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.

✓ 29 results · all verified · 13 also independently AI-judged
Every result on this page is machine-checked by a proof checker and read in full and owner-audited; the judge is an additional, independent cross-model AI review of the proofs. The 16 not AI-judged were verified by owner audit (typically over a confirmed judge false positive), not failures.

Riemann Roch for Curves via Euler Characteristics

1 · Prerequisites

2 · Summary

This page proves the Riemann–Roch theorem for a smooth proper geometrically integral curve C over a field k in its Euler-characteristic form, ℓ(D)−i(D)=deg⁡k(D)+1−g, together with the consequences that can be reached without Serre duality. Throughout, divisors are finite integral combinations of closed points with the degree weighted by residue degrees, OC(D) is the associated invertible sheaf, L(D) is the space of rational functions whose poles are bounded by D, and hi(D) denotes dim⁡kHi(C,OC(D)). No identification of H1(C,OC(D)) with the sections of a complementary invertible sheaf is made on this page: the index of speciality stays an unknown nonnegative integer, and the sharp classical thresholds are deferred to the pair on residues, Serre duality and the full Riemann–Roch theorem.

The first block sets up the cohomological bookkeeping. The integer ℓ(D)=h0(D) is defined through the divisor space, and it is finite for every divisor because the invertible sheaf OC(D) is coherent and the cohomology of a coherent sheaf on a proper curve over a field is finite-dimensional and vanishes above degree one. The natural inclusion L(D)⊆L(E) for D≤E is recorded together with the exact sequence for adding one point, whose cokernel is a skyscraper at the point with cohomology concentrated in degree zero, of dimension the residue degree of the point. Iterating gives the Euler-characteristic shift χ(OC(D+E))=χ(OC(D))+deg⁡k(E) for effective E, and every divisor is a finite signed sum of closed points, so the shift extends to all divisors: χ(OC(D))−χ(OC)=deg⁡k(D). The genus is defined by g(C)=h1(C,OC), so that χ(C,OC)=1−g(C), and the Riemann–Roch theorem is the combination of the two identities. The Riemann inequality ℓ(D)≥deg⁡k(D)+1−g and the vanishing of L(D) in negative degree are immediate consequences; the latter is proved by the effective-divisor argument rather than through the inequality, whose nonpositive lower bound cannot ensure a nonzero section.

The second block derives what can be said about h1 without duality. Adding points never raises h1: the long exact sequence of the one-point sequence identifies H1(OC(D+p)) with a quotient of H1(OC(D)), so n↦h1(OC(D0+nA)) is non-increasing and stabilizes for every fixed divisor D0 and effective A. From the Riemann inequality one obtains nonconstant rational functions with bounded pole at a single closed point, then the finite morphism φf:C→Pk1 attached to such a function, of degree [k(C):k(f)] and with fibre over infinity the pole divisor of f. A finite morphism to the projective line pulls the ample twisting sheaf back to an ample invertible sheaf, and Serre vanishing along that fixed ample direction gives the vanishing of H1(OC(D0+nA+E)) for all n≥n0 and every effective E, where n0 may depend on D0 and on the chosen morphism. Combining this with Riemann–Roch yields the fixed-direction form of Riemann's theorem, ℓ(D)=deg⁡k(D)+1−g for every divisor D≥D0+n0A. The same circle of ideas gives the genus-zero criterion: a curve of genus zero carrying a divisor of degree one is isomorphic to the projective line.

The third block computes on the projective line and carries an appendix on vector bundles. Divisors on Pk1 are classified up to linear equivalence by their degree, so that the degree homomorphism induces an isomorphism Pic⁡(Pk1)≅Z sending O(d) to d. On the projective line the Euler-characteristic form of Riemann–Roch becomes an identity between explicitly known numbers. The appendix proves the Birkhoff–Grothendieck splitting theorem: a finite locally free sheaf of rank r≥1 on Pk1 is a direct sum of line bundles O(a1)⊕⋯⊕O(ar), and the multiset of degrees is determined by the sheaf. The route passes through three lemmas — a nonzero morphism from an invertible sheaf to a finite locally free sheaf on an integral scheme is injective; a nonzero vector bundle on the projective line has a line subbundle of maximal degree; and the quotient by such a maximal line subbundle is again finite locally free — together with the twisting computation showing that extensions of line bundles whose summand degrees are at most the subbundle degree split.

The final block records the degree-zero and positivity statements. An effective divisor of degree zero is empty; an invertible sheaf of degree zero admitting a nonzero global section is trivial; and a degree-zero invertible sheaf that is not trivial has no nonzero global section, with the plane-cubic instance discharging the promise recorded by the counterexample of the preceding pair. The index of speciality i(D)=h1(D) is defined and the theorem is restated as ℓ(D)−i(D)=deg⁡k(D)+1−g, so that a divisor is nonspecial exactly when the Riemann inequality is an equality, and the Riemann inequality is strict by exactly i(D) otherwise. When the complete linear system ∣D∣ is nonempty, its dimension is dim⁡k∣D∣=ℓ(D)−1=deg⁡k(D)−g+i(D). In a fixed ample direction, sufficiently positive divisors are nonspecial, and Riemann's theorem computes their spaces of sections. A closing remark states precisely which classical thresholds — deg⁡KC=2g−2, h0(C,KC)=g, vanishing above degree 2g−2, base-point-freeness at degree 2g and very ampleness at degree 2g+1 — are not available on this page and belong to the following duality pair.

Several items inherit the Axiom of Choice from the published sheaf-cohomology, proper-cohomology and ampleness suppliers, to which the finite-dimensionality, vanishing and ample-direction arguments appeal; each such item names the supplier and carries the assumption in its contract. The local calculations use at most finitely many selections. Their items retain the assumptions of the actual divisor, DVR, coherence, twisting-sheaf and affine-correspondence suppliers they invoke.

3 · Logical flowchart

4 · Definitions, theorems and proofs

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

The Riemann-Roch dimension l(D)

Definition

Assume the Axiom of Choice (The Axiom of Choice). It supplies the Dependent Choice premise of the current curve Cartier-to-Weil result by AC implies DC implies countable choice. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field) and let D be a divisor on C (Divisors on a smooth proper curve) with associated invertible sheaf OC(D). For every such k, C and D, the dimensions below are finite; define l(D):=dim⁡kL(D)=dim⁡kH0(C,OC(D))=h0(D) to be the dimension of the Riemann-Roch space, and define hi(D):=dim⁡kHi(C,OC(D)),i≥0, for the dimensions of the sheaf cohomology groups (Sheaf cohomology as right derived global sections, Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis).

For finiteness, OC(D) is invertible by Invertible sheaf of cartier divisor, The sheaf of a Cartier divisor is invertible and Cartier and Weil divisors agree on a smooth curve. An invertible sheaf is locally free of rank one, hence quasi-coherent and of finite type. The curve is locally Noetherian because it is of finite type over the Noetherian field k; therefore a finite-type quasi-coherent sheaf on it is coherent (Invertible sheaves, Locally free sheaves of finite rank, Quasi-coherent module on a scheme, Finite type and finitely presented module sheaves, Locally finite type and finite type morphisms, A field has only the zero ideal and itself, hence is Noetherian, Every algebra of finite type over a Noetherian ring is a Noetherian ring, Locally Noetherian and Noetherian schemes, Coherent sheaves on a locally Noetherian scheme). The published proper coherent-cohomology theorem then makes every Hi(C,OC(D)) finite-dimensional over k (Finite-dimensional coherent cohomology over a field). Thus l(D) and each hi(D) are always nonnegative integers in this setting. For the zero divisor, l(0)=dim⁡kH0(C,OC)=1, because H0(C,OC) is canonically k (Functions on a proper curve); more generally l(D) depends only on the divisor D through its associated invertible sheaf.

The current The space L(D) supplies the order-defined space L(D) and identifies it with H0(C,OC(D)), using the rational-section dictionary Rational sections of line bundles are Cartier divisors. The Cartier-to-Weil route requires Dependent Choice, which the stated Axiom of Choice supplies through AC implies DC implies countable choice. This finiteness route uses the published proper cohomology result directly and does not depend on the later Riemann-Roch-space finite-dimensionality lemma.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

Finite-dimensionality of the Riemann-Roch space

Statement

Assume the Axiom of Choice as inherited from the proper finiteness theorem. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field) and let D be a divisor on C. Then:

  1. the invertible sheaf OC(D) is a coherent OC-module (Coherent module sheaves);
  2. L(D)=H0(C,OC(D)) is a finite-dimensional k-vector space;
  3. Hq(C,OC(D)) is a finite-dimensional k-vector space (Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis) for every q≥0, and vanishes for every q≥2;
  4. the Euler characteristic χ(C,OC(D))=h0(D)−h1(D) is defined (Euler characteristic of a coherent sheaf);
  5. consequently l(D):=dim⁡kL(D)=h0(D) is a nonnegative integer.

The divisor D is first identified as a Cartier divisor by Cartier and Weil divisors agree on a smooth curve. The associated sheaf is constructed by Invertible sheaf of cartier divisor and proved invertible by The sheaf of a Cartier divisor is invertible. The Riemann-Roch space and its identification with H0(C,OC(D)) are supplied by The space L(D), using the rational-section dictionary Rational sections of line bundles are Cartier divisors; these are the inputs used in [F8]. The stated Axiom of Choice supplies the Dependent Choice premise of the curve Cartier-to-Weil result through AC implies DC implies countable choice.

Facts & Assumptions

Given: a field k, a smooth proper geometrically integral curve C over k, and a divisor D on C.

[F1]

A curve over k is geometrically integral, separated and of finite type over k, and its underlying topological space has chain dimension one; being geometrically integral, C is an integral k-scheme, so it is nonempty, reduced and irreducible, and the adjectives smooth and proper mean that the structure morphism C→Spec⁡k is smooth and proper (Curves over a field, Integral schemes).

[F2]

For an integral k-scheme of finite type with underlying space of chain dimension one, the underlying space is Noetherian, and every proper closed subset is a finite set of closed points; the chain dimension is the Krull dimension of Chain dimension and the empty-space convention, so a curve has dimension at most one in the sense required for vanishing theorems (Proper closed subsets of a curve are finite, Chain dimension and the empty-space convention).

[F3]

If X is a Noetherian topological space with dim⁡X≤d for an integer d≥0, then Hq(X,F)=0 for every sheaf of abelian groups F on X and every integer q>d (Grothendieck vanishing on a Noetherian space).

[F4]

If X is a scheme proper over a field k and F a coherent OX-module, then Hq(X,F) is a finite-dimensional k-vector space for every q≥0, and only finitely many of the groups are nonzero: for a finite affine open cover of X with n members, Hq(X,F)=0 for every q≥n (Finite-dimensional coherent cohomology over a field).

[F5]

An invertible OX-module is locally free of rank one, and a locally free module is quasi-coherent; a locally free module of rank r is of finite type, since on a chart E∣U≅OUr=Ar~ with Ar a finitely generated module; on a locally Noetherian scheme a quasi-coherent module is coherent if and only if it is of finite type (Invertible sheaves, Locally free sheaves of finite rank, Quasi-coherent module on a scheme, Finite type and finitely presented module sheaves, Coherent sheaves on a locally Noetherian scheme).

[F6]

A scheme is locally Noetherian if it has an affine open cover by spectra of Noetherian rings; a morphism locally of finite type provides, around every point, an affine chart Spec⁡B with B a finitely generated algebra over the coordinate ring of an affine open of the target; a field is a Noetherian ring, and a finitely generated algebra over a Noetherian ring is Noetherian (Locally finite type and finite type morphisms, Locally Noetherian and Noetherian schemes, A field has only the zero ideal and itself, hence is Noetherian, Every algebra of finite type over a Noetherian ring is a Noetherian ring).

[F7]

For a coherent module F on a scheme proper over k whose cohomology is finite-dimensional with only finitely many nonzero groups, the Euler characteristic χ(X,F)=∑q≥0(−1)qdim⁡kHq(X,F) is an integer; the dimension dim⁡k of a finite-dimensional k-vector space is a nonnegative integer, and Hq denotes sheaf cohomology of the underlying sheaf of abelian groups (Euler characteristic of a coherent sheaf, Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis, Sheaf cohomology as right derived global sections).

[F8]

Weil-to-Cartier, invertibility, and the Riemann-Roch space. The divisor D is a Weil divisor on the smooth curve, so Cartier and Weil divisors agree on a smooth curve identifies it with a Cartier divisor. The local-equation construction of Invertible sheaf of cartier divisor defines OC(D), and The sheaf of a Cartier divisor is invertible proves it is invertible. The actual definition The space L(D) identifies L(D)={f∈k(C)×:div⁡(f)+D≥0}∪{0} with the image of H0(C,OC(D)) in k(C), using Rational sections of line bundles are Cartier divisors. These interfaces supply the uses at steps 1.3 and 5.1.

[F9]

The Axiom of Choice enters through the proper finiteness theorem [F4], the coherence and Noetherian suppliers [F5] and [F6], and the choice premises of the curve Cartier-to-Weil route [F8]. In ZF, AC implies DC by AC implies DC implies countable choice, so the DC premise of [F8] is available from the stated assumption (The Axiom of Choice, The axiom of dependent choice: a relation in which every element is related to something admits an N-indexed chain); no further selection is made below.

Proof

technique · direct; combine the coherence of $\mathcal O_C(D)$ on the locally Noetherian curve with the proper finiteness theorem for all $q$ and the Noetherian-dimension vanishing theorem for $q\ge2$
1.1F1F2

The curve has the required global shape. By [F1] the curve C is an integral k-scheme of finite type whose structure morphism is proper, and by [F2] its underlying space is Noetherian of dimension at most one; in particular C is nonempty.

1.2F1F6

The curve is locally Noetherian. Let x∈C be a point. Since C→Spec⁡k is of finite type, [F6] gives an affine open neighbourhood U=Spec⁡B of x with B a finitely generated k-algebra; the field k is Noetherian and a finitely generated algebra over a Noetherian ring is Noetherian, so B is Noetherian by [F6]. Therefore C has an affine open cover by spectra of Noetherian rings, i.e. C is locally Noetherian.

1.3F5F8

The associated sheaf is invertible. The given divisor is a Weil divisor; by [F8] the curve Cartier-to-Weil result first realizes it as a Cartier divisor. The local-equation construction then gives OC(D), and [F8] supplies its invertibility; by [F5] it is locally free of rank one.

2.1F5step 1.3

The associated sheaf is quasi-coherent of finite type. By [F5] a locally free module is quasi-coherent, and of finite type because its charts are free modules of finite rank; hence OC(D) is a quasi-coherent OC-module of finite type.

2.2F3step 1.1

Vanishing above degree one. By step 1.1 the underlying space of C is Noetherian of dimension at most one, so [F3] with d=1 gives Hq(C,OC(D))=0 for every integer q>1, that is, for every q≥2.

3.1F5step 1.2step 2.1

The associated sheaf is coherent. By step 1.2 the curve C is locally Noetherian, so [F5] applies in the form: a quasi-coherent module of finite type on a locally Noetherian scheme is coherent. With step 2.1, OC(D) is a coherent OC-module.

4.1F4step 1.1step 3.1

Finite-dimensionality in every degree. Apply [F4] to the scheme C proper over the field k and the coherent OC-module OC(D) of step 3.1: for every q≥0 the k-vector space Hq(C,OC(D)) is finite-dimensional, and only finitely many of these groups are nonzero.

5.1F8step 4.1

The Riemann-Roch space is the space of global sections. By [F8], the divisor space L(D) is identified with H0(C,OC(D)) as k-subspaces of k(C); hence L(D) is a finite-dimensional k-vector space by step 4.1, and l(D)=dim⁡kL(D)=h0(D) with h0(D)=dim⁡kH0(C,OC(D)).

5.2F7step 4.1step 2.2

The Euler characteristic. By [F7] the Euler characteristic of the coherent module OC(D) on the proper k-scheme C is the alternating sum ∑q≥0(−1)qdim⁡kHq(C,OC(D)) of finite dimensions, an integer; by step 2.2 only q=0 and q=1 contribute, so χ(C,OC(D))=h0(D)−h1(D) with both terms finite-dimensional by step 4.1.

6.1F7step 4.1step 5.1

The integer l(D). By step 5.1 l(D)=dim⁡kL(D)=dim⁡kH0(C,OC(D))=h0(D); this is the dimension of the finite-dimensional k-vector space H0(C,OC(D)) of step 4.1, hence a nonnegative integer by [F7].

7.1F4F5F6F8F9step 3.1step 4.1step 2.2step 5.1step 5.2step 6.1∎

Conclusion and choice accounting. Step 3.1 establishes (1), steps 4.1 and 2.2 establish (3), step 5.1 establishes (2), step 5.2 establishes (4) and step 6.1 establishes (5). The Axiom of Choice is used only through the proper finiteness theorem [F4], the suppliers of [F5] and [F6], and the flagged suppliers of [F8], as recorded in [F9]; the argument above makes no further selection.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Monotonicity of L(D) in the divisor

Statement

Assume the Axiom of Choice as inherited from the local-DVR, divisor and finite-dimensionality suppliers. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field) and let D≤E be divisors on C (Divisors on a smooth proper curve), so that E−D is effective. Then L(D)⊆L(E) as k-subspaces of the function field k(C); equivalently, the natural morphism of invertible subsheaves of the constant sheaf of rational functions OC(D)→OC(E) is injective.

If in addition E=D+p for a single closed point p, choose a uniformizer t of OC,p and put a=np(D). The canonical evaluation map from L(D+p)=H0(C,OC(D+p)) to the fiber OC(D+p)∣p has, in the frame t−a−1, the coordinate L(D+p)⟶κ(p),f⟼ta+1f mod (t). In this chosen coordinate its kernel is L(D), so the quotient L(D+p)/L(D) embeds k-linearly into κ(p). The coordinate map depends on the chosen uniformizer; the kernel and dimension bound do not. Consequently dim⁡kL(D+p)/L(D)≤[κ(p):k].

In particular l(D)≤l(E)≤l(D)+deg⁡k(E−D) for all divisors D≤E (The Riemann-Roch dimension l(D), Degree divisor proper curve).

The current interfaces The space L(D), Principal weil divisor and class group, Invertible sheaf of cartier divisor and Cartier and Weil divisors agree on a smooth curve supply the spaces, divisors and sheaves used below. The Cartier-to-Weil route requires Dependent Choice, which the stated Axiom of Choice supplies through AC implies DC implies countable choice. The rational-section identification uses Rational sections of line bundles are Cartier divisors.

Facts & Assumptions

Given: the Axiom of Choice inherited from the local-DVR, divisor and finite-dimensionality suppliers; a field k, a smooth proper geometrically integral curve C over k, and divisors D≤E on C.

[F1]

A divisor on C is a finite formal sum D=∑xnx[x] over the closed points of C with integer coefficients; D≤E means that the coefficients satisfy nx≤mx for all x, equivalently that E−D is effective; the degree is additive, deg⁡k(E−D)=∑x(mx−nx)[κ(x):k], the sum over the finite support, and each residue field κ(x) is a finite extension of k with [κ(x):k]=dim⁡kκ(x)≥1 (Divisors on a smooth proper curve, Degree divisor proper curve, Divisor support positive negative parts).

[F2]

The current The space L(D) identifies L(D)={f∈k(C)×:div⁡(f)+D≥0}∪{0} as a k-subspace of k(C) with the image of H0(C,OC(D)). It uses the principal Weil divisor interface Principal weil divisor and class group, the local-equation construction Invertible sheaf of cartier divisor, and the curve Cartier-to-Weil identification Cartier and Weil divisors agree on a smooth curve. The latter's Dependent Choice premise is supplied by AC through AC implies DC implies countable choice. The rational-section dictionary Rational sections of line bundles are Cartier divisors identifies the global sections with the stated rational functions. These interfaces give the section and stalk descriptions used below.

[F3]

For a normal locally Noetherian integral scheme the order along a prime divisor is a group homomorphism ord⁡Z:K(X)×→Z with ord⁡Z(fg)=ord⁡Z(f)+ord⁡Z(g) and ord⁡Z(f−1)=−ord⁡Z(f), and on an integral scheme ord⁡Z(f)≥0 if and only if f lies in the local ring, with equality to zero exactly for units (Order codimension one rational function). The closed points of the smooth curve C are its codimension-one points (Divisors on a smooth proper curve). For the order inequalities below only, put ord⁡x(0)=+∞; zero belongs to every L(D) by [F2].

[F4]

The local ring OC,p of a closed point of the smooth curve C is a discrete valuation ring with maximal ideal generated by a uniformizer t; every nonzero f∈k(C)× is f=tmu with m=ord⁡p(f)∈Z and u a unit, and κ(p)=OC,p/(t) is a field, the residue field, of k-dimension [κ(p):k] (Local rings at closed points of smooth curves are discrete valuation rings, The residue field at a point of an affine scheme, Degree divisor proper curve).

[F5]

L(D)=H0(C,OC(D)) is a finite-dimensional k-vector space and l(D)=dim⁡kL(D)=h0(D) is a nonnegative integer; the same holds with D replaced by D+p (Finite-dimensionality of the Riemann-Roch space, The Riemann-Roch dimension l(D), Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis).

[F6]

Linear algebra over k: for a linear map T:V→W with V finite-dimensional, dim⁡kV=dim⁡kker⁡T+dim⁡kim⁡T (Rank-nullity: dim⁡FV=nullity⁡T+rank⁡T); the formula T~(v+ker⁡T)=T(v) defines a linear isomorphism V/ker⁡T→im⁡T (First isomorphism theorem for vector spaces: V/ker⁡T is isomorphic to im⁡T); a subspace of a finite-dimensional space has dimension at most that of the ambient space (If dim⁡FV=n and U is a linear subspace of V, then U is finite-dimensional, dim⁡FU≤n, and dim⁡FU=n if and only if U=V); and for W≤V with V finite-dimensional, dim⁡k(V/W)=dim⁡kV−dim⁡kW (A quotient basis lifts to a basis adapted to W).

[F7]

The Axiom of Choice enters through the DVR supplier of [F4], the finiteness suppliers of [F5] and the Cartier-to-Weil route of [F2]. In ZF, AC implies DC by AC implies DC implies countable choice, supplying the DC premise of Cartier and Weil divisors agree on a smooth curve. The chosen uniformizer in step 2.3 only specifies a coordinate on the fiber; the kernel is independent of it, and no additional choice principle is used (The Axiom of Choice, The axiom of dependent choice: a relation in which every element is related to something admits an N-indexed chain).

Proof

technique · direct; read $L(D)$ through orders of vanishing at closed points, prove the containment and the one-point kernel computation via the local uniformizer, and iterate the one-point bound over the finite support of $E-D$
1.1F1

Set-up and coefficients. By [F1] write D=∑xnx[x] and E=∑xmx[x] with nx≤mx for every closed point x, both sums having finite support, and write cx:=mx−nx≥0 for the coefficient of E−D; the degree is deg⁡k(E−D)=∑xcx[κ(x):k].

1.2F2F3

Order description of the Riemann-Roch space. By [F2], for f∈k(C)× one has f∈L(D) if and only if div⁡(f)+D≥0, and by [F3] this is equivalent to the coefficientwise condition ord⁡x(f)+nx≥0 for every closed point x; moreover L(D) is a k-subspace of k(C) and 0∈L(D).

2.1F2step 1.2

The sheaf picture. By [F2] the attached invertible sheaves are subsheaves OC(D)⊆OC(E)⊆KC of the constant sheaf of rational functions, with stalks cut out by the very order conditions of step 1.2 and with H0(C,OC(D))=L(D) and H0(C,OC(E))=L(E).

2.2F2step 1.1step 1.2

Monotonicity of the spaces. Let f∈L(D). By step 1.2, ord⁡x(f)+nx≥0 for every closed point x; since nx≤mx by step 1.1, also ord⁡x(f)+mx≥0 for every x, so f∈L(E) by step 1.2 again. Hence L(D)⊆L(E) as subsets of k(C), and both are k-subspaces by [F2].

2.3F3F4step 1.2

The one-point case: the evaluation map. Now let E=D+p for a single closed point p, and let a:=np be the coefficient of D at p. Choose a uniformizer t of the discrete valuation ring OC,p. The canonical evaluation of sections of OC(D+p) at p has target fiber OC(D+p)∣p; in the local frame t−a−1, its coordinate is φt(f)=ta+1f mod (t)∈κ(p). This coordinate description depends on t, but is well defined: f∈L(D+p) gives ord⁡p(f)≥−a−1, hence ord⁡p(ta+1f)≥0 and ta+1f∈OC,p by [F3] and [F4]. The map φt:L(D+p)→κ(p) is k-linear, since multiplication by ta+1 and reduction modulo (t) are k-linear on OC,p.

2.4F3F4step 1.2

Its kernel is L(D). If f∈L(D) then ord⁡p(f)≥−a, so ta+1f∈(t), and φ(f)=0. Conversely, if φ(f)=0 then ta+1f∈(t), that is, ord⁡p(ta+1f)≥1, so ord⁡p(f)≥−a by [F3]; for points x≠p the conditions f∈L(D) and f∈L(D+p) coincide because D and D+p have the same coefficients away from p; hence f∈L(D) by step 1.2. Therefore ker⁡φ=L(D).

3.1F2F5F6step 2.1step 2.2

The morphism of invertible subsheaves. Both OC(D) and OC(E) are invertible subsheaves of the constant sheaf KC by [F2]; the stalkwise inclusion of step 2.1, given by the coefficientwise comparison of step 2.2, defines the natural morphism OC(D)→OC(E), which is injective on every stalk and hence on sections; the induced map on global sections is the inclusion L(D)⊆L(E), and the dimension formula l(D)≤l(E) follows from [F5] and [F6].

3.2F4F5F6step 2.4

The quotient embeds in the residue field. By step 2.4 and [F6] the first isomorphism theorem gives a k-linear isomorphism L(D+p)/L(D)→im⁡φ, so L(D+p)/L(D) embeds k-linearly into κ(p); since L(D+p) is finite-dimensional by [F5], rank-nullity together with the quotient formula gives dim⁡kL(D+p)/L(D)=dim⁡kL(D+p)−dim⁡kL(D)=dim⁡kim⁡φ≤dim⁡kκ(p)=[κ(p):k], the inequality by subspace monotonicity and the last equality by [F4]. In particular l(D+p)=dim⁡kL(D+p)≤l(D)+[κ(p):k].

4.1F1step 1.1step 3.2

Iteration over the support. Enumerate the finite support of E−D as points x1,…,xr and let ci=cxi≥1 be the multiplicities of step 1.1. Consider the finite chain of divisors starting at D and adding one copy of [xi] at a time, ci times for each i, ending at E; every successive difference is a single closed point q, so step 3.2 applied to the pair of consecutive divisors gives an increase of l by at most [κ(q):k]. Summing the r chains of inequalities gives l(E)≤l(D)+∑ici[κ(xi):k]=l(D)+deg⁡k(E−D) by step 1.1.

5.1F7step 2.2step 3.1step 3.2step 4.1∎

Conclusion and choice accounting. Step 2.2 gives L(D)⊆L(E) and step 3.1 the injective morphism of invertible subsheaves together with l(D)≤l(E); steps 2.3 and 2.4 identify L(D) as the kernel of the evaluation L(D+p)→κ(p), step 3.2 embeds the quotient in κ(p) with the bound dim⁡kL(D+p)/L(D)≤[κ(p):k], and step 4.1 gives l(E)≤l(D)+deg⁡k(E−D) in general. The Axiom of Choice is used only through the suppliers recorded in [F7], namely the DVR structure of [F4], the finiteness results of [F5], and the Cartier-to-Weil route of [F2]; no further selection is made above.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

The exact sequence for adding one point to a divisor

Statement

Assume the Axiom of Choice, inherited through the sheaf-cohomology and Cartier-divisor suppliers of this page. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field), let D be a divisor on C (Divisors on a smooth proper curve), let p∈C be a closed point with residue field κ(p) and residue degree d:=[κ(p):k] (The residue field at a point of an affine scheme, Degree divisor proper curve), and let ip,∗κ(p) be the skyscraper sheaf at p with value κ(p) (A skyscraper sheaf of abelian groups at a point).

  1. The natural morphism OC(D)→OC(D+p) of invertible subsheaves of the constant sheaf of rational functions is injective (Monotonicity of L(D) in the divisor), and it fits into a short exact sequence of coherent OC-modules 0→OC(D)→OC(D+p)→ip,∗κ(p)→0; thus the cokernel of the inclusion is the skyscraper sheaf at p with value κ(p) (Coherent module sheaves). This realises the promised third term i∗(OC(D+p)∣p): the restriction of the invertible sheaf OC(D+p) to the closed point p is a one-dimensional κ(p)-vector space, and the pushforward of that value is the skyscraper sheaf of the exact sequence.
  2. H0(C,ip,∗κ(p))≅κ(p), so dim⁡kH0(C,ip,∗κ(p))=d, and Hq(C,ip,∗κ(p))=0 for every q≥1 (Sheaf cohomology as right derived global sections).
  3. Iterating: for every effective divisor E≥0 on C the inclusion OC(D)→OC(D+E) has cokernel QE fitting into a short exact sequence of coherent OC-modules 0→OC(D)→OC(D+E)→QE→0 with (QE)x=0 for every closed point x∉Supp⁡(E); moreover Hq(C,QE)=0 for every q≥1 and dim⁡kH0(C,QE)=deg⁡k(E), so in particular 0≤dim⁡kH0(C,QE)≤deg⁡k(E), the upper bound being the one promised (Divisor support positive negative parts, Effective divisors have nonnegative degree).

The sheaves OC(D′) used here are constructed from the actual Weil-to-Cartier and Cartier-sheaf interfaces in [F2]. Under the stated Axiom of Choice, [F12] supplies the Dependent Choice premise of the curve Cartier-to-Weil result. The open-set order description in [F2] is stated for nonempty opens; the section group on the empty open is zero.

Facts & Assumptions

Given: a field k, a smooth proper geometrically integral curve C over k, a divisor D on C, a closed point p∈C, and, for part (3), an effective divisor E on C.

[F1]

Divisors and degrees. A divisor on C is a finite formal sum D=∑xnx[x] over the closed points, Supp⁡(E) is the finite set of closed points with nonzero coefficient, D is effective when all coefficients are ≥0, the residue field κ(x) of a closed point is a finite extension of k with [κ(x):k]=dim⁡kκ(x)≥1, and deg⁡k(E)=∑xnx[κ(x):k]; C is geometrically integral, separated and of finite type over k, and it is nonempty (Divisors on a smooth proper curve, Divisor support positive negative parts, Degree divisor proper curve, The residue field at a point of an affine scheme, Curves over a field). Moreover deg⁡k(E)≥0 for effective E, with deg⁡k(E)=0 exactly for E=0 (Effective divisors have nonnegative degree).

[F2]

Weil divisors, their Cartier sheaves, and the order description. In order inequalities below, use the convention ord⁡x(0)=+∞; this is notation for the zero section, not an extension of the valuation homomorphism domain. The divisor D′ on the smooth curve is a Weil divisor. By Cartier and Weil divisors agree on a smooth curve it is represented by a Cartier divisor whose cycle is D′. The local-equation construction of Invertible sheaf of cartier divisor gives the subsheaf OC(D′)⊆KC, and The sheaf of a Cartier divisor is invertible proves it invertible. The generic sheaf KC is constant with value k(C) (Sheaf total quotient rings). For a nonempty open U⊆C, its sections identify with k(C); a rational function f is a section of OC(D′) on U exactly when it belongs to the stalk at every point of U. At the generic point the stalk is k(C) and imposes no condition. At each closed point x, the local equation has order nx(D′) by the cycle identification, so the DVR stalk condition is ord⁡x(f)+nx(D′)≥0 (Local rings at closed points of smooth curves are discrete valuation rings, Order codimension one rational function). Locality of the subsheaf then gives OC(D′)(U)={f∈k(C):ord⁡x(f)+nx(D′)≥0 for every closed point x∈U}. For U=∅, OC(D′)(U)=0. Thus D′≤D′′ gives the inclusion of these subsheaves and the stated closed-point stalk descriptions. The global-section instance is also the identification of The space L(D) using the rational-section dictionary Rational sections of line bundles are Cartier divisors.

