1 / 12

Locale pullback via dcpos

Locale pullback via dcpos. Dr Christopher Townsend (Open University). Main Idea. THESIS: When changing base it is only really the directed joins that need to be modelled/worried about. All the rest of the finitary data takes care of itself.

rance
Download Presentation

Locale pullback via dcpos

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Locale pullback via dcpos Dr Christopher Townsend (Open University)

  2. Main Idea • THESIS: When changing base it is only really the directed joins that need to be modelled/worried about. All the rest of the finitary data takes care of itself. • Study locales only, i.e. frames. I.e. the data is finite meets and arbitrary joins. Equivalently: finite meets, finite joins and DIRECTED joins. • The definition of geometric morphism (suggests atleast) that the finitary structure is preserved.

  3. Technical Aims • Given f:E -> E’, a geometric morphism. Then the direct image functor preserves dcpo structure. I.e. f*:dcpoE -> dcpoE’ is well defined. • Further f*:dcpoE -> dcpoE’ has a left adjoint f#. • In any topos Dlat(dcpo)=Frames. Where we look at order-internal distributive lattices in the order enriched category dcpo. • The left adjoint f# restricts to a functor Dlat(dcpoE’)-> Dlat(dcpoE) left adjoint to f*:FrE -> FrE’ That is: locale pullback.

  4. What is known already? • This trick has been done by Joyal and Tierney already with suplattices: - f# Joyal and Tierney ‘84 dcpoE dcpoE’ NEW! f* f# supE supE’ Frames as ring objects in sup f* Frames as order-internal dlats in dcpo f# FrE FrE’ f*

  5. Is f*:dcpoE -> dcpoE’ well defined? x fiber directed iff x-1(j) directed for all j. (I,J posets.) As with sup lattices: - • YES. Use external definition of dcpo. For every fiber directed x:I->J, the map x*:PosE (J,A)-> PosE(I,A) has a left adjoint (and Beck-Chevalley holds). There exists V:IdlA->A left adjoint to A->IdlA Internal Definition of dcpo External Definition • Then unravel the adjunction of the geometric morphism with the external definition to prove f* is well defined. This works as fiber directedness is stable under the inverse image. (Known?)

  6. Defining f#:dcpoE’ -> dcpoE TRICK: Use presentations. For every dcpo A, there exists posets G and R and dcpo maps e1 and e2 such that: e1 is a coequalizer A Idl(R) Idl(G) e2 Dcpo coequalizer well defined? Folklore, or adapt Johnstone & Vickers 91 f# f*e1 f# A Idl(f* R) Idl(f* G) defines f# A. f*e2 Note: e1and e2 are suitably geometric and so f*e1 and f*e1 well defined

  7. Frames as Order-Internal DLats DEFINITION: For any order enriched category C (with lax products), an object X is an order-internal meet semilattice iff !:X->1 and Δ:X>->XxX have right adjoints. … in other words, iff finite complete wrt to the order enrichment • Define order-internal distributive lattice in the standard way from this. Then: - THEOREM: Fr=Dlat(dcpo) Proof:A in Dlat(dcpo), then  A a dcpo A directed cocomplete A a DLat /\:AxA->A a dcpo hom. A finite joins/meets + finite dist. dcpo Finite meet distr. over directed joins since AxA=A(x)A in dcpo

  8. Defining f#:FrE’ -> FrE • (Certainly f*:FrE -> FrE’ since the direct image preserves finite products) • So to prove that f#:dcpoE’ -> dcpoE restricts to f#:FrE’ -> FrE it must be verified that • (a) f#(A)x f#(A)= f#(AxA) and • (b) f# preserves the order enrichment. • PROOF: (b) is immediate from construction since f* certainly preserves the order enrichment. • (a) follows since product in dcpo is tensor and so is a coequalizer construction preserved by left adjoint.  Detailed construction available

  9. Applications (Locale Pullback) If f:X->Y is locale map then f*:FrSX -> FrSY is equivalent to Σf:Loc/X->Loc/Y (by the [JT84] equivalence LocSX=Loc/X). A left adjoint to f* is therefore right adjoint to Σf i.e. pullback. (Triquotient Assignments) If p:Z->Y is a locale map then a triquotient assignment for p is a dcpo map p#:ΩZ-> ΩY, satisfying a mixed Frobenius/coFrobenius condition with Ωp: p#[c/\(d\/Ωp(e)]=(p#c/\e)\/p#(c/\d). Using the dcpo description of pullback it can be shown that maps with triquotient assignment are pullback stable. This implies the well known pullback stability results for both proper and open maps.

  10. Applications: Double Power Locale (Double Power Locale) If Z is a locale map then the double power locale on Z, denoted PZ is define by Ω PZ=Fr< Ω Z qua dcpo>. The points of PZ at stage Y (i.e. the locale maps Y-> PZ) are therefore exactly dcpo maps ΩZ-> ΩY. Using the dcpo description of change of base it can be shown (Townsend/Vickers 03) that dcpo(ΩZ, ΩY)=Nat(Loc(_xZ,$),Loc(_xY,$)) where Loc(_xZ,$):Locop->Set is the presheaf, Nat(_) the collection of natural transformations and $ the Sierpiński locale. This gives a universal description of the double power locale.

  11. Further Work • Beck-Chevalley for f#:dcpoE’ -> dcpoE? • (Following on from pullback stability of triquotient assignments): Decent for Triquotient Surjections? • Topos theoretic version. (With toposes taking the place of locales.) Using pretopos sites to describe filtered cocontinuous maps between toposes.

  12. Summary • An external description of dcpos is available showing the dcpo structure is preserved by the direct image of a geometric morphism. • Since dcpo presentations are models of geometric theories they are preserved by the left adjoints of geometric morphisms. This defines a left adjoint to the direct image functor on dcpos. • Frames are exactly order-internal distributive lattices in the category of dcpos since dcpo tensor is given by set-theoretic product. • The left adjoint f#:dcpoE’ -> dcpoE preserves order and tensor and so preserves order internal distributive lattices. It therefore defines a left adjoint to f*:FrE -> FrE’ which must be equivalent to locale pullback. • Well known pullback stability results (open/proper maps) can be reproved by a single appeal to the pullback stability of maps with triquotient assignment.

More Related