We will now work with the real numbers R\mathbb{R} instead of an arbitrary continuum CC. Accordingly, let us now use the usual notation (a,b)(a,b) for the region ab={x∈R∣a<x<b}ab = \{x \in \mathbb{R} \mid a < x < b\}. This is not the ordered pair (a,b)(a,b), and we must be vigilant that we do not confuse the two concepts.

Definition 9.1. A sequence is a function a:N⟶Ra: \mathbb{N} \longrightarrow \mathbb{R} from the natural numbers to the real numbers.

By setting an=a(n)a_n = a(n), we can think of a sequence aa as a list a1,a2,a3,…a_1, a_2, a_3, \ldots of real numbers. We use the notation (an)n=1∞(a_n)_{n=1}^{\infty} for such a sequence, or if there is no possibility of confusion, we sometimes abbreviate this and write simply (an)(a_n). More generally, we also use the term sequence to refer to a function defined on {n∈N∣n≥n0}\{n \in \mathbb{N} \mid n \geq n_0\} for any fixed n0∈Nn_0 \in \mathbb{N}. We notate this by writing (an)n=n0∞(a_n)_{n=n_0}^{\infty} for such a sequence.

Definition 9.2. We say that a sequence (an)(a_n) converges to a point p∈Rp \in \mathbb{R} if, for every region RR containing pp, there exists N∈NN \in \mathbb{N} such that if n≥Nn \geq N, then an∈Ra_n \in R. If (an)(a_n) converges to pp, we write this as:

lim⁡n→∞an=p,\lim_{n \to \infty} a_n = p,

and call pp the limit of (an)(a_n). If (an)(a_n) does not converge to any point pp, we say that the sequence diverges.

Exercise 9.3. Show that if a sequence (an)(a_n) converges to pp, then any region containing pp contains all but finitely many terms in the sequence.

Exercise 9.4. Which of the following sequences converge? Which diverge? If one converges, what does it converge to? If one diverges, what can you say about the nature of its divergence?

  1. an=5a_n = 5
  2. an=na_n = n
  3. an=1/na_n = 1/n
  4. an=(−1)na_n = (-1)^n
  5. an=(−1)n⋅na_n = (-1)^n \cdot n
  6. an=(−1)n⋅1na_n = (-1)^n \cdot \frac{1}{n}

Suppose the region (a,b)(a,b)

The next theorem tells us that the limits of sequences are unique:

Theorem 9.5. Suppose that

lim⁡n→∞an=pandlim⁡n→∞an=p′.\lim_{n \to \infty} a_n = p \quad \text{and} \quad \lim_{n \to \infty} a_n = p'.

Then p=p′p = p'.

Sometimes divergent sequences have points that behave like limits, but are not necessarily unique:

Definition 9.6. A point p∈Rp \in \mathbb{R} is a sublimit point of (an)(a_n) if, for every region RR containing pp, there exist infinitely many n∈Nn \in \mathbb{N} with an∈Ra_n \in R.

Exercise 9.7. Construct a sequence with two distinct sublimit points. Construct a sequence with infinitely many sublimit points.

Theorem 9.8. Suppose that lim⁡n→∞an=p\lim_{n \to \infty} a_n = p. Then pp is the only sublimit point of the sequence (an)(a_n).

Definition 9.9. Let (an)(a_n) be a sequence. A subsequence of (an)(a_n) is a sequence b:N→Rb: \mathbb{N} \to \mathbb{R} defined by the composition b=a∘ib = a \circ i, where i:N→Ni: \mathbb{N} \to \mathbb{N} is an injective, order-preserving function. (By order-preserving, we mean that ii has the property that if n<mn < m, then i(n)<i(m)i(n) < i(m).)

If we let nk=i(k)∈Nn_k = i(k) \in \mathbb{N}, we can write bk=ankb_k = a_{n_k}, so that (bk)(b_k) is the sequence b1,b2,b3,…b_1, b_2, b_3, \ldots which is equal to the sequence an1,an2,an3,…a_{n_1}, a_{n_2}, a_{n_3}, \ldots, where n1<n2<n3<⋯n_1 < n_2 < n_3 < \cdots.

Theorem 9.10. If (an)(a_n) converges to pp, then so do all of its subsequences.

Theorem 9.11. Let (an)(a_n) be a sequence and suppose that there exists a subsequence (bk=ank)(b_k = a_{n_k}) that converges to pp. Then pp is a sublimit point of (an)(a_n).

The next two lemmas will help prove the converse to Theorem 9.11. Let p∈Rp \in \mathbb{R} and for each natural number k≥1k \geq 1, define RkR_k to be the region (p−1k,p+1k)(p - \frac{1}{k}, p + \frac{1}{k}).

