MATH2111 Higher Several Variable Calculus

Analysis

Dr. Jonathan Kress

School of Mathematics and Statistics

University of New South Wales

Semester 1, 2014

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 1 / 52

Analysis
Concepts from real one-variable calculus

f :R→R
limits
continuity
differentiability
integrability
Theorems
Min/Max
A continuous function on a closed interval attains a max and min value.
Intermediate Value Theorem
A continuous function on [a, b] attains all values in [f (a), f (b)].
Mean Value Theorem
Connects the instantaneous rate of change of a differentiable function to its
change over a finite closed interval.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 2 / 52

Functions from Rn to Rm

We want to study functions with domain D ⊂ Rn

f :D→R scalar fields

f : D → Rm vector fields

Examples

f(x) = Ax where A is a matrix

f(x1 , x2 , x3 ) = (x12 − x22 , − sin x2 )

A vector field can be thought of as m scalar fields, its components, that is

 
f(x) = f1 (x), f2 (x), . . . , fm (x) .

f1 , f2 , . . . , fm are the components of f.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 3 / 52

Distance Functions (metrics)

The usual Euclidean distance between x and y in Rn is
v
u n
uX
d (x, y) = ||x − y|| = t (xi − yi )2 .
i=1

It’s easy to check that this satisfies

z
n
1 d (x, y) ≥ 0 ∀x, y ∈ R and d (x, y) = 0 ⇔ x = y
)
z

d (x, y) = d (y, x) ∀x, y ∈ Rn

x,

2
d (y , z
d(

3 d (x, z) ≤ d (x, y) + d (y, z) ∀x, y, z ∈ Rn .

)

x
A function d : Rn × Rn → R statisfying these properties is d (x, y
called a metric. ) y

A metric is (1) positive definite, (2) symmetric and (3)

satisfies the triangle inequality.
JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 4 / 52
Distance Functions (metrics)
Examples

n
! p1
X
dp (x, y) = |xi − yi |p for 1 ≤ p < ∞.
i=1

eg d1 (x, y) = |x1 − y1 | + |x2 − y2 | + · · · + |xn − yn |

p
d2 (x, y) = (x1 − y1 )2 + (x2 − y2 )2 + · · · + (xn − yn )2
 
d∞ = max |x1 − y1 |, |x2 − y2 |, . . . , |xn − yn |

Eg,
 
d1 (1, 2, 3), (−1, 2, 4) = |1 − (−1)| + |2 − 2| + |3 − 4| =3
  p √
d2 (1, 2, 3), (−1, 2, 4) = (1 − (−1))2 + (2 − 2)2 + (3 − 4)2 = 5
 
d∞ (1, 2, 3), (−1, 2, 4) = max(2, 0, 1) =2

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 5 / 52

Metrics
A related concept of a norm (length of an element in a vector space) will be used
in the Fourier series section of the course. It’s definition is not given here.

If ||x|| is the norm of x ∈ Rn , then for x, y ∈ Rn , d (x, y) = ||x − y|| defines a

metric.

Definition
Two metrics d and δ are equivalent if there exists constants

0<c <C <∞

such that
cδ(x, y) ≤ d (x, y) ≤ C δ(x.y) ∀x, y ∈ Rn .

[This is an equivalence relation as studied in MATH1081.]

Problem 8 on tutorial sheet 2 shows d2 and d∞ are equivalent. (dp and d∞ are
also equivalent.)
JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 6 / 52
Limits of sequences

Definition
A ball around a ∈ Rn of radius  > 0 is the set

B(a, ) = {x ∈ Rn : d (a, x) < }.

Think of “x is close to a” as meaning that x ∈ B(a, ) for some small postive .

Definition
For a sequence {xi } of points in Rn we say x is the limit of the sequence {xi } if
and only if
∀ > 0 ∃N such that n ≥ N ⇒ d (x, xn ) < 
or equivalently

∀ > 0 ∃N such that n ≥ N ⇒ xn ∈ B(x, ).

