Jump to content

Unbounded sequences diverge – "Math for Non-Geeks"

From Wikibooks, open books for an open world

In this chapter, we will see that unbounded sequences must diverge. From this, we can deduce that convergent sequences must be bounded.

Unbounded sequences diverge

[edit | edit source]

In the chapter How to prove convergence and divergence we have already seen that the sequence (2n)n∈ℕ diverges. We used the property that the sequence grows beyond any boundary. That is to say, if we have a∈ℝ fixed, then there exists N∈ℕ with 2N≥a+1. Also for all n∈ℕ with n≥N we have 2n≥a+1 and thus

|2n−a|=2n−a≥(a+1)−a=1

Infinitely many members of (2n)n∈ℕ lie outside the ϵ-neighbourhood (a−1;a+1). Therefore (2n)n∈ℕ cannot converge towards a. If that were the case, almost all members of (2n)n∈ℕ would have to be contained inside (a−1;a+1), which is not the case. Because a was chosen arbitrarily, the sequence (2n)n∈ℕ cannot have a limit and is therefore divergent.

We can extend this argument to any sequence that is unbounded, since we only used the property that (2n)n∈ℕ becomes arbitrarily large. Remember the definition of an unbounded sequence:

A sequence (an)n∈ℕ is unbounded, if for all S≥0 there are infinitely many members an with |an|≥S.

We can use this property to prove the following theorem:

Theorem (Unbounded sequences diverge)

Let (an)n∈ℕ be an unbounded sequence. Then the sequence (an)n∈ℕ is divergent.

With this theorem, we can prove that a sequence is divergent. If we can show that a sequence is unbounded, this immediately implies that it is divergent.

How to get to the proof? (Unbounded sequences diverge)

As in the argument from above, we choose a∈ℝ arbitrarily and show that the unbounded sequence (an)n∈ℕ cannot converge towards a. For this, we show that |an−a| is bigger than a certain ϵ>0 for infinitely many n∈ℕ. We must find a lower bound of |an−a|. Here we can use the inverted triangle inequality:

|an−a|≥||an|−|a||

We know that infinitely many |an| are larger than any fixed bound S≥0, since (an)n∈ℕ is unbounded. We set S=|a|+1. Now we can proceed as we did in the example above with (2n)n∈ℕ:

|an−a|≥||an|−|a||↓ |an|≥|a|+1≥|a|≥|an|−|a|↓ |an|≥|a|+1≥(|a|+1)−|a|=1

From |an−a|≥1 it follows that an is not contained in the ϵ-neighbourhood (a−1;a+1) with ϵ=1. But this means that a cannot be the limit of (an)n∈ℕ, which is precisely what we wanted to prove.

Proof (Unbounded sequences diverge)

Let (an)n∈ℕ be an unbounded sequence. Let a∈ℝ be arbitrary. Since (an)n∈ℕ is unbounded, there are infinitely many an with |an|≥|a|+1. For these members, it follows:

|an−a|↓ inverted triangle inequality≥||an|−|a||↓ |an|≥|a|+1≥|a|≥|an|−|a|↓ |an|≥|a|+1≥(|a|+1)−|a|=1

These infinitely many members an are outside the interval (a−1;a+1). Hence (an)n∈ℕ cannot converge to a. Because the choice of a was arbitrary, (an)n∈ℕ must diverge.

Example (Geometric progression)

A generalisation of the introductory example: For |q|>1 the geometric progression (qn)n∈ℕ diverges. From one of the consequences of Bernoulli's inequality, for every S>0 there are infinitely many n∈ℕ with |qn|=|q|n>S. So (qn)n∈ℕ is unbounded and consequently divergent.

Convergent sequences are bounded

[edit | edit source]

Proof by contraposition

[edit | edit source]

The above theorem tells us that unbounded sequences are divergent. With the aid of logical contraposition, we can deduce that convergent sequences must be bounded. The principle of contraposition is:

(A⟹B)⟺(¬B⟹¬A)

The theorem is the following implication

(an)n∈ℕ is unbounded⟹(an)n∈ℕ diverges

Using contraposition, we obtain the equivalent statement:

¬((an)n∈ℕ diverges)⟹¬((an)n∈ℕ is unbounded)

But this means that

(an)n∈ℕ converges⟹(an)n∈ℕ is bounded

If you are not sure that contraposition works, just make the truth table of (A⟹B) and compare it to that of (¬B⟹¬A). As an example: "If it rains (A), the ground is wet (B)." The contraposition is: "If the ground is not wet (¬B), it doesn't rain (¬A)." These two statements are logically equivalent.

So by contraposition the following theorem is true, which we will need to prove further results later on:

Theorem (Convergent sequences are bounded)

Every convergent sequence is bounded. Thus, if a sequence (an)n∈ℕ is convergent, then there exists S≥0 with |an|≤S for all n∈ℕ.

Hint

The converse of the above statement is not true. This means: A bounded sequence need not be convergent. And a divergent sequence need not be unbounded.

As a counterexample, consider ((−1)n)n∈ℕ. This sequence is bounded, but not convergent.

Alternative direct proof

[edit | edit source]

We also want to show an alternative, direct way of proving that a convergent sequence is bounded. This proof is often given in other textbooks. It shows how one can use the ϵ-definition of the limit in a proof.

Proof (Convergent sequences are bounded)

Let (an)n∈ℕ be a convergent sequence. Then there must be an a∈ℝ, so that for all ϵ>0 there exists an index N∈ℕ with |an−a|<ϵ for all n≥N (this is the ϵ-definition of the limit).

We fix ϵ=1 (we could have chosen any other ϵ>0). Now there exists an index N∈ℕ, which we may choose to satisfy N≥2, so that |an−a|<ϵ=1 for all n≥N. All the members following aN are contained inside the ϵ-neighbourhood (a−1;a+1). Thus, all members starting from aN must be smaller than a+1:

(an)n∈ℕ=(a1,a2,a3,…,aN,aN+1,aN+2,aN+3,…⏟These elements are smaller than a+1)

Before aN there are only finitely many members, so there is a maximum M=max⁡{a1,a2,a3,…,aN−1}. So now we have:

(an)n∈ℕ=(a1,a2,a3,…,⏟These elements are less than or equal to MaN,aN+1,aN+2,aN+3,…⏟These elements are less than a+1)

Altogether, the elements of the sequence are bounded above by the maximum of M and a+1, so the sequence is bounded above. Analogously, we can show that the sequence is bounded below. Therefore, the sequence (an)n∈ℕ is bounded.