[F3]

Local structure at p. The local ring OC,p is a discrete valuation ring with maximal ideal generated by a uniformizer t, so that κ(p)=OC,p/(t) and every nonzero element of OC,p is a unit times a power of t (Local rings at closed points of smooth curves are discrete valuation rings). For f∈k(C)× the order ord⁡p(f) is additive, ord⁡p(f)≥0 if and only if f∈OC,p, and ord⁡p(f)=0 if and only if f is a unit of OC,p (Order codimension one rational function); consequently tmf∈OC,p if and only if ord⁡p(f)≥−m, and tmf∈tOC,p if and only if ord⁡p(f)≥−m+1, for every m∈Z.

[F4]

Monotonicity supplier. For divisors D≤E the natural morphism of invertible subsheaves OC(D)→OC(E) is injective, and L(D)⊆L(E); for E=D+p the quotient L(D+p)/L(D) embeds in κ(p) with dimension at most [κ(p):k] (Monotonicity of L(D) in the divisor). Only the injectivity assertion is used below.

[F5]

Exactness and stalks of sheaves. A sequence of sheaves of OC-modules is exact exactly when it is exact in the abelian category of sheaves of abelian groups, in the sense of Exact sequences of sheaves; the kernel sheaf of a morphism is computed objectwise while the cokernel sheaf is the sheafification of the objectwise cokernel, with the same formulas taken in the module categories on each open set, so cokernels and exactness of OC-module morphisms are computed on underlying sheaves of abelian groups (Kernel sheaves are objectwise, while cokernels and images are sheafified). Since kernels are objectwise, stalks are filtered colimits of section groups (The stalk of a presheaf at a point) and filtered colimits of abelian groups are exact (Filtered colimits of abelian groups are exact), the stalk of the kernel of a morphism is the kernel of the stalk map, and by the same exactness and Sheafification preserves stalks the stalk of the cokernel is the cokernel of the stalk map. Consequently a sequence of sheaves of abelian groups is exact if and only if it is exact on every stalk (A sequence of abelian sheaves is exact exactly when it is exact on every stalk), and the category of sheaves of abelian groups is abelian (Sheaves of abelian groups, and likewise sheaves of modules on a ringed space, form abelian categories).

[F6]

Abelian-category algebra. For a morphism f:A→B of an abelian category there is a canonical isomorphism A/ker⁡(f)≅im⁡(f) (First isomorphism theorem in an abelian category), and for subobjects C≤B≤A there is a canonical isomorphism (A/C)/(B/C)≅A/B (Third isomorphism theorem in an abelian category); the quotient A/B is the cokernel of the representing monomorphism B↣A (The quotient of an object by a subobject).

[F7]

Skyscraper sheaves. For a point x∈X of a topological space and an abelian group A, the skyscraper sheaf ix,∗A has (ix,∗A)(V)=A for x∈V with identity restrictions and value 0 for x∉V (A skyscraper sheaf of abelian groups at a point). For a closed subset Z⊆X with the subspace topology and inclusion i:Z↪X, the direct image i∗F of a sheaf F of abelian groups on Z is given by (i∗F)(V)=F(i−1V) (Direct image of a sheaf along a continuous map, Subspace topology: the traces of the open sets, its closed sets and its bases, the continuity of the inclusion, and the characteristic property of a map into a subspace), and Hq(Z,F)≅Hq(X,i∗F) for every q≥0 (Pushforward along a closed immersion preserves sheaf cohomology). On the one-point space Z={∗} one has H0(Z,F)≅F(Z) and Hq(Z,F)=0 for every q>0 (A point has no higher sheaf cohomology).

[F8]

Long exact sequence. For every short exact sequence of abelian sheaves on C there is a natural long exact sequence of sheaf cohomology groups ⋯→Hq(C,F′)→Hq(C,F)→Hq(C,F′′)→Hq+1(C,F′)→⋯ (Long exact sequence of sheaf cohomology, Sheaf cohomology as right derived global sections).

[F9]

Coherence. A curve is of finite type over the field k (Curves over a field); a morphism of finite type provides affine charts Spec⁡B with B a finitely generated algebra over the coordinate ring of an affine open of the target (Locally finite type and finite type morphisms), a field is Noetherian (A field has only the zero ideal and itself, hence is Noetherian) and a finitely generated algebra over a Noetherian ring is Noetherian (Every algebra of finite type over a Noetherian ring is a Noetherian ring), so C is locally Noetherian: it has an affine open cover by spectra of Noetherian rings (Locally Noetherian and Noetherian schemes). For every divisor D′ the sheaf OC(D′) is a coherent OC-module (Finite-dimensionality of the Riemann-Roch space), and on a locally Noetherian scheme the kernel, image and cokernel of a morphism of coherent OC-modules are coherent (Coherent sheaves on a locally Noetherian scheme, Coherent module sheaves).

[F10]

Linear algebra over k. For a linear map T:V→W with V finite-dimensional, dim⁡kV=dim⁡kker⁡T+dim⁡kim⁡T (Rank-nullity: dim⁡FV=nullity⁡T+rank⁡T); the formula T~(v+ker⁡T)=T(v) defines an isomorphism V/ker⁡T→im⁡T (First isomorphism theorem for vector spaces: V/ker⁡T is isomorphic to im⁡T); for W≤V with V finite-dimensional, dim⁡k(V/W)=dim⁡kV−dim⁡kW (A quotient basis lifts to a basis adapted to W); and dim⁡k of a finite-dimensional k-vector space is a nonnegative integer (Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis). Hence for an exact sequence of k-vector spaces 0→A→B→C→0 with A and C finite-dimensional one has dim⁡kB=dim⁡kA+dim⁡kC.

[F11]

The Axiom of Choice enters through the sheaf-cohomology suppliers of [F7] and [F8], the coherence supplier [F9], the curve Cartier-to-Weil supplier in [F2], and the local-DVR supplier [F3]. The only additional premise needed there is Dependent Choice, which follows from the stated Axiom of Choice by [F12]; the argument below makes no further selection (The Axiom of Choice, The axiom of dependent choice: a relation in which every element is related to something admits an N-indexed chain).

[F12]

In ZF, the Axiom of Choice implies Dependent Choice (AC implies DC implies countable choice). Hence the stated Choice assumption supplies the Dependent Choice premise of Cartier and Weil divisors agree on a smooth curve.

Proof

technique · direct; construct the evaluation morphism carrying $\mathcal O_C(D+p)$ onto the skyscraper at $p$ with kernel $\mathcal O_C(D)$, verify the resulting sequence stalkwise, compute the cohomology of the skyscraper by pushing forward from the one-point space, and iterate the single-point sequence along the finite support of an effective divisor
1.1F1F3

Setup and local structure at p. Write D=∑xnx[x], put a:=np(D), and recall d=[κ(p):k]=dim⁡kκ(p) by [F1]; the divisor D+p has coefficient a+1 at p and the same coefficient as D at every other closed point. By [F3], OC,p is a discrete valuation ring with maximal ideal (t) for a uniformizer t, the residue field is κ(p)=OC,p/(t), ord⁡p is additive with ord⁡p(f)≥0 if and only if f∈OC,p, and tmf∈OC,p if and only if ord⁡p(f)≥−m for m∈Z.

1.2F9

The curve is locally Noetherian. By [F9] the structure morphism C→Spec⁡k is of finite type, so every point of C has an affine open neighbourhood Spec⁡B with B a finitely generated k-algebra; k is Noetherian and a finitely generated algebra over a Noetherian ring is Noetherian, so each such B is Noetherian, and C is locally Noetherian by [F9].

1.3F2F4

The two subsheaves and their order conditions. By [F2] the divisors D and D+p carry inclusions of invertible subsheaves OC(D)⊆OC(D+p)⊆KC of the constant sheaf of rational functions such that, for every nonempty open U, OC(D)(U) consists of the f∈k(C) with ord⁡x(f)+nx≥0 for every closed point x∈U, and likewise for OC(D+p) with the coefficient at p raised by one; their section groups on the empty open are zero. In particular, if p∈U and f∈OC(D+p)(U), then ord⁡p(f)≥−a−1. By [F4] the natural morphism OC(D)→OC(D+p) of OC-modules is injective, given by the inclusion of subsheaves of KC.

2.1F7step 1.1step 1.3

Definition of the evaluation morphism ψ. Define for every open U⊆C a map ψU:OC(D+p)(U)→(ip,∗κ(p))(U) by ψU:=0 when p∉U, and by ψU(f):=ta+1f mod (t)∈κ(p) when p∈U, where t is the uniformizer of step 1.1. This is well defined: for f∈OC(D+p)(U) with p∈U, step 1.3 gives ord⁡p(f)≥−a−1, hence ord⁡p(ta+1f)≥0, so ta+1f∈OC,p by step 1.1 and its class in κ(p)=OC,p/(t) is defined.

2.2F7step 1.1

Cohomology of the skyscraper. Let Z={p} be the one-point space with the subspace topology, with inclusion i:Z↪C, and let F be the sheaf of abelian groups on Z with F(Z)=κ(p); by [F7] the direct image is (i∗F)(V)=F(i−1V), which equals κ(p)=(ip,∗κ(p))(V) for opens V containing p and 0 otherwise, so i∗F=ip,∗κ(p) and Hq(C,ip,∗κ(p))≅Hq(Z,F) for every q≥0 by [F7]. By [F7] again H0(Z,F)≅F(Z)=κ(p) and Hq(Z,F)=0 for q>0; hence H0(C,ip,∗κ(p))≅κ(p), of k-dimension dim⁡kκ(p)=d by step 1.1, and Hq(C,ip,∗κ(p))=0 for every q≥1. This is part (2) of the Statement.

3.1F5F7step 1.1step 2.1

ψ is a morphism of sheaves of OC-modules. For every open U the map ψU is additive, because multiplication by ta+1 and reduction modulo (t) are additive on the groups involved; it is OC(U)-linear, because for g∈OC(U) the germ of g at p lies in OC,p with class gˉ∈κ(p) and ta+1(gf)=g⋅ta+1f holds in OC,p, so ψU(gf)=gˉ ψU(f); and the maps are compatible with restrictions: for V⊆U with p∉U both maps are zero, for p∈V⊆U the two subsheaves of KC have the same element f and ip,∗κ(p) has identity restrictions on opens containing p, while for p∈U, p∉V the target (ip,∗κ(p))(V) is the zero group. Hence ψ is a morphism of sheaves of OC-modules.

3.2F2F5step 1.1step 1.3step 2.1

The kernel of ψ is OC(D). At an open U with p∉U, step 1.3 gives OC(D)(U)=OC(D+p)(U) and ψU=0, so ker⁡ψU=OC(D)(U). At an open U with p∈U, step 1.1 shows that f∈ker⁡ψU is equivalent to ta+1f∈tOC,p, equivalently to ord⁡p(f)≥−a, and together with the order conditions at the other points of U, which are the same for D and D+p, this is exactly the condition f∈OC(D)(U); the converse is immediate. Therefore ker⁡ψ=OC(D) as subsheaves of OC(D+p).

3.3F2F3F7step 2.1

Stalks of the skyscraper and surjectivity of ψ. By [F7] the skyscraper ip,∗κ(p) has value κ(p) on the opens containing p, with identity restrictions, and value 0 on the opens not containing p; hence its stalk at p is κ(p), while at a point x≠p every section over an open containing x restricts to zero on the open complement of the closed set {p}, which still contains x, so the stalk at x is 0. The map ψx is a map into the zero group for x≠p, and at p the map ψp is surjective: for u∈OC,p the rational function t−a−1u satisfies ord⁡p(t−a−1u)≥−a−1, hence lies in OC(D+p)p by the stalk description of [F2], so it is the germ of a section of OC(D+p) near p, and ψp(t−a−1u)=u mod (t) because ψ is defined by f↦ta+1f mod (t) on sections over opens containing p as in step 2.1.

4.1F5step 1.3step 3.2step 3.3

Exactness of the single-point sequence. Consider the sequence of sheaves of OC-modules 0→OC(D)→OC(D+p)→ψip,∗κ(p)→0. At a point x≠p the stalk sequence is 0→Ax→Ax→0→0, exact because the two subsheaves agree away from p by step 1.3 and the target stalk is 0; at p the stalk sequence is 0→t−aOC,p→t−a−1OC,p→κ(p)→0 by steps 1.1 and 3.3, and it is exact: the first map is the inclusion of the subgroup t−aOC,p, so its kernel vanishes; the kernel of ψp is the image of that inclusion, because the kernel sheaf of ψ is OC(D) by step 3.2 and the stalk of a kernel is the kernel of the stalk map by [F5]; and ψp is surjective by step 3.3. By the stalkwise criterion for exactness [F5] the sequence is a short exact sequence, so the cokernel of the inclusion OC(D)→OC(D+p) is its third term ip,∗κ(p).

5.1F5F6F9step 1.2step 4.1

Coherence of the third term. By [F9] the curve C is locally Noetherian, and by [F9] the invertible sheaves OC(D) and OC(D+p) are coherent OC-modules; the present sequence is a short exact sequence of OC-modules whose left-hand map is a morphism of coherent modules with cokernel ip,∗κ(p), so the cokernel is a coherent OC-module by [F9]. This and step 4.1 give part (1) of the Statement, with the cokernel identified as the skyscraper at p with value κ(p).

5.2F5F6step 1.3step 4.1

The extension step for the iteration. Let p be a closed point with E≥p and put E′:=E−p; set A:=OC(D), B:=OC(D+E′) and C′:=OC(D+E)=OC(D+E′+p), so that A⊆B⊆C′ are nested subsheaves of KC by steps 1.3 with injective inclusion morphisms by [F4]. Define QE′:=coker⁡(A→B) and QE:=coker⁡(A→C′). By the third isomorphism theorem in an abelian category [F6], applied to the subobjects A≤B≤C′, the quotient QE/QE′ is canonically isomorphic to C′/B=coker⁡(B→C′), and by step 4.1 applied to the divisor D+E′ and the point p the latter is ip,∗κ(p); hence 0→QE′→QE→ip,∗κ(p)→0 is a short exact sequence of sheaves of OC-modules.

6.1F5step 1.3step 5.2

Support of QE. Let x be a closed point with x∉Supp⁡(E); then the coefficients of D and D+E at x coincide, so over an open neighbourhood of x the order conditions defining the two subsheaves of step 1.3 are the same and the inclusion A⊆C′ induces an isomorphism of stalks at x; since the stalk of a cokernel is the cokernel of the stalk map by [F5], the stalk (QE)x is the cokernel of an isomorphism and hence 0. Therefore QE is supported on the finite set Supp⁡(E).

6.2F9step 1.2step 5.2

Coherence of QE. The sheaves OC(D) and OC(D+E) are coherent OC-modules by [F9], the curve is locally Noetherian by step 1.2, and QE is their cokernel by step 5.2; hence QE is a coherent OC-module by [F9].

6.3F1F8F10step 2.2step 5.2

The induction on deg⁡kE. We prove by induction on the nonnegative integer deg⁡kE that dim⁡kH0(C,QE)=deg⁡k(E) and Hq(C,QE)=0 for every q≥1, for every effective divisor E. If deg⁡kE=0 then E=0 by [F1] and QE is the cokernel of the identity of OC(D), hence the zero sheaf, so both assertions hold. If E≠0, choose a closed point p∈Supp⁡(E) and put E′:=E−p, an effective divisor with deg⁡kE′=deg⁡kE−[κ(p):k]<deg⁡kE by [F1]; by step 5.2 there is a short exact sequence 0→QE′→QE→ip,∗κ(p)→0, whose long exact sequence [F8] contains the exact segment H0(QE′)→H0(QE)→κ(p)→H1(QE′)→H1(QE)→H1(ip,∗κ(p))=0, where Hq(ip,∗κ(p))=0 for q≥1 by step 2.2. Since H1(QE′)=0 by the induction hypothesis, the segment collapses to the exact sequence 0→H0(QE′)→H0(QE)→κ(p)→0, so [F10] gives dim⁡kH0(QE)=dim⁡kH0(QE′)+dim⁡kκ(p)=deg⁡kE′+[κ(p):k]=deg⁡kE; and for q≥1 the exactness of Hq(QE′)→Hq(QE)→Hq(ip,∗κ(p)), with both outer groups zero by the induction hypothesis and step 2.2, gives Hq(QE)=0.

7.1F2F3F7F8F9F11F12step 4.1step 5.1step 2.2step 6.1step 6.2step 6.3∎

Conclusion and choice accounting. Step 4.1 gives the short exact sequence of part (1) with its cokernel identified, and step 5.1 the coherence of its terms; step 2.2 gives part (2); and steps 6.1, 6.2 and 6.3 give, for every effective divisor E, the short exact sequence with cokernel QE, its support on Supp⁡(E), its coherence and the dimension formula dim⁡kH0(C,QE)=deg⁡k(E), hence the promised bounds 0≤dim⁡kH0(C,QE)≤deg⁡k(E). The Axiom of Choice is used only through the suppliers recorded in [F11], namely the sheaf-cohomology results of [F7] and [F8], the coherence supplier of [F9], the flagged Cartier-divisor suppliers of [F2], and the local-DVR supplier of [F3]; the only selections made above are the uniformizer t supplied by [F3] and the point p chosen in the finite set Supp⁡(E).

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

Euler characteristic changes by the residue degree

Statement

Assume the Axiom of Choice, inherited from the Euler-characteristic, finiteness and skyscraper suppliers below. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field), let D be a divisor on C (Divisors on a smooth proper curve) and let p∈C be a closed point with residue degree d=[κ(p):k] (Degree divisor proper curve). Then χ(C,OC(D+p))=χ(C,OC(D))+d, and more generally χ(C,OC(D+E))=χ(C,OC(D))+deg⁡k(E) for every effective divisor E≥0 on C, where χ is the Euler characteristic of coherent sheaves on the proper k-scheme C (Euler characteristic of a coherent sheaf). Both identities hold in Z.

The short exact sequence 0→OC(D)→OC(D+p)→ip,∗κ(p)→0, the cokernel QE of OC(D)→OC(D+E) together with dim⁡kH0(C,QE)=deg⁡k(E) and the higher vanishing, and the coherence of ip,∗κ(p) and of QE are supplied by The exact sequence for adding one point to a divisor. That current supplier states and proves the Cartier-to-Weil, Cartier-sheaf and local-order interfaces used to construct the sequence; its Cartier-to-Weil route obtains Dependent Choice from the Axiom of Choice by AC implies DC implies countable choice. This corollary uses the stated sequence interface and does not certify its separate suppliers.

Facts & Assumptions

Given: the Axiom of Choice inherited from the Euler-characteristic, finiteness and skyscraper suppliers; a field k, a smooth proper geometrically integral curve C over k, a divisor D on C, a closed point p∈C and an effective divisor E≥0 on C.

[F1]

The curve C is proper and geometrically integral over the field k, and deg⁡k is the group homomorphism on divisors with deg⁡k(D′)=∑xnx[κ(x):k] (Curves over a field, Divisors on a smooth proper curve, Degree divisor proper curve, Effective divisors have nonnegative degree).

[F2]

For every divisor D′ the sheaf OC(D′) is a coherent OC-module (Finite-dimensionality of the Riemann-Roch space, Coherent module sheaves).

[F3]

Adding one point: 0→OC(D)→OC(D+p)→ip,∗κ(p)→0 is a short exact sequence of coherent OC-modules, H0(C,ip,∗κ(p))≅κ(p) has k-dimension d=[κ(p):k] while Hq(C,ip,∗κ(p))=0 for every q≥1; and for every effective divisor E≥0 the cokernel QE of OC(D)→OC(D+E) is a coherent OC-module with Hq(C,QE)=0 for every q≥1 and dim⁡kH0(C,QE)=deg⁡k(E) (The exact sequence for adding one point to a divisor).

[F4]

The Euler characteristic of a coherent module F on a scheme X proper over k is χ(X,F)=∑q≥0(−1)qdim⁡kHq(X,F), a finite alternating sum of finite dimensions, and χ(X,0)=0 for the zero sheaf (Euler characteristic of a coherent sheaf, Sheaf cohomology as right derived global sections, Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis).

[F5]

If 0→F′→F→F′′→0 is a short exact sequence of coherent OX-modules on a scheme X proper over k, then χ(X,F)=χ(X,F′)+χ(X,F′′) (Euler characteristic is additive in short exact sequences).

[F6]

The Axiom of Choice is used exactly through the Euler-characteristic supplier [F4], the finiteness and coherence suppliers [F2] and the single-point sequence supplier [F3]; no further selection is made below (The Axiom of Choice).

Proof

technique · direct; apply additivity of the Euler characteristic to the single-point sequence, evaluate $\chi$ on the skyscraper through its cohomology, and repeat for the iterated cokernel
1.1F1F2F3

Set-up. By [F1] the curve C is proper over k, and by [F2] the sheaves OC(D) and OC(D+p) are coherent OC-modules; consequently all Euler characteristics below are those of [F4]. By [F3] the sequence 0→OC(D)→OC(D+p)→ip,∗κ(p)→0 is a short exact sequence of coherent OC-modules with H0(C,ip,∗κ(p))≅κ(p) of k-dimension d and Hq(C,ip,∗κ(p))=0 for every q≥1.

1.2F2F3F4F5

The general effective shift. Let E≥0 be effective. By [F2] the sheaves OC(D) and OC(D+E) are coherent, and by [F3] the cokernel QE of OC(D)→OC(D+E) is coherent with Hq(C,QE)=0 for q≥1 and dim⁡kH0(C,QE)=deg⁡k(E); hence χ(C,QE)=dim⁡kH0(C,QE)=deg⁡k(E) by the definition of χ in [F4]. Applying additivity [F5] to 0→OC(D)→OC(D+E)→QE→0 gives χ(C,OC(D+E))=χ(C,OC(D))+χ(C,QE)=χ(C,OC(D))+deg⁡k(E).

2.1F3F4step 1.1

The Euler characteristic of the skyscraper. By [F4] the Euler characteristic of the coherent module ip,∗κ(p) on the proper k-scheme C is the alternating sum ∑q≥0(−1)qdim⁡kHq(C,ip,∗κ(p)); by step 1.1 every term with q≥1 vanishes and the remaining term is dim⁡kH0(C,ip,∗κ(p))=dim⁡kκ(p)=d. Hence χ(C,ip,∗κ(p))=d.

3.1F5step 1.1step 2.1

The one-point shift. Applying additivity [F5] to the short exact sequence of step 1.1 gives χ(C,OC(D+p))=χ(C,OC(D))+χ(C,ip,∗κ(p))=χ(C,OC(D))+d by step 2.1.

4.1F2F3F4F6step 3.1step 1.2∎

Conclusion and choice accounting. Step 3.1 gives the one-point identity and step 1.2 the identity for every effective divisor E, both in Z because they are alternating sums of finite dimensions and the degree deg⁡k(E) is an integer. The Axiom of Choice is used only through the suppliers recorded in [F6], namely the Euler-characteristic definition of [F4] and the finiteness, coherence and skyscraper suppliers of [F2] and [F3], which themselves inherit it; no further selection is made above.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

Every divisor is a finite signed sum of points

Statement

Assume the Axiom of Choice, inherited from the Euler-characteristic suppliers below. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field) and let D=∑xnx[x] be a divisor on C (Divisors on a smooth proper curve). Write D+ and D− for the positive and negative parts of D (Divisor support positive negative parts).

  1. (Decomposition.) D is a finite Z-linear combination of closed points, and D=D+−D− with D+ and D− effective divisors of disjoint support; consequently deg⁡k(D)=deg⁡k(D+)−deg⁡k(D−) (Degree divisor proper curve, Effective divisors have nonnegative degree).
  2. (Chain.) Let p1,…,pm be a listing of the points of Supp⁡(D+) in which each point x occurs exactly nx times, and let q1,…,qn be a listing of the points of Supp⁡(D−) in which each point y occurs exactly −ny times. For every ordering of the m+n signed symbols +[p1],…,+[pm],−[q1],…,−[qn], the chain of divisors M0=0, Mk=Mk−1±[rk], where the k-th symbol is ±[rk], has successive differences ±[rk] at closed points and ends at Mm+n=∑i[pi]−∑j[qj]=D.
  3. (Order-independent iterated computation.) Computing χ along this chain by the one-point shift of Euler characteristic changes by the residue degree, every step changes the value by +[κ(rk):k] for an addition and by −[κ(rk):k] for a removal, so the telescoping total is χ(C,OC(D))=χ(C,OC(0))+∑i[κ(pi):k]−∑j[κ(qj):k]=χ(C,OC(0))+deg⁡k(D+)−deg⁡k(D−)=χ(C,OC(0))+deg⁡k(D), a value that depends only on the multiset of listed points and not on the chosen ordering; identifying OC(0)≅OC with the structure sheaf, this reads χ(C,OC(D))=χ(C,OC)+deg⁡k(D). Here χ is the Euler characteristic of coherent sheaves on the proper k-scheme C (Euler characteristic of a coherent sheaf), and every sheaf appearing is coherent (Finite-dimensionality of the Riemann-Roch space).

The attachment of the invertible sheaf OC(D) to the divisor D and the identification OC(0)≅OC use the current interfaces of Invertible sheaf of cartier divisor and Cartier and Weil divisors agree on a smooth curve. The latter has an explicit Dependent Choice premise, supplied by the stated Axiom of Choice through AC implies DC implies countable choice; these premises are recorded in [F6] and [F7]. (Scaffold repair: the scaffold's phrase "enumeration of the support" is read as a listing with repetitions, each point occurring exactly as often as its coefficient, which is what makes the chain end at D; the statement above says this explicitly.)

Facts & Assumptions

Given: the Axiom of Choice inherited from the Euler-characteristic and Cartier-divisor suppliers; a field k, a smooth proper geometrically integral curve C over k, a divisor D=∑xnx[x] on C, and listings p1,…,pm, q1,…,qn as in part 2.

[F1]

Divisors, parts and degree. The divisors on C are the finite formal integral combinations of closed points and form the free abelian group Div⁡(C); the support Supp⁡(D)={x:nx≠0} is finite, D+=∑xmax⁡(nx,0)[x], D−=∑xmax⁡(−nx,0)[x] are effective with disjoint supports, and D=D+−D−; the k-degree is deg⁡k(D)=∑xnx[κ(x):k] and deg⁡k:Div⁡(C)→Z is a group homomorphism, while deg⁡k(E)≥0 for every effective E≥0 (Divisors on a smooth proper curve, Divisor support positive negative parts, Degree divisor proper curve, Effective divisors have nonnegative degree).

[F2]

The one-point shift. For every divisor M on C and every closed point r∈C, χ(C,OC(M+[r]))=χ(C,OC(M))+[κ(r):k], and more generally χ(C,OC(M+E))=χ(C,OC(M))+deg⁡k(E) for every effective divisor E≥0; both identities hold in Z (Euler characteristic changes by the residue degree).

[F3]

Coherence and finiteness. For every divisor M on C the invertible sheaf OC(M) is a coherent OC-module, and Hq(C,OC(M)) is a finite-dimensional k-vector space for every q≥0 that vanishes for q≥2 (Finite-dimensionality of the Riemann-Roch space).

[F4]

The Euler characteristic of a coherent module F on the proper k-scheme C is χ(C,F)=∑q≥0(−1)qdim⁡kHq(C,F), a finite alternating sum of finite dimensions and hence an element of Z (Euler characteristic of a coherent sheaf, Sheaf cohomology as right derived global sections, Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis).

[F5]

Functoriality in the sheaf. For each q≥0 the assignment F↦Hq(C,F) is a covariant additive functor on abelian sheaves on C, so a morphism φ:F→G induces Hq(C,φ):Hq(C,F)→Hq(C,G) compatibly with identities and composites (Variance of sheaf cohomology); a functor sends isomorphisms to isomorphisms (Every functor preserves isomorphisms), so isomorphic sheaves have isomorphic cohomology groups in every degree and equal Euler characteristics.

[F6]

The current Cartier-to-Weil interface identifies the Weil divisors of this smooth proper curve with Cartier divisors and preserves their principal divisors (Cartier and Weil divisors agree on a smooth curve). The associated sheaf is constructed with OC(D)⊆KC and OC(0)≅OC by Invertible sheaf of cartier divisor. These are the interfaces for every sheaf OC(M) in the chain.

[F7]

The Axiom of Choice is used through the Euler-characteristic supplier [F2], the finiteness and coherence supplier [F3], the Euler-characteristic definition [F4] and the Cartier-to-Weil interface [F6]. In ZF, AC implies DC by AC implies DC implies countable choice, so the DC premise of Cartier and Weil divisors agree on a smooth curve is available from the stated assumption; no further selection is made below (The Axiom of Choice, The axiom of dependent choice: a relation in which every element is related to something admits an N-indexed chain).

Proof

technique · direct; split the divisor into positive and negative parts, expand each coefficient into that many single-point operations, walk the resulting finite chain through the divisor group while telescoping the one-point shifts of $\chi$, and observe that the total is a signed sum of residue degrees over a multiset that does not depend on the ordering
1.1F1

Decomposition. By [F1] the support of D=∑xnx[x] is a finite set of closed points, so D is a finite Z-linear combination of closed points. By [F1] the parts D+=∑xmax⁡(nx,0)[x] and D−=∑xmax⁡(−nx,0)[x] are effective divisors with disjoint supports and D=D+−D−. Since deg⁡k is a group homomorphism [F1], deg⁡k(D)=deg⁡k(D+)−deg⁡k(D−), and both terms are nonnegative integers because D+ and D− are effective [F1].

2.1F1step 1.1

The chain. In the free abelian group Div⁡(C) of [F1] one has ∑i[pi]=∑x∈Supp⁡(D+)nx[x]=D+ and ∑j[qj]=∑y∈Supp⁡(D−)(−ny)[y]=D−, because the listings repeat each point with its coefficient; hence ∑i[pi]−∑j[qj]=D+−D−=D by step 1.1. Define M0=0 and, for k=1,…,m+n, Mk=Mk−1+[rk] if the k-th symbol is +[rk] and Mk=Mk−1−[rk] if it is −[rk]; each Mk is an element of Div⁡(C), each successive difference Mk−Mk−1 is ±[rk] at the closed point rk, and the final term is Mm+n=∑i[pi]−∑j[qj]=D for every ordering, because addition in the abelian group Div⁡(C) is commutative and associative.

3.1F2F3F4step 2.1

The shift at each step. Let M∈Div⁡(C) and let r∈C be a closed point. Applying the one-point identity of [F2] to M gives χ(C,OC(M+[r]))=χ(C,OC(M))+[κ(r):k], and applying it to M−[r] in place of M gives χ(C,OC(M))=χ(C,OC(M−[r]))+[κ(r):k], that is, χ(C,OC(M−[r]))=χ(C,OC(M))−[κ(r):k]. All these values are defined: for every divisor M′ the sheaf OC(M′) is coherent by [F3], so [F4] applies to the proper k-scheme C. Consequently each step of the chain of step 2.1 changes the Euler characteristic by +[κ(rk):k] for a symbol +[rk] and by −[κ(rk):k] for a symbol −[rk].

4.1F1step 1.1step 2.1step 3.1

Telescoping. Induction on k using step 3.1 gives χ(C,OC(Mk))=χ(C,OC(0))+∑j=1kεj[κ(rj):k], where εj=1 if the j-th symbol is an addition and εj=−1 if it is a removal. At k=m+n this is χ(C,OC(D))=χ(C,OC(0))+∑i[κ(pi):k]−∑j[κ(qj):k] by step 2.1, and the two sums are ∑x∈Supp⁡(D+)nx[κ(x):k]=deg⁡k(D+) and ∑y∈Supp⁡(D−)(−ny)[κ(y):k]=deg⁡k(D−) by the definition of the listings, so χ(C,OC(D))=χ(C,OC(0))+deg⁡k(D+)−deg⁡k(D−)=χ(C,OC(0))+deg⁡k(D) by step 1.1. The multiset of signed residue degrees {εj[κ(rj):k]} is determined by the listings alone, so this total, and hence the iterated value, is independent of the chosen ordering.

