Metric Spaces

Definition

Let E be or
nonempty set .
A distance on E is a
function ed : Ex E
→ IR that

satisfies the
following properties
:

1 dex x ) =o t x E E and d I x ) t E (positivity)

.
, , , -1 > o
,
x
, y E
,
X
t y

2 d (x ol Fx E (symmetry)
.

,
-1 ) =
ly
,
x
) , , >
E

3 .
d I x
, -1 ) E old 't ,
) +014,7 ) t x. y 't E E (triangular inequality)
, ,

If d is distance the ( E d ) is called metric and its

a on E
, pair ,
a
space

elements are called points .

direct
A
consequence of the
triangular inequality is

4 .
014 xn ) , ,
e DX,xz
, ) + 0142,43 ) t . . .
t
d kn
-
I
,
xn ) ,
t ×
,
,×2 ,
. . .

,
Xu C- E

Idk 014,7) I

ol
Vx E
5 .
,
t ) - E
(x , y ) ,
, -1 ,
Z E

Indeed have
,
for any
x
, -1
,
z E E
,
we

d ( 014-1 ) +014,2-3 d Hy ) 014,7

x
,
t I s
,
⇒ 014,7 ) - E )
and

dfz , ) ,
E d C 2-
,
x
) t 014,7) ⇒ -
dixie ) E 014,7) -
0117,71

Hence - dlx
, y ) E dixit ) -
014,7) E 014,7) ⇒ I dlx ,
7 ) -
01177 ) I Ed Lxii )

Notice that
if (E ,
ol ) is a metric
space ,
and
if S is a subset of E
,

then (s d ) is also a metric

, space .

We will call
(s ,
d) a metric
subspace of
the metric
space (E ,
ol ) .
Examples of metric spaces

ol
1 .
: IR x IR IR
defined by 014,7 ) =
I x -
71 ,
×
, -1
E IR .

is distance IR Hence ( ol ) is metric ok 7) =/ yl

IR
space
a on x
or
x.
-
.
.

is called the natural metric of IR .

d olcx y I where Ix the modulus

defined by
is
2 . : axe IR
, > I = Ix -

,
-

yl

metric
of the complex number
,
is a distance on a .
Hence Ca ,
d ) is a

space
.
014,7 ) = -

y I is called the Ixnatural metric of a .

that continuous
3 . Let Cla ,
b ] be the set of all
functions f : Ca ,
b ] → IR are .
The

function ol : Cca ,
b) a Cle ,
b ] IR
defined by dlf , g ) =
sup got
If-

,
is a

[ a
,
b ]

distance ( metric ol called

on Cca ,b ] .
Hence cc a
.bz ,
ol ) is a
space
. is
usually

the metric
of uniform convergence
.

4 .
Let E be an
arbitrary set . The function ol : Ex E IR
defined by

dlx ,
-1 ) = I
if xty and dlx
,
x ) = o
mortifies the three
properties of

a distance .
ol is called or discrete metric
 Normed Vector Spaces

Let E be a vector
space
over the
field of reel numbers IR or over the
field of

complex numbers a .
He will use 1k to denote either IR or a .

Definition

satisfies
A norm in the vector
space
E is a
function y
: E IR that the
properties

I .
4 H ) 3 0
,
it x E E and
4K ) = o x = O (Positivity & Separation)

2 .
y ( Xx ) = 121 4 C x ) ,
it X E E
,
F X Elk (Homothety)

3 .
4 ( xty ) E 4 ( x ) t 4 ly ) ,
Tx
, -1
E E (Triangular Inequality)

the called normed vector

If y )
is E is
y a norm on
, pair ( E
,
a
space .

The elements of a normed vector

space will be called vectors .

It is
customary to denote the norm
y by the
symbol 11.11 .
with this notation

4 30 is written as 1141130

Xx ) 1714 is written 117×11=17111×11

41 = as

4 I Xt -1 ) E 41×7+4 L y ) is written as
11×+711 E 11×11+11-111

consequences
direct have
As
of properties 2 and 3
,
we

It + + Xn 11 E It x
11+11×211 + + llxnll V Xi Xz Xn E E
g
. . . . . -

, . . .

,
.

, , ,

5 Il x -111 E 11×11 11-111 H x E E

11×11=11*-17+-111
t
-1
-
.
, ,

G
I 11×11 It -111 I E 11 x
y
11 Fx C- E
y
. - -

, ,

establish last
To the
property write

E 11×-711 t " T " " × " -

" t " ' " × -

" "

11-111 = 11 x -
( X -
T ) 11 E 11×11+11 x -
-111 11-111-11×11 E 11×-711
} 1,1×1 , -11,111 s n × -

yn
Examples of Normed vector spaces

1 .
The function y
: IR →
IR
defined by 4*7=1×1 , defines a norm on IR ,

considered
considered vector 1k IR
as a
space over = .

2 .
The
function y : a → IR
defined by ylx ) = modulus ( x )
, defines or norm

on a vector 1k a
,
as a
space over =

check that
3 . It is
easy
to Cla ,
b ]
,
the set of all
functions f : Ca ,
b ] → IR that

continuous vector the usual addition scalar

are in ca ,
b ]
,
is a
space for and

multiplication of functions .
The function Hill : Cle ,
b ] IR
defined by

11 f It =
sup I fall , f C- C Ca
,
b ]
C or b ]
,

is Cca ] Hence ( cc ,b3 H H ) is normed vector

norm b
space
a on . a .
a .

, ,

4 . The function Hill : Cle ,

b ] IR
defined by

It fell =
?
Jo If I dx , f E Cca ,
b ]

is Cca ,b ] Hence ( cc ,b3 H H ) is normed vector

norm
space
a on . a .
a .

Let
integer
"
5 . n > I be an .
IR is the set of n -

tuples cxyxs ,
. . .

,
xn )
,
where

" "
EIR j The
of X C Xi ) EIR and ( EIR
Xj is En sum Xz xn
y T Tn )
= =
-12
. . -

,
.
,
. . .
, , , , , ,

defined by multiplication
"
is xty =L a ty xztyz .
. .

, xnttn ) EIR .
The scalar
, , ,

defined by
"

of
"
A IR is
by ( )
C E Xx Xx 7×2 a EIR
x = x ,
,
xz ,
. . .

,
xn ) EIR = ,
, ,
. . .

,
xn

With
operations field
"
these two IR becomes vector the IR
space
a over .

function defined for

"
The "
• It .
11
,
: IR IR x = C x
,
,
Xs ,
. . .

,
xn ) C- IR
by
n

11×11 = I x
, I +1×21 + . . .
t I xnl = I lxjl
is or norm on IR ?
,
j=i
 function defined for
"
The "
• 11.11 : IR IR × C x x xn ) C- IR
by

needs
=
,
, , ,
. . .

,
A

11×11 I is IR ?
N
= max
xjl a norm on

is jen

function

112
defined for
"
The "
• It .
: IR IR × = C x
, Xs . . .

,
xn ) C- IR
by

12+1×212-1
11×112
-21%12
)¥=
, ,

' "
=
(I x. . . .
+ lxnl ( ) is a norm on IR ?
j=i

nontrivial
The
property to establish
,
is the
triangular inequality .
To
prove

it
,
one the Cauchy - Schwarz - Buniakovskii inequality
"

II
"

II
'

lxjllyjl s ( lxjl ) l Hill ) , V-xj.yjc.IR ,

5=1,3 . .

,
n

, ,
,

To
prove
the inequality start with

&

JI ( lxjlttlyjl ) 3 o
,
Tt EIR
,
it x
,
-

EIR
,
it
Tj
EIR

to
Expand get

jI +24¥ IIIHjk )
'
'
lxjl ) t lxjlltjl ) t + so . . .
#
,
,

'
that
Recalling for a
quadratic at + btec to
always be
nonnegative ,
its discriminant

'
b -
aac must be so .

Applying it to 1*7 ,
we
get

KII II.
's IIin II. IItill
'
's
till
kill ) -
a mil " so ⇒
till
kill
II's mil '

both sides leads C.

inequality
Taking the
square
root
of to S B . .
 Convex Functions
Definition

Let I be an interval and let

f : I → IR be a
function defined in I .

lxle
say
that f is convex or I if t x
,
,
xz E I
,
we have

f (X x
,
t 4
- X 3×2 ) E X
fix , ) to
-
x ) f ⇐ z ) ,
FT E Co , IT

Geometrically ,
this means the point " ""

( Tx t t Ha f ( xx tf a) xz ) ) the
graph of f
lies
-

, on
-
•
,
,
•

•
.

the point ( below

Xx , ti a) xz
-

,
a fix , ) taH -
fan ) on the line : .

b
X Xq
a
,
x XX tl7) Xz
-
.
-
,

segment
with
endpoints C x ,
,
fix , , ) and
42fixes
,
)

The
following is an easy test to determine when a
differentiable function is convex .

Theorem

Let
f : I IR be
differentiable in I .
Then
f is convex on I

'
and
if only if f is
increasing
in I .

Proof '

Let
suppose f is convex on I . us show that f is
increasing .

Let x
,
ex
,
eI . we want to show that ffx , ) c
f # z ) .
Let x
,
s
xcxz .

f- have
Then x = Xx , + I I 7) xz
-

for some X C- ( 0,1 ) .

being convex
,
we

f( x ) =
f I Xx , th -
x ) xz ) E 7 f- C x , ) ti . a ) f . . . CA )

both sides
subtracting fcx ) from ,
of CA ) ,
we
get

f- ex ) -

fix , ) EH1) -

f 4,7+47) fcxz -
) =L
I-

7) (fcxz)
) fix -

sides
subtracting ) from both
fine of 1*7 ,
we
get
 fix ) -

fcxz ) E X fix , ) -

Xfc ) = - X ( fixe ) XDf- -

It
follows (f ))
f XI -

fi , ) fix ) -

fkn ) ( l 7) -
FK
-

, fcxz
) fix -

,
)
= s =

X -
Xi (I - X ) ( Xz -
×
,
) (i -
a ) ( xz x
-
) Xz -
Xi
,

f- CX ) -

ftxz ) -
xffcxz
, fly -
, ) -
x If
) ca -
fix ,
) )
, =

x -
Xz X Xz Xz )
-

( X
× ,
-

the limit
Kassing to
,
we
get
f I ) fix fling f f 1×2 )
fix ) fix
¥
X ) ) ) ( x ) flxz fix )
)
-

him
,
and him
-
,
-

f
,
-

, = e z ) =

? X × '
Xi 42 A Xz Az
×
×→×I
-

x
- - -

→ x i

fix
'

Hence
,
we have
proved that it
x. cxzt
I
,
, ) E f Cx
2)

'

Let the f is
suppose
I
increasing
converse
prove in
us .

Let x
cxz E I let X Elo ' ] and set x= Xx , + ( I 73×2 -

By the MVT we have

, ,
.

, ,

0 Oz

;!
,

=f' to ) f
II floe )
⇐×
-177102
f- HI -
f HI " and
' '' ' ' " )
,
I I
-

= = -

X - x
,
X X
, Xz
'

't
'
I
since
f = , f , ) E f coz ) .
It
follows

44=7*4
fix fo
t
f I
" '
-

' "
e
Xz - X

Multiplying by ( x - x , ) ( xz - x ) > o
,
leads to

( x
,
-
x ) fcx ) - ( xz - x
) f- Hi ) E (x -
x , ) fcxz ) - Cx - x
, ) ft )

which (X fez else

we rewrite as
( ×
,
- x + x -
x
, ) f Cx ) E 1×2 - x ) ffx , )
+ -
Xi ) ) or

Cxzx ) -
, fix ) E ( -
TX , -17×2 ) f Hi ) + ( X 1) - x
,
th -
7) Xe ) f Hz ) = ilxz -
x
, ) FH ) t l I -
7) HEX ) fcxz , )

fix and
Hence f C x ) E 7 , ) + C I -
x ) f cxz ) so
f is convex on I

Consequence

If f is two times
differentiable in our interval I
,
then f is convex on I

"

and
if only if f as so ,
it x c- I

If
"

(f) If
'

f is convex on I
,
then f is
increasing
in I
,
therefore must be

" '

positive .
Conversely , if f ex ) 30
,
tx c- I
,
then f is
increasing
in I .
Theorem

Let f be a
function that is convex on our interval I .
Then t x
,
,x2 ,
. . .

