riemann

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

Named results

The Lean code gives some of its inputs the names of published theorems. This page lists the ones that the top of the proof uses. For each one it tells what the Lean statement says, whether the proof reads the hypothesis structure, and where the result is in the literature.

Most of these theorems do not read their argument, the structure OpenLiteratureStatements. The code proves them from other declarations, so the name says where the idea comes from, not that the code assumes the result. The column "reads the hypothesis" tells which ones use the argument. The citation is the work in the literature that the name points to. A citation with the mark not verified is still to be checked.

Angehrn-Siu point separation for adjoint bundles

Lean: UpperTheory.LiteratureTheorems.point_separation_pair_of_literature, which proves GlobalTheory.Published.PointSeparationPair. The proof does not read the hypothesis structure.

Let XX be a normal projective complex variety of dimension nn with a klt boundary Δ\Delta. Let MM be a line bundle numerically equal to G/m+N0/rG/m + N_0/r, where GG is a log canonical multiple and N0N_0 is nef and big. Assume that every subvariety of dimension ee through x1x_1 or x2x_2 has N0N_0-degree large compared with ceec_e^e, and that the sum of 21/ee/ce2^{1/e} e / c_e is at most 11. Then some section of MM vanishes at exactly one of the two distinct closed points x1x_1 and x2x_2.

This is a variant of the Angehrn-Siu theorem, written with a ℚ\mathbb{Q}-divisor Δ\Delta and numerical constants cec_e. The exact constants were not compared with the published statement. The proof is a direct term and does not use O.

Arithmetic Hilbert-Samuel theorem (upper bound) and arithmetic nonnegativity for nef bundles

Lean: PositivityTheory.Pf1.HilbertSamuelMoriwaki.hsUse_outright, which proves PositivityTheory.Pf1.HilbertSamuelMoriwaki.HSUse. The proof does not read the hypothesis structure.

HSUse is three statements about arithmetic subvarieties of a product of projective spaces over ℤ\mathbb{Z}. ArithKleiman says a nef hermitian line bundle with nonnegative heights on arithmetic curves has nonnegative self-intersection. G7 and ArithmeticVolumeNef say that the number of small sections of tensor powers grows at most like the arithmetic self-intersection times nD+1/(D+1)!n^{D+1}/(D+1)!, as a lim sup⁡\limsup bound.

The Lean form has only the upper bound (lim sup⁡≤\limsup \leq), not the asymptotic equality of the published theorem. Intersection numbers come from an abstract IntersectionSystem that satisfies the axioms IsIntersectionSystem. They are not the Gillet-Soulé numbers, so the result depends on those axioms. This entry takes no O argument. It is proved from resK, the Poincaré-Lelong result and DenseCurves.

Asymptotic Riemann-Roch for nef divisors

Lean: UpperTheory.LiteratureTheorems.nef_rr_of_literature, which proves GlobalTheory.CompanionTheory.Ar.Pub.NefRR. The proof does not read the hypothesis structure.

Let XX be a variety of dimension DD in a product of projective spaces, and let aa be a multidegree that is nef on the generic fibre. Then the spaces of sections of 𝒪(τa)\mathcal{O}(\tau a) are finite-dimensional over ℚ\mathbb{Q}. Also D!D! times their dimension divided by τD\tau^D tends to the degree of aa on XX.

In Lazarsfeld this is Corollary 1.4.41 (the corollary number is from memory and was not checked). In Lean, sections are modelled as the ℚ\mathbb{Q}-span of a lattice of elements integral over the chart rings, so the model is the normalisation of the generic fibre. The proof is a direct term and does not use O.

Birational invariance of arithmetic self-intersection, and the Chern-Weil degree formula

Lean: UpperTheory.LiteratureTheorems.ar_degree_of_literature, which proves GlobalTheory.Published.ArDegree. The proof does not read the hypothesis structure.

ArDegree is the conjunction of ArBirational and ChernDegree. ArBirational says that the arithmetic self-intersection number of a hermitian line bundle does not change under pullback by a birational morphism of generically smooth arithmetic varieties. ChernDegree says that for a smooth metric on a line bundle given by a divisor, the integral of the top power of its Chern form equals the degree of the divisor.

This is two standard facts in one statement, not one named theorem. The first part is a case of the projection formula of Gillet-Soulé. The proof does not use O. It uses the in-repository Poincaré-Lelong result and resolution.

Exactness of the Hodge-Tate sequence for p-divisible groups over the ring of integers of an algebraically closed p-adic field (Fargues)

Lean: UpperTheory.LiteratureTheorems.fargues_exact_of_literature, which proves GlobalTheory.CompanionTheory.Ar.HT.FarguesExact. The proof does not read the hypothesis structure.

