Preimage transfer of slicings #
For a functor F : C ⥤ D and a slicing s on D, the raw phase collection
P_F(phi)(E) := s.P phi (F.obj E)
is the common categorical core of both constructions in arXiv:2607.28411v1,
Definitions 3.1 and 3.6. Remarks 3.2 and 3.7 explicitly warn that this raw
collection need not be a slicing, even when F is conservative. Accordingly,
Slicing.PreimageData records precisely the two slicing axioms which do not
follow formally from functoriality and shift compatibility: Hom-vanishing and
HN existence.
This factorization is deliberately honest about the theorem boundary. The
geometric inducing results (Propositions 3.3 and 3.8, via Appendix A) are
expected to construct PreimageData; bare adjunction and conservativity do not.
Main definitions #
Slicing.preimagePhase,Slicing.PreimageData,Slicing.preimage: the raw phase collection, its two non-formal axioms, and the slicing they produce.Slicing.pullback,Slicing.pushforward: the source-facing names of Definitions 3.1 and 3.6.
Main results #
Slicing.preimage_id,Slicing.preimage_comp,Slicing.preimage_iso,Slicing.preimage_square: the preimage slicing is functorial in the detecting functor up to natural isomorphism; the square is the shape of Lemma 3.10(1.b)/(2.b).
References #
- arXiv:2607.28411v1, Definitions 3.1 and 3.6, Remarks 3.2 and 3.7, Lemma 3.10.
The raw inverse-image phase collection along a functor.
It is intentionally only an ObjectProperty, not a Slicing: conservativity
alone does not supply Hom-vanishing or HN filtrations.
Equations
- s.preimagePhase F phi E = s.P phi (F.obj E)
Instances For
Source-facing name for Definition 3.1 of arXiv:2607.28411v1. A geometric pullback slicing is computed using the direct-image functor.
Equations
- s.pullbackPhaseCollection push = s.preimagePhase push
Instances For
Source-facing name for Definition 3.6 of arXiv:2607.28411v1. A geometric pushforward slicing is computed using the inverse-image functor.
Equations
- s.pushforwardPhaseCollection pull = s.preimagePhase pull
Instances For
The genuinely missing axioms for turning the raw inverse-image phase collection into a slicing.
Closure under isomorphisms, zero membership, and the shift law follow from
s and F.CommShift. The two fields here are exactly what remains.
- hom_vanishing (phi₁ phi₂ : ℝ) (A B : C) : phi₂ < phi₁ → s.P phi₁ (F.obj A) → s.P phi₂ (F.obj B) → ∀ (g : A ⟶ B), g = 0
Hom-vanishing for objects whose images lie in separated phase slices.
- hn_exists (E : C) : Nonempty (HNFiltration C (s.preimagePhase F) E)
HN filtrations in the source with factors detected by
F.
Instances For
The identity functor satisfies the two non-formal preimage axioms.
Construct the genuine inverse-image slicing once the two non-formal axioms have been supplied.
Equations
- s.preimage F h = { P := s.preimagePhase F, closedUnderIso := ⋯, zero_mem := ⋯, shift_iff := ⋯, hom_vanishing := ⋯, hn_exists := ⋯ }
Instances For
Preimage witnesses compose in the same contravariant order as their functors. The second witness is measured against the slicing constructed by the first stage, so its HN filtrations already have exactly the phases needed for the composite.
A natural isomorphism of detecting functors transports the two non-formal preimage axioms. In particular, the HN factors are unchanged because phase membership is invariant under the component isomorphisms.
Constructing a preimage slicing in two stages agrees with constructing it along the composite functor. The statement is independent of the proof terms used to witness the slicing axioms.
Preimage slicings are invariant under a natural isomorphism of the detecting functors.
A triangulated equivalence supplies the two non-formal preimage axioms.
Unlike Slicing.mapEquiv, this theorem allows the source and target
categories to differ. HN filtrations are transported through the inverse
functor and then identified with the original object by the unit isomorphism.
The preimage slicing along a triangulated equivalence is detected exactly by the equivalence functor.
For a faithful functor, target Hom-vanishing supplies the Hom component of
PreimageData; only HN existence remains to be proved.
Equations
- ⋯ = ⋯
Instances For
The preimage lifting criterion is stable under a uniform translation of all phases.
Equations
- ⋯ = ⋯
Instances For
Source-facing genuine pullback name. Its explicit PreimageData argument
is the formal reminder that conservativity alone is insufficient.
Instances For
Source-facing genuine pushforward name.
Equations
- s.pushforward pull h = s.preimage pull h
Instances For
Two-step preimage slicings around a commuting square of detecting functors agree: the
slicing pulled back along G₂ and then along F₁ is the slicing pulled back along F₂ and
then along G₁, when F₁ ⋙ G₂ ≅ G₁ ⋙ F₂. This is the shape of Lemma 3.10(1.b) and (2.b) of
arXiv:2607.28411v1, the compatibility of pullback and pushforward of slicings with base change
along a field extension: base change and the geometric transfer are the two ways around the
square, and the slicing does not see which is taken first. Here F₁ : C ⥤ D and
F₂ : C' ⥤ E are the transfer functors (f_{ℓ*}, f_* for (1.b); f_ℓ^*, f^* for (2.b))
and G₁, G₂ the base-change direct images π_*; all four run between bounded categories
carrying slicings, which is the finite-extension case. For an infinite extension the
base-change slicing is not the preimage of a Slicing along π_*, whose target is the
Ind-extension in Dqc, and only the shape is captured here.