Alphabeta Math
Pipeline-generated
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

Surface Riemann-Roch and the Hodge Index Theorem

1 · Prerequisites

2 · Summary

This page develops the Riemann-Roch theorem and the Hodge index theorem for an integral smooth projective surface over a field. It opens with the canonical divisor, defined as the divisor of a rational section of the dualizing line ωX=⋀2ΩX/k1, and with numerical equivalence and the real Neron-Severi space. Two ampleness tools follow: positivity of an ample class against a nonzero effective divisor, proved through the leading coefficient of the Hilbert polynomial, and the fact that a sufficiently positive twist of any line bundle by an ample bundle is very ample. The adjunction formula for an effective divisor is derived from the surface intersection pairing, its structure sequence and Serre duality, which also gives the vanishing of top cohomology above a canonical ample threshold. These inputs feed the Riemann-Roch theorem χ(OX(D))=12D⋅(D−KX)+χ(OX) for a Cartier divisor D on a smooth projective surface. From Riemann-Roch the page proves that an invertible class with positive square and positive intersection with an ample class has an effective multiple, then the Hodge index theorem first in the ample case and then for an arbitrary class of positive square through expansion in a fixed ample class, and finally the negative definiteness of the intersection form on a primitive part. The closing remark records the base-field and smoothness conventions, the distinction between numerical and linear equivalence, and which tools are deliberately not used. The companion examples page carries the Picard computations behind the worked examples and counterexamples.

3 · Logical flowchart

4 · Definitions, theorems and proofs

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)Open item page →

The canonical divisor of a smooth projective surface

Definition

Assume the Axiom of Choice, inherited from the dualizing-bundle supplier and the Euler-characteristic and Serre-duality suppliers below (The Axiom of Choice). Let k be a field and let X be an integral (Integral schemes), smooth (Smoothness over a field by geometric regularity), projective (Projective morphisms before Proj) k-scheme of pure dimension two (Chain dimension and the empty-space convention). Such an X is an integral regular projective surface over k (Intersection numbers of Cartier divisors on a smooth projective surface): it is of finite type over k and smoothness makes every local ring regular, so all the intersection-theoretic and cohomological constructions of that item apply.

Let ωX=⋀2ΩX/k1 be the dualizing line bundle (Dualizing line bundle and trace datum of a smooth projective variety), an invertible OX-module, and recall that all its cohomology groups are finite-dimensional and the Euler characteristic χ(X,−) of Euler characteristic of a coherent sheaf is defined on it.

Canonical divisor. A canonical divisor on X is a Cartier divisor KX (Cartier divisor) whose associated invertible sheaf satisfies OX(KX)≅ωX (Invertible sheaf of cartier divisor).

One exists. The generic stalk of ωX is one-dimensional over the function field K(X), so ωX has a rational section s (Rational section line bundle); by Rational sections of line bundles are Cartier divisors the divisor KX=div⁡C(s) is a well-defined Cartier divisor and the canonical section 1KX identifies OX(KX)≅ωX carrying 1KX to s. Conversely, for every Cartier divisor D the canonical section 1D is a rational section of OX(D) with div⁡C(1D)=D.

Any two canonical divisors are linearly equivalent (Linear equivalence cartier divisors), and every Cartier divisor linearly equivalent to a canonical divisor is canonical. Indeed, if D,D′ are Cartier divisors with OX(D)≅OX(D′)≅ωX, transport the canonical rational section 1D across an isomorphism φ:OX(D)→OX(D′). There is a unique g∈K(X)× with φ(1D)=g 1D′, since two rational sections of one invertible sheaf differ by a meromorphic unit (Rational section line bundle), and comparing local equations in trivializations shows div⁡C(g 1D′)=div⁡C(g)+D′ (Cartier divisor). Divisors of sections are preserved by a sheaf isomorphism, and div⁡C(1D)=D and div⁡C(1D′)=D′ by Rational sections of line bundles are Cartier divisors, so D−D′=div⁡C(g) is principal and D∼D′; the converse is immediate from the same computation.

Basic properties.

  1. The intersection numbers KX⋅D and D⋅KX depend only on the linear equivalence classes of KX and D and are therefore independent of the chosen canonical divisor (Intersection numbers of Cartier divisors on a smooth projective surface, The surface intersection product is symmetric and bilinear); in particular KX⋅D=D⋅KX. For an invertible OX-module L one writes L⋅KX:=L⋅ωX and KX⋅L:=ωX⋅L; these depend only on the classes.
  2. Serre duality (Serre duality for locally free sheaves on a smooth projective variety) gives χ(X,ωX)=χ(X,OX) and, for every invertible OX-module L, the identity χ(X,L⊗ωX)=χ(X,L∨): dualizing E=OX and E=L∨ pairs Hq with H2−q and preserves the alternating sum because the pairings are perfect and all groups are finite-dimensional. Equivalently χ(X,OX(KX+D))=χ(X,OX(−D)) for every Cartier divisor D on X.
  3. The Euler characteristic of the structure sheaf is the integer χ(X,OX)=h0(X,OX)−h1(X,OX)+h2(X,OX) with hq=dim⁡kHq(X,OX); by property 2 it equals χ(X,ωX). The surface arithmetic genus is pa(X)=χ(X,OX)−1 (Vakil, §18.4.4). No topological description of this integer is used on this page. The expression becomes 1−h1+h2 when H0(X,OX)=k; integrality alone over an arbitrary field does not imply that condition.
DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)Open item page →

Numerical equivalence and the Neron-Severi space of a surface

Definition

Assume the Axiom of Choice, inherited from the intersection product and the Euler-characteristic suppliers (The Axiom of Choice). Let k be a field and let X be an integral regular projective surface over k (Intersection numbers of Cartier divisors on a smooth projective surface).

Numerical equivalence. Two invertible OX-modules L,M (Invertible sheaves) are numerically equivalent, written L≡M, if L⋅N=M⋅Nfor every invertible OX-module N. By Z-bilinearity and symmetry of the intersection product (The surface intersection product is symmetric and bilinear) this is equivalent to (L⊗M∨)⋅N=0 for every invertible N. An invertible sheaf is numerically trivial if it is numerically equivalent to OX.

For Cartier divisors C,D on X (Cartier divisor) write C≡D when OX(C)≡OX(D). Because D↦[OX(D)] induces an isomorphism CaDiv⁡(X)/Prin⁡C(X)→∼Pic⁡(X) and every invertible sheaf is OX(C) for a Cartier divisor C (On an integral scheme, Cartier divisors modulo principal divisors compute the Picard group, Rational sections of line bundles are Cartier divisors), this is equivalent to C⋅E=D⋅E for every Cartier divisor E; by linear-equivalence invariance and the fact that every Cartier divisor is linearly equivalent to a difference of effective Cartier divisors (Intersection numbers of Cartier divisors on a smooth projective surface, The surface intersection product is symmetric and bilinear), it suffices to test E running over effective Cartier divisors. For an effective Cartier divisor E the number C⋅E is the degree of OX(C)∣E (Intersection with a curve is the degree of the restriction), which is the formulation used in the sources.

Basic facts.

  1. Numerical equivalence is an equivalence relation: the defining equalities of integers are reflexive, symmetric and transitive. It is compatible with tensor products and duality, and the numerically trivial classes form a subgroup of the Picard group Pic⁡(X) (Picard group of a scheme): if L≡L′ and M≡M′ then L⊗M≡L′⊗M′ because ⋅ is bilinear, and L∨≡L′∨ because bilinearity gives L∨⋅N=L′∨⋅N (Dual of a line bundle is its tensor inverse).
  2. Linearly equivalent Cartier divisors are numerically equivalent (Intersection numbers of Cartier divisors on a smooth projective surface): the intersection product depends only on the isomorphism classes of the associated invertible sheaves.
  3. Consequently the intersection pairing descends to a well-defined symmetric Z-bilinear pairing on the quotient N⁡1(X):=Pic⁡(X)/{numerically trivial classes}, and L⋅M depends only on the classes of L and M: if L≡L′ and M≡M′ then L⋅M=L′⋅M′ by testing L≡L′ against M and M≡M′ against L′.

