Chapter-7

Non-linear Programming Problem (NLP)

7.1 Introduction: To solve an economic, financial or business problem we formulate the

problem mathematically and find the optimum value of a given function which called objective
function under some restrictions termed constraint functions. If the objective function or some
constraint functions or both are non-linear functions of the decision variables then the problem
are called non-linear programming problem. The mathematical technique for determining the
optimum solutions of non-linear programming problem is called non-linear programming.
Knowledge of differential calculus is assumed for better understanding the non-linear
programming problem.

7.2 General non-linear programming problem:

General form of non-linear programming problem is

Maximize or minimize z=f ( x 1 , x 2 , … … … , x n)

1
Subjects to g ( x1 , x 2 , … … … , x n ) ≤ ,=,≥ b 1

g2 ( x1 , x 2 , … … … , x n ) ≤ ,=,≥ b 2

………………………… …………

gm ( x 1 , x 2 ,… … … , x n ) ≤,=, ≥ bm

where z and gi’s are real valued functions of n variables and x j ≥ 0 , j=1 ,2 , … . , n and either

f (x 1 , x 2 , … … … , x n) ore some gi ( x 1 , x 2 , … … … , x n ) ,i=1 , 2 ,… . , m or both are non-linear. The

problem is to determining the n-tuple ( x 1 , x 2 ,… … … , x n ) which makes z a maximum or

minimum and satisfies all the constraints.

7.3 Mathematical formulation of non-linear programming problem:

In practical many business, economic and financial problem contains non-linear cost/profit
(objective) function.
Example-1. A company manufactures two products A and B. It takes 30 minutes to process one
unit of product A and 15 minutes for each unit of B and maximum machine time available is 35
hours per week. Product A and B require 2kgs and 3kgs of raw material per unit respectively.
The available quantity of raw materials is 180 kgs per week. Market price of product A is 20
taka and product B is 50 taka per unit. All products may sold in the market. If the manufacturing
cost per unit of product A is the double or square of its quantity is made and that of product B is
the three times of square of its quantity is made. How much of each product should be produced
per week to maximize the profit. Formulate the problem as a non-linear programming problem.

Solution: Formulation:

Step-1. Here key decision is that how many units of product A and B should be produced to
maximize the profit.

Step-2. Let x 1 units of product A and x 2 units of product B should be produced to maximize the
profit.

Step-3. We can never be produced any negative unit of products. So x 1 ≥ 0 and x 2 ≥ 0.

Step-4. Here total cost ¿ 2 x21 +3 x 22

And total sell price ¿ 20 x 1+50 x 2

∴ Total profit¿ 20 x 1+50 x 2−2 x21 −3 x 22

∴ Objective function z=20 x 1+ 50 x 2−2 x 21−3 x 22

which will be maximize.

Step-5. Restriction of total machine time is

30 15
x 1+ x2 ≤35 ⇒ 2 x 1+ x2 ≤140
60 60

and restriction of raw material is

2 x1 +3 x 2 ≤ 180
Step-6. The non-linear programming problem of the given problem is

Maximize z=20 x 1+ 50 x 2−2 x 21−3 x 22

Subjects to 2 x1 + x 2 ≤ 140

2 x1 +3 x 2 ≤ 180

and x 1 ≥ 0 , x2 ≥ 0

Example-2. A company decide to allocate amounts x 1 and x 2 units on the two media for
advertisement of their products. The relationship between the sales S and the amounts x 1 and x 2

20 x 1 10 x 2
spent on two advertising media is given by S= + . The net profit is one fifth of the
4+ x 1 8+ x 1
sales minus the advertising cost. The budget for advertising has a maximum limitation of 25
units. The company desires to know how it should allocate the available amount between the two
media so as to maximize the net profit. Formulate the problem as a non-linear programming
problem.

Solution: Formulation:

Total advertising cost C=x 1+ x 2

20 x 1 10 x 2
and total sales S= + units
4+ x 1 8+ x 1

1 4 x1 2 x2
∴ Profit ¿ S−C= + −x −x
5 4 + x 1 8+ x 1 1 2

4 x1 2 x2
∴ Objective function z= + −x −x
4 + x 1 8+ x 1 1 2

which will be maximize.

