Assignment 5 MATH2320

Jaron Lee
May 19, 2014
1
1.i
We rst observe that because each f F is a continuous function over a compact set X, we have
that each f is uniformly continuous. Hence: f F > 0
f,
> 0 : d(x, x

) <
f,
=
|f(x) f(x

)| < .
To show that the set F is uniformly equicontinuous, we must prove that there exists a single that
works for every function in F.
We choose an open cover of F such that

fF
B

(f) F. Because F is compact, there exists a

nite subcover of this cover such that B

(f
1
) . . . B

(f
n
) F.
Dene := min{
f
1
, . . . ,
f
n
}.
Pick some f F. Then it must be true that f B

(f
i
) for some i. Hence it is true that
d
u
(f, f
i
) < .
Assume that d(x, x

) < . Then |f
i
(x) f
i
(x

)| < . We now want to show that it follows that

|f(x) f(x

)| < .
|f(x) f(x

)| = |f(x) f
i
(x) + f
i
(x) f
i
(x

) + f
i
(x

) f(x

)|
|f(x) f
i
(x)| +|f
i
(x) f
i
(x

)| +|f
i
(x

) f(x

)|
< + +

=
The rst and third terms are less than because each f is in a ball centred around a certain f
i
.
The second term is less than because each f
i
is continuous.
Hence there exists some which works for every f F. As each f is uniformly continuous, we
have that F is uniformly equicontinuous.
1.ii
First, we must show that if (X, d) is a compact metric space and (f
n
) a sequence of continuous real
valued functions converging uniformly on X to a continuous function, then {f
n
} is equicontinuous
on X.
1
Since (f
n
) is uniformly convergent it is uniformly Cauchy. Then |f
n
(x) f
N
(x)| < x
X, n > N.
Continuous functions are uniformly continuous on compact sets. Then there is a > 0 such that
|f
i
(x) f
i
(y)| < for all i = 1, . . . , N and all x, y X with d(x, y) < .
If n > N and d(x, y) < then:
|f
n
(x) f
n
(y)| |f
n
(x) f
N
(x)| +|f
N
(x) f
N
(y)| +|f
N
(y) f
n
(y)|
3
Thus we select a
eq
= min(,
1
, . . . ,
N
) where each
1
, . . . ,
N
corresponds to the delta for each
uniformly continuous f
1
, . . . , f
N
.
Therefore {f
n
} is equicontinuous on X.
We consider the family of functions f
n
(x) =
1
n
sin(n
2
x).
As n then f
n
(x) 0. This function is continuous.
We want to show that f
n
is uniformly convergent. We want to show that N N x X :
n > N = |f
n
(x) f(x)| < .
First assume n > N. Then:
|f
n
(x) f(x)| = |
1
n
sin(n
2
x) 0|
= |
1
n
sin(n
2
x)|
|
1
n
|
=
1
n
Hence if we set n =
1

we are done. Hence f

n
is uniformly convergent, and so by the above lemma
we have that {f
n
} is equicontinuous. By Theorem 15.6.2, we know that since each f is over the
compact interval [0, 2], the equicontinuous set {f
n
} is uniformly equicontinuous.
2
2.i
We have that Q = {(a
n
) l
1
:

n|a
n
| 1}.
First we observe that

n|a
n
| =

N
n=1
n|a
n
| 1. Hence

N
n=1
n|a
n
| > 1.
Pick (b
n
) / Q. Then

n|b
n
| > 1. We are able to pick a corresponding delta such that

n|b
n
| >
1 + , > 0.
We pick N such that

N
n=1
n|b
n
| 1 + .
2
Then combining

N
n=1
n|b
n
| 1 + and

N
n=1
n|a
n
| > 1 we obtain:
N

n=1
n|b
n
| +
N

n=1
n|a
n
| 1 + 1

N

n=1
n(|b
n
| |a
n
|)

N

n=1
n(|b
n
| |a
n
|)

n=1
N(|b
n
| |a
n
|)

n=1
N||b
n
| |a
n
|| by triangle inequality
This nal result allows us to state:

N

N

n=1
||b
n
| |a
n
||

n=1
|b
n
a
n
| by the reverse triangle inequality

n=1
|b
n
a
n
|
= ||b
n
a
n
||
We have just shown that the distance between a point a
n
Q and b
n
Q
c
is always greater than
some xed number

N
. Thus we can always take some open ball of nonzero radius around any
point in Q
c
that is contained entirely in Q
c
. Thus Q
c
is open and therefore Q is closed.
We now show that the set Q is totally bounded.
Let > 0. Then choose N() :
1
N+1
<

