Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Theorem 0.1. Let x ∈ R. If 0 ≤ x < ϵ for all ϵ > 0, then x = 0.

Proof. Suppose the contrary, x > 0. Take ϵ := x/2, then 0 < ϵ < x. This
contradicts to 0 ≤ x < ϵ for all ϵ > 0.

Corollary 0.1. Let x ∈ R. If 0 ≤ x ≤ ϵ for all ϵ > 0, then x = 0.

Proof. Suppose the contrary, x > 0. Take ϵ := x/2, then 0 < ϵ < x. This
contradicts to 0 ≤ x ≤ ϵ for all ϵ > 0.

Definition 0.1. Let a ∈ R and ϵ > 0. Then the ϵ-neighborhood of a is the


set Vϵ (a) := {x ∈ R : |x − a| < ϵ}.

Theorem 0.2. Let a, b ∈ R. Suppose that a ≤ b + ϵ for all ϵ > 0.


Then a ≤ b.

Proof. Suppose the contrary, a > b. Take ϵ := (a−b)/2 > 0, then a ≤ b+ϵ =
b + (a − b)/2 = (a + b)/2 < a.Contradiction!!
Or we can ues the above corollary to prove. Suppose the contrary, a > b.
Then 0 < a − b ≤ ϵ for all ϵ > 0. Then by corollary, a=b, contradiction!!

Definition 0.2. Let X be a nonempty subset of R.

(a) u ∈ R is an upper bound of X if x ≤ u for all x ∈ S.

(b) l ∈ R is an lower bound of X if l ≤ x for all x ∈ S.

(c) X is bounded above if it has an upper bound.

(d) X is bounded below if it has a lower bound.

(e) X is bounded if it is both bounded above and bounded below. X is


unbounded if it is not bounded.

Definition 0.3. Let X be a nonempty subset of R.

(a) If X is bounded above, then u ∈ R is a supremum (or least upper


bound) of X if it satisfies the conditions:

(1) u is an upper bound of X.


(2) if v is any upper bound of X, then u ≤ v.

1
(b) If X is bounded below, then l ∈ R is a infimum (or greatest lower
bound) of X if it satisfies the conditions:

(1) l is an lower bound of X.


(2) if w is any lower bound of X, then w ≤ l.

Remark. (1) There can be only one supremum of X ⊆ R. If a supremum


exists, then it is unique.

(2) Suppose the supremum of X exists, it may or may not be an element of


X.

There are alternative ways of expressing the idea that u is the ”least” of the
upper bounds of X.

Proposition 0.1. Let X be a nonempty subset of R and X is bounded above.


Let u be an upper bound of X. Then the following statements are equivalent:

(i) if v is any upper bound of S, then u ≤ v.

(ii) if z < u, then z is not an upper bound of X.

(iii) if z < u, then there exists xz ∈ X such that z < xz .

(iv) if ϵ > 0, then there exists xϵ ∈ X such that u − ϵ < xϵ .

Proof. (i) =⇒ (ii). Suppose not, z is an upper bound of X. By (i), u ≤ z,


contradiction!!.
(ii) =⇒ (iii). definition of upper bound
(iii) =⇒ (iv). Let ϵ > 0. As u − ϵ < u ,by (iii), there exists x ∈ X such
that u − ϵ < x.
(iv) =⇒ (i). Let v be an upper bound of X. Suppose v ≤ u. Take ϵ = u−v >
0. By (iv), there exists x ∈ X such that u − ϵ < x, i.e. v = u − (u − v) < x.
This contradicts v is an upper bound of X.
Therefore, we can have alternate formulations for the supremum. When work
with limits, it is useful to express the condition in terms of ϵ > 0.

Corollary 0.2. Let X be a nonempty subset of R and u ∈ R.


Then u is the supremum of X if and only if

(1) x ≤ u for all x ∈ X

2
(2) For all ϵ > 0, there exists x ∈ X s.t. u − ϵ ≤ x.

The Completeness Property of R. Every nonempty set of real numbers


that has an upper bound also has a supremum in R.
Archimedean Property. If x ∈ R, there exists n ∈ N such that x ≤ n.
Proof. Let x ∈ R. Assume that n < x for all n ∈ N. Then x is an upper
bound of N. By the Completeness Property of R, N has a supremum u in
R. Then u-1 is not an upper bound of N. There exists N ∈ N s.t. u−1 < N .
So u < N + 1. By the Inductive Property of N, N + 1 ∈ N. The inequality
u < N + 1 contradicts that u is an upper bound of N.

Remark. 1. n depends on x.
2. N is not bounded in R. We must use the Completeness Property of R as
well as the Inductive Property of N (that is, if n ∈ N, then n + 1 ∈ N).