Let GG be a pp-divisible group over the ring of integers 𝒪\mathcal{O} of a complete algebraically closed field CC of characteristic zero. Let BB be the 𝒪\mathcal{O}-linear extension of the Weil pairing between the Tate modules of GG and its dual. Then an element xx of 𝒪⊗TpG\mathcal{O} \otimes T_p G is in the kernel of the Hodge-Tate map if and only if B(x,y)=0B(x, y) = 0 for all yy in the kernel of the dlog⁡\operatorname{dlog} map of the dual.

In Math 2.0: p-adic Hodge theory.

The Lean form is an orthogonality statement about kernels. It is not the exact sequence itself. Over 𝒪C\mathcal{O}_C the integral Hodge-Tate sequence is exact only up to torsion killed by p1/(p−1)p^{1/(p-1)}. Check that the integral if-and-only-if in Lean is true as stated. The exact section of Fargues' chapter was not checked. The proof is a direct term and does not use O.

Faltings-Chai toroidal compactification of the Siegel moduli space over the integers with the n-th roots of unity and 1/n

Lean: UpperTheory.LiteratureTheorems.siegel_toroidal_integral_of_literature, which proves GlobalTheory.Published.SiegelToroidalIntegral. The proof does not read the hypothesis structure.

For each g≥1g \geq 1 there is a polarised fan with the following property. For each odd n≥3n \geq 3 the moduli space of principally polarised abelian schemes with level nn has a compactification SS, which is projective and smooth over ℤ[ζn,1/n]\mathbb{Z}[\zeta_n, 1/n] of relative dimension g(g+1)/2g(g+1)/2. The open part is dense in every fibre. The universal abelian scheme extends to a semi-abelian scheme, and the boundary has toroidal charts. A power of the Hodge bundle is generated by global sections, and large powers of it twisted by minus the boundary are ample.

The Lean form asks only for odd n≥3n \geq 3. The published theorem covers all n≥3n \geq 3. The relevant results are in Faltings-Chai Chapters IV and V. The theorem numbers were not checked. The proof is a direct term and does not use O.

Galois representations attached to regular algebraic polarised cuspidal automorphic representations of GL(n) over CM fields, with local-global compatibility

Lean: UpperTheory.LiteratureTheorems.OpenImageCondition, which proves structure with one field galoisRepOfPolarizedPairLit : UpperTheory.V.Z.Win.RowT. The proof does not read the hypothesis structure.

For each regular algebraic cuspidal automorphic representation π\pi of GLa\mathrm{GL}_a over a CM field that is part of a polarised pair, and each prime ℓ\ell, there is a continuous semisimple ℓ\ell-adic Galois representation. It has the expected Hodge-Tate weights, and they are symmetric under complex conjugation. Away from ℓ\ell it is compatible with the local Langlands map rcrc, and at places above ℓ\ell it has the expected pp-adic Hodge theoretic behaviour.

In Math 2.0: Galois representations attached to modular and automorphic forms.

This is a hypothesis structure, not a theorem. On the trunk, vwin_of_row builds it as OpenImageCondition.mk h, where h is the hypothesis Row0b, and the Lean code proves Row0b with galoisLocalGlobalStrong_recParameter. So the proof on the trunk does not assume it. The name is misleading, because the field states the existence of Galois representations, not an open image condition. The compatibility is stated for the map rcrc given by IsRecT, so it depends on the Lean local Langlands statement. The 'ℓ=p\ell = p, I' paper of the same authors was not checked.

Grothendieck-Messing deformation theory of abelian schemes

Lean: UpperTheory.LiteratureTheorems.grothendieck_deformation_of_literature, which proves GlobalTheory.Published.GrothendieckDeformation. The proof does not read the hypothesis structure.

Let R→R0R \to R_0 be a surjection of rings whose kernel II has nilpotent divided powers. For abelian schemes over R0R_0 with a lift to RR, the crystal gives a canonical identification with de Rham cohomology. Lifts of an abelian scheme correspond to admissible lifts of its Hodge filtration. A homomorphism lifts, and lifts uniquely, when it respects the lifted filtrations.

In Math 2.0: Deformation theory.

The proof passes O to abelian_de_rham_of_literature and abelian_lifting_of_literature. Both ignore O, so O is used in name only. The Lean form has no hypothesis that pp is nilpotent in RR. The derived GrothendieckMessing has that hypothesis. So the Lean form is the general form for abelian schemes, as in Mazur-Messing. It is not Messing's pp-nilpotent setting. Check that the published results cover this generality.

Hironaka's resolution of singularities in characteristic zero

Lean: UpperTheory.LiteratureTheorems.resolution_char_zero_of_literature, which proves GlobalTheory.CompanionTheory.Hypotheses₂.ResolutionCharZero. The proof does not read the hypothesis structure.