Real Neron-Severi space. Set N⁡R1(X):=N⁡1(X)⊗ZR, the extension of scalars along Z→R (Restriction of scalars and extension of scalars S⊗RM along a ring homomorphism R→S). It is a real vector space; no finiteness theorem is used in this definition. The pairing extends uniquely to a symmetric R-bilinear form N⁡R1(X)×N⁡R1(X)⟶R,(x,y)⟼x⋅y, the intersection form. For an invertible sheaf H we write [H]∈N⁡R1(X) for its class.

The form remains nondegenerate after extending scalars. Indeed, N⁡1(X) is torsion-free: if nv=0 with n≠0, then n(v⋅w)=0 for every integral class w, so v is numerically trivial and v=0. Put VQ=N⁡1(X)⊗Q. Every rational class has an integral multiple, so the pairing on VQ is nondegenerate by the same vanishing criterion. For any finite-dimensional rational subspace W⊆VQ, the functionals v↦v⋅w, with w∈VQ, span W∗: otherwise their common kernel in W would contain a nonzero vector, contradicting nondegeneracy. Choose a basis of these functionals. Its evaluation matrix on a rational basis of W is invertible over Q, hence over R. Every real class is a finite real combination of vectors in a rational basis of some such W. If it pairs to zero with every class, this invertible evaluation matrix forces all its coefficients to be zero. In particular [H]=0 exactly when H is numerically trivial.

LemmaStatement: Literature-sourcedProof: AI-adaptedjudge pass (gpt-6.1-sol)Open item page →

Ample divisors meet nonzero effective divisors positively

Statement

Assume the Axiom of Choice (The Axiom of Choice). Let k be a field, let X be an integral smooth projective surface over k (Intersection numbers of Cartier divisors on a smooth projective surface) and let H be an ample invertible OX-module (Absolute ampleness by affine section opens).

  1. If D⊆X is a nonzero effective Cartier divisor (Effective cartier divisor), then H⋅D>0.
  2. Consequently H⋅H>0.
  3. If M is an invertible OX-module (Invertible sheaves) and s∈Γ(X,M) is a nonzero global section with zero scheme Z(s) (Zero scheme of a line-bundle section), then M⋅H≥0, and M⋅H>0 unless Z(s)=∅ and M≅OX. In particular, if h0(X,M)>0 and M is not numerically trivial, then M⋅H>0.

Facts & Assumptions

Given: a field k, an integral smooth projective surface X over k, an ample invertible OX-module H, a nonzero effective Cartier divisor D⊆X, and (for part 3) an invertible sheaf M with a nonzero global section s.

[F1]

X is projective over k in the H-projective convention, hence proper over k (Projective morphisms are proper) and of finite type, so it is Noetherian and locally Noetherian; coherent OX-modules have finite-dimensional cohomology and a well-defined Euler characteristic χ (Euler characteristic of a coherent sheaf, Intersection numbers of Cartier divisors on a smooth projective surface). The intersection product on invertible sheaves and Cartier divisors is symmetric and Z-bilinear, vanishes against OX, and depends only on the isomorphism classes of the entries (Intersection numbers of Cartier divisors on a smooth projective surface, The surface intersection product is symmetric and bilinear).

[F2]

For an effective Cartier divisor C on X and any Cartier divisor E one has C⋅E=deg⁡C(OX(E)∣C), the degree being deg⁡C(N)=χ(C,N)−χ(C,OC) of Degree of an invertible sheaf on a proper one-dimensional scheme on the proper curve C (Intersection with a curve is the degree of the restriction).

[F3]

A nonzero effective Cartier divisor D on the surface X is nonempty and has pure dimension one. Nonemptiness: the vanishing subscheme is empty exactly for the zero divisor (Effective cartier divisor). Dimension: on an affine chart U=Spec⁡A meeting Supp⁡D, the coordinate ring A is a finite-type k-domain of dimension two, and at a closed point x∈U with maximal ideal mx one has ht⁡(mx)=dim⁡A=2 (Maximal ideals of an affine domain have full height); the local equation fx of D at x is a nonzerodivisor and a nonunit, so every prime minimal over (fx) has height one (A minimal prime over a principal nonzerodivisor has height one) and dim⁡A/p=dim⁡A−ht⁡(p)=1 (Height plus quotient dimension equals ambient dimension in an affine domain); hence Supp⁡D has dimension one, and it has dimension at most one everywhere by the same height computation on the remaining charts (Chain dimension and the empty-space convention, A Zariski-closed subset is irreducible exactly when its radical defining ideal is prime, and then it has a unique generic point). Thus D is a proper k-scheme of pure dimension one (Projective morphisms are proper), and for the very ample embedding used below the coherent sheaf F=i∗OD is nonzero with dim⁡Supp⁡F=1 (Integral schemes).

[F4]

Very ample powers: since Spec⁡k is Noetherian, X→Spec⁡k is proper of finite type and H is ample, there is an integer m≥1 such that H⊗m is closed H-very ample relative to Spec⁡k (High powers of an ample line bundle embed a proper scheme); by definition this means that for some N≥0 there is a closed immersion i:X↪PkN with i∗OPN(1)≅H⊗m (Relative very ampleness in the finite projective-space convention). Fix such m,i,N and write OX(1):=H⊗m for the embedding i.

[F5]

Hilbert polynomial: for a coherent OX-module F and the twist F(n)=F⊗OX(1)⊗n the function n↦χ(X,F(n)) is a polynomial PF of degree dim⁡Supp⁡F, with exact degree and nonzero leading coefficient when F≠0, and PF(n)=h0(X,F(n)) for n≫0 (Hilbert function and Euler characteristic on a projective scheme, Euler characteristic is a Hilbert polynomial, Degree of the coherent Hilbert polynomial). Moreover for n≫0 all higher cohomology of F(n) vanishes (Serre vanishing for coherent sheaves and ample twists).

[F6]

For a closed immersion j:D↪X and an invertible sheaf A on X, the projection formula gives A⊗n⊗j∗OD≅j∗(A⊗n∣D) and χ(X,j∗(A⊗n∣D))=χ(D,A⊗n∣D) for every integer n (Projection formula for a closed immersion and an invertible sheaf).

[F7]

A nonzero global section of an invertible sheaf on the integral scheme X is a regular section, its zero scheme Z(s) is an effective Cartier divisor with OX(Z(s))≅M and Z(s)=∅ exactly when s is nowhere vanishing, in which case M≅OX (A regular global section of an invertible sheaf glues to an effective Cartier divisor, Zero scheme of a line-bundle section).

[F8]

The Axiom of Choice is inherited from the cohomology, Hilbert-polynomial and ample-embedding suppliers of [F4]–[F6]; the divisor D, the section s and the embedding i are given data, and only finitely many sheaves and Hilbert-polynomial values are used below.

Proof

technique · direct: replace $H$ by a very ample power, express its intersection with the curve $D$ as the first difference of the Hilbert polynomial of $\mathcal O_D$, and read positivity off the leading coefficient
1.1F1F4

Reduction to a very ample power. Fix m≥1, N and the closed immersion i:X↪PkN with i∗O(1)≅H⊗m supplied by [F4], and put A:=H⊗m. Since the intersection product is Z-bilinear, A⋅D=m (H⋅D) for every Cartier divisor D; hence H⋅D=m−1(A⋅D), and for part 1 it suffices to prove A⋅D>0 for every nonzero effective Cartier divisor D.