Limitation of advertising cost is x 1+ x2 ≤25 and x 1 ≥ 0 and x 2 ≥ 0.

∴ The non-linear programming problem is

 4 x1 2 x2
Maximize z= + −x −x
4 + x 1 8+ x 1 1 2

Subjects to x 1+ x2 ≤25
and x 1 ≥ 0 , x2 ≥ 0

Example-3. A manufacturing company produces two products Radio and TV sets, The sales
price relationships for these products are given below.

Products Quantity Demanded Unit Price

Radio 1500−5 p1 p1
TV 3800−10 p2 p2

The total cost functions for these two products are given by 200 x 1+ 0.1 x 21 and 300 x 2+ 0.1 x 22
respectively. The production takes place on two assembly lines. Radio sets are assembled on
assembly line I and TV sets are assembled on assembly line II. Because of the limitations of the
assembly line capacities, the daily production is limited to no more than 80 Radio sets and 60 TV
sets. The production of both types of products requires electronic components. The production of
each of these sets requires five units and six units of electric equipment respectively. The
electronic components are supplied by another manufacturer and the supply is limited to 600
units per day. The company has 160 man day. The production of one unit of Radio sets requires
1 man day labor where as 2 man days of labor are required for a TV set. How many units of
Radio and TV sets should the company produce in order to maximize the total profit. Formulate
the problem as a non-linear programming problem.

Solution: Formulation:

Let x 1 and x 2 units of Radio and TV sets to be produced to maximize the profit and all products
is sold in the market.

∴ x1=1500−5 p1 ; x 2=3800−10 p 2

p1=300−0.2 x1 ; p 2=380−0.1 x 2

Let c 1 and c 2 are the total cost for Radio and TV sets then
c 1=200 x 1+ 0.1 x 21

c 2=300 x 2+ 0.1 x 22

Total sells price for x 1 units of Radio and x 2 units of TV set.

R=p 1 x 1+ p 2 x 2

⇒ R=( 300−0.2 x 1 ) x1 +(380−0.1 x 2) x2

∴ R=300 x 1+380 x 2−0.2 x 21−0.1 x22

∴ Total profit z=R−c 1−c 2

⇒ z=300 x 1+ 380 x 2−0.2 x 21−0.1 x 22−200 x1 −0.1 x 21−300 x 2−0.1 x 22

⇒ z=100 x 1+ 80 x2 −0.3 x 21−0.2 x 22

which is the objective function and is ti be maximize.

Restriction for product of Radio is x 1 ≤ 80

Restriction for product of TV is x 2 ≤ 60

Restriction for electronic parts 5 x 1+6 x 2 ≤600

Restriction for electronic labor x 1+ 2 x 2 ≤ 160

Since negative number of products can be produced so x 1 ≥ 0 , x2 ≥ 0.

Hence the non-linear programming problem to the given problem is

Maximize z=100 x 1+ 80 x 2−0.3 x 21−0.2 x 22

Subjects to 5 x 1+6 x 2 ≤600

x 1+ 2 x 2 ≤ 160

x 1 ≤ 80
x 2 ≤ 60

and x 1 ≥ 0 , x2 ≥ 0.

7.4 Graphical solution: The non-linear programming problem involving only two variables can
be solved by graphical solution. Method based on a set of logical steps. Graphically the optimum
solution of a linear programming problem was found at one of the vertex of the solution region
generated by the constraints of the problem.

But in case of non-linear programming problem, the optimum solution may or may not be found
at one of the vertex of the solution region generated by the constraints and objective function of
the given problem. i.e, optimum solution of the non-linear programming problem c an exist at
the boundary or in any interior point of the solution region. We will illustrate this in given
examples.

7.5 Algorithm of graphical solution: The graphical method for solving a non-linear
programming problem involves the following steps:

Step-1. Set the each inequality constraint to an equality and draw the graph of constraints and
objective function. Let any numerical value of the objective function for which the graph of
objective function touches the solution region.

