MSA / L-7 CSE-3512 - Numerical Methods 18th February, 2005

System of Linear Equations

A linear equation involving two variables x and y has the standard form ax + by = c, where
a, b and c are real number and a and b cannot both equal zero. Notice that the exponent
(power) of variables is one. Some examples of linear equations are:
i) 4x + 7y = 15 ii) -x – 2/3 y = 0 iii) 3u – 2v = -1/2
The equation becomes nonlinear if any of the variables has the exponent other then one.
Similarly, equations containing terms involving a product of two variables also considered
nonlinear. Some examples of nonlinear equations are:
i) 2x – xy + y = 2 ii) x2 + y2 = 25 iii) x + √x = 6
In practice, linear equations occur in more than two variables. A linear equation with n
variables has the form a1x1 + a2x2 + a3x3 + ... ... + anxn = b where ai (i = 1, 2, 3, ..., n)
are real numbers and at least one of them is not zero. The main concern here is to solve for x i
(i = 1, 2, 3, ..., n), given the values of ai and b. Note that an infinite set of xi values will satisfy
the above equation. There is no unique solution. If we need a unique solution of an equation
with n variables (unknowns), then we need a set of n such independent equations. This set of
equations is known as system of simultaneous linear equations (or simply, system of linear
equations).
A system of n linear equations is represented generally
a11x1 + a12x2 + a13x3 + ... ... + a1nxn = b1
a21x1 + a22x2 + a23x3 + ... ... + a2nxn = b2
... ... ... ... ... ... ... ... ... … …
... ... ... ... ... ... ... ... ... … …
an1x1 + an2x2 + an3x3 + ... ... + annxn = bn
In matrix notion, we can express as Ax = b where A is an n x n matrix, b is
an n vector and x is a vector of n unknowns.
The techniques and methods for solving system of linear algebraic equations belong to two
fundamentally different approaches:
a. Elimination approach: Elimination approach, also known as direct method, reduces the
given system of linear equations to a form from which the solution can be obtained by simple
substitution.
b. Iterative approach: Iterative approach, as usual, involves assumption of some initial values
which are then refined repeatedly till they reach some accepted level of accuracy.
Existence of Solution
In solving systems of linear equations, we are interested in identifying values of the variables
that satisfy all equations in the system simultaneously. Given an arbitrary system of
equations, it is difficult to say whether the system has a solution or not. Sometimes there may
be solution but it may not be unique. There are four possibilities:
a. System has a unique solution
b. System has no solution
c. System has a solution but not a unique one (i.e. it has infinite solutions)
d. System is ill – conditioned
Unique solution
Consider the system x + 2y = 9 and 2x – 3y = 4. The system has a solution x = 5 and y = 2.
Since no other pair of values of x and y would satisfy the equation, the solution is said to be
unique. The system is illustrated in Fig 7.1(a). [Page 208, Balagurusamy].

