Chapter 13

unctions
Functi of Several Variables
aINTS TO BE LEARNT WITH
sUITABLE
Point Sets in two and three dimensions:
EXAMPLES Func
Elementary Topology.
tions. Limits: Double Limit. Repeated limits. Continuity in R. Direc
tional Derivative.
tional Derivative. Partial Derivatives. Sufficient condition for Continuity.
Differentiability-Sufficient condition. Chain Rule. Homogeneous Func
tions: Euler's Theorem. Mixed Derivative: Schwarz Theorem. Young's
Theorem. Jacobians. Functional Dependence. Implicit Functions. Perfect
Differential. d'z at a point.

13.1 Introduction: Point Sets in Higher Dimensions

In Chapter 4, we have already discussed Point sets on the real line (R). Our readers
Rte. already acquainted with such point sets, which we called Linear Point Sets.

1. A point in two-dimensional space is ordered pair of real numbers

an
(c, y),
represented by a point P in the ry-plane.

The sets of all such points in two-dimension forms what call

we a two-dimensional
Euclidean space R2.
2. A point in three-dimensional space is an ordered triple of three real numbers
, ,2). We can represent it
by a point P in our usual ryz-space.
The set of all such
points in three-dimension forms what we call the three-dimensional
Euclidean space R".
3. We generalise this notion and consider an ordered n-tuple of real numbers
can

1,2, 03,." ,Tn) and call it a point P in n-dimensional space. The set of all such
T-CImensional points is called n-dimensional Euclidean space R",
We often write
7 =(r1, 02, , an) and call it a point T ¬ R" with n co-ordinates

739
 DIFFED.
NTIALCALC
ANALYSIS:
TO
INTRODUCTION

AN
740
T1, 12,*** , n (Th is th e kth cotr
n-components
vector d with
T1, T2 , dn Or a of the point x).
atistical and quanr
co-ordinate
Higher
of the dimensional
vector + or kth spaces arise in the study of relativity, Statistiealnics.
occur in
Quantum Mechani
infinite-dimensional spaces
mechanics. Even follows:
defined as
operations in
R" are
4. Algebraic

where
Equality: Two points f and ù

i= (1, 2, T3,** ,Tn)

and y (1,32,V3, ,)

V1, T2 V2***, Tk =
y,
=ý) if zI
= ..

are said to beequal (i.e., ï

Sum: T+ ù = (T1 +1, 22 +y2* ** ,Tn t Un)

Multiplication by real numbers (scalars):

If a is a real number and + ¬ R" then af == (a71,ar2, , axn)
Difference: - ù = (T1 - y1,72 - 2 , * , Tn - Jn).

Zero Vector or Origin: (0,0, ,0) = 0.

The two operations Sum and Muliplkcation bgy a scalar satisfy commutatiw
sociative and distribr.tive laws and there by make R" a vector space over the .
field. We also define:

Inner Product or Dot product

T y= *1y1 +2y2 + ***+TnVn (a Scalar).

Norm or length of a vector e R

The norm |Ë- `|| is called

the distance between T
In fact, the set of all
and y.
points R" in n-dimensional together with
product defines what is known as euclidean a suitable
5. n-space.
Properties of Norm: Let and ù denote
points in E. Then we nave
1.
||20 and ||E|| =0 iff ~ 0. =

2. 1a@|| la||T||, where a is any real

=

number.
3.
2-l =|lJ-T||.
 13: FUNCTIO OF
4 P T E R

SEVERAL VARIABLES
741
1< |F||1|l (Cauchy-Schwarz Inequality).
Hints:(4) can be ritten as

(r1y1+202 t*** * *nbn) S (rf +

ality holds for (1) any one
The equality
ah+.+7)+y2++Y)
kkER. Both (1) and (i)
E R. Bot
of f and ù or both are null vectors or,
may be stated by the single statement " ()andTy are
some

for
jinearly d e p e n d e n t . "

This inequality follows from the fact that

the
quadratic in t given by
ft) =
(T1 +ty1)" + (z2
+ty2)* +..+ (Tn + tyn)
nositive for all real t when T and
s po j are
independent and hence its discriminant
negative, therwise
is negative, othe some real t
would make f(t) 0. This will give
=
the desired
For details consult
neqalality.
more
any Higher Algebra book e.g., Higher Algebra
Ghosh and Maity.]

5. ||7+ s |E||+ I5|| [Triangle Inequality)

T

I17+1 =a+ue =i+2z + k=1 k=l k=l k=l

=
||T||*+2(7 ) +||| [Now using (d)]
IF+2||||+|15 =(|1E||+|10.
We also have

IiF-1|E-
(Deduce it from the Triangle Inequality).
6. There are n unit vectors (i.e., vectors of length = one) in Ez, namely,

=
(1,0,0, ,0), k2 =
(0, 1,0, ,0), .

n=(0,0,0, ,1).

