solid arrows, then subtract them from the products of elements along the
dashed arrow.

,*)
",'5'
~
~ff";-7i~,
8

""",

-' 3

"

150
837
6 -2 9

(1)(3)(9) + (5)(7)(6) + (0)(8)( -2) - (5)(8)(9)

- (1)(7)(-2) - (0)(3)(6) = -109

Next, we use the minor to compute the determinant of [A]. For this example, we eliminate the elements in the first row and in first , second , and third
columns, as shown.

~ ~
~ ~2 ~

~ +2 ~

~
~ ~2 ~

: ~2 ~ = 11~2 ~H~ ~I + l~ ~21

= (1)[(3)(9)-(7)( -2)]-(5)[(8)(9)-(7)(6)] = -109
Alternatively, to compute the determinant of the [A] matrix, we can eliminate the elements in the second row, and first , second, and third column as shown:

~
~

~
~

42

~
~ ~2 ~

Section 2.6

Determinant of a Matrix

85

1
8
6

= - (8)[ (5)(9) - (0)( -2)]+(3)[ (1 )(9) - (0)( 6) ]-(7)[ (1)( -2) -(5)( 6)] = -109
b. As already mentioned, the determinant of [AYis equal to the determinant of [A].
There fore, there is no need to perform any additional calculations. However, as a
means of verifying this identity, we will compute and compare determinant of
[AY to the determinant of [A]. Recall that [AY is obtained by interchanging the
first , second, and third rows of [A] into the first , second, and third columns of the

[1 8 6]

[AY, leading to [AY = 5 3 -2. Using minors, we get

079

~
~ ; ~~
1
5

8
3

~
~ ; 1~

~
~ t ~~

= (1)[(3)(9)-(-2)(7)]-(8)[(5)(9)-(-2)(0)] + (6)[(5)(7)-(3)(0)] = -109

When the determinant of a matrix is zero, the matrix is called a singlliar. A singular matrLx results when the elements in two or more rows of a given matrix are identical. For
example, consider the following matrix:

[2 1 4]

[A] = 2

1
3

4
5

whose rows one and two are identical. As shown below, the determinant of [A] is zero.

2
2
1

1
1
3

4
4
5

(2)(1)(5) + (1)(4)(1) + (4)(2)(3) - (1)(2)(5)

- (2)(4)(3) - (4)(1)(1) = 0

Matrix singularity can also occur when the elements in two or more rows of a matrix are
linearly dependent. For example, if we multiply the elements of the second row of matrix [A] by a scalar factor such as 7 , then the resulting matrix,

[A]

[2 1 4]
14 7 28
135

86

Chapter 2

Matrix Algebra

is singular because rows one and two are now linearly dependent. As shown below, the
determinant of the new [A] matrix is zero.

14

1
7

4
28

2.7

(2)(7)(5) + (1)(28)(1) + (4)(14)(3) - (1)(14)(5)

(2)(28)(3) - (4)(7)(1) = 0

=-

SOLUTIONS OF SIMULTANEOUS LINEAR EQUATIONS

As you saw in Chapter 1, the finite element formulation leads to a system of algebraic
equations. Recall that for Example 1.1, the bar with a variable cross section supporting
a load, the finite element approximation and the application of the boundary condition
and the load resulted in a set of four linear equations:
1820

-845
103[

o
o

-845
1560
-715

o
-715
1300
-585

-~85] {:::} ~
={

585

llS

103

In the sections that follow we discuss two methods that you can use to obtain solutions to a set of linear equations.
Gauss Elimination Method

We begin our discussion by demonstrating the Gauss elimination method using an

example. Consider the following three linear equations with three unknowns, Xl> X2 ,
and X3'

+ X3 = 13
3x l + 2x2 + 4x3 = 32
5Xl - X2 + 3X3 = 17
2Xl

X2

(2.19a)
(2.19b)
(2.19c)

