Want to take part in these discussions? Sign in if you have an account, or apply for one below
Vanilla 1.1.10 is a product of Lussumo. More Information: Documentation, Community Support.
added to locale a section relation to toposes stating localic reflection
thanks, fixed.
Added to locale remakrs on how the frame of subobjects is the subcategory of -truncated objects and how this is the beginning of a pattern continued at n-localic (infinity,1)-topos.
Have added to locale in the section Category of locales two theorems from the Elephant on the externalization of internal locales (needed in the discussion of Bohrification)
At locale in the section relation to topological spaces I have tried to make some of the statements more pronounced.
I have made more explicit in the discussion of localic reflection at locale that is fully faithful in fact as a 2-functor.
The discussion of this point in the entry would deserve a bit of expansion. Maybe I find time to expand on it later.
I realized that we never stated the definition of the 2-category (or (1,2)-category) of locales. So I have now added it to the Definition-section at Locale.
Oh. dear, I should take a break :-) I did look for it, but didn’t see it! Thanks.
Anyway, I have now added to Locale, too, the precise definition, and some more pointers.
Added to locale the observation that frames are the same as lex total posets, by way of introducing connections between locales and toposes.
Nice, thanks! Should we have a page lex-total category with Street’s theorem on it?
The second paragraph of Locale says:
For example, there is a locale of all surjections from natural numbers (thought of as forming the discrete space ) to real numbers (forming the real line ): the locale of real numbers. This local has no points, since there are no such surjections, but it contains many nontrivial open subspaces; these open subspaces are generated by a family parametrised by and ; the basic open associated to and may be described as .
Can someone explain exactly what this means, or at least give a reference?
Capitalised ’Locale’ is redirected to Loc, the category. So you mean the lower case version locale.
Does the description at locale of real numbers help? It seems to be largely the work of Toby.
Yes I mean locale. The locale of real numbers doesn’t help at all. In any case I would guess that the particular sets (locales?) and are basically irrelevant for this example. (Except for having , which is supposed to imply there are no points.) I just want to know what definition is underlying this example.
Here’s my speculation, though it doesn’t seem to work right. Let and be sets, and let be the poset of “finite partial graphs” in , i.e., finite subsets such that the composite is injective. Let be the collection of upward-closed subsets of : this set is supposed to be the collection of “opens” in the locale.
Given any function let be the set of all such that is not the graph of any function: it’s the set of finite partial graphs inconsistent with . Note that since , and I can show that is non-trivially an intersection of non-maximal elements of if and only if is not surjective. That is, surjective functions give rise to points in the locale . What I can’t prove is that these are the only points in the locale.
Edit. Nevermind that doesn’t make sense anyway. But I expect they mean some kind of example along these lines.
The locale of surjections is the classifying locale of the propositional geometric theory of surjections . There are basic propositions “” (an atomic formula) for each and , and then axioms like and to make it a function and to make it surjective. So the opens of the locale are generated by these propositions “” under finite meet and arbitrary join, modulo those axioms. This can be re-expressed in terms of partial functions in a way kind of like you say, but I don’t remember (and don’t have time to work out) exactly how it goes. But with this description, the points of the locale are, by the universal property of classifying locales (or equivalently, by the universal property of a frame presented by generators and relations), precisely surjections .
Thanks Mike. I would still love a reference for this. It looks to me that what I was describing above is the locale of “partial graphs in ”; at least, the points are precisely the partial graphs. I assume the additional axioms are imposed by choosing appropriate sublocales.
It’s C1.2.8 in Sketches of an Elephant.
Thanks Mike.
Here’s a brief sketch of the construction (where I’ve unwound most of the terminology Johnstone uses here).
Fix discrete spaces and , and let with the usual product topology. For a finite subset write , the subset of functions which map onto . Let be the collection of open subsets such that
for all open and finite , we have that implies .
The claims are that:
is a complete lattice satisfying the infinite distributive law, i.e., it corresponds to a locale.
If is infinite then and , e.g., is a non-trivial locale when is infinite and is non-empty.
The points of correspond exactly to surjective functions .
In particular, if then is a non-trivial pointless locale. (Johnstone in the example assumes , but as far as I can tell the construction is entirely general.)
It is easy to see that is closed under pairwise intersections of open sets and contains , so has pairwise meets and a top element. To show it is a locale, you need to know that for every open set there is a smallest open set which contains and is an element of . The recipe for is to define for any open set ,
where the union ranges over all open and finite satisfying the condition. Then iterate some possibly transfinite number of times to get . Then is the bottom element and . The key observation is that preserves pairwise intersection and has exactly elements of as its fixed points, whence has these same properties.
More conceptually: is the limit in Locale of the family of subspaces of indexed by the poset of all finite ; compare with the limit in Top, which is just the subspace of surjective functions. (Johnntone says “intersection of sublocales” here, but I think it is also an example of a limit in Locale this context, as is the intersection subspace in Top.)
Unrelated remark: Apparently the TeX engine on this thing displays the same character for () and (), which is unfortunate since is so ugly.
Thanks for writing that out! It would be useful to add it to some nlab page, maybe locale?
I generally prefer \emptyset to \varnothing; I find the latter uglier than the former. But if you really want the latter you may be able to get it with unicode ∅ (∅).
