riemann

The Lean code claims a proof of the Riemann Hypothesis. This site has not checked the claim yet. See what is checked.

The Galois row

The theorem everyCharacter_of_row needs one hypothesis, Row0b with the local notion IsRecT. This page says what that hypothesis states and how the Lean code proves it, in three large parts.

draft

Result

The top of the proof is the theorem everyCharacter (see the map of the trunk). It is one line:

theorem UpperTheory.FinalTheorems.everyCharacter_of_row
    (h : UpperTheory.GalStar.Row0b fun h rc => UpperTheory.Zero.LLC.IsRecT h rc) :
    UpperTheory.EveryCharacter := ...

theorem UpperTheory.FinalTheorems.everyCharacter : UpperTheory.EveryCharacter :=
  UpperTheory.FinalTheorems.everyCharacter_of_row UpperTheory.Gal.Int.galoisLocalGlobalStrong_recParameter

The hypothesis h is used twice in everyCharacter_of_row (MainConclusion.lean#L7): once as a field of the structure of literature statements, and once as the input of vwin_of_row. This page is about the proof of that hypothesis:

theorem galoisLocalGlobalStrong_recParameter : UpperTheory.GalStar.Row0b recT :=
  galoisRepresentations_of_thetaLift UpperTheory.TP.K.tLiftK

Here recT is fun h rc => UpperTheory.Zero.LLC.IsRecT h rc (Development.lean#L18). In standard notation, the Lean statement says this. Let EE be a CM field with maximal real subfield E+E^+. Let ℓ\ell be a prime, let ι:ℚ‾ℓ≅ℂ\iota : \overline{\mathbb{Q}}_\ell \cong \mathbb{C} and let a≥1a \geq 1. Let π\pi be a cuspidal automorphic representation of GLa(𝔸E)\mathrm{GL}_a(\mathbb{A}_E) that is regular algebraic, and let χ\chi be a Hecke character of E+E^+ such that (π,χ)(\pi, \chi) is a polarized pair. Then there is a continuous semisimple representation r:Gal(E‾/E)→GLa(ℚ‾ℓ)r : \mathrm{Gal}(\overline{E}/E) \to \mathrm{GL}_a(\overline{\mathbb{Q}}_\ell), with an integer wtwt and Hodge-Tate multisets HTτHT_\tau, that satisfy a list of local conditions at every finite place. The conditions compare rr at a place ww with the Weil-Deligne parameter rc(πw)rc(\pi_w), for each local map rcrc that satisfies IsRecT.

The whole branch has 51,591 declarations in its dominator tree. This is the second largest branch below the trunk.

Idea

The statement is a version of the existence of Galois representations attached to regular algebraic, conjugate self-dual cuspidal representations over CM fields, with local-global compatibility. The Lean names use words of that literature ("galois local global", "row"). This page says only what the Lean statements say.

The proof has a short top and three large parts. The top first proves a variant Row0 with a slightly stronger local clause, and then weakens it to Row0b by one structural lemma. Row0 itself comes from a theorem galoisLocalGlobal_of_localGlobalCompatibility. That theorem takes three inputs: CoreAll (existence of rr and its properties at places where πw\pi_w is spherical), R1 (compatibility at places where πw\pi_w has a support of characters with no links) and LocalGlobalCompatAtIwahoriFixedPlaces. The last two come from CoreAll again.

So almost all the work is in CoreAll. The Lean code gets CoreAll from four inputs. Two of them are large:

  1. GeoOne, a statement about Galois representations cut out of geometric towers (43,297 declarations). Its proof uses a holomorphic Lefschetz formula for finite étale correspondences of smooth projective schemes over ℂ\mathbb{C}.
  2. TLiftK, a theta lift statement between definite unitary groups (2,709 declarations).

The rest (about 5,400 declarations) assembles CoreAll, then R1 and the Iwahori clause, then Row0.

Before you read

Plan

  1. Say what IsRecT states about a map from local representations to Weil-Deligne classes.
  2. Say what Row0b states, and how it differs from Row0.
  3. Show the top of the proof: galoisLocalGlobalStrong_recParameter and galoisRepresentations_of_thetaLift.
  4. Show how Row0 comes from CoreAll (the rest step).
  5. Show how CoreAll comes from GeoOne, TLiftK and two other inputs (the coreAll_of step).
  6. Describe the geometric branch geoOne_holds.
  7. Describe the theta lift branch tLiftK.

