Documentation

TauCeti.RingTheory.Huber.Basic

Huber rings and Tate rings #

A pair of definition for a topological ring A is an open subring A₀ ⊆ A together with a finitely generated ideal I ⊆ A₀ whose adic topology is the subspace topology of A₀. A ring admitting one is a Huber ring (Wedhorn's f-adic ring), and a Huber ring containing a topologically nilpotent unit is a Tate ring.

Everything the later layers use about the topology of a Huber ring comes from one statement: the images in A of the powers Iⁿ are a neighbourhood basis of zero (TauCeti.Huber.PairOfDefinition.hasBasis_nhds_zero). They are open additive subgroups, so a Huber ring is nonarchimedean, which is exactly the hypothesis under which TauCeti/RingTheory/Huber/PowerBounded.lean makes A° a subring.

Main definitions #

Main results #

Provenance #

PairOfDefinition and IsHuberRing follow the shape of sfingali's mathlib4#42312, which bundles the same five fields; the name PairOfDefinition is used rather than that PR's RingOfDefinition because the structure carries the pair (A₀, I), which is what Wedhorn calls a pair of definition — a ring of definition is the A₀ alone. Everything else here is new. The selection of results follows AdicSpaces/Suggested.lean in the roadmap.

Implementation notes #

The pair of definition is data, not a Prop-valued field of the ring: a Huber ring has many pairs of definition and later layers choose between them. IsHuberRing is the Prop asserting that the type of pairs is nonempty, in the shape used by mathlib4#42312.

References #

theorem TauCeti.Huber.exists_sum_eq_of_mem_span_mul {R : Type u_1} [CommSemiring R] (G : Finset R) (K : Ideal R) {b : R} (hb : b ∈ Ideal.span ↑G * K) :
∃ (c : R → R), (∀ (z : R), c z ∈ K) ∧ ∑ z ∈ G, z * c z = b

An element of (G) * K is a K-linear combination of the finite family G.

This is Mathlib's Submodule.mem_ideal_smul_span_iff_exists_sum' in the form the Huber theory uses it: a Finset.sum over G itself, with cofactors given by a function on all of R, rather than a Finsupp on the subtype ↥G. It is what bounds, uniformly in k, the number of terms needed to write an element of Iⁿ⁺ᵏ = Iⁿ * Iᵏ over generators of Iⁿ, both in Wedhorn Remark 6.8 (TauCeti.RingTheory.Huber.Completion) and in the identification of the neighbourhood subgroups of A⟨X⟩_T with the powers of one finitely generated ideal (TauCeti.RingTheory.Huber.WeightedRestrictedSeries.PairOfDefinition).

A pair of definition (A₀, I) for a topological ring A: an open subring A₀ together with a finitely generated ideal I of A₀ whose adic topology is the subspace topology.

This is data rather than a proposition, because a Huber ring generally has many pairs of definition and the later theory chooses among them.

Instances For

    A topological ring is a Huber ring — Wedhorn's f-adic ring — if it admits a pair of definition.

    Instances

      A pseudouniformiser of a topological ring is a topologically nilpotent unit.

      This is the topological notion, unrelated to Mathlib's Valuation.IsUniformizer, which asks a discretely valued ring's element to have valuation the generator of the value group. A Tate ring need carry no valuation at all.

      Equations
      Instances For

        A pseudouniformiser is a unit.

        A pseudouniformiser is topologically nilpotent.

        theorem TauCeti.Huber.IsPseudoUniformizer.map {A : Type u_1} {B : Type u_2} {F : Type u_3} [MonoidWithZero A] [TopologicalSpace A] [MonoidWithZero B] [TopologicalSpace B] [FunLike F A B] [MonoidWithZeroHomClass F A B] {φ : F} (hφ : Continuous ⇑φ) {a : A} (ha : IsPseudoUniformizer a) :

        A continuous morphism of topological monoids with zero sends pseudouniformisers to pseudouniformisers.

        A Tate ring is a Huber ring containing a pseudouniformiser, that is, a topologically nilpotent unit.

        Instances

          An adic topology transports along a ring equivalence that is also an inducing map.

          This is what lets a ring of definition carry an ideal of definition: PairOfDefinition asks for an Ideal A₀ whose adic topology is the subspace topology, while the ideal at hand usually lives in a ring that is only equivalent to A₀.

          A continuous homomorphism out of a Tate ring makes a Huber target Tate (Wedhorn, Adic Spaces, Proposition 6.25): the image of a pseudouniformiser of A is a pseudouniformiser of B.

          The target must already be known to be Huber; this supplies only the pseudouniformiser.

          The image in A of the n-th power of the ideal of definition. These sets are the neighbourhood basis of zero of a Huber ring.

          Equations
          Instances For
            @[simp]

            As a set, Iⁿ's image in A is the image of Iⁿ ⊆ A₀ under the inclusion A₀ → A.

