nLab
gauge space

Gauge spaces

Idea

A gauge space is a topological space (necessary completely regular) whose topology is given by a family of pseudometrics. More generally, a quasigauge space is a space (not necessarily completely regular) whose topology is given by a family of quasipseudometrics.

Actually, a gauge space has additional structure, so that it can be seen as giving a (completely regular) Cauchy space, a uniform space, or even a generalisation of a metric space in which the category MetMet of metric spaces and short maps is a full subcategory.

Please note that, while this is based on the presentation in HAF, the precise definitions of the objects and morphisms of the category of gauge spaces below constitute original research. (In particular, HAF really considers the category of pregauge spaces and uniformly continuous maps, which is equivalent to the category of uniform spaces, since it uses them only to study that category.)

Definitions

Given a set XX, a pregauge on XX is simply a family of pseudometrics on XX. A gauge is a \geq-filter of pseudometrics on XX, that is a collection GG of gauging distances such that

  1. There is an element of GG; in the light of (3), the zero pseudometric (x,y0)( x, y \mapsto 0 ) is a gauging distance.
  2. Given d,eGd, e \in G, some fGf \in G satisfies
    d(x,y),e(x,y)f(x,y) d(x,y), e(x,y) \leq f(x,y)

    for all x,yx, y in XX; in the light of (3), the pseudometric (x,ymax(d(x,y),e(x,y)))( x, y \mapsto max(d(x,y), e(x,y)) ) is a gauging distance.

  3. Given dGd \in G and any pseudometric ee on XX, if
    e(x,y)d(x,y) e(x,y) \leq d(x,y)

    for all x,yx, y in XX, then eGe \in G.

A pregauge satisfying axioms (1&2) is a base for a gauge; a base is precisely what generates a gauge by taking the downward closure. Any pregauge whatsoever is a subbase for a gauge; a subbase is precisely what generates a base by closing under finitary joins.

A gauge space is a set equipped with a gauge. A quasigauge is a collection of quasipseudometrics satisfying (1–3); a quasigauge space is a set equipped with a quasigauge.

Given (quasi)gauge spaces XX and XX', a short map from XX to XX' is a function FF (on their underlying sets) such that the composite with FF (or with F×FF \times F, to be precise) of any gauging distance on XX' is a gauging distance on XX. That is,

  • for every eGe \in G', there is a dGd \in G such that
    e(F(x),F(y))d(x,y) e(F(x),F(y)) \leq d(x,y)

    for all x,yx, y in XX; in the light of (3), the pseudometric (x,ye(F(x),F(y)))( x, y \mapsto e(F(x),F(y)) ) (this is the composite) is a gauging distance.

(Warning: this definition of short map is probably the most significant original research on this page.)

Gauge spaces and short maps between them form the category GauGau of gauge spaces; quasigauge spaces and short maps between them form the category QGauQGau of quasigauge spaces. Note that any gauge is a base for a quasigauge; in this way, GauGau is (equivalent to) a full subcategory of QGauQGau.

Examples

The categories GauGau and QGauQGau are not well known, but some of their subcategories are.

  • A metric space (or pseudometric space) defines a gauge space, taking its single (pseudo)metric as a base; similarly, a quasi(psuedo)metric space defines a quasigauge space. Taking short maps (in the usual sense, that is distance-nonincreasing functions) as morphisms, the categories MetMet and PsMetPsMet are full subcategories of GauGau; similarly, QMetQMet and QPsMetQPsMet are full subcategories of QGauQGau.

  • A uniform space XX defines a gauge space, consisting of all of the pseudometrics on XX that are uniformly continuous as maps from X×XX \times X to the real line; similarly, a quasiuniform space defines a quasigauge space. Taking uniformly continuous maps as morphisms, the category UnifUnif is a full subcategory of GauGau; similarly, QUnifQUnif is a full subcategory of QGauQGau. Thus, uniform spaces can be viewed as gauge spaces whose collection of gauging distances is “saturated” under uniform continuity.