Page 1 of 12
MSA / L-7 CSE-3512 - Numerical Methods 18th February, 2005

No Solution
The equations 2x – y = 5 and 3x – 3/2y = 4 have no solution. These two lines are parallel as
shown in Fig 7.1(b). and, therefore, they never meet. Such equations are called inconsistent
equations.
No Unique Solution
The system -2x + 3y = 6 and 4x – 6y = -12 has many different solutions. We can see that
these are two different forms of the same equation and, therefore, they represent the same line
[Fig 7.1(c)]. Such equations are called dependent equations.
Ill-Conditioned Systems
The system x - 2y = -2 and 0.45x – 0.91y = -1 has a solution but it is very difficult to
identify the exact point at which the lines intersect [Fig 7.1(d)]. Such systems are said to be
ill-conditioned. Ill-conditioned systems are very sensitive to roundoff errors and, therefore,
may pose problems during computation of the solution.
Let us consider a general form of a system of linear equations of size m x n.
a11x1 + a12x2 + a13x3 + ... ... + a1nxn = b1
a21x1 + a22x2 + a23x3 + ... ... + a2nxn = b2
... ... ... ... ... ... ... ... ... … …
... ... ... ... ... ... ... ... ... … …
am1x1 + an2x2 + an3x3 + ... ... + amnxn = bm
In order to effect a unique solution, the number of equations m should be equal to the number
of unknowns n. If m < n, the system is said to be under-determined and a unique solution for
all unknowns is not possible. On the other hand, if the number of equations is larger than the
number of unknowns, then the set is said to be over-determined, and a solution may or may
not exist.
The system is said to be homogeneous when the constants bi are all zero. Otherwise, the
system is called a non-homogeneous system (at least one bi ≠ 0).
Overview of Matrices
Matrix
A matrix is a rectangular array of elements arranged in rows and columns. If the matrix
contains m rows and n columns, the matrix is said to be of size m x n. The element in the ith
row and jth column of the matrix A is denoted by aij. For example,
a11 a12 a13
A = a21 a22 a23 is a 3 x 3 matrix.
a31 a32 a33
Square Matrix: A matrix in which rows m is equal to columns n is called square matrix.
Identity Matrix: A square matrix in which all the diagonal elements are one and other
elements are zero is called identity matrix. That is aij = 1 for i = j and aij = 0 for i ≠ j.
Identity matrices are denoted by I.
Row Vector: A matrix with one row and n columns is called row vector.
Column Vector: A matrix with m rows and one column is called column vector.
Transpose Matrix: The matrix AT is called the transpose of A if the element aij in A is equal
to element aji in AT for all i and j. For example, if
1 2
A= 3 4 then AT = 1 3 4
5 6 2 4 6
It is clear that AT is obtained by interchanging the rows and the columns of A.

Page 2 of 12
MSA / L-7 CSE-3512 - Numerical Methods 18th February, 2005

Upper and Lower Triangular Matrix: A square matrix is said to be an upper triangular
matrix if aij = 0 for i > j and a lower triangular matrix if aij = 0 for i < j. For example:

1 2 3 1 0 0
A= 0 2 4 is an upper triangular matrix and B = 2 2 0 an lower triangular matrix.
0 0 3 3 4 3
Matrix Algebra
We can perform only three operations, namely addition, subtraction and multiplication on the
matrices. The division, although not defined, is replaced by the concept of inversion.
Two matrices A and B can be added together (or subtracted from each other) if they are of
the same order. Then, Sum C = A + B can be obtained by adding the corresponding
elements. That is, cij = aij + bij for all i and j
Similarly, Difference E = A – B can be obtained by subtracting the corresponding
elements. That is, cij = aij - bij for all i and j
Two matrices A and B can be multiplied in the order AB if and only if the number of
columns of A is equal to the number of rows of B. That is, if A is of order m x r, then B
should be of order r x n, where m and n are arbitrary values. In such cases, we may obtain,
P = AB which is order of m x n.
r
Elements of the matrix P is given by pij = 
k 1
aik bkj for all i and j

Determinants
Every square matrix A is associated with a number called its determinant, which is denoted
by | A |. The determinant of a 2 x 2 matrix, say A, is written in the form a11 a12
a21 a22
and the determinant is computed as | A | = a11 a22 - a12 a21.
Similarly, for a 3 x 3 matrix the determinant is given by
| A | = a11a22a33 + a12a23a31 + a13a21a32 – a11a32a23 – a21a12a33 – a31a22a13
For large matrices, the determinant is much more difficult to define and compute manually.
In general, for n x n matrix, the determinant will contain a sum of n! signed product terms,
each having n elements. [See page-508, Balagurusamy]
 For a triangular matrix (in which all the elements below (or above) the diagonal are
zero), the determinant is the product of the diagonal elements.
 If | A | = 0, then A said to be a singular matrix; otherwise, it is said to be nonsingular.

