lacunary - Mathnotes

Sequences

Review the definition of a sequence:

Definition: Sequence \@{sequence}

A sequence is a function f defined on the set J of all positive integers.

Referenced by (52 direct, 36 transitive)

Direct references:

Definition: Converge \@{converge}

A sequence {pn} in a metric space X is said to converge if there is a point pX with the following property: For every ϵ>0, there is an integer N such that nN implies that d(pn,p)<ϵ.

Referenced by (4 direct)
Definition: Limit (Sequence) \@{limit-sequence}

If a sequence {pn} converges to p, we say that p is the limit of {pn}, denoted as:

limnpn=p.

Definition: Diverge \@{diverge}

If a sequence {pn} does not converge, it is said to diverge.

Referenced by (1 direct)

Direct references:

Definition: Range (sequence) \@{range-sequence}

The set of all points pn of a sequence {pn}(n=1,2,3,) is the range of {pn}.

Note \@{sequence-range-cardinality}

The range (sequence) of a sequence may be finite or it may be infinite.

Definition: Bounded (sequence) \@{bounded-sequence}

The sequence {pn} is said to be bounded if its range (sequence) is bounded.

Note \@{sequence-theorems-context}

In the following theorems, let {pn} be a sequence in a metric space X.

Theorem \@{sequence-converges-iff-neighborhood-contains-all-but-finitely-many-points}

{pn} converges to pX if and only if every neighborhood of p contains pn for all but finitely many n.

Proof \@{proof-of-sequence-converges-iff-neighborhood-contains-all-but-finitely-many-points}

Suppose {pn} converges to pX. Let ϵ>0. For some integer N, d(p,pn)<ϵ when n>N. Therefore, pnBϵ(p) for all but the finitely many pn where nN.

Conversely, suppose every neighborhood of p contains all but finitely many pn, i.e., for all but N elements of {pn}. Let ϵ>0. Then, pnBϵ(p) whenever nN, therefore, d(p,pn)<ϵ.

Theorem \@{limits-of-sequences-are-unique}

If pX,pX and {pn} converges to p and p, then p=p.

Proof \@{proof-of-limits-of-sequences-are-unique}

Suppose, for contradiction, that pp. Then, ϵ=d(p,p)>0. Let δ=ϵ/2 and Bδ(p),Bδ(p) be balls around p and p, respectively. This means that only finitely many points from {pn} are not in Bδ(p). However, since Bδ(p) and Bδ(p) are disjoint by construction, this means only finitely many points of {pn} are in Bδ(p), a contradiction. Therefore, our assumption that pp is incorrect, so p=p.

Theorem \@{convergent-sequences-are-bounded}

If {pn} converges, then {pn} is bounded (sequence).

Proof \@{proof-of-convergent-sequences-are-bounded}

Let ϵ>0. Only finitely many points in {pn} lie outside of Bϵ(p). That is, for some integer N, only the points pn where nN lie outside of Bϵ(p). Let δ=max{ϵ,d(p,p1),d(p,p2),,d(p,pn)},n=1,2,,N. Then, d(p,pn)<δ for all n=1,2,3,.

Theorem \@{limit-point-implies-convergent-sequence}

If EX and if p is a limit point of E, then there is a sequence {pn} in E such that p=limnpn.

Proof \@{proof-of-limit-point-implies-convergent-sequence}

For each n=1,2,3, there is a point pnE such that d(p,pn)<1/n. Let ϵ>0, and pick N so that Nϵ>1. Then, if n>N, d(p,pn)<ϵ, so limnpn=p.

Theorem \@{sums-and-products-of-sequences}

Suppose {sn} and {tn} are complex sequences, and limnsn=s,limntn=t. Then

(a) limn(sn+tn)=s+t;

(b) limncsn=cs, for any number c;

(c) limn(c+sn)=c+s, for any number c;

(d) limnsntn=st;

(e) limn1sn=1s,sn0,s0.

Proof \@{proof-of-sums-and-products-of-sequences}

For (a), let ϵ>0, pick Ns such that |sns|<ϵ/2 when nNs, and pick Nt such that |tns|<ϵ/2 when nNt. Then, let N=max{Ns,Nt}. Then, when nN,

|(sn+tn)(s+t)|=|(sns)+(tnt)||sns|+|tnt|<ϵ.

