Understanding the Formal Definition of Sheaf

121 Views Asked by At

A few questions based on the definitions at the end:

  1. The meaning of $a \rightsquigarrow b$ (#7 in abstract simplical complex definition). Not sure what "that is compatible with the inclusions of faces" means.
  2. If the sheaf vector spaces can be different types among the faces. Or they all have to be the same type (#1 in first sheaf definition).
  3. If an abstract simplical complex is an open cover or related. The second sheaf definition says a sheaf is on an open cover, while the first says it's on an abstract simplical complex.
  4. What the inverse limit part is saying (the second sheaf definition at the bottom): $$F(Q)\to\varprojlim_{I\in N(Q)}F(U_I) =: F[Q]$$

The definitions of sheaf are from here.

An abstract simplical complex $X$ is:

  1. Of a set $A$
  2. A collection $X$ of subsets of $A$
  3. Every element $x$ of $X$ (subset of $A$) implies that every subset of $x$ is in $X$.
  4. Every $x$ in $X$ with $k+1$ elements is a k-face of $X$. This is the dimension.
  5. 0-dimensional faces are vertices.
  6. 1-dimensional faces are edges.
  7. A face $a$ "includes onto" $b$ ($a \rightsquigarrow b$) whenever $a$ is a proper subset of $b$.

A sheaf $F$ on an abstract simplical complex $X$ consists of:

  1. A vector space $F(a)$ to each face of a of $X$ (the stalk at $a$).
  2. A linear map $F(a \rightsquigarrow b) : F(a) \to F(b)$ (the restriction along $a \rightsquigarrow b$).
  3. $F(b \rightsquigarrow c) \circ F(a \rightsquigarrow b) = F(a \rightsquigarrow c)$ whenever $a \rightsquigarrow b \rightsquigarrow c$.

Given a simplicial complex, a sheaf is merely the assignment of vector-valued data to each face that is compatible with the inclusions of faces.

So to each face we assign a vector space. The face is a vertex of a graph if the face is 0-dimensional. The face is an edge of the graph if the face is 1-dimensional. So we assign vector space to each node and/or edge (and/or more if there's higher dimension). I think I understand that part. I don't understand the "is compatible with the inclusion of faces" part.


From the full book, the definitions are:

A sheaf is a functor that commutes with limits coming from open covers.

Suppose $Λ = \{1, \dotsc , n\}$ is a finite discrete category.

The open set category is:

  • Associated to a topological space $X$
  • Denoted $\mathbf{Open}(X)$
  • Whose objects are the open sets of $X$
  • And whose morphisms $U \to V$ are for each pair related by inclusion $U \subseteq V$.

An open cover $Q$ is:

  • Given a topological space $X$
  • And an open set $U$ on $X$
  • It's a collection of open sets on the open set $U$, $Q = \{U_i\}_{i\in Λ}$ whose union is $U$.

The nerve $N(Q)$ of an open cover $Q$:

  • Gives us an abstract simplical complex $N(Q)$
  • Elements are subsets of $Q$, $I = \{i_0, \dots ,i_n\}$
  • (for which $U_I := U_{i_0} \cap \dots \cap U_{i_n} \neq \emptyset$

A pre-sheaf is:

  • A functor $F : \mathbf{Open}(X)^{op} \to \mathbf{D}$

A pre-cosheaf is:

  • A functor $\hat{F} : \mathbf{Open}(X) \to \mathbf{D}$

A sheaf $F$ is

  • On $Q$ (open cover).
  • If the unique map from $F(U)$ to the limit of $F \circ \rho^{op}_Q$ is an isomorphism, written: $$F(Q)\to\varprojlim_{I\in N(Q)}F(U_I) =: F[Q]$$