1.2F3F5F6

The Hilbert polynomial of the curve. Let j:D↪X be the nonzero effective Cartier divisor and put F=j∗OD, a nonzero coherent sheaf on X with support of dimension one by [F3]. For the fixed embedding with OX(1)=A, [F5] and [F6] give PF(n)=χ(X,F⊗A⊗n)=χ(D,A⊗n∣D) for every integer n. This polynomial has exact degree one, so PF(n)=cn+b with c≠0. For large n it equals h0(X,F⊗A⊗n)≥0, so c>0.

2.1F2step 1.1step 1.2

The intersection number is the leading coefficient. Since PF is linear, c=PF(1)−PF(0). By step 1.2, PF(1)=χ(D,A∣D) and PF(0)=χ(D,OD); by the degree formula of [F2] applied to the effective Cartier divisor D, A⋅D=deg⁡D(A∣D)=χ(D,A∣D)−χ(D,OD)=c>0. This proves part 1 for A and hence, by step 1.1, for the given ample H: H⋅D=m−1(A⋅D)>0.

2.2F1F5step 1.1

Positive self-intersection. Apply [F5] to OX with the same embedding and A=OX(1). Its support is the surface X, so its Hilbert polynomial P(n)=χ(X,A⊗n) has exact degree two, say P(n)=cn2+bn+a with c≠0. Since P(n)=h0(X,A⊗n)≥0 for large n, one has c>0. The defining intersection expression gives A⋅A=χ(X,OX)−2χ(X,A∨)+χ(X,A∨⊗2)=P(0)−2P(−1)+P(−2)=2c>0. Bilinearity then gives H⋅H=m−2(A⋅A)>0. This proves part 2 over every field without choosing a rational point or a hyperplane through one.

3.1F1F7step 2.1

The section criterion. Let M be invertible with a nonzero global section s. By [F7] the zero scheme Z(s) is an effective Cartier divisor with OX(Z(s))≅M. If Z(s)≠∅, then Z(s) is a nonzero effective Cartier divisor and part 1 gives M⋅H=Z(s)⋅H=H⋅Z(s)>0 by symmetry. If Z(s)=∅, then s is nowhere vanishing, so s trivialises M, M≅OX, and M⋅H=0; in this case M is numerically trivial. Hence M⋅H≥0 always, with equality only in the stated case, and if h0(X,M)>0 for a numerically nontrivial M then any nonzero section has nonempty zero scheme and M⋅H>0.

4.1F8step 2.1step 2.2step 3.1∎

Choice accounting and conclusion. Steps 1.1 and 2.1 prove part 1, step 2.2 proves part 2, and step 3.1 proves part 3. The Axiom of Choice is used through the cohomology, Serre-vanishing and Hilbert-polynomial suppliers recorded in [F8], which underlie the very ample embedding and the finiteness of cohomology; the divisors and the section s are single given objects, and no family is selected.

LemmaStatement: Literature-sourcedProof: AI-adaptedjudge pass (gpt-6.1-sol)Open item page →

Large ample twists of a line bundle are very ample

Statement

Assume the Axiom of Choice (The Axiom of Choice). Let k be a field, let X be an integral smooth projective surface over k, let L be an ample invertible OX-module and let M be an invertible OX-module (Absolute ampleness by affine section opens, Invertible sheaves). Then there is an integer n0 such that for every n≥n0 the twist M⊗OXL⊗n is closed H-very ample relative to Spec⁡k (Relative very ampleness in the finite projective-space convention); in particular M⊗L⊗n is H-very ample, hence very ample, relative to Spec⁡k for all n≥n0. No hypothesis is imposed on M.

Facts & Assumptions

Given: a field k, an integral smooth projective surface X over k, an ample invertible sheaf L and an invertible sheaf M on X.

[F1]

Serre's global-generation criterion: on the Noetherian scheme X the invertible sheaf L is ample if and only if for every coherent OX-module F the twist F⊗L⊗n is globally generated for all sufficiently large n (Serre global-generation criterion for ampleness, Locally Noetherian and Noetherian schemes). The invertible sheaf M is coherent on the locally Noetherian scheme X, and X is quasi-compact, so global generation of an invertible sheaf is witnessed by finitely many global sections (Global generation by the evaluation map, Invertible sheaves).

[F2]

Ample powers embed: Spec⁡k is Noetherian, X→Spec⁡k is proper of finite type, and L is ample, so there is an integer d0≥1 such that L⊗d is closed H-very ample relative to Spec⁡k for every d≥d0 (High powers of an ample line bundle embed a proper scheme, Projective morphisms before Proj). Concretely this means that for each such d there is a closed immersion id:X↪PkNd with id∗O(1)≅L⊗d (Relative very ampleness in the finite projective-space convention).

[F3]

Generating sections define a morphism: if s0,…,sp are global sections generating an invertible sheaf G, there is a unique k-morphism φ:X→Pkp with φ∗O(1)≅G and φ∗(xi)=si (Generating line-bundle sections define a morphism to projective space).

[F4]

Graphs and immersions: for a k-morphism u:X→Y with Y separated over k the graph Γu:X→X×kY is a closed immersion, because the defining square is a base change of the diagonal ΔY/k (The graph is a pullback of the diagonal, Separated morphism of schemes) and base changes of closed immersions are closed immersions (Closed immersions are affine quotients and survive base change); the projective space Pkp is proper over k, hence separated (Finite-dimensional projective space is proper over every base, Separated morphism of schemes), and a composite of closed immersions is a closed immersion, since the composite of homeomorphisms onto closed subsets is again one and a composite of surjective sheaf maps is surjective (Closed immersions of schemes). The canonical swap X×kPkp→Pkp×kX is an isomorphism, so it preserves closed immersions.

[F5]

The Segre embedding: with p,q≥0 and P=Pkp×kPkq with projections pr1,pr2, there is a closed immersion σ:P↪Pk(p+1)(q+1)−1 with σ∗O(1)≅pr1∗O(1)⊗pr2∗O(1) (Segre embedding and its line bundle).

[F6]

The Axiom of Choice is inherited from the Proj, ample-embedding and Segre suppliers of [F2], [F4] and [F5]; the finitely many generating sections chosen in step 3.1 are a finite family, and no infinite selection occurs.

Proof

technique · direct: generate one factor by global sections, embed the other by an ample power, and combine the two morphisms through the graph immersion and the Segre embedding
1.1F1F2

The two thresholds. By [F1] applied to the coherent module M there is an integer n1 with G:=M⊗L⊗n1 globally generated. By [F2] there is d0≥1 with L⊗d closed H-very ample for every d≥d0. Put n0:=n1+d0.

2.1F1F2step 1.1

The splitting of the twist. Let n≥n0 and put V:=L⊗(n−n1), so that n−n1≥d0 and V is closed H-very ample by step 1.1; tensoring the identity M⊗L⊗n=(M⊗L⊗n1)⊗L⊗(n−n1) and using associativity and commutativity of the tensor product of invertible sheaves gives M⊗L⊗n≅G⊗V, a tensor product of the globally generated invertible sheaf G and the closed H-very ample invertible sheaf V.

3.1F1F2F3step 2.1

The two morphisms. By [F3] the finite generating family of G (which exists by [F1]) defines a k-morphism φ:X→Pkp with φ∗O(1)≅G. By [F2] applied to d=n−n1 the sheaf V is closed H-very ample, so there is a closed immersion ψ:X↪Pkq with ψ∗O(1)≅V.

