Real Analysis Notes by Prof Syed Gul Shah

Chapter 1 – Real Number System
Subject: Real Analysis (Mathematics) Level: M.Sc.

Collected & Composed by: Atiq ur Rehman (mathcity@gmail.com), http://www.mathcity.org
The rational number system is inadequate for many purposes, both as a field
and as an order set for many purpose. This leads to introduction of so called irrational
numbers. We can prove in many ways that the rational number system has certain
gaps and hence we fail to use it as an ordered set and as a field. q
Í Theorem
There is no rational p such that p 2 = 2 .
Proof
Let us suppose that there exists a rational p such that p 2 = 2 .
This implies we can write
m
p= where m, n ∈ ¢ & m, n have no common factor.
n
m2
Then p = 2 ⇒ 2 = 2 ⇒ m2 = 2n 2
2
n
⇒ m is even
2
⇒ m is even
⇒ m is divisible by 2 and so m 2 is divisible by 4.
⇒ 2n2 is divisible by 4 and so n 2 is divisible by 2. Q m 2 = 2n 2
i.e. n 2 is even ⇒ n is even
⇒ m and n both have common factor 2.
Which is contradiction. (because m and n have no common factor.)
Hence p 2 = 2 is impossible for rational p. q
Í Theorem
Let A be the set of all positive rationals p such that p 2 < 2 and let B consist of
all positive rationals p such that p 2 > 2 then A contain no largest member and B
contains no smallest member.
Proof
We are to show that for every p in A there exists a rational q ∈ A such that
p < q and for all p ∈ B we can find rational q ∈ B such that q 0 the number
p2 − 2 2p + 2
q= p− = ……..……. (i)
p+2 p+2
2
 2p + 2 2( p 2 − 2)
Then q − 2 = 
2
 − 2 = ( p + 2) 2 ……………(ii)
 p+2 
Now if p ∈ A then p 2 < 2 ⇒ p 2 − 2 < 0
p2 − 2
Since from (i) q= p− ⇒ q> p
p+2
2( p 2 − 2)
And < 0 ⇒ q2 − 2 < 0 ⇒ q2 < 2 ⇒ q∈ A
( p + 2) 2
Now if p ∈ B then p 2 > 2 ⇒ p 2 − 2 > 0

Ch. 01 - Real Number System - 2
p2 − 2
Since form (i) q= p− ⇒ q0 ⇒ q2 − 2 > 0 ⇒ q2 > 2 ⇒ q ∈ B
( p + 2) 2
The purpose of above discussion is simply to show that the rational number
system has certain gaps, in spite of the fact that the set of rationals is dense i.e. we
can always find a rational between any two given rational numbers. These gaps are
r+s
filled by the irrational number. (e.g. if r < s then r < < s .) q
2
Í Order on a set
Let S be a non-empty set. An order on a set S is a relation denoted by “ < ” with
the following two properties
(i) If x ∈ S and y ∈ S ,
then one and only one of the statement x < y , x = y , y < x is true.
(ii) If x, y, z ∈ S and if x < y , y < z then x < z .
Í Ordered Set
A set S is said to be ordered set if an order is defined on S.
Í Bound
Let S be an ordered set and E ⊂ S . If there exists a β ∈ S such that
x ≤ β ∀ x ∈ E , then we say that E is bounded above, and β is known as upper
bound of E.
Lower bound can be define in the same manner with ≥ in place of ≤ .
Í Least Upper Bound (Supremum)
Suppose S is an ordered set, E ⊂ S and E is bounded above. Suppose there
exists an α ∈ S such that
(i) α is an upper bound of E.
(ii) If γ < α then γ is not an upper bound of E.
Then α is called the least upper bound of E or supremum of E and is written as
sup E = α .
In other words α is the least member of the set of upper bound of E.
We can define the greatest lower bound or infimum of a set E , which is bounded
below, in the same manner. q
Í Example
Consider the sets
A = { p : p ∈ ¤ ∧ p 2 < 2}
B = { p : p ∈ ¤ ∧ p 2 > 2}
where ¤ is set of rational numbers.
Then the set A is bounded above. The upper bound of A are the exactly the members
of B. Since B contain no smallest member therefore A has no supremum in ¤ .
Similarly B is bounded below. The set of all lower bounds of B consists of A and
r ∈¤ with r ≤ 0 . Since A has no largest member, therefore, B has no infimum in ¤ .
Í Example
If α is supremum of E then α may or may not belong to E.
Let E1 = {r : r ∈ ¤ ∧ r < 0}
E2 = {r : r ∈ ¤ ∧ r ≥ 0}
then sup E1 = inf E2 = 0 and 0 ∉ E1 and 0 ∈ E2 . q
Í Example
1
Let E be the set of all numbers of the form , where n is the natural numbers.
n
 1 1 1 
i.e. E = 1, , , ,.........
 2 3 4 
Then sup E = 1 which is in E, but inf E = 0 which is not in E. q
Í Least Upper Bound Property
A set S is said to have the least upper bound property if the followings is true
(i) S is non-empty and ordered.
(ii) If E ⊂ S and E is non-empty and bounded above then supE exists in S.
Greatest lower bound property can be defined in a similar manner. q
Í Example
Let S be set of rational numbers and
E = { p : p ∈ ¤ ∧ p 2 < 2}
then E ⊂ ¤ , E is non-empty and also bounded above but supremum of E is not in S,
this implies that ¤ the set of rational numbers does not posses the least upper bound
property. q
Í Theorem
Suppose S is an ordered set with least upper bound property. B ⊂ S , B is non-
empty and is bounded below. Let L be set of all lower bounds of B then α = sup L
exists in S and also α = inf B .
In particular infimum of B exists in S.
OR
An ordered set which has the least upper bound property has also the greatest
lower bound property.
Proof
Since B is bounded below; therefore, L is non-empty.
Since L consists of exactly those y ∈ S which satisfy the inequality.
y≤x ∀ x∈B
We see that every x ∈ B is an upper bound of L.
⇒ L is bounded above.
Since S is ordered and non-empty therefore L has a supremum in S. Let us call it α .
If γ < α , then γ is not upper bound of L.
⇒ γ ∉B L B
⇒ α≤x ∀ x∈ B ⇒ α ∈L γ α
Now if α < β then β ∉ L because α = sup L .
We have shown that α ∈ L but β ∉ L if β > α . In other words, α is a lower
bound of B, but β is not if β > α . This means that α = inf B . q
………………………
Í Field
A set F with two operations called addition and multiplication satisfying the
following axioms is known to be field.
Axioms for Addition:
(i) If x, y ∈ F then x + y ∈ F . Closure Law
(ii) x + y = y + x ∀ x, y ∈ F . Commutative Law
(iii) x + ( y + z ) = ( x + y ) + z ∀ x, y, z ∈ F . Associative Law
(iv) For any x ∈ F , ∃ 0 ∈ F such that x + 0 = 0 + x = x Additive Identity
(v) For any x ∈ F , ∃ − x ∈ F such that x + (− x ) = (− x) + x = 0 +tive Inverse
Axioms for Multiplication:
(i) If x, y ∈ F then x y ∈ F . Closure Law
(ii) x y = y x ∀ x, y ∈ F Commutative Law
(iii) x ( y z ) = ( x y ) z ∀ x, y, z ∈ F
(iv) For any x ∈ F , ∃ 1∈ F such that x ⋅1 = 1⋅ x = x Multiplicative Identity
1 1 1
(v) For any x ∈ F , x ≠ 0 , ∃ ∈ F , such that x   =   x = 1 × tive Inverse.
x x x
Distributive Law
For any x, y , z ∈ F , (i) x ( y + z ) = xy + xz
(ii) ( x + y ) z = xz + yz q
Í Theorem
The axioms for addition imply the following:
(a) If x + y = x + z then y = z
(b) If x + y = x then y = 0
(c) If x + y = 0 then y = − x .
(d) −(− x) = x
Proof
(a) Suppose x + y = x + z .
Since y = 0 + y
= ( − x + x) + y Q −x+x=0
= − x + ( x + y) by Associative law
= − x + ( x + z) by supposition
= (− x + x ) + z by Associative law
= (0) + z Q −x+x=0
=z
(b) Take z = 0 in (a)
x+ y = x+0
⇒ y =0
(c) Take z = − x in (a)
x + y = x + ( − x)
⇒ y = −x
(d) Since (− x) + x = 0
then (c) gives x = −(− x) q
Í Theorem
Axioms of multiplication imply the following.
(a) If x ≠ 0 and x y = x z then y = z .
(b) If x ≠ 0 and x y = x then y = 1 .
1
(c) If x ≠ 0 and x y = 1 then y = .
x
1
(d) If x ≠ 0 , then = x.
1
x
Proof
(a) Suppose x y = x z
1  1
Since y = 1 ⋅ y =  ⋅ x  y Q ⋅x =1
x  x
1
= ( x y) by associative law
x
1
= (x z) Qxy = xz
x
1 
=  ⋅ xz by associative law
x 
= 1⋅ z = z
(b) Take z = 1 in (a)
x y = x ⋅1 ⇒ y =1
1
(c) Take z = in (a)
x
1
x y = x⋅ i.e. x y = 1
x
1
⇒ y=
x
1
(d) Since ⋅ x =1
x
then (c) give
1
x= q
1
x
Í Theorem
The field axioms imply the following.
(i) 0 ⋅ x = 0
(ii) if x ≠ 0 , y ≠ 0 then xy ≠ 0 .
(iii) (− x) y = − ( xy ) = x (− y )
(iv) (− x)(− y ) = xy
Proof
(i) Since 0 x + 0 x = (0 + 0) x
⇒ 0x + 0x = 0x
⇒ 0x = 0 Qx+ y=x ⇒ y=0
(ii) Suppose x ≠ 0, y ≠ 0 but x y = 0
1
Since 1 = ⋅xy
( x)( y )
1
⇒ 1= (0) Q xy = 0 , x ≠ 0 , y ≠ 0
( x)( y )
⇒ 1= 0 from (i) Q x 0 = 0
a contradiction, thus (ii) is true.
(iii) Since (− x) y + xy = (− x + x ) y = 0 y = 0 …….. (1)
Also x (− y ) + xy = x (− y + y ) = x 0 = 0 ……… (2)
Also − ( xy ) + xy = 0 …………. (3)
Combining (1) and (2)
(− x) y + xy = x (− y ) + xy
⇒ (− x) y = x (− y ) ………… (4)
Combining (2) and (3)
x (− y ) + xy = − ( xy ) + xy
⇒ x(− y ) = − xy …………. (5)
From (4) and (5)
(− x) y = x (− y ) = − xy
(iv) (− x)(− y ) = − [ x(− y ) ] = − [ − xy ] = xy using (iii) q
7⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅8
Í Ordered Field
An ordered field is a field F which is also an ordered set such that
i) x + y < x + z if x, y , z ∈ F and y < z .
ii) xy > 0 if x, y ∈ F , x > 0 and y > 0 .
e.g. the set ¤ of rational number is an ordered field. q
Í Theorem
The following statements are true in every ordered field.
i) If x > 0 then − x < 0 and vice versa.
ii) If x > 0 and y < z then xy < xz .
iii) If x < 0 and y < z then xy > xz .
iv) If x ≠ 0 then x 2 > 0 in particular 1 > 0 .
1 1
v) If 0 < x < y then 0 < < .
y x
Proof
i) If x > 0 then 0 = − x + x > − x + 0 so that − x < 0 .
If x < 0 then 0 = − x + x < − x + 0 so that − x > 0 .
ii) Since z > y we have z − y > y − y = 0
which means that z − y > 0 , Also x > 0
∴ x( z − y) > 0
⇒ xz − xy > 0
⇒ xz − xy + xy > 0 + xy
⇒ xz + 0 > 0 + xy
⇒ xz > xy
iii) Since y < z⇒ −y + y < −y + z

⇒ z− y>0
Also x < 0 ⇒ − x > 0
Therefore − x( z − y ) > 0
⇒ − xz + xy > 0 ⇒ − xz + xy + xz > 0 + xz
⇒ xy > xz
iv) If x > 0 then x ⋅ x > 0 ⇒ x 2 > 0

If x < 0 then − x > 0 ⇒ (− x)(− x) > 0 ⇒ (− x) 2 > 0 ⇒ x2 > 0
i.e. if x > 0 then x 2 > 0 , since 12 = 1 then 1 > 0 .
1 1
v) If y > 0 and v ≤ 0 then y v ≤ 0 , But y   = 1 > 0 ⇒ >0
 y y
1
Likewise > 0 as x > 0
x
 1  1 
If we multiply both sides of the inequality x < y by the positive quantity   
 x  y 
 1  1   1  1 
we obtain     x <     y
 x  y   x  y 
1 1
i.e. <
y x
1 1
finally 0< < q
y x
Í Existence of Real Field

There exists an ordered field ¡ (set of reals) which has the least upper bound
property and it contain ¤ (set of rationals) as a subfield. q
Í Theorem
a) If x ∈ ¡ , y ∈ ¡ and x > 0 then there exists a positive integer n such that
nx > y . (Archimedean Property)
b) If x ∈ ¡ , y ∈ ¡ and x < y then there exists p ∈ ¤ such that x 0, x ∈ ¡}
Suppose the given statement is false i.e. nx ≤ y .
⇒ y is an upper bound of A.
Since we are dealing with a set of reals, therefore, it has the least upper bound
property.
Let α = sup A
⇒ α − x is not an upper bound of A.
⇒ α − x < mx where mx ∈ A for some positive integer m.
⇒ α < (m + 1) x where m + 1 is integer, therefore (m + 1) x ∈ A
Which is impossible because α is least upper bound of A i.e. α = sup A .
Hence we conclude that the given statement is true i.e. nx > y .
b) Since x < y , therefore y − x > 0
⇒ ∃ a +ive integer n such that
n ( y − x) > 1 (by Archimedean Property)
⇒ ny > 1 + nx …………… (i)
We apply (a) part of the theorem again to obtain two +ive integers m1 and m2
such that m1 ⋅ 1 > nx and m2 ⋅1 > − nx
⇒ − m2 < nx < m1
then there exists an integers m(− m2 ≤ m ≤ m1 ) such that
m − 1 ≤ nx < m
⇒ nx < m and m ≤ 1 + nx
⇒ nx < m < 1 + nx
⇒ nx < m < ny from (i)
m
⇒ x< < y
n
⇒ x 0, y > 0
Then ∃ a rational number q such that
x y
0< <q< where α is an irrational.
α α
⇒ x < αq < y
⇒ x<u< y
Where u = α q is an irrational as product of rational and irrational is irrational. q
Í Theorem
For every real number x there is a set E of rational number such that x = sup E .
Proof
Take E = {q ∈ ¤ : q < x} where x is a real.
Then E is bounded above. Since E ⊂ ¡ therefore supremum of E exists in ¡ .
Suppose sup E = λ .
It is clear that λ ≤ x .
If λ = x then there is nothing to prove.
If λ < x then ∃ q ∈ ¤ such that λ < q < x
Which can not happen. Hence we conclude that real x is supE. q
Í Theorem
For every real x > 0 and every integer n > 0 , there is one and only one real y such
that y n = x .
1
This number y is written n x or x n .
Proof
Take y1 , y2 ∈ ¡ such that 0 < y1 < y2 . Then y1n < y2n i.e. there is at most one y ∈ ¡
such that y n = x . This shows the uniqueness of y.
Let us suppose E be the set of all positive real numbers t such that t n < x .
i.e. E = {t : t ∈ ¡ ∧ t n < x}
x
Take t = then 0 < t < 1 .
1+ x
Hence t n < t and we have t n < x
⇒ tn < t < x
⇒ t ∈ E and E is non-empty .
If t > 1 + x then t n > t > x so that t ∉ E .
Thus 1 + x is an upper bound of E.
Since E is non-empty and bounded above therefore sup E exists.
Take y = sup E
To show that y n = x we will show that each of the inequality y n < x and y n > x
leads to contradiction.
Consider
bn − a n = (b − a)(b n−1 + bn −2 a + bn−3a 2 + ⋅⋅ ⋅⋅ ⋅⋅ ⋅⋅ ⋅⋅ + a n −1 ) where n ∈¢ + .
Which yields the inequality (each a is replaced by b on R.H.S of above)
bn − a n < (b − a)(nbn −1 ) ..................(i) where 0 < a < b .
Now assume y n < x
x − yn
Choose h so that 0 < h < 1 and h <
n( y + 1)n −1
Put a = y and b = y + h in (i)
( y + h) − y n < nh ( y + h )
n n−1
Then
< nh( y + 1) n−1 Q h <1
< x − yn
⇒ ( y + h) < x
n
⇒ y + h∈E
Since y + h > y therefore it contradict the fact that y is sup E .
Hence y n < x is impossible.
Now suppose y n > x

yn − x
Put k = , then 0 < k < y
ny n−1
Now if t ≥ y − k we get
y n − t n < y n − ( y − k )n < y n − ( y n − nky n −1 ) by binomial expansion
< kny n−1 = y n − x
⇒ − t n < − x ⇒ t n > x and t ∉ E
It follows that y − k is an upper bound of E but y − k < y , which contradict the
fact that y is sup E .
Hence we conclude that y n = x . q
Í The Extended Real Numbers
The extended real number system consists of real field ¡ and two symbols + ∞
and − ∞ , We preserve the original order in ¡ and define
−∞ < x < +∞ ∀ x ∈ ¡ .
The extended real number system does not form a field. Mostly we write + ∞ = ∞ .
We make following conventions
x x
i) If x is real then x + ∞ = ∞ , x − ∞ = − ∞ , = =0
∞ −∞
ii) If x > 0 then x (∞) = ∞ , x(− ∞) = − ∞ .
iii) If x < 0 then x (∞) = − ∞ , x(− ∞) = ∞ .
Í Euclidean Space
For each positive integer k, let ¡ k be the set of all ordered k-tuples
x = ( x1 , x2 ,............., xk )
where x1 , x2 ,............, xk are real numbers, called the coordinates of x . The elements
of ¡ k are called points, or vectors, especially when k > 1 .
If y = ( y1 , y2 ,..........., yn ) and α is a real number, put
x + y = ( x1 + y1 , x2 + y2 ,.............., xk + yk )
and α x = (α x1 ,α x2 ,...........,α xk )
So that x + y ∈ ¡ k and α x ∈ ¡ k . These operations make ¡ k into a vector space
over the real field.
The inner product or scalar product of x and y is defined as
k
x . y = ∑ xi yi = ( x1 y1 + x2 y2 + .......... + xk yk )
i =1
And the norm of x is defined by
1
 k 2 2
x = ( x ⋅ x ) =  ∑ xi 
1
2
 1 
The vector space ¡ with the above inner product and norm is called
k
Euclidean k-space. q
š…………………..…………..›
Í Theorem
Let x , y ∈ ¡ n then
= x⋅x
2
i) x
ii) x ⋅ y ≤ x y (Cauchy-Schwarz’s inequality)
Proof
1
i) Since x = ( x ⋅ x ) therefore x = x⋅x
2 2
ii) For λ ∈ ¡ we have

( )( )
2
0≤ x−λy = x−λy ⋅ x−λy
= x ⋅ ( x − λ y ) + ( −λ y ) ⋅ ( x − λ y )
= x ⋅ x + x ⋅ (−λ y ) + (−λ y ) ⋅ x + (−λ y ) ⋅ (−λ y )
2
= x − 2λ ( x ⋅ y ) + λ 2 y
2
x⋅ y
Now put λ = 2 (certain real number)
y
( x ⋅ y )( x ⋅ y ) + ( x ⋅ y ) (x ⋅ y)
2 2
2
⇒ 0≤ x −2 ⇒ 0≤ x −
2 2
2 4
y 2
y y y
2 2
⇒ 0≤ x − x⋅ y
2
y
⇒ 0≤ ( x y + x⋅ y )( x y − x⋅ y )
Which hold if and only if
0≤ x y − x⋅ y
i.e. x ⋅ y ≤ x y q
Í Question
Suppose x , y , z ∈ ¡ n the prove that
a) x + y ≤ x + y
b) x − z ≤ x − y + y − z
Proof
( )( )
2
a) Consider x+ y = x+ y ⋅ x+ y
= x⋅x + x⋅ y + y⋅x + y⋅ y
( )
2
= x + 2 x⋅ y + y
2
2
≤ x +2 x y + y Q x
2
y ≥ x. y
( )
2
= x + y
⇒ x+ y ≤ x + y …………. (i)
b) We have x−z = x− y+ y−z
≤ x− y + y−z from (i) q
Í Question
If r is rational and x is irrational then prove that r + x and r x are irrational.
Proof
Let r + x be rational.
a
⇒ r+x= where a , b ∈ ¢ , b ≠ 0 such that ( a, b ) = 1
b
a
⇒ x= −r
b
c
Since r is rational therefore r = where c, d ∈¢ , d ≠ 0 such that ( c, d ) = 1
d
a c ad − bc
⇒ x= − ⇒ x=
b d bd
Which is rational, which can not happened because x is given to be irrational.
Similarly let us suppose that r x is rational then
a
rx= for some a , b ∈ ¢ , b ≠ 0 such that ( a, b ) = 1
b
a 1
⇒ x= ⋅
b r
c
Since r is rational therefore r = where c, d ∈¢ , d ≠ 0 such that ( c, d ) = 1
d
a 1 a d ad
⇒ x= ⋅ = ⋅ =
b c b c bc
d
Which shows that x is rational, which is again contradiction; hence we conclude
that r + x and r x are irrational. q
Í Question
If n is a positive integer which is not perfect square then prove that n is irrational
number.
Solution
There will be two cases
Case I. When n contain no square factor greater then 1.
Let us suppose that n is a rational number.
p
⇒ n= where p, q ∈¢ , q ≠ 0 and ( p, q ) = 1
q
p2
⇒ n = 2 ⇒ p 2 = nq 2 ...............(i )
q
p2
⇒ q = 2
n
⇒ n p 2 ⇒ n p ................(ii) ( n p means “ n divides p” )
p
Now suppose = c where c∈ ¢
n
⇒ p = nc ⇒ p 2 = n2c 2
Putting this value of p 2 in equation (i)
n 2c 2 = nq 2
q2
⇒ nc = q ⇒ c =
2 2 2
n
⇒ n q ⇒ n q .................(iii)
2
From (ii) and (iii) we get p and q both have common factor n i.e. ( p, q ) = n
Which is a contradiction.
Hence our supposition is wrong.
Case II When n contain a square factor greater then 1.
Let us suppose n = k 2 m > 1
⇒ n =k m
Where k is rational and m is irrational because m has no square factor greater than
one, this implies n , the product of rational and irrational, is irrational. q
Í Question
Prove that 12 is irrational.
Proof
Suppose 12 is rational.
p
⇒ 12 = where p, q ∈¢ , q ≠ 0 and ( p, q ) = 1
q
p2
⇒ 12 = 2 ⇒ p 2 = 12q 2 ………….. (i)
q
p2 p2
⇒ q2 = ⇒ q2 = 2
12 2 ⋅3
⇒ 2 p and 3 p
2 2 2
⇒ 2 p and 3 p
⇒ 2 and 3 are prime divisor of p.
⇒ 2 ⋅ 3 p i.e. 6 p
p
⇒ = c , where c is an integer.
6
⇒ p = 6c
Put this value of p in equation (i) to get
36 c 2 = 12 q 2
q2
⇒ 3c 2 = q 2 ⇒ c 2 =
3
⇒ 3q 2
⇒ 3q
⇒ ( p, q ) = 3 , which is a contradiction.
Hence 12 is an irrational number. q
Í Question
Let E be a non-empty subset of an ordered set, suppose α is a lower bound of E
and β is an upper bound then prove that α ≤ β .
Proof
Since E is a subset of an ordered set S i.e. E ⊆ S .
Also α is a lower bound of E therefore by definition of lower bound
α ≤ x ∀ x ∈ E …………… (i)
Since β is an upper bound of E therefore by the definition of upper bound
x ≤ β ∀ x ∈ E …………… (ii)
Combining (i) and (ii)
α ≤ x≤β
⇒ α ≤ β as required. q
References: (1) Lectures (2003-04)

Prof. Syed Gull Shah
Chairman, Department of Mathematics.
University of Sargodha, Sargodha.
(2) Book
Principles of Mathematical Analysis
Walter Rudin (McGraw-Hill, Inc.)
Collected and Composed by: Atiq ur Rehman (mathcity@gmail.com)

Available online at http://www.mathcity.org in PDF Format.
Page Setup used Legal ( 8′′ 1 2 × 14′′ )
Printed: February 20, 2004. Updated: October 20, 2005
Chapter 2 – Sequences and Series
Subject: Real Analysis Level: M.Sc.
Source: Syed Gul Shah (Chairman, Department of Mathematics, UoS Sargodha)
Collected & Composed by: Atiq ur Rehman(mathcity@gmail.com), http://www.mathcity.org
Sequence
A sequence is a function whose domain of definition is the set of natural
numbers.
Or it can also be defined as an ordered set.
Notation:
An infinite sequence is denoted as
∞
{S n } or {Sn : n ∈ ¥} or {S1, S2 , S3 ,...........} or simply as {Sn }
n=1
e.g. i) {n} = {1, 2,3,..........}

1  1 1 
ii)   = 1, , ,...............
n  2 3 
iii) {(−1) } = {1, −1,1, −1,.............}
n+1
Subsequence
It is a sequence whose terms are contained in given sequence.
∞ ∞
A subsequence of {S n } is usually written as {S nk } .
n=1
Increasing Sequence
A sequence {Sn } is said to be an increasing sequence if S n +1 ≥ S n ∀ n ≥ 1 .
Decreasing Sequence
A sequence {Sn } is said to be an decreasing sequence if S n +1 ≤ S n ∀ n ≥ 1 .
Monotonic Sequence
A sequence {Sn } is said to be monotonic sequence if it is either increasing or
decreasing.
S
{Sn } is monotonically increasing if Sn+1 − Sn ≥ 0 or n+1 ≥ 1, ∀ n ≥ 1
Sn
S
{Sn } is monotonically decreasing if Sn − Sn+1 ≥ 0 or n ≥ 1, ∀ n ≥ 1
S n+1
Strictly Increasing or Decreasing
{Sn } is called strictly increasing or decreasing according as
S n+1 > S n or S n+1 < S n ∀ n ≥ 1 .
Bernoulli’s Inequality
Let p ∈ ¡ , p ≥ −1 and p ≠ 0 then for n ≥ 2 we have
(1 + p ) > 1 + np
n
Proof:
We shall use mathematical induction to prove this inequality.
If n = 2
L.H.S = (1 + p) 2 = 1 + 2 p + p 2
R.H.S = 1 + 2 p
⇒ L.H .S > R.H .S
Sequences and Series -2-
i.e. condition I of mathematical induction is satisfied.
Suppose (1 + p ) > 1 + kp .................. (i ) where k ≥ 2

k
(1 + p ) = (1 + p )(1 + p )
k +1 k
Now
> (1 + p )(1 + kp ) using (i)
= 1 + kp + p + kp 2
= 1 + (k + 1) p + kp 2
≥ 1 + (k + 1) p ignoring kp 2 ≥ 0
⇒ (1 + p ) > 1 + ( k + 1) p
k +1
Since the truth for n = k implies the truth for n = k + 1 therefore condition II of
mathematical induction is satisfied. Hence we conclude that (1 + p ) > 1 + np .
n
Example
n
 1
Let S n = 1 +  where n ≥ 1
 n
−1
To prove that this sequence is an increasing sequence, we use p = , n ≥ 2 in
n2
Bernoulli’s inequality to have
n
 1  n
1 − 2  > 1 − 2
 n  n
n
  1  1   1
⇒  1 − 1 +   > 1 −
  n  n   n
1− n 1− n n−1 n −1
 n −1
n
 1  1  n   1 
⇒ 1 +  > 1 −  =   =  = 1 + 
 n  n  n   n −1  n −1 
⇒ Sn > S n −1 ∀ n ≥1
which shows that {Sn } is increasing sequence.
Example
n+1
 1
Let tn = 1 +  ; n ≥1
 n
then the sequence is decreasing sequence.
1
We use p = 2 in Bernoulli’s inequality.
n −1
n
 1  n
1 + 2  >1+ 2 ……….. (i)
 n −1 n −1
where
1 n2  n  n 
1+ 2 = 2 =  
n − 1 n − 1  n − 1  n + 1 
 1  n + 1   n 
⇒ 1 + 2  =  …………… (ii)
 n − 1  n   n − 1 
n n
 1   n 
Now tn−1 = 1 +  = 
 n − 1   n −1
n
 1  n + 1  
=  1 + 2   from (ii)
  n − 1  n  
1   n +1
n n

= 1 + 2   
 n −1  n 
n  n + 1 
n

> 1 + 2   from (i)
 n − 1  n 
 1  n + 1 
n
n n 1
> 1 +   Q 2 > 2=
 n  n  n −1 n n
n+1
 n +1
=  = tn
 n 
i.e. tn−1 > t n
Hence the given sequence is decreasing sequence.
Bounded Sequence
A sequence {Sn } is said to be bounded if there exists a positive real number λ
such that Sn < λ ∀ n ∈ ¥
If S and s are the supremum and infimum of elements forming the bounded
sequence {Sn } we write S = sup S n and s = inf S n
All the elements of the sequence S n such that Sn < λ ∀ n ∈ ¥ lie with in the
strip { y : − λ < y < λ} . But the elements of the unbounded sequence can not be
contained in any strip of a finite width.
Examples
 (−1)n 
(i) {U n} =   is a bounded sequence
 n 
(ii) {Vn } = {sin nx} is also bounded sequence. Its supremum is 1 and infimum is −1 .
(iii) The geometric sequence {ar n −1} , r > 1 is an unbounded above sequence. It is
bounded below by a.
 nπ 
(iv)  tan  is an unbounded sequence.
 2 
Convergence of the Sequence
A sequence {Sn } of real numbers is said to convergent to limit ‘s’ as n → ∞ , if
for every positive real number ε > 0 , however small, there exists a positive integer n0 ,
depending uponε , such that Sn − s < ε ∀ n > n0 .
Theorem
A convergent sequence of real number has one and only one limit (i.e. Limit of
the sequence is unique.)
Proof:
Suppose {Sn } converges to two limits s and t, where s ≠ t .
s−t
Put ε = then there exits two positive integers n1 and n2 such that
2
Sn − s < ε ∀ n > n1
and Sn − t < ε ∀ n > n2
⇒ Sn − s < ε and Sn − t < ε hold simultaneously ∀ n > max(n1 , n2 ) .
Thus for all n > max(n1 , n2 ) we have
s − t = s − S n + Sn − t
≤ Sn − s + S n − t
< ε + ε = 2ε
 s−t 
⇒ s − t < 2 
 2 
⇒ s −t < s −t
Which is impossible, therefore the limit of the sequence is unique.
Note: If {Sn } converges to s then all of its infinite subsequence converge to s.
Cauchy Sequence
A sequence { xn } of real number is said to be a Cauchy sequence if for given
positive real number ε , ∃ a positive integer n0 (ε ) such that
xn − xm < ε ∀ m, n > n0
Theorem
A Cauchy sequence of real numbers is bounded.
Proof
Let {Sn } be a Cauchy sequence.
Take ε = 1 , then there exits a positive integers n0 such that
Sn − Sm < 1 ∀ m, n > n0 .
Fix m = n0 + 1 then
Sn = S n − S n0 +1 + S n0 +1
≤ S n − S n0 +1 + S n0 +1
< 1 + S n0 +1 ∀ n > n0
<λ ∀ n > 1 , and λ = 1 + S n0 +1 ( n0 changes as ε changes)
Hence we conclude that {Sn } is a Cauchy sequence, which is bounded one.
Note:
(i) Convergent sequence is bounded.
(ii) The converse of the above theorem does not hold.
i.e. every bounded sequence is not Cauchy.
Consider the sequence {Sn } where Sn = (−1)n , n ≥ 1 . It is bounded sequence because
(−1) n = 1 < 2 ∀ n ≥1
But it is not a Cauchy sequence if it is then for ε = 1 we should be able to find a
positive integer n0 such that Sn − Sm < 1 for all m, n > n0
But with m = 2k + 1 , n = 2k + 2 when 2 k + 1 > n0 , we arrive at
Sn − S m = (−1)2 n+2 − (−1) 2 k +1
= 1 + 1 = 2 < 1 is absurd.
Hence {Sn } is not a Cauchy sequence. Also this sequence is not a convergent
sequence. (it is an oscillatory sequence)
……………………………
Divergent Sequence
A {Sn } is said to be divergent if it is not convergent or it is unbounded.
e.g. { n } is divergent, it is unbounded.
2
(ii) {(−1) } tends to 1 or -1 according as n is even or odd. It oscillates finitely.

n
(iii) {(−1) n} is a divergent sequence. It oscillates infinitely.

n
Note: If two subsequence of a sequence converges to two different limits then the
sequence itself is a divergent.
Theorem
If S n < U n < t n ∀ n ≥ n0 and if both the {Sn } and {tn } converge to same limits as
s, then the sequence {U n } also converges to s.
Proof
Since the sequence {Sn } and {tn } converge to the same limit s, therefore, for
given ε > 0 there exists two positive integers n1 , n2 > n0 such that
Sn − s < ε ∀ n > n1
tn − s < ε ∀ n > n2
i.e. s − ε < S n < s + ε ∀ n > n1
s − ε < tn < s + ε ∀ n > n2
Since we have given
Sn n0
∴ s − ε < Sn max(n0 , n1 , n2 )
⇒ s − ε < Un < s + ε ∀ n > max(n0 , n1 , n2 )
i.e. Un − s < ε ∀ n > max(n0 , n1 , n2 )
i.e. lim U n = s
n→∞
Example
1
Show that lim n = 1 n
n→∞
Solution
Using Bernoulli’s Inequality
n
 1  n
1 +  ≥1+ ≥ n ≥1 ∀ n.
 n n
Also
2
1   1  
2 n

( n)
n 2 1
+ =  +   > > n ≥1
n n
 1   1
 n   n  
2
 1 
1
⇒ 1 ≤ n < 1 +
n

 n
2
 1 
1
⇒ lim1 ≤ lim n < lim 1 + 
n
n→∞ n→∞ n→∞
 n
1
⇒ 1 ≤ lim n < 1 n
n→∞
1
i.e. lim n = 1 . n
n→∞
……………………..
Example
 1 1 1 
Show that lim  + + ............ + =0
n→∞ ( n + 1)

2
(n + 2) 2
(2 n)2 
Solution
We have
 1 1 1 
Sn =  + + ............ + 
 (n + 1) (n + 2)2
2
(2n) 2 
and
n n
< S n <
(2n) 2 n2
1 1
⇒ < Sn <
4n n
1 1
⇒ lim < lim S n < lim
n→∞ 4n n→∞ n→∞ n
⇒ 0 < lim S n < 0

n→∞
⇒ lim S n = 0
n→∞
Theorem
If the sequence {Sn } converges to s then ∃ a positive integer n
1
such that Sn > s .
2
Proof
1
We fix ε = s > 0
2
⇒ ∃ a positive integer n1 such that
Sn − s < ε for n > n1
1
⇒ Sn − s < s
2
Now
1 1
s = s − s
2 2
< s − Sn − s ≤ s + ( S n − s )
1
⇒ s < Sn
2
Theorem
Let a and b be fixed real numbers if {Sn } and {tn } converge to s and t
respectively, then
(i) {aSn + btn } converges to as + bt.
(ii) {Sn tn } converges to st.
S  s
(iii)  n  converges to , provided tn ≠ 0 ∀ n and t ≠ 0 .
 tn  t
Proof
Since {Sn } and {tn } converge to s and t respectively,
∴ Sn − s < ε ∀ n > n1 ∈ ¥
tn − t < ε ∀ n > n2 ∈ ¥
Also ∃ λ > 0 such that Sn < λ ∀ n > 1 ( Q {Sn } is bounded )
(i) We have
( aSn + btn ) − ( as + bt ) = a ( Sn − s ) + b ( tn − t )
≤ a ( S n − s ) + b ( tn − t )
< a ε+ bε ∀ n > max (n1 , n2 )
= ε1 Where ε1 = a ε + b ε a certain number.
This implies {aSn + btn } converges to as + bt.
(ii) Sn tn − st = Sn tn − Sn t + Sn t − s t
= S n ( tn − t ) + t ( S n − s ) ≤ S n ⋅ ( tn − t ) + t ⋅ ( S n − s )
< λε + t ε ∀ n > max (n1 , n2 )
= ε2 where ε 2 = λ ε + t ε a certain number.
This implies {Sn tn } converges to st.
1 1 t − tn
(iii) − =
tn t tn t
tn − t ε 1
= < 1 ∀ n > max (n1 , n2 ) Q tn > t
tn t 2 t t 2
ε ε
= = ε3 where ε 3 = a certain number.
1 t 2 1 t
2
2 2
1 1
This implies   converges to .
 tn  t
S   1 1 s
Hence  n  =  Sn ⋅  converges to s ⋅ = . ( from (ii) )
 tn   tn  t t
Theorem
For each irrational number x, there exists a sequence {rn } of distinct rational
numbers such that lim rn = x .
n→∞
Proof
Since x and x + 1 are two different real numbers
Q ∃ a rational number r1 such that
x < r1 < x + 1
Similarly ∃ a rational number r2 ≠ r1 such that
 1
x < r2 < min  r1 , x +  < x + 1
 2
Continuing in this manner we have
 1
x < r3 < min  r2 , x +  < x + 1
 3
 1
x < r4 < min  r3 , x +  < x + 1
 4
……………………………..........
……………………………..........
……………………………..........
 1
x < rn < min  rn−1 , x +  < x + 1
 n
This implies that ∃ a sequence {rn } of the distinct rational number such that
1 1
x − < x < rn < x +
n n
Since
 1  1
lim  x −  = lim  x +  = x
n→∞
 n  n→∞  n
Therefore
lim rn = x
n→∞
Theorem
Let a sequence {Sn } be a bounded sequence.
(i) If {Sn } is monotonically increasing then it converges to its supremum.
(ii) If {Sn } is monotonically decreasing then it converges to its infimum.
Proof
Let S = sup S n and s = inf S n
Take ε > 0
(i) Since S = sup S n
∴ ∃ Sn0 such that S − ε < Sn0
Since {Sn } is ↑ ( ↑ stands for monotonically increasing )
∴ S − ε < Sn0 < Sn < S < S + ε for n > n0
⇒ S − ε < Sn < S + ε for n > n0
⇒ Sn − S < ε for n > n0
⇒ lim S n = S
n→∞
(ii) Since s = inf S n
∴ ∃ Sn1 such that Sn1 < s + ε
Since {Sn } is ↓ . ( ↓ stands for monotonically decreasing )
∴ s − ε < s < Sn < Sn1 < s + ε for n > n1
⇒ s − ε < Sn < s + ε for n > n1
⇒ Sn − s < ε for n > n1
Thus lim S n = s
n→∞
Note
A monotonic sequence can not oscillate infinitely.
Example:
 1 n 
Consider {S n } = 1 +  
 n  
As shown earlier it is an increasing sequence
2n
 1 
Take S 2 n = 1 + 
 2n 
n
 1 
Then S 2 n = 1 + 
 2n 
n n
1  2n  1  1 
⇒ =  ⇒ = 1 − 
S 2 n  2n + 1  S 2 n  2n + 1 
Using Bernoulli’s Inequality we have
n
Q  1 −
1 n n 1 1  n
⇒ ≥ 1− > 1− =  ≥ 1−
S2 n 2n + 1 2n 2  2n + 1  2n + 1
⇒ S2n < 2 ∀ n = 1,2,3,..........

⇒ S2n < 4 ∀ n = 1,2,3,..........
⇒ Sn < S2 n < 4 ∀ n = 1,2,3,..........
Which show that the sequence {Sn } is bounded one.
Hence {Sn } is a convergent sequence the number to which it converges is its
supremum, which is denoted by ‘e’ and 2 < e < 3 .
Recurrence Relation
A sequence is said to be defined recursively or by recurrence relation if the
general term is given as a relation of its preceding and succeeding terms in the sequence
together with some initial condition.
Example
1
Let t1 > 0 and let {tn } be defined by tn+1 > 2 − ; n ≥1
tn
⇒ tn > 0 ∀ n ≥1
1
Also tn − tn+1 = tn − 2 +
tn
( t − 1) > 0
2
t 2 − 2tn + 1
= n = n
tn tn
⇒ tn > tn+1 ∀ n ≥1
This implies that tn is monotonically decreasing.
Since tn > 1 ∀ n ≥1
⇒ t n is bounded below ⇒ tn is convergent.
Let us suppose lim t n = t
n→∞
Then lim tn+1 = lim tn
n→∞ n→∞
 1
⇒ lim  2 −  = lim tn
n→∞ t n  n→∞

1 2t − 1
⇒ 2− =t ⇒ =t ⇒ 2t − 1 = t 2 ⇒ t 2 − 2t + 1 = 0
t t
⇒ ( t − 1) = 0 ⇒ t =1
2
Example
Let {Sn } be defined by S n +1 = S n + b ; n ≥ 1 and S1 = a > b .
It is clear that S n > 0 ∀ n ≥ 1 and S 2 > S1 and
Sn2+1 − Sn2 = ( Sn + b ) − ( Sn−1 + b )
= S n − S n−1
⇒ ( Sn+1 + Sn )( Sn +1 − Sn ) = Sn − Sn−1
S − S n −1
⇒ S n+1 − Sn = n
S n+1 + S n
Since S n +1 + S n > 0 ∀ n ≥ 1
Therefore S n +1 − S n and S n − S n−1 have the same sign.
i.e. S n+1 > S n if and only if S n > S n−1 and
S n+1 < S n if and only if S n < S n−1 .
Sequences and Series - 10 -
But we know that S 2 > S1 therefore S3 > S 2 , S 4 > S3 , and so on.

