Processing math: 100%

Presheaf . topology

Collection

context X,T … topological space
context C … category
definiendum COp(X)op

Discussion

Motivation via fibre bundles

Let X,T be a topological space. A fibre bundle

FibEX

is given by a projection map p:EX together with, for each open set UT, the possibility to locally identify E as U×Fib (pic related). With this fibre point of view, we want to speak of particular functions XE, namely the sections σ which satisfy pσ=idX.

In the simplest case, when globally E=X×Fib, we can define a section σ:XE of the bundle via σ(x):=x,s(x) and here we can simply take p:=π1. Clearly, σ is determined by a function s:XFib into the fibre.

If the the bundle has non-trivial topology and the fibres over X are twisted, then there can't be a global function s:XFib determining the section σ. However, each section is still fully determined by a collection of local such functions s|U:UFib. Note that the function space UFib is the object image of U under the contravariant hom-functor HomSet(,Fib) to Set.

A sheaf F is a particular kind of contravariant functor which helps to capture sections. The object image FU of a sheaf of a an open set U is a local function space.

Examples

  • Consider the sheaf C of all smooth functions over X. If U is an open set of X, then the objects CU:={f:UR | fsmooth} is the restriction of such functions to U.
  • Any continuous function. Take X=C{0}, E=C, and p=exp. Define a the sheaf Γ(E/X) via Γ(E/X)U:={s:UE | ps=idU}. Then Γ(E/X)U is the set of branches of the logarithm on U.

Sheaf theory is used for discussing the idea of analytical continuation and applying it too subjects which don't necessarily have to do with complex numbers. The image of the functor can contain other objects than sections (e.g. the spaces may be vector bundles as in K-theory, or spaces of differential forms, etc.) and the category itself doesn't have to be a classical topological space (see Grothendieck topology).

Notation

For the purpose of defining seperated presheaves of sections, we use the following notation:

If iOp(X)op[V,U] is an inclusion i:VU of a smaller open set VU in U, then it's fmap image F(i):C[FU,FV] should be understood as restriction of function domain. So if sFU, we write s|V for F(i)(s).

Reference

Wikipedia: Sheaf

Parents

Requirements

Link to graph
Log In
Improvements of the human condition