Single Variable Optimization

Discipline Courses-I
Semester-I
Paper II: Mathematical Methods for Economics: Preliminaries-I
Unit-III
Lesson: Single Variable Optimization
Lesson Developer: Himanshu Singh
College/Department: Satyawati College, University of Delhi

Institute of Lifelong Learning,

1 University of Delhi
 Single Variable Optimization

Content:

1. Learning Outcome
2. Introduction of Geometric Properties of Functions

2.1 Linearity

2.2 Differentiability

3. Optimization

3.1 First Derivative Test for Maxima and Minima

4. Geometric Interpretation of Rolle's Theorem and the Mean Value Theorem.

4.1 The mean value Theorem for Derivatives

5. Existence, Global and Uniqueness of solution.

5.1 Weierstass Theorem (Existence)

5.2 Global Maximum

5.3 Uniqueness of Solution
5.4 Related examples and their solutions

6. Complete criteria for Maximum, Minimum and inflectional values

7. End Chapter Exercise

8. References

1. Learning Outcome
After reading this lesson you should be able to learn about the geometric properties of
functions (linearity and differentiability); optimization of function (maxima and
minima); geometric interpretation of Rolle’s theorem and mean value theorem;
existence, global and uniqueness of solution.

Institute of Lifelong Learning,

2 University of Delhi
 Single Variable Optimization

2. Introduction of Geometric properties of functions

If an economic variable lives in one set, and charges in that variable help to explain
changes in another economic variable, two variables are related. There is a
correspondence between the two sets of variables. If the first set is denoted by S1 and
the second set by S2, the correspondence (which is defined by f) is written f : S 1 
S2 to biggest that f associates elements in S1 with elements in S2. Here f "sends" or
"Transforms" or "maps" x in S1 into y and z in S2 (as shown in the following
diagram).

If each element in S1 gets sent to exactly one element of S2 (see fig. 3.1.b), then f is
called a function. Notice that more than one element in S1 may go tvo a single
element of S2.

The main problem is to find ways of combining functions and describing their
properties. Thus a diagram like that in fig. 3.2.a does not represent a continuous
function. Formally, continuity requires that for any distance  > 0 around f(x), there is

Institute of Lifelong Learning,

3 University of Delhi
 Single Variable Optimization

a distance  > 1 around x, so that points within distance  of x get sent to points
within distance  of f(x) as in fig. 3.2.b.

Geometrically speaking, a continuous function preserves some of the structure of the

domain set as this domain set gets transformed into the range. The structure that is
preserved is the "togetherness" of points. If, of course, a function preserves the
"nearbyness" of points, it is reasonable that property of "being a closed set" is
preserved by a continuous function, because a closed set contains its limit points, and
limit points are characterised by the "nearness" of an infinite number of points of a
sequence. If the property of "closedness" of set is preserved by a continuous function,
all structures such as open sets, or compact sets that are define by likewise preserved
under action of continuous function. In brief, a continuous function from R to R
preserves the topological structure of the set it transforms.

Simply stated, a function transforms a set of real numbers into another set of real
numbers. But a set of real numbers may have a certain coherence or structure. This
set, for example, may be open, or compact, or convex, etc. Because that set is
transformed by a rule or function representing and explanation or theory, the structure
of set S1 ought to have an analogue in the structure of the set S2.

Thus the important question to ask is whether or not a function preserves the relevant
structure of the set being transformed. Functions are appropriately classified by the
kinds of structures they preserve under the transformations they represent. For
example, continuous functions transform "nearby" points into "nearby" points. A
function is continuous at a point in the domain, called x, if points close to x get sent to
points close to the image or transformation of x, called f(x).

Institute of Lifelong Learning,

4 University of Delhi
 Single Variable Optimization

2.1 Linearity

A function f : R  R is linear if for a1b1    , v

(i) f(a+b) = f (a) + f(b) and

( a)  f (a) First condition con notes that linear function preserves the additive
structure of the real numbers while second condition implies that a stretching of an
arrow from 0 to a by a factor  is preserved under the action of a linear function. All
linear functions from R to R are continuous. In fact, linear functions were created to
preserve arithmetic properties, while continuous functions were created to preserves
topological (or geometric closedness) properties of the real numbers.

2.2 Differentiability: The differential calculus, when all is said and done, is the
study of linear approximation to nonlinear functions. If any nonlinear function has an
associated linear function that approximates it closely, then analysis of non-linear
functions is rather easy. However, it is not possible to associate a linear function with
an arbitrary function from R to R. It is sometimes possible, though, to carve up the
domain of the function to perform a local approximation. Gaining the simplicity of
linearity requires forsaking the global picture. If it is feasible to do a local
approximation in all the chunks of the carved up domain, it may be possible to patch
together the local approximations into some coherent picture. Differential calculus is
after all a local analysis.

Institute of Lifelong Learning,

5 University of Delhi
 Single Variable Optimization

In general given an arbitrary function f : R  R, suppose there exist a number a such

f ( x  h) – f ( x ) – ah
that lim
h o  0 . This is read as "a is the derivative of f at x ". If a
h
function is differentiable (i.e. possesses a derivative) at some point in its domain, then
it is continuous at that point.

Linearity, continuity and differentiability are useful characteristics of functions.