This implies the sequence is an increasing sequence.
( )
2
Also S n2+1 − S n2 = Sn + b − S n2 = Sn + b − S n2
= − ( S n2 − S n − b )
Since S n > 0 ∀ n ≥ 1 , therefore S n is the root (+ive) of the
Sn2 − Sn − b = 0
1 + 1 + 4b
Take this value of S n as α where α = For equation ax 2 + bx + c = 0
2
−b The product of roots is αβ = c
the other root of equation is therefore a
α c
Since S n+1 > S n ∀ n ≥ 1 i.e. the other root β =
aα
 b
Also − ( S n − α )  S n +  = Sn2+1 − S n2 > 0
 α
b
∴ S n + > 0 or − ( Sn − α ) ≥ 0
α
⇒ Sn < α ∀ n ≥ 1
which shows that S n is bounded and hence it is convergent.
Suppose lim S n = s
n→∞
Then lim( S n+1 )2 = lim( S n + b)

n→∞ n→∞
⇒ s = s+b2
⇒ s2 − s − b = 0
1 + 1 + 4b
Which shows that α = is the limit of the sequence.
2
Theorem
Every Cauchy sequence of real numbers has a convergent subsequence.
Proof
Suppose {Sn } is a Cauchy sequence.
Let ε > 0 then ∃ a positive integer n0 ≥ 1 such that
ε
Snk − Snk −1 < k ∀ nk , nk −1 , k = 1,2,3,.........
2
Put ( ) ( ) (
bk = S n1 − S n0 + S n2 − S n1 + ............ + S nk − S nk −1 )
⇒ bk = (S n1 − S n0 ) + ( S − S ) + ............ + ( S − S )
n2 n1 nk nk −1
≤ (S n1 − S n0 ) + ( S − S ) + ............ + ( S − S )
n2 n1 nk nk −1
ε ε ε
< + 2 + ............ + k
2 2 2
1 1 1  1 1 − 1k
= ε  + 2 + ............ + k  = ε 
2 2 ( )  = ε 1 − 1 
 
2 2 2   1− 2 1   2k 
 
⇒ bk < ε ∀ k ≥1
⇒ {bk } is convergent
Q bk = Snk − Sn0 ∴ Snk = bk + Sn0
Where Sn0 is a certain fix number therefore S nk { } which is a subsequence of {S } is
n
convergent.
Theorem (Cauchy’s General Principle for Convergence)

A sequence of real number is convergent if and only if it is a Cauchy sequence.
Proof
Necessary Condition
Let {Sn } be a convergent sequence, which converges to s .
Then for given ε > 0 ∃ a positive integer n0 , such that
ε
Sn − s < ∀ n > n0
2
Now for n > m > n0
Sn − S m = S n − s + S m − s
≤ Sn − s + Sm − s
ε ε
< + =ε
2 2
Which shows that {Sn } is a Cauchy sequence.
Sufficient Condition
Let us suppose that {Sn } is a Cauchy sequence then for ε > 0 , ∃ a positive
integer m1 such that
ε
Sn − S m < ∀ n, m > m1 ……….. (i)
2
Since {Sn } is a Cauchy sequence
therefore it has a subsequence S nk{ } converging to s (say).
⇒ ∃ a positive integer m2 such that
ε
S nk − s < ∀ n > m2 ……….. (ii)
2
Now
Sn − s = S n − S nk + S nk − s
≤ S n − S nk + S nk − s
ε ε
<+ =ε ∀ n > max(m1 , m2 )
2 2
which shows that {Sn } is a convergent sequence.
Example
Let {Sn } be define by 0 < a < S1 < S 2 2 ………. (i)
Here S n > 0 , ∀ n ≥ 1 and a < S1 2
a < Sk 
Sn +1 b
S a
∴ n +1 > +1
Sn +1 b
S n + S n+1 a+b
⇒ >
S n +1 b
S + S n+1 a+b
⇒ n > S n +1 is replace by S n ∴ S n < S n+1
Sn b
And Sn2+1 − Sn2 = Sn ⋅ Sn−1 − Sn2 Q Sn +1 = Sn S n−1
= Sn ( Sn−1 − Sn )
Sn
⇒ S n+1 − Sn = S n−1 − S n
S n + S n+1
b
< S n−1 − S n
a+b
b
⇒ S n+1 − Sn < Sn − S n−1 Q S n−1 − S n = S n − S n−1
a+b
2
 b 
<  S n−1 − S n−2
a+b
3
 b 
<  S n − 2 − S n −3
a+b
………………………
………………………
………………………
n −1
 b 
<  (b − a )
a+b
b
Take r = <1
a+b
Then for n > m we have
Sn − Sm = Sn − Sn −1 + Sn −1 − Sn −2 + .............. + Sm+1 − Sm
≤ Sn − Sn−1 + Sn−1 − Sn−2 + .............. + Sm+1 − Sm
< ( r n −2 + r n −3 + ............... + r m−1 ) (b − a )
=ε
This implies that {Sn } is a Cauchy sequence, therefore it is convergent.
Example
Let {tn } be defined by
1 1 1
tn = 1 + + + ............... +
2 3 n
For m, n ∈ ¥ , n > m we have
1 1 1
tn − tm = + + ............. +
m +1 m + 2 n
1 m
> ( n − m) = 1 −
n n
In particular if n = 2m then
1
tn − tm >
2
This implies that {tn } is not a Cauchy sequence therefore it is divergent.
7⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅8
Theorem (nested intervals)

Suppose that {I n } is a sequence of the closed interval such that I n = [ an , bn ] ,
I n+1 ⊂ I n ∀ n ≥ 1 , and ( bn − an ) → 0 as n → ∞ then I I n contains one and only one
point.
Proof
Since I n+1 ⊂ I n
∴ a1 < a2 < a3 < ............. < an−1 < an < bn < bn−1 < .......... < b3 < b2 < b1
{an } is increasing sequence, bounded above by b1 and bounded below by a1 .
And {bn } is decreasing sequence bounded below by a1 and bounded above by b1 .
⇒ {an } and {bn } both are convergent.
Suppose {an } converges to a and {bn } converges to b.
But a − b = a − an + an − bn + bn − b
≤ an − a + an − bn + bn − b → 0 as n → ∞ .
⇒ a=b
and an < a < bn ∀ n ≥ 1 .
Theorem (Bolzano-Weierstrass theorem)
Every bounded sequence has a convergent subsequence.
Proof
Let {Sn } be a bounded sequence.
Take a1 = inf S n and b1 = sup S n
Then a1 < S n < b1 ∀ n ≥ 1 .
Now bisect interval [ a1 , b1 ] such that at least one of the two sub-intervals contains
infinite numbers of terms of the sequence.
Denote this sub-interval by [ a2 , b2 ] .
If both the sub-intervals contain infinite number of terms of the sequence then choose
the one on the right hand.
Then clearly a1 ≤ a2 < b2 ≤ b1 .
Suppose there exist a subinterval [ ak , bk ] such that
a1 ≤ a2 ≤ ........... ≤ ak < bk ≤ ........... ≤ b2 ≤ b1
1
⇒ ( bk − ak ) = k ( b1 − a1 )
2
Bisect the interval [ ak , bk ] in the same manner and choose [ ak +1 , bk +1 ] to have
a1 ≤ a2 ≤ ........... ≤ ak ≤ ak +1 < bk +1 ≤ bk ≤ ........... ≤ b2 ≤ b1
1
and bk +1 − ak +1 = k +1 ( b1 − a1 )
2
This implies that we obtain a sequence of interval [ an , bn ] such that
1
bn − an = n (b1 − a1 ) → 0 as n → ∞ .
2
⇒ we have a unique point s such that
s = I [ an , bn ]
there are infinitely many terms of the sequence whose length is ε > 0 that contain s.
For ε = 1 there are infinitely many values of n such that
Sn − s < 1
Let n1 be one of such value then
S n1 − s < 1
Again choose n2 > n1 such that

1
S n2 − s <
2
Continuing in this manner we find a sequence S nk { } for each positive integer k such
that nk < nk +1 and
1
S nk − s < ∀ k = 1, 2,3,............
k
Hence there is a subsequence S nk { } which converges to s.
Limit Inferior of the sequence

Suppose {Sn } is bounded then we define limit inferior of {Sn } as follow
lim ( inf S n ) = lim U k where U k = inf {Sn : n ≥ k }
n→∞ n→∞
If S n is bounded below then
lim ( inf S n ) = − ∞
n→∞
Limit Superior of the sequence

Suppose {Sn } is bounded above then we define limit superior of {Sn } as follow
lim ( sup S n ) = lim Vk where Vk = inf {Sn : n ≥ k }
n→∞ n→∞
If S n is not bounded above then we have
lim ( sup S n ) = + ∞
n→∞
Note:
(i) A bounded sequence has unique limit inferior and superior
(ii) Let {Sn } contains all the rational numbers, then every real number is a
subsequencial limit then limit superior of S n is + ∞ and limit inferior of S n is −∞
 1
(iii) Let {S n } = (−1)n 1 + 
 n
then limit superior of S n is 1 and limit inferior of S n is −1 .
(iv) Let U k = inf {Sn : n ≥ k }
 1   1   1  
= inf 1 +  cos kπ , 1 +  cos( k + 1)π ,  1 +  cos(k + 2)π ,................
 k   k +1  k +2 
 1 
 1 + k  cos kπ if k is odd
=  
1 + 1  cos(k + 1)π if k is even
 k + 1 
⇒ lim ( inf Sn ) = lim U k = −1
n→∞ n→∞
Also Vk = sup{Sn : n ≥ k }
 1 
 k + 1  cos(k + 1)π
1 + if k is odd
=  
1 + 1  cos kπ if k is even
 k 
⇒ lim ( inf Sn ) = lim Vk = 1
n→∞ n→∞
ï⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅ð
Theorem
If {Sn } is a convergent sequence then
lim S n = lim ( inf S n ) = lim ( sup S n )
n→∞ n→∞ n→∞
Proof
Let lim S n = s then for a real number ε > 0 , ∃ a positive integer n0 such that
n→∞
Sn − s < ε ∀ n ≥ n0 ……….. (i)

i.e. s − ε < Sn < s + ε ∀ n ≥ n0
If Vk = sup{Sn : n ≥ k }
Then s − ε < Vn < s + ε ∀ k ≥ n0
⇒ s − ε < lim Vn < s + ε ∀ k ≥ n0 …………. (ii)
k →∞
from (i) and (ii) we have
s = lim sup {S n }
k →∞
We can have the same result for limit inferior of {Sn } by taking
U k = inf {Sn : n ≥ k }
ï⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅ð
Infinite Series
∞
Given a sequence {an } , we use the notation ∑a
i =1
n or simply ∑an to denotes the
sum a1 + a2 + a3 + .............. and called a infinite series or just series.

n
The numbers S n = ∑ ak are called the partial sum of the series.
k =1
If the sequence {Sn } converges to s, we say that the series converges and write
∞
∑a
n=1
n = s , the number s is called the sum of the series but it should be clearly
understood that the ‘s’ is the limit of the sequence of sums and is not obtained simply
by addition.
If the sequence {Sn } diverges then the series is said to be diverge.
Note:
The behaviors of the series remain unchanged by addition or deletion of the certain
terms
Theorem
∞
If ∑a
n=1
n converges then lim an = 0 .
n→∞
Proof
Let S n = a1 + a2 + a3 + .......... + an
Take lim S n = s = ∑ an
n→∞
Since an = Sn − S n −1
Therefore lim an = lim ( S n − S n−1 )
n→∞ n→∞
= lim S n − lim S n−1
n→∞ n→∞
=s−s=0
Note:
The converse of the above theorem is false
Example
∞
1
Consider the series ∑n.
n=1
1 1 1
We know that the sequence {Sn } where S n = 1 + + + ............. + is divergent
2 3 n
∞
1
therefore ∑n
n=1
is divergent series, although lim an = 0 .
n→∞
This implies that if lim an ≠ 0 , then

n→∞
∑a n is divergent.
It is know as basic divergent test.
Theorem (General Principle of Convergence)
A series ∑ an is convergent if and only if for any real number ε > 0 , there exists a
positive integer n0 such that
∞
∑a
i = m +1
i <ε ∀ n > m > n0
Proof
Let S n = a1 + a2 + a3 + ............. + an
then {Sn } is convergent if and only if for ε > 0 ∃ a positive integer n0 such that
Sn − Sm < ε ∀ n > m > n0

∞
⇒ ∑a
i = m +1
i = Sn − Sm < ε
Example
∞
1
If x < 1 then ∑x
n= 0
n
=
1− x
∞
And if x ≥ 1 then ∑x
n =0
n
is divergent.
Theorem
Let ∑ an be an infinite series of non-negative terms and let {Sn } be a sequence of
its partial sums then ∑ an is convergent if {Sn } is bounded and it diverges if {Sn } is
unbounded.
Proof
Since an ≥ 0 ∀ n ≥ 0
S n = S n−1 + an > S n−1 ∀ n≥0
therefore the sequence {Sn } is monotonic increasing and hence it is converges if
{Sn } is bounded and it will diverge if it is unbounded.
Hence we conclude that ∑a n is convergent if {Sn } is bounded and it divergent if
{Sn } is unbounded.
Theorem (Comparison Test)
Suppose ∑ an and ∑ bn are infinite series such that an > 0 , bn > 0 ∀ n . Also
suppose that for a fixed positive number λ and positive integer k , an < λ bn ∀ n ≥ k
Then ∑a n converges if ∑b n is converges and ∑b n is diverges if ∑a n is diverges.
Proof
Suppose ∑b is convergent and
n
an < λ bn ∀ n ≥ k …………. (i)

then for any positive number ε > 0 there exists n0 such that
n
ε
∑
i = m +1
bi <
λ
n > m > n0
from (i)
n n
∑ ⇒
i =m +1
ai < λ ∑ b <ε
i = m +1
i , n > m > n0
⇒ ∑a n is convergent.
Now suppose ∑ a is divergent then {Sn } is unbounded.
n
⇒ ∃ a real number β > 0 such that

n
∑ b >λβ
i = m +1
i , n>m
from (i)
n
1 n
⇒ ∑ bi > ∑ ai > β , n>m
i =m +1 λ i =m+1
⇒ ∑ bn is convergent.
Example
1
We know that ∑n is divergent and
n≥ n ∀ n ≥1
1 1
⇒ ≤
n n
1 1
⇒ ∑
n
is divergent as ∑n is divergent.
Example
1
The series ∑n α
is convergent if α > 1 and diverges if α ≤ 1 .
1 1 1
Let Sn = 1 + + + .................. +
2α 3α nα
If α > 1 then
1 1
Sn < S2n and <
nα (n − 1)α
 1 1 1 1 
Now S 2 n = 1 + α + α + α ............ +
 2 3 4 (2n)α 
 1 1 1  1 1 1 1 
= 1 + α + α + ............ + + + + + +
(2n − 1)α   2α 4α 6α (2n)α 
............
 3 5
 1 1 1  1  1 1 1 
= 1 + α + α + ............ + α 
+ α 1 + α + α + ............ + α 
 3 5 (2n − 1)  2  2 3 ( n) 
 1 1 1  1
< 1 + α + α + ............ + α 
+ Sn
 2 4 (2n − 2)  2α
replacing 3 by 2, 5 by 4 and so on.
1  1 1  1
=1 + + + + +
2α  2α (n − 1)α  2α
1 ............ Sn
1 1 1 1
=1 + α Sn −1 + α S n =1 + α S 2 n + α S2 n Q Sn −1 < S n < S 2 n
2 2 2 2
2
=1 + α S 2 n
2
1
⇒ S 2 n < 1 + α −1 S 2 n
2
 1   2α −1 − 1  2α −1
⇒ 1 − α −1  S 2 n < 1 ⇒  α −1  S 2 n < 1 ⇒ S 2 n < α −1
 2   2  2 −1
2α −1
i.e. S n < S 2 n < α −1
2 −1
1
⇒ {Sn } is bounded and also monotonic. Hence we conclude that ∑ α is
n
convergent when α > 1 .
If α ≤ 1 then
nα ≤ n ∀ n ≥ 1
1 1
⇒ α ≥ ∀ n ≥1
n n
1 1
Q ∑ is divergent therefore ∑ α is divergent when α ≤ 1.
n n
Theorem
an
Let an > 0 , bn > 0 and lim
n→∞ b
= λ ≠ 0 then the series ∑a n and ∑b n behave alike.
n
Proof
an
Since lim =λ
n→∞ b
n
an
⇒ − λ < ε ∀ n ≥ n0 .
bn
λ
Use ε =
2
a λ
⇒ n −λ < ∀ n ≥ n0 .
bn 2
λ a λ
⇒ λ− < n <λ+
2 bn 2
λ an 3λ
⇒ < <
2 bn 2
then we got
3λ 2
an < bn and bn < an
2 λ
Hence by comparison test we conclude that ∑a n and ∑b n converge or diverge
together.
Example
1 x
To check ∑ n sin n
2
diverges or converges consider
1 x 1
an = sin 2 and take bn = 3
n n n
an x
then = n 2 sin 2
bn n
2
x 2  x
sin  sin 
= n = x2 n
1  x 
 
n2  n 
Applying limit as n → ∞
2 2
 x  x
sin  sin 
an 2 n 2 n = x 2 (1) = x 2
lim = lim x   = x  lim 
n→∞ b n→∞ x n→∞ x
n    
 n   n 
⇒ ∑ an and ∑ bn have the similar behavior ∀ finite values of x except x = 0.
1
Since ∑n 3
is convergent series therefore the given series is also convergent for
finite values of x except x = 0.
ï⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅ð
Theorem ( Cauchy Condensation Test )

Let an ≥ 0 , an > an +1 ∀ n ≥ 1 , then the series ∑ an and ∑2 n −1
a2n −1 converges or
diverges together.
Proof
Let us suppose
S n = a1 + a2 + a3 + ............... + an
and tn = a1 + 2a2 + 2 2 a22 + ................ + 2 n−1 a2n−1 .
Q an ≥ 0 and n < 2n−1 < 2n − 1
∴ Sn < S2n−1 < S2n −1 for n > 2
then
S2n −1 = a1 + a2 + a3 + ...... + a2n −1
(
= a1 + ( a2 + a3 ) + ( a4 + a5 + a6 + a7 ) + ....... + a2n−1 + a2n −1+1 + a2n −1+ 2 + ..... + a2n −1 )
< a1 + ( a2 + a2 ) + ( a4 + a4 + a4 + a ) + ....... + ( a
4 2n −1
+ a2n −1 + a2n−1 + ..... + a2n−1 )
< a1 + 2 a2 + 22 a4 + ....... + 2 n−1 a2n −1 = tn
⇒ S n < tn
⇒ Sn < tn < 2S2n …………. (i)
Now consider
S2n = a1 + a2 + a3 + ............... + a2n
(
= a1 + a2 + ( a3 + a4 ) + ( a5 + a6 + a7 + a8 ) + ....... + a2n−1+1 + a2n−1+ 2 + a2n−1+3 + .... + a2n )
1
(
> a1 + a2 + ( a4 + a4 ) + ( a8 + a8 + a8 + a8 ) + ...... + a2n + a2n + a2n + ..... + a2n
2
)
1
= a1 + a2 + 2 a4 + 2 2 a8 + ................ + 2 n−1 a2n
2
1
(
= a1 + 2 a2 + 2 3a4 + 23 a8 + ................ + 2 n a2n
2
)
1
⇒ S 2 n > t n ………… (ii)
2
⇒ 2 S 2 n > tn
From (i) and (ii) we see that the sequence S n and tn are either both bounded or both
unbounded, implies that ∑a n and ∑2 n −1
a2n −1 converges or diverges together.
Example
1
Consider the series ∑n p
1
If p ≤ 0 then lim ≠0
n→∞ n p
therefore the series diverges when p ≤ 0 .

If p > 0 then the condensation test is applicable and we are lead to the series
∞ ∞
1 1
∑
k =0
2 k p = ∑ kp −k
k
(2 ) k =0 2
∞ ∞ k
1  1 
=∑ = ∑  ( p −1) 
k =0 2( p −1) k k =0  2 
∞
= ∑2
k =0
(1− p ) k
1− p
Now 2 < 1 iff 1 − p < 0 i.e. when p > 1
And the result follows by comparing this series with the geometric series having
common ratio less than one.
The series diverges when 21− p = 1 ( i.e. when p = 1 )
The series is also divergent if 0 1 , p
converges and
n= 2
If p ≤ 1 the series is divergent.

 1 
Q {ln n} is increasing ∴   decreases
 n ln n 
and we can use the condensation test to the above series.
1
We have an =
n ( ln n )
p
1 1
⇒ a2n = ⇒ 2 n a2n =
( ) ( n ln 2 )
p p
2 n ln 2 n
⇒ we have the series
1 1 1
∑ 2n a2n = ∑ (n ln 2) p = ( ln 2 ) p ∑ n p
which converges when p > 1 and diverges when p ≤ 1 .
Example
1
Consider ∑ ln n
 1 
Since {ln n} is increasing there   decreases.
 ln n 
And we can apply the condensation test to check the behavior of the series
1 1
Q an = ∴ a2n =
ln n ln 2 n
2n 2n
so 2 a2n =
n
⇒ 2 a2n =
n
ln 2n n ln 2
n
2 1
since > ∀ n ≥1
n n
1
and ∑ is diverges therefore the given series is also diverges.
n
ï⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅ð
Alternating Series
A series in which successive terms have opposite signs is called an alternating series.
(−1)n +1 1 1 1
e.g. ∑ = 1 − + − + ............... is an alternating series.
n 2 3 4
Theorem (Alternating Series Test or Leibniz Test)
Let {an } be a decreasing sequence of positive numbers such that lim an = 0 then the
n→∞
∞
alternating series ∑ (−1)
n=1
n+1
an = a1 − a2 + a3 − a4 + ................ converges.
Proof
Looking at the odd numbered partial sums of this series we find that
S 2 n +1 = (a1 − a2 ) + (a3 − a4 ) + (a5 − a6 ) + ........... + (a2 n−1 − a2 n ) + a2 n+1
Since {an } is decreasing therefore all the terms in the parenthesis are non-negative
⇒ S 2 n +1 > 0 ∀ n
Moreover
S 2 n +3 = S 2 n+1 − a2 n+2 + a2 n+3
= S2 n +1 − ( a2 n +2 − a2 n+3 )
Since a2 n+ 2 − a2 n +3 ≥ 0 therefore S 2 n +3 ≤ S 2 n +1
Hence the sequence of odd numbered partial sum is decreasing and is bounded below
by zero. (as it has +ive terms)
It is therefore convergent.
Thus S 2 n+1 converges to some limit l (say).
Now consider the even numbered partial sum. We find that
S 2 n +2 = S2 n+1 − a2 n+ 2
and
lim S 2 n+ 2 = lim ( S 2 n+1 − a2 n +2 )
n→∞ n→∞
= lim S 2 n+1 − lim a2 n+2
n→∞ n→∞
=l −0 =l Q lim an = 0
n→∞
so that the even partial sum is also convergent to l .
⇒ both sequences of odd and even partial sums converge to the same limit.
Hence we conclude that the corresponding series is convergent.
Absolute Convergence
∑ an is said to converge absolutely if ∑a n converges.
Theorem
An absolutely convergent series is convergent.
Proof:
If ∑ an is convergent then for a real number ε > 0 , ∃ a positive integer n0 such that
n n
∑a
i = m +1
i < ∑
i = m +1
ai < ε ∀ n, m > n0
⇒ the series ∑a n is convergent. (Cauchy Criterion has been used)

Note
The converse of the above theorem does not hold.
(−1)n +1 1
e.g. ∑ is convergent but ∑ is divergent.
n n
Theorem (The Root Test)

1
Let lim Sup an n
=p
n→∞
Then ∑a n converges absolutely if p < 1 and it diverges if p > 1 .

Proof
Let p < 1 then we can find the positive number ε > 0 such that p + ε < 1
1
⇒ an n
 n0
⇒ an < ( p + ε ) < 1
n
Q ∑ ( p + ε ) is convergent because it is a geometric series with

n
r < 1.
∴ ∑ a is convergent
n
⇒ ∑ a converges absolutely.
n
Now let p > 1 then we can find a number ε1 > 0 such that p − ε1 > 1 .
1
⇒ an n
> p +ε > 1
⇒ an > 1 for infinitely many values of n.
⇒ lim an ≠ 0
n→∞
⇒ ∑a n is divergent.
Note:
The above test give no information when p = 1 .
1 1
e.g. Consider the series ∑ and ∑ 2 .
n n
1 1
For each of these series p = 1 , but ∑ is divergent and ∑n 2
is convergent.
n
Theorem (Ratio Test)
The series ∑ an
an+1
(i) Converges if lim Sup <1
n→∞ an
an +1
(ii) Diverges if > 1 for n ≥ n0 , where n0 is some fixed integer.
an
Proof
If (i) holds we can find β < 1 and integer N such that
an +1
< β for n ≥ N
an
In particular
a N +1
< β
aN
⇒ aN +1 < β aN
⇒ aN +2 < β aN +1 < β 2 aN
⇒ a N +3 < β 3 a N
…………………….
…………………….
…………………….
⇒ a N + p < β p aN
⇒ an < β n − N aN we put N + p = n .
i.e. an < aN β − N β n for n ≥ N .
Q ∑β n
is convergent because it is geometric series with common ration < 1 .
Therefore ∑a n is convergent (by comparison test)
Now if
an+1 ≥ an for n ≥ n0
then lim an ≠ 0
n→∞
⇒ ∑a n is divergent.
Note
an +1
The knowledge = 1 implies nothing about the convergent or divergent of series.
an
Example
−n
 n  n  
n+1
Consider the series ∑ an with an =  −  

 n + 1  n + 1  
n
Q <1 ∴ an > 0 ∀ n .
n +1
−1
1  n  n  
n+1
Also ( an ) n =  −  
 n + 1  n + 1  
−1 −1
 n +1   n    n +1   n +1 
n −n
=   1 −    =   1 −   
 n    n + 1    n    n  
−1
 1   1 
−n
= 1 +  1 − 1 +  
 n    n  
−1
 1   1 
−n
lim an = lim 1 +  1 − 1 +  
n→∞ n→∞
 n    n  
−1
 1   1 − n 
= lim 1 +  lim 1 − 1 +  
n→∞
 n  n→∞   n  
−1 −1
−1 −1  1  e − 1  e 
= 1 ⋅ 1 − e  = 1 −  =  =  >1
 e  e   e − 1 
⇒ the series is divergent.
Theorem (Dirichlet)
Suppose that {Sn } , S n = a1 + a2 + a3 + ............. + an is bounded. Let { bn } be positive
term decreasing sequence such that lim bn = 0 , then
n→∞
∑a n bn is convergent.
Proof
Q {Sn } is bounded
∴ ∃ a positive number λ such that
Sn < λ ∀ n ≥ 1.
Then ai bi = ( Si − Si −1 ) bi for i ≥ 2
= Si bi − Si−1 bi
= Si bi − Si−1 bi + Si bi +1 − Si bi +1
= Si ( bi − bi+1 ) − Si −1 bi + Si bi +1
n n
⇒ ∑
i = m +1
ai bi = ∑ S (b − b ) − ( S
i = m +1
i i i +1 m bm+1 − S n bn +1 )
Q { bn } is decreasing
n n
∴ ∑ab
i = m +1
i i = ∑ S (b − b ) − S
i = m+1
i i i +1 m bm+1 + Sn bn+1
n
< ∑ { S (b − b )} +
i = m +1
i i i +1 S m bm+1 + S n bn +1
n
< ∑ {λ (b − b )} + λ b
i = m +1
i i +1 m +1 + λ bn+1 Q Si < λ
 n 
= λ  ∑ ( bi − bi+1 ) + bm+1 + bn +1 
 i =m+1 
= λ ( ( bm+1 − bn+1 ) + bm+1 + bn +1 ) = 2λ ( bm+1 )
n
⇒ ∑ab
i = m +1
i i <ε where ε = 2λ ( bm+1 ) a certain number
⇒ The ∑a n bn is convergent. ( We have use Cauchy Criterion here. )
Theorem
Suppose that ∑a n is convergent and that { bn } is monotonic convergent sequence
then ∑a n bn is also convergent.
Proof
Suppose { bn } is decreasing and it converges to b .
Put cn = bn − b
⇒ cn ≥ 0 and lim cn = 0
n→∞
Q ∑a n is convergent
∴ {Sn } , S n = a1 + a2 + a3 + ................. + an is convergent
⇒ It is bounded
⇒ ∑ an cn is bounded.
Q an bn = an cn + an b and ∑a n cn and ∑a n b are convergent.
∴ ∑a n bn is convergent.
Now if { bn } is increasing and converges to b then we shall put cn = b − bn .
Example
1
∑ (n ln n) α
is convergent if α > 1 and divergent if α ≤ 1.
To see this we proceed as follows

1
an =
(n ln n)α
2n 2n
Take bn = 2 a2n = n
=
( 2n ln 2n ) (2 n ln 2 )
α n α
2n 1
= =
2nα nα ( ln 2 ) 2nα −n nα ( ln 2 )
α α
(α −1) n
1
 
⋅ α
1 2
=
( ln 2 )
α
n
(α −1) n
1 1
Since ∑ α is convergent when α > 1 and   is decreasing for α > 1 and it
n 2
converges to 0. Therefore ∑ bn is convergent
⇒∑ a is also convergent.
n
Now ∑ b is divergent for α ≤ 1 therefore ∑ a

n n diverges for α ≤ 1.
Example
1
To check ∑n α
ln n
is convergent or divergent.
1
We have an = α
n ln n
2n 2n
Take bn = 2 a2n n
= = nα
(2 n )α (ln 2 n ) 2 (n ln 2)
(α −1) n
1
1 2(1−α ) n 1  2 
= ⋅ = ⋅
ln 2 n ln 2 n
1  1 n (α −1) 
Q ∑ is divergent although    is decreasing, tending to zero for α > 1
n  2  
∑ b is divergent.
therefore n
⇒ ∑ a is divergent.
n
The series also divergent if α ≤ 1.

i.e. it is always divergent.

Prof. Syyed Gull Shah
(2) Book
Made by: Atiq ur Rehman (mathcity@gmail.com)

Page Setup: Legal ( 8′′ 1 2 × 14′′ )
Printed: October 20, 2004. Updated: October 11, 2005
Chapter 3 – Limit and Continuity
Source: Syyed Gul Shah (Chairman, Department of Mathematics, US Sargodha)
Collected & Composed by: Atiq ur Rehman (mathcity@gmail.com), http://www.mathcity.org
v Limit of the function

Suppose
(i) ( X , d X ) and (Y , dY ) be two metric spaces
(ii) E ⊂ X
(iii) f : E → Y i.e. f maps E into Y .
(iv) p is the limit point of E .
We write f ( x ) → q as x → p or lim f ( x ) = q , if there is a point q with the
x→ p
following property;
For every ε > 0 , there exists a δ > 0 such that dY ( f ( x), q ) < ε for all points
x ∈ E for which d X ( x, p ) < δ .
If X and Y are replaced by a real line, complex plane or by Euclidean space ¡ k ,
then the distances d X and dY are replaced by absolute values or by appropriate
norms. q
Note: i) It is to be noted that p ∈ X but that p need not a point of E in the above
definition ( p is a limit point of E which may or may not belong to E .)
ii) Even if p ∈ E , we may have f ( p ) ≠ lim f ( x) . q
x→ p
v Example
2x
lim =2
x →∞ 1 + x
2x 2x − 2 − 2x −2 2
We have −2 = = <
x −1 1+ x 1+ x x
2
Now if ε > 0 is given we can find δ = so that
ε
2x
− 2 < ε whenever x > δ . q
1+ x
v Example
x2 − 1
Consider the function f ( x ) = .
x −1
It is to be noted that f is not defined at x = 1 but if x ≠ 1 and is very close to 1 or
less then f ( x) equals to 2. q
v Definitions
i) Let X and Y be subsets of ¡ , a function f : X → Y is said to tend to limit l
as x → ∞ , if for a real number ε > 0 however small, ∃ a positive number δ which
depends upon ε such that distance
f ( x ) − l < ε when x > δ and we write lim f ( x) = l .
x →∞
ii) f is said to tend to a right limit l as x → c if for ε > 0 , ∃ δ > 0 such that
f ( x ) − l < ε whenever x ∈ G and 0 < x < c + δ .
And we write f (c + ) = lim f ( x) = l
x →c +
iii) f is said to tend to a left limit l as x → c if for ε > 0 , ∃ a δ > 0 such that
f ( x ) − l < ε whenever x ∈ G and 0 < c − δ < x < c .
And we write f (c −) = lim f ( x ) = l . q
x →c −
2
v Theorem
Suppose
(i) ( X , d x ) and (Y , d y ) be two metric spaces
(ii) E ⊂ X
(iii) f : E → Y i.e. f maps E into Y .
(iv) p is the limit point of E .
Then lim f ( x) = q iff lim f ( pn ) = q for every sequence { pn } in E such that
x→ p n→∞
pn ≠ p , lim pn = p .
n→∞
Proof
Suppose lim f ( x) = q holds.
x→ p
Choose { pn } in E such that pn ≠ p , lim pn = p , we are to show that

n→∞
lim f ( pn ) = q
n→∞
Then there exists a δ > 0 such that
d y ( f ( x), q ) < ε if x ∈ E and 0 < d x ( x, p ) < δ ………. (i)
Also ∃ a positive integer n0 such that n > n0
⇒ d x ( pn , p ) < δ ………….. (ii)
from (i) and (ii), we have for n > n0
d y ( f ( pn ), q ) < ε
Which shows that limit of the sequence
lim f ( pn ) = q
n→∞
Conversely, suppose that lim f ( pn ) = q is false.
n→∞
Then ∃ some ε > 0 such that for every δ > 0 , there is a point x ∈ E for which
d y ( f ( x), q ) ≥ ε but 0 < d x ( x, p ) < δ .
1
In particular, taking δ n = , n = 1,2,3,......
n
We find a sequence in E satisfied pn ≠ p , lim pn = p for which lim f ( pn ) = q
n→∞ n→∞
is false. q
v Example
1
lim sin does not exist.
x →∞ x
1
Suppose that lim sin exists and take it to be l, then there exist a positive real
x →∞ x
number δ such that
1
sin − l < 1 when 0 < x − 0 < δ (we take ε = 1 > 0 here)
x
We can find a positive integer n such that
2 2 2
< δ then < δ and <δ
nπ (4n + 1)π (4n + 3)π
It thus follows
(4n + 1)π
sin −l < 1 ⇒ 1− l < 1
2
(4n + 3)π
and sin − l < 1 ⇒ − 1 − l < 1 or 1 + l < 1
2
3
So that
2 = 1+ l +1− l ≤ 1+ l + 1− l < 1+1 ⇒ 2 < 2
This is impossible; hence limit of the function does not exist. q
Alternative:
2
Consider xn = then lim xn = 0
(2n − 1)π x →∞
 1
But { f ( xn )} i.e. sin  is an oscillatory sequence
 xn 
 1
i.e. {1, −1,1, −1,..........} therefore sin  diverges.
 xn 
1
Hence we conclude that lim sin does not exit. q
x →∞ x
v Example
Consider the function
 x ; x <1
f ( x) = 
 2 + ( x − 1) ; x ≥1
2
We show that lim f ( x) does not exist.

x→1
1
To prove this take xn = 1 − , then lim xn = 1 and lim f ( xn ) = 1
n x →∞ n→∞
1
But if we take xn = 1 + then xn → 1 as n → ∞
n
2
 1 
and lim f ( xn ) = lim 2 + 1 + − 1 = 2
x →∞ x→∞
 n 
This show that { f ( xn )} does not tend to a same limit as for all sequences { Sn }
such that xn → 1 .
Hence this limit does not exist. q
v Example
Consider the function f :[0,1] → ¡ defined as
0 if x is rational
f ( x) = 
1 if x is irratioanl
Show that lim f ( x) where p ∈ [0,1] does not exist.
x→ p
Solution
Let lim f ( x ) = q , if given ε > 0 we can find δ > 0 such that
x→ p
f ( x ) − q < ε whenever x − p < δ .

Consider the irrational (r − s, r + s ) ⊂ [0,1] such that r is rational and s is
irrational.
Then f (r ) = 0 & f (s ) = 1
Suppose lim f ( x ) = q then
x→ p
f (s ) = 1
⇒ 1 = f (s) − q + q
= ( f (s ) − q + q − 0
= f (s ) − q + q − f (r ) Q 0 = f (r )
4
≤ f ( s) − q + f (r ) − q < ε + ε
i.e. 1 < ε + ε
1 1 1
⇒ 1< + if ε =
4 4 4
Which is absurd.
Hence the limit of the function does not exist. q
v Exercise
1
lim x sin =0
x →0 x
We have
1
x sin − 0 < ε where ε > 0 is a pre-assigned positive number.
x
1
⇒ x sin < ε
x
1
⇒ x sin < ε
x
1
⇒ x <ε Q sin ≤1
x
⇒ x−0 <ε =δ
1
It shows that lim x sin = 0 .
x →0 x
1
Same the case for function for f ( x ) = x cos
x
1
Also we can derived the result that lim x 2 sin = 0. q
x →0 x
v Theorem
If lim f ( x) exists then it is unique.
x→c
Proof
Suppose lim f ( x) is not unique.
x→c
Take lim f ( x) = l1 and lim f ( x ) = l2 where l1 ≠ l2 .
x →c x →c
⇒ ∃ real numbers δ1 and δ 2 such that
f ( x ) − l1 < ε whenever x − c < δ1
& f ( x ) − l2 < ε whenever x − c < δ2
Now l1 − l2 = ( f ( x) − l1 ) − ( f ( x) − l2 )
≤ f ( x ) − l1 + f ( x) − l2
< ε + ε whenever x − c < min(δ1 ,δ 2 )
⇒ l1 = l2 q
…………………
5
v Theorem
Suppose that a real valued function f is defined on an open interval G except
possibly at c ∈ G . Then lim f ( x) = l if and only if for every positive real number
x→c
ε , there is δ > 0 such that f (t ) − f (s ) < ε whenever s & t are in

{x : x − c < δ }.
Proof
Suppose lim f ( x) = l
x→c
∴ for every ε > 0 , ∃ δ > 0 such that
1
f ( s) − l < ε whenever 0 < s − c < δ
2
1
& f (t ) − l < ε whenever 0 < t − c < δ
2
⇒ f ( s ) − f (t ) ≤ f ( s ) − l + f (t ) − l
ε ε
< + whenever s − c < δ & t − c < δ
2 2
f (t ) − f ( s ) < ε whenever s & t are in { x : x − c < δ } .
Conversely, suppose that the given condition holds.
Let { xn } be a sequence of distinct elements of G such that xn → c as n → ∞ .
Then for δ > 0 ∃ a natural number n0 such that
xn − l < δ and xm − l < δ ∀ m, n > n0 .
And for ε > 0
f ( xn ) − f ( xm ) < ε whenever m, n > n0
⇒ { f ( xn )} is a Cauchy sequence and therefore it is convergent. q
v Theorem (Sandwiching Theorem)

Suppose that f , g and h are functions defined on an open interval G except
possibly at c ∈ G . Let f ≤ h ≤ g on G .
If lim f ( x ) = lim g ( x) = l , then lim h( x ) = l .
x →c x→c x →c
Proof
For ε > 0 ∃ δ 1 ,δ 2 > 0 such that
f ( x ) − l < ε whenever 0 < x − c < δ1
& g ( x) − l < ε whenever 0 < x − c < δ 2
⇒ l − ε < f ( x) < l + ε for 0 < x − c < δ1
& l − ε < g ( x) < l + ε for 0 < x − c < δ2
⇒ l − ε < f ( x ) ≤ h( x ) ≤ g ( x) < l + ε
⇒ l − ε < h( x ) < l + ε for 0 < x − c < min(δ 1 , δ 2 )
⇒ lim h( x) = l q
x →c
……………………….
6
v Theorem
Let (i) ( X , d ) , (Y , d y ) be two metric spaces.
(ii) E ⊂ X
(iii) p is a limit point of E .
(iv) f : E → Y .
(v) g : E → Y
and lim f ( x ) = A and lim g ( x) = B then
x→ p x→ p
i- lim ( f ( x) ± g ( x) ) = A ± B
x→ p
ii- lim( fg )( x) = AB
x→ p
 f ( x)  A
iii- lim  = provided B ≠ 0 .

x → p g ( x)
 B
Proof
Do yourself q
v Continuity
Suppose
i) ( X , d X ) , (Y , dY ) are two metric spaces
ii) E ⊂ X
iii) p ∈ E
iv) f : E → Y
Then f is said to be continuous at p if for every ε > 0 ∃ a δ > 0 such that
dY ( f ( x), f ( p) ) < ε for all points x ∈ E for which d X ( x, p ) < δ .
Note:
(i) If f is continuous at every point of E . Then f is said to be continuous on E .
(ii) It is to be noted that f has to be defined at p iff lim f ( x) = f ( p ) . q
x→ p
v Examples
f ( x) = x 2 is continuous ∀ x ∈ ¡ .
Here f ( x) = x 2 , Take p ∈ ¡
Then f ( x) − f ( p) < ε
⇒ x2 − p 2 < ε
⇒ ( x − p)( x + p) < ε
⇒ x− p < ε =δ
Q p is arbitrary real number
∴ the function f ( x) is continuous ∀ real numbers. q
………………………..
7
v Theorem
Let
i) X ,Y , Z be metric spaces
ii) E ⊂ X
iii) f : E → Y , g : f ( E ) → Z and h : E → Z defined by h( x ) = g ( f ( x) )
If f is continuous at p ∈ E and if g is continuous at the point f ( p) , then h is
continuous at p.
Proof
h
f g
x f(x) h(x) = g(f(x))
p
f(p)
h(p) = g(f(p))
E f g
Y Z
X
h
Q g is continuous at f ( p)
∴ for every ε > 0 , ∃ a δ > 0 such that
d Z ( g ( y ), g ( f ( p ) ) ) < ε whenever dY ( y , f ( p) ) < δ1 ……….. (i)
Q f is continuous at p ∈ E
∴ ∃ a δ > 0 such that
dY ( f ( x ), f ( p) ) < δ 1 whenever d X ( x, p ) < δ ………… (ii)
Combining (i ) and (ii ) , we have
d Z ( g ( y ), g ( f ( p ) ) ) < ε whenever d X ( x, p ) < δ
⇒ d Z ( h( x), h( p) ) < ε whenever d X ( x, p ) < δ
which shows that the function h is continuous at p . q
v Example
(i) f ( x ) = (1 − x 2 ) is continuous ∀ x ∈ ¡ and g ( x) = x is continuous
∀ x ∈ [0, ∞ ] , then g ( f ( x) ) = 1 − x 2 is continuous x ∈ ( −1,1) .
{
(ii) Let g ( x) = sin x and f ( x) = x − π , x ≤ 0
x +π , x > 0
Then g ( f ( x) ) = − sin x ∀ x
Then the function g ( f ( x) ) is continuous at x = 0 , although f is discontinuous
at x = 0 . q
v Theorem
Let f be defined on X . If f is continuous at c ∈ X then ∃ a number δ > 0
such that f is bounded on the open interval (c − δ , c + δ ) .
Proof
Since f is continuous at c ∈ X .
Therefore for a real number ε > 0, ∃ a real number δ > 0 such that
f ( x ) − f (c) < ε whenever x ∈ X and x − c < δ .
⇒ f ( x ) = f ( x ) − f (c ) + f (c )
8
≤ f ( x ) − f (c ) − f (c )
< ε + f (c) whenever x − c < δ .
It shows that f is bounded on the open interval ]c − δ , c + δ [ . q
v Theorem
Suppose f is continuous on [ a, b] . If f (c ) > 0 for some c ∈ [ a, b ] then there
exist an open interval G ⊂ [ a, b] such that f ( x ) > 0 ∀ x ∈ G .
Proof
1
Take ε = f (c )
2
Q f is continuous on [ a, b]
∴ f ( x) − f (c) < ε whenever x − c < δ , x ∈ [ a, b]
Take G = { x ∈ [ a, b ] : x − c < δ }
⇒ f ( x ) = f ( x ) − f (c ) + f (c )
≤ f ( x ) − f (c ) + f (c )
< ε + f (c) whenever x − c < δ
For x ∈ G , we have
f ( x ) = f ( c ) − ( f (c ) − f ( x ) ) ≥ f (c ) − f (c ) − f ( x )
1
≥ f (c ) − f ( x ) − f (c ) > f ( c ) − f ( c )
2
1
⇒ f ( x ) > f (c ) > 0 q
2
v Example
Define a function f by
 x cos x ; x ≠ o
f ( x) = 
 0 ; x=0
This function is continuous at x = 0 because
f ( x ) − f (0) = x cos x ≤ x ( Q cos x ≤ 1 )
Which shows that for ε > 0 , we can find δ > 0 such that
f ( x ) − f (0) < ε whenever 0 < x − c < δ = ε q
v Example
f ( x) = x is continuous on [ 0,∞ [ .
Let c be an arbitrary point such that 0 < c < ∞
For ε > 0 , we have
x−c x−c
f ( x ) − f (c ) = x − c = <
x+ c c
x−c
⇒ f ( x) − f (c) < ε whenever <ε
c
i.e. x − c < c ε = δ
⇒ f is continuous for x = c .
Q c is an arbitrary point lying in [0,∞ [
∴ f ( x) = x is continuous on [0,∞ [ q
………………………..
9
v Example
Consider the function f defined on ¡ such that
 1 , x is rational
f ( x) = 
 −1 , x is irrational
This function is discontinuous every where but f ( x) is continuous on ¡ . q
v Theorem
A mapping of a metric space X into a metric space Y is continuous on X iff
f −1 (V ) is open in X for every open set V in Y .
Proof
Suppose f is continuous on X and V is open in Y .
We are to show that f −1 (V ) is open in X i.e. every point of f −1 (V ) is an
interior point of f −1 (V ) .
Let p ∈ X and f ( p ) ∈V
Q V is open
∴ ∃ ε > 0 such that y ∈V if dY ( y, f ( p) ) < ε …….. (i)
Q f is continuous at p
∴ ∃ δ > 0 such that dY ( f ( x), f ( p) ) < ε when d X ( x, p ) < δ ……… (ii)
From (i) and (ii), we conclude that
x ∈ f −1 (V ) as soon as d X ( x, p ) < δ
Which shows that f −1 (V ) is open in X .
Conversely, suppose f −1 (V ) is open in X for every open set V in Y .
We are to prove that f is continuous for this.
Fix p ∈ X and ε > 0 .
Let V be the set of all y ∈ Y such that dY ( y, f ( p) ) < ε
V is open, f −1 (V ) is open
⇒ ∃ δ > 0 such that x ∈ f −1 (V ) as soon as d X ( x, p ) < δ .
But if x ∈ f −1 (V ) then f ( x ) ∈V so that dY ( f ( x ), f ( y ) ) < ε
Which proves that f is continuous. q
Note
The above theorem can also be stated as a mapping f : X → Y is continuous iff
f −1 (C ) is closed in X for every closed set C in Y . q
v Theorem
Let f1 , f 2 , f3 ,...., f k be real valued functions on a metric space X and f be a
mapping from X on to ¡ k defined by
f ( x ) = ( f1 ( x), f2 ( x ), f3 ( x),....., f k ( x) ) , x ∈ X
then f is continuous on X if and only if f1, f 2 , f 3 ,....., f k are continuous on X .
Proof
Let us suppose that the function f is continuous on X , we are to show that
f1 , f 2 , f3 ,......, f k are continuous on X .
(
If p ∈ X , then d ¡k f ( x), f ( p ) < ε ) whenever d X ( x, p ) < δ
⇒ f ( x) − f ( p ) < ε whenever x− p < δ
10
⇒ f1 ( x) − f1 ( p), f1 ( x) − f1 ( p),...... f k ( x) − f k ( p) < ε whenever x− p < δ
1
⇒ ( f1 ( x) − f1 ( p ) ) , ( f 2 ( x) − f 2 ( p ) ) ,......, ( f k ( x) − f k ( p ) )  < ε
2 2 2 2
 
whenever x − p < δ
1
 k 2
2
i.e. ⇒ ∑ ( f i ( x) − f i ( p ) )  < ε whenever x − p < δ
 i =1 
⇒ f1 ( x) − f1 ( p) < ε whenever x − p < δ
f 2 ( x) − f 2 ( p) < ε whenever x − p < δ
……………………
……………………
……………………
f k ( x) − fk ( x ) < ε whenever x − p < δ
⇒ all the functions f1, f 2 , f 3 ,....., f k are continuous at p .
Q p is arbitrary point of x , therefore f1, f 2 , f 3 ,....., f k are continuous on X .
Conversely, suppose that the function f1, f 2 , f 3 ,....., f k are continuous on X , we
are to show that f is continuous on X .
For p ∈ X and given ε i > 0 , i = 1,2,.....k ∃ δ i > 0 , i = 1,2,...., k
Such that
f1 ( x) − f1 ( p) < ε1 whenever x − p < δ1
f 2 ( x) − f2 ( p) < ε 2 whenever x − p < δ2
……………………
……………………
……………………
f k ( x) − f k ( x) < ε k whenever x − p < δk
Take δ = min (δ 1, δ 2 ,δ 3 ,...., δ k ) then
fi ( x) − fi ( p) < ε i whenever x− p < δ
( )2
1 1
⇒ ( f1 ( x) − f1 ( p ) ) + ( f 2 ( x) − f 2 ( p ) ) + .... + ( f k ( x) − f k ( p ) ) 
2
< ε12 + ε 22 + .... + ε k2
2 2 2
 