Lemma 9.12. The RkR_k form a descending collection of regions R1⊃R2⊃R3⊃⋯R_1 \supset R_2 \supset R_3 \supset \cdots whose intersection is the point pp:

⋂k≥1Rk={p}.\bigcap_{k \geq 1} R_k = \{p\}.

Lemma 9.13. Let RR be a region containing pp. Then there exists a natural number NN such that Rk⊂RR_k \subset R for all k≥Nk \geq N.

Theorem 9.14. Let (an)(a_n) be a sequence with a sublimit point pp. Then there exists a subsequence (bk=ank)(b_k = a_{n_k}) that converges to pp.

Here is another nice application of Lemmas 9.12 and 9.13:

Theorem 9.15. Let A⊂RA \subset \mathbb{R}. Then p∈A‾p \in \overline{A} if and only if there exists a sequence (an)(a_n), with each an∈Aa_n \in A, that converges to pp.

Definition 9.16. A sequence (an)(a_n) is bounded above if there exists u∈Ru \in \mathbb{R} such that u≥anu \geq a_n for all n∈Nn \in \mathbb{N}. Similarly, (an)(a_n) is bounded below if there exists l∈Rl \in \mathbb{R} such that l≤anl \leq a_n for all n∈Nn \in \mathbb{N}. We say that (an)(a_n) is bounded if it is both bounded above and bounded below.

Remark 9.17. Note that the sequence (an)(a_n) is bounded precisely when the set {an∣n∈N}\{a_n \mid n \in \mathbb{N}\} is bounded.

Theorem 9.18. Every convergent sequence is bounded.

The converse is not always true, but we do have the following:

Theorem 9.19. Every bounded sequence has a convergent subsequence.
(Hint: sometimes this is called the Bolzano–Weierstrass theorem.)

The rest of the sheet is a natural continuation of the idea behind Lemmas 9.12 and 9.13. Mathematicians often use the letters δ\delta and ϵ\epsilon to denote "small" positive numbers. For example, 1k\frac{1}{k} is small when kk is large. Keeping this in mind, we can rephrase Lemma 9.13 as follows:

Lemma 9.20. Let RR be a region containing the point pp. Then there exists a number δ>0\delta > 0 such that (p−δ,p+δ)⊂R(p - \delta, p + \delta) \subset R.

Theorem 9.21. A function f:R⟶Rf: \mathbb{R} \longrightarrow \mathbb{R} is continuous if and only if the following condition holds for each point p∈Rp \in \mathbb{R}: for all ϵ>0\epsilon > 0, there exists δ>0\delta > 0 such that

(p−δ,p+δ)⊂f−1((f(p)−ϵ,f(p)+ϵ)).(p - \delta, p + \delta) \subset f^{-1}\big( (f(p) - \epsilon, f(p) + \epsilon) \big).

By Theorem 5.4
There is a slightly nicer way of writing this theorem. First, we define the absolute value of a real number:

Definition 9.22. The absolute value of a real number xx is the non-negative number ∣x∣|x| defined by:

∣x∣={x,if x≥0,−x,if x<0.|x| = \begin{cases} x, & \text{if } x \geq 0, \\ -x, & \text{if } x < 0. \end{cases}

Exercise 9.23. Prove that:

(p−δ,p+δ)={x∈R∣∣x−p∣<δ}.(p - \delta, p + \delta) = \{ x \in \mathbb{R} \mid |x - p| < \delta \}.

Our reformulation of Theorem 9.21 is then:

Theorem 9.24. A function f:R⟶Rf: \mathbb{R} \longrightarrow \mathbb{R} is continuous if and only if the following condition holds for each point p∈Rp \in \mathbb{R}: for every ϵ>0\epsilon > 0, there exists δ>0\delta > 0 such that

if ∣x−p∣<δ, then ∣f(x)−f(p)∣<ϵ.\text{if } |x - p| < \delta, \text{ then } |f(x) - f(p)| < \epsilon.

Calculus textbooks often give the condition in this theorem as the definition of continuity. Notice that we have shifted focus to the local behavior of ff at a point pp. This suggests that we can define continuity "one point at a time":

Definition 9.25. A function f:R⟶Rf: \mathbb{R} \longrightarrow \mathbb{R} is continuous at pp if, for every ϵ>0\epsilon > 0, there exists δ>0\delta > 0 such that

if ∣x−p∣<δ, then ∣f(x)−f(p)∣<ϵ.\text{if } |x - p| < \delta, \text{ then } |f(x) - f(p)| < \epsilon.