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.

✓ 11 results · all verified · 3 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 8 not AI-judged were verified by owner audit (typically over a confirmed judge false positive), not failures.

Residues Serre Duality for Curves and the Full Riemann Roch Theorem — Examples

1 · Prerequisites

2 · Summary

These examples and counterexamples run the page's residue, duality and Riemann–Roch results on explicit curves and divisors.

On the projective line with affine coordinate t, the residues of f dt at the points cut out by irreducible polynomials are computed from the Laurent expansion in a local parameter, and their sum is checked to vanish, including the point at infinity. Duality is unpacked for the twists O(−d−2) and O(d): the monomial bases pair into the coefficient of t−1, exhibiting the perfect pairing in coordinates. The full Riemann–Roch theorem is verified on P1 for every degree, and on a genus-one curve it gives h0=deg⁡ for positive-degree line bundles.

Adjunction is applied to plane cubics and plane quartics: a smooth cubic has trivial canonical bundle while a smooth quartic has ωC≅OC(1) with genus three, recovering the hyperplane class. Three counterexamples record the sharpness of the standard thresholds: the canonical map of a hyperelliptic curve factors through the degree-two map to P1 and so is not an embedding, a degree-2g−1 line bundle on a hyperelliptic curve need not be base-point-free, and a degree-2g line bundle need not be very ample. A tame double cover of P1 with 2r simple branch points has its genus computed from the complete Riemann–Hurwitz formula with the different divisor, and one explicit principal part is carried through the residue pairing to compute a class of H1 on the projective line.

3 · Logical flowchart

4 · Definitions, theorems and proofs

None yet.

5 · Examples, counterexamples and false statements

ExampleConstruction: Literature-sourcedVerification: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Residues on the projective line and the vanishing of their sum

Example

Assume the Axiom of Choice as inherited from the cited cohomology and residue suppliers (The Axiom of Choice). Let k be a perfect field and let C=Pk1 have homogeneous coordinates x0,x1, affine coordinate t=x1/x0 on U0=Spec⁡k[t], and point at infinity ∞=[0:1]. Every rational differential is f(t) dt with f∈k(t).

(1) Finite points. Let p=V(g) be a closed point of U0 defined by a monic irreducible g∈k[t], and put u=g(t). Since k is perfect, g is separable, g′(t) is a unit at p, and u is a local parameter. If the completed local expansion is f(t)=∑n≥Nanun with an∈κ(p), then the coefficient-trace definition gives res⁡p(f(t) dt)=Tr⁡κ(p)/k ⁣([u−1]f(t)g′(t)). Here 1/g′(t) is expanded as a unit power series in u, since du=g′(t) dt. For a simple pole f=c/u+regular this is Tr⁡κ(p)/k(c/g′(tˉ)); for higher-order poles the positive powers in the unit expansion can also contribute. When k is algebraically closed and p=a is a rational point, this is the coefficient of (t−a)−1 in the Laurent expansion of f.

(2) The point at infinity. On U1=Spec⁡k[s] with s=x0/x1=1/t, the local parameter is s and t=s−1,dt=−s−2 ds. Thus res⁡∞(f dt) is the coefficient of s−1 in −f(1/s)s−2 ds. In particular, res⁡∞(tn dt)={−1,n=−1,0,n≠−1.

(3) Vanishing of the sum. The monomials tn dt have possible poles only at 0 and ∞; their residues there cancel for n=−1 and are both zero otherwise. For every rational differential on Pk1, the sum of its residues at all closed points is zero by the global residue theorem (The global residue theorem on a smooth proper curve over a perfect field). For example, dtt(t−1)=−dtt+dtt−1 has residues −1 at 0, +1 at 1, and 0 at infinity.

(4) The Laurent-tail class. The principal part t−1 dt at the origin represents a class [t−1dt]∈H1(Pk1,ω), where ω=ΩPk1/k1. Its positive residue sum is +1, so the class is nonzero. The space H1(Pk1,ω) is one-dimensional. Under the fixed Gysin trace of Serre duality, its value is −1, the negative of the positive residue sum; in characteristic two these scalars coincide.

Facts & Assumptions

Given: the Axiom of Choice, a perfect field k, C=Pk1 with coordinate t=x1/x0, a monic irreducible g∈k[t], and the finite-support principal part t−1dt at the origin.

[F1]

The Axiom of Choice is inherited from the cited cohomology, principal-parts, global-residue, and duality suppliers; the computations here make no additional choices. (The Axiom of Choice)

[F2]

The standard charts are U0=Spec⁡k[t] and U1=Spec⁡k[s], with s=1/t on their overlap, and infinity has local parameter s. Closed points in U0 correspond to monic irreducible polynomials g, with residue field k[t]/(g). (Relative projective space from standard charts, Divisors on the projective line are classified by degree, Degree divisor proper curve)

[F3]

At p=V(g), u=g(t) is a uniformizer. Since k is perfect, the irreducible polynomial g is separable, so g′(t) is a unit modulo (g); the chain rule gives du=g′(t)dt. (Local rings at closed points of smooth curves are discrete valuation rings, Perfect fields: every irreducible polynomial is separable)

[F4]

At a closed point with finite separable residue field, residue is the field trace of the coefficient of u−1du in a local parameter u; the residue is independent of the chosen parameter. (Residue of a rational differential at a separable closed point, The residue is independent of the uniformizer)

[F5]

Formal Laurent-series residue extracts the coefficient of exponent −1. (Formal Laurent series K((x)), their order, derivative, and residue)

[F6]

Over a perfect field, the sum of residues of a rational differential on a smooth proper geometrically integral curve is zero. (The global residue theorem on a smooth proper curve over a perfect field)

[F7]

The principal-parts presentation represents H1(C,ω) by finite-support local principal parts modulo rational principal parts. In particular, a single local Laurent tail at the rational origin gives a class. (H^1 of a line bundle on a curve as principal parts modulo meromorphic and regular sections, Canonical bundle and canonical divisors)

[F8]

For Pk1, h1(ω)=h0(O)=1. The fixed normalized Gysin trace on H1(C,ω) is the negative of the positive residue-sum functional over a perfect field. (h^1 of a line bundle equals the dimension of the space of dual sections, Normalization of the trace for Serre duality on a curve)

Verification

Proof technique: compute local coefficients in the two standard charts; use the global residue theorem for arbitrary rational differentials.

1.1F1F3F4F5

At p=V(g), [F3] makes u=g(t) a uniformizer and g′(t) a unit. Since du=g′(t)dt, one has f(t)dt=(f(t)/g′(t))du; the coefficient-trace formula [F4] gives res⁡p(f(t)dt)=Tr⁡κ(p)/k([u−1](f(t)/g′(t))). For a simple pole f=c/u+regular the coefficient is c/g′(tˉ); higher-order poles use the full unit expansion.

1.2F1F2F4F5algebra

At 0, u=t and κ(0)=k, so [F4] gives res⁡0(tndt)=1 for n=−1 and 0 otherwise. At infinity, tndt=−s−n−2ds, whose s−1 coefficient is −1 exactly for n=−1 and 0 otherwise. Thus the monomial residues sum to zero.

2.1F1F6step 1.2

The global residue theorem [F6] supplies the vanishing of the residue sum for every rational differential; the monomial calculation in step 1.2 is only the displayed special case. This does not require writing an arbitrary rational differential as a finite Laurent polynomial plus exact terms.

2.2F2F5step 1.2algebra

For ω=dt/(t(t−1)), 1/(t(t−1))=−1/t+1/(t−1) gives residues −1 and +1 at 0 and 1. At infinity, −f(1/s)s−2ds=−(1−s)−1ds=−(1+s+s2+⋯ )ds, which has no s−1 term. The sum is therefore zero.

3.1F1F6F7F8step 1.2∎

The single principal part at 0 has positive residue sum +1 by step 1.2. By [F6], rational principal parts have total residue zero; regular local parts have zero residue, so this functional detects a nonzero cohomology class by [F7]. Since h1(ω)=1 by [F8], it generates the group. The fixed trace is −1 on this class by [F8], not +1 except in characteristic two.

ExampleConstruction: Literature-sourcedVerification: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Serre duality on the projective line, twist by twist

Example

Assume the Axiom of Choice as inherited from the cited cohomology and duality suppliers (The Axiom of Choice). Let k be any field, let Pk1 have homogeneous coordinates x0,x1 and affine coordinate t=x1/x0 on U0, fix d≥0, and set L=O(−d−2).

The groups H1(Pk1,L) and H0(Pk1,O(d)) have dimension d+1; all other cohomology groups of these two twists vanish. For 1≤j≤d+1, let cj=[t−j⊗x0−d−2]∈H1(Pk1,L), where the displayed expression is the local principal part in the frame x0−d−2 on U0. For 0≤m≤d, the corresponding global section of ωP1⊗L−1 is σm=tmdt⊗x0d+2on U0. On U1, with s=1/t, its expression is σm=−sd−mds⊗x1d+2, so it is regular for exactly the stated range m≤d; under dt↦x0−2 and ds↦−x1−2 it corresponds to x0d−mx1m.

At the rational origin, the positive local residue of the product is [t−1](tm−j)=δj,m+1. This is the identity matrix when the classes are ordered by j=1,…,d+1 and sections by m=0,…,d. Reversing the section order to m=d,…,0 displays the same positive matrix as anti-diagonal. The fixed normalized Serre pairing is the negative of this matrix. This remains perfect over every field; for d=0 its value is −1, which equals 1 in characteristic two.

Facts & Assumptions

Given: the Axiom of Choice, a field k, Pk1 with coordinate t=x1/x0, an integer d≥0, and L=O(−d−2).

