04.04.01 · algebraic-geometry / riemann-roch

Riemann-Roch theorem for curves

shipped3 tiersLean: partial

Anchor (Master): Hartshorne §IV; Mumford *Curves and their Jacobians*; Arbarello-Cornalba-Griffiths-Harris

Intuition [Beginner]

The Riemann-Roch theorem tells you exactly how many functions exist on a curve with prescribed poles and zeros. Imagine you specify: "I want a function with at most a double pole at point $P$ and a simple pole at point $Q$ , and that vanishes at points $R, S$ ." Riemann-Roch counts the dimension of the space of such functions in terms of the curve's genus and the divisor data.

The headline formula: on a smooth projective curve $C$ of genus $g$ , for a divisor $D$ of degree $d$ ,

ℓ (D) - ℓ (K - D) = d - g + 1,

where $ℓ (D)$ is the dimension of the function space and $K$ is the canonical divisor.

The genus $g$ is the only invariant of the curve that appears: it's the topological "number of holes" of the underlying Riemann surface. Riemann-Roch makes geometry computable through arithmetic on divisors.

Visual [Beginner]

A genus- $g$ curve with marked points where divisor data is specified; the dimension of the function space depends only on the divisor and the genus.

Worked example [Beginner]

The simplest curve is the projective line $P^{1}$ , which has genus $g = 0$ . For any divisor $D$ of degree $d \geq 0$ on $P^{1}$ , the function space has dimension $ℓ (D) = d + 1$ — for example, polynomials of degree $\leq d$ form a $(d + 1)$ -dimensional space.

Riemann-Roch on $P^{1}$ : $ℓ (D) - ℓ (K - D) = d - 0 + 1 = d + 1$ . Since $K = - 2$ (degree of canonical) on $P^{1}$ and $ℓ$ of a negative-degree divisor vanishes, the formula simplifies to $ℓ (D) = d + 1$ . Match.

For an elliptic curve (genus $g = 1$ ): a divisor of degree $1$ at one point has $ℓ (D) = 1$ (only constants and one extra function, by Riemann-Roch). A divisor of degree $\geq 2$ has $ℓ (D) = d$ , so the function space grows linearly with the degree past genus.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $C$ be a smooth projective curve over an algebraically closed field $k$ , of genus $g = dim_{k} H^{1} (C; O_{C})$ . For a divisor $D$ on $C$ , denote by $O (D)$ the corresponding invertible sheaf (line bundle).

Definitions:

$ℓ (D) := dim_{k} H^{0} (C; O (D)) = dim_{k} {rational functions f with (f) + D \geq 0} \cup {0}$ .
$h^{1} (D) := dim_{k} H^{1} (C; O (D))$ .
The canonical divisor $K_{C}$ is any divisor associated to the canonical line bundle $ω_{C} = Ω_{C / k}^{1}$ of holomorphic 1-forms.
$de g (K_{C}) = 2 g - 2$ (a foundational computation).

Theorem (Riemann-Roch). For any divisor $D$ on a smooth projective curve $C$ of genus $g$ ,

ℓ (D) - ℓ (K_{C} - D) = de g (D) - g + 1.

By Serre duality, $H^{1} (C; O (D)) ≅ H^{0} (C; ω_{C} \otimes O (- D))^{\lor} = H^{0} (C; O (K_{C} - D))^{\lor}$ , so $h^{1} (D) = ℓ (K_{C} - D)$ . The Riemann-Roch formula becomes:

h^{0} (D) - h^{1} (D) = de g (D) - g + 1

— a statement about the Euler characteristic of the line bundle $O (D)$ :

χ (O (D)) = h^{0} (D) - h^{1} (D) = de g (D) - g + 1 = de g (D) + χ (O_{C}) .

Riemann-Roch in this form generalises directly: for any line bundle $L$ on $C$ , $χ (L) = de g (L) + 1 - g$ . The Euler characteristic is additive in line bundles, with the genus playing the role of a baseline.

Corollaries.

