Čech Cohomology 00: Roadmap for Calculations
Čech cohomology / 00
Roadmap for calculations
Čech cohomology turns sheaf cohomology into a finite-looking problem: choose a cover, write local data on intersections, take coboundaries, and identify what cannot be patched away.
The practical problem is simple to state. A sheaf \(\mathcal F\) gives local data on open sets. Sheaf cohomology measures the failure of local data to become global data. Čech cohomology makes that failure explicit by replacing the space with an open cover
\[\mathfrak U=\{U_i\}_{i\in I}\]and replacing global questions by sections on intersections
\[U_{i_0\cdots i_q}=U_{i_0}\cap\cdots\cap U_{i_q}.\]The slogan for computation is:
Calculation principle
Choose a cover whose finite intersections are easy for the sheaf. Then compute the cohomology of the Čech complex
\(C^0(\mathfrak U,\mathcal F) \xrightarrow{\delta} C^1(\mathfrak U,\mathcal F) \xrightarrow{\delta} C^2(\mathfrak U,\mathcal F) \xrightarrow{\delta}\cdots .\)
Calculus intuition
If you know de Rham cohomology from vector calculus, the same obstruction pattern is already familiar:
\[C^\infty \xrightarrow{\nabla} \text{vector fields} \xrightarrow{\nabla\times} \text{vector fields} \xrightarrow{\nabla\cdot} C^\infty.\]The equations \(\nabla\times\nabla f=0\) and \(\nabla\cdot(\nabla\times A)=0\) are the calculus face of \(d^2=0\). Čech cohomology has the parallel identity \(\delta^2=0\). The difference is that de Rham starts from differential equations whose local solutions may fail to patch, while Čech records that patching failure directly.
For example, a closed \(1\)-form \(\alpha\) is locally exact:
\[\alpha=df_i\qquad\text{on }U_i.\]On overlaps, \(d(f_j-f_i)=0\), so the differences \(f_j-f_i\) are locally constant. These constants form a Čech \(1\)-cocycle. If the cocycle is a coboundary, the local potentials patch to one global potential; if not, \(\alpha\) is closed but not exact.
What each degree means
The low degrees are the ones that appear constantly in geometry.
| Degree | Data | Practical meaning |
|---|---|---|
| \(0\) | sections \(s_i\in\mathcal F(U_i)\) | local sections that agree on overlaps glue to a global section |
| \(1\) | sections \(a_{ij}\in\mathcal F(U_{ij})\) | overlap errors, transition functions, and first obstruction classes |
| \(2\) | sections \(b_{ijk}\in\mathcal F(U_{ijk})\) | compatibility of overlap data, Chern classes, and connecting maps |
Degree zero recovers global sections. Degree one records whether overlap data can be written as differences of local choices. Degree two appears naturally when a multiplicative cocycle is lifted locally to additive logarithms.
The article sequence
The order is intentional: first import the calculus intuition, then turn it into Čech notation, then use the notation for actual sheaf-cohomology computations.
What to calculate first
For a sheaf cohomology computation, do not start by invoking abstract definitions. Start with these questions:
- What sheaf is being computed: constant, holomorphic, smooth, units, line-bundle sections, or a quotient sheaf?
- What cover makes the sections on intersections explicit?
- Are the finite intersections acyclic for the sheaf?
- What are the cochains in degree \(0\), \(1\), and maybe \(2\)?
- Which cocycles are coboundaries?
If the cover is Leray for the sheaf, the answer from this cover is already the sheaf cohomology group \(H^q(X,\mathcal F)\). If not, refine the cover or move to an exact sequence.
Connection to sheaf cohomology
The companion sheaf-cohomology notes explain derived functors, exact sequences, and fine resolutions. This series keeps the computational front end visible: where the cocycles live, how signs work, and how a class is represented in practice.