Minors and Cofactors

The minor of a particular element is the determinant that remains after the row and the
column that contain the element have been deleted.
The cofactor of an element is its minor with a sign attached. The cofactor dij of an element aij
is given by dij = (-1)i+j Mij
Adjoint Matrix
If dij is the cofactor of the element aij of the square matrix A, then, by definition, the adjoint
matrix of A is given by adj (A) = DT
Inverse Matrix
If B and C are two n x n square matrices such that BC = CB = I (identity matrix) then, B is
called the inverse of C and C is the inverse of B. The common notation for inverses is B-1 and
C-1. That is, B-1B = I and C-1C = I.

Page 3 of 12
MSA / L-7 CSE-3512 - Numerical Methods 18th February, 2005

Solution of Linear Systems by Matrix Inversion Method

A linear system of n equations in n unknowns can be represented in matrix form as AX = B,
where A, X and B are matrices and are given by
a11 a12 a13 ... ... … a1n x1 b1
a21 a22 a23 ... ... … a2n x2 b2
A = ... ... ... ... ... ... ... X= . and B = .
... ... ... ... ... ... ... . .
an1 an2 an3 ... ... … ann xn bn

A is called coefficient matrix and B is known as constant vector. X is the required solution
and, therefore, it is called the solution vector. If we multiply the matrix equation AX = B
By A-1 on the both sides, we get A-1A X = A-1B
A-1 is the inverse matrix of A. We know that A-1A = I is the identity matrix.
Therefore, A-1A X = I X = X is the solution of the system of equations and is obtained from
X = A-1B = CB
If we know the inverse of the matrix A, we can obtain the solution vector X by post-
multiplying it by B.
 The inverse A of a square matrix A exists, if and only if, A is nonsingular (i.e. det A ≠ 0).
-1
-1
A is the matrix obtained from A by replacing each element aij by its cofactor dij and then
transposing the resulting matrix and dividing it by the determinant of A.
adj (A)
-1
C = A = -----------
det A
where adj (A) is the adjoint matrix of A and is given by the transpose of the cofactor
matrix of A. Then, adj (A) = DT where
d11 d12 d13 ... ... … d1n
d21 d22 d23 ... ... … d2n
D= ... ... ... ... ... ... ...
... ... ... ... ... ... ...
dn1 dn2 dn3 ... ... … dnn

dij is the cofactor of the element aij and is given by dij = (-1)i+j Mij
Mij is called the minor of aij and is taken as the determinant of matrix A after deleting i th
row and j th column.
 This method is obviously unsuitable for solving large systems, since the computation of
A-1by cofactors will then become exceedingly difficult.
Example: Solve the following system of linear equations using matrix inversion method
3x + y + 2z = 3
2x - 3y - z = -3
x + 2y + z = 4
Solution: Here
3 1 2 3
A = 2 -3 -1 B = -3
1 2 1 4
3 1 2
det A = 2 -3 -1 =8 since |A| ≠ 0, hence A-1 exists.
1 2 1

Page 4 of 12
MSA / L-7 CSE-3512 - Numerical Methods 18th February, 2005

3 2 1 -1 3 5 -1 3 5
1
AT = 1 -3 2 adj (A) = -3 1 7 A-1 = -3 1 7
8
2 -1 1 7 -5 -11 7 -5 -11
x -1 3 5 3 8 1
1 1
X= y = A-1 B = -3 1 7 -3 = 16 = 2
8 8
z 7 -5 -11 4 -8 -1
 x = 1, y = 2 and z = -1

Cramer’s Rule
A linear system of n equations in n unknowns can be represented in matrix form as AX = B,
where A, X and B are matrices and are given by
a11 a12 a13 ... ... … a1n x1 b1
a21 a22 a23 ... ... … a2n x2 b2
A = ... ... ... ... ... ... ... X= . and B = .
... ... ... ... ... ... ... . .
an1 an2 an3 ... ... … ann xn bn

