logo

Limit of Functions in Metric Spaces 📂MetricSpace

Limit of Functions in Metric Spaces

Definition

Let (X,dX)(X,d_{X}), (Y,dY)(Y,d_{Y}) be a metric space. Suppose EXE\subset X, f:EYf: E\rightarrow Y, and that pp is a limit point of EE. Then, for every positive number ε\varepsilon,

xE and dX(x,p)<δ    dY(f(x),q)<ε x \in E \ \text{and} \ d_{X}(x,p)<\delta \implies d_{Y}(f(x),q) <\varepsilon

exists δ>0\delta>0 such that,

f(x)q as xp f(x)\rightarrow q\ \mathrm{as}\ x\to p

or

limxpf(x)=q \lim \limits_{x\to p}f(x)=q

is denoted and ff is said to have the limit qq at pp.

Explanation

Now that we have learned the epsilon-delta argument, we can rigorously define the limit of a function in a metric space as above. It is important to note that by this definition, pXp \in X need not be equal to pEp \in E. Hence, for some pEp \in E,

f(p)limxpf(x) f(p) \ne \lim \limits_{x\to p}f(x)

may hold.

Theorem

Assume XX, YY, EE, ff, pp are as described in the definition. Then, the following two propositions are equivalent.

(a)

limxpf(x)=q \lim \limits_{x\to p}f(x) = q

(b)

For all sequences EE of pnpp_{n}\ne p that are limnpn=p\lim \limits_{n\to\infty}p_{n}=p,

limnf(pn)=q \lim \limits_{n\to\infty}f(p_{n})=q


Explained, (b)(b) means ‘for all sequences {pn}\left\{ p_{n} \right\} converging to pp, the sequence {f(pn)}\left\{ f(p_{n}) \right\} converges to qq’.

Proof

(a) \Longrightarrow (b)

Assume (a) holds. Choose any sequence {pn}\left\{ p_{n} \right\} such that limnpn=p\lim \limits_{n\to\infty} p_{n}=p. Given any positive number ε>0\varepsilon >0, by assumption, there exists some positive number δ\delta such that,

xEanddX(x,p)<δ    dY(f(x),q)<ε \begin{equation} x \in E \quad \text{and} \quad d_{X}(x,p)<\delta \implies d_{Y}(f(x),q)<\varepsilon \label{eq1} \end{equation}

Also, since {pn}\left\{ p_{n} \right\} is a sequence converging to pp, there exists a positive number NN such that

nN    dX(pn,p)<δ \begin{equation} n\ge N \implies d_{X}(p_{n},p) <\delta \label{eq2} \end{equation}

Therefore, by (eq1)\eqref{eq1} and (eq2)\eqref{eq2}, the following holds.

nN    dX(pn,p)<δ    dY(f(pn),q)<ε n\ge N \implies d_{X}(p_{n},p)<\delta \implies d_{Y}(f(p_{n}),q)<\varepsilon

This condition signifies that the sequence {f(pn)}\left\{ f(p_{n}) \right\} converges to qq, thus

limnf(pn)=q \lim \limits_{n\to\infty} f(p_{n})=q

Before going into the proof, note that the following is slightly complicated. By contrapositive, if ‘(b) then (a)’ is true, it means that ‘if not (a) then not (b)’ is also true. So, we will show that ‘if limxpf(x)=q\lim \limits_{x\to p}f(x)=q does not hold, then for some sequence {pn}\left\{ p_{n} \right\} converging to pp, the sequence {f(pn)}\left\{ f(p_{n}) \right\} does not converge to qq’.

(a) \Longleftarrow (b)

Assume (a) does not hold. Then by definition, for some ε\varepsilon,

xEanddX(x,p)<δ    dY(f(x),q)ε \begin{equation} x\in E \quad \text{and} \quad d_{X}(x,p)<\delta \implies d_{Y}(f(x),q) \ge \varepsilon \label{eq3} \end{equation}

exists a positive number δ\delta. Consider any sequence {pn}\left\{ p_{n} \right\} converging to pp. Then,

nN    dX(pn,p)<δ \begin{equation} n\ge N \implies d_{X}(p_{n},p) <\delta \label{eq4} \end{equation}

exists sufficiently large positive number NN. Therefore, by (eq3)\eqref{eq3}, (eq4)\eqref{eq4}, for some ε\varepsilon,

nN    dX(pn,p)<δ    dY(f(pn),q)ε n \ge N \implies d_{X}(p_{n},p)<\delta \implies d_{Y}(f(p_{n}),q) \ge \varepsilon

does not hold, thus the sequence {f(pn)}\left\{ f(p_{n}) \right\} does not converge to qq.

Corollary

If ff has a limit at pp, that limit is unique.


This follows from properties of sequences converging in a metric space1 and the theorem above.


  1. See (b) for reference. ↩︎