,
xn C- I
,

we have f (
,
xjxj ) e

j
xjfcxj ) ,
3 o such
taj that
xj=i
II.
Proof
For x
, Xz Xz E I and X
, 30 7230 7330 such that X +72+73=1 we have
, , , , ,
,

I
T
(( ) I
I

2×2+73×3
2×2
fax , +72×2+73×3 ) =
f x
, -172 ) .
'
E , H-172 ) f ( '
) +73 flxz )

E Hit T2 )
fly ,
, ,
At
, , ,
×
2) +
13 f l ×3 )

E 2)
Hitt
( , ,
, ,
f l Xi ) +

, , ,
fixe )
) t 73 f I ×3 )

E X , fix , ) t 72 f l ta ) t 73 fcxs )

The done induction

general case can
using .

Holder Inequality

II.IIIxil.rs
*
Let and be such that '
to =L For Xi C- IR
psi g > I
any Xz he y In
-
t
-12
.
. . . .

,
.

, , , , ,
p ,
, ,

" "
' " e have s
II.
'skill . 's
i i

Proof
"
Consider As result
the
function f CN = x n since
p > I
, f is convex on Co , + a
) .
a
,

for any so and

for
ajc any Tjco ,
to ) ,
is
jen we have

E. naan E
nai
"
a÷÷ is
.

n .

n
n

Zai
Factoring
out -2 an and
rewriting it as I a
j ,
we
get
k = '
j = I
y
p y
-

(
p
2 Tj aj ) I Tj aj

Ii j"
"

§
'
'

af
j=
( aj aj ) ( )
i
s s I Tj ) Tj
n n
p ,

( I )
,

a
j
j=i j ,
-

-
 pH sides
Taking the root of both
,
we obtain

ajajl
TI II
"P " "
a
ja ; s I a
j ) . . .
#
, ,

p
af
q
select and that and
so
if we
xj aj such Tj = It
;
I Dj =
lxlj
9
I
9
-

lxjl lxjl "

then and
aj Xjaj =L
Ij I
=
lxjl I Yjl =
lxjl I
Yjl
-
=
, o
q" "P
,
I
-1J I
P
lyjl

"

lxjllyjl.IE?ixjiP
II II.
" " "
"
④ becomes ) ( it
,
)

Remark

Holder becomes
When
p=2 , inequality

II)kik
" .rs
"
'
II " l ' ' it s c is

and called the SchwarzBuniakovskii

is
usually Cauchy - -

inequality

Minkowski Inequality

such

HIIt.IE
Let p > I
, q > I be that
¥ + I = I . Then
for any Xj ,
Yj
E IR
,
is jen ,
we have
,

"
I
" " " "
's
''
ix. in
.
+ 'snot hot
.

Proof

lxityi
Hit
Kil
Yi
" "

lxityil
Til
ly
it
P I
" " '
-

xillxityil
xityil
lxil
hit
til
!
YE
)
I He have I I I E I I

hit
=
-1 x t
, ;

.fi?.ixityii
Adding and
using
Holder
inequality ,
we
get

÷ II
" " "
I I
Yi I
I
E +

, ,

" "

=L
" " "
" "
( II
" " "

) 't
" " '

.
) )

IIIt)
" " " " " "
)
'

(( )
" P
e
.
)
 P '
But
¥ t to
,
⇒ ⇒ to
,
= I -

To =

j ⇒
q
=

tf ,
⇒ lp -
is g =p . It
follows

Hittite
lyin
lxil
tilt
#yil
"
't

HE II.
" "P
Is ) I ) ( Ee )
"
)
.

I
, ,

'
¥
(¥
'

)
"
leads the result
Dividing by I to .

Theorem
"
Let The function It IR
Ilpdefined for
"
EIR
p I : IR x C x Xs xn ) C- IR
p
> .
.
= , . . .

, , ,

1%1×214
Hxllp
,lxjlP
help
! )
" " P
( ( ) IR ?
by I is
=
Ix ,
. . .
+ a norm on

We will call it the

p - norm .

Proof

• 11×11=0 ⇒
,lxjlP"P=o
) ( ⇐ Kj 1=0 , iejsn ⇐ 4=0

"P
1214¥ )
""

4¥
" " "

• 112×11 =
lxxjl ) = lxjl =
# 11×11 . ,
it X EIR
,
it x EIR
,
,
,

"

2
" "

=L II
" " " "
" "

Hxeyllp lxjtyjl ) lxjl I C Hjl )

• e + =
nxnptnynp
, ,
,

"
with
In all that
follows ,
we assume that IR is
equipped a norm It . 11

Definition (Open and Closed Ball)

"
Let c C- IR and

Bcc
r > o .

I } called ball
"
The set ) =/ x E IR 11 all er is the
open with

centre
r x -
c
,

and radius r .

Bcc
Her
=/ I } called closed ball
"
The set ,
r ) x e IR 11 x - c is the with c

and radius r
.

set
Ix I called sphere and
"
The Scc ,
r ) = EIR u x -
all = r
} is with c

radius r
 "
n
'

Assuming 11×11
11 =
x
112 =
( I Hit )
i . i

then this
•
If n=1
,
X
,
C C- IR and 11×11=1×1
,
Il x -
all = Ix -
cl . In case the

ball closed ball and center and

open ,
the the sphere with c radius r > o are

( c - r
,
ctr ) ,
( c - r
,
ctr ] and I c- r
,
ctr
} , respectively .

I • I

c- r C
Ctr

• If n= 2
,
then x = C x ,
,
xz ) ,
e = Ca
,
Cs ) and 11×11=1×7+17
,
H x -
ell = ,/¢,-DIHz
'
the the circle 2+1×2 r2 the ball
In this case
,
sphere is (x ,
-
a) -
ca ) =
, open
is

the
region
within the circle and the closed ball is the circle plus the
region
Xz
enclosed by it .
r >

Cz . . . - . . . . . . . .

c , µ a) 2+1×2
- -
Cz } =P
:

C
, X
,

then and / Xf -1×22-1×32

•
If n =3
,
X = l x
,
,
x2
,
x 37
,
C = C a
,
Cz
,
Cz ) 11×11 = -

✓ sphere
2
" x -
Cll =
Cx
,
-

a) 2+1×2 -

Cz , 2x I xz -

g) .
In this case the is the usual
,

't
2

sphere X a) 2+1×2 r ? the ball is the within the

ca ) Ix G) =

open region
- -

,
-

enclosed
sphere and the closed ball is the
sphere plus the
region by it .

Definition
"
said
set each
point ?
"
to
A be
if
u IR IR
for
in is in e EIR
open ,

there exists r > o such that Bcc r ) C 0

,

The interval I o I ) is in IR while the interval [ o I ) is not

open
.

, , ,

The set '

xihtxz 1122 but not
I ,xz ) EIR I 4
} is in
} 11221 } is
a
ex l L Ix xg ) x x
open E 30 ? o
, ,
.

, , , ,

The 1123
I IR ?
set
} Lxi
,
xz
,
xz ) E x
, > o
} is
open
in
Theorem (properties of open sets)

" "
entire
•
The
empty set of and the IR are
open in IR
two

"
intersection
•
The
of a
finite number
of open
sets is
open
in IR

of any number
finite infinite of sets
"
The union IR
open open
• or is in

For Y and such that

"

there sets
"
b E IR exist IN two in IR
open
at
any
•
,

and A txt
at Y
,
b E txt Y =
of .

Proof

the
of it doesn't contain element
By definition empty

0=0
set

02
-4
Oz
Uz
is
open any
• as .

ball
" "
the
"

( )
for any
IR is since IR B I C IR
open
at a
open
.

, , ,

Let two sets

"
and let
be
If
-

in IR h a e U
open
.
•
,
,

then and Because 0 and there exist

- -

U are
a e E
open
a
, . , ,

Oe
Ve
Ok
-4
such

Oz
that
Letting

Cuz
Bla 0 and Bla
r
,
so
,
R2 > o
,
r
, ) C
, ,
rz ) C .

02=0
have Bla ) Blair ) C 0 and Bca r ) C Bla re )
mingy
r C
re re } ,
we
, , , , ,
,

Hence Blair ) C 0 A .

collection let
'
set
•
Let
, ,
. . .
,
,
. . . be a
of open
in IR and

¥ If for
0 then
c- some l Now
being
=
.
a a e .

exists such that

-

there 0
open ,
r > o Bla ,
r ) c C

• Let r = It a -
bit V = Bla 43 ) and IN = BC b 43 ) . We have 11 A txt = to .

, , ,

Indeed if YA then It b It
fr
txt chill
11 b 11
; y
c E a -
E a c E = r
+
- -
, ,

That is
impossible .

Exercise Show that an

open
ball BC a
,
r ) ,
r > o is an
open
set .
Exercise that

D=
"
Let and I
"
the set
a EIR r > o .
Show
I x EIR 11 x - all > r
}

is in IR
open .

Definition

set complement
" "

subset The called the

Let U be
of IR I I I
-

EIR U
-

a .
x x
} is

'
denote
of by
"
will it "
-

U
-

U in We
-

IR . or IR , U .

Definition

said closed if
"

be
" '
to
"
A set IR in IR U is IR
-

U is
-

in
open in .

Examples

closed its
• is in IR
,
since
complement C-
a
,
o ) U ( I
,
+ no ) is
open .

• I o
,
+ a ) is closed in IR
,
since its complement f-
a
,
o
) is
open .

) I and closed ?
•
} ( Xi ,
Xz X
, 30 Xz > o
}
and
f (x , ,xz )
/ x
, = xz } are in IR

1×330 ) closed ?
•
11×4×2 ,
Xz ) is in IR

"
subset number
A of R that consists of single point finite of
-

• U a or a

points is closed in IR ?

• The closed ball B ( a r

) is closed in IR ?
,

Lemma 1 (De Morgan Laws)

If A
,
B
, ,
132 ,
. . .

,
Bm are sets
,
then

•
All Bj ) =
CABj
- ) • •
As
Bj =
Hi Bj )

Proof
M in

Let A U Bj Then A and ¢ U EA and jem

x E n x c x
Bj ⇒ x x ¢ Bj is
.

• .

,
j I
j
-

I
- -
-

× E Ai
Bj is
jem x E N ( An
Bj )
,
j I
-
-
 m

¢ Bj

x.cn
conversely let As Bj Then Ai
j ⇒ A jem
,
.
x E
Bj ,
is em x E
,
X
,
is

j I
-
-

Hence x EA
, x-fj.IN
Bj
,
⇒ x E A -
Bj .

• • Let XE An ( Bnc ) .
Then
X C- A X EB and x Et C
'
x E Anc
,

{
or or

and # and
)
EA x Bnc C- A X C- C X Et B
X x
,
x E A - B ⇒ x E ( A B)
- U CA - c )
or
or

X EA X Et B and X Etc x C- and

, A -
B XE An ,
(

let ( B) ) Then
conversely x E A UCA - c
-
.

B
x E A -
x C- A X Ct B

{
,

×
Farc / × c. ! a × ¢ ,
Xt A and × Ct Bnc x e A -
( Bnc )

The
following is a
generalization of Lemma I .

