Download as pdf or txt
Download as pdf or txt
You are on page 1of 42

Dedekind Cuts

Arhum Khawaja
November 2022

1 Introduction
Real numbers R constitute an ordered field with an additional property:
completeness. It is by this unique property that complete ordered fields only
have one isomorphism type: real numbers.

German mathematician, Richard Dedekind, constructs real numbers from


rational numbers Q by constructing partitions with specific properties. These
partitions are known as Dedekind cuts and were used to construct a Dedekind-
complete-ordered-field, which serves as a construction of the real numbers.

This project seeks to explore firstly, the construction of Dedekind cuts


and the construction of arithmetic operators and order on Dedekind cuts.
Further, the project outlines how the set of all Dedekind cuts is a complete
ordered field and therefore, constructs the real numbers.

1
2 Dedekind Cuts
Now that we’ve constructed the set of rational numbers Q, we can begin
to define a system that will allow us to fulfill the final property needed to
construct R. In order to do this, we establish a system based on operations
defined on cuts known as Dedekind cuts.

We define a cut as a partition of Q into A, Ac ⊂ Q, A, Ac ̸= ∅ with the


following properties:

A is closed downward if and only if:


∀ y ∈ A, x < y ⇒ x ∈ A.

A has no greatest element if and only if:


∄x ∈ A such that x ≥ a ∀a ∈ A. [3]

Additionally, we make the observation that if A ̸= Q, is closed downward


and has no greatest element, then A is bounded above.

We now consider the set of all Dedekind cuts, D. We will now aim to show
that the set of Dedekind cuts forms a complete ordered field by definition
addition and multiplication operations, along with order.

2.1 Defining Operations and Order


We will now define the addition and multiplication for cuts, along with show-
ing that the set of Dedekind cuts is closed under both operations. We will
also define order for the cuts.
Let A, B ∈ D, we define addition and order as follows:

A + B := {a + b | a ∈ A, b ∈ B}.

A < B : ⇐⇒ A ⊂ B.
Whilst it is clear that additive associativity and commutativity follow
from the definition, we further need to define the additive inverse and identity
in order to fulfill the addition axioms of a field.
The additive identity and additive inverses are as follows:

0 := {x ∈ Q | x < 0}.

−A := {a′ ∈ Q | ∃ − b ∈ Ac : b > a′ }. [3]

2
We can show informally that A+(−A) = 0 ∀ A ∈ D since ac < 0−a ∀ a ∈
A, ac ∈ Ac ⇒ a′ < 0 − a ⇒ a′ + a < 0. By properties of cuts, we can
always choose a larger a, a′ ⇒ A + (−A) = 0.

Let A, B ∈ D, we can now define multiplication for A, B ∈ D, A, B ≥ 0


as follows:
AB := {ab |a ∈ A\0 and b ∈ B\0} ∪ 0.
This can be further extended to for all A, B ∈ D as follows:
(
−(|A||B|) if A < 0 or B < 0
AB :=
|A||B| if A < 0 and B < 0
(
A if A ≥ 0
where |A| := . [1]
−A if A < 0
Again, it is clear that associativity and commutativity follow from the defi-
nition, we can define a multiplicative inverse A−1 and identity 1 as follows:

A−1 := {a′ ∈ Q\{0} | ∃b ∈ Ac : a′ < (b)−1 } ∪ 0 ∪ {0} if A > 0. [3]

A−1 := −|A|−1 if A < 0


1 := {x ∈ Q | x < 1}.
We can see informally that AA−1 = 1 ∀ A ⊂ D since a1c < a1 ∀ a ∈ A and
a′ < a1c ∀ ac ∈ Ac ⇒ aa−1 < 1. By properties of cuts, we can always choose
a larger a, a′ , ⇒ AA−1 = 1.

Having defined both addition and multiplication, we can also see intu-
itively that the distributivity axioms hold. However, although it seems as
though D is an ordered field, we can only conclude this if D is closed under
addition and multiplication.

2.2 Closure
This next section aims to prove that D is closed under both operations

Note that A ∈ D always has an upper bound by definition. Therefore,


for any A, B ∈ D, let u, v be their positive upper bounds. Since A, B are
non-empty, A + B and AB are non-empty and have upper bounds u + v and

3
uv respectively.

Assume on the contrary that A + B and AB are not closed downward.


Therefore, ∃ s < a + b and t < ab such that s, t ∈ Q, t > 0 do not belong to
A + B and AB respectively.
t t
∵ s − b < a ⇒ s − b ∈ A and < a ⇒ ∈ A,
b b
t
=⇒ s = (s − b) + b =⇒ s ∈ A + B and t = b =⇒ t ∈ AB.
b
Assume on the contrary that A + B and AB have greatest elements
m = a + b, n = ab ∈ Q. Since A has no greatest element, ∃ a′ ∈ A such that
a′ > a ⇒ u′ = a′ + b > u, v ′ = a′ b > v.

q.e.d

3 Constructing the Real Numbers


So far, by defining Dedekind cuts and defining addition, multiplication and
order for them, we have established that the set of all Dedekind cuts, D is
an ordered field. We now aim to continue further and show that D is also
complete, making it a complete ordered field.

An ordered field F is Dedekind complete if and only if every A ⊂ D has a


least upper bound in F [3]

With this definition of completeness, we can now begin to show that the
Dedekind real number system is complete.

Consider
S X ⊂ D such that X is non-empty and bounded above. Let
S = A ∈ A. Since X is non-empty and bounded above ⇒ S is non-empty
and bounded above. Furthermore, since S is the union of closed downward
sets ⇒ S is closed downward. Likewise, since S is the union of sets with no
greatest element, it has no greatest element.

We can now show that S is the least upper bound for X

By definition, S ≥ A ∀ A ∈ X. Further, assume that ∃P ∈ D such that


P < S and P is an upper bound of X.

∵P <S⇒P ⊂S

4
⇒ ∃Ai ∈ X such that Ai ∩ P ̸= Ai
⇒ P ≱ A for some A ∈ X

Contradiction

Therefore, S is the least upper bound for X.

q.e.d

Since every subset of D has a supremum in D, D is a complete ordered


field. Although we know that the Dedekind complete ordered field is isomor-
phic to R, we can outline this map to explicitly construct R.

Let f : D → R be a function such that f (A) = r, where r is the supremum of A in R.


For every such b, there are 2 cases: r ∈ Ac or r ∈ / Ac . Since A, Ac ∈ Q,
c
r ∈ A ⇒ r ∈ Q. Note that this case constructs all rational numbers in R.
Similarly, r ∈/ Ac ⇒ r ∈ / Q ⇒ r ∈ Q. We can see that this case constructs
the set of all irrational numbers.

References
[1] Stephen Abbott. Understanding Analysis. Springer, 2 edition, 2016.

[2] Robert G. Bartle and Donald R. Sherbert. Introduction to Real Analysis.


John Wiley & Sons, Inc., 4 edition, 1992.

[3] Lothar Sebastian Krapp. Construction of the real numbers, a set theoret-
ical approach. http://www.math.uni-konstanz.de/~krapp/research/
Constructions_of_the_real_numbers.pdf, 2014.