5.1F2F3F4F5F6F7step 1.1step 2.1step 4.1∎

The base term and conclusion. The chain of step 2.1 starts at the zero divisor 0, whose attached sheaf is OC(0); the current dictionary [F6] identifies OC(0)≅OC with the structure sheaf, and by [F5] isomorphic sheaves have isomorphic cohomology in every degree, hence equal Euler characteristics, so χ(C,OC(0))=χ(C,OC). With step 4.1 this gives the asserted identity χ(C,OC(D))=χ(C,OC)+deg⁡k(D). Steps 1.1, 2.1 and 4.1 prove the decomposition, chain and order-independence clauses, so all three parts of the Statement hold. The Axiom of Choice is used only through the suppliers recorded in [F7], namely the one-point shift [F2], the coherence and finiteness of [F3], the Euler-characteristic definition [F4] and the current dictionary [F6], with DC supplied by AC as recorded there; no further selection is made, the listings of part 2 being finite and fixed.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

Riemann-Roch in Euler-characteristic form: the degree shift

Statement

Assume the Axiom of Choice, inherited from the Euler-characteristic and finiteness suppliers below. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field) and let D be a divisor on C (Divisors on a smooth proper curve). Then χ(C,OC(D))−χ(C,OC)=deg⁡k(D), an identity in Z, and equivalently h0(D)−h1(D)=deg⁡k(D)+h0(0)−h1(0), where hi(D)=dim⁡kHi(C,OC(D))≥0 is the notation of The Riemann-Roch dimension l(D) and χ is the Euler characteristic of coherent sheaves on the proper k-scheme C (Euler characteristic of a coherent sheaf), a finite alternating sum of finite dimensions. No Serre duality is used: both forms leave the index of speciality h1(D) as an unknown nonnegative integer.

The attachment of OC(D), the identity OC(0)≅OC, and the global-section identification use the current Cartier-divisor interfaces Invertible sheaf of cartier divisor, Rational sections of line bundles are Cartier divisors and Cartier and Weil divisors agree on a smooth curve, inherited through Every divisor is a finite signed sum of points.

Facts & Assumptions

Given: a field k, a smooth proper geometrically integral curve C over k, and a divisor D on C.

[F1]

The curve C is proper, separated and of finite type over the field k, geometrically integral and of chain dimension one; a divisor on C is a finite formal integral combination of closed points, D=∑xnx[x], and the k-degree is deg⁡k(D)=∑xnx[κ(x):k], a group homomorphism Div⁡(C)→Z (Curves over a field, Divisors on a smooth proper curve, Degree divisor proper curve).

[F2]

The decomposition lemma: D=D+−D− with D+,D− effective of disjoint support and deg⁡k(D)=deg⁡k(D+)−deg⁡k(D−); for every listing of the points of Supp⁡(D+) repeated with their coefficients, followed by the points of Supp⁡(D−) repeated with the coefficients of D−, and for every ordering of the resulting signed symbols, the chain M0=0, Mk=Mk−1±[rk] ends at D, and telescoping the one-point shift gives χ(C,OC(D))=χ(C,OC(0))+deg⁡k(D), a value independent of the ordering; identifying OC(0)≅OC with the structure sheaf this reads χ(C,OC(D))=χ(C,OC)+deg⁡k(D) (Every divisor is a finite signed sum of points).

[F3]

Coherence, finiteness and the dimension form of χ: for every divisor D′ on C the invertible sheaf OC(D′) is a coherent OC-module, Hq(C,OC(D′)) is a finite-dimensional k-vector space for every q≥0 and vanishes for every q≥2, and the Euler characteristic satisfies χ(C,OC(D′))=h0(D′)−h1(D′) (Finite-dimensionality of the Riemann-Roch space, Euler characteristic of a coherent sheaf, Sheaf cohomology as right derived global sections, Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis).

[F4]

Notation: hi(D′)=dim⁡kHi(C,OC(D′)) for every i≥0 and divisor D′, so in particular h0(D′) and h1(D′) are the dimensions appearing in [F3], and l(D′)=h0(D′) (The Riemann-Roch dimension l(D)).

[F5]

The current Cartier-divisor interfaces Invertible sheaf of cartier divisor, Rational sections of line bundles are Cartier divisors and Cartier and Weil divisors agree on a smooth curve attach OC(D) to D, identify L(D) with H0(C,OC(D)), and give OC(0)≅OC. The proof uses the last identification for its base term.

[F6]

The Axiom of Choice is used exactly through the decomposition lemma [F2], the finiteness and coherence supplier [F3] and the flagged dictionary [F5]; no further selection is made below (The Axiom of Choice).

Proof

technique · direct; read the degree shift off the point-by-point decomposition of the divisor, convert both Euler characteristics into their $h^0-h^1$ forms, and compare the resulting identities to obtain the two displayed forms and their equivalence
1.1F1F3F4F5

Set-up. By [F1] the curve C is proper over k, so the Euler characteristic of [F3] is defined for coherent sheaves on C. By [F3] the sheaves OC(D) and OC(0) are coherent OC-modules, so χ(C,OC(D)), χ(C,OC(0)) and, through the flagged dictionary [F5], χ(C,OC)=χ(C,OC(0)) are all integers; by [F4] the symbols hi(D) are the dimensions of [F3].

1.2F2F5

The degree shift. By part 3 of the decomposition lemma [F2], applied to the divisor D, the telescoping of the one-point shifts along any ordering of the listed signed points gives χ(C,OC(D))=χ(C,OC(0))+deg⁡k(D); by the flagged identification OC(0)≅OC of [F5] this reads χ(C,OC(D))=χ(C,OC)+deg⁡k(D). Rearranging in Z gives the first displayed identity χ(C,OC(D))−χ(C,OC)=deg⁡k(D); the ordering-independence asserted in [F2] shows that the value does not depend on how the listing is traversed.

2.1F3F4step 1.2

The dimension forms of the two Euler characteristics. By [F3] applied to the divisor D, χ(C,OC(D))=h0(D)−h1(D); by [F3] applied to the zero divisor, χ(C,OC(0))=h0(0)−h1(0). Substituting the second identity into the un-flagged form of step 1.2 gives h0(D)−h1(D)=deg⁡k(D)+h0(0)−h1(0), the second displayed identity; no Serre duality is involved, the terms h1(D) and h1(0) being the nonnegative dimensions of [F3] and [F4].

3.1F3F5step 1.2step 2.1

Equivalence of the two forms. Assume first the first displayed identity. By step 2.1, h0(D)−h1(D)=χ(C,OC(D)) and h0(0)−h1(0)=χ(C,OC(0)); by the flagged dictionary [F5], χ(C,OC(0))=χ(C,OC), so the first identity rearranges to the second. Conversely, assume the second displayed identity; then by step 2.1, χ(C,OC(D))−χ(C,OC(0))=deg⁡k(D), and by the flagged dictionary [F5] the second term is χ(C,OC), giving the first identity. Thus the two displayed forms are equivalent under the flagged identification, and each of them is proved: the first in step 1.2 with the flag, the second in step 2.1 without it.

4.1F2F3F5F6step 1.2step 2.1step 3.1∎

Conclusion and choice accounting. Step 1.2 gives the degree-shift identity and step 2.1 the equivalent h0−h1 identity, both in Z because they are rearrangements of identities between finite alternating sums of finite dimensions and the integer deg⁡k(D); step 3.1 records the equivalence. The Axiom of Choice is used only through the suppliers recorded in [F6], namely the decomposition lemma [F2], the finiteness and coherence supplier [F3] and the flagged dictionary [F5]; in particular no ordering of the divisor support is chosen in a way that needs any choice principle, the listings being finite and fixed by the divisor, and the ordering-independence clause of [F2] holds for every ordering.

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

Genus via the Euler characteristic

Definition

Assume the Axiom of Choice for proper-cohomology finiteness and the global functions theorem (The Axiom of Choice). Let k be a field and let C be a smooth proper geometrically integral curve over k (Curves over a field). The current Finite-dimensionality of the Riemann-Roch space proves under this assumption that H1(C,OC) is finite-dimensional. The genus of C is the nonnegative integer g(C):=h1(C,OC)=dim⁡kH1(C,OC) (The Riemann-Roch dimension l(D), Sheaf cohomology as right derived global sections, Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis).

The global functions theorem Functions on a proper curve gives H0(C,OC)=k. Consequently the Euler characteristic χ(C,OC)=h0(C,OC)−h1(C,OC) of Euler characteristic of a coherent sheaf satisfies χ(C,OC)=1−g(C),equivalentlyg(C)=1−χ(C,OC). This is the arithmetic genus pa(C)=1−χ(OC) of Genus and arithmetic genus of a curve for a smooth geometrically connected proper curve. The equality of the two definitions here follows from H0(C,OC)=k and finite-dimensionality of H1.

For the projective line, the direct published cohomology calculation Top cohomology of projective twists gives H1(Pk1,O)=0 (take projective dimension 1 and twist d=0), so g(Pk1)=0. No Serre duality is used in this definition.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

Riemann-Roch for curves: the Euler-characteristic form

Statement

Assume the Axiom of Choice, inherited from the Euler-characteristic, genus and finiteness suppliers below. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field) with genus g=g(C)=h1(C,OC)=1−χ(C,OC) (Genus via the Euler characteristic), and let D be a divisor on C (Divisors on a smooth proper curve). Then h0(D)−h1(D)=χ(C,OC(D))=deg⁡k(D)+1−g, where hi(D)=dim⁡kHi(C,OC(D)) and χ is the Euler characteristic of coherent sheaves on the proper k-scheme C (The Riemann-Roch dimension l(D), Euler characteristic of a coherent sheaf). No Serre duality is used: the index of speciality h1(D) is left as an unknown nonnegative integer, and the theorem is a statement about the Euler characteristic and the k-degree alone.

The attachment of OC(D) and the identity OC(0)≅OC use the current Cartier-divisor interfaces Invertible sheaf of cartier divisor, Rational sections of line bundles are Cartier divisors and Cartier and Weil divisors agree on a smooth curve, inherited through Riemann-Roch in Euler-characteristic form: the degree shift and Finite-dimensionality of the Riemann-Roch space.

Facts & Assumptions

Given: a field k, a smooth proper geometrically integral curve C over k with genus g=g(C), and a divisor D on C.

[F1]

The curve C is proper, separated and of finite type over the field k, geometrically integral and of chain dimension one, so the Euler characteristic of coherent sheaves on C is defined; a divisor on C is a finite formal integral combination of closed points and deg⁡k:Div⁡(C)→Z is a group homomorphism (Curves over a field, Divisors on a smooth proper curve, Degree divisor proper curve, Euler characteristic of a coherent sheaf).

[F2]

The genus: g=g(C)=h1(C,OC)=dim⁡kH1(C,OC) and, since H0(C,OC) is canonically k with Euler characteristic χ(C,OC)=h0(C,OC)−h1(C,OC), one has χ(C,OC)=1−g, equivalently g=1−χ(C,OC) (Genus via the Euler characteristic, Sheaf cohomology as right derived global sections, Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis).

[F3]

The degree shift: for every divisor D′ on C, χ(C,OC(D′))−χ(C,OC)=deg⁡k(D′), and equivalently h0(D′)−h1(D′)=deg⁡k(D′)+h0(0)−h1(0); no Serre duality is used (Riemann-Roch in Euler-characteristic form: the degree shift).

[F4]

Finiteness and the dimension form: for every divisor D′ the sheaf OC(D′) is coherent, Hq(C,OC(D′)) is finite-dimensional over k for every q≥0 and vanishes for q≥2, and χ(C,OC(D′))=h0(D′)−h1(D′) (Finite-dimensionality of the Riemann-Roch space, The Riemann-Roch dimension l(D)).

[F5]

The current Cartier-divisor interfaces Invertible sheaf of cartier divisor, Rational sections of line bundles are Cartier divisors and Cartier and Weil divisors agree on a smooth curve supply the attachment of OC(D), the global-section identification, and OC(0)≅OC used in [F3].

[F6]

The Axiom of Choice is used exactly through the degree shift [F3], the genus definition [F2] and the finiteness supplier [F4], which inherit it from the proper-cohomology suppliers; no further selection is made below (The Axiom of Choice).

Proof

technique · direct; substitute the identity $\chi(C,\mathcal O_C)=1-g$ of the genus definition into the degree shift and read off the two equalities
1.1F1F2F4

Set-up. By [F1] the curve C is proper over k, so the Euler characteristics of [F3] and [F4] are defined. By [F2] the genus satisfies χ(C,OC)=1−g; by [F4] the sheaf OC(D) is coherent and χ(C,OC(D))=h0(D)−h1(D), the dimensions being finite and the higher cohomology vanishing.

1.2F2F3

The degree shift. By [F3], applied to the divisor D, χ(C,OC(D))−χ(C,OC)=deg⁡k(D). Substituting χ(C,OC)=1−g from [F2] gives χ(C,OC(D))=deg⁡k(D)+1−g, the second asserted equality.

2.1F4step 1.2

The dimension form. By [F4] the Euler characteristic of OC(D) is h0(D)−h1(D), so combining with step 1.2 gives h0(D)−h1(D)=χ(C,OC(D))=deg⁡k(D)+1−g, which is the full displayed chain of the Statement; the index of speciality h1(D) is a nonnegative integer by [F4] and is not identified with any other expression, so no Serre duality is used.

3.1F2F3F4F5F6step 1.2step 2.1∎

Conclusion and choice accounting. Steps 1.2 and 2.1 give both asserted equalities for every divisor D on C, with g=g(C) as in [F2]. The Axiom of Choice is used only through the suppliers recorded in [F6], namely the degree shift [F3], the genus definition [F2] and the finiteness supplier [F4]; the flagged dictionary [F5] is the inherited obligation on the sheaf OC(D), and no further selection is made above.

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

The Riemann inequality

Statement

Assume the Axiom of Choice, inherited from the Riemann-Roch and finiteness suppliers below. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field) with genus g=h1(C,OC)=1−χ(C,OC) (Genus via the Euler characteristic), and let D be a divisor on C (Divisors on a smooth proper curve). Then l(D)=h0(D)≥deg⁡k(D)+1−g, where l(D)=dim⁡kL(D)=h0(D) and hi(D)=dim⁡kHi(C,OC(D)) are the integers of The Riemann-Roch dimension l(D). The inequality is the Riemann inequality; it is generally strict, the excess being the index of speciality, and it is not used here to produce sections.

The attachment of OC(D) and the identification of L(D) with H0(C,OC(D)) use the current interfaces Invertible sheaf of cartier divisor, Rational sections of line bundles are Cartier divisors and Cartier and Weil divisors agree on a smooth curve, inherited through Riemann-Roch for curves: the Euler-characteristic form.

Facts & Assumptions

Given: a field k, a smooth proper geometrically integral curve C over k with genus g=h1(C,OC), and a divisor D on C.

[F1]

The curve C is proper, of finite type and of chain dimension one over the field k; a divisor on C is a finite formal integral combination of closed points and deg⁡k:Div⁡(C)→Z is the k-degree homomorphism (Curves over a field, Divisors on a smooth proper curve).

[F2]

The integers l(D) and hi(D): l(D)=dim⁡kL(D)=dim⁡kH0(C,OC(D))=h0(D) is a nonnegative integer, and hi(D)=dim⁡kHi(C,OC(D)) for every i≥0; in particular h1(D) is the dimension of a k-vector space (The Riemann-Roch dimension l(D), Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis).

[F3]

Riemann-Roch in Euler-characteristic form: for the genus g=g(C)=h1(C,OC) and every divisor D on C, h0(D)−h1(D)=χ(C,OC(D))=deg⁡k(D)+1−g; no Serre duality is used (Riemann-Roch for curves: the Euler-characteristic form).

[F4]

The current interfaces Invertible sheaf of cartier divisor, Rational sections of line bundles are Cartier divisors and Cartier and Weil divisors agree on a smooth curve attach OC(D) and identify L(D) with its global sections; this use is inherited from [F3].

[F5]

The Axiom of Choice is used exactly through the Riemann-Roch supplier [F3] and the finiteness supplier [F2], which inherit it from the proper-cohomology suppliers; no further selection is made below (The Axiom of Choice).

Proof

technique · direct; rearrange the Riemann-Roch identity into $h^0(D)=$ right-hand side $+$ $h^1(D)$ and use that a dimension is nonnegative
1.1F1F2F3

Set-up. By [F1] the curve C is proper over k and D is a divisor on C with k-degree deg⁡k(D). By [F2] the integer l(D) equals h0(D)=dim⁡kH0(C,OC(D)), and h1(D)=dim⁡kH1(C,OC(D)) is the dimension of a k-vector space, hence nonnegative by [F2]. By [F3] the identity h0(D)−h1(D)=deg⁡k(D)+1−g holds for the divisor D and the genus g=h1(C,OC) of the curve.

2.1F2step 1.1

The inequality. Adding h1(D) to both sides of the identity of step 1.1 gives h0(D)=deg⁡k(D)+1−g+h1(D); since h1(D)≥0, the right-hand side is at least deg⁡k(D)+1−g, so h0(D)≥deg⁡k(D)+1−g. As l(D)=h0(D) by [F2], this is exactly the asserted inequality l(D)≥deg⁡k(D)+1−g.

3.1F2F3F4F5step 2.1∎

Conclusion and choice accounting. Step 2.1 proves l(D)=h0(D)≥deg⁡k(D)+1−g for every divisor D on C, the identity l(D)=h0(D) being the definitional identification of [F2]. The Axiom of Choice is used only through the suppliers recorded in [F5], namely the Riemann-Roch theorem [F3] and the finiteness supplier [F2]; the deduction itself makes no selection, and the flagged dictionary [F4] records the inherited obligation on OC(D).

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

No sections in negative degree

Statement

Assume the Axiom of Choice as inherited from the divisor and degree suppliers. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field) and let D be a divisor on C (Divisors on a smooth proper curve). If deg⁡k(D)<0 then L(D)=0 and l(D)=0 (The space L(D), The Riemann-Roch dimension l(D)). The proof given is the effective-divisor argument: a nonzero f∈L(D) would exhibit the effective divisor div⁡(f)+D linearly equivalent to D, whose degree is therefore deg⁡k(D), contradicting the nonnegativity of the degree of an effective divisor. It does not appeal to the Riemann inequality, which for negative degree would only give the vacuous bound l(D)≥deg⁡k(D)+1−g with a nonpositive right-hand side; it cannot ensure a nonzero section.

Facts & Assumptions

Given: a field k, a smooth proper geometrically integral curve C over k, a divisor D on C with deg⁡k(D)<0, and an element f∈L(D).

[F1]

The Riemann-Roch space: L(D)={f∈k(C)×:div⁡(f)+D≥0}∪{0} is the k-subspace of functions whose poles are no worse than −D allows, membership being read coefficientwise as ord⁡x(f)+nx≥0 at every closed point x, and div⁡(f)=∑xord⁡x(f)[x] (The space L(D)).

[F2]

The dimension: l(D)=dim⁡kL(D) is the dimension of the Riemann-Roch space, so L(D)=0 exactly when l(D)=0 (The Riemann-Roch dimension l(D)).

[F3]

Divisors and degree: a divisor on C is a finite formal Z-linear combination of closed points, the k-degree is deg⁡k(D)=∑xnx[κ(x):k], and deg⁡k is a group homomorphism on the divisor group (Divisors on a smooth proper curve, Degree divisor proper curve).

[F4]

Effective divisors: if D=∑xnx[x] is effective then deg⁡k(D)≥0, and deg⁡k(D)=0 only for D=0; conversely an effective divisor of negative degree cannot exist (Effective divisors have nonnegative degree).

[F5]

Sections versus effective divisors: for every nonzero f∈L(D) the divisor div⁡(f)+D is an effective divisor on C linearly equivalent to D; in particular L(D)=0 if and only if no effective divisor is linearly equivalent to D (Effective divisors linearly equivalent to D are sections modulo scalars).

[F6]

The current Principal divisors on a normal proper curve have degree zero states that a principal divisor div⁡(f) of a nonzero rational function on a normal proper curve has degree 0, equivalently that linearly equivalent divisors have equal degree. The smooth curve C is normal by the local-ring and normality clause of Divisors on a smooth proper curve, so this theorem applies at step 2.1.

[F7]

The Axiom of Choice is available and is inherited only through the suppliers named above, in particular the principal-divisor degree theorem of [F6]; the proof below makes no selection (The Axiom of Choice).

Proof

technique · direct contradiction; a nonzero section would produce an effective divisor of the same negative degree, and effective divisors have nonnegative degree
1.1F1F5

A nonzero section and its effective divisor. Suppose f∈L(D) with f≠0. By [F1] membership means div⁡(f)+D≥0 coefficientwise, so E:=div⁡(f)+D is an effective divisor on C; by [F5] the divisor E is linearly equivalent to D, the difference being the principal divisor div⁡(f).

2.1F3F6step 1.1

Its degree. By [F3] the degree is additive on the divisor group, so deg⁡k(E)=deg⁡k(div⁡(f))+deg⁡k(D), and by the principal-divisor degree theorem [F6] the principal divisor div⁡(f) has degree 0; hence deg⁡k(E)=deg⁡k(D), which is negative by hypothesis.

3.1F2F4step 1.1step 2.1

Contradiction and vanishing. By [F4] the effective divisor E of step 1.1 has deg⁡k(E)≥0, contradicting deg⁡k(E)=deg⁡k(D)<0 from step 2.1. Hence L(D) contains no nonzero element, that is L(D)=0, and then l(D)=dim⁡kL(D)=0 by [F2].

4.1F3F6F7step 2.1step 3.1∎

Conclusion and choice accounting. Steps 1.1 through 3.1 show that deg⁡k(D)<0 forces L(D)=0 and l(D)=0 by the effective-divisor argument; the Riemann inequality is not used, and indeed for negative degree it would only bound l(D) from below by the nonpositive number deg⁡k(D)+1−g. The principal-divisor degree theorem [F6], used at step 2.1, expresses that degree is well defined on linear-equivalence classes; the Axiom of Choice is inherited from that theorem and the divisor suppliers, and no further selection is made, since the contradiction argument chooses nothing beyond the given f.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

Adding points never raises h^1, and h^1 stabilizes

Statement

Assume the Axiom of Choice, inherited from the sheaf-cohomology suppliers below. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field), let D be a divisor on C (Divisors on a smooth proper curve) and let p∈C be a closed point with residue degree d=[κ(p):k] (Degree divisor proper curve). Write hi(D)=dim⁡kHi(C,OC(D)) (The Riemann-Roch dimension l(D)).

  1. The long exact sequence of the short exact sequence 0→OC(D)→OC(D+p)→ip,∗κ(p)→0 contains, after identifying H0(C,ip,∗κ(p)) with κ(p), the exact sequence of finite-dimensional k-vector spaces and k-linear maps 0→H0(C,OC(D))→H0(C,OC(D+p))→κ(p)→ ∂ H1(C,OC(D))→H1(C,OC(D+p))→0, whose middle arrow ∂ is the connecting map. Consequently H1(C,OC(D+p))≅H1(C,OC(D))/im⁡∂, so that h1(D+p)≤h1(D), with equality if and only if ∂=0, and h1(D)−h1(D+p)=dim⁡kim⁡∂≤d.
  2. Consequently for every effective divisor E≥0 one has h1(D+E)≤h1(D): the integer h1 is antitone in the divisor (Divisor support positive negative parts).
  3. In particular, for a fixed divisor D0 and a fixed effective divisor A≥0, the sequence n↦h1(D0+nA) is non-increasing and therefore stabilizes: it is constant for all sufficiently large n. Equivalently, removing points from a divisor can only raise or preserve h1: if D′≤D′′ then h1(D′′)≤h1(D′).

The short exact sequence and the cohomology of the skyscraper ip,∗κ(p) are supplied by The exact sequence for adding one point to a divisor. Its construction uses the current Cartier-divisor and order interfaces Invertible sheaf of cartier divisor, Rational sections of line bundles are Cartier divisors and Cartier and Weil divisors agree on a smooth curve.

Facts & Assumptions

Given: a field k, a smooth proper geometrically integral curve C over k, divisors D, D0, A on C with A≥0, and a closed point p∈C.

[F1]

Divisors and degrees. A divisor on C is a finite formal sum ∑xnx[x] over the closed points, D≤E means that E−D is effective, deg⁡k is additive and deg⁡k(E)≥0 for effective E, and the residue degree satisfies [κ(p):k]=dim⁡kκ(p)≥1 for every closed point p (Divisors on a smooth proper curve, Divisor support positive negative parts, Degree divisor proper curve, Effective divisors have nonnegative degree).

[F2]

Adding one point. The sequence 0→OC(D)→OC(D+p)→ip,∗κ(p)→0 is a short exact sequence of coherent OC-modules, H0(C,ip,∗κ(p))≅κ(p) with k-dimension d=[κ(p):k], and Hq(C,ip,∗κ(p))=0 for every q≥1 (The exact sequence for adding one point to a divisor).

[F3]

Long exact sequence. A short exact sequence of abelian sheaves on C gives a natural long exact sequence of sheaf cohomology groups Hq, with the connecting map H0→H1 (Long exact sequence of sheaf cohomology, Sheaf cohomology as right derived global sections).

[F4]

Finiteness and the notation hi. For every divisor D′ the groups Hq(C,OC(D′)) are finite-dimensional k-vector spaces for all q≥0, vanishing for q≥2, and hi(D′)=dim⁡kHi(C,OC(D′)) is a nonnegative integer (The Riemann-Roch dimension l(D), Finite-dimensionality of the Riemann-Roch space, Finite-dimensional coherent cohomology over a field, Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis).

[F5]

k-linearity of the long exact sequence. For an OC-module F the cohomology groups are the right derived objects of the global-sections functor of Sheaf cohomology as right derived global sections, and multiplication by a scalar c∈k on F is the OC-module endomorphism given by multiplication by the global function c, so functoriality of the derived objects turns it into the scalar action on Hq(C,F); a morphism of OC-modules commutes with these endomorphisms and hence induces a k-linear map on cohomology. Consequently the maps and the connecting map in the long exact sequence of [F3], applied to a short exact sequence of OC-modules, are k-linear, and the terms are the finite-dimensional k-vector spaces of the proper finiteness theorem (Finite-dimensional coherent cohomology over a field, Long exact sequence of sheaf cohomology).

[F6]

Linear algebra over k. For a linear map T:V→W with V finite-dimensional one has dim⁡kV=dim⁡kker⁡T+dim⁡kim⁡T (Rank-nullity: dim⁡FV=nullity⁡T+rank⁡T), the formula T~(v+ker⁡T)=T(v) defines an isomorphism V/ker⁡T→im⁡T (First isomorphism theorem for vector spaces: V/ker⁡T is isomorphic to im⁡T), and for W≤V with V finite-dimensional, dim⁡k(V/W)=dim⁡kV−dim⁡kW (A quotient basis lifts to a basis adapted to W).

[F7]

The Axiom of Choice is used exactly through the sheaf-cohomology suppliers [F2], [F3] and the finiteness supplier [F4]; no further selection is made below (The Axiom of Choice).

Proof

technique · direct; read the segment of the long exact sequence around the connecting map at one point, compute the dimension drop through the first isomorphism theorem, iterate over the points of an effective divisor, and observe that a non-increasing sequence of nonnegative integers is eventually constant
1.1F2F3F4F5

The exact segment. By [F2] the sequence 0→OC(D)→OC(D+p)→ip,∗κ(p)→0 is a short exact sequence of coherent OC-modules with H0(C,ip,∗κ(p))≅κ(p) and H1(C,ip,∗κ(p))=0. The long exact sequence of [F3] therefore contains the exact segment H0(C,OC(D))→H0(C,OC(D+p))→H0(C,ip,∗κ(p))→∂H1(C,OC(D))→H1(C,OC(D+p))→H1(C,ip,∗κ(p)); substituting the identification H0(C,ip,∗κ(p))≅κ(p) of [F2] and its vanishing in degree one, and using the injectivity of the first map and the exactness at H0(C,OC(D+p)), gives the exact sequence 0→H0(C,OC(D))→H0(C,OC(D+p))→κ(p)→∂H1(C,OC(D))→H1(C,OC(D+p))→0, with ∂ the connecting map; by [F4] and [F5] all six terms are finite-dimensional k-vector spaces and all maps are k-linear.

2.1F4F5F6step 1.1

The dimension drop. Exactness of the sequence of step 1.1 at H1(C,OC(D)) identifies im⁡∂ with the kernel of the map H1(C,OC(D))→H1(C,OC(D+p)), and exactness at H1(C,OC(D+p)) says that this map is surjective; hence the first isomorphism theorem of [F6] identifies the k-vector spaces H1(C,OC(D+p)) and H1(C,OC(D))/im⁡∂. Taking dimensions with the quotient formula of [F6] gives h1(D+p)=dim⁡kH1(C,OC(D))−dim⁡kim⁡∂=h1(D)−dim⁡kim⁡∂, so h1(D+p)≤h1(D), with equality exactly when im⁡∂=0, that is, exactly when the connecting map ∂ is zero. Since ∂ is k-linear out of the finite-dimensional space κ(p) of dimension d, rank-nullity gives dim⁡kim⁡∂=dim⁡kκ(p)−dim⁡kker⁡∂≤dim⁡kκ(p)=d by [F4], so the drop is at most d.

3.1F1step 2.1

Antitonicity in the divisor. Let E≥0 be effective and write its finite support with multiplicities as E=∑ici[xi]; consider the finite chain of divisors D≤D+[x1]≤D+2[x1]≤⋯≤D+E that adds one copy of a closed point at a time. Every consecutive pair is of the form D′≤D′+q for a closed point q, so step 2.1 applied to the divisor D′ and the point q gives h1(D′+q)≤h1(D′); chaining these inequalities along the finite chain gives h1(D+E)≤h1(D).

4.1F1F4step 3.1

Stabilization and the downward reading. Fix a divisor D0 and an effective divisor A≥0. For every n≥0 one has D0+nA≤D0+(n+1)A, so step 3.1 applied to the divisor D0+nA gives h1(D0+(n+1)A)≤h1(D0+nA): the sequence n↦h1(D0+nA) is non-increasing. Its values are nonnegative integers bounded above by h1(D0) by step 3.1; a strict decrease lowers the value by at least one, so the sequence has at most h1(D0) strict decreases and is therefore constant for all sufficiently large n, which is the asserted stabilization. Finally, if D′≤D′′ are divisors, then D′′=D′+E for the effective divisor E=D′′−D′ and step 3.1 gives h1(D′′)≤h1(D′); reading D′ as obtained from D′′ by removing the points of E, removing points from a divisor can only raise or preserve h1.

5.1F2F3F4F7step 1.1step 2.1step 3.1step 4.1∎

Conclusion and choice accounting. Step 1.1 gives the exact sequence and the identification of the connecting map, step 2.1 the inequality, the equality criterion and the bound d on the drop, step 3.1 the antitonicity for every effective divisor, and step 4.1 the stabilization and the equivalent downward reading; this proves parts (1), (2) and (3) of the Statement. The Axiom of Choice is used only through the sheaf-cohomology suppliers of [F2] and [F3] and the finiteness supplier of [F4], as recorded in [F7]; the point q added at each stage of step 3.1 is one of the finitely many points of E, so no further selection is made.

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Rational functions with poles bounded at one point

Statement

Assume the Axiom of Choice, inherited from the Riemann-Roch, proper-functions and curve suppliers below. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field) with genus g=g(C) (Genus via the Euler characteristic), and let p∈C be a closed point of residue degree d=[κ(p):k]≥1 (Degree divisor proper curve). Closed points of C exist: the curve is nonempty of chain dimension one, so besides its unique generic point it contains a point, and every such point is closed (Proper closed subsets of a curve are finite). For every integer n≥1 with nd+1−g≥2 there is a function f∈L(np) that is not constant (The space L(D)). Every such f is nonconstant, every pole of f lies at p and has order at most n, and the pole divisor (f)∞ of Order codimension one rational function is a nonzero effective divisor supported at p (Divisor support positive negative parts); in particular f has at least one pole at p.

The identification L(D)=H0(C,OC(D)) and the sheaf OC(D) use the current interfaces Invertible sheaf of cartier divisor, Rational sections of line bundles are Cartier divisors and Cartier and Weil divisors agree on a smooth curve, inherited through The Riemann inequality.

Facts & Assumptions

Given: a field k, a smooth proper geometrically integral curve C over k with genus g=g(C), a closed point p∈C of residue degree d=[κ(p):k], and an integer n≥1 with nd+1−g≥2.

