© 2000 Copyright by Andrej Bauer. All rights reserved.
| When does the 0-dimensional Reflection Preserve Products? Andrej Bauer | |||||||||||
We say that a topological space is zero-dimensional when the collection of its clopen (closed and open) subsets forms a basis for its topology.
For a topological space X, let z(X) be its zero-dimensional reflection, i.e., the same underlying set X but with the topology generated by the collection of the clopen subsets of X.
Question 1: Does z preserve products, i.e., is z(X ×Y) homeomorphic to z(X) ×z(Y)?
The question can be stated equivalently as follows.
Question 1': Is every clopen subset of X ×Y a union of the sets of the form U ×V, where U 'subset or equal' X is clopen and V 'subset or equal' Y is clopen?
We know that the answer to this is negative. Raouchan Bouziakova constructed a couple of counter-examples. Alexander Arhangel'skii observed that the answer is affirmative if X is compact.
For the purposes of the original motivation, it would be useful to know the answer to the following question.
Question 2: Does Question 1 have an affirmative answer if X and Y are homogeneous, and are topological quotients of homogenous, countably based, T0-spaces?
The motivation for this question comes from theoretical computer science. Consider the category of equilogical spaces [1], originally defined by Dana Scott, which is a cartesian closed extension of the category of T0-spaces. Briefly, the objects are pairs (X, ~ ) where X is a T0-spaces and ~ is an equivalence relation on X. A morphism is (an equivalence class of) an equivalence preserving continuous maps, i.e., a map f\colon X --> Y such that a ~ X b implies f(a) ~ Y f(b). See the reference for details.
Every T0-space X can be viewed as an equilogical space equipped with the identity equivalence relation. Because the category of equilogical spaces is cartesian closed, we can iterate exponentials of the space of the real numbers with Euclidean topology, R:
|
|
|
|
Martín Escardó, Alex Simpson, and I checked that this coincidence extends to levels one and two, i.e., there is a natural bijective correspondence between RR and RIRI, and also between RRR and RIRIRI. We do not know the answer for higher types.
Alex Simpson and I observed the following: if the zero-dimensional reflection preserves products, then the hierarchies of Euclidean reals and intensional reals coincide.
For other background material on this topic see [3].