1
i.e. ⇒ ( f1 ( x) − f1 ( p ) ) + ( f 2 ( x) − f 2 ( p ) ) + ..... + ( f k ( x) − f k ( p ) )  < ε
2 2 2 2
 
whenever x − p < δ
(ε ) 2= ε
1
where 2
1 + ε 22 + ..... + ε k2
(
Then d ¡k f ( x), f ( p ) < ε ) whenever d X ( x, p ) < δ
⇒ f ( x) is continuous at p .
Q p is an arbitrary point therefore we conclude that f is continuous on X . q
…………………………
11
v Theorem
Suppose f is continuous on [ a, b]
i) If f (a ) < 0 and f (b ) > 0 then there is a point c , a < c 0 and f (b ) < 0 , then there is a point c , a < c < b such that f (c ) = 0 .
Proof
i) Bisect [ a, b] then f must satisfy the given condition on at least one of the
sub-interval so obtained. Denote this interval by [ a2 , b2 ]
If f satisfies the condition on both sub-interval then choose the right hand one
[ a2 , b2 ] .
It is obvious that a ≤ a2 ≤ b2 ≤ b . By repeated bisection we can find nested
intervals {I n } , I n+1 ⊆ I n , I n = [ an , bn ] so that f satisfies the given condition on
[ an , bn ] , n = 1,2,.......
And a = a1 ≤ a2 ≤ a3 ≤ ..... ≤ an ≤ bn ≤ ..... ≤ b2 ≤ b1 = b
n
1
Where bn − an =   ( b − a )
2
II
n
Then n contain one and only one point. Let that point be c such that
i =1
f (c ) = 0
If f (c ) ≠ 0 , let f (c ) > 0 then there is a subinterval [ am , bm ] such that am < bm < c
Which can not happen. Hence f (c ) = 0
ii) Do yourself as above q
v Example
Show that x 3 − 2 x 2 − 3 x + 1 = 0 has a solution c∈ [ −1,1]
Solution
Let f ( x) = x 3 − 2 x 2 − 3x + 1
Q f ( x) is polynomial
∴ it is continuous everywhere. (for being a polynomial continuous everywhere)
Now f (−1) = (−1)3 − 2(−1)2 − 3(−1) + 1
= −1 − 2 + 3 + 1 = 1 > 0
f (1) = (1)3 − 2(1)2 − 3(1) + 1
= 1 − 2 − 3 + 1 = −3 < 0
Therefore there is a point c∈ [ −1,1] such that f (c ) = 0
i.e. c is the root of the equation. q
v Theorem (The intermediate value theorem)
Suppose f is continuous on [ a, b] and f (a ) ≠ f (b ) , then given a number λ that
lies between f (a ) and f (b) , ∃ a point c , a < c 0
⇒ ∃ a point c between a and b such that g (c) = 0
⇒ f (c ) − λ = 0 ⇒ f (c ) = λ
If f (a ) > f (b) then take g ( x) = λ − f ( x) to obtain the required result. q
12
v Theorem
Suppose f is continuous on [ a, b] , then f is bounded on [ a, b]
(Continuity implies boundedness)
Proof
Suppose that f is not bounded on [ a, b] ,
We can, therefore, find a sequence { xn } in the interval [ a, b] such that
f ( xn ) > n for all n ≥ 1 .
⇒ { f ( xn )} diverges.
But a ≤ xn ≤ b ; n ≥ 1
⇒ ∃ a subsequence { x } such that { x } converges to λ .
nk nk
⇒ { f ( x )}
nk also converges to λ .
⇒ { f ( xn )} converges to λ .
Which is contradiction
Hence our supposition is wrong. q
v Uniform continuity
Let f be a mapping of a metric space X into a metric space Y . We say that f
is uniformly continuous on X if for every ε > 0 there exists δ > 0 such that
dY ( f ( p), f (q) ) < ε ∀ p, q ∈ X for which d x ( p, q ) < δ
The uniform continuity is a property of a function on a set i.e. it is a global
property but continuity can be defined at a single point i.e. it is a local property.
Uniform continuity of a function at a point has no meaning.
If f is continuous on X then it is possible to find for each ε > 0 and for each
point p of X , a number δ > 0 such that dY ( f ( x), f ( p) ) < ε whenever
d X ( x, p ) < δ . Then number δ depends upon ε and on p in this case but if f is
uniformly continuous on X then it is possible for each ε > 0 to find one number
δ > 0 which will do for all point p of X .
It is evident that every uniformly continuous function is continuous.
To emphasize a difference between continuity and uniform continuity on set S ,
we consider the following examples. q
v Example
Let S be a half open interval 0 < x ≤ 1 and let f be defined for each x in S by
the formula f ( x) = x 2 . It is uniformly continuous on S . To prove this observe that
we have
f ( x) − f ( y) = x 2 − y 2
= x− y x+ y
< 2 x− y
If x − y < δ then f ( x ) − f ( y ) < 2δ = ε
ε
Hence if ε is given we need only to take δ = to guarantee that
2
f ( x ) − f ( y ) < ε for every pair x, y with x − y < δ
Thus f is uniformly continuous on the set S . q
13
v Example
f ( x) = x n , n ≥ 0 is uniformly continuous of [0,1]
Solution
For any two values x1, x2 in [0,1] we have
x1n − x2n = ( x1 − x2 ) ( x1n−1 + x1n−2 x2 + x1n−3 x22 + ..... + x2n−1 )
≤ n x1 − x2
ε
Given ε > 0 , we can find δ = independent of x1 and x2 such that
n
ε
x12 − x22 < n x1 − x2 < ε whenever x1, x2 ∈ [0,1] and x1 − x2 < δ =
n
Hence the function f is uniformly continuous on [0,1] . q
v Example
Let S be the half open interval 0 < x ≤ 1 and let a function f be defined for each
1
x in S by the formula f ( x ) = . This function is continuous on the set S ,
x
however we shall prove that this function is not uniformly continuous on S .
Solution
Let suppose ε = 10 and suppose we can find a δ , 0 < δ < 1 , to satisfy the
condition of the definition.
δ
Taking x = δ , y = , we obtain
11
10δ
x− y = < δ
11
and
1 11 10
f ( x) − f ( y) = − = > 10
δ δ δ
Hence for these two points we have f ( x ) − f ( y ) > 10 (always)
Which contradict the definition of uniform continuity.
Hence the given function being continuous on a set S is not uniformly
continuous on S . q
v Example
1
f ( x ) = sin ; x ≠ 0 . is not uniformly continuous on 0 < x ≤ 1 i.e (0,1].
x
Proof
Suppose that f is uniformly continuous on the given interval then for ε = 1 ,
there is δ > 0 such that
f ( x1 ) − f ( x2 ) < 1 whenever x1 − x2 < δ
1 1
Take x1 = and x2 = , n ≥1.
( n − 2 )π
1
3 ( n − 12 )π
2
So that x1 − x2 < δ =
3 ( n − 12 )π
But f ( x1 ) − f ( x2 ) = sin ( n − 12 )π − sin 3 ( n − 12 )π = 2 >1
Which contradict the assumption.
Hence f is not uniformly continuous on the interval. q
14
v Example
Prove that f ( x) = x is uniformly continuous on [0,1] .
Solution
Suppose ε = 1 and suppose we can find δ , 0 < δ < 1 to satisfy the condition of
the definition.
δ2
Taking x = δ , y =
2
4
δ2 3δ 2
Then x − y = δ − 2
= <δ
4 4
δ2
And f ( x) − f ( y) = δ −
2
4
δ δ
= δ− = < 1=ε
2 2
Hence f is uniformly continuous on [0,1] . q
v Theorem
If f is continuous on a closed and bounded interval [ a, b] , then f is uniformly
continuous on [ a, b] .
Proof
Suppose that f is not uniformly continuous on [ a, b] then ∃ a real number
ε > 0 such that for every real number δ > 0 .
We can find a pair u , v satisfying
u − v < δ but f (u ) − f (v) ≥ ε > 0
1
If δ = , n = 1,2,3,....
n
We can determine two sequence {un } and {vn } such that
1
u n − vn < but f (un ) − f (vn ) ≥ ε
n
Q a ≤ un ≤ b ∀ n = 1,2,3.......
∴ there is a subsequence {u } which converges to some number
nk u0 in [ a, b]
⇒ for some λ > 0 , we can find an integer n0 such that
unk − u0 < λ ∀ n ≥ n0
1
⇒ vnk − u0 ≤ vnk − unk + unk − u0 < +λ
n
{ }
⇒ vnk also converges to u0 .
⇒ { f ( u )} and { f ( v )} converge to f (u ) .
nk nk 0
Consequently, f ( u ) − f ( v ) < ε whenever

nk nk unk − vnk < ε
Which contradict our supposition.
Hence we conclude that f is uniformly continuous on [ a, b] . q
…………………………….
15
v Theorem
Let f and g be two continuous mappings from a metric space X into ¡ k , then
the mappings f + g and f ⋅ g are also continuous on X .
i.e. the sum and product of two continuous vector valued function are also
continuous.
Proof
i) Q f & g are continuous on X .
∴ by the definition of continuity, we have for a point p ∈ X .
ε
f ( x) − f ( p ) < whenever x − p < δ 1
2
ε
and g ( x) − g ( p) < whenever x − p < δ2
2
Now consider
f ( x) + g ( x) − f ( x) − g ( p )
= f ( x ) − f ( p ) + g ( x) − g ( p )
≤ f ( x) − f ( p ) + g ( x) − g ( p)
ε ε
+ = ε whenever x − p < δ where δ = min (δ 1, δ 2 )
<
2 2
which shows that the vector valued function f + g is continuous at x = p and
hence on X .
k
ii) f ⋅g = ∑ f ⋅g
i =1
i i
= f1g1 + f 2 g 2 + f 3 g 3 + ..... + f k g k
Q the function f and g are continuous on X
∴ their components f i and g i are continuous on X . q
v Question
Suppose f is a real valued function define on ¡ which satisfies
lim [ f ( x + h) − f ( x − h)] = 0 ∀ x ∈ ¡
h→0
Does this imply that the function f is continuous on ¡ .
Solution
Q lim [ f ( x + h) − f ( x − h) ] = 0 ∀ x∈¡
h→0
⇒ lim f ( x + h) = lim f ( x − h)
h→0 h→0
⇒ f ( x + 0) = f ( x − 0) ∀ x ∈ ¡
Also it is given that f ( x) = f ( x + 0) = f ( x − 0)
It means f is continuous on x ∈ ¡ . q
……………………………
16
v Discontinuities
If x is a point in the domain of definition of the function f at which f is not
continuous, we say that f is discontinuous at x or that f has a discontinuity at
x.
If the function f is defined on an interval, the discontinuity is divided into two
types
1. Let f be defined on ( a, b ) . If f is discontinuous at a point x and if f ( x+ ) and
f ( x−) exist then f is said to have a discontinuity of first kind or a simple
discontinuity at x .
2. Otherwise the discontinuity is said to be second kind.
For simple discontinuity
i. either f ( x +) ≠ f ( x −) [ f ( x) is immaterial]
ii. or f ( x + ) = f ( x −) ≠ f ( x) q
v Example
1 , x is rational
i) Define f ( x) = 
0 , x is irrational
The function f has discontinuity of second kind on every point x because
neither f ( x+ ) nor f ( x−) exists. q
 x , x is rational
ii) Define f ( x) = 
 0 , x is irrational
Then f is continuous at x = 0 and has a discontinuity of the second kind at
every other point. q
 x+2 (−3 < x < −2)
iii) Define f ( x ) =  − x − 2 (−2 < x < 0)
 x+2 (0 < x < 1)

The function has simple discontinuity at x = 0 and it is continuous at every other
point of the interval (−3,1) q
 1
sin , x≠0
iv) Define f ( x ) =  x
 0 , x=0

Q neither f (0+ ) nor f (0−) exists, therefore the function f has discontinuity
of second kind.
f is continuous at every point except x = 0 . q

Prof. Syyed Gull Shah
(2) Book
Collected and composed by: Atiq ur Rehman (mathcity@gmail.com)

Page Setup: Legal ( 8′′ 1 2 × 14′′ )
Printed: October 20, 2004. Updated: November 03, 2005
Chapter 4 – Differentiation
Source: Syed Gul Shah (Chairman, Department of Mathematics, UoS Sargodha)
Collected & Composed by: Atiq ur Rehman (mathcity@gmail.com), http://www.mathcity.org/msc
v Derivative of a function:
Let f be defined and real valued on [ a, b] . For any point c Î [ a, b ] , form the
quotient
f ( x ) - f (c )
x-c
and define
f ( x ) - f (c )
f ¢(c ) = lim
x ®c x-c
provided this limit exits.
We thus associate a function f ¢ with the function f , where domain of f ¢ is the
set of points at which the above limit exists.
The function f ¢ is so defined is called the derivative of f .
(i) If f ¢ is defined at point x, we say that f is differentiable at x.
(ii) f ¢(c) exists if and only if for a real number e > 0 , $ a real number d > 0
such that
f ( x ) - f (c )
- f ¢(c) < e whenever x - c < d
x-c
(iii) If x - c = h then we have
f (c + h ) - f ( c )
f ¢(c) = lim
h®0 h
(iv) f is differentiable at c if and only if c is a removable discontinuity of the
f ( x ) - f (c )
function j ( x) = .
x-c
v Example
(i) A function f : ¡ ® ¡ defined by
ì x 2 sin 1x ; x¹0
f ( x) = í
î 0 ; x=0
This function is differentiable at x = 0 because
f ( x) - f (0) x 2 sin 1x - 0
lim = lim
x ®0 x-0 x ®0 x-0
2
x sin 1x
= lim = lim x sin 1x = 0
x ®0 x x ®0
(ii) Let f ( x) = x n ; n ³ 0 (n is integer), x Î ¡ .

Then
f ( x ) - f (c ) xn - cn
lim = lim
x®c x-c x ®c x - c
( x - c )( x n-1 + cx n-2 + .......... + c n-2 x + c n-1 )

= lim
x ®c x-c
= lim ( x + cx + .......... + c n-2 x + c n-1 )
n -1 n-2
x ®c
n -1
= nc
implies that f is differentiable every where and f ¢( x) = nx n-1 .
2 Chap 4 – Differentiation
v Theorem
Let f be defined on [ a, b] , if f is differentiable at a point x Î [ a, b] , then f is
continuous at x. (Differentiability implies continuity)
Proof
We know that
f (t ) - f ( x)
lim = f ¢( x) where t ¹ x and a < t < b
t®x t-x
Now
æ f (t ) - f ( x) ö
lim ( f (t ) - f ( x) ) = lim ç ÷ lim (t - x )
t®x t®x è t-x ø t®x
= f ¢( x ) × 0
=0
Þ lim f (t ) = f ( x ) .
t ®x
Which show that f is continuous at x.
Note
(i) The converse of the above theorem does not hold.
ìx if x ³ 0
Consider f ( x) = x = í
î- x if x < 0
f ¢(0) does not exists but f ( x) is continuous at x = 0
(ii) If f is discontinuous at c Î D f then f ¢(c) does not exists.

e.g.
ì1 if x > 0
f ( x) = í
î0 if x £ 0
is discontinuous at x = 0 therefore it is not differentiable at x = 0 .
(iii) f is differentiable at a point c if and only if D+ f (c) (right derivative) and
D- f (c) (left derivative) exists and equal.
i.e. D+ f (c) = D- f (c) = Df (c)
v Example
Let f : ¡ ® ¡ be defined by
ì x2 if x > 1
f ( x) = í 3
îx if x £ 1
f ( x) - f (1)
then D+ f (1) = lim
x®1+ h
h ®0
x -1
f (1 + h) - f (1) (1 + h) 2 - 1
= lim = lim
h ®0 1 + h -1 h ®0 h
1 + 2h + h - 1
2
= lim = lim (2 + h) = 2
h ®0 h h ®0
and
f ( x) - f (1)
D- f (1) = lim
x ®1- h
h®0
x -1
f (1 - h) - f (1) (1 - h)3 - 1
= lim = lim
h ®0 1 - h -1 h ®0 -h
1 - 3h + 3h - h - 1
( )
2 3
= lim = lim 3 - 3h + h 2 = 3
h ®0 -h h ®0
Chap 4 – Differentiation 3
Since D+ f (1) ¹ D- f (1) Þ f ¢(1) does not exist even though f is continuous at
x = 1 . f ¢( x ) exist for all other values of x.
v Theorem
Suppose f and g are defined on [a, b] and are differentiable at a point
f
x Î [a, b] , then f + g , fg and are differentiable at x and
g
(i) ( f + g )¢( x ) = f ¢( x ) + g ¢( x )
(ii) ( fg )¢( x ) = f ¢( x ) g ( x) + f ( x ) g ¢( x)
æ f ö¢ g ( x) f ¢( x) - f ( x ) g ¢( x)
(iii) ç ÷ ( x) =
ègø g 2 ( x)
The proof of this theorem can be get from any F.Sc or B.Sc text book.
Note
The derivative of any constant is zero.
And if f is defined by f ( x ) = x then f ¢( x ) = 1
And for f ( x) = x n then f ¢( x ) = n x n-1 where n is positive integer, if n < 0 we
have to restrict ourselves to x = 0 .
Thus every polynomial Pn ( x) = a0 + a1x + a2 x 2 + .......... + an x n is differentiable
every where and so every rational function except at the point where denominator is
zero.
v Theorem (Chain Rule)
Suppose f is continuous on [a, b] , f ¢( x ) exists at some point x Î [a, b] . A
function g is defined on an interval I which contains the range of f , and g is
differentiable at the point f ( x) .
If h(t ) = g ( f (t ) ) ; a £ t £ b
Then h is differentiable at x and h¢( x ) = g ¢ ( f ( x) ) × f ¢( x) .
Proof
Let y = f ( x)
By the definition of the derivative we have
f (t ) - f ( x ) = (t - x) [ f ¢( x) + u (t )] ………... (i)
and g ( s) - g ( y ) = ( s - y ) [ g ¢( y ) + v( s )] ……….. (ii)
where t Î [ a, b] , s Î I and u (t ) ® 0 as t ® x and v( s ) ® 0 as s ® y .
Let us suppose s = f (t ) then
h(t ) - h( x) = g ( f (t ) ) - g ( f ( x) )
= [ f (t ) - f ( x )][ g ¢( y ) + v( s) ] by (ii)
= (t - x) [ f ¢( x ) + u (t )][ g ¢( y ) + v (s )] by (i)
or if t ¹ x
h(t ) - h( x)
= [ f ¢( x) + u (t )][ g ¢( y ) + v( s )]
t-x
taking the limit as t ® x we have
h¢( x) = [ f ¢( x) + 0][ g ¢( y ) + 0]
= g ¢ ( f ( x ) ) × f ¢( x) Q y = f ( x)
which is the required result.
It is known as chain rule.
………………………………………………..
v Example
Let f be defined by
ì 1
ï x sin ; x¹0
f ( x) = í x
ïî 0 ; x=0
1 1 1
Þ f ¢( x) = sin - cos where x ¹ 0 .
x x x
1
Q at x = 0 , is not defined.
x
\ Applying the definition of the derivative we have
1
t sin
f (t ) - f (0) t = lim sin 1
f ¢(0) = lim = lim
t ®0 t -0 t ®0 t t ®0 t
which does not exit.
The derivative of the function f ( x) does not exist at x = 0 but it is continuous at
x = 0 (i.e. it is not differentiable although it is continuous at x = 0 )
Same the case with absolute value function.
v Example
Let f be defined by
ì 2 1
ï x sin ; x¹0
f ( x) = í x
ïî 0 ; x=0
1 1
We have f ¢( x ) = 2 x sin - cos where x ¹ 0 .
x x
1
Q at x = 0 , is not defined.
x
\ Applying the definition of the derivative we have
f (t ) - f (0) 1
= t sin £ t , (t ¹ 0)
t -0 t
Taking limit as t ® 0 we see that f ¢(0) = 0
Thus f is differentiable at points x but f ¢ is not a continuous function, since
1
cos does not tend to a limit as x ® 0 .
x
v Local Maximum
Let f be a real valued function defined on a metric space X , we say that f
has a local maximum at a point p Î X if there exist d > 0 such that f (q ) £ f ( p )
" q Î X with d ( p, q ) < d .
Local minimum is defined likewise.
…………………………………………..
v Theorem
Let f be defined on [ a, b] , if f has a local maximum at a point x Î [ a, b] and
if f ¢( x ) exist then f ¢( x ) = 0 .
(The analogous for local minimum is of course also true)
Proof
Choose d such that
a < x -d < x < x + d < b
Now if x - d < t < x then f(x)
f (t ) - f ( x)
³0
t-x
Taking limit as t ® x we get
f ¢( x ) ³ 0 …………. (i) a x – d t x t x + d b
If x < t < x + d
Then
f (t ) - f ( x)
£0
t-x
Again taking limit when t ® x we get
f ¢( x ) £ 0 ……………. (ii)
Combining (i) and (ii) we have
f ¢( x ) = 0
v Generalized Mean Value Theorem
If f and g are continuous real valued functions on closed interval [ a, b] , then
there is a point x Î ( a, b ) at which
[ f (b) - f (a)] g ¢( x) = [ g (b) - g (a)] f ¢( x)
The differentiability is not required at the end point.
Proof
Let
h(t ) = [ f (b) - f (a)] g (t ) - [ g (b) - g (a) ] f (t ) ( a £ t £ b)
Q h involves f and g therefore h is
i) Continuous on close interval [ a, b] .
ii) Differentiable on open interval (a, b) .
iii) and h (a) = h(b) .
To prove the theorem we have to show that h¢( x ) = 0 for some x Î (a, b )
There are two cases to be discussed
(i) h is constant function.
Þ h¢( x) = 0 " x Î (a, b)
(ii) If h is not constant.
then h (t ) > h(a ) for some t Î (a, b)
Let x be the point in the interval (a, b) at which h attain its maximum,
then h¢( x ) = 0
Similarly,
if h (t ) < h(a ) for some t Î (a, b) then $ a point x Î (a, b ) at which the
function h attain its minimum and since the derivative at a local minimum is
zero therefore we get h¢( x ) = 0
Hence
h¢( x ) = [ f (b) - f (a)] g ¢( x ) - [ g (b) - g (a)] f ¢( x) = 0
This gives the desire result.
v Geometric Interpretation of M.V.T.

Consider a plane curve C represented by x = f (t ) , y = g (t ) then theorem
states that there is a point S on C between two points P ( f (a), g (a) ) and
Q ( f (b), g (b) ) of C such that the tangent at S to the curve C is parallel to the
chord PQ .
v Lagrange’s M.V.T.
Let f be
i) continuous on [a, b]
ii) differentiable on (a, b)
f (b) - f (a )
then $ a point x Î (a, b ) such that = f ¢( x) .
b-a
Proof
Let us design a new function
h(t ) = [ f (b) - f (a)]t - (b - a) f (t ) , ( a £ t £ b)
then clearly h (a) = h(b)
Since h(t ) depends upon t and f (t ) therefore it possess all the properties of f .
Now there are two cases
i) h is a constant.
implies that h¢( x ) = 0 " x Î ( a, b)
ii) h is not a constant, then
if h (t ) > h(a ) for some t Î (a, b)
then $ a point x Î (a, b ) at which h attains its maximum
implies that h¢( x ) = 0
and if h (t ) < h(a )
then $ a point x Î (a, b ) at which h attain its minimum
implies that h¢( x ) = 0
Q h(t ) = [ f (b) - f (a)]t - (b - a) f (t )
\ h¢( x) = [ f (b) - f (a)] - (b - a) f ¢( x )
Which gives
f (b) - f (a )
= f ¢( x) as desired.
b-a
v Theorem (Intermediate Value Theorem or Darboux,s Theorem)
Suppose f is a real differentiable function on some interval [a, b] and
suppose f ¢(a ) < l < f ¢(b) then there exist a point x Î (a, b ) such that f ¢( x) = l .
A similar result holds if f ¢(a ) > f ¢(b) .
Proof
Put g (t ) = f (t ) - lt
Then g ¢(t ) = f ¢(t ) - l
If t = a we have
g ¢(a ) = f ¢(a ) - l
Q f ¢(a ) - l < 0 \ g ¢(a ) < 0
implies that g is monotonically decreasing at a .
a t1 t2 b
Þ $ a point t1 Î (a, b) such that g (a ) > g (t1 ) .
Similarly,
g ¢(b) = f ¢(b) - l
Q f ¢(b) - l > 0 \ g ¢(b) > 0
implies that g is monotonically increasing at b .

Þ $ a point t2 Î (a, b) such that g (t2 ) < g (b)
Þ the function attain its minimum on (a, b) at a point x (say)
such that g ¢( x) = 0 Þ f ¢( x ) - l = 0
Þ f ¢( x) = l .
Note
We know that a function f may have a derivative f ¢ which exist at every point
but is discontinuous at some point however not every function is a derivative. In
particular derivatives which exist at every point on the interval have one important
property in common with function which are continuous on an interval is that
intermediate value are assumed.
The above theorem relates to this fact.
v Question
If a and c are real numbers, c > 0 and f is defined on [-1,1] by
ì x a sin x -c ; x ¹ 0
f ( x) = í
î 0 ; x=0
then discuss the differentiability as well as continuity at x = 0 .
Solution
f (t ) - f ( x)
f ¢( x) = lim
t®x t-x
t sin t - c - x a sin x - c
a
= lim
t®x t-x
a -c
t sin t
Þ f ¢(0) = lim
t ®0 t
= lim t sin t - c
a -1
t ®0
If a - 1 > 0 , then lim t a -1 sin t - c = 0 Þ f ¢(0) = 0 when a > 0 .

t ®0
If a - 1 < 0 i.e. when a < 1 we have t a -1 = t - b where b > 0

And lim t a -1 sin t -c = lim t - b sin t - c
t ®0 t ®0
Which does not exist.
If a - 1 = 0 , we get lim sin t - c
t ®¥
Which also does not exist.
Hence f ¢(0) exists if and only if a > 1 .
Also lim x a sin x - c exist and zero when a > 0 , which equals the actual value of
x ®0
the function f ( x) at zero.
Hence the function is continuous at x = 0 .
v Question
Let f be defined for all real x and suppose that
f ( x ) - f ( y ) £ ( x - y )2 " real x & y . Prove that f is constant.
Solution
Since f ( x ) - f ( y ) £ ( x - y )2
Therefore
- ( x - y ) 2 £ f ( x) - f ( y ) £ ( x - y ) 2
Dividing throughout by x - y , we get
f ( x) - f ( y )
- ( x - y) £ £ ( x - y) when x > y
x- y
and
f ( x) - f ( y )
- ( x - y) ³ ³ ( x - y ) when x < y
x- y
Taking limit as x ® y , we get
0 £ f ¢( y ) £ 0 ù
Þ f ¢( y ) = 0
0 ³ f ¢( y ) ³ 0 úû
which shows that function is constant.
v Question
If f ¢( x ) > 0 in (a, b) then prove that f is strictly increasing in (a, b) and let g
be its inverse function, prove that the function g is differentiable and that
1
g ¢ ( f ( x) ) = ; a< x<b
f ( x)
Solution
Let y Î ( f (a), f (b) )
Þ y = f ( x) for some x Î (a, b )
g ( z) - g ( y)
Þ g ¢( y ) = lim
z®y z-y
g ( f ( xz ) ) - g ( f ( x) )
= lim g ( f ( xz ) ) =
xz ® x f ( xz ) - f ( x )
f -1 ( f ( xz ) ) - f -1 ( f ( x) )
= lim
xz ® x f ( xz ) - f ( x )
xz - x 1 1
= lim = =
xz ® x f ( x ) - f ( x ) f ( xz ) - f ( x ) f ¢( x)
z lim
xz ® x xz - x
v Question
Suppose f is defined and differentiable for every x > 0 and f ¢( x ) ® 0 as
x ® +¥ put g ( x) = f ( x + 1) - f ( x) . Prove that g ( x ) ® 0 as x ® +¥ .
Solution
Since f is defined and differentiable for x > 0 therefore we can apply the
Lagrange’s M.V. T. to have
f ( x + 1) - f ( x) = ( x + 1 - x) f ¢( x1 ) where x < x1 .
Q f ¢( x) ® 0 as x ® ¥
\ f ¢( x1 ) ® 0 as x ® ¥
Þ f ( x + 1) - f ( x ) ® 0 as x ® 0
Þ g ( x) ® 0 as x ® 0
……………………………
v Question (L Hospital Rule)

Suppose f ¢( x ), g ¢( x) exist, g ¢( x) ¹ 0 and f ( x ) = g ( x) = 0 .
f (t ) f ¢( x )
Prove that lim =
t ® x g (t ) g ¢( x)
Proof
f (t ) f (t ) - 0 f (t ) - f ( x)
lim = lim = lim Q f ( x) = g ( x ) = 0
t ® x g (t ) t ® x g (t ) - 0 t ® x g (t ) - ( x )
f (t ) - f ( x) t-x
= lim ×
t®x t-x g (t ) - ( x)
f (t ) - f ( x) 1
= lim × lim
t®x t-x t ® x g (t ) - ( x )
t-x
f (t ) - f ( x) 1 1 f ¢( x)
= lim × = f ¢( x) × =
t®x t-x g (t ) - ( x) g ¢( x) g ¢( x)
lim
t®x t-x
Q.E.D.
v Question
Suppose f is defined in the neighborhood of a point x and f ¢¢( x) exists.
f ( x + h) + f ( x - h ) - 2 f ( x)
Show that lim = f ¢¢( x)
h®0 h2
Solution
By use of Lagrange’s Mean Value Theorem
f ( x + h) + f ( x) = hf ¢( x1 ) where x < x1 < x + h …………… (i)
and
- [ f ( x - h) - f ( x)] = hf ¢( x2 ) where x - h < x2 < x …………… (ii)
Subtract (ii) from (i) to get
f ( x + h) + f ( x - h) - 2 f ( x) = h [ f ¢( x1 ) - f ¢( x2 )]
f ( x + h) + f ( x - h) - 2 f ( x) f ¢( x1 ) - f ¢( x2 )
Þ =
h2 h
Q x2 - x1 ® 0 as h ® 0
therefore
f ( x + h) + f ( x - h) - 2 f ( x ) f ¢( x1 ) - f ¢( x2 )
\ lim = lim
h®0 h 2 x1 ® x2 x1 - x2
= f ¢¢( x2 )
v Question
c c c c
If c0 + 1 + 2 + ......... + n -1 + n = 0
2 3 n n +1
Where c0 , c1 , c2 ,......., cn are real constants.
Prove that c0 + c1 x + c2 x 2 + ......... + cn x n = 0 has at least one real root between 0
and 1.
Solution
c c
Suppose f ( x ) = c0 x + 1 x 2 + .......... + n x n+1
2 n +1
c c c
Then f (0) = 0 and f (1) = c0 + 1 + 2 + .......... + n = 0
2 3 n +1
Þ f (0) = f (1) = 0
Q f ( x) is a polynomial therefore we have
i) It is continuous on [0,1]
ii) It is differentiable on (0,1)
iii) And f (a ) = 0 = f (b)
Þ the function f has local maximum or a local minimum at some point x Î ( 0,1)
Þ f ¢( x) = c0 + c1 x + c2 x 2 + ......... + cn x n = 0 for some x Î ( 0,1)
Þ the given equation has real root between 0 and 1.
v Riemann Differentiation of Vector valued function
If f (t ) = f1 (t ) + i f 2 (t )
f ¢(t ) = f1¢(t ) + i f 2¢ (t )
where f1 (t ) and f 2 (t ) are the real and imaginary part of f (t ) .
The Rule of differentiation of real valued functions are valid in case of vector valued
function but the situation changes in the case of Mean Value Theorem.
v Example
Take f ( x ) = eix = cos x + i sin x in (0,2p ) .
Then f (2p ) = cos 2p + i sin 2p = 1
f (0) = cos(0) + i sin(0) = 1
Þ f (2p ) - f (0) = 0 but f ¢( x ) = i eix
f (2p ) - f (0)
Þ ¹ i eix (there is no such x )
2p - 0
Þ the M.V.T. fails.
In case of vector valued functions, the M.V.T. is not of the form as in the case of
real valued function.
v Theorem
Let f be a continuous mapping of the interval [a, b] into a space ¡ k and f be
differentiable in (a, b) then $ x Î (a, b ) such that f (b) - f (a) £ (b - a ) f ¢( x ) .
Proof
Put z = f (b) - f (a)
And suppose j (t ) = z × f (t ) (a £ t £ b)
j (t ) so defined is a real valued function and it possess the properties of f (t ) .
Þ M.V.T. is applicable to j (t ) .
We have j (b) - j (a ) = (b - a )j ¢( x)
i.e. j (b) - j (a) = (b - a) z × f ¢( x ) for some x Î (a, b ) ……….. (i)
Also j (b) = z × f (b) and j (a) = z × f (a)
( )
Þ j (b) - j (a ) = z × f (b) - f (a ) …………. (ii)
from (i) and (ii)
z × z = (b - a ) z × f ¢( x )
£ (b - a ) z f ¢( x )
Þ z £ (b - a ) z f ¢( x)
2
Þ z £ (b - a ) f ¢( x )
i.e. f (b) - f (a) £ (b - a ) f ¢( x ) Q z = f (b) - f (a)
which is the required result.
v Question
If f ( x ) = x 3 , then compute f ¢( x ), f ¢¢( x ) and f ¢¢¢( x) , and show that f ¢¢¢(0) does
not exist.
Solution
ì x 3 if x ³ 0
f ( x) = x = í 3
3
î- x if x < 0
f ( x ) - f (0) x3 - 0
Now D+ f (0) = lim = lim = lim x 2 = 0
x ®0 + 0 x-0 x ® 0 + 0 x-0 x ®0+ 0
f ( x) - f (0) - x3 - 0
& D- f (0) = lim = lim = lim (- x 2 ) = 0
x®0- 0 x-0 x®0- 0 x - 0 x ® 0 -0
Q D+ f ( x) = D- f ( x)
\ f ¢( x ) exists at x = 0 & f ¢(0) = 0 .
Now if x ¹ 0 and x > 0 then
f ( x) = x 3 Þ f ¢( x) = 3x 2
and if x ¹ 0 and x < 0 then
f ( x ) = - x3 Þ f ¢( x) = -3x 2
ì 3x2 if x > 0
ï
i.e. f ¢( x) = í 0 if x = 0
ïî -3 x 2 if x < 0
f ¢( x) - f ¢(0) 3x 2 - 0
Now D+ f ¢(0) = lim = lim
x ®0 + 0 x-0 x ®0+ 0 x - 0
= lim 3 x = 0
x ®0+ 0
f ¢( x ) - f ¢(0) -3 x 2 - 0
And Now D- f ¢(0) = lim = lim
x ®0- 0 x-0 x ® 0 -0 x-0
= lim (-3x ) = 0
x ®0 + 0
Q D+ f ¢( x) = D- f ¢( x)
\ f ¢¢( x) exists at x = 0 & f ¢¢(0) = 0 .
Now if x ¹ 0 and x > 0 then
f ¢( x ) = 3x 2 Þ f ¢¢( x) = 6 x
and if x ¹ 0 and x < 0 then
f ¢( x) = -3x 2 Þ f ¢¢( x) = - 6 x
ì 6x if x > 0
ï
i.e. f ¢¢( x ) = í 0 if x = 0
ïî- 6 x if x < 0
f ¢¢( x) - f ¢¢(0) 6x - 0
Now D+ f ¢¢(0) = lim = lim =6
x ®0 + 0 x-0 x ® 0+ 0 x - 0
f ¢¢( x) - f ¢¢(0) - 6x - 0
And D- f ¢¢(0) = lim = lim = -6
x®0- 0 x-0 x ® 0- 0 x - 0
Q D+ f ¢¢(0) ¹ D- f ¢¢(0)
\ f ¢¢¢(0) doest not exist.
But f ¢¢¢(0) exist if x ¹ 0 , and equal to 6 if x > 0 and equal to - 6 if x < 0 .
……………………………….

Chapter 5 – Function of Several Variables
Source: Syed Gul Shah (Chairman, Department of Mathematics, US Sargodha)
Collected & Composed by: Atiq ur Rehman (atiq@mathcity.tk), http://www.mathcity.tk
v Introduction
There is the basic difference between the calculus of functions of one variable and
the calculus of functions of two variables. But there is a slight difference between the
calculus of two variable and the calculus of functions of three, four or of many
variables. Therefore we shall emphasise mainly on the study of functions of two
variables.
v Function of two variables
If to each point ( x, y ) of a certain part of xy - plane, there is assigned a real number
z , then z is known to be a function of two variable x and y .
e.g. z = x 2 - y 2 , z = x 2 + y 2 , z = xy etc.
v Neighbourhood (nhood)
A neighbourhood of radius d of a point ( x0 , y0 ) of the xy - plane is the set of
points which lies inside a circle with centre at ( x0 , y0 ) and has radius d .
Nd ( x0 , y0 ) = ( x - x0 ) + ( y - y0 ) < d 2
2 2
Similarly, a nhood of a radius d of a point ( x0 , y0 , z0 ) of a space is a sphere with

centre at ( x0 , y0 , z0 ) and radius d .
Nd ( x0 , y0 , z0 ) = ( x - x0 ) + ( y - y0 ) + ( z - z0 ) < d 2
2 2 2
This definition can be extended to the definition of a nhood of a point of a space of

any dimension.
v Open Set
A set is known to be open set if each point ( x0 , y0 ) of the set has a nhood which
totally lies inside the set.
v Domain
A set D which is not empty and open is known to be a domain, if any two points
of the set can be joined by a broken line which lies completely with in D .
v Region
A domain D is known to be a region if some or all of the boundary points are
contained in D .
v Closed Region
A region is known to be closed if it contains all the boundary points.
e.g. i) x 2 + y 2 < 1 (Domain) ii) x y < 1 (Domain)
x + y = 1 (Boundary)
2 2
xy=2 (Boundary)
x2 + y2 £ 1 (Closed region) xy £ 1 (Closed Region)
v Limit & Continuity
Let z = f ( x, y ) be a function of two variables defined in a domain D . Suppose
there is a point ( x0 , y0 ) Î D or is a boundary point then
lim f ( x, y ) = c
x® x0
y ® y0
It means that given e > 0 $ a d > 0 such that

f ( x, y ) - c < e whenever ( x, y ) - ( x0 , y0 ) < d " ( x, y ) Î Nd ( x0 , y0 )
2 Chap 5 – Function of several variables
If limit of a function is equal to actual value of function then f is said to be

continuous at the point ( x0 , y0 )
lim f ( x, y ) = f ( x0 , y0 )
x ® x0
y ® y0
If f is continuous at every point of D , then f is said to be continuous on D .

v Theorem
Let f ( x, y ) & g ( x, y ) be defined in a domain D and suppose that
lim f ( x, y ) = u1 & lim g ( x, y ) = v1
x ® x0 x® x0
y ® y0 y ® y0
a) then (i) lim [ f ( x, y ) + g ( x, y )] = u1 + v1

x ® x0
y ® y0
(ii) lim [ f ( x, y ) × g ( x, y )] = u1v1

x ® x0
y ® y0
f ( x, y ) u1
(iii) lim =
x ® x0 g ( x, y ) v1
y ® y0
b) If f ( x, y ) & g ( x, y ) are defined in D , then

lim f ( x, y ) = f ( x0 , y0 ) & lim g ( x, y ) = g ( x0 , y0 )
x ® x0 x ® x0
y ® y0 y ® y0
i.e. f ( x, y ) , g ( x, y ) are continuous at ( x0 , y0 ) then so are the functions

f ( x, y )
f ( x, y ) + g ( x, y ) , f ( x, y ) g ( x, y ) and , provided g ( x, y ) ¹ 0 .
g ( x, y )
Proof
a) (i) Q lim f ( x, y ) = u1 , lim g ( x, y ) = v1
x ® x0 x® x0
y ® y0 y ® y0
e
\ given > 0 $ a d1 , d 2 > 0 such that
2
e
f ( x, y ) - u1 < " ( x, y ) Î Nd1 ( x0 , y0 )
2
e
& g ( x, y ) - v1 < " ( x, y ) Î Nd 2 ( x0 , y0 )
2
then [ f ( x, y ) + g ( x, y )] - [u1 + v1 ] = [ f ( x, y ) - u1 ] + [ g ( x, y ) - v1 ]
£ f ( x, y ) - u1 + g ( x, y ) - v1
e e
< + " ( x, y ) Î Nd ( x0 , y0 )
2 2
where d = min (d 1, d 2 )
Which show that
lim [ f ( x, y ) + g ( x, y )] = u1 + v1
x ® x0
y ® y0
(ii) f ( x, y ) × g ( x, y ) - u1v1 = f ( x, y ) × g ( x, y ) - u1 g ( x, y ) + u1g ( x, y ) - u1v1

= g ( x, y ) [ f ( x, y ) - u1 ] + u1 [ g ( x, y ) - v1 ]
£ g ( x, y ) [ f ( x, y ) - u1 ] + u1 [ g ( x, y ) - v1 ]
e e
< g ( x, y ) + u1 = e1 " ( x, y ) Î Nd ( x0 , y0 )
2 2
Chap 5 – Function of several variables 3
Þ lim f ( x, y ) × g ( x, y ) = u1v1
x ® x0
y ® y0
1 1
iii) We prove that lim =
x® x0 g ( x, y ) v1
y ® y0
1 1 v - g ( x, y )
- = 1
g ( x, y ) v1 v1 g ( x, y )
g ( x, y ) - v1 e
= < 2
v1 g ( x, y ) v1 g ( x, y )
e e
< 2 < 2
v1 g ( x, y ) - v1 + v1 v1 ( g ( x, y ) - v1 + v1 )
e
< 2 = e1 " ( x, y ) Î Nd 2 ( x0 , y0 )
v1 ( e +v
2 1 )
1 1
Þ lim =
x ® x0 g ( x, y ) v1
y ® y0
Q Þ lim f ( x, y ) = u1 & lim g ( x, y ) = v1

x ® x0 x® x0
y ® y0 y ® y0
By (ii) of theorem
1 f ( x, y ) u1
lim f ( x, y ) × = lim =
x ® x0 g ( x, y ) x® x0 g ( x, y ) v1
y ® y0 y ® y0
b) Since it is given that the limiting values are the same as the actual values of the
f ( x, y )
functions f ( x, y ) + g ( x, y ) , f ( x, y ) × g ( x, y ) and at the point ( x0 , y0 )
g ( x, y )
therefore these function are continuous on ( x0 , y0 ) .
Note
It is to be noted that there is a difference between lim f ( x, y ) and lim f ( x, y )
( x , y ) ®( a ,b ) x ®a
y ®b
i.e.
x ®a
y ®b
(
lim f ( x, y ) = lim lim f ( x, y )
y ®b x ®a
) or
x ®a
y ®b
x ®a (
lim f ( x, y ) = lim lim f ( x, y )
y ®b )
Obviously in the two cases limits are taken first w.r.t one variable and then w.r.t
other variable. These limits are called the repeated limits. Since these are taken along
the special path, therefore repeated limits are the special cases of limits.
lim f ( x, y ) exists if and only if limiting vales are not depend upon any path
( x , y ) ®( a ,b )
along which ( x, y ) ® (a, b) .