1. We begin by dividing the first equation, Eq. (2.19a), by 2, the coefficient of

term. This operation leads to

Xl

(2.20)
2. We multiply Eq. (2.20) by 3, the coefficient of Xl in Eq. (2.19b).
(2.21)

Section 2.7

Solutions of Simultaneous linear Equations

We then subtract Eq. (2.21) from Eq. (2.19b). This step eliminates
Eq. (2.19b). This operation leads to
3xI

87

from

XI

+ 2X2 + 4X3 = 32

+ 1.
x, + ~ x- = 39)
2 - 2 ~
2

- (3XI

25

2"X 2 + 2"X 3 = 2

(2.22)

3. Similarly, to eliminate XI from Eq. (2.19c), we multiply Eq. (2.20) by 5, the coefficient of XI in Eq. (2.19c)
(2.23)
We then subtract the above equation from Eq. (2.19c), which eliminates
Eq. (2.19c). This operation leads to
5xI -

X2

65)
=2

from

+ 3X3 = 17

5
5
( 5xI + 2"
X 2 + 2"X 3
-"2 X2

XI

+ "2X3 =

31

-2

(2.24)

Let us summarize the results of the operations performed during steps 1 through
3. These operations eliminated the XI from Eqs. (2.1%) and (2.19c).
1

13

25

+ "2 X2 + "2 X3 = 2

XI

1
"2 X2

7
--x
2 2

+ "2X3 = 2
1

+ -x2 ~

31
2

=--

(2.25a)
(2.25b)
(2.25c)

4. To eliminate X2 from Eq. (2.25c), first we divide Eq. (2.25b) by 112, the coefficient
of X2 .
X2

+ 5X3 = 25

(2.26)

Then we multiply Eq. (2.26) by -7/2, the coefficient of X2 in Eq. (2.25c), and subtract that equation from Eq. (2.25c). These operations lead to

7
-"2 X2

- (

-~

X2 -

+ "2X3 =

31

X3

= -

1~X3

= 72

~)
(2.27)

88

Chapter 2

Matrix Algebra

Dividing both sides of Eq. (2.27) by 18, we get

X3

=4

Summarizing the results of the previous steps, we have

x,

X2

13

="2

(2.28)

+ 5X3 = 25

(2.29)

+ '2X2 + '2X3

(2.30)

Note Eqs. (2.28) and (2.29) are the same as Eqs. (2.25a) ancl (2.26), which are
renumbered for cOllvenience. Now we can use back substitution to compute the
values of X2 and x,. We substitute for X3 in Eq. (2.29) and solve for X2'
x2

Next, we substitute for

+ 5(4) = 25

X3 and X2

x2

=5

in Eq. (2.28) and solve for

x,.

The Lower Triangular, Upper Triangular (LU)

Decomposition Method

When designing structures, it is often necessary to change the load and consequently the
load matrix to determine its effect on the resulting displacements and stresses. Some heat
transfer analysis also requires experimenting with the heat load in reaching the desirable
temperature distribution within the medium. The Gauss elimination method requires full
implementation of the coefficient matrix (the stiffness or the conductance matrix) and
the right-hand side matrix (load matrix) in order to solve for the unknown displacements (or temperatures). When using Gauss elimination, the entire process must be repeated each time a change in the load matrix is made. Whereas the Gauss elimination
is not well suited for such situations, the LU method handles any changes in the load
matrix much more efficiently. The LU method consists of two major parts: a decomposition part and a solution part. We explain the LU method using the following three
equations

a 2'x,

+
+

a 31x l

+ a 32 x 2 + a 33 x 3

a"x,

a'2x 2
a n X2

+
+

aUx3 =

b,

(2.31)

a n X3 =

b2

(2.32)

b3

(2.33)

or in a matrix form ,

(2.34)

Section 2.7

solutions of Simultaneous Linear Equations

89

Decomposition Part The main idea behind[ ;he LOU ~e]thOd is first to decompose the coefficient matrix [A] into lower [L] =
Illl