However, an arbitrary function does not have to either linear, continuous or
differentiable. Most economic models will require that associated functions to be
differentiable functions. Their "almost" linear form, or locally linear structure, permits
visualisation almost like linear functions, which are continuous. Discontinuity suggest
that small changes in the variables x1....., xn induce large changes in the variable y.
The use of continuous functions to model economic relationships is an expression of
faith by economists

Before elaborating on convex functions and concave functions. We will define

convex set. A convex set in R is a set of points in R such that a line segment drawn
between any two of its points lies wholly within the set. That is, S is convex if x, y
 S C R, and  x  (1 –  ) y  z for 0    1 implies zs. If the function is defined
over some convex set, then the function has some useful properties. In the previous
paragraph we defined linear function f: R  R by two conditions additivity (f (x + y)
= f(x) + f(y), x, y R) and homogeneity ( f (x) , R, xR). These properties taken
individually, define classes of functions that include linear functions. What functions
satisfy additivity alone or a generalisation of it? What functions satisfy homogeneity
property alone or its generalisation?

The "Concavity" generalises the additivity property of linear functions. Intuitively, a

strictly concave function has the property that a line joining two points on the graph
lies below the graph between those two points, as can be seen from the diagram given
below.

Institute of Lifelong Learning,

6 University of Delhi
 Single Variable Optimization

Now, we shall examine special class of functions that have significant economic
applications.

Now, we will examine concavity and convexity of a function more formally. Suppose
f : SCR  R, So that f is defined on open set S1 and suppose that S is a convex set; f
is a concave function if, given any x and x̂ is S, and for all
0    1, (i) f (x  (1 –  ) xˆ )   f ( x)  (1 –  ) f ( xˆ )

Because  x  (1 –  ) xˆ = zs, , inequality (i) means that the value of  or the value
of f at some print between x and x̂ , is greater than or equal to the value of f
represented by a point on the line connecting f (x) and f ( x̂ ).

A function f : R  R, defined or an open convex set, is 'strictly' concave if the

inequality (i) is a strict inequality i.e.

f ( x  (1 –  ) xˆ )   f ( x)  (1 –  ) f ( xˆ ) )

Convex function:

A function f: R  R, defined on an open convex set SCR, is convex if, for any x
and

Institute of Lifelong Learning,

7 University of Delhi
 Single Variable Optimization

x̂  s, and any   [0,1],

[2] f (  x  (1 –  ) xˆ )   f ( x)  (1 –  ) f ( xˆ ).

If the inequality sign in (2) is a strict inequality, f(x) is strictly convex.

A convex function has the property that a line drawn between two points on the graph
lies on or above the graph between those two points. Thus concave functions look like
parabolas opening downward, and convex functions look like parabolas opening
upward. Linear functions are certainly both concave and convex, but neither strictly
concave nor strictly convex.

Test for concavity and convexity of a function : f" test. Intuitively, if the graph of f
lies above all of its tangents or an interval I, then it is called concave upward or
convex function on I. If the graph of f lies below all of its tangents or I, it is called
concave downward or simply concave function.

Let us now see how the second derivative of the given function helps determine the
intervals of concavity. For a concave function or some interval I, slope of the tangent
line falls continuously over I and for a convex function slope of the tangent line
drawn rises.

This mean that for a concave function, the derivative f1 is decreasing, and therefore f"
is negative. For a convex function, the derivative f1 is increasing throughout the

Institute of Lifelong Learning,

8 University of Delhi
 Single Variable Optimization

interval I, and hence f11 is positive. This reasoning can be reversed and suggests that
the following theorem is true [Proof can be provided with the help of the Mean Value
Theorem].

Test :

If f11 >0 for all x in I, the graph of f is convex (concave upward) on I.

If f11(x)<0 for all x in I, then the graph of f is concave downward on I.

Definition:

A point P on a curve f = f (x) is called an inflection point if f is continuous there and

the curve change from convex to concave or concave to convex at P.

In view of the Concavity Test, there is a point of inflection at any point where the
second derivative changes sign.

Example : Sketch a possible graph of a function f that satisfies the following

conditions :

(i) f(0) = 0 , f(2) = 3, f(4) = 6, f1(0) = f1 (4) = 0

(ii) f1 (x) > 0 for 0 < x < 4, f1 (x) < 0 for x < 0 and for x > 4

(iii) f11 (x) > 0 for x < 2, f11 (x) < 0 for x > 2.

Sketch of the solution:

Condition (i) tells us that the graph has horizontal tangents at the points (0,0) and (4,
6). Condition (ii) says that f is increasing (as f1 > o) on the interval (0,4) and
decreasing on the intervals (– , 0) and (4,  ). Condition (iii) Says that the graph is
convex (concave upward) on the interval (–  , 2) and concave on (2,  ). Because the
curve changes from convex to concave when x = 2, the point (2, 3) is an inflection
point. We use this information to sketch the graph in fig.3.5. Notice that we made the
curve bend upward when x < 2 and bend downward x > 2.

Institute of Lifelong Learning,

9 University of Delhi
 Single Variable Optimization

Another related application of second derivative is the following test for maximum
and minimum values. It is a consequences of the concavity test. The Second
Derivative Test : Suppose f11 is continuous near C.

(a) If f1 (c) = 0 and f11(c) > o), then f has a local minimum at c.

(b) If f1(c) = 0 and f11(c) < 0, then f has a local maximum at c.

For instance, part (a) is true because f11(x) > 0 near c and so f is convex near c.

This means that the graph1 of lies above its horizontal tangent at c and so f has a local
minimum at c. This may be seen from the diagram below:

Examples: 2

Institute of Lifelong Learning,

10 University of Delhi
 Single Variable Optimization

Discuss the curve Discuss the curve y = x4 – 4x3 with respect to concavity, points of
inflection and local maxima and mininima. Use this information to sketch the curve.