[F1]

Curve and closed points: C is nonempty, geometrically integral, separated, of finite type and of chain dimension one over k; its underlying space is Noetherian, it has a unique generic point ηC, and every point x≠ηC is a closed point of C (Curves over a field, Proper closed subsets of a curve are finite). Since chain dimension one means that there is a strict chain of two nonempty irreducible closed subsets, C has at least two points, hence a point different from ηC, and therefore at least one closed point.

[F2]

Divisor and degree: a divisor on C is a finite formal integral combination of closed points, and for a closed point x the residue field κ(x) is finite over k with [κ(x):k]≥1 and deg⁡k([x])=[κ(x):k] (Divisors on a smooth proper curve, Degree divisor proper curve).

[F3]

The Riemann-Roch space: for a divisor D on C with function field k(C), L(D)={f∈k(C)×:div⁡(f)+D≥0}∪{0} is a k-subspace of k(C), where div⁡(f)=∑xord⁡x(f)[x] uses the order of vanishing at each closed point; L(D) is a k-vector space with dim⁡kL(D)=l(D)=h0(D) and l is a nonnegative integer (The space L(D), Order codimension one rational function, The Riemann-Roch dimension l(D)).

[F4]

The Riemann inequality: l(D)≥deg⁡k(D)+1−g for every divisor D; for D=np this gives l(np)≥nd+1−g≥2 by the hypothesis on n, using deg⁡k(np)=nd from [F2] (The Riemann inequality, Degree divisor proper curve).

[F5]

Constants: H0(C,OC) is canonically k, and for every effective divisor D≥0 each nonzero constant c∈k× has div⁡(c)=0, while 0∈L(D) by definition; hence k⋅1⊆L(D); every f∈L(0) is constant, since a nonzero such f has div⁡(f)≥0 and hence is regular everywhere, and zero is constant (Functions on a proper curve, The space L(D), The Riemann-Roch dimension l(D)).

[F6]

The dimension of a k-vector space: if dim⁡kV=m>1 and W⊆V is a subspace of dimension one, then W≠V and there is v∈V∖W; dimensions of vector spaces are compared by inclusion and equality (Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis).

[F7]

The current interfaces Invertible sheaf of cartier divisor, Rational sections of line bundles are Cartier divisors and Cartier and Weil divisors agree on a smooth curve supply OC(D) and the identification L(D)=H0(C,OC(D)); the use of l and the inequality below is inherited from [F4].

[F8]

The Axiom of Choice is inherited from the curve, divisor, dimension, Riemann-inequality and proper-functions suppliers recorded in [F1]-[F7]; choosing a single function outside the constants needs no additional choice principle (The Axiom of Choice).

Proof

technique · direct; find a closed point, use the Riemann inequality to make $L(np)$ at least two-dimensional, take a function outside the one-dimensional space of constants, and read the pole conditions off the membership in $L(np)$
1.1F1F2

Existence of closed points and the residue degree. By [F1] the curve C is nonempty with a unique generic point ηC, and its chain dimension one provides a strict chain of two nonempty irreducible closed subsets, so C has at least two points and we may fix a point p0≠ηC; by [F1] every point other than ηC is closed, so p0 is a closed point of C. In particular closed points exist, and for every closed point x, such as the given p, the residue field is finite over k with [κ(x):k]≥1 by [F2]; the given residue degree is d=[κ(p):k]≥1.

1.2F2F3F4

The dimension bound. By [F4] applied to the divisor np, whose degree is deg⁡k(np)=nd by [F2], one has l(np)≥nd+1−g≥2; by [F3] the integer l(np) is the k-dimension of L(np), so L(np) is a k-vector space of dimension at least two.

2.1F3F5F6step 1.2

A nonconstant function with poles only at p. By [F5] the constants form the subspace k⋅1⊆L(np) of dimension one; since dim⁡kL(np)≥2>1, [F6] provides f∈L(np) with f∉k⋅1, so f is not a constant function. By [F3] membership f∈L(np) means div⁡(f)+np≥0, so ord⁡x(f)≥0 for every closed point x≠p and ord⁡p(f)≥−n; that is, every pole of f lies at p with order at most n.

3.1F3F5step 2.1

f has a pole, and the pole divisor. Since f is nonconstant by step 2.1, f∉L(0): otherwise div⁡(f)≥0 and [F5] would exhibit f as a constant. Hence some order ord⁡x(f) is negative; by step 2.1 the only point where this can happen is p, so f has a pole at p. Therefore the pole divisor (f)∞=∑ord⁡x(f)<0(−ord⁡x(f))[x] is a nonzero effective divisor supported at p, with coefficient −ord⁡p(f) between 1 and n.

4.1F1F4F5F7F8step 2.1step 3.1∎

Conclusion and choice accounting. For the given closed point p and every n≥1 with nd+1−g≥2, steps 2.1 and 3.1 produce f∈L(np) that is nonconstant, has all its poles at p of order at most n, and has at least one pole there; moreover every f∈L(np)∖k⋅1 has the same properties by steps 2.1 and 3.1 applied to it. The Axiom of Choice is inherited from the suppliers recorded in [F8], including the divisor and dimension interfaces [F2], [F3], [F6], [F7]; choosing the single function f outside the one-dimensional subspace k⋅1 is a single selection from a nonempty set and needs no choice principle, and the flagged dictionary [F7] records the inherited obligation on OC(D).

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Finite morphisms from a curve to the projective line

Statement

Assume the Axiom of Choice as inherited from the rational-map extension and finite-map suppliers. Let k be a field and let C be a smooth proper geometrically integral curve over k (Curves over a field), and let f∈k(C)× be a nonconstant rational function, for instance one produced by Rational functions with poles bounded at one point. Then f defines a finite locally free k-morphism φf:C⟶Pk1 of degree [k(C):k(f)]≥1, whose fibre over infinity is the pole divisor (f)∞=∑ord⁡x(f)<0(−ord⁡x(f))[x] ≥ 0 of degree [k(C):k(f)] (A nonconstant rational function defines a finite map to the projective line, Divisor support positive negative parts). In particular every smooth proper geometrically integral curve over k admits a finite k-morphism to Pk1, and if A:=(f)∞ then A is an effective divisor with OC(A)≅φf∗OPk1(1).

Facts & Assumptions

Given: a field k, a smooth proper geometrically integral curve C over k, and a nonconstant rational function f∈k(C)×.

[F1]

The map attached to a rational function: f defines a finite locally free morphism φf:C→Pk1 of degree [k(C):k(f)], whose fibre over infinity is the pole divisor (f)∞=∑ord⁡x(f)<0(−ord⁡x(f))[x] of degree [k(C):k(f)] and whose fibre over zero is the zero divisor (f)0 of the same degree; a nonzero rational function with no poles is algebraic over k and a global unit (A nonconstant rational function defines a finite map to the projective line).

[F2]

The function-field construction: if f is transcendental over k then there is a finite locally free morphism φf:C→Pk1 of degree [k(C):k(f)] with φf#(t)=f for the standard coordinate t=x1(0) of Pk1; if f is algebraic over k then f and f−1 are global units, that is f∈Γ(C,OC)× (Proper normal curve rational function map, Relative projective space from standard charts).

[F3]

Units are constants: the canonical map k→H0(C,OC) is an isomorphism, so Γ(C,OC)×=k× and a global unit is a constant function (Functions on a proper curve).

[F4]

Degree of a morphism: for a nonconstant morphism φ:C→D of smooth proper geometrically integral curves the degree is deg⁡(φ)=[k(C):k(D)], a positive integer; for φf with φf#(t)=f the target function field is k(Pk1)=k(t) identified with k(f), so this agrees with the degree [k(C):k(f)] of [F1] and [F2] (Degree of a nonconstant morphism of curves).

[F5]

Nonconstant functions exist under the numerical hypothesis: for a closed point p and an integer n≥1 with n[κ(p):k]+1−g≥2 there is a nonconstant f∈L(np), with (f)∞ a nonzero effective divisor supported at p (Rational functions with poles bounded at one point).

[F6]

Divisors of functions and the projective line: on Pk1 with coordinate t one has div⁡(t)=[V(t)]−[∞], OPk1(1)≅OPk1([∞]) with deg⁡kOPk1(1)=1, and the divisors div⁡(f) of rational functions form a subgroup of the divisor group; the zero and pole parts of div⁡(f) are effective divisors with (f)0−(f)∞=div⁡(f) (Divisors on the projective line are classified by degree, Divisors on a smooth proper curve, Order codimension one rational function).

[F7]

A flat morphism has defined pullbacks of Cartier divisors, computed by pulling back their local equations. Whenever the pullback is defined, there is a canonical isomorphism OC(φf∗D)≅φf∗OPk1(D). (Pullback of a Cartier divisor, Pullback of a Cartier divisor computes the pullback of its line bundle)

[F8]

The Axiom of Choice is available and is inherited through the suppliers named above; the proof below uses the maps and divisors attached to the given function f and, for the existence clause, one function produced by [F5] (The Axiom of Choice).

[F9]

Under AC, every proper closed subset of an integral finite-type curve is a finite set of closed points. A dimension-one curve has a strict chain of nonempty irreducible closed subsets Z0⊊Z1; hence Z0 is a nonempty proper closed subset of C and contains a closed point. (Proper closed subsets of a curve are finite)

Proof

technique · direct; rule out the algebraic case for a nonconstant function so that the function-field construction applies, read the degree and the fibre over infinity off the rational-map lemma, and identify the pullback of $\mathcal O(1)$ with the sheaf of the pole divisor through the local-equation Cartier pullback dictionary
1.1F2F3

Nonconstant functions are transcendental, hence define the map. Suppose first that f is algebraic over k. By [F2] both f and f−1 are global units, so f∈Γ(C,OC)×, and by [F3] this group is k×, so f is a constant function, contrary to the hypothesis. Hence f is transcendental over k, and the transcendental clause of [F2] provides a finite locally free morphism φf:C→Pk1 of degree [k(C):k(f)] with φf#(t)=f for the standard coordinate t of Pk1.

2.1F1F2F4step 1.1

Degree and the fibre over infinity. By [F1] the morphism φf is finite locally free of degree [k(C):k(f)], its fibre over infinity is exactly the pole divisor (f)∞=∑ord⁡x(f)<0(−ord⁡x(f))[x], and this divisor has degree [k(C):k(f)]; by [F4] the integer [k(C):k(f)] is the degree of the nonconstant morphism φf in the sense of the curve-degree definition, since φf#(t)=f identifies the function field of the target with k(f). In particular [k(C):k(f)]≥1, a degree of a finite field extension being positive, and (f)∞≥0 is an effective divisor.

3.1F5F9step 1.1step 2.1

The existence clause. Let C be any smooth proper geometrically integral curve over k, take a closed point p, which exists by [F9], and an integer n≥1 with n[κ(p):k]+1−g≥2; by [F5] there is a nonconstant f∈L(np) with (f)∞ a nonzero effective divisor supported at p, and by steps 1.1 and 2.1 the function f defines a finite locally free k-morphism φf:C→Pk1 of positive degree. Hence every such curve admits a finite k-morphism to the projective line, for instance one obtained from the bounded-pole corollary.

3.2F1F6F7step 2.1

The sheaf of the pole divisor is the pullback of O(1). The map φf is flat by [F1], so [F7] defines the Cartier pullback. On the chart about infinity use the equation s=1/t for [∞] and on its complement use the equation 1. Their pullbacks are 1/f on φf−1(U∞) and 1 on the complement of the infinity fibre. At a point of that fibre, f has negative order, so the pulled-back equation has order −ord⁡x(f)>0; outside the fibre the local equation is 1 and its order is zero. Thus these are exactly the local Cartier equations of the pole divisor from [F1], and φf∗[∞]=(f)∞=A. Now [F6] and [F7] give φf∗OPk1(1)≅φf∗OPk1([∞])≅OC(φf∗[∞])=OC(A). The degree of A is the weighted fibre degree in step 2.1; no claim that pullback preserves degree is needed.

4.1F1F5F7F8F9step 1.1step 2.1step 3.1step 3.2∎

Conclusion and choice accounting. Steps 1.1 and 2.1 show that every nonconstant f∈k(C)× defines a finite locally free morphism φf:C→Pk1 of degree [k(C):k(f)]≥1 whose fibre over infinity is the pole divisor (f)∞ of that degree; step 3.1 shows that each smooth proper geometrically integral curve over k carries such a function, hence admits a finite k-morphism to Pk1; and step 3.2 identifies the sheaf of the pole divisor with φf∗OPk1(1). The fibre-degree clause is the actual pole-map interface [F1]; the sheaf identity follows from the explicit local Cartier pullback and its canonical line-bundle isomorphism [F7]. AC is inherited through the stated suppliers as in [F8]; the existence clause uses one closed point supplied by [F9] and one function from [F5].

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

Vanishing of H^1 in a fixed ample direction

Statement

Assume the Axiom of Choice as inherited from Serre vanishing and the ample-powers theorem. It also supplies the Dependent Choice premise of the curve Cartier-to-Weil result through AC implies DC implies countable choice. Let k be a field and let C be a smooth proper geometrically integral curve over k (Curves over a field). Let φ:C→Pk1 be a finite k-morphism (such a morphism exists by Finite morphisms from a curve to the projective line) and let A be an effective divisor on C with OC(A)≅φ∗OPk1(1) (Divisors on a smooth proper curve). Write L=OC(A) (The Riemann-Roch dimension l(D)).

Then for every divisor D0 on C there is an integer n0 such that H1(C,OC(D0+nA+E))=0for every n≥n0 and every effective divisor E, equivalently h1(D)=0 for every divisor D with D≥D0+n0A. The bound n0 may depend on D0 and on the fixed morphism φ, but not on E or on the degree of E. No Serre duality and no Riemann-Roch threshold 2g−2 are used.

The divisor-to-sheaf interface used below first identifies the Weil divisors D0 and A as Cartier divisors, constructs their sheaves, and applies the Cartier addition/tensor isomorphism; the actual interfaces and their uses are recorded in [F6]. The finite morphism and pole-divisor realization in [F1] are the stated interface of Finite morphisms from a curve to the projective line. The proof uses the closed H-very ample witness from [F4] and the established affine-base implication in [F9] for Serre vanishing, and uses Finite-dimensionality of the Riemann-Roch space to establish coherence of OC(D0) as recorded in [F10].

Facts & Assumptions

Given: a field k, a smooth proper geometrically integral curve C over k, a finite k-morphism φ:C→Pk1, an effective divisor A on C with OC(A)≅φ∗O(1), the invertible sheaf L=OC(A), and a divisor D0.

[F1]

The morphism and its twist: every smooth proper geometrically integral curve over k admits a finite locally free k-morphism to Pk1, and for a nonconstant f∈k(C)× with pole divisor A=(f)∞ the sheaf OC(A) is isomorphic to φf∗OPk1(1) (Finite morphisms from a curve to the projective line, Finite morphisms of schemes).

[F2]

Ampleness of the twisting sheaf on Pk1: the identity of Pk1 over Spec⁡k is a closed immersion pulling O(1) back to O(1), so O(1) is H-very ample relative to Spec⁡k; since the base is affine, Relative very ampleness implies relative ampleness makes O(1) ample in the absolute sense of Absolute ampleness by affine section opens (Relative very ampleness in the finite projective-space convention, Twists of a quasi-coherent sheaf, Invertible sheaves).

[F3]

Ampleness pulls back along finite morphisms: the pullback g∗L of an ample invertible sheaf along a finite morphism g is ample (Finite pullback preserves absolute ampleness).

[F4]

Ample powers are closed H-very ample over a proper finite-type base: for a proper finite-type morphism X→S with S Noetherian and L an ample invertible OX-module, there is d0≥1 such that L⊗d is closed H-very ample relative to S for every d≥d0, witnessed by a closed immersion into a relative projective space with O(1) pulling back to L⊗d (High powers of an ample line bundle embed a proper scheme, Locally Noetherian and Noetherian schemes, Proper morphisms, Locally finite type and finite type morphisms, Relative very ampleness in the finite projective-space convention).

[F5]

Serre vanishing: for a Noetherian commutative ring A, a scheme X projective over A in the finite-dimensional H-projective convention, an ample invertible OX-module L and a coherent OX-module F, there is m0 with Hq(X,F⊗L⊗m)=0 for every q>0 and every m≥m0 (Serre vanishing for coherent sheaves and ample twists, Projective morphisms before Proj, Coherent module sheaves, Sheaf cohomology as right derived global sections); a field is Noetherian and its spectrum is Noetherian (A field has only the zero ideal and itself, hence is Noetherian, The spectrum of a Noetherian ring is a Noetherian topological space).

[F6]

Divisor and tensor dictionary. The divisors D0 and A on the smooth curve are Weil divisors. The curve Cartier-to-Weil result Cartier and Weil divisors agree on a smooth curve identifies them as Cartier divisors; Invertible sheaf of cartier divisor constructs their associated sheaves, and The sheaf of a Cartier divisor is invertible proves those sheaves invertible. For every integer n≥0, repeated application of Addition of Cartier divisors is tensor product of their sheaves gives OC(D0+nA)≅OC(D0)⊗OC(A)⊗n≅OC(D0)⊗L⊗n. The rational-section identification with L(D0+nA) is the one in The space L(D), using Rational sections of line bundles are Cartier divisors. This is the actual interface used in step 4.1.

[F7]

Monotonicity of h1: for every divisor D and every effective divisor E≥0 one has h1(D+E)≤h1(D), where h1(D)=dim⁡kH1(C,OC(D)); equivalently h1 is antitone in the divisor (Adding points never raises h^1, and h^1 stabilizes, The Riemann-Roch dimension l(D)).

[F8]

The Axiom of Choice is inherited from the ample-powers theorem [F4] and Serre vanishing [F5]. In ZF, AC implies DC by AC implies DC implies countable choice, so the stated assumption supplies the Dependent Choice premise of the curve Cartier-to-Weil result in [F6] (The Axiom of Choice, The axiom of dependent choice: a relation in which every element is related to something admits an N-indexed chain). Apart from the integers supplied by the cited existence theorems, the proof makes no further selection.

[F9]

A closed H-very ample invertible sheaf on a scheme over the affine base Spec⁡k is ample in the absolute sense by Relative very ampleness implies relative ampleness; the implication applies to the closed immersion and pulled-back O(1) witness supplied in step 2.1. This is the route establishing ampleness of L⊗d for Serre vanishing, rather than an assumed closure of ampleness under tensor powers.

[F10]

The actual lemma Finite-dimensionality of the Riemann-Roch space applies to the given smooth proper geometrically integral curve and divisor D0; it proves that OC(D0) is coherent. This supplies the coherent-sheaf hypothesis of [F5] in step 3.1.

Proof

technique · pull the ample twisting sheaf on $\mathbb P^1_k$ back along the finite morphism, use that an ample power of the pullback embeds $C$ projectively, apply Serre vanishing to the twist of $\mathcal O_C(D_0)$, and remove the extra effective summand by monotonicity of $h^1$
1.1F1F2F3

Ampleness of L. By [F2] the twisting sheaf O(1) is ample on Pk1, and by the hypothesis φ is a finite k-morphism with OC(A)≅φ∗O(1); hence L=OC(A) is isomorphic to the pullback of an ample invertible sheaf along a finite morphism, and [F3] makes L ample.

2.1F1F4F5step 1.1

An ample power embeds C projectively. The curve C is proper and of finite type over the field k by [F1], so the structure morphism C→Spec⁡k is proper and of finite type, and Spec⁡k is Noetherian by [F5]; with L ample by step 1.1, [F4] provides an integer d≥1 such that L⊗d is closed H-very ample relative to Spec⁡k, witnessed by a closed immersion C↪PkN with O(1) pulling back to L⊗d. In particular C is projective over the Noetherian ring k in the H-projective convention of [F5].

3.1F5F9F10step 2.1

Serre vanishing for the twist by a power of L. The field k and Spec⁡k are Noetherian by [F5]. The sheaf OC(D0) is coherent by [F10], and the closed H-very ample witness for L⊗d from step 2.1 makes L⊗d ample by [F9]. Applying [F5] to the projective k-scheme C, this ample invertible sheaf and the coherent sheaf OC(D0) gives an integer m1 such that H1(C,OC(D0)⊗OCL⊗dm)=0for every m≥m1. Set m0:=max⁡(m1,0). Since the vanishing holds for every m≥m1, it holds for every m≥m0, and now all exponents used in the divisor translation are nonnegative.

4.1F6step 3.1

Translation to divisors. By [F6] one has OC(D0)⊗L⊗dm≅OC(D0+dmA) for every m≥0, with L=OC(A); since m0≥0, step 3.1 gives H1(C,OC(D0+dmA))=0 for every m≥m0.

5.1F1F7step 4.1

The stated form. Put n0:=dm0, which is nonnegative since d≥1 and m0≥0. By step 4.1 one has H1(C,OC(D0+n0A))=0, and this vanishing spreads to all n≥n0 and all effective E: write n=n0+r with r≥0, so D0+nA+E=D0+n0A+(rA+E), where rA+E is effective because A and E are effective; by [F7] applied to the divisor D0+n0A and the effective divisor rA+E one gets h1(D0+nA+E)≤h1(D0+n0A)=0. Thus H1(C,OC(D0+nA+E))=0 for every n≥n0 and every effective E.

6.1F7step 5.1

The equivalent form. If D is a divisor with D≥D0+n0A, then E:=D−(D0+n0A) has nonnegative coefficients, that is, E is effective, and D=D0+n0A+E; step 5.1 at n=n0 gives h1(D)=0. Conversely, if h1(D)=0 for every D≥D0+n0A, then for every n≥n0 and every effective E the divisor D0+nA+E satisfies D0+nA+E≥D0+n0A, so h1(D0+nA+E)=0; the two formulations are therefore equivalent.

7.1F1F4F5F6F7F8F9F10step 1.1step 2.1step 3.1step 4.1step 5.1step 6.1∎

Conclusion and choice accounting. Step 1.1 makes L=OC(A) ample as the pullback of the ample twisting sheaf of Pk1 along the finite morphism φ; step 2.1 exhibits C as projective over the Noetherian ring k through a closed H-very ample witness; step 3.1 applies Serre vanishing to the coherent twist by the ample sheaf L⊗d and replaces its bound by m0≥0; step 4.1 translates that vanishing to the divisors D0+dmA; and steps 5.1 and 6.1 spread it to all n≥n0 and all effective summands E, proving both formulations with n0=dm0. The integers d and m0 depend only on the morphism φ (through L and the closed immersion witness) and on D0, not on E or on deg⁡kE. No Serre duality and no threshold 2g−2 is used anywhere; the Choice premise used by the Cartier-to-Weil supplier is explicitly obtained from the stated Axiom of Choice by [F8].

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Riemann's theorem for sufficiently positive divisors

Statement

Assume the Axiom of Choice as inherited from the vanishing theorem. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field) of genus g=g(C) (Genus via the Euler characteristic), let φ:C→Pk1 be a finite k-morphism and let A be an effective divisor with OC(A)≅φ∗OPk1(1). Let D0 be a divisor on C (Divisors on a smooth proper curve) and let n0 be the integer supplied for D0 by Vanishing of H^1 in a fixed ample direction for this fixed morphism φ and divisor A. Then every divisor D on C with D≥D0+n0A satisfies h1(D)=0andl(D)=deg⁡k(D)+1−g, where l(D)=h0(D)=dim⁡kH0(C,OC(D)) and h1(D)=dim⁡kH1(C,OC(D)) (The Riemann-Roch dimension l(D)). In particular both conclusions hold for every divisor of the form D=D0+nA+E with n≥n0 and E effective.

Facts & Assumptions

Given: a field k, a smooth proper geometrically integral curve C over k of genus g, a finite k-morphism φ:C→Pk1, an effective divisor A with OC(A)≅φ∗O(1), a divisor D0 on C, and the integer n0 supplied for D0 by the vanishing theorem for this φ and A.

[F1]

Vanishing theorem: for the fixed curve, morphism φ and effective divisor A, the integer n0 satisfies H1(C,OC(D0+nA+E))=0 for every n≥n0 and every effective divisor E; equivalently h1(D)=0 for every divisor D with D≥D0+n0A (Vanishing of H^1 in a fixed ample direction).

[F2]

Riemann-Roch in Euler-characteristic form: for every divisor D on C one has h0(D)−h1(D)=χ(C,OC(D))=deg⁡k(D)+1−g, with no Serre duality used (Riemann-Roch for curves: the Euler-characteristic form).

[F3]

Notations: l(D)=h0(D)=dim⁡kH0(C,OC(D)) and hi(D)=dim⁡kHi(C,OC(D)) for i≥0, and g=g(C)=h1(C,OC)=1−χ(C,OC) is the genus (The Riemann-Roch dimension l(D), Genus via the Euler characteristic, Divisors on a smooth proper curve).

[F4]

The Axiom of Choice is available and is inherited from the vanishing theorem [F1] (through ampleness and Serre vanishing), the Riemann-Roch theorem [F2], and the dimension, genus and divisor interfaces [F3]; the argument below evaluates the two statements at the given divisor and selects nothing beyond the integer n0 supplied by [F1] (The Axiom of Choice).

Proof

technique · read the vanishing of $h^1$ off the vanishing theorem, substitute it into the Euler-characteristic form of Riemann-Roch, and record the explicit form $D_0+nA+E$
1.1F1

Vanishing above the threshold. Let D be a divisor with D≥D0+n0A. Writing E:=D−(D0+n0A)≥0 exhibits D=D0+n0A+E with E effective, so [F1] gives H1(C,OC(D))=0, that is h1(D)=0. In particular, for every n≥n0 and every effective E the divisor D=D0+nA+E satisfies D≥D0+n0A and therefore h1(D)=0.

2.1F2F3step 1.1

The dimension formula. For every divisor D with D≥D0+n0A, combining h1(D)=0 of step 1.1 with the Riemann-Roch identity [F2] gives l(D)=h0(D)=h1(D)+deg⁡k(D)+1−g=deg⁡k(D)+1−g by the notation [F3]: the section space has dimension exactly deg⁡k(D)+1−g, with no correction term.

3.1F1F4step 1.1step 2.1∎

The explicit form and choice accounting. Every divisor D=D0+nA+E with n≥n0 and E effective satisfies D≥D0+n0A, so step 1.1 gives h1(D)=0 and step 2.1 gives l(D)=deg⁡k(D)+1−g; the general divisor D≥D0+n0A is of this form with n=n0, so both formulations coincide. The integer n0 is the one supplied by the vanishing theorem for D0 and the fixed morphism φ; it is not chosen here, and the Axiom of Choice is inherited through [F1], [F2] and [F3], as recorded in [F4].

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

A genus-zero curve with a degree-one divisor is the projective line

Statement

Assume the Axiom of Choice as inherited from the curve, divisor and cohomology suppliers. Let k be a field and let C be a smooth proper geometrically integral curve over k (Curves over a field) with genus g(C)=0 (Genus via the Euler characteristic). Suppose that C admits a divisor of degree 1 (Degree divisor proper curve). Then C is isomorphic to Pk1; equivalently, if C has a k-rational closed point p, that is a closed point with [κ(p):k]=1 (The residue field at a point of an affine scheme), then C≅Pk1. The two hypotheses are equivalent by step 1.1.

The current Principal divisors on a normal proper curve have degree zero supplies deg⁡k(div⁡f)=0 in step 1.1. The divisor and function-space interfaces are Principal weil divisor and class group and The space L(D). The finite map and its pole-fibre degree in steps 3.1–4.1 are supplied by A nonconstant rational function defines a finite map to the projective line, which uses the current finite-flat fibre-degree result.

Facts & Assumptions

Given: a field k, a smooth proper geometrically integral curve C over k of genus g=g(C)=0, and a divisor D on C with deg⁡k(D)=1.

[F1]

The Riemann inequality: for every divisor E on C one has l(E)≥deg⁡k(E)+1−g, so here l(E)≥deg⁡k(E)+1 (The Riemann inequality, Genus via the Euler characteristic).

[F2]

Divisors, degrees and rational points: a divisor on C is a finite formal combination E=∑xnx[x] of closed points x, it is effective exactly when every nx≥0, and deg⁡k(E)=∑xnx[κ(x):k] with [κ(x):k]≥1 for every closed point; a closed point x is k-rational exactly when [κ(x):k]=1, and then deg⁡k([x])=1 (Divisors on a smooth proper curve, Degree divisor proper curve, The residue field at a point of an affine scheme).

[F3]

The Riemann-Roch space and principal divisors: for a divisor E, L(E)={f∈k(C)×:div⁡(f)+E≥0}∪{0} is a k-subspace of k(C), where div⁡(f)=∑xord⁡x(f)[x] uses the order of vanishing at each closed point; the constant functions c∈k× have div⁡(c)=0 and therefore lie in L(E) whenever E is effective, dim⁡kL(E)=l(E) (The space L(D), The Riemann-Roch dimension l(D)), and for every nonzero f∈L(E) the divisor div⁡(f)+E is effective and linearly equivalent to E (Effective divisors linearly equivalent to D are sections modulo scalars).

[F4]

Principal divisors on a proper curve have degree zero: for every nonzero rational function f∈k(C)×, deg⁡k(div⁡f)=0 (Principal divisors on a normal proper curve have degree zero). This is used at step 1.1.

[F5]

The map attached to a nonconstant function: for nonconstant f∈k(C)× there is a finite locally free morphism φf:C→Pk1 of degree [k(C):k(f)], a positive integer, whose fibre over infinity is the pole divisor (f)∞=∑ord⁡x(f)<0(−ord⁡x(f))[x] of degree [k(C):k(f)]; a nonzero rational function with no poles is algebraic over k and is a global unit (A nonconstant rational function defines a finite map to the projective line, Degree of a nonconstant morphism of curves).

[F6]

Birational curves: every birational rational map C⇢Pk1, that is, every dominant rational map whose pullback on function fields is an isomorphism, is represented by a k-isomorphism C→Pk1 (Birational smooth proper curves are isomorphic).

[F7]

Vector-space dimension: if dim⁡kV≥2 and W⊆V is a subspace of dimension one, then W≠V and there exists v∈V∖W (Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis).

[F8]

The Axiom of Choice is inherited from the curve, divisor and cohomology suppliers recorded above; the argument below works with the given curve and divisor, chooses one nonzero f∈L(D) and one h∈L([q])∖k, and selects nothing else (The Axiom of Choice).

Proof

technique · reduce a degree-one divisor to a rational point through a section of $L(D)$, then read a nonconstant function with pole divisor a single rational point, whose associated morphism to $\mathbb P^1_k$ has degree one and is therefore birational
1.1F1F2F3F4

From a degree-one divisor to a rational point. Let D be a divisor on C with deg⁡k(D)=1. By [F1] one has l(D)≥deg⁡k(D)+1−g=2, so dim⁡kL(D)≥2 and there is a nonzero f∈L(D). By [F3] the divisor E:=div⁡(f)+D is effective and linearly equivalent to D, and by [F4] deg⁡k(div⁡f)=0, so deg⁡k(E)=deg⁡k(D)=1. Write E=∑xnx[x] with nx≥0 by [F2]; then 1=deg⁡k(E)=∑xnx[κ(x):k] is a sum of nonnegative terms, so exactly one closed point q has nq≥1, with nq[κ(q):k]=1 and nx=0 for x≠q; hence nq=1 and [κ(q):k]=1, that is, q is a k-rational closed point by [F2]. Conversely, if p is a closed point with [κ(p):k]=1 then deg⁡k([p])=1 by [F2], so C admits a divisor of degree 1; this proves the equivalence of the two hypotheses of the statement.

2.1F1F2F3F7step 1.1

A nonconstant function with poles at most at q. With q as in step 1.1 one has deg⁡k([q])=[κ(q):k]=1 by [F2], so [F1] gives l([q])≥1+1−g=2. The divisor [q] is effective, so every constant c∈k× satisfies div⁡(c)+[q]=[q]≥0 and lies in L([q]) by [F3]; thus k⋅1⊆L([q]) is a subspace of dimension one, and since dim⁡kL([q])≥2 it is proper, so by [F7] there is h∈L([q])∖k⋅1. The function h is nonconstant, and h∈L([q]) means div⁡(h)+[q]≥0 by [F3].