The steps

Step 1. The local notion IsRecT

Fix a finite place ww of EE and a valuation subring AA of Ω\Omega above ww (the proof h : IsPlaceOver A w). The Lean type IrrGL (w.adicCompletion E) is the set of pairs (n,x)(n, x) where xx is an isomorphism class of irreducible smooth representations of GLn(Ew)\mathrm{GL}_n(E_w) (NonemptyEquivMultiProdOne.lean#L179). The type WD h is the set of isomorphism classes of finite-dimensional Weil-Deligne representations of the Weil group Weil h (MonodromyZeroOfWeilDeligneRepClassIff.lean#L380).

def IsRecT (rc : IrrGL (w.adicCompletion E) → WD h) : Prop := ∃ a ha, LawsT (h := h) a ha rc

So IsRecT h rc says: there is a map a:WEw→Ew×a : W_{E_w} \to E_w^\times with IsLocalArtin h a (a local Artin map; this page does not unfold IsLocalArtin), such that the twelve laws of LawsT hold (IsRecTOfIsRec.lean#L55). In words, with rcrc the map:

  • dim: rc(x)rc(x) has dimension nn when xx is a representation of GLn\mathrm{GL}_n.
  • one: on a character χ\chi of GL1(Ew)=Ew×\mathrm{GL}_1(E_w) = E_w^\times, rc(χ)rc(\chi) is the character χ∘a\chi \circ a.
  • langlands: if a list of segments is related to xx by IsL (by its name, a Langlands classification; this page does not unfold it), then rc(x)rc(x) is the sum segSum over the segments.
  • cusp: rcrc sends each xx with IsCusp to an irreducible Weil-Deligne class.
  • nu: rcrc commutes with the twist by ∣det⁡∣s|\det|^s.
  • absCentral: if xx has central character ω\omega, then ∣det⁡rc(x)(σ)∣=∣ω(a(σ))∣|\det rc(x)(\sigma)| = |\omega(a(\sigma))|.
  • twist, central, dual, inj, depth: for xx in a class Good (below), rcrc commutes with twists by characters, has determinant ω∘a\omega \circ a, sends the contragredient to the dual, is injective, and sends depth zero (a nonzero vector fixed by the first congruence subgroup KOne) to a tame class.
  • type: the law TypeLaw, a formula for the characteristic polynomial of inertia on rc(σ)rc(\sigma) when Ew≅ℚgE_w \cong \mathbb{Q}_g and σ\sigma inflates a cuspidal representation of GLn(𝔽gn)\mathrm{GL}_n(\mathbb{F}_{g^n}) given by a regular character θ\theta. The formula is ∏j<n(X−θ(zgj))\prod_{j<n} (X - \theta(z^{g^j})).

The class Good (IsRecTOfIsRec.lean#L24) holds when the supercuspidal support of xx has only characters, or (over ℚg\mathbb{Q}_g) twists of depth zero supercuspidals. So the laws for all of GLn\mathrm{GL}_n are only the first six. The other laws hold only on this subclass. The site does not claim that IsRecT is the local Langlands correspondence. It is a list of properties that the local Langlands correspondence has. A separate theorem recTExists gives a map with these properties (it is the second field passed in everyCharacter_of_row).

Step 2. The global statement Row0b

def Row0b : Prop :=
  ∀ (E Ω : Type) [Field E] [NumberField E] [IsCMField E] [Field Ω] [Algebra E Ω] [IsAlgClosure E Ω]
    (l : ℕ) [Fact l.Prime] (ι : Qbar l ≃+* ℂ) (a : ℕ), 1 ≤ a →
    ∀ (π : CuspSub E a) (χ : HChar (maximalRealSubfield E)), (CG.mk π).IsRegAlg → GlobalTheory.ContraPair.PolarizedPair π χ →
      ∃ (r : G E Ω →* GL (Fin a) (Qbar l)) (wt : ℤ) (HT : (E →+* Qbar l) → Multiset ℤ), PropsB Rc π ι r wt HT

The objects are these.

