Professional Documents
Culture Documents
(Mathematics) Semester-I
MATHEMATICAL ANALYSIS
Structure
1.0 Introduction
1.1 Unit Objectives
1.2 Riemann-Stieltjes integral
1.2.1 Definitions and Notations
Partition P, P* finer than P, Common refinement, Norm (or Mesh)
Lower and Upper Riemann-Stieltjes Sums and Integrals
Riemann-Stieltjes integral
1.3 Existence and properties
1.3.1 Characterization of upper and lower Stieltjes sums and upper and lower Stieltjes integrals
1.3.2 Integrability of continuous and monotonic functions along with properties of Riemann-
Stieltjes integrals
1.3.3 Riemann-Stieltjes integral as limit of sums.
1.4 Integration and Differentiation
1.5 Fundamental Theorem of the Integral Calculus
1.5.1 Theorem on Integration by parts or Partial Integration Formula
1.5.2 Mean Value Theorems for Riemann-Stieltjes Integrals.
First Mean Value Theorem for Riemann-Stieltjes Integral
Second Mean Value Theorem for Riemann-Stieltjes Integral
1.5.3 Change of variables
1.6 Integration of Vector –Valued Functions
1.6.1 Fundamental theorem of integral calculus for vector valued function
1.7 Rectifiable Curves
1.8 References
1.0 Introduction
In this unit, we will deal with the Riemann-Stieltjes integral and study its existence and properties. The
Riemann-Stieltjes integral is a generalization of Riemann integral named after Bernhard Riemann and
Thomas Joannes Stieltjes. The reason for introducing this concept is to get a more unified approach to
the theory of random variables. Fundamental Theorem of the Integral Calculus is discussed later on.
2 The Riemann-Stieltjes Integral
1.1 Unit Objectives
After going through this unit, one will be able to
define Riemann-Stieltjes integral and characterize its properties.
recognize Riemann-Stieltjes integral as a limit of sums.
know about Fundamental Theorem of the Integral Calculus and Mean Value Theorems .
understand the concept of Rectifiable Curves
1.2 Riemann-Stieltjes integral
We have already studied the Riemann integrals in our undergraduate level studies in Mathematics. Now
we consider a more general concept than that of Riemann. This concept is known as Riemann-Stieltjes
integral which involve two functions f and . In what follows, we shall consider only real-valued
functions.
1.2.1 Definitions and Notations
Definition1. Let [ a, b] be a given interval. By a partition (or subdivision) P of [ a, b] , we mean a finite
set of points
P {x0 , x1 ,........, xn }
such that
a x0 x1 x2 ........xn 1 xn b .
Definition 3. The P1 and P2 be two partitions of an interval [ a, b] . Then a partition P* is called their
common refinement of P1 and P2 if P* P1 P2 .
Definition 4. The length of the largest subinterval of a partition P {x0 , x1 ,...., xn } of [ a, b] is called the
Norm (or Mesh) of P. We denote norm of P by P . Thus
Let be monotonically increasing function on [ a, b] . Then is bounded on [ a, b] since (a) and (b)
are finite.
Corresponding to each partition P={x0,x1,…………,xn } of [a, b], we put
i ( xi ) ( xi 1 )
.
The monotonicity of implies that i 0 .
n
U ( P , f , ) M i i ,
i 1
where mi and M i are bounds of f defined above. The sums L( P, f , ) and U ( P, f , ) are respectively
called Lower Stieltjes sum and Upper Stieltjes sum corresponding to the partition P. We further define
b
fd lub L( P, f , )
a
fd glbU ( P, f , ) ,
a
b b
where lub and glb are taken over all possible partitions P of [a, b] . Then fd and fd are
a a
If the lower and upper integrals are equal, then their common value, denoted by fd , is called the
a
Riemann-Stieltjes integral of f with respect to , over [a, b] and in that case we say that f is
integrable with respect to , in the Riemann sense and we write f ( ) .
The functions f and are known as the integrand and the integrator respectively.
In the special case, when ( x) x , the Riemann-Stieltjes integral reduces to Riemann-integral. In such
b b
Clearly, the numerical value of fd depends only on f , , a and b and does not depend on the symbol
x. In fact x is a “dummy variable” and may be replaced by any other convenient symbol.
1.3 Existence and properties
1.3.1 In this section, we shall study characterization of upper and lower Stieltjes sums, and upper and
lower Stieltjes integrals.
The next theorem shows that for increasing function , refinement of the partition increases the lower
sums and decreases the upper sums.
Theorem1. If P is a refinement of P, f is bounded real valued function on a, b and is
monotonically increasing function defined on a, b . Then,
L P , f , LP, f ,
U P , f , U P, f , .
mi g.l.b. of f in x i 1 , xi
w1 g.l.b. of f in xi 1 , x*
w2 g.l.b of f in x* , xi
.
Obviously, mi w1 ; mi w2 .
Then,
L P, f , m11 m2 2 ............ mi 1 i 1 mi i ............ mn n
L P , f , m11 m2 2 ............ mi 1 i 1 w1 x xi 1 w2 xi x
mi 1 i 1 .......... ......... m n n
Thus,
L P , f , L P , f , w1 x x i 1 w 2 x i x m i x i x i 1
w m x x w
1 i
i 1 2
mi xi x
Now, w1 mi 0 ; w2 mi 0
Mathematical Analysis 5
x xi 1 0
xi x 0
L P , f , LP, f , 0
LP , f , LP, f ,
Similarly, U P , f , U P, f , .
If P contains more points, then similar process holds and so the result follows.
Theorem 2. For any two partitions P1 and P2 of a, b , let f be a bounded real valued function defined
on a, b and is monotonically increasing function defined on a, b , then
LP1 , f , U P2 , f , .
Proof. Let P be the common refinement of P1 and P2 , that is, P P1 P2 . Then, using Theorem 1, we
have
L ( P1 , f , ) L ( P, f , ) U ( P, f , ) U ( P2 , f , ) .
Remark 1. If m f x M . Then,
m i mi i M i i M i
n n n n
m i mi i M i i M i
i 1 i 1 i 1 i 1
m b a L P , f , U P , f , M b a .
Theorem 3. If f is bounded real valued function defined on a, b and is monotonic function defined
on a, b . Then,
b b
fd fd
a a
.
Proof . Let Pa, b denotes the set of all partition of a, b . For P1 , P2 Pa, b , we know that
6 The Riemann-Stieltjes Integral
L P1, f , U P2 , f , 1
This holds for each P1 Pa, b , keeping P2 fixed, it follows from 1 that U P2 , f , is an upper
bound of the set {L P1 , f , : P1 P a, b} .
b
But least upper bound of this set is f x d .
a
b
i.e, f x d l.u.b.{L P , f , : P P a, b}
1 1
a
f x d U P , f ,
1 2
a
b
This holds for each P2 Pa, b . So, it follows from 2 that f x d is a lower bound of the set
a
U P2 , f , : P2 Pa, b.
b
But greatest lower bound of this set is fd .
a
b
i.e, fd g.l.b.U P , f , : P P a, b
a
2 2
1, x Q
f ( x)
0, x Q
Then for every partition P of [0,1] , we have
0
n
U ( P , f , ) M i xi
i 1
n
( xi xi 1 ) xn x0 1 0 1
i 1
fd fd .
Theorem 4. Let is monotonically increasing on a, b then f ( ) iff for any 0 , there exists a
partition P of a, b such that
U P, f , LP, f ,
Proof. The condition is necessary:
Let f be integrable on a, b i.e, f on a, b ,
b b b
so that fd fd fd 1
a a a
Let 0 be given.
b
Since fd sup{L P, f , : P is a partition of [a,b]}
a
b
L P1 , f , fd
a
2
b
L P1 , f , fd 2
2 a
Again,
8 The Riemann-Stieltjes Integral
b
Since fd inf{U P, f , : P is a partition of [a,b]}.
a
b b
U P2 , f , fd fd
a
2 a
2
b
U P2 , f , fd 3
a
2
U P, f , U P2 , f , 4
And L P1, f , L P, f , 5
Now, we have
b
U P, f , U P2 , f , fd (by (3))
a
2
LP1 , f , (by (2))
2 2
L P1, f ,
L P, f , (by (5))
U P, f , L P, f ,
or U P, f , L P, f , .
U P, f , L P, f , 6
Since lower integral condition exceed the upper integral.
b b
So, fd fd .
a a
Mathematical Analysis 9
b b
fd fd 0 7
a a
b b
fd fd U P, f , L P, f , 8
a a
b b
0 fd fd
a a
b b
The non – negative number fd fd being less than every positive number must be zero,
a a
b b
i.e, fd fd 0
a a
b b
fd fd .
a a
1.3.2 In this section, we shall discuss integrability of continuous and monotonic functions along with
properties of Riemann-Stieltjes integrals.
Theorem 1. If f is continuous on [ a, b] , then
(i) f ( )
M i mi (i 1, 2,......, n )
Hence
n n
U ( P , f , ) L ( P , f , ) M i i mi i
i 1 i 1
n n
( M i mi ) i i
i 1 i 1
n
[ i ( xi ) ( xi 1 )]
i 1
[ (b ) ( a )]
. ,
which is necessary and sufficient condition for f ( ) .
(ii) We have
n
L ( P , f , ) f ( ti ) i U ( P , f , )
i 1
and
b
L( P, f , ) fd U ( P, f , )
a
Since f ( ) , for each 0 there exists 0 such that for all partition P with P , we have
U ( P , f , ) L ( P , f , )
Thus
n b
n b
Thus for continuous functions f , lim P 0 f (ti ) i exists and is equal to fd .
i 1 a
Proof. Let be a given positive number. For any positive integer n, choose a partition P of [ a, b] such
that
(b) (a)
i (i 1, 2,......, n) .
n
This is possible since is continuous and monotonic on [ a, b] and so assumes every value between its
bounds ( a ) and (b) . It is sufficient to prove the result for monotonically increasing function f , the
proof for monotonically decreasing function being analogous. The bounds of f in [ xi 1 , xi ] are then
mi f ( xi 1 ) , M i f ( xi ) , i 1, 2,......, n .
Hence
n
U ( P, f , ) L ( P, f , ) ( M i mi ) i
i 1
(b ) ( a ) n
n
(M
i 1
i mi )
(b ) ( a ) n
n
[ f (x ) f (x
i 1
i i 1 )]
(b) (a)
[ f (b) f (a)]
n
for large n.
Hence f ( ) .
f ( x* ) 1 and f ( x ) 0 for x x* , a x* b .
b
Suppose is increasing on [ a, b] and is continuous at x * . Then f ( ) over [ a, b] and fd 0 .
a
Solution. Let P {x0 , x1 ,......, xn } be a partition of [ a, b] and let x * xi . Since is continuous at x * ,
to each 0 there exists 0 such that
12 The Riemann-Stieltjes Integral
( x ) ( x* ) whenever x x*
2
Again since is an increasing function,
( x ) ( x* ) for 0 x x*
2
and
( x* ) ( x ) for 0 x x*
2
Then for a partition P of [ a, b] ,
i ( xi ) ( xi 1 )
( xi ) ( x * ) ( x * ) ( xi 1 )
.
2 2
n 0, ti x*
Therefore f (ti ) i
i , ti x
*
i 1
that is,
n
f (t )
i 1
i i 0
Hence
n b
lim P 0 f (ti ) i fd 0
i 1 a
( f
a
1 f 2 )d f1d f 2 d
a a
Proof. Let P {a x0 , x1 ,......, xn b} be any partition of [ a, b] . Suppose further that M i' , mi' , M i" , mi" and
M i , mi are the bounds of f1 , f 2 and f1 f 2 respectively in the subinterval [ xi 1 , xi ] . If 1 , 2 [ xi 1 , xi ] ,
then
U ( P1 , f1 , ) L( P1 , f1 , ) 2
(2)
U ( P , f , ) L( P , f , )
2 2 2 2
2
These inequalities hold if P1 and P2 are replaced by their common refinement P.
n n
( M i' mi' ) i ( M i" mi" ) i
i 1 i 1
(using (2))
2 2
.
Hence f f1 f 2 ( ) .
L ( P, f1 , ) L ( P, f 2 , ) L ( P, f , ) U ( P, f , )
U ( P, f1 , ) U ( P, f1 , ) U ( P, f 2 , ) (3)
Also
b
U ( P, f1 , ) f1d (4)
a
2
b
U ( P, f 2 , ) f 2 d (5)
a
2
14 The Riemann-Stieltjes Integral
Combining (3), (4) and (5), we have
b
fd U ( P, f , ) U ( P, f , ) U ( P, f , )
a
1 2
b b
f1d f 2 d
a a
2 2
a
fd f1d f 2 d
a a
(6)
or
b b b
a
fd f1d f 2 d
a a
(7)
fd ( f
a a
1 f 2 )d f1d f 2 d .
a a
a
fd ( ) fd fd .
a a
Proof. Since f ( ) and f ( ) , there exist partitions P1 and P2 such that
U ( P1 , f , ) L( P1 , f , )
2
U ( P2 , f , ) L( P2 , f , )
2
These inequalities hold if P1 and P2 are replaced by their common refinement P.
Also
( i i ) [ ( xi ) ( xi 1 )] [ ( xi ) ( xi 1 )]
n n
( M i mi ) i ( M i mi ) i
i 1 i 1
.
2 2
Hence f ( ) .
Further
b
U ( P, f , ) fd
a
2
b
U ( P, f , ) fd
a
2
and
U (P, f , ) Mi i Mi i
Also, then
b
fd ( ) U ( P, f , ) U ( P, f , ) U ( P, f , )
a
b b
fd fd
a
2 a 2
b b
fd fd
a a
a
fd ( ) fd fd .
a a
a
fd ( ) fd fd .
a a
a
fd fd fd .
a c
16 The Riemann-Stieltjes Integral
Proof. Since f ( ) , there exists a partition P such that
U ( P , f , ) L ( P , f , ) , 0 .
L ( P, f , ) L ( P* , f , ) U ( P* , f , ) U ( P, f , )
which yields
U ( P* , f , ) L ( P*, f , ) U ( P, f , ) L ( P, f , ) (1)
Let P1 and P2 denote the sets of points of P* between [ a, c ] , [c, b] respectively. Then P1 and P2 are
partitions of [ a, c ] and [c, b] and P* P1 P2 . Also
U ( P * , f , ) U ( P1 , f , ) U ( P2 , f , ) (2)
and
L ( P * , f , ) L ( P1 , f , ) L ( P2 , f , ) (3)
Since each of U ( P1 , f , ) L ( P1 , f , ) and U ( P2 , f , ) L ( P2 , f , ) is non-negative, it follows that
U ( P1 , f , ) L ( P1 , f , )
and
U ( P2 , f , ) L ( P2 , f , )
a
fd fd fd
a c
(5)
fd
a
K [ (b) (a)] .
n
c M i i
i 1
cU ( P , f , )
Similarly
L ( P , g , ) cL ( P , f , )
U ( P , f , ) L ( P, f , )
c
Hence
U ( P , g , ) L ( P , g , ) c[U ( P, f , ) L ( P , f , )]
c. .
c
Hence g cf ( ) .
b
Further, since U ( P, f , ) fd ,
a
2
18 The Riemann-Stieltjes Integral
b
gd U ( P, g , ) cU ( P, f , )
a
b
c fd
a 2
Since is arbitrary
b b
gd c fd
a a
Replacing f by f , we get
b b
gd c fd
a a
b b
b
fd
a
U ( P, f , ) M i i
M [ (b ) ( a )]
which yields
b
m[ (b) (a)] fd M [(b) (a)] (2)
a
K f ( x) K
M [ (b ) ( a )] K [ (b ) ( a )]
Hence
b
fd
a
K [ (b) (a)] .
(s) (t ) if s t , s, t [ m, M ] .
U ( P , f , ) L ( P, f , ) 2 (1).
Let M i , mi and M i* , mi* be the lub, glb of f ( x ) and ( x) respectively in [ xi 1 , xi ] . Divide the number
1,2,……,n into two classes:
i A if M i mi
and
i B if M i mi .
For i A , our choice of implies that M i* mi* . Also, for i B , M i* mi* 2 k where k lub (t ) ,
t [ m, M ] . Hence, using (1), we have
i (M i mi ) i 2 (2)
iB iB
so that
iB
i . Then we have
U ( P, h, ) L( P, h, ) ( M i* mi* ) i ( M i* mi* ) i
i A iB
[ (b ) ( a )] 2 k
[[ (b ) ( a )] 2k ]
20 The Riemann-Stieltjes Integral
Since was arbitrary,
U ( P, h, ) L( P, h, ) * , * 0 .
Hence h f ( ) .
a
fd f d .
a
c fd 0
Then
because cf f .
1.3.3. Riemann-Stieltjes integral as limit of sums. In this section, we shall show that Riemann-
Stieltjes integral fd can be considered as the limit of a sequence of sums in which M , m i i involved in
n
S ( P , f , ) f (t i ) i
i 1
S ( P, f , ) A .
Proof. Suppose lim P 0 S ( P, f , ) exists and is equal to A. Then given 0 there exists a 0 such
that P implies
S ( P, f , ) A
2
or
A S ( P, f , ) A (1)
2 2
If we choose partition P satisfying P and if we allow the points ti to range over [ xi 1 , xi ] , taking lub
and glb of the numbers S ( P , f , ) obtained in this way, the relation (1) gives
A L ( P, f , ) U ( P, f , ) A
2 2
and so
U ( P, f , ) L ( P, f , )
2 2
Hence f ( ) . Further
A L( P, f , ) fd U ( P, f , ) A
2 2
which yields
A fd A
2 2
or
fd A lim P 0
S ( P, f , ) .
22 The Riemann-Stieltjes Integral
Theorem 2. If
(i) f is continuous, then
b
lim P 0 S ( P, f , ) fd
a
Proof. Part (i) is already proved in Theorem 1(ii) of section 1.3.2 of this unit.
(ii) Let f ( ) , be continuous and 0 . Then there exists a partition P* such that
U ( P* , f , ) fd (1)
4
Now, being uniformly continuous, there exists 1 0 such that for any partition P of [ a, b] with
P 1 , we have
i ( xi ) ( xi 1 ) for all i
4Mn
where n is the number of intervals into which P* divides [ a, b] . Consider the sum U ( P, f , ) . Those
intervals of P which contains a point of P* in their interior contribute no more than:
(n 1) M
(n 1) max i .M to U(P*,f, ) (2)
4Mn 4
Then (1) and (2) yield
U ( P, f , ) fd (3)
2
L( P, f , ) fd (4)
2
Taking min{1 , 2 } , it follows that (2) and (3) hold for every P such that P .
Since
L( P, f , ) S ( P, f , ) U ( P, f , )
Mathematical Analysis 23
(3) and (4) yield
S ( P, f , ) fd
2
and
S ( P, f , ) > fd
2
Hence
S ( P, f , ) fd
2
for all P such that P and so
lim P 0 S ( P, f , ) fd
An a0 a1 ........ an ( A1 0) ,
then
q q 1
1.4 Integration and Differentiation. In this section, we show that integration and differentiation are
inverse operations.