Step-2. Identify solution region generated by the constraints and objective function which satisfy
all the inequality of constraints.

Step-3. Determine the point in the solution region at which the objective function has optimum
value.

Step-4. Interpret the optimum solution, so obtained.

Example-4. Solve graphically the following non-linear programming problem:

Maximize z=2 x 1+3 x 2

Subjects to x 21+ x22 ≤20

x1 x2 ≤ 8

and x 1 ≥ 0 , x2 ≥ 0

Solution:

Step-1. Set the inequality of the constraint to equality:

x 21+ x22 ≤20 ⇒ x 21 + x 22=20(1)

x 1 x 2 ≤ 8 ⇒ x 1 x 2=8 (2)

Let x 1 along x-axis and x 2 along y-axis and draw the graph of ( 1 ) ,(2) and objective function.
Since x 1 ≥ 0 , x2 ≥ 0 graph lies in 1st quadrant.

x2

x 1 x 2=8

D C

x 21+ x22 =20

B 2 x1 +3 x 2=z

O A x1

Step-2. Here OABCDO is the solution region which satisfies all the inequality of constraints.

Setp-3. Now x 1 x 2=8 represents a rectangular hyperbola and x 21+ x22 =20 represents a circle of

radius √ 20 with origin as its center. Thus, since x 1 x 2 ≤ 8 and x 21+ x22 ≤20, the desired point may be
some where in the solution region OABCDO for which objective function has maximum value.
Since, our search is for such a pair ( x 1 , x 2) which gives a maximum value of 2 x1 +3 x 2 and lies in
the solution region. The desired point can be obtained by moving parallel to 2 x1 +3 x 2=k touches
the extreme boundary point of the feasible region.

8
( 2 ) ⇒ x 2=
x1

8 2
(1)⇒ x 21 + ( )
x1
=20

⇒ x 41 −20 x21 +64=0

⇒ ( x21 −4 )( x 21−16 ) =0

⇒ x 1=2 or x 1=4

∴ x1=2⇒ x 2=4 and x 1=4 ⇒ x 2=2

∴ Point of intersection of (1) and (2) is B ( 4 , 2 ) ,C (2 , 4)

∴ Extreme point of the feasible region is O , A , B ,C , D

Coordinates of the point A is ( 0 , √ 20 )

Coordinates of the point O is ( 0 , 0 )

Coordinates of the point B is ( 4 , 2 )

Coordinates of the point C is ( 2 , 4 )

Coordinates of the point D is ( √ 20 , 0 )

Value of z at O is z=0

Value of z at A is z=3 √ 20

Value of z at B is z=14

Value of z at C is z=16

Value of z at D is z=2 √ 20
Maximum value of z occurs at C=( 2, 4 ) is 16

Step-4. Hence the optimum solution attains at x 1=2 , x 2=4 and z opt =16

Example-5. Solve graphically the following non-linear programming problem:

Minimize z=x 21 + x 22

Subjects to x 1+ x2 ≥ 4

2 x1 + x 2 ≥ 5

and x 1 ≥ 0 , x2 ≥ 0

Solution:

Step-1. Set the inequality of the constraint to equality:

x 1+ x2 ≥ 4 ⇒ x 1 + x 2=4 (1)

2 x1 + x 2 ≥ 5 ⇒ 2 x1 + x 2=5( 2)

and x 21+ x22 =r 2=z (3)

Let x 1 along x-axis and x 2 along y-axis, since x 1 ≥ 0 , x2 ≥ 0 graph lies in 1st quadrant. Draw the
graph of ( 1 ) ,(2) and (3) obtaining the value of r for which the graph of (3) touches the convex
region formed by the constraints.

x2

C 2 x1 + x 2=5

B x 21+ x22 =r 2=z

x 1+ x2=4

O A x1
Step-2. Here unbounded convex region ABC is the solution region which satisfies all the
inequalities of constraints.