[U]= [ 0

1112
1122

1l13]
1123

121

131

In

and upper triangular

matrices so that

1133

a21

a22

a,,] = [ 1
a23

121

0
1

a 31

a32

a33

131

132

a12

[a"

~]['~'

u,,]

1112

1122

ll23

(2.35)

ll33

Carrying out the multiplication operation, we get

[a"

al2

a21

a22

a,,] = [
a23

/21

0
1

a31

an

a33

131

In

[""

nu'

1112

12 III 11

U"

121 1113 +

+ ll22
+ 1321122

1211112

13111ll

1311112

13 III 13

1112

1122

",,]
1123
1133

(2.36)

1123

+ ll33

Inll23

Now let us compare the elements in the first row of [A] matrix in Eg. (2.36) to the
elements in the first row of the [L][U] multiplication results. From this comparison we
can see the following relationships:
llll

= all and lll2 = al2 and 1113 = au

Now, by comparing the elements in the first column of [A] matrix in Eg. (2.36) to
the elements in the first column of the [LHu] product, we can obtain the values of 121

and 131 :
1211111

1311111

= a21
= a31

21 -

a21 _ a21
all

(2.37)

a31 _ a31
Illl
all

(2.38)

Illl

31 -

Note the value of Illl was determined in the previous step. That is, 1111 = all' We can obtain the values of ll22 and liz3 by comparing the elements in the second rows of the matrices in Eg. (2.36).
1211112

12111 13

+ ll22 = a22
+ ll23 = a23

llzz

= a22

lzlllt2

(2.39)

Un

= a23

1211113

(2.40)

When examining Egs. (2.39) and (2.40) remember that the values of 121> 1l12, and ll13 are
known from previous steps. Now we compare the elements in the second columns of

90

Chapter 2

Matrix Algebra

Eq. (2.36). This comparison leads to the value of 132 Note, we already know values of
from previous steps.

U12 , ~h U22 , and 131

132

an - 1311112
= --:.::'---.....::..:.~

(2.41)

1122

Finally, the comparison of elements in the third rows leads to the value of 1l33.
(2.42)
We used a simple 3 X 3 matrix to show how the LU decomposition is performed.
We can now generalize the scheme for a square matrix of any size n in the following
manner:
Step 1. TIle values of the elements in the first row of the [V] matrix are obtained from

= al;

Ill;

for j

= 1 to n

(2.43)

Step 2. The unknown values of the elements in the first column of the [L] matrix are
obtained from
Iii

= -ail

1111

for i

= 2 to n

(2.44)

Step 3. The unknown values of the elements in the second row of the [V] matrix are
computed from
(2.45)
Step 4. TIle values of the elements in the second column of [L] matrix are calculated from
ai2 -

In

lilll12

1122

for i

= 3 to n

(2.46)

Next, we determine the unknown values of the elements in the third row of the [V]
matrix and the third column of [L]. By now you should see a clear pattern. We evaluate the values of the elements in a row first and then switch to a column. This procedure
is repeated until all the unknown elements are computed. \Ve can generalize the above
steps in the following way. To obtain the values of the elements in the kth row of [V] matrix, we use
k-I

Ilki

= ak;

"LlkPllp;

for j

=k

to n

(2.47)

p=1

We will then switch to the kth column of [L] and determine the unknown values in that
column.
k-I

aik lik

"L/ipllpk

p=1

= ---'----Ilkk

for i

= k + 1 to n

(2.48)

Solution Purf So far you have seen how to decompose a square coefficient
matrix [A] into lower ancl upper triangular [L] and [V] matrices. Next, we use the [L]

Section 2.7

Solutions of Simultaneous Linear Equations

91

and the [U] matrices to solve a set of linear equations. Let's turn our attention back to
the three equations and three unknowns example and replace the coefficient matrix
[A] with the [L] and [U] matrices:

