Sequences

A sequence is a function whose domain is N
f:NR,f(n)

Convergence of a Sequence

A sequence (an) converges to aR if for every ϵ>0 there exists a NN such that if n>N then |ana|<ϵ

ϵ-Neighbourhood of a

Vϵ={xR:|xa|<ϵ}

So an equivalent definition is:
limnan=a iff ε>0,Bε={m:anVε(a)} is finite.

Thm 2.2.7 Uniqueness of Limits

If (an)a and (an)b then a=b

Proof

There exists and Na and Nb where
for all ϵ2>0 |ana|<ϵ2 and |anb|<ϵ2

|anaan+b||ana|+|anb|<ϵ
|ba|<ϵ


Algebraic and Order Limit Theorems

Algebraic Theorems

limnan=a and limnbn=b then

(i)limn(can)=ca, for all cR(ii)limn(an+bn)=a+b(iii)limn(anbn)=ab(iv)limn(anbn)=ab,if b0

Proofs

Proof i
There is an N so that n>N

|ana|<ϵ|c||canca|=|c||ana|<|c|ϵ|c|=ϵ

Proof ii

We know for some Na that |ana|<ε2 and some Nb that |bnb|ϵ2. Take max{Na,Nb} and the rest is trivial

Proof iii

|anbnab|=|anbnabn+abnab||anbnabn|+|abnab|=|bn||ana|+|a||bnb|.

We know there is a Nb so that

nNb|bnb|<1|a|ϵ2

Also bn converges so it is bounded |bn|M

nNa|ana|<1Mϵ2

nmax{Na,Nb}

|anbnab||bn||ana|+|a||bnb|M|ana|+|bnb|M(ϵ2M)+|a|(ϵ|a|2)=ϵ

Proof iv
(bn)b(iii)(1bn)1b if b0

Since (bn)b there is a N1 so that |bnb|<|b|/2 for any n>N1 so bn<|b|/2

there is an N2 where |bnb|<ϵ|b|22 so take max{N1,N2}

|1bn1b|=|bbn|1|b||bn|<ϵ|b|221|b||b|2=ϵ.

Order Theorems

Assume limnan=a and limnbn=b

(i)n,an0a0(ii)n,anbnab(iii)n,cbncbancac

Proofs

Proof i
For contradiction assume that a<0
ϵ=|a| so for some N, nN, |ana|<|a| so aN<0 which is a contradiction

Proof ii
by the algebraic limit theorems (bnan)ba. Since bnan0 using i we know ba0

Proof iii
Take an=c for all nN and apply ii


Bounded Sequence

A sequence (xn) is bounded if there exists a number M>0 so that |xn|M for all nN

Thm 2.3.2 Convergent then bounded

Proof

Assume (xn)l. So ϵ>0N so that when n>N|xnl|<ϵ

So let ϵ=1, for some N, all n>N we have

|xnl|<1|xn|<|l|+1

Now the set of xi not in Vε=1 is finite. Let

M=max{|x1|,|x2|,,|xN1|,|l|+1}

therefore |xn|M for all nN

Monotone Sequence

A sequence (an) is increasing if anan+1 for all nN and decreasing if anan+1 and monotonic if either.


Thm Squeeze

ymxnzn for all nN
Assume (yn)L and (zn)L. Then (xn)L

Proof

ϵ>0 since (yn)L there exists an NyN so that for all n>Ny, Lϵ<yn<Lϵ

Since (zn)L there exists an NzN so that for all n>Ny we have Lϵ<zn<L+ϵ

let N=max{Ny,Nz}
nN
Lϵ<ynxnzn<L+ϵ

Thm Ratio Test

Let xn>0 for all nN and limxn+1xn=L exists. If L<1 then (xn) converges and (xn)0

Proof

Let rR so that L<r<1. Set ϵ=rL, there exists kN so that xn+1xn<L+ϵ=r for all n>k

So, 0<xk+1<rxk so

xk+1<rxkxk+2<rxk+1<r2xkxk+m<rmxk(0<r<1)rm0

So (xk+m)0
(xn)0

Thm 2.4.2 Monotone Convergence

If (an) is monotonic bounded then it converges if and only if it is bounded

Proof

Assume an is increasing and bounded and set s=sup{an:nN},

sϵ<aNans<s+ϵ

by lemma 1.3.8

so |ans|<ϵ

by thm 2.3.2 it is bounded.