Setp-3. Since the points for which x 1+ x2 ≥ 4 and 2 x1 + x 2 ≥ 5. z=x 21 + x 22 has the minimum values
so the desired point will be that point of the convex region at which any side of the convex
region ABC will touches, i.e, will be tangent of the circle ( 3 ) .

Differential equation (3)

⇒ 2 x 1 d x1 +2 x 2 d x 2=0

d x 2 −x 1
⇒ = ( 4)
d x1 x2

Slope of the constraint (1)

⇒ d x 1+ d x 2=0

d x2
⇒ =−1(5)
d x1

Slope of the constraint (2)

⇒ 2 d x1 +d x 2=0

d x2
⇒ =−2(6)
d x1

From ( 4) and (5)

⇒ x 1=x 2 (7)

From (1) and (7)

⇒ x 1=2 ⇒ x 2=2

∴ x1 + x 2=4 is a tangent of (3) at (2 , 2)

¿ P(let) at (2 , 2) is z=22 +22=8

From ( 4) and (6)

⇒ x 1=2 x 2( 8)

From (8) and (2)

⇒ x 1=2 ⇒ x 2=1

∴ 2 x 1 + x 2=5 is a tangent of (3) at (2 , 1)

¿ Q(let) at (2 , 1) is z=22 +12=5

But (2 , 1) is out of the convex region ABC so it is discarded.

∴ Minimum value of z=8 is found at (2 , 2) in the convex region ABC .

Step-4. Hence the optimum solution is Min.z=8 and x 1=2 , x 2=2

Example-6. Solve graphically the following non-linear programming problem:

Minimize z=x 21 + x 22

Subjects to x 1+ x2 ≥ 8

x 1+ 2 x 2 ≥ 10

2 x1 + x 2 ≥ 10

and x 1 ≥ 0 , x2 ≥ 0

Solution:

Set the inequality of the constraint to equality:

x 1+ x2 ≥ 8⇒ x1 + x 2=8( 1)

x 1+ 2 x 2 ≥ 10 ⇒ x 1+ 2 x 2=10(2)

2 x1 + x 2 ≥ 10 ⇒2 x1 + x 2=10(3)
and x 21+ x22 =r 2 (4)

where r is constant.

Let x 1 along x-axis and x 2 along y-axis and value of r is any constant for which ( 4) touches the
convex region. Since x 1 ≥ 0 , x2 ≥ 0 graph lies in 1st quadrant. Draw the graph of ( 1 ) , ( 2 ) ,(3) and ( 4)
we get

x2

D 2 x1 + x 2=10

C x 21+ x22 =r 2=z

B x 1+ x2=8 x 1+ 2 x 2=10

O A x1

From the graph the unbounded region ABCD is the solution region and touches the circle ( 4) at
such a point ( 4) has a tangent any one of ( 1 ) ,(2) or (3) gradient of any tangent of ( 4) is given by

⇒ 2 x 1 d x1 +2 x 2 d x 2=0

d x 2 −x 1
⇒ = (5)
d x1 x2

Slope of the constraint (1)

⇒ d x 1+ d x 2=0

d x2
⇒ =−1(6)
d x1
Slope of the constraint (2)

⇒ d x 1+ 2d x 2=0

d x 2 −1
⇒ = (7)
d x1 2

Slope of the constraint (3)

⇒ 2 d x1 +d x 2=0

d x2
⇒ =−2( 8)
d x1

From (5) and (6) ⇒ x 1=x 2

∴ ( 1 ) ⇒ x 1 + x 2=8 ⇒ x 1=4 , x 2=4

∴(1) is a tangent of ( 4) at ( 4 , 4) i.e, at the point ( 4 , 4) is z=32

From (5) and (7) ⇒ x 1=2 x 2

∴ ( 2 ) ⇒ x 1 +2 x 2=10 ⇒ x 2=4 , x 1=2

∴(2) is a tangent of ( 4) at (2 , 4) but it is out of the region and hence discarded.

From (5) and (8) ⇒ x 2=2 x 1

∴ ( 3 ) ⇒ 2 x 1+ x2 =10⇒ x2 =2, x 1=4

∴(3) is a tangent of ( 4) at ( 4 , 2) but it is out of the region and hence discarded.

