Strongly locally contractible spaces #
This file defines LocallyContractibleSpace and StronglyLocallyContractibleSpace.
Main definitions #
LocallyContractibleSpace X: classical local contractibility (null-homotopic inclusions).StronglyLocallyContractibleSpace X: a space where each point has a neighborhood basis consisting of contractible sets (not necessarily open).
Main results #
StronglyLocallyContractibleSpace.locallyContractible: SLC implies classical LCinstLocPathConnectedSpace: strongly locally contractible spaces are locally path-connectedStronglyLocallyContractibleSpace.of_bases: a helper to construct strongly locally contractible spaces from a neighborhood basiscontractible_subset_basis: basis of contractible neighborhoods contained in an open setIsOpenEmbedding.stronglyLocallyContractibleSpace: open embeddings preserve strong local contractibilityIsOpen.stronglyLocallyContractibleSpace: open subsets of strongly locally contractible spaces are strongly locally contractible
TODO #
- Define contractible components and prove they are open in strongly locally contractible spaces
- Add examples: convex sets, real vector spaces, star-shaped sets
- Products of strongly locally contractible spaces
Notes #
Terminology: The classical definition of locally contractible (LC) requires that for every
point x and neighborhood U ∋ x, there exists a neighborhood V ∋ x with V ⊆ U such that the
inclusion V ↪ U is null-homotopic. The definition here is strictly stronger: we require
contractible neighborhoods to form a neighborhood basis. This is often called strongly locally
contractible (SLC).
Hierarchy of notions:
- "Basis of open contractible neighborhoods" (strongest)
- "Basis of contractible neighborhoods" (this file, SLC)
- "Null-homotopic inclusions" (classical LC, weakest)
This naming is not used uniformly: according to https://ncatlab.org/nlab/show/locally+contractible+space the second and third notion here could also be called "locally contractible" and "semilocally contractible" respectively. We've enquired at https://math.stackexchange.com/questions/5109428/terminology-for-local-contractibility-locally-contractible-vs-strongly-local in the hope of gettting definitive naming advice.
The Borsuk-Mazurkiewicz counterexample [BM34] shows that classical LC does not
imply SLC. Moreover, from a contractible neighborhood S one generally cannot shrink to an open
V ⊆ S that remains contractible, so requiring neighborhoods to be open is potentially strictly
stronger than SLC.
Classical local contractibility: for every point and every neighborhood U, there exists a neighborhood V ⊆ U such that the inclusion V ↪ U is null-homotopic.
This is weaker than StronglyLocallyContractibleSpace.
Equations
- LocallyContractibleSpace X = ∀ (x : X), ∀ U ∈ nhds x, ∃ (V : Set X) (hVU : V ⊆ U), V ∈ nhds x ∧ (ContinuousMap.inclusion hVU).Nullhomotopic
Instances For
A topological space is strongly locally contractible if, at every point, contractible neighborhoods form a neighborhood basis. Here "contractible" means contractible as a subspace.
This is strictly stronger than the classical notion of locally contractible, which only requires null-homotopic inclusions.
- contractible_basis (x : X) : (nhds x).HasBasis (fun (s : Set X) => s ∈ nhds x ∧ ContractibleSpace ↑s) id
Each neighborhood filter has a basis of contractible subspace neighborhoods.
Instances
A helper to construct a strongly locally contractible space from a neighborhood basis where each basis element is contractible as a subspace.
In a strongly locally contractible space, for any open set U containing x, there is a basis
of contractible neighborhoods of x contained in U.
Strongly locally contractible spaces are locally path-connected.
Open embeddings preserve strong local contractibility: if X is strongly locally contractible
and e : Y → X is an open embedding, then Y is strongly locally contractible.
Open subsets of strongly locally contractible spaces are strongly locally contractible.
The strong notion (contractible neighborhood basis) implies the classical notion (null-homotopic inclusions). The converse is false by the Borsuk-Mazurkiewicz counterexample [BM34]; see also [Mel12] for discussion and the Whitehead manifold example.