The handle-to-cell comparison across a critical level #
Let f be a continuous function on M and let φ i be finitely many charts with values in
ℝⁿ, with pairwise disjoint sources, in which f has the normal form
f = c - Σ_{j < k i} yⱼ² + Σ_{j ≥ k i} yⱼ²,
and whose targets contain the closed ball of radius √ε. Each chart carries the characteristic
map χᵢ : (D^{k i}, ∂D^{k i}) → (M_{c+ε}, M_{c-ε}), w ↦ φ i⁻¹(√ε • (w, 0)), of its cell, and
its attaching map αᵢ : ∂D^{k i} → M_{c-ε} (OpenPartialHomeomorph.morseCharacteristicMap and
OpenPartialHomeomorph.morseAttachingMap), where M_a = {f ≤ a}.
This file forms the space M_{c-ε} ∪_α ⨆ᵢ D^{k i} obtained by attaching one disc along each αᵢ,
as a pushout in TopCat (TauCeti.morseCellAttachment), and the comparison map
Φ : M_{c-ε} ∪_α ⨆ᵢ D^{k i} → M_{c+ε}
induced on the pushout by the inclusion of the lower sublevel and the characteristic maps
(TauCeti.morseCellComparison). When M is Hausdorff, f is continuous, ε > 0 and the
charts have pairwise disjoint sources, Φ is a closed embedding: it is injective because the open
discs are carried into the disjoint chart sources and off the lower sublevel, and closed because
the lower sublevel is closed and the discs are compact. Its image is the lower sublevel together
with the cells {‖φ i x‖² ≤ ε, (φ i x)ⱼ = 0 for j ≥ k i}.
Milnor's cell attachment theorem (TauCeti.exists_sublevel_union_cells_deformation) strong
deformation retracts M_{c+ε} onto exactly this image when f is smooth, the charts are smooth
and their targets contain the closed ball of radius √(2ε), and the band f⁻¹[c - ε, c + ε] is
compact with its critical points in the charts. Transported along Φ, this gives a map
g : M_{c+ε} → M_{c-ε} ∪_α ⨆ᵢ D^{k i} with g ∘ Φ = id and a homotopy from Φ ∘ g to the
identity fixing the image of Φ, in particular the whole lower sublevel. So Φ is a homotopy
equivalence relative to M_{c-ε}; this is the form of the attachment in which a sublevel of a
Morse function is built from the previous one by attaching cells. The charts may have any indices,
including 0 (a disc attached along the empty sphere) and n, and several critical points on one
level are handled simultaneously.
Main definitions #
TauCeti.morseCellAttachment: the spaceM_{c-ε} ∪_α ⨆ᵢ D^{k i}.TauCeti.morseCellAttachmentIncl,TauCeti.morseCellAttachmentCell: its inclusion of the lower sublevel and the characteristic maps of its cells.TauCeti.morseCellAttachmentDesc: the map out ofTauCeti.morseCellAttachmentinduced by compatible maps on the lower sublevel and on the discs.TauCeti.morseCellAttachmentProdDesc: the mapI × (M_{c-ε} ∪_α ⨆ᵢ D^{k i}) → Yinduced by compatible maps onI × M_{c-ε}and on the cylindersI × D^{k i}.TauCeti.morseCellComparison: the comparison mapM_{c-ε} ∪_α ⨆ᵢ D^{k i} → M_{c+ε}.TauCeti.morseCellAttachmentIso: when the attaching maps of two chart families agree strictly up to homeomorphismsh iof the discs, the homeomorphism between their cell attachments which is the identity on the lower sublevel andh ion the cells.
Main results #
TauCeti.isPushout_morseCellAttachment: the pushout square defining the attachment.TauCeti.isClosedEmbedding_morseCellComparison: the comparison map is a closed embedding.TauCeti.range_morseCellComparison: its image is the lower sublevel together with the cells.TauCeti.morseCellAttachmentIso_hom_comp_morseCellComparison: when the characteristic maps are also strictly related,χ₁ i = χ₀ i ∘ h i, the strict reattachment homeomorphism intertwines the comparison maps.TauCeti.exists_morseCellAttachment_deformation: the comparison map is the inclusion of a strong deformation retract ofM_{c+ε}.TauCeti.IsMorse.exists_morseCellAttachment_deformation: the same for a Morse function at any valuecwith a compact band (the cells are indexed by the critical points on the level), with the normalized Morse charts and the widthεsupplied byTauCeti.IsMorse.exists_admissible_morseCharts.