Any vector + = (T1,T2, T3 Cn) can be written as a linear combination of k1,k2,"

n uniquely.
ese n unit vectors are linearly independent and any vector d E E, can be ex-
pre
asa linear combination of i,ü2,*** ,in
and hence {i1,U2,** ,kn} forms what
we call a basis of
E-
 DIFFERENTIAL CAT
742 AN INTRODUCTION TO
ANALYSIS:
iLCULUS
7. Open ball in R
positive real number.
Then the set points e of all R
a) Let a E R" and let. & be a 0 and centre a, denote
Such that | -||< is called an open n-ball of radius by
B(a;6). 0,a +o|. We cal
lr al < d Ja
-

b) In R' this is
simply an open interval -

it a or

of a, denoted
centre a is not
by Ns(a)]. In
case, thetab
aken
consideration, i.e.,(8-nbd
-neighbourhood of a
into Ja-8, a[U]a, a +Q[ then the neighbourhood of a is called deleted

by Ng(a).
6-nbd of a, denoted
a and radius o. We call
)In R,|E- d|| < 6 gives circular disc with centre
a it
a Circular &-neighbourhood of ä. Suppose ä (a,b)
E R° and=
z =
(7,y) E R2 th
then
T-||< Ñgives
(-a)+(y-6) <6* (Fig. 13.1)
the circle ( a)" + (y- 6)* = 2. -

1.e, the set of points that(z,y)

lie within
f the point (a, b) is excluded, then the deleted circular d-nbd of (a, b) is given

by
<8.
(z-a)+ (y- o)
0<

In R we shall also use other neighbourhoods

tectangularneighbourhood: a < r <b and c <y<d (Fig. 13.2) or

Square neighbourhood: Jz - al < and ly-b| < (Fig. 13.3).

AY AY Y

C
b+8
(a,b) b a,b)

b
Oa- a a+ô
Fig. 13.1 Circular Fig. 13.2 Rectangular Fig. 13.3 Square
Neighbourhood Neighbourhood 6-Neighbourhood
of (a,6) of (a, b)
d) In R , | T - | | < d gives a spherical solid with centre at and radius . We call

it a Spherical d-nbd of ä. Suppose ã = (a,b, e) ¬ RS and7= (z, y, z) E R°.

Then ||7-a|| 6 gives (- a)
< + (y-b2 + (z -c) < 8*,
i.e., the of points (z, y, 2) within the sphere
set
(z a)2 +(y- 6)}2 + (z c) =*', -
-

the centre (a,b, c) is excluded then the deleted

spherical nbd of (a, b, c) is written &5
0< (a)+ (y -b)2 + (z -

c <8*.
 H A R
R 13: FUNCTIGNS OF
SEVERAL VARIABLEs 743

Definition of an interior point and an

open set
.a Let S be a subset of R", and assume
that eS. Then d is Called an interor
point ofSif if 3 an open n-ball with centre at
d, nll of whone points belong to S. Every
Doint d ES can be surrounded by an
intes
open n-bal1 Ba,5) C S.
he set of all interior points of S is called
The the interior of S and is denoted by int
h)4 set SR 15 Called an open set if all its points are interior potnts.
In R an open interTval =
(a, b) is an open set, but a closed interval (a,bj 18
s e t be
set because the end pointsa and b are not interior points of the no
val
open
87
ery nbd ot these points contains points of the set la, bl and points outside this set,
hc are in fact the boundary pts).

In R: the set S of points satisfying the inequality (r - a) + (y - b) <0* dethnes

an Open set jan open circle ot radius ó, with centre at (a, b)].

Remember. Note that an

open interval in R' is no longer an open
set when it is considered as a subset of the plane (i.c., a subset of R*).

See that the subset S = {{(a,0): a E R} of the plane Rx R is not an open subset

ofRx R: no point is an interior point as any of its neighbourhoods will contain points
outside the set S.
Note: R is the same R

Properties of open sets

)o(Null set) and the whole space R" are open sets.

(3 Union of any family of open sets (in R") is open

T h e intersection of a finite number of open sets is again an open set.

v) A sct SCRT is open if and only if S = intS.

Proof of (ii). Let F be any arbitrary family of open sets and let S = U A.
AEF
ASSume ES. Then i must belong to at least one of the sets of F, say E A. Since
A IS open, an open n-ball B(T) C A. But A C S, so B(T) C S, and hence + is an

NTErior point of S. Since every point of S is an interior point, S is open.

Proof of (ii). Let S A r , where each A, is open. If S = d, there is nothing

T=1
to prove
See (i)).
 DIFFERENTIAT

CALCULUSs
ANALYSIS:
TO
744
I N T R O D U C T I O N

AN

and hence 3onon

cach set A1,A2, A3, , An en n-balls
ASSme S , then e

B(r, TA)Ax( 1,2,3, n); cach r k .

=

S. That 1s any f E S is:

is an interio
E
t
min.{ri,r2, ,}. Then f E B(7;r)
point of S and so S is open.

union of an arbZUTary ection ofoj open Collection on

13.1.1. We have proved that the sets

is an ets
Note
collection of open sets is
open sets
intersection of a finite
4
an open set and also the
collection of open
sets may or
may
open
However, the intersection an arbitrary
of
Set. Tnot
be open: we give an example in R' e.g.,

each open interval .