  • CuspSub E a is a nonzero subspace VV of cusp forms in the smooth functions on GLa(𝔸E)\mathrm{GL}_a(\mathbb{A}_E). It is stable under the action of GLa(𝔸E,f)\mathrm{GL}_a(\mathbb{A}_{E,f}), the maximal compact subgroup and the Lie algebra at infinity, and it is irreducible for that action (ExpoConjOfIsCsd.lean#L209).
  • HChar L is the group of continuous characters 𝔸L×→ℂ×\mathbb{A}_L^\times \to \mathbb{C}^\times that are trivial on L×L^\times.
  • IsRegAlg says: for each embedding τ:E→ℂ\tau : E \to \mathbb{C}, the infinitesimal character multiset of π\pi at τ\tau has distinct entries, and each entry is n+a−12n + \frac{a-1}{2} for an integer nn.
  • PolarizedPair π χ says: χv(−1)=(−1)a\chi_v(-1) = (-1)^a at each real place vv of E+E^+, and π∨\pi^\vee matches the complex conjugate of π\pi after a twist by χ∘NE/E+\chi \circ N_{E/E^+} (DualIsConjByOne.lean#L53).

The conclusion PropsB (WeightsConjBis.lean#L54) has six fields:

  • cont, ss: rr is continuous and semisimple (each invariant subspace has an invariant complement).
  • away: at each place ww not above ℓ\ell, for each rcrc with IsRecT, and each local component σ\sigma of π\pi at ww, the class wdOf h hl ι r (the Weil-Deligne class of rr at ww) and rc(σ⊗∣det⁡∣(1−a)/2)rc(\sigma \otimes |\det|^{(1-a)/2}) satisfy AwayB. AwayB x y says: equal dimension; x=yx = y if yy has N=0N = 0 and is trivial on commutators of the Weil group (Tor); and in a tame case, xx has N=0N = 0 and the same characteristic polynomials as yy on degree zero elements.
  • weights: HTτHT_\tau, read in ℂ\mathbb{C}, is {a−12−z}\{\frac{a-1}{2} - z\} over the infinitesimal character of π\pi at ι∘τ\iota \circ \tau.
  • weights_conj: HTτ∘c={wt−h:h∈HTτ}HT_{\tau \circ c} = \{wt - h : h \in HT_\tau\}, with cc the complex conjugation of EE.
  • above: at places above ℓ\ell, crystalline, de Rham and Hodge-Tate jump conditions when σ\sigma is spherical, and de Rham, potentially semistable and AwayB with rc(σ)rc(\sigma) in the cases of AboveB.nice.

Row0 (WeightsConjAlt.lean#L85) is the same statement with Props and Away in place of PropsB and AwayB. The only change is in the tame clause. Away asks for it when Ew≅ℚgE_w \cong \mathbb{Q}_g (IsQp). AwayB asks for it only when Ew≅ℚgE_w \cong \mathbb{Q}_g and also n=1n = 1 or 16n2<g16 n^2 < g (IsQpBig). So Row0b is weaker than Row0, and the theorem Row0.row0b (WeightsConjBis.lean#L84) proves Row0 Rc → Row0b Rc by IsQpBig.isQp.

Step 3. The top of the proof

theorem galoisRepresentations_of_thetaLift (up : UpperTheory.TP.TLiftK) : UpperTheory.GalStar.Row0b recT :=
  galoisLocalGlobalStrong_of_geometry_lift UpperTheory.Gal.Int.Geo.geoOne_holds up

theorem galoisLocalGlobalStrong_of_geometry_lift (geo : GeoOne) (up : UpperTheory.TP.TLiftK) : Row0b recT :=
  UpperTheory.GalStar.Row0.row0b recT (galoisLocalGlobal_of_geometry_lift geo up)

theorem galoisLocalGlobal_of_geometry_lift (geo : GeoOne) (up : UpperTheory.TP.TLiftK) : Row0 recT :=
  rest (coreAll_of (UpperTheory.Gal.Int.Div.seedFinOdd_of_geoOne geo)
    (UpperTheory.Gal.Int.Cong.congFinOdd_of_ex UpperTheory.Gal.Wt.wtLacEx_holds) up)

So the direct uses of galoisLocalGlobalStrong_recParameter are galoisRepresentations_of_thetaLift (48,870 declarations below it) and the theta lift theorem UpperTheory.TP.K.tLiftK (2,709). The first one feeds the proof geoOne_holds of GeoOne into a conditional theorem. It is a plain chain of implications. GeoOne enters at one point only: seedFinOdd_of_geoOne turns it into SeedFinOdd. TLiftK enters at one point only, inside coreAll_of.