[4] Walter Rudin. Principles of Mathematical Analysis. McGraw Hill, 3


edition, 1976.

5
MATH 309: Introduction to Analysis II
Equivalence of the Completeness of R and the Bolzano-Weierstrass Theorem

by Abdullah Ahmed
2024-10-0035
November 19, 2022

Completeness of R ⇒ Bolzano-Weierstrass Theorem


Proof Let {xn } be a bounded sequence in R such that |xn | < M ∀n ∈ N. This implies sup{xn } = r ∈ R
exists.

Case 1: r 6= xn ∀n ∈ N.

min{|xi − r|}
Consider the construction of the following subsequence: Take xn1 = xk for any k ∈ N. Set δ1 =
2
where 1 ≤ i ≤ n1 . We know that δ1 > 0 as r > xn ∀n ∈ N by assumption. Furthermore as r =
sup{xn } ⇒ ∃ ñ ∈ N suchthat |xñ − r| < δ1 . Clearly, by construction, ñ > n1 . Set xn2 = xñ . Set
min{|xi − r|} δ1
δ2 = ≤ where 1 ≤ i ≤ n2 and repeat to build a subsequence xnk . We can see that
2 2
δk−1
|xnk − r| < δk−1 ∀k ∈ N. But δk ≤ ∀k ∈ N ⇒ δk → 0 ⇒ xnk → r.
2
Case 2: r = xm for some m ∈ N.

Sub-Case 1: ∃ñ ∈ N such that sup{xn |n > ñ} ∈


/ {xn |n > ñ}.

Repeat case 1 for the sequence xn where n > ñ to find a subsequence xnk that converges to sup{xn |n > ñ} ∈ R
as {xn |n > ñ} is also bounded.

Sub-Case 2: ∀ñ ∈ N sup{xn |n > ñ} ∈ {xn |n > ñ}.

Consider the following subsequence: Take xn1 = xm . Set xn2 = sup{xn |n > n1 (= m)} ∈ {xn }. Set
xn3 = sup{xn |n > n2 } ∈ {xn }. Similarly xnk = sup{xn |n > nk−1 }. This is a monotonically non-increasing
subsequence in {xn } which is bounded below ⇒ xnk is convergent by the Monotone Convergence Theorem.

Thus ∃ a subsequence xnk of xn that converges.

1
Bolzano-Weierstrass Theorem ⇒ Completeness of R
Proof Consider any set A ⊆ R that is bounded ⇒ ∃ an a ∈ R such that a is not an upper bound of A
but a + 1 is. It is assumed A does not have a maximal element, otherwise sup(A) = max(A) ∈ R trivially.
Consider the construction of the following sequence:

a
 x=0
1
xn = xn−1 + 2n if xn−1 is not an upper bound of A
xn−1 − 21n if xn−1 is an upper bound of A

Firstly we shall prove that {xn } is a Cauchy sequence. Consider any  > 0 ⇒ ∃n0 ∈ N such that
1 P∞ 1
n
<  ∀n ≥ n0 ⇒ l=n+1 2−l = n <  ∀n > n0 .
2 P2m P∞
−l −l
We can see that |xn − xm | ≤ l=n+1 2 < l=n+1 2 < , where m > n > n0 ⇒ {xn } is a Cauchy
sequence. By assumption of the Bolzano-Weierstrass Theorem, xn is convergent and {xn } → r for some
r ∈ R.

Proposition 1: If {xn } is strictly increasing then sup(A) = (a + 1) ∈ R. If not, then @ m ∈ N such


that {xn } is strictly increasing ∀n > m.

Proof Assume
Pn the sequence is strictly increasing ⇒ xn is not an upper bound of A ∀n ∈ N, and thus
xn = a + l=1 2−l ⇒ xn → (a + 1). We already have that (a + 1) is an upper bound of A by assumption.
Clearly now, for any  > 0 ∃ n̄ ∈ N such that (a + 1) − xn <  ∀n > n̄, where xn are not an upper bounds of
A ⇒ [(a + 1) − ] is not an upper bound of A ∀ > 0 ⇒ sup(A) = (a + 1) ∈ R.
Assume now that it is not strictly increasing and that on the contrary ∃ an m ∈ N>1 such that {xn } is
strictly increasing ∀n > m. Thus we have ∃m̃ ∈ N such that {xn } is strictly increasing ∀n > m̃ and
xm̃−1 > xm̃ . That is, there is an index where the sequence starts to become strictly increasing. This can
1
be argued as m < ∞ by assumption. This gives us that xm̃−1 is an upper bound and xm̃ = xm̃−1 − m̃ .
2
Also as xm̃−1 is not equal to sup(A) by assumption, ⇒ ∃ an upper bound ã such that xm̃ < ã < xm̃−1 ⇒
1 Pn
xm̃−1 − m̃ < ã < xm̃−1 . As the sequence is now strictly increasing ⇒ xn = xm̃ + l=m̃+1 2−l where
2  
1
n > m ⇒ xn → xm̃ + m̃ = xm̃−1 ⇒ ∃ñ ∈ N>m̃ such that ã < xñ < xm̃−1 and xñ is not an upper bound
2
as the sequence is now strictly increasing. This is a contradiction as ã is an upper bound. Thus the sequence
never becomes strictly increasing.

Proposition 2: @ m ∈ N such that {xn } is strictly decreasing ∀n > m.

Proof Assume on the contrary such an m exists ⇒ ∃m̃ ∈ N such that {xn } is strictly decreasing ∀n > m̃
1
and xm̃−1 < xm̃ ⇒ xm̃−1 is not an upper bound and xm̃ = xm̃−1 + m̃ ⇒ ∃ã ∈ A such that xm̃ > ã > xm̃−1 .
2
As the sequence is strictly decreasing now, we can use a similar argument to the one in Proposition 1
⇒ xn → xm̃−1 ⇒ ∃ñ ∈ N>m̃ such that ã > xñ > xm̃−1 and xñ is an upper bound as the sequence is strictly
decreasing now. This is a contradiction as ã ∈ A. Hence the sequence never becomes strictly decreasing.

Proposition 3: If xm is an upper bound ⇒ xn < xm ∀n > m.

2
1 Pm+1+k
Proof Clearly xm+1 = xm − m+1 . Now we also have x(m+1)+k ≤ xm+1 + l=m+2 2−l < xm+1 +
2
P∞ −l 1
l=m+2 2 = xm+1 + m+1 = xm ⇒ xn < xm ∀n > m.
2
Proposition 4: If xm is not an upper bound ⇒ xn > xm ∀n > m.

Proof The proof follows from almost an identical argument to Proposition 3.

Now using Proposition 1 and 2, we can build the following subsequences:

xnk = subsequence of all upper bounds in xn


xnl = subsequence of all non-upper bounds in xn
as we can always find elements of {xn } that are upper bounds or non-upper bounds due to {xn } never
becoming a strictly increasing or decreasing sequence. Both of these are also respectively strictly decreasing
and strictly increasing subsequences by Proposition 3 and 4. As xn → r ⇒ xnk , xnl → r.

Proposition 5: r is an upper bound of A.

