nLab
discrete fibration

Discrete fibrations

Definition

A functor F:CB is a discrete fibration if for every object c in C, and every morphism of the form g:bF(c) in B there is a unique morphism h:dc in C such that F(h)=g.

A functor F:CB is a discrete opfibration if F op:C opB op is a discrete fibration.

A discrete fibration is a special case of a Grothendieck fibration.

Given a cartesian category E, internal categories C,B in E, an internal functor F:CB is a discrete fibration of internal categories if the square

C 1 F 1 B 1 d 0 d 0 C 0 F 0 B 0\begin{matrix} C_1 &\stackrel{F_1}\to& B_1\\ d_0\downarrow && \downarrow d_0\\ C_0 &\stackrel{F_0}\to& B_0 \end{matrix}

is cartesian.

Discussion via category of elements

Given a discrete fibration F:CB, define a functor F *:B opSet as follows:

  • For x an object of B, let F *(x) be the set of objects y of C such that F(y)=x.
  • For g:xy a morphism of B, let F *(g):F *(y)F *(x) be the function that maps each element of F *(y) to the unique morphism h determined by the definition of discrete fibration above.

There is a size issue here, is F *(x) in fact small? We say that the fibration has small fibres if so; else we must pass to a larger universe when we define Set.

Conversely, give a functor F *:B opSet, define a category C and a discrete fibration F:CB as follows:

  • Let C be the category of elements of the functor F; that is:

    • an object of C is a pair consisting of an object x of B and an element of F *(x),
    • a morphism from (x,a) to (y,b) in C is a morphism g:xy in B such that F *(g) assigns a to b.
  • The functor from C to B is the obvious forgetful functor.

If you start from F *, construct C and F, and then construct a new F *, it will be equal to the original F *. Conversely, if you start with C and F, construct F *, and then construct a new C and F, then there will be an isomorphism of categories between C and C, relative to which F and F are equal.

Invariance under equivalence

Note that the definition of fibration refers to equality of morphisms without previously assuming that the sources match, while the construction of F * from F refers to equality of objects. This is also why we get equality of functors and isomorphism of categories in the immediately preceding paragraph. So the only non-evil thing on this page is the idea of a functor to Set. That is the fundamental invariant notion; a discrete fibration is just a convenient way of talking about it.

Generalization for spans internal to a category

Let E be a cartesian category. A span of internal categories ApCqB in Cat(E) is called a discrete fibration from A to B if in the diagram

A 0 C l i l A p C i r C r q B B 0\begin{matrix} A_0 & \leftarrow & C_l && \\ \downarrow &&\downarrow i_l &&\\ A &\stackrel{p}\leftarrow & C &\stackrel{i_r}\leftarrow & C_r\\ &&q \downarrow && \downarrow \\ && B &\leftarrow & B_0 \end{matrix}

in which the two squares are the cartesian satisfies the following 3 properties:

  • pi r:C 1A is a discrete fibration

  • qi l:C lB is a discrete opfibration

  • Let X be defined as the pullback

    X (C r) 1 (C l) 1 C 0\begin{matrix} X & \to & (C_r)_1 \\ \downarrow &&\downarrow \\ (C_l)_1 &\to & C_0 \end{matrix}

    and j:XC 1× C 0C 1 the canonical inclusion. Then the morphism cj:XC 1, where c:C 1× C 0C 1C 1 is the composition morphism of internal category C, is invertible.

Example. Given internal functors a:AD and b:BD in E, the obvious span AabB is a discrete fibration from A to B.

Revised on April 27, 2013 00:55:16 by David Carchedi (88.153.46.18)