Totally bounded iff every sequence has a Cauchy subsequence

A metric set admits finite ε-nets for all ε>0 exactly when sequences have Cauchy subsequences
Totally bounded iff every sequence has a Cauchy subsequence

Let (X,d)(X,d) be a and let EXE\subseteq X.

The set EE is if for every ε>0\varepsilon>0 there exist points x1,,xNXx_1,\dots,x_N\in X such that Ej=1NHAHAHUGOSHORTCODE818s2HBHB(xj,ε). E\subseteq \bigcup_{j=1}^N (x_j,\varepsilon).

Proposition: EE is totally bounded if and only if every sequence in EE has a .

This proposition is the “sequential form” of total boundedness and is the key step in proving that .

Proof sketch: (\Rightarrow) Let (xn)(x_n) be a sequence in EE. Cover EE by finitely many balls of radius 11, so one ball contains infinitely many terms; pick a subsequence in that ball. Then cover EE by finitely many balls of radius 1/21/2 and refine to a further subsequence inside one of those balls. Continue with radii 1/2k1/2^k. The diagonal subsequence (xnk)(x_{n_k}) satisfies d(xnk,xn)<2k+1d(x_{n_k},x_{n_\ell})<2^{-k+1} for all k\ell\ge k, hence is Cauchy.

(\Leftarrow) If EE is not totally bounded, there exists ε>0\varepsilon>0 such that no finite collection of ε\varepsilon-balls covers EE. Construct a sequence (xn)(x_n) inductively by choosing xn+1Ej=1nB(xj,ε)x_{n+1}\in E\setminus\bigcup_{j=1}^n B(x_j,\varepsilon). Then d(xn,xm)εd(x_n,x_m)\ge \varepsilon for nmn\neq m, so no subsequence can be Cauchy.