Proof Assume it is not ⇒ ∃a0 ∈ A such that a0 > r. Take  = a0 − r. By convergence of xnk ∃k 0 ∈ N such
that xnk − r < a0 − r ∀k > k 0 ⇒ xnk < a0 . This is a contradiction as xnk is an upper bound of A. Hence r is
an upper bound of A.

Now consider any  > 0. We have that ∃l0 ∈ N such that r − xnl <  ∀ l > l0 ⇒ xnl > r −  ⇒ as
xnl is not an upper bound of A, r −  is not either ∀ > 0. Combining this with Proposition 5 that r is an
upper bound of A, we finally have that r = sup(A) ∈ R. Thus R is complete.

q.e.d.

3
Completeness of the Reals via Cauchy Sequences
Course Project for MATH 309: Introduction to Analysis II

Name: Hassan Mehmood


ID: 23100127

November 2022

Abstract
That the real numbers form a complete field is the edifice of analysis. Thus in order to
do analysis, it is important to demonstrate the existence of a field that is complete in an
appropriate sense. There are a number of ways of defining an adequate notion of completeness
and thence to show the existence of a complete field. This essay will show that via Cauchy
sequences of rational numbers, it is possible to characterise completeness in a way that is
equivalent to its being identified with the supremum property, and that a field which is
complete in the former sense exists and can thus be identified with the real numbers.

Definition 1 (Ordered Sets). An order on a set S, denoted by <, is a binary relation that
satisfies the following two properties.
(i) For all x, y ∈ S, exactly one of the statements

x < y, x = y, y<x

is true.
(ii) For all x, y, z ∈ S, if x < y and y < z, then x < z.
Note that x < y and y > x mean the same thing. Also, the notation x ≤ y indicates that x < y
or x = y. A set on which an order is defined is called an ordered set.
Definition 2 (Field). A set F is called a field if it has two operations on it, called multiplication
(denoted ‘·’) and addition (denotes ‘+’), which satisfy the following field axioms.
(A) Axioms for addition
(1) If x ∈ F and y ∈ F , then x + y ∈ F .
(2) x + y = y + x for all x, y ∈ F (commutativity).
(3) (x + y) + z = x + (y + z) for all x, y, z ∈ F (associativity).
(4) There exists an element 0 ∈ F , called the additive identity, such that 0 + x = x for all
x ∈ F.

1
(5) For every x ∈ F , there exists an element −x ∈ F , called the additive inverse of x, such
that x + (−x) = 0.
(M) Axioms for multiplication
(1) If x ∈ F and y ∈ F , then x · y ∈ F .
(2) x · y = y · x for all x, y ∈ F (commutativity).
(3) (x · y) · z = x · (y · z) for all x, y, z ∈ F .
(4) There exists an element 1 ̸= 0 in F , called the multiplicative identity, such that 1·x = x
for all x ∈ F .
(5) For every x ∈ F such that x ̸= 0, there exists an element 1/x ∈ F , called the
multiplicative inverse of x, such that x · (1/x) = 1.
(D) The distributive law
x · (y + z) = x · y + x · z
for all x, y, z ∈ F .
For simplicity, we will just write x − y, xy, x/y etc. in place of x + (−y), x · y, x · (1/y) etc. Using
the foregoing axioms, it is possible to derive all the familiar algebraic manipulations, which we
will now use without justification.
Definition 3 (Ordered Field). An ordered set F is an ordered field if it satisfies the field
axioms and
(i) If x, y, z ∈ F and y < z, then x + y < x + z.
(ii) For all x, y ∈ F , if x > 0 and y > 0, then xy > 0.
Example 4. The set of all rational numbers, Q = {m/n : m ∈ Z, n ∈ Z\{0}}, forms an ordered
field.
Definition 5. The absolute value of Q is the function | · | : Q → Q≥0 such that

x if x ≥ 0,
|x| =
−x if x < 0.

Lemma 6 (Archemdian Property of Q). For every x ∈ Q, there exists n ∈ N such that
n > x.

Proof. If x ≤ 0, then any n works. If x > 0, then x = k/l for some k, l ∈ N. Thus k = lx. Let
n = k + 1.

Definition 7 (Cauchy Sequences in Q). A sequence of rational numbers (x1 , x2 , x3 , ...) :=


(xn ) is called Cauchy if for every ϵ > 0 in Q, there exists Nϵ ∈ N such that

|xm − xn | < ϵ

2
for all m, n ≥ Nϵ .
Example 8.
(i) (1/n), n ∈ N is a Cauchy sequence. Let m > n and Nϵ > 2/ϵ.
(ii) ( nk=1 1/k) is not a Cauchy sequence since it is unbounded (see the next lemma).
P

Lemma 9. Every Cauchy sequence (xn ) is bounded.

Proof. Let ϵ = 1 and M = max{|xn | : n ≤ Nϵ } + 1. Then |xn | ≤ M for all n ≤ Nϵ , and for all
n > Nϵ ,
|xn | = |xn + xNϵ − xNϵ | ≤ |xn − xNϵ | + |xNϵ | < 1 + M − 1 = M.

Lemma 10. If (xn ) and (yn ) are Cauchy sequences, then (xn yn ) and (xn + yn ) are also Cauchy
sequences.

Proof. For the sum, observe that

|(xm + ym ) − (xn + yn )| = |(xm − xn ) + (ym − yn )| ≤ |xm − xn | + |ym − yn |

Thus for any ϵ > 0, pick m, n > Nϵ/2 , where Nϵ/2 exists because (xn ) and (yn ) are Cauchy.
For the product, first note that by Lemma 9, there exists M ∈ Q such that |xn | ≤ M and
|yn | ≤ M for all n ∈ N. Then, for any ϵ > 0, take m, n > Nϵ/2M . We get

|xm ym − xn yn | = |xm ym − xm yn + xm yn − xn yn |
= |xm (ym − yn ) + (xm − xn )yn |
≤ |xm ||ym − yn | + |xm − xn ||yn |
ϵ ϵ
<M + M = ϵ.
2M 2M

We can thus define addition and multiplication on Cauchy sequences in the usual way, i.e.
(xn ) + (yn ) = (xn + yn ) and (xn )(yn ) = (xn yn ). We then immediately see that the sequences
(0, 0, ...) and (1, 1, ...) are the additive and multiplicative inverses, respectively, on the set of
all Cauchy sequences. Further, every Cauchy sequence also has an additive inverse, namely
(−xn ). However, Cauchy sequences still fall short of being a field, since not all of them have
multiplicative inverses, and the product of two non-zero Cauchy sequences may yield 0, e.g.
(1, 0, 0, ...) · (0, 1, 0, 0, ...) = (0, 0, ...). But we can form a field out of the set of Cauchy sequences.
To this end, we first need to define a notion of equivalence between two Cauchy sequences.
Definition 11 (Convergence of a Sequence). A sequence (xn ) converges to a limit x ∈ Q
if for every ϵ > 0 in Q, there exists Nϵ ∈ N such that for all n ≥ Nϵ

|xn − x| < ϵ.

3
We write x = limn→∞ xn or xn → x
Definition 12. Two Cauchy sequences (xn ) and (yn ) are said to be equivalent if and only if
|xn − yn | → 0. We write (xn ) ∼ (yn ).
Lemma 13. The relation ∼ is an equivalence relation.