  • A completely regular topological space XX defines a gauge space, consisting of all the pseudometrics on XX that are continuous as maps from X×XX \times X to the real line. In this way the category CRegTopCReg Top of completely regular spaces and continuous maps is a full subcategory of GauGau; it is in fact contained in UnifUnif. (In general, a completely regular space can be uniformized in many ways; this inclusion corresponds to the “initial” uniformity.)

  • An arbitrary topological space defines a quasigauge space in a more complicated way. Given any open set UU in a space XX, define the pseudometric (in fact a pseudoultrametric) d Ud_U as follows:

    d U(x,y)={0 ifxUyU, 1 ifxUyU; d_U(x,y) = \begin{cases} 0 & if\; x \in U \;\Rightarrow\; y \in U ,\\ 1 & if\; x \in U \;\wedge\; y \notin U ;\end{cases}

    then the d Ud_U form a base for a quasigauge, which induces the original topology on XX. In other words, every space is “quasigaugeable.” In this way TopTop also becomes a full subcategory of QGauQGau. Note, though that replacing 11 by any other positive real number gives a different embedding of TopTop in QGauQGau.

  • Another way to get a quasigauge space from a topological space is to take as a base the set of all quasi-pseudometrics dd on XX such that for each xXx\in X, the function d(x,):Xd(x,-):X\to \mathbb{R} is upper semicontinuous. This is also a full embedding of TopTop in QGauQGau, which is perhaps more “canonical.”

Mike: Back atcha… do you have any good examples of gauge spaces that are not of one of these types? And is there any value in embedding metric spaces, uniform spaces, and topological spaces into this mysterious larger category QGauQGau, in ways so that their images are essentially disjoint? Do we ever, for instance, want to talk about a short map from a metric space to a topological space, or vice versa? I would like the answer to be “yes,” but I haven’t managed to make it come out that way myself yet.

Toby: I don't know any examples of gauge spaces that arise naturally (you can make them artificially, of course, say as disjoint unions) and don't correspond to some other more familiar type of space. However, I can give examples that don't belong to MetMet, UnifUnif, or TopTop … because they're Cauchy spaces, which aren't listed above yet!

Incidentally, I didn't list Cauchy spaces yet (and didn't finish describing general topological spaces), since I haven't checked yet that things behave correctly. (In particular, I haven't checked that TopTop becomes a full subcategory of QGauQGau —did you?—, although I hope that it will.)

I think that it's nice to be able to see these all as special cases of one kind of thing; then the usual ways of finding the ‘underlying’ foo of a bar become reflections (and sometimes coreflections). This is basically how Lowen-Colebunder's book (which I've referenced, for example, on convergence space) works (although her big category of everything is the category of mereological spaces, which includes ConvConv, rather than QGauQGau).

Seeing this as inclusions and (co)reflections prevents any expectation that the diagrams above should commute, since they mix different things in the wrong way. On the other hand, it also shows us that (for example) any topological space has an underlying uniform space … if you want it.

Mike: I believe I did check that TopQGauTop\to QGau is full, but you should verify it.

I removed my comments about triangles commuting; go ahead and write about the reflections and I’ll see whether I still want to make that comment. I can see thinking of uniform spaces as “saturated” gauge spaces, so that the uniform space underlying a metric space is its “saturation” (reflection) in the larger category GauGau. Perhaps Cauchy spaces can also be thought of this way. I will be kind of surprised if TopTop turns out to be reflective in QGauQGau, but if it were that would also be pretty neat.

Mike: I added a new embedding of TopTop in QGauQGau, which seems to me more likely to be (co)reflective.

Toby: Ah yes, I think that that's the good one that I wasn't finding.

Reflections

Many of these full subcategories of GauGau and QGauQGau are reflective.

Details to come

References

Revised on October 10, 2012 13:55:19 by Toby Bartels (69.171.187.18)