[A]{x} = {b}

(2.49)

= {b}

(2.50)

[L][U]{x}

We now replace the product of [U]{x} by a column matrix {z} such that

= {z}

(2.51)

= {b} ~ O[L]{z} = {b}

(2.52)

[U]{x}
{z}
~

[L][U]{x}

Because [L] is a lower triangular matrix, we can easily solve for the values of the
elements in the {z} matrix, and then use the known values of the {z} matrix to solve for
the unknowns in the {x} from the relationship [U] {x} = {z}. These steps are demonstrated next.
(2.53)

From Eq. (2.53), it is clear that

Zl

= bl

Z2

= b2

Z3

= b3 -

(2.54)
-

(2.55)

12\zl
131Z1 -

(2.56)

132z2

Now that the values of the elements in the {z} matrix are known , we can solve for the
unknown matrix {x} using
til 2

[I

ti?J.

lh ,
II"

1133

Jr' } {" }
x?

X3

Z3
X3 = 1133
Z2 -

z?

(2.57)

Z3

(2.58)
1123X3

(2.59)

X2 =
1122

Zl - 1112X2 XI =

/l\3X3

(2.60)

llll

Here we used a simple three equations and three unknowns to demonstrate how
best to proceed to obtain solutions; we can now generalize the scheme to obtain the solutions for a set of 11 equations anclll unknown.

92

Chapter 2

Matrix Algebra
i-I

ZI

= bl

and

= bi

Zi

~/ijZj

fori = 2,3,4, ... ,n

(2.61)

i~1

for i

=n

- 1, n - 2, n - 3, ... ,3,2,1

(2.62)

Next, we apply the LU method to the set of equations that we used to demonstrate
the Gauss elimination method.
EXAMPLE 2.5

Apply the LU decomposition method to the following three equations and three
unknowns set of equations:
2xI +
3xI +
5xI -

[A]

+ X3 = 13
2X2 + 4X3 = 32
x2 + 3x3 = 17
X2

~ G ~1 ~}nd ~
(b)

Note that for the given problem, n

nn

= 3.

Decomposition Part
Step 1. The values of the elements in the first row of the [V] matrix are obtained from
llli
llll

= ali
= all = 2

= 1 to n
lll2 = al2 = 1

for j

Step 2. The unknown values of the elements in the first column of the [L] matrix are
obtained from

III =~
Ull
121 --

for i

a21 _ 3
-Ull
2

= 2 to n

131 --

a31 Ull

Step 3. The unknown values of the elements in the second row of the [V] matrix are
computed from
U2j

= a2j

U22

a22 -

1211112

U23

a23 -

121ll\3

= 4

121 ulj

for j

= 2 to n

2-(%)(1) ~
-(%)(1) &
=

Section 2.7

Step 4.

Solutions of Simultaneous Linear Equations

The unknown values of th e elements in the second column of the [L] matrix
are determined from

= 3 to n

for i

a 32 -

32 -

Step 5.

93

= -1-(-25)(1) = -7

13\ll\ 2

ll 22

2
Compute the remaining unknown elements in the [U] and [L] matrices.
k-\

llkj

ll33

= akj

= a 33

for j

2:)k p llpj

r\
(l3\ll\3

= k to n

+ In U23 ) =

3- ((%)(1) + (-7)(%) ) = 18

Because of the size of this problem (n = 3) and the fact that the elements along
the main diagonal of the [L] matrix have values of 1, that is, 133 = 1, we do not need to
proceed any furthel: Therefore, the application of the last step
k-\

~/iPllpk

aik -

p=1

lik

= ---'----

for i

= k + 1 to n

llkk

is omitted. We have now decomposed the coefficient matrix [A] into the following lower
and upper triangular [L] and [U] matrices:
1