            @[simp]
            theorem TauCeti.Huber.PairOfDefinition.mem_idealImage {A : Type u_1} [CommRing A] [TopologicalSpace A] (P : PairOfDefinition A) (n : ℕ) {x : A} :
            x ∈ P.idealImage n ↔ ∃ y ∈ P.idealOfDefinition ^ n, ↑y = x

            Membership in the image of Iⁿ.

            The images of the powers of the ideal of definition are nested.

            theorem TauCeti.Huber.PairOfDefinition.mul_mem_idealImage {A : Type u_1} [CommRing A] [TopologicalSpace A] (P : PairOfDefinition A) {n : ℕ} {a x : A} (ha : a ∈ P.ringOfDefinition) (hx : x ∈ P.idealImage n) :
            a * x ∈ P.idealImage n

            Iⁿ ⊆ A absorbs multiplication by an element of the ring of definition.

            theorem TauCeti.Huber.PairOfDefinition.mul_mem_idealImage_add {A : Type u_1} [CommRing A] [TopologicalSpace A] (P : PairOfDefinition A) {a b : ℕ} {x y : A} (hx : x ∈ P.idealImage a) (hy : y ∈ P.idealImage b) :
            x * y ∈ P.idealImage (a + b)

            The images of the powers of the ideal of definition multiply: Iᵃ · Iᵇ ⊆ Iᵃ⁺ᵇ.

            theorem TauCeti.Huber.PairOfDefinition.exists_sum_eq_of_mem_idealImage_succ {A : Type u_1} [CommRing A] [TopologicalSpace A] (P : PairOfDefinition A) {G : Finset ↥P.ringOfDefinition} (hG : Ideal.span ↑G = P.idealOfDefinition) (n : ℕ) {x : A} (hx : x ∈ P.idealImage (n + 1)) :
            ∃ (c : ↥P.ringOfDefinition → A), (∀ (z : ↥P.ringOfDefinition), c z ∈ P.idealImage n) ∧ ∑ z ∈ G, ↑z * c z = x

            One coefficient, decomposed. An element of Iⁿ⁺¹ ⊆ A is a combination of a finite generating set G of I with cofactors in Iⁿ.

            The ideal I · A of A generated by the ideal of definition I ⊆ A₀. This is core data of a pair of definition; TauCeti/RingTheory/Huber/OpenIdeal.lean characterises the open ideals of A in terms of its powers.

            Equations
            Instances For
              @[simp]

              Membership in I · A is membership in the ideal spanned by the image of I.

              There is no simpler characterisation: the image of I under A₀ → A is an additive subgroup but not in general an ideal of A, so x ∈ I · A is strictly weaker than ∃ y ∈ I, ↑y = x. For A₀ = ℤ_[p] ⊆ A = ℚ_[p] and I = p • ℤ_[p] the image is p • ℤ_[p] while the ideal it generates is all of ℚ_[p].

              The extended ideal I · A is finitely generated, because I is.

              A finite generating set of the ideal of definition, mapped into A, generates the extended ideal of definition.

              theorem TauCeti.Huber.PairOfDefinition.hasBasis_nhds_zero {A : Type u_1} [CommRing A] [TopologicalSpace A] (P : PairOfDefinition A) :
              (nhds 0).HasBasis (fun (x : ℕ) => True) fun (n : ℕ) => ↑(P.idealImage n)

              Wedhorn Proposition and Definition 6.1: the images in A of the powers of the ideal of definition are a neighbourhood basis of zero.

              If two adjacent powers of an ideal of definition agree, then the closure of zero in the ambient Huber ring is open. The stable image of that power is both a basic open neighbourhood and the intersection of all basic neighbourhoods.

              A power of the ideal of definition multiplies any element into any open subring. Multiplication by x is continuous, so the preimage of B is a neighbourhood of 0, and the images of the powers of I are a neighbourhood basis there (TauCeti.Huber.PairOfDefinition.hasBasis_nhds_zero).

              theorem TauCeti.Huber.PairOfDefinition.exists_idealImage_subset_image_mul_left {A : Type u_1} [CommRing A] [TopologicalSpace A] [IsTopologicalRing A] (P : PairOfDefinition A) {u : A} (hu : IsUnit u) (N : ℕ) :
              ∃ (N' : ℕ), ↑(P.idealImage N') ⊆ (fun (x : A) => u * x) '' ↑(P.idealImage N)

              A unit multiple of one level of the filtration contains a deeper level: for a unit u of A and any N, some Iᴺ' lands inside u · Iᴺ.

              This is what lets a denominator be rescaled by a unit: the standing hypotheses of a presentation are stated at some level of the filtration, and rescaling moves that level by a unit.

              An element of the ideal of definition is topologically nilpotent. Its powers lie in the successive Iⁿ, whose images are a neighbourhood basis of zero (TauCeti.Huber.PairOfDefinition.hasBasis_nhds_zero), so they converge to 0 in A.

              This is the property Wedhorn uses throughout §7.2: it is what forces a continuous valuation to have cofinal values on I, and hence v a < 1 there (Theorem 7.10).