A is called coefficient matrix and B is known as constant vector. X is the required solution
and, therefore, it is called the solution vector.
According to Cramer’s rule, the solution can be given in terms of determinants, i.e.
det Ai
Xi = ---------------
det A
where Ai is the matrix obtained from A by replacing the ith column with the vector B.
 Obviously, Cramer’s rule only works for det A ≠ 0. Onecan show that the system has a
unique solution if and only if det A ≠ 0. Otherwise, the system has either no solution or
infinitely many solutions.
 One would not use Cramer’s rule to solve a large system of linear equations, simply
because calculating a single determinant is as time consuming as solving the complete
system by a more efficient algorithm.
Example: Solve the following system of equations using Cramer’s rule:
2x1 + 3x2 = 12
4x1 – x2 = 10
Solution: 12 3
10 -1 -12 - 30 42
x1 = --------- = ------------ = ------- = 3
2 3 -2 - 12 14
4 -1
2 12
4 10 20 - 48 28
x2 = --------- = ------------ = ------- = 2
2 3 -2 - 12 14
4 -1
Triangular Factorization Method (Dolittle Decomposition method)
Self Study. See Page 230 – 235 [E. Balagurusamy].

Page 5 of 12
MSA / L-7 CSE-3512 - Numerical Methods 18th February, 2005

Basic Gauss Elimination Method

This is the elementary elimination method and it reduces the system of equations to an
equivalent upper triangular system, which can be solved by back substitution. Although quite
general, we shall describe this method by considering a system of three equations for the sake
of clarity and simplicity. Let the system be
a11x1 + a12x2 + a13x3 = b1
a21x1 + a22x2 + a23x3 = b2 ... ... ... ... ... ... ... … (1)
a31x1 + a32x2 + a33x3 = b3
We first form the augmented matrix of the system (1)
a11 a12 a13 b1
a21 a22 a23 b2 ... ... ... ... ... ... ... … (2)
a31 a32 a33 b3
To eliminate x1 from the second equation, we multiply the first equation by - a21 / a11 and
then add it to the second equation. Similarly, to eliminate x 1 from the third equation, we
multiply the first equation by - a31 / a11 and then add it to the third equation. - a21 / a11 and
- a31 / a11 are called the multipliers for the first stage of elimination. In this stage, we have
assumed that a11 ≠ 0. At the end of the first stage of elimination, the augment matrix becomes
a11 a12 a13 b1
0 a′22 a′23 b′2 ... ... ... ... ... ... ... … (3)
0 a′32 a′33 b′3
where a′22, a′23 ... are all changed elements.
Now, to eliminate x2 from the third equation, we multiply the second equation by - a′32 / a′22
and then add it to the third equation. Again, we assume that a22 ≠ 0. At the end of the second
stage, we have the upper triangular system
a11 a12 a13 b1
0 a′22 a′23 b′2 ...............................(4)
0 0 a″33 b″3
from which the values of x1, x2 and x3 can be obtained by the back substitution.
Example 6.11: Solve the following system by using Gauss elimination method
2x + y + z = 10
3x + 2y + 3z = 18
x + 4y + 9z = 16
Solution: In the first stage, the multipliers are –3/2 and –1/2. We multiply the first equation
successively by –3/2 and –1/2 and add it to the second and third equations respectively to
obtain the equations: 1/2 y + 3/2 z = 3 and 7/2 y + 17/2 z = 11
The augmented matrix therefore becomes
2 1 1 10
0 ½ 3/2 3
0 7/2 17/2 11
At the second stage, we eliminate y from the third equation by multiplying the second
equation by –7 and adding it to the third. The resulting system will be upper triangular
2x +y +z = 10
1/2 y + 3/2 z = 3
- 2 z = -10
Back substitution gives the solution:
x = 7, y = -9 and z = 5
Algorithm 7.1: Gauss Elimination (basic) Method
See Page-212 [Balagurusamy].