∴ Minimum value of z=8 is foun( 4 , 4) in the convex region ABCD .

Hence the optimum solution is Min.z=32 and x 1=4 , x 2=4

Example-7. Solve graphically the following non-linear programming problem:

Maximize z=8 x 1−x 21+ 8 x2 −x22

Subjects to x 1+ x2 ≤12

x 1−x 2 ≥ 4

and x 1 ≥ 0 , x2 ≥ 0

Solution:

Set the inequality of the constraint to equality:

x 1+ x2 ≤12 ⇒ x 1+ x2 =12(1)

x 1−x 2 ≥ 4 ⇒ x1 −x2 =4 (2)

2 2 2 2
and z=8 x 1−x 1+ 8 x2 −x2 ⇒ z=32−( x1 −4 ) −( x 2−4 )

∴ ( x1 −4 )2 + ( x 2−4 )2=32+ k (3)

where k is constant.

Let x 1 along x-axis and x 2 along y-axis and draw the graph of ( 1 )and (2) also draw the graph of
(3) for any value of k such that (3) touches the convex region formed by the constraints. Since
x 1 ≥ 0 , x2 ≥ 0 graph lies in 1st quadrant. Here ABC is the feasible solution region.

x2

x 1+ x2=12

2 2
B ( x 1−4 ) + ( x 2−4 ) =32+k

x 1−x 2=4

O A C x1

Thus, the optimal point ( x 1 , x 2) must be some where in the convex region ABC. However, the
desired point would be that at which a side of the convex region is tangent to the circle (3). The
gradient of tangent of (3) is given by differential equation (3)
8 d x1 −2 x 1 d x 1 +8 x 2−2 x 2 d x 2=0

d x 2 2 x 1−8
⇒ = ( 4)
d x 1 8−2 x 2

Slope of the constraint (1)

⇒ d x 1+ d x 2=0

d x2
⇒ =−1(5)
d x1

Slope of the constraint (2)

⇒ d x 1−d x2 =0

d x2
⇒ =1(6)
d x1

From ( 4) and (5) ⇒ x 1=x 2

∴ ( 1 ) ⇒ x 1 + x 2=12⇒ x 1=6 , x 2=6

∴(1) is a tangent of (3) at (6 , 6) but it is out of the region and hence discarded.

From ( 4) and ( 6 ) ⇒ x 1=8−x 2

∴ ( 2 ) ⇒ x 1−x 2=4 ⇒ x 2=2 , x1=6

∴(2) is a tangent of (3) at (6 , 2) i.e, at the point (6 , 2) is z=24.

∴ Maximum value of z=24 at (6 , 2) in the convex region.

Hence the optimum solution is Max.z=24 and x 1=6 , x 2=2

Example-8. Solve graphically the following non-linear programming problem:

2 2
Minimize z=( x 1−4 ) + ( x 2−4 )

Subjects to 2 x1 +3 x 2 ≥ 6
3 x 1+2 x 2 ≤ 12

and x 1 ≥ 0 , x2 ≥ 0

Solution:

Set the inequality of the constraint to equality:

2 x1 +3 x 2 ≥ 6 ⇒ 2 x1 +3 x 2=6(1)

3 x 1+2 x 2 ≤ 12⇒ 3 x 1 +2 x2=12( 2)

2 2
and( x 1−4 ) + ( x 2−4 ) =k (3)

where k is constant.

Let x 1 along x-axis and x 2 along y-axis and draw the graph of ( 1 )and (2) also draw the graph of
(3) for any value of k such that (3) touches the convex region formed by the constraints.

x2 3 x 1+2 x 2=12

2 2
C ( x 1−4 ) + ( x 2−4 ) =k

D 2 x1 +3 x 2=6

O A B x1

Since x 1 ≥ 0 , x2 ≥ 0 graph lies in 1st quadrant. Here the solution region is the convex region ABCD
.