If $de g (D) > 2 g - 2$ , then $ℓ (K_{C} - D) = 0$ (negative degree), so $ℓ (D) = de g (D) - g + 1$ .
If $de g (D) < 0$ , then $ℓ (D) = 0$ .
Canonical divisor: $ℓ (K_{C}) = g$ (apply Riemann-Roch with $D = K_{C}$ , get $ℓ (K_{C}) - 1 = (2 g - 2) - g + 1 = g - 1$ ).

Key theorem with proof [Intermediate+]

Theorem (Riemann-Roch on curves). Let $C$ be a smooth projective curve of genus $g$ over an algebraically closed field $k$ , and $D$ a divisor on $C$ . Then $χ (O (D)) = de g (D) + 1 - g$ .

Proof sketch (induction on the support of $D$ ).

Base case. For $D = 0$ , $χ (O_{C}) = h^{0} (O_{C}) - h^{1} (O_{C}) = 1 - g$ (since $C$ is connected reduced projective and $g$ is the dimension of $H^{1}$ ). This matches $de g (0) + 1 - g = 1 - g$ . ✓

Inductive step. Suppose Riemann-Roch holds for some divisor $D$ . Consider $D^{'} = D + P$ for a closed point $P \in C$ . There is a short exact sequence of sheaves on $C$ :

0 \to O (D) \to O (D + P) \to k_{P} \to 0,

where $k_{P}$ is the skyscraper sheaf at $P$ with stalk $k$ . The long exact sequence of cohomology gives:

\dots \to H^{0} (O (D)) \to H^{0} (O (D + P)) \to H^{0} (k_{P}) \to H^{1} (O (D)) \to H^{1} (O (D + P)) \to 0.

Since $H^{0} (k_{P}) = k$ and $H^{i} (k_{P}) = 0$ for $i \geq 1$ , the alternating sum of dimensions gives:

χ (O (D + P)) - χ (O (D)) = χ (k_{P}) = 1.

Combined with $de g (D + P) = de g (D) + 1$ , by induction:

χ (O (D + P)) = χ (O (D)) + 1 = (de g (D) + 1 - g) + 1 = de g (D + P) + 1 - g .

The same argument with subtracted points (using $0 \to O (D - P) \to O (D) \to k_{P} \to 0$ ) extends to all divisors. By induction, Riemann-Roch holds for any divisor.

The Serre duality identification $h^{1} (D) = ℓ (K - D)$ then gives the classical form. $□$

Bridge. The construction here builds toward later units of the strand, where the same pattern is taken up at higher structure. The defining pattern appears again in those units in a sharpened form, where the local data is glued or quotiented. Putting these together, the foundational insight is that the data of this unit gives the structural signature that the rest of the strand reads off.

Exercises [Intermediate+]

Exercise 5 (medium, conceptual).

Why is Riemann-Roch on a curve the correct generalisation of "polynomials of degree $\leq d$ form a $(d + 1)$ -dimensional space"?

Hint

On $P^{1}$ (genus 0), Riemann-Roch reduces to exactly the polynomial-degree statement.

Answer

On $P^{1}$ , polynomials of degree $\leq d$ form $H^{0} (P^{1}; O (d))$ . Riemann-Roch gives $ℓ (O (d)) = d - 0 + 1 = d + 1$ for $d \geq 0$ , exactly recovering the polynomial-counting statement.

The generalisation to higher genus introduces a correction term $ℓ (K - D)$ — the obstruction to having "as many functions as you'd expect." This obstruction is the genus- $g$ defect: on a curve of genus $g$ , you have fewer functions than the naive degree-counting predicts, and the Serre-dual term $h^{0} (K - D)$ measures exactly that defect. Riemann-Roch corrects polynomial counting for nonzero topology.

Exercise 6 (hard, proof).

Show that on a smooth projective curve, two linearly equivalent divisors have the same dimension of associated function space: $D_{1} \sim D_{2} \Rightarrow ℓ (D_{1}) = ℓ (D_{2})$ .

Hint

Linear equivalence means $D_{1} - D_{2} = (f)$ for a rational function $f$ .