v Example
Consider f : ¡ 2 ® ¡ given by
ì x2 y2
ï , ( x, y ) ¹ (0,0)
f ( x, y ) = í x 4 + y 4
îï 0 , ( x, y ) = (0,0)
Now lim f ( x, y ) = lim f ( x, y ) = 0
x®0 y ®0
y ®0 x ®0
However along the straight line y = mx , we have
m4
lim f ( x, y ) =
( x , y ) ®(0,0) 1 + m4
which is different for different values of m . Hence lim f ( x, y ) does not exist.
( x , y ) ®(0,0)
v Example
ì x 2 cos x - y 2 cos y
ï , ( x, y ) ¹ 0
f ( x, y ) = í x2 + y 2
ïî 0 , ( x, y ) = 0
then lim élim f ( x, y ) ù = lim cos x = 1

x ®0 ê
ë y ®0 úû x®0
and lim élim f ( x, y ) ù = lim(- cos y ) = -1
y ®0 ë x ®0 û y ®0
Þ lim f ( x, y ) does not exist.
x ®0
y ®0
v Example
ì sin( x 2 + y 2 )
ï( x + y ) , ( x, y ) ¹ (0,0)
f ( x, y ) = í x2 + y2
ïî 0 , ( x, y ) = (0,0)
sin x
Use < 1 to get
x
f ( x, y ) - 0 £ x + y < x + y
e e
Thus f ( x, y ) - 0 < e whenever x < , y <
2 2
e
Take d = ,
2
It follows that for given e > 0 , we can find d > 0 such that
f ( x, y ) - f (0,0) < e whenever ( x - 0)2 + ( y - 0)2 < d
i.e. " ( x, y ) Î Nd (0,0)
Limit of the function at (0,0) is equal to actual value of function at (0,0) .
Hence f is continuous at (0,0) .
v Partial Derivative
Let z = f ( x, y ) be defined in a domain D of xy -plane and take ( x0 , y0 ) Î D ,
then f ( x, y0 ) is a function of x alone and its derivative may exist. If it exists then its
value at ( x0 , y0 ) is known to be the partial derivative of f ( x, y ) at ( x0 , y0 ) and is
¶f ¶z
denoted as or
¶x ( x0 , y0 ) ¶x ( x0 , y0 )
The other notations are z x , f x , f1 .
¶f f ( x + Dx, y0 ) - f ( x, y0 )
= lim
¶x ( x0 , y0 ) Dx®0 Dx
¶f
We can define in the same manner.
¶y
v Geometrical Interpretation
z = f ( x, y ) represents a surface in space. y = y0 is a plane. z = ( x, y0 ) is the curve
¶f
which arises when y = y0 cuts the surface z = f ( x, y ) . Thus denotes the
¶x ( x0 , y0 )
¶f
slope of tangent to the curve z = f ( x, y0 ) at x = x0 . Similarly denotes the
¶y ( x0 , y0 )
slope of the tangent to the curve z = f ( x0 , y ) at y = y0 .
If the point ( x0 , y0 ) varies, then f x & f y are themselves functions of x & y .
In the case of functions of more than three variables it is necessary to indicate the
variable held constant during the process of differentiation as a suffix to avoid the
confusion.
æ ¶z ö æ ¶z ö
For example, z = f ( x, y, u , v) , then partial derivatives are written as ç ÷ , ç ÷
è ¶u ø x è ¶y øv
and so on. We take an example: x = u + v , y = u - v
æ ¶x ö æ ¶x ö æ ¶y ö æ ¶y ö
ç ÷ = 1 , ç ÷ = 1 , ç ÷ = 1 , ç ÷ = -1
è ¶u øv è ¶v øu è ¶u øv è ¶v øu
Also x + y = 2u and x = 2u - y , then
æ ¶x ö æ ¶x ö
ç ÷ = 2 & ç y ÷ = -1 and so on.
è ¶u ø y è ¶ øu
v Total Differential
In the case of partial derivative we have considered increments Dx & Dy
separately.
Now take ( x, y ) & ( x + Dx, y + Dy ) two points in the domain of definition of z
then if ( z + Dz ) correspond to the point ( x + Dx, y + Dy ) we have
Dz = f ( x + Dx, y + Dy ) - f ( x, y )
If the increment Dz can be expressed as
Dz = aDx + bDy + e1Dx + e 2Dy
and e1 , e 2 ® 0 as Dx, Dy ® 0 , then aDx + bDy is known to be the total differential
of z denoted by dz , and we write
Dz = dz + e1Dx + e 2 Dy
In case when z is differentiable function dz gives very close approximation of Dz .
v Theorem
If z = f ( x, y ) has a total differential at a point ( x, y ) Î D , then
¶z ¶z
a= & b= .
¶x ¶y
Proof
We have
Dz = dz + e1Dx + e 2 Dy where e1 , e 2 ® 0 as Dx, Dy ® 0
Let us suppose that Dy = 0
then Dz = aDx + e1Dx
Taking the limit as Dx ® 0
¶z
=a
¶x
¶z
Similarly we can get =b.
¶y
v Theorem (Fundamental Lemma)

If z = f ( x, y ) has a continuous first order partial derivative in D then z has total
¶z ¶z
differential dz = Dx + Dy at every point ( x, y ) Î D .
¶x ¶y
Proof
Take a point ( x, y ) as a fixed point in the domain D . Suppose x changes alone.
Then we have
Dz = f ( x + Dx, y ) - f ( x, y )
= f x ( x1 , y )Dx ( x < x1 < x + Dx ) ( It is by M. V. Theorem)
Q f x is continuous
\ e1 = f x ( x1, y ) - f x ( x, y ) ® 0 as Dx ® 0
Þ f ( x + Dx, y ) - f ( x, y ) = f x ( x, y )Dx + e1Dx ………. (i)
Now if both x, y changes, we obtain a change Dz in z as
Dz = f ( x + Dx, y + Dy ) - f ( x, y )
= [ f ( x + Dx, y ) - f ( x, y ) ] + [ f ( x + Dx, y + Dy ) - f ( x + Dx, y )]
that is we have expressed Dz as the sum of terms representing the effect of a
change in x alone and subsequent change in y alone.
Now f ( x + Dx, y + Dy ) - f ( x + Dx, y ) = f y ( x + Dx, y1 ) Dy ( y < y1 < y + Dy )
(It is by use of M.V. theorem)
Q f y is given to be continuous
\ e 2 = f y ( x + Dx, y1 ) - f y ( x, y ) ® 0 as Dx, Dy ® 0
Þ f ( x + Dx, y + Dy ) - f ( x + Dx, y ) = f y ( x, y )Dy + e 2 Dy ……….. (ii)
Using (i) & (ii), we have
Dz = f x ( x, y )Dx + f y ( x, y )Dy + e1Dx + e 2Dy where e1 , e 2 ® 0 as Dx, Dy ® 0
which shows that the total differential dz of z exist & is given by
dz = f x ( x, y )Dx + f y ( x, y )Dy ………….. (iii)
Note
(a) For reasons to be explained later; Dx & Dy can be replaced by dx & dy in (iii).
¶z ¶z
Thus we have dz = dx + dy
¶x ¶y
Which is the customary way of writing the differential. The preceding analysis
extends at once to functions of three or more variables. For example, if
¶w ¶w ¶w ¶w
w = f ( x, y , u , v) , then dw = dx + dy + du + dv .
¶x ¶y ¶u ¶v
(b) In the following discussion, the function and their Ist order partial derivatives
will be considered to be continuous in their respective domain of definition.
Example
If z = x 2 - y 2 , then dz = 2 xdx - 2 ydy .
Example:
xy y x xy
If w = , then dw = dx + dy - 2 dz
z x z z
PROBLEMS
¶z ¶z
1) Evaluate and if
¶x ¶y
x ¶z y 2 - x2 ¶z -2 xy
a) z= 2 Ans: = , =
x + y2 ¶x ( ) ¶x ( )
2 2
x2 + y 2 x2 + y 2
¶z ¶z
b) z = x sin xy Ans: = sin xy + xy cos xy , = x 2 cos xy
¶x ¶y
¶z 3 x 2 + y 2 - 2 xz ¶z e x +2 y - y
c) x 3 + xy 2 - x 2 z + z 3 - 2 = 0 Ans: = , =
¶x x 2 - 3z 2 ¶y e x+2 y - y 2
2) Evaluate the indicated partial derivatives:
æ ¶u ö æ ¶v ö
a) ç ÷ and ç ÷ if u = x 2 - y 2 , v = x + 2 y
è ¶x ø y è ¶y ø x
æ ¶x ö æ ¶y ö æ ¶x ö æ ¶y ö 1
b) ç ÷ and ç ÷ if u = x - 2 y , v = u + 2 y Ans: ç ÷ = 1 , ç ÷ =
è ¶u ø y è ¶v øu è ¶u ø y è ¶v øu 2
3) Find the differentials of the following functions
x ydx - xdy
a) z = Ans:
y y2
xdx + ydy
b) z = log x 2 + y 2 Ans:
x2 + y2
æ yö - ydx + ydy
c) z = tan -1 ç ÷ Ans:
èxø x2 + y 2
1 -( xdx + ydy + zdz )
d) u= Ans: 3
x2 + y2 + z 2 ( x2 + y 2 + z 2 ) 2
4) If z = x 2 + 2 xy , find Dz in terms of Dx , Dy for x = 1 , y = 1 .
Ans: Dz = 4Dx + 2Dy + Dx 2 + 2DxDy , dz = 4Dx + 2Dy , dz = 4Dx + 2Dy .
…………………………………..
v Derivative and Differential of functions of functions

In the following discussion, the function and their first order partial derivatives
will be considered to be continuous in their respective domain of definitions.
v Theorem (Chain Rule I)
Let z = f ( x, y ) , x = g (t ) & y = h(t ) be defined in a domain D , then
dz ¶z dx ¶z dy
= × + ×
dt ¶x dt ¶y dt
Proof
Q z = f ( x, y ) , x = g (t ) , y = h(t ) are defined in D , are continuous and have Ist
order partial derivatives.
\ By using the fundamental lemma we have
¶z ¶z
Dz = Dx + Dy + e1Dx + e 2 Dy ………… (i)
¶x ¶y
where e1 , e 2 ® 0 as Dx, Dy ® 0
Also Dx = g (t + Dt ) - g (t )
Dy = h(t + Dt ) - h(t )
Dividing (i) by Dt , we get
Dz ¶z Dx ¶z Dy Dx Dy
= × + × + e1 + e2
Dt ¶x Dt ¶y Dt Dt Dt
Take the limit as Dt ® 0 , we get
dz ¶z dx ¶z dy
= × + × as desired.
dt ¶x dt ¶y dt
v Theorem (Chain Rule II)
Let z = f ( x, y ) , x = g (u , v) , y = h(u , v ) be defined in a domain D and have
continuous first order partial derivative in D , then
¶z ¶z ¶x ¶z ¶y
= × + ×
¶u ¶x ¶u ¶y ¶u
¶z ¶z ¶x ¶z ¶y
and = × + ×
¶v ¶x ¶v ¶y ¶v
Proof
Q the functions are continuous having first order partial derivatives in D ,
therefore by the fundamental lemma, we have
¶z ¶z
Dz = Dx + Dy + e1Dx + e 2 Dy ………… (i)
¶x ¶y
where Dx = g (u + Du , v ) - g (u , v ) , Dy = h(u + Du , v) - h(u , v)
and e1 , e 2 ® 0 as Dx, Dy ® 0 i.e. Du ® 0
Dividing (i) by Du throughout to have
= × + × + e1 + e2
Du ¶x Du ¶y Du Du Du
Taking the limit as Du ® 0 i.e. Dx, Dy ® 0 , we have
¶z ¶z ¶x ¶z ¶y
= × + ×
¶u ¶x ¶u ¶y ¶u
Similarly if Dx = g (u , v + Dv) - g (u , v)
Dy = h(u , v + Dv) - h(u , v)
Then dividing (i) by Dv throughout, we obtain
= × + × + e1 + e2
Dv ¶x Dv ¶y Dv Dv Dv
Taking the limit as Dv ® 0 , we have
¶z ¶z ¶x ¶z ¶y
= × + ×
¶v ¶x ¶v ¶y ¶v
v Note
We have proved in chain rule I, that if z = f ( x, y ) , x = g (t ) , y = h(t ) , then
dz ¶z dx ¶z dy
= × + × ……….. (i)
dt ¶x dt ¶y dt
The three functions of t considered here: x = g (t ) , y = h(t ) , z = f ( g (t ), h(t ) )
dx dy dz
have differentials dx = Dt , dy = Dt , dz = Dt .
dt dt dt
From (i) we conclude that
dz ¶z æ dx ö ¶z æ dy ö
Dt = ç Dt ÷ + ç Dt ÷
dt ¶x è dt ø ¶y è dt ø
¶z ¶z
Þ dz = dx + dy ………… (ii)
¶x ¶y
¶x ¶x
Similarly, dx = Du + Dv
¶u ¶v
¶y ¶y
dy = Du + Dv
¶u ¶v
¶z ¶z
dz = Du + Dv
¶u ¶v
are the corresponding differentials when z = f ( x, y ) , x = g (u , v) , y = h(u , v )
æ ¶z ¶x ¶z ¶y ö æ ¶z ¶x ¶z ¶y ö
Þ dz = ç × + × ÷ Du + ç × + × ÷ Dv
è ¶x ¶u ¶y ¶u ø è ¶x ¶v ¶y ¶v ø
¶z æ ¶x ¶x ö ¶z æ ¶y ¶y ö
= ç Du + Dv ÷ + ç Du + Dv ÷
¶x è ¶u ¶v ø ¶y è ¶u ¶v ø
¶z ¶z
= dx + dy
¶x ¶y
which is again (ii)
The generalization of this permits to conclude that:
The differential formula
¶z ¶z ¶z
dz = dx + dy + dt + ......
¶x ¶y ¶t
which holds when z = f ( x, y , t ,....) and dx = Dx , dy = Dy , dt = Dt ,…..., remain the
true when x, y , t ,..... , and hence z , are all functions of other independent variables
and dx, dy, dt ,....., dz are the corresponding differentials.
As a consequence we can conclude:
Any equation in differentials which is correct for one choice of independent
variables remains true for any other choice. Another way of saying this is that any
equation in differentials treats all variables on an equal basis.
1 3
Thus, if dz = 2dx - 3dy at a given point, then dx = dz + dy is the
2 2
corresponding differentials of x in terms of y and z .
v Example
x2 - 1 2 xy dx - ( x 2 - 1)dy
If z = , then dz =
y y2
¶z 2 x ¶z 1 - x 2
Hence = , =
¶x y ¶y y2
v Example
If r 2 = x 2 + y 2 , then rdr = xdx + ydy
æ ¶r ö x æ ¶r ö y æ ¶x ö r
and ç ÷ = , ç ÷ = , ç ÷ = , etc.
è ¶x ø y r è ¶y ø x r è ¶r ø y x
v Example
æ yö
If z = tan -1 ç ÷ ( x ¹ 0) , then
èxø
1 æ yö xdy - ydx
dz = .d ç ÷ = 2
æ yö èxø
2
x + y2
1+ ç ÷
èxø
and hence
¶z y ¶z x
=- 2 , =
¶x x + y2 ¶y x 2 + y 2
…………………………………..
v Implicit Function
If F ( x, y, z ) is a given function of x, y & z , then the equation F ( x, y , z ) = 0 is a
relation which may describe one or several functions z of x & y .
Thus if x 2 + y 2 + z 2 - 1 = 0 , then
z = 1 - x2 - y 2 or z = - 1 - x2 - y2
Where both functions being defined for x 2 + y 2 £ 1 . Either function is said to be
implicitly defined by the equation x 2 + y 2 + z 2 - 1 = 0 .
Similarly, an equation F ( x, y, z, w) = 0 may define one or more implicit functions
w of x, y , z . If two such equations are given;
F ( x, y , z , w) = 0 , G ( x, y , z , w) = 0 ,
It is in general possible (at least in theory) to reduce the equations by elimination to
the form
w = f ( x , y ) , z = g ( x, y )
i.e. to obtain two functions of two variables. In general, if m equations in n
unknown are given (m < n) , it is possible to solve for m of the variables in terms of
the remaining n - m variables; the number of dependent variables equals the number
of equations
v Example
If 3 x + 2 y + z + 2w = 0
2x + 3y - z - w = 0
then w = f ( x, y ) = -5 x - 5 y & z = g ( x, y ) = 7 x + 8 y
v Example
Suppose that the functions w = f ( x, y ) & z = g ( x, y ) are implicitly defined by
2 x 2 + y 2 + z 2 - zw = 0
x 2 + y 2 + 2 z 2 - 8 + zw = 0
Then taking the differentials, we obtain
4 xdx + 2 ydy + 2 zdz - wdz - zdw = 0 ……… (i)
wdz + zdw + 2 xdx + 2 ydy + 4 zdz = 0 ……… (ii)
Eliminate dw between (i ) and (ii ) to have
6 xdx + 4 ydy + 6 zdz = 0
x 2y
Þ dz = - dx - dy
z 3x
¶z x ¶z 2y
Þ =- , =-
¶x z ¶y 3z
Eliminating of dz from (i ) and (ii ) gives
6 x ( 2 z + w ) dx + 4 y ( z + w ) dy - 6 z 2 dw = 0
x ( 2 z + w) 2 y ( z + w)
Þ dw = dx + dy
z2 3z 2
¶w x (2 x + w) ¶w 2 y ( z + w)
= , =
¶x x 2
¶y z2
v Examples
Suppose that the functions w = f ( x, y ) & z = g ( x, y ) are implicitly define by
F ( x, y, z, w) = 0 and G ( x, y , z ) = 0 , then
Fx dx + Fy dy + Fz dz + Fwdw = 0
and Gx dx + Gy dy + Gz dz + Gwdw = 0
Þ Fz dz + Fwdw = - éë Fx dx + Fy dy ùû
and Gz dz + Gwdw = - éëGx dx + G y dy ùû
Then by crammer rule, we have
Fx dx + Fy dy Fw Fx Fw Fy Fw
-
Gx dx + G y dy Gw Gx Gw G y Gw
dz = =- dx - dy
Fz Fw Fz Fw Fz Fw
Gz Gw Gz Gw Gz Gw
Fx Fw Fy Fw
¶z G Gw ¶z G y Gw
Þ =- x , =-
¶x Fz Fw ¶y Fz Fw
Gz Gw Gz Gw
¶( F , G) ¶( F , G )
¶z ¶ ( x, w) ¶z ¶ ( y, w) ¶( F , G)
Þ =- , =- provided ¹0
¶x ¶( F , G) ¶y ¶( F , G ) ¶ ( z , w)
¶ ( z , w) ¶ ( z , w)
Similarly, we have
Fz Fx dx + Fy dy
Gz Gx dx + G y dy
dw = -
Fz Fw
Gz Gw
¶w ¶w
and we can find and in the same manner.
¶x ¶y
v Particular Cases
i) One equation in 2 unknowns i.e. F ( x, y ) = 0
Þ Fx dx + Fy dy = 0
Þ
dy
dx
F
=- x
Fy
(F y ¹ 0)
ii) One equation in 3 unknowns i.e. F ( x, y , z ) = 0

Fx dx + Fy dy + Fz dz = 0
¶z F ¶z F
Þ =- x , = - y ( Fz ¹ 0)
¶x Fz ¶y Fz
iii) 2 equations in 3 unknown
F ( x, y , z ) = 0 , G ( x , y , z ) = 0
¶( F , G ) ¶( F , G )
¶z ¶ ( y, x ) ¶z ¶ ( w, y )
=- , =-
¶x ¶( F , G ) ¶y ¶( F , G )
¶ ( y, z ) ¶ ( z , w)
………………………………
v Example
Find the partial derivatives w.r.t x & y , when
u + 2v - x 2 + y 2 = 0
2u - v - 2 xy = 0
Solution
Take the differentials
du + 2 dv - 2 x dx + 2 y dy = 0 ………… (i)
2 du - dv - 2 x dy - 2 y dx = 0 …………. (ii)
Eliminating dv between (i ) and (ii ) , we have
5du - (2 x + 4 y )dx + (2 y - 4 x )dy = 0
1 1
Þ du = ( 2 x + 4 y ) dx - ( 2 y - 4 x ) dy
5 5
¶u 1 ¶u 1
Þ = ( 2x + 4 y ) & = - (2 y - 4x )
¶x 5 ¶y 5
Eliminating du between (i ) and (ii ) , we get
5dv - ( 4 x - 2 y ) dx + ( 4 y + 2 x ) dy = 0
1 1
Þ dv = ( 4 x - 2 y ) dx - ( 4 y + 2 x ) dy
5 5
¶v 1 ¶v 1
Þ = (4x - 2 y ) & = - (4 y + 2x)
¶x 5 ¶y 5
v Question
Give that
2 x + y - 3 z - 2u = 0
& x + 2y + z + u = 0
æ ¶x ö æ ¶y ö æ ¶z ö æ ¶y ö
Find ç ÷ , ç ÷ , ç ÷ , ç ÷
è ¶y ø z è ¶x øu è ¶u ø x è ¶z ø x
Solution
Take the differentials
2 dx + dy - 3dz - 2du = 0 ………… (i)
dx + 2dy + dz + du = 0 ………...… (ii)
Eliminating du between (i) and (ii), we have
4 dx + 5dy - dz = 0 ………. (iii)
5 1
Þ dx = - dy + dz
4 4
æ ¶x ö 5
Þ ç ÷ =-
è ¶y ø z 4
From (iii), we have
5dy = dz - 4dx
1 4
Þ dy = dz - dx
5 5
æ ¶y ö 1
Þ ç ÷ =
è ¶z ø x 5
Eliminating dz between (i ) & (ii ) , we get
5dx + 7 dy + du = 0
5 1
Þ dy = - dx - du
7 7
æ ¶y ö 5
Þ ç ÷ =-
è ¶x øu 7
Now eliminating dy between (i ) & (ii ) , we get
-3dx - 5dz - 3du = 0
3 3
Þ dz = - dx - du
5 5
æ ¶z ö 3
Þ ç ÷ =-
è ¶u ø x 5
v Question
Given that
x 2 + y 2 + z 2 - u 2 + v 2 = 1 …………... (i)
x 2 - y 2 + z 2 + u 2 + 2v 2 = 2 ………… (ii)
a) Find du & dv in terms of dx, dy & dz at the point
x =1 , y =1 , z = 2 , u = 3 & v = 2 .
æ ¶u ö æ ¶v ö
b) Find ç ÷ , ç ÷ at the point given above.
è ¶x ø( y , z ) è ¶y ø( x , z )
c) Find approximately the values of u & v for x = 1 × 1 , y = 1.2 , z = 1.8
Solutions
Differential gives
2 xdx + 2 ydy + 2 zdz - 2udu + 2vdv = 0 ……… (iii)
2 xdx - 2 ydy + 2 zdz + 2udu + 2vdv = 0 ……… (iv)
a) Putting x = 1 , y = 1 , z = 2 , u = 3 & v = 2 in (iii) & (iv), we obtain
2 dx + 2dy + 4dz - 6du + 4dv = 0 ………… (v)
& 2 dx - 2dy + 4dz + 6du + 8dv = 0 ………… (vi)
Adding gives
12dv = - ( 4dx + 8dz )
1
Þ dv = - ( dx + 0 × dy + 2dz )
3
Similarly eliminating dv between (v ) and (vi ) , we get
1
du = ( dx + 3dy + 2dz )
9
1
b) Q du = ( dx + 3dy + 2dz )
9
æ ¶u ö 1
\ ç ÷ =
è ¶x ø y , z 9
1
& Q dv = - ( dx + 0 × dy + 2dz )
3
æ ¶v ö
\ ç ÷ =0
è ¶y ø x , z
…………………………………..
v Question
Find the transformation of x = r cosq , y = r sin q from rectangular to polar
coordinates. Verify the relations
a) dx = cosq dr - r sinq dq
dy = sin q dr + r cosq dq
b) dr = cosq dx + sin q dy
sin q cosq
dq = - dx + dy
r r
æ ¶x ö æ ¶x ö ¶ (r ,q ) 1
c) ç ÷ = cosq , ç ÷ = secq , =
è ¶r øq è ¶r ø y ¶ ( x, y ) r
Solutions
Given that x = r cosq & y = r sin q
a) Differential gives
dx = cosq dr - r sinq dq ………. (i)
dy = sin q dr + r cosq dq ………. (ii)
b) Multiplying (i) by cosq & (ii) by sinq and adding, we get
dr = cosq dx + sin q dy
Now multiply (i) by sinq & (ii) by cosq and subtract to obtain
sin q cosq
dq = - dx + dy
r r
c) Given x = r cosq
æ ¶x ö
Þ ç ÷ = cosq
è ¶r øq
We have already shown that dr = cosq dx + sin q dy
dr
Which can be written as dx = - tan q dy
cosq
æ ¶x ö
Þ ç ÷ = secq
è ¶r ø y
¶x ¶x
¶ ( x, y ) ¶r ¶q cosq - r sin q
= = = r cos 2 q + r sin 2 q = r
¶ (r ,q ) ¶y ¶y sin q r cosq
¶r ¶q
¶r ¶r
cosq sin q
¶ (r ,q ) ¶x ¶y 1 1 1
and = = sin q cosq = cos 2 q + sin 2 q =
¶ ( x, y ) ¶q ¶q - r r r
r r
¶x ¶y
v Question
Given that x 2 - y 2 cos uv + z 2 = 0
x 2 + y 2 - sin uv + 2 z 2 = 2
and xy - sin u cos v + z = 0
æ ¶x ö æ ¶x ö p
Find ç ÷ , ç ÷ at x = 1 , y = 1 , u = , v = 0 , z = 0
è ¶u øv è ¶v øu 2
Solution
Differential gives
2 x dx - 2 y cos uv dy + y 2 sin uv × udv + y 2 sin uv × vdu + 2 zdz = 0 ……. (i)
2 x dx + 2 y dy - cos uv × udv - cos uv × vdu + 4 zdz = 0 ……… (ii)

& x dy + y dx - cos u × cos v du + sin u × sin v dv + dz = 0 ……... (iii)
At the given point, these equations reduce to
2 dx - 2dy = 0 …………..…. (iv)
p
2 dx + 2dy - dv = 0 ……… (v)
2
& dx + dy + dz = 0 …………... (vi)
Adding (iv) & (v), we have
p
4 dx - dv = 0
2
p æ ¶x ö æ ¶x ö p
Þ dx = dv + 0 × du Þ ç ÷ = 0 , ç ÷ =
8 è ¶u øv è ¶v øu 8
v Question
æ ¶u ö
Find ç ÷ if x 2 - y 2 + u 2 + 2v 2 = 1
è ¶x ø y
x 2 + y 2 - u 2 - v2 = 2
Solution
Taking the differentials, we have
2 x dx - 2 y dy + 2u du + 4v dv = 0
2 x dx + 2 y dy - 2u du - 2v dv = 0
Eliminating dv , we get
6 x dx + 2 y dy - 2u du = 0
3x y
Þ du = dx + dy
u u
æ ¶u ö 3x
Þ ç ÷ =
è ¶x ø y u
v Question
Given the transformation
x = u - 2v
y = 2u + v
a) Write the equations of the inverse transformation
b) Evaluate the Jacobian of the transformation and that of the inverse
transformation.
Solution
a) From the equations, we have
1 2
u= x+ y
5 5
2 1
v=- x+ y
5 5
which are the equations of the inverse transformation.
¶x ¶x
¶ ( x, y ) ¶u ¶v
b) Jacobian of the given transformation = =
¶ (u , v) ¶y ¶y
¶u ¶v
1 -2
= =5
2 1
¶u ¶u
¶ (u , v) ¶x ¶y
Jacobian of the inverse transformation = =
¶ ( x, y ) ¶v ¶v
¶x ¶y
1 2
5 5 1
= =
2 1 5
-
5 5
v Question
¶ ( x, y )
Given the transformation x = f (u , v) , y = g (u , v) with Jacobian J = , show
¶ (u , v)
that for the inverse transformation one has
¶u 1 ¶y ¶u 1 ¶x ¶v 1 ¶y ¶u 1 ¶x
= , =- , =- , =
¶x J ¶v ¶y J ¶v ¶x J ¶u ¶y J ¶u
Solution
The given equations are
f (u , v ) - x = 0 ………. (i)
g (u , v) - y = 0 ………. (ii)
Differentiating w.r.t. x , we get
¶u ¶v
fu + fv - 1 = 0
¶x ¶x
¶u ¶v
gu + gv -0=0
¶x ¶x
Solving these equations by Crammer’s rule, we have
-1 f v
¶u 0 gv gv 1 ¶y æ ¶y ö
=- = = çQ = gv ÷
¶x fu fv J J ¶v è ¶v ø
gu gv
fu -1
¶v g 0 g 1 ¶y
=- u =- u =-
¶x J J J ¶u
Differentiating (i) & (ii) w.r.t. y , we have
¶u ¶v
fu + fv -0=0
¶y ¶y
¶u ¶v
gu + gv -1 = 0
¶y ¶y
Solving these equations by Crammer’s rule, we get
0 fv
¶u -1 g v f 1 ¶x
=- =- v =-
¶y J J J ¶v
fu 0
¶v g -1 f 1 ¶x
=- u = u =
¶y J J J ¶u
v Question
Given the transformation
x = u 2 - v2
y = 2uv
a) Compute its Jacobian.
æ ¶u ö æ ¶v ö
b) Evaluate ç ÷ & ç ÷
è ¶x ø y è ¶x ø y
Solution
The given equations can be written as
u 2 - v 2 - x = 0 ……….. (i)
2uv - y = 0 ……...…… (ii)
Differentiating (i) & (ii) partially w.r.t. x , we have
¶u ¶v
2u - 2v - 1 = 0 …………. (iii)
¶x ¶x
¶u ¶v
2v + 2u - 0 = 0 …………. (iv)
¶x ¶x
¶x ¶x
¶ ( x, y ) ¶u ¶v 2u -2v
J= = = = 4(u 2 + v 2 )
¶ (u , v ) ¶y ¶y 2v 2u
¶u ¶v
Solving (iii) & (iv) by Crammer’s rule, we have
-1 -2v
æ ¶u ö 0 2u 2u u
ç ÷ =- = =
è ¶x ø y 4(u + v 2 ) 2(u + v 2 )
2 2
J
2u -1
æ ¶v ö 2v 0 -2v -v
ç ÷ =- = =
è ¶x ø y 4(u 2 + v 2 ) 2(u + v 2 )
2
J
Note
æ ¶u ö æ ¶v ö
ç ¶y ÷ & ç ¶y ÷ can be determined in the same manner.
è øx è øx
v Question
Prove that if F ( x, y , z ) = 0 , then
æ ¶z ö æ ¶x ö æ ¶y ö
ç ÷ × ç ÷ × ç ÷ = -1
è ¶x ø y è ¶y ø z è ¶z ø x
Solution
F ( x, y , z ) = 0
Þ Fx dx + Fy dy + Fz dz = 0
Fy Fz æ ¶x ö F
Þ dx = - dy - dz Þ ç ÷ =- y
Fx Fx è ¶y ø z Fx
Fx F æ ¶y ö F
& dy = - dx - z dz Þ ç ÷ =- z
Fy Fy è ¶z ø x Fy
Fx F æ ¶z ö F
dz = - dx - y dy Þ ç ÷ =- x
Fz Fz è ¶x ø y Fz
Hence
æ ¶z ö æ ¶x ö æ ¶y ö æ Fx ö æ Fy ö æ Fz ö
ç ÷ ×ç ÷ ×ç ÷ = ç - ÷×ç - ÷ × çç - ÷÷ = -1
è ¶x ø y è ¶y ø z è ¶z ø x è Fz ø è Fx ø è Fy ø
v Question
Prove that, if x = f (u , v) , y = g (u , v) , then
æ ¶x ö æ ¶u ö æ ¶y ö æ ¶v ö
ç ÷ ç ÷ =ç ÷ ç ÷
è ¶u øv è ¶x ø y è ¶v øu è ¶y ø x
æ ¶x ö æ ¶v ö æ ¶u ö æ ¶y ö
and ç ÷ ç ÷ =ç ÷ ç ÷
è ¶v øu è ¶x ø y è ¶y ø x è ¶u øv
æ ¶x ö æ ¶y ö
also that ç ÷ ç ÷ = 1
è ¶y øu è ¶x øu
Solution
Q f (u , v ) - x = 0
g (u , v) - y = 0
æ ¶u ö g æ ¶v ö gu
\ ç ÷ = v , ç ÷ =-
è ¶x ø y J è ¶x ø y J
æ ¶u ö fv æ ¶v ö f
ç ÷ =- , ç ÷ = u as already shown
è ¶y ø x J è ¶y ø x J
Taking differentials of the given equations, we have
f u du + f v dv - dx = 0
g u du + g v dv - dy = 0
Þ dx = f u du + f v dv ………… (i)
dy = g u du + g v dv ……..…. (ii)
æ ¶x ö æ ¶x ö
Þ ç ÷ = fu , ç ÷ = f v
è ¶u øv è ¶v øu
æ ¶y ö æ ¶y ö
ç ÷ = gu , ç ÷ = gv
è ¶u øv è ¶v øu
æ ¶x ö æ ¶u ö æ ¶y ö æ ¶v ö
Now ç ÷ ×ç ÷ = ç ÷ ×ç ÷
è ¶u øv è ¶x ø y è ¶v øu è ¶y ø x
g f
Þ f u × v = g v × u , which is true
J J
Similarly, we have the second relation.
Eliminating dv between (i ) & (ii ) , we get
( fu × gv - fv × gu ) du - gv dx + f v dy = 0
f g - fv gu f
Þ dx = u v × du + v dy
gv gv
g f g - f v gu
and dy = v dx - u v du
fv fv
æ ¶x ö f æ ¶y ö g
Þ ç ÷ = v & ç ÷ = v
è ¶y øu g v è ¶x øu f v
æ ¶x ö æ ¶y ö f g
Þ ç ÷ ×ç ÷ = v × v =1
è ¶y øu è ¶x øu g v f v
v Question
Given that x = f (u , v, w) , y = g (u , v, w) , z = h(u , v, w) with the Jacobian
¶ ( x, y , z )
J= , show that for the inverse transformation one has
¶ (u , v, w)
¶u 1 ¶ ( y, z ) ¶u 1 ¶ ( z, x ) ¶u 1 ¶ ( x, y )
i) = , = , =
¶x J ¶ (v, w) ¶y J ¶ (v, w) ¶z J ¶ (v, w)
¶v 1 ¶ ( y, z ) ¶v 1 ¶ ( z , x ) ¶v 1 ¶ ( x, y )
ii) = , = , =
¶x J ¶ ( w, u ) ¶y J ¶ ( w, u ) ¶z J ¶ ( w, u )
¶w 1 ¶ ( y, z ) ¶w 1 ¶ ( z , x) ¶w 1 ¶ ( x, y )
iii) = , = , =
¶x J ¶ (u, v) ¶y J ¶ (u , v ) ¶z J ¶ (u , v )
Solution
We have f (u , v, w) - x = 0
g (u , v, w) - y = 0
h (u , v, w) - z = 0
Differentiating w.r.t. to x , we get
¶u ¶v ¶w
fu + fv + f w -1 = 0
¶x ¶x ¶x
¶u ¶v ¶w
gu + gv + gw -0=0
¶x ¶x ¶x
¶u ¶v ¶w
hu + hv + hw -0=0
¶x ¶x ¶x
By Crammer’s rule, we have
-1 f v f w
0 gv g w gv g w
¶u 0 hv hw hv hw 1 ¶ ( g , h) 1 ¶( y, z)
=- = = =
¶x J J J ¶ ( v, w) J ¶ (v, w)
fu -1 f w
gu 0 gw gu gw
¶v h 0 hw h hw 1 ¶ ( g , h) 1 ¶( y, z)
=- u =- u = =
¶x J J J ¶ ( w, u ) J ¶ ( w, u )
fu f v -1
gu gv 0 gu g v
¶w h hv 0 h hv 1 ¶ ( g , h) 1 ¶ ( y, z )
=- u = u = =
¶x J J J ¶ (u , v) J ¶ (u , v)
We can find the other relations in the same way by differentiating given relation
w.r.t. y and w.r.t. z respectively.
………………………………….
v Partial Derivative of Higher Order

¶z ¶z
Let a function z = f ( x, y ) be given. Then its two partial derivatives &
¶x ¶y
are themselves functions of x & y .
¶z ¶z
i.e. = f x ( x, y ) , = f y ( x, y )
¶x ¶y
Hence each can be differentiable w.r.t. x & y .
Thus, we obtain four partial derivatives
¶2 z ¶2z
= f xx ( x, y ) , = f xy ( x, y )
¶x 2 ¶x ¶y
¶2z ¶2z
= f yx ( x, y ) , = f yy ( x, y )
¶y ¶x ¶y 2
¶2 z ¶z ¶2 z
is the result of differentiating w.r.t. x , where is the result of
¶x 2 ¶x ¶y ¶x
¶z
differentiating w.r.t. y . If all the derivatives concerned are continuous in the
¶x
¶2z ¶2z
domain considered, then = i.e. order of differentiation is immaterial.
¶x ¶y ¶y ¶x
Third and higher order partial derivatives are defined in the same manner and
under appropriate assumptions of continuity the order of differentiation does not
matter.
v Laplacian of z
If z = f ( x, y ) , then the Laplacian of z is denoted by Ñ 2 z is the expression
¶2 z ¶2 z
Ñ z= 2 + 2
2
¶x ¶y
if w = f ( x, y , z ) , the Laplacian of w is the expression
¶ 2 w ¶ 2w ¶ 2w
Ñ w= 2 + 2 + 2
2
¶x ¶y ¶z
The symbol “ Ñ ” is a vector differential operator define as
¶ ¶ ˆ ¶ ˆ
Ñ = iˆ + j+ k
¶x ¶y ¶x
We then have symbolically
¶2 ¶2 ¶2
Ñ = Ñ×Ñ = 2 + 2 + 2
2
¶x ¶y ¶z
v Harmonic Function
If z = f ( x, y ) has continuous second order derivatives in a domain D and Ñ 2 z = 0
in D , then z is said to be Harmonic in D . The same term is used for the function of
three variables which has continuous 2nd derivatives in a domain D in space and
whose Laplacian is zero in D . The two equations for harmonic functions
¶2 z ¶2 z
Ñ z= 2 + 2 =0
2
¶x ¶y
¶ 2 w ¶ 2 w ¶ 2w
Ñ w= 2 + 2 + 2 =0
2
¶x ¶y ¶z
are known as the Laplace equations in two and three dimensions respectively.
v Bi-Harmonic Equations
Another important combination of derivatives occurs in the equation
¶4 z ¶4 z ¶4 z
+ 2 + =0
¶x 4 ¶x 2¶y 2 ¶y 4
which is known to be the Bi-harmonic equation. This combination can be
expressed in terms of Laplacian as
( )
Ñ2 Ñ2 z = Ñ4 z = 0
The solutions of Ñ 4 z = 0 are termed as Pri-harmonic functions.
v Higher Derivatives of Functions of Functions
(1) Let z = f ( x, y ) and x = g (t ) , y = h(t ) so that z can be expressed in terms of t
alone. Then
dz ¶z dx ¶z dy
= + ……….. (i)
dt ¶x dt ¶y dt
d 2 z d æ dz ö ¶z d 2 x dx d æ ¶z ö ¶z d 2 y dy d æ ¶z ö
= ç ÷= + ç ÷+ + ç ÷ ……… (ii)
dt 2 dt è dt ø ¶x dt 2 dt dt è ¶x ø ¶y dt 2 dt dt è ¶y ø
Using (i), we have
d æ ¶z ö ¶ 2 z dx ¶ 2 z dy
ç ÷= +
dt è ¶x ø ¶x 2 dt ¶y ¶x dt
d æ ¶z ö ¶ 2 z dx ¶ 2 z dy
& ç ÷= +
dt è ¶y ø ¶x¶z dt ¶y 2 dt
Putting these values in (ii ) , we have
2 2
d 2 z ¶z d 2 x ¶ 2 z æ dx ö ¶ 2 z dx dy ¶ 2 z æ dy ö ¶z d 2 y
= + ç ÷ +2 × + ç ÷ +
dt 2 ¶x dt 2 ¶x 2 è dt ø ¶x ¶y dt dt ¶y 2 è dt ø ¶y dt 2
(2) If z = f ( x, y ) and x = g (u , v) , y = h(u , v ) , then
¶z ¶z ¶x ¶z ¶y
= × + × ……….. (iii)
¶u ¶x ¶u ¶y ¶u
¶z ¶z ¶x ¶z ¶y
= × + × ……….. (iv)
¶v ¶x ¶v ¶y ¶v
¶ 2 z ¶z ¶ 2 x ¶ æ ¶z ö ¶x ¶z æ ¶ 2 y ö ¶y ¶ æ ¶z ö
= × + ç ÷× + ç ÷+ ×
¶u 2 ¶x ¶u 2 ¶u è ¶x ø ¶u ¶y è ¶u 2 ø ¶u ¶u çè ¶y ÷ø
…………. (iv)
Using (iii), we have

¶ æ ¶z ö ¶ 2 z ¶x ¶ 2 z ¶y
ç ÷= × + ×
¶u è ¶x ø ¶x 2 ¶u ¶y ¶x ¶u
¶ æ ¶z ö ¶ 2 z ¶x ¶ 2 z ¶y
= +
¶u èç ¶y ÷ø ¶x ¶y ¶u ¶y 2 ¶u
and
Putting these values in (iv ) , we get

2
¶ 2 z ¶z ¶ 2 x ¶ 2 z æ ¶x ö ¶ 2 z ¶x ¶y ¶ 2 z æ ¶y ö ¶z ¶ 2 y
= + ç ÷ + 2 × + ç ÷ + ×
¶u 2 ¶x ¶u 2 ¶x 2 è ¶u ø ¶x ¶y ¶u ¶u ¶y 2 è ¶u ø ¶y ¶u 2
¶2z ¶2 z
We can find the values of & in the same manner.
¶u ¶v ¶v 2
………………………………………..
v The Laplacian in Polar, Cylindrical and Spherical Co-ordinate

We consider first the two-dimensional Laplacian
¶ 2 w ¶ 2w
Ñ w= 2 + 2
2
¶x ¶y
and its expression in terms of polar co-ordinates r & q .
Thus we are given w = f ( x, y ) and x = r cosq , y = r sin q and we wish to express
Ñ 2 w in terms of r , q and derivatives of w with respect to r and q . The solution is
as follows. One has
¶w ¶w ¶r ¶w ¶q
= × + ×
¶x ¶r ¶x ¶q ¶x
¶w ¶w ¶r ¶w ¶q
= × + × by chain rule
¶y ¶r ¶y ¶q ¶y
¶r ¶q ¶r ¶q
To evaluate , , , , we use the equations
¶x ¶x ¶y ¶y
dx = cosq dr - r sin q dq
dy = sin q dr + r cosq dq
These can be solved for dr and dq by determinants or by elimination to give
dr = cosq dx + sin q dy
sin q cosq
dq = - dx + dy
r r
¶r ¶r ¶q sin q ¶q cosq
Hence = cosq , = sin q , =- and =
¶x ¶y ¶x r ¶y r
¶w ¶w
Putting these values above in expressions of & , we have
¶x ¶y
¶w ¶w sin q ¶w ü
= cosq -
¶x ¶r r ¶q ïï
¶w cosq ¶w ýï
.............. (i )
¶w
= sin q +
¶y ¶r r ¶q ïþ
These equations provide general rules for expressing derivatives w.r.t. x or y in
terms of derivatives w.r.t. r and q . By applying the first equation to the function
¶w
, one finds that
¶x
¶ 2 w ¶ æ ¶w ö ¶ æ ¶w ö sin q ¶ æ ¶w ö
= ç ÷ = cosq ç ÷- ç ÷
¶x 2
¶x è ¶x ø ¶r è ¶x ø r ¶q è ¶x ø
By (i) this can be written as follows:
¶2w ¶æ ¶w sin q ¶w ö sin q ¶ æ ¶w sin q ¶w ö
= cosq ç cosq - ÷- ç cosq - ÷
¶x 2
¶r è ¶r r ¶q ø r ¶q è ¶r r ¶q ø
The rule for differentiation of a product gives finally
¶2w ¶ 2 w 2sin q cosq ¶ 2 w sin 2 q ¶ 2 w
= cos q × 2 -
2
× + 2 × 2
¶x 2 ¶r r ¶r ¶q r ¶q
sin 2 q ¶w 2sin q cosq ¶w
+ × + × ……. (ii)
r ¶q r2 ¶q
In the same manner one finds
¶ 2 w ¶ æ ¶w ö ¶æ ¶w cosq ¶w ö cosq ¶ æ ¶w cosq ¶w ö
= ç ÷ = sin q ç sin q + ÷+ ç sin q + ÷
¶y 2
¶y è ¶y ø ¶r è ¶r r ¶q ø r ¶q è ¶r r ¶q ø
¶ 2 w 2sin q cosq ¶ 2 w cos 2 q ¶ 2 w