Thus, the optimal point ( x 1 , x 2) must be some where in the convex region ABCD. However, the
desired point would be that at which a side of the convex region is tangent to the circle (3).The
value of z=0 at the centre of the circle but it is out of the solution region. The gradient of tangent
of (3) is given by differential equation (3)

( x 1−4 ) d x1 + ( x 1−4 ) d x 2=0

d x 2 x1 −4
⇒ = (4)
d x 1 4−x 2

Slope of the constraint (1)

⇒ 2 d x1 +3 d x2 =0

d x 2 −2
⇒ = (5)
d x1 3

Slope of the constraint (2)

⇒ 3 d x 1+2 d x2 =0

d x 2 −3
⇒ = (6)
d x1 2

From ( 4) and (5)

x 1−4 −2 4+ 2 x 2
⇒ = ⇒ x 1=
4−x 2 3 3

∴ ( 1 ) ⇒ 2 x 1+3 x 2=6

8+ 4 x 2
⇒ +3 x 2=6
3

10 24
⇒ x 2= , x 1=
13 13

∴(1) is a tangent of (3) at ( 2413 , 1013 ) i.e, at the point ( 2413 , 1013 ) is z= 2548
169
.
From ( 4 ) and ( 6 )

x 1−4 −3 3 x 2−4
⇒ = ⇒ x 1=
4−x 2 2 2

∴ ( 2 ) ⇒ 3 x 1+ 2 x 2=12

9 x 2−12
⇒ +2 x 2=12
2

36 10
⇒ x 2= , x 1=
13 13

∴(2) is a tangent of (3) at ( 1013 , 3613 ) i.e, at the point ( 1013 , 3613 ) is z= 2020
169
.

2020 10 36
∴ Minimum value of z=
169
at ,(
13 13 )
in the convex region.

2020 10 36
Hence the optimum solution is Min. z= and x 1= , x 2=
169 13 13

Example-9. Solve graphically the following non-linear programming problem:

2 2
Minimize z=( x 1−4 ) + ( x 2−4 )

Subjects to x 1+ x2 ≥5

x1 ≤ 6

2 x2 ≤11

and x 1 ≥ 0 , x2 ≥ 0

Solution:
Set the inequality of the constraint to equality:

x 1+ x2 ≥5 ⇒ x 1 + x 2=5(1)

x 1 ≤ 6 ⇒ x 1=6( 2)

11
2 x2 ≤11 ⇒ x2 = (3)
2

2 2
Let ( x 1−4 ) + ( x 2−4 ) =k ( 4)

where k is any constant.

Let x 1 along x-axis and x 2 along y-axis and draw the graph of ( 1 ) ,(2)and (3) also draw the graph
of ( 4) for any value of k. Since x 1 ≥ 0 , x2 ≥ 0 graph lies in 1st quadrant.

D x2 C x 1=6

11
x 2=
2

x 1+ x2=5

O A B x1

Here the feasible solution region is ABCDE. Thus, the optimal point Thus, the optimal point
( x 1 , x 2) must be somewhere in the convex region ABCDE. Since the centre of the circle

z=( x 1−4 ) 2+ ( x 2−4 )2 lies in the region and the values of z=0, minimum at centre ( 4 , 4). Hence

the optimal solution is x 1=x 2=4 and z=0.

Example-10. Solve graphically the following non-linear programming problem:

Maximize z=2 x 1+ x 2

Subjects to −x 21+ 4 x 1−x 2 ≤ 0

2 x1 +3 x 2 ≤ 12

and x 1 ≥ 0 , x2 ≥ 0

Solution: Set the inequality of the constraint to equality:

−x 21+ 4 x 1−x 2 ≤ 0 ⇒−x 21+ 4 x 1−x 2=0 (1)

2 x1 +3 x 2 ≤ 12⇒ 2 x 1 +3 x2=12( 2)

Objective function 2 x1 + x 2=k ( ¿ z ) (3)

Let x 1 along x-axis and x 2 along y-axis and draw the graph of ( 1 ) ,(2) and objective function (3).
Since x 1 ≥ 0 , x2 ≥ 0 graph lies in 1st quadrant.