4.1F4step 3.1

The product morphism is a closed immersion. Consider (φ,ψ):X→Pkp×kPkq. The graph Γφ:X→X×kPkp is a closed immersion by [F4]; composing with the swap isomorphism gives the closed immersion γ:X↪Pkp×kX, x↦(φ(x),x). The morphism id×ψ:Pkp×kX→Pkp×kPkq is the base change of the closed immersion ψ along the first projection, hence a closed immersion by [F4], and its composite with γ is (φ,ψ). A composite of closed immersions is a closed immersion by [F4], so (φ,ψ) is a closed immersion.

5.1F5step 2.1step 3.1step 4.1

The Segre composite. Let σ be the Segre closed immersion of [F5]. The composite θ:=σ∘(φ,ψ):X↪PkN, N=(p+1)(q+1)−1, is a composite of closed immersions, hence a closed immersion, and θ∗O(1)≅(φ,ψ)∗σ∗O(1)≅(φ,ψ)∗(pr1∗O(1)⊗pr2∗O(1))≅φ∗O(1)⊗ψ∗O(1)≅G⊗V≅M⊗L⊗n.

6.1F6step 5.1∎

Conclusion. The closed immersion θ exhibits M⊗L⊗n≅θ∗O(1) as closed H-very ample relative to Spec⁡k for every n≥n0, hence H-very ample. The Axiom of Choice is inherited from the suppliers recorded in [F6]; the construction selects only the finite generating family of G and the fixed integer parameters.

LemmaStatement: Literature-sourcedProof: AI-adaptedjudge pass (gpt-6.1-sol)Open item page →

Adjunction formula for effective divisors on a smooth projective surface

Statement

Assume the Axiom of Choice, inherited from the intersection, Euler-characteristic and Serre-duality suppliers (The Axiom of Choice). Let k be a field, let X be an integral smooth projective surface over k (Intersection numbers of Cartier divisors on a smooth projective surface, The canonical divisor of a smooth projective surface), let KX be a canonical divisor and let C⊆X be a nonzero effective Cartier divisor (Effective cartier divisor), viewed as a proper curve of dimension one. Then C⋅(KX+C)=−2 χ(C,OC), where χ is the Euler characteristic (Euler characteristic of a coherent sheaf). If C is integral, so that its arithmetic genus is pa(C)=1−χ(C,OC) (Genus and arithmetic genus of a curve), this reads C⋅(KX+C)=2pa(C)−2. No smoothness, reducedness or irreducibility of C is assumed, and deg⁡C(OX(KX+C)∣C)=C⋅(KX+C) for the degree of Degree of an invertible sheaf on a proper one-dimensional scheme.

Facts & Assumptions

Given: a field k, an integral smooth projective surface X over k, a canonical divisor KX with OX(KX)≅ωX, and a nonzero effective Cartier divisor C⊆X.

[F1]

X is proper over k and Noetherian; the intersection product on invertible sheaves and Cartier divisors is defined by the alternating sum of Intersection numbers of Cartier divisors on a smooth projective surface, is symmetric and Z-bilinear, and depends only on the linear equivalence classes of its entries (The surface intersection product is symmetric and bilinear). For an effective Cartier divisor C and any Cartier divisor D one has C⋅D=deg⁡C(OX(D)∣C) with deg⁡C(N)=χ(C,N)−χ(C,OC) (Intersection with a curve is the degree of the restriction, Degree of an invertible sheaf on a proper one-dimensional scheme).

[F2]

Serre duality: for every invertible OX-module E and every q the cup-product pairing Hq(X,E)×H2−q(X,E∨⊗ωX)→k is perfect and all groups are finite-dimensional (Serre duality for locally free sheaves on a smooth projective variety). Consequently χ(X,E)=χ(X,E∨⊗ωX) and χ(X,ωX)=χ(X,OX); moreover OX(KX)≅ωX, so OX(KX+D)≅ωX⊗OX(D) and (ωX⊗OX(D))∨≅OX(−KX−D) for every Cartier divisor D (The canonical divisor of a smooth projective surface, Dual of a line bundle is its tensor inverse, Tensor product of sheaves of modules, Invertible sheaves).

[F3]

Structure sequence and additivity: for the nonzero effective Cartier divisor C there is a short exact sequence of coherent OX-modules 0→OX(−C)→OX→i∗OC→0 (Effective Cartier divisors give a short exact sequence); the Euler characteristic is additive on short exact sequences of coherent modules on the proper scheme X (Euler characteristic is additive in short exact sequences); and χ(X,i∗OC)=χ(C,OC) (Euler characteristic of a coherent sheaf, the projection-formula identification of Intersection with a curve is the degree of the restriction in the effective case). Hence χ(X,OX)−χ(X,OX(−C))=χ(C,OC).

[F4]

The divisor −KX−C is a Cartier divisor and OX(−KX−C)≅(ωX⊗OX(C))∨, while OX(−C)≅OX(C)∨; the canonical divisor is a Cartier divisor and linear equivalence may be replaced by the associated invertible sheaves in all intersection computations (The canonical divisor of a smooth projective surface, Cartier divisor, Invertible sheaf of cartier divisor).

[F5]

The Axiom of Choice is inherited from the Serre-duality and Euler-characteristic suppliers of [F2] and [F3]; the divisor C and the canonical divisor are given data.

Proof

technique · direct: compute $C\cdot(-K_X-C)$ from the defining alternating sum, replace two of its terms by Serre duality, and read the remaining difference off the structure sequence
1.1F1F4

The defining alternating sum. By [F1] and [F4], C⋅(−KX−C)=OX(C)⋅OX(−KX−C)=χ(X,OX)−χ(X,OX(−C))−χ(X,OX(KX+C))+χ(X,OX(KX)), since the duals of OX(C) and OX(−KX−C) are OX(−C) and OX(KX+C), and their tensor product is OX(KX).

1.2F2F4

Serre duality in the two middle terms. By [F2] with E=OX(−C), whose dual twisted by ωX is ωX⊗OX(C), we have χ(X,ωX⊗OX(C))=χ(X,OX(−C)), that is χ(X,OX(KX+C))=χ(X,OX(−C)); and χ(X,OX(KX))=χ(X,ωX)=χ(X,OX).

1.3F3

The structure-sequence difference. By [F3], χ(X,OX)−χ(X,OX(−C))=χ(C,OC).

2.1F1step 1.1step 1.2step 1.3

Conclusion of the computation. Substituting steps 1.2 and 1.3 into step 1.1 gives C⋅(−KX−C)=(χ(X,OX)−χ(X,OX(−C)))+(χ(X,OX(KX))−χ(X,OX(KX+C)))=χ(C,OC)+(χ(X,OX)−χ(X,OX(−C)))=2χ(C,OC), where the second bracket was rewritten using χ(X,OX(KX))=χ(X,OX) and χ(X,OX(KX+C))=χ(X,OX(−C)) from steps 1.2 and 1.3. Since the intersection product is Z-bilinear, C⋅(−KX−C)=−C⋅(KX+C); hence C⋅(KX+C)=−2χ(C,OC).

3.1F5step 2.1∎

The genus form and the degree identity. If C is integral, its arithmetic genus is pa(C)=1−χ(C,OC) by definition (Genus and arithmetic genus of a curve), so the formula becomes C⋅(KX+C)=2pa(C)−2. Since C is effective, the restriction-degree theorem gives deg⁡C(OX(KX+C)∣C)=C⋅(KX+C) (Intersection with a curve is the degree of the restriction). The Axiom of Choice is inherited from [F5]; no further selection is made.

