Limits and Continuous Functions

Arun Ram
Department of Mathematics and Statistics
University of Melbourne
Parkville, VIC 3010 Australia
aram@unimelb.edu.au

Last updates: 25 January 2014

Filters and limits

Let X be a set.

Let Y be a topological space. Let yY.

Let Y be a topological space and let 𝒢 be a filter on Y.

The issue of which properites of a space guarantee that limit points are unique (when they exist) is treated in Section ??: Hausdorff spaces, and the the issue of which properties of a space guarantee that cluster points exist is treated in Section ??: Compact spaces.

HW: Let Y be a topological space and let 𝒢 and be filters on Y such that 𝒢.

  1. (a) Show that if y is a limit point of then y is a limit point of 𝒢.
  2. (b) Show that if y is a cluster point of 𝒢 then y is a cluster point of .

HW: Let Y be a topological space and let 𝒢 be a filter on Y. Let yY. Show that y is a cluster point of 𝒢 if and only if there exists a filter on Y such that 𝒢 and 𝒩(y).

HW: Let Y be a topological space and let 𝒢 be a filter on Y. Let yY. Show that y is a cluster point of 𝒢 if and only if there exists a filter 𝒢 such that y is a limit point of .

HW: Let Y be a topological space and let 𝒢 be a filter on Y. Let yY. Show that if y is a limit point of 𝒢 then y is a cluster point of 𝒢.

HW: Let Y be a topological space and let 𝒢 be an ultrafilter on Y. Let yY. Show that y is a limit point of 𝒢 if and only if y is a cluster point of 𝒢.

Let X be a set with a filter and let Y be a topological space. Let f:XY be a function.

Sequences

Let Y be a topological space.

Write     y=lim nyn,    if y is a limit point of the sequence y1 y2 y3 .

Let Y be a topological space and let y1 y2 y3 be a sequence in Y. Let yY.

  1. y is a limit point of y1 y2 y3 if and only if
    1. if Ny is a neighbourhood of y then there exists n0 >0 such that if n >0 and n>n0 then ynNy.
  2. y is a cluster point of y1 y2 y3 if and only if
    1. if Ny is a neighbourhood of y and n0 >0 then there exists n >0 such that n>n0 and ynNy.

Proof.
  1. ⇒) Assume y is a limit point of (y1,y2,...).
    So y = limn yn = lim f(n), where f: >0 y n yn and is the Fréchet filter on >0.
    So 𝒩(y) f().
    So 𝒩(y) f ( complements of finite subsets of>0 ).
    To show: If Ny is a neighbourhood of y then there exists n0>0 such that if n>0 and n>n0 then ynNy.
    Assume Ny is a neighbourhood of y.
    So Ny𝒩(y).
    So there exists a finite set S>0 such that Ny f(Sc).
    Let n0=sup(S).
    To show: If n>0 and n>n0 then ynNy.
    Assume n>0 and n>n0.
    Then nSc.
    So yn = f(n) f(Sc).
    Since f(Sc)Ny,  ynNy.

    ⇐) Assume that if Ny is a neighbourhood of y then there exists n0>0 such that if n>0 and n>n0 then ynNy.
    To show: y = limnyn. To show: 𝒩(y) f ( complements of finite subsets of>0 ). To show: If Ny𝒩(y) then there exists a finite subset S>0 such that Nf(Sc).
    Assume Ny𝒩(y).
    To show: There exists a finite subset S>0 such that Nyf(Sc).
    Let n0>0 such that if n>0 and n>n0 then ynNy.
    Let S={1,2,...,n0} so that S is a finite subset of >0.
    Then f(Sc) = f({n0+1,n0+2,...}) = {yn0+1,yn0+2,...} = {yn  |  n>n0} Ny.
  2. ⇒) Assume y is a cluster point of (y1,y2,...).
    To show: If Ny is a neighbourhood of y and n0>0 then there exists n>0 such that n>n0 and ynNy.
    We know: If U f ( complements of finite subsets of>0 ). then yU_.
    Assume Ny is a neighbourhood of y and n0>0.
    To show: There exists n>0 such that n>n0 and ynNy.
    To show: There exists n{1,...,n0}c such that ynNy.
    To show: f({1,...,n0}c) Ny . We know: y f({1,...,n0}c) _. So Ny f({1,...,n0}c) .

    ⇐) Assume that if Ny is a neighbourhood of y and n0>0 then there exists n>0 such that n>n0 and ynNy.
    To show: If U f ( complements of finite subsets of>0 ). then yU_.
    Assume U f ( complements of finite subsets of>0 ).
    So Uf(Sc) for a finite set S>0.
    Let n0=sup(S).
    Then Uf(Sc) {yn0+1,yn0+2,...} = {yn  |  n>n0}.
    To show: yU_.
    To show: If Ny is a neighbourhood of y then NyU.
    Assume Ny is a neighbourhood of y.
    Since Ny {yn  |  n>n0} and {yn  |  n>n0} U, then NyU.