Solution:

f(x) = x4 – 4x3, then

f1(x)= 4x3 – 12x2 = 4x2 (x –3)

f11(x) = 12x2 – 24x = 12x(x–z)

To find the critical numbers we set f1(x) = 0 and obtain x = 0 and x = 3. To use the
second derivative test we evaluate f11 at these critical points :

f11 (0) = 0 f11(3) = 36 >0.

Since f1(3) = 0 and f11(3) > 0, f(3) = — 27 is a local minimum. Since f11(0) > 0, the
second derivative test give no information about the critical number 0. But since f1(x)
<0 for x < 0 and also for 0<x<3, the first derivative test tells us that f does not have a
local maximum or minimum at 0.

Since f11(x) = 0 when x = 0 or x = 2, we divide the real line into intervals with these
numbers as and points end complete the following schedule.

Interval f11(x) = 12x (x-2) concavity

(– , 0) + convex

(0, 2) – concave

(2,  ) + convex

The point (0,0) is an inflection point since the curve changes from convex to concave
there. Also, (2, – 16) is an inflection point since the curve changes from concave to
convex there.

Using the local minimum, the intervals of concavity and convexity and the points of
inflection, we plot the curve in the diagram given below:—

Institute of Lifelong Learning,

11 University of Delhi
 Single Variable Optimization

Note : The Second derivative test is inconclusive when f11(c) = 0. At such a point
there might be a maximum, or minimum or neither. This test fails when f11(c) does not
exist. In all such cases the first derivative test must be used. In fact, even when both
tests apply, the first derivative test is often the easier one to use.

Example :

Discuss the curve with respect to maxima, concavity and points of inflection. Also
sketch the graph of the function given below:

f(x) = x2/3 (6 – x)1/3

4–x
Solution : f1(x) =
x 1/ 3
(6 – x ) 2 / 3

–8
f11(x) =
x 4/3
(6 – x ) 5 / 3

Since, f1(x) = 0 when x = 4 and f1(x) does not exist when x = 0 or x = 6, the critical
numbers are 0, 4, and 6.

Interval 4–x x1/3 (6–x)2/3 f1(x) f

x<0 + – + – decreasing

Institute of Lifelong Learning,

12 University of Delhi
 Single Variable Optimization

on (–,0)

0<x < 4 + + + + increasing

on (0, 4)

4<x<6 – + + – decreasing
on (4, 6)

x>6 – + + – decreasing
on (6.  )

We now apply the first derivation test to find local extreme values (maxima or minima).
Since f1 change sign from positive to negative at x = 4, f(4) = 25/3 is a local maximum. The sign
of f1 does not change at 6. So there is no maximum or minimum there

[The second derivative test could be used at x = 4, but not at x = 0 or x = 6, since

f11(0) and f11 (b) does not exist.

Looking at the expression for f11(x) and noting that x 4 / 3  0 for all x, we have f11 (x)
< 0 for x < o and for 0 < x < 6, and f11(x) > 0 for x > 6. so f is concave on (– – , o)
and (0, 6) and convex on (6,  00) and the only point of inflection is (6, 0). The graph
is sketched below. Note that the curve has vertical tangents at (0,0) and (6,0) because
f 1 ( x)   as x  0 and as x  6.

Strategy for Graphing y = f(x)

Institute of Lifelong Learning,

13 University of Delhi
 Single Variable Optimization

(1) Identify the domain of f and any symmetries the curve may have.
(2) Find f1 and f11.
(3) Find the critical points of f, and identify the function's behaviour at each
one.
(4) Find intervals where the curve is increasing and when it is decreasing.
(5) Find the points of inflection, if any occur, and determine the concavity of
the curve.
(6) Identify any asymptotes.
(7) Plot the key points, such as the intercepts and the points found in steps
3,4,5 and sketch the curve.

Illustration : Using the Graphing Strategy:

( x  1) 2
Sketch the graph of f (x) =
1  x2

1. The domain of f is (– , ) and there are no symmetries about either axis or the
origin.
2. Find f1 and f11 x – intercept at x = – 1

y – intercept at y = 1 at x = o

(1  x 2 ) 2( x  1) – ( x  1) 2 2 x 2 (1 – x 2 )
f1  
(1  x 2 ) 2 (1  x 2 ) 2

Critical points: x = – 1, x = 1

(1  x 2 ) 2(–2 x) – 2(1 – x 2 ) [2(1  x 2 ) 2 x]

f ( x) 
11

(1  x 2 ) 4

4 x ( x 2 – 3)
= 
(1  x 2 ) 3

3. Behavior at critical points: The critical points occur only at x =  1 where f` (x)
= 0 since f` exists everywhere over the domain of f. x  – 1, f11 (–1) = 1 > 0
yielding a local minimum by the second derivative test.

Institute of Lifelong Learning,

14 University of Delhi
 Single Variable Optimization

At x = 1, f11 (1) = –1 < 0 yielding a local maximum by the second derivative test.

4. Increasing and Decreasing: We see that on the internal (– , – 1) the derivative f1(x) < 0,
and the curve is decreasing on the interval (–1, 1), f1(x) > 0 and the curve is increasing; it is
decreasing on ( 1, ) where f1(x) < 0 again.

5. Inflection Points: It may noted that denominator of the second derivative (step 2) is

always positive. The second derivative f11 is zero when x  3, 0 or – 3 . f11

changes sign at each of these points : negative on (–  , – 3), positive on (– 3, 0),

negative on (0, 3) and positive again on ( 3, ). Thus, each point is a point of


inflection. The curve is concave on the internal – , – 3 . Convex on   
3, 0 .,

concave on (0, 3 ), and again convex on ( 3, )

6. Asymptotes: Expanding the numerator of f(x) and then dividing both numerator
and denominator by x2 yields:

( x  1) 2 x 2  2 x  1
f ( x)   (Expanding Numerator)
1  x2 1  x2

2 1
1 
= x x2 (Dividing by x2)
1
1
x2

We see that f(x)  1+ as x   and that f(x)  1 – as x  –  . Thus, the line y =1 is

a horizontal asymptote.