[F1]

The Axiom of Choice is inherited from the projective cohomology, principal-parts, field-extension, and duality suppliers, and is used to take an algebraic closure in step 2.1. No other selection is made. (The Axiom of Choice)

[F2]

On Pk1, H0(O(d)) has basis x0d−mx1m for 0≤m≤d, and H1(O(−d−2)) has Laurent basis x0e0x1e1 with e0,e1<0 and e0+e1=−d−2; both groups have dimension d+1. (Cohomology of O(d) on projective space)

[F3]

The standard frames satisfy x0d+2=sd+2x1d+2 on the overlap, s=1/t, and dt=−s−2ds. Also dt maps to x0−2 and ds maps to −x1−2 under ωP1≅O(−2). Thus tmdt⊗x0d+2=−sd−mds⊗x1d+2 and corresponds to x0d−mx1m. The sheaves O(r) are invertible and O(r)−1≅O(−r). (Twisting sheaf on Proj, Relative projective space from standard charts, Invertible sheaves, Canonical bundle and canonical divisors)

[F4]

Principal parts give the cokernel description of H1; in particular, each finite-support tail t−j in the frame x0−d−2 represents the class cj. (H^1 of a line bundle on a curve as principal parts modulo meromorphic and regular sections)

[F5]

At the rational origin with parameter t, the local residue of tm−jdt is its t−1 coefficient. Over a perfect field, the positive residue pairing is the sum of these local residues. (Residue of a rational differential at a separable closed point, The residue pairing of a line bundle with the dual canonical twist)

[F6]

For a smooth proper geometrically integral curve, the fixed normalized Gysin trace and its Serre pairing are defined over every field. Over a perfect field the trace pairing is the negative of the positive residue pairing. (Serre duality for line bundles on a smooth proper curve, and the residue realization, Normalization of the trace for Serre duality on a curve)

[F7]

For a proper scheme, coherent cohomology commutes with arbitrary field extension. The fixed Gysin trace and its cup/evaluation pairing also commute with extension of the base field. (Flat field extension commutes with coherent cohomology, Embedding compatibility of smooth-projective Gysin traces)

[F8]

On Pk1, ωP1≅O(−2): its canonical degree is −2, and the Picard group is classified by degree. (The canonical divisor has degree 2g - 2, The Picard group of the projective line, Divisors on the projective line are classified by degree, Degree divisor proper curve, Canonical bundle and canonical divisors)

Verification

Proof technique: identify the Laurent-tail and section bases, compute the positive residue matrix at the rational origin, then descend the fixed-trace sign from an algebraic closure.

1.1F2F4

By [F2], H1(P1,L) and H0(P1,O(d)) have dimension d+1. For 1≤j≤d+1, the single tail t−j in the frame x0−d−2 is a finite-support principal part at the rational origin, and by [F4] it represents the class cj.

1.2F2F3

By [F3], σm=tmdt⊗x0d+2 has U1 expression −sd−mds⊗x1d+2, hence is regular for 0≤m≤d. Under dt↦x0−2 and ds↦−x1−2 it corresponds to x0d−mx1m, so these sections form the basis of the dual canonical twist.

2.1F1F5F6step 1.1step 1.2

At the rational point 0, the product of the local principal part and section is tm−jdt, whose positive local residue is δj,m+1 by [F5]. Choose an algebraic closure K=kˉ; it is perfect, so [F6] gives tPK1(cj,K∪σm,K)=−δj,m+1. The class and section defined over k pull back to the same expressions over K.

3.1F2F7step 1.1step 1.2step 2.1algebra

By [F7], cohomology classes, cup products and the fixed Gysin trace commute with k→K. Thus tPk1(cj∪σm) maps to −δj,m+1 in K; injectivity of k↪K gives the same scalar over k. In ascending orders the normalized matrix is −Id+1; reversing the section order to m=d,…,0 makes it anti-diagonal with entries −1. Because the σm form a basis by step 1.2, invertibility of this matrix proves that the d+1 classes cj are independent; their number equals dim⁡kH1=d+1 by step 1.1, so they form a basis.

4.1F6F8step 1.1step 1.2step 3.1∎

By [F8], ωP1⊗L−1≅O(d), so the bases above are those of the two Serre-dual spaces. The explicit normalized matrix is perfect, in agreement with the arbitrary-field duality pairing [F6]. For d=0 it is [−1], and characteristic two identifies −1 with 1.

ExampleConstruction: Literature-sourcedVerification: AI-adaptedprecheck passaudited 2026-10-02Open item page →

The full Riemann-Roch theorem on the projective line, in every degree

Example

Assume the Axiom of Choice; it supplies Dependent Choice through AC implies DC implies countable choice. Let k be a field, let C=Pk1 with point at infinity ∞=[0:1], and let D be a divisor on C of degree d=deg⁡kD.

Every divisor on C is linearly equivalent to d[∞], and OC([∞])≅OC(1) with [O(1)] mapping to 1 under Pic⁡(Pk1)≅Z. Hence OC(D)≅OC(d) and ℓ(D)=h0(C,O(d))={d+1,d≥0,0,d<0, because H0(Pk1,O(d)) is the degree-d part of k[x0,x1] for d≥0 and vanishes for d<0.

The canonical divisor satisfies KC∼−2[∞] and deg⁡kKC=2g−2=−2 with g=0, so KC−D∼(−2−d)[∞] and ℓ(KC−D)=h0(C,O(−d−2))={−d−1,d≤−2,0,d≥−1.

Subtracting, ℓ(D)−ℓ(KC−D)={(d+1)−0=d+1,d≥0,0−0=0=d+1,d=−1,0−(−d−1)=d+1,d≤−2, so the full Riemann-Roch identity ℓ(D)−ℓ(KC−D)=deg⁡k(D)+1−g holds on the projective line for every integer degree, including the negative range. The special divisors are exactly those of degree d≤−2, with index of speciality i(D)=ℓ(KC−D)=−d−1, while the divisors of degree d≥−1 are nonspecial with i(D)=0.

Facts & Assumptions

Given: the Axiom of Choice and its consequence Dependent Choice; a field k, the projective line C=Pk1 with point at infinity ∞, and a divisor D of degree d=deg⁡kD.

[F1]

Every divisor on Pk1 is linearly equivalent to (deg⁡kD)[∞], and the associated invertible sheaf of [∞] is O(1); the divisor and invertible-sheaf dictionaries agree on Pk1. (Divisors on the projective line are classified by degree, Cartier and Weil divisors agree on a smooth curve, Invertible sheaf of cartier divisor)

[F2]

Pic⁡(Pk1)≅Z, the class of O(1) corresponding to 1; hence O(D)≅O(d) for D of degree d, and ℓ(D)=h0(C,O(D)) is the dimension of the Riemann-Roch space of D. (The Picard group of the projective line, The Riemann-Roch dimension l(D))

[F3]

On Pk1, H0(Pk1,O(m)) is the degree-m part of k[x0,x1], of dimension m+1 for m≥0, and is 0 for m<0; the twisting sheaves O(m) are the ones attached to the standard charts. (Cohomology of O(d) on projective space, Twisting sheaf on Proj)

[F4]

For a smooth proper geometrically integral curve of genus g, the canonical divisor has degree 2g−2; on Pk1 the genus is 0, so deg⁡kKC=−2, and O(KC)=ωC with KC∼−2[∞] because every degree-(−2) divisor is linearly equivalent to −2[∞]. (The canonical divisor has degree 2g - 2, Canonical bundle and canonical divisors, Degree divisor proper curve, [F1])

[F5]

The full Riemann-Roch theorem states ℓ(D)−ℓ(KC−D)=deg⁡k(D)+1−g for a divisor D on a smooth proper geometrically integral curve of genus g over an arbitrary field, and the index of speciality is i(D)=ℓ(KC−D), with D nonspecial when i(D)=0. (The full Riemann-Roch theorem for divisors on a smooth proper curve, The index of speciality i(D), Special and nonspecial divisors)

[F6]

The Axiom of Choice: every family of nonempty sets has a choice function. (The Axiom of Choice)

[F7]

In ZF, the Axiom of Choice implies Dependent Choice; this supplies the Dependent Choice premise of the Cartier-to-Weil divisor dictionary used in [F1] and [F4]. (AC implies DC implies countable choice, The axiom of dependent choice: a relation in which every element is related to something admits an N-indexed chain)

Verification

Proof technique: reduce an arbitrary divisor to the model O(d) and read off both dimensions from the twisting-sheaf cohomology.

1.1F1F2

By [F1] and [F2], D∼d[∞], so OC(D)≅O(d) and ℓ(D)=h0(C,O(d)).

2.1F1F3F4step 1.1

By [F3] applied to m=d and m=−d−2, ℓ(D)=d+1 for d≥0 and ℓ(D)=0 for d<0, while [F4] gives KC−D∼(−d−2)[∞] and hence ℓ(KC−D)=h0(C,O(−d−2)) equals −d−1 when −d−2≥0 (that is d≤−2) and 0 when d≥−1; the case d=−1 is the boundary m=−1<0, where the section space vanishes.

3.1F4F5step 2.1algebra

Taking the three ranges separately: for d≥0 the difference is (d+1)−0=d+1; for d=−1 it is 0−0=0=d+1; for d≤−2 it is 0−(−d−1)=d+1. In every case ℓ(D)−ℓ(KC−D)=d+1=deg⁡k(D)+1−g because g=0 by [F4], which is exactly the Riemann-Roch identity of [F5].

3.2F5step 2.1

The index of speciality of [F5] is i(D)=ℓ(KC−D), which is −d−1>0 precisely for d≤−2 and 0 for d≥−1; hence the special divisors of Pk1 are exactly the divisors of degree at most −2, and all divisors of degree at least −1 are nonspecial.

4.1F6F7F1F4F5step 3.2∎

The Axiom of Choice is used through the divisor dictionary and the full Riemann-Roch supplier; [F7] supplies the Dependent Choice premise required by the Cartier-to-Weil part of that dictionary.

ExampleConstruction: Literature-sourcedVerification: AI-adaptedprecheck passaudited 2026-10-02Open item page →

A degree-n line bundle on a genus-one curve has an n-dimensional space of sections for n > 0

Example

Assume the Axiom of Choice; it supplies Dependent Choice through AC implies DC implies countable choice. Let C be a smooth proper geometrically integral curve of genus one over a field k, and let L be an invertible sheaf of degree n≥1, with associated divisor D under the divisor--invertible-sheaf dictionary.

Since n≥1>0=2g−2, the line bundle L is nonspecial: H1(C,L)=0,ℓ(D)=h0(C,L)=deg⁡kD+1−g=n. Equivalently, since ωC≅OC for a genus-one curve, Serre duality reads h1(C,L)=h0(C,ωC⊗L−1)=h0(C,L−1)=0, because deg⁡L−1=−n<0 forces the vanishing of the sections of the negative-degree dual.

