lacunary - Mathnotes

Metric Spaces

Note \@{metric-spaces-reference-note}

This section was developed by following Rudin, Principles of Mathematical Analysis, Chapter 2.

Definition of a Metric Space

Definition: Metric Space \@{metric-space}

A set X, whose elements we'll call points, together with a distance function d:X×XR is called a metric space and the distance function d is called a metric, if the following conditions, called the metric axioms, hold for p,q,rX:

  • If pq,d(p,q)>0. (distance is always positive between two distinct points.)
  • d(p,p)=0. (distance is always zero between a point and itself.)
  • d(p,q)=d(q,p). (the distance from p to q is the same as the distance from q to p.)
  • d(p,q)d(p,r)+d(r,p) (triangle inequality.)

We can denote a metric space on set X with metric d as the tuple (X,d).

Referenced by (47 direct, 64 transitive)

Direct references:

Transitive (depth 1):

Transitive (depth 2):

Transitive (depth 3):

Examples of Metric Spaces

The most important metric spaces are the euclidean spaces Rn together with a norm.

Theorem \@{euclidean-space-is-metric-space}

(Rn,d:RnR=|xy|) is a metric space for any n0.

Proof \@{proof-of-euclidean-space-is-metric-space}

First, for n=0, R0 is just the empty set, so the metric axioms are vacuously satisfied for all points in the set. Now, for n1,

  • Let p,qRn,pq. Then |pq|<0, so d(p,q)>0.
  • Let pRn. Now, pp=0, so |pp|=0, so d(p,p)=0.
  • Let p,qRn. |pq|=|qp|, so d(p,q)=d(q,p).
  • Let p,q,rRn. |pq||pr|+|rp|, so d(p,q)d(p,r)+d(r,p).

Therefore, d is a metric on Rn and (Rn,d) is a metric space.

Remark \@{rn-vector-space-vs-metric-space}

Note that Rn by itself is just a vector space, but together with a norm forms a metric space. The proof above relies on the properties of norms without going into detail - norms are similar to distance metrics but they are even stricter because they require homogeneity - a|x|=|ax|.

We could define other, non-norm metrics on Rn, such as the discrete metric, which works for any set X.

Also note that every subset of a Y of a metric space X is a metric space on its own, using the same distance function.

Definition: Discrete Metric \@{discrete-metric}

The discrete metric is defined as:

d(x,y)={0if x=y1if xy

Referenced by (1 direct)
Remark \@{discrete-metric-satisfies-axioms}

It's easy to see that this satisfies the metric axioms above for any set.

Also note that every subset Y of a metric space X is also a metric space, since if the metric axioms hold for all points in X, that necessarily includes the points in Y.

Segments, Intervals, Cells and Balls

Definition: Segment \@{segment}

A segment (a,b) is the set of all real numbers x such that a<x<b.

Definition: Half-open Interval \@{half-open-interval}

A half-open interval (a,b] or [a,b) is the set of all real numbers such that a<xb or ax<b, respectively.

Definition: k-cell \@{k-cell}

Given a,bRk, if ai<bi for all i=1,2,,k, then the set of all points x who satisfy aixibi, i=1,2,,k, is called a k-cell. So, a 1-cell is an interval, a 2-cell is a rectangle, and so on.

Referenced by (1 direct)

Direct references:

Definition: Ball \@{ball}

Given xRk,r>0, the open or closed ball with center x and radius r is defined as the set of points y such that |xy|<r or |xy|r, respectively.

Referenced by (8 direct, 200 transitive)

Transitive (depth 1):

Transitive (depth 2):

Transitive (depth 3):

Transitive (depth 4):

Transitive (depth 5):

Transitive (depth 6):

Transitive (depth 7):

Convex Sets

Definition: Convex Set \@{convex-set}

A set ERk is said to be convex if

λx+(1λ)yE

whenever x,yE, and 0<λ<1.

In geometric terms, this means a set is convex if we can connect any two points in the set with a line segment whose points are all within the set.

Theorem \@{balls-are-convex}

All balls are convex.

Proof \@{proof-of-balls-are-convex}

Let y,z be points in a ball with center x and radius r. Then, by definition, |yx|<r, and |zx|<r. Suppose p{λy+(1λ)z|0<λ<1}. We will show that |xp|<r.

|xp|=|x(λy+(1λ)z)|(substitute definition of p)=|xλyz+λz|(expand)=|xλyz+λz+λxλx|(add and subtract λx)=|λ(xy)+(1λ)(xz)|(factor λ and 1λ)λ|xy|+(1λ)|xz|(triangle inequality)<λr+(1λ)r(since |xy|,|xz|<r)=r(because λ+(1λ)=1).