Lemma 2

If I Ax } is
,
a
family of sets then

'

Ai
'

(y Aa ) ? I I ) Ai
• = . . Aa =

Proof

Let x E (Y An )! Then x
¢ U Aa x ¢ Aa V-2 x E AI V-2 x AI
c-
an
•
, ,
a

'

conversely ,
let x C-
an AI .
Then x E Ad ,
V-2 x ¢ An ,
ta x Et
Y Aa x c- ( YAa )

• . Let x E ( I Aa ) ? Then x ¢ ?Aa tho such that x at Aa

.
x E Aino x E
YAI
let Aac
conversely ,
x E
y .
Then

'

that Ado I ( I Aa )
such ¢
Fao x E x
Ayo x
I Aa x C-

Theorem

and entire ?
"

set closed
• The
empty the IR are in IR

number closed closed

of finite
?
The sets
• union a
of is in IR

intersection of any number infinite of closed sets closed "

The
•
, finite or
,
is in IR .
Proof

Okc
D=
VE
VI
tfI D=
Of
Oj
C C
both conclude
CR )
y
Because
of and
"
that both of
• = IR =
of are
open ,
we

and IR
"
are closed in IR
"
.

closed sets
"
Let u v u be IR and let U V the
•
I
I 2 ,
. . .
' m
in
j = I
.

J
.

By previous

lemmas
"
have
"
we IR - 0 = IN - = IR - =
,

m
C C

it also
Now since , ja ,
2
,
. .

,
m are
open , follows that
? is
open .

• Let u v u be a collection of closed sets in IR

"
and let
k
' . .

21
.

I ,
. -

y ,

A each
¥
0 Then by Lemme 2 we have Now since
k
.

,
.

k >
,
, ,

is that Hence O closed

open
it
follows is
open
is .

,
.

Definition
"
contains
"

Let
neighborhood of U
of
-

IR A is subset IR that
• a E . a
any

set that contains

an
open a .

subset
"
Let be of
• A a
nonempty IR .
A
neighborhood of A is
any
subset r

contains that contains

"
set
of IR that an
open
A .

Examples

=/
>
'
V ( I and
} is
neighborhood ball

neighborhood
x IR
•
,
,
xg ) E x ,
o
xz o a
open
. . .
I

of the point a =L ; ) , .
Indeed the
open
ball Bla
,
r
) with r :

small is set that contains a and is contained in 0

open
an .

• B ( co ,
o ) ,
2 ) is a
of the set A =/IRI
+21×21
( ) x
,
,
x
,
E I Ix ,
I I
}
 Indeed
if

s
A then and
,
X= I x
,
Xz ) C-
,
IX. Is I 1×21
EL .

-00
1/2
I

BO
42
•
2
It
follows 11×112 = 1×112+1×212 it
I, =
I, ⇒ 11×112 E -2/5 2
1×11+21×21=1

Hence A C B ( c 0,0 ) , 2)

-
I

Definition
"
distance
let of from
"
• Let a EIR and B be a subset IR .
The
point a to B

is the number olla B ) =

inf 11 X - all .

X C- B

of distance
subsets
from the
"
•
Let A and B be two IR .
The A to B is

number dl A B ) = 1lb inf - all .

AEA BCB
,

doesn't
Notice that
if a E B ,
then olla ,
B) = o .
This
imply that if olla
,
B) =o
,
we

'
instance and the
must have at B .
Take
for a =L I
,
I ) EIR B to square lo ,
1) x
,
I ) .

clearly a ¢ B even olla

,
B) =o .

though
Example Let A be or non
empty set of IR ? show that

)1dL d Cy I
"
x A B ) t IR
y 11
E 11 x x E
y
- -

, ,
, ,

We have

fz
11 X 2- It E 11 X -111 t Hy 711 Ht E A 11×-2-11 E It x It t It y Z It it 2- E A
y
- - - - -

, ,

Hence

dcx
,
A) E 11×-711 t in
f Ily - 2-11 = 11 x -

y Il t
ol Cy A ) dcx
,
A ) -
Olly A ) E Hx -
-111 . . . # I
, ,

2- c. A

similarly ,
we have

Ily - 2- II E It y - all + It x -
711 t t C- A
inf Ily .
2- It E 11 x -
711 + It x -
711 it 2- E A
, ,

2- E A
Hence

d ly A) Ell x Ht in
f 11 x 2- It tix dcx A ) d A )Cy ol ( x A ) Hx I )
y =
y 11 E -111 * *
- -

+
- - - . . .

, , , ,
2- E A

and leads to the result

combining # c * * ) ,
.
 Definition
"
Let be IR

A)
A a set in .

diameter diam
• The
of A is
defined by ( =
sup It x -

y
11

77 E A

We that A is bounded if diam I A )

say
• a to .

That is
equivalent to
say ,
there exists R > o such that A C B I O
,
R
) ,

where O is the zero element

of IR ?

If diam A that A is unbounded

•
( ) = to
,
we
say
.

the
Using
2- norm
,

the diameter of the interval

• A =
[ 2,5 ) is
equal to 5-2=3 .

the diameter of the toe x. } II

•
square region { ex ,
,x2 ) E 2
,
o e xze 2 is 2 -

'
the diameter
•
of the
strip A =
fix ,
,x , ) EIR I 1×21 E I
} is + a .

Exercise compute the diameter

of the ball BC o
,
r
) if

1 .
11 . 11 = 11 .

112 2 - H - It = It .
It
,
3 . Il .
11 = 11 .

Ha Answer : 2r

Exercise bounded bounded

"
Show that the union
of two set A and B
of IR is .

Definition
Let A be a subset
of IR ?

called interior
point of if neighborhood of
"
• A
point a EIR is our A
,
A is a a .

set of all interior called the interior of and denoted I

• The points of A is A is .

Theorem If A is
open
then It = A .

"
Ao contained
1 ,
If A is a subset of IR
,
then is the
largest open
set in A .

It Is
"
that
2 .
If A
,
B are subsets
of IR such A a B
,
then a .

O
O O
-
"

3 .
If A
,
B are subsets
of IR ,
then An B = An B
Proof
It of Rn
1 •
clearly is an
open set .
Let B be an
open subset such that B CA .

Ao
want interior
We to show that B C
,
i. e.
, every
x c- B is an
point of A .

let B it of B A
So x E B -

being open ,
is an
open neighborhood x .
But C
,

it
follows that A is or
neighborhood of x .
Hence x is an interior
point of A .

have Ao the theorem ti Is

2 . We C A C B .
the
By
first
part of ,
we have c .

It and Bo it that I B
3 .
From CA c B
, follows n c An B .
Now since

tin Bo
is
open
must have tin Bo
C
AIT Conversely the
using
we .

, ,

second
part of the theorem
,
we can
successively write :

O 0
- o - 0

An B C A ⇒ An B C A and An B C B ⇒ An B C B

O
0

I
-

Hence An B C n B

Definition
"
Let subset IR
A be a
of .

'

interior
"
called it
• A
point a EIR is an exterior
point of A if is an
point of IR - A .

• The set of all

points of A
exterior
called
exterior is the of A .

"
Be
careful ! IR - A is not
always equal to the exterior of A .

closed "
It is
if A is
,
since in this case IR - A is
open .

Theorem
"
subset
Let A be a
of IR .

and
"
at IR is an exterior point of A if only if d la
,
A ) > o .

Proof
'

If a is an exterior
point of A then a is in the interior of IR - A .

"

such that )
Therefore Fr , o BC a
,
r C IR -
A .
 ¢ ) Hence
If x E A
,
then x BC a
,
r
⇒ 11 x -
all > r > o

11 x - all 3
r
,
it x E A ⇒ olla A ) =
inf 11 x -
all > r > o .

X E A

olla and consider ) want to show that

conversely ,
suppose
r=
,
A) so Bca
,
42 .
We

"

let
Bla
,
42 ) C IR -
A .
so x E B ( a
,
r 12 ) .
Then It x -
all Erle .
Consequently x
of A
,

not and 't

because
if 11 x -
all 3 olla ,
A) = r
we can have re It x -
at Erk .

Thus established that CITRA Thus exterior

we have Bla
,
42 ) . a is an
point of A .

Definition
"

Let
"
called cluster
A be a subset
of IR .
A
point a e IR is a
point of A
,

if neighborhood U of has
nonempty intersection with A
-

every
a a .

Notice that if then ofis cluster

a E A
,
a a
point A . The converse is not

cluster doesn't have

necessarily true
,
i. e.
,
a
point of A to
belong to A .

the 1,1 )
For instance corner
point a
of the
=L square region I

A = Lo ,
, ) x
( o
,
i ) is a cluster point of A but it doesn't belong to A .
•-q •

Definition
"

Let set cluster called the

A be a subset
of IR .
The
of all
points of A is

closure denoted I
of A and is
by .

For instance the closure

of A = I o
,
I ) x
( o I ) is It = I o
,
I ] x [ 0
,
I ] -

,
Theorem

Let A be a subset of TK ,
then

At closed subset IR ?
1 .
is a
of
-

"
A is closed subset of then A A
2 .
If a IR
,
=

At closed
containing
3 .
is the smallest set A .

"
Let A
,
B be subsets of IR then
,

ItI
4 .
If A C B
,
then C .

5 .
A U B = A U B

Proof

"
1 We need to show that IR At is
open
.
, .

If a At ,
then I r > o suchEt that Blair ) A A =
of .
we claim that

At Indeed At But
Bla
,
r ) n = to .

, if 7 be Blair ) n .

• b

be At B ( b
,
Ha -
b 11/2 ) A A
t of Bla ,
r ) h At of .
OF
That is impossible since Blair ) A A =
of .

"
At IR At This shows that
"
At
How Bla
,
r ) n = of Bla
,
r ) C ' . IR , is
open

At closed
and
therefore is .

He want to show that A' A that not the Then there exists
2 .
= .

suppose is case .

"
II ofA that
"
and )
b E b . IR , A
being open ,
Fr > o such BC b
,
r C IR , A

that contradicts
Hence BC b. r
) n A =p .
But the
fact that be At ,
since

At
be Bcb
,
r ) A A to .
} .
We want to show that At is the smallest closed set that contains A .

closed that
"
subset such Let that I
Let be
of IR A show
-

t U U
-

a c .
us c .

Let It tr Bca ) A At lo Bla ) n Utd A

- -

a e Then > o r r since C U

, , .
,
.

This shows that I Thus At -

U U
-

a e = .
c .

want show ItI At

4 .
We to that
if A c B
,
then c .
Let at .

If
E A E B C
55 e B
a ⇒ a ⇒ a

If a f- A ,
then Hr > o
,
B ( a
,
r
) n A to Bca ,
r ) A B I to a C- B .

This shows that I C Is we want to show that AUB

AUB
= .

5 .
We start with A AUB
UBC .

A' Ui
Clearly AUB C .

Making use
of property 14 ) .

,
we have

AU B C AUB .
But since AUB is closed we must have AUB = AUB .

Hence AUB C AUB .

it
follows

Alc
A B
,

It remains to AUB C AUB

prove .

From AUB and and of property get

B C AUB
making use C4.1
,
we

At C AUB and B C AUB AUB C AUB

Theorem
"
Let subset IR
A be a
of .

"
cluster of and A)
at IR is a
point A if only if olla
,
= o .

Proof

suppose
a is a cluster point of A .

If a E A
,
then olla
,
A) =o .

If act A ,
then ten C- IN
,
3- xn E Bla
, I ) n A to ,
with
 olla inf limit
Anta .