Since f decreases on (– , – 1) and then increases on (–1, 1), we know that f(–1) = 0
is a local minimum. Although f decreases on (1, ) , it never crosses the horizontal
asymptote y = 1 on that interval (it approaches the asymptote from above). So the
graph never becomes negative, and f(–1) = 0 is an absolute minimum as well.
Likewise, f(1) = 2is an absolute maximum because the graph never crosses the
asymptote y =1 on the interval (–, – 1) , approaching it from below. Therefore, there
are no vertical asymptotes (the range of f is 0  y  2).

Institute of Lifelong Learning,

15 University of Delhi
 Single Variable Optimization

The graph of f is sketched in fig. 3.9.

The graph is concave down as it approaches the horizontal asymptote y = 1 as

x  – , and convex is its approach to y =1 as x   .

3. OPTIMIZATION

The concept of rationality, inter alia, means that a decision maker (consumer) firm,
government etc.) tries to find the best alternative out of those available to him. That
is, he tries to optimize.

An optimization problem consists of choice variable, an objective function, and a

feasible set. The problem to choose the most preferred alternative in the feasible set of
alternatives. Our theory, in general allows us to represent this as the problem of
finding the maximum or minimum of the objective function, subject to some
constraints. For this reason, the problem of optimization is Synonymous with finding
constrained maxima or minimum of the given function.

Solution of an optimization problem:

It is the values of the choice variables which is in the feasible set (Set of alternatives
available) and which yields a maximum or minimum value of the objective function
over the feasible set.

Institute of Lifelong Learning,

16 University of Delhi
 Single Variable Optimization

Suppose y = f(x) is the objective function and the problem is to maximize f. The
feasible set is S. Then a solution to the problem is the choice variable x* having the
property that:

f ( x * )  f ( x) for all x s .

In also frequency happens that we are concerned with the greatest or least value over
a certain neighborhood in the domain of the function rather than with the absolutely
greatest or least value over the entire domain. The next definition makes this idea
more precise.

Definition: A function is said to have a relative (local) maximum at a point no of the

domain of f if there is a neighborhood N(xo) such that f(xo)  f(x) for all x  N ( xo ) .

A relative minimum is defined in the similar manner. The greatest value (if there is
one) or the global maximum on its entire domain is sometimes called absolute
maximum. The maxima and minima of a function are called the extremes of a
function. Note that the existence of a relative maximum (or minimum) at xo implies
that the function is defined in some neighborhood of xo, N(0o). If xo is an end point of
the domain of f, then the neighborhood is a left or right neighborhood, and the
extreme is sometimes called an end point (boundary point) extreme. The following
figure illustrates some of the ways in which extremes can occur.

The graph is that of a function with relative maximum at each of the points x = x1, x3
and x4, and with a relative (local) minimum for each of x = x2 and x5. The extreme at

Institute of Lifelong Learning,

17 University of Delhi
 Single Variable Optimization

x1 is and end point maximum. The extreme at x2 is an (global) minimum. There is no

absolute maximum as the curve is indicated to be rising indefinitely for x>x5.

The figure also suggests that if f(xo) is an extreme and f1(xo) exists, then f1(x0) = o.
This seems to be so at x = x3 and x = x5. At x1, x2 and x5, the derivative apparently does
not exist (only one sided derivatives can exist at such points).

The following theorem covers these conjectures.

Theorem: If a function has an extreme f(xo), then either

(i)f1(xo) = 0, or else

(ii)f1(x0) does not exist.

The converse of the theorem is not true. For example the function defined by f(x) = x3
has a zero derivative for x = 0 since f1(n) =3 x2. Yet f has no extreme at x = o as f1 (x)
> 0 for x  0, so that f is an increasing function over its entire domain, the set of all
real numbers. One can easily verify that the curve y = x3 has a point of inflection with
a horizontal tangent at (0,0).

As an illustration of an instance where the non-existence of the derivative does not

imply an extreme. We will discuss the following function.

A direct consequence of the preceding discussion is started in the following theorems.

Theorem : If the function f is continuous on a closed interval a  x  b and if

f (a)  f (b), then there exist at least one critical point xc in the open interval a<x<b.

If in addition to the continuity of on the closed interval a  x  b, it is known that f1

exists on the open interval, then for the xc of the theorem, we have f1(xc) = 0. We state
this result in the theorem below.

Theorem: Rolle's Theorem.

If the function f is continuous on a  x  b with f(a) = f(b), and if f1(x) exists

everywhere on the open interval a<x<b, then there is at least one number xc, a<xc<b
such that f1(xc) = 0.

Illustration: Suppose we have a function of the form

Institute of Lifelong Learning,

18 University of Delhi
 Single Variable Optimization

f ( x)  1 – x 2 |, –2 x2


 1 – x2 for | x | 1 

since f ( x)   
– (1 – x  x – 1 for

2 2
1 | x | 2 


It follows that f is continuous on – 2  x  2. Moreover,

– 2x for | x | 1
f 1 ( x) 
2x for 1| x |  2
1
1/3
Let the function f (x) = (x–2) , . We first find the derivatives : f (x) 1 = ( x – 2) – 2 / 3`
3
and

f11-(x) =
5
–
–2
3

( x – 2)
9
In this case, there is no value of x such that f11(x)=0; however f11(x) fails to exist for x
= 2 also f ` (2) does not exist. The point (2,0) is on the curve (because when x = 2, f(x)
= 0, and it is wident from the expression for f11(x) that the second derivatives function
is continuous except at
x = 2. Furthermore, f11(x) >0 for x<2 and

f11(x) <0 for x > 2.

