Properties of limits

We see some basic properties of the limit of a sequence, such as

  • The limit of a sequence is unique.
  • A convergent sequence is bounded.
  • The squeeze theorem.
  • Limiting operation is linear.
  • The limit of the product of sequences is the product of the limits.
  • etc.


Theorem (Uniqueness of limit)

If the sequence {an} converges, its limit is unique. In other words, if we have limnan=α and limnan=β, then α=β.

Proof. Suppose that {an} converges to α as well as to β. Take any positive real number ε

Since limnan=α, there exists NN such that for any nN, |anα|<ε.

Similarly, there exists NN such that for any nN, |anβ|<ε.

Let N=max{N,N}. Then, if nN,

|αβ|=|αan+anβ||αan|+|anβ|<ε+ε=2ε.

Since ε>0 is arbitrary, it must be that α=β. ■

Definition (Bounded sequence)

Given a sequence {an}, we can define a set of real numbers that appear in the sequence: {annN}R. If this set is bounded above, bounded below, or bounded, we say that the sequence {an} is bounded above, bounded below, or bounded, respectively.

Theorem (Convergent sequences are bounded)

Any convergent sequence (i.e., a sequence that converges) is bounded.

Proof. Let {an} be a converging sequence such that limnan=α. Let ε=1. Then there exists NN such that for all nN, |anα|<1. Then, by the triangle inequality,

|an||α||anα|<1

so that

|an|<|α|+1.

Now, let us define

M=max{|α|+1,|a1|,|a2|,,|aN1|}.

Then, whether nN or n<N, we always have |an|M. Hence {an} is bounded. ■

Remark. The converse of this theorem is not necessarily true. e.g., an=(1)n is bounded, but not convergent. □

Theorem (Squeeze Theorem)

Let {an},{bn},{cn} be sequences. Suppose that for all nN, we have anbncn and limnan=limncn=α. Then limnbn=α.

Proof. Let ε>0 be an arbitrary positive real number. Since anα, there exists NN such that for any nN|anα|<ε. In particular, for any nN,

ε<anα.

Similarly, there exists NN such that for any nN,

cnα<ε.

Let N=max{N,N}. Then for any nN, we have

ε<anαcnα<ε.

However, since anαbnαcnα, we have

|bnα|<ε.

Therefore limnbn=α. ■

Let {an} be a sequence. Let's pick a subset {nk} of N which we arrange in increasing order:

n1<n2<n3<nk<nk+1<.

Then, for k=1,2, we obtain a sequence

an1,an2,,ank,ank+1,

This new sequence {ank} is called a subsequence of {an}.

Theorem (Convergence of subsequence)

Let {an} be a sequence that converges to αR. Then, any subsequence {ank} also converges to α.

Proof. Let ε>0 be any positive real number. Since limnan=α, there exists NN such that for all nN, |anα|<ε

Now take KN such that nKN. Then, for any kK, nkN. Thus |ankα|<ε

That is, for any ε>0, there exists KN such that for all kK, |ankα|<ε. This means limkank=α. ■

Theorem

Let {an} be a sequence that converges to αR.

  1. If there exists a real number a such that ana holds for infinitely many nN, then αa.
  2. If there exists a real number b such that anb holds for infinitely many nN, then αb.
Proof. We only prove part 1. Part 2 is similar.

Let us define a subsequence of {an} such that {anana}, and re-index this subsequence so that we may assume ana for all nN in this subsequence. We prove it by contradiction.

Suppose α<a. Let ε=aα(>0). Since limnan=α, there exists NN such that for all nN, |anα|<ε. But then, an<α+ϵ=a, which is a contradiction. ■

Theorem

Let {an} and {bn} be sequences such that limnan=α and limnbn=β.

  1. limnkan=kα where k is a constant.
  2. limn(an+bn)=α+β.
  3. limnanbn=αβ.
  4. limnanbn=αβ given β0.
Proof. Let ε>0. There exists N1N such that for any nN1, |anα|<ε.

Also, there exists N2N such that for any nN2, |bnβ|<ε. Let N=max{N1,N2}. Then for nN, we have both |anα|<ε and |bnβ|<ε.
  1. This is a special case of Part 3 (where bn=k).
  2. By the triangular inequality, |(an+bn)(α+β)|=|(anα)+(bnβ)||anα|+|bnβ|<ε+ε=2ε. Note that 2ε is also an arbitrary positive real number. Therefore, for any nN, we have |(an+bn)(α+β)|<2ε. Thus, limn(an+bn)=α+β.
  3. {an} and {bn} are convergent, and are hence bounded. Therefore, there exists some real number M>0 such that |an|M, |bn|M. By the above theorem, we also have |α|M, |β|M. Then, |anbnαβ|=|anbnanβ+anβαβ|=|an(bnβ)+(anα)β||an||bnβ|+|anα||β|<Mε+εM=2Mε. Note that 2Mε is an arbitrary positive real number. Thus, limnanbn=αβ.
  4. First we show that there exists NN such that for any nN, bn0. Since β0, |β|>0. There exists N such that for any nN, |bnβ|<12|β|. By the triangular inequality, |β||bn||βbn|<12|β| so that |bn|>|β|12|β|=12|β|>0. Therefore, for nN, bn0. For any ε>0 and a sufficiently large n, we have |bnβ|<ε, and |1bn1β|=|βbn||bnβ|<2εβ2. Thus, {1bn} converges to 1β. Therefore, using Part 3, we have limnanbn=limnan1bn=α1β=αβ.

Comments

Popular posts from this blog

Birth process

Branching processes: Mean and variance

Informal introduction to formal logic