Corollary 0.3. If x ∈ R, there exists n ∈ N such that x < n.

Corollary 0.4. If ϵ > 0, there exists n ∈ N such that 0 < 1/n < ϵ.

Proof. Let ϵ > 0. From Corollary 0.3, there exists n ∈ N such that 1/ϵ < n,
i.e. 0 < 1/n < ϵ.

Characterization of intervals. An obvious property of intervals is that


if two points x, y with x < y belong to an interval I, then any point lying
between them also belongs to I. That is, if x < t < y, then the point t
belongs to the same interval as x and y. In other words, if x and y belong to
an interval I, then the interval [x, y] is contained in I.

Theorem 0.3 (Characterization of intervals). If S is a subset of R that


contains at least two points and has the property

if x, y ∈ S and x < y, then [x, y] ⊆ S (1),

then S is an interval.

Definition 0.4. A sequence of intervals In , n ∈ N, is nested if the following


chain of inclusions holds:

I1 ⊇ I2 ⊇ · · · ⊇ In ⊇ In+1 ⊇ · · ·

3
Theorem 0.4 (Nested Intervals Theorem). If In = [an , bn ], n ∈ N, is a
nested sequence of closed and bounded intervals, then there exists a number
ξ ∈ R such that ξ ∈ In for all n ∈ N.
Remark. It only shows the existence of a real number ξ ∈ In for all n ∈ N.
We need more conditions to guarantee the uniqueness.
Example. Let In = [0, 1 + 1/n], n ∈ N. Then we can show that

\
Ii = [0, 1]
i=1

Theorem 0.5 (Nested Intervals Theorem). (Existence and Uniqueness)


If In := [an , bn ], n ∈ N, is a nested sequence of closed and bounded intervals
such that the lengths bn − an of In satisfy

inf {bn − an : n ∈ N} = 0,

then there exists an unique number ξ ∈ In for all n ∈ N.


Theorem 0.6 (Characterization of Closed Sets). Let A ⊆ R. Then the
following statements are equivalent:
(i) A is closed.

(ii) If (xn ) is a sequence in A and limxn exists, then limxn ∈ A.


Proof. (i) =⇒ (ii). Let (xn ) be a sequence in A and x = limxn . Aims to
show x ∈ A. Suppose to the contrary that x ∈ / A. As R \ A is open, there
exists ϵ > 0 s.t. (x − ϵ, x + ϵ) ⊆ R \ A. Since x = limxn , there exists N ∈ N
s.t. xN ∈ (x − ϵ, x + ϵ), so xN ∈ R \ A. This contradicts to the assumption
xn ∈ A for all n ∈ N.
(ii) =⇒ (i). Suppose to the contrary that A is not closed, i.e. R \ A is
not open. Then there exists some x ∈ R \ A s.t. for all n ∈ N, there exists
xn ∈ C( R \ A) = A s.t. |xn − x| < 1/n. So limxn = x. Since xn ∈ A for all
n ∈ N, by assumption, x ∈ A, contradiction!!
The Characterization of Open Sets. The idea of an open set in R is a
generalization of the notion of an open interval.
Theorem 0.7 (Characterization of Open Sets). Let U ⊆ R. U is open if
and only if it is the union of countably many disjoint open intervals in R.

4
Definition 0.5. Let A ⊆ R. An open cover of A is a collection G = {Gα }
of open sets in R s.t. [
A⊆ Gα
α

Definition 0.6. Let A ⊆ R. A is said to be compact if every open cover


has a finite subcover.

Proposition 0.2. [a,b] is compact.

Proof. (First proof) Using the supremum (or least-upper-bound or complete-


ness) property of R.
For case a=b, the result is trivial. Assume a < b. Let {Ui }i∈I be any open
cover of [a,b], [
[a, b] ⊆ Ui
i∈I