Accordingly, the curve is convex for x < 2 and concave for x > 2, so that point (2,0)
is a point of inflection. We also see that f11(x)   as x  2, s that the
inflectional tangent is vertical (see figure below).

The main point to note here is that f1(2)does not exist, although (2,0) does not
correspond to extreme point.

Hence f1(x) = 0 for x = 0, and f1(x) does not exist at x = –1 or at x = 1.

Thus, points is an example in which both types of critical points occur.

Furthermore, since f(1) = f(–1) and f1(x) exists for –1 < x < 1, we may regard

Institute of Lifelong Learning,

19 University of Delhi
 Single Variable Optimization

the function on the interval – 1  x  1 as an illustration of the Rolle's Theorem

(see figure below):

3.1 First Derivative Test for Maxima and Minima

Theorem : Let f be a continuous function over a N(xc, h) where xc is a critical point of

the domain of f1, and let f1(x) exist for x  N* ( xe– , h) Then

(i) f’(x) > 0 for x  N* ( xe– , h) and f’(x)>0 for x  N* ( xe , h)  f ( xc ) is a maximum.



(ii) f1 (x) < 0 for x N * ( xc , h) and f1(x) < 0 for x N * ( xc , h)  f ( xc ) is a minimum.

–

(iii) f1(x) of constant sign (negative or positive) for x N * ( xc , h)  f ( xc )

Example: Find the extremes of the function defined by f(x) =2x3 + 3x2 – 12x

Solution: f(x) = 6x2 + 6x – 12 = 6( x 2  x – 2)  6 ( x  2) ( x – 1)

If f1 (x) = 0 then x = – 2 or 1. Since f1 is continuous for all x, the only possible

extremes are at these critical values of x. We also find that:

f1(x) < 0 for x < – 2 and for x < 1

f1(x) < 0 for –2 <x < 1.

We see that f1(–2) = 0, and f1(x) changes sign from plus to minus as x increases
through the value –x. The point (–2, 20) is a maximum point on the graph of f. Also,
f1(1) = 0 and f1(x) changes sign from minus to plus as x increases through value 1.
The point (1– 7) is a minimum point on the graph.

Example: Find the extremes of the function defined by f(x) = x2/3.

Institute of Lifelong Learning,

20 University of Delhi
 Single Variable Optimization

–1
2
Solution: 1
f (x) = x 3
,x0
3
The domain of f is all xR, yet here is no value of n for which f ` (x) = 0.

The only possible critical value of x is x = 0, for which f1(x) does not exist. Since
f1(x) < 0 for x < 0 and f1(x) > 0 for x> 0, f(o) = 0 is a relative (local) minimum of the
function.

Fig.

In the discussion of concavity of the function, we have already mentioned second

Derivative Test for Maxima and Minima.

4. Geometric Interpretation of Rolle's Theorem and the Mean Value

Theorem.

To explain, geometrically, Rolle's Theorem, let there be y = f(x) with f being

continuous on a  x  b. If f(a)= f(b), then there is at least one point (c, f(c) with
a<c<b, where the tangent to the curve is horizontal (see figure below).

Clearly, if the curve of the function in panel a were turned so that the line AB became
not parallel to the x–axis, the geometric content of Rolle's Theorem would still be
true; the tangent line through point P (in panel b) would still be parallel to the recent
line AB. This apparently evident result is analytically formalised in the Mean value
Theorem.

Institute of Lifelong Learning,

21 University of Delhi
 Single Variable Optimization

4.1 The mean value Theorem for Derivatives

Let f be a continuous function on the closed interval a  x  b. and let f ` exist

everywhere on the open interval a < x < b then there exist at least one point c  (a,b)
f (b) – f (a)
s/+ f1(c) =
b –a

This equation is often written in the form obtained by solving for f(b). f(b) = f (a) + f1
(c) (b–a), c  (a,b).

Example : Find the value(s) of c which the mean value Theorem predicts if f(x) = x3 –
x, a = 0, b = 3.

Solution : f(a) = 0, f(b) = 24, f1(x) = 3x2 –1

Hence, for the required value of c, we must have

f (b) – f (a)
f1(c) = Mean value theorem (henceforth MVT)
b–a

24 – 0
3c2–1 =  8, or , c 2  3 or c  3 or – 3
3–0

Since C  (0,3) or 0 < c < 3, the positive root 3 is the desired value of c.

Example: Estimate the value of 110 by using MVT.

Solution: Let (f(x)... = x and use the MVT in the form f(x + h) = f(x) + h f1(x + oh), 0<
1 1
0    1. Since 
2 x and f (x + h)
1 = 1
f (x) 2 x  h , we have

h
xh  x 
2 x  h

5
With x = 100 and h = 10, this formula gives 110  10 
100  100

Since f1 (x) > 0, we have

100  100  100  110

Institute of Lifelong Learning,

22 University of Delhi
 Single Variable Optimization

From the preceding two inequalities we have

5
110  10   10.5
100

5 110
Similarly, 110  10   10 
110 22

21
or,  110  10, so that
22

220
110   10.476, Finally, we have
21

10.476 < 110  10.5

Example* : Discuss and sketch the curve :

x3 + xy2 + ay2– 3ax2 = 0, a > 0

Solution :

1. We see that if x = 0, then y = 0 and if y = 0, then x3 – 3ax2 = 0 so that x = 0 or

x = 3a.

