Stellar subdivision at a face #
Starring a complex K at one of its faces σ, with a fresh vertex v, replaces the closed
star of σ by the cone with apex v on the boundary of that closed star. It is the combinatorial
model of the geometric move in which a new vertex placed in the interior of σ cones off the
boundary of the closed star; suitably realized, the source and the target of that move are
PL-homeomorphic, but nothing of the kind is asserted here, since the construction below supplies
no placement of v (see the paragraph on realizations).
This file builds that move for PreAbstractSimplicialComplex, Mathlib's downward-closed
collections of nonempty finite faces. The precomplex type is the right home: a starring changes
which vertices are used, since σ itself stops being a face while v becomes one.
The definition is the usual explicit description of the faces, split by whether the new vertex occurs. A face is either
- a face
τofKwithv ∉ τthat does not containσ; whenvis fresh, these are the faces of the deletiondeletion K σ; or insert v ρ, whereρdoes not containσandρ ∪ σis a face ofK— that is,vjoined with a face of the boundary∂σ ∗ link K σof the closed star.
Phrasing the second clause through ρ ∪ σ ∈ K rather than through an explicit join keeps the
whole construction on the original vertex type, so starrings can be iterated. The face collection
is downward closed with no hypothesis at all on σ or v. For a genuine stellar move, the
intended hypotheses are that σ is a face and that v is fresh, meaning {v} is not a face of
K (equivalently, by notMem_of_singleton_notMem, that v occurs in no face); individual
theorems state only the assumptions they need.
Stellar subdivision is the combinatorial substitute for a general subdivision that layer 11 of the
geometric-topology roadmap (TauCetiRoadmap/GeometricTopology/README.md) needs before
combinatorial spheres and balls can be defined: those are complexes that are PL-homeomorphic,
after subdivision, to the boundary of the standard simplex, respectively to the standard
simplex, and the barycentric subdivision alone (Subdivision.Basic) is too rigid to serve, since
it cannot be applied at a single face. The equivalence relation generated by the move is
Subdivision.Stellar.Equivalence, and the ball, sphere and manifold predicates built from it are
in CombinatorialManifold. The definition and the standard-model computations follow
Rourke--Sanderson, Introduction to Piecewise-Linear Topology, Chapter 2 (starring and stellar
moves).
No claim is made here that a starring preserves the geometric realization; that identification is
realization work, parallel to the corresponding open question for Subdivision.Basic. What is
proved is the combinatorial shadow of it: the dimension is unchanged, the part of the subdivision
away from the new vertex is exactly deletion K σ, and the link of the new vertex is exactly the
boundary closedStar K σ ⊓ deletion K σ of the closed star, so the subdivision is glued from the
same two pieces the geometric picture uses.
Main definitions #
PreAbstractSimplicialComplex.stellarSubdivision K σ v: the complex obtained fromKby starring the faceσat the fresh vertexv.
Main results #
PreAbstractSimplicialComplex.mem_stellarSubdivision_iff_of_notMemandinsert_mem_stellarSubdivision_iff: the two halves of the face description, one for faces missing the new vertex and one for faces containing it.PreAbstractSimplicialComplex.singleton_mem_stellarSubdivision_iff: the new vertex is a vertex of the subdivision exactly whenσwas a face, andPreAbstractSimplicialComplex.self_notMem_stellarSubdivision: the starred face is destroyed, as long as the new vertex does not already lie in it.PreAbstractSimplicialComplex.deletion_stellarSubdivision_singletonandlink_stellarSubdivision_singleton: away from the new vertex the subdivision isdeletion K σ, and the link of the new vertex is the boundaryclosedStar K σ ⊓ deletion K σof the closed star.PreAbstractSimplicialComplex.map_stellarSubdivision: injective relabeling commutes with stellar subdivision.PreAbstractSimplicialComplex.dimension_stellarSubdivision: starring preserves the dimension.PreAbstractSimplicialComplex.finite_faces_stellarSubdivision_iff: a genuine starring preserves and reflects finiteness.PreAbstractSimplicialComplex.isCone_stellarSubdivision_of_closedStar_eq_self: starring a complex that is its own closed star atσproduces a cone with apex the new vertex; for a simplex,PreAbstractSimplicialComplex.link_stellarSubdivision_simplex_selfidentifies the link of that apex with the simplex boundary.