Definition 1. If f on [a, b] , then the function F defined by
t
F (t ) f ( x)dx , t [ a, b]
a
Proof. We have
t
F (t ) f ( x)dx
a
Since f , it is bounded and therefore there exists a number M such that for all x in [a, b] ,
f ( x) M .
24 The Riemann-Stieltjes Integral
Let be any positive number and c be any point of [a, b] . Then
c ch
F (c) f ( x)dx , F (c h) f ( x)dx
a a
Therefore
ch c
F ( c h ) F (c )
a
f ( x)dx f ( x)dx
a
ch
c
f ( x)dx
M h
if h .
M
Thus (c h) c implies F (c h) F (c) . Hence F is continuous at any point c [ a, b]
M
and is so continuous in the interval [a, b] .
F ( x0 ) f ( x0 ) , x [ a, b] .
Proof. Let f be continuous at x0 in [a, b] . Then there exists 0 for every 0 such that
f (t ) f ( x0 ) (1)
t t
1 1
ts s
[ f ( x) f ( x0 )]dx
t s s
[ f ( x) f ( x0 )]dx ,
(using (1)).
Hence F ( x0 ) f ( x0 ) . This completes the proof of the theorem.
Definition 2. A derivable function F such that F ' is equal to a given function f in [a, b] is called
Primitive of f .
Mathematical Analysis 25
Thus the above theorem asserts that “Every continuous function f possesses a Primitive, viz the integral
t
function f ( x)dx .”
a
Furthermore, the continuity of a function is not necessary for the existence of primitive. In other words,
the function possessing primitive is not necessary continuous. For example, consider the function f on
[0,1] defined by
1 1
2 x sin cos , x 0
f ( x) x x
0, x 0
It has primitive
2 1
x sin , x 0
F ( x) x
0, x 0
Proof. Let P be a partition of [a, b] and choose ti , (i 1, 2,......, n) such that xi 1 ti xi . Then, by
Lagrange’s Mean Value Theorem, we have
F ( xi ) F ( xi 1 ) ( xi xi 1 ) F (ti ) ( xi xi 1 ) f (ti ) (Since F f ).
Further
n
F (b) F (a ) [ F ( xi ) F ( xi 1 )]
i 1
n
f (ti )( xi xi 1 )
i 1
n
f (ti ) xi
i 1
b
and the last sum tends to f ( x)dx as
a
P 0 , by theorem 1 of section 1.3.3, taking ( x ) x . Hence
26 The Riemann-Stieltjes Integral
b
integral.
Theorem 2. If f and on [a, b] , then f ( ) and
b b
Proof. Since f , , it follows that their product f . Let 0 be given. Choose M such
that f M . Since f and , using Theorem 2(ii) of section 1.3.3 for integrator as x, we
have
f (t ) (t )x f
i i i (1).
if P 1 and xi 1 ti xi and
(t )x
i i (2).
(s )x
i i (3).
2
or
(t ) (s ) x
i i i 2 (4).
if P 2 and xi 1 ti xi , xi 1 si xi .
i ( xi ) ( xi 1 ) ( si )( xi xi 1 )
( si )xi
Then, we have
2 M (1 2 M )
Hence
b
lim P 0 f (ti )xi f ( x) ( x)dx
a
or
b b
2 2
x dx [ x]dx
2 2 2
Example 1. Evaluate (i) , (ii) .
0 0
Therefore
2 2 2
x dx x (2 x)dx 2 x dx
2 2 2 3
0 0 0
2
x4
2 8 .
4 0
and
2 2
0 0
1 2
[ x]2 xdx [ x]2 xdx
0 1
2 2
x2
0 2 xdx 0 2
1
2 1
28 The Riemann-Stieltjes Integral
=0+3 = 3.
We now establish a connection between the integrand and the integrator in a Riemann-Stieltjes integral.
We shall show that existence of fd implies the existence of df .
We recall that Abel’s transformation (Partial Summation Formula) for sequences reads as follows:
“Let an and bn be two sequences and let An a0 a1 ...... an ( A1 0) . Then
q q 1
since ti 1 xi 1 ti . If P 0 , Q 0 , then
Hence
Proof. If (a ) (b) , the theorem holds trivially, both sides being 0 in that case ( become constant
and so d 0 ). Hence we assume that ( a ) (b) . Let
M lub f ( x ) , m glb f ( x ) . a x b
Mathematical Analysis 29
Then
m f ( x) M
or
The use of First Mean Value Theorem for Riemann-Stieltjes Integral yields that there is x in [a, b] such
that
b
f ( x)dx f ( ( y))d ( y)
a
1 (a ) , 1 (b) .
Let P {a x0 , x1 ,......, xn b} be any partition of [a, b] and Q { y0 , y1 ,......, yn } be the
corresponding partition of [ , ] , where yi 1 ( xi ) . Then
xi xi xi 1
( yi ) ( yi 1 )
i .
f ( (di ))i
i
g (di )i
i
S (Q , g , )
b
Continuity of f implies that S ( P, f ) f ( x)dx as P 0 and continuity of g implies that
a
S (Q, g , ) g ( y )d as Q 0 .
a
f ( x)dx g ( y )d f ( ( y ))d ( y )
Let be monotonically increasing function on [a, b] . If fi ( ) for i 1, 2,......, k , we say that
f ( ) and then the integral of f is defined as
b b b b
fd ( f d , f d ,........, f d ) .
a a
1
a
2
a
k
b b
fd f d .
k
Thus is the point in R whose ith coordinate is i
a a
a
fd (1 2 ) fd1 fd 2
a a
To prove these results, we have to apply earlier results to each coordinate of f . Also, fundamental
theorem of integral calculus holds for vector valued function f . We have
Theorem 2. If f maps [a, b] into R k and if f R ( ) for some monotonically increasing function on
[a, b] , then f R( ) and
b b
a
fd f d .
a
Proof. Let
f ( f1 ,......, f k ) .
Then
f ( f12 ...... f k2 )1/ 2
32 The Riemann-Stieltjes Integral
Since each fi R( ) , the function f i 2 R ( ) and so their sum f12 ...... f k2 R ( ) . Since x 2 is a
continuous function of x, the square root function is continuous on [0, M ] for every real M. Therefore
f R( ) .
y fd
and
y yi2 yi fi d
2
( yi fi )d
y f (t )
i i y f (t ) , (a t b)
Then
2
y y f d (1)
y f d
b b
or
a
fd f d .
a
1.7 Rectifiable Curves. The aim of this section is to consider application of results studied in this
chapter to geometry.
Definition 1. A continuous mapping of an interval [a, b] into R k is called a curve in R k .
If :[a, b] R k is continuous and one-to-one, then it is called an arc.
If for a curve :[a, b] R k ,
( a ) (b )
but
(t1 ) (t2 )
for every other pair of distinct points t1 , t2 in [a, b] , then the curve is called a simple closed curve.
The ith term ( xi ) ( xi 1 ) in this sum is the distance in R k between the points ( xi 1 ) and ( xi ) .
n
Further (x ) (x
i 1
i i 1 ) is the length of a polygon whose vertices are at the points
( x0 ), ( x1 ),...., ( xn ) . As the norm of our partition tends to zero, then those polygons approach the
range of more and more closely.
Theorem 1. Let be a curve in R k . If is continuous on [a, b] , then is rectifiable and has length
b
(t ) dt .
a
(x ) (x
i 1
i i 1 ) (t )dt
i 1 xi 1
n xi
(t ) dt
i 1 xi 1
b
(t ) dt
a
Thus
V ( , a, b) (1). .
To prove the reverse inequality, let be a positive number. Since is uniformly continuous on [a, b] ,
there exists 0 such that
( s) (t ) , if s t .
(t ) ( xi ) ,
34 The Riemann-Stieltjes Integral
so that
xi
xi 1
(t ) dt xi ( xi ) xi
xi
xi xi
xi 1
(t )dt [ ( x ) (t )]dt
xi 1
i
( xi ) ( xi 1 ) xi
Adding these inequalities for i 1, 2,...., n , we get
b n
(t ) dt ( xi ) ( xi 1 ) 2 (b a)
a i 1
V ( , a, b) 2 (b a )
Since is arbitrary, it follows that
b
(t ) dt V ( , a, b)
a
(2).
(t ) dt V ( , a, b)
a
1.8 References
1. Walter Rudin, Principles of Mathematical Analysis (3rd edition) McGraw-Hill, Kogakusha, 1976,
International Student Edition.
2. T. M. Apostol, Mathematical Analysis, Narosa Publishing House, New Delhi, 1974.
3. H.L. Royden, Real Analysis, Macmillan Pub. Co., Inc. 4th Edition, New York, 1993.
4. G. De Barra, Measure Theory and Integration, Wiley Eastern Limited, 1981.
5. R.R. Goldberg, Methods of Real Analysis, Oxford & IBH Pub. Co. Pvt. Ltd, 1976.
6. R. G. Bartle, The Elements of Real Analysis, Wiley International Edition, 2011.
7. S.C. Malik and Savita Arora, Mathematical Analysis, New Age International Limited, New Delhi,
2012.
U N IT – II
SEQUENCE AND SERIES OF FUNCTIONS
Structure
2.0 Introduction
2.1 Unit Objectives
2.2 Sequence and Series of Functions
2.3 Pointwise and Uniform Convergence
2.4 Cauchy Criterion for Uniform Convergence
2.5 Test for Uniform Convergence
2.6 Uniform Convergence and Continuity
2.7 Uniform Convergence and Integration
2.8 Uniform Convergence and Differentiation
2.9 Weierstrass’s Approximation Theorem
2.10 References
2.0 Introduction
In this unit, we will consider sequence and series of functions whose terms depend on a variable.
Uniform convergence of sequence or series is a concept of great importance in its domain. With the help
of tests for uniform convergence, we will naturally inquire how we can determine whether the given
sequence or series does or does not converge uniformly in a given interval. The Weierstrass
approximation theorem describes that every continuous function can be “uniformly approximated” by
polynomials to within any degree of accuracy.
2.1 Unit Objectives
After going through this unit, one will be able to
learn about pointwise and uniform convergence of sequence and series of functions
examine uniform convergence through various tests for uniform convergence.
study uniform convergence and continuity.
understand importance of Weierstrass approximation theorem.
2.2 Sequence and Series of Functions
Let fn be a real valued function defined on an interval I (or on a subset D of R ) and for each n N , then
<f1, f2,……… ,fn,……..> is called a sequence of real valued functions on I. It is denoted by {fn} or <fn>.
36 Sequence & Series of Functions
If <fn> is a sequence of real valued functions on an interval I, then f1+ f2+……… +fn+……..is called a
series of real valued functions defined on I. This series is denoted by f
n 1
n or simply f n . That is, we
shall consider sequences whose terms are functions rather than real numbers. These sequences are useful
in obtaining approximations to a given function.
2.3 Pointwise and Uniform Convergence of Sequences of Functions
We shall study two different notations of convergence for a sequence of functions: Pointwise
convergence and uniform convergence.
Definition 1. Let A R and suppose that for each n N there is a function f n : A R . Then f n is
called a sequence of functions on A. For each x A , this sequence gives rise to a sequence of real
numbers, namely the sequence fn x .
When such a function f exists, we say that the sequence f n is convergent on A 0 or that f n
converges pointwise on A 0 to f and we write f x limn fn x , x A 0 .
The question arises: If each function of a sequence fn has certain property, such as continuity,
differentiability or integrability, then to what extent is this property transferred to the limit function? For
example, if each function f n is continuous at a point x 0 , is the limit function f also continuous at x 0 ? In
general, it is not true. Thus, pointwise convergence is not so strong concept which transfers above
mentioned property to the limit function. Therefore some stronger methods of convergence are needed.
One of these methods is the notion of uniform convergence:
Let
x 2n
fn x , x R, n = 1, 2, ………
1 x 2n
We note that
0 if | x | 1
1
limn f n x f x if | x | 1.
2
1 if | x | 1
m
Smn .
mn
limm Smn 1
and so
lim n lim m Smn 1
On the other hand, for every fixed m, we have
1
lim n Smn lim n 0
n
1
m
and so
f n x n 2 x 1 x , x R, n 1, 2,......
n
If 0 x 1, then
f x limn fn x 0
and so
1
f (x) dx 0.
0
1 1
f x 1 x
n
But n (x) dx n 2
dx
0 0
n2 n2
n 1 n 2
n2
(n 1) n 2
.
1
and so lim n f n (x) dx 1
0
1 1
Hence lim n f n (x) dx (lim n f n (x)) dx .
0 0
Example 4. A sequence of differentiable functions { f n } with limit 0 for which { f n ' } diverges.
Let
sin nx
fn x if x R, n = 1, 2,
n
If each term of the sequence f n is real-valued, then the expression (*) can be written as
This shows that the entire graph of f n lies between a “band” of height 2ε situated symmetrically about
the graph of f.
Theorem 1. Every uniformly convergent sequence is pointwise convergent but not conversely.
Proof. Let { f n } be a sequence of functions which converges uniformly to f on E.
For given 0 , there exists a positive integer N (depending only on ) such that
f n x f x for all n N
1
f n x ,0 x 1
nx 1
1
Then, f ( x) lim f n x lim 0
n n nx 1
Hence, f n converges pointwise to 0 for all 0 x 1 .
f n x f x , n n0
1
or 0 , n n0 .
nx 1
1
or .
nx 1
1
or
nx
1
or nx .
1
or n .
x
1
If n 0 is taken as integer greater than , then
x
f n x f x for all n n0 .
Since n 0 depends both on & x in (0,1) , so f n does not converge uniformly on (0,1).
1
Example 6. Consider the sequence Sn defined by Sn x in any interval [a, b], a > 0. Then
xn
1
S x lim Sn x lim 0
n n x n
Sn x S x , n n0
(1)
1
or 0 , n n0
xn
Mathematical Analysis 41
1
or
xn
1
or xn
1
or n x
1 1
If we select n 0 as integer next higher to , then (1) is satisfied for m(integer) greater than which
does not depend on x [a, b] . Hence the sequence Sn is uniformly convergent to S(x) in [a, b].
x
fn x , x0
1 nx
Then
x
f x lim 0 for all x 0 .
n 1 nx
Then f n converges pointwise to 0 for all x 0 . Let 0 , then for convergence we must have
f n x f x , n n 0
x
or 0 , n n 0
1 nx
x
1 nx
x nx
nx x
x
n
x
x 1
n
x
1
If n 0 is taken as integer greater than , then
fn x xn , 0 x 1
Then
0 if 0 x 1
f n x lim x n
n
1 if x 1
f n x f x , n n 0
or xn
n
1 1
or
x
1
log
or n .
1
log
x
1
log
Thus we should take n 0 to be an integer next higher to . If we take x = 1, then m does not exist.
1
log
x
Thus the sequence in question is not uniformly convergent to f in the interval which contains 1.
Definition 5 (Point of non–uniform convergence). A point which is such as the sequence is non –
uniformly convergent in any interval containing that point is called a point of non–uniform convergence.
In the following example x = 0 is a point of non–uniform convergence.
nx
Example 9. Consider the sequence f n defined by f n x , 0xa.
1 n2 x2
Then if x = 0, then fn x 0
and so f x limn fn x 0.
If x 0 , then
nx
f x limn f n x limn 0.
1 n2x2
Thus f is continuous at x = 0. For convergence, we must have
f n x f x , n n 0 .
Mathematical Analysis 43
nx
or .
1 n2x2
nx
or 1 n2x2 0.
1 1 1
or nx 4.
2 2 2
Thus we can find an upper bound for n in any interval 0 a x b , but the upper bound is infinite if the
interval includes 0. Hence the given sequence is non-uniformly convergent in any interval which
includes the origin. So 0 is the point of non-uniform convergence for this sequence.
Example 10. Consider the sequence f n defined by
f n x tan 1 nx, 0 x a.
Then
if x 0
f x lim f n x 2 .
n
0 if x 0
f n x f x , n n 0
or tan 1 nx
2
or cot 1 nx
1
or nx
tan
1 1
or n
tan x
Thus no upper bound can be found for the function on the right if 0 is an end point of the interval. Hence
the convergence is non-uniform in any interval which includes 0. So, here 0 is the point of non-uniform
convergence.
M > 0 such that f n x M for all x in E and all n. The number M is called a uniform bound for fn .
44 Sequence & Series of Functions
For example, the sequence f n defined by fn x sin nx, x R is uniformly bounded. Infact,
If each individual function is bounded and if f n f uniformly on E, then it can be shown that fn is
uniformly bounded on E. This result generally helps us to conclude that a sequence is not uniformly
convergent.
2.4 Cauchy Criterion for Uniform Convergence
We now find necessary and sufficient condition for uniform convergence of a sequence of
functions.
Theorem 1 (Cauchy criterion for uniform convergence). The sequence of functions fn , defined on
E, converges uniformly if and only if for every 0 there exists an integer N such that
m N, n N, x E imply f n x f m x .
Proof. Suppose first that f n converges uniformly on E to f. Then to each 0 there exists an integer
N such that n > N implies
f n x f x , for all x E
2
Similarly for m > N implies
fm x f x , for all x E
2
Hence, for n > N, m > N, we have
fn x fm x fn x f x f x fm x
fn x f x fm x f x
/ 2 / 2 for all x E
Hence the condition is necessary.
Conversely, suppose that the given condition holds. Therefore {fn(x)} is a Cauchy sequence in R for
each x E. Since R is complete, it follows that {fn(x)} converges to some value f(x), for each x E &
{fn} converges to f pointwise. We need only to show that the convergence is uniform. to show this let
>0 be given, then by hypothesis, n0 N (depending only on ) such that
f n x f m x ,
n, m > N and x E
Let n be fixed and let m , then we have
fn x f x x E
Hence f n f uniformly on E.
Mathematical Analysis 45
We now find necessary and sufficient condition for uniform convergence of a series of functions.
Theorem 2 (Cauchy criterion for uniform convergence). A series of real functions f n , each
defined on a set X converges uniformly on X iff for every 0, n0 N (depending only on )
such that
f x f x f x ............... f x ,
i 1
i 1 2 n x X
so that S n x is a sequence of partial sums of the series f
n 1
n . Now the series f n is uniformly
Proof. We have
lub f n x f x M n 0 as n .
xE
fn x f x
Hence f n f uniformly on E.
46 Sequence & Series of Functions
nx
f n x is not uniformly convergent on any interval containing 0.
1 n2 x2
nx
Solution. Here f x lim f n x lim
n n 1 n2 x2
x/n
lim 0
n 1/ n 2 x 2
Thus the sequence f n converges pointwise to the function f identically 0.
Now M n sup f n x f x
xa ,b
nx nx
sup 0 sup
xa ,b 1 n x xa ,b 1 n x
2 2 2 2
nx
Let us find the maximum value of by second derivative test.
1 n2 x2
nx
Let x
1 n2 x2
' x
1 n x n nx.2n x
2 2 2
1 n x 2 2 2
Put ' x 0 .
