Normal space

Jump to: navigation, search
Topological spaces in
separation axiom
Kolmogorov (T0) version
T0 | T1 | T2 | T | completely T2
T3 | T | T4 | T5 | T6

In topology and related branches of mathematics, normal spaces, T4 spaces, T5 spaces, and T6 spaces are particularly nice kinds of topological spaces. These conditions are examples of separation axioms.


Suppose that X is a topological space. X is a normal space if and only if, given any disjoint closed sets E and F, there are neighbourhoods U of E and V of F that are also disjoint. In fancier terms, this condition says that E and F can be separated by neighbourhoods.

The closed sets E and F, here represented by closed disks on opposite sides of the picture, are separated by their respective neighbourhoods U and V, here represented by larger, but still disjoint, open disks.

X is a T4 space, if it is both normal and Hausdorff.

X is a completely normal space or a hereditarily normal space if every subspace of X is normal. It turns out that X is completely normal if and only if every two separated sets can be separated by neighbourhoods.

X is a T5 space, or completely T4 space, if it is both completely normal and Hausdorff, or equivalently, if every subspace of X is T4.

X is a perfectly normal space if every two disjoint closed sets can be precisely separated by a function. That is, given disjoint closed sets E and F, there is a continuous function f from X to the real line R such the preimages of {0} and {1} under f are E and F respectively. The real line can be replaced with the unit interval [0,1] in this definition; the result is the same. It turns out that X is perfectly normal if and only if X is normal and every closed set is a Gδ set. Equivalently, X is perfectly normal if and only if every closed set is a zero set. Every perfectly normal space is automatically completely normal[citation needed].

X is a T6 space, or perfectly T4 space if it is both perfectly normal and Hausdorff.

Note that some mathematical literature uses different definitions for the terms "normal" and "T4", and the terms containing those words. The definitions that we have given here are the ones usually used today, and the ones used in Wikipedia. However, some authors switch the meanings of the two terms in a given pair, or use both terms synonymously for only one condition, and one should take care to find out which definitions the author is using when reading mathematical literature. (But "T5" always means the same as "completely T4", whatever that may be.) For more on this issue, see History of the separation axioms.

Terms like normal regular space and normal Hausdorff space also turn up in the literature; these simply mean that the space both is normal and satisfies the other condition mentioned. In particular, a normal Hausdorff space is the same thing as a T4 space. These phrases are useful, since they are less ambiguous given the historical confusion of the terms' meanings. In this encyclopedia, we prefer these phrases when applicable; that is, "normal Hausdorff" instead of "T4", or "completely normal Hausdorff" instead of "T5".

Fully normal spaces and fully T4 spaces are discussed elsewhere; they are related to paracompactness.

A locally normal space is a topological space where every point has an open neighbourhood that is normal. Every normal space is locally normal, but the converse is not true. A classical example of a completely regular locally normal space that is not normal is the Niemitzki plane.

Examples of normal spaces

Most spaces encountered in mathematical analysis are normal Hausdorff spaces, or at least normal regular spaces:

  • All metric spaces (and hence all metrizable spaces) are perfectly normal Hausdorff;
  • All pseudometric spaces (and hence all pseudometrisable spaces) are perfectly normal regular, although not in general Hausdorff;
  • All compact Hausdorff spaces are normal;
  • In particular, the Stone-Cech compactification of a Tychonoff space is normal Hausdorff;
  • Generalizing the above examples, all paracompact Hausdorff spaces are normal, and all paracompact regular spaces are normal;
  • All paracompact topological manifolds are perfectly normal Hausdorff. However, there exist non-paracompact manifolds which are not even normal.
  • All order topologies on totally ordered sets are hereditarily normal and Hausdorff.
  • Every regular second-countable space is completely normal, and every regular Lindelöf space is normal.

Also, all fully normal spaces are normal (even if not regular). Sierpinski space is an example of a normal space that is not regular.

Examples of non-normal spaces

An important example of a non-normal topology is given by the Zariski topology on an algebraic variety or on the spectrum of a ring, which is used in algebraic geometry.

A non-normal space of some relevance to analysis is the topological vector space of all functions from the real line R to itself, with the topology of pointwise convergence. More generally, a theorem of A. H. Stone states that the product of uncountably many non-compact Hausdorff spaces is never normal.


The main significance of normal spaces lies in the fact that they admit "enough" continuous real-valued functions, as expressed by the following theorems valid for any normal space X.

Urysohn's lemma: If A and B are two disjoint closed subsets of X, then there exists a continuous function f from X to the real line R such that f(x) = 0 for all x in A and f(x) = 1 for all x in B. In fact, we can take the values of f to be entirely within the unit interval [0,1]. (In fancier terms, disjoint closed sets are not only separated by neighbourhoods, but also function.)

More generally, the Tietze extension theorem: If A is a closed subset of X and f is a continuous function from A to R, then there exists a continuous function F: XR which extends f in the sense that F(x) = f(x) for all x in A.

If U is a locally finite open cover of a normal space X, then there is a partition of unity precisely subordinate to U. (This shows the relationship of normal spaces to paracompactness.)

In fact, any space that satisfies any one of these conditions must be normal.

A product of normal spaces is not necessarily normal. This fact was first proved by Robert Sorgenfrey. An example of this phenomenon is the Sorgenfrey plane. Also, a subset of a normal space need not be normal (i.e. not every normal Hausdorff space is a completely normal Hausdorff space), since every Tychonoff space is a subset of its Stone-Cech compactification (which is normal Hausdorff). A more explicit example is the Tychonoff plank.

Relationships to other separation axioms

If a normal space is R0, then it is in fact completely regular. Thus, anything from "normal R0" to "normal completely regular" is the same as what we normally call normal regular. Taking Kolmogorov quotients, we see that all normal T1 spaces are Tychonoff. These are what we normally call normal Hausdorff spaces.

Counterexamples to some variations on these statements can be found in the lists above. Specifically, Sierpinski space is normal but not regular, while the space of functions from R to itself is Tychonoff but not normal.


  • Kemoto, Nobuyuki (2004). "Higher Separation Axioms". in K.P. Hart, J. Nagata, and J.E. Vaughan. Encyclopedia of General Topology. Amsterdam: Elsevier Science. ISBN 0-444-50355-2. 
  • Willard, Stephen (1970). General Topology. Reading, Massachusetts: Addison-Wesley. ISBN 0-486-43479-6 (Dover edition).