Comparing two presentations of a rational localisation #
A rational subset U = R(T/s) of Spa(A,A⁺) has many presentations (T,s), so the completed
localisations they give must be compared by canonical isomorphisms, compatible for three
presentations. This file supplies the conditional half: given comparison maps compatible with the
structure maps from A, they are mutually inverse and compose correctly.
The other half is the passage from an equality of two rational subsets to the existence of those
maps — the step Mathlib-side rationalSubset_subset_rationalSubset_iff stops short of, in its own
words "leaving the passage from those facts to invertibility of s and power-boundedness of t/s
in the coordinate ring as a separate, genuinely algebraic step" (Wedhorn §8.2). That passage is
TauCeti.ValuationSpectrum.presentationRingEquivOfEq, which instantiates the comparison theory
here through Wedhorn's Proposition 8.2(1).
The file has two halves. The first bundles a presentation as Presentation and orders those
bundles by refinement — the Preorder and IsDirected instances,
Presentation.commonRefinement as the common refinement, and le_def as the route from p ≤ q
to a cofactor.
The second is the comparison theory, and there everything is uniqueness: no comparison map is
constructed. Given maps in both directions that commute with the structure maps, they are mutually
inverse, and given three presentations the comparison through the middle one is the direct
comparison. That is exactly what
TauCeti.Huber.PairOfDefinition.eq_id_of_comp_toCompletionLoc_eq_self and
…eq_comp_of_comp_toCompletionLoc_eq say about maps out of A⟨T/s⟩, so each proof in that half
is a single application of one of them.
Main definitions #
TauCeti.Huber.PairOfDefinition.Presentation: the bundle(num, den, hasDenominatorPower)of a presentation, with the refinement preorderPresentation.RefinedByand the common refinementPresentation.commonRefinementmaking refinement directed.Presentation.commonRefinement_numandPresentation.commonRefinement_denare its projection equations, since the body is not exported.TauCeti.Huber.PairOfDefinition.presentationRingEquiv: the canonical isomorphismA⟨T/s⟩ ≃+* A⟨T'/s'⟩assembled from compatible comparison maps in both directions.
Main results #
TauCeti.Huber.PairOfDefinition.comp_eq_id_of_comp_toCompletionLoc_eq: compatible comparison maps in both directions compose to the identity.TauCeti.Huber.PairOfDefinition.eq_comp_of_comp_toCompletionLoc_eq_three: compatibility for three presentations — the comparison from the first to the third is the composite through the second.TauCeti.Huber.PairOfDefinition.presentationRingEquiv_coeand…_symm_coe: the characteristic equations — the isomorphism is the forward map it was built from, and its inverse is the backward one, so it introduces nothing new.TauCeti.Huber.PairOfDefinition.continuous_presentationRingEquiv, its…_symmcounterpart, and…_coe_comp_toCompletionLoc: it is an isomorphism of topological rings, and compatible with the structure maps fromA— the property that determines it.
Provenance #
The bundle-and-refinement section is adapted from AINTLIB (see References): the idea of indexing
the structure presheaf by presentation data rather than by rational subsets, and the refinement
relation between presentations, are StructurePresheafLimit.lean's. The bundling differs
deliberately — AINTLIB threads RationalLocData records through explicit hypotheses, while here
Presentation packs (num, den, hasDenominatorPower) so that refinement is a Preorder and
downstream constructions can be functorial. The comparison theory in the rest of this file is
original to this repository.
References #
- C. Birkbeck, AINTLIB, branch
dev/adic-spaces, commit37bbdaeb,projects/AdicSpaces/Adic spaces/StructurePresheafLimit.lean, Apache-2.0. - T. Wedhorn, Adic Spaces, §8.1–8.2.
The bundle of presentations, and refinement #
The comparison theory below works with two presentations given separately; consumers indexing a
construction by all presentations — the intended structure presheaf — need them bundled and
ordered. Refinement here is a cofactor condition on the presentation data. It is sufficient for
containment of the rational subsets. The converse is not claimed, and the gap is not the
comparison-map step described above: exists_refinement_of_subset already produces numerator and
denominator data from a containment. What it does not supply is the standing HasDenominatorPower
hypothesis for the re-presented pair, which is what Presentation requires.
A presentation of a rational localization: a numerator finset, a denominator, and the
standing HasDenominatorPower hypothesis for the localization away from the denominator.