Answer

If $D_{1} \sim D_{2}$ , then $D_{1} - D_{2} = (f)$ for some rational function $f \neq = 0$ . Multiplication by $f$ defines a map of sheaves $O (D_{2}) \to O (D_{1})$ sending a section $s$ with $(s) + D_{2} \geq 0$ to $f s$ with $(f s) + D_{1} = (s) + (f) + D_{1} = (s) + D_{2} \geq 0$ . This is an isomorphism of line bundles (with inverse given by multiplication by $f^{- 1}$ ).

So $H^{0} (C; O (D_{1})) ≅ H^{0} (C; O (D_{2}))$ as $k$ -vector spaces, hence $ℓ (D_{1}) = ℓ (D_{2})$ . $□$

The dimension $ℓ$ depends only on the linear-equivalence class, justifying viewing it as a function on the divisor class group $Cl (C) = Pic (C)$ .

Exercise 7 (hard, conceptual).

State the Riemann-Roch theorem in the form $χ (L) = de g (L) + 1 - g$ and explain why this is the more "modern" formulation.

Hint

The Euler characteristic packages $h^{0}$ and $h^{1}$ together; Serre duality recovers the classical form.

Answer

For any line bundle $L$ on a smooth projective curve $C$ ,

χ (L) := h^{0} (L) - h^{1} (L) = de g (L) + 1 - g .

This is the modern form because:

Euler characteristic is additive. On any short exact sequence $0 \to L_{1} \to L_{2} \to L_{3} \to 0$ of coherent sheaves, $χ (L_{2}) = χ (L_{1}) + χ (L_{3})$ by the long exact sequence. So $χ$ is a "topological invariant."
The formula isolates the genus. $χ (O_{C}) = 1 - g$ , and the line bundle contributes $de g (L)$ on top.
Generalisation. For higher-rank vector bundles $E$ , $χ (E) = de g (E) + r (1 - g)$ where $r = rank (E)$ . For higher-dimensional varieties, Hirzebruch-Riemann-Roch and Grothendieck-Riemann-Roch give analogous formulas.

The classical $ℓ (D) - ℓ (K - D)$ form is then recovered via Serre duality $H^{1} (L) ≅ H^{0} (K - L)^{\lor}$ . The modern form makes the theorem categorical (Euler characteristic is a homomorphism $K_{0} (Coh (C)) \to Z$ ) and generalises immediately.

Lean formalization [Intermediate+]

lean_status: partial — Mathlib has the categorical infrastructure (sheaves, divisors, line bundles on schemes) but the Riemann-Roch theorem itself is not yet formalised in full generality.

[object Promise]

Advanced results [Master]

Hirzebruch-Riemann-Roch. For a smooth projective variety $X$ over $C$ of dimension $n$ and a coherent sheaf $F$ ,

χ (X; F) = \int_{X} ch (F) \cdot td (X),

where $ch$ is the Chern character and $td$ the Todd class. For curves ( $n = 1$ ): $ch (L) = 1 + c_{1} (L)$ , $td (C) = 1 + \frac{1}{2} c_{1} (T_{C})$ , giving the classical Riemann-Roch upon integration.

Grothendieck-Riemann-Roch. For a proper morphism $f : X \to Y$ between smooth projective varieties, the Euler characteristic transforms by

ch (f_{!} F) \cdot td (Y) = f_{*} (ch (F) \cdot td (X)),

a far-reaching categorification of the classical theorem.

Brill-Noether theory. The dimension of the space of "special" divisors — those with $ℓ (D) > de g (D) - g + 1$ — is studied by Brill-Noether theory. The expected dimension of the space of divisor classes $W_{d}^{r} := {D : de g (D) = d, ℓ (D) \geq r + 1} \subset Pic^{d} (C)$ is the Brill-Noether number $ρ (g, r, d) := g - (r + 1) (g - d + r)$ .

Clifford's theorem. For a special divisor $D$ on a curve (i.e., $ℓ (K - D) > 0$ ), $ℓ (D) \leq \frac{1}{2} de g (D) + 1$ , with equality iff $D = 0$ , $D = K$ , or $C$ is hyperelliptic. This bounds how special a divisor can be.

