skeletal geometric morphism



Topos Theory

topos theory



Internal Logic

Topos morphisms

Extra stuff, structure, properties

Cohomology and homotopy

In higher category theory




Skeletal geometric morphisms are those geometric morphisms that preserve double negation sheaves and therefore play a role in the descriptions of classes of toposes like e.g. Boolean or De Morgan toposes in whose definitions the negation participates.

The notion of a skeletal geometric morphism can be viewed as a weakening of the notion of open geometric morphism.


A geometric morphism f:f : \mathcal{F} \to \mathcal{E} is called skeletal if the following equivalent conditions hold:

  • ff restricts to a geometric morphism Sh ¬¬()Sh ¬¬()Sh_{\neg\neg}(\mathcal{F}) \to Sh_{\neg\neg}(\mathcal{E}) .

  • The inverse image f *f^\ast preserves ¬¬\neg\neg-dense monomorphisms.

  • For any subobject AAA'\rightarrowtail A in \mathcal{E}: ¬¬f *(A)=¬f *(¬A)\neg\neg f^\ast(A')=\neg f^\ast(\neg A') in Sub (f *(A))Sub_\mathcal{F}(f^\ast(A)).


  • Inverse images of open geometric morphisms are Heyting functors, hence commute with ¬\neg and, therefore, open geometric morphisms are skeletal. In particular, geometric morphisms with Boolean codomain are open (Johnstone 2002, p.612), hence skeletal.

  • Dense subtoposesi:Sh j()i:Sh_j(\mathcal{E})\hookrightarrow \mathcal{E} are precisely those subtoposes with Sh ¬¬(Sh j())=Sh ¬¬()Sh_{\neg\neg}(Sh_j(\mathcal{E}))=Sh_{\neg\neg}(\mathcal{E}) (cf. this proposition) and, therefore, are skeletal.


The following two propositions concern skeletal inclusions (cf. Johnstone (2002, p.1007)):


An inclusion i:Sh j()i:Sh_j(\mathcal{E})\hookrightarrow \mathcal{E} is skeletal iff ext(j)ext(j) , the jj-closure of 010\rightarrowtail 1 , is a ¬¬\neg\neg-closed subterminal object of \mathcal{E}.

Proof: First, notice that in general for a topology jj a subterminal UU is jj-closed iff it is a jj-sheaf since 11 is always a jj-sheaf. Hence 00 is ¬¬\neg\neg-closed precisely when it is a ¬¬\neg\neg-sheaf.

Now assume ii skeletal. Since ext(j)ext(j) is a ¬¬\neg\neg-sheaf in Sh j()Sh_j(\mathcal{E}) and ii preserves them, it is also a ¬¬\neg\neg-sheaf in \mathcal{E}.

Conversely, assume ext(j)ext(j) is a ¬¬\neg\neg-sheaf in \mathcal{E}. Since it is also a jj-sheaf, it is contained in Sh j()Sh ¬¬()Sh_j(\mathcal{E})\cap Sh_{\neg\neg}(\mathcal{E}) but this coincides with Sh ¬¬(Sh j())Sh_{\neg\neg}(Sh_j(\mathcal{E})) because as a subtopos of the Boolean Sh ¬¬()Sh_{\neg\neg}(\mathcal{E}) the intersection Sh j()Sh ¬¬()Sh_j(\mathcal{E})\cap Sh_{\neg\neg}(\mathcal{E}) is Boolean and Sh j()Sh ¬¬()Sh_j(\mathcal{E})\cap Sh_{\neg\neg}(\mathcal{E}), since it contains ext(j)ext(j), is dense in Sh j()Sh_j(\mathcal{E}) and there can be only one such dense Boolean subtopos. \qed


The class Σ\Sigma of skeletal inclusions is the smallest class Γ\Gamma of geometric morphisms such that:

  • Γ\Gamma contains open inclusions and,

  • Γ\Gamma is closed under precomposition with dense inclusions: from gg dense, fΓf\in\Gamma and f,gf,g composable, follows fgΓfg\in\Gamma. \qed

The following exhibits the link between skeletal morphisms and Booleanness:


A topos \mathcal{E} is Boolean iff all geometric morphisms \mathcal{F}\to\mathcal{E} to \mathcal{E} are skeletal.

Proof: When \mathcal{E} is Boolean it coincides with Sh ¬¬()Sh_{\neg\neg}(\mathcal{E}) hence ¬¬\neg\neg-sheaves of \mathcal{F} trivially have to land there.

Conversely, assume all \mathcal{F}\to\mathcal{E} are skeletal. By Barr's theorem, \mathcal{E} receives a surjective f:f:\mathcal{B}\to\mathcal{E} from a Boolean topos. ff being skeletal and surjective implies that im(f)=im(f)=\mathcal{E} is Boolean. \qed

A pullback characterisation of open geometric morphisms from Johnstone (2006, cor. 4.9):


A geometric morphism f:f:\mathcal{F}\to\mathcal{E} is open iff the pullback of any bounded geometric morphism with codomain \mathcal{E} is skeletal. \qed


Skeletal morphisms between frames are studied in Banaschewski-Pultr (1994,1996), called weakly open there.

The equivalent concept for topological spaces appears in Mioduszewski-Rudolf (1969).

It is possible to define an analogous concept of m-skeletal geometric morphism using the De Morgan topology on a topos \mathcal{E} instead of ¬¬\neg\neg.


  • B. Banaschewski, A. Pultr, Variants of openness , Appl. Cat. Struc. 2 (1994) pp.1-21.

  • B. Banaschewski, A. Pultr, Booleanization , Cah. Top. Géom. Diff. Cat. XXXVII no.1 (1996) pp.41-60. (numdam)

  • Peter Johnstone, Factorization theorems for geometric morphisms II , pp.216-233 in LNM 915 Springer Heidelberg 1982.

  • Peter Johnstone, Sketches of an Elephant vol.II , Oxford UP 2002. (section D4.6, pp.1006-1010)

  • Peter Johnstone, Complemented sublocales and open maps , Annals of Pure and Applied Logic 137 (2006) pp.240–255.

  • Peter Johnstone, The Gleason Cover of a Realizability Topos , TAC 28 no.32 (2013) pp.1139-1152. (abstract)

  • J. Mioduszewski, L. Rudolf, H-closed and extremally disconnected Hausdorff spaces , Dissertationes Math. 66 1969. (toc)

Last revised on August 24, 2015 at 05:45:23. See the history of this page for a list of all contributions to it.