Page 6 of 12
MSA / L-7 CSE-3512 - Numerical Methods 18th February, 2005

Gauss Elimination Method with Pivoting

In the basic Gauss elimination method, the element aij when i = j is known as a pivot element.
Each row is normalized by dividing the coefficients of that row by its pivot element i.e.
akj = akj / akk where j = 1, 2, ….. , n
If akk = 0, kth row cannot be normalized. Therefore, the procedure fails. One way to
overcome this problem is to interchange this row with another row below it which does not
have a zero element in that position. If this is impossible, then the matrix is singular and the
equations have no solution. [In particular, the number of nonzero diagonal elements in (4)
will represent the rank of original matrix.]
From the given set of equations, it is possible to reorder the equations such that a11 is not
zero. But subsequently, the values of akk are continuously modified during the elimination
process and, therefore, it is not possible to predict their values beforehand.
The reordering of the rows is done such that akk of the row to be normalized is not zero. There
may be more than one non-zero values in the kth column below the element akk. It is
suggested that the row with zero pivot element should be interchanged with the row having
the largest (absolute value) coefficient in that position. [It can be proved that roundoff errors
would be reduced if the absolute value of the pivot element is large.] In general: the
reordering of equations is done to improve accuracy, even if the pivot element is not zero.
The procedure of reordering involves the following steps:
1. Search and locate the largest absolute value among the coefficients in the first column.
2. Exchange the first row with the row containing that element.
3. Then eliminate the first variable in the other equations as explained earlier.
4. When the second row becomes the pivot row, search for the coefficients in the second
column from the second row to the nth row and locate the largest coefficient. Exchange
the second row with the row containing the large coefficient.
5. Continue this procedure till (n-1) unknowns are eliminated.
This process is referred to as partial pivoting. There is an alternative scheme known as
complete pivoting in which, at each stage, the largest element at any of the remaining rows is
used as the pivot.
Example 7.3: Page 220 [Balagurusamy].
Algorithm 7.2: Gauss Elimination Method with Partial Pivoting
See Page-220 [Balagurusamy].
Gauss-Jordan Method
Gauss-Jordan method is another popular method used for solving a system of linear
equations. Like Gauss elimination method, Gauss-Jordan method also uses the process of
elimination of variables, but there is a major difference between them. In Gauss elimination
method, a variable is eliminated from the rows below the pivot equation. But in Gauss-Jordan
method, it is eliminated from all other rows (both below and above). This process thus
eliminates all the off-diagonal terms producing a diagonal matrix rather than a triangular
matrix. Further, all rows are normalized by dividing them by their pivot elements. [This is
illustrated in Fig 7.4 (page 228, Balagurusamy).] Consequently, we can obtain the values of
unknowns directly from the b vector, without employing back substitution.
Example 7.4: Page 228 [Balagurusamy].
Algorithm 7.3: Gauss-Jordan Elimination Method
See Page-228 [Balagurusamy].
 The Gauss-Jordan method requires approximately 50 percent more arithmetic operations
compared to Gauss method. Therefore, this method is rarely used.
 See Table 7.2 [page 230, Balagurusamy] for the comparison of computational effort.

Page 7 of 12
MSA / L-7 CSE-3512 - Numerical Methods 18th February, 2005

Gauss-Jordan Matrix Inversion Method

See page 513 [E. Balagurusamy]
Example A.3: Page 514 [E. Balagurusamy]