Clearly ,
A) = 11 x -
all ell xn - all s
tf . Ias
sing
to the as n - star

x C- A

we
get dca
,
A) = o .

olla and let show that cluster point of

conversely , suppose ,
A) = o us a is a A .

that
If a c- A
,
there is
nothing to
prove .

suppose a at A .
The fact

olla each
,
A ) =
inf Hx - all = O
, guarantees for NEIN
,
the existence of xn E A such

XEA

that Hxn all because otherwise would have all A which

-
E In , ,
we tix -
>
In ,
txt
,

would
imply that olla ,
A) = in
f It x -
all 3
In .
How
, given
r > o
,
there exists
x C- A

such that ) contains

he IN
Incr ⇒ Blair ) A A > Bla , I AA Xu .

Hence cluster for

a is a
point A .

Definition

"
Let subset IR
A be a
of .

called
boundary point tea
"
and
A
point a c- IR is or
of A
, if a act .

The set Fr CA ) that consists all the called the

of boundary points of A is

other words consists

boundary or
frontier of A . In Fr CA ) of all the

points in IR
"
that neither interior nor exterior to A .

Notice that
by definition I Fr C A ) )
'

=
It UTI is
open .

consequently Fr LA )

closed subset IR ?
is a
of

Exercise
show that A is a
if
subset of IR
"

,
then Fr CA ) =
I ' I
Definition

called
"

point isolated
"
Let A be a subset
of IR . A a E IR is an
point of A

if A and
if there exists
neighborhood U
of such that A
-

n }
-

a e a a U =L a .

Definition

limit accumulation
point
'
Let A be a subset
of IR ? or EIR is called a
point or an

of if every neighborhood of contains at least point from other

-

A U a one A
,

than a .

Notice that a limit

point of A be an
cannot
isolated
point of A .
I could

interior
be a
boundary point or an
point .

For instance if A
=L ix.
,
xD lot x
,
s I
, ocxzc I
} then both co
,
o
) E A and

( limit points of A
kik )
' are .

Theorem

Let be a subset of IR
"
.
The
following two statements are
equivalent .

converges
limit accumulation
1. a is I point of

I am } distinct
2 .
there exists or
sequence of pairwise points in A

that to a
,
in e. lim 11am
all -
= o .

,
m → a

Proof
(I ) . ( 2 .
) )

Blairs
is accumulation of A Then
suppose
a an
; ; ; ;
.

B l a r
3 )
,
'

for
'
r = r
,
=L there exists a in Bla r
, ) n A
,
with a t a
, ,
B ( a
,
ri )
 ja of
" "
of
I
a
there exists A with
for
'
r -

ra in Bla re ) n to
g
- = .

, , ,

j
' " a " " r
Notice there
that BC la ) and 't
: Iz
re ) B
for
a C r r r
e e
-

, , , a a .
, =
,

exists of in Bla vz ) A A with of tea

Notice that Blais ) C Bla rz ) and of t
of .

, , ,
.

this construct of balls Blair ) Bla re ) Bla

continuing way ,
we a
sequence , s
,
>
,
Rs ) s . . .

each with radii

containing the and
'
next
Is
one I
,
r
,
=
, ra e
z ,
r
,
e ,
. . . a
sequence

distinct of as
of
'

pairwise points a
, , ,
. . .
such that

'll '
Haa' H 'll Hence him any
Is
a a s E Ha a E 11
Lz
11 - - -
a -
= o
z
. . . .

, , ,

m → a

&)
.
⇒ Ct I .

that
there } of
'

Let be such exists 3 am distinct

a e IR a
sequence pairwise

points in A such that him 11 a .

em 11=0 .
We want to show that a is a

m → to

limit
point of A let U be
neighborhood of U
being open
there exists
- -

. so a a
,
.

r > o such that B la r ) C

-

U . Now because lim It a -

any = o there exists N EIN
, ,

n → N
t

"
am
particular
"
such that and
Ha -

u e r
,
t m
> N .
In a
tea a E
Blair ) A r .

Hence a is limit
point of A .

direct of result have established

A
consequence
the we
just is the

fact that if limit point of then neighborhood of

-

is A U
every
a a a
,

contains
infinitely many
distinct points of A .
Definition

1 Let A and B be two subsets of IR

"
He that A is dense in B
say
.
.
,

if It > B .

"

dense
"
I
'
Let subset IR We that A if
2 . A be a
of .

say
is in IR
,
= IR .

that
if
"
This means x EIR
,
then tr > o
,
Bex
,
r ) A A to

Example

dense
"
• A is in IR ?

£
• A =
I (x ,
-1 ) E f -
le x El and -
is yet } is dense in the
square
B = [ o
,
Da [ o
,
I ]

Sequences

Definition

xp
"
countable called
of points
' '
A set IR
I x
,
x
,
. . .

, ,
. . .

} in
,
is a
sequence

"
in IR ? It is
customary
to denote it
by I x
} .

natural
numbers then the
If I pn
}n
is an
increasing sequence of , sequence
, ,

P
keg called subsequence
} x is a
of the
sequence I X
3 .

↳ I

Definition

"
said to
"
to
"

Hxn
A I IR IR and write
sequence
x
} in is
converge
x c-
,
we

lim xn = x
, if He > o Z N -
-
H such that n ? N ⇒ - x It s E
,
n → to

This is
equivalent to that line Hxn
say x11 = o
-

n → too
 Definition
"
if
"
A I x
} in IR is called Cauchy
sequence or
sequence ,

" "
HE Z N NCE ) such that p
N ⇒ It x x H e
g > 3 s
> o -
-
-

Theorem

XP
xp
x'
Let XP be hi
I ) or
sequence
in .

P P
converges
1
If I X
} then
I I is X or
Cauchy sequence
.
, .

k
2 .

If I } is or
Cauchy sequence ,
then
any subsequence I } of } }

is well
a
Cauchy sequence
as .

leave the
We
proof as an exercise .

not
In a
general normed vector
space
the converse
of I I ) . is
always true .

Definition

Let normed vector

xD
be with H
If
E It
every Cauchy
norm
a
space sequence
-

I } in E to E that E is complete for the It II

converges
e norm
say
x we .

, .

it
If E is
complete ,
we call a Banach space .

Theorem
"
IR is Banach
or
space .

Proof

Let } be the canonical basis

I 9=11 ,
0
,
. .

,
o )
,
ez = I o
,
I
,
o
,
. . .
,
o ) ,
. . .

,
en = Co ,
. - .
,
o
,
, )

-211
"
of HI
and
define let
"
IR M =
ejll .
} x
} be or
Cauchy sequence
in
,
i. e.
,

jet
" ''
HE > o I HCE ) such that > nice , 11 X x 11 s E
p ⇒
o >
,
-
.
,
 "

¥! LEY
" "
want show that
We to
f x
) converges
.

Rewriting x -
x = xp ) -

ej ,
we have

xjlIIlies
II.
I
'

yy.gg
!g xjl
"
?
11 - x 'll E I - 'll e lies 'll )
Eyefulxj
-

xjl E M .
-

shows that
This the
sequences I xp } ,
is jen are
Cauchy sequences in IR .

"
IR
being complete ,
each
of s xj 3 converges .
Let
xj =

lying
xp ,
ja ,
2
,
.

,
n
and set
,

× =L Xi
,
Xz ,
. . .

,
Xu ) .

Clearly

11×7×11 = It ( xj 'll xj )
e E M I xjl
Iggy
- -

"
"
him
T.fr#lxj-xjl=o that and have
since
phiz ,
it follows I x
} converges we x -
- x .

, p -7 to

result another characterization

The
following is
of complete spaces .

Theorem (Nested Balls)

Bjcxt
Let
} be
of nonempty closed and nested balls ?
}j
IR
rj ) sequence in
a
, ,

, ,

i. e.
,
B >
Bz 0
Bz > . . -
> Bn > . . -
If fiI r = o there exists a
unique
;
, ,

j , ,

!
"
that n
point x E IR such
B.J ( x
rj ) =
1×3 .

jet

Proof

Let
peg
be
integers .
From Bo C x
,
! ro )
,
C
Bp ( x ? rp ) ⇒ x' e BC x
'
? rp ) ⇒
11×9 -
x Pll
Erp .

conclude that the of centres 1×53 Cauchy

jhjz.rs
since we is
. = o
, sequence a

j 31

IR
" "

I
"
Let xi have
being complete
IR We must
sequence in x
} converges
x=
If
. . .

; , ,

't
Indeed 11×9 Pll x'
'

Bj rj ) V-j letting get il

E C x
X x
, ,
31 .

, g → + • in - x
Erp ,
we 11 -
E
rp
a

Thus x c BC XP , Rp ) , p = ,
,
2
,
3
,
. . .
Hence × E A
j=I
Bj I x ! rj ) .
If y E
II Bj I x ! rj ) ,

then limit
11 x
y H E
rj tj=i 2 Ias
sing to the as
j → to we
get x
y
-
- .
. -
,
. .

, , ,
Theorem (Bolzano-Weierstrass)

"

Every infinite ,
bounded set in IR has at least a limit
point .

Proof

be such set Include it dimensional closed

cubes
Let cube Divide
-

U a in our n
y, J,
. -
.

into closed
2 sub
by bisecting each of its sides .
At least one
of those

closed will contain and

sub cubes
,
call it
Jz , infinitely many points of
tr we have

by bisecting
"

J ,
-
J
,
.
Next divide
Jz
into 2 closed sub cubes each of its sides .

least call will contain

Again ,
at one
of those closed sub cubes
,
it
Jz , infinitely

points of and have this construct

many
to we
J ,
>
J
,
>
J
,
.

Continuing way ,
we a

of empty closed nested cubes J J

sequence J Jk
non > > > . - . o > . - .

, , ,

xk center
k
Let be the
of the cube
Jk ,
k -
-
I
,
2
,
3
,
. . .

clearly the x are

pairwise distinct s .
By the theorem of nested balls
,
there exists a
unique x E
§ ,
Jn .

limit that
point of U Indeed there exists k such
-

x is a .
or > o
, ,

( ) n U with
-

) B r
Jk Xp E
B C ×
C x
,
r ,
, xk t x .

Definition

Let be set ? collection of sets I }

"
called

EC
E a in IR A Ga in IR is an

open covering of
E
, if for every
a
,
Ga is
open and
if E c
YGa .

consider
"

Notice that set E IR has Simply the

covering
in
every an
open .

family of open
ball
I Bla ,
1)
fae
.

clearly µ C- E
Bla
,
D
E
Definition

Let E be a set in IR ? We
say
that E is
compact if every open covering

of E can be reduced to a
finite covering ,
i. e.
, if I Ga 3 is an
open
k to
there such that
covering of exist Gai
E E U
,
a
,
q ,
. .

, an C
i =L

Example

Gj
Let I } be
of real number such that him Un C IR
un a
sequence u
-
- .

n → to

Show that E
=/ a
,
u
,
,
us ,
. . .

,
Un
,
. . .

} is a compact set .

Let } } that

Gi
Gi
i
covering of
be E Let be such since
an
open . u e .

is set there exists such that @ ) Now because

open
an r > o - r u tr C .

, ,

him Un 7 H such that N

I un ul
Thus C- it N
U no
⇒ a r
un n ?
-
- -
, .
,
.

,
n -7 to

Let
Gjn
be such that
une GiGjn ,
I e k EN .
Then
f , Gjy
is he it -
I
} is a

that set
finite sub
covering of compact
E This E is
open
.

proves a .

Example
"
that not
show IR is
compact .

consider balls

an
the
sequence of nested open ( fB
o
,
n )
} ,
They define infinite open
,

"
number
covering of
IR .

Any finite of them

} B ( o
, uh , } must be contained in B ( o
, nm ) .

I e k E N

"
doesn't contain conclude
"

( that IR not
compact
B IR
nm )
since o
,
we is .

Exercise unbounded
"
it not
show that
if E C IR is then is
compact .

Definition

Let E be a set in IR ?

that if
pre compact finite covering of
is
1 . we
say
E
,
V-E> o
,
a
I Gj } a E

I In

such that diam ( Gj ) c E ie je H .