Proof. |xn − xn | = 0, so ∼ is reflexive.


Since |xn − yn | → 0 is the same thing as |yn − xn | → 0, ∼ is symmetric.
Suppose (xn ) ∼ (yn ) and (yn ) ∼ (zn ). But |xn − zn | ≤ |xn − yn | + |yn − zn |. Thus (xn ) ∼ (zn )
and so, ∼ is transitive.

Thus, every Cauchy sequence (xn ) defines an equivalence class [(xn )] of Cauchy sequences. We
will demonstrate that the set E of all equivalence classes of Cauchy sequences in Q can be
identified with the real numbers. Let us proceed in steps and first establish that E is an ordered
field. Since (zn ) ∼ (xn ) implies that (zn + yn ) ∼ (xn + yn ) and (zn yn ) ∼ (xn yn ), if (wn ) ∈ [(xn )]
and (zn ) ∈ [(yn )], then (wn + zn ) ∈ [(xn + yn )] and (wn zn ) ∈ [(xn yn )]. Therefore, E inherits
multiplicative and additive structure of the set of all Cauchy sequences. This entails that E
contains the multiplicative and additive identities, and the additive inverse of every element.
The only non-trivial fact to establish is thus the existence of a multiplicative inverse in E. This
we now do.
Lemma 14. Every nonzero equivalence class of Cauchy sequences has a multiplicative inverse.

Proof. Let [(xn )] ̸= 0, where 0 stands for the set of all Cauchy sequences equivalent to 0. Thus
for every (zn ) ∈ [(xn )], there exists N ∈ N such that zn ̸= 0 for all n ≥ N . Let yn = 0 for n < N
and yn = 1/xn for n ≥ N . Then clearly, (zn )(yn ) ∼ 1. Hence, [(yn )] is the multiplicative inverse
of [(xn )].

We can define an order on E as follows: [(xn )] < [(yn )] if and only if, for every (wn ) ∈ [(xn )]
and (zn ) ∈ [(yn )], wn < zn for all n ∈ N. This, combined with Lemma 14 and the definitions of
addition and multiplication on E, suffices to prove the next result.
Corollary 15. The set of all equivalence classes of Cauchy sequences in Q forms a field.
Definition 16. The set of real numbers, R, is the set of equivalence classes of Cauchy sequences
in Q.
Lemma 17. R contains a subfield that is isomorphic to Q. Thus R contains Q.

Proof. Define a function f : Q → R such that for all x ∈ Q

f (x) = [(x, x, x, ...)].

4
This map is clearly injective, since for every x, y ∈ Q such that x ̸= y, limn→∞ |x−y| = |x−y| =
̸ 0
and so, (x, x, ...) ̸∼ (y, y, ...). Furthermore, it is easy to see that

f (xy) = [(xy, xy, ...)] = [(x, x, ...)(y, y, ...)] = [(x, x, ...)][(y, y, ...)] = f (x)f (y)
f (x + y) = [(x + y, x + y, ...)] = [(x, x, ...) + (y, y, ...)] = [(x, x, ...)] + [(y, y, ...)] = f (x) + f (y).

Therefore, f restricted to Range(f ), is an isomporphism from Q to R.


That Range(f ) forms a subfield of R follows from the fact that Q is a field.

We can extend the absolute value of Q (Definition 5) to an absolute value on R via

|[(xn )]| := [(|xn |)].

This, combined with the order structure on R, enables us to define Cauchy sequences in R just
as we did in Q. Now, every element of the sequence will itself be an equivalence class of Cauchy
sequences in Q. Similarly, we can define the convergence of sequences in R. With this, the
following lemma and its proof become meaningful.
Lemma 18. A Cauchy sequence of rational numbers (xn ) (appropriately embedded in R via
Lemma 17) converges to [(xn )].

Proof. Given ϵ > 0 in Q, pick an Nϵ such that for all m, n ≥ Nϵ , |xm − xn | < ϵ. But this implies
that given ϵ > 0 in R, |f (xn ) − [(xn )]| < ϵ for all n ≥ Nϵ , where f is the function defined in
Lemma 17.

Lemma 19 (Density of Q in R). For all x ∈ R and ϵ > 0 in Q, there exists r ∈ Q such that
|x − r| < ϵ.

Proof. Let r = xNϵ . Then by the definition of [(xn )], |r − [(xn )]| < ϵ.

We are now ready to prove the central theorem of this essay.


Theorem 20. Every Cauchy sequence of real numbers (xn ) converges to a real number.

Proof. By Lemma 19, there exists a sequence of rationals (rn ) such that |xn − rn | < 1/n for all
n ∈ N. Thus (rn ) is Cauchy, which, in view of Lemma 18, entails that it converges to [(rn )].
But since |xn − [(rn )]| ≤ |xn − rn | + |rn − [(rn )]|, (xn ) converges to [(rn )] too.

Definition 21 (The Completeness Axiom). A field is said to be complete if every Cauchy


sequence in it converges to an element in the field.
By Theorem 20 and Definition 21, R is complete. It just remains to be seen that this definition of
completeness is equivalent to the least-upper-bound property. The reverse direction was proved
in Analysis I. For the forward direction, we can show that Theorem 20 implies the Bolzano-
Weirestrass theorem, which in turn implies the Nested Interval Property, which in turn leads to
the least-upper-bound property.

5
References
Andrew Sutherland, 18.095 Lecture Series in Mathematics (IAP 2015), Lecture 1, Lecture Notes,
MIT.

6
Analysis-II Presentation
Bolzano Intermediate Value Theorem
Sana Waqas - 23100292
November 25, 2022

1 Bolzano’s Intermediate Value Theorem


Let I be the interval [a,b] and let f : I → IR be continuous on I. If there exists a k ∈ IR which
satisfies f (a) < k < f (b) then there exists a point c ∈ (a, b) such that f (c) = k.

A visual representation:

2 Proof
Proof. Let g(x) = f (x) − k
Then
g(a) = f (a) − k < 0
g(b) = f (b) − k > 0
⇒ g(a) < 0 < g(b)
Then by the Location of Roots Theorem* for g(a) < 0 < g(b) ∃c ∈ (a, b) such that g(c) = 0
⇒ f (c) − k = 0
⇒ f (c) = k

*The Location of Roots Theorem states: let I = [a, b] and f : I → IR be continuous. If


f (a) < 0 < f (b) or f (a) > 0 > f (b) then ∃c ∈ (a, b) such that f (c) = 0.

1
3 Corollary
Let I be the closed and bounded interval [a, b] and let f : I → IR be continuous in I. If k ∈ IR is
any number that satisfies inf(f (I)) ≤ k ≤ sup(f (I)) then ∃c ∈ I such that f (c) = k

Proof. It follows from the Max-Min Theorem that there are points c1 and c2 in I such that
inf(f (I)) = f (c1 ) ≤ k ≤ f (c2 ) = sup(f (I))
Then using Bolzano’s Intermediate Value Theorem we can conclude
∃c ∈ I such that f (c) = k

4 Applications
Application 1:
Theorem: Let I be a closed and bounded interval and let f : I → IR be continuous on I. Then the
set f (I) := {f (x) : x ∈ I} is a closed and bounded interval.