If x is the limit of the sequence {xi } then for each postive  there is a point in the
sequence beyond which all points of the sequence are inside B(x, ).
JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 7 / 52

Limits of sequences

x is the limit of the sequence of

blue dots.

y is the not limit of the

sequence of blue dots.
y
If x is the limit of the sequence
x xi then for all positive  there
are always infinitely many points
of the sequence inside B(x, ).

[The converse is not true. Can

you think of a counter example?]

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 8 / 52

Limits of sequences

Sketch the following.

The unit ball in R2 using d1
The unit ball in R2 using d2
The unit ball in R2 using d3
The unit ball in R2 using d4
The unit ball in R2 using d∞
The unit ball in R3 using d2
The unit ball in R3 using d∞

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 9 / 52

Limit example
 
1 −k
Let xk = 2 − ,e , k = 1, 2, 3, . . .. Show that lim xk = (2, 0).
k k→∞

Let x = (2, 0). How big does k need to be to ensure that xk ∈ B(x, ),
xk+1 ∈ B(x, ), . . . etc?
s
 2 r
1 2 1
d (x, xk ) = 2 − − 2 + (e −k − 0) = + e −2k
k k2
[We don’t need the smallest k that makes xk , xk+1 , xk+2 ∈ B(x, ).]
1
Since 2 > e −2k for k ≥ 1,
k
r √
1 1 2
d (x, xk ) < + = .
k2 k2 k
l√ m
So if we take K = 2 then
√ √
2 2
k > K ⇒ d (x, xk ) < < ≤  ⇒ xk ∈ B(x, ).
k K
JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 10 / 52
Limit example
 
0 1 −k
Show that x = (0, 0) is not the limit of xk = 2 − , e .
k

x1
x2
x′ x3
∀ > 0 ∃K such that
1 k ≥ K ⇒ xk ∈ B(x, ).

Using d1 : B1 (x, 12 ) does not contain infinitely many members of the sequence.
Using d2 : B2 (x, 12 ) does not contain infinitely many members of the sequence.
Using d∞ : B∞ (x, 12 ) does not contain infinitely many members of the sequence.

The limit of the sequence is not x0 for any of these equivalent metrics.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 11 / 52

Limits of sequences

Theorem
A sequence xk converges to a limit x
⇔ the components of xk converge to the components of x.
⇔ d (xk , x) → 0.

We can use any equivalent distance function (metric). Why? See tutorial sheet 2
problems 7 and 8.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 12 / 52

Limits and equivalent metrics

Suppose d and δ are two equivalent metrics. That is,

cd (x, y) ≤ δ(x, y) ≤ Cd (x, y)

for some strictly positive constants c and C .

Now, using d as the metric, suppose

xk → x for xk , x ∈ Rn

that is,
∀ > 0 ∃K such that k ≥ K ⇒ d (xk , x) <  (∗)
We want to make a similar statement using δ.
∀0 > 0 choose  so that 0 = C . Since  > 0, (∗) says ∃K such that

k ≥ K ⇒ d (xk , x) <  ⇒ δ(xk , x) ≤ Cd (xk , x) < C  = 0

that is δ(xk , x) < 0 . Hence xk → x using the metric δ.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 13 / 52

Proof of first part of theorem (⇒)

We can use any equivalent metric. Let’s use d∞ .

Suppose xk → x for xk , x ∈ Rn , that is

∀ > 0 ∃K such that k ≥ K ⇒ d∞ (xk , x) < .

Now, for any i = 1, 2, . . . , n, (ie for any component)

 
|xk,i − xi | ≤ max |xk,1 − x1 |, |xk,2 − x2 |, . . . , |xk,n − xk |
= d∞ (xk , x).

Hence ∀ > 0, there is a K (the same K as above) such that

k ≥ K ⇒ |xk,i − xi | < 

and so
xk,i → xi .