References #
- J. Milnor, Morse Theory, Annals of Mathematics Studies 51, Princeton University Press, 1963, Theorem 3.2 and Remark 3.3.
The space M_{c-ε} ∪_α ⨆ᵢ D^{k i} obtained from the lower sublevel M_{c-ε} = {f ≤ c - ε}
by attaching one k i-disc along the attaching map
OpenPartialHomeomorph.morseAttachingMap (φ i) of each chart φ i, in which
f = c - Σ_{j < k i} yⱼ² + Σ_{j ≥ k i} yⱼ² and whose target contains the closed ball of radius
√ε. It is the pushout in TopCat of the coproduct of the attaching maps and of the boundary
inclusions ∂D^{k i} → D^{k i} (TauCeti.isPushout_morseCellAttachment).
Instances For
The inclusion of the lower sublevel M_{c-ε} into TauCeti.morseCellAttachment.
Instances For
The characteristic map D^{k i} → TauCeti.morseCellAttachment of the cell attached along the
attaching map of the chart φ i.
Instances For
TauCeti.morseCellAttachment is the pushout of the coproduct of the attaching maps along the
coproduct of the boundary inclusions ∂D^{k i} → D^{k i}, with the inclusion of the lower sublevel
and the characteristic maps of the cells as its legs.
In TauCeti.morseCellAttachment, the characteristic map of a cell restricts on the boundary
sphere to the attaching map followed by the inclusion of the lower sublevel.
In TauCeti.morseCellAttachment, the characteristic map of a cell restricts on the boundary
sphere to the attaching map followed by the inclusion of the lower sublevel.
A map on the lower sublevel and maps on the discs which agree on the boundary spheres induce a
map out of TauCeti.morseCellAttachment.
Instances For
The map out of TauCeti.morseCellAttachment restricts to the prescribed map on the lower
sublevel.
The map out of TauCeti.morseCellAttachment restricts to the prescribed map on the lower
sublevel.
The map out of TauCeti.morseCellAttachment restricts to the prescribed map on each disc.
The map out of TauCeti.morseCellAttachment restricts to the prescribed map on each disc.
Maps out of TauCeti.morseCellAttachment are determined by their restrictions to the lower
sublevel and to the cells.
Every point of TauCeti.morseCellAttachment lies in the lower sublevel or in a cell.
The map I × (M_{c-ε} ∪_α ⨆ᵢ D^{k i}) → Y glued from a map L : I × M_{c-ε} → Y and maps
C i : I × D^{k i} → Y that agree with L along the attaching maps: the form of
TauCeti.morseCellAttachmentDesc for homotopies out of TauCeti.morseCellAttachment, given by
CategoryTheory.IsPushout.cylinderDesc.
Instances For
On the lower sublevel, the map glued by TauCeti.morseCellAttachmentProdDesc is L.
On the cell of the chart φ i, the map glued by TauCeti.morseCellAttachmentProdDesc is
C i.
Strict reattachment. Let φ₀ and φ₁ be two families of normal-form charts at the same
level, and let h i be homeomorphisms of the discs D^{k i} restricting to homeomorphisms
g i of their boundary spheres. If the attaching maps agree up to these reparametrizations,
α₁ i = α₀ i ∘ g i, then the map which is the identity on the lower sublevel M_{c-ε} and h i
on the i-th disc is a homeomorphism
M_{c-ε} ∪_{α₁} ⨆ᵢ D^{k i} ≅ M_{c-ε} ∪_{α₀} ⨆ᵢ D^{k i} relative to M_{c-ε}
(TauCeti.morseCellAttachmentIncl_comp_morseCellAttachmentIso_hom). For an orthogonal h i, take
LinearIsometryEquiv.diskIso and LinearIsometryEquiv.diskBoundaryIso.