For n=1 the space of sections is one-dimensional. Under an isomorphism L≅OC(D), a nonzero section corresponds to a nonzero rational function f∈L(D), and its zero divisor is E=D+div⁡(f). This is effective by the definition of L(D) and has degree 1, since principal divisors have degree zero. Thus E=[p] for a closed point p with [κ(p):k]=1, so p is k-rational. The same argument applies to every degree-one divisor D; its complete linear system has the single effective member [p]. The dimension formula dim⁡∣D∣=deg⁡kD−g+i(D)=n−1+0 recovers this for n=1 and shows that the complete linear system grows by exactly one dimension for each added degree.

Facts & Assumptions

Given: the Axiom of Choice and its consequence Dependent Choice; a smooth proper geometrically integral curve C of genus one over a field k, an invertible sheaf L of degree n≥1, and the associated divisor D.

[F1]

For an invertible sheaf of degree >2g−2 on a smooth proper geometrically integral curve of genus g, H1(C,L)=0 and h0(C,L)=deg⁡L+1−g; equivalently ℓ(D)=deg⁡kD+1−g for divisors of degree >2g−2. (H^1 of a line bundle vanishes above degree 2g - 2, Riemann-Roch in exact form for divisors of degree above 2g - 2)

[F2]

The full Riemann-Roch theorem reads ℓ(D)−ℓ(KC−D)=deg⁡kD+1−g; the line bundle associated with D has h0(C,O(D))=ℓ(D), and the degree of L−1 is −deg⁡L. (The full Riemann-Roch theorem for divisors on a smooth proper curve, The Riemann-Roch dimension l(D), Invertible sheaf of cartier divisor, Cartier and Weil divisors agree on a smooth curve)

[F3]

On a genus-one curve ωC≅OC, and Serre duality for invertible sheaves gives h1(C,L)=h0(C,ωC⊗L−1); a line bundle of degree <0 has no nonzero global section. (The canonical bundle of a genus-one curve is trivial, Serre duality for line bundles on a smooth proper curve, and the residue realization, Negative-degree line bundles have no nonzero sections)

[F4]

The dimension of the complete linear system of a divisor is dim⁡∣D∣=ℓ(D)−1=deg⁡kD−g+i(D), where i(D)=ℓ(KC−D) is the index of speciality. Under an isomorphism L≅OC(D), a nonzero global section corresponds to a nonzero f∈L(D); its zero divisor is the effective divisor D+div⁡(f), of degree deg⁡kD because principal divisors have degree zero. (The dimension of a complete linear system, Complete linear system, The space L(D), Degree divisor proper curve, Rational sections of line bundles are Cartier divisors, Principal divisors on a normal proper curve have degree zero)

[F5]

The Axiom of Choice: every family of nonempty sets has a choice function. (The Axiom of Choice)

[F6]

In ZF, the Axiom of Choice implies Dependent Choice; this supplies the Dependent Choice premise of the Cartier-to-Weil dictionary used in [F2] and [F4]. (AC implies DC implies countable choice, The axiom of dependent choice: a relation in which every element is related to something admits an N-indexed chain)

Verification

Proof technique: apply the nonspecial Riemann-Roch formula and check the degree-one case separately, with Serre duality as an independent computation of H1.

1.1F1

Since g=1 we have 2g−2=0, and n≥1 gives deg⁡L=n>0; by [F1] applied to L, H1(C,L)=0 and ℓ(D)=h0(C,L)=n+1−1=n.

2.1F3step 1.1

Independently, [F3] gives h1(C,L)=h0(C,L−1) up to the identification ωC≅OC, and deg⁡L−1=−n<0, so [F3] again gives h0(C,L−1)=0; this agrees with Step 1.1 and confirms that L is nonspecial.

2.2F4F5step 1.1

For n=1, Step 1.1 gives ℓ(D)=1. Choose a nonzero section s∈H0(C,L) and an isomorphism L≅OC(D); the section corresponds to a nonzero f∈L(D), and [F4] gives the effective zero divisor E=D+div⁡(f). By [F4] and the degree-zero theorem for principal divisors, deg⁡k(E)=deg⁡k(D)=1. Hence E=[p] for a closed point with residue degree one, so p is k-rational and D∼[p]. Since ℓ(D)=1, the complete linear system has exactly the single effective member [p]. This argument applies to every degree-one divisor.

3.1F4step 1.1step 2.2algebra

The dimension formula of [F4] gives dim⁡∣D∣=deg⁡kD−g+i(D)=n−1+0=n−1, with i(D)=h1(C,L)=0 by Step 1.1; for n=1 this says dim⁡∣D∣=0 in agreement with Step 2.2, and each increase of the degree by one increases dim⁡∣D∣ by exactly one.

4.1F5F6F2F3step 3.1∎

The Axiom of Choice is used through the duality, degree, and divisor suppliers; [F6] supplies the Dependent Choice premise required by the Cartier-to-Weil divisor dictionary.

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

Adjunction on a smooth plane cubic: the canonical bundle is trivial

Example

Assume the Axiom of Choice; it supplies Dependent Choice through AC implies DC implies countable choice. Let k be a field and let C=V+(F)⊆Pk2 be a smooth plane cubic: a smooth projective plane curve of degree d=3, so C is a smooth projective geometrically integral curve over k.

Adjunction for smooth plane curves gives ωC≅OC(d−3)=OC(0)=OC, and the genus formula gives g(C)=(d−1)(d−2)/2=1, consistently with deg⁡kωC=d(d−3)=0=2g−2. Since h0(C,ωC)=g=1 and ωC has degree 0 with a nonzero global section, ωC is trivial; equivalently, every canonical divisor of C is a principal divisor, and the canonical class is the zero element of Pic⁡0(C).

The complete canonical linear system therefore has dimension 0 and its associated canonical morphism has target Pk0. Thus the complete canonical linear system has no positive-dimensional projective target. This is the exceptional case of the genus-one behaviour: the degree-three line bundle OC(3p0) at a k-rational point p0 is what embeds such a curve as a plane cubic, conversely to the computation above.

Facts & Assumptions

Given: the Axiom of Choice and its consequence Dependent Choice; a field k and a smooth plane cubic C=V+(F)⊆Pk2 of degree d=3.

[F1]

For a smooth plane curve C=V+(F)⊆Pk2 of degree d, adjunction gives ωC≅OC(d−3), and the genus is (d−1)(d−2)/2; the degree of the degree-d hypersurface is deg⁡F=d. (Adjunction for smooth plane curves, The genus of a smooth plane curve in terms of its degree, degree projective hypersurface)

[F2]

For a smooth proper geometrically integral curve of genus g, deg⁡kωC=2g−2 and h0(C,ωC)=g, with ωC=OC(KC) the canonical bundle. (The canonical divisor has degree 2g - 2, The canonical bundle has exactly g independent sections, Canonical bundle and canonical divisors, Degree divisor proper curve)

[F3]

An invertible sheaf of degree 0 on a smooth proper curve that has a nonzero global section is trivial; equivalently, an effective divisor of degree 0 is 0, so a degree-zero divisor whose sheaf has a section is principal. (A degree-zero line bundle with a nonzero section is trivial, Rational sections of line bundles are Cartier divisors, Invertible sheaves)

[F4]

A genus-one curve over k with a k-rational point p0 embeds as a smooth plane cubic via the degree-three very ample invertible sheaf OC(3p0); and on a genus-one curve the canonical bundle is trivial. (A genus-one curve with a rational point embeds as a plane cubic, The canonical bundle of a genus-one curve is trivial)

[F5]

The Axiom of Choice: every family of nonempty sets has a choice function. (The Axiom of Choice)

[F6]

In ZF, the Axiom of Choice implies Dependent Choice; this supplies the Dependent Choice premise of the Cartier-to-Weil dictionary used in [F3] and the cited genus-one and degree suppliers. (AC implies DC implies countable choice, The axiom of dependent choice: a relation in which every element is related to something admits an N-indexed chain)

Verification

Proof technique: specialize adjunction and the genus formula to d=3, then apply the degree-zero triviality criterion.

1.1F1

By [F1] with d=3, ωC≅OC(0)=OC and g(C)=(3−1)(3−2)/2=1.

2.1F2F3step 1.1

By [F2] with g=1, deg⁡kωC=0 and h0(C,ωC)=g=1; the unique one-dimensional space of sections is nonzero, so [F3] applies to the degree-zero invertible sheaf ωC and shows ωC≅OC; equivalently every canonical divisor is principal and the canonical class is 0 in Pic⁡0(C).

2.2F4step 1.1

Conversely, if C is a genus-one curve over k with a k-rational point p0, then OC(3p0) has degree 3=2g+1, so by [F4] it is very ample and embeds C as a plane cubic; thus every genus-one curve with a rational point has a smooth plane-cubic model. The forward calculation applies to every smooth plane cubic without assuming a rational point: its canonical class is zero in Pic⁡0. This proves the stated converse implication and does not imply that every smooth plane cubic has a rational point.

3.1F2F4step 2.1

Since h0(C,ωC)=1, the complete canonical linear system has dimension ℓ(KC)−1=0 and its associated complete canonical morphism has target Pk0; it supplies no positive-dimensional canonical target. This matches the direct computation ωC≅OC of Step 2.1 and the general genus-one statement [F4].

4.1F5F6F1F2F3F4step 2.2∎

The Axiom of Choice is used through the cohomology, degree, and divisor suppliers; [F6] supplies the Dependent Choice premise used by the Cartier-to-Weil divisor dictionary.

ExampleConstruction: Literature-sourcedVerification: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Adjunction on a smooth plane quartic: the canonical bundle is the hyperplane bundle

Example

Assume AC, as required by the cited canonical-map criterion. Let k be a field and let C=V+(F)⊆Pk2 be a smooth plane quartic, a smooth projective plane curve of degree d=4.

Adjunction gives ωC≅OC(d−3)=OC(1), the restriction to C of the hyperplane bundle OP2(1); the genus is g(C)=(4−1)(4−2)/2=3, consistently with deg⁡kωC=4⋅1=4=2g−2. Since h0(C,ωC)=g=3, the space of canonical sections has dimension three, so the complete canonical linear system has projective dimension two. The restriction argument below identifies it with the coordinate sections, which generate OC(1) at every point and hence define the canonical morphism to P2.