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 14 / 52

Proof of first part of theorem (⇐)
If all of the components of xk converge, then ∀ > 0,
∃K1 such that k ≥ K1 ⇒ |xk,1 − x1 | < 
∃K2 such that k ≥ K2 ⇒ |xk,2 − x2 | < 
..
.
∃Kn such that k ≥ Kn ⇒ |xk,n − xn | < 
If we take K = max(K1 , K2 , . . . , Kn ) then
k ≥ K ⇒ |xk,1 − x1 | < 
k ≥ K ⇒ |xk,2 − x2 | < 
..
.
k ≥ K ⇒ |xk,n − xn | < 
 
⇒ d∞ (xk , x) = max |xk,1 − x1 |, |xk,2 − x2 |, . . . , |xk,n − xk | < .
Hence
∀ > 0 ∃K such that k ≥ K ⇒ d∞ (xk , x) < .
JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 15 / 52

Cauchy sequences

We can define convergence without knowing the limit of a sequence.

Definition
A sequence {xk } in Rn is a Cauchy sequence if

∀ > 0 ∃K such that k, l > K ⇒ d (xk , xl ) < .

Theorem
A sequence {xk } converges in Rn ⇔ {xk } is a Cauchy sequence.

Proof.
 
“⇒” d (xk , xl ) ≤ d (xk , x) + d (x, xl ) <
+ =
2 2
“⇐” This depends on how R is constructed. One common definition of R is
that it is the set of limits of Cauchy sequences.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 16 / 52

Open and closed sets

Definition
Consider Ω ⊂ Rn .
x0 ∈ Ω is an interior point of Ω if there is a
ball around x0 contained in Ω.
Ω is open if every point of Ω is an interior
point. Ω
Ω is closed if its complement is open.
x0 ∈ Rn is a boundary point of Ω if every
ball around x0 contains both points in Ω
and points not in Ω.

Theorem
Ω ⊂ Rn is closed ⇔ it contains all of its boundary points.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 17 / 52

Examples of open and closed sets

on R [a, b] closed
(a, b) open
R open and closed
∅ open and closed
[a, b) neither
Q neither
 −1
k : k ∈ Z+ neither
on Rn B(x0 , ) open

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 18 / 52

Open and closed sets

Prove that for x0 ∈ Rn and  ∈ R+

B(x0 , ) = {x ∈ Rn : d (x, x0 ) < } δ

x′
is an open subset of Rn . x
r
For each x ∈ B(x0 , ) we need to show that there
is a δ such that B(x, δ) ⊂ B(x0 , ). x0
ǫ
Choose δ =  − r where r = d (x0 , x).

Want to show x0 ∈ B(x, δ) ⇒ x0 ∈ B(x0 , ).

Now, by the triangle inequality,

d (x0 , x0 ) ≤ d (x0 , x) + d (x, x0 ) < r + δ = r +  − r = .

So x0 ∈ B(x0 , ) and hence B(x, δ) ⊂ B(x0 , ).

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 19 / 52

Open and closed sets

Definition
x0 is a limit point (or accumulation point) of Ω if there is
a sequence {xi } in Ω with limit x0 and xi 6= x0 .
Ω
Every interior point of Ω is a limit point of Ω.
x0 is not necessarily in Ω.
A set is closed ⇔ it contains all of its limit points.
Definition
The interior of Ω is the set of all interior points of Ω.
The boundary of Ω is the set of boundary points of Ω (denoted ∂Ω).
The closure of Ω is Ω ∪ ∂Ω (denoted Ω̄).

Eg, Q̄ = R.
The interior of Ω is the largest open subset of Ω.
The closure of Ω is the smallest closed set containing Ω.
JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 20 / 52
Open and closed sets

Tutorial sheet 2 Q4
i) The interesection and union of two open sets is open.
ii) The intersection and union of two closed sets is closed.
What about countable intersections and unions?

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 21 / 52

Open and closed sets

Tutorial sheet 2 Q5: S closed ⇔ S contains all of its limit points.

Let’s prove “⇐”

Suppose S contains all of its limit points.

[We want to show that S c is open, ie S c contains only interior points.]

Let x ∈ S c . Now, x is either an interior point of S c or a boundary point of S c .

Assume x is a boundary point of S c . That is, all balls around x contain a point in
S c and a point in (S c )c = S. So x is also a boundary point of S.