For (b), If c=0, let ϵ>0, and then |csncs|=0<ϵ. Otherwise,

Let ϵ/|c|>0. For some N, when nN, we have

|sns|<ϵ/|c||c||sns|<ϵ|csncs|<ϵ.

For (c), let ϵ>0. For some N, when nN, we have

|(sn+c)(s+c)|=|sns|<ϵ.

For (d), first note the identity

sntnst=(sns)(tnt)+s(tnt)+t(sns).(1)

Now, let ϵ>0. Pick Ns such that |sns|<ϵ when nNs, and pick Nt such that |tns|<ϵ when nNt. Then, let N=max{Ns,Nt}. Then, when nN,

|(sns)(tnt)|<ϵ,

which means

limn(sns)(tnt)=0.

Applying this, along with the results of (a) and (b) to (1) gives:

limn(sntnst)=limn((sns)(tnt)+s(tnt)+t(sns))=limn(s(tnt)+t(sns))=limn(s(tnt))+limn(t(sns))=s0+t0=0.

For (e), pick m such that when nm,|sns|<12|s|, so we have that

|sns|<12|s||sn|+|s|<12|s||sn|<12|s||sn|>12|s|12|sn||s|<1.

Now, let ϵ>0. For some integer N>m, when nM, we have that

|sns|<12|s|2ϵ.

Thus, when nN,

|1sn1s|=|snssns|=|sns||sns|<12|sn||s||s|2ϵ=12|sn||s|ϵ<ϵ.

Referenced by (1 direct)

Direct references:

Theorem: A sequence in Rk converges iff its components converge \@{sequence-in-rk-converges-iff-its-components-converge}

(a) Suppose xnRk,(n=1,2,3,) and xn=(α1,n,,αk,n).

Then, {xn} converges to x=(α1,,αk) if and only if

limnαj,n=αj,(1jk).

(b) Suppose {xn},{yn} are sequences in Rk, {βn} is a sequence of real numbers, and xnx,yny,βnβ. Then

limn(xn+yn)=x+y,limnxnyn=xy,limnβnxn=βx.

Proof \@{proof-of-sequence-in-rk-converges-iff-its-components-converge}

For (a), assume xnx. Then, from the definition of the norm,

|αj,nαj||xnx|,

that is, the distance from αk,n to αn is always less than or equal to the distance from xn to x. Therefore, for ϵ>0, |xnx|<ϵ|αj,nαj|<ϵ, and we can pick n to make this true for as small of ϵ as we'd like. Therefore, limnαj,n=αj.

Conversely, assume limnαj,n=αj. Let ϵ>0. For some integer N, when nN we have

|αj,nαj|ϵk,(1jk).

Therefore, nN implies that

|xnx|=j=1k|αj,nαj|2<ϵ,

so xnx.

Part (b) follows from part (a) and A sequence in Rk converges iff its components converge.

Subsequences

Definition: Subsequential limit \@{subsequential-limit}

If a subsequence {pni} of {pn} converges, its limit (sequence) is called a subsequential limit of {pn}.

Theorem \@{theorem-23}

A sequence {pn} converges to p if and only if every subsequence of {pn} converges to p.

Proof \@{proof-of-theorem-23}

Suppose that {pn} converges to p. Suppose some subsequence {pni} converges to q, and suppose, for contradiction, that qp. Now, following an argument similar to the proof that limits of sequences are unique, we can see that if pq, arbitrary neighborhoods around both can't contain all but finitely many points, so we have a contradiction, and p=q.

Conversely, suppose every subsequence of {pn} converges to p. Then, {pn} is a subsequence of itself, so it converges to p.

Theorem \@{sequence-in-compact-metric-space-has-a-convergent-subsequence}

If {pn} is a sequence in a compact metric space X, then some subsequence of {pn} converges to a point in X.

Proof \@{proof-of-sequence-in-compact-metric-space-has-a-convergent-subsequence}

Let E be the range of {pn}. If E is finite, then at least one point p in E must be repeated infinitely many times in {pn}. If we let {ni} be the indices of the occurrences of p in {pn}:

pn1=pn2==p,

then the subsequence {pni} converges to p.