That map is the given inclusion C↪P2: the twisted hypersurface sequence and its low-degree cohomology sequence, together with H0(P2,O(−3))=H1(P2,O(−3))=0, show that the restriction map H0(P2,O(1))→H0(C,OC(1)) is an isomorphism; hence ∣KC∣=∣H∣ is the linear system of lines and the canonical image of C is the plane quartic itself. In particular C is geometrically nonhyperelliptic, and the canonical bundle of a plane quartic is the hyperplane bundle; the canonical model of this genus-three curve is the plane quartic.

Facts & Assumptions

Given: AC, a field k and a smooth plane quartic C=V+(F)⊆Pk2 of degree d=4.

[F1]

For a smooth plane curve C=V+(F)⊆Pk2 of degree d, ωC≅OC(d−3) and g(C)=(d−1)(d−2)/2; the degree of the hypersurface is deg⁡F=d. (Adjunction for smooth plane curves, The genus of a smooth plane curve in terms of its degree, degree projective hypersurface)

[F2]

For a smooth proper geometrically integral curve of genus g, deg⁡kωC=2g−2 and h0(C,ωC)=g, and, when the complete canonical system ∣KC∣ is base-point-free and its section space is nonzero, it defines the canonical morphism (The canonical divisor has degree 2g - 2, The canonical bundle has exactly g independent sections, Canonical bundle and canonical divisors, Complete linear system, A base-point-free linear system defines a morphism to projective space).

[F3]

On Pk2 one has H1(P2,O(m))=0 for every m, H0(P2,O(−3))=0, and H0(P2,O(1)) is the space of linear forms; the twisting sheaves are the ones attached to the standard graded presentation. (Cohomology of O(d) on projective space, Twisting sheaf on Proj)

[F4]

If i:C↪Pk2 is the plane quartic, the published hypersurface sequence gives 0→OP2(−4)→⋅FOP2→i∗OC→0: the nonzero dehomogenizations of F are nonzerodivisors in the polynomial domain rings of the standard charts. The standard graded polynomial ring generated in degree one makes OP2(1) invertible, so tensoring preserves exactness. On each standard chart the quotient by the local equation F with the restricted twist identifies the last term with i∗OC(1), yielding 0→OP2(−3)→⋅FOP2(1)→i∗OC(1)→0. (Hypersurface cohomology sequence, Invertible twists for degree-one generated rings, Relative projective space from standard charts, Closed immersions are affine quotients and survive base change, Direct image of a sheaf along a continuous map, Twisting sheaf on Proj)

[F5]

The short exact sequence in [F4] gives a long exact sequence in sheaf cohomology, and closed-immersion pushforward identifies Hq(P2,i∗OC(1)) with Hq(C,OC(1)). In particular its low-degree segment is 0→H0(P2,O(−3))→H0(P2,O(1))→H0(C,OC(1))→H1(P2,O(−3)). (Long exact sequence of sheaf cohomology, Closed immersion preserves cohomology and coherent pushforward)

[F6]

For a smooth proper geometrically integral curve of genus g≥2, the canonical map is a closed immersion exactly when g≥3 and the curve is geometrically nonhyperelliptic, meaning that its algebraic-closure base change has no degree-two map to P1. (The canonical map: base-point-freeness and the hyperelliptic exception, Hyperelliptic curves and hyperelliptic maps)

[F7]

For a divisor D on a smooth proper curve, the complete linear system ∣D∣ is the set of effective divisors linearly equivalent to D, and its dimension is ℓ(D)−1. (Complete linear system, Degree divisor proper curve)

Verification

Proof technique: specialize adjunction to d=4 and identify the canonical linear system with the linear system of lines via the restriction map.

1.1F1

By [F1] with d=4, ωC≅OC(1) and g(C)=(3)(2)/2=3.

2.1F2F3step 1.1

By [F2] and [F3], deg⁡kωC=4=2⋅3−2 and h0(C,ωC)=3; thus its complete canonical system has projective dimension two.

2.2F2F3F4F5step 1.1

The low-degree segment in [F5], together with the two vanishings in [F3], makes the restriction map H0(P2,O(1))→H0(C,OC(1)) an isomorphism. By Step 1.1, ωC≅OC(1), so the complete canonical system ∣KC∣=∣OC(1)∣ is exactly the linear system of lines cut on C by H0(P2,O(1)). The three restricted coordinate sections generate OC(1): at each point at least one coordinate is nonzero, and on that standard chart its section is a local frame. Hence the canonical system is base-point-free and its nonzero three-dimensional section space defines a morphism to P2 by [F2].

3.1F6step 2.2

Because ∣KC∣ is the restriction of the linear system of lines, the canonical map of C is the restriction of the inclusion C↪P2: it is a closed embedding, and its image is the plane quartic C itself. This embedding remains a closed embedding after extending k to kˉ, so [F6] shows that C is geometrically nonhyperelliptic. In particular it has no degree-two map to the split Pk1; its canonical model is the plane quartic.

4.1F2F7step 2.1∎

As a consistency check, the canonical divisor of C is cut by a line: deg⁡kKC=4 equals 2g−2 by Step 2.1, and the isomorphism ωC≅OC(1) of Step 1.1 is the statement that the canonical divisor class is the class of a hyperplane section, i.e. the hyperplane bundle.

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

The canonical map of a hyperelliptic curve is not an embedding

Statement refuted

Assume AC, as required by the cited canonical-map theorem. For a smooth proper geometrically integral curve C of genus g≥2, geometric hyperellipticity prevents the canonical map from being a closed immersion. After extending to an algebraic closure, the canonical map factors through a degree-two map to P1 and the (g−1)-fold Veronese embedding. It has generic degree two onto a rational normal curve; its generic geometric fiber has two distinct points, which the canonical map identifies. This does not assert that every closed fiber has two distinct points. If the degree-two map is defined over k with target Pk1, the same factorization holds over k.

Facts & Assumptions

Given: AC, a field k, a smooth proper geometrically integral curve C/k of genus g≥2, and, after base extension to kˉ, a degree-two map φ:Ckˉ→Pkˉ1. In the split case the map may already be given over k.

[F1]

The degree-two map is finite and surjective, and L=φ∗OP1(1) has degree two. (Hyperelliptic curves and hyperelliptic maps, Degree of a nonconstant morphism of curves)

[F2]

The canonical bundle and canonical map commute with field extension. The canonical bundle is generated for g≥2, and the degree-two map gives L⊗(g−1)≅ωCkˉ; the canonical map is the composition of φ with the (g−1)-fold Veronese map. (The canonical map: base-point-freeness and the hyperelliptic exception, A base-point-free linear system defines a morphism to projective space)

[F3]

h0(C,ωC)=g and deg⁡ωC=2g−2. (The canonical bundle has exactly g independent sections, The canonical divisor has degree 2g - 2)

[F4]

The Veronese map is a closed immersion. The actual target Pkˉ1 is reduced, since its standard affine charts have polynomial-domain coordinate rings. A closed immersion into this target that is surjective on points is an isomorphism: on each affine chart its defining ideal lies in every prime, hence in the nilradical, which is zero. (The Veronese map is a well-defined closed immersion, Closed immersions of schemes, Relative projective space from standard charts)

[F5]

Smooth proper birational curves over a field are isomorphic. (Birational smooth proper curves are isomorphic)

Counterexample

Let C/k be as in the given data. By [F2], after base extension the canonical map is φK,kˉ=vg−1∘φ, where vg−1:Pkˉ1→Pkˉg−1 is the Veronese closed immersion. Its image is a rational normal curve, and the composite has generic degree two. If the composite were a closed immersion, then φ would be a closed immersion into the Veronese image: the surjection on coordinate rings for the composite factors through the coordinate ring of that image, which is isomorphic to the reduced scheme Pkˉ1. Since φ is also finite and surjective, [F4] would make it an isomorphism, contradicting its degree two. Thus the canonical map is not a closed immersion. A k-map that became a closed immersion would remain one after base extension, so the same conclusion holds over k.

Proof technique: use the canonical Veronese factorization. Separability identifies the generic geometric fiber as two distinct points; it is not needed for the non-embedding argument.

1.1F1F2F4

(Veronese factorization and nonembedding.) By [F1] and [F2], L has degree two, L⊗(g−1)≅ωCkˉ, and the canonical map factors through vg−1∘φ. Its generic degree onto the rational normal image is two. If this composite were a closed immersion, the coordinate-ring surjections would make φ a closed immersion into its Veronese image, which is isomorphic to the reduced scheme Pkˉ1. The map φ is finite and surjective by [F1]; [F4] then makes it an isomorphism, contradicting degree two. This proves the nonembedding without assuming separability.

1.2F1F2F5

(Generic geometric fiber.) To describe the generic geometric fiber, work over Ω=kˉ. In characteristic different from two, a degree-two extension Ω(C)/Ω(t) is separable. In characteristic two, a degree-two extension is either separable or purely inseparable. Suppose it were purely inseparable. Choose a generator α with α2=h(t)∈Ω(t), and write h(t)=P(t)/Q(t) with P,Q∈Ω[t], Q≠0. Since Ω is perfect, choose P0,Q0∈Ω[z] such that P(z2)=P0(z)2 and Q(z2)=Q0(z)2. The embedding Ω(t)↪Ω(z), t↦z2, extends to Ω(C) by sending α to P0(z)/Q0(z). Indeed, this element squares to h(z2), and X2−h(t) is irreducible because the extension is purely inseparable of degree two. The resulting field embedding has image of degree two over Ω(z2); since [Ω(z):Ω(z2)]=2, its image is all of Ω(z). Thus Ckˉ is birational to Pkˉ1. Smooth proper birational curves are isomorphic, contradicting g≥2 by [F5]. The degree-two map is therefore separable in characteristic two as well. Its generic geometric fiber consists of two distinct points, and the Veronese factorization identifies them under the canonical map. Special fibers may be ramified and need not have two distinct points.

2.1F2F3step 1.1

(Numerical canonical data.) The canonical space has dimension g, the canonical bundle has degree 2g−2, and it is globally generated. Since the canonical map is not a closed immersion by step 1.1, base-point-freeness and these numerical data alone do not imply the embedding conclusion.

2.2F2step 1.1∎

(Genus two.) For g=2, the Veronese map in the factorization is the identity of P1. Thus after base extension the canonical map is the degree-two map φ itself, not an embedding.