So, p is within our ball and therefore all balls are convex.

Note \@{balls-are-convex-note}

Similar proofs can be used to show that closed balls and k-cells are also convex.

Elements and Subsets of a Metric Space

Note \@{metric-space-context}

Assume we have a metric space (X,d).

Definition: Neighborhood \@{neighborhood}

A neighborhood, or r-neighborhood of p is a set Nr(p) consisting of all q such that d(p,q)<r for some r>0. This subset of X is all the points within a circle of radius r - the open ball of radius r centered at p.

Referenced by (35 direct, 169 transitive)

Transitive (depth 1):

Transitive (depth 2):

Transitive (depth 3):

Transitive (depth 4):

Transitive (depth 5):

Transitive (depth 6):

Definition: Limit Point \@{limit-point}

A point p is a limit point of the set E if every neighborhood of p contains a point qp such that qE.

Referenced by (30 direct, 88 transitive)

Transitive (depth 1):

Transitive (depth 2):

Transitive (depth 3):

Transitive (depth 4):

Definition: Isolated Point \@{isolated-point}

If pE and p is not a limit point of E, then p is called an isolated point of E.

Referenced by (1 direct)

Direct references:

Definition: Closed \@{closed}

A set E is closed if every limit point of E is a point of E.

Referenced by (35 direct, 16 transitive)
Definition: Interior Point \@{interior-point}

A point p is an interior point of E if there is a neighborhood N of p such that NE.

Referenced by (11 direct, 132 transitive)

Transitive (depth 1):

Transitive (depth 2):

Transitive (depth 3):

Transitive (depth 4):

Transitive (depth 5):

Definition: Open \@{open}

A set E is open if every point of E is an interior point of E.

Referenced by (34 direct, 112 transitive)

Transitive (depth 1):

Transitive (depth 2):

Transitive (depth 3):

Transitive (depth 4):

Theorem \@{every-neighborhood-is-an-open-set}

Every neighborhood is an open set.

Proof \@{proof-of-every-neighborhood-is-an-open-set}

Suppose Nr(p) is a neighborhood in X. Let qNr(p). We need to show that q is an interior point of Nr(p). Let s=rd(p,q); because d(p,q)<r, we have s>0. Now let Ns(q) be the neighborhood of radius s around q. We need to show that Ns(q)Nr(p). Suppose xNs(q). First note that because s=rd(p,q), d(p,q)=rs. Now,

d(p,x)d(p,q)+d(q,x)<rs+s=r.

Therefore, Ns(q)Nr(p), so q is an interior point of Nr(p), and since q was arbitrary, every point of Nr(p) is interior. Hence, Nr(p) is open.

Referenced by (1 direct)
Theorem \@{neighborhood-of-limit-point-contains-infinitely-many-points}

If p is a limit point of a set E, then every neighborhood of p contains infinitely many points of E.

Proof \@{proof-of-neighborhood-of-limit-point-contains-infinitely-many-points}

Let p be a limit point of E and let Nr(p) be a neighborhood of p. Suppose that Nr(p) contains only finitely many points of E. Since we have finitely many points, we can inspect each and find the minimum distance from p to any point in Nr(p)E{p} and call it s. Now, we can make a new neighborhood Ns(p), which contains none of the points in Nr(p)E{p} since they're all at least s away from p, by construction. But then, p is not a limit point of E, since it has a neighborhood that contains no points of E{p} Therefore, we have a contradiction, and Nr(p) must therefore contain infinitely many points.

Corollary \@{only-infinite-sets-have-limit-points}

From this, it's evident that a finite set of points has no limit points. That is, if a set has a limit point, then the set if infinite.

Theorem \@{complement-of-union-is-intersection-of-complements-general}

Let Eα be a collection of sets. Then

(αEα)c=α(Eαc).

Proof \@{proof-of-complement-of-union-is-intersection-of-complements-general}

Suppose x(αEα)c. Then, xαEα, so x is not in any Eα. Therefore, for every Eα, xEαc, and thus xα(Eαc). Conversely, suppose xα(Eαc). Then, x is in every Eαc, that is, x is not in any Eα. Therefore, x(αEα)c.

Note \@{complement-of-union-is-intersection-of-complements-general-note}

This is just De Morgan's law extended to arbitrary indexed collections.