2 .
We
say
that E is
relatively compact , if E is compact .
Theorem

Let set ? bounded and closed

E be a
compact in IR Then E is .

Proof
show that bounded consider balls B ( ) 3
To E is
,
the
sequence of open f o
,
n .

h 31

••k
which there
They define an
open covering of
E
,
is
compact , therefore exist

rn
such that B (
m ,
nz ,
. . .

, ,
E IN
f o
nu
,
) } is an
open covering of
E
,
i. e.
,

I Ek EN
H µ

E U B ( the balls nested have U (

nu ) Since B
me ) BC
nm )
C o are we o 0
=
,
. .
, , ,
k .
-

I k .
-

Hence E C B ( o
uh ) which shows that E is bounded
,
.

, ,

b Il
11 a
Ec
-

Let show that E is closed i. e. show that is

us now
, , open .

'

Let E For each be E let all We have

13
1lb
a c .

, rb = -
.

C
-

t
E
h Indeed the intersection
Bl b
, rb ) BC a
, rb ) =
of .

, if a c in

then all Wich

11 b -
E 11 b - ell + Hc - cell Erb +
rb =
2311
b. all
,
is
impossible .

The
family of open
balls
I B ( b
, rb ) } is an
open covering of E .

Being
b E E

compact ,
we can extract
from it a
finite covering I B
Lbj, 3 .

,
sj , ,

construction
Let r =

Ijn rbj
.

By BC bj , rb )
;
n BC oh
,
rbj ) = ol .
It
follows
,

BC bj h BC ol je Hence
) r
) le
p
rbj
oh = .

,
, ,

E A Blair ) C ( BC bj , rj ) ) h Blair ) = U B
Lbj , rj )
n Blair ) = to
; j=l

This that Ec
shows
arbitrary that
'
Bla ,
r
) C E . since a is in
,
it
follows E

is This shows that E closed

open .
is .
Theorem

Let be set ?
L or in IR The
following statements are
equivalent .

1 .
L is
compact

2
infinite subset U of L has limit
point that
belongs to L
-

every a u .

Proof
.
2 .

doesn't
suppose L is
compact and let V be an
infinite subset of L .
If V

have limit then that

point Vx E such to
-

F
or
,
E
,
r
,
> o Bex
,
r× ) n U =
if

V and if contains at
-

I ) B C
) U
- -

U U n
× Et B C x
,
r× ) n = x x E
,
i. e.
,
x
,
r×

most a
point from V .

Clearly , fee
B ( x
,
r× ) is an
open covering of
E .

Being

compact ,
there exist x.
,
xz ,
. . .

,
A ,
E E such that
II
Blxj ,
r× ; ) is a
finite

covering of
E V E BC xj since each contains
But
r× )
C
Blxj
C
r× ; )
.
.

,
; ; ,

it that has at most

points
at most one
point from V
, follows V N .
That is

assumption that doesn't have limit point incorrect

impossible .
so our to a is .

has limit
This shows that U
point clearly E I E E
-

a u .
u e since u e c = .

2 . .

infinite subset of limit point

En
Ei
has and let
-

U E A
suppose every a
x.ve

µ Ga be an
open covering of E .

Claim : ⇒ r > o
,
such that H x E E
,
B ( x
,
r ) C G
, for some a .

the claim holds

Assuming , pick q e E .

Then Bla ,
,
r ) C
Ga ,
.

If Gg > E
,
then Ga
,
is our
finite covering of E .

If not consider and Then (

, Ga pick Eze Ga
,
.
B as ,
r
) C
Gq .

If Gg .
J E
,
then I Ga
; ! ,
is our
finite covering of E .
 2

not consider E and asE E Then

If ,
- U
Gg .
pick i

Gg .
.
Blas ,
r
) C
Gq .

ja

then
finite covering of
E
If E I Ga
!
is
Gg
> our .

,
.

;
,

If not continue the

,
we
process . . .

of
' "
Let a
,
,
. . .

,
a
,
. . .
be the
points determined by that
process .

These
points at distance Indeed construction
are
pairwise a > r .
, by

at ad'll
'
BC ) and E
j Hoi j
Ga E 2 r
Ga
or r C - 3 ⇒ -
> > 2
, , ,
, ,

j
Ha
.

33
2
Bla ) G and E U
j Ila 'll r
j
Gq
r C or E -
) ⇒ -
a > > 3
, , ,
, ,
2

- . - - - -
- - - - - - - - -

Notice that ball contains ah

for any x E E
,
the B ( x
, I ) at most one
of the .

Indeed
, if I ,
al E B ( x
, I ) ,
then Hak dell -
E
Hak -
x11 thx -
al lls
I +
I
r

This shows that the subset at doesn't have limit point

I } a a e E .

would contain
Because
if it did
,
then
any
Bca ,
s ) ,
S > o
infinitely many points

ai number
points
shows
finite
oui This that the has of
from } } .

sequence I I a :

oh ? ok
,
a . . .

,
.
Hence E C
Gq .
and so
from the
open covering YG , of E
,

able extract
we were to a
finite open covering Gq .
.
Hence E is
compact .

Claim
"
subset of and subset has
Let A be that
infinite of
-

IR U A
suppose
a
every

or limit point in A .
If Y Ga is an
open covering of A
,
there exists r > o and

such that Blx )

do Vx E A
, ,
r C
Ga .

Proof
Indeed if claim then such that
) not
the were
false ,
tn c- in
,
a xy E A B ( xn
,
I is

contained the I ) subset

in
any of Ga .
The
sequence xn is an
infinite of A
,
it limit to
must have or
point x c- A .
x must
belong some
Gg

::÷:÷
÷: "i:
.

"

large enough diagram ) such that It

s and
s
I s
I
11
see
xn - x
g
3 ,

then
If y E B ( x
, , L, ) ,

Y H B ( )
} }
H H H
H Xp, H § }
x E X
-

y -
t
Xm -
E + E + = s ⇒
y
E x
,
s
, ,

Hence ( construction
B x
, , L, )
C B Cx
,
s ) C
Gao . But that is
impossible ,
since by

B ( is not included of the

x
, , I, ) in
any Ga .

Remark
The results stated the two theorems valid IN normed
in
previous are in
,
in
any

vector and and in metric The next result is

only valid in
space any space .

dimensional normed vector

finite space .

Theorem
"
bounded and closed subset
If L is a
of IR
,
then L is
compact .

Proof

is
If L has a
finite number of points ,
then it
clearly compact .

If L has
infinitely many points ,
then
making use
of Bolzano - Weierstrass ,

infinite subset U of L will have limit point since E closed

-

is
every or u .

must have Based theorem

we u e E .
on the
previous ,
E is
compact .
Real Functions of Several Variables

"
Let rule
•
E be a subset of IR .

suppose we are
given a
f that

associates with E real number

thenfunction of
Cx ) C
every
x
,x2 xn
-
-
, a u we
,
. .

, ,

write
fix call real several variables
u = ) .
He
f a .

also
"
write
We will
f : IR →
IR .

called
of explicitly

the
function f when
•
E is the domain the . we want to

mention domain
the
of f ,
we will write
f : E → IR .

The
of values called
• set R all u =
fix , as x varies in E is the
range

of function f .

"

)
"

•
The set G of all
points ( x
,
fix , =
( x , ,xz ,
. .

,
xn
,
fix I ) in IR is

function f
usually
called the
graph of
the When n 1 the
graph a is
-
-

.
,

usually a curve in the

plane 1122
. txt hen n= 2
,
the
graph is a

the 1123
surface in
space .

Example
Find the domain the and the
graph of
the
,
range

⇒
A

function I '
f defined by
'
fix , y ) = - I - x -

y x' ty
Ii

' ' '

XII -141
defined
x2
Domain
f is
if i -
x -

y z ⇐ t
y e I
⇒

That the closed disk centered at co o

) with radius 2 x' +
'
I
y >
.
,

For 2+-12 1×2+-12 ) 1×473 l ?

yds
Range
x I so I I E I E I to o E I x I
o E e I
s ⇐ - - -
-
- -

the
range of f
Hence is co , ] .

,
 Graph Let Ix y t ) be point in the
graph of if then
'

a
,
, ,

:÷:i II:c
:* '

ii.It: I:÷÷÷÷ I
'

⇒
I :: " .
⇐
.
. ⇐ .

2- 3 O

z -

axis

Hence the
graph of f the
half of the sphere
t
is
upper .

z -1*42
,
.

centered at lo
,
o
,
o
) with radius I
e
y . axis
1

x - axis

defined
"
that
•
Let
f : IR
→
IR with domain E .
We
say f is in

n
the
neighborhood of a e IR
, if E contains a
neighborhood of a .

Notice variables
"
that
if f : IR → IR
,
then
by fixing p among
the

n - variables x
,
,
xa ,
. . .

,
xn ,
we
get or
function of n -

p variables .

For instance , if f- Lxi

,
xz
,
xs ) = In (421,2×22) ,
then
setting x
,
=L
,

variables ft : )
defines the
function of two
g
I x
,
,
xz ) =
f ( x
, xs
,
I ) = en
(x x .

function domain
"
with
•
Let
f : IR → IR be or E .
If r E E
,
we

define f Cr ) as the set

of all fix ) as x varies in r .

fer )

is called the
image of r
by f .
In
particular the
range of
the
function f is
simply f LE ) .

Let
function with domain
If f IE )
"
be
•
f : IR →
IR a E .
to E
,
we

'
all that
define f- @) the set
of E such fa
-

as x e E U .

"
the
and then
functions
"
•
If f : IR → IR
,
g
: IR → IR a E IR
,

f- t
g , f -

g ,
a
f ,
f. g , tag
are
defined in the same
way

functions of a
single variable were
defined .
 f function with domain
"
•
Let : IR → IR be a E SIR
"
and let

g
: IR → IR be
defined in I E IR .

If f CE ) A I
t to ,
we

"

define g ( fix ) )
the
composition go f
: IR → IR
by Cg
of I = .

Definition

let
"
Let and
f defined
"
: IR →
IR be in E E IR c E E .

We
say
that
f is continuous at c
, if

that and
HE > o
,
7 8 so
,
such if x e E it x -
ell a O
,
then If f -

/ e e

I
f f is continuous at
every re we that f is continuous in r
c e E
, say .

Example
' 'd
Show that f continuous
: IR → IR
defined by f ex , y ) = 2×+3
y
is in IR

da
Let
,
b) c- 1122 and let E > o . axle have

f- Cx -1 )
,
-
f la
,
b ) =
(2×+3-1) -

Her
b)
-13 = 2 I x -
a) +3 ly b )
- .
It
follows

I fix -

D fcab) I E 21 x at +31 T bl E 12¥32 l K

a) 44 bi s 115 4 I xx ) la b)
112
- -
- - - .
- - -

, , ,

have it to have -115

so to I fix > ) f la b) I EE
suffices 114,7 ) ( a. b) 112 e E
-
-

, , ,

8
IF
Thus 78 that
, for E > o
,
= such if 11 exit ) -
la ,
b) 112 E ⇒ If IX. 7) -

fcab) Is ,
E .

Example

=/
if " id # 90 )
2

IRI
show that the
function f : , re
defined by fc× , > ,
if c x
,
-1 ) = ( o
,
o )

is continuous at lo ,
o
) .
 Let E > o . We have

I !y I !y
I
x

If
4,7 )fest - =

, ,
- o =

z
I -11 El 'll EH kid -

1903112
×

have have
As a result
,
to ex
, y ) If fi-

go > I e E
,
it
suffices to

114,7 ) -
( 0,911 EE .
Hence
,

V-E > o t D= E such that Hex , > ) -

led 112=0 ⇒ I fix ,7 )
-

fco.SI EE .