1 n x x 2 xn
2 2 2
1 2 x 2 n2 n2 x 2 1 n2 x 2
1 1
x2 2
x .
n n
1 1
or x or .
n n
Also,
1 n x .n 2n x n 1 n x 4n x 4n x
2
2 2 2 2 2 2 4 3
" x
1 n x
4
2 2
Mathematical Analysis 47
.
2
2n3 x 1 n 2 x 2 n 1 n 2 x 2 4n 2 x 4n 4 x 3
1 n x
4
2 2
1 1
At x , " x 0 . Therefore d(x) is maximum when x .
n n
1 1
Also
n 2
Thus we take an interval [a,b] containing zero ,then
nx 1
M n sup f n x f x sup ,
x a ,b 1 n x
2 2
x a ,b 2
x
f n x converges uniformly on R.
1 nx 2
Solution. Here pointwise limit is
x
f x lim 0 x R.
n 1 nx 2
x
Let x f n x f x .
1 nx 2
For maximum & minimum of x , we have
1 nx 2 2nx 2
' x 0 0
1 nx 2 2
1 nx 2 1
0 1 nx 2 0 x .
1 nx
2
2 n
Now, ' x
1 nx 2
1 nx 2 2
.
2 2 2
1 nx 2 1 nx 2 1 nx 2
1 2
.
1 nx 1 nx2 2
2
2nx 8nx
'' x ..
1 nx2 1 nx2
2 3
48 Sequence & Series of Functions
1
Put x ,
n
1 2 n 8 n n n
'' 2 2 23 2 n 2
n
1
' ' ve maximum.
n
1/ n 1
Hence max. x .
1 n 1/ n 2 n
Thus M n sup f n x f x
xR
x
sup 0 sup x 1
xR 1 nx 2 xR 2 n
1
Also so lim M n lim 0
n n
2 n
Hence by M n test, the sequence f n x uniformly converges on R.
Example 3. Show that 0 is a point of non – uniformly convergent of the sequence f n x , where
f n x nxe nx ; x 0 .
x f n x f x nxe nx
' x nx ne nx ne nx
' x 0 n 2 xenx ne
nx
Now 0
ne nx 1 1
x 2 nx
x
n e n n
n 3 xe nx 2n 2 e nx
1 1 1 2n 2 n2
' ' n3. .
n n e e e
1
' ' x ve i.e., maximum at x
n
1 1
Hence max. of x n. e 1 .
n e
Thus M n sup f n x f x
xR
1
sup nxe nx 0 sup x
xR xR e
1 1
So lim M n lim 0
n n e e
Hence by M n test, the sequence of function is not uniform convergent on R.
Weierstrass contributed a very convenient test for the uniformly convergence of infinite series of
functions.
Theorem 2 (Weierstrass M-test). Let f n be a sequence of functions defined on E and suppose
f n x M n ( x E , n = 1, 2, 3,……), where M n is independent of x. Then f n converges uniformly
f n 1 x f n 2 x .... f m x
M n 1 M n 2 ... M m .
Mn 1 Mn 2 ... Mm , n N.
Hence
Sm x Sn x , n N, x E
50 Sequence & Series of Functions
cos n
Example 4. Consider the series . We observe that
n 1 np
cos n 1
p.
np n
1 cos n
Also, we know that n
n 1
p
is convergent if p > 1. Hence, by Weierstrass M-Test, the series np
sin n
converges absolutely and uniformly for all real values of θ if p > 1. Similarly, the series
n 1 np
converges absolutely and uniformly by Weierstrass’s M-Test.
Example 5. Taking M n r n , 0 < r < 1, it can be shown by Weierstrass’s M-Test that the series
r n
cos n, r n
sin n, r n
cos2 n, r n
sin 2 n converge uniformly and absolutely.
x
Example 6. Consider , xR .
n 1 n 1 nx 2
We assume that x is positive, for if x is negative, we can change signs of all the terms. We have
x 1
fn x and fn ' x 0 implies nx 2 = 1. Thus maximum value of f n x is .
n 1 nx 2 2n3/2
1
Hence fn x
2n 3/ 2
1 x
Since n3/2 is convergent, Weierstrass’s M-Test implies that n 1 nx is uniformly convergent
n 1 2
for all x R .
x
Example 7. Consider the series , x R . We have
n 1 n x 2 2
x
fn x
n x 2 2
and so fn ' x
n x 2 2
2x n x 2 2x
n x 2 4
x 4 n 2 2nx 2 4nx 2 4x 4 0
n 2 2nx 2 3x 4 0
3x 4 2nx 2 n 2 0
n n
x2 or x .
3 3
3 3 1
Also f n '' x is negative. Hence maximum value of f n x is . Since n 3/2
is convergent, it
16n 3/ 2
follows by Weierstrass’s M-Test that the given series is uniformly convergent.
an xn
a n x 2n
Example 8. The series and
n 1 1 x n 1 1 x
2n 2n
converge uniformly for all real values of x and a n is absolutely convergent. The solution follow the
same line as for example 7.
Lemma 1 (Abel’s Lemma). If v1 , v 2 ,...., vn be positive and decreasing, the sum u1 v1 u 2 v 2 .... u n v n
lies between A v1 and B v1 , where A and B are the greatest and least of the quantities
u1 , u1 u 2 , u1 u 2 u 3 ,....., u1 u 2 .... u n .
Proof. Write
Sn u1 u 2 .... u n .
Therefore
u1 S1 , u 2 S2 S1 ,....., u n Sn Sn 1.
Hence
n
u v
i 1
i i u1v1 u 2 v2 .... u n vn .
S1 (v1 v2 ) S2 v2 v3 ..... Sn 1 vn 1 vn Sn vn
A v1 v2 v2 v3 .... vn 1 vn vn .
Av1.
Similarly, we can show that
n
u v
i 1
i i Bv1.
u n x , u n x u n 1 x ,...., u n x u n 1 x .... u m x .
Clearly A is function of x.
u n x u n 1 x .... u m x for all n > N, x E
M
and so A for all n > N (independent of x) and for all x E . Also, since vn x is decreasing,
M
vn x v1 x M since v1 x is bounded for all x E
Hence
u n x v n x u n 1 x v n 1 x .... u m x v m x
for n > N and all x E and so u x v x is uniformly convergent.
n 1
n n
1
n
x 2n
n 1 np 1 x 2n
.
1
n
1
n
If 0 p 1 , the series np
is convergent but not absolutely. Let
Mathematical Analysis 53
x 2n
vn x
1 x 2n
x 2n x 2n 2
v n x v n 1 x
1 x 2n 1 x 2n 2
x 2n 1 x 2
(+ve)
(1 x 2n ) 1 x 2n 2
x2
Also v1 x 1.
1 x2
1
n
x 2n
Hence, by Abel’s Test, the series
n 1 np
.
1 x 2n
is uniformly convergent for 0 p 1 and |x| < 1.
xn
Example 10. Consider the series an . 1 x 2n
, under the condition that a n is convergent. Let
xn
vn x
1 x 2n
Then
vn x 1 x 2n 2
v n 1 x x(1 x 2n )
and so
vn x 1 x 1 x 2n 1
1
v n 1 x x(1 x 2n )
which is positive if 0 < x < 1. Hence v n v n 1 and so vn x is monotonically decreasing and positive.
x xn
Also v1 x is bounded. Hence, by Abel’s test, the series a n . 1 x 2n is uniformly convergent
1 x2
in (0, 1) if a n is convergent.
nx n 1 1 x
Example 11. Consider the series a n
1 xn
under the condition that a n is convergent. We
have
nx n 1 1 x
vn x .
1 xn
Then
54 Sequence & Series of Functions
vn x n 1 x n 1
. .
v n 1 x (n 1)x (1 x n )
n
Since 0 as n , taking n sufficient large
(n 1)
vn x
1 x 1 if 0 < x < 1.
n 1
v n 1 x (1 x n )
Hence <vn(x)> is monotonically decreasing and positive. Hence, by Abel’s Test, the given series
converges uniformly in (0, 1).
Theorem 4. (Dirichlet’s Test for uniform convergence). The series u x v x converges
n 1
n n
uniformly on E if
(i) v x
n is a positive decreasing sequence for all values of x E , which tends to zero
uniformly on E
(ii) u x
n oscillates or converges in such a way that the moduli of its limits of oscillation
remains less than a fixed number M for all x E .
Proof. Consider the series u ( x)v ( x)
n n where vn x is a positive decreasing sequence tending to
n 1
zero uniformly on E. By Abel’s Lemma
u n x v n x u n 1 x v n 1 x .... u m x v m x Av n x ,
where A is greatest of the magnitudes
u n x , u n x u n 1 x ,...., u n x u n 1 x .... u m x
and A is a function of x.
s
Since un x converges or oscillates finitely in such a way that u x M for all x E ,
r
n
Proof. Since V n x converges uniformly to zero thus for any 0, an integer N (Independent of x)
s.t. for all x a, b
Vn x n N ...................... 1 .
4M
n
Let S n U r x n N & x a , b
r 1
so that Sn x M n........................ 2 .
n p
Now consider V xU x V xU x ........... V xU x
r n 1
r r n 1 n 1 n p n p
Vn1x Sn Vn 1 x Vn 2 x S n1 ........... Vn p1 x Vn p x S n p 1 Vn p x S n p
n+p-1
Vr x Vr 1 x S r x Vn 1 x S n x Vn p x S n p x
r n1
np n p 1
1
Vr x U r x Vr x Vr 1 x S r x Vn 1 x S n x Vn p x S n p x
r
n 1
r n
n p 1
n 1 Vr x Vr 1 x M 4M .M 4M .M n N
r
(By (1)&(2))
M Vn 1 x Vn p x
2
M Vn 1 x Vn p x
2
M .
4M 4M 2
56 Sequence & Series of Functions
n p
Hence by Cauchy Criteria, the series V x U x
r n 1
r r converges uniformly on [a,b].
n
Remark 1. The statement U x K x a, b
r 1
r & n is equivalent to saying that the sequence of
This fact is also stated as the partial sum of the series is uniformly bounded.
This, in turm is equivalent to saying that the series u x either converges uniformly or oscillates
n
finitely.
So Dirichlet’s test can be states also as “If Vn x is a monotonic function of n for each fixed value of x
in [a,b] and V n x converges uniformly to zero for x a, b and if Un(x) either uniformly converges
to zero or oscillates finitely in [a,b]. Then the series V xU x is uniformly convergent on [a,b].
n n
Cosn Sinn
Example 12. Prove that the series
np
and
n p
converges uniformly for all values of p > 0
Solution. When, p > 1 , By Weierstrass M-test at once prove both the series uniformly converge for all
values of .
When 0 p 1 , U r Cosr
1
Take bn p and U n Cosn or Sinn
n
1
Then by Dirichlet’s test p is positive and monotonic decreasing and uniformly tending to zero with
n
n n
U
r 1
r Cosn
r 1
Cos Cos2 .............. Cosn
Sinn
2 .Cos Ist.angle Last.angle
Sin 2
2
Sinn
2 .Cos n
Sin 2
2
Co sec n ( Sin n 1 and Cos 1 ).
2
Mathematical Analysis 57
Cosn Sinn
Thus all the conditions of Dirichlet’s test are fulfilled and the series np
and np
converges on ,2 .
f x f c f x f M x f M x f M c f M c f (c)
f x f M x f M x f M c f M c f (c)
3 3 3
Hence
f x f c , x S E
2 1
n x for 0 x
n
2 1 2
f n x n 2 x for x .
n n n
2
0 for x 1
n
Each of the function fn is continuous on [0, 1]. Also f n x 0 as n for all x 0,1 . Hence the
limit function f vanishes identically and is continuous. But the convergence f n f is non-uniform.
Proof. Let Sn x fni x , n N and let 0 . Since f n converges uniformly to f on [a, b], there
i 1
Sn x f1 x f 2 x .... f n x
is also continuous on [a, b] for all n. Hence to each 0 there exists a 0 such that
Sn x Sn c whenever x c (3).
3
Now, by triangle inequality, and using (1), (2) and (3), we have
f x f c f x Sn x Sn x Sn c Sn c f (c)
f x Sn x Sn x Sn c Sn c f (c)
, whenever x c .
3 3 3
Hence f is continuous at c. Since c is arbitrary point in [a, b], f is continuous on [a, b].
However, the converse of Theorem 1 is true with some additional condition on the sequence f n of
continuous functions. The required result goes as follows:
Mathematical Analysis 59
Theorem 3. (Dini‘s theorem on uniform convergence of subsequences (first form)). Let E be
compact and let f n be a sequence of functions continuous on E which converges to a continuous
function f on E. If f n x f n 1 x for n = 1, 2, 3, …, and for every x E , then f n f uniformly on E.
Proof. Take
gn x fn x f x
.
Being the difference of two continuous functions g n x is continuous. Also g n 0 and g n g n 1 . We
shall show that g n 0 uniformly on E.
Let 0 be given. Since g n 0 , there exists an integer n N x such that
gn x 0 / 2
In particular
g Nx x 0 / 2
i.e. 0 g Nx x / 2.
The continuity and monotonicity of the sequence g n imply that there exists an open set J(x) containing
x such that
0 gn t
if t J(x) and n N x .
Since E is compact, there exists a finite set of points x1 , x 2 ,...., x m such that
E J(x1 ) J(x 2 ) ... J(x m ).
Taking
N max N x1 , N x 2 ,..., N x m .
it follows that
0 gn t
for all t E and n N . Hence g n 0 uniformly on E and so f n f uniformly on E.
Theorem 4. If a sequence f n of real valued function converges uniformly to f in [a,b] and let x 0 be a
point of [a,b] s.t. lim f n x an ; n 1, 2,......... . .
x x0
Proof. (i) The sequence f n converges uniformly on [a,b] .Therefore for 0 , there exists an
integer m (independent of x ) s.t. for all x a, b
an p an n m, p 1
Thus for given 0 , there exists an integer N 1 s.t. for all x a, b .
f n x f x n N1 1
3
Now the sequence a n converges to A. So there exists an integer N 2 s.t.
an A n N2 (2)
3
Now take a no. N such that N max.N1 , N 2
Since we have ,
lim f n x an
x x0
In particular, lim f N x a N
x x0
f N x a N x x0 3
3 whenever
Now, f x A f x f N x f N x a N a N A
lim f n x a n ; n 1,2,...............
x x0
Then (i) a
n 1
n converges
(ii) lim f x a n
x x0
n 1
Proof. (i) Given that the series f n converges uniformly on [a,b], for given 0 , there exists an
integer m such that for all x a, b
n p
f x n m, p 1
r
r n 1
(By Cauchy’s Criterion)
Keeping n,p fixed and taking the limits x x 0 , we obtain
n p
a x
r n 1
r
Again a n converges to A.
n
a r A n N 2 .................2
r 1 3
f n x a n x x0 i
Such that 3N whenever .
f n x an for x x0
3N
N N N
Thus f x a
r 1
r
r 1
r f r x ar N .
r 1
............3
3N 3
f x A
We have seen earlier that if sequence f n is a sequence of continuous functions which converges
pointwise to the function f, then it is not necessary for f to be continuous. However, the concept of
uniform convergence is of much importance as the property of continuity transfers to the limit function
if the given sequence converges.
Theorem 6. If the sequence of continuous function f n is uniformly convergent to a function f on
[a,b] then f is continuous on [a,b].
Proof. Let 0 be given.
Now given that sequence f n is uniformly convergent to f on [a, b], then there exists a positive integer
m such that
f n x f x n m & x a, b 1
3
Let x 0 be any point of [a, b].
f x f x 0 f x f n x f n x f n x0 f n x 0 f x0
f x f n x f n x f n x0 f n x 0 f x0
(from (1), (2)& (3))
3 3 3
x 0 a, b
Hence f is continuous at
f is continuous on [a,b].
Theorem 7. If a series f
n 1
n of continuous function is uniformly convergent to a function f on [a, b],
Hence for x x0 ,
N N N N
f x f x 0 f x f r x f r x f r x 0 f r x 0 f x0
r 1 r 1 r 1 r 1
64 Sequence & Series of Functions
N N N N
f r x f x
r 1
f r x0 f x0
r 1
f r x f r x0
r 1 r 1
f x f x0
f is continuous at x0 on [a,b]. Since x0 was chosen arbitrary.
Remark 1. (i) Uniform convergence of the sequence f n is sufficient but not a necessary condition for
the limit function to be continuous. This means that a sequence of continuous functions may have a
continuous limit function without uniform convergence.
However the above theorem yields a negative test for uniform convergence of a sequence namely “If the
sequence of continuous functions is discontinuous, the sequence cannot be uniformly convergent.”
(ii) The same argument hold good in the case of infinite series fn 1
n .
(1) The sequence x of continuous functions has a discontinuous limit function f which is given by
n
0, if 0 x 1
f x
1, if x 1 .
nx
(2) The sequence 2 2 of continuous functions has a continuous limit function but the given
1 n x
sequence is not uniformly convergent.
1, if x 0
f x
0, if x 0
which is discontinuous on [0,1]. Therefore the series is not uniformly convergent on [0,1].
n 1
Mathematical Analysis 65
1
1 x 1.
1 x
(1) Every monotonically increasing sequence bounded above converges to the least upper bound
(l.u.b.).
(2) Every monotonically decreasing sequence bounded below converges to greatest lower bound
(g.l.b).
(3) A real no. is said to be a limit point of a sequence a n if given any 0 and a +ve integer
m, there exists a +ve integer k > m such that ak .
(ii) diverges to if r 1.
(i) a n 1 a n n
(ii) a n 0 as n
(8) For every limit point of a sequence we can form a subsequence converging to limit point. Limit
point is also called subsequential limit.
Theorem 8 (Dini’s theorem on uniform convergence of subsequences(2nd form)). If a sequence of
continuous function f n defined on [a,b] is monotonically increasing & converges pointwise to a
continuous function f, then the convergence is uniform on [a,b].
66 Sequence & Series of Functions
Therefore, for any 0 and for a point x a, b there is an integer N s.t.
0 f x fn x n N 1
We consider Rn f x f n x ; n = 1,2,…………..
Since the sequence f n is monotonically increasing. So, the seq. R n x is monotonically decreasing.
i.e, R1 x R2 x R3 x ........... Rn x 2 .
Also, the sequence R n x is bounded below by 0.
Suppose if possible for a fixed a 0 0, no integer N which works for all x a, b .
Rn xn a0 3
The seq. x n of points belonging to the interval [a,b] is bounded and thus has atleast one limit say ' '
in [a,b].
i.e, x n as k .
k
k
lim Rm x nk Rm x nk as k .
But this is contradiction to the fact that sequence Rm converges pointwise to 0 i.e.,
Mathematical Analysis 67
lim Rm 0
n
Thus the convergence must be uniform and this completes the proof.
Theorem 9 (Dini’s theorem on uniform convergence for series). If the sum function of a series
f n with non negative terms defined on an interval [a, b] is continuous on [a,b], then the series is
uniformly convergent on the interval [a, b].
Proof. Consider the partial sum of the given series
n
S n x f r x
r 1
Since all the function f r are non –ve . So, the seq. of partial sum S n should be increasing.
Therefore, Sn x Sn1 x n
x4 x4
x
4
...................