Proof. let m = inf(f (I)) and M = sup(f (I))


We know from Max-Min Theorem that m, M ∈ f (I)
f is continuous so we have f (I) ⊆ [m, M ] — (1)
Let k ∈ [m, M ]
⇒ inf(f (I)) = m ≤ k ≤ M = sup(f (I))
Then it follows from the corollary that ∈ I such that f (c) = k
Hence k ∈ f (I)
⇒ [m, M ] ⊆ f (I) — (2)
Then by (1) and (2) ⇒ f (I) = [m, M ]

Application 2:
Preservation of Intervals Theorem: Let I be an interval and let f : I → IR be continuous on I
then the set f (I) is an interval.
Proof. Let a, b ∈ f (I) with a < b
Then ∃ points x, y ∈ I such that a = f (x) and b = f (y)
Further it follows from Bolzano’s Intermediate Value Theorem that if k ∈ (a, b) then ∃c ∈ I with
k = f (c) ∈ f (I)
⇒ [a, b] ⊆ f (I)
This shows that f (I) possesses the following property: If S ⊆ IR containing at least 2 points and if
x, y ∈ S, x < y then [x, y] ⊆ S, then S is an interval.
Therefore, f (I) is an interval.

5 Example
Let f be continuous on the interval [0, 1] to IR and such that f (0) = f (1). Prove that there exists a
point c ∈ [0, 21 ] such that f (c) = f (c + 12 )
Solution:
Given: f s continuous on [0, 1] and f (0) = f (1)
Goal: ∃c ∈ [0, 12 ] such that f (c) = f (c + 21 )

Let g(x) = f (x) − f (x + 12 )

2
And let f (0) = f (1) = a
g(0) = f (0) − f ( 12 ) ⇒ g(0) = a − f ( 12 )
g( 21 ) = f ( 12 ) − f (1) ⇒ g( 12 ) = f ( 12 ) − a
Then
g(0) ∗ g( 12 ) = (a − f ( 12 ))(f ( 12 ) − a)
= −(a − f ( 12 ))2 ≤ 0
⇒ One of the terms g(0), g( 12 ) must be positive and one must be negative

f (x) is continuous on [0, 1] so g(x) is continuous on [0, 12 ]


Then recall if f (a) < 0 < f (b) or f (a) > 0 > f (b) then ∃c ∈ (a, b) such that f (c) = 0

So ∃ a point c ∈ [0, 21 ] such that g(c) = 0


g(c) = 0
f (c) − f (c + 21 ) = 0
f (c) = f (c + 12 )

6 References
[1] Sherbert, Donald R., and Robert G. Bartle. ”Introduction to Real Analysis”, Fourth Edition.
2020.
[2] Trench, William F. ”Introduction to Real Analysis.” Digital commons at Trinity, 2013.

3
Limit Point, Lim inf, Lim sup Abdul Basit
2024-10-0261
BS MATH
Limit Point (def. by sequence)
Let A be a subset of R. A point c belongs to A is called a limit point of A if there is a sequence
(𝑎𝑛 ) in A such that 𝑎𝑛 ≠ 𝑐 𝑓𝑜𝑟 𝑒𝑣𝑒𝑟𝑦 𝑛 ∈ 𝑁 and such that (𝑎𝑛 ) → c
For example
1 1 1 1
𝐴 = {2 , 3 , 4 , 5 , … }
1
We can find a sequence 𝑥𝑛 = 𝑛 which converges to 0 as n goes to infinity. Clearly, 𝑥𝑛 ≠
0 𝑓𝑜𝑟 𝑎𝑛𝑦 𝑛 ∈ 𝑁. Hence 0 is a limit point of this set. We cannot find any other limit point
for this set and hence it only has a single limit point.

Let’s take another example,


B = (1,2]
We can get sequences for every point in this set as follows:

1
𝑎𝑛 = {𝑥 + 𝑛} , 1 ≤ 𝑥 ≤ 3/2 , it converges to x
1
𝑎𝑛 = {𝑥 − 𝑛} , 3/2 ≤ 𝑥 ≤ 2 , it converges to x

Hence, Limit points of B = [1,2]

It is interesting to note that 1 is also a limit point of set B which does not even belong to B.
Also, our set of limit points is even bigger than the original set.

Limit Point (def. by neighbourhood)


Let S be a subset of R, a real number l is called a limit point of S if every neighbourhood of l
contain at least one member of S other than l itself.

For example

C = {1,2,3,4,5,6,…n) be a finite set. We know that we can take 𝜀 =1/2 such that the
neighbourhood of every point l of this set contains no member of A other than l itself. Hence
every finite set contains no limit point.

For set A discussed above (previous heading) we can apply this concept to see that we can
1
find an 𝜀 = 𝑥+2 for every 𝑥𝜖𝐴 hence no member of A can be a limit point. We can also see
that there are infinite entries of A near 0. Hence by neighbourhood definition,
neighbourhood of 0 will always contain at least one member of A other than 0 itself. Hence 0
is the only limit point of A.
Lim Sup and Lim Inf
The limit superior of 𝑥𝑛 is the infimum of the set V of v belongs to R such that v < 𝑥𝑛 for at most a
finite number of n belongs to N. Type equation here.

The limit inferior of (𝑥𝑛 ) is the supremum of the set of w belongs to R such that 𝑥𝑚 < w for at most a
finite number of m belongs to N.

Let’s consider this sequence


𝑧𝑛 = (−1)𝑛
We know that this sequence does not converge to any point. Let S be a set which contains
supremum of 𝑧𝑛 and its sub-sequences such that each new sub sequence is constructed by deleting
one entry from previous one. We know that supremum of all these sequence is 1 infimum of S is 1.
i.e limsup = 1 and it converges. Similar is the case with infimum and lim inf = -1 . Amazingly as we can
see that the sequence itself does not converge but its lim sup and lim inf converges.

Theorem: A bounded sequence (𝑥𝑛 ) is convergent if and only if lim sup(𝑥𝑛 ) = lim inf(𝑥𝑛 ) .
Now consider a sequence (𝑥𝑛 ) is convergent to a point c ↔ ∃𝑚 ≥ 𝑘 𝑤ℎ𝑒𝑟𝑒 𝑘 ∈
𝑁 𝑠𝑢𝑐ℎ 𝑡ℎ𝑎𝑡 mod(𝑥𝑚 − x) < ε , ∀ε > 0 ↔ The sup and inf of that sub sequence will only differ by
𝜀 ↔hence by definition of limsup and lim inf, when n approaches infinity 𝑙𝑖𝑚𝑠𝑢𝑝 − 𝑙𝑖𝑚𝑖𝑛𝑓 =
mod(𝑥𝑚 − x) < 𝜀 ∀ε > 0
↔ lim sup(𝑥𝑛 ) = lim inf(𝑥𝑛 ) .
Weierstrass Approximation Theorem

Description:
The Weierstrass Approximation Theorem is used to approximate continuous functions that are
not well behaved (differentiable at every point) using polynomials (on a compact interval).
Statement:
If [a, b] is an interval, f : [a, b] → R is a continuous function, and ε > 0, then there exists a
polynomial P on [a, b] such that d∞(P, f) ≤ ε (i.e., |P(x) − f(x)| ≤ ε for all x ∈ [a, b]).