Hence the curve crosses the axes at (0,0) and (3a, 0).

2. Since y appears y appears to even powers only, the curve must be symmetric
to the x–axis. This is the only simple symmetry the curve has.

3a – x
3. We solve for y to obtain y   x which show that there is a vertical
x–a
asymptote at x = a provided that x = – a within or on the boundary of the curve. As
we shall notice in the extent of the curve is –a < x  3a,the curve lies to the right of
and is asymptotic to the line x = – a.

4.To determine the extent of the curve (the set of values of x for which y is real), we
3a – x
solve the inequality  0 . The required solution is – a  x  3a .
xa

Institute of Lifelong Learning,

23 University of Delhi
 Single Variable Optimization

Note : Solving for x in terms of y involves solving a general cubic equation. Since a
cubic equation. with real coefficients always has one or three distinct real roots, it is
clear that for each real value of y, there is always at least one corresponding real
value of x and there may be three such values. Consequently, there is no restriction
on the extent of the curve in the y-direction.

5.Maximum and minimum points on the curve : Let us consider the two separate
equations :

3a – x 3a – x
y1  x , y2  – x
xa x a

4. These two formulas represent two branches of the curve. The two equations
can now be properly associated with functions rather than relations. Because of the
symmetry of the curve, one branch is symmetric to other with respect to the x-axis. so
that we may restrict the discussion to y1 Since,

3a 2 – x 2
y1`  , we see that x = 3a is a critical value of x.
( x  a) 3 (3a – x)

The nature of the point at x  3 a can be investigated easily by determining the

intervals over which the curve is rising or falling. It is easy to see that :

y1`   0 for – a  x  3 a and,

y1`  0 for 3 a  x  3 a, so that the curve is rising and falling over the

respective intervals. This show that  

3 a, 1.2a is a minimum point.

We have already seen that x= –a is a vertical asymptote. The fact that

| y11 |   as x  a  is consistent with this result. However, at x = 3a, as

x  3a – , y1'  –  . This indicates that there is a vertical asymptote at (3a,0) or

that the curve approaches the x – axis vertically
as x  3a .

Institute of Lifelong Learning,

24 University of Delhi
 Single Variable Optimization

12a 3
We see that y  –
"
. It is clear here that the curve is concave
(3a – x) 3 / 2 ( x  a) 5 / 2
1

everywhere and hence there are no points of inflection:

Graph of x 3  xy3  xy 2 _ ay2 – 3ax2  0, a  0.

In the above figure, concave portion of the curve corresponds to y1 and the convex
portion, which is obtained by symmetry, correspond to y2,

Note: The curve is known as a tri-sectrix because of the property that the angle  is
one-third of angle  if x > 0, y> 0 for p(x, y).

Example: Find extremes of the function:

f ( x) | x`|  | x 2 – 1 | For all x  R.

Solution: f 1 ( x)  – 1  2 x for x  (– , – 1)

= – 1 – 2x for x (–1, 0)

= 1 – 2x for x  (0, 1)

= 1  2x for x  (1, )

Institute of Lifelong Learning,

25 University of Delhi
 Single Variable Optimization

Note that, here, f1 does not exist at –1, 0,1 and f1 = 0 at .

1 1 1 1
x– and . Also, f "  – 2  0 at x  – , Hence the function has maximum
2 2 2 2
5 1 1
value at x  – , .
4 2 2

1
To investigation the existence of extreme values of f at –1, 0, 1 for 0    , we
4
note that f(–1) = 1, f(–1 –  ) = 1 + (1+  )2–1 = 1+ 3  +  2

f (–1   ) = 1 –   1 – (1 –  ) 2  1  –  2

f(0) = 1, f(0–  ) =  +1–  2, f(0+  ) =  +1 –  2

f(1) = 1, f(1–  ) = 1 –  +1 –(1–  )2 = 1 +  –  2

f(1+  ) =1 +  + (1 +  )2 – 1 = 1 + 3  +  2

These show that the function f has minimum values 1 at x = –1, 0, 1.

5. Existence, Global and Uniqueness of solution.

5.1 Weierstass Theorem (Existence)

An optimization problem always has a solution if:

(a) An optimization the objective function is continuous, and
(b) the feasible set is non-empty
(c) closed
(d) bounded
It should be noted that the conditions of continuity of the objective function and
closedness and boundedness of the feasible set are sufficient but not necessary
conditions for existence of a solution. In other words, solution may exist if they are
not satisfied, but solutions may also not exist. Satisfaction of the conditions, however,
rules out all possible cases of non-existence. Non-emptiness of the feasible set is a
necessary condition for existence of a solution: any problem in which no point is
feasible cannot have a solution.

Institute of Lifelong Learning,

26 University of Delhi
 Single Variable Optimization

5.2 Global Maximum

A local maximum is always a global maximum if
(a) the objective function is concave
(b) feasible set is convex.

Example: Show that the curve y = ax3 + bx2 + cx can have only one point of inflexion.
If a is positive, show the curvature changes from concave to convex from below as we
pass through the point of inflection from left to right. Deduce that the point of
inflection is also a stationary point if b2 = 3ac.

Solution: f(x)= y = ax3 + bx2+ cx

f1 = 3ax2 + 2bx +c

f11 = 6a x + 2b

Necessary condition for point of inflection: f11 = 0

–b
6ax + 2b = 0  x 
3a

Sufficient condition for point of inflection: f 111  0.

f 111  6a  0 (if a  0).