1 x4 1 x4
2
Now, Let us find the sum function for the given series. Let, f x denotes the sum function of the given
series.
1 1
If x 0 then the series is a geometric series with common ratio 4 and
1x 0,1 .
1 x 1 x4
Hence the sum function is given by
x4
f x 1 x4
1
1
1 x4
1 x 4 if x 0
Thus, f x
0, if x 0
which is discontinuous on 0 and hence on [0,1] . So, the series cannot converge uniformly on [0,1].
68 Sequence & Series of Functions
x
Example 2. Show that the series nx 1n 1x 1 is uniformly convergent on any interval [a, b],
0 < a < b, but only pointwise on [0, b].
x 1 1
Solution. Let f n x
nx 1n 1x 1 n 1x 1 nx 1
Therefore nth partial sum is
n
S n x f r x f 1 x f 2 x .......... f n x
r 1
1 1 1 1 1
1 ..........
x 1 x 1 2x 1 n 1x 1 nx 1
1
1
nx 1
1
lim 1
n
nx 1
1 if x 0
0, if x 0
Clearly f is discontinuous at x = 0 and hence discontinuous on [0, b].
This implies that the convergence is not uniform on [0, b] i.e, it is only pointwise.
Now take the interval [a,b] such that 0 < a < b, then the given series is uniformly convergent on [a,b] if
for given 0 .
1
S n x f x
nx 1
11
i.e, if n 1
x
11
Now, 1 decreasing with x and its maximum value is
x
11
1 m0 (say).
a
values of x.
1n1
Solution. The given series is n x
n 1
2 .
1
Let an .
n x2
a n 1 a n n and a n 0 as n .
1n1
Hence by Leibnitz’s rule, the alternative series is convergent.
n 1 n x2
1
n 1
1
Now, n 1 n x 2
n 1 n x
2 which behaves like 1n and hence is divergent.
It remains to prove that the given series is uniformly convergent.
Let S n x denotes the partial sum and S x denote the sum of the series.
Now, consider
1 1 1 1 1
S2n x .............
1 x 2
2 x 3 x
2 2
4 x 2
2n x 2
1 1 1 1 1 1
S2n x .................
1 x
2
2 x2 3 x2 4 x2 2n 1 x 2 2n x 2
Now, note that each bracket in the above expression is positive. Hence S 2 n x is positive and increasing
to the sum S x .
S x S 2 n x 0
1 1 1
Also S x S 2 n x ......
2n 1 x 2n 2 x 2n 3 x 3
2 2
1 1 1
...........
2
2n 1 x 2n 2 x 2n 3 x
2 2
70 Sequence & Series of Functions
1
2n 1 x 2
1
.
2n 1
1
So, 0 S x S2 n x 1 .
2n 1
Also, consider
1 1 1
S 2 n 1 x S x .......... ....
2n 2 x 2n 3 x 2n 4 x 2
2 2
1 1 1 1
2
..............
2
2n 2 x 2n 3 x 2n 4 x 2n 5 x
2 2
1 1
S 2 n 1 x S x .
2n 2 x 2n 2
2
1 1
0 S2 n 1 x S x 2 .
2n 2 2n 1
S x S n x n m
Show that the sequence of differentiable functions f n does not converge uniformly in an interval
containing zero.
nx
Solution. Here f n x
1 n2 x2
f x lim f n x 0
n
' x
1 n x n 2nx.n x n
2 2 2 3 2
1/ n x 2 x
2 2
Now, f n
1 n x
1/ n 2 x 2 2
2 4
2 2n
Hence the sequence f n ' does not converges uniformly in an interval that contains zero.
2.7 Uniform Convergence and Integrability.
We know that if f and g are integrable, then f g f g and this result holds for the sum
of a finite number of functions.
The aim of this section is to find sufficient condition to extend this result to an infinite number of
functions.
Theorem 1. Let α be monotonically increasing on [a, b]. Suppose that each term of the sequence
f n is a real valued function such that f n R() on [a, b] for n = 1, 2,.. and suppose f n f uniformly
on [a, b]. Then f R() on [a, b] and
b b
f d lim f
a
n
a
n d ,
b b
that is, lim f n (x) d(x) lim f n (x) d(x)
a
n n
a
(Thus limit and integral can be interchanged in this case. This property is generally described by saying
that a uniformly convergent sequence can be integrated term by term).
Proof. Let be a positive number. Choose 0 such that
b a 1
3
This is possible since is monotonically increasing. Since f n f uniformly on [a, b], to each 0
there exists an integer n such that
f n x f x , x a, b (2)
U P, f , U P, f n , (4)
3
Similarly, f x f n x implies
L P, f , L P, f n , (5)
3
Combining (3), (4) and (5), we get
U P, f , L P, f ,
Further uniform convergence implies that to each 0 , there exists an integer N such that n N
fn x f x , x a, b
b a
f d f n d
a a
f f n d f f n
a a
d
b
d x dx
b a a
b a
.
b a
b b
Hence f d lim f
a
n
a
n d
Proof. Let Sn denotes the sequence of partial sums of f n . Since f n converges uniformly to f on
[a, b], the sequence Sn converges uniformly to f. Then Sn being the sum of n integrable functions is
integrable for each n. Therefore, by theorem 1, f is also integrable in Riemann sense and
b b
f x dx lim S x dx
a
n
a
n
b b b b
n b
fi x dx
i 1 a
b n b
f x dx lim f x dx
a
n
i 1 a
i
b
f i x d
i 1 a
Example 1. Consider the sequence f n for which fn x nxe , n N, x 0,1 . We note that
2
nx
f x lim f n x
n
nx
lim 2
0, x 0,1
n nx n2x4
1 ......
1! 2!
Then
1
f x dx 0
0
1 1
f n x dx nxe nx dx
2
0 0
n
1 t
20
e dt, t nx 2
1
1 e n
2
Therefore
1 1
lim f n x dx lim 1 e n .
n n 2 2
74 Sequence & Series of Functions
1
If f n were uniformly convergent, then f x dx should have been equal to lim f n x dx .
n
0
But it is not the case. Hence the given sequence is not uniformly convergent to f. In fact, x = 0 is the
point of non-uniform convergence.
x
Example 2. Consider the series . This series is uniformly convergent and so is integrable
n x2
2
n 1
1 m 1
x x
0
n 1 n x
2 2
dx lim
m
n 1 0 n x
2 2
m 1
lim x n x 2 dx
2
m
n 1 0
1
n x 2 1
m
lim
m
n 1 2
0
m
11 1
lim
m
n 1 2 n n 1
1 1 1 1 1 1
lim 1 ...
m 2
2 2 3 m m 1
1 1 1
lim 1
m 2
m 1 2
Example 3. Consider the series
nx
n 1 x , a x 1.
n 1 1 n x
2 2
1 n 1 x 2
2
nx
Sn x
1 n x
2 2
As we know that 0 is point of non-uniform convergence of the sequence Sn x , the given series is not
uniformly convergent on [0, 1]. But
Mathematical Analysis 75
1 1
f x dx 0 dx 0
0 0
1 1
nx
and S x dx 1 n x dx
0
n
0
2 2
1
1 2n 2 x
2n 0 1 n 2 x 2
dx
1
1
log 1 n 2 2
x
2n 0
1
log 1 n 2 .
2n
Hence
1
lim Sn x dx lim
1
log 1 n 2
form
n 2n
n
0
n
lim 2
form
n 1 n
1
lim 0.
n 2n
Thus
1 1
f x dx lim S x dx ,
0
n
0
n
and so the series is integrable term by term although 0 is a point of non-uniform convergence.
lim V g g n 0
n
and g(a) = 0. Then for every continuous function f on [a, b], we have
b b
lim f dg n lim f dg
n n
a a
V g V gn V g gn
Since g n is of bounded variation and lim V g g n 0 it follows that total variation of g is finite and
n
so g is of bounded variation on [a, b]. Thus the integrals in the assertion of the theorem exist.
b b b
f dg f dg
a a
n f d g g
a
n
M V g gn
.
f dg lim f dg n .
a
n
a
Furthermore,
g x gn x V g gn , axb
Therefore, as n , we have
g n g uniformly.
fn x0 fm x0 (1)
2
Mathematical Analysis 77
and
fn ' t fm ' t a t b (2).
2b a
x t
fn x fm x f n t fm t (3)
2 b a 2
fn x fm x fn x fm x fn x0 fm x0 fn x0 fm x0 .
fn x fm x / 2 / 2 a x b .
Hence, by Cauchy criterion for uniform convergence, it follows that f n converges uniformly on [a, b].
Let
f x lim f n x a x b .
n
fn t fn x f t f x
n t , t (4)
tx tx
for a t b, t x . Then
fn t fn x
lim n t lim fn ' x (n = 1, 2,…) (5)
tx tx tx
Further, (3) implies
n t m t n N, m N .
2(b a)
Hence n converges uniformly for t x . We have proved just now that f n converges to f uniformly
on [a, b]. Therefore (4) implies that
lim n t t (6)
n
lim lim n t
n t x
78 Sequence & Series of Functions
lim f n ' x .
n
Remark 1. If in addition to the above hypothesis, each f n ' is continuous, then the proof becomes
simpler. Infact, we have then
Theorem 2. Let f n be a sequence of functions such that
(iv) f n ' converges uniformly to g on [a, b], then f is differentiable and f n ' x g x for all
x a, b .
Proof. Since each f n ' is continuous on [a, b] and f n ' converges uniformly to g on [a, b], the
application of Theorem 1 of section 2.6 of this unit implies that g is continuous and hence Riemann
integrable. Therefore, Theorem 1 of section 2.7 of this unit implies
t t
g x dx lim f n ' x dx
a
n
a
f ' x dx f t f a
a
n n n
Hence
t
Hence
t
g x dx f t f a
a
and so
Mathematical Analysis 79
d
t
g x dx f ' t
dt a
or g t f ' t , t a, b .
the sequence Sn converges to f on [a, b]. Further, since f n ' converges uniformly to g on [a, b], the
sequence Sn ' of partial sums converges uniformly to g on [a, b].
Example 1. Consider the series
nx
n 1 x .
n 1 1 n x
2 2
1 n 1 x 2
2
For this series, we have
nx
Sn x , 0 x 1
1 n x
2 2
We have seen that 0 is a point of non-uniform convergence for this sequence. We have
nx
f x limSn x lim
n n
1 n x
2 2
0 for 0 x 1.
Therefore
f ' 0 0
80 Sequence & Series of Functions
Sn 0 h Sn 0
Sn ' 0 lim
h h
n
lim n
h 0 1 n 2 h 2
Hence
limSn ' 0 .
n
Then
sin nx
Example 2. Consider the series , x R . We have
n 1 n3
sin nx
fn x
n3
cos nx
fn ' x .
n2
Thus
cos nx
f ' x n
n 2
cos nx 1 1
Since
n 2
2 and
n
n 2
is convergent, therefore, by Weierstrass’s M-test the series f ' x
n is
uniformly as well as absolutely convergent for all x R and so f n can be differentiated term by
term.
'
Hence n fn '
f
n 1 n 1
'
sin nx
cos nx
or n3 n 2
n 1 n 1
2.9 Weierstrass’s Approximation Theorem
Weierstrass proved an important result regarding approximation of continuous function which has many
applications in numerical methods and other branches of mathematics.
The following computation shall be required for the proof of Weierstrass’s approximation theroem.
For any p,q R , we have, by Binomial Theorem
Mathematical Analysis 81
n
n
k p q k nk
( p q)n , , n I , (1)
k 0
where
n n!
k k ! n k !
Differentiating with respect to p, we obtain
n
n
k kp k 1 n k
q n( p q ) n 1 ,
k 0
which implies
n
k n
n k p q k nk
p( p q ) n 1 , n I (2).
k 0
Differentiating once more, we have
n
k 2 n k 1 n k
n2
p q p(n 1)( p q) ( p q)
k 0 n k
n 1
and so
n
k 2 n k nk 1 p
2
k 0 n k
p q p 2 (1 )( p q) n 2 ( p q) n 1
n n
(3).
Now if x [0,1] , take p = x and q = 1-x. Then (1), (2) and (3) yield
n n k
x 1 x 1
nk
k 0 k
n k n
x 1 x x
k nk
(4).
k 0 n k
n k2 n
2 x k 1 x n k x 2 1
1 x
k 0 n k n n
2
k
On expanding x , it follows from (4) that
n
x 1 x
2
n
k n k
x k x 1 x
nk
0 x 1 (5).
k 0 n n
k n
The polynomial defined in (*) is called Bernstain polynomial as shown in above equation (6).
We shall prove that certain Bernstain polynomial exists which uniformly converges to f on [0,1].
Now consider the identity
n
n
k x 1 x x 1 x 1 1
k n k n
k 0
k kx 1 x 0
n
n
n k x k 1 x
k 1 nk n k 1
k 0
n
n
x k 1 1 x k nx 0
n k 1
k 0 k
k 0 k
k nx 1 x k nx2 0
n
n
x k 1 1 x
k nk n k 1
k 0
which on applying (1),we get
k x 1 x k nx2 n
n
n k 1 n k 1
k 0
Multiplying by x(1-x), we get
n
n
k x 1 x k nx nx 1 x
k nk 2
k 0
x 1 x
2
n
n
nk k
x k 1 x x 3
k 0 k n n
Since the maximum value of x(1-x) in [0,1] is ¼.
f x x 1 x , f ' x 1 2 x
f ' x 0 1 2x 0
x 1/ 2 f 1/ 2 1/ 4.
So, (3) can be written as
2
n
n
nk k 1
x k 1 x x 4
k 0 k n 4n
Now f is continuous on [0,1]. So, f is bounded and uniformly continuous on [0,1].
K 0 such that
f x K x 0,1
and by uniform continuity for given 0 , there exists 0 such that for all x 0,1 .
84 Sequence & Series of Functions
k k
f x f whenever x 5.
n 2 n
Now for any fixed but arbitrary x in [0,1], then n- values 0,1,2,………..,n of k can be divided into two
parts as follows:
k
Let A be the set of values of k for which x and B be the set of remaining values for which
n
k
x .
n
kB k kB k n 4n
n 1
x k 1 x 6
nk
kB k 4n 2
Now
n
n k
f x Bn x 1. f x x k 1 x f
nk
k 0 k n
(By (1))
n
n
k
k x 1 x f x
nk
k
f
k 0 n
n
n k
f x Bn x x k 1 x f x f
n k
k 0 k n
We split the summation on R.H.S into two parts accordingly as
k k
x or x
n n
Let k A or k B .
Thus we have
n k n k
f x Bn x x k 1 x f x f x k 1 x f x f
nk k
k A k n kB k n
n n
x 1 x 2 K x k 1 x
nk nk
k
2 k
k A kB k
2K K
for all values of n 2 .
2 4n 2
Mathematical Analysis 85
on [0,1], then
1 1
0 p x f x dx 0 a n x n f x dx
n 0
1
an x n f x dx an .0 0.
n 0 0 n 0
0
f 2 dx lim p n. . f dx 0
n
0
Therefore, f
2
x 0 on [0,1].
Hence f(x) = 0 on [0,1].
2.10 References
1. Walter Rudin, Principles of Mathematical Analysis (3rd edition) McGraw-Hill, Kogakusha, 1976,
International Student Edition.
2. T. M. Apostol, Mathematical Analysis, Narosa Publishing House, New Delhi, 1974.
3. H.L. Royden, Real Analysis, Macmillan Pub. Co., Inc. 4th Edition, New York, 1993.
4. G. De Barra, Measure Theory and Integration, Wiley Eastern Limited, 1981.
5. R.R. Goldberg, Methods of Real Analysis, Oxford & IBH Pub. Co. Pvt. Ltd, 1976.
6. R. G. Bartle, The Elements of Real Analysis, Wiley International Edition, 2011.
7. S.C. Malik and Savita Arora, Mathematical Analysis, New Age International Limited, New
Delhi, 2012.
U N IT – III
POWER SERIES AND FUNCTION OF SEVERAL VARIABLES
Structure
3.0 Introduction
3.1 Unit Objectives
3.2 Power Series
3.2.1 Power series
3.2.2 Uniform convergence and uniqueness theorem
3.2.3 Abel theorem
3.2.4 Tauber theorem
3.3 Function of several variables
3.3.1 Linear transformation
Euclidean space Rn
3.3.2 Derivatives in an open subset E of Rn
Chain rule
3.3.3 Partial derivatives
Continuously differentiable mapping
Young theorem
Schwarz theorem
3.4 References
3.0 Introduction
In this unit, we study convergence and divergence of a power series and applications of Abel’s theorem.
Tauber showed that the converse of Abel’s theorem can be obtained by imposing additional condition on
coefficients, whenever the converse of Abel’s theorem is false in general. Many of the concepts i.e.,
continuity, differentiability, chain rule, partial derivatives etc are extended to functions of more than one
independent variable.
3.1 Unit Objectives
After going through this unit, one will be able to
understand the concept of power series and radius of convergence.
identify the notation associated with functions of several variables
familiar with the chain rule, partial derivatives and concept of derivation in an open subset of Rn.
know the features of Young and Schwarz’s Theorems.
Mathematical Analysis 87
3.2 Power Series
A very important class of series to study is power series. A power series is a type of series with terms
involving a variable. Evidently, if the variable is x, then all the terms of the series involve powers of x.
So we can say that a power series can be design of as an infinite polynomial. In this section we will give
the definition of the power series as well as the definition of the radius of convergence, uniform
convergence and uniqueness theorem, Abel and Tauber theorems.
Definition 1. A power series is an infinite series of the form a
n 0
n x n where a n ' s are called its
coefficients.
Definition 2 (Convergence of power series). It is clear that for x = 0, every power series is convergent,
independent of the values of the coefficients. Now, we are given three possible cases about the
convergence of a power series.
(a) The series converges for only x = 0 which is trivial point of convergence, then it is called
“nowhere convergent”
e.g. n! x n
converges only for x = 0 and for x 0 , we have
lim n! x n .
n
Thus the terms of the series do not converge for x 0 and thus the series converges only for x =
0. Hence it is ‘Nowhere convergent’ series.
(b) The series converges absolutely for all values of x, then it is called “Everywhere convergent”.
e.g. The series converges absolutely for all values of x,
xn x n 1
un , un 1
n! n 1!
un x n (n 1)! n 1
lim n 1 .
n un 1 n! x x
By D-Ratio test, the series converges for all values of x. So, it is called “Everywhere convergent”
series.
(c) The series converges for some values of x and diverges for others.
e.g. The series x
n 0
n
converges for x < 1 and diverges for x > 1.
The collection of points x for which the series is convergent is called its “Region of
convergence”.
88 Power Series & Function of Several Variables
Definition 3. Let a
n 0
n x n be a power series. Then, applying Cauchy’s root test, we observe that the
power series a
n 0
n x n is convergent if
1
x ,
L
where
1/n
L lim a n .
1
The series is divergent if x .
L
Taking
1
R 1/ n
.
lim a n
We will prove that the power series is absolutely convergent if x R and divergent if x R . If
a 0 , a1 ,.... are all real and if x is real, we get an interval R x R inside which the series is convergent.