LemmaStatement: Literature-sourcedProof: AI-adaptedjudge pass (gpt-6.1-sol)Open item page →

Vanishing of top cohomology past the canonical threshold

Statement

Assume the Axiom of Choice (The Axiom of Choice). Let k be a field, let X be an integral smooth projective surface over k, let KX be a canonical divisor (The canonical divisor of a smooth projective surface), let H be an ample invertible OX-module (Absolute ampleness by affine section opens) and let M be an invertible OX-module with M⋅H>KX⋅H. Then H2(X,M)=0. Equivalently H0(X,ωX⊗M∨)=0.

Facts & Assumptions

Given: a field k, an integral smooth projective surface X over k, a canonical divisor KX, an ample invertible sheaf H, and an invertible sheaf M with M⋅H>KX⋅H.

[F1]

Serre duality: for the invertible sheaf M the pairing H2(X,M)×H0(X,M∨⊗ωX)→H2(X,ωX)→tXk is perfect and both groups are finite-dimensional (Serre duality for locally free sheaves on a smooth projective variety); hence h2(X,M)=h0(X,ωX⊗M∨) and H2(X,M)=0 if and only if H0(X,ωX⊗M∨)=0 (Euler characteristic of a coherent sheaf, Invertible sheaves, Tensor product of sheaves of modules).

[F2]

Positivity of ample classes: if D is a nonzero effective Cartier divisor, then H⋅D>0; and if an invertible sheaf N has a nonzero global section whose zero scheme is empty, then N≅OX (Ample divisors meet nonzero effective divisors positively).

[F3]

Zero schemes of sections: a nonzero global section of an invertible sheaf on the integral scheme X is regular, its zero scheme Z(s) is an effective Cartier divisor with OX(Z(s))≅N, and Z(s) is empty if and only if s is nowhere vanishing (A regular global section of an invertible sheaf glues to an effective Cartier divisor, Zero scheme of a line-bundle section).

[F4]

The intersection numbers KX⋅H and M⋅H are computed through the associated invertible sheaves ωX and M; they depend only on the linear equivalence classes and are additive, so that (KX−M)⋅H=KX⋅H−M⋅H (The canonical divisor of a smooth projective surface, Intersection numbers of Cartier divisors on a smooth projective surface, The surface intersection product is symmetric and bilinear).

[F5]

The Axiom of Choice is inherited from the Serre-duality and ample-positivity suppliers of [F1] and [F2]; the section used below, if it exists, is a single object.

Proof

technique · direct: pass to the Serre-dual space, suppose it has a nonzero section, and split into the nowhere-vanishing and effective-divisor cases
1.1F1F4

Duality reduction. Put N:=ωX⊗M∨. By [F1] it suffices to prove H0(X,N)=0.

1.2F2F3F4

A nonzero section leads to a contradiction. Suppose 0≠s∈H0(X,N) and let Z:=Z(s) be its zero scheme. By [F3], Z is either empty or an effective Cartier divisor with OX(Z)≅N. If Z=∅, then s is nowhere vanishing, so N≅OX by [F2]; tensoring with M gives ωX≅M, hence M⋅H=KX⋅H, contradicting the hypothesis. If Z≠∅, then Z is a nonzero effective Cartier divisor whose sheaf is OX(Z)≅ωX⊗M∨; by linear-equivalence invariance of the intersection product and positivity [F2], [F4], 0<Z⋅H=(KX−M)⋅H=KX⋅H−M⋅H, so M⋅H<KX⋅H, again contradicting the hypothesis.

2.1F5step 1.1step 1.2∎

Conclusion. Both cases of step 1.2 are impossible, so H0(X,ωX⊗M∨)=0 and hence H2(X,M)=0 by step 1.1. The Axiom of Choice is inherited from the suppliers recorded in [F5]; no family of sections is selected.

TheoremStatement: Literature-sourcedProof: AI-adaptedjudge pass (gpt-6.1-sol)Open item page →

Riemann-Roch for smooth projective surfaces

Statement

Assume the Axiom of Choice (The Axiom of Choice). Let k be a field, let X be an integral smooth projective surface over k, and let KX be a canonical divisor (The canonical divisor of a smooth projective surface). Then for every invertible OX-module L (Invertible sheaves) χ(X,L)=χ(X,OX)+12(L⋅L−L⋅KX), and equivalently, for every Cartier divisor D on X (Cartier divisor) with OX(D)≅L, χ(X,OX(D))=χ(X,OX)+12 D⋅(D−KX). Here χ is the Euler characteristic (Euler characteristic of a coherent sheaf). The right-hand side is an integer and is independent of the choice of canonical divisor and of the representative divisor of L (The canonical divisor of a smooth projective surface). No ampleness, effectivity or vanishing hypothesis is imposed on L.

Facts & Assumptions

Given: a field k, an integral smooth projective surface X over k, a canonical divisor KX, and an invertible sheaf L.

[F1]

The intersection product is defined by the alternating sum L1⋅L2=χ(X,OX)−χ(X,L1∨)−χ(X,L2∨)+χ(X,L1∨⊗L2∨) of Intersection numbers of Cartier divisors on a smooth projective surface, is symmetric and Z-bilinear, and satisfies L1∨⊗L1≅OX (The surface intersection product is symmetric and bilinear, Dual of a line bundle is its tensor inverse, Tensor product of sheaves of modules, Invertible sheaves).

[F2]

The canonical sheaf satisfies OX(KX)≅ωX=⋀2ΩX/k1, and KX⋅L=ωX⋅L for every invertible L (The canonical divisor of a smooth projective surface).

[F3]

Serre duality: for every invertible sheaf E the cup-product pairing Hq(X,E)×H2−q(X,E∨⊗ωX)→k is perfect with finite-dimensional groups, so χ(X,E)=χ(X,E∨⊗ωX); in particular χ(X,ωX)=χ(X,OX) (Serre duality for locally free sheaves on a smooth projective variety, Euler characteristic of a coherent sheaf).

[F4]

The Axiom of Choice is inherited from the Serre-duality and Euler-characteristic suppliers of [F3]; the sheaf L and the divisor KX are given data.

Proof

technique · direct: evaluate the defining alternating sum on the pair $(\mathcal L^{\vee},\mathcal L\otimes\omega_X^{\vee})$ and replace two terms by Serre duality
1.1F1

The defining sum. Put A:=L∨ and B:=L⊗ωX∨. By [F1], A⋅B=χ(X,OX)−χ(X,A∨)−χ(X,B∨)+χ(X,A∨⊗B∨)=χ(X,OX)−χ(X,L)−χ(X,L∨⊗ωX)+χ(X,ωX), because A∨≅L, B∨≅L∨⊗ωX and A∨⊗B∨≅ωX (Dual of a line bundle is its tensor inverse).

2.1F3step 1.1

Serre duality in the two last terms. By [F3] with E=L∨⊗ωX, whose dual twisted by ωX is L, we have χ(X,L∨⊗ωX)=χ(X,L); and χ(X,ωX)=χ(X,OX). Hence A⋅B=2(χ(X,OX)−χ(X,L)).

3.1F1F2step 1.1step 2.1

Bilinearity. By [F1] and [F2], A⋅B=L∨⋅(L⊗ωX∨)=−L⋅L+L⋅ωX=L⋅KX−L⋅L, where L⋅KX is the intersection with ωX by [F2]. Comparing with step 2.1 and rearranging, 2(χ(X,OX)−χ(X,L))=L⋅KX−L⋅L,that isχ(X,L)=χ(X,OX)+12(L⋅L−L⋅KX). The right-hand side is an integer because χ is; it depends only on the isomorphism class of L and on the canonical class, hence is independent of the chosen representative divisor and canonical divisor (The canonical divisor of a smooth projective surface).

