Axiom of completeness

Any nonempty set of real numbers that is bounded above has an least upper bound.

Monotone Convergence Theorem

Let (an)(a_n) be an increasing sequence that is bounded above, then it converges.

Proof:

Consider the set of point {an:n∈N}\{ a_n : n \in N \}. By assumption, it is bounded. By Axiom of Completeness, the set {an}\{ a_n \} has an least upper bound, denoted as a=sup⁡{an:n∈N}a = \sup \{ a_n : n \in N \}. Because aa is a least upper bound, ∀ϵ>0\forall \epsilon > 0, there is an element aNa_N, such that aN≥a−ϵa_N \ge a - \epsilon. But (an)(a_n) is an increasing sequence. Therefore, for any n≥Nn \ge N, we have a≥an≥aN≥a−ϵa \ge a_n \ge a_N \ge a - \epsilon, which implies that (an)→a(a_n) \rightarrow a.

Corollary:

Let (bn)(b_n) be a decreasing sequence that is bounded below, then it converges.

Nested Interval Property

Let I1⊃I2⊃I3...I_1 \supset I_2 \supset I_3 ... be a collection of closed intervals. Then ∩nIn≠∅\cap_{n} I_n \neq \emptyset.

Proof 1 (By Axiom of Completeness):

Denote In=[an,bn]={x∈R:an≤x≤bn}I_n = [a_n, b_n]= \{ x \in R : a_n \le x \le b_n \}. By definition, we have

a1≤a2≤a3≤...an≤...≤bn≤...≤b3≤b2≤b1 a_1 \le a_2 \le a_3 \le ... a_n \le ... \le b_n \le ... \le b_3 \le b_2 \le b_1

We are going to use Axiom of Completeness to produce a single real number a∈∩nIna \in \cap_n I_n. The set we consider is

A≐{a1,a2,...} A \doteq \{ a_1, a_2, ... \}

It is bounded above by any bnb_n for n≥1n \ge 1. Therefore we are justified in the setting a=sup⁡Aa = \sup A. Because aa is an upper bound, we have an≤aa_n \le a. The fact bnb_n is an upper bound and that aa is the least upper bound implies a≤bna \le b_n. Putting together, a∈Ina \in I_n for n≥1n \ge 1. Hence, a∈∩nIna \in \cap_n I_n.

Proof 2 (By Monotone Convergence Theorem):

(an)(a_n) is a monotone increasing sequence and bounded above. By Monotone Convergence Theorem, there exists a real number a=lim⁡nana = \lim_n a_n. Moreover, it holds that bm≥anb_m \ge a_n for any integer mm and nn, which implies that bm≥lim⁡nan=ab_m \ge \lim_n a_n = a, as desired.

Remark:

We may not necessarily choose Axiom of Completeness as our starting point. For example, if we assume that Nested Interval Property is true and lim⁡1/2n→0\lim 1/2^n \rightarrow 0, it can be used to prove Axiom of Completeness.

Proof: Let AA be an nonempty set that is bounded above, let b1b_1 be an upper bound and a1a_1 be an element in AA. Necessarily, a1<b1a_1 < b_1 and we can define the closed interval I1=[a1,b1]I_1 = [a_1, b_1].

The idea is to construct a sequence of closed intervals and use Nested Interval Property. Now we bisect I1I_1 into two closed intervals of equal length let c1c_1 be the midpoint. If c1c_1 is an upper bound for AA, we define I2=[a1,c1]I_2 = [a_1, c_1]. Otherwise, let I2=[c1,b1]I_2 = [c_1, b_1]. In general, we construct In=[an,bn]I_n = [a_n, b_n] by taking the half from In−1I_{n - 1} such that, ana_n is no greater than any upper bound of AA and bnb_n is no less than any element of AA.

By Nested Interval Property, there exists a∈∩nIna \in \cap_n I_n. Also we notice that the length of In→0I_n \rightarrow 0, which implies lim⁡an=lim⁡bn=s\lim a_n = \lim b_n = s.

Because for all n∈Nn \in N, and any element a∈Aa \in A, we have a≤bna \le b_n. Taking limit of the right hand side, we obtain a≤sa \le s, which shows ss is an upper bound for AA.