On the other hand, if E is infinite, then E has a limit point pX. Pick n1 so that d(p,pn1<1. Now, after picking n1,,ni1, we can pick ni>ni1 such that d(p,pni)<1/i, so {pni} converges to p.

Referenced by (1 direct)

Direct references:

Theorem: Bolzano-Weierstrass \@{theorem-27}

Every bounded sequence in Rk contains a convergent subsequence.

Proof \@{proof-of-theorem-27}

Note that any bounded sequence {pn}Rk is a subset of some closed set, bounded and thus compact k-cell in Rk. Therefore, {pn} is a sequence in a compact metric space, and has a convergent subsequence.

Theorem \@{subsequential-limits-of-a-metric-space-form-a-closed-set}

The subsequential limits of a sequence {pn} in a metric space X form a closed set subset of X.

Proof \@{proof-of-subsequential-limits-of-a-metric-space-form-a-closed-set}

Let E be the set of all subsequential limits of {pn} and let q be a limit point of E. We want to show that qE.

First, note that if the range of {pn} is just {q}, then q is the only subsequential limit of {pn}. In this case, E={q} is a singleton and is closed set, as it vacuously contains all of its limit points. So, assume this is not the case.

Choose n1 so that pn1q, and let δ=d(q,pn1). Suppose n1,,ni1 are chosen. Since q is a limit point of E, there is an xE with d(q,x)<δ2i. Since xE and is thus the limit (sequence) of some subsequence of {pn}, there is an ni>ni1 such that d(x,pni)<δ2i. Now, via the triangle inequality,

d(q,pni)d(q,x)+d(x,pni)<δ2i+δ2i=2δ2i=δ2i1,i=1,2,3,.

This means that {pni} converges to q, because we can find a pni as close as desired to q. Therefore q is a subsequential limit of {pn} and so qE, and E is closed.

Referenced by (1 direct)

Direct references:

Note \@{note-31}

The theorem above tells us about the long term behavior of a sequence, even if it doesn't converge. The set of all subsequential limits of {pn} gives us the set of all points that are approached arbitrarily closely infinitely often in {pn}. It's basically the set of points that {pn} likes to hang out around! E being closed means that if there is a point in X that the points of E get arbitrarily close to, then it is also a point {pn} likes to hang out around.

Cauchy Sequences

Definition: Diameter \@{diameter}

Let E be a nonempty subset of a metric space X, and let S be the set of all real numbers of the form d(p,q), with p,qE. The supremum of S is called the diameter of E.

Referenced by (1 direct)
Theorem \@{limit-of-diameter-of-remaining-points-in-cauchy-sequence-is-zero}

if {pn} is a sequence in X and if EN consists of the points pN,pN+1,pN+2,, then {pn} is a Cauchy sequence if and only if

limNdiamEN=0.

Proof \@{proof-of-limit-of-diameter-of-remaining-points-in-cauchy-sequence-is-zero}

Suppose {pn} is a Cauchy sequence. Then, let ϵ>0. For some integer N, d(pn,pm)<ϵ when n,mN. Therefore, diamEN<ϵ. Since ϵ was arbitrary, we can see that the sequence {diamEN} converges to 0.

Conversely, suppose limNdiamEN=0. Then, every neighborhood of 0 contains {diamEN} for all but finitely many N. Let ϵ>0. Then, pick N such that diamEn<ϵ. Letting m,nN, we have that d(pm,pn)diamEn<ϵ.

Referenced by (1 direct, 1 transitive)
Theorem \@{diameter-of-set-equals-diameter-of-closure}

If E is the closure of a set E in a metric space X, then

diamE=diamE.

Proof \@{proof-of-diameter-of-set-equals-diameter-of-closure}

Because EE, diamEdiamE.

Conversely, Let p,qE, and p,qE, such that d(p,p)<ϵ,d(q,q)<ϵ. Therefore, by the triangle inequality,

d(p,q)d(p,p)+d(p,q)+d(q,q)<2ϵ+d(p,q)2ϵ+diamE.

Therefore, diamE2ϵ+diamE, and since ϵ was arbitrary, diamE=diamE.