x2

D E −x 21+ 4 x 1−x 2=0

2 x1 +3 x 2=12

O A B x1

Here ADF is the convex solution region which satisfies all the inequality of constraints. Thus,
the optimal point ( x 1 , x 2) must be somewhere in the convex region ADF. However, the desired
point lie on the straight line parallel to 2 x1 + x 2=k ( ¿ 4 )

Coordinates of the point A is ( 0 , 0 )

Coordinates of the point F is ( 0 , 4 )

Coordinates of the point D is ( 7−3√13 , 22+29 √13 )

Value of z at A is z=0
Value of z at F is z=4

Value of z at D is z=2 ( 7−3√ 13 )+ 22+29√ 13 =5.5086

∴ Maximum value of z occurs at D= ( 7−3√13 , 22+29 √13 ) is 5.5086
7− √ 13 22+2 √ 13
Hence the optimum solution attains at x 1= , x 2= and Max. z=5.5086
3 9

7.6 Some definitions:

(1) Quadratic forms: A quadratic form on Rn is a function Q defined on Rn whose value at a

vector x in Rn can be computed by on expression of the form

Q ( x ) =x T A x

n m
⇒ Q ( x )=∑ ∑ a ij x i x j (1)
i=1 j=i

T
where x=[ x 1 , x 2 ,… … , x n ] and A is a n × n symmetric matrix whose diagonal elements in i-th

row and i-th column is the coefficient of x 2i and the off diagonal elements in i-th row and j-th
column is halt the co-efficient of the product x i x j .

An equation of the form

a x 2+ b y 2+ c z 2 +2uxy +2 vyz+ 2 wzx+ fx+ gy+ hz+ d=0 (2)

where a , b , … … , h are not all zero constant is called a quadratic equation in x , y and z the
expression

a x 2+ b y 2+ c z 2 +2uxy +2 vyz+ 2 wzx

is called the associated quadrate forms. Matrix form of the equation (2) is
 a u v x x

[ ][ ]
( x , y , z ) u b w y +[ f
v w c z []
g h ] y +d=0
z

⇒ x T A x+ c x +d =0

⇒ c x+ xT A x +d =0

x a u v
where c= [ f
[] [
g h ] , x = y and A= u b w
z v w c ]
A quadratic form Q ( x ) =x T A x is said to be

(i) Positive definite:

(a) If Q ( x ) > 0 , ∀ x ≠ 0

or (b) If the values of all principal minor determinant of A are positive.

or (c) If all eigenvalue of A are positive.

(ii) Negative definite:

(a) If Q ( x ) < 0 , ∀ x ≠ 0

or (b) If the value of i-th principal minor determinant of A has the sign (−1 )i where
i=1 , 2, … … . , n. i.e, has alternative sign starting with negative sign.

or (c) If all eigenvalue of A are negative.

(iii) Positive semi-definite:

(a) If Q ( x ) ≥ 0 , ∀ x ≠ 0

or (b) If the values of all principal minor determinant of A are non-negative.

or (c) If all eigenvalue of A are non-negaitive.

(iv) Negative semi-definite:

 (a) If Q ( x ) ≤ 0 , ∀ x ≠ 0

or (b) If the value of i-th principal minor determinant of A is either zero or has the sign of
(−1 )i where i=1 , 2, … … . , n.

or (c) If all eigenvalue of A are non-positive.

(v) Indefinite:

(a) If Q ( x ) has both positive and negative value.

or (b) If the value of i-th principal minor determinant of A has no sign of (−1 )i where

i=1 , 2, … … . , n and non-zero.

or (c) If A has both positive and negative eigenvalues.

Example-11. (i) Q ( x )=2 x 21 +2 x22 +3 x 23+ 2 x 1 x 2 +2 x2 x 3

2 1 0 x1
[ ][ ]
¿ [ x 1 x 2 x 3 ] 1 2 1 x 2 =x T A x
0 1 3 x3

Here values of principal minor determinant of A are

2 1 0
|1 2| [ ]
|2|=2 , 2 1 =3 , 1 2 1 =7
0 1 3

Since values of principal minor determinants are positive. So Q ( x ) is positive definite.