If x is replaced by a complex number z, the power series a
n 0
n z n converges absolutely at all points z
inside the circle z R and does not converge at any point outside this circle. The circle is known as
circle of convergence and R is called radius of convergence. In case of real power series, the interval
(-R, R) is called interval of convergence.
1/n
If lim a n 0, then R and the power series converges for all finite values of x. The function
represented by the sum of series is then called an Entire function or an integral function. For example,
e z , sin z and cos z are integral functions.
1/n
If lim a n , R 0, the power series does not converge for any value of x except x = 0.
Definition 4. Let R be the radius of convergence of the power series a
n 0
n z n , then the open interval
(-R, R) is called the interval of convergence for the given power series.
1
a
1/ n
Theorem 1. Let n x n be a power series such that lim a n .
n R
Then the power series is convergent with radius of convergence R.
Mathematical Analysis 89
Proof. Given that
1/ n 1
lim a n
n R
1/ n x
So, lim a n x n
n R
x x
Hence by Cauchy’s Root test, the series a x n
n
is convergence if
R
1 and divergent if
R
1 i.e,
Remark 1. (i) From the proof of above theorem, it follows that if for the series a x n
n
,
1/ n 1
lim a n
n R
then the series is absolutely convergent.
(ii) In view of the last theorem, we define the power series of convergence in the following way:
Consider the power series an x n , then the radius of convergence of this series is given by
1 1/ n
R 1/ n
when lim a n 0
lim a n
1/ n
0 when lim a n
1/ n
when lim an 0.
Thus a n x 0n 0 as n .
x1
Now x1 x0 1.
x0
n
x1
Thus x0
is geometric series with common ratio less than 1. So, it is convergent. By comparison
Let if possible, the series is convergent for x x" , then by theorem 2, it must be convergent for all x
such that x x" .
Hence the series diverges for every x x" , where x" x' .
an
For the power series a x n
n
, the radius of convergence is also defined by the relation R lim
n a
,
n 1
x2 x3
(1) x ...............
2! 3!
(2) 1 x 2! x 2 3! x 3 ....................
Mathematical Analysis 91
1 1.3 2 1.3.5 3
(3) x x x .................
2 2.5 2.5.8
1 2 1.2 3 1.2.3 4
(4) x x 3 x 4 x ...............
22 3 4
1
Solution. (1) Here a n
n!
an 1
R lim lim n 1!
n a n n!
n 1
n!
R lim 0
n n 1!
So, the series converges for no value of x other than zero. So, it is nowhere convergent series.
1.3.5...................2n 1
(3) Here a n
2.5.8.................3n 1
3 2/n 3
lim
n 2 1/ n 2
3
So series converges for all x where x .
2
(4) Here a n
n 1!
nn
n 1! n 1
n 1 n 1
1
R lim lim 1 e.
n nn n! n
n
Definition 6. Let f(x) be a function which can be express in terms of the power series as
f x an x n ,
n 0
f (x) a n x n x R .
n 0
Then
(i) a n 0
n x n converges uniformly on [R , R ], 0.
a n x n a n R
n
Since every power series converges absolutely in interior of its interval of convergence by Cauchy’s root
test, the series a n R converges absolutely and so, by Weierstrass’s M-test, a n x n converges
n
(ii) Also then the sum f(x) of a n x n is continuous and differentiable on (-R, R) and a x n
n
is
uniformly convergent on R , R .
Since n
1/n
1 as n , we have
lim n | a n | lim | a n |
1/n 1/n
Hence the series a n x n and na n x n 1 have the same interval of convergence. Since na n x n 1 is a
power series, it converges uniformly in R , R for every 0. Then, by term by term
differentiation yields
Mathematical Analysis 93
na n x n 1 f '(x) if x R .
But, given any x such that x R we can find an 0 such that x R. Hence
na n x n 1 f '(x) if x R.
Note. It follows from the above theorem 4 that by repeated application of the theorem f can be
differentiable any number of time and series obtained by differentiation at each step has the same radius
of convergence as series a x n
n
.
Theorem 5. Under the hypothesis of Theorem 4, f has derivative of all orders in (-R, R) which are given
by
f (k ) (x) n(n 1)(n 2)...(n k 1)a n x n k .
n k
In particular
f ( k ) (0) k ak , k 0,1, 2,...........
Proof. Let
f ( x ) nan x n .
n 0
Then by theorem 4,
f ' ( x ) n an x n 1.
n 1
........................................
........................................
........................................
f ( k ) ( x) n(n 1)(n 2)......(n k 1) an x n k .
nk
Also we can find coefficient from the values at origin of f , f ', f ",...
94 Power Series & Function of Several Variables
Theorem 6 (Uniqueness theorem). If a n x n and b n x n converge on some interval (-R, R), R>0
to some function f, then
a n b n for all n N.
Proof. Under the given condition, the function f have derivatives of all order in (-R, R) given by
f ( k ) ( x ) n( n 1)( n 2)......( n k 1) an x n k .
nk
interval of convergence (-R, R), then it uniformly converges in the interval [0, R].
Proof. Consider the sum
Then, we have
S n ,1 a n 1 R n 1
S n , 2 a n 1 R n 1 a n 2 R n 2
and so on.
This gives
an 1 R n 1 Sn ,1
an 2 R n 2 Sn ,2 Sn ,1
.............................. (1)
.............................
an p R n p Sn , p Sn , p 1
Let є > 0 be given.
Now the series a
n 0
n x n is convergent at x = R.
Mathematical Analysis 95
The series of numbers a R
n 0
n
n
is convergent and hence by Cauchy’s general principle of convergence,
an 1 R n 1 an 2 R n 2 ....... an q R n q n N q 1, 2,..........
a n 1 x n 1 a n 2 x n 2 ............... a n p x n p
n 1 n2 n p
n 1 x n2 x n p x
an 1 R an 2 R ............ an p R
R R R
n 1 n2 n p
x x x
Sn ,1 Sn ,2 Sn ,1 ............ Sn , p Sn , p 1
R R R
x n 1 x n 2 x n 2 x n 3 x
n p
Sn ,1 Sn ,2 ........ Sn , p
R R R R R
x n 1 x n 2 x n 2 x n 3 x
n p
Sn ,1 Sn ,2 ......... Sn , p
R R R R R
x x
n 1 n2 n2 n3 n p 1 n p n p
x x x x x
...........
R R R R R R R
x
.1 (by (3))
R
Thus we have proved that
an 1 x n 1 an 2 x n 2 ......... an p x n p p 1, x 0, R .
Hence by Cauchy’s criterion of convergence of series, the series a x
n 0
n
n
converges uniformly on [0, R].
Remark 4. (i) In case, a power series with interval of convergence (-R, R) converges at x R , then
the series is uniformly convergent in [-R, 0].
96 Power Series & Function of Several Variables
Similarly, if a series convergent at the end points -R and R, then the series is uniformly convergent on [-
R, R].
(ii) If a power series with interval of convergence (-R, R) diverges at end point x R , then it cannot be
uniformly convergent on [0, R].
For, if the series is uniformly convergent on [0, R], it will converge at x = R. A contradiction to the
given hypothesis.
lim f x an R n .
x R
a xn
n
an R n y n bn y n where bn a n R n
.
Now, this is a power series with radius R’, where
1 1 R
R' 1.
n 1/ n 1/ n
R
lim an R lim an R
So, if any series is given, we can transform it in another power series with unit radius of convergence.
Hence we can take R = 1.
Thus, now it is sufficient to prove that let a x
n 0
n
n
be a power series with unit radius of convergence and
let f x an x n ; x 1, if the series an converges then lim f x an . .Let us proceed to prove
x 1
n 0
the same.
S n a 0 a1 .......... .. a n
S 1 0 and a
n0
n S.
m m
Then a n x n S n S n 1 x n
n 0 n0
m m
S n x n S n 1 x n
n 0 n 0
Mathematical Analysis 97
m 1 m
S n x n S m x m x S n 1 x n 1
n 0 n 0
m 1 m
S n x n x S n 1 x n 1 S m x m
n 0 n 0
m 1 m 1
Sn x n x Sn x n Sm x m S1 0
n 0 n 0
m 1
1 x S n x n S m x m .
n 0
m m 1
lim an x n lim 1 x Sn x n lim Sm x m
m m m
n 0 n 0
f x 1 x Sn x n 1
n 0
Sn S n N (2)
2
Also, we have
1 x x n 1 3
n 0
1 x S n x n 1 x Sx n (by (3))
n 0 n 0
98 Power Series & Function of Several Variables
1 x Sn S x n
n 0
N
1 x Sn S x n (1 x) x n
n 0 2 n N 1
N
1 x Sn S x n
n 0 2
N
Now for a fixed N, 1 x S n S x n is continuous function of x having zero value at x = 1.
n 0
1 x S n S x n
N
n0 2
f x S whenever 1 x 1
2 2
Hence, lim f x S an .
x 1
n 0
Remark 5. We state some result related to Cauchy product of two series which will use in following
theorem, which is infact an application of Abel’s theorem.
(i) Let an and
n0
bn , then the series
n0
c
n 0
n where cn a0bn a1bn 1 .......... an b0 is called
Cauchy product of series a
n0
n & bn .
n 0
(ii) Cauchy’s Theorem. Let an &
n0
b
n 0
n be absolutely convergent series such that
c
n
n AB .
Theorem 9. If a n & bn and
n0 n 0
c
n 0
n converges to sum A, B & C respectively and if c n be
c n a 0 bn a1bn 1 .......... a n b0
Let f x a n x n , g x bn x n and h x c n x n ; 0 x 1 .
n 0 n0 n 0
c x n
n
f x g x (By Cauchy’s theorem in Remark 5(ii))
h x f x .g x ;0 x 1. (1)
Similarly, g x B, h x C as x 1 2
Thus from (1) & (2), we have
AB = C.
Example 1. Show that
x3 x5 x7
(i ) tan 1 x x
..............
3 5 7
1 1 1
(ii ) 1 .........
4 3 5 7
Solution. (i) We know that
1 x
1
2
1 x 2 x 4 x 6 ........; x 1 1
The series on the right is a power series with radius of convergence 1, so it is absolutely convergent in (-
1, 1) and uniformly convergent in [-k, k] where │k│< 1.
Now integrating (1), we get
x3 x5 x7
tan 1 x c x ............; x 1
3 5 7
Putting x = 0, we obtain c = 0, so that
x3 x5 x7
tan 1 x x ...............; x 1
3 5 7
The series on R.H.S is a power series with radius of convergence equal to 1. However, the series
x3 x5 x7
x ...........
3 5 7
is convergent at ±1.
100 Power Series & Function of Several Variables
1 1 1
1 ..........
4 3 5 7
Example 2. Show that for -1≤ x ≤ 1,
x2 x3 x4
(i ) log(1 x) x .........
2 3 4
1 1 1
(ii ) log 2 1 ...........
2 3 4
Solution. (i) We know that
1 x 1 1 x x 2 x 3 .......... ......; 1 x 1
On integrating, we get
x2 x3
log 1 x x .................;1 x 1
2 3
The power series on R.H.S. converges at x = 1.
So, by Abel’s theorem
x2 x3
log1 x x ...............;1 x 1
2 3
(ii) Put x = 1, in above series we get result
1 1 1
log 1 1 1 ............; 1 x 1
2 3 4
1 1 1
log 2 1 ................; 1 x 1.
2 3 4
Tauber’s Theorem. The converse of Abel’s theorem proved above is false in general. If f is given by
f (x) a n x n , r x r
n 0
the limit f (r ) may exists but yet the series a
n 0
n r n may fail to converge. For example, if
a n 1 , f x 1 x n
n n
n0
Mathematical Analysis 101
x , x 1
n
n 0
1 x x 2 x 3 .............
Then
1
f ( x) , 1 x 1.
1 x
1
f 1 lim f x lim
x 1 x 1 1 x
.
f 1 exists.
However,
Tauber showed that the converse of Abel’s theorem can be obtained by imposing additional condition on
coefficients a n . A large number of such results are known now a days as Tauberian Theorems. We
present here only Tauber’s first theorem.
Theorem 10 (Tauber). Let f (x) a n x n , for 1 x 1 and suppose that lim n na n 0. If
n 0
f (x) S as x 1 , then a
n 0
n converges and has the sum S.
Proof. Let n n k a k . Then n 0 as n . (1)
k 0
1
Also, lim f ( xn ) S where xn 1 . (2)
n n
when
n , xn 1 , f ( xn ) S .
n 0 , f ( xn ) S , nan 0
3 3 3
102 Power Series & Function of Several Variables
i.e.,
n , f ( xn ) S , n an n N . (3)
3 3 3
n
Let S n ak . Then for 1 x 1, we have
k 0
n
S n S ak S
k 0
n
ak S f ( x) ak x k
k 0 k 0
n n
f ( x) S ak ak x k
ax k
k
k 0 k 0 k n 1
n
f ( x) S ak (1 x k ) ax. k
k
k 0 k n 1
n
S n S f ( x) S ak (1 x k ) ax k
k
. (4)
k 0 k n 1
(1 x k ) (1 x )(1 x ........ x k 1 ) k (1 x )
k 0 k n 1
n
f ( x) S ak (1 x k )
k 0
ax
k n 1
k
k
n
f ( x) S ak k (1 x)
k 0
ax
k n 1
k
k
n
f ( x) S (1 x) k ak ax k
k
k 0 k n 1
n
k
f ( x) S (1 x) k ak x
k 0 3n k n 1
n
f ( x) S (1 x) k ak .
k 0 3n(1 x)
1
Putting x x n 1 , we find that
n
1
(1 x)
n
Mathematical Analysis 103
n
k ak
Sn S f ( x) S
k 0 n 1
3n.
n
.
3 3 3
a
n0
k converges & has sum S, which completes the proof.
Theorem 1. A linear operator T on a finite dimensional vector space X is one-to-one if and only if the
range of T is equal to X. i.e, T(X) = X.
Proof. Let R(T) denotes range of T. Let x1, x 2 ,..., x n be basis of X. Since T is linear the set
Tx1,Tx2 ,...,Tx n spans R(T). The range of T will be whole of X if and only if Tx1,Tx2 ,...,Tx n is
linearly independent.
So, suppose first that T is one-to-one. We shall prove that Tx1 ,Tx 2 ,...,Tx n is linearly independent.
Hence, let
c1Tx1 c 2 Tx 2 ... c n Tx n 0
T(c1x1 c 2 x 2 ... c n x n ) 0
and so c1 x1 c 2 x 2 ... c n x n 0
T(c1x1 c 2 x 2 ... c n x n ) 0
c1x1 c 2 x 2 ... c n x n 0
Definition 2. Let L(X, Y) be the set of all linear transformations of the vector space X into the vector
space Y. If T1, T2 ∈ L(X,Y) and if c1,c2 are scalars, then
Definition 3. Let X, Y and Z be vector spaces over the same field. If S, T ∈ L(X,Y), then we define their
product ST by
ST(x) = S(T(x)); x ∈ X.
Also, ST L( X , Y ).
Euclidean space Rn. A point in two dimensional space is an ordered pair of real no. (x1, x2). Similarly, a
point in three dimensional space is an ordered triplet of real no. (x1, x2, x3). It is just as easy to consider
an ordered n-tuple of real no. (x1, x2,……, xn) and refer to this as a point in n-dimensional space.
Definition 4. Let n > 0 be an integer. An ordered set of n real no. (x1, x2,…….., xn) is called an n-
dimensional point or a vector with n-component points. Vector will usually be denoted by single bold
face letter.
e.g. x = (x1, x2,………., xn)
y = (y1, y2,…………, yn)
The number xk is called the kth co-ordinate of point x or kth component of the vector x.
The set of all n-dimensional point is called n-dimensional Euclidean space or n-space and is denoted by Rn.
Mathematical Analysis 105
Algebraic operations in Rn - n-dimensional Euclidean space are as follow:
Let x = (x1,x2,……,xn) and y = (y1,y2,…………,yn) be in Rn.
We define
Tx | x |, x R n , then T .
lub Tx : x R n , x 1
and
1/2
n
x xk2 .
k 1
(i) Also, Let x and y denote points in Rn, then the following results hold:
x 0 and x 0 iff x 0.
(i)
(iii) x y yx
(iv) Cauchy Schwarz Inequality:
2
x. y x y.
(v) x y x y .
xz x y yz .
This follows form (v) by replacing x by x-y and y by y-z. We also have
x y x y
Definition 5. The unit co-ordinate vector uk in Rn is the vector whose kth component is 1 and remaining
components are zero. Then
u1 1, 0, 0, ......., 0
u2 0, 1, 0,........, 0
..........................
..........................
un 0, 0, .............. ,1 .
If x x1 , x2 ,............, xn , then
Te i
Mathematical Analysis 107
Further
Tx Ty T ( x y) T x y ; x, y Rn
So if x y , then
T
Tx Ty ; x, y Rn .
cT c T .
Hence, d is a metric.
Theorem 3. If T ϵ L(Rn, Rm) and S ϵ L(Rn, Rm), then
ST S T
Proof. We have
(ST)x S(Tx) S Tx
S T x
108 Power Series & Function of Several Variables
In theorem 2, we have seen that the set of linear transformation form a metric space. Hence the concepts
of convergence, continuity, open sets etc. make sense in Rn.
Theorem 4. Let C be the collection of all invertible linear operators on Rn.
1
(a) If T C, T 1 , S L(R n , R m ) and S T , then S C.
(b) C is an open subset of L(R n , R m ) and mapping T T 1 is continuous on C.
Thus kernel of S consists of 0 only. Hence S is one-to-one. Then Theorem 1 implies that S is also onto.
Hence S is invertible and so S ϵ C. But this holds for all S satisfying ||S - T|| < α. Hence every point of C
is an interior point and so C is open.
Replacing x by S-1y in (1), we have
( ) S1 y SS1 y y
y
or S1 y
1
and so S1
sin ce S1 T 1 S1 (T S)T 1
We have
S 1 T 1 S 1 T S T 1
. (2)
( )
Thus if f is the mapping which maps T → T-1, then (2) implies
Mathematical Analysis 109
ST
f (S) f (T)
( ) .
Hence, if ||S –T|| → 0 then f(S) → f(T) and so f is continuous. This completes the proof of the theorem.
3.3.2 Derivatives in an open subset E of Rn
In one-dimensional case, a function f with a derivative at c can be approximated by a linear
polynomial. In fact if fʹ(c) exists, let r(h) denotes the difference
f ( x h) f ( x )
r ( h) f ' ( x) if h 0 (1)
h
and let r(0) = 0. Then we have
f ( x h) f ( x ) h f ' ( x ) hr ( h), (2)
an equation which holds also for h = 0. The equation (2) is called the First order Taylor formula for
approximating f(x + h) – f(x) by h fʹ(x). The error committed in this approximation is h r(h). From (1),
we observe that r(h) → 0 as h → 0. The error h r(h) is said to be of smaller order than h as h → 0. We
also note that h fʹ(x) is a linear function of h. Thus, if we write Ah = h fʹ(x), then
A(ah1 bh 2 ) aAh1 bAh 2
Here, the aim is to study total derivative of a function f from Rn to Rm in such a way that the above said
properties of hfʹ(x) and hr(h) are preserved.