Referenced by (1 direct, 1 transitive)
Theorem \@{nested-sequence-of-compact-sets-with-lim-diam-zero-has-singleton-intersection}

If Kn is a sequence of nonempty compact sets in X such that Kn+1Kn,(n=1,2,3,) and if

limndiamKn=0,

then 1Kn consists of exactly one point.

Proof \@{proof-of-nested-sequence-of-compact-sets-with-lim-diam-zero-has-singleton-intersection}

Let K=1Kn. Then K is not empty. Assume for the sake of contradiction that K contains more than one point. Then, diamK>0. But, for each n,KKn, so that diamKndiamK. But, this contradicts our given that lim+ndiamKn=0, so our assumption that K contains more than one point must be invalid, and thus K contains exactly one point.

Referenced by (1 direct, 1 transitive)
Theorem \@{every-convergent-sequence-in-a-metric-space-is-a-cauchy-sequence}

Every convergent sequence in a metric space X is a Cauchy sequence.

Proof \@{proof-of-every-convergent-sequence-in-a-metric-space-is-a-cauchy-sequence}

Suppose {pn} is a convergent sequence in a metric space X. Let ϵ>0. Then for some N,d(p,pn)<ϵ when nN. Thus,

d(pn,pm)d(p,pn)+d(p,pm)<2ϵ

whenever n,mN, and so {pn} is Cauchy.

Referenced by (1 direct)
Theorem \@{compact-metric-spaces-are-complete}

Compact metric spaces are complete.

That is, if X is a compact metric space and if {pn} is a cauchy sequence in X, then {pn} converges to some point of X.

Proof \@{proof-of-compact-metric-spaces-are-complete}

Let {pn} be a cauchy sequence in the compact metric space X. For N=1,2,3,, let En bet the set consisting of pN,pN+1,pN+2,. Then

limNdiamEN=0,

by two theorems above. Each En is a closed subset of the compact space X, and is thus compact. Also, EN+1EN, which implies that EN+1EN. Now, we have that there is a unique pX which lies in every En.

Let ϵ>0. Since limNdiamEN=0, there is an integer N0 such that diamEn<ϵ if NN0. Since pEn, we have that d(p,q)<ϵ for every qEN and thus for every qEn. That is, d(p,pn)<ϵ if nN0, so {pn} converges to p.

Referenced by (1 direct)
Referenced by (2 direct)
Theorem \@{euclidean-spaces-are-complete}

All Euclidean spaces are complete

That is, in Rk, every cauchy sequence converges.

Proof \@{proof-of-euclidean-spaces-are-complete}

Let {xn} be a cauchy sequence in Rk. Define EN as in the proof above, but with xi in place of pi. For some N,diamEn<1. The range (sequence) of {xn} is the union of En and the finite set {x1,,xN1}. Hence, {xn} is bounded (sequence) (since the finite set of points can be contained in some bounding box, and the remaining points can be contained in some ball with diameter 1.) Since {xn} is bounded, it is compact, and thus {xn} is a subset of a compact metric space (its closure), and so converges.

Referenced by (1 direct)
Theorem: Cauchy criterion for convergence \@{cauchy-criterion-for-convergence}

A sequence converges in Rk if and only if it is a cauchy sequence.

Proof \@{proof-of-cauchy-criterion-for-convergence}

Suppose {pn}Rk converges. Then, because Rk is a metric space, {pn} is cauchy.

Conversely, suppose {pn}Rk is cauchy. Then, {pn} converges.

Note \@{note-49}

An important difference between the definition of a convergent sequence and a cauchy sequence is that the limit (sequence) is explicitly involved in the former, but not the latter. Thus, we may be able to decide whether or not a given sequence converges without knowledge of the limit (sequence) to which it may converge.

Theorem \@{theorem-50}

Every closed subset E of a complete metric space X is complete.

Proof \@{proof-of-theorem-50}

Let {pn}E be a cauchy sequence. Then, it converges to some point pX, and actually pE, because E is closed. Therefore, E is complete.

Example \@{example-52}

Not all metric spaces are complete. For example, the space of all rationals with d(x,y)=|xy| is not complete.

For more content specifically on real sequences, see Real Sequences.