= sin q × 2 +
2
× + × 2
¶r r ¶r ¶q r2 ¶q
cos 2 q ¶w 2sin q cosq ¶w
+ × - × …… (iii)
r ¶r r2 ¶q
Adding (ii) & (iii), we conclude
¶ 2 w ¶ 2 w ¶ 2 w 1 ¶ 2 w 1 ¶w
Ñ w= 2 + 2 = 2 + 2
2
+ ………. (iv)
¶x ¶y ¶r r ¶q 2 r ¶r
This is the desired result.
Equation (iv) at once permits one to write the expression for the 3-demensional
Laplacian in cylindrical co-ordinates for the transformation of coordinates
x = r cosq , y = r sin q , z = z
involves only x & y . In the same way as above, we have
¶ 2 w ¶ 2w ¶ 2w
Ñ2w = + +
¶x 2 ¶y 2 ¶z 2
¶ 2 w 1 ¶ 2 w 1 ¶w ¶ 2 w
= 2 + 2 + +
¶r r ¶q 2 r ¶r ¶z 2
v Laplacian in Spherical Polar Coordinates
The transformation form rectangular to spherical polar coordinates is
x = r sin j cosq , y = r sin j sin q , z = r cos j
Writing r = r sin j , we have
x = r cosq , y = r sin q , z = z
Which can be considered as a transformation from rectangular to cylindrical
coordinates (r ,q , z )
We have
¶ 2 w 1 ¶ 2 w 1 ¶w ¶ 2 w
Ñ2w = 2 + 2 + + ……….. (i)
¶r r ¶q 2 r ¶r ¶z 2
where z = r cos j ü
ý ………. (ii)
r = r sin j þ
We have transformation from ( x, y ) to (r ,q ) as
¶ 2 w ¶ 2 w ¶ 2 w 1 ¶ 2 w 1 ¶w
+ = + +
¶x 2 ¶y 2 ¶r 2 r 2 ¶q 2 r ¶q
Now if we take transformation from ( z , r ) to ( r ,j ) , then
¶ 2 w ¶ 2 w ¶ 2 w 1 ¶ 2 w 1 ¶w
Þ + = + +
¶z 2 ¶r 2 ¶r 2 r 2 ¶j 2 r ¶j
¶w ¶w ¶r ¶w ¶j
Also = × + ×
¶r ¶r ¶r ¶j ¶r
r
Where r 2 = z 2 + r 2 , tan j =
z
¶r ¶r r r sin j
Þ 2r = 2r Þ = = = sin j
¶r ¶r r r
¶j 1 ¶j cos 2 j cos 2 j cos j
& sec 2 j × = Þ = = =
¶r z ¶r z r cos j r
¶w ¶w ¶w cos j
Þ = × sin j + × ………… (iv)
¶r ¶r ¶j r
Substituting (iii) & (iv) in (i), we have
æ ¶ 2 w ¶ 2 w ö 1 ¶ 2 w 1 ¶w
Ñ2w = ç 2 + 2 ÷ + 2 +
è ¶ z ¶r ø r ¶ q 2
r ¶r
¶ 2 w 1 ¶ 2 w 1 ¶w 1 ¶ 2w 1 æ ¶w ¶w cos j ö
= 2+ 2 + + 2 2 × 2 + ç sin j +
¶r r ¶j 2
r ¶r r sin j ¶q r sin j è ¶r ¶j r ÷ø
¶2w 1 ¶ 2w 1 ¶w 1 ¶ 2 w 1 ¶w cot j ¶w
= 2+ × 2+ × + 2 2 × 2+ × + 2 ×
¶r r2 ¶j r ¶r r sin j ¶q r ¶r r ¶j
¶2w 1 ¶ 2w 2 ¶w 1 ¶ 2 w cot j ¶w
= 2+ × 2+ × + 2 2 × 2+ 2 ×
¶r r2 ¶j r ¶r r sin j ¶q r ¶j
v Question
If u & v are functions of x & y defined by the equations
xy + uv = 1 , xu + yv = 1
¶ 2u
then find .
¶x 2
Solution
y dx + x dy + v du + u dv = 0 ……… (i)
u dx + v dy + x du + y dv = 0 ……... (ii)
Eliminating dv between (i) & (ii)
( )
y 2 - u 2 dx + ( xy - uv ) dy + ( vy - ux ) du = 0
u 2 - y2 uv - xy
Þ du = dx + dy
vy - ux vy - ux
¶u u 2 - y 2 u 2 - y 2
Þ = = ( using given eq. )
¶x vy - ux 1 - 2ux
¶u é ¶u ù
(1 - 2ux) × 2u × - (u 2 - y 2 ) ê(-2u ) - 2 x ú
¶u
2
¶x ë ¶x û
Þ =
¶x (1 - 2ux )
2 2
v Question
¶ 2w ¶ 2w
Find , when
¶x 2 ¶y 2
1
i) w =
x2 + y2
y
ii) w = tan -1
x
x2 - y 2
iii) w = e
v Question
Show that the following functions are harmonic in x & y
i) e x cos y
ii) x 3 - 3xy 2
iii) log x 2 + y 2
………………………………………
v Sufficient Condition for the Validity of Reversal in the Order

of Derivation
We now prove two theorems which lay sufficient conditions for the equality of f xy
and f yx .
v Schawarz’s Theorem
If (a, b) be a point of the domain of a function f ( x, y ) such that
i) f x ( x, y ) exists in a certain nhood of (a, b) .
ii) f xy ( x, y ) is continuous at (a, b) .
then f yx (a, b) exists and is equal to f xy (a, b) .
Proof
The given conditions imply that there exists a certain nhood of (a, b) at every point
( x, y ) of which f x ( x, y ) , f y ( x, y ) and f xy ( x, y ) exist. Let (a + h, b + k ) be any point
of this nhood. We write
f (h, k ) = f (a + h, b + k ) - f (a + h, b) - f (a, b + k ) + f (a, b)
g ( y ) = f (a + h, y ) - f (a, y )
so that f (h, k ) = g (b + k ) - g (b) ……….. (i)
Q f y exists in a nhood of ( a, b ) , the function g ( y ) in derivable in [b, b + k ] , and,
therefore, by applying the M.V. theorem to the expression on R.H.S of (i ) , we have
f (h, k ) = kg ¢(b + q k ) ( 0 < q < 1)
= k ( f y (a + h, b + q k ) - f y (a, b + q k ) ) ……….. (ii)
Again since f xy exists in a nhood of (a, b) , the function f y ( x, b + q k ) of x is
derivable w.r.t. x in interval (a, a + h) and, therefore, by applying the M.V. theorem
to the right of (ii ) , we have
f (h, k ) = hk f xy (a + q ¢h, b + q k ) ( 0 < q ¢ < 1)
1 æ f ( a + h, b + k ) - f ( a , b + k ) f ( a + h , b ) - f ( a , b ) ö
or ç - ÷ = f xy (a + q ¢h, b + q k )
kè h h ø
Since f x ( x, y ) exists in a nhood of (a, b) , this gives when h ® 0 ,
f x (a , b + k ) - f x ( a , b )
= lim f xy (a + q ¢h, b + q k )
k h®0
Let, now, k ® 0 . Since f xy ( x, y ) is continuous at (a, b) , we obtain

f yx (a, b) = lim lim f xy (a + q ¢h, b + q k ) = f xy (a, b)
k ®0 h ®0
v Young’s Theorem
If (a, b) be a point of the domain of definition of a function f ( x, y ) such that
f x ( x, y ) and f y ( x, y ) are both differentiable at (a, b) , then
f xy (a, b) = f yx (a, b)
Proof
The differentiability of f x and f y at (a, b) implies that they exist in a certain
nhood of (a, b) and that f xx , f yx , f xy , f yy exist at (a, b) .
Let (a + h, b + h) be a point of this nhood. We write
f (h, h) = f (a + h, b + h) - f (a + h, b) - f (a , b + h) + f (a, b)
g ( y ) = f (a + h, y ) - f (a, y )
so that f (h, h) = g (b + h) - g (b) ………. (i )
Since f y exists in a nhood of (a, b) , the function g ( y ) is derivable in (b, b + h) ,

and, therefore, by applying the M.V. theorem to the expression on the right of (i ) ,
we have
f (h, h) = h g ¢(b + q h) ( 0 < q < 1)
= h ( f y (a + h, b + q h) - f y (a, b + q h) ) ……….. (ii )
Since f y ( x, y ) is differentiable at (a, b) , we have, by definition,
f y (a + h, b + q h) - f y (a, b) = h f xy (a, b) + q h f yy (a, b)
+ h j1 (h, h) + q hy 1 (h, h) …… (iii)
and f y (a, b + q h) - f y (a, b) = q h f yy (a, b) + q hy 2 (h, h) ….….. (iv )
where j1 , y 1 , y 2 all ® 0 as h ® 0
From (ii ) , (iii ) and (iv ) , we obtain
f (h, h)
= f xy (a, b ) + f1 (h, h) + qy 1 (h, h) - q y 2 (h, h) …..…. (v )
h2
By a similar argument and on considering
g ( x ) = f ( x, b + k ) - f ( x, b )
We can show that
f (h, h)
= f yx (a, b) + y 3 (h, h) + q ¢j 2 (h, h) - q ¢j 3 (h, h) ……….. (vi )
h2
where j 2 , j 3 ,y 3 all ® 0 as h ® 0
Equating the right hand side of (v ) and (vi ) and making h ® 0 , we obtain
f xy (a, b) = f yx (a, b)
………………………………….
v Maxima and Minima for Functions of Two Variables

Let ( x0 . y0 ) be the point of the domain of a function f ( x, y ) , then f ( x0 , y0 ) said to
an extreme value of the function f ( x, y ) , if the expression
Df = f ( x0 + h, y0 + h) - f ( x0 , y0 )
preserves its sign for all h and k .
The extreme value of f ( x0 , y0 ) being called a maximum or a minimum value
according as this difference is positive or negative respectively.
Necessary Condition
The Necessary Condition for f ( x0 , y0 ) to be an extreme value of function f ( x, y )
is that f x ( x0 , y0 ) = 0 = f y ( x0 , y0 ) , provided that these partial derivatives exist.
It is to be noted that it is impossible to determine the nature of a critical point by
studying the function f ( x, y0 ) and f ( x0 , y ) .
e.g. Let f ( x, y ) = 1 + x 2 - y 2
then f (0, y ) = 1 - y 2 Þ f ¢(0, y ) = -2 y = 0 Þ (0,0) is a turning point.
Now f ¢¢(0, y ) = -2 Þ (0,0) is a point of maximum value.
But f ( x,0) = 1 + x 2
Þ f ¢( x,0) = 2 x = 0 Þ x = 0 Þ (0,0) is the critical point
Þ f ¢¢( x,0) = 2 > 0 Þ (0,0) is the maximum value
Hence we fail to decide the nature of the critical point in this way.
Sufficient Condition
Let z = f ( x, y ) be defined and have continuous 1st and 2nd order partial derivatives
in a domain D . Suppose ( x0 , y0 ) is a point of D for which f x and f y are both zero.
Let A = f xx ( x0 , y0 ) , B = f xy ( x0 , y0 ) , C = f yy ( x0 , y0 ) ,
then we have the following cases
i) B 2 - AC < 0 and A + C < 0 Þ relative maximum at ( x0 , y0 ) .
ii) B 2 - AC < 0 and A + C > 0 Þ relative minimum at ( x0 , y0 )
iii) B 2 - AC > 0 Þ saddle point at ( x0 , y0 )
iv) B 2 - AC = 0 Þ nature of the critical point is undetermined
Proof
By the application of M.V. theorem for function of two variables we have
Df = hf x ( x0 + q h, y0 + q k ) + kf y ( x0 + q h, y0 + q k ) (0 < q < 1)
= h [ f x ( x0 + q h, y0 + q k ) - f x ( x0 , y0 ) ] + k éë f y ( x0 + q h, y0 + q k ) - f y ( x0 , y0 ) ùû
(it is because f x ( x0 , y0 ) = f y ( x0 , y0 ) = 0 , a turning point)
= h éëq hf xx ( x0 , y0 ) + q kf yx ( x0 , y0 ) + e1q h + e 2q k ùû
+ k éëq hf xy ( x0 , y0 ) + q kf yy ( x0 , y0 ) + e 3q h + e 4q k ùû
where e1 , e 2 , e 3 & e 4 ® 0 as h, k ® 0
Df = h 2 f xx ( x0 , y0 ) + 2hk f xy ( x0 , y0 ) + k 2 f yy ( x0 , y0 ) +e1h 2 + (e 2 + e 3 )hk + e 4k 2
Þ Df = h 2 A + 2hkB + k 2C + e1h2 + (e 2 + e 3 )hk + e 4 k 2
The sign of Df depends upon the quadratic d 2 f = h2 A + 2hk B + k 2C
i & ii) Let B 2 - AC < 0 , ( A ¹ 0)
Þ d2 f =
1 2
A
(
h A + 2hk AB + k 2 AC )
=
A
(
1 2 2
)
h A + 2hk AB + k 2 B 2 + (k 2 AC - k 2 B 2 )
=
1
A
( )
(hA + kB )2 + k 2 ( AC - B 2 )
Since (hA + kB )2 is positive and AC - B 2 (supposed) is +ive, therefore the sign of

d 2 f depends upon the sign of A .
Þ Df > 0 if A > 0 & Df < 0 if A < 0
Again, since B 2 - AC < 0 Þ B 2 < AC Þ AC > 0
Þ A and C are either both +ive or both –ive.
If A > 0 , C > 0 then A + C > 0 and if A < 0 , C < 0 then A + C < 0 .
Hence we have the following result
a) Df > 0 when A + C > 0 Þ ( x0 , y0 ) is a point of minimum value.
b) Df < 0 when A + C < 0 Þ ( x0 , y0 ) is a point of maximum value.
iii) Let B 2 - AC > 0 , then
d2 f =
1
A
( )
(hA + kB) 2 + k 2 ( AC - B 2 )
=
1
A
( )
(hA + kB )2 - k 2 ( B 2 - AC )
which may be +ive or –ive for certain value of h & k , therefore ( x0 , y0 ) is a
saddle point.
iv) Let B 2 - AC = 0 , A ¹ 0
1
Þ d 2 f = ( hA + kB )
2
A
which may vanish for certain values of h and k , implies that nature of the point
remain undetermined.
v Question
Test for maxima and minima
z = 1 - x2 - y2
Solution
¶z
= -2 x = 0 Þ x = 0
¶x
¶z
= -2 y = 0 Þ y = 0
¶y
Þ (0,0) is the only critical point.
¶2z ¶2z ¶2z
A = 2 = -2 , B = = 0 , C = 2 = -2
¶x ¶x ¶y ¶y
B - AC = 0 - 4 = -4 < 0 and A + C = -2 - 2 = -4 < 0
2
Þ the function has maximum value at (0,0) .

v Question
z = x 3 - 3xy 2
Solution
¶z
= 3x 2 - 3 y 2 = 0 Þ x = - y & x = y
¶y
¶z
= -6 xy = 0 Þ xy = 0
¶y
Þ (0,0) is the critical point.

¶2 z
A = 2 = 6 x = 0 at (0,0)
¶x
¶2 z
B= = -6 y = 0 at (0,0)
¶x¶y
¶2z
C = 2 = -6 x = 0 at (0,0)
¶y
B - 4 AC = 0 also A + C = 0
2
Therefore we need further consideration for the nature of point

Dz = z (0 + h,0 + k ) - z (0,0)
= z (h, k ) - z (0,0)
= h 3 - 2hk 2
For h = k
Dz = h 3 - 3h3 = -2h3
Þ Dz > 0 if h < 0 & Dz < 0 if h > 0
Hence (0,0) is a saddle point.
v Question
Examine the function
z = f ( x, y ) = x 2 y 2
Solution
f x = 0 Þ 2 xy 2 = 0
f y = 0 Þ 2 yx 2 = 0
implies that (0,0) is the critical point
A = f xx = 2 y 2 = 0 at (0,0)
B = f xy = -4 xy = 0 at (0,0)
C = f yy = 2 x 2 = 0 at (0,0)
Since B 2 - 4 AC = 0 and also A + C = 0
Therefore we need further consideration for the nature of point.
Df = f (h, k ) - f (0,0)
= h 2k 2
Df > 0 for all h & k
Hence (0,0) is the point where function has minimum value.
……………………………………..
v Lagrange’s Multiplier
(Maxima & Minima for Function with Side Condition)
A problem of considerable importance for application is that of maximizing and
minimizing of function (optimization) of several variables where the variables are
related by one or more equations, which are turned as side condition. e.g. the
problem of finding the radius of largest sphere inscribable in the ellipsoid
x 2 + 2 y 2 + 3z 2 = 6 is equivalent to minimizing the function w = x 2 + y 2 + z 2 with the
side condition x 2 + 2 y 2 + z 2 = 6 .
To handle such problem, we can, if possible, eliminate some of the variables by
using the side conditions and reduce the problem to an ordinary maximum and
minimum problem such as that consider previously.
This procedure is not always feasible and following procedure often is more
convenient which treat the variable in more symmetrical manner, so that various
simplifications may be possible.
Consider the problem of finding the extreme values of the function f ( x1, x2 ,...., xn )
when the variable are restricted by a certain number of side conditions say
g1 ( x1, x2 ,...., xn ) = 0
g 2 ( x1, x2 ,...., xn ) = 0
……………………
…………………....
……………………
g m ( x1, x2 ,...., xn ) = 0
We then form the linear combination
j ( x1 ,..., xn ) = f ( x1 ,..., xn ) + l1g1 ( x1 ,..., xn ) + l2 g 2 ( x2 ,..., xn ) + ....... + lm g m ( x1,..., xn )
where l1 , l2 ,...., lm are m constants.
We then differentiate j w.r.t. each coordinate and consider the following system
of n + m equations.
Drj ( x1 , x2 ,...., xn ) = 0 , r = 1,2,...., n
g k ( x1, x2 ,...., xn ) = 0 , k = 1, 2,...., m
Lagrange discovered that if the point ( x1, x2 ,...., xn ) is a solution of the extreme
problem then it will also satisfy the system of n + m equation.
In practise, we attempt to solve this system for n + m unknowns, which are
l1 , l2 ,...., lm & x1, x2 ,...., xn
The point so obtain must then be tested to determine whether they yield a
maximum, a minimum or neither.
The numbers l1 , l2 ,...., lm , which are introduced only to help to solve the system
for x1, x2 ,...., xn are known as Lagrange’s multiplier. One multiplier is introduced for
each side condition.
v Question
Find the critical points of w = xyz , subject to condition x 2 + y 2 + z 2 = 1 .
Solution
We form the function
j = xyz + l ( x 2 + y 2 + z 2 - 1)
then
¶j
= yz + 2l x = 0
¶x
¶j
= xz + 2l y = 0
¶y
¶j
= xy + 2l z = 0
¶z
& x2 + y 2 + z 2 -1 = 0
Multiplying the first three equations by x, y & z respectively, adding and using
the fourth equation, we find
3x y z
l=-
2
using this relation we find that ( 0,0, ±1) , ( 0, ±1,0 ) , ( ±1,0,0 ) and
æ 1 1 1 ö
ç± ,± ,± ÷ are the critical points.
è 3 3 3ø
v Question
Find the critical points of w = xyz , where x 2 + y 2 = 1 & x - z = 0 . Also test for
maxima and minima.
Solution
( )
Consider F = xyz + l1 x 2 + y 2 + 1 + l2 ( x - z )
For the critical points, we have
Fx = yz + 2l1 x + l2 = 0 ……….. (i)
Fy = xz + 2l1 y = 0 ………...….. (ii)
Fz = xy - l2 = 0 …………...….. (iii)
and x 2 + y 2 = 1 ………...….…. (iv)
x - z = 0 …………………. (v)
xz
From (iii ) , l2 = xy & from (ii ) l1 = -
2y
Use these values in equation (i ) to have
x2 z
yz - + xy = 0
y
Þ y 2 z - x 2 z + xy 2 = 0
Q x = z from (v )
\ y 2 x - x 3 + xy 2 = 0 Þ 2 xy 2 - x 3 = 0
But y 2 = 1 - x 2 , from (iv )
( )
\ 2 x 1 - x 2 - x3 = 0 Þ 2 x - 3 x3 = 0 Þ x=0 , ±
2
3
æ 2 1 2ö æ 2 1 2ö
This implies the critical points are ç ± , ,± ÷ ç
, ± , - , ± ÷,
è 3 3 3 ø è 3 3 3 ø
( 0,1,0 ) , ( 0, -1,0 )
A = Fxx = 2l1
B = Fxy = z
C = Fyy = 2l1
B 2 - AC = z 2 - 4l12
x2 z 2
= z -4 2 =
2 (
z2 y 2 - x2 )
4y y2
æ 2 1 2ö
(i) At ç ± , ,± ÷ , we have
è 3 3 3 ø
2æ1 2ö
ç - ÷
3è3 3ø
B - AC =
2
<0
1
3
æ 2 ö
& A = Fxx = 2l1 = - = - ç 3 ÷ < 0
xz
ç 1 ÷
y ç ÷
è 3ø
æ 2 1 2ö
Þ function has maximum value at ç ± , , ± ÷
è 3 3 3 ø
Similarly, we can show that F is also maximum at (0, -1,0) and is minimum at
remaining points. (Check yourself)
v Question
Find the point of the curve
x 2 - xy + y 2 - z 2 = 1 , x 2 + y 2 = 1
which is nearest to the origin.
Solution
Let a point on a given curve be ( x, y, z )
Implies that we are to minimize the function
f = d 2 = x2 + y 2 + z 2
subject to the conditions
x 2 - xy + y 2 - z 2 = 1
x2 + y2 = 1
Consider
( ) ( )
F = x 2 + y 2 + z 2 + l1 x 2 - xy + y 2 - z 2 - 1 + l2 x 2 + y 2 - 1
For the critical points
Fx = 2 x (1 + l1 + l2 ) - l1 y = 0 ……….. (i )
Fy = 2 y (1 + l1 + l2 ) - l1x = 0 ……….. (ii )
Fz = 2 z (1 - l1 ) = 0 ………… (iii )
x 2 - xy + y 2 - z 2 = 1 ………… (iv )
x 2 + y 2 = 1 ………….. (v )
From equation (iii ) , we have
z = 0 and l1 = 1
Put z = 0 in equation (iv ) , gives
x 2 - xy + y 2 - 1 = 0
Þ xy = x 2 + y 2 - 1
Þ xy = 0 by (v )
Þ x = 0 or y = 0 or both are zero.
z = 0 , x = 0 in (v ) gives, y 2 = 1 Þ y = ±1
Þ ( 0, ±1,0 ) are the critical points.
z = 0 , y = 0 Þ x = ±1 Þ ( ±1,0,0 ) are the critical points.
We can not take x = 0 , y = 0 at the same time, because it gives ( 0,0,0 ) which is
origin itself as a critical point.
Q d 2 = 1 at all these four points.
\ these are the required point at which function is nearest to origin.
v Question
Find the point on the curve
x2 + y 2 + z 2 = 1
which is farthest from the point (1,2,3)
Solution
We are the maximize the function
f = ( x - 1) + ( y - 2 ) + ( z - 3 )
2 2 2
subject to the condition

x2 + y 2 + z 2 = 1
Let
2 2 2
( )
F = ( x - 1) + ( y - 2 ) + ( z - 3) + l x 2 + y 2 + z 2 - 1
For the critical points, we have
x - 1 + l x = 0 …………. (i )
y - 2 + l y = 0 …..…..… (ii )
z - 3 + l z = 0 ………… (iii )
& x 2 + y 2 + z 2 = 1 ……….. (iv )
1 2 1
Þ x= , y= , z=
1+ l 1+ l 3+l
Putting in (iv )
2
æ 1 ö
÷ (1 + 4 + 9 ) = 1 Þ (1 + l ) = 14 Þ l = -1 ± 14
2
ç
è1+ l ø
1 2 3
Þ x= , y= , z=
± 14 ± 14 ± 14
Þ critical points are
æ 1 2 3 ö
ç± ,± ,± ÷
è 14 14 14 ø
Its clear that the required point which is farthest from the point (1,2,3) is
æ 1 2 3 ö
ç- ,- ,- ÷
è 14 14 14 ø
………………………………………
v Directional Derivative
i) Let f : V ® ¡ , where V Ì ¡ n , is nhood of a Î ¡ n . Then the directional
derivative Db f at a in the direction of b Î ¡ n , is defined by the limit, if it exists,
Db f (a ) = lim
(
f a + hb - f ( a ) )
hh ®0
ii) The directional derivative of f ( x1, x2 ,..., xi ,..., xn ) at a = ( a1 , a2 ,..., ai ,..., an ) in

the direction of the unit vector ( 0,0,...,1,0,0,...,0 ) is called partial derivative of f at
a w.r.t. the ith component xi and is denoted by
¶f ( a )
Di f ( a ) or or f xi (a)
¶x
f ( a1, a2 ,..., ai + h,..., an ) - f ( a1, a2 ,..., ai ,..., an )
where Di f ( a ) = lim
h®0 h
v Example
Let f ( x, y ) = x 2 + y 2 + x + y , then f has a directional derivative in every direction
and at every point in ¡ 2 .
Since, if b = ( a, b ) Î ¡ 2 , we have
( x + ha )
+ ( y + hb ) + ( x + ha ) + ( y + hb ) - x 2 - y 2 - x - y
2 2
Db f ( x, y ) = lim
h®0 h
h ®0
(
= lim 2ax + 2by + ha + hb + a + b
2 2
)
= 2 ( ax + by ) + a + b
v Exercise
ï
(
ì xy x 2 - y 2 ) ; x4 + y 4 ¹ 0
Let f ( x, y ) = í x 4 + y 4
ï ; ( x, y ) ¹ (0,0)
î 0
Note that if b = ( a, b ) Î ¡ 2 ,
( 0 + ah )( 0 + bh ) éë( 0 + ah ) - ( 0 + bh ) ù
2 2
Db f (0,0) = lim û
h é( 0 + ah ) + ( 0 + bh ) ù
h ®0 4 4
ë û
= lim
(
ab a 2 - b 2 )
h ®0 h(a 4
+ b4 )
This limit obviously exists only if b = (1,0 ) or ( 0,1) . Hence the directional
derivatives of f at ( 0,0 ) that exists are the partial derivatives f x and f y given by
fx = 0 , fy = 0.
v Example
Let
ì xy 2
ï ; ( x, y ) = ( 0,0 )
f ( x, y ) = í x 4 + y 4
ï 0 ; ( x, y ) ¹ ( 0,0 )
î
It is discontinuous at ( 0,0 ) . To see it, note that
1
lim f ( x, y ) is zero along y = 0 and is along y 2 = x .
( x , y ) ®(0,0) 2
However, if b = ( a, b ) , then
( 0 + ah )( 0 + bh )
2
f b ( 0,0 ) = lim
h ( 0 + ah ) + ( 0 + bh ) ù
h ®0 é 2 4
ë û
ah × b 2 h 2 ab 2
= lim = lim
h ®0 h é a 2 h 2 + b 4 h 4 ù h ®0 a 2 + h 2b 4
ë û
ïìb a , a ¹ 0
2
=í
ïî 0 , a = 0
Hence the directional derivative of f at ( 0,0 ) exists in every direction.
v Question
Let z = f ( x, y ) , x = u 2 - v 2 , y = 2uv . Then show that
ìïæ ¶z ö2 æ ¶z ö2 üï
2
æ ¶z ö æ ¶z ö
2
1
ç ÷ +ç ÷ = íç ÷ + ç ÷ ý
è ¶x ø è ¶y ø 4 u 2 + v 2 ( ) îïè ¶u ø è ¶v ø þï
Solution
We have
¶x ¶x ¶y ¶y
= 2u , = -2v , = 2v , = 2v
¶u ¶v ¶u ¶v
Also
¶u ¶v ¶u ¶v
1 = 2u - 2v - 2v
, 0 = 2u
¶x ¶x ¶y ¶y
¶u ¶v ¶u ¶v
and 0 = 2v + 2u , 1 = 2v + 2u
¶x ¶x ¶y ¶y
¶u ¶v ¶u ¶v
Solving these four equations for , , & , we get
¶x ¶x ¶y ¶y
¶u u ¶v -v
= , =
(
¶x 2 u 2 + v 2 )
¶x 2 u 2 + v 2 ( )
¶u v ¶v u
= , =
(
¶y 2 u + v 2
2
) (
¶y 2 u + v 2
2
)
And
¶z ¶z ¶u ¶z ¶v
= × + ×
¶x ¶u ¶x ¶v ¶x
1 é ¶z ¶z ù
= u× -v× ú
( 2 ê
2 u + v ë ¶u
2
)
¶v û
¶z ¶z ¶u ¶z ¶v
& = × + ×
¶y ¶u ¶y ¶v ¶y
1 é ¶z ¶z ù
= v × + u ×
(
2 u 2 + v 2 êë ¶u ) ¶v úû
Hence
ïìæ ¶z ö æ ¶z ö ïü
2
æ ¶z ö æ ¶z ö
2 2 2
1
ç ÷ +ç ÷ = íç ÷ + ç ÷ ý
è ¶x ø è ¶y ø 4 u 2 + v 2 ( ) ïîè ¶u ø è ¶v ø ïþ
…………………………………………………..
v Question
Let f : ¡ 2 ® ¡ be given by
ì xy
ï ; ( x, y ) ¹ ( 0,0 )
f ( x, y ) = í x 2 + y 2
ï 0 ; ( x, y ) = ( 0,0 )
î
Show that f x , f y exist at ( 0,0 ) but f is discontinuous at ( 0,0 ) .
Solution
f b ( 0,0 ) = lim
( ah )( bh ) where b = ( a, b )
h ( ah ) + ( bh ) ù
h®0 é 2 2
ë û
ab
= lim
(
h ®0 h a + b 2
2
)
Which exists only when b = (1,0 ) or ( 0,1) .
Þ f x & f y exist at ( 0,0 )
Now
xy
lim f ( x, y ) = lim
( x , y ) ®(0,0) ( x , y ) ®(0,0) x 2 + y 2
Let y = mx , then
xy mx 2
lim = lim 2
( x , y ) ®(0,0) x 2 + y 2 x ®0 x + m 2 x 2
m
= lim
x ®0 1 + m 2
Which is different for different m .

Þ f ( x, y ) is discontinuous at ( 0,0 ) .
v Question
Let f : ¡ 2 ® ¡ be given by
ì x2 y
ï ; ( x, y ) ¹ ( 0,0 )
f ( x, y ) = í x 4 + y 2
ï 0 ; ( x, y ) = ( 0,0 )
î
Show that f x , f y exist at ( 0,0 ) but f is discontinuous at ( 0,0 ) .
Solution
f b ( 0,0 ) = lim
( a h ) ( bh )
2 2
, b = ( a, b )
h®0 h éë a 4 h 4 + b 2h 2 ùû
a 2b
= lim 4 2
h ®0 a h + b 2
ì a2
ï , b¹0
=í b
ï 0 , b=0
î
1
Now lim f ( x, y ) is zero along x = 0 and is along y = x 2
( x , y ) ®(0,0) 2
Þ it is discontinuous at (0,0) .
…………………………………..
v Question
Find the greatest volume of the box contained in the ellipsoid 3x 2 + 2 y 2 + z 2 = 18 ,
when each of its edges is parallel to one of the coordinate axes.
Solution
V = volume of the box = (12 x )( 2 y )( 2 z ) = 8xyz
We need to find maximum of V subject to 3x 2 + 2 y 2 + z 2 - 18 = 0
( )
Consider j ( x, y, z ) = 8 xyz + l 3 x 2 + 2 y 2 + z 2 - 18 = 0
Then
j x = 8 yz + 6l x = 0
j y = 8 xz + 4l y = 0
j z = 8 xy + 2l z = 0
Þ 4 xyz + 3l x 2 = 0
2 xyz + l y 2 = 0
4 xyz + l z 2 = 0
(
Þ l 3x 2 - 2 y 2 = 0 )
l (3x 2
)
- z2 = 0
2 y2 z2
Þ x = 2
=
3 3
Substituting these values in
3x 2 + 2 y 2 + z 2 - 18 = 0
We get
3 x 2 + 3 x 2 + 3 x 2 = 18 Þ 9 x 2 = 18
Þ x = 2 , y = 3 and z = 6
Which gives
f ( x, y , z ) = 8 xyz = 48
……………………………………..
v Definition
If f : ¡ n ® ¡ , a Î ¡ n then
n
¶f ( a ) ¶f ( a ) ¶f ( a ) ¶f ( a )
Ñf ( a ) = å = + + .... +
k =1 ¶xk ¶x1 ¶x2 ¶xn
v Definition
Let f : G ® ¡ , G is an open set in ¡ n .
i) f is said to have a local maximum at a Î G , if there is a nhood Ve ( a ) such that
f ( x ) £ f ( a ) " x ÎVe .
ii) f is said to have a local minimum at a Î G , if there is a nhood Ve ( a ) such that
f ( x ) ³ f ( a ) " x ÎVe .
v Theorem
Let f : G ® ¡ , G is an open set in ¡ n . If f has a local extremum at a Î G , then
Ñf ( a ) = 0 .
Proof
¶(a)
It is clear that Ñf ( a ) = 0 iff = 0 , i = 1,2,3,...., n
¶xi
Write f ( xi + t ) = f ( x1 , x2 ,...., xi + t ,...., xn ) = f ( x )
If f has a local maximum at a , then
f ( ai + t ) - f ( ai )
£ 0 if t > 0
t
f ( ai + t ) - f ( ai )
Þ lim £ 0 if t > 0
t ®0 t
¶f ( a )
So that £ 0
¶xi
Similarly,
f ( ai + t ) - f ( ai )
lim ³ 0 if t < 0
t ®0 t
¶f ( a )
So that ³0
¶xi
¶f ( a )
Hence = 0 , i = 1,2,3,...., n
¶xi
Þ Ñf ( a ) = 0
Note
There are situations when Ñf ( a ) = 0 but f has no local maximum or minimum at
a . If so and if the sign of f ( x ) - f ( a ) depends upon the direction of x and a , f is
said to have a saddle point at a .
={ END }=
Maxima and Minima for Functions with side Remarks
conditions. Lagrange’s Multiplier.
Question
Find the critical points of w = xyz subject to the condition
x2 + y 2 + z 2 = 1.
Solution
We form the function
ϕ = f + λ g = xyz + λ ( x 2 + y 2 + z 2 − 1)
and obtain four equations
∂ϕ
= yz + 2λ x = 0
∂x
∂ϕ
= xz + 2λ y = 0
∂y
∂ϕ
= xy + 2λ z = 0
∂z
& x2 + y 2 + z 2 − 1 = 0
Multiplying the first three equations by x, y , z respectively, adding
3 xyz
and using in fourth equation we find λ = − .
2
Using this relation we have (0,0, ±1) , (0, ±1,0) , (±1,0,0) , and
 1 1 1 
± ,± ,±  as the critical points.
 3 3 3
Question
Find the critical points of the function z = x 2 + 24 xy + 8 y 2
where x 2 + y 2 = 25 . Test for maxima & minima.
Solution
F ( x, y, λ ) = x 2 + 24 xy + 8 y 2 + λ ( x 2 + y 2 − 25)
Fx = 2 x + 24 y + 2λ x = 0 ……….…… (i)
Fy = 24 x + 16 y + 2λ y = 0 …………… (ii)
& x 2 + y 2 − 25 = 0 …………….….. (iii)
(i) ⇒ (1 + λ ) x + 12 y = 0 ………..…. (iv)
(ii) ⇒ 12 x + (8 + λ ) y = 0 ………...…. (v)
Multiplying equation (iv) by 12, (v) by (1 + λ ) and adding
12(1 + λ ) x + 144 y = 0
12(1 + λ ) x + (1 + λ )(8 + λ ) y = 0
− −
144 y − (1 + λ )(8 + λ ) y = 0
⇒ y = 0 or λ 2 + 9λ − 136 = 0
⇒ y = 0 , λ = 8, − 17
From (ii), y = 0 ⇒ x = 0
 (0,0) does not satisfy (iii) ∴ It is not a critical point.
4y
λ =8 ⇒ x = − form (iv)
3
Put this value of x in (iii)
2
16 y 2 Remarks
⇒ + y 2 = 25 ⇒ y = ± 3
9
⇒ (−4,3) & (4, −3) are the critical points.
3y
Similarly when λ = −17 , we have x = from (iv)
4
And putting the value of x in (iii) we get y = ± 4
⇒ (± 3, ± 4) are the other two critical points.
A = Fxx = 2 + 2λ
B = Fxy = 24
C = Fyy = 16 + 2λ
When λ = 8
A = 2 + 16 = 18 , B = 24 , C = 16 + 16 = 32
and so B 2 − AC = 576 − 576 = 0
F ( x, y, λ ) = x 2 + 24 xy + 8 y 2 + 8( x 2 + y 2 − 25) when λ = 8
⇒ F ( x, y ) = 9 x 2 + 24 xy + 16 y 2 − 200
At (−4,3)
∆F = F (−4 + h,3 + h) − F (−4,3)
= 9(−4 + h) 2 + 24(−4 + h)(3 + h) + 16(3 + h) 2 − 200
−9(−4) 2 + 24(−4)(−3) − 16(3) 2 + 200
= 9(16 − 8h + h2 ) + 24( h2 − h − 12) + 16(9 + 6h + h 2 )
−144 + 288 − 144
= 144 − 72h + 9h + 24h − 24h − 288 + 144 + 96h + 26h 2
2 2
−144 + 288 − 144

= 49h ≥ 0
2
⇒ (−4,3) is the point of minimum value.

Similarly (4, −3) gives a point of minimum value.
And when λ = −17 , (± 3, ± 4) are the point of maximum value.
Question
Find the critical points of w = x + z , where x 2 + y 2 + z 2 = 1 .
Test for a maxima and minima.
Solution
Consider the function
F ( x, y, z ) = x + z + λ ( x 2 + y 2 + z 2 − 1)
Fx = 1 + 2λ x , Fy = 2λ y , Fz = 1 + 2λ z
For critical points, we have
1 + 2λ x = 0 …………...… (i)
2λ y = 0 ……………..…. (ii)
1 + 2λ z = 0 …………….. (iii)
and x 2 + y 2 + z 2 = 1 ………… (iv)
1
Solving these equations we have λ = ±
2
1  −1 −1 
λ= gives  ,0,  as the critical point.
2  2 2
1  1 1 
λ=− gives  ,0,  as the critical point.
2  2 2
3
A = Fxx = 2λ , B = Fxy = 0 , C = Fyy = 2λ Remarks
1
i) λ = ⇒ A = 2 , B = 0, C = 2
2
so B 2 − AC = 0 − 2 < 0 and A > 0
 −1 −1 
⇒  ,0,  is a point of relative minimum value.
 2 2 
1
ii) λ = − ⇒ A = − 2 , B = 0, C = − 2
2
so B 2 − AC = 0 − 2 < 0 and A < 0
 1 1 
⇒  ,0,  is a point of relative maximum value.
 2 2
Question
Find the critical points of w = xyz where x 2 + y 2 = 1 &
x − z = 0 . Test for the maxima and minima.
Solution
Consider F = xyz + λ1 ( x 2 + y 2 − 1) + λ2 ( x − z )
For critical points
Fx = yz + 2λ1 x + λ2 = 0 …………. (i)
Fy = xz + 2λ1 y = 0 ………….….. (ii)
Fz = xy − λ2 = 0 ……………….. (iii)
& x 2 + y 2 = 1 …………..……. (iv)
x − z = 0 …………….……. (v)
xz
From (iii) λ2 = xy and from (ii) λ1 = −
2y
Putting in (i), we get
x2 z
yz − + xy = 0
y
⇒ y 2 z − x 2 z + xy 2 = 0
 x = z form (iv) ∴ y 2 x − x3 + xy 2 = 0
⇒ 2 xy 2 − x 3 = 0
But from (iv), y 2 = 1 − x 2 ⇒ 2 x(1 − x 2 ) − x3 = 0
2
⇒ 3x3 − 2 x = 0 ⇒ x(3x 2 − 2) = 0 ⇒ x = 0, ±
3
 2 1 2   2 −1 2
⇒ The critical points are  ± , ,± , ± , ,± ,
 3 3 3   3 3 3 
( 0,1,0 ) and (0, −1,0) .
A = Fxx = 2λ1 , B = Fxy = z , C = Fyy = 2λ1
B 2 − AC = z 2 − 4λ12
x2z 2 z 2 x2 z 2 ( y 2 − x 2 )
From (ii) λ12 = ⇒ B 2 − AC = z 2 −
=
4 y2 y2 y2
 2 1 2 3(3 3)
2 1−2
i) At  ± , ,±  , we have B − AC =
2
<0
1
 3 3 3  3
4
xz − 2 3 Remarks
And A = Fxx = 2λ1 = − = <0
y 1
3
 2 1 2
⇒ Function is maximum at  ± , ,± .
 3 3 3 
Similarly we can show that w is maximum at (0, −1,0) and
 2 −1 2
minimum at  ± , ,±  & (0,1,0).
 3 3 3 
Question
Find the point to the curves x 2 − xy + y 2 − z 2 = 1, x 2 + y 2 = 1
nearest to the origin (0,0,0,) .
Solution
Let ( x, y , z ) be a point on the curve. Then its distance from the
origin is given by x2 + y 2 + z 2
We are to minimize f = d 2 = x 2 + y 2 + z 2
subject to the conditions x 2 − xy + y 2 − z 2 = 1 , x 2 + y 2 = 1
Consider
F = x 2 + y 2 + z 2 + λ1 ( x 2 − xy + y 2 − z 2 − 1) + λ2 ( x 2 + y 2 − 1)
Fx = 2 x + (2 x − y )λ1 + 2λ2 x
Fy = 2 y + (2 y − x)λ1 + 2λ2 y
Fz = 2 z + λ1 (−2 z )
2 x (1 + λ1 + λ2 ) − λ1 y = 0 ………… (i)
2 y (1 + λ1 + λ2 ) − λ1 x = 0 ………….(ii)
2 z (1 − λ1 ) = 0 …………………..... (iii)
x 2 − xy + y 2 − z 2 − 1 = 0 …………. (iv)
x 2 + y 2 − 1 = 0 ………………..….. (v)
From (iii), we have z = 0 & λ1 = 1 .
z = 0 in (iv) gives x 2 − xy + y 2 − 1 = 0 ⇒ xy = x 2 + y 2 − 1
But x 2 + y 2 − 1 = 0 ⇒ xy = 0
⇒ x = 0 or y = 0 or both are zero.
We can not take x = 0 , y = 0 at a same time because it gives
(0,0,0) which is origin itself.
z = 0, x = 0 in (v) ⇒ y 2 = 1 ⇒ y = ±1
⇒ (0, ±1,0) are the critical points
& z = 0, y = 0 in (v) ⇒ x 2 = 1 ⇒ x = ±1
⇒ ( ±1,0,0 ) are the other critical points.
 f = d 2 = 1 at these four points
∴ These are the required points at which function is nearest to origin.
5
Question Remarks
Find the shortest distance from the origin to the curve
x 2 + 8 xy + 7 y 2 = 225
Solution
We are to find the minimum value of f = d 2 = x 2 + y 2
subject to the condition x 2 + 8 xy + 7 y 2 = 225 .
Consider F = x 2 + y 2 + λ ( x 2 + 8xy + 7 y 2 − 225)
Fx = 2 x + λ (2 x + 8 y )
Fy = 2 y + λ (8 x + 14 y )
For critical points
x + λ ( x + 4 y ) = 0 ……….……. (i)
y + λ (4 x + 7 y ) = 0 ………..…. (ii)
x 2 + 8 xy + 7 y 2 − 225 = 0 …….. (iii)
x 4λ
(i) ⇒ (1 + λ ) x + 4λ y = 0 ⇒ =−
y 1+ λ
x 1 + 7λ
(ii) ⇒ 4λ x + (1 + 7 λ ) y = 0 ⇒ =−
y 4λ
x 4λ 1 + 7λ
⇒ =− =− ⇒ 16λ 2 = (1 + λ )(1 + 7λ )
y 1+ λ 4λ
⇒ 16λ 2 = 1 + λ + 7λ + 7λ 2 ⇒ 9λ 2 − 8λ − 1 = 0
1
⇒ (λ − 1)(9λ + 1) = 0 ⇒ λ = 1, −
9
x
λ =1 ⇒ = −2 ⇒ x = −2 y
y
Putting this value of x in equation (iii) we have
(−2 y )2 + 8(−2 y ) y + 7 y 2 = 225
⇒ 4 y 2 − 16 y 2 + 7 y 2 = 225 ⇒ − 5 y 2 = 225
which gives imaginary values of y.
1 x 4 ( − 19 ) 4 9 1
λ =− ⇒ =− = = ⇒ y = 2x
9 y 1 − 19 8
9 2
Putting in (iii), we have
x 2 + 8 x(2 x) + 7(2 x) 2 = 225
⇒ x 2 + 16 x 2 + 28x 2 = 225
⇒ 45 x 2 = 225 ⇒ x 2 = 5 ⇒ x = ± 5
x= 5 ⇒ y=2 5
& x = − 5 ⇒ y = −2 5
∴ The critical points are ( ) (
5,2 5 & − 5, −2 5 . )
d (2± 5,± 2 5 )
= 25
⇒ Shortest distance = d = 5
6
Question
Find a point ( x, y , z ) on the sphere x 2 + y 2 + z 2 = 1 which is Remarks
farthest from the point (1, 2,3) .
Solution
We are to maximize
f ( x, y, z ) = ( x − 1)2 + ( y − 2)2 + ( z − 3)2
subject to the condition x 2 + y 2 + z 2 = 1
Let F = ( x − 1)2 + ( y − 2)2 + ( z − 3)2 + λ ( x 2 + y 2 + z 2 − 1)
For critical points
Fx = 2( x − 1) + 2λ x = 0
Fy = 2( y − 2) + 2λ y = 0
Fz = 2( z − 3) + 2λ z = 0 and x 2 + y 2 + z 2 = 1
⇒ x − 1 + λ x = 0 …………….. (i)
y − 2 + λ y = 0 …………… (ii)
z − 3 + λ z = 0 ……….…... (iii)
x 2 + y 2 + z 2 = 1 ………….. (iv)
1 2 3
⇒ x= , y= , z=
1+ λ 1+ λ 1+ λ
Putting in (iv)
2 2 2
 1   2   3 
  +  +  =1
 1 + λ   1 + λ   1 + λ 
⇒ 14 = (1 + λ ) 2
⇒ λ + 1 = ± 14 ⇒ λ = −1 ± 14
1 2 3
⇒ x= , y= , z=
± 14 ± 14 ± 14
 −1 −2 −3 
Clearly  , ,  is the point which is farthest from (1, 2,3) .
 14 14 14 
Question
Find the extreme values of z = 6 − 4 x − 3 y , provided x & y
satisfy x 2 + y 2 = 1 .
Solution
Define F = 6 − 4 x − 3 y + λ ( x 2 + y 2 − 1)
Fx = − 4 + 2λ x = 0 …………. (i)
Fy = −3 + 2λ y = 0 …………. (ii)
and x 2 + y 2 = 1 …..………. (iii)
2 3
From (i) and (ii) we have x = , y =
λ 2λ
5
Putting these values in (iii) we get λ = ±
2
5 2 4 3 3
λ= ⇒ x= = & y= =
2 5
2 5 2 ⋅ 52 5
7
5 2 4 3 3 Remarks
λ =− ⇒ x= =− & y= =−
2 − 52 5 2 ⋅ (− 52 ) 5
 4 3  4 3
⇒  ,  &  − , −  are the critical points.
 5 5  5 5
A = Fxx = 2λ , B = Fxy = 0 , C = Fyy = 2λ
2
 5
⇒ B − AC = 0 − 4λ = − 4  ±  = −25 < 0
2 2
 2
⇒ F is maximum or minimum at the critical points.
 4 3
Now at  ,  , we have A = 5 > 0
5 5
 4 3
And at  − , −  , we have A = −5 < 0
 5 5
 4 3  4 3
⇒ The function is min. at  ,  and max. at  − , −  .
5 5  5 5
Question
Find the critical point of f ( x, y ) = x 2 + 2 y 2 + 2 xy + 2 x + 3 y .
Where x 2 − y = 1 . Test for maxima and minima.
Solution
Define F = x 2 + 2 y 2 + 2 xy + 2 x + 3 y + λ ( x 2 − y − 1)
Fx = 2 x + 2 y + 2 + 2λ x = 0 …..….. (i)
Fy = 4 y + 2 x + 3 − λ = 0 ………… (ii)
and x 2 − y − 1 = 0 ……….……… (iii)
−x − y −1
From (i) λ =
x
From (ii) λ = 2 x + 4 y + 3
−x − y −1
⇒ = 2x + 4 y + 3
x
⇒ − x − y − 1 = 2 x 2 + 4 xy + 3x
⇒ 2 x 2 + 4 x + 4 xy + y + 1 = 0
But from (iii) x 2 = 1 + y
⇒ 2(1 + y ) + 4 x + 4 xy + y + 1 = 0
⇒ 4 x + 4 xy + 3 y + 3 = 0
⇒ 4 x(1 + y ) + 3( y + 1) = 0
⇒ ( y + 1)(4 x + 3) = 0
3
⇒ Either y = −1 or x = −
4
If y = −1 , we get x = 0 from (iii)
2
⇒ (0, −1) is a critical point and λ = −1 in this case.