Page 8 of 12
MSA / L-7 CSE-3512 - Numerical Methods 18th February, 2005

Iterative Solution of Linear Equations

Direct methods have some problems when the systems grow larger or when most of the
coefficients are zero. They require prohibitively large number of floating point operations
and, therefore, not only become time consuming but also severely affect the accuracy of the
solution due to roundoff error. In such cases, iterative methods provide an alternative.
Jacobi’s Method
Let us consider a system of n equations in n unknowns.
a11 x1 + a12 x2 + a13 x3 + …… + a1n xn = b1
a21 x1 + a22 x2 + a23 x3 + …… + a2n xn = b2
… … … … …. …. … … … … … … …
a31 x1 + a32 x2 + a33x3 + …… + ann xn = bn
We rewrite the original system as
1
x1 = [b1 - a12 x2 - a13 x3 - …… - a1n xn]
a11
1
x2 = [b2 - a21 x1 – a23 x3 - …… - a2n xn]
a 22
…. …. …. …. …. …. …. …. .. … … … …
1
xn = [b3 – a31 x1 – a32 x2 –…… - ann-1 xn]
ann
Now, we cam compute x1, x2, …., xn by using initial guesses for these values. These new
values are again used to compute the next set of x values. The process can continue till we
obtain a desired level of accuracy in the x values.
In general, an iteration for xi can be obtained from the ith equation as follows
1
xi (k+1) = [bi –ai1 x1(k) - …. – aii-1 xi-1(k) – aii+1 xi+1(k) - … - ain xn(k)]
aii
 Condition for Convergence
For each row, the absolute value of the diagonal element should be greater than the sum
of absolute values of the other elements in the equation. i.e.
n
| aii | > 
i 1
| aij | for i  j

Remember that this condition is sufficient, but not necessary, for convergence. Some
systems may converge even if this condition is not satisfied.
Systems that satisfy the above condition are called diagonally dominant systems.
Convergence of such systems are guaranteed.
Example: Solve the following system of linear equations by Jacobi’s method
5x + 2y + z = 12
x + 4y + 2z = 15
x + 2y + 5z = 20
Solution: The above system is diagonally dominant i.e. in each equation the absolute value of
the largest coefficient is greater than the sum of the absolute values of the other coefficients.
The given equations can be written as
x = 1/5 [12 – 2y –z ]
y = 1/4 [15 – x – 2z]
z = 1/5 [20 – x – 2y]
We start the iteration by putting x = 0, y = 0, z = 0

Page 9 of 12
MSA / L-7 CSE-3512 - Numerical Methods 18th February, 2005

 For the first iteration we get

x (1) = 1/5 [12 – 0 – 0] = 2.40
y (1) = 1/4 [15 – 0 – 0] = 3.75
z (1) = 1/5 [20 – 0 – 0] = 4.00
Substituting the values x (1), y (1) and z (1) we get,
x (2) = 1/5 [12 – 2(3.75) – 4.00 ] = 0.10
y (2) = 1/4 [15 – 2.40 – 2(4.00)] = 1.15
z (2) = 1/5 [20 – 2.40 – 2(3.75)] = 2.02
The iteration process is continued and results are tabulated as follows
Iteration 1 2 3 4 5 6 7 8
x 2.40 0.10 1.54 0.61 1.41 0.80 1.08 1.084
y 3.75 1.15 1.72 1.17 2.29 1.69 1.95 1.95
z 4.00 2.02 3.57 2.60 3.41 3.20 3.16 3.164
The values of x, y, z at the end of the 8th iteration are
x = 1.084, y = 1.95 and z = 3.164
Algorithm 8.1: Jacobi Iteration Method
Gauss-Seidel method
Gauss-Seidel method is an improved version of Jacobi iteration method. In Jacobi method,
we begin with the initial values x1(0), x2(0), …., xn(0) and obtain next approximation x1(1), x2(1),
…., xn(1) . Note that, in computing x2(1), we used x1(0) and not x1(1) which has just been
computed. Since, at that point, both x1(0) and x1(1) are available, we can use x1(1) which is a
better approximation. For computing x2(1). Similarly, for computing x3(1), we can use x1(1) and
x2(1) along with x4(0), …, xn(0). This idea can be extended to all subsequent computations. This
approach is called the Gauss-Seidel method.
The Gauss-Seidel method uses the most recent values of x as soon as they becomes available
at any point of iteration process. During the (k+10 th iteration of Gauss-Seidel method, xi
takes the form
1
xi (k+1) = [bi –ai1 x1(k+1) - …. – aii-1 xi-1(k+1) – aii+1 xi+1(k) - … - ain xn(k)]
aii
 Condition for Convergence
Same as Jacobi’s method. But the convergence in Gauss-Seidel method is more rapid than
in Jacobi’s method.
Example 8.2: Page 261 [Balagurusamy].
Algorithm 8.2: Gauss-Seidel Method
See Page-261 [Balagurusamy].