Step 4. From CoreAll to Row0 (the step rest)

theorem rest (h : GlobalTheory.RoadU.CoreAll) : UpperTheory.GalStar.Row0 recT :=
  galoisLocalGlobal_of_localGlobalCompatibility h (galoisCompat_unlinkedSupport_of_spherical h)
    (localGlobalCompatAtIwahoriFixed_of_unlinked (galoisCompat_unlinkedSupport_of_spherical h))

CoreAll (CyclicStrongAtCusp.lean#L96) is Ex .all ∧ ∀ c, SphericalLevelCondition .all c. In words: for each (E,ℓ,ι,a,π,χ)(E, \ell, \iota, a, \pi, \chi) as in Row0b, there is a continuous rr that is attached to π\pi (Weak: rr is semisimple, and at all but finitely many places ww there are Satake parameters BB of π\pi such that rr is trivial on inertia and the characteristic polynomial of r(Frobw−1)r(\mathrm{Frob}_w^{-1}), through ι\iota, is ∏(X−qw(a−1)/2b)\prod (X - q_w^{(a-1)/2} b) over b∈Bb \in B). And for each such rr, at each place ww where πw\pi_w has Satake parameters BB, four clauses hold: unramified with matching Frobenius eigenvalues (away from ℓ\ell), de Rham with the right Hodge-Tate jumps, crystalline, and a Frobenius characteristic polynomial equal to the one of the unramified class built from BB (above ℓ\ell).

R1 (IsQg.lean#L13) says: when σ⊗∣det⁡∣(1−a)/2\sigma \otimes |\det|^{(1-a)/2} has support of characters with no two linked by ∣⋅∣|\cdot| (PlainTor), the class of rr at ww equals rcrc of it for each rcrc with IsRecT (away from ℓ\ell), and the analogous statement for the class wdAt above ℓ\ell. LocalGlobalCompatAtIwahoriFixedPlaces (LocalGlobalCompatAtIwahoriFixedPlaces.lean#L16) gives N=0N = 0 and equal inertia characteristic polynomials when σ\sigma has a KOne-fixed vector and either PlainTor holds or Ew≅ℚgE_w \cong \mathbb{Q}_g with a separation condition Apart.

The proof of galoisLocalGlobal_of_localGlobalCompatibility (GaloisLocalGlobalOfLocalGlobalCompatibility.lean#L192) takes rr from CoreAll.1, takes wtwt and HTHT from a lemma exists_weights, and fills each field of Props from the three inputs. The sizes are: localGlobalCompatAtIwahoriFixed_of_unlinked 1,522, galoisCompat_unlinkedSupport_of_spherical 238, and the assembly 116. The step localGlobalCompatAtIwahoriFixed_of_unlinked itself uses UpperTheory.Zero.LLC.recTExists.

Step 5. CoreAll from four inputs (the step coreAll_of)

theorem coreAll_of (seed : SeedFinOdd) (cong : CongFinOdd) (up : UpperTheory.TP.TLiftK) : GlobalTheory.RoadU.CoreAll :=
  UpperTheory.Gal.Int.Inp.coreAll_of_core
    (UpperTheory.Gal.Int.Core.toCoreK up UpperTheory.Gal.ClosedPhiFromClosedTrace.congPhi_holds
      UpperTheory.Gal.Desc.descOffT (UpperTheory.Gal.Lim.limFinOdd_of seed cong))

The Lean code first proves Core, the same as CoreAll but restricted to a smaller scope (Level.csd: EE has an element xx with x2=−dx^2 = -d, [E+:ℚ][E^+ : \mathbb{Q}] even, conditions at non-split places, χ\chi a fixed power of a quadratic character, and only split places). toCoreK gives Core from TLiftK, ClosedPhi, DescOffT and LimFinOdd. Then coreAll_of_core extends Core to CoreAll. LimFinOdd comes from SeedFinOdd and CongFinOdd. Sizes: coreAll_of 1,760 in total, toCoreK 496, seedFinOdd_of_geoOne 718, congFinOdd_of_ex 495, wtLacEx_holds 183. This page does not unfold SeedFinOdd, CongFinOdd, LimFinOdd, ClosedPhi and DescOffT. They are defined in LimWU4OfLimWU3.lean#L103 and nearby files.