1. {0, not an open set, though is an
open set.

which is an opem set.

1(
3.- 0,1 is not an open set.
7

9. Every open n-ball is an open set and every open interval in R" in
n
open.
Proof of 2nd part: Let a,b,(i = 1,2,3, ,n) be real nuumbers such that a; <h
for all i. Then the set of all points =
(T1, ¬2,**. n) E R" such that a; <
aji <b
for
by
i=1,2,3,
the points d
,7, is called an n-dimensional, open interval in R", determined
= (a1, @2,*. ,an) and b = (b1, b2, ,n) of R". We may write this
n-dimensional open interval by the symbol (,b). This is an open set in R".

10. Closed Sets in R"

We shall give two définitions for closed sets and show that they are equivalent.
Definition 13.1.1. A set S in R" is called closed set if its
ifR"- S is open).
a
complement is open (1.e.,

Definition 13.1.2. A set S in R" is a closed set iff it contains all its accumulaton
points.
The set of all
accumulation points of a set S is called the derived set of S,
denoted by S'. Thus S is closed iff S' and
C S.
What is accumulation point of a set S in R"?
an

We shall first introduce a term

callei Adherent
point of a set S.
 cHAPTER 1 2 .

FUNCTIONS OF
SEVERAL VARIABLEs
745
(a) Defnition
of an adherent point
S.
tn S. 7Then f 15 an of a set: Let SR" and let f
ssarily in adherent point of ¬
0
S if
very n-bal B(T) contains at least
one
Clearly,
wiv, if Fif E S then adheres to S because
point of S.
S where every n-ball B(f) contains
here to every ball B(F) T
p o i n t s

These are cali aTe called accumulation points contains point

Contains point of
of S,
S, other than
than the point
of S.
Definition oj an acumulation point of a set: If S CR and 7 eR, not
sarily in SS, then r is called an accumulation point of S, if every n-ball B(T)
necessarily
wntainsat least one point of S, other than . Thus F is
an accumulation
: point of S
E adheres to S- {E}.
i fE S, but r is not an
acumulation point of S, then 7 is called an Isolated po
ofS

Tt follows from the above definition that (The proof is given later as an
exampie)
orem 13.1.1. If T is an accumulation point of S CR", then every n-ball b )
tains nfinitely many points of S.
Accordingly, a set cannot have an accumulation point unless it contains infinitety
many points. But this does not mean that every infinite set must have some accumu-

lation point e.g, in R, the set of positive integers {1,2,3,,}is an infinite set but

it has no accumulation point.

Bolzano-Weierstrass Theorem, however, proves that a bounded infinite set S in R
has at least one point in R" which is an accumulation point of S.
By a bounded set S in R" we mean a set which lies entirely within a certain n-Ball

Example 13.1.1. Assuming that a set S in R" is closed if only its complement R"- S
is open, prove the following:

its adherent points.

(a) If S C R" is a closed set, then it contains all
S is closed.
b) If however, S contains all its adheent points then
Solution: (a) Let S be a closed set and let some point ï adheres to S. We wish to prove

ie S.
R" -S. Now S is a closed set, hence R" S is
-

lt g S then f should belong to

open, i.e., some n-ball B(E) lies completely in R"

-S. Thus B(E) does not contain any
Hence etc.
point of S, then i cannot adhere to S, a contradiction.
We wish to prove that S is a closed
(b) Suppose S contains all its adherent points.
Set. Let i E R" - S. Then S and so T does not adhere to S. Hence some ball
lies in R" S. Therefore R" S is
B() with no point of S and so B(T) completely
-
-

open and hence S is closed.

 INTRODUCTION TO
ANALYSIS: DIFFERENTIAL.
rIaL ALCULUSs
AN
746
sct of all adherent points af c
a set S in R": The
Definition of Closure of
by S.
is called the closure of S, denoted
S adheres to S
and S contains all adherent noi ints
Since every point E S
above that set S is closed then it contai ns all its
if a
thereforeSCS. We have proved
contains all its adherent points then S is. a
adherent points and conversely if S closed
is closed. Hence
only if S
set. Hence SCSif and
A set is closed if and ony ifS = S.
Definition of deried Set: The set of all accmulation points of S C R" is

the derived set of S, denoted by S'.

alled
See that S contains all adherent points (they max be points of the set S or ac
accu-
mulation points of S). We write 5 =SUS'.
We therefore conclude:
A set is closed if and only if S' E S
Or, A set S in R" is closed if and only if it contains all its accumulation points
11. Bolzano-Weierstrass Theorem: Every bounded infinite set S in R2 has at
least one accumulation point. at
We have already proved the theorem in R. We now prove it in R2.
We must produce a point p with the
property that every neighbourbood of p contains
infnitely many points S (because then p would be an accumulation
of
point of S).
Since the set S is bounded we may assume that it is
square Co of side L (Fig 13.4).
completely contained in a

We divide Co into four

equal squares by intersecting lines. One of these
squares must contain infinitely many points of S, for
otherwise S itself
smaller
& Sinite set. would have been
We choose such a
subsquare. We call it C (i.e., C1 contains
of S). We have
C C Co and both are closed and infinitely many points
bounded.