Moreover, for any upper bound ll for AA, we have an≤la_n \le l. Again taking limit of the left hand side, we conclude s≤ls \le l, which shows that ss is the least upper bound.

Bolzano-Weierstrass Theorem

Let (an)(a_n) be a bounded sequence, then it has a convergent sub-sequence.

Proof:
Since (an)(a_n) is bounded there exists M>0M > 0 such that ∣an∣≤M|a_n| \le M for all n∈Nn \in N . Define I0=[−M,M]I_0 = [-M, M]. Then at least one of the intervals [−M,0][-M, 0] and [0,M][0, M] contains infinite number of terms from the sequence (an)(a_n). Select such half interval and label it I1I_1. Let an1a_{n_1} be some term from (an)(a_n) belonging to I1I_1.

Next, we continue to bisect I1I_1 into two closed intervals of equal length. Choose the half that contains infinite number of terms from (an)(a_n) and label it as I2I_2. We can select an element an2∈I2a_{n_2} \in I_2 from (an)(a_n) with n2≥n1n_2 \ge n_1. In general, we construct IkI_k by taking the half from Ik−1I_{k - 1} that contains infinite number of terms of (an)(a_n) and then choose ank∈Ika_{n_k} \in I_k so that nk>nk−1>...>n2>n1n_k > n_{k - 1} > ... > n_2 > n_1. The sets {In:n∈N}\{ I_n : n \in N \} forms a sequence of nested closed intervals

I1⊃I2⊃I3... I_1 \supset I_2 \supset I_3 ...

By Nested Interval Property, there is at least on point a∈∩nIna \in \cap_n I_n. On the other hand, the lengths of the intervals shrink to 00, which implies that ank→aa_{n_k} \rightarrow a, as desired.

Cauchy Criterion

Let (an)(a_n) be a sequence which satisfies

∀ϵ>0,∃N∈N,s.t.,∀n≥m≥N→∣an−am∣≤ϵ \forall \epsilon > 0, \exists N \in \mathbf{N}, s.t., \forall n \ge m \ge N \rightarrow |a_n - a_m| \le \epsilon

Then (an)(a_n) converges.

Proof:

Step 1: (an)(a_n) is bounded. To see this, let N1N_1 be the integer that for n≥Nn \ge N, we have ∣an−aN1∣≤1|a_n - a_{N_1} | \le 1 and so ∣an∣≤∣aN1∣+1|a_n| \le |a_{N_1} | + 1. It can be verified max⁡{∣a1∣,∣a2∣,...,∣aN1−1∣,∣aN1∣+1}\max \{ |a_1|, |a_2|, ..., |a_{N_1 - 1} |, |a_{N_1} | + 1 \} is an upper bound for the sequence (an)( a_n ).

Step 2: Now, by Bolzano-Weierstrass Theorem, (an)(a_n) has an convergent sub-sequence (ank)→a( a_{n_k} ) \rightarrow a. We show that (an)(a_n) converges to the same limit. Let ϵ>0\epsilon > 0, because (an)(a_n) is Cauchy,

∃N>0,s.t.,∀n≥m≥N→∣an−am∣≤ϵ/2 \exists N > 0,s.t., \forall n \ge m \ge N \rightarrow |a_n - a_m| \le \epsilon / 2

We also know (ank)→a(a_{n_k}) \rightarrow a, so choose a term anka_{n_k}, with nk>Nn_k > N and

∣ank−a∣≤ϵ/2 |a_{n_k} - a| \le \epsilon / 2

Putting together,

∀n>N,∣an−a∣≤∣an−ank∣+∣ank−a∣≤ϵ \forall n > N, |a_n - a| \le |a_n - a_{n_k} | + |a_{n_k} - a| \le \epsilon

as desired.

Remark:

Cauchy Criterion can be used to prove Bolzano-Weierstrass Theorem. The key is to construct a sequence of nested closed intervals I1⊃I2⊃I3...I_1 \supset I_2 \supset I_3 ... as before and obtain a sub-sequence (ank)(a_{n_k}). Then, instead of using Nested Interval Property, we observe that (ank)(a_{n_k}) is Cauchy as the length of IkI_k shrinks to zero. We know (ank)(a_{n_k}) converges by Cauchy Criterion.

Reference: Stephen Abbott, Understanding Analysis