Example

-0=10,03×3-2,3
that
function
'
Show
f : IR → IR is continuous out c x
,

=/
× 273
if

=/
-

( xn ) # I 0,0 )
if ( xn ) # 90 )
×2txyty2
" "
f t
" " " '
a - ' " ' "

O if " , ,= , o.o , o
if ix.⇒ a o.o ,

Hint :
%
1×3-2-131 1×13+21-113 ( x 't -12 ) 't
2+2
' 'd
b
Cx
f)
E 3
a .
s .
x
xy + y = +
y
,

The
following gives equivalent definitions of the
continuity at a
point .

Theorem
defined
"

f Eand let
"
Let : IR →
IR be in E IR c E E .

The
following statements are
equivalent .

continuous
I .
f is at x = c .

exists
for every neighborhood of fcc neighborhood of
"
2 I ) C- IR there a O c e IR
,
.

such that fl on E ) CI .

3 .

for any sequence I ah ) of points in E that

converges
to c
,
the
sequence

} feck , } converges
to
fee ) .

4 .

for every neighborhood I

of fcc , ,
there exists a
neighborhood V
of c

such that ft I ) = Vh E .
Proof As an illustration we show that H ) . ⇐ 12 ) .

ft ) .
⇒ 12 ) .

Let and
I be a
neighborhood of fcc ) E > o be such that [ f as - e
,
fate ] C I .

exists 8>0 such that fell other

By a) ,
there x E B ( e
,
8) ME ⇒ If .
e E .
In

words
f ( Bcc ,
8) A E ) C ( f -
e
fate , ] a I .
so
for any neighborhood I
of fcc ) ,
,

exists such that

there V= Ble ,
o ) a
neighborhood of c

f ( Vh E ) =
f ( Bcc
,
O) A E ) C
I fcc ) -
e
, fate ] C I

(2)
⇒ a ) .

let For
fate ) neighborhood of f there
Conversely ,
E > o .
I =
(f - e
,
or
,

exists V a
neighborhood of c such that
f funE ) C I .
Let O > o be such

that ) V then
f ) 454,0
Bcc
n E )
f ( ) I
( feel fate )
'

O one c
-

C c = - e
, , ,

T ET and
If f
so
for e > o we have
found > o such that H x all x C- E ⇒ Cx ) I E
-
-
e
,
,

Theorem
"
and
Let
defined
" "
be let
f IR IR E IR
IR IR in E a e IR
g
: :
→ → .

suppose f and
g
are continuous at c c. E .
Then

f and continuous at
•
f t
g ,
a
f. g are c

• If y H ) to
,
t x E E
,
then
Ig is
defined in E and is continuous at c .

Theorem

f function with domain

" "
Let IR IR be E EIR and
suppose
: → a
,

continuous at
it is x= c e E .
Let
g
: IR → IR be
defined and

continuous at fcc ) Then continuous

g of
t = .
is at x = c .

variable
The
proofs
of both theorems are similar to that
of single functions .
 For instance
"
the and
"
1122
, f i IR → IR
, g
: IR → IR h : →
IR
defined
"
xnin In ( 't 't txt
by f X ) =
Z Ai iz ,
. . .
in
x , . . .

, gcx ) = it x
, xz . . .
) and

ii. iz ,
. .

;
in C- IN

-1×2
he ⇐ )
× '
all continuous in their domains

functions
= are .

X
,
2
t 422

Definition

function defined EIR

" "

Let f : IR → IR be a in E

that bounded above exists such that

We
say f is
,
if there Me IR

f Cx ) E M
,
F x E E

that bounded below there exists such that

He
say f if E IR
i
is m
,

f- X 7 3 m
,
t x E E

it bounded above and below

We
say
that f is bounded
if is .

that bounded above then

defined
'
f
'
IR IR on E E IR is
suppose
: →
, ,

its ) bounded above The of LE )

range f CE E IR is .

supremum
M
f

called the
f and denoted fix
supremum of by
is is
sup ,
X C- E

M =
sup fix ) is characterized by the
following two
properties :

Xt E

f- H) M Fx E and V-E F such that M f He )

ee
E C- > o x E -
E a
, ,

If there exists Xm E E such that

fix ) E
f ( xn ) t x E E then
, ,

f called the value

f I Xm ) =
sup fix , .
In this case C
Xm ) is maximum

Xt E

and that reaches its

of f on E we
say f maximum on E .
Example

function f
'
consider

fix
'
the and
g defined in E IR
by )
-
- =

It 4,4×22
44×22
and
gcx ,
= n

It Xf -1×22

From f KI E I
,
Vx E E and
fco , = I
,
it
follows that f has or

maximum on E and 1 is its maximum value

From
gas
e I
,
txt E and the
fact that
for any
e > o
,

2
xp -1×2
xftxz
' E

glx )
-

I -
E c ⇐ I - E a ⇒ >
'
tix ,
-1×22 E

it that doesn't have value

gcx
and
follows sup , =L
g
a maximum .

Xf E

infimum defined
"
Let Its
defined
be
f
"
: IR IR E IR is in a
- s in E .

similar bounded
way
as the
supremum
.
More
specifically , if f CE ) is

below the called the

, infimum m
of f CE ) s R is
infimum of f and is

denoted f f characterized by
by in Cx ) . m = in
f fix , is

XE E X EE

fix ) 3 m txt E and V-E > o F x C- E such that f e)

Cx a mte
, , ,

If there exists such that f E

xn e E
fix , s
,
Km )
,
tx C-

then f inf fix , this fl called the

minimum
its
Hm ) In is minimum
= .
case x
m )
Xt E

value and that reaches

of f on E we
say f on E .
Theorem

Let
"
continuous closed bounded set ?
f : IR IR be in a and E CIR

closed and bounded

1 .

f LE ) is in IR

2 There exist and in E such that

xm Xm
.

inf fix , =
fcxm , and
sup fix ) =
fkn )
X E E
X C- E

Proof

not bounded Then

suppose fl E ) is .

'

for k=i such that f

'
For EE ca ) 31
,

at
'
of that f of )
for k=2 such L
7 C- E # a z 2
, ,

for k =3 7 LEE ,
oft a
'
of such that
flop ) 33
, ,

continuing this
way ,
we construct a
sequence fork } of pairwise

distinct
points in E such that flock ) 3 k
,
k=i
,
2
,
.
. .

ork

=/
has
infinitely

points
The set A I k=i
,
2
,
. . .

} CE
many
and

is bounded Bolzano Weierstrass theorem

By
-
.

"
" "
accumulation
Every infinite ,
bounded set in IR has at least a
limit point

A has limit I accumulation extract ak }

a
point a .
Hence we can
from }

orb aki
'

subsequence such that him closed

or
I f = a .
Since E is

j Sta
j
-

3 I

or = him aki C- E .
Now
f is continuous on E
,
we have
j -
star

flats
'

him ) =
far )
j -7 to

'

This contradicts the

fact that flocks ) >
, Kj , j
-
-
I
,
2
,
. . .

and shows that fl E ) is bounded .

 that closed f

bn
will that
•
To show ft ) is
,
we show
f LE ) ) .

limit f CE )
so let b be a
point of .
There exists a
sequence I } of

points in
ft 't ) that are
pairwise different such that
Eye ,
bn = b .

since each bk C- f- CE ) ,
C- E mech Tak
that bk= flak) .

The k distinct the keel

bn ,
=L
, 2,3 ,
. . .

being pairwise , ok ,
, 2,3 ,
. . .

also distinct
are
pairwise . The set A =
fank
,
> i } has
infinitely
many points and is bounded , by Bolzano - Weierstrass Theorem it has

accumulation extract
point

akif
an or .
Hence we can a
convergent subsequence
aki
I from the sequence fork } such that him = a .
Because
i star
, k > I
-

that continuous
E is closed
,
we must have a e E .

Using the
fact f is in E

"
La )
ki
write b
finna t.iq f f ) for
bn )
we
air
can -
- = = = .

;
,

That not b Ct f LE )
is
possible as .

show that exists such

How we there Xm C- E that
sup f ex ) =
fkn )
X E E

By It ) .
f CE ) is bounded
four above
, therefore sup fix , exists .
Let

X C- E

property characterizes
hype
fix
M
By the that the
supremum
= , . :

'
He > o
,
t a E E such that M - e e
f Lol ) E M

of distinct points 3 ok
}↳
there exists in E
a
sequence pairwise
,

such that M -

th a flak ) E M .

Again by Bolzano - Weierstrass

we can extract a
convergent subsequence y
aki } such that
; 31
ki
Lie flak
)
lim E Ias to the limit in
'
M
sing
a = a c- .
M -
s

i -
in

and
using
the continuity of f in E
,
we
get f Ca ) = M .
Consequence

norms
equivalent
"
two
Any on IR are .

Proof
n

By transitivity ,
it is
enough to show that
every
norm y on IR is

the
equivalent to one - norm Hell
, ,
i. e.
,
show the existence of two positive constants

"
K ,
a K2 such K ,
11×11 f 4 C x ) E K2 11×11 t x E IR .
, , ,

"

continuous
"
First
,
the norm
y : IR - s IR as a
function ,
it is on IR
equipped with the

"
Indeed have
one - norm .

, if c e IR
,
we

141×7
401 -
E 4 l x - c ) a
y l
,
ki -

Ci ) ei ) E
÷ ,
Hi -

Cil Heil E C
, n
y Cei ) ) . tix - ell
,

such that 8
Hence
, given
E , o
,
I
0=4 wax y Ceil ,
H x - all
,
s ⇒

is Isn

let
41913
l " l )
14k I '
Neil ) 11 x c E
Neil E
) 4
?rn
- .
-
= .

,
, n , , n

1=1
closed IR ?
"
Let x E IR I 11×11
,
= I
3 .
This is or and bounded set in
By the
previous

theorem there exist and such that '

'
in is
, xn xn ycxm ) E
y E
ycxm )
,
Vx C- S

"
×
have It
follows
'
Now
, for x EIR
,
X # 0
,
we
11×11
E S .

4 ( )
nd 4 (
yin
) YCXM ) 4km ) 11×11 , )
X
4 Hml E E
4km ) ⇐ 4km ) E E ⇐ f 4C x s ycxm ) 11×11
,
,,
, ,

and have
Letting
K
, =
4C Xm ) =
inf y K2 =
ycxm ) =
sup ylx ) ,
we

x C- § XE is
'

K
,
11×11 E 4 CX ) f Kz 11×11 , V x C IR ?
,
,

Definition

that
'
Let subset of uniformly
"

f
'
IR be continuous in E IR We f is
IR a
say
:
→ .

continuous in E
, if

He o I 8=5 CE ) such that x C- E and Hx -111 ET ⇒ I fix ) fly ) Is E

>
y
- -

, ,
Example
"
" "

suppose
IR is
equipped with a norm it .
It .
Let
y : IR → IR be a norm on IR .

"
show that
y
is
uniformly continuous in IR .

have equivalent
"
For x
, Y
EIR
,
we
14 -

ya ) I E 4 I x -
y ) . since 4*1 and 11×11 are ,

there

Ekg
"
constants

Kell
exist o c K ,
a
kg such that K ,
11×11 E
y 11×11
,
VX C- IR

It Hence
follows 14 x ) -

y I s y I x -
y ) s x -

y 11 .

K2
ke 11×-711 ¥
"
and
for x
, Y
EIR 11×-711 E 8 ⇒ Il x -

y
H s
¥ ,
⇒ 14414 -
I e E
,
= E

Theorem

"
continuous closed and bounded set
"

If f : IR → IR is in a E in IR
,

then it is
uniformly continuous in E .