–b
Therefore, we have a point of inflection at x =
3a

a  0  f 111  0, Hence, the curvature changes from concave to convex.

[Note: if f111 < 0, curvature changes from convex to concave].

The point of inflection is also a stationary point if f 1 is also zero at that point.
2
–b  –b  –b
f 1
 3a    26 .  c  0
3a  3a   3a 

b 2 2b 2
Or, – c0
3a 3a

Institute of Lifelong Learning,

27 University of Delhi
 Single Variable Optimization

– b2
Or, c0
3a
Or, b2 = 3ac.

–b
Hence, The point of inflection is also a stationary point if b2 = 3ac.
3a
5.3 Uniqueness of Solution

Given an optimization problem in which the feasible is convex and the objective
function is non-constant and quasi-concave, a solution is unique if :

(a) the feasible set is strictly convex, or

(b) the objective function is strictly quasi-concave, or

(c) both

5.4 Related examples and their solutions

Example: Show that the maximum value of average product of a is a constant

H 2 – AB
which is independent of fixed amount of b used for the production
B

function : x = 2 Hab – Aa 2 – Bb2

Solution: APa = x  1 2 Hab – Aa 2 – Bb 2

a a
APa will be maximum, if (i) f1= 0 and f11 < 0.

1  a(2 Hb – 2 Aa 
f1 =   – 2 Hab – Aa – Bb  0
2 2

a2  2 2 Hab – Aa – Bb 
2 2

1  – 2 Hab  2 Bb 2 
Or,  0
a 2  2 2 Hab – Aa 2 – Bb 2 


B aH
Or, –2 Hab +2Bb2 = 0 or a =  b or b 
H B

Institute of Lifelong Learning,

28 University of Delhi
 Single Variable Optimization

B
It is easy to check that at a   b, f 11  0.
H
6
B
Thus APa will be maximum at a   
H

aH
To get APmax, we substitute b = , and simplify.
B

2 Hab Bb 2
APmax = – A– 2
a2 a

2 Ha aH aH B a2 H 2
= – A–
a2 B a2B2

2 H 2 – AB – H 2
=
B

H 2 – AB
APmax = Hence Proved.
B

Example :

k
Determine the constant K so that the function f(x) = x2 + may have a (i) minimum
x
at x = 2, (ii) a minimum at x = –3. Show that that the function cannot have a local
maximum for any value of K.

K k
Solution: f1(x) = 2x – 2
 0  x3 
x 2

2k
f11(x) = 2 + >0 (for minimum)
x3

k
(by substitutions x3 = )
2
1/ 2
k 
Therefore, x=   is a point of minimum.
2 

x = 2  k = 16

Institute of Lifelong Learning,

29 University of Delhi
 Single Variable Optimization

k
(ii) x=–3   – 27  k  – 54
2
1  k 
f (x) = 2  x 3 
11

  13  
 k 
  2    2(1  2)  6  0. Which is always positive.
11
f
 

Therefore, the function cannot have a local maximum for any value k.

Example: A wire of length L is cut into two piece, one being bent to form a square
and the other to form an equilateral triangle. How should the wires be cut if the sum
of the two areas is minimum?

Solution: Let the wire be cut at length  and is used to form the triangle. Side =
 / 3.

3 3  
Altitude = Side  . 
2 2 3 2 3

3
Area of the triangle (equilateral) = . ( Side) 2
4
2
3 
= . 
4  3

2
=
12 3
2
L – 
A2 = Area of the square =  
 4 

2 ( L – ) 2
A = Total Area = A1 + A2 = 
12 3 16

dA 2l 2( L – l )
 – 0
dl 12 3 16

Institute of Lifelong Learning,

30 University of Delhi
 Single Variable Optimization

 L–
Or, – 0
6 3 8

4 – 3 3 L  3 3 
Or,  0
24 3

Or,
4  3 3  – 3 3L
0
24 3

Or, (4  3 3)   3 3 L

 3 3 
Or,   L

 4  3 3 

d2A 4  3 3
 0
dr 2 24 3

3 3
Hence, Using fraction of the total length
43 3

L and rest for the square gives the minimum total area.

6. Complete criteria for Maximum, Minimum and inflectional values.

(1) If f(x) has an extreme value at x = a then f1(a) = 0

(2) If f1(a) = f11(a) = ... = f(n–1) (a) = 0, fn) (a)  0, then f(x)has a stationary value
at
x = a which is an inflectional value if n is odd, a maximum value if n is even
and f(n) (a) <0 and a minimum value if n is even and f(n) (a) > 0.

This criterion is complete and so both necessary and sufficient, subject to the
condition that the derivatives involved are finite and continuous.

There is no case of failure; unless the function is a constant (and hence without
maxima and minima) there must always be some derivative which is not zero.

Illustration : Let y = (x –1)5. Now we investigate this function with respect to maxima,
minima or point of inflection.

Institute of Lifelong Learning,

31 University of Delhi
 Single Variable Optimization

f(x) = y = (x –1)5

f1 = 5(x –1)4 = 0  x = 1

f11 = 20 (x–1)3 = 0 at x = 1

f111 = 60 (x –1)2 = 0 at x = 1

f(4) = 120(x–1) = 0 at x = 1

f(5) = 120  0 at x = 1

Now, we apply nth derivative test.

Here, the first non-zero derivative at x = 1 is the 5th – order derivative (Odd – ordered
derivative). Hence at x = 1, the function has a point of inflection.

at x= 1, y = (1–1)5 = 0

Thus, (1,0) is an inflection point.