Example: The sequence yn = { 1-1n, if nis even, 0, if nis odd, in has no limit point, but has cluster points at 1 and 0.

Limits and continuity

Let X and Y be topological spaces and let f:XY be a function. Let aX.
A limit of f as x approaches a is a limit point of f with respect to the neighbourhood filter 𝒩(a).

Write     y= limxa f(x) ,    if y is a limit of f as x approaches a.

The function f:XY is continuous at x=a if it satisfies

  1. if N is a neighbourhood of f(a) in Y then f-1(N) is a neighbourhood of a in X.

Let X and Y be topological spaces and let aX. A function f:XY is continuous at x=a if and only if limxa f(x) = f(a).

Proof.
⇒) Assume f:XY is continuous at x=a.
To show: limxaf(x)=f(a)
To show: 𝒩(f(a)) f(𝒩(a)).
To show: If N𝒩(f(a)) then there exists U𝒩(a) such that f(U)N.
Assume N𝒩(f(a)).
To show: There exists U𝒩(a) such that f(U)N.
Let U=f-1(N). Since f is continuous at x=a, U𝒩(a).
To show: f(U)N.
Since f(U) = f(f-1(N)) = N,  f(U) N. So limxa f(x) = f(a).

⇐) Assume f(a) = limxa f(x). So 𝒩(f(a)) f(𝒩(a)).
To show: If N𝒩(f(a)) then f-1(N) 𝒩(a).
Assume N𝒩(f(a)).
To show: f-1(N) 𝒩(a).
Since 𝒩(f(a)) f(𝒩(a)) we know:
There exists U𝒩(a) such that Nf(U).
So f-1(N) U.
So f-1(N) 𝒩(a).

Generating filters

Let X be a set and let 1 and 2 be filters on X. The filter 1 is finer than 2 if 12.

If is a collection of subsets of X that satisfies

(a)   if B1, B2 then there exists B such that BB1 B2, and
(b)   and ,
then
= { subsets ofX that contain a set in }
is the coarsest filter containing .

Let X be a set and let 𝒮 be a collection of subsets of X. If

= {finite intersections of elements of 𝒮 }     and    = {subsets ofX containing a set in }
then satisfies (a) and (b) and is the coarsest filter on X containing 𝒮.

Metric spaces

Let X and Y be metric spaces and let f:XY be a function. Let aX and yY. Then limxa f(x) = y if and only if if ε>0 then there exists δ>0 such that if d(x,a) <δ then d(f(x),y) <ε.

Let Y be a metric space with the metric space uniformity.
Let (y1, y2, ) be a sequence in Y and let (y1, y2, ) be the filter associated to (y1, y2, ) . Then (y1, y2, ) is a Cauchy sequence in Y if and only if (y1, y2, ) is a Cauchy filter on Y,

Notes and References

Historically, the mathematical community became infatuated by limits, partly because of the many applications of "calculus" and the ideas of infinitesimals, but also partly because they weren't very well understood. This infatuation has often focused on the epsilon-delta definition of limits, which has advantages and disadvantages.

The thorough and easy definition of limits given in this exposition, not depending on a metric space, follows [Bou, Top., Ch.I, §6-7]. In particular, the definition of filter, neighbourhood filter, and Fréchet filter, are in [Bou, Top., Ch.I, §6 No 1] and the definition of limit point and cluster point of a filter are [Bou, Top., Ch.I, §7 Def. 1,2] and the definition of limit point and cluster point of a function are [Bou, Top., Ch.I, §7 Def 3].

Theorem 2.1 is [Bou, Top., Ch.I, §7 Prop 9] and Theorem 3.1 is Example 1 in [Bou, Top., Ch.I, §7 No 3]. The discussion in §4 Generating filters is based on [Bou, Top., Ch.I, §6 No. 2-3].

The second and third conditions in the definition of a filter say that finite intersections of elements of a filter cannot be empty. This is the rigidity condition that plays an important role in arguments about limit points and compactness.

A point y is a cluster point of 𝒢 if y is a close point for all the sets C𝒢.

The partial order on filters by inclusion is fundamental. A point y is a cluster point if the supremum of 𝒢 and the neighborhood filter of y exists. If filters are analogous to sequences then inclusion of filters is analogous to subsequences. In particular, the existence of an ultrafilter with a limit point is analogous to a subsequence with a limit point.

References

[Bou] N. Bourbaki, General Topology, Springer-Verlag, 1989. MR1726779.

[BR] W. Rudin, Principles of mathematical analysis, Third edition, International Series in Pure and Applied Mathematics, McGraw-Hill 1976. MR0385023.

[Ru] W. Rudin, Real and complex analysis, Third edition, McGraw-Hill, 1987. MR0924157.

page history