Consider the set S of all the elements x ∈ [a, b] such that [a,x] is covered by
finite many Ui ’s. Then S ̸= ∅ as a ∈ S. And b is an upper bound of S, may
be or may not be in S. By completeness property of R, S has the least upper
bound. Let s := supS ∈ [a, b]. Aims to show s = b.
Assume s < b.
Show s > a. As there exists some j ∈ I s.t. a ∈ Uj , we can find ϵ > 0
s.t. (a − ϵ, a + ϵ) ⊆ Uj and therefore [a, a + ϵ/2] ⊆ Uj . So a + ϵ/2 ∈ S and
s ≥ a + ϵ/2 > a.
Now a < s < b from above and by assumption. Then there exists some i0 ∈ I
s.t. s ∈ Ui0 . Pick ϵ > 0 s.t. a ≤ s − ϵ < s < s + ϵ ≤ b and [s − ϵ, s + ϵ] ⊆ Ui0 .
As s − ϵ is not an upper bound of S, there exists x ∈ S s.t. s − ϵ S ≤ x ≤ s.
As x ∈ S, [a,x]Scan be covered by finitely many Ui ’s, i.e. [a, x] ⊆ nj=1 Uij .
So [a, s + ϵ] ⊆ nj=1 Uij ∪ Ui0 . It follows that s + ϵ ∈ S, contradicts that s is
the least upper bound of S. So s = b.
Now b = supS. We need to show that b ∈ S. There exists some i0 ∈ I s.t.
b ∈ Ui0 . Pick ϵ > 0 s.t. [b − ϵ, b] ⊆ Ui0 . As b − ϵ is not an upper bound of
S, there exists x ∈ S and b − ϵ ≤ x ≤ b. So [a,x] is covered by finite many
Ui ’s. As [a, b] = [a, x] ∪ [b − ϵ, b] ⊆ [a, x] ∪ Ui0 , [a,b] must be covered by finite
many Ui ’s. So b ∈ S.
Proof. (Second proof) Using the nested intervals theorem.
Let G = {Gα } be an open cover of [a,b]. Suppose the contrary that [a,b] is

5
not compact, we assum that it satisfies the following condition:

[a,b] is not contained in the union of any finite number of sets in G (1)
Below the proof, we say a set satisfies condition (1) if it is not contained in
the union of any finite number of sets in G.

Now, construct the nested sequence of closed and bounded intervals.


Let a0 := a, b0 := b, I0 := [a, b]. Bisect I0 into two closed subintervals
′ ′′
I0 := [a, a+b 2
] and I0 := [ a+b
2
, b]. At least one of the two intervals satisfies
′ ′
the condiotion (1). If I0 satisfies condiotion (1), we let I1 := I0 , otherwise
′′
let I1 := I0 . And we let a1 and b1 be the left and right end points of I1
respectively.
′ ′′
Similarly, Bisect I1 into two closed subintervals I1 := [a1 , a1 +b 2
1
] and I1 :=
′ ′ ′′
[ a1 +b
2
1
, b1 ]. If I1 satisfies condiotion (1), we let I2 := I1 , otherwise let I2 := I1 .
And we let a2 and b2 be the left and right end points of I2 respectively.
Continuing this process, we obtain a nested sequence of intervals (In ). By the
nested intervals theorem, there exists a number ξ ∈ R such that ξ ∈ In for
all n ∈ N. As each interval In satisfies condiotion (1), In contains infinitely
many points in [a,b] for all n ∈ N. Therefore, ξ is a cluster point of [a,b], we
can show that ξ ∈ [a, b] from the Characterization of Closed Sets and [a,b] is
closed.
Since ξ ∈ [a, b], there exists Gα ∈ G such that ξ ∈ Gα . As Gα is open, there
exists ϵ > 0 such that (ξ − ϵ, ξ + ϵ) ⊆ Gα . From construction,
bn−1 − an−1
length of In = bn − an =
2
b−a
=
2n

By Archimedean Property, there exists N ∈ N s.t. b−a 2N


< ϵ. So IN ⊆
(ξ − b−a
2N
, ξ + b−a
2N
) ⊆ (ξ − ϵ, ξ + ϵ) ⊆ Gα . This leads to contradiction that In
satisfies condiotion (1) for all n ∈ N.
Theorem 0.8 (Monotone Convergence Theorem). A monotone sequence of
real numbers is convergent if and only if it is bounded. Further:
(a) If (xn ) is a bounded increasing sequence, then
lim xn = sup{xn : n ∈ N}
n→∞

6
.

(b) If (yn ) is a bounded decreasing sequence, then

lim yn = inf{yn : n ∈ N}
n→∞

.
Proof. =⇒ A convergent sequence is bounded.

⇐= (a). Let (xn )n∈ N be a bounded increasing sequence of R. By complete-


ness property, the supremum x∗ := sup{xn : n ∈ N} exists in R.
Claim: x∗ = lim xn .
Let ϵ > 0. Then x∗ − ϵ is not an upper bound of the set {xn : n ∈ N}. There
exists some N ∈ N s.t. x∗ − ϵ < xN . As (xn ) is increasing, xN ≤ xn for all
n ≥ N , so x∗ − ϵ < xN ≤ xn ≤ x∗ < x∗ + ϵ for all n ≥ N .
(b). Let (yn )n∈ N be a bounded decreasing sequence of R. and let xn := −yn .
Then (xn )n∈ N is a bounded increasing sequence of R. By (a), lim xn =
sup{xn : n ∈ N}. As lim yn = − lim xn and sup{−yn : n ∈ N} = − inf{yn :
n ∈ N}, lim yn = inf{yn : n ∈ N}.
Theorem 0.9 (Monotone Subsequence Theorem). Every sequence of real
numbers has a monotone subsequence.
Remark. It shows the existence of a monotone subsequence but that subse-
quence is not unique. The sequence may have other monotone subsequence.
Example. X = (xn ) := ((−1)n )
X ′ = (x2n ) = (1)
X ′′ = (x2n−1 ) = (−1)
Theorem 0.10 (The Bolzano-Weierstrass Theorem). Every bounded sequence
of real numbers has a convergent subsequence.
Theorem 0.11 (Heine-Bore Theorem). Let A ⊆ R. Then A is compact if
and only if it is closed and bounded.
Theorem 0.12. Let A ⊆ R. Then the following statements are equivalent:
(i) A is compact