Definition 1(Open ball and open sets in Rn). Let ‘a’ be a given point in Rn and let r be a given positive
number, then the set of all points x in Rn such that
xa r
is called an open n-ball of radius ‘r’ and centre ‘a’.
We denote this set by B(a) or B(a, r) . The B (a, r) consists of all points whose distance from ‘a’ is less
than r.
In R1, this is simply an open interval with centre at a.
In R2, it is a circular disc.
In R3, it is a spherical solid with centre at a and radius r.
Definition 2 (Interior point). Let E be a subset of Rn and assume that a ∈ E, then a is called an interior
point of E if there is an open ball with centre surrounded by an n-ball. i.e.,
B a E.
The set of all interior points of E, is called the interior of E and is denoted by int E.
Any set containing a ball with centre ‘a’ is sometime called a neighbourhood of a.
Definition 3 (Open set). A set E in Rn is called open if all points are interior points.
110 Power Series & Function of Several Variables
Note 1. A set E is open if and only if E = interior of E.
Every open n-ball is an open set in Rn.
The cartesian product ( a1 , b1 ) a 2 , b2 .......... .... a n , bn of n-dimensional open interval
(a1,b1),…………,(an,bn) is an open set is Rn called n-dimensional open interval, we denote it by (a, b)
where
a a1 , a2 ,............., an
b b1 , b2 ,.................., bn .
1 1
Gn , ; n N
n n
Clearly each Gn is open set but G1 G2 .............Gn {0} which is being a finite set is not open.
,
Definition 4 (The structure of open sets in R’). In R’ the union of countable collection of disjoint
open interval is an open set in R’ can be obtained in this way.
First we introduce the concept of a component interval.
Definition 5 (Component interval). Let E be an open subset in R’ and open interval I (which may be
finite or infinite) is called a component interval of E
If I E and if there is no interval J I s.t. I J E.
In other words, a component interval of E is not a proper subset of any other open interval contained in E.
Remark 2. (i) Every point of a nonempty open set E belongs to one and only one component interval of
E.
(ii) Representative theorem for open sets on the real line.
Every nonempty open set E in R’ is the union of a countable collection of disjoint intervals of E.
Definition 6(Closed set). A set in Rn is called closed if and only if its complement Rn - E is open.
Remark 3. (i) The union of a finite collection of closed sets is closed and the intersection of an arbitrary
collection of closed set is closed.
(ii) If A is open and B is closed, then A - B is open and B-A is closed.
A B A Bc .
Mathematical Analysis 111
Definition 7 (Adherent point). Let E be a subset of Rn and x is point in Rn, x is not necessary in E.
Then x is said to be adherent to E if every n-ball B(x) contains atleast one point of E.
E.g. (i) If xєE, then x adherenes to E for the trivial reason that every n-ball B(x) contains x.
(ii) If E is a subset of R which is bounded above. Then sup.E is adherent to E.
Some points adheres to E because every ball B(x) contains points of E distinct from x these are called
adherent points.
Definition 8 (Accumulation point/Limit point). Let E be a subset of Rn and x is a point in Rn, then x is
called an accumulation point of E if every n-ball B(x) contains atleast one point of E distinct from x.
In other words, x is an accumulation point of E if and only if x adheres to E-{x}.
If x є E, but x is not an accumulation point of E, then x is called an isolated point of E.
e.g. (i) The set of numbers of the form 1/n (n=1,2,……..) has 0 as an accumulation point.
( ii) The set of rational numbers has every real number as accumulation point.
(iii) Every point of the closed interval [a, b] is an accumulation point of the set of numbers in the open
interval (a, b).
Remark 4. If x is an accumulation point of E, then every n-ball B(x) contains infinitely many points of
E.
Definition 9 (Closure of a set). The set of all adherent points of a set E is called a closure of E
__
and is denoted by E .
Definition 10 (Derived set). The set of all accumulation points of a set E is called the derived set of E
and is denoted by E’.
Remark 5. (i) A set E in Rn is closed if and only if it contains all its adherent points.
__
(ii) A set E is closed iff E E .
(iii) A set E in Rn is closed iff it contains all its accumulation points.
Definition 11. Suppose E is an open set in Rn and let f : E → Rn be a function defined on a set E in Rn
with values in Rm. Let x ϵ E and h be a point in Rn such that |h| < r and x + h ϵ B(x, r). Then f is said to
be differentiable at x if there exists a linear transformation A of Rn into Rn such that
f(x+h)=f(x)+Ah+r(h) (1)
where the reminder r(h) is small in the sense that
| r(h) |
lim h 0 0.
|h|
We write fʹ(x) = A.
The equation (1) is called a First order Taylor formula.
112 Power Series & Function of Several Variables
f ( x h) f ( x) Ah
0. (2)
h→0 h
The equation (2) thus can be interpreted as “For fixed x and small h, f(x + h) – f(x) is approximately
equal to fʹ(x)h, that is, the value of a linear function applied to h.”
Also (1) shows that f is continuous at any point at which f is differentiable.
The derivatives Ah derived by (1) or (2) is called total derivative of f at x or the differential of f at x.
In particular, let f be a real valued function of three variables x, y, z say. Then f is differentiable at the
point (x, y, z) if it possesses a determinant value in the neighbourhood of this point and if
f f (x x, y y, z z) f (x, y, z) Ax By Cz , where | x | | y | | z |,
0 as 0 and A, B, C are independent of x, y, z. In this case Ax By Cz is called
differential of f at (x, y, z).
Theorem 1 (Uniqueness of derivative of a function). Let E be an open set in Rn and f maps E in Rm
and x ϵ E. Suppose h ϵ Rn is small enough such that x + h ϵ E. Then f has a unique derivative.
Proof. If possible, let there are two derivatives A1 and A2. Therefore
f (x h) f (x) A1 h
lim h 0 0
h
and
f (x h) f (x) A 2 h
lim h 0 0
h
Bh f ( x h) f ( x) A2 h f ( x h) f ( x) A1h
which implies
Bh f (x h) f (x) A1h f (x h) f (x) A 2 h
lim h 0 lim h 0
|h| |h| |h|
0
or r2 ( y) y y0 as y y0 i.e., f ( x) f ( x0 )
and r1 x x x0 if x x0 .
Hence
r2 ( f ( x)) f ( x) f ( x0 )
r1 ( x ) A( x x0 ) (3)
2 x x0 A ( x x0 )
114 Power Series & Function of Several Variables
and
Br1 x B r1 x
B x x0 if x x0 (4)
Hence,
r ( x)
0 as x x0
x x0
fi ( x tek ) fi ( x)
( Dk fi )( x) lim
t 0 t (1)
The next theorem shows that A h f x (h) is a linear combination of partial derivatives of f.
Theorem 1. Let E Rn and let f : E → Rn be differentiable at x (interior point of open set E). If
h c1e1 c 2 e 2 ... c n e n where e1 ,e2 ,...,en is a standard basis for Rn, then
n
f x (h) c k D k f ( x).
k 1
But, by (2),
f x e k (D k f )(x)
n
Hence f x (h) c k D k (f )(x )
k 1
Proof. Suppose first that f is continuously differentiable in E. Therefore to each x ϵ E and ϵ > 0, there
exists a δ > 0 such that
f (y) f (x) if y E and y x .
We have then
Since f is differentiable, partial derivatives Djfi exist. Taking components of vectors in (1), it follows that
Conversely, suppose that D jf i are continuous on E for 1 i m,1 j n. It is sufficient to consider one-
dimensional case, i.e., the case m = 1. Fix x ϵ E and ϵ > 0. Since E is open, x is an interior point of E and
so there is an open ball B E with centre at x and radius R. The continuity of D jf implies that R can be
chosen so that
( D j f )( y ) ( D j f )( x) if y B, 1 j n. (2)
n
Suppose h h je j ,| h | R, and take v0 0
and vk h1e1 h 2e2 .... h k ek for 1 k n.
Then
n
f ( x h) f ( x ) [ f ( x v j ) f ( x v j 1 )].
j 1
(3)
n
f s v
j 1
j i h j e j f ( x v j 1 )
n
subtracting h D f x
j 1
j j
n
h ( D j f ) x v j 1 j h j e j D j f x
j
j 1
n
| hj |
j 1 n
n
n j 1
| h | .n | h | | h |
n
Hence f is differentiable at x and f'(x) is the linear function which assigns the number
f ' x h h j D j f x when f'(x) is applied on h. Since D1 f x , D2 f x ..., Dn f x are
j 1
Hence f is differentiable at x and f′(x) is the linear function which assigns the number h j (D i f )(x) to
the vector h h je j . The matrix [f′(x)] consists of the row ((D1f )(x), (D 2 f )(x)...., (D n f )(x)) . Since
(D1f )(x), (D 2 f )(x)...., (D n f )(x) are continuous functions on E, it follows that f′ is continuous and hence
f C(E).
Classical theory for functions of more than one variable
Consider a variable u connected with the three independent variables x, y and z by the functional
relation
u = u(x, y, z)
If arbitrary increment Δx, Δy, Δz are given to the independent variables, the corresponding increment
Δu of the dependent variable of course depends upon three increments assigned to x, y, z.
x 2 y 2 z 2 .
So, if u : Rn R be a function, then u is said to be differentiable at a point x x1 , x 2 ,......... .., x n R n if
there exist constants A1 , A2 ,......... ....., An such that for given 0
we suppose that y z 0, then, on the assumption that u is differentiable at the point (x, y, z),
u u(x x, y, z) u(x, y, z)
A x x
u
and by the taking limit as x 0, since 0 as x 0, we get A.
x
u u
Similarly B and C.
y z
Mathematical Analysis 119
u u u
Hence, when the function u = u(x, y, z) is differentiable, the partial derivatives , , are
x y z
respectively the differential coefficients A, B, C and so
u u u
u x y z
x y z
The differential of the dependent variable du is defined to be the principal part of Δu so that the above
expression may be written as
u du .
Now as in the case of functions of one variable, the differentials of the independent variables are
identical with the arbitrary increment of these variables. If we write u = x, u = y, u = z respectively, it
follows that
dx x, dy y, dz z
Therefore, expression for du reduces to
u u u
du dx dy dz
x y z .
where xi
1/ 2
and 0 as 0 .
Remark 1. If the function u u x1 , x 2 ,......... ...., x n is differentiable at point x1 , x 2 ,......... , x n then the
partial derivative of u w.r.t. x1 , x 2 ,......... ...., x n certainly exist and are finite at this point, because by the
above proposition, they are identical to constants A1 , A2 ,......... .., An respectively.
However converse of this is not true, i.e., partial derivatives may exist at a point but the function need
not be differential at that point.
In other words, we can say partial derivatives need not always be differential coefficients.
The distinction between derivatives and differential coefficients
We know that the necessary and sufficient condition that the function y = f(x) should be differentiable at
the point x is that it possesses a finite definite derivative at that point. Thus for functions of one variable,
the existence of derivative f′(x) implies the differentiability of f(x) at any given point.
For functions of more than one variable this is not true. If the function u = u(x, y, z) is differentiable at
the point (x, y, z), the partial derivatives of u with respect to x, y and z certainly exist and are finite at
this point, for then they are identical with differential coefficients A, B and C respectively. The partial
derivatives, however, may exist at a point when the function is not differentiable at that point. In other
words, the partial derivatives need not always be differential coefficients.
x 3 y3
Example 1. Let f be a function defined by f (x, y) , where x and y are not simultaneously zero,
x 2 y2
f(0, 0) = 0.
If this function is differentiable at the origin, then, by definition,
f (h, k ) f (0, 0) Ah Bk (1)
where h 2 k 2 and 0 as 0.
Mathematical Analysis 121
Putting h cos , k sin in (1) and dividing through by and taking limit as 0 , we get
xy
Example 2. if x 2 y 2 0
Let f (x, y) x 2 y 2 .
0 if x 0, y 0
Then f x (0, 0) 0 f y (0, 0)
This yields
hk
h 2 k 2 , h 2 k 2
h k
2 2
or hk (h 2 k 2 )
m
or
1 m2
m
Hence lim k 0 0, which is impossible. Hence the function is not differentiable at the origin.
1 m2
Remark 2. (i)Thus the information given by the existence of the two first partial derivatives is limited.
The values of f x (x, y) and f y (x, y) depend only on the values of f(x, y) along two lines through the
point (x, y) respectively parallel to the axes of x and y. This information is incomplete and tells us
nothing at all about the behavior of the function f(x, y) as the point (x, y) is approached along a line
which is inclined to the axis of x at any given angle θ which is not equal to 0 or / 2.
(ii) Partial derivatives are also in general functions of x, y and z which may possess partial derivatives
with respect to each of the three independent variables, we have the definition
122 Power Series & Function of Several Variables
u u x (x x, y, z) u x (x, y, z)
a) lim x 0
x x x
u u x x, y, z z u x x, y, z
c) lim z 0
z x z
provided that each of these limits exist. We shall denote the second order partial derivatives by
2u 2u 2u
or u xx , or u yx and or u zx .
x 2 yx zx
u u
Similarly we may define higher order partial derivatives of and .
y z
The following example shows that certain second partial derivatives of a function may exist at a point at
which the function is not continuous.
x 3 y3
when (x, y) (0, 0)
Example 3. Let x, y x y
0 when (x, y) (0, 0).
This function is discontinuous at the origin. To show this it is sufficient to prove that if the origin is
approached along different paths, ϕ(x, y) does not tend to the same definite limit. For, if ϕ(x, y) were
continuous at (0, 0), ϕ(x, y) would tend to zero (the value of the function at the origin) by whatever path
the origin were approached.
Let the origin be approached along the three curves
(i) y x x 2 (ii) y x x 3 (iii) y x x 4 ;
Then we have
2x 3 0(x 4 )
(i) x, y 0 as x 0
x2
2x 3 0(x 4 )
(ii) x, y 2 as x 0
x3
2x 3 0(x 4 )
(iii) x, y as x 0
x4
Certain partial derivatives, however, exist at (0, 0), for if xx denote we have, for example
x x
Mathematical Analysis 123
h, 0 0, 0 h2
x 0, 0 lim h0 lim h0 0
h h
h, 0 x 0, 0 2h
xx 0, 0 lim h0 lim h0 2,
h h
since (x, 0) x 2 , x (x, 0) 2x when x 0.
The following example shows that uxy is not always equal to uyx.
xy(x 2 y 2 )
when (x, y) (0, 0)
Example 4. Let f x, y x 2 y2
0 when (x, y) (0, 0).
When the point (x, y) is not the origin, then
f x2 y 2 4x2 y 2
y 2 2 2 2
x x y (x y )
2
(1)
f x2 y 2 4x2 y 2
x 2 2 2
y x y (x y )
2 2
(2)
while at origin,
f (h, 0) f (0, 0)
f x (0, 0) lim 0
h 0 h (3)
and similarly f y (0, 0) 0.
f (x, y) xy
124 Power Series & Function of Several Variables
f f
is not differentiable at the point (0, 0), but that and both exist at the origin and have the value
x y
zero.
Hence deduce that these two partial derivatives are continuous except at the origin.
Solution. We have
f f (h, 0) f (0, 0)
(0, 0) lim h 0 0
x h
f f (0, k) f (0, 0)
(0, 0) lim k 0 0
y k
where 0 as h2 k2 0 .
hk
Now
h2 k2
Putting h cos , k sin , we get
cos sin
lim 0 cos sin cos sin 0 which is impossible for arbitrary .
f f (x h, y) f (x, y)
lim h 0
x h
(x h)y xy xh x
lim h 0 lim h 0 | y |
h (x h)y xy h xh x
y 1 y
Now, we can take h so small that x + h and x have the same sign. Hence the limit is or .
2 xy 2 x
f x 1 x
Similarly, or . Both of these are continuous except at (0,0). We now prove two
y 2 xy 2 y
theorems, the object of which is to set out precisely under what conditions it is allowable to assume that
f xy (a, b) f yx (a, b).
Mathematical Analysis 125
Theorem 3 (Young). If (i) f x and f y exist in the neighbourhood of the point (a, b) and (ii) f x and f y are
differentiable at (a, b); then
f xy f yx .
Proof. We shall prove this theorem by taking equal increment h for both x and y and calculating 2 f in
two different ways, where
2 f f (a h, b h) f (a h, b) f (a, b h) f (a, b).
Let
H(x) f (x, b h) f (x, b)
Then
2 f H(a h) H(a).
Since f x exists in the neighbourhood of (a, b), the function H(x) is derivable in (a, a h). Applying mean
value theorem to H(x) for 0 1, we obtain
and
f x (a h, b) f x (a, b) hf xx (a, b) h,
where and tend to zero as h → 0. Thus, we get (on subtracting)
f x (a h, b h) f x (a h, b) hf yx (a, b) ( ) h
Proof. Let (a h, b k) be point in neighbourhood of (a, b). Let (as in the above theorem)
2 f f (a h, b k) f (a h, b) f (a, b k) f (a, b).
and
H(x) f (x, b k) f (x, b)
so that we have
2 f H(a h) H(a).
Since f x exists in the neighbourhood of (a, b), H(x) is derivable in (a, a h). Applying Mean value
theorem to H(x) for 0 1, we have
and therefore
Now, since f yx exists in the neighbourhood of (a, b), the function f x is derivable with respect to y in
(b, b k). Applying mean value theorem, we have
That is
1 f (a h, b k) f (a h, b) f (a, b k) f (a, b)
h k
k f yx (a h, b k )
f yx (a h, b) f yx (a, b) ,
where 0 and h 0.
Hence taking the limit h 0 in (1), we have
f y (a h, b) f y (a, b)
lim h 0 lim h 0 [f yx (a, b) ]
h
that is, f xy (a, b) f yx (a, b)
To see that conditions of Young’s theorem are also not satisfied, we notice that
f x (h, 0) f x (0, 0)
f xx (0, 0) limh 0 0.
h
If f x is differentiable at (0, 0), we should have
where h 2 k 2 and 0 as 0.
so we have
2 cos . 4 sin 4
4
2 cos .sin 4
Taking limit as ρ → 0, we have
2 cos .sin 4 0
which is impossible for arbitrary θ.
3.4 References
1. Walter Rudin, Principles of Mathematical Analysis (3rd edition) McGraw-Hill, Kogakusha,
1976, International Student Edition.
2. T. M. Apostol, Mathematical Analysis, Narosa Publishing House, New Delhi, 1974.
3. H.L. Royden, Real Analysis, Macmillan Pub. Co., Inc. 4th Edition, New York, 1993.
4. G. De Barra, Measure Theory and Integration, Wiley Eastern Limited, 1981.
5. R.R. Goldberg, Methods of Real Analysis, Oxford & IBH Pub. Co. Pvt. Ltd, 1976.
6. R. G. Bartle, The Elements of Real Analysis, Wiley International Edition, 2011.
7. S.C. Malik and Savita Arora, Mathematical Analysis, New Age International Limited, New
Delhi, 2012.