Proof

We need to show

He 8=8 such that and 11×-711 ET

> o
,
I so ,
x
, -1
E E
I fix ) -

f 11 E E

So let
f continuous at E Ole such that E and
being p)
e > e a x e
o .
p , , ,

See If
fell )
pH
¥eB(
H
clearly Ole p ) 12
x s p) x I E 42 is
⇒ p an
open cover
- -

.
, , ,

extract from the

¥eB(

exists
of E .
E
being compact ,
we can
covering p ,
see .pl/2 ) of E a

finite covering it ,
B
Cpj ,
die
, Pjs 12 ) .
Let Ice ) =

If JEN
min Ole Pj ) 12 ,
.
If ×
, -1
C- E are

that 8 that

ke
such H x -
-111 s
,
then there is H such x EB ( p ,
TCE p 312 ) ⇒
,
k k

8kg
'
Elly 4th 86 ) Pk TCE pre ) B ( p OLE P ) ) Hence
Hy -

pull
-
x x -

Pk " s t e
, ⇒ y
E
, ,
.

k k

×
-1
E E and H x -111 ETC E ) ⇒ x
-1 E B ( p OLE P ) ) Now
f is continuous at
ph
-

, , ,
-

, ,
k k

it
follows x
Ifl
fall
) -
e I fix ) -

ftp.sl-lfcpni -
fail s
{ +
{ = E
Definition

let
"
Let and
f defined
"
E
: IR →
IR be in E E IR c E .
We
say
that
f

osllx
has a limit heir at x=c
,
and we write
¥zfix ) =L
, if

such and O
HE > o
,
78 so
,
that
if x C- E -
ell a
,
then If -
L
fee

Notice that whether f not at doesn't role

,
is
defined or x=c
play any

limit at Notice that with definition

in
f having a x=c .
also this ,

continuous
"

f : IR → IR is at x=c
if and only if fig fix ) =
fcc ) .

Let e
,
= Cl 0
. .
o ) ez = I O
1,0 .
o ) . . . en = ( o o
. . .
o i ) be the canonical basis
,
. .

,
.
,
, , , ,
, , , , , ,

"

of IR ? Notice that
if x -
- Lxi
,
xz
,
. . .

,
xn ) C- IR
,
then

x t tei =
l Xi
,
. - .

,
Xi
,
. . .

,
xn ) + t ( o
,
. . .

,
I
,
. . .

,
o
) = ( Xi ,
. . .

,
Kitt ,
- - -

,
Xu )

Definition

Let let
"
set
"

f : IR → IR be
defined in an
open
r e IR and c c- r .

derivatives of f
• The
first partial at x = c are
defined by

f I Ct tei ft ) )
Jq f
him
-

i =L 2
.
I c ) = , , ,
. . .

,
n

t -
so t

provided the limitsexist notations

Otherfor if and
.
2x cc ) are
fq.cc ) 3¥.cc )
•
If 2x , f ,
I si en exist
for every c c- r
,
this
defines the
first
derivatives
partial functions of f
"
.

They are
Kif : IR →
R : x .

2. fix )

derivative
•
The
first partial derivatives
of the
first partial functions

2x if ,
I e ien are called the second
partial derivative
functions of f
'
second derivatives
⇐f ) partial
These are I si
jen There are n

2g
. .
,
.

22 f
2g .

# if ) is usually denoted
by 2
xjf
or
fxixj
or

z×j*
.
 derivatives
•
The third
partial of f are
defined as the
first partial
2
derivatives second partial derivatives
of the
2×i×jf ,
is ijen
,

2 '

These are 2
,'
k
¥g.
.
f ) is is heh
j ,
. There are n third
partial derivatives .

andf)g. % Ej
usually denoted
¥×j×nf k
fq×j×n
is
,
is iii. th or or
a sq
.

derivatives
•
Higher order
partial are
defined recursively by

¥ (d !
Ptl

f Kip xipf ) ' '

it Eh
=

,
.

, Xis . . .

xipxip ,
ii xiz . . .

"
that class write
• We will
say f is
of C in r
,
and we will

"

partial derivatives
f- E C l r
) , if all the
functions of f , up
to

the order exist and continuous in r

p
are .

Example

=/
×
y
1×2-72 )
( x
,7 ) I I 0,0 )
Let R2 +2+42
f : -
s IR be
defined by f- Hii ) -

Compute
, , , , , , , ,
2 2 O

2x f Try
, f ,
2×yf and 2y×f .

have
x
"
t 4×42-74
At 1971*10,0 ) ,
we 2x fix ,
y ) =
y 2×(43-1422) =
y
2+-12 )
2
( x

ft:D
If
D
I
At have " °
to
,
o
) ,
we
2×710,0) =
lying =
thing
= o

2! 99
4×42
" "

jk
?
( )
x T
At 1471*10,03 have fix
-
-

we y ) =
× 2x = ×
, ,
,
2+-12 )
2
( x

ft
If
D
I
At have ↳ °
to
,
o
) ,
we a
y
flop ) =
lying =
lying
= 0

TY
5)?I!
+49×42-9×4
"
Tb
23 , f #fD
ix
-

Atlan ) have ix. D= =3

tied ,
we
ay ,
>
,
=
3
2
⇐
2)y +

f ¥
!
" to 't f
aiyflo
-

-
" ' " " '
- °

At have
10,0 ) ,
we
,
D= hey , .
=
thy , . ,
=
hey , ofD=
- -
I
 2x 4ft
1×2+-12,3
.li
¥"
-4×41×74 ×b+qx4y ? 9×44--16

)=2x(
Hy
24
's

)
x
At c xn ) t C 0,03 , we have f- Hi ) = f kid =

2++2,2 ⇐
,

5
t
"
22
He
°=
"
-
-
'

At 10,0 ) ,
we have f 10,0) =
lim =
thing hey ) =L
t → o e . ,

Exercise

=/
XT lylflxl
Let
find - > IR
defined by fcxie ) -
Show that
zyfco.DE?,xflo.o)
-

xy I -11 > txt

171=1×1
141=1×1
HI > txt

171 CIXI lylslxl

HHH
I -1171×1
1-11=1×1

Theorem (Schwarz)

Let f
" "
be
defined open set
: IR → IR in an r SIR

If 2x f ,
2g.f
and I f exist and are continuous in r
,
then
, , g.

22g.×if exists in r and

22g ,
fix ) =
21 f Ix )
,
it x E r .

g.
.

, ,

Proof
kart
of assignment # 2 .
Definition

defined
"
Let
f : IR IR be in a
neighborhood of a
point c e IR ?

We
say
that f is
differentiable at x=c
, if there exists a linear

transformation such that

for h small have
"
T : IR →
IR we
,

f C c th ) =
fcc ) t TK ) t 11h11 4th ) . - - .
#

where tech ) is a real

function of he that
satisfies hey oath ) = o

The linear
transformation tint → IR is called the
differential of

and denoted by dfc With this notation becomes

f out x=c will be .
et ) .

f-
4th ) =
fcc ) +
dfc K ) +
11h11 4th )

f- 4th ) f TK )
him
- -

Notice that * is
equivalent to = o .

h → o
11h11

Theorem

defined neighborhood of
"
Let
f : IR IR be in a x -
- C .

If f is
differentiable at x=c
,
then

has derivatives
I .
f first partial at x=c and we have

dfclh )

=L
=
2x fly ,
.
h, + 2×2 ftp.h2 + . . .
+ Inf Cc) .hn

2x ,
f I c )
Isf cc ) . . .

2µF Cc ) ]
I ) §gkn

continuous at
2 .
f is x=c

matrix called
The row I 2x ,
f le )
Isf c c ) . . .

2µF cc ) ] is
usually the
gradient
sometimes denoted
of the
function f : Rn → IR at x=c .
It is
by

< 2x
,
ft )
, 2×2 fcc ) ,
. . .

,
dxnfcc ) > or
Ff I c )
Proof

IF
to
f has partial derivative out ×=c

at
f being differentiable x c we have
f 4th ) fest dfdh ) + 11h11 inch )
-
. =
,

him
small satisfies
"

for where

transformation
h
, do : IR → IR

h - so
49=0 .

Writing has

canonical basis
-2 where the of
"

h = hiei ,
-

- u
,
o
,
. . .

,
o )
, fee
. . .

,
en = I o
,
. . .

,
o
,
is } is IR

it

and linear have

using
the
fact that
dfc : IRN → IR is a
,
we

dfdh ) =
dfc I
,
¥hiei ) =

II ,
dfclhiei ) =

,
hi dflei ) * ,

It
follows f @
th ) =
fee , tdfch ) + 11h11 Ldh ) can be rewritten as

f 4th ) = fcc ) + I hidfcei ) + 11h11 Ldh ) and we have

i= ,

tdfdei
I
de ""
tei
Kt f ) It

ehof
) -
c) t " ei " ' '
dfdei ) after)
figo
.

et
=
hey ,
= + vein
,

=
dfdei)

back write
Hence
2×2 fcc ) .
exists and is
equal to
olfdei ) .

Going
to # )
,
we can

dfdh ) = Z hi 2x feel
,
it

2 .

f is continuous at x=c

established
Using the
result just , if f is
differentiable at x=c
,
then

h small have where IRI

for ,
we
f Cc
th ) =
fee ) + dfdh ) + 11h11 21h ) ,
do : IR

satisfies hairy
.
497=0 ,
and where
dfth ) =

=2 2- =)
hi 2x ; f Cc )

absolute values
Taking the and
working use
of the CSB
inequality ,
we have

" "
II II taxi fault )
'
I hit taxi fol I hit ) (
I dfdh ) I e I hi axiflcs I E e

, , , ,

" "
)
El ?K
'
s ( ⇐ I 2x ; fest 11h11
,
taxi f 12 ) .
11h11
,
,

"

Z
' 2
E L 11h11
,
where L = K I dxifccsl )
,

Now because him tech ) = o

,
there To so such exists
that 11h11 ED ⇒ the Is I

h - so
 Let x be in the
neighborhood of c - Then

) -
=
@tCx.c) ) -

fell
IfI
f Hfc
Ifk =
K c)
-
t 11 x - all
do ( x -
c) I

E I dfccx ) -
c I + H x - Cll I he C x -
c) I E L 11 X -
Cll + It x -
all .
I E 4th ) 11 x -
all

Hence
for E > o
,
there exists 8 =
¥ ,
,
such that
if tix -
c " e 8
,
then

If fcc ) I 4+4 ( ) T ( )
¥
ex ) -
E
11 x -
all E HL = HL = E

and so
f is continuous at x=c .

Theorem

Let
f defined neighborhood of
"
: IR →
IR be in a or
point c .

first partial derivatives

all the continuous at
If 2x ; f are x=c
,

then f is
differentiable at x= c .

Proof

For
clarity let 's have
suppose
n 2 For h
@he ) small
enough we
-

-
=
, ,
.

f- 4th ) -

f = f @the , Cs tha ) -
f Ca
,
Cs ) =

th
=
f @ , , Sth ) .

fla ,
cat he ) t Ha
,
Sethe ) f -

4G)
,

variable
Making use
of the mean value theorem
for functions of or
single :

'

y ( ) y ( )
att t Ot a I
-

y Loi ) = at o s c

write
we can
successively

Costly
Cathy
) f- k ) h 2x f ( hi

he
f- ( 9th ,
-

,
=
, at 0
, , cat ) ,
o a O
,
c I
,

f la
,
Cz ) -

ftha kCs ) ,
= h2 2×2 f ( a , Cz t 02h27 ,
o a 02<1

It
follows
 tha
hi
f 4th ) fcc ) =
hi 2x f ( Gt Q G ) t h2 2×2 f I 9 Cz t 0292 )

Czthg
Ethos
02h21
-

, , ,
,

cztozhz
=
he 2x , f C a
,
Ca ) + he ,
f Cato , , §× ) -
2x f- ,
C9
,
be ) )

+ 922×2 f ( a ,
G ) t h2 (dxzf ( a
,
Cz t -

2×2 f ( a. Cs ) )

Hence
f Cc
th ) =
fcc ) + I 2x )
fcc
fcc
2×2 , ]
I Ye ) + 11h11 Lech ) where

fly
, ,

4th ) =
Iq , (2x ,
f- Cato , , ) -
2x f- (
,
9,47 ) +
hsdxzf.ca
) 2×274,4) ) )
,
-

Taking the absolute value

, using
the
triangle inequality and the
fact that

I hit s 11h Ha ,
1h21 s 11h Ha ,
we
get

024
that
0,4
thhi
14911 E
je ( that flat G
12) 2x f- Ca g. t 1h21 ) fI a
2×2719
2g
cat ) -
I )
Cztozhz
Cathy
Ethos
-
,
.
.