Next, construct a sequence {xk } by choosing xk ∈ B x, k1 ∩ S

[Note xk 6= x because x 6∈ S. Also, this makes sense because we have already

shown that every ball around x contains a point in S.]

Since k1 → 0 as k → ∞, it is clear that xk → x, that is, x is a limit point of S and

hence x ∈ S.

This is a contraction and so S c is open and hence S is closed.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 22 / 52
Limit of a function at a point
Definition
For f : Ω ⊂ Rn → Rm , lim f(x) = b means
x→x0

∀  > 0 ∃ δ > 0 such that for x ∈ Ω

0 < d (x, x0 ) < δ ⇒ d (f(x), b) < 

or alternatively
x ∈ B(x0 , δ) \ {x0 } ⇒ f(x) ∈ B(b, ).

f
Ω
x b
f(x)
x0

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 23 / 52

Limits example
Show that for f : R2 \ {(0.0)} → R with

x4 + x2 + y2 + y4
f (x, y ) =
x2 + y2

lim f (x, y ) = 1.
(x,y )→(0,0)

x4 + x2 + y2 + y4 x4 + y4
d (f (x, y ), 1) = −1 = 2
x2 + y2 x + y2
p
d ((x, y ), (0, 0)) = x2 + y2

x4 + y4 x 4 + 2x 2 y 2 + y 4 (x 2 + y 2 )2 2 2
p
2 2
2
≤ = =x +y = x +y
x2 + y2 x2 + y2 x2 + y2
√
If we choose δ =  then

0 < d ((x, y ), (0, 0)) < δ ⇒ d (f (x, y ), 1) < .

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 24 / 52

Limits

Note: if a limit exists for f , then f approaches that limit along any path.

This can be used to show a limit does not exist, eg,

Show that
lim f (x, y )
(x,y )→(0,0)

does not exist for

xy
f (x, y ) =
x2 + y2
x 2y
f (x, y ) = 4
x + y2

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 25 / 52

Limits
xy
f (x, y ) = , f : R2 \ 0 → R.
x2 + y2

Approach along positive x-axis (ie y = 0)

xy 0
lim = lim = 0.
(x,y )→(0,0) x 2 + y 2 x→0+ x 2

Similarly for approaching along the positive y -axis.

Approach along the line y = x in the first quadrant
xy x x 1 1
lim = lim = lim = .
(x,y )→(0,0) x 2 + y 2 x→0+ x 2 + x 2 x→0+ 2 2
Approach along the line y = mx in the first quadrant
xy x mx m m
lim = lim = lim = .
(x,y )→(0,0) x 2 + y 2 x→0+ x 2 + m2 x 2 x→0+ m2 + 1 m2 + 1
A different limit is reached approaching (0, 0) along different straight lines.
Hence the limit of f (x, y ) as (x, y ) → (0, 0) does not exist.
JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 26 / 52
Limits
x 2y
f (x, y ) = 4 , f : R2 \ 0 → R.
x + y2

x 2 mx mx 3
f (x, mx) = = 4
x 4 + m2 x 2 x + m2 x 2
mx
= → 0 as x → 0.
x + m2
2

2 x 2 ax 2 a
f (x, ax ) = = .
x 4 + a2 x 4 1 + a2
Hence a different limit is attained by approaching along different parabolas and so
lim f (x, y ) does not exist.
(x,y )→(0,0)

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 27 / 52

Limits
For f : Rn → Rm it is sufficient to consider the limits of the components of f.