U N IT – IV
TAYLOR THEOREM
Structure
4.0 Introduction
4.1 Unit Objectives
4.2 Taylor Theorem
4.3 Explicit and Implicit Functions
4.3.1 Implicit function theorem
4.3.2 Inverse function theorem
4.4 Higher Order Differentials
4.4.1 Choice of independent variables
4.4.2 Higher order derivatives of implicit functions
4.5 Change of Variables
4.6 Extreme Values of Explicit Functions
4.7 Stationary Values of Implicit Functions
4.8 Lagrange Multipliers Method
4.9 Jacobian and its Properties
4.9.1 Jacobian
4.9.2 Properties of Jacobian
4.10 References
4.0 Introduction
In this unit, we study most important mathematical tool of analysis i.e. Taylor theorem. As we know,
Taylor series is an expression of a function as an infinite series whose terms are expressed in term of the
values of the function’s derivatives at a single point. Also we shall be mainly concerned with the
applications of differential calculus to functions of more than one variable such as how to find stationary
points and extreme values of implicit functions, implicit function theorem, Jacobian and its properties etc.
4.1 Unit Objectives
After going through this unit, one will be able to
solve Taylor series expansions.
find the stationary points and extreme values of implicit functions.
understand Jacobian and its properties.
know about the local character of Implicit function i.e. the implicit function is a unique solution
of a function f(x, y)=0 in a certain neighbourhood.
130 Taylor Theorem
4.2 Taylor Theorem
In view of Taylor’s theorem for functions of one variable, it is not unnatural to expect the possibility of
expanding a function of more than one variable f (x h, y k, z m), in a series of ascending powers of
h, k, m. To fix the ideas, consider a function of two variables only; the reasoning in general case is
precisely the same.
Theorem 1 (Taylor’s theorem). If f(x, y) and all its partial derivatives of order n are finite and
continuous for all point (x, y) in domain a x a h, b y b k, then
1 2 1
f (a h, b k) f (a, b) df (a, b) d f (a, b) ... d n 1f (a, b) R n
2! (n 1)!
1 n
where Rn d f (a h, b k), 0 1.
n!
Proof. Consider a circular domain of centre (a, b) and radius large enough for the point (a h, b k) to
be also with in domain. Suppose that f(x, y) is a function such that all the partial derivatives of order n of
f(x, y) are continuous in the domain. Write
x a ht, y b kt,
so that, as t ranges from 0 to 1, the point (x, y) moves along the line joining the point (a, b) to the point
(a h, b k) ; then
f (x, y) f (a ht, b kt) (t).
f dx f dy f f
Now, (t) . . h k df
x dt y dt x y
and
f f dx f f dy
"(t ) h k h k
x x y dt y x y dt
2 f dx 2 f dx 2 f dy 2 f dy
h k h k
x 2 dt xy dt yx dt y 2 dt
2 2 f 2 f 2 f 2 f
2
h hk hk k
x xy yx y 2
2
2 2 2
h2 2 2hk k2 2 f (by Schwarz ' s theorem)
x xy y
2
h k f (a ht , b kt )
x y
and hence, similarly we get
(t) d 2 f ,..., (n ) (t) d n f
Mathematical Analysis 131
Also, ϕ(t) and its n derivatives are continuous functions of t in the interval 0 t 1, and so, by
Maclaurin’s theorem
t2 t n (n )
(t) (0) t (0) (0) ... ( t) (1)
2! n!
where 0 1. Now put t = 1 and observe that
(1) f (a h, b k),
(0) f (a, b),
(0) df (a, b),
(0) d 2 f (a, b),
.....
(n ) ( t) d n f (a h, b k).
It follows immediately from (1) that
1 2 1
f (a h, b k) f (a, b) df (a, b) d f (a, b) ... d n 1f (a, b) R n (2)
2! (n 1)!
1 n
where Rn d f (a h, b k), 0 1.
n!
Here, we assumed that all the partial derivatives of order n are continuous in the domain. Taylor
expansion does not necessarily hold if these derivatives are not continuous.
Remark 1. If we put a b 0, h x, k y, from the equation (2), we get
1 2 1
f (x, y) f (0, 0) df (0, 0) d f (0, 0) ... d n 1f (0, 0) R n
2! (n 1)!
1 n
where Rn d f ( x, y), 0 1.
n!
This is known as Maclaurin’s theorem.
2. If we put a h x, b k y, we get
f ( x, y) f (a, b) ( x a) ( y b) f (a, b) ........
x y
n 1
1
( x a ) ( y b) f (a, b) Rn ,
(n 1)! x y
n
1
where Rn ( x a) ( y b) f (a ( x a) , b ( y b) ) .
n! x y
This is called Taylor’s expansion of f ( x, y ) about the point (a, b) in power of ( x a) and ( y b) .
132 Taylor Theorem
Example 1. If f (x, y) xy , prove that Taylor’s expansion about the point (x, x) is not valid in any
domain which includes the origin.
1 x
,y0
2 y
Also f y ( x, y )
1 x
2 y , y 0
1
2 , x 0
Thus, f x ( x , x ) f y ( x, x )
1 , x 0
2
f ( x h, x h) f ( x, x) h f x ( x h, x h) f y ( x h, x h)
x h, x h 0
x h x h, x h 0 (1)
x , x h 0.
If the domain (( x, x), ( x h, x h)) contains origin then x and x h must be of opposite sign i.e.
xh xh, x x
or x h ( x h) , x x
f x ( x, y ) 2 xy , f x (1, 2) 4
f y ( x, y) x2 3 , f y (1, 2) 4
f xx ( x, y ) 2 y , f xx (1, 2) 4
f xy ( x , y ) 2 x , f xy (1, 2) 2
f yy ( x , y ) 0 , f yy (1, 2) 0
u f x1 , x2 , x3......, xn (*)
denotes the functional relation. In this case if y1 , y2 , y3 ......, yn are the n arbitrarily assigned values of the
independent variables, the corresponding values of the dependent variable u are determined by the
functional relation.
134 Taylor Theorem
The function represented by equation * is an Explicit function but where several variables are
involved, then it is difficult to express one variable explicitly in terms of the others. Thus most of the
functions of more than one variable are implicit function, that is to say we are given a functional relation
x1 , x2 , x3 ......, xn 0
connecting the n variables x1 , x2 , x3 ......, xn and is not in general possible to solve this equation to find an
explicit function which expresses one of these variables say x1 , in terms of the other n 1 variables.
Implicit function
Let F(x1 , x 2 ,..., x n , u) 0 (1)
between the n m variables x1 ,..., x n , u1 ,..., u m and suppose that the set of equations (3) are such that
there are points (x1 , x 2 ,..., x n ) for which these m equations are satisfied for at least one set of values
u1 , u 2 ,..., u m We may consider the u’s as function of x’s.
u p p (x 1 , x 2 ,..., x n ) (p 1, 2,..., m)
where the function ϕ have assigned to them at the point (x1 , x 2 ,..., x n ) the values of the roots
u1 , u 2 ,..., u m at this point. We say that u1 , u 2 ,..., u m constitute a system of implicit functions defined by
the set of equation (3). These functions are in general many valued.
Definition 1 (Implicit function of two variables). Let f ( x, y ) be a function of two variables and
y ( x) be a function of x such that for every value of x for which ( x) is defined, f ( x, ( x))
vanishes identically i.e., y ( x) is a root of the functional equation f ( x, y) 0 . Then, y ( x) is an
implicit function defined by the functional equation f ( x, y) 0 .
F
(iii) The partial derivative (u0 , a, b) 0.
. u
Then there exists at least one function u = u(x, y) reducing to u 0 at the point (a, b) and which, in the
neighbourhood of this point, satisfies the equation F(u, x, y) = 0 identically.
Also, every function u which possesses these two properties is continuous and differentiable at the point
(a, b).
F
Proof. Since F(u 0 , a, b) 0 and(u 0 , a, b) 0, the function F is either an increasing or decreasing
u
function of u when u u 0 . Thus there exists a positive number δ such that F(u 0 , a, b) and
F(u 0 , a, b) have opposite signs. Since F is given to be continuous, a positive number η can be found
so that the functions
F(u 0 , x, y) and F(u 0 , x, y)
Let x, y be any two values satisfying the above conditions. Then F(u, x, y) is a continuous function of u
which changes sign between u 0 and u 0 and so vanishes somewhere in this interval. Thus for
these x and y there is a u in [u 0 , u 0 ] for which F(u, x, y) = 0. Thus u is a function of x and y, say
u(x, y) which reduces to u 0 at the point (a, b).
Suppose that u, x, y are the increments of such function u and of the variables x and y measured
from the point (a, b). Since F is differentiable at (u 0 , a, b) we have
F [ Fu (u0 , a, b) ]u [ Fx (u0 , a, b) ' ]x [ Fy (u0 , a, b) "]y 0.
Since F 0 because of F = 0. The numbers ,' ," tend to zero with u, x & y and can be made
as small as we please with & . Let and be so small that the numbers ,' ," are all less than
1
Fu u0 , a, b , which is not zero by our hypothesis. The above equation then shows that u 0 as
2
x 0 and y 0 which means that the function u u x, y is continuous at (a, b).
136 Taylor Theorem
Moreover, we have
Fx u0 , a, b ' x Fy u0 , a, b " y
u
Fu u0 , a, b
Fx u 0 , a, b Fy u 0 , a, b
x y 1 x 2 y ,
Fu u 0 , a, b Fu u 0 , a, b
0 F u1 , x, y F u2 , x, y u1 u2 Fu u ', x, y ,
and so Fu u, x, y would vanish at some point in the neighbourhood of u 0 , a, b which is contrary to
our hypothesis.
Corollary 2. If F u, x, y is differentiable in the neighbourhood of u 0 , a, b , the function u u x, y
is differentiable in the neighbourhood of the point (a, b).
This is immediate, because the preceding proof is then application at every point u , x, y in that
neighbourhood.
4.3.2 Inverse function theorem.
Corollary 1 is of great importance, for a function of two variables only, F u, x 0 and taking
F u, x f u x , we can express the fundamental theorem on inverse functions as follows:
(ii) f ' u 0
in the neighbourhood of the point u u 0 , then there exists a unique continuous function u x ,
which is equal to u 0 when x = a, and which satisfies identically the equation
f u x 0 ,
in the neighbourhood at the point x = a.
The function u x thus defined is called the inverse function of x f u .
Mathematical Analysis 137
4.4 Higher Order Differentials
Let z f ( x, y ) be a function of two independent variables x and y defined in a certain domain
and let it be differentiable at the point ( x, y) of the domain. The first differential coefficient of z at the
point ( x, y) is defined as
z z
dz dx dy (1)
x y
z z
and if and are differentiable at the point ( x, y) , then the differential coefficient of dz is called
x y
second differential coefficient of z and is denoted by d 2 z and is given by
z z
d 2 z d dx dy
x y
z z
d dx d dy (2)
x y
z z 2 z
2
Now, d 2 dx dy
x x yx
z 2 z 2 z
and d dx 2 dy.
y xy y .
Putting these values in (2), we get
2 z 2 z 2 z
2
dy
2 2
d 2z dx 2 dxdy
x 2
yx y
2
Thus, d z dx dy z
2
x y
3
Similarly, d z dx dy z
3
x y
Thus, the differential coefficient of nth order d z exists if d n 1 z is differentiable i.e. if all the partial
n
n z n z n( n 1) n z n z
n
dy
n n 1 n2 2 n
dnz dx n n 1
dx dy n2
dx dy ........
x n
x y 2! x y 2
y
138 Taylor Theorem
n
dx dy z .
x y
4.4.1. Choice of independent variables
Let F ( x, y, z ) 0 (1)
Differentiate (1), we get
F F F
dx dy dz 0 (2)
x y z
Now, if z is dependent on the two independent variables x and y in such a way that the equation
F ( x, y, z) 0 is satisfied by z z ( x, y ) , then
z z
dz dx dy (3)
x y
Now, equation (2) can be written as
Fx F
dz dx y dy (4)
Fz Fz
Comparing (3) and (4), we get
z F z F
x , y
x Fz y Fz
Similarly, if x is dependent on y and z then
x Fy x F
, z
y Fx z Fx
Similarly, if y is dependent on z and x , then
y F y F
x , z .
x Fy z Fy
z z 2 z 2 z 2 z
We denote the partial derivatives , , , , by p, q, r, s, t respectively.
x y x 2 xy y 2
Now, we suppose that x is dependent variable so that x x( y, z) . Then, we will show that how to
express partial derivatives of first and second order w.r.t. y and z in terms of p, q, r, s and t.
Since z z( x, y)
Mathematical Analysis 139
z z
dz dx dy. (1)
x y
z z z
0 d dx d 2 x d dy
x x y
2 z 2 z z 2 2 z 2 z
2
dy
2 2
dx dydx d x dxdy
x 2
xy x xy y
r dx 2 sdx.dy t dy pd 2 x.
2 2
(2)
1
dx dz qdy (4)
p
p p
pd 2 x r.
1
p 2
2
1
p
2
dz q 2 dy 2qdz.dy 2s dzdy q dy t dy
2 2
r rq 2 2sq 2s 2qr
2 2 t dy 2 dzdy
2 2
dz
p p p p p
r
2 dz
rq 2 2spq tp 2 2sp 2qr dzdy
dy
2 2
2
p p p2
r
d x 3 dz
2 pqs rq 2 tp 2 2qr 2sp dzdy.
2 2
2
dy (5)
p p3 p3
From (4), we have
x Coefficient of 1
dz in (4) ;
z p
x q
Coefficient of dy in (4) .
y p
2 x r
Coefficient of dz in (5) 3
2
z 2
p
2 x 2 pqs rq 2 p 2t
2
Coefficient of dy in (5)
y 2 p3
2 x 1 1 2qr 2 sp
Coefficient of dydz in (5)
yz 2 2 p3
qr sp
.
p3
2w 2w
Example1. Let w be a function of two variables x and y , then transform the expression by
x 2 y 2
the formula of polar transformation x u cos v , y u sin v .
sin v
cos v w (5)
u u v
w cos v
Similarly sin v w (6)
y u u v
2 w w
Now
x 2 x x
sin v w sin v w
cos v cos v
u u v u u v
z z
2
2 z
2
z
2
x a x y
2 2
x y x y
Solution. We wish to express z as a function of x and y where x and y are the functions of r and
i.e. z z( x, y) and given x x(r, ) , y y(r, ) .
142 Taylor Theorem
x x
dx dr d
r
cos dr ( sin .r)d
cos dr sin .rd (1)
y y
Similarly dy dr d
r
sin dr r cos d (2)
Multiplying (1) by cos and (2) by sin and adding, we get
dr cos dx sin dy (3)
cos sin
,
y r x r
z z r z
Now,
x r x x
z sin z
cos (5)
r r
2 2
z z sin z
cos
x r r
2 2
z sin z 2sin cos z
2 2
cos 2
2
(6)
r r r r
z z r z
Similarly
y r y y
z cos z
sin (7)
r r
2 2 2
z z cos z 2sin cos z
2 2
sin 2
(8)
y r r 2 r r
Mathematical Analysis 143
Adding (6) and (8), we get
2 2 2 2
z z z 1 z
2
x y r r
z z
2
z 1 z
2 2 2
( a r ) ( a 2 r 2 ) 2
2 2
(9)
x y r r
z z z 1 z
x y x cos y sin y cos x sin
x y r r
z r z
r cos 2 r sin 2 sin cos cos sin
r r
z
r
r
Squaring on both sides,
2 2
z z 2 z
x y r (10)
x y r
Adding (9) and (10), we get required result
z z
2
2 z
2
z
2
2 z
2
1 z z
2 2
x y ( a r ) a 2
2
x y x y r r
r 2
x2 y2
2u 2u 2u 2u u 2u
x 2
y 2 2 2 4 xy r2 2 r 2
r
x y xy r
sin cos
,
x r y r
u u r u
Now,
x r x x
u u sin
cos
r r
sin
cos u (5)
r r
u cos
Similarly sin u (6)
y r r
2u u
Now,
x 2 x x
sin u sin u
cos cos
r r r r
x 2
y 2 r 2 cos 2 sin 2 r 2 cos 2
(9)
2u u
Now,
xy x y
sin u cos u
cos sin
r r r r
2
r 2 cos 2 .
xy
Solution. Here, x x(r, )
x x
dx dr d
r
146 Taylor Theorem
cos dr r sin d (1)
Since y y(r, )
y y
dy dr d
r
sin dr r cos d (2)
Multiplying (1) by cos and (2) by sin and adding, we get
dr cos dx sin dy (3)
Multiplying (1) by sin and (2) by cos and subtracting, we get
sin cos
d dx dy (4)
r r
From (3) and (4), we get
r r
cos , sin
x y
sin cos
,
x r y r
From (4), rd sin dx cos dy.
Now differentiating, we get
drd rd 2 cos ddx sin ddy
cosdx sin dy d
rd 2 cos dx sin dy d drd
cos dx sin dy d cos dx sin dy d
sin cos
2cos dx sin dy dx dy
r r
2
d 2 cosdx sin dy sin dx cosdy
r2
2
r 2
sin cos dx 2 cos 2 dxdy sin cos dy 2 .
2 2 2 2 2
As d 2 dx 2 dxdy dy , so we get
x2 xy y 2
Mathematical Analysis 147
2
r 2 cos 2 .
xy
4.6 Extreme Values of Explicit Functions
We now investigate the theory of extreme values for explicit functions of more than one variable.
Definition 1. Let u f ( x, y) be the equation which defines u as a function of two independent variables
x and y . Then, the function u f ( x, y) has an extreme value at the point (a, b) if the increment
f f (a h, b k ) f (a, b) preserves the same sign for all values of h and k such that h , k
where is a sufficiently small positive number. If f is negative then the value is maximum and if f
is positive then the value is minimum.
Necessary condition for extreme value
The necessary condition that f (a, b) should be an extreme value is that both f x ( a , b ) and f y ( a , b ) are
zero. Values of ( x, y) at which df 0 are called stationary values.
But the necessary condition that these have extreme values at x a and y b respectively is
f x (a, b) 0 and f y ( a , b ) 0 .
2
The value f (a, b) is an extreme value of f ( x, y) if f x (a, b) 0 , f y ( a , b ) 0 and also f xx . f yy f xy
and the value is maximum or minimum according as f xx or f yy is negative or positive respectively.
Here, A f xx , C f yy , B f xy
Solution. Here, f ( x, y ) x 2 xy y 2 3 x 2 y 1
fx 2x y 3 , f xx 2
148 Taylor Theorem
fy x 2y 2 , f yy 2 , f xy 1 .
2x y 3 and x 2 y 2
1 4
y , x
3 3
4 1
Thus, the extreme point is , .
3 3
4 1
At , , A f xx 2 , B f xy 1 , C f xy 2
3 3
Now, AC B2 4 1 3 0 and A 2 0
4 1 4 1
, is a point of minimum and minimum value f ,
3 3 3 3
16 4 1 2 4
4 1 .
9 9 9 3 3
Example 2. Show that f ( x, y ) 2 x 4 3 x 2 y y 2 has neither maximum nor minimum at (0, 0).
