The classical definition is:
A topological space is completely regular if, given a point of and a closed subset of , if , then there is a continuous map from to the real line such that is constant when restricted to but takes a different value at .
(Traditionally, one requires and , but this can always be forced by postcomposition with an affine linear map. One sometimes sees, even more specifically, the requirement that land in the unit interval , but this too can be forced since is a retract of .)
An equivalent definition focussing on open sets is this:
(Here is the preimage under of the complement of the singleton of . Again, one can force , , and even if desired.) This definition is suitable for constructive mathematics based on weak countable choice, Markov's principle, and the fan theorem (all of which follow from excluded middle), or in any case if the following interpretations are made:
There is also a definition entirely in terms of the lattice of open sets, suitable for locales, which I need to look up again. (It's similar to the localic definition of regular space, but using a stronger notion than well-inside.)
When a space is completely regular, we also often ask it to be :
Given any two points, if each neighbourhood of either is a neighbourhood of the other, then they are equal.
A completely regular -space is variously called a -space (or -space), a completely regular Hausdorff space (since it is a theorem that such a space is Hausdorff), or a Tychonoff space (or using other spellings of Tychonoff?'s name).
As is usual with the separation axioms, some authors while conflate the meanings of ‘completely regular’ and ‘’ either way, or even reverse them. In contrast, the terms ‘completely regular Hausdorff’ and ‘Tychonoff’ are unambiguous. The only unambiguous term for a completely regular space in general seems to be ‘’, but this is not widely recognised.
Every completely regular space is preregular, so every completely regular Hausdorff space (as defined above) really is Hausdorff.
Every completely regular space is regular.
The natural map from a Tychonoff space to its Stone-Cech compactification (the unit of the adjunction) really is a compactification (the inclusion of a dense subspace into a compact space). Conversely, any subspace of a compact Hausdorff space must be Tychonoff.