The stellar subdivision (or starring) of K at the face σ, using the new vertex v.
Its faces are the faces τ of K with v ∉ τ that do not contain σ, together with the sets
insert v ρ for which ρ does not contain σ and ρ ∪ σ is a face of K. Under the intended
freshness hypothesis, equivalently, the closed star of σ is removed and replaced by the cone
with apex v on the boundary of that closed star.
The intended hypotheses are that σ is a face of K and that v is fresh, i.e. {v} ∉ K; they
are not part of the definition, and the degenerate values are pinned by
stellarSubdivision_empty and stellarSubdivision_eq_self_of_notMem.
Equations
Instances For
The defining description of the faces of a stellar subdivision, split by whether the new vertex occurs.
A set missing the new vertex is a face of the stellar subdivision exactly when it is a face of
K not containing the starred face: these are precisely the faces of deletion K σ.
A set containing the new vertex is a face of the stellar subdivision exactly when the rest of
it is a face of the boundary of the closed star of σ: it must not contain σ, while its union
with σ must remain a face.
The new vertex becomes a vertex of the stellar subdivision exactly when the starred set was a face to begin with.
Starring destroys the starred face, as long as the new vertex does not already lie in σ:
every face of the subdivision either avoids σ or contains the new vertex, and in the latter case
avoids σ after erasing it. The hypothesis v ∉ σ is needed, since for v ∈ σ one has
σ.erase v ∪ σ = σ, so σ survives the starring whenever it was a face.
Starring a genuine face at a vertex outside that face genuinely changes the complex.
Starring at the empty set gives the bottom precomplex.
Starring at a nonempty set that is not a face changes nothing, provided the new vertex is
fresh: no face of K can contain the starred set, and freshness makes the new vertex occur in no
face of K and in no face of the subdivision. Freshness is needed: without it, starring can
destroy the faces of K that contain v.
The two pieces glued by a starring #
Away from the new vertex a stellar subdivision is the deletion deletion K σ, and the link of
the new vertex is the boundary closedStar K σ ⊓ deletion K σ of the closed star of σ. Together
with closedStar_sup_deletion these say that the subdivision is the union of deletion K σ and
the cone with apex v on that boundary, which is the geometric description of the move.
The faces of a stellar subdivision missing the new vertex form exactly the deletion of the starred face.
The deletion of the starred face survives untouched inside the stellar subdivision.
The link of the new vertex in a stellar subdivision is the boundary of the closed star of
the starred face, namely closedStar K σ ⊓ deletion K σ: the faces ρ with ρ ∪ σ a face of K
that do not themselves contain σ. This is the combinatorial form of "v is coned off over
∂σ ∗ link K σ".
Relabeling #
An injective relabeling commutes with stellar subdivision.
Dimension and finiteness #
Starring a nonempty set at a fresh vertex preserves the dimension.
Starring a complex with finitely many faces again gives finitely many faces: a face either is
a face of K, or is the new vertex adjoined to a subset of a face of K.
A genuine stellar subdivision preserves and reflects finiteness of the face collection.
Starring a closed star, and the standard model #
When K is its own closed star at σ — the case of a simplex starred at its top face — every
face can absorb the new vertex, so the subdivision is a cone with apex v.
Starring a complex that is its own closed star at σ produces a cone with apex the new
vertex.
Starring a simplex at its whole vertex set produces a cone with apex the new vertex.
The starred simplex is the cone on the simplex boundary: the link of the new vertex is
exactly simplexBoundary V. This is the standard model of the move, and pins the convention.
Starring a simplex at its whole vertex set leaves the dimension at V.card - 1.