Co

Fig. 13.4
Fig. 13.5
 1
ER 13:
FUNCTIONS
FUNC" OF
(WAPTE SEVERAL VARIABIES 747

a repeat the
process: Divide Ci into four of
W smaller sequares and seleet o
hich cotains imhnitely many points of the set 5. By continuing this, we
in a sequence of closed sqmares {Cn} with C% C1 C2 C and such
t h e m

ench
one ontains infinitely many points of S.
that

ction
3y the bisecti process we see that C, has sides of length L/2". Therefore, by
Nested Set
et Property', 3 a
point p which lies
in all the sets Cn We wish to prove
is A point of accumiuiation of S. Note that p may ot belong to S.
that

Take a neighbo bourhoodo of p. a circle with centre p and radius e. When n is suf
large, say
fciently Jarge, say with L/2"< e/2. the set Cn which contains p must in fact lie in
. Sine
Cn contains inhnitely many points of S, so will the set a. Hence every neigh-
ourhood of b contains infintely many points of S and p is an accumulation point for

s (Fig 13.5).

ertension of the
Not 13.1.2. The proof of Bolzano-Weierstrass Theorei for R" is an

deas used
m treating R2,

13.2 Functions in Higher Dimensional Space

We are acquainted with the meaning of a function of a single real variable. We shall

function of several real variables what comes to the same

extend this idea to define a or

main concern will

thing, a function of a point in Euclidean Space R". However, our

three-dimensional spaces.
be with two and
in an
Definition 13.2.1. Let S1 be of real numbers and S2 a set of pornts
a set
there corresponds a unique w
Eucladean Space R". Suppose that for each pointP¬ S2
value in
7 S. The set of
all pairs (P, w), whee P e S2 and w is the correspondirng
denoted by f. defined on R"
S, 18 called a FUNCTION,

The functional values are denoted by w =

f(P), or equivalently w= f(r. y).w =

n)according to the dimension of the space.

2), w=f(r1, T2, T3,
denotedd
The set of P's for which f is defined is called the domain of the function,
hy
which correspond to sone P in D)
Theof funetional vaues (i.e., the set of w's
set
8 nled the range of f. denoted by R )

three variables
EXanples of functions of two or

fr, y)= r+y is defined over theentire ry-plane.

The Nested Set Property: Let Ci.Cz.C'3,be a sequence of non-emnpty closed and bounded
sets Ci C2 C3 Then 3 at least one point p which belongs to all the sets C.
 CHAPTER 1 3 :
3: FUNCTIONS OF SEVERAL VARIABLES 749

f(a,,
3)
=x?
2y+y'z+zr is a function of three
independent variables a, y, 2.It
.
is
defined for all points (x,y, 2) E R3

Limits of Functions
133

defined on a domain D and let Po be an accumulation point of D. we tneu

for which
fhas a limit A as P~ Po if, for each e> 0, there exists a Q(e, Po)
that

f P ) - A| <e whenever |P Pol <Q; and PeD,PzPo

this symbolically as
We epress

lim f(P) =A or, f(P) A as P Po.

P-Po

of neighbourhood (nbd) we can state thus:

In the language
A For each >0, 3a deleted nbd N of Po
lim f(P) > e
=

A| < 6, if P e DnN'.
|5(P) -

Such that

Simultaneous limit or Double limit

variables:
Functions of two
and Po by (a,b)]
Replace P by (7,y)
variables a and y, is said to tend
two independent
Definition 13.3.1. A function f of 36>0
tends to a given point (a,b), if, to each e> 0,
toalimit A as the point (x,y)
such that
flav)-A<e
to some deleted 6-nbd N' of (a, b). The
of the domain which belongs
or all points (, ) (a,b).
mepuality may not be satisfied when (2, y) =

0< ly b| < Ñ
| -al < ô,
-

N may be a square nbd. 0 <

(r- a) + (y b) -
<82
N may be a circular nbd. like 0 <
or,
Or,
nbd.
or, Nmay be any other

We then write
lim f(,y) = A.
=A or,
lim f(z,y) (a.)(a,b)

A.LT,DCnn
 ANALYSIS:
L CALCULUS
D I F F E R E N T I A L . .

TO
INTRODUCTION

750 AN

Examples on Limits

Example 13.3.1. Let f(P) = f(r. 3 ) 2 +y2

y 0.
Using e-6 definition establish: lim 2
in somedeleted 6-nbd of
me del
We are to find a d >0
s.t.
(0,01
Solution: Let e > 0 be given.
for all points (7, y),
<E, Or,2 2E.

Now clearly,
s+ <+

so that
2+y
if 0< +<0 where = Ve.

