Let (X,d)(X,d) be a , and define

τd={UX:for every xU there is r>0 with Bd(x,r)U},\tau_d=\{U\subseteq X:\text{for every }x\in U\text{ there is }r>0\text{ with }B_d(x,r)\subseteq U\},

where Bd(x,r)B_d(x,r) is an . Then:

  1. ,Xτd\varnothing,X\in\tau_d.
  2. Any union of members of τd\tau_d belongs to τd\tau_d.
  3. Any finite intersection of members of τd\tau_d belongs to τd\tau_d.

Thus τd\tau_d is a on XX.

Proof

The empty set and XX lie in τd\tau_d. An arbitrary union of members lies in τd\tau_d, since each point belongs to one of them and has a ball inside it. For a finite intersection, take the minimum of the finitely many positive radii supplied at a point. The empty intersection is XX.

Resulting structure

This is the . The proof starts with the metric ball condition, rather than assuming an existing topology and its open sets.