Reattaching Morse cells along a homotopy of characteristic maps #
Passing a critical level c of a Morse function attaches one cell at each critical point on the
level (TauCeti.morseCellAttachment), but the attaching maps depend on the chosen normal-form
charts. This file compares the attachments built from two chart families φ₀ and φ₁ with the
same indices, following Milnor's Lemma 3.6.
Write M_a = {f ≤ a}, χⱼ i : (D^{k i}, ∂D^{k i}) → (M_{c+ε}, M_{c-ε}) for the characteristic
maps of the cells and αⱼ i for their attaching maps. Suppose given, for each chart index i, a
self-map h i of the pair (D^{k i}, ∂D^{k i}) and a homotopy of maps of pairs
K i : χ₁ i ≃ χ₀ i ∘ h i, that is, a homotopy whose restriction H i to the boundary sphere stays
in the lower sublevel M_{c-ε}. Then H i is a homotopy from α₁ i to α₀ i ∘ h i, and the
reattachment map
E : M_{c-ε} ∪_{α₁} ⨆ᵢ D^{k i} → M_{c-ε} ∪_{α₀} ⨆ᵢ D^{k i}
(TauCeti.morseCellReattachment) is the identity on M_{c-ε}, sends a point x of the disc of
i with ‖x‖ ≤ 1/2 to the point h i (2x) of the disc of i, and sends a point with
1/2 ≤ ‖x‖ to the point H i (2 - 2‖x‖, x/‖x‖) of M_{c-ε}. These pieces agree on the seam
‖x‖ = 1/2, because H i (1, ·) = α₀ i ∘ h i and h i preserves the boundary sphere, and on the
attaching sphere ‖x‖ = 1, because H i (0, ·) = α₁ i. It is the fiberwise reattachment map
TauCeti.fiberwiseCellReattachment over a point. When 0 < ε, any self-map h i of the
disc admitting such a homotopy K i preserves the boundary sphere: at the end of K i the sphere
is sent by χ₀ i ∘ h i into M_{c-ε}, which χ₀ i meets exactly on the sphere.
Composed with the handle-to-cell comparison Φ₀ : M_{c-ε} ∪_{α₀} ⨆ᵢ D^{k i} → M_{c+ε}
(TauCeti.morseCellComparison), the map E reads K i along the map
x ↦ (τ x, ρ x) from the disc to the cylinder I × D^{k i}, where τ x = 1 and ρ x = 2x for
‖x‖ ≤ 1/2 and τ x = 2 - 2‖x‖ and ρ x = x/‖x‖ for 1/2 ≤ ‖x‖. On the boundary sphere this
map is x ↦ (0, x), so the straight-line homotopy to x ↦ (0, x) in the convex cylinder fixes
the boundary sphere. Through K i, it gives a homotopy from Φ₀ ∘ E to Φ₁ relative to the
lower sublevel (TauCeti.homotopicRel_morseCellReattachment_comp_morseCellComparison), which is
the homotopy TauCeti.fiberwiseCellReattachmentHomotopy over a point.
Consequently, if g₀ is a left inverse of Φ₀, then E is homotopic to g₀ ∘ Φ₁ relative to the
lower sublevel. Milnor's cell attachment theorem supplies such left inverses for admissible charts
(TauCeti.exists_morseCellAttachment_deformation), with Φⱼ ∘ gⱼ homotopic to the identity
relative to the image of Φⱼ. From this:
Eis a homotopy equivalence whose homotopy inverse and both homotopies fix the lower sublevel;- the relative homotopy class of
Edoes not depend on the homotopiesK i; - for three chart families the reattachment maps compose up to homotopy relative to the lower sublevel.
Main definitions #
TauCeti.morseCellReattachment: the reattachment mapM_{c-ε} ∪_{α₁} ⨆ᵢ D^{k i} → M_{c-ε} ∪_{α₀} ⨆ᵢ D^{k i}of a family of homotopies of pairsχ₁ i ≃ χ₀ i ∘ h i.
Main results #
TauCeti.homotopicRel_morseCellReattachment_comp_morseCellComparison:Φ₀ ∘ Eis homotopic toΦ₁relative to the lower sublevel.TauCeti.homotopicRel_morseCellReattachment_of_leftInverse: for a left inverseg₀ofΦ₀,Eis homotopic tog₀ ∘ Φ₁relative to the lower sublevel.TauCeti.homotopicRel_morseCellReattachment_morseCellReattachment: two reattachment maps between the same attachments are homotopic relative to the lower sublevel, providedΦ₀has a left inverse.TauCeti.homotopicRel_morseCellReattachment_comp_morseCellReattachment: reattachment maps compose up to homotopy relative to the lower sublevel, providedΦ₀has a left inverse.TauCeti.exists_homotopicRel_morseCellReattachment_comp: Milnor's Lemma 3.6 for Morse cells. For admissible chart families, the reattachment map is a homotopy equivalence relative to the lower sublevel.