3.1F3F5step 2.1

The pole divisor of h is [q]. From div⁡(h)+[q]≥0 at every closed point x≠q the order ord⁡x(h) is ≥0, and at q it is ≥−1; hence the pole divisor (h)∞=∑ord⁡x(h)<0(−ord⁡x(h))[x] of [F5] satisfies (h)∞≤[q]. Since h is nonconstant, [F5] exhibits the finite locally free morphism φh whose fibre over infinity is (h)∞, of degree [k(C):k(h)]≥1; in particular (h)∞≠0. As (h)∞≤[q] and [q] has coefficient one at q and zero elsewhere, the only nonzero effective divisor dominated by [q] that is nonzero is [q] itself, so (h)∞=[q].

4.1F2F5step 3.1

Degree one and birationality. By [F5] the degree of φh is deg⁡k(h)∞=deg⁡k[q]=1 by step 3.1 and [F2], that is [k(C):k(h)]=1; the pullback of the coordinate function of Pk1 is h, so the image of k(Pk1)→k(C) is k(h) and the extension k(C)/k(h) is trivial, k(C)=k(h). Hence φh is dominant with pullback an isomorphism of function fields, that is, φh is birational.

5.1F2F5F6F8step 1.1step 2.1step 3.1step 4.1∎

Conclusion. By [F6] the birational map φh:C⇢Pk1 of step 4.1 is represented by a k-isomorphism C→Pk1; hence C≅Pk1, which is the claim. The proof used one nonzero f∈L(D) in step 1.1, one nonconstant h∈L([q]) in step 2.1, and the morphism φh supplied by [F5]; the Axiom of Choice is inherited only through the suppliers recorded in [F8].

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Divisors on the projective line are classified by degree

Statement

Assume the Axiom of Choice as inherited from the cohomology, DVR/normality, degree and Cartier-divisor suppliers. It supplies the Dependent Choice premise of the curve Cartier-to-Weil route through AC implies DC implies countable choice. Let k be a field and let Pk1 be the projective line over k with standard affine chart U0=Spec⁡k[t], coordinate t=x1/x0, and point at infinity ∞=[0:1]=V(x0), the pole of t; the origin is the point [1:0]=V(x1) of U0, where t=0. Then:

  1. Pk1 is a smooth proper geometrically integral curve over k of genus 0;
  2. for every monic irreducible polynomial g∈k[t] of degree d, the closed point p=V(g) of U0 has [κ(p):k]=d and div⁡(g)=[p]−d[∞] for the divisor of the rational function g, so [p] is linearly equivalent to d[∞];
  3. the coordinate section x0 of O(1) vanishes exactly at infinity with multiplicity one, so div⁡(x0)=[∞] and O(1)≅O(∞) with deg⁡kO(1)=1, while the other coordinate section x1 vanishes exactly at the origin, with div⁡(x1)=[ [1:0] ]=[V(x1)];
  4. every divisor D on Pk1 is linearly equivalent to deg⁡k(D)[∞]; consequently the degree homomorphism deg⁡k ⁣:CaDiv⁡(Pk1)/Prin⁡(Pk1)→Z is an isomorphism.

Scaffold repair, recorded for the owner. The frozen scaffold statement wrote t=x1/x0 together with ∞=[1:0]=V(x1) and "the coordinate section x1". Those clauses are not simultaneously satisfiable: for g=t clause 2 would then read div⁡(t)=[p]−[∞]=0 at p=∞, and t is not constant. The statement above keeps every promised claim with the labels corrected to the running convention ∞=[0:1]=V(x0) of this page, and keeps the true statement about x1 as the final clause of (3).

Clauses 3 and 4 use the current interfaces Invertible sheaf of cartier divisor, Rational sections of line bundles are Cartier divisors, Cartier and Weil divisors agree on a smooth curve, The degree of a divisor descends to the Picard group of a normal proper curve and Principal weil divisor and class group. Their roles are recorded in [F14].

Facts & Assumptions

Given: the Axiom of Choice inherited from the cohomology, DVR/normality, degree and Cartier-divisor suppliers, with Dependent Choice supplied by AC implies DC implies countable choice; a field k, the projective line Pk1=P1 with its two standard charts U0=Spec⁡k[t] and U1=Spec⁡k[u], related on the overlap by t=u−1, and a monic irreducible polynomial g∈k[t] of degree d.

[F1]

A curve over k is a k-scheme that is geometrically integral, separated, of finite type and of chain dimension one; a smooth curve is a curve whose structure morphism is smooth in the local-standard-smooth convention, and a proper curve is one whose structure morphism is proper (Curves over a field, Smooth morphisms via local standard smooth presentations).

[F2]

The projective line Pk1 is the relative projective space Pk1 of Relative projective space from standard charts with standard charts U0=Spec⁡k[x1(0)] and U1=Spec⁡k[x0(1)] glued along D(x1(0)) and D(x0(1)) by x1(0)↦1/x0(1), and the charts and transitions are stable under base change; it is canonically Proj⁡k[x0,x1] (Projective space is Proj of a polynomial ring). The two-affine model of Two-affine projective line and its twists is the gluing of the same two affine schemes along the same open subschemes by the same transition isomorphism and is therefore canonically isomorphic to Pk1 by the uniqueness clause of Gluing affine schemes along compatible open isomorphisms; on the overlap the twists are glued with frames e0 on U0 and e∞ on U1 related by e∞=tne0 (Two-affine projective line and its twists, The twist index on the projective line is an isomorphism invariant, Twisting sheaf on Proj). In the identification with Proj⁡k[x0,x1], the origin [1:0]=V(x1) is the point t=0 of U0, and the point at infinity ∞=[0:1]=V(x0) is the point u=0 of U1, outside U0.

[F3]

PSn→S is proper and of finite type for every scheme S and every n≥0, and a proper morphism is separated (Finite-dimensional projective space is proper over every base, Projective space is of finite type over its base, Proper morphisms). So Pk1→Spec⁡k is proper, separated and of finite type.

[F4]

A polynomial ring R[x1,…,xn] is a standard smooth R-algebra through the presentation with c=0 variables and g=1, and a morphism of finite-type k-schemes is smooth in the local-standard-smooth convention when every source point has affine neighbourhoods on which the induced ring map is standard smooth at the corresponding prime (Standard smooth presentations and locally standard smooth maps, Smooth morphisms via local standard smooth presentations).

[F5]

dim⁡A=trdeg⁡kFrac⁡(A) for a finite-type k-domain A (Affine-domain dimension equals transcendence degree), in particular dim⁡k[t]=1 (A polynomial ring in n variables over a field has dimension n, Krull dimension of a nonzero ring); on a finite affine cover of a Noetherian space the chain dimension is the supremum of the chart dimensions (Dimension can be computed on an open cover, Chain dimension and the empty-space convention). Strict chains of nonempty irreducible closed subsets of Spec⁡A correspond to strict chains of prime ideals, since an irreducible closed Z is V(p) for its prime defining ideal and Z⊆Z′ matches the reverse inclusion of the ideals (A Zariski-closed subset is irreducible exactly when its radical defining ideal is prime, and then it has a unique generic point).

[F6]

An affine scheme is integral when its coordinate ring is a nonzero domain; such a scheme is nonempty, reduced and irreducible, and a finite-type algebra over a field is Noetherian (Integral affine schemes, Reduced affine schemes). A space is irreducible exactly when it is nonempty and every two nonempty open subsets meet, equivalently when every nonempty open subset is dense; a nonempty open subspace of an irreducible space is irreducible (Irreducibility via nonempty open subsets, connectedness and open subspaces). Geometric integrality means integrality of the algebraic-closure fibre (Curves over a field).

[F7]

For a point x of a scheme, κ(x)=OX,x/mx, and for p∈Spec⁡A one has κ(p)≅Frac⁡(A/p) (The residue field at a point of an affine scheme); a point of Spec⁡R is closed exactly when it is a maximal ideal (The closed points of the prime spectrum are exactly the maximal ideals); F[x] is a principal ideal domain and (p) is maximal exactly when the nonconstant p is irreducible (For every field F, F[x] is a principal ideal domain, For a nonconstant p in F[x], the ideal (p) is maximal and F[x]/(p) is a field exactly when p is irreducible, and the quotient of F[x] by a maximal ideal is identified with the residue field by The residue field at a point of an affine scheme). For an algebraic element with minimal polynomial of degree n, the simple extension has degree n and power basis 1,a,…,an−1; the degree [K:F] is dim⁡FK (A simple algebraic extension is its minimal-polynomial quotient and has power basis 1,a,…,an−1 and degree n, The degree [K:F]=dim⁡FK of a finite field extension).

[F8]

A divisor on a curve is a finite formal Z-linear combination of closed points, its support is the finite set of points with nonzero coefficient, D=D+−D− for the positive and negative parts, D≥0 means D effective, and deg⁡kD=∑xnx[κ(x):k] is a group homomorphism (Degree divisor proper curve, Divisors on a smooth proper curve, Divisor support positive negative parts).

[F9]

For a closed point x of a smooth curve C over k the local ring OC,x is a discrete valuation ring with a uniformizer tx; every nonzero rational function f∈k(C)× has a well-defined order ord⁡x(f)∈Z, every nonzero element of OC,x is a unit times a power of tx, and the order is a group homomorphism, so it is additive in products and vanishes on units (Local rings at closed points of smooth curves are discrete valuation rings, Discrete valuations).

[F10]

For an integral finite-type k-scheme X the stalk at the generic point is the function field and is canonically Frac⁡Γ(U,OX) for every nonempty affine open U, so k(P1)=k(t)=Frac⁡k[t] here (Function field of an integral finite-type scheme); k[t] is a unique factorisation domain, so every nonzero element of k(t) is a unit of k times a finite product of powers of monic irreducible polynomials (For every field F, F[x] is a principal ideal domain, Every principal ideal domain is a unique factorisation domain).

[F11]

H0(Pk1,O(d)) is the degree-d part k[x0,x1]d of the polynomial ring for d≥0 and vanishes for d<0; the coordinate forms x0,x1 are global sections of O(1) (Global sections of projective twists, Twisting sheaf on Proj); and H1(Pk1,O(d))=0 for every d≥−1, in particular H1(Pk1,O)=0 (Top cohomology of projective twists). The genus of a smooth proper geometrically integral curve is g(C)=h1(C,OC)=dim⁡kH1(C,OC) (Genus via the Euler characteristic).

[F12]

A rational section of an invertible sheaf L on an integral scheme is a nonzero element of the one-dimensional K(X)-vector space Lη: a nonzero global section whose stalk at the generic point is nonzero qualifies (Rational section line bundle). On U0 the twist O(1) is free with frame x0 and on U1 with frame x1, since x1=tx0 on the overlap; under the identification of the two models with the two-affine frames e0,e∞ of [F2], these frames correspond as x0↔e0 and x1↔e∞, so x1=tx0 is the relation e∞=te0 [F2].

[F13]

Cartier divisors form a group CaDiv⁡(X) whose elements have local meromorphic equations, and the principal Cartier divisors — the images of global meromorphic units — form a subgroup Prin⁡(X); the quotient CaDiv⁡(X)/Prin⁡(X) is the group of linear equivalence classes (Cartier divisor).

[F14]

Cartier, Weil, and Picard interfaces. The current Invertible sheaf of cartier divisor constructs OX(D) by local equations, gives OX(0)≅OX, and uses the sign convention positive for zeros. The current Rational sections of line bundles are Cartier divisors associates to a nonzero rational section s its Cartier divisor div⁡(s) and an isomorphism O(div⁡(s))≅L carrying the canonical section to s. The current Cartier and Weil divisors agree on a smooth curve identifies Cartier divisors and Weil divisors on a smooth proper geometrically integral curve and preserves principal divisors. The degree of an invertible sheaf is the homomorphism supplied by The degree of a divisor descends to the Picard group of a normal proper curve, with deg⁡kOC(D)=deg⁡kD; the principal Weil divisor subgroup is defined in Principal weil divisor and class group. Regular local domains are integrally closed by regular local rings are normal, as used in [F9] to establish normality before applying the degree homomorphism. AC supplies the DC premise in the Cartier-to-Weil source through AC implies DC implies countable choice.

Proof

Given: the Axiom of Choice inherited from the cohomology, DVR/normality, degree and Cartier-divisor suppliers, with Dependent Choice supplied by AC implies DC implies countable choice; a field k, the projective line Pk1=P1 with its two standard charts U0=Spec⁡k[t] and U1=Spec⁡k[u], related on the overlap by t=u−1, and a monic irreducible polynomial g∈k[t] of degree d.

[F1] A curve over k is a k-scheme that is geometrically integral, separated, of finite type and of chain dimension one; a smooth curve is a curve whose structure morphism is smooth in the local-standard-smooth convention, and a proper curve is one whose structure morphism is proper (Curves over a field, Smooth morphisms via local standard smooth presentations).

[F2] The projective line Pk1 is the relative projective space Pk1 of Relative projective space from standard charts with standard charts U0=Spec⁡k[x1(0)] and U1=Spec⁡k[x0(1)] glued along D(x1(0)) and D(x0(1)) by x1(0)↦1/x0(1), and the charts and transitions are stable under base change; it is canonically Proj⁡k[x0,x1] (Projective space is Proj of a polynomial ring). The two-affine model of Two-affine projective line and its twists is the gluing of the same two affine schemes along the same open subschemes by the same transition isomorphism and is therefore canonically isomorphic to Pk1 by the uniqueness clause of Gluing affine schemes along compatible open isomorphisms; on the overlap the twists are glued with frames e0 on U0 and e∞ on U1 related by e∞=tne0 (Two-affine projective line and its twists, The twist index on the projective line is an isomorphism invariant, Twisting sheaf on Proj). In the identification with Proj⁡k[x0,x1], the origin [1:0]=V(x1) is the point t=0 of U0, and the point at infinity ∞=[0:1]=V(x0) is the point u=0 of U1, outside U0.

[F3] PSn→S is proper and of finite type for every scheme S and every n≥0, and a proper morphism is separated (Finite-dimensional projective space is proper over every base, Projective space is of finite type over its base, Proper morphisms). So Pk1→Spec⁡k is proper, separated and of finite type.

[F4] A polynomial ring R[x1,…,xn] is a standard smooth R-algebra through the presentation with c=0 variables and g=1, and a morphism of finite-type k-schemes is smooth in the local-standard-smooth convention when every source point has affine neighbourhoods on which the induced ring map is standard smooth at the corresponding prime (Standard smooth presentations and locally standard smooth maps, Smooth morphisms via local standard smooth presentations).

[F5] dim⁡A=trdeg⁡kFrac⁡(A) for a finite-type k-domain A (Affine-domain dimension equals transcendence degree), in particular dim⁡k[t]=1 (A polynomial ring in n variables over a field has dimension n, Krull dimension of a nonzero ring); on a finite affine cover of a Noetherian space the chain dimension is the supremum of the chart dimensions (Dimension can be computed on an open cover, Chain dimension and the empty-space convention). Strict chains of nonempty irreducible closed subsets of Spec⁡A correspond to strict chains of prime ideals, since an irreducible closed Z is V(p) for its prime defining ideal and Z⊆Z′ matches the reverse inclusion of the ideals (A Zariski-closed subset is irreducible exactly when its radical defining ideal is prime, and then it has a unique generic point).

[F6] An affine scheme is integral when its coordinate ring is a nonzero domain; such a scheme is nonempty, reduced and irreducible, and a finite-type algebra over a field is Noetherian (Integral affine schemes, Reduced affine schemes). A space is irreducible exactly when it is nonempty and every two nonempty open subsets meet, equivalently when every nonempty open subset is dense; a nonempty open subspace of an irreducible space is irreducible (Irreducibility via nonempty open subsets, connectedness and open subspaces). Geometric integrality means integrality of the algebraic-closure fibre (Curves over a field).

[F7] For a point x of a scheme, κ(x)=OX,x/mx, and for p∈Spec⁡A one has κ(p)≅Frac⁡(A/p) (The residue field at a point of an affine scheme); a point of Spec⁡R is closed exactly when it is a maximal ideal (The closed points of the prime spectrum are exactly the maximal ideals); F[x] is a principal ideal domain and (p) is maximal exactly when the nonconstant p is irreducible (For every field F, F[x] is a principal ideal domain, For a nonconstant p in F[x], the ideal (p) is maximal and F[x]/(p) is a field exactly when p is irreducible, and the quotient of F[x] by a maximal ideal is identified with the residue field by The residue field at a point of an affine scheme). For an algebraic element with minimal polynomial of degree n, the simple extension has degree n and power basis 1,a,…,an−1; the degree [K:F] is dim⁡FK (A simple algebraic extension is its minimal-polynomial quotient and has power basis 1,a,…,an−1 and degree n, The degree [K:F]=dim⁡FK of a finite field extension).

[F8] A divisor on a curve is a finite formal Z-linear combination of closed points, its support is the finite set of points with nonzero coefficient, D=D+−D− for the positive and negative parts, D≥0 means D effective, and deg⁡kD=∑xnx[κ(x):k] is a group homomorphism (Degree divisor proper curve, Divisors on a smooth proper curve, Divisor support positive negative parts).

[F9] For a closed point x of a smooth curve C over k the local ring OC,x is a discrete valuation ring with a uniformizer tx; every nonzero rational function f∈k(C)× has a well-defined order ord⁡x(f)∈Z, every nonzero element of OC,x is a unit times a power of tx, and the order is a group homomorphism, so it is additive in products and vanishes on units (Local rings at closed points of smooth curves are discrete valuation rings, Discrete valuations).

[F10] For an integral finite-type k-scheme X the stalk at the generic point is the function field and is canonically Frac⁡Γ(U,OX) for every nonempty affine open U, so k(P1)=k(t)=Frac⁡k[t] here (Function field of an integral finite-type scheme); k[t] is a unique factorisation domain, so every nonzero element of k(t) is a unit of k times a finite product of powers of monic irreducible polynomials (For every field F, F[x] is a principal ideal domain, Every principal ideal domain is a unique factorisation domain).

[F11] H0(Pk1,O(d)) is the degree-d part k[x0,x1]d of the polynomial ring for d≥0 and vanishes for d<0; the coordinate forms x0,x1 are global sections of O(1) (Global sections of projective twists, Twisting sheaf on Proj); and H1(Pk1,O(d))=0 for every d≥−1, in particular H1(Pk1,O)=0 (Top cohomology of projective twists). The genus of a smooth proper geometrically integral curve is g(C)=h1(C,OC)=dim⁡kH1(C,OC) (Genus via the Euler characteristic).

[F12] A rational section of an invertible sheaf L on an integral scheme is a nonzero element of the one-dimensional K(X)-vector space Lη: a nonzero global section whose stalk at the generic point is nonzero qualifies (Rational section line bundle). On U0 the twist O(1) is free with frame x0 and on U1 with frame x1, since x1=tx0 on the overlap; under the identification of the two models with the two-affine frames e0,e∞ of [F2], these frames correspond as x0↔e0 and x1↔e∞, so x1=tx0 is the relation e∞=te0 [F2].

[F13] Cartier divisors form a group CaDiv⁡(X) whose elements have local meromorphic equations, and the principal Cartier divisors — the images of global meromorphic units — form a subgroup Prin⁡(X); the quotient CaDiv⁡(X)/Prin⁡(X) is the group of linear equivalence classes (Cartier divisor).

[F14] Cartier, Weil, and Picard interfaces. The current Invertible sheaf of cartier divisor constructs OX(D) by local equations, gives OX(0)≅OX, and uses the sign convention positive for zeros. The current Rational sections of line bundles are Cartier divisors associates to a nonzero rational section s its Cartier divisor div⁡(s) and an isomorphism O(div⁡(s))≅L carrying the canonical section to s. The current Cartier and Weil divisors agree on a smooth curve identifies Cartier divisors and Weil divisors on a smooth proper geometrically integral curve and preserves principal divisors. The degree of an invertible sheaf is the homomorphism supplied by The degree of a divisor descends to the Picard group of a normal proper curve, with deg⁡kOC(D)=deg⁡kD; the principal Weil divisor subgroup is defined in Principal weil divisor and class group. Regular local domains are integrally closed by regular local rings are normal, as used in [F9] to establish normality before applying the degree homomorphism. AC supplies the DC premise in the Cartier-to-Weil source through AC implies DC implies countable choice.

Proof technique: establish the projective-line curve hypotheses first; then compute the closed-point orders, the coordinate divisors and their degrees, and finally reduce every divisor to a multiple of infinity.

1.1F2

Chart data and the two special points. By [F2] the charts U0 and U1 cover Pk1, with overlap Spec⁡k[t,t−1]=Spec⁡k[u,u−1] and t=u−1. The complement of U0 is the point u=0 in U1, namely ∞=[0:1]=V(x0); the origin [1:0]=V(x1) is the point t=0 in U0.

1.2F2F7

Closed points. The closed points of U0=Spec⁡k[t] are the maximal ideals (g) generated by monic irreducible polynomials g∈k[t], by [F7]. For such a g of degree d, its homogenization x0dg(x1/x0) defines a closed point in U0 and does not vanish at ∞, since its value at [0:1] is 1. The only point outside U0 is ∞, which is closed on U1. Thus the closed points of Pk1 are the points p=V(g) for monic irreducible g, together with ∞. Every nonempty open subset meets U0, since {∞} is not open.

1.3F3

Properness and finite type. By [F3] the structure morphism Pk1→Spec⁡k is proper and of finite type; a proper morphism is separated, so Pk1 is separated over k.

1.4F2F5F6

Chain dimension one. The two charts are spectra of the Noetherian rings k[t] and k[u], each of dimension one by [F5]. The finite affine cover makes Pk1 Noetherian, and [F5] computes its chain dimension as the supremum of the chart dimensions, namely one.

1.5F1F2F3F4

Smoothness. Each chart ring k[t] or k[u] is standard smooth over k through the presentation with no equations and one free variable, by [F4]. The charts cover the source, so the structure morphism is smooth in the convention of [F1].

2.1F2F6step 1.2

Integrality and geometric integrality. Any two nonempty open subsets of Pk1 meet U0 by step 1.2, and their intersections with U0 meet because k[t] is a domain. Hence Pk1 is irreducible. Its local rings are localizations of k[t] or k[u], so they are domains and the scheme is reduced; it is therefore integral. After base change to an algebraic closure kˉ, [F2] gives the same two-chart description with kˉ[t] and kˉ[u], so the same irreducibility and reducedness proof shows that the base change is integral. Thus Pk1 is geometrically integral.

2.2F7step 1.2

Residue degrees of finite points. Let g∈k[t] be monic irreducible of degree d and let p=V(g). Its residue field is k[t]/(g), a simple extension generated by the class of t with minimal polynomial g; by [F7], [κ(p):k]=d. At infinity the residue field is k, since ∞ is the maximal ideal (u) of k[u].

3.1F11step 1.3step 1.4step 1.5step 2.1

Genus zero. Steps 1.3, 1.4, 1.5 and 2.1 show that Pk1 is a smooth proper geometrically integral curve. By [F11], H1(Pk1,OPk1)=H1(Pk1,OPk1(0))=0; the genus definition in [F11] therefore gives g(Pk1)=0.

3.2F9F10F14step 1.3step 2.1

Normality. For each closed point the local ring is a DVR by [F9], and the generic local ring is the function field k(t), a field by [F10]. These local rings are regular; by the regular-local normality theorem in [F14] they are integrally closed. Hence Pk1 is normal as well as proper, so the degree homomorphism on its Picard group in [F14] applies.

3.3F7F9F10step 1.1step 1.5step 2.1

Local orders of g. For p=V(g), the local ring is k[t](g); the maximal ideal is generated by the uniformizer g, so ord⁡p(g)=1. At every other finite point V(g′), with g′ a distinct monic irreducible, g is a unit and its order is zero. At infinity put u=t−1. Writing g(t)=td+cd−1td−1+⋯+c0=u−dh(u), where h(0)=1, shows that h is a unit in k[u](u) and ord⁡∞(g)=−d. These are the DVR orders of [F9], applicable now that the curve hypotheses have been established.

3.4F2F11F12F14step 1.1step 2.1

Coordinate sections and their divisors. By [F11] the global sections x0,x1 lie in H0(Pk1,O(1)). The sheaf O(1) has frame x0 on U0 and frame x1 on U1, with x1=tx0. Therefore x0 has local coefficients 1 on U0 and u on U1, while x1 has coefficients t on U0 and 1 on U1; these nonzero global sections qualify as rational sections by [F12]. By the rational-section dictionary of [F14], div⁡(x0)=[∞] and div⁡(x1)=[1:0]=[V(x1)], with the latter the origin. The same dictionary identifies O(div⁡(x0)) with O(1).

3.5F8F9F14step 1.5step 2.1

Additivity of the divisor map. For f1,f2∈k(t)×, additivity of each DVR order in [F9] gives div⁡(f1f2)=div⁡(f1)+div⁡(f2); constants in k× are units at every point and have zero divisor.

4.1F8F14step 3.2step 2.2step 3.4

Degree of O(1). By step 3.2 the proper curve Pk1 is normal, so [F14] gives deg⁡kO(1)=deg⁡kdiv⁡(x0)=deg⁡k[∞]. Since [κ(∞):k]=1 by step 2.2, this degree is 1. Thus O(1)≅O(∞) and has degree one, as in clause 3.

4.2F8F13F14step 2.2step 3.3

The divisor of a monic irreducible. By step 3.3 the only nonzero orders of g occur at p=V(g) and ∞, with orders 1 and −d. Thus div⁡(g)=[p]−d[∞]. Its degree is deg⁡k([p]−d[∞])=[κ(p):k]−d[κ(∞):k]=d−d=0, using [κ(p):k]=d and [κ(∞):k]=1 from step 2.2. The divisor is principal by [F14], so [p] is linearly equivalent to d[∞]. This proves clause 2.

5.1F8F10step 2.2step 4.2step 3.5

Principal divisors have degree zero. By [F10], every nonzero f∈k(t) has a finite factorization c∏igini with c∈k×, distinct monic irreducibles gi, and integers ni. By step 3.5 and step 4.2, div⁡(f)=∑ini([pi]−di[∞]). Each summand has degree [κ(pi):k]−di[κ(∞):k]=di−di=0, so additivity of divisor degree gives deg⁡kdiv⁡(f)=0.

5.2F8F13step 1.2step 4.2step 3.5

Every divisor is linearly equivalent to its degree times infinity. Let D=∑xnx[x] be any divisor on Pk1. By step 1.2, its finite support consists of points pi=V(gi) for monic irreducibles gi of degrees di, together with a possible term m[∞]. Thus D=∑ini[pi]+m[∞], where each ni∈Z, and deg⁡kD=∑inidi+m by [F8]. Using step 4.2 for each gi and the finite product ∏igini∈k(t)×, including negative exponents, gives D−deg⁡k(D)[∞]=∑ini([pi]−di[∞])=div⁡ ⁣(∏igini). Hence D∼deg⁡k(D)[∞], proving the first assertion of clause 4.

6.1F8F14step 4.1step 5.1step 5.2

The degree isomorphism on Weil divisor classes. Degree is surjective because deg⁡k(m[∞])=m for every m∈Z. It vanishes on principal divisors by step 5.1, so it descends to Div⁡(Pk1)/Prin⁡(Pk1)→Z. If a divisor has degree zero, step 5.2 makes it principal; thus the descended map is injective and is an isomorphism.

7.1F13F14step 1.3step 1.4step 1.5step 2.1step 6.1

The Cartier divisor-class statement. By the Cartier-to-Weil isomorphism in [F14], established for the smooth proper geometrically integral curve in steps 1.3, 1.4, 1.5 and 2.1, the map CaDiv⁡(Pk1)→Div⁡(Pk1) is an isomorphism compatible with principal divisors. It therefore induces an isomorphism of the corresponding divisor-class groups. Transporting the degree isomorphism of step 6.1 proves deg⁡k:CaDiv⁡(Pk1)/Prin⁡(Pk1)⟶Z is an isomorphism.

8.1F5F9F10F11F14step 1.3step 1.4step 1.5step 2.1step 3.1step 3.4step 4.1step 4.2step 7.1∎

Conclusion and choice accounting. Steps 1.3, 1.4, 1.5, 2.1 and 3.1 prove clause 1, step 4.2 proves clause 2, steps 3.4 and 4.1 prove clause 3, and step 7.1 proves clause 4. The Axiom of Choice enters through the cohomology, local DVR, UFD and Cartier/Picard suppliers; AC supplies the Dependent Choice premise of the curve Cartier-to-Weil and degree routes by AC implies DC implies countable choice. All other listings and products are finite and all fields k and polynomial degrees d≥1 are allowed.

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

The Picard group of the projective line

Statement

Assume the Axiom of Choice as inherited from the divisor, degree and twisting-sheaf suppliers. For every field k the degree homomorphism induces an isomorphism Pic⁡(Pk1)→Z. Under it the class of the twisting sheaf OPk1(d) corresponds to d, so [O(1)] is a generator, and every invertible sheaf on Pk1 is isomorphic to OPk1(d) for a unique integer d.

Facts & Assumptions

Given: a field k, the projective line Pk1 with structure sheaf O=OPk1, and the twisting sheaves O(d) for d∈Z.

[F1]

An invertible OX-module is one locally isomorphic to OX; the Picard group Pic⁡(X) is the abelian group of isomorphism classes [L] of invertible modules under [L]⋅[M]=[L⊗OXM], with identity [OX] and inverse [L∨] (Picard group of a scheme, Invertible sheaves).

[F2]

Pk1 is a smooth proper geometrically integral curve over k; for every divisor D on Pk1 the difference D−deg⁡k(D)[∞] is a principal Cartier divisor, so the degree homomorphism deg⁡k ⁣:CaDiv⁡(Pk1)/Prin⁡(Pk1)→Z is an isomorphism of groups; and O(1)≅O(∞)=O([∞]) with deg⁡kO(1)=deg⁡k[∞]=[κ(∞):k]=1 (Divisors on the projective line are classified by degree).

[F3]

Cartier divisors on a scheme X form a group CaDiv⁡(X), a Cartier divisor being represented by local meromorphic equations; the principal Cartier divisors form a subgroup Prin⁡(X) which is the image of the global meromorphic units, so two Cartier divisors are linearly equivalent exactly when their difference is principal (Cartier divisor).

[F4]

For a commutative nonnegatively graded ring S and X=Proj⁡S, the twisting sheaf is OX(n)=S(n)~ with Γ(D+(f),OX(n))=S(n)(f), multiplication of the graded ring gives morphisms OX(m)⊗OX(n)→OX(m+n), and OX(0)⊗OX(n)→OX(n) is the canonical identification; over a field F and the standard charts D+(xi) of PF1=Proj⁡F[x0,x1], the localised degree-zero part S(0)(xi) is the polynomial ring F[xj/xi] in the ratio variable and the degree-n part S(n)(xi) is its free module of rank one on the generator xin, so the displayed multiplication morphisms are isomorphisms on each chart (Twisting sheaf on Proj, Tensor product of sheaves of modules). Compatible local sheaves with their overlap identifications glue uniquely, and invertible sheaves glued from free rank-one sheaves with matching frames and transition units are isomorphic (Compatible local sheaves glue uniquely up to unique isomorphism). On Pk1 the twists defined on the standard charts by the prescription e1=une0 satisfy O(n)≅O(m) if and only if n=m (The twist index on the projective line is an isomorphism invariant).

[F5]

The actual Cartier-to-Picard dictionary sends a Cartier divisor D to [OX(D)], is a group homomorphism with kernel the principal Cartier divisors, and is surjective when X is integral (On an integral scheme, Cartier divisors modulo principal divisors compute the Picard group). The associated sheaf has local frame fi−1 for local equation fi, and OX(0)≅OX (Invertible sheaf of cartier divisor). Addition of Cartier divisors corresponds to tensor product, and OX(−D)≅OX(D)∨ (Addition of Cartier divisors is tensor product of their sheaves). On a normal proper integral curve C over k, [OC(D)]↦deg⁡kD is a well-defined group homomorphism Pic⁡(C)→Z (The degree of a divisor descends to the Picard group of a normal proper curve); [F2] verifies that X=Pk1 satisfies these hypotheses.

[F6]

The Axiom of Choice is inherited from the divisor classification [F2], the twisting-sheaf construction [F4], and the degree-descent theorem in [F5]. The Cartier-to-Picard dictionary, associated Cartier-divisor sheaf and addition/tensor identifications in [F5] use no choice principle; no additional choice principle is needed below (The Axiom of Choice).