2
. We dene a set of ball centres A = {(k
1

2N
, k
2

2N
, . . . , k
N

2N
, 0, . . .) :
k
i

2N
[1, 1], k
i
Z}.
Let (a
n
) Q be arbitrary. For each a
n
, pick k
n
: |k
n

2N
a
n
| <

2N
n N.
Now we can assemble a sequence out of these k
n
we have just chosen; call it (b
n
). This is the centre
of the ball that contains (a
n
). Then (b
n
) = (k
1

2N
, k
2

2N
, . . . , k
N

2N
, 0, . . .). This does not have to
be in Q.
We now want to show that ||(a
n
) (b
n
)|| < in order to prove that (a
n
) indeed resides within the
ball centred at (b
n
) of radius .
First, a lemma. Let (a
n
) Q. Assume for the sake of contradiction that

n=N
|a
n
| >
1
N
. Then
this implies that

n=N
N|a
n
| > 1 which implies that

n=N
n|a
n
| > 1 since each n > N. But this
is a contradiction, since for (a
n
) Q it must be that

n=1
n|a
n
| 1. Therefore we conclude that

n=N
|a
n
|
1
N
and therefore

n=N+1
|a
n
|
1
N+1
.
3
||(a
n
) (b
n
)|| =

n=1
|a
n
b
n
|
=
N

n=1
|a
n
b
n
| +

n=N+1
|a
n
b
n
|
Note that |a
n
b
n
| = |k
n

2N
a
n
| <

2N
and that b
n
= 0 n > N.
<
N

n=1

2N
+

n=N+1
|a
n
|
Computing the rst summation and using the lemma for the second summation:
< N

2N
+
1
N + 1
=

2
+

2
=
2.ii
Dene a sequence of sequences in Q as (a
i
n
), where n denotes elements of the ith sequence, and
where i denotes the position of a sequence in the overall sequence of sequences.
Consider (a
i
1
), the sequence of the rst components of all sequences in (a
i
n
). We note that (a
i
1
) is a
sequence in the compact interval [1, 1] since |a
n
| 1. Therefore a convergent subsequence exists.
This implies a Cauchy subsequence exists.
We now select a subsequence of (a
i
n
). We do this by throwing away sequences in (a
i
n
) until we
obtain the Cauchy sequence specied above. Call this subsequence (a
i
n
) for ease of notation. Then
we conclude that (a
i
1
) is a Cauchy sequence of real numbers.
We create a new sequence of sequences (b
i
n
) by picking the rst sequence to be (b
1
n
) = (a
1
n
).
Consider (a
i
2
). This sequence lies in [1, 1] and a Cauchy subsequence exists by the above argument.
We then take (a
i
n
) which has already been altered by the above process and throw away sequences
to achieve the above Cauchy sequence. Call this new sequence (a
i
n
), again for notational ease.
(a
i
2
) is now a Cauchy sequence of real numbers.
Pick the second term of (b
i
n
) to be (b
2
n
) = (a
2
n
).
Repeat this process to ll out the sequence of sequences (b
i
n
). This sequence is a subsequence of
the original, by construction.
Moreover, if we inspect each sequence of components in (b
i
n
) we nd that it is Cauchy.
This means for each (b
j
n
) if we consider the sequences of sequences (b
i
n
) for i j then that is a
subsequence of a sequence whose sequence of jth components is Cauchy.
For each particular n, it follows that the sequence of nth components in (b
i
n
) is Cauchy.
4
We now want to show that (b
i
n
) Q is Cauchy. Since Q is closed if (b
i
n
) converges, the limit point
will be in Q.
Let > 0. Choose a nite N large enough such that
1
N+1
<

2
.
For 1 n N: pick I
n
large enough so that i, j I
n
= |b
i
n
b
j
n
| <

2N
.
This is possible since (b
i
n
) is Cauchy for each xed n.
Dene I := max(I
1
, . . . , I
N
). This is possible since N is nite.
Suppose i, j > I. Then:
||(b
i
n
) (b
j
n
|| =

n=1
|b
i
n
b
j
n
|
=
N

n=1
|b
i
n
b
j
n
| +

n=N+1
|b
i
n
b
j
n
|
Each term within the rst summation is bounded by

2N
by the selection of I. The second sum-
mation is bounded by
1
N+1
, as per the lemma in the previous part of this question.