Let XX be a reduced scheme of finite type, quasi-compact and separated over a field of characteristic zero. Then a finite sequence of blow-ups gives a smooth X′→XX' \to X. This map is an isomorphism over the smooth locus of XX, and the reduced preimage of the singular locus is a simple normal crossings divisor.

In Math 2.0: Resolution of singularities.

The statement matches the published theorem. The proof is a direct term and does not use O.

Local Langlands correspondence for GL(n) over p-adic fields (Harris-Taylor, Henniart)

Lean: UpperTheory.LiteratureTheorems.local_langlands_gl_recT_exists_of_literature, which proves UpperTheory.Zero.LLC.RecTExists. The proof reads the hypothesis structure.

At each finite place ww of a number field there is a map rcrc from irreducible admissible representations of GLn(Ew)\mathrm{GL}_n(E_w) to Weil-Deligne representations. The map keeps dimension. It agrees with local class field theory for n=1n = 1 and is compatible with Langlands classification by segments. It sends cuspidals to irreducibles, respects twists, central characters and contragredients, and is injective on a class of good representations. It sends depth-zero representations to tame ones and follows a law for depth-zero types.

In Math 2.0: Local Langlands conjectures.

The proof is llOf_alt applied to O.localLanglandsExists, and llOf_alt is fun h _ ... _ => h. So the theorem returns the field localLanglandsExists of O without change. On the trunk, everyCharacter_of_row fills this field with the theorem recTExists, which the Lean code proves. So the proof on the trunk does not assume local Langlands. The Lean statement is weaker than the published theorem. It states existence only, with no uniqueness and no surjectivity. It does not match LL-factors or ε\varepsilon-factors of pairs, and injectivity holds only on good representations.

Moriwaki's bigness / volume of nef hermitian line bundles (with Yuan)

Lean: PositivityTheory.Pf1.HilbertSamuelMoriwaki.moriwakiBig_outright, which proves ReciprocityTheory.ChartRestriction.G.ArakelovGeometry.MoriwakiBigStatement. The proof does not read the hypothesis structure.

Let AA be a hermitian line bundle on an arithmetic variety of dimension D+1D+1 in a product of projective spaces. Assume AA is semipositive, conjugation-invariant, nef on fibres and nef on the generic fibre. Then its arithmetic self-intersection number is at most its arithmetic volume.

The Lean form is the inequality self-intersection ≤\leq volume. For nef bundles the published result gives equality, so the Lean form is weaker. Positive self-intersection then implies bigness. As in hsUse_outright, the intersection numbers come from an abstract IntersectionSystem. This entry takes no O argument.

Not identified (theta integrals for unitary dual pairs, with Siegel-Weil inputs)

Lean: UpperTheory.LiteratureTheorems.theta_integrals_agree_and_local_matching_of_literature, which proves GlobalTheory.IndefiniteRowSideFullHecke.WeilTensorModels.ThetaIntegralsAgree ∧ GlobalTheory.IndefiniteRowSideFullHecke.WeilTensorModels.LocalMatching. The proof does not read the hypothesis structure.

LocalMatching says that for each admissible triple of geometric data, characters and test data for a unitary dual pair over a CM extension, there are matched local data. ThetaIntegralsAgree says three things. The theta integrals decay. Matched data give one constant and coinvariance. The constants for signature (a,b+2)(a, b+2) and (a,b)(a, b), or (b+2,a′)(b+2, a') and (b,a′)(b, a'), differ by an explicit factor that involves a partial Hecke LL-value.

This is a project-specific statement, not a published theorem we could identify. The proof is theoremK_head2 applied to eis_bl_of_literature O and siegel_weil_std_of_literature O. Both ignore O, so O is used in name only. The citations are only the probable sources of those two inputs (Eisenstein series and Siegel-Weil). They are not a source for this statement.

Poincaré-Lelong formula (integrated form used in arithmetic intersection theory)

Lean: UpperTheory.LiteratureTheorems.poincare_lelong_inv_of_literature, which proves GlobalTheory.CompanionTheory.Hypotheses₂.PoincareLelongInv. The proof does not read the hypothesis structure.

Let XX be an arithmetic subvariety of a product of projective spaces, with smooth conjugation-invariant metrized line bundles LjL_j, AA and BB, where BB has weight zero. Let ss be a section of AA. Then the integral of log⁡∥s∥\log \lVert s \rVert against the Chern forms of BB and the LjL_j equals two terms. The first is the sum over horizontal components WW of div⁡(s)\operatorname{div}(s) of their multiplicity times the integral of log⁡hB\log h_B over WW. The second, which is subtracted, is the integral of log⁡hB\log h_B against the Chern form of AA.

In Math 2.0: Poincaré–Lelong formula.