Example

f : R2 \ {(0, 0)} → R2

x3 x 2 + y 2 + x 2y 2
 
f (x, y ) = ,
x2 + y2 x2 + y2

If we are given that

x3
lim = 0
(x,y )→(0,0) x 2 + y 2

x 2 + y 2 + x 2y 2
lim = 1
(x,y )→(0,0) x2 + y2

then
lim f (x, y ) = (0, 1).
(x,y )→(0,0)

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 28 / 52

Algebra of limits

Theorem
For
f : Rn → R and g : Rn → R
with
lim f (x) = a and lim g (x) = b.
x→x0 x→x0

Then

lim (f + g )(x) = a+b

x→x0
lim (fg )(x) = ab
x→x0
lim (f /g )(x) = a/b provided b 6= 0.
x→x0

For f : Rm → R, g : Ω ⊂ Rn → Rm , is it true that lim (f ◦ g )(x) = f (g (x0 ))?

x→x0
For this we need continuity.
JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 29 / 52

Algebra of limits
If we can prove
lim c = c and lim xi = ai
x→a x→a

then we can use the algebra of limits to find limits for rational functions.

To prove lim c = c, for each  > 0 choose δ = 1.

x→a

To prove lim xi = ai , for each  > 0 choose δ = .

x→a

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 30 / 52

Continuity

Definition
f : Ω ⊂ Rn → Rm is continuous at x0 ∈ Ω means either
i) x0 is a limit point of Ω, lim f(x) exists and equals f(x0 ), or
x→x0
ii) x0 is not a limit point of Ω.
f is continuous on Ω if it is continuous at each point of Ω.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 31 / 52

Continuity

Theorem
Suppose that f : Ω ⊂ Rn → Rm and x0 ∈ Ω. The following are equivalent.
i) f is continuous at x0 ∈ Ω.
ii) ∀ > 0 ∃δ > 0 such that for x ∈ Ω, d (x, x0 ) < δ ⇒ d (f(x), f(x0 )) < .
[Ie x ∈ B(x0 , δ) ⇒ f(x) ∈ B(f(x0 ), ).]
iii) ∀ sequences {xk } in Ω with limit x0 , {f(xk )} converges to f(x0 ).
iv) f(x0 ) is an interior point of f(Ω) ⇒ x0 is an interior point of Ω.

Theorem
Suppose that f : Ω ⊂ Rn → Rm . The following two statements are equivalent.
f is continuous on Ω.
U is open in Rm ⇒ f −1 (U) is open in Rn .

The preimage f −1 (U) = {y ∈ Rn : f(y) ∈ U}.

The second statement is an alternative definition of continuity. (Tutorial sheet 2.)
JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 32 / 52
Continuity (iii) ⇒ (i) [¬(i) ⇒ ¬(iii)]
 
¬ ∀ > 0 ∃δ > 0 ∀x ∈ B(x0 , δ) ⇒ f(x) ∈ B(f(x0 , )
 
∃ > 0 ¬ ∃δ > 0 ∀x ∈ B(x0 , δ) ⇒ f(x) ∈ B(f(x0 , )
 
∃ > 0 ∀δ > 0 ¬ ∀x ∈ B(x0 , δ) ⇒ f(x) ∈ B(f(x0 , )

∃ > 0 ∀δ > 0 ∃x ∈ B(x0 , δ) f(x) 6∈ B(f(x0 , )

That is,

∃ > 0 such that ∀δ > 0 ∃x ∈ B(x0 , δ) such that f(x) 6∈ B(f(x0 , ) (∗)

Choose an  verifying (∗).

In each ball B(x0 , k1 ) choose xk ∈ B(x0 , k1 ) ∩ Ω such that f(xk ) 6∈ B(f(x0 ), ).
[Can do this because of (∗).]
This is a sequence with xk → x0 but f(xk ) 6→ f(x0 ).
JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 33 / 52

Continuity (iii) ⇒ (i) [¬(i) ⇒ ¬(iii)]

Rm

Rn

f(x0 )

x0

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 34 / 52

Continuity — alternative definition
Tutorial sheet 2 Q15 gives another Note: a continuous function can map
definition of continuity. open sets to closed sets.

y y

f (V )
U

f −1 (U) x V x

f is not continuous. U is open but f is continuous. V is open but f (V ) is

f −1 (U) is not open. not open.
JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 35 / 52

Algebra of continuous functions

Theorem
A function f : Ω ⊂ Rn → Rm is continuous on Ω if and only if its component
functions are continuous.