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

Degree 2g-1 does not force base-point-freeness

Statement refuted

Assume the Axiom of Choice; it supplies Dependent Choice through AC implies DC implies countable choice. The theorem that a line bundle of degree at least 2g on a curve of genus g is base-point-free is sharp in its degree bound: a line bundle of degree 2g−1 need not be base-point-free.

Facts & Assumptions

Given: the Axiom of Choice and its consequence Dependent Choice; a field k, a smooth proper geometrically integral curve C of genus g≥1 over k with a k-rational point p, and the invertible sheaf L=OC(KC+p) with associated divisor D=KC+p.

[F1]

deg⁡kKC=2g−2, so deg⁡kD=2g−1; and the divisor--invertible-sheaf dictionary identifies L(−p)=OC(KC) and H0(C,L(−p))⊆H0(C,L). (The canonical divisor has degree 2g - 2, Canonical bundle and canonical divisors, Degree divisor proper curve, Invertible sheaf of cartier divisor, Cartier and Weil divisors agree on a smooth curve)

[F2]

For a divisor of degree >2g−2 on a curve of genus g the nonspecial formula gives ℓ(D)=deg⁡kD+1−g and H1=0; equivalently L has h0(C,L)=deg⁡L+1−g. (Riemann-Roch in exact form for divisors of degree above 2g - 2, H^1 of a line bundle vanishes above degree 2g - 2, The full Riemann-Roch theorem for divisors on a smooth proper curve, The Riemann-Roch dimension l(D))

[F3]

ℓ(KC)=h0(C,ωC)=g, where ωC=OC(KC) is the canonical bundle; the Riemann-Roch space of KC is the space of holomorphic differentials. (The canonical bundle has exactly g independent sections, The space L(D))

[F4]

A closed point q is a base point of the complete linear system ∣D∣ of a divisor D exactly when every global section of O(D) vanishes at q, equivalently H0(C,O(D−q))=H0(C,O(D)); the sheaf is base-point-free when no closed point is a base point. (Base points and base-point-free linear systems)

[F5]

A line bundle of degree at least 2g on a curve of genus g is base-point-free; this is the statement whose degree bound is tested here. (Line bundles of degree at least 2g are base-point-free)

[F6]

On a genus-one curve, the canonical bundle is trivial and every canonical divisor is principal; hence KC∼0 and OC(KC+p)≅OC(p). The divisor-line-bundle dictionary identifies this linear equivalence with the corresponding isomorphism of invertible sheaves. (The canonical bundle of a genus-one curve is trivial, Rational sections of line bundles are Cartier divisors, Cartier and Weil divisors agree on a smooth curve)

[F7]

The Axiom of Choice: every family of nonempty sets has a choice function. (The Axiom of Choice)

[F8]

In ZF, the Axiom of Choice implies Dependent Choice; this supplies the Dependent Choice premise of the Cartier-to-Weil divisor dictionary used in [F1] and [F6]. (AC implies DC implies countable choice, The axiom of dependent choice: a relation in which every element is related to something admits an N-indexed chain)

Counterexample

Assume the Axiom of Choice; it supplies Dependent Choice through AC implies DC implies countable choice. Let k be a field, let C be a smooth proper geometrically integral curve of genus g≥1 over k with a k-rational point p, let KC be a canonical divisor and put L=OC(KC+p),D=KC+p. Then deg⁡kD=2g−2+1=2g−1, and since 2g−1>2g−2 the nonspecial formula gives ℓ(D)=deg⁡kD+1−g=g,H1(C,L)=0. On the other hand L(−p)=OC(KC) has ℓ(KC)=h0(C,ωC)=g by the canonical-sections computation, and H0(C,L(−p))⊆H0(C,L) is an inclusion of spaces of the same dimension g; hence the two spaces are equal and every global section of L vanishes at p. Therefore p is a base point of the complete linear system ∣D∣=∣L∣, and L is not base-point-free even though its degree is the largest value below the safe bound 2g of the base-point-freeness theorem. For g=1, [F6] gives KC∼0, so L≅OC(p); its unique section has zero divisor [p] and vanishes at p.

Proof technique: compute both section spaces and observe that the evaluation at p has no room to be nonzero.

1.1F1F2

By [F1], deg⁡kD=2g−1>2g−2, so [F2] applies and gives ℓ(D)=g together with H1(C,L)=0.

2.1F1F3step 1.1

Since L(−p)=OC(KC) by [F1] and ℓ(KC)=g by [F3], the space H0(C,L(−p)) has dimension g; by the inclusion of [F1] and Step 1.1, H0(C,L(−p))⊆H0(C,L) is an inclusion of k-vector spaces of the same dimension g, hence an equality.

3.1F4step 2.1

The equality of Step 2.1 says exactly that every global section of L vanishes at the k-rational point p; by the base-point criterion [F4], p is a base point of the complete linear system ∣D∣ and L=OC(KC+p) is not base-point-free.

4.1F5step 3.1

Since deg⁡L=2g−1<2g, this counterexample shows that the degree bound in the base-point-freeness theorem [F5] cannot be lowered from 2g to 2g−1: the bound is sharp.

5.1F1F2F6F7F8step 1.1step 3.1∎

For g=1, [F6] gives KC∼0 (the chosen representative need not equal the zero divisor), so L≅OC(p). Step 1.1 gives h0(C,L)=1, and this isomorphism makes H0(C,OC(p)) one-dimensional. Its canonical section has zero divisor [p], so the unique one-dimensional section space vanishes at p, in agreement with Step 3.1. The Axiom of Choice is used through the degree, duality, and divisor suppliers, and [F8] supplies the Dependent Choice premise of the Cartier-to-Weil route.

CounterexampleConstruction: Literature-sourcedVerification: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Degree 2g does not force very ampleness

Statement refuted

Assume the Axiom of Choice; it supplies Dependent Choice through AC implies DC implies countable choice. The theorem that every line bundle of degree at least 2g+1 on a curve of genus g is very ample is sharp over any field k when the curve has a k-rational point: for a smooth proper geometrically integral curve C/k of genus g≥1 with p∈C(k), the line bundle OC(KC+2p) has degree 2g and need not be very ample. When k is algebraically closed, every closed point is k-rational, so this includes the algebraically closed-field case.

Facts & Assumptions

Given: the Axiom of Choice and its consequence Dependent Choice; a field k, a smooth proper geometrically integral curve C of genus g≥1 over k, a k-rational point p∈C(k), a canonical divisor KC, and the invertible sheaf L=OC(KC+2p).

[F1]

On an integral proper curve over a field, divisors are finite integral sums of closed points and deg⁡kD=∑xnx[κ(x):k]; since the given point p is k-rational, deg⁡k(p)=1. Moreover deg⁡k(KC)=2g−2 for every choice of the nonzero rational differential defining KC, the sheaf ωC=ΩC/k1 satisfies ωC≅OC(KC), and under the divisor--invertible-sheaf dictionary L(−p)=OC(KC+p) and L(−2p)=OC(KC), with the corresponding inclusions of spaces of global sections. (Degree divisor proper curve, The canonical divisor has degree 2g - 2, Canonical bundle and canonical divisors, Invertible sheaf of cartier divisor, Cartier and Weil divisors agree on a smooth curve)

[F2]

For a divisor D of degree >2g−2 on a smooth proper geometrically integral curve of genus g the nonspecial formula holds: ℓ(D)=deg⁡kD+1−g and H1(C,OC(D))=0, equivalently h0(C,OC(D))=deg⁡kD+1−g. (Riemann-Roch in exact form for divisors of degree above 2g - 2, H^1 of a line bundle vanishes above degree 2g - 2, The full Riemann-Roch theorem for divisors on a smooth proper curve, The Riemann-Roch dimension l(D))

[F3]

ℓ(KC)=h0(C,ωC)=g for any canonical divisor KC. (The canonical bundle has exactly g independent sections, The Riemann-Roch dimension l(D))

[F4]

A closed point q is a base point of the complete linear system ∣D∣ of a divisor D exactly when H0(C,OC(D−q))=H0(C,OC(D)), equivalently when every global section of OC(D) vanishes at q; D is base-point-free when it has no base point. A base-point-free subspace V⊆L(D) of dimension r+1≥1 defines a morphism φV:C→Pkr with φV∗O(1)≅OC(D); its system members are the pullbacks of hyperplanes. Conversely, if a morphism i:C→Pkn is equipped with an isomorphism i∗O(1)≅OC(D), its pulled-back coordinate sections form an ordered generating tuple, and the projective-data theorem reconstructs i from that tuple. Their span is a base-point-free subspace of L(D), but may have dimension less than n+1; equality with the basis morphism from the span up to a projective-linear change applies when the pulled-back coordinates are linearly independent. (Base points and base-point-free linear systems, Complete linear system, A base-point-free linear system defines a morphism to projective space, Maps to projective space equal generating line-bundle data)

[F5]

An invertible OC-module L is closed H-very ample relative to Spec⁡k when there is a closed immersion i:C→Pkn with i∗O(1)≅L; H-very ample relative to Spec⁡k means such an immersion exists that is quasi-compact. A locally closed immersion has injective differential at every point of its source (for a closed immersion this is the injectivity of the induced maps of Zariski tangent spaces, and an open immersion is an isomorphism onto its image); in particular an immersion i:C→Pkn has dix≠0 at every closed point x of C, because the tangent space TxC of the smooth curve C is one-dimensional. (Relative very ampleness in the finite projective-space convention)

[F6]

The very-ampleness theorem whose degree bound is tested here: if deg⁡kL≥2g+1, then L is closed H-very ample relative to Spec⁡k and the associated morphism φL is a closed immersion. (Line bundles of degree at least 2g+1 are very ample)

[F7]

The Axiom of Choice: every family of nonempty sets has a choice function. (The Axiom of Choice)

[F8]

In ZF, the Axiom of Choice implies Dependent Choice; this supplies the Dependent Choice premise of the Cartier-to-Weil divisor dictionary used in [F1] and the cited divisor suppliers. (AC implies DC implies countable choice, The axiom of dependent choice: a relation in which every element is related to something admits an N-indexed chain)

[F9]

On a genus-one curve, every canonical divisor satisfies KC∼0 by The canonical bundle of a genus-one curve is trivial. A nonconstant morphism of smooth proper curves is finite, and the degree of a pulled-back line bundle is the map degree times its degree; on Pk1, deg⁡O(1)=1. Also H0(C,OC)=k, so a constant k-morphism from C to Pk1 has image a k-rational point and pulls O(1) back to a trivial line bundle. (Degree of a nonconstant morphism of curves, Fibres, pullbacks and degrees of divisors under a finite morphism of curves, Divisors on the projective line are classified by degree, Functions on a proper curve)

[F10]

Assuming AC as inherited from the duality suppliers, if C is a smooth proper geometrically integral curve of genus g over any field k and an invertible sheaf E has degree at least 2g, then E is base-point-free and its complete linear system defines a k-morphism to Pkh0(C,E)−1 pulling back O(1) to E. (Line bundles of degree at least 2g are base-point-free)

Counterexample

Proof technique: compute the image of the evaluation map on two-jets at p; it is one-dimensional, so no immersion with pullback L can be nonzero on tangent spaces at p.