Page 10 of 12
MSA / L-7 CSE-3512 - Numerical Methods 18th February, 2005

The Eigen value Problems

Let A be a square matrix of order n with elements aij. We wish to find a column vector X and
a constant λ such that AX = X ..................... (1)
In the equation (1),  is called the eigenvalue and X is called the corresponding eigenvector.
The matrix equation (1), when written out in full, represents a set of homogeneous linear
equations
(a11 - ) x1 + a12 x2 + a13x 3 + ... ... a1n xn = 0
a21 x1 + (a22 - ) x2 + a23 x3 + ... ... a2n xn = 0
... ... ... ... ... ... ... ... ... ... ... ............................ (2)
... ... ... ... ... ... ... ... ... ... ...
an1 x1 + an2 x2 + an3 x3 + .… + (ann - ) xn = 0
A nontrivial solution exists only when the coefficient determinant in (2) vanishes. Hence, we
have
a11 -  a12 a13 ... ... a1n
a21 a22 -  a23 ... .. a2n
... ... ... ... ... ... ... .… = 0 ............................ (3)
... ... ... ... ... ... ... .…
an1 an2 an3 ... ... ann - 

This equation, called the characteristic equation of the matrix A, is a polynomial equation of
degree n in , the polynomial being called the characteristic polynomial of A. If The roots of
(3) be given by i (i = 1, 2, ... , n), then for each value of i, there exists a corresponding Xi
such that A Xi = iXi .............................. (4)
The eigenvalues i may be either distinct (i.e. all different) or repeated. The evaluation of
eigenvectors in the case of the repeated roots is a much involved process and will not be
discussed here. The set of all eigenvalues, i, of a matrix A is called the spectrum of A and
the largest of |i | is called the spectral radius of A.

Example: Find the eigenvalues and eigenvectors of the matrix

5 0 1
A = 0 -2 0
1 0 5
Solution: The characteristic equation of this matrix is given by
5- 0 1
A= 0 -2 -  0 =0
1 0 5-
which gives 1 = - 2, 2 = 4 and 3 = 6. The corresponding eigenvectors are obtained thus
a. 1 = - 2
Let the eigenvector be x1
X1 = x2
x3
Then we have,
x1 x1
A x2 = -2 x2
x3 x3

Page 11 of 12
MSA / L-7 CSE-3512 - Numerical Methods 18th February, 2005

which gives the equations

7x1 + x3 = 0
x1 + 7x3 = 0
The solution is x1 = x3 = 0 with x2 arbitrary. In particular, we take x2 = 1 and the eigenvector is
0
X1 = 1
0
b. 2 = 4
With x1 as the eigenvector, we get the equations are
X2 = x2
x3
x1 + x3 = 0
6x2 = 0
from which we obtain x1 = - x3 and x2 = 0
We choose, in particular, x1 = 1/2 and x3 = - 1/2 so that x12 + x22 + x32 = 1. The eigenvector
chosen in this way is said to be normalized. We have therefore
1/2
X2 = 0
- 1/2
c. 3 = 6
With x1 as the eigenvector, we get the equations are
X3 = x2
x3

- x 1 + x3 = 0
- 8x2 = 0
x1 - x3 = 0
which gives x1 = - x3 and x2 = 0
Choosing, x1 = x3 = 1/2, the normalized eigenvector is given by
1/2
X3 = 0
1/2
 The above method for solving eigenvalues problem is unstable for matrices of higher
order and better methods must be applied. For symmetric matrices, in particular, several
methods are available and a recent method, known as Householder’s method, is used.

Page 12 of 12