Theorem (Algebra of continuous functions)

For two functions continuous on Ω

f : Ω ⊂ Rn → R and g : Ω ⊂ Rn → R

f + g , fg and f /g are continuous. [The domain of f /g must exclude points where

g (x) = 0.]

Note also that for

f : Rm → R and g : Ω ⊂ Rn → Rm

f ◦ g is continuous where it makes sense.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 36 / 52

Compact and connected sets

Definition
A set Ω ⊂ R n is bounded if there is an M such that d (x, 0) ≤ M for all x ∈ Ω.

Example: Example:

Ω ⊂ B(0, M) B(x0 , ) is bounded ∀x0 ∈ Rn and

∀ > 0.

Proof:
Ω
for x ∈ B(x0 , ) choose
M =  + d (x0 , 0).

d (x, 0) ≤ d (x, x0 ) + d (x0 , 0)

M
<  + d (x0 , 0)
= M.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 37 / 52

Compact and connected sets

Example:

The set
Ω = {(x, y ) ∈ R2 : xy ≤ 1}
is not bounded.
Suppose Ω ⊂ B(0, M).

Now, (M + 1, 0) ∈ Ω since
Ω (M + 1).0 = 0 ≤ 1. But
 
d (M + 1, 0), (0, 0) = M + 1 > M

so (M + 1, 0) 6∈ B(0, M) and hence

(M + 1, 0) 6∈ Ω.

This is a contradiction and so

Ω 6⊂ B(0, M). Hence Ω is not bounded.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 38 / 52

Monotone convergence theorem

Theorem (Monotone convergence theorem)

A bounded monotone sequence in R converges to a limit in R.

This relies on the existence of a least upper bound for bounded set in R. (Note
that Q does not have this property.)

Lemma
Every bounded sequence in R has a monotone subsequence.

Theorem
Every bounded sequence in R has a convergent subsequence with a limit in R.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 39 / 52

Bolzano-Weierstrass theorem

Theorem
For Ω ⊂ Rn , the following are equivalent.
(i) Ω is closed and bounded.
(ii) Every sequence in Ω has a subsequence that converges to an element of Ω.

A third equivalent statement that is beyond the scope of this course is given by
the Heine-Borel theorem.
(iii) Whenever the union of a collection of open sets contains Ω there is always a
finite sub-collection thats union also contains Ω.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 40 / 52

Bolzano-Weierstrass theorem, proof of (i) ⇒ (ii)
1, 0, 21 , −1, 12 , − 12 ,



Suppose Ω is closed and bounded and let {xk } be a
sequence in Ω. 1

3


, 0, 0 , −1, 0, − 4 ,
The first components form a bounded sequence in R. 2
1 2 7 7



Choose a subsequence for which the first ,
3 3 8 , , −1, 0, − 8 ,
components converge (to x0,1 ). 1 15

3 15


4 , 0, 16 , −1, 4 , − 16 ,
Choose a subsequence of this subsequence for which
1 31 31



the second components converge (to x0,2 ). 5 , 0, 32 , −1, 0, − 32 ,
1 4 63



Repeat for each component. ,
6 5 , 0 , −1, 0, − 64 ,
We now have a subsequence {xkl } that converges to 1 , 0, 127
,
7 128
x0 = (x0,1 , x0,2 , . . . , x0,n ). But is x0 ∈ Ω?
−1, 56 , − 127


c 128 ,
Suppose x0 ∈ Ω which is open because Ω is closed.
1 255


, 0, 256 ,

So, ∃ > 0 such that B(x0 , ) ⊂ Ωc and so no 8
255
element of {xkl } is in B(x0 , ). −1, 0, − 256 ,
1 6 511


This contradicts xkl → x0 . So x0 ∈ Ω. 9 7 512 ,
, ,
−511


That is, {xk } has a convergent subsequence with limit in −1, 0, − 512 , . . .