1.1F1

Since p is k-rational, deg⁡k(p)=1, so deg⁡k(KC+2p)=(2g−2)+2=2g and deg⁡k(KC+p)=2g−1; moreover L(−p)=OC(KC+p) and L(−2p)=OC(KC)=ωC as subsheaves of the sheaf of rational sections.

1.2F4

Let i:C→Pkn be any k-morphism with an isomorphism α:i∗O(1)→L. Put tj=α(i∗xj) for the projective coordinates x0,…,xn. The projective-data theorem in [F4] says that these sections form an ordered generating tuple and reconstruct i from that tuple. Their span Wi⊆H0(C,L) is base-point-free, but its dimension may be less than n+1 if the coordinate sections are linearly dependent.

1.3F7F8F1F2F4F6

The Axiom of Choice is used through the degree, duality, linear-system, and projective-data suppliers; [F8] supplies Dependent Choice for the Cartier-to-Weil divisor route.

2.1F4algebra

Local computation at p: since the tuple in Step 1.2 generates L, one of its coordinate sections is nonzero at p; after renumbering coordinates call it t0. Use t0 as a local frame to identify Lp/mp2Lp with OC,p/mp2, where the jet of t0 is the class of 1. On the affine chart of Pkn where the corresponding coordinate is nonzero, i is given by the ratios tj/t0; therefore dip=0 exactly when every ratio has zero differential, equivalently when the jet of each tj lies in the line k⋅[t0]. This says precisely that the image of Wi in Lp/mp2Lp is one-dimensional.

2.2F2step 1.1

Since deg⁡kL=2g>2g−2, the divisor D=KC+2p is nonspecial and [F2] gives h0(C,L)=ℓ(D)=deg⁡kD+1−g=g+1 together with H1(C,L)=0.

2.3F1F2F3step 1.1

By Step 1.1, L(−2p)=OC(KC)=ωC and L(−p)=OC(KC+p); hence h0(C,L(−2p))=h0(C,ωC)=g by [F3], and h0(C,L(−p))=g by [F2] applied to the divisor KC+p of degree 2g−1>2g−2.

2.4F10step 1.1

Since deg⁡kL=2g, theorem [F10] applies over the given arbitrary field k and shows that L is base-point-free. Thus the complete linear system ∣L∣ is defined over k.

3.1F2step 2.2step 2.3algebra

Let A=OC,p and m=(t), where t is a uniformizer; choose a local frame e of Lp. Since A/m=κ(p)=k and m/m2 is one-dimensional over k, the exact sequence 0→m/m2→A/m2→A/m→0 gives dim⁡k(A/m2)=2. Hence Lp/m2Lp≅e(A/m2) is two-dimensional, with value and first-order classes represented by e and te. The kernel of ev:H0(C,L)→Lp/m2Lp consists of sections whose stalk at p lies in m2Lp; since L(−2p) has stalk m2Lp at p and agrees with L away from p, this kernel is exactly H0(C,L(−2p)). Therefore dim⁡kim⁡ev=h0(C,L)−h0(C,L(−2p))=(g+1)−g=1.

3.2F4step 2.2step 2.4

By Step 2.4 the space V=H0(C,L) is base-point-free of dimension g+1, so [F4] provides the morphism φL:C→Pkg of the complete linear system, with φL∗O(1)≅L; its target is Pkh0(C,L)−1=Pkg.

4.1step 3.1step 1.2step 2.1

For the arbitrary morphism i of Step 1.2, Wi⊆H0(C,L), so its image in Lp/mp2Lp is contained in the one-dimensional image of H0(C,L) from Step 3.1. It contains the nonzero jet of t0 from Step 2.1, so its image is exactly one-dimensional and Step 2.1 gives dip=0.

5.1F5step 4.1

By [F5] an immersion has injective differential at every point and the tangent space TpC is one-dimensional, so dip=0 means that i is not an immersion at p; Steps 1.2--4.1 apply to every k-morphism i:C→Pkn with i∗O(1)≅L and every n. Thus no such morphism is a closed or locally closed immersion, and L is not closed H-very ample relative to Spec⁡k.

6.1F4step 3.2step 5.1

The morphism φL of Step 3.2 is one of the morphisms covered by Step 5.1, so dφL=0 at p and φL is not a closed immersion; its differential vanishes at the point p of the base-point-free system ∣L∣.

7.1F1F3F4F9step 2.2step 6.1

For g=1, [F9] gives KC∼0 (the chosen representative need not equal the zero divisor), hence L≅OC(2p). Step 2.2 gives h0(C,L)=2, so the complete linear system morphism has target Pk1 and pulls back O(1) to L. It is nonconstant because a constant map to a k-rational point pulls O(1) back to a trivial line bundle, whereas deg⁡kL=2. By [F9], it is finite and deg⁡kL=deg⁡(φL)deg⁡OP1(1)=deg⁡(φL), so it is a morphism of degree two. The divisors in the pencil ∣2p∣ are pullbacks of k-rational points of Pk1 and have degree two; a closed point q pulls back to degree 2[κ(q):k] by [F9].

8.1F6step 5.1step 7.1∎

In summary L has deg⁡kL=2g, is base-point-free by Step 2.4, and is not closed H-very ample relative to Spec⁡k by Step 5.1; since [F6] gives closed H-very ampleness for every line bundle of degree at least 2g+1, the hypothesis deg⁡≥2g+1 cannot be weakened to deg⁡≥2g and the bound 2g+1 is sharp.

ExampleConstruction: Literature-sourcedVerification: AI-adaptedprecheck passaudited 2026-10-02Open item page →

Riemann-Hurwitz for a tame double cover with 2r branch points

Example

Let k be a field of characteristic not two and let f:C→Pk1 be a finite surjective morphism of degree n=2 between smooth proper geometrically integral curves over k (Curves over a field) that is tamely ramified with branch locus exactly 2r distinct k-rational points of Pk1, the ramification index being ep=2 at each of the points p of C lying over the branch points. Because the residue characteristic is not two, every residue extension of degree at most two is separable and every ramification index that occurs is invertible, so the tameness hypothesis is automatic here.

The fibre over a branch point qi is computed by the pullback-degree identity f∗[qi]=∑pep[p]: since deg⁡kf∗[qi]=2 and some p over qi has ep=2, the branch point carries a single point pi with epi=2 and residue degree 1, so κ(pi)=k. Over a non-branch point every point is unramified, and with the tame different formula ℓp=ep−1 the different divisor is Rf=∑i=12r[pi],deg⁡kRf=2r. With g(Pk1)=0 the complete Riemann-Hurwitz formula 2g(C)−2=n(2g(Pk1)−2)+deg⁡kRf (The Riemann-Hurwitz formula with the different) reads 2g(C)−2=2(−2)+2r, that is g(C)=r−1.

Such a cover is realized concretely by the smooth projective model Ch of y2=h(x) with h∈k[x] monic and squarefree of degree 2r over a perfect field k of characteristic not two (Perfect fields: every irreducible polynomial is separable, Smooth proper curves, dominant morphisms and function fields, Every smooth proper curve admits a projective embedding). Its two affine charts and their overlap are constructed in Verification, step 1.4; this makes the projection π:Ch→Pk1 and the relative differential module explicit. A point p is ramified exactly when y vanishes at p, in which case ep=2, κ(p) is the residue field of the corresponding root, and ℓp=1. The roots of h number 2r counted with their residue degrees, so deg⁡kRπ=deg⁡h=2r and again g(Ch)=r−1. The infinity chart has two k-rational points over ∞, both unramified. When h splits over k with distinct roots c1,…,c2r∈k, the branch locus is exactly the 2r distinct k-rational points c1,…,c2r and the model is an example of the abstract cover above.

The small cases confirm the formula: r=1 gives g=0, as for y2=x2−1, whose smooth projective conic has a rational point and is a projective line by A genus-zero curve with a degree-one divisor is the projective line; r=2 gives g=1, the genus-one quartic family y2=(x2−1)(x2−c) with c∈k∖{0,1}; and r=3 gives g=2, as for a squarefree sextic. In general the genus of the model of y2=h(x) is r−1.

The proof explicitly assumes the Axiom of Choice. By [F19], it supplies the Dependent Choice required by the Cartier-to-Weil cycle argument in the divisor suppliers; the other Choice-bearing uses are through (Smooth proper curves, dominant morphisms and function fields, Fibres, pullbacks and degrees of divisors under a finite morphism of curves, Local support and index bound for the different of a curve map, Divisors on the projective line are classified by degree, Cartier and Weil divisors agree on a smooth curve, The Riemann-Hurwitz formula with the different, Local rings at closed points of smooth curves are discrete valuation rings, Relative Jacobian criterion with its presentation hypothesis, Composite of a finite morphism and a proper morphism is proper, Modules over a field are projective, flat, and injective, Every smooth proper curve admits a projective embedding, and A genus-zero curve with a degree-one divisor is the projective line. The computations below make no further choice.

Facts & Assumptions

Given: the Axiom of Choice; a field k of characteristic ≠2, an integer r≥1, and either (i) a finite surjective degree-two morphism f:C→Pk1 of smooth proper geometrically integral curves over k, tamely ramified with branch locus exactly 2r distinct k-rational points and ramification index 2 at each point of C over them, or (ii) a perfect such field k together with a monic squarefree polynomial h∈k[x] of degree 2r, whose model Ch with projection π:Ch→Pk1 is constructed below.