Theorem \@{open-iff-complement-closed}

A set E is open iff its complement is closed.

Proof \@{proof-of-open-iff-complement-closed}

First, consider the case that Ec is empty, and therefore open. If Ec has no limit points, it is vacuously closed. Suppose Ec has a limit point x. Since E is empty, x must be in Ec, therefore x is closed. Now, consider the case that Ec is empty, and therefore closed. If E is empty, it is open, and the theorem is satisfied. If E is not empty, a point x in E has only points in E in any neighborhood, since all points are in E, and therefore E is open.

Now we deal with the cases where neither E nor Ec are empty.

Now, let Ec be closed. Let xE. Since Ec is closed, x is not a limit point of Ec, that is x has some neighborhood that doesn't contain a point in Ec and must therefore be a subset of E. Therefore, x is an interior point of E, and E is open.

Conversely, assume E is open. Let x be a limit point of Ec. Suppose, for the sake of contradiction, that xE. Then, since E is open, x is an interior point of E and has some neighborhood that is a subset of E. This is a contradiction, since every neighborhood of x must contain at least one point of Ec to be a limit point of Ec. Therefore, x must be in Ec, and it follows that Ec is closed.

Remark \@{open-closed-intuition}

For some intuition, consider U=[0,4],E=[0,3]. Any point x in E has a neighborhood Nr(x) that contains only points in E with as r<d(x,3), so it is open. Since 3 must be in Ec, we have Ec=[3,4], which is obviously closed. Here, 3 is the boundary between the two sets - it is a limit point for both sets - and it has to be in one set or the other. So, E is open, but not closed, since it doesn't contain one of its limit points, 3. Ec is closed because if x is a limit point of Ec, it would be a contradiction for it to be in E, which is open and only contains interior points of E. Ec is also not open, because 3 has no neighborhoods that contain only points of Ec.

Theorem \@{union-and-intersection-of-open-and-closed-sets}

(a) - For any collection {Ga} of open sets, αGα is open.

(b) - For any collection {Fa} of closed sets, αFα is closed.

(c) - For any finite collection G1,,Gn of open sets, i=1nGi is open.

(d) - For any finite collection F1,,Fn of closed sets, i=1nFi is closed.

Proof \@{proof-of-union-and-intersection-of-open-and-closed-sets}

Let G=αGα,xG. Then x is in some Gα for some α, and is an interior point of that Gα, since Gα is open. Therefore, x has some neighborhood that is a subset of Gα and therefore of G, so x is an interior point of G and G is open - this shows (a).

Note that

(αFα)c=αFαc,(e)

and by (a) above, (e) is open. Then its complement, αFα is closed, and we've shown (b).

Now, let x be in G=i=1nGi, so x is in every Gi, and has a neighborhood Ni in every Gi with radius ri>0. Let r be min{r1,,rn}. Then, x has a neighborhood N of radius r in every Gi, and thus in G, so x is an interior point of G, and we've shown (c).

Now, (i=1nFi)c=i=1nFic is open by (c), so its complement, i=1nFi is closed, and we've shown (d).

Note \@{union-and-intersection-of-open-and-closed-sets-note}

In parts (c) and (d) of the above theorem, finiteness of the collections of sets is required - the property do not necessarily hold for infinite collections of sets.

Theorem \@{limit-points-form-closed-set}

Let E be the set of all limit points of a set E in space X. Then E is closed.

Proof \@{proof-of-limit-points-form-closed-set}

Let pX,pE. Then some neighborhood N of p contains no points in E, other than possibly p itself. If N contains only p, then p is not a limit point of E. Suppose, for the sake of contradiction, some point qN,qp is a limit point of E. Then, every neighborhood of q contains some point m in E. Let MN,pM, be such a neighborhood, and let mM,mq be a limit point of E. Now, m has neighborhoods wholly in N, and such neighborhoods can have no point in E, so we have a contradiction, and therefore q is not a limit point of E. Hence, p is an interior point of Ec, and Ec is open, and therefore E is closed.

Theorem \@{set-is-its-closure-iff-it-is-closed}

If X is a metric space and EX, then

(a) E is closed.

(b) E=E iff E is closed.

(c) EF for every closed set FX such that EF.

By (a) and (c), E is the smallest closed subset of X that contains E.