3
1
2
5
-7
2
When performing this method by hand, here is a good place to check the decomposition results by premultiplying the [L] matrix by the [U] matrix to see if the [A] matrix is recovered.
We now proceed with the solution phase of the LU method, Eg. (2.61).
Solution Part
i-I

ZI

= bl

and

Zi

= bi

~/ijZj

for i

= 2 to n

j=\

13

Z\

Z3

= b3 -

Z2

(l3\Z\

= b2 + 132 z2 )

32 - (~)(13) = 2;
= 17 - ((%)(13) + (_7)(2;)) = 72

12\z\

94

Chapter 2

Matrix Algebra

The solution is obtained from Eq. (2.62).

n

Zi Xn = -

Zn

and

llijXi

;=i+1

for i

Xi = -----'----

=n-

1, n - 2, n - 3, ... ,3, 2,1

Uii

linn

Note for this Problem 11 = 3, therefore i = 2, 1.

X3

= -Z3 = -72 = 4
18

1133

X2

Z2 -

5 )(4)
225 - (-2

1123 X 3
1122

=5

1
2

XI

2.8

ZI - lll2x2 - ll\3x3
llll

13 - 1)(5)
2

+ (1)(4

=2

INVERSE OF A MATRIX
In the previous sections we discussed matrix addition , subtraction, and multiplication ,
but you may have noticed that we did not say anything about matrix division. That is because such an operation is not formally defined. Instead, we define an inverse of a matrix in such a way that when it is multiplied by the original matrix , the identity matrix is
obtained.

(2.63)

In Eq. (2.63), [A l is called the inverse of [A]. Only a square and nonsingular matrix has
an inverse. In Section 2.7 we explained the Gauss elimination and the LU methods that
you can use to obtain solutions to a set of linear equations. Matrix inversion allows for
yet another way of solving for the solutions of a set of linear equations. Once again , recall from our discussion in Chapter 1 that the finite element formulation of an engineering problem leads to a set of linear equations, and the solution of these equations
render the nodal values. For instance, formulation of the problem in Example 1.1 led
to the set of linear equations given by

[KHII}

= {F}

(2.64)

To obtain the nodal displacement values {II}, we premultiply Eq. (2.64) by [Kr l ,
which leads to
[I)
~

[Krl[K]{II}
[/HII}
and noting that [/HII}

= [Krl{F}

= [Krl{F}

(2.65)
(2.66)

= {II} and simplifying,

{II} = [Kj-I{F}

(2.67)

Section 2.8

Inverse of a Matrix

95

From the matrix relationship given by Eq. (2.67), you can see that the nodal solutions can be easily obtained, provided the value of [Ktl is known. This example shows
the important role of the inverse of a matrix in obtaining the solution to a set of linear
equations. Now that you see why the inverse of a matrix is important , the next question
is, How do we compute the inverse of a square and nonsingular matrix? There are a
number of established methods that we can use to determine the inverse of a matrix.
Here we discuss a procedure based on the LU decomposition method. Let us refer back
to the relationship given by Eq. (2.63), and decompose matrix [A] into lower and upper
triangular [L] and [U] matrices.
[r\)
~

[L][U]

(2.68)

[Atl = [I]

Next, we represent the product of [U][Atl by another matrix, say matrix [Y]:

= [Y]

[U][Atl

(2.69)

and substitute for [U][Afl in terms of [Y] in Eq. (2.68), which leads to

[L][Y]

= [I]

(2.70)

We then use the relationships given by Eq. (2.70) to solve for the unknown values of elements in matrLx [Y] , and then use Eq. (2.69) to solve for the values of the elements in
matrix [Atl. These steps are demonstrated using Example 2.6.

EXAMPLE 2.6
Given [A]
Step 1.

~G ~
1
2
-1

compute [Ar',

Decompose the given matrix into lower and upper triangular matrices. In
Example 2.5 we showed the procedure for decomposing the [A] matrix into
lower and upper triangular [L] and [U] matrices.
1

1
1
2

-7

[L] =

Step 2.