In other words, we can always find a polynomial that can approximate a continuous function on
a given interval.
It can also be stated in the following way:

P([a, b] → R) = C([a, b] → R)

This means that the closure of space of the polynomials is the space of continuous functions.

History:
This theorem was established by Weierstrass in 1885, and like most of his work, it wasn’t
discovered until after his death (most of his work wasn’t published). This theorem has become a
staple in approximation theory and has been the basis for multiple further theorems/results.
There is one in particular, the Stone-Weierstrass theorem that is of particular interest here. The
Stone-Weierstrass theorem is a more general version of the Weierstrass approximation
theorem, as it generalizes the results to 2 dimensions (we will not see the formal definition here
as it is beyond the scope of this course).
There have been multiple proofs for the Weierstrass approximation theorem, two of which we
will see here:

Proof:

Proof 1 uses Bernstein polynomials (Next Page contains the proof)

Proof 2 uses multiple other results and corrolaries (Next page)

Applications:

- Space C[a, b] is separable


- Polynomial functions are dense
- Each polynomial function can be uniformly approximated by one
with rational coefficients

*Note: The Proofs have taken more space as the writing size is large (the document has
exceeded 3 pages due to this
ate

Roo 1
Becastein ne N Ronren
oMods Pynenol O a unchoa
is dekes
fE CP, B)
K

&e Clesal,R)-
Xhea oee iSa Seaene x o
ha

Consides fest fel),R)

S-t

ee,4

t:).3..>...-. A ...LA9.i .
***** .
Date

e 2 Lo,1

- 2 2 . S, en

aPpoxiacte on \91-

i e t

binem eae as n e2
Date
B-)-Ae\= \8.(3,f-AE)|

2 8.(xaT)2
2 ne
OB f)-
Con be proven Aroh e
denea_ak t))

So
2 2 LG

Pcicdor

A snpe coddadon
2-22L

********************************************************
*************************** ******* *
Date

e(2A)-fa)\<h 2S

PoNeS

\e
Lasside
aLlab,R)
3p-a)a)
is a oncOaerpSm

peci
LAnnAusu
jenetase

* ***************** ******
Date
Paoof
Arof of
s roof hs onstMe A Sarne t
anad eolokched eidts

le a,b R be caetiauous
R
le a lo1 R denote teCtios

e)=ar Coa) eo,A).

CA earlier resut mar

e i R Cahiuaa on e,)
o a oma

P:p1)R Sun Pe) -ka\SE

, 1}
pnanial Q-s,11R s

Q
** --.
Date
Partile e hae

e-)-gy-Cy-b-)|SE
Sek P Q(rab))
hseue P 3aka a
olunemiaA

desi e re)ur
Saif Ur Rehman
23100270
Riemann’s Rearrangement Theorem
The outcomes of their work with infinite series frequently baffled mathematicians in the late
seventeenth and early eighteenth century. Peter Lejeune-Dirichlet made this startling discovery
in 1827 while researching the circumstances that made Fourier series convergence inevitable. He
was the first to realise that the sum of some series, now known as conditionally convergent
series, might be altered by rearranging the terms. Why is this outcome conceivable? Dirichlet
was never in a position to answer this question. Riemann quickly discovered a wonderful
explanation for this peculiar behaviour, which is now known as Riemann's rearrangement
theorem.
As an example of what can go wrong, suppose we let S represent the sum of the
alternating harmonic series, that is:

𝑆 = ∑((−1)𝑛−1 )/𝑛
𝑛=1

Fig. 1
See figure 1. What’s wrong here? (This series, by the way, is not divergent. Its sum is
ln 2, which is easy to determine. Find the Taylor series of and let x equal 1.)
It seems that although we merely rearranged the terms of an infinite series (equations 3
and 4), its sum has changed from 2S to S!
Process of rearrangement:
To get at Riemann’s theorem we will use the definition of the sum of an infinite series
and seven theorems that are part of a standard first course on infinite series. We will also
need to distinguish between two types of convergent series.
The Sum of an Infinite Series:
“If the limit of the nth partial sum of the infinite series exists and equals S, then we say
converges and its sum is S. If, as n approaches infinity, the limit of the nth partial sum does not
exist, then we say the series diverges and has no sum”.
Theorem 1: The sum of two convergent series is a convergent series. If ∑ 𝑎𝑛 =S and ∑ 𝑏𝑛 =T
then
∑(𝑎𝑛 + 𝑏𝑛 )= S + T
Theorem 2: The sum of a convergent series and a divergent series is a divergent series.

Theorem 3: ∑∞
1 𝑎𝑛 and ∑𝑖 𝑏𝑛 both converge or both diverge. (In other words, the first

finite number of terms do not determine the convergence of a series.)


Theorem 4: If the series ∑ 𝑎𝑛 converges, then lim (𝑎)𝑛 = 0
𝑛→∞

Theorem 5: If ∑ |𝑎𝑛 | converges, then ∑ 𝑎𝑛 converges.