Proof

technique · direct; transport the divisor-class isomorphism of [F2] across the Cartier-to-Picard dictionary of [F5], then identify the integer attached to $\mathcal O(d)$ by the transition computation of [F4]
1.1F2F3

The degree isomorphism on divisor classes. By [F2] the homomorphism deg⁡k from the group CaDiv⁡(Pk1)/Prin⁡(Pk1) of Cartier divisor classes to Z is an isomorphism: it is well defined by [F3], surjective because m[∞] has degree m, and injective because a degree-zero divisor is principal.

1.2F5

The Cartier-to-Picard dictionary. Since Pk1 is integral, the actual dictionary [F5] induces an isomorphism CaDiv⁡(Pk1)/Prin⁡(Pk1)→Pic⁡(Pk1) with inverse [O(D)]↦ the class of D. Its compatibility with addition and duals is given by [F5] and the local tensor identifications.

2.1F5step 1.1step 1.2

The composite isomorphism. Composing the inverse of the isomorphism of step 1.2 with the degree isomorphism of step 1.1 gives a group isomorphism Pic⁡(Pk1)→Z. By the degree-descent supplier [F5], it is given on classes by [O(D)]↦deg⁡kD.

3.1F2F4F5step 1.2step 2.1

The twists and the degree. Let d∈Z. By [F2], O(1)≅O([∞])=O(∞), and by the addition and dual isomorphisms of [F5] applied to the multiple d[∞] one has O(d[∞])≅O([∞])⊗d≅O(1)⊗d. To compare this with the twisting sheaf, let Ui=D+(xi) be the standard charts of Pk1=Proj⁡k[x0,x1]; by [F4] the module S(n)(xi)=xinS(0)(xi) is free of rank one over S(0)(xi) on the generator xin, so O(1)⊗d is free of rank one on Ui with frame xid, and on the overlap the frames satisfy x1d=(x1/x0)dx0d with the unit (x1/x0)d; this is exactly the transition unit of the twist of index d in [F4], so the gluing uniqueness of [F4] gives O(1)⊗d≅O(d) for every d (for d<0 duals invert the transition units, and the identification of [F5] provides the dual isomorphism O(−D)≅O(D)∨). Hence O(d)≅O(d[∞]), and the isomorphism of step 2.1 sends [O(d)] to deg⁡k(d[∞])=d⋅[κ(∞):k]=d.

4.1F1F4F5step 2.13.1

Generator and uniqueness. By step 3.1 the class [O(1)] maps to 1, so it generates Pic⁡(Pk1) under the isomorphism of step 2.1, and d↦[O(d)] is a two-sided inverse Z→Pic⁡(Pk1): it is a group homomorphism because O(d)⊗O(e)≅O(d+e) by the same comparison with the addition isomorphisms of [F5], and it is inverse to the isomorphism of step 2.1. Consequently every invertible sheaf L on Pk1 satisfies L≅O(d) for the integer d determined by [L], and this d is unique by [F4] (no two distinct twists are isomorphic).

5.1F2F4F6step 2.13.14.1∎

Conclusion and choice accounting. Steps 2.1, 3.1 and 4.1 prove the statement: the degree homomorphism induces an isomorphism Pic⁡(Pk1)→Z, [O(d)]↦d, so [O(1)] is a generator and every invertible sheaf is isomorphic to a unique twist. The Axiom of Choice is inherited through the divisor classification [F2], the twisting-sheaf construction [F4] and the degree-descent theorem [F5], as recorded in [F6]; the remaining arguments multiply finitely many transition units, use the single chart cover of Pk1, and select the integer d determined by a class, so no further choice is made.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Torsion-free coherent modules on a smooth curve are locally free

Statement

Assume the Axiom of Choice as inherited from the DVR and coherence suppliers. Let k be a field, let C be a smooth curve over k (Curves over a field) and let M be a coherent OC-module (Coherent module sheaves). Then:

  1. M is finite locally free (Locally free sheaves of finite rank) if and only if M is torsion-free in the following local sense: there are no open U⊆C, nonzero m∈M(U) and nonzero a∈OC(U) with a m=0.
  2. If M is not torsion-free in that sense, then M has a nonzero global section; in particular a nonzero coherent module killed locally by a nonzerodivisor has a nonzero global section.
  3. A subsheaf of a locally free OC-module is torsion-free, and a coherent torsion-free OC-module is finite locally free.

Facts & Assumptions

Given: a field k, a smooth curve C over k, and a coherent OC-module M.

[F1]

C is a nonempty integral scheme, so C is irreducible and every two nonempty open subsets of C meet. Every nonempty open W⊆C has an injective restriction map Γ(W,OC)→k(C): on any affine open Spec⁡A⊆W, this is the injection from the domain A into its fraction field. Thus a nonzero regular section on W has nonzero image in k(C), and its germ at every point of W maps to that same nonzero element. Every point of C is either its generic point or a closed point (Proper closed subsets of a curve are finite); at a closed point x the local ring OC,x is a discrete valuation ring, hence a principal ideal domain (Curves over a field, Integral schemes, Local rings at closed points of smooth curves are discrete valuation rings, Every DVR is a PID), while at the generic point η the local ring OC,η=k(C) is the function field, a field and hence also a principal ideal domain; in either case OC,x is a principal ideal domain.

[F2]

A finitely generated torsion-free module over a principal ideal domain is free (Every finitely generated torsion-free module over a PID is free).

[F3]

M is quasi-coherent of finite type; a stalk relation axmx=0 with ax,mx≠0 is represented by sections a,m over some open neighbourhood with am=0, and a section with nonzero germ at a point is a nonzero section (The stalk of a presheaf at a point, Modules on a ringed space).

[F4]

Coherent modules on the locally Noetherian scheme C are finitely presented, and for a finitely presented quasi-coherent module the locus of points at which the stalk is free of a given rank is open, with the module free of that rank on a neighbourhood of each such point (Coherent sheaves on a locally Noetherian scheme, Finite type and finitely presented module sheaves, Locally Noetherian and Noetherian schemes, Openness of the finite free locus).

[F5]

Sections of a sheaf over two open sets that agree on the intersection glue to a section over the union, and a section is nonzero once it is nonzero on one member of a cover (A sheaf on a topological space, Subsheaves).

[F6]

The Axiom of Choice enters only through the coherence and local-ring suppliers of [F1] and [F4]; the points selected below are chosen from sets known to be nonempty, and the proof makes no further choice (The Axiom of Choice).

[F7]

Every proper closed subset of an integral finite-type curve is a finite set of closed points, and every point other than the generic point is closed (Proper closed subsets of a curve are finite).

Proof

technique · direct; reduce the torsion-freeness question to the stalks, which are finitely generated modules over principal ideal domains (discrete valuation rings at closed points, the function field at the generic point), and use the structure of modules over a principal ideal domain
1.1F1

Local freeness implies torsion-freeness. Suppose M is finite locally free and let U, m≠0, a≠0 satisfy am=0. Since m≠0, choose x∈U with mx≠0. By [F1], the nonzero section a maps to a nonzero element of k(C), and its germ ax maps to that same element, so ax≠0. Shrink to an open neighbourhood of x on which M is free. The relation gives axmx=0 in a free module over the domain OC,x; multiplication by the nonzero scalar ax is injective coordinatewise, forcing mx=0, a contradiction. So a finite locally free module has no such a,m.

1.2F1F2F3

Torsion-freeness implies free stalks. Suppose M has no relation am=0 as in the statement, and let x∈C. The stalk Mx is a finitely generated module over the principal ideal domain OC,x (a discrete valuation ring when x is closed, the field k(C) when x is the generic point) [F1], and it is torsion-free: if axmx=0 in Mx with ax≠0 and mx≠0, then by [F3] the relation is represented over an open U by nonzero sections a,m with am=0, contradicting the hypothesis. By [F2] the stalk Mx is free, say of rank rx≥0.

1.3F1F5F7

Part (2). Suppose am=0 with U open, m∈M(U) nonzero and a∈OC(U) nonzero. Choose x∈U with mx≠0, and then an affine open neighbourhood W=Spec⁡A with x∈W⊆U. The restriction a∣W is nonzero by [F1]. Let Z={y∈W:ay∈my}, the vanishing locus of the residue of a∣W; it is the proper closed subset V(a∣W) of W, proper because a∣W has nonzero image in the function field [F1]. Since W is an integral curve, [F7] says that Z is a finite set of closed points of W. None is the generic point of C, and [F7] also says every such point is closed in C, so Z is a finite closed subset of C. Put V=C∖Z. Then V is open and W∪V=C. On W∩V=W∖Z, the residue of a is nonzero at every point, hence a is a unit in every stalk and m∣W∩V=0. The sections m∣W and 0 over V agree on the intersection, so by [F5] they glue to a global section of M, which is nonzero because mx≠0.

2.1F4step 1.1step 1.2

Conclusion of (1). If M is torsion-free then by step 1.2 every stalk is free; by [F4] each point has an open neighbourhood on which M is free of the rank of its stalk, so M is finite locally free. With step 1.1 this proves the equivalence (1).

3.1step 1.1step 2.1

Part (3). A subsheaf N⊆E of a locally free module E is torsion-free: a relation am=0 with m≠0 and a≠0 in N is also a relation in E, which is impossible by step 1.1; and a coherent torsion-free module is finite locally free by step 2.1.

4.1F6step 1.1step 2.1step 1.3step 3.1∎

Conclusion. Step 2.1 proves (1), step 1.3 proves (2) and step 3.1 proves (3). The points chosen in steps 1.1 and 1.3 exist because the corresponding sections are nonzero; the Axiom of Choice enters only through the suppliers recorded in [F6].

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

Nonzero maps from an invertible sheaf to a locally free sheaf are injective

Statement

Let X be an integral scheme (Integral schemes), let L be an invertible OX-module (Invertible sheaves) and let M be a locally free OX-module of finite rank (Locally free sheaves of finite rank). Then every nonzero morphism of OX-modules φ:L→M (Modules on a ringed space) is injective (Kernel sheaves are objectwise, while cokernels and images are sheafified).

Facts & Assumptions

Given: An integral scheme X, an invertible OX-module L, a locally free OX-module M of finite rank, and a morphism φ:L→M with φ≠0.

[F1]

X is nonempty, reduced and irreducible, and every nonempty affine open subset of X is the spectrum of a domain (Integral schemes).

[F2]

A topological space is irreducible if and only if it is nonempty and every two of its nonempty open subsets have nonempty intersection (Irreducible topological spaces and irreducible subsets in the subspace topology, Irreducibility via nonempty open subsets, connectedness and open subspaces).

[F3]

L is locally free of rank 1, so every point of X has an open neighbourhood on which L is isomorphic to the structure sheaf (Invertible sheaves), and M is locally free of finite rank, so every point of X has an open neighbourhood on which M is isomorphic to OU r for some r≥0 (Locally free sheaves of finite rank).

[F4]

For f∈A the sections of OSpec⁡A on the distinguished open D(f) are Af and restriction is the canonical localisation map A→Af (Sections and restrictions on distinguished opens of an affine scheme); in a localisation r1=0 holds exactly when ur=0 for some u in the multiplicative set, and the localisation map of a commutative ring R is injective exactly when no element of the multiplicative set annihilates a nonzero element (Equality, vanishing, and the kernel of the localisation map).

[F5]

If U=Spec⁡A is affine and W⊆U is open with p∈W, then there is h∈A with p∈D(h)⊆W (Every point of a Zariski-open set has a distinguished-open neighbourhood inside it), and the morphism induced by A→Ah identifies Spec⁡(Ah) with the open subscheme D(h) of U, so D(h) is affine with ring Ah (A principal localization identifies its spectrum with a distinguished open).

[F6]

A sequence of sheaves of abelian groups is exact if and only if every stalk sequence is exact (A sequence of abelian sheaves is exact exactly when it is exact on every stalk), and the kernel sheaf of a morphism is computed objectwise (Kernel sheaves are objectwise, while cokernels and images are sheafified).

[F7]

Two morphisms of sheaves with equal stalk maps are equal (Morphisms of sheaves are determined by their maps on stalks), so a morphism is zero if and only if all of its stalk maps are zero, and a morphism that vanishes on every member of an open cover is zero (The stalk of a presheaf at a point).

[F8]

The Axiom of Choice is not used: the arguments below select a chart through each individual point and use only the localisation criteria of [F4]; no family of choices over an infinite index set is made.

Proof

technique · direct; reduce to affine charts, propagate vanishing of the morphism from one chart to all charts by irreducibility, then test injectivity on stalks
1.1F1F2F3F5

Setup. By [F1] the scheme X is nonempty and irreducible, so by [F2] any two nonempty open subsets of X meet; by [F3] the open sets on which L is trivial and the open sets on which M is free form two open covers of X, so for a given x∈X we may choose an affine open U0∋x, intersect it with such a trivialising and such a freeing open set, and apply [F5] to obtain a distinguished open D(h)⊆U0 containing x on which both L and M are free; D(h) is affine with ring a domain by [F1]. The resulting adapted charts U=Spec⁡A, with A a domain, L∣U≅OU and M∣U≅OU r, cover X.

1.2F4

Chart dictionary. Fix an adapted chart U=Spec⁡A and trivialisations L∣U≅OU, M∣U≅OU r; these identify Hom⁡OU(L∣U,M∣U) with OU(U) r=A r, and φ∣U corresponds to an element m=(a1,…,ar)∈A r acting by multiplication; thus φ∣U=0 if and only if m=0. If m≠0, choose i with ai≠0 and let D(f)⊆U be distinguished; A is a domain and A→Af is injective when f≠0 by [F4], so the image of ai in the domain Af is nonzero, and a section s∈OU(D(f))=Af with s⋅ai=0 must be s=0; as the distinguished opens cover every open subset of U, multiplication by m is injective, that is, φ∣U is injective.

2.1F2F4F5step 1.2

Vanishing propagates. Suppose φ∣U0=0 for one adapted chart U0; let V=Spec⁡B be any adapted chart and let mV∈B s correspond to φ∣V as in step 1.2. Since V∩U0 is nonempty by [F2], [F5] supplies a distinguished open D(h)⊆V∩U0; there φ∣D(h)=0, and restriction of the element mV to D(h) is its image in (B s)h by [F4], so that image is zero; the zero criterion of [F4] gives hkmV=0 in B s for some k≥0, and since B is a domain and D(h)≠∅ forces h≠0, this yields mV=0 and hence φ∣V=0 by step 1.2. The adapted charts cover X, so φ=0 by [F7].

3.1step 1.2step 2.1

Stalks are injective. Now suppose φ≠0. By the contrapositive of step 2.1, φ∣U≠0 for every adapted chart U; by step 1.2 the corresponding element m is nonzero and φ∣U is injective. Every point x∈X lies in an adapted chart U, and the stalk map φx is the stalk of φ∣U at x, hence is injective.

4.1F6F8step 3.1∎

Conclusion. The kernel sheaf ker⁡φ is the sheaf of abelian groups with (ker⁡φ)x=ker⁡(φx) by [F6]; all these stalks are zero by step 3.1, so every section of ker⁡φ over every open set is zero and φ is injective; the argument made no use of the Axiom of Choice beyond the fixed data recorded in [F8].

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

A vector bundle on the projective line has a line subbundle of maximal degree

Statement

Assume the Axiom of Choice as inherited from the cohomology, local-DVR, and divisor-degree suppliers. Let k be a field, put X=Pk1, and let E be a nonzero finite locally free OX-module of rank r≥1 (Locally free sheaves of finite rank). Then the set of integers n with H0(X,E(n)) nonzero is nonempty and bounded below, so it has a minimum a; putting b:=−a, one has H0(X,E(−b)) nonzero, H0(X,E(−b−1))=0, and every nonzero morphism OX(b)→E is injective, so its image is a line subbundle of E of degree b. Moreover no line subbundle of E has degree greater than b: E contains a line subbundle of maximal degree b.

Facts & Assumptions

Given: a field k, the scheme X=Pk1, and a nonzero finite locally free OX-module E of rank r≥1.

[F1]

X=Pk1≅Proj⁡k[x0,x1] is projective over Spec⁡k in the H-projective convention, by the identity embedding X↪Pk1 (Projective space is Proj of a polynomial ring, Projective morphisms before Proj, Twisting sheaf on Proj).

[F2]

The twisting sheaf OX(1) is invertible (Invertible twists for degree-one generated rings, Twisting sheaf on Proj). The identity presentation of [F1] exhibits OX(1) as relatively very ample over Spec⁡k, the sections x0,x1 giving the identity morphism X→Pk1 (Relative very ampleness in the finite projective-space convention, Relative projective space from standard charts); hence OX(1) is ample (Relative very ampleness implies relative ampleness, Absolute ampleness by affine section opens).

[F3]

For a coherent OX-module F there is m0 with F⊗OX(m) globally generated for every m≥m0 (Eventual generation of coherent projective twists, Global generation by the evaluation map). A globally generated nonzero module has a nonzero global section, because global generation says that the images of the global sections generate every stalk, and F≠0 has a nonzero stalk (Global generation by the evaluation map).

[F4]

E is coherent and H0(X,E) is a finite-dimensional k-vector space: E is locally free hence quasi-coherent of finite type, X is finite type over the field k, hence locally Noetherian, and coherent modules on a locally Noetherian scheme form an abelian category (Coherent module sheaves, Locally Noetherian and Noetherian schemes, Coherent sheaves on a locally Noetherian scheme); finiteness of H0 is the proper finiteness statement (Finite-dimensional coherent cohomology over a field).

[F5]

H0(X,F)≅Γ(X,F) (Degree-zero sheaf cohomology is global sections), a global section s of a module F gives the morphism s♯:OX→F, a↦a⋅s∣U, and conversely φ↦φX(1); these are inverse, so Γ(X,F)≅Hom⁡OX(OX,F) (Modules on a ringed space). Global sections form a left exact functor: an injective morphism F→G induces an injective map H0(X,F)→H0(X,G) (Global sections are left exact but need not preserve epimorphisms).

[F6]

For every d∈Z, dim⁡kH0(X,OX(d))=d+1 for d≥0 and =0 for d<0 (Global sections of projective twists).

[F7]

Every nonzero morphism from an invertible sheaf to the finite locally free module E is injective (Nonzero maps from an invertible sheaf to a locally free sheaf are injective, Invertible sheaves).

[F8]

Twists are defined by F(m)=F⊗OX(m), with F(m)⊗OX(n)≅F(m+n) and (F(m))(n)≅F(m+n), so twisting by OX(m) is functorial and carries nonzero morphisms to nonzero morphisms (Twists of a quasi-coherent sheaf, Invertible twists for degree-one generated rings).

[F9]

The Axiom of Choice is inherited through the cohomology and global-generation suppliers [F3], [F4] and through the smooth-curve DVR and divisor-degree/Picard suppliers in [F10]; no additional choice is used in the local extension or basis argument (The Axiom of Choice).

[F10]

For every closed point p of the smooth proper curve X, the local ring OX,p is a discrete valuation ring with a uniformizer π (Local rings at closed points of smooth curves are discrete valuation rings). The closed point divisor [p] is effective Cartier; near p its equation can be taken to be π, and away from p its equation is 1. Its associated invertible sheaf OX(p) is locally π−1OX near p and is OX off p (Cartier divisor, Effective cartier divisor, Invertible sheaf of cartier divisor). The degree homomorphism sends [OX(p)] to d=[κ(p):k]>0, and every invertible sheaf of degree j on X is isomorphic to OX(j) (The degree of a divisor descends to the Picard group of a normal proper curve, The Picard group of the projective line). In particular OX(b)≅OX(b[∞]), the Cartier tensor/addition supplier identifies OX(b[∞])⊗OX(p) with OX(b[∞]+[p]), and that line has degree b+d; Picard classification identifies it with OX(b+d) (Addition of Cartier divisors is tensor product of their sheaves, Divisors on the projective line are classified by degree, The Picard group of the projective line).

Proof

technique · direct; produce sections in high degree by global generation, bound the degrees carrying sections below by comparison with $H^0(E)$, and take the minimum
1.1F1F2F3F4

Nonemptiness of the section degrees. By [F4] the module E is coherent, so [F3] applies with the ample invertible sheaf OX(1) of [F2] and provides m0 with E(m) globally generated for all m≥m0. Since E≠0 and X is nonempty, some stalk of E(m) is nonzero, and global generation exhibits a global section with nonzero germ; hence H0(X,E(m))≠0 for every m≥m0.

2.1F4F5F6F7F8

Boundedness below. Suppose H0(X,E(n))≠0 and let s≠0 be a global section. By [F5] the section s is a nonzero morphism s♯:OX→E(n); twisting by OX(−n) yields a nonzero morphism OX(−n)→E by [F8], which is injective by [F7]. The left exact functor H0 of [F5] therefore gives an injection H0(X,OX(−n))↪H0(X,E), so dim⁡kH0(X,OX(−n))≤h0(E) with h0(E):=dim⁡kH0(X,E)<∞ by [F4]. If n≤0, then by [F6] the left side is −n+1, so −n+1≤h0(E), that is n≥1−h0(E). If n≥1, then n≥1; and since h0(E)≥0 we get n≥1−h0(E) in this case too. Hence every n with H0(X,E(n))≠0 satisfies n≥B for the integer B:=1−h0(E), and the set of such n is nonempty by step 1.1 and bounded below.

3.1step 2.1

The extremal degree. The set S={n∈Z:H0(X,E(n))≠0} is a nonempty subset of Z bounded below, so it has a minimum a. Put b:=−a. Then H0(X,E(−b))=H0(X,E(a))≠0, and H0(X,E(−b−1))=H0(X,E(a−1))=0, since a−1∉S by minimality.

4.1F7F10step 3.1

The maximal map is a subbundle. Fix any nonzero morphism φ:OX(b)→E. It is injective by [F7]. Let p be a closed point and choose local frames for OX(b) and E near p; write the coefficient vector of φ in these frames as (q1,…,qr). Suppose every qi lies in the maximal ideal of OX,p. By [F10], this local ring is a DVR with uniformizer π, so every qi/π is regular at p. After shrinking a neighborhood U of p, these quotients are regular sections and OX(p)∣U=π−1OU. Define a morphism OX(b)⊗OX(p)∣U→E∣U by sending the frame e⊗π−1 to φ(e)/π. On X∖{p} use the identification OX(p)=OX and the original φ. On the overlap U∖{p}, π is a unit and the maps agree, so they glue to a global morphism OX(b)⊗OX(p)→E. It is nonzero because its restriction at the generic point agrees with φ. By [F10] its source is isomorphic to OX(b+d) for d=[κ(p):k]>0. Thus H0(X,E(−b−d))≠0, contradicting minimality of a=−b. Therefore at least one qi is a unit at every closed point p. On a neighborhood where that coefficient remains a unit, elementary row operations make the image a direct summand of E, so the quotient is locally free there. At the generic point the nonzero map is an inclusion of a one-dimensional subspace into a vector space and is likewise a direct summand after restricting to a neighborhood. Hence φ has locally free quotient and its image is a line subbundle of degree b. Since φ was arbitrary, every nonzero map OX(b)→E has this property.

4.2F5F8F10step 3.1

Maximality. Let L⊆E be any line subbundle of degree d. By the Picard classification in [F10], L≅OX(d). Its inclusion is nonzero, so after twisting by OX(−d) it gives a nonzero morphism OX→E(−d) by [F8], hence a nonzero global section of E(−d) by [F5]. Thus −d∈S. By step 3.1, −d≥a=−b, so d≤b: no line subbundle has degree greater than b.

5.1F9step 1.1step 2.1step 3.1step 4.1step 4.2∎

Conclusion. Steps 1.1 and 2.1 show that S is nonempty and bounded below, step 3.1 produces its minimum a with b=−a and the two vanishing statements, step 4.1 proves that every nonzero maximal-degree map has locally free quotient, and step 4.2 proves maximality among line subbundles. Choice is inherited only from the suppliers recorded in [F9].

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

The quotient by a maximal line subbundle is locally free

Statement

Assume the Axiom of Choice as inherited from the cohomology and curve suppliers. Let E be a finite locally free OPk1-module of rank r≥2 and let φ:OPk1(b)→E be a nonzero morphism with b maximal as in A vector bundle on the projective line has a line subbundle of maximal degree, so that H0(Pk1,E(−b−1))=0. Let F:=E/OPk1(b) be the cokernel of φ. Then F is a finite locally free OPk1-module of rank r−1.

Facts & Assumptions

Given: a field k, a finite locally free module E of rank r≥2 on X=Pk1, and a nonzero maximal-degree morphism φ:OX(b)→E.

[F1]

φ is injective, H0(X,E(−b))≠0 and H0(X,E(−b−1))=0 (A vector bundle on the projective line has a line subbundle of maximal degree).

[F2]

Twisting is F(m)=F⊗OX(m), with canonical isomorphisms F(m)⊗OX(n)≅F(m+n); each OX(m) is invertible with OX(m)⊗OX(−m)≅OX, so twisting by OX(m) is an equivalence of categories with inverse twisting by OX(−m) and preserves exact sequences of OX-modules; if L is invertible and L∣U is free on a frame g, then s↦s⊗g is an isomorphism F∣U→(F⊗L)∣U (Twists of a quasi-coherent sheaf, Invertible sheaves, Dual of a line bundle is its tensor inverse, Tensor product of sheaves of modules, A sequence of abelian sheaves is exact exactly when it is exact on every stalk, Exact sequences of sheaves).

[F3]

H0(X,OX(−1))=0 and H1(X,OX(−1))=0 (Global sections of projective twists, Top cohomology of projective twists), and a short exact sequence of modules gives a long exact sequence of cohomology (Long exact sequence of sheaf cohomology).

[F4]

X is locally Noetherian, being covered by the spectra of the Noetherian rings k[t] and k[u] (Locally Noetherian and Noetherian schemes, Every algebra of finite type over a Noetherian ring is a Noetherian ring, A field has only the zero ideal and itself, hence is Noetherian, Two-affine projective line and its twists). On a locally Noetherian scheme finite locally free modules and invertible modules are coherent; cokernels of morphisms of coherent modules are coherent; and every twist of a coherent module is coherent, because coherence is local and on an open set on which OX(1) is trivial the twist is isomorphic to the original module (Coherent module sheaves, Coherent sheaves on a locally Noetherian scheme, Hilbert function and Euler characteristic on a projective scheme, Finite type and finitely presented module sheaves).

[F5]

X is covered by the two standard charts U0=Spec⁡k[t] and U1=Spec⁡k[u] with u=t−1, and on each chart every twisting sheaf OX(d) is free on a frame, in particular OX(−1) has a nowhere-vanishing frame on each chart (Two-affine projective line and its twists); the polynomial rings k[t], k[u] are principal ideal domains (For every field F, F[x] is a principal ideal domain); a finitely generated torsion-free module over a principal ideal domain is free (Every finitely generated torsion-free module over a PID is free).

[F6]

At each point x, the stalk sequence of a short exact sequence of finite locally free modules is exact. Once F is known to be locally free, the surjection Ex→Fx splits because Fx is a free module over the local ring OX,x and a basis can be lifted; hence the stalk ranks add (A sequence of abelian sheaves is exact exactly when it is exact on every stalk, Locally free sheaves of finite rank, The direct sum of an indexed family of modules).

[F7]

For every quasi-coherent F and every affine open U=Spec⁡A of X, the canonical comparison F∣U≅Γ(U,F)~ is an isomorphism, compatibly with restrictions to smaller affine opens; on an associated sheaf M~ one has M~(D(f))=Mf with restriction the localisation map; the basic opens D(f) form a basis of the topology of Spec⁡A; and an element of Mf is zero exactly when some power of f kills a numerator (Checking quasi-coherence on an affine cover, Sections of the associated sheaf on basic opens, The underlying space of an affine spectrum, A localised module fraction is zero exactly when one denominator kills its numerator).

[F8]

Affine schemes are quasi-compact, so every open cover of Spec⁡A has a finite subcover (Every affine scheme is quasi-compact, Quasi-compact and quasi-separated schemes).

[F9]

If z∈F(U) and a∈OX(U) satisfy a⋅z=0 and W⊆U is an open set on which a is a unit of OX(W), then z∣W=0; on the nonvanishing locus of a the section a is invertible; and sections of a sheaf of modules over U and over W that agree on U∩W glue to a unique section over U∪W (A sheaf on a topological space, Modules on a ringed space, A locally ringed space, A line-bundle section cuts an affine open inside an affine scheme).

[F10]

The Axiom of Choice is inherited from the twisting supplier [F2], the cohomology suppliers [F1], [F3], the coherence supplier [F4], the affine correspondence [F7], and the curve supplier [F11] (The Axiom of Choice).

[F11]

X=Pk1 is an integral finite-type curve. For every nonempty open V⊆X, restriction embeds Γ(V,OX) into the function field k(t); thus a nonzero regular section remains nonzero on every nonempty open and has nonzero germ at every point there. The residue-zero locus of a nonzero regular function on a nonempty affine open U⊆X is a proper closed subset, hence is finite, and each of its points is closed in X (Two-affine projective line and its twists, Integral schemes, Proper closed subsets of a curve are finite).

Proof

technique · direct; compute the cohomology of the quotient twisted by $\mathcal O_X(-b-1)$ to force torsion-freeness, then read local freeness off the two affine charts
1.1F1F2

The twisted extension. By [F1] the morphism φ is injective, so 0→OX(b)→E→F→0 is exact with F=E/OX(b). Twisting by OX(−b) and using OX(b)⊗OX(−b)≅OX gives the exact sequence 0→OX→M→F(−b)→0 with M:=E(−b); by [F1] H0(X,M)≠0, while H0(X,M(−1))=H0(X,E(−b−1))=0 because M(−1)≅E(−b−1).

2.1F1F2F3step 1.1

The quotient has no sections in the next twist. Twisting the sequence of step 1.1 by OX(−1) gives the exact sequence 0→OX(−1)→M(−1)→F(−b−1)→0. Its long exact sequence [F3] reads 0=H0(X,M(−1))→H0(X,F(−b−1))→H1(X,OX(−1))=0, because H0(X,OX(−1))=0 as well; hence H0(X,F(−b−1))=0.

3.1F2F5F7F9F11step 2.1

F(−b) is torsion-free. Suppose there are an open U0⊆X, nonzero m∈F(−b)(U0) and nonzero a∈OX(U0) with a m=0. Choose a point x∈U0 with mx≠0; this only uses that m is a nonzero section, and the nonzero-germ locus is not asserted to be open. Choose one of the two standard charts T containing x, then a principal affine open U⊆U0∩T containing x. The restricted sections remain nonzero, and OX(−1)∣U has a nowhere-vanishing frame g by [F5, F11]. The assignment s↦s⊗g is an isomorphism F(−b)∣U→F(−b−1)∣U, so m′:=m∣U⊗g∈F(−b−1)(U) is nonzero and a∣U m′=0. Let Z={y∈U:ay∈my}=V(a∣U), the residue-zero locus. It is a proper closed subset of the integral curve U because a∣U is nonzero [F11]; by [F11] it is finite and each point of Z is closed in X. Thus W:=X∖Z is open and U∪W=X. On U∩W=U∖Z, the residue of a is nonzero at each point, so a is a unit in every stalk and m′∣U∩W=0 by [F9]. The sections m′ over U and 0 over W agree on the overlap and glue to a nonzero global section of F(−b−1), contradicting step 2.1. Hence F(−b) is torsion-free.

4.1F2F4F5F7F8step 3.1

Local freeness. Fix a standard chart U=Spec⁡A, A=k[t] or k[u], and put M:=F(−b)(U). By [F4], F(−b) is coherent, hence quasi-coherent, so [F7] applies and F(−b)∣U≅M~, so F(−b)(D(f))≅Mf for every f∈A. By [F4] F(−b) is coherent, hence of finite type, so every point x∈U has an affine open Vx⊆X with F(−b)∣Vx≅Nx~ for a finitely generated module Nx; choosing fx∈A with x∈D(fx)⊆Vx∩U [F7], putting Bx:=Γ(Vx,OX), the affine restriction isomorphism in [F7] identifies Mfx with Afx⊗BxNx, a finitely generated Afx-module whose generators are the images of any finite generating set of Nx. Since U is affine, hence quasi-compact [F8], finitely many such D(f1),…,D(fk) cover U, so f1,…,fk generate the unit ideal of A and each Mfi is finitely generated. Choose finitely many elements of M whose images generate each Mfi and let N⊆M be the submodule they generate. For every z∈M/N, the vanishing (M/N)fi=0 and [F7] give, for each i, a power fini(z) that annihilates this element z; the powers may depend on z. Since (f1,…,fk)=A, also (f1n1(z),…,fknk(z))=A (their generated ideal has the same radical as the unit ideal). Thus z=0, and M/N=0: M is a finitely generated A-module. By step 3.1 the module M is torsion-free, and A=k[t] (respectively k[u]) is a principal ideal domain [F5], so M is free [F5]. As the two charts cover X, F(−b) is finite locally free; twisting back by OX(b), an equivalence by [F2], F is finite locally free as well.