4.1F4step 3.1∎

The divisor form. If D is a Cartier divisor with OX(D)≅L, then L⋅L=D⋅D and L⋅KX=D⋅KX by the dictionary of Intersection numbers of Cartier divisors on a smooth projective surface and The canonical divisor of a smooth projective surface, so χ(X,OX(D))=χ(X,OX)+12D⋅(D−KX). The Axiom of Choice is inherited from [F4]; no further selection is made.

LemmaStatement: Literature-sourcedProof: AI-adaptedjudge pass (gpt-6.1-sol)Open item page →

Positive square and positive ample intersection force an effective multiple

Statement

Assume the Axiom of Choice (The Axiom of Choice). Let k be a field, let X be an integral smooth projective surface over k, let H be an ample invertible OX-module (Absolute ampleness by affine section opens) and let L be an invertible OX-module with L⋅L>0,L⋅H>0. Then there exists an integer n≥1 with H0(X,L⊗n)≠0.

Facts & Assumptions

Given: a field k, an integral smooth projective surface X over k, an ample invertible sheaf H, and an invertible sheaf L with L⋅L>0 and L⋅H>0.

[F1]

Bilinearity: Z-bilinearity of the intersection product gives L⊗n⋅H=n(L⋅H) and L⊗n⋅L⊗n=n2(L⋅L) for every n≥0; in particular L⊗n⋅H>KX⋅H for all sufficiently large n, with KX⋅H=ωX⋅H (Intersection numbers of Cartier divisors on a smooth projective surface, The surface intersection product is symmetric and bilinear, The canonical divisor of a smooth projective surface, Invertible sheaves, Tensor product of sheaves of modules, Dual of a line bundle is its tensor inverse).

[F2]

Threshold vanishing: if M⋅H>KX⋅H for an invertible sheaf M, then H2(X,M)=0 (Vanishing of top cohomology past the canonical threshold, The canonical divisor of a smooth projective surface).

[F3]

Riemann-Roch: for every invertible sheaf M, χ(X,M)=χ(X,OX)+12(M⋅M−M⋅KX) (Riemann-Roch for smooth projective surfaces). The Euler characteristic is the alternating sum χ=h0−h1+h2 of the dimensions hq=dim⁡kHq(X,M), so h0=χ+h1−h2 with h1,h2≥0 (Euler characteristic of a coherent sheaf).

[F4]

The Axiom of Choice is inherited from the Riemann-Roch and vanishing suppliers of [F2] and [F3]; the integer n is chosen below from the growth of a quadratic polynomial, no family is selected.

Proof

technique · direct: for large $n$ the top cohomology of $L^{\otimes n}$ vanishes and Riemann-Roch makes the Euler characteristic positive; the sign in $h^0=\chi+h^1-h^2$ then gives a section
1.1F1F2

Vanishing of the top cohomology for large n. By [F1], L⊗n⋅H=n(L⋅H)→+∞ as n→∞ because L⋅H>0, so for all n≫0 we have L⊗n⋅H>KX⋅H; by [F2], H2(X,L⊗n)=0 for all such n.

1.2F1F3

Growth of the Euler characteristic. By [F3] and [F1], χ(X,L⊗n)=χ(X,OX)+n22(L⋅L)−n2(L⋅KX)⟶+∞(n→∞), because the quadratic term has positive leading coefficient L⋅L>0.

2.1F3F4step 1.1step 1.2∎

A positive Euler characteristic gives a section. Choose n≫0 so large that both step 1.1 applies and χ(X,L⊗n)>0, which is possible by steps 1.1 and 1.2. By [F3], h0(X,L⊗n)=χ(X,L⊗n)+h1(X,L⊗n)−h2(X,L⊗n), and h2=0 by step 1.1 while h1≥0; hence h0(X,L⊗n)≥χ(X,L⊗n)>0 and H0(X,L⊗n)≠0. The Axiom of Choice is inherited from [F4].

TheoremStatement: Literature-sourcedProof: AI-adaptedjudge pass (gpt-6.1-sol)Open item page →

The Hodge index theorem for an ample class

Statement

Assume the Axiom of Choice (The Axiom of Choice). Let k be a field, let X be an integral smooth projective surface over k, let H be an ample invertible OX-module (Absolute ampleness by affine section opens) and let L be an invertible OX-module with L⋅H=0. Then:

(a) L⋅L≤0; (b) L⋅L=0 if and only if L is numerically trivial (Numerical equivalence and the Neron-Severi space of a surface).

Facts & Assumptions

Given: a field k, an integral smooth projective surface X over k, an ample invertible sheaf H, and an invertible sheaf L with L⋅H=0.

[F1]

The intersection product is symmetric and Z-bilinear, so (L⊗H⊗m)⋅(L⊗H⊗m)=L⋅L+m2(H⋅H) because L⋅H=0, and (L⊗H⊗m)⋅L=L⋅L; the product is trivial against the structure sheaf and depends only on isomorphism classes (Intersection numbers of Cartier divisors on a smooth projective surface, The surface intersection product is symmetric and bilinear, Invertible sheaves, Tensor product of sheaves of modules, Dual of a line bundle is its tensor inverse).

[F2]

Large ample twists of L are very ample: there is m0 such that L⊗H⊗m is closed H-very ample, hence H-very ample and ample, for every m≥m0 (Large ample twists of a line bundle are very ample, Relative very ampleness implies relative ampleness, Absolute ampleness by affine section opens). In particular H⋅H>0 (Ample divisors meet nonzero effective divisors positively).

[F3]

Positive square and positive ample intersection force an effective multiple: if A is ample and M is invertible with M⋅M>0 and M⋅A>0, then H0(X,M⊗n)≠0 for some n≥1 (Positive square and positive ample intersection force an effective multiple).

[F4]

Sections and positivity: a nonzero global section of an invertible sheaf M on the integral X is regular, its zero scheme Z is an effective Cartier divisor with OX(Z)≅M, and if Z=∅ then M≅OX (A regular global section of an invertible sheaf glues to an effective Cartier divisor, Zero scheme of a line-bundle section, Ample divisors meet nonzero effective divisors positively). A nonzero effective Cartier divisor D satisfies H⋅D>0, and H⋅D=deg⁡D(H∣D) (Ample divisors meet nonzero effective divisors positively, Intersection with a curve is the degree of the restriction).

[F5]

Numerical triviality: L is numerically trivial exactly when L⋅N=0 for every invertible N; otherwise there is an invertible Q with Q⋅L≠0 (Numerical equivalence and the Neron-Severi space of a surface).

[F6]

The Axiom of Choice is inherited from the embedding, positive-square and positivity suppliers of [F2]–[F4]; the integer m and the sign of n below are chosen from finitely described data.

Proof

technique · direct: reduce part (a) to the positive-square lemma via a large ample twist, then handle the equality case of part (b) by a two-parameter perturbation
1.1F1F2F3

Part (a), first reduction. Suppose L⋅L>0. By [F2] choose m≥m0 and put L′:=L⊗H⊗m, an ample invertible sheaf. By [F1], L′⋅L′=L⋅L+m2(H⋅H)>0 and L′⋅L=L⋅L>0. Applying [F3] with the ample class L′ and the sheaf L, we obtain n≥1 with H0(X,L⊗n)≠0.

2.1F1F4step 1.1

Part (a), contradiction. Let n≥1 and let s be a nonzero global section of L⊗n; by [F4] its zero scheme Z is an effective Cartier divisor with OX(Z)≅L⊗n, and L⊗n⋅H=n(L⋅H)=0 by [F1]. If Z≠∅, then H⋅Z>0 by [F4], contradicting H⋅Z=L⊗n⋅H=0. Hence Z=∅ and L⊗n≅OX by [F4]; then L⋅L=n−2(L⊗n⋅L⊗n)=0 by [F1], contradicting L⋅L>0. Therefore L⋅L≤0, proving (a).