Theorem 6: The comparison test. If the series ∑ 𝑎𝑛 and ∑ 𝑏𝑛 have only positive terms
with 𝑎𝑛 < 𝑏𝑛 for all n ≥ 1 and,
(1) if ∑ 𝑏𝑛 converges, then ∑ 𝑎𝑛 converges;
(2) if ∑ 𝑎𝑛 diverges, then ∑ 𝑏𝑛 diverges.
Theorem 7: Leibniz’ alternating series test. The alternating series ∑(−1)𝑛−1 𝑎𝑛
converges if the sequence {𝑎𝑛 } is monotone decreasing to 0.
We see that although the alternating harmonic series converges, the series obtained by
replacing each term by its absolute value diverges. This result shows that the
convergence of ∑ 𝑎𝑛 does not imply the convergence of ∑ |𝑎𝑛 |.
Two Types of Infinite Series:
A series ∑ 𝑎𝑛 is called absolutely convergent if ∑ 𝑎𝑛 and ∑ |𝑎𝑛 | converges. A series is
called conditionally convergent if ∑ 𝑎𝑛 converges but ∑ |𝑎𝑛 | diverges.
For Example:
∑∞
𝑛=1((−1)
𝑛−1
)/𝑛 is conditionally convergent because ∑∞
𝑛=1 |(−1)
𝑛−1
/𝑛| diverges!
∑∞
𝑛=1((−1)
𝑛−1
)/𝑛2 is absolutely convergent because ∑∞
𝑛=1 |(−1)
𝑛−1
)/𝑛2 | converges!
Riemann’s Rearrangement Theorem—Part 1
In a conditionally convergent series, the sum of the positive terms and the sum of the negative
terms of the series is a divergent series.
Proof:
Take a conditionally convergent series ∑ 𝑎𝑛 . Divide it into finite series of +ve and -ve
terms, such that:
∑ 𝑎𝑛 = ∑ 𝑎𝑛+ + ∑ 𝑎𝑛− ;
𝑎𝑛+ = {𝑎𝑛 𝑎𝑛 > 0 ; 0 𝑎𝑛 < 0 and 𝑎𝑛− = {0 𝑎𝑛 > 0; 𝑎𝑛 𝑎𝑛 < 0
For example: Consider a series:
∑ 𝑎𝑛 = 1-1/2-1/3+1/4-1/5-1/6+1/7-…….
∑ 𝑎𝑛+ =1+0+0+1/4+0+0+1/7+……
∑ 𝑎𝑛− =0-1/2-1/3+0-1/5-1/6+0-…….
To check for the conditional convergence of ∑ 𝑎𝑛 , we have 4 possible cases:
(1) ∑ 𝑎𝑛+ converges and ∑ 𝑎𝑛− converges -> not possible as it makes the series ∑ 𝑎𝑛
absolutely convergent.
(2) ∑ 𝑎𝑛+ converges and ∑ 𝑎𝑛− divergent -> not possible as it makes the series ∑ 𝑎𝑛 divergent.
(3) ∑ 𝑎𝑛+ diverges and ∑ 𝑎𝑛− converges -> not possible as it makes the series ∑ 𝑎𝑛 divergent.
(4) ∑ 𝑎𝑛+ converges and ∑ 𝑎𝑛− converges -> Only possible case of two divergent series that
make the series ∑ 𝑎𝑛 conditionally convergent.
 𝐷𝑖𝑣𝑒𝑟𝑔𝑒𝑛𝑐𝑒 𝑜𝑓 𝑡𝑤𝑜 𝑠𝑒𝑟𝑖𝑒𝑠 𝑖𝑠 𝑡ℎ𝑒 𝑘𝑒𝑦 𝑖𝑑𝑒𝑎 𝑖𝑛 𝑝𝑟𝑜𝑣𝑖𝑛𝑔 𝑝𝑎𝑟𝑡 −
𝐼𝐼 𝑜𝑓 𝑅𝑒𝑖𝑚𝑎𝑛𝑛′ 𝑠 𝑟𝑒𝑎𝑟𝑟𝑎𝑛𝑔𝑒𝑚𝑒𝑛𝑡 𝑡ℎ𝑒𝑜𝑟𝑒𝑚.
 𝑇ℎ𝑒 𝑡𝑒𝑟𝑚𝑠 𝑐𝑎𝑛 𝑏𝑒 𝑟𝑒𝑎𝑟𝑟𝑎𝑛𝑔𝑒𝑑 𝑡𝑜 𝑎𝑑𝑑 𝑢𝑝 𝑎𝑛𝑦 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑒𝑟𝑚𝑠 𝑡𝑜 𝑔𝑒𝑡 𝑎𝑛𝑦 𝑔𝑖𝑣𝑒𝑛 𝑠𝑢𝑚
Riemann’s Rearrangement Theorem—Part 2
Let ∑ 𝑎𝑛 be a conditionally convergent series, and let S be a given real number. Then a
rearrangement of the terms of ∑ 𝑎𝑛 exists that converges to S.
Proof:
We can rearrange terms of ∑ 𝑎𝑛 to form series whose sum is any given S
 Add just enough number of +ve terms, say 𝑛1 terms, of ∑ 𝑎𝑛 so that their sum is larger
than S.
𝑆𝑛1 > S
(As we know the sum of +ve series diverges to +∞).
 Add just enough -ve terms , say 𝑛2 terms, of ∑ 𝑎𝑛 to the sum so that it is smaller than S.
𝑆𝑛1 +𝑛2 < S

(As we know the sum of +ve series diverges to -∞).


 Repeat this process, add +ve terms to make sum larger than S and -ve terms to make the
sum smaller than S.
𝑆𝑛1 +𝑛2 +𝑛3 > S
𝑆𝑛1 +𝑛2 +𝑛3+𝑛4 < S

 All these partial sums differ from S by at most 1 positive or negative term. These partial
sums must be closing in on S, since the original series converges (𝑎𝑛 → 0 𝑎𝑠 𝑛 → ∞).
 So, the partial sums converge to S which proves that the series of rearranged terms
converges to S!

• The rearrangements are not unique! It depends on the S we want to get by rearranging
the terms.
• We can rearrange the terms of any conditionally convergent series so that it will
diverge!
One such rearrangement is to pick positive terms to add to a million, then add
on one negative term, then add on positive terms to reach a trillion, then add on
another negative term, then add positive terms till we are beyond a googolplex, then
add on a negative term.
• That’s Riemann’s rearrangement theorem. It’s really a grand counterexample to
the seemingly plausible idea that we can rearrange the terms of any infinite
series and be sure that we will not alter its sum.

Bibliography
Apostol, Tom M. Calculus. Vol. 1. New York: John Wiley & Sons, 1967.
Grattan-Guinness, Ivor. The Development of the Foundations of Mathematical Analysis
from Euler to Riemann. Cambridge: MIT Press, 1970.
Kline, Morris. Mathematical Thought from Ancient to Modern Times. New York: Oxford
University Press, 1972.
Riemann, Bernhard. “Uber die Darstellbarkeit einer Function durch eine
trigonometrische Reihe.” Gesammelte Mathematische Werke (Leipzig 1876): 213–53
Existence of the Exponential and Logarithmic Functions

M. Sameer Abbas„

November 25, 2022

„ Roll No. 24100131 - LUMS

1
Contents
1 Introduction 2

2 The Logarithmic Function 2

3 The Exponential Function 3

List of Theorems and Definitions


2.1 Theorem (Riemann Integrability) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1 Definition (The Logarithmic Function) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
3.1 Theorem (Surjectivity) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3.2 Theorem (Injectivity) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3.1 Definition (The Exponential Function) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1
1 Introduction
The following contains a proof of the existence of the logarithmic and exponential functions, denoted by
log x and ex . We start by exploring the logarithmic function by the usual definition as an integral, and
we show its existence over its domain. Then, we prove the logarithm’s bijectivity, thereby guaranteeing
an inverse function. Lastly, we claim that the exponential function is the logarithm’s inverse, thereby
guaranteeing its existence.

2 The Logarithmic Function

Theorem 2.1 (Riemann Integrability). Let f : [a, b] → R be continuous. Then it is integrable on


[a, b].

Proof. Since [a, b] is a closed and bounded subset of R, by the Heine-Borel theorem, it is compact.
It follows from the Heine-Cantor theorem that f is uniformly continuous on [a, b]. For any partition
a = x0 , . . . , xn = b, define

∆xi = [xi−1 , xi ]; Mk = sup f (x); mk = inf f (x)


∆xk ∆xk

Mk − m k = sup |f (x) − f (y)|


x,y∈∆xk

ϵ
Given ϵ > 0, we can find δ > 0 such that |x − y| < δ =⇒ |f (x) − f (y)| < . So if Pϵ is a
2(b − a)
partition with ∥Pϵ ∥ < δ, then for any finer partition P , we must have:
n n
X ϵ X ϵ
U (f ; P ) − L(f ; P ) = (Mi − mi )∆xi ≤ ∆xi = < ϵ
i=1
2(b − a) i=1 2

So f is integrable on [a, b] ■

Consider the function f (t) = 1t . Clearly f is continuous only any [a, b] with 0 < a ≤ b. It follows from
theorem 2.1 that its integral over any such [a, b] exists. We define the logarithmic function as follows:

Definition 2.1 (The Logarithmic Function).