3 9 7
If x = − , we get − 1 = y i.e. y = −
4 16 16
 3 7 1
⇒  − , −  is the other critical point and λ = − in this. case.
 4 16  4
8
Now A = Fxx = 2 + 2λ , B = Fxy = 2 , C = Fyy = 4 Remarks
⇒ B 2 − AC = 4 − 4(2 + 2λ ) = − 4 − 8λ
λ = −1 ⇒ B 2 − AC = 4 > 0 ⇒ f is neither maximum nor
minimum at (0,1) .
1  1
λ =− ⇒ B 2 − AC = − 4 − 8  −  = − 4 + 2 = −2 < 0
4  4
 1 1
and A = 2 + 2λ = 2 + 2  −  = 2 − > 0
 4 2
 3 7
⇒  − , −  is the point of minimum value.
 4 16 
Question
Find the critical points of z = x 2 + y 2 when x 3 + y 3 = 6 xy ,
Also test for maxima and minima.
Solution
Define F = x 2 + y 2 + λ ( x3 + y 3 − 6 xy )
For critical points we have
Fx = 2 x + 3λ x 2 − 6λ y = 0 …..……. (i)
Fy = 2 y + 3λ y 2 − 6λ x = 0 ……….. (ii)
and x 3 + y 3 − 6 xy = 0 …………… (iii)
−2 x
from (i) λ = 2
3x − 6 y
−2 y
from (ii) λ = 2
3 y − 6x
−2 x −2 y
⇒ = 2
3x − 6 y 3 y − 6 x
2
⇒ x (3 y 2 − 6 x ) = y (3x 2 − 6 y )
⇒ x ( y 2 − 2 x) = y ( x2 − 2 y)
⇒ xy 2 − 2 x 2 = x 2 y − 2 y 2
⇒ x 2 y − xy 2 + 2 x 2 − 2 y 2 = 0
⇒ xy( x − y) + 2( x − y )( x + y) = 0
⇒ ( x − y )(2 x + 2 y + xy ) = 0
⇒ Either x − y = 0 or 2 x + 2 y + xy = 0
If x − y = 0 then (iii) becomes x 3 + x 3 − 6 x 2 = 0
⇒ 2 x 3 − 6 x 2 = 0 ⇒ x 2 ( x − 3) = 0
⇒ x = 0, 3
⇒ x = 0, y = 0 & x = 3, y = 3
⇒ (0,0) & (3,3) are the critical points.
− 2x − 2x
At (0,0), λ = 2 = 2  x− y=0 ⇒ x= y
3x − 6 y 3x − 6 x
−2 −2 1
= = =
3 x − 6 3(0) − 6 3
2
And at (3,3) , λ = −
3
9
Remarks
A = Fxx = 2 + 6λ x
B = Fxy = −6λ
C = Fyy = 2 + 6λ y
At (0,0), we have A = 2 , B = −2 , C = 2
And ∴ B 2 − AC = 0
Consider ∆z = z (h, h) − z (0,0) = h 2 + h 2 = 2h 2 ≥ 0
⇒ (0,0) is the point of minimum value.
 2
At (3,3) , we have A = 2 + 6  −  (3) = −10
 3
 2
B = −6  −  = 4
 3
 2
C = 2 + 6  −  (3) = −10
 3
and ∴ B − AC = 16 − 100 < 0 and A = −10 < 0
2
⇒ (3,3) is a point of maximum value.
Question
Find the points in the plane 2 x + 3 y − z = 5 nearest to the origin.
Solution
We are to minimize f = d 2 = x 2 + y 2 + z 2
subject to 2 x + 3 y − z − 5 = 0 .
Define F = x 2 + y 2 + z 2 + λ ( 2 x + 3 y − z − 5)
Fx = 2 x + 2λ = 0 …………….... (i)
Fy = 2 y + 3λ = 0 ……………... (ii)
Fz = 2 z − λ = 0 ………………..(iii)
and 2 x + 3 y − z − 5 = 0 …...….. (iv)
− 3λ λ
x = −λ , y = , z= from (i), (ii) & (iii) resp.
2 2
9λ λ
(iv) becomes −2λ − − −5= 0
2 2
⇒ 4λ + 9λ + λ = −10
10 5
⇒ λ =− =−
14 7
5 15 5
⇒ x= , y= , z=−
7 14 14
 5 15 − 5 
⇒  , ,  is the critical point.
 7 14 14 
A = Fxx = 2 , B = Fxy = 0 , C = Fyy = 2
B 2 − AC = 0 − 4 < 0 A=2>0
and
 5 15 − 5 
⇒ F is relative minimum at  , ,  so this is the required
 7 14 14 
point.
http://www.mathcity.tk
1
Maxima and Minima for Functions of Two Variable Remarks
Question
(i) z = 1 − x 2 − y 2 (ii) z = x2 + y2
(iii) z = xy (iv) z = x 3 − 3xy 2
(v) z = x2 y 2 (vi) z = 4 − y 2
Solution
(i) z = 1 − x 2 − y 2
∂z ∂z
= −2 x , = −2 y
∂x ∂y
∂z ∂z
For critical points =0=
∂x ∂y
⇒ x = 0, y = 0 ⇒ (0,0) is the critical point.
∂2 z ∂2 z ∂2z
A = 2 = −2 , B= = 0, C = 2 = −2
∂x ∂x ∂y ∂y
B 2 − AC = 0 − 4 = − 4 < 0 and A + C = −2 − 2 = −4 < 0
⇒ (0,0) is the point of maximum value
and maximum value of z at ( 0,0 ) is 1.
(ii) Do yourself as above
(iii) z = xy
∂z ∂z
=y , =x
∂x ∂y
∂z ∂z
For critical points =0=
∂x ∂y
⇒ y = 0 and x = 0 ⇒ (0,0) is the critical point.
∂2 z ∂2 z ∂2z
A= = 0, B= =1, C= =0
∂x 2 ∂x ∂y ∂y 2
B 2 − AC = (1) 2 − (0)(0) = 1 > 0
Therefore (0,0) is a saddle point.
(iv)z = x 3 − 3xy 2
∂z
= 0 ⇒ 3x2 − 3 y 2 = 0 ⇒ x = − y & x = y
∂x
∂z
= 0 ⇒ − 6 xy = 0 ⇒ xy = 0
∂y
⇒ either x = 0 or y = 0 or both are zero
⇒ (0,0) is the only critical point.
∂2 z
A = 2 = 6x = 0 at (0,0)
∂x
∂2 z
B= = − 6 y = 0 at (0,0)
∂x ∂y
∂2z
C = 2 = −6x = 0 at (0,0)
∂y
2
Remarks
⇒ B − AC = 0 and A + C = 0
2
so we need further consideration for the nature of point.

∆z = z (0 + h,0 + k ) − z (0,0)
= z (h, k ) − z (0,0)
= z (h, k ) = h 3 − 3hk
For h = k we have
> 0 if h < 0
∆z = h 3 − 3h 3 = −2h 3
< 0 if h > 0
⇒ (0,0) is a saddle point.
(v) z = f ( x, y ) = x 2 y 2
f x = 0 ⇒ 2 xy 2 = 0 , f y = 0 ⇒ 2 x2 y = 0
⇒ (0,0) is the critical point.
A = f xx = 2 y 2 = 0 at (0,0)
B = f xy = 4 xy = 0 at (0,0)
C = f yy = 2 x 2 = 0 at (0,0)
⇒ B 2 − AC = 0 and A + C = 0
so we need further consideration
∆f = f ( x0 + h, y0 + h) − f ( x0 , y0 )
= f (h, k ) − f (0,0) = h 2 k 2
If h = k , we have
∆f = h4 ≥ 0 ∀ h
Thus (0,0) is the point of minimum value.
Question
Find the critical points of the following functions and test for
maxima and minima.
(a) z = 1 − x2 − y2
(b)z = 2 x 2 − xy − 3 y 2 − 3x + 7 y
(c) z = 1 + x 2 + y 2
(d) z = x 2 − 5 xy − y 2
(e) z = x 2 − 2 xy + y 2
(f) z = x 3 − 3xy 2 + y 3
Solution
(a) z = 1 − x 2 − y 2
∂z 1 1
−x
= (1 − x − y ) ( −2 x ) =
2 2 −2
=0 ⇒ x=0
∂x 2 1 − x2 − y 2
∂z −y
= =0 ⇒ y =0
∂y 1− x − y
2 2

  −x 
−  1 − x2 − y 2 − x ⋅  
  1 − x2 − y 2 
∂2 z  
=
∂x 2
1 − x − y2
2
3
− 1 − x − y + x 
2 2 2
−1 + y 2 Remarks
= =
(1 − x 2 − y 2 ) 2 (1 − x − y2 )
3 3
2 2
∂2z
⇒ A = 2 = −1 at (0,0)
∂x
∂2 z ∂  ∂z  ∂  −y 
=  =  
∂x ∂y ∂x  ∂y  ∂x  1 − x 2 − y 2 

 1 − xy
( ) ( −2 x )
3
−
= − y ⋅  −  1 − x2 − y2 2
=
 2
3
(1 − x 2
−y 2
) 2
∂2z
⇒ B= =0 at (0,0)
∂x ∂y
 1  −y 
( )
−  1 − x 2 − y 2 2 (1) − y 
 1 − x2 − y 2


∂2z    −1 + x 2
= =
∂y 1 − x − y2
( )
2 2 3
1 − x2 − y 2 2
∂2 z
⇒ C= = −1 at (0,0)
∂y 2
⇒ B 2 − AC = 0 − (−1)(−1) = −1 < 0 and A + C = −1 − 1 = −2 < 0

⇒ z has a relative maxima at (0,0).
(b) z = 2 x 2 − xy − 3 y 2 − 3x + 7 y
∂z ∂z
= 4x − y − 3 , = −x − 6 y + 7
∂x ∂y
∂z ∂z
For critical points = 0, =0
∂x ∂y
⇒ 4 x − y − 3 = 0 …..………… (i)
& x + 6 y − 7 = 0 …………… (ii)
Multiplying equation (i) by 6 and adding in (ii)
24 x − 6 y − 18 = 0
x + 6y − 7 = 0
25 x − 25 = 0
⇒ x =1 ⇒ y =1
⇒ (1,1) is the critical point
∂2 z ∂2 z ∂2z
A = 2 = 4, B = = −1 , C = 2 = −6
∂x ∂x ∂y ∂y
B 2 − AC = (−1) 2 − (−4)(−6) = 25 > 0
⇒ There is a saddle point at (1,1) .
(c) z = 1 + x2 + y 2
∂z ∂z
= 2x , = 2y
∂x ∂y
∂z ∂z
For critical points = 0, =0 ⇒ (0,0) is the critical point.
∂x ∂y
4
Remarks
∂ z
2
∂ z 2
∂ z 2
A = 2 = 2, B= = 0, C = 2 = 2
∂x ∂x ∂y ∂y
⇒ B 2 − AC = (0) 2 − (2)(2) = − 4 < 0 and A + C = 2 + 2 = 4 > 0
⇒ The function has a relative minima at (0,0).
(d) z = x 2 − 5 xy − y 2
∂z ∂z
= 2x − 5 y , = −5 x − 2 y
∂x ∂y
∂z
= 0 ⇒ 2 x − 5 y = 0 …..……. (i)
∂x
∂z
= 0 ⇒ − 5 x − 2 y = 0 ……… (ii)
∂y
(i) and (ii) gives (0,0) is the critical point.
∂2 z ∂2 z ∂2z
A= 2 =2 , B= = −5 , C = 2 = −2
∂x ∂x ∂y ∂y
⇒ B 2 − AC = (−5)2 − (2)(−2) = 25 + 4 = 29 > 0
⇒ There is a saddle point at (0,0).
(e) z = x 2 − 2 xy + y 2
∂z ∂z
= 2x − 2 y , = 2 y − 2x
∂x ∂y
∂z ∂z
= 0, =0 ⇒ x− y=0 ⇒ x= y
∂x ∂y
⇒ Every point on the line y = x is a critical point.
∂2 z ∂2 z ∂2z
A = 2 = 2, B= = −2 , C = 2 =2
∂x ∂x ∂y ∂y
⇒ B 2 − AC = (−2)2 − (2)(2) = 4 − 4 = 0
Consider ∆z = z ( x + h , y + k ) − z ( x , y )
 x = y ∴ ∆z = z ( x + h, x + k ) − z ( x, x)
= ( x + h)2 − 2( x + h)( x + k ) + ( x + k )2
= [ ( x + h) − ( x + k ) ] ≥ 0
2
⇒ Each point on the line y = x gives a relative minimum.
(f) z = x 3 − 3xy 2 + y 3
∂z ∂z
= 3x 2 − 3 y 2 , = − 6 xy + 3 y 2
∂x ∂y
∂z
= 0 ⇒ 3x 2 − 3 y 2 = 0 …………... (i)
∂x
∂z
= 0 ⇒ − 6 xy + 3 y 2 = 0 ....……… (ii)
∂y
From (i) and (ii), we have
3x 2 − 6 xy = 0 ⇒ x( x − 2 y) = 0 ⇒ x = 0, x = 2 y
Now x = 0 ⇒ y = 0
And x = 2 y ⇒ (2 y )2 − y 2 = 0 ⇒ y = 0
Hence (0,0) is the only critical point.
5
Remarks
∂ z2
A = 2 = 6x = 0 at (0,0)
∂x
∂2 z
B= = −6 y = 0 at (0,0)
∂x ∂y
∂2 z
C = 2 = −6 x + 6 y = 0 at (0,0)
∂y
⇒ B − AC = 0
2
Consider ∆z = z (h, k ) − z (0,0)

= h 3 − 3hk 2 + k 3 = h 3 − 3h 3 + h 3 when h = k
< 0 when h > 0
= − h3
> 0 when h < 0
⇒ There is a saddle point at (0,0)
Note : (i) If for a point A = B = C = 0 and ∆z ≥ 0 , then z is minimum
at that point and if ∆z ≤ 0 , then z is maximum at that point.
(ii) If A, B, C are not zero and B 2 − AC = 0 then z is neither
maximum nor minimum.
Question
Find the critical points of the following functions and test for
maxima and minima.
(a) z = x 3 − 2 xy 2 + y 3
(b) z = x 3 + y 3 − 3x − 12 y + 20
(c) z = x 3 + y 3 − 63( x + y ) + 12 xy
(d) z = xy (a − x − y )
(e) z = x 2 − 2 xy + y 2 + x 3 − y 3 + 25
(f) z = x 2 y 2 − 5x 2 − 8 xy − 5 y 2
(g) z = 2( x − y )2 − x 4 − y 4
(h) z = 2( x − y )3 − ( x 4 − y 4 )
(i) z = x 2 − 5 xy − y 3
Solution
(a) z = x 3 − 2 xy 2 + y 3
∂z ∂z
= 3x 2 − 2 y 2 , = −4 xy + 3 y 2
∂x ∂y
∂z
= 0 ⇒ 3x 2 − 2 y 2 = 0 …….…….... (i)
∂x
∂z
= 0 ⇒ − 4 xy + 3 y 2 = 0 …………. (ii)
∂y
Adding (i) and (ii), we get
3x 2 − 4 xy + y 2 = 0 ⇒ 3x 2 − 3xy − xy + y 2 = 0
⇒ 3 x( x − y ) − y ( x − y ) = 0 ⇒ ( x − y )(3 x − y ) = 0
If x − y = 0 , then x = y in (i) gives
3x2 − 2 x2 = 0 ⇒ x = 0 ⇒ y = 0 .
And if 3 x − y = 0 , then y = 3 x in (i) gives
3x 2 − 2(3x)2 = 0 ⇒ x = 0 ⇒ y = 0
6
Remarks
∂ z
2
A = 2 = 6x = 0 at (0,0)
∂x
∂2 z
B= = −4 y = 0 at (0,0)
∂x ∂y
∂2z
C = 2 = 6 y − 4 x = 0 at (0,0)
∂y
⇒ A = B = C = 0 at (0,0) and hence B 2 − AC = 0
Now consider ∆z = z (h, k ) − z (0,0)
= h3 − 2hk 2 + k 3
= h 3 − 2h 3 + h 3 = 0 when h = k
⇒ The nature of the point is undetermined.
(b) z = x 3 + y 3 − 3x − 12 y + 20
∂z ∂z
= 3x 2 − 3 , = 3 y 2 − 12
∂x ∂y
∂z
= 0 ⇒ x2 − 1 = 0
∂x
∂z
= 0 ⇒ y2 − 4 = 0
∂y
⇒ x = ±1 , y = ± 2 , and the critical points are
(1, 2) , (1, −2) , (−1, 2) , (−1, −2)
∂2 z ∂2 z ∂2z
A = 2 = 6x , B= = 0, C = 2 = 6y
∂x ∂x ∂y ∂y
⇒ B 2 − AC = −36 xy
B 2 − AC = −36(1)(2) = −72 < 0 at (1, 2)
B 2 − AC = −36(1)(−2) = 72 > 0 at (1, −2)
B 2 − AC = −36(−1)(2) = 72 > 0 at (−1, 2)
B − AC = −36(−1)(−2) = − 72 < 0
2
at (−1, −2)
⇒ There is a saddle point at (1, −2) and (−1, 2) .
B 2 − AC < 0 while A = 6 > 0 at (1, 2)
and A = − 6 < 0 at (−1, −2)
⇒ z has relative minima at (1, 2) & relative maxima at (−1, −2) .
(c) z = x 3 + y 3 − 63( x + y ) + 12 xy
∂z ∂z
= 3 x 2 − 63 + 12 y , = 3 y 2 − 63 + 12 x
∂x ∂y
∂z ∂z
For critical points = 0, =0.
∂x ∂y
⇒ 3x 2 + 12 y − 63 = 0 ……………… (i)
& 3 y 2 + 12 x − 63 = 0 …….……….. (ii)
Subtracting (ii) from (i) , we get
3x 2 − 3 y 2 + 12 y − 12 x = 0
⇒ x 2 − y 2 + 4( y − x ) = 0
⇒ ( x − y )( x + y ) − 4( x − y ) = 0
⇒ ( x − y )( x + y − 4) = 0
7
Remarks
If x − y = 0 then (i) gives 3 x + 12 x − 63 = 0
2
⇒ x 2 + 4 x − 21 = 0
⇒ ( x + 7)( x − 3) = 0 ⇒ x = −7, 3
⇒ The critical points are (−7, −7) & (3,3) .
If x + y − 4 = 0 then x = 4 − y
Put this value of x in (ii) , we have
3 y 2 + 12(4 − y ) − 63 = 0
⇒ y 2 + 4(4 − y ) − 21 = 0
⇒ y2 − 4 y − 5 = 0
⇒ ( y − 5)( y + 1) = 0 ⇒ y = 5, − 1
y = 5 ⇒ x = −1 & y = −1 ⇒ x = 5
⇒ (−1,5) and (5, −1) are the other two critical points.
∂2 z ∂2 z ∂2z
A = 2 = 6x , B = = 12 , C = 2 = 6 y
∂x ∂x ∂y ∂y
⇒ B − AC = (12) − 36 xy = 144 − 36 xy
2 2
At (−7, −7) , we have

B 2 − AC = 144 − 36(−7)( −7) < 0 and A < 0
⇒ (−7,7) is a point of relative maximum value.
At (3,3) , we have
B 2 − AC = 144 − 36(3)(3) = 144 − 324 < 0 and A > 0 .
⇒ (3,3) is a point of relative minimum value.
At (−1,5) , we have
B 2 − AC = 144 − 36(−1)(5) > 0
⇒ (−1,5) is a saddle point.
B 2 − AC = 144 − ( −5)(1) > 0
⇒ (−5,1) is also a saddle point.
(d) z = xy (a − x − y ) = axy − x 2 y − xy 2
∂z
= ay − 2 xy − y 2
∂x
∂z
= ax − x 2 − 2 xy
∂y
∂z
= 0 ⇒ ay − 2 xy − y 2 = 0 ………………. (i)
∂x
∂z
= 0 ⇒ ax − x 2 − 2 xy = 0 ……………… (ii)
∂y
Subtracting (i) and (ii)
ay − 2 xy − y 2 = 0
ax − 2 xy − x 2 = 0
− + +
ay − ax − y 2 + x 2 = 0
⇒ ( x 2 − y 2 ) − a( x − y ) = 0
⇒ ( x − y )( x + y ) − a( x − y ) = 0
⇒ ( x − y )( x + y − a) = 0
8
If x − y = 0 ⇒ x = y then (i) give Remarks
ax − 2 x − x = 0
2 2
⇒ ax − 3x = 0 2
⇒ x (a − 3 x ) = 0 ⇒ x = 0 , a
3
⇒ (0,0) & a , a (3 3 )
are the critical points.
If x + y − a = 0 then y = a − x and (i) gives
a ( a − x ) − 2 x ( a − x) − ( a − x) 2 = 0
⇒ a 2 − ax − 2ax + 2 x 2 − a 2 − x 2 + 2ax = 0
⇒ x 2 − ax = 0 ⇒ x ( x − a ) = 0 ⇒ x = 0, a
⇒ (0, a ) & (a,0) are the other two critical points.
∂2 z ∂2 z ∂2 z
A = 2 = −2 y , B= = a − 2x − 2 y , C= = −2 x
∂x ∂x ∂y ∂x ∂y
⇒ B 2 − AC = ( a − 2 x − 2 y) 2 − 4 xy
At (0,0), we have B 2 − AC = a 2 > 0 ⇒ (0,0) is a saddle point.
( )
At a , a , we have
3 3
( ) ( )( )
2
B 2 − AC = a − 2a − 2a −4 a a
3 3 3 3
2 2
a 4a
= − < 0 and A < 0
9 9
a a
⇒  ,  is a point of maximum value.
3 3
At (0, a) , we have B 2 − AC = (a − 2a )2 − 4(0)(a) = a 2 > 0
⇒ (0, a ) is a saddle point.
At (a,0) , we have B 2 − AC = (a − 2a )2 − 4(a)(0) = a 2 > 0
⇒ (a,0) is also a saddle point.
(e) z = x 2 − 2 xy + y 2 + x 3 − y 3 + 25
∂z
= 2 x − 2 y + 3x2
∂x
∂z
= −2 x + 2 y − 3 y 2
∂y
∂z
= 0 ⇒ 3x 2 + 2 x − 2 y = 0 ……………. (i)
∂x
∂z
= 0 ⇒ 3 y 2 + 2 x − 2 y = 0 ………….…(ii)
∂y
Subtracting (i) and (ii), we have
3x2 − 3 y 2 = 0
⇒ 3( x − y )( x + y ) = 0
x − y = 0 ⇒ x = y , using in (i) we have
3x2 + 2 x − 2 x = 0 ⇒ x = 0
And x + y = 0 ⇒ x = − y , using in (i) we have
3x2 + 2 x + 2 x = 0
⇒ 3x 2 + 4 x = 0 ⇒ x (3x + 4) = 0
⇒ x = 0, x = − 4
3
9
4 4 Remarks
x=0 ⇒ y = 0 and x = − ⇒ y=
3 3
 4 4
⇒ The critical points are (0,0) &  − , 
 3 3
∂2 z
A = 2 = 2 + 6x
∂x
∂2 z
B= = −2
∂x ∂y
∂2 z
C = 2 = 2 − 6y
∂y
B − AC = 4 − (2 + 6 x )(2 − 6 y )
2
At (0,0), we have B 2 − AC = 4 − 4 = 0 ⇒ Nature undetermined

 4 4
At  − ,  , we have
 3 3
B 2 − AC = 4 − (2 − 8)(2 − 8) = 4 − (−6)(−6) < 0 and A < 0
 4 4
∴ Relative maximum at  − ,  .
 3 3
(f) z = x 2 y 2 − 5x 2 − 8 xy − 5 y 2
∂z
= 2 xy 2 − 10 x − 8 y
∂x
∂z
= 2 x 2 y − 10 y − 8 x
∂y
xy 2 − 5x − 4 y = 0 …………. (i)
x 2 y − 5 y − 4 x = 0 …………. (ii)
Adding (i) and (ii), we have
xy 2 + x 2 y − 9 x − 9 y = 0
⇒ xy ( y + x) − 9( x + y ) = 0
⇒ ( x + y )( xy − 9) = 0
x + y = 0 ⇒ y = − x in (i) gives
x3 − 5 x + 4 x = 0
⇒ x3 − x = 0 ⇒ x( x − 1)( x + 1) = 0
⇒ x = 0, 1, − 1
x=0 ⇒ y =0
x = 1 ⇒ y = −1
x = −1 ⇒ y = 1
⇒ (0,0), (1, −1), (−1,1) are the critical points.
9
If xy − 9 = 0 , then y = in (i) gives x 2 − 9 = 0 ⇒ x = ± 3
x
x = 3 ⇒ y = 3 and x = −3 ⇒ y = −3
⇒ (3,3) & (−3, −3) are also the critical points.
∂2 z ∂2 z ∂2z
A= = 2 y 2
− 10 , B = = 4 xy − 8 , C = = 2 x 2 − 10
∂x 2
∂x ∂y ∂y 2
B 2 − AC = (4 xy − 8)2 − (2 y 2 − 10)(2 x 2 − 10)

10
Remarks
At (0,0), we have
B 2 − AC = 64 − (−10)(−10) < 0 and A = −10 < 0
⇒ (0,0) is the point of maximum value.
At (1, −1) , we have
B 2 − AC = (−4 − 8) 2 − (2 − 10)(2 − 10) = 144 − 64 > 0
⇒ (1, −1) is a saddle point.
B 2 − AC = (−4 − 8) 2 − (2 − 10)(2 − 10) = 144 − 64 > 0
⇒ (−1,1) is a saddle point.
At (3,3) , we have
B 2 − AC = (36 − 8)2 − (18 − 10)(18 − 10) = (24)2 − 64 > 0
⇒ (3,3) is a saddle point.
At (−3, −3) , we have
B 2 − AC = (36 − 8)2 − (8)(8) > 0
⇒ (−3, −3) is again a saddle point.
(g) z = 2( x − y ) 2 − x 4 − y 4
∂z
= 4( x − y ) − 4 x 3
∂x
∂z
= −4( x − y ) − 4 y 3
∂y
For critical points
∂z
= 0 ⇒ x − y − x 3 = 0 ……………. (i)
∂x
∂z
= 0 ⇒ − x + y − y 3 = 0 ………..… (ii)
∂y
Addition of (i) and (ii) gives
x3 + y 3 = 0
⇒ ( x + y )( x 2 − xy + y 2 ) = 0
⇒ x + y = 0 or x 2 − xy + y 2 = 0 which gives imaginary values.
x + y = 0 ⇒ y = − x in (i) gives
x + x − x 3 = 0 ⇒ 2 x − x3 = 0
⇒ x(2 − x 2 ) = 0 ⇒ x = 0, ± 2
x=0 ⇒ y =0
x= 2 ⇒ y=− 2
x=− 2 ⇒ y= 2
⇒ The critical points are ( 0,0 ) , ( ) (
2, − 2 , − 2, 2 . )
∂2 z ∂2 z ∂2z
A = 2 = 4 − 12 x 2 , B= = −4 , C = 2 = 4 − 12 y 2
∂x ∂x ∂y ∂y
B 2 − AC = 16 − (4 − 12 x 2 )(4 − 12 y 2 )
At (0,0), we have B 2 − AC = 0
Consider ∆z = z (h, k ) − z (0,0)
= 2(h − k )2 − h 4 − k 4 = −2h 4 ≤ 0 if h = k
⇒ (0,0) is the points of maximum value.
11
At ( )
2, − 2 , we have Remarks
B 2 − AC = 16 − (4 − 24)(4 − 24)
= 16 − (−20)(−20) < 0 and A < 0 .
⇒ ( )
2, − 2 is a point of maximum value.
(
At − 2, 2 , we have)
B 2 − AC = 16 − (4 − 24)(4 − 24) < 0 and A < 0 .
( )
⇒ − 2, 2 is also a point of maximum vale.
(h) z = 2( x − y )3 − ( x 4 − y 4 )
∂z
= 6( x − y )2 − 4 x 3 = 0 ……..…….. (i)
∂x
∂z
= −6( x − y )2 + 4 y 3 = 0 …………. (ii)
∂y
Adding (i) and (ii), we get
y 3 − x3 = 0 ⇒ ( y − x)( y 2 + xy + x 2 ) = 0
y − x = 0 ⇒ y = x in (i) gives
4 x3 = 0 ⇒ x = 0 ⇒ y = 0
x 2 + xy + y 2 = 0 gives imaginary values
⇒ (0,0) is the only critical point
∂2 z
A = 2 = 12( x − y ) − 12 x 2
∂x
∂2 z
B= = −12( x − y )
∂x ∂y
∂2 z
C= = 12( x − y) + 12 y 2
∂y 2
at (0,0), A = B = C = 0 ⇒ B 2 − AC = 0
Consider ∆z = z (h, h ) − z (0,0) = 0
⇒ Nature undecided.
(i) z = x 2 − 5 xy − y 3
∂z
= 2 x − 5 y = 0 ……………….. (i)
∂x
∂z
= −5 x − 3 y 2 = 0 …………….. (ii)
∂y
2x
From (i) y =
5
 4x2 
(ii) becomes −5 x − 3  =0
 25 
⇒ − 125 x − 12 x 2 = 0
⇒ 12 x 2 + 125 x = 0
125
⇒ x (12 x + 125) = 0 ⇒ x = 0, −
12
12
125 2  125  25 Remarks
x=0 ⇒ y =0 & x=− ⇒ y = − =−
12 5  12  6
 125 25 
⇒ (0,0) &  − , −  are the critical points
 12 6 
∂ z
2
∂ z
2
∂2 z
A = 2 = 2, B = = −5 , C = 2 = −6 y
∂x ∂x ∂y ∂y
B 2 − AC = 25 + 12 y
At (0,0), we have B 2 − AC = 25 > 0 ⇒ (0,0) is a saddle point.
 125 25 
At  − , −  , we have
 12 6 
 25 
B 2 − AC = 25 + 12  −  = −25 < 0 and A = 2 > 0
 6 
 125 25 
∴ − , −  is a point of maximum value.
 12 6 
http://www.mathcity.tk
Chapter 6 – Riemann-Stieltjes Integral.
Source: Syyed Gul Shah (Chairman, Department of Mathematics, US Sargodha)
Collected & Composed by: Atiq ur Rehman (atiq@mathcity.org), http://www.mathcity.org
Ø Introduction
In elementary treatment of Integral Calculus the subject of integration is
treated as inverse of differentiation. The subject arose in connection with the
determination of areas of plane regions and was based on the notion of the limit of a
type of sum when the number of terms in the sum tends to infinity and each term
tends to zero. In fact the name Integral Calculus has its origin in this process of
summation. It was only afterwards that it was seen that the subject of integration
can also be viewed from the point of the inverse of differentiation.
Ø Partition
Let [a, b] be a given interval. A finite set P = {a = x0 , x1 , x2 ,...., xk ,...., xn = b} is
said to be a partition of [a, b] which divides it into n such intervals
[ x0 , x 1 ] , [ x1, x 2 ] , [ x2 , x 3 ] ,......, [ xn-1, x n ]
Each sub-interval is called a component of the partition.
Obviously, corresponding to different choices of the points xi we shall have
different partition.
The maximum of the length of the components is defined as the norm of the
partition.
Ø Riemann Integral
Let f be a real-valued function defined and bounded on [a, b] . Corresponding to
each partition P of [a, b] , we put
M i = sup f ( x) ( xi-1 £ x £ xi )
mi = inf f ( x) ( xi-1 £ x £ xi )
We define upper and lower sums as Mi
n
U ( P, f ) = å M i Dxi mi
i =1 xi – 1 xi
n
and L ( P, f ) = å mi Dxi
i =1
where Dxi = xi - xi-1 (i = 1,2,...., n)
b
and finally ò f dx = inf U ( P, f ) ………….… (i)
a
b
ò f dx = sup L( P, f ) ……………..(ii)
a
Where the infimum and the supremum are taken over all partitions P of [a, b] .
b b
Then ò f dx
a
and ò f dx
a
are called the upper and lower Riemann Integrals of f
over [a, b] respectively.

In case the upper and lower integrals are equal, we say that f is Riemann-
Integrable on [a, b] and we write f ÎR , where R denotes the set of Riemann
integrable functions.
2 Riemann-Stieltjes Integral
b b
The common value of (i ) and (ii ) is denoted by ò f dx
a
or by ò f ( x) dx .
a
Which is known as the Riemann integral of f over [a, b] .
Ø Theorem
The upper and lower integrals are defined for every bounded function f .
Proof
Take M and m to be the upper and lower bounds of f ( x) in [a, b] .
Þ m £ f ( x) £ M (a £ x £ b )
Then M i £ M and mi ³ m (i = 1,2,....., n)
Where M i and mi denote the supremum and infimum of f ( x) in ( xi-1 , xi ) for
certain partition P of [ a, b] .
n n
Þ L ( P, f ) = å mi Dxi ³ å m Dxi (Dxi = xi -1 - xi )
i =1 i =1
n
Þ L ( P, f ) ³ må Dxi
i =1
n
But å Dx
i =1
i = ( x1 - x0 ) + ( x2 - x1 ) + ( x3 - x2 ) + .... + ( xn - xn -1 )
= xn - x0 = b - a
Þ L ( P, f ) ³ m(b - a)
Similarity U ( P, f ) £ M (b - a)
Þ m(b - a) £ L ( P, f ) £ U ( P, f ) £ M (b - a)
Which shows that the numbers L ( P, f ) and U ( P, f ) form a bounded set.
Þ The upper and lower integrals are defined for every bounded function f . ¤
Ø Riemann-Stieltjes Integral
It is a generalization of the Riemann Integral. Let a ( x) be a monotonically
increasing function on [a, b] . a (a ) and a (b) being finite, it follows that a ( x)
is bounded on [a, b] . Corresponding to each partition P of [a, b] , we write
Da i = a ( xi ) - a ( xi -1 )
( Difference of values of a at xi & xi -1 )
Q a ( x) is monotonically increasing.
\ Da i ³ 0
Let f be a real function which is bounded on [a, b] .
n
Put U ( P, f ,a ) = å M i Da i
i =1
n
L ( P, f ,a ) = å mi Da i
i =1
Where M i and mi have their usual meanings.
Define
b
ò f da = inf U ( P, f ,a ) ……………. (i)

a
b
ò f da = sup L ( P, f ,a ) ……………. (ii)

a
Riemann-Stieltjes Integral 3
Where the infimum and supremum are taken over all partitions of [a, b] .
b b b b
If ò f da = ò f da ,
a a
we denote their common value by òa f da or òa f ( x) da ( x) .
This is the Riemann-Stieltjes integral or simply the Stieltjes Integral of f w.r.t.
a over [a, b] .
b
If òa f da exists, we say that f is integrable w.r.t. a , in the Riemann sense,
and write f ÎR(a ) .
Ø Note
The Riemann-integral is a special case of the Riemann-Stieltjes integral when we
take a ( x ) = x .
Q The integral depends upon f ,a , a and b but not on the variable of integration.
b b
\ We can omit the variable and prefer to write òa f da instead of òa f ( x) da ( x) .
In the following discussion f will be assume to be real and bounded, and a
monotonically increasing on [a, b] .
Ø Refinement of a Partition
Let P and P* be two partitions of an interval [ a, b] such that P Ì P* i.e. every
point of P is a point of P* , then P* is said to be a refinement of P .
Ø Common Refinement
Let P1 and P2 be two partitions of [a, b] . Then a partition P* is said to be their
common refinement if P* = P1 È P2 .
Ø Theorem
If P* is a refinement of P , then
(
L ( P, f ,a ) £ L P* , f ,a ……..………… (i) )
and U ( P, f ,a ) ³ U ( P , f ,a ) ………………. (ii)
*
Proof
Let us suppose that P* contains just one point x * more than P such that
xi-1 < x* < xi where xi -1 and xi are two consecutive points of P .
Put
w1 = inf f ( x) (
xi-1 £ x £ x* )
xi – 1 x* xi
w2 = inf f ( x ) (x *
£ x £ xi )
It is clear that w1 ³ mi & w2 ³ mi where mi = inf f ( x ) , ( xi -1 £ x £ xi ) .
Hence
( )
L P* , f ,a - L ( P, f ,a ) = w1 éëa ( x* ) - a ( xi -1 ) ùû + w2 éëa ( xi ) - a ( x* ) ùû
-mi [a ( xi ) - a ( xi -1 )]
= w1 éëa ( x* ) - a ( xi -1 ) ùû + w2 éëa ( xi ) - a ( x* ) ùû
- mi éëa ( xi ) - a ( x* ) + a ( x* ) - a ( xi -1 ) ùû
= ( w1 - mi ) éëa ( x* ) - a ( xi-1 ) ùû + ( w2 - mi ) éëa ( xi ) - a ( x* ) ùû
Q a is a monotonically increasing function.