3.1F1F2F5step 2.1

Part (b). If L is numerically trivial then L⋅L=0 by definition [F5]. Conversely assume L⋅L=0 and suppose L is not numerically trivial; by [F5] choose an invertible Q with Q⋅L≠0. Put R:=Q⊗(H⋅H)⊗H⊗−(Q⋅H), an invertible sheaf with R⋅H=(H⋅H)(Q⋅H)−(Q⋅H)(H⋅H)=0,R⋅L=(H⋅H)(Q⋅L)−(Q⋅H)(H⋅L)=(H⋅H)(Q⋅L)≠0, using H⋅L=0, H⋅H>0 and bilinearity [F1, F2]. For an integer n put L′:=L⊗n⊗R; then L′⋅H=n(L⋅H)+R⋅H=0 and L′⋅L′=n2(L⋅L)+2n(L⋅R)+R⋅R=2n(L⋅R)+R⋅R, which is positive for a suitable sign of n because L⋅R≠0. Part (a) applied to L′ then gives L′⋅L′≤0, a contradiction. Hence L is numerically trivial, and part (b) follows in both directions.

4.1F6step 2.1step 3.1∎

Conclusion and choice accounting. Step 2.1 proves (a) and step 3.1 proves (b); the Axiom of Choice is inherited from the suppliers recorded in [F6].

TheoremStatement: Literature-sourcedProof: AI-adaptedjudge pass (gpt-6.1-sol)Open item page →

The Hodge index theorem for smooth projective surfaces

Statement

Assume the Axiom of Choice (The Axiom of Choice). Let k be a field, let X be an integral smooth projective surface over k, and let H,L be invertible OX-modules (Invertible sheaves) with H⋅H>0,L⋅H=0. Then:

(a) L⋅L≤0; (b) L⋅L=0 if and only if L is numerically trivial (Numerical equivalence and the Neron-Severi space of a surface).

No ampleness of H is assumed, and the field k is arbitrary; in particular the statement applies to classes with positive self-intersection that are not ample.

Facts & Assumptions

Given: a field k, an integral smooth projective surface X over k, and invertible sheaves H,L with H⋅H>0 and L⋅H=0.

[F1]

Ample classes exist on X: the projective hypothesis provides a closed immersion X↪Pkn in the H-projective convention, and OX(1)=i∗OPn(1) is closed H-very ample relative to Spec⁡k, hence H-very ample and ample (Relative very ampleness in the finite projective-space convention, Relative very ampleness implies relative ampleness, Absolute ampleness by affine section opens). Such an ample class A satisfies A⋅A>0 (Ample divisors meet nonzero effective divisors positively).

[F2]

The ample case of the Hodge index theorem: for an ample invertible sheaf A and an invertible sheaf L′ with L′⋅A=0 one has L′⋅L′≤0, with equality if and only if L′ is numerically trivial (The Hodge index theorem for an ample class).

[F3]

Bilinearity: the intersection product is symmetric and Z-bilinear, and intersection numbers depend only on isomorphism classes, so expressions such as (H⊗a⊗L⊗b)⋅A=a(H⋅A)+b(L⋅A) and expansions of self-intersections of tensor products may be computed term by term (The surface intersection product is symmetric and bilinear, Invertible sheaves).

[F4]

Numerical triviality is characterized by vanishing against every invertible sheaf (Numerical equivalence and the Neron-Severi space of a surface).

[F5]

The Axiom of Choice is inherited from the ample-embedding and positivity suppliers of [F1]; all sheaves below are fixed tensor products of the given H,L and of one ample class A.

Proof

technique · direct: fix an ample reference class and reduce to the ample case, once directly and once through a perturbation $M$ depending on $H$ and $L$
1.1F1F2

The case A⋅L=0. Let A be an ample class as in [F1]. If A⋅L=0, then [F2] applied to A and L gives L⋅L≤0 with equality if and only if L is numerically trivial; this is exactly (a) and (b) in this case.

2.1F1F2F3step 1.1

The case A⋅L≠0: a nonzero auxiliary class. Assume now A⋅L≠0. If A⋅H=0, then [F2] applied to A and H gives H⋅H≤0, contradicting H⋅H>0; hence A⋅H≠0. Define M:=H⊗(A⋅L)⊗L⊗−(A⋅H), an invertible sheaf whose class is (A⋅L)[H]−(A⋅H)[L]. By [F3], M⋅A=(A⋅L)(H⋅A)−(A⋅H)(L⋅A)=0, so [F2] applies to M and gives M⋅M≤0. Expanding with [F3] and using H⋅L=0, M⋅M=(A⋅L)2(H⋅H)+(A⋅H)2(L⋅L). Since (A⋅L)2>0 and H⋅H>0, if L⋅L≥0 then M⋅M>0, a contradiction. Hence L⋅L<0 in this case; in particular L is not numerically trivial.

3.1F4F5step 1.1step 2.1∎

Parts (a) and (b). In the case of step 1.1, (a) and (b) are proved there. In the case of step 2.1 we have L⋅L<0, so (a) holds, and (b) holds because a numerically trivial L would have L⋅L=0 by definition, while conversely L⋅L=0 is false here. More explicitly, for (b) in the mixed case: if L is numerically trivial then L⋅L=0 by [F4]; if L⋅L=0 then step 2.1 forces A⋅L=0, so step 1.1 gives that L is numerically trivial. Thus (b) holds in all cases. The Axiom of Choice is inherited from [F5].

CorollaryStatement: Literature-sourcedProof: AI-adaptedjudge pass (gpt-6.1-sol)Open item page →

Negative definiteness of the primitive part of the Neron-Severi space

Statement

Assume the Axiom of Choice (The Axiom of Choice). Let k be a field, let X be an integral smooth projective surface over k and let H be an invertible OX-module with H⋅H>0 (Invertible sheaves). Write h:=[H]∈N⁡R1(X) for its class and let h⊥={ x∈N⁡R1(X):x⋅h=0 } be the primitive part (Numerical equivalence and the Neron-Severi space of a surface).

  1. Every class of N⁡R1(X) has a unique orthogonal decomposition x=a h+x0 with a∈R and x0∈h⊥; equivalently N⁡R1(X)=Rh⊕h⊥ is an orthogonal direct sum.
  2. The intersection form is negative definite on h⊥ (Positive and negative definiteness, the inertia (p,q,r), rank p+q, and signature p−q of a real symmetric bilinear or quadratic form): for x∈h⊥ one has x⋅x≤0, with equality if and only if x=0.
  3. If the Picard number ρ(X)=dim⁡RN⁡R1(X) is finite, then the intersection form on N⁡R1(X) is nondegenerate of dimension ρ(X) and has inertia (1,ρ(X)−1,0), equivalently index (1,ρ(X)−1) and signature 2−ρ(X) (Positive and negative definiteness, the inertia (p,q,r), rank p+q, and signature p−q of a real symmetric bilinear or quadratic form). Finiteness of ρ(X) is not proved here; the negative definiteness in (2) is unconditional.

Facts & Assumptions

Given: a field k, an integral smooth projective surface X over k, an invertible sheaf H with H⋅H>0, and its class h∈N⁡R1(X).

[F1]

N⁡R1(X)=N⁡1(X)⊗ZR carries the R-bilinear extension of the intersection pairing, which is symmetric; the rational evaluation-matrix argument in the numerical-space definition proves that the extended form is nondegenerate: x⋅y=0 for all y forces x=0 (Numerical equivalence and the Neron-Severi space of a surface, Restriction of scalars and extension of scalars S⊗RM along a ring homomorphism R→S, Intersection numbers of Cartier divisors on a smooth projective surface, The surface intersection product is symmetric and bilinear).