Solution. Here, f ( x, y ) 2 x 4 3 x 2 y y 2
f x 8 x 3 6 xy , f xx 24 x 2 6 y
f y 3x2 2 y , f yy 2 , f xy 6 x
8 x 3 6 xy 0 and 3 x 2 2 y 0
3x 2
2 x 4 x 2 3 y 0 and y
2
4x2
x 0 or y
3
If x 0 y 0
4x2 3x 2 4 x 2 3x 2
If y and y , which is not possible.
3 2 3 2
So, stationary point is (0,0).
Now, A f xx (0, 0) 0 , B f xy (0, 0) 0 , C f yy (0, 0) 2
Mathematical Analysis 149
AC B2 0 0 0
So, doubtful case and further investigation is required.
Now, f f (0 h,0 k ) f (0,0)
f (h, k ) f (0,0)
2h 4 3h 2 k k 2 (2h 2 k )( h 2 k )
k
If h k 0 and 2h k 0 i.e. h k and h2 then f 0 .
2 2 2
2
k
If h k 0 and 2h k 0 i.e. h k and h2 then f 0 .
2 2 2
2
So, for different values of h and k , f does not have the same sign. Hence, f has neither maximum
nor minimum at (0, 0).
Example 3. Find the extreme value of x 3 3axy y 3 ; a 0. .
f x 3 x 2 3ay , f xx 6 x
f y 3 y 2 3ax , f yy 6 y , f xy 3a
x2 y2
y and x
a a
After solving, the stationary points are (0, 0) and (a, a) .
AC B2 9a2 0 at (0,0).
So, given function has no extreme value at (0,0).
Now, A f xx ( a , a ) 6 a , B f xy ( a , a ) 3a , C f yy ( a , a ) 6 a
& A 6a 0
a3 a3
Example 4. Let u xy ,
x y
u a3
y 2
x x
u a3
x 2 .
y y
Putting x= a, y = a
2u 2a3 2u 2u 2a3
Hence 2 3 2, 1, 3 2.
x x xy y 2 y
Therefore r and t are positive when x = a = y and rt- = 2.2-1= 3 (positive). Therefore, there is a
minimum value of u viz. u = 3 .
Example 5. Let
, .
It can be verified that
0, 0 0, 0, 0 0
0, 0 0, 0, 0 2
0, 0 0.
However, on writing
Δ , 0, 0
If the m variables u1 , u 2 ,......... ...., u m are determinate as functions of x1 , x 2 ,......... .., x n from the system of
m equations of (2), then f can be regarded as a function of n independent variables x1 , x 2 ,........, x n .
At a stationary point of f, df 0 .
(3)
Again differentiating the equation (2), we get
1
dx1 ......... 1 dx n 1 du1 ............ 1 du m 0
x1 x n u1 u m
2 2 2 2
dx1 ......... dx n du1 .......... du m 0
x1 x n u1 u m
..................................................................................... (4)
....................................................................................
m m m m
dx1 ........ dx n du1 .......... du m 0
x1 x n u1 u m
From these m equations of (4), the differentials du1 , du 2 ,......... ......, du m of the m dependent variables
may be found in terms of the n differentials dx1 , dx 2 ,........., dx n and are substituted in (3). This way df
has been expressed in terms of the differentials of the independent variables, and since the differentials
of the independent variables are arbitrary and df 0 , the coefficients of each of these n differentials may
be equated to zero. These n equations together with the m equations of (2) constitute a system of n m
equations to determine the n m coordinates of the stationary points of f.
G x dx G y dy G z dz 0. (2)
Putting the values of dz from (2) into (1), or what is same thing, eliminating dz from (1) and (2), we get
152 Taylor Theorem
F x G z G x F z dx F y G z G y F z dy 0
FxGz Gx Fz 0
Fy Gz Gy Fz 0
which gives Fx G y G x Fy 0.
x x x ⋯ x 0
Also x x x ⋯ x 0
x x x ⋯ x 0
…………………………………………………………………….
……………………………………………………………………..
x x x ⋯ x 0
Multiplying all these respectively by 1, , ,…, and adding, we get a result which may be written
x x x ⋯ x 0,
Where ⋯
The m quantities , ,…, are of our choice. Let us choose them so as to satisfy the m linear
equations
P1 = P2 = ………….= Pm.
The above equation is now reduced to
⋯ 0
Mathematical Analysis 153
It is indifferent which n-m of the n variables are regarded as independent. Let them be
, , … , . Then since n-m quantities , ,… are all independent, their
coefficients must be separately zero. Thus we obtain the additional n-m equations
Pm+1 = Pm+2 = ………….= Pn = 0.
Thus the m+n equations f1 = f2 = ………….= fm=0 and P1 = P2 = ………….= Pn=0 determine the m
multipliers , , … , and values of n variables x , x , … , x for which maximum and minimum
values of u are possible.
Example 1. Find the length of the axes of the section of the ellipsoid 1 by the plane
0.
Solution. We have to find the extreme values of the function where , subject to
the equations of the condition
1 0,
0.
Then 0 (1)
0, (2)
0 (3)
Multiplying these equations by 1, , and adding we get
x
x 1 2l 0 (4)
a2
y
y 1 2 m 0 (5)
b2
z
z 1 2 n 0 (6)
c2
Multiplying (4), (5) and (6) by x, y, z and adding we get
or 0 ⟹ .
From (4), (5) and (6), we have
, ,
154 Taylor Theorem
l 2a2 m 2b 2 n2c 2
But 0 2 0 and since 2 0 the equation giving the
r a r 2 b2 r 2 c2
2 2
values of which are the squares of the length of semi-axes required (quadratic in r2) is
l 2a2 m 2b 2 n2c 2
2 0.
r a r 2 b2 r 2 c2
2
Example 2. Investigate the maximum and minimum radii vector of the sector of “surface of elasticity”
(x2+y2+z2)2=a2x2+y2b2+z2c2 made by the plane lx + my + nz = 0.
Solution. We have
xdx ydy zdz 0 (1)
x a 2 x1 l 2 0 (4)
x 2
y 2 z 2 1 a 2 x 2 y 2b 2 z 2 c 2 2 lx my nz 0
1
r 2 1r 4 0 1
r2
2lr 2 2 mr 2 2 nr 2
x , y , z .
a2 r 2 b2 r 2 c2 r 2
2l 2 r 2 2 m 2 r 2 2 n 2 r 2
Then lx + my + nz = 0 0.
a2 r 2 b2 r 2 c2 r 2
l2 m2 n2
2
r a 2 r 2 b2 r 2 c 2
It is quadratic in r2 and give its required values.
Example 3. Prove that the volume of the greatest rectangular parallelepiped that can be inscribed in the
1 1 1
,
ellipsoid .
, ,
Mathematical Analysis 155
Solution. Volume of the parallelepiped . Its maximum value is to find under the condition that it
2 2 2
x y z
is inscribed in the ellipsoid 2 2 2 1 , we have
a b c
Therefore
(1)
2x 2y 2z
df1 2
dx 2 dy 2 dz 0 (2)
a b c
Multiplying (1) by 1 and (2) by and adding we get
x
yz 0 (3)
a2
y
zx 2 0 (4)
b
z
xy 0 (5)
c2
From (3), (4) and (5), we get
a 2 yz b2 zx c 2 xy
x y z
a 2 yz b 2 zx c 2 xy
and so
x y z
Dividing throughout by xyz we get
a 2 b2 c 2 x2 y 2 z 2
2 2 2 1 .
x2 y 2 z 2 a b c
3x 2 a b c
Hence 2
1 or x .Similarly y , z
x 3 3 3
8abc
It follows therefore that u 8 xyz
3 3
Example 4. Find the point of the circle x2 +y2 + z2 = 1, lx my nz 0 at which the function
u = ax2 + by2 + cz2 + 2fyz + 2gzx + 2hxy attains its greatest and least value.
Solution. We have
u = ax2 + by2 + cz2 + 2fyz + 2gzx + 2hxy
f1 lx my nz
156 Taylor Theorem
f2 x2 y 2 z 2 1
Then
axdx bydy czdz fydz fzdy ... gzdx gxdz hxdy hydx
.
Multiplying by x, y, z and adding we get
= -u.
Putting all the values in the above equation we have
hx y (b u ) fz m1 0
.
Eliminating x, y, z and ,we get
= 0.
Solution. We have
(1)
(2)
Mathematical Analysis 157
Differentiating (1) we get
1 b2 c 2
x3 z 2 y 2 dx 0
which on multiplication with yields
1
x b 2
y 2 c 2 z 2 dx 0 (3)
(4)
i.e. (5)
(6)
(7)
= -
Therefore
Similarly and .
which equals to
(8)
Since a, b, c are positive, any one of (5), (6), (7) shows that must be positive. Hence is a positive
root of (8).
158 Taylor Theorem
4.9 Jacobian and its Properties
4.9.1 Jacobian
If be n differentiable functions of the n variables then the determinant
vanishes identically provided that as a function of the u’s has no stationary values in the domain
considered.
Proof. Since
We have
(1)
But
(2)
In other words
(4)
On eliminating the partial derivatives of from the set of equation (4) we get
If v does not depend on y, then and so either or else In the former case u and v are
the functions of x only, and the functional relation sought is obtained from
By regarding x as a function of v and substituting in In the latter case v is constant, and the
functional relation is v = a. If v does depend on y, since the equation v = g(x, y) defines y as a
function of x and v, say
(obtained on multiplying the second row by and subtracting from the first ) and so, either
which is contrary to hypothesis or else so that F is a function of v only; hence the functional
relation is
u = F(v)
Now assume that the theorem holds for n 1.
Now u n must involve one of the variables at least, for if not there is a functional relation u n a. Let one
u n
such variable be called x n since 0 we can solve the equation
x n
u n f n (x1 , x 2 ,..., x n )
for x n in terms of x1 , x 2 ,..., x n 1 and u n , and on substituting this value in each of the other equations we
get n 1 equations of the form
u r g r (x1 , x 2 ,..., x n 1 , u n ), (r 1, 2,..., n 1) (1)
If now we substitute f n (x1 , x 2 ,..., x n ) for u n the functions g r (x1 , x 2 ,..., x n 1 , u n ) become
Then
f1 f1 f
,..., 1
x1 x 2 x n
f 2 f 2 f
,..., 2
0 x1 x 2 x n
...........................
f n f n f
,..., n
x1 x 2 x n
Mathematical Analysis 161
g1 g1
,..., , 0
x1 x n 1
g 2 g 2
,..., , 0
x1 x n 1
...........................
u n u n u n
,..., ,
x1 x n 1 x n
by subtracting the elements of the last row multiplied by
g1 g 2 g
, ..., n
u n u n u n
from each of the others. Hence
u n (g1 , g 2 ,..., g n 1 )
. 0.
x n (x1 , x 2 ,..., x n 1 )
u n (g1 , g 2 ,..., g n 1 )
Since 0 we must have 0, and so by hypothesis there is a functional relation
x n (x1 , x 2 ,..., x n 1 )
between g1 , g 2 ,..., g n 1 , that is between u1 , u 2 ,..., u n 1 into which u n may enter, because u n may occur in
set of equation (1) as an auxiliary variable. We have therefore proved by induction that there is a relation
between u1 , u 2 ,..., u n .
4.9.2 Properties of Jacobian
Lemma 1. If U and V are functions of u and v, where u and v are themselves functions of x and y, we
have
(U, V) (U, V) (u, v)
.
(x, y) (u, v) (x, y)
U U u U v
Then . .
x u x v x
162 Taylor Theorem
U U u U v
. .
y u y v y
V V u V v
. .
x u x v x
V V u V v
. .
y u y v y
and
U U u u
(U, V) (u, v) u v x y
.
(u, v) (x, y) V V v v
u u x y
U u U v U u U v
. . . .
u x v x u y v y
V u V v V u V v
. . . .
u x v x u y v y
U U
x y (U, V)
V V (x, y)
x y
The same method of proof applies if there are several functions and the same number of variables.
Lemma 2. If J is the Jacobian of system u, v with regard to x, y and J′ the Jacobian of x, y with regard to
u, v, then J J′ = 1.
Proof. Let u = f(x, y) and v = F(x, y), and suppose that these are solved for x and y giving
x (u, v) and y (u, v),
we then have differentiating u = f(x, y) w.r.t u and v; v = F(x, y) w.r.t u and v
u x u y
1 . .
x u y u
obtained from u = f(x, y)
u x u y
0 . .
x v y v
v x v y
0 . .
x u y u
obtained from v = F(x, y).
v x v y
1 . .
x v y v
Mathematical Analysis 163
u u x x
x y u v
Also J J
v v y y
x y u v
u x u y u x u y
. . . .
x u y u x v y v
v x v y v x v y
. . . .
x u y u x v y v
1 0
1
0 1
Example 1. If u x 2y z, v x 2y 3z
w 2xy xz 4yz 2z 2 ,
(u, v, w)
prove that 0, and find a relation between u, v, w.
(x, y, z)
Solution. We have
u u u
x y z
(u, v, w) v v v
(x, y, z) x y z
w w w
x y z
1 2 1
1 2 3
2y z 2x 4z x 4y 4z
1 0 0
1 4 2
2y z 2x 6z 4y x 2y 3z
4 2 0 2
2x 6z 4y x 2y 3z 0 x 2y 3z
0.
Performing c1 c1 2c 2
Hence a relation between u, v and w exists.
164 Taylor Theorem
Now,
u v 2x 4z
u v 4y 2z
w x(2y z) 2z(2y z)
(x 2z)(2y z)
4w (u v)(u v)
4w u 2 v2
Solution. Let
u px qy rz
v px q y rz
w ax 2 by 2 cz 2 2fyz 2gzx 2hxy
Therefore
u u u
x y z
v v v
0.
x y z
w w w
x y z
But
u u u
p, q, r
x y z
v v v
p, q, r .
x y z
w
2ax 2hy 2gz
x
w
2hx 2by 2fz
y
Mathematical Analysis 165
w
2gx 2fy 2cz
z
Therefore
p q r
p' q' r' =0
2ax 2hy 2gz 2hx 2by 2fz 2gx 2fy 2cz
p q r p q r p q r
p' q' r ' 0, p ' q' r' 0 , p' q' r' 0
a h g h b f g f c
xy
f (x) f (y) f .
1 xy
u u
x y
Now J u, v
v v
x y
1 1
1 x2 1 y2
0
1 y2 1 x2
(1 xy) 2 (1 xy) 2
xy
f (x) f (y)
1 xy
Putting y = 0, we get
f (x) f (0) (x)
xy
Hence f (x) f (y) f .
1 xy
x y z 0
3 3 3
u, v, w (y z)(z x)(x y)
are u, v, w. Prove that 2 .
x, y,z (v w)(w u)(u v)
x y z
Example 5. If , , are the roots of the equation 1 in k,
ak bk ck
then
(x, y, z) ( )( )( )
.
(, , ) (a b)(b c)(c a)
x y z
1 1 1
x y z
Now, (b c) (c a) (a b)
bc ca ab
x y z
1 1 1
Hence
(x, y, z)
(b c)(c a)(a b) ( )( )( )
(, , )
(x, y, z) ( )( )( )
(, , ) (b c)(c a)(a b)
168 Taylor Theorem
Second Method. After the equation (1),
let a b c (x y z)
ab bc ca x(b c) y(c a) z(a b)
and
1 1 1
,
, ,
( )( )( )
Example 6. Prove that the three functions U, V, W are connected by an identical functional relation if
U = x + y – z, V = x – y + z, W = x2 +y2+z2-2yz
and find the functional relation.
Solution. Here
U U U
x y z
(U, V, W) V V V
(x, y, z) x y z
W W W
x y z
1 1 1
1 1 1
2x 2(y z) 2(z y)
Performing c3 c3 c 2
1 1 0
1 1 0 0
2x 2(y z) 0
Solution. We are given a function V which is function of x and y and therefore it is a function of r and θ.
From x r cos , y r sin , we have
r x 2 y 2 , tan 1 y / x.
Now
V V r V
. .
x r x x
V sin V r sin
cos x cos , x r
r r
V V r V
And . .
y r y y
V cos V r cos
sin sin ,
r r y y r
sin
Therefore cos
x r r
cos
sin
y r r
2V sin V sin V
Hence cos cos
x 2
r r r r
V sin V sin V sin V
cos cos cos
r r r r r r
170 Taylor Theorem
2V sin V sin 2V
cos cos 2 2
r r r r
2V cos V cos 2V
sin sin 2 2
r r r r
Z Z
2
2 Z
2
Z
2
x y ( a x y )
2 2
x y x y
by the substitution x r cos , y r sin .
Solution. If V is a function of x, y, then
Mathematical Analysis 171
V V x V y x V y V
r x r y r r x r y
V V V
r x y x y V
r x y x y
r x y
r x y
Similarly x y
y x
Z Z r Z Z sin Z
Now . . cos (1)
x r x x r r
Z Z cos Z
sin (2)
y r r
2 2 2 2
Z Z Z 1 Z
Therefore 2
x y r r
and the given expression is equal to
Z
2
Z 2 1 Z 2
( a r ) 2
2 2
r
r r r
2 2
Z a Z
2
a2 2 1
.
r r
Example 9. If x r cos , y r sin , prove that
2u 2 u 2u 2u u 2u
( x 2 y 2 ) 2 2 4 xy r2 2 r 2
x y xy r r
where u is any twice differentiable function of x and y.
Solution. We have
u u x u y
. .
r x r y r
u u x u y u
cos sin
x y r x r y
u u u
r x y (1)
r x y
u u u
Therefore r r x y x y
r r x y x y
172 Taylor Theorem
u u u u
x x y y x y
x x y y x y
2u 2u 2u 2 u
2
u u
x2 xy xy y x y
x 2
xy yx y 2
x y
Therefore
2u u 2 u
2
2u 2 u
2
u u
r2 r x 2 xy y x y (2)
r 2
r x 2
xy y 2
x y
2u 2 u
2
2u 2 u
2
r2
x 2 xy y (using (1))
r 2 x 2 xy y 2
u u x u y
Again, . .
x y
u u
x y
y x
2u u u
Therefore x y x y
2
y x y x
u u u u u
x x y y x y
y y x x y x
2u 2u 2 u
2
u u
x2 2 xy y x y (3)
y 2
yx y 2
x y
From (1), (2) and (3), we get the required result.
4.10 References
1. Walter Rudin, Principles of Mathematical Analysis (3rd edition) McGraw-Hill, Kogakusha,
1976, International Student Edition.
2. T. M. Apostol, Mathematical Analysis, Narosa Publishing House, New Delhi, 1974.
3. H.L. Royden, Real Analysis, Macmillan Pub. Co., Inc. 4th Edition, New York, 1993.
4. G. De Barra, Measure Theory and Integration, Wiley Eastern Limited, 1981.
5. R.R. Goldberg, Methods of Real Analysis, Oxford & IBH Pub. Co. Pvt. Ltd, 1976.
6. R. G. Bartle, The Elements of Real Analysis, Wiley International Edition, 2011.
7. S.C. Malik and Savita Arora, Mathematical Analysis, New Age International Limited, New
Delhi, 2012.