Z x
1
log x := dt; (x > 0)
1 t

Therefore, the existence of log x over its entire domain is trivial. To show, however, that log x is a new
function is much less elementary. That is to say that it is not among the class of rational and algebraical
functions already well known to us. Perhaps a good starting point for such a discussion begins with
exploring the properties that log x possesses. For example, making the substitution t = yu, we see that
Z xy Z x Z x Z y−1
1 1 1 1
log xy = dt = du = du − du
1 t y −1 u 1 u 1 u
Z x Z y
1
u7→ u 1 1
= du + du = log x + log y
1 u 1 u
So log x satisfies the functional equation

f (xy) = f (x) + f (y) (1)

In fact, it can be shown that there is no solution of the equation 1 which possesses a differential coefficient
and is fundamentally distinct from log x. For when we differential the functional equation, first with
respect to x and then with respect to y, we obtain the two equations:

yf ′ (xy) = f ′ (x), xf ′ (xy) = f ′ (y)

2
and so, eliminating f ′ (xy), xf ′ (x) = yf ′ (y). Since this is true for each choice of x and y, we must have
xf ′ (x) = C. This gives us: Z
C
f (x) = dx = C log x + C ′
x
and it is easy to see that C ′ = 0. Thus there is no solution fundamentally distinct from log x, except the
trivial solution f = 0 obtained by taking C = 0. So it seems as though log x truly defines a new class
of functions (those that satisfy the functional equation given above). Though this does not prove the
transcendence of the logarithm (which is not the aim of this paper), it does give a flavor of log x being
a new kind of function.

3 The Exponential Function

Theorem 3.1 (Surjectivity). Let f : (a, b) → R be continuous. If f (x) → ±∞ as x → a and


f (x) → ∓∞ as x → b, then f is surjective.

Proof. Suppose WLOG that x → a =⇒ f → −∞ and x → b =⇒ f → ∞. Then for any


c ∈ R, we will show that ∃x0 ∈ (a, b) : f (x0 ) = c. First, notice that since x → a =⇒ f → −∞,
∃a0 ∈ (a, b) : f (a0 ) < c. Also notice that since x → b =⇒ f → ∞, ∃b0 ∈ (a, b) : f (b0 ) > c. Then
notice that f : [a0 , b0 ] 7→ [f (a0 ), f (b0 )] is continuous and c is number between f (a0 ) and f (b0 ), so
by the intermediate value theorem, ∃x0 ∈ [a0 , b0 ] ⊆ (a, b) : f (x0 ) = c. So f is surjective. ■

Theorem 3.2 (Injectivity). Let f : (a, b) → R be continuous. If f ′ > 0 for every x ∈ (a, b), or if
f ′ < 0 for every x ∈ (a, b), then f is injective.

Proof. WLOG, Suppose f ′ > 0 for each x ∈ (a, b). Then for any x, y ∈ (a, b) with x < y, we have
by the mean value theorem:

f (y) − f (x)
= f ′ (c) > 0 (c ∈ (x, y))
y−x

Which implies f (y) > f (x). So x ̸= y =⇒ f (x) ̸= f (y) for each x, y ∈ (a, b). Therefore f is
injective. ■

Our goal is to prove the bijectivity of log : (0, ∞) → R, thereby guaranteeing the existence of an inverse
function. Then we define the exponential function as the inverse of the logarithmic function so that its
existence follows by definition. First, notice that log is differentiable over its domain and therefore is
continuous. Next, we explore the behavior of log near the boundary of its domain. Notice that if x > 2n ,
we have: Z x Z 2n Z 2 Z 4 Z 2n
dt dt dt dt dt
log x = > = + + ... +
1 t 1 t 1 t 2 t 2n−1 t
But by putting x = 2r u, we obtain
Z 2r+1 Z 2
dt du
=
2r t 1 u
Z 2
du
and so log x > n . Therefore log x → ∞ as n → ∞ =⇒ x → ∞ Moreover, making the substitution
1 u
1
t 7→ u gives us
Z x Z x−1  
dt du 1
log x = =− = − log
1 t 1 u x
1
So as x → 0, → ∞, therefore log x → −∞. Observing the continuity of log and using theorem 3.1, we
x
see that log is surjective. Moreover, its derivative x1 > 0 is positive over its entire domain. Theorem 3.2
shows us log is injective. Therefore we have that it is bijective and has an inverse. We define this inverse
function as the exponential function ex by the following definition:

3
Definition 3.1 (The Exponential Function).

ey := x (where y = log x)

Then it follows by definition that the exponential function exists over its entire domain.

4
RESENTATION

WRITE UP

Ahmer Nadeem Khan

24100142

I
#

*Set
Cantor Set

Oection
Co C0.13
=

c. =

(0.5)U/E,y
( (x
)v(j) -(z)
=

I
·
v(8,1)

zu(z c)
a =
+ -11

ken,
ci=tcn i.e. what is left

iterative process.
This
b

Co A ⑭

Cz
O
i
I -
↑ - 11)
Obviously. He boundary points on

each division are in the set e.g.

13 and 213 (They are never removed).

&
Length 0
=

Co has length

is
removed for C.
Length

C for Co,
tx2forc.... T
Note length removed:

=
1
= ya...
+ +

-
-SE

-SECE,this
-- 1

Geometric series
with v 2
=

5 -
Then, Cantor set length is 0.(1 -1
0)
=

ten Cantor set cannot contain any


interval of non-zero
length (Def. of

Lebesque Measure).

&
Each number in the set has a

binary (ternary) form.

Let 0 denote a left and Idevote

a right, then

J 01.
=
... .

0 00000....
=

He cantor set is all possible infinite

numbers (this set is also


binary
uncountable -

consider
any enumeration (counting) of

these numbers:
I
> 0 1 0 1 I) . .
.

I
000010 ...

& 1 010100. 11100...


1
1000100...


50 1, I, I 1 . . .

11001100
100 ...

- :

Forres t i
steerbscio m e
it
-
lipping e.g.
pen, this number won't be in the set

i.e. C is uncountable.
As,
Is in

C
No

Y 010101010 ...
=
te
ternary
expansion if x<(*)

* Te set of endpoints is countable.

* It is not dense. (nowhere dense).

* If << (0, 2], ken Ja,be(s.t.


x
=
a+ b
* C"Ci.e. multidimensional cantorset)

& Cantor set is the complement of

a union of
open sets, so it is closed.

Since (C [0,1], ten C is bounded

Cinfact, totally bounded). C is closed.

Heine Borel
Hence, I is compact by
-

Reorem. B

& Co =0

Assume sce CO. Then 5030 s.t. B(or)

(x C. But then 3
r,x r) open
- - +
an

&measure)
C (x +r)
interval in of
length -
(x -

r)

In 30 which is a contradiction.
-
D

* Every Metric space


is a
subspace
of 0.
O
once Dimension: -

· c is kind of like a fractal.

It is of itself but occled


~ a
copy

(self-similar)

3 -
- 3

3C,
:I
6 -t
-

23

-
& Cantor Sets

ag()
D 0.61)
=

* Fractals have fractional

dimension.

c-

You might also like