Example : (Curious Case) Find maxima and Minima of the function f(x) = y =
2–x
x x –2
2

x ( x – 4)
f1 
( x  x – 2) 2
2

–1
There are stationary values y = –1 at x = 0 and y = at x = 4.
9
To check sign of the derivative near these stationary values x = 0, 4, we get the
following :

h(h – 4)
f1 = at x = 0+h
(h  h – 2) 2
2

h (h  4)
f1  at x = 4 + h
(h  9h  18) 2
2

As may be easy to see, the first expression changes from positive to negative as h is
given small values changing from negative to positive. The second expression

Institute of Lifelong Learning,

32 University of Delhi
 Single Variable Optimization

changes in the opposite sense as h is varied similarly. The function thus has a
1
maximum value –1 at x = 0 and a minimum value – at x = 4.
9
The curious feature of this case is that the maximum value of the function is smaller
than the minimum value. This apparently paradoxical result is due to the fact that the
function has infinite values at x = 1 and at x = – 2. (At each of these values the
denominator of y is zero). The graph of the function illustrates how the presence of
infinities influences the maximum and minimum values.

Example : If a monopolist has a total cot function c  ax2  bx  c and if the demand

law is p   – x 2 , then show that the output for maximum net revenue is

a 2  3 x (  – b) – a
x=
3x
Solution:

Net Revenue =   p. x – c( x)

= ( – x 2 ) x – (ax2  bx  c)

= x – x 3 – ax2 – bx – c

Institute of Lifelong Learning,

33 University of Delhi
 Single Variable Optimization

d
Necessary condition for maxima 0
dx

Or,  – 3x 2 – 2ax – b  0

Or, 3x 2  2ax  (b –  )  0

– 2a  4a 2 – 12 (b –  )
Or, x 
6

– 2a  4[a 2  (  – b) 3
Or, x 
b

x : negative ruled out.

– a  a 2  3 (  – b) a 2  3 (  – b) – a
Hence : x  
3 3

d2 
Sufficient condition for maxima: 0
dx2

d2 
 0.
dx2
Hence the output for maximum net revenue,

a 2  3 (  – b) – a
x
3

7. End Chapter Exercise:

1. Without appealing to graphical ideas, find the location and nature of the
extreme of the following two functions and determine if they are differentiable
at these points :

1 3
(a) f ( x)  x  2 x 2  3x  1
2

(b) f ( x)  (2 x – 5) x 2 / 3

Institute of Lifelong Learning,

34 University of Delhi
 Single Variable Optimization

1
2. Show that the curve of y  2 x – 3  is convex from below for all positive
x
c
values of x. Is same true for y  ax  b  ?
x

a
3. Show that the demand curve p  – c is downward sloping and
( x  b)
convex from below. Do the same properties hold for marginal revenue curve?

4. If the supply function is x  a p – b  c , where a,b,c are positive constants,

show that the supply curve is upward sloping and concave to axis OP(price
axis) at all points.

5. Find, on the part of rectangular hyperbola x y = 4 in the positive quadrant, the

point which is nearest to the origin and show that the shortest distance is
perpendicular to the tangent at this point.

1
6. Slow that y  x  has one maximum and one minimum value and the latter
x
is larger than the former.

7. A firm has the following total cost and demand functions :

1 3
c Q – 7Q 2  111 Q  50
3
Q = 100 – p.

(a) Find profit maximizing level of output.

(b) Find Equilibrium Level of output (profit maximizing) if the firm is

assumed to fix the price.

8. A radio manufactures produces 'x' sets per week at a total cost of

 x2 
Rs.   3x  100  . He is a monopolist and the demand for his product is
 25 
x = 75 – 3p; p = price/set. Show that the maximum set revenue is obtained

Institute of Lifelong Learning,

35 University of Delhi
 Single Variable Optimization

when about 30 sets are produced per week. What is the monopoly price and
net revenue at this level of output?

2x 2 4 x (1  x 2 ) (1  x) (1 – x)
9. Show that f ( x)  4  f ( x) 
1

x 1 ( x 4  1) 2

Also find the maximum value of f on (0, ) . Show that f(–x) = f(x), for all x.
What are the maximum points for f on (– , ) ?

10. Find two positive numbers whose sum 16 and whose product is as large as
possible.

11. Let C(Q) = a Qb + c, for a>0, b > 1, and c  o, be cost function. Prove that the
average cost function has a minimum on (0, ), and find it.

6x3
12. Classify the stationary points of f(x) = , with respect to maxima,
x4  x2  2
maxima and point(s) of inflection.

13. Let f be defined for all x by f(x) = (x2–1)2/3.

(a) Compute f1(x) and f11(x).

(b) Find local extreme points of f, and draw the graph of f.

14. Find possible inflection points for f(x) = x2ex. Draw its graph.

15. Find the intervals where the following Cubic cost function is convex and
where it is concave, and find the unique inflection point:

(Q) = aQ3 + bQ2 + cQ + d, a>0, b< 0, c>0, d>0.

(16) Are the following functions concave or convex (assuming x > 0 in parts (b)
and (c) ?

1 x 1 –x
(a) e  e (b) 2x – 3 + 4 lnx
2 2
1

(c) 5 x 2 – 10 x 3 / 2 (d) 3x2 – 2x + 1 + e–x – 3

Institute of Lifelong Learning,

36 University of Delhi
 Single Variable Optimization

8. References
K. Sydsaeter and P. Hammond, Mathematics for Economic Analysis,
Pearson Educational Asia, Delhi, 2002.

Institute of Lifelong Learning,

37 University of Delhi