, , ,
,
,
,,

Eff (I )
"
E 2x f- Cato
, , ,
) -
2x , f- la ,
) It 12×2 f ( a
,
) -

2×2 f la ,
Cz ) I

follows
" " h "
being equivalent have
How all norms on IR we L
a
k h to
, s e ,
,
11h11

constants
for some L
,
K EIR .
It

Cztozhg
Cathy
hi 49=0
4th ) I s K
(I 2x
, f I Gto , ,
) -
2x , f- I a
,
4th ,
) I t 12×2 f ( a
,
) -

2×2 f 19 ,
(2) I )
Now
passing
to the limit as h → o and
using
the
fact that both 2x , f

and continuous at that leads

2×2 f figg
are a = c and o so , 0-2<1 to .

, ,

Remark

derivatives
Notice that the mere existence
of the
first partial doesn't
guarantee

differentiability .
Actually it doesn't even
guarantee continuity .

Y
if
×

03×2-42 ) 147 If

{
,

For instance
f H il ) = has
partial derivatives
) )
o
if ex ,y = co
,
o
 !
-142×2 ) if
× 1×42 )
ix. ) tho if ix. ) ¥40 )

I
y y

{
,
2
I2)TY 444232
2×fcx y ) = and 2
fix ,y ) =

!
,

o if 471=190 o if ix. 7) =
Igo )

(
verify ) but f is not continuous , , ,,, , , . Indeed ,
we have

fifo flx.tl Iyo 11×2 ¥7

'
= =

z
=
'
z t flo ,
03=0
.

Theorem (Operations on Differentiable Functions)

differentiable
"
Let R' that at
f : IR → IR
, g
: → IR be two
functions are s

x=c ,and let a EIR .

Then

out
•
ftg , af ,
f. g are
differentiable x = c and we have

d
Lf
g) t =
dfc t
Olga , diff = a
dfc ,
a

ol ( f.g) e
=
get dfa t
fadge
at
•
fog
is
differentiable x=c
, provided g to and we have

8 "
dfa
to fadge
-

) ,
=

94 )

The
proof is left as an exercise .
 Definition

tix
z
LbOD
-

•
+
a
b
•

II =/
"
Let a
,
b E IR .
The set a
,
b I at t Ib -
a) I o et si } a
line
segment

called the line end

is
segment with
points a and b .

m " " set

Let be set IR ? We that

r a in
say
r is convex
, if

ta
,
be r II a b I C r .

, ,

nonconvex set

Mean Value Theorem

differentiable EIR
"
set
"
Then
Let
f : IR -
s IR be in an
open
r .

it x E r and Hh EIR
"
such that xth It C A we have
, ,

f 4th ) -

f K ) =
Z 2 flxtoh ) hi
I for some o E C o
,
i )
Xi
i= I

Proof

xthtt consider
function of
"
For and he such that Ex the
-

x E U IR C r
, ,

variable
a
single y : IR → IR
defined by yet ) =
f (x + th ) .
Let t
,
to E ( 0,1 ) .

Notice that close to close to xttoh

if t is to
,
then x - ith =
Cxtoe )
+ + Lt -
to ) h is .

The
function f being differentiable at xttoh
,
we have

f- I xtth ) f ft toh ) ol f Kc
to ) h ) It Ct tbh II
2×+1*44
tbh )
-

+
- -

= +
* ↳ ,

which we can rewrite as

f- I x + th ) -

f Cx t to h ) =
(t -
to ) Of (9) + It -

tot 11h11
2x + toes
( Ct
-

tbh )
xttoh

It
follows
yes -

y to ) f- I # th ) -

f ' '' t to " I

= = if toes
) +
I l ( t tbh
-

)
t -
to t -
to , , + toe

or else
 I
yes -

4 Lto) It -
tot
hi
t .
to
= 2 Kif txt to h ) t
t .
to I a-!
tbh
.io
)
in

Kassing to the limit shows that

y
is
differentiable at t= to and we have

n
'

4 to ) =
-2 hi .
flxttoh )
i= I

know and
differentiable
that continuous
Now we
yet ) =
f Cx
t th ) is in co ,
D

in co ,
' )
,
we can
apply the Mean Value Theorem to yet ) to
get
n

Z
'

I ) I f 4th ) f- ( X ) hi 2x , fl x oh ) osoci
(I
=

4 I ) y @) o
y
o cool ⇐ t
-

- = -

, ,

2 I

Example

continuous
f be
function with
first partial derivatives
"
Let : IR → IR a

"
set
2×2 f- Lx ) for
in r SIR Hx r it
suppose
o
open E
or convex = 2
,n
. . .
.
, , ,

Show that f Cx ) = o Vx Er .

Let a Er be
fixed point in r For
any x er the line
segment

ca
flat
a .

II II contained the Value Theorem have

a
,
x is in r .
By Mean ,
we

n n

f- Cx) -

f ) =
I Z ,
O ( x -
al ) .
Hi Qi )
-
=
-2 O .
Lxi -

Qi ) = o

it I I
2

Hence fees
fix ) =
, for any x er .

Now that f IR
'
IR has continuous partial derivatives
to order
suppose pti
: →
up

set EIR
'
induction and did
in an
open
t . Then
using proceeding as we in the

establish Taylor Theorem

proof of the Mean Value Theorem ,
one can a

variables
for functions of several .
More
specifically ,
we have
Taylor Theorem

differentiable EIR
"
set
"

tix
Then
Let
f : IR -
s IR be in an
open
r .

it x E r and Hh EIR
"
such that xth It C A we have
, ,

I
"

II ! 2x
2 ]

f- Cx th ) =
f A +
i, ( II ,
hi 2x ; fix ) x
L, (
,
hi ) fix ) + . . . .

axil ix.
÷!
axis
"

p
! II. hi fix , +

, .
hi on ,
o.o . ,

"

fix )
" '
Here
4¥ 2x
,
hi
i
) fix ) = hi
Kif K )
, (
,
hi 2x
i
) is
computed by first
" ] '

expanding (
,
hi 2x
; ) as one would
expand ( ÷ ,
hi di ) with the convention

that
any product ai g- is replaced by 2
×j
to
get ( hi 2x ; ) = hi hi I. xj ,

, ,
" '
then
( II 4.)
hi
,
f =

,
§ ,
hi hj Bigfix .
) .

For instance

" '

(h , 2x
,
+ h2 2×2 ) f =
( K2x 12+2, the
2x ,
) l hzdxz ) + 2×25 )
( he f Cx )

= (hi ¥ ×
x 2h he
× ,
the 22×2×2 ) f- Cx )
,

=
hid
2¥ ×
fix ) t 2h he
*
f K ) +
h : Imf I x
)
,

"
More
generally ( II ,
hi Ki fix ) is
computed by first expanding (
,
hi 2x
, )
k

2µm
as one would
expand ( ÷ hi di ) with the convention that
any product hi
,
hi
,
. . .

air
"
!
,

( II
k

replaced by get dxi ) .si?g,...,infjihii..hikK!xiz

is to hi '
×
xin
. . .
. .

,
.

, ,

then

( II kill hi
.si?g.......n.hniihii..hinK!xi....xinf
" =

"f "
Example

Theorem f it 2×7+3-12 point

Write
Taylor for Cx -1 )
,
= - -

6×-27 -
4 at the C -

2,1 )

We start
by computing the partial derivatives of f and
evaluating them at C- 2. is

We have dxflx -1 ) = -
2×+2
y
- 6 2 flxie ) = 2×+67-2
, , ,

2×2 f ix. y ) = - 2
,
27 f- 4,7 ) =
27 fix ,7 ) = 2 27 f- C xn ) = 6
, , , ,
,

All
higher order partial derivatives of fun , are
equal to zero .

Furthermore f I -2,1 ) = I 2x ft -2,1 ) = o

Lyft -2 1) = o
, , , ,

0¥ ft -

2,1 ) = - 2
,
Sky fl -

2. D= 2
,
24 , ft -

2,1 ) = 6

and order derivative equal to

all
higher of f at c- 2,1 ) are o .

It
follows

f- I x
,
y ) =
f l -

2,1 ) +
÷, ( Kt 2) 2×+4 1) Try ) -
I -2,1 ) +
Is (4+2)
2x t ly -

Day }¥ C- 2,1 ) t Ot Ot . . -

tf
=
ft -

2,1 ) + + HI )
-

Lyft -2,1 ) ]

(4+272×71-2,1) 2¥ ft
+ 4+21243×71-2,1 ) -121×+2 ) ly D -

-2,1 ) t if
LT -
ft -2,1 )
2

?C ?
1
( I ) (y D ] I ( I x -12 ) 2) +21×+2 )( D (2) (T ) (6)
x -12 t
O o y t
- -

= + + +
-
- .
. .

&
= I -
4+212+21×+2 ) ( y -
I ) + 3 ( y -
1)

Example

Apply Taylor theorem

of order to
function fix
thy
the

fix
2 T ) =
,

,
,

at the
point c 1,1 ) .

2X 2-1
f
- -

We have 2x ,y ) = 2 Lxii ) =
, ,
2 2
( x 2+-12 ) ( x
2
2)
+
y

6×2-2-12 2×2+6-12
&xx f 2%71×9 ) ¥7 ft
-

Kil ) =
i
=
I TY =

1×2+12,3
3
2)
z ( ×z+y 2)
 -24×1×2 2) 8715×2 -13
2¥ fix
T
23×+1 fix ? ,×fCx , ) ,,f
-

=3
- .

* , ) =
' ,y ) =
2,3 ix. y ) =
4
( 2+7
2)
, 4
⇐
2+-12 ) x

£ 8×1×2-572 )
yyxflxi ) =
¥ ,
fix ,Y ) =
¥yf I xp ) =

( x 2x2)y
4

2471×2--12
¥yyf I
xn ) =
2
,

( x 2)
+ y
4

Evaluating at c 1,1 )
,
we
get fu , ) .
-

I ,
2x fly )
=2yf
11,1 ) = -
t
2 I

¥ fly ) =3 ? fly ) =
I ,
23 , fell ,
' ) =
27×711,1) = I .
It
follows
,

¥11,1
I ()"
"

f I xn ) =
f Cl
,
' ) +
÷ ( 4-172×+4 -
D2 , ) + HD
-
2×+4 D dy -

) If I 1,1 )
,

(3 ]
( H dy ) f ( )
f,
+
-
t ) 2x + lT -
D it OC x-D It -0 Cy - I )
,

2
=
'
z
t I X D 2x
-

fly ) t HDdyfll
-

,
' ) I +
'
z I D2
Cx -
23×711,1 ) + 2K -

DH -

D fly ) t HD-

¥ fl ;D ]

(
3
21*+31×-1124 D2 !× D2 21 4-1132? ) flit )
+
If K is t 3 Cx icy -0k it Holy it
-

+
-
- - . -

, , , , ,
,

] 2+4 D2
I I K t) )
¥ I ] Rl )
'
= t
- -
H -
t t ( x -
1) I x -
1) IT - t ) + Cy
-

+ x. 7 o
z ,