Enriques surfaces #
An Enriques surface is a smooth projective surface whose canonical class is 2-torsion but not trivial, with no first cohomology. This file says exactly that, against the objects this repository already has:
- the surface is
X : Schemewith[IsSmoothProperVariety k X]carrying aCanonicalSheafData k X 2, exactly as forIsK3Surface— the dimension is the2in that package'sSmoothOfRelativeDimension; - "2-torsion canonical class" is the pair of group equations
C.canonicalClass ^ 2 = 1andC.canonicalClass ≠ 1inPic X, which need no torsion-Picard API:Picis a group of units and the conditions are ordinary equations in it; - "no first cohomology" is vanishing of
H¹(X, O_X), using the constructed functorCohomology.coherentH, verbatim theIsK3Surfaceclause.
The characteristic is not a field of the structure #
Over a field of characteristic 2 the surfaces classically called Enriques
split into classical, singular, and supersingular types, and for some of them
ω_X is trivial, contradicting canonicalClass_ne_one. The definition here
is the characteristic ≠ 2 one, but the exclusion is not recorded as a
field: the source papers (Li--Nuer--Stellari--Zhao, arXiv:1912.04332, and
Li--Stellari--Zhao, arXiv:2104.13610) assume an algebraically closed field of
characteristic different from 2 theorem by theorem, and that is where the
hypothesis belongs. A characteristic field here would make every downstream
statement carry it even when unused.
What this file does not do #
- It exhibits no Enriques surface. Nothing here constructs an
X, aCanonicalSheafData k X 2, or anIsEnriquesSurface; every statement is conditional on data nobody has yet produced, exactly as forIsK3Surface. - It states no geometric numerical consequence. Both
H²(X, O_X) = 0andχ(O_X) = 1reduce to one missing input: Serre duality against the non-triviality ofω_X. It givesH²(X, O_X) ≅ H⁰(X, ω_X)^∨ = 0, because a non-trivial torsion line bundle has no sections, and thenχ(O_X) = h⁰ - h¹ + h² = 1needs nothing further.Duality/Serre/carries that duality asDerivedStatementrealization data rather than as a theorem, so neither geometric consequence is available at the pin. A separate numerical shadow now exists inNumerical/Models/Surface/Enriques.lean, including a degree-one witness oftd₁ = 0and∫td₂ = 1; no bridge from that numerical model to a geometricSchemeor toIsEnriquesSurfaceis claimed. - It does not construct the K3 double cover. The 2-torsion class classically determines an étale double cover with trivial canonical class; no covering machinery exists at the pin and none is stated.
Main definitions #
IsEnriquesSurface— theProp-valued definition, over a supplied canonical-sheaf package.
Main results #
IsEnriquesSurface.canonicalSheaf_tensor_self_iso_unit— the sheaf-level trivializationω_X ⊗ ω_X ≅ O_Xextracted from the class equation.IsEnriquesSurface.canonicalSheaf_not_iso_unit— non-triviality at the sheaf level, throughcanonicalClass_eq_one_iff.IsEnriquesSurface.antiCanonicalClass_eq_canonicalClass— a 2-torsion class is its own inverse.IsEnriquesSurface.not_isK3Surface— the same canonical-sheaf package cannot witness both definitions.
References #
- Li, Nuer, Stellari, Zhao, A refined Derived Torelli Theorem for Enriques surfaces, arXiv:1912.04332.
- Li, Stellari, Zhao, A refined Derived Torelli Theorem for Enriques surfaces, II: the non-generic case, arXiv:2104.13610.
Tags #
Enriques surface, canonical class, torsion, Picard group
The definition #
An Enriques surface: a smooth projective surface whose canonical class is 2-torsion but not trivial, with vanishing first cohomology.
Prop-valued over a chosen CanonicalSheafData k X 2, for the same reason as
IsK3Surface: the canonical-sheaf package is genuine data, so bundling it
into the predicate would make "being an Enriques surface" depend on a choice.
The torsion conditions are ordinary group equations in Pic X; no
torsion-subgroup API is involved. The characteristic ≠ 2 hypothesis of the
classical theory is deliberately absent — see the module docstring.
- projective : Variety.IsProjective k X
The surface is projective. Stronger than the properness already carried by
SmoothProperVariety, and the standard hypothesis of the source papers. ω_Xis 2-torsion inPic X. Equivalently, bycanonicalSheaf_tensor_self_iso_unit, there is an isomorphismω_X ⊗ ω_X ≅ O_X.ω_Xis not trivial inPic X. This is what separates an Enriques surface from a K3 surface; seenot_isK3Surface.- h1_vanishing : CategoryTheory.Limits.IsZero ((Cohomology.coherentH X 1).obj X.structureSheafCoh)
H¹(X, O_X) = 0.
Instances
The square of the canonical sheaf of an Enriques surface is trivial, at the sheaf level.
The class equation κ² = 1 is unpacked through
Scheme.Modules.LineBundleData.toPic_tensor and
Scheme.Modules.LineBundleData.toPic_eq_iff; the recorded tensor inverses
play no part, which is why a bare Nonempty of isomorphisms comes out.
The canonical sheaf of an Enriques surface is not trivial, at the sheaf
level. The contrapositive of canonicalClass_eq_one_iff applied to
canonicalClass_ne_one.
A 2-torsion canonical class is its own inverse: κ⁻¹ = κ, since
κ * κ = 1.
Stated for Serre duality, whose surface-level statements twist by κ * L⁻¹
(Duality.Serre.SurfacePicardSymmetry): on an Enriques surface the ω_X and
ω_X⁻¹ twists coincide, so the two directions of the duality agree without a
case split. Nothing consumes it yet — those statements take the canonical
class as a bare Pic parameter and never see a CanonicalSheafData.
No canonical-sheaf package witnesses both definitions: an Enriques surface is not a K3 surface. The two definitions disagree exactly at triviality of the canonical class, and this pins them against each other on the shared carrier.
Scheme-facing convenience namespace #
As for K3 surfaces, "let Y be an Enriques surface" means
(Y : Scheme) [Y.Over (Spec k)] [IsSmoothProperVariety k Y] (C : CanonicalSheafData k Y 2) [IsEnriquesSurface k Y C]. There is no bundled
surface carrier. The namespace below only gives Scheme-facing convenience theorems.
An Enriques surface is projective. Not an instance, for the reason given at
K3Surface.isProjective.