              An element of the image of Iⁿ is topologically nilpotent, for n ≠ 0. Unpacking the membership gives an element of Iⁿ ⊆ I, so isTopologicallyNilpotent_of_mem_idealOfDefinition applies. This is the form consumers meet, idealImage being what TauCeti.Huber.PairOfDefinition.hasBasis_nhds_zero is stated with.

              n ≠ 0 is needed, not incidental: I ^ 0 = ⊤, so idealImage 0 is the image of the whole ring of definition and its elements are not topologically nilpotent in general.

              A ring admitting a pair of definition is nonarchimedean.

              Wedhorn Corollary 6.4: a ring of definition is bounded.

              A ring of definition consists of power-bounded elements: A₀ ≤ A°. The nonarchimedean hypothesis is only needed to state it, since P itself supplies one.

              Some power of a topologically nilpotent s carries any c : A into the ring of definition. The ring of definition is open and sⁿ c → 0, so sⁿ c is eventually inside it.

              This is the arbitrary-c generalisation of TauCeti.Huber.IsPseudoUniformizer.eventually_pow_mem_ringOfDefinition, which is the case c = 1; it also asks only for topological nilpotence rather than for a pseudouniformiser.

              A topologically nilpotent element rescales any element to a topologically nilpotent one. In a Huber ring, for topologically nilpotent ϖ and any s, some ϖ ^ i * s is topologically nilpotent.

              Use it to make a hypothesis of topological nilpotence available for an element that need not have it: replace s by ϖ ^ i * s, which differs from it by a factor drawn from the topology rather than an arbitrary one. When ϖ is a pseudouniformiser that factor is a unit, so the replacement is an associate of s; TauCeti.Huber.IsTateRing.exists_isTopologicallyNilpotent_pow_mul is the form that records this.

              A pseudouniformiser rescales any element to a topologically nilpotent one. For s : A in a Tate ring there are a pseudouniformiser ϖ and an exponent i with ϖ ^ i * s topologically nilpotent.

              ϖ ^ i is a unit, so s and ϖ ^ i * s are associated. Any construction depending on an element only up to associates is therefore unchanged by the rescaling, while hypotheses asking topological nilpotence of that element become available; IsLocalization.Away is the case that matters, by IsLocalization.Away.of_associated.

              Rescaling a denominator alone is not such a construction: the fractions t / s are not the fractions t / (ϖ ^ i * s), so a rational localisation A⟨T/s⟩ is preserved only if the numerators are rescaled by the same unit.

              A caller who already holds a topologically nilpotent element should use IsTopologicallyNilpotent.exists_pow_mul, which neither asks for a Tate ring nor chooses the multiplier.

              Quotients of Huber rings, with the quotient topology, are Huber rings.

              The pair of definition of a discrete ring: the whole ring, with the zero ideal.

              Equations
              Instances For
                @[instance 100]

                A discrete ring is Huber, with (A, 0) as a pair of definition. This is the first of the roadmap's Layer-0 examples, and the witness that IsHuberRing is not vacuous.

                @[instance 100]

                Every Huber ring is nonarchimedean. This is what makes A° a subring.

                The neighbourhoods of zero in a Huber ring are countably generated. The images of the powers of an ideal of definition are a basis of 𝓝 0 indexed by ℕ (TauCeti.Huber.PairOfDefinition.hasBasis_nhds_zero), and a basis indexed by a countable type generates a countably generated filter.

                This is one of the two hypotheses Henkel's open mapping theorem asks of a topological group, the other being TauCeti.Huber.IsHuberRing.toNonarchimedeanRing above: nonarchimedean makes the open subgroups a basis at zero, and countable generation extracts an antitone sequence from that basis (NonarchimedeanAddGroup.exists_antitone_basis_openAddSubgroup). Being an instance is the point — it is what lets a Huber ring be handed to that theorem without the caller discharging anything.

                Wedhorn Corollary 6.4: the power-bounded subring of a Huber ring is open.

                Wedhorn: sufficiently high powers of a pseudouniformiser lie in a given ring of definition.

                The scaled copy ϖⁿ A₀ of a ring of definition is a neighbourhood of zero.

                Only continuity of multiplication by a constant is needed: multiplication by the unit ϖⁿ is a homeomorphism, so it carries the neighbourhood A₀ of zero to a neighbourhood of zero.

                Wedhorn: in a Tate ring the sets ϖⁿ A₀ are a neighbourhood basis of zero, for any pseudouniformiser ϖ and any ring of definition A₀.

                Cofinality is the boundedness of A₀ (PairOfDefinition.isBounded_ringOfDefinition) together with ϖⁿ → 0; that each ϖⁿ A₀ is itself a neighbourhood is smul_ringOfDefinition_mem_nhds_zero.

                In a Tate ring one may choose a pseudouniformiser ϖ and a ring of definition A₀ whose scaled copies ϖⁿ A₀ are a neighbourhood basis of zero.