Thus, we have found out =

VE to satisfy the requirements of the definition
Therefore,
lim 0.
Example 13.3.2. Establish that limaya .

y0

Solution: Let e > 0 be given. To find a d-nbd of (0,0)(6 > 0) such that in that nbd
N, V(z,y)

eie. E.
But
4l

Clearly, lal < yr?+y, ll< Vr+ and Y<1.

r+y4
.2

+y2
i.e
ie., + <E,
if0< #+<Q8 where = ve.
CHAPTER 2 FUNCTIONS OF SEVERAL VARIABLES 751

requirements of the definition of limit are met and hence the limit exists

qual to
zero.

nd
is

of limit
existence
Non-e

Example
nle 13.3.3. To prove lim-2+2UOES not ezist.
V0

soution: Let f ( , y ) = - +

domain of f is the whole zy-plane punctured at the origin (0,0). For existence
The
imit we are to examne the values of f near (0,0), i.e., in some deleted nbd. of (0,0).

Tf we introduce polar co-ordinates t =

pcos0, , =
psin 0, then f takes the form

p cos6 1
psino sin 26.
sin29.

Here, of course, p= y2 +y?. As (,y) - (0,0),p0.

f is constant on each straight line through the
From this formula, it is clear that
and that, in general, the constant is different on different lines.
Origin (y t tan 6)
=

constant A by simply
Clearly, then, the values off cannot be made close to any
rest..cting p. Thus, lim -
r-0 a2
does not exist in this case.
+

Then have
Alternative Method. u t y ma. we
=

f(,ma) 1+m?
constant values on different lines through the
Again, we see that f has different
(0,0) along a suitable line through the origin, f(7, y)
will
origin; so by letting (r, y) »

approac: any value k specified at pleasure.

one constant A in some nbd of (0,0);
So the values of f cannot be made close to any
mit cannot exist in this case. The limit must have to be unique.

Note 13.3.1. It is easy to see that

=
A,
im f(z,y)
y0

and whatsoever, such that o(z) b as * a then

=o(r), any function of r
,olz)) » A as a~ a.
 752 AN INTRODUCTION TO ANALYSIS: DIFFERENTIAL CALO.
CALCULUs
So, ifwe can determine two functions d1 (r), dz(r) such that the limits ofimultaneous
f(z,
and f(x, d2(r)) as x a are different, then we can conclude that the simultane
limit
lim
(7u)-(a,b)
f(a,u)
does not exist.
Similarly, i f z = v{) be any function such that v ( y ) > a as b and

f(U(v), u) A as
y b, the

im f(7,y) =
lim f(v(y), v) = A

So, to show the non-existence of the limit we choose b1(y), V2() both tend to a as
as
y b but

lim f(eb1(y), y)and lim f(v»(v),v)

are different. See the next example.
Example 13.3.4. The function f, defined over the whole ry-plane, is given by

f(a, y) ={al/, when y#0;

0, when y=0.
Discuss the eristence of limit of f(c,y) as (c, y) -
(0,0).
Solution: On the line z =
0, we have f(0,y) =
0, Vy and the limit is zero as (r, y)
(0,0) along this line.
On any other line z =
ay through the origin we have

flay, y) =-laul/v=-a/
y2
and the limit of this expression as y -0 is easily seen to be 0, by
rule. using (say) L'Hospital's
Thus (r, y)
as -
(0,0) along any straight line through the origin, we have fl,9)
tending to zero.
But see that along the parabolar =y, we have
y2

f) e-/)=e1,
Hence f is always e along a parabola passing through the
lim f(r, y) is not unique and origin and clearly theu
the limit, in
)-(0.0 fact, does not exist.
CHAPTER 13: FUN
FUNCTIONS OP SEVERAL 753
VARIABLES

ratnple 13.3.5. f(z, u) =

Discuss the eristence of the limit of f(r, y) as (r,y) (0,0).

ion: Let us consider the double limit as (r, y)» (0,0). If we approach othe origin
Solun

any axis,
alongany axis, f ( , ) = 0 and the limit is zero. If we approach along any line y=
(7m 0 ) ther

a.ma m

tends to zero as c
fa)=m
» 0.
m2
which
Thus, when we approach the origin along any line y = ma, the limit = 0. But e
asapproach the origin along a parabolic path y = ma2; here also y 0 as

m2 m
But, fz,y) =f(z,mr) A+m?zi1+ m=

T* -

which has different values for different m. Hence the double limit

lim r'y does not exist.

(ag)-(0.0) rd + 12
Example 13.3.6. lim(z+2y) = 5, uhen (7,y) -(1,2). Verify

Solution:

+2-5- l-1+2y -4| -11+2-2= I+1|a -11+2ly-2

<26 +26 when 0 < la 1| <Q and 0 < y-2 < 6 if 0<6S1
-

+2y 5E, wheree is any given positive number,

if40<E, i.e., if ó < e/4 or 1, whichever is smaller, i.e., d < min{{,1}.
Thus, the definition requirements are met and the statement is TRUE.