Proof \@{proof-of-set-is-its-closure-iff-it-is-closed}

(a) Suppose pX and pE. Then p is not in E and is not in E, and is in fact in Ec. Now, since p is not a limit point of E, it has some neighborhood Nr(p) that does not intersect E. Any point x in Nr(p) is an interior point of Nr(p), and therefore has its own neighborhood Nϵ(x) that does not intersect E, and therefore x is not a limit point of E. Thus, any point in Ec is an interior point of Ec, and Ec is therefore open and its complement, E, is closed (since a set is open iff its complement is closed.)

(b) Suppose E is closed. Then it contains its limit points, so E=EE=E. Conversely, suppose E=E. By (a), E is closed.

(c) Suppose that EFX, and that F is closed. Suppose pE. If pE, then pF because EF. If pE, then it must be in F also, since F contains all points of E, and is closed, and thus must contain the limit points of E.

Theorem \@{set-and-closure-have-same-limit-points}

If E is a set in a metric space, then E and E have the same limit points.

Proof \@{proof-of-set-and-closure-have-same-limit-points}

If E is closed, then we are done, because a set equals its closure if it is closed.

Suppose p is a limit point of E. Then every neighborhood of p contains some qE,qp. Since qE, qE, so p is a limit point of E.

Conversely, suppose p is a limit point of E. Then, every neighborhood N of p contains a point of qE,qp. If qE, then N clearly contains a point in E. Otherwise, qE,qnE. Now, since qN and every neighborhood is an open set, q has some neighborhood MN. Since qE, M contains some point sE. Since MN, sN, and therefore N contains a point in E. Thus, all neighborhoods of p contain some point in E, and p is a limit point of E.

Theorem \@{set-and-its-limit-points-may-have-different-limit-points}

A set E and its limit points E do not necessarily have the same limit points.

Proof \@{proof-of-set-and-its-limit-points-may-have-different-limit-points}

Consider P={1n|nN}. Then P has one limit point, 0, but P={0} has no limit points, since the only number that contains 0 in all its neighborhoods is 0 itself (see From this, it's evident that a finite....)

Theorem \@{closure-distributes-over-finite-unions}

Closure distributes over finite unions.

If Bn=i=inAi, then Bn=i=1nAi,n=1,2,3,.

Proof \@{proof-of-closure-distributes-over-finite-unions}

Let pi=1nAi. Then, pAi for some i. If pAi, then pi=inAi=Bn, so pBn. If p is in some Ai, then every neighborhood of p contains some point qAi,qp, and since AiBn, qBn, so pBn.

Conversely, let pBn. If pBn, pAi for some i, and thus pAii=1nAi. If p is only in Bn, suppose for contradiction that pi=1nAi. For each i, let Ni be a neighborhood centered at p with NiAi=. Then, N=i=1nNi is a neighborhood of p (see (a) - For any collection {Ga} of...,) but NNi= for all i, so NBn=. But, this contradicts our hypothesis that pBn, so our contradictory assumption must be invalid, and pi=1nAi.

Theorem \@{sup-is-in-closure-of-bounded-nonempty-set-of-reals}

Let E be a nonempty set of real numbers which is bounded above. Let y=supE. Then yE. Hence yE if E is closed.

Proof \@{proof-of-sup-is-in-closure-of-bounded-nonempty-set-of-reals}

Suppose yE. Then yE. Suppose yE. Now, by hypothesis, for every h>0, there is some x such that yh<x<y, because otherwise, x would be an upper bound on E. Therefore, every neighborhood Nh(y) contains some xE, and thus y is a limit point of E and yE.

Referenced by (1 direct)
Definition: Open Relative \@{open-relative}

Suppose EYX, and X is a metric space. We say that E is open relative to Y if to each pE there is associated an r>0 such that qE,qY whenever d(p,q)<r.

Theorem \@{open-relative-iff-intersection-with-open-subset}

Suppose YX. A subset E of Y is open relative to Y iff E=YG for some open subset G of X.

Proof \@{proof-of-open-relative-iff-intersection-with-open-subset}

Suppose E is open relative to Y. Then, for every pE, there is some rp>0 such that d(p,q)<rp, qX implies that qY. Let Vp be the set of all qX where d(p,q)<rp (and thus qY) and let

G=pEVp.

Then, since each Vp is an open subset of X, so is G. Now, since pVp for each pE, EGY. Also, since VpYE for every pE, GYE, and E=GY.

Conversely, suppose E=YG for some open subset G of X. Now, suppose pE. Then, pG, and there is some neighborhood VpG. Then, VpYE, so E is open relative to Y.

Referenced by (1 direct)