Instances For
The strict reattachment homeomorphism is the identity on the lower sublevel.
The strict reattachment homeomorphism is the identity on the lower sublevel.
The strict reattachment homeomorphism is h i on the i-th cell.
The strict reattachment homeomorphism is h i on the i-th cell.
The inverse of the strict reattachment homeomorphism is the identity on the lower sublevel.
The inverse of the strict reattachment homeomorphism is the identity on the lower sublevel.
The inverse of the strict reattachment homeomorphism is (h i)⁻¹ on the i-th cell.
The inverse of the strict reattachment homeomorphism is (h i)⁻¹ on the i-th cell.
The strict reattachment homeomorphism sends a point of the lower sublevel to itself.
The strict reattachment homeomorphism sends the point w of the disc of i to the point
h i w of the disc of i.
The comparison map M_{c-ε} ∪_α ⨆ᵢ D^{k i} → M_{c+ε} from TauCeti.morseCellAttachment to
the sublevel M_{c+ε} = {f ≤ c + ε}, induced on the pushout by the inclusion of the lower sublevel
M_{c-ε} and by the characteristic maps OpenPartialHomeomorph.morseCharacteristicMap of the
cells.
Instances For
On the lower sublevel, the comparison map is the inclusion M_{c-ε} → M_{c+ε}.
On the lower sublevel, the comparison map is the inclusion M_{c-ε} → M_{c+ε}.
On the cell of the chart φ i, the comparison map is the characteristic map of the cell.
On the cell of the chart φ i, the comparison map is the characteristic map of the cell.
The comparison map sends a point of the lower sublevel to itself.
The comparison map sends the point w of the cell of the chart φ i to
φ i⁻¹(√ε • (w, 0)).
If the characteristic maps are strictly related, χ₁ i = χ₀ i ∘ h i, then the strict
reattachment homeomorphism intertwines the handle-to-cell comparisons: Φ₀ ∘ (id ⊔ h) = Φ₁.
The image of the comparison map is the lower sublevel M_{c-ε} together with the cells
{‖φ i x‖² ≤ ε, (φ i x)ⱼ = 0 for j ≥ k i} of the charts.
The comparison map is injective: the open discs are carried off the lower sublevel and into the pairwise disjoint chart sources.
The handle-to-cell comparison is a closed embedding. For a continuous f on a Hausdorff
space, the comparison map M_{c-ε} ∪_α ⨆ᵢ D^{k i} → M_{c+ε} of finitely many normal-form charts
with disjoint sources is a closed embedding.
The handle-to-cell comparison is a homotopy equivalence relative to the lower sublevel.
Let f be smooth and let ψ i be finitely many charts with pairwise disjoint sources, whose
targets contain the closed ball of radius √(2ε) about 0, in which
f = c - Σ_{j < k i} yⱼ² + Σ_{j ≥ k i} yⱼ². If the band f⁻¹[c - ε, c + ε] is compact and its
critical points lie in the charts, then the comparison map
Φ : M_{c-ε} ∪_α ⨆ᵢ D^{k i} → M_{c+ε} has a left inverse g such that Φ ∘ g is homotopic to
the identity of M_{c+ε} relative to the image of Φ, which contains the lower sublevel
M_{c-ε}.
Milnor's cell attachment theorem, in pushout form. Let f be a Morse function and c a
value with a compact band f⁻¹[c - δ, c + δ]. Then for some 0 < ε ≤ δ there are normalized Morse
charts ψ p at the critical points p on the level c, with pairwise disjoint sources, such that
the comparison map Φ : M_{c-ε} ∪_α ⨆ₚ D^{λ p} → M_{c+ε} from the lower sublevel with one cell of
dimension the Morse index λ p attached at each p has a left inverse g, and Φ ∘ g is
homotopic to the identity of M_{c+ε} relative to the image of Φ, which contains the lower
sublevel M_{c-ε}.