Repeated or Iterated Limits

Let f(,y) be defined in a certain nbd. of (a, b).

Then lim f(r, y),y remaining constant, when exists, will be different for different
values of y and in fact, this limit will be a function of y, say o(y).
If, then, lim oly) exists and is equal to l1, we write

limlim fr,v)=h
 AN INTRODUCTION TO ANALYSIS: DIFFERENTIAL CA
754
CALCUULUS
i s then called repeated or iterated limit of f(*, ) as a a and then
b.
A change in the order of obtaining limits may produce a change in the final
say l (instead of l,).
Thus we may obtain another repeated limit l2 (say) thus:

lim lim f(x, y)= l2

There is no reason why l1 should be always equal to l2. What is worth notini.
that even when both g is
repeated limits exist and have the same value, the
or double limit may not exist. simultane.
ultaneous
Example 13.3.7. Verify that the double limit

lim
0 T- y
0

does not erist.

Solution: For along

the line= mx, f(1,y)
y =

Thus, by setting (x, y) -y11-mm

value k(# -1)
(0,0) along suitable ine, f{2,y) will
»
a
specified pleasure. Hence f(z,y) docs not tend
at approach anvany
x. y (0,0). to a
unique linmit as
. The double limit does not
exist.
But we observe that
both repeated
different values). (or iterated) limits exist
(although they have

lim lim

Even when both

may not exist. Next repeated limits exist and have the
two
examples illustrate this fact same value, the double limit
Example 13.3.8. Verify that
the double limit

lim
does not erist.
ay2 + ( - y2
0
 .oTER 13: FUNCTIONS OF SEVERAL VARIABLESs
755
C H A P

Solution: The
ion: The double limit does not exist, for f(z, y) y =1,
the
= on

Jiney =
2,2 =

+(t - y)
ay2+(-y) 1 along
= tr but
but f(z,y) == 0 on the lines z = 0 or y = 0, i . e . , l i m f ) =

r ( 0 . 0 J ( T ,
V)
= -r and (z,u)-(0,0) along r =0 or along y =0.
Nevertheless, both repeated limits exist and equal to zero (Check)

Example 13.3.9. f(z,y)= TY

+y
Solution: See that lim lim f(7,y) = 0 ==lim lim f(z, y).
T-0y0 y-0r-0
m
But lim f(T, y) does not exist; for along the line y = mr, f ( 7 , y) = 1
, different
y-0
constant value along difterent lines through the origin.
does not exist«
Here repeated limits exist and are equal but still the double limit
limits do
Example 13.3.10. This is an example uhere double imits ezist but repeated
not erist

+ysin
sin when cy # 0;
fz.v)== {
Sz,y)
when ry =0.
0,

Solution:
Here lim f(c, y) (y remaining constant)
does not exist.
Or, lim f(, y) (T remaining constant)

Remember. lim sin- does not exist.

T+0

and im lim f(z, y) do not exist.

Thus repeated limits lim lim f(r. y)
U-0z-0 t-0y-0
exists and 0.
lim f(z, y)
=

But we see that the double limit

rsin+ysin-o =|rsin+ysins
|

ll+lul in1ins1)
< + + V+
= 2T2+ y2
 756 AN INTRODUCTION TO ANALYSIS: DIFFERENTIAL CAT
ALCULUS
his can be made less than any preassigned positive number e& if

+< (
i.e., if a2+< 8 where 6 = e/2,
i.e., we have a deletec circular 6-nbd. of (0,0) where |f(z,y)-0| < e.

lim f(z, y) = 0.
y0

Thus double limit exists even though repeated limits do not

exist.
Note 13.3.2. What have observed is that
we

repeated limits, if erist, must necessarily be

in case
the double limit
erists, the.
e
equal but the converse is not true. HoUentuo
the repeated limits are not
equal then the double limit cannot erist. vever,
Example 13.3.11.
lim lim ( /l+z\
0y-0 +T lim
T0 (1+o) = -1
and lim lim Y-* ( )1+a 1

See that the

repeated limits exist but are not
1+v =+1.
conclude that double limit cannot equal. We can, therefore, at
exist; this can be easily verified by
once
S(,y) m-1 putting y
m +1
as
(z,y)-(0,0). =
ma:

Another Approach of Limit based on the notion of

limits of sequences
Consider a function
f (z, y) defined is some nbd of
Let the
sequence of points (zn,
(a, 6).
Then the distance Yn)» (a,b) as n o .
between (Tns Vn) and
(a, b) approach zero
i.., (Tn-a) +
(yn- b)* 0 as n > o . -

Now, if
its f(Tn, ¥n) A for
every such
limit, then we say that
point sequence {(zn,
yn)} that has (a,0) as

fr, ) tends to the limit A as

and we write: (r, y)- (a, b)
lim f(r,y) = A.
*a

With such (13.3.1)

a
definition in mind we
prove the
following theorem;
CHAP
13: FUNCTiON OP 8EVERAI, VARIASs1LS 157

13.3.1. In order that f(z, y) -A un (a,/)

orem 13. (a, b) is hoth nsessury and
Tn hat, for eeT}, E> 0, we cun find a uh that
suflicien

ponitine nurnber ii nuchu

f(.u) A| <E, uhen ) < Jza<h (13.3.2)
und 0< y- b6
Denof: The condition (13.3.2) is sufficient enure (13.3.1).
For, if (a Un)* (a, b), then,

0<ln-a<6, Vn > N

0<lyn-b < 6, Vn> N,

both inoqualities hold Yn> N =max {Ni, Na}

From condition(13.3.2), it then follows

f( n)- A| <e when n> N.

f r V) As T .