2
5
2

1]

Use Eq. (2.70) to determine the unknown values of the elements in the [Y] matrix.
[I.)
~

~ ~

: [;.

Pl
-------...

~~: ~~ ~~:] =

Y31

-7

Y32

Y33

1
[00

100

~1]

96

Chapter 2

Matrix Algebra

First, let us consider the multiplication results pertaining to the first column of
the [Y] matrix, as shown.

3
2

-7

5
2

u:}~m

The solution of this system of equations leads to

YII

=1

Y21

= --2

Y31

= -13

Next, consider the multiplication results pertaining to the second column of [Y]:

o o

1
3
2
5
2

-7

The solution of this system of equation yields

Y12

=0

Y22

=1

=7

Y32

Similarly, solve for the unknown values of the elements in the remaining
column of the [Y] matrix:

1
3
2
5
2
Y13

-7

=0

Y 23

{2}~m
=0

Y33

=1

Now that the values of elements of the [YJ are known, we can proceed with calculation
of the values of the elements comprising the [Atl, denoted by XII, X12," . , and so on ,
as shown. Using the relationship given by Eq. (2.69) , we have

[U][Ar l

= [Y]

[: ~ i][;::

:::] = [
X33

~& ~

-13

:]
1

Section 2.8

Inverse of a Matrix

97

Again , we consider multiplication results pertaining to one column at a time. Considering the first column of the [x] matrix,

[: ~ i]{;: }
o

18

= -18

~&}

-13

31

13

X 31

={

11

X 21

= 18

10

XlI

= 18

Multiplication results for the second column render

= 18

X 32

X22

= 18

XI 2

= -18

TIle multiplication results of the third column yield

X\3

= 18

Therefore, the inverse of the [A] matrix is

[Ar l

10
[
= J.- 11
18 -13

-4
1
7

We can check the result of our calculations by verifying that [A][Arl

1
2
-1

1][
1011
4
3

-13

= [I].

-4
1
7

Finally, it is worth noting that the inversion of a diagonal matrix is computed simply by
inversine its elements. That is, the inverse of a diagonal mat.rix is also a diagonal mat.rix

98

Chapter 2

Matrix Algebra

with its elements being the inverse of the elements of the original matrix. For example,
the inverse of the 4 x 4 diagonal matrix

la,
[AJ = ~

0
a2
0
0

0
0

a3

qS[At' =

a4

1
al

0
0

llz
0

1
a3

a4

This property of a diagonal matrix should be obvious because [Arl[A]

1

al
0

[Arl[A]

=
0
0

0
1

a2
0
0

a3

l~'

a2
0
0

0
0

a3
0

= [I].

~}l~ ~]
0
1
0
0

0
0
1
0

a4

2.9 EIGENVALUES AND EIGENVECTORS

Up to this point we have discussed some of the methods that you can use to solve a set
of linear equations of the form

[A]{x}

= {b}

(2.71)

For the set of linear equations that we have considered so far, the values of the elements
of the {b} matrix were typically nonzero. This type of system of linear equations is commonly referred to as l1onhomogenous. For a nonhomogenous system, unique solutions
exist as long as the determinant of the coefficient matrix [A] is nonzero. We now discuss
the type of problems that render a set of linear equations of the form

[A]{X} - A{X} = 0

(2.72)

This type of problem, called an eigenvalue problem, occurs in analysis of buckling problems, vibration of elastic structures, and electrical systems. In general, this class of problems has nonunique solutions. That is, we can establish relationships among the
unknowns, and many values can satisfy these relationships. It is common practice to
write Eq. (2.72) as

[[A] - A[I]]{X}

=0

(2.73)

where [I] is the identity matrix having the same dimension as the [A] matrix. In
Eg. (2.73), the unknown matrix {X} is called the eigenvector. We demonstrate how to
obtain the eigenvectors using the following vibration example.