2
+

2
=
Therefore this sequence is Cauchy. By completeness it converges. By the closedness of Q the limit
must be in Q. Hence proven.
3
3.i
Intuitively we can see that R
2
\{x} is path connected, because we can form a line between any two
points in R
2
without needing to pass through a particular point
The description of the path is as follows. Let two points a, b R
2
be given. Consider the point
x R
2
for the moment.
If it is the case that f : [0, 1] R
2
such that f(t) = a + t(b a) and f(t) = x t [0, 1], then
the path is given by f(t).
If it is true that f(t) = x for some value of t. Dene h(t) : R R
2
such that h(t) = a +
t(b a). Pick some point in R
2
\{h(t)}. Call this c. Then we create a new path g(t) where:

g(t) = a + 2t(c a), t [0, 0.5)

g(t) = c + 2t(b c), [0.5, 1]
.
Hence we have shown constructively that R
2
\{x} is path connected, since every point in R
2
that
is not {x} is path connected under our path system.
Path connectedness implies connectedness.
5
3.ii
We assume that a homeomorphism f exists from R to R
2
for the sake of contradiction.
We remove a point from R to create R\{y}. Then by the bijectivity of f we know that R
2
must
also be missing a point - call it R
2
\{x}.
But by Theorem 15.6.2 we know that the continuous image of a connected set is connected. The
continuous image of R
2
\{x} via f
1
is therefore R\{y}. But this is not connected as we can write
it as a union of two non-empty disjoint open sets, so we arrive at a contradiction.
3.iii
We assume for the sake of contradiction that a homeomorphism f exists from S
1
to S
2
. Then
by bijectivity if we remove a point from S
1
we also remove a point from S
2
. The result is a
homeomorphism f

from S
1
\{x} to S
2
\{y}.
Consider rst S
1
\{x}. We claim it is homeomorphic to the open interval (0, 1). Observe the
following parametrisation:
(0, 1) S
1
\{x}
g : t e
2it
This function is continuous since exponentials are continuous. The inverse is given by g
1
(t) =
ln(t)
2i
.
But this function is continuous since ln is continuous. This function is one-to-one and onto.
Consider the claim that the open interval (0, 1) is homeomorphic to R. This can be done with the
appropriate trigonometric function (shown in lectures).
Thus, we conclude that S
1
\{x} is homeomorphic to R.
We now claim that S
2
\{y} is homeomorphic to R
2
. This can be done with a stereographic projection
from the sphere into the plane.
However, we reach a contradiction because the above implies that R is homeomorphic to R
2
. Hence
no such homeomorphism exists from S
1
to S
2
.
4
4.i
Dene the unit ball centred at 0 as B
1
(0) = {y : (

n
k=1
x
2
k
)
1/2
1}. Let x = (x
1
, . . . , x
n
).
For property (1): (

n
k=1
(x
k
)
2
)
1/2
= (

k = 1
n
(x
k
)
2
)
1/2
. Hence x K = x K.
For property (2):s
For property (3): This is true since the centre of the ball 0 is located in the ball.
For property (4): This is true since by denition of a closed unit ball ||x|| 1 = M.
6
4.ii
||x||
k
= inf{t > 0 :
1
t
x K}.
If ||x||
k
is a norm then it must satisfy these three properties:
1. ||x||
k
0; ||x||
k
x = 0
2. ||x||
k
= ||||x||
k
; R
3. ||x + y||
k
||x||
k
+||y||
k
For property (1): ||x||
k
0 since t 0.
Want to show that x = 0 = ||x||
k
= 0.
x = 0
=
1
t
(0) = 0 K
= t > 0
= inf{t > 0 :
1
t
x K} = 0
= ||x||
k
= 0
Want to show that ||x||
k
= 0 = x = 0.
||x||
k
= 0
= inf{t > 0 :
1
t
x K} = 0
= any t > 0 will satisfy the equation
= x = 0 otherwise there would be some t large enough so that
1
t
x / K
For property (2):
||x||
k
= inf{t > 0 :
1
t
x K}
= inf{||t > 0 :
1
t
x K} using property (1) from part a
= || inf{t > 0 :
1
t
x K}
= ||||x||
k
For property (3):
||x + y||
k
= inf{t > 0 :
1
t
(x + y) K
= inf{t > 0 : (
1
t
x +
1
t
y) K}
inf{t > 0 :
1
t
x K} + inf{t > 0 :
1
t
y K}
7
We now want to show K is the unit ball. Consider that the unit ball has the property ||x||
k
1.
This is true i inf{t > 0 :
1
t
x K} 1. But then this implies
1
t
x K t 1.
4.iii
We are given that (x, y) R
E
x y E.
Equivalence relations must satisfy three properties.
1. (x, x) R
E
2. (x, y) R
E
(y, x) R
E
3.

(x, y) R
E
(y, z) R
E
= (x, z) R
E
For property (1):
x x = 0 E since a vector subspace must contain the zero vector.
For property (2):
Let xy E. But a subspace is closed under scalar multiplication so (xy) E which implies
that y x E. Hence (y, x) R
E
.
For property (3):
If x y E and y z E then since vector subspaces are closed under addition of vectors
belonging to the subspace we obtain x y + y z E. Therefore x z E which implies
(x, z) R
E
.
8