This leads to the definition (13.3.1).

Condition (13.3.2) is also a necessary consequence of (13.3.1).
If not, there is an e for which no corresponding 6 can be found, such that ior every
null sequence
01,02,3, }
there exists at least one r in lz -

al < ón, say an and there exists at least one y in

iy-bl<on, Say yn, for which |f(n, Vn) - Al2E.
Now the sequence {(n,Yn) (a,b) but the numbers

f(n n)-A|}
have e as a lower bound and cannot from a null sequence.

But this contradicts the hypothesis that flzns Un) A whenever (zns Vn) (a,b)
Hence the necessity of the condition is established.

Theorem 13.3.2. Algebra of Limits

(a) If f(z, y) A as (r, y) - (a,b), then f(r,y) will have the same sign in some
circle (or square) about (a, b).

Take E 2 if A>0 use the definition of limit.

=
-A, if A<0
 758 AN INTRODUCTION TO
ANALYS19: DIPFEREMTIAt , GA.

(b) If f(r, y ) - A, g(r, y) B a s r , y) (a, b) then B6 (1, U) a,

A + B3
S(7, y) + 9(x, y)
f(, u)g(r,y) AB
S(,v)/g(z,y) A/B, if B40.
The proofs are similar to these ofthe corresponding theorens for single 1eal

Example 13.3.12. Show that

iakle
sin (ry - 2) 1
lim
(,u)-(2,1) tan-(3æy - 6) 3

Solution: Given limit =

lim sin(t)
As (r, y)- (2, 1), we
tan-1(3t
t0 tan(3t)
where t = r y - 2.

see thatt ^0 and given limit is of the form

0/01
= lim 1/V1-2
t-0 3/1+9t2

13.4
Continuity of a Function of Several
Variables
Definition 13.4.1. A function f defned in a domain
Euclidean Space) is said to be continuous at the DCR (n-dimensioma
point Po ¬ D if
in
P--Po fP) f(Po).
It is to be
understood that Po is
point a limit of
In e- notation this becomes: D.
a
o(e, Po) for which f is continuous at Po if, to each e > 0, there exiss

whenever P -Pol <ó

SP)-f(Po) <e
and PED.
If is f continuous at every
The concept of point Po ¬ D, then we
uniform say f is continuous in
Ifin the continuity carries D.
for every preceding
pair of definition, to each e >0
over
immediately:
points. P,Po 36(e) such that
|f(P)
uniformly continuous in D.
in D for
which |P - Pol <
f{P}0
-

then we say tna

CHAPTER1 3 a.
FUNCTIONS OF SEVERAL VARIABLES 759

ial Cases: Contimuity in R2

5pecia

Continuit
for a function f of two variables a and u at a point (a, b) in the o1
f definition oJ the function f:
efinition
hat
ssume that (a,b) is a limit point of D so that in any ubd. of (a, b) 3 infinite
mber points of D where f is defined.
o 1p o i n t s

aid to be continuous at (a, b), if to eache>03a tive number ô such

Now
at all oints (r. y) in some ð-nbd N of (a,b)
t h a t

say, in (- a) + (y- b)2 <8

or, i n r al <8 and ly- b|<

weh a v e

S,y)-f(a, b)| < E. (13.4.1)

13.4.1. This condition (13.4.1) is satisfied if (a., b) E D but not a limit poin
Note then f 13
Dtinuous at
i.e., if (a,0) 15 aTm 1s0lated point of D. If (a, b) is a limit point of D
(a,b) if lin f(a, y) = f(a, b).
y

A function f
is said to be continuous in a certain region if it is continuous at every

point of the region.

Points to remember
about Continuity (A -

E)
of two variables a, y is continuous at (a, b), then the function
A. If a function f
constant at b) is
f, b), which is a function of single variable a (y remaining
a
continuous
continuous at t = a. Similarly, fla,y), a function ofsingle variable y is
of f(T, y) at
at y = b. The proof is immediate from the definition of continuity
b in (13.4.1), then to each e > 0,3 such that
a,b); for example put y =

fz,6)-f(a, b)| < e Vz E \t al < . -

f(r,b) is continuous at a = a.

from the illustrations given below.