\ a ( x* ) - a ( xi -1 ) ³ 0 , a ( xi ) - a ( x* ) ³ 0
( )
Þ L P * , f ,a - L ( P, f ,a ) ³ 0
(
Þ L ( P , f ,a ) £ L P * , f ,a ) which is (i)
If P* contains k points more than P , we repeat this reasoning k times and
arrive at (i).
Now put
W1 = sup f ( x ) ( xi-1 £ x £ x* )
and W2 = sup f ( x) ( x* £ x £ xi )
Clearly M i ³ W1 & M i ³ W2
Consider
( )
U ( P, f ,a ) - U P* , f ,a = M i [a ( xi ) - a ( xi -1 )]
- W1 éëa ( x* ) - a ( xi -1 ) ùû - W2 éëa ( xi ) - a ( x* ) ùû
= M i éëa ( xi ) - a ( x* ) + a ( x* ) - a ( xi -1 ) ùû
- W1 éëa ( x* ) - a ( xi -1 ) ùû - W2 éëa ( xi ) - a ( x* ) ùû
= ( M i - W1 ) éëa ( x* ) - a ( xi -1 ) ùû + ( M i - W2 ) éëa ( xi ) - a ( x* ) ùû ³ 0
(Q a is - )
Þ U ( P , f ,a ) ³ U P * , f ,a ( ) which is (ii)
¤
Ø Theorem
Let f be a real valued function defined on [a, b] and a be a monotonically
increasing function on [a, b] . Then
sup L ( P, f ,a ) £ inf U ( P, f ,a )
b b
i.e. ò f da
a
£ ò f da
a
Proof
Let P* be the common refinement of two partitions P1 and P2 . Then
(
L ( P1, f ,a ) £ L P* , f ,a ) (
£ U P * , f ,a ) £ U ( P2 , f ,a )
Hence L ( P1 , f ,a ) £ U ( P2 , f ,a ) …………. (i)
If P2 is fixed and the supremum is taken over all P1 then (i) gives
b
ò f da £ U ( P2 , f ,a )
a
Now take the infimum over all P2

b b
Þ ò f da
a
£ ò f da
a
¤
{{{{{{{{{{{{{{{{{{{{
Ø Theorem (Condition of Integrability or Cauchy’s Criterion for

Integrability.)
f ÎR(a ) on [a, b] iff for every e > 0 there exists a partition P such that
U ( P, f ,a ) - L ( P, f ,a ) < e
Proof
Let U ( P, f ,a ) - L ( P, f ,a ) < e …………. (i)
b b
Then L ( P, f ,a ) £ ò f da £ ò f d a £ U ( P , f ,a )
a a
b b
Þ ò f da - L ( P, f ,a ) ³ 0 and U ( P, f ,a ) - ò f da ³ 0
a a
Adding these two results, we have

b b
ò f da - ò f da - L ( P, f ,a ) + U ( P, f ,a ) ³ 0
a a
b b
Þ ò f da - ò f da £ U ( P, f ,a ) - L ( P, f ,a ) < e from (i)
a a
b b
i.e. 0£ ò f da - ò f da
a a
<e for every e > 0 .
b b
Þ ò f da
a
= ò f da
a
i.e. f ÎR(a )
Conversely, let f ÎR(a ) and let e > 0

b b b
Þ ò f da = ò f da = ò f da
a a a
b b
Now ò f da = inf U ( P, f ,a )
a
and ò f da = sup L ( P, f ,a )
a
There exist partitions P1 and P2 such that

e
b
U ( P2 , f ,a ) - ò f da < ……..…… (ii) U ( P2 , f , a ) - e < ò f da
a
2 2
e ò f da < L ( P1 , f ,a ) + e 2
b
and òa f da - L ( P1 , f ,a ) <
……..…… (iii)
2
We choose P to be the common refinement of P1 and P2 .
Then
e
b
U ( P, f ,a ) £ U ( P2 , f ,a ) < ò f da + < L ( P1 , f ,a ) + e £ L ( P, f ,a ) + e
a
2
So that
U ( P, f ,a ) - L ( P, f ,a ) < e ¤
•-----------------------------œ
Ø Theorem
a) If U ( P, f ,a ) - L ( P, f ,a ) < e holds for some P and some e , then it holds
(with the same e ) for every refinement of P .
b) If U ( P, f ,a ) - L ( P, f ,a ) < e holds for P = { x0 ,...., xn } and si , ti are arbitrary
points in [ xi -1, xi ] , then
n
å
i =1
f ( si ) - f (ti ) Da i < e
c) If f ÎR(a ) and the hypotheses of (b) holds, then

n b
å f (t )Da - ò f da
i =1
i i <e
a
Proof
a) Let P* be a refinement of P . Then
L ( P, f ,a ) £ L P * , f ,a ( )
and (
U P * , f ,a ) £ U ( P, f ,a )
Þ L ( P , f ,a ) + U P * , f ,a ( ) ( )
£ L P * , f ,a + U ( P , f ,a )
(
Þ U P * , f ,a - L P * , f ,a ) ( ) £ U ( P , f ,a ) - L ( P , f ,a )
Q U ( P, f ,a ) - L ( P, f ,a ) < e
( ) (
\ U P * , f ,a - L P * , f ,a < e )
b) P = { x0 ,...., xn } and si , ti are arbitrary points in [ xi -1, xi ] .
Þ f ( si ) and f (ti ) both lie in [ mi , M i ] .
Þ f ( si ) - f (ti ) £ M i - mi xi – 1 si ti xi
Þ f ( si ) - f (ti ) Da i £ M i Da i - mi Da i
n n n
Þ åi =1
f ( si ) - f (ti ) Da i £ åM i =1
i Da i - å mi Da i
i =1
n
Þ åi =1
f ( si ) - f (ti ) Da i £ U ( P, f ,a ) - L ( P, f ,a )
Q U ( P , f ,a ) - L ( P , f ,a ) < e
n
\ å
i =1
f (si ) - f (ti ) Da i < e
c) Q mi £ f (ti ) £ M i
\ å m Da i å f (t ) Da £ å M Da
i £ i i i i
Þ L ( P, f ,a ) £ å f (t ) Da £ U ( P, f ,a ) i i
b
and also L ( P, f ,a ) £ òa f da £ U ( P, f ,a )
Using (b), we have

b
å f (t ) Da - ò f da
i i < e ¤
a
˜-----------------------------™
Ø Theorem
If f is continuous on [a, b] then f ÎR(a ) on [a, b] .
Proof
Let e > 0 be given. Choose b > 0 so that
[a (b) - a (a)] b < e
f is continuous on [a, b] Þ f is uniformly continuous on [a, b] .
Þ There exists a d > 0 such that
f ( s) - f (t ) < b if x Î [a, b] , t Î [a, b] and x - t < d ………..(i)
If P is any partition of [a, b] such that Dxi < d for all i
then (i) implies that M i - mi £ b , (i = 1,2,...., n)
Þ U ( P, f ,a ) - L ( P, f ,a ) = å M i Da i - å mi Da i
= å ( M i - mi ) Da i
£ b å Da i = b [a (b) - a (a )] < e
Þ f ÎR (a ) by Cauchy Criterion. ¤
Ø Theorem
If f is monotonic on [a, b] , and if a is continuous on [a, b] , then f ÎR(a ) .
( Monotonicity of a still assumed. )
Proof
Let e > 0 be a given positive number.
For any positive integer n , choose a partition P = { x0 , x1,....., xn } of [a, b]
such that
a (b) - a (a )
Da i = , i = 1,2,...., n
n
This is possible because a is continuous and monotonic increasing on the closed
interval [a, b] and thus assumes every value between its bounds, a (a ) and a (b) .
Let f be monotonic increasing on [a, b] , so that its lower and upper bounds
mi , M i in [ xi -1, xi ] are given by
mi = f ( xi -1 ) , M i = f ( xi ) , i = 1,2,...., n
n
\ U ( P, f ,a ) - L ( P, f ,a ) = å ( M i - mi ) Da i
i =1
a (b) - a (a ) n
= å [ f ( xi ) - f ( xi-1 )]
n i =1
a (b) - a (a )
= [ f (b) - f (a)]
n
<e if n is taken large enough.
Þ f Î R(a ) on [a, b] .
¤
Note: f ÎR(a ) when either
i) f is continuous and a is monotonic, or
ii) f is monotonic and a is continuous, of course a is still monotonic.
Ø Properties of Integral
i) If f ÎR(a ) on [a, b] , then cf ÎR(a ) for every constant c and
b b
òa cf da = c òa f da .
Proof
Q f Î R (a )
\ $ a partition P such that
U ( P, f ,a ) - L ( P, f ,a ) < e , where e is an arbitrary +ive number.
n n
Now U ( P, cf ,a ) = å cM i Da i = c å M i Da i
i =1 i =1
n n
& L ( P, cf ,a ) = å cmi Da i = c å mi Da i
i =1 i =1
Þ U ( P, f ,a ) - L ( P, f ,a ) = c éëå M i Da i - å mi Da i ùû
= c éëU ( P, f ,a ) - L ( P, f ,a ) ùû
< ce = e1
Þ cf ÎR(a )
Q U ( P, cf ,a ) = c éëU ( P, f ,a ) ùû & L ( P, cf ,a ) = c éë L ( P, f ,a ) ùû
\ inf U ( P, cf ,a ) = c éëinf U ( P, f ,a ) ùû & sup L ( P, cf ,a ) = c éësup L ( P, f ,a ) ùû
where infimum and supremum are taken over all P on [a, b] .
b b b b
Þ ò cf da = c ò f da & ò cf da = c ò f da
a a a a
b b b b
Q ò cf da = ò cf da
a a
and ò f da = ò f d a
a a
b b
\ òa cf da = c òa f da ¤
ii) If f1 ÎR (a ) and f 2 ÎR (a ) on [a, b] , then f1 + f 2 ÎR (a ) and

b b b
òa ( f1 + f 2 ) da = òa f1 da + òa f 2 da .
Proof
If f = f1 + f 2 and P is any partition of [a, b] , we have
mi¢ + mi¢¢ £ mi £ M i £ M i¢ + M i¢¢
where M i¢ , mi¢ , M i¢¢, mi¢¢ and M i , mi are the bounds of f1 , f 2 and f respectively in
[ xi-1, xi ] .
Multiplying throughout by Da i and adding the inequalities for i = 1,2,...., n ,
we get
L ( P , f1 ,a ) + L ( P , f 2 ,a ) £ L ( P , f ,a ) £ U ( P, f ,a ) £ U ( P, f1 ,a ) + U ( P , f 2 ,a ) ….....(i)
Since f1 ÎR (a ) and f 2 ÎR (a ) on [a, b] therefore $ e > 0 and there are
partitions P1 and P2 such that
U ( P1 , f1 ,a ) - L ( P1 , f1 ,a ) < e
ü
ý ………….. (ii)
and U ( P2 , f 2 ,a ) - L ( P2 , f 2 ,a ) < e þ
These inequalities hold if P1 and P2 are replaced by their common refinement P .
(ii) Þ éëU ( P, f1 ,a ) + U ( P, f 2 ,a ) ùû - éë L ( P, f1,a ) + L ( P, f 2 ,a ) ùû < 2e

Using (i) we have
U ( P, f ,a ) - L ( P, f ,a ) < 2e
which proves that f ÎR(a ) on [a, b]
With the same partition P , we have
b
U ( P, f1,a ) < òa f1 da + e
b
and U ( P, f 2 ,a ) < òa f 2 da + e
Hence (i) implies that
b b b
òa f da £ U ( P, f ,a ) < òa f1 da + òa f 2 da + 2e
Q e is arbitrary, we conclude that
b b b
òa f da £ òa f1 da + òa f 2 da
Similarly if we consider the lower sums we arrive at
b b b
òa f da ³ òa f1 da + òa f 2 da
Combining the above two results, we have
b b b
òa f da = òa f1 da + òa f 2 da ¤
iii) If f1 ( x) £ f 2 ( x) on [a, b] , then

b b
òa f1 da £ òa f 2 da
Proof
Let f ( x ) ³ 0 , then M i ³ 0 Þ U ( P , f ,a ) ³ 0
b
and \ òa f da ³ 0
Q f1 £ f 2 \ f 2 - f1 ³ 0
b b b
Þ òa ( f 2 - f1 ) da ³ 0 Þ òa f 2 da - òa f1 da ³ 0
b b
Þ òa f1 da £ òa f 2 da ¤
Ø Note
(i) (f+ g ) ( x ) = f ( x ) + g ( x ) £ sup f + sup g
Þ sup ( f + g ) £ sup f + sup g
(ii) (f+ g ) ( x) = f ( x) + g ( x) ³ inf f + inf g

Þ inf ( f + g ) ³ inf f + inf g
iv) If f ÎR(a ) on [a, b] and if a < c 0 , $ a partition P such that
U ( P, f ,a ) - L ( P, f ,a ) < e
Let P* be the refinement of P such that P* = P È {c}
( ) ( )
\ L ( P, f ,a ) £ L P* , f ,a £ U P* , f ,a £ U ( P, f ,a ) ….…...... (i)
Þ U ( P , f ,a ) - L ( P , f ,a ) £ U ( P , f ,a ) - L ( P, f ,a ) < e
* *
……….(ii)
Let P1 , P2 denote the sets of points of P* between [a, c ] , [c, b] respectively.
Clearly P1 , P2 are partitions of [a, c ] , [c, b] respectively and P* = P1 È P2 .
Also ( )
U P* , f ,a = U ( P1 , f ,a ) + U ( P2 , f ,a ) ……….. (iii)
and L ( P , f ,a ) = L ( P , f ,a ) + L ( P , f ,a ) ………… (iv)
*
1 2
\ {U ( P1 , f ,a ) - L ( P1, f ,a )} + {U ( P2 , f ,a ) - L ( P2 , f ,a )}
( ) ( )
= U P * , f ,a - L P * , f ,a < e
Since each bracket on the left is non-negative, it follows that
U ( P1, f ,a ) - L ( P1, f ,a ) < e
and U ( P2 , f ,a ) - L ( P2 , f ,a ) < e
Þ f ÎR (a ) on [a, c ] and on [c, b] .
We know that for any functions f1 and f 2 , if f = f1 + f 2 , then
inf f ³ inf f1 + inf f 2
and sup f £ sup f1 + sup f 2
Now for any partitions P1 , P2 of [a, c ] , [c, b] respectively, if P* = P1 È P2 , then
( )
U P* , f ,a = U ( P1 , f ,a ) + U ( P2 , f ,a )
Hence on taking the infimum for all partitions, we get
b c b
ò f da
a
³ ò f da + ò f da
a c
But since f ÎR(a ) on [a, c ] , [c, b] , [a, b ]
b c b
\ òa f da ³ òa f da + òc f da …………. (v)
Again ( )
L P* , f ,a = L ( P1 , f ,a ) + L ( P2 , f ,a )
and on taking the supremum for all partitions, we get
b c b
ò f da
a
£ ò f da + ò f da
a c
But since f ÎR(a ) on [a, c ] , [c, b] , [a, b ]

b c b
\ òa f da £ òa f da + òc f da ……………(vi)
(v) and (vi) imply that

b c b
òa f da = òa f da + òc f da ¤
v) If f ÎR(a ) on [a, b] and f ( x ) £ M on [a, b] , then

b
ò f da £ M [a (b) - a (a )]
a
Proof
We know that
b
òa f da £ U ( P, f ,a )
= å M i Da i £ M å Da i
But
å Da i = a (b) - a (a )
b
Þ ò f da £ M [a (b) - a (a ) ] ¤
a
vi) If f ÎR (a1 ) and f ÎR(a 2 ) , then f Î R(a1 + a 2 ) and

b b b
òa f d (a1 + a 2 ) = òa f da1 + òa f da 2
and if f ÎR(a ) and c is a positive constant, then f ÎR(ca ) and
b b
òa f d (ca ) = c òa f da
Proof
Since f ÎR (a1 ) and f ÎR(a 2 ) , therefore for e > 0 , there exists partitions
P1, P2 of [a, b] such that
e
U ( P1, f ,a1 ) - L ( P1 , f ,a1 ) <
2
e
and U ( P2 , f ,a 2 ) - L ( P2 , f ,a 2 ) <
2
Let P = P1 È P2
e ü
\ U ( P, f ,a1 ) - L ( P, f ,a1 ) <
2 ïï ………….. (i)
ý
e ï
& U ( P, f ,a 2 ) - L ( P, f ,a 2 ) <
2 ïþ
Let mi , M i be bounds of f in [ xi -1, xi ]
Take a = a1 + a 2
Þ Da i = Da1i + Da 2 i
\ U ( P , f ,a ) = å M i Da i
(
= å M i Da1i + Da 2 i )
= U ( P, f ,a1 ) + U ( P, f ,a 2 )
Similarly
L ( P, f ,a ) = L ( P, f ,a1 ) + L ( P, f ,a 2 )
\ U ( P, f ,a ) - L ( P, f ,a ) = U ( P, f ,a1 ) - L ( P, f ,a1 ) + U ( P, f ,a 2 ) - L ( P, f ,a 2 )
e e
< + = e by (i)
2 2
Þ f ÎR (a ) where a = a1 + a 2
To prove the second part, we notice that

b
òa f da = inf U ( P, f ,a )
= inf {U ( P, f ,a1 ) + U ( P, f ,a 2 )}
³ inf U ( P, f ,a1 ) + inf U ( P, f ,a 2 )
b b
= òa f da1 + òa f da 2 ………….... (ii)
Similarly by taking the supremum of lower sum of partition we arrive that

b b b
òa f da £ òa f da1 + òa f da 2 ……...…….. (iii)
From (ii) and (iii)

b b b
òa f da = òa f da1 + òa f da 2
b b b
i.e. òa f d (a1 + a 2 ) = òa f da1 + òa f da 2 Q a = a1 + a 2
Now Q f Î R (a ) \ for e > 0 , $ a partition P of [a, b] such that

U ( P, f ,a ) - L ( P, f ,a ) < e …………. (iv)
Let a ¢ = ca then Da i¢ = D (ca i ) = cDa i
Þ U ( P, f ,a ¢ ) = å M i Da i¢
= å M i ( cD a i )
= c å M i Da i
= c U ( P , f ,a )
Similarly, L ( P, f ,a ¢) = c L ( P, f ,a )
Þ U ( P, f ,a ¢ ) - L ( P, f ,a ¢) = c {U ( P, f ,a ) - L ( P, f ,a )} < c e by (iv)
Þ f ÎR (a ¢) where a ¢ = ca
b
Also òa f da ¢ = inf U ( P, f ,a ¢ )
= inf c U ( P, f ,a )
= c inf U ( P, f ,a )
b
= c ò f da
a
and
b
òa f da ¢ = sup L ( P, f ,a ¢ )
= sup cU ( P, f ,a )
= c supU ( P, f ,a )
b
= c ò f da
a
Hence
b b
òa f da ¢ = c òa f da where a ¢ = ca ¤
Ø Lemma
If M & m are the supremum and infimum of f and M ¢ , m¢ are the
supremum & infimum of f on [a, b] then M ¢ - m¢ £ M - m .
Proof
Let x1, x2 Î [a, b] , then
f ( x1 ) - f ( x2 ) £ f ( x1 ) - f ( x2 ) ………………(A)
Q M and m denote the supremum and infimum of f ( x) on [a, b]
\ f ( x) £ M & f ( x ) ³ m " x Î [ a, b]
Q x1 , x2 Î [a, b]
\ f ( x1 ) £ M and f ( x2 ) ³ m
Þ f ( x1 ) £ M and - f ( x2 ) £ - m
Þ f ( x1 ) - f ( x2 ) £ M - m ……………... (i)
Interchanging x1 & x2 , we get
- [ f ( x1 ) - f ( x2 )] £ M - m ………….. (ii)
(i) & (ii) Þ f ( x1 ) - f ( x2 ) £ M - m
Þ f ( x1 ) - f ( x2 ) £ M - m by eq. (A) …….…..(I)
Q M ¢ and m¢ denote the supremum and infimum of f ( x ) on [a, b]
\ f ( x) £ M ¢ and f ( x) ³ m¢ " x Î [a, b]
Þ $ e > 0 such that
f ( x1 ) > M ¢ - e ………….. (iii)
and f ( x2 ) < m¢ + e Þ - f ( x2 ) + e > - m¢ ………….. (iv)
From (iii) and (iv), we get
f ( x1 ) - f ( x2 ) + e > M ¢ - m¢ - e
Þ 2e + f ( x1 ) - f ( x2 ) > M ¢ - m¢
Q e is arbitrary \ M ¢ - m¢ £ f ( x1 ) - f ( x2 ) …….…..…. (v)
Interchanging x1 & x2 , we get
M ¢ - m¢ £ - ( f ( x1 ) - f ( x2 ) ) …………… (vi)
Combining (v) and (vi), we get
M ¢ - m¢ £ f ( x1 ) - f ( x2 ) ……………. (II)
From (I) and (II), we have the require result
M ¢ - m¢ £ M - m ¤
Ø Theorem
b b
If f ÎR(a ) on [a, b] , then f ÎR (a ) on [a, b] and ò f da
a
£ ò
a
f da .
Proof
Q f Î R (a )
\ given e > 0 $ a partition P of [a, b] such that
U ( P, f ,a ) - L ( P, f ,a ) < e
i.e. åM i Da i - å mi Da i = å ( M i - mi ) Da i < e
Where M i and mi are supremum and infimum of f on [ xi -1, xi ]
Now if M i¢ and mi¢ are supremum and infimum of f on [ xi -1, xi ] then
M i¢ - mi¢ £ M i - mi
Þ å ( M ¢ - m¢ ) Da £ å ( M - m ) Da
i i i i i i
Þ U ( P , f ,a ) - L ( P, f ,a ) £ U ( P , f ,a ) - L ( P , f ,a ) < e
Þ f ÎR (a ) .
Take c = +1 or -1 to make c ò f da ³ 0
b b
Then ò f da
a
= c ò f da …….……… (i)
a
Also c f ( x) £ f ( x) " x Î [ a, b]
b b b b
Þ òa c f da £ òa f da Þ c ò f da £
a
òa f da ………. (ii)
From (i) and (ii), we have

b b
ò f da
a
£ ò
a
f da ¤
Ø Theorem
If f ÎR(a ) on [a, b] , then f 2 ÎR (a ) on [a, b] .
Proof
Q f ÎR(a ) Þ f ÎR (a )
Þ f ( x) < M " x Î [ a, b]
Q f Î R (a ) \ given e > 0 , $ a partition P of [a, b] such that
U ( P, f ,a ) - L ( P, f ,a ) < e ………… (i)
2M
If M i & mi denote the sup. & inf. of f on [ xi -1, xi ] then M i2 & mi2 are the
sup. & inf. of f 2 on [ xi -1, xi ] .
( ) (
Þ U P, f 2 ,a - L P, f 2 ,a = å M i2 - mi2 Da i ) ( )
= å ( M i + mi )( M i - mi ) Da i
Q f ( x) £ f ( x) £ M " x Î [ a, b]
2
and f 2 = f
\ M i £ M & mi £ M
( ) ( )
Þ U P, f 2 ,a - L P, f 2 ,a £ å ( M + M ) ( M i - mi ) Da i
= 2 M å ( M i - mi ) Da i
e
= 2 M éëU ( P, f ,a ) - L ( P, f ,a ) ùû < 2 M × =e
2M
Þ f 2 ÎR(a ) ¤
Ø Corollary
If f ÎR(a ) & g ÎR (a ) on [a, b] then fg ÎR(a ) on [a, b] .
Proof
Q f Î R (a ) , g ÎR (a )
\ f + g ÎR(a ) , f - g ÎR (a )
Þ ( f + g )2 ÎR(a ) , ( f - g ) 2 ÎR (a )
Þ ( f + g )2 - ( f - g ) 2 ÎR (a ) Þ 4 fg ÎR (a )
and ultimately
fg ÎR(a ) on [a, b] ¤
Ø Theorem
Assume a increases monotonically and a ¢ ÎR on [a, b] . Let f be bounded
real function on [a, b] . Then f ÎR(a ) iff f a ¢ ÎR . In that case
b b
òa f da = òa f ( x) × a ¢( x) dx
Proof
Q a ¢ Î R on [a, b]
\ given e > 0 $ a partition P of [a, b] such that
U ( P,a ¢ ) - L ( P,a ¢) < e ….……..…. (i)
The Mean-value theorem furnishes point ti Î [ xi -1, xi ] such that
Da i = a ( xi ) - a ( xi -1 )
= a ¢(ti ) Dxi for i = 1,2,...., n …………. (ii)
If si Î [ xi -1, xi ] , then form (i) we have
åa ¢( si ) Dxi - åa ¢(ti ) Dxi < e | Previously proved at page 6
Þ å a ¢(si ) - a ¢(ti ) Dxi < e …………… (iii)
Put M = sup f ( x) and consider
å f ( si ) Dai -å f ( si )a ¢( si ) Dxi …………….. (A)
= å f ( si )a ¢(ti ) Dxi -å f ( si )a ¢( si ) Dxi by (ii)
= å f ( si ) (a ¢(ti ) - a ¢( si ) ) Dxi
£ å M (a ¢(ti ) - a ¢( si ) ) Dxi
£ Me …………..…….. (iv) by (iii)
Þ å f ( si ) Da i £ å f ( si )a ¢( si ) Dxi + M e for all choices of si Î [ xi -1, xi ]
Þ U ( P, f ,a ) £ U ( P, f a ¢) + M e
The same arguments leads from (A) to
U ( P, f a ¢ ) £ U ( P, f ,a ) + M e
Thus U ( P, f ,a ) - U ( P, f a ¢) £ M e ………..…. (v)
Q (i) remains true if P is replaced by any refinement
\ (v) also remains true
b b
Þ ò f da - ò f ( x)a ¢( x) dx
a a
£ Me
Q e was arbitrary
b b
\ òa f da = òa f ( x)a ¢( x) dx for any bounded f .
Using the same argument, we can prove from (iv ) by considering the infimum of
f ( x) that
b b
ò f da = ò f ( x)a ¢( x) dx
a a
Hence
b b b b
ò f da = ò f da
a a
Û ò f ( x)a ¢( x) dx = ò f ( x)a ¢( x) dx
a a
Equivalently f ÎR(a ) Û f a ¢ ÎR(a ) . ¤

Ø Theorem (Change of Variable)

Suppose j is a strictly increasing continuous function that maps an interval
[ A, B] onto [a, b] . Suppose a is monotonically increasing on [a, b] and f ÎR(a )
on [a, b] . Define b and g on [ A, B ] by
b ( y ) = a ( j ( y ) ) , g ( y ) = f (j ( y ) )
B b
then g ÎR ( b ) and òA g d b = òa f da .
Proof
b
f a(f(y))= b(y)
y x=f(y)
f f(f(y))= g(y)
[a,b]
[A,B]
R
g
To each partition P = { x0 ,....., xn } of [a, b] corresponds a partition

Q = { y0 ,....., yn } of [ A, B ] because j maps [ A, B ] onto [a, b] .
Þ xi = j ( yi )
All partitions of [ A, B ] are obtained in this way.
Q The value taken by f on [ xi -1, xi ] are exactly the same as those taken by g
on [ yi -1 , yi ] , we see that
U ( Q, g , b ) = U ( P, f ,a )
and L ( Q, g , b ) = L ( P, f ,a )
Q f Î R(a ) on [a, b]
\ given e > 0 , we have
U ( P, f ,a ) - L ( P, f ,a ) < e
Þ U ( Q, g , b ) - L ( Q, g , b ) < e
B b
Þ g ÎR( b ) and òA g d b = òa f da ¤
{{{{{{{{{{{{{{{{{{{{{{{{{{{{{{{{{{{{
INTEGRATION AND DIFFERENTIATION

Ø Theorem (Ist Fundamental Theorem of Calculus)
x
Let f ÎR on [a, b] . For a £ x £ b , put F ( x) = ò f (t ) dt , then F is continuous
a
on [a, b] ; furthermore, if f is continuous at point x0 of [a, b] , then F is
differentiable at x0 , and F ¢( x0 ) = f ( x0 ) .
Proof
Q f ÎR
\ f is bounded.
Let f (t ) £ M for t Î [a, b]
If a £ x < y £ b , then a x y b
y x
F ( y) - F ( x) = ò f (t ) dt - ò f (t ) dt
a a
x y x
= ò f (t ) dt + ò f (t ) dt - ò f (t ) dt
a x a
y y y
= ò f (t ) dt
x
£ òx f (t ) dt £ M ò dt = M ( y - x)
x
Þ F ( y ) - F ( x) < e for e > 0 provided M y - x < e

e
i.e. F ( y ) - F ( x ) < e whenever y - x <
M
This proves the continuity (and, in fact, uniform continuity) of F on [a, b] .
Next, we have to prove that if f is continuous at x0 Î [a , b] then F is
differentiable at x0 and F ¢( x0 ) = f ( x0 )
F (t ) - F ( x0 )
i.e. lim = f ( x0 )
t ® x0 t - x0
Suppose f is continuous at x0 . Given e > 0 , $ d > 0 such that
f (t ) - f ( x0 ) < e if t - x0 < d where t Î [a, b]
Þ f ( x0 ) - e < f (t ) < f ( x0 ) + e if x0 - d < t < x0 + d
t t t t
Þ ò ( f ( x ) - e ) dt
x0
0 < ò f (t ) dt
x0
< ò ( f ( x ) + e ) dt
x0
0
a x0 – d x0 x0+d
t t t
Þ ( f ( x0 ) - e ) ò dt < ò f (t ) dt < ( f ( x0 ) + e ) ò dt
x0 x0 x0
Þ ( f ( x0 ) - e ) (t - x0 )
< F (t ) - F ( x0 ) < ( f ( x0 ) + e ) (t - x0 )
F (t ) - F ( x0 )
Þ f ( x0 ) - e < < f ( x0 ) + e
t - x0
F (t ) - F ( x0 )
Þ - f ( x0 ) < e
t - x0
F (t ) - F ( x0 )
Þ lim = f ( x0 )
t ® x0 t - x0
Þ F ¢( x0 ) = f ( x0 ) ¤
˜ --------------------------------- ™
Ø Theorem (IInd Fundamental Theorem of Calculus)

If f ÎR on [a, b] and if there is a differentiable function F on [a, b] such that
F ¢ = f , then
b
òa f ( x) dx = F (b) - F (a)
Proof
Q f Î R on [a, b]
\ given e > 0 , $ a partition P of [a, b] such that
U ( P, f ) - L ( P, f ) < e
Q F is differentiable on [a, b]
\ $ ti Î [ xi-1 , xi ] such that
F ( xi ) - F ( xi -1 ) = F ¢(ti )Dxi
Þ F ( xi ) - F ( xi -1 ) = f (ti ) Dxi for i = 1,2,...., n Q F¢ = f
n
Þ å f (t ) Dx
i =1
i i = F (b) - F (a )
Q if f Î R (a ) then
å f (t ) Da - ò
b b
f da < e
Þ F (b) - F (a ) - ò f ( x) dx < e i i
a
a
Q e is arbitrary
b
\ òa f ( x) dx = F (b) - F (a) ¤
Ø Theorem (Integration by Parts)

Suppose F and G are differentiable function on [a, b] , F ¢ = f ÎR and
G¢ = g ÎR then
b b
òa F ( x) g ( x) dx = F (b) G (b) - F (a) G(a) - òa f ( x) G ( x) dx

Proof
Put H ( x) = F ( x) G ( x )
Þ H ¢ = F ¢( x) G ( x) + F ( x ) G¢( x) = h
Now Q H Î R and h ÎR on [a, b]
\ By applying the fundamental theorem of calculus to H and its derivative h ,
we have
b
òa h dx = H (b) - H (a)
b
Þ òa [ F ¢( x) G ( x) + F ( x) G¢( x)] dx = H (b) - H (a)
b b
Þ òa f ( x) G ( x) dx + òa F ( x) g ( x) dx = F (b) G (b) - F (a) G (a)
b b
Þ òa F ( x) g ( x) dx = F (b) G(b) - F (a) G (a) - òa f ( x) G ( x) dx ¤
˜ ---------------------------------- ™
Made by: Atiq ur Rehman (atiq@mathcity.tk)
Available online at http://www.mathcity.tk in PDF Format.
Page Setup: Legal ( 8¢¢ ´ 14¢¢ )
1
2
Printed: 15 April 2004 (Revised: June 08, 2004.)

Ø Question
Show that the function f defined on [ 0,1] by
ì1 ; x is rational
f ( x) = í
î0 ; x is irrational
is not integrable on [ 0,1]
Solution
For any partition P of [ 0,1] , mk = 0 , M k = 1
n n
Þ S ( P, f ) = å M k Dxk = å Dxk = 1 - 0 = 1
k =1 k =1
n
and L ( P, f ) = å mk Dxk = 0
k =1
1 1
so that ò f dx = 1 , ò f dx = 0
0 0
1 1
i.e. ò f dx ¹ ò f dx Þ f is not integrable on [ 0,1] . ¤
0 0
Ø Question
é pù
Show that f ( x ) = sin x is Riemann integrable over ê0, ú .
ë 2û
Solution
ì p p 3p np ü é pù
Take P = í0, , , ,....., ý by dividing ê0, ú into n equal parts.
î 2n n 2n 2n þ ë 2û
kp (k - 1)p
Then M k = sin , mk = sin
2n 2n
æ kp (k - 1)p ö p
Þ S ( P, f ) - L ( P, f ) = å ç sin - sin ÷
è 2n 2n ø 2n
p p
£ <e for n > n0 =
2n 2e
é pù
Þ f is Riemann integrable over ê0, ú . ¤
ë 2û
Ø Question
ìï 1 ; x is rational , 0 < x £ 1
Show that f ( x) = í x
ïî 0 ; x is irrational
is integrable on [ 0,1] .
Solution
f is continuous at each irrational. And rational numbers are dense in [ 0,1] .
1
Also L ( P, f ) = 0 for any partition P of [ 0,1] so that ò f dx = 0
0
1
Q f ³ 0 \ S ( P, f ) ³ 0 Þ ò f da ³ 0 ……..…. (i)
0
p æ pö q e
Q There are only finite number of points (rationals) for which f ç ÷ = ³
q èqø p 2
e
\ Suppose f ( x ) ³ for k values of x in [ 0,1]
2
e
Take P1 such that P1 < .
2k
n
Consider S ( P1 , f ) = å M i ( xi - xi -1 )
i =1
e e
There are at most k values for which £ M i £ 1 . For all other values M i > .
2 2
Þ S ( P1, f ) = å
k values
M i ( xi - xi -1 ) + å M i ( xi - xi -1 )
other values
e e e e
£ × k + å ( xi - xi -1 ) < + = e
2k 2 2 2
Q e is arbitrary
1
\ S ( P1 , f ) £ 0 and ò f dx £ 0 ………… (ii)

0
By (i) and (ii), we have
1
ò f dx = 0
0
1
Hence ò0 f dx = 0 ¤
Ø Note
If f is integrable then f is also integrable but the converse is false.
For example, let f be a function defined on [a, b] by
ì 1 ; x Î ¤ Ç [ a, b]
f ( x) = í
î -1 ; otherwise
Then f is Riemann-integrable but f is not.
References: (1) Lectures (Year 2003-04)

Prof. Syyed Gul Shah
(2) Book
Mathematical Analysis
Tom M. Apostol (John Wiley & Sons, Inc.)
Made by: Atiq ur Rehman (atiq@mathcity.org)
Page Setup: Legal ( 8¢¢ 1 2 ´ 14¢¢ )
Printed: 15 April 2004 (Revised: March 19, 2006.)
Submit error or mistake at http://www.mathcity.org/error
Chapter 7 – Functions of Bounded Variation.
We shall now discuss the concept of functions of bounded variation which is

closely associated to the concept of monotonic functions and has wide application in
mathematics. These functions are used in Riemann-Stieltjes integrals and Fourier
series.
Let a function f be defined on an interval [a, b] and P = {a = x0 , x1 ,........, xn = b}
n
be a partition of [a, b] . Consider the sum åi =1
f ( xi ) - f ( xi-1 ) . The set of these sums
is infinite. It changes when we make a refinement in a partition. If this set of sums is

bounded above then the function f is said to be a bounded variation and the
supremum of the set is called the total variation of the function f on [a, b] , and is
denoted by V ( f ; a, b ) or V f ( a, b ) and it is also affiliated as V ( f ) or V f .
Thus
n
V ( f ; a, b ) = sup å f ( xi ) - f ( xi -1 )
i =1
The supremum being taken over all the partition of [a, b] .
Hence the function f is said to be of bounded variation on [a, b] if, and only, if
its total variation is finite i.e. V ( f ; a, b ) < ¥ .
Ð Note
Since for x £ c £ y , we have
f ( y ) - f ( x ) £ f ( y ) - f (c ) + f (c ) - f ( x )
Therefore the sum å f ( xi ) - f ( xi-1 ) can not be decrease (it can, in fact only
increase) by the refinement of the partition.
Ð Theorem
A bounded monotonic function is a function of bounded variation.
Proof
Suppose a function f is monotonically increasing on [a, b] and P is any partition
of [a, b] then
n n
å
i =1
f ( xi ) - f ( xi-1 ) = å ( f ( xi ) - f ( xi -1 ) ) = f (b) - f (a )
i =1
\ V ( f ; a, b ) = sup å f ( xi ) - f ( xi -1 ) = f (b) - f (a ) (finite)

Hence the function f is of bounded variation on [a, b] .
Similarly a monotonically decreasing bounded function is of bounded variation
with total variation = f (a ) - f (b ) .
Thus for a bounded monotonic function f
V ( f ) = f (b) - f (a) q
{{{{{{{{{{{{{{{{{{{{{{{{{{{
2 Chap 7 – Functions of bounded variation.
Ð Example
A continuous function may not be a function of bounded variation.
e.g. Consider a function f , where
ì p
ï x sin ; when 0 < x £ 1
f ( x) í x
ïî 0 ; when x = 0
It is clear that f is continuous on [0,1] .

ì 2 2 2 2 ü
Let us choose the partition P = í0, , ,........, , ,1ý
î 2n + 1 2n - 1 5 3 þ
Then
å f ( xi ) - f ( xi-1 ) = f (1) - f æçè 3 ö÷ø + f æçè 3 ö÷ø - f æçè 5 ö÷ø + ....... + f æçè 2n + 1 ö÷ø - f ( 0 )
2 2 2 2
2 æ 3p ö 2 æ 3p ö 2 æ 5p ö
= sin p - sin ç ÷ + sin ç ÷ - sin ç ÷ + ..........
3 è 2 ø 3 è 2 ø 5 è 2 ø
2 æ (2n + 1)p ö
..... + sin ç ÷-0
2n + 1 è 2 ø
2 æ2 2ö æ2 2ö æ 2 2 ö 2
= + ç + ÷ + ç + ÷ + ............ + ç + ÷+
3 è3 5ø è5 7ø è 2n - 1 2n + 1 ø 2 n + 1
æ æ2ö æ2ö æ2ö æ 2 öö
= ç 2 ç ÷ + 2 ç ÷ + 2 ç ÷ + ............. + 2 ç ÷÷
è è3ø è5ø è7ø è 2n + 1 ø ø
æ1 1 1 1 ö
= 4 ç + + + ............. + ÷
è3 5 7 2n + 1 ø
1 1 1
Since the infinite series + + + ........... is divergent, therefore its partial sums
3 5 7
1 1 1 1
sequence {Sn } , where S n = + + + ........ + , is not bounded above.
3 5 7 2n + 1
Thus å f ( xi ) - f ( xi-1 ) can be made arbitrarily large by taking n sufficiently
large.
Þ V ( f ;0,1) ® ¥ and so f is not of bounded variation. q
Ð Remarks
A function of bounded variation is not necessarily continuous.
e.g. the step-function f ( x ) = [ x] , where [ x] denotes the greatest integer not
greater than x, is a function of bounded variation on [0, 2] but is not continuous.
Ð Theorem
If the derivative of the function f exists and is bounded on [a, b] , then f is of
bounded variation on [a, b] .
Proof
Q f ¢ is bounded on [a, b]
\ $ k such that f ¢( x ) £ k " x Î [ a, b] .
Let P be any partition of the interval [a, b] then
å f ( xi ) - f ( xi-1 ) = å xi - xi -1 f ¢(c) , c Î [ a, b] (by M.V.T)
£ k b-a
Þ V ( f ; a, b ) is finite. Þ f is of bounded variation. q
Chap 7 – Functions of bounded variation. 3
Note
Boundedness of f ¢ is a sufficient condition for V ( f ) to be finite and is not
necessary.
Ð Theorem
A function of bounded variation is necessarily bounded.
Proof
Suppose f is of bounded variation on [a, b] .
For any x Î [a, b] , consider the partition {a, x, b} , consisting of just three points
then
f ( x ) - f (a) + f (b) - f ( x ) £ V ( f ; a, b)
Þ f ( x ) - f ( a ) £ V ( f ; a, b)
Again
f ( x ) = f (a) + f ( x ) - f (a)
£ f (a ) + f ( x ) - f ( a )
£ f (a ) + V ( f ; a , b ) < ¥
Þ f is bounded on [a, b] . q
Ð Properties of functions of bounded variation

1) The sum (difference) of two functions of bounded variation is also of bounded
variation.
Proof
Let f and g be two functions of bounded variation on [a, b] . Then for any
partition P of [a, b] we have
å ( f + g )(x ) - ( f + g )(x
i i -1 ) =å { f ( xi ) + g ( xi )} - { f ( xi -1 ) + g ( xi -1 )}
= å f ( xi ) - f ( xi -1 ) + g ( xi ) - g ( xi-1 )
£ å f ( xi ) - f ( xi-1 ) + å g ( xi ) - g ( xi-1 )
£ V ( f ; a, b) + V ( g ; a, b)
Þ V ( f + g ; a, b ) £ V ( f ; a, b ) + V ( g ; a, b)
This show that the function f + g is of bounded variation.
Similarly it can be shown that f - g is also of bounded variation.
i.e. V ( f - g ) £ V ( f ) + V ( g ) q
Note
(i) If f and g are monotonic increasing on [a, b] then ( f - g ) is of bounded
variation on [a, b] .
(ii) If c is constant, the sums å f ( xi ) - f ( xi -1 ) and therefore the total variation
function, V ( f ) is same for f and f - c .
2) The product of two functions of bounded variation is also of bounded

variation.
Proof
Let f and g be two functions of bounded variation on [a, b] .
Þ f and g are bounded and $ a number k such that
f ( x ) £ k & g ( x) £ k " x Î [ a, b] .
For any partition P of [a, b] we have
å ( fg ) ( x ) - ( fg ) ( x )
i i -1
= å f (x )g (x ) - f (x )g (x )
i i i -1 i -1
= å f (x )g(x ) - f (x )g (x ) + f (x )g(x ) - f (x )g (x
i i i i -1 i i -1 i -1 i -1 )
= å f ( x ){ g ( x ) - g ( x )} + g ( x ){ f ( x ) - f ( x )}
i i i -1 i -1 i i -1
£ å f (x ) g (x ) - g (x ) + å g (x ) f (x ) - f (x )
i i i -1 i -1 i i -1
£ k å g(x ) - g(x ) + k å f (x ) - f (x )
i i -1 i i -1
£ k V (g ) + k V ( f )
Þ fg is of bounded variation on [a, b] . q
Ð Note
Theorems like the above, could not be applied to quotients of functions because the
reciprocal of a function of bounded variation need not be of bounded variation.
1
e.g. if f ( x ) ® 0 as x ® x0 , then will not be bounded and therefore can not
f ( x)
be of bounded variation on any interval which contains x0 .
Therefore to consider quotient, we avoid functions whose values becomes
arbitrarily close to zero.
3) If f is a function of bounded variation on [a, b] and if $ a positive number k

1
such that f ( x ) ³ k " x Î [a, b] then is also of bounded variation on [a, b] .
f
Proof
For any partition P of [a, b] , we have
1 1 1 1
å f
( xi ) - ( xi-1 )
f
=å -
f ( xi ) f ( xi-1 )
f ( xi -1 ) - f ( xi )
=å
f ( xi ) f ( xi-1 )
1 1
£
k2
å f ( xi-1 ) - f ( xi ) £
k2
V ( f ; a, b )
1
Þ is of bounded variation on [a, b] . q
f
4) If f is of bounded variation on [a, b] , then it is also of bounded variation on

[a, c ] and [c, b] , where c is a point of [a, b] , and conversely. Also
V ( f ; a , b ) = V ( f ; a , c ) + V ( f ; c, b ) .
Proof
a) Let, first, f be of bounded variation on [a, b] .
Take P1 = {a = x0 , x1 ,....., xm = c} & P2 = {c = y0 , y1 ,....., yn = b} any two
partitions of [a, c ] and [c, b] respectively.
Evidently, P = P1 È P2 = {a = x0 ,...., xm , y0 ,...., yn = b} is a partition of [a, b] .
We have
ìm n
ü
íå f ( xi ) - f ( xi -1 ) + å f ( yi ) - f ( yi-1 ) ý £ V ( f ; a, b )
î i =1 i =1 þ
m
Þ åi =1
f ( xi ) - f ( xi -1 ) £ V ( f ; a, b )
n
and åi =1
f ( yi ) - f ( yi -1 ) £ V ( f ; a, b )
Þ f is of bounded variation on [a, c ] and [c, b] both.

b) Let, now, f be of bounded variation on [a, c ] and [c, b] both.
Let P = {a = z0 , z1 ,....., zn = b} be a partition of [a, b] .
If it does not contain the point c , let us consider the partition P* = P È {c}
Let c Î [ zr -1 , zr ] i.e. zr -1 £ c £ zr , r < n c
a z1 zr – 1 zr
Then
n r -1 n
å
i =1
f ( zi ) - f ( zi -1 ) = å f ( zi ) - f ( zi -1 ) + f ( zr ) - f ( zr -1 ) +
i =1
å
i = r +1
f ( zi ) - f ( zi-1 )
r -1
£ å f ( zi ) - f ( zi -1 ) + f (c) - f ( zr -1 )
i =1
n
+ f ( zr ) - f ( c ) + å
i = r +1
f ( zi ) - f ( zi -1 )
£ V ( f ; a, c ) + V ( f ; c, b )
Þ f is of bounded variation on [a, b] if it is of bounded variation on [ a, c ] &
[c, b] both, then
V ( f ; a, b ) £ V ( f ; a, c ) + V ( f ; c, b ) …………. (i)
Now let e > 0 be any arbitrary number.
Since V ( f ; a, c ) and V ( f ; c, b ) are the total variation of f on [ a, c ] & [c, b]
respectively therefore $ partition P1 = {a = x0 , x1 , x2 ,....., xm = c} and
P2 = {c = y0 , y1 , y3 ,......, yn = b} of [ a, c ] & [c, b] respectively such that
m
e
i =1
å f ( xi ) - f ( xi-1 ) > V ( f ; a, c ) -
2
……………. (ii)
n
e
& åi =1
f ( yi ) - f ( yi -1 ) > V ( f ; c, b ) - …………… (iii)
2
Adding (ii) and (iii) we get
m n
åi =1
f ( xi ) - f ( xi -1 ) + å f ( yi ) - f ( yi-1 ) > V ( f ; a, c ) + V ( f ; c, b ) - e
i =1
Þ V ( f ; a, b ) > V ( f ; a, c ) + V ( f ; c, b ) - e
But e is arbitrary positive number therefore we get
V ( f ; a, b ) ³ V ( f ; a, c ) + V ( f ; c, b ) …………….. (iv)
From (i) and (iv), we get
V ( f ; a, b ) = V ( f ; a, c ) + V ( f ; c , b ) q
||||||||||||||||||||||||||||
Ð Variation Function
Let f be a function of bounded variation on [a, b] and x is a point of [a, b] .
Then the total variation of f is V ( f ; a, x ) on [a, x] , which is clearly a function of x,
is called the total variation function or simply the variation function of f and is
denoted by V f ( x) , and when there is no scope for confusion, it is simply written as
V ( x) .
Thus V f ( x) = V ( f ; a, x ) ; (a £ x £ b)
If x1 , x2 are two points of the interval [a, b] such that x2 > x1 , then
0 £ f ( x2 ) - f ( x1 ) £ V ( f ; x1 , x2 )
= V ( f ; a, x1 ) - V ( f ; a, x2 )
= V f ( x2 ) - V f ( x1 )
Þ V f ( x2 ) ³ V f ( x1 )
implies that the variation function is monotonically increasing function on [a, b] .
CHARACTERIZATION OF FUNCTIONS OF BOUNDED VARIATION
Ð Theorem
A function of bounded variation is expressible as the difference of two
monotonically increasing function.
Proof
We have
1 1
f ( x) = (V ( x) + f ( x) ) - (V ( x) - f ( x) )
2 2
= G ( x) - H ( x ) (say)
We shall prove that these two functions G ( x) and H ( x ) are monotonically
increasing on [a, b] .
Now, if x2 > x1 , we have
1
G ( x2 ) - G ( x1 ) = [V ( x2 ) - V ( x1 ) + f ( x2 ) - f ( x1 )]
2
1
= éëV ( f ; x1 , x2 ) - ( f ( x1 ) - f ( x2 ) ) ùû
2
Since V ( f ; x1 , x2 ) ³ f ( x1 ) - f ( x2 )
Þ G ( x2 ) - G ( x1 ) ³ 0 i.e. G ( x2 ) ³ G ( x1 )
so that the function G ( x) is monotonically increasing on [a, b] .
Again, we have
1
H ( x2 ) - H ( x1 ) = éë(V ( x2 ) - V ( x1 ) ) - ( f ( x2 ) - f ( x1 ) ) ùû
2
1
= éëV ( f ; x1 , x2 ) - ( f ( x2 ) - f ( x1 ) ) ùû
2
so that as before
H ( x2 ) - H ( x1 ) ³ 0 i.e. H ( x2 ) ³ H ( x1 ) .
i.e. H ( x ) is also monotonically increasing function.
Hence the result. q
Ð Note
A function f ( x) is of bounded variation over the interval [a, b] iff it can be
expressed as the difference of two monotonically functions.
Ð Theorem
Let f be of bounded variation on [a, b] . Let V be defined on [a, b] as follows:
V ( x ) = V f ( x) = V ( f ; a, x) ) if a < x £ b , V (a ) = 0 .
Then
i) V is an increasing function on [a, b] .
ii) (V - f ) is an increasing function on [a, b] .
Proof
If a < x < y £ b , we can write
V ( f ; a, y ) = V ( f ; a, x) - V ( f ; x, y ) a x y b
Þ V ( y ) - V ( x ) = V ( f ; x, y )
Q V ( f ; x, y ) ³ 0
\ V ( y ) - V ( x ) ³ 0 Þ V ( x) £ V ( y ) and (i) holds.
To prove (ii), let D ( x) = V ( x) - f ( x) if x Î [a, b] .
Then, if a £ x < y £ b , we have
D( y ) - D( x) = [V ( y ) - V ( x) ] - [ f ( y ) - f ( x )]
= V ( f ; x, y ) - [ f ( y ) - f ( x) ]
But from the definition of V ( f ; x, y ) , it follows that
f ( y ) - f ( x ) £ V ( f ; x, y )
This means that D ( y ) - D ( x) ³ 0 and (ii) holds. q
Ð Theorem
If c be any point of [a, b] , then V ( x ) is continuous at c if and only if
f ( x) is continuous at c .
i.e. A point of continuity of f ( x) is also a point of continuity of V ( x ) and
conversely.
Proof
Firstly suppose that V ( x ) is continuous at c .
Let e > 0 be given, then $ d > 0 such that
V ( x ) - V (c ) < e for x - c < d ………… (i)
Also, we have
f ( x ) - f (c ) £ V ( x ) - V (c ) if x > c …………… (ii)
And
f ( x ) - f (c ) £ V (c ) - V ( x ) if x < c …………… (iii)
From (i), (ii) and (iii), we deduce that
f ( x ) - f (c ) £ V ( x ) - V (c ) < e for x - c < d
Which shows that f ( x) is continuous at c .
Now suppose that c is a point of continuity of f ( x) and let e > 0 be given,
then $ d > 0 such that
e
f ( x ) - f (c ) < for x - c < d
2
Also $ a partition P = {c = y0 , y1 ,....., yq -1 , yq ,....., yn = b} of [c, b] such that
n
1
å
q =1
f ( yq ) - f ( yq -1 ) > V ( f ; c, b) - e …………… (iv)
2
Since as a result of introducing addition points to the partition P , the
corresponding sum of the moduli of the differences of the function values at end
points will not be decreased, therefore we may assume that
0 < y1 - c < d
e
so that f ( y1 ) - f (c) < ……………. (v)
2
Thus (iv) becomes
n
1 1 1
V ( f ; c, b) - e < e + å f ( yq ) - f ( yq -1 ) < e + V ( f ; y1 , b)
2 2 q =2 2
Þ V ( f ; c, b ) - V ( f ; y1 , b) < e
Þ V ( y1 ) - V (c) < e
Thus for 0 < y1 - c < d , we have 0 < V ( y1 ) - V (c) < e
\ lim V ( x ) = V (c)
x ®c + 0
Similarly, we can have
lim V ( x) = V (c )
x ®c -0
Which shows that V ( x ) is continuous at c . q
Ð Note
V ( x ) is continuous in [a, b] iff f ( x) is continuous in [a, b] .
Ð Corollary
A function f is of bounded variation on [a, b] iff there is a bounded increasing
function g on [a, b] such that for any two points x¢ and x¢¢ in [a, b] , x¢ < x¢¢ ,
we have
f ( x¢¢) - f ( x¢) £ g ( x¢¢) - g ( x¢)
Moreover, if g is continuous at x¢ , so is f .
Proof
ìV x , a < x £ b
Take g ( x) = í a
î 0 , x=a
Then g is increasing and bounded on [a, b] .
Also, f ( x¢) - f ( x¢¢) £ Vxx¢ ¢¢ ( f ) = g ( x¢¢) - g ( x¢)
Which also yields that if g is continuous at x¢ , so is f . q
Ð Question
Show that the function f defined by
ì 2 1
ï x sin ; x¹0
f ( x) = í x
ïî 0 ; x=0
is of bounded variation on [ 0,1] .