- num : Finset A
The numerators.
- den : A
The denominator.
- hasDenominatorPower : P.HasDenominatorPower self.num self.den (Localization.Away self.den)
The denominator-power hypothesis for
Localization.Away den.
Instances For
The refinement relation: q refines p when some cofactor r has
q.den = p.den * r and carries every numerator of p into a numerator of q.
Instances For
The witness facts for the trivial refinement, with cofactor 1.
The witness facts for a composite refinement: the cofactors multiply.
Every presentation refines itself, with cofactor 1.
Refinements compose: the cofactors multiply.
Refinement is a preorder, by Presentation.RefinedBy.refl and
Presentation.RefinedBy.trans.
These two stay as named theorems rather than being inlined into the fields below. The instance is
public, so its body is exposed for typeclass resolution and cannot unfold RefinedBy, whose
body this file deliberately does not export — inlining gives
Invalid ⟨...⟩ notation: The expected type p.RefinedBy p is not an inductive type. A theorem body
is not exposed, so it may unfold it; that asymmetry is what forces the two names.
Equations
- TauCeti.Huber.PairOfDefinition.instPreorderPresentation = { le := TauCeti.Huber.PairOfDefinition.Presentation.RefinedBy, le_refl := ⋯, le_trans := ⋯, lt_iff_le_not_ge := ⋯ }
The refinement preorder, unfolded in one step: the single introduction/elimination
lemma for ≤. The body of RefinedBy is not exported, so this is the route from p ≤ q to a
cofactor.
le_def, not le_iff: this is the one-step definitional unfolding of a custom ≤, which is
what Mathlib names le_def (Order/Quotient.lean, Order/Hom/Basic.lean,
Order/Preorder/Finsupp.lean, several of them .rfl as here). Bare le_iff there is reserved
for characterisations that are not the definition.
The common refinement, refining both factors: the numerators are the pairwise
products of the factors' numerator sets augmented by their own denominators, and the denominator
is the product. The augmentation matches rationalSubset_inter's presentation of an
intersection.
Presentation carries no openness or admissibility field, so nothing here tracks or preserves
openness of the numerator ideal; a consumer that needs it — the structure presheaf's index — must
carry and re-establish it itself.
Equations
Instances For
The numerator equation for the common refinement. The body of commonRefinement is not
exported, so this is how a consumer computes with it — in particular, how it is matched against
rationalSubset_inter's presentation of an intersection.
The denominator equation for the common refinement.
The common refinement refines its left factor, with cofactor the right denominator.
The common refinement refines its right factor, with cofactor the left denominator.
The refinement preorder is directed: any two presentations admit a common refinement,
namely Presentation.commonRefinement. So presentations of the same rational subset never sit as
independent factors in a limit over this order — both map onwards to a common refinement. The
existential form is exists_ge_ge, which this instance supplies generically.
Comparing two presentations of the same subset #
Compatible comparison maps between two presentations are mutually inverse. If g carries
the structure map of A⟨T/s⟩ to that of A⟨T'/s'⟩ and h carries it back, then h ∘ g fixes
the structure map of A⟨T/s⟩, so it is the identity.
The comparison isomorphism between two presentations. Comparison maps in both directions
that are compatible with the structure maps from A assemble into a ring isomorphism, because
each composite fixes a structure map and is therefore the identity.
This is the canonical half of presentation independence: it says the comparison is an
isomorphism and is determined by compatibility, not that the compatibility hypotheses hold for
two presentations of the same rational subset. Supplying those is a separate step:
TauCeti.ValuationSpectrum.presentationRingEquivOfEq derives them from an equality of rational
subsets through Wedhorn's Proposition 8.2(1).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Compatibility for three presentations. A comparison map from the first presentation to
the third is the composite of the comparisons through the second, whenever all three are
compatible with the structure maps from A. This is the cocycle condition that makes the
comparisons a coherent system rather than a family of unrelated isomorphisms.
The comparison isomorphism is the map it was built from. The characteristic equation of
presentationRingEquiv: it does not introduce a new map, it packages g together with the
inverse supplied by h.
The inverse of the comparison isomorphism is the backward map it was built from.
The comparison isomorphism is continuous, being g.
The comparison isomorphism has a continuous inverse, being h. With
continuous_presentationRingEquiv this makes it an isomorphism of topological rings.
The comparison isomorphism is compatible with the structure maps from A, which is the
property that determines it.