This is a consequence of the Poincaré-Lelong formula ddclog⁡∣s∣2=[div⁡s]−c1(A)dd^c \log \lvert s \rvert^2 = [\operatorname{div} s] - c_1(A), combined with Stokes. It is not the formula itself. In Lean the integral over the complex points uses a projection chosen with Classical.choice. A Lebesgue integral of a non-integrable function is 00, so check that the model integral is the right one. The proof does not use O.

Principalisation of ideals in characteristic zero (Hironaka)

Lean: UpperTheory.LiteratureTheorems.principalisation_char_zero_of_literature, which proves GlobalTheory.CompanionTheory.Hypotheses₂.PrincipalisationCharZero. The proof does not read the hypothesis structure.

Let XX be smooth, quasi-compact and separated over a field of characteristic zero. Let II be an ideal sheaf that contains no irreducible component of XX in its support, and let EE be a simple normal crossings divisor. Then a sequence of blow-ups in smooth centres with normal crossings with EE makes the pullback of II an effective Cartier divisor with snc support. The sequence is an isomorphism outside the support of II.

In Math 2.0: Resolution of singularities.

The Math 2.0 node is for resolution of singularities in general. No node for principalisation was found. The proof is a direct term and does not use O.

Seminegativity of the Lie algebra of a semi-abelian scheme over a curve (dually, nefness of the Hodge bundle)

Lean: UpperTheory.LiteratureTheorems.semi_abelian_lie_seminegative_of_literature, which proves GlobalTheory.Published.SemiAbelianLieSeminegative. The proof does not read the hypothesis structure.

Let CC be a smooth projective geometrically connected curve over a field of characteristic zero, and let GG be a semi-abelian group scheme over CC. Then every vector bundle WW that injects into Lie⁡(G)\operatorname{Lie}(G) has det⁡W\det W of degree at most 00.

Moret-Bailly (on Numdam, no DOI) proves that the determinant of the relative Lie algebra is negative. The Lean form is stronger: it covers every subsheaf that is a vector bundle, which is seminegativity of Lie⁡(G)\operatorname{Lie}(G) as a bundle. In characteristic zero this follows from the semipositivity of the Hodge bundle (Griffiths, Fujita). The proof is a direct term and does not use O.

Toroidal compactifications of PEL-type unitary Shimura varieties with the log Kodaira-Spencer isomorphism

Lean: UpperTheory.LiteratureTheorems.toroidal_for_fan_ks_of_literature, which proves GlobalTheory.Published.ToroidalForFanKS. The proof does not read the hypothesis structure.

Take a unitary Shimura variety of signature (p,q)(p, q) over a CM field, a level, and a smooth polarised fan with no self-intersection. Then there is a toroidal model over the coefficient field. Its generic part is smooth, its boundary has simple normal crossings and lies outside the abelian locus, and it has toroidal charts. A twist of the Hodge bundle by the boundary is ample, and the log Kodaira-Spencer map is an isomorphism.

The Lean form is over the coefficient field (characteristic zero), so it is weaker than Lan's integral result. The proof is a direct term and does not use O.

Unramified Howe duality (the spherical vector generates the invariants)

Lean: GlobalTheory.Hk3.howeUnram_holds, which proves GlobalTheory.Hk3.HoweUnram. The proof does not read the hypothesis structure.

Take an unramified local unitary dual pair over a pp-adic field of characteristic zero, with a Weil representation MM and a nonzero vacuum vector ss that is cyclic. Then every vector fixed by the integral points of the first group lies in the span of the second group's orbit of ss. The same holds with the two groups exchanged.

In Math 2.0: Theta correspondence.

This theorem is proved in Lean and takes no O argument. It is a lemma of the unramified theta correspondence, not full Howe duality. The exact place of the statement in Howe or Rallis was not checked.

Zarhin's trick

Lean: UpperTheory.LiteratureTheorems.zarhin_statement, which proves GlobalTheory.CompanionTheory.Ar.GR.ZarhinStatement. The proof does not read the hypothesis structure.

Let RR be the valuation ring of an algebraically closed valued field, and let AA be an abelian scheme over RR with a relatively ample line bundle. Then AA is a retract of a principally polarised abelian scheme over RR, through group homomorphisms ii and rr with ii followed by rr equal to the identity.

The Lean form is a consequence of Zarhin's trick, which says (A×A∨)4(A \times A^\vee)^4 has a principal polarisation. It is weaker, because it does not name the principally polarised scheme. The DOI 10.1007/BF01388979 is not this paper. It resolves to Hendriks, 'Hyperplane complements of large type'. The correct DOI for Zarhin, Invent. Math. 79 (1985) is 10.1007/BF01388976. The proof passes O to dual_abelian_scheme_of_literature, which ignores O, so O is used in name only.

Markdown. Something is wrong? Report it.