Step 6. The geometric branch geoOne_holds

This is the largest part (43,297 declarations).

theorem weight_one_geometric_realization : LocalSmoothExtension.HolomorphicLefschetzFormula := by
  unfold LocalSmoothExtension.HolomorphicLefschetzFormula
  exact GlobalTheory.T.lefschetz_fixedPoint_finite_etale_correspondence

theorem geoOne_holds : GeoOne := geometric_realization_of_weight_one weight_one_geometric_realization

theorem geometric_realization_of_weight_one (hW : HolomorphicLefschetzFormula) : GeoOne :=
  geoOne_of_parts steps_holds (kodaira_spencer_of_weight_one hW) Pl.planEx_holds

GeoOne (HodgeTrAt.lean#L79) is a statement about an odd aa, an integer dd and x∈Ex \in E with x2=−dx^2 = -d, places w0,v′w_0, v' above ℓ\ell, finite sets of places T,ST, S, and a space UU of automorphic functions on the unitary group Ua(𝔸E)U_a(\mathbb{A}_E) with conditions at those places. It concludes that there are Galois representations ρ\rho with VBase ι d U ρ, with matching Frobenius eigenvalues at good places away from ℓ\ell, and with crystalline, Hodge-Tate jump and Frobenius conditions at places above ℓ\ell. This page does not unfold VBase, SatU, AutOffT, TyEll and SphOffU.

HolomorphicLefschetzFormula (HodgeTraceFromLefschetzAndGroupSide.lean#L102) states: let ZZ be smooth of relative dimension dd and projective over ℂ\mathbb{C}, let c1,c2:Y→Zc_1, c_2 : Y \to Z be finite étale, let FF be the finite set of complex points yy with c1(y)=c2(y)c_1(y) = c_2(y), and let AyA_y be a d×dd \times d matrix with IsTan and det⁡(1−Ay)≠0\det(1 - A_y) \neq 0. Then for each aa,

∑b=0d(−1)btr⁡(c1∗∘c2∗∣Ha,b)=∑y∈F[Xa]det⁡(1+XAy)det⁡(1−Ay),\sum_{b=0}^{d} (-1)^b \operatorname{tr}\big(c_{1*} \circ c_2^* \mid H^{a,b}\big) = \sum_{y \in F} \frac{[X^a] \det(1 + X A_y)}{\det(1 - A_y)} ,

where Ha,bH^{a,b} is the Lean object hodgeH s a b and the spaces are finite-dimensional. This has the shape of the holomorphic Lefschetz fixed-point formula. This page does not unfold hodgeH, hodgeTrace, hodgePull and IsTan. The proof is UpperTheory.Cb.lefschetz_fixedPoint_finite_etale_correspondence with 1,812 declarations below it.

The three parts of geoOne_of_parts are:

  • steps_holds : Steps (26,256). Steps is a structure of seventeen fields built by steps_of. Three are the main local steps: CrysStep (a piece of the cohomology of a tower is crystalline at places above ℓ\ell where a cover is onto), JumpStep (Hodge-Tate jumps occur where the Hodge numbers say) and PhiStep (the Frobenius characteristic polynomial divides a power of a root polynomial). The largest children are crysStep_holds (4,188), jumpStep_holds (2,709), a projective bundle formula (1,595), UpperTheory.Et.proper_base_change (1,247) and phiStep (1,031).
  • kodaira_spencer_of_weight_one hW : KSAll (11,907). KSAll has six fields: StdEx, KSEx, KSNice, KSGood, KSCong, KSHodge. The Lefschetz formula enters only in the field KSHodge, through ksHodge_of_geo hW FixK_holds. The largest children are model_nonempty (2,597), ksHodge_of_geo (1,902) and FixK_holds (1,789).
  • Pl.planEx_holds : PlanEx (1,934). PlanEx has the same hypotheses as GeoOne and gives "plans" with Hodge-Tate and congruence properties. Plan.pack turns a plan into a Pack by KSAll.

Step 7. The theta lift branch tLiftK

theorem UpperTheory.TP.K.tLiftK : UpperTheory.TP.TLiftK := UpperTheory.TP.tLift_holds_k

TLiftK (TLiftK.lean#L12) says, in words: let EE be a CM field with a good base B. Let XX be a hermitian lattice space of rank nn with Gram matrix J=1J = 1, positive definite at each infinite place of E+E^+. Then there is d0d_0 such that for each d≥d0d \geq d_0, each skew-hermitian space WW of rank dd with J=δ⋅1J = \delta \cdot 1, definite in the same sense, each pair of Hecke characters χX,χW\chi_X, \chi_W with conditions ResEtaPow and exponents mX,mWm_X, m_W, and each automorphic representation UU of Un(𝔸E)U_n(\mathbb{A}_E) with dominant weights λw\lambda_w that satisfy d+mW,w≤2λw,kd + m_{W,w} \leq 2 \lambda_{w,k}, there is an automorphic representation U′U' of Ud(𝔸E)U_d(\mathbb{A}_E) with IsLift. IsLift (SchurStep.lean#L106) asks three things: U′U' is automorphic, its weights λ′\lambda' satisfy