(ii) Q ( x )=3 x21 +7 x 22

¿ [ x1 x2 ] [30 07] [ xx ]=x A x

1

2
T

Here values of principal minor determinant of A are

|3|=3 , 3 0 =21
|0 7|
Since values of principal minor determinants are positive. So Q ( x ) is positive definite.

(iii) Q ( x )=−3 x12−7 x22

¿ [ x1 x2 ] [−30 −70 ] [ xx ]=x A x

1

2
T

Here values of principal minor determinant of A are

0 =21
|−3|=−3 , −3 |0 −7 |
has sign (−1 )i of i-th principal minor determinant. So Q ( x ) is negative definite.

(iv) Q ( x )=3 x21 −7 x22

¿ [ x1 x2 ] [30 −70 ] [ xx ]=x A x

1

2
T

Here values of principal minor determinant of A are

|3|=3 , 3
|0 −70 |=−21
has no sign of the form (−1 )i for i-th principal minor determinant and all are non-zero. So Q ( x ) is
indefinite.

2 2 2
(v) Q ( x )=( x 1−x 2 ) =x 1−2 x 1 x2 + x 2

¿ [ x1 x2 ] [−11 −11 ][ xx ]=x A x

1

2
T

Here values of principal minor determinant of A are

|1|=1 , 1
|−1 −11 |=0
Since the values of all principal minor determinants are non-negative. So Q ( x ) is positive
semidefinite.
(vi) Q ( x )=−x 21+ 2 x 1 x 2−x 22

¿ [ x1 x2 ] [−11 −11 ][ xx ]=x A x

1

2
T

Here values of principal minor determinant of A are

1 =0
|−1|=−1 , −1 |1 −1 |
Since the values of all principal minor determinants are non-positive. So Q ( x ) is negative
semidefinite.

(2) Principal minor: If A be a square matrix then a minor of A whose diagonal elements are
also the diagonal elements of A is called a principal minor of A.

(3) Convex functions: Let S be a non-empty convex subset of Rn. A function f (x) on S is said
to be convex function if for any two vectors x 1 and x 2 in S.

f [ λ x 1 +(1−λ)x 2 ] < λf ( x 1 ) + ( 1−λ ) x 2 ; 0< λ<1

y=f ( x )

o Convex function x

( 4) Strictly convex functions: Let S be a non-empty convex subset of Rn. A function f ( x) on S

is said to be strictly convex function if for any two vectors x 1 and x 2 in S.

f [ λ x 1 +(1−λ)x 2 ] < λf ( x 1 ) + ( 1−λ ) x 2 ; 0< λ<1

every strictly convex function is also convex.

y
 y=f ( x )

o Strictly convex function x

(5) Concave or strictly concave function: A function f ( x) on a non-empty subset S of Rn is

said to be concave if −f (x) is convex and strictly concave if −f (x) is strictly convex.

y=f ( x )

o Strictly concave function x

(6) Both convex and concave function: A function f (x) to be both convex and concave if it is
linear. For example f ( x )=x is such a function.

y=f ( x )

o Both convex and concave function x

(7) Global minima: A global minimum of the function f (x) is said to be attained at x 0 if
f (x 0) ≤ f ( x ) for all x in the feasible region. Which is also called absolute minima.

(8) Local minima: A local minima of the function f ( x) is said to be attained at x 0 if there exist

ε > 0, such that f (x 0) ≤ f ( x ) and |x 0−x|≤ ε , for all x in the feasible region.

y Global maxima

Local maxima

Local minima
 Global minima

o x

( 9 ) Global maxima: A global maxima of the function f ( x) is said to be attained at x 0 if

f (x 0) ≥ f ( x ) for all x in the feasible region. Which is also called absolute maxima.

(10) Local maxima: A local maxima of the function f ( x) is said to be attained at x 0 if there

exist ε > 0, such that f (x 0)≥ f ( x ) and |x 0−x|≤ ε , for all x in the feasible region.