1, 0, 12 ,


Ω. , , ,
1



JM Kress (UNSW Maths & Stats) MATH2111 Analysis 2 , 0, 0 ,
Semester 1, ,
2014 , 41, / 52
1 2 7



,
3 3 8 , , , , ,
1 15



, 0, , , , ,
Bolzano-Weierstrass theorem, ¬(i) ⇒ ¬(ii),4 ie (ii) 16 ⇒ (i)
1 31



5 , 0, 32 , , , ,
If Ω is not bounded, 1 4
, , 0


, , ,


,
6 5
I choose a sequence {x } with d (x , 0) > k.
k k
1 127


I This sequence has no convergent subsequence.
7 , 0, 128 ,

If Ω is not closed, , , ,
I it has a boundary point x not in Ω. 1 255


0
8 , 0, 256 ,
1
I Construct a sequence {x } by choosing x
k k to be any point in B(x0 , k ) ∩ Ω
, , ,
which must be non-empty.
I Clearly {x } converges with limit x and hence any subsequence 1 6 511 converges


9 , 7 , 512 ,
k 0
with limit x0 .

, ,
, ...

That is, we have constructed a sequence with no convergent 1, 0, 2 ,1 subsequence
, , ,
with limit in Ω.
1



2 , 0, 0 , , , ,
1 2 7



,
3 3 8 , , , , ,



, , , , , ,



, , , , , ,
1 4



,
6 5 , 0 , , , ,


JM Kress (UNSW Maths & Stats) MATH2111 Analysis , , Semester, 1,
2014 42 / 52
Compact sets

Definition
A set Ω is compact if it satisfies either property from the Bolzano-Weierstrass
theorem.

Examples:

∅ compact
R not compact
(0, 1) not compact
[0, 1] compact
[0, 1] ∪ [3, 4] compact
[0, 1] × [0, 1] compact
S 2 (the 2-sphere) compact
Cantor set compact

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 43 / 52

Path connected sets

Definition
A continuous path between x, y ∈ Rn is a function φ : [0, 1] → Rn such that φ is
continuous and φ(0) = x, φ(1) = y.

Definition
A set Ω ⊂ Rn is said to be path connected if, given any x, y ∈ Ω, there is a
continuous path between them that lies entirely in Ω (ie ∀t ∈ [0, 1] φ(t) ∈ Ω).

Example: For Ω = {(x, y ) : 1 < x 2 + y 2 < 4} show that

Ω is path connected.
  x
 r1 cos θ(t), r1 sin θ(t) for 0 ≤ t ≤ 12
φ(t) =  
 r (t) cos θ , r (t) sin θ
2 2 for 12 < t ≤ 1

with θ(t) = θ1 + 2(θ2 − θ1 )t, r (t) = r1 + 2(r2 − r1 )(t − 12 ) y

and x = (r1 cos θ1 , r1 sin θ1 ), y = (r2 cos θ2 , r2 sin θ2 ).
JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 44 / 52
Path connected sets
Show that Ω = {(x, y ) : xy > 1} is not path connected.

Suppose that Ω is path connected. So there must be a continuous path between

(−2, −2) and (2, 2) that lies entirely in Ω. That is, there is a continuous function
φ : [0, 1] → R2 with φ(0) = (−2, −2), φ(1) = (2, 2), φ(t) ∈ Ω ∀t ∈ [0, 1].

The first component of φ

Ω
ie φ1 : [0, 1] → R
(2, 2)
must be continuous on [0, 1] with φ1 (0) = −2
and φ1 (1) = 2. So the Intermediate Value
Theorem says
(−2, −2)
∃ c ∈ [0, 1] such that φ1 (c) = 0

and hence φ(c) 6∈ Ω. This is a contradiction and

hence Ω can not be path connected.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 45 / 52

Big theorems

Theorem
Let f : Ω ⊂ Rn → Rm be continuous. Then
(i) K ⊂ Ω and K is compact ⇒ f(K ) is compact.
(ii) B ⊂ Ω and B is path connected ⇒ f(B) is path connected.