Subsequences

Let (an) be a sequence and let n1<n2<<n5 be a sequence of increasing natural numbers. The then sequence

(an1,an2,)

is a subsequence of (an) and is denoted by ank where kN indexes the subsequence.

Thm Convergence and subsequences (divergence)

The subsequences of a convergent sequence converges to the same limit as the original sequence.

Proof

Assume (an)a and let(ank) be a subsequence.

For any ε>0 there exist an NN so that |ana|<ε for any n>N. nkk for all k, then |anka|<ε for any kN

Thm Bolzano Weierstrass

Every bounded Sequence contains a convergent sequence

Proofs

Using Nested Interval Property

Let (an) be bounded so that there exists an M>0 satisfying |an|M for all nN.
Take [M,0] and [0,M]. At least one interval must have an infinite number of terms outside the epsilon neighbourhood.

Let I1 be the the interval with infinite elements. Let an1 be a term in the sequence (an) such that an1I1.

Bisect I1 and let I2 have the infinite number of terms.

Prick an2 from the original sequence with n2>n1 and an2I2 In general, construct the closed interval Ik taking a half of Ik1 containing an infinite number of terms of (an) and then select nk>nk1>···>n2>n1 so that ankIk

I1I2I3

By the the nested interval property there exists at least one xR for every Ik. WTS that (ank)x

lets ε>0 the length of Ik is M(12)k1 which converges to 0. Choose N so that kN the length of Ik<ε. Since x,ankIk it follows that |ankx|<ε

Using Monotone Convergence Theorem

Lemma:
Every Sequence has a monotone subsequence
Let element an of (an) be called a castle if anak for all kn

Case 1: There are infinitely many castles
an1,an2, then an1an2an3
and so (ani) is a decreasing sequence

Case 2: Only finitely many castles in (an)
There an1>ak for every castle ak where any element past an1 is not a castle.
so an2>an1 and an2 is not a castle. an3>an2 and so on, so strictly increasing.

Monotone + bounded = convergent

Limit Superior

Let (xn) be bounded
is the infimum of the set V of vR such that v<xn for at most finitely many nN
denoted lim supan or limxn

Alternatively lim supxn=infsupxm for n=1,2,3,m=n,n+1,

Limit Inferior

Let (xn) be bounded
is the supremum of the W of wR such that x>an for at most finitely many nN

Alternatively lim infxn=supinfxm for n=1,2,3,m=n,n+1,

Example:

(an)=212,6+13,214,6+15,

infan=212

throw away 2 terms
inf an=214

throw away 2n terms
inf an=212n
take the collection and lim inf = 2

Now:

{ak:k2}sup:6+13{ak:k4}sup:6+15{ak:k2n}sup:6+12n+1infnNsupm=n,n+1,=6

Thm Limit with limsup and liminf

Let an be bounded then lim(an)=a iff lim sup(an)=liminf(an)=a


Cauchy:

if the terms of (an) become "closer" as n then an should converge

A sequence is a Cauchy Sequence if ε>0 there exists kN so that if n,mk then |anam|<ε

note m and n are independent can assume m>n

Thm Convergent iff Cauchy

Proof

Converges then Cauchy

If (an)l then an is Cauchy
Let ε>0 and K be such that |anL|<ε2 for nK

If m,nk

|anam|=|anL+Lam||anL|+|anL|<ε2+ε2=ε

Lemma Cauchy then Bounded

If (an) is Cauchy then it is Bounded

From the definition of a Cauchy Sequence then there exists kN so that for nk that |anak|<ε=set1

By triangle

|an||anak|+|ak|
|an||ak|+1 for all nk
Set M=max{|a1|,|a2|,,|ak|,|ak+1|}

Then |an|M for all nN


Since (an) is Cauchy it is bounded so by BW there exists and convergent subsequence (ank)L

Let ε>0 since an is Cauchy, there exists MN so that if m,nM, then |anam|<ε/2

Also since (ank)L there exists nlM so that |anlL|ε2

So if

nM|anL|=|ananl+anlL||ananl|+|anlL|ε2+ε2=ε

Example

(1+(1)n)

There exists some value ε0>0 so that for every kN there exists at least one value nk and mk so that

|anam|>ε0

Have a2n=2 and a2n+1=0. Take ε0=1 then for any kN choose n even and m=n+1 then

|anan+1|=2>1

Thm Archimedean and Cauchy to Completeness

Suppose every Cauchy Sequence in R converges to an element in R and that the Archimedean property holds. Then R satisfies the completeness property.

completeness -> MCT -> BW -> Cauchy