5.1F6step 1.1step 4.1

The rank. At each point x∈X, the stalk sequence from step 1.1 is exact and all three stalks are free after step 4.1. The surjection Ex→Fx splits because Fx is free, so the ranks add: r=1+rk⁡Fx. Hence F is locally free of rank r−1 everywhere.

6.1F10step 4.1step 5.1∎

Conclusion. The module F is finite locally free of rank r−1 by steps 4.1 and 5.1. The Axiom of Choice enters only through the suppliers recorded in [F10].

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Extensions of line bundles on the projective line split after ordering

Statement

Assume the Axiom of Choice as inherited from the sheaf-cohomology suppliers. Let k be a field and put X=Pk1. Let 0→OX→M→W→0 be a short exact sequence of finite locally free OX-modules (Locally free sheaves of finite rank) in which W is isomorphic to a finite direct sum of twisting sheaves OX(ni) (Twisting sheaf on Proj) with ni≤0 for every i. Then M is isomorphic to OX directly summed with W, M≅OX⊕W. More generally, if 0→OX(b)→E→F→0 is a short exact sequence of finite locally free sheaves with F isomorphic to a direct sum of line bundles OX(bi), bi≤b, then E≅OX(b)⊕F.

Facts & Assumptions

Given: a field k, the scheme X=Pk1, and the two short exact sequences of the statement.

[F1]

X=Pk1≅Proj⁡k[x0,x1]. The twisting sheaves satisfy OX(0)=OX, each OX(d) is invertible, and the multiplication maps OX(m)⊗OX(n)→OX(m+n) are isomorphisms (Projective space is Proj of a polynomial ring, Twisting sheaf on Proj, Invertible twists for degree-one generated rings). The twist of an OX-module is F(d)=F⊗OX(d) with F(0)≅F and F(m)⊗OX(n)≅F(m+n) (Twists of a quasi-coherent sheaf); in particular (F(−c))(c)≅F for every integer c, and twisting is functorial, so it carries isomorphisms to isomorphisms (Tensor product of sheaves of modules).

[F2]

For every d∈Z one has H0(X,OX(d))=0 for d<0 and H0(X,OX(d))≅k d+1 for d≥0 (Global sections of projective twists), and H1(X,OX(d))=0 for d≥−1 (Top cohomology of projective twists). In particular H0(X,OX)≅k with unit section 1.

[F3]

A short exact sequence of OX-modules induces a long exact sequence of cohomology groups (Long exact sequence of sheaf cohomology), and H0(X,F)≅Γ(X,F) is the module of global sections of F (Degree-zero sheaf cohomology is global sections).

[F4]

(Sections as morphisms.) For an OX-module F every global section s∈Γ(X,F) determines a morphism of OX-modules s♯:OX→F by sU♯(a)=a⋅s∣U for opens U⊆X and a∈OX(U), and conversely φ↦φX(1) inverts this assignment; the maps are mutual inverses, so Γ(X,F)≅Hom⁡OX(OX,F), and for a morphism ψ:F→G one has (ψ∘s♯)X(1)=ψX(s). The construction uses only that F(U) is an OX(U)-module with restriction maps linear over the ring maps, and that the unit section 1 generates OX as an OX-module (Modules on a ringed space, Sections, restrictions, and global sections of a presheaf).

[F5]

A sequence of sheaves of modules is exact if and only if all its stalk sequences are exact; stalk formation preserves kernels, commutes with the tensor product of OX-modules and turns an invertible factor into a free module of rank one over the local ring, so tensoring with an invertible sheaf preserves exactness (A sequence of abelian sheaves is exact exactly when it is exact on every stalk, Kernel sheaves are objectwise, while cokernels and images are sheafified, The stalk of a tensor product sheaf is the tensor product of the stalks, Exact sequences of sheaves). For a finite family the direct sum of modules is a coproduct with the coordinate injections and a product with the coordinate projections, and 0→F1→F1⊕F2→F2→0 is exact (The direct sum of an indexed family of modules).

[F6]

The Axiom of Choice is inherited from the Proj and twisting-sheaf suppliers [F1] and the cohomology suppliers [F2] and [F3]; the selection made below is the selection of one lift, and the induction makes finitely many such selections (The Axiom of Choice).

Proof

technique · direct; prove by induction on the number of line-bundle summands of the quotient that an extension of a direct sum of line bundles of degrees at most $c$ by $\mathcal O_X(c)$ splits, splitting off a summand of minimal degree
1.1F1F5

The induction statement. We prove, for every r≥0, the assertion P(r): for every c∈Z and every short exact sequence 0→OX(c)→iE→pF→0 of finite locally free OX-modules with F≅⨁i=1rOX(ci) and ci≤c for all i, one has E≅OX(c)⊕F. Since X≅Proj⁡k[x0,x1] and OX(1) generates the twisting sheaves with OX(m)⊗OX(n)≅OX(m+n) by [F1], the reindexing of the ci and the canonical identifications of direct sums do not change the conclusion, and P(r) for all r and all c gives both assertions of the Statement: the first with c=0.

1.2F5

Base case. For r=0 one has F=0, so p=0 and i is an isomorphism; hence E≅OX(c)≅OX(c)⊕F.

1.3F2F3F5

Vanishing of the relevant H1. Assume r≥1 and that P(r−1) holds. Choose an enumeration with cr=min⁡ici, put F′=⨁i<rOX(ci), so that F≅F′⊕OX(cr), and let q:F→OX(cr) be the projection. Twist the given sequence by OX(−cr): by [F1] and [F5] the result is the short exact sequence 0→OX(c−cr)→E(−cr)→F(−cr)→0, with F(−cr)≅F′(−cr)⊕OX. Since ci−cr≥0 for all i<r, each summand OX(ci−cr) has vanishing H1 by [F2], and H1 vanishes on the finite direct sum F′(−cr) by induction on the number of summands: for a summand split injection 0→G→G⊕OX(cj−cr)→OX(cj−cr)→0 of [F5] the long exact sequence [F3] yields the exact portion H1(X,G)→H1(X,G⊕OX(cj−cr))→H1(X,OX(cj−cr)) with both outer groups zero. Also c−cr≥0≥−1, so H1(X,OX(c−cr))=0 by [F2].

2.1F2F3F4

A lift of the unit section. The long exact sequence [F3] of the twisted sequence begins H0(X,E(−cr))→H0(X,F(−cr))→H1(X,OX(c−cr))=0, so the first map is surjective. The projection q twisted by OX(−cr) is a surjection F(−cr)→OX whose kernel is F′(−cr); by step 1.3 and the long exact sequence of 0→F′(−cr)→F(−cr)→OX→0, the induced map on H0 is surjective. Composing the two surjections there is s∈H0(X,E(−cr)) with image the unit section 1∈H0(X,OX)≅k. By [F4] the section s corresponds to a morphism σ=s♯:OX→E(−cr) with p′∘σ=id, where p′:E(−cr)→OX is the composite q∘p twisted.

3.1F5step 2.1

Splitting off the minimal summand. Let K=ker⁡p′⊆E(−cr). The morphism Ψ:K⊕OX→E(−cr) defined on the summands by the inclusion of K and by σ is an isomorphism. Indeed, on stalks at a point x the argument is the elementary module argument: if k+σ(l)=0 with k∈Kx, l∈OX,x, then applying px′ gives l=0 and then k=0, so Ψx is injective; and for m∈E(−cr)x one has m−σ(px′(m))∈Kx, so Ψx is surjective. By [F5] a morphism of sheaves that is stalkwise bijective is an isomorphism. Hence E(−cr)≅K⊕OX.

4.1F5step 1.3step 3.1

The complement is again an extension of the same shape. Let r′:F(−cr)→F′(−cr) be the projection of step 1.3. The sequence 0→OX(c−cr)→K→F′(−cr)→0 is exact, where the first map is the restriction of the inclusion of OX(c−cr) and the second is the restriction of r′∘p to K. Stalks at x: the sequence 0→OX(c−cr)x→E(−cr)x→F(−cr)x→0 is exact and F(−cr)x=F′(−cr)x⊕OX,x; an element of F′(−cr)x lifted to E(−cr)x can be corrected by an element with the same OX,x-component to lie in Kx, because E(−cr)x→OX,x is surjective, so Kx→F′(−cr)x is surjective; and its kernel is the kernel of E(−cr)x→F(−cr)x, namely OX(c−cr)x, the inclusion being injective. Exactness of the displayed sequence follows from [F5]. Also K is finite locally free: near any point, trivialize the kernel line bundle and the finite locally free quotient F′(−cr), lift the finitely many quotient basis germs to sections of K, and shrink so that their images equal the basis sections. These lifts define a local splitting. Together with a frame of the kernel they identify K locally with a finite free sheaf, as required for P(r−1).

5.1F1step 1.3step 3.1step 4.1

Induction step. In the exact sequence of step 4.1 the quotient F′(−cr)≅⨁i<rOX(ci−cr) has r−1 summands. Since cr=min⁡ici and ci≤c, their exponents satisfy 0≤ci−cr≤c−cr. Thus P(r−1), applied with kernel exponent c−cr, gives K≅OX(c−cr)⊕F′(−cr). Combining with step 3.1 and twisting back by OX(cr), which preserves direct sums and isomorphisms by [F1] and inverts (−cr), E≅E(−cr)(cr)≅K(cr)⊕OX(cr)≅OX(c)⊕F′⊕OX(cr)≅OX(c)⊕F. This proves P(r).

6.1F6step 1.1step 1.2step 2.1step 5.1∎

Conclusion. By steps 1.2 and 5.1 the assertion P(r) holds for every r≥0 and every c∈Z; taking c=0 gives M≅OX⊕W for the first sequence of the Statement, and the general assignment c=b gives E≅OX(b)⊕F whenever all bi≤b. Every selection made was the choice of one lift of a specified element in step 2.1 and finitely many such selections occur, so the Axiom of Choice enters through the Proj, twisting-sheaf and cohomology suppliers recorded in [F6].

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Birkhoff-Grothendieck: vector bundles on the projective line split

Statement

Assume the Axiom of Choice as inherited from the cohomology and splitting suppliers. Let E be a finite locally free OPk1-module of rank r≥1 on the projective line over a field k. Then E is isomorphic to a direct sum of line bundles, E≅O(a1)⊕⋯⊕O(ar) with integers a1≤⋯≤ar. The multiset {a1,…,ar} is determined by E; equivalently, the function m↦h0(Pk1,E(m)) determines it, and the decomposition is unique up to permutation. In the language of geometric vector bundles, every vector bundle on Pk1 is a direct sum of line bundles of uniquely determined degrees.

Facts & Assumptions

Given: a field k, the projective line X=Pk1, and a finite locally free OX-module E of rank r≥1.

[F1]

A finite locally free OX-module of rank r is an OX-module locally isomorphic to OX⊕r; such modules are the sheaves of sections of geometric vector bundles, and the two descriptions determine each other, so a splitting statement for finite locally free modules is a splitting statement for vector bundles (Locally free sheaves of finite rank, Finite locally free sheaves and geometric vector bundles).

[F2]

Every invertible sheaf on X is isomorphic to OX(d) for a unique integer d, and OX(d)≅OX(e) if and only if d=e; equivalently Pic⁡(X)≅Z with generator [OX(1)] (The Picard group of the projective line).

[F3]

h0(X,OX(d))=dim⁡kH0(X,OX(d))=d+1 for d≥0 and =0 for d<0 (Global sections of projective twists, Twisting sheaf on Proj), and H1(X,OX(d))=0 for every d≥−1 (Top cohomology of projective twists). In particular h0(OX(−1))=H1(X,OX(−1))=0.

[F4]

If E is nonzero then the set of integers n with H0(X,E(n))≠0 is nonempty and bounded below; with b its negative minimum, H0(X,E(−b))≠0, H0(X,E(−b−1))=0, and no line subbundle of E has degree greater than b, while E contains a line subbundle of degree b (A vector bundle on the projective line has a line subbundle of maximal degree).

[F5]

Let M be a finite locally free OX-module of rank r≥2 and let φ:OX→M be a nonzero morphism with H0(X,M(−1))=0 (the case b=0 of the maximality condition). Then the cokernel W=M/OX is finite locally free of rank r−1 (The quotient by a maximal line subbundle is locally free, Locally free sheaves of finite rank).

[F6]

Every nonzero morphism from an invertible sheaf to a finite locally free module is injective; a global section s of a module F is the same thing as the morphism s♯:OX→F, a↦a⋅s∣U (Nonzero maps from an invertible sheaf to a locally free sheaf are injective, Invertible sheaves).

[F7]

Let 0→OX→M→W→0 be a short exact sequence of finite locally free sheaves with W≅⨁iOX(ni) and ni≤0 for every i. Then M≅OX⊕W (Extensions of line bundles on the projective line split after ordering).

[F8]

Twisting is F(m)=F⊗OXOX(m), with F(m)⊗OX(n)≅F(m+n) and (F(m))(n)≅F(m+n); twisting is functorial, carries nonzero morphisms to nonzero morphisms, and preserves exactness because OX(m) is invertible (Twists of a quasi-coherent sheaf, Invertible twists for degree-one generated rings, Twisting sheaf on Proj).

[F9]

H0(X,F)=Γ(X,F) and the functor H0 is left exact; for a short exact sequence of sheaves 0→F→G→H→0 there is a long exact sequence of cohomology ⋯→H0(F)→H0(G)→H0(H)→H1(F)→⋯ (Sheaf cohomology as right derived global sections, Long exact sequence of sheaf cohomology, Global sections are left exact but need not preserve epimorphisms).

[F10]

A finite direct sum of modules is both a coproduct and a product: a section of ⨁iFi is a finite tuple of sections of the Fi, so H0(X,⨁iFi)≅⨁iH0(X,Fi) and dimensions add; and the direct sum of finite locally free sheaves is finite locally free of the summed rank (The direct sum of an indexed family of modules, Locally free sheaves of finite rank).

[F11]

The Axiom of Choice is assumed and is used only through the suppliers named in the facts above; the induction below makes no further infinite selection (The Axiom of Choice).

Proof

technique · induction on the rank; split off a line subbundle of maximal degree and apply the extension-splitting lemma to the quotient, then recover the multiset of degrees from the function $m\mapsto h^0(E(m))$
1.1F2

Base case. If r=1 then E is invertible, so by [F2] there is a unique integer a1 with E≅OX(a1); this is a direct sum of one line bundle, and the multiset {a1} is determined by E.

1.2F4

The maximal line subbundle. Let r≥2 and assume the theorem known for all finite locally free modules of rank r−1. By [F4], applied to the nonzero module E, there is an integer b with H0(X,E(−b))≠0 and H0(X,E(−b−1))=0, no line subbundle of E has degree greater than b, and E contains a line subbundle of degree b.

2.1F5F6F8step 1.2

The normalized extension. Put M:=E(−b), so that H0(X,M)≠0 and H0(X,M(−1))=0 by [F8] and step 1.2. Choose a nonzero global section s of M; by [F6] the corresponding morphism s♯:OX→M is injective, and by [F5] (with b=0, its maximality hypothesis being exactly H0(X,M(−1))=0) its cokernel W:=M/OX is a finite locally free OX-module of rank r−1. Thus 0→OX→M→W→0 is a short exact sequence.

3.1F3F8F9step 1.2step 2.1

The vanishing on the quotient. Twist the sequence of step 2.1 by OX(−1): by [F8] this gives the short exact sequence 0→OX(−1)→M(−1)→W(−1)→0, whose long exact cohomology sequence by [F9] begins 0→H0(OX(−1))→H0(M(−1))→H0(W(−1))→H1(OX(−1)). The two outer terms vanish by [F3] and the middle term vanishes by [F8] and step 1.2; exactness therefore forces H0(X,W(−1))=0.

4.1F3F10step 1.2step 2.1step 3.1

The quotient splits into twists of nonpositive degree. The module W is finite locally free of rank r−1 by step 2.1, so the induction hypothesis of step 1.2 applied to W gives W≅⨁i=1r−1OX(ni) for integers ni. By [F10] and [F3], h0(X,W(−1))=∑i=1r−1h0(X,OX(ni−1)), and each summand equals ni when ni≥1 and 0 when ni≤0. Since h0(X,W(−1))=0 by step 3.1 and all summands are nonnegative, every summand vanishes, so ni≤0 for every i.

5.1F7F8F10step 1.1step 2.1step 4.1

Splitting off a line subbundle. The extension 0→OX→M→W→0 of step 2.1 has W≅⨁iOX(ni) with ni≤0 by step 4.1, so [F7] gives M≅OX⊕W. Twisting by OX(b), which commutes with finite direct sums and satisfies OX(b)⊗OX(n)≅OX(n+b) by [F8], yields E≅M(b)≅OX(b)⊕⨁i=1r−1OX(ni+b), a direct sum of r line bundles; this is the induction step, and with the base case of step 1.1 it proves that every finite locally free OX-module of rank r≥1 is a direct sum of line bundles.

6.1F3F8F10step 5.1

Uniqueness of the multiset. Suppose E≅⨁i=1rOX(ai). Twisting by OX(m) and using [F8], [F10] and [F3] gives h0(X,E(m))=∑i=1rh0(X,OX(ai+m))=∑i=1rmax⁡(ai+m+1,0). Consequently the difference of consecutive values is h0(X,E(m))−h0(X,E(m−1))=#{ i:ai+m+1>0 }=#{ i:ai≥−m }, so for every integer t the function determines the counting number #{i:ai≥t} by evaluation at m=−t. The finitely many counting numbers #{i:ai≥t} determine the multiset {ai}, so the multiset is determined by E, equivalently by the function m↦h0(X,E(m)); in particular the decomposition is unique up to permutation of the summands.

7.1F1F11step 1.1step 5.1step 6.1∎

Conclusion and choice accounting. Steps 1.1 and 5.1 prove the existence of the direct-sum decomposition for every rank r≥1, and step 6.1 proves that the multiset of degrees is determined by E, hence unique up to permutation; the translation to geometric vector bundles is [F1]. The Axiom of Choice is inherited only through the suppliers of the cited facts, as recorded in [F11]; the proof selects a section of a nonzero finite-dimensional space and a finite tuple of integers, and repeatedly reduces the rank by one, so no further infinite selection occurs.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

An effective divisor of degree zero is empty

Statement

Let C be a proper geometrically integral curve over a field k and let D be an effective divisor on C. Then deg⁡k(D)≥0, and deg⁡k(D)=0 if and only if D=0. In particular if D≤D′ are effective divisors with deg⁡k(D)=deg⁡k(D′) then D=D′.

Facts & Assumptions

Given: a field k, a proper geometrically integral curve C over k, and an effective divisor D on C.

[F1]

A curve over k is geometrically integral, separated and of finite type of chain dimension one; a proper curve is a curve whose structure morphism is proper. In particular C is an integral k-scheme of dimension one and the empty scheme is not a curve (Curves over a field).

[F2]

A divisor on C is a finite formal sum D=∑xnx[x] over the closed points x of C with integer coefficients; each residue field κ(x) is a finite extension of k with [κ(x):k]=dim⁡kκ(x)≥1; the degree is deg⁡kD=∑xnx[κ(x):k], and deg⁡k is a group homomorphism Div⁡(C)→Z (Degree divisor proper curve).

[F3]

For a divisor D the support is finite, the positive and negative parts satisfy D=D+−D− with disjoint supports, both parts have nonnegative coefficients, and D is effective exactly when all its coefficients are nonnegative, equivalently exactly when D−=0; for two divisors one writes D≥D′ when D−D′ is effective, so D≤D′ means that D′−D has nonnegative coefficients (Divisor support positive negative parts, Divisors on a smooth proper curve).

[F4]

For an effective divisor on a proper geometrically integral curve, the degree is nonnegative and vanishes exactly for the zero divisor; the argument is the finite sum deg⁡k(D)=∑xnx[κ(x):k] of nonnegative terms with every [κ(x):k]≥1 (Effective divisors have nonnegative degree).

Proof

technique · direct; the nonnegativity and vanishing statement is the cited effective-divisor lemma, and the comparison clause reduces the difference to that lemma
1.1F1F2F3

Unwinding. By [F1] the curve C is an integral k-scheme of dimension one, so the divisor formalism of [F2] applies. By [F2] the divisor D has finite support and is written as a finite sum D=∑xnx[x] with integer coefficients; by [F3] effectiveness of D says that all coefficients nx are nonnegative. The degree is the finite sum deg⁡k(D)=∑xnx[κ(x):k] of [F2].

1.2F2

The residue degrees are positive. For each closed point x of C the residue field κ(x) is a finite extension of k, so [κ(x):k]=dim⁡kκ(x) is a positive integer, at least one.

2.1F2F4step 1.1step 1.2

Nonnegativity and vanishing. Apply [F4] to the proper geometrically integral curve C and the effective divisor D: the degree deg⁡k(D) is a nonnegative integer, and deg⁡k(D)=0 if and only if D=0. In unfolded terms, each summand nx[κ(x):k] is a product of the nonnegative coefficient of step 1.1 and the positive integer of step 1.2, so the sum is nonnegative and can vanish only when every coefficient vanishes.

3.1F2F3step 2.1

The comparison clause. Let D≤D′ be effective divisors with deg⁡k(D)=deg⁡k(D′), and put E:=D′−D. By [F3] the relation D≤D′ says that E has nonnegative coefficients, i.e. E is effective; by the additivity in [F2], deg⁡k(E)=deg⁡k(D′)−deg⁡k(D)=0. Applying step 2.1 to the effective divisor E gives E=0, hence D=D′.

4.1F2step 2.1step 3.1∎

Conclusion. For every effective divisor D on the proper geometrically integral curve C one has deg⁡k(D)≥0 with equality exactly for D=0 by step 2.1, and two effective divisors with D≤D′ and equal degree coincide by step 3.1. No choice principle is used: only coefficients of a finite sum, integer degrees of finite field extensions and the cited degree-additivity are involved.

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

A degree-zero line bundle with a nonzero section is trivial

Statement

Assume the Axiom of Choice as inherited from the degree homomorphism on Pic⁡ and the structure-sheaf cohomology supplier. It supplies the Dependent Choice premise of the Cartier-to-Weil route through AC implies DC implies countable choice. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field) and let L be an invertible sheaf on C (Invertible sheaves) with deg⁡kL=0. If H0(C,L)≠0 then L≅OC; equivalently, an invertible sheaf of degree zero that is not trivial has no nonzero global section, that is H0(C,L)=0 (Sheaf cohomology as right derived global sections).

The degree, Cartier-sheaf and rational-section interfaces used here are The degree of a divisor descends to the Picard group of a normal proper curve, Invertible sheaf of cartier divisor, Rational sections of line bundles are Cartier divisors and Cartier and Weil divisors agree on a smooth curve. The proof verifies that a nonzero global section has nonzero generic germ before applying the rational-section interface; the facts and their exact uses are recorded below.

Facts & Assumptions

Given: the Axiom of Choice inherited from the degree, Cartier and structure-sheaf cohomology suppliers, with Dependent Choice supplied through AC implies DC implies countable choice; a field k, a smooth proper geometrically integral curve C over k, an invertible sheaf L on C with deg⁡kL=0, and the hypothesis H0(C,L)≠0.

[F1]

Cohomology and global sections: H0(C,L)=Γ(C,L) is the group of global sections of the abelian sheaf L, so a nonzero cohomology group gives a nonzero section s∈Γ(C,L); L is invertible, and C is integral (Sheaf cohomology as right derived global sections, Global sections of an abelian sheaf, Invertible sheaves, Integral schemes). Such an s has nonzero generic germ. Indeed, if sη=0, the definition of a stalk gives a nonempty open U on which s vanishes. On any nonempty affine open V=Spec⁡A trivializing L, write s∣V=fe in a frame e. Since C is integral, A is a domain; since C is irreducible, U∩V is a nonempty open and contains a nonempty distinguished open D(a) for some a≠0. The coefficient f becomes zero in Aa. The localization map A→Aa is injective because A is a domain and a≠0, so f=0. Thus s vanishes on every such V, hence globally, a contradiction.

[F2]

The current Rational sections of line bundles are Cartier divisors says that a nonzero rational section s of an invertible sheaf L determines a Cartier divisor div⁡C(s) with OC(div⁡C(s))≅L. By [F1] a nonzero global section has a nonzero generic germ, so this interface applies to it; its regularity makes the associated Cartier divisor effective. The current Invertible sheaf of cartier divisor constructs OC(D) and gives OC(0)≅OC. The degree homomorphism The degree of a divisor descends to the Picard group of a normal proper curve satisfies deg⁡kOC(D)=deg⁡kD for every divisor D on the normal proper curve.

[F3]

The current Cartier-to-Weil theorem identifies Cartier divisors on the smooth proper geometrically integral curve C with the divisors of Divisors on a smooth proper curve, compatibly with principal divisors and local orders. Its local rings at closed points are DVRs and its generic local ring is a field, so C is normal; the normal proper curve hypothesis of [F2] is therefore met. In particular, the effective Cartier divisor div⁡C(s) of [F2] gives an effective divisor on C (Cartier and Weil divisors agree on a smooth curve, Divisors on a smooth proper curve, Principal weil divisor and class group).

[F4]

Effective divisors: for an effective divisor D=∑xnx[x] on a proper geometrically integral curve, deg⁡k(D)=∑xnx[κ(x):k] is a nonnegative integer, and deg⁡k(D)=0 if and only if D=0 (Effective divisors have nonnegative degree).

[F5]

The structure sheaf: the canonical map k→H0(C,OC) is an isomorphism, so H0(C,OC)≅k≠0 and the structure sheaf has a nonzero global section; and deg⁡kOC=deg⁡k(0)=0 under the dictionary of [F2] (Functions on a proper curve, Sheaf cohomology as right derived global sections).

[F6]

The Axiom of Choice is inherited through the degree, Cartier-to-Weil and structure-sheaf cohomology suppliers of [F2], [F3] and [F5]. The Cartier-to-Weil supplier requires Dependent Choice, supplied from AC by AC implies DC implies countable choice. The proof selects one nonzero section of the given nonzero space and makes no further selection (The Axiom of Choice, The axiom of dependent choice: a relation in which every element is related to something admits an N-indexed chain).

Proof

technique · direct; convert a nonzero global section of $\mathcal L$ into an effective divisor of degree $\deg_k\mathcal L=0$, conclude that the divisor is zero, and read off the triviality of $\mathcal L$; the converse uses the identity section of the structure sheaf
1.1F1F2

The nonzero section and its effective divisor. Since H0(C,L)≠0, [F1] provides a nonzero global section s∈Γ(C,L) and proves its generic germ is nonzero. Thus s is a nonzero rational section; by [F2] its associated Cartier divisor div⁡C(s) is effective and satisfies OC(div⁡C(s))≅L. Effectiveness means that in every local frame the coefficient of s is regular, so all orders of vanishing are nonnegative.

2.1F2step 1.1

The degree of the divisor is zero. By the degree homomorphism in [F2], deg⁡kOC(D)=deg⁡kD for every divisor D; applying this to D=div⁡C(s) and using the isomorphism OC(div⁡C(s))≅L of step 1.1 gives deg⁡kdiv⁡C(s)=deg⁡kOC(div⁡C(s))=deg⁡kL=0, the last equality being the hypothesis.

3.1F3F4step 2.1

The divisor vanishes. By [F3] the effective Cartier divisor div⁡C(s) corresponds to an effective divisor on C of the same degree 0 computed in step 2.1; by [F4] an effective divisor of degree zero is the zero divisor. Hence div⁡C(s)=0.

4.1F2step 1.1step 3.1

The invertible sheaf is trivial. Combining the isomorphism of step 1.1 with the vanishing div⁡C(s)=0 of step 3.1 and the identity OC(0)≅OC of [F2], L≅OC(div⁡C(s))=OC(0)≅OC.

5.1F1F2F5F6step 1.1step 4.1∎

The converse and the contrapositive. Conversely, if L≅OC then deg⁡kL=deg⁡kOC=0 by [F5], and H0(C,L)≅H0(C,OC)≅k≠0 by [F5]; so for invertible sheaves of degree zero the existence of a nonzero global section is equivalent to triviality. Contrapositively, if L has degree zero and is not isomorphic to OC, then H0(C,L)=0: a nonzero group would produce a nonzero section and force L≅OC by steps 1.1 through 4.1. Choice is used through the current degree, Cartier-to-Weil and structure-sheaf cohomology interfaces, including the AC-to-DC route recorded in [F6]. The only selection is the single nonzero section s of the given nonzero group H0(C,L).

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Nontrivial degree-zero line bundles have no sections

Statement

Assume the Axiom of Choice as inherited from the degree, Cartier-to-Weil, smooth-base-change, weighted-Bezout and structure-sheaf cohomology suppliers. It supplies the Dependent Choice premise of the Cartier-to-Weil route through AC implies DC implies countable choice. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field) and let L be an invertible sheaf on C (Invertible sheaves) with deg⁡kL=0 which is not isomorphic to OC. Then H0(C,L)=0.

The plane-cubic instance. Moreover, let E=V+(F)⊆Pk2 be a plane cubic smooth over k and of pure dimension one, where F is a nonzero homogeneous form of degree three, and let P≠Q be k-rational points of E. The proof shows that this smooth plane cubic is geometrically integral, so the arithmetic-genus computation of Arithmetic genus of a plane curve applies. Then the invertible sheaf OE(P−Q) has degree zero, is not trivial, and H0(E,OE(P−Q))=0. The triviality obstruction is the classical one: a trivialization would exhibit a rational function with divisor P−Q, hence a degree-one morphism E→Pk1 and an isomorphism E≅Pk1, contradicting g(E)=1 and g(Pk1)=0. This discharges, without Serre duality, the promise recorded by the preceding-pair counterexample item cex-degree-zero-line-bundle-no-section (batch 6 of this run).

Facts & Assumptions

Given: the Axiom of Choice inherited from the degree, Cartier-to-Weil, smooth-base-change, weighted-Bezout and structure-sheaf cohomology suppliers; a field k; a smooth proper geometrically integral curve C over k; an invertible sheaf L on C with deg⁡kL=0; and, for the instance, a plane cubic E=V+(F)⊆Pk2 smooth over k and of pure dimension one with P≠Q rational over k.

[F1]

Contrapositive of the section-triviality corollary: if L is an invertible sheaf of degree zero on C with H0(C,L)≠0, then L≅OC; equivalently, an invertible sheaf of degree zero that is not trivial has no nonzero global section (A degree-zero line bundle with a nonzero section is trivial, Sheaf cohomology as right derived global sections).

[F2]

On a normal proper curve, the degree of the attached invertible sheaf satisfies deg⁡kOC(D)=deg⁡kD for every divisor D (The degree of a divisor descends to the Picard group of a normal proper curve, Invertible sheaf of cartier divisor). A divisor D=∑xnx[x] has degree deg⁡kD=∑xnx[κ(x):k]; in particular each k-rational point has residue degree one (Degree divisor proper curve, Divisors on a smooth proper curve, The residue field at a point of an affine scheme).

[F3]

The Cartier-to-Weil/Picard dictionary on a smooth proper geometrically integral curve identifies the attached sheaf classes with divisor classes and preserves principal divisors (Cartier and Weil divisors agree on a smooth curve, On an integral scheme, Cartier divisors modulo principal divisors compute the Picard group, Cartier divisor, Principal weil divisor and class group). The rational-section interface identifies the divisor of a nonzero rational section with its line bundle (Rational sections of line bundles are Cartier divisors). Thus if OE(P−Q)≅OE, then P−Q is a principal divisor on E.