2λk′={2λk−d−mW+n+mXk<n,n+mXk≥n,2\lambda'_k = \begin{cases} 2\lambda_k - d - m_W + n + m_X & k < n, \\ n + m_X & k \geq n, \end{cases}

and at each split place the local components are related by a nonzero map from a Schwartz-Bruhat space that intertwines the two actions (LocPair). The Lean name points to the theta lift. The statement is an existence statement of this form; this page does not claim it is a published theorem.

The proof tLift_holds_k is tLift_of_tools_k tools_holds_k (hgtEx_of_k stab2Cpt_k). The two parts are tools_holds_k (2,169, with siegelWeilStd_k 1,680, a Siegel-Weil statement by its name, and weilEx_k 395) and tLift_of_tools_k (503, with realPair_k 140). The chapter block determinants and theta Parseval follows tLift_of_tools_k in detail. The chapter the theta matching treats theta lifts in more detail.

Where it is in the code

Declaration Module Size
galoisLocalGlobalStrong_recParameter …IntegralStructures.RowSrc 51,591
galoisRepresentations_of_thetaLift …IntegralStructures.GaloisRepresentationsOfThetaLift 48,870
Geo.geoOne_holds …GeometricFlagBases.WeightOneGeometricRealization 43,297
Geo.geometric_realization_of_weight_one …GeometricFlagBases.GeometricRealizationOfWeightOne 41,363
Geo.steps_holds same 26,256
Geo.kodaira_spencer_of_weight_one same 11,907
Geo.Pl.planEx_holds …PrimeLevelLocalization.SpecTopHolds 1,934
Geo.weight_one_geometric_realization …WeightOneGeometricRealization 1,814
galoisLocalGlobalStrong_of_geometry_lift …IntegralStructures.Development 5,415
galoisLocalGlobal_of_geometry_lift same 5,414
rest same 1,944
coreAll_of same 1,760
Div.seedFinOdd_of_geoOne …ZetaCharacterDivisibility.MainFin 718
Cong.congFinOdd_of_ex …CongruenceOddLevel.CongFinOddOfLvLac 495
UpperTheory.TP.K.tLiftK …ThetaParseval.ThetaKernels.TLiftK 2,709
UpperTheory.TP.tLift_of_tools_k …ThetaKernels.TPGTop 503
UpperTheory.TP.tools_holds_k …ThetaKernels.TPGHolds 2,169

The prefix … is Development.AutomorphicArithmetic.Foundations.GaloisRepresentations. Names without a namespace are in UpperTheory.Gal.Int. "Size" is the number of declarations in the dominator subtree: the declarations that the trunk reaches only through this one.

Notes

  • IsRecT is a hypothesis inside Row0b. The local clauses of Row0b hold for each rcrc with IsRecT. If no such rcrc existed, those clauses would hold for no rcrc and say nothing. The Lean code closes this gap with a separate theorem recTExists, which everyCharacter_of_row also uses. This page does not describe that theorem.
  • wdOf and infChar are defined by choice. wdOf returns the trivial class if no Weil-Deligne class of rr at ww exists. infChar returns aa copies of 00 if no infinitesimal character exists. A reader who checks the statement must check that the hypotheses force the intended values.
  • The difference between Row0 and Row0b is the bound 16n2<g16 n^2 < g in IsQpBig. The proof goes through the stronger Row0, so the bound is not used in this branch. The other users of Row0b may need only the weaker form.
  • GeoOne is very large (43,297 declarations), but the Lean code uses it at one point only, to prove SeedFinOdd. This page does not unfold SeedFinOdd, so it does not say why that statement needs the geometry.
  • This page does not unfold IsLocalArtin, segSum, IsL, IsCrysAt, IsdRAt, IsPstAt, IsJumpsAt, wdAt, LocBridge, VBase, hodgeH, ResEtaPow, IsAutRepU and LocPair. Their names suggest standard notions, but the words above use those notions only where the code shows them.
  • The theorem galoisLocalGlobalStrong_recParameter has one other user, UpperTheory.V.SmallClassNumbers.litN. This page does not describe it.

The source of this page. Markdown. Something is wrong? Report it.