[F2]

The Hodge index theorem: for invertible sheaves H′,L′ with H′⋅H′>0 and L′⋅H′=0 one has L′⋅L′≤0, and equality holds if and only if L′ is numerically trivial (The Hodge index theorem for smooth projective surfaces).

[F3]

Structure of N⁡R1(X): it is the extension of scalars of the quotient of Pic⁡(X) by the subgroup of numerically trivial classes (Numerical equivalence and the Neron-Severi space of a surface, Restriction of scalars and extension of scalars S⊗RM along a ring homomorphism R→S). Hence every element is a finite real linear combination of classes [Li] of invertible sheaves, the Q-span of finitely many such classes consists of rational combinations ∑qi[Li], and clearing denominators turns a rational combination into an integral combination ∑ai[Li], ai∈Z, which is the class of the invertible sheaf ⨂iLi⊗ai (negative exponents meaning duals, Invertible sheaves, Dual of a line bundle is its tensor inverse). On each finite-dimensional real span the bilinear form is continuous in its usual Euclidean topology and the rational span of finitely many classes is dense in their real span.

[F4]

Inertia, rank, index and signature of a symmetric bilinear form on a finite-dimensional real vector space: a diagonalizing basis with p positive, q negative and r zero diagonal entries gives inertia (p,q,r), rank p+q and signature p−q, independent of the basis (Positive and negative definiteness, the inertia (p,q,r), rank p+q, and signature p−q of a real symmetric bilinear or quadratic form).

[F5]

The Axiom of Choice is inherited from the Hodge index and extension-of-scalars suppliers of [F2] and [F3]; the finite families of classes used below are chosen finitely many at a time.

Proof

technique · direct: split off the positive line $\mathbb R h$, reduce the negativity on the primitive part to the integral Hodge index theorem by rational approximation, and handle the equality case with a two-dimensional indefinite plane
1.1F1F3

The orthogonal decomposition. For x∈N⁡R1(X) put a:=(x⋅h)/(h⋅h), a real number because h⋅h>0 by hypothesis (Positive and negative definiteness, the inertia (p,q,r), rank p+q, and signature p−q of a real symmetric bilinear or quadratic form). Then (x−ah)⋅h=x⋅h−a(h⋅h)=0, so x0:=x−ah∈h⊥ and x=ah+x0. If x=a′h+x0′ is another such decomposition, then (a−a′)h=x0′−x0∈h⊥ and both summands are orthogonal to h, so (a−a′)(h⋅h)=((a−a′)h)⋅h=0 and a=a′, whence x0=x0′; orthogonality of Rh and h⊥ is the definition of the latter and h⋅h≠0. This proves (1).

1.2F2F3

Negative semidefiniteness on the primitive part. Suppose x∈h⊥ has x⋅x>0. Write x as a finite real combination of classes [L1],…,[Lr] of invertible sheaves, consider the real span V of h,[L1],…,[Lr] and the rational span W of the [Li] inside V; the map p(y):=y−y⋅hh⋅hh is R-linear with rational coefficients on W, so p(W) is a dense Q-subspace of p(V)=V∩h⊥. The set {y∈V∩h⊥:y⋅y>0} is open in V∩h⊥ and nonempty because it contains x, hence contains some y=p(w)∈p(W). Clearing denominators of the rational coefficients of y produces an integer N≥1 and an element z=Ny∈N⁡1(X) of the form ∑ai[Li]+a0h with ai,a0∈Z, which is the class of the invertible sheaf ⨂iLi⊗ai⊗H⊗a0; moreover z⋅h=N(y⋅h)=0 and z⋅z=N2(y⋅y)>0. The Hodge index theorem [F2] applied to H and this invertible sheaf gives z⋅z≤0, a contradiction. Hence x⋅x≤0 for every x∈h⊥.

2.1F1step 1.2

The equality case. Let x∈h⊥ satisfy x⋅x=0. If x≠0, nondegeneracy [F1] supplies z with z⋅x≠0. Put z′:=z−z⋅hh⋅hh∈h⊥, so z′⋅x=z⋅x≠0. For t∈R, the class z′+tx lies in h⊥ and has square z′⋅z′+2t(z′⋅x), which is positive for t of a suitable sign and sufficiently large magnitude. This contradicts step 1.2. Hence x⋅x=0 forces x=0; the converse is immediate by bilinearity. Together with step 1.2, this proves negative definiteness.

3.1F4F5step 1.1step 2.1∎

The signature statement. Assume ρ(X)=dim⁡RN⁡R1(X) is finite. By (1) the form decomposes as the orthogonal direct sum of Rh, on which it is positive definite because h⋅h>0, and of h⊥, on which it is negative definite by step 2.1; the decomposition is nondegenerate with dim⁡h⊥=ρ(X)−1 (if ρ(X)=0 the space is zero and the statement is vacuous, while ρ(X)≥1 here because h≠0). Hence the inertia is (1,ρ(X)−1,0), the rank is ρ(X) and the signature is 1−(ρ(X)−1)=2−ρ(X) by [F4]. The Axiom of Choice is inherited from [F5].

RemarkRemark: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)Open item page →

Conventions and hypothesis bookkeeping for surface Riemann-Roch and Hodge index

Remark

Assume the Axiom of Choice where the cited cohomology, Euler-characteristic and duality suppliers do (The Axiom of Choice). The following conventions govern the items of this page.

Base field and smoothness. The theorems Riemann-Roch for smooth projective surfaces, The Hodge index theorem for smooth projective surfaces and Negative definiteness of the primitive part of the Neron-Severi space are stated for an integral smooth projective surface over an arbitrary field k, following Vakil's Theorem 20.2.13 and Exercise 20.2.B. Smoothness over k (Smoothness over a field by geometric regularity) is used through the dualizing line bundle ωX=⋀2ΩX/k1 and Serre duality (Dualizing line bundle and trace datum of a smooth projective variety, Serre duality for locally free sheaves on a smooth projective variety); over an imperfect field it is strictly stronger than regularity, and no claim is made here for regular non-smooth surfaces. A smooth surface over a field is regular, so all intersection-theoretic items of Intersection numbers of Cartier divisors on a smooth projective surface apply.

Numerical versus linear equivalence. Numerical equivalence is defined by vanishing of all intersection numbers (Numerical equivalence and the Neron-Severi space of a surface); it is coarser than linear equivalence, and the equality case in the Hodge index theorem is numerical triviality, not linear triviality. The real Neron-Severi space N⁡R1(X) is finite-dimensional by the Neron-Severi theorem, which is not proved or used on this page. Accordingly, the signature statement in Negative definiteness of the primitive part of the Neron-Severi space is conditional on finiteness of ρ(X) while its negative-definiteness statement is unconditional.

Positive-square hypothesis, and tools not used. The Hodge index theorem assumes H⋅H>0, not ampleness of H; this generality is used in the companion example on the blowup, where the class H=π∗O(1) of the blowup π:X′→P2 of a rational point has H⋅H=1>0 yet is not ample, since H⋅E=0 for the exceptional curve E. Positivity of ample divisors against nonzero effective divisors is supplied by Ample divisors meet nonzero effective divisors positively and rests on the Hilbert-polynomial leading coefficient, not on any resolution or Bertini input. No Bertini theorem, no resolution of singularities and no Nakai-Moishezon criterion is invoked anywhere on this page: surface Riemann-Roch and both Hodge index statements are derived here from the cited duality, intersection and cohomology suppliers alone.

5 · Examples, counterexamples and false statements

None yet.

Sources