Synthesis. This construction generalises the pattern fixed in 04.01.01 (sheaf), with the symmetric data replaced by its skew or twisted analogue. Read in the opposite direction, the construction is dual to the metric story: complements and orthogonality are taken with respect to the bilinear datum of this unit, not a metric, and the resulting category of subobjects is the one the rest of the strand classifies. The central insight is that this datum identifies algebra with geometry: functions become vector fields, subspaces become quotients, and invariants become cohomology classes — and that identification is the engine driving every theorem downstream.

Full proof set [Master]

The standard inductive proof on the support of $D$ is given in the formal-definition section. A higher proof uses the Cousin sequence and the long exact sequence in sheaf cohomology, identifying everything in terms of the Euler characteristic of $O (D)$ and exploiting Serre duality. A proof via adjunction on the surface $C \times C$ is also possible but more sophisticated.

Connections [Master]

Sheaf cohomology 04.03.01 — Riemann-Roch is fundamentally a statement about Euler characteristics of coherent sheaf cohomology.
Sheaf 04.01.01 — line bundles, divisors, and the structure sheaf are the protagonists.
Scheme 04.02.01 — the algebraic-geometric setting; smooth projective curves are 1-dimensional smooth proper schemes.
Riemann-Roch for compact Riemann surfaces 06.04.01 — the analytic version, equivalent via Serre's GAGA.
Hirzebruch-Riemann-Roch — generalisation to higher-dimensional smooth projective varieties.
Grothendieck-Riemann-Roch — relative version, the deep statement of intersection theory and K-theory.
Atiyah-Singer index theorem 03.09.10 — the analytic generalisation of Riemann-Roch to elliptic operators on manifolds.
Serre duality — the key duality $H^{1} (C; O (D)) ≅ H^{0} (C; ω_{C} \otimes O (- D))^{\lor}$ that powers the classical form.

Historical & philosophical context [Master]

Bernhard Riemann proved one half of the theorem in 1857 — the inequality $ℓ (D) \geq de g (D) - g + 1$ — in the context of compact Riemann surfaces, in his pioneering work on multivalued functions. His student Gustav Roch supplied the equality (the correction term $ℓ (K - D)$ ) in 1865, completing the theorem just before his early death.

The theorem was reformulated and generalised through the 20th century: Friedrich Hirzebruch's 1954 Hirzebruch-Riemann-Roch theorem extended it to higher-dimensional complex manifolds using Chern classes. Alexander Grothendieck's relative version (1957–61) — Grothendieck-Riemann-Roch — placed it in the framework of K-theory and made it work for arbitrary proper morphisms. The Atiyah-Singer index theorem (1963) gave the full analytic generalisation to elliptic operators on smooth manifolds, identifying Riemann-Roch as a special case.

Conceptually, Riemann-Roch is the prototype of a cohomological dimension formula: it expresses a transcendental invariant (the dimension of a function space) in terms of topological data (genus, divisor degree). This pattern recurs throughout modern mathematics: the Euler characteristic, the Hirzebruch signature theorem, the Atiyah-Singer index theorem, and the Eichler-Selberg trace formula all share Riemann-Roch's underlying philosophy that index = topology.

Bibliography [Master]

Hartshorne, Algebraic Geometry — §IV.1 is the standard algebro-geometric proof.
Vakil, The Rising Sea: Foundations of Algebraic Geometry — §18, modern scheme-theoretic treatment.
Forster, Lectures on Riemann Surfaces — §16, classical analytic perspective.
Griffiths & Harris, Principles of Algebraic Geometry — Ch. 2, Hodge-theoretic Riemann-Roch.
Mumford, Curves and their Jacobians — geometric perspective, especially on Brill-Noether and theta divisors.
Arbarello, Cornalba, Griffiths & Harris, Geometry of Algebraic Curves — comprehensive treatment of Brill-Noether theory.
Hirzebruch, Topological Methods in Algebraic Geometry — the original Hirzebruch-Riemann-Roch.