Solution
1
f is differentiable on [ 0,1] and f ¢( x ) = 2 x sin - sin x for 0 £ x £ 1 .
x
Also
1
f ¢( x ) £ 2 x sin + sin x £ 2 + 1 = 3
x
i.e. f ¢( x ) is bounded on [ 0,1]
Hence f is of bounded variation on [ 0,1] . q
Ð Question
ì px
ï x cos , 0 < x £1
Show that g ( x) = í 2 is not of bounded variation on [ 0,1]
ïî 0 , x=0
Solution
ì 1 1 1 ü
,....., , ,1ý be a partition of [ 0,1] .
1
Let P = í0, ,
î 2n 2n - 1 3 2 þ
Then
å f ( xi ) - f ( xi-1)
æ1ö æ1ö æ1ö æ1ö æ1ö æ 1 ö
= f (1) - f ç ÷ + f ç ÷ - f ç ÷ + f ç ÷ - f ç ÷ + ...... + f ç ÷ - f ( 0 )
è2ø è2ø è3ø è3ø è4ø è 2n ø
p 1 p 1 p 1 p 1 p 1 p 1 p
= cos - cos + cos - cos + cos - cos + ..... + cos - 0
2 2 4 2 4 3 6 3 6 4 8 2n 4n
æ1 pö æ1 p ö æ1 pö æ 1 p ö
= 2 ç cos ÷ + 2 ç cos ÷ + 2 ç cos ÷ + ..... + 2 ç cos ÷
è2 4ø è3 6ø è4 8ø è 2n 4n ø
æ1 p 1 p 1 p 1 p ö
= 2 ç cos + cos + cos + ...... + cos ÷
è2 4 3 6 4 8 2n 4n ø
which is not bounded.
Hence f ( x) is not of bounded variation on [ 0,1] . q
Alternative
We have
g ( xk +1 ) - g ( xk ) + g ( xk ) - g ( xk -1 )
1 (k + 1)p 1 kp 1 kp 1 (k - 1) p
= cos - cos + cos - cos
k +1 2 k 2 k 2 k -1 2
ì 2 ; if k is even
ïï k
=í
ï 1 + 1 ; if k is odd
îï k + 1 k - 1
n ¥
1 1
Þ Vab ( g ) £ å
k =1 k
£ åk
k =1
¥
1
Q åk
k =1
is divergent \ Vab ( g ) is not finite.
Hence g is not of bounded variation. q
References: (1) Lectures & Notes (Year 2003-04)

Page Setup: Legal ( 8¢¢ 1 2 ´ 14¢¢ )
Chapter 8 – Improper Integrals.
We discussed Riemann-Stieltjes’s integrals of the form ò a

b
f da under the
restrictions that both f and a are defined and bounded on a finite interval [a, b] .
To extend the concept, we shall relax these restrictions on f and a .
Ø Definition
b
The integral ò a
f da is called an improper integral of first kind if a = -¥ or
b = + ¥ or both i.e. one or both integration limits is infinite.
Ø Definition
b
The integral ò a
f da is called an improper integral of second kind if f ( x) is
unbounded at one or more points of a £ x £ b . Such points are called singularities
of f ( x) .
Ø Notations
We shall denote the set of all functions f such that f ÎR(a ) on [a, b] by
R(a ; a, b) . When a ( x ) = x , we shall simply write R(a, b) for this set. The notation
a - on [a, ¥) will mean that a is monotonically increasing on [a, ¥) .
Ø Definition
Assume that f ÎR(a ; a, b) for every b ³ a . Keep a,a and f fixed and define
a function I on [a, ¥) as follows:
b
I (b) = ò f ( x ) da ( x) if b ³ a ………… (i)
a
The function I so defined is called an infinite ( or an improper ) integral of first
¥ ¥
kind and is denoted by the symbol ò a
f ( x) da ( x ) or by ò a
f da .
¥
The integral ò a
f da is said to converge if the limit
lim I (b) ………… (ii)
b®¥
¥
exists (finite). Otherwise, ò a
f da is said to diverge.
If the limit in (ii) exists and equals A , the number A is called the value of the
¥
integral and we write ò a
f da = A
Ø Example
b
Consider ò 1
x - p dx .
b
(1 - b ) 1- p ¥
òx òx
-p -p
dx = if p ¹ 1 , the integral dx diverges if p < 1 . When
1
p -1 1
1
p > 1 , it converges and has the value .
p -1
b ¥
If p = 1 , we get ò 1
x -1 dx = log b ® ¥ as b ® ¥ . Þ ò 1
x -1 dx diverges.
2 Chap. 8 – Improper Integrals.
Ø Example
b
Consider ò sin 2p x dx
0
(1 - cos 2p b)
b
Q ò sin 2p x dx = ® ¥ as b ® ¥ .
0
2p
¥
\ the integral ò sin 2p x dx diverges.
0
Ø Note
a ¥
If ò
-¥
f da and ò f da
a
are both convergent for some value of a , we say that
¥
the integral ò
-¥
f da is convergent and its value is defined to be the sum
¥ a ¥
ò
-¥
f da = ò
-¥
f da + ò f da
a
The choice of the point a is clearly immaterial.

¥ b
If the integral ò
-¥
f da converges, its value is equal to the limit: lim
b®+ ¥ ò f da .
-b
Ø Theorem
Assume that a - on [a, + ¥) and suppose that f ÎR(a ; a, b) for every b ³ a .
¥
Assume that f ( x ) ³ 0 for each x ³ a . Then ò a
f da converges if, and only if,
there exists a constant M > 0 such that
b
ò f da
a
£ M for every b ³ a .
Proof
b
We have I (b) = ò f ( x ) da ( x) , b³a
a
Þ I - on [a, + ¥)
Then lim I (b) = sup {I (b) | b ³ a} = M > 0 and the theorem follows
b®+¥
b
Þ ò f da £ M
a
for every b ³ a whenever the integral converges.
]]]]]]]]]]]]]]]]
Chap. 8 – Improper Integrals. 3
Ø Theorem: (Comparison Test)

Assume that a - on [a, + ¥) . If f ÎR(a ; a, b) for every b ³ a , if
¥ ¥
0 £ f ( x ) £ g ( x) for every x ³ a , and if ò g da
a
converges, then ò f da
a
converges
and we have
¥ ¥
ò f da
a
£ ò g da
a
Proof
b b
Let I1 (b) = ò f da and I 2 (b) = ò g da , b³a
a a
Q 0 £ f ( x) £ g ( x ) for every x ³ a
\ I1 (b) £ I 2 (b) …………………. (i)
¥
Q ò g da converges \ $ a constant M > 0 such that
a
¥
ò g da £ M
a
, b ³ a …………………(ii)
From (i) and (ii) we have I1 (b) £ M , b ³ a.

Þ lim I1 (b) exists and is finite.
b®¥
¥
Þ ò f da
a
converges.
Also lim I1 (b) £ lim I 2 (b) £ M

b®¥ b®¥
¥ ¥
Þ ò f da £ ò g da .
a a
Ø Theorem (Limit Comparison Test)

Assume that a - on [a, + ¥) . Suppose that f ÎR(a ; a, b) and that
g ÎR (a ; a, b) for every b ³ a , where f ( x ) ³ 0 and g ( x ) ³ 0 if x ³ a . If
f ( x)
lim =1
x ®¥ g ( x )
¥ ¥
then ò f da
a
and ò g da
a
both converge or both diverge.
Proof
For all b ³ a , we can find some N > 0 such that
f ( x)
-1 < e " x ³ N for every e > 0 .
g ( x)
f ( x)
Þ 1-e < <1+ e
g ( x)
1
Let e = , then we have
2
1 f ( x) 3
< <
2 g ( x) 2
Þ g ( x ) < 2 f ( x ) …..…..(i) and 2 f ( x ) < 3 g ( x ) ……....(ii)
¥ ¥
From (i) ò g da
a
< 2 ò f da
a
¥ ¥ ¥ ¥
Þ ò g da
a
converges if ò f da
a
converges and ò f da
a
diverges if ò f da
a
diverges.
¥ ¥
From (ii) 2 ò f da < 3ò g da
a a
¥ ¥ ¥ ¥
Þ ò f da
a
converges if ò g da
a
converges and ò g da
a
diverges if ò f da
a
diverges.
¥ ¥
Þ The integrals ò f da
a
and ò g da
a
converge or diverge together.
Ø Note
f ( x)
The above theorem also holds if lim = c , provided that c ¹ 0 . If c = 0 ,
x ®¥ g ( x )
¥ ¥
we can only conclude that convergence of ò g da
a
implies convergence of ò f da .
a
Ø Example
¥
òe
-x
For every real p , the integral x p dx converges.
1
¥
1
This can be seen by comparison of this integral with òx
1
2
dx .
-x p
f ( x) e x 1
Since lim = lim where f ( x) = e - x x p and g ( x) = 2 .
x ®¥ g ( x ) x®¥ 1 x
x2
f ( x) x p+2
Þ lim = lim e - x x p+ 2 = lim x = 0
x®¥ g ( x ) x ®¥ x ®¥ e
¥
1
and Q òx 2
dx is convergent
1
¥
òe
-x
\ the given integral x p dx is also convergent.
1
Ø Theorem
¥
Assume a - on [a, + ¥) . If f ÎR(a ; a, b) for every b ³ a and if ò
a
f da
¥
converges, then ò f da
a
also converges.
Or: An absolutely convergent integral is convergent.

Proof
If x ³ a , ± f ( x) £ f ( x )
Þ f ( x ) - f ( x) ³ 0
Þ 0 £ f ( x) - f ( x) £ 2 f ( x )
¥
Þ ò(
a
f - f ) da converges.
¥ ¥
Subtracting from ò
a
f da we find that ò f da
a
converges.
( Q Difference of two convergent integrals is convergent )
Ø Note
¥ ¥
ò f da
a
is said to converge absolutely if ò
a
f da converges. It is said to be
¥ ¥
convergent conditionally if ò f da
a
converges but ò
a
f da diverges.
Ø Remark
Every absolutely convergent integral is convergent.
Ø Theorem
Let f be a positive decreasing function defined on [a, + ¥) such that
f ( x ) ® 0 as x ® +¥ . Let a be bounded on [a, + ¥) and assume that
¥
f ÎR(a ; a, b) for every b ³ a . Then the integral ò
a
f da is convergent.
Proof
Integration by parts gives
b b
ò f da = f ( x ) × a ( x) a - ò a ( x ) df
b
a a
b
= f (b) × a (b) - f (a ) × a (a ) + ò a d (- f )
a
It is obvious that f (b)a (b) ® 0 as b ® + ¥
(Q a is bounded and f ( x ) ® 0 as x ® +¥ )
and f (a )a (a ) is finite.
b b
\ the convergence of ò f da
a
depends upon the convergence of ò a d (- f ) .
a
Actually, this integral converges absolutely. To see this, suppose a ( x) £ M for

all x ³ a ( Q a ( x ) is given to be bounded )
b b
Þ
a
ò a ( x) d ( - f ) £ ò M d ( - f )
a
b
But ò M d (- f ) = M - f = M f (a ) - M f (b) ® M f (a ) as b ® ¥ .
b
a
a
¥
Þ ò M d (- f ) is convergent.
a
Q - f is an increasing function.
¥
\ òa
a
d (- f ) is convergent. (Comparison Test)
¥
Þ ò f da
a
is convergent.
]]]]]]]]]]]]]]]]
Ø Theorem (Cauchy condition for infinite integrals)

¥
Assume that f ÎR(a ; a, b) for every b ³ a . Then the integral ò f da
a
converges
if, and only if, for every e > 0 there exists a B > 0 such that c > b > B implies
c
ò f ( x) da ( x)
b
<e
Proof
¥
Let ò f da
a
be convergent. Then $ B > 0 such that
B b c
¥
e
b
ò f da - ò f da
a a
<
2
for every b ³ B ………..(i)
Also for c > b > B ,

¥
e
c
ò f da - ò f da
a a
<
2
…………….. (ii)
Consider
c c b
ò f da
b
= ò f da - ò f da
a a
c ¥ ¥ b
= ò f da - ò f da + ò f da - ò f da
a a a a
¥ ¥
e e
c b
£ ò f da - ò f da
a a
+ ò f da - ò f da
a a
< + =e
2 2
c
Þ ò f da
b
<e when c > b > B .
Conversely, assume that the Cauchy condition holds.

a +n
Define an = ò
a
f da if n = 1,2,......
The sequence {an } is a Cauchy sequence Þ it converges.

Let lim an = A
n®¥
e
c
Given e > 0 , choose B so that ò
b
f da <
2
if c > b > B .
e a+N
and also that an - A < whenever a + n ³ B . a
2 B b c
Choose an integer N such that a + N > B i.e. N > B - a
Then, if b > a + N , we have
b a+ N b
ò f da - A
a
= ò
a
f da - A + ò
a+ N
f da
e e
b
£ aN - A + ò
a+ N
f da < + =e
2 2
¥
Þ ò f da = A
a
This completes the proof.
Ø Remarks
¥
It follows from the above theorem that convergence of ò a
f da implies
b +e
lim ò f da = 0 for every fixed e > 0 .
b®¥ b
However, this does not imply that f ( x ) ® 0 as x ® ¥ .
Ø Theorem
¥
Every convergent infinite integral ò a
f ( x) da ( x) can be written as a convergent
infinite series. In fact, we have
¥ ¥ a +k
ò f ( x) da ( x) = å a
a k =1
k where ak = ò
a + k -1
f ( x) da ( x) ……….. (1)
Proof
{ò }
¥
a +n
Q ò f da converges, the sequence f da also converges.
a
a
a +n n ¥ ¥
But ò f da = å ak . Hence the series åa k converges and equals ò f da .
a k =1 k =1 a
Ø Remarks
It is to be noted that the convergence of the series in (1) does not always imply
k
convergence of the integral. For example, suppose ak = ò sin 2p x dx . Then each
k -1
ak = 0 and åa k converges.
¥
1 - cos 2p b
b
However, the integral ò sin 2p x dx = lim ò sin 2p x dx = lim
0
b®¥
0
b®¥ 2p
diverges.
IMPROPER INTEGRAL OF THE SECOND KIND
Ø Definition
Let f be defined on the half open interval ( a, b] and assume that f ÎR(a ; x, b)
for every x Î ( a, b] . Define a function I on ( a, b] as follows:
b
I ( x ) = ò f da if x Î ( a, b] ……….. (i)
x
The function I so defined is called an improper integral of the second kind and
b b
is denoted by the symbol ò f (t ) da (t )
a+
or
a+
ò f da .
b
The integral ò f da
a+
is said to converge if the limit
lim I ( x ) ……...(ii) exists (finite).

x ®a +
b
Otherwise, ò f da
a+
is said to diverge. If the limit in (ii) exists and equals A , the
b
number A is called the value of the integral and we write
a+
ò f da = A .
Similarly, if f is defined on [a, b) and f ÎR(a ; a, x) " x Î [a, b) then

x
I ( x ) = ò f da if x Î [a, b) is also an improper integral of the second kind and is
a
b-
denoted as òa
f da and is convergent if lim I ( x ) exists (finite).
x ® b-
Ø Example
f ( x) = x - p is defined on (0, b] and f ÎR ( x, b) for every x Î (0, b] .
b
I ( x ) = ò x - p dx if x Î (0, b]
x
b b
òx ò
-p
= dx = lim x - p dx
e ®0
0+ 0 +e
b
x1- p b1- p - e 1- p
= lim = lim , ( p ¹ 1)
e ®0 1 - p e ®0 1- p
e
é finite , p < 1
=ê
ë infinite , p > 1
b
1
When p = 1 , we get ò dx = log b - log e ® ¥ as e ® 0 .
e
x
b
òx
-1
Þ dx also diverges.
0+
Hence the integral converges when p < 1 and diverges when p ³ 1 .
Ø Note
c b-
If the two integrals ò f da
a+
and ò f da
c
both converge, we write
b- c b-
a+
ò f da = ò f da + ò f da
a+ c
The definition can be extended to cover the case of any finite number of sums.
We can also consider mixed combinations such as
b ¥ ¥
ò f da + ò f da
a+ b
which can be written as ò f da .
a+
Ø Example
¥
òe
-x
Consider x p -1 dx , ( p > 0)
0+
This integral must be interpreted as a sum as

¥ 1 ¥
òe òe dx + ò e - x x p -1 dx
-x p -1 -x p -1
x dx = x
0+ 0+ 1
= I1 + I 2 ………..….…… (i)
I 2 , the second integral, converges for every real p as proved earlier.
1 1
To test I1 , put t = Þ dx = - 2 dt
x t
1
e
æ 1 ö
1 1
- 1 1- p - 1 - p -1
Þ I1 = lim ò e- x x p -1 dx = lim òe tt ç - 2 dt ÷ = lim òe tt dt
e ®0
e
e ®0
1 è t ø e ®0
1
e
-1
Take f (t ) = e t t - p -1 and g (t ) = t - p -1
-1 ¥
f (t ) e t × t - p -1
òt
- p -1
Then lim = lim - p -1 = 1 and since dt converges when p > 0
t ®¥ g (t ) t ®¥ t 1
¥
- 1 - p -1
\ òe
1
t t dt converges when p > 0
¥
òe
-x
Thus x p -1 dx converges when p > 0 .
0+
When p > 0 , the value of the sum in (i) is denoted by G( p ) . The function so
defined is called the Gamma function.
Ø Note
The tests developed to check the behaviour of the improper integrals of Ist kind
are applicable to improper integrals of IInd kind after making necessary
modifications.
Ø A Useful Comparison Integral

b
dx
ò ( x - a )n
a
We have, if n ¹ 1 ,
b b
dx 1
ò ( x - a )n
a +e
=
(1 - n)( x - a )n-1 a +e
1 æ 1 1 ö
= ç (b - a ) n-1 - e n-1 ÷
(1 - n) è ø
1
Which tends to or + ¥ according as n < 1 or n > 1 , as e ® 0 .
(1 - n)(b - a) n-1
Again, if n = 1 ,
b
dx
òa+e x - a = log(b - a) - log e ® + ¥ as e ® 0 .
b
dx
Hence the improper integral ò ( x - a )n
a
converges iff n < 1 .
]]]]]]]]]]]]]]]]
Ø Question
Examine the convergence of
1 1 1
dx dx dx
(i) ò 1 (ii) ò0 x 2 (1 + x)2 (iii) òx
( ) (1 - x )
1 1
0 x
3
1+ x 2
0
2 3
Solution
1
dx
(i) ò0 x 13
(1 + x ) 2
Here ‘0’ is the only point of infinite discontinuity of the integrand.

We have
1
f ( x) = 1
x 3 1 + x2 ( )
1
Take g ( x) = 1
x 3
f ( x) 1
Then lim = lim =1
x ®0 g ( x ) x ®0 1 + x 2
1 1
Þ ò
0
f ( x) dx and ò0
g ( x ) dx have identical behaviours.
1 1
dx dx
Q ò0 x 13 converges \ ò0 x 13 also converges.
(1 + x )
2
1
dx
(ii) ò0 x 2 (1 + x)2
Here ‘0’ is the only point of infinite discontinuity of the given integrand.
We have
1
f ( x) = 2
x (1 + x )2
1
Take g ( x) = 2
x
f ( x) 1
Then lim = lim =1
( )
2
x ®0 g ( x ) x ®0
1 + x
1 1
Þ ò 0
f ( x) dx and ò
0
g ( x) dx behave alike.
1
But n = 2 being greater than 1, the integral ò0
g ( x) dx does not converge. Hence
the given integral also does not converge.
1
dx
(iii) òx
(1 - x )
1 1
2 3
0
Here ‘0’ and ‘1’ are the two points of infinite discontinuity of the integrand.
We have
1
f ( x) = 1
x 2 (1 - x ) 3
1
We take any number between 0 and 1, say 1 , and examine the convergence of
2
1
2 1
the improper integrals ò0 f ( x ) dx and ò
1
f ( x ) dx .
2
1
2
1 1
To examine the convergence of òx 1
2 (1 - x)
1
3
dx , we take g ( x) =
x
1
2
0
Then
f ( x) 1
lim = lim =1
x ®0 g ( x) x®0 (1 - x) 13
1 1
2 2
1 1
Q ò0 x 12 dx converges \ ò 1
x (1 - x)
2
1
3
dx is convergent.
0
1
1 1
To examine the convergence of ò
1
1
x 2 (1 - x)
1
3
dx , we take g ( x) =
(1 - x)
1
3
2
Then
f ( x) 1
lim = lim 1 = 1
x ®1 g ( x) x®1 x 2
1 1
1 1
Q ò 1
dx converges Q ò 1 1
dx is convergent.
1 (1 - x ) 3 1 x 2 (1 - x ) 3
2 2
1
Hence ò 0
f ( x ) dx converges.
Ø Question
1
ò x (1 - x ) dx exists iff m , n are both positive.

m -1 n-1
Show that
0
Solution
The integral is proper if m ³ 1 and n ³ 1 .
The number ‘0’ is a point of infinite discontinuity if m < 1 and the number ‘1’ is
a point of infinite discontinuity if n < 1 .
Let m < 1 and n < 1 .
We take any number, say 1 , between 0 & 1 and examine the convergence of
2
1
2 1
ò0 x (1 - x ) ò1 x (1 - x ) dx
m-1 n -1 m-1 n -1
the improper integrals dx and at ‘0’ and ‘1’
2
respectively.
Convergence at 0:
We write
m-1 n -1 (1 - x) n-1 1
f ( x) = x (1 - x ) = and take g ( x) =
x1-m x1- m
f ( x)
Then ® 1 as x ® 0
g ( x)
1
2
1
As ò0 x1-m dx is convergent at 0 iff 1 - m < 1 i.e. m > 0
1
2
ò0 x (1 - x ) dx
m-1 n -1
We deduce that the integral is convergent at 0, iff m is +ive.
Convergence at 1:
m-1 n -1 x m-1 1
We write f ( x ) = x (1 - x ) = and take g ( x) =
(1 - x)1-n (1 - x )1- n
f ( x)
Then ® 1 as x ® 1
g ( x)
1
1
As ò
1 (1 - x )
1-n
dx is convergent, iff 1 - n < 1 i.e. n > 0 .
2
1
ò1 x (1 - x ) dx converges iff
m-1 n -1
We deduce that the integral n > 0.
2
1
ò0 x (1 - x ) dx exists for positive values of m , n only.

m -1 n-1
Thus
It is a function which depends upon m & n and is defined for all positive values
of m & n . It is called Beta function.
Ø Question
Show that the following improper integrals are convergent.
¥ ¥ 1 1
2 1 sin 2 x x log x
(i) ò sin dx (ii) ò 2 dx (iii) ò dx (iv) ò0 log x × log(1 + x) dx
1
x 1
x 0
(1 + x ) 2
Solution
1 1
(i) Let f ( x ) = sin 2 and g ( x) = 2
x x
2
f ( x) sin 2 1x æ sin y ö
then lim = lim 1 = lim ç ÷ =1
x ®¥ g ( x ) x ®¥
2 x
y ®0
è y ø
¥ ¥
1
Þ ò1 f ( x) dx and ò1 x 2 dx behave alike.
¥ ¥
1 1
Q ò 2 dx is convergent \ ò sin 2 dx is also convergent.
1
x 1
x
¥
sin 2 x
(ii) ò 2 dx
1
x
sin 2 x 1
Take f ( x ) = 2
and g ( x) = 2
x x
2
sin x 1
sin 2 x £ 1 Þ £ 2 " x Î (1, ¥ )
x2 x
¥ ¥
1 sin 2 x
and ò 2 dx converges \ ò 2 dx converges.
1
x 1
x
Ø Note
1
sin 2 x sin 2 x
ò0 x2 dx is a proper integral because lim
x ®0 x2
= 1 so that ‘0’ is not a point
¥
sin 2 x
of infinite discontinuity. Therefore ò 2 dx is convergent.
0
x
1
x log x
(iii) ò0 (1 + x)2 dx
Q log x < x , x Î (0,1)
\ x log x < x 2
x log x x2
Þ <
(1 + x ) (1 + x )
2 2
1
x2
Now ò (1 + x ) 2
dx is a proper integral.
0
1
x log x
\ ò (1 + x )
0
2
dx is convergent.
1
(iv) ò0 log x × log(1 + x) dx
Q log x < x \ log( x + 1) < x + 1
Þ log x × log(1 + x) < x ( x + 1)
1 1
Q ò0 x ( x + 1) dx is a proper integral \ ò0 log x × log(1 + x) dx is convergent.
Ø Note
a
1
(i) ò0 x p dx diverges when p ³ 1 and converges when p < 1 .
¥
1
(ii) òa x p dx converges iff p > 1 .
UNIFORM CONVERGENCE OF IMPROPER INTEGRALS
Ø Definition
Let f be a real valued function of two variables x & y , x Î [a, + ¥) , y ÎS
¥
where S Ì ¡ . Suppose further that, for each y in S , the integral ò a
f ( x, y ) da ( x )
is convergent. If F denotes the function defined by the equation
¥
F ( y ) = ò f ( x, y ) da ( x) if y ÎS
a
the integral is said to converge pointwise to F on S
Ø Definiton
¥
Assume that the integral òa
f ( x, y ) da ( x ) converges pointwise to F on S . The
integral is said to converge Uniformly on S if, for every e > 0 there exists a B > 0
(depending only on e ) such that b > B implies
b
F ( y ) - ò f ( x, y ) da ( x) < e " y ÎS .
a
( Pointwise convergence means convergence when y is fixed but uniform
convergence is for every y ÎS ).
Ø Theorem (Cauchy condition for uniform convergence.)

¥
The integral ò
a
f ( x, y ) da ( x ) converges uniformly on S , iff, for every e > 0
there exists a B > 0 (depending on e ) such that c > b > B implies
c
ò f ( x, y ) da ( x)
b
<e " y ÎS .
Proof
¥
Proceed as in the proof for Cauchy condition for infinite integral ò a
f da .
Ø Theorem (Weierstrass M-test)

b
Assume that a - on [a, + ¥) and suppose that the integral òa
f ( x , y ) da ( x )
exists for every b ³ a and for every y in S . If there is a positive function M
¥
defined on [a, + ¥) such that the integral òa
M ( x) da ( x) converges and
f ( x, y ) £ M ( x) for each x ³ a and every y in S , then the integral
¥
òa
f ( x, y ) da ( x ) converges uniformly on S .
Proof
Q f ( x, y ) £ M ( x) for each x ³ a and every y in S .
\ For every c ³ b , we have
c c c
ò f ( x, y) da ( x) £ ò
b b
f ( x, y ) da ( x) £ ò M da ………… (i)
b
¥
Q I = ò M da is convergent
a
\ given e > 0 , $ B > 0 such that b > B implies
b
ò M da - I < e …………… (ii)

2
a
Also if c > b > B , then
c
ò M da - I < e …………… (iii)

2
a
c c b
Then ò M da
b
= ò M da - ò M da
a a
c b
= ò M da - I + I - ò M da
a a
c b
£ ò M da - I + ò M da - I < e +e =e (By ii & iii)
2 2
a a
c
Þ ò f ( x, y) da ( x)
b
< e , c > b > B & for each y ÎS
Cauchy condition for convergence (uniform) being satisfied.

¥
Therefore the integral òa
]]]]]]]]]]]]]]]]
Ø Example
¥
ò0 e
- xy
Consider sin x dx
e - xy sin x £ e - xy = e - xy (Q sin x £ 1 )
and e- xy £ e - xc if c£ y
Now take M ( x ) = e - cx
¥ ¥
1
ò0 M ( x) dx = ò0 e
- cx
The integral dx is convergent & converging to .
c
¥
ò0 e
- xy
\ The conditions of M-test are satisfied and sin x dx converges
uniformly on [c, + ¥) for every c > 0 .
Ø Theorem (Dirichlet’s test for uniform convergence)

b
Assume that a is bounded on [a, + ¥) and suppose the integral òa f ( x, y) da ( x)
exists for every b ³ a and for every y in S . For each fixed y in S , assume that
f ( x, y ) £ f ( x¢, y ) if a £ x¢ < x < + ¥ . Furthermore, suppose there exists a positive
function g , defined on [a, + ¥) , such that g ( x ) ® 0 as x ® +¥ and such that
x ³ a implies
f ( x, y ) £ g ( x) for every y in S .
¥
Then the integral ò a
Proof
Let M > 0 be an upper bound for a on [ a, +¥ ) .
Given e > 0 , choose B > a such that x ³ B implies
e
g ( x) <
4M
e
( Q g ( x ) is +ive and ® 0 as x ® ¥ \ g ( x ) - 0 < for x ³ B )
4M
If c > b , integration by parts yields
c c
òb f da = f ( x, y ) × a ( x) b - ò a df
c
b
c
= f (c, y )a (c) - f (b, y )a (b) + ò a d (- f ) ………… (i)
b
But, since - f is increasing (for each fixed y ), we have
c c
òa d (- f ) £ M ò d (- f ) ( Q upper bound of a is M )
b b
= M f (b, y ) - M f (c, y ) …………… (ii)
Now if c > b > B , we have from (i) and (ii)
c c
ò f da
b
£ f (c, y )a (c) - f (b, y )a (b) + òa d (- f )
b
£ a (c) f (c, y ) + f (b, y ) a (b) + M f (b, y ) - f (c, y )

£ a (c) f (c, y ) + a (b) f (b, y ) + M f (b, y ) + M f (c, y )
£ M g (c) + M g (b) + M g (b) + M g (c)

= 2M [ g (b) + g (c)]
é e e ù
< 2M ê + ú =e
ë 4 M 4 M û
c
Þ ò f da
b
<e for every y in S .
¥
Therefore the Cauchy condition is satisfied and òa f ( x, y) da ( x) converges
uniformly on S .
Ø Example
¥
e - xy
Consider ò0 x
sin x dx
e - xy
Take a ( x) = cos x and f ( x, y ) = if x > 0 , y ³ 0 .
x
1
If S = [0, +¥ ) and g ( x) = on [e , +¥ ) for every e > 0 then
x
i) f ( x, y ) £ f ( x¢, y ) if x¢ £ x and a ( x) is bounded on [e , +¥ ) .
ii) g ( x ) ® 0 as x ® +¥
e - xy 1
iii) f ( x, y ) = £ = g ( x) " y ÎS .
x x
So that the conditions of Dirichlet’s theorem are satisfied.
Hence
¥ - xy ¥ - xy
òe x sin x dx = + òe x d (- cos x) converges uniformly on [e , +¥ ) if e > 0 .

e e
e
sin x sin x
Q lim =1 \ òe
- xy
dx converges being a proper integral.
x ®0 x x
0
¥
dx also converges uniformly on [0, +¥ ) .
sin x
ò0 e
- xy
Þ
x
Ø Remarks
Dirichlet’s test can be applied to test the convergence of the integral of a
product. For this purpose the test can be modified and restated as follows:
Let f ( x) be bounded and monotonic in [ a, +¥ ) and let f ( x) ® 0 , when
X
x ® ¥ . Also let òa f ( x) dx be bounded when X ³ a .
¥
Then òa f ( x)f ( x) dx is convergent.
Ø Example
¥
sin x
Consider
x ò0
dx
sin x
Q ® 1 as x ® 0 .
x
\ 0 is not a point of infinite discontinuity.

¥
sin x
Now consider the improper integral ò1 x
dx .
1
The factor of the integrand is monotonic and ® 0 as x ® ¥ .
x
X
Also ò sin x dx
1
= - cos X + cos(1) £ cos X + cos(1) < 2
X
So that ò1 sin x dx is bounded above for every X ³ 1 .
¥ 1
sin x sin x
Þ ò dx is convergent. Now since ò0 x dx is a proper integral, we see
1
x
¥
sin x
that ò0 x
dx is convergent.
Ø Example
¥
ò0 sin x
2
Consider dx .
1
We write sin x 2 = × 2 x × sin x 2
2x
¥ ¥
1
Now ò sin x 2 dx = ò × 2 x × sin x 2 dx
1 1
2x
1
is monotonic and ® 0 as x ® ¥ .
2x
X
ò 2 x sin x dx = - cos X 2 + cos(1) < 2

2
Also
1
X
ò1 2 x sin x dx is bounded for X ³ 1 .

2
So that
¥ ¥
1
Hence ò × 2 x × sin x 2 dx i.e. ò1 sin x
2
dx is convergent.
1
2x
1
ò0 sin x
2
Since dx is only a proper integral, we see that the given integral is
convergent.
Ø Example
¥
sin x
Consider ò e - a x dx , a > 0
0
x
¥
sin x
ò0 x dx is convergent.
-a x
Here e is monotonic and bounded and
¥
sin x
ò0 e
-a x
Hence dx is convergent.
x
]]]]]]]]]]]]]]]]
Ø Example
¥
sin x
Show that ò0 x
dx is not absolutely convergent.
Solution
np We need not
sin x
Consider the proper integral ò dx take x
x
0 because x ³ 0 .
where n is a positive integer. We have
np n rp
sin x sin x
ò x dx = å =
ò x dx
0 r 1 ( r -1)p
Put x = (r - 1)p + y so that y varies in [0,p ] .

We have sin[(r - 1)p + y ] = (-1)r -1 sin y = sin y
rp p
sin x sin y
\ ò x
dx = ò
( r - 1)p + y
dy
( r -1)p 0
Q rp is the max. value of [(r - 1)p + y ] in [0,p ]

p p
sin y 1 2 é Division by max. value
\ ò dy ³ ò sin y dy = êë Q
0
(r - 1)p + y rp 0 rp will lessen the value
np n
sin x 2 2 n 1
Þ ò0 dx ³ å = å
x 1 rp p 1 r
n
1
Q år ®¥
1
as n ® ¥ , we see that
np
sin x
ò0 x
dx ® ¥ as n ® ¥ .
Let, now, X be any real number.

There exists a +tive integer n such that np £ X < (n + 1)p .
X np
sin x sin x
We have ò0 x
dx ³ ò0 x
dx
X
sin x
Let X ® ¥ so that n also ® ¥ . Then we see that ò0 x
dx ® ¥
¥
sin x
So that ò0 x
dx does not converge.
Ø Questions
Examine the convergence of
¥ ¥ ¥
x 1 dx
(i) ò
(1 + x)3
dx (ii) ò1 (1 + x) x dx (iii) òx
(1 + x )
1 1
3 2
1 1
Solution
x x 1
(i) Let f ( x) = and take g ( x) = 3 = 2
(1 + x) 3
x x
3
f ( x) x
As lim = lim =1
x ®¥ g ( x ) x®¥ (1 + x ) 3
¥ ¥
x 1
Therefore the two integrals ò dx and ò 2 dx have identical behaviour
1
(1 + x) 3
1
x
for convergence at ¥ .
¥ ¥
1 x
Q ò 2 dx is convergent \ ò dx is convergent.
1
x 1
(1 + x ) 3
1 1 1
(ii) Let f ( x ) = and take g ( x) = = 3
(1 + x) x x x x 2
f ( x) x
We have lim = lim =1
x ®¥ g ( x ) x ®¥ 1 + x
¥ ¥
1 1
and ò1 x 3 2 dx is convergent. Thus ò1 (1 + x) x
dx is convergent.
1
(iii) Let f ( x ) =
(1 + x )
1 1
x 3 2
1 1
we take g ( x) = 1 1
= 5
x 3 ×x 2 x 6
¥ ¥
f ( x) 1
We have lim = 1 and ò 5 dx is convergent \ ò1 f ( x) dx is convergent.
x ®¥ g ( x ) 6
1 x
Ø Question
¥
1
Show that
-¥
ò 1+ x 2
dx is convergent.
Solution
We have
¥
1 é0 1 a
1 ù
ò 1 + x2 dx = alim
-¥
®¥ ò 1 + x 2
ê
êë - a
dx + ò
0
dx ú
1 + x 2 úû
éa 1 a
1 ù éa 1 ù
= lim ê ò dx + ò dx ú = 2 lim ê ò dx ú
ë0 1+ x 1+ x ë 01+ x
a ®¥ 2 2 a®¥ 2
0 û û
a æp ö
= 2 lim tan -1 x = 2 ç ÷ = p
a ®¥ 0
è2ø
therefore the integral is convergent.
Ø Question
¥
tan -1 x
Show that ò0 1 + x 2 dx is convergent.
Solution
tan -1 x p -1
Q (1 + x 2 ) × = tan -1
x ® as x ® ¥ Here f ( x ) =
tan x
(1 + x 2 ) 2 1+ x
2
¥ ¥
tan -1 x 1 g (x) = 1 + x
2
ò0 1 + x 2 dx ò0 1 + x 2 dx behave alike.
and
&
¥
1
Q ò0 1 + x 2 dx is convergent \ A given integral is convergent.
Ø Question
¥
sin x
Show that ò0 (1 + x)a dx converges for a > 0 .
Solution
¥ X
ò0 sin x dx is bounded because ò0 sin x dx £ 2 " x >0.
, a > 0 is monotonic on [0, +¥ ) .

1
Furthermore the function
(1 + x)a
¥
sin x
Þ the integral ò0 (1 + x)a dx is convergent.
Ø Question
¥
ò0 e
-x
Show that cos x dx is absolutely convergent.
Solution
¥
Q e cos x < e ò0 e
-x -x -x
and dx = 1
\ the given integral is absolutely convergent. (comparison test)
Ø Question
e- x
1
Show that ò0 1- x 4
dx is convergent.
Solution
Q e - x < 1 and 1 + x 2 > 1
e- x 1 1
\ < <
1 - x4 (1 - x 2 )(1 + x 2 ) 1 - x2
1 1-e
1 1
Also ò0 1 - x2
dx = lim
e ®0 ò
0 1 - x2
dx
p
= lim sin -1 (1 - e ) =
e ®0 2
e- x
1
Þ ò0 1- x 4
dx is convergent. (by comparison test)
References: (1) Lectures (Year 2003-04)

(2) Book
Mathematical Analysis
Tom M. Apostol (John Wiley & Sons, Inc.)
Page Setup: Legal ( 8¢¢ 1 2 ´ 14¢¢ )

Real Analysis Notes by Prof Syed Gul Shah

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Real Analysis Notes by Prof Syed Gul Shah

Uploaded by

Copyright:

Available Formats

Chapter 1 – Real Number System

Subject: Real Analysis (Mathematics) Level: M.Sc.

Now if p ∈ B then p 2 > 2 ⇒ p 2 − 2 > 0

iii) Since y < z⇒ −y + y < −y + z

iv) If x > 0 then x ⋅ x > 0 ⇒ x 2 > 0

Í Existence of Real Field

Now suppose y n > x

ii) For λ ∈ ¡ we have

References: (1) Lectures (2003-04)

Collected and Composed by: Atiq ur Rehman (mathcity@gmail.com)

e.g. i) {n} = {1, 2,3,..........}

i.e. condition I of mathematical induction is satisfied.

Suppose (1 + p ) > 1 + kp .................. (i ) where k ≥ 2

(ii) {(−1) } tends to 1 or -1 according as n is even or odd. It oscillates finitely.

(iii) {(−1) n} is a divergent sequence. It oscillates infinitely.

⇒ 0 < lim S n < 0

⇒ S2n < 2 ∀ n = 1,2,3,..........

But we know that S 2 > S1 therefore S3 > S 2 , S 4 > S3 , and so on.

Then lim( S n+1 )2 = lim( S n + b)

Theorem (Cauchy’s General Principle for Convergence)

Theorem (nested intervals)

Again choose n2 > n1 such that

Limit Inferior of the sequence

Limit Superior of the sequence

Sn − s < ε ∀ n ≥ n0 ……….. (i)

sum a1 + a2 + a3 + .............. and called a infinite series or just series.

This implies that if lim an ≠ 0 , then

Sn − Sm < ε ∀ n > m > n0

an < λ bn ∀ n ≥ k …………. (i)

⇒ ∃ a real number β > 0 such that

Theorem ( Cauchy Condensation Test )

therefore the series diverges when p ≤ 0 .

If p ≤ 1 the series is divergent.

⇒ the series ∑a n is convergent. (Cauchy Criterion has been used)

Theorem (The Root Test)

Then ∑a n converges absolutely if p < 1 and it diverges if p > 1 .

Q ∑ ( p + ε ) is convergent because it is a geometric series with

Consider the series ∑ an with an =  −  

⇒ The ∑a n bn is convergent. ( We have use Cauchy Criterion here. )

To see this we proceed as follows

Now ∑ b is divergent for α ≤ 1 therefore ∑ a

The series also divergent if α ≤ 1.

References: (1) Lectures (2003-04)

Made by: Atiq ur Rehman (mathcity@gmail.com)

v Limit of the function

Choose { pn } in E such that pn ≠ p , lim pn = p , we are to show that

We show that lim f ( x) does not exist.

f ( x ) − q < ε whenever x − p < δ .

ε , there is δ > 0 such that f (t ) − f (s ) < ε whenever s & t are in

v Theorem (Sandwiching Theorem)

Consequently, f ( u ) − f ( v ) < ε whenever

References: (1) Lectures (2003-04)

Collected and composed by: Atiq ur Rehman (mathcity@gmail.com)

(ii) Let f ( x) = x n ; n ³ 0 (n is integer), x Î ¡ .

( x - c )( x n-1 + cx n-2 + .......... + c n-2 x + c n-1 )

(ii) If f is discontinuous at c Î D f then f ¢(c) does not exists.

v Geometric Interpretation of M.V.T.

implies that g is monotonically increasing at b .

If a - 1 > 0 , then lim t a -1 sin t - c = 0 Þ f ¢(0) = 0 when a > 0 .

If a - 1 < 0 i.e. when a < 1 we have t a -1 = t - b where b > 0

v Question (L Hospital Rule)

Similarly, a nhood of a radius d of a point ( x0 , y0 , z0 ) of a space is a sphere with

This definition can be extended to the definition of a nhood of a point of a space of

It means that given e > 0 $ a d > 0 such that