[F4]

Every nonconstant rational function f∈k(E)× on a smooth proper geometrically integral curve defines a finite locally free morphism φf:E→Pk1 of degree [k(E):k(f)], and its fiber over infinity is the pole divisor (f)∞ with that same degree (A nonconstant rational function defines a finite map to the projective line). A birational morphism between smooth proper geometrically integral curves is an isomorphism (Birational smooth proper curves are isomorphic).

[F5]

If X=V+(F)⊆Pk2 is an integral plane curve cut out by a homogeneous form of degree d≥1, then H0(X,OX)=k and pa(X)=1−χ(OX)=(d−1)(d−2)/2 (Arithmetic genus of a plane curve). For a smooth proper geometrically integral curve the arithmetic genus is its genus (Genus via the Euler characteristic). Once [F7] and [F8] establish that E is such a curve, this gives g(E)=1.

[F6]

The projective line has genus zero by Divisors on the projective line are classified by degree and Genus via the Euler characteristic. Genus is invariant under isomorphism because scheme isomorphisms induce isomorphisms on cohomology (Variance of sheaf cohomology, Every functor preserves isomorphisms).

[F7]

Smoothness is preserved under arbitrary base change (Smoothness survives base change and composition). Over the algebraic closure kˉ, coprime positive-degree plane forms have a nonempty projective intersection by Algebraic Bezout formula as a sum of local scheme lengths. On a standard affine chart the hypersurface is given by the dehomogenized equation (projective hypersurface affine pieces); if that equation lies in the square of a closed point's maximal ideal, all its partial derivatives vanish there, contradicting the smoothness criterion for a one-equation presentation (Relative Jacobian criterion with its presentation hypothesis). The polynomial ring over kˉ is a UFD, so an irreducible equation generates a prime ideal (Every finite-variable polynomial ring over a field is a UFD, with prime irreducibles and principal height-one primes).

[F8]

Projective space over k is proper, closed immersions are proper, and proper morphisms compose (Finite-dimensional projective space is proper over every base, Closed immersions are proper, Properness survives composition). Hence the closed plane subscheme E is proper over k.

[F9]

Every smooth proper geometrically integral curve considered here is normal: its local ring at a closed point is a DVR by Local rings at closed points of smooth curves are discrete valuation rings, and the generic local ring is a field; these are regular local domains and hence integrally closed by regular local rings are normal. This supplies the normality hypothesis of the degree homomorphism in [F2].

[F10]

The Axiom of Choice supplies the Dependent Choice premise of the Cartier-to-Weil route through AC implies DC implies countable choice. The stated Choice assumption also covers the degree, smooth-base-change, weighted-Bezout, Jacobian, finite-morphism and cohomology suppliers used here (The Axiom of Choice, The axiom of dependent choice: a relation in which every element is related to something admits an N-indexed chain).

Proof

technique · prove the general statement as the contrapositive of the section-triviality corollary, show that a smooth pure-dimension-one plane cubic is geometrically integral, and then use the degree-one-map obstruction for the cubic instance
1.1F1

The general statement. Let L be an invertible sheaf on C with deg⁡kL=0 and L≇OC. If H0(C,L)≠0, then [F1] gives L≅OC, contrary to the hypothesis. Hence H0(C,L)=0.

1.2F7

Base change and factorization. Write F as a homogeneous cubic and pass to kˉ. By [F7], Ekˉ=V+(F)⊆Pkˉ2 is smooth. We show that F is irreducible and squarefree over kˉ. If an irreducible factor G occurs with multiplicity at least two, then 2deg⁡G≤3, so G is linear. At least one coordinate linear form L is not proportional to G; thus G and L are coprime. By weighted Bezout [F7], they meet at a closed point z∈V+(G,L). Choose a standard affine chart containing z. The dehomogenized equation f of Ekˉ is divisible by the square of the dehomogenized G, so f∈mz2 and every first partial derivative of f vanishes at z. By [F7] the one-equation chart is not smooth at z, contradicting the smoothness of Ekˉ. If F is squarefree but reducible, factor it as F=GH with G,H coprime homogeneous forms of positive degree. Weighted Bezout [F7] gives a closed point z∈V+(G,H). In a standard affine chart containing z, the dehomogenized equation is f=gh∈mz2, so all its first partial derivatives vanish at z. The same Jacobian criterion contradicts smoothness. Thus F is irreducible and squarefree over kˉ. Since the polynomial ring is a UFD, (F) is prime; consequently Ekˉ is integral and E is geometrically integral. The given pure dimension one then makes E an integral plane curve.

2.1F2F8F9step 1.2

The degree of OE(P−Q). Since P and Q are k-rational, [κ(P):k]=[κ(Q):k]=1 by [F2]. Therefore deg⁡k(P−Q)=1−1=0. The closed plane subscheme E is proper by [F8], and its smoothness and geometric integrality from step 1.2 make it normal by [F9]. Hence [F2] applies and gives deg⁡kOE(P−Q)=0.

2.2F5F8step 1.2

Properness and genus. By [F8], E is proper as a closed subscheme of projective space. It is smooth by hypothesis and geometrically integral by step 1.2, hence a smooth proper geometrically integral curve. Apply [F5] to the integral plane cubic E to get arithmetic genus one; smoothness and properness identify this with its genus, so g(E)=1.

3.1F3F4F6step 2.2

Nontriviality of OE(P−Q). Suppose OE(P−Q)≅OE. By [F3], P−Q is principal, so some f∈k(E)× satisfies div⁡(f)=P−Q. Since P≠Q, f is nonconstant, and its pole divisor is (f)∞=[Q], of degree one. By [F4], f defines a finite morphism φf:E→Pk1 whose degree is [k(E):k(f)]=deg⁡k(f)∞=1. Thus the morphism is birational, and [F4] makes it an isomorphism. This contradicts g(E)=1 from step 2.2 and g(Pk1)=0 from [F6]. Hence OE(P−Q)≇OE.

4.1F1step 1.1step 2.1step 3.1

The vanishing for the cubic. By step 2.1 the invertible sheaf OE(P−Q) has degree zero, and by step 3.1 it is not trivial. Step 1.1 applied to C=E and L=OE(P−Q) gives H0(E,OE(P−Q))=0.

5.1F1F2F3F4F5F6F7F8F9F10step 1.1step 1.2step 2.1step 2.2step 3.1step 4.1∎

Conclusion and Choice accounting. Step 1.1 proves the general statement: a nontrivial degree-zero invertible sheaf on a smooth proper geometrically integral curve has no nonzero global section, by the contrapositive of the section-triviality corollary and without Serre duality. Steps 1.2, 2.1, 2.2, and 3.1 show that every smooth pure-dimension-one plane cubic in the stated scope is a smooth proper geometrically integral genus-one curve and that OE(P−Q) is degree zero and nontrivial; step 4.1 gives the promised vanishing. AC supplies DC through [F10], and the other inherited Choice uses are exactly those named there. No additional selection is made beyond the given data and the finite nonempty intersections in the Bezout argument.

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

The index of speciality i(D)

Definition

Assume the Axiom of Choice for proper-cohomology finiteness, the curve Cartier-to-Weil identification, and the cohomological Riemann-Roch theorem (The Axiom of Choice). AC supplies the Dependent Choice premise of the curve Cartier-to-Weil result through AC implies DC implies countable choice. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field), let D be a divisor on C (Divisors on a smooth proper curve) and let OC(D) be its associated invertible sheaf. The current Finite-dimensionality of the Riemann-Roch space proves that H1(C,OC(D)) is finite-dimensional under this assumption. The index of speciality of D is the nonnegative integer i(D):=h1(C,OC(D))=dim⁡kH1(C,OC(D)), the first cohomology dimension of the attached invertible sheaf (Sheaf cohomology as right derived global sections, Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis, The Riemann-Roch dimension l(D)).

The index of speciality depends only on the linear equivalence class of D. Indeed, Cartier and Weil divisors agree on a smooth curve identifies the Weil divisors on C with Cartier divisors and preserves principal divisors; Linear equivalence cartier divisors gives the Cartier equivalence relation; and the current Cartier-sheaf and addition-tensor dictionaries (Invertible sheaf of cartier divisor, The sheaf of a Cartier divisor is invertible, Addition of Cartier divisors is tensor product of their sheaves, Rational sections of line bundles are Cartier divisors) identify linearly equivalent divisors with isomorphic invertible sheaves. Their first cohomology groups therefore have the same dimension.

For the zero divisor, i(0)=h1(C,OC)=g(C), by Genus via the Euler characteristic. The independent Riemann-Roch for curves: the Euler-characteristic form gives h0(C,OC(D))−h1(C,OC(D))=deg⁡k(D)+1−g(C). By the definitions of l(D) and i(D) this is l(D)−i(D)=deg⁡k(D)+1−g(C). Thus i(D)≥0 gives the Riemann inequality l(D)≥deg⁡k(D)+1−g(C), and equality holds exactly when i(D)=0.

This definition is deliberately one-sided. No identification of H1(C,OC(D)) with the sections of a complementary invertible sheaf, and no Serre-duality statement, is made or used here. The Cartier, finite-dimensionality, and Euler-characteristic suppliers named above are present in the working tree as draft or published items as indicated by their frontmatter; their presence alone does not certify a mathematical review.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

Riemann-Roch as l minus i

Statement

Assume the Axiom of Choice, inherited from the Riemann-Roch, genus and finiteness suppliers below. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field) with genus g=g(C)=h1(C,OC) (Genus via the Euler characteristic), and let D be a divisor on C (Divisors on a smooth proper curve). Then l(D)−i(D)=h0(C,OC(D))−h1(C,OC(D))=deg⁡k(D)+1−g, where l(D)=h0(D) is the dimension of the Riemann-Roch space, i(D)=h1(D) the index of speciality (The Riemann-Roch dimension l(D), The index of speciality i(D)). Moreover i(D)≥0, so l(D)≥deg⁡k(D)+1−g, with equality if and only if i(D)=0; a divisor is called nonspecial exactly when i(D)=0, a terminology fixed in def-nonspecial-divisor, which follows on this page and consumes the present theorem. For the zero divisor the identity reads 1−g=0+1−g with i(0)=g. The index of speciality remains an unknown defect: no duality identifies it with the space of sections of a complementary divisor, and no threshold statement about 2g−2 is made.

The attachment of OC(D) and the identification of L(D) with H0(C,OC(D)) use the current interfaces Invertible sheaf of cartier divisor, Rational sections of line bundles are Cartier divisors and Cartier and Weil divisors agree on a smooth curve, inherited through Riemann-Roch for curves: the Euler-characteristic form.

Facts & Assumptions

Given: a field k, a smooth proper geometrically integral curve C over k with genus g=g(C)=h1(C,OC), and a divisor D on C.

[F1]

The curve C is proper, of finite type and of chain dimension one over the field k, and a divisor on C is a finite formal integral combination of closed points with k-degree deg⁡k(D) (Curves over a field, Divisors on a smooth proper curve).

[F2]

Notation: l(D)=h0(D)=dim⁡kH0(C,OC(D)) and hi(D)=dim⁡kHi(C,OC(D)) for every i≥0; both are nonnegative integers, and for the zero divisor l(0)=dim⁡kH0(C,OC)=1 (The Riemann-Roch dimension l(D)).

[F3]

The index of speciality: i(D)=h1(C,OC(D))=dim⁡kH1(C,OC(D)) is a nonnegative integer, and i(0)=h1(C,OC)=g(C) is the genus; it depends only on the linear equivalence class of D (The index of speciality i(D), Genus via the Euler characteristic).

[F4]

Riemann-Roch in Euler-characteristic form: h0(D)−h1(D)=χ(C,OC(D))=deg⁡k(D)+1−g; no Serre duality is used, and h1(D) is left as an unknown nonnegative integer (Riemann-Roch for curves: the Euler-characteristic form).

[F5]

The current interfaces Invertible sheaf of cartier divisor, Rational sections of line bundles are Cartier divisors and Cartier and Weil divisors agree on a smooth curve supply the attachment of OC(D) and the identification of L(D) with its global sections; this use is inherited from [F4].

[F6]

The Axiom of Choice is used exactly through the Riemann-Roch supplier [F4], the genus definition [F3] and the finiteness supplier [F2]; no further selection is made below (The Axiom of Choice).

Proof

technique · direct; substitute the definitions $l(D)=h^0(D)$ and $i(D)=h^1(D)$ into the Euler-characteristic Riemann-Roch identity, and read off the inequality and its equality case from $i(D)\ge0$
1.1F1F2F3F4

Set-up. By [F1] the curve C is proper over k and D is a divisor on C with k-degree deg⁡k(D). By [F2] l(D)=h0(D) and by [F3] i(D)=h1(C,OC(D)), a nonnegative integer; by [F4] applied to the divisor D and the genus g=g(C) of [F3], h0(D)−h1(D)=deg⁡k(D)+1−g.

2.1F2F3step 1.1

The defect identity. Substituting the definitions of l and i from step 1.1 into the Riemann-Roch identity gives l(D)−i(D)=h0(D)−h1(D)=deg⁡k(D)+1−g, the first displayed identity; all three expressions are integers, the left side because it is a difference of dimensions.

3.1F2F3step 2.1

The inequality and the equality case. Solving the identity of step 2.1 for l(D) gives l(D)=deg⁡k(D)+1−g+i(D); since i(D)≥0 by [F3], this gives l(D)≥deg⁡k(D)+1−g. If l(D)=deg⁡k(D)+1−g, then subtracting gives i(D)=0; conversely if i(D)=0, the same identity gives l(D)=deg⁡k(D)+1−g. Hence equality holds if and only if i(D)=0, and i(D)≥0 is the nonnegativity of the dimension h1(D) of [F2].

3.2F2F3step 2.1

The zero divisor. By [F2] l(0)=1, and by [F3] i(0)=g(C)=g; the identity of step 2.1 at D=0 therefore reads 1−g=0+1−g, so the zero divisor is nonspecial exactly when g=0, and it is special with defect g when g>0.

4.1F2F3F4F5F6step 2.1step 3.1step 3.2∎

Conclusion and choice accounting. Steps 2.1, 3.1 and 3.2 give the defect identity, the inequality with its equality case, and the zero-divisor reading, all for an arbitrary divisor D on C; the index of speciality appears only as the dimension i(D) of [F3], with no duality identification and no threshold statement. The Axiom of Choice is used only through the suppliers recorded in [F6], namely the Riemann-Roch theorem [F4], the genus definition [F3] and the finiteness supplier [F2]; the flagged dictionary [F5] records the inherited obligation on OC(D).

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

Special and nonspecial divisors

Definition

Assume the Axiom of Choice inherited from proper-cohomology finiteness and the Riemann-Roch theorem (The Axiom of Choice). Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field) with genus g=g(C) (Genus via the Euler characteristic), and let D be a divisor on C (Divisors on a smooth proper curve) with index of speciality i(D)=h1(C,OC(D)) and l(D)=h0(D) (The index of speciality i(D), The Riemann-Roch dimension l(D)).

Call D nonspecial when i(D)=0,that is, whenH1(C,OC(D))=0, and call it special otherwise, so that D is special exactly when i(D)≥1. These are the classical names for the two cases of the index of speciality: vanishing first cohomology, and nonvanishing first cohomology.

The current Riemann-Roch as l minus i gives l(D)−i(D)=deg⁡k(D)+1−g, and i(D)≥0, so the Riemann inequality l(D)≥deg⁡k(D)+1−g always holds. Consequently, D nonspecial  ⟺  l(D)=deg⁡k(D)+1−g  ⟺  the Riemann inequality is an equality for D, and D special  ⟺  l(D)>deg⁡k(D)+1−g, because the difference l(D)−(deg⁡k(D)+1−g)=i(D) is exactly the index of speciality. Thus a divisor is special precisely when its space of sections is larger than the Riemann inequality requires, and the equality case of that inequality is the vanishing of H1(C,OC(D)).

Speciality is a property of the divisor class and not of the individual divisor: the index of speciality depends only on the linear equivalence class of D (The index of speciality i(D)), so linearly equivalent divisors are special or nonspecial together. For the zero divisor the definition gives i(0)=h1(C,OC)=g, the genus of C (The index of speciality i(D), Genus via the Euler characteristic), so 0 is nonspecial exactly when g=0 and special exactly when g≥1; equivalently, the equality reading l(0)=1=deg⁡k(0)+1−g holds exactly when g=0, in agreement with l(0)=1 for the zero divisor (The Riemann-Roch dimension l(D)).

The finite-valued cohomology dimensions and Riemann-Roch identity used here are supplied by the current Finite-dimensionality of the Riemann-Roch space, The index of speciality i(D), Genus via the Euler characteristic and Riemann-Roch as l minus i. Their source files are present in the working tree; their mathematical status remains the status recorded in their frontmatter.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

Sufficiently positive divisors in a fixed direction are nonspecial

Statement

Assume the Axiom of Choice as inherited from the vanishing theorem. Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field), let φ:C→Pk1 be a finite k-morphism and let A be an effective divisor on C with OC(A)≅φ∗OPk1(1) — for instance A=(f)∞ for a nonconstant f∈k(C)×, as produced by Finite morphisms from a curve to the projective line (Divisors on a smooth proper curve). Let D0 be a divisor on C. Then there is an integer n0, depending on D0 and on the fixed morphism φ (through A), such that every divisor D with D≥D0+n0A is nonspecial: H1(C,OC(D))=0. Explicitly, every divisor of the form D0+nA+E with n≥n0 and E effective is nonspecial.

Fixed direction only. This is only a statement about the fixed ample direction A: it is not claimed that every divisor of degree greater than 2g−2 is nonspecial, which is a different statement requiring the duality pair that follows this page. No threshold in terms of deg⁡k(D) alone and no Serre duality is used or asserted here.

The example uses the finite-map construction Finite morphisms from a curve to the projective line and the fixed-direction vanishing theorem Vanishing of H^1 in a fixed ample direction, as stated in [F1].

Facts & Assumptions

Given: a field k, a smooth proper geometrically integral curve C over k, a finite k-morphism φ:C→Pk1, an effective divisor A on C with OC(A)≅φ∗OPk1(1), and a divisor D0 on C.

[F1]

Vanishing theorem: for the fixed curve, morphism φ and effective divisor A, and for the given D0, there is an integer n0 such that H1(C,OC(D0+nA+E))=0 for every n≥n0 and every effective divisor E; equivalently h1(D)=0 for every divisor D with D≥D0+n0A (Vanishing of H^1 in a fixed ample direction).

[F2]

Nonspecial divisors: i(D)=h1(C,OC(D))=dim⁡kH1(C,OC(D)) is the index of speciality, and D is nonspecial exactly when i(D)=0, that is, exactly when H1(C,OC(D))=0; D is special exactly when i(D)≥1 (Special and nonspecial divisors, The index of speciality i(D), The Riemann-Roch dimension l(D)).

[F3]

Divisors: a divisor on C is a finite formal integral combination of closed points, effective when all coefficients are nonnegative, and D≥D′ means that D−D′ is effective (Divisors on a smooth proper curve).

[F4]

The genus g=g(C)=h1(C,OC) is a nonnegative integer; the threshold 2g−2 and the duality pair are not part of this lemma and no statement about them is made here (Genus via the Euler characteristic).

[F5]

The Axiom of Choice is available and is inherited only through the vanishing theorem [F1]; the argument below transforms the vanishing statement into the definition of nonspeciality and selects nothing beyond the integer n0 supplied by [F1] (The Axiom of Choice).

Proof

technique · apply the fixed-direction vanishing theorem to $D_0$, then translate $h^1(D)=0$ into nonspeciality
1.1F1

The fixed-direction threshold. By [F1] applied to the given divisor D0 and the fixed morphism φ and divisor A, there is an integer n0 with H1(C,OC(D0+nA+E))=0 for every n≥n0 and every effective E, equivalently h1(D)=0 for every divisor D with D≥D0+n0A; this integer depends only on D0 and on φ through the fixed sheaf OC(A)≅φ∗O(1), not on D or E.

2.1F2step 1.1

Nonspeciality above the threshold. Let D be a divisor with D≥D0+n0A. By step 1.1, h1(D)=0, and by [F2] the index of speciality vanishes exactly for the nonspecial divisors, so D is nonspecial, that is, H1(C,OC(D))=0. The same applies to every D=D0+nA+E with n≥n0 and E effective, since such a divisor dominates D0+n0A and is of the form covered by [F1].

3.1F1F2F3F4F5step 1.1step 2.1∎

Conclusion, the fixed-direction restriction and choice accounting. Steps 1.1 and 2.1 show that every divisor D≥D0+n0A, in particular every D0+nA+E with n≥n0 and E effective, is nonspecial with H1(C,OC(D))=0. Nothing is asserted about divisors merely of large degree: the lemma provides no universal bound in terms of deg⁡k(D) and 2g−2, and no Serre duality enters, as [F4] records. The integer n0 is the one supplied by [F1] for D0 and the fixed morphism; it is not chosen, and the Axiom of Choice is inherited only through [F1], as recorded in [F5].

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

The dimension of a complete linear system

Statement

Assume the Axiom of Choice as inherited from proper-cohomology, projective-space and Riemann-Roch suppliers (The Axiom of Choice). Let k be a field, let C be a smooth proper geometrically integral curve over k (Curves over a field) of genus g=g(C) (Genus via the Euler characteristic), let D be a divisor on C (Divisors on a smooth proper curve) and let ∣D∣ be its complete linear system (Complete linear system). For a finite-dimensional k-vector space W, write Pk(W):=Proj⁡(Sym⁡k(W∨)) for the projective scheme parameterizing one-dimensional subspaces of W. When ∣D∣ is nonempty, put PD:=Pk(L(D)). The section-divisor correspondence identifies the set ∣D∣ with the set of k-rational points PD(k).

Then:

  1. ∣D∣ is nonempty if and only if l(D)≥1 (The Riemann-Roch dimension l(D)); equivalently ∣D∣ is empty exactly when L(D)=0;
  2. if ∣D∣ is nonempty and l(D)=r≥1, a choice of basis of L(D) identifies PD with the projective scheme Pkr−1, and we define dim⁡k∣D∣:=dim⁡PD. Its dimension is dim⁡k∣D∣=l(D)−1=deg⁡k(D)−g+i(D), with i(D) the index of speciality (The index of speciality i(D));
  3. if D is nonspecial (Special and nonspecial divisors) then ∣D∣ is nonempty if and only if deg⁡k(D)≥g, and in that case dim⁡k∣D∣=deg⁡k(D)−g;
  4. for the zero divisor, ∣0∣={0} is a single k-rational point, P0≅Pk0, and dim⁡k∣0∣=0=l(0)−1. The nonspecial value deg⁡k(0)−g=−g is attained as dim⁡k∣0∣ exactly when g=0.

Thus dim⁡k∣D∣ means the Krull dimension of the projectivization scheme, not the dimension of its set of k-rational points. No algebraic-closure hypothesis on k is used.

Facts & Assumptions

Given: the Axiom of Choice; a field k; a smooth proper geometrically integral curve C over k of genus g=g(C); a divisor D on C; the complete linear system ∣D∣; and the index of speciality i(D)=h1(C,OC(D)).

[F1]

The assignment f↦div⁡(f)+D induces a bijection (L(D)∖{0})/k×→∣D∣. Hence ∣D∣ is identified with the set of k-lines in L(D) and is empty exactly when L(D)=0 (Effective divisors linearly equivalent to D are sections modulo scalars, Complete linear system).

[F2]

Under the stated Axiom of Choice, the current Finite-dimensionality of the Riemann-Roch space proves that L(D) and all Hq(C,OC(D)) are finite-dimensional. Thus l(D) and i(D) are nonnegative integers (The Riemann-Roch dimension l(D), The index of speciality i(D), Finite-dimensional vector space, and its dimension dim⁡FV; infinite-dimensional means having no finite basis). For D=0, H0(C,OC)=k, so l(0)=1 (Functions on a proper curve).

[F3]

The cohomological Riemann-Roch identity is l(D)−i(D)=deg⁡k(D)+1−g, and a divisor is nonspecial exactly when i(D)=0 (Riemann-Roch as l minus i, Special and nonspecial divisors).

[F4]

For a nonzero finite-dimensional k-vector space W of dimension r, Pk(W):=Proj⁡(Sym⁡k(W∨)) is the projective scheme parameterizing lines in W. A basis of W identifies it with Pkr−1; its k-rational points are exactly the one-dimensional k-subspaces of W: if e0,…,er−1 is a basis, the points [a0:⋯:ar−1] with ai∈k and not all ai=0, modulo common nonzero scalar, correspond to the line spanned by ∑iaiei. The case r=1 is Pk0 (Symmetric algebra of a vector space, Relative projective space from standard charts, Projective space is Proj of a polynomial ring).

[F5]

The projective scheme Pkr−1 has r standard open charts, each isomorphic to Spec⁡k[y1,…,yr−1] (Relative projective space from standard charts, Projective space is Proj of a polynomial ring). The coordinate ring has Krull dimension r−1 (A polynomial ring in n variables over a field has dimension n). Since a field is Noetherian, these polynomial rings are Noetherian (A field has only the zero ideal and itself, hence is Noetherian, Every algebra of finite type over a Noetherian ring is a Noetherian ring), and their spectra are Noetherian spaces (The spectrum of a Noetherian ring is a Noetherian topological space). Any descending chain of closed subsets of Pkr−1 stabilizes after restriction to each standard chart. Since there are finitely many charts, the maximum of their stabilization indices works on the whole projective space, so it is Noetherian. Its dimension is the supremum of the chart dimensions by Dimension can be computed on an open cover and Chain dimension and the empty-space convention. Consequently dim⁡Pkr−1=r−1 for every field k.

[F6]

The Axiom of Choice is inherited from the finiteness, projective-space and Riemann-Roch suppliers. The only choice made below is a basis of the finite-dimensional space L(D) (The Axiom of Choice).

[F7]

The k-degree is deg⁡kD=∑xnx[κ(x):k], a sum over the finite support of D=∑xnx[x]; the zero divisor has empty support and deg⁡k(0)=0 (Degree divisor proper curve, Divisors on a smooth proper curve).

Proof

technique · read the complete linear system off the bijection with the $k$-lines in $L(D)$, use its projectivization scheme for dimension, and substitute the Riemann-Roch identity
1.1F1F2

The empty case. By [F1], ∣D∣ is in bijection with the k-lines in L(D). It is empty exactly when L(D)=0, which by [F2] is equivalent to l(D)=0. Therefore ∣D∣ is nonempty exactly when l(D)≥1.

1.2F1F2F4F5

The projective parameter scheme. Suppose ∣D∣ is nonempty, so L(D)≠0 by [F1]. By [F2], L(D) is finite-dimensional; put r=l(D)=dim⁡kL(D)≥1. Choose a basis. By [F4] it identifies PD=Proj⁡(Sym⁡k(L(D)∨)) with Pkr−1, whose k-rational points are the k-lines in L(D). By [F1], these points are exactly ∣D∣. The standard-chart calculation [F5] then gives dim⁡k∣D∣:=dim⁡PD=r−1=l(D)−1.

1.3F1F2F3F4F5F7

The zero divisor. By [F2], L(0)=H0(C,OC)=k and l(0)=1. Its unique k-line maps under [F1] to div⁡(1)+0=0, so ∣0∣={0} and P0≅Pk0 by [F4]. Therefore dim⁡k∣0∣=0=l(0)−1. By [F7], deg⁡k(0)=0, and by [F3], i(0)=g. The nonspecial value deg⁡k(0)−g=−g equals the actual dimension zero exactly when g=0.

2.1F3step 1.2

Riemann-Roch substitution. By [F3], l(D)=deg⁡k(D)+1−g+i(D). Combining this with step 1.2 gives dim⁡k∣D∣=l(D)−1=deg⁡k(D)−g+i(D).

3.1F3step 1.1step 2.1

The nonspecial case. If D is nonspecial then i(D)=0 by [F3], so l(D)=deg⁡k(D)+1−g. By step 1.1, ∣D∣ is nonempty exactly when this integer is at least one, equivalently when deg⁡k(D)≥g. When this holds, step 2.1 gives dim⁡k∣D∣=deg⁡k(D)−g.

4.1F1F2F3F4F5F6F7step 1.1step 1.2step 2.1step 3.1step 1.3

Conclusion and choice accounting. Steps 1.1 and 1.2 give ∣D∣≠∅  ⟺  l(D)≥1 and define its dimension as the Krull dimension of the projectivization scheme. Steps 2.1 and 3.1 give the general and nonspecial dimension formulas; step 1.3 verifies all four zero-divisor claims. The projective-space dimension computation uses standard relative Proj charts and holds over every field, including fields that are not algebraically closed. Choice is used only for the cohomology and Riemann-Roch suppliers and the basis of L(D) in step 1.2.

5.1F1F2F3F4F5F6F7∎

Current supplier boundary. The section-to-divisor set bijection is given by the current Effective divisors linearly equivalent to D are sections modulo scalars and Complete linear system. The Cartier, divisor-space, finite-dimension and Riemann-Roch sources cited above are present in the working tree; their draft or published status remains as recorded in their frontmatter. The projective parameter scheme is defined in this item and its dimension is proved from the published relative-Proj charts and affine polynomial-ring dimension theorem. The construction and dimension calculation apply over arbitrary fields k.

RemarkRemark: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)audited 2026-10-02Open item page →

Why the sharp degree thresholds wait for the duality pair

Remark

Let k be a field and let C be a smooth proper geometrically integral curve over k (Curves over a field) of genus g=g(C) (Genus via the Euler characteristic). This page proves the Euler-characteristic form of Riemann-Roch, l(D)−i(D)=h0(C,OC(D))−h1(C,OC(D))=deg⁡k(D)+1−g, with i(D)≥0, in the notation l(D)=h0(D) and i(D)=h1(D) of The Riemann-Roch dimension l(D) and The index of speciality i(D) (Riemann-Roch as l minus i, Riemann-Roch for curves: the Euler-characteristic form), and it proves the fixed-direction Serre vanishing theorem Vanishing of H^1 in a fixed ample direction: for a fixed finite k-morphism φ:C→Pk1 and a fixed effective divisor A with OC(A)≅φ∗OPk1(1), and for every divisor D0 on C, there is an integer n0=n0(D0,φ) such that H1(C,OC(D0+nA+E))=0 for every n≥n0 and every effective divisor E, equivalently h1(D)=0 for every divisor D≥D0+n0A. Through Riemann's theorem for sufficiently positive divisors and Sufficiently positive divisors in a fixed direction are nonspecial this yields the exact count l(D)=deg⁡k(D)+1−g and nonspeciality, again only for D≥D0+n0A: the bound is a bound along one fixed ample direction, it depends on D0 and on φ, and it is not a bound in deg⁡k(D).

The classical degree thresholds are not available on this page and must not be quoted from it. Each of the following rests on the identification i(D)=l(KC−D) supplied by Serre duality, in the pair on residues, Serre duality and the full Riemann-Roch theorem that follows this page:

  1. the canonical identities deg⁡kKC=2g−2 and h0(C,KC)=g for a canonical divisor KC;
  2. vanishing i(D)=0, equivalently l(D)=deg⁡k(D)+1−g, for every divisor of degree greater than 2g−2 — not merely along one fixed ample direction;
  3. base-point-freeness of every invertible sheaf of degree at least 2g;
  4. very ampleness of every invertible sheaf of degree at least 2g+1.

In the vocabulary of the index of speciality (The index of speciality i(D)) the missing ingredient is exactly a description of the dual space of H1(C,OC(D)): nothing on this page identifies that space with the space of sections of a complementary invertible sheaf, defines a canonical divisor, or proves a threshold in terms of deg⁡k(D) alone. The batch-8 items cor-canonical-degree-two-g-minus-two, cor-h0-canonical-differentials-genus, cor-h1-line-bundle-vanishes-degree-over-two-g-minus-two, thm-degree-two-g-line-bundle-basepoint-free and thm-degree-two-g-plus-one-line-bundle-very-ample are the destinations of that material in the following duality pair. This remark does not use those results; it records that the present page proves only the fixed-direction form stated above.

The Axiom of Choice is inherited here from the suppliers named above and is not otherwise used: the remark selects nothing and adds no choice principle of its own (The Axiom of Choice).

5 · Examples, counterexamples and false statements

None yet.

Sources