References #
- J. Milnor, Morse Theory, Annals of Mathematics Studies 51, Princeton University Press, 1963, Lemma 3.6 and Remark 3.3.
The cell attachment of a chart family is the fiberwise cell attachment over a point, with the
lower sublevel M_{c-ε} viewed inside the upper sublevel M_{c+ε}.
The reattachment map E : M_{c-ε} ∪_{α₁} ⨆ᵢ D^{k i} → M_{c-ε} ∪_{α₀} ⨆ᵢ D^{k i} of a
family of self-maps h i of the disc pairs (D^{k i}, ∂D^{k i}) and of homotopies of maps of pairs
K i : χ₁ i ≃ χ₀ i ∘ h i between the characteristic maps of the cells of the chart families φ₁
and φ₀. It is the identity on the lower sublevel M_{c-ε}; on the disc of i it sends x to
the point h i (2x) of the disc of i when ‖x‖ ≤ 1/2, and to the point
K i (2 - 2‖x‖, x/‖x‖) of M_{c-ε} when 1/2 ≤ ‖x‖. It is the fiberwise reattachment map
TauCeti.fiberwiseCellReattachment over a point.
Instances For
The reattachment map sends a point of the lower sublevel to itself.
The reattachment map is the identity on the lower sublevel.
The reattachment map is the identity on the lower sublevel.
On the disc of radius 1/2 of the cell of i, the reattachment map is x ↦ h i (2x), read in
the cell of i.
Outside the disc of radius 1/2 of the cell of i, the reattachment map is
x ↦ K i (2 - 2‖x‖, x/‖x‖), read in the lower sublevel.
Φ₀ ∘ E is homotopic to Φ₁ relative to the lower sublevel. The reattachment map followed
by the handle-to-cell comparison Φ₀ is homotopic to the handle-to-cell comparison Φ₁, by a
homotopy fixing the lower sublevel M_{c-ε}: the homotopy
TauCeti.fiberwiseCellReattachmentHomotopy over a point.
For a left inverse g₀ of the handle-to-cell comparison Φ₀, the reattachment map is
homotopic to g₀ ∘ Φ₁ relative to the lower sublevel.
The reattachment map does not depend on the homotopy of characteristic maps. Two
reattachment maps between the same attachments, built from any self-maps of the disc pairs and any
homotopies of maps of pairs, are homotopic relative to the lower sublevel, provided the target
handle-to-cell comparison Φ₀ has a left inverse.
Reattachment maps compose up to homotopy. For three chart families φ₀, φ₁, φ₂ and
reattachment maps E₀₁, E₁₂, E₀₂ between their attachments, E₀₁ ∘ E₁₂ is homotopic to E₀₂
relative to the lower sublevel, provided Φ₀ has a left inverse.
The reattachment map is a homotopy equivalence relative to the lower sublevel. If both
handle-to-cell comparisons Φⱼ have left inverses gⱼ, and Φ₁ ∘ g₁ is homotopic to the
identity relative to the image of Φ₁, then the reattachment map E has a homotopy inverse E'
which is the identity on the lower sublevel, with E' ∘ E and E ∘ E' homotopic to the identities
relative to the lower sublevel. The inverse is g₁ ∘ Φ₀.
Milnor's Lemma 3.6 for Morse cells. Let f be smooth and let ψ₀, ψ₁ be two families of
normal-form charts with the same indices k i, each with pairwise disjoint sources, targets
containing the closed ball of radius √(2ε), and sources covering the critical points of the
compact band f⁻¹[c - ε, c + ε]. Given self-maps h i of the discs and homotopies of maps of
pairs K i : χ₁ i ≃ χ₀ i ∘ h i between the characteristic maps of the cells, the reattachment map
E : M_{c-ε} ∪_{α₁} ⨆ᵢ D^{k i} → M_{c-ε} ∪_{α₀} ⨆ᵢ D^{k i} is a homotopy equivalence relative to
the lower sublevel: it has a homotopy inverse which is the identity on M_{c-ε}, and both
composites are homotopic to the identities relative to M_{c-ε}. The maps h i need not be given
as maps of pairs: since 0 < ε, the homotopies K i force them to preserve the boundary spheres
(TauCeti.mapsTo_range_diskBoundaryInclusion_of_homotopy_morseCharacteristicMap).