[F1]

A nonconstant morphism f:C→D of smooth proper geometrically integral curves over k has degree deg⁡(f)=[k(C):k(D)], a positive integer; the field extension k(C)/k(D) has degree two in our setting, and every extension of degree at most two in characteristic ≠2 is separable, since an irreducible quadratic has nonzero derivative when 2≠0. (Degree of a nonconstant morphism of curves)

[F2]

For a nonconstant morphism of smooth proper geometrically integral curves the index-ramification locus is {p:ep>1} and its image is the index branch locus; f is unramified at p exactly when ΩC/D,p=0; a nonconstant morphism of smooth proper curves is finite and surjective. (Ramification points, branch points and unramifiedness)

[F3]

The ramification index at p over q is ep=ord⁡p(f∗tq) for a uniformizer tq of OD,q, and it is independent of the uniformizer. (Ramification index of a morphism of curves)

[F4]

For a finite surjective morphism of smooth proper geometrically integral curves with separable function-field extension, the different length ℓp=length⁡OC,p(ΩC/D,p) satisfies ℓp≥ep−1; moreover ℓp=0 if and only if ep=1 and κ(p)/κ(q) is separable, and ℓp=ep−1 if and only if κ(p)/κ(q) is separable and ep is invertible in κ(q) (tame ramification); in particular Supp⁡(ΩC/D) consists exactly of the points with ep>1 or inseparable residue extension. (Local support and index bound for the different of a curve map, Ramification points, branch points and unramifiedness)

[F5]

The different divisor is the effective divisor Rf=∑pℓp[p] on C; its support is the differential-ramification locus. (The different divisor of a generically separable morphism of curves)

[F6]

For a nonconstant morphism f:C→D of degree n of smooth proper geometrically integral curves, f is finite and flat, the pullback of Cartier divisors is defined and additive, f∗[q]=∑p∈f−1(q)ep[p] for closed points, and deg⁡k(f∗E)=ndeg⁡kE for every divisor E on D; consequently ∑p∈f−1(q)epfp=n for every closed point q, where fp=[κ(p):κ(q)] is the residue degree. (Fibres, pullbacks and degrees of divisors under a finite morphism of curves, Ramification index of a morphism of curves, Degree divisor proper curve)

[F7]

For an integral proper curve over k, a divisor is a finite integral sum D=∑xnx[x] of closed points and deg⁡kD=∑xnx[κ(x):k]. The degree of a principal divisor is zero, so linearly equivalent divisors have equal degree, and the divisor of a rational function z on a smooth proper curve has deg⁡kdiv⁡(z)=0. (Degree divisor proper curve, Principal divisors on a normal proper curve have degree zero)

[F8]

Pk1 is a smooth proper geometrically integral curve over k of genus 0, with affine coordinate t=x1/x0 and the point ∞=[0:1]; every divisor on Pk1 is linearly equivalent to deg⁡k(D)[∞]. (Divisors on the projective line are classified by degree, Genus via the Euler characteristic)

[F9]

At every closed point of a smooth curve the local ring is a DVR, so the order of a rational function there is defined and additive; the local ring at the generic point is the function field, hence a field, not a DVR. Every invertible sheaf on a smooth proper curve is OC(D) for a divisor D well defined modulo linear equivalence. (Local rings at closed points of smooth curves are discrete valuation rings, Cartier and Weil divisors agree on a smooth curve)

[F10]

The genus of a smooth proper geometrically integral curve is g(C)=h1(C,OC)=1−χ(C,OC). (Genus via the Euler characteristic)

[F11]

Riemann-Hurwitz: for a finite surjective morphism f:C→D of smooth proper geometrically integral curves with separable function-field extension and different divisor Rf, one has 2g(C)−2=n(2g(D)−2)+deg⁡kRf with n=deg⁡(f), equivalently KC∼f∗KD+Rf. (The Riemann-Hurwitz formula with the different)

[F12]

Let K/k be a finitely generated field extension of transcendence degree one in which k is relatively algebraically closed and let k be perfect. Then there is a smooth proper geometrically integral curve C over k together with a fixed k-isomorphism K≅k(C); any two such identified models are related by a unique k-isomorphism inducing the prescribed function-field identification; and for smooth proper geometrically integral curves C,D over k the assignment f↦f∗ is a bijection from dominant k-morphisms C→D onto injective k-algebra homomorphisms k(D)↪k(C). (Smooth proper curves, dominant morphisms and function fields)

[F13]

A curve over k is a geometrically integral, separated, finite-type k-scheme of chain dimension one; smooth and proper are extra adjectives. (Curves over a field)

[F14]

For a finitely generated extension K/k, saying that k is relatively algebraically closed in K means that every element of K algebraic over k lies in k. A perfect field has only separable finite algebraic extensions; for each finite separable extension L/k, L⊗kkˉ is a product of [L:k] copies of kˉ. The extension kˉ/k is flat. (The relative algebraic closure of F in an extension K, An algebraic closure of a field, Perfect fields: every irreducible polynomial is separable, Modules over a field are projective, flat, and injective)

[F15]

In case (ii), write h(x)=x2r+a1x2r−1+⋯+a2r and put w(s)=s2rh(s−1)=1+a1s+⋯+a2rs2r. The two chart rings B0=k[x,y]/(y2−h(x)) and B∞=k[s,z]/(z2−w(s)) are glued on D(x) and D(s) by s=x−1 and z=y/xr. A morphism is finite when the inverse images of an affine target cover are affine with module-finite coordinate algebras; each displayed monic quadratic quotient is free of rank two over its coordinate polynomial ring. (Finite morphisms of schemes, algebra)

[F16]

If h is squarefree and 2 is invertible, the hypersurface chart y2=h(x) is smooth: at any prime either y or h′(x) is a unit. The same criterion applies to z2=w(s) when w is squarefree. For the finite chart over k[x], the relative differential module is ΩB0/k[x]≅(B0/(2y)) dy, and for the infinity chart it is ΩB∞/k[s]≅(B∞/(2z)) dz. (Relative Jacobian criterion with its presentation hypothesis, Locally finite presentation morphisms, Jacobian presentation of Ω)

[F17]

A finite morphism to a proper scheme is proper. Every smooth proper geometrically integral curve over k admits a closed immersion into some PkN. (Composite of a finite morphism and a proper morphism is proper, Every smooth proper curve admits a projective embedding)

[F18]

If A is a finite-type k-domain, then dim⁡A=trdeg⁡kFrac⁡(A). (Affine-domain dimension equals transcendence degree)

[F19]

The Axiom of Choice implies Dependent Choice, which is the additional choice assumption carried by the Cartier-to-Weil supplier used in [F9] and by the Cartier-to-Weil arguments in the finite-morphism and projective-line divisor suppliers. (AC implies DC implies countable choice, The axiom of dependent choice: a relation in which every element is related to something admits an N-indexed chain)

[F20]

Under the Axiom of Choice assumed here, if a smooth proper geometrically integral curve over k has genus zero and admits a divisor of degree one (equivalently, a k-rational closed point), then it is isomorphic to Pk1. (A genus-zero curve with a degree-one divisor is the projective line)

Verification

Proof technique: push the degree-two pullback identity through the fibre of each closed point of Pk1 to pin all ramification indices, then apply the tame different formula and Riemann-Hurwitz; for the concrete model, glue its two standard affine charts and compute their relative differential modules.

1.1F1F3F6

For every closed point q of Pk1 and every closed point p of C with f(p)=q, [F6] gives deg⁡kf∗[q]=2deg⁡kq and f∗[q]=∑p∈f−1(q)ep[p], hence ∑p∈f−1(q)epfp=2 with fp=[κ(p):κ(q)]; consequently epfp≤2, so ep∈{1,2}, fp≤2, every residue extension is separable by [F1], and all ramification is tame.

1.2F13algebra

In case (ii), h has an irreducible factor g of multiplicity one. The g-adic valuation of h in k(x) is therefore 1, so h is not a square in k(x) and y2−h(x) is irreducible over k(x). Thus K=k(x)[y]/(y2−h(x)) is a field, finitely generated of transcendence degree one over k.

1.3F14algebra

The field k is relatively algebraically closed in K. Let A=k[x,y]/(y2−h(x)), so K=Frac⁡(A). Over kˉ, h remains squarefree and has a simple root; its valuation there is odd, so it is not a square in kˉ(x). Hence A⊗kkˉ≅kˉ[x,y]/(y2−h(x)) is a domain. Since kˉ/k is flat, A↪A⊗kkˉ is injective and K⊗kkˉ≅S−1(A⊗kkˉ), where S=A∖{0}, is a domain. If α∈K is algebraic over k but not in k, then L=k(α) is a finite extension of degree greater than one. Since k is perfect, L/k is separable; therefore L⊗kkˉ is a product of [L:k]>1 copies of kˉ, not a domain. Flatness makes L⊗kkˉ↪K⊗kkˉ, a contradiction. Thus k is relatively algebraically closed in K.

1.4F12F15F16F17F18F19algebra

Construct the model explicitly. Write h=x2r+a1x2r−1+⋯+a2r and define w(s)=s2rh(s−1)=1+a1s+⋯+a2rs2r. Glue U0=Spec⁡B0, B0=k[x,y]/(y2−h(x)), to U∞=Spec⁡B∞, B∞=k[s,z]/(z2−w(s)), on D(x) and D(s) by s=x−1 and z=y/xr. The equations agree on this overlap. Each chart is a hypersurface, hence finitely presented over k; the resulting map π:Ch→Pk1 has inverse images U0,U∞ over the standard affine charts, and each chart ring is free of rank two over its base coordinate ring, so π is finite of degree two. The polynomial w is squarefree: its roots are the reciprocals of the nonzero roots of h, and w(0)=1; there is at least one nonzero root since h has 2r≥2 distinct roots and at most one is zero. On either chart, at a prime containing y (respectively z), the derivative h′(x) (respectively w′(s)) is a unit because the polynomial is squarefree; away from those primes, 2y (respectively 2z) is a unit. The relative Jacobian criterion therefore makes both charts smooth over k. Over kˉ, each chart ring is a domain because its squarefree polynomial has a simple root and is not a square in the rational function field. The charts meet in a nonempty open, so their gluing remains integral after base change. Their common function field is K, and both chart rings have dimension one by [F18], so this is a curve; it is smooth and geometrically integral. The finite map to the proper curve Pk1 makes it proper by [F17]. By [F12] it is the smooth proper geometrically integral model of K, unique up to the isomorphism compatible with the specified identification of its function field. The projective-embedding result in [F17] makes this model projective. The Given Axiom of Choice supplies Dependent Choice by [F19] for the Cartier-to-Weil divisor suppliers used below.