(ii) A is closed and bounded

7
(iii) Every sequence in A has a subsequence that converges to a point in A

Proof. (i) =⇒ (ii). (ii) =⇒ (iii). (iii) =⇒ (i).


(i) =⇒ (iii). (iii) =⇒ (ii). (ii) =⇒ (i).

Corollary 0.5. Summarize the relationship between each property or theo-


rem.

(i) Completeness Property

(ii) Archimedean Property

(iii) Nested Intervals Theorem

(iv) Bolzano-Weierstrass Theorem

Proof. 1. Completeness =⇒ Archimedean

2. Completeness =⇒ Nested Intervals

3. Nested Intervals =⇒ Bolzano-Weierstrass

4. Nested Intervals + Archimedean =⇒ Completeness

5. Bolzano-Weierstrass =⇒ Nested Intervals

6. Monotone Convergence property (think later)

7. Cauchy Criterion property (think later)

Show Nested Intervals + Archimedean =⇒ Completeness.


Proof. Let X be a nonempty set of real numbers that has an upper bound b
in R. and let a ∈ X.
Let I0 := [a0 , b0 ] = [a, b], c0 = a+b
2
.
For n ≥ 1, if cn−1 is an upper bound of X, take an = an−1 and bn = cn−1 ,
otherwise take an = cn−1 and bn = bn−1 . For this construction, an is not
an upper bound of X and bn is an upper bound of X for all n ∈ N. Then
(In := [an , bn ]) is a nested sequence of closed and bounded intervals. By
Nested Intervals theorem, there exists some z ∈ In for all n ∈ N.

8
Claim z is an upper bound of X. Suppose the contrary, there exists some
x ∈ X s.t. z < x. (Idea: find bn ∈ (z, x)). Let ϵ = x−z
2
. Using Archimedean
property, there exists N ∈ N s.t. b−a2N
< ϵ. Since length of IN = b−a
2N
and
x−z x+z
z ∈ IN , IN ⊆ (z − ϵ, z + ϵ). Hence, bN < z + ϵ = z + 2 = 2 < x. This
contradicts bN is an upper bound of X.

Claim for all ϵ > 0, there exists x ∈ X s.t. z − ϵ < x.


Let ϵ > 0. Using Archimedean property, there exists N ∈ N s.t. b−a 2N
< ϵ.
b−a
Since length of IN = 2N and z ∈ IN , IN ⊆ (z−ϵ, z+ϵ). Hence, z−ϵ < aN . As
aN is not an upper bound, there exists x ∈ X s.t. aN < x ,i.e. z−ϵ < aN < x.

Show Bolzano-Weierstrass =⇒ Nested Intervals.


Proof. Let In = [an , bn ], n ∈ N be a nested sequence of closed and bounded
intervals. Aims to show there exists a number ξ ∈ R such that ξ ∈ In for
all n ∈ N. Note that a1 ≤ a2 ≤ · · · ≤ an ≤ an+1 ≤ · · · ≤ bn+1 ≤ bn ≤ · · · ≤
b2 ≤ b1 . Both (an ) and (bn ) are bounded sequence. By Bolzano-Weierstrass
Theorem, (an ) has a convergent subsequence (anj ) such that anj → a ∈ R.
Show a ∈ In for all n ∈ N.
(Idea: IN is closed and contain infinitely many anj ’s. If a ∈ / IN , a has
ϵ-neighborhood not intersect IN .) Suppose the contrary that, there exists
some N ∈ N s.t. a ∈ / IN . Let k > N , i.e. nk > N and therefore a ∈ / Ink .
Then either a > bnk or a < ank . Take ϵ = min(|bnk − a|, |ank − a|). So
(a − ϵ, a + ϵ) ∩ Ink = ∅. As In is nested, anj ∈ Ink for all j ≥ k and therefore
an j ∈
/ (a − ϵ, a + ϵ) for all j ≥ k. This contradicts that a is the limit of the
subsequence (anj ).

You might also like