But the converse is not true as we shall see

and f(a, b) # 0 then there exists a certain neigh-

5, if
f(z,y) is continuous at (a, b)
that of f(a, b) for every
bourhood of (a, b) such that f(7, v) has the same sign as
point (, y) of that nbd. We need only take e
=
a,6)| in the definition (13.4.1).
centre where
Then a square of side 26 about (a,b) as

Sa,b)-fla,b)| < flz,v) < fla,b) + lsa,5)|.

 CAV
ALCULUN
A N A L Y S I S : |

TO
INTRODUCTION

760 AN

of the
square
is positive;
Now, if f(a, 6) is
positive
then every
f(r, y)
square
is
negative. fla,b) i
of the
y)
negative then every (*, then ftg and
region R,
continuous
functions
in a
where g # 0.
are also
C. If f and g are .
at all
points
continuous

and f/g is
continuous;

continuous.
Polynomials are
everywhere
where g =
0.
continuous
except
f/g are
thito
attonal functions
continuous:

continuous
functions
are
sin cy and
functions of
ontinuous
e-+y*)are continuous.

Examples on Continuity

Example 13.4.1. The function

-y where 2+y#0,
f , y) =
where a = 0, y = 0;
0,

is continuous at (0, 0).

Solution: For

l l < Vr2+?Vz? +y? <a

Sz)-f0,0) I=vl =

when +y<0,6= yE. (See Ex. 13.3.2)

lim f,y) = f(0.0) i.e., f is continuous at (0,0).

y0

Example 13.4.2. Discuss the continuity of the function

2.cy where either c or # 0;

f(,v) = T2+2 y

0, where both r = 0 and y =

0.
Solution: This function is
continuous function of either variable when the otaer
a
given a fixed value.
However, f(z, y) is not a continuous function of both r and
along the ine y =t tan v y.
2 tan v
J(T,Y)=Ltand. sin 2, except at (0,0) where
f(0,0) =F0.
 3: FUNCTIONS OF
1APTER SEVERAL VARIABLES
761
T h es u r f a c e z =
f(c,y) is a
cONOID who8e
parametric equations are
=p co8 U, y =
p8in v, z= sin 2.
surfa is generated by a
This line, perpendicular to z-axis,
along this axis so that its which revolves
axis, which revolves about
a n ds l i d e s ,

with the z-axis.

height above the zy-plane is 8in
sin 20 when
when it makes
man
angle
v
an
ect
Theaspect of this surtace near
the z-axis shows
the nature of the remarkabDie dis
of f(,v)
(0,0). at On any circle
2 +?=
continuity

between - and +1, no matter

hetween-1 the 2, function assumes a
v a l u e s
how 6 is chosen.
13.4.3. Prove that the
E x a m p l e

function
f,y)= (, ) # (0,0),
0, (,) = (0.0);

is continauous
at (0,0).

Solution: We can prove

imf(z,y) =0= f(0,0).
V0

cy
ellu v T +
ie,<r2 + < e (where e is ay preassigned positive number) if 2 +y < 6,
where = E.

Thus given any e> 0,38 >0

, y ) -f(0,0) <e v(«,v) ¬+f< where 6 =¬.
flz,y) is continuous at (0,0).
Alternative Method. Let = rcos 0,y = r sin.

Then -0 r|cos6sin6| r =
yr?+y' <e
yr2+2

ie, if al<ul
V r 4 U=f0,0).
lim
y0

Hence f is continuous at (0,0).

 AN INTRODUCTION TO

762 functio7
the
that
Prove
13.4.4. when t # ,
Xample

r . y ) = - y
whenr =Y:
0.

s not
continuous at (0, 0).
not
exist
and as suchch fl,y) cannot
does

shall prove
lim f(x, y)
Sohution:
We 0

continuous at (0,0).
m r " _1+ (1- nr2)3
mr2)3
+_r+(T-
m
f.y)=
-Y
putting y
= t
-
en as z
mx"; then -0,y0)
1+(1-mr"_.
m
lim f(z,y)= lim m

y0

different choices of m (i.., the limit

limit can o.

different for
So that the limit is
at pleasure, not unique).
become
any arbitrary value k specified
is not continuous at (0,0).
the limit does not exist andf

D. We know that a function f is said to be continuous in a region (open or

(open or close
if it is continuous at every point of the region. A tunction f, continnous
in &
closed region R has properties analogous to those of continuous functions one
variable in a closed interval (a, bj:

A functionf of two variables r and y is said to be

R, open or closed, when for any
unitornly continuous in a regin
e> 0, we can find d > 0 so that if
(T1, y1) and (z2,
are
any two points of R for which |z1 - #2l < d and ly1-y2< ð then

f1,V1)- f(#2, v2)| <e.

A function f of two
variablesr and y which
R is is
uniformly continuous there, that continuous
1s in
in aa closed
closed icregut
R,f(1,n) -f(z2y2)|<e when (1,yand (r2.y2) are p is, if ints «

arbitrary preassigned lr1 t2l <Ñ -

positive number and ó and Iy1-y2<0 2<o wi where eisa

A function
f(r,y)
depends one alone.
which is
takes its
least and continuous in a closed
greatest values at least region oundedthere
R, is bou
onmce in the repion