1.5F6F7F9algebra

For every closed point p of Ch with q=π(p), the divisor identity π∗div⁡(h)=2div⁡(y) and the additivity and closed-point formula of [F6] give 2ord⁡p(y)=epord⁡q(h), where ord⁡q(h) is the order of the rational function h at q; since h is monic of degree 2r, its only pole is at ∞ and there ord⁡∞(h)=−2r, while at a finite point q=V(g) one has ord⁡q(h)=1 if g∣h (multiplicity of the irreducible factor, h squarefree) and ord⁡q(h)=0 otherwise.

2.1F2F4F16step 1.4step 1.5

If ord⁡p(y)=0, Step 1.5 shows that q≠∞ and that h has order zero at q. Thus p lies on the finite chart, where the actual relative-differentials presentation is ΩB0/k[x]≅(B0/(2y)) dy by [F16]. The element y is a unit at p, so this localized module is zero; hence ℓp=0 and p is unramified with separable residue extension, that is ep=1 by [F4].

2.2F2F3step 1.1

In case (i), the branch locus is exactly {q1,…,q2r} and every point p over qi has ep=2; by Step 1.1 there is exactly one such p=pi, with fpi=1, so κ(pi)=κ(qi)=k and epi=2, while every point over a non-branch point has ep=1.

2.3F3F4F9step 1.4step 1.5

If ord⁡p(y)=m>0, then Step 1.5 shows that q=V(g) for an irreducible factor g of h, with ord⁡q(h)=1, and ep=2m≥2. The fibre of the actual finite chart over q is κ(q)[y]/(y2), so it has a unique point with residue field κ(q). Locally write h=gu with u a unit. The maximal ideal at that point is (g,y) and g=y2/u, hence it is (y); by [F9], y is a uniformizer. Therefore ep=ord⁡p(g)=2 and m=1. The point is tame and has ℓp=ep−1=1 by [F4].

2.4F4F16step 1.4step 1.5algebra

If ord⁡p(y)<0, then Step 1.5 forces q=∞. On the infinity chart the coordinates are s=1/x and z=y/xr, with z2=w(s). The fibre over s=0 is k[z]/(z2−1)≅k×k, so it consists of two distinct k-rational points. At each, z is a unit and [F16] gives ΩB∞/k[s]≅(B∞/(2z)) dz=0 locally; each is unramified, hence has ep=1. These are the two rational points at infinity.

3.1F4F5F7step 2.2

In case (i), the ramified points pi are tame with separable residue extension, so [F4] gives ℓpi=epi−1=1, and every unramified point has ℓp=0 by the criterion ℓp=0⇔ep=1 and separable residue; hence Rf=∑i=12r[pi] with deg⁡kRf=∑ideg⁡k(pi)=2r.

3.2F4F5F7step 2.1step 2.3step 2.4

Combining Steps 2.1, 2.3 and 2.4, the ramified points of π are exactly the points pg over the closed points V(g) for the irreducible factors g of h, each unique with epg=2, fpg=1, hence κ(pg)=k[x]/(g) and ℓpg=1; every other point is unramified with ℓp=0; therefore Rπ=∑g∣h[pg] and deg⁡kRπ=∑g∣hdeg⁡(g)=deg⁡(h)=2r.

4.1F1F8F10F11step 3.1

In case (i), k(C)/k(Pk1) has degree 2 and is separable by [F1], so Riemann-Hurwitz applies: 2g(C)−2=2(2⋅0−2)+deg⁡kRf=−4+2r, and therefore g(C)=r−1 with g as in [F10].

4.2F1F8F10F11step 3.2

In case (ii) the extension k(Ch)/k(Pk1) is separable by [F1], so Riemann-Hurwitz gives 2g(Ch)−2=2(2⋅0−2)+deg⁡kRπ=−4+2r, that is g(Ch)=r−1, with g as in [F10]; when h splits over k with distinct roots c1,…,c2r∈k the branch locus is exactly those 2r distinct k-rational points, so Ch→Pk1 is a cover of case (i) and the two computations agree.

5.1F8F10F20step 4.2∎

The small cases: for r=1 the model of y2=x2−1 is the smooth plane conic Y2=X2−Z2 with rational point [1:0:1]; its partial derivatives −2X,2Y,2Z cannot vanish simultaneously at a projective point. Step 4.2 gives genus zero, and the rational point defines a degree-one divisor, so [F20] gives Ch≅Pk1. The projection identification is explicit: [U:V]⟼[U2+V2:U2−V2:2UV]. On X+Y≠0 its inverse is [X+Y:Z], and on X−Y≠0 it is [Z:X−Y]; the formulas agree on the overlap. For r=2 and c∈k∖{0,1} the polynomial (x2−1)(x2−c) is monic squarefree of degree 4, so the model gives a quartic family of genus 1. For r=3 every squarefree sextic gives a model of genus 2. Each case is consistent with g(Ch)=r−1.

ExampleConstruction: AI-generatedVerification: AI-adaptedprecheck passaudited 2026-10-02Open item page →

One cocycle carried through the residue realization of Serre duality

Example

Assume the Axiom of Choice as inherited from the cited duality, residue and cohomology suppliers (The Axiom of Choice). Let k be a perfect field, let C=Pk1 have coordinate t=x1/x0 on U0, fix d≥0, and put L=O(−d−2). Its dual canonical twist is ωC⊗L−1≅O(d).

For 1≤j≤d+1, the local tail in the U0 frame of L, cj=[t−j⊗x0−d−2]∈H1(C,L), is a finite-support principal-part class. For 0≤m≤d, the section of the dual twist is σm=tmdt⊗x0d+2on U0,σm=−sd−mds⊗x1d+2on U1, where s=1/t. The second expression shows it is regular at infinity and corresponds to x0d−mx1m under ωC≅O(−2).

The positive residue pairing evaluates at the supported rational point: ⟨cj,σm⟩=res⁡0(tm−jdt)=δj,m+1. With rows j=1,…,d+1 and columns m=0,…,d, this is the diagonal identity matrix. Reversing the section order to m=d,…,0 makes it anti-diagonal. Since the sections form a basis, this invertible matrix also proves that the classes cj form a basis. The fixed normalized Serre pairing is its negative, so its matrix has entries −δj,m+1; it is perfect. At d=0 its value on c1 and σ0 is −1, which equals 1 in characteristic two.

The Axiom of Choice enters through the cited suppliers; the displayed computations make no additional choices.

Facts & Assumptions

Given: the Axiom of Choice, a perfect field k, C=Pk1 with coordinate t=x1/x0, an integer d≥0, and L=O(−d−2).

[F1]

The Axiom of Choice is inherited through the duality, residue and projective-cohomology suppliers; this computation makes no further selection. (The Axiom of Choice)

[F2]

The canonical bundle is ωC≅O(−2). Under the standard charts, dt=−s−2ds and the canonical-bundle identification sends dt to x0−2 and ds to −x1−2. The sheaves O(r) are invertible, satisfy O(r)−1≅O(−r), and their frames obey x0r=srx1r on the overlap. (Canonical bundle and canonical divisors, Twisting sheaf on Proj)

[F3]

The cohomology of the twists gives H1(O(−d−2)) the Laurent basis x0e0x1e1 with e0,e1<0 and e0+e1=−d−2, and gives H0(O(d)) the basis x0d−mx1m, 0≤m≤d. (Cohomology of O(d) on projective space)

[F4]

Principal parts compute H1 as finite-support local tails modulo principal parts of global meromorphic sections (including zero); in particular each tail t−j in the frame x0−d−2 represents the class cj. (H^1 of a line bundle on a curve as principal parts modulo meromorphic and regular sections)

[F5]

At a rational point with parameter t, residue is the coefficient of t−1dt. Over a perfect field, the positive residue pairing is the sum of these local residues. (Residue of a rational differential at a separable closed point, The residue pairing of a line bundle with the dual canonical twist, Perfect fields: every irreducible polynomial is separable)

[F6]

Over a perfect field, the fixed normalized Serre pairing is the negative of the positive residue pairing. (Serre duality for line bundles on a smooth proper curve, and the residue realization, Normalization of the trace for Serre duality on a curve)

[F7]

For a smooth proper geometrically integral curve, h1(C,ωC)=h0(C,OC). (h^1 of a line bundle equals the dimension of the space of dual sections)

Verification

Proof technique: represent the tails and global sections in their actual line-bundle frames, evaluate the positive local coefficient, and apply the fixed-trace sign comparison.

1.1F1F2F3F4

By [F2], ωC⊗L−1≅O(d), and [F3] gives dim⁡kH1(C,L)=dim⁡kH0(C,O(d))=d+1. For each 1≤j≤d+1, the tail t−j in the U0 frame has finite support at the rational origin and represents cj by [F4].

2.1F1F2F3step 1.1

The standard section x0d−mx1m is tmx0d on U0, so under dt↦x0−2 it is represented by σm=tmdt⊗x0d+2. With s=1/t, dt=−s−2ds and x0d+2=sd+2x1d+2, this becomes −sd−mds⊗x1d+2, regular for 0≤m≤d. Thus the σm form the displayed section basis.

3.1F1F5F6step 1.1step 2.1

Multiplying the local representative of cj by σm cancels the line-bundle frames and gives tm−jdt at the origin. By [F5], its positive residue is 1 when m−j=−1 and 0 otherwise, namely δj,m+1. By [F6], the fixed normalized Serre value is −δj,m+1.

4.1F1F3step 1.1step 2.1step 3.1algebra

For ascending section order m=0,…,d, the matrix δj,m+1 is diagonal; for reversed order m=d,…,0 it is anti-diagonal. Since the sections form a basis by step 2.1, invertibility of this positive residue matrix proves that the cj are independent; their number is d+1=dim⁡kH1(C,L) by step 1.1, so they form a basis. The normalized matrix is its negative and is invertible, so both pairings are perfect.

5.1F1F2F6F7step 3.1∎

For d=0, [F7] gives h1(C,ωC)=1, and step 3.1 gives normalized trace −1 on [t−1⊗x0−2] paired with dt⊗x02. In characteristic two this value is 1.

Sources