I added a bit about the locale of surjections discussed in comments #13–#21, only at the level of detail of Mike #17 (actually less detail than that), not Charles #20.
Toby: no. This is not “direct image”, since direct image maps do not (generally) take opens to opens. Rather, we are taking an adjoint to a frame map , and since frame maps are left exact left adjoints, we want here its right adjoint, playing a role analogous to geometric morphisms between toposes.
(Rolled back to previous version.)
Sorry, I thought that it was a typo (well, something analogous to that, involving thinking about things backwards), but it was me who was thinking backwards! (left vs right adjoint).
Although a rollback is a dangerous tool, since I also made some other (more minor) edits. I just used that tool myself to restore my other edits; but if all that you did was to roll back my edit and add your remark (edit: and change ‘This category’ to ‘The category ’), then everything is OK now.
Sorry if I missed those other edits. Looks good now.
Added the following text:
As explained above, the functor from topological spaces to locales provides a very large collection of examples. When restricted to sober topological spaces, this functor becomes fully faihtful and its essential images consists of spatial locales.
If excluded middle fails, many classical examples, such as the locale of real numbers, can cease to be spatial.
Naturally, one is interested in examples of locales that are not spatial, and a few are given below.
An amazing feature of pointfree topology is that it contains not only point-set general topology, but also point-set measure theory, again as a full subcategory.
Specifically, recall from the duality between geometry and algebra that various categories of commutative algebras are contravariantly equivalent to certain corresponding categories of spaces. The category of algebras relevant for measure theory is the category of commutative von Neumann algebras and ultraweak *-homomorphisms; it is widely accepted that dropping the commutativity condition and passing to the opposite category yields the correct category of noncommutative measurable spaces.
The category of commutative von Neumann algebras is contravariantly equivalent to several other categories:
compact strictly localizable enhanced measurable spaces and measurable maps;
measurable locales and maps of locales.
The last category is particularly interesting: it is a full subcategory of locales, which means that measure theory embeds into pointfree topology, which means that methods and results from pointfree topology can be used right away in measure theory. This stands in contrast to the traditional point-set treatments, where two rather different formalisms must be developed for point-set topological spaces and for point-set measurable spaces.
With the exception of discrete locales, measurable locales are never spatial.
Another important source of nonspatial sublocales is given by intersections (of arbitrary cardinality) of dense sublocales.
Again, in contrast to the point-set setting, where the Baire category theorem identifies the rather restrictive conditions under which the intersection of dense topological subspaces is again dense, in the pointfree setting arbitrary intersections of dense sublocales are always dense.
In particular, one can intersection all dense sublocales of a given locale, which always produces a nonspatial locale, unless the original locale is discrete. This is the double negation sublocale.
But there are other interesting examples. Connecting to measure theory, we can consider a valuation on a given locale and take the intersection of all sublocales such that . The resulting sublocale can be seen as the smallest sublocale with a measure 0 complement.
Thanks!! That’s excellent.
(We should really have an entry noncommutative measure theory…)
Does Johnstone’s example here have anything to do with Andrej Bauer’s recent work on a topos where the reals are countable? I.e., should the fact that the classifying space for a surjection from N to R is non-trivial already be a proof that there is a non-trivial topos where the reals are countable? Though this is a geometric theory and Andrej’s is an elementary topos so I’m not sure what the relationship is.
More generally a geometric theory having a classifying locale with no points just means there’s no propositional models in Set, but it suggests that there are alternative toposes where it would have points. This seems to be a nice thing to mention (if I didn’t botch it).
should the fact that the classifying space for a surjection from N to R is non-trivial already be a proof that there is a non-trivial topos where the reals are countable?
I don’t think so. The “real numbers” in the base topos (e.g. Set) become countable in the topos of sheaves on that classifying locale, but they won’t be the same as the internally defined real numbers in that topos.
More generally a geometric theory having a classifying locale with no points just means there’s no propositional models in Set, but it suggests that there are alternative toposes where it would have points.
Yes, that’s right. (Although terminology-wise, being “propositional” is a property of a theory, not of its models.) In particular, the classifying topos of that theory itself contains a generic model of the theory, corresponding to the identity geometric morphism.
Added reference
Anonymous
That it’s a “bad choice” is your opinion. I’m not convinced “space” makes “more sense” than “locale”; one could argue that “space” is a poor choice because it is apt to be confused with an ordinary topological space.
In any case, the real point is that what “we” at the nLab call a frame is what everyone else besides Joyal and Tierney calls a frame, and what we call a locale is what everyone but J and T calls a locale. So we’re just sticking with the established convention and terminology – for better or worse. This makes it easier and generally less confusing for others to compare what we do with what other people do, save for some outliers (like J and T in this case).
(Full disclosure: as I was a Tierney student, I often slip and write “locale” when I mean “frame”, and then I’ll catch myself later.)
It is brilliant (and influential), but in some places it’s a bit terse and demands a persistent effort. Some time ago I saw here mention of an expository paper that goes through the Joyal-Tierney monograph more gently. There may be some nLab articles on this in the foreseeable future.
The word “space” is overloaded in mathematics. Alongside topological spaces and locales, there are also formal spaces. There is no reason why locales should take precedent over the others.
added pointer to the recent
establishing a model category of locales which makes the reflection of sober topological spaces a Quillen adjunction to the sober-restriction of the classical model structure on topological spaces.
1 to 44 of 44