Proof (i):
Let f be continuous, K compact and {yk } be a sequence in f(K ).
So there is {xk } such that xk ∈ K and yk = f(xk ).
K compact ⇒ there is a convergent subsequence {xkl } with limit x ∈ K .
f continuous ⇒ {f(xkl )}, that is, {ykl } is a convergent subsequence of {yk } with
limit f(x) = y ∈ f(K ).
That is, we have shown that for any sequence in f(K ), there exists a convergent
subsequence with limit in f(K ) and hence f(K ) is compact.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 46 / 52

Big theorems
Proof (ii):
Let f be continuous, B path connected and y1 , y2 ∈ f(B).
So there are x1 , x2 ∈ B such that y1 = f(x1 ) and y2 = f(x2 ).
B is path connected means there is a continuous function φ : [0, 1] → B such that

φ(0) = x1 , φ(1) = x2 and φ(t) ∈ B ∀t ∈ [0, 1].

Since f is continuous, f ◦ φ : [0, 1] → f(B) is continuous with

(f ◦ φ)(0) = f(φ(0)) = f(x1 ) = y1

(f ◦ φ)(1) = f(φ(1)) = f(x2 ) = y2
(f ◦ φ)(t) ∈ f(B) for t ∈ [0, 1].

That is, f ◦ φ is a continuous path between y1 and y2 contained in f(B).

Hence f(B) is path connected.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 47 / 52

Min/max theorem for f : R → R

For f : K ⊂ R → R For f : K ⊂ R → R not For f : Ω ⊂ R → R

continuous on a compact continuous on a compact continuous on a
set K , maximum and set K , maximum and non-compact set Ω,
minimum values are minimum values may or maximum and minimum
attained. may not be attained. values may or may not be
attained.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 48 / 52

Intermediate Value Theorem for f : R → R

For f : B ⊂ R → R For f : B ⊂ R → R For f : B ⊂ R → R not

continuous on a path continuous on a not path continuous on a path
connected set B, f (B) is connected set B, f (B) is connected set B, f (B) is
path connected. not necessarily path not necessarily path
connected. connected.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 49 / 52

Big theorems
Consider

S1 = {(x, y ) : x 2 + y 2 ≤ 1} S2 = {(x, y ) : x 2 + y 2 < 1}

Is there a continuous function f : R2 → R2 such that

1 f(S1 ) = S2 ?
2 f(S2 ) = S1 ?
3 f(R2 ) = S2 ?
4 f(R2 ) = S1 ?
5 f(S2 ) = R2 ?
6 f(S1 ) = R2 ?

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 50 / 52

Big theorems
Consider S1 = {(x, y ) : x 2 + y 2 ≤ 1} and S2 = {(x, y ) : x 2 + y 2 < 1}.
S1 , S2 and R2 are path connected but only S1 is compact.

1. S1 is compact and S2 is not. So 3. Consider the function f3 : R2 → R2

there can not be a continuous described in polar coordinates by
function f1 with f1 (S1 ) = S2 .  
2
2. Consider the function f2 : R2 → R2 (r , θ) → tan−1 r , θ .
described in terms of polar π
coordinates by
This is continuous and f3 (R2 ) = S2 .
4. f2 ◦ f3 is a continuous function that
(
(2r , θ) for r < 12
(r , θ) → maps R2 to S1 .
(1, θ) for r ≥ 12 .
5. f5 = f3−1 is a continuous function
This is continuous and f2 (S2 ) = S1 . with f (S2 ) = R2 .
6. S1 is compact and R2 is not. So
there can not be a continuous
function f6 with f6 (R2 ) = S1 .
JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 51 / 52

Example
Prove that if the temperature is above 0 somewhere on the Earth’s surface and
below 0 somewhere else, then there must be a third point where it is exactly 0.

The surface of the Earth S 2 is compact and path connected and (assume) that
the tempature T : S 2 → R is continuous.

So the image of S 2 under T , T (S 2 ), must compact and path connected. That is

T (S 2 ) is a closed bounded interval [a, b].

There is a point x where T (x) < 0 and y where T (y) > 0. That is, [a, b] contains
both positive and negative values and hence 0 ∈ [a, b].

Hence there is u ∈ S 2 such that T (u) = 0.

JM Kress